diff --git "a/checkpoint-49491/trainer_state.json" "b/checkpoint-49491/trainer_state.json" new file mode 100644--- /dev/null +++ "b/checkpoint-49491/trainer_state.json" @@ -0,0 +1,297043 @@ +{ + "best_metric": 0.5778843760490417, + "best_model_checkpoint": "bird_sounds_classification/checkpoint-49491", + "epoch": 9.0, + "global_step": 49491, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 6.000000000000001e-08, + "loss": 2.1692, + "step": 1 + }, + { + "epoch": 0.0, + "learning_rate": 1.2000000000000002e-07, + "loss": 1.1512, + "step": 2 + }, + { + "epoch": 0.0, + "learning_rate": 1.8e-07, + "loss": 1.7151, + "step": 3 + }, + { + "epoch": 0.0, + "learning_rate": 2.4000000000000003e-07, + "loss": 0.133, + "step": 4 + }, + { + "epoch": 0.0, + "learning_rate": 3.0000000000000004e-07, + "loss": 0.7116, + "step": 5 + }, + { + "epoch": 0.0, + "learning_rate": 3.6e-07, + "loss": 1.4782, + "step": 6 + }, + { + "epoch": 0.0, + "learning_rate": 4.2000000000000006e-07, + "loss": 0.2438, + "step": 7 + }, + { + "epoch": 0.0, + "learning_rate": 4.800000000000001e-07, + "loss": 0.3106, + "step": 8 + }, + { + "epoch": 0.0, + "learning_rate": 5.4e-07, + "loss": 1.7486, + "step": 9 + }, + { + "epoch": 0.0, + "learning_rate": 6.000000000000001e-07, + "loss": 1.0698, + "step": 10 + }, + { + "epoch": 0.0, + "learning_rate": 6.6e-07, + "loss": 0.3339, + "step": 11 + }, + { + "epoch": 0.0, + "learning_rate": 7.2e-07, + "loss": 1.407, + "step": 12 + }, + { + "epoch": 0.0, + "learning_rate": 7.8e-07, + "loss": 2.013, + "step": 13 + }, + { + "epoch": 0.0, + "learning_rate": 8.400000000000001e-07, + "loss": 0.0807, + "step": 14 + }, + { + "epoch": 0.0, + "learning_rate": 9e-07, + "loss": 1.02, + "step": 15 + }, + { + "epoch": 0.0, + "learning_rate": 9.600000000000001e-07, + "loss": 1.3393, + "step": 16 + }, + { + "epoch": 0.0, + "learning_rate": 1.0200000000000002e-06, + "loss": 0.8432, + "step": 17 + }, + { + "epoch": 0.0, + "learning_rate": 1.08e-06, + "loss": 1.0313, + "step": 18 + }, + { + "epoch": 0.0, + "learning_rate": 1.14e-06, + "loss": 1.3397, + "step": 19 + }, + { + "epoch": 0.0, + "learning_rate": 1.2000000000000002e-06, + "loss": 3.7463, + "step": 20 + }, + { + "epoch": 0.0, + "learning_rate": 1.26e-06, + "loss": 0.6277, + "step": 21 + }, + { + "epoch": 0.0, + "learning_rate": 1.32e-06, + "loss": 0.7714, + "step": 22 + }, + { + "epoch": 0.0, + "learning_rate": 1.3800000000000001e-06, + "loss": 3.0711, + "step": 23 + }, + { + "epoch": 0.0, + "learning_rate": 1.44e-06, + "loss": 0.666, + "step": 24 + }, + { + "epoch": 0.0, + "learning_rate": 1.5e-06, + "loss": 1.185, + "step": 25 + }, + { + "epoch": 0.0, + "learning_rate": 1.56e-06, + "loss": 0.7573, + "step": 26 + }, + { + "epoch": 0.0, + "learning_rate": 1.6200000000000002e-06, + "loss": 1.4296, + "step": 27 + }, + { + "epoch": 0.01, + "learning_rate": 1.6800000000000002e-06, + "loss": 1.4161, + "step": 28 + }, + { + "epoch": 0.01, + "learning_rate": 1.7399999999999999e-06, + "loss": 0.0748, + "step": 29 + }, + { + "epoch": 0.01, + "learning_rate": 1.8e-06, + "loss": 1.0158, + "step": 30 + }, + { + "epoch": 0.01, + "learning_rate": 1.86e-06, + "loss": 1.8429, + "step": 31 + }, + { + "epoch": 0.01, + "learning_rate": 1.9200000000000003e-06, + "loss": 1.2369, + "step": 32 + }, + { + "epoch": 0.01, + "learning_rate": 1.98e-06, + "loss": 0.0828, + "step": 33 + }, + { + "epoch": 0.01, + "learning_rate": 2.0400000000000004e-06, + "loss": 0.4729, + "step": 34 + }, + { + "epoch": 0.01, + "learning_rate": 2.1e-06, + "loss": 0.4093, + "step": 35 + }, + { + "epoch": 0.01, + "learning_rate": 2.16e-06, + "loss": 2.0908, + "step": 36 + }, + { + "epoch": 0.01, + "learning_rate": 2.22e-06, + "loss": 1.0054, + "step": 37 + }, + { + "epoch": 0.01, + "learning_rate": 2.28e-06, + "loss": 0.9275, + "step": 38 + }, + { + "epoch": 0.01, + "learning_rate": 2.34e-06, + "loss": 0.3824, + "step": 39 + }, + { + "epoch": 0.01, + "learning_rate": 2.4000000000000003e-06, + "loss": 0.3715, + "step": 40 + }, + { + "epoch": 0.01, + "learning_rate": 2.4599999999999997e-06, + "loss": 0.5357, + "step": 41 + }, + { + "epoch": 0.01, + "learning_rate": 2.52e-06, + "loss": 1.1678, + "step": 42 + }, + { + "epoch": 0.01, + "learning_rate": 2.58e-06, + "loss": 0.2686, + "step": 43 + }, + { + "epoch": 0.01, + "learning_rate": 2.64e-06, + "loss": 1.3976, + "step": 44 + }, + { + "epoch": 0.01, + "learning_rate": 2.7e-06, + "loss": 0.3518, + "step": 45 + }, + { + "epoch": 0.01, + "learning_rate": 2.7600000000000003e-06, + "loss": 0.7257, + "step": 46 + }, + { + "epoch": 0.01, + "learning_rate": 2.82e-06, + "loss": 2.7868, + "step": 47 + }, + { + "epoch": 0.01, + "learning_rate": 2.88e-06, + "loss": 0.7075, + "step": 48 + }, + { + "epoch": 0.01, + "learning_rate": 2.9400000000000002e-06, + "loss": 0.9932, + "step": 49 + }, + { + "epoch": 0.01, + "learning_rate": 3e-06, + "loss": 1.343, + "step": 50 + }, + { + "epoch": 0.01, + "learning_rate": 2.999945394976338e-06, + "loss": 2.6099, + "step": 51 + }, + { + "epoch": 0.01, + "learning_rate": 2.9998907899526755e-06, + "loss": 1.5674, + "step": 52 + }, + { + "epoch": 0.01, + "learning_rate": 2.9998361849290135e-06, + "loss": 0.2964, + "step": 53 + }, + { + "epoch": 0.01, + "learning_rate": 2.9997815799053514e-06, + "loss": 1.4369, + "step": 54 + }, + { + "epoch": 0.01, + "learning_rate": 2.999726974881689e-06, + "loss": 0.8677, + "step": 55 + }, + { + "epoch": 0.01, + "learning_rate": 2.999672369858027e-06, + "loss": 0.5638, + "step": 56 + }, + { + "epoch": 0.01, + "learning_rate": 2.999617764834365e-06, + "loss": 0.0694, + "step": 57 + }, + { + "epoch": 0.01, + "learning_rate": 2.9995631598107028e-06, + "loss": 1.3163, + "step": 58 + }, + { + "epoch": 0.01, + "learning_rate": 2.9995085547870407e-06, + "loss": 0.1027, + "step": 59 + }, + { + "epoch": 0.01, + "learning_rate": 2.9994539497633782e-06, + "loss": 2.0331, + "step": 60 + }, + { + "epoch": 0.01, + "learning_rate": 2.999399344739716e-06, + "loss": 0.6172, + "step": 61 + }, + { + "epoch": 0.01, + "learning_rate": 2.999344739716054e-06, + "loss": 0.7194, + "step": 62 + }, + { + "epoch": 0.01, + "learning_rate": 2.9992901346923916e-06, + "loss": 2.3816, + "step": 63 + }, + { + "epoch": 0.01, + "learning_rate": 2.9992355296687296e-06, + "loss": 1.9861, + "step": 64 + }, + { + "epoch": 0.01, + "learning_rate": 2.9991809246450675e-06, + "loss": 0.951, + "step": 65 + }, + { + "epoch": 0.01, + "learning_rate": 2.999126319621405e-06, + "loss": 1.6874, + "step": 66 + }, + { + "epoch": 0.01, + "learning_rate": 2.999071714597743e-06, + "loss": 1.2229, + "step": 67 + }, + { + "epoch": 0.01, + "learning_rate": 2.999017109574081e-06, + "loss": 1.8053, + "step": 68 + }, + { + "epoch": 0.01, + "learning_rate": 2.9989625045504185e-06, + "loss": 0.461, + "step": 69 + }, + { + "epoch": 0.01, + "learning_rate": 2.9989078995267564e-06, + "loss": 0.5853, + "step": 70 + }, + { + "epoch": 0.01, + "learning_rate": 2.9988532945030943e-06, + "loss": 0.2402, + "step": 71 + }, + { + "epoch": 0.01, + "learning_rate": 2.9987986894794323e-06, + "loss": 1.921, + "step": 72 + }, + { + "epoch": 0.01, + "learning_rate": 2.9987440844557702e-06, + "loss": 0.2355, + "step": 73 + }, + { + "epoch": 0.01, + "learning_rate": 2.9986894794321078e-06, + "loss": 0.6991, + "step": 74 + }, + { + "epoch": 0.01, + "learning_rate": 2.9986348744084457e-06, + "loss": 0.3046, + "step": 75 + }, + { + "epoch": 0.01, + "learning_rate": 2.9985802693847836e-06, + "loss": 0.6173, + "step": 76 + }, + { + "epoch": 0.01, + "learning_rate": 2.998525664361121e-06, + "loss": 1.955, + "step": 77 + }, + { + "epoch": 0.01, + "learning_rate": 2.998471059337459e-06, + "loss": 2.4006, + "step": 78 + }, + { + "epoch": 0.01, + "learning_rate": 2.998416454313797e-06, + "loss": 0.3752, + "step": 79 + }, + { + "epoch": 0.01, + "learning_rate": 2.9983618492901346e-06, + "loss": 0.4353, + "step": 80 + }, + { + "epoch": 0.01, + "learning_rate": 2.9983072442664725e-06, + "loss": 0.8027, + "step": 81 + }, + { + "epoch": 0.01, + "learning_rate": 2.9982526392428105e-06, + "loss": 1.3555, + "step": 82 + }, + { + "epoch": 0.02, + "learning_rate": 2.998198034219148e-06, + "loss": 1.2083, + "step": 83 + }, + { + "epoch": 0.02, + "learning_rate": 2.998143429195486e-06, + "loss": 0.3872, + "step": 84 + }, + { + "epoch": 0.02, + "learning_rate": 2.998088824171824e-06, + "loss": 0.6844, + "step": 85 + }, + { + "epoch": 0.02, + "learning_rate": 2.998034219148162e-06, + "loss": 1.165, + "step": 86 + }, + { + "epoch": 0.02, + "learning_rate": 2.9979796141244997e-06, + "loss": 1.022, + "step": 87 + }, + { + "epoch": 0.02, + "learning_rate": 2.9979250091008373e-06, + "loss": 1.1632, + "step": 88 + }, + { + "epoch": 0.02, + "learning_rate": 2.997870404077175e-06, + "loss": 1.1239, + "step": 89 + }, + { + "epoch": 0.02, + "learning_rate": 2.997815799053513e-06, + "loss": 1.098, + "step": 90 + }, + { + "epoch": 0.02, + "learning_rate": 2.9977611940298507e-06, + "loss": 0.8895, + "step": 91 + }, + { + "epoch": 0.02, + "learning_rate": 2.9977065890061886e-06, + "loss": 0.6392, + "step": 92 + }, + { + "epoch": 0.02, + "learning_rate": 2.9976519839825266e-06, + "loss": 1.1739, + "step": 93 + }, + { + "epoch": 0.02, + "learning_rate": 2.997597378958864e-06, + "loss": 0.7522, + "step": 94 + }, + { + "epoch": 0.02, + "learning_rate": 2.997542773935202e-06, + "loss": 0.3061, + "step": 95 + }, + { + "epoch": 0.02, + "learning_rate": 2.99748816891154e-06, + "loss": 0.8269, + "step": 96 + }, + { + "epoch": 0.02, + "learning_rate": 2.997433563887878e-06, + "loss": 0.9363, + "step": 97 + }, + { + "epoch": 0.02, + "learning_rate": 2.997378958864216e-06, + "loss": 0.6306, + "step": 98 + }, + { + "epoch": 0.02, + "learning_rate": 2.9973243538405534e-06, + "loss": 1.0612, + "step": 99 + }, + { + "epoch": 0.02, + "learning_rate": 2.9972697488168913e-06, + "loss": 0.6667, + "step": 100 + }, + { + "epoch": 0.02, + "learning_rate": 2.9972151437932293e-06, + "loss": 2.2683, + "step": 101 + }, + { + "epoch": 0.02, + "learning_rate": 2.9971605387695668e-06, + "loss": 2.3802, + "step": 102 + }, + { + "epoch": 0.02, + "learning_rate": 2.9971059337459047e-06, + "loss": 0.722, + "step": 103 + }, + { + "epoch": 0.02, + "learning_rate": 2.9970513287222427e-06, + "loss": 1.3615, + "step": 104 + }, + { + "epoch": 0.02, + "learning_rate": 2.99699672369858e-06, + "loss": 0.2915, + "step": 105 + }, + { + "epoch": 0.02, + "learning_rate": 2.996942118674918e-06, + "loss": 1.2603, + "step": 106 + }, + { + "epoch": 0.02, + "learning_rate": 2.996887513651256e-06, + "loss": 2.245, + "step": 107 + }, + { + "epoch": 0.02, + "learning_rate": 2.9968329086275936e-06, + "loss": 1.9905, + "step": 108 + }, + { + "epoch": 0.02, + "learning_rate": 2.9967783036039315e-06, + "loss": 1.1718, + "step": 109 + }, + { + "epoch": 0.02, + "learning_rate": 2.9967236985802695e-06, + "loss": 1.1217, + "step": 110 + }, + { + "epoch": 0.02, + "learning_rate": 2.9966690935566074e-06, + "loss": 0.4619, + "step": 111 + }, + { + "epoch": 0.02, + "learning_rate": 2.9966144885329454e-06, + "loss": 0.6483, + "step": 112 + }, + { + "epoch": 0.02, + "learning_rate": 2.996559883509283e-06, + "loss": 1.2079, + "step": 113 + }, + { + "epoch": 0.02, + "learning_rate": 2.996505278485621e-06, + "loss": 1.3388, + "step": 114 + }, + { + "epoch": 0.02, + "learning_rate": 2.9964506734619588e-06, + "loss": 2.2373, + "step": 115 + }, + { + "epoch": 0.02, + "learning_rate": 2.9963960684382963e-06, + "loss": 0.0693, + "step": 116 + }, + { + "epoch": 0.02, + "learning_rate": 2.9963414634146342e-06, + "loss": 0.9476, + "step": 117 + }, + { + "epoch": 0.02, + "learning_rate": 2.996286858390972e-06, + "loss": 1.7786, + "step": 118 + }, + { + "epoch": 0.02, + "learning_rate": 2.9962322533673097e-06, + "loss": 0.6775, + "step": 119 + }, + { + "epoch": 0.02, + "learning_rate": 2.9961776483436476e-06, + "loss": 0.0872, + "step": 120 + }, + { + "epoch": 0.02, + "learning_rate": 2.9961230433199856e-06, + "loss": 1.41, + "step": 121 + }, + { + "epoch": 0.02, + "learning_rate": 2.996068438296323e-06, + "loss": 0.61, + "step": 122 + }, + { + "epoch": 0.02, + "learning_rate": 2.996013833272661e-06, + "loss": 0.1387, + "step": 123 + }, + { + "epoch": 0.02, + "learning_rate": 2.995959228248999e-06, + "loss": 0.7241, + "step": 124 + }, + { + "epoch": 0.02, + "learning_rate": 2.995904623225337e-06, + "loss": 1.862, + "step": 125 + }, + { + "epoch": 0.02, + "learning_rate": 2.995850018201675e-06, + "loss": 0.6098, + "step": 126 + }, + { + "epoch": 0.02, + "learning_rate": 2.9957954131780124e-06, + "loss": 2.3538, + "step": 127 + }, + { + "epoch": 0.02, + "learning_rate": 2.9957408081543503e-06, + "loss": 2.1592, + "step": 128 + }, + { + "epoch": 0.02, + "learning_rate": 2.9956862031306883e-06, + "loss": 0.8298, + "step": 129 + }, + { + "epoch": 0.02, + "learning_rate": 2.995631598107026e-06, + "loss": 1.6141, + "step": 130 + }, + { + "epoch": 0.02, + "learning_rate": 2.9955769930833638e-06, + "loss": 1.5525, + "step": 131 + }, + { + "epoch": 0.02, + "learning_rate": 2.9955223880597017e-06, + "loss": 1.948, + "step": 132 + }, + { + "epoch": 0.02, + "learning_rate": 2.9954677830360392e-06, + "loss": 1.2537, + "step": 133 + }, + { + "epoch": 0.02, + "learning_rate": 2.995413178012377e-06, + "loss": 0.9385, + "step": 134 + }, + { + "epoch": 0.02, + "learning_rate": 2.995358572988715e-06, + "loss": 2.1847, + "step": 135 + }, + { + "epoch": 0.02, + "learning_rate": 2.9953039679650526e-06, + "loss": 3.0623, + "step": 136 + }, + { + "epoch": 0.02, + "learning_rate": 2.9952493629413906e-06, + "loss": 0.7333, + "step": 137 + }, + { + "epoch": 0.03, + "learning_rate": 2.9951947579177285e-06, + "loss": 0.7077, + "step": 138 + }, + { + "epoch": 0.03, + "learning_rate": 2.9951401528940665e-06, + "loss": 2.6214, + "step": 139 + }, + { + "epoch": 0.03, + "learning_rate": 2.9950855478704044e-06, + "loss": 0.9258, + "step": 140 + }, + { + "epoch": 0.03, + "learning_rate": 2.995030942846742e-06, + "loss": 0.5185, + "step": 141 + }, + { + "epoch": 0.03, + "learning_rate": 2.99497633782308e-06, + "loss": 1.0169, + "step": 142 + }, + { + "epoch": 0.03, + "learning_rate": 2.994921732799418e-06, + "loss": 0.666, + "step": 143 + }, + { + "epoch": 0.03, + "learning_rate": 2.9948671277757553e-06, + "loss": 1.5745, + "step": 144 + }, + { + "epoch": 0.03, + "learning_rate": 2.9948125227520933e-06, + "loss": 1.6427, + "step": 145 + }, + { + "epoch": 0.03, + "learning_rate": 2.994757917728431e-06, + "loss": 1.622, + "step": 146 + }, + { + "epoch": 0.03, + "learning_rate": 2.9947033127047687e-06, + "loss": 1.589, + "step": 147 + }, + { + "epoch": 0.03, + "learning_rate": 2.9946487076811067e-06, + "loss": 0.163, + "step": 148 + }, + { + "epoch": 0.03, + "learning_rate": 2.9945941026574446e-06, + "loss": 0.446, + "step": 149 + }, + { + "epoch": 0.03, + "learning_rate": 2.994539497633782e-06, + "loss": 0.1297, + "step": 150 + }, + { + "epoch": 0.03, + "learning_rate": 2.99448489261012e-06, + "loss": 3.4558, + "step": 151 + }, + { + "epoch": 0.03, + "learning_rate": 2.994430287586458e-06, + "loss": 1.401, + "step": 152 + }, + { + "epoch": 0.03, + "learning_rate": 2.994375682562796e-06, + "loss": 0.3302, + "step": 153 + }, + { + "epoch": 0.03, + "learning_rate": 2.994321077539134e-06, + "loss": 0.9294, + "step": 154 + }, + { + "epoch": 0.03, + "learning_rate": 2.9942664725154714e-06, + "loss": 1.6694, + "step": 155 + }, + { + "epoch": 0.03, + "learning_rate": 2.9942118674918094e-06, + "loss": 2.0929, + "step": 156 + }, + { + "epoch": 0.03, + "learning_rate": 2.9941572624681473e-06, + "loss": 0.9457, + "step": 157 + }, + { + "epoch": 0.03, + "learning_rate": 2.994102657444485e-06, + "loss": 0.3696, + "step": 158 + }, + { + "epoch": 0.03, + "learning_rate": 2.9940480524208228e-06, + "loss": 0.9275, + "step": 159 + }, + { + "epoch": 0.03, + "learning_rate": 2.9939934473971607e-06, + "loss": 1.0101, + "step": 160 + }, + { + "epoch": 0.03, + "learning_rate": 2.9939388423734982e-06, + "loss": 1.3145, + "step": 161 + }, + { + "epoch": 0.03, + "learning_rate": 2.993884237349836e-06, + "loss": 0.5611, + "step": 162 + }, + { + "epoch": 0.03, + "learning_rate": 2.993829632326174e-06, + "loss": 0.8808, + "step": 163 + }, + { + "epoch": 0.03, + "learning_rate": 2.9937750273025116e-06, + "loss": 1.2432, + "step": 164 + }, + { + "epoch": 0.03, + "learning_rate": 2.9937204222788496e-06, + "loss": 1.768, + "step": 165 + }, + { + "epoch": 0.03, + "learning_rate": 2.9936658172551875e-06, + "loss": 0.4574, + "step": 166 + }, + { + "epoch": 0.03, + "learning_rate": 2.9936112122315255e-06, + "loss": 0.7146, + "step": 167 + }, + { + "epoch": 0.03, + "learning_rate": 2.9935566072078634e-06, + "loss": 0.7418, + "step": 168 + }, + { + "epoch": 0.03, + "learning_rate": 2.993502002184201e-06, + "loss": 4.6949, + "step": 169 + }, + { + "epoch": 0.03, + "learning_rate": 2.993447397160539e-06, + "loss": 0.4883, + "step": 170 + }, + { + "epoch": 0.03, + "learning_rate": 2.993392792136877e-06, + "loss": 0.3216, + "step": 171 + }, + { + "epoch": 0.03, + "learning_rate": 2.9933381871132143e-06, + "loss": 0.8024, + "step": 172 + }, + { + "epoch": 0.03, + "learning_rate": 2.9932835820895523e-06, + "loss": 0.1847, + "step": 173 + }, + { + "epoch": 0.03, + "learning_rate": 2.9932289770658902e-06, + "loss": 1.5905, + "step": 174 + }, + { + "epoch": 0.03, + "learning_rate": 2.9931743720422278e-06, + "loss": 1.3885, + "step": 175 + }, + { + "epoch": 0.03, + "learning_rate": 2.9931197670185657e-06, + "loss": 0.6115, + "step": 176 + }, + { + "epoch": 0.03, + "learning_rate": 2.9930651619949036e-06, + "loss": 1.0431, + "step": 177 + }, + { + "epoch": 0.03, + "learning_rate": 2.9930105569712416e-06, + "loss": 0.7625, + "step": 178 + }, + { + "epoch": 0.03, + "learning_rate": 2.9929559519475795e-06, + "loss": 0.0391, + "step": 179 + }, + { + "epoch": 0.03, + "learning_rate": 2.992901346923917e-06, + "loss": 2.1114, + "step": 180 + }, + { + "epoch": 0.03, + "learning_rate": 2.992846741900255e-06, + "loss": 2.6923, + "step": 181 + }, + { + "epoch": 0.03, + "learning_rate": 2.992792136876593e-06, + "loss": 0.7562, + "step": 182 + }, + { + "epoch": 0.03, + "learning_rate": 2.9927375318529305e-06, + "loss": 1.0831, + "step": 183 + }, + { + "epoch": 0.03, + "learning_rate": 2.9926829268292684e-06, + "loss": 0.8389, + "step": 184 + }, + { + "epoch": 0.03, + "learning_rate": 2.9926283218056063e-06, + "loss": 0.9981, + "step": 185 + }, + { + "epoch": 0.03, + "learning_rate": 2.992573716781944e-06, + "loss": 0.6063, + "step": 186 + }, + { + "epoch": 0.03, + "learning_rate": 2.992519111758282e-06, + "loss": 0.1634, + "step": 187 + }, + { + "epoch": 0.03, + "learning_rate": 2.9924645067346197e-06, + "loss": 2.3038, + "step": 188 + }, + { + "epoch": 0.03, + "learning_rate": 2.9924099017109573e-06, + "loss": 0.3854, + "step": 189 + }, + { + "epoch": 0.03, + "learning_rate": 2.992355296687295e-06, + "loss": 0.0874, + "step": 190 + }, + { + "epoch": 0.03, + "learning_rate": 2.992300691663633e-06, + "loss": 0.055, + "step": 191 + }, + { + "epoch": 0.03, + "learning_rate": 2.992246086639971e-06, + "loss": 0.3981, + "step": 192 + }, + { + "epoch": 0.04, + "learning_rate": 2.992191481616309e-06, + "loss": 0.7466, + "step": 193 + }, + { + "epoch": 0.04, + "learning_rate": 2.9921368765926466e-06, + "loss": 0.8288, + "step": 194 + }, + { + "epoch": 0.04, + "learning_rate": 2.9920822715689845e-06, + "loss": 1.3679, + "step": 195 + }, + { + "epoch": 0.04, + "learning_rate": 2.9920276665453224e-06, + "loss": 0.8257, + "step": 196 + }, + { + "epoch": 0.04, + "learning_rate": 2.99197306152166e-06, + "loss": 0.7174, + "step": 197 + }, + { + "epoch": 0.04, + "learning_rate": 2.991918456497998e-06, + "loss": 0.5527, + "step": 198 + }, + { + "epoch": 0.04, + "learning_rate": 2.991863851474336e-06, + "loss": 1.1633, + "step": 199 + }, + { + "epoch": 0.04, + "learning_rate": 2.9918092464506734e-06, + "loss": 2.1671, + "step": 200 + }, + { + "epoch": 0.04, + "learning_rate": 2.9917546414270113e-06, + "loss": 0.9674, + "step": 201 + }, + { + "epoch": 0.04, + "learning_rate": 2.9917000364033493e-06, + "loss": 0.8842, + "step": 202 + }, + { + "epoch": 0.04, + "learning_rate": 2.9916454313796868e-06, + "loss": 0.9043, + "step": 203 + }, + { + "epoch": 0.04, + "learning_rate": 2.9915908263560247e-06, + "loss": 0.6933, + "step": 204 + }, + { + "epoch": 0.04, + "learning_rate": 2.9915362213323627e-06, + "loss": 2.0229, + "step": 205 + }, + { + "epoch": 0.04, + "learning_rate": 2.9914816163087006e-06, + "loss": 1.6724, + "step": 206 + }, + { + "epoch": 0.04, + "learning_rate": 2.9914270112850386e-06, + "loss": 0.4769, + "step": 207 + }, + { + "epoch": 0.04, + "learning_rate": 2.991372406261376e-06, + "loss": 1.6237, + "step": 208 + }, + { + "epoch": 0.04, + "learning_rate": 2.991317801237714e-06, + "loss": 2.6205, + "step": 209 + }, + { + "epoch": 0.04, + "learning_rate": 2.991263196214052e-06, + "loss": 0.6015, + "step": 210 + }, + { + "epoch": 0.04, + "learning_rate": 2.9912085911903895e-06, + "loss": 0.8163, + "step": 211 + }, + { + "epoch": 0.04, + "learning_rate": 2.9911539861667274e-06, + "loss": 2.0459, + "step": 212 + }, + { + "epoch": 0.04, + "learning_rate": 2.9910993811430654e-06, + "loss": 1.2603, + "step": 213 + }, + { + "epoch": 0.04, + "learning_rate": 2.991044776119403e-06, + "loss": 1.1164, + "step": 214 + }, + { + "epoch": 0.04, + "learning_rate": 2.990990171095741e-06, + "loss": 1.3802, + "step": 215 + }, + { + "epoch": 0.04, + "learning_rate": 2.9909355660720788e-06, + "loss": 1.4584, + "step": 216 + }, + { + "epoch": 0.04, + "learning_rate": 2.9908809610484163e-06, + "loss": 0.9949, + "step": 217 + }, + { + "epoch": 0.04, + "learning_rate": 2.9908263560247542e-06, + "loss": 0.2623, + "step": 218 + }, + { + "epoch": 0.04, + "learning_rate": 2.990771751001092e-06, + "loss": 0.983, + "step": 219 + }, + { + "epoch": 0.04, + "learning_rate": 2.99071714597743e-06, + "loss": 3.307, + "step": 220 + }, + { + "epoch": 0.04, + "learning_rate": 2.990662540953768e-06, + "loss": 1.5972, + "step": 221 + }, + { + "epoch": 0.04, + "learning_rate": 2.9906079359301056e-06, + "loss": 1.1724, + "step": 222 + }, + { + "epoch": 0.04, + "learning_rate": 2.9905533309064435e-06, + "loss": 0.6152, + "step": 223 + }, + { + "epoch": 0.04, + "learning_rate": 2.9904987258827815e-06, + "loss": 1.4518, + "step": 224 + }, + { + "epoch": 0.04, + "learning_rate": 2.990444120859119e-06, + "loss": 0.5853, + "step": 225 + }, + { + "epoch": 0.04, + "learning_rate": 2.990389515835457e-06, + "loss": 0.56, + "step": 226 + }, + { + "epoch": 0.04, + "learning_rate": 2.990334910811795e-06, + "loss": 0.4368, + "step": 227 + }, + { + "epoch": 0.04, + "learning_rate": 2.9902803057881324e-06, + "loss": 0.5082, + "step": 228 + }, + { + "epoch": 0.04, + "learning_rate": 2.9902257007644703e-06, + "loss": 2.9001, + "step": 229 + }, + { + "epoch": 0.04, + "learning_rate": 2.9901710957408083e-06, + "loss": 2.947, + "step": 230 + }, + { + "epoch": 0.04, + "learning_rate": 2.990116490717146e-06, + "loss": 1.1549, + "step": 231 + }, + { + "epoch": 0.04, + "learning_rate": 2.9900618856934838e-06, + "loss": 1.2564, + "step": 232 + }, + { + "epoch": 0.04, + "learning_rate": 2.9900072806698217e-06, + "loss": 0.8689, + "step": 233 + }, + { + "epoch": 0.04, + "learning_rate": 2.9899526756461596e-06, + "loss": 0.4304, + "step": 234 + }, + { + "epoch": 0.04, + "learning_rate": 2.9898980706224976e-06, + "loss": 0.1459, + "step": 235 + }, + { + "epoch": 0.04, + "learning_rate": 2.989843465598835e-06, + "loss": 1.4742, + "step": 236 + }, + { + "epoch": 0.04, + "learning_rate": 2.989788860575173e-06, + "loss": 1.2041, + "step": 237 + }, + { + "epoch": 0.04, + "learning_rate": 2.989734255551511e-06, + "loss": 0.6161, + "step": 238 + }, + { + "epoch": 0.04, + "learning_rate": 2.9896796505278485e-06, + "loss": 3.9504, + "step": 239 + }, + { + "epoch": 0.04, + "learning_rate": 2.9896250455041865e-06, + "loss": 2.3776, + "step": 240 + }, + { + "epoch": 0.04, + "learning_rate": 2.9895704404805244e-06, + "loss": 1.1377, + "step": 241 + }, + { + "epoch": 0.04, + "learning_rate": 2.989515835456862e-06, + "loss": 0.92, + "step": 242 + }, + { + "epoch": 0.04, + "learning_rate": 2.9894612304332e-06, + "loss": 0.7379, + "step": 243 + }, + { + "epoch": 0.04, + "learning_rate": 2.989406625409538e-06, + "loss": 0.4387, + "step": 244 + }, + { + "epoch": 0.04, + "learning_rate": 2.9893520203858753e-06, + "loss": 1.4414, + "step": 245 + }, + { + "epoch": 0.04, + "learning_rate": 2.9892974153622133e-06, + "loss": 0.4327, + "step": 246 + }, + { + "epoch": 0.04, + "learning_rate": 2.989242810338551e-06, + "loss": 0.8047, + "step": 247 + }, + { + "epoch": 0.05, + "learning_rate": 2.989188205314889e-06, + "loss": 2.1015, + "step": 248 + }, + { + "epoch": 0.05, + "learning_rate": 2.989133600291227e-06, + "loss": 1.1239, + "step": 249 + }, + { + "epoch": 0.05, + "learning_rate": 2.9890789952675646e-06, + "loss": 0.0311, + "step": 250 + }, + { + "epoch": 0.05, + "learning_rate": 2.9890243902439026e-06, + "loss": 0.3483, + "step": 251 + }, + { + "epoch": 0.05, + "learning_rate": 2.9889697852202405e-06, + "loss": 0.6896, + "step": 252 + }, + { + "epoch": 0.05, + "learning_rate": 2.988915180196578e-06, + "loss": 2.6758, + "step": 253 + }, + { + "epoch": 0.05, + "learning_rate": 2.988860575172916e-06, + "loss": 1.1837, + "step": 254 + }, + { + "epoch": 0.05, + "learning_rate": 2.988805970149254e-06, + "loss": 0.7391, + "step": 255 + }, + { + "epoch": 0.05, + "learning_rate": 2.9887513651255914e-06, + "loss": 1.4117, + "step": 256 + }, + { + "epoch": 0.05, + "learning_rate": 2.9886967601019294e-06, + "loss": 0.1456, + "step": 257 + }, + { + "epoch": 0.05, + "learning_rate": 2.9886421550782673e-06, + "loss": 0.8746, + "step": 258 + }, + { + "epoch": 0.05, + "learning_rate": 2.988587550054605e-06, + "loss": 1.1749, + "step": 259 + }, + { + "epoch": 0.05, + "learning_rate": 2.9885329450309428e-06, + "loss": 2.7803, + "step": 260 + }, + { + "epoch": 0.05, + "learning_rate": 2.9884783400072807e-06, + "loss": 1.4111, + "step": 261 + }, + { + "epoch": 0.05, + "learning_rate": 2.9884237349836187e-06, + "loss": 1.2873, + "step": 262 + }, + { + "epoch": 0.05, + "learning_rate": 2.9883691299599566e-06, + "loss": 0.9672, + "step": 263 + }, + { + "epoch": 0.05, + "learning_rate": 2.988314524936294e-06, + "loss": 1.2286, + "step": 264 + }, + { + "epoch": 0.05, + "learning_rate": 2.988259919912632e-06, + "loss": 1.5252, + "step": 265 + }, + { + "epoch": 0.05, + "learning_rate": 2.98820531488897e-06, + "loss": 2.15, + "step": 266 + }, + { + "epoch": 0.05, + "learning_rate": 2.9881507098653075e-06, + "loss": 1.6712, + "step": 267 + }, + { + "epoch": 0.05, + "learning_rate": 2.9880961048416455e-06, + "loss": 0.9963, + "step": 268 + }, + { + "epoch": 0.05, + "learning_rate": 2.9880414998179834e-06, + "loss": 2.8993, + "step": 269 + }, + { + "epoch": 0.05, + "learning_rate": 2.987986894794321e-06, + "loss": 2.0765, + "step": 270 + }, + { + "epoch": 0.05, + "learning_rate": 2.987932289770659e-06, + "loss": 1.7782, + "step": 271 + }, + { + "epoch": 0.05, + "learning_rate": 2.987877684746997e-06, + "loss": 0.559, + "step": 272 + }, + { + "epoch": 0.05, + "learning_rate": 2.9878230797233348e-06, + "loss": 0.9201, + "step": 273 + }, + { + "epoch": 0.05, + "learning_rate": 2.9877684746996727e-06, + "loss": 0.361, + "step": 274 + }, + { + "epoch": 0.05, + "learning_rate": 2.9877138696760102e-06, + "loss": 1.513, + "step": 275 + }, + { + "epoch": 0.05, + "learning_rate": 2.987659264652348e-06, + "loss": 0.9958, + "step": 276 + }, + { + "epoch": 0.05, + "learning_rate": 2.987604659628686e-06, + "loss": 1.5439, + "step": 277 + }, + { + "epoch": 0.05, + "learning_rate": 2.9875500546050236e-06, + "loss": 0.3045, + "step": 278 + }, + { + "epoch": 0.05, + "learning_rate": 2.9874954495813616e-06, + "loss": 2.1196, + "step": 279 + }, + { + "epoch": 0.05, + "learning_rate": 2.9874408445576995e-06, + "loss": 0.4796, + "step": 280 + }, + { + "epoch": 0.05, + "learning_rate": 2.987386239534037e-06, + "loss": 1.1776, + "step": 281 + }, + { + "epoch": 0.05, + "learning_rate": 2.987331634510375e-06, + "loss": 0.4671, + "step": 282 + }, + { + "epoch": 0.05, + "learning_rate": 2.987277029486713e-06, + "loss": 0.8933, + "step": 283 + }, + { + "epoch": 0.05, + "learning_rate": 2.9872224244630505e-06, + "loss": 0.3449, + "step": 284 + }, + { + "epoch": 0.05, + "learning_rate": 2.9871678194393884e-06, + "loss": 1.1821, + "step": 285 + }, + { + "epoch": 0.05, + "learning_rate": 2.9871132144157263e-06, + "loss": 3.5541, + "step": 286 + }, + { + "epoch": 0.05, + "learning_rate": 2.9870586093920643e-06, + "loss": 0.9707, + "step": 287 + }, + { + "epoch": 0.05, + "learning_rate": 2.9870040043684022e-06, + "loss": 0.9626, + "step": 288 + }, + { + "epoch": 0.05, + "learning_rate": 2.9869493993447397e-06, + "loss": 1.3094, + "step": 289 + }, + { + "epoch": 0.05, + "learning_rate": 2.9868947943210777e-06, + "loss": 0.11, + "step": 290 + }, + { + "epoch": 0.05, + "learning_rate": 2.9868401892974156e-06, + "loss": 0.7097, + "step": 291 + }, + { + "epoch": 0.05, + "learning_rate": 2.986785584273753e-06, + "loss": 1.4984, + "step": 292 + }, + { + "epoch": 0.05, + "learning_rate": 2.986730979250091e-06, + "loss": 1.5047, + "step": 293 + }, + { + "epoch": 0.05, + "learning_rate": 2.986676374226429e-06, + "loss": 2.8089, + "step": 294 + }, + { + "epoch": 0.05, + "learning_rate": 2.9866217692027666e-06, + "loss": 2.685, + "step": 295 + }, + { + "epoch": 0.05, + "learning_rate": 2.9865671641791045e-06, + "loss": 1.2786, + "step": 296 + }, + { + "epoch": 0.05, + "learning_rate": 2.9865125591554424e-06, + "loss": 0.4703, + "step": 297 + }, + { + "epoch": 0.05, + "learning_rate": 2.98645795413178e-06, + "loss": 2.14, + "step": 298 + }, + { + "epoch": 0.05, + "learning_rate": 2.986403349108118e-06, + "loss": 1.5662, + "step": 299 + }, + { + "epoch": 0.05, + "learning_rate": 2.986348744084456e-06, + "loss": 2.1308, + "step": 300 + }, + { + "epoch": 0.05, + "learning_rate": 2.986294139060794e-06, + "loss": 0.0734, + "step": 301 + }, + { + "epoch": 0.05, + "learning_rate": 2.9862395340371317e-06, + "loss": 0.5396, + "step": 302 + }, + { + "epoch": 0.06, + "learning_rate": 2.9861849290134693e-06, + "loss": 0.2219, + "step": 303 + }, + { + "epoch": 0.06, + "learning_rate": 2.986130323989807e-06, + "loss": 0.9931, + "step": 304 + }, + { + "epoch": 0.06, + "learning_rate": 2.986075718966145e-06, + "loss": 2.1521, + "step": 305 + }, + { + "epoch": 0.06, + "learning_rate": 2.9860211139424827e-06, + "loss": 2.7448, + "step": 306 + }, + { + "epoch": 0.06, + "learning_rate": 2.9859665089188206e-06, + "loss": 0.5044, + "step": 307 + }, + { + "epoch": 0.06, + "learning_rate": 2.9859119038951586e-06, + "loss": 0.5769, + "step": 308 + }, + { + "epoch": 0.06, + "learning_rate": 2.985857298871496e-06, + "loss": 1.0114, + "step": 309 + }, + { + "epoch": 0.06, + "learning_rate": 2.985802693847834e-06, + "loss": 0.5307, + "step": 310 + }, + { + "epoch": 0.06, + "learning_rate": 2.985748088824172e-06, + "loss": 0.2442, + "step": 311 + }, + { + "epoch": 0.06, + "learning_rate": 2.9856934838005095e-06, + "loss": 1.0447, + "step": 312 + }, + { + "epoch": 0.06, + "learning_rate": 2.9856388787768474e-06, + "loss": 3.0317, + "step": 313 + }, + { + "epoch": 0.06, + "learning_rate": 2.9855842737531854e-06, + "loss": 1.4896, + "step": 314 + }, + { + "epoch": 0.06, + "learning_rate": 2.9855296687295233e-06, + "loss": 2.612, + "step": 315 + }, + { + "epoch": 0.06, + "learning_rate": 2.9854750637058613e-06, + "loss": 1.2261, + "step": 316 + }, + { + "epoch": 0.06, + "learning_rate": 2.9854204586821988e-06, + "loss": 1.4028, + "step": 317 + }, + { + "epoch": 0.06, + "learning_rate": 2.9853658536585367e-06, + "loss": 0.4865, + "step": 318 + }, + { + "epoch": 0.06, + "learning_rate": 2.9853112486348747e-06, + "loss": 3.164, + "step": 319 + }, + { + "epoch": 0.06, + "learning_rate": 2.985256643611212e-06, + "loss": 2.0928, + "step": 320 + }, + { + "epoch": 0.06, + "learning_rate": 2.98520203858755e-06, + "loss": 0.8046, + "step": 321 + }, + { + "epoch": 0.06, + "learning_rate": 2.985147433563888e-06, + "loss": 1.3557, + "step": 322 + }, + { + "epoch": 0.06, + "learning_rate": 2.9850928285402256e-06, + "loss": 1.815, + "step": 323 + }, + { + "epoch": 0.06, + "learning_rate": 2.9850382235165635e-06, + "loss": 2.0869, + "step": 324 + }, + { + "epoch": 0.06, + "learning_rate": 2.9849836184929015e-06, + "loss": 0.9576, + "step": 325 + }, + { + "epoch": 0.06, + "learning_rate": 2.984929013469239e-06, + "loss": 0.4615, + "step": 326 + }, + { + "epoch": 0.06, + "learning_rate": 2.984874408445577e-06, + "loss": 1.652, + "step": 327 + }, + { + "epoch": 0.06, + "learning_rate": 2.984819803421915e-06, + "loss": 2.3108, + "step": 328 + }, + { + "epoch": 0.06, + "learning_rate": 2.984765198398253e-06, + "loss": 1.2138, + "step": 329 + }, + { + "epoch": 0.06, + "learning_rate": 2.9847105933745908e-06, + "loss": 1.0024, + "step": 330 + }, + { + "epoch": 0.06, + "learning_rate": 2.9846559883509283e-06, + "loss": 0.9707, + "step": 331 + }, + { + "epoch": 0.06, + "learning_rate": 2.9846013833272662e-06, + "loss": 1.3537, + "step": 332 + }, + { + "epoch": 0.06, + "learning_rate": 2.984546778303604e-06, + "loss": 3.0601, + "step": 333 + }, + { + "epoch": 0.06, + "learning_rate": 2.9844921732799417e-06, + "loss": 1.219, + "step": 334 + }, + { + "epoch": 0.06, + "learning_rate": 2.9844375682562796e-06, + "loss": 1.7781, + "step": 335 + }, + { + "epoch": 0.06, + "learning_rate": 2.9843829632326176e-06, + "loss": 2.3294, + "step": 336 + }, + { + "epoch": 0.06, + "learning_rate": 2.984328358208955e-06, + "loss": 1.1651, + "step": 337 + }, + { + "epoch": 0.06, + "learning_rate": 2.984273753185293e-06, + "loss": 1.4017, + "step": 338 + }, + { + "epoch": 0.06, + "learning_rate": 2.984219148161631e-06, + "loss": 0.7239, + "step": 339 + }, + { + "epoch": 0.06, + "learning_rate": 2.9841645431379685e-06, + "loss": 1.7129, + "step": 340 + }, + { + "epoch": 0.06, + "learning_rate": 2.9841099381143065e-06, + "loss": 0.3286, + "step": 341 + }, + { + "epoch": 0.06, + "learning_rate": 2.9840553330906444e-06, + "loss": 1.0801, + "step": 342 + }, + { + "epoch": 0.06, + "learning_rate": 2.9840007280669823e-06, + "loss": 0.4647, + "step": 343 + }, + { + "epoch": 0.06, + "learning_rate": 2.9839461230433203e-06, + "loss": 2.0866, + "step": 344 + }, + { + "epoch": 0.06, + "learning_rate": 2.983891518019658e-06, + "loss": 1.3799, + "step": 345 + }, + { + "epoch": 0.06, + "learning_rate": 2.9838369129959957e-06, + "loss": 0.9881, + "step": 346 + }, + { + "epoch": 0.06, + "learning_rate": 2.9837823079723337e-06, + "loss": 1.4527, + "step": 347 + }, + { + "epoch": 0.06, + "learning_rate": 2.983727702948671e-06, + "loss": 1.9795, + "step": 348 + }, + { + "epoch": 0.06, + "learning_rate": 2.983673097925009e-06, + "loss": 2.5974, + "step": 349 + }, + { + "epoch": 0.06, + "learning_rate": 2.983618492901347e-06, + "loss": 0.5438, + "step": 350 + }, + { + "epoch": 0.06, + "learning_rate": 2.9835638878776846e-06, + "loss": 0.6764, + "step": 351 + }, + { + "epoch": 0.06, + "learning_rate": 2.9835092828540226e-06, + "loss": 0.9812, + "step": 352 + }, + { + "epoch": 0.06, + "learning_rate": 2.9834546778303605e-06, + "loss": 1.5627, + "step": 353 + }, + { + "epoch": 0.06, + "learning_rate": 2.9834000728066984e-06, + "loss": 0.1872, + "step": 354 + }, + { + "epoch": 0.06, + "learning_rate": 2.9833454677830364e-06, + "loss": 0.5184, + "step": 355 + }, + { + "epoch": 0.06, + "learning_rate": 2.983290862759374e-06, + "loss": 1.3999, + "step": 356 + }, + { + "epoch": 0.06, + "learning_rate": 2.983236257735712e-06, + "loss": 1.5437, + "step": 357 + }, + { + "epoch": 0.07, + "learning_rate": 2.98318165271205e-06, + "loss": 0.1432, + "step": 358 + }, + { + "epoch": 0.07, + "learning_rate": 2.9831270476883873e-06, + "loss": 0.4452, + "step": 359 + }, + { + "epoch": 0.07, + "learning_rate": 2.9830724426647253e-06, + "loss": 1.1858, + "step": 360 + }, + { + "epoch": 0.07, + "learning_rate": 2.983017837641063e-06, + "loss": 2.4206, + "step": 361 + }, + { + "epoch": 0.07, + "learning_rate": 2.9829632326174007e-06, + "loss": 0.8721, + "step": 362 + }, + { + "epoch": 0.07, + "learning_rate": 2.9829086275937387e-06, + "loss": 0.1135, + "step": 363 + }, + { + "epoch": 0.07, + "learning_rate": 2.9828540225700766e-06, + "loss": 1.7062, + "step": 364 + }, + { + "epoch": 0.07, + "learning_rate": 2.982799417546414e-06, + "loss": 2.2479, + "step": 365 + }, + { + "epoch": 0.07, + "learning_rate": 2.982744812522752e-06, + "loss": 0.841, + "step": 366 + }, + { + "epoch": 0.07, + "learning_rate": 2.98269020749909e-06, + "loss": 0.5775, + "step": 367 + }, + { + "epoch": 0.07, + "learning_rate": 2.982635602475428e-06, + "loss": 0.7531, + "step": 368 + }, + { + "epoch": 0.07, + "learning_rate": 2.982580997451766e-06, + "loss": 0.6448, + "step": 369 + }, + { + "epoch": 0.07, + "learning_rate": 2.9825263924281034e-06, + "loss": 2.2061, + "step": 370 + }, + { + "epoch": 0.07, + "learning_rate": 2.9824717874044414e-06, + "loss": 1.0608, + "step": 371 + }, + { + "epoch": 0.07, + "learning_rate": 2.9824171823807793e-06, + "loss": 1.1346, + "step": 372 + }, + { + "epoch": 0.07, + "learning_rate": 2.982362577357117e-06, + "loss": 0.7021, + "step": 373 + }, + { + "epoch": 0.07, + "learning_rate": 2.9823079723334548e-06, + "loss": 0.1281, + "step": 374 + }, + { + "epoch": 0.07, + "learning_rate": 2.9822533673097927e-06, + "loss": 1.8289, + "step": 375 + }, + { + "epoch": 0.07, + "learning_rate": 2.9821987622861302e-06, + "loss": 2.0955, + "step": 376 + }, + { + "epoch": 0.07, + "learning_rate": 2.982144157262468e-06, + "loss": 0.2107, + "step": 377 + }, + { + "epoch": 0.07, + "learning_rate": 2.982089552238806e-06, + "loss": 0.3628, + "step": 378 + }, + { + "epoch": 0.07, + "learning_rate": 2.9820349472151436e-06, + "loss": 1.2399, + "step": 379 + }, + { + "epoch": 0.07, + "learning_rate": 2.9819803421914816e-06, + "loss": 1.6739, + "step": 380 + }, + { + "epoch": 0.07, + "learning_rate": 2.9819257371678195e-06, + "loss": 1.2921, + "step": 381 + }, + { + "epoch": 0.07, + "learning_rate": 2.9818711321441575e-06, + "loss": 0.4113, + "step": 382 + }, + { + "epoch": 0.07, + "learning_rate": 2.9818165271204954e-06, + "loss": 3.0145, + "step": 383 + }, + { + "epoch": 0.07, + "learning_rate": 2.981761922096833e-06, + "loss": 2.4546, + "step": 384 + }, + { + "epoch": 0.07, + "learning_rate": 2.981707317073171e-06, + "loss": 0.1051, + "step": 385 + }, + { + "epoch": 0.07, + "learning_rate": 2.981652712049509e-06, + "loss": 1.4639, + "step": 386 + }, + { + "epoch": 0.07, + "learning_rate": 2.9815981070258463e-06, + "loss": 1.8325, + "step": 387 + }, + { + "epoch": 0.07, + "learning_rate": 2.9815435020021843e-06, + "loss": 1.5397, + "step": 388 + }, + { + "epoch": 0.07, + "learning_rate": 2.9814888969785222e-06, + "loss": 1.2867, + "step": 389 + }, + { + "epoch": 0.07, + "learning_rate": 2.9814342919548597e-06, + "loss": 1.1751, + "step": 390 + }, + { + "epoch": 0.07, + "learning_rate": 2.9813796869311977e-06, + "loss": 2.0222, + "step": 391 + }, + { + "epoch": 0.07, + "learning_rate": 2.9813250819075356e-06, + "loss": 2.3381, + "step": 392 + }, + { + "epoch": 0.07, + "learning_rate": 2.981270476883873e-06, + "loss": 2.1074, + "step": 393 + }, + { + "epoch": 0.07, + "learning_rate": 2.981215871860211e-06, + "loss": 0.7101, + "step": 394 + }, + { + "epoch": 0.07, + "learning_rate": 2.981161266836549e-06, + "loss": 2.9177, + "step": 395 + }, + { + "epoch": 0.07, + "learning_rate": 2.981106661812887e-06, + "loss": 0.7427, + "step": 396 + }, + { + "epoch": 0.07, + "learning_rate": 2.981052056789225e-06, + "loss": 1.0021, + "step": 397 + }, + { + "epoch": 0.07, + "learning_rate": 2.9809974517655624e-06, + "loss": 1.1259, + "step": 398 + }, + { + "epoch": 0.07, + "learning_rate": 2.9809428467419004e-06, + "loss": 1.9746, + "step": 399 + }, + { + "epoch": 0.07, + "learning_rate": 2.9808882417182383e-06, + "loss": 0.9371, + "step": 400 + }, + { + "epoch": 0.07, + "learning_rate": 2.980833636694576e-06, + "loss": 1.3733, + "step": 401 + }, + { + "epoch": 0.07, + "learning_rate": 2.980779031670914e-06, + "loss": 0.6086, + "step": 402 + }, + { + "epoch": 0.07, + "learning_rate": 2.9807244266472517e-06, + "loss": 1.6791, + "step": 403 + }, + { + "epoch": 0.07, + "learning_rate": 2.9806698216235893e-06, + "loss": 1.6923, + "step": 404 + }, + { + "epoch": 0.07, + "learning_rate": 2.980615216599927e-06, + "loss": 2.0209, + "step": 405 + }, + { + "epoch": 0.07, + "learning_rate": 2.980560611576265e-06, + "loss": 0.1009, + "step": 406 + }, + { + "epoch": 0.07, + "learning_rate": 2.9805060065526027e-06, + "loss": 0.1538, + "step": 407 + }, + { + "epoch": 0.07, + "learning_rate": 2.9804514015289406e-06, + "loss": 0.9717, + "step": 408 + }, + { + "epoch": 0.07, + "learning_rate": 2.9803967965052786e-06, + "loss": 0.3154, + "step": 409 + }, + { + "epoch": 0.07, + "learning_rate": 2.9803421914816165e-06, + "loss": 1.2343, + "step": 410 + }, + { + "epoch": 0.07, + "learning_rate": 2.9802875864579544e-06, + "loss": 0.1033, + "step": 411 + }, + { + "epoch": 0.07, + "learning_rate": 2.980232981434292e-06, + "loss": 0.834, + "step": 412 + }, + { + "epoch": 0.08, + "learning_rate": 2.98017837641063e-06, + "loss": 1.1077, + "step": 413 + }, + { + "epoch": 0.08, + "learning_rate": 2.980123771386968e-06, + "loss": 0.3826, + "step": 414 + }, + { + "epoch": 0.08, + "learning_rate": 2.9800691663633054e-06, + "loss": 2.0291, + "step": 415 + }, + { + "epoch": 0.08, + "learning_rate": 2.9800145613396433e-06, + "loss": 1.4407, + "step": 416 + }, + { + "epoch": 0.08, + "learning_rate": 2.9799599563159813e-06, + "loss": 1.9213, + "step": 417 + }, + { + "epoch": 0.08, + "learning_rate": 2.9799053512923188e-06, + "loss": 1.527, + "step": 418 + }, + { + "epoch": 0.08, + "learning_rate": 2.9798507462686567e-06, + "loss": 0.6119, + "step": 419 + }, + { + "epoch": 0.08, + "learning_rate": 2.9797961412449947e-06, + "loss": 1.5264, + "step": 420 + }, + { + "epoch": 0.08, + "learning_rate": 2.979741536221332e-06, + "loss": 0.3318, + "step": 421 + }, + { + "epoch": 0.08, + "learning_rate": 2.97968693119767e-06, + "loss": 0.397, + "step": 422 + }, + { + "epoch": 0.08, + "learning_rate": 2.979632326174008e-06, + "loss": 2.7232, + "step": 423 + }, + { + "epoch": 0.08, + "learning_rate": 2.979577721150346e-06, + "loss": 0.1474, + "step": 424 + }, + { + "epoch": 0.08, + "learning_rate": 2.979523116126684e-06, + "loss": 0.5597, + "step": 425 + }, + { + "epoch": 0.08, + "learning_rate": 2.9794685111030215e-06, + "loss": 1.3936, + "step": 426 + }, + { + "epoch": 0.08, + "learning_rate": 2.9794139060793594e-06, + "loss": 1.4083, + "step": 427 + }, + { + "epoch": 0.08, + "learning_rate": 2.9793593010556974e-06, + "loss": 1.0996, + "step": 428 + }, + { + "epoch": 0.08, + "learning_rate": 2.979304696032035e-06, + "loss": 1.8221, + "step": 429 + }, + { + "epoch": 0.08, + "learning_rate": 2.979250091008373e-06, + "loss": 1.0265, + "step": 430 + }, + { + "epoch": 0.08, + "learning_rate": 2.9791954859847108e-06, + "loss": 0.2511, + "step": 431 + }, + { + "epoch": 0.08, + "learning_rate": 2.9791408809610483e-06, + "loss": 1.2038, + "step": 432 + }, + { + "epoch": 0.08, + "learning_rate": 2.9790862759373862e-06, + "loss": 1.0254, + "step": 433 + }, + { + "epoch": 0.08, + "learning_rate": 2.979031670913724e-06, + "loss": 2.531, + "step": 434 + }, + { + "epoch": 0.08, + "learning_rate": 2.978977065890062e-06, + "loss": 1.8889, + "step": 435 + }, + { + "epoch": 0.08, + "learning_rate": 2.9789224608664e-06, + "loss": 2.5037, + "step": 436 + }, + { + "epoch": 0.08, + "learning_rate": 2.9788678558427376e-06, + "loss": 0.1841, + "step": 437 + }, + { + "epoch": 0.08, + "learning_rate": 2.9788132508190755e-06, + "loss": 1.6687, + "step": 438 + }, + { + "epoch": 0.08, + "learning_rate": 2.9787586457954135e-06, + "loss": 3.1398, + "step": 439 + }, + { + "epoch": 0.08, + "learning_rate": 2.978704040771751e-06, + "loss": 2.1109, + "step": 440 + }, + { + "epoch": 0.08, + "learning_rate": 2.978649435748089e-06, + "loss": 0.433, + "step": 441 + }, + { + "epoch": 0.08, + "learning_rate": 2.978594830724427e-06, + "loss": 1.4674, + "step": 442 + }, + { + "epoch": 0.08, + "learning_rate": 2.9785402257007644e-06, + "loss": 1.9827, + "step": 443 + }, + { + "epoch": 0.08, + "learning_rate": 2.9784856206771023e-06, + "loss": 0.2964, + "step": 444 + }, + { + "epoch": 0.08, + "learning_rate": 2.9784310156534403e-06, + "loss": 0.7274, + "step": 445 + }, + { + "epoch": 0.08, + "learning_rate": 2.978376410629778e-06, + "loss": 1.2262, + "step": 446 + }, + { + "epoch": 0.08, + "learning_rate": 2.9783218056061157e-06, + "loss": 0.8901, + "step": 447 + }, + { + "epoch": 0.08, + "learning_rate": 2.9782672005824537e-06, + "loss": 0.4223, + "step": 448 + }, + { + "epoch": 0.08, + "learning_rate": 2.9782125955587916e-06, + "loss": 1.1614, + "step": 449 + }, + { + "epoch": 0.08, + "learning_rate": 2.9781579905351296e-06, + "loss": 1.2633, + "step": 450 + }, + { + "epoch": 0.08, + "learning_rate": 2.978103385511467e-06, + "loss": 1.3857, + "step": 451 + }, + { + "epoch": 0.08, + "learning_rate": 2.978048780487805e-06, + "loss": 1.711, + "step": 452 + }, + { + "epoch": 0.08, + "learning_rate": 2.977994175464143e-06, + "loss": 1.4926, + "step": 453 + }, + { + "epoch": 0.08, + "learning_rate": 2.9779395704404805e-06, + "loss": 0.9174, + "step": 454 + }, + { + "epoch": 0.08, + "learning_rate": 2.9778849654168184e-06, + "loss": 0.9942, + "step": 455 + }, + { + "epoch": 0.08, + "learning_rate": 2.9778303603931564e-06, + "loss": 0.0925, + "step": 456 + }, + { + "epoch": 0.08, + "learning_rate": 2.977775755369494e-06, + "loss": 0.4576, + "step": 457 + }, + { + "epoch": 0.08, + "learning_rate": 2.977721150345832e-06, + "loss": 0.1882, + "step": 458 + }, + { + "epoch": 0.08, + "learning_rate": 2.97766654532217e-06, + "loss": 0.7668, + "step": 459 + }, + { + "epoch": 0.08, + "learning_rate": 2.9776119402985073e-06, + "loss": 1.4846, + "step": 460 + }, + { + "epoch": 0.08, + "learning_rate": 2.9775573352748453e-06, + "loss": 0.3552, + "step": 461 + }, + { + "epoch": 0.08, + "learning_rate": 2.977502730251183e-06, + "loss": 0.6346, + "step": 462 + }, + { + "epoch": 0.08, + "learning_rate": 2.977448125227521e-06, + "loss": 1.2715, + "step": 463 + }, + { + "epoch": 0.08, + "learning_rate": 2.977393520203859e-06, + "loss": 1.8845, + "step": 464 + }, + { + "epoch": 0.08, + "learning_rate": 2.9773389151801966e-06, + "loss": 0.0658, + "step": 465 + }, + { + "epoch": 0.08, + "learning_rate": 2.9772843101565346e-06, + "loss": 1.6045, + "step": 466 + }, + { + "epoch": 0.08, + "learning_rate": 2.9772297051328725e-06, + "loss": 1.8064, + "step": 467 + }, + { + "epoch": 0.09, + "learning_rate": 2.97717510010921e-06, + "loss": 1.9028, + "step": 468 + }, + { + "epoch": 0.09, + "learning_rate": 2.977120495085548e-06, + "loss": 1.2388, + "step": 469 + }, + { + "epoch": 0.09, + "learning_rate": 2.977065890061886e-06, + "loss": 0.8358, + "step": 470 + }, + { + "epoch": 0.09, + "learning_rate": 2.9770112850382234e-06, + "loss": 0.4642, + "step": 471 + }, + { + "epoch": 0.09, + "learning_rate": 2.9769566800145614e-06, + "loss": 1.4104, + "step": 472 + }, + { + "epoch": 0.09, + "learning_rate": 2.9769020749908993e-06, + "loss": 0.9037, + "step": 473 + }, + { + "epoch": 0.09, + "learning_rate": 2.976847469967237e-06, + "loss": 1.8288, + "step": 474 + }, + { + "epoch": 0.09, + "learning_rate": 2.9767928649435748e-06, + "loss": 1.0714, + "step": 475 + }, + { + "epoch": 0.09, + "learning_rate": 2.9767382599199127e-06, + "loss": 1.9357, + "step": 476 + }, + { + "epoch": 0.09, + "learning_rate": 2.9766836548962507e-06, + "loss": 1.7033, + "step": 477 + }, + { + "epoch": 0.09, + "learning_rate": 2.9766290498725886e-06, + "loss": 1.6362, + "step": 478 + }, + { + "epoch": 0.09, + "learning_rate": 2.976574444848926e-06, + "loss": 1.7232, + "step": 479 + }, + { + "epoch": 0.09, + "learning_rate": 2.976519839825264e-06, + "loss": 1.0029, + "step": 480 + }, + { + "epoch": 0.09, + "learning_rate": 2.976465234801602e-06, + "loss": 0.107, + "step": 481 + }, + { + "epoch": 0.09, + "learning_rate": 2.9764106297779395e-06, + "loss": 0.5101, + "step": 482 + }, + { + "epoch": 0.09, + "learning_rate": 2.9763560247542775e-06, + "loss": 0.0918, + "step": 483 + }, + { + "epoch": 0.09, + "learning_rate": 2.9763014197306154e-06, + "loss": 1.5692, + "step": 484 + }, + { + "epoch": 0.09, + "learning_rate": 2.976246814706953e-06, + "loss": 2.9436, + "step": 485 + }, + { + "epoch": 0.09, + "learning_rate": 2.976192209683291e-06, + "loss": 1.2087, + "step": 486 + }, + { + "epoch": 0.09, + "learning_rate": 2.976137604659629e-06, + "loss": 3.015, + "step": 487 + }, + { + "epoch": 0.09, + "learning_rate": 2.9760829996359663e-06, + "loss": 2.8106, + "step": 488 + }, + { + "epoch": 0.09, + "learning_rate": 2.9760283946123043e-06, + "loss": 1.395, + "step": 489 + }, + { + "epoch": 0.09, + "learning_rate": 2.9759737895886422e-06, + "loss": 0.649, + "step": 490 + }, + { + "epoch": 0.09, + "learning_rate": 2.97591918456498e-06, + "loss": 1.4251, + "step": 491 + }, + { + "epoch": 0.09, + "learning_rate": 2.975864579541318e-06, + "loss": 1.4131, + "step": 492 + }, + { + "epoch": 0.09, + "learning_rate": 2.9758099745176556e-06, + "loss": 1.1041, + "step": 493 + }, + { + "epoch": 0.09, + "learning_rate": 2.9757553694939936e-06, + "loss": 1.9382, + "step": 494 + }, + { + "epoch": 0.09, + "learning_rate": 2.9757007644703315e-06, + "loss": 1.1813, + "step": 495 + }, + { + "epoch": 0.09, + "learning_rate": 2.975646159446669e-06, + "loss": 0.5994, + "step": 496 + }, + { + "epoch": 0.09, + "learning_rate": 2.975591554423007e-06, + "loss": 0.733, + "step": 497 + }, + { + "epoch": 0.09, + "learning_rate": 2.975536949399345e-06, + "loss": 2.4371, + "step": 498 + }, + { + "epoch": 0.09, + "learning_rate": 2.9754823443756825e-06, + "loss": 0.0842, + "step": 499 + }, + { + "epoch": 0.09, + "learning_rate": 2.9754277393520204e-06, + "loss": 0.787, + "step": 500 + }, + { + "epoch": 0.09, + "learning_rate": 2.9753731343283583e-06, + "loss": 2.5006, + "step": 501 + }, + { + "epoch": 0.09, + "learning_rate": 2.975318529304696e-06, + "loss": 0.6975, + "step": 502 + }, + { + "epoch": 0.09, + "learning_rate": 2.975263924281034e-06, + "loss": 1.5164, + "step": 503 + }, + { + "epoch": 0.09, + "learning_rate": 2.9752093192573717e-06, + "loss": 1.0925, + "step": 504 + }, + { + "epoch": 0.09, + "learning_rate": 2.9751547142337097e-06, + "loss": 2.9295, + "step": 505 + }, + { + "epoch": 0.09, + "learning_rate": 2.9751001092100476e-06, + "loss": 0.9253, + "step": 506 + }, + { + "epoch": 0.09, + "learning_rate": 2.975045504186385e-06, + "loss": 0.6316, + "step": 507 + }, + { + "epoch": 0.09, + "learning_rate": 2.974990899162723e-06, + "loss": 1.7724, + "step": 508 + }, + { + "epoch": 0.09, + "learning_rate": 2.974936294139061e-06, + "loss": 1.336, + "step": 509 + }, + { + "epoch": 0.09, + "learning_rate": 2.9748816891153986e-06, + "loss": 1.0048, + "step": 510 + }, + { + "epoch": 0.09, + "learning_rate": 2.9748270840917365e-06, + "loss": 1.8142, + "step": 511 + }, + { + "epoch": 0.09, + "learning_rate": 2.9747724790680744e-06, + "loss": 0.4202, + "step": 512 + }, + { + "epoch": 0.09, + "learning_rate": 2.974717874044412e-06, + "loss": 0.9853, + "step": 513 + }, + { + "epoch": 0.09, + "learning_rate": 2.97466326902075e-06, + "loss": 0.3574, + "step": 514 + }, + { + "epoch": 0.09, + "learning_rate": 2.974608663997088e-06, + "loss": 2.1146, + "step": 515 + }, + { + "epoch": 0.09, + "learning_rate": 2.974554058973426e-06, + "loss": 0.9518, + "step": 516 + }, + { + "epoch": 0.09, + "learning_rate": 2.9744994539497633e-06, + "loss": 0.7479, + "step": 517 + }, + { + "epoch": 0.09, + "learning_rate": 2.9744448489261013e-06, + "loss": 0.2197, + "step": 518 + }, + { + "epoch": 0.09, + "learning_rate": 2.974390243902439e-06, + "loss": 0.1648, + "step": 519 + }, + { + "epoch": 0.09, + "learning_rate": 2.974335638878777e-06, + "loss": 0.8783, + "step": 520 + }, + { + "epoch": 0.09, + "learning_rate": 2.9742810338551147e-06, + "loss": 0.4891, + "step": 521 + }, + { + "epoch": 0.09, + "learning_rate": 2.9742264288314526e-06, + "loss": 0.8804, + "step": 522 + }, + { + "epoch": 0.1, + "learning_rate": 2.9741718238077906e-06, + "loss": 0.5651, + "step": 523 + }, + { + "epoch": 0.1, + "learning_rate": 2.974117218784128e-06, + "loss": 0.1688, + "step": 524 + }, + { + "epoch": 0.1, + "learning_rate": 2.974062613760466e-06, + "loss": 0.5991, + "step": 525 + }, + { + "epoch": 0.1, + "learning_rate": 2.974008008736804e-06, + "loss": 1.0054, + "step": 526 + }, + { + "epoch": 0.1, + "learning_rate": 2.9739534037131415e-06, + "loss": 1.1638, + "step": 527 + }, + { + "epoch": 0.1, + "learning_rate": 2.9738987986894794e-06, + "loss": 1.5509, + "step": 528 + }, + { + "epoch": 0.1, + "learning_rate": 2.9738441936658174e-06, + "loss": 2.0955, + "step": 529 + }, + { + "epoch": 0.1, + "learning_rate": 2.9737895886421553e-06, + "loss": 1.0453, + "step": 530 + }, + { + "epoch": 0.1, + "learning_rate": 2.9737349836184933e-06, + "loss": 0.7044, + "step": 531 + }, + { + "epoch": 0.1, + "learning_rate": 2.9736803785948308e-06, + "loss": 1.3964, + "step": 532 + }, + { + "epoch": 0.1, + "learning_rate": 2.9736257735711687e-06, + "loss": 1.4311, + "step": 533 + }, + { + "epoch": 0.1, + "learning_rate": 2.9735711685475067e-06, + "loss": 1.4214, + "step": 534 + }, + { + "epoch": 0.1, + "learning_rate": 2.973516563523844e-06, + "loss": 0.6203, + "step": 535 + }, + { + "epoch": 0.1, + "learning_rate": 2.973461958500182e-06, + "loss": 1.8366, + "step": 536 + }, + { + "epoch": 0.1, + "learning_rate": 2.97340735347652e-06, + "loss": 1.2888, + "step": 537 + }, + { + "epoch": 0.1, + "learning_rate": 2.9733527484528576e-06, + "loss": 1.5895, + "step": 538 + }, + { + "epoch": 0.1, + "learning_rate": 2.9732981434291955e-06, + "loss": 1.3397, + "step": 539 + }, + { + "epoch": 0.1, + "learning_rate": 2.9732435384055335e-06, + "loss": 0.8638, + "step": 540 + }, + { + "epoch": 0.1, + "learning_rate": 2.973188933381871e-06, + "loss": 0.6343, + "step": 541 + }, + { + "epoch": 0.1, + "learning_rate": 2.973134328358209e-06, + "loss": 1.2441, + "step": 542 + }, + { + "epoch": 0.1, + "learning_rate": 2.973079723334547e-06, + "loss": 0.3392, + "step": 543 + }, + { + "epoch": 0.1, + "learning_rate": 2.973025118310885e-06, + "loss": 1.2516, + "step": 544 + }, + { + "epoch": 0.1, + "learning_rate": 2.9729705132872228e-06, + "loss": 2.4395, + "step": 545 + }, + { + "epoch": 0.1, + "learning_rate": 2.9729159082635603e-06, + "loss": 1.523, + "step": 546 + }, + { + "epoch": 0.1, + "learning_rate": 2.9728613032398982e-06, + "loss": 2.1792, + "step": 547 + }, + { + "epoch": 0.1, + "learning_rate": 2.972806698216236e-06, + "loss": 1.2725, + "step": 548 + }, + { + "epoch": 0.1, + "learning_rate": 2.9727520931925737e-06, + "loss": 2.1682, + "step": 549 + }, + { + "epoch": 0.1, + "learning_rate": 2.9726974881689116e-06, + "loss": 1.3773, + "step": 550 + }, + { + "epoch": 0.1, + "learning_rate": 2.9726428831452496e-06, + "loss": 0.4908, + "step": 551 + }, + { + "epoch": 0.1, + "learning_rate": 2.972588278121587e-06, + "loss": 0.7731, + "step": 552 + }, + { + "epoch": 0.1, + "learning_rate": 2.972533673097925e-06, + "loss": 0.6758, + "step": 553 + }, + { + "epoch": 0.1, + "learning_rate": 2.972479068074263e-06, + "loss": 1.0384, + "step": 554 + }, + { + "epoch": 0.1, + "learning_rate": 2.9724244630506005e-06, + "loss": 2.1615, + "step": 555 + }, + { + "epoch": 0.1, + "learning_rate": 2.9723698580269384e-06, + "loss": 0.8837, + "step": 556 + }, + { + "epoch": 0.1, + "learning_rate": 2.9723152530032764e-06, + "loss": 0.4609, + "step": 557 + }, + { + "epoch": 0.1, + "learning_rate": 2.9722606479796143e-06, + "loss": 0.4677, + "step": 558 + }, + { + "epoch": 0.1, + "learning_rate": 2.9722060429559523e-06, + "loss": 2.101, + "step": 559 + }, + { + "epoch": 0.1, + "learning_rate": 2.97215143793229e-06, + "loss": 1.7523, + "step": 560 + }, + { + "epoch": 0.1, + "learning_rate": 2.9720968329086277e-06, + "loss": 1.1541, + "step": 561 + }, + { + "epoch": 0.1, + "learning_rate": 2.9720422278849657e-06, + "loss": 1.4073, + "step": 562 + }, + { + "epoch": 0.1, + "learning_rate": 2.971987622861303e-06, + "loss": 0.5807, + "step": 563 + }, + { + "epoch": 0.1, + "learning_rate": 2.971933017837641e-06, + "loss": 2.8393, + "step": 564 + }, + { + "epoch": 0.1, + "learning_rate": 2.971878412813979e-06, + "loss": 1.2873, + "step": 565 + }, + { + "epoch": 0.1, + "learning_rate": 2.9718238077903166e-06, + "loss": 1.4802, + "step": 566 + }, + { + "epoch": 0.1, + "learning_rate": 2.9717692027666546e-06, + "loss": 0.1934, + "step": 567 + }, + { + "epoch": 0.1, + "learning_rate": 2.9717145977429925e-06, + "loss": 1.7146, + "step": 568 + }, + { + "epoch": 0.1, + "learning_rate": 2.97165999271933e-06, + "loss": 2.7439, + "step": 569 + }, + { + "epoch": 0.1, + "learning_rate": 2.971605387695668e-06, + "loss": 0.9016, + "step": 570 + }, + { + "epoch": 0.1, + "learning_rate": 2.971550782672006e-06, + "loss": 1.7779, + "step": 571 + }, + { + "epoch": 0.1, + "learning_rate": 2.971496177648344e-06, + "loss": 0.516, + "step": 572 + }, + { + "epoch": 0.1, + "learning_rate": 2.971441572624682e-06, + "loss": 0.2199, + "step": 573 + }, + { + "epoch": 0.1, + "learning_rate": 2.9713869676010193e-06, + "loss": 0.9214, + "step": 574 + }, + { + "epoch": 0.1, + "learning_rate": 2.9713323625773573e-06, + "loss": 1.1077, + "step": 575 + }, + { + "epoch": 0.1, + "learning_rate": 2.971277757553695e-06, + "loss": 0.2947, + "step": 576 + }, + { + "epoch": 0.1, + "learning_rate": 2.9712231525300327e-06, + "loss": 1.0809, + "step": 577 + }, + { + "epoch": 0.11, + "learning_rate": 2.9711685475063707e-06, + "loss": 1.0213, + "step": 578 + }, + { + "epoch": 0.11, + "learning_rate": 2.9711139424827086e-06, + "loss": 0.5028, + "step": 579 + }, + { + "epoch": 0.11, + "learning_rate": 2.971059337459046e-06, + "loss": 0.4814, + "step": 580 + }, + { + "epoch": 0.11, + "learning_rate": 2.971004732435384e-06, + "loss": 1.592, + "step": 581 + }, + { + "epoch": 0.11, + "learning_rate": 2.970950127411722e-06, + "loss": 0.3565, + "step": 582 + }, + { + "epoch": 0.11, + "learning_rate": 2.9708955223880595e-06, + "loss": 0.9211, + "step": 583 + }, + { + "epoch": 0.11, + "learning_rate": 2.9708409173643975e-06, + "loss": 1.1368, + "step": 584 + }, + { + "epoch": 0.11, + "learning_rate": 2.9707863123407354e-06, + "loss": 3.5467, + "step": 585 + }, + { + "epoch": 0.11, + "learning_rate": 2.9707317073170734e-06, + "loss": 0.5811, + "step": 586 + }, + { + "epoch": 0.11, + "learning_rate": 2.9706771022934113e-06, + "loss": 0.8972, + "step": 587 + }, + { + "epoch": 0.11, + "learning_rate": 2.970622497269749e-06, + "loss": 2.4861, + "step": 588 + }, + { + "epoch": 0.11, + "learning_rate": 2.9705678922460868e-06, + "loss": 0.2572, + "step": 589 + }, + { + "epoch": 0.11, + "learning_rate": 2.9705132872224247e-06, + "loss": 2.0007, + "step": 590 + }, + { + "epoch": 0.11, + "learning_rate": 2.9704586821987622e-06, + "loss": 1.1135, + "step": 591 + }, + { + "epoch": 0.11, + "learning_rate": 2.9704040771751e-06, + "loss": 1.9895, + "step": 592 + }, + { + "epoch": 0.11, + "learning_rate": 2.970349472151438e-06, + "loss": 2.1645, + "step": 593 + }, + { + "epoch": 0.11, + "learning_rate": 2.9702948671277756e-06, + "loss": 1.7544, + "step": 594 + }, + { + "epoch": 0.11, + "learning_rate": 2.9702402621041136e-06, + "loss": 0.5942, + "step": 595 + }, + { + "epoch": 0.11, + "learning_rate": 2.970185657080451e-06, + "loss": 1.0052, + "step": 596 + }, + { + "epoch": 0.11, + "learning_rate": 2.970131052056789e-06, + "loss": 1.0894, + "step": 597 + }, + { + "epoch": 0.11, + "learning_rate": 2.970076447033127e-06, + "loss": 1.097, + "step": 598 + }, + { + "epoch": 0.11, + "learning_rate": 2.970021842009465e-06, + "loss": 0.4952, + "step": 599 + }, + { + "epoch": 0.11, + "learning_rate": 2.969967236985803e-06, + "loss": 0.8215, + "step": 600 + }, + { + "epoch": 0.11, + "learning_rate": 2.969912631962141e-06, + "loss": 1.8314, + "step": 601 + }, + { + "epoch": 0.11, + "learning_rate": 2.9698580269384783e-06, + "loss": 1.8294, + "step": 602 + }, + { + "epoch": 0.11, + "learning_rate": 2.9698034219148163e-06, + "loss": 1.2225, + "step": 603 + }, + { + "epoch": 0.11, + "learning_rate": 2.9697488168911542e-06, + "loss": 1.5149, + "step": 604 + }, + { + "epoch": 0.11, + "learning_rate": 2.9696942118674917e-06, + "loss": 0.9738, + "step": 605 + }, + { + "epoch": 0.11, + "learning_rate": 2.9696396068438297e-06, + "loss": 1.8546, + "step": 606 + }, + { + "epoch": 0.11, + "learning_rate": 2.9695850018201676e-06, + "loss": 1.3722, + "step": 607 + }, + { + "epoch": 0.11, + "learning_rate": 2.969530396796505e-06, + "loss": 1.6757, + "step": 608 + }, + { + "epoch": 0.11, + "learning_rate": 2.969475791772843e-06, + "loss": 0.5149, + "step": 609 + }, + { + "epoch": 0.11, + "learning_rate": 2.969421186749181e-06, + "loss": 1.5728, + "step": 610 + }, + { + "epoch": 0.11, + "learning_rate": 2.969366581725519e-06, + "loss": 1.0201, + "step": 611 + }, + { + "epoch": 0.11, + "learning_rate": 2.969311976701857e-06, + "loss": 1.5437, + "step": 612 + }, + { + "epoch": 0.11, + "learning_rate": 2.9692573716781944e-06, + "loss": 0.9177, + "step": 613 + }, + { + "epoch": 0.11, + "learning_rate": 2.9692027666545324e-06, + "loss": 0.6756, + "step": 614 + }, + { + "epoch": 0.11, + "learning_rate": 2.9691481616308703e-06, + "loss": 0.6143, + "step": 615 + }, + { + "epoch": 0.11, + "learning_rate": 2.969093556607208e-06, + "loss": 0.5739, + "step": 616 + }, + { + "epoch": 0.11, + "learning_rate": 2.969038951583546e-06, + "loss": 1.4547, + "step": 617 + }, + { + "epoch": 0.11, + "learning_rate": 2.9689843465598837e-06, + "loss": 1.1793, + "step": 618 + }, + { + "epoch": 0.11, + "learning_rate": 2.9689297415362213e-06, + "loss": 0.5268, + "step": 619 + }, + { + "epoch": 0.11, + "learning_rate": 2.968875136512559e-06, + "loss": 1.8643, + "step": 620 + }, + { + "epoch": 0.11, + "learning_rate": 2.968820531488897e-06, + "loss": 0.6181, + "step": 621 + }, + { + "epoch": 0.11, + "learning_rate": 2.9687659264652347e-06, + "loss": 1.1711, + "step": 622 + }, + { + "epoch": 0.11, + "learning_rate": 2.9687113214415726e-06, + "loss": 1.3192, + "step": 623 + }, + { + "epoch": 0.11, + "learning_rate": 2.9686567164179106e-06, + "loss": 0.6562, + "step": 624 + }, + { + "epoch": 0.11, + "learning_rate": 2.9686021113942485e-06, + "loss": 0.9239, + "step": 625 + }, + { + "epoch": 0.11, + "learning_rate": 2.9685475063705864e-06, + "loss": 1.2271, + "step": 626 + }, + { + "epoch": 0.11, + "learning_rate": 2.968492901346924e-06, + "loss": 0.4942, + "step": 627 + }, + { + "epoch": 0.11, + "learning_rate": 2.968438296323262e-06, + "loss": 0.6918, + "step": 628 + }, + { + "epoch": 0.11, + "learning_rate": 2.9683836912996e-06, + "loss": 1.1429, + "step": 629 + }, + { + "epoch": 0.11, + "learning_rate": 2.9683290862759374e-06, + "loss": 0.5276, + "step": 630 + }, + { + "epoch": 0.11, + "learning_rate": 2.9682744812522753e-06, + "loss": 1.1377, + "step": 631 + }, + { + "epoch": 0.11, + "learning_rate": 2.9682198762286133e-06, + "loss": 0.6269, + "step": 632 + }, + { + "epoch": 0.12, + "learning_rate": 2.9681652712049508e-06, + "loss": 0.9446, + "step": 633 + }, + { + "epoch": 0.12, + "learning_rate": 2.9681106661812887e-06, + "loss": 0.9368, + "step": 634 + }, + { + "epoch": 0.12, + "learning_rate": 2.9680560611576267e-06, + "loss": 1.2303, + "step": 635 + }, + { + "epoch": 0.12, + "learning_rate": 2.968001456133964e-06, + "loss": 0.3369, + "step": 636 + }, + { + "epoch": 0.12, + "learning_rate": 2.967946851110302e-06, + "loss": 0.6063, + "step": 637 + }, + { + "epoch": 0.12, + "learning_rate": 2.96789224608664e-06, + "loss": 0.3105, + "step": 638 + }, + { + "epoch": 0.12, + "learning_rate": 2.967837641062978e-06, + "loss": 2.3038, + "step": 639 + }, + { + "epoch": 0.12, + "learning_rate": 2.967783036039316e-06, + "loss": 1.869, + "step": 640 + }, + { + "epoch": 0.12, + "learning_rate": 2.9677284310156535e-06, + "loss": 0.9822, + "step": 641 + }, + { + "epoch": 0.12, + "learning_rate": 2.9676738259919914e-06, + "loss": 0.7886, + "step": 642 + }, + { + "epoch": 0.12, + "learning_rate": 2.9676192209683294e-06, + "loss": 1.8126, + "step": 643 + }, + { + "epoch": 0.12, + "learning_rate": 2.967564615944667e-06, + "loss": 1.7645, + "step": 644 + }, + { + "epoch": 0.12, + "learning_rate": 2.967510010921005e-06, + "loss": 3.0221, + "step": 645 + }, + { + "epoch": 0.12, + "learning_rate": 2.9674554058973428e-06, + "loss": 0.6241, + "step": 646 + }, + { + "epoch": 0.12, + "learning_rate": 2.9674008008736803e-06, + "loss": 1.596, + "step": 647 + }, + { + "epoch": 0.12, + "learning_rate": 2.9673461958500182e-06, + "loss": 1.4545, + "step": 648 + }, + { + "epoch": 0.12, + "learning_rate": 2.967291590826356e-06, + "loss": 0.4474, + "step": 649 + }, + { + "epoch": 0.12, + "learning_rate": 2.9672369858026937e-06, + "loss": 1.8524, + "step": 650 + }, + { + "epoch": 0.12, + "learning_rate": 2.9671823807790316e-06, + "loss": 1.2019, + "step": 651 + }, + { + "epoch": 0.12, + "learning_rate": 2.9671277757553696e-06, + "loss": 0.9641, + "step": 652 + }, + { + "epoch": 0.12, + "learning_rate": 2.9670731707317075e-06, + "loss": 1.7136, + "step": 653 + }, + { + "epoch": 0.12, + "learning_rate": 2.9670185657080455e-06, + "loss": 1.0452, + "step": 654 + }, + { + "epoch": 0.12, + "learning_rate": 2.966963960684383e-06, + "loss": 1.0816, + "step": 655 + }, + { + "epoch": 0.12, + "learning_rate": 2.966909355660721e-06, + "loss": 0.536, + "step": 656 + }, + { + "epoch": 0.12, + "learning_rate": 2.966854750637059e-06, + "loss": 1.7447, + "step": 657 + }, + { + "epoch": 0.12, + "learning_rate": 2.9668001456133964e-06, + "loss": 1.6886, + "step": 658 + }, + { + "epoch": 0.12, + "learning_rate": 2.9667455405897343e-06, + "loss": 0.1567, + "step": 659 + }, + { + "epoch": 0.12, + "learning_rate": 2.9666909355660723e-06, + "loss": 0.2031, + "step": 660 + }, + { + "epoch": 0.12, + "learning_rate": 2.96663633054241e-06, + "loss": 0.1498, + "step": 661 + }, + { + "epoch": 0.12, + "learning_rate": 2.9665817255187477e-06, + "loss": 1.2271, + "step": 662 + }, + { + "epoch": 0.12, + "learning_rate": 2.9665271204950857e-06, + "loss": 0.0344, + "step": 663 + }, + { + "epoch": 0.12, + "learning_rate": 2.966472515471423e-06, + "loss": 0.8937, + "step": 664 + }, + { + "epoch": 0.12, + "learning_rate": 2.966417910447761e-06, + "loss": 0.3609, + "step": 665 + }, + { + "epoch": 0.12, + "learning_rate": 2.966363305424099e-06, + "loss": 1.2617, + "step": 666 + }, + { + "epoch": 0.12, + "learning_rate": 2.966308700400437e-06, + "loss": 0.2714, + "step": 667 + }, + { + "epoch": 0.12, + "learning_rate": 2.966254095376775e-06, + "loss": 0.0595, + "step": 668 + }, + { + "epoch": 0.12, + "learning_rate": 2.9661994903531125e-06, + "loss": 0.3238, + "step": 669 + }, + { + "epoch": 0.12, + "learning_rate": 2.9661448853294504e-06, + "loss": 0.6843, + "step": 670 + }, + { + "epoch": 0.12, + "learning_rate": 2.9660902803057884e-06, + "loss": 2.6054, + "step": 671 + }, + { + "epoch": 0.12, + "learning_rate": 2.966035675282126e-06, + "loss": 0.6123, + "step": 672 + }, + { + "epoch": 0.12, + "learning_rate": 2.965981070258464e-06, + "loss": 1.7799, + "step": 673 + }, + { + "epoch": 0.12, + "learning_rate": 2.965926465234802e-06, + "loss": 1.713, + "step": 674 + }, + { + "epoch": 0.12, + "learning_rate": 2.9658718602111393e-06, + "loss": 1.7712, + "step": 675 + }, + { + "epoch": 0.12, + "learning_rate": 2.9658172551874773e-06, + "loss": 0.7829, + "step": 676 + }, + { + "epoch": 0.12, + "learning_rate": 2.9657626501638148e-06, + "loss": 2.831, + "step": 677 + }, + { + "epoch": 0.12, + "learning_rate": 2.9657080451401527e-06, + "loss": 2.4311, + "step": 678 + }, + { + "epoch": 0.12, + "learning_rate": 2.9656534401164907e-06, + "loss": 1.0772, + "step": 679 + }, + { + "epoch": 0.12, + "learning_rate": 2.9655988350928286e-06, + "loss": 1.9698, + "step": 680 + }, + { + "epoch": 0.12, + "learning_rate": 2.9655442300691665e-06, + "loss": 0.3368, + "step": 681 + }, + { + "epoch": 0.12, + "learning_rate": 2.9654896250455045e-06, + "loss": 2.5859, + "step": 682 + }, + { + "epoch": 0.12, + "learning_rate": 2.965435020021842e-06, + "loss": 0.553, + "step": 683 + }, + { + "epoch": 0.12, + "learning_rate": 2.96538041499818e-06, + "loss": 0.7136, + "step": 684 + }, + { + "epoch": 0.12, + "learning_rate": 2.965325809974518e-06, + "loss": 0.8892, + "step": 685 + }, + { + "epoch": 0.12, + "learning_rate": 2.9652712049508554e-06, + "loss": 1.6595, + "step": 686 + }, + { + "epoch": 0.12, + "learning_rate": 2.9652165999271934e-06, + "loss": 1.437, + "step": 687 + }, + { + "epoch": 0.13, + "learning_rate": 2.9651619949035313e-06, + "loss": 0.7633, + "step": 688 + }, + { + "epoch": 0.13, + "learning_rate": 2.965107389879869e-06, + "loss": 1.4544, + "step": 689 + }, + { + "epoch": 0.13, + "learning_rate": 2.9650527848562068e-06, + "loss": 1.9576, + "step": 690 + }, + { + "epoch": 0.13, + "learning_rate": 2.9649981798325447e-06, + "loss": 1.3275, + "step": 691 + }, + { + "epoch": 0.13, + "learning_rate": 2.9649435748088827e-06, + "loss": 1.7216, + "step": 692 + }, + { + "epoch": 0.13, + "learning_rate": 2.9648889697852206e-06, + "loss": 0.0536, + "step": 693 + }, + { + "epoch": 0.13, + "learning_rate": 2.964834364761558e-06, + "loss": 1.6132, + "step": 694 + }, + { + "epoch": 0.13, + "learning_rate": 2.964779759737896e-06, + "loss": 0.2729, + "step": 695 + }, + { + "epoch": 0.13, + "learning_rate": 2.964725154714234e-06, + "loss": 0.5666, + "step": 696 + }, + { + "epoch": 0.13, + "learning_rate": 2.9646705496905715e-06, + "loss": 0.6017, + "step": 697 + }, + { + "epoch": 0.13, + "learning_rate": 2.9646159446669095e-06, + "loss": 0.4321, + "step": 698 + }, + { + "epoch": 0.13, + "learning_rate": 2.9645613396432474e-06, + "loss": 1.0557, + "step": 699 + }, + { + "epoch": 0.13, + "learning_rate": 2.964506734619585e-06, + "loss": 2.7315, + "step": 700 + }, + { + "epoch": 0.13, + "learning_rate": 2.964452129595923e-06, + "loss": 1.6479, + "step": 701 + }, + { + "epoch": 0.13, + "learning_rate": 2.964397524572261e-06, + "loss": 0.3913, + "step": 702 + }, + { + "epoch": 0.13, + "learning_rate": 2.9643429195485983e-06, + "loss": 0.1017, + "step": 703 + }, + { + "epoch": 0.13, + "learning_rate": 2.9642883145249363e-06, + "loss": 2.4901, + "step": 704 + }, + { + "epoch": 0.13, + "learning_rate": 2.9642337095012742e-06, + "loss": 2.1353, + "step": 705 + }, + { + "epoch": 0.13, + "learning_rate": 2.964179104477612e-06, + "loss": 0.2567, + "step": 706 + }, + { + "epoch": 0.13, + "learning_rate": 2.96412449945395e-06, + "loss": 0.8329, + "step": 707 + }, + { + "epoch": 0.13, + "learning_rate": 2.9640698944302876e-06, + "loss": 0.6303, + "step": 708 + }, + { + "epoch": 0.13, + "learning_rate": 2.9640152894066256e-06, + "loss": 0.9866, + "step": 709 + }, + { + "epoch": 0.13, + "learning_rate": 2.9639606843829635e-06, + "loss": 0.9145, + "step": 710 + }, + { + "epoch": 0.13, + "learning_rate": 2.963906079359301e-06, + "loss": 1.1519, + "step": 711 + }, + { + "epoch": 0.13, + "learning_rate": 2.963851474335639e-06, + "loss": 1.3231, + "step": 712 + }, + { + "epoch": 0.13, + "learning_rate": 2.963796869311977e-06, + "loss": 1.972, + "step": 713 + }, + { + "epoch": 0.13, + "learning_rate": 2.9637422642883144e-06, + "loss": 0.5062, + "step": 714 + }, + { + "epoch": 0.13, + "learning_rate": 2.9636876592646524e-06, + "loss": 0.3928, + "step": 715 + }, + { + "epoch": 0.13, + "learning_rate": 2.9636330542409903e-06, + "loss": 1.881, + "step": 716 + }, + { + "epoch": 0.13, + "learning_rate": 2.963578449217328e-06, + "loss": 0.9357, + "step": 717 + }, + { + "epoch": 0.13, + "learning_rate": 2.963523844193666e-06, + "loss": 2.1717, + "step": 718 + }, + { + "epoch": 0.13, + "learning_rate": 2.9634692391700037e-06, + "loss": 1.9032, + "step": 719 + }, + { + "epoch": 0.13, + "learning_rate": 2.9634146341463417e-06, + "loss": 1.8051, + "step": 720 + }, + { + "epoch": 0.13, + "learning_rate": 2.9633600291226796e-06, + "loss": 0.0657, + "step": 721 + }, + { + "epoch": 0.13, + "learning_rate": 2.963305424099017e-06, + "loss": 0.0839, + "step": 722 + }, + { + "epoch": 0.13, + "learning_rate": 2.963250819075355e-06, + "loss": 0.7676, + "step": 723 + }, + { + "epoch": 0.13, + "learning_rate": 2.963196214051693e-06, + "loss": 0.6881, + "step": 724 + }, + { + "epoch": 0.13, + "learning_rate": 2.9631416090280306e-06, + "loss": 1.7361, + "step": 725 + }, + { + "epoch": 0.13, + "learning_rate": 2.9630870040043685e-06, + "loss": 1.2237, + "step": 726 + }, + { + "epoch": 0.13, + "learning_rate": 2.9630323989807064e-06, + "loss": 1.3523, + "step": 727 + }, + { + "epoch": 0.13, + "learning_rate": 2.962977793957044e-06, + "loss": 1.31, + "step": 728 + }, + { + "epoch": 0.13, + "learning_rate": 2.962923188933382e-06, + "loss": 1.3578, + "step": 729 + }, + { + "epoch": 0.13, + "learning_rate": 2.96286858390972e-06, + "loss": 0.0868, + "step": 730 + }, + { + "epoch": 0.13, + "learning_rate": 2.9628139788860574e-06, + "loss": 0.7539, + "step": 731 + }, + { + "epoch": 0.13, + "learning_rate": 2.9627593738623953e-06, + "loss": 2.2706, + "step": 732 + }, + { + "epoch": 0.13, + "learning_rate": 2.9627047688387333e-06, + "loss": 1.3553, + "step": 733 + }, + { + "epoch": 0.13, + "learning_rate": 2.962650163815071e-06, + "loss": 0.8863, + "step": 734 + }, + { + "epoch": 0.13, + "learning_rate": 2.962595558791409e-06, + "loss": 1.5222, + "step": 735 + }, + { + "epoch": 0.13, + "learning_rate": 2.9625409537677467e-06, + "loss": 1.6614, + "step": 736 + }, + { + "epoch": 0.13, + "learning_rate": 2.9624863487440846e-06, + "loss": 2.1311, + "step": 737 + }, + { + "epoch": 0.13, + "learning_rate": 2.9624317437204225e-06, + "loss": 1.4951, + "step": 738 + }, + { + "epoch": 0.13, + "learning_rate": 2.96237713869676e-06, + "loss": 0.1495, + "step": 739 + }, + { + "epoch": 0.13, + "learning_rate": 2.962322533673098e-06, + "loss": 0.2753, + "step": 740 + }, + { + "epoch": 0.13, + "learning_rate": 2.962267928649436e-06, + "loss": 1.5404, + "step": 741 + }, + { + "epoch": 0.13, + "learning_rate": 2.9622133236257735e-06, + "loss": 0.8157, + "step": 742 + }, + { + "epoch": 0.14, + "learning_rate": 2.9621587186021114e-06, + "loss": 1.4251, + "step": 743 + }, + { + "epoch": 0.14, + "learning_rate": 2.9621041135784494e-06, + "loss": 1.0473, + "step": 744 + }, + { + "epoch": 0.14, + "learning_rate": 2.962049508554787e-06, + "loss": 2.0148, + "step": 745 + }, + { + "epoch": 0.14, + "learning_rate": 2.961994903531125e-06, + "loss": 1.4472, + "step": 746 + }, + { + "epoch": 0.14, + "learning_rate": 2.9619402985074628e-06, + "loss": 1.6008, + "step": 747 + }, + { + "epoch": 0.14, + "learning_rate": 2.9618856934838007e-06, + "loss": 2.6746, + "step": 748 + }, + { + "epoch": 0.14, + "learning_rate": 2.9618310884601387e-06, + "loss": 2.0028, + "step": 749 + }, + { + "epoch": 0.14, + "learning_rate": 2.961776483436476e-06, + "loss": 2.0214, + "step": 750 + }, + { + "epoch": 0.14, + "learning_rate": 2.961721878412814e-06, + "loss": 1.6736, + "step": 751 + }, + { + "epoch": 0.14, + "learning_rate": 2.961667273389152e-06, + "loss": 0.716, + "step": 752 + }, + { + "epoch": 0.14, + "learning_rate": 2.9616126683654896e-06, + "loss": 0.2339, + "step": 753 + }, + { + "epoch": 0.14, + "learning_rate": 2.9615580633418275e-06, + "loss": 2.2946, + "step": 754 + }, + { + "epoch": 0.14, + "learning_rate": 2.9615034583181655e-06, + "loss": 0.3971, + "step": 755 + }, + { + "epoch": 0.14, + "learning_rate": 2.961448853294503e-06, + "loss": 0.3116, + "step": 756 + }, + { + "epoch": 0.14, + "learning_rate": 2.961394248270841e-06, + "loss": 0.5069, + "step": 757 + }, + { + "epoch": 0.14, + "learning_rate": 2.9613396432471784e-06, + "loss": 0.6083, + "step": 758 + }, + { + "epoch": 0.14, + "learning_rate": 2.9612850382235164e-06, + "loss": 1.7593, + "step": 759 + }, + { + "epoch": 0.14, + "learning_rate": 2.9612304331998543e-06, + "loss": 0.2105, + "step": 760 + }, + { + "epoch": 0.14, + "learning_rate": 2.9611758281761923e-06, + "loss": 0.5179, + "step": 761 + }, + { + "epoch": 0.14, + "learning_rate": 2.9611212231525302e-06, + "loss": 1.3439, + "step": 762 + }, + { + "epoch": 0.14, + "learning_rate": 2.961066618128868e-06, + "loss": 2.1452, + "step": 763 + }, + { + "epoch": 0.14, + "learning_rate": 2.9610120131052057e-06, + "loss": 0.4385, + "step": 764 + }, + { + "epoch": 0.14, + "learning_rate": 2.9609574080815436e-06, + "loss": 1.6785, + "step": 765 + }, + { + "epoch": 0.14, + "learning_rate": 2.9609028030578816e-06, + "loss": 0.7915, + "step": 766 + }, + { + "epoch": 0.14, + "learning_rate": 2.960848198034219e-06, + "loss": 0.555, + "step": 767 + }, + { + "epoch": 0.14, + "learning_rate": 2.960793593010557e-06, + "loss": 1.2837, + "step": 768 + }, + { + "epoch": 0.14, + "learning_rate": 2.960738987986895e-06, + "loss": 1.0571, + "step": 769 + }, + { + "epoch": 0.14, + "learning_rate": 2.9606843829632325e-06, + "loss": 0.7631, + "step": 770 + }, + { + "epoch": 0.14, + "learning_rate": 2.9606297779395704e-06, + "loss": 0.276, + "step": 771 + }, + { + "epoch": 0.14, + "learning_rate": 2.9605751729159084e-06, + "loss": 0.7443, + "step": 772 + }, + { + "epoch": 0.14, + "learning_rate": 2.9605205678922463e-06, + "loss": 0.1255, + "step": 773 + }, + { + "epoch": 0.14, + "learning_rate": 2.9604659628685843e-06, + "loss": 1.0438, + "step": 774 + }, + { + "epoch": 0.14, + "learning_rate": 2.960411357844922e-06, + "loss": 1.1311, + "step": 775 + }, + { + "epoch": 0.14, + "learning_rate": 2.9603567528212597e-06, + "loss": 2.1666, + "step": 776 + }, + { + "epoch": 0.14, + "learning_rate": 2.9603021477975977e-06, + "loss": 2.1573, + "step": 777 + }, + { + "epoch": 0.14, + "learning_rate": 2.960247542773935e-06, + "loss": 1.2507, + "step": 778 + }, + { + "epoch": 0.14, + "learning_rate": 2.960192937750273e-06, + "loss": 1.1352, + "step": 779 + }, + { + "epoch": 0.14, + "learning_rate": 2.960138332726611e-06, + "loss": 1.7223, + "step": 780 + }, + { + "epoch": 0.14, + "learning_rate": 2.9600837277029486e-06, + "loss": 0.9326, + "step": 781 + }, + { + "epoch": 0.14, + "learning_rate": 2.9600291226792866e-06, + "loss": 1.8211, + "step": 782 + }, + { + "epoch": 0.14, + "learning_rate": 2.9599745176556245e-06, + "loss": 0.6297, + "step": 783 + }, + { + "epoch": 0.14, + "learning_rate": 2.959919912631962e-06, + "loss": 0.7146, + "step": 784 + }, + { + "epoch": 0.14, + "learning_rate": 2.9598653076083e-06, + "loss": 1.404, + "step": 785 + }, + { + "epoch": 0.14, + "learning_rate": 2.959810702584638e-06, + "loss": 1.9647, + "step": 786 + }, + { + "epoch": 0.14, + "learning_rate": 2.959756097560976e-06, + "loss": 0.8943, + "step": 787 + }, + { + "epoch": 0.14, + "learning_rate": 2.9597014925373138e-06, + "loss": 2.1091, + "step": 788 + }, + { + "epoch": 0.14, + "learning_rate": 2.9596468875136513e-06, + "loss": 0.8071, + "step": 789 + }, + { + "epoch": 0.14, + "learning_rate": 2.9595922824899893e-06, + "loss": 0.5771, + "step": 790 + }, + { + "epoch": 0.14, + "learning_rate": 2.959537677466327e-06, + "loss": 0.1029, + "step": 791 + }, + { + "epoch": 0.14, + "learning_rate": 2.9594830724426647e-06, + "loss": 0.8255, + "step": 792 + }, + { + "epoch": 0.14, + "learning_rate": 2.9594284674190027e-06, + "loss": 0.8856, + "step": 793 + }, + { + "epoch": 0.14, + "learning_rate": 2.9593738623953406e-06, + "loss": 1.5344, + "step": 794 + }, + { + "epoch": 0.14, + "learning_rate": 2.959319257371678e-06, + "loss": 0.5149, + "step": 795 + }, + { + "epoch": 0.14, + "learning_rate": 2.959264652348016e-06, + "loss": 0.9704, + "step": 796 + }, + { + "epoch": 0.14, + "learning_rate": 2.959210047324354e-06, + "loss": 2.5066, + "step": 797 + }, + { + "epoch": 0.15, + "learning_rate": 2.9591554423006915e-06, + "loss": 2.2022, + "step": 798 + }, + { + "epoch": 0.15, + "learning_rate": 2.9591008372770295e-06, + "loss": 1.9915, + "step": 799 + }, + { + "epoch": 0.15, + "learning_rate": 2.9590462322533674e-06, + "loss": 1.2531, + "step": 800 + }, + { + "epoch": 0.15, + "learning_rate": 2.9589916272297054e-06, + "loss": 0.6217, + "step": 801 + }, + { + "epoch": 0.15, + "learning_rate": 2.9589370222060433e-06, + "loss": 2.1522, + "step": 802 + }, + { + "epoch": 0.15, + "learning_rate": 2.958882417182381e-06, + "loss": 2.0401, + "step": 803 + }, + { + "epoch": 0.15, + "learning_rate": 2.9588278121587188e-06, + "loss": 0.4329, + "step": 804 + }, + { + "epoch": 0.15, + "learning_rate": 2.9587732071350567e-06, + "loss": 1.6369, + "step": 805 + }, + { + "epoch": 0.15, + "learning_rate": 2.9587186021113942e-06, + "loss": 1.8644, + "step": 806 + }, + { + "epoch": 0.15, + "learning_rate": 2.958663997087732e-06, + "loss": 2.8692, + "step": 807 + }, + { + "epoch": 0.15, + "learning_rate": 2.95860939206407e-06, + "loss": 0.6426, + "step": 808 + }, + { + "epoch": 0.15, + "learning_rate": 2.9585547870404076e-06, + "loss": 0.0409, + "step": 809 + }, + { + "epoch": 0.15, + "learning_rate": 2.9585001820167456e-06, + "loss": 2.2336, + "step": 810 + }, + { + "epoch": 0.15, + "learning_rate": 2.9584455769930835e-06, + "loss": 1.6213, + "step": 811 + }, + { + "epoch": 0.15, + "learning_rate": 2.958390971969421e-06, + "loss": 1.0497, + "step": 812 + }, + { + "epoch": 0.15, + "learning_rate": 2.958336366945759e-06, + "loss": 0.7726, + "step": 813 + }, + { + "epoch": 0.15, + "learning_rate": 2.958281761922097e-06, + "loss": 0.5773, + "step": 814 + }, + { + "epoch": 0.15, + "learning_rate": 2.958227156898435e-06, + "loss": 1.0544, + "step": 815 + }, + { + "epoch": 0.15, + "learning_rate": 2.958172551874773e-06, + "loss": 1.6246, + "step": 816 + }, + { + "epoch": 0.15, + "learning_rate": 2.9581179468511103e-06, + "loss": 1.9757, + "step": 817 + }, + { + "epoch": 0.15, + "learning_rate": 2.9580633418274483e-06, + "loss": 0.9596, + "step": 818 + }, + { + "epoch": 0.15, + "learning_rate": 2.9580087368037862e-06, + "loss": 1.2205, + "step": 819 + }, + { + "epoch": 0.15, + "learning_rate": 2.9579541317801237e-06, + "loss": 1.5842, + "step": 820 + }, + { + "epoch": 0.15, + "learning_rate": 2.9578995267564617e-06, + "loss": 0.9447, + "step": 821 + }, + { + "epoch": 0.15, + "learning_rate": 2.9578449217327996e-06, + "loss": 0.6575, + "step": 822 + }, + { + "epoch": 0.15, + "learning_rate": 2.957790316709137e-06, + "loss": 2.4566, + "step": 823 + }, + { + "epoch": 0.15, + "learning_rate": 2.957735711685475e-06, + "loss": 0.3791, + "step": 824 + }, + { + "epoch": 0.15, + "learning_rate": 2.957681106661813e-06, + "loss": 0.6183, + "step": 825 + }, + { + "epoch": 0.15, + "learning_rate": 2.9576265016381506e-06, + "loss": 0.6699, + "step": 826 + }, + { + "epoch": 0.15, + "learning_rate": 2.9575718966144885e-06, + "loss": 0.5989, + "step": 827 + }, + { + "epoch": 0.15, + "learning_rate": 2.9575172915908264e-06, + "loss": 0.8496, + "step": 828 + }, + { + "epoch": 0.15, + "learning_rate": 2.9574626865671644e-06, + "loss": 2.7162, + "step": 829 + }, + { + "epoch": 0.15, + "learning_rate": 2.9574080815435023e-06, + "loss": 0.6079, + "step": 830 + }, + { + "epoch": 0.15, + "learning_rate": 2.95735347651984e-06, + "loss": 1.6463, + "step": 831 + }, + { + "epoch": 0.15, + "learning_rate": 2.957298871496178e-06, + "loss": 0.2229, + "step": 832 + }, + { + "epoch": 0.15, + "learning_rate": 2.9572442664725157e-06, + "loss": 0.7543, + "step": 833 + }, + { + "epoch": 0.15, + "learning_rate": 2.9571896614488533e-06, + "loss": 1.6838, + "step": 834 + }, + { + "epoch": 0.15, + "learning_rate": 2.957135056425191e-06, + "loss": 0.7997, + "step": 835 + }, + { + "epoch": 0.15, + "learning_rate": 2.957080451401529e-06, + "loss": 1.1602, + "step": 836 + }, + { + "epoch": 0.15, + "learning_rate": 2.9570258463778667e-06, + "loss": 1.5578, + "step": 837 + }, + { + "epoch": 0.15, + "learning_rate": 2.9569712413542046e-06, + "loss": 0.7802, + "step": 838 + }, + { + "epoch": 0.15, + "learning_rate": 2.956916636330542e-06, + "loss": 1.181, + "step": 839 + }, + { + "epoch": 0.15, + "learning_rate": 2.95686203130688e-06, + "loss": 0.9525, + "step": 840 + }, + { + "epoch": 0.15, + "learning_rate": 2.956807426283218e-06, + "loss": 0.9313, + "step": 841 + }, + { + "epoch": 0.15, + "learning_rate": 2.956752821259556e-06, + "loss": 1.5721, + "step": 842 + }, + { + "epoch": 0.15, + "learning_rate": 2.956698216235894e-06, + "loss": 0.978, + "step": 843 + }, + { + "epoch": 0.15, + "learning_rate": 2.956643611212232e-06, + "loss": 1.0741, + "step": 844 + }, + { + "epoch": 0.15, + "learning_rate": 2.9565890061885694e-06, + "loss": 0.7573, + "step": 845 + }, + { + "epoch": 0.15, + "learning_rate": 2.9565344011649073e-06, + "loss": 0.9267, + "step": 846 + }, + { + "epoch": 0.15, + "learning_rate": 2.9564797961412452e-06, + "loss": 0.2381, + "step": 847 + }, + { + "epoch": 0.15, + "learning_rate": 2.9564251911175828e-06, + "loss": 2.3484, + "step": 848 + }, + { + "epoch": 0.15, + "learning_rate": 2.9563705860939207e-06, + "loss": 3.4164, + "step": 849 + }, + { + "epoch": 0.15, + "learning_rate": 2.9563159810702587e-06, + "loss": 0.657, + "step": 850 + }, + { + "epoch": 0.15, + "learning_rate": 2.956261376046596e-06, + "loss": 0.1214, + "step": 851 + }, + { + "epoch": 0.15, + "learning_rate": 2.956206771022934e-06, + "loss": 2.5639, + "step": 852 + }, + { + "epoch": 0.16, + "learning_rate": 2.956152165999272e-06, + "loss": 0.0854, + "step": 853 + }, + { + "epoch": 0.16, + "learning_rate": 2.9560975609756096e-06, + "loss": 0.5002, + "step": 854 + }, + { + "epoch": 0.16, + "learning_rate": 2.9560429559519475e-06, + "loss": 0.6486, + "step": 855 + }, + { + "epoch": 0.16, + "learning_rate": 2.9559883509282855e-06, + "loss": 0.4433, + "step": 856 + }, + { + "epoch": 0.16, + "learning_rate": 2.9559337459046234e-06, + "loss": 1.5369, + "step": 857 + }, + { + "epoch": 0.16, + "learning_rate": 2.9558791408809614e-06, + "loss": 0.1268, + "step": 858 + }, + { + "epoch": 0.16, + "learning_rate": 2.955824535857299e-06, + "loss": 2.0844, + "step": 859 + }, + { + "epoch": 0.16, + "learning_rate": 2.955769930833637e-06, + "loss": 1.3528, + "step": 860 + }, + { + "epoch": 0.16, + "learning_rate": 2.9557153258099748e-06, + "loss": 1.4226, + "step": 861 + }, + { + "epoch": 0.16, + "learning_rate": 2.9556607207863123e-06, + "loss": 0.6084, + "step": 862 + }, + { + "epoch": 0.16, + "learning_rate": 2.9556061157626502e-06, + "loss": 0.093, + "step": 863 + }, + { + "epoch": 0.16, + "learning_rate": 2.955551510738988e-06, + "loss": 1.5283, + "step": 864 + }, + { + "epoch": 0.16, + "learning_rate": 2.9554969057153257e-06, + "loss": 2.2161, + "step": 865 + }, + { + "epoch": 0.16, + "learning_rate": 2.9554423006916636e-06, + "loss": 0.2556, + "step": 866 + }, + { + "epoch": 0.16, + "learning_rate": 2.9553876956680016e-06, + "loss": 0.6867, + "step": 867 + }, + { + "epoch": 0.16, + "learning_rate": 2.9553330906443395e-06, + "loss": 1.273, + "step": 868 + }, + { + "epoch": 0.16, + "learning_rate": 2.9552784856206775e-06, + "loss": 2.2984, + "step": 869 + }, + { + "epoch": 0.16, + "learning_rate": 2.955223880597015e-06, + "loss": 0.4078, + "step": 870 + }, + { + "epoch": 0.16, + "learning_rate": 2.955169275573353e-06, + "loss": 1.6587, + "step": 871 + }, + { + "epoch": 0.16, + "learning_rate": 2.955114670549691e-06, + "loss": 0.5698, + "step": 872 + }, + { + "epoch": 0.16, + "learning_rate": 2.9550600655260284e-06, + "loss": 0.2482, + "step": 873 + }, + { + "epoch": 0.16, + "learning_rate": 2.9550054605023663e-06, + "loss": 1.13, + "step": 874 + }, + { + "epoch": 0.16, + "learning_rate": 2.9549508554787043e-06, + "loss": 1.9434, + "step": 875 + }, + { + "epoch": 0.16, + "learning_rate": 2.954896250455042e-06, + "loss": 0.6925, + "step": 876 + }, + { + "epoch": 0.16, + "learning_rate": 2.9548416454313797e-06, + "loss": 0.5341, + "step": 877 + }, + { + "epoch": 0.16, + "learning_rate": 2.9547870404077177e-06, + "loss": 0.7753, + "step": 878 + }, + { + "epoch": 0.16, + "learning_rate": 2.954732435384055e-06, + "loss": 1.4176, + "step": 879 + }, + { + "epoch": 0.16, + "learning_rate": 2.954677830360393e-06, + "loss": 1.2696, + "step": 880 + }, + { + "epoch": 0.16, + "learning_rate": 2.954623225336731e-06, + "loss": 0.7854, + "step": 881 + }, + { + "epoch": 0.16, + "learning_rate": 2.954568620313069e-06, + "loss": 0.0418, + "step": 882 + }, + { + "epoch": 0.16, + "learning_rate": 2.954514015289407e-06, + "loss": 0.1493, + "step": 883 + }, + { + "epoch": 0.16, + "learning_rate": 2.9544594102657445e-06, + "loss": 2.8342, + "step": 884 + }, + { + "epoch": 0.16, + "learning_rate": 2.9544048052420824e-06, + "loss": 1.7822, + "step": 885 + }, + { + "epoch": 0.16, + "learning_rate": 2.9543502002184204e-06, + "loss": 0.967, + "step": 886 + }, + { + "epoch": 0.16, + "learning_rate": 2.954295595194758e-06, + "loss": 0.7213, + "step": 887 + }, + { + "epoch": 0.16, + "learning_rate": 2.954240990171096e-06, + "loss": 1.0803, + "step": 888 + }, + { + "epoch": 0.16, + "learning_rate": 2.9541863851474338e-06, + "loss": 1.1196, + "step": 889 + }, + { + "epoch": 0.16, + "learning_rate": 2.9541317801237713e-06, + "loss": 1.2627, + "step": 890 + }, + { + "epoch": 0.16, + "learning_rate": 2.9540771751001093e-06, + "loss": 0.1076, + "step": 891 + }, + { + "epoch": 0.16, + "learning_rate": 2.954022570076447e-06, + "loss": 0.4868, + "step": 892 + }, + { + "epoch": 0.16, + "learning_rate": 2.9539679650527847e-06, + "loss": 3.0788, + "step": 893 + }, + { + "epoch": 0.16, + "learning_rate": 2.9539133600291227e-06, + "loss": 2.2321, + "step": 894 + }, + { + "epoch": 0.16, + "learning_rate": 2.9538587550054606e-06, + "loss": 1.0111, + "step": 895 + }, + { + "epoch": 0.16, + "learning_rate": 2.9538041499817985e-06, + "loss": 0.1631, + "step": 896 + }, + { + "epoch": 0.16, + "learning_rate": 2.9537495449581365e-06, + "loss": 1.3026, + "step": 897 + }, + { + "epoch": 0.16, + "learning_rate": 2.953694939934474e-06, + "loss": 1.885, + "step": 898 + }, + { + "epoch": 0.16, + "learning_rate": 2.953640334910812e-06, + "loss": 1.0145, + "step": 899 + }, + { + "epoch": 0.16, + "learning_rate": 2.95358572988715e-06, + "loss": 3.3421, + "step": 900 + }, + { + "epoch": 0.16, + "learning_rate": 2.9535311248634874e-06, + "loss": 0.0855, + "step": 901 + }, + { + "epoch": 0.16, + "learning_rate": 2.9534765198398254e-06, + "loss": 0.728, + "step": 902 + }, + { + "epoch": 0.16, + "learning_rate": 2.9534219148161633e-06, + "loss": 0.2103, + "step": 903 + }, + { + "epoch": 0.16, + "learning_rate": 2.953367309792501e-06, + "loss": 1.1232, + "step": 904 + }, + { + "epoch": 0.16, + "learning_rate": 2.9533127047688388e-06, + "loss": 1.1956, + "step": 905 + }, + { + "epoch": 0.16, + "learning_rate": 2.9532580997451763e-06, + "loss": 2.4672, + "step": 906 + }, + { + "epoch": 0.16, + "learning_rate": 2.9532034947215142e-06, + "loss": 0.0688, + "step": 907 + }, + { + "epoch": 0.17, + "learning_rate": 2.953148889697852e-06, + "loss": 0.5254, + "step": 908 + }, + { + "epoch": 0.17, + "learning_rate": 2.95309428467419e-06, + "loss": 1.2627, + "step": 909 + }, + { + "epoch": 0.17, + "learning_rate": 2.953039679650528e-06, + "loss": 1.6421, + "step": 910 + }, + { + "epoch": 0.17, + "learning_rate": 2.952985074626866e-06, + "loss": 0.9274, + "step": 911 + }, + { + "epoch": 0.17, + "learning_rate": 2.9529304696032035e-06, + "loss": 3.6787, + "step": 912 + }, + { + "epoch": 0.17, + "learning_rate": 2.9528758645795415e-06, + "loss": 2.9265, + "step": 913 + }, + { + "epoch": 0.17, + "learning_rate": 2.9528212595558794e-06, + "loss": 1.5383, + "step": 914 + }, + { + "epoch": 0.17, + "learning_rate": 2.952766654532217e-06, + "loss": 1.1374, + "step": 915 + }, + { + "epoch": 0.17, + "learning_rate": 2.952712049508555e-06, + "loss": 1.7596, + "step": 916 + }, + { + "epoch": 0.17, + "learning_rate": 2.952657444484893e-06, + "loss": 0.7845, + "step": 917 + }, + { + "epoch": 0.17, + "learning_rate": 2.9526028394612303e-06, + "loss": 1.3766, + "step": 918 + }, + { + "epoch": 0.17, + "learning_rate": 2.9525482344375683e-06, + "loss": 1.4038, + "step": 919 + }, + { + "epoch": 0.17, + "learning_rate": 2.952493629413906e-06, + "loss": 3.4192, + "step": 920 + }, + { + "epoch": 0.17, + "learning_rate": 2.9524390243902437e-06, + "loss": 0.9787, + "step": 921 + }, + { + "epoch": 0.17, + "learning_rate": 2.9523844193665817e-06, + "loss": 1.283, + "step": 922 + }, + { + "epoch": 0.17, + "learning_rate": 2.9523298143429196e-06, + "loss": 1.2396, + "step": 923 + }, + { + "epoch": 0.17, + "learning_rate": 2.9522752093192576e-06, + "loss": 2.4784, + "step": 924 + }, + { + "epoch": 0.17, + "learning_rate": 2.9522206042955955e-06, + "loss": 1.8479, + "step": 925 + }, + { + "epoch": 0.17, + "learning_rate": 2.952165999271933e-06, + "loss": 0.6585, + "step": 926 + }, + { + "epoch": 0.17, + "learning_rate": 2.952111394248271e-06, + "loss": 0.6022, + "step": 927 + }, + { + "epoch": 0.17, + "learning_rate": 2.952056789224609e-06, + "loss": 1.2554, + "step": 928 + }, + { + "epoch": 0.17, + "learning_rate": 2.9520021842009464e-06, + "loss": 1.2969, + "step": 929 + }, + { + "epoch": 0.17, + "learning_rate": 2.9519475791772844e-06, + "loss": 1.1342, + "step": 930 + }, + { + "epoch": 0.17, + "learning_rate": 2.9518929741536223e-06, + "loss": 1.701, + "step": 931 + }, + { + "epoch": 0.17, + "learning_rate": 2.95183836912996e-06, + "loss": 0.8072, + "step": 932 + }, + { + "epoch": 0.17, + "learning_rate": 2.951783764106298e-06, + "loss": 1.8466, + "step": 933 + }, + { + "epoch": 0.17, + "learning_rate": 2.9517291590826353e-06, + "loss": 0.9951, + "step": 934 + }, + { + "epoch": 0.17, + "learning_rate": 2.9516745540589733e-06, + "loss": 1.1498, + "step": 935 + }, + { + "epoch": 0.17, + "learning_rate": 2.951619949035311e-06, + "loss": 1.1729, + "step": 936 + }, + { + "epoch": 0.17, + "learning_rate": 2.951565344011649e-06, + "loss": 4.3715, + "step": 937 + }, + { + "epoch": 0.17, + "learning_rate": 2.951510738987987e-06, + "loss": 1.9196, + "step": 938 + }, + { + "epoch": 0.17, + "learning_rate": 2.951456133964325e-06, + "loss": 1.6298, + "step": 939 + }, + { + "epoch": 0.17, + "learning_rate": 2.9514015289406625e-06, + "loss": 2.8931, + "step": 940 + }, + { + "epoch": 0.17, + "learning_rate": 2.9513469239170005e-06, + "loss": 1.031, + "step": 941 + }, + { + "epoch": 0.17, + "learning_rate": 2.9512923188933384e-06, + "loss": 2.834, + "step": 942 + }, + { + "epoch": 0.17, + "learning_rate": 2.951237713869676e-06, + "loss": 1.3073, + "step": 943 + }, + { + "epoch": 0.17, + "learning_rate": 2.951183108846014e-06, + "loss": 1.1025, + "step": 944 + }, + { + "epoch": 0.17, + "learning_rate": 2.951128503822352e-06, + "loss": 1.1245, + "step": 945 + }, + { + "epoch": 0.17, + "learning_rate": 2.9510738987986894e-06, + "loss": 0.2525, + "step": 946 + }, + { + "epoch": 0.17, + "learning_rate": 2.9510192937750273e-06, + "loss": 3.4846, + "step": 947 + }, + { + "epoch": 0.17, + "learning_rate": 2.9509646887513652e-06, + "loss": 1.1124, + "step": 948 + }, + { + "epoch": 0.17, + "learning_rate": 2.950910083727703e-06, + "loss": 0.9983, + "step": 949 + }, + { + "epoch": 0.17, + "learning_rate": 2.950855478704041e-06, + "loss": 0.2779, + "step": 950 + }, + { + "epoch": 0.17, + "learning_rate": 2.9508008736803787e-06, + "loss": 1.5949, + "step": 951 + }, + { + "epoch": 0.17, + "learning_rate": 2.9507462686567166e-06, + "loss": 0.2156, + "step": 952 + }, + { + "epoch": 0.17, + "learning_rate": 2.9506916636330545e-06, + "loss": 1.8812, + "step": 953 + }, + { + "epoch": 0.17, + "learning_rate": 2.950637058609392e-06, + "loss": 1.0901, + "step": 954 + }, + { + "epoch": 0.17, + "learning_rate": 2.95058245358573e-06, + "loss": 0.7758, + "step": 955 + }, + { + "epoch": 0.17, + "learning_rate": 2.950527848562068e-06, + "loss": 1.3003, + "step": 956 + }, + { + "epoch": 0.17, + "learning_rate": 2.9504732435384055e-06, + "loss": 2.3154, + "step": 957 + }, + { + "epoch": 0.17, + "learning_rate": 2.9504186385147434e-06, + "loss": 1.6577, + "step": 958 + }, + { + "epoch": 0.17, + "learning_rate": 2.9503640334910814e-06, + "loss": 1.2545, + "step": 959 + }, + { + "epoch": 0.17, + "learning_rate": 2.950309428467419e-06, + "loss": 0.9147, + "step": 960 + }, + { + "epoch": 0.17, + "learning_rate": 2.950254823443757e-06, + "loss": 0.8486, + "step": 961 + }, + { + "epoch": 0.17, + "learning_rate": 2.9502002184200948e-06, + "loss": 1.478, + "step": 962 + }, + { + "epoch": 0.18, + "learning_rate": 2.9501456133964327e-06, + "loss": 0.6057, + "step": 963 + }, + { + "epoch": 0.18, + "learning_rate": 2.9500910083727706e-06, + "loss": 1.595, + "step": 964 + }, + { + "epoch": 0.18, + "learning_rate": 2.950036403349108e-06, + "loss": 1.0408, + "step": 965 + }, + { + "epoch": 0.18, + "learning_rate": 2.949981798325446e-06, + "loss": 1.5185, + "step": 966 + }, + { + "epoch": 0.18, + "learning_rate": 2.949927193301784e-06, + "loss": 2.0519, + "step": 967 + }, + { + "epoch": 0.18, + "learning_rate": 2.9498725882781216e-06, + "loss": 0.5436, + "step": 968 + }, + { + "epoch": 0.18, + "learning_rate": 2.9498179832544595e-06, + "loss": 0.6263, + "step": 969 + }, + { + "epoch": 0.18, + "learning_rate": 2.9497633782307975e-06, + "loss": 0.5505, + "step": 970 + }, + { + "epoch": 0.18, + "learning_rate": 2.949708773207135e-06, + "loss": 1.6213, + "step": 971 + }, + { + "epoch": 0.18, + "learning_rate": 2.949654168183473e-06, + "loss": 0.3889, + "step": 972 + }, + { + "epoch": 0.18, + "learning_rate": 2.949599563159811e-06, + "loss": 1.3778, + "step": 973 + }, + { + "epoch": 0.18, + "learning_rate": 2.9495449581361484e-06, + "loss": 1.5732, + "step": 974 + }, + { + "epoch": 0.18, + "learning_rate": 2.9494903531124863e-06, + "loss": 0.1381, + "step": 975 + }, + { + "epoch": 0.18, + "learning_rate": 2.9494357480888243e-06, + "loss": 1.2364, + "step": 976 + }, + { + "epoch": 0.18, + "learning_rate": 2.9493811430651622e-06, + "loss": 0.7221, + "step": 977 + }, + { + "epoch": 0.18, + "learning_rate": 2.9493265380415e-06, + "loss": 0.6803, + "step": 978 + }, + { + "epoch": 0.18, + "learning_rate": 2.9492719330178377e-06, + "loss": 1.3, + "step": 979 + }, + { + "epoch": 0.18, + "learning_rate": 2.9492173279941756e-06, + "loss": 1.3177, + "step": 980 + }, + { + "epoch": 0.18, + "learning_rate": 2.9491627229705136e-06, + "loss": 0.9752, + "step": 981 + }, + { + "epoch": 0.18, + "learning_rate": 2.949108117946851e-06, + "loss": 0.95, + "step": 982 + }, + { + "epoch": 0.18, + "learning_rate": 2.949053512923189e-06, + "loss": 2.2573, + "step": 983 + }, + { + "epoch": 0.18, + "learning_rate": 2.948998907899527e-06, + "loss": 1.2123, + "step": 984 + }, + { + "epoch": 0.18, + "learning_rate": 2.9489443028758645e-06, + "loss": 1.123, + "step": 985 + }, + { + "epoch": 0.18, + "learning_rate": 2.9488896978522024e-06, + "loss": 0.689, + "step": 986 + }, + { + "epoch": 0.18, + "learning_rate": 2.94883509282854e-06, + "loss": 0.0385, + "step": 987 + }, + { + "epoch": 0.18, + "learning_rate": 2.948780487804878e-06, + "loss": 2.489, + "step": 988 + }, + { + "epoch": 0.18, + "learning_rate": 2.948725882781216e-06, + "loss": 1.6283, + "step": 989 + }, + { + "epoch": 0.18, + "learning_rate": 2.948671277757554e-06, + "loss": 1.2148, + "step": 990 + }, + { + "epoch": 0.18, + "learning_rate": 2.9486166727338917e-06, + "loss": 0.3783, + "step": 991 + }, + { + "epoch": 0.18, + "learning_rate": 2.9485620677102297e-06, + "loss": 1.1762, + "step": 992 + }, + { + "epoch": 0.18, + "learning_rate": 2.948507462686567e-06, + "loss": 0.7421, + "step": 993 + }, + { + "epoch": 0.18, + "learning_rate": 2.948452857662905e-06, + "loss": 1.3281, + "step": 994 + }, + { + "epoch": 0.18, + "learning_rate": 2.948398252639243e-06, + "loss": 1.2404, + "step": 995 + }, + { + "epoch": 0.18, + "learning_rate": 2.9483436476155806e-06, + "loss": 1.0298, + "step": 996 + }, + { + "epoch": 0.18, + "learning_rate": 2.9482890425919185e-06, + "loss": 0.5909, + "step": 997 + }, + { + "epoch": 0.18, + "learning_rate": 2.9482344375682565e-06, + "loss": 1.1213, + "step": 998 + }, + { + "epoch": 0.18, + "learning_rate": 2.948179832544594e-06, + "loss": 0.4847, + "step": 999 + }, + { + "epoch": 0.18, + "learning_rate": 2.948125227520932e-06, + "loss": 0.9046, + "step": 1000 + }, + { + "epoch": 0.18, + "learning_rate": 2.9480706224972695e-06, + "loss": 2.0238, + "step": 1001 + }, + { + "epoch": 0.18, + "learning_rate": 2.9480160174736074e-06, + "loss": 0.653, + "step": 1002 + }, + { + "epoch": 0.18, + "learning_rate": 2.9479614124499454e-06, + "loss": 0.7815, + "step": 1003 + }, + { + "epoch": 0.18, + "learning_rate": 2.9479068074262833e-06, + "loss": 1.3728, + "step": 1004 + }, + { + "epoch": 0.18, + "learning_rate": 2.9478522024026212e-06, + "loss": 1.5473, + "step": 1005 + }, + { + "epoch": 0.18, + "learning_rate": 2.947797597378959e-06, + "loss": 0.2146, + "step": 1006 + }, + { + "epoch": 0.18, + "learning_rate": 2.9477429923552967e-06, + "loss": 1.171, + "step": 1007 + }, + { + "epoch": 0.18, + "learning_rate": 2.9476883873316347e-06, + "loss": 2.2267, + "step": 1008 + }, + { + "epoch": 0.18, + "learning_rate": 2.9476337823079726e-06, + "loss": 0.7526, + "step": 1009 + }, + { + "epoch": 0.18, + "learning_rate": 2.94757917728431e-06, + "loss": 1.6309, + "step": 1010 + }, + { + "epoch": 0.18, + "learning_rate": 2.947524572260648e-06, + "loss": 1.3997, + "step": 1011 + }, + { + "epoch": 0.18, + "learning_rate": 2.947469967236986e-06, + "loss": 0.0848, + "step": 1012 + }, + { + "epoch": 0.18, + "learning_rate": 2.9474153622133235e-06, + "loss": 0.4292, + "step": 1013 + }, + { + "epoch": 0.18, + "learning_rate": 2.9473607571896615e-06, + "loss": 1.3929, + "step": 1014 + }, + { + "epoch": 0.18, + "learning_rate": 2.947306152165999e-06, + "loss": 1.1234, + "step": 1015 + }, + { + "epoch": 0.18, + "learning_rate": 2.947251547142337e-06, + "loss": 0.4357, + "step": 1016 + }, + { + "epoch": 0.18, + "learning_rate": 2.947196942118675e-06, + "loss": 0.4709, + "step": 1017 + }, + { + "epoch": 0.19, + "learning_rate": 2.947142337095013e-06, + "loss": 4.0583, + "step": 1018 + }, + { + "epoch": 0.19, + "learning_rate": 2.9470877320713508e-06, + "loss": 0.2479, + "step": 1019 + }, + { + "epoch": 0.19, + "learning_rate": 2.9470331270476887e-06, + "loss": 1.1993, + "step": 1020 + }, + { + "epoch": 0.19, + "learning_rate": 2.9469785220240262e-06, + "loss": 0.6557, + "step": 1021 + }, + { + "epoch": 0.19, + "learning_rate": 2.946923917000364e-06, + "loss": 0.905, + "step": 1022 + }, + { + "epoch": 0.19, + "learning_rate": 2.946869311976702e-06, + "loss": 0.1644, + "step": 1023 + }, + { + "epoch": 0.19, + "learning_rate": 2.9468147069530396e-06, + "loss": 0.85, + "step": 1024 + }, + { + "epoch": 0.19, + "learning_rate": 2.9467601019293776e-06, + "loss": 0.8201, + "step": 1025 + }, + { + "epoch": 0.19, + "learning_rate": 2.9467054969057155e-06, + "loss": 1.0207, + "step": 1026 + }, + { + "epoch": 0.19, + "learning_rate": 2.946650891882053e-06, + "loss": 0.537, + "step": 1027 + }, + { + "epoch": 0.19, + "learning_rate": 2.946596286858391e-06, + "loss": 1.1748, + "step": 1028 + }, + { + "epoch": 0.19, + "learning_rate": 2.946541681834729e-06, + "loss": 0.3862, + "step": 1029 + }, + { + "epoch": 0.19, + "learning_rate": 2.946487076811067e-06, + "loss": 0.7893, + "step": 1030 + }, + { + "epoch": 0.19, + "learning_rate": 2.946432471787405e-06, + "loss": 1.4236, + "step": 1031 + }, + { + "epoch": 0.19, + "learning_rate": 2.9463778667637423e-06, + "loss": 0.8245, + "step": 1032 + }, + { + "epoch": 0.19, + "learning_rate": 2.9463232617400803e-06, + "loss": 0.3872, + "step": 1033 + }, + { + "epoch": 0.19, + "learning_rate": 2.9462686567164182e-06, + "loss": 0.6639, + "step": 1034 + }, + { + "epoch": 0.19, + "learning_rate": 2.9462140516927557e-06, + "loss": 0.3898, + "step": 1035 + }, + { + "epoch": 0.19, + "learning_rate": 2.9461594466690937e-06, + "loss": 0.0611, + "step": 1036 + }, + { + "epoch": 0.19, + "learning_rate": 2.9461048416454316e-06, + "loss": 3.0981, + "step": 1037 + }, + { + "epoch": 0.19, + "learning_rate": 2.946050236621769e-06, + "loss": 1.3213, + "step": 1038 + }, + { + "epoch": 0.19, + "learning_rate": 2.945995631598107e-06, + "loss": 2.822, + "step": 1039 + }, + { + "epoch": 0.19, + "learning_rate": 2.945941026574445e-06, + "loss": 0.374, + "step": 1040 + }, + { + "epoch": 0.19, + "learning_rate": 2.9458864215507825e-06, + "loss": 0.9485, + "step": 1041 + }, + { + "epoch": 0.19, + "learning_rate": 2.9458318165271205e-06, + "loss": 1.2164, + "step": 1042 + }, + { + "epoch": 0.19, + "learning_rate": 2.9457772115034584e-06, + "loss": 0.3997, + "step": 1043 + }, + { + "epoch": 0.19, + "learning_rate": 2.9457226064797964e-06, + "loss": 0.6999, + "step": 1044 + }, + { + "epoch": 0.19, + "learning_rate": 2.9456680014561343e-06, + "loss": 0.8071, + "step": 1045 + }, + { + "epoch": 0.19, + "learning_rate": 2.945613396432472e-06, + "loss": 0.4854, + "step": 1046 + }, + { + "epoch": 0.19, + "learning_rate": 2.9455587914088098e-06, + "loss": 0.5732, + "step": 1047 + }, + { + "epoch": 0.19, + "learning_rate": 2.9455041863851477e-06, + "loss": 1.3509, + "step": 1048 + }, + { + "epoch": 0.19, + "learning_rate": 2.9454495813614852e-06, + "loss": 2.9287, + "step": 1049 + }, + { + "epoch": 0.19, + "learning_rate": 2.945394976337823e-06, + "loss": 0.049, + "step": 1050 + }, + { + "epoch": 0.19, + "learning_rate": 2.945340371314161e-06, + "loss": 1.6306, + "step": 1051 + }, + { + "epoch": 0.19, + "learning_rate": 2.9452857662904987e-06, + "loss": 0.9738, + "step": 1052 + }, + { + "epoch": 0.19, + "learning_rate": 2.9452311612668366e-06, + "loss": 2.3241, + "step": 1053 + }, + { + "epoch": 0.19, + "learning_rate": 2.9451765562431745e-06, + "loss": 1.4361, + "step": 1054 + }, + { + "epoch": 0.19, + "learning_rate": 2.945121951219512e-06, + "loss": 2.3852, + "step": 1055 + }, + { + "epoch": 0.19, + "learning_rate": 2.94506734619585e-06, + "loss": 3.3932, + "step": 1056 + }, + { + "epoch": 0.19, + "learning_rate": 2.945012741172188e-06, + "loss": 0.5389, + "step": 1057 + }, + { + "epoch": 0.19, + "learning_rate": 2.944958136148526e-06, + "loss": 0.6967, + "step": 1058 + }, + { + "epoch": 0.19, + "learning_rate": 2.944903531124864e-06, + "loss": 1.0048, + "step": 1059 + }, + { + "epoch": 0.19, + "learning_rate": 2.9448489261012014e-06, + "loss": 0.4138, + "step": 1060 + }, + { + "epoch": 0.19, + "learning_rate": 2.9447943210775393e-06, + "loss": 0.8955, + "step": 1061 + }, + { + "epoch": 0.19, + "learning_rate": 2.9447397160538772e-06, + "loss": 2.4851, + "step": 1062 + }, + { + "epoch": 0.19, + "learning_rate": 2.9446851110302148e-06, + "loss": 1.148, + "step": 1063 + }, + { + "epoch": 0.19, + "learning_rate": 2.9446305060065527e-06, + "loss": 1.2997, + "step": 1064 + }, + { + "epoch": 0.19, + "learning_rate": 2.9445759009828907e-06, + "loss": 0.7295, + "step": 1065 + }, + { + "epoch": 0.19, + "learning_rate": 2.944521295959228e-06, + "loss": 2.5747, + "step": 1066 + }, + { + "epoch": 0.19, + "learning_rate": 2.944466690935566e-06, + "loss": 1.0418, + "step": 1067 + }, + { + "epoch": 0.19, + "learning_rate": 2.9444120859119036e-06, + "loss": 1.9512, + "step": 1068 + }, + { + "epoch": 0.19, + "learning_rate": 2.9443574808882416e-06, + "loss": 0.5865, + "step": 1069 + }, + { + "epoch": 0.19, + "learning_rate": 2.9443028758645795e-06, + "loss": 1.6722, + "step": 1070 + }, + { + "epoch": 0.19, + "learning_rate": 2.9442482708409175e-06, + "loss": 0.8361, + "step": 1071 + }, + { + "epoch": 0.19, + "learning_rate": 2.9441936658172554e-06, + "loss": 1.9455, + "step": 1072 + }, + { + "epoch": 0.2, + "learning_rate": 2.9441390607935934e-06, + "loss": 1.2842, + "step": 1073 + }, + { + "epoch": 0.2, + "learning_rate": 2.944084455769931e-06, + "loss": 0.7804, + "step": 1074 + }, + { + "epoch": 0.2, + "learning_rate": 2.944029850746269e-06, + "loss": 0.3684, + "step": 1075 + }, + { + "epoch": 0.2, + "learning_rate": 2.9439752457226068e-06, + "loss": 1.2273, + "step": 1076 + }, + { + "epoch": 0.2, + "learning_rate": 2.9439206406989443e-06, + "loss": 1.4146, + "step": 1077 + }, + { + "epoch": 0.2, + "learning_rate": 2.9438660356752822e-06, + "loss": 4.7049, + "step": 1078 + }, + { + "epoch": 0.2, + "learning_rate": 2.94381143065162e-06, + "loss": 1.5416, + "step": 1079 + }, + { + "epoch": 0.2, + "learning_rate": 2.9437568256279577e-06, + "loss": 1.064, + "step": 1080 + }, + { + "epoch": 0.2, + "learning_rate": 2.9437022206042956e-06, + "loss": 1.0132, + "step": 1081 + }, + { + "epoch": 0.2, + "learning_rate": 2.943647615580633e-06, + "loss": 1.7281, + "step": 1082 + }, + { + "epoch": 0.2, + "learning_rate": 2.943593010556971e-06, + "loss": 0.9633, + "step": 1083 + }, + { + "epoch": 0.2, + "learning_rate": 2.943538405533309e-06, + "loss": 1.2133, + "step": 1084 + }, + { + "epoch": 0.2, + "learning_rate": 2.943483800509647e-06, + "loss": 1.0203, + "step": 1085 + }, + { + "epoch": 0.2, + "learning_rate": 2.943429195485985e-06, + "loss": 0.8201, + "step": 1086 + }, + { + "epoch": 0.2, + "learning_rate": 2.943374590462323e-06, + "loss": 1.1463, + "step": 1087 + }, + { + "epoch": 0.2, + "learning_rate": 2.9433199854386604e-06, + "loss": 0.2924, + "step": 1088 + }, + { + "epoch": 0.2, + "learning_rate": 2.9432653804149983e-06, + "loss": 0.7078, + "step": 1089 + }, + { + "epoch": 0.2, + "learning_rate": 2.9432107753913363e-06, + "loss": 0.7465, + "step": 1090 + }, + { + "epoch": 0.2, + "learning_rate": 2.943156170367674e-06, + "loss": 0.6773, + "step": 1091 + }, + { + "epoch": 0.2, + "learning_rate": 2.9431015653440117e-06, + "loss": 2.2552, + "step": 1092 + }, + { + "epoch": 0.2, + "learning_rate": 2.9430469603203497e-06, + "loss": 1.3008, + "step": 1093 + }, + { + "epoch": 0.2, + "learning_rate": 2.942992355296687e-06, + "loss": 0.7863, + "step": 1094 + }, + { + "epoch": 0.2, + "learning_rate": 2.942937750273025e-06, + "loss": 1.0988, + "step": 1095 + }, + { + "epoch": 0.2, + "learning_rate": 2.9428831452493627e-06, + "loss": 1.5269, + "step": 1096 + }, + { + "epoch": 0.2, + "learning_rate": 2.9428285402257006e-06, + "loss": 1.5221, + "step": 1097 + }, + { + "epoch": 0.2, + "learning_rate": 2.9427739352020385e-06, + "loss": 1.432, + "step": 1098 + }, + { + "epoch": 0.2, + "learning_rate": 2.9427193301783765e-06, + "loss": 0.8745, + "step": 1099 + }, + { + "epoch": 0.2, + "learning_rate": 2.9426647251547144e-06, + "loss": 0.4561, + "step": 1100 + }, + { + "epoch": 0.2, + "learning_rate": 2.9426101201310524e-06, + "loss": 0.1717, + "step": 1101 + }, + { + "epoch": 0.2, + "learning_rate": 2.94255551510739e-06, + "loss": 0.8259, + "step": 1102 + }, + { + "epoch": 0.2, + "learning_rate": 2.942500910083728e-06, + "loss": 0.0426, + "step": 1103 + }, + { + "epoch": 0.2, + "learning_rate": 2.9424463050600658e-06, + "loss": 0.7838, + "step": 1104 + }, + { + "epoch": 0.2, + "learning_rate": 2.9423917000364033e-06, + "loss": 0.4044, + "step": 1105 + }, + { + "epoch": 0.2, + "learning_rate": 2.9423370950127412e-06, + "loss": 0.606, + "step": 1106 + }, + { + "epoch": 0.2, + "learning_rate": 2.942282489989079e-06, + "loss": 1.0402, + "step": 1107 + }, + { + "epoch": 0.2, + "learning_rate": 2.9422278849654167e-06, + "loss": 0.3295, + "step": 1108 + }, + { + "epoch": 0.2, + "learning_rate": 2.9421732799417547e-06, + "loss": 1.2387, + "step": 1109 + }, + { + "epoch": 0.2, + "learning_rate": 2.9421186749180926e-06, + "loss": 0.6123, + "step": 1110 + }, + { + "epoch": 0.2, + "learning_rate": 2.9420640698944305e-06, + "loss": 1.726, + "step": 1111 + }, + { + "epoch": 0.2, + "learning_rate": 2.9420094648707685e-06, + "loss": 0.8106, + "step": 1112 + }, + { + "epoch": 0.2, + "learning_rate": 2.941954859847106e-06, + "loss": 0.6276, + "step": 1113 + }, + { + "epoch": 0.2, + "learning_rate": 2.941900254823444e-06, + "loss": 3.0847, + "step": 1114 + }, + { + "epoch": 0.2, + "learning_rate": 2.941845649799782e-06, + "loss": 0.0529, + "step": 1115 + }, + { + "epoch": 0.2, + "learning_rate": 2.9417910447761194e-06, + "loss": 0.6651, + "step": 1116 + }, + { + "epoch": 0.2, + "learning_rate": 2.9417364397524574e-06, + "loss": 0.5615, + "step": 1117 + }, + { + "epoch": 0.2, + "learning_rate": 2.9416818347287953e-06, + "loss": 2.8283, + "step": 1118 + }, + { + "epoch": 0.2, + "learning_rate": 2.941627229705133e-06, + "loss": 0.5657, + "step": 1119 + }, + { + "epoch": 0.2, + "learning_rate": 2.9415726246814708e-06, + "loss": 0.549, + "step": 1120 + }, + { + "epoch": 0.2, + "learning_rate": 2.9415180196578087e-06, + "loss": 1.6639, + "step": 1121 + }, + { + "epoch": 0.2, + "learning_rate": 2.9414634146341462e-06, + "loss": 0.5508, + "step": 1122 + }, + { + "epoch": 0.2, + "learning_rate": 2.941408809610484e-06, + "loss": 0.9435, + "step": 1123 + }, + { + "epoch": 0.2, + "learning_rate": 2.941354204586822e-06, + "loss": 1.8586, + "step": 1124 + }, + { + "epoch": 0.2, + "learning_rate": 2.94129959956316e-06, + "loss": 0.7451, + "step": 1125 + }, + { + "epoch": 0.2, + "learning_rate": 2.941244994539498e-06, + "loss": 1.3242, + "step": 1126 + }, + { + "epoch": 0.2, + "learning_rate": 2.9411903895158355e-06, + "loss": 1.0123, + "step": 1127 + }, + { + "epoch": 0.21, + "learning_rate": 2.9411357844921735e-06, + "loss": 1.008, + "step": 1128 + }, + { + "epoch": 0.21, + "learning_rate": 2.9410811794685114e-06, + "loss": 1.7107, + "step": 1129 + }, + { + "epoch": 0.21, + "learning_rate": 2.941026574444849e-06, + "loss": 0.8767, + "step": 1130 + }, + { + "epoch": 0.21, + "learning_rate": 2.940971969421187e-06, + "loss": 0.188, + "step": 1131 + }, + { + "epoch": 0.21, + "learning_rate": 2.940917364397525e-06, + "loss": 0.6567, + "step": 1132 + }, + { + "epoch": 0.21, + "learning_rate": 2.9408627593738623e-06, + "loss": 1.0766, + "step": 1133 + }, + { + "epoch": 0.21, + "learning_rate": 2.9408081543502003e-06, + "loss": 0.8157, + "step": 1134 + }, + { + "epoch": 0.21, + "learning_rate": 2.9407535493265382e-06, + "loss": 0.895, + "step": 1135 + }, + { + "epoch": 0.21, + "learning_rate": 2.9406989443028757e-06, + "loss": 1.9535, + "step": 1136 + }, + { + "epoch": 0.21, + "learning_rate": 2.9406443392792137e-06, + "loss": 1.9496, + "step": 1137 + }, + { + "epoch": 0.21, + "learning_rate": 2.9405897342555516e-06, + "loss": 0.1625, + "step": 1138 + }, + { + "epoch": 0.21, + "learning_rate": 2.9405351292318896e-06, + "loss": 1.4565, + "step": 1139 + }, + { + "epoch": 0.21, + "learning_rate": 2.9404805242082275e-06, + "loss": 1.4389, + "step": 1140 + }, + { + "epoch": 0.21, + "learning_rate": 2.940425919184565e-06, + "loss": 0.8975, + "step": 1141 + }, + { + "epoch": 0.21, + "learning_rate": 2.940371314160903e-06, + "loss": 1.4234, + "step": 1142 + }, + { + "epoch": 0.21, + "learning_rate": 2.940316709137241e-06, + "loss": 2.7405, + "step": 1143 + }, + { + "epoch": 0.21, + "learning_rate": 2.9402621041135784e-06, + "loss": 0.5535, + "step": 1144 + }, + { + "epoch": 0.21, + "learning_rate": 2.9402074990899164e-06, + "loss": 0.5035, + "step": 1145 + }, + { + "epoch": 0.21, + "learning_rate": 2.9401528940662543e-06, + "loss": 1.7929, + "step": 1146 + }, + { + "epoch": 0.21, + "learning_rate": 2.940098289042592e-06, + "loss": 0.6907, + "step": 1147 + }, + { + "epoch": 0.21, + "learning_rate": 2.9400436840189298e-06, + "loss": 0.8299, + "step": 1148 + }, + { + "epoch": 0.21, + "learning_rate": 2.9399890789952673e-06, + "loss": 1.2637, + "step": 1149 + }, + { + "epoch": 0.21, + "learning_rate": 2.9399344739716053e-06, + "loss": 0.0615, + "step": 1150 + }, + { + "epoch": 0.21, + "learning_rate": 2.939879868947943e-06, + "loss": 1.8947, + "step": 1151 + }, + { + "epoch": 0.21, + "learning_rate": 2.939825263924281e-06, + "loss": 0.1222, + "step": 1152 + }, + { + "epoch": 0.21, + "learning_rate": 2.939770658900619e-06, + "loss": 1.421, + "step": 1153 + }, + { + "epoch": 0.21, + "learning_rate": 2.939716053876957e-06, + "loss": 0.2322, + "step": 1154 + }, + { + "epoch": 0.21, + "learning_rate": 2.9396614488532945e-06, + "loss": 1.4256, + "step": 1155 + }, + { + "epoch": 0.21, + "learning_rate": 2.9396068438296325e-06, + "loss": 1.208, + "step": 1156 + }, + { + "epoch": 0.21, + "learning_rate": 2.9395522388059704e-06, + "loss": 0.9818, + "step": 1157 + }, + { + "epoch": 0.21, + "learning_rate": 2.939497633782308e-06, + "loss": 1.0521, + "step": 1158 + }, + { + "epoch": 0.21, + "learning_rate": 2.939443028758646e-06, + "loss": 0.1393, + "step": 1159 + }, + { + "epoch": 0.21, + "learning_rate": 2.939388423734984e-06, + "loss": 1.5098, + "step": 1160 + }, + { + "epoch": 0.21, + "learning_rate": 2.9393338187113214e-06, + "loss": 0.8088, + "step": 1161 + }, + { + "epoch": 0.21, + "learning_rate": 2.9392792136876593e-06, + "loss": 0.7254, + "step": 1162 + }, + { + "epoch": 0.21, + "learning_rate": 2.939224608663997e-06, + "loss": 1.7678, + "step": 1163 + }, + { + "epoch": 0.21, + "learning_rate": 2.9391700036403348e-06, + "loss": 2.7266, + "step": 1164 + }, + { + "epoch": 0.21, + "learning_rate": 2.9391153986166727e-06, + "loss": 0.1606, + "step": 1165 + }, + { + "epoch": 0.21, + "learning_rate": 2.9390607935930107e-06, + "loss": 0.9223, + "step": 1166 + }, + { + "epoch": 0.21, + "learning_rate": 2.9390061885693486e-06, + "loss": 2.2397, + "step": 1167 + }, + { + "epoch": 0.21, + "learning_rate": 2.9389515835456865e-06, + "loss": 0.0967, + "step": 1168 + }, + { + "epoch": 0.21, + "learning_rate": 2.938896978522024e-06, + "loss": 0.7307, + "step": 1169 + }, + { + "epoch": 0.21, + "learning_rate": 2.938842373498362e-06, + "loss": 1.2148, + "step": 1170 + }, + { + "epoch": 0.21, + "learning_rate": 2.9387877684747e-06, + "loss": 1.415, + "step": 1171 + }, + { + "epoch": 0.21, + "learning_rate": 2.9387331634510375e-06, + "loss": 0.9456, + "step": 1172 + }, + { + "epoch": 0.21, + "learning_rate": 2.9386785584273754e-06, + "loss": 0.6818, + "step": 1173 + }, + { + "epoch": 0.21, + "learning_rate": 2.9386239534037134e-06, + "loss": 0.2457, + "step": 1174 + }, + { + "epoch": 0.21, + "learning_rate": 2.938569348380051e-06, + "loss": 0.3798, + "step": 1175 + }, + { + "epoch": 0.21, + "learning_rate": 2.938514743356389e-06, + "loss": 0.5738, + "step": 1176 + }, + { + "epoch": 0.21, + "learning_rate": 2.9384601383327263e-06, + "loss": 2.3764, + "step": 1177 + }, + { + "epoch": 0.21, + "learning_rate": 2.9384055333090643e-06, + "loss": 2.4391, + "step": 1178 + }, + { + "epoch": 0.21, + "learning_rate": 2.9383509282854022e-06, + "loss": 0.1782, + "step": 1179 + }, + { + "epoch": 0.21, + "learning_rate": 2.93829632326174e-06, + "loss": 1.7839, + "step": 1180 + }, + { + "epoch": 0.21, + "learning_rate": 2.938241718238078e-06, + "loss": 2.887, + "step": 1181 + }, + { + "epoch": 0.21, + "learning_rate": 2.938187113214416e-06, + "loss": 0.3674, + "step": 1182 + }, + { + "epoch": 0.22, + "learning_rate": 2.9381325081907536e-06, + "loss": 1.7996, + "step": 1183 + }, + { + "epoch": 0.22, + "learning_rate": 2.9380779031670915e-06, + "loss": 0.9102, + "step": 1184 + }, + { + "epoch": 0.22, + "learning_rate": 2.9380232981434295e-06, + "loss": 1.3625, + "step": 1185 + }, + { + "epoch": 0.22, + "learning_rate": 2.937968693119767e-06, + "loss": 1.5273, + "step": 1186 + }, + { + "epoch": 0.22, + "learning_rate": 2.937914088096105e-06, + "loss": 1.2069, + "step": 1187 + }, + { + "epoch": 0.22, + "learning_rate": 2.937859483072443e-06, + "loss": 0.1573, + "step": 1188 + }, + { + "epoch": 0.22, + "learning_rate": 2.9378048780487804e-06, + "loss": 1.5465, + "step": 1189 + }, + { + "epoch": 0.22, + "learning_rate": 2.9377502730251183e-06, + "loss": 1.4757, + "step": 1190 + }, + { + "epoch": 0.22, + "learning_rate": 2.937695668001456e-06, + "loss": 0.6042, + "step": 1191 + }, + { + "epoch": 0.22, + "learning_rate": 2.937641062977794e-06, + "loss": 1.5351, + "step": 1192 + }, + { + "epoch": 0.22, + "learning_rate": 2.9375864579541317e-06, + "loss": 0.6033, + "step": 1193 + }, + { + "epoch": 0.22, + "learning_rate": 2.9375318529304697e-06, + "loss": 0.1196, + "step": 1194 + }, + { + "epoch": 0.22, + "learning_rate": 2.9374772479068076e-06, + "loss": 1.1151, + "step": 1195 + }, + { + "epoch": 0.22, + "learning_rate": 2.9374226428831456e-06, + "loss": 0.404, + "step": 1196 + }, + { + "epoch": 0.22, + "learning_rate": 2.937368037859483e-06, + "loss": 2.1587, + "step": 1197 + }, + { + "epoch": 0.22, + "learning_rate": 2.937313432835821e-06, + "loss": 1.676, + "step": 1198 + }, + { + "epoch": 0.22, + "learning_rate": 2.937258827812159e-06, + "loss": 0.0867, + "step": 1199 + }, + { + "epoch": 0.22, + "learning_rate": 2.9372042227884965e-06, + "loss": 2.6285, + "step": 1200 + }, + { + "epoch": 0.22, + "learning_rate": 2.9371496177648344e-06, + "loss": 2.0229, + "step": 1201 + }, + { + "epoch": 0.22, + "learning_rate": 2.9370950127411724e-06, + "loss": 1.1177, + "step": 1202 + }, + { + "epoch": 0.22, + "learning_rate": 2.93704040771751e-06, + "loss": 2.3439, + "step": 1203 + }, + { + "epoch": 0.22, + "learning_rate": 2.936985802693848e-06, + "loss": 0.4813, + "step": 1204 + }, + { + "epoch": 0.22, + "learning_rate": 2.9369311976701858e-06, + "loss": 0.9145, + "step": 1205 + }, + { + "epoch": 0.22, + "learning_rate": 2.9368765926465237e-06, + "loss": 1.0875, + "step": 1206 + }, + { + "epoch": 0.22, + "learning_rate": 2.9368219876228617e-06, + "loss": 1.4439, + "step": 1207 + }, + { + "epoch": 0.22, + "learning_rate": 2.936767382599199e-06, + "loss": 2.3866, + "step": 1208 + }, + { + "epoch": 0.22, + "learning_rate": 2.936712777575537e-06, + "loss": 2.4005, + "step": 1209 + }, + { + "epoch": 0.22, + "learning_rate": 2.936658172551875e-06, + "loss": 0.7257, + "step": 1210 + }, + { + "epoch": 0.22, + "learning_rate": 2.9366035675282126e-06, + "loss": 0.5266, + "step": 1211 + }, + { + "epoch": 0.22, + "learning_rate": 2.9365489625045505e-06, + "loss": 1.3377, + "step": 1212 + }, + { + "epoch": 0.22, + "learning_rate": 2.9364943574808885e-06, + "loss": 1.1635, + "step": 1213 + }, + { + "epoch": 0.22, + "learning_rate": 2.936439752457226e-06, + "loss": 2.0403, + "step": 1214 + }, + { + "epoch": 0.22, + "learning_rate": 2.936385147433564e-06, + "loss": 1.3117, + "step": 1215 + }, + { + "epoch": 0.22, + "learning_rate": 2.9363305424099015e-06, + "loss": 0.5542, + "step": 1216 + }, + { + "epoch": 0.22, + "learning_rate": 2.9362759373862394e-06, + "loss": 1.4041, + "step": 1217 + }, + { + "epoch": 0.22, + "learning_rate": 2.9362213323625774e-06, + "loss": 0.0916, + "step": 1218 + }, + { + "epoch": 0.22, + "learning_rate": 2.9361667273389153e-06, + "loss": 2.6, + "step": 1219 + }, + { + "epoch": 0.22, + "learning_rate": 2.9361121223152532e-06, + "loss": 1.1124, + "step": 1220 + }, + { + "epoch": 0.22, + "learning_rate": 2.936057517291591e-06, + "loss": 2.3454, + "step": 1221 + }, + { + "epoch": 0.22, + "learning_rate": 2.9360029122679287e-06, + "loss": 0.3118, + "step": 1222 + }, + { + "epoch": 0.22, + "learning_rate": 2.9359483072442666e-06, + "loss": 0.6112, + "step": 1223 + }, + { + "epoch": 0.22, + "learning_rate": 2.9358937022206046e-06, + "loss": 3.1157, + "step": 1224 + }, + { + "epoch": 0.22, + "learning_rate": 2.935839097196942e-06, + "loss": 0.7653, + "step": 1225 + }, + { + "epoch": 0.22, + "learning_rate": 2.93578449217328e-06, + "loss": 0.7838, + "step": 1226 + }, + { + "epoch": 0.22, + "learning_rate": 2.935729887149618e-06, + "loss": 0.7078, + "step": 1227 + }, + { + "epoch": 0.22, + "learning_rate": 2.9356752821259555e-06, + "loss": 0.7552, + "step": 1228 + }, + { + "epoch": 0.22, + "learning_rate": 2.9356206771022935e-06, + "loss": 0.5157, + "step": 1229 + }, + { + "epoch": 0.22, + "learning_rate": 2.935566072078631e-06, + "loss": 0.809, + "step": 1230 + }, + { + "epoch": 0.22, + "learning_rate": 2.935511467054969e-06, + "loss": 0.5465, + "step": 1231 + }, + { + "epoch": 0.22, + "learning_rate": 2.935456862031307e-06, + "loss": 0.9115, + "step": 1232 + }, + { + "epoch": 0.22, + "learning_rate": 2.935402257007645e-06, + "loss": 1.1207, + "step": 1233 + }, + { + "epoch": 0.22, + "learning_rate": 2.9353476519839828e-06, + "loss": 1.4806, + "step": 1234 + }, + { + "epoch": 0.22, + "learning_rate": 2.9352930469603207e-06, + "loss": 0.7334, + "step": 1235 + }, + { + "epoch": 0.22, + "learning_rate": 2.9352384419366582e-06, + "loss": 1.0456, + "step": 1236 + }, + { + "epoch": 0.22, + "learning_rate": 2.935183836912996e-06, + "loss": 1.7637, + "step": 1237 + }, + { + "epoch": 0.23, + "learning_rate": 2.935129231889334e-06, + "loss": 1.8008, + "step": 1238 + }, + { + "epoch": 0.23, + "learning_rate": 2.9350746268656716e-06, + "loss": 2.3588, + "step": 1239 + }, + { + "epoch": 0.23, + "learning_rate": 2.9350200218420096e-06, + "loss": 1.7962, + "step": 1240 + }, + { + "epoch": 0.23, + "learning_rate": 2.9349654168183475e-06, + "loss": 1.9676, + "step": 1241 + }, + { + "epoch": 0.23, + "learning_rate": 2.934910811794685e-06, + "loss": 1.0876, + "step": 1242 + }, + { + "epoch": 0.23, + "learning_rate": 2.934856206771023e-06, + "loss": 1.7319, + "step": 1243 + }, + { + "epoch": 0.23, + "learning_rate": 2.9348016017473605e-06, + "loss": 1.5325, + "step": 1244 + }, + { + "epoch": 0.23, + "learning_rate": 2.9347469967236984e-06, + "loss": 1.0523, + "step": 1245 + }, + { + "epoch": 0.23, + "learning_rate": 2.9346923917000364e-06, + "loss": 0.1803, + "step": 1246 + }, + { + "epoch": 0.23, + "learning_rate": 2.9346377866763743e-06, + "loss": 0.08, + "step": 1247 + }, + { + "epoch": 0.23, + "learning_rate": 2.9345831816527123e-06, + "loss": 0.7393, + "step": 1248 + }, + { + "epoch": 0.23, + "learning_rate": 2.9345285766290502e-06, + "loss": 0.4246, + "step": 1249 + }, + { + "epoch": 0.23, + "learning_rate": 2.9344739716053877e-06, + "loss": 1.0125, + "step": 1250 + }, + { + "epoch": 0.23, + "learning_rate": 2.9344193665817257e-06, + "loss": 1.7755, + "step": 1251 + }, + { + "epoch": 0.23, + "learning_rate": 2.9343647615580636e-06, + "loss": 1.4899, + "step": 1252 + }, + { + "epoch": 0.23, + "learning_rate": 2.934310156534401e-06, + "loss": 1.4543, + "step": 1253 + }, + { + "epoch": 0.23, + "learning_rate": 2.934255551510739e-06, + "loss": 1.5957, + "step": 1254 + }, + { + "epoch": 0.23, + "learning_rate": 2.934200946487077e-06, + "loss": 0.9339, + "step": 1255 + }, + { + "epoch": 0.23, + "learning_rate": 2.9341463414634145e-06, + "loss": 1.2454, + "step": 1256 + }, + { + "epoch": 0.23, + "learning_rate": 2.9340917364397525e-06, + "loss": 0.4287, + "step": 1257 + }, + { + "epoch": 0.23, + "learning_rate": 2.93403713141609e-06, + "loss": 1.2703, + "step": 1258 + }, + { + "epoch": 0.23, + "learning_rate": 2.933982526392428e-06, + "loss": 0.3599, + "step": 1259 + }, + { + "epoch": 0.23, + "learning_rate": 2.933927921368766e-06, + "loss": 0.997, + "step": 1260 + }, + { + "epoch": 0.23, + "learning_rate": 2.933873316345104e-06, + "loss": 2.2492, + "step": 1261 + }, + { + "epoch": 0.23, + "learning_rate": 2.9338187113214418e-06, + "loss": 0.4571, + "step": 1262 + }, + { + "epoch": 0.23, + "learning_rate": 2.9337641062977797e-06, + "loss": 1.8625, + "step": 1263 + }, + { + "epoch": 0.23, + "learning_rate": 2.9337095012741172e-06, + "loss": 0.5352, + "step": 1264 + }, + { + "epoch": 0.23, + "learning_rate": 2.933654896250455e-06, + "loss": 1.5612, + "step": 1265 + }, + { + "epoch": 0.23, + "learning_rate": 2.933600291226793e-06, + "loss": 1.9513, + "step": 1266 + }, + { + "epoch": 0.23, + "learning_rate": 2.9335456862031307e-06, + "loss": 1.0245, + "step": 1267 + }, + { + "epoch": 0.23, + "learning_rate": 2.9334910811794686e-06, + "loss": 0.9398, + "step": 1268 + }, + { + "epoch": 0.23, + "learning_rate": 2.9334364761558065e-06, + "loss": 1.2568, + "step": 1269 + }, + { + "epoch": 0.23, + "learning_rate": 2.933381871132144e-06, + "loss": 1.8061, + "step": 1270 + }, + { + "epoch": 0.23, + "learning_rate": 2.933327266108482e-06, + "loss": 1.1644, + "step": 1271 + }, + { + "epoch": 0.23, + "learning_rate": 2.9332726610848195e-06, + "loss": 0.1001, + "step": 1272 + }, + { + "epoch": 0.23, + "learning_rate": 2.9332180560611575e-06, + "loss": 2.7606, + "step": 1273 + }, + { + "epoch": 0.23, + "learning_rate": 2.9331634510374954e-06, + "loss": 0.9059, + "step": 1274 + }, + { + "epoch": 0.23, + "learning_rate": 2.9331088460138334e-06, + "loss": 1.2998, + "step": 1275 + }, + { + "epoch": 0.23, + "learning_rate": 2.9330542409901713e-06, + "loss": 1.0261, + "step": 1276 + }, + { + "epoch": 0.23, + "learning_rate": 2.9329996359665092e-06, + "loss": 1.1444, + "step": 1277 + }, + { + "epoch": 0.23, + "learning_rate": 2.9329450309428468e-06, + "loss": 0.133, + "step": 1278 + }, + { + "epoch": 0.23, + "learning_rate": 2.9328904259191847e-06, + "loss": 2.2653, + "step": 1279 + }, + { + "epoch": 0.23, + "learning_rate": 2.9328358208955226e-06, + "loss": 0.602, + "step": 1280 + }, + { + "epoch": 0.23, + "learning_rate": 2.93278121587186e-06, + "loss": 0.4261, + "step": 1281 + }, + { + "epoch": 0.23, + "learning_rate": 2.932726610848198e-06, + "loss": 1.6646, + "step": 1282 + }, + { + "epoch": 0.23, + "learning_rate": 2.932672005824536e-06, + "loss": 1.4547, + "step": 1283 + }, + { + "epoch": 0.23, + "learning_rate": 2.9326174008008736e-06, + "loss": 0.2682, + "step": 1284 + }, + { + "epoch": 0.23, + "learning_rate": 2.9325627957772115e-06, + "loss": 0.5084, + "step": 1285 + }, + { + "epoch": 0.23, + "learning_rate": 2.9325081907535495e-06, + "loss": 0.6663, + "step": 1286 + }, + { + "epoch": 0.23, + "learning_rate": 2.9324535857298874e-06, + "loss": 1.0605, + "step": 1287 + }, + { + "epoch": 0.23, + "learning_rate": 2.9323989807062253e-06, + "loss": 0.5247, + "step": 1288 + }, + { + "epoch": 0.23, + "learning_rate": 2.932344375682563e-06, + "loss": 2.7604, + "step": 1289 + }, + { + "epoch": 0.23, + "learning_rate": 2.932289770658901e-06, + "loss": 0.8332, + "step": 1290 + }, + { + "epoch": 0.23, + "learning_rate": 2.9322351656352388e-06, + "loss": 1.5897, + "step": 1291 + }, + { + "epoch": 0.23, + "learning_rate": 2.9321805606115763e-06, + "loss": 0.6922, + "step": 1292 + }, + { + "epoch": 0.24, + "learning_rate": 2.9321259555879142e-06, + "loss": 1.3272, + "step": 1293 + }, + { + "epoch": 0.24, + "learning_rate": 2.932071350564252e-06, + "loss": 2.1455, + "step": 1294 + }, + { + "epoch": 0.24, + "learning_rate": 2.9320167455405897e-06, + "loss": 1.1626, + "step": 1295 + }, + { + "epoch": 0.24, + "learning_rate": 2.9319621405169276e-06, + "loss": 0.656, + "step": 1296 + }, + { + "epoch": 0.24, + "learning_rate": 2.931907535493265e-06, + "loss": 0.6462, + "step": 1297 + }, + { + "epoch": 0.24, + "learning_rate": 2.931852930469603e-06, + "loss": 0.9209, + "step": 1298 + }, + { + "epoch": 0.24, + "learning_rate": 2.931798325445941e-06, + "loss": 0.3773, + "step": 1299 + }, + { + "epoch": 0.24, + "learning_rate": 2.931743720422279e-06, + "loss": 1.0945, + "step": 1300 + }, + { + "epoch": 0.24, + "learning_rate": 2.931689115398617e-06, + "loss": 0.2643, + "step": 1301 + }, + { + "epoch": 0.24, + "learning_rate": 2.931634510374955e-06, + "loss": 1.1894, + "step": 1302 + }, + { + "epoch": 0.24, + "learning_rate": 2.9315799053512924e-06, + "loss": 0.9616, + "step": 1303 + }, + { + "epoch": 0.24, + "learning_rate": 2.9315253003276303e-06, + "loss": 2.4924, + "step": 1304 + }, + { + "epoch": 0.24, + "learning_rate": 2.9314706953039683e-06, + "loss": 2.7214, + "step": 1305 + }, + { + "epoch": 0.24, + "learning_rate": 2.9314160902803058e-06, + "loss": 0.9752, + "step": 1306 + }, + { + "epoch": 0.24, + "learning_rate": 2.9313614852566437e-06, + "loss": 0.806, + "step": 1307 + }, + { + "epoch": 0.24, + "learning_rate": 2.9313068802329817e-06, + "loss": 0.4023, + "step": 1308 + }, + { + "epoch": 0.24, + "learning_rate": 2.931252275209319e-06, + "loss": 1.5316, + "step": 1309 + }, + { + "epoch": 0.24, + "learning_rate": 2.931197670185657e-06, + "loss": 2.4086, + "step": 1310 + }, + { + "epoch": 0.24, + "learning_rate": 2.9311430651619947e-06, + "loss": 1.6652, + "step": 1311 + }, + { + "epoch": 0.24, + "learning_rate": 2.9310884601383326e-06, + "loss": 0.0866, + "step": 1312 + }, + { + "epoch": 0.24, + "learning_rate": 2.9310338551146705e-06, + "loss": 0.974, + "step": 1313 + }, + { + "epoch": 0.24, + "learning_rate": 2.9309792500910085e-06, + "loss": 1.2698, + "step": 1314 + }, + { + "epoch": 0.24, + "learning_rate": 2.9309246450673464e-06, + "loss": 1.8906, + "step": 1315 + }, + { + "epoch": 0.24, + "learning_rate": 2.9308700400436844e-06, + "loss": 0.7882, + "step": 1316 + }, + { + "epoch": 0.24, + "learning_rate": 2.930815435020022e-06, + "loss": 1.2598, + "step": 1317 + }, + { + "epoch": 0.24, + "learning_rate": 2.93076082999636e-06, + "loss": 1.8011, + "step": 1318 + }, + { + "epoch": 0.24, + "learning_rate": 2.9307062249726978e-06, + "loss": 0.8394, + "step": 1319 + }, + { + "epoch": 0.24, + "learning_rate": 2.9306516199490353e-06, + "loss": 1.1129, + "step": 1320 + }, + { + "epoch": 0.24, + "learning_rate": 2.9305970149253732e-06, + "loss": 0.7214, + "step": 1321 + }, + { + "epoch": 0.24, + "learning_rate": 2.930542409901711e-06, + "loss": 0.9736, + "step": 1322 + }, + { + "epoch": 0.24, + "learning_rate": 2.9304878048780487e-06, + "loss": 0.7185, + "step": 1323 + }, + { + "epoch": 0.24, + "learning_rate": 2.9304331998543866e-06, + "loss": 0.2651, + "step": 1324 + }, + { + "epoch": 0.24, + "learning_rate": 2.930378594830724e-06, + "loss": 0.1029, + "step": 1325 + }, + { + "epoch": 0.24, + "learning_rate": 2.930323989807062e-06, + "loss": 1.131, + "step": 1326 + }, + { + "epoch": 0.24, + "learning_rate": 2.9302693847834e-06, + "loss": 0.3936, + "step": 1327 + }, + { + "epoch": 0.24, + "learning_rate": 2.930214779759738e-06, + "loss": 0.7548, + "step": 1328 + }, + { + "epoch": 0.24, + "learning_rate": 2.930160174736076e-06, + "loss": 1.0649, + "step": 1329 + }, + { + "epoch": 0.24, + "learning_rate": 2.930105569712414e-06, + "loss": 2.1364, + "step": 1330 + }, + { + "epoch": 0.24, + "learning_rate": 2.9300509646887514e-06, + "loss": 3.3475, + "step": 1331 + }, + { + "epoch": 0.24, + "learning_rate": 2.9299963596650893e-06, + "loss": 0.6484, + "step": 1332 + }, + { + "epoch": 0.24, + "learning_rate": 2.9299417546414273e-06, + "loss": 0.5357, + "step": 1333 + }, + { + "epoch": 0.24, + "learning_rate": 2.929887149617765e-06, + "loss": 0.5034, + "step": 1334 + }, + { + "epoch": 0.24, + "learning_rate": 2.9298325445941028e-06, + "loss": 0.5772, + "step": 1335 + }, + { + "epoch": 0.24, + "learning_rate": 2.9297779395704407e-06, + "loss": 0.6226, + "step": 1336 + }, + { + "epoch": 0.24, + "learning_rate": 2.9297233345467782e-06, + "loss": 0.1471, + "step": 1337 + }, + { + "epoch": 0.24, + "learning_rate": 2.929668729523116e-06, + "loss": 1.2149, + "step": 1338 + }, + { + "epoch": 0.24, + "learning_rate": 2.9296141244994537e-06, + "loss": 0.4917, + "step": 1339 + }, + { + "epoch": 0.24, + "learning_rate": 2.9295595194757916e-06, + "loss": 2.5066, + "step": 1340 + }, + { + "epoch": 0.24, + "learning_rate": 2.9295049144521296e-06, + "loss": 1.2576, + "step": 1341 + }, + { + "epoch": 0.24, + "learning_rate": 2.9294503094284675e-06, + "loss": 1.1464, + "step": 1342 + }, + { + "epoch": 0.24, + "learning_rate": 2.9293957044048055e-06, + "loss": 1.1806, + "step": 1343 + }, + { + "epoch": 0.24, + "learning_rate": 2.9293410993811434e-06, + "loss": 1.7771, + "step": 1344 + }, + { + "epoch": 0.24, + "learning_rate": 2.929286494357481e-06, + "loss": 1.5361, + "step": 1345 + }, + { + "epoch": 0.24, + "learning_rate": 2.929231889333819e-06, + "loss": 0.9998, + "step": 1346 + }, + { + "epoch": 0.24, + "learning_rate": 2.929177284310157e-06, + "loss": 0.7872, + "step": 1347 + }, + { + "epoch": 0.25, + "learning_rate": 2.9291226792864943e-06, + "loss": 1.029, + "step": 1348 + }, + { + "epoch": 0.25, + "learning_rate": 2.9290680742628323e-06, + "loss": 0.3756, + "step": 1349 + }, + { + "epoch": 0.25, + "learning_rate": 2.9290134692391702e-06, + "loss": 1.1268, + "step": 1350 + }, + { + "epoch": 0.25, + "learning_rate": 2.9289588642155077e-06, + "loss": 1.2034, + "step": 1351 + }, + { + "epoch": 0.25, + "learning_rate": 2.9289042591918457e-06, + "loss": 0.9925, + "step": 1352 + }, + { + "epoch": 0.25, + "learning_rate": 2.928849654168183e-06, + "loss": 2.3601, + "step": 1353 + }, + { + "epoch": 0.25, + "learning_rate": 2.928795049144521e-06, + "loss": 1.0985, + "step": 1354 + }, + { + "epoch": 0.25, + "learning_rate": 2.928740444120859e-06, + "loss": 0.1589, + "step": 1355 + }, + { + "epoch": 0.25, + "learning_rate": 2.928685839097197e-06, + "loss": 0.8947, + "step": 1356 + }, + { + "epoch": 0.25, + "learning_rate": 2.928631234073535e-06, + "loss": 1.8837, + "step": 1357 + }, + { + "epoch": 0.25, + "learning_rate": 2.928576629049873e-06, + "loss": 0.6971, + "step": 1358 + }, + { + "epoch": 0.25, + "learning_rate": 2.9285220240262104e-06, + "loss": 1.3197, + "step": 1359 + }, + { + "epoch": 0.25, + "learning_rate": 2.9284674190025484e-06, + "loss": 1.6395, + "step": 1360 + }, + { + "epoch": 0.25, + "learning_rate": 2.9284128139788863e-06, + "loss": 1.7253, + "step": 1361 + }, + { + "epoch": 0.25, + "learning_rate": 2.928358208955224e-06, + "loss": 0.1246, + "step": 1362 + }, + { + "epoch": 0.25, + "learning_rate": 2.9283036039315618e-06, + "loss": 0.9812, + "step": 1363 + }, + { + "epoch": 0.25, + "learning_rate": 2.9282489989078997e-06, + "loss": 1.0285, + "step": 1364 + }, + { + "epoch": 0.25, + "learning_rate": 2.9281943938842372e-06, + "loss": 0.3853, + "step": 1365 + }, + { + "epoch": 0.25, + "learning_rate": 2.928139788860575e-06, + "loss": 0.4869, + "step": 1366 + }, + { + "epoch": 0.25, + "learning_rate": 2.928085183836913e-06, + "loss": 0.4105, + "step": 1367 + }, + { + "epoch": 0.25, + "learning_rate": 2.928030578813251e-06, + "loss": 1.9341, + "step": 1368 + }, + { + "epoch": 0.25, + "learning_rate": 2.927975973789589e-06, + "loss": 1.0484, + "step": 1369 + }, + { + "epoch": 0.25, + "learning_rate": 2.9279213687659265e-06, + "loss": 0.4008, + "step": 1370 + }, + { + "epoch": 0.25, + "learning_rate": 2.9278667637422645e-06, + "loss": 2.2185, + "step": 1371 + }, + { + "epoch": 0.25, + "learning_rate": 2.9278121587186024e-06, + "loss": 1.5251, + "step": 1372 + }, + { + "epoch": 0.25, + "learning_rate": 2.92775755369494e-06, + "loss": 0.9281, + "step": 1373 + }, + { + "epoch": 0.25, + "learning_rate": 2.927702948671278e-06, + "loss": 0.3186, + "step": 1374 + }, + { + "epoch": 0.25, + "learning_rate": 2.927648343647616e-06, + "loss": 2.5679, + "step": 1375 + }, + { + "epoch": 0.25, + "learning_rate": 2.9275937386239534e-06, + "loss": 0.9265, + "step": 1376 + }, + { + "epoch": 0.25, + "learning_rate": 2.9275391336002913e-06, + "loss": 0.367, + "step": 1377 + }, + { + "epoch": 0.25, + "learning_rate": 2.927484528576629e-06, + "loss": 1.9514, + "step": 1378 + }, + { + "epoch": 0.25, + "learning_rate": 2.9274299235529668e-06, + "loss": 0.7325, + "step": 1379 + }, + { + "epoch": 0.25, + "learning_rate": 2.9273753185293047e-06, + "loss": 3.3148, + "step": 1380 + }, + { + "epoch": 0.25, + "learning_rate": 2.9273207135056426e-06, + "loss": 0.7711, + "step": 1381 + }, + { + "epoch": 0.25, + "learning_rate": 2.9272661084819806e-06, + "loss": 0.1234, + "step": 1382 + }, + { + "epoch": 0.25, + "learning_rate": 2.9272115034583185e-06, + "loss": 1.1183, + "step": 1383 + }, + { + "epoch": 0.25, + "learning_rate": 2.927156898434656e-06, + "loss": 0.7126, + "step": 1384 + }, + { + "epoch": 0.25, + "learning_rate": 2.927102293410994e-06, + "loss": 1.0338, + "step": 1385 + }, + { + "epoch": 0.25, + "learning_rate": 2.927047688387332e-06, + "loss": 0.0307, + "step": 1386 + }, + { + "epoch": 0.25, + "learning_rate": 2.9269930833636695e-06, + "loss": 1.9812, + "step": 1387 + }, + { + "epoch": 0.25, + "learning_rate": 2.9269384783400074e-06, + "loss": 1.6128, + "step": 1388 + }, + { + "epoch": 0.25, + "learning_rate": 2.9268838733163453e-06, + "loss": 0.0874, + "step": 1389 + }, + { + "epoch": 0.25, + "learning_rate": 2.926829268292683e-06, + "loss": 0.6942, + "step": 1390 + }, + { + "epoch": 0.25, + "learning_rate": 2.926774663269021e-06, + "loss": 0.4437, + "step": 1391 + }, + { + "epoch": 0.25, + "learning_rate": 2.9267200582453583e-06, + "loss": 1.4794, + "step": 1392 + }, + { + "epoch": 0.25, + "learning_rate": 2.9266654532216963e-06, + "loss": 0.2957, + "step": 1393 + }, + { + "epoch": 0.25, + "learning_rate": 2.9266108481980342e-06, + "loss": 1.1862, + "step": 1394 + }, + { + "epoch": 0.25, + "learning_rate": 2.926556243174372e-06, + "loss": 1.4925, + "step": 1395 + }, + { + "epoch": 0.25, + "learning_rate": 2.92650163815071e-06, + "loss": 1.8411, + "step": 1396 + }, + { + "epoch": 0.25, + "learning_rate": 2.926447033127048e-06, + "loss": 1.7646, + "step": 1397 + }, + { + "epoch": 0.25, + "learning_rate": 2.9263924281033856e-06, + "loss": 1.2807, + "step": 1398 + }, + { + "epoch": 0.25, + "learning_rate": 2.9263378230797235e-06, + "loss": 0.5226, + "step": 1399 + }, + { + "epoch": 0.25, + "learning_rate": 2.9262832180560615e-06, + "loss": 0.5376, + "step": 1400 + }, + { + "epoch": 0.25, + "learning_rate": 2.926228613032399e-06, + "loss": 0.9927, + "step": 1401 + }, + { + "epoch": 0.25, + "learning_rate": 2.926174008008737e-06, + "loss": 4.2672, + "step": 1402 + }, + { + "epoch": 0.26, + "learning_rate": 2.926119402985075e-06, + "loss": 3.447, + "step": 1403 + }, + { + "epoch": 0.26, + "learning_rate": 2.9260647979614124e-06, + "loss": 0.8423, + "step": 1404 + }, + { + "epoch": 0.26, + "learning_rate": 2.9260101929377503e-06, + "loss": 1.8727, + "step": 1405 + }, + { + "epoch": 0.26, + "learning_rate": 2.925955587914088e-06, + "loss": 0.6792, + "step": 1406 + }, + { + "epoch": 0.26, + "learning_rate": 2.9259009828904258e-06, + "loss": 0.5272, + "step": 1407 + }, + { + "epoch": 0.26, + "learning_rate": 2.9258463778667637e-06, + "loss": 1.4837, + "step": 1408 + }, + { + "epoch": 0.26, + "learning_rate": 2.9257917728431017e-06, + "loss": 1.1793, + "step": 1409 + }, + { + "epoch": 0.26, + "learning_rate": 2.9257371678194396e-06, + "loss": 1.5515, + "step": 1410 + }, + { + "epoch": 0.26, + "learning_rate": 2.9256825627957776e-06, + "loss": 1.4827, + "step": 1411 + }, + { + "epoch": 0.26, + "learning_rate": 2.925627957772115e-06, + "loss": 1.8537, + "step": 1412 + }, + { + "epoch": 0.26, + "learning_rate": 2.925573352748453e-06, + "loss": 1.237, + "step": 1413 + }, + { + "epoch": 0.26, + "learning_rate": 2.925518747724791e-06, + "loss": 0.7092, + "step": 1414 + }, + { + "epoch": 0.26, + "learning_rate": 2.9254641427011285e-06, + "loss": 1.4081, + "step": 1415 + }, + { + "epoch": 0.26, + "learning_rate": 2.9254095376774664e-06, + "loss": 0.7711, + "step": 1416 + }, + { + "epoch": 0.26, + "learning_rate": 2.9253549326538044e-06, + "loss": 1.3218, + "step": 1417 + }, + { + "epoch": 0.26, + "learning_rate": 2.925300327630142e-06, + "loss": 2.5654, + "step": 1418 + }, + { + "epoch": 0.26, + "learning_rate": 2.92524572260648e-06, + "loss": 0.3369, + "step": 1419 + }, + { + "epoch": 0.26, + "learning_rate": 2.9251911175828174e-06, + "loss": 0.6102, + "step": 1420 + }, + { + "epoch": 0.26, + "learning_rate": 2.9251365125591553e-06, + "loss": 0.7378, + "step": 1421 + }, + { + "epoch": 0.26, + "learning_rate": 2.9250819075354932e-06, + "loss": 0.7911, + "step": 1422 + }, + { + "epoch": 0.26, + "learning_rate": 2.925027302511831e-06, + "loss": 0.3807, + "step": 1423 + }, + { + "epoch": 0.26, + "learning_rate": 2.924972697488169e-06, + "loss": 0.9897, + "step": 1424 + }, + { + "epoch": 0.26, + "learning_rate": 2.924918092464507e-06, + "loss": 1.978, + "step": 1425 + }, + { + "epoch": 0.26, + "learning_rate": 2.9248634874408446e-06, + "loss": 1.723, + "step": 1426 + }, + { + "epoch": 0.26, + "learning_rate": 2.9248088824171825e-06, + "loss": 1.4834, + "step": 1427 + }, + { + "epoch": 0.26, + "learning_rate": 2.9247542773935205e-06, + "loss": 1.1962, + "step": 1428 + }, + { + "epoch": 0.26, + "learning_rate": 2.924699672369858e-06, + "loss": 0.984, + "step": 1429 + }, + { + "epoch": 0.26, + "learning_rate": 2.924645067346196e-06, + "loss": 1.9264, + "step": 1430 + }, + { + "epoch": 0.26, + "learning_rate": 2.924590462322534e-06, + "loss": 1.8262, + "step": 1431 + }, + { + "epoch": 0.26, + "learning_rate": 2.9245358572988714e-06, + "loss": 2.0129, + "step": 1432 + }, + { + "epoch": 0.26, + "learning_rate": 2.9244812522752094e-06, + "loss": 1.1387, + "step": 1433 + }, + { + "epoch": 0.26, + "learning_rate": 2.924426647251547e-06, + "loss": 1.2666, + "step": 1434 + }, + { + "epoch": 0.26, + "learning_rate": 2.924372042227885e-06, + "loss": 0.6292, + "step": 1435 + }, + { + "epoch": 0.26, + "learning_rate": 2.9243174372042228e-06, + "loss": 2.7434, + "step": 1436 + }, + { + "epoch": 0.26, + "learning_rate": 2.9242628321805607e-06, + "loss": 0.3579, + "step": 1437 + }, + { + "epoch": 0.26, + "learning_rate": 2.9242082271568986e-06, + "loss": 0.7057, + "step": 1438 + }, + { + "epoch": 0.26, + "learning_rate": 2.9241536221332366e-06, + "loss": 2.0441, + "step": 1439 + }, + { + "epoch": 0.26, + "learning_rate": 2.924099017109574e-06, + "loss": 1.1061, + "step": 1440 + }, + { + "epoch": 0.26, + "learning_rate": 2.924044412085912e-06, + "loss": 1.7523, + "step": 1441 + }, + { + "epoch": 0.26, + "learning_rate": 2.92398980706225e-06, + "loss": 0.9887, + "step": 1442 + }, + { + "epoch": 0.26, + "learning_rate": 2.9239352020385875e-06, + "loss": 1.6879, + "step": 1443 + }, + { + "epoch": 0.26, + "learning_rate": 2.9238805970149255e-06, + "loss": 1.166, + "step": 1444 + }, + { + "epoch": 0.26, + "learning_rate": 2.9238259919912634e-06, + "loss": 0.0864, + "step": 1445 + }, + { + "epoch": 0.26, + "learning_rate": 2.923771386967601e-06, + "loss": 0.6506, + "step": 1446 + }, + { + "epoch": 0.26, + "learning_rate": 2.923716781943939e-06, + "loss": 1.3266, + "step": 1447 + }, + { + "epoch": 0.26, + "learning_rate": 2.923662176920277e-06, + "loss": 3.4464, + "step": 1448 + }, + { + "epoch": 0.26, + "learning_rate": 2.9236075718966148e-06, + "loss": 0.7181, + "step": 1449 + }, + { + "epoch": 0.26, + "learning_rate": 2.9235529668729523e-06, + "loss": 1.4977, + "step": 1450 + }, + { + "epoch": 0.26, + "learning_rate": 2.9234983618492902e-06, + "loss": 0.4709, + "step": 1451 + }, + { + "epoch": 0.26, + "learning_rate": 2.923443756825628e-06, + "loss": 2.5069, + "step": 1452 + }, + { + "epoch": 0.26, + "learning_rate": 2.923389151801966e-06, + "loss": 0.4071, + "step": 1453 + }, + { + "epoch": 0.26, + "learning_rate": 2.9233345467783036e-06, + "loss": 0.0692, + "step": 1454 + }, + { + "epoch": 0.26, + "learning_rate": 2.9232799417546416e-06, + "loss": 0.5653, + "step": 1455 + }, + { + "epoch": 0.26, + "learning_rate": 2.9232253367309795e-06, + "loss": 0.8719, + "step": 1456 + }, + { + "epoch": 0.26, + "learning_rate": 2.923170731707317e-06, + "loss": 0.1018, + "step": 1457 + }, + { + "epoch": 0.27, + "learning_rate": 2.923116126683655e-06, + "loss": 1.8007, + "step": 1458 + }, + { + "epoch": 0.27, + "learning_rate": 2.9230615216599925e-06, + "loss": 0.7609, + "step": 1459 + }, + { + "epoch": 0.27, + "learning_rate": 2.9230069166363304e-06, + "loss": 0.5846, + "step": 1460 + }, + { + "epoch": 0.27, + "learning_rate": 2.9229523116126684e-06, + "loss": 0.2615, + "step": 1461 + }, + { + "epoch": 0.27, + "learning_rate": 2.9228977065890063e-06, + "loss": 1.0062, + "step": 1462 + }, + { + "epoch": 0.27, + "learning_rate": 2.9228431015653443e-06, + "loss": 2.2337, + "step": 1463 + }, + { + "epoch": 0.27, + "learning_rate": 2.922788496541682e-06, + "loss": 1.5725, + "step": 1464 + }, + { + "epoch": 0.27, + "learning_rate": 2.9227338915180197e-06, + "loss": 1.4215, + "step": 1465 + }, + { + "epoch": 0.27, + "learning_rate": 2.9226792864943577e-06, + "loss": 2.6899, + "step": 1466 + }, + { + "epoch": 0.27, + "learning_rate": 2.9226246814706956e-06, + "loss": 1.7751, + "step": 1467 + }, + { + "epoch": 0.27, + "learning_rate": 2.922570076447033e-06, + "loss": 0.8035, + "step": 1468 + }, + { + "epoch": 0.27, + "learning_rate": 2.922515471423371e-06, + "loss": 0.8858, + "step": 1469 + }, + { + "epoch": 0.27, + "learning_rate": 2.922460866399709e-06, + "loss": 0.6483, + "step": 1470 + }, + { + "epoch": 0.27, + "learning_rate": 2.9224062613760465e-06, + "loss": 0.9872, + "step": 1471 + }, + { + "epoch": 0.27, + "learning_rate": 2.9223516563523845e-06, + "loss": 0.6114, + "step": 1472 + }, + { + "epoch": 0.27, + "learning_rate": 2.922297051328722e-06, + "loss": 2.5537, + "step": 1473 + }, + { + "epoch": 0.27, + "learning_rate": 2.92224244630506e-06, + "loss": 1.3899, + "step": 1474 + }, + { + "epoch": 0.27, + "learning_rate": 2.922187841281398e-06, + "loss": 0.5153, + "step": 1475 + }, + { + "epoch": 0.27, + "learning_rate": 2.922133236257736e-06, + "loss": 0.9407, + "step": 1476 + }, + { + "epoch": 0.27, + "learning_rate": 2.9220786312340738e-06, + "loss": 1.3126, + "step": 1477 + }, + { + "epoch": 0.27, + "learning_rate": 2.9220240262104117e-06, + "loss": 2.7552, + "step": 1478 + }, + { + "epoch": 0.27, + "learning_rate": 2.9219694211867492e-06, + "loss": 0.2805, + "step": 1479 + }, + { + "epoch": 0.27, + "learning_rate": 2.921914816163087e-06, + "loss": 0.9434, + "step": 1480 + }, + { + "epoch": 0.27, + "learning_rate": 2.921860211139425e-06, + "loss": 0.5209, + "step": 1481 + }, + { + "epoch": 0.27, + "learning_rate": 2.9218056061157626e-06, + "loss": 1.5922, + "step": 1482 + }, + { + "epoch": 0.27, + "learning_rate": 2.9217510010921006e-06, + "loss": 1.6175, + "step": 1483 + }, + { + "epoch": 0.27, + "learning_rate": 2.9216963960684385e-06, + "loss": 2.1873, + "step": 1484 + }, + { + "epoch": 0.27, + "learning_rate": 2.921641791044776e-06, + "loss": 1.1779, + "step": 1485 + }, + { + "epoch": 0.27, + "learning_rate": 2.921587186021114e-06, + "loss": 0.705, + "step": 1486 + }, + { + "epoch": 0.27, + "learning_rate": 2.9215325809974515e-06, + "loss": 1.1242, + "step": 1487 + }, + { + "epoch": 0.27, + "learning_rate": 2.9214779759737895e-06, + "loss": 1.7634, + "step": 1488 + }, + { + "epoch": 0.27, + "learning_rate": 2.9214233709501274e-06, + "loss": 2.0044, + "step": 1489 + }, + { + "epoch": 0.27, + "learning_rate": 2.9213687659264653e-06, + "loss": 0.5408, + "step": 1490 + }, + { + "epoch": 0.27, + "learning_rate": 2.9213141609028033e-06, + "loss": 1.6707, + "step": 1491 + }, + { + "epoch": 0.27, + "learning_rate": 2.9212595558791412e-06, + "loss": 0.8415, + "step": 1492 + }, + { + "epoch": 0.27, + "learning_rate": 2.9212049508554788e-06, + "loss": 1.1686, + "step": 1493 + }, + { + "epoch": 0.27, + "learning_rate": 2.9211503458318167e-06, + "loss": 0.0332, + "step": 1494 + }, + { + "epoch": 0.27, + "learning_rate": 2.9210957408081546e-06, + "loss": 1.8146, + "step": 1495 + }, + { + "epoch": 0.27, + "learning_rate": 2.921041135784492e-06, + "loss": 0.373, + "step": 1496 + }, + { + "epoch": 0.27, + "learning_rate": 2.92098653076083e-06, + "loss": 0.1234, + "step": 1497 + }, + { + "epoch": 0.27, + "learning_rate": 2.920931925737168e-06, + "loss": 0.7653, + "step": 1498 + }, + { + "epoch": 0.27, + "learning_rate": 2.9208773207135056e-06, + "loss": 1.6764, + "step": 1499 + }, + { + "epoch": 0.27, + "learning_rate": 2.9208227156898435e-06, + "loss": 1.3132, + "step": 1500 + }, + { + "epoch": 0.27, + "learning_rate": 2.920768110666181e-06, + "loss": 1.2066, + "step": 1501 + }, + { + "epoch": 0.27, + "learning_rate": 2.920713505642519e-06, + "loss": 0.7262, + "step": 1502 + }, + { + "epoch": 0.27, + "learning_rate": 2.920658900618857e-06, + "loss": 0.6512, + "step": 1503 + }, + { + "epoch": 0.27, + "learning_rate": 2.920604295595195e-06, + "loss": 1.2712, + "step": 1504 + }, + { + "epoch": 0.27, + "learning_rate": 2.920549690571533e-06, + "loss": 0.3879, + "step": 1505 + }, + { + "epoch": 0.27, + "learning_rate": 2.9204950855478707e-06, + "loss": 1.1798, + "step": 1506 + }, + { + "epoch": 0.27, + "learning_rate": 2.9204404805242083e-06, + "loss": 0.6186, + "step": 1507 + }, + { + "epoch": 0.27, + "learning_rate": 2.920385875500546e-06, + "loss": 1.6751, + "step": 1508 + }, + { + "epoch": 0.27, + "learning_rate": 2.920331270476884e-06, + "loss": 0.2316, + "step": 1509 + }, + { + "epoch": 0.27, + "learning_rate": 2.9202766654532217e-06, + "loss": 0.044, + "step": 1510 + }, + { + "epoch": 0.27, + "learning_rate": 2.9202220604295596e-06, + "loss": 1.3377, + "step": 1511 + }, + { + "epoch": 0.27, + "learning_rate": 2.9201674554058976e-06, + "loss": 1.2188, + "step": 1512 + }, + { + "epoch": 0.28, + "learning_rate": 2.920112850382235e-06, + "loss": 1.5428, + "step": 1513 + }, + { + "epoch": 0.28, + "learning_rate": 2.920058245358573e-06, + "loss": 3.2367, + "step": 1514 + }, + { + "epoch": 0.28, + "learning_rate": 2.9200036403349105e-06, + "loss": 1.5487, + "step": 1515 + }, + { + "epoch": 0.28, + "learning_rate": 2.9199490353112485e-06, + "loss": 0.7941, + "step": 1516 + }, + { + "epoch": 0.28, + "learning_rate": 2.9198944302875864e-06, + "loss": 0.5506, + "step": 1517 + }, + { + "epoch": 0.28, + "learning_rate": 2.9198398252639244e-06, + "loss": 0.0723, + "step": 1518 + }, + { + "epoch": 0.28, + "learning_rate": 2.9197852202402623e-06, + "loss": 2.5705, + "step": 1519 + }, + { + "epoch": 0.28, + "learning_rate": 2.9197306152166003e-06, + "loss": 3.0957, + "step": 1520 + }, + { + "epoch": 0.28, + "learning_rate": 2.9196760101929378e-06, + "loss": 0.6028, + "step": 1521 + }, + { + "epoch": 0.28, + "learning_rate": 2.9196214051692757e-06, + "loss": 1.4269, + "step": 1522 + }, + { + "epoch": 0.28, + "learning_rate": 2.9195668001456137e-06, + "loss": 3.7945, + "step": 1523 + }, + { + "epoch": 0.28, + "learning_rate": 2.919512195121951e-06, + "loss": 2.2851, + "step": 1524 + }, + { + "epoch": 0.28, + "learning_rate": 2.919457590098289e-06, + "loss": 0.9831, + "step": 1525 + }, + { + "epoch": 0.28, + "learning_rate": 2.9194029850746267e-06, + "loss": 1.1441, + "step": 1526 + }, + { + "epoch": 0.28, + "learning_rate": 2.9193483800509646e-06, + "loss": 0.9796, + "step": 1527 + }, + { + "epoch": 0.28, + "learning_rate": 2.9192937750273025e-06, + "loss": 0.6925, + "step": 1528 + }, + { + "epoch": 0.28, + "learning_rate": 2.91923917000364e-06, + "loss": 0.4293, + "step": 1529 + }, + { + "epoch": 0.28, + "learning_rate": 2.919184564979978e-06, + "loss": 1.5784, + "step": 1530 + }, + { + "epoch": 0.28, + "learning_rate": 2.919129959956316e-06, + "loss": 1.0114, + "step": 1531 + }, + { + "epoch": 0.28, + "learning_rate": 2.919075354932654e-06, + "loss": 0.1568, + "step": 1532 + }, + { + "epoch": 0.28, + "learning_rate": 2.919020749908992e-06, + "loss": 1.0104, + "step": 1533 + }, + { + "epoch": 0.28, + "learning_rate": 2.9189661448853298e-06, + "loss": 2.055, + "step": 1534 + }, + { + "epoch": 0.28, + "learning_rate": 2.9189115398616673e-06, + "loss": 2.0236, + "step": 1535 + }, + { + "epoch": 0.28, + "learning_rate": 2.9188569348380052e-06, + "loss": 1.5123, + "step": 1536 + }, + { + "epoch": 0.28, + "learning_rate": 2.918802329814343e-06, + "loss": 2.2372, + "step": 1537 + }, + { + "epoch": 0.28, + "learning_rate": 2.9187477247906807e-06, + "loss": 2.6597, + "step": 1538 + }, + { + "epoch": 0.28, + "learning_rate": 2.9186931197670186e-06, + "loss": 2.2432, + "step": 1539 + }, + { + "epoch": 0.28, + "learning_rate": 2.918638514743356e-06, + "loss": 2.0237, + "step": 1540 + }, + { + "epoch": 0.28, + "learning_rate": 2.918583909719694e-06, + "loss": 0.3466, + "step": 1541 + }, + { + "epoch": 0.28, + "learning_rate": 2.918529304696032e-06, + "loss": 0.7617, + "step": 1542 + }, + { + "epoch": 0.28, + "learning_rate": 2.91847469967237e-06, + "loss": 0.4355, + "step": 1543 + }, + { + "epoch": 0.28, + "learning_rate": 2.918420094648708e-06, + "loss": 0.2544, + "step": 1544 + }, + { + "epoch": 0.28, + "learning_rate": 2.918365489625046e-06, + "loss": 0.4655, + "step": 1545 + }, + { + "epoch": 0.28, + "learning_rate": 2.9183108846013834e-06, + "loss": 0.2223, + "step": 1546 + }, + { + "epoch": 0.28, + "learning_rate": 2.9182562795777213e-06, + "loss": 0.2841, + "step": 1547 + }, + { + "epoch": 0.28, + "learning_rate": 2.9182016745540593e-06, + "loss": 0.5875, + "step": 1548 + }, + { + "epoch": 0.28, + "learning_rate": 2.918147069530397e-06, + "loss": 1.7523, + "step": 1549 + }, + { + "epoch": 0.28, + "learning_rate": 2.9180924645067348e-06, + "loss": 0.4948, + "step": 1550 + }, + { + "epoch": 0.28, + "learning_rate": 2.9180378594830727e-06, + "loss": 3.6054, + "step": 1551 + }, + { + "epoch": 0.28, + "learning_rate": 2.9179832544594102e-06, + "loss": 1.049, + "step": 1552 + }, + { + "epoch": 0.28, + "learning_rate": 2.917928649435748e-06, + "loss": 2.104, + "step": 1553 + }, + { + "epoch": 0.28, + "learning_rate": 2.9178740444120857e-06, + "loss": 3.0181, + "step": 1554 + }, + { + "epoch": 0.28, + "learning_rate": 2.9178194393884236e-06, + "loss": 1.1148, + "step": 1555 + }, + { + "epoch": 0.28, + "learning_rate": 2.9177648343647616e-06, + "loss": 0.3642, + "step": 1556 + }, + { + "epoch": 0.28, + "learning_rate": 2.9177102293410995e-06, + "loss": 0.8111, + "step": 1557 + }, + { + "epoch": 0.28, + "learning_rate": 2.9176556243174375e-06, + "loss": 1.701, + "step": 1558 + }, + { + "epoch": 0.28, + "learning_rate": 2.9176010192937754e-06, + "loss": 1.119, + "step": 1559 + }, + { + "epoch": 0.28, + "learning_rate": 2.917546414270113e-06, + "loss": 1.4423, + "step": 1560 + }, + { + "epoch": 0.28, + "learning_rate": 2.917491809246451e-06, + "loss": 3.5895, + "step": 1561 + }, + { + "epoch": 0.28, + "learning_rate": 2.917437204222789e-06, + "loss": 0.7004, + "step": 1562 + }, + { + "epoch": 0.28, + "learning_rate": 2.9173825991991263e-06, + "loss": 1.0491, + "step": 1563 + }, + { + "epoch": 0.28, + "learning_rate": 2.9173279941754643e-06, + "loss": 1.3385, + "step": 1564 + }, + { + "epoch": 0.28, + "learning_rate": 2.917273389151802e-06, + "loss": 0.7604, + "step": 1565 + }, + { + "epoch": 0.28, + "learning_rate": 2.9172187841281397e-06, + "loss": 1.4884, + "step": 1566 + }, + { + "epoch": 0.28, + "learning_rate": 2.9171641791044777e-06, + "loss": 1.2806, + "step": 1567 + }, + { + "epoch": 0.29, + "learning_rate": 2.917109574080815e-06, + "loss": 2.4462, + "step": 1568 + }, + { + "epoch": 0.29, + "learning_rate": 2.917054969057153e-06, + "loss": 1.7043, + "step": 1569 + }, + { + "epoch": 0.29, + "learning_rate": 2.917000364033491e-06, + "loss": 0.1511, + "step": 1570 + }, + { + "epoch": 0.29, + "learning_rate": 2.916945759009829e-06, + "loss": 0.8761, + "step": 1571 + }, + { + "epoch": 0.29, + "learning_rate": 2.916891153986167e-06, + "loss": 1.1052, + "step": 1572 + }, + { + "epoch": 0.29, + "learning_rate": 2.916836548962505e-06, + "loss": 1.0871, + "step": 1573 + }, + { + "epoch": 0.29, + "learning_rate": 2.9167819439388424e-06, + "loss": 0.9687, + "step": 1574 + }, + { + "epoch": 0.29, + "learning_rate": 2.9167273389151804e-06, + "loss": 0.676, + "step": 1575 + }, + { + "epoch": 0.29, + "learning_rate": 2.9166727338915183e-06, + "loss": 0.4246, + "step": 1576 + }, + { + "epoch": 0.29, + "learning_rate": 2.916618128867856e-06, + "loss": 1.8398, + "step": 1577 + }, + { + "epoch": 0.29, + "learning_rate": 2.9165635238441938e-06, + "loss": 1.1237, + "step": 1578 + }, + { + "epoch": 0.29, + "learning_rate": 2.9165089188205317e-06, + "loss": 0.9754, + "step": 1579 + }, + { + "epoch": 0.29, + "learning_rate": 2.9164543137968692e-06, + "loss": 0.1543, + "step": 1580 + }, + { + "epoch": 0.29, + "learning_rate": 2.916399708773207e-06, + "loss": 1.906, + "step": 1581 + }, + { + "epoch": 0.29, + "learning_rate": 2.9163451037495447e-06, + "loss": 0.4931, + "step": 1582 + }, + { + "epoch": 0.29, + "learning_rate": 2.9162904987258826e-06, + "loss": 1.1706, + "step": 1583 + }, + { + "epoch": 0.29, + "learning_rate": 2.9162358937022206e-06, + "loss": 0.7307, + "step": 1584 + }, + { + "epoch": 0.29, + "learning_rate": 2.9161812886785585e-06, + "loss": 0.6397, + "step": 1585 + }, + { + "epoch": 0.29, + "learning_rate": 2.9161266836548965e-06, + "loss": 2.215, + "step": 1586 + }, + { + "epoch": 0.29, + "learning_rate": 2.9160720786312344e-06, + "loss": 1.5076, + "step": 1587 + }, + { + "epoch": 0.29, + "learning_rate": 2.916017473607572e-06, + "loss": 1.0941, + "step": 1588 + }, + { + "epoch": 0.29, + "learning_rate": 2.91596286858391e-06, + "loss": 2.0258, + "step": 1589 + }, + { + "epoch": 0.29, + "learning_rate": 2.915908263560248e-06, + "loss": 1.4113, + "step": 1590 + }, + { + "epoch": 0.29, + "learning_rate": 2.9158536585365853e-06, + "loss": 1.6275, + "step": 1591 + }, + { + "epoch": 0.29, + "learning_rate": 2.9157990535129233e-06, + "loss": 0.5769, + "step": 1592 + }, + { + "epoch": 0.29, + "learning_rate": 2.9157444484892612e-06, + "loss": 1.2418, + "step": 1593 + }, + { + "epoch": 0.29, + "learning_rate": 2.9156898434655988e-06, + "loss": 1.8259, + "step": 1594 + }, + { + "epoch": 0.29, + "learning_rate": 2.9156352384419367e-06, + "loss": 0.0606, + "step": 1595 + }, + { + "epoch": 0.29, + "learning_rate": 2.9155806334182742e-06, + "loss": 0.6411, + "step": 1596 + }, + { + "epoch": 0.29, + "learning_rate": 2.915526028394612e-06, + "loss": 0.4673, + "step": 1597 + }, + { + "epoch": 0.29, + "learning_rate": 2.91547142337095e-06, + "loss": 1.4265, + "step": 1598 + }, + { + "epoch": 0.29, + "learning_rate": 2.915416818347288e-06, + "loss": 0.3727, + "step": 1599 + }, + { + "epoch": 0.29, + "learning_rate": 2.915362213323626e-06, + "loss": 0.9312, + "step": 1600 + }, + { + "epoch": 0.29, + "learning_rate": 2.915307608299964e-06, + "loss": 1.6291, + "step": 1601 + }, + { + "epoch": 0.29, + "learning_rate": 2.9152530032763015e-06, + "loss": 0.5584, + "step": 1602 + }, + { + "epoch": 0.29, + "learning_rate": 2.9151983982526394e-06, + "loss": 1.2906, + "step": 1603 + }, + { + "epoch": 0.29, + "learning_rate": 2.9151437932289773e-06, + "loss": 0.7008, + "step": 1604 + }, + { + "epoch": 0.29, + "learning_rate": 2.915089188205315e-06, + "loss": 0.3528, + "step": 1605 + }, + { + "epoch": 0.29, + "learning_rate": 2.915034583181653e-06, + "loss": 0.6045, + "step": 1606 + }, + { + "epoch": 0.29, + "learning_rate": 2.9149799781579903e-06, + "loss": 1.3414, + "step": 1607 + }, + { + "epoch": 0.29, + "learning_rate": 2.9149253731343283e-06, + "loss": 0.4171, + "step": 1608 + }, + { + "epoch": 0.29, + "learning_rate": 2.9148707681106662e-06, + "loss": 1.7234, + "step": 1609 + }, + { + "epoch": 0.29, + "learning_rate": 2.9148161630870037e-06, + "loss": 0.6242, + "step": 1610 + }, + { + "epoch": 0.29, + "learning_rate": 2.9147615580633417e-06, + "loss": 0.9916, + "step": 1611 + }, + { + "epoch": 0.29, + "learning_rate": 2.9147069530396796e-06, + "loss": 0.8707, + "step": 1612 + }, + { + "epoch": 0.29, + "learning_rate": 2.9146523480160176e-06, + "loss": 0.9862, + "step": 1613 + }, + { + "epoch": 0.29, + "learning_rate": 2.9145977429923555e-06, + "loss": 0.5019, + "step": 1614 + }, + { + "epoch": 0.29, + "learning_rate": 2.9145431379686934e-06, + "loss": 0.8143, + "step": 1615 + }, + { + "epoch": 0.29, + "learning_rate": 2.914488532945031e-06, + "loss": 2.7915, + "step": 1616 + }, + { + "epoch": 0.29, + "learning_rate": 2.914433927921369e-06, + "loss": 1.1111, + "step": 1617 + }, + { + "epoch": 0.29, + "learning_rate": 2.914379322897707e-06, + "loss": 2.8096, + "step": 1618 + }, + { + "epoch": 0.29, + "learning_rate": 2.9143247178740444e-06, + "loss": 2.1763, + "step": 1619 + }, + { + "epoch": 0.29, + "learning_rate": 2.9142701128503823e-06, + "loss": 0.4747, + "step": 1620 + }, + { + "epoch": 0.29, + "learning_rate": 2.91421550782672e-06, + "loss": 1.7641, + "step": 1621 + }, + { + "epoch": 0.29, + "learning_rate": 2.9141609028030578e-06, + "loss": 2.1462, + "step": 1622 + }, + { + "epoch": 0.3, + "learning_rate": 2.9141062977793957e-06, + "loss": 0.3546, + "step": 1623 + }, + { + "epoch": 0.3, + "learning_rate": 2.9140516927557337e-06, + "loss": 1.2307, + "step": 1624 + }, + { + "epoch": 0.3, + "learning_rate": 2.9139970877320716e-06, + "loss": 0.5478, + "step": 1625 + }, + { + "epoch": 0.3, + "learning_rate": 2.9139424827084096e-06, + "loss": 2.5413, + "step": 1626 + }, + { + "epoch": 0.3, + "learning_rate": 2.913887877684747e-06, + "loss": 0.6674, + "step": 1627 + }, + { + "epoch": 0.3, + "learning_rate": 2.913833272661085e-06, + "loss": 1.5513, + "step": 1628 + }, + { + "epoch": 0.3, + "learning_rate": 2.913778667637423e-06, + "loss": 1.0803, + "step": 1629 + }, + { + "epoch": 0.3, + "learning_rate": 2.9137240626137605e-06, + "loss": 0.9788, + "step": 1630 + }, + { + "epoch": 0.3, + "learning_rate": 2.9136694575900984e-06, + "loss": 1.3583, + "step": 1631 + }, + { + "epoch": 0.3, + "learning_rate": 2.9136148525664364e-06, + "loss": 2.0988, + "step": 1632 + }, + { + "epoch": 0.3, + "learning_rate": 2.913560247542774e-06, + "loss": 0.3499, + "step": 1633 + }, + { + "epoch": 0.3, + "learning_rate": 2.913505642519112e-06, + "loss": 1.0167, + "step": 1634 + }, + { + "epoch": 0.3, + "learning_rate": 2.9134510374954494e-06, + "loss": 0.1925, + "step": 1635 + }, + { + "epoch": 0.3, + "learning_rate": 2.9133964324717873e-06, + "loss": 1.1294, + "step": 1636 + }, + { + "epoch": 0.3, + "learning_rate": 2.9133418274481252e-06, + "loss": 0.5266, + "step": 1637 + }, + { + "epoch": 0.3, + "learning_rate": 2.913287222424463e-06, + "loss": 0.7766, + "step": 1638 + }, + { + "epoch": 0.3, + "learning_rate": 2.913232617400801e-06, + "loss": 1.4648, + "step": 1639 + }, + { + "epoch": 0.3, + "learning_rate": 2.913178012377139e-06, + "loss": 0.0673, + "step": 1640 + }, + { + "epoch": 0.3, + "learning_rate": 2.9131234073534766e-06, + "loss": 0.71, + "step": 1641 + }, + { + "epoch": 0.3, + "learning_rate": 2.9130688023298145e-06, + "loss": 1.7849, + "step": 1642 + }, + { + "epoch": 0.3, + "learning_rate": 2.9130141973061525e-06, + "loss": 0.4654, + "step": 1643 + }, + { + "epoch": 0.3, + "learning_rate": 2.91295959228249e-06, + "loss": 0.7691, + "step": 1644 + }, + { + "epoch": 0.3, + "learning_rate": 2.912904987258828e-06, + "loss": 0.8028, + "step": 1645 + }, + { + "epoch": 0.3, + "learning_rate": 2.912850382235166e-06, + "loss": 0.145, + "step": 1646 + }, + { + "epoch": 0.3, + "learning_rate": 2.9127957772115034e-06, + "loss": 2.1003, + "step": 1647 + }, + { + "epoch": 0.3, + "learning_rate": 2.9127411721878413e-06, + "loss": 0.8596, + "step": 1648 + }, + { + "epoch": 0.3, + "learning_rate": 2.912686567164179e-06, + "loss": 3.0909, + "step": 1649 + }, + { + "epoch": 0.3, + "learning_rate": 2.912631962140517e-06, + "loss": 0.5157, + "step": 1650 + }, + { + "epoch": 0.3, + "learning_rate": 2.9125773571168548e-06, + "loss": 1.1401, + "step": 1651 + }, + { + "epoch": 0.3, + "learning_rate": 2.9125227520931927e-06, + "loss": 1.9643, + "step": 1652 + }, + { + "epoch": 0.3, + "learning_rate": 2.9124681470695306e-06, + "loss": 1.182, + "step": 1653 + }, + { + "epoch": 0.3, + "learning_rate": 2.9124135420458686e-06, + "loss": 3.4458, + "step": 1654 + }, + { + "epoch": 0.3, + "learning_rate": 2.912358937022206e-06, + "loss": 0.447, + "step": 1655 + }, + { + "epoch": 0.3, + "learning_rate": 2.912304331998544e-06, + "loss": 1.2151, + "step": 1656 + }, + { + "epoch": 0.3, + "learning_rate": 2.912249726974882e-06, + "loss": 0.5502, + "step": 1657 + }, + { + "epoch": 0.3, + "learning_rate": 2.9121951219512195e-06, + "loss": 0.6024, + "step": 1658 + }, + { + "epoch": 0.3, + "learning_rate": 2.9121405169275575e-06, + "loss": 0.3443, + "step": 1659 + }, + { + "epoch": 0.3, + "learning_rate": 2.9120859119038954e-06, + "loss": 1.3934, + "step": 1660 + }, + { + "epoch": 0.3, + "learning_rate": 2.912031306880233e-06, + "loss": 1.7919, + "step": 1661 + }, + { + "epoch": 0.3, + "learning_rate": 2.911976701856571e-06, + "loss": 0.3023, + "step": 1662 + }, + { + "epoch": 0.3, + "learning_rate": 2.9119220968329084e-06, + "loss": 0.0883, + "step": 1663 + }, + { + "epoch": 0.3, + "learning_rate": 2.9118674918092463e-06, + "loss": 0.6973, + "step": 1664 + }, + { + "epoch": 0.3, + "learning_rate": 2.9118128867855843e-06, + "loss": 1.3216, + "step": 1665 + }, + { + "epoch": 0.3, + "learning_rate": 2.911758281761922e-06, + "loss": 1.1521, + "step": 1666 + }, + { + "epoch": 0.3, + "learning_rate": 2.91170367673826e-06, + "loss": 0.9045, + "step": 1667 + }, + { + "epoch": 0.3, + "learning_rate": 2.911649071714598e-06, + "loss": 1.0403, + "step": 1668 + }, + { + "epoch": 0.3, + "learning_rate": 2.9115944666909356e-06, + "loss": 0.2692, + "step": 1669 + }, + { + "epoch": 0.3, + "learning_rate": 2.9115398616672736e-06, + "loss": 1.8504, + "step": 1670 + }, + { + "epoch": 0.3, + "learning_rate": 2.9114852566436115e-06, + "loss": 0.6759, + "step": 1671 + }, + { + "epoch": 0.3, + "learning_rate": 2.911430651619949e-06, + "loss": 1.9942, + "step": 1672 + }, + { + "epoch": 0.3, + "learning_rate": 2.911376046596287e-06, + "loss": 1.763, + "step": 1673 + }, + { + "epoch": 0.3, + "learning_rate": 2.911321441572625e-06, + "loss": 1.4443, + "step": 1674 + }, + { + "epoch": 0.3, + "learning_rate": 2.9112668365489624e-06, + "loss": 1.4278, + "step": 1675 + }, + { + "epoch": 0.3, + "learning_rate": 2.9112122315253004e-06, + "loss": 0.875, + "step": 1676 + }, + { + "epoch": 0.3, + "learning_rate": 2.911157626501638e-06, + "loss": 1.6395, + "step": 1677 + }, + { + "epoch": 0.31, + "learning_rate": 2.911103021477976e-06, + "loss": 1.1771, + "step": 1678 + }, + { + "epoch": 0.31, + "learning_rate": 2.9110484164543138e-06, + "loss": 1.1428, + "step": 1679 + }, + { + "epoch": 0.31, + "learning_rate": 2.9109938114306517e-06, + "loss": 1.5366, + "step": 1680 + }, + { + "epoch": 0.31, + "learning_rate": 2.9109392064069897e-06, + "loss": 0.1671, + "step": 1681 + }, + { + "epoch": 0.31, + "learning_rate": 2.9108846013833276e-06, + "loss": 1.8628, + "step": 1682 + }, + { + "epoch": 0.31, + "learning_rate": 2.910829996359665e-06, + "loss": 0.6863, + "step": 1683 + }, + { + "epoch": 0.31, + "learning_rate": 2.910775391336003e-06, + "loss": 0.3972, + "step": 1684 + }, + { + "epoch": 0.31, + "learning_rate": 2.910720786312341e-06, + "loss": 1.3162, + "step": 1685 + }, + { + "epoch": 0.31, + "learning_rate": 2.9106661812886785e-06, + "loss": 0.088, + "step": 1686 + }, + { + "epoch": 0.31, + "learning_rate": 2.9106115762650165e-06, + "loss": 1.9242, + "step": 1687 + }, + { + "epoch": 0.31, + "learning_rate": 2.910556971241354e-06, + "loss": 0.998, + "step": 1688 + }, + { + "epoch": 0.31, + "learning_rate": 2.910502366217692e-06, + "loss": 0.0777, + "step": 1689 + }, + { + "epoch": 0.31, + "learning_rate": 2.91044776119403e-06, + "loss": 0.5002, + "step": 1690 + }, + { + "epoch": 0.31, + "learning_rate": 2.9103931561703674e-06, + "loss": 3.051, + "step": 1691 + }, + { + "epoch": 0.31, + "learning_rate": 2.9103385511467053e-06, + "loss": 0.0706, + "step": 1692 + }, + { + "epoch": 0.31, + "learning_rate": 2.9102839461230433e-06, + "loss": 1.2268, + "step": 1693 + }, + { + "epoch": 0.31, + "learning_rate": 2.9102293410993812e-06, + "loss": 1.0772, + "step": 1694 + }, + { + "epoch": 0.31, + "learning_rate": 2.910174736075719e-06, + "loss": 3.1288, + "step": 1695 + }, + { + "epoch": 0.31, + "learning_rate": 2.910120131052057e-06, + "loss": 1.257, + "step": 1696 + }, + { + "epoch": 0.31, + "learning_rate": 2.9100655260283946e-06, + "loss": 1.1935, + "step": 1697 + }, + { + "epoch": 0.31, + "learning_rate": 2.9100109210047326e-06, + "loss": 1.1442, + "step": 1698 + }, + { + "epoch": 0.31, + "learning_rate": 2.9099563159810705e-06, + "loss": 0.0491, + "step": 1699 + }, + { + "epoch": 0.31, + "learning_rate": 2.909901710957408e-06, + "loss": 1.4033, + "step": 1700 + }, + { + "epoch": 0.31, + "learning_rate": 2.909847105933746e-06, + "loss": 0.3275, + "step": 1701 + }, + { + "epoch": 0.31, + "learning_rate": 2.9097925009100835e-06, + "loss": 0.6118, + "step": 1702 + }, + { + "epoch": 0.31, + "learning_rate": 2.9097378958864215e-06, + "loss": 0.1293, + "step": 1703 + }, + { + "epoch": 0.31, + "learning_rate": 2.9096832908627594e-06, + "loss": 0.8841, + "step": 1704 + }, + { + "epoch": 0.31, + "learning_rate": 2.9096286858390973e-06, + "loss": 2.0731, + "step": 1705 + }, + { + "epoch": 0.31, + "learning_rate": 2.9095740808154353e-06, + "loss": 2.7859, + "step": 1706 + }, + { + "epoch": 0.31, + "learning_rate": 2.9095194757917732e-06, + "loss": 0.9235, + "step": 1707 + }, + { + "epoch": 0.31, + "learning_rate": 2.9094648707681108e-06, + "loss": 0.5507, + "step": 1708 + }, + { + "epoch": 0.31, + "learning_rate": 2.9094102657444487e-06, + "loss": 0.6069, + "step": 1709 + }, + { + "epoch": 0.31, + "learning_rate": 2.9093556607207866e-06, + "loss": 1.0493, + "step": 1710 + }, + { + "epoch": 0.31, + "learning_rate": 2.909301055697124e-06, + "loss": 0.2547, + "step": 1711 + }, + { + "epoch": 0.31, + "learning_rate": 2.909246450673462e-06, + "loss": 0.7407, + "step": 1712 + }, + { + "epoch": 0.31, + "learning_rate": 2.9091918456498e-06, + "loss": 0.6542, + "step": 1713 + }, + { + "epoch": 0.31, + "learning_rate": 2.9091372406261376e-06, + "loss": 1.2389, + "step": 1714 + }, + { + "epoch": 0.31, + "learning_rate": 2.9090826356024755e-06, + "loss": 0.4702, + "step": 1715 + }, + { + "epoch": 0.31, + "learning_rate": 2.909028030578813e-06, + "loss": 1.0851, + "step": 1716 + }, + { + "epoch": 0.31, + "learning_rate": 2.908973425555151e-06, + "loss": 1.1223, + "step": 1717 + }, + { + "epoch": 0.31, + "learning_rate": 2.908918820531489e-06, + "loss": 2.255, + "step": 1718 + }, + { + "epoch": 0.31, + "learning_rate": 2.908864215507827e-06, + "loss": 3.3151, + "step": 1719 + }, + { + "epoch": 0.31, + "learning_rate": 2.908809610484165e-06, + "loss": 0.7115, + "step": 1720 + }, + { + "epoch": 0.31, + "learning_rate": 2.9087550054605027e-06, + "loss": 2.1222, + "step": 1721 + }, + { + "epoch": 0.31, + "learning_rate": 2.9087004004368403e-06, + "loss": 1.1819, + "step": 1722 + }, + { + "epoch": 0.31, + "learning_rate": 2.908645795413178e-06, + "loss": 1.182, + "step": 1723 + }, + { + "epoch": 0.31, + "learning_rate": 2.908591190389516e-06, + "loss": 0.081, + "step": 1724 + }, + { + "epoch": 0.31, + "learning_rate": 2.9085365853658537e-06, + "loss": 2.6635, + "step": 1725 + }, + { + "epoch": 0.31, + "learning_rate": 2.9084819803421916e-06, + "loss": 0.3939, + "step": 1726 + }, + { + "epoch": 0.31, + "learning_rate": 2.9084273753185296e-06, + "loss": 0.1801, + "step": 1727 + }, + { + "epoch": 0.31, + "learning_rate": 2.908372770294867e-06, + "loss": 1.1028, + "step": 1728 + }, + { + "epoch": 0.31, + "learning_rate": 2.908318165271205e-06, + "loss": 0.325, + "step": 1729 + }, + { + "epoch": 0.31, + "learning_rate": 2.9082635602475425e-06, + "loss": 2.0689, + "step": 1730 + }, + { + "epoch": 0.31, + "learning_rate": 2.9082089552238805e-06, + "loss": 1.3983, + "step": 1731 + }, + { + "epoch": 0.31, + "learning_rate": 2.9081543502002184e-06, + "loss": 2.8431, + "step": 1732 + }, + { + "epoch": 0.32, + "learning_rate": 2.9080997451765564e-06, + "loss": 0.7123, + "step": 1733 + }, + { + "epoch": 0.32, + "learning_rate": 2.9080451401528943e-06, + "loss": 1.9354, + "step": 1734 + }, + { + "epoch": 0.32, + "learning_rate": 2.9079905351292323e-06, + "loss": 0.4606, + "step": 1735 + }, + { + "epoch": 0.32, + "learning_rate": 2.9079359301055698e-06, + "loss": 1.2375, + "step": 1736 + }, + { + "epoch": 0.32, + "learning_rate": 2.9078813250819077e-06, + "loss": 1.8524, + "step": 1737 + }, + { + "epoch": 0.32, + "learning_rate": 2.9078267200582457e-06, + "loss": 1.4485, + "step": 1738 + }, + { + "epoch": 0.32, + "learning_rate": 2.907772115034583e-06, + "loss": 1.503, + "step": 1739 + }, + { + "epoch": 0.32, + "learning_rate": 2.907717510010921e-06, + "loss": 1.4059, + "step": 1740 + }, + { + "epoch": 0.32, + "learning_rate": 2.907662904987259e-06, + "loss": 0.5222, + "step": 1741 + }, + { + "epoch": 0.32, + "learning_rate": 2.9076082999635966e-06, + "loss": 0.7711, + "step": 1742 + }, + { + "epoch": 0.32, + "learning_rate": 2.9075536949399345e-06, + "loss": 1.2874, + "step": 1743 + }, + { + "epoch": 0.32, + "learning_rate": 2.907499089916272e-06, + "loss": 0.6294, + "step": 1744 + }, + { + "epoch": 0.32, + "learning_rate": 2.90744448489261e-06, + "loss": 0.6907, + "step": 1745 + }, + { + "epoch": 0.32, + "learning_rate": 2.907389879868948e-06, + "loss": 0.873, + "step": 1746 + }, + { + "epoch": 0.32, + "learning_rate": 2.907335274845286e-06, + "loss": 2.7042, + "step": 1747 + }, + { + "epoch": 0.32, + "learning_rate": 2.907280669821624e-06, + "loss": 1.6541, + "step": 1748 + }, + { + "epoch": 0.32, + "learning_rate": 2.9072260647979618e-06, + "loss": 1.0285, + "step": 1749 + }, + { + "epoch": 0.32, + "learning_rate": 2.9071714597742993e-06, + "loss": 0.4571, + "step": 1750 + }, + { + "epoch": 0.32, + "learning_rate": 2.9071168547506372e-06, + "loss": 2.0454, + "step": 1751 + }, + { + "epoch": 0.32, + "learning_rate": 2.907062249726975e-06, + "loss": 0.099, + "step": 1752 + }, + { + "epoch": 0.32, + "learning_rate": 2.9070076447033127e-06, + "loss": 1.3116, + "step": 1753 + }, + { + "epoch": 0.32, + "learning_rate": 2.9069530396796506e-06, + "loss": 0.2154, + "step": 1754 + }, + { + "epoch": 0.32, + "learning_rate": 2.9068984346559886e-06, + "loss": 1.199, + "step": 1755 + }, + { + "epoch": 0.32, + "learning_rate": 2.906843829632326e-06, + "loss": 0.468, + "step": 1756 + }, + { + "epoch": 0.32, + "learning_rate": 2.906789224608664e-06, + "loss": 2.6957, + "step": 1757 + }, + { + "epoch": 0.32, + "learning_rate": 2.9067346195850016e-06, + "loss": 0.5804, + "step": 1758 + }, + { + "epoch": 0.32, + "learning_rate": 2.9066800145613395e-06, + "loss": 0.5715, + "step": 1759 + }, + { + "epoch": 0.32, + "learning_rate": 2.9066254095376775e-06, + "loss": 1.9492, + "step": 1760 + }, + { + "epoch": 0.32, + "learning_rate": 2.9065708045140154e-06, + "loss": 0.7622, + "step": 1761 + }, + { + "epoch": 0.32, + "learning_rate": 2.9065161994903533e-06, + "loss": 2.1156, + "step": 1762 + }, + { + "epoch": 0.32, + "learning_rate": 2.9064615944666913e-06, + "loss": 0.9669, + "step": 1763 + }, + { + "epoch": 0.32, + "learning_rate": 2.906406989443029e-06, + "loss": 0.6195, + "step": 1764 + }, + { + "epoch": 0.32, + "learning_rate": 2.9063523844193667e-06, + "loss": 0.6847, + "step": 1765 + }, + { + "epoch": 0.32, + "learning_rate": 2.9062977793957047e-06, + "loss": 1.2418, + "step": 1766 + }, + { + "epoch": 0.32, + "learning_rate": 2.906243174372042e-06, + "loss": 1.3702, + "step": 1767 + }, + { + "epoch": 0.32, + "learning_rate": 2.90618856934838e-06, + "loss": 1.5874, + "step": 1768 + }, + { + "epoch": 0.32, + "learning_rate": 2.9061339643247177e-06, + "loss": 2.0752, + "step": 1769 + }, + { + "epoch": 0.32, + "learning_rate": 2.9060793593010556e-06, + "loss": 0.7608, + "step": 1770 + }, + { + "epoch": 0.32, + "learning_rate": 2.9060247542773936e-06, + "loss": 1.7152, + "step": 1771 + }, + { + "epoch": 0.32, + "learning_rate": 2.905970149253731e-06, + "loss": 2.0451, + "step": 1772 + }, + { + "epoch": 0.32, + "learning_rate": 2.905915544230069e-06, + "loss": 0.1953, + "step": 1773 + }, + { + "epoch": 0.32, + "learning_rate": 2.905860939206407e-06, + "loss": 0.181, + "step": 1774 + }, + { + "epoch": 0.32, + "learning_rate": 2.905806334182745e-06, + "loss": 0.5756, + "step": 1775 + }, + { + "epoch": 0.32, + "learning_rate": 2.905751729159083e-06, + "loss": 2.4955, + "step": 1776 + }, + { + "epoch": 0.32, + "learning_rate": 2.905697124135421e-06, + "loss": 0.0856, + "step": 1777 + }, + { + "epoch": 0.32, + "learning_rate": 2.9056425191117583e-06, + "loss": 0.4219, + "step": 1778 + }, + { + "epoch": 0.32, + "learning_rate": 2.9055879140880963e-06, + "loss": 1.548, + "step": 1779 + }, + { + "epoch": 0.32, + "learning_rate": 2.905533309064434e-06, + "loss": 0.2599, + "step": 1780 + }, + { + "epoch": 0.32, + "learning_rate": 2.9054787040407717e-06, + "loss": 0.0841, + "step": 1781 + }, + { + "epoch": 0.32, + "learning_rate": 2.9054240990171097e-06, + "loss": 1.6274, + "step": 1782 + }, + { + "epoch": 0.32, + "learning_rate": 2.905369493993447e-06, + "loss": 2.7823, + "step": 1783 + }, + { + "epoch": 0.32, + "learning_rate": 2.905314888969785e-06, + "loss": 0.3986, + "step": 1784 + }, + { + "epoch": 0.32, + "learning_rate": 2.905260283946123e-06, + "loss": 0.1628, + "step": 1785 + }, + { + "epoch": 0.32, + "learning_rate": 2.905205678922461e-06, + "loss": 2.1276, + "step": 1786 + }, + { + "epoch": 0.32, + "learning_rate": 2.9051510738987985e-06, + "loss": 0.974, + "step": 1787 + }, + { + "epoch": 0.33, + "learning_rate": 2.9050964688751365e-06, + "loss": 0.697, + "step": 1788 + }, + { + "epoch": 0.33, + "learning_rate": 2.9050418638514744e-06, + "loss": 0.1466, + "step": 1789 + }, + { + "epoch": 0.33, + "learning_rate": 2.9049872588278124e-06, + "loss": 0.7832, + "step": 1790 + }, + { + "epoch": 0.33, + "learning_rate": 2.9049326538041503e-06, + "loss": 0.5952, + "step": 1791 + }, + { + "epoch": 0.33, + "learning_rate": 2.904878048780488e-06, + "loss": 0.0385, + "step": 1792 + }, + { + "epoch": 0.33, + "learning_rate": 2.9048234437568258e-06, + "loss": 0.0552, + "step": 1793 + }, + { + "epoch": 0.33, + "learning_rate": 2.9047688387331637e-06, + "loss": 0.9923, + "step": 1794 + }, + { + "epoch": 0.33, + "learning_rate": 2.9047142337095012e-06, + "loss": 1.1078, + "step": 1795 + }, + { + "epoch": 0.33, + "learning_rate": 2.904659628685839e-06, + "loss": 0.9645, + "step": 1796 + }, + { + "epoch": 0.33, + "learning_rate": 2.9046050236621767e-06, + "loss": 0.377, + "step": 1797 + }, + { + "epoch": 0.33, + "learning_rate": 2.9045504186385146e-06, + "loss": 0.7546, + "step": 1798 + }, + { + "epoch": 0.33, + "learning_rate": 2.9044958136148526e-06, + "loss": 1.3446, + "step": 1799 + }, + { + "epoch": 0.33, + "learning_rate": 2.9044412085911905e-06, + "loss": 0.8333, + "step": 1800 + }, + { + "epoch": 0.33, + "learning_rate": 2.9043866035675285e-06, + "loss": 2.7683, + "step": 1801 + }, + { + "epoch": 0.33, + "learning_rate": 2.9043319985438664e-06, + "loss": 0.9302, + "step": 1802 + }, + { + "epoch": 0.33, + "learning_rate": 2.904277393520204e-06, + "loss": 1.4543, + "step": 1803 + }, + { + "epoch": 0.33, + "learning_rate": 2.904222788496542e-06, + "loss": 0.1756, + "step": 1804 + }, + { + "epoch": 0.33, + "learning_rate": 2.90416818347288e-06, + "loss": 3.4352, + "step": 1805 + }, + { + "epoch": 0.33, + "learning_rate": 2.9041135784492173e-06, + "loss": 0.942, + "step": 1806 + }, + { + "epoch": 0.33, + "learning_rate": 2.9040589734255553e-06, + "loss": 0.3109, + "step": 1807 + }, + { + "epoch": 0.33, + "learning_rate": 2.9040043684018932e-06, + "loss": 0.352, + "step": 1808 + }, + { + "epoch": 0.33, + "learning_rate": 2.9039497633782308e-06, + "loss": 0.793, + "step": 1809 + }, + { + "epoch": 0.33, + "learning_rate": 2.9038951583545687e-06, + "loss": 2.0683, + "step": 1810 + }, + { + "epoch": 0.33, + "learning_rate": 2.9038405533309062e-06, + "loss": 1.0378, + "step": 1811 + }, + { + "epoch": 0.33, + "learning_rate": 2.903785948307244e-06, + "loss": 2.1021, + "step": 1812 + }, + { + "epoch": 0.33, + "learning_rate": 2.903731343283582e-06, + "loss": 0.1198, + "step": 1813 + }, + { + "epoch": 0.33, + "learning_rate": 2.90367673825992e-06, + "loss": 1.1034, + "step": 1814 + }, + { + "epoch": 0.33, + "learning_rate": 2.903622133236258e-06, + "loss": 1.08, + "step": 1815 + }, + { + "epoch": 0.33, + "learning_rate": 2.903567528212596e-06, + "loss": 2.1451, + "step": 1816 + }, + { + "epoch": 0.33, + "learning_rate": 2.9035129231889335e-06, + "loss": 0.0204, + "step": 1817 + }, + { + "epoch": 0.33, + "learning_rate": 2.9034583181652714e-06, + "loss": 0.2051, + "step": 1818 + }, + { + "epoch": 0.33, + "learning_rate": 2.9034037131416093e-06, + "loss": 1.0558, + "step": 1819 + }, + { + "epoch": 0.33, + "learning_rate": 2.903349108117947e-06, + "loss": 2.2962, + "step": 1820 + }, + { + "epoch": 0.33, + "learning_rate": 2.903294503094285e-06, + "loss": 1.0924, + "step": 1821 + }, + { + "epoch": 0.33, + "learning_rate": 2.9032398980706227e-06, + "loss": 0.2912, + "step": 1822 + }, + { + "epoch": 0.33, + "learning_rate": 2.9031852930469603e-06, + "loss": 1.5393, + "step": 1823 + }, + { + "epoch": 0.33, + "learning_rate": 2.903130688023298e-06, + "loss": 0.6878, + "step": 1824 + }, + { + "epoch": 0.33, + "learning_rate": 2.9030760829996357e-06, + "loss": 2.8386, + "step": 1825 + }, + { + "epoch": 0.33, + "learning_rate": 2.9030214779759737e-06, + "loss": 0.6352, + "step": 1826 + }, + { + "epoch": 0.33, + "learning_rate": 2.9029668729523116e-06, + "loss": 3.4241, + "step": 1827 + }, + { + "epoch": 0.33, + "learning_rate": 2.9029122679286496e-06, + "loss": 1.6185, + "step": 1828 + }, + { + "epoch": 0.33, + "learning_rate": 2.9028576629049875e-06, + "loss": 1.8201, + "step": 1829 + }, + { + "epoch": 0.33, + "learning_rate": 2.9028030578813254e-06, + "loss": 1.5519, + "step": 1830 + }, + { + "epoch": 0.33, + "learning_rate": 2.902748452857663e-06, + "loss": 2.0428, + "step": 1831 + }, + { + "epoch": 0.33, + "learning_rate": 2.902693847834001e-06, + "loss": 1.2048, + "step": 1832 + }, + { + "epoch": 0.33, + "learning_rate": 2.902639242810339e-06, + "loss": 0.4, + "step": 1833 + }, + { + "epoch": 0.33, + "learning_rate": 2.9025846377866764e-06, + "loss": 0.3581, + "step": 1834 + }, + { + "epoch": 0.33, + "learning_rate": 2.9025300327630143e-06, + "loss": 1.9731, + "step": 1835 + }, + { + "epoch": 0.33, + "learning_rate": 2.902475427739352e-06, + "loss": 0.3415, + "step": 1836 + }, + { + "epoch": 0.33, + "learning_rate": 2.9024208227156898e-06, + "loss": 0.5071, + "step": 1837 + }, + { + "epoch": 0.33, + "learning_rate": 2.9023662176920277e-06, + "loss": 1.0927, + "step": 1838 + }, + { + "epoch": 0.33, + "learning_rate": 2.9023116126683652e-06, + "loss": 1.4217, + "step": 1839 + }, + { + "epoch": 0.33, + "learning_rate": 2.902257007644703e-06, + "loss": 1.3002, + "step": 1840 + }, + { + "epoch": 0.33, + "learning_rate": 2.902202402621041e-06, + "loss": 0.5757, + "step": 1841 + }, + { + "epoch": 0.33, + "learning_rate": 2.902147797597379e-06, + "loss": 0.6986, + "step": 1842 + }, + { + "epoch": 0.34, + "learning_rate": 2.902093192573717e-06, + "loss": 2.2428, + "step": 1843 + }, + { + "epoch": 0.34, + "learning_rate": 2.902038587550055e-06, + "loss": 0.8958, + "step": 1844 + }, + { + "epoch": 0.34, + "learning_rate": 2.9019839825263925e-06, + "loss": 1.738, + "step": 1845 + }, + { + "epoch": 0.34, + "learning_rate": 2.9019293775027304e-06, + "loss": 2.1254, + "step": 1846 + }, + { + "epoch": 0.34, + "learning_rate": 2.9018747724790684e-06, + "loss": 0.91, + "step": 1847 + }, + { + "epoch": 0.34, + "learning_rate": 2.901820167455406e-06, + "loss": 1.0947, + "step": 1848 + }, + { + "epoch": 0.34, + "learning_rate": 2.901765562431744e-06, + "loss": 0.9324, + "step": 1849 + }, + { + "epoch": 0.34, + "learning_rate": 2.9017109574080813e-06, + "loss": 0.4146, + "step": 1850 + }, + { + "epoch": 0.34, + "learning_rate": 2.9016563523844193e-06, + "loss": 1.2582, + "step": 1851 + }, + { + "epoch": 0.34, + "learning_rate": 2.9016017473607572e-06, + "loss": 0.0373, + "step": 1852 + }, + { + "epoch": 0.34, + "learning_rate": 2.9015471423370948e-06, + "loss": 0.2488, + "step": 1853 + }, + { + "epoch": 0.34, + "learning_rate": 2.9014925373134327e-06, + "loss": 0.8904, + "step": 1854 + }, + { + "epoch": 0.34, + "learning_rate": 2.9014379322897706e-06, + "loss": 1.9762, + "step": 1855 + }, + { + "epoch": 0.34, + "learning_rate": 2.9013833272661086e-06, + "loss": 0.0857, + "step": 1856 + }, + { + "epoch": 0.34, + "learning_rate": 2.9013287222424465e-06, + "loss": 0.2668, + "step": 1857 + }, + { + "epoch": 0.34, + "learning_rate": 2.9012741172187845e-06, + "loss": 0.1555, + "step": 1858 + }, + { + "epoch": 0.34, + "learning_rate": 2.901219512195122e-06, + "loss": 1.9618, + "step": 1859 + }, + { + "epoch": 0.34, + "learning_rate": 2.90116490717146e-06, + "loss": 0.583, + "step": 1860 + }, + { + "epoch": 0.34, + "learning_rate": 2.901110302147798e-06, + "loss": 1.662, + "step": 1861 + }, + { + "epoch": 0.34, + "learning_rate": 2.9010556971241354e-06, + "loss": 0.9713, + "step": 1862 + }, + { + "epoch": 0.34, + "learning_rate": 2.9010010921004733e-06, + "loss": 1.2609, + "step": 1863 + }, + { + "epoch": 0.34, + "learning_rate": 2.900946487076811e-06, + "loss": 0.7688, + "step": 1864 + }, + { + "epoch": 0.34, + "learning_rate": 2.900891882053149e-06, + "loss": 0.131, + "step": 1865 + }, + { + "epoch": 0.34, + "learning_rate": 2.9008372770294867e-06, + "loss": 2.8621, + "step": 1866 + }, + { + "epoch": 0.34, + "learning_rate": 2.9007826720058243e-06, + "loss": 0.5729, + "step": 1867 + }, + { + "epoch": 0.34, + "learning_rate": 2.900728066982162e-06, + "loss": 1.6394, + "step": 1868 + }, + { + "epoch": 0.34, + "learning_rate": 2.9006734619585e-06, + "loss": 1.2481, + "step": 1869 + }, + { + "epoch": 0.34, + "learning_rate": 2.900618856934838e-06, + "loss": 0.9356, + "step": 1870 + }, + { + "epoch": 0.34, + "learning_rate": 2.900564251911176e-06, + "loss": 0.8891, + "step": 1871 + }, + { + "epoch": 0.34, + "learning_rate": 2.900509646887514e-06, + "loss": 0.049, + "step": 1872 + }, + { + "epoch": 0.34, + "learning_rate": 2.9004550418638515e-06, + "loss": 0.0909, + "step": 1873 + }, + { + "epoch": 0.34, + "learning_rate": 2.9004004368401894e-06, + "loss": 0.3892, + "step": 1874 + }, + { + "epoch": 0.34, + "learning_rate": 2.9003458318165274e-06, + "loss": 0.2485, + "step": 1875 + }, + { + "epoch": 0.34, + "learning_rate": 2.900291226792865e-06, + "loss": 2.7968, + "step": 1876 + }, + { + "epoch": 0.34, + "learning_rate": 2.900236621769203e-06, + "loss": 0.9346, + "step": 1877 + }, + { + "epoch": 0.34, + "learning_rate": 2.9001820167455404e-06, + "loss": 2.6668, + "step": 1878 + }, + { + "epoch": 0.34, + "learning_rate": 2.9001274117218783e-06, + "loss": 0.539, + "step": 1879 + }, + { + "epoch": 0.34, + "learning_rate": 2.9000728066982163e-06, + "loss": 2.6155, + "step": 1880 + }, + { + "epoch": 0.34, + "learning_rate": 2.900018201674554e-06, + "loss": 0.4101, + "step": 1881 + }, + { + "epoch": 0.34, + "learning_rate": 2.899963596650892e-06, + "loss": 0.2318, + "step": 1882 + }, + { + "epoch": 0.34, + "learning_rate": 2.89990899162723e-06, + "loss": 1.1737, + "step": 1883 + }, + { + "epoch": 0.34, + "learning_rate": 2.8998543866035676e-06, + "loss": 1.0083, + "step": 1884 + }, + { + "epoch": 0.34, + "learning_rate": 2.8997997815799056e-06, + "loss": 1.1377, + "step": 1885 + }, + { + "epoch": 0.34, + "learning_rate": 2.8997451765562435e-06, + "loss": 0.5157, + "step": 1886 + }, + { + "epoch": 0.34, + "learning_rate": 2.899690571532581e-06, + "loss": 0.6704, + "step": 1887 + }, + { + "epoch": 0.34, + "learning_rate": 2.899635966508919e-06, + "loss": 2.2367, + "step": 1888 + }, + { + "epoch": 0.34, + "learning_rate": 2.899581361485257e-06, + "loss": 0.5008, + "step": 1889 + }, + { + "epoch": 0.34, + "learning_rate": 2.8995267564615944e-06, + "loss": 1.6511, + "step": 1890 + }, + { + "epoch": 0.34, + "learning_rate": 2.8994721514379324e-06, + "loss": 1.6347, + "step": 1891 + }, + { + "epoch": 0.34, + "learning_rate": 2.89941754641427e-06, + "loss": 0.9434, + "step": 1892 + }, + { + "epoch": 0.34, + "learning_rate": 2.899362941390608e-06, + "loss": 0.625, + "step": 1893 + }, + { + "epoch": 0.34, + "learning_rate": 2.8993083363669458e-06, + "loss": 1.0955, + "step": 1894 + }, + { + "epoch": 0.34, + "learning_rate": 2.8992537313432837e-06, + "loss": 1.305, + "step": 1895 + }, + { + "epoch": 0.34, + "learning_rate": 2.8991991263196217e-06, + "loss": 0.9728, + "step": 1896 + }, + { + "epoch": 0.34, + "learning_rate": 2.8991445212959596e-06, + "loss": 0.1817, + "step": 1897 + }, + { + "epoch": 0.35, + "learning_rate": 2.899089916272297e-06, + "loss": 1.8465, + "step": 1898 + }, + { + "epoch": 0.35, + "learning_rate": 2.899035311248635e-06, + "loss": 2.6309, + "step": 1899 + }, + { + "epoch": 0.35, + "learning_rate": 2.898980706224973e-06, + "loss": 0.318, + "step": 1900 + }, + { + "epoch": 0.35, + "learning_rate": 2.8989261012013105e-06, + "loss": 1.3902, + "step": 1901 + }, + { + "epoch": 0.35, + "learning_rate": 2.8988714961776485e-06, + "loss": 1.1305, + "step": 1902 + }, + { + "epoch": 0.35, + "learning_rate": 2.8988168911539864e-06, + "loss": 1.3461, + "step": 1903 + }, + { + "epoch": 0.35, + "learning_rate": 2.898762286130324e-06, + "loss": 1.1179, + "step": 1904 + }, + { + "epoch": 0.35, + "learning_rate": 2.898707681106662e-06, + "loss": 0.5508, + "step": 1905 + }, + { + "epoch": 0.35, + "learning_rate": 2.8986530760829994e-06, + "loss": 0.7755, + "step": 1906 + }, + { + "epoch": 0.35, + "learning_rate": 2.8985984710593373e-06, + "loss": 1.3733, + "step": 1907 + }, + { + "epoch": 0.35, + "learning_rate": 2.8985438660356753e-06, + "loss": 0.9649, + "step": 1908 + }, + { + "epoch": 0.35, + "learning_rate": 2.8984892610120132e-06, + "loss": 1.4318, + "step": 1909 + }, + { + "epoch": 0.35, + "learning_rate": 2.898434655988351e-06, + "loss": 1.6913, + "step": 1910 + }, + { + "epoch": 0.35, + "learning_rate": 2.898380050964689e-06, + "loss": 1.2739, + "step": 1911 + }, + { + "epoch": 0.35, + "learning_rate": 2.8983254459410266e-06, + "loss": 0.1311, + "step": 1912 + }, + { + "epoch": 0.35, + "learning_rate": 2.8982708409173646e-06, + "loss": 1.0691, + "step": 1913 + }, + { + "epoch": 0.35, + "learning_rate": 2.8982162358937025e-06, + "loss": 0.9636, + "step": 1914 + }, + { + "epoch": 0.35, + "learning_rate": 2.89816163087004e-06, + "loss": 1.1859, + "step": 1915 + }, + { + "epoch": 0.35, + "learning_rate": 2.898107025846378e-06, + "loss": 0.5799, + "step": 1916 + }, + { + "epoch": 0.35, + "learning_rate": 2.8980524208227155e-06, + "loss": 1.1577, + "step": 1917 + }, + { + "epoch": 0.35, + "learning_rate": 2.8979978157990535e-06, + "loss": 0.886, + "step": 1918 + }, + { + "epoch": 0.35, + "learning_rate": 2.8979432107753914e-06, + "loss": 0.4681, + "step": 1919 + }, + { + "epoch": 0.35, + "learning_rate": 2.897888605751729e-06, + "loss": 1.2696, + "step": 1920 + }, + { + "epoch": 0.35, + "learning_rate": 2.897834000728067e-06, + "loss": 2.7333, + "step": 1921 + }, + { + "epoch": 0.35, + "learning_rate": 2.897779395704405e-06, + "loss": 0.8476, + "step": 1922 + }, + { + "epoch": 0.35, + "learning_rate": 2.8977247906807427e-06, + "loss": 1.7343, + "step": 1923 + }, + { + "epoch": 0.35, + "learning_rate": 2.8976701856570807e-06, + "loss": 0.6476, + "step": 1924 + }, + { + "epoch": 0.35, + "learning_rate": 2.8976155806334186e-06, + "loss": 0.7418, + "step": 1925 + }, + { + "epoch": 0.35, + "learning_rate": 2.897560975609756e-06, + "loss": 0.5507, + "step": 1926 + }, + { + "epoch": 0.35, + "learning_rate": 2.897506370586094e-06, + "loss": 3.697, + "step": 1927 + }, + { + "epoch": 0.35, + "learning_rate": 2.897451765562432e-06, + "loss": 0.1779, + "step": 1928 + }, + { + "epoch": 0.35, + "learning_rate": 2.8973971605387696e-06, + "loss": 1.2149, + "step": 1929 + }, + { + "epoch": 0.35, + "learning_rate": 2.8973425555151075e-06, + "loss": 0.4951, + "step": 1930 + }, + { + "epoch": 0.35, + "learning_rate": 2.897287950491445e-06, + "loss": 0.6325, + "step": 1931 + }, + { + "epoch": 0.35, + "learning_rate": 2.897233345467783e-06, + "loss": 0.3593, + "step": 1932 + }, + { + "epoch": 0.35, + "learning_rate": 2.897178740444121e-06, + "loss": 0.565, + "step": 1933 + }, + { + "epoch": 0.35, + "learning_rate": 2.8971241354204584e-06, + "loss": 0.5589, + "step": 1934 + }, + { + "epoch": 0.35, + "learning_rate": 2.8970695303967964e-06, + "loss": 0.6665, + "step": 1935 + }, + { + "epoch": 0.35, + "learning_rate": 2.8970149253731343e-06, + "loss": 1.1949, + "step": 1936 + }, + { + "epoch": 0.35, + "learning_rate": 2.8969603203494723e-06, + "loss": 0.3498, + "step": 1937 + }, + { + "epoch": 0.35, + "learning_rate": 2.89690571532581e-06, + "loss": 0.1969, + "step": 1938 + }, + { + "epoch": 0.35, + "learning_rate": 2.896851110302148e-06, + "loss": 1.0749, + "step": 1939 + }, + { + "epoch": 0.35, + "learning_rate": 2.8967965052784857e-06, + "loss": 1.8445, + "step": 1940 + }, + { + "epoch": 0.35, + "learning_rate": 2.8967419002548236e-06, + "loss": 1.443, + "step": 1941 + }, + { + "epoch": 0.35, + "learning_rate": 2.8966872952311616e-06, + "loss": 1.7418, + "step": 1942 + }, + { + "epoch": 0.35, + "learning_rate": 2.896632690207499e-06, + "loss": 0.8536, + "step": 1943 + }, + { + "epoch": 0.35, + "learning_rate": 2.896578085183837e-06, + "loss": 0.8044, + "step": 1944 + }, + { + "epoch": 0.35, + "learning_rate": 2.8965234801601745e-06, + "loss": 1.475, + "step": 1945 + }, + { + "epoch": 0.35, + "learning_rate": 2.8964688751365125e-06, + "loss": 0.3097, + "step": 1946 + }, + { + "epoch": 0.35, + "learning_rate": 2.8964142701128504e-06, + "loss": 0.8125, + "step": 1947 + }, + { + "epoch": 0.35, + "learning_rate": 2.896359665089188e-06, + "loss": 2.0919, + "step": 1948 + }, + { + "epoch": 0.35, + "learning_rate": 2.896305060065526e-06, + "loss": 2.1871, + "step": 1949 + }, + { + "epoch": 0.35, + "learning_rate": 2.896250455041864e-06, + "loss": 0.0576, + "step": 1950 + }, + { + "epoch": 0.35, + "learning_rate": 2.8961958500182018e-06, + "loss": 1.9992, + "step": 1951 + }, + { + "epoch": 0.35, + "learning_rate": 2.8961412449945397e-06, + "loss": 1.4882, + "step": 1952 + }, + { + "epoch": 0.36, + "learning_rate": 2.8960866399708777e-06, + "loss": 0.9976, + "step": 1953 + }, + { + "epoch": 0.36, + "learning_rate": 2.896032034947215e-06, + "loss": 0.8362, + "step": 1954 + }, + { + "epoch": 0.36, + "learning_rate": 2.895977429923553e-06, + "loss": 1.0448, + "step": 1955 + }, + { + "epoch": 0.36, + "learning_rate": 2.895922824899891e-06, + "loss": 1.1962, + "step": 1956 + }, + { + "epoch": 0.36, + "learning_rate": 2.8958682198762286e-06, + "loss": 0.3855, + "step": 1957 + }, + { + "epoch": 0.36, + "learning_rate": 2.8958136148525665e-06, + "loss": 1.7284, + "step": 1958 + }, + { + "epoch": 0.36, + "learning_rate": 2.895759009828904e-06, + "loss": 1.2863, + "step": 1959 + }, + { + "epoch": 0.36, + "learning_rate": 2.895704404805242e-06, + "loss": 1.9724, + "step": 1960 + }, + { + "epoch": 0.36, + "learning_rate": 2.89564979978158e-06, + "loss": 0.1302, + "step": 1961 + }, + { + "epoch": 0.36, + "learning_rate": 2.895595194757918e-06, + "loss": 1.0029, + "step": 1962 + }, + { + "epoch": 0.36, + "learning_rate": 2.895540589734256e-06, + "loss": 0.8, + "step": 1963 + }, + { + "epoch": 0.36, + "learning_rate": 2.8954859847105938e-06, + "loss": 0.7513, + "step": 1964 + }, + { + "epoch": 0.36, + "learning_rate": 2.8954313796869313e-06, + "loss": 2.2739, + "step": 1965 + }, + { + "epoch": 0.36, + "learning_rate": 2.8953767746632692e-06, + "loss": 0.7929, + "step": 1966 + }, + { + "epoch": 0.36, + "learning_rate": 2.895322169639607e-06, + "loss": 2.6691, + "step": 1967 + }, + { + "epoch": 0.36, + "learning_rate": 2.8952675646159447e-06, + "loss": 1.3063, + "step": 1968 + }, + { + "epoch": 0.36, + "learning_rate": 2.8952129595922826e-06, + "loss": 1.0177, + "step": 1969 + }, + { + "epoch": 0.36, + "learning_rate": 2.8951583545686206e-06, + "loss": 0.7687, + "step": 1970 + }, + { + "epoch": 0.36, + "learning_rate": 2.895103749544958e-06, + "loss": 2.1725, + "step": 1971 + }, + { + "epoch": 0.36, + "learning_rate": 2.895049144521296e-06, + "loss": 0.7084, + "step": 1972 + }, + { + "epoch": 0.36, + "learning_rate": 2.8949945394976336e-06, + "loss": 0.9374, + "step": 1973 + }, + { + "epoch": 0.36, + "learning_rate": 2.8949399344739715e-06, + "loss": 0.9717, + "step": 1974 + }, + { + "epoch": 0.36, + "learning_rate": 2.8948853294503094e-06, + "loss": 1.8987, + "step": 1975 + }, + { + "epoch": 0.36, + "learning_rate": 2.8948307244266474e-06, + "loss": 0.8769, + "step": 1976 + }, + { + "epoch": 0.36, + "learning_rate": 2.8947761194029853e-06, + "loss": 2.487, + "step": 1977 + }, + { + "epoch": 0.36, + "learning_rate": 2.8947215143793233e-06, + "loss": 2.0438, + "step": 1978 + }, + { + "epoch": 0.36, + "learning_rate": 2.894666909355661e-06, + "loss": 0.0438, + "step": 1979 + }, + { + "epoch": 0.36, + "learning_rate": 2.8946123043319987e-06, + "loss": 0.9814, + "step": 1980 + }, + { + "epoch": 0.36, + "learning_rate": 2.8945576993083367e-06, + "loss": 0.3717, + "step": 1981 + }, + { + "epoch": 0.36, + "learning_rate": 2.894503094284674e-06, + "loss": 0.7086, + "step": 1982 + }, + { + "epoch": 0.36, + "learning_rate": 2.894448489261012e-06, + "loss": 1.002, + "step": 1983 + }, + { + "epoch": 0.36, + "learning_rate": 2.89439388423735e-06, + "loss": 1.6262, + "step": 1984 + }, + { + "epoch": 0.36, + "learning_rate": 2.8943392792136876e-06, + "loss": 1.1854, + "step": 1985 + }, + { + "epoch": 0.36, + "learning_rate": 2.8942846741900256e-06, + "loss": 4.5036, + "step": 1986 + }, + { + "epoch": 0.36, + "learning_rate": 2.894230069166363e-06, + "loss": 1.3189, + "step": 1987 + }, + { + "epoch": 0.36, + "learning_rate": 2.894175464142701e-06, + "loss": 1.2269, + "step": 1988 + }, + { + "epoch": 0.36, + "learning_rate": 2.894120859119039e-06, + "loss": 1.2665, + "step": 1989 + }, + { + "epoch": 0.36, + "learning_rate": 2.894066254095377e-06, + "loss": 2.1866, + "step": 1990 + }, + { + "epoch": 0.36, + "learning_rate": 2.894011649071715e-06, + "loss": 1.294, + "step": 1991 + }, + { + "epoch": 0.36, + "learning_rate": 2.893957044048053e-06, + "loss": 0.7528, + "step": 1992 + }, + { + "epoch": 0.36, + "learning_rate": 2.8939024390243903e-06, + "loss": 0.5665, + "step": 1993 + }, + { + "epoch": 0.36, + "learning_rate": 2.8938478340007283e-06, + "loss": 0.7999, + "step": 1994 + }, + { + "epoch": 0.36, + "learning_rate": 2.893793228977066e-06, + "loss": 1.118, + "step": 1995 + }, + { + "epoch": 0.36, + "learning_rate": 2.8937386239534037e-06, + "loss": 1.0957, + "step": 1996 + }, + { + "epoch": 0.36, + "learning_rate": 2.8936840189297417e-06, + "loss": 0.4578, + "step": 1997 + }, + { + "epoch": 0.36, + "learning_rate": 2.893629413906079e-06, + "loss": 2.4658, + "step": 1998 + }, + { + "epoch": 0.36, + "learning_rate": 2.893574808882417e-06, + "loss": 0.3907, + "step": 1999 + }, + { + "epoch": 0.36, + "learning_rate": 2.893520203858755e-06, + "loss": 2.618, + "step": 2000 + }, + { + "epoch": 0.36, + "learning_rate": 2.8934655988350926e-06, + "loss": 1.3146, + "step": 2001 + }, + { + "epoch": 0.36, + "learning_rate": 2.8934109938114305e-06, + "loss": 0.9601, + "step": 2002 + }, + { + "epoch": 0.36, + "learning_rate": 2.8933563887877685e-06, + "loss": 1.0428, + "step": 2003 + }, + { + "epoch": 0.36, + "learning_rate": 2.8933017837641064e-06, + "loss": 1.4093, + "step": 2004 + }, + { + "epoch": 0.36, + "learning_rate": 2.8932471787404444e-06, + "loss": 0.951, + "step": 2005 + }, + { + "epoch": 0.36, + "learning_rate": 2.8931925737167823e-06, + "loss": 0.9763, + "step": 2006 + }, + { + "epoch": 0.36, + "learning_rate": 2.89313796869312e-06, + "loss": 0.0603, + "step": 2007 + }, + { + "epoch": 0.37, + "learning_rate": 2.8930833636694578e-06, + "loss": 0.6424, + "step": 2008 + }, + { + "epoch": 0.37, + "learning_rate": 2.8930287586457957e-06, + "loss": 0.4144, + "step": 2009 + }, + { + "epoch": 0.37, + "learning_rate": 2.8929741536221332e-06, + "loss": 1.5435, + "step": 2010 + }, + { + "epoch": 0.37, + "learning_rate": 2.892919548598471e-06, + "loss": 2.0296, + "step": 2011 + }, + { + "epoch": 0.37, + "learning_rate": 2.8928649435748087e-06, + "loss": 1.5944, + "step": 2012 + }, + { + "epoch": 0.37, + "learning_rate": 2.8928103385511466e-06, + "loss": 0.9145, + "step": 2013 + }, + { + "epoch": 0.37, + "learning_rate": 2.8927557335274846e-06, + "loss": 1.9823, + "step": 2014 + }, + { + "epoch": 0.37, + "learning_rate": 2.892701128503822e-06, + "loss": 1.9845, + "step": 2015 + }, + { + "epoch": 0.37, + "learning_rate": 2.89264652348016e-06, + "loss": 0.9525, + "step": 2016 + }, + { + "epoch": 0.37, + "learning_rate": 2.892591918456498e-06, + "loss": 1.0917, + "step": 2017 + }, + { + "epoch": 0.37, + "learning_rate": 2.892537313432836e-06, + "loss": 0.8435, + "step": 2018 + }, + { + "epoch": 0.37, + "learning_rate": 2.892482708409174e-06, + "loss": 3.7641, + "step": 2019 + }, + { + "epoch": 0.37, + "learning_rate": 2.892428103385512e-06, + "loss": 0.9084, + "step": 2020 + }, + { + "epoch": 0.37, + "learning_rate": 2.8923734983618493e-06, + "loss": 0.8687, + "step": 2021 + }, + { + "epoch": 0.37, + "learning_rate": 2.8923188933381873e-06, + "loss": 2.5531, + "step": 2022 + }, + { + "epoch": 0.37, + "learning_rate": 2.8922642883145252e-06, + "loss": 1.5196, + "step": 2023 + }, + { + "epoch": 0.37, + "learning_rate": 2.8922096832908627e-06, + "loss": 0.4051, + "step": 2024 + }, + { + "epoch": 0.37, + "learning_rate": 2.8921550782672007e-06, + "loss": 0.5789, + "step": 2025 + }, + { + "epoch": 0.37, + "learning_rate": 2.892100473243538e-06, + "loss": 0.532, + "step": 2026 + }, + { + "epoch": 0.37, + "learning_rate": 2.892045868219876e-06, + "loss": 0.3539, + "step": 2027 + }, + { + "epoch": 0.37, + "learning_rate": 2.891991263196214e-06, + "loss": 2.4419, + "step": 2028 + }, + { + "epoch": 0.37, + "learning_rate": 2.8919366581725516e-06, + "loss": 1.6716, + "step": 2029 + }, + { + "epoch": 0.37, + "learning_rate": 2.8918820531488896e-06, + "loss": 1.4723, + "step": 2030 + }, + { + "epoch": 0.37, + "learning_rate": 2.8918274481252275e-06, + "loss": 1.5522, + "step": 2031 + }, + { + "epoch": 0.37, + "learning_rate": 2.8917728431015654e-06, + "loss": 0.3352, + "step": 2032 + }, + { + "epoch": 0.37, + "learning_rate": 2.8917182380779034e-06, + "loss": 1.4489, + "step": 2033 + }, + { + "epoch": 0.37, + "learning_rate": 2.8916636330542413e-06, + "loss": 1.0887, + "step": 2034 + }, + { + "epoch": 0.37, + "learning_rate": 2.891609028030579e-06, + "loss": 0.9443, + "step": 2035 + }, + { + "epoch": 0.37, + "learning_rate": 2.891554423006917e-06, + "loss": 2.0954, + "step": 2036 + }, + { + "epoch": 0.37, + "learning_rate": 2.8914998179832547e-06, + "loss": 3.3095, + "step": 2037 + }, + { + "epoch": 0.37, + "learning_rate": 2.8914452129595923e-06, + "loss": 0.9373, + "step": 2038 + }, + { + "epoch": 0.37, + "learning_rate": 2.89139060793593e-06, + "loss": 0.9768, + "step": 2039 + }, + { + "epoch": 0.37, + "learning_rate": 2.8913360029122677e-06, + "loss": 0.8436, + "step": 2040 + }, + { + "epoch": 0.37, + "learning_rate": 2.8912813978886057e-06, + "loss": 0.7707, + "step": 2041 + }, + { + "epoch": 0.37, + "learning_rate": 2.8912267928649436e-06, + "loss": 0.0333, + "step": 2042 + }, + { + "epoch": 0.37, + "learning_rate": 2.8911721878412816e-06, + "loss": 1.3608, + "step": 2043 + }, + { + "epoch": 0.37, + "learning_rate": 2.8911175828176195e-06, + "loss": 3.2655, + "step": 2044 + }, + { + "epoch": 0.37, + "learning_rate": 2.8910629777939574e-06, + "loss": 0.749, + "step": 2045 + }, + { + "epoch": 0.37, + "learning_rate": 2.891008372770295e-06, + "loss": 2.6787, + "step": 2046 + }, + { + "epoch": 0.37, + "learning_rate": 2.890953767746633e-06, + "loss": 1.3792, + "step": 2047 + }, + { + "epoch": 0.37, + "learning_rate": 2.890899162722971e-06, + "loss": 0.4859, + "step": 2048 + }, + { + "epoch": 0.37, + "learning_rate": 2.8908445576993084e-06, + "loss": 0.908, + "step": 2049 + }, + { + "epoch": 0.37, + "learning_rate": 2.8907899526756463e-06, + "loss": 2.353, + "step": 2050 + }, + { + "epoch": 0.37, + "learning_rate": 2.8907353476519843e-06, + "loss": 0.2218, + "step": 2051 + }, + { + "epoch": 0.37, + "learning_rate": 2.8906807426283218e-06, + "loss": 0.667, + "step": 2052 + }, + { + "epoch": 0.37, + "learning_rate": 2.8906261376046597e-06, + "loss": 0.4597, + "step": 2053 + }, + { + "epoch": 0.37, + "learning_rate": 2.8905715325809972e-06, + "loss": 1.4823, + "step": 2054 + }, + { + "epoch": 0.37, + "learning_rate": 2.890516927557335e-06, + "loss": 1.4032, + "step": 2055 + }, + { + "epoch": 0.37, + "learning_rate": 2.890462322533673e-06, + "loss": 1.8354, + "step": 2056 + }, + { + "epoch": 0.37, + "learning_rate": 2.890407717510011e-06, + "loss": 1.3691, + "step": 2057 + }, + { + "epoch": 0.37, + "learning_rate": 2.890353112486349e-06, + "loss": 3.1388, + "step": 2058 + }, + { + "epoch": 0.37, + "learning_rate": 2.890298507462687e-06, + "loss": 2.0183, + "step": 2059 + }, + { + "epoch": 0.37, + "learning_rate": 2.8902439024390245e-06, + "loss": 0.9947, + "step": 2060 + }, + { + "epoch": 0.37, + "learning_rate": 2.8901892974153624e-06, + "loss": 0.6064, + "step": 2061 + }, + { + "epoch": 0.37, + "learning_rate": 2.8901346923917004e-06, + "loss": 0.8055, + "step": 2062 + }, + { + "epoch": 0.38, + "learning_rate": 2.890080087368038e-06, + "loss": 0.1209, + "step": 2063 + }, + { + "epoch": 0.38, + "learning_rate": 2.890025482344376e-06, + "loss": 0.6986, + "step": 2064 + }, + { + "epoch": 0.38, + "learning_rate": 2.8899708773207138e-06, + "loss": 1.9435, + "step": 2065 + }, + { + "epoch": 0.38, + "learning_rate": 2.8899162722970513e-06, + "loss": 2.9326, + "step": 2066 + }, + { + "epoch": 0.38, + "learning_rate": 2.8898616672733892e-06, + "loss": 2.0975, + "step": 2067 + }, + { + "epoch": 0.38, + "learning_rate": 2.8898070622497268e-06, + "loss": 0.6026, + "step": 2068 + }, + { + "epoch": 0.38, + "learning_rate": 2.8897524572260647e-06, + "loss": 1.8264, + "step": 2069 + }, + { + "epoch": 0.38, + "learning_rate": 2.8896978522024026e-06, + "loss": 0.7379, + "step": 2070 + }, + { + "epoch": 0.38, + "learning_rate": 2.8896432471787406e-06, + "loss": 1.2248, + "step": 2071 + }, + { + "epoch": 0.38, + "learning_rate": 2.8895886421550785e-06, + "loss": 2.9779, + "step": 2072 + }, + { + "epoch": 0.38, + "learning_rate": 2.8895340371314165e-06, + "loss": 1.1169, + "step": 2073 + }, + { + "epoch": 0.38, + "learning_rate": 2.889479432107754e-06, + "loss": 0.9418, + "step": 2074 + }, + { + "epoch": 0.38, + "learning_rate": 2.889424827084092e-06, + "loss": 0.078, + "step": 2075 + }, + { + "epoch": 0.38, + "learning_rate": 2.88937022206043e-06, + "loss": 1.0644, + "step": 2076 + }, + { + "epoch": 0.38, + "learning_rate": 2.8893156170367674e-06, + "loss": 0.4817, + "step": 2077 + }, + { + "epoch": 0.38, + "learning_rate": 2.8892610120131053e-06, + "loss": 0.687, + "step": 2078 + }, + { + "epoch": 0.38, + "learning_rate": 2.889206406989443e-06, + "loss": 2.5976, + "step": 2079 + }, + { + "epoch": 0.38, + "learning_rate": 2.889151801965781e-06, + "loss": 1.4437, + "step": 2080 + }, + { + "epoch": 0.38, + "learning_rate": 2.8890971969421187e-06, + "loss": 1.5171, + "step": 2081 + }, + { + "epoch": 0.38, + "learning_rate": 2.8890425919184563e-06, + "loss": 3.557, + "step": 2082 + }, + { + "epoch": 0.38, + "learning_rate": 2.888987986894794e-06, + "loss": 1.1445, + "step": 2083 + }, + { + "epoch": 0.38, + "learning_rate": 2.888933381871132e-06, + "loss": 2.8984, + "step": 2084 + }, + { + "epoch": 0.38, + "learning_rate": 2.88887877684747e-06, + "loss": 0.7566, + "step": 2085 + }, + { + "epoch": 0.38, + "learning_rate": 2.888824171823808e-06, + "loss": 0.9165, + "step": 2086 + }, + { + "epoch": 0.38, + "learning_rate": 2.888769566800146e-06, + "loss": 1.2524, + "step": 2087 + }, + { + "epoch": 0.38, + "learning_rate": 2.8887149617764835e-06, + "loss": 0.6667, + "step": 2088 + }, + { + "epoch": 0.38, + "learning_rate": 2.8886603567528214e-06, + "loss": 1.5176, + "step": 2089 + }, + { + "epoch": 0.38, + "learning_rate": 2.8886057517291594e-06, + "loss": 1.5393, + "step": 2090 + }, + { + "epoch": 0.38, + "learning_rate": 2.888551146705497e-06, + "loss": 0.965, + "step": 2091 + }, + { + "epoch": 0.38, + "learning_rate": 2.888496541681835e-06, + "loss": 1.261, + "step": 2092 + }, + { + "epoch": 0.38, + "learning_rate": 2.8884419366581724e-06, + "loss": 1.6922, + "step": 2093 + }, + { + "epoch": 0.38, + "learning_rate": 2.8883873316345103e-06, + "loss": 0.1178, + "step": 2094 + }, + { + "epoch": 0.38, + "learning_rate": 2.8883327266108483e-06, + "loss": 0.805, + "step": 2095 + }, + { + "epoch": 0.38, + "learning_rate": 2.8882781215871858e-06, + "loss": 0.2408, + "step": 2096 + }, + { + "epoch": 0.38, + "learning_rate": 2.8882235165635237e-06, + "loss": 1.4871, + "step": 2097 + }, + { + "epoch": 0.38, + "learning_rate": 2.8881689115398617e-06, + "loss": 0.7234, + "step": 2098 + }, + { + "epoch": 0.38, + "learning_rate": 2.8881143065161996e-06, + "loss": 1.5057, + "step": 2099 + }, + { + "epoch": 0.38, + "learning_rate": 2.8880597014925376e-06, + "loss": 0.3199, + "step": 2100 + }, + { + "epoch": 0.38, + "learning_rate": 2.8880050964688755e-06, + "loss": 0.1853, + "step": 2101 + }, + { + "epoch": 0.38, + "learning_rate": 2.887950491445213e-06, + "loss": 1.171, + "step": 2102 + }, + { + "epoch": 0.38, + "learning_rate": 2.887895886421551e-06, + "loss": 0.7408, + "step": 2103 + }, + { + "epoch": 0.38, + "learning_rate": 2.887841281397889e-06, + "loss": 1.351, + "step": 2104 + }, + { + "epoch": 0.38, + "learning_rate": 2.8877866763742264e-06, + "loss": 0.0927, + "step": 2105 + }, + { + "epoch": 0.38, + "learning_rate": 2.8877320713505644e-06, + "loss": 0.5931, + "step": 2106 + }, + { + "epoch": 0.38, + "learning_rate": 2.887677466326902e-06, + "loss": 1.3979, + "step": 2107 + }, + { + "epoch": 0.38, + "learning_rate": 2.88762286130324e-06, + "loss": 1.1572, + "step": 2108 + }, + { + "epoch": 0.38, + "learning_rate": 2.8875682562795778e-06, + "loss": 1.2715, + "step": 2109 + }, + { + "epoch": 0.38, + "learning_rate": 2.8875136512559153e-06, + "loss": 2.88, + "step": 2110 + }, + { + "epoch": 0.38, + "learning_rate": 2.8874590462322532e-06, + "loss": 0.0471, + "step": 2111 + }, + { + "epoch": 0.38, + "learning_rate": 2.887404441208591e-06, + "loss": 2.8927, + "step": 2112 + }, + { + "epoch": 0.38, + "learning_rate": 2.887349836184929e-06, + "loss": 0.1506, + "step": 2113 + }, + { + "epoch": 0.38, + "learning_rate": 2.887295231161267e-06, + "loss": 1.2334, + "step": 2114 + }, + { + "epoch": 0.38, + "learning_rate": 2.887240626137605e-06, + "loss": 2.2161, + "step": 2115 + }, + { + "epoch": 0.38, + "learning_rate": 2.8871860211139425e-06, + "loss": 1.5298, + "step": 2116 + }, + { + "epoch": 0.38, + "learning_rate": 2.8871314160902805e-06, + "loss": 0.0705, + "step": 2117 + }, + { + "epoch": 0.39, + "learning_rate": 2.8870768110666184e-06, + "loss": 1.2666, + "step": 2118 + }, + { + "epoch": 0.39, + "learning_rate": 2.887022206042956e-06, + "loss": 0.8247, + "step": 2119 + }, + { + "epoch": 0.39, + "learning_rate": 2.886967601019294e-06, + "loss": 0.1478, + "step": 2120 + }, + { + "epoch": 0.39, + "learning_rate": 2.8869129959956314e-06, + "loss": 1.0434, + "step": 2121 + }, + { + "epoch": 0.39, + "learning_rate": 2.8868583909719693e-06, + "loss": 2.081, + "step": 2122 + }, + { + "epoch": 0.39, + "learning_rate": 2.8868037859483073e-06, + "loss": 1.4824, + "step": 2123 + }, + { + "epoch": 0.39, + "learning_rate": 2.886749180924645e-06, + "loss": 1.5691, + "step": 2124 + }, + { + "epoch": 0.39, + "learning_rate": 2.8866945759009827e-06, + "loss": 0.42, + "step": 2125 + }, + { + "epoch": 0.39, + "learning_rate": 2.8866399708773207e-06, + "loss": 0.5354, + "step": 2126 + }, + { + "epoch": 0.39, + "learning_rate": 2.8865853658536586e-06, + "loss": 1.202, + "step": 2127 + }, + { + "epoch": 0.39, + "learning_rate": 2.8865307608299966e-06, + "loss": 1.3161, + "step": 2128 + }, + { + "epoch": 0.39, + "learning_rate": 2.8864761558063345e-06, + "loss": 0.0685, + "step": 2129 + }, + { + "epoch": 0.39, + "learning_rate": 2.886421550782672e-06, + "loss": 0.2343, + "step": 2130 + }, + { + "epoch": 0.39, + "learning_rate": 2.88636694575901e-06, + "loss": 1.5334, + "step": 2131 + }, + { + "epoch": 0.39, + "learning_rate": 2.886312340735348e-06, + "loss": 0.8108, + "step": 2132 + }, + { + "epoch": 0.39, + "learning_rate": 2.8862577357116854e-06, + "loss": 1.7363, + "step": 2133 + }, + { + "epoch": 0.39, + "learning_rate": 2.8862031306880234e-06, + "loss": 1.1414, + "step": 2134 + }, + { + "epoch": 0.39, + "learning_rate": 2.886148525664361e-06, + "loss": 0.5058, + "step": 2135 + }, + { + "epoch": 0.39, + "learning_rate": 2.886093920640699e-06, + "loss": 3.1744, + "step": 2136 + }, + { + "epoch": 0.39, + "learning_rate": 2.886039315617037e-06, + "loss": 1.9064, + "step": 2137 + }, + { + "epoch": 0.39, + "learning_rate": 2.8859847105933747e-06, + "loss": 1.1957, + "step": 2138 + }, + { + "epoch": 0.39, + "learning_rate": 2.8859301055697127e-06, + "loss": 1.5522, + "step": 2139 + }, + { + "epoch": 0.39, + "learning_rate": 2.8858755005460506e-06, + "loss": 1.8688, + "step": 2140 + }, + { + "epoch": 0.39, + "learning_rate": 2.885820895522388e-06, + "loss": 1.3832, + "step": 2141 + }, + { + "epoch": 0.39, + "learning_rate": 2.885766290498726e-06, + "loss": 0.4686, + "step": 2142 + }, + { + "epoch": 0.39, + "learning_rate": 2.885711685475064e-06, + "loss": 0.8392, + "step": 2143 + }, + { + "epoch": 0.39, + "learning_rate": 2.8856570804514016e-06, + "loss": 0.2203, + "step": 2144 + }, + { + "epoch": 0.39, + "learning_rate": 2.8856024754277395e-06, + "loss": 0.4774, + "step": 2145 + }, + { + "epoch": 0.39, + "learning_rate": 2.885547870404077e-06, + "loss": 0.9946, + "step": 2146 + }, + { + "epoch": 0.39, + "learning_rate": 2.885493265380415e-06, + "loss": 0.3352, + "step": 2147 + }, + { + "epoch": 0.39, + "learning_rate": 2.885438660356753e-06, + "loss": 2.0715, + "step": 2148 + }, + { + "epoch": 0.39, + "learning_rate": 2.8853840553330904e-06, + "loss": 1.0778, + "step": 2149 + }, + { + "epoch": 0.39, + "learning_rate": 2.8853294503094284e-06, + "loss": 0.5126, + "step": 2150 + }, + { + "epoch": 0.39, + "learning_rate": 2.8852748452857663e-06, + "loss": 0.1583, + "step": 2151 + }, + { + "epoch": 0.39, + "learning_rate": 2.8852202402621043e-06, + "loss": 0.5785, + "step": 2152 + }, + { + "epoch": 0.39, + "learning_rate": 2.885165635238442e-06, + "loss": 1.5727, + "step": 2153 + }, + { + "epoch": 0.39, + "learning_rate": 2.88511103021478e-06, + "loss": 0.8408, + "step": 2154 + }, + { + "epoch": 0.39, + "learning_rate": 2.8850564251911177e-06, + "loss": 0.887, + "step": 2155 + }, + { + "epoch": 0.39, + "learning_rate": 2.8850018201674556e-06, + "loss": 1.6913, + "step": 2156 + }, + { + "epoch": 0.39, + "learning_rate": 2.8849472151437935e-06, + "loss": 1.3147, + "step": 2157 + }, + { + "epoch": 0.39, + "learning_rate": 2.884892610120131e-06, + "loss": 1.6703, + "step": 2158 + }, + { + "epoch": 0.39, + "learning_rate": 2.884838005096469e-06, + "loss": 1.5359, + "step": 2159 + }, + { + "epoch": 0.39, + "learning_rate": 2.8847834000728065e-06, + "loss": 0.8019, + "step": 2160 + }, + { + "epoch": 0.39, + "learning_rate": 2.8847287950491445e-06, + "loss": 2.5774, + "step": 2161 + }, + { + "epoch": 0.39, + "learning_rate": 2.8846741900254824e-06, + "loss": 0.9009, + "step": 2162 + }, + { + "epoch": 0.39, + "learning_rate": 2.88461958500182e-06, + "loss": 1.0543, + "step": 2163 + }, + { + "epoch": 0.39, + "learning_rate": 2.884564979978158e-06, + "loss": 1.0066, + "step": 2164 + }, + { + "epoch": 0.39, + "learning_rate": 2.884510374954496e-06, + "loss": 0.3454, + "step": 2165 + }, + { + "epoch": 0.39, + "learning_rate": 2.8844557699308338e-06, + "loss": 1.0309, + "step": 2166 + }, + { + "epoch": 0.39, + "learning_rate": 2.8844011649071717e-06, + "loss": 0.034, + "step": 2167 + }, + { + "epoch": 0.39, + "learning_rate": 2.8843465598835097e-06, + "loss": 0.062, + "step": 2168 + }, + { + "epoch": 0.39, + "learning_rate": 2.884291954859847e-06, + "loss": 0.252, + "step": 2169 + }, + { + "epoch": 0.39, + "learning_rate": 2.884237349836185e-06, + "loss": 0.8612, + "step": 2170 + }, + { + "epoch": 0.39, + "learning_rate": 2.884182744812523e-06, + "loss": 0.356, + "step": 2171 + }, + { + "epoch": 0.39, + "learning_rate": 2.8841281397888606e-06, + "loss": 0.7561, + "step": 2172 + }, + { + "epoch": 0.4, + "learning_rate": 2.8840735347651985e-06, + "loss": 1.461, + "step": 2173 + }, + { + "epoch": 0.4, + "learning_rate": 2.884018929741536e-06, + "loss": 0.7222, + "step": 2174 + }, + { + "epoch": 0.4, + "learning_rate": 2.883964324717874e-06, + "loss": 0.4505, + "step": 2175 + }, + { + "epoch": 0.4, + "learning_rate": 2.883909719694212e-06, + "loss": 0.5929, + "step": 2176 + }, + { + "epoch": 0.4, + "learning_rate": 2.8838551146705495e-06, + "loss": 2.6868, + "step": 2177 + }, + { + "epoch": 0.4, + "learning_rate": 2.8838005096468874e-06, + "loss": 0.1527, + "step": 2178 + }, + { + "epoch": 0.4, + "learning_rate": 2.8837459046232253e-06, + "loss": 0.2177, + "step": 2179 + }, + { + "epoch": 0.4, + "learning_rate": 2.8836912995995633e-06, + "loss": 1.2883, + "step": 2180 + }, + { + "epoch": 0.4, + "learning_rate": 2.8836366945759012e-06, + "loss": 0.8375, + "step": 2181 + }, + { + "epoch": 0.4, + "learning_rate": 2.883582089552239e-06, + "loss": 0.4997, + "step": 2182 + }, + { + "epoch": 0.4, + "learning_rate": 2.8835274845285767e-06, + "loss": 0.857, + "step": 2183 + }, + { + "epoch": 0.4, + "learning_rate": 2.8834728795049146e-06, + "loss": 1.0567, + "step": 2184 + }, + { + "epoch": 0.4, + "learning_rate": 2.8834182744812526e-06, + "loss": 1.6184, + "step": 2185 + }, + { + "epoch": 0.4, + "learning_rate": 2.88336366945759e-06, + "loss": 1.4027, + "step": 2186 + }, + { + "epoch": 0.4, + "learning_rate": 2.883309064433928e-06, + "loss": 0.2624, + "step": 2187 + }, + { + "epoch": 0.4, + "learning_rate": 2.8832544594102656e-06, + "loss": 1.0919, + "step": 2188 + }, + { + "epoch": 0.4, + "learning_rate": 2.8831998543866035e-06, + "loss": 0.8386, + "step": 2189 + }, + { + "epoch": 0.4, + "learning_rate": 2.8831452493629414e-06, + "loss": 0.0523, + "step": 2190 + }, + { + "epoch": 0.4, + "learning_rate": 2.883090644339279e-06, + "loss": 1.099, + "step": 2191 + }, + { + "epoch": 0.4, + "learning_rate": 2.883036039315617e-06, + "loss": 0.0718, + "step": 2192 + }, + { + "epoch": 0.4, + "learning_rate": 2.882981434291955e-06, + "loss": 2.7151, + "step": 2193 + }, + { + "epoch": 0.4, + "learning_rate": 2.882926829268293e-06, + "loss": 1.8857, + "step": 2194 + }, + { + "epoch": 0.4, + "learning_rate": 2.8828722242446307e-06, + "loss": 0.8542, + "step": 2195 + }, + { + "epoch": 0.4, + "learning_rate": 2.8828176192209687e-06, + "loss": 0.333, + "step": 2196 + }, + { + "epoch": 0.4, + "learning_rate": 2.882763014197306e-06, + "loss": 2.1748, + "step": 2197 + }, + { + "epoch": 0.4, + "learning_rate": 2.882708409173644e-06, + "loss": 0.176, + "step": 2198 + }, + { + "epoch": 0.4, + "learning_rate": 2.882653804149982e-06, + "loss": 0.6402, + "step": 2199 + }, + { + "epoch": 0.4, + "learning_rate": 2.8825991991263196e-06, + "loss": 2.275, + "step": 2200 + }, + { + "epoch": 0.4, + "learning_rate": 2.8825445941026576e-06, + "loss": 0.1576, + "step": 2201 + }, + { + "epoch": 0.4, + "learning_rate": 2.882489989078995e-06, + "loss": 1.6155, + "step": 2202 + }, + { + "epoch": 0.4, + "learning_rate": 2.882435384055333e-06, + "loss": 1.2696, + "step": 2203 + }, + { + "epoch": 0.4, + "learning_rate": 2.882380779031671e-06, + "loss": 1.7551, + "step": 2204 + }, + { + "epoch": 0.4, + "learning_rate": 2.8823261740080085e-06, + "loss": 1.4085, + "step": 2205 + }, + { + "epoch": 0.4, + "learning_rate": 2.8822715689843464e-06, + "loss": 0.9332, + "step": 2206 + }, + { + "epoch": 0.4, + "learning_rate": 2.8822169639606844e-06, + "loss": 2.6914, + "step": 2207 + }, + { + "epoch": 0.4, + "learning_rate": 2.8821623589370223e-06, + "loss": 1.992, + "step": 2208 + }, + { + "epoch": 0.4, + "learning_rate": 2.8821077539133603e-06, + "loss": 0.7799, + "step": 2209 + }, + { + "epoch": 0.4, + "learning_rate": 2.882053148889698e-06, + "loss": 0.8425, + "step": 2210 + }, + { + "epoch": 0.4, + "learning_rate": 2.8819985438660357e-06, + "loss": 0.9934, + "step": 2211 + }, + { + "epoch": 0.4, + "learning_rate": 2.8819439388423737e-06, + "loss": 0.7764, + "step": 2212 + }, + { + "epoch": 0.4, + "learning_rate": 2.8818893338187116e-06, + "loss": 0.5564, + "step": 2213 + }, + { + "epoch": 0.4, + "learning_rate": 2.881834728795049e-06, + "loss": 0.0516, + "step": 2214 + }, + { + "epoch": 0.4, + "learning_rate": 2.881780123771387e-06, + "loss": 1.8337, + "step": 2215 + }, + { + "epoch": 0.4, + "learning_rate": 2.8817255187477246e-06, + "loss": 1.7885, + "step": 2216 + }, + { + "epoch": 0.4, + "learning_rate": 2.8816709137240625e-06, + "loss": 0.4701, + "step": 2217 + }, + { + "epoch": 0.4, + "learning_rate": 2.8816163087004005e-06, + "loss": 3.7771, + "step": 2218 + }, + { + "epoch": 0.4, + "learning_rate": 2.8815617036767384e-06, + "loss": 1.2422, + "step": 2219 + }, + { + "epoch": 0.4, + "learning_rate": 2.8815070986530764e-06, + "loss": 0.1158, + "step": 2220 + }, + { + "epoch": 0.4, + "learning_rate": 2.8814524936294143e-06, + "loss": 0.6154, + "step": 2221 + }, + { + "epoch": 0.4, + "learning_rate": 2.881397888605752e-06, + "loss": 1.2944, + "step": 2222 + }, + { + "epoch": 0.4, + "learning_rate": 2.8813432835820898e-06, + "loss": 0.5739, + "step": 2223 + }, + { + "epoch": 0.4, + "learning_rate": 2.8812886785584277e-06, + "loss": 0.1426, + "step": 2224 + }, + { + "epoch": 0.4, + "learning_rate": 2.8812340735347652e-06, + "loss": 1.2824, + "step": 2225 + }, + { + "epoch": 0.4, + "learning_rate": 2.881179468511103e-06, + "loss": 0.7244, + "step": 2226 + }, + { + "epoch": 0.4, + "learning_rate": 2.8811248634874407e-06, + "loss": 0.894, + "step": 2227 + }, + { + "epoch": 0.41, + "learning_rate": 2.8810702584637786e-06, + "loss": 0.9294, + "step": 2228 + }, + { + "epoch": 0.41, + "learning_rate": 2.8810156534401166e-06, + "loss": 0.6715, + "step": 2229 + }, + { + "epoch": 0.41, + "learning_rate": 2.880961048416454e-06, + "loss": 0.834, + "step": 2230 + }, + { + "epoch": 0.41, + "learning_rate": 2.880906443392792e-06, + "loss": 1.951, + "step": 2231 + }, + { + "epoch": 0.41, + "learning_rate": 2.88085183836913e-06, + "loss": 0.4693, + "step": 2232 + }, + { + "epoch": 0.41, + "learning_rate": 2.880797233345468e-06, + "loss": 0.4842, + "step": 2233 + }, + { + "epoch": 0.41, + "learning_rate": 2.880742628321806e-06, + "loss": 0.2552, + "step": 2234 + }, + { + "epoch": 0.41, + "learning_rate": 2.880688023298144e-06, + "loss": 1.4403, + "step": 2235 + }, + { + "epoch": 0.41, + "learning_rate": 2.8806334182744813e-06, + "loss": 0.6144, + "step": 2236 + }, + { + "epoch": 0.41, + "learning_rate": 2.8805788132508193e-06, + "loss": 0.2975, + "step": 2237 + }, + { + "epoch": 0.41, + "learning_rate": 2.8805242082271572e-06, + "loss": 0.8126, + "step": 2238 + }, + { + "epoch": 0.41, + "learning_rate": 2.8804696032034947e-06, + "loss": 1.9209, + "step": 2239 + }, + { + "epoch": 0.41, + "learning_rate": 2.8804149981798327e-06, + "loss": 0.0638, + "step": 2240 + }, + { + "epoch": 0.41, + "learning_rate": 2.88036039315617e-06, + "loss": 0.9697, + "step": 2241 + }, + { + "epoch": 0.41, + "learning_rate": 2.880305788132508e-06, + "loss": 0.7335, + "step": 2242 + }, + { + "epoch": 0.41, + "learning_rate": 2.880251183108846e-06, + "loss": 0.3625, + "step": 2243 + }, + { + "epoch": 0.41, + "learning_rate": 2.8801965780851836e-06, + "loss": 0.5742, + "step": 2244 + }, + { + "epoch": 0.41, + "learning_rate": 2.8801419730615216e-06, + "loss": 2.3564, + "step": 2245 + }, + { + "epoch": 0.41, + "learning_rate": 2.8800873680378595e-06, + "loss": 0.7727, + "step": 2246 + }, + { + "epoch": 0.41, + "learning_rate": 2.8800327630141974e-06, + "loss": 1.2582, + "step": 2247 + }, + { + "epoch": 0.41, + "learning_rate": 2.8799781579905354e-06, + "loss": 1.3393, + "step": 2248 + }, + { + "epoch": 0.41, + "learning_rate": 2.8799235529668733e-06, + "loss": 2.0321, + "step": 2249 + }, + { + "epoch": 0.41, + "learning_rate": 2.879868947943211e-06, + "loss": 1.0707, + "step": 2250 + }, + { + "epoch": 0.41, + "learning_rate": 2.879814342919549e-06, + "loss": 0.1019, + "step": 2251 + }, + { + "epoch": 0.41, + "learning_rate": 2.8797597378958867e-06, + "loss": 0.4355, + "step": 2252 + }, + { + "epoch": 0.41, + "learning_rate": 2.8797051328722243e-06, + "loss": 1.31, + "step": 2253 + }, + { + "epoch": 0.41, + "learning_rate": 2.879650527848562e-06, + "loss": 2.2022, + "step": 2254 + }, + { + "epoch": 0.41, + "learning_rate": 2.8795959228248997e-06, + "loss": 2.5515, + "step": 2255 + }, + { + "epoch": 0.41, + "learning_rate": 2.8795413178012377e-06, + "loss": 1.9524, + "step": 2256 + }, + { + "epoch": 0.41, + "learning_rate": 2.8794867127775756e-06, + "loss": 1.7569, + "step": 2257 + }, + { + "epoch": 0.41, + "learning_rate": 2.879432107753913e-06, + "loss": 1.635, + "step": 2258 + }, + { + "epoch": 0.41, + "learning_rate": 2.879377502730251e-06, + "loss": 2.027, + "step": 2259 + }, + { + "epoch": 0.41, + "learning_rate": 2.879322897706589e-06, + "loss": 1.8191, + "step": 2260 + }, + { + "epoch": 0.41, + "learning_rate": 2.879268292682927e-06, + "loss": 2.1221, + "step": 2261 + }, + { + "epoch": 0.41, + "learning_rate": 2.879213687659265e-06, + "loss": 0.4966, + "step": 2262 + }, + { + "epoch": 0.41, + "learning_rate": 2.879159082635603e-06, + "loss": 0.6646, + "step": 2263 + }, + { + "epoch": 0.41, + "learning_rate": 2.8791044776119404e-06, + "loss": 2.0935, + "step": 2264 + }, + { + "epoch": 0.41, + "learning_rate": 2.8790498725882783e-06, + "loss": 0.8024, + "step": 2265 + }, + { + "epoch": 0.41, + "learning_rate": 2.8789952675646162e-06, + "loss": 0.7077, + "step": 2266 + }, + { + "epoch": 0.41, + "learning_rate": 2.8789406625409538e-06, + "loss": 2.4148, + "step": 2267 + }, + { + "epoch": 0.41, + "learning_rate": 2.8788860575172917e-06, + "loss": 1.3154, + "step": 2268 + }, + { + "epoch": 0.41, + "learning_rate": 2.8788314524936292e-06, + "loss": 0.9511, + "step": 2269 + }, + { + "epoch": 0.41, + "learning_rate": 2.878776847469967e-06, + "loss": 0.5522, + "step": 2270 + }, + { + "epoch": 0.41, + "learning_rate": 2.878722242446305e-06, + "loss": 1.3678, + "step": 2271 + }, + { + "epoch": 0.41, + "learning_rate": 2.8786676374226426e-06, + "loss": 1.5022, + "step": 2272 + }, + { + "epoch": 0.41, + "learning_rate": 2.8786130323989806e-06, + "loss": 0.0635, + "step": 2273 + }, + { + "epoch": 0.41, + "learning_rate": 2.8785584273753185e-06, + "loss": 0.4607, + "step": 2274 + }, + { + "epoch": 0.41, + "learning_rate": 2.8785038223516565e-06, + "loss": 1.6619, + "step": 2275 + }, + { + "epoch": 0.41, + "learning_rate": 2.8784492173279944e-06, + "loss": 0.7369, + "step": 2276 + }, + { + "epoch": 0.41, + "learning_rate": 2.8783946123043324e-06, + "loss": 1.7663, + "step": 2277 + }, + { + "epoch": 0.41, + "learning_rate": 2.87834000728067e-06, + "loss": 0.8439, + "step": 2278 + }, + { + "epoch": 0.41, + "learning_rate": 2.878285402257008e-06, + "loss": 0.721, + "step": 2279 + }, + { + "epoch": 0.41, + "learning_rate": 2.8782307972333458e-06, + "loss": 1.5171, + "step": 2280 + }, + { + "epoch": 0.41, + "learning_rate": 2.8781761922096833e-06, + "loss": 1.0418, + "step": 2281 + }, + { + "epoch": 0.41, + "learning_rate": 2.8781215871860212e-06, + "loss": 1.1319, + "step": 2282 + }, + { + "epoch": 0.42, + "learning_rate": 2.8780669821623587e-06, + "loss": 2.3019, + "step": 2283 + }, + { + "epoch": 0.42, + "learning_rate": 2.8780123771386967e-06, + "loss": 2.4645, + "step": 2284 + }, + { + "epoch": 0.42, + "learning_rate": 2.8779577721150346e-06, + "loss": 0.8178, + "step": 2285 + }, + { + "epoch": 0.42, + "learning_rate": 2.877903167091372e-06, + "loss": 0.0992, + "step": 2286 + }, + { + "epoch": 0.42, + "learning_rate": 2.87784856206771e-06, + "loss": 1.2489, + "step": 2287 + }, + { + "epoch": 0.42, + "learning_rate": 2.877793957044048e-06, + "loss": 0.2184, + "step": 2288 + }, + { + "epoch": 0.42, + "learning_rate": 2.877739352020386e-06, + "loss": 1.5203, + "step": 2289 + }, + { + "epoch": 0.42, + "learning_rate": 2.877684746996724e-06, + "loss": 2.4842, + "step": 2290 + }, + { + "epoch": 0.42, + "learning_rate": 2.877630141973062e-06, + "loss": 1.3715, + "step": 2291 + }, + { + "epoch": 0.42, + "learning_rate": 2.8775755369493994e-06, + "loss": 1.0018, + "step": 2292 + }, + { + "epoch": 0.42, + "learning_rate": 2.8775209319257373e-06, + "loss": 1.6826, + "step": 2293 + }, + { + "epoch": 0.42, + "learning_rate": 2.8774663269020753e-06, + "loss": 1.1497, + "step": 2294 + }, + { + "epoch": 0.42, + "learning_rate": 2.877411721878413e-06, + "loss": 1.0906, + "step": 2295 + }, + { + "epoch": 0.42, + "learning_rate": 2.8773571168547507e-06, + "loss": 0.5456, + "step": 2296 + }, + { + "epoch": 0.42, + "learning_rate": 2.8773025118310883e-06, + "loss": 1.134, + "step": 2297 + }, + { + "epoch": 0.42, + "learning_rate": 2.877247906807426e-06, + "loss": 3.1515, + "step": 2298 + }, + { + "epoch": 0.42, + "learning_rate": 2.877193301783764e-06, + "loss": 2.8333, + "step": 2299 + }, + { + "epoch": 0.42, + "learning_rate": 2.877138696760102e-06, + "loss": 1.9726, + "step": 2300 + }, + { + "epoch": 0.42, + "learning_rate": 2.87708409173644e-06, + "loss": 0.3147, + "step": 2301 + }, + { + "epoch": 0.42, + "learning_rate": 2.877029486712778e-06, + "loss": 0.4177, + "step": 2302 + }, + { + "epoch": 0.42, + "learning_rate": 2.8769748816891155e-06, + "loss": 1.2762, + "step": 2303 + }, + { + "epoch": 0.42, + "learning_rate": 2.8769202766654534e-06, + "loss": 1.3666, + "step": 2304 + }, + { + "epoch": 0.42, + "learning_rate": 2.8768656716417914e-06, + "loss": 0.4517, + "step": 2305 + }, + { + "epoch": 0.42, + "learning_rate": 2.876811066618129e-06, + "loss": 1.9536, + "step": 2306 + }, + { + "epoch": 0.42, + "learning_rate": 2.876756461594467e-06, + "loss": 0.5661, + "step": 2307 + }, + { + "epoch": 0.42, + "learning_rate": 2.8767018565708044e-06, + "loss": 3.1654, + "step": 2308 + }, + { + "epoch": 0.42, + "learning_rate": 2.8766472515471423e-06, + "loss": 2.0474, + "step": 2309 + }, + { + "epoch": 0.42, + "learning_rate": 2.8765926465234803e-06, + "loss": 1.1203, + "step": 2310 + }, + { + "epoch": 0.42, + "learning_rate": 2.8765380414998178e-06, + "loss": 0.5672, + "step": 2311 + }, + { + "epoch": 0.42, + "learning_rate": 2.8764834364761557e-06, + "loss": 0.1455, + "step": 2312 + }, + { + "epoch": 0.42, + "learning_rate": 2.8764288314524937e-06, + "loss": 2.0494, + "step": 2313 + }, + { + "epoch": 0.42, + "learning_rate": 2.8763742264288316e-06, + "loss": 0.3608, + "step": 2314 + }, + { + "epoch": 0.42, + "learning_rate": 2.8763196214051695e-06, + "loss": 0.6761, + "step": 2315 + }, + { + "epoch": 0.42, + "learning_rate": 2.8762650163815075e-06, + "loss": 4.7566, + "step": 2316 + }, + { + "epoch": 0.42, + "learning_rate": 2.876210411357845e-06, + "loss": 2.6867, + "step": 2317 + }, + { + "epoch": 0.42, + "learning_rate": 2.876155806334183e-06, + "loss": 1.3663, + "step": 2318 + }, + { + "epoch": 0.42, + "learning_rate": 2.876101201310521e-06, + "loss": 3.5211, + "step": 2319 + }, + { + "epoch": 0.42, + "learning_rate": 2.8760465962868584e-06, + "loss": 1.272, + "step": 2320 + }, + { + "epoch": 0.42, + "learning_rate": 2.8759919912631964e-06, + "loss": 1.2316, + "step": 2321 + }, + { + "epoch": 0.42, + "learning_rate": 2.875937386239534e-06, + "loss": 0.9126, + "step": 2322 + }, + { + "epoch": 0.42, + "learning_rate": 2.875882781215872e-06, + "loss": 1.2586, + "step": 2323 + }, + { + "epoch": 0.42, + "learning_rate": 2.8758281761922098e-06, + "loss": 1.2217, + "step": 2324 + }, + { + "epoch": 0.42, + "learning_rate": 2.8757735711685473e-06, + "loss": 1.0503, + "step": 2325 + }, + { + "epoch": 0.42, + "learning_rate": 2.8757189661448852e-06, + "loss": 0.5738, + "step": 2326 + }, + { + "epoch": 0.42, + "learning_rate": 2.875664361121223e-06, + "loss": 1.7501, + "step": 2327 + }, + { + "epoch": 0.42, + "learning_rate": 2.875609756097561e-06, + "loss": 0.1059, + "step": 2328 + }, + { + "epoch": 0.42, + "learning_rate": 2.875555151073899e-06, + "loss": 1.5002, + "step": 2329 + }, + { + "epoch": 0.42, + "learning_rate": 2.875500546050237e-06, + "loss": 1.8046, + "step": 2330 + }, + { + "epoch": 0.42, + "learning_rate": 2.8754459410265745e-06, + "loss": 0.5538, + "step": 2331 + }, + { + "epoch": 0.42, + "learning_rate": 2.8753913360029125e-06, + "loss": 2.2709, + "step": 2332 + }, + { + "epoch": 0.42, + "learning_rate": 2.8753367309792504e-06, + "loss": 1.0834, + "step": 2333 + }, + { + "epoch": 0.42, + "learning_rate": 2.875282125955588e-06, + "loss": 0.4322, + "step": 2334 + }, + { + "epoch": 0.42, + "learning_rate": 2.875227520931926e-06, + "loss": 0.3539, + "step": 2335 + }, + { + "epoch": 0.42, + "learning_rate": 2.8751729159082634e-06, + "loss": 0.9797, + "step": 2336 + }, + { + "epoch": 0.42, + "learning_rate": 2.8751183108846013e-06, + "loss": 0.9852, + "step": 2337 + }, + { + "epoch": 0.43, + "learning_rate": 2.8750637058609393e-06, + "loss": 0.1519, + "step": 2338 + }, + { + "epoch": 0.43, + "learning_rate": 2.875009100837277e-06, + "loss": 0.503, + "step": 2339 + }, + { + "epoch": 0.43, + "learning_rate": 2.8749544958136147e-06, + "loss": 0.6049, + "step": 2340 + }, + { + "epoch": 0.43, + "learning_rate": 2.8748998907899527e-06, + "loss": 0.5082, + "step": 2341 + }, + { + "epoch": 0.43, + "learning_rate": 2.8748452857662906e-06, + "loss": 1.0096, + "step": 2342 + }, + { + "epoch": 0.43, + "learning_rate": 2.8747906807426286e-06, + "loss": 1.3581, + "step": 2343 + }, + { + "epoch": 0.43, + "learning_rate": 2.8747360757189665e-06, + "loss": 1.0037, + "step": 2344 + }, + { + "epoch": 0.43, + "learning_rate": 2.874681470695304e-06, + "loss": 0.7709, + "step": 2345 + }, + { + "epoch": 0.43, + "learning_rate": 2.874626865671642e-06, + "loss": 0.5968, + "step": 2346 + }, + { + "epoch": 0.43, + "learning_rate": 2.87457226064798e-06, + "loss": 0.0765, + "step": 2347 + }, + { + "epoch": 0.43, + "learning_rate": 2.8745176556243174e-06, + "loss": 2.0559, + "step": 2348 + }, + { + "epoch": 0.43, + "learning_rate": 2.8744630506006554e-06, + "loss": 1.76, + "step": 2349 + }, + { + "epoch": 0.43, + "learning_rate": 2.874408445576993e-06, + "loss": 0.5847, + "step": 2350 + }, + { + "epoch": 0.43, + "learning_rate": 2.874353840553331e-06, + "loss": 1.5309, + "step": 2351 + }, + { + "epoch": 0.43, + "learning_rate": 2.874299235529669e-06, + "loss": 0.4781, + "step": 2352 + }, + { + "epoch": 0.43, + "learning_rate": 2.8742446305060063e-06, + "loss": 1.4067, + "step": 2353 + }, + { + "epoch": 0.43, + "learning_rate": 2.8741900254823443e-06, + "loss": 0.7565, + "step": 2354 + }, + { + "epoch": 0.43, + "learning_rate": 2.874135420458682e-06, + "loss": 0.5701, + "step": 2355 + }, + { + "epoch": 0.43, + "learning_rate": 2.87408081543502e-06, + "loss": 0.7015, + "step": 2356 + }, + { + "epoch": 0.43, + "learning_rate": 2.874026210411358e-06, + "loss": 1.3524, + "step": 2357 + }, + { + "epoch": 0.43, + "learning_rate": 2.873971605387696e-06, + "loss": 1.748, + "step": 2358 + }, + { + "epoch": 0.43, + "learning_rate": 2.8739170003640336e-06, + "loss": 0.4523, + "step": 2359 + }, + { + "epoch": 0.43, + "learning_rate": 2.8738623953403715e-06, + "loss": 2.3272, + "step": 2360 + }, + { + "epoch": 0.43, + "learning_rate": 2.8738077903167094e-06, + "loss": 1.0009, + "step": 2361 + }, + { + "epoch": 0.43, + "learning_rate": 2.873753185293047e-06, + "loss": 0.5867, + "step": 2362 + }, + { + "epoch": 0.43, + "learning_rate": 2.873698580269385e-06, + "loss": 0.7785, + "step": 2363 + }, + { + "epoch": 0.43, + "learning_rate": 2.8736439752457224e-06, + "loss": 1.2067, + "step": 2364 + }, + { + "epoch": 0.43, + "learning_rate": 2.8735893702220604e-06, + "loss": 1.1726, + "step": 2365 + }, + { + "epoch": 0.43, + "learning_rate": 2.8735347651983983e-06, + "loss": 1.4714, + "step": 2366 + }, + { + "epoch": 0.43, + "learning_rate": 2.873480160174736e-06, + "loss": 0.9088, + "step": 2367 + }, + { + "epoch": 0.43, + "learning_rate": 2.8734255551510738e-06, + "loss": 0.6542, + "step": 2368 + }, + { + "epoch": 0.43, + "learning_rate": 2.8733709501274117e-06, + "loss": 0.8294, + "step": 2369 + }, + { + "epoch": 0.43, + "learning_rate": 2.8733163451037497e-06, + "loss": 0.7067, + "step": 2370 + }, + { + "epoch": 0.43, + "learning_rate": 2.8732617400800876e-06, + "loss": 3.8098, + "step": 2371 + }, + { + "epoch": 0.43, + "learning_rate": 2.8732071350564255e-06, + "loss": 1.7198, + "step": 2372 + }, + { + "epoch": 0.43, + "learning_rate": 2.873152530032763e-06, + "loss": 1.1559, + "step": 2373 + }, + { + "epoch": 0.43, + "learning_rate": 2.873097925009101e-06, + "loss": 0.7628, + "step": 2374 + }, + { + "epoch": 0.43, + "learning_rate": 2.873043319985439e-06, + "loss": 1.9073, + "step": 2375 + }, + { + "epoch": 0.43, + "learning_rate": 2.8729887149617765e-06, + "loss": 0.9182, + "step": 2376 + }, + { + "epoch": 0.43, + "learning_rate": 2.8729341099381144e-06, + "loss": 1.5309, + "step": 2377 + }, + { + "epoch": 0.43, + "learning_rate": 2.872879504914452e-06, + "loss": 0.5694, + "step": 2378 + }, + { + "epoch": 0.43, + "learning_rate": 2.87282489989079e-06, + "loss": 0.4662, + "step": 2379 + }, + { + "epoch": 0.43, + "learning_rate": 2.872770294867128e-06, + "loss": 0.8799, + "step": 2380 + }, + { + "epoch": 0.43, + "learning_rate": 2.8727156898434658e-06, + "loss": 0.702, + "step": 2381 + }, + { + "epoch": 0.43, + "learning_rate": 2.8726610848198037e-06, + "loss": 0.748, + "step": 2382 + }, + { + "epoch": 0.43, + "learning_rate": 2.8726064797961412e-06, + "loss": 1.1714, + "step": 2383 + }, + { + "epoch": 0.43, + "learning_rate": 2.872551874772479e-06, + "loss": 3.0736, + "step": 2384 + }, + { + "epoch": 0.43, + "learning_rate": 2.872497269748817e-06, + "loss": 1.3956, + "step": 2385 + }, + { + "epoch": 0.43, + "learning_rate": 2.872442664725155e-06, + "loss": 1.1122, + "step": 2386 + }, + { + "epoch": 0.43, + "learning_rate": 2.8723880597014926e-06, + "loss": 0.5246, + "step": 2387 + }, + { + "epoch": 0.43, + "learning_rate": 2.8723334546778305e-06, + "loss": 0.538, + "step": 2388 + }, + { + "epoch": 0.43, + "learning_rate": 2.872278849654168e-06, + "loss": 0.2469, + "step": 2389 + }, + { + "epoch": 0.43, + "learning_rate": 2.872224244630506e-06, + "loss": 1.5023, + "step": 2390 + }, + { + "epoch": 0.43, + "learning_rate": 2.872169639606844e-06, + "loss": 1.0319, + "step": 2391 + }, + { + "epoch": 0.43, + "learning_rate": 2.8721150345831814e-06, + "loss": 1.223, + "step": 2392 + }, + { + "epoch": 0.44, + "learning_rate": 2.8720604295595194e-06, + "loss": 1.1255, + "step": 2393 + }, + { + "epoch": 0.44, + "learning_rate": 2.8720058245358573e-06, + "loss": 1.2949, + "step": 2394 + }, + { + "epoch": 0.44, + "learning_rate": 2.8719512195121953e-06, + "loss": 0.4864, + "step": 2395 + }, + { + "epoch": 0.44, + "learning_rate": 2.8718966144885332e-06, + "loss": 1.2969, + "step": 2396 + }, + { + "epoch": 0.44, + "learning_rate": 2.871842009464871e-06, + "loss": 1.6518, + "step": 2397 + }, + { + "epoch": 0.44, + "learning_rate": 2.8717874044412087e-06, + "loss": 2.4664, + "step": 2398 + }, + { + "epoch": 0.44, + "learning_rate": 2.8717327994175466e-06, + "loss": 1.6513, + "step": 2399 + }, + { + "epoch": 0.44, + "learning_rate": 2.8716781943938846e-06, + "loss": 0.1697, + "step": 2400 + }, + { + "epoch": 0.44, + "learning_rate": 2.871623589370222e-06, + "loss": 0.6335, + "step": 2401 + }, + { + "epoch": 0.44, + "learning_rate": 2.87156898434656e-06, + "loss": 1.0911, + "step": 2402 + }, + { + "epoch": 0.44, + "learning_rate": 2.8715143793228976e-06, + "loss": 0.7154, + "step": 2403 + }, + { + "epoch": 0.44, + "learning_rate": 2.8714597742992355e-06, + "loss": 2.4498, + "step": 2404 + }, + { + "epoch": 0.44, + "learning_rate": 2.8714051692755734e-06, + "loss": 2.1708, + "step": 2405 + }, + { + "epoch": 0.44, + "learning_rate": 2.871350564251911e-06, + "loss": 0.2037, + "step": 2406 + }, + { + "epoch": 0.44, + "learning_rate": 2.871295959228249e-06, + "loss": 2.4159, + "step": 2407 + }, + { + "epoch": 0.44, + "learning_rate": 2.871241354204587e-06, + "loss": 1.4595, + "step": 2408 + }, + { + "epoch": 0.44, + "learning_rate": 2.871186749180925e-06, + "loss": 1.0495, + "step": 2409 + }, + { + "epoch": 0.44, + "learning_rate": 2.8711321441572627e-06, + "loss": 0.0665, + "step": 2410 + }, + { + "epoch": 0.44, + "learning_rate": 2.8710775391336007e-06, + "loss": 0.7824, + "step": 2411 + }, + { + "epoch": 0.44, + "learning_rate": 2.871022934109938e-06, + "loss": 0.6412, + "step": 2412 + }, + { + "epoch": 0.44, + "learning_rate": 2.870968329086276e-06, + "loss": 0.883, + "step": 2413 + }, + { + "epoch": 0.44, + "learning_rate": 2.870913724062614e-06, + "loss": 0.3059, + "step": 2414 + }, + { + "epoch": 0.44, + "learning_rate": 2.8708591190389516e-06, + "loss": 0.5613, + "step": 2415 + }, + { + "epoch": 0.44, + "learning_rate": 2.8708045140152895e-06, + "loss": 0.4654, + "step": 2416 + }, + { + "epoch": 0.44, + "learning_rate": 2.870749908991627e-06, + "loss": 0.4527, + "step": 2417 + }, + { + "epoch": 0.44, + "learning_rate": 2.870695303967965e-06, + "loss": 0.7327, + "step": 2418 + }, + { + "epoch": 0.44, + "learning_rate": 2.870640698944303e-06, + "loss": 0.8642, + "step": 2419 + }, + { + "epoch": 0.44, + "learning_rate": 2.8705860939206405e-06, + "loss": 0.5479, + "step": 2420 + }, + { + "epoch": 0.44, + "learning_rate": 2.8705314888969784e-06, + "loss": 0.8828, + "step": 2421 + }, + { + "epoch": 0.44, + "learning_rate": 2.8704768838733164e-06, + "loss": 0.0853, + "step": 2422 + }, + { + "epoch": 0.44, + "learning_rate": 2.8704222788496543e-06, + "loss": 1.9554, + "step": 2423 + }, + { + "epoch": 0.44, + "learning_rate": 2.8703676738259922e-06, + "loss": 0.7484, + "step": 2424 + }, + { + "epoch": 0.44, + "learning_rate": 2.87031306880233e-06, + "loss": 0.6185, + "step": 2425 + }, + { + "epoch": 0.44, + "learning_rate": 2.8702584637786677e-06, + "loss": 0.8412, + "step": 2426 + }, + { + "epoch": 0.44, + "learning_rate": 2.8702038587550057e-06, + "loss": 0.1527, + "step": 2427 + }, + { + "epoch": 0.44, + "learning_rate": 2.8701492537313436e-06, + "loss": 0.4352, + "step": 2428 + }, + { + "epoch": 0.44, + "learning_rate": 2.870094648707681e-06, + "loss": 1.7473, + "step": 2429 + }, + { + "epoch": 0.44, + "learning_rate": 2.870040043684019e-06, + "loss": 1.4003, + "step": 2430 + }, + { + "epoch": 0.44, + "learning_rate": 2.8699854386603566e-06, + "loss": 0.2514, + "step": 2431 + }, + { + "epoch": 0.44, + "learning_rate": 2.8699308336366945e-06, + "loss": 0.5219, + "step": 2432 + }, + { + "epoch": 0.44, + "learning_rate": 2.8698762286130325e-06, + "loss": 2.3453, + "step": 2433 + }, + { + "epoch": 0.44, + "learning_rate": 2.86982162358937e-06, + "loss": 0.7927, + "step": 2434 + }, + { + "epoch": 0.44, + "learning_rate": 2.869767018565708e-06, + "loss": 1.059, + "step": 2435 + }, + { + "epoch": 0.44, + "learning_rate": 2.869712413542046e-06, + "loss": 1.7419, + "step": 2436 + }, + { + "epoch": 0.44, + "learning_rate": 2.869657808518384e-06, + "loss": 0.0608, + "step": 2437 + }, + { + "epoch": 0.44, + "learning_rate": 2.8696032034947218e-06, + "loss": 0.6231, + "step": 2438 + }, + { + "epoch": 0.44, + "learning_rate": 2.8695485984710597e-06, + "loss": 0.558, + "step": 2439 + }, + { + "epoch": 0.44, + "learning_rate": 2.8694939934473972e-06, + "loss": 1.137, + "step": 2440 + }, + { + "epoch": 0.44, + "learning_rate": 2.869439388423735e-06, + "loss": 0.6202, + "step": 2441 + }, + { + "epoch": 0.44, + "learning_rate": 2.869384783400073e-06, + "loss": 0.9509, + "step": 2442 + }, + { + "epoch": 0.44, + "learning_rate": 2.8693301783764106e-06, + "loss": 0.4575, + "step": 2443 + }, + { + "epoch": 0.44, + "learning_rate": 2.8692755733527486e-06, + "loss": 1.166, + "step": 2444 + }, + { + "epoch": 0.44, + "learning_rate": 2.869220968329086e-06, + "loss": 0.6417, + "step": 2445 + }, + { + "epoch": 0.44, + "learning_rate": 2.869166363305424e-06, + "loss": 0.8255, + "step": 2446 + }, + { + "epoch": 0.44, + "learning_rate": 2.869111758281762e-06, + "loss": 1.1824, + "step": 2447 + }, + { + "epoch": 0.45, + "learning_rate": 2.8690571532580995e-06, + "loss": 1.9569, + "step": 2448 + }, + { + "epoch": 0.45, + "learning_rate": 2.8690025482344374e-06, + "loss": 1.0639, + "step": 2449 + }, + { + "epoch": 0.45, + "learning_rate": 2.8689479432107754e-06, + "loss": 0.7387, + "step": 2450 + }, + { + "epoch": 0.45, + "learning_rate": 2.8688933381871133e-06, + "loss": 0.6523, + "step": 2451 + }, + { + "epoch": 0.45, + "learning_rate": 2.8688387331634513e-06, + "loss": 2.1588, + "step": 2452 + }, + { + "epoch": 0.45, + "learning_rate": 2.8687841281397892e-06, + "loss": 2.7459, + "step": 2453 + }, + { + "epoch": 0.45, + "learning_rate": 2.8687295231161267e-06, + "loss": 1.023, + "step": 2454 + }, + { + "epoch": 0.45, + "learning_rate": 2.8686749180924647e-06, + "loss": 0.8579, + "step": 2455 + }, + { + "epoch": 0.45, + "learning_rate": 2.868620313068802e-06, + "loss": 1.1278, + "step": 2456 + }, + { + "epoch": 0.45, + "learning_rate": 2.86856570804514e-06, + "loss": 1.9453, + "step": 2457 + }, + { + "epoch": 0.45, + "learning_rate": 2.868511103021478e-06, + "loss": 0.8695, + "step": 2458 + }, + { + "epoch": 0.45, + "learning_rate": 2.8684564979978156e-06, + "loss": 1.7788, + "step": 2459 + }, + { + "epoch": 0.45, + "learning_rate": 2.8684018929741536e-06, + "loss": 0.6074, + "step": 2460 + }, + { + "epoch": 0.45, + "learning_rate": 2.8683472879504915e-06, + "loss": 0.284, + "step": 2461 + }, + { + "epoch": 0.45, + "learning_rate": 2.868292682926829e-06, + "loss": 1.3841, + "step": 2462 + }, + { + "epoch": 0.45, + "learning_rate": 2.868238077903167e-06, + "loss": 0.2696, + "step": 2463 + }, + { + "epoch": 0.45, + "learning_rate": 2.868183472879505e-06, + "loss": 1.3881, + "step": 2464 + }, + { + "epoch": 0.45, + "learning_rate": 2.868128867855843e-06, + "loss": 0.5571, + "step": 2465 + }, + { + "epoch": 0.45, + "learning_rate": 2.8680742628321808e-06, + "loss": 0.0811, + "step": 2466 + }, + { + "epoch": 0.45, + "learning_rate": 2.8680196578085187e-06, + "loss": 1.6458, + "step": 2467 + }, + { + "epoch": 0.45, + "learning_rate": 2.8679650527848563e-06, + "loss": 1.3797, + "step": 2468 + }, + { + "epoch": 0.45, + "learning_rate": 2.867910447761194e-06, + "loss": 0.3265, + "step": 2469 + }, + { + "epoch": 0.45, + "learning_rate": 2.8678558427375317e-06, + "loss": 0.6424, + "step": 2470 + }, + { + "epoch": 0.45, + "learning_rate": 2.8678012377138697e-06, + "loss": 0.8688, + "step": 2471 + }, + { + "epoch": 0.45, + "learning_rate": 2.8677466326902076e-06, + "loss": 0.1693, + "step": 2472 + }, + { + "epoch": 0.45, + "learning_rate": 2.867692027666545e-06, + "loss": 1.1743, + "step": 2473 + }, + { + "epoch": 0.45, + "learning_rate": 2.867637422642883e-06, + "loss": 3.8814, + "step": 2474 + }, + { + "epoch": 0.45, + "learning_rate": 2.867582817619221e-06, + "loss": 0.3682, + "step": 2475 + }, + { + "epoch": 0.45, + "learning_rate": 2.867528212595559e-06, + "loss": 1.2434, + "step": 2476 + }, + { + "epoch": 0.45, + "learning_rate": 2.867473607571897e-06, + "loss": 1.8057, + "step": 2477 + }, + { + "epoch": 0.45, + "learning_rate": 2.867419002548235e-06, + "loss": 1.7022, + "step": 2478 + }, + { + "epoch": 0.45, + "learning_rate": 2.8673643975245724e-06, + "loss": 0.3467, + "step": 2479 + }, + { + "epoch": 0.45, + "learning_rate": 2.8673097925009103e-06, + "loss": 1.4123, + "step": 2480 + }, + { + "epoch": 0.45, + "learning_rate": 2.8672551874772482e-06, + "loss": 2.4151, + "step": 2481 + }, + { + "epoch": 0.45, + "learning_rate": 2.8672005824535858e-06, + "loss": 0.8842, + "step": 2482 + }, + { + "epoch": 0.45, + "learning_rate": 2.8671459774299237e-06, + "loss": 2.4615, + "step": 2483 + }, + { + "epoch": 0.45, + "learning_rate": 2.8670913724062612e-06, + "loss": 3.947, + "step": 2484 + }, + { + "epoch": 0.45, + "learning_rate": 2.867036767382599e-06, + "loss": 0.6008, + "step": 2485 + }, + { + "epoch": 0.45, + "learning_rate": 2.866982162358937e-06, + "loss": 1.8221, + "step": 2486 + }, + { + "epoch": 0.45, + "learning_rate": 2.8669275573352746e-06, + "loss": 1.2939, + "step": 2487 + }, + { + "epoch": 0.45, + "learning_rate": 2.8668729523116126e-06, + "loss": 3.292, + "step": 2488 + }, + { + "epoch": 0.45, + "learning_rate": 2.8668183472879505e-06, + "loss": 0.8496, + "step": 2489 + }, + { + "epoch": 0.45, + "learning_rate": 2.8667637422642885e-06, + "loss": 0.7372, + "step": 2490 + }, + { + "epoch": 0.45, + "learning_rate": 2.8667091372406264e-06, + "loss": 1.3108, + "step": 2491 + }, + { + "epoch": 0.45, + "learning_rate": 2.8666545322169644e-06, + "loss": 0.7531, + "step": 2492 + }, + { + "epoch": 0.45, + "learning_rate": 2.866599927193302e-06, + "loss": 0.9882, + "step": 2493 + }, + { + "epoch": 0.45, + "learning_rate": 2.86654532216964e-06, + "loss": 1.9415, + "step": 2494 + }, + { + "epoch": 0.45, + "learning_rate": 2.8664907171459778e-06, + "loss": 1.4884, + "step": 2495 + }, + { + "epoch": 0.45, + "learning_rate": 2.8664361121223153e-06, + "loss": 0.2155, + "step": 2496 + }, + { + "epoch": 0.45, + "learning_rate": 2.8663815070986532e-06, + "loss": 2.1614, + "step": 2497 + }, + { + "epoch": 0.45, + "learning_rate": 2.8663269020749907e-06, + "loss": 1.6114, + "step": 2498 + }, + { + "epoch": 0.45, + "learning_rate": 2.8662722970513287e-06, + "loss": 0.865, + "step": 2499 + }, + { + "epoch": 0.45, + "learning_rate": 2.8662176920276666e-06, + "loss": 2.4227, + "step": 2500 + }, + { + "epoch": 0.45, + "learning_rate": 2.866163087004004e-06, + "loss": 1.4788, + "step": 2501 + }, + { + "epoch": 0.45, + "learning_rate": 2.866108481980342e-06, + "loss": 0.8041, + "step": 2502 + }, + { + "epoch": 0.46, + "learning_rate": 2.86605387695668e-06, + "loss": 0.7344, + "step": 2503 + }, + { + "epoch": 0.46, + "learning_rate": 2.865999271933018e-06, + "loss": 0.0584, + "step": 2504 + }, + { + "epoch": 0.46, + "learning_rate": 2.865944666909356e-06, + "loss": 0.3507, + "step": 2505 + }, + { + "epoch": 0.46, + "learning_rate": 2.865890061885694e-06, + "loss": 0.2221, + "step": 2506 + }, + { + "epoch": 0.46, + "learning_rate": 2.8658354568620314e-06, + "loss": 0.1321, + "step": 2507 + }, + { + "epoch": 0.46, + "learning_rate": 2.8657808518383693e-06, + "loss": 0.8783, + "step": 2508 + }, + { + "epoch": 0.46, + "learning_rate": 2.8657262468147073e-06, + "loss": 1.6915, + "step": 2509 + }, + { + "epoch": 0.46, + "learning_rate": 2.865671641791045e-06, + "loss": 0.3772, + "step": 2510 + }, + { + "epoch": 0.46, + "learning_rate": 2.8656170367673827e-06, + "loss": 1.6855, + "step": 2511 + }, + { + "epoch": 0.46, + "learning_rate": 2.8655624317437203e-06, + "loss": 1.9469, + "step": 2512 + }, + { + "epoch": 0.46, + "learning_rate": 2.865507826720058e-06, + "loss": 1.1727, + "step": 2513 + }, + { + "epoch": 0.46, + "learning_rate": 2.865453221696396e-06, + "loss": 1.3204, + "step": 2514 + }, + { + "epoch": 0.46, + "learning_rate": 2.8653986166727337e-06, + "loss": 0.0699, + "step": 2515 + }, + { + "epoch": 0.46, + "learning_rate": 2.8653440116490716e-06, + "loss": 0.666, + "step": 2516 + }, + { + "epoch": 0.46, + "learning_rate": 2.8652894066254095e-06, + "loss": 0.4778, + "step": 2517 + }, + { + "epoch": 0.46, + "learning_rate": 2.8652348016017475e-06, + "loss": 0.281, + "step": 2518 + }, + { + "epoch": 0.46, + "learning_rate": 2.8651801965780854e-06, + "loss": 1.312, + "step": 2519 + }, + { + "epoch": 0.46, + "learning_rate": 2.8651255915544234e-06, + "loss": 0.0394, + "step": 2520 + }, + { + "epoch": 0.46, + "learning_rate": 2.865070986530761e-06, + "loss": 3.4277, + "step": 2521 + }, + { + "epoch": 0.46, + "learning_rate": 2.865016381507099e-06, + "loss": 0.6957, + "step": 2522 + }, + { + "epoch": 0.46, + "learning_rate": 2.8649617764834368e-06, + "loss": 1.2973, + "step": 2523 + }, + { + "epoch": 0.46, + "learning_rate": 2.8649071714597743e-06, + "loss": 0.8656, + "step": 2524 + }, + { + "epoch": 0.46, + "learning_rate": 2.8648525664361122e-06, + "loss": 0.351, + "step": 2525 + }, + { + "epoch": 0.46, + "learning_rate": 2.8647979614124498e-06, + "loss": 0.3634, + "step": 2526 + }, + { + "epoch": 0.46, + "learning_rate": 2.8647433563887877e-06, + "loss": 0.5142, + "step": 2527 + }, + { + "epoch": 0.46, + "learning_rate": 2.8646887513651257e-06, + "loss": 0.6984, + "step": 2528 + }, + { + "epoch": 0.46, + "learning_rate": 2.864634146341463e-06, + "loss": 1.3728, + "step": 2529 + }, + { + "epoch": 0.46, + "learning_rate": 2.864579541317801e-06, + "loss": 0.7444, + "step": 2530 + }, + { + "epoch": 0.46, + "learning_rate": 2.864524936294139e-06, + "loss": 1.5218, + "step": 2531 + }, + { + "epoch": 0.46, + "learning_rate": 2.864470331270477e-06, + "loss": 1.3916, + "step": 2532 + }, + { + "epoch": 0.46, + "learning_rate": 2.864415726246815e-06, + "loss": 0.0525, + "step": 2533 + }, + { + "epoch": 0.46, + "learning_rate": 2.864361121223153e-06, + "loss": 0.5093, + "step": 2534 + }, + { + "epoch": 0.46, + "learning_rate": 2.8643065161994904e-06, + "loss": 0.7439, + "step": 2535 + }, + { + "epoch": 0.46, + "learning_rate": 2.8642519111758284e-06, + "loss": 0.3407, + "step": 2536 + }, + { + "epoch": 0.46, + "learning_rate": 2.864197306152166e-06, + "loss": 1.5751, + "step": 2537 + }, + { + "epoch": 0.46, + "learning_rate": 2.864142701128504e-06, + "loss": 0.5567, + "step": 2538 + }, + { + "epoch": 0.46, + "learning_rate": 2.8640880961048418e-06, + "loss": 0.0712, + "step": 2539 + }, + { + "epoch": 0.46, + "learning_rate": 2.8640334910811793e-06, + "loss": 1.3561, + "step": 2540 + }, + { + "epoch": 0.46, + "learning_rate": 2.8639788860575172e-06, + "loss": 0.5933, + "step": 2541 + }, + { + "epoch": 0.46, + "learning_rate": 2.863924281033855e-06, + "loss": 0.7758, + "step": 2542 + }, + { + "epoch": 0.46, + "learning_rate": 2.8638696760101927e-06, + "loss": 1.1412, + "step": 2543 + }, + { + "epoch": 0.46, + "learning_rate": 2.8638150709865306e-06, + "loss": 2.2625, + "step": 2544 + }, + { + "epoch": 0.46, + "learning_rate": 2.8637604659628686e-06, + "loss": 0.0786, + "step": 2545 + }, + { + "epoch": 0.46, + "learning_rate": 2.8637058609392065e-06, + "loss": 0.1398, + "step": 2546 + }, + { + "epoch": 0.46, + "learning_rate": 2.8636512559155445e-06, + "loss": 0.2754, + "step": 2547 + }, + { + "epoch": 0.46, + "learning_rate": 2.8635966508918824e-06, + "loss": 1.2807, + "step": 2548 + }, + { + "epoch": 0.46, + "learning_rate": 2.86354204586822e-06, + "loss": 1.3264, + "step": 2549 + }, + { + "epoch": 0.46, + "learning_rate": 2.863487440844558e-06, + "loss": 3.2846, + "step": 2550 + }, + { + "epoch": 0.46, + "learning_rate": 2.8634328358208954e-06, + "loss": 1.4474, + "step": 2551 + }, + { + "epoch": 0.46, + "learning_rate": 2.8633782307972333e-06, + "loss": 1.0458, + "step": 2552 + }, + { + "epoch": 0.46, + "learning_rate": 2.8633236257735713e-06, + "loss": 0.159, + "step": 2553 + }, + { + "epoch": 0.46, + "learning_rate": 2.863269020749909e-06, + "loss": 1.2033, + "step": 2554 + }, + { + "epoch": 0.46, + "learning_rate": 2.8632144157262467e-06, + "loss": 1.6927, + "step": 2555 + }, + { + "epoch": 0.46, + "learning_rate": 2.8631598107025847e-06, + "loss": 1.0255, + "step": 2556 + }, + { + "epoch": 0.46, + "learning_rate": 2.8631052056789226e-06, + "loss": 1.0383, + "step": 2557 + }, + { + "epoch": 0.47, + "learning_rate": 2.8630506006552606e-06, + "loss": 0.2206, + "step": 2558 + }, + { + "epoch": 0.47, + "learning_rate": 2.8629959956315985e-06, + "loss": 0.0853, + "step": 2559 + }, + { + "epoch": 0.47, + "learning_rate": 2.862941390607936e-06, + "loss": 1.0553, + "step": 2560 + }, + { + "epoch": 0.47, + "learning_rate": 2.862886785584274e-06, + "loss": 1.0295, + "step": 2561 + }, + { + "epoch": 0.47, + "learning_rate": 2.862832180560612e-06, + "loss": 0.0671, + "step": 2562 + }, + { + "epoch": 0.47, + "learning_rate": 2.8627775755369494e-06, + "loss": 2.5835, + "step": 2563 + }, + { + "epoch": 0.47, + "learning_rate": 2.8627229705132874e-06, + "loss": 1.1176, + "step": 2564 + }, + { + "epoch": 0.47, + "learning_rate": 2.862668365489625e-06, + "loss": 0.0344, + "step": 2565 + }, + { + "epoch": 0.47, + "learning_rate": 2.862613760465963e-06, + "loss": 2.1793, + "step": 2566 + }, + { + "epoch": 0.47, + "learning_rate": 2.862559155442301e-06, + "loss": 2.8125, + "step": 2567 + }, + { + "epoch": 0.47, + "learning_rate": 2.8625045504186383e-06, + "loss": 0.225, + "step": 2568 + }, + { + "epoch": 0.47, + "learning_rate": 2.8624499453949763e-06, + "loss": 2.611, + "step": 2569 + }, + { + "epoch": 0.47, + "learning_rate": 2.862395340371314e-06, + "loss": 0.0609, + "step": 2570 + }, + { + "epoch": 0.47, + "learning_rate": 2.862340735347652e-06, + "loss": 1.6179, + "step": 2571 + }, + { + "epoch": 0.47, + "learning_rate": 2.86228613032399e-06, + "loss": 1.1829, + "step": 2572 + }, + { + "epoch": 0.47, + "learning_rate": 2.862231525300328e-06, + "loss": 0.5734, + "step": 2573 + }, + { + "epoch": 0.47, + "learning_rate": 2.8621769202766655e-06, + "loss": 0.58, + "step": 2574 + }, + { + "epoch": 0.47, + "learning_rate": 2.8621223152530035e-06, + "loss": 0.6733, + "step": 2575 + }, + { + "epoch": 0.47, + "learning_rate": 2.8620677102293414e-06, + "loss": 1.2493, + "step": 2576 + }, + { + "epoch": 0.47, + "learning_rate": 2.862013105205679e-06, + "loss": 2.2568, + "step": 2577 + }, + { + "epoch": 0.47, + "learning_rate": 2.861958500182017e-06, + "loss": 1.7352, + "step": 2578 + }, + { + "epoch": 0.47, + "learning_rate": 2.8619038951583544e-06, + "loss": 1.1175, + "step": 2579 + }, + { + "epoch": 0.47, + "learning_rate": 2.8618492901346924e-06, + "loss": 0.3205, + "step": 2580 + }, + { + "epoch": 0.47, + "learning_rate": 2.8617946851110303e-06, + "loss": 0.3445, + "step": 2581 + }, + { + "epoch": 0.47, + "learning_rate": 2.861740080087368e-06, + "loss": 1.0218, + "step": 2582 + }, + { + "epoch": 0.47, + "learning_rate": 2.8616854750637058e-06, + "loss": 0.2384, + "step": 2583 + }, + { + "epoch": 0.47, + "learning_rate": 2.8616308700400437e-06, + "loss": 0.3164, + "step": 2584 + }, + { + "epoch": 0.47, + "learning_rate": 2.8615762650163817e-06, + "loss": 2.2589, + "step": 2585 + }, + { + "epoch": 0.47, + "learning_rate": 2.8615216599927196e-06, + "loss": 0.9092, + "step": 2586 + }, + { + "epoch": 0.47, + "learning_rate": 2.8614670549690575e-06, + "loss": 0.1877, + "step": 2587 + }, + { + "epoch": 0.47, + "learning_rate": 2.861412449945395e-06, + "loss": 1.5441, + "step": 2588 + }, + { + "epoch": 0.47, + "learning_rate": 2.861357844921733e-06, + "loss": 0.5643, + "step": 2589 + }, + { + "epoch": 0.47, + "learning_rate": 2.861303239898071e-06, + "loss": 0.4182, + "step": 2590 + }, + { + "epoch": 0.47, + "learning_rate": 2.8612486348744085e-06, + "loss": 1.7443, + "step": 2591 + }, + { + "epoch": 0.47, + "learning_rate": 2.8611940298507464e-06, + "loss": 0.5442, + "step": 2592 + }, + { + "epoch": 0.47, + "learning_rate": 2.861139424827084e-06, + "loss": 3.1129, + "step": 2593 + }, + { + "epoch": 0.47, + "learning_rate": 2.861084819803422e-06, + "loss": 0.5985, + "step": 2594 + }, + { + "epoch": 0.47, + "learning_rate": 2.86103021477976e-06, + "loss": 0.7413, + "step": 2595 + }, + { + "epoch": 0.47, + "learning_rate": 2.8609756097560973e-06, + "loss": 1.9935, + "step": 2596 + }, + { + "epoch": 0.47, + "learning_rate": 2.8609210047324353e-06, + "loss": 1.2745, + "step": 2597 + }, + { + "epoch": 0.47, + "learning_rate": 2.8608663997087732e-06, + "loss": 1.0743, + "step": 2598 + }, + { + "epoch": 0.47, + "learning_rate": 2.860811794685111e-06, + "loss": 2.2323, + "step": 2599 + }, + { + "epoch": 0.47, + "learning_rate": 2.860757189661449e-06, + "loss": 2.3996, + "step": 2600 + }, + { + "epoch": 0.47, + "learning_rate": 2.860702584637787e-06, + "loss": 0.2725, + "step": 2601 + }, + { + "epoch": 0.47, + "learning_rate": 2.8606479796141246e-06, + "loss": 0.9866, + "step": 2602 + }, + { + "epoch": 0.47, + "learning_rate": 2.8605933745904625e-06, + "loss": 1.1105, + "step": 2603 + }, + { + "epoch": 0.47, + "learning_rate": 2.8605387695668005e-06, + "loss": 1.4054, + "step": 2604 + }, + { + "epoch": 0.47, + "learning_rate": 2.860484164543138e-06, + "loss": 1.0888, + "step": 2605 + }, + { + "epoch": 0.47, + "learning_rate": 2.860429559519476e-06, + "loss": 1.089, + "step": 2606 + }, + { + "epoch": 0.47, + "learning_rate": 2.8603749544958134e-06, + "loss": 0.6844, + "step": 2607 + }, + { + "epoch": 0.47, + "learning_rate": 2.8603203494721514e-06, + "loss": 3.3121, + "step": 2608 + }, + { + "epoch": 0.47, + "learning_rate": 2.8602657444484893e-06, + "loss": 1.942, + "step": 2609 + }, + { + "epoch": 0.47, + "learning_rate": 2.860211139424827e-06, + "loss": 0.73, + "step": 2610 + }, + { + "epoch": 0.47, + "learning_rate": 2.860156534401165e-06, + "loss": 0.3382, + "step": 2611 + }, + { + "epoch": 0.47, + "learning_rate": 2.8601019293775027e-06, + "loss": 1.0002, + "step": 2612 + }, + { + "epoch": 0.48, + "learning_rate": 2.8600473243538407e-06, + "loss": 0.3007, + "step": 2613 + }, + { + "epoch": 0.48, + "learning_rate": 2.8599927193301786e-06, + "loss": 1.4533, + "step": 2614 + }, + { + "epoch": 0.48, + "learning_rate": 2.8599381143065166e-06, + "loss": 0.5516, + "step": 2615 + }, + { + "epoch": 0.48, + "learning_rate": 2.859883509282854e-06, + "loss": 2.0758, + "step": 2616 + }, + { + "epoch": 0.48, + "learning_rate": 2.859828904259192e-06, + "loss": 0.6282, + "step": 2617 + }, + { + "epoch": 0.48, + "learning_rate": 2.8597742992355295e-06, + "loss": 0.8457, + "step": 2618 + }, + { + "epoch": 0.48, + "learning_rate": 2.8597196942118675e-06, + "loss": 1.4106, + "step": 2619 + }, + { + "epoch": 0.48, + "learning_rate": 2.8596650891882054e-06, + "loss": 0.0289, + "step": 2620 + }, + { + "epoch": 0.48, + "learning_rate": 2.859610484164543e-06, + "loss": 0.7282, + "step": 2621 + }, + { + "epoch": 0.48, + "learning_rate": 2.859555879140881e-06, + "loss": 0.5693, + "step": 2622 + }, + { + "epoch": 0.48, + "learning_rate": 2.859501274117219e-06, + "loss": 0.1037, + "step": 2623 + }, + { + "epoch": 0.48, + "learning_rate": 2.8594466690935564e-06, + "loss": 0.4649, + "step": 2624 + }, + { + "epoch": 0.48, + "learning_rate": 2.8593920640698943e-06, + "loss": 1.1107, + "step": 2625 + }, + { + "epoch": 0.48, + "learning_rate": 2.8593374590462322e-06, + "loss": 0.2124, + "step": 2626 + }, + { + "epoch": 0.48, + "learning_rate": 2.85928285402257e-06, + "loss": 0.592, + "step": 2627 + }, + { + "epoch": 0.48, + "learning_rate": 2.859228248998908e-06, + "loss": 0.4197, + "step": 2628 + }, + { + "epoch": 0.48, + "learning_rate": 2.859173643975246e-06, + "loss": 0.7195, + "step": 2629 + }, + { + "epoch": 0.48, + "learning_rate": 2.8591190389515836e-06, + "loss": 0.7964, + "step": 2630 + }, + { + "epoch": 0.48, + "learning_rate": 2.8590644339279215e-06, + "loss": 1.018, + "step": 2631 + }, + { + "epoch": 0.48, + "learning_rate": 2.859009828904259e-06, + "loss": 0.1193, + "step": 2632 + }, + { + "epoch": 0.48, + "learning_rate": 2.858955223880597e-06, + "loss": 1.0492, + "step": 2633 + }, + { + "epoch": 0.48, + "learning_rate": 2.858900618856935e-06, + "loss": 0.849, + "step": 2634 + }, + { + "epoch": 0.48, + "learning_rate": 2.8588460138332725e-06, + "loss": 1.6355, + "step": 2635 + }, + { + "epoch": 0.48, + "learning_rate": 2.8587914088096104e-06, + "loss": 0.5055, + "step": 2636 + }, + { + "epoch": 0.48, + "learning_rate": 2.8587368037859484e-06, + "loss": 1.4665, + "step": 2637 + }, + { + "epoch": 0.48, + "learning_rate": 2.8586821987622863e-06, + "loss": 2.3431, + "step": 2638 + }, + { + "epoch": 0.48, + "learning_rate": 2.8586275937386242e-06, + "loss": 1.0153, + "step": 2639 + }, + { + "epoch": 0.48, + "learning_rate": 2.858572988714962e-06, + "loss": 0.8245, + "step": 2640 + }, + { + "epoch": 0.48, + "learning_rate": 2.8585183836912997e-06, + "loss": 0.3654, + "step": 2641 + }, + { + "epoch": 0.48, + "learning_rate": 2.8584637786676376e-06, + "loss": 1.7205, + "step": 2642 + }, + { + "epoch": 0.48, + "learning_rate": 2.8584091736439756e-06, + "loss": 0.3757, + "step": 2643 + }, + { + "epoch": 0.48, + "learning_rate": 2.858354568620313e-06, + "loss": 0.5386, + "step": 2644 + }, + { + "epoch": 0.48, + "learning_rate": 2.858299963596651e-06, + "loss": 0.4116, + "step": 2645 + }, + { + "epoch": 0.48, + "learning_rate": 2.8582453585729886e-06, + "loss": 1.0755, + "step": 2646 + }, + { + "epoch": 0.48, + "learning_rate": 2.8581907535493265e-06, + "loss": 2.4595, + "step": 2647 + }, + { + "epoch": 0.48, + "learning_rate": 2.8581361485256645e-06, + "loss": 1.8995, + "step": 2648 + }, + { + "epoch": 0.48, + "learning_rate": 2.858081543502002e-06, + "loss": 0.1516, + "step": 2649 + }, + { + "epoch": 0.48, + "learning_rate": 2.85802693847834e-06, + "loss": 0.8998, + "step": 2650 + }, + { + "epoch": 0.48, + "learning_rate": 2.857972333454678e-06, + "loss": 1.214, + "step": 2651 + }, + { + "epoch": 0.48, + "learning_rate": 2.857917728431016e-06, + "loss": 0.2651, + "step": 2652 + }, + { + "epoch": 0.48, + "learning_rate": 2.8578631234073538e-06, + "loss": 0.5251, + "step": 2653 + }, + { + "epoch": 0.48, + "learning_rate": 2.8578085183836917e-06, + "loss": 0.0402, + "step": 2654 + }, + { + "epoch": 0.48, + "learning_rate": 2.8577539133600292e-06, + "loss": 0.0786, + "step": 2655 + }, + { + "epoch": 0.48, + "learning_rate": 2.857699308336367e-06, + "loss": 0.6064, + "step": 2656 + }, + { + "epoch": 0.48, + "learning_rate": 2.857644703312705e-06, + "loss": 0.0327, + "step": 2657 + }, + { + "epoch": 0.48, + "learning_rate": 2.8575900982890426e-06, + "loss": 1.0091, + "step": 2658 + }, + { + "epoch": 0.48, + "learning_rate": 2.8575354932653806e-06, + "loss": 0.0785, + "step": 2659 + }, + { + "epoch": 0.48, + "learning_rate": 2.857480888241718e-06, + "loss": 1.7804, + "step": 2660 + }, + { + "epoch": 0.48, + "learning_rate": 2.857426283218056e-06, + "loss": 1.2301, + "step": 2661 + }, + { + "epoch": 0.48, + "learning_rate": 2.857371678194394e-06, + "loss": 0.458, + "step": 2662 + }, + { + "epoch": 0.48, + "learning_rate": 2.8573170731707315e-06, + "loss": 1.9993, + "step": 2663 + }, + { + "epoch": 0.48, + "learning_rate": 2.8572624681470694e-06, + "loss": 1.4411, + "step": 2664 + }, + { + "epoch": 0.48, + "learning_rate": 2.8572078631234074e-06, + "loss": 0.6741, + "step": 2665 + }, + { + "epoch": 0.48, + "learning_rate": 2.8571532580997453e-06, + "loss": 0.1733, + "step": 2666 + }, + { + "epoch": 0.48, + "learning_rate": 2.8570986530760833e-06, + "loss": 0.0781, + "step": 2667 + }, + { + "epoch": 0.49, + "learning_rate": 2.8570440480524212e-06, + "loss": 0.2735, + "step": 2668 + }, + { + "epoch": 0.49, + "learning_rate": 2.8569894430287587e-06, + "loss": 0.9444, + "step": 2669 + }, + { + "epoch": 0.49, + "learning_rate": 2.8569348380050967e-06, + "loss": 1.975, + "step": 2670 + }, + { + "epoch": 0.49, + "learning_rate": 2.8568802329814346e-06, + "loss": 3.4754, + "step": 2671 + }, + { + "epoch": 0.49, + "learning_rate": 2.856825627957772e-06, + "loss": 0.7003, + "step": 2672 + }, + { + "epoch": 0.49, + "learning_rate": 2.85677102293411e-06, + "loss": 0.1157, + "step": 2673 + }, + { + "epoch": 0.49, + "learning_rate": 2.8567164179104476e-06, + "loss": 1.159, + "step": 2674 + }, + { + "epoch": 0.49, + "learning_rate": 2.8566618128867855e-06, + "loss": 1.3351, + "step": 2675 + }, + { + "epoch": 0.49, + "learning_rate": 2.8566072078631235e-06, + "loss": 1.0145, + "step": 2676 + }, + { + "epoch": 0.49, + "learning_rate": 2.856552602839461e-06, + "loss": 0.7229, + "step": 2677 + }, + { + "epoch": 0.49, + "learning_rate": 2.856497997815799e-06, + "loss": 1.8494, + "step": 2678 + }, + { + "epoch": 0.49, + "learning_rate": 2.856443392792137e-06, + "loss": 2.3045, + "step": 2679 + }, + { + "epoch": 0.49, + "learning_rate": 2.856388787768475e-06, + "loss": 0.6958, + "step": 2680 + }, + { + "epoch": 0.49, + "learning_rate": 2.8563341827448128e-06, + "loss": 0.5539, + "step": 2681 + }, + { + "epoch": 0.49, + "learning_rate": 2.8562795777211507e-06, + "loss": 0.7375, + "step": 2682 + }, + { + "epoch": 0.49, + "learning_rate": 2.8562249726974882e-06, + "loss": 2.0095, + "step": 2683 + }, + { + "epoch": 0.49, + "learning_rate": 2.856170367673826e-06, + "loss": 1.8744, + "step": 2684 + }, + { + "epoch": 0.49, + "learning_rate": 2.856115762650164e-06, + "loss": 0.5394, + "step": 2685 + }, + { + "epoch": 0.49, + "learning_rate": 2.8560611576265017e-06, + "loss": 2.8054, + "step": 2686 + }, + { + "epoch": 0.49, + "learning_rate": 2.8560065526028396e-06, + "loss": 1.6477, + "step": 2687 + }, + { + "epoch": 0.49, + "learning_rate": 2.855951947579177e-06, + "loss": 0.8844, + "step": 2688 + }, + { + "epoch": 0.49, + "learning_rate": 2.855897342555515e-06, + "loss": 0.781, + "step": 2689 + }, + { + "epoch": 0.49, + "learning_rate": 2.855842737531853e-06, + "loss": 0.5745, + "step": 2690 + }, + { + "epoch": 0.49, + "learning_rate": 2.8557881325081905e-06, + "loss": 0.6307, + "step": 2691 + }, + { + "epoch": 0.49, + "learning_rate": 2.8557335274845285e-06, + "loss": 2.9276, + "step": 2692 + }, + { + "epoch": 0.49, + "learning_rate": 2.8556789224608664e-06, + "loss": 0.7511, + "step": 2693 + }, + { + "epoch": 0.49, + "learning_rate": 2.8556243174372044e-06, + "loss": 0.9215, + "step": 2694 + }, + { + "epoch": 0.49, + "learning_rate": 2.8555697124135423e-06, + "loss": 0.6897, + "step": 2695 + }, + { + "epoch": 0.49, + "learning_rate": 2.8555151073898802e-06, + "loss": 2.5875, + "step": 2696 + }, + { + "epoch": 0.49, + "learning_rate": 2.8554605023662178e-06, + "loss": 2.0374, + "step": 2697 + }, + { + "epoch": 0.49, + "learning_rate": 2.8554058973425557e-06, + "loss": 0.6195, + "step": 2698 + }, + { + "epoch": 0.49, + "learning_rate": 2.8553512923188932e-06, + "loss": 0.85, + "step": 2699 + }, + { + "epoch": 0.49, + "learning_rate": 2.855296687295231e-06, + "loss": 0.4146, + "step": 2700 + }, + { + "epoch": 0.49, + "learning_rate": 2.855242082271569e-06, + "loss": 2.8695, + "step": 2701 + }, + { + "epoch": 0.49, + "learning_rate": 2.8551874772479066e-06, + "loss": 0.6846, + "step": 2702 + }, + { + "epoch": 0.49, + "learning_rate": 2.8551328722242446e-06, + "loss": 2.3231, + "step": 2703 + }, + { + "epoch": 0.49, + "learning_rate": 2.8550782672005825e-06, + "loss": 0.5139, + "step": 2704 + }, + { + "epoch": 0.49, + "learning_rate": 2.85502366217692e-06, + "loss": 1.8623, + "step": 2705 + }, + { + "epoch": 0.49, + "learning_rate": 2.854969057153258e-06, + "loss": 1.0533, + "step": 2706 + }, + { + "epoch": 0.49, + "learning_rate": 2.854914452129596e-06, + "loss": 0.9518, + "step": 2707 + }, + { + "epoch": 0.49, + "learning_rate": 2.854859847105934e-06, + "loss": 0.7597, + "step": 2708 + }, + { + "epoch": 0.49, + "learning_rate": 2.854805242082272e-06, + "loss": 0.4245, + "step": 2709 + }, + { + "epoch": 0.49, + "learning_rate": 2.8547506370586098e-06, + "loss": 0.7161, + "step": 2710 + }, + { + "epoch": 0.49, + "learning_rate": 2.8546960320349473e-06, + "loss": 1.9949, + "step": 2711 + }, + { + "epoch": 0.49, + "learning_rate": 2.8546414270112852e-06, + "loss": 1.4351, + "step": 2712 + }, + { + "epoch": 0.49, + "learning_rate": 2.8545868219876227e-06, + "loss": 0.8493, + "step": 2713 + }, + { + "epoch": 0.49, + "learning_rate": 2.8545322169639607e-06, + "loss": 0.4828, + "step": 2714 + }, + { + "epoch": 0.49, + "learning_rate": 2.8544776119402986e-06, + "loss": 0.6623, + "step": 2715 + }, + { + "epoch": 0.49, + "learning_rate": 2.854423006916636e-06, + "loss": 0.944, + "step": 2716 + }, + { + "epoch": 0.49, + "learning_rate": 2.854368401892974e-06, + "loss": 0.7844, + "step": 2717 + }, + { + "epoch": 0.49, + "learning_rate": 2.854313796869312e-06, + "loss": 0.2617, + "step": 2718 + }, + { + "epoch": 0.49, + "learning_rate": 2.85425919184565e-06, + "loss": 0.284, + "step": 2719 + }, + { + "epoch": 0.49, + "learning_rate": 2.8542045868219875e-06, + "loss": 0.107, + "step": 2720 + }, + { + "epoch": 0.49, + "learning_rate": 2.8541499817983254e-06, + "loss": 0.6964, + "step": 2721 + }, + { + "epoch": 0.49, + "learning_rate": 2.8540953767746634e-06, + "loss": 0.1074, + "step": 2722 + }, + { + "epoch": 0.5, + "learning_rate": 2.8540407717510013e-06, + "loss": 2.1224, + "step": 2723 + }, + { + "epoch": 0.5, + "learning_rate": 2.8539861667273393e-06, + "loss": 1.0612, + "step": 2724 + }, + { + "epoch": 0.5, + "learning_rate": 2.8539315617036768e-06, + "loss": 0.5694, + "step": 2725 + }, + { + "epoch": 0.5, + "learning_rate": 2.8538769566800147e-06, + "loss": 0.9835, + "step": 2726 + }, + { + "epoch": 0.5, + "learning_rate": 2.8538223516563523e-06, + "loss": 1.7971, + "step": 2727 + }, + { + "epoch": 0.5, + "learning_rate": 2.85376774663269e-06, + "loss": 0.7337, + "step": 2728 + }, + { + "epoch": 0.5, + "learning_rate": 2.853713141609028e-06, + "loss": 0.0787, + "step": 2729 + }, + { + "epoch": 0.5, + "learning_rate": 2.8536585365853657e-06, + "loss": 0.4947, + "step": 2730 + }, + { + "epoch": 0.5, + "learning_rate": 2.8536039315617036e-06, + "loss": 1.3957, + "step": 2731 + }, + { + "epoch": 0.5, + "learning_rate": 2.8535493265380415e-06, + "loss": 0.626, + "step": 2732 + }, + { + "epoch": 0.5, + "learning_rate": 2.8534947215143795e-06, + "loss": 0.8488, + "step": 2733 + }, + { + "epoch": 0.5, + "learning_rate": 2.8534401164907174e-06, + "loss": 2.7337, + "step": 2734 + }, + { + "epoch": 0.5, + "learning_rate": 2.8533855114670554e-06, + "loss": 0.4379, + "step": 2735 + }, + { + "epoch": 0.5, + "learning_rate": 2.853330906443393e-06, + "loss": 0.8196, + "step": 2736 + }, + { + "epoch": 0.5, + "learning_rate": 2.853276301419731e-06, + "loss": 2.2913, + "step": 2737 + }, + { + "epoch": 0.5, + "learning_rate": 2.8532216963960688e-06, + "loss": 1.9653, + "step": 2738 + }, + { + "epoch": 0.5, + "learning_rate": 2.8531670913724063e-06, + "loss": 1.2113, + "step": 2739 + }, + { + "epoch": 0.5, + "learning_rate": 2.8531124863487442e-06, + "loss": 2.0365, + "step": 2740 + }, + { + "epoch": 0.5, + "learning_rate": 2.8530578813250818e-06, + "loss": 0.2032, + "step": 2741 + }, + { + "epoch": 0.5, + "learning_rate": 2.8530032763014197e-06, + "loss": 2.2774, + "step": 2742 + }, + { + "epoch": 0.5, + "learning_rate": 2.8529486712777577e-06, + "loss": 1.108, + "step": 2743 + }, + { + "epoch": 0.5, + "learning_rate": 2.852894066254095e-06, + "loss": 0.7229, + "step": 2744 + }, + { + "epoch": 0.5, + "learning_rate": 2.852839461230433e-06, + "loss": 1.6787, + "step": 2745 + }, + { + "epoch": 0.5, + "learning_rate": 2.852784856206771e-06, + "loss": 1.2307, + "step": 2746 + }, + { + "epoch": 0.5, + "learning_rate": 2.852730251183109e-06, + "loss": 1.9725, + "step": 2747 + }, + { + "epoch": 0.5, + "learning_rate": 2.852675646159447e-06, + "loss": 2.7597, + "step": 2748 + }, + { + "epoch": 0.5, + "learning_rate": 2.852621041135785e-06, + "loss": 0.6227, + "step": 2749 + }, + { + "epoch": 0.5, + "learning_rate": 2.8525664361121224e-06, + "loss": 2.4038, + "step": 2750 + }, + { + "epoch": 0.5, + "learning_rate": 2.8525118310884604e-06, + "loss": 1.3008, + "step": 2751 + }, + { + "epoch": 0.5, + "learning_rate": 2.8524572260647983e-06, + "loss": 0.7935, + "step": 2752 + }, + { + "epoch": 0.5, + "learning_rate": 2.852402621041136e-06, + "loss": 2.2619, + "step": 2753 + }, + { + "epoch": 0.5, + "learning_rate": 2.8523480160174738e-06, + "loss": 0.7768, + "step": 2754 + }, + { + "epoch": 0.5, + "learning_rate": 2.8522934109938113e-06, + "loss": 2.5855, + "step": 2755 + }, + { + "epoch": 0.5, + "learning_rate": 2.8522388059701492e-06, + "loss": 0.7743, + "step": 2756 + }, + { + "epoch": 0.5, + "learning_rate": 2.852184200946487e-06, + "loss": 0.9098, + "step": 2757 + }, + { + "epoch": 0.5, + "learning_rate": 2.8521295959228247e-06, + "loss": 0.9097, + "step": 2758 + }, + { + "epoch": 0.5, + "learning_rate": 2.8520749908991626e-06, + "loss": 3.0021, + "step": 2759 + }, + { + "epoch": 0.5, + "learning_rate": 2.8520203858755006e-06, + "loss": 0.6059, + "step": 2760 + }, + { + "epoch": 0.5, + "learning_rate": 2.8519657808518385e-06, + "loss": 1.6598, + "step": 2761 + }, + { + "epoch": 0.5, + "learning_rate": 2.8519111758281765e-06, + "loss": 0.0563, + "step": 2762 + }, + { + "epoch": 0.5, + "learning_rate": 2.8518565708045144e-06, + "loss": 1.3721, + "step": 2763 + }, + { + "epoch": 0.5, + "learning_rate": 2.851801965780852e-06, + "loss": 0.5809, + "step": 2764 + }, + { + "epoch": 0.5, + "learning_rate": 2.85174736075719e-06, + "loss": 0.0249, + "step": 2765 + }, + { + "epoch": 0.5, + "learning_rate": 2.8516927557335274e-06, + "loss": 2.187, + "step": 2766 + }, + { + "epoch": 0.5, + "learning_rate": 2.8516381507098653e-06, + "loss": 1.8244, + "step": 2767 + }, + { + "epoch": 0.5, + "learning_rate": 2.8515835456862033e-06, + "loss": 0.2847, + "step": 2768 + }, + { + "epoch": 0.5, + "learning_rate": 2.851528940662541e-06, + "loss": 1.191, + "step": 2769 + }, + { + "epoch": 0.5, + "learning_rate": 2.8514743356388787e-06, + "loss": 1.6063, + "step": 2770 + }, + { + "epoch": 0.5, + "learning_rate": 2.8514197306152167e-06, + "loss": 3.2978, + "step": 2771 + }, + { + "epoch": 0.5, + "learning_rate": 2.851365125591554e-06, + "loss": 0.9738, + "step": 2772 + }, + { + "epoch": 0.5, + "learning_rate": 2.851310520567892e-06, + "loss": 0.6252, + "step": 2773 + }, + { + "epoch": 0.5, + "learning_rate": 2.85125591554423e-06, + "loss": 0.8848, + "step": 2774 + }, + { + "epoch": 0.5, + "learning_rate": 2.851201310520568e-06, + "loss": 1.1555, + "step": 2775 + }, + { + "epoch": 0.5, + "learning_rate": 2.851146705496906e-06, + "loss": 0.9026, + "step": 2776 + }, + { + "epoch": 0.51, + "learning_rate": 2.851092100473244e-06, + "loss": 1.546, + "step": 2777 + }, + { + "epoch": 0.51, + "learning_rate": 2.8510374954495814e-06, + "loss": 0.5513, + "step": 2778 + }, + { + "epoch": 0.51, + "learning_rate": 2.8509828904259194e-06, + "loss": 0.7481, + "step": 2779 + }, + { + "epoch": 0.51, + "learning_rate": 2.850928285402257e-06, + "loss": 0.5746, + "step": 2780 + }, + { + "epoch": 0.51, + "learning_rate": 2.850873680378595e-06, + "loss": 0.9614, + "step": 2781 + }, + { + "epoch": 0.51, + "learning_rate": 2.8508190753549328e-06, + "loss": 2.9812, + "step": 2782 + }, + { + "epoch": 0.51, + "learning_rate": 2.8507644703312703e-06, + "loss": 1.8138, + "step": 2783 + }, + { + "epoch": 0.51, + "learning_rate": 2.8507098653076082e-06, + "loss": 0.9346, + "step": 2784 + }, + { + "epoch": 0.51, + "learning_rate": 2.850655260283946e-06, + "loss": 1.2705, + "step": 2785 + }, + { + "epoch": 0.51, + "learning_rate": 2.8506006552602837e-06, + "loss": 1.534, + "step": 2786 + }, + { + "epoch": 0.51, + "learning_rate": 2.8505460502366217e-06, + "loss": 0.1603, + "step": 2787 + }, + { + "epoch": 0.51, + "learning_rate": 2.8504914452129596e-06, + "loss": 0.6995, + "step": 2788 + }, + { + "epoch": 0.51, + "learning_rate": 2.8504368401892975e-06, + "loss": 0.2278, + "step": 2789 + }, + { + "epoch": 0.51, + "learning_rate": 2.8503822351656355e-06, + "loss": 1.0037, + "step": 2790 + }, + { + "epoch": 0.51, + "learning_rate": 2.8503276301419734e-06, + "loss": 1.5099, + "step": 2791 + }, + { + "epoch": 0.51, + "learning_rate": 2.850273025118311e-06, + "loss": 0.0699, + "step": 2792 + }, + { + "epoch": 0.51, + "learning_rate": 2.850218420094649e-06, + "loss": 2.3805, + "step": 2793 + }, + { + "epoch": 0.51, + "learning_rate": 2.8501638150709864e-06, + "loss": 0.2978, + "step": 2794 + }, + { + "epoch": 0.51, + "learning_rate": 2.8501092100473244e-06, + "loss": 0.0639, + "step": 2795 + }, + { + "epoch": 0.51, + "learning_rate": 2.8500546050236623e-06, + "loss": 1.5005, + "step": 2796 + }, + { + "epoch": 0.51, + "learning_rate": 2.85e-06, + "loss": 0.9682, + "step": 2797 + }, + { + "epoch": 0.51, + "learning_rate": 2.8499453949763378e-06, + "loss": 2.5698, + "step": 2798 + }, + { + "epoch": 0.51, + "learning_rate": 2.8498907899526757e-06, + "loss": 0.5922, + "step": 2799 + }, + { + "epoch": 0.51, + "learning_rate": 2.8498361849290132e-06, + "loss": 2.0674, + "step": 2800 + }, + { + "epoch": 0.51, + "learning_rate": 2.849781579905351e-06, + "loss": 0.5929, + "step": 2801 + }, + { + "epoch": 0.51, + "learning_rate": 2.849726974881689e-06, + "loss": 0.5906, + "step": 2802 + }, + { + "epoch": 0.51, + "learning_rate": 2.849672369858027e-06, + "loss": 1.2728, + "step": 2803 + }, + { + "epoch": 0.51, + "learning_rate": 2.849617764834365e-06, + "loss": 1.5922, + "step": 2804 + }, + { + "epoch": 0.51, + "learning_rate": 2.849563159810703e-06, + "loss": 0.5008, + "step": 2805 + }, + { + "epoch": 0.51, + "learning_rate": 2.8495085547870405e-06, + "loss": 1.6708, + "step": 2806 + }, + { + "epoch": 0.51, + "learning_rate": 2.8494539497633784e-06, + "loss": 0.9471, + "step": 2807 + }, + { + "epoch": 0.51, + "learning_rate": 2.849399344739716e-06, + "loss": 0.0682, + "step": 2808 + }, + { + "epoch": 0.51, + "learning_rate": 2.849344739716054e-06, + "loss": 0.2869, + "step": 2809 + }, + { + "epoch": 0.51, + "learning_rate": 2.849290134692392e-06, + "loss": 0.4895, + "step": 2810 + }, + { + "epoch": 0.51, + "learning_rate": 2.8492355296687293e-06, + "loss": 0.7694, + "step": 2811 + }, + { + "epoch": 0.51, + "learning_rate": 2.8491809246450673e-06, + "loss": 0.1011, + "step": 2812 + }, + { + "epoch": 0.51, + "learning_rate": 2.8491263196214052e-06, + "loss": 0.1126, + "step": 2813 + }, + { + "epoch": 0.51, + "learning_rate": 2.849071714597743e-06, + "loss": 1.5194, + "step": 2814 + }, + { + "epoch": 0.51, + "learning_rate": 2.849017109574081e-06, + "loss": 1.7318, + "step": 2815 + }, + { + "epoch": 0.51, + "learning_rate": 2.848962504550419e-06, + "loss": 0.656, + "step": 2816 + }, + { + "epoch": 0.51, + "learning_rate": 2.8489078995267566e-06, + "loss": 1.1603, + "step": 2817 + }, + { + "epoch": 0.51, + "learning_rate": 2.8488532945030945e-06, + "loss": 1.1, + "step": 2818 + }, + { + "epoch": 0.51, + "learning_rate": 2.8487986894794325e-06, + "loss": 1.2658, + "step": 2819 + }, + { + "epoch": 0.51, + "learning_rate": 2.84874408445577e-06, + "loss": 2.3474, + "step": 2820 + }, + { + "epoch": 0.51, + "learning_rate": 2.848689479432108e-06, + "loss": 1.6783, + "step": 2821 + }, + { + "epoch": 0.51, + "learning_rate": 2.8486348744084454e-06, + "loss": 1.0172, + "step": 2822 + }, + { + "epoch": 0.51, + "learning_rate": 2.8485802693847834e-06, + "loss": 1.26, + "step": 2823 + }, + { + "epoch": 0.51, + "learning_rate": 2.8485256643611213e-06, + "loss": 0.7922, + "step": 2824 + }, + { + "epoch": 0.51, + "learning_rate": 2.848471059337459e-06, + "loss": 1.7822, + "step": 2825 + }, + { + "epoch": 0.51, + "learning_rate": 2.848416454313797e-06, + "loss": 1.2264, + "step": 2826 + }, + { + "epoch": 0.51, + "learning_rate": 2.8483618492901347e-06, + "loss": 3.1196, + "step": 2827 + }, + { + "epoch": 0.51, + "learning_rate": 2.8483072442664727e-06, + "loss": 0.7239, + "step": 2828 + }, + { + "epoch": 0.51, + "learning_rate": 2.8482526392428106e-06, + "loss": 0.029, + "step": 2829 + }, + { + "epoch": 0.51, + "learning_rate": 2.8481980342191486e-06, + "loss": 1.8625, + "step": 2830 + }, + { + "epoch": 0.51, + "learning_rate": 2.848143429195486e-06, + "loss": 0.2837, + "step": 2831 + }, + { + "epoch": 0.52, + "learning_rate": 2.848088824171824e-06, + "loss": 0.2578, + "step": 2832 + }, + { + "epoch": 0.52, + "learning_rate": 2.848034219148162e-06, + "loss": 1.377, + "step": 2833 + }, + { + "epoch": 0.52, + "learning_rate": 2.8479796141244995e-06, + "loss": 0.5938, + "step": 2834 + }, + { + "epoch": 0.52, + "learning_rate": 2.8479250091008374e-06, + "loss": 1.4054, + "step": 2835 + }, + { + "epoch": 0.52, + "learning_rate": 2.847870404077175e-06, + "loss": 0.4145, + "step": 2836 + }, + { + "epoch": 0.52, + "learning_rate": 2.847815799053513e-06, + "loss": 1.9536, + "step": 2837 + }, + { + "epoch": 0.52, + "learning_rate": 2.847761194029851e-06, + "loss": 1.3715, + "step": 2838 + }, + { + "epoch": 0.52, + "learning_rate": 2.8477065890061884e-06, + "loss": 0.2831, + "step": 2839 + }, + { + "epoch": 0.52, + "learning_rate": 2.8476519839825263e-06, + "loss": 0.6242, + "step": 2840 + }, + { + "epoch": 0.52, + "learning_rate": 2.8475973789588642e-06, + "loss": 0.5807, + "step": 2841 + }, + { + "epoch": 0.52, + "learning_rate": 2.847542773935202e-06, + "loss": 0.9221, + "step": 2842 + }, + { + "epoch": 0.52, + "learning_rate": 2.84748816891154e-06, + "loss": 0.1365, + "step": 2843 + }, + { + "epoch": 0.52, + "learning_rate": 2.847433563887878e-06, + "loss": 1.3774, + "step": 2844 + }, + { + "epoch": 0.52, + "learning_rate": 2.8473789588642156e-06, + "loss": 1.0591, + "step": 2845 + }, + { + "epoch": 0.52, + "learning_rate": 2.8473243538405535e-06, + "loss": 0.4638, + "step": 2846 + }, + { + "epoch": 0.52, + "learning_rate": 2.847269748816891e-06, + "loss": 0.5403, + "step": 2847 + }, + { + "epoch": 0.52, + "learning_rate": 2.847215143793229e-06, + "loss": 0.0987, + "step": 2848 + }, + { + "epoch": 0.52, + "learning_rate": 2.847160538769567e-06, + "loss": 1.0072, + "step": 2849 + }, + { + "epoch": 0.52, + "learning_rate": 2.8471059337459045e-06, + "loss": 2.1498, + "step": 2850 + }, + { + "epoch": 0.52, + "learning_rate": 2.8470513287222424e-06, + "loss": 1.7988, + "step": 2851 + }, + { + "epoch": 0.52, + "learning_rate": 2.8469967236985804e-06, + "loss": 1.3296, + "step": 2852 + }, + { + "epoch": 0.52, + "learning_rate": 2.846942118674918e-06, + "loss": 0.597, + "step": 2853 + }, + { + "epoch": 0.52, + "learning_rate": 2.846887513651256e-06, + "loss": 4.11, + "step": 2854 + }, + { + "epoch": 0.52, + "learning_rate": 2.8468329086275938e-06, + "loss": 1.5975, + "step": 2855 + }, + { + "epoch": 0.52, + "learning_rate": 2.8467783036039317e-06, + "loss": 1.9319, + "step": 2856 + }, + { + "epoch": 0.52, + "learning_rate": 2.8467236985802696e-06, + "loss": 1.4613, + "step": 2857 + }, + { + "epoch": 0.52, + "learning_rate": 2.8466690935566076e-06, + "loss": 0.6326, + "step": 2858 + }, + { + "epoch": 0.52, + "learning_rate": 2.846614488532945e-06, + "loss": 0.5084, + "step": 2859 + }, + { + "epoch": 0.52, + "learning_rate": 2.846559883509283e-06, + "loss": 0.6582, + "step": 2860 + }, + { + "epoch": 0.52, + "learning_rate": 2.8465052784856206e-06, + "loss": 1.5776, + "step": 2861 + }, + { + "epoch": 0.52, + "learning_rate": 2.8464506734619585e-06, + "loss": 2.8041, + "step": 2862 + }, + { + "epoch": 0.52, + "learning_rate": 2.8463960684382965e-06, + "loss": 0.9002, + "step": 2863 + }, + { + "epoch": 0.52, + "learning_rate": 2.846341463414634e-06, + "loss": 1.2089, + "step": 2864 + }, + { + "epoch": 0.52, + "learning_rate": 2.846286858390972e-06, + "loss": 0.8984, + "step": 2865 + }, + { + "epoch": 0.52, + "learning_rate": 2.84623225336731e-06, + "loss": 1.6464, + "step": 2866 + }, + { + "epoch": 0.52, + "learning_rate": 2.8461776483436474e-06, + "loss": 1.4772, + "step": 2867 + }, + { + "epoch": 0.52, + "learning_rate": 2.8461230433199853e-06, + "loss": 0.6643, + "step": 2868 + }, + { + "epoch": 0.52, + "learning_rate": 2.8460684382963233e-06, + "loss": 1.1525, + "step": 2869 + }, + { + "epoch": 0.52, + "learning_rate": 2.8460138332726612e-06, + "loss": 4.6629, + "step": 2870 + }, + { + "epoch": 0.52, + "learning_rate": 2.845959228248999e-06, + "loss": 0.7945, + "step": 2871 + }, + { + "epoch": 0.52, + "learning_rate": 2.845904623225337e-06, + "loss": 1.3119, + "step": 2872 + }, + { + "epoch": 0.52, + "learning_rate": 2.8458500182016746e-06, + "loss": 1.5248, + "step": 2873 + }, + { + "epoch": 0.52, + "learning_rate": 2.8457954131780126e-06, + "loss": 0.0702, + "step": 2874 + }, + { + "epoch": 0.52, + "learning_rate": 2.84574080815435e-06, + "loss": 0.7451, + "step": 2875 + }, + { + "epoch": 0.52, + "learning_rate": 2.845686203130688e-06, + "loss": 1.3752, + "step": 2876 + }, + { + "epoch": 0.52, + "learning_rate": 2.845631598107026e-06, + "loss": 1.3107, + "step": 2877 + }, + { + "epoch": 0.52, + "learning_rate": 2.8455769930833635e-06, + "loss": 0.5256, + "step": 2878 + }, + { + "epoch": 0.52, + "learning_rate": 2.8455223880597014e-06, + "loss": 0.1103, + "step": 2879 + }, + { + "epoch": 0.52, + "learning_rate": 2.8454677830360394e-06, + "loss": 0.6464, + "step": 2880 + }, + { + "epoch": 0.52, + "learning_rate": 2.845413178012377e-06, + "loss": 0.9704, + "step": 2881 + }, + { + "epoch": 0.52, + "learning_rate": 2.845358572988715e-06, + "loss": 0.5771, + "step": 2882 + }, + { + "epoch": 0.52, + "learning_rate": 2.8453039679650528e-06, + "loss": 3.1016, + "step": 2883 + }, + { + "epoch": 0.52, + "learning_rate": 2.8452493629413907e-06, + "loss": 2.1441, + "step": 2884 + }, + { + "epoch": 0.52, + "learning_rate": 2.8451947579177287e-06, + "loss": 0.9097, + "step": 2885 + }, + { + "epoch": 0.52, + "learning_rate": 2.8451401528940666e-06, + "loss": 0.0725, + "step": 2886 + }, + { + "epoch": 0.53, + "learning_rate": 2.845085547870404e-06, + "loss": 0.3238, + "step": 2887 + }, + { + "epoch": 0.53, + "learning_rate": 2.845030942846742e-06, + "loss": 0.3566, + "step": 2888 + }, + { + "epoch": 0.53, + "learning_rate": 2.8449763378230796e-06, + "loss": 0.7621, + "step": 2889 + }, + { + "epoch": 0.53, + "learning_rate": 2.8449217327994175e-06, + "loss": 2.0395, + "step": 2890 + }, + { + "epoch": 0.53, + "learning_rate": 2.8448671277757555e-06, + "loss": 1.0002, + "step": 2891 + }, + { + "epoch": 0.53, + "learning_rate": 2.844812522752093e-06, + "loss": 0.6759, + "step": 2892 + }, + { + "epoch": 0.53, + "learning_rate": 2.844757917728431e-06, + "loss": 0.3774, + "step": 2893 + }, + { + "epoch": 0.53, + "learning_rate": 2.844703312704769e-06, + "loss": 0.6321, + "step": 2894 + }, + { + "epoch": 0.53, + "learning_rate": 2.844648707681107e-06, + "loss": 1.7814, + "step": 2895 + }, + { + "epoch": 0.53, + "learning_rate": 2.8445941026574448e-06, + "loss": 1.9451, + "step": 2896 + }, + { + "epoch": 0.53, + "learning_rate": 2.8445394976337827e-06, + "loss": 1.1482, + "step": 2897 + }, + { + "epoch": 0.53, + "learning_rate": 2.8444848926101202e-06, + "loss": 0.914, + "step": 2898 + }, + { + "epoch": 0.53, + "learning_rate": 2.844430287586458e-06, + "loss": 1.5184, + "step": 2899 + }, + { + "epoch": 0.53, + "learning_rate": 2.844375682562796e-06, + "loss": 1.214, + "step": 2900 + }, + { + "epoch": 0.53, + "learning_rate": 2.8443210775391336e-06, + "loss": 0.2934, + "step": 2901 + }, + { + "epoch": 0.53, + "learning_rate": 2.8442664725154716e-06, + "loss": 0.2603, + "step": 2902 + }, + { + "epoch": 0.53, + "learning_rate": 2.844211867491809e-06, + "loss": 0.4118, + "step": 2903 + }, + { + "epoch": 0.53, + "learning_rate": 2.844157262468147e-06, + "loss": 0.5681, + "step": 2904 + }, + { + "epoch": 0.53, + "learning_rate": 2.844102657444485e-06, + "loss": 2.0221, + "step": 2905 + }, + { + "epoch": 0.53, + "learning_rate": 2.8440480524208225e-06, + "loss": 0.9554, + "step": 2906 + }, + { + "epoch": 0.53, + "learning_rate": 2.8439934473971605e-06, + "loss": 0.5132, + "step": 2907 + }, + { + "epoch": 0.53, + "learning_rate": 2.8439388423734984e-06, + "loss": 2.0912, + "step": 2908 + }, + { + "epoch": 0.53, + "learning_rate": 2.8438842373498363e-06, + "loss": 1.2735, + "step": 2909 + }, + { + "epoch": 0.53, + "learning_rate": 2.8438296323261743e-06, + "loss": 0.4611, + "step": 2910 + }, + { + "epoch": 0.53, + "learning_rate": 2.8437750273025122e-06, + "loss": 1.4192, + "step": 2911 + }, + { + "epoch": 0.53, + "learning_rate": 2.8437204222788498e-06, + "loss": 0.4289, + "step": 2912 + }, + { + "epoch": 0.53, + "learning_rate": 2.8436658172551877e-06, + "loss": 2.2922, + "step": 2913 + }, + { + "epoch": 0.53, + "learning_rate": 2.8436112122315256e-06, + "loss": 2.7495, + "step": 2914 + }, + { + "epoch": 0.53, + "learning_rate": 2.843556607207863e-06, + "loss": 0.5199, + "step": 2915 + }, + { + "epoch": 0.53, + "learning_rate": 2.843502002184201e-06, + "loss": 0.9381, + "step": 2916 + }, + { + "epoch": 0.53, + "learning_rate": 2.8434473971605386e-06, + "loss": 1.2276, + "step": 2917 + }, + { + "epoch": 0.53, + "learning_rate": 2.8433927921368766e-06, + "loss": 1.1375, + "step": 2918 + }, + { + "epoch": 0.53, + "learning_rate": 2.8433381871132145e-06, + "loss": 0.8615, + "step": 2919 + }, + { + "epoch": 0.53, + "learning_rate": 2.843283582089552e-06, + "loss": 0.2552, + "step": 2920 + }, + { + "epoch": 0.53, + "learning_rate": 2.84322897706589e-06, + "loss": 0.6254, + "step": 2921 + }, + { + "epoch": 0.53, + "learning_rate": 2.843174372042228e-06, + "loss": 2.342, + "step": 2922 + }, + { + "epoch": 0.53, + "learning_rate": 2.843119767018566e-06, + "loss": 0.4029, + "step": 2923 + }, + { + "epoch": 0.53, + "learning_rate": 2.843065161994904e-06, + "loss": 2.0754, + "step": 2924 + }, + { + "epoch": 0.53, + "learning_rate": 2.8430105569712417e-06, + "loss": 1.256, + "step": 2925 + }, + { + "epoch": 0.53, + "learning_rate": 2.8429559519475793e-06, + "loss": 1.743, + "step": 2926 + }, + { + "epoch": 0.53, + "learning_rate": 2.8429013469239172e-06, + "loss": 0.0887, + "step": 2927 + }, + { + "epoch": 0.53, + "learning_rate": 2.8428467419002547e-06, + "loss": 0.9894, + "step": 2928 + }, + { + "epoch": 0.53, + "learning_rate": 2.8427921368765927e-06, + "loss": 0.8945, + "step": 2929 + }, + { + "epoch": 0.53, + "learning_rate": 2.8427375318529306e-06, + "loss": 1.0849, + "step": 2930 + }, + { + "epoch": 0.53, + "learning_rate": 2.842682926829268e-06, + "loss": 0.5719, + "step": 2931 + }, + { + "epoch": 0.53, + "learning_rate": 2.842628321805606e-06, + "loss": 0.7726, + "step": 2932 + }, + { + "epoch": 0.53, + "learning_rate": 2.842573716781944e-06, + "loss": 0.3946, + "step": 2933 + }, + { + "epoch": 0.53, + "learning_rate": 2.8425191117582815e-06, + "loss": 0.5927, + "step": 2934 + }, + { + "epoch": 0.53, + "learning_rate": 2.8424645067346195e-06, + "loss": 0.5114, + "step": 2935 + }, + { + "epoch": 0.53, + "learning_rate": 2.8424099017109574e-06, + "loss": 0.9124, + "step": 2936 + }, + { + "epoch": 0.53, + "learning_rate": 2.8423552966872954e-06, + "loss": 1.6641, + "step": 2937 + }, + { + "epoch": 0.53, + "learning_rate": 2.8423006916636333e-06, + "loss": 0.3355, + "step": 2938 + }, + { + "epoch": 0.53, + "learning_rate": 2.8422460866399713e-06, + "loss": 0.412, + "step": 2939 + }, + { + "epoch": 0.53, + "learning_rate": 2.8421914816163088e-06, + "loss": 0.493, + "step": 2940 + }, + { + "epoch": 0.53, + "learning_rate": 2.8421368765926467e-06, + "loss": 1.8725, + "step": 2941 + }, + { + "epoch": 0.54, + "learning_rate": 2.8420822715689842e-06, + "loss": 1.408, + "step": 2942 + }, + { + "epoch": 0.54, + "learning_rate": 2.842027666545322e-06, + "loss": 0.5775, + "step": 2943 + }, + { + "epoch": 0.54, + "learning_rate": 2.84197306152166e-06, + "loss": 1.9408, + "step": 2944 + }, + { + "epoch": 0.54, + "learning_rate": 2.8419184564979977e-06, + "loss": 0.8108, + "step": 2945 + }, + { + "epoch": 0.54, + "learning_rate": 2.8418638514743356e-06, + "loss": 0.8337, + "step": 2946 + }, + { + "epoch": 0.54, + "learning_rate": 2.8418092464506735e-06, + "loss": 1.1009, + "step": 2947 + }, + { + "epoch": 0.54, + "learning_rate": 2.841754641427011e-06, + "loss": 0.3164, + "step": 2948 + }, + { + "epoch": 0.54, + "learning_rate": 2.841700036403349e-06, + "loss": 1.2397, + "step": 2949 + }, + { + "epoch": 0.54, + "learning_rate": 2.841645431379687e-06, + "loss": 0.0217, + "step": 2950 + }, + { + "epoch": 0.54, + "learning_rate": 2.841590826356025e-06, + "loss": 2.1434, + "step": 2951 + }, + { + "epoch": 0.54, + "learning_rate": 2.841536221332363e-06, + "loss": 1.3286, + "step": 2952 + }, + { + "epoch": 0.54, + "learning_rate": 2.8414816163087008e-06, + "loss": 1.5508, + "step": 2953 + }, + { + "epoch": 0.54, + "learning_rate": 2.8414270112850383e-06, + "loss": 0.705, + "step": 2954 + }, + { + "epoch": 0.54, + "learning_rate": 2.8413724062613762e-06, + "loss": 0.4198, + "step": 2955 + }, + { + "epoch": 0.54, + "learning_rate": 2.8413178012377138e-06, + "loss": 0.7759, + "step": 2956 + }, + { + "epoch": 0.54, + "learning_rate": 2.8412631962140517e-06, + "loss": 0.894, + "step": 2957 + }, + { + "epoch": 0.54, + "learning_rate": 2.8412085911903896e-06, + "loss": 0.6146, + "step": 2958 + }, + { + "epoch": 0.54, + "learning_rate": 2.841153986166727e-06, + "loss": 1.887, + "step": 2959 + }, + { + "epoch": 0.54, + "learning_rate": 2.841099381143065e-06, + "loss": 1.0746, + "step": 2960 + }, + { + "epoch": 0.54, + "learning_rate": 2.841044776119403e-06, + "loss": 0.5985, + "step": 2961 + }, + { + "epoch": 0.54, + "learning_rate": 2.8409901710957406e-06, + "loss": 1.8068, + "step": 2962 + }, + { + "epoch": 0.54, + "learning_rate": 2.8409355660720785e-06, + "loss": 0.0381, + "step": 2963 + }, + { + "epoch": 0.54, + "learning_rate": 2.8408809610484165e-06, + "loss": 0.403, + "step": 2964 + }, + { + "epoch": 0.54, + "learning_rate": 2.8408263560247544e-06, + "loss": 1.1425, + "step": 2965 + }, + { + "epoch": 0.54, + "learning_rate": 2.8407717510010923e-06, + "loss": 1.2667, + "step": 2966 + }, + { + "epoch": 0.54, + "learning_rate": 2.8407171459774303e-06, + "loss": 0.4622, + "step": 2967 + }, + { + "epoch": 0.54, + "learning_rate": 2.840662540953768e-06, + "loss": 0.2348, + "step": 2968 + }, + { + "epoch": 0.54, + "learning_rate": 2.8406079359301058e-06, + "loss": 0.3055, + "step": 2969 + }, + { + "epoch": 0.54, + "learning_rate": 2.8405533309064433e-06, + "loss": 0.3799, + "step": 2970 + }, + { + "epoch": 0.54, + "learning_rate": 2.8404987258827812e-06, + "loss": 1.8889, + "step": 2971 + }, + { + "epoch": 0.54, + "learning_rate": 2.840444120859119e-06, + "loss": 0.4304, + "step": 2972 + }, + { + "epoch": 0.54, + "learning_rate": 2.8403895158354567e-06, + "loss": 3.006, + "step": 2973 + }, + { + "epoch": 0.54, + "learning_rate": 2.8403349108117946e-06, + "loss": 1.8779, + "step": 2974 + }, + { + "epoch": 0.54, + "learning_rate": 2.8402803057881326e-06, + "loss": 0.4038, + "step": 2975 + }, + { + "epoch": 0.54, + "learning_rate": 2.8402257007644705e-06, + "loss": 0.7579, + "step": 2976 + }, + { + "epoch": 0.54, + "learning_rate": 2.8401710957408085e-06, + "loss": 0.565, + "step": 2977 + }, + { + "epoch": 0.54, + "learning_rate": 2.8401164907171464e-06, + "loss": 1.5381, + "step": 2978 + }, + { + "epoch": 0.54, + "learning_rate": 2.840061885693484e-06, + "loss": 0.9719, + "step": 2979 + }, + { + "epoch": 0.54, + "learning_rate": 2.840007280669822e-06, + "loss": 0.4847, + "step": 2980 + }, + { + "epoch": 0.54, + "learning_rate": 2.83995267564616e-06, + "loss": 1.4932, + "step": 2981 + }, + { + "epoch": 0.54, + "learning_rate": 2.8398980706224973e-06, + "loss": 0.6354, + "step": 2982 + }, + { + "epoch": 0.54, + "learning_rate": 2.8398434655988353e-06, + "loss": 0.5824, + "step": 2983 + }, + { + "epoch": 0.54, + "learning_rate": 2.8397888605751728e-06, + "loss": 0.5287, + "step": 2984 + }, + { + "epoch": 0.54, + "learning_rate": 2.8397342555515107e-06, + "loss": 1.2343, + "step": 2985 + }, + { + "epoch": 0.54, + "learning_rate": 2.8396796505278487e-06, + "loss": 1.7453, + "step": 2986 + }, + { + "epoch": 0.54, + "learning_rate": 2.839625045504186e-06, + "loss": 0.3786, + "step": 2987 + }, + { + "epoch": 0.54, + "learning_rate": 2.839570440480524e-06, + "loss": 0.3754, + "step": 2988 + }, + { + "epoch": 0.54, + "learning_rate": 2.839515835456862e-06, + "loss": 0.429, + "step": 2989 + }, + { + "epoch": 0.54, + "learning_rate": 2.8394612304332e-06, + "loss": 1.0922, + "step": 2990 + }, + { + "epoch": 0.54, + "learning_rate": 2.839406625409538e-06, + "loss": 0.9852, + "step": 2991 + }, + { + "epoch": 0.54, + "learning_rate": 2.839352020385876e-06, + "loss": 1.1919, + "step": 2992 + }, + { + "epoch": 0.54, + "learning_rate": 2.8392974153622134e-06, + "loss": 1.983, + "step": 2993 + }, + { + "epoch": 0.54, + "learning_rate": 2.8392428103385514e-06, + "loss": 1.0478, + "step": 2994 + }, + { + "epoch": 0.54, + "learning_rate": 2.8391882053148893e-06, + "loss": 0.0469, + "step": 2995 + }, + { + "epoch": 0.54, + "learning_rate": 2.839133600291227e-06, + "loss": 0.4999, + "step": 2996 + }, + { + "epoch": 0.55, + "learning_rate": 2.8390789952675648e-06, + "loss": 2.3108, + "step": 2997 + }, + { + "epoch": 0.55, + "learning_rate": 2.8390243902439023e-06, + "loss": 0.9072, + "step": 2998 + }, + { + "epoch": 0.55, + "learning_rate": 2.8389697852202402e-06, + "loss": 1.337, + "step": 2999 + }, + { + "epoch": 0.55, + "learning_rate": 2.838915180196578e-06, + "loss": 1.6418, + "step": 3000 + }, + { + "epoch": 0.55, + "learning_rate": 2.8388605751729157e-06, + "loss": 0.9773, + "step": 3001 + }, + { + "epoch": 0.55, + "learning_rate": 2.8388059701492537e-06, + "loss": 0.4956, + "step": 3002 + }, + { + "epoch": 0.55, + "learning_rate": 2.8387513651255916e-06, + "loss": 0.3712, + "step": 3003 + }, + { + "epoch": 0.55, + "learning_rate": 2.8386967601019295e-06, + "loss": 0.8214, + "step": 3004 + }, + { + "epoch": 0.55, + "learning_rate": 2.8386421550782675e-06, + "loss": 2.506, + "step": 3005 + }, + { + "epoch": 0.55, + "learning_rate": 2.8385875500546054e-06, + "loss": 1.2509, + "step": 3006 + }, + { + "epoch": 0.55, + "learning_rate": 2.838532945030943e-06, + "loss": 0.9781, + "step": 3007 + }, + { + "epoch": 0.55, + "learning_rate": 2.838478340007281e-06, + "loss": 0.038, + "step": 3008 + }, + { + "epoch": 0.55, + "learning_rate": 2.8384237349836184e-06, + "loss": 1.1859, + "step": 3009 + }, + { + "epoch": 0.55, + "learning_rate": 2.8383691299599564e-06, + "loss": 1.2579, + "step": 3010 + }, + { + "epoch": 0.55, + "learning_rate": 2.8383145249362943e-06, + "loss": 0.6434, + "step": 3011 + }, + { + "epoch": 0.55, + "learning_rate": 2.838259919912632e-06, + "loss": 0.4877, + "step": 3012 + }, + { + "epoch": 0.55, + "learning_rate": 2.8382053148889698e-06, + "loss": 1.7121, + "step": 3013 + }, + { + "epoch": 0.55, + "learning_rate": 2.8381507098653077e-06, + "loss": 0.5333, + "step": 3014 + }, + { + "epoch": 0.55, + "learning_rate": 2.8380961048416452e-06, + "loss": 2.7293, + "step": 3015 + }, + { + "epoch": 0.55, + "learning_rate": 2.838041499817983e-06, + "loss": 0.7426, + "step": 3016 + }, + { + "epoch": 0.55, + "learning_rate": 2.837986894794321e-06, + "loss": 0.6099, + "step": 3017 + }, + { + "epoch": 0.55, + "learning_rate": 2.837932289770659e-06, + "loss": 1.9515, + "step": 3018 + }, + { + "epoch": 0.55, + "learning_rate": 2.837877684746997e-06, + "loss": 1.3632, + "step": 3019 + }, + { + "epoch": 0.55, + "learning_rate": 2.837823079723335e-06, + "loss": 1.2838, + "step": 3020 + }, + { + "epoch": 0.55, + "learning_rate": 2.8377684746996725e-06, + "loss": 1.2738, + "step": 3021 + }, + { + "epoch": 0.55, + "learning_rate": 2.8377138696760104e-06, + "loss": 1.4178, + "step": 3022 + }, + { + "epoch": 0.55, + "learning_rate": 2.837659264652348e-06, + "loss": 2.0573, + "step": 3023 + }, + { + "epoch": 0.55, + "learning_rate": 2.837604659628686e-06, + "loss": 0.4038, + "step": 3024 + }, + { + "epoch": 0.55, + "learning_rate": 2.837550054605024e-06, + "loss": 1.6234, + "step": 3025 + }, + { + "epoch": 0.55, + "learning_rate": 2.8374954495813613e-06, + "loss": 0.8118, + "step": 3026 + }, + { + "epoch": 0.55, + "learning_rate": 2.8374408445576993e-06, + "loss": 0.4392, + "step": 3027 + }, + { + "epoch": 0.55, + "learning_rate": 2.8373862395340372e-06, + "loss": 0.9046, + "step": 3028 + }, + { + "epoch": 0.55, + "learning_rate": 2.8373316345103747e-06, + "loss": 2.5018, + "step": 3029 + }, + { + "epoch": 0.55, + "learning_rate": 2.8372770294867127e-06, + "loss": 0.9881, + "step": 3030 + }, + { + "epoch": 0.55, + "learning_rate": 2.8372224244630506e-06, + "loss": 0.1831, + "step": 3031 + }, + { + "epoch": 0.55, + "learning_rate": 2.8371678194393886e-06, + "loss": 1.7839, + "step": 3032 + }, + { + "epoch": 0.55, + "learning_rate": 2.8371132144157265e-06, + "loss": 2.0693, + "step": 3033 + }, + { + "epoch": 0.55, + "learning_rate": 2.8370586093920645e-06, + "loss": 1.3491, + "step": 3034 + }, + { + "epoch": 0.55, + "learning_rate": 2.837004004368402e-06, + "loss": 0.6368, + "step": 3035 + }, + { + "epoch": 0.55, + "learning_rate": 2.83694939934474e-06, + "loss": 0.5693, + "step": 3036 + }, + { + "epoch": 0.55, + "learning_rate": 2.8368947943210774e-06, + "loss": 2.4142, + "step": 3037 + }, + { + "epoch": 0.55, + "learning_rate": 2.8368401892974154e-06, + "loss": 0.2604, + "step": 3038 + }, + { + "epoch": 0.55, + "learning_rate": 2.8367855842737533e-06, + "loss": 2.4612, + "step": 3039 + }, + { + "epoch": 0.55, + "learning_rate": 2.836730979250091e-06, + "loss": 0.7566, + "step": 3040 + }, + { + "epoch": 0.55, + "learning_rate": 2.8366763742264288e-06, + "loss": 0.4335, + "step": 3041 + }, + { + "epoch": 0.55, + "learning_rate": 2.8366217692027667e-06, + "loss": 1.5381, + "step": 3042 + }, + { + "epoch": 0.55, + "learning_rate": 2.8365671641791042e-06, + "loss": 1.7487, + "step": 3043 + }, + { + "epoch": 0.55, + "learning_rate": 2.836512559155442e-06, + "loss": 0.3441, + "step": 3044 + }, + { + "epoch": 0.55, + "learning_rate": 2.83645795413178e-06, + "loss": 0.7518, + "step": 3045 + }, + { + "epoch": 0.55, + "learning_rate": 2.836403349108118e-06, + "loss": 1.1385, + "step": 3046 + }, + { + "epoch": 0.55, + "learning_rate": 2.836348744084456e-06, + "loss": 0.9954, + "step": 3047 + }, + { + "epoch": 0.55, + "learning_rate": 2.836294139060794e-06, + "loss": 1.4871, + "step": 3048 + }, + { + "epoch": 0.55, + "learning_rate": 2.8362395340371315e-06, + "loss": 1.3421, + "step": 3049 + }, + { + "epoch": 0.55, + "learning_rate": 2.8361849290134694e-06, + "loss": 1.2965, + "step": 3050 + }, + { + "epoch": 0.55, + "learning_rate": 2.836130323989807e-06, + "loss": 0.8147, + "step": 3051 + }, + { + "epoch": 0.56, + "learning_rate": 2.836075718966145e-06, + "loss": 1.0102, + "step": 3052 + }, + { + "epoch": 0.56, + "learning_rate": 2.836021113942483e-06, + "loss": 0.2462, + "step": 3053 + }, + { + "epoch": 0.56, + "learning_rate": 2.8359665089188204e-06, + "loss": 0.7449, + "step": 3054 + }, + { + "epoch": 0.56, + "learning_rate": 2.8359119038951583e-06, + "loss": 0.6785, + "step": 3055 + }, + { + "epoch": 0.56, + "learning_rate": 2.8358572988714962e-06, + "loss": 0.4606, + "step": 3056 + }, + { + "epoch": 0.56, + "learning_rate": 2.8358026938478338e-06, + "loss": 1.2557, + "step": 3057 + }, + { + "epoch": 0.56, + "learning_rate": 2.8357480888241717e-06, + "loss": 0.8432, + "step": 3058 + }, + { + "epoch": 0.56, + "learning_rate": 2.8356934838005096e-06, + "loss": 1.692, + "step": 3059 + }, + { + "epoch": 0.56, + "learning_rate": 2.8356388787768476e-06, + "loss": 4.58, + "step": 3060 + }, + { + "epoch": 0.56, + "learning_rate": 2.8355842737531855e-06, + "loss": 1.8083, + "step": 3061 + }, + { + "epoch": 0.56, + "learning_rate": 2.8355296687295235e-06, + "loss": 0.1017, + "step": 3062 + }, + { + "epoch": 0.56, + "learning_rate": 2.835475063705861e-06, + "loss": 0.7207, + "step": 3063 + }, + { + "epoch": 0.56, + "learning_rate": 2.835420458682199e-06, + "loss": 0.5004, + "step": 3064 + }, + { + "epoch": 0.56, + "learning_rate": 2.8353658536585365e-06, + "loss": 0.7338, + "step": 3065 + }, + { + "epoch": 0.56, + "learning_rate": 2.8353112486348744e-06, + "loss": 0.3343, + "step": 3066 + }, + { + "epoch": 0.56, + "learning_rate": 2.8352566436112123e-06, + "loss": 0.9209, + "step": 3067 + }, + { + "epoch": 0.56, + "learning_rate": 2.83520203858755e-06, + "loss": 0.4508, + "step": 3068 + }, + { + "epoch": 0.56, + "learning_rate": 2.835147433563888e-06, + "loss": 2.2361, + "step": 3069 + }, + { + "epoch": 0.56, + "learning_rate": 2.8350928285402258e-06, + "loss": 2.281, + "step": 3070 + }, + { + "epoch": 0.56, + "learning_rate": 2.8350382235165637e-06, + "loss": 1.0652, + "step": 3071 + }, + { + "epoch": 0.56, + "learning_rate": 2.8349836184929016e-06, + "loss": 0.5383, + "step": 3072 + }, + { + "epoch": 0.56, + "learning_rate": 2.8349290134692396e-06, + "loss": 0.8238, + "step": 3073 + }, + { + "epoch": 0.56, + "learning_rate": 2.834874408445577e-06, + "loss": 0.1412, + "step": 3074 + }, + { + "epoch": 0.56, + "learning_rate": 2.834819803421915e-06, + "loss": 0.6295, + "step": 3075 + }, + { + "epoch": 0.56, + "learning_rate": 2.834765198398253e-06, + "loss": 0.1693, + "step": 3076 + }, + { + "epoch": 0.56, + "learning_rate": 2.8347105933745905e-06, + "loss": 0.8847, + "step": 3077 + }, + { + "epoch": 0.56, + "learning_rate": 2.8346559883509285e-06, + "loss": 0.945, + "step": 3078 + }, + { + "epoch": 0.56, + "learning_rate": 2.834601383327266e-06, + "loss": 0.5379, + "step": 3079 + }, + { + "epoch": 0.56, + "learning_rate": 2.834546778303604e-06, + "loss": 1.1738, + "step": 3080 + }, + { + "epoch": 0.56, + "learning_rate": 2.834492173279942e-06, + "loss": 0.9586, + "step": 3081 + }, + { + "epoch": 0.56, + "learning_rate": 2.8344375682562794e-06, + "loss": 0.5108, + "step": 3082 + }, + { + "epoch": 0.56, + "learning_rate": 2.8343829632326173e-06, + "loss": 4.3894, + "step": 3083 + }, + { + "epoch": 0.56, + "learning_rate": 2.8343283582089553e-06, + "loss": 2.0332, + "step": 3084 + }, + { + "epoch": 0.56, + "learning_rate": 2.834273753185293e-06, + "loss": 1.5688, + "step": 3085 + }, + { + "epoch": 0.56, + "learning_rate": 2.834219148161631e-06, + "loss": 1.8452, + "step": 3086 + }, + { + "epoch": 0.56, + "learning_rate": 2.834164543137969e-06, + "loss": 0.1288, + "step": 3087 + }, + { + "epoch": 0.56, + "learning_rate": 2.8341099381143066e-06, + "loss": 0.8171, + "step": 3088 + }, + { + "epoch": 0.56, + "learning_rate": 2.8340553330906446e-06, + "loss": 2.2711, + "step": 3089 + }, + { + "epoch": 0.56, + "learning_rate": 2.834000728066982e-06, + "loss": 0.2527, + "step": 3090 + }, + { + "epoch": 0.56, + "learning_rate": 2.83394612304332e-06, + "loss": 0.377, + "step": 3091 + }, + { + "epoch": 0.56, + "learning_rate": 2.833891518019658e-06, + "loss": 2.3497, + "step": 3092 + }, + { + "epoch": 0.56, + "learning_rate": 2.8338369129959955e-06, + "loss": 2.3249, + "step": 3093 + }, + { + "epoch": 0.56, + "learning_rate": 2.8337823079723334e-06, + "loss": 1.4676, + "step": 3094 + }, + { + "epoch": 0.56, + "learning_rate": 2.8337277029486714e-06, + "loss": 0.459, + "step": 3095 + }, + { + "epoch": 0.56, + "learning_rate": 2.833673097925009e-06, + "loss": 1.0608, + "step": 3096 + }, + { + "epoch": 0.56, + "learning_rate": 2.833618492901347e-06, + "loss": 2.8628, + "step": 3097 + }, + { + "epoch": 0.56, + "learning_rate": 2.8335638878776848e-06, + "loss": 0.6216, + "step": 3098 + }, + { + "epoch": 0.56, + "learning_rate": 2.8335092828540227e-06, + "loss": 1.2619, + "step": 3099 + }, + { + "epoch": 0.56, + "learning_rate": 2.8334546778303607e-06, + "loss": 2.4358, + "step": 3100 + }, + { + "epoch": 0.56, + "learning_rate": 2.8334000728066986e-06, + "loss": 1.8361, + "step": 3101 + }, + { + "epoch": 0.56, + "learning_rate": 2.833345467783036e-06, + "loss": 1.0776, + "step": 3102 + }, + { + "epoch": 0.56, + "learning_rate": 2.833290862759374e-06, + "loss": 1.2827, + "step": 3103 + }, + { + "epoch": 0.56, + "learning_rate": 2.8332362577357116e-06, + "loss": 0.6405, + "step": 3104 + }, + { + "epoch": 0.56, + "learning_rate": 2.8331816527120495e-06, + "loss": 1.6886, + "step": 3105 + }, + { + "epoch": 0.56, + "learning_rate": 2.8331270476883875e-06, + "loss": 0.0516, + "step": 3106 + }, + { + "epoch": 0.57, + "learning_rate": 2.833072442664725e-06, + "loss": 1.49, + "step": 3107 + }, + { + "epoch": 0.57, + "learning_rate": 2.833017837641063e-06, + "loss": 0.9882, + "step": 3108 + }, + { + "epoch": 0.57, + "learning_rate": 2.832963232617401e-06, + "loss": 3.2185, + "step": 3109 + }, + { + "epoch": 0.57, + "learning_rate": 2.8329086275937384e-06, + "loss": 2.5982, + "step": 3110 + }, + { + "epoch": 0.57, + "learning_rate": 2.8328540225700764e-06, + "loss": 0.5482, + "step": 3111 + }, + { + "epoch": 0.57, + "learning_rate": 2.8327994175464143e-06, + "loss": 0.4266, + "step": 3112 + }, + { + "epoch": 0.57, + "learning_rate": 2.8327448125227522e-06, + "loss": 1.0954, + "step": 3113 + }, + { + "epoch": 0.57, + "learning_rate": 2.83269020749909e-06, + "loss": 1.2623, + "step": 3114 + }, + { + "epoch": 0.57, + "learning_rate": 2.832635602475428e-06, + "loss": 0.5114, + "step": 3115 + }, + { + "epoch": 0.57, + "learning_rate": 2.8325809974517656e-06, + "loss": 0.5917, + "step": 3116 + }, + { + "epoch": 0.57, + "learning_rate": 2.8325263924281036e-06, + "loss": 0.7349, + "step": 3117 + }, + { + "epoch": 0.57, + "learning_rate": 2.832471787404441e-06, + "loss": 1.5303, + "step": 3118 + }, + { + "epoch": 0.57, + "learning_rate": 2.832417182380779e-06, + "loss": 1.6332, + "step": 3119 + }, + { + "epoch": 0.57, + "learning_rate": 2.832362577357117e-06, + "loss": 0.7596, + "step": 3120 + }, + { + "epoch": 0.57, + "learning_rate": 2.8323079723334545e-06, + "loss": 0.1841, + "step": 3121 + }, + { + "epoch": 0.57, + "learning_rate": 2.8322533673097925e-06, + "loss": 0.6331, + "step": 3122 + }, + { + "epoch": 0.57, + "learning_rate": 2.8321987622861304e-06, + "loss": 3.4382, + "step": 3123 + }, + { + "epoch": 0.57, + "learning_rate": 2.832144157262468e-06, + "loss": 0.7958, + "step": 3124 + }, + { + "epoch": 0.57, + "learning_rate": 2.832089552238806e-06, + "loss": 2.2552, + "step": 3125 + }, + { + "epoch": 0.57, + "learning_rate": 2.832034947215144e-06, + "loss": 1.2539, + "step": 3126 + }, + { + "epoch": 0.57, + "learning_rate": 2.8319803421914818e-06, + "loss": 0.1643, + "step": 3127 + }, + { + "epoch": 0.57, + "learning_rate": 2.8319257371678197e-06, + "loss": 2.6818, + "step": 3128 + }, + { + "epoch": 0.57, + "learning_rate": 2.8318711321441576e-06, + "loss": 1.4281, + "step": 3129 + }, + { + "epoch": 0.57, + "learning_rate": 2.831816527120495e-06, + "loss": 1.2964, + "step": 3130 + }, + { + "epoch": 0.57, + "learning_rate": 2.831761922096833e-06, + "loss": 0.5424, + "step": 3131 + }, + { + "epoch": 0.57, + "learning_rate": 2.8317073170731706e-06, + "loss": 0.5443, + "step": 3132 + }, + { + "epoch": 0.57, + "learning_rate": 2.8316527120495086e-06, + "loss": 1.6075, + "step": 3133 + }, + { + "epoch": 0.57, + "learning_rate": 2.8315981070258465e-06, + "loss": 0.8524, + "step": 3134 + }, + { + "epoch": 0.57, + "learning_rate": 2.831543502002184e-06, + "loss": 1.7852, + "step": 3135 + }, + { + "epoch": 0.57, + "learning_rate": 2.831488896978522e-06, + "loss": 0.3744, + "step": 3136 + }, + { + "epoch": 0.57, + "learning_rate": 2.83143429195486e-06, + "loss": 0.308, + "step": 3137 + }, + { + "epoch": 0.57, + "learning_rate": 2.8313796869311974e-06, + "loss": 0.549, + "step": 3138 + }, + { + "epoch": 0.57, + "learning_rate": 2.8313250819075354e-06, + "loss": 0.5623, + "step": 3139 + }, + { + "epoch": 0.57, + "learning_rate": 2.8312704768838733e-06, + "loss": 0.7856, + "step": 3140 + }, + { + "epoch": 0.57, + "learning_rate": 2.8312158718602113e-06, + "loss": 0.8301, + "step": 3141 + }, + { + "epoch": 0.57, + "learning_rate": 2.831161266836549e-06, + "loss": 1.1675, + "step": 3142 + }, + { + "epoch": 0.57, + "learning_rate": 2.831106661812887e-06, + "loss": 2.0376, + "step": 3143 + }, + { + "epoch": 0.57, + "learning_rate": 2.8310520567892247e-06, + "loss": 1.4025, + "step": 3144 + }, + { + "epoch": 0.57, + "learning_rate": 2.8309974517655626e-06, + "loss": 1.1007, + "step": 3145 + }, + { + "epoch": 0.57, + "learning_rate": 2.8309428467419e-06, + "loss": 0.9841, + "step": 3146 + }, + { + "epoch": 0.57, + "learning_rate": 2.830888241718238e-06, + "loss": 1.1862, + "step": 3147 + }, + { + "epoch": 0.57, + "learning_rate": 2.830833636694576e-06, + "loss": 0.3493, + "step": 3148 + }, + { + "epoch": 0.57, + "learning_rate": 2.8307790316709135e-06, + "loss": 1.8974, + "step": 3149 + }, + { + "epoch": 0.57, + "learning_rate": 2.8307244266472515e-06, + "loss": 1.0344, + "step": 3150 + }, + { + "epoch": 0.57, + "learning_rate": 2.8306698216235894e-06, + "loss": 0.0512, + "step": 3151 + }, + { + "epoch": 0.57, + "learning_rate": 2.8306152165999274e-06, + "loss": 0.0572, + "step": 3152 + }, + { + "epoch": 0.57, + "learning_rate": 2.8305606115762653e-06, + "loss": 2.6859, + "step": 3153 + }, + { + "epoch": 0.57, + "learning_rate": 2.8305060065526033e-06, + "loss": 0.0877, + "step": 3154 + }, + { + "epoch": 0.57, + "learning_rate": 2.8304514015289408e-06, + "loss": 0.0721, + "step": 3155 + }, + { + "epoch": 0.57, + "learning_rate": 2.8303967965052787e-06, + "loss": 0.694, + "step": 3156 + }, + { + "epoch": 0.57, + "learning_rate": 2.8303421914816162e-06, + "loss": 0.9986, + "step": 3157 + }, + { + "epoch": 0.57, + "learning_rate": 2.830287586457954e-06, + "loss": 0.7903, + "step": 3158 + }, + { + "epoch": 0.57, + "learning_rate": 2.830232981434292e-06, + "loss": 0.1652, + "step": 3159 + }, + { + "epoch": 0.57, + "learning_rate": 2.8301783764106296e-06, + "loss": 1.4351, + "step": 3160 + }, + { + "epoch": 0.57, + "learning_rate": 2.8301237713869676e-06, + "loss": 0.7786, + "step": 3161 + }, + { + "epoch": 0.58, + "learning_rate": 2.8300691663633055e-06, + "loss": 0.5187, + "step": 3162 + }, + { + "epoch": 0.58, + "learning_rate": 2.830014561339643e-06, + "loss": 2.7731, + "step": 3163 + }, + { + "epoch": 0.58, + "learning_rate": 2.829959956315981e-06, + "loss": 0.4346, + "step": 3164 + }, + { + "epoch": 0.58, + "learning_rate": 2.829905351292319e-06, + "loss": 1.069, + "step": 3165 + }, + { + "epoch": 0.58, + "learning_rate": 2.829850746268657e-06, + "loss": 3.3014, + "step": 3166 + }, + { + "epoch": 0.58, + "learning_rate": 2.829796141244995e-06, + "loss": 1.4589, + "step": 3167 + }, + { + "epoch": 0.58, + "learning_rate": 2.8297415362213328e-06, + "loss": 0.7926, + "step": 3168 + }, + { + "epoch": 0.58, + "learning_rate": 2.8296869311976703e-06, + "loss": 0.6773, + "step": 3169 + }, + { + "epoch": 0.58, + "learning_rate": 2.8296323261740082e-06, + "loss": 1.0317, + "step": 3170 + }, + { + "epoch": 0.58, + "learning_rate": 2.8295777211503458e-06, + "loss": 2.4142, + "step": 3171 + }, + { + "epoch": 0.58, + "learning_rate": 2.8295231161266837e-06, + "loss": 2.0744, + "step": 3172 + }, + { + "epoch": 0.58, + "learning_rate": 2.8294685111030216e-06, + "loss": 1.5061, + "step": 3173 + }, + { + "epoch": 0.58, + "learning_rate": 2.829413906079359e-06, + "loss": 1.0149, + "step": 3174 + }, + { + "epoch": 0.58, + "learning_rate": 2.829359301055697e-06, + "loss": 1.2628, + "step": 3175 + }, + { + "epoch": 0.58, + "learning_rate": 2.829304696032035e-06, + "loss": 2.3206, + "step": 3176 + }, + { + "epoch": 0.58, + "learning_rate": 2.8292500910083726e-06, + "loss": 0.8199, + "step": 3177 + }, + { + "epoch": 0.58, + "learning_rate": 2.8291954859847105e-06, + "loss": 1.1499, + "step": 3178 + }, + { + "epoch": 0.58, + "learning_rate": 2.8291408809610485e-06, + "loss": 1.6592, + "step": 3179 + }, + { + "epoch": 0.58, + "learning_rate": 2.8290862759373864e-06, + "loss": 2.516, + "step": 3180 + }, + { + "epoch": 0.58, + "learning_rate": 2.8290316709137243e-06, + "loss": 0.4795, + "step": 3181 + }, + { + "epoch": 0.58, + "learning_rate": 2.8289770658900623e-06, + "loss": 0.2858, + "step": 3182 + }, + { + "epoch": 0.58, + "learning_rate": 2.8289224608664e-06, + "loss": 1.5258, + "step": 3183 + }, + { + "epoch": 0.58, + "learning_rate": 2.8288678558427377e-06, + "loss": 0.6613, + "step": 3184 + }, + { + "epoch": 0.58, + "learning_rate": 2.8288132508190753e-06, + "loss": 1.2871, + "step": 3185 + }, + { + "epoch": 0.58, + "learning_rate": 2.8287586457954132e-06, + "loss": 1.2162, + "step": 3186 + }, + { + "epoch": 0.58, + "learning_rate": 2.828704040771751e-06, + "loss": 0.9962, + "step": 3187 + }, + { + "epoch": 0.58, + "learning_rate": 2.8286494357480887e-06, + "loss": 0.8477, + "step": 3188 + }, + { + "epoch": 0.58, + "learning_rate": 2.8285948307244266e-06, + "loss": 1.0462, + "step": 3189 + }, + { + "epoch": 0.58, + "learning_rate": 2.8285402257007646e-06, + "loss": 1.6387, + "step": 3190 + }, + { + "epoch": 0.58, + "learning_rate": 2.828485620677102e-06, + "loss": 1.127, + "step": 3191 + }, + { + "epoch": 0.58, + "learning_rate": 2.82843101565344e-06, + "loss": 1.1648, + "step": 3192 + }, + { + "epoch": 0.58, + "learning_rate": 2.828376410629778e-06, + "loss": 0.8834, + "step": 3193 + }, + { + "epoch": 0.58, + "learning_rate": 2.828321805606116e-06, + "loss": 0.3906, + "step": 3194 + }, + { + "epoch": 0.58, + "learning_rate": 2.828267200582454e-06, + "loss": 0.2029, + "step": 3195 + }, + { + "epoch": 0.58, + "learning_rate": 2.828212595558792e-06, + "loss": 1.0837, + "step": 3196 + }, + { + "epoch": 0.58, + "learning_rate": 2.8281579905351293e-06, + "loss": 1.5442, + "step": 3197 + }, + { + "epoch": 0.58, + "learning_rate": 2.8281033855114673e-06, + "loss": 1.975, + "step": 3198 + }, + { + "epoch": 0.58, + "learning_rate": 2.8280487804878048e-06, + "loss": 3.3194, + "step": 3199 + }, + { + "epoch": 0.58, + "learning_rate": 2.8279941754641427e-06, + "loss": 1.7534, + "step": 3200 + }, + { + "epoch": 0.58, + "learning_rate": 2.8279395704404807e-06, + "loss": 1.188, + "step": 3201 + }, + { + "epoch": 0.58, + "learning_rate": 2.827884965416818e-06, + "loss": 1.6408, + "step": 3202 + }, + { + "epoch": 0.58, + "learning_rate": 2.827830360393156e-06, + "loss": 0.1816, + "step": 3203 + }, + { + "epoch": 0.58, + "learning_rate": 2.827775755369494e-06, + "loss": 1.1395, + "step": 3204 + }, + { + "epoch": 0.58, + "learning_rate": 2.8277211503458316e-06, + "loss": 0.8129, + "step": 3205 + }, + { + "epoch": 0.58, + "learning_rate": 2.8276665453221695e-06, + "loss": 2.048, + "step": 3206 + }, + { + "epoch": 0.58, + "learning_rate": 2.8276119402985075e-06, + "loss": 1.5038, + "step": 3207 + }, + { + "epoch": 0.58, + "learning_rate": 2.8275573352748454e-06, + "loss": 1.3954, + "step": 3208 + }, + { + "epoch": 0.58, + "learning_rate": 2.8275027302511834e-06, + "loss": 1.6701, + "step": 3209 + }, + { + "epoch": 0.58, + "learning_rate": 2.8274481252275213e-06, + "loss": 0.0773, + "step": 3210 + }, + { + "epoch": 0.58, + "learning_rate": 2.827393520203859e-06, + "loss": 0.3387, + "step": 3211 + }, + { + "epoch": 0.58, + "learning_rate": 2.8273389151801968e-06, + "loss": 1.0777, + "step": 3212 + }, + { + "epoch": 0.58, + "learning_rate": 2.8272843101565343e-06, + "loss": 0.287, + "step": 3213 + }, + { + "epoch": 0.58, + "learning_rate": 2.8272297051328722e-06, + "loss": 1.0072, + "step": 3214 + }, + { + "epoch": 0.58, + "learning_rate": 2.82717510010921e-06, + "loss": 2.8229, + "step": 3215 + }, + { + "epoch": 0.58, + "learning_rate": 2.8271204950855477e-06, + "loss": 1.4361, + "step": 3216 + }, + { + "epoch": 0.59, + "learning_rate": 2.8270658900618856e-06, + "loss": 0.36, + "step": 3217 + }, + { + "epoch": 0.59, + "learning_rate": 2.8270112850382236e-06, + "loss": 0.4885, + "step": 3218 + }, + { + "epoch": 0.59, + "learning_rate": 2.826956680014561e-06, + "loss": 1.5065, + "step": 3219 + }, + { + "epoch": 0.59, + "learning_rate": 2.826902074990899e-06, + "loss": 0.8062, + "step": 3220 + }, + { + "epoch": 0.59, + "learning_rate": 2.826847469967237e-06, + "loss": 0.0644, + "step": 3221 + }, + { + "epoch": 0.59, + "learning_rate": 2.826792864943575e-06, + "loss": 0.1796, + "step": 3222 + }, + { + "epoch": 0.59, + "learning_rate": 2.826738259919913e-06, + "loss": 1.4979, + "step": 3223 + }, + { + "epoch": 0.59, + "learning_rate": 2.826683654896251e-06, + "loss": 1.3509, + "step": 3224 + }, + { + "epoch": 0.59, + "learning_rate": 2.8266290498725883e-06, + "loss": 1.1966, + "step": 3225 + }, + { + "epoch": 0.59, + "learning_rate": 2.8265744448489263e-06, + "loss": 1.3933, + "step": 3226 + }, + { + "epoch": 0.59, + "learning_rate": 2.826519839825264e-06, + "loss": 0.4658, + "step": 3227 + }, + { + "epoch": 0.59, + "learning_rate": 2.8264652348016018e-06, + "loss": 1.0735, + "step": 3228 + }, + { + "epoch": 0.59, + "learning_rate": 2.8264106297779397e-06, + "loss": 0.4787, + "step": 3229 + }, + { + "epoch": 0.59, + "learning_rate": 2.8263560247542772e-06, + "loss": 0.264, + "step": 3230 + }, + { + "epoch": 0.59, + "learning_rate": 2.826301419730615e-06, + "loss": 0.6664, + "step": 3231 + }, + { + "epoch": 0.59, + "learning_rate": 2.826246814706953e-06, + "loss": 0.9758, + "step": 3232 + }, + { + "epoch": 0.59, + "learning_rate": 2.826192209683291e-06, + "loss": 1.2767, + "step": 3233 + }, + { + "epoch": 0.59, + "learning_rate": 2.826137604659629e-06, + "loss": 2.3695, + "step": 3234 + }, + { + "epoch": 0.59, + "learning_rate": 2.826082999635967e-06, + "loss": 2.5105, + "step": 3235 + }, + { + "epoch": 0.59, + "learning_rate": 2.8260283946123045e-06, + "loss": 0.9466, + "step": 3236 + }, + { + "epoch": 0.59, + "learning_rate": 2.8259737895886424e-06, + "loss": 0.4707, + "step": 3237 + }, + { + "epoch": 0.59, + "learning_rate": 2.82591918456498e-06, + "loss": 1.0274, + "step": 3238 + }, + { + "epoch": 0.59, + "learning_rate": 2.825864579541318e-06, + "loss": 2.0297, + "step": 3239 + }, + { + "epoch": 0.59, + "learning_rate": 2.825809974517656e-06, + "loss": 1.5368, + "step": 3240 + }, + { + "epoch": 0.59, + "learning_rate": 2.8257553694939933e-06, + "loss": 0.9796, + "step": 3241 + }, + { + "epoch": 0.59, + "learning_rate": 2.8257007644703313e-06, + "loss": 0.0256, + "step": 3242 + }, + { + "epoch": 0.59, + "learning_rate": 2.825646159446669e-06, + "loss": 0.8489, + "step": 3243 + }, + { + "epoch": 0.59, + "learning_rate": 2.8255915544230067e-06, + "loss": 1.3382, + "step": 3244 + }, + { + "epoch": 0.59, + "learning_rate": 2.8255369493993447e-06, + "loss": 2.9581, + "step": 3245 + }, + { + "epoch": 0.59, + "learning_rate": 2.8254823443756826e-06, + "loss": 1.5093, + "step": 3246 + }, + { + "epoch": 0.59, + "learning_rate": 2.8254277393520206e-06, + "loss": 1.8839, + "step": 3247 + }, + { + "epoch": 0.59, + "learning_rate": 2.8253731343283585e-06, + "loss": 2.0895, + "step": 3248 + }, + { + "epoch": 0.59, + "learning_rate": 2.8253185293046964e-06, + "loss": 0.6052, + "step": 3249 + }, + { + "epoch": 0.59, + "learning_rate": 2.825263924281034e-06, + "loss": 0.0652, + "step": 3250 + }, + { + "epoch": 0.59, + "learning_rate": 2.825209319257372e-06, + "loss": 0.6295, + "step": 3251 + }, + { + "epoch": 0.59, + "learning_rate": 2.8251547142337094e-06, + "loss": 0.1787, + "step": 3252 + }, + { + "epoch": 0.59, + "learning_rate": 2.8251001092100474e-06, + "loss": 1.8962, + "step": 3253 + }, + { + "epoch": 0.59, + "learning_rate": 2.8250455041863853e-06, + "loss": 0.855, + "step": 3254 + }, + { + "epoch": 0.59, + "learning_rate": 2.824990899162723e-06, + "loss": 0.8504, + "step": 3255 + }, + { + "epoch": 0.59, + "learning_rate": 2.8249362941390608e-06, + "loss": 3.6703, + "step": 3256 + }, + { + "epoch": 0.59, + "learning_rate": 2.8248816891153987e-06, + "loss": 1.4324, + "step": 3257 + }, + { + "epoch": 0.59, + "learning_rate": 2.8248270840917362e-06, + "loss": 2.0806, + "step": 3258 + }, + { + "epoch": 0.59, + "learning_rate": 2.824772479068074e-06, + "loss": 1.7586, + "step": 3259 + }, + { + "epoch": 0.59, + "learning_rate": 2.824717874044412e-06, + "loss": 0.5074, + "step": 3260 + }, + { + "epoch": 0.59, + "learning_rate": 2.82466326902075e-06, + "loss": 1.492, + "step": 3261 + }, + { + "epoch": 0.59, + "learning_rate": 2.824608663997088e-06, + "loss": 2.0588, + "step": 3262 + }, + { + "epoch": 0.59, + "learning_rate": 2.824554058973426e-06, + "loss": 0.4229, + "step": 3263 + }, + { + "epoch": 0.59, + "learning_rate": 2.8244994539497635e-06, + "loss": 0.4054, + "step": 3264 + }, + { + "epoch": 0.59, + "learning_rate": 2.8244448489261014e-06, + "loss": 0.9249, + "step": 3265 + }, + { + "epoch": 0.59, + "learning_rate": 2.824390243902439e-06, + "loss": 0.9888, + "step": 3266 + }, + { + "epoch": 0.59, + "learning_rate": 2.824335638878777e-06, + "loss": 1.6043, + "step": 3267 + }, + { + "epoch": 0.59, + "learning_rate": 2.824281033855115e-06, + "loss": 0.3348, + "step": 3268 + }, + { + "epoch": 0.59, + "learning_rate": 2.8242264288314523e-06, + "loss": 1.9428, + "step": 3269 + }, + { + "epoch": 0.59, + "learning_rate": 2.8241718238077903e-06, + "loss": 0.5618, + "step": 3270 + }, + { + "epoch": 0.59, + "learning_rate": 2.8241172187841282e-06, + "loss": 0.0717, + "step": 3271 + }, + { + "epoch": 0.6, + "learning_rate": 2.8240626137604658e-06, + "loss": 2.0243, + "step": 3272 + }, + { + "epoch": 0.6, + "learning_rate": 2.8240080087368037e-06, + "loss": 0.9009, + "step": 3273 + }, + { + "epoch": 0.6, + "learning_rate": 2.8239534037131416e-06, + "loss": 0.3678, + "step": 3274 + }, + { + "epoch": 0.6, + "learning_rate": 2.8238987986894796e-06, + "loss": 0.9129, + "step": 3275 + }, + { + "epoch": 0.6, + "learning_rate": 2.8238441936658175e-06, + "loss": 2.1191, + "step": 3276 + }, + { + "epoch": 0.6, + "learning_rate": 2.8237895886421555e-06, + "loss": 2.2775, + "step": 3277 + }, + { + "epoch": 0.6, + "learning_rate": 2.823734983618493e-06, + "loss": 0.2524, + "step": 3278 + }, + { + "epoch": 0.6, + "learning_rate": 2.823680378594831e-06, + "loss": 0.9471, + "step": 3279 + }, + { + "epoch": 0.6, + "learning_rate": 2.8236257735711685e-06, + "loss": 1.6917, + "step": 3280 + }, + { + "epoch": 0.6, + "learning_rate": 2.8235711685475064e-06, + "loss": 0.6937, + "step": 3281 + }, + { + "epoch": 0.6, + "learning_rate": 2.8235165635238443e-06, + "loss": 1.3803, + "step": 3282 + }, + { + "epoch": 0.6, + "learning_rate": 2.823461958500182e-06, + "loss": 0.3002, + "step": 3283 + }, + { + "epoch": 0.6, + "learning_rate": 2.82340735347652e-06, + "loss": 1.0403, + "step": 3284 + }, + { + "epoch": 0.6, + "learning_rate": 2.8233527484528578e-06, + "loss": 0.3586, + "step": 3285 + }, + { + "epoch": 0.6, + "learning_rate": 2.8232981434291953e-06, + "loss": 1.4083, + "step": 3286 + }, + { + "epoch": 0.6, + "learning_rate": 2.8232435384055332e-06, + "loss": 0.9055, + "step": 3287 + }, + { + "epoch": 0.6, + "learning_rate": 2.823188933381871e-06, + "loss": 2.3269, + "step": 3288 + }, + { + "epoch": 0.6, + "learning_rate": 2.823134328358209e-06, + "loss": 1.5799, + "step": 3289 + }, + { + "epoch": 0.6, + "learning_rate": 2.823079723334547e-06, + "loss": 1.062, + "step": 3290 + }, + { + "epoch": 0.6, + "learning_rate": 2.823025118310885e-06, + "loss": 1.3854, + "step": 3291 + }, + { + "epoch": 0.6, + "learning_rate": 2.8229705132872225e-06, + "loss": 2.8796, + "step": 3292 + }, + { + "epoch": 0.6, + "learning_rate": 2.8229159082635605e-06, + "loss": 1.0014, + "step": 3293 + }, + { + "epoch": 0.6, + "learning_rate": 2.822861303239898e-06, + "loss": 0.1898, + "step": 3294 + }, + { + "epoch": 0.6, + "learning_rate": 2.822806698216236e-06, + "loss": 0.2666, + "step": 3295 + }, + { + "epoch": 0.6, + "learning_rate": 2.822752093192574e-06, + "loss": 2.2017, + "step": 3296 + }, + { + "epoch": 0.6, + "learning_rate": 2.8226974881689114e-06, + "loss": 1.2771, + "step": 3297 + }, + { + "epoch": 0.6, + "learning_rate": 2.8226428831452493e-06, + "loss": 3.3705, + "step": 3298 + }, + { + "epoch": 0.6, + "learning_rate": 2.8225882781215873e-06, + "loss": 0.0489, + "step": 3299 + }, + { + "epoch": 0.6, + "learning_rate": 2.8225336730979248e-06, + "loss": 0.229, + "step": 3300 + }, + { + "epoch": 0.6, + "learning_rate": 2.8224790680742627e-06, + "loss": 0.9013, + "step": 3301 + }, + { + "epoch": 0.6, + "learning_rate": 2.8224244630506007e-06, + "loss": 1.0291, + "step": 3302 + }, + { + "epoch": 0.6, + "learning_rate": 2.8223698580269386e-06, + "loss": 0.6273, + "step": 3303 + }, + { + "epoch": 0.6, + "learning_rate": 2.8223152530032766e-06, + "loss": 0.1859, + "step": 3304 + }, + { + "epoch": 0.6, + "learning_rate": 2.8222606479796145e-06, + "loss": 0.691, + "step": 3305 + }, + { + "epoch": 0.6, + "learning_rate": 2.822206042955952e-06, + "loss": 1.762, + "step": 3306 + }, + { + "epoch": 0.6, + "learning_rate": 2.82215143793229e-06, + "loss": 2.7936, + "step": 3307 + }, + { + "epoch": 0.6, + "learning_rate": 2.8220968329086275e-06, + "loss": 1.0442, + "step": 3308 + }, + { + "epoch": 0.6, + "learning_rate": 2.8220422278849654e-06, + "loss": 1.9889, + "step": 3309 + }, + { + "epoch": 0.6, + "learning_rate": 2.8219876228613034e-06, + "loss": 0.9469, + "step": 3310 + }, + { + "epoch": 0.6, + "learning_rate": 2.821933017837641e-06, + "loss": 0.1499, + "step": 3311 + }, + { + "epoch": 0.6, + "learning_rate": 2.821878412813979e-06, + "loss": 0.9036, + "step": 3312 + }, + { + "epoch": 0.6, + "learning_rate": 2.8218238077903168e-06, + "loss": 0.5158, + "step": 3313 + }, + { + "epoch": 0.6, + "learning_rate": 2.8217692027666547e-06, + "loss": 0.979, + "step": 3314 + }, + { + "epoch": 0.6, + "learning_rate": 2.8217145977429927e-06, + "loss": 0.7349, + "step": 3315 + }, + { + "epoch": 0.6, + "learning_rate": 2.82165999271933e-06, + "loss": 0.8164, + "step": 3316 + }, + { + "epoch": 0.6, + "learning_rate": 2.821605387695668e-06, + "loss": 0.5828, + "step": 3317 + }, + { + "epoch": 0.6, + "learning_rate": 2.821550782672006e-06, + "loss": 0.5788, + "step": 3318 + }, + { + "epoch": 0.6, + "learning_rate": 2.8214961776483436e-06, + "loss": 0.992, + "step": 3319 + }, + { + "epoch": 0.6, + "learning_rate": 2.8214415726246815e-06, + "loss": 2.6928, + "step": 3320 + }, + { + "epoch": 0.6, + "learning_rate": 2.8213869676010195e-06, + "loss": 1.586, + "step": 3321 + }, + { + "epoch": 0.6, + "learning_rate": 2.821332362577357e-06, + "loss": 1.8169, + "step": 3322 + }, + { + "epoch": 0.6, + "learning_rate": 2.821277757553695e-06, + "loss": 1.4082, + "step": 3323 + }, + { + "epoch": 0.6, + "learning_rate": 2.821223152530033e-06, + "loss": 2.8495, + "step": 3324 + }, + { + "epoch": 0.6, + "learning_rate": 2.8211685475063704e-06, + "loss": 1.1115, + "step": 3325 + }, + { + "epoch": 0.6, + "learning_rate": 2.8211139424827083e-06, + "loss": 0.5807, + "step": 3326 + }, + { + "epoch": 0.61, + "learning_rate": 2.8210593374590463e-06, + "loss": 0.4281, + "step": 3327 + }, + { + "epoch": 0.61, + "learning_rate": 2.8210047324353842e-06, + "loss": 1.5023, + "step": 3328 + }, + { + "epoch": 0.61, + "learning_rate": 2.820950127411722e-06, + "loss": 1.1494, + "step": 3329 + }, + { + "epoch": 0.61, + "learning_rate": 2.82089552238806e-06, + "loss": 1.1545, + "step": 3330 + }, + { + "epoch": 0.61, + "learning_rate": 2.8208409173643976e-06, + "loss": 0.1228, + "step": 3331 + }, + { + "epoch": 0.61, + "learning_rate": 2.8207863123407356e-06, + "loss": 0.7107, + "step": 3332 + }, + { + "epoch": 0.61, + "learning_rate": 2.820731707317073e-06, + "loss": 1.1782, + "step": 3333 + }, + { + "epoch": 0.61, + "learning_rate": 2.820677102293411e-06, + "loss": 0.8615, + "step": 3334 + }, + { + "epoch": 0.61, + "learning_rate": 2.820622497269749e-06, + "loss": 0.9344, + "step": 3335 + }, + { + "epoch": 0.61, + "learning_rate": 2.8205678922460865e-06, + "loss": 1.7216, + "step": 3336 + }, + { + "epoch": 0.61, + "learning_rate": 2.8205132872224245e-06, + "loss": 1.2249, + "step": 3337 + }, + { + "epoch": 0.61, + "learning_rate": 2.8204586821987624e-06, + "loss": 0.4421, + "step": 3338 + }, + { + "epoch": 0.61, + "learning_rate": 2.8204040771751e-06, + "loss": 1.9536, + "step": 3339 + }, + { + "epoch": 0.61, + "learning_rate": 2.820349472151438e-06, + "loss": 0.509, + "step": 3340 + }, + { + "epoch": 0.61, + "learning_rate": 2.820294867127776e-06, + "loss": 0.4277, + "step": 3341 + }, + { + "epoch": 0.61, + "learning_rate": 2.8202402621041137e-06, + "loss": 0.9232, + "step": 3342 + }, + { + "epoch": 0.61, + "learning_rate": 2.8201856570804517e-06, + "loss": 1.4008, + "step": 3343 + }, + { + "epoch": 0.61, + "learning_rate": 2.8201310520567896e-06, + "loss": 0.3823, + "step": 3344 + }, + { + "epoch": 0.61, + "learning_rate": 2.820076447033127e-06, + "loss": 0.7673, + "step": 3345 + }, + { + "epoch": 0.61, + "learning_rate": 2.820021842009465e-06, + "loss": 0.4428, + "step": 3346 + }, + { + "epoch": 0.61, + "learning_rate": 2.8199672369858026e-06, + "loss": 0.3453, + "step": 3347 + }, + { + "epoch": 0.61, + "learning_rate": 2.8199126319621406e-06, + "loss": 0.3177, + "step": 3348 + }, + { + "epoch": 0.61, + "learning_rate": 2.8198580269384785e-06, + "loss": 0.4952, + "step": 3349 + }, + { + "epoch": 0.61, + "learning_rate": 2.819803421914816e-06, + "loss": 1.6269, + "step": 3350 + }, + { + "epoch": 0.61, + "learning_rate": 2.819748816891154e-06, + "loss": 0.9326, + "step": 3351 + }, + { + "epoch": 0.61, + "learning_rate": 2.819694211867492e-06, + "loss": 0.4966, + "step": 3352 + }, + { + "epoch": 0.61, + "learning_rate": 2.8196396068438294e-06, + "loss": 1.5733, + "step": 3353 + }, + { + "epoch": 0.61, + "learning_rate": 2.8195850018201674e-06, + "loss": 1.3896, + "step": 3354 + }, + { + "epoch": 0.61, + "learning_rate": 2.8195303967965053e-06, + "loss": 3.0796, + "step": 3355 + }, + { + "epoch": 0.61, + "learning_rate": 2.8194757917728433e-06, + "loss": 0.794, + "step": 3356 + }, + { + "epoch": 0.61, + "learning_rate": 2.819421186749181e-06, + "loss": 1.6779, + "step": 3357 + }, + { + "epoch": 0.61, + "learning_rate": 2.819366581725519e-06, + "loss": 0.3838, + "step": 3358 + }, + { + "epoch": 0.61, + "learning_rate": 2.8193119767018567e-06, + "loss": 1.7935, + "step": 3359 + }, + { + "epoch": 0.61, + "learning_rate": 2.8192573716781946e-06, + "loss": 0.2904, + "step": 3360 + }, + { + "epoch": 0.61, + "learning_rate": 2.819202766654532e-06, + "loss": 2.2265, + "step": 3361 + }, + { + "epoch": 0.61, + "learning_rate": 2.81914816163087e-06, + "loss": 2.0032, + "step": 3362 + }, + { + "epoch": 0.61, + "learning_rate": 2.819093556607208e-06, + "loss": 2.1821, + "step": 3363 + }, + { + "epoch": 0.61, + "learning_rate": 2.8190389515835455e-06, + "loss": 2.8884, + "step": 3364 + }, + { + "epoch": 0.61, + "learning_rate": 2.8189843465598835e-06, + "loss": 1.7689, + "step": 3365 + }, + { + "epoch": 0.61, + "learning_rate": 2.8189297415362214e-06, + "loss": 0.7334, + "step": 3366 + }, + { + "epoch": 0.61, + "learning_rate": 2.818875136512559e-06, + "loss": 1.0697, + "step": 3367 + }, + { + "epoch": 0.61, + "learning_rate": 2.818820531488897e-06, + "loss": 1.0276, + "step": 3368 + }, + { + "epoch": 0.61, + "learning_rate": 2.818765926465235e-06, + "loss": 0.6104, + "step": 3369 + }, + { + "epoch": 0.61, + "learning_rate": 2.8187113214415728e-06, + "loss": 0.1257, + "step": 3370 + }, + { + "epoch": 0.61, + "learning_rate": 2.8186567164179107e-06, + "loss": 0.2014, + "step": 3371 + }, + { + "epoch": 0.61, + "learning_rate": 2.8186021113942487e-06, + "loss": 1.101, + "step": 3372 + }, + { + "epoch": 0.61, + "learning_rate": 2.818547506370586e-06, + "loss": 1.1977, + "step": 3373 + }, + { + "epoch": 0.61, + "learning_rate": 2.818492901346924e-06, + "loss": 0.382, + "step": 3374 + }, + { + "epoch": 0.61, + "learning_rate": 2.8184382963232616e-06, + "loss": 1.0717, + "step": 3375 + }, + { + "epoch": 0.61, + "learning_rate": 2.8183836912995996e-06, + "loss": 1.6844, + "step": 3376 + }, + { + "epoch": 0.61, + "learning_rate": 2.8183290862759375e-06, + "loss": 1.2664, + "step": 3377 + }, + { + "epoch": 0.61, + "learning_rate": 2.818274481252275e-06, + "loss": 2.7602, + "step": 3378 + }, + { + "epoch": 0.61, + "learning_rate": 2.818219876228613e-06, + "loss": 1.5966, + "step": 3379 + }, + { + "epoch": 0.61, + "learning_rate": 2.818165271204951e-06, + "loss": 2.4559, + "step": 3380 + }, + { + "epoch": 0.61, + "learning_rate": 2.8181106661812885e-06, + "loss": 0.3938, + "step": 3381 + }, + { + "epoch": 0.62, + "learning_rate": 2.8180560611576264e-06, + "loss": 0.5647, + "step": 3382 + }, + { + "epoch": 0.62, + "learning_rate": 2.8180014561339643e-06, + "loss": 1.3076, + "step": 3383 + }, + { + "epoch": 0.62, + "learning_rate": 2.8179468511103023e-06, + "loss": 1.6602, + "step": 3384 + }, + { + "epoch": 0.62, + "learning_rate": 2.8178922460866402e-06, + "loss": 0.9449, + "step": 3385 + }, + { + "epoch": 0.62, + "learning_rate": 2.817837641062978e-06, + "loss": 1.3863, + "step": 3386 + }, + { + "epoch": 0.62, + "learning_rate": 2.8177830360393157e-06, + "loss": 1.1823, + "step": 3387 + }, + { + "epoch": 0.62, + "learning_rate": 2.8177284310156536e-06, + "loss": 1.64, + "step": 3388 + }, + { + "epoch": 0.62, + "learning_rate": 2.817673825991991e-06, + "loss": 0.4053, + "step": 3389 + }, + { + "epoch": 0.62, + "learning_rate": 2.817619220968329e-06, + "loss": 0.8838, + "step": 3390 + }, + { + "epoch": 0.62, + "learning_rate": 2.817564615944667e-06, + "loss": 1.2647, + "step": 3391 + }, + { + "epoch": 0.62, + "learning_rate": 2.8175100109210046e-06, + "loss": 1.2717, + "step": 3392 + }, + { + "epoch": 0.62, + "learning_rate": 2.8174554058973425e-06, + "loss": 0.7229, + "step": 3393 + }, + { + "epoch": 0.62, + "learning_rate": 2.8174008008736805e-06, + "loss": 1.1202, + "step": 3394 + }, + { + "epoch": 0.62, + "learning_rate": 2.817346195850018e-06, + "loss": 0.2791, + "step": 3395 + }, + { + "epoch": 0.62, + "learning_rate": 2.817291590826356e-06, + "loss": 0.745, + "step": 3396 + }, + { + "epoch": 0.62, + "learning_rate": 2.817236985802694e-06, + "loss": 1.5448, + "step": 3397 + }, + { + "epoch": 0.62, + "learning_rate": 2.817182380779032e-06, + "loss": 0.843, + "step": 3398 + }, + { + "epoch": 0.62, + "learning_rate": 2.8171277757553697e-06, + "loss": 0.9523, + "step": 3399 + }, + { + "epoch": 0.62, + "learning_rate": 2.8170731707317073e-06, + "loss": 0.8234, + "step": 3400 + }, + { + "epoch": 0.62, + "learning_rate": 2.817018565708045e-06, + "loss": 0.329, + "step": 3401 + }, + { + "epoch": 0.62, + "learning_rate": 2.816963960684383e-06, + "loss": 3.3786, + "step": 3402 + }, + { + "epoch": 0.62, + "learning_rate": 2.8169093556607207e-06, + "loss": 2.0702, + "step": 3403 + }, + { + "epoch": 0.62, + "learning_rate": 2.8168547506370586e-06, + "loss": 1.4144, + "step": 3404 + }, + { + "epoch": 0.62, + "learning_rate": 2.8168001456133966e-06, + "loss": 0.3398, + "step": 3405 + }, + { + "epoch": 0.62, + "learning_rate": 2.816745540589734e-06, + "loss": 1.4481, + "step": 3406 + }, + { + "epoch": 0.62, + "learning_rate": 2.816690935566072e-06, + "loss": 0.2453, + "step": 3407 + }, + { + "epoch": 0.62, + "learning_rate": 2.81663633054241e-06, + "loss": 0.5917, + "step": 3408 + }, + { + "epoch": 0.62, + "learning_rate": 2.816581725518748e-06, + "loss": 1.022, + "step": 3409 + }, + { + "epoch": 0.62, + "learning_rate": 2.816527120495086e-06, + "loss": 2.202, + "step": 3410 + }, + { + "epoch": 0.62, + "learning_rate": 2.816472515471424e-06, + "loss": 1.8914, + "step": 3411 + }, + { + "epoch": 0.62, + "learning_rate": 2.8164179104477613e-06, + "loss": 0.0658, + "step": 3412 + }, + { + "epoch": 0.62, + "learning_rate": 2.8163633054240993e-06, + "loss": 2.0966, + "step": 3413 + }, + { + "epoch": 0.62, + "learning_rate": 2.8163087004004368e-06, + "loss": 2.0835, + "step": 3414 + }, + { + "epoch": 0.62, + "learning_rate": 2.8162540953767747e-06, + "loss": 1.381, + "step": 3415 + }, + { + "epoch": 0.62, + "learning_rate": 2.8161994903531127e-06, + "loss": 0.6614, + "step": 3416 + }, + { + "epoch": 0.62, + "learning_rate": 2.81614488532945e-06, + "loss": 0.2355, + "step": 3417 + }, + { + "epoch": 0.62, + "learning_rate": 2.816090280305788e-06, + "loss": 0.9431, + "step": 3418 + }, + { + "epoch": 0.62, + "learning_rate": 2.816035675282126e-06, + "loss": 0.4633, + "step": 3419 + }, + { + "epoch": 0.62, + "learning_rate": 2.8159810702584636e-06, + "loss": 1.5951, + "step": 3420 + }, + { + "epoch": 0.62, + "learning_rate": 2.8159264652348015e-06, + "loss": 0.5262, + "step": 3421 + }, + { + "epoch": 0.62, + "learning_rate": 2.8158718602111395e-06, + "loss": 2.7063, + "step": 3422 + }, + { + "epoch": 0.62, + "learning_rate": 2.8158172551874774e-06, + "loss": 1.3707, + "step": 3423 + }, + { + "epoch": 0.62, + "learning_rate": 2.8157626501638154e-06, + "loss": 1.7419, + "step": 3424 + }, + { + "epoch": 0.62, + "learning_rate": 2.8157080451401533e-06, + "loss": 0.2834, + "step": 3425 + }, + { + "epoch": 0.62, + "learning_rate": 2.815653440116491e-06, + "loss": 0.6609, + "step": 3426 + }, + { + "epoch": 0.62, + "learning_rate": 2.8155988350928288e-06, + "loss": 1.1979, + "step": 3427 + }, + { + "epoch": 0.62, + "learning_rate": 2.8155442300691663e-06, + "loss": 1.3906, + "step": 3428 + }, + { + "epoch": 0.62, + "learning_rate": 2.8154896250455042e-06, + "loss": 1.0903, + "step": 3429 + }, + { + "epoch": 0.62, + "learning_rate": 2.815435020021842e-06, + "loss": 2.1024, + "step": 3430 + }, + { + "epoch": 0.62, + "learning_rate": 2.8153804149981797e-06, + "loss": 0.545, + "step": 3431 + }, + { + "epoch": 0.62, + "learning_rate": 2.8153258099745176e-06, + "loss": 1.9389, + "step": 3432 + }, + { + "epoch": 0.62, + "learning_rate": 2.8152712049508556e-06, + "loss": 1.0027, + "step": 3433 + }, + { + "epoch": 0.62, + "learning_rate": 2.815216599927193e-06, + "loss": 1.4245, + "step": 3434 + }, + { + "epoch": 0.62, + "learning_rate": 2.815161994903531e-06, + "loss": 1.671, + "step": 3435 + }, + { + "epoch": 0.62, + "learning_rate": 2.815107389879869e-06, + "loss": 0.1931, + "step": 3436 + }, + { + "epoch": 0.63, + "learning_rate": 2.815052784856207e-06, + "loss": 2.467, + "step": 3437 + }, + { + "epoch": 0.63, + "learning_rate": 2.814998179832545e-06, + "loss": 0.2035, + "step": 3438 + }, + { + "epoch": 0.63, + "learning_rate": 2.814943574808883e-06, + "loss": 0.281, + "step": 3439 + }, + { + "epoch": 0.63, + "learning_rate": 2.8148889697852203e-06, + "loss": 0.7211, + "step": 3440 + }, + { + "epoch": 0.63, + "learning_rate": 2.8148343647615583e-06, + "loss": 0.937, + "step": 3441 + }, + { + "epoch": 0.63, + "learning_rate": 2.814779759737896e-06, + "loss": 1.2271, + "step": 3442 + }, + { + "epoch": 0.63, + "learning_rate": 2.8147251547142337e-06, + "loss": 0.1889, + "step": 3443 + }, + { + "epoch": 0.63, + "learning_rate": 2.8146705496905717e-06, + "loss": 0.1644, + "step": 3444 + }, + { + "epoch": 0.63, + "learning_rate": 2.814615944666909e-06, + "loss": 0.0731, + "step": 3445 + }, + { + "epoch": 0.63, + "learning_rate": 2.814561339643247e-06, + "loss": 0.1253, + "step": 3446 + }, + { + "epoch": 0.63, + "learning_rate": 2.814506734619585e-06, + "loss": 0.6739, + "step": 3447 + }, + { + "epoch": 0.63, + "learning_rate": 2.8144521295959226e-06, + "loss": 1.039, + "step": 3448 + }, + { + "epoch": 0.63, + "learning_rate": 2.8143975245722606e-06, + "loss": 0.6264, + "step": 3449 + }, + { + "epoch": 0.63, + "learning_rate": 2.8143429195485985e-06, + "loss": 0.9135, + "step": 3450 + }, + { + "epoch": 0.63, + "learning_rate": 2.8142883145249364e-06, + "loss": 1.621, + "step": 3451 + }, + { + "epoch": 0.63, + "learning_rate": 2.8142337095012744e-06, + "loss": 1.2398, + "step": 3452 + }, + { + "epoch": 0.63, + "learning_rate": 2.8141791044776123e-06, + "loss": 1.3551, + "step": 3453 + }, + { + "epoch": 0.63, + "learning_rate": 2.81412449945395e-06, + "loss": 0.7974, + "step": 3454 + }, + { + "epoch": 0.63, + "learning_rate": 2.814069894430288e-06, + "loss": 0.5529, + "step": 3455 + }, + { + "epoch": 0.63, + "learning_rate": 2.8140152894066253e-06, + "loss": 0.568, + "step": 3456 + }, + { + "epoch": 0.63, + "learning_rate": 2.8139606843829633e-06, + "loss": 0.7106, + "step": 3457 + }, + { + "epoch": 0.63, + "learning_rate": 2.813906079359301e-06, + "loss": 0.6921, + "step": 3458 + }, + { + "epoch": 0.63, + "learning_rate": 2.8138514743356387e-06, + "loss": 1.0432, + "step": 3459 + }, + { + "epoch": 0.63, + "learning_rate": 2.8137968693119767e-06, + "loss": 0.5075, + "step": 3460 + }, + { + "epoch": 0.63, + "learning_rate": 2.8137422642883146e-06, + "loss": 1.3208, + "step": 3461 + }, + { + "epoch": 0.63, + "learning_rate": 2.813687659264652e-06, + "loss": 0.4174, + "step": 3462 + }, + { + "epoch": 0.63, + "learning_rate": 2.81363305424099e-06, + "loss": 1.8495, + "step": 3463 + }, + { + "epoch": 0.63, + "learning_rate": 2.813578449217328e-06, + "loss": 0.7068, + "step": 3464 + }, + { + "epoch": 0.63, + "learning_rate": 2.813523844193666e-06, + "loss": 1.9651, + "step": 3465 + }, + { + "epoch": 0.63, + "learning_rate": 2.813469239170004e-06, + "loss": 1.1515, + "step": 3466 + }, + { + "epoch": 0.63, + "learning_rate": 2.8134146341463414e-06, + "loss": 1.6069, + "step": 3467 + }, + { + "epoch": 0.63, + "learning_rate": 2.8133600291226794e-06, + "loss": 0.3372, + "step": 3468 + }, + { + "epoch": 0.63, + "learning_rate": 2.8133054240990173e-06, + "loss": 1.1402, + "step": 3469 + }, + { + "epoch": 0.63, + "learning_rate": 2.813250819075355e-06, + "loss": 2.5114, + "step": 3470 + }, + { + "epoch": 0.63, + "learning_rate": 2.8131962140516928e-06, + "loss": 0.6854, + "step": 3471 + }, + { + "epoch": 0.63, + "learning_rate": 2.8131416090280307e-06, + "loss": 0.5523, + "step": 3472 + }, + { + "epoch": 0.63, + "learning_rate": 2.8130870040043682e-06, + "loss": 1.1252, + "step": 3473 + }, + { + "epoch": 0.63, + "learning_rate": 2.813032398980706e-06, + "loss": 1.6815, + "step": 3474 + }, + { + "epoch": 0.63, + "learning_rate": 2.812977793957044e-06, + "loss": 0.26, + "step": 3475 + }, + { + "epoch": 0.63, + "learning_rate": 2.8129231889333816e-06, + "loss": 2.6713, + "step": 3476 + }, + { + "epoch": 0.63, + "learning_rate": 2.8128685839097196e-06, + "loss": 0.418, + "step": 3477 + }, + { + "epoch": 0.63, + "learning_rate": 2.8128139788860575e-06, + "loss": 0.6602, + "step": 3478 + }, + { + "epoch": 0.63, + "learning_rate": 2.8127593738623955e-06, + "loss": 0.5959, + "step": 3479 + }, + { + "epoch": 0.63, + "learning_rate": 2.8127047688387334e-06, + "loss": 1.4637, + "step": 3480 + }, + { + "epoch": 0.63, + "learning_rate": 2.812650163815071e-06, + "loss": 0.3123, + "step": 3481 + }, + { + "epoch": 0.63, + "learning_rate": 2.812595558791409e-06, + "loss": 0.7514, + "step": 3482 + }, + { + "epoch": 0.63, + "learning_rate": 2.812540953767747e-06, + "loss": 0.9455, + "step": 3483 + }, + { + "epoch": 0.63, + "learning_rate": 2.8124863487440843e-06, + "loss": 1.6085, + "step": 3484 + }, + { + "epoch": 0.63, + "learning_rate": 2.8124317437204223e-06, + "loss": 0.1801, + "step": 3485 + }, + { + "epoch": 0.63, + "learning_rate": 2.8123771386967602e-06, + "loss": 1.4139, + "step": 3486 + }, + { + "epoch": 0.63, + "learning_rate": 2.8123225336730978e-06, + "loss": 0.597, + "step": 3487 + }, + { + "epoch": 0.63, + "learning_rate": 2.8122679286494357e-06, + "loss": 0.9663, + "step": 3488 + }, + { + "epoch": 0.63, + "learning_rate": 2.8122133236257736e-06, + "loss": 1.3865, + "step": 3489 + }, + { + "epoch": 0.63, + "learning_rate": 2.8121587186021116e-06, + "loss": 1.6503, + "step": 3490 + }, + { + "epoch": 0.63, + "learning_rate": 2.8121041135784495e-06, + "loss": 1.3783, + "step": 3491 + }, + { + "epoch": 0.64, + "learning_rate": 2.8120495085547875e-06, + "loss": 1.5321, + "step": 3492 + }, + { + "epoch": 0.64, + "learning_rate": 2.811994903531125e-06, + "loss": 0.8037, + "step": 3493 + }, + { + "epoch": 0.64, + "learning_rate": 2.811940298507463e-06, + "loss": 1.1863, + "step": 3494 + }, + { + "epoch": 0.64, + "learning_rate": 2.8118856934838005e-06, + "loss": 1.9804, + "step": 3495 + }, + { + "epoch": 0.64, + "learning_rate": 2.8118310884601384e-06, + "loss": 1.3076, + "step": 3496 + }, + { + "epoch": 0.64, + "learning_rate": 2.8117764834364763e-06, + "loss": 0.2908, + "step": 3497 + }, + { + "epoch": 0.64, + "learning_rate": 2.811721878412814e-06, + "loss": 1.527, + "step": 3498 + }, + { + "epoch": 0.64, + "learning_rate": 2.811667273389152e-06, + "loss": 1.6371, + "step": 3499 + }, + { + "epoch": 0.64, + "learning_rate": 2.8116126683654897e-06, + "loss": 1.2031, + "step": 3500 + }, + { + "epoch": 0.64, + "learning_rate": 2.8115580633418273e-06, + "loss": 1.1599, + "step": 3501 + }, + { + "epoch": 0.64, + "learning_rate": 2.811503458318165e-06, + "loss": 1.4977, + "step": 3502 + }, + { + "epoch": 0.64, + "learning_rate": 2.811448853294503e-06, + "loss": 0.7495, + "step": 3503 + }, + { + "epoch": 0.64, + "learning_rate": 2.811394248270841e-06, + "loss": 1.5999, + "step": 3504 + }, + { + "epoch": 0.64, + "learning_rate": 2.811339643247179e-06, + "loss": 2.8003, + "step": 3505 + }, + { + "epoch": 0.64, + "learning_rate": 2.811285038223517e-06, + "loss": 0.7893, + "step": 3506 + }, + { + "epoch": 0.64, + "learning_rate": 2.8112304331998545e-06, + "loss": 0.3142, + "step": 3507 + }, + { + "epoch": 0.64, + "learning_rate": 2.8111758281761924e-06, + "loss": 1.4037, + "step": 3508 + }, + { + "epoch": 0.64, + "learning_rate": 2.81112122315253e-06, + "loss": 1.3412, + "step": 3509 + }, + { + "epoch": 0.64, + "learning_rate": 2.811066618128868e-06, + "loss": 0.0441, + "step": 3510 + }, + { + "epoch": 0.64, + "learning_rate": 2.811012013105206e-06, + "loss": 0.4793, + "step": 3511 + }, + { + "epoch": 0.64, + "learning_rate": 2.8109574080815434e-06, + "loss": 0.68, + "step": 3512 + }, + { + "epoch": 0.64, + "learning_rate": 2.8109028030578813e-06, + "loss": 1.1555, + "step": 3513 + }, + { + "epoch": 0.64, + "learning_rate": 2.8108481980342193e-06, + "loss": 0.5628, + "step": 3514 + }, + { + "epoch": 0.64, + "learning_rate": 2.8107935930105568e-06, + "loss": 0.3691, + "step": 3515 + }, + { + "epoch": 0.64, + "learning_rate": 2.8107389879868947e-06, + "loss": 1.586, + "step": 3516 + }, + { + "epoch": 0.64, + "learning_rate": 2.8106843829632327e-06, + "loss": 0.5975, + "step": 3517 + }, + { + "epoch": 0.64, + "learning_rate": 2.8106297779395706e-06, + "loss": 0.5436, + "step": 3518 + }, + { + "epoch": 0.64, + "learning_rate": 2.8105751729159086e-06, + "loss": 0.6086, + "step": 3519 + }, + { + "epoch": 0.64, + "learning_rate": 2.8105205678922465e-06, + "loss": 1.5614, + "step": 3520 + }, + { + "epoch": 0.64, + "learning_rate": 2.810465962868584e-06, + "loss": 0.8842, + "step": 3521 + }, + { + "epoch": 0.64, + "learning_rate": 2.810411357844922e-06, + "loss": 1.1436, + "step": 3522 + }, + { + "epoch": 0.64, + "learning_rate": 2.8103567528212595e-06, + "loss": 0.9182, + "step": 3523 + }, + { + "epoch": 0.64, + "learning_rate": 2.8103021477975974e-06, + "loss": 1.1209, + "step": 3524 + }, + { + "epoch": 0.64, + "learning_rate": 2.8102475427739354e-06, + "loss": 0.7713, + "step": 3525 + }, + { + "epoch": 0.64, + "learning_rate": 2.810192937750273e-06, + "loss": 1.7254, + "step": 3526 + }, + { + "epoch": 0.64, + "learning_rate": 2.810138332726611e-06, + "loss": 0.1447, + "step": 3527 + }, + { + "epoch": 0.64, + "learning_rate": 2.8100837277029488e-06, + "loss": 0.6128, + "step": 3528 + }, + { + "epoch": 0.64, + "learning_rate": 2.8100291226792863e-06, + "loss": 0.1951, + "step": 3529 + }, + { + "epoch": 0.64, + "learning_rate": 2.8099745176556242e-06, + "loss": 0.8426, + "step": 3530 + }, + { + "epoch": 0.64, + "learning_rate": 2.809919912631962e-06, + "loss": 0.0752, + "step": 3531 + }, + { + "epoch": 0.64, + "learning_rate": 2.8098653076083e-06, + "loss": 0.804, + "step": 3532 + }, + { + "epoch": 0.64, + "learning_rate": 2.809810702584638e-06, + "loss": 0.9421, + "step": 3533 + }, + { + "epoch": 0.64, + "learning_rate": 2.809756097560976e-06, + "loss": 1.2209, + "step": 3534 + }, + { + "epoch": 0.64, + "learning_rate": 2.8097014925373135e-06, + "loss": 1.4681, + "step": 3535 + }, + { + "epoch": 0.64, + "learning_rate": 2.8096468875136515e-06, + "loss": 0.8374, + "step": 3536 + }, + { + "epoch": 0.64, + "learning_rate": 2.809592282489989e-06, + "loss": 1.9438, + "step": 3537 + }, + { + "epoch": 0.64, + "learning_rate": 2.809537677466327e-06, + "loss": 1.2304, + "step": 3538 + }, + { + "epoch": 0.64, + "learning_rate": 2.809483072442665e-06, + "loss": 0.552, + "step": 3539 + }, + { + "epoch": 0.64, + "learning_rate": 2.8094284674190024e-06, + "loss": 0.3569, + "step": 3540 + }, + { + "epoch": 0.64, + "learning_rate": 2.8093738623953403e-06, + "loss": 0.9486, + "step": 3541 + }, + { + "epoch": 0.64, + "learning_rate": 2.8093192573716783e-06, + "loss": 3.2794, + "step": 3542 + }, + { + "epoch": 0.64, + "learning_rate": 2.809264652348016e-06, + "loss": 0.6599, + "step": 3543 + }, + { + "epoch": 0.64, + "learning_rate": 2.8092100473243537e-06, + "loss": 0.0516, + "step": 3544 + }, + { + "epoch": 0.64, + "learning_rate": 2.8091554423006917e-06, + "loss": 0.9983, + "step": 3545 + }, + { + "epoch": 0.64, + "learning_rate": 2.8091008372770296e-06, + "loss": 0.6777, + "step": 3546 + }, + { + "epoch": 0.65, + "learning_rate": 2.8090462322533676e-06, + "loss": 0.4931, + "step": 3547 + }, + { + "epoch": 0.65, + "learning_rate": 2.808991627229705e-06, + "loss": 1.811, + "step": 3548 + }, + { + "epoch": 0.65, + "learning_rate": 2.808937022206043e-06, + "loss": 0.1094, + "step": 3549 + }, + { + "epoch": 0.65, + "learning_rate": 2.808882417182381e-06, + "loss": 0.1576, + "step": 3550 + }, + { + "epoch": 0.65, + "learning_rate": 2.8088278121587185e-06, + "loss": 0.5813, + "step": 3551 + }, + { + "epoch": 0.65, + "learning_rate": 2.8087732071350564e-06, + "loss": 0.4013, + "step": 3552 + }, + { + "epoch": 0.65, + "learning_rate": 2.8087186021113944e-06, + "loss": 1.2778, + "step": 3553 + }, + { + "epoch": 0.65, + "learning_rate": 2.808663997087732e-06, + "loss": 1.441, + "step": 3554 + }, + { + "epoch": 0.65, + "learning_rate": 2.80860939206407e-06, + "loss": 1.8729, + "step": 3555 + }, + { + "epoch": 0.65, + "learning_rate": 2.808554787040408e-06, + "loss": 1.4981, + "step": 3556 + }, + { + "epoch": 0.65, + "learning_rate": 2.8085001820167453e-06, + "loss": 0.419, + "step": 3557 + }, + { + "epoch": 0.65, + "learning_rate": 2.8084455769930833e-06, + "loss": 1.567, + "step": 3558 + }, + { + "epoch": 0.65, + "learning_rate": 2.808390971969421e-06, + "loss": 1.1936, + "step": 3559 + }, + { + "epoch": 0.65, + "learning_rate": 2.808336366945759e-06, + "loss": 0.3984, + "step": 3560 + }, + { + "epoch": 0.65, + "learning_rate": 2.808281761922097e-06, + "loss": 0.0399, + "step": 3561 + }, + { + "epoch": 0.65, + "learning_rate": 2.8082271568984346e-06, + "loss": 0.9287, + "step": 3562 + }, + { + "epoch": 0.65, + "learning_rate": 2.8081725518747726e-06, + "loss": 0.5523, + "step": 3563 + }, + { + "epoch": 0.65, + "learning_rate": 2.8081179468511105e-06, + "loss": 0.5494, + "step": 3564 + }, + { + "epoch": 0.65, + "learning_rate": 2.808063341827448e-06, + "loss": 1.4763, + "step": 3565 + }, + { + "epoch": 0.65, + "learning_rate": 2.808008736803786e-06, + "loss": 0.4614, + "step": 3566 + }, + { + "epoch": 0.65, + "learning_rate": 2.807954131780124e-06, + "loss": 1.4769, + "step": 3567 + }, + { + "epoch": 0.65, + "learning_rate": 2.8078995267564614e-06, + "loss": 1.6088, + "step": 3568 + }, + { + "epoch": 0.65, + "learning_rate": 2.8078449217327994e-06, + "loss": 0.152, + "step": 3569 + }, + { + "epoch": 0.65, + "learning_rate": 2.8077903167091373e-06, + "loss": 1.5217, + "step": 3570 + }, + { + "epoch": 0.65, + "learning_rate": 2.8077357116854753e-06, + "loss": 0.9495, + "step": 3571 + }, + { + "epoch": 0.65, + "learning_rate": 2.807681106661813e-06, + "loss": 1.5524, + "step": 3572 + }, + { + "epoch": 0.65, + "learning_rate": 2.807626501638151e-06, + "loss": 0.809, + "step": 3573 + }, + { + "epoch": 0.65, + "learning_rate": 2.8075718966144887e-06, + "loss": 1.8327, + "step": 3574 + }, + { + "epoch": 0.65, + "learning_rate": 2.8075172915908266e-06, + "loss": 1.5692, + "step": 3575 + }, + { + "epoch": 0.65, + "learning_rate": 2.807462686567164e-06, + "loss": 1.5399, + "step": 3576 + }, + { + "epoch": 0.65, + "learning_rate": 2.807408081543502e-06, + "loss": 0.1019, + "step": 3577 + }, + { + "epoch": 0.65, + "learning_rate": 2.80735347651984e-06, + "loss": 3.3499, + "step": 3578 + }, + { + "epoch": 0.65, + "learning_rate": 2.8072988714961775e-06, + "loss": 0.3528, + "step": 3579 + }, + { + "epoch": 0.65, + "learning_rate": 2.8072442664725155e-06, + "loss": 2.1908, + "step": 3580 + }, + { + "epoch": 0.65, + "learning_rate": 2.8071896614488534e-06, + "loss": 0.961, + "step": 3581 + }, + { + "epoch": 0.65, + "learning_rate": 2.807135056425191e-06, + "loss": 0.1071, + "step": 3582 + }, + { + "epoch": 0.65, + "learning_rate": 2.807080451401529e-06, + "loss": 2.6052, + "step": 3583 + }, + { + "epoch": 0.65, + "learning_rate": 2.807025846377867e-06, + "loss": 1.4498, + "step": 3584 + }, + { + "epoch": 0.65, + "learning_rate": 2.8069712413542048e-06, + "loss": 2.1686, + "step": 3585 + }, + { + "epoch": 0.65, + "learning_rate": 2.8069166363305427e-06, + "loss": 1.5658, + "step": 3586 + }, + { + "epoch": 0.65, + "learning_rate": 2.8068620313068807e-06, + "loss": 0.9308, + "step": 3587 + }, + { + "epoch": 0.65, + "learning_rate": 2.806807426283218e-06, + "loss": 3.2909, + "step": 3588 + }, + { + "epoch": 0.65, + "learning_rate": 2.806752821259556e-06, + "loss": 1.3554, + "step": 3589 + }, + { + "epoch": 0.65, + "learning_rate": 2.8066982162358936e-06, + "loss": 0.0663, + "step": 3590 + }, + { + "epoch": 0.65, + "learning_rate": 2.8066436112122316e-06, + "loss": 0.5436, + "step": 3591 + }, + { + "epoch": 0.65, + "learning_rate": 2.8065890061885695e-06, + "loss": 0.1024, + "step": 3592 + }, + { + "epoch": 0.65, + "learning_rate": 2.806534401164907e-06, + "loss": 1.39, + "step": 3593 + }, + { + "epoch": 0.65, + "learning_rate": 2.806479796141245e-06, + "loss": 0.7471, + "step": 3594 + }, + { + "epoch": 0.65, + "learning_rate": 2.806425191117583e-06, + "loss": 1.651, + "step": 3595 + }, + { + "epoch": 0.65, + "learning_rate": 2.8063705860939205e-06, + "loss": 1.7115, + "step": 3596 + }, + { + "epoch": 0.65, + "learning_rate": 2.8063159810702584e-06, + "loss": 1.1746, + "step": 3597 + }, + { + "epoch": 0.65, + "learning_rate": 2.8062613760465963e-06, + "loss": 0.0587, + "step": 3598 + }, + { + "epoch": 0.65, + "learning_rate": 2.8062067710229343e-06, + "loss": 0.0156, + "step": 3599 + }, + { + "epoch": 0.65, + "learning_rate": 2.8061521659992722e-06, + "loss": 0.9204, + "step": 3600 + }, + { + "epoch": 0.65, + "learning_rate": 2.80609756097561e-06, + "loss": 0.2004, + "step": 3601 + }, + { + "epoch": 0.66, + "learning_rate": 2.8060429559519477e-06, + "loss": 1.2442, + "step": 3602 + }, + { + "epoch": 0.66, + "learning_rate": 2.8059883509282856e-06, + "loss": 0.6546, + "step": 3603 + }, + { + "epoch": 0.66, + "learning_rate": 2.805933745904623e-06, + "loss": 0.928, + "step": 3604 + }, + { + "epoch": 0.66, + "learning_rate": 2.805879140880961e-06, + "loss": 0.4796, + "step": 3605 + }, + { + "epoch": 0.66, + "learning_rate": 2.805824535857299e-06, + "loss": 0.313, + "step": 3606 + }, + { + "epoch": 0.66, + "learning_rate": 2.8057699308336366e-06, + "loss": 0.6364, + "step": 3607 + }, + { + "epoch": 0.66, + "learning_rate": 2.8057153258099745e-06, + "loss": 0.0982, + "step": 3608 + }, + { + "epoch": 0.66, + "learning_rate": 2.8056607207863124e-06, + "loss": 0.6253, + "step": 3609 + }, + { + "epoch": 0.66, + "learning_rate": 2.80560611576265e-06, + "loss": 0.0255, + "step": 3610 + }, + { + "epoch": 0.66, + "learning_rate": 2.805551510738988e-06, + "loss": 0.8637, + "step": 3611 + }, + { + "epoch": 0.66, + "learning_rate": 2.805496905715326e-06, + "loss": 1.1436, + "step": 3612 + }, + { + "epoch": 0.66, + "learning_rate": 2.805442300691664e-06, + "loss": 1.4243, + "step": 3613 + }, + { + "epoch": 0.66, + "learning_rate": 2.8053876956680017e-06, + "loss": 1.576, + "step": 3614 + }, + { + "epoch": 0.66, + "learning_rate": 2.8053330906443397e-06, + "loss": 3.8692, + "step": 3615 + }, + { + "epoch": 0.66, + "learning_rate": 2.805278485620677e-06, + "loss": 1.5016, + "step": 3616 + }, + { + "epoch": 0.66, + "learning_rate": 2.805223880597015e-06, + "loss": 0.8316, + "step": 3617 + }, + { + "epoch": 0.66, + "learning_rate": 2.8051692755733527e-06, + "loss": 1.1912, + "step": 3618 + }, + { + "epoch": 0.66, + "learning_rate": 2.8051146705496906e-06, + "loss": 0.235, + "step": 3619 + }, + { + "epoch": 0.66, + "learning_rate": 2.8050600655260286e-06, + "loss": 0.508, + "step": 3620 + }, + { + "epoch": 0.66, + "learning_rate": 2.805005460502366e-06, + "loss": 1.5783, + "step": 3621 + }, + { + "epoch": 0.66, + "learning_rate": 2.804950855478704e-06, + "loss": 2.0431, + "step": 3622 + }, + { + "epoch": 0.66, + "learning_rate": 2.804896250455042e-06, + "loss": 1.869, + "step": 3623 + }, + { + "epoch": 0.66, + "learning_rate": 2.8048416454313795e-06, + "loss": 2.7977, + "step": 3624 + }, + { + "epoch": 0.66, + "learning_rate": 2.8047870404077174e-06, + "loss": 1.7513, + "step": 3625 + }, + { + "epoch": 0.66, + "learning_rate": 2.8047324353840554e-06, + "loss": 1.9079, + "step": 3626 + }, + { + "epoch": 0.66, + "learning_rate": 2.8046778303603933e-06, + "loss": 0.7521, + "step": 3627 + }, + { + "epoch": 0.66, + "learning_rate": 2.8046232253367313e-06, + "loss": 1.2761, + "step": 3628 + }, + { + "epoch": 0.66, + "learning_rate": 2.8045686203130688e-06, + "loss": 2.0029, + "step": 3629 + }, + { + "epoch": 0.66, + "learning_rate": 2.8045140152894067e-06, + "loss": 1.3928, + "step": 3630 + }, + { + "epoch": 0.66, + "learning_rate": 2.8044594102657447e-06, + "loss": 0.9095, + "step": 3631 + }, + { + "epoch": 0.66, + "learning_rate": 2.804404805242082e-06, + "loss": 0.1211, + "step": 3632 + }, + { + "epoch": 0.66, + "learning_rate": 2.80435020021842e-06, + "loss": 0.6517, + "step": 3633 + }, + { + "epoch": 0.66, + "learning_rate": 2.804295595194758e-06, + "loss": 0.7115, + "step": 3634 + }, + { + "epoch": 0.66, + "learning_rate": 2.8042409901710956e-06, + "loss": 1.7438, + "step": 3635 + }, + { + "epoch": 0.66, + "learning_rate": 2.8041863851474335e-06, + "loss": 1.7051, + "step": 3636 + }, + { + "epoch": 0.66, + "learning_rate": 2.8041317801237715e-06, + "loss": 1.7223, + "step": 3637 + }, + { + "epoch": 0.66, + "learning_rate": 2.804077175100109e-06, + "loss": 0.5883, + "step": 3638 + }, + { + "epoch": 0.66, + "learning_rate": 2.804022570076447e-06, + "loss": 1.1991, + "step": 3639 + }, + { + "epoch": 0.66, + "learning_rate": 2.803967965052785e-06, + "loss": 2.123, + "step": 3640 + }, + { + "epoch": 0.66, + "learning_rate": 2.803913360029123e-06, + "loss": 1.6779, + "step": 3641 + }, + { + "epoch": 0.66, + "learning_rate": 2.8038587550054608e-06, + "loss": 1.585, + "step": 3642 + }, + { + "epoch": 0.66, + "learning_rate": 2.8038041499817983e-06, + "loss": 1.2526, + "step": 3643 + }, + { + "epoch": 0.66, + "learning_rate": 2.8037495449581362e-06, + "loss": 2.1799, + "step": 3644 + }, + { + "epoch": 0.66, + "learning_rate": 2.803694939934474e-06, + "loss": 1.2459, + "step": 3645 + }, + { + "epoch": 0.66, + "learning_rate": 2.8036403349108117e-06, + "loss": 1.6259, + "step": 3646 + }, + { + "epoch": 0.66, + "learning_rate": 2.8035857298871496e-06, + "loss": 1.1473, + "step": 3647 + }, + { + "epoch": 0.66, + "learning_rate": 2.8035311248634876e-06, + "loss": 0.757, + "step": 3648 + }, + { + "epoch": 0.66, + "learning_rate": 2.803476519839825e-06, + "loss": 0.4691, + "step": 3649 + }, + { + "epoch": 0.66, + "learning_rate": 2.803421914816163e-06, + "loss": 3.6832, + "step": 3650 + }, + { + "epoch": 0.66, + "learning_rate": 2.803367309792501e-06, + "loss": 0.0303, + "step": 3651 + }, + { + "epoch": 0.66, + "learning_rate": 2.803312704768839e-06, + "loss": 0.92, + "step": 3652 + }, + { + "epoch": 0.66, + "learning_rate": 2.8032580997451765e-06, + "loss": 1.221, + "step": 3653 + }, + { + "epoch": 0.66, + "learning_rate": 2.8032034947215144e-06, + "loss": 0.7318, + "step": 3654 + }, + { + "epoch": 0.66, + "learning_rate": 2.8031488896978523e-06, + "loss": 2.2516, + "step": 3655 + }, + { + "epoch": 0.66, + "learning_rate": 2.8030942846741903e-06, + "loss": 2.5181, + "step": 3656 + }, + { + "epoch": 0.67, + "learning_rate": 2.803039679650528e-06, + "loss": 0.1149, + "step": 3657 + }, + { + "epoch": 0.67, + "learning_rate": 2.8029850746268657e-06, + "loss": 0.6194, + "step": 3658 + }, + { + "epoch": 0.67, + "learning_rate": 2.8029304696032037e-06, + "loss": 2.2129, + "step": 3659 + }, + { + "epoch": 0.67, + "learning_rate": 2.802875864579541e-06, + "loss": 2.3919, + "step": 3660 + }, + { + "epoch": 0.67, + "learning_rate": 2.802821259555879e-06, + "loss": 1.4876, + "step": 3661 + }, + { + "epoch": 0.67, + "learning_rate": 2.802766654532217e-06, + "loss": 0.621, + "step": 3662 + }, + { + "epoch": 0.67, + "learning_rate": 2.8027120495085546e-06, + "loss": 0.1148, + "step": 3663 + }, + { + "epoch": 0.67, + "learning_rate": 2.8026574444848926e-06, + "loss": 2.4791, + "step": 3664 + }, + { + "epoch": 0.67, + "learning_rate": 2.8026028394612305e-06, + "loss": 1.6809, + "step": 3665 + }, + { + "epoch": 0.67, + "learning_rate": 2.8025482344375684e-06, + "loss": 0.1681, + "step": 3666 + }, + { + "epoch": 0.67, + "learning_rate": 2.8024936294139064e-06, + "loss": 0.7782, + "step": 3667 + }, + { + "epoch": 0.67, + "learning_rate": 2.8024390243902443e-06, + "loss": 1.6996, + "step": 3668 + }, + { + "epoch": 0.67, + "learning_rate": 2.802384419366582e-06, + "loss": 0.6213, + "step": 3669 + }, + { + "epoch": 0.67, + "learning_rate": 2.80232981434292e-06, + "loss": 0.7563, + "step": 3670 + }, + { + "epoch": 0.67, + "learning_rate": 2.8022752093192573e-06, + "loss": 0.8266, + "step": 3671 + }, + { + "epoch": 0.67, + "learning_rate": 2.8022206042955953e-06, + "loss": 0.5744, + "step": 3672 + }, + { + "epoch": 0.67, + "learning_rate": 2.802165999271933e-06, + "loss": 2.3292, + "step": 3673 + }, + { + "epoch": 0.67, + "learning_rate": 2.8021113942482707e-06, + "loss": 2.0221, + "step": 3674 + }, + { + "epoch": 0.67, + "learning_rate": 2.8020567892246087e-06, + "loss": 0.4433, + "step": 3675 + }, + { + "epoch": 0.67, + "learning_rate": 2.8020021842009466e-06, + "loss": 1.2623, + "step": 3676 + }, + { + "epoch": 0.67, + "learning_rate": 2.801947579177284e-06, + "loss": 0.2694, + "step": 3677 + }, + { + "epoch": 0.67, + "learning_rate": 2.801892974153622e-06, + "loss": 0.5044, + "step": 3678 + }, + { + "epoch": 0.67, + "learning_rate": 2.80183836912996e-06, + "loss": 0.8702, + "step": 3679 + }, + { + "epoch": 0.67, + "learning_rate": 2.801783764106298e-06, + "loss": 2.1645, + "step": 3680 + }, + { + "epoch": 0.67, + "learning_rate": 2.801729159082636e-06, + "loss": 0.3673, + "step": 3681 + }, + { + "epoch": 0.67, + "learning_rate": 2.801674554058974e-06, + "loss": 1.2468, + "step": 3682 + }, + { + "epoch": 0.67, + "learning_rate": 2.8016199490353114e-06, + "loss": 1.6417, + "step": 3683 + }, + { + "epoch": 0.67, + "learning_rate": 2.8015653440116493e-06, + "loss": 0.1882, + "step": 3684 + }, + { + "epoch": 0.67, + "learning_rate": 2.801510738987987e-06, + "loss": 0.9047, + "step": 3685 + }, + { + "epoch": 0.67, + "learning_rate": 2.8014561339643248e-06, + "loss": 1.2363, + "step": 3686 + }, + { + "epoch": 0.67, + "learning_rate": 2.8014015289406627e-06, + "loss": 1.9858, + "step": 3687 + }, + { + "epoch": 0.67, + "learning_rate": 2.8013469239170002e-06, + "loss": 1.6637, + "step": 3688 + }, + { + "epoch": 0.67, + "learning_rate": 2.801292318893338e-06, + "loss": 1.0635, + "step": 3689 + }, + { + "epoch": 0.67, + "learning_rate": 2.801237713869676e-06, + "loss": 0.6208, + "step": 3690 + }, + { + "epoch": 0.67, + "learning_rate": 2.8011831088460136e-06, + "loss": 0.5867, + "step": 3691 + }, + { + "epoch": 0.67, + "learning_rate": 2.8011285038223516e-06, + "loss": 1.5213, + "step": 3692 + }, + { + "epoch": 0.67, + "learning_rate": 2.8010738987986895e-06, + "loss": 2.1498, + "step": 3693 + }, + { + "epoch": 0.67, + "learning_rate": 2.8010192937750275e-06, + "loss": 0.3595, + "step": 3694 + }, + { + "epoch": 0.67, + "learning_rate": 2.8009646887513654e-06, + "loss": 1.1482, + "step": 3695 + }, + { + "epoch": 0.67, + "learning_rate": 2.8009100837277034e-06, + "loss": 1.1379, + "step": 3696 + }, + { + "epoch": 0.67, + "learning_rate": 2.800855478704041e-06, + "loss": 2.3603, + "step": 3697 + }, + { + "epoch": 0.67, + "learning_rate": 2.800800873680379e-06, + "loss": 2.6598, + "step": 3698 + }, + { + "epoch": 0.67, + "learning_rate": 2.8007462686567163e-06, + "loss": 1.0074, + "step": 3699 + }, + { + "epoch": 0.67, + "learning_rate": 2.8006916636330543e-06, + "loss": 0.1725, + "step": 3700 + }, + { + "epoch": 0.67, + "learning_rate": 2.8006370586093922e-06, + "loss": 0.6137, + "step": 3701 + }, + { + "epoch": 0.67, + "learning_rate": 2.8005824535857297e-06, + "loss": 1.7454, + "step": 3702 + }, + { + "epoch": 0.67, + "learning_rate": 2.8005278485620677e-06, + "loss": 1.1114, + "step": 3703 + }, + { + "epoch": 0.67, + "learning_rate": 2.8004732435384056e-06, + "loss": 1.6445, + "step": 3704 + }, + { + "epoch": 0.67, + "learning_rate": 2.800418638514743e-06, + "loss": 1.2158, + "step": 3705 + }, + { + "epoch": 0.67, + "learning_rate": 2.800364033491081e-06, + "loss": 0.4794, + "step": 3706 + }, + { + "epoch": 0.67, + "learning_rate": 2.800309428467419e-06, + "loss": 0.5688, + "step": 3707 + }, + { + "epoch": 0.67, + "learning_rate": 2.800254823443757e-06, + "loss": 1.4109, + "step": 3708 + }, + { + "epoch": 0.67, + "learning_rate": 2.800200218420095e-06, + "loss": 0.5258, + "step": 3709 + }, + { + "epoch": 0.67, + "learning_rate": 2.8001456133964324e-06, + "loss": 0.4647, + "step": 3710 + }, + { + "epoch": 0.67, + "learning_rate": 2.8000910083727704e-06, + "loss": 4.1478, + "step": 3711 + }, + { + "epoch": 0.68, + "learning_rate": 2.8000364033491083e-06, + "loss": 0.6216, + "step": 3712 + }, + { + "epoch": 0.68, + "learning_rate": 2.799981798325446e-06, + "loss": 1.0427, + "step": 3713 + }, + { + "epoch": 0.68, + "learning_rate": 2.799927193301784e-06, + "loss": 0.8985, + "step": 3714 + }, + { + "epoch": 0.68, + "learning_rate": 2.7998725882781217e-06, + "loss": 0.6341, + "step": 3715 + }, + { + "epoch": 0.68, + "learning_rate": 2.7998179832544593e-06, + "loss": 0.4689, + "step": 3716 + }, + { + "epoch": 0.68, + "learning_rate": 2.799763378230797e-06, + "loss": 2.8803, + "step": 3717 + }, + { + "epoch": 0.68, + "learning_rate": 2.799708773207135e-06, + "loss": 0.8492, + "step": 3718 + }, + { + "epoch": 0.68, + "learning_rate": 2.7996541681834727e-06, + "loss": 1.4393, + "step": 3719 + }, + { + "epoch": 0.68, + "learning_rate": 2.7995995631598106e-06, + "loss": 0.8479, + "step": 3720 + }, + { + "epoch": 0.68, + "learning_rate": 2.7995449581361486e-06, + "loss": 0.1685, + "step": 3721 + }, + { + "epoch": 0.68, + "learning_rate": 2.7994903531124865e-06, + "loss": 1.2539, + "step": 3722 + }, + { + "epoch": 0.68, + "learning_rate": 2.7994357480888244e-06, + "loss": 0.8069, + "step": 3723 + }, + { + "epoch": 0.68, + "learning_rate": 2.799381143065162e-06, + "loss": 0.5084, + "step": 3724 + }, + { + "epoch": 0.68, + "learning_rate": 2.7993265380415e-06, + "loss": 1.6013, + "step": 3725 + }, + { + "epoch": 0.68, + "learning_rate": 2.799271933017838e-06, + "loss": 1.482, + "step": 3726 + }, + { + "epoch": 0.68, + "learning_rate": 2.7992173279941754e-06, + "loss": 1.0739, + "step": 3727 + }, + { + "epoch": 0.68, + "learning_rate": 2.7991627229705133e-06, + "loss": 0.7554, + "step": 3728 + }, + { + "epoch": 0.68, + "learning_rate": 2.7991081179468513e-06, + "loss": 1.7493, + "step": 3729 + }, + { + "epoch": 0.68, + "learning_rate": 2.7990535129231888e-06, + "loss": 0.1851, + "step": 3730 + }, + { + "epoch": 0.68, + "learning_rate": 2.7989989078995267e-06, + "loss": 0.3687, + "step": 3731 + }, + { + "epoch": 0.68, + "learning_rate": 2.7989443028758647e-06, + "loss": 1.2831, + "step": 3732 + }, + { + "epoch": 0.68, + "learning_rate": 2.798889697852202e-06, + "loss": 1.4757, + "step": 3733 + }, + { + "epoch": 0.68, + "learning_rate": 2.79883509282854e-06, + "loss": 1.2554, + "step": 3734 + }, + { + "epoch": 0.68, + "learning_rate": 2.798780487804878e-06, + "loss": 2.8627, + "step": 3735 + }, + { + "epoch": 0.68, + "learning_rate": 2.798725882781216e-06, + "loss": 0.2223, + "step": 3736 + }, + { + "epoch": 0.68, + "learning_rate": 2.798671277757554e-06, + "loss": 0.6147, + "step": 3737 + }, + { + "epoch": 0.68, + "learning_rate": 2.7986166727338915e-06, + "loss": 1.824, + "step": 3738 + }, + { + "epoch": 0.68, + "learning_rate": 2.7985620677102294e-06, + "loss": 0.6081, + "step": 3739 + }, + { + "epoch": 0.68, + "learning_rate": 2.7985074626865674e-06, + "loss": 2.0928, + "step": 3740 + }, + { + "epoch": 0.68, + "learning_rate": 2.798452857662905e-06, + "loss": 1.8589, + "step": 3741 + }, + { + "epoch": 0.68, + "learning_rate": 2.798398252639243e-06, + "loss": 0.1072, + "step": 3742 + }, + { + "epoch": 0.68, + "learning_rate": 2.7983436476155808e-06, + "loss": 1.4482, + "step": 3743 + }, + { + "epoch": 0.68, + "learning_rate": 2.7982890425919183e-06, + "loss": 0.6322, + "step": 3744 + }, + { + "epoch": 0.68, + "learning_rate": 2.7982344375682562e-06, + "loss": 0.4116, + "step": 3745 + }, + { + "epoch": 0.68, + "learning_rate": 2.798179832544594e-06, + "loss": 1.6258, + "step": 3746 + }, + { + "epoch": 0.68, + "learning_rate": 2.798125227520932e-06, + "loss": 0.4029, + "step": 3747 + }, + { + "epoch": 0.68, + "learning_rate": 2.79807062249727e-06, + "loss": 0.4607, + "step": 3748 + }, + { + "epoch": 0.68, + "learning_rate": 2.798016017473608e-06, + "loss": 0.3412, + "step": 3749 + }, + { + "epoch": 0.68, + "learning_rate": 2.7979614124499455e-06, + "loss": 1.0935, + "step": 3750 + }, + { + "epoch": 0.68, + "learning_rate": 2.7979068074262835e-06, + "loss": 0.3066, + "step": 3751 + }, + { + "epoch": 0.68, + "learning_rate": 2.797852202402621e-06, + "loss": 0.3061, + "step": 3752 + }, + { + "epoch": 0.68, + "learning_rate": 2.797797597378959e-06, + "loss": 0.5913, + "step": 3753 + }, + { + "epoch": 0.68, + "learning_rate": 2.797742992355297e-06, + "loss": 1.2685, + "step": 3754 + }, + { + "epoch": 0.68, + "learning_rate": 2.7976883873316344e-06, + "loss": 1.3468, + "step": 3755 + }, + { + "epoch": 0.68, + "learning_rate": 2.7976337823079723e-06, + "loss": 0.8431, + "step": 3756 + }, + { + "epoch": 0.68, + "learning_rate": 2.7975791772843103e-06, + "loss": 0.7795, + "step": 3757 + }, + { + "epoch": 0.68, + "learning_rate": 2.797524572260648e-06, + "loss": 0.0814, + "step": 3758 + }, + { + "epoch": 0.68, + "learning_rate": 2.7974699672369857e-06, + "loss": 1.4304, + "step": 3759 + }, + { + "epoch": 0.68, + "learning_rate": 2.7974153622133237e-06, + "loss": 1.2665, + "step": 3760 + }, + { + "epoch": 0.68, + "learning_rate": 2.7973607571896616e-06, + "loss": 1.4909, + "step": 3761 + }, + { + "epoch": 0.68, + "learning_rate": 2.7973061521659996e-06, + "loss": 0.6081, + "step": 3762 + }, + { + "epoch": 0.68, + "learning_rate": 2.7972515471423375e-06, + "loss": 0.7681, + "step": 3763 + }, + { + "epoch": 0.68, + "learning_rate": 2.797196942118675e-06, + "loss": 1.1428, + "step": 3764 + }, + { + "epoch": 0.68, + "learning_rate": 2.797142337095013e-06, + "loss": 0.443, + "step": 3765 + }, + { + "epoch": 0.68, + "learning_rate": 2.7970877320713505e-06, + "loss": 1.2886, + "step": 3766 + }, + { + "epoch": 0.69, + "learning_rate": 2.7970331270476884e-06, + "loss": 1.1301, + "step": 3767 + }, + { + "epoch": 0.69, + "learning_rate": 2.7969785220240264e-06, + "loss": 0.0473, + "step": 3768 + }, + { + "epoch": 0.69, + "learning_rate": 2.796923917000364e-06, + "loss": 1.6239, + "step": 3769 + }, + { + "epoch": 0.69, + "learning_rate": 2.796869311976702e-06, + "loss": 0.334, + "step": 3770 + }, + { + "epoch": 0.69, + "learning_rate": 2.79681470695304e-06, + "loss": 0.0768, + "step": 3771 + }, + { + "epoch": 0.69, + "learning_rate": 2.7967601019293773e-06, + "loss": 1.0132, + "step": 3772 + }, + { + "epoch": 0.69, + "learning_rate": 2.7967054969057153e-06, + "loss": 0.4696, + "step": 3773 + }, + { + "epoch": 0.69, + "learning_rate": 2.796650891882053e-06, + "loss": 2.029, + "step": 3774 + }, + { + "epoch": 0.69, + "learning_rate": 2.796596286858391e-06, + "loss": 0.0769, + "step": 3775 + }, + { + "epoch": 0.69, + "learning_rate": 2.796541681834729e-06, + "loss": 1.6824, + "step": 3776 + }, + { + "epoch": 0.69, + "learning_rate": 2.7964870768110666e-06, + "loss": 0.6564, + "step": 3777 + }, + { + "epoch": 0.69, + "learning_rate": 2.7964324717874046e-06, + "loss": 2.0001, + "step": 3778 + }, + { + "epoch": 0.69, + "learning_rate": 2.7963778667637425e-06, + "loss": 1.0627, + "step": 3779 + }, + { + "epoch": 0.69, + "learning_rate": 2.79632326174008e-06, + "loss": 1.3478, + "step": 3780 + }, + { + "epoch": 0.69, + "learning_rate": 2.796268656716418e-06, + "loss": 1.7139, + "step": 3781 + }, + { + "epoch": 0.69, + "learning_rate": 2.796214051692756e-06, + "loss": 0.8687, + "step": 3782 + }, + { + "epoch": 0.69, + "learning_rate": 2.7961594466690934e-06, + "loss": 2.0925, + "step": 3783 + }, + { + "epoch": 0.69, + "learning_rate": 2.7961048416454314e-06, + "loss": 2.4921, + "step": 3784 + }, + { + "epoch": 0.69, + "learning_rate": 2.7960502366217693e-06, + "loss": 0.7891, + "step": 3785 + }, + { + "epoch": 0.69, + "learning_rate": 2.795995631598107e-06, + "loss": 2.3069, + "step": 3786 + }, + { + "epoch": 0.69, + "learning_rate": 2.7959410265744448e-06, + "loss": 0.9887, + "step": 3787 + }, + { + "epoch": 0.69, + "learning_rate": 2.7958864215507827e-06, + "loss": 0.5798, + "step": 3788 + }, + { + "epoch": 0.69, + "learning_rate": 2.7958318165271207e-06, + "loss": 1.0558, + "step": 3789 + }, + { + "epoch": 0.69, + "learning_rate": 2.7957772115034586e-06, + "loss": 0.7955, + "step": 3790 + }, + { + "epoch": 0.69, + "learning_rate": 2.795722606479796e-06, + "loss": 1.5251, + "step": 3791 + }, + { + "epoch": 0.69, + "learning_rate": 2.795668001456134e-06, + "loss": 1.8784, + "step": 3792 + }, + { + "epoch": 0.69, + "learning_rate": 2.795613396432472e-06, + "loss": 0.1308, + "step": 3793 + }, + { + "epoch": 0.69, + "learning_rate": 2.7955587914088095e-06, + "loss": 3.1946, + "step": 3794 + }, + { + "epoch": 0.69, + "learning_rate": 2.7955041863851475e-06, + "loss": 1.1494, + "step": 3795 + }, + { + "epoch": 0.69, + "learning_rate": 2.7954495813614854e-06, + "loss": 0.9384, + "step": 3796 + }, + { + "epoch": 0.69, + "learning_rate": 2.795394976337823e-06, + "loss": 0.5254, + "step": 3797 + }, + { + "epoch": 0.69, + "learning_rate": 2.795340371314161e-06, + "loss": 1.4696, + "step": 3798 + }, + { + "epoch": 0.69, + "learning_rate": 2.795285766290499e-06, + "loss": 1.0196, + "step": 3799 + }, + { + "epoch": 0.69, + "learning_rate": 2.7952311612668363e-06, + "loss": 0.6651, + "step": 3800 + }, + { + "epoch": 0.69, + "learning_rate": 2.7951765562431743e-06, + "loss": 0.1098, + "step": 3801 + }, + { + "epoch": 0.69, + "learning_rate": 2.7951219512195122e-06, + "loss": 0.9543, + "step": 3802 + }, + { + "epoch": 0.69, + "learning_rate": 2.79506734619585e-06, + "loss": 0.838, + "step": 3803 + }, + { + "epoch": 0.69, + "learning_rate": 2.795012741172188e-06, + "loss": 0.9435, + "step": 3804 + }, + { + "epoch": 0.69, + "learning_rate": 2.7949581361485256e-06, + "loss": 0.0735, + "step": 3805 + }, + { + "epoch": 0.69, + "learning_rate": 2.7949035311248636e-06, + "loss": 2.7512, + "step": 3806 + }, + { + "epoch": 0.69, + "learning_rate": 2.7948489261012015e-06, + "loss": 0.9843, + "step": 3807 + }, + { + "epoch": 0.69, + "learning_rate": 2.794794321077539e-06, + "loss": 1.4483, + "step": 3808 + }, + { + "epoch": 0.69, + "learning_rate": 2.794739716053877e-06, + "loss": 0.6167, + "step": 3809 + }, + { + "epoch": 0.69, + "learning_rate": 2.794685111030215e-06, + "loss": 1.193, + "step": 3810 + }, + { + "epoch": 0.69, + "learning_rate": 2.7946305060065524e-06, + "loss": 0.5965, + "step": 3811 + }, + { + "epoch": 0.69, + "learning_rate": 2.7945759009828904e-06, + "loss": 0.2946, + "step": 3812 + }, + { + "epoch": 0.69, + "learning_rate": 2.7945212959592283e-06, + "loss": 0.6905, + "step": 3813 + }, + { + "epoch": 0.69, + "learning_rate": 2.794466690935566e-06, + "loss": 1.4992, + "step": 3814 + }, + { + "epoch": 0.69, + "learning_rate": 2.794412085911904e-06, + "loss": 0.9472, + "step": 3815 + }, + { + "epoch": 0.69, + "learning_rate": 2.7943574808882417e-06, + "loss": 0.2275, + "step": 3816 + }, + { + "epoch": 0.69, + "learning_rate": 2.7943028758645797e-06, + "loss": 0.6253, + "step": 3817 + }, + { + "epoch": 0.69, + "learning_rate": 2.7942482708409176e-06, + "loss": 2.5969, + "step": 3818 + }, + { + "epoch": 0.69, + "learning_rate": 2.794193665817255e-06, + "loss": 1.5576, + "step": 3819 + }, + { + "epoch": 0.69, + "learning_rate": 2.794139060793593e-06, + "loss": 0.9546, + "step": 3820 + }, + { + "epoch": 0.69, + "learning_rate": 2.794084455769931e-06, + "loss": 1.051, + "step": 3821 + }, + { + "epoch": 0.7, + "learning_rate": 2.7940298507462686e-06, + "loss": 0.9824, + "step": 3822 + }, + { + "epoch": 0.7, + "learning_rate": 2.7939752457226065e-06, + "loss": 1.2329, + "step": 3823 + }, + { + "epoch": 0.7, + "learning_rate": 2.7939206406989444e-06, + "loss": 0.5843, + "step": 3824 + }, + { + "epoch": 0.7, + "learning_rate": 2.793866035675282e-06, + "loss": 2.0656, + "step": 3825 + }, + { + "epoch": 0.7, + "learning_rate": 2.79381143065162e-06, + "loss": 0.4302, + "step": 3826 + }, + { + "epoch": 0.7, + "learning_rate": 2.793756825627958e-06, + "loss": 2.336, + "step": 3827 + }, + { + "epoch": 0.7, + "learning_rate": 2.793702220604296e-06, + "loss": 1.9958, + "step": 3828 + }, + { + "epoch": 0.7, + "learning_rate": 2.7936476155806337e-06, + "loss": 1.3245, + "step": 3829 + }, + { + "epoch": 0.7, + "learning_rate": 2.7935930105569717e-06, + "loss": 1.1986, + "step": 3830 + }, + { + "epoch": 0.7, + "learning_rate": 2.793538405533309e-06, + "loss": 1.1508, + "step": 3831 + }, + { + "epoch": 0.7, + "learning_rate": 2.793483800509647e-06, + "loss": 2.0145, + "step": 3832 + }, + { + "epoch": 0.7, + "learning_rate": 2.7934291954859847e-06, + "loss": 2.2476, + "step": 3833 + }, + { + "epoch": 0.7, + "learning_rate": 2.7933745904623226e-06, + "loss": 1.3315, + "step": 3834 + }, + { + "epoch": 0.7, + "learning_rate": 2.7933199854386605e-06, + "loss": 0.5736, + "step": 3835 + }, + { + "epoch": 0.7, + "learning_rate": 2.793265380414998e-06, + "loss": 0.3715, + "step": 3836 + }, + { + "epoch": 0.7, + "learning_rate": 2.793210775391336e-06, + "loss": 0.0529, + "step": 3837 + }, + { + "epoch": 0.7, + "learning_rate": 2.793156170367674e-06, + "loss": 1.6567, + "step": 3838 + }, + { + "epoch": 0.7, + "learning_rate": 2.7931015653440115e-06, + "loss": 0.7043, + "step": 3839 + }, + { + "epoch": 0.7, + "learning_rate": 2.7930469603203494e-06, + "loss": 1.6114, + "step": 3840 + }, + { + "epoch": 0.7, + "learning_rate": 2.7929923552966874e-06, + "loss": 1.4401, + "step": 3841 + }, + { + "epoch": 0.7, + "learning_rate": 2.7929377502730253e-06, + "loss": 2.8925, + "step": 3842 + }, + { + "epoch": 0.7, + "learning_rate": 2.7928831452493632e-06, + "loss": 1.5696, + "step": 3843 + }, + { + "epoch": 0.7, + "learning_rate": 2.792828540225701e-06, + "loss": 0.5222, + "step": 3844 + }, + { + "epoch": 0.7, + "learning_rate": 2.7927739352020387e-06, + "loss": 1.0248, + "step": 3845 + }, + { + "epoch": 0.7, + "learning_rate": 2.7927193301783767e-06, + "loss": 0.6642, + "step": 3846 + }, + { + "epoch": 0.7, + "learning_rate": 2.792664725154714e-06, + "loss": 0.1523, + "step": 3847 + }, + { + "epoch": 0.7, + "learning_rate": 2.792610120131052e-06, + "loss": 2.7956, + "step": 3848 + }, + { + "epoch": 0.7, + "learning_rate": 2.79255551510739e-06, + "loss": 1.485, + "step": 3849 + }, + { + "epoch": 0.7, + "learning_rate": 2.7925009100837276e-06, + "loss": 1.0674, + "step": 3850 + }, + { + "epoch": 0.7, + "learning_rate": 2.7924463050600655e-06, + "loss": 2.4855, + "step": 3851 + }, + { + "epoch": 0.7, + "learning_rate": 2.7923917000364035e-06, + "loss": 0.7557, + "step": 3852 + }, + { + "epoch": 0.7, + "learning_rate": 2.792337095012741e-06, + "loss": 0.9247, + "step": 3853 + }, + { + "epoch": 0.7, + "learning_rate": 2.792282489989079e-06, + "loss": 1.2112, + "step": 3854 + }, + { + "epoch": 0.7, + "learning_rate": 2.792227884965417e-06, + "loss": 0.4817, + "step": 3855 + }, + { + "epoch": 0.7, + "learning_rate": 2.792173279941755e-06, + "loss": 1.7177, + "step": 3856 + }, + { + "epoch": 0.7, + "learning_rate": 2.7921186749180928e-06, + "loss": 1.4633, + "step": 3857 + }, + { + "epoch": 0.7, + "learning_rate": 2.7920640698944303e-06, + "loss": 1.2663, + "step": 3858 + }, + { + "epoch": 0.7, + "learning_rate": 2.7920094648707682e-06, + "loss": 1.2615, + "step": 3859 + }, + { + "epoch": 0.7, + "learning_rate": 2.791954859847106e-06, + "loss": 1.9713, + "step": 3860 + }, + { + "epoch": 0.7, + "learning_rate": 2.7919002548234437e-06, + "loss": 1.4162, + "step": 3861 + }, + { + "epoch": 0.7, + "learning_rate": 2.7918456497997816e-06, + "loss": 0.7732, + "step": 3862 + }, + { + "epoch": 0.7, + "learning_rate": 2.7917910447761196e-06, + "loss": 1.6554, + "step": 3863 + }, + { + "epoch": 0.7, + "learning_rate": 2.791736439752457e-06, + "loss": 0.6158, + "step": 3864 + }, + { + "epoch": 0.7, + "learning_rate": 2.791681834728795e-06, + "loss": 1.9357, + "step": 3865 + }, + { + "epoch": 0.7, + "learning_rate": 2.791627229705133e-06, + "loss": 0.5192, + "step": 3866 + }, + { + "epoch": 0.7, + "learning_rate": 2.7915726246814705e-06, + "loss": 1.1512, + "step": 3867 + }, + { + "epoch": 0.7, + "learning_rate": 2.7915180196578084e-06, + "loss": 1.4088, + "step": 3868 + }, + { + "epoch": 0.7, + "learning_rate": 2.7914634146341464e-06, + "loss": 0.0267, + "step": 3869 + }, + { + "epoch": 0.7, + "learning_rate": 2.7914088096104843e-06, + "loss": 0.8123, + "step": 3870 + }, + { + "epoch": 0.7, + "learning_rate": 2.7913542045868223e-06, + "loss": 0.2959, + "step": 3871 + }, + { + "epoch": 0.7, + "learning_rate": 2.79129959956316e-06, + "loss": 1.6381, + "step": 3872 + }, + { + "epoch": 0.7, + "learning_rate": 2.7912449945394977e-06, + "loss": 0.9012, + "step": 3873 + }, + { + "epoch": 0.7, + "learning_rate": 2.7911903895158357e-06, + "loss": 0.6993, + "step": 3874 + }, + { + "epoch": 0.7, + "learning_rate": 2.791135784492173e-06, + "loss": 3.5317, + "step": 3875 + }, + { + "epoch": 0.7, + "learning_rate": 2.791081179468511e-06, + "loss": 0.5695, + "step": 3876 + }, + { + "epoch": 0.71, + "learning_rate": 2.791026574444849e-06, + "loss": 1.6293, + "step": 3877 + }, + { + "epoch": 0.71, + "learning_rate": 2.7909719694211866e-06, + "loss": 1.0568, + "step": 3878 + }, + { + "epoch": 0.71, + "learning_rate": 2.7909173643975246e-06, + "loss": 0.4475, + "step": 3879 + }, + { + "epoch": 0.71, + "learning_rate": 2.7908627593738625e-06, + "loss": 0.4705, + "step": 3880 + }, + { + "epoch": 0.71, + "learning_rate": 2.7908081543502e-06, + "loss": 0.7538, + "step": 3881 + }, + { + "epoch": 0.71, + "learning_rate": 2.790753549326538e-06, + "loss": 0.8433, + "step": 3882 + }, + { + "epoch": 0.71, + "learning_rate": 2.790698944302876e-06, + "loss": 1.7945, + "step": 3883 + }, + { + "epoch": 0.71, + "learning_rate": 2.790644339279214e-06, + "loss": 0.5268, + "step": 3884 + }, + { + "epoch": 0.71, + "learning_rate": 2.790589734255552e-06, + "loss": 0.1914, + "step": 3885 + }, + { + "epoch": 0.71, + "learning_rate": 2.7905351292318893e-06, + "loss": 0.1361, + "step": 3886 + }, + { + "epoch": 0.71, + "learning_rate": 2.7904805242082273e-06, + "loss": 2.5633, + "step": 3887 + }, + { + "epoch": 0.71, + "learning_rate": 2.790425919184565e-06, + "loss": 1.2663, + "step": 3888 + }, + { + "epoch": 0.71, + "learning_rate": 2.7903713141609027e-06, + "loss": 0.8569, + "step": 3889 + }, + { + "epoch": 0.71, + "learning_rate": 2.7903167091372407e-06, + "loss": 0.8088, + "step": 3890 + }, + { + "epoch": 0.71, + "learning_rate": 2.7902621041135786e-06, + "loss": 1.7674, + "step": 3891 + }, + { + "epoch": 0.71, + "learning_rate": 2.790207499089916e-06, + "loss": 1.1444, + "step": 3892 + }, + { + "epoch": 0.71, + "learning_rate": 2.790152894066254e-06, + "loss": 0.0485, + "step": 3893 + }, + { + "epoch": 0.71, + "learning_rate": 2.790098289042592e-06, + "loss": 1.7871, + "step": 3894 + }, + { + "epoch": 0.71, + "learning_rate": 2.7900436840189295e-06, + "loss": 2.1036, + "step": 3895 + }, + { + "epoch": 0.71, + "learning_rate": 2.7899890789952675e-06, + "loss": 0.0418, + "step": 3896 + }, + { + "epoch": 0.71, + "learning_rate": 2.7899344739716054e-06, + "loss": 0.8181, + "step": 3897 + }, + { + "epoch": 0.71, + "learning_rate": 2.7898798689479434e-06, + "loss": 1.0414, + "step": 3898 + }, + { + "epoch": 0.71, + "learning_rate": 2.7898252639242813e-06, + "loss": 0.062, + "step": 3899 + }, + { + "epoch": 0.71, + "learning_rate": 2.789770658900619e-06, + "loss": 0.706, + "step": 3900 + }, + { + "epoch": 0.71, + "learning_rate": 2.7897160538769568e-06, + "loss": 0.6214, + "step": 3901 + }, + { + "epoch": 0.71, + "learning_rate": 2.7896614488532947e-06, + "loss": 0.2327, + "step": 3902 + }, + { + "epoch": 0.71, + "learning_rate": 2.7896068438296322e-06, + "loss": 0.8582, + "step": 3903 + }, + { + "epoch": 0.71, + "learning_rate": 2.78955223880597e-06, + "loss": 1.2524, + "step": 3904 + }, + { + "epoch": 0.71, + "learning_rate": 2.789497633782308e-06, + "loss": 0.1635, + "step": 3905 + }, + { + "epoch": 0.71, + "learning_rate": 2.7894430287586456e-06, + "loss": 0.701, + "step": 3906 + }, + { + "epoch": 0.71, + "learning_rate": 2.7893884237349836e-06, + "loss": 0.3708, + "step": 3907 + }, + { + "epoch": 0.71, + "learning_rate": 2.7893338187113215e-06, + "loss": 0.3543, + "step": 3908 + }, + { + "epoch": 0.71, + "learning_rate": 2.7892792136876595e-06, + "loss": 0.058, + "step": 3909 + }, + { + "epoch": 0.71, + "learning_rate": 2.7892246086639974e-06, + "loss": 2.1672, + "step": 3910 + }, + { + "epoch": 0.71, + "learning_rate": 2.7891700036403354e-06, + "loss": 0.745, + "step": 3911 + }, + { + "epoch": 0.71, + "learning_rate": 2.789115398616673e-06, + "loss": 0.7237, + "step": 3912 + }, + { + "epoch": 0.71, + "learning_rate": 2.789060793593011e-06, + "loss": 2.0731, + "step": 3913 + }, + { + "epoch": 0.71, + "learning_rate": 2.7890061885693483e-06, + "loss": 0.934, + "step": 3914 + }, + { + "epoch": 0.71, + "learning_rate": 2.7889515835456863e-06, + "loss": 0.8935, + "step": 3915 + }, + { + "epoch": 0.71, + "learning_rate": 2.7888969785220242e-06, + "loss": 1.1692, + "step": 3916 + }, + { + "epoch": 0.71, + "learning_rate": 2.7888423734983617e-06, + "loss": 0.7084, + "step": 3917 + }, + { + "epoch": 0.71, + "learning_rate": 2.7887877684746997e-06, + "loss": 1.9708, + "step": 3918 + }, + { + "epoch": 0.71, + "learning_rate": 2.7887331634510376e-06, + "loss": 0.409, + "step": 3919 + }, + { + "epoch": 0.71, + "learning_rate": 2.788678558427375e-06, + "loss": 1.1123, + "step": 3920 + }, + { + "epoch": 0.71, + "learning_rate": 2.788623953403713e-06, + "loss": 0.0769, + "step": 3921 + }, + { + "epoch": 0.71, + "learning_rate": 2.788569348380051e-06, + "loss": 1.0697, + "step": 3922 + }, + { + "epoch": 0.71, + "learning_rate": 2.788514743356389e-06, + "loss": 0.4657, + "step": 3923 + }, + { + "epoch": 0.71, + "learning_rate": 2.788460138332727e-06, + "loss": 0.9553, + "step": 3924 + }, + { + "epoch": 0.71, + "learning_rate": 2.788405533309065e-06, + "loss": 0.7162, + "step": 3925 + }, + { + "epoch": 0.71, + "learning_rate": 2.7883509282854024e-06, + "loss": 0.8511, + "step": 3926 + }, + { + "epoch": 0.71, + "learning_rate": 2.7882963232617403e-06, + "loss": 1.1487, + "step": 3927 + }, + { + "epoch": 0.71, + "learning_rate": 2.788241718238078e-06, + "loss": 0.8886, + "step": 3928 + }, + { + "epoch": 0.71, + "learning_rate": 2.788187113214416e-06, + "loss": 1.7891, + "step": 3929 + }, + { + "epoch": 0.71, + "learning_rate": 2.7881325081907537e-06, + "loss": 0.8169, + "step": 3930 + }, + { + "epoch": 0.71, + "learning_rate": 2.7880779031670913e-06, + "loss": 0.122, + "step": 3931 + }, + { + "epoch": 0.72, + "learning_rate": 2.788023298143429e-06, + "loss": 0.7266, + "step": 3932 + }, + { + "epoch": 0.72, + "learning_rate": 2.787968693119767e-06, + "loss": 0.4994, + "step": 3933 + }, + { + "epoch": 0.72, + "learning_rate": 2.7879140880961047e-06, + "loss": 1.3062, + "step": 3934 + }, + { + "epoch": 0.72, + "learning_rate": 2.7878594830724426e-06, + "loss": 0.0979, + "step": 3935 + }, + { + "epoch": 0.72, + "learning_rate": 2.7878048780487806e-06, + "loss": 2.1142, + "step": 3936 + }, + { + "epoch": 0.72, + "learning_rate": 2.7877502730251185e-06, + "loss": 0.8546, + "step": 3937 + }, + { + "epoch": 0.72, + "learning_rate": 2.7876956680014564e-06, + "loss": 0.5928, + "step": 3938 + }, + { + "epoch": 0.72, + "learning_rate": 2.787641062977794e-06, + "loss": 1.2339, + "step": 3939 + }, + { + "epoch": 0.72, + "learning_rate": 2.787586457954132e-06, + "loss": 0.4998, + "step": 3940 + }, + { + "epoch": 0.72, + "learning_rate": 2.78753185293047e-06, + "loss": 0.6988, + "step": 3941 + }, + { + "epoch": 0.72, + "learning_rate": 2.7874772479068074e-06, + "loss": 0.3539, + "step": 3942 + }, + { + "epoch": 0.72, + "learning_rate": 2.7874226428831453e-06, + "loss": 1.1527, + "step": 3943 + }, + { + "epoch": 0.72, + "learning_rate": 2.7873680378594833e-06, + "loss": 1.6511, + "step": 3944 + }, + { + "epoch": 0.72, + "learning_rate": 2.7873134328358208e-06, + "loss": 0.0821, + "step": 3945 + }, + { + "epoch": 0.72, + "learning_rate": 2.7872588278121587e-06, + "loss": 0.526, + "step": 3946 + }, + { + "epoch": 0.72, + "learning_rate": 2.7872042227884967e-06, + "loss": 0.793, + "step": 3947 + }, + { + "epoch": 0.72, + "learning_rate": 2.787149617764834e-06, + "loss": 1.8984, + "step": 3948 + }, + { + "epoch": 0.72, + "learning_rate": 2.787095012741172e-06, + "loss": 2.2127, + "step": 3949 + }, + { + "epoch": 0.72, + "learning_rate": 2.78704040771751e-06, + "loss": 1.1807, + "step": 3950 + }, + { + "epoch": 0.72, + "learning_rate": 2.786985802693848e-06, + "loss": 0.4588, + "step": 3951 + }, + { + "epoch": 0.72, + "learning_rate": 2.786931197670186e-06, + "loss": 2.5588, + "step": 3952 + }, + { + "epoch": 0.72, + "learning_rate": 2.7868765926465235e-06, + "loss": 0.7428, + "step": 3953 + }, + { + "epoch": 0.72, + "learning_rate": 2.7868219876228614e-06, + "loss": 0.8595, + "step": 3954 + }, + { + "epoch": 0.72, + "learning_rate": 2.7867673825991994e-06, + "loss": 1.9668, + "step": 3955 + }, + { + "epoch": 0.72, + "learning_rate": 2.786712777575537e-06, + "loss": 0.7331, + "step": 3956 + }, + { + "epoch": 0.72, + "learning_rate": 2.786658172551875e-06, + "loss": 2.1322, + "step": 3957 + }, + { + "epoch": 0.72, + "learning_rate": 2.7866035675282128e-06, + "loss": 0.1144, + "step": 3958 + }, + { + "epoch": 0.72, + "learning_rate": 2.7865489625045503e-06, + "loss": 3.6387, + "step": 3959 + }, + { + "epoch": 0.72, + "learning_rate": 2.7864943574808882e-06, + "loss": 3.9484, + "step": 3960 + }, + { + "epoch": 0.72, + "learning_rate": 2.786439752457226e-06, + "loss": 0.0992, + "step": 3961 + }, + { + "epoch": 0.72, + "learning_rate": 2.7863851474335637e-06, + "loss": 2.0838, + "step": 3962 + }, + { + "epoch": 0.72, + "learning_rate": 2.7863305424099016e-06, + "loss": 1.3818, + "step": 3963 + }, + { + "epoch": 0.72, + "learning_rate": 2.7862759373862396e-06, + "loss": 0.1197, + "step": 3964 + }, + { + "epoch": 0.72, + "learning_rate": 2.7862213323625775e-06, + "loss": 0.5535, + "step": 3965 + }, + { + "epoch": 0.72, + "learning_rate": 2.7861667273389155e-06, + "loss": 0.6173, + "step": 3966 + }, + { + "epoch": 0.72, + "learning_rate": 2.786112122315253e-06, + "loss": 2.9423, + "step": 3967 + }, + { + "epoch": 0.72, + "learning_rate": 2.786057517291591e-06, + "loss": 0.2184, + "step": 3968 + }, + { + "epoch": 0.72, + "learning_rate": 2.786002912267929e-06, + "loss": 1.6123, + "step": 3969 + }, + { + "epoch": 0.72, + "learning_rate": 2.7859483072442664e-06, + "loss": 0.599, + "step": 3970 + }, + { + "epoch": 0.72, + "learning_rate": 2.7858937022206043e-06, + "loss": 1.695, + "step": 3971 + }, + { + "epoch": 0.72, + "learning_rate": 2.7858390971969423e-06, + "loss": 1.9292, + "step": 3972 + }, + { + "epoch": 0.72, + "learning_rate": 2.78578449217328e-06, + "loss": 0.8501, + "step": 3973 + }, + { + "epoch": 0.72, + "learning_rate": 2.7857298871496177e-06, + "loss": 1.3777, + "step": 3974 + }, + { + "epoch": 0.72, + "learning_rate": 2.7856752821259557e-06, + "loss": 1.0206, + "step": 3975 + }, + { + "epoch": 0.72, + "learning_rate": 2.785620677102293e-06, + "loss": 0.6757, + "step": 3976 + }, + { + "epoch": 0.72, + "learning_rate": 2.785566072078631e-06, + "loss": 0.09, + "step": 3977 + }, + { + "epoch": 0.72, + "learning_rate": 2.785511467054969e-06, + "loss": 1.6801, + "step": 3978 + }, + { + "epoch": 0.72, + "learning_rate": 2.785456862031307e-06, + "loss": 2.0303, + "step": 3979 + }, + { + "epoch": 0.72, + "learning_rate": 2.785402257007645e-06, + "loss": 2.5293, + "step": 3980 + }, + { + "epoch": 0.72, + "learning_rate": 2.7853476519839825e-06, + "loss": 3.606, + "step": 3981 + }, + { + "epoch": 0.72, + "learning_rate": 2.7852930469603204e-06, + "loss": 1.3997, + "step": 3982 + }, + { + "epoch": 0.72, + "learning_rate": 2.7852384419366584e-06, + "loss": 1.437, + "step": 3983 + }, + { + "epoch": 0.72, + "learning_rate": 2.785183836912996e-06, + "loss": 2.4472, + "step": 3984 + }, + { + "epoch": 0.72, + "learning_rate": 2.785129231889334e-06, + "loss": 2.5842, + "step": 3985 + }, + { + "epoch": 0.72, + "learning_rate": 2.785074626865672e-06, + "loss": 0.3344, + "step": 3986 + }, + { + "epoch": 0.73, + "learning_rate": 2.7850200218420093e-06, + "loss": 0.5068, + "step": 3987 + }, + { + "epoch": 0.73, + "learning_rate": 2.7849654168183473e-06, + "loss": 1.7267, + "step": 3988 + }, + { + "epoch": 0.73, + "learning_rate": 2.784910811794685e-06, + "loss": 2.3376, + "step": 3989 + }, + { + "epoch": 0.73, + "learning_rate": 2.7848562067710227e-06, + "loss": 1.2658, + "step": 3990 + }, + { + "epoch": 0.73, + "learning_rate": 2.7848016017473607e-06, + "loss": 0.3319, + "step": 3991 + }, + { + "epoch": 0.73, + "learning_rate": 2.7847469967236986e-06, + "loss": 2.1793, + "step": 3992 + }, + { + "epoch": 0.73, + "learning_rate": 2.7846923917000365e-06, + "loss": 0.8321, + "step": 3993 + }, + { + "epoch": 0.73, + "learning_rate": 2.7846377866763745e-06, + "loss": 2.227, + "step": 3994 + }, + { + "epoch": 0.73, + "learning_rate": 2.784583181652712e-06, + "loss": 0.1478, + "step": 3995 + }, + { + "epoch": 0.73, + "learning_rate": 2.78452857662905e-06, + "loss": 1.6857, + "step": 3996 + }, + { + "epoch": 0.73, + "learning_rate": 2.784473971605388e-06, + "loss": 0.0456, + "step": 3997 + }, + { + "epoch": 0.73, + "learning_rate": 2.7844193665817254e-06, + "loss": 0.2059, + "step": 3998 + }, + { + "epoch": 0.73, + "learning_rate": 2.7843647615580634e-06, + "loss": 0.246, + "step": 3999 + }, + { + "epoch": 0.73, + "learning_rate": 2.7843101565344013e-06, + "loss": 0.7199, + "step": 4000 + }, + { + "epoch": 0.73, + "learning_rate": 2.784255551510739e-06, + "loss": 0.5252, + "step": 4001 + }, + { + "epoch": 0.73, + "learning_rate": 2.7842009464870768e-06, + "loss": 0.4011, + "step": 4002 + }, + { + "epoch": 0.73, + "learning_rate": 2.7841463414634147e-06, + "loss": 1.0774, + "step": 4003 + }, + { + "epoch": 0.73, + "learning_rate": 2.7840917364397527e-06, + "loss": 1.1597, + "step": 4004 + }, + { + "epoch": 0.73, + "learning_rate": 2.7840371314160906e-06, + "loss": 1.0269, + "step": 4005 + }, + { + "epoch": 0.73, + "learning_rate": 2.7839825263924285e-06, + "loss": 0.924, + "step": 4006 + }, + { + "epoch": 0.73, + "learning_rate": 2.783927921368766e-06, + "loss": 1.3939, + "step": 4007 + }, + { + "epoch": 0.73, + "learning_rate": 2.783873316345104e-06, + "loss": 0.5852, + "step": 4008 + }, + { + "epoch": 0.73, + "learning_rate": 2.7838187113214415e-06, + "loss": 1.7527, + "step": 4009 + }, + { + "epoch": 0.73, + "learning_rate": 2.7837641062977795e-06, + "loss": 2.1453, + "step": 4010 + }, + { + "epoch": 0.73, + "learning_rate": 2.7837095012741174e-06, + "loss": 1.1524, + "step": 4011 + }, + { + "epoch": 0.73, + "learning_rate": 2.783654896250455e-06, + "loss": 0.6642, + "step": 4012 + }, + { + "epoch": 0.73, + "learning_rate": 2.783600291226793e-06, + "loss": 1.1573, + "step": 4013 + }, + { + "epoch": 0.73, + "learning_rate": 2.783545686203131e-06, + "loss": 1.3688, + "step": 4014 + }, + { + "epoch": 0.73, + "learning_rate": 2.7834910811794683e-06, + "loss": 3.624, + "step": 4015 + }, + { + "epoch": 0.73, + "learning_rate": 2.7834364761558063e-06, + "loss": 1.6909, + "step": 4016 + }, + { + "epoch": 0.73, + "learning_rate": 2.7833818711321442e-06, + "loss": 1.6756, + "step": 4017 + }, + { + "epoch": 0.73, + "learning_rate": 2.783327266108482e-06, + "loss": 1.6548, + "step": 4018 + }, + { + "epoch": 0.73, + "learning_rate": 2.78327266108482e-06, + "loss": 0.7349, + "step": 4019 + }, + { + "epoch": 0.73, + "learning_rate": 2.7832180560611576e-06, + "loss": 0.8365, + "step": 4020 + }, + { + "epoch": 0.73, + "learning_rate": 2.7831634510374956e-06, + "loss": 1.7373, + "step": 4021 + }, + { + "epoch": 0.73, + "learning_rate": 2.7831088460138335e-06, + "loss": 2.2689, + "step": 4022 + }, + { + "epoch": 0.73, + "learning_rate": 2.783054240990171e-06, + "loss": 0.0412, + "step": 4023 + }, + { + "epoch": 0.73, + "learning_rate": 2.782999635966509e-06, + "loss": 0.3521, + "step": 4024 + }, + { + "epoch": 0.73, + "learning_rate": 2.782945030942847e-06, + "loss": 2.8728, + "step": 4025 + }, + { + "epoch": 0.73, + "learning_rate": 2.7828904259191844e-06, + "loss": 0.233, + "step": 4026 + }, + { + "epoch": 0.73, + "learning_rate": 2.7828358208955224e-06, + "loss": 0.462, + "step": 4027 + }, + { + "epoch": 0.73, + "learning_rate": 2.7827812158718603e-06, + "loss": 1.1099, + "step": 4028 + }, + { + "epoch": 0.73, + "learning_rate": 2.782726610848198e-06, + "loss": 0.7707, + "step": 4029 + }, + { + "epoch": 0.73, + "learning_rate": 2.782672005824536e-06, + "loss": 0.1299, + "step": 4030 + }, + { + "epoch": 0.73, + "learning_rate": 2.7826174008008737e-06, + "loss": 0.4699, + "step": 4031 + }, + { + "epoch": 0.73, + "learning_rate": 2.7825627957772117e-06, + "loss": 2.2989, + "step": 4032 + }, + { + "epoch": 0.73, + "learning_rate": 2.7825081907535496e-06, + "loss": 0.1115, + "step": 4033 + }, + { + "epoch": 0.73, + "learning_rate": 2.782453585729887e-06, + "loss": 1.563, + "step": 4034 + }, + { + "epoch": 0.73, + "learning_rate": 2.782398980706225e-06, + "loss": 2.1335, + "step": 4035 + }, + { + "epoch": 0.73, + "learning_rate": 2.782344375682563e-06, + "loss": 2.0489, + "step": 4036 + }, + { + "epoch": 0.73, + "learning_rate": 2.7822897706589006e-06, + "loss": 1.2684, + "step": 4037 + }, + { + "epoch": 0.73, + "learning_rate": 2.7822351656352385e-06, + "loss": 0.589, + "step": 4038 + }, + { + "epoch": 0.73, + "learning_rate": 2.7821805606115764e-06, + "loss": 0.6116, + "step": 4039 + }, + { + "epoch": 0.73, + "learning_rate": 2.782125955587914e-06, + "loss": 2.8245, + "step": 4040 + }, + { + "epoch": 0.73, + "learning_rate": 2.782071350564252e-06, + "loss": 1.1249, + "step": 4041 + }, + { + "epoch": 0.74, + "learning_rate": 2.78201674554059e-06, + "loss": 0.1964, + "step": 4042 + }, + { + "epoch": 0.74, + "learning_rate": 2.7819621405169274e-06, + "loss": 0.5033, + "step": 4043 + }, + { + "epoch": 0.74, + "learning_rate": 2.7819075354932653e-06, + "loss": 2.9752, + "step": 4044 + }, + { + "epoch": 0.74, + "learning_rate": 2.7818529304696033e-06, + "loss": 0.4398, + "step": 4045 + }, + { + "epoch": 0.74, + "learning_rate": 2.781798325445941e-06, + "loss": 2.0734, + "step": 4046 + }, + { + "epoch": 0.74, + "learning_rate": 2.781743720422279e-06, + "loss": 1.3706, + "step": 4047 + }, + { + "epoch": 0.74, + "learning_rate": 2.7816891153986167e-06, + "loss": 0.7064, + "step": 4048 + }, + { + "epoch": 0.74, + "learning_rate": 2.7816345103749546e-06, + "loss": 1.2653, + "step": 4049 + }, + { + "epoch": 0.74, + "learning_rate": 2.7815799053512925e-06, + "loss": 0.1865, + "step": 4050 + }, + { + "epoch": 0.74, + "learning_rate": 2.78152530032763e-06, + "loss": 1.5819, + "step": 4051 + }, + { + "epoch": 0.74, + "learning_rate": 2.781470695303968e-06, + "loss": 2.113, + "step": 4052 + }, + { + "epoch": 0.74, + "learning_rate": 2.781416090280306e-06, + "loss": 2.7724, + "step": 4053 + }, + { + "epoch": 0.74, + "learning_rate": 2.7813614852566435e-06, + "loss": 1.3925, + "step": 4054 + }, + { + "epoch": 0.74, + "learning_rate": 2.7813068802329814e-06, + "loss": 0.777, + "step": 4055 + }, + { + "epoch": 0.74, + "learning_rate": 2.7812522752093194e-06, + "loss": 1.5158, + "step": 4056 + }, + { + "epoch": 0.74, + "learning_rate": 2.781197670185657e-06, + "loss": 0.9722, + "step": 4057 + }, + { + "epoch": 0.74, + "learning_rate": 2.781143065161995e-06, + "loss": 1.1555, + "step": 4058 + }, + { + "epoch": 0.74, + "learning_rate": 2.7810884601383328e-06, + "loss": 1.07, + "step": 4059 + }, + { + "epoch": 0.74, + "learning_rate": 2.7810338551146707e-06, + "loss": 0.2779, + "step": 4060 + }, + { + "epoch": 0.74, + "learning_rate": 2.7809792500910087e-06, + "loss": 0.3316, + "step": 4061 + }, + { + "epoch": 0.74, + "learning_rate": 2.780924645067346e-06, + "loss": 0.5029, + "step": 4062 + }, + { + "epoch": 0.74, + "learning_rate": 2.780870040043684e-06, + "loss": 0.9675, + "step": 4063 + }, + { + "epoch": 0.74, + "learning_rate": 2.780815435020022e-06, + "loss": 0.2553, + "step": 4064 + }, + { + "epoch": 0.74, + "learning_rate": 2.7807608299963596e-06, + "loss": 0.1133, + "step": 4065 + }, + { + "epoch": 0.74, + "learning_rate": 2.7807062249726975e-06, + "loss": 0.5982, + "step": 4066 + }, + { + "epoch": 0.74, + "learning_rate": 2.7806516199490355e-06, + "loss": 0.1136, + "step": 4067 + }, + { + "epoch": 0.74, + "learning_rate": 2.780597014925373e-06, + "loss": 1.6144, + "step": 4068 + }, + { + "epoch": 0.74, + "learning_rate": 2.780542409901711e-06, + "loss": 0.0835, + "step": 4069 + }, + { + "epoch": 0.74, + "learning_rate": 2.780487804878049e-06, + "loss": 0.8359, + "step": 4070 + }, + { + "epoch": 0.74, + "learning_rate": 2.7804331998543864e-06, + "loss": 1.0261, + "step": 4071 + }, + { + "epoch": 0.74, + "learning_rate": 2.7803785948307243e-06, + "loss": 0.7699, + "step": 4072 + }, + { + "epoch": 0.74, + "learning_rate": 2.7803239898070623e-06, + "loss": 1.1827, + "step": 4073 + }, + { + "epoch": 0.74, + "learning_rate": 2.7802693847834002e-06, + "loss": 0.882, + "step": 4074 + }, + { + "epoch": 0.74, + "learning_rate": 2.780214779759738e-06, + "loss": 1.2573, + "step": 4075 + }, + { + "epoch": 0.74, + "learning_rate": 2.7801601747360757e-06, + "loss": 0.3704, + "step": 4076 + }, + { + "epoch": 0.74, + "learning_rate": 2.7801055697124136e-06, + "loss": 0.8345, + "step": 4077 + }, + { + "epoch": 0.74, + "learning_rate": 2.7800509646887516e-06, + "loss": 0.4333, + "step": 4078 + }, + { + "epoch": 0.74, + "learning_rate": 2.779996359665089e-06, + "loss": 2.2255, + "step": 4079 + }, + { + "epoch": 0.74, + "learning_rate": 2.779941754641427e-06, + "loss": 2.1884, + "step": 4080 + }, + { + "epoch": 0.74, + "learning_rate": 2.779887149617765e-06, + "loss": 1.2595, + "step": 4081 + }, + { + "epoch": 0.74, + "learning_rate": 2.7798325445941025e-06, + "loss": 1.6328, + "step": 4082 + }, + { + "epoch": 0.74, + "learning_rate": 2.7797779395704404e-06, + "loss": 1.0626, + "step": 4083 + }, + { + "epoch": 0.74, + "learning_rate": 2.7797233345467784e-06, + "loss": 0.7479, + "step": 4084 + }, + { + "epoch": 0.74, + "learning_rate": 2.7796687295231163e-06, + "loss": 0.5841, + "step": 4085 + }, + { + "epoch": 0.74, + "learning_rate": 2.7796141244994543e-06, + "loss": 1.4472, + "step": 4086 + }, + { + "epoch": 0.74, + "learning_rate": 2.779559519475792e-06, + "loss": 0.5723, + "step": 4087 + }, + { + "epoch": 0.74, + "learning_rate": 2.7795049144521297e-06, + "loss": 0.2515, + "step": 4088 + }, + { + "epoch": 0.74, + "learning_rate": 2.7794503094284677e-06, + "loss": 0.3253, + "step": 4089 + }, + { + "epoch": 0.74, + "learning_rate": 2.779395704404805e-06, + "loss": 1.2332, + "step": 4090 + }, + { + "epoch": 0.74, + "learning_rate": 2.779341099381143e-06, + "loss": 1.1148, + "step": 4091 + }, + { + "epoch": 0.74, + "learning_rate": 2.779286494357481e-06, + "loss": 0.1923, + "step": 4092 + }, + { + "epoch": 0.74, + "learning_rate": 2.7792318893338186e-06, + "loss": 0.4052, + "step": 4093 + }, + { + "epoch": 0.74, + "learning_rate": 2.7791772843101565e-06, + "loss": 1.2768, + "step": 4094 + }, + { + "epoch": 0.74, + "learning_rate": 2.7791226792864945e-06, + "loss": 0.9107, + "step": 4095 + }, + { + "epoch": 0.74, + "learning_rate": 2.779068074262832e-06, + "loss": 3.724, + "step": 4096 + }, + { + "epoch": 0.75, + "learning_rate": 2.77901346923917e-06, + "loss": 2.0155, + "step": 4097 + }, + { + "epoch": 0.75, + "learning_rate": 2.778958864215508e-06, + "loss": 0.9711, + "step": 4098 + }, + { + "epoch": 0.75, + "learning_rate": 2.778904259191846e-06, + "loss": 1.7292, + "step": 4099 + }, + { + "epoch": 0.75, + "learning_rate": 2.7788496541681838e-06, + "loss": 1.8838, + "step": 4100 + }, + { + "epoch": 0.75, + "learning_rate": 2.7787950491445213e-06, + "loss": 0.6856, + "step": 4101 + }, + { + "epoch": 0.75, + "learning_rate": 2.7787404441208592e-06, + "loss": 0.5632, + "step": 4102 + }, + { + "epoch": 0.75, + "learning_rate": 2.778685839097197e-06, + "loss": 1.1868, + "step": 4103 + }, + { + "epoch": 0.75, + "learning_rate": 2.7786312340735347e-06, + "loss": 1.2313, + "step": 4104 + }, + { + "epoch": 0.75, + "learning_rate": 2.7785766290498727e-06, + "loss": 1.3221, + "step": 4105 + }, + { + "epoch": 0.75, + "learning_rate": 2.7785220240262106e-06, + "loss": 0.861, + "step": 4106 + }, + { + "epoch": 0.75, + "learning_rate": 2.778467419002548e-06, + "loss": 1.288, + "step": 4107 + }, + { + "epoch": 0.75, + "learning_rate": 2.778412813978886e-06, + "loss": 0.8081, + "step": 4108 + }, + { + "epoch": 0.75, + "learning_rate": 2.778358208955224e-06, + "loss": 1.9387, + "step": 4109 + }, + { + "epoch": 0.75, + "learning_rate": 2.7783036039315615e-06, + "loss": 1.2323, + "step": 4110 + }, + { + "epoch": 0.75, + "learning_rate": 2.7782489989078995e-06, + "loss": 0.3086, + "step": 4111 + }, + { + "epoch": 0.75, + "learning_rate": 2.7781943938842374e-06, + "loss": 1.5292, + "step": 4112 + }, + { + "epoch": 0.75, + "learning_rate": 2.7781397888605754e-06, + "loss": 1.3376, + "step": 4113 + }, + { + "epoch": 0.75, + "learning_rate": 2.7780851838369133e-06, + "loss": 0.6524, + "step": 4114 + }, + { + "epoch": 0.75, + "learning_rate": 2.778030578813251e-06, + "loss": 1.6876, + "step": 4115 + }, + { + "epoch": 0.75, + "learning_rate": 2.7779759737895888e-06, + "loss": 0.8905, + "step": 4116 + }, + { + "epoch": 0.75, + "learning_rate": 2.7779213687659267e-06, + "loss": 0.741, + "step": 4117 + }, + { + "epoch": 0.75, + "learning_rate": 2.7778667637422642e-06, + "loss": 1.0823, + "step": 4118 + }, + { + "epoch": 0.75, + "learning_rate": 2.777812158718602e-06, + "loss": 0.5891, + "step": 4119 + }, + { + "epoch": 0.75, + "learning_rate": 2.77775755369494e-06, + "loss": 1.3684, + "step": 4120 + }, + { + "epoch": 0.75, + "learning_rate": 2.7777029486712776e-06, + "loss": 2.217, + "step": 4121 + }, + { + "epoch": 0.75, + "learning_rate": 2.7776483436476156e-06, + "loss": 0.8849, + "step": 4122 + }, + { + "epoch": 0.75, + "learning_rate": 2.7775937386239535e-06, + "loss": 1.6222, + "step": 4123 + }, + { + "epoch": 0.75, + "learning_rate": 2.777539133600291e-06, + "loss": 0.4329, + "step": 4124 + }, + { + "epoch": 0.75, + "learning_rate": 2.777484528576629e-06, + "loss": 2.484, + "step": 4125 + }, + { + "epoch": 0.75, + "learning_rate": 2.777429923552967e-06, + "loss": 0.5006, + "step": 4126 + }, + { + "epoch": 0.75, + "learning_rate": 2.777375318529305e-06, + "loss": 0.1532, + "step": 4127 + }, + { + "epoch": 0.75, + "learning_rate": 2.777320713505643e-06, + "loss": 0.0702, + "step": 4128 + }, + { + "epoch": 0.75, + "learning_rate": 2.7772661084819803e-06, + "loss": 0.0436, + "step": 4129 + }, + { + "epoch": 0.75, + "learning_rate": 2.7772115034583183e-06, + "loss": 1.804, + "step": 4130 + }, + { + "epoch": 0.75, + "learning_rate": 2.7771568984346562e-06, + "loss": 0.3977, + "step": 4131 + }, + { + "epoch": 0.75, + "learning_rate": 2.7771022934109937e-06, + "loss": 0.58, + "step": 4132 + }, + { + "epoch": 0.75, + "learning_rate": 2.7770476883873317e-06, + "loss": 2.0006, + "step": 4133 + }, + { + "epoch": 0.75, + "learning_rate": 2.7769930833636696e-06, + "loss": 1.0217, + "step": 4134 + }, + { + "epoch": 0.75, + "learning_rate": 2.776938478340007e-06, + "loss": 1.201, + "step": 4135 + }, + { + "epoch": 0.75, + "learning_rate": 2.776883873316345e-06, + "loss": 1.022, + "step": 4136 + }, + { + "epoch": 0.75, + "learning_rate": 2.776829268292683e-06, + "loss": 1.2338, + "step": 4137 + }, + { + "epoch": 0.75, + "learning_rate": 2.7767746632690206e-06, + "loss": 1.362, + "step": 4138 + }, + { + "epoch": 0.75, + "learning_rate": 2.7767200582453585e-06, + "loss": 0.1607, + "step": 4139 + }, + { + "epoch": 0.75, + "learning_rate": 2.7766654532216964e-06, + "loss": 0.0792, + "step": 4140 + }, + { + "epoch": 0.75, + "learning_rate": 2.7766108481980344e-06, + "loss": 1.3272, + "step": 4141 + }, + { + "epoch": 0.75, + "learning_rate": 2.7765562431743723e-06, + "loss": 1.4613, + "step": 4142 + }, + { + "epoch": 0.75, + "learning_rate": 2.77650163815071e-06, + "loss": 1.1601, + "step": 4143 + }, + { + "epoch": 0.75, + "learning_rate": 2.776447033127048e-06, + "loss": 0.4058, + "step": 4144 + }, + { + "epoch": 0.75, + "learning_rate": 2.7763924281033857e-06, + "loss": 0.7437, + "step": 4145 + }, + { + "epoch": 0.75, + "learning_rate": 2.7763378230797233e-06, + "loss": 0.8024, + "step": 4146 + }, + { + "epoch": 0.75, + "learning_rate": 2.776283218056061e-06, + "loss": 1.2769, + "step": 4147 + }, + { + "epoch": 0.75, + "learning_rate": 2.776228613032399e-06, + "loss": 1.4227, + "step": 4148 + }, + { + "epoch": 0.75, + "learning_rate": 2.7761740080087367e-06, + "loss": 1.7214, + "step": 4149 + }, + { + "epoch": 0.75, + "learning_rate": 2.7761194029850746e-06, + "loss": 1.7998, + "step": 4150 + }, + { + "epoch": 0.75, + "learning_rate": 2.7760647979614125e-06, + "loss": 2.3234, + "step": 4151 + }, + { + "epoch": 0.76, + "learning_rate": 2.77601019293775e-06, + "loss": 1.0307, + "step": 4152 + }, + { + "epoch": 0.76, + "learning_rate": 2.775955587914088e-06, + "loss": 1.1445, + "step": 4153 + }, + { + "epoch": 0.76, + "learning_rate": 2.775900982890426e-06, + "loss": 1.2631, + "step": 4154 + }, + { + "epoch": 0.76, + "learning_rate": 2.775846377866764e-06, + "loss": 2.1066, + "step": 4155 + }, + { + "epoch": 0.76, + "learning_rate": 2.775791772843102e-06, + "loss": 1.9111, + "step": 4156 + }, + { + "epoch": 0.76, + "learning_rate": 2.7757371678194394e-06, + "loss": 1.0958, + "step": 4157 + }, + { + "epoch": 0.76, + "learning_rate": 2.7756825627957773e-06, + "loss": 1.2514, + "step": 4158 + }, + { + "epoch": 0.76, + "learning_rate": 2.7756279577721152e-06, + "loss": 0.474, + "step": 4159 + }, + { + "epoch": 0.76, + "learning_rate": 2.7755733527484528e-06, + "loss": 0.8631, + "step": 4160 + }, + { + "epoch": 0.76, + "learning_rate": 2.7755187477247907e-06, + "loss": 0.8538, + "step": 4161 + }, + { + "epoch": 0.76, + "learning_rate": 2.7754641427011287e-06, + "loss": 0.346, + "step": 4162 + }, + { + "epoch": 0.76, + "learning_rate": 2.775409537677466e-06, + "loss": 0.9808, + "step": 4163 + }, + { + "epoch": 0.76, + "learning_rate": 2.775354932653804e-06, + "loss": 0.622, + "step": 4164 + }, + { + "epoch": 0.76, + "learning_rate": 2.775300327630142e-06, + "loss": 1.1946, + "step": 4165 + }, + { + "epoch": 0.76, + "learning_rate": 2.77524572260648e-06, + "loss": 0.0294, + "step": 4166 + }, + { + "epoch": 0.76, + "learning_rate": 2.775191117582818e-06, + "loss": 1.2101, + "step": 4167 + }, + { + "epoch": 0.76, + "learning_rate": 2.7751365125591555e-06, + "loss": 0.1353, + "step": 4168 + }, + { + "epoch": 0.76, + "learning_rate": 2.7750819075354934e-06, + "loss": 0.4256, + "step": 4169 + }, + { + "epoch": 0.76, + "learning_rate": 2.7750273025118314e-06, + "loss": 1.5671, + "step": 4170 + }, + { + "epoch": 0.76, + "learning_rate": 2.774972697488169e-06, + "loss": 0.1541, + "step": 4171 + }, + { + "epoch": 0.76, + "learning_rate": 2.774918092464507e-06, + "loss": 0.3761, + "step": 4172 + }, + { + "epoch": 0.76, + "learning_rate": 2.7748634874408448e-06, + "loss": 0.3451, + "step": 4173 + }, + { + "epoch": 0.76, + "learning_rate": 2.7748088824171823e-06, + "loss": 0.7026, + "step": 4174 + }, + { + "epoch": 0.76, + "learning_rate": 2.7747542773935202e-06, + "loss": 4.3575, + "step": 4175 + }, + { + "epoch": 0.76, + "learning_rate": 2.774699672369858e-06, + "loss": 0.6366, + "step": 4176 + }, + { + "epoch": 0.76, + "learning_rate": 2.7746450673461957e-06, + "loss": 0.1082, + "step": 4177 + }, + { + "epoch": 0.76, + "learning_rate": 2.7745904623225336e-06, + "loss": 0.0711, + "step": 4178 + }, + { + "epoch": 0.76, + "learning_rate": 2.7745358572988716e-06, + "loss": 1.7763, + "step": 4179 + }, + { + "epoch": 0.76, + "learning_rate": 2.7744812522752095e-06, + "loss": 0.629, + "step": 4180 + }, + { + "epoch": 0.76, + "learning_rate": 2.7744266472515475e-06, + "loss": 1.9598, + "step": 4181 + }, + { + "epoch": 0.76, + "learning_rate": 2.774372042227885e-06, + "loss": 1.2211, + "step": 4182 + }, + { + "epoch": 0.76, + "learning_rate": 2.774317437204223e-06, + "loss": 1.7928, + "step": 4183 + }, + { + "epoch": 0.76, + "learning_rate": 2.774262832180561e-06, + "loss": 3.1423, + "step": 4184 + }, + { + "epoch": 0.76, + "learning_rate": 2.7742082271568984e-06, + "loss": 0.2109, + "step": 4185 + }, + { + "epoch": 0.76, + "learning_rate": 2.7741536221332363e-06, + "loss": 1.0801, + "step": 4186 + }, + { + "epoch": 0.76, + "learning_rate": 2.7740990171095743e-06, + "loss": 3.3619, + "step": 4187 + }, + { + "epoch": 0.76, + "learning_rate": 2.774044412085912e-06, + "loss": 0.8318, + "step": 4188 + }, + { + "epoch": 0.76, + "learning_rate": 2.7739898070622497e-06, + "loss": 0.0802, + "step": 4189 + }, + { + "epoch": 0.76, + "learning_rate": 2.7739352020385877e-06, + "loss": 1.7328, + "step": 4190 + }, + { + "epoch": 0.76, + "learning_rate": 2.773880597014925e-06, + "loss": 3.2874, + "step": 4191 + }, + { + "epoch": 0.76, + "learning_rate": 2.773825991991263e-06, + "loss": 0.6101, + "step": 4192 + }, + { + "epoch": 0.76, + "learning_rate": 2.773771386967601e-06, + "loss": 1.2551, + "step": 4193 + }, + { + "epoch": 0.76, + "learning_rate": 2.773716781943939e-06, + "loss": 1.162, + "step": 4194 + }, + { + "epoch": 0.76, + "learning_rate": 2.773662176920277e-06, + "loss": 2.3081, + "step": 4195 + }, + { + "epoch": 0.76, + "learning_rate": 2.7736075718966145e-06, + "loss": 0.3956, + "step": 4196 + }, + { + "epoch": 0.76, + "learning_rate": 2.7735529668729524e-06, + "loss": 0.4517, + "step": 4197 + }, + { + "epoch": 0.76, + "learning_rate": 2.7734983618492904e-06, + "loss": 0.3388, + "step": 4198 + }, + { + "epoch": 0.76, + "learning_rate": 2.773443756825628e-06, + "loss": 1.4126, + "step": 4199 + }, + { + "epoch": 0.76, + "learning_rate": 2.773389151801966e-06, + "loss": 0.3812, + "step": 4200 + }, + { + "epoch": 0.76, + "learning_rate": 2.7733345467783038e-06, + "loss": 1.0628, + "step": 4201 + }, + { + "epoch": 0.76, + "learning_rate": 2.7732799417546413e-06, + "loss": 0.0977, + "step": 4202 + }, + { + "epoch": 0.76, + "learning_rate": 2.7732253367309792e-06, + "loss": 0.7127, + "step": 4203 + }, + { + "epoch": 0.76, + "learning_rate": 2.773170731707317e-06, + "loss": 1.8997, + "step": 4204 + }, + { + "epoch": 0.76, + "learning_rate": 2.7731161266836547e-06, + "loss": 1.9282, + "step": 4205 + }, + { + "epoch": 0.76, + "learning_rate": 2.7730615216599927e-06, + "loss": 0.7551, + "step": 4206 + }, + { + "epoch": 0.77, + "learning_rate": 2.7730069166363306e-06, + "loss": 0.7653, + "step": 4207 + }, + { + "epoch": 0.77, + "learning_rate": 2.7729523116126685e-06, + "loss": 0.6344, + "step": 4208 + }, + { + "epoch": 0.77, + "learning_rate": 2.7728977065890065e-06, + "loss": 1.8703, + "step": 4209 + }, + { + "epoch": 0.77, + "learning_rate": 2.772843101565344e-06, + "loss": 0.6401, + "step": 4210 + }, + { + "epoch": 0.77, + "learning_rate": 2.772788496541682e-06, + "loss": 2.1083, + "step": 4211 + }, + { + "epoch": 0.77, + "learning_rate": 2.77273389151802e-06, + "loss": 0.8341, + "step": 4212 + }, + { + "epoch": 0.77, + "learning_rate": 2.7726792864943574e-06, + "loss": 1.1106, + "step": 4213 + }, + { + "epoch": 0.77, + "learning_rate": 2.7726246814706954e-06, + "loss": 2.0261, + "step": 4214 + }, + { + "epoch": 0.77, + "learning_rate": 2.7725700764470333e-06, + "loss": 3.3398, + "step": 4215 + }, + { + "epoch": 0.77, + "learning_rate": 2.772515471423371e-06, + "loss": 2.3325, + "step": 4216 + }, + { + "epoch": 0.77, + "learning_rate": 2.7724608663997088e-06, + "loss": 1.8932, + "step": 4217 + }, + { + "epoch": 0.77, + "learning_rate": 2.7724062613760467e-06, + "loss": 0.6751, + "step": 4218 + }, + { + "epoch": 0.77, + "learning_rate": 2.7723516563523842e-06, + "loss": 1.0659, + "step": 4219 + }, + { + "epoch": 0.77, + "learning_rate": 2.772297051328722e-06, + "loss": 0.0971, + "step": 4220 + }, + { + "epoch": 0.77, + "learning_rate": 2.77224244630506e-06, + "loss": 2.9903, + "step": 4221 + }, + { + "epoch": 0.77, + "learning_rate": 2.772187841281398e-06, + "loss": 0.5228, + "step": 4222 + }, + { + "epoch": 0.77, + "learning_rate": 2.772133236257736e-06, + "loss": 1.1361, + "step": 4223 + }, + { + "epoch": 0.77, + "learning_rate": 2.7720786312340735e-06, + "loss": 0.6148, + "step": 4224 + }, + { + "epoch": 0.77, + "learning_rate": 2.7720240262104115e-06, + "loss": 0.6295, + "step": 4225 + }, + { + "epoch": 0.77, + "learning_rate": 2.7719694211867494e-06, + "loss": 0.5321, + "step": 4226 + }, + { + "epoch": 0.77, + "learning_rate": 2.771914816163087e-06, + "loss": 1.031, + "step": 4227 + }, + { + "epoch": 0.77, + "learning_rate": 2.771860211139425e-06, + "loss": 1.4101, + "step": 4228 + }, + { + "epoch": 0.77, + "learning_rate": 2.771805606115763e-06, + "loss": 0.6474, + "step": 4229 + }, + { + "epoch": 0.77, + "learning_rate": 2.7717510010921003e-06, + "loss": 1.3489, + "step": 4230 + }, + { + "epoch": 0.77, + "learning_rate": 2.7716963960684383e-06, + "loss": 1.1796, + "step": 4231 + }, + { + "epoch": 0.77, + "learning_rate": 2.7716417910447762e-06, + "loss": 0.3176, + "step": 4232 + }, + { + "epoch": 0.77, + "learning_rate": 2.7715871860211137e-06, + "loss": 1.1557, + "step": 4233 + }, + { + "epoch": 0.77, + "learning_rate": 2.7715325809974517e-06, + "loss": 0.725, + "step": 4234 + }, + { + "epoch": 0.77, + "learning_rate": 2.7714779759737896e-06, + "loss": 1.1397, + "step": 4235 + }, + { + "epoch": 0.77, + "learning_rate": 2.7714233709501276e-06, + "loss": 1.3048, + "step": 4236 + }, + { + "epoch": 0.77, + "learning_rate": 2.7713687659264655e-06, + "loss": 1.5273, + "step": 4237 + }, + { + "epoch": 0.77, + "learning_rate": 2.771314160902803e-06, + "loss": 1.4862, + "step": 4238 + }, + { + "epoch": 0.77, + "learning_rate": 2.771259555879141e-06, + "loss": 0.9061, + "step": 4239 + }, + { + "epoch": 0.77, + "learning_rate": 2.771204950855479e-06, + "loss": 0.5262, + "step": 4240 + }, + { + "epoch": 0.77, + "learning_rate": 2.7711503458318164e-06, + "loss": 0.6458, + "step": 4241 + }, + { + "epoch": 0.77, + "learning_rate": 2.7710957408081544e-06, + "loss": 1.3663, + "step": 4242 + }, + { + "epoch": 0.77, + "learning_rate": 2.7710411357844923e-06, + "loss": 0.8464, + "step": 4243 + }, + { + "epoch": 0.77, + "learning_rate": 2.77098653076083e-06, + "loss": 0.9067, + "step": 4244 + }, + { + "epoch": 0.77, + "learning_rate": 2.770931925737168e-06, + "loss": 0.0882, + "step": 4245 + }, + { + "epoch": 0.77, + "learning_rate": 2.7708773207135057e-06, + "loss": 1.6116, + "step": 4246 + }, + { + "epoch": 0.77, + "learning_rate": 2.7708227156898437e-06, + "loss": 1.2155, + "step": 4247 + }, + { + "epoch": 0.77, + "learning_rate": 2.7707681106661816e-06, + "loss": 0.8669, + "step": 4248 + }, + { + "epoch": 0.77, + "learning_rate": 2.770713505642519e-06, + "loss": 1.111, + "step": 4249 + }, + { + "epoch": 0.77, + "learning_rate": 2.770658900618857e-06, + "loss": 1.5988, + "step": 4250 + }, + { + "epoch": 0.77, + "learning_rate": 2.770604295595195e-06, + "loss": 0.7914, + "step": 4251 + }, + { + "epoch": 0.77, + "learning_rate": 2.7705496905715325e-06, + "loss": 2.2207, + "step": 4252 + }, + { + "epoch": 0.77, + "learning_rate": 2.7704950855478705e-06, + "loss": 1.1996, + "step": 4253 + }, + { + "epoch": 0.77, + "learning_rate": 2.7704404805242084e-06, + "loss": 1.3838, + "step": 4254 + }, + { + "epoch": 0.77, + "learning_rate": 2.770385875500546e-06, + "loss": 1.387, + "step": 4255 + }, + { + "epoch": 0.77, + "learning_rate": 2.770331270476884e-06, + "loss": 0.7483, + "step": 4256 + }, + { + "epoch": 0.77, + "learning_rate": 2.770276665453222e-06, + "loss": 0.3159, + "step": 4257 + }, + { + "epoch": 0.77, + "learning_rate": 2.7702220604295594e-06, + "loss": 2.1549, + "step": 4258 + }, + { + "epoch": 0.77, + "learning_rate": 2.7701674554058973e-06, + "loss": 0.7847, + "step": 4259 + }, + { + "epoch": 0.77, + "learning_rate": 2.7701128503822352e-06, + "loss": 0.7461, + "step": 4260 + }, + { + "epoch": 0.77, + "learning_rate": 2.770058245358573e-06, + "loss": 1.5346, + "step": 4261 + }, + { + "epoch": 0.78, + "learning_rate": 2.770003640334911e-06, + "loss": 0.0759, + "step": 4262 + }, + { + "epoch": 0.78, + "learning_rate": 2.7699490353112487e-06, + "loss": 0.1624, + "step": 4263 + }, + { + "epoch": 0.78, + "learning_rate": 2.7698944302875866e-06, + "loss": 1.5053, + "step": 4264 + }, + { + "epoch": 0.78, + "learning_rate": 2.7698398252639245e-06, + "loss": 0.6118, + "step": 4265 + }, + { + "epoch": 0.78, + "learning_rate": 2.769785220240262e-06, + "loss": 2.03, + "step": 4266 + }, + { + "epoch": 0.78, + "learning_rate": 2.7697306152166e-06, + "loss": 0.841, + "step": 4267 + }, + { + "epoch": 0.78, + "learning_rate": 2.769676010192938e-06, + "loss": 0.9655, + "step": 4268 + }, + { + "epoch": 0.78, + "learning_rate": 2.7696214051692755e-06, + "loss": 0.7134, + "step": 4269 + }, + { + "epoch": 0.78, + "learning_rate": 2.7695668001456134e-06, + "loss": 2.2989, + "step": 4270 + }, + { + "epoch": 0.78, + "learning_rate": 2.7695121951219514e-06, + "loss": 0.0853, + "step": 4271 + }, + { + "epoch": 0.78, + "learning_rate": 2.769457590098289e-06, + "loss": 0.0842, + "step": 4272 + }, + { + "epoch": 0.78, + "learning_rate": 2.769402985074627e-06, + "loss": 2.063, + "step": 4273 + }, + { + "epoch": 0.78, + "learning_rate": 2.7693483800509648e-06, + "loss": 2.2539, + "step": 4274 + }, + { + "epoch": 0.78, + "learning_rate": 2.7692937750273027e-06, + "loss": 0.8842, + "step": 4275 + }, + { + "epoch": 0.78, + "learning_rate": 2.7692391700036406e-06, + "loss": 0.3643, + "step": 4276 + }, + { + "epoch": 0.78, + "learning_rate": 2.769184564979978e-06, + "loss": 1.444, + "step": 4277 + }, + { + "epoch": 0.78, + "learning_rate": 2.769129959956316e-06, + "loss": 0.5725, + "step": 4278 + }, + { + "epoch": 0.78, + "learning_rate": 2.769075354932654e-06, + "loss": 0.3666, + "step": 4279 + }, + { + "epoch": 0.78, + "learning_rate": 2.7690207499089916e-06, + "loss": 1.5342, + "step": 4280 + }, + { + "epoch": 0.78, + "learning_rate": 2.7689661448853295e-06, + "loss": 0.5736, + "step": 4281 + }, + { + "epoch": 0.78, + "learning_rate": 2.7689115398616675e-06, + "loss": 0.2877, + "step": 4282 + }, + { + "epoch": 0.78, + "learning_rate": 2.768856934838005e-06, + "loss": 3.3402, + "step": 4283 + }, + { + "epoch": 0.78, + "learning_rate": 2.768802329814343e-06, + "loss": 0.0838, + "step": 4284 + }, + { + "epoch": 0.78, + "learning_rate": 2.768747724790681e-06, + "loss": 0.1666, + "step": 4285 + }, + { + "epoch": 0.78, + "learning_rate": 2.7686931197670184e-06, + "loss": 2.0361, + "step": 4286 + }, + { + "epoch": 0.78, + "learning_rate": 2.7686385147433563e-06, + "loss": 0.3308, + "step": 4287 + }, + { + "epoch": 0.78, + "learning_rate": 2.7685839097196943e-06, + "loss": 0.5377, + "step": 4288 + }, + { + "epoch": 0.78, + "learning_rate": 2.7685293046960322e-06, + "loss": 1.3377, + "step": 4289 + }, + { + "epoch": 0.78, + "learning_rate": 2.76847469967237e-06, + "loss": 2.2041, + "step": 4290 + }, + { + "epoch": 0.78, + "learning_rate": 2.7684200946487077e-06, + "loss": 1.0072, + "step": 4291 + }, + { + "epoch": 0.78, + "learning_rate": 2.7683654896250456e-06, + "loss": 0.9164, + "step": 4292 + }, + { + "epoch": 0.78, + "learning_rate": 2.7683108846013836e-06, + "loss": 1.6203, + "step": 4293 + }, + { + "epoch": 0.78, + "learning_rate": 2.768256279577721e-06, + "loss": 1.2981, + "step": 4294 + }, + { + "epoch": 0.78, + "learning_rate": 2.768201674554059e-06, + "loss": 0.3214, + "step": 4295 + }, + { + "epoch": 0.78, + "learning_rate": 2.768147069530397e-06, + "loss": 0.4464, + "step": 4296 + }, + { + "epoch": 0.78, + "learning_rate": 2.7680924645067345e-06, + "loss": 0.6482, + "step": 4297 + }, + { + "epoch": 0.78, + "learning_rate": 2.7680378594830724e-06, + "loss": 1.0396, + "step": 4298 + }, + { + "epoch": 0.78, + "learning_rate": 2.7679832544594104e-06, + "loss": 0.053, + "step": 4299 + }, + { + "epoch": 0.78, + "learning_rate": 2.767928649435748e-06, + "loss": 0.4283, + "step": 4300 + }, + { + "epoch": 0.78, + "learning_rate": 2.767874044412086e-06, + "loss": 0.5184, + "step": 4301 + }, + { + "epoch": 0.78, + "learning_rate": 2.7678194393884238e-06, + "loss": 1.5593, + "step": 4302 + }, + { + "epoch": 0.78, + "learning_rate": 2.7677648343647617e-06, + "loss": 1.6993, + "step": 4303 + }, + { + "epoch": 0.78, + "learning_rate": 2.7677102293410997e-06, + "loss": 0.5289, + "step": 4304 + }, + { + "epoch": 0.78, + "learning_rate": 2.767655624317437e-06, + "loss": 0.9507, + "step": 4305 + }, + { + "epoch": 0.78, + "learning_rate": 2.767601019293775e-06, + "loss": 0.3969, + "step": 4306 + }, + { + "epoch": 0.78, + "learning_rate": 2.767546414270113e-06, + "loss": 0.39, + "step": 4307 + }, + { + "epoch": 0.78, + "learning_rate": 2.7674918092464506e-06, + "loss": 0.6676, + "step": 4308 + }, + { + "epoch": 0.78, + "learning_rate": 2.7674372042227885e-06, + "loss": 0.6976, + "step": 4309 + }, + { + "epoch": 0.78, + "learning_rate": 2.7673825991991265e-06, + "loss": 2.3708, + "step": 4310 + }, + { + "epoch": 0.78, + "learning_rate": 2.767327994175464e-06, + "loss": 0.0447, + "step": 4311 + }, + { + "epoch": 0.78, + "learning_rate": 2.767273389151802e-06, + "loss": 0.2132, + "step": 4312 + }, + { + "epoch": 0.78, + "learning_rate": 2.76721878412814e-06, + "loss": 0.1649, + "step": 4313 + }, + { + "epoch": 0.78, + "learning_rate": 2.7671641791044774e-06, + "loss": 1.1409, + "step": 4314 + }, + { + "epoch": 0.78, + "learning_rate": 2.7671095740808154e-06, + "loss": 0.8231, + "step": 4315 + }, + { + "epoch": 0.78, + "learning_rate": 2.7670549690571533e-06, + "loss": 0.6175, + "step": 4316 + }, + { + "epoch": 0.79, + "learning_rate": 2.7670003640334912e-06, + "loss": 0.3613, + "step": 4317 + }, + { + "epoch": 0.79, + "learning_rate": 2.766945759009829e-06, + "loss": 0.384, + "step": 4318 + }, + { + "epoch": 0.79, + "learning_rate": 2.7668911539861667e-06, + "loss": 2.1343, + "step": 4319 + }, + { + "epoch": 0.79, + "learning_rate": 2.7668365489625047e-06, + "loss": 1.6808, + "step": 4320 + }, + { + "epoch": 0.79, + "learning_rate": 2.7667819439388426e-06, + "loss": 0.5032, + "step": 4321 + }, + { + "epoch": 0.79, + "learning_rate": 2.76672733891518e-06, + "loss": 3.4253, + "step": 4322 + }, + { + "epoch": 0.79, + "learning_rate": 2.766672733891518e-06, + "loss": 0.4564, + "step": 4323 + }, + { + "epoch": 0.79, + "learning_rate": 2.766618128867856e-06, + "loss": 0.4166, + "step": 4324 + }, + { + "epoch": 0.79, + "learning_rate": 2.7665635238441935e-06, + "loss": 0.5133, + "step": 4325 + }, + { + "epoch": 0.79, + "learning_rate": 2.7665089188205315e-06, + "loss": 1.3354, + "step": 4326 + }, + { + "epoch": 0.79, + "learning_rate": 2.7664543137968694e-06, + "loss": 0.5522, + "step": 4327 + }, + { + "epoch": 0.79, + "learning_rate": 2.766399708773207e-06, + "loss": 0.3654, + "step": 4328 + }, + { + "epoch": 0.79, + "learning_rate": 2.766345103749545e-06, + "loss": 1.0112, + "step": 4329 + }, + { + "epoch": 0.79, + "learning_rate": 2.766290498725883e-06, + "loss": 0.4684, + "step": 4330 + }, + { + "epoch": 0.79, + "learning_rate": 2.7662358937022208e-06, + "loss": 1.2014, + "step": 4331 + }, + { + "epoch": 0.79, + "learning_rate": 2.7661812886785587e-06, + "loss": 2.5241, + "step": 4332 + }, + { + "epoch": 0.79, + "learning_rate": 2.7661266836548962e-06, + "loss": 2.3872, + "step": 4333 + }, + { + "epoch": 0.79, + "learning_rate": 2.766072078631234e-06, + "loss": 0.7331, + "step": 4334 + }, + { + "epoch": 0.79, + "learning_rate": 2.766017473607572e-06, + "loss": 0.3179, + "step": 4335 + }, + { + "epoch": 0.79, + "learning_rate": 2.7659628685839096e-06, + "loss": 2.0575, + "step": 4336 + }, + { + "epoch": 0.79, + "learning_rate": 2.7659082635602476e-06, + "loss": 1.4914, + "step": 4337 + }, + { + "epoch": 0.79, + "learning_rate": 2.7658536585365855e-06, + "loss": 1.1518, + "step": 4338 + }, + { + "epoch": 0.79, + "learning_rate": 2.765799053512923e-06, + "loss": 0.9389, + "step": 4339 + }, + { + "epoch": 0.79, + "learning_rate": 2.765744448489261e-06, + "loss": 1.4528, + "step": 4340 + }, + { + "epoch": 0.79, + "learning_rate": 2.765689843465599e-06, + "loss": 2.7315, + "step": 4341 + }, + { + "epoch": 0.79, + "learning_rate": 2.765635238441937e-06, + "loss": 1.3807, + "step": 4342 + }, + { + "epoch": 0.79, + "learning_rate": 2.765580633418275e-06, + "loss": 0.4033, + "step": 4343 + }, + { + "epoch": 0.79, + "learning_rate": 2.7655260283946123e-06, + "loss": 0.0512, + "step": 4344 + }, + { + "epoch": 0.79, + "learning_rate": 2.7654714233709503e-06, + "loss": 1.0207, + "step": 4345 + }, + { + "epoch": 0.79, + "learning_rate": 2.7654168183472882e-06, + "loss": 1.8048, + "step": 4346 + }, + { + "epoch": 0.79, + "learning_rate": 2.7653622133236257e-06, + "loss": 0.6069, + "step": 4347 + }, + { + "epoch": 0.79, + "learning_rate": 2.7653076082999637e-06, + "loss": 0.6197, + "step": 4348 + }, + { + "epoch": 0.79, + "learning_rate": 2.7652530032763016e-06, + "loss": 0.7982, + "step": 4349 + }, + { + "epoch": 0.79, + "learning_rate": 2.765198398252639e-06, + "loss": 1.6075, + "step": 4350 + }, + { + "epoch": 0.79, + "learning_rate": 2.765143793228977e-06, + "loss": 0.0525, + "step": 4351 + }, + { + "epoch": 0.79, + "learning_rate": 2.765089188205315e-06, + "loss": 0.8182, + "step": 4352 + }, + { + "epoch": 0.79, + "learning_rate": 2.7650345831816525e-06, + "loss": 0.1477, + "step": 4353 + }, + { + "epoch": 0.79, + "learning_rate": 2.7649799781579905e-06, + "loss": 1.395, + "step": 4354 + }, + { + "epoch": 0.79, + "learning_rate": 2.7649253731343284e-06, + "loss": 1.2416, + "step": 4355 + }, + { + "epoch": 0.79, + "learning_rate": 2.7648707681106664e-06, + "loss": 0.7747, + "step": 4356 + }, + { + "epoch": 0.79, + "learning_rate": 2.7648161630870043e-06, + "loss": 1.3134, + "step": 4357 + }, + { + "epoch": 0.79, + "learning_rate": 2.764761558063342e-06, + "loss": 0.6119, + "step": 4358 + }, + { + "epoch": 0.79, + "learning_rate": 2.7647069530396798e-06, + "loss": 0.322, + "step": 4359 + }, + { + "epoch": 0.79, + "learning_rate": 2.7646523480160177e-06, + "loss": 1.3736, + "step": 4360 + }, + { + "epoch": 0.79, + "learning_rate": 2.7645977429923552e-06, + "loss": 0.0805, + "step": 4361 + }, + { + "epoch": 0.79, + "learning_rate": 2.764543137968693e-06, + "loss": 0.2088, + "step": 4362 + }, + { + "epoch": 0.79, + "learning_rate": 2.764488532945031e-06, + "loss": 0.1702, + "step": 4363 + }, + { + "epoch": 0.79, + "learning_rate": 2.7644339279213687e-06, + "loss": 1.0512, + "step": 4364 + }, + { + "epoch": 0.79, + "learning_rate": 2.7643793228977066e-06, + "loss": 0.6601, + "step": 4365 + }, + { + "epoch": 0.79, + "learning_rate": 2.7643247178740445e-06, + "loss": 1.588, + "step": 4366 + }, + { + "epoch": 0.79, + "learning_rate": 2.764270112850382e-06, + "loss": 1.0559, + "step": 4367 + }, + { + "epoch": 0.79, + "learning_rate": 2.76421550782672e-06, + "loss": 1.5483, + "step": 4368 + }, + { + "epoch": 0.79, + "learning_rate": 2.764160902803058e-06, + "loss": 0.9445, + "step": 4369 + }, + { + "epoch": 0.79, + "learning_rate": 2.764106297779396e-06, + "loss": 0.8799, + "step": 4370 + }, + { + "epoch": 0.79, + "learning_rate": 2.764051692755734e-06, + "loss": 0.4559, + "step": 4371 + }, + { + "epoch": 0.8, + "learning_rate": 2.7639970877320714e-06, + "loss": 0.0453, + "step": 4372 + }, + { + "epoch": 0.8, + "learning_rate": 2.7639424827084093e-06, + "loss": 2.0518, + "step": 4373 + }, + { + "epoch": 0.8, + "learning_rate": 2.7638878776847472e-06, + "loss": 0.2545, + "step": 4374 + }, + { + "epoch": 0.8, + "learning_rate": 2.7638332726610848e-06, + "loss": 0.1241, + "step": 4375 + }, + { + "epoch": 0.8, + "learning_rate": 2.7637786676374227e-06, + "loss": 0.5356, + "step": 4376 + }, + { + "epoch": 0.8, + "learning_rate": 2.7637240626137606e-06, + "loss": 3.5928, + "step": 4377 + }, + { + "epoch": 0.8, + "learning_rate": 2.763669457590098e-06, + "loss": 0.7638, + "step": 4378 + }, + { + "epoch": 0.8, + "learning_rate": 2.763614852566436e-06, + "loss": 0.5541, + "step": 4379 + }, + { + "epoch": 0.8, + "learning_rate": 2.763560247542774e-06, + "loss": 0.4481, + "step": 4380 + }, + { + "epoch": 0.8, + "learning_rate": 2.7635056425191116e-06, + "loss": 1.0062, + "step": 4381 + }, + { + "epoch": 0.8, + "learning_rate": 2.7634510374954495e-06, + "loss": 0.514, + "step": 4382 + }, + { + "epoch": 0.8, + "learning_rate": 2.7633964324717875e-06, + "loss": 1.4713, + "step": 4383 + }, + { + "epoch": 0.8, + "learning_rate": 2.7633418274481254e-06, + "loss": 1.1222, + "step": 4384 + }, + { + "epoch": 0.8, + "learning_rate": 2.7632872224244633e-06, + "loss": 1.2533, + "step": 4385 + }, + { + "epoch": 0.8, + "learning_rate": 2.763232617400801e-06, + "loss": 1.3539, + "step": 4386 + }, + { + "epoch": 0.8, + "learning_rate": 2.763178012377139e-06, + "loss": 1.1863, + "step": 4387 + }, + { + "epoch": 0.8, + "learning_rate": 2.7631234073534768e-06, + "loss": 0.3473, + "step": 4388 + }, + { + "epoch": 0.8, + "learning_rate": 2.7630688023298143e-06, + "loss": 1.2079, + "step": 4389 + }, + { + "epoch": 0.8, + "learning_rate": 2.7630141973061522e-06, + "loss": 0.5688, + "step": 4390 + }, + { + "epoch": 0.8, + "learning_rate": 2.76295959228249e-06, + "loss": 1.4715, + "step": 4391 + }, + { + "epoch": 0.8, + "learning_rate": 2.7629049872588277e-06, + "loss": 1.8196, + "step": 4392 + }, + { + "epoch": 0.8, + "learning_rate": 2.7628503822351656e-06, + "loss": 1.2441, + "step": 4393 + }, + { + "epoch": 0.8, + "learning_rate": 2.7627957772115036e-06, + "loss": 0.5553, + "step": 4394 + }, + { + "epoch": 0.8, + "learning_rate": 2.762741172187841e-06, + "loss": 1.535, + "step": 4395 + }, + { + "epoch": 0.8, + "learning_rate": 2.762686567164179e-06, + "loss": 1.1767, + "step": 4396 + }, + { + "epoch": 0.8, + "learning_rate": 2.762631962140517e-06, + "loss": 1.9014, + "step": 4397 + }, + { + "epoch": 0.8, + "learning_rate": 2.762577357116855e-06, + "loss": 2.5825, + "step": 4398 + }, + { + "epoch": 0.8, + "learning_rate": 2.762522752093193e-06, + "loss": 0.2531, + "step": 4399 + }, + { + "epoch": 0.8, + "learning_rate": 2.7624681470695304e-06, + "loss": 0.4257, + "step": 4400 + }, + { + "epoch": 0.8, + "learning_rate": 2.7624135420458683e-06, + "loss": 1.5505, + "step": 4401 + }, + { + "epoch": 0.8, + "learning_rate": 2.7623589370222063e-06, + "loss": 0.8935, + "step": 4402 + }, + { + "epoch": 0.8, + "learning_rate": 2.762304331998544e-06, + "loss": 1.296, + "step": 4403 + }, + { + "epoch": 0.8, + "learning_rate": 2.7622497269748817e-06, + "loss": 0.9828, + "step": 4404 + }, + { + "epoch": 0.8, + "learning_rate": 2.7621951219512197e-06, + "loss": 1.8342, + "step": 4405 + }, + { + "epoch": 0.8, + "learning_rate": 2.762140516927557e-06, + "loss": 1.1723, + "step": 4406 + }, + { + "epoch": 0.8, + "learning_rate": 2.762085911903895e-06, + "loss": 1.5952, + "step": 4407 + }, + { + "epoch": 0.8, + "learning_rate": 2.762031306880233e-06, + "loss": 0.032, + "step": 4408 + }, + { + "epoch": 0.8, + "learning_rate": 2.7619767018565706e-06, + "loss": 0.8222, + "step": 4409 + }, + { + "epoch": 0.8, + "learning_rate": 2.7619220968329085e-06, + "loss": 1.1832, + "step": 4410 + }, + { + "epoch": 0.8, + "learning_rate": 2.7618674918092465e-06, + "loss": 0.4359, + "step": 4411 + }, + { + "epoch": 0.8, + "learning_rate": 2.7618128867855844e-06, + "loss": 3.0793, + "step": 4412 + }, + { + "epoch": 0.8, + "learning_rate": 2.7617582817619224e-06, + "loss": 2.4455, + "step": 4413 + }, + { + "epoch": 0.8, + "learning_rate": 2.76170367673826e-06, + "loss": 0.974, + "step": 4414 + }, + { + "epoch": 0.8, + "learning_rate": 2.761649071714598e-06, + "loss": 1.3881, + "step": 4415 + }, + { + "epoch": 0.8, + "learning_rate": 2.7615944666909358e-06, + "loss": 1.8779, + "step": 4416 + }, + { + "epoch": 0.8, + "learning_rate": 2.7615398616672733e-06, + "loss": 1.0937, + "step": 4417 + }, + { + "epoch": 0.8, + "learning_rate": 2.7614852566436112e-06, + "loss": 1.8761, + "step": 4418 + }, + { + "epoch": 0.8, + "learning_rate": 2.761430651619949e-06, + "loss": 0.4455, + "step": 4419 + }, + { + "epoch": 0.8, + "learning_rate": 2.7613760465962867e-06, + "loss": 1.6194, + "step": 4420 + }, + { + "epoch": 0.8, + "learning_rate": 2.7613214415726247e-06, + "loss": 0.6054, + "step": 4421 + }, + { + "epoch": 0.8, + "learning_rate": 2.7612668365489626e-06, + "loss": 0.4862, + "step": 4422 + }, + { + "epoch": 0.8, + "learning_rate": 2.7612122315253005e-06, + "loss": 1.1336, + "step": 4423 + }, + { + "epoch": 0.8, + "learning_rate": 2.7611576265016385e-06, + "loss": 1.6099, + "step": 4424 + }, + { + "epoch": 0.8, + "learning_rate": 2.761103021477976e-06, + "loss": 1.2964, + "step": 4425 + }, + { + "epoch": 0.8, + "learning_rate": 2.761048416454314e-06, + "loss": 2.1718, + "step": 4426 + }, + { + "epoch": 0.81, + "learning_rate": 2.760993811430652e-06, + "loss": 0.0459, + "step": 4427 + }, + { + "epoch": 0.81, + "learning_rate": 2.7609392064069894e-06, + "loss": 1.7954, + "step": 4428 + }, + { + "epoch": 0.81, + "learning_rate": 2.7608846013833274e-06, + "loss": 0.5696, + "step": 4429 + }, + { + "epoch": 0.81, + "learning_rate": 2.7608299963596653e-06, + "loss": 1.0949, + "step": 4430 + }, + { + "epoch": 0.81, + "learning_rate": 2.760775391336003e-06, + "loss": 1.5568, + "step": 4431 + }, + { + "epoch": 0.81, + "learning_rate": 2.7607207863123408e-06, + "loss": 0.4913, + "step": 4432 + }, + { + "epoch": 0.81, + "learning_rate": 2.7606661812886787e-06, + "loss": 0.352, + "step": 4433 + }, + { + "epoch": 0.81, + "learning_rate": 2.7606115762650162e-06, + "loss": 0.26, + "step": 4434 + }, + { + "epoch": 0.81, + "learning_rate": 2.760556971241354e-06, + "loss": 0.4555, + "step": 4435 + }, + { + "epoch": 0.81, + "learning_rate": 2.760502366217692e-06, + "loss": 1.6752, + "step": 4436 + }, + { + "epoch": 0.81, + "learning_rate": 2.76044776119403e-06, + "loss": 0.625, + "step": 4437 + }, + { + "epoch": 0.81, + "learning_rate": 2.760393156170368e-06, + "loss": 1.6068, + "step": 4438 + }, + { + "epoch": 0.81, + "learning_rate": 2.7603385511467055e-06, + "loss": 1.1006, + "step": 4439 + }, + { + "epoch": 0.81, + "learning_rate": 2.7602839461230435e-06, + "loss": 0.2947, + "step": 4440 + }, + { + "epoch": 0.81, + "learning_rate": 2.7602293410993814e-06, + "loss": 1.6736, + "step": 4441 + }, + { + "epoch": 0.81, + "learning_rate": 2.760174736075719e-06, + "loss": 0.6254, + "step": 4442 + }, + { + "epoch": 0.81, + "learning_rate": 2.760120131052057e-06, + "loss": 1.3286, + "step": 4443 + }, + { + "epoch": 0.81, + "learning_rate": 2.760065526028395e-06, + "loss": 0.8369, + "step": 4444 + }, + { + "epoch": 0.81, + "learning_rate": 2.7600109210047323e-06, + "loss": 0.8007, + "step": 4445 + }, + { + "epoch": 0.81, + "learning_rate": 2.7599563159810703e-06, + "loss": 1.0898, + "step": 4446 + }, + { + "epoch": 0.81, + "learning_rate": 2.7599017109574082e-06, + "loss": 0.8744, + "step": 4447 + }, + { + "epoch": 0.81, + "learning_rate": 2.7598471059337457e-06, + "loss": 1.5082, + "step": 4448 + }, + { + "epoch": 0.81, + "learning_rate": 2.7597925009100837e-06, + "loss": 0.0566, + "step": 4449 + }, + { + "epoch": 0.81, + "learning_rate": 2.7597378958864216e-06, + "loss": 2.3583, + "step": 4450 + }, + { + "epoch": 0.81, + "learning_rate": 2.7596832908627596e-06, + "loss": 1.0838, + "step": 4451 + }, + { + "epoch": 0.81, + "learning_rate": 2.7596286858390975e-06, + "loss": 0.8751, + "step": 4452 + }, + { + "epoch": 0.81, + "learning_rate": 2.759574080815435e-06, + "loss": 0.9684, + "step": 4453 + }, + { + "epoch": 0.81, + "learning_rate": 2.759519475791773e-06, + "loss": 1.1399, + "step": 4454 + }, + { + "epoch": 0.81, + "learning_rate": 2.759464870768111e-06, + "loss": 0.0895, + "step": 4455 + }, + { + "epoch": 0.81, + "learning_rate": 2.7594102657444484e-06, + "loss": 2.1968, + "step": 4456 + }, + { + "epoch": 0.81, + "learning_rate": 2.7593556607207864e-06, + "loss": 3.3297, + "step": 4457 + }, + { + "epoch": 0.81, + "learning_rate": 2.7593010556971243e-06, + "loss": 0.5388, + "step": 4458 + }, + { + "epoch": 0.81, + "learning_rate": 2.759246450673462e-06, + "loss": 0.4615, + "step": 4459 + }, + { + "epoch": 0.81, + "learning_rate": 2.7591918456497998e-06, + "loss": 0.5398, + "step": 4460 + }, + { + "epoch": 0.81, + "learning_rate": 2.7591372406261377e-06, + "loss": 0.0258, + "step": 4461 + }, + { + "epoch": 0.81, + "learning_rate": 2.7590826356024752e-06, + "loss": 1.5838, + "step": 4462 + }, + { + "epoch": 0.81, + "learning_rate": 2.759028030578813e-06, + "loss": 0.3953, + "step": 4463 + }, + { + "epoch": 0.81, + "learning_rate": 2.758973425555151e-06, + "loss": 2.6413, + "step": 4464 + }, + { + "epoch": 0.81, + "learning_rate": 2.758918820531489e-06, + "loss": 1.657, + "step": 4465 + }, + { + "epoch": 0.81, + "learning_rate": 2.758864215507827e-06, + "loss": 0.7898, + "step": 4466 + }, + { + "epoch": 0.81, + "learning_rate": 2.7588096104841645e-06, + "loss": 0.6905, + "step": 4467 + }, + { + "epoch": 0.81, + "learning_rate": 2.7587550054605025e-06, + "loss": 2.3803, + "step": 4468 + }, + { + "epoch": 0.81, + "learning_rate": 2.7587004004368404e-06, + "loss": 0.466, + "step": 4469 + }, + { + "epoch": 0.81, + "learning_rate": 2.758645795413178e-06, + "loss": 0.8009, + "step": 4470 + }, + { + "epoch": 0.81, + "learning_rate": 2.758591190389516e-06, + "loss": 1.2401, + "step": 4471 + }, + { + "epoch": 0.81, + "learning_rate": 2.758536585365854e-06, + "loss": 1.6857, + "step": 4472 + }, + { + "epoch": 0.81, + "learning_rate": 2.7584819803421914e-06, + "loss": 0.5568, + "step": 4473 + }, + { + "epoch": 0.81, + "learning_rate": 2.7584273753185293e-06, + "loss": 0.9279, + "step": 4474 + }, + { + "epoch": 0.81, + "learning_rate": 2.7583727702948672e-06, + "loss": 0.8422, + "step": 4475 + }, + { + "epoch": 0.81, + "learning_rate": 2.7583181652712048e-06, + "loss": 1.6374, + "step": 4476 + }, + { + "epoch": 0.81, + "learning_rate": 2.7582635602475427e-06, + "loss": 1.6651, + "step": 4477 + }, + { + "epoch": 0.81, + "learning_rate": 2.7582089552238806e-06, + "loss": 1.837, + "step": 4478 + }, + { + "epoch": 0.81, + "learning_rate": 2.7581543502002186e-06, + "loss": 3.0161, + "step": 4479 + }, + { + "epoch": 0.81, + "learning_rate": 2.7580997451765565e-06, + "loss": 0.9443, + "step": 4480 + }, + { + "epoch": 0.81, + "learning_rate": 2.758045140152894e-06, + "loss": 2.314, + "step": 4481 + }, + { + "epoch": 0.82, + "learning_rate": 2.757990535129232e-06, + "loss": 1.162, + "step": 4482 + }, + { + "epoch": 0.82, + "learning_rate": 2.75793593010557e-06, + "loss": 0.9189, + "step": 4483 + }, + { + "epoch": 0.82, + "learning_rate": 2.7578813250819075e-06, + "loss": 0.9055, + "step": 4484 + }, + { + "epoch": 0.82, + "learning_rate": 2.7578267200582454e-06, + "loss": 0.0481, + "step": 4485 + }, + { + "epoch": 0.82, + "learning_rate": 2.7577721150345833e-06, + "loss": 1.3415, + "step": 4486 + }, + { + "epoch": 0.82, + "learning_rate": 2.757717510010921e-06, + "loss": 0.5679, + "step": 4487 + }, + { + "epoch": 0.82, + "learning_rate": 2.757662904987259e-06, + "loss": 1.2015, + "step": 4488 + }, + { + "epoch": 0.82, + "learning_rate": 2.7576082999635968e-06, + "loss": 0.4068, + "step": 4489 + }, + { + "epoch": 0.82, + "learning_rate": 2.7575536949399343e-06, + "loss": 1.3255, + "step": 4490 + }, + { + "epoch": 0.82, + "learning_rate": 2.7574990899162722e-06, + "loss": 0.4932, + "step": 4491 + }, + { + "epoch": 0.82, + "learning_rate": 2.75744448489261e-06, + "loss": 0.3777, + "step": 4492 + }, + { + "epoch": 0.82, + "learning_rate": 2.757389879868948e-06, + "loss": 1.1505, + "step": 4493 + }, + { + "epoch": 0.82, + "learning_rate": 2.757335274845286e-06, + "loss": 0.4136, + "step": 4494 + }, + { + "epoch": 0.82, + "learning_rate": 2.7572806698216236e-06, + "loss": 3.711, + "step": 4495 + }, + { + "epoch": 0.82, + "learning_rate": 2.7572260647979615e-06, + "loss": 0.8696, + "step": 4496 + }, + { + "epoch": 0.82, + "learning_rate": 2.7571714597742995e-06, + "loss": 2.1565, + "step": 4497 + }, + { + "epoch": 0.82, + "learning_rate": 2.757116854750637e-06, + "loss": 1.5746, + "step": 4498 + }, + { + "epoch": 0.82, + "learning_rate": 2.757062249726975e-06, + "loss": 1.4758, + "step": 4499 + }, + { + "epoch": 0.82, + "learning_rate": 2.757007644703313e-06, + "loss": 0.8829, + "step": 4500 + }, + { + "epoch": 0.82, + "learning_rate": 2.7569530396796504e-06, + "loss": 1.853, + "step": 4501 + }, + { + "epoch": 0.82, + "learning_rate": 2.7568984346559883e-06, + "loss": 1.6057, + "step": 4502 + }, + { + "epoch": 0.82, + "learning_rate": 2.7568438296323263e-06, + "loss": 1.116, + "step": 4503 + }, + { + "epoch": 0.82, + "learning_rate": 2.7567892246086642e-06, + "loss": 0.648, + "step": 4504 + }, + { + "epoch": 0.82, + "learning_rate": 2.756734619585002e-06, + "loss": 1.7488, + "step": 4505 + }, + { + "epoch": 0.82, + "learning_rate": 2.7566800145613397e-06, + "loss": 1.9951, + "step": 4506 + }, + { + "epoch": 0.82, + "learning_rate": 2.7566254095376776e-06, + "loss": 2.1164, + "step": 4507 + }, + { + "epoch": 0.82, + "learning_rate": 2.7565708045140156e-06, + "loss": 0.8562, + "step": 4508 + }, + { + "epoch": 0.82, + "learning_rate": 2.756516199490353e-06, + "loss": 0.7026, + "step": 4509 + }, + { + "epoch": 0.82, + "learning_rate": 2.756461594466691e-06, + "loss": 0.5685, + "step": 4510 + }, + { + "epoch": 0.82, + "learning_rate": 2.756406989443029e-06, + "loss": 0.3112, + "step": 4511 + }, + { + "epoch": 0.82, + "learning_rate": 2.7563523844193665e-06, + "loss": 1.0311, + "step": 4512 + }, + { + "epoch": 0.82, + "learning_rate": 2.7562977793957044e-06, + "loss": 0.0322, + "step": 4513 + }, + { + "epoch": 0.82, + "learning_rate": 2.7562431743720424e-06, + "loss": 2.0067, + "step": 4514 + }, + { + "epoch": 0.82, + "learning_rate": 2.75618856934838e-06, + "loss": 2.1722, + "step": 4515 + }, + { + "epoch": 0.82, + "learning_rate": 2.756133964324718e-06, + "loss": 1.0407, + "step": 4516 + }, + { + "epoch": 0.82, + "learning_rate": 2.7560793593010558e-06, + "loss": 0.0508, + "step": 4517 + }, + { + "epoch": 0.82, + "learning_rate": 2.7560247542773937e-06, + "loss": 0.9119, + "step": 4518 + }, + { + "epoch": 0.82, + "learning_rate": 2.7559701492537317e-06, + "loss": 1.076, + "step": 4519 + }, + { + "epoch": 0.82, + "learning_rate": 2.755915544230069e-06, + "loss": 0.5987, + "step": 4520 + }, + { + "epoch": 0.82, + "learning_rate": 2.755860939206407e-06, + "loss": 0.7691, + "step": 4521 + }, + { + "epoch": 0.82, + "learning_rate": 2.755806334182745e-06, + "loss": 0.9136, + "step": 4522 + }, + { + "epoch": 0.82, + "learning_rate": 2.7557517291590826e-06, + "loss": 1.327, + "step": 4523 + }, + { + "epoch": 0.82, + "learning_rate": 2.7556971241354205e-06, + "loss": 0.1086, + "step": 4524 + }, + { + "epoch": 0.82, + "learning_rate": 2.7556425191117585e-06, + "loss": 1.5178, + "step": 4525 + }, + { + "epoch": 0.82, + "learning_rate": 2.755587914088096e-06, + "loss": 0.9187, + "step": 4526 + }, + { + "epoch": 0.82, + "learning_rate": 2.755533309064434e-06, + "loss": 0.4919, + "step": 4527 + }, + { + "epoch": 0.82, + "learning_rate": 2.755478704040772e-06, + "loss": 0.6971, + "step": 4528 + }, + { + "epoch": 0.82, + "learning_rate": 2.7554240990171094e-06, + "loss": 0.6482, + "step": 4529 + }, + { + "epoch": 0.82, + "learning_rate": 2.7553694939934474e-06, + "loss": 0.44, + "step": 4530 + }, + { + "epoch": 0.82, + "learning_rate": 2.7553148889697853e-06, + "loss": 1.8795, + "step": 4531 + }, + { + "epoch": 0.82, + "learning_rate": 2.7552602839461232e-06, + "loss": 0.8563, + "step": 4532 + }, + { + "epoch": 0.82, + "learning_rate": 2.755205678922461e-06, + "loss": 0.6342, + "step": 4533 + }, + { + "epoch": 0.82, + "learning_rate": 2.7551510738987987e-06, + "loss": 0.3984, + "step": 4534 + }, + { + "epoch": 0.82, + "learning_rate": 2.7550964688751366e-06, + "loss": 0.6103, + "step": 4535 + }, + { + "epoch": 0.82, + "learning_rate": 2.7550418638514746e-06, + "loss": 0.0716, + "step": 4536 + }, + { + "epoch": 0.83, + "learning_rate": 2.754987258827812e-06, + "loss": 0.5591, + "step": 4537 + }, + { + "epoch": 0.83, + "learning_rate": 2.75493265380415e-06, + "loss": 2.3986, + "step": 4538 + }, + { + "epoch": 0.83, + "learning_rate": 2.754878048780488e-06, + "loss": 0.2068, + "step": 4539 + }, + { + "epoch": 0.83, + "learning_rate": 2.7548234437568255e-06, + "loss": 1.3431, + "step": 4540 + }, + { + "epoch": 0.83, + "learning_rate": 2.7547688387331635e-06, + "loss": 0.8326, + "step": 4541 + }, + { + "epoch": 0.83, + "learning_rate": 2.7547142337095014e-06, + "loss": 0.2014, + "step": 4542 + }, + { + "epoch": 0.83, + "learning_rate": 2.754659628685839e-06, + "loss": 0.827, + "step": 4543 + }, + { + "epoch": 0.83, + "learning_rate": 2.754605023662177e-06, + "loss": 0.6948, + "step": 4544 + }, + { + "epoch": 0.83, + "learning_rate": 2.754550418638515e-06, + "loss": 1.6257, + "step": 4545 + }, + { + "epoch": 0.83, + "learning_rate": 2.7544958136148528e-06, + "loss": 0.2866, + "step": 4546 + }, + { + "epoch": 0.83, + "learning_rate": 2.7544412085911907e-06, + "loss": 3.5513, + "step": 4547 + }, + { + "epoch": 0.83, + "learning_rate": 2.7543866035675282e-06, + "loss": 0.266, + "step": 4548 + }, + { + "epoch": 0.83, + "learning_rate": 2.754331998543866e-06, + "loss": 1.0418, + "step": 4549 + }, + { + "epoch": 0.83, + "learning_rate": 2.754277393520204e-06, + "loss": 1.4102, + "step": 4550 + }, + { + "epoch": 0.83, + "learning_rate": 2.7542227884965416e-06, + "loss": 0.437, + "step": 4551 + }, + { + "epoch": 0.83, + "learning_rate": 2.7541681834728796e-06, + "loss": 2.3985, + "step": 4552 + }, + { + "epoch": 0.83, + "learning_rate": 2.7541135784492175e-06, + "loss": 0.5474, + "step": 4553 + }, + { + "epoch": 0.83, + "learning_rate": 2.754058973425555e-06, + "loss": 1.0839, + "step": 4554 + }, + { + "epoch": 0.83, + "learning_rate": 2.754004368401893e-06, + "loss": 0.7799, + "step": 4555 + }, + { + "epoch": 0.83, + "learning_rate": 2.753949763378231e-06, + "loss": 1.8558, + "step": 4556 + }, + { + "epoch": 0.83, + "learning_rate": 2.7538951583545684e-06, + "loss": 0.8289, + "step": 4557 + }, + { + "epoch": 0.83, + "learning_rate": 2.7538405533309064e-06, + "loss": 0.4311, + "step": 4558 + }, + { + "epoch": 0.83, + "learning_rate": 2.7537859483072443e-06, + "loss": 0.7015, + "step": 4559 + }, + { + "epoch": 0.83, + "learning_rate": 2.7537313432835823e-06, + "loss": 0.1782, + "step": 4560 + }, + { + "epoch": 0.83, + "learning_rate": 2.75367673825992e-06, + "loss": 0.6106, + "step": 4561 + }, + { + "epoch": 0.83, + "learning_rate": 2.7536221332362577e-06, + "loss": 0.184, + "step": 4562 + }, + { + "epoch": 0.83, + "learning_rate": 2.7535675282125957e-06, + "loss": 0.118, + "step": 4563 + }, + { + "epoch": 0.83, + "learning_rate": 2.7535129231889336e-06, + "loss": 1.9081, + "step": 4564 + }, + { + "epoch": 0.83, + "learning_rate": 2.753458318165271e-06, + "loss": 0.7973, + "step": 4565 + }, + { + "epoch": 0.83, + "learning_rate": 2.753403713141609e-06, + "loss": 0.969, + "step": 4566 + }, + { + "epoch": 0.83, + "learning_rate": 2.753349108117947e-06, + "loss": 0.3623, + "step": 4567 + }, + { + "epoch": 0.83, + "learning_rate": 2.7532945030942845e-06, + "loss": 3.0606, + "step": 4568 + }, + { + "epoch": 0.83, + "learning_rate": 2.7532398980706225e-06, + "loss": 0.9226, + "step": 4569 + }, + { + "epoch": 0.83, + "learning_rate": 2.7531852930469604e-06, + "loss": 0.0602, + "step": 4570 + }, + { + "epoch": 0.83, + "learning_rate": 2.753130688023298e-06, + "loss": 0.2264, + "step": 4571 + }, + { + "epoch": 0.83, + "learning_rate": 2.753076082999636e-06, + "loss": 1.0194, + "step": 4572 + }, + { + "epoch": 0.83, + "learning_rate": 2.753021477975974e-06, + "loss": 2.3153, + "step": 4573 + }, + { + "epoch": 0.83, + "learning_rate": 2.7529668729523118e-06, + "loss": 1.4445, + "step": 4574 + }, + { + "epoch": 0.83, + "learning_rate": 2.7529122679286497e-06, + "loss": 0.7363, + "step": 4575 + }, + { + "epoch": 0.83, + "learning_rate": 2.7528576629049872e-06, + "loss": 1.2162, + "step": 4576 + }, + { + "epoch": 0.83, + "learning_rate": 2.752803057881325e-06, + "loss": 0.4182, + "step": 4577 + }, + { + "epoch": 0.83, + "learning_rate": 2.752748452857663e-06, + "loss": 1.3465, + "step": 4578 + }, + { + "epoch": 0.83, + "learning_rate": 2.7526938478340007e-06, + "loss": 1.7628, + "step": 4579 + }, + { + "epoch": 0.83, + "learning_rate": 2.7526392428103386e-06, + "loss": 1.2284, + "step": 4580 + }, + { + "epoch": 0.83, + "learning_rate": 2.7525846377866765e-06, + "loss": 0.9248, + "step": 4581 + }, + { + "epoch": 0.83, + "learning_rate": 2.752530032763014e-06, + "loss": 1.348, + "step": 4582 + }, + { + "epoch": 0.83, + "learning_rate": 2.752475427739352e-06, + "loss": 0.6257, + "step": 4583 + }, + { + "epoch": 0.83, + "learning_rate": 2.75242082271569e-06, + "loss": 1.4208, + "step": 4584 + }, + { + "epoch": 0.83, + "learning_rate": 2.752366217692028e-06, + "loss": 0.8475, + "step": 4585 + }, + { + "epoch": 0.83, + "learning_rate": 2.7523116126683654e-06, + "loss": 1.1641, + "step": 4586 + }, + { + "epoch": 0.83, + "learning_rate": 2.7522570076447034e-06, + "loss": 1.5767, + "step": 4587 + }, + { + "epoch": 0.83, + "learning_rate": 2.7522024026210413e-06, + "loss": 1.9004, + "step": 4588 + }, + { + "epoch": 0.83, + "learning_rate": 2.7521477975973792e-06, + "loss": 0.6817, + "step": 4589 + }, + { + "epoch": 0.83, + "learning_rate": 2.7520931925737168e-06, + "loss": 3.0572, + "step": 4590 + }, + { + "epoch": 0.83, + "learning_rate": 2.7520385875500547e-06, + "loss": 0.0725, + "step": 4591 + }, + { + "epoch": 0.84, + "learning_rate": 2.7519839825263926e-06, + "loss": 1.411, + "step": 4592 + }, + { + "epoch": 0.84, + "learning_rate": 2.75192937750273e-06, + "loss": 0.1858, + "step": 4593 + }, + { + "epoch": 0.84, + "learning_rate": 2.751874772479068e-06, + "loss": 0.776, + "step": 4594 + }, + { + "epoch": 0.84, + "learning_rate": 2.751820167455406e-06, + "loss": 0.9768, + "step": 4595 + }, + { + "epoch": 0.84, + "learning_rate": 2.7517655624317436e-06, + "loss": 0.8891, + "step": 4596 + }, + { + "epoch": 0.84, + "learning_rate": 2.7517109574080815e-06, + "loss": 0.4969, + "step": 4597 + }, + { + "epoch": 0.84, + "learning_rate": 2.7516563523844195e-06, + "loss": 0.6298, + "step": 4598 + }, + { + "epoch": 0.84, + "learning_rate": 2.7516017473607574e-06, + "loss": 1.2216, + "step": 4599 + }, + { + "epoch": 0.84, + "learning_rate": 2.7515471423370953e-06, + "loss": 0.2035, + "step": 4600 + }, + { + "epoch": 0.84, + "learning_rate": 2.751492537313433e-06, + "loss": 1.9324, + "step": 4601 + }, + { + "epoch": 0.84, + "learning_rate": 2.751437932289771e-06, + "loss": 0.1238, + "step": 4602 + }, + { + "epoch": 0.84, + "learning_rate": 2.7513833272661088e-06, + "loss": 0.4987, + "step": 4603 + }, + { + "epoch": 0.84, + "learning_rate": 2.7513287222424463e-06, + "loss": 0.3471, + "step": 4604 + }, + { + "epoch": 0.84, + "learning_rate": 2.7512741172187842e-06, + "loss": 0.2362, + "step": 4605 + }, + { + "epoch": 0.84, + "learning_rate": 2.751219512195122e-06, + "loss": 1.2375, + "step": 4606 + }, + { + "epoch": 0.84, + "learning_rate": 2.7511649071714597e-06, + "loss": 0.0689, + "step": 4607 + }, + { + "epoch": 0.84, + "learning_rate": 2.7511103021477976e-06, + "loss": 1.5581, + "step": 4608 + }, + { + "epoch": 0.84, + "learning_rate": 2.7510556971241356e-06, + "loss": 1.7504, + "step": 4609 + }, + { + "epoch": 0.84, + "learning_rate": 2.751001092100473e-06, + "loss": 1.7351, + "step": 4610 + }, + { + "epoch": 0.84, + "learning_rate": 2.750946487076811e-06, + "loss": 0.7736, + "step": 4611 + }, + { + "epoch": 0.84, + "learning_rate": 2.750891882053149e-06, + "loss": 1.3121, + "step": 4612 + }, + { + "epoch": 0.84, + "learning_rate": 2.750837277029487e-06, + "loss": 0.63, + "step": 4613 + }, + { + "epoch": 0.84, + "learning_rate": 2.750782672005825e-06, + "loss": 1.1648, + "step": 4614 + }, + { + "epoch": 0.84, + "learning_rate": 2.7507280669821624e-06, + "loss": 1.1343, + "step": 4615 + }, + { + "epoch": 0.84, + "learning_rate": 2.7506734619585003e-06, + "loss": 1.2866, + "step": 4616 + }, + { + "epoch": 0.84, + "learning_rate": 2.7506188569348383e-06, + "loss": 1.701, + "step": 4617 + }, + { + "epoch": 0.84, + "learning_rate": 2.7505642519111758e-06, + "loss": 0.8577, + "step": 4618 + }, + { + "epoch": 0.84, + "learning_rate": 2.7505096468875137e-06, + "loss": 0.8859, + "step": 4619 + }, + { + "epoch": 0.84, + "learning_rate": 2.7504550418638517e-06, + "loss": 2.2626, + "step": 4620 + }, + { + "epoch": 0.84, + "learning_rate": 2.750400436840189e-06, + "loss": 1.0505, + "step": 4621 + }, + { + "epoch": 0.84, + "learning_rate": 2.750345831816527e-06, + "loss": 0.6197, + "step": 4622 + }, + { + "epoch": 0.84, + "learning_rate": 2.750291226792865e-06, + "loss": 2.1632, + "step": 4623 + }, + { + "epoch": 0.84, + "learning_rate": 2.7502366217692026e-06, + "loss": 1.3604, + "step": 4624 + }, + { + "epoch": 0.84, + "learning_rate": 2.7501820167455405e-06, + "loss": 0.4642, + "step": 4625 + }, + { + "epoch": 0.84, + "learning_rate": 2.7501274117218785e-06, + "loss": 0.2304, + "step": 4626 + }, + { + "epoch": 0.84, + "learning_rate": 2.7500728066982164e-06, + "loss": 0.8195, + "step": 4627 + }, + { + "epoch": 0.84, + "learning_rate": 2.7500182016745544e-06, + "loss": 0.9565, + "step": 4628 + }, + { + "epoch": 0.84, + "learning_rate": 2.749963596650892e-06, + "loss": 1.5026, + "step": 4629 + }, + { + "epoch": 0.84, + "learning_rate": 2.74990899162723e-06, + "loss": 1.1899, + "step": 4630 + }, + { + "epoch": 0.84, + "learning_rate": 2.7498543866035678e-06, + "loss": 1.0984, + "step": 4631 + }, + { + "epoch": 0.84, + "learning_rate": 2.7497997815799053e-06, + "loss": 1.7667, + "step": 4632 + }, + { + "epoch": 0.84, + "learning_rate": 2.7497451765562432e-06, + "loss": 0.7893, + "step": 4633 + }, + { + "epoch": 0.84, + "learning_rate": 2.749690571532581e-06, + "loss": 1.3874, + "step": 4634 + }, + { + "epoch": 0.84, + "learning_rate": 2.7496359665089187e-06, + "loss": 1.0752, + "step": 4635 + }, + { + "epoch": 0.84, + "learning_rate": 2.7495813614852566e-06, + "loss": 2.2277, + "step": 4636 + }, + { + "epoch": 0.84, + "learning_rate": 2.7495267564615946e-06, + "loss": 0.8064, + "step": 4637 + }, + { + "epoch": 0.84, + "learning_rate": 2.749472151437932e-06, + "loss": 0.9128, + "step": 4638 + }, + { + "epoch": 0.84, + "learning_rate": 2.74941754641427e-06, + "loss": 1.1432, + "step": 4639 + }, + { + "epoch": 0.84, + "learning_rate": 2.749362941390608e-06, + "loss": 1.9354, + "step": 4640 + }, + { + "epoch": 0.84, + "learning_rate": 2.749308336366946e-06, + "loss": 0.4545, + "step": 4641 + }, + { + "epoch": 0.84, + "learning_rate": 2.749253731343284e-06, + "loss": 0.0884, + "step": 4642 + }, + { + "epoch": 0.84, + "learning_rate": 2.7491991263196214e-06, + "loss": 0.2252, + "step": 4643 + }, + { + "epoch": 0.84, + "learning_rate": 2.7491445212959593e-06, + "loss": 1.6692, + "step": 4644 + }, + { + "epoch": 0.84, + "learning_rate": 2.7490899162722973e-06, + "loss": 1.0171, + "step": 4645 + }, + { + "epoch": 0.84, + "learning_rate": 2.749035311248635e-06, + "loss": 0.0444, + "step": 4646 + }, + { + "epoch": 0.85, + "learning_rate": 2.7489807062249728e-06, + "loss": 1.5357, + "step": 4647 + }, + { + "epoch": 0.85, + "learning_rate": 2.7489261012013107e-06, + "loss": 0.7134, + "step": 4648 + }, + { + "epoch": 0.85, + "learning_rate": 2.7488714961776482e-06, + "loss": 0.6845, + "step": 4649 + }, + { + "epoch": 0.85, + "learning_rate": 2.748816891153986e-06, + "loss": 1.1261, + "step": 4650 + }, + { + "epoch": 0.85, + "learning_rate": 2.748762286130324e-06, + "loss": 0.6305, + "step": 4651 + }, + { + "epoch": 0.85, + "learning_rate": 2.7487076811066616e-06, + "loss": 0.3837, + "step": 4652 + }, + { + "epoch": 0.85, + "learning_rate": 2.7486530760829996e-06, + "loss": 0.8971, + "step": 4653 + }, + { + "epoch": 0.85, + "learning_rate": 2.7485984710593375e-06, + "loss": 1.3943, + "step": 4654 + }, + { + "epoch": 0.85, + "learning_rate": 2.7485438660356755e-06, + "loss": 1.8035, + "step": 4655 + }, + { + "epoch": 0.85, + "learning_rate": 2.7484892610120134e-06, + "loss": 1.1282, + "step": 4656 + }, + { + "epoch": 0.85, + "learning_rate": 2.748434655988351e-06, + "loss": 0.025, + "step": 4657 + }, + { + "epoch": 0.85, + "learning_rate": 2.748380050964689e-06, + "loss": 0.6862, + "step": 4658 + }, + { + "epoch": 0.85, + "learning_rate": 2.748325445941027e-06, + "loss": 0.6569, + "step": 4659 + }, + { + "epoch": 0.85, + "learning_rate": 2.7482708409173643e-06, + "loss": 1.0747, + "step": 4660 + }, + { + "epoch": 0.85, + "learning_rate": 2.7482162358937023e-06, + "loss": 2.5048, + "step": 4661 + }, + { + "epoch": 0.85, + "learning_rate": 2.74816163087004e-06, + "loss": 0.1026, + "step": 4662 + }, + { + "epoch": 0.85, + "learning_rate": 2.7481070258463777e-06, + "loss": 0.9603, + "step": 4663 + }, + { + "epoch": 0.85, + "learning_rate": 2.7480524208227157e-06, + "loss": 1.1753, + "step": 4664 + }, + { + "epoch": 0.85, + "learning_rate": 2.7479978157990536e-06, + "loss": 2.1913, + "step": 4665 + }, + { + "epoch": 0.85, + "learning_rate": 2.747943210775391e-06, + "loss": 1.2054, + "step": 4666 + }, + { + "epoch": 0.85, + "learning_rate": 2.747888605751729e-06, + "loss": 2.7015, + "step": 4667 + }, + { + "epoch": 0.85, + "learning_rate": 2.747834000728067e-06, + "loss": 1.0782, + "step": 4668 + }, + { + "epoch": 0.85, + "learning_rate": 2.747779395704405e-06, + "loss": 0.7548, + "step": 4669 + }, + { + "epoch": 0.85, + "learning_rate": 2.747724790680743e-06, + "loss": 0.7356, + "step": 4670 + }, + { + "epoch": 0.85, + "learning_rate": 2.7476701856570804e-06, + "loss": 0.2214, + "step": 4671 + }, + { + "epoch": 0.85, + "learning_rate": 2.7476155806334184e-06, + "loss": 1.5154, + "step": 4672 + }, + { + "epoch": 0.85, + "learning_rate": 2.7475609756097563e-06, + "loss": 0.777, + "step": 4673 + }, + { + "epoch": 0.85, + "learning_rate": 2.747506370586094e-06, + "loss": 0.1018, + "step": 4674 + }, + { + "epoch": 0.85, + "learning_rate": 2.7474517655624318e-06, + "loss": 1.9598, + "step": 4675 + }, + { + "epoch": 0.85, + "learning_rate": 2.7473971605387697e-06, + "loss": 0.0265, + "step": 4676 + }, + { + "epoch": 0.85, + "learning_rate": 2.7473425555151072e-06, + "loss": 1.2006, + "step": 4677 + }, + { + "epoch": 0.85, + "learning_rate": 2.747287950491445e-06, + "loss": 0.6897, + "step": 4678 + }, + { + "epoch": 0.85, + "learning_rate": 2.747233345467783e-06, + "loss": 2.1896, + "step": 4679 + }, + { + "epoch": 0.85, + "learning_rate": 2.747178740444121e-06, + "loss": 0.1856, + "step": 4680 + }, + { + "epoch": 0.85, + "learning_rate": 2.747124135420459e-06, + "loss": 0.1749, + "step": 4681 + }, + { + "epoch": 0.85, + "learning_rate": 2.7470695303967965e-06, + "loss": 1.1467, + "step": 4682 + }, + { + "epoch": 0.85, + "learning_rate": 2.7470149253731345e-06, + "loss": 0.4189, + "step": 4683 + }, + { + "epoch": 0.85, + "learning_rate": 2.7469603203494724e-06, + "loss": 0.1177, + "step": 4684 + }, + { + "epoch": 0.85, + "learning_rate": 2.74690571532581e-06, + "loss": 0.6629, + "step": 4685 + }, + { + "epoch": 0.85, + "learning_rate": 2.746851110302148e-06, + "loss": 0.6007, + "step": 4686 + }, + { + "epoch": 0.85, + "learning_rate": 2.746796505278486e-06, + "loss": 0.8662, + "step": 4687 + }, + { + "epoch": 0.85, + "learning_rate": 2.7467419002548234e-06, + "loss": 1.0427, + "step": 4688 + }, + { + "epoch": 0.85, + "learning_rate": 2.7466872952311613e-06, + "loss": 1.3888, + "step": 4689 + }, + { + "epoch": 0.85, + "learning_rate": 2.7466326902074992e-06, + "loss": 0.2634, + "step": 4690 + }, + { + "epoch": 0.85, + "learning_rate": 2.7465780851838368e-06, + "loss": 0.0906, + "step": 4691 + }, + { + "epoch": 0.85, + "learning_rate": 2.7465234801601747e-06, + "loss": 0.8709, + "step": 4692 + }, + { + "epoch": 0.85, + "learning_rate": 2.7464688751365126e-06, + "loss": 1.0328, + "step": 4693 + }, + { + "epoch": 0.85, + "learning_rate": 2.7464142701128506e-06, + "loss": 0.9779, + "step": 4694 + }, + { + "epoch": 0.85, + "learning_rate": 2.7463596650891885e-06, + "loss": 1.3004, + "step": 4695 + }, + { + "epoch": 0.85, + "learning_rate": 2.746305060065526e-06, + "loss": 0.9437, + "step": 4696 + }, + { + "epoch": 0.85, + "learning_rate": 2.746250455041864e-06, + "loss": 2.1817, + "step": 4697 + }, + { + "epoch": 0.85, + "learning_rate": 2.746195850018202e-06, + "loss": 0.9748, + "step": 4698 + }, + { + "epoch": 0.85, + "learning_rate": 2.7461412449945395e-06, + "loss": 0.7299, + "step": 4699 + }, + { + "epoch": 0.85, + "learning_rate": 2.7460866399708774e-06, + "loss": 0.0221, + "step": 4700 + }, + { + "epoch": 0.85, + "learning_rate": 2.7460320349472153e-06, + "loss": 1.5136, + "step": 4701 + }, + { + "epoch": 0.86, + "learning_rate": 2.745977429923553e-06, + "loss": 0.839, + "step": 4702 + }, + { + "epoch": 0.86, + "learning_rate": 2.745922824899891e-06, + "loss": 0.4256, + "step": 4703 + }, + { + "epoch": 0.86, + "learning_rate": 2.7458682198762288e-06, + "loss": 3.0742, + "step": 4704 + }, + { + "epoch": 0.86, + "learning_rate": 2.7458136148525663e-06, + "loss": 0.6013, + "step": 4705 + }, + { + "epoch": 0.86, + "learning_rate": 2.7457590098289042e-06, + "loss": 0.6596, + "step": 4706 + }, + { + "epoch": 0.86, + "learning_rate": 2.745704404805242e-06, + "loss": 1.5472, + "step": 4707 + }, + { + "epoch": 0.86, + "learning_rate": 2.74564979978158e-06, + "loss": 1.196, + "step": 4708 + }, + { + "epoch": 0.86, + "learning_rate": 2.745595194757918e-06, + "loss": 0.2628, + "step": 4709 + }, + { + "epoch": 0.86, + "learning_rate": 2.7455405897342556e-06, + "loss": 1.1895, + "step": 4710 + }, + { + "epoch": 0.86, + "learning_rate": 2.7454859847105935e-06, + "loss": 1.9993, + "step": 4711 + }, + { + "epoch": 0.86, + "learning_rate": 2.7454313796869315e-06, + "loss": 0.6738, + "step": 4712 + }, + { + "epoch": 0.86, + "learning_rate": 2.745376774663269e-06, + "loss": 1.5537, + "step": 4713 + }, + { + "epoch": 0.86, + "learning_rate": 2.745322169639607e-06, + "loss": 1.4312, + "step": 4714 + }, + { + "epoch": 0.86, + "learning_rate": 2.745267564615945e-06, + "loss": 0.3632, + "step": 4715 + }, + { + "epoch": 0.86, + "learning_rate": 2.7452129595922824e-06, + "loss": 1.2903, + "step": 4716 + }, + { + "epoch": 0.86, + "learning_rate": 2.7451583545686203e-06, + "loss": 0.3735, + "step": 4717 + }, + { + "epoch": 0.86, + "learning_rate": 2.7451037495449583e-06, + "loss": 1.5928, + "step": 4718 + }, + { + "epoch": 0.86, + "learning_rate": 2.7450491445212958e-06, + "loss": 1.1677, + "step": 4719 + }, + { + "epoch": 0.86, + "learning_rate": 2.7449945394976337e-06, + "loss": 1.8562, + "step": 4720 + }, + { + "epoch": 0.86, + "learning_rate": 2.7449399344739717e-06, + "loss": 2.2502, + "step": 4721 + }, + { + "epoch": 0.86, + "learning_rate": 2.7448853294503096e-06, + "loss": 0.7178, + "step": 4722 + }, + { + "epoch": 0.86, + "learning_rate": 2.7448307244266476e-06, + "loss": 2.7701, + "step": 4723 + }, + { + "epoch": 0.86, + "learning_rate": 2.744776119402985e-06, + "loss": 0.7129, + "step": 4724 + }, + { + "epoch": 0.86, + "learning_rate": 2.744721514379323e-06, + "loss": 1.77, + "step": 4725 + }, + { + "epoch": 0.86, + "learning_rate": 2.744666909355661e-06, + "loss": 2.0127, + "step": 4726 + }, + { + "epoch": 0.86, + "learning_rate": 2.7446123043319985e-06, + "loss": 1.7183, + "step": 4727 + }, + { + "epoch": 0.86, + "learning_rate": 2.7445576993083364e-06, + "loss": 1.9989, + "step": 4728 + }, + { + "epoch": 0.86, + "learning_rate": 2.7445030942846744e-06, + "loss": 0.8955, + "step": 4729 + }, + { + "epoch": 0.86, + "learning_rate": 2.744448489261012e-06, + "loss": 1.109, + "step": 4730 + }, + { + "epoch": 0.86, + "learning_rate": 2.74439388423735e-06, + "loss": 0.6509, + "step": 4731 + }, + { + "epoch": 0.86, + "learning_rate": 2.7443392792136878e-06, + "loss": 0.4772, + "step": 4732 + }, + { + "epoch": 0.86, + "learning_rate": 2.7442846741900253e-06, + "loss": 0.3834, + "step": 4733 + }, + { + "epoch": 0.86, + "learning_rate": 2.7442300691663632e-06, + "loss": 1.1679, + "step": 4734 + }, + { + "epoch": 0.86, + "learning_rate": 2.744175464142701e-06, + "loss": 1.0916, + "step": 4735 + }, + { + "epoch": 0.86, + "learning_rate": 2.744120859119039e-06, + "loss": 0.353, + "step": 4736 + }, + { + "epoch": 0.86, + "learning_rate": 2.744066254095377e-06, + "loss": 1.8537, + "step": 4737 + }, + { + "epoch": 0.86, + "learning_rate": 2.7440116490717146e-06, + "loss": 0.0898, + "step": 4738 + }, + { + "epoch": 0.86, + "learning_rate": 2.7439570440480525e-06, + "loss": 2.535, + "step": 4739 + }, + { + "epoch": 0.86, + "learning_rate": 2.7439024390243905e-06, + "loss": 1.5749, + "step": 4740 + }, + { + "epoch": 0.86, + "learning_rate": 2.743847834000728e-06, + "loss": 1.0012, + "step": 4741 + }, + { + "epoch": 0.86, + "learning_rate": 2.743793228977066e-06, + "loss": 0.0468, + "step": 4742 + }, + { + "epoch": 0.86, + "learning_rate": 2.743738623953404e-06, + "loss": 1.5618, + "step": 4743 + }, + { + "epoch": 0.86, + "learning_rate": 2.7436840189297414e-06, + "loss": 0.4803, + "step": 4744 + }, + { + "epoch": 0.86, + "learning_rate": 2.7436294139060793e-06, + "loss": 0.7202, + "step": 4745 + }, + { + "epoch": 0.86, + "learning_rate": 2.7435748088824173e-06, + "loss": 0.5242, + "step": 4746 + }, + { + "epoch": 0.86, + "learning_rate": 2.743520203858755e-06, + "loss": 1.387, + "step": 4747 + }, + { + "epoch": 0.86, + "learning_rate": 2.7434655988350928e-06, + "loss": 0.5619, + "step": 4748 + }, + { + "epoch": 0.86, + "learning_rate": 2.7434109938114307e-06, + "loss": 1.1171, + "step": 4749 + }, + { + "epoch": 0.86, + "learning_rate": 2.7433563887877686e-06, + "loss": 0.8092, + "step": 4750 + }, + { + "epoch": 0.86, + "learning_rate": 2.7433017837641066e-06, + "loss": 0.9181, + "step": 4751 + }, + { + "epoch": 0.86, + "learning_rate": 2.743247178740444e-06, + "loss": 0.5472, + "step": 4752 + }, + { + "epoch": 0.86, + "learning_rate": 2.743192573716782e-06, + "loss": 3.7638, + "step": 4753 + }, + { + "epoch": 0.86, + "learning_rate": 2.74313796869312e-06, + "loss": 1.8476, + "step": 4754 + }, + { + "epoch": 0.86, + "learning_rate": 2.7430833636694575e-06, + "loss": 0.1631, + "step": 4755 + }, + { + "epoch": 0.86, + "learning_rate": 2.7430287586457955e-06, + "loss": 0.3403, + "step": 4756 + }, + { + "epoch": 0.87, + "learning_rate": 2.7429741536221334e-06, + "loss": 2.2169, + "step": 4757 + }, + { + "epoch": 0.87, + "learning_rate": 2.742919548598471e-06, + "loss": 0.5797, + "step": 4758 + }, + { + "epoch": 0.87, + "learning_rate": 2.742864943574809e-06, + "loss": 0.5822, + "step": 4759 + }, + { + "epoch": 0.87, + "learning_rate": 2.742810338551147e-06, + "loss": 0.8061, + "step": 4760 + }, + { + "epoch": 0.87, + "learning_rate": 2.7427557335274847e-06, + "loss": 1.1335, + "step": 4761 + }, + { + "epoch": 0.87, + "learning_rate": 2.7427011285038227e-06, + "loss": 0.6607, + "step": 4762 + }, + { + "epoch": 0.87, + "learning_rate": 2.7426465234801602e-06, + "loss": 0.0632, + "step": 4763 + }, + { + "epoch": 0.87, + "learning_rate": 2.742591918456498e-06, + "loss": 1.8069, + "step": 4764 + }, + { + "epoch": 0.87, + "learning_rate": 2.742537313432836e-06, + "loss": 1.1272, + "step": 4765 + }, + { + "epoch": 0.87, + "learning_rate": 2.7424827084091736e-06, + "loss": 0.062, + "step": 4766 + }, + { + "epoch": 0.87, + "learning_rate": 2.7424281033855116e-06, + "loss": 0.261, + "step": 4767 + }, + { + "epoch": 0.87, + "learning_rate": 2.7423734983618495e-06, + "loss": 0.8383, + "step": 4768 + }, + { + "epoch": 0.87, + "learning_rate": 2.742318893338187e-06, + "loss": 2.0698, + "step": 4769 + }, + { + "epoch": 0.87, + "learning_rate": 2.742264288314525e-06, + "loss": 1.6716, + "step": 4770 + }, + { + "epoch": 0.87, + "learning_rate": 2.742209683290863e-06, + "loss": 2.2368, + "step": 4771 + }, + { + "epoch": 0.87, + "learning_rate": 2.7421550782672004e-06, + "loss": 1.3978, + "step": 4772 + }, + { + "epoch": 0.87, + "learning_rate": 2.7421004732435384e-06, + "loss": 1.105, + "step": 4773 + }, + { + "epoch": 0.87, + "learning_rate": 2.7420458682198763e-06, + "loss": 2.7112, + "step": 4774 + }, + { + "epoch": 0.87, + "learning_rate": 2.7419912631962143e-06, + "loss": 0.724, + "step": 4775 + }, + { + "epoch": 0.87, + "learning_rate": 2.741936658172552e-06, + "loss": 1.2927, + "step": 4776 + }, + { + "epoch": 0.87, + "learning_rate": 2.7418820531488897e-06, + "loss": 0.4531, + "step": 4777 + }, + { + "epoch": 0.87, + "learning_rate": 2.7418274481252277e-06, + "loss": 0.7151, + "step": 4778 + }, + { + "epoch": 0.87, + "learning_rate": 2.7417728431015656e-06, + "loss": 1.5877, + "step": 4779 + }, + { + "epoch": 0.87, + "learning_rate": 2.741718238077903e-06, + "loss": 1.4075, + "step": 4780 + }, + { + "epoch": 0.87, + "learning_rate": 2.741663633054241e-06, + "loss": 0.967, + "step": 4781 + }, + { + "epoch": 0.87, + "learning_rate": 2.741609028030579e-06, + "loss": 0.518, + "step": 4782 + }, + { + "epoch": 0.87, + "learning_rate": 2.7415544230069165e-06, + "loss": 0.2727, + "step": 4783 + }, + { + "epoch": 0.87, + "learning_rate": 2.7414998179832545e-06, + "loss": 0.0554, + "step": 4784 + }, + { + "epoch": 0.87, + "learning_rate": 2.7414452129595924e-06, + "loss": 1.4177, + "step": 4785 + }, + { + "epoch": 0.87, + "learning_rate": 2.74139060793593e-06, + "loss": 2.1873, + "step": 4786 + }, + { + "epoch": 0.87, + "learning_rate": 2.741336002912268e-06, + "loss": 0.6273, + "step": 4787 + }, + { + "epoch": 0.87, + "learning_rate": 2.741281397888606e-06, + "loss": 0.7715, + "step": 4788 + }, + { + "epoch": 0.87, + "learning_rate": 2.7412267928649438e-06, + "loss": 1.3006, + "step": 4789 + }, + { + "epoch": 0.87, + "learning_rate": 2.7411721878412817e-06, + "loss": 1.2484, + "step": 4790 + }, + { + "epoch": 0.87, + "learning_rate": 2.7411175828176192e-06, + "loss": 1.0124, + "step": 4791 + }, + { + "epoch": 0.87, + "learning_rate": 2.741062977793957e-06, + "loss": 2.5755, + "step": 4792 + }, + { + "epoch": 0.87, + "learning_rate": 2.741008372770295e-06, + "loss": 1.3833, + "step": 4793 + }, + { + "epoch": 0.87, + "learning_rate": 2.7409537677466326e-06, + "loss": 0.2745, + "step": 4794 + }, + { + "epoch": 0.87, + "learning_rate": 2.7408991627229706e-06, + "loss": 0.0682, + "step": 4795 + }, + { + "epoch": 0.87, + "learning_rate": 2.7408445576993085e-06, + "loss": 0.3327, + "step": 4796 + }, + { + "epoch": 0.87, + "learning_rate": 2.740789952675646e-06, + "loss": 1.4141, + "step": 4797 + }, + { + "epoch": 0.87, + "learning_rate": 2.740735347651984e-06, + "loss": 2.3468, + "step": 4798 + }, + { + "epoch": 0.87, + "learning_rate": 2.740680742628322e-06, + "loss": 0.2661, + "step": 4799 + }, + { + "epoch": 0.87, + "learning_rate": 2.7406261376046595e-06, + "loss": 0.0558, + "step": 4800 + }, + { + "epoch": 0.87, + "learning_rate": 2.7405715325809974e-06, + "loss": 2.5867, + "step": 4801 + }, + { + "epoch": 0.87, + "learning_rate": 2.7405169275573353e-06, + "loss": 0.5728, + "step": 4802 + }, + { + "epoch": 0.87, + "learning_rate": 2.7404623225336733e-06, + "loss": 2.0231, + "step": 4803 + }, + { + "epoch": 0.87, + "learning_rate": 2.7404077175100112e-06, + "loss": 1.5627, + "step": 4804 + }, + { + "epoch": 0.87, + "learning_rate": 2.7403531124863488e-06, + "loss": 0.8419, + "step": 4805 + }, + { + "epoch": 0.87, + "learning_rate": 2.7402985074626867e-06, + "loss": 0.0503, + "step": 4806 + }, + { + "epoch": 0.87, + "learning_rate": 2.7402439024390246e-06, + "loss": 0.5027, + "step": 4807 + }, + { + "epoch": 0.87, + "learning_rate": 2.740189297415362e-06, + "loss": 2.9373, + "step": 4808 + }, + { + "epoch": 0.87, + "learning_rate": 2.7401346923917e-06, + "loss": 0.4374, + "step": 4809 + }, + { + "epoch": 0.87, + "learning_rate": 2.740080087368038e-06, + "loss": 1.5132, + "step": 4810 + }, + { + "epoch": 0.87, + "learning_rate": 2.7400254823443756e-06, + "loss": 0.732, + "step": 4811 + }, + { + "epoch": 0.88, + "learning_rate": 2.7399708773207135e-06, + "loss": 0.9389, + "step": 4812 + }, + { + "epoch": 0.88, + "learning_rate": 2.7399162722970515e-06, + "loss": 1.2559, + "step": 4813 + }, + { + "epoch": 0.88, + "learning_rate": 2.739861667273389e-06, + "loss": 0.2369, + "step": 4814 + }, + { + "epoch": 0.88, + "learning_rate": 2.739807062249727e-06, + "loss": 0.3904, + "step": 4815 + }, + { + "epoch": 0.88, + "learning_rate": 2.739752457226065e-06, + "loss": 0.9256, + "step": 4816 + }, + { + "epoch": 0.88, + "learning_rate": 2.739697852202403e-06, + "loss": 1.318, + "step": 4817 + }, + { + "epoch": 0.88, + "learning_rate": 2.7396432471787407e-06, + "loss": 1.4873, + "step": 4818 + }, + { + "epoch": 0.88, + "learning_rate": 2.7395886421550783e-06, + "loss": 1.0855, + "step": 4819 + }, + { + "epoch": 0.88, + "learning_rate": 2.739534037131416e-06, + "loss": 0.3957, + "step": 4820 + }, + { + "epoch": 0.88, + "learning_rate": 2.739479432107754e-06, + "loss": 0.8253, + "step": 4821 + }, + { + "epoch": 0.88, + "learning_rate": 2.7394248270840917e-06, + "loss": 0.0405, + "step": 4822 + }, + { + "epoch": 0.88, + "learning_rate": 2.7393702220604296e-06, + "loss": 0.3096, + "step": 4823 + }, + { + "epoch": 0.88, + "learning_rate": 2.7393156170367676e-06, + "loss": 1.7598, + "step": 4824 + }, + { + "epoch": 0.88, + "learning_rate": 2.739261012013105e-06, + "loss": 0.7236, + "step": 4825 + }, + { + "epoch": 0.88, + "learning_rate": 2.739206406989443e-06, + "loss": 0.5163, + "step": 4826 + }, + { + "epoch": 0.88, + "learning_rate": 2.739151801965781e-06, + "loss": 2.1936, + "step": 4827 + }, + { + "epoch": 0.88, + "learning_rate": 2.7390971969421185e-06, + "loss": 0.7022, + "step": 4828 + }, + { + "epoch": 0.88, + "learning_rate": 2.7390425919184564e-06, + "loss": 1.2885, + "step": 4829 + }, + { + "epoch": 0.88, + "learning_rate": 2.7389879868947944e-06, + "loss": 0.344, + "step": 4830 + }, + { + "epoch": 0.88, + "learning_rate": 2.7389333818711323e-06, + "loss": 0.718, + "step": 4831 + }, + { + "epoch": 0.88, + "learning_rate": 2.7388787768474703e-06, + "loss": 1.4486, + "step": 4832 + }, + { + "epoch": 0.88, + "learning_rate": 2.7388241718238078e-06, + "loss": 0.8779, + "step": 4833 + }, + { + "epoch": 0.88, + "learning_rate": 2.7387695668001457e-06, + "loss": 1.5381, + "step": 4834 + }, + { + "epoch": 0.88, + "learning_rate": 2.7387149617764837e-06, + "loss": 1.0423, + "step": 4835 + }, + { + "epoch": 0.88, + "learning_rate": 2.738660356752821e-06, + "loss": 1.2587, + "step": 4836 + }, + { + "epoch": 0.88, + "learning_rate": 2.738605751729159e-06, + "loss": 3.2164, + "step": 4837 + }, + { + "epoch": 0.88, + "learning_rate": 2.738551146705497e-06, + "loss": 1.9046, + "step": 4838 + }, + { + "epoch": 0.88, + "learning_rate": 2.7384965416818346e-06, + "loss": 1.5724, + "step": 4839 + }, + { + "epoch": 0.88, + "learning_rate": 2.7384419366581725e-06, + "loss": 2.3939, + "step": 4840 + }, + { + "epoch": 0.88, + "learning_rate": 2.7383873316345105e-06, + "loss": 1.1644, + "step": 4841 + }, + { + "epoch": 0.88, + "learning_rate": 2.7383327266108484e-06, + "loss": 0.0374, + "step": 4842 + }, + { + "epoch": 0.88, + "learning_rate": 2.7382781215871864e-06, + "loss": 0.3523, + "step": 4843 + }, + { + "epoch": 0.88, + "learning_rate": 2.738223516563524e-06, + "loss": 0.5119, + "step": 4844 + }, + { + "epoch": 0.88, + "learning_rate": 2.738168911539862e-06, + "loss": 3.535, + "step": 4845 + }, + { + "epoch": 0.88, + "learning_rate": 2.7381143065161998e-06, + "loss": 0.0881, + "step": 4846 + }, + { + "epoch": 0.88, + "learning_rate": 2.7380597014925373e-06, + "loss": 0.4037, + "step": 4847 + }, + { + "epoch": 0.88, + "learning_rate": 2.7380050964688752e-06, + "loss": 2.2219, + "step": 4848 + }, + { + "epoch": 0.88, + "learning_rate": 2.737950491445213e-06, + "loss": 1.4974, + "step": 4849 + }, + { + "epoch": 0.88, + "learning_rate": 2.7378958864215507e-06, + "loss": 0.6794, + "step": 4850 + }, + { + "epoch": 0.88, + "learning_rate": 2.7378412813978886e-06, + "loss": 1.4425, + "step": 4851 + }, + { + "epoch": 0.88, + "learning_rate": 2.7377866763742266e-06, + "loss": 1.3154, + "step": 4852 + }, + { + "epoch": 0.88, + "learning_rate": 2.737732071350564e-06, + "loss": 1.1865, + "step": 4853 + }, + { + "epoch": 0.88, + "learning_rate": 2.737677466326902e-06, + "loss": 1.0554, + "step": 4854 + }, + { + "epoch": 0.88, + "learning_rate": 2.73762286130324e-06, + "loss": 0.5213, + "step": 4855 + }, + { + "epoch": 0.88, + "learning_rate": 2.737568256279578e-06, + "loss": 0.0371, + "step": 4856 + }, + { + "epoch": 0.88, + "learning_rate": 2.737513651255916e-06, + "loss": 1.9584, + "step": 4857 + }, + { + "epoch": 0.88, + "learning_rate": 2.7374590462322534e-06, + "loss": 0.7126, + "step": 4858 + }, + { + "epoch": 0.88, + "learning_rate": 2.7374044412085913e-06, + "loss": 0.5611, + "step": 4859 + }, + { + "epoch": 0.88, + "learning_rate": 2.7373498361849293e-06, + "loss": 0.9623, + "step": 4860 + }, + { + "epoch": 0.88, + "learning_rate": 2.737295231161267e-06, + "loss": 0.9315, + "step": 4861 + }, + { + "epoch": 0.88, + "learning_rate": 2.7372406261376047e-06, + "loss": 2.3832, + "step": 4862 + }, + { + "epoch": 0.88, + "learning_rate": 2.7371860211139427e-06, + "loss": 0.071, + "step": 4863 + }, + { + "epoch": 0.88, + "learning_rate": 2.7371314160902802e-06, + "loss": 0.7231, + "step": 4864 + }, + { + "epoch": 0.88, + "learning_rate": 2.737076811066618e-06, + "loss": 1.0941, + "step": 4865 + }, + { + "epoch": 0.88, + "learning_rate": 2.737022206042956e-06, + "loss": 0.5704, + "step": 4866 + }, + { + "epoch": 0.89, + "learning_rate": 2.7369676010192936e-06, + "loss": 1.3532, + "step": 4867 + }, + { + "epoch": 0.89, + "learning_rate": 2.7369129959956316e-06, + "loss": 0.3856, + "step": 4868 + }, + { + "epoch": 0.89, + "learning_rate": 2.7368583909719695e-06, + "loss": 2.3717, + "step": 4869 + }, + { + "epoch": 0.89, + "learning_rate": 2.7368037859483075e-06, + "loss": 0.9415, + "step": 4870 + }, + { + "epoch": 0.89, + "learning_rate": 2.7367491809246454e-06, + "loss": 0.5144, + "step": 4871 + }, + { + "epoch": 0.89, + "learning_rate": 2.736694575900983e-06, + "loss": 0.3867, + "step": 4872 + }, + { + "epoch": 0.89, + "learning_rate": 2.736639970877321e-06, + "loss": 0.2626, + "step": 4873 + }, + { + "epoch": 0.89, + "learning_rate": 2.736585365853659e-06, + "loss": 1.6056, + "step": 4874 + }, + { + "epoch": 0.89, + "learning_rate": 2.7365307608299963e-06, + "loss": 1.836, + "step": 4875 + }, + { + "epoch": 0.89, + "learning_rate": 2.7364761558063343e-06, + "loss": 0.4858, + "step": 4876 + }, + { + "epoch": 0.89, + "learning_rate": 2.736421550782672e-06, + "loss": 0.8036, + "step": 4877 + }, + { + "epoch": 0.89, + "learning_rate": 2.7363669457590097e-06, + "loss": 0.6625, + "step": 4878 + }, + { + "epoch": 0.89, + "learning_rate": 2.7363123407353477e-06, + "loss": 0.807, + "step": 4879 + }, + { + "epoch": 0.89, + "learning_rate": 2.7362577357116856e-06, + "loss": 1.2321, + "step": 4880 + }, + { + "epoch": 0.89, + "learning_rate": 2.736203130688023e-06, + "loss": 1.021, + "step": 4881 + }, + { + "epoch": 0.89, + "learning_rate": 2.736148525664361e-06, + "loss": 0.5475, + "step": 4882 + }, + { + "epoch": 0.89, + "learning_rate": 2.736093920640699e-06, + "loss": 3.4475, + "step": 4883 + }, + { + "epoch": 0.89, + "learning_rate": 2.736039315617037e-06, + "loss": 0.4259, + "step": 4884 + }, + { + "epoch": 0.89, + "learning_rate": 2.735984710593375e-06, + "loss": 2.3809, + "step": 4885 + }, + { + "epoch": 0.89, + "learning_rate": 2.7359301055697124e-06, + "loss": 1.1496, + "step": 4886 + }, + { + "epoch": 0.89, + "learning_rate": 2.7358755005460504e-06, + "loss": 0.7495, + "step": 4887 + }, + { + "epoch": 0.89, + "learning_rate": 2.7358208955223883e-06, + "loss": 0.7142, + "step": 4888 + }, + { + "epoch": 0.89, + "learning_rate": 2.735766290498726e-06, + "loss": 0.3448, + "step": 4889 + }, + { + "epoch": 0.89, + "learning_rate": 2.7357116854750638e-06, + "loss": 1.0468, + "step": 4890 + }, + { + "epoch": 0.89, + "learning_rate": 2.7356570804514017e-06, + "loss": 0.0325, + "step": 4891 + }, + { + "epoch": 0.89, + "learning_rate": 2.7356024754277392e-06, + "loss": 1.5866, + "step": 4892 + }, + { + "epoch": 0.89, + "learning_rate": 2.735547870404077e-06, + "loss": 1.4011, + "step": 4893 + }, + { + "epoch": 0.89, + "learning_rate": 2.735493265380415e-06, + "loss": 0.7795, + "step": 4894 + }, + { + "epoch": 0.89, + "learning_rate": 2.7354386603567526e-06, + "loss": 0.4611, + "step": 4895 + }, + { + "epoch": 0.89, + "learning_rate": 2.7353840553330906e-06, + "loss": 1.5386, + "step": 4896 + }, + { + "epoch": 0.89, + "learning_rate": 2.7353294503094285e-06, + "loss": 0.4545, + "step": 4897 + }, + { + "epoch": 0.89, + "learning_rate": 2.7352748452857665e-06, + "loss": 1.0879, + "step": 4898 + }, + { + "epoch": 0.89, + "learning_rate": 2.7352202402621044e-06, + "loss": 1.7974, + "step": 4899 + }, + { + "epoch": 0.89, + "learning_rate": 2.735165635238442e-06, + "loss": 1.1519, + "step": 4900 + }, + { + "epoch": 0.89, + "learning_rate": 2.73511103021478e-06, + "loss": 1.0372, + "step": 4901 + }, + { + "epoch": 0.89, + "learning_rate": 2.735056425191118e-06, + "loss": 1.0984, + "step": 4902 + }, + { + "epoch": 0.89, + "learning_rate": 2.7350018201674553e-06, + "loss": 1.251, + "step": 4903 + }, + { + "epoch": 0.89, + "learning_rate": 2.7349472151437933e-06, + "loss": 0.7319, + "step": 4904 + }, + { + "epoch": 0.89, + "learning_rate": 2.7348926101201312e-06, + "loss": 2.0746, + "step": 4905 + }, + { + "epoch": 0.89, + "learning_rate": 2.7348380050964688e-06, + "loss": 0.9187, + "step": 4906 + }, + { + "epoch": 0.89, + "learning_rate": 2.7347834000728067e-06, + "loss": 0.095, + "step": 4907 + }, + { + "epoch": 0.89, + "learning_rate": 2.7347287950491446e-06, + "loss": 0.2307, + "step": 4908 + }, + { + "epoch": 0.89, + "learning_rate": 2.734674190025482e-06, + "loss": 0.91, + "step": 4909 + }, + { + "epoch": 0.89, + "learning_rate": 2.73461958500182e-06, + "loss": 0.0695, + "step": 4910 + }, + { + "epoch": 0.89, + "learning_rate": 2.734564979978158e-06, + "loss": 0.2023, + "step": 4911 + }, + { + "epoch": 0.89, + "learning_rate": 2.734510374954496e-06, + "loss": 2.3209, + "step": 4912 + }, + { + "epoch": 0.89, + "learning_rate": 2.734455769930834e-06, + "loss": 0.3487, + "step": 4913 + }, + { + "epoch": 0.89, + "learning_rate": 2.7344011649071715e-06, + "loss": 0.6811, + "step": 4914 + }, + { + "epoch": 0.89, + "learning_rate": 2.7343465598835094e-06, + "loss": 4.2086, + "step": 4915 + }, + { + "epoch": 0.89, + "learning_rate": 2.7342919548598473e-06, + "loss": 0.3882, + "step": 4916 + }, + { + "epoch": 0.89, + "learning_rate": 2.734237349836185e-06, + "loss": 1.7774, + "step": 4917 + }, + { + "epoch": 0.89, + "learning_rate": 2.734182744812523e-06, + "loss": 0.6778, + "step": 4918 + }, + { + "epoch": 0.89, + "learning_rate": 2.7341281397888607e-06, + "loss": 0.5737, + "step": 4919 + }, + { + "epoch": 0.89, + "learning_rate": 2.7340735347651983e-06, + "loss": 1.8083, + "step": 4920 + }, + { + "epoch": 0.89, + "learning_rate": 2.734018929741536e-06, + "loss": 0.0193, + "step": 4921 + }, + { + "epoch": 0.9, + "learning_rate": 2.733964324717874e-06, + "loss": 1.1502, + "step": 4922 + }, + { + "epoch": 0.9, + "learning_rate": 2.7339097196942117e-06, + "loss": 0.6113, + "step": 4923 + }, + { + "epoch": 0.9, + "learning_rate": 2.7338551146705496e-06, + "loss": 1.1096, + "step": 4924 + }, + { + "epoch": 0.9, + "learning_rate": 2.7338005096468876e-06, + "loss": 1.03, + "step": 4925 + }, + { + "epoch": 0.9, + "learning_rate": 2.7337459046232255e-06, + "loss": 0.9827, + "step": 4926 + }, + { + "epoch": 0.9, + "learning_rate": 2.7336912995995634e-06, + "loss": 1.0248, + "step": 4927 + }, + { + "epoch": 0.9, + "learning_rate": 2.733636694575901e-06, + "loss": 1.5763, + "step": 4928 + }, + { + "epoch": 0.9, + "learning_rate": 2.733582089552239e-06, + "loss": 1.8122, + "step": 4929 + }, + { + "epoch": 0.9, + "learning_rate": 2.733527484528577e-06, + "loss": 0.0711, + "step": 4930 + }, + { + "epoch": 0.9, + "learning_rate": 2.7334728795049144e-06, + "loss": 1.7574, + "step": 4931 + }, + { + "epoch": 0.9, + "learning_rate": 2.7334182744812523e-06, + "loss": 0.905, + "step": 4932 + }, + { + "epoch": 0.9, + "learning_rate": 2.7333636694575903e-06, + "loss": 0.0373, + "step": 4933 + }, + { + "epoch": 0.9, + "learning_rate": 2.7333090644339278e-06, + "loss": 2.5324, + "step": 4934 + }, + { + "epoch": 0.9, + "learning_rate": 2.7332544594102657e-06, + "loss": 1.2117, + "step": 4935 + }, + { + "epoch": 0.9, + "learning_rate": 2.7331998543866037e-06, + "loss": 0.4016, + "step": 4936 + }, + { + "epoch": 0.9, + "learning_rate": 2.7331452493629416e-06, + "loss": 0.9419, + "step": 4937 + }, + { + "epoch": 0.9, + "learning_rate": 2.7330906443392796e-06, + "loss": 0.9956, + "step": 4938 + }, + { + "epoch": 0.9, + "learning_rate": 2.733036039315617e-06, + "loss": 2.2057, + "step": 4939 + }, + { + "epoch": 0.9, + "learning_rate": 2.732981434291955e-06, + "loss": 1.3228, + "step": 4940 + }, + { + "epoch": 0.9, + "learning_rate": 2.732926829268293e-06, + "loss": 0.5245, + "step": 4941 + }, + { + "epoch": 0.9, + "learning_rate": 2.7328722242446305e-06, + "loss": 0.308, + "step": 4942 + }, + { + "epoch": 0.9, + "learning_rate": 2.7328176192209684e-06, + "loss": 0.3151, + "step": 4943 + }, + { + "epoch": 0.9, + "learning_rate": 2.7327630141973064e-06, + "loss": 0.1723, + "step": 4944 + }, + { + "epoch": 0.9, + "learning_rate": 2.732708409173644e-06, + "loss": 0.3162, + "step": 4945 + }, + { + "epoch": 0.9, + "learning_rate": 2.732653804149982e-06, + "loss": 1.5242, + "step": 4946 + }, + { + "epoch": 0.9, + "learning_rate": 2.7325991991263198e-06, + "loss": 0.6397, + "step": 4947 + }, + { + "epoch": 0.9, + "learning_rate": 2.7325445941026573e-06, + "loss": 0.8302, + "step": 4948 + }, + { + "epoch": 0.9, + "learning_rate": 2.7324899890789952e-06, + "loss": 0.8939, + "step": 4949 + }, + { + "epoch": 0.9, + "learning_rate": 2.732435384055333e-06, + "loss": 1.4732, + "step": 4950 + }, + { + "epoch": 0.9, + "learning_rate": 2.732380779031671e-06, + "loss": 0.4613, + "step": 4951 + }, + { + "epoch": 0.9, + "learning_rate": 2.732326174008009e-06, + "loss": 2.6429, + "step": 4952 + }, + { + "epoch": 0.9, + "learning_rate": 2.7322715689843466e-06, + "loss": 0.1377, + "step": 4953 + }, + { + "epoch": 0.9, + "learning_rate": 2.7322169639606845e-06, + "loss": 0.5692, + "step": 4954 + }, + { + "epoch": 0.9, + "learning_rate": 2.7321623589370225e-06, + "loss": 1.1249, + "step": 4955 + }, + { + "epoch": 0.9, + "learning_rate": 2.73210775391336e-06, + "loss": 1.2518, + "step": 4956 + }, + { + "epoch": 0.9, + "learning_rate": 2.732053148889698e-06, + "loss": 1.1891, + "step": 4957 + }, + { + "epoch": 0.9, + "learning_rate": 2.731998543866036e-06, + "loss": 0.3164, + "step": 4958 + }, + { + "epoch": 0.9, + "learning_rate": 2.7319439388423734e-06, + "loss": 1.85, + "step": 4959 + }, + { + "epoch": 0.9, + "learning_rate": 2.7318893338187113e-06, + "loss": 0.5268, + "step": 4960 + }, + { + "epoch": 0.9, + "learning_rate": 2.7318347287950493e-06, + "loss": 0.0822, + "step": 4961 + }, + { + "epoch": 0.9, + "learning_rate": 2.731780123771387e-06, + "loss": 1.4321, + "step": 4962 + }, + { + "epoch": 0.9, + "learning_rate": 2.7317255187477248e-06, + "loss": 0.8573, + "step": 4963 + }, + { + "epoch": 0.9, + "learning_rate": 2.7316709137240627e-06, + "loss": 1.2384, + "step": 4964 + }, + { + "epoch": 0.9, + "learning_rate": 2.7316163087004006e-06, + "loss": 1.3224, + "step": 4965 + }, + { + "epoch": 0.9, + "learning_rate": 2.7315617036767386e-06, + "loss": 0.2114, + "step": 4966 + }, + { + "epoch": 0.9, + "learning_rate": 2.731507098653076e-06, + "loss": 1.103, + "step": 4967 + }, + { + "epoch": 0.9, + "learning_rate": 2.731452493629414e-06, + "loss": 1.8396, + "step": 4968 + }, + { + "epoch": 0.9, + "learning_rate": 2.731397888605752e-06, + "loss": 1.3375, + "step": 4969 + }, + { + "epoch": 0.9, + "learning_rate": 2.7313432835820895e-06, + "loss": 1.5727, + "step": 4970 + }, + { + "epoch": 0.9, + "learning_rate": 2.7312886785584275e-06, + "loss": 0.7164, + "step": 4971 + }, + { + "epoch": 0.9, + "learning_rate": 2.7312340735347654e-06, + "loss": 0.7925, + "step": 4972 + }, + { + "epoch": 0.9, + "learning_rate": 2.731179468511103e-06, + "loss": 0.3937, + "step": 4973 + }, + { + "epoch": 0.9, + "learning_rate": 2.731124863487441e-06, + "loss": 1.7784, + "step": 4974 + }, + { + "epoch": 0.9, + "learning_rate": 2.731070258463779e-06, + "loss": 1.8702, + "step": 4975 + }, + { + "epoch": 0.9, + "learning_rate": 2.7310156534401163e-06, + "loss": 1.5441, + "step": 4976 + }, + { + "epoch": 0.91, + "learning_rate": 2.7309610484164543e-06, + "loss": 4.9607, + "step": 4977 + }, + { + "epoch": 0.91, + "learning_rate": 2.730906443392792e-06, + "loss": 1.5872, + "step": 4978 + }, + { + "epoch": 0.91, + "learning_rate": 2.73085183836913e-06, + "loss": 0.6, + "step": 4979 + }, + { + "epoch": 0.91, + "learning_rate": 2.730797233345468e-06, + "loss": 0.1229, + "step": 4980 + }, + { + "epoch": 0.91, + "learning_rate": 2.7307426283218056e-06, + "loss": 1.5527, + "step": 4981 + }, + { + "epoch": 0.91, + "learning_rate": 2.7306880232981436e-06, + "loss": 0.091, + "step": 4982 + }, + { + "epoch": 0.91, + "learning_rate": 2.7306334182744815e-06, + "loss": 1.9676, + "step": 4983 + }, + { + "epoch": 0.91, + "learning_rate": 2.730578813250819e-06, + "loss": 0.7866, + "step": 4984 + }, + { + "epoch": 0.91, + "learning_rate": 2.730524208227157e-06, + "loss": 1.5656, + "step": 4985 + }, + { + "epoch": 0.91, + "learning_rate": 2.730469603203495e-06, + "loss": 2.2949, + "step": 4986 + }, + { + "epoch": 0.91, + "learning_rate": 2.7304149981798324e-06, + "loss": 0.965, + "step": 4987 + }, + { + "epoch": 0.91, + "learning_rate": 2.7303603931561704e-06, + "loss": 1.0023, + "step": 4988 + }, + { + "epoch": 0.91, + "learning_rate": 2.7303057881325083e-06, + "loss": 0.4794, + "step": 4989 + }, + { + "epoch": 0.91, + "learning_rate": 2.730251183108846e-06, + "loss": 1.3349, + "step": 4990 + }, + { + "epoch": 0.91, + "learning_rate": 2.7301965780851838e-06, + "loss": 1.0255, + "step": 4991 + }, + { + "epoch": 0.91, + "learning_rate": 2.7301419730615217e-06, + "loss": 0.2637, + "step": 4992 + }, + { + "epoch": 0.91, + "learning_rate": 2.7300873680378597e-06, + "loss": 1.5955, + "step": 4993 + }, + { + "epoch": 0.91, + "learning_rate": 2.7300327630141976e-06, + "loss": 0.4604, + "step": 4994 + }, + { + "epoch": 0.91, + "learning_rate": 2.729978157990535e-06, + "loss": 0.1338, + "step": 4995 + }, + { + "epoch": 0.91, + "learning_rate": 2.729923552966873e-06, + "loss": 0.5728, + "step": 4996 + }, + { + "epoch": 0.91, + "learning_rate": 2.729868947943211e-06, + "loss": 1.0277, + "step": 4997 + }, + { + "epoch": 0.91, + "learning_rate": 2.7298143429195485e-06, + "loss": 0.815, + "step": 4998 + }, + { + "epoch": 0.91, + "learning_rate": 2.7297597378958865e-06, + "loss": 1.3249, + "step": 4999 + }, + { + "epoch": 0.91, + "learning_rate": 2.7297051328722244e-06, + "loss": 1.4939, + "step": 5000 + }, + { + "epoch": 0.91, + "learning_rate": 2.729650527848562e-06, + "loss": 1.2343, + "step": 5001 + }, + { + "epoch": 0.91, + "learning_rate": 2.7295959228249e-06, + "loss": 0.7773, + "step": 5002 + }, + { + "epoch": 0.91, + "learning_rate": 2.729541317801238e-06, + "loss": 0.7059, + "step": 5003 + }, + { + "epoch": 0.91, + "learning_rate": 2.7294867127775753e-06, + "loss": 0.2202, + "step": 5004 + }, + { + "epoch": 0.91, + "learning_rate": 2.7294321077539133e-06, + "loss": 1.8118, + "step": 5005 + }, + { + "epoch": 0.91, + "learning_rate": 2.7293775027302512e-06, + "loss": 0.5257, + "step": 5006 + }, + { + "epoch": 0.91, + "learning_rate": 2.729322897706589e-06, + "loss": 0.9699, + "step": 5007 + }, + { + "epoch": 0.91, + "learning_rate": 2.729268292682927e-06, + "loss": 1.2904, + "step": 5008 + }, + { + "epoch": 0.91, + "learning_rate": 2.7292136876592646e-06, + "loss": 0.3076, + "step": 5009 + }, + { + "epoch": 0.91, + "learning_rate": 2.7291590826356026e-06, + "loss": 0.8605, + "step": 5010 + }, + { + "epoch": 0.91, + "learning_rate": 2.7291044776119405e-06, + "loss": 0.8328, + "step": 5011 + }, + { + "epoch": 0.91, + "learning_rate": 2.729049872588278e-06, + "loss": 1.2682, + "step": 5012 + }, + { + "epoch": 0.91, + "learning_rate": 2.728995267564616e-06, + "loss": 2.2278, + "step": 5013 + }, + { + "epoch": 0.91, + "learning_rate": 2.728940662540954e-06, + "loss": 0.0683, + "step": 5014 + }, + { + "epoch": 0.91, + "learning_rate": 2.7288860575172915e-06, + "loss": 0.4824, + "step": 5015 + }, + { + "epoch": 0.91, + "learning_rate": 2.7288314524936294e-06, + "loss": 1.7431, + "step": 5016 + }, + { + "epoch": 0.91, + "learning_rate": 2.7287768474699673e-06, + "loss": 2.2334, + "step": 5017 + }, + { + "epoch": 0.91, + "learning_rate": 2.7287222424463053e-06, + "loss": 0.4952, + "step": 5018 + }, + { + "epoch": 0.91, + "learning_rate": 2.7286676374226432e-06, + "loss": 1.7637, + "step": 5019 + }, + { + "epoch": 0.91, + "learning_rate": 2.7286130323989807e-06, + "loss": 1.0713, + "step": 5020 + }, + { + "epoch": 0.91, + "learning_rate": 2.7285584273753187e-06, + "loss": 0.0384, + "step": 5021 + }, + { + "epoch": 0.91, + "learning_rate": 2.7285038223516566e-06, + "loss": 0.9081, + "step": 5022 + }, + { + "epoch": 0.91, + "learning_rate": 2.728449217327994e-06, + "loss": 1.8958, + "step": 5023 + }, + { + "epoch": 0.91, + "learning_rate": 2.728394612304332e-06, + "loss": 0.774, + "step": 5024 + }, + { + "epoch": 0.91, + "learning_rate": 2.72834000728067e-06, + "loss": 3.1266, + "step": 5025 + }, + { + "epoch": 0.91, + "learning_rate": 2.7282854022570076e-06, + "loss": 1.0573, + "step": 5026 + }, + { + "epoch": 0.91, + "learning_rate": 2.7282307972333455e-06, + "loss": 1.9771, + "step": 5027 + }, + { + "epoch": 0.91, + "learning_rate": 2.7281761922096834e-06, + "loss": 0.4412, + "step": 5028 + }, + { + "epoch": 0.91, + "learning_rate": 2.728121587186021e-06, + "loss": 2.2094, + "step": 5029 + }, + { + "epoch": 0.91, + "learning_rate": 2.728066982162359e-06, + "loss": 0.5672, + "step": 5030 + }, + { + "epoch": 0.91, + "learning_rate": 2.728012377138697e-06, + "loss": 0.7858, + "step": 5031 + }, + { + "epoch": 0.92, + "learning_rate": 2.727957772115035e-06, + "loss": 1.1946, + "step": 5032 + }, + { + "epoch": 0.92, + "learning_rate": 2.7279031670913727e-06, + "loss": 0.0365, + "step": 5033 + }, + { + "epoch": 0.92, + "learning_rate": 2.7278485620677103e-06, + "loss": 2.3431, + "step": 5034 + }, + { + "epoch": 0.92, + "learning_rate": 2.727793957044048e-06, + "loss": 2.1613, + "step": 5035 + }, + { + "epoch": 0.92, + "learning_rate": 2.727739352020386e-06, + "loss": 0.569, + "step": 5036 + }, + { + "epoch": 0.92, + "learning_rate": 2.7276847469967237e-06, + "loss": 0.9061, + "step": 5037 + }, + { + "epoch": 0.92, + "learning_rate": 2.7276301419730616e-06, + "loss": 0.6149, + "step": 5038 + }, + { + "epoch": 0.92, + "learning_rate": 2.7275755369493996e-06, + "loss": 0.8916, + "step": 5039 + }, + { + "epoch": 0.92, + "learning_rate": 2.727520931925737e-06, + "loss": 0.6406, + "step": 5040 + }, + { + "epoch": 0.92, + "learning_rate": 2.727466326902075e-06, + "loss": 1.3989, + "step": 5041 + }, + { + "epoch": 0.92, + "learning_rate": 2.727411721878413e-06, + "loss": 0.8319, + "step": 5042 + }, + { + "epoch": 0.92, + "learning_rate": 2.7273571168547505e-06, + "loss": 1.261, + "step": 5043 + }, + { + "epoch": 0.92, + "learning_rate": 2.7273025118310884e-06, + "loss": 0.1454, + "step": 5044 + }, + { + "epoch": 0.92, + "learning_rate": 2.7272479068074264e-06, + "loss": 0.5396, + "step": 5045 + }, + { + "epoch": 0.92, + "learning_rate": 2.7271933017837643e-06, + "loss": 0.126, + "step": 5046 + }, + { + "epoch": 0.92, + "learning_rate": 2.7271386967601023e-06, + "loss": 0.612, + "step": 5047 + }, + { + "epoch": 0.92, + "learning_rate": 2.7270840917364398e-06, + "loss": 0.4071, + "step": 5048 + }, + { + "epoch": 0.92, + "learning_rate": 2.7270294867127777e-06, + "loss": 1.9768, + "step": 5049 + }, + { + "epoch": 0.92, + "learning_rate": 2.7269748816891157e-06, + "loss": 0.5633, + "step": 5050 + }, + { + "epoch": 0.92, + "learning_rate": 2.726920276665453e-06, + "loss": 0.4353, + "step": 5051 + }, + { + "epoch": 0.92, + "learning_rate": 2.726865671641791e-06, + "loss": 1.4669, + "step": 5052 + }, + { + "epoch": 0.92, + "learning_rate": 2.726811066618129e-06, + "loss": 0.7608, + "step": 5053 + }, + { + "epoch": 0.92, + "learning_rate": 2.7267564615944666e-06, + "loss": 1.5931, + "step": 5054 + }, + { + "epoch": 0.92, + "learning_rate": 2.7267018565708045e-06, + "loss": 0.5331, + "step": 5055 + }, + { + "epoch": 0.92, + "learning_rate": 2.7266472515471425e-06, + "loss": 0.5122, + "step": 5056 + }, + { + "epoch": 0.92, + "learning_rate": 2.72659264652348e-06, + "loss": 2.0583, + "step": 5057 + }, + { + "epoch": 0.92, + "learning_rate": 2.726538041499818e-06, + "loss": 0.8765, + "step": 5058 + }, + { + "epoch": 0.92, + "learning_rate": 2.726483436476156e-06, + "loss": 1.0039, + "step": 5059 + }, + { + "epoch": 0.92, + "learning_rate": 2.726428831452494e-06, + "loss": 1.6904, + "step": 5060 + }, + { + "epoch": 0.92, + "learning_rate": 2.7263742264288318e-06, + "loss": 0.1307, + "step": 5061 + }, + { + "epoch": 0.92, + "learning_rate": 2.7263196214051693e-06, + "loss": 1.4878, + "step": 5062 + }, + { + "epoch": 0.92, + "learning_rate": 2.7262650163815072e-06, + "loss": 0.2062, + "step": 5063 + }, + { + "epoch": 0.92, + "learning_rate": 2.726210411357845e-06, + "loss": 1.1233, + "step": 5064 + }, + { + "epoch": 0.92, + "learning_rate": 2.7261558063341827e-06, + "loss": 0.705, + "step": 5065 + }, + { + "epoch": 0.92, + "learning_rate": 2.7261012013105206e-06, + "loss": 0.5637, + "step": 5066 + }, + { + "epoch": 0.92, + "learning_rate": 2.7260465962868586e-06, + "loss": 0.0536, + "step": 5067 + }, + { + "epoch": 0.92, + "learning_rate": 2.725991991263196e-06, + "loss": 1.224, + "step": 5068 + }, + { + "epoch": 0.92, + "learning_rate": 2.725937386239534e-06, + "loss": 1.4031, + "step": 5069 + }, + { + "epoch": 0.92, + "learning_rate": 2.725882781215872e-06, + "loss": 1.6204, + "step": 5070 + }, + { + "epoch": 0.92, + "learning_rate": 2.7258281761922095e-06, + "loss": 0.46, + "step": 5071 + }, + { + "epoch": 0.92, + "learning_rate": 2.7257735711685475e-06, + "loss": 3.3128, + "step": 5072 + }, + { + "epoch": 0.92, + "learning_rate": 2.7257189661448854e-06, + "loss": 1.9683, + "step": 5073 + }, + { + "epoch": 0.92, + "learning_rate": 2.7256643611212233e-06, + "loss": 0.7104, + "step": 5074 + }, + { + "epoch": 0.92, + "learning_rate": 2.7256097560975613e-06, + "loss": 0.7808, + "step": 5075 + }, + { + "epoch": 0.92, + "learning_rate": 2.725555151073899e-06, + "loss": 0.4199, + "step": 5076 + }, + { + "epoch": 0.92, + "learning_rate": 2.7255005460502367e-06, + "loss": 1.7127, + "step": 5077 + }, + { + "epoch": 0.92, + "learning_rate": 2.7254459410265747e-06, + "loss": 0.0891, + "step": 5078 + }, + { + "epoch": 0.92, + "learning_rate": 2.725391336002912e-06, + "loss": 0.7567, + "step": 5079 + }, + { + "epoch": 0.92, + "learning_rate": 2.72533673097925e-06, + "loss": 0.4629, + "step": 5080 + }, + { + "epoch": 0.92, + "learning_rate": 2.725282125955588e-06, + "loss": 0.0779, + "step": 5081 + }, + { + "epoch": 0.92, + "learning_rate": 2.7252275209319256e-06, + "loss": 1.1689, + "step": 5082 + }, + { + "epoch": 0.92, + "learning_rate": 2.7251729159082636e-06, + "loss": 1.4423, + "step": 5083 + }, + { + "epoch": 0.92, + "learning_rate": 2.7251183108846015e-06, + "loss": 1.6285, + "step": 5084 + }, + { + "epoch": 0.92, + "learning_rate": 2.725063705860939e-06, + "loss": 1.9069, + "step": 5085 + }, + { + "epoch": 0.92, + "learning_rate": 2.725009100837277e-06, + "loss": 0.3162, + "step": 5086 + }, + { + "epoch": 0.93, + "learning_rate": 2.724954495813615e-06, + "loss": 3.8896, + "step": 5087 + }, + { + "epoch": 0.93, + "learning_rate": 2.724899890789953e-06, + "loss": 0.2887, + "step": 5088 + }, + { + "epoch": 0.93, + "learning_rate": 2.724845285766291e-06, + "loss": 1.5085, + "step": 5089 + }, + { + "epoch": 0.93, + "learning_rate": 2.7247906807426283e-06, + "loss": 0.3453, + "step": 5090 + }, + { + "epoch": 0.93, + "learning_rate": 2.7247360757189663e-06, + "loss": 0.7845, + "step": 5091 + }, + { + "epoch": 0.93, + "learning_rate": 2.724681470695304e-06, + "loss": 0.8224, + "step": 5092 + }, + { + "epoch": 0.93, + "learning_rate": 2.7246268656716417e-06, + "loss": 1.037, + "step": 5093 + }, + { + "epoch": 0.93, + "learning_rate": 2.7245722606479797e-06, + "loss": 0.844, + "step": 5094 + }, + { + "epoch": 0.93, + "learning_rate": 2.7245176556243176e-06, + "loss": 1.3472, + "step": 5095 + }, + { + "epoch": 0.93, + "learning_rate": 2.724463050600655e-06, + "loss": 0.3105, + "step": 5096 + }, + { + "epoch": 0.93, + "learning_rate": 2.724408445576993e-06, + "loss": 0.6615, + "step": 5097 + }, + { + "epoch": 0.93, + "learning_rate": 2.724353840553331e-06, + "loss": 0.0503, + "step": 5098 + }, + { + "epoch": 0.93, + "learning_rate": 2.724299235529669e-06, + "loss": 1.3783, + "step": 5099 + }, + { + "epoch": 0.93, + "learning_rate": 2.724244630506007e-06, + "loss": 0.682, + "step": 5100 + }, + { + "epoch": 0.93, + "learning_rate": 2.7241900254823444e-06, + "loss": 1.517, + "step": 5101 + }, + { + "epoch": 0.93, + "learning_rate": 2.7241354204586824e-06, + "loss": 2.5755, + "step": 5102 + }, + { + "epoch": 0.93, + "learning_rate": 2.7240808154350203e-06, + "loss": 2.1433, + "step": 5103 + }, + { + "epoch": 0.93, + "learning_rate": 2.724026210411358e-06, + "loss": 2.118, + "step": 5104 + }, + { + "epoch": 0.93, + "learning_rate": 2.7239716053876958e-06, + "loss": 0.0867, + "step": 5105 + }, + { + "epoch": 0.93, + "learning_rate": 2.7239170003640337e-06, + "loss": 2.0719, + "step": 5106 + }, + { + "epoch": 0.93, + "learning_rate": 2.7238623953403712e-06, + "loss": 0.7703, + "step": 5107 + }, + { + "epoch": 0.93, + "learning_rate": 2.723807790316709e-06, + "loss": 1.2808, + "step": 5108 + }, + { + "epoch": 0.93, + "learning_rate": 2.723753185293047e-06, + "loss": 2.844, + "step": 5109 + }, + { + "epoch": 0.93, + "learning_rate": 2.7236985802693846e-06, + "loss": 0.7056, + "step": 5110 + }, + { + "epoch": 0.93, + "learning_rate": 2.7236439752457226e-06, + "loss": 1.4818, + "step": 5111 + }, + { + "epoch": 0.93, + "learning_rate": 2.7235893702220605e-06, + "loss": 1.3634, + "step": 5112 + }, + { + "epoch": 0.93, + "learning_rate": 2.7235347651983985e-06, + "loss": 1.2339, + "step": 5113 + }, + { + "epoch": 0.93, + "learning_rate": 2.7234801601747364e-06, + "loss": 0.0615, + "step": 5114 + }, + { + "epoch": 0.93, + "learning_rate": 2.723425555151074e-06, + "loss": 0.94, + "step": 5115 + }, + { + "epoch": 0.93, + "learning_rate": 2.723370950127412e-06, + "loss": 1.4484, + "step": 5116 + }, + { + "epoch": 0.93, + "learning_rate": 2.72331634510375e-06, + "loss": 0.1801, + "step": 5117 + }, + { + "epoch": 0.93, + "learning_rate": 2.7232617400800873e-06, + "loss": 1.5727, + "step": 5118 + }, + { + "epoch": 0.93, + "learning_rate": 2.7232071350564253e-06, + "loss": 1.117, + "step": 5119 + }, + { + "epoch": 0.93, + "learning_rate": 2.7231525300327632e-06, + "loss": 0.9351, + "step": 5120 + }, + { + "epoch": 0.93, + "learning_rate": 2.7230979250091007e-06, + "loss": 0.6628, + "step": 5121 + }, + { + "epoch": 0.93, + "learning_rate": 2.7230433199854387e-06, + "loss": 0.6552, + "step": 5122 + }, + { + "epoch": 0.93, + "learning_rate": 2.7229887149617766e-06, + "loss": 3.7322, + "step": 5123 + }, + { + "epoch": 0.93, + "learning_rate": 2.722934109938114e-06, + "loss": 0.1763, + "step": 5124 + }, + { + "epoch": 0.93, + "learning_rate": 2.722879504914452e-06, + "loss": 0.1865, + "step": 5125 + }, + { + "epoch": 0.93, + "learning_rate": 2.72282489989079e-06, + "loss": 1.6939, + "step": 5126 + }, + { + "epoch": 0.93, + "learning_rate": 2.722770294867128e-06, + "loss": 1.6472, + "step": 5127 + }, + { + "epoch": 0.93, + "learning_rate": 2.722715689843466e-06, + "loss": 0.5856, + "step": 5128 + }, + { + "epoch": 0.93, + "learning_rate": 2.7226610848198034e-06, + "loss": 1.4865, + "step": 5129 + }, + { + "epoch": 0.93, + "learning_rate": 2.7226064797961414e-06, + "loss": 1.2273, + "step": 5130 + }, + { + "epoch": 0.93, + "learning_rate": 2.7225518747724793e-06, + "loss": 1.4447, + "step": 5131 + }, + { + "epoch": 0.93, + "learning_rate": 2.722497269748817e-06, + "loss": 0.6341, + "step": 5132 + }, + { + "epoch": 0.93, + "learning_rate": 2.722442664725155e-06, + "loss": 0.0766, + "step": 5133 + }, + { + "epoch": 0.93, + "learning_rate": 2.7223880597014927e-06, + "loss": 0.7655, + "step": 5134 + }, + { + "epoch": 0.93, + "learning_rate": 2.7223334546778303e-06, + "loss": 1.1429, + "step": 5135 + }, + { + "epoch": 0.93, + "learning_rate": 2.722278849654168e-06, + "loss": 1.1817, + "step": 5136 + }, + { + "epoch": 0.93, + "learning_rate": 2.722224244630506e-06, + "loss": 1.1905, + "step": 5137 + }, + { + "epoch": 0.93, + "learning_rate": 2.7221696396068437e-06, + "loss": 0.7327, + "step": 5138 + }, + { + "epoch": 0.93, + "learning_rate": 2.7221150345831816e-06, + "loss": 0.0885, + "step": 5139 + }, + { + "epoch": 0.93, + "learning_rate": 2.7220604295595196e-06, + "loss": 1.5319, + "step": 5140 + }, + { + "epoch": 0.93, + "learning_rate": 2.7220058245358575e-06, + "loss": 1.1524, + "step": 5141 + }, + { + "epoch": 0.94, + "learning_rate": 2.7219512195121954e-06, + "loss": 1.3976, + "step": 5142 + }, + { + "epoch": 0.94, + "learning_rate": 2.721896614488533e-06, + "loss": 1.2038, + "step": 5143 + }, + { + "epoch": 0.94, + "learning_rate": 2.721842009464871e-06, + "loss": 2.2061, + "step": 5144 + }, + { + "epoch": 0.94, + "learning_rate": 2.721787404441209e-06, + "loss": 2.7425, + "step": 5145 + }, + { + "epoch": 0.94, + "learning_rate": 2.7217327994175464e-06, + "loss": 2.3293, + "step": 5146 + }, + { + "epoch": 0.94, + "learning_rate": 2.7216781943938843e-06, + "loss": 0.9874, + "step": 5147 + }, + { + "epoch": 0.94, + "learning_rate": 2.7216235893702223e-06, + "loss": 1.4374, + "step": 5148 + }, + { + "epoch": 0.94, + "learning_rate": 2.7215689843465598e-06, + "loss": 1.9528, + "step": 5149 + }, + { + "epoch": 0.94, + "learning_rate": 2.7215143793228977e-06, + "loss": 0.3015, + "step": 5150 + }, + { + "epoch": 0.94, + "learning_rate": 2.7214597742992357e-06, + "loss": 0.3019, + "step": 5151 + }, + { + "epoch": 0.94, + "learning_rate": 2.721405169275573e-06, + "loss": 0.8785, + "step": 5152 + }, + { + "epoch": 0.94, + "learning_rate": 2.721350564251911e-06, + "loss": 2.4153, + "step": 5153 + }, + { + "epoch": 0.94, + "learning_rate": 2.721295959228249e-06, + "loss": 1.4353, + "step": 5154 + }, + { + "epoch": 0.94, + "learning_rate": 2.721241354204587e-06, + "loss": 0.8539, + "step": 5155 + }, + { + "epoch": 0.94, + "learning_rate": 2.721186749180925e-06, + "loss": 3.6029, + "step": 5156 + }, + { + "epoch": 0.94, + "learning_rate": 2.7211321441572625e-06, + "loss": 2.0331, + "step": 5157 + }, + { + "epoch": 0.94, + "learning_rate": 2.7210775391336004e-06, + "loss": 1.5499, + "step": 5158 + }, + { + "epoch": 0.94, + "learning_rate": 2.7210229341099384e-06, + "loss": 0.6656, + "step": 5159 + }, + { + "epoch": 0.94, + "learning_rate": 2.720968329086276e-06, + "loss": 0.3386, + "step": 5160 + }, + { + "epoch": 0.94, + "learning_rate": 2.720913724062614e-06, + "loss": 0.143, + "step": 5161 + }, + { + "epoch": 0.94, + "learning_rate": 2.7208591190389518e-06, + "loss": 3.1278, + "step": 5162 + }, + { + "epoch": 0.94, + "learning_rate": 2.7208045140152893e-06, + "loss": 1.7743, + "step": 5163 + }, + { + "epoch": 0.94, + "learning_rate": 2.7207499089916272e-06, + "loss": 0.3556, + "step": 5164 + }, + { + "epoch": 0.94, + "learning_rate": 2.720695303967965e-06, + "loss": 1.6055, + "step": 5165 + }, + { + "epoch": 0.94, + "learning_rate": 2.7206406989443027e-06, + "loss": 0.7611, + "step": 5166 + }, + { + "epoch": 0.94, + "learning_rate": 2.7205860939206406e-06, + "loss": 0.4853, + "step": 5167 + }, + { + "epoch": 0.94, + "learning_rate": 2.7205314888969786e-06, + "loss": 0.8273, + "step": 5168 + }, + { + "epoch": 0.94, + "learning_rate": 2.7204768838733165e-06, + "loss": 1.2117, + "step": 5169 + }, + { + "epoch": 0.94, + "learning_rate": 2.7204222788496545e-06, + "loss": 0.9881, + "step": 5170 + }, + { + "epoch": 0.94, + "learning_rate": 2.720367673825992e-06, + "loss": 2.3441, + "step": 5171 + }, + { + "epoch": 0.94, + "learning_rate": 2.72031306880233e-06, + "loss": 1.8898, + "step": 5172 + }, + { + "epoch": 0.94, + "learning_rate": 2.720258463778668e-06, + "loss": 0.729, + "step": 5173 + }, + { + "epoch": 0.94, + "learning_rate": 2.7202038587550054e-06, + "loss": 1.7239, + "step": 5174 + }, + { + "epoch": 0.94, + "learning_rate": 2.7201492537313433e-06, + "loss": 1.7901, + "step": 5175 + }, + { + "epoch": 0.94, + "learning_rate": 2.7200946487076813e-06, + "loss": 0.3711, + "step": 5176 + }, + { + "epoch": 0.94, + "learning_rate": 2.720040043684019e-06, + "loss": 1.536, + "step": 5177 + }, + { + "epoch": 0.94, + "learning_rate": 2.7199854386603567e-06, + "loss": 2.1099, + "step": 5178 + }, + { + "epoch": 0.94, + "learning_rate": 2.7199308336366947e-06, + "loss": 0.4208, + "step": 5179 + }, + { + "epoch": 0.94, + "learning_rate": 2.7198762286130326e-06, + "loss": 1.1423, + "step": 5180 + }, + { + "epoch": 0.94, + "learning_rate": 2.7198216235893706e-06, + "loss": 0.5751, + "step": 5181 + }, + { + "epoch": 0.94, + "learning_rate": 2.719767018565708e-06, + "loss": 1.6608, + "step": 5182 + }, + { + "epoch": 0.94, + "learning_rate": 2.719712413542046e-06, + "loss": 0.7693, + "step": 5183 + }, + { + "epoch": 0.94, + "learning_rate": 2.719657808518384e-06, + "loss": 1.4558, + "step": 5184 + }, + { + "epoch": 0.94, + "learning_rate": 2.7196032034947215e-06, + "loss": 0.8688, + "step": 5185 + }, + { + "epoch": 0.94, + "learning_rate": 2.7195485984710594e-06, + "loss": 0.5261, + "step": 5186 + }, + { + "epoch": 0.94, + "learning_rate": 2.7194939934473974e-06, + "loss": 0.6685, + "step": 5187 + }, + { + "epoch": 0.94, + "learning_rate": 2.719439388423735e-06, + "loss": 0.7419, + "step": 5188 + }, + { + "epoch": 0.94, + "learning_rate": 2.719384783400073e-06, + "loss": 2.4759, + "step": 5189 + }, + { + "epoch": 0.94, + "learning_rate": 2.719330178376411e-06, + "loss": 1.1632, + "step": 5190 + }, + { + "epoch": 0.94, + "learning_rate": 2.7192755733527483e-06, + "loss": 2.2899, + "step": 5191 + }, + { + "epoch": 0.94, + "learning_rate": 2.7192209683290863e-06, + "loss": 0.3902, + "step": 5192 + }, + { + "epoch": 0.94, + "learning_rate": 2.719166363305424e-06, + "loss": 0.8765, + "step": 5193 + }, + { + "epoch": 0.94, + "learning_rate": 2.719111758281762e-06, + "loss": 0.3191, + "step": 5194 + }, + { + "epoch": 0.94, + "learning_rate": 2.7190571532581e-06, + "loss": 1.4793, + "step": 5195 + }, + { + "epoch": 0.94, + "learning_rate": 2.7190025482344376e-06, + "loss": 0.99, + "step": 5196 + }, + { + "epoch": 0.95, + "learning_rate": 2.7189479432107756e-06, + "loss": 0.9277, + "step": 5197 + }, + { + "epoch": 0.95, + "learning_rate": 2.7188933381871135e-06, + "loss": 0.6162, + "step": 5198 + }, + { + "epoch": 0.95, + "learning_rate": 2.718838733163451e-06, + "loss": 1.5882, + "step": 5199 + }, + { + "epoch": 0.95, + "learning_rate": 2.718784128139789e-06, + "loss": 1.1256, + "step": 5200 + }, + { + "epoch": 0.95, + "learning_rate": 2.718729523116127e-06, + "loss": 0.0437, + "step": 5201 + }, + { + "epoch": 0.95, + "learning_rate": 2.7186749180924644e-06, + "loss": 4.1014, + "step": 5202 + }, + { + "epoch": 0.95, + "learning_rate": 2.7186203130688024e-06, + "loss": 1.1126, + "step": 5203 + }, + { + "epoch": 0.95, + "learning_rate": 2.7185657080451403e-06, + "loss": 0.1553, + "step": 5204 + }, + { + "epoch": 0.95, + "learning_rate": 2.718511103021478e-06, + "loss": 0.7463, + "step": 5205 + }, + { + "epoch": 0.95, + "learning_rate": 2.7184564979978158e-06, + "loss": 0.4729, + "step": 5206 + }, + { + "epoch": 0.95, + "learning_rate": 2.7184018929741537e-06, + "loss": 0.4332, + "step": 5207 + }, + { + "epoch": 0.95, + "learning_rate": 2.7183472879504917e-06, + "loss": 0.4821, + "step": 5208 + }, + { + "epoch": 0.95, + "learning_rate": 2.7182926829268296e-06, + "loss": 1.071, + "step": 5209 + }, + { + "epoch": 0.95, + "learning_rate": 2.718238077903167e-06, + "loss": 1.3881, + "step": 5210 + }, + { + "epoch": 0.95, + "learning_rate": 2.718183472879505e-06, + "loss": 1.1241, + "step": 5211 + }, + { + "epoch": 0.95, + "learning_rate": 2.718128867855843e-06, + "loss": 2.9197, + "step": 5212 + }, + { + "epoch": 0.95, + "learning_rate": 2.7180742628321805e-06, + "loss": 0.0321, + "step": 5213 + }, + { + "epoch": 0.95, + "learning_rate": 2.7180196578085185e-06, + "loss": 0.1062, + "step": 5214 + }, + { + "epoch": 0.95, + "learning_rate": 2.7179650527848564e-06, + "loss": 2.1188, + "step": 5215 + }, + { + "epoch": 0.95, + "learning_rate": 2.717910447761194e-06, + "loss": 0.05, + "step": 5216 + }, + { + "epoch": 0.95, + "learning_rate": 2.717855842737532e-06, + "loss": 1.5029, + "step": 5217 + }, + { + "epoch": 0.95, + "learning_rate": 2.71780123771387e-06, + "loss": 0.7212, + "step": 5218 + }, + { + "epoch": 0.95, + "learning_rate": 2.7177466326902073e-06, + "loss": 0.4483, + "step": 5219 + }, + { + "epoch": 0.95, + "learning_rate": 2.7176920276665453e-06, + "loss": 0.0827, + "step": 5220 + }, + { + "epoch": 0.95, + "learning_rate": 2.7176374226428832e-06, + "loss": 2.2887, + "step": 5221 + }, + { + "epoch": 0.95, + "learning_rate": 2.717582817619221e-06, + "loss": 0.5285, + "step": 5222 + }, + { + "epoch": 0.95, + "learning_rate": 2.717528212595559e-06, + "loss": 0.7034, + "step": 5223 + }, + { + "epoch": 0.95, + "learning_rate": 2.7174736075718966e-06, + "loss": 1.7891, + "step": 5224 + }, + { + "epoch": 0.95, + "learning_rate": 2.7174190025482346e-06, + "loss": 1.0482, + "step": 5225 + }, + { + "epoch": 0.95, + "learning_rate": 2.7173643975245725e-06, + "loss": 0.9035, + "step": 5226 + }, + { + "epoch": 0.95, + "learning_rate": 2.71730979250091e-06, + "loss": 0.3201, + "step": 5227 + }, + { + "epoch": 0.95, + "learning_rate": 2.717255187477248e-06, + "loss": 0.1503, + "step": 5228 + }, + { + "epoch": 0.95, + "learning_rate": 2.717200582453586e-06, + "loss": 0.5439, + "step": 5229 + }, + { + "epoch": 0.95, + "learning_rate": 2.7171459774299235e-06, + "loss": 1.6665, + "step": 5230 + }, + { + "epoch": 0.95, + "learning_rate": 2.7170913724062614e-06, + "loss": 1.0312, + "step": 5231 + }, + { + "epoch": 0.95, + "learning_rate": 2.7170367673825993e-06, + "loss": 0.5792, + "step": 5232 + }, + { + "epoch": 0.95, + "learning_rate": 2.716982162358937e-06, + "loss": 0.8157, + "step": 5233 + }, + { + "epoch": 0.95, + "learning_rate": 2.716927557335275e-06, + "loss": 0.8478, + "step": 5234 + }, + { + "epoch": 0.95, + "learning_rate": 2.7168729523116127e-06, + "loss": 0.4787, + "step": 5235 + }, + { + "epoch": 0.95, + "learning_rate": 2.7168183472879507e-06, + "loss": 0.0812, + "step": 5236 + }, + { + "epoch": 0.95, + "learning_rate": 2.7167637422642886e-06, + "loss": 0.9928, + "step": 5237 + }, + { + "epoch": 0.95, + "learning_rate": 2.716709137240626e-06, + "loss": 1.9322, + "step": 5238 + }, + { + "epoch": 0.95, + "learning_rate": 2.716654532216964e-06, + "loss": 1.8923, + "step": 5239 + }, + { + "epoch": 0.95, + "learning_rate": 2.716599927193302e-06, + "loss": 0.1485, + "step": 5240 + }, + { + "epoch": 0.95, + "learning_rate": 2.7165453221696396e-06, + "loss": 0.7482, + "step": 5241 + }, + { + "epoch": 0.95, + "learning_rate": 2.7164907171459775e-06, + "loss": 2.0046, + "step": 5242 + }, + { + "epoch": 0.95, + "learning_rate": 2.7164361121223154e-06, + "loss": 0.585, + "step": 5243 + }, + { + "epoch": 0.95, + "learning_rate": 2.716381507098653e-06, + "loss": 1.1537, + "step": 5244 + }, + { + "epoch": 0.95, + "learning_rate": 2.716326902074991e-06, + "loss": 0.4255, + "step": 5245 + }, + { + "epoch": 0.95, + "learning_rate": 2.716272297051329e-06, + "loss": 1.6395, + "step": 5246 + }, + { + "epoch": 0.95, + "learning_rate": 2.7162176920276664e-06, + "loss": 1.3232, + "step": 5247 + }, + { + "epoch": 0.95, + "learning_rate": 2.7161630870040043e-06, + "loss": 0.6199, + "step": 5248 + }, + { + "epoch": 0.95, + "learning_rate": 2.7161084819803423e-06, + "loss": 0.9756, + "step": 5249 + }, + { + "epoch": 0.95, + "learning_rate": 2.71605387695668e-06, + "loss": 0.2971, + "step": 5250 + }, + { + "epoch": 0.95, + "learning_rate": 2.715999271933018e-06, + "loss": 2.1801, + "step": 5251 + }, + { + "epoch": 0.96, + "learning_rate": 2.7159446669093557e-06, + "loss": 1.2898, + "step": 5252 + }, + { + "epoch": 0.96, + "learning_rate": 2.7158900618856936e-06, + "loss": 0.9626, + "step": 5253 + }, + { + "epoch": 0.96, + "learning_rate": 2.7158354568620316e-06, + "loss": 1.4112, + "step": 5254 + }, + { + "epoch": 0.96, + "learning_rate": 2.715780851838369e-06, + "loss": 1.4128, + "step": 5255 + }, + { + "epoch": 0.96, + "learning_rate": 2.715726246814707e-06, + "loss": 2.0051, + "step": 5256 + }, + { + "epoch": 0.96, + "learning_rate": 2.715671641791045e-06, + "loss": 0.3878, + "step": 5257 + }, + { + "epoch": 0.96, + "learning_rate": 2.7156170367673825e-06, + "loss": 0.1232, + "step": 5258 + }, + { + "epoch": 0.96, + "learning_rate": 2.7155624317437204e-06, + "loss": 0.6879, + "step": 5259 + }, + { + "epoch": 0.96, + "learning_rate": 2.715507826720058e-06, + "loss": 1.4598, + "step": 5260 + }, + { + "epoch": 0.96, + "learning_rate": 2.715453221696396e-06, + "loss": 1.1048, + "step": 5261 + }, + { + "epoch": 0.96, + "learning_rate": 2.715398616672734e-06, + "loss": 1.0481, + "step": 5262 + }, + { + "epoch": 0.96, + "learning_rate": 2.7153440116490718e-06, + "loss": 1.4204, + "step": 5263 + }, + { + "epoch": 0.96, + "learning_rate": 2.7152894066254097e-06, + "loss": 0.3874, + "step": 5264 + }, + { + "epoch": 0.96, + "learning_rate": 2.7152348016017477e-06, + "loss": 0.4452, + "step": 5265 + }, + { + "epoch": 0.96, + "learning_rate": 2.715180196578085e-06, + "loss": 0.4425, + "step": 5266 + }, + { + "epoch": 0.96, + "learning_rate": 2.715125591554423e-06, + "loss": 0.6773, + "step": 5267 + }, + { + "epoch": 0.96, + "learning_rate": 2.715070986530761e-06, + "loss": 1.2846, + "step": 5268 + }, + { + "epoch": 0.96, + "learning_rate": 2.7150163815070986e-06, + "loss": 0.6121, + "step": 5269 + }, + { + "epoch": 0.96, + "learning_rate": 2.7149617764834365e-06, + "loss": 0.3978, + "step": 5270 + }, + { + "epoch": 0.96, + "learning_rate": 2.7149071714597745e-06, + "loss": 0.2558, + "step": 5271 + }, + { + "epoch": 0.96, + "learning_rate": 2.714852566436112e-06, + "loss": 0.4889, + "step": 5272 + }, + { + "epoch": 0.96, + "learning_rate": 2.71479796141245e-06, + "loss": 2.0194, + "step": 5273 + }, + { + "epoch": 0.96, + "learning_rate": 2.714743356388788e-06, + "loss": 1.2191, + "step": 5274 + }, + { + "epoch": 0.96, + "learning_rate": 2.714688751365126e-06, + "loss": 4.0731, + "step": 5275 + }, + { + "epoch": 0.96, + "learning_rate": 2.7146341463414638e-06, + "loss": 0.1262, + "step": 5276 + }, + { + "epoch": 0.96, + "learning_rate": 2.7145795413178013e-06, + "loss": 0.4409, + "step": 5277 + }, + { + "epoch": 0.96, + "learning_rate": 2.7145249362941392e-06, + "loss": 0.0894, + "step": 5278 + }, + { + "epoch": 0.96, + "learning_rate": 2.714470331270477e-06, + "loss": 2.1925, + "step": 5279 + }, + { + "epoch": 0.96, + "learning_rate": 2.7144157262468147e-06, + "loss": 0.0358, + "step": 5280 + }, + { + "epoch": 0.96, + "learning_rate": 2.7143611212231526e-06, + "loss": 0.9585, + "step": 5281 + }, + { + "epoch": 0.96, + "learning_rate": 2.7143065161994906e-06, + "loss": 0.0603, + "step": 5282 + }, + { + "epoch": 0.96, + "learning_rate": 2.714251911175828e-06, + "loss": 3.724, + "step": 5283 + }, + { + "epoch": 0.96, + "learning_rate": 2.714197306152166e-06, + "loss": 0.4009, + "step": 5284 + }, + { + "epoch": 0.96, + "learning_rate": 2.714142701128504e-06, + "loss": 0.0692, + "step": 5285 + }, + { + "epoch": 0.96, + "learning_rate": 2.7140880961048415e-06, + "loss": 1.0374, + "step": 5286 + }, + { + "epoch": 0.96, + "learning_rate": 2.7140334910811794e-06, + "loss": 1.1752, + "step": 5287 + }, + { + "epoch": 0.96, + "learning_rate": 2.7139788860575174e-06, + "loss": 1.4025, + "step": 5288 + }, + { + "epoch": 0.96, + "learning_rate": 2.7139242810338553e-06, + "loss": 1.15, + "step": 5289 + }, + { + "epoch": 0.96, + "learning_rate": 2.7138696760101933e-06, + "loss": 1.2712, + "step": 5290 + }, + { + "epoch": 0.96, + "learning_rate": 2.713815070986531e-06, + "loss": 1.5132, + "step": 5291 + }, + { + "epoch": 0.96, + "learning_rate": 2.7137604659628687e-06, + "loss": 1.6515, + "step": 5292 + }, + { + "epoch": 0.96, + "learning_rate": 2.7137058609392067e-06, + "loss": 0.0662, + "step": 5293 + }, + { + "epoch": 0.96, + "learning_rate": 2.713651255915544e-06, + "loss": 0.2216, + "step": 5294 + }, + { + "epoch": 0.96, + "learning_rate": 2.713596650891882e-06, + "loss": 2.7026, + "step": 5295 + }, + { + "epoch": 0.96, + "learning_rate": 2.71354204586822e-06, + "loss": 1.9455, + "step": 5296 + }, + { + "epoch": 0.96, + "learning_rate": 2.7134874408445576e-06, + "loss": 1.5415, + "step": 5297 + }, + { + "epoch": 0.96, + "learning_rate": 2.7134328358208956e-06, + "loss": 0.5205, + "step": 5298 + }, + { + "epoch": 0.96, + "learning_rate": 2.7133782307972335e-06, + "loss": 1.3646, + "step": 5299 + }, + { + "epoch": 0.96, + "learning_rate": 2.713323625773571e-06, + "loss": 0.9858, + "step": 5300 + }, + { + "epoch": 0.96, + "learning_rate": 2.713269020749909e-06, + "loss": 1.0433, + "step": 5301 + }, + { + "epoch": 0.96, + "learning_rate": 2.713214415726247e-06, + "loss": 1.3637, + "step": 5302 + }, + { + "epoch": 0.96, + "learning_rate": 2.713159810702585e-06, + "loss": 1.4462, + "step": 5303 + }, + { + "epoch": 0.96, + "learning_rate": 2.713105205678923e-06, + "loss": 2.1202, + "step": 5304 + }, + { + "epoch": 0.96, + "learning_rate": 2.7130506006552603e-06, + "loss": 0.5466, + "step": 5305 + }, + { + "epoch": 0.96, + "learning_rate": 2.7129959956315983e-06, + "loss": 1.436, + "step": 5306 + }, + { + "epoch": 0.97, + "learning_rate": 2.712941390607936e-06, + "loss": 1.681, + "step": 5307 + }, + { + "epoch": 0.97, + "learning_rate": 2.7128867855842737e-06, + "loss": 0.2672, + "step": 5308 + }, + { + "epoch": 0.97, + "learning_rate": 2.7128321805606117e-06, + "loss": 0.0829, + "step": 5309 + }, + { + "epoch": 0.97, + "learning_rate": 2.7127775755369496e-06, + "loss": 0.656, + "step": 5310 + }, + { + "epoch": 0.97, + "learning_rate": 2.712722970513287e-06, + "loss": 1.2524, + "step": 5311 + }, + { + "epoch": 0.97, + "learning_rate": 2.712668365489625e-06, + "loss": 0.3235, + "step": 5312 + }, + { + "epoch": 0.97, + "learning_rate": 2.712613760465963e-06, + "loss": 1.5753, + "step": 5313 + }, + { + "epoch": 0.97, + "learning_rate": 2.7125591554423005e-06, + "loss": 0.717, + "step": 5314 + }, + { + "epoch": 0.97, + "learning_rate": 2.7125045504186385e-06, + "loss": 2.7721, + "step": 5315 + }, + { + "epoch": 0.97, + "learning_rate": 2.7124499453949764e-06, + "loss": 0.6825, + "step": 5316 + }, + { + "epoch": 0.97, + "learning_rate": 2.7123953403713144e-06, + "loss": 0.6528, + "step": 5317 + }, + { + "epoch": 0.97, + "learning_rate": 2.7123407353476523e-06, + "loss": 2.7413, + "step": 5318 + }, + { + "epoch": 0.97, + "learning_rate": 2.71228613032399e-06, + "loss": 0.1316, + "step": 5319 + }, + { + "epoch": 0.97, + "learning_rate": 2.7122315253003278e-06, + "loss": 1.4259, + "step": 5320 + }, + { + "epoch": 0.97, + "learning_rate": 2.7121769202766657e-06, + "loss": 3.5632, + "step": 5321 + }, + { + "epoch": 0.97, + "learning_rate": 2.7121223152530032e-06, + "loss": 0.765, + "step": 5322 + }, + { + "epoch": 0.97, + "learning_rate": 2.712067710229341e-06, + "loss": 1.2889, + "step": 5323 + }, + { + "epoch": 0.97, + "learning_rate": 2.712013105205679e-06, + "loss": 0.9408, + "step": 5324 + }, + { + "epoch": 0.97, + "learning_rate": 2.7119585001820166e-06, + "loss": 1.6518, + "step": 5325 + }, + { + "epoch": 0.97, + "learning_rate": 2.7119038951583546e-06, + "loss": 0.514, + "step": 5326 + }, + { + "epoch": 0.97, + "learning_rate": 2.711849290134692e-06, + "loss": 0.8358, + "step": 5327 + }, + { + "epoch": 0.97, + "learning_rate": 2.71179468511103e-06, + "loss": 0.8473, + "step": 5328 + }, + { + "epoch": 0.97, + "learning_rate": 2.711740080087368e-06, + "loss": 0.501, + "step": 5329 + }, + { + "epoch": 0.97, + "learning_rate": 2.711685475063706e-06, + "loss": 0.6314, + "step": 5330 + }, + { + "epoch": 0.97, + "learning_rate": 2.711630870040044e-06, + "loss": 0.4157, + "step": 5331 + }, + { + "epoch": 0.97, + "learning_rate": 2.711576265016382e-06, + "loss": 1.5236, + "step": 5332 + }, + { + "epoch": 0.97, + "learning_rate": 2.7115216599927193e-06, + "loss": 1.3857, + "step": 5333 + }, + { + "epoch": 0.97, + "learning_rate": 2.7114670549690573e-06, + "loss": 1.0028, + "step": 5334 + }, + { + "epoch": 0.97, + "learning_rate": 2.7114124499453952e-06, + "loss": 0.4834, + "step": 5335 + }, + { + "epoch": 0.97, + "learning_rate": 2.7113578449217327e-06, + "loss": 1.082, + "step": 5336 + }, + { + "epoch": 0.97, + "learning_rate": 2.7113032398980707e-06, + "loss": 0.582, + "step": 5337 + }, + { + "epoch": 0.97, + "learning_rate": 2.7112486348744086e-06, + "loss": 0.0945, + "step": 5338 + }, + { + "epoch": 0.97, + "learning_rate": 2.711194029850746e-06, + "loss": 3.0577, + "step": 5339 + }, + { + "epoch": 0.97, + "learning_rate": 2.711139424827084e-06, + "loss": 0.0576, + "step": 5340 + }, + { + "epoch": 0.97, + "learning_rate": 2.7110848198034216e-06, + "loss": 0.2445, + "step": 5341 + }, + { + "epoch": 0.97, + "learning_rate": 2.7110302147797596e-06, + "loss": 1.51, + "step": 5342 + }, + { + "epoch": 0.97, + "learning_rate": 2.7109756097560975e-06, + "loss": 0.4696, + "step": 5343 + }, + { + "epoch": 0.97, + "learning_rate": 2.7109210047324354e-06, + "loss": 2.328, + "step": 5344 + }, + { + "epoch": 0.97, + "learning_rate": 2.7108663997087734e-06, + "loss": 2.0235, + "step": 5345 + }, + { + "epoch": 0.97, + "learning_rate": 2.7108117946851113e-06, + "loss": 0.9427, + "step": 5346 + }, + { + "epoch": 0.97, + "learning_rate": 2.710757189661449e-06, + "loss": 0.4608, + "step": 5347 + }, + { + "epoch": 0.97, + "learning_rate": 2.710702584637787e-06, + "loss": 1.4279, + "step": 5348 + }, + { + "epoch": 0.97, + "learning_rate": 2.7106479796141247e-06, + "loss": 0.6341, + "step": 5349 + }, + { + "epoch": 0.97, + "learning_rate": 2.7105933745904623e-06, + "loss": 0.8639, + "step": 5350 + }, + { + "epoch": 0.97, + "learning_rate": 2.7105387695668e-06, + "loss": 0.6736, + "step": 5351 + }, + { + "epoch": 0.97, + "learning_rate": 2.710484164543138e-06, + "loss": 0.7926, + "step": 5352 + }, + { + "epoch": 0.97, + "learning_rate": 2.7104295595194757e-06, + "loss": 2.2659, + "step": 5353 + }, + { + "epoch": 0.97, + "learning_rate": 2.7103749544958136e-06, + "loss": 1.4184, + "step": 5354 + }, + { + "epoch": 0.97, + "learning_rate": 2.7103203494721516e-06, + "loss": 0.7989, + "step": 5355 + }, + { + "epoch": 0.97, + "learning_rate": 2.7102657444484895e-06, + "loss": 1.0043, + "step": 5356 + }, + { + "epoch": 0.97, + "learning_rate": 2.7102111394248274e-06, + "loss": 0.4296, + "step": 5357 + }, + { + "epoch": 0.97, + "learning_rate": 2.710156534401165e-06, + "loss": 0.2431, + "step": 5358 + }, + { + "epoch": 0.97, + "learning_rate": 2.710101929377503e-06, + "loss": 1.4641, + "step": 5359 + }, + { + "epoch": 0.97, + "learning_rate": 2.710047324353841e-06, + "loss": 0.0917, + "step": 5360 + }, + { + "epoch": 0.97, + "learning_rate": 2.7099927193301784e-06, + "loss": 3.3562, + "step": 5361 + }, + { + "epoch": 0.98, + "learning_rate": 2.7099381143065163e-06, + "loss": 3.1434, + "step": 5362 + }, + { + "epoch": 0.98, + "learning_rate": 2.7098835092828543e-06, + "loss": 0.1111, + "step": 5363 + }, + { + "epoch": 0.98, + "learning_rate": 2.7098289042591918e-06, + "loss": 0.7567, + "step": 5364 + }, + { + "epoch": 0.98, + "learning_rate": 2.7097742992355297e-06, + "loss": 0.571, + "step": 5365 + }, + { + "epoch": 0.98, + "learning_rate": 2.7097196942118677e-06, + "loss": 0.5081, + "step": 5366 + }, + { + "epoch": 0.98, + "learning_rate": 2.709665089188205e-06, + "loss": 0.5317, + "step": 5367 + }, + { + "epoch": 0.98, + "learning_rate": 2.709610484164543e-06, + "loss": 0.599, + "step": 5368 + }, + { + "epoch": 0.98, + "learning_rate": 2.709555879140881e-06, + "loss": 1.161, + "step": 5369 + }, + { + "epoch": 0.98, + "learning_rate": 2.709501274117219e-06, + "loss": 1.3872, + "step": 5370 + }, + { + "epoch": 0.98, + "learning_rate": 2.709446669093557e-06, + "loss": 0.9104, + "step": 5371 + }, + { + "epoch": 0.98, + "learning_rate": 2.7093920640698945e-06, + "loss": 0.6129, + "step": 5372 + }, + { + "epoch": 0.98, + "learning_rate": 2.7093374590462324e-06, + "loss": 1.2037, + "step": 5373 + }, + { + "epoch": 0.98, + "learning_rate": 2.7092828540225704e-06, + "loss": 0.0861, + "step": 5374 + }, + { + "epoch": 0.98, + "learning_rate": 2.709228248998908e-06, + "loss": 0.4428, + "step": 5375 + }, + { + "epoch": 0.98, + "learning_rate": 2.709173643975246e-06, + "loss": 0.5166, + "step": 5376 + }, + { + "epoch": 0.98, + "learning_rate": 2.7091190389515838e-06, + "loss": 0.6311, + "step": 5377 + }, + { + "epoch": 0.98, + "learning_rate": 2.7090644339279213e-06, + "loss": 2.492, + "step": 5378 + }, + { + "epoch": 0.98, + "learning_rate": 2.7090098289042592e-06, + "loss": 0.0291, + "step": 5379 + }, + { + "epoch": 0.98, + "learning_rate": 2.708955223880597e-06, + "loss": 0.0437, + "step": 5380 + }, + { + "epoch": 0.98, + "learning_rate": 2.7089006188569347e-06, + "loss": 2.6539, + "step": 5381 + }, + { + "epoch": 0.98, + "learning_rate": 2.7088460138332726e-06, + "loss": 2.3733, + "step": 5382 + }, + { + "epoch": 0.98, + "learning_rate": 2.7087914088096106e-06, + "loss": 0.2341, + "step": 5383 + }, + { + "epoch": 0.98, + "learning_rate": 2.7087368037859485e-06, + "loss": 0.5884, + "step": 5384 + }, + { + "epoch": 0.98, + "learning_rate": 2.7086821987622865e-06, + "loss": 2.2765, + "step": 5385 + }, + { + "epoch": 0.98, + "learning_rate": 2.708627593738624e-06, + "loss": 0.6517, + "step": 5386 + }, + { + "epoch": 0.98, + "learning_rate": 2.708572988714962e-06, + "loss": 2.0766, + "step": 5387 + }, + { + "epoch": 0.98, + "learning_rate": 2.7085183836913e-06, + "loss": 1.1644, + "step": 5388 + }, + { + "epoch": 0.98, + "learning_rate": 2.7084637786676374e-06, + "loss": 0.4632, + "step": 5389 + }, + { + "epoch": 0.98, + "learning_rate": 2.7084091736439753e-06, + "loss": 0.5717, + "step": 5390 + }, + { + "epoch": 0.98, + "learning_rate": 2.7083545686203133e-06, + "loss": 1.0543, + "step": 5391 + }, + { + "epoch": 0.98, + "learning_rate": 2.708299963596651e-06, + "loss": 0.395, + "step": 5392 + }, + { + "epoch": 0.98, + "learning_rate": 2.7082453585729887e-06, + "loss": 2.3456, + "step": 5393 + }, + { + "epoch": 0.98, + "learning_rate": 2.7081907535493267e-06, + "loss": 0.4565, + "step": 5394 + }, + { + "epoch": 0.98, + "learning_rate": 2.708136148525664e-06, + "loss": 0.1332, + "step": 5395 + }, + { + "epoch": 0.98, + "learning_rate": 2.708081543502002e-06, + "loss": 1.5358, + "step": 5396 + }, + { + "epoch": 0.98, + "learning_rate": 2.70802693847834e-06, + "loss": 0.3141, + "step": 5397 + }, + { + "epoch": 0.98, + "learning_rate": 2.707972333454678e-06, + "loss": 1.5002, + "step": 5398 + }, + { + "epoch": 0.98, + "learning_rate": 2.707917728431016e-06, + "loss": 0.3722, + "step": 5399 + }, + { + "epoch": 0.98, + "learning_rate": 2.7078631234073535e-06, + "loss": 0.6908, + "step": 5400 + }, + { + "epoch": 0.98, + "learning_rate": 2.7078085183836914e-06, + "loss": 0.0724, + "step": 5401 + }, + { + "epoch": 0.98, + "learning_rate": 2.7077539133600294e-06, + "loss": 0.7408, + "step": 5402 + }, + { + "epoch": 0.98, + "learning_rate": 2.707699308336367e-06, + "loss": 0.2373, + "step": 5403 + }, + { + "epoch": 0.98, + "learning_rate": 2.707644703312705e-06, + "loss": 0.4497, + "step": 5404 + }, + { + "epoch": 0.98, + "learning_rate": 2.707590098289043e-06, + "loss": 0.9494, + "step": 5405 + }, + { + "epoch": 0.98, + "learning_rate": 2.7075354932653803e-06, + "loss": 1.0656, + "step": 5406 + }, + { + "epoch": 0.98, + "learning_rate": 2.7074808882417183e-06, + "loss": 1.5226, + "step": 5407 + }, + { + "epoch": 0.98, + "learning_rate": 2.7074262832180558e-06, + "loss": 0.7201, + "step": 5408 + }, + { + "epoch": 0.98, + "learning_rate": 2.7073716781943937e-06, + "loss": 0.9334, + "step": 5409 + }, + { + "epoch": 0.98, + "learning_rate": 2.7073170731707317e-06, + "loss": 1.0554, + "step": 5410 + }, + { + "epoch": 0.98, + "learning_rate": 2.7072624681470696e-06, + "loss": 0.2011, + "step": 5411 + }, + { + "epoch": 0.98, + "learning_rate": 2.7072078631234075e-06, + "loss": 0.0501, + "step": 5412 + }, + { + "epoch": 0.98, + "learning_rate": 2.7071532580997455e-06, + "loss": 0.4499, + "step": 5413 + }, + { + "epoch": 0.98, + "learning_rate": 2.707098653076083e-06, + "loss": 0.5437, + "step": 5414 + }, + { + "epoch": 0.98, + "learning_rate": 2.707044048052421e-06, + "loss": 0.3221, + "step": 5415 + }, + { + "epoch": 0.98, + "learning_rate": 2.706989443028759e-06, + "loss": 2.2223, + "step": 5416 + }, + { + "epoch": 0.99, + "learning_rate": 2.7069348380050964e-06, + "loss": 0.5478, + "step": 5417 + }, + { + "epoch": 0.99, + "learning_rate": 2.7068802329814344e-06, + "loss": 1.7756, + "step": 5418 + }, + { + "epoch": 0.99, + "learning_rate": 2.7068256279577723e-06, + "loss": 1.307, + "step": 5419 + }, + { + "epoch": 0.99, + "learning_rate": 2.70677102293411e-06, + "loss": 0.2486, + "step": 5420 + }, + { + "epoch": 0.99, + "learning_rate": 2.7067164179104478e-06, + "loss": 1.3149, + "step": 5421 + }, + { + "epoch": 0.99, + "learning_rate": 2.7066618128867853e-06, + "loss": 0.2032, + "step": 5422 + }, + { + "epoch": 0.99, + "learning_rate": 2.7066072078631232e-06, + "loss": 0.3655, + "step": 5423 + }, + { + "epoch": 0.99, + "learning_rate": 2.706552602839461e-06, + "loss": 0.7065, + "step": 5424 + }, + { + "epoch": 0.99, + "learning_rate": 2.706497997815799e-06, + "loss": 0.2737, + "step": 5425 + }, + { + "epoch": 0.99, + "learning_rate": 2.706443392792137e-06, + "loss": 3.2087, + "step": 5426 + }, + { + "epoch": 0.99, + "learning_rate": 2.706388787768475e-06, + "loss": 0.6157, + "step": 5427 + }, + { + "epoch": 0.99, + "learning_rate": 2.7063341827448125e-06, + "loss": 0.4631, + "step": 5428 + }, + { + "epoch": 0.99, + "learning_rate": 2.7062795777211505e-06, + "loss": 0.5082, + "step": 5429 + }, + { + "epoch": 0.99, + "learning_rate": 2.7062249726974884e-06, + "loss": 0.9551, + "step": 5430 + }, + { + "epoch": 0.99, + "learning_rate": 2.706170367673826e-06, + "loss": 0.4137, + "step": 5431 + }, + { + "epoch": 0.99, + "learning_rate": 2.706115762650164e-06, + "loss": 0.4333, + "step": 5432 + }, + { + "epoch": 0.99, + "learning_rate": 2.706061157626502e-06, + "loss": 0.9198, + "step": 5433 + }, + { + "epoch": 0.99, + "learning_rate": 2.7060065526028393e-06, + "loss": 0.2682, + "step": 5434 + }, + { + "epoch": 0.99, + "learning_rate": 2.7059519475791773e-06, + "loss": 1.3089, + "step": 5435 + }, + { + "epoch": 0.99, + "learning_rate": 2.7058973425555152e-06, + "loss": 0.3259, + "step": 5436 + }, + { + "epoch": 0.99, + "learning_rate": 2.705842737531853e-06, + "loss": 0.6917, + "step": 5437 + }, + { + "epoch": 0.99, + "learning_rate": 2.705788132508191e-06, + "loss": 0.6296, + "step": 5438 + }, + { + "epoch": 0.99, + "learning_rate": 2.7057335274845286e-06, + "loss": 1.3806, + "step": 5439 + }, + { + "epoch": 0.99, + "learning_rate": 2.7056789224608666e-06, + "loss": 1.2993, + "step": 5440 + }, + { + "epoch": 0.99, + "learning_rate": 2.7056243174372045e-06, + "loss": 1.3349, + "step": 5441 + }, + { + "epoch": 0.99, + "learning_rate": 2.705569712413542e-06, + "loss": 1.6741, + "step": 5442 + }, + { + "epoch": 0.99, + "learning_rate": 2.70551510738988e-06, + "loss": 1.0221, + "step": 5443 + }, + { + "epoch": 0.99, + "learning_rate": 2.705460502366218e-06, + "loss": 1.11, + "step": 5444 + }, + { + "epoch": 0.99, + "learning_rate": 2.7054058973425554e-06, + "loss": 0.5083, + "step": 5445 + }, + { + "epoch": 0.99, + "learning_rate": 2.7053512923188934e-06, + "loss": 1.4472, + "step": 5446 + }, + { + "epoch": 0.99, + "learning_rate": 2.7052966872952313e-06, + "loss": 0.6472, + "step": 5447 + }, + { + "epoch": 0.99, + "learning_rate": 2.705242082271569e-06, + "loss": 1.5237, + "step": 5448 + }, + { + "epoch": 0.99, + "learning_rate": 2.705187477247907e-06, + "loss": 2.8269, + "step": 5449 + }, + { + "epoch": 0.99, + "learning_rate": 2.7051328722242447e-06, + "loss": 2.1803, + "step": 5450 + }, + { + "epoch": 0.99, + "learning_rate": 2.7050782672005827e-06, + "loss": 0.2866, + "step": 5451 + }, + { + "epoch": 0.99, + "learning_rate": 2.7050236621769206e-06, + "loss": 2.1845, + "step": 5452 + }, + { + "epoch": 0.99, + "learning_rate": 2.704969057153258e-06, + "loss": 0.6848, + "step": 5453 + }, + { + "epoch": 0.99, + "learning_rate": 2.704914452129596e-06, + "loss": 0.4739, + "step": 5454 + }, + { + "epoch": 0.99, + "learning_rate": 2.704859847105934e-06, + "loss": 1.0416, + "step": 5455 + }, + { + "epoch": 0.99, + "learning_rate": 2.7048052420822716e-06, + "loss": 2.5257, + "step": 5456 + }, + { + "epoch": 0.99, + "learning_rate": 2.7047506370586095e-06, + "loss": 0.7659, + "step": 5457 + }, + { + "epoch": 0.99, + "learning_rate": 2.7046960320349474e-06, + "loss": 1.9718, + "step": 5458 + }, + { + "epoch": 0.99, + "learning_rate": 2.704641427011285e-06, + "loss": 2.4782, + "step": 5459 + }, + { + "epoch": 0.99, + "learning_rate": 2.704586821987623e-06, + "loss": 1.9683, + "step": 5460 + }, + { + "epoch": 0.99, + "learning_rate": 2.704532216963961e-06, + "loss": 2.2164, + "step": 5461 + }, + { + "epoch": 0.99, + "learning_rate": 2.7044776119402984e-06, + "loss": 1.8285, + "step": 5462 + }, + { + "epoch": 0.99, + "learning_rate": 2.7044230069166363e-06, + "loss": 0.2352, + "step": 5463 + }, + { + "epoch": 0.99, + "learning_rate": 2.7043684018929743e-06, + "loss": 0.9042, + "step": 5464 + }, + { + "epoch": 0.99, + "learning_rate": 2.704313796869312e-06, + "loss": 1.0092, + "step": 5465 + }, + { + "epoch": 0.99, + "learning_rate": 2.70425919184565e-06, + "loss": 1.26, + "step": 5466 + }, + { + "epoch": 0.99, + "learning_rate": 2.7042045868219877e-06, + "loss": 2.0064, + "step": 5467 + }, + { + "epoch": 0.99, + "learning_rate": 2.7041499817983256e-06, + "loss": 1.2511, + "step": 5468 + }, + { + "epoch": 0.99, + "learning_rate": 2.7040953767746635e-06, + "loss": 1.1384, + "step": 5469 + }, + { + "epoch": 0.99, + "learning_rate": 2.704040771751001e-06, + "loss": 0.0461, + "step": 5470 + }, + { + "epoch": 0.99, + "learning_rate": 2.703986166727339e-06, + "loss": 1.9841, + "step": 5471 + }, + { + "epoch": 1.0, + "learning_rate": 2.703931561703677e-06, + "loss": 1.1338, + "step": 5472 + }, + { + "epoch": 1.0, + "learning_rate": 2.7038769566800145e-06, + "loss": 0.7464, + "step": 5473 + }, + { + "epoch": 1.0, + "learning_rate": 2.7038223516563524e-06, + "loss": 2.1266, + "step": 5474 + }, + { + "epoch": 1.0, + "learning_rate": 2.7037677466326904e-06, + "loss": 1.2722, + "step": 5475 + }, + { + "epoch": 1.0, + "learning_rate": 2.703713141609028e-06, + "loss": 0.3608, + "step": 5476 + }, + { + "epoch": 1.0, + "learning_rate": 2.703658536585366e-06, + "loss": 0.0933, + "step": 5477 + }, + { + "epoch": 1.0, + "learning_rate": 2.7036039315617038e-06, + "loss": 0.1979, + "step": 5478 + }, + { + "epoch": 1.0, + "learning_rate": 2.7035493265380417e-06, + "loss": 0.836, + "step": 5479 + }, + { + "epoch": 1.0, + "learning_rate": 2.7034947215143797e-06, + "loss": 1.5889, + "step": 5480 + }, + { + "epoch": 1.0, + "learning_rate": 2.703440116490717e-06, + "loss": 0.667, + "step": 5481 + }, + { + "epoch": 1.0, + "learning_rate": 2.703385511467055e-06, + "loss": 0.4544, + "step": 5482 + }, + { + "epoch": 1.0, + "learning_rate": 2.703330906443393e-06, + "loss": 0.0364, + "step": 5483 + }, + { + "epoch": 1.0, + "learning_rate": 2.7032763014197306e-06, + "loss": 0.1621, + "step": 5484 + }, + { + "epoch": 1.0, + "learning_rate": 2.7032216963960685e-06, + "loss": 2.9108, + "step": 5485 + }, + { + "epoch": 1.0, + "learning_rate": 2.7031670913724065e-06, + "loss": 0.1066, + "step": 5486 + }, + { + "epoch": 1.0, + "learning_rate": 2.703112486348744e-06, + "loss": 3.2892, + "step": 5487 + }, + { + "epoch": 1.0, + "learning_rate": 2.703057881325082e-06, + "loss": 1.4026, + "step": 5488 + }, + { + "epoch": 1.0, + "learning_rate": 2.7030032763014194e-06, + "loss": 0.9754, + "step": 5489 + }, + { + "epoch": 1.0, + "learning_rate": 2.7029486712777574e-06, + "loss": 3.554, + "step": 5490 + }, + { + "epoch": 1.0, + "learning_rate": 2.7028940662540953e-06, + "loss": 0.4115, + "step": 5491 + }, + { + "epoch": 1.0, + "learning_rate": 2.7028394612304333e-06, + "loss": 0.5012, + "step": 5492 + }, + { + "epoch": 1.0, + "learning_rate": 2.7027848562067712e-06, + "loss": 0.252, + "step": 5493 + }, + { + "epoch": 1.0, + "learning_rate": 2.702730251183109e-06, + "loss": 2.7843, + "step": 5494 + }, + { + "epoch": 1.0, + "learning_rate": 2.7026756461594467e-06, + "loss": 0.7842, + "step": 5495 + }, + { + "epoch": 1.0, + "learning_rate": 2.7026210411357846e-06, + "loss": 0.229, + "step": 5496 + }, + { + "epoch": 1.0, + "learning_rate": 2.7025664361121226e-06, + "loss": 2.4133, + "step": 5497 + }, + { + "epoch": 1.0, + "learning_rate": 2.70251183108846e-06, + "loss": 1.1649, + "step": 5498 + }, + { + "epoch": 1.0, + "learning_rate": 2.702457226064798e-06, + "loss": 1.8254, + "step": 5499 + }, + { + "epoch": 1.0, + "eval_accuracy": 0.7557283142389526, + "eval_loss": 0.969521701335907, + "eval_runtime": 135.542, + "eval_samples_per_second": 18.031, + "eval_steps_per_second": 4.508, + "step": 5499 + }, + { + "epoch": 1.0, + "learning_rate": 2.702402621041136e-06, + "loss": 1.1157, + "step": 5500 + }, + { + "epoch": 1.0, + "learning_rate": 2.7023480160174735e-06, + "loss": 0.2808, + "step": 5501 + }, + { + "epoch": 1.0, + "learning_rate": 2.7022934109938114e-06, + "loss": 0.7003, + "step": 5502 + }, + { + "epoch": 1.0, + "learning_rate": 2.702238805970149e-06, + "loss": 0.7606, + "step": 5503 + }, + { + "epoch": 1.0, + "learning_rate": 2.702184200946487e-06, + "loss": 1.263, + "step": 5504 + }, + { + "epoch": 1.0, + "learning_rate": 2.702129595922825e-06, + "loss": 0.6325, + "step": 5505 + }, + { + "epoch": 1.0, + "learning_rate": 2.702074990899163e-06, + "loss": 2.6072, + "step": 5506 + }, + { + "epoch": 1.0, + "learning_rate": 2.7020203858755007e-06, + "loss": 0.5554, + "step": 5507 + }, + { + "epoch": 1.0, + "learning_rate": 2.7019657808518387e-06, + "loss": 2.8088, + "step": 5508 + }, + { + "epoch": 1.0, + "learning_rate": 2.701911175828176e-06, + "loss": 1.9115, + "step": 5509 + }, + { + "epoch": 1.0, + "learning_rate": 2.701856570804514e-06, + "loss": 1.0147, + "step": 5510 + }, + { + "epoch": 1.0, + "learning_rate": 2.701801965780852e-06, + "loss": 1.4866, + "step": 5511 + }, + { + "epoch": 1.0, + "learning_rate": 2.7017473607571896e-06, + "loss": 0.072, + "step": 5512 + }, + { + "epoch": 1.0, + "learning_rate": 2.7016927557335276e-06, + "loss": 0.0303, + "step": 5513 + }, + { + "epoch": 1.0, + "learning_rate": 2.7016381507098655e-06, + "loss": 1.8426, + "step": 5514 + }, + { + "epoch": 1.0, + "learning_rate": 2.701583545686203e-06, + "loss": 0.4026, + "step": 5515 + }, + { + "epoch": 1.0, + "learning_rate": 2.701528940662541e-06, + "loss": 0.0573, + "step": 5516 + }, + { + "epoch": 1.0, + "learning_rate": 2.701474335638879e-06, + "loss": 0.9338, + "step": 5517 + }, + { + "epoch": 1.0, + "learning_rate": 2.701419730615217e-06, + "loss": 0.6123, + "step": 5518 + }, + { + "epoch": 1.0, + "learning_rate": 2.7013651255915544e-06, + "loss": 0.3401, + "step": 5519 + }, + { + "epoch": 1.0, + "learning_rate": 2.7013105205678923e-06, + "loss": 0.0716, + "step": 5520 + }, + { + "epoch": 1.0, + "learning_rate": 2.7012559155442303e-06, + "loss": 0.3414, + "step": 5521 + }, + { + "epoch": 1.0, + "learning_rate": 2.701201310520568e-06, + "loss": 2.5675, + "step": 5522 + }, + { + "epoch": 1.0, + "learning_rate": 2.7011467054969057e-06, + "loss": 0.3401, + "step": 5523 + }, + { + "epoch": 1.0, + "learning_rate": 2.7010921004732437e-06, + "loss": 0.4541, + "step": 5524 + }, + { + "epoch": 1.0, + "learning_rate": 2.7010374954495816e-06, + "loss": 1.502, + "step": 5525 + }, + { + "epoch": 1.0, + "learning_rate": 2.700982890425919e-06, + "loss": 1.0788, + "step": 5526 + }, + { + "epoch": 1.01, + "learning_rate": 2.700928285402257e-06, + "loss": 1.0645, + "step": 5527 + }, + { + "epoch": 1.01, + "learning_rate": 2.700873680378595e-06, + "loss": 0.7839, + "step": 5528 + }, + { + "epoch": 1.01, + "learning_rate": 2.7008190753549325e-06, + "loss": 1.4465, + "step": 5529 + }, + { + "epoch": 1.01, + "learning_rate": 2.7007644703312705e-06, + "loss": 1.1557, + "step": 5530 + }, + { + "epoch": 1.01, + "learning_rate": 2.7007098653076084e-06, + "loss": 1.2354, + "step": 5531 + }, + { + "epoch": 1.01, + "learning_rate": 2.7006552602839464e-06, + "loss": 0.2284, + "step": 5532 + }, + { + "epoch": 1.01, + "learning_rate": 2.7006006552602843e-06, + "loss": 0.0454, + "step": 5533 + }, + { + "epoch": 1.01, + "learning_rate": 2.700546050236622e-06, + "loss": 2.4706, + "step": 5534 + }, + { + "epoch": 1.01, + "learning_rate": 2.7004914452129598e-06, + "loss": 0.8464, + "step": 5535 + }, + { + "epoch": 1.01, + "learning_rate": 2.7004368401892977e-06, + "loss": 0.5179, + "step": 5536 + }, + { + "epoch": 1.01, + "learning_rate": 2.7003822351656352e-06, + "loss": 0.5453, + "step": 5537 + }, + { + "epoch": 1.01, + "learning_rate": 2.700327630141973e-06, + "loss": 1.3831, + "step": 5538 + }, + { + "epoch": 1.01, + "learning_rate": 2.700273025118311e-06, + "loss": 0.9718, + "step": 5539 + }, + { + "epoch": 1.01, + "learning_rate": 2.7002184200946486e-06, + "loss": 0.4424, + "step": 5540 + }, + { + "epoch": 1.01, + "learning_rate": 2.7001638150709866e-06, + "loss": 0.1937, + "step": 5541 + }, + { + "epoch": 1.01, + "learning_rate": 2.7001092100473245e-06, + "loss": 0.1824, + "step": 5542 + }, + { + "epoch": 1.01, + "learning_rate": 2.700054605023662e-06, + "loss": 2.6774, + "step": 5543 + }, + { + "epoch": 1.01, + "learning_rate": 2.7e-06, + "loss": 1.3331, + "step": 5544 + }, + { + "epoch": 1.01, + "learning_rate": 2.699945394976338e-06, + "loss": 1.1466, + "step": 5545 + }, + { + "epoch": 1.01, + "learning_rate": 2.699890789952676e-06, + "loss": 0.0668, + "step": 5546 + }, + { + "epoch": 1.01, + "learning_rate": 2.699836184929014e-06, + "loss": 0.6967, + "step": 5547 + }, + { + "epoch": 1.01, + "learning_rate": 2.6997815799053513e-06, + "loss": 1.969, + "step": 5548 + }, + { + "epoch": 1.01, + "learning_rate": 2.6997269748816893e-06, + "loss": 1.8632, + "step": 5549 + }, + { + "epoch": 1.01, + "learning_rate": 2.6996723698580272e-06, + "loss": 1.3362, + "step": 5550 + }, + { + "epoch": 1.01, + "learning_rate": 2.6996177648343647e-06, + "loss": 0.5915, + "step": 5551 + }, + { + "epoch": 1.01, + "learning_rate": 2.6995631598107027e-06, + "loss": 0.4389, + "step": 5552 + }, + { + "epoch": 1.01, + "learning_rate": 2.6995085547870406e-06, + "loss": 2.9388, + "step": 5553 + }, + { + "epoch": 1.01, + "learning_rate": 2.699453949763378e-06, + "loss": 1.8906, + "step": 5554 + }, + { + "epoch": 1.01, + "learning_rate": 2.699399344739716e-06, + "loss": 1.368, + "step": 5555 + }, + { + "epoch": 1.01, + "learning_rate": 2.699344739716054e-06, + "loss": 1.1708, + "step": 5556 + }, + { + "epoch": 1.01, + "learning_rate": 2.6992901346923916e-06, + "loss": 1.6922, + "step": 5557 + }, + { + "epoch": 1.01, + "learning_rate": 2.6992355296687295e-06, + "loss": 0.2672, + "step": 5558 + }, + { + "epoch": 1.01, + "learning_rate": 2.6991809246450674e-06, + "loss": 0.1167, + "step": 5559 + }, + { + "epoch": 1.01, + "learning_rate": 2.6991263196214054e-06, + "loss": 0.9244, + "step": 5560 + }, + { + "epoch": 1.01, + "learning_rate": 2.6990717145977433e-06, + "loss": 1.3272, + "step": 5561 + }, + { + "epoch": 1.01, + "learning_rate": 2.699017109574081e-06, + "loss": 0.2242, + "step": 5562 + }, + { + "epoch": 1.01, + "learning_rate": 2.698962504550419e-06, + "loss": 1.2379, + "step": 5563 + }, + { + "epoch": 1.01, + "learning_rate": 2.6989078995267567e-06, + "loss": 0.8259, + "step": 5564 + }, + { + "epoch": 1.01, + "learning_rate": 2.6988532945030943e-06, + "loss": 1.2886, + "step": 5565 + }, + { + "epoch": 1.01, + "learning_rate": 2.698798689479432e-06, + "loss": 0.9178, + "step": 5566 + }, + { + "epoch": 1.01, + "learning_rate": 2.69874408445577e-06, + "loss": 1.7909, + "step": 5567 + }, + { + "epoch": 1.01, + "learning_rate": 2.6986894794321077e-06, + "loss": 1.2955, + "step": 5568 + }, + { + "epoch": 1.01, + "learning_rate": 2.6986348744084456e-06, + "loss": 2.1205, + "step": 5569 + }, + { + "epoch": 1.01, + "learning_rate": 2.698580269384783e-06, + "loss": 1.7411, + "step": 5570 + }, + { + "epoch": 1.01, + "learning_rate": 2.698525664361121e-06, + "loss": 1.1468, + "step": 5571 + }, + { + "epoch": 1.01, + "learning_rate": 2.698471059337459e-06, + "loss": 0.83, + "step": 5572 + }, + { + "epoch": 1.01, + "learning_rate": 2.698416454313797e-06, + "loss": 2.048, + "step": 5573 + }, + { + "epoch": 1.01, + "learning_rate": 2.698361849290135e-06, + "loss": 0.8473, + "step": 5574 + }, + { + "epoch": 1.01, + "learning_rate": 2.698307244266473e-06, + "loss": 0.8235, + "step": 5575 + }, + { + "epoch": 1.01, + "learning_rate": 2.6982526392428104e-06, + "loss": 1.0377, + "step": 5576 + }, + { + "epoch": 1.01, + "learning_rate": 2.6981980342191483e-06, + "loss": 1.3688, + "step": 5577 + }, + { + "epoch": 1.01, + "learning_rate": 2.6981434291954862e-06, + "loss": 1.5651, + "step": 5578 + }, + { + "epoch": 1.01, + "learning_rate": 2.6980888241718238e-06, + "loss": 0.7514, + "step": 5579 + }, + { + "epoch": 1.01, + "learning_rate": 2.6980342191481617e-06, + "loss": 0.6476, + "step": 5580 + }, + { + "epoch": 1.01, + "learning_rate": 2.6979796141244997e-06, + "loss": 0.6208, + "step": 5581 + }, + { + "epoch": 1.02, + "learning_rate": 2.697925009100837e-06, + "loss": 2.4935, + "step": 5582 + }, + { + "epoch": 1.02, + "learning_rate": 2.697870404077175e-06, + "loss": 1.3112, + "step": 5583 + }, + { + "epoch": 1.02, + "learning_rate": 2.6978157990535126e-06, + "loss": 0.9379, + "step": 5584 + }, + { + "epoch": 1.02, + "learning_rate": 2.6977611940298506e-06, + "loss": 0.2767, + "step": 5585 + }, + { + "epoch": 1.02, + "learning_rate": 2.6977065890061885e-06, + "loss": 0.8319, + "step": 5586 + }, + { + "epoch": 1.02, + "learning_rate": 2.6976519839825265e-06, + "loss": 0.6651, + "step": 5587 + }, + { + "epoch": 1.02, + "learning_rate": 2.6975973789588644e-06, + "loss": 0.5684, + "step": 5588 + }, + { + "epoch": 1.02, + "learning_rate": 2.6975427739352024e-06, + "loss": 0.8747, + "step": 5589 + }, + { + "epoch": 1.02, + "learning_rate": 2.69748816891154e-06, + "loss": 1.1765, + "step": 5590 + }, + { + "epoch": 1.02, + "learning_rate": 2.697433563887878e-06, + "loss": 0.4922, + "step": 5591 + }, + { + "epoch": 1.02, + "learning_rate": 2.6973789588642158e-06, + "loss": 1.5671, + "step": 5592 + }, + { + "epoch": 1.02, + "learning_rate": 2.6973243538405533e-06, + "loss": 0.284, + "step": 5593 + }, + { + "epoch": 1.02, + "learning_rate": 2.6972697488168912e-06, + "loss": 1.7301, + "step": 5594 + }, + { + "epoch": 1.02, + "learning_rate": 2.697215143793229e-06, + "loss": 0.6783, + "step": 5595 + }, + { + "epoch": 1.02, + "learning_rate": 2.6971605387695667e-06, + "loss": 0.1582, + "step": 5596 + }, + { + "epoch": 1.02, + "learning_rate": 2.6971059337459046e-06, + "loss": 0.8507, + "step": 5597 + }, + { + "epoch": 1.02, + "learning_rate": 2.697051328722242e-06, + "loss": 0.3794, + "step": 5598 + }, + { + "epoch": 1.02, + "learning_rate": 2.69699672369858e-06, + "loss": 1.8309, + "step": 5599 + }, + { + "epoch": 1.02, + "learning_rate": 2.696942118674918e-06, + "loss": 1.5687, + "step": 5600 + }, + { + "epoch": 1.02, + "learning_rate": 2.696887513651256e-06, + "loss": 0.6351, + "step": 5601 + }, + { + "epoch": 1.02, + "learning_rate": 2.696832908627594e-06, + "loss": 0.0464, + "step": 5602 + }, + { + "epoch": 1.02, + "learning_rate": 2.696778303603932e-06, + "loss": 1.0871, + "step": 5603 + }, + { + "epoch": 1.02, + "learning_rate": 2.6967236985802694e-06, + "loss": 2.9073, + "step": 5604 + }, + { + "epoch": 1.02, + "learning_rate": 2.6966690935566073e-06, + "loss": 1.2649, + "step": 5605 + }, + { + "epoch": 1.02, + "learning_rate": 2.6966144885329453e-06, + "loss": 1.2827, + "step": 5606 + }, + { + "epoch": 1.02, + "learning_rate": 2.696559883509283e-06, + "loss": 1.0421, + "step": 5607 + }, + { + "epoch": 1.02, + "learning_rate": 2.6965052784856207e-06, + "loss": 4.4613, + "step": 5608 + }, + { + "epoch": 1.02, + "learning_rate": 2.6964506734619587e-06, + "loss": 0.6112, + "step": 5609 + }, + { + "epoch": 1.02, + "learning_rate": 2.696396068438296e-06, + "loss": 0.3299, + "step": 5610 + }, + { + "epoch": 1.02, + "learning_rate": 2.696341463414634e-06, + "loss": 0.8025, + "step": 5611 + }, + { + "epoch": 1.02, + "learning_rate": 2.696286858390972e-06, + "loss": 1.1269, + "step": 5612 + }, + { + "epoch": 1.02, + "learning_rate": 2.69623225336731e-06, + "loss": 0.0312, + "step": 5613 + }, + { + "epoch": 1.02, + "learning_rate": 2.696177648343648e-06, + "loss": 0.6422, + "step": 5614 + }, + { + "epoch": 1.02, + "learning_rate": 2.6961230433199855e-06, + "loss": 1.5742, + "step": 5615 + }, + { + "epoch": 1.02, + "learning_rate": 2.6960684382963234e-06, + "loss": 0.4609, + "step": 5616 + }, + { + "epoch": 1.02, + "learning_rate": 2.6960138332726614e-06, + "loss": 0.1938, + "step": 5617 + }, + { + "epoch": 1.02, + "learning_rate": 2.695959228248999e-06, + "loss": 1.0937, + "step": 5618 + }, + { + "epoch": 1.02, + "learning_rate": 2.695904623225337e-06, + "loss": 1.1278, + "step": 5619 + }, + { + "epoch": 1.02, + "learning_rate": 2.6958500182016748e-06, + "loss": 0.3007, + "step": 5620 + }, + { + "epoch": 1.02, + "learning_rate": 2.6957954131780123e-06, + "loss": 1.6668, + "step": 5621 + }, + { + "epoch": 1.02, + "learning_rate": 2.6957408081543503e-06, + "loss": 0.2322, + "step": 5622 + }, + { + "epoch": 1.02, + "learning_rate": 2.695686203130688e-06, + "loss": 0.1946, + "step": 5623 + }, + { + "epoch": 1.02, + "learning_rate": 2.6956315981070257e-06, + "loss": 2.9966, + "step": 5624 + }, + { + "epoch": 1.02, + "learning_rate": 2.6955769930833637e-06, + "loss": 1.0296, + "step": 5625 + }, + { + "epoch": 1.02, + "learning_rate": 2.6955223880597016e-06, + "loss": 1.408, + "step": 5626 + }, + { + "epoch": 1.02, + "learning_rate": 2.6954677830360395e-06, + "loss": 0.037, + "step": 5627 + }, + { + "epoch": 1.02, + "learning_rate": 2.6954131780123775e-06, + "loss": 0.8948, + "step": 5628 + }, + { + "epoch": 1.02, + "learning_rate": 2.695358572988715e-06, + "loss": 0.911, + "step": 5629 + }, + { + "epoch": 1.02, + "learning_rate": 2.695303967965053e-06, + "loss": 1.1012, + "step": 5630 + }, + { + "epoch": 1.02, + "learning_rate": 2.695249362941391e-06, + "loss": 2.4145, + "step": 5631 + }, + { + "epoch": 1.02, + "learning_rate": 2.6951947579177284e-06, + "loss": 0.4553, + "step": 5632 + }, + { + "epoch": 1.02, + "learning_rate": 2.6951401528940664e-06, + "loss": 1.4914, + "step": 5633 + }, + { + "epoch": 1.02, + "learning_rate": 2.6950855478704043e-06, + "loss": 0.5601, + "step": 5634 + }, + { + "epoch": 1.02, + "learning_rate": 2.695030942846742e-06, + "loss": 2.0604, + "step": 5635 + }, + { + "epoch": 1.02, + "learning_rate": 2.6949763378230798e-06, + "loss": 0.3019, + "step": 5636 + }, + { + "epoch": 1.03, + "learning_rate": 2.6949217327994177e-06, + "loss": 0.998, + "step": 5637 + }, + { + "epoch": 1.03, + "learning_rate": 2.6948671277757552e-06, + "loss": 0.6098, + "step": 5638 + }, + { + "epoch": 1.03, + "learning_rate": 2.694812522752093e-06, + "loss": 0.4631, + "step": 5639 + }, + { + "epoch": 1.03, + "learning_rate": 2.694757917728431e-06, + "loss": 0.081, + "step": 5640 + }, + { + "epoch": 1.03, + "learning_rate": 2.694703312704769e-06, + "loss": 0.2743, + "step": 5641 + }, + { + "epoch": 1.03, + "learning_rate": 2.694648707681107e-06, + "loss": 1.1033, + "step": 5642 + }, + { + "epoch": 1.03, + "learning_rate": 2.6945941026574445e-06, + "loss": 0.8832, + "step": 5643 + }, + { + "epoch": 1.03, + "learning_rate": 2.6945394976337825e-06, + "loss": 0.5144, + "step": 5644 + }, + { + "epoch": 1.03, + "learning_rate": 2.6944848926101204e-06, + "loss": 0.8703, + "step": 5645 + }, + { + "epoch": 1.03, + "learning_rate": 2.694430287586458e-06, + "loss": 2.5827, + "step": 5646 + }, + { + "epoch": 1.03, + "learning_rate": 2.694375682562796e-06, + "loss": 1.3178, + "step": 5647 + }, + { + "epoch": 1.03, + "learning_rate": 2.694321077539134e-06, + "loss": 0.6226, + "step": 5648 + }, + { + "epoch": 1.03, + "learning_rate": 2.6942664725154713e-06, + "loss": 1.8293, + "step": 5649 + }, + { + "epoch": 1.03, + "learning_rate": 2.6942118674918093e-06, + "loss": 0.9867, + "step": 5650 + }, + { + "epoch": 1.03, + "learning_rate": 2.694157262468147e-06, + "loss": 0.1401, + "step": 5651 + }, + { + "epoch": 1.03, + "learning_rate": 2.6941026574444847e-06, + "loss": 0.048, + "step": 5652 + }, + { + "epoch": 1.03, + "learning_rate": 2.6940480524208227e-06, + "loss": 2.4736, + "step": 5653 + }, + { + "epoch": 1.03, + "learning_rate": 2.6939934473971606e-06, + "loss": 1.369, + "step": 5654 + }, + { + "epoch": 1.03, + "learning_rate": 2.6939388423734986e-06, + "loss": 0.2555, + "step": 5655 + }, + { + "epoch": 1.03, + "learning_rate": 2.6938842373498365e-06, + "loss": 0.4221, + "step": 5656 + }, + { + "epoch": 1.03, + "learning_rate": 2.693829632326174e-06, + "loss": 1.2901, + "step": 5657 + }, + { + "epoch": 1.03, + "learning_rate": 2.693775027302512e-06, + "loss": 0.6383, + "step": 5658 + }, + { + "epoch": 1.03, + "learning_rate": 2.69372042227885e-06, + "loss": 0.524, + "step": 5659 + }, + { + "epoch": 1.03, + "learning_rate": 2.6936658172551874e-06, + "loss": 0.7921, + "step": 5660 + }, + { + "epoch": 1.03, + "learning_rate": 2.6936112122315254e-06, + "loss": 1.0779, + "step": 5661 + }, + { + "epoch": 1.03, + "learning_rate": 2.6935566072078633e-06, + "loss": 1.2157, + "step": 5662 + }, + { + "epoch": 1.03, + "learning_rate": 2.693502002184201e-06, + "loss": 1.254, + "step": 5663 + }, + { + "epoch": 1.03, + "learning_rate": 2.693447397160539e-06, + "loss": 0.2707, + "step": 5664 + }, + { + "epoch": 1.03, + "learning_rate": 2.6933927921368763e-06, + "loss": 2.2814, + "step": 5665 + }, + { + "epoch": 1.03, + "learning_rate": 2.6933381871132143e-06, + "loss": 0.8819, + "step": 5666 + }, + { + "epoch": 1.03, + "learning_rate": 2.693283582089552e-06, + "loss": 1.2586, + "step": 5667 + }, + { + "epoch": 1.03, + "learning_rate": 2.69322897706589e-06, + "loss": 0.6612, + "step": 5668 + }, + { + "epoch": 1.03, + "learning_rate": 2.693174372042228e-06, + "loss": 1.348, + "step": 5669 + }, + { + "epoch": 1.03, + "learning_rate": 2.693119767018566e-06, + "loss": 0.1499, + "step": 5670 + }, + { + "epoch": 1.03, + "learning_rate": 2.6930651619949035e-06, + "loss": 1.9083, + "step": 5671 + }, + { + "epoch": 1.03, + "learning_rate": 2.6930105569712415e-06, + "loss": 0.8056, + "step": 5672 + }, + { + "epoch": 1.03, + "learning_rate": 2.6929559519475794e-06, + "loss": 1.2059, + "step": 5673 + }, + { + "epoch": 1.03, + "learning_rate": 2.692901346923917e-06, + "loss": 0.999, + "step": 5674 + }, + { + "epoch": 1.03, + "learning_rate": 2.692846741900255e-06, + "loss": 2.5588, + "step": 5675 + }, + { + "epoch": 1.03, + "learning_rate": 2.692792136876593e-06, + "loss": 0.7677, + "step": 5676 + }, + { + "epoch": 1.03, + "learning_rate": 2.6927375318529304e-06, + "loss": 2.3093, + "step": 5677 + }, + { + "epoch": 1.03, + "learning_rate": 2.6926829268292683e-06, + "loss": 2.3492, + "step": 5678 + }, + { + "epoch": 1.03, + "learning_rate": 2.692628321805606e-06, + "loss": 0.5904, + "step": 5679 + }, + { + "epoch": 1.03, + "learning_rate": 2.6925737167819438e-06, + "loss": 2.7962, + "step": 5680 + }, + { + "epoch": 1.03, + "learning_rate": 2.6925191117582817e-06, + "loss": 1.1264, + "step": 5681 + }, + { + "epoch": 1.03, + "learning_rate": 2.6924645067346197e-06, + "loss": 2.6561, + "step": 5682 + }, + { + "epoch": 1.03, + "learning_rate": 2.6924099017109576e-06, + "loss": 0.0744, + "step": 5683 + }, + { + "epoch": 1.03, + "learning_rate": 2.6923552966872955e-06, + "loss": 0.5751, + "step": 5684 + }, + { + "epoch": 1.03, + "learning_rate": 2.692300691663633e-06, + "loss": 1.3324, + "step": 5685 + }, + { + "epoch": 1.03, + "learning_rate": 2.692246086639971e-06, + "loss": 0.1235, + "step": 5686 + }, + { + "epoch": 1.03, + "learning_rate": 2.692191481616309e-06, + "loss": 0.3071, + "step": 5687 + }, + { + "epoch": 1.03, + "learning_rate": 2.6921368765926465e-06, + "loss": 1.9056, + "step": 5688 + }, + { + "epoch": 1.03, + "learning_rate": 2.6920822715689844e-06, + "loss": 0.7428, + "step": 5689 + }, + { + "epoch": 1.03, + "learning_rate": 2.6920276665453224e-06, + "loss": 1.5748, + "step": 5690 + }, + { + "epoch": 1.03, + "learning_rate": 2.69197306152166e-06, + "loss": 2.1226, + "step": 5691 + }, + { + "epoch": 1.04, + "learning_rate": 2.691918456497998e-06, + "loss": 2.6828, + "step": 5692 + }, + { + "epoch": 1.04, + "learning_rate": 2.6918638514743358e-06, + "loss": 0.0605, + "step": 5693 + }, + { + "epoch": 1.04, + "learning_rate": 2.6918092464506737e-06, + "loss": 0.3186, + "step": 5694 + }, + { + "epoch": 1.04, + "learning_rate": 2.6917546414270116e-06, + "loss": 0.4385, + "step": 5695 + }, + { + "epoch": 1.04, + "learning_rate": 2.691700036403349e-06, + "loss": 1.7409, + "step": 5696 + }, + { + "epoch": 1.04, + "learning_rate": 2.691645431379687e-06, + "loss": 0.3824, + "step": 5697 + }, + { + "epoch": 1.04, + "learning_rate": 2.691590826356025e-06, + "loss": 0.7787, + "step": 5698 + }, + { + "epoch": 1.04, + "learning_rate": 2.6915362213323626e-06, + "loss": 0.1052, + "step": 5699 + }, + { + "epoch": 1.04, + "learning_rate": 2.6914816163087005e-06, + "loss": 1.85, + "step": 5700 + }, + { + "epoch": 1.04, + "learning_rate": 2.6914270112850385e-06, + "loss": 3.9936, + "step": 5701 + }, + { + "epoch": 1.04, + "learning_rate": 2.691372406261376e-06, + "loss": 0.751, + "step": 5702 + }, + { + "epoch": 1.04, + "learning_rate": 2.691317801237714e-06, + "loss": 0.5504, + "step": 5703 + }, + { + "epoch": 1.04, + "learning_rate": 2.691263196214052e-06, + "loss": 0.2781, + "step": 5704 + }, + { + "epoch": 1.04, + "learning_rate": 2.6912085911903894e-06, + "loss": 1.4662, + "step": 5705 + }, + { + "epoch": 1.04, + "learning_rate": 2.6911539861667273e-06, + "loss": 0.5166, + "step": 5706 + }, + { + "epoch": 1.04, + "learning_rate": 2.6910993811430653e-06, + "loss": 0.4307, + "step": 5707 + }, + { + "epoch": 1.04, + "learning_rate": 2.6910447761194032e-06, + "loss": 0.1582, + "step": 5708 + }, + { + "epoch": 1.04, + "learning_rate": 2.690990171095741e-06, + "loss": 0.4114, + "step": 5709 + }, + { + "epoch": 1.04, + "learning_rate": 2.6909355660720787e-06, + "loss": 0.3881, + "step": 5710 + }, + { + "epoch": 1.04, + "learning_rate": 2.6908809610484166e-06, + "loss": 2.7254, + "step": 5711 + }, + { + "epoch": 1.04, + "learning_rate": 2.6908263560247546e-06, + "loss": 1.097, + "step": 5712 + }, + { + "epoch": 1.04, + "learning_rate": 2.690771751001092e-06, + "loss": 2.6597, + "step": 5713 + }, + { + "epoch": 1.04, + "learning_rate": 2.69071714597743e-06, + "loss": 0.6608, + "step": 5714 + }, + { + "epoch": 1.04, + "learning_rate": 2.690662540953768e-06, + "loss": 1.3735, + "step": 5715 + }, + { + "epoch": 1.04, + "learning_rate": 2.6906079359301055e-06, + "loss": 0.9516, + "step": 5716 + }, + { + "epoch": 1.04, + "learning_rate": 2.6905533309064434e-06, + "loss": 1.3472, + "step": 5717 + }, + { + "epoch": 1.04, + "learning_rate": 2.690498725882781e-06, + "loss": 0.483, + "step": 5718 + }, + { + "epoch": 1.04, + "learning_rate": 2.690444120859119e-06, + "loss": 2.9393, + "step": 5719 + }, + { + "epoch": 1.04, + "learning_rate": 2.690389515835457e-06, + "loss": 0.8169, + "step": 5720 + }, + { + "epoch": 1.04, + "learning_rate": 2.690334910811795e-06, + "loss": 1.7941, + "step": 5721 + }, + { + "epoch": 1.04, + "learning_rate": 2.6902803057881327e-06, + "loss": 0.7395, + "step": 5722 + }, + { + "epoch": 1.04, + "learning_rate": 2.6902257007644707e-06, + "loss": 1.2374, + "step": 5723 + }, + { + "epoch": 1.04, + "learning_rate": 2.690171095740808e-06, + "loss": 0.5615, + "step": 5724 + }, + { + "epoch": 1.04, + "learning_rate": 2.690116490717146e-06, + "loss": 0.0827, + "step": 5725 + }, + { + "epoch": 1.04, + "learning_rate": 2.690061885693484e-06, + "loss": 2.2032, + "step": 5726 + }, + { + "epoch": 1.04, + "learning_rate": 2.6900072806698216e-06, + "loss": 0.5201, + "step": 5727 + }, + { + "epoch": 1.04, + "learning_rate": 2.6899526756461595e-06, + "loss": 3.408, + "step": 5728 + }, + { + "epoch": 1.04, + "learning_rate": 2.6898980706224975e-06, + "loss": 1.2709, + "step": 5729 + }, + { + "epoch": 1.04, + "learning_rate": 2.689843465598835e-06, + "loss": 1.0144, + "step": 5730 + }, + { + "epoch": 1.04, + "learning_rate": 2.689788860575173e-06, + "loss": 0.3858, + "step": 5731 + }, + { + "epoch": 1.04, + "learning_rate": 2.6897342555515105e-06, + "loss": 0.4513, + "step": 5732 + }, + { + "epoch": 1.04, + "learning_rate": 2.6896796505278484e-06, + "loss": 0.5889, + "step": 5733 + }, + { + "epoch": 1.04, + "learning_rate": 2.6896250455041864e-06, + "loss": 0.7003, + "step": 5734 + }, + { + "epoch": 1.04, + "learning_rate": 2.6895704404805243e-06, + "loss": 0.1742, + "step": 5735 + }, + { + "epoch": 1.04, + "learning_rate": 2.6895158354568622e-06, + "loss": 0.1254, + "step": 5736 + }, + { + "epoch": 1.04, + "learning_rate": 2.6894612304332e-06, + "loss": 0.6952, + "step": 5737 + }, + { + "epoch": 1.04, + "learning_rate": 2.6894066254095377e-06, + "loss": 0.5165, + "step": 5738 + }, + { + "epoch": 1.04, + "learning_rate": 2.6893520203858757e-06, + "loss": 0.9101, + "step": 5739 + }, + { + "epoch": 1.04, + "learning_rate": 2.6892974153622136e-06, + "loss": 1.9058, + "step": 5740 + }, + { + "epoch": 1.04, + "learning_rate": 2.689242810338551e-06, + "loss": 0.0177, + "step": 5741 + }, + { + "epoch": 1.04, + "learning_rate": 2.689188205314889e-06, + "loss": 2.9153, + "step": 5742 + }, + { + "epoch": 1.04, + "learning_rate": 2.689133600291227e-06, + "loss": 0.8758, + "step": 5743 + }, + { + "epoch": 1.04, + "learning_rate": 2.6890789952675645e-06, + "loss": 0.5121, + "step": 5744 + }, + { + "epoch": 1.04, + "learning_rate": 2.6890243902439025e-06, + "loss": 0.4084, + "step": 5745 + }, + { + "epoch": 1.04, + "learning_rate": 2.68896978522024e-06, + "loss": 0.6908, + "step": 5746 + }, + { + "epoch": 1.05, + "learning_rate": 2.688915180196578e-06, + "loss": 0.1902, + "step": 5747 + }, + { + "epoch": 1.05, + "learning_rate": 2.688860575172916e-06, + "loss": 0.8195, + "step": 5748 + }, + { + "epoch": 1.05, + "learning_rate": 2.688805970149254e-06, + "loss": 1.3802, + "step": 5749 + }, + { + "epoch": 1.05, + "learning_rate": 2.6887513651255918e-06, + "loss": 0.7612, + "step": 5750 + }, + { + "epoch": 1.05, + "learning_rate": 2.6886967601019297e-06, + "loss": 1.7452, + "step": 5751 + }, + { + "epoch": 1.05, + "learning_rate": 2.6886421550782672e-06, + "loss": 0.1181, + "step": 5752 + }, + { + "epoch": 1.05, + "learning_rate": 2.688587550054605e-06, + "loss": 2.8505, + "step": 5753 + }, + { + "epoch": 1.05, + "learning_rate": 2.688532945030943e-06, + "loss": 0.5969, + "step": 5754 + }, + { + "epoch": 1.05, + "learning_rate": 2.6884783400072806e-06, + "loss": 1.3568, + "step": 5755 + }, + { + "epoch": 1.05, + "learning_rate": 2.6884237349836186e-06, + "loss": 0.262, + "step": 5756 + }, + { + "epoch": 1.05, + "learning_rate": 2.6883691299599565e-06, + "loss": 4.2878, + "step": 5757 + }, + { + "epoch": 1.05, + "learning_rate": 2.688314524936294e-06, + "loss": 0.3851, + "step": 5758 + }, + { + "epoch": 1.05, + "learning_rate": 2.688259919912632e-06, + "loss": 0.5687, + "step": 5759 + }, + { + "epoch": 1.05, + "learning_rate": 2.6882053148889695e-06, + "loss": 0.9327, + "step": 5760 + }, + { + "epoch": 1.05, + "learning_rate": 2.6881507098653074e-06, + "loss": 1.0095, + "step": 5761 + }, + { + "epoch": 1.05, + "learning_rate": 2.6880961048416454e-06, + "loss": 1.1017, + "step": 5762 + }, + { + "epoch": 1.05, + "learning_rate": 2.6880414998179833e-06, + "loss": 0.7677, + "step": 5763 + }, + { + "epoch": 1.05, + "learning_rate": 2.6879868947943213e-06, + "loss": 1.2562, + "step": 5764 + }, + { + "epoch": 1.05, + "learning_rate": 2.6879322897706592e-06, + "loss": 1.2758, + "step": 5765 + }, + { + "epoch": 1.05, + "learning_rate": 2.6878776847469967e-06, + "loss": 0.7469, + "step": 5766 + }, + { + "epoch": 1.05, + "learning_rate": 2.6878230797233347e-06, + "loss": 0.2073, + "step": 5767 + }, + { + "epoch": 1.05, + "learning_rate": 2.6877684746996726e-06, + "loss": 1.5909, + "step": 5768 + }, + { + "epoch": 1.05, + "learning_rate": 2.68771386967601e-06, + "loss": 0.5764, + "step": 5769 + }, + { + "epoch": 1.05, + "learning_rate": 2.687659264652348e-06, + "loss": 2.6613, + "step": 5770 + }, + { + "epoch": 1.05, + "learning_rate": 2.687604659628686e-06, + "loss": 0.3319, + "step": 5771 + }, + { + "epoch": 1.05, + "learning_rate": 2.6875500546050235e-06, + "loss": 2.4099, + "step": 5772 + }, + { + "epoch": 1.05, + "learning_rate": 2.6874954495813615e-06, + "loss": 1.5487, + "step": 5773 + }, + { + "epoch": 1.05, + "learning_rate": 2.6874408445576994e-06, + "loss": 1.243, + "step": 5774 + }, + { + "epoch": 1.05, + "learning_rate": 2.6873862395340374e-06, + "loss": 1.2661, + "step": 5775 + }, + { + "epoch": 1.05, + "learning_rate": 2.6873316345103753e-06, + "loss": 0.303, + "step": 5776 + }, + { + "epoch": 1.05, + "learning_rate": 2.687277029486713e-06, + "loss": 1.1288, + "step": 5777 + }, + { + "epoch": 1.05, + "learning_rate": 2.6872224244630508e-06, + "loss": 0.7194, + "step": 5778 + }, + { + "epoch": 1.05, + "learning_rate": 2.6871678194393887e-06, + "loss": 0.1483, + "step": 5779 + }, + { + "epoch": 1.05, + "learning_rate": 2.6871132144157262e-06, + "loss": 1.7387, + "step": 5780 + }, + { + "epoch": 1.05, + "learning_rate": 2.687058609392064e-06, + "loss": 1.224, + "step": 5781 + }, + { + "epoch": 1.05, + "learning_rate": 2.687004004368402e-06, + "loss": 0.9864, + "step": 5782 + }, + { + "epoch": 1.05, + "learning_rate": 2.6869493993447397e-06, + "loss": 0.7071, + "step": 5783 + }, + { + "epoch": 1.05, + "learning_rate": 2.6868947943210776e-06, + "loss": 0.9337, + "step": 5784 + }, + { + "epoch": 1.05, + "learning_rate": 2.6868401892974155e-06, + "loss": 1.1587, + "step": 5785 + }, + { + "epoch": 1.05, + "learning_rate": 2.686785584273753e-06, + "loss": 0.4968, + "step": 5786 + }, + { + "epoch": 1.05, + "learning_rate": 2.686730979250091e-06, + "loss": 0.4165, + "step": 5787 + }, + { + "epoch": 1.05, + "learning_rate": 2.686676374226429e-06, + "loss": 0.9211, + "step": 5788 + }, + { + "epoch": 1.05, + "learning_rate": 2.686621769202767e-06, + "loss": 0.5546, + "step": 5789 + }, + { + "epoch": 1.05, + "learning_rate": 2.686567164179105e-06, + "loss": 0.7087, + "step": 5790 + }, + { + "epoch": 1.05, + "learning_rate": 2.6865125591554424e-06, + "loss": 0.7044, + "step": 5791 + }, + { + "epoch": 1.05, + "learning_rate": 2.6864579541317803e-06, + "loss": 0.7349, + "step": 5792 + }, + { + "epoch": 1.05, + "learning_rate": 2.6864033491081182e-06, + "loss": 1.3541, + "step": 5793 + }, + { + "epoch": 1.05, + "learning_rate": 2.6863487440844558e-06, + "loss": 0.5522, + "step": 5794 + }, + { + "epoch": 1.05, + "learning_rate": 2.6862941390607937e-06, + "loss": 2.5696, + "step": 5795 + }, + { + "epoch": 1.05, + "learning_rate": 2.6862395340371316e-06, + "loss": 0.2577, + "step": 5796 + }, + { + "epoch": 1.05, + "learning_rate": 2.686184929013469e-06, + "loss": 0.4308, + "step": 5797 + }, + { + "epoch": 1.05, + "learning_rate": 2.686130323989807e-06, + "loss": 0.9985, + "step": 5798 + }, + { + "epoch": 1.05, + "learning_rate": 2.6860757189661446e-06, + "loss": 1.1244, + "step": 5799 + }, + { + "epoch": 1.05, + "learning_rate": 2.6860211139424826e-06, + "loss": 1.9021, + "step": 5800 + }, + { + "epoch": 1.05, + "learning_rate": 2.6859665089188205e-06, + "loss": 1.3663, + "step": 5801 + }, + { + "epoch": 1.06, + "learning_rate": 2.6859119038951585e-06, + "loss": 1.9338, + "step": 5802 + }, + { + "epoch": 1.06, + "learning_rate": 2.6858572988714964e-06, + "loss": 1.4951, + "step": 5803 + }, + { + "epoch": 1.06, + "learning_rate": 2.6858026938478343e-06, + "loss": 1.9571, + "step": 5804 + }, + { + "epoch": 1.06, + "learning_rate": 2.685748088824172e-06, + "loss": 0.9074, + "step": 5805 + }, + { + "epoch": 1.06, + "learning_rate": 2.68569348380051e-06, + "loss": 0.1535, + "step": 5806 + }, + { + "epoch": 1.06, + "learning_rate": 2.6856388787768478e-06, + "loss": 1.0656, + "step": 5807 + }, + { + "epoch": 1.06, + "learning_rate": 2.6855842737531853e-06, + "loss": 0.1142, + "step": 5808 + }, + { + "epoch": 1.06, + "learning_rate": 2.6855296687295232e-06, + "loss": 0.423, + "step": 5809 + }, + { + "epoch": 1.06, + "learning_rate": 2.685475063705861e-06, + "loss": 0.3762, + "step": 5810 + }, + { + "epoch": 1.06, + "learning_rate": 2.6854204586821987e-06, + "loss": 1.8356, + "step": 5811 + }, + { + "epoch": 1.06, + "learning_rate": 2.6853658536585366e-06, + "loss": 1.7201, + "step": 5812 + }, + { + "epoch": 1.06, + "learning_rate": 2.685311248634874e-06, + "loss": 2.225, + "step": 5813 + }, + { + "epoch": 1.06, + "learning_rate": 2.685256643611212e-06, + "loss": 0.2985, + "step": 5814 + }, + { + "epoch": 1.06, + "learning_rate": 2.68520203858755e-06, + "loss": 1.1515, + "step": 5815 + }, + { + "epoch": 1.06, + "learning_rate": 2.685147433563888e-06, + "loss": 1.5974, + "step": 5816 + }, + { + "epoch": 1.06, + "learning_rate": 2.685092828540226e-06, + "loss": 0.5451, + "step": 5817 + }, + { + "epoch": 1.06, + "learning_rate": 2.685038223516564e-06, + "loss": 0.7001, + "step": 5818 + }, + { + "epoch": 1.06, + "learning_rate": 2.6849836184929014e-06, + "loss": 2.1982, + "step": 5819 + }, + { + "epoch": 1.06, + "learning_rate": 2.6849290134692393e-06, + "loss": 0.6491, + "step": 5820 + }, + { + "epoch": 1.06, + "learning_rate": 2.6848744084455773e-06, + "loss": 0.3346, + "step": 5821 + }, + { + "epoch": 1.06, + "learning_rate": 2.684819803421915e-06, + "loss": 0.8149, + "step": 5822 + }, + { + "epoch": 1.06, + "learning_rate": 2.6847651983982527e-06, + "loss": 0.3841, + "step": 5823 + }, + { + "epoch": 1.06, + "learning_rate": 2.6847105933745907e-06, + "loss": 1.0337, + "step": 5824 + }, + { + "epoch": 1.06, + "learning_rate": 2.684655988350928e-06, + "loss": 0.1469, + "step": 5825 + }, + { + "epoch": 1.06, + "learning_rate": 2.684601383327266e-06, + "loss": 0.7769, + "step": 5826 + }, + { + "epoch": 1.06, + "learning_rate": 2.6845467783036037e-06, + "loss": 0.806, + "step": 5827 + }, + { + "epoch": 1.06, + "learning_rate": 2.6844921732799416e-06, + "loss": 0.7679, + "step": 5828 + }, + { + "epoch": 1.06, + "learning_rate": 2.6844375682562795e-06, + "loss": 1.3516, + "step": 5829 + }, + { + "epoch": 1.06, + "learning_rate": 2.6843829632326175e-06, + "loss": 1.2605, + "step": 5830 + }, + { + "epoch": 1.06, + "learning_rate": 2.6843283582089554e-06, + "loss": 0.0965, + "step": 5831 + }, + { + "epoch": 1.06, + "learning_rate": 2.6842737531852934e-06, + "loss": 0.7684, + "step": 5832 + }, + { + "epoch": 1.06, + "learning_rate": 2.684219148161631e-06, + "loss": 0.3423, + "step": 5833 + }, + { + "epoch": 1.06, + "learning_rate": 2.684164543137969e-06, + "loss": 1.225, + "step": 5834 + }, + { + "epoch": 1.06, + "learning_rate": 2.6841099381143068e-06, + "loss": 0.5946, + "step": 5835 + }, + { + "epoch": 1.06, + "learning_rate": 2.6840553330906443e-06, + "loss": 2.6604, + "step": 5836 + }, + { + "epoch": 1.06, + "learning_rate": 2.6840007280669822e-06, + "loss": 0.3507, + "step": 5837 + }, + { + "epoch": 1.06, + "learning_rate": 2.68394612304332e-06, + "loss": 1.2198, + "step": 5838 + }, + { + "epoch": 1.06, + "learning_rate": 2.6838915180196577e-06, + "loss": 1.0649, + "step": 5839 + }, + { + "epoch": 1.06, + "learning_rate": 2.6838369129959957e-06, + "loss": 0.8393, + "step": 5840 + }, + { + "epoch": 1.06, + "learning_rate": 2.683782307972333e-06, + "loss": 1.0455, + "step": 5841 + }, + { + "epoch": 1.06, + "learning_rate": 2.683727702948671e-06, + "loss": 0.0555, + "step": 5842 + }, + { + "epoch": 1.06, + "learning_rate": 2.683673097925009e-06, + "loss": 0.2592, + "step": 5843 + }, + { + "epoch": 1.06, + "learning_rate": 2.683618492901347e-06, + "loss": 0.7697, + "step": 5844 + }, + { + "epoch": 1.06, + "learning_rate": 2.683563887877685e-06, + "loss": 0.0478, + "step": 5845 + }, + { + "epoch": 1.06, + "learning_rate": 2.683509282854023e-06, + "loss": 1.9965, + "step": 5846 + }, + { + "epoch": 1.06, + "learning_rate": 2.6834546778303604e-06, + "loss": 1.7588, + "step": 5847 + }, + { + "epoch": 1.06, + "learning_rate": 2.6834000728066984e-06, + "loss": 0.3376, + "step": 5848 + }, + { + "epoch": 1.06, + "learning_rate": 2.6833454677830363e-06, + "loss": 1.4263, + "step": 5849 + }, + { + "epoch": 1.06, + "learning_rate": 2.683290862759374e-06, + "loss": 0.0909, + "step": 5850 + }, + { + "epoch": 1.06, + "learning_rate": 2.6832362577357118e-06, + "loss": 2.3954, + "step": 5851 + }, + { + "epoch": 1.06, + "learning_rate": 2.6831816527120497e-06, + "loss": 0.3302, + "step": 5852 + }, + { + "epoch": 1.06, + "learning_rate": 2.6831270476883872e-06, + "loss": 1.504, + "step": 5853 + }, + { + "epoch": 1.06, + "learning_rate": 2.683072442664725e-06, + "loss": 1.6601, + "step": 5854 + }, + { + "epoch": 1.06, + "learning_rate": 2.6830178376410627e-06, + "loss": 0.408, + "step": 5855 + }, + { + "epoch": 1.06, + "learning_rate": 2.6829632326174006e-06, + "loss": 2.4437, + "step": 5856 + }, + { + "epoch": 1.07, + "learning_rate": 2.6829086275937386e-06, + "loss": 0.2977, + "step": 5857 + }, + { + "epoch": 1.07, + "learning_rate": 2.6828540225700765e-06, + "loss": 1.168, + "step": 5858 + }, + { + "epoch": 1.07, + "learning_rate": 2.6827994175464145e-06, + "loss": 0.5492, + "step": 5859 + }, + { + "epoch": 1.07, + "learning_rate": 2.6827448125227524e-06, + "loss": 1.6689, + "step": 5860 + }, + { + "epoch": 1.07, + "learning_rate": 2.68269020749909e-06, + "loss": 1.9091, + "step": 5861 + }, + { + "epoch": 1.07, + "learning_rate": 2.682635602475428e-06, + "loss": 1.3528, + "step": 5862 + }, + { + "epoch": 1.07, + "learning_rate": 2.682580997451766e-06, + "loss": 1.3293, + "step": 5863 + }, + { + "epoch": 1.07, + "learning_rate": 2.6825263924281033e-06, + "loss": 0.4712, + "step": 5864 + }, + { + "epoch": 1.07, + "learning_rate": 2.6824717874044413e-06, + "loss": 0.5956, + "step": 5865 + }, + { + "epoch": 1.07, + "learning_rate": 2.6824171823807792e-06, + "loss": 0.5225, + "step": 5866 + }, + { + "epoch": 1.07, + "learning_rate": 2.6823625773571167e-06, + "loss": 0.7598, + "step": 5867 + }, + { + "epoch": 1.07, + "learning_rate": 2.6823079723334547e-06, + "loss": 0.2204, + "step": 5868 + }, + { + "epoch": 1.07, + "learning_rate": 2.6822533673097926e-06, + "loss": 0.132, + "step": 5869 + }, + { + "epoch": 1.07, + "learning_rate": 2.6821987622861306e-06, + "loss": 1.9779, + "step": 5870 + }, + { + "epoch": 1.07, + "learning_rate": 2.6821441572624685e-06, + "loss": 0.4818, + "step": 5871 + }, + { + "epoch": 1.07, + "learning_rate": 2.682089552238806e-06, + "loss": 1.3286, + "step": 5872 + }, + { + "epoch": 1.07, + "learning_rate": 2.682034947215144e-06, + "loss": 0.3437, + "step": 5873 + }, + { + "epoch": 1.07, + "learning_rate": 2.681980342191482e-06, + "loss": 0.8856, + "step": 5874 + }, + { + "epoch": 1.07, + "learning_rate": 2.6819257371678194e-06, + "loss": 0.6922, + "step": 5875 + }, + { + "epoch": 1.07, + "learning_rate": 2.6818711321441574e-06, + "loss": 0.3133, + "step": 5876 + }, + { + "epoch": 1.07, + "learning_rate": 2.6818165271204953e-06, + "loss": 0.276, + "step": 5877 + }, + { + "epoch": 1.07, + "learning_rate": 2.681761922096833e-06, + "loss": 2.5047, + "step": 5878 + }, + { + "epoch": 1.07, + "learning_rate": 2.6817073170731708e-06, + "loss": 0.4333, + "step": 5879 + }, + { + "epoch": 1.07, + "learning_rate": 2.6816527120495083e-06, + "loss": 0.1125, + "step": 5880 + }, + { + "epoch": 1.07, + "learning_rate": 2.6815981070258463e-06, + "loss": 0.7399, + "step": 5881 + }, + { + "epoch": 1.07, + "learning_rate": 2.681543502002184e-06, + "loss": 0.9246, + "step": 5882 + }, + { + "epoch": 1.07, + "learning_rate": 2.681488896978522e-06, + "loss": 1.357, + "step": 5883 + }, + { + "epoch": 1.07, + "learning_rate": 2.68143429195486e-06, + "loss": 1.8646, + "step": 5884 + }, + { + "epoch": 1.07, + "learning_rate": 2.681379686931198e-06, + "loss": 1.2812, + "step": 5885 + }, + { + "epoch": 1.07, + "learning_rate": 2.6813250819075355e-06, + "loss": 0.7955, + "step": 5886 + }, + { + "epoch": 1.07, + "learning_rate": 2.6812704768838735e-06, + "loss": 0.8731, + "step": 5887 + }, + { + "epoch": 1.07, + "learning_rate": 2.6812158718602114e-06, + "loss": 1.4729, + "step": 5888 + }, + { + "epoch": 1.07, + "learning_rate": 2.681161266836549e-06, + "loss": 0.9708, + "step": 5889 + }, + { + "epoch": 1.07, + "learning_rate": 2.681106661812887e-06, + "loss": 2.2405, + "step": 5890 + }, + { + "epoch": 1.07, + "learning_rate": 2.681052056789225e-06, + "loss": 0.1472, + "step": 5891 + }, + { + "epoch": 1.07, + "learning_rate": 2.6809974517655624e-06, + "loss": 1.0589, + "step": 5892 + }, + { + "epoch": 1.07, + "learning_rate": 2.6809428467419003e-06, + "loss": 1.6768, + "step": 5893 + }, + { + "epoch": 1.07, + "learning_rate": 2.680888241718238e-06, + "loss": 2.1884, + "step": 5894 + }, + { + "epoch": 1.07, + "learning_rate": 2.6808336366945758e-06, + "loss": 0.4321, + "step": 5895 + }, + { + "epoch": 1.07, + "learning_rate": 2.6807790316709137e-06, + "loss": 1.6542, + "step": 5896 + }, + { + "epoch": 1.07, + "learning_rate": 2.6807244266472517e-06, + "loss": 0.4344, + "step": 5897 + }, + { + "epoch": 1.07, + "learning_rate": 2.6806698216235896e-06, + "loss": 0.029, + "step": 5898 + }, + { + "epoch": 1.07, + "learning_rate": 2.6806152165999275e-06, + "loss": 1.2959, + "step": 5899 + }, + { + "epoch": 1.07, + "learning_rate": 2.680560611576265e-06, + "loss": 0.9129, + "step": 5900 + }, + { + "epoch": 1.07, + "learning_rate": 2.680506006552603e-06, + "loss": 0.3782, + "step": 5901 + }, + { + "epoch": 1.07, + "learning_rate": 2.680451401528941e-06, + "loss": 0.3274, + "step": 5902 + }, + { + "epoch": 1.07, + "learning_rate": 2.6803967965052785e-06, + "loss": 1.0502, + "step": 5903 + }, + { + "epoch": 1.07, + "learning_rate": 2.6803421914816164e-06, + "loss": 0.5977, + "step": 5904 + }, + { + "epoch": 1.07, + "learning_rate": 2.6802875864579544e-06, + "loss": 0.5905, + "step": 5905 + }, + { + "epoch": 1.07, + "learning_rate": 2.680232981434292e-06, + "loss": 0.2122, + "step": 5906 + }, + { + "epoch": 1.07, + "learning_rate": 2.68017837641063e-06, + "loss": 0.759, + "step": 5907 + }, + { + "epoch": 1.07, + "learning_rate": 2.6801237713869673e-06, + "loss": 2.2879, + "step": 5908 + }, + { + "epoch": 1.07, + "learning_rate": 2.6800691663633053e-06, + "loss": 0.2936, + "step": 5909 + }, + { + "epoch": 1.07, + "learning_rate": 2.6800145613396432e-06, + "loss": 0.8922, + "step": 5910 + }, + { + "epoch": 1.07, + "learning_rate": 2.679959956315981e-06, + "loss": 0.7988, + "step": 5911 + }, + { + "epoch": 1.08, + "learning_rate": 2.679905351292319e-06, + "loss": 0.5669, + "step": 5912 + }, + { + "epoch": 1.08, + "learning_rate": 2.679850746268657e-06, + "loss": 1.8753, + "step": 5913 + }, + { + "epoch": 1.08, + "learning_rate": 2.6797961412449946e-06, + "loss": 0.0686, + "step": 5914 + }, + { + "epoch": 1.08, + "learning_rate": 2.6797415362213325e-06, + "loss": 0.4377, + "step": 5915 + }, + { + "epoch": 1.08, + "learning_rate": 2.6796869311976705e-06, + "loss": 0.2504, + "step": 5916 + }, + { + "epoch": 1.08, + "learning_rate": 2.679632326174008e-06, + "loss": 2.3354, + "step": 5917 + }, + { + "epoch": 1.08, + "learning_rate": 2.679577721150346e-06, + "loss": 0.7247, + "step": 5918 + }, + { + "epoch": 1.08, + "learning_rate": 2.679523116126684e-06, + "loss": 0.253, + "step": 5919 + }, + { + "epoch": 1.08, + "learning_rate": 2.6794685111030214e-06, + "loss": 0.3747, + "step": 5920 + }, + { + "epoch": 1.08, + "learning_rate": 2.6794139060793593e-06, + "loss": 0.036, + "step": 5921 + }, + { + "epoch": 1.08, + "learning_rate": 2.679359301055697e-06, + "loss": 1.0979, + "step": 5922 + }, + { + "epoch": 1.08, + "learning_rate": 2.679304696032035e-06, + "loss": 1.294, + "step": 5923 + }, + { + "epoch": 1.08, + "learning_rate": 2.6792500910083727e-06, + "loss": 0.022, + "step": 5924 + }, + { + "epoch": 1.08, + "learning_rate": 2.6791954859847107e-06, + "loss": 1.3318, + "step": 5925 + }, + { + "epoch": 1.08, + "learning_rate": 2.6791408809610486e-06, + "loss": 0.6534, + "step": 5926 + }, + { + "epoch": 1.08, + "learning_rate": 2.6790862759373866e-06, + "loss": 2.047, + "step": 5927 + }, + { + "epoch": 1.08, + "learning_rate": 2.679031670913724e-06, + "loss": 1.5286, + "step": 5928 + }, + { + "epoch": 1.08, + "learning_rate": 2.678977065890062e-06, + "loss": 0.4577, + "step": 5929 + }, + { + "epoch": 1.08, + "learning_rate": 2.6789224608664e-06, + "loss": 2.9441, + "step": 5930 + }, + { + "epoch": 1.08, + "learning_rate": 2.6788678558427375e-06, + "loss": 1.5135, + "step": 5931 + }, + { + "epoch": 1.08, + "learning_rate": 2.6788132508190754e-06, + "loss": 1.0801, + "step": 5932 + }, + { + "epoch": 1.08, + "learning_rate": 2.6787586457954134e-06, + "loss": 0.7669, + "step": 5933 + }, + { + "epoch": 1.08, + "learning_rate": 2.678704040771751e-06, + "loss": 0.1851, + "step": 5934 + }, + { + "epoch": 1.08, + "learning_rate": 2.678649435748089e-06, + "loss": 1.5414, + "step": 5935 + }, + { + "epoch": 1.08, + "learning_rate": 2.6785948307244264e-06, + "loss": 0.4505, + "step": 5936 + }, + { + "epoch": 1.08, + "learning_rate": 2.6785402257007643e-06, + "loss": 1.0774, + "step": 5937 + }, + { + "epoch": 1.08, + "learning_rate": 2.6784856206771022e-06, + "loss": 1.3924, + "step": 5938 + }, + { + "epoch": 1.08, + "learning_rate": 2.67843101565344e-06, + "loss": 0.5958, + "step": 5939 + }, + { + "epoch": 1.08, + "learning_rate": 2.678376410629778e-06, + "loss": 1.0952, + "step": 5940 + }, + { + "epoch": 1.08, + "learning_rate": 2.678321805606116e-06, + "loss": 0.7583, + "step": 5941 + }, + { + "epoch": 1.08, + "learning_rate": 2.6782672005824536e-06, + "loss": 0.2389, + "step": 5942 + }, + { + "epoch": 1.08, + "learning_rate": 2.6782125955587915e-06, + "loss": 1.0255, + "step": 5943 + }, + { + "epoch": 1.08, + "learning_rate": 2.6781579905351295e-06, + "loss": 0.714, + "step": 5944 + }, + { + "epoch": 1.08, + "learning_rate": 2.678103385511467e-06, + "loss": 0.2763, + "step": 5945 + }, + { + "epoch": 1.08, + "learning_rate": 2.678048780487805e-06, + "loss": 0.3226, + "step": 5946 + }, + { + "epoch": 1.08, + "learning_rate": 2.677994175464143e-06, + "loss": 1.3523, + "step": 5947 + }, + { + "epoch": 1.08, + "learning_rate": 2.6779395704404804e-06, + "loss": 0.5516, + "step": 5948 + }, + { + "epoch": 1.08, + "learning_rate": 2.6778849654168184e-06, + "loss": 0.9008, + "step": 5949 + }, + { + "epoch": 1.08, + "learning_rate": 2.6778303603931563e-06, + "loss": 0.7671, + "step": 5950 + }, + { + "epoch": 1.08, + "learning_rate": 2.6777757553694942e-06, + "loss": 1.1204, + "step": 5951 + }, + { + "epoch": 1.08, + "learning_rate": 2.677721150345832e-06, + "loss": 0.8438, + "step": 5952 + }, + { + "epoch": 1.08, + "learning_rate": 2.6776665453221697e-06, + "loss": 0.5373, + "step": 5953 + }, + { + "epoch": 1.08, + "learning_rate": 2.6776119402985076e-06, + "loss": 1.0305, + "step": 5954 + }, + { + "epoch": 1.08, + "learning_rate": 2.6775573352748456e-06, + "loss": 0.6953, + "step": 5955 + }, + { + "epoch": 1.08, + "learning_rate": 2.677502730251183e-06, + "loss": 1.4934, + "step": 5956 + }, + { + "epoch": 1.08, + "learning_rate": 2.677448125227521e-06, + "loss": 0.0745, + "step": 5957 + }, + { + "epoch": 1.08, + "learning_rate": 2.677393520203859e-06, + "loss": 2.8017, + "step": 5958 + }, + { + "epoch": 1.08, + "learning_rate": 2.6773389151801965e-06, + "loss": 3.0558, + "step": 5959 + }, + { + "epoch": 1.08, + "learning_rate": 2.6772843101565345e-06, + "loss": 1.1737, + "step": 5960 + }, + { + "epoch": 1.08, + "learning_rate": 2.677229705132872e-06, + "loss": 1.493, + "step": 5961 + }, + { + "epoch": 1.08, + "learning_rate": 2.67717510010921e-06, + "loss": 1.6176, + "step": 5962 + }, + { + "epoch": 1.08, + "learning_rate": 2.677120495085548e-06, + "loss": 0.8734, + "step": 5963 + }, + { + "epoch": 1.08, + "learning_rate": 2.677065890061886e-06, + "loss": 0.8097, + "step": 5964 + }, + { + "epoch": 1.08, + "learning_rate": 2.6770112850382238e-06, + "loss": 0.2222, + "step": 5965 + }, + { + "epoch": 1.08, + "learning_rate": 2.6769566800145617e-06, + "loss": 1.0668, + "step": 5966 + }, + { + "epoch": 1.09, + "learning_rate": 2.6769020749908992e-06, + "loss": 3.1608, + "step": 5967 + }, + { + "epoch": 1.09, + "learning_rate": 2.676847469967237e-06, + "loss": 0.4447, + "step": 5968 + }, + { + "epoch": 1.09, + "learning_rate": 2.676792864943575e-06, + "loss": 0.708, + "step": 5969 + }, + { + "epoch": 1.09, + "learning_rate": 2.6767382599199126e-06, + "loss": 0.0417, + "step": 5970 + }, + { + "epoch": 1.09, + "learning_rate": 2.6766836548962506e-06, + "loss": 2.2365, + "step": 5971 + }, + { + "epoch": 1.09, + "learning_rate": 2.6766290498725885e-06, + "loss": 0.1997, + "step": 5972 + }, + { + "epoch": 1.09, + "learning_rate": 2.676574444848926e-06, + "loss": 0.5431, + "step": 5973 + }, + { + "epoch": 1.09, + "learning_rate": 2.676519839825264e-06, + "loss": 0.4542, + "step": 5974 + }, + { + "epoch": 1.09, + "learning_rate": 2.6764652348016015e-06, + "loss": 2.5869, + "step": 5975 + }, + { + "epoch": 1.09, + "learning_rate": 2.6764106297779394e-06, + "loss": 0.2811, + "step": 5976 + }, + { + "epoch": 1.09, + "learning_rate": 2.6763560247542774e-06, + "loss": 0.0761, + "step": 5977 + }, + { + "epoch": 1.09, + "learning_rate": 2.6763014197306153e-06, + "loss": 1.6327, + "step": 5978 + }, + { + "epoch": 1.09, + "learning_rate": 2.6762468147069533e-06, + "loss": 2.1674, + "step": 5979 + }, + { + "epoch": 1.09, + "learning_rate": 2.6761922096832912e-06, + "loss": 0.5789, + "step": 5980 + }, + { + "epoch": 1.09, + "learning_rate": 2.6761376046596287e-06, + "loss": 1.6526, + "step": 5981 + }, + { + "epoch": 1.09, + "learning_rate": 2.6760829996359667e-06, + "loss": 1.182, + "step": 5982 + }, + { + "epoch": 1.09, + "learning_rate": 2.6760283946123046e-06, + "loss": 3.8619, + "step": 5983 + }, + { + "epoch": 1.09, + "learning_rate": 2.675973789588642e-06, + "loss": 0.6321, + "step": 5984 + }, + { + "epoch": 1.09, + "learning_rate": 2.67591918456498e-06, + "loss": 0.0752, + "step": 5985 + }, + { + "epoch": 1.09, + "learning_rate": 2.675864579541318e-06, + "loss": 1.1836, + "step": 5986 + }, + { + "epoch": 1.09, + "learning_rate": 2.6758099745176555e-06, + "loss": 4.3227, + "step": 5987 + }, + { + "epoch": 1.09, + "learning_rate": 2.6757553694939935e-06, + "loss": 1.7241, + "step": 5988 + }, + { + "epoch": 1.09, + "learning_rate": 2.675700764470331e-06, + "loss": 0.073, + "step": 5989 + }, + { + "epoch": 1.09, + "learning_rate": 2.675646159446669e-06, + "loss": 2.048, + "step": 5990 + }, + { + "epoch": 1.09, + "learning_rate": 2.675591554423007e-06, + "loss": 1.5305, + "step": 5991 + }, + { + "epoch": 1.09, + "learning_rate": 2.675536949399345e-06, + "loss": 0.0579, + "step": 5992 + }, + { + "epoch": 1.09, + "learning_rate": 2.6754823443756828e-06, + "loss": 1.3969, + "step": 5993 + }, + { + "epoch": 1.09, + "learning_rate": 2.6754277393520207e-06, + "loss": 1.5894, + "step": 5994 + }, + { + "epoch": 1.09, + "learning_rate": 2.6753731343283582e-06, + "loss": 0.1326, + "step": 5995 + }, + { + "epoch": 1.09, + "learning_rate": 2.675318529304696e-06, + "loss": 0.5013, + "step": 5996 + }, + { + "epoch": 1.09, + "learning_rate": 2.675263924281034e-06, + "loss": 0.6537, + "step": 5997 + }, + { + "epoch": 1.09, + "learning_rate": 2.6752093192573717e-06, + "loss": 3.3629, + "step": 5998 + }, + { + "epoch": 1.09, + "learning_rate": 2.6751547142337096e-06, + "loss": 0.4751, + "step": 5999 + }, + { + "epoch": 1.09, + "learning_rate": 2.6751001092100475e-06, + "loss": 0.6976, + "step": 6000 + }, + { + "epoch": 1.09, + "learning_rate": 2.675045504186385e-06, + "loss": 0.2597, + "step": 6001 + }, + { + "epoch": 1.09, + "learning_rate": 2.674990899162723e-06, + "loss": 1.8621, + "step": 6002 + }, + { + "epoch": 1.09, + "learning_rate": 2.6749362941390605e-06, + "loss": 1.0689, + "step": 6003 + }, + { + "epoch": 1.09, + "learning_rate": 2.6748816891153985e-06, + "loss": 1.8465, + "step": 6004 + }, + { + "epoch": 1.09, + "learning_rate": 2.6748270840917364e-06, + "loss": 0.8958, + "step": 6005 + }, + { + "epoch": 1.09, + "learning_rate": 2.6747724790680744e-06, + "loss": 1.3403, + "step": 6006 + }, + { + "epoch": 1.09, + "learning_rate": 2.6747178740444123e-06, + "loss": 3.2667, + "step": 6007 + }, + { + "epoch": 1.09, + "learning_rate": 2.6746632690207502e-06, + "loss": 0.0283, + "step": 6008 + }, + { + "epoch": 1.09, + "learning_rate": 2.6746086639970878e-06, + "loss": 0.2424, + "step": 6009 + }, + { + "epoch": 1.09, + "learning_rate": 2.6745540589734257e-06, + "loss": 0.4703, + "step": 6010 + }, + { + "epoch": 1.09, + "learning_rate": 2.6744994539497636e-06, + "loss": 1.5251, + "step": 6011 + }, + { + "epoch": 1.09, + "learning_rate": 2.674444848926101e-06, + "loss": 1.1914, + "step": 6012 + }, + { + "epoch": 1.09, + "learning_rate": 2.674390243902439e-06, + "loss": 1.1176, + "step": 6013 + }, + { + "epoch": 1.09, + "learning_rate": 2.674335638878777e-06, + "loss": 1.0355, + "step": 6014 + }, + { + "epoch": 1.09, + "learning_rate": 2.6742810338551146e-06, + "loss": 0.0905, + "step": 6015 + }, + { + "epoch": 1.09, + "learning_rate": 2.6742264288314525e-06, + "loss": 0.0933, + "step": 6016 + }, + { + "epoch": 1.09, + "learning_rate": 2.67417182380779e-06, + "loss": 1.2792, + "step": 6017 + }, + { + "epoch": 1.09, + "learning_rate": 2.674117218784128e-06, + "loss": 0.8697, + "step": 6018 + }, + { + "epoch": 1.09, + "learning_rate": 2.674062613760466e-06, + "loss": 0.0743, + "step": 6019 + }, + { + "epoch": 1.09, + "learning_rate": 2.674008008736804e-06, + "loss": 1.4394, + "step": 6020 + }, + { + "epoch": 1.09, + "learning_rate": 2.673953403713142e-06, + "loss": 0.5693, + "step": 6021 + }, + { + "epoch": 1.1, + "learning_rate": 2.6738987986894798e-06, + "loss": 0.7072, + "step": 6022 + }, + { + "epoch": 1.1, + "learning_rate": 2.6738441936658173e-06, + "loss": 1.0942, + "step": 6023 + }, + { + "epoch": 1.1, + "learning_rate": 2.6737895886421552e-06, + "loss": 0.3531, + "step": 6024 + }, + { + "epoch": 1.1, + "learning_rate": 2.673734983618493e-06, + "loss": 1.9691, + "step": 6025 + }, + { + "epoch": 1.1, + "learning_rate": 2.6736803785948307e-06, + "loss": 1.0951, + "step": 6026 + }, + { + "epoch": 1.1, + "learning_rate": 2.6736257735711686e-06, + "loss": 1.7471, + "step": 6027 + }, + { + "epoch": 1.1, + "learning_rate": 2.673571168547506e-06, + "loss": 1.0742, + "step": 6028 + }, + { + "epoch": 1.1, + "learning_rate": 2.673516563523844e-06, + "loss": 1.0366, + "step": 6029 + }, + { + "epoch": 1.1, + "learning_rate": 2.673461958500182e-06, + "loss": 1.7055, + "step": 6030 + }, + { + "epoch": 1.1, + "learning_rate": 2.67340735347652e-06, + "loss": 0.7777, + "step": 6031 + }, + { + "epoch": 1.1, + "learning_rate": 2.673352748452858e-06, + "loss": 2.6177, + "step": 6032 + }, + { + "epoch": 1.1, + "learning_rate": 2.673298143429196e-06, + "loss": 0.6969, + "step": 6033 + }, + { + "epoch": 1.1, + "learning_rate": 2.6732435384055334e-06, + "loss": 1.1778, + "step": 6034 + }, + { + "epoch": 1.1, + "learning_rate": 2.6731889333818713e-06, + "loss": 1.7681, + "step": 6035 + }, + { + "epoch": 1.1, + "learning_rate": 2.6731343283582093e-06, + "loss": 0.9364, + "step": 6036 + }, + { + "epoch": 1.1, + "learning_rate": 2.6730797233345468e-06, + "loss": 0.7174, + "step": 6037 + }, + { + "epoch": 1.1, + "learning_rate": 2.6730251183108847e-06, + "loss": 0.5343, + "step": 6038 + }, + { + "epoch": 1.1, + "learning_rate": 2.6729705132872227e-06, + "loss": 0.8926, + "step": 6039 + }, + { + "epoch": 1.1, + "learning_rate": 2.67291590826356e-06, + "loss": 1.1566, + "step": 6040 + }, + { + "epoch": 1.1, + "learning_rate": 2.672861303239898e-06, + "loss": 0.167, + "step": 6041 + }, + { + "epoch": 1.1, + "learning_rate": 2.6728066982162357e-06, + "loss": 1.9401, + "step": 6042 + }, + { + "epoch": 1.1, + "learning_rate": 2.6727520931925736e-06, + "loss": 0.8179, + "step": 6043 + }, + { + "epoch": 1.1, + "learning_rate": 2.6726974881689115e-06, + "loss": 0.6734, + "step": 6044 + }, + { + "epoch": 1.1, + "learning_rate": 2.6726428831452495e-06, + "loss": 1.9854, + "step": 6045 + }, + { + "epoch": 1.1, + "learning_rate": 2.6725882781215874e-06, + "loss": 0.5315, + "step": 6046 + }, + { + "epoch": 1.1, + "learning_rate": 2.6725336730979254e-06, + "loss": 1.2428, + "step": 6047 + }, + { + "epoch": 1.1, + "learning_rate": 2.672479068074263e-06, + "loss": 0.8796, + "step": 6048 + }, + { + "epoch": 1.1, + "learning_rate": 2.672424463050601e-06, + "loss": 2.1297, + "step": 6049 + }, + { + "epoch": 1.1, + "learning_rate": 2.6723698580269388e-06, + "loss": 0.4232, + "step": 6050 + }, + { + "epoch": 1.1, + "learning_rate": 2.6723152530032763e-06, + "loss": 0.0326, + "step": 6051 + }, + { + "epoch": 1.1, + "learning_rate": 2.6722606479796142e-06, + "loss": 0.3046, + "step": 6052 + }, + { + "epoch": 1.1, + "learning_rate": 2.672206042955952e-06, + "loss": 1.7339, + "step": 6053 + }, + { + "epoch": 1.1, + "learning_rate": 2.6721514379322897e-06, + "loss": 2.0913, + "step": 6054 + }, + { + "epoch": 1.1, + "learning_rate": 2.6720968329086276e-06, + "loss": 1.1836, + "step": 6055 + }, + { + "epoch": 1.1, + "learning_rate": 2.672042227884965e-06, + "loss": 0.1462, + "step": 6056 + }, + { + "epoch": 1.1, + "learning_rate": 2.671987622861303e-06, + "loss": 2.1997, + "step": 6057 + }, + { + "epoch": 1.1, + "learning_rate": 2.671933017837641e-06, + "loss": 0.7044, + "step": 6058 + }, + { + "epoch": 1.1, + "learning_rate": 2.671878412813979e-06, + "loss": 1.3285, + "step": 6059 + }, + { + "epoch": 1.1, + "learning_rate": 2.671823807790317e-06, + "loss": 1.4543, + "step": 6060 + }, + { + "epoch": 1.1, + "learning_rate": 2.671769202766655e-06, + "loss": 1.8841, + "step": 6061 + }, + { + "epoch": 1.1, + "learning_rate": 2.6717145977429924e-06, + "loss": 0.1442, + "step": 6062 + }, + { + "epoch": 1.1, + "learning_rate": 2.6716599927193303e-06, + "loss": 1.7184, + "step": 6063 + }, + { + "epoch": 1.1, + "learning_rate": 2.6716053876956683e-06, + "loss": 0.0701, + "step": 6064 + }, + { + "epoch": 1.1, + "learning_rate": 2.671550782672006e-06, + "loss": 1.2325, + "step": 6065 + }, + { + "epoch": 1.1, + "learning_rate": 2.6714961776483438e-06, + "loss": 0.3273, + "step": 6066 + }, + { + "epoch": 1.1, + "learning_rate": 2.6714415726246817e-06, + "loss": 3.1252, + "step": 6067 + }, + { + "epoch": 1.1, + "learning_rate": 2.6713869676010192e-06, + "loss": 0.4762, + "step": 6068 + }, + { + "epoch": 1.1, + "learning_rate": 2.671332362577357e-06, + "loss": 0.3924, + "step": 6069 + }, + { + "epoch": 1.1, + "learning_rate": 2.6712777575536947e-06, + "loss": 1.7681, + "step": 6070 + }, + { + "epoch": 1.1, + "learning_rate": 2.6712231525300326e-06, + "loss": 2.0205, + "step": 6071 + }, + { + "epoch": 1.1, + "learning_rate": 2.6711685475063706e-06, + "loss": 1.661, + "step": 6072 + }, + { + "epoch": 1.1, + "learning_rate": 2.6711139424827085e-06, + "loss": 0.5007, + "step": 6073 + }, + { + "epoch": 1.1, + "learning_rate": 2.6710593374590465e-06, + "loss": 1.4263, + "step": 6074 + }, + { + "epoch": 1.1, + "learning_rate": 2.6710047324353844e-06, + "loss": 0.8803, + "step": 6075 + }, + { + "epoch": 1.1, + "learning_rate": 2.670950127411722e-06, + "loss": 0.3658, + "step": 6076 + }, + { + "epoch": 1.11, + "learning_rate": 2.67089552238806e-06, + "loss": 1.0284, + "step": 6077 + }, + { + "epoch": 1.11, + "learning_rate": 2.670840917364398e-06, + "loss": 0.1742, + "step": 6078 + }, + { + "epoch": 1.11, + "learning_rate": 2.6707863123407353e-06, + "loss": 0.6722, + "step": 6079 + }, + { + "epoch": 1.11, + "learning_rate": 2.6707317073170733e-06, + "loss": 0.0699, + "step": 6080 + }, + { + "epoch": 1.11, + "learning_rate": 2.6706771022934112e-06, + "loss": 1.1483, + "step": 6081 + }, + { + "epoch": 1.11, + "learning_rate": 2.6706224972697487e-06, + "loss": 0.0537, + "step": 6082 + }, + { + "epoch": 1.11, + "learning_rate": 2.6705678922460867e-06, + "loss": 1.0654, + "step": 6083 + }, + { + "epoch": 1.11, + "learning_rate": 2.670513287222424e-06, + "loss": 0.7791, + "step": 6084 + }, + { + "epoch": 1.11, + "learning_rate": 2.670458682198762e-06, + "loss": 0.6637, + "step": 6085 + }, + { + "epoch": 1.11, + "learning_rate": 2.6704040771751e-06, + "loss": 2.2176, + "step": 6086 + }, + { + "epoch": 1.11, + "learning_rate": 2.670349472151438e-06, + "loss": 0.5228, + "step": 6087 + }, + { + "epoch": 1.11, + "learning_rate": 2.670294867127776e-06, + "loss": 0.4237, + "step": 6088 + }, + { + "epoch": 1.11, + "learning_rate": 2.670240262104114e-06, + "loss": 0.7213, + "step": 6089 + }, + { + "epoch": 1.11, + "learning_rate": 2.6701856570804514e-06, + "loss": 0.9663, + "step": 6090 + }, + { + "epoch": 1.11, + "learning_rate": 2.6701310520567894e-06, + "loss": 0.1875, + "step": 6091 + }, + { + "epoch": 1.11, + "learning_rate": 2.6700764470331273e-06, + "loss": 3.6239, + "step": 6092 + }, + { + "epoch": 1.11, + "learning_rate": 2.670021842009465e-06, + "loss": 0.3969, + "step": 6093 + }, + { + "epoch": 1.11, + "learning_rate": 2.6699672369858028e-06, + "loss": 0.6122, + "step": 6094 + }, + { + "epoch": 1.11, + "learning_rate": 2.6699126319621407e-06, + "loss": 2.136, + "step": 6095 + }, + { + "epoch": 1.11, + "learning_rate": 2.6698580269384782e-06, + "loss": 0.5054, + "step": 6096 + }, + { + "epoch": 1.11, + "learning_rate": 2.669803421914816e-06, + "loss": 0.5103, + "step": 6097 + }, + { + "epoch": 1.11, + "learning_rate": 2.6697488168911537e-06, + "loss": 0.4067, + "step": 6098 + }, + { + "epoch": 1.11, + "learning_rate": 2.6696942118674917e-06, + "loss": 1.968, + "step": 6099 + }, + { + "epoch": 1.11, + "learning_rate": 2.6696396068438296e-06, + "loss": 0.6857, + "step": 6100 + }, + { + "epoch": 1.11, + "learning_rate": 2.6695850018201675e-06, + "loss": 1.2912, + "step": 6101 + }, + { + "epoch": 1.11, + "learning_rate": 2.6695303967965055e-06, + "loss": 1.8873, + "step": 6102 + }, + { + "epoch": 1.11, + "learning_rate": 2.6694757917728434e-06, + "loss": 0.1806, + "step": 6103 + }, + { + "epoch": 1.11, + "learning_rate": 2.669421186749181e-06, + "loss": 1.6603, + "step": 6104 + }, + { + "epoch": 1.11, + "learning_rate": 2.669366581725519e-06, + "loss": 0.6961, + "step": 6105 + }, + { + "epoch": 1.11, + "learning_rate": 2.669311976701857e-06, + "loss": 1.2574, + "step": 6106 + }, + { + "epoch": 1.11, + "learning_rate": 2.6692573716781944e-06, + "loss": 0.2534, + "step": 6107 + }, + { + "epoch": 1.11, + "learning_rate": 2.6692027666545323e-06, + "loss": 0.3002, + "step": 6108 + }, + { + "epoch": 1.11, + "learning_rate": 2.66914816163087e-06, + "loss": 0.5879, + "step": 6109 + }, + { + "epoch": 1.11, + "learning_rate": 2.6690935566072078e-06, + "loss": 0.1226, + "step": 6110 + }, + { + "epoch": 1.11, + "learning_rate": 2.6690389515835457e-06, + "loss": 0.7503, + "step": 6111 + }, + { + "epoch": 1.11, + "learning_rate": 2.6689843465598836e-06, + "loss": 0.6586, + "step": 6112 + }, + { + "epoch": 1.11, + "learning_rate": 2.6689297415362216e-06, + "loss": 0.67, + "step": 6113 + }, + { + "epoch": 1.11, + "learning_rate": 2.6688751365125595e-06, + "loss": 1.7843, + "step": 6114 + }, + { + "epoch": 1.11, + "learning_rate": 2.668820531488897e-06, + "loss": 1.4911, + "step": 6115 + }, + { + "epoch": 1.11, + "learning_rate": 2.668765926465235e-06, + "loss": 1.456, + "step": 6116 + }, + { + "epoch": 1.11, + "learning_rate": 2.668711321441573e-06, + "loss": 1.8767, + "step": 6117 + }, + { + "epoch": 1.11, + "learning_rate": 2.6686567164179105e-06, + "loss": 0.023, + "step": 6118 + }, + { + "epoch": 1.11, + "learning_rate": 2.6686021113942484e-06, + "loss": 1.028, + "step": 6119 + }, + { + "epoch": 1.11, + "learning_rate": 2.6685475063705863e-06, + "loss": 0.4493, + "step": 6120 + }, + { + "epoch": 1.11, + "learning_rate": 2.668492901346924e-06, + "loss": 0.1848, + "step": 6121 + }, + { + "epoch": 1.11, + "learning_rate": 2.668438296323262e-06, + "loss": 1.0288, + "step": 6122 + }, + { + "epoch": 1.11, + "learning_rate": 2.6683836912995993e-06, + "loss": 1.0806, + "step": 6123 + }, + { + "epoch": 1.11, + "learning_rate": 2.6683290862759373e-06, + "loss": 2.37, + "step": 6124 + }, + { + "epoch": 1.11, + "learning_rate": 2.6682744812522752e-06, + "loss": 1.1367, + "step": 6125 + }, + { + "epoch": 1.11, + "learning_rate": 2.668219876228613e-06, + "loss": 1.7931, + "step": 6126 + }, + { + "epoch": 1.11, + "learning_rate": 2.668165271204951e-06, + "loss": 1.2887, + "step": 6127 + }, + { + "epoch": 1.11, + "learning_rate": 2.668110666181289e-06, + "loss": 1.2552, + "step": 6128 + }, + { + "epoch": 1.11, + "learning_rate": 2.6680560611576266e-06, + "loss": 0.1089, + "step": 6129 + }, + { + "epoch": 1.11, + "learning_rate": 2.6680014561339645e-06, + "loss": 0.9918, + "step": 6130 + }, + { + "epoch": 1.11, + "learning_rate": 2.6679468511103025e-06, + "loss": 0.9817, + "step": 6131 + }, + { + "epoch": 1.12, + "learning_rate": 2.66789224608664e-06, + "loss": 0.4392, + "step": 6132 + }, + { + "epoch": 1.12, + "learning_rate": 2.667837641062978e-06, + "loss": 2.0594, + "step": 6133 + }, + { + "epoch": 1.12, + "learning_rate": 2.667783036039316e-06, + "loss": 0.7997, + "step": 6134 + }, + { + "epoch": 1.12, + "learning_rate": 2.6677284310156534e-06, + "loss": 1.5751, + "step": 6135 + }, + { + "epoch": 1.12, + "learning_rate": 2.6676738259919913e-06, + "loss": 2.3204, + "step": 6136 + }, + { + "epoch": 1.12, + "learning_rate": 2.667619220968329e-06, + "loss": 0.6401, + "step": 6137 + }, + { + "epoch": 1.12, + "learning_rate": 2.6675646159446668e-06, + "loss": 1.1093, + "step": 6138 + }, + { + "epoch": 1.12, + "learning_rate": 2.6675100109210047e-06, + "loss": 0.5232, + "step": 6139 + }, + { + "epoch": 1.12, + "learning_rate": 2.6674554058973427e-06, + "loss": 0.8841, + "step": 6140 + }, + { + "epoch": 1.12, + "learning_rate": 2.6674008008736806e-06, + "loss": 3.336, + "step": 6141 + }, + { + "epoch": 1.12, + "learning_rate": 2.6673461958500186e-06, + "loss": 1.1558, + "step": 6142 + }, + { + "epoch": 1.12, + "learning_rate": 2.667291590826356e-06, + "loss": 1.393, + "step": 6143 + }, + { + "epoch": 1.12, + "learning_rate": 2.667236985802694e-06, + "loss": 3.0118, + "step": 6144 + }, + { + "epoch": 1.12, + "learning_rate": 2.667182380779032e-06, + "loss": 0.6152, + "step": 6145 + }, + { + "epoch": 1.12, + "learning_rate": 2.6671277757553695e-06, + "loss": 0.2659, + "step": 6146 + }, + { + "epoch": 1.12, + "learning_rate": 2.6670731707317074e-06, + "loss": 1.9167, + "step": 6147 + }, + { + "epoch": 1.12, + "learning_rate": 2.6670185657080454e-06, + "loss": 0.4994, + "step": 6148 + }, + { + "epoch": 1.12, + "learning_rate": 2.666963960684383e-06, + "loss": 1.3883, + "step": 6149 + }, + { + "epoch": 1.12, + "learning_rate": 2.666909355660721e-06, + "loss": 1.1107, + "step": 6150 + }, + { + "epoch": 1.12, + "learning_rate": 2.6668547506370584e-06, + "loss": 0.4127, + "step": 6151 + }, + { + "epoch": 1.12, + "learning_rate": 2.6668001456133963e-06, + "loss": 2.6726, + "step": 6152 + }, + { + "epoch": 1.12, + "learning_rate": 2.6667455405897342e-06, + "loss": 0.1799, + "step": 6153 + }, + { + "epoch": 1.12, + "learning_rate": 2.666690935566072e-06, + "loss": 0.3685, + "step": 6154 + }, + { + "epoch": 1.12, + "learning_rate": 2.66663633054241e-06, + "loss": 0.5131, + "step": 6155 + }, + { + "epoch": 1.12, + "learning_rate": 2.666581725518748e-06, + "loss": 1.2241, + "step": 6156 + }, + { + "epoch": 1.12, + "learning_rate": 2.6665271204950856e-06, + "loss": 1.4977, + "step": 6157 + }, + { + "epoch": 1.12, + "learning_rate": 2.6664725154714235e-06, + "loss": 0.4992, + "step": 6158 + }, + { + "epoch": 1.12, + "learning_rate": 2.6664179104477615e-06, + "loss": 1.478, + "step": 6159 + }, + { + "epoch": 1.12, + "learning_rate": 2.666363305424099e-06, + "loss": 0.4147, + "step": 6160 + }, + { + "epoch": 1.12, + "learning_rate": 2.666308700400437e-06, + "loss": 0.9971, + "step": 6161 + }, + { + "epoch": 1.12, + "learning_rate": 2.666254095376775e-06, + "loss": 0.7528, + "step": 6162 + }, + { + "epoch": 1.12, + "learning_rate": 2.6661994903531124e-06, + "loss": 2.3599, + "step": 6163 + }, + { + "epoch": 1.12, + "learning_rate": 2.6661448853294504e-06, + "loss": 2.169, + "step": 6164 + }, + { + "epoch": 1.12, + "learning_rate": 2.666090280305788e-06, + "loss": 0.0447, + "step": 6165 + }, + { + "epoch": 1.12, + "learning_rate": 2.666035675282126e-06, + "loss": 0.304, + "step": 6166 + }, + { + "epoch": 1.12, + "learning_rate": 2.6659810702584638e-06, + "loss": 0.3239, + "step": 6167 + }, + { + "epoch": 1.12, + "learning_rate": 2.6659264652348017e-06, + "loss": 2.7801, + "step": 6168 + }, + { + "epoch": 1.12, + "learning_rate": 2.6658718602111396e-06, + "loss": 0.625, + "step": 6169 + }, + { + "epoch": 1.12, + "learning_rate": 2.6658172551874776e-06, + "loss": 0.6995, + "step": 6170 + }, + { + "epoch": 1.12, + "learning_rate": 2.665762650163815e-06, + "loss": 1.2594, + "step": 6171 + }, + { + "epoch": 1.12, + "learning_rate": 2.665708045140153e-06, + "loss": 0.8824, + "step": 6172 + }, + { + "epoch": 1.12, + "learning_rate": 2.665653440116491e-06, + "loss": 0.9063, + "step": 6173 + }, + { + "epoch": 1.12, + "learning_rate": 2.6655988350928285e-06, + "loss": 2.2162, + "step": 6174 + }, + { + "epoch": 1.12, + "learning_rate": 2.6655442300691665e-06, + "loss": 0.7044, + "step": 6175 + }, + { + "epoch": 1.12, + "learning_rate": 2.6654896250455044e-06, + "loss": 0.3918, + "step": 6176 + }, + { + "epoch": 1.12, + "learning_rate": 2.665435020021842e-06, + "loss": 0.2136, + "step": 6177 + }, + { + "epoch": 1.12, + "learning_rate": 2.66538041499818e-06, + "loss": 0.5029, + "step": 6178 + }, + { + "epoch": 1.12, + "learning_rate": 2.6653258099745174e-06, + "loss": 1.58, + "step": 6179 + }, + { + "epoch": 1.12, + "learning_rate": 2.6652712049508553e-06, + "loss": 2.3217, + "step": 6180 + }, + { + "epoch": 1.12, + "learning_rate": 2.6652165999271933e-06, + "loss": 0.8235, + "step": 6181 + }, + { + "epoch": 1.12, + "learning_rate": 2.6651619949035312e-06, + "loss": 2.5724, + "step": 6182 + }, + { + "epoch": 1.12, + "learning_rate": 2.665107389879869e-06, + "loss": 0.0645, + "step": 6183 + }, + { + "epoch": 1.12, + "learning_rate": 2.665052784856207e-06, + "loss": 0.352, + "step": 6184 + }, + { + "epoch": 1.12, + "learning_rate": 2.6649981798325446e-06, + "loss": 0.4608, + "step": 6185 + }, + { + "epoch": 1.12, + "learning_rate": 2.6649435748088826e-06, + "loss": 0.7409, + "step": 6186 + }, + { + "epoch": 1.13, + "learning_rate": 2.6648889697852205e-06, + "loss": 0.2028, + "step": 6187 + }, + { + "epoch": 1.13, + "learning_rate": 2.664834364761558e-06, + "loss": 0.2838, + "step": 6188 + }, + { + "epoch": 1.13, + "learning_rate": 2.664779759737896e-06, + "loss": 1.6816, + "step": 6189 + }, + { + "epoch": 1.13, + "learning_rate": 2.6647251547142335e-06, + "loss": 1.5512, + "step": 6190 + }, + { + "epoch": 1.13, + "learning_rate": 2.6646705496905714e-06, + "loss": 0.0192, + "step": 6191 + }, + { + "epoch": 1.13, + "learning_rate": 2.6646159446669094e-06, + "loss": 0.9766, + "step": 6192 + }, + { + "epoch": 1.13, + "learning_rate": 2.664561339643247e-06, + "loss": 0.2425, + "step": 6193 + }, + { + "epoch": 1.13, + "learning_rate": 2.664506734619585e-06, + "loss": 1.5139, + "step": 6194 + }, + { + "epoch": 1.13, + "learning_rate": 2.6644521295959228e-06, + "loss": 1.0781, + "step": 6195 + }, + { + "epoch": 1.13, + "learning_rate": 2.6643975245722607e-06, + "loss": 0.5763, + "step": 6196 + }, + { + "epoch": 1.13, + "learning_rate": 2.6643429195485987e-06, + "loss": 1.1733, + "step": 6197 + }, + { + "epoch": 1.13, + "learning_rate": 2.6642883145249366e-06, + "loss": 0.1413, + "step": 6198 + }, + { + "epoch": 1.13, + "learning_rate": 2.664233709501274e-06, + "loss": 0.2234, + "step": 6199 + }, + { + "epoch": 1.13, + "learning_rate": 2.664179104477612e-06, + "loss": 0.2256, + "step": 6200 + }, + { + "epoch": 1.13, + "learning_rate": 2.66412449945395e-06, + "loss": 0.8018, + "step": 6201 + }, + { + "epoch": 1.13, + "learning_rate": 2.6640698944302875e-06, + "loss": 0.0728, + "step": 6202 + }, + { + "epoch": 1.13, + "learning_rate": 2.6640152894066255e-06, + "loss": 2.1261, + "step": 6203 + }, + { + "epoch": 1.13, + "learning_rate": 2.663960684382963e-06, + "loss": 0.0602, + "step": 6204 + }, + { + "epoch": 1.13, + "learning_rate": 2.663906079359301e-06, + "loss": 0.0384, + "step": 6205 + }, + { + "epoch": 1.13, + "learning_rate": 2.663851474335639e-06, + "loss": 1.3076, + "step": 6206 + }, + { + "epoch": 1.13, + "learning_rate": 2.663796869311977e-06, + "loss": 0.3071, + "step": 6207 + }, + { + "epoch": 1.13, + "learning_rate": 2.6637422642883148e-06, + "loss": 0.9721, + "step": 6208 + }, + { + "epoch": 1.13, + "learning_rate": 2.6636876592646527e-06, + "loss": 0.4075, + "step": 6209 + }, + { + "epoch": 1.13, + "learning_rate": 2.6636330542409902e-06, + "loss": 1.1133, + "step": 6210 + }, + { + "epoch": 1.13, + "learning_rate": 2.663578449217328e-06, + "loss": 0.6795, + "step": 6211 + }, + { + "epoch": 1.13, + "learning_rate": 2.663523844193666e-06, + "loss": 3.929, + "step": 6212 + }, + { + "epoch": 1.13, + "learning_rate": 2.6634692391700036e-06, + "loss": 0.3189, + "step": 6213 + }, + { + "epoch": 1.13, + "learning_rate": 2.6634146341463416e-06, + "loss": 1.5985, + "step": 6214 + }, + { + "epoch": 1.13, + "learning_rate": 2.6633600291226795e-06, + "loss": 0.9287, + "step": 6215 + }, + { + "epoch": 1.13, + "learning_rate": 2.663305424099017e-06, + "loss": 0.8988, + "step": 6216 + }, + { + "epoch": 1.13, + "learning_rate": 2.663250819075355e-06, + "loss": 0.4291, + "step": 6217 + }, + { + "epoch": 1.13, + "learning_rate": 2.6631962140516925e-06, + "loss": 0.0934, + "step": 6218 + }, + { + "epoch": 1.13, + "learning_rate": 2.6631416090280305e-06, + "loss": 4.7532, + "step": 6219 + }, + { + "epoch": 1.13, + "learning_rate": 2.6630870040043684e-06, + "loss": 0.9306, + "step": 6220 + }, + { + "epoch": 1.13, + "learning_rate": 2.6630323989807063e-06, + "loss": 0.0891, + "step": 6221 + }, + { + "epoch": 1.13, + "learning_rate": 2.6629777939570443e-06, + "loss": 1.0282, + "step": 6222 + }, + { + "epoch": 1.13, + "learning_rate": 2.6629231889333822e-06, + "loss": 3.42, + "step": 6223 + }, + { + "epoch": 1.13, + "learning_rate": 2.6628685839097198e-06, + "loss": 1.4373, + "step": 6224 + }, + { + "epoch": 1.13, + "learning_rate": 2.6628139788860577e-06, + "loss": 0.9342, + "step": 6225 + }, + { + "epoch": 1.13, + "learning_rate": 2.6627593738623956e-06, + "loss": 0.7861, + "step": 6226 + }, + { + "epoch": 1.13, + "learning_rate": 2.662704768838733e-06, + "loss": 0.6058, + "step": 6227 + }, + { + "epoch": 1.13, + "learning_rate": 2.662650163815071e-06, + "loss": 0.2147, + "step": 6228 + }, + { + "epoch": 1.13, + "learning_rate": 2.662595558791409e-06, + "loss": 0.7961, + "step": 6229 + }, + { + "epoch": 1.13, + "learning_rate": 2.6625409537677466e-06, + "loss": 1.9168, + "step": 6230 + }, + { + "epoch": 1.13, + "learning_rate": 2.6624863487440845e-06, + "loss": 1.3946, + "step": 6231 + }, + { + "epoch": 1.13, + "learning_rate": 2.662431743720422e-06, + "loss": 1.1524, + "step": 6232 + }, + { + "epoch": 1.13, + "learning_rate": 2.66237713869676e-06, + "loss": 0.4852, + "step": 6233 + }, + { + "epoch": 1.13, + "learning_rate": 2.662322533673098e-06, + "loss": 1.2874, + "step": 6234 + }, + { + "epoch": 1.13, + "learning_rate": 2.662267928649436e-06, + "loss": 0.7275, + "step": 6235 + }, + { + "epoch": 1.13, + "learning_rate": 2.662213323625774e-06, + "loss": 0.6526, + "step": 6236 + }, + { + "epoch": 1.13, + "learning_rate": 2.6621587186021117e-06, + "loss": 1.0522, + "step": 6237 + }, + { + "epoch": 1.13, + "learning_rate": 2.6621041135784493e-06, + "loss": 0.6347, + "step": 6238 + }, + { + "epoch": 1.13, + "learning_rate": 2.662049508554787e-06, + "loss": 1.6983, + "step": 6239 + }, + { + "epoch": 1.13, + "learning_rate": 2.661994903531125e-06, + "loss": 0.6245, + "step": 6240 + }, + { + "epoch": 1.13, + "learning_rate": 2.6619402985074627e-06, + "loss": 2.2336, + "step": 6241 + }, + { + "epoch": 1.14, + "learning_rate": 2.6618856934838006e-06, + "loss": 1.1422, + "step": 6242 + }, + { + "epoch": 1.14, + "learning_rate": 2.6618310884601386e-06, + "loss": 0.3106, + "step": 6243 + }, + { + "epoch": 1.14, + "learning_rate": 2.661776483436476e-06, + "loss": 0.0662, + "step": 6244 + }, + { + "epoch": 1.14, + "learning_rate": 2.661721878412814e-06, + "loss": 0.5016, + "step": 6245 + }, + { + "epoch": 1.14, + "learning_rate": 2.6616672733891515e-06, + "loss": 1.5353, + "step": 6246 + }, + { + "epoch": 1.14, + "learning_rate": 2.6616126683654895e-06, + "loss": 0.9787, + "step": 6247 + }, + { + "epoch": 1.14, + "learning_rate": 2.6615580633418274e-06, + "loss": 0.0673, + "step": 6248 + }, + { + "epoch": 1.14, + "learning_rate": 2.6615034583181654e-06, + "loss": 0.6048, + "step": 6249 + }, + { + "epoch": 1.14, + "learning_rate": 2.6614488532945033e-06, + "loss": 0.3817, + "step": 6250 + }, + { + "epoch": 1.14, + "learning_rate": 2.6613942482708413e-06, + "loss": 0.277, + "step": 6251 + }, + { + "epoch": 1.14, + "learning_rate": 2.6613396432471788e-06, + "loss": 0.9058, + "step": 6252 + }, + { + "epoch": 1.14, + "learning_rate": 2.6612850382235167e-06, + "loss": 1.6291, + "step": 6253 + }, + { + "epoch": 1.14, + "learning_rate": 2.6612304331998547e-06, + "loss": 0.782, + "step": 6254 + }, + { + "epoch": 1.14, + "learning_rate": 2.661175828176192e-06, + "loss": 0.5715, + "step": 6255 + }, + { + "epoch": 1.14, + "learning_rate": 2.66112122315253e-06, + "loss": 1.2754, + "step": 6256 + }, + { + "epoch": 1.14, + "learning_rate": 2.661066618128868e-06, + "loss": 0.8943, + "step": 6257 + }, + { + "epoch": 1.14, + "learning_rate": 2.6610120131052056e-06, + "loss": 0.0845, + "step": 6258 + }, + { + "epoch": 1.14, + "learning_rate": 2.6609574080815435e-06, + "loss": 1.1864, + "step": 6259 + }, + { + "epoch": 1.14, + "learning_rate": 2.660902803057881e-06, + "loss": 1.0003, + "step": 6260 + }, + { + "epoch": 1.14, + "learning_rate": 2.660848198034219e-06, + "loss": 0.4768, + "step": 6261 + }, + { + "epoch": 1.14, + "learning_rate": 2.660793593010557e-06, + "loss": 1.0676, + "step": 6262 + }, + { + "epoch": 1.14, + "learning_rate": 2.660738987986895e-06, + "loss": 2.1294, + "step": 6263 + }, + { + "epoch": 1.14, + "learning_rate": 2.660684382963233e-06, + "loss": 1.355, + "step": 6264 + }, + { + "epoch": 1.14, + "learning_rate": 2.6606297779395708e-06, + "loss": 0.8383, + "step": 6265 + }, + { + "epoch": 1.14, + "learning_rate": 2.6605751729159083e-06, + "loss": 2.5238, + "step": 6266 + }, + { + "epoch": 1.14, + "learning_rate": 2.6605205678922462e-06, + "loss": 0.3292, + "step": 6267 + }, + { + "epoch": 1.14, + "learning_rate": 2.660465962868584e-06, + "loss": 0.5999, + "step": 6268 + }, + { + "epoch": 1.14, + "learning_rate": 2.6604113578449217e-06, + "loss": 1.1345, + "step": 6269 + }, + { + "epoch": 1.14, + "learning_rate": 2.6603567528212596e-06, + "loss": 2.8517, + "step": 6270 + }, + { + "epoch": 1.14, + "learning_rate": 2.660302147797597e-06, + "loss": 0.6915, + "step": 6271 + }, + { + "epoch": 1.14, + "learning_rate": 2.660247542773935e-06, + "loss": 0.1996, + "step": 6272 + }, + { + "epoch": 1.14, + "learning_rate": 2.660192937750273e-06, + "loss": 0.5252, + "step": 6273 + }, + { + "epoch": 1.14, + "learning_rate": 2.6601383327266106e-06, + "loss": 0.4319, + "step": 6274 + }, + { + "epoch": 1.14, + "learning_rate": 2.6600837277029485e-06, + "loss": 1.147, + "step": 6275 + }, + { + "epoch": 1.14, + "learning_rate": 2.6600291226792865e-06, + "loss": 2.1935, + "step": 6276 + }, + { + "epoch": 1.14, + "learning_rate": 2.6599745176556244e-06, + "loss": 0.1229, + "step": 6277 + }, + { + "epoch": 1.14, + "learning_rate": 2.6599199126319623e-06, + "loss": 1.1084, + "step": 6278 + }, + { + "epoch": 1.14, + "learning_rate": 2.6598653076083003e-06, + "loss": 0.9085, + "step": 6279 + }, + { + "epoch": 1.14, + "learning_rate": 2.659810702584638e-06, + "loss": 0.1048, + "step": 6280 + }, + { + "epoch": 1.14, + "learning_rate": 2.6597560975609758e-06, + "loss": 0.8633, + "step": 6281 + }, + { + "epoch": 1.14, + "learning_rate": 2.6597014925373137e-06, + "loss": 0.5208, + "step": 6282 + }, + { + "epoch": 1.14, + "learning_rate": 2.6596468875136512e-06, + "loss": 0.3683, + "step": 6283 + }, + { + "epoch": 1.14, + "learning_rate": 2.659592282489989e-06, + "loss": 1.5568, + "step": 6284 + }, + { + "epoch": 1.14, + "learning_rate": 2.6595376774663267e-06, + "loss": 1.2641, + "step": 6285 + }, + { + "epoch": 1.14, + "learning_rate": 2.6594830724426646e-06, + "loss": 0.4687, + "step": 6286 + }, + { + "epoch": 1.14, + "learning_rate": 2.6594284674190026e-06, + "loss": 1.1818, + "step": 6287 + }, + { + "epoch": 1.14, + "learning_rate": 2.6593738623953405e-06, + "loss": 0.6033, + "step": 6288 + }, + { + "epoch": 1.14, + "learning_rate": 2.6593192573716785e-06, + "loss": 0.693, + "step": 6289 + }, + { + "epoch": 1.14, + "learning_rate": 2.6592646523480164e-06, + "loss": 2.1094, + "step": 6290 + }, + { + "epoch": 1.14, + "learning_rate": 2.659210047324354e-06, + "loss": 0.897, + "step": 6291 + }, + { + "epoch": 1.14, + "learning_rate": 2.659155442300692e-06, + "loss": 3.2619, + "step": 6292 + }, + { + "epoch": 1.14, + "learning_rate": 2.65910083727703e-06, + "loss": 0.1138, + "step": 6293 + }, + { + "epoch": 1.14, + "learning_rate": 2.6590462322533673e-06, + "loss": 0.5281, + "step": 6294 + }, + { + "epoch": 1.14, + "learning_rate": 2.6589916272297053e-06, + "loss": 2.254, + "step": 6295 + }, + { + "epoch": 1.14, + "learning_rate": 2.658937022206043e-06, + "loss": 0.2142, + "step": 6296 + }, + { + "epoch": 1.15, + "learning_rate": 2.6588824171823807e-06, + "loss": 4.2777, + "step": 6297 + }, + { + "epoch": 1.15, + "learning_rate": 2.6588278121587187e-06, + "loss": 2.7956, + "step": 6298 + }, + { + "epoch": 1.15, + "learning_rate": 2.658773207135056e-06, + "loss": 0.3094, + "step": 6299 + }, + { + "epoch": 1.15, + "learning_rate": 2.658718602111394e-06, + "loss": 2.8009, + "step": 6300 + }, + { + "epoch": 1.15, + "learning_rate": 2.658663997087732e-06, + "loss": 1.3293, + "step": 6301 + }, + { + "epoch": 1.15, + "learning_rate": 2.65860939206407e-06, + "loss": 0.1724, + "step": 6302 + }, + { + "epoch": 1.15, + "learning_rate": 2.658554787040408e-06, + "loss": 0.9231, + "step": 6303 + }, + { + "epoch": 1.15, + "learning_rate": 2.658500182016746e-06, + "loss": 0.2742, + "step": 6304 + }, + { + "epoch": 1.15, + "learning_rate": 2.6584455769930834e-06, + "loss": 0.609, + "step": 6305 + }, + { + "epoch": 1.15, + "learning_rate": 2.6583909719694214e-06, + "loss": 3.0696, + "step": 6306 + }, + { + "epoch": 1.15, + "learning_rate": 2.6583363669457593e-06, + "loss": 1.3185, + "step": 6307 + }, + { + "epoch": 1.15, + "learning_rate": 2.658281761922097e-06, + "loss": 0.3897, + "step": 6308 + }, + { + "epoch": 1.15, + "learning_rate": 2.6582271568984348e-06, + "loss": 0.9469, + "step": 6309 + }, + { + "epoch": 1.15, + "learning_rate": 2.6581725518747727e-06, + "loss": 0.6912, + "step": 6310 + }, + { + "epoch": 1.15, + "learning_rate": 2.6581179468511102e-06, + "loss": 0.9973, + "step": 6311 + }, + { + "epoch": 1.15, + "learning_rate": 2.658063341827448e-06, + "loss": 0.1659, + "step": 6312 + }, + { + "epoch": 1.15, + "learning_rate": 2.6580087368037857e-06, + "loss": 0.6594, + "step": 6313 + }, + { + "epoch": 1.15, + "learning_rate": 2.6579541317801236e-06, + "loss": 4.2169, + "step": 6314 + }, + { + "epoch": 1.15, + "learning_rate": 2.6578995267564616e-06, + "loss": 1.7115, + "step": 6315 + }, + { + "epoch": 1.15, + "learning_rate": 2.6578449217327995e-06, + "loss": 0.5762, + "step": 6316 + }, + { + "epoch": 1.15, + "learning_rate": 2.6577903167091375e-06, + "loss": 0.347, + "step": 6317 + }, + { + "epoch": 1.15, + "learning_rate": 2.6577357116854754e-06, + "loss": 1.2274, + "step": 6318 + }, + { + "epoch": 1.15, + "learning_rate": 2.657681106661813e-06, + "loss": 1.1418, + "step": 6319 + }, + { + "epoch": 1.15, + "learning_rate": 2.657626501638151e-06, + "loss": 0.4779, + "step": 6320 + }, + { + "epoch": 1.15, + "learning_rate": 2.657571896614489e-06, + "loss": 2.2991, + "step": 6321 + }, + { + "epoch": 1.15, + "learning_rate": 2.6575172915908263e-06, + "loss": 1.0292, + "step": 6322 + }, + { + "epoch": 1.15, + "learning_rate": 2.6574626865671643e-06, + "loss": 0.8138, + "step": 6323 + }, + { + "epoch": 1.15, + "learning_rate": 2.6574080815435022e-06, + "loss": 0.5568, + "step": 6324 + }, + { + "epoch": 1.15, + "learning_rate": 2.6573534765198398e-06, + "loss": 2.1207, + "step": 6325 + }, + { + "epoch": 1.15, + "learning_rate": 2.6572988714961777e-06, + "loss": 1.1694, + "step": 6326 + }, + { + "epoch": 1.15, + "learning_rate": 2.6572442664725152e-06, + "loss": 1.3721, + "step": 6327 + }, + { + "epoch": 1.15, + "learning_rate": 2.657189661448853e-06, + "loss": 0.6627, + "step": 6328 + }, + { + "epoch": 1.15, + "learning_rate": 2.657135056425191e-06, + "loss": 0.9162, + "step": 6329 + }, + { + "epoch": 1.15, + "learning_rate": 2.657080451401529e-06, + "loss": 1.6214, + "step": 6330 + }, + { + "epoch": 1.15, + "learning_rate": 2.657025846377867e-06, + "loss": 0.2189, + "step": 6331 + }, + { + "epoch": 1.15, + "learning_rate": 2.656971241354205e-06, + "loss": 0.5445, + "step": 6332 + }, + { + "epoch": 1.15, + "learning_rate": 2.6569166363305425e-06, + "loss": 0.1672, + "step": 6333 + }, + { + "epoch": 1.15, + "learning_rate": 2.6568620313068804e-06, + "loss": 0.1273, + "step": 6334 + }, + { + "epoch": 1.15, + "learning_rate": 2.6568074262832183e-06, + "loss": 1.4687, + "step": 6335 + }, + { + "epoch": 1.15, + "learning_rate": 2.656752821259556e-06, + "loss": 2.6675, + "step": 6336 + }, + { + "epoch": 1.15, + "learning_rate": 2.656698216235894e-06, + "loss": 1.2242, + "step": 6337 + }, + { + "epoch": 1.15, + "learning_rate": 2.6566436112122313e-06, + "loss": 0.366, + "step": 6338 + }, + { + "epoch": 1.15, + "learning_rate": 2.6565890061885693e-06, + "loss": 0.5738, + "step": 6339 + }, + { + "epoch": 1.15, + "learning_rate": 2.6565344011649072e-06, + "loss": 1.7332, + "step": 6340 + }, + { + "epoch": 1.15, + "learning_rate": 2.6564797961412447e-06, + "loss": 0.791, + "step": 6341 + }, + { + "epoch": 1.15, + "learning_rate": 2.6564251911175827e-06, + "loss": 1.9926, + "step": 6342 + }, + { + "epoch": 1.15, + "learning_rate": 2.6563705860939206e-06, + "loss": 0.3101, + "step": 6343 + }, + { + "epoch": 1.15, + "learning_rate": 2.6563159810702586e-06, + "loss": 0.8318, + "step": 6344 + }, + { + "epoch": 1.15, + "learning_rate": 2.6562613760465965e-06, + "loss": 1.7235, + "step": 6345 + }, + { + "epoch": 1.15, + "learning_rate": 2.6562067710229344e-06, + "loss": 0.7759, + "step": 6346 + }, + { + "epoch": 1.15, + "learning_rate": 2.656152165999272e-06, + "loss": 1.5106, + "step": 6347 + }, + { + "epoch": 1.15, + "learning_rate": 2.65609756097561e-06, + "loss": 1.6458, + "step": 6348 + }, + { + "epoch": 1.15, + "learning_rate": 2.656042955951948e-06, + "loss": 1.0426, + "step": 6349 + }, + { + "epoch": 1.15, + "learning_rate": 2.6559883509282854e-06, + "loss": 0.086, + "step": 6350 + }, + { + "epoch": 1.15, + "learning_rate": 2.6559337459046233e-06, + "loss": 0.158, + "step": 6351 + }, + { + "epoch": 1.16, + "learning_rate": 2.655879140880961e-06, + "loss": 1.3076, + "step": 6352 + }, + { + "epoch": 1.16, + "learning_rate": 2.6558245358572988e-06, + "loss": 0.3541, + "step": 6353 + }, + { + "epoch": 1.16, + "learning_rate": 2.6557699308336367e-06, + "loss": 1.2871, + "step": 6354 + }, + { + "epoch": 1.16, + "learning_rate": 2.6557153258099742e-06, + "loss": 1.2501, + "step": 6355 + }, + { + "epoch": 1.16, + "learning_rate": 2.655660720786312e-06, + "loss": 0.8657, + "step": 6356 + }, + { + "epoch": 1.16, + "learning_rate": 2.65560611576265e-06, + "loss": 0.8876, + "step": 6357 + }, + { + "epoch": 1.16, + "learning_rate": 2.655551510738988e-06, + "loss": 1.2585, + "step": 6358 + }, + { + "epoch": 1.16, + "learning_rate": 2.655496905715326e-06, + "loss": 0.0661, + "step": 6359 + }, + { + "epoch": 1.16, + "learning_rate": 2.655442300691664e-06, + "loss": 0.1399, + "step": 6360 + }, + { + "epoch": 1.16, + "learning_rate": 2.6553876956680015e-06, + "loss": 2.3367, + "step": 6361 + }, + { + "epoch": 1.16, + "learning_rate": 2.6553330906443394e-06, + "loss": 1.4359, + "step": 6362 + }, + { + "epoch": 1.16, + "learning_rate": 2.6552784856206774e-06, + "loss": 2.8874, + "step": 6363 + }, + { + "epoch": 1.16, + "learning_rate": 2.655223880597015e-06, + "loss": 1.1807, + "step": 6364 + }, + { + "epoch": 1.16, + "learning_rate": 2.655169275573353e-06, + "loss": 0.7783, + "step": 6365 + }, + { + "epoch": 1.16, + "learning_rate": 2.6551146705496904e-06, + "loss": 1.2631, + "step": 6366 + }, + { + "epoch": 1.16, + "learning_rate": 2.6550600655260283e-06, + "loss": 0.7856, + "step": 6367 + }, + { + "epoch": 1.16, + "learning_rate": 2.6550054605023662e-06, + "loss": 1.7412, + "step": 6368 + }, + { + "epoch": 1.16, + "learning_rate": 2.654950855478704e-06, + "loss": 0.3353, + "step": 6369 + }, + { + "epoch": 1.16, + "learning_rate": 2.654896250455042e-06, + "loss": 0.9057, + "step": 6370 + }, + { + "epoch": 1.16, + "learning_rate": 2.65484164543138e-06, + "loss": 0.076, + "step": 6371 + }, + { + "epoch": 1.16, + "learning_rate": 2.6547870404077176e-06, + "loss": 1.2782, + "step": 6372 + }, + { + "epoch": 1.16, + "learning_rate": 2.6547324353840555e-06, + "loss": 0.7311, + "step": 6373 + }, + { + "epoch": 1.16, + "learning_rate": 2.6546778303603935e-06, + "loss": 0.2446, + "step": 6374 + }, + { + "epoch": 1.16, + "learning_rate": 2.654623225336731e-06, + "loss": 1.516, + "step": 6375 + }, + { + "epoch": 1.16, + "learning_rate": 2.654568620313069e-06, + "loss": 0.4573, + "step": 6376 + }, + { + "epoch": 1.16, + "learning_rate": 2.654514015289407e-06, + "loss": 0.368, + "step": 6377 + }, + { + "epoch": 1.16, + "learning_rate": 2.6544594102657444e-06, + "loss": 0.9933, + "step": 6378 + }, + { + "epoch": 1.16, + "learning_rate": 2.6544048052420823e-06, + "loss": 1.9926, + "step": 6379 + }, + { + "epoch": 1.16, + "learning_rate": 2.65435020021842e-06, + "loss": 0.5211, + "step": 6380 + }, + { + "epoch": 1.16, + "learning_rate": 2.654295595194758e-06, + "loss": 0.5418, + "step": 6381 + }, + { + "epoch": 1.16, + "learning_rate": 2.6542409901710958e-06, + "loss": 1.4598, + "step": 6382 + }, + { + "epoch": 1.16, + "learning_rate": 2.6541863851474337e-06, + "loss": 1.2196, + "step": 6383 + }, + { + "epoch": 1.16, + "learning_rate": 2.6541317801237716e-06, + "loss": 0.1833, + "step": 6384 + }, + { + "epoch": 1.16, + "learning_rate": 2.6540771751001096e-06, + "loss": 0.8726, + "step": 6385 + }, + { + "epoch": 1.16, + "learning_rate": 2.654022570076447e-06, + "loss": 0.4071, + "step": 6386 + }, + { + "epoch": 1.16, + "learning_rate": 2.653967965052785e-06, + "loss": 0.3791, + "step": 6387 + }, + { + "epoch": 1.16, + "learning_rate": 2.653913360029123e-06, + "loss": 2.6736, + "step": 6388 + }, + { + "epoch": 1.16, + "learning_rate": 2.6538587550054605e-06, + "loss": 2.9899, + "step": 6389 + }, + { + "epoch": 1.16, + "learning_rate": 2.6538041499817985e-06, + "loss": 0.2361, + "step": 6390 + }, + { + "epoch": 1.16, + "learning_rate": 2.6537495449581364e-06, + "loss": 0.2066, + "step": 6391 + }, + { + "epoch": 1.16, + "learning_rate": 2.653694939934474e-06, + "loss": 0.7687, + "step": 6392 + }, + { + "epoch": 1.16, + "learning_rate": 2.653640334910812e-06, + "loss": 2.1234, + "step": 6393 + }, + { + "epoch": 1.16, + "learning_rate": 2.6535857298871494e-06, + "loss": 1.2343, + "step": 6394 + }, + { + "epoch": 1.16, + "learning_rate": 2.6535311248634873e-06, + "loss": 1.3301, + "step": 6395 + }, + { + "epoch": 1.16, + "learning_rate": 2.6534765198398253e-06, + "loss": 1.7694, + "step": 6396 + }, + { + "epoch": 1.16, + "learning_rate": 2.653421914816163e-06, + "loss": 0.4044, + "step": 6397 + }, + { + "epoch": 1.16, + "learning_rate": 2.653367309792501e-06, + "loss": 0.4084, + "step": 6398 + }, + { + "epoch": 1.16, + "learning_rate": 2.653312704768839e-06, + "loss": 1.1725, + "step": 6399 + }, + { + "epoch": 1.16, + "learning_rate": 2.6532580997451766e-06, + "loss": 0.7793, + "step": 6400 + }, + { + "epoch": 1.16, + "learning_rate": 2.6532034947215146e-06, + "loss": 1.0539, + "step": 6401 + }, + { + "epoch": 1.16, + "learning_rate": 2.6531488896978525e-06, + "loss": 0.6308, + "step": 6402 + }, + { + "epoch": 1.16, + "learning_rate": 2.65309428467419e-06, + "loss": 1.3871, + "step": 6403 + }, + { + "epoch": 1.16, + "learning_rate": 2.653039679650528e-06, + "loss": 1.2546, + "step": 6404 + }, + { + "epoch": 1.16, + "learning_rate": 2.652985074626866e-06, + "loss": 0.0288, + "step": 6405 + }, + { + "epoch": 1.16, + "learning_rate": 2.6529304696032034e-06, + "loss": 1.34, + "step": 6406 + }, + { + "epoch": 1.17, + "learning_rate": 2.6528758645795414e-06, + "loss": 0.7517, + "step": 6407 + }, + { + "epoch": 1.17, + "learning_rate": 2.652821259555879e-06, + "loss": 0.5462, + "step": 6408 + }, + { + "epoch": 1.17, + "learning_rate": 2.652766654532217e-06, + "loss": 0.9834, + "step": 6409 + }, + { + "epoch": 1.17, + "learning_rate": 2.6527120495085548e-06, + "loss": 0.7394, + "step": 6410 + }, + { + "epoch": 1.17, + "learning_rate": 2.6526574444848927e-06, + "loss": 0.9086, + "step": 6411 + }, + { + "epoch": 1.17, + "learning_rate": 2.6526028394612307e-06, + "loss": 0.467, + "step": 6412 + }, + { + "epoch": 1.17, + "learning_rate": 2.6525482344375686e-06, + "loss": 0.0428, + "step": 6413 + }, + { + "epoch": 1.17, + "learning_rate": 2.652493629413906e-06, + "loss": 0.7659, + "step": 6414 + }, + { + "epoch": 1.17, + "learning_rate": 2.652439024390244e-06, + "loss": 0.9768, + "step": 6415 + }, + { + "epoch": 1.17, + "learning_rate": 2.652384419366582e-06, + "loss": 0.3188, + "step": 6416 + }, + { + "epoch": 1.17, + "learning_rate": 2.6523298143429195e-06, + "loss": 0.6352, + "step": 6417 + }, + { + "epoch": 1.17, + "learning_rate": 2.6522752093192575e-06, + "loss": 0.8673, + "step": 6418 + }, + { + "epoch": 1.17, + "learning_rate": 2.652220604295595e-06, + "loss": 1.319, + "step": 6419 + }, + { + "epoch": 1.17, + "learning_rate": 2.652165999271933e-06, + "loss": 2.3629, + "step": 6420 + }, + { + "epoch": 1.17, + "learning_rate": 2.652111394248271e-06, + "loss": 0.1893, + "step": 6421 + }, + { + "epoch": 1.17, + "learning_rate": 2.6520567892246084e-06, + "loss": 0.6408, + "step": 6422 + }, + { + "epoch": 1.17, + "learning_rate": 2.6520021842009463e-06, + "loss": 1.4144, + "step": 6423 + }, + { + "epoch": 1.17, + "learning_rate": 2.6519475791772843e-06, + "loss": 0.9192, + "step": 6424 + }, + { + "epoch": 1.17, + "learning_rate": 2.6518929741536222e-06, + "loss": 0.73, + "step": 6425 + }, + { + "epoch": 1.17, + "learning_rate": 2.65183836912996e-06, + "loss": 0.678, + "step": 6426 + }, + { + "epoch": 1.17, + "learning_rate": 2.651783764106298e-06, + "loss": 1.3899, + "step": 6427 + }, + { + "epoch": 1.17, + "learning_rate": 2.6517291590826356e-06, + "loss": 1.6634, + "step": 6428 + }, + { + "epoch": 1.17, + "learning_rate": 2.6516745540589736e-06, + "loss": 0.5641, + "step": 6429 + }, + { + "epoch": 1.17, + "learning_rate": 2.6516199490353115e-06, + "loss": 0.4358, + "step": 6430 + }, + { + "epoch": 1.17, + "learning_rate": 2.651565344011649e-06, + "loss": 1.7831, + "step": 6431 + }, + { + "epoch": 1.17, + "learning_rate": 2.651510738987987e-06, + "loss": 2.1971, + "step": 6432 + }, + { + "epoch": 1.17, + "learning_rate": 2.6514561339643245e-06, + "loss": 0.4791, + "step": 6433 + }, + { + "epoch": 1.17, + "learning_rate": 2.6514015289406625e-06, + "loss": 0.0834, + "step": 6434 + }, + { + "epoch": 1.17, + "learning_rate": 2.6513469239170004e-06, + "loss": 0.4923, + "step": 6435 + }, + { + "epoch": 1.17, + "learning_rate": 2.651292318893338e-06, + "loss": 0.7502, + "step": 6436 + }, + { + "epoch": 1.17, + "learning_rate": 2.651237713869676e-06, + "loss": 0.4036, + "step": 6437 + }, + { + "epoch": 1.17, + "learning_rate": 2.651183108846014e-06, + "loss": 0.9591, + "step": 6438 + }, + { + "epoch": 1.17, + "learning_rate": 2.6511285038223517e-06, + "loss": 0.6392, + "step": 6439 + }, + { + "epoch": 1.17, + "learning_rate": 2.6510738987986897e-06, + "loss": 0.0978, + "step": 6440 + }, + { + "epoch": 1.17, + "learning_rate": 2.6510192937750276e-06, + "loss": 0.9005, + "step": 6441 + }, + { + "epoch": 1.17, + "learning_rate": 2.650964688751365e-06, + "loss": 2.5931, + "step": 6442 + }, + { + "epoch": 1.17, + "learning_rate": 2.650910083727703e-06, + "loss": 0.0843, + "step": 6443 + }, + { + "epoch": 1.17, + "learning_rate": 2.650855478704041e-06, + "loss": 0.0818, + "step": 6444 + }, + { + "epoch": 1.17, + "learning_rate": 2.6508008736803786e-06, + "loss": 0.7809, + "step": 6445 + }, + { + "epoch": 1.17, + "learning_rate": 2.6507462686567165e-06, + "loss": 1.4025, + "step": 6446 + }, + { + "epoch": 1.17, + "learning_rate": 2.650691663633054e-06, + "loss": 1.1307, + "step": 6447 + }, + { + "epoch": 1.17, + "learning_rate": 2.650637058609392e-06, + "loss": 1.4159, + "step": 6448 + }, + { + "epoch": 1.17, + "learning_rate": 2.65058245358573e-06, + "loss": 2.1158, + "step": 6449 + }, + { + "epoch": 1.17, + "learning_rate": 2.650527848562068e-06, + "loss": 0.2712, + "step": 6450 + }, + { + "epoch": 1.17, + "learning_rate": 2.650473243538406e-06, + "loss": 0.0354, + "step": 6451 + }, + { + "epoch": 1.17, + "learning_rate": 2.6504186385147433e-06, + "loss": 0.5928, + "step": 6452 + }, + { + "epoch": 1.17, + "learning_rate": 2.6503640334910813e-06, + "loss": 0.6806, + "step": 6453 + }, + { + "epoch": 1.17, + "learning_rate": 2.650309428467419e-06, + "loss": 2.0632, + "step": 6454 + }, + { + "epoch": 1.17, + "learning_rate": 2.650254823443757e-06, + "loss": 0.1724, + "step": 6455 + }, + { + "epoch": 1.17, + "learning_rate": 2.6502002184200947e-06, + "loss": 0.6874, + "step": 6456 + }, + { + "epoch": 1.17, + "learning_rate": 2.6501456133964326e-06, + "loss": 0.8823, + "step": 6457 + }, + { + "epoch": 1.17, + "learning_rate": 2.6500910083727706e-06, + "loss": 0.8985, + "step": 6458 + }, + { + "epoch": 1.17, + "learning_rate": 2.650036403349108e-06, + "loss": 2.6038, + "step": 6459 + }, + { + "epoch": 1.17, + "learning_rate": 2.649981798325446e-06, + "loss": 0.8472, + "step": 6460 + }, + { + "epoch": 1.17, + "learning_rate": 2.6499271933017835e-06, + "loss": 0.4702, + "step": 6461 + }, + { + "epoch": 1.18, + "learning_rate": 2.6498725882781215e-06, + "loss": 1.0839, + "step": 6462 + }, + { + "epoch": 1.18, + "learning_rate": 2.6498179832544594e-06, + "loss": 1.6515, + "step": 6463 + }, + { + "epoch": 1.18, + "learning_rate": 2.6497633782307974e-06, + "loss": 0.4756, + "step": 6464 + }, + { + "epoch": 1.18, + "learning_rate": 2.6497087732071353e-06, + "loss": 1.464, + "step": 6465 + }, + { + "epoch": 1.18, + "learning_rate": 2.6496541681834733e-06, + "loss": 1.1714, + "step": 6466 + }, + { + "epoch": 1.18, + "learning_rate": 2.6495995631598108e-06, + "loss": 0.1477, + "step": 6467 + }, + { + "epoch": 1.18, + "learning_rate": 2.6495449581361487e-06, + "loss": 0.9665, + "step": 6468 + }, + { + "epoch": 1.18, + "learning_rate": 2.6494903531124867e-06, + "loss": 1.2396, + "step": 6469 + }, + { + "epoch": 1.18, + "learning_rate": 2.649435748088824e-06, + "loss": 0.8488, + "step": 6470 + }, + { + "epoch": 1.18, + "learning_rate": 2.649381143065162e-06, + "loss": 2.0693, + "step": 6471 + }, + { + "epoch": 1.18, + "learning_rate": 2.6493265380415e-06, + "loss": 0.8792, + "step": 6472 + }, + { + "epoch": 1.18, + "learning_rate": 2.6492719330178376e-06, + "loss": 0.1491, + "step": 6473 + }, + { + "epoch": 1.18, + "learning_rate": 2.6492173279941755e-06, + "loss": 1.2954, + "step": 6474 + }, + { + "epoch": 1.18, + "learning_rate": 2.649162722970513e-06, + "loss": 0.1742, + "step": 6475 + }, + { + "epoch": 1.18, + "learning_rate": 2.649108117946851e-06, + "loss": 0.7445, + "step": 6476 + }, + { + "epoch": 1.18, + "learning_rate": 2.649053512923189e-06, + "loss": 1.2637, + "step": 6477 + }, + { + "epoch": 1.18, + "learning_rate": 2.648998907899527e-06, + "loss": 0.051, + "step": 6478 + }, + { + "epoch": 1.18, + "learning_rate": 2.648944302875865e-06, + "loss": 0.698, + "step": 6479 + }, + { + "epoch": 1.18, + "learning_rate": 2.6488896978522028e-06, + "loss": 2.0561, + "step": 6480 + }, + { + "epoch": 1.18, + "learning_rate": 2.6488350928285403e-06, + "loss": 1.9659, + "step": 6481 + }, + { + "epoch": 1.18, + "learning_rate": 2.6487804878048782e-06, + "loss": 1.2028, + "step": 6482 + }, + { + "epoch": 1.18, + "learning_rate": 2.648725882781216e-06, + "loss": 1.7308, + "step": 6483 + }, + { + "epoch": 1.18, + "learning_rate": 2.6486712777575537e-06, + "loss": 1.7598, + "step": 6484 + }, + { + "epoch": 1.18, + "learning_rate": 2.6486166727338916e-06, + "loss": 0.712, + "step": 6485 + }, + { + "epoch": 1.18, + "learning_rate": 2.6485620677102296e-06, + "loss": 1.3288, + "step": 6486 + }, + { + "epoch": 1.18, + "learning_rate": 2.648507462686567e-06, + "loss": 0.3182, + "step": 6487 + }, + { + "epoch": 1.18, + "learning_rate": 2.648452857662905e-06, + "loss": 0.7231, + "step": 6488 + }, + { + "epoch": 1.18, + "learning_rate": 2.6483982526392426e-06, + "loss": 1.779, + "step": 6489 + }, + { + "epoch": 1.18, + "learning_rate": 2.6483436476155805e-06, + "loss": 1.5341, + "step": 6490 + }, + { + "epoch": 1.18, + "learning_rate": 2.6482890425919185e-06, + "loss": 0.7617, + "step": 6491 + }, + { + "epoch": 1.18, + "learning_rate": 2.6482344375682564e-06, + "loss": 0.1081, + "step": 6492 + }, + { + "epoch": 1.18, + "learning_rate": 2.6481798325445943e-06, + "loss": 0.3839, + "step": 6493 + }, + { + "epoch": 1.18, + "learning_rate": 2.6481252275209323e-06, + "loss": 0.4305, + "step": 6494 + }, + { + "epoch": 1.18, + "learning_rate": 2.64807062249727e-06, + "loss": 1.611, + "step": 6495 + }, + { + "epoch": 1.18, + "learning_rate": 2.6480160174736077e-06, + "loss": 0.6488, + "step": 6496 + }, + { + "epoch": 1.18, + "learning_rate": 2.6479614124499457e-06, + "loss": 0.7726, + "step": 6497 + }, + { + "epoch": 1.18, + "learning_rate": 2.647906807426283e-06, + "loss": 0.2976, + "step": 6498 + }, + { + "epoch": 1.18, + "learning_rate": 2.647852202402621e-06, + "loss": 0.9938, + "step": 6499 + }, + { + "epoch": 1.18, + "learning_rate": 2.6477975973789587e-06, + "loss": 0.7037, + "step": 6500 + }, + { + "epoch": 1.18, + "learning_rate": 2.6477429923552966e-06, + "loss": 0.4335, + "step": 6501 + }, + { + "epoch": 1.18, + "learning_rate": 2.6476883873316346e-06, + "loss": 0.9045, + "step": 6502 + }, + { + "epoch": 1.18, + "learning_rate": 2.647633782307972e-06, + "loss": 0.7856, + "step": 6503 + }, + { + "epoch": 1.18, + "learning_rate": 2.64757917728431e-06, + "loss": 1.2243, + "step": 6504 + }, + { + "epoch": 1.18, + "learning_rate": 2.647524572260648e-06, + "loss": 1.1763, + "step": 6505 + }, + { + "epoch": 1.18, + "learning_rate": 2.647469967236986e-06, + "loss": 1.2351, + "step": 6506 + }, + { + "epoch": 1.18, + "learning_rate": 2.647415362213324e-06, + "loss": 0.6129, + "step": 6507 + }, + { + "epoch": 1.18, + "learning_rate": 2.647360757189662e-06, + "loss": 0.4233, + "step": 6508 + }, + { + "epoch": 1.18, + "learning_rate": 2.6473061521659993e-06, + "loss": 4.4049, + "step": 6509 + }, + { + "epoch": 1.18, + "learning_rate": 2.6472515471423373e-06, + "loss": 0.4852, + "step": 6510 + }, + { + "epoch": 1.18, + "learning_rate": 2.647196942118675e-06, + "loss": 1.6069, + "step": 6511 + }, + { + "epoch": 1.18, + "learning_rate": 2.6471423370950127e-06, + "loss": 1.0224, + "step": 6512 + }, + { + "epoch": 1.18, + "learning_rate": 2.6470877320713507e-06, + "loss": 1.4132, + "step": 6513 + }, + { + "epoch": 1.18, + "learning_rate": 2.647033127047688e-06, + "loss": 0.533, + "step": 6514 + }, + { + "epoch": 1.18, + "learning_rate": 2.646978522024026e-06, + "loss": 2.7081, + "step": 6515 + }, + { + "epoch": 1.18, + "learning_rate": 2.646923917000364e-06, + "loss": 2.5652, + "step": 6516 + }, + { + "epoch": 1.19, + "learning_rate": 2.6468693119767016e-06, + "loss": 0.6113, + "step": 6517 + }, + { + "epoch": 1.19, + "learning_rate": 2.6468147069530395e-06, + "loss": 0.451, + "step": 6518 + }, + { + "epoch": 1.19, + "learning_rate": 2.6467601019293775e-06, + "loss": 1.6139, + "step": 6519 + }, + { + "epoch": 1.19, + "learning_rate": 2.6467054969057154e-06, + "loss": 1.0959, + "step": 6520 + }, + { + "epoch": 1.19, + "learning_rate": 2.6466508918820534e-06, + "loss": 0.567, + "step": 6521 + }, + { + "epoch": 1.19, + "learning_rate": 2.6465962868583913e-06, + "loss": 1.5778, + "step": 6522 + }, + { + "epoch": 1.19, + "learning_rate": 2.646541681834729e-06, + "loss": 0.0912, + "step": 6523 + }, + { + "epoch": 1.19, + "learning_rate": 2.6464870768110668e-06, + "loss": 0.9889, + "step": 6524 + }, + { + "epoch": 1.19, + "learning_rate": 2.6464324717874047e-06, + "loss": 2.0258, + "step": 6525 + }, + { + "epoch": 1.19, + "learning_rate": 2.6463778667637422e-06, + "loss": 1.1426, + "step": 6526 + }, + { + "epoch": 1.19, + "learning_rate": 2.64632326174008e-06, + "loss": 0.5818, + "step": 6527 + }, + { + "epoch": 1.19, + "learning_rate": 2.6462686567164177e-06, + "loss": 1.6798, + "step": 6528 + }, + { + "epoch": 1.19, + "learning_rate": 2.6462140516927556e-06, + "loss": 1.0848, + "step": 6529 + }, + { + "epoch": 1.19, + "learning_rate": 2.6461594466690936e-06, + "loss": 1.1219, + "step": 6530 + }, + { + "epoch": 1.19, + "learning_rate": 2.646104841645431e-06, + "loss": 0.1195, + "step": 6531 + }, + { + "epoch": 1.19, + "learning_rate": 2.646050236621769e-06, + "loss": 1.5885, + "step": 6532 + }, + { + "epoch": 1.19, + "learning_rate": 2.645995631598107e-06, + "loss": 0.9965, + "step": 6533 + }, + { + "epoch": 1.19, + "learning_rate": 2.645941026574445e-06, + "loss": 1.1186, + "step": 6534 + }, + { + "epoch": 1.19, + "learning_rate": 2.645886421550783e-06, + "loss": 0.665, + "step": 6535 + }, + { + "epoch": 1.19, + "learning_rate": 2.645831816527121e-06, + "loss": 1.4364, + "step": 6536 + }, + { + "epoch": 1.19, + "learning_rate": 2.6457772115034583e-06, + "loss": 2.8439, + "step": 6537 + }, + { + "epoch": 1.19, + "learning_rate": 2.6457226064797963e-06, + "loss": 0.5175, + "step": 6538 + }, + { + "epoch": 1.19, + "learning_rate": 2.6456680014561342e-06, + "loss": 0.9978, + "step": 6539 + }, + { + "epoch": 1.19, + "learning_rate": 2.6456133964324718e-06, + "loss": 1.7804, + "step": 6540 + }, + { + "epoch": 1.19, + "learning_rate": 2.6455587914088097e-06, + "loss": 1.1518, + "step": 6541 + }, + { + "epoch": 1.19, + "learning_rate": 2.6455041863851472e-06, + "loss": 0.3765, + "step": 6542 + }, + { + "epoch": 1.19, + "learning_rate": 2.645449581361485e-06, + "loss": 0.9795, + "step": 6543 + }, + { + "epoch": 1.19, + "learning_rate": 2.645394976337823e-06, + "loss": 1.7242, + "step": 6544 + }, + { + "epoch": 1.19, + "learning_rate": 2.645340371314161e-06, + "loss": 0.7811, + "step": 6545 + }, + { + "epoch": 1.19, + "learning_rate": 2.645285766290499e-06, + "loss": 0.3722, + "step": 6546 + }, + { + "epoch": 1.19, + "learning_rate": 2.645231161266837e-06, + "loss": 0.8946, + "step": 6547 + }, + { + "epoch": 1.19, + "learning_rate": 2.6451765562431745e-06, + "loss": 0.0295, + "step": 6548 + }, + { + "epoch": 1.19, + "learning_rate": 2.6451219512195124e-06, + "loss": 0.1112, + "step": 6549 + }, + { + "epoch": 1.19, + "learning_rate": 2.6450673461958503e-06, + "loss": 1.0847, + "step": 6550 + }, + { + "epoch": 1.19, + "learning_rate": 2.645012741172188e-06, + "loss": 1.0571, + "step": 6551 + }, + { + "epoch": 1.19, + "learning_rate": 2.644958136148526e-06, + "loss": 0.5629, + "step": 6552 + }, + { + "epoch": 1.19, + "learning_rate": 2.6449035311248637e-06, + "loss": 0.9092, + "step": 6553 + }, + { + "epoch": 1.19, + "learning_rate": 2.6448489261012013e-06, + "loss": 0.0598, + "step": 6554 + }, + { + "epoch": 1.19, + "learning_rate": 2.644794321077539e-06, + "loss": 1.5365, + "step": 6555 + }, + { + "epoch": 1.19, + "learning_rate": 2.6447397160538767e-06, + "loss": 1.449, + "step": 6556 + }, + { + "epoch": 1.19, + "learning_rate": 2.6446851110302147e-06, + "loss": 0.3206, + "step": 6557 + }, + { + "epoch": 1.19, + "learning_rate": 2.6446305060065526e-06, + "loss": 0.7627, + "step": 6558 + }, + { + "epoch": 1.19, + "learning_rate": 2.6445759009828906e-06, + "loss": 0.4264, + "step": 6559 + }, + { + "epoch": 1.19, + "learning_rate": 2.6445212959592285e-06, + "loss": 1.2946, + "step": 6560 + }, + { + "epoch": 1.19, + "learning_rate": 2.6444666909355664e-06, + "loss": 0.8716, + "step": 6561 + }, + { + "epoch": 1.19, + "learning_rate": 2.644412085911904e-06, + "loss": 0.0434, + "step": 6562 + }, + { + "epoch": 1.19, + "learning_rate": 2.644357480888242e-06, + "loss": 0.3457, + "step": 6563 + }, + { + "epoch": 1.19, + "learning_rate": 2.64430287586458e-06, + "loss": 0.1615, + "step": 6564 + }, + { + "epoch": 1.19, + "learning_rate": 2.6442482708409174e-06, + "loss": 0.6678, + "step": 6565 + }, + { + "epoch": 1.19, + "learning_rate": 2.6441936658172553e-06, + "loss": 0.3905, + "step": 6566 + }, + { + "epoch": 1.19, + "learning_rate": 2.6441390607935933e-06, + "loss": 1.2413, + "step": 6567 + }, + { + "epoch": 1.19, + "learning_rate": 2.6440844557699308e-06, + "loss": 1.6147, + "step": 6568 + }, + { + "epoch": 1.19, + "learning_rate": 2.6440298507462687e-06, + "loss": 1.1729, + "step": 6569 + }, + { + "epoch": 1.19, + "learning_rate": 2.6439752457226062e-06, + "loss": 1.7758, + "step": 6570 + }, + { + "epoch": 1.19, + "learning_rate": 2.643920640698944e-06, + "loss": 1.8196, + "step": 6571 + }, + { + "epoch": 1.2, + "learning_rate": 2.643866035675282e-06, + "loss": 0.891, + "step": 6572 + }, + { + "epoch": 1.2, + "learning_rate": 2.64381143065162e-06, + "loss": 1.9568, + "step": 6573 + }, + { + "epoch": 1.2, + "learning_rate": 2.643756825627958e-06, + "loss": 0.9125, + "step": 6574 + }, + { + "epoch": 1.2, + "learning_rate": 2.643702220604296e-06, + "loss": 0.9661, + "step": 6575 + }, + { + "epoch": 1.2, + "learning_rate": 2.6436476155806335e-06, + "loss": 1.8178, + "step": 6576 + }, + { + "epoch": 1.2, + "learning_rate": 2.6435930105569714e-06, + "loss": 1.4869, + "step": 6577 + }, + { + "epoch": 1.2, + "learning_rate": 2.6435384055333094e-06, + "loss": 1.2704, + "step": 6578 + }, + { + "epoch": 1.2, + "learning_rate": 2.643483800509647e-06, + "loss": 0.9298, + "step": 6579 + }, + { + "epoch": 1.2, + "learning_rate": 2.643429195485985e-06, + "loss": 0.4583, + "step": 6580 + }, + { + "epoch": 1.2, + "learning_rate": 2.6433745904623223e-06, + "loss": 2.5147, + "step": 6581 + }, + { + "epoch": 1.2, + "learning_rate": 2.6433199854386603e-06, + "loss": 1.2672, + "step": 6582 + }, + { + "epoch": 1.2, + "learning_rate": 2.6432653804149982e-06, + "loss": 1.5622, + "step": 6583 + }, + { + "epoch": 1.2, + "learning_rate": 2.6432107753913358e-06, + "loss": 0.6302, + "step": 6584 + }, + { + "epoch": 1.2, + "learning_rate": 2.6431561703676737e-06, + "loss": 1.0058, + "step": 6585 + }, + { + "epoch": 1.2, + "learning_rate": 2.6431015653440116e-06, + "loss": 0.4464, + "step": 6586 + }, + { + "epoch": 1.2, + "learning_rate": 2.6430469603203496e-06, + "loss": 0.8896, + "step": 6587 + }, + { + "epoch": 1.2, + "learning_rate": 2.6429923552966875e-06, + "loss": 0.0954, + "step": 6588 + }, + { + "epoch": 1.2, + "learning_rate": 2.6429377502730255e-06, + "loss": 0.9602, + "step": 6589 + }, + { + "epoch": 1.2, + "learning_rate": 2.642883145249363e-06, + "loss": 2.5009, + "step": 6590 + }, + { + "epoch": 1.2, + "learning_rate": 2.642828540225701e-06, + "loss": 0.5251, + "step": 6591 + }, + { + "epoch": 1.2, + "learning_rate": 2.642773935202039e-06, + "loss": 1.8231, + "step": 6592 + }, + { + "epoch": 1.2, + "learning_rate": 2.6427193301783764e-06, + "loss": 0.3155, + "step": 6593 + }, + { + "epoch": 1.2, + "learning_rate": 2.6426647251547143e-06, + "loss": 0.1471, + "step": 6594 + }, + { + "epoch": 1.2, + "learning_rate": 2.642610120131052e-06, + "loss": 2.1156, + "step": 6595 + }, + { + "epoch": 1.2, + "learning_rate": 2.64255551510739e-06, + "loss": 0.9989, + "step": 6596 + }, + { + "epoch": 1.2, + "learning_rate": 2.6425009100837277e-06, + "loss": 1.0708, + "step": 6597 + }, + { + "epoch": 1.2, + "learning_rate": 2.6424463050600653e-06, + "loss": 0.2254, + "step": 6598 + }, + { + "epoch": 1.2, + "learning_rate": 2.642391700036403e-06, + "loss": 0.5453, + "step": 6599 + }, + { + "epoch": 1.2, + "learning_rate": 2.642337095012741e-06, + "loss": 3.4057, + "step": 6600 + }, + { + "epoch": 1.2, + "learning_rate": 2.642282489989079e-06, + "loss": 0.0358, + "step": 6601 + }, + { + "epoch": 1.2, + "learning_rate": 2.642227884965417e-06, + "loss": 0.6419, + "step": 6602 + }, + { + "epoch": 1.2, + "learning_rate": 2.642173279941755e-06, + "loss": 2.3374, + "step": 6603 + }, + { + "epoch": 1.2, + "learning_rate": 2.6421186749180925e-06, + "loss": 0.6623, + "step": 6604 + }, + { + "epoch": 1.2, + "learning_rate": 2.6420640698944304e-06, + "loss": 0.8182, + "step": 6605 + }, + { + "epoch": 1.2, + "learning_rate": 2.6420094648707684e-06, + "loss": 1.4591, + "step": 6606 + }, + { + "epoch": 1.2, + "learning_rate": 2.641954859847106e-06, + "loss": 0.034, + "step": 6607 + }, + { + "epoch": 1.2, + "learning_rate": 2.641900254823444e-06, + "loss": 0.0527, + "step": 6608 + }, + { + "epoch": 1.2, + "learning_rate": 2.6418456497997814e-06, + "loss": 2.4643, + "step": 6609 + }, + { + "epoch": 1.2, + "learning_rate": 2.6417910447761193e-06, + "loss": 1.2952, + "step": 6610 + }, + { + "epoch": 1.2, + "learning_rate": 2.6417364397524573e-06, + "loss": 1.0063, + "step": 6611 + }, + { + "epoch": 1.2, + "learning_rate": 2.6416818347287948e-06, + "loss": 0.8563, + "step": 6612 + }, + { + "epoch": 1.2, + "learning_rate": 2.6416272297051327e-06, + "loss": 0.0472, + "step": 6613 + }, + { + "epoch": 1.2, + "learning_rate": 2.6415726246814707e-06, + "loss": 0.6388, + "step": 6614 + }, + { + "epoch": 1.2, + "learning_rate": 2.6415180196578086e-06, + "loss": 0.167, + "step": 6615 + }, + { + "epoch": 1.2, + "learning_rate": 2.6414634146341466e-06, + "loss": 0.9559, + "step": 6616 + }, + { + "epoch": 1.2, + "learning_rate": 2.6414088096104845e-06, + "loss": 0.9903, + "step": 6617 + }, + { + "epoch": 1.2, + "learning_rate": 2.641354204586822e-06, + "loss": 0.3683, + "step": 6618 + }, + { + "epoch": 1.2, + "learning_rate": 2.64129959956316e-06, + "loss": 1.1934, + "step": 6619 + }, + { + "epoch": 1.2, + "learning_rate": 2.641244994539498e-06, + "loss": 0.3323, + "step": 6620 + }, + { + "epoch": 1.2, + "learning_rate": 2.6411903895158354e-06, + "loss": 0.3463, + "step": 6621 + }, + { + "epoch": 1.2, + "learning_rate": 2.6411357844921734e-06, + "loss": 0.433, + "step": 6622 + }, + { + "epoch": 1.2, + "learning_rate": 2.641081179468511e-06, + "loss": 0.8733, + "step": 6623 + }, + { + "epoch": 1.2, + "learning_rate": 2.641026574444849e-06, + "loss": 0.4615, + "step": 6624 + }, + { + "epoch": 1.2, + "learning_rate": 2.6409719694211868e-06, + "loss": 2.3091, + "step": 6625 + }, + { + "epoch": 1.2, + "learning_rate": 2.6409173643975247e-06, + "loss": 0.4339, + "step": 6626 + }, + { + "epoch": 1.21, + "learning_rate": 2.6408627593738627e-06, + "loss": 1.2336, + "step": 6627 + }, + { + "epoch": 1.21, + "learning_rate": 2.6408081543502006e-06, + "loss": 0.3219, + "step": 6628 + }, + { + "epoch": 1.21, + "learning_rate": 2.640753549326538e-06, + "loss": 0.1253, + "step": 6629 + }, + { + "epoch": 1.21, + "learning_rate": 2.640698944302876e-06, + "loss": 0.242, + "step": 6630 + }, + { + "epoch": 1.21, + "learning_rate": 2.640644339279214e-06, + "loss": 1.6517, + "step": 6631 + }, + { + "epoch": 1.21, + "learning_rate": 2.6405897342555515e-06, + "loss": 0.4232, + "step": 6632 + }, + { + "epoch": 1.21, + "learning_rate": 2.6405351292318895e-06, + "loss": 1.0307, + "step": 6633 + }, + { + "epoch": 1.21, + "learning_rate": 2.6404805242082274e-06, + "loss": 0.9809, + "step": 6634 + }, + { + "epoch": 1.21, + "learning_rate": 2.640425919184565e-06, + "loss": 0.4788, + "step": 6635 + }, + { + "epoch": 1.21, + "learning_rate": 2.640371314160903e-06, + "loss": 0.7548, + "step": 6636 + }, + { + "epoch": 1.21, + "learning_rate": 2.6403167091372404e-06, + "loss": 0.3335, + "step": 6637 + }, + { + "epoch": 1.21, + "learning_rate": 2.6402621041135783e-06, + "loss": 1.0781, + "step": 6638 + }, + { + "epoch": 1.21, + "learning_rate": 2.6402074990899163e-06, + "loss": 0.6462, + "step": 6639 + }, + { + "epoch": 1.21, + "learning_rate": 2.6401528940662542e-06, + "loss": 1.2293, + "step": 6640 + }, + { + "epoch": 1.21, + "learning_rate": 2.640098289042592e-06, + "loss": 1.087, + "step": 6641 + }, + { + "epoch": 1.21, + "learning_rate": 2.64004368401893e-06, + "loss": 0.1719, + "step": 6642 + }, + { + "epoch": 1.21, + "learning_rate": 2.6399890789952676e-06, + "loss": 0.8655, + "step": 6643 + }, + { + "epoch": 1.21, + "learning_rate": 2.6399344739716056e-06, + "loss": 0.5547, + "step": 6644 + }, + { + "epoch": 1.21, + "learning_rate": 2.6398798689479435e-06, + "loss": 1.1375, + "step": 6645 + }, + { + "epoch": 1.21, + "learning_rate": 2.639825263924281e-06, + "loss": 1.7047, + "step": 6646 + }, + { + "epoch": 1.21, + "learning_rate": 2.639770658900619e-06, + "loss": 0.7581, + "step": 6647 + }, + { + "epoch": 1.21, + "learning_rate": 2.6397160538769565e-06, + "loss": 1.3189, + "step": 6648 + }, + { + "epoch": 1.21, + "learning_rate": 2.6396614488532945e-06, + "loss": 0.0586, + "step": 6649 + }, + { + "epoch": 1.21, + "learning_rate": 2.6396068438296324e-06, + "loss": 0.6314, + "step": 6650 + }, + { + "epoch": 1.21, + "learning_rate": 2.63955223880597e-06, + "loss": 0.9105, + "step": 6651 + }, + { + "epoch": 1.21, + "learning_rate": 2.639497633782308e-06, + "loss": 0.1039, + "step": 6652 + }, + { + "epoch": 1.21, + "learning_rate": 2.639443028758646e-06, + "loss": 0.4714, + "step": 6653 + }, + { + "epoch": 1.21, + "learning_rate": 2.6393884237349837e-06, + "loss": 0.7433, + "step": 6654 + }, + { + "epoch": 1.21, + "learning_rate": 2.6393338187113217e-06, + "loss": 0.6245, + "step": 6655 + }, + { + "epoch": 1.21, + "learning_rate": 2.6392792136876596e-06, + "loss": 0.4427, + "step": 6656 + }, + { + "epoch": 1.21, + "learning_rate": 2.639224608663997e-06, + "loss": 1.0364, + "step": 6657 + }, + { + "epoch": 1.21, + "learning_rate": 2.639170003640335e-06, + "loss": 1.2368, + "step": 6658 + }, + { + "epoch": 1.21, + "learning_rate": 2.639115398616673e-06, + "loss": 0.307, + "step": 6659 + }, + { + "epoch": 1.21, + "learning_rate": 2.6390607935930106e-06, + "loss": 1.114, + "step": 6660 + }, + { + "epoch": 1.21, + "learning_rate": 2.6390061885693485e-06, + "loss": 1.1889, + "step": 6661 + }, + { + "epoch": 1.21, + "learning_rate": 2.638951583545686e-06, + "loss": 0.1164, + "step": 6662 + }, + { + "epoch": 1.21, + "learning_rate": 2.638896978522024e-06, + "loss": 2.2444, + "step": 6663 + }, + { + "epoch": 1.21, + "learning_rate": 2.638842373498362e-06, + "loss": 3.0895, + "step": 6664 + }, + { + "epoch": 1.21, + "learning_rate": 2.6387877684746994e-06, + "loss": 1.8558, + "step": 6665 + }, + { + "epoch": 1.21, + "learning_rate": 2.6387331634510374e-06, + "loss": 1.7275, + "step": 6666 + }, + { + "epoch": 1.21, + "learning_rate": 2.6386785584273753e-06, + "loss": 0.6562, + "step": 6667 + }, + { + "epoch": 1.21, + "learning_rate": 2.6386239534037133e-06, + "loss": 1.3931, + "step": 6668 + }, + { + "epoch": 1.21, + "learning_rate": 2.638569348380051e-06, + "loss": 0.5887, + "step": 6669 + }, + { + "epoch": 1.21, + "learning_rate": 2.638514743356389e-06, + "loss": 1.0085, + "step": 6670 + }, + { + "epoch": 1.21, + "learning_rate": 2.6384601383327267e-06, + "loss": 0.972, + "step": 6671 + }, + { + "epoch": 1.21, + "learning_rate": 2.6384055333090646e-06, + "loss": 1.7919, + "step": 6672 + }, + { + "epoch": 1.21, + "learning_rate": 2.6383509282854026e-06, + "loss": 0.584, + "step": 6673 + }, + { + "epoch": 1.21, + "learning_rate": 2.63829632326174e-06, + "loss": 0.2302, + "step": 6674 + }, + { + "epoch": 1.21, + "learning_rate": 2.638241718238078e-06, + "loss": 0.2179, + "step": 6675 + }, + { + "epoch": 1.21, + "learning_rate": 2.6381871132144155e-06, + "loss": 0.0432, + "step": 6676 + }, + { + "epoch": 1.21, + "learning_rate": 2.6381325081907535e-06, + "loss": 0.5495, + "step": 6677 + }, + { + "epoch": 1.21, + "learning_rate": 2.6380779031670914e-06, + "loss": 0.6578, + "step": 6678 + }, + { + "epoch": 1.21, + "learning_rate": 2.638023298143429e-06, + "loss": 0.3714, + "step": 6679 + }, + { + "epoch": 1.21, + "learning_rate": 2.637968693119767e-06, + "loss": 1.989, + "step": 6680 + }, + { + "epoch": 1.21, + "learning_rate": 2.637914088096105e-06, + "loss": 0.8593, + "step": 6681 + }, + { + "epoch": 1.22, + "learning_rate": 2.6378594830724428e-06, + "loss": 1.494, + "step": 6682 + }, + { + "epoch": 1.22, + "learning_rate": 2.6378048780487807e-06, + "loss": 1.286, + "step": 6683 + }, + { + "epoch": 1.22, + "learning_rate": 2.6377502730251187e-06, + "loss": 0.0927, + "step": 6684 + }, + { + "epoch": 1.22, + "learning_rate": 2.637695668001456e-06, + "loss": 0.0416, + "step": 6685 + }, + { + "epoch": 1.22, + "learning_rate": 2.637641062977794e-06, + "loss": 2.2094, + "step": 6686 + }, + { + "epoch": 1.22, + "learning_rate": 2.637586457954132e-06, + "loss": 2.1366, + "step": 6687 + }, + { + "epoch": 1.22, + "learning_rate": 2.6375318529304696e-06, + "loss": 0.74, + "step": 6688 + }, + { + "epoch": 1.22, + "learning_rate": 2.6374772479068075e-06, + "loss": 0.3859, + "step": 6689 + }, + { + "epoch": 1.22, + "learning_rate": 2.637422642883145e-06, + "loss": 1.2022, + "step": 6690 + }, + { + "epoch": 1.22, + "learning_rate": 2.637368037859483e-06, + "loss": 0.7329, + "step": 6691 + }, + { + "epoch": 1.22, + "learning_rate": 2.637313432835821e-06, + "loss": 1.5433, + "step": 6692 + }, + { + "epoch": 1.22, + "learning_rate": 2.6372588278121585e-06, + "loss": 2.2002, + "step": 6693 + }, + { + "epoch": 1.22, + "learning_rate": 2.6372042227884964e-06, + "loss": 0.8336, + "step": 6694 + }, + { + "epoch": 1.22, + "learning_rate": 2.6371496177648343e-06, + "loss": 1.9036, + "step": 6695 + }, + { + "epoch": 1.22, + "learning_rate": 2.6370950127411723e-06, + "loss": 1.2127, + "step": 6696 + }, + { + "epoch": 1.22, + "learning_rate": 2.6370404077175102e-06, + "loss": 1.9126, + "step": 6697 + }, + { + "epoch": 1.22, + "learning_rate": 2.636985802693848e-06, + "loss": 2.4194, + "step": 6698 + }, + { + "epoch": 1.22, + "learning_rate": 2.6369311976701857e-06, + "loss": 0.3418, + "step": 6699 + }, + { + "epoch": 1.22, + "learning_rate": 2.6368765926465236e-06, + "loss": 0.5903, + "step": 6700 + }, + { + "epoch": 1.22, + "learning_rate": 2.6368219876228616e-06, + "loss": 0.3123, + "step": 6701 + }, + { + "epoch": 1.22, + "learning_rate": 2.636767382599199e-06, + "loss": 0.5732, + "step": 6702 + }, + { + "epoch": 1.22, + "learning_rate": 2.636712777575537e-06, + "loss": 2.5825, + "step": 6703 + }, + { + "epoch": 1.22, + "learning_rate": 2.6366581725518746e-06, + "loss": 1.1522, + "step": 6704 + }, + { + "epoch": 1.22, + "learning_rate": 2.6366035675282125e-06, + "loss": 0.9849, + "step": 6705 + }, + { + "epoch": 1.22, + "learning_rate": 2.6365489625045504e-06, + "loss": 0.9408, + "step": 6706 + }, + { + "epoch": 1.22, + "learning_rate": 2.6364943574808884e-06, + "loss": 0.9518, + "step": 6707 + }, + { + "epoch": 1.22, + "learning_rate": 2.6364397524572263e-06, + "loss": 1.304, + "step": 6708 + }, + { + "epoch": 1.22, + "learning_rate": 2.6363851474335643e-06, + "loss": 0.7216, + "step": 6709 + }, + { + "epoch": 1.22, + "learning_rate": 2.636330542409902e-06, + "loss": 0.0541, + "step": 6710 + }, + { + "epoch": 1.22, + "learning_rate": 2.6362759373862397e-06, + "loss": 0.0829, + "step": 6711 + }, + { + "epoch": 1.22, + "learning_rate": 2.6362213323625777e-06, + "loss": 0.06, + "step": 6712 + }, + { + "epoch": 1.22, + "learning_rate": 2.636166727338915e-06, + "loss": 2.0735, + "step": 6713 + }, + { + "epoch": 1.22, + "learning_rate": 2.636112122315253e-06, + "loss": 1.138, + "step": 6714 + }, + { + "epoch": 1.22, + "learning_rate": 2.636057517291591e-06, + "loss": 0.8439, + "step": 6715 + }, + { + "epoch": 1.22, + "learning_rate": 2.6360029122679286e-06, + "loss": 1.1169, + "step": 6716 + }, + { + "epoch": 1.22, + "learning_rate": 2.6359483072442666e-06, + "loss": 0.3419, + "step": 6717 + }, + { + "epoch": 1.22, + "learning_rate": 2.635893702220604e-06, + "loss": 2.5435, + "step": 6718 + }, + { + "epoch": 1.22, + "learning_rate": 2.635839097196942e-06, + "loss": 0.0973, + "step": 6719 + }, + { + "epoch": 1.22, + "learning_rate": 2.63578449217328e-06, + "loss": 1.0338, + "step": 6720 + }, + { + "epoch": 1.22, + "learning_rate": 2.635729887149618e-06, + "loss": 0.7569, + "step": 6721 + }, + { + "epoch": 1.22, + "learning_rate": 2.635675282125956e-06, + "loss": 0.5085, + "step": 6722 + }, + { + "epoch": 1.22, + "learning_rate": 2.635620677102294e-06, + "loss": 0.7919, + "step": 6723 + }, + { + "epoch": 1.22, + "learning_rate": 2.6355660720786313e-06, + "loss": 2.0203, + "step": 6724 + }, + { + "epoch": 1.22, + "learning_rate": 2.6355114670549693e-06, + "loss": 0.8557, + "step": 6725 + }, + { + "epoch": 1.22, + "learning_rate": 2.635456862031307e-06, + "loss": 0.1158, + "step": 6726 + }, + { + "epoch": 1.22, + "learning_rate": 2.6354022570076447e-06, + "loss": 1.8128, + "step": 6727 + }, + { + "epoch": 1.22, + "learning_rate": 2.6353476519839827e-06, + "loss": 0.0432, + "step": 6728 + }, + { + "epoch": 1.22, + "learning_rate": 2.63529304696032e-06, + "loss": 1.9734, + "step": 6729 + }, + { + "epoch": 1.22, + "learning_rate": 2.635238441936658e-06, + "loss": 1.1473, + "step": 6730 + }, + { + "epoch": 1.22, + "learning_rate": 2.635183836912996e-06, + "loss": 0.0927, + "step": 6731 + }, + { + "epoch": 1.22, + "learning_rate": 2.6351292318893336e-06, + "loss": 2.408, + "step": 6732 + }, + { + "epoch": 1.22, + "learning_rate": 2.6350746268656715e-06, + "loss": 1.5039, + "step": 6733 + }, + { + "epoch": 1.22, + "learning_rate": 2.6350200218420095e-06, + "loss": 0.8262, + "step": 6734 + }, + { + "epoch": 1.22, + "learning_rate": 2.6349654168183474e-06, + "loss": 0.0545, + "step": 6735 + }, + { + "epoch": 1.22, + "learning_rate": 2.6349108117946854e-06, + "loss": 0.7414, + "step": 6736 + }, + { + "epoch": 1.23, + "learning_rate": 2.6348562067710233e-06, + "loss": 1.4522, + "step": 6737 + }, + { + "epoch": 1.23, + "learning_rate": 2.634801601747361e-06, + "loss": 1.0778, + "step": 6738 + }, + { + "epoch": 1.23, + "learning_rate": 2.6347469967236988e-06, + "loss": 1.0298, + "step": 6739 + }, + { + "epoch": 1.23, + "learning_rate": 2.6346923917000367e-06, + "loss": 0.3252, + "step": 6740 + }, + { + "epoch": 1.23, + "learning_rate": 2.6346377866763742e-06, + "loss": 1.3305, + "step": 6741 + }, + { + "epoch": 1.23, + "learning_rate": 2.634583181652712e-06, + "loss": 0.6281, + "step": 6742 + }, + { + "epoch": 1.23, + "learning_rate": 2.6345285766290497e-06, + "loss": 0.4888, + "step": 6743 + }, + { + "epoch": 1.23, + "learning_rate": 2.6344739716053876e-06, + "loss": 0.0962, + "step": 6744 + }, + { + "epoch": 1.23, + "learning_rate": 2.6344193665817256e-06, + "loss": 2.3827, + "step": 6745 + }, + { + "epoch": 1.23, + "learning_rate": 2.634364761558063e-06, + "loss": 1.801, + "step": 6746 + }, + { + "epoch": 1.23, + "learning_rate": 2.634310156534401e-06, + "loss": 1.456, + "step": 6747 + }, + { + "epoch": 1.23, + "learning_rate": 2.634255551510739e-06, + "loss": 0.8602, + "step": 6748 + }, + { + "epoch": 1.23, + "learning_rate": 2.634200946487077e-06, + "loss": 0.086, + "step": 6749 + }, + { + "epoch": 1.23, + "learning_rate": 2.634146341463415e-06, + "loss": 0.493, + "step": 6750 + }, + { + "epoch": 1.23, + "learning_rate": 2.634091736439753e-06, + "loss": 1.8236, + "step": 6751 + }, + { + "epoch": 1.23, + "learning_rate": 2.6340371314160903e-06, + "loss": 4.0481, + "step": 6752 + }, + { + "epoch": 1.23, + "learning_rate": 2.6339825263924283e-06, + "loss": 1.6233, + "step": 6753 + }, + { + "epoch": 1.23, + "learning_rate": 2.6339279213687662e-06, + "loss": 1.6618, + "step": 6754 + }, + { + "epoch": 1.23, + "learning_rate": 2.6338733163451037e-06, + "loss": 0.9996, + "step": 6755 + }, + { + "epoch": 1.23, + "learning_rate": 2.6338187113214417e-06, + "loss": 1.2062, + "step": 6756 + }, + { + "epoch": 1.23, + "learning_rate": 2.633764106297779e-06, + "loss": 0.3245, + "step": 6757 + }, + { + "epoch": 1.23, + "learning_rate": 2.633709501274117e-06, + "loss": 0.5193, + "step": 6758 + }, + { + "epoch": 1.23, + "learning_rate": 2.633654896250455e-06, + "loss": 0.4927, + "step": 6759 + }, + { + "epoch": 1.23, + "learning_rate": 2.6336002912267926e-06, + "loss": 0.4544, + "step": 6760 + }, + { + "epoch": 1.23, + "learning_rate": 2.6335456862031306e-06, + "loss": 0.1355, + "step": 6761 + }, + { + "epoch": 1.23, + "learning_rate": 2.6334910811794685e-06, + "loss": 0.0547, + "step": 6762 + }, + { + "epoch": 1.23, + "learning_rate": 2.6334364761558064e-06, + "loss": 0.5683, + "step": 6763 + }, + { + "epoch": 1.23, + "learning_rate": 2.6333818711321444e-06, + "loss": 0.1799, + "step": 6764 + }, + { + "epoch": 1.23, + "learning_rate": 2.6333272661084823e-06, + "loss": 0.0739, + "step": 6765 + }, + { + "epoch": 1.23, + "learning_rate": 2.63327266108482e-06, + "loss": 2.0708, + "step": 6766 + }, + { + "epoch": 1.23, + "learning_rate": 2.633218056061158e-06, + "loss": 0.4777, + "step": 6767 + }, + { + "epoch": 1.23, + "learning_rate": 2.6331634510374957e-06, + "loss": 2.1575, + "step": 6768 + }, + { + "epoch": 1.23, + "learning_rate": 2.6331088460138333e-06, + "loss": 1.8573, + "step": 6769 + }, + { + "epoch": 1.23, + "learning_rate": 2.633054240990171e-06, + "loss": 1.6484, + "step": 6770 + }, + { + "epoch": 1.23, + "learning_rate": 2.6329996359665087e-06, + "loss": 1.5761, + "step": 6771 + }, + { + "epoch": 1.23, + "learning_rate": 2.6329450309428467e-06, + "loss": 0.6643, + "step": 6772 + }, + { + "epoch": 1.23, + "learning_rate": 2.6328904259191846e-06, + "loss": 0.8834, + "step": 6773 + }, + { + "epoch": 1.23, + "learning_rate": 2.632835820895522e-06, + "loss": 1.1616, + "step": 6774 + }, + { + "epoch": 1.23, + "learning_rate": 2.63278121587186e-06, + "loss": 0.5011, + "step": 6775 + }, + { + "epoch": 1.23, + "learning_rate": 2.632726610848198e-06, + "loss": 1.2277, + "step": 6776 + }, + { + "epoch": 1.23, + "learning_rate": 2.632672005824536e-06, + "loss": 2.2642, + "step": 6777 + }, + { + "epoch": 1.23, + "learning_rate": 2.632617400800874e-06, + "loss": 1.617, + "step": 6778 + }, + { + "epoch": 1.23, + "learning_rate": 2.632562795777212e-06, + "loss": 1.3133, + "step": 6779 + }, + { + "epoch": 1.23, + "learning_rate": 2.6325081907535494e-06, + "loss": 1.4823, + "step": 6780 + }, + { + "epoch": 1.23, + "learning_rate": 2.6324535857298873e-06, + "loss": 0.2822, + "step": 6781 + }, + { + "epoch": 1.23, + "learning_rate": 2.6323989807062253e-06, + "loss": 0.5607, + "step": 6782 + }, + { + "epoch": 1.23, + "learning_rate": 2.6323443756825628e-06, + "loss": 0.0609, + "step": 6783 + }, + { + "epoch": 1.23, + "learning_rate": 2.6322897706589007e-06, + "loss": 1.6624, + "step": 6784 + }, + { + "epoch": 1.23, + "learning_rate": 2.6322351656352382e-06, + "loss": 0.9355, + "step": 6785 + }, + { + "epoch": 1.23, + "learning_rate": 2.632180560611576e-06, + "loss": 0.1219, + "step": 6786 + }, + { + "epoch": 1.23, + "learning_rate": 2.632125955587914e-06, + "loss": 1.1917, + "step": 6787 + }, + { + "epoch": 1.23, + "learning_rate": 2.6320713505642516e-06, + "loss": 1.111, + "step": 6788 + }, + { + "epoch": 1.23, + "learning_rate": 2.6320167455405896e-06, + "loss": 0.9649, + "step": 6789 + }, + { + "epoch": 1.23, + "learning_rate": 2.6319621405169275e-06, + "loss": 0.4856, + "step": 6790 + }, + { + "epoch": 1.23, + "learning_rate": 2.6319075354932655e-06, + "loss": 1.1796, + "step": 6791 + }, + { + "epoch": 1.24, + "learning_rate": 2.6318529304696034e-06, + "loss": 0.5748, + "step": 6792 + }, + { + "epoch": 1.24, + "learning_rate": 2.6317983254459414e-06, + "loss": 1.5048, + "step": 6793 + }, + { + "epoch": 1.24, + "learning_rate": 2.631743720422279e-06, + "loss": 0.5978, + "step": 6794 + }, + { + "epoch": 1.24, + "learning_rate": 2.631689115398617e-06, + "loss": 0.5193, + "step": 6795 + }, + { + "epoch": 1.24, + "learning_rate": 2.6316345103749548e-06, + "loss": 1.7803, + "step": 6796 + }, + { + "epoch": 1.24, + "learning_rate": 2.6315799053512923e-06, + "loss": 2.3678, + "step": 6797 + }, + { + "epoch": 1.24, + "learning_rate": 2.6315253003276302e-06, + "loss": 1.7348, + "step": 6798 + }, + { + "epoch": 1.24, + "learning_rate": 2.6314706953039678e-06, + "loss": 0.596, + "step": 6799 + }, + { + "epoch": 1.24, + "learning_rate": 2.6314160902803057e-06, + "loss": 0.2594, + "step": 6800 + }, + { + "epoch": 1.24, + "learning_rate": 2.6313614852566436e-06, + "loss": 2.3881, + "step": 6801 + }, + { + "epoch": 1.24, + "learning_rate": 2.6313068802329816e-06, + "loss": 0.4838, + "step": 6802 + }, + { + "epoch": 1.24, + "learning_rate": 2.6312522752093195e-06, + "loss": 1.1087, + "step": 6803 + }, + { + "epoch": 1.24, + "learning_rate": 2.6311976701856575e-06, + "loss": 1.0221, + "step": 6804 + }, + { + "epoch": 1.24, + "learning_rate": 2.631143065161995e-06, + "loss": 0.0512, + "step": 6805 + }, + { + "epoch": 1.24, + "learning_rate": 2.631088460138333e-06, + "loss": 1.7144, + "step": 6806 + }, + { + "epoch": 1.24, + "learning_rate": 2.631033855114671e-06, + "loss": 1.3695, + "step": 6807 + }, + { + "epoch": 1.24, + "learning_rate": 2.6309792500910084e-06, + "loss": 0.9215, + "step": 6808 + }, + { + "epoch": 1.24, + "learning_rate": 2.6309246450673463e-06, + "loss": 0.9207, + "step": 6809 + }, + { + "epoch": 1.24, + "learning_rate": 2.630870040043684e-06, + "loss": 0.533, + "step": 6810 + }, + { + "epoch": 1.24, + "learning_rate": 2.630815435020022e-06, + "loss": 0.7489, + "step": 6811 + }, + { + "epoch": 1.24, + "learning_rate": 2.6307608299963597e-06, + "loss": 0.3515, + "step": 6812 + }, + { + "epoch": 1.24, + "learning_rate": 2.6307062249726973e-06, + "loss": 0.2282, + "step": 6813 + }, + { + "epoch": 1.24, + "learning_rate": 2.630651619949035e-06, + "loss": 0.4511, + "step": 6814 + }, + { + "epoch": 1.24, + "learning_rate": 2.630597014925373e-06, + "loss": 3.1644, + "step": 6815 + }, + { + "epoch": 1.24, + "learning_rate": 2.630542409901711e-06, + "loss": 0.9256, + "step": 6816 + }, + { + "epoch": 1.24, + "learning_rate": 2.630487804878049e-06, + "loss": 1.2424, + "step": 6817 + }, + { + "epoch": 1.24, + "learning_rate": 2.630433199854387e-06, + "loss": 1.6051, + "step": 6818 + }, + { + "epoch": 1.24, + "learning_rate": 2.6303785948307245e-06, + "loss": 2.1967, + "step": 6819 + }, + { + "epoch": 1.24, + "learning_rate": 2.6303239898070624e-06, + "loss": 2.2163, + "step": 6820 + }, + { + "epoch": 1.24, + "learning_rate": 2.6302693847834004e-06, + "loss": 0.3817, + "step": 6821 + }, + { + "epoch": 1.24, + "learning_rate": 2.630214779759738e-06, + "loss": 2.6337, + "step": 6822 + }, + { + "epoch": 1.24, + "learning_rate": 2.630160174736076e-06, + "loss": 0.4819, + "step": 6823 + }, + { + "epoch": 1.24, + "learning_rate": 2.6301055697124134e-06, + "loss": 0.5201, + "step": 6824 + }, + { + "epoch": 1.24, + "learning_rate": 2.6300509646887513e-06, + "loss": 2.3672, + "step": 6825 + }, + { + "epoch": 1.24, + "learning_rate": 2.6299963596650893e-06, + "loss": 1.7304, + "step": 6826 + }, + { + "epoch": 1.24, + "learning_rate": 2.6299417546414268e-06, + "loss": 0.5031, + "step": 6827 + }, + { + "epoch": 1.24, + "learning_rate": 2.6298871496177647e-06, + "loss": 0.9073, + "step": 6828 + }, + { + "epoch": 1.24, + "learning_rate": 2.6298325445941027e-06, + "loss": 0.193, + "step": 6829 + }, + { + "epoch": 1.24, + "learning_rate": 2.6297779395704406e-06, + "loss": 2.3546, + "step": 6830 + }, + { + "epoch": 1.24, + "learning_rate": 2.6297233345467786e-06, + "loss": 0.4814, + "step": 6831 + }, + { + "epoch": 1.24, + "learning_rate": 2.6296687295231165e-06, + "loss": 1.0735, + "step": 6832 + }, + { + "epoch": 1.24, + "learning_rate": 2.629614124499454e-06, + "loss": 1.4469, + "step": 6833 + }, + { + "epoch": 1.24, + "learning_rate": 2.629559519475792e-06, + "loss": 0.7433, + "step": 6834 + }, + { + "epoch": 1.24, + "learning_rate": 2.62950491445213e-06, + "loss": 0.2509, + "step": 6835 + }, + { + "epoch": 1.24, + "learning_rate": 2.6294503094284674e-06, + "loss": 1.1868, + "step": 6836 + }, + { + "epoch": 1.24, + "learning_rate": 2.6293957044048054e-06, + "loss": 0.5534, + "step": 6837 + }, + { + "epoch": 1.24, + "learning_rate": 2.629341099381143e-06, + "loss": 0.3671, + "step": 6838 + }, + { + "epoch": 1.24, + "learning_rate": 2.629286494357481e-06, + "loss": 0.7012, + "step": 6839 + }, + { + "epoch": 1.24, + "learning_rate": 2.6292318893338188e-06, + "loss": 1.1578, + "step": 6840 + }, + { + "epoch": 1.24, + "learning_rate": 2.6291772843101563e-06, + "loss": 0.6289, + "step": 6841 + }, + { + "epoch": 1.24, + "learning_rate": 2.6291226792864942e-06, + "loss": 1.7741, + "step": 6842 + }, + { + "epoch": 1.24, + "learning_rate": 2.629068074262832e-06, + "loss": 2.5085, + "step": 6843 + }, + { + "epoch": 1.24, + "learning_rate": 2.62901346923917e-06, + "loss": 1.3442, + "step": 6844 + }, + { + "epoch": 1.24, + "learning_rate": 2.628958864215508e-06, + "loss": 0.7653, + "step": 6845 + }, + { + "epoch": 1.24, + "learning_rate": 2.628904259191846e-06, + "loss": 1.2659, + "step": 6846 + }, + { + "epoch": 1.25, + "learning_rate": 2.6288496541681835e-06, + "loss": 2.1983, + "step": 6847 + }, + { + "epoch": 1.25, + "learning_rate": 2.6287950491445215e-06, + "loss": 0.6803, + "step": 6848 + }, + { + "epoch": 1.25, + "learning_rate": 2.6287404441208594e-06, + "loss": 0.1007, + "step": 6849 + }, + { + "epoch": 1.25, + "learning_rate": 2.628685839097197e-06, + "loss": 1.8334, + "step": 6850 + }, + { + "epoch": 1.25, + "learning_rate": 2.628631234073535e-06, + "loss": 0.8235, + "step": 6851 + }, + { + "epoch": 1.25, + "learning_rate": 2.6285766290498724e-06, + "loss": 1.534, + "step": 6852 + }, + { + "epoch": 1.25, + "learning_rate": 2.6285220240262103e-06, + "loss": 1.7294, + "step": 6853 + }, + { + "epoch": 1.25, + "learning_rate": 2.6284674190025483e-06, + "loss": 0.65, + "step": 6854 + }, + { + "epoch": 1.25, + "learning_rate": 2.628412813978886e-06, + "loss": 1.538, + "step": 6855 + }, + { + "epoch": 1.25, + "learning_rate": 2.6283582089552237e-06, + "loss": 0.2279, + "step": 6856 + }, + { + "epoch": 1.25, + "learning_rate": 2.6283036039315617e-06, + "loss": 0.5762, + "step": 6857 + }, + { + "epoch": 1.25, + "learning_rate": 2.6282489989078996e-06, + "loss": 1.6285, + "step": 6858 + }, + { + "epoch": 1.25, + "learning_rate": 2.6281943938842376e-06, + "loss": 1.862, + "step": 6859 + }, + { + "epoch": 1.25, + "learning_rate": 2.6281397888605755e-06, + "loss": 0.7796, + "step": 6860 + }, + { + "epoch": 1.25, + "learning_rate": 2.628085183836913e-06, + "loss": 1.2099, + "step": 6861 + }, + { + "epoch": 1.25, + "learning_rate": 2.628030578813251e-06, + "loss": 1.2605, + "step": 6862 + }, + { + "epoch": 1.25, + "learning_rate": 2.627975973789589e-06, + "loss": 0.9392, + "step": 6863 + }, + { + "epoch": 1.25, + "learning_rate": 2.6279213687659264e-06, + "loss": 0.9262, + "step": 6864 + }, + { + "epoch": 1.25, + "learning_rate": 2.6278667637422644e-06, + "loss": 0.2918, + "step": 6865 + }, + { + "epoch": 1.25, + "learning_rate": 2.627812158718602e-06, + "loss": 1.148, + "step": 6866 + }, + { + "epoch": 1.25, + "learning_rate": 2.62775755369494e-06, + "loss": 3.4704, + "step": 6867 + }, + { + "epoch": 1.25, + "learning_rate": 2.627702948671278e-06, + "loss": 0.569, + "step": 6868 + }, + { + "epoch": 1.25, + "learning_rate": 2.6276483436476153e-06, + "loss": 0.4668, + "step": 6869 + }, + { + "epoch": 1.25, + "learning_rate": 2.6275937386239533e-06, + "loss": 2.1002, + "step": 6870 + }, + { + "epoch": 1.25, + "learning_rate": 2.627539133600291e-06, + "loss": 0.5977, + "step": 6871 + }, + { + "epoch": 1.25, + "learning_rate": 2.627484528576629e-06, + "loss": 0.3262, + "step": 6872 + }, + { + "epoch": 1.25, + "learning_rate": 2.627429923552967e-06, + "loss": 0.652, + "step": 6873 + }, + { + "epoch": 1.25, + "learning_rate": 2.627375318529305e-06, + "loss": 1.7968, + "step": 6874 + }, + { + "epoch": 1.25, + "learning_rate": 2.6273207135056426e-06, + "loss": 0.9702, + "step": 6875 + }, + { + "epoch": 1.25, + "learning_rate": 2.6272661084819805e-06, + "loss": 0.129, + "step": 6876 + }, + { + "epoch": 1.25, + "learning_rate": 2.6272115034583184e-06, + "loss": 1.0687, + "step": 6877 + }, + { + "epoch": 1.25, + "learning_rate": 2.627156898434656e-06, + "loss": 1.7254, + "step": 6878 + }, + { + "epoch": 1.25, + "learning_rate": 2.627102293410994e-06, + "loss": 3.1757, + "step": 6879 + }, + { + "epoch": 1.25, + "learning_rate": 2.6270476883873314e-06, + "loss": 0.5971, + "step": 6880 + }, + { + "epoch": 1.25, + "learning_rate": 2.6269930833636694e-06, + "loss": 0.219, + "step": 6881 + }, + { + "epoch": 1.25, + "learning_rate": 2.6269384783400073e-06, + "loss": 1.0764, + "step": 6882 + }, + { + "epoch": 1.25, + "learning_rate": 2.6268838733163453e-06, + "loss": 0.397, + "step": 6883 + }, + { + "epoch": 1.25, + "learning_rate": 2.626829268292683e-06, + "loss": 1.9298, + "step": 6884 + }, + { + "epoch": 1.25, + "learning_rate": 2.626774663269021e-06, + "loss": 1.899, + "step": 6885 + }, + { + "epoch": 1.25, + "learning_rate": 2.6267200582453587e-06, + "loss": 1.6198, + "step": 6886 + }, + { + "epoch": 1.25, + "learning_rate": 2.6266654532216966e-06, + "loss": 0.2729, + "step": 6887 + }, + { + "epoch": 1.25, + "learning_rate": 2.6266108481980345e-06, + "loss": 0.9876, + "step": 6888 + }, + { + "epoch": 1.25, + "learning_rate": 2.626556243174372e-06, + "loss": 1.4173, + "step": 6889 + }, + { + "epoch": 1.25, + "learning_rate": 2.62650163815071e-06, + "loss": 0.4887, + "step": 6890 + }, + { + "epoch": 1.25, + "learning_rate": 2.6264470331270475e-06, + "loss": 2.8586, + "step": 6891 + }, + { + "epoch": 1.25, + "learning_rate": 2.6263924281033855e-06, + "loss": 0.2535, + "step": 6892 + }, + { + "epoch": 1.25, + "learning_rate": 2.6263378230797234e-06, + "loss": 0.2587, + "step": 6893 + }, + { + "epoch": 1.25, + "learning_rate": 2.626283218056061e-06, + "loss": 0.0593, + "step": 6894 + }, + { + "epoch": 1.25, + "learning_rate": 2.626228613032399e-06, + "loss": 1.653, + "step": 6895 + }, + { + "epoch": 1.25, + "learning_rate": 2.626174008008737e-06, + "loss": 0.0176, + "step": 6896 + }, + { + "epoch": 1.25, + "learning_rate": 2.6261194029850748e-06, + "loss": 1.2123, + "step": 6897 + }, + { + "epoch": 1.25, + "learning_rate": 2.6260647979614127e-06, + "loss": 0.8092, + "step": 6898 + }, + { + "epoch": 1.25, + "learning_rate": 2.6260101929377507e-06, + "loss": 0.2573, + "step": 6899 + }, + { + "epoch": 1.25, + "learning_rate": 2.625955587914088e-06, + "loss": 0.5476, + "step": 6900 + }, + { + "epoch": 1.25, + "learning_rate": 2.625900982890426e-06, + "loss": 1.2339, + "step": 6901 + }, + { + "epoch": 1.26, + "learning_rate": 2.625846377866764e-06, + "loss": 2.0798, + "step": 6902 + }, + { + "epoch": 1.26, + "learning_rate": 2.6257917728431016e-06, + "loss": 0.0401, + "step": 6903 + }, + { + "epoch": 1.26, + "learning_rate": 2.6257371678194395e-06, + "loss": 0.3332, + "step": 6904 + }, + { + "epoch": 1.26, + "learning_rate": 2.625682562795777e-06, + "loss": 0.157, + "step": 6905 + }, + { + "epoch": 1.26, + "learning_rate": 2.625627957772115e-06, + "loss": 0.0734, + "step": 6906 + }, + { + "epoch": 1.26, + "learning_rate": 2.625573352748453e-06, + "loss": 2.0685, + "step": 6907 + }, + { + "epoch": 1.26, + "learning_rate": 2.6255187477247905e-06, + "loss": 2.0379, + "step": 6908 + }, + { + "epoch": 1.26, + "learning_rate": 2.6254641427011284e-06, + "loss": 0.0842, + "step": 6909 + }, + { + "epoch": 1.26, + "learning_rate": 2.6254095376774663e-06, + "loss": 1.73, + "step": 6910 + }, + { + "epoch": 1.26, + "learning_rate": 2.6253549326538043e-06, + "loss": 0.9233, + "step": 6911 + }, + { + "epoch": 1.26, + "learning_rate": 2.6253003276301422e-06, + "loss": 1.5718, + "step": 6912 + }, + { + "epoch": 1.26, + "learning_rate": 2.62524572260648e-06, + "loss": 0.5395, + "step": 6913 + }, + { + "epoch": 1.26, + "learning_rate": 2.6251911175828177e-06, + "loss": 1.6808, + "step": 6914 + }, + { + "epoch": 1.26, + "learning_rate": 2.6251365125591556e-06, + "loss": 0.5296, + "step": 6915 + }, + { + "epoch": 1.26, + "learning_rate": 2.6250819075354936e-06, + "loss": 0.1162, + "step": 6916 + }, + { + "epoch": 1.26, + "learning_rate": 2.625027302511831e-06, + "loss": 1.2532, + "step": 6917 + }, + { + "epoch": 1.26, + "learning_rate": 2.624972697488169e-06, + "loss": 1.1618, + "step": 6918 + }, + { + "epoch": 1.26, + "learning_rate": 2.6249180924645066e-06, + "loss": 1.8048, + "step": 6919 + }, + { + "epoch": 1.26, + "learning_rate": 2.6248634874408445e-06, + "loss": 1.288, + "step": 6920 + }, + { + "epoch": 1.26, + "learning_rate": 2.6248088824171824e-06, + "loss": 0.9104, + "step": 6921 + }, + { + "epoch": 1.26, + "learning_rate": 2.62475427739352e-06, + "loss": 1.156, + "step": 6922 + }, + { + "epoch": 1.26, + "learning_rate": 2.624699672369858e-06, + "loss": 0.3048, + "step": 6923 + }, + { + "epoch": 1.26, + "learning_rate": 2.624645067346196e-06, + "loss": 0.7712, + "step": 6924 + }, + { + "epoch": 1.26, + "learning_rate": 2.624590462322534e-06, + "loss": 0.5185, + "step": 6925 + }, + { + "epoch": 1.26, + "learning_rate": 2.6245358572988717e-06, + "loss": 0.6146, + "step": 6926 + }, + { + "epoch": 1.26, + "learning_rate": 2.6244812522752097e-06, + "loss": 0.2215, + "step": 6927 + }, + { + "epoch": 1.26, + "learning_rate": 2.624426647251547e-06, + "loss": 2.385, + "step": 6928 + }, + { + "epoch": 1.26, + "learning_rate": 2.624372042227885e-06, + "loss": 2.3727, + "step": 6929 + }, + { + "epoch": 1.26, + "learning_rate": 2.624317437204223e-06, + "loss": 1.4977, + "step": 6930 + }, + { + "epoch": 1.26, + "learning_rate": 2.6242628321805606e-06, + "loss": 0.0356, + "step": 6931 + }, + { + "epoch": 1.26, + "learning_rate": 2.6242082271568986e-06, + "loss": 0.0919, + "step": 6932 + }, + { + "epoch": 1.26, + "learning_rate": 2.624153622133236e-06, + "loss": 0.0491, + "step": 6933 + }, + { + "epoch": 1.26, + "learning_rate": 2.624099017109574e-06, + "loss": 2.1155, + "step": 6934 + }, + { + "epoch": 1.26, + "learning_rate": 2.624044412085912e-06, + "loss": 1.0812, + "step": 6935 + }, + { + "epoch": 1.26, + "learning_rate": 2.6239898070622495e-06, + "loss": 0.9311, + "step": 6936 + }, + { + "epoch": 1.26, + "learning_rate": 2.6239352020385874e-06, + "loss": 0.7579, + "step": 6937 + }, + { + "epoch": 1.26, + "learning_rate": 2.6238805970149254e-06, + "loss": 0.9272, + "step": 6938 + }, + { + "epoch": 1.26, + "learning_rate": 2.6238259919912633e-06, + "loss": 0.7762, + "step": 6939 + }, + { + "epoch": 1.26, + "learning_rate": 2.6237713869676013e-06, + "loss": 0.3362, + "step": 6940 + }, + { + "epoch": 1.26, + "learning_rate": 2.623716781943939e-06, + "loss": 2.0173, + "step": 6941 + }, + { + "epoch": 1.26, + "learning_rate": 2.6236621769202767e-06, + "loss": 0.4237, + "step": 6942 + }, + { + "epoch": 1.26, + "learning_rate": 2.6236075718966147e-06, + "loss": 1.6044, + "step": 6943 + }, + { + "epoch": 1.26, + "learning_rate": 2.6235529668729526e-06, + "loss": 0.7834, + "step": 6944 + }, + { + "epoch": 1.26, + "learning_rate": 2.62349836184929e-06, + "loss": 1.1832, + "step": 6945 + }, + { + "epoch": 1.26, + "learning_rate": 2.623443756825628e-06, + "loss": 0.2313, + "step": 6946 + }, + { + "epoch": 1.26, + "learning_rate": 2.6233891518019656e-06, + "loss": 0.5874, + "step": 6947 + }, + { + "epoch": 1.26, + "learning_rate": 2.6233345467783035e-06, + "loss": 2.9013, + "step": 6948 + }, + { + "epoch": 1.26, + "learning_rate": 2.6232799417546415e-06, + "loss": 0.5177, + "step": 6949 + }, + { + "epoch": 1.26, + "learning_rate": 2.623225336730979e-06, + "loss": 0.2214, + "step": 6950 + }, + { + "epoch": 1.26, + "learning_rate": 2.623170731707317e-06, + "loss": 1.5248, + "step": 6951 + }, + { + "epoch": 1.26, + "learning_rate": 2.623116126683655e-06, + "loss": 2.193, + "step": 6952 + }, + { + "epoch": 1.26, + "learning_rate": 2.623061521659993e-06, + "loss": 2.549, + "step": 6953 + }, + { + "epoch": 1.26, + "learning_rate": 2.6230069166363308e-06, + "loss": 1.1581, + "step": 6954 + }, + { + "epoch": 1.26, + "learning_rate": 2.6229523116126687e-06, + "loss": 0.1219, + "step": 6955 + }, + { + "epoch": 1.26, + "learning_rate": 2.6228977065890062e-06, + "loss": 1.5983, + "step": 6956 + }, + { + "epoch": 1.27, + "learning_rate": 2.622843101565344e-06, + "loss": 0.2885, + "step": 6957 + }, + { + "epoch": 1.27, + "learning_rate": 2.6227884965416817e-06, + "loss": 1.623, + "step": 6958 + }, + { + "epoch": 1.27, + "learning_rate": 2.6227338915180196e-06, + "loss": 0.6143, + "step": 6959 + }, + { + "epoch": 1.27, + "learning_rate": 2.6226792864943576e-06, + "loss": 1.0462, + "step": 6960 + }, + { + "epoch": 1.27, + "learning_rate": 2.622624681470695e-06, + "loss": 0.1214, + "step": 6961 + }, + { + "epoch": 1.27, + "learning_rate": 2.622570076447033e-06, + "loss": 0.3871, + "step": 6962 + }, + { + "epoch": 1.27, + "learning_rate": 2.622515471423371e-06, + "loss": 1.0247, + "step": 6963 + }, + { + "epoch": 1.27, + "learning_rate": 2.622460866399709e-06, + "loss": 0.9932, + "step": 6964 + }, + { + "epoch": 1.27, + "learning_rate": 2.622406261376047e-06, + "loss": 0.2951, + "step": 6965 + }, + { + "epoch": 1.27, + "learning_rate": 2.622351656352385e-06, + "loss": 1.0626, + "step": 6966 + }, + { + "epoch": 1.27, + "learning_rate": 2.6222970513287223e-06, + "loss": 1.1063, + "step": 6967 + }, + { + "epoch": 1.27, + "learning_rate": 2.6222424463050603e-06, + "loss": 1.9452, + "step": 6968 + }, + { + "epoch": 1.27, + "learning_rate": 2.6221878412813982e-06, + "loss": 2.8697, + "step": 6969 + }, + { + "epoch": 1.27, + "learning_rate": 2.6221332362577357e-06, + "loss": 0.0522, + "step": 6970 + }, + { + "epoch": 1.27, + "learning_rate": 2.6220786312340737e-06, + "loss": 1.0503, + "step": 6971 + }, + { + "epoch": 1.27, + "learning_rate": 2.622024026210411e-06, + "loss": 0.23, + "step": 6972 + }, + { + "epoch": 1.27, + "learning_rate": 2.621969421186749e-06, + "loss": 0.4233, + "step": 6973 + }, + { + "epoch": 1.27, + "learning_rate": 2.621914816163087e-06, + "loss": 1.0862, + "step": 6974 + }, + { + "epoch": 1.27, + "learning_rate": 2.6218602111394246e-06, + "loss": 0.3212, + "step": 6975 + }, + { + "epoch": 1.27, + "learning_rate": 2.6218056061157626e-06, + "loss": 0.2984, + "step": 6976 + }, + { + "epoch": 1.27, + "learning_rate": 2.6217510010921005e-06, + "loss": 1.5399, + "step": 6977 + }, + { + "epoch": 1.27, + "learning_rate": 2.6216963960684384e-06, + "loss": 1.6941, + "step": 6978 + }, + { + "epoch": 1.27, + "learning_rate": 2.6216417910447764e-06, + "loss": 2.3338, + "step": 6979 + }, + { + "epoch": 1.27, + "learning_rate": 2.6215871860211143e-06, + "loss": 1.1303, + "step": 6980 + }, + { + "epoch": 1.27, + "learning_rate": 2.621532580997452e-06, + "loss": 0.9653, + "step": 6981 + }, + { + "epoch": 1.27, + "learning_rate": 2.62147797597379e-06, + "loss": 0.8075, + "step": 6982 + }, + { + "epoch": 1.27, + "learning_rate": 2.6214233709501277e-06, + "loss": 1.9734, + "step": 6983 + }, + { + "epoch": 1.27, + "learning_rate": 2.6213687659264653e-06, + "loss": 0.1682, + "step": 6984 + }, + { + "epoch": 1.27, + "learning_rate": 2.621314160902803e-06, + "loss": 0.0408, + "step": 6985 + }, + { + "epoch": 1.27, + "learning_rate": 2.6212595558791407e-06, + "loss": 0.6316, + "step": 6986 + }, + { + "epoch": 1.27, + "learning_rate": 2.6212049508554787e-06, + "loss": 2.5841, + "step": 6987 + }, + { + "epoch": 1.27, + "learning_rate": 2.6211503458318166e-06, + "loss": 1.0518, + "step": 6988 + }, + { + "epoch": 1.27, + "learning_rate": 2.621095740808154e-06, + "loss": 0.5502, + "step": 6989 + }, + { + "epoch": 1.27, + "learning_rate": 2.621041135784492e-06, + "loss": 2.8608, + "step": 6990 + }, + { + "epoch": 1.27, + "learning_rate": 2.62098653076083e-06, + "loss": 2.8428, + "step": 6991 + }, + { + "epoch": 1.27, + "learning_rate": 2.620931925737168e-06, + "loss": 1.0193, + "step": 6992 + }, + { + "epoch": 1.27, + "learning_rate": 2.620877320713506e-06, + "loss": 1.6241, + "step": 6993 + }, + { + "epoch": 1.27, + "learning_rate": 2.620822715689844e-06, + "loss": 0.7886, + "step": 6994 + }, + { + "epoch": 1.27, + "learning_rate": 2.6207681106661814e-06, + "loss": 3.2722, + "step": 6995 + }, + { + "epoch": 1.27, + "learning_rate": 2.6207135056425193e-06, + "loss": 0.4298, + "step": 6996 + }, + { + "epoch": 1.27, + "learning_rate": 2.6206589006188572e-06, + "loss": 0.0633, + "step": 6997 + }, + { + "epoch": 1.27, + "learning_rate": 2.6206042955951948e-06, + "loss": 0.033, + "step": 6998 + }, + { + "epoch": 1.27, + "learning_rate": 2.6205496905715327e-06, + "loss": 0.4533, + "step": 6999 + }, + { + "epoch": 1.27, + "learning_rate": 2.6204950855478702e-06, + "loss": 1.2611, + "step": 7000 + }, + { + "epoch": 1.27, + "learning_rate": 2.620440480524208e-06, + "loss": 0.3728, + "step": 7001 + }, + { + "epoch": 1.27, + "learning_rate": 2.620385875500546e-06, + "loss": 1.7941, + "step": 7002 + }, + { + "epoch": 1.27, + "learning_rate": 2.6203312704768836e-06, + "loss": 1.6299, + "step": 7003 + }, + { + "epoch": 1.27, + "learning_rate": 2.6202766654532216e-06, + "loss": 0.5365, + "step": 7004 + }, + { + "epoch": 1.27, + "learning_rate": 2.6202220604295595e-06, + "loss": 0.0318, + "step": 7005 + }, + { + "epoch": 1.27, + "learning_rate": 2.6201674554058975e-06, + "loss": 0.4963, + "step": 7006 + }, + { + "epoch": 1.27, + "learning_rate": 2.6201128503822354e-06, + "loss": 1.5134, + "step": 7007 + }, + { + "epoch": 1.27, + "learning_rate": 2.6200582453585734e-06, + "loss": 0.9428, + "step": 7008 + }, + { + "epoch": 1.27, + "learning_rate": 2.620003640334911e-06, + "loss": 1.2811, + "step": 7009 + }, + { + "epoch": 1.27, + "learning_rate": 2.619949035311249e-06, + "loss": 1.2486, + "step": 7010 + }, + { + "epoch": 1.27, + "learning_rate": 2.6198944302875868e-06, + "loss": 1.033, + "step": 7011 + }, + { + "epoch": 1.28, + "learning_rate": 2.6198398252639243e-06, + "loss": 1.3199, + "step": 7012 + }, + { + "epoch": 1.28, + "learning_rate": 2.6197852202402622e-06, + "loss": 1.7204, + "step": 7013 + }, + { + "epoch": 1.28, + "learning_rate": 2.6197306152165997e-06, + "loss": 2.2382, + "step": 7014 + }, + { + "epoch": 1.28, + "learning_rate": 2.6196760101929377e-06, + "loss": 1.2886, + "step": 7015 + }, + { + "epoch": 1.28, + "learning_rate": 2.6196214051692756e-06, + "loss": 0.5261, + "step": 7016 + }, + { + "epoch": 1.28, + "learning_rate": 2.619566800145613e-06, + "loss": 0.5671, + "step": 7017 + }, + { + "epoch": 1.28, + "learning_rate": 2.619512195121951e-06, + "loss": 0.9081, + "step": 7018 + }, + { + "epoch": 1.28, + "learning_rate": 2.619457590098289e-06, + "loss": 1.7718, + "step": 7019 + }, + { + "epoch": 1.28, + "learning_rate": 2.619402985074627e-06, + "loss": 0.6967, + "step": 7020 + }, + { + "epoch": 1.28, + "learning_rate": 2.619348380050965e-06, + "loss": 0.3335, + "step": 7021 + }, + { + "epoch": 1.28, + "learning_rate": 2.619293775027303e-06, + "loss": 1.1571, + "step": 7022 + }, + { + "epoch": 1.28, + "learning_rate": 2.6192391700036404e-06, + "loss": 2.1886, + "step": 7023 + }, + { + "epoch": 1.28, + "learning_rate": 2.6191845649799783e-06, + "loss": 2.8583, + "step": 7024 + }, + { + "epoch": 1.28, + "learning_rate": 2.6191299599563163e-06, + "loss": 0.542, + "step": 7025 + }, + { + "epoch": 1.28, + "learning_rate": 2.619075354932654e-06, + "loss": 0.8107, + "step": 7026 + }, + { + "epoch": 1.28, + "learning_rate": 2.6190207499089917e-06, + "loss": 0.2401, + "step": 7027 + }, + { + "epoch": 1.28, + "learning_rate": 2.6189661448853293e-06, + "loss": 3.0574, + "step": 7028 + }, + { + "epoch": 1.28, + "learning_rate": 2.618911539861667e-06, + "loss": 0.403, + "step": 7029 + }, + { + "epoch": 1.28, + "learning_rate": 2.618856934838005e-06, + "loss": 0.4183, + "step": 7030 + }, + { + "epoch": 1.28, + "learning_rate": 2.6188023298143427e-06, + "loss": 0.8476, + "step": 7031 + }, + { + "epoch": 1.28, + "learning_rate": 2.6187477247906806e-06, + "loss": 0.0398, + "step": 7032 + }, + { + "epoch": 1.28, + "learning_rate": 2.6186931197670186e-06, + "loss": 0.0451, + "step": 7033 + }, + { + "epoch": 1.28, + "learning_rate": 2.6186385147433565e-06, + "loss": 1.1735, + "step": 7034 + }, + { + "epoch": 1.28, + "learning_rate": 2.6185839097196944e-06, + "loss": 0.3688, + "step": 7035 + }, + { + "epoch": 1.28, + "learning_rate": 2.6185293046960324e-06, + "loss": 0.4304, + "step": 7036 + }, + { + "epoch": 1.28, + "learning_rate": 2.61847469967237e-06, + "loss": 1.3761, + "step": 7037 + }, + { + "epoch": 1.28, + "learning_rate": 2.618420094648708e-06, + "loss": 1.1789, + "step": 7038 + }, + { + "epoch": 1.28, + "learning_rate": 2.6183654896250454e-06, + "loss": 0.2904, + "step": 7039 + }, + { + "epoch": 1.28, + "learning_rate": 2.6183108846013833e-06, + "loss": 0.0391, + "step": 7040 + }, + { + "epoch": 1.28, + "learning_rate": 2.6182562795777213e-06, + "loss": 0.349, + "step": 7041 + }, + { + "epoch": 1.28, + "learning_rate": 2.6182016745540588e-06, + "loss": 0.3455, + "step": 7042 + }, + { + "epoch": 1.28, + "learning_rate": 2.6181470695303967e-06, + "loss": 0.0485, + "step": 7043 + }, + { + "epoch": 1.28, + "learning_rate": 2.6180924645067347e-06, + "loss": 1.6003, + "step": 7044 + }, + { + "epoch": 1.28, + "learning_rate": 2.6180378594830726e-06, + "loss": 1.339, + "step": 7045 + }, + { + "epoch": 1.28, + "learning_rate": 2.6179832544594105e-06, + "loss": 1.3807, + "step": 7046 + }, + { + "epoch": 1.28, + "learning_rate": 2.6179286494357485e-06, + "loss": 1.3556, + "step": 7047 + }, + { + "epoch": 1.28, + "learning_rate": 2.617874044412086e-06, + "loss": 0.3502, + "step": 7048 + }, + { + "epoch": 1.28, + "learning_rate": 2.617819439388424e-06, + "loss": 0.5949, + "step": 7049 + }, + { + "epoch": 1.28, + "learning_rate": 2.617764834364762e-06, + "loss": 0.3544, + "step": 7050 + }, + { + "epoch": 1.28, + "learning_rate": 2.6177102293410994e-06, + "loss": 0.488, + "step": 7051 + }, + { + "epoch": 1.28, + "learning_rate": 2.6176556243174374e-06, + "loss": 1.1647, + "step": 7052 + }, + { + "epoch": 1.28, + "learning_rate": 2.617601019293775e-06, + "loss": 0.856, + "step": 7053 + }, + { + "epoch": 1.28, + "learning_rate": 2.617546414270113e-06, + "loss": 0.3747, + "step": 7054 + }, + { + "epoch": 1.28, + "learning_rate": 2.6174918092464508e-06, + "loss": 0.669, + "step": 7055 + }, + { + "epoch": 1.28, + "learning_rate": 2.6174372042227883e-06, + "loss": 1.4094, + "step": 7056 + }, + { + "epoch": 1.28, + "learning_rate": 2.6173825991991262e-06, + "loss": 0.3769, + "step": 7057 + }, + { + "epoch": 1.28, + "learning_rate": 2.617327994175464e-06, + "loss": 0.2065, + "step": 7058 + }, + { + "epoch": 1.28, + "learning_rate": 2.617273389151802e-06, + "loss": 3.714, + "step": 7059 + }, + { + "epoch": 1.28, + "learning_rate": 2.61721878412814e-06, + "loss": 0.9447, + "step": 7060 + }, + { + "epoch": 1.28, + "learning_rate": 2.617164179104478e-06, + "loss": 1.2855, + "step": 7061 + }, + { + "epoch": 1.28, + "learning_rate": 2.6171095740808155e-06, + "loss": 0.81, + "step": 7062 + }, + { + "epoch": 1.28, + "learning_rate": 2.6170549690571535e-06, + "loss": 2.0951, + "step": 7063 + }, + { + "epoch": 1.28, + "learning_rate": 2.6170003640334914e-06, + "loss": 3.5877, + "step": 7064 + }, + { + "epoch": 1.28, + "learning_rate": 2.616945759009829e-06, + "loss": 0.0571, + "step": 7065 + }, + { + "epoch": 1.28, + "learning_rate": 2.616891153986167e-06, + "loss": 0.7347, + "step": 7066 + }, + { + "epoch": 1.29, + "learning_rate": 2.6168365489625044e-06, + "loss": 0.4838, + "step": 7067 + }, + { + "epoch": 1.29, + "learning_rate": 2.6167819439388423e-06, + "loss": 0.7973, + "step": 7068 + }, + { + "epoch": 1.29, + "learning_rate": 2.6167273389151803e-06, + "loss": 0.789, + "step": 7069 + }, + { + "epoch": 1.29, + "learning_rate": 2.616672733891518e-06, + "loss": 0.8001, + "step": 7070 + }, + { + "epoch": 1.29, + "learning_rate": 2.6166181288678557e-06, + "loss": 0.5454, + "step": 7071 + }, + { + "epoch": 1.29, + "learning_rate": 2.6165635238441937e-06, + "loss": 1.0286, + "step": 7072 + }, + { + "epoch": 1.29, + "learning_rate": 2.6165089188205316e-06, + "loss": 0.3894, + "step": 7073 + }, + { + "epoch": 1.29, + "learning_rate": 2.6164543137968696e-06, + "loss": 0.3829, + "step": 7074 + }, + { + "epoch": 1.29, + "learning_rate": 2.6163997087732075e-06, + "loss": 1.1259, + "step": 7075 + }, + { + "epoch": 1.29, + "learning_rate": 2.616345103749545e-06, + "loss": 1.6731, + "step": 7076 + }, + { + "epoch": 1.29, + "learning_rate": 2.616290498725883e-06, + "loss": 2.1517, + "step": 7077 + }, + { + "epoch": 1.29, + "learning_rate": 2.616235893702221e-06, + "loss": 0.0242, + "step": 7078 + }, + { + "epoch": 1.29, + "learning_rate": 2.6161812886785584e-06, + "loss": 1.4031, + "step": 7079 + }, + { + "epoch": 1.29, + "learning_rate": 2.6161266836548964e-06, + "loss": 1.2081, + "step": 7080 + }, + { + "epoch": 1.29, + "learning_rate": 2.616072078631234e-06, + "loss": 1.1995, + "step": 7081 + }, + { + "epoch": 1.29, + "learning_rate": 2.616017473607572e-06, + "loss": 0.5969, + "step": 7082 + }, + { + "epoch": 1.29, + "learning_rate": 2.61596286858391e-06, + "loss": 0.4166, + "step": 7083 + }, + { + "epoch": 1.29, + "learning_rate": 2.6159082635602473e-06, + "loss": 0.7122, + "step": 7084 + }, + { + "epoch": 1.29, + "learning_rate": 2.6158536585365853e-06, + "loss": 1.1685, + "step": 7085 + }, + { + "epoch": 1.29, + "learning_rate": 2.615799053512923e-06, + "loss": 0.1092, + "step": 7086 + }, + { + "epoch": 1.29, + "learning_rate": 2.615744448489261e-06, + "loss": 1.1968, + "step": 7087 + }, + { + "epoch": 1.29, + "learning_rate": 2.615689843465599e-06, + "loss": 1.7123, + "step": 7088 + }, + { + "epoch": 1.29, + "learning_rate": 2.615635238441937e-06, + "loss": 0.1082, + "step": 7089 + }, + { + "epoch": 1.29, + "learning_rate": 2.6155806334182745e-06, + "loss": 0.6954, + "step": 7090 + }, + { + "epoch": 1.29, + "learning_rate": 2.6155260283946125e-06, + "loss": 0.0473, + "step": 7091 + }, + { + "epoch": 1.29, + "learning_rate": 2.6154714233709504e-06, + "loss": 0.8474, + "step": 7092 + }, + { + "epoch": 1.29, + "learning_rate": 2.615416818347288e-06, + "loss": 1.3945, + "step": 7093 + }, + { + "epoch": 1.29, + "learning_rate": 2.615362213323626e-06, + "loss": 2.0329, + "step": 7094 + }, + { + "epoch": 1.29, + "learning_rate": 2.6153076082999634e-06, + "loss": 0.8642, + "step": 7095 + }, + { + "epoch": 1.29, + "learning_rate": 2.6152530032763014e-06, + "loss": 0.0532, + "step": 7096 + }, + { + "epoch": 1.29, + "learning_rate": 2.6151983982526393e-06, + "loss": 0.5456, + "step": 7097 + }, + { + "epoch": 1.29, + "learning_rate": 2.615143793228977e-06, + "loss": 0.7157, + "step": 7098 + }, + { + "epoch": 1.29, + "learning_rate": 2.6150891882053148e-06, + "loss": 0.0432, + "step": 7099 + }, + { + "epoch": 1.29, + "learning_rate": 2.6150345831816527e-06, + "loss": 0.0813, + "step": 7100 + }, + { + "epoch": 1.29, + "learning_rate": 2.6149799781579907e-06, + "loss": 1.7993, + "step": 7101 + }, + { + "epoch": 1.29, + "learning_rate": 2.6149253731343286e-06, + "loss": 1.9509, + "step": 7102 + }, + { + "epoch": 1.29, + "learning_rate": 2.6148707681106665e-06, + "loss": 0.6196, + "step": 7103 + }, + { + "epoch": 1.29, + "learning_rate": 2.614816163087004e-06, + "loss": 0.8788, + "step": 7104 + }, + { + "epoch": 1.29, + "learning_rate": 2.614761558063342e-06, + "loss": 0.5427, + "step": 7105 + }, + { + "epoch": 1.29, + "learning_rate": 2.61470695303968e-06, + "loss": 1.3526, + "step": 7106 + }, + { + "epoch": 1.29, + "learning_rate": 2.6146523480160175e-06, + "loss": 1.7839, + "step": 7107 + }, + { + "epoch": 1.29, + "learning_rate": 2.6145977429923554e-06, + "loss": 2.4515, + "step": 7108 + }, + { + "epoch": 1.29, + "learning_rate": 2.614543137968693e-06, + "loss": 1.2464, + "step": 7109 + }, + { + "epoch": 1.29, + "learning_rate": 2.614488532945031e-06, + "loss": 1.8589, + "step": 7110 + }, + { + "epoch": 1.29, + "learning_rate": 2.614433927921369e-06, + "loss": 1.6496, + "step": 7111 + }, + { + "epoch": 1.29, + "learning_rate": 2.6143793228977063e-06, + "loss": 0.4932, + "step": 7112 + }, + { + "epoch": 1.29, + "learning_rate": 2.6143247178740443e-06, + "loss": 0.6, + "step": 7113 + }, + { + "epoch": 1.29, + "learning_rate": 2.6142701128503822e-06, + "loss": 0.3708, + "step": 7114 + }, + { + "epoch": 1.29, + "learning_rate": 2.61421550782672e-06, + "loss": 2.989, + "step": 7115 + }, + { + "epoch": 1.29, + "learning_rate": 2.614160902803058e-06, + "loss": 0.1455, + "step": 7116 + }, + { + "epoch": 1.29, + "learning_rate": 2.614106297779396e-06, + "loss": 2.0969, + "step": 7117 + }, + { + "epoch": 1.29, + "learning_rate": 2.6140516927557336e-06, + "loss": 1.5859, + "step": 7118 + }, + { + "epoch": 1.29, + "learning_rate": 2.6139970877320715e-06, + "loss": 0.3972, + "step": 7119 + }, + { + "epoch": 1.29, + "learning_rate": 2.613942482708409e-06, + "loss": 1.547, + "step": 7120 + }, + { + "epoch": 1.29, + "learning_rate": 2.613887877684747e-06, + "loss": 0.7201, + "step": 7121 + }, + { + "epoch": 1.3, + "learning_rate": 2.613833272661085e-06, + "loss": 2.1724, + "step": 7122 + }, + { + "epoch": 1.3, + "learning_rate": 2.6137786676374224e-06, + "loss": 0.7431, + "step": 7123 + }, + { + "epoch": 1.3, + "learning_rate": 2.6137240626137604e-06, + "loss": 0.3485, + "step": 7124 + }, + { + "epoch": 1.3, + "learning_rate": 2.6136694575900983e-06, + "loss": 0.6946, + "step": 7125 + }, + { + "epoch": 1.3, + "learning_rate": 2.613614852566436e-06, + "loss": 0.6643, + "step": 7126 + }, + { + "epoch": 1.3, + "learning_rate": 2.613560247542774e-06, + "loss": 3.0338, + "step": 7127 + }, + { + "epoch": 1.3, + "learning_rate": 2.6135056425191117e-06, + "loss": 2.6688, + "step": 7128 + }, + { + "epoch": 1.3, + "learning_rate": 2.6134510374954497e-06, + "loss": 0.107, + "step": 7129 + }, + { + "epoch": 1.3, + "learning_rate": 2.6133964324717876e-06, + "loss": 0.945, + "step": 7130 + }, + { + "epoch": 1.3, + "learning_rate": 2.6133418274481256e-06, + "loss": 2.1816, + "step": 7131 + }, + { + "epoch": 1.3, + "learning_rate": 2.613287222424463e-06, + "loss": 0.0735, + "step": 7132 + }, + { + "epoch": 1.3, + "learning_rate": 2.613232617400801e-06, + "loss": 1.7275, + "step": 7133 + }, + { + "epoch": 1.3, + "learning_rate": 2.6131780123771386e-06, + "loss": 1.9285, + "step": 7134 + }, + { + "epoch": 1.3, + "learning_rate": 2.6131234073534765e-06, + "loss": 0.0685, + "step": 7135 + }, + { + "epoch": 1.3, + "learning_rate": 2.6130688023298144e-06, + "loss": 1.4755, + "step": 7136 + }, + { + "epoch": 1.3, + "learning_rate": 2.613014197306152e-06, + "loss": 1.8295, + "step": 7137 + }, + { + "epoch": 1.3, + "learning_rate": 2.61295959228249e-06, + "loss": 1.1831, + "step": 7138 + }, + { + "epoch": 1.3, + "learning_rate": 2.612904987258828e-06, + "loss": 2.0871, + "step": 7139 + }, + { + "epoch": 1.3, + "learning_rate": 2.612850382235166e-06, + "loss": 0.7573, + "step": 7140 + }, + { + "epoch": 1.3, + "learning_rate": 2.6127957772115037e-06, + "loss": 1.85, + "step": 7141 + }, + { + "epoch": 1.3, + "learning_rate": 2.6127411721878417e-06, + "loss": 1.0394, + "step": 7142 + }, + { + "epoch": 1.3, + "learning_rate": 2.612686567164179e-06, + "loss": 0.7333, + "step": 7143 + }, + { + "epoch": 1.3, + "learning_rate": 2.612631962140517e-06, + "loss": 1.385, + "step": 7144 + }, + { + "epoch": 1.3, + "learning_rate": 2.612577357116855e-06, + "loss": 1.6078, + "step": 7145 + }, + { + "epoch": 1.3, + "learning_rate": 2.6125227520931926e-06, + "loss": 0.9272, + "step": 7146 + }, + { + "epoch": 1.3, + "learning_rate": 2.6124681470695305e-06, + "loss": 0.3602, + "step": 7147 + }, + { + "epoch": 1.3, + "learning_rate": 2.612413542045868e-06, + "loss": 0.5474, + "step": 7148 + }, + { + "epoch": 1.3, + "learning_rate": 2.612358937022206e-06, + "loss": 0.9489, + "step": 7149 + }, + { + "epoch": 1.3, + "learning_rate": 2.612304331998544e-06, + "loss": 1.3377, + "step": 7150 + }, + { + "epoch": 1.3, + "learning_rate": 2.6122497269748815e-06, + "loss": 1.0775, + "step": 7151 + }, + { + "epoch": 1.3, + "learning_rate": 2.6121951219512194e-06, + "loss": 1.0285, + "step": 7152 + }, + { + "epoch": 1.3, + "learning_rate": 2.6121405169275574e-06, + "loss": 1.9679, + "step": 7153 + }, + { + "epoch": 1.3, + "learning_rate": 2.6120859119038953e-06, + "loss": 0.7066, + "step": 7154 + }, + { + "epoch": 1.3, + "learning_rate": 2.6120313068802332e-06, + "loss": 0.4251, + "step": 7155 + }, + { + "epoch": 1.3, + "learning_rate": 2.611976701856571e-06, + "loss": 0.6718, + "step": 7156 + }, + { + "epoch": 1.3, + "learning_rate": 2.6119220968329087e-06, + "loss": 1.2193, + "step": 7157 + }, + { + "epoch": 1.3, + "learning_rate": 2.6118674918092467e-06, + "loss": 1.272, + "step": 7158 + }, + { + "epoch": 1.3, + "learning_rate": 2.6118128867855846e-06, + "loss": 0.3801, + "step": 7159 + }, + { + "epoch": 1.3, + "learning_rate": 2.611758281761922e-06, + "loss": 1.7959, + "step": 7160 + }, + { + "epoch": 1.3, + "learning_rate": 2.61170367673826e-06, + "loss": 0.8171, + "step": 7161 + }, + { + "epoch": 1.3, + "learning_rate": 2.6116490717145976e-06, + "loss": 1.5355, + "step": 7162 + }, + { + "epoch": 1.3, + "learning_rate": 2.6115944666909355e-06, + "loss": 2.7446, + "step": 7163 + }, + { + "epoch": 1.3, + "learning_rate": 2.6115398616672735e-06, + "loss": 0.8016, + "step": 7164 + }, + { + "epoch": 1.3, + "learning_rate": 2.611485256643611e-06, + "loss": 0.4351, + "step": 7165 + }, + { + "epoch": 1.3, + "learning_rate": 2.611430651619949e-06, + "loss": 1.7049, + "step": 7166 + }, + { + "epoch": 1.3, + "learning_rate": 2.611376046596287e-06, + "loss": 0.3036, + "step": 7167 + }, + { + "epoch": 1.3, + "learning_rate": 2.611321441572625e-06, + "loss": 0.1366, + "step": 7168 + }, + { + "epoch": 1.3, + "learning_rate": 2.6112668365489628e-06, + "loss": 0.6564, + "step": 7169 + }, + { + "epoch": 1.3, + "learning_rate": 2.6112122315253007e-06, + "loss": 0.1006, + "step": 7170 + }, + { + "epoch": 1.3, + "learning_rate": 2.6111576265016382e-06, + "loss": 2.8943, + "step": 7171 + }, + { + "epoch": 1.3, + "learning_rate": 2.611103021477976e-06, + "loss": 0.3109, + "step": 7172 + }, + { + "epoch": 1.3, + "learning_rate": 2.611048416454314e-06, + "loss": 2.3014, + "step": 7173 + }, + { + "epoch": 1.3, + "learning_rate": 2.6109938114306516e-06, + "loss": 1.4971, + "step": 7174 + }, + { + "epoch": 1.3, + "learning_rate": 2.6109392064069896e-06, + "loss": 0.4528, + "step": 7175 + }, + { + "epoch": 1.3, + "learning_rate": 2.610884601383327e-06, + "loss": 1.6268, + "step": 7176 + }, + { + "epoch": 1.31, + "learning_rate": 2.610829996359665e-06, + "loss": 3.2647, + "step": 7177 + }, + { + "epoch": 1.31, + "learning_rate": 2.610775391336003e-06, + "loss": 1.4296, + "step": 7178 + }, + { + "epoch": 1.31, + "learning_rate": 2.6107207863123405e-06, + "loss": 0.9485, + "step": 7179 + }, + { + "epoch": 1.31, + "learning_rate": 2.6106661812886784e-06, + "loss": 1.0514, + "step": 7180 + }, + { + "epoch": 1.31, + "learning_rate": 2.6106115762650164e-06, + "loss": 0.4469, + "step": 7181 + }, + { + "epoch": 1.31, + "learning_rate": 2.6105569712413543e-06, + "loss": 2.0004, + "step": 7182 + }, + { + "epoch": 1.31, + "learning_rate": 2.6105023662176923e-06, + "loss": 0.2974, + "step": 7183 + }, + { + "epoch": 1.31, + "learning_rate": 2.6104477611940302e-06, + "loss": 1.7288, + "step": 7184 + }, + { + "epoch": 1.31, + "learning_rate": 2.6103931561703677e-06, + "loss": 0.781, + "step": 7185 + }, + { + "epoch": 1.31, + "learning_rate": 2.6103385511467057e-06, + "loss": 1.3792, + "step": 7186 + }, + { + "epoch": 1.31, + "learning_rate": 2.6102839461230436e-06, + "loss": 3.3093, + "step": 7187 + }, + { + "epoch": 1.31, + "learning_rate": 2.610229341099381e-06, + "loss": 1.2088, + "step": 7188 + }, + { + "epoch": 1.31, + "learning_rate": 2.610174736075719e-06, + "loss": 1.4995, + "step": 7189 + }, + { + "epoch": 1.31, + "learning_rate": 2.6101201310520566e-06, + "loss": 0.3786, + "step": 7190 + }, + { + "epoch": 1.31, + "learning_rate": 2.6100655260283946e-06, + "loss": 1.3067, + "step": 7191 + }, + { + "epoch": 1.31, + "learning_rate": 2.6100109210047325e-06, + "loss": 0.954, + "step": 7192 + }, + { + "epoch": 1.31, + "learning_rate": 2.60995631598107e-06, + "loss": 0.4953, + "step": 7193 + }, + { + "epoch": 1.31, + "learning_rate": 2.609901710957408e-06, + "loss": 0.3788, + "step": 7194 + }, + { + "epoch": 1.31, + "learning_rate": 2.609847105933746e-06, + "loss": 1.0253, + "step": 7195 + }, + { + "epoch": 1.31, + "learning_rate": 2.609792500910084e-06, + "loss": 2.1471, + "step": 7196 + }, + { + "epoch": 1.31, + "learning_rate": 2.6097378958864218e-06, + "loss": 1.4388, + "step": 7197 + }, + { + "epoch": 1.31, + "learning_rate": 2.6096832908627597e-06, + "loss": 0.9583, + "step": 7198 + }, + { + "epoch": 1.31, + "learning_rate": 2.6096286858390973e-06, + "loss": 0.0335, + "step": 7199 + }, + { + "epoch": 1.31, + "learning_rate": 2.609574080815435e-06, + "loss": 1.1494, + "step": 7200 + }, + { + "epoch": 1.31, + "learning_rate": 2.6095194757917727e-06, + "loss": 0.9869, + "step": 7201 + }, + { + "epoch": 1.31, + "learning_rate": 2.6094648707681107e-06, + "loss": 0.8949, + "step": 7202 + }, + { + "epoch": 1.31, + "learning_rate": 2.6094102657444486e-06, + "loss": 1.8566, + "step": 7203 + }, + { + "epoch": 1.31, + "learning_rate": 2.609355660720786e-06, + "loss": 1.6259, + "step": 7204 + }, + { + "epoch": 1.31, + "learning_rate": 2.609301055697124e-06, + "loss": 1.9423, + "step": 7205 + }, + { + "epoch": 1.31, + "learning_rate": 2.609246450673462e-06, + "loss": 1.0361, + "step": 7206 + }, + { + "epoch": 1.31, + "learning_rate": 2.6091918456497995e-06, + "loss": 0.5027, + "step": 7207 + }, + { + "epoch": 1.31, + "learning_rate": 2.6091372406261375e-06, + "loss": 0.4108, + "step": 7208 + }, + { + "epoch": 1.31, + "learning_rate": 2.6090826356024754e-06, + "loss": 1.397, + "step": 7209 + }, + { + "epoch": 1.31, + "learning_rate": 2.6090280305788134e-06, + "loss": 0.111, + "step": 7210 + }, + { + "epoch": 1.31, + "learning_rate": 2.6089734255551513e-06, + "loss": 2.1909, + "step": 7211 + }, + { + "epoch": 1.31, + "learning_rate": 2.6089188205314892e-06, + "loss": 0.5243, + "step": 7212 + }, + { + "epoch": 1.31, + "learning_rate": 2.6088642155078268e-06, + "loss": 0.7288, + "step": 7213 + }, + { + "epoch": 1.31, + "learning_rate": 2.6088096104841647e-06, + "loss": 1.3568, + "step": 7214 + }, + { + "epoch": 1.31, + "learning_rate": 2.6087550054605022e-06, + "loss": 0.0759, + "step": 7215 + }, + { + "epoch": 1.31, + "learning_rate": 2.60870040043684e-06, + "loss": 1.5435, + "step": 7216 + }, + { + "epoch": 1.31, + "learning_rate": 2.608645795413178e-06, + "loss": 0.6328, + "step": 7217 + }, + { + "epoch": 1.31, + "learning_rate": 2.6085911903895156e-06, + "loss": 3.0058, + "step": 7218 + }, + { + "epoch": 1.31, + "learning_rate": 2.6085365853658536e-06, + "loss": 0.7113, + "step": 7219 + }, + { + "epoch": 1.31, + "learning_rate": 2.6084819803421915e-06, + "loss": 1.0115, + "step": 7220 + }, + { + "epoch": 1.31, + "learning_rate": 2.6084273753185295e-06, + "loss": 0.0775, + "step": 7221 + }, + { + "epoch": 1.31, + "learning_rate": 2.6083727702948674e-06, + "loss": 1.0278, + "step": 7222 + }, + { + "epoch": 1.31, + "learning_rate": 2.6083181652712054e-06, + "loss": 0.5313, + "step": 7223 + }, + { + "epoch": 1.31, + "learning_rate": 2.608263560247543e-06, + "loss": 1.5035, + "step": 7224 + }, + { + "epoch": 1.31, + "learning_rate": 2.608208955223881e-06, + "loss": 1.8505, + "step": 7225 + }, + { + "epoch": 1.31, + "learning_rate": 2.6081543502002188e-06, + "loss": 0.8173, + "step": 7226 + }, + { + "epoch": 1.31, + "learning_rate": 2.6080997451765563e-06, + "loss": 1.0663, + "step": 7227 + }, + { + "epoch": 1.31, + "learning_rate": 2.6080451401528942e-06, + "loss": 0.6901, + "step": 7228 + }, + { + "epoch": 1.31, + "learning_rate": 2.6079905351292317e-06, + "loss": 0.2544, + "step": 7229 + }, + { + "epoch": 1.31, + "learning_rate": 2.6079359301055697e-06, + "loss": 0.1079, + "step": 7230 + }, + { + "epoch": 1.31, + "learning_rate": 2.6078813250819076e-06, + "loss": 1.1527, + "step": 7231 + }, + { + "epoch": 1.32, + "learning_rate": 2.607826720058245e-06, + "loss": 0.7954, + "step": 7232 + }, + { + "epoch": 1.32, + "learning_rate": 2.607772115034583e-06, + "loss": 1.2597, + "step": 7233 + }, + { + "epoch": 1.32, + "learning_rate": 2.607717510010921e-06, + "loss": 0.5806, + "step": 7234 + }, + { + "epoch": 1.32, + "learning_rate": 2.607662904987259e-06, + "loss": 0.4588, + "step": 7235 + }, + { + "epoch": 1.32, + "learning_rate": 2.607608299963597e-06, + "loss": 0.0362, + "step": 7236 + }, + { + "epoch": 1.32, + "learning_rate": 2.607553694939935e-06, + "loss": 1.3117, + "step": 7237 + }, + { + "epoch": 1.32, + "learning_rate": 2.6074990899162724e-06, + "loss": 1.3502, + "step": 7238 + }, + { + "epoch": 1.32, + "learning_rate": 2.6074444848926103e-06, + "loss": 0.3174, + "step": 7239 + }, + { + "epoch": 1.32, + "learning_rate": 2.6073898798689483e-06, + "loss": 0.948, + "step": 7240 + }, + { + "epoch": 1.32, + "learning_rate": 2.607335274845286e-06, + "loss": 0.5403, + "step": 7241 + }, + { + "epoch": 1.32, + "learning_rate": 2.6072806698216237e-06, + "loss": 0.3405, + "step": 7242 + }, + { + "epoch": 1.32, + "learning_rate": 2.6072260647979613e-06, + "loss": 0.334, + "step": 7243 + }, + { + "epoch": 1.32, + "learning_rate": 2.607171459774299e-06, + "loss": 1.5059, + "step": 7244 + }, + { + "epoch": 1.32, + "learning_rate": 2.607116854750637e-06, + "loss": 1.0157, + "step": 7245 + }, + { + "epoch": 1.32, + "learning_rate": 2.6070622497269747e-06, + "loss": 2.0089, + "step": 7246 + }, + { + "epoch": 1.32, + "learning_rate": 2.6070076447033126e-06, + "loss": 0.745, + "step": 7247 + }, + { + "epoch": 1.32, + "learning_rate": 2.6069530396796505e-06, + "loss": 0.5957, + "step": 7248 + }, + { + "epoch": 1.32, + "learning_rate": 2.6068984346559885e-06, + "loss": 0.8739, + "step": 7249 + }, + { + "epoch": 1.32, + "learning_rate": 2.6068438296323264e-06, + "loss": 0.7706, + "step": 7250 + }, + { + "epoch": 1.32, + "learning_rate": 2.6067892246086644e-06, + "loss": 0.5328, + "step": 7251 + }, + { + "epoch": 1.32, + "learning_rate": 2.606734619585002e-06, + "loss": 1.5715, + "step": 7252 + }, + { + "epoch": 1.32, + "learning_rate": 2.60668001456134e-06, + "loss": 0.4334, + "step": 7253 + }, + { + "epoch": 1.32, + "learning_rate": 2.6066254095376778e-06, + "loss": 1.8997, + "step": 7254 + }, + { + "epoch": 1.32, + "learning_rate": 2.6065708045140153e-06, + "loss": 0.4555, + "step": 7255 + }, + { + "epoch": 1.32, + "learning_rate": 2.6065161994903532e-06, + "loss": 0.8151, + "step": 7256 + }, + { + "epoch": 1.32, + "learning_rate": 2.6064615944666908e-06, + "loss": 0.7939, + "step": 7257 + }, + { + "epoch": 1.32, + "learning_rate": 2.6064069894430287e-06, + "loss": 0.3521, + "step": 7258 + }, + { + "epoch": 1.32, + "learning_rate": 2.6063523844193667e-06, + "loss": 1.5091, + "step": 7259 + }, + { + "epoch": 1.32, + "learning_rate": 2.606297779395704e-06, + "loss": 0.6845, + "step": 7260 + }, + { + "epoch": 1.32, + "learning_rate": 2.606243174372042e-06, + "loss": 0.5106, + "step": 7261 + }, + { + "epoch": 1.32, + "learning_rate": 2.60618856934838e-06, + "loss": 0.3005, + "step": 7262 + }, + { + "epoch": 1.32, + "learning_rate": 2.606133964324718e-06, + "loss": 0.9394, + "step": 7263 + }, + { + "epoch": 1.32, + "learning_rate": 2.606079359301056e-06, + "loss": 1.0296, + "step": 7264 + }, + { + "epoch": 1.32, + "learning_rate": 2.606024754277394e-06, + "loss": 0.1985, + "step": 7265 + }, + { + "epoch": 1.32, + "learning_rate": 2.6059701492537314e-06, + "loss": 2.3756, + "step": 7266 + }, + { + "epoch": 1.32, + "learning_rate": 2.6059155442300694e-06, + "loss": 0.4936, + "step": 7267 + }, + { + "epoch": 1.32, + "learning_rate": 2.605860939206407e-06, + "loss": 1.2497, + "step": 7268 + }, + { + "epoch": 1.32, + "learning_rate": 2.605806334182745e-06, + "loss": 0.4843, + "step": 7269 + }, + { + "epoch": 1.32, + "learning_rate": 2.6057517291590828e-06, + "loss": 0.0108, + "step": 7270 + }, + { + "epoch": 1.32, + "learning_rate": 2.6056971241354203e-06, + "loss": 0.8603, + "step": 7271 + }, + { + "epoch": 1.32, + "learning_rate": 2.6056425191117582e-06, + "loss": 0.5882, + "step": 7272 + }, + { + "epoch": 1.32, + "learning_rate": 2.605587914088096e-06, + "loss": 1.5561, + "step": 7273 + }, + { + "epoch": 1.32, + "learning_rate": 2.6055333090644337e-06, + "loss": 1.6991, + "step": 7274 + }, + { + "epoch": 1.32, + "learning_rate": 2.6054787040407716e-06, + "loss": 1.0638, + "step": 7275 + }, + { + "epoch": 1.32, + "learning_rate": 2.6054240990171096e-06, + "loss": 0.9401, + "step": 7276 + }, + { + "epoch": 1.32, + "learning_rate": 2.6053694939934475e-06, + "loss": 0.0416, + "step": 7277 + }, + { + "epoch": 1.32, + "learning_rate": 2.6053148889697855e-06, + "loss": 1.0773, + "step": 7278 + }, + { + "epoch": 1.32, + "learning_rate": 2.6052602839461234e-06, + "loss": 1.0758, + "step": 7279 + }, + { + "epoch": 1.32, + "learning_rate": 2.605205678922461e-06, + "loss": 1.7924, + "step": 7280 + }, + { + "epoch": 1.32, + "learning_rate": 2.605151073898799e-06, + "loss": 0.0569, + "step": 7281 + }, + { + "epoch": 1.32, + "learning_rate": 2.6050964688751364e-06, + "loss": 0.5038, + "step": 7282 + }, + { + "epoch": 1.32, + "learning_rate": 2.6050418638514743e-06, + "loss": 0.9899, + "step": 7283 + }, + { + "epoch": 1.32, + "learning_rate": 2.6049872588278123e-06, + "loss": 1.0664, + "step": 7284 + }, + { + "epoch": 1.32, + "learning_rate": 2.60493265380415e-06, + "loss": 0.601, + "step": 7285 + }, + { + "epoch": 1.32, + "learning_rate": 2.6048780487804877e-06, + "loss": 0.0617, + "step": 7286 + }, + { + "epoch": 1.33, + "learning_rate": 2.6048234437568257e-06, + "loss": 0.2037, + "step": 7287 + }, + { + "epoch": 1.33, + "learning_rate": 2.604768838733163e-06, + "loss": 1.7405, + "step": 7288 + }, + { + "epoch": 1.33, + "learning_rate": 2.604714233709501e-06, + "loss": 1.2547, + "step": 7289 + }, + { + "epoch": 1.33, + "learning_rate": 2.604659628685839e-06, + "loss": 0.4302, + "step": 7290 + }, + { + "epoch": 1.33, + "learning_rate": 2.604605023662177e-06, + "loss": 0.1222, + "step": 7291 + }, + { + "epoch": 1.33, + "learning_rate": 2.604550418638515e-06, + "loss": 1.8363, + "step": 7292 + }, + { + "epoch": 1.33, + "learning_rate": 2.604495813614853e-06, + "loss": 0.4812, + "step": 7293 + }, + { + "epoch": 1.33, + "learning_rate": 2.6044412085911904e-06, + "loss": 0.7258, + "step": 7294 + }, + { + "epoch": 1.33, + "learning_rate": 2.6043866035675284e-06, + "loss": 2.9495, + "step": 7295 + }, + { + "epoch": 1.33, + "learning_rate": 2.604331998543866e-06, + "loss": 1.2227, + "step": 7296 + }, + { + "epoch": 1.33, + "learning_rate": 2.604277393520204e-06, + "loss": 1.0308, + "step": 7297 + }, + { + "epoch": 1.33, + "learning_rate": 2.604222788496542e-06, + "loss": 0.4675, + "step": 7298 + }, + { + "epoch": 1.33, + "learning_rate": 2.6041681834728793e-06, + "loss": 0.7056, + "step": 7299 + }, + { + "epoch": 1.33, + "learning_rate": 2.6041135784492173e-06, + "loss": 0.4918, + "step": 7300 + }, + { + "epoch": 1.33, + "learning_rate": 2.604058973425555e-06, + "loss": 1.8624, + "step": 7301 + }, + { + "epoch": 1.33, + "learning_rate": 2.604004368401893e-06, + "loss": 0.0643, + "step": 7302 + }, + { + "epoch": 1.33, + "learning_rate": 2.603949763378231e-06, + "loss": 1.0242, + "step": 7303 + }, + { + "epoch": 1.33, + "learning_rate": 2.603895158354569e-06, + "loss": 0.8135, + "step": 7304 + }, + { + "epoch": 1.33, + "learning_rate": 2.6038405533309065e-06, + "loss": 1.4633, + "step": 7305 + }, + { + "epoch": 1.33, + "learning_rate": 2.6037859483072445e-06, + "loss": 0.2898, + "step": 7306 + }, + { + "epoch": 1.33, + "learning_rate": 2.6037313432835824e-06, + "loss": 0.6272, + "step": 7307 + }, + { + "epoch": 1.33, + "learning_rate": 2.60367673825992e-06, + "loss": 0.7536, + "step": 7308 + }, + { + "epoch": 1.33, + "learning_rate": 2.603622133236258e-06, + "loss": 0.5694, + "step": 7309 + }, + { + "epoch": 1.33, + "learning_rate": 2.6035675282125954e-06, + "loss": 0.5169, + "step": 7310 + }, + { + "epoch": 1.33, + "learning_rate": 2.6035129231889334e-06, + "loss": 1.7395, + "step": 7311 + }, + { + "epoch": 1.33, + "learning_rate": 2.6034583181652713e-06, + "loss": 1.1038, + "step": 7312 + }, + { + "epoch": 1.33, + "learning_rate": 2.603403713141609e-06, + "loss": 0.6527, + "step": 7313 + }, + { + "epoch": 1.33, + "learning_rate": 2.6033491081179468e-06, + "loss": 0.3219, + "step": 7314 + }, + { + "epoch": 1.33, + "learning_rate": 2.6032945030942847e-06, + "loss": 0.5404, + "step": 7315 + }, + { + "epoch": 1.33, + "learning_rate": 2.6032398980706227e-06, + "loss": 1.2662, + "step": 7316 + }, + { + "epoch": 1.33, + "learning_rate": 2.6031852930469606e-06, + "loss": 0.0574, + "step": 7317 + }, + { + "epoch": 1.33, + "learning_rate": 2.6031306880232985e-06, + "loss": 1.0301, + "step": 7318 + }, + { + "epoch": 1.33, + "learning_rate": 2.603076082999636e-06, + "loss": 1.2196, + "step": 7319 + }, + { + "epoch": 1.33, + "learning_rate": 2.603021477975974e-06, + "loss": 0.529, + "step": 7320 + }, + { + "epoch": 1.33, + "learning_rate": 2.602966872952312e-06, + "loss": 1.6388, + "step": 7321 + }, + { + "epoch": 1.33, + "learning_rate": 2.6029122679286495e-06, + "loss": 1.4386, + "step": 7322 + }, + { + "epoch": 1.33, + "learning_rate": 2.6028576629049874e-06, + "loss": 1.1727, + "step": 7323 + }, + { + "epoch": 1.33, + "learning_rate": 2.602803057881325e-06, + "loss": 3.0333, + "step": 7324 + }, + { + "epoch": 1.33, + "learning_rate": 2.602748452857663e-06, + "loss": 0.0801, + "step": 7325 + }, + { + "epoch": 1.33, + "learning_rate": 2.602693847834001e-06, + "loss": 0.349, + "step": 7326 + }, + { + "epoch": 1.33, + "learning_rate": 2.6026392428103383e-06, + "loss": 1.0422, + "step": 7327 + }, + { + "epoch": 1.33, + "learning_rate": 2.6025846377866763e-06, + "loss": 0.0799, + "step": 7328 + }, + { + "epoch": 1.33, + "learning_rate": 2.6025300327630142e-06, + "loss": 0.3996, + "step": 7329 + }, + { + "epoch": 1.33, + "learning_rate": 2.602475427739352e-06, + "loss": 0.5992, + "step": 7330 + }, + { + "epoch": 1.33, + "learning_rate": 2.60242082271569e-06, + "loss": 1.1765, + "step": 7331 + }, + { + "epoch": 1.33, + "learning_rate": 2.602366217692028e-06, + "loss": 2.0752, + "step": 7332 + }, + { + "epoch": 1.33, + "learning_rate": 2.6023116126683656e-06, + "loss": 2.2653, + "step": 7333 + }, + { + "epoch": 1.33, + "learning_rate": 2.6022570076447035e-06, + "loss": 1.4196, + "step": 7334 + }, + { + "epoch": 1.33, + "learning_rate": 2.6022024026210415e-06, + "loss": 0.2499, + "step": 7335 + }, + { + "epoch": 1.33, + "learning_rate": 2.602147797597379e-06, + "loss": 1.3026, + "step": 7336 + }, + { + "epoch": 1.33, + "learning_rate": 2.602093192573717e-06, + "loss": 0.7811, + "step": 7337 + }, + { + "epoch": 1.33, + "learning_rate": 2.6020385875500544e-06, + "loss": 0.6692, + "step": 7338 + }, + { + "epoch": 1.33, + "learning_rate": 2.6019839825263924e-06, + "loss": 1.071, + "step": 7339 + }, + { + "epoch": 1.33, + "learning_rate": 2.6019293775027303e-06, + "loss": 0.0457, + "step": 7340 + }, + { + "epoch": 1.33, + "learning_rate": 2.601874772479068e-06, + "loss": 1.005, + "step": 7341 + }, + { + "epoch": 1.34, + "learning_rate": 2.601820167455406e-06, + "loss": 1.6928, + "step": 7342 + }, + { + "epoch": 1.34, + "learning_rate": 2.6017655624317437e-06, + "loss": 1.3075, + "step": 7343 + }, + { + "epoch": 1.34, + "learning_rate": 2.6017109574080817e-06, + "loss": 2.3538, + "step": 7344 + }, + { + "epoch": 1.34, + "learning_rate": 2.6016563523844196e-06, + "loss": 1.9183, + "step": 7345 + }, + { + "epoch": 1.34, + "learning_rate": 2.6016017473607576e-06, + "loss": 0.5816, + "step": 7346 + }, + { + "epoch": 1.34, + "learning_rate": 2.601547142337095e-06, + "loss": 0.2961, + "step": 7347 + }, + { + "epoch": 1.34, + "learning_rate": 2.601492537313433e-06, + "loss": 2.3081, + "step": 7348 + }, + { + "epoch": 1.34, + "learning_rate": 2.6014379322897705e-06, + "loss": 0.0746, + "step": 7349 + }, + { + "epoch": 1.34, + "learning_rate": 2.6013833272661085e-06, + "loss": 0.3591, + "step": 7350 + }, + { + "epoch": 1.34, + "learning_rate": 2.6013287222424464e-06, + "loss": 1.7799, + "step": 7351 + }, + { + "epoch": 1.34, + "learning_rate": 2.601274117218784e-06, + "loss": 2.7336, + "step": 7352 + }, + { + "epoch": 1.34, + "learning_rate": 2.601219512195122e-06, + "loss": 1.1242, + "step": 7353 + }, + { + "epoch": 1.34, + "learning_rate": 2.60116490717146e-06, + "loss": 0.3921, + "step": 7354 + }, + { + "epoch": 1.34, + "learning_rate": 2.6011103021477974e-06, + "loss": 2.4567, + "step": 7355 + }, + { + "epoch": 1.34, + "learning_rate": 2.6010556971241353e-06, + "loss": 1.3314, + "step": 7356 + }, + { + "epoch": 1.34, + "learning_rate": 2.6010010921004732e-06, + "loss": 1.1357, + "step": 7357 + }, + { + "epoch": 1.34, + "learning_rate": 2.600946487076811e-06, + "loss": 0.521, + "step": 7358 + }, + { + "epoch": 1.34, + "learning_rate": 2.600891882053149e-06, + "loss": 0.2515, + "step": 7359 + }, + { + "epoch": 1.34, + "learning_rate": 2.600837277029487e-06, + "loss": 0.5011, + "step": 7360 + }, + { + "epoch": 1.34, + "learning_rate": 2.6007826720058246e-06, + "loss": 1.1267, + "step": 7361 + }, + { + "epoch": 1.34, + "learning_rate": 2.6007280669821625e-06, + "loss": 0.3698, + "step": 7362 + }, + { + "epoch": 1.34, + "learning_rate": 2.6006734619585e-06, + "loss": 1.0251, + "step": 7363 + }, + { + "epoch": 1.34, + "learning_rate": 2.600618856934838e-06, + "loss": 0.7458, + "step": 7364 + }, + { + "epoch": 1.34, + "learning_rate": 2.600564251911176e-06, + "loss": 0.0412, + "step": 7365 + }, + { + "epoch": 1.34, + "learning_rate": 2.6005096468875135e-06, + "loss": 0.0472, + "step": 7366 + }, + { + "epoch": 1.34, + "learning_rate": 2.6004550418638514e-06, + "loss": 0.8349, + "step": 7367 + }, + { + "epoch": 1.34, + "learning_rate": 2.6004004368401894e-06, + "loss": 1.2427, + "step": 7368 + }, + { + "epoch": 1.34, + "learning_rate": 2.600345831816527e-06, + "loss": 0.6128, + "step": 7369 + }, + { + "epoch": 1.34, + "learning_rate": 2.600291226792865e-06, + "loss": 1.3066, + "step": 7370 + }, + { + "epoch": 1.34, + "learning_rate": 2.6002366217692028e-06, + "loss": 0.4821, + "step": 7371 + }, + { + "epoch": 1.34, + "learning_rate": 2.6001820167455407e-06, + "loss": 0.6398, + "step": 7372 + }, + { + "epoch": 1.34, + "learning_rate": 2.6001274117218786e-06, + "loss": 0.979, + "step": 7373 + }, + { + "epoch": 1.34, + "learning_rate": 2.6000728066982166e-06, + "loss": 2.0505, + "step": 7374 + }, + { + "epoch": 1.34, + "learning_rate": 2.600018201674554e-06, + "loss": 0.3344, + "step": 7375 + }, + { + "epoch": 1.34, + "learning_rate": 2.599963596650892e-06, + "loss": 0.6926, + "step": 7376 + }, + { + "epoch": 1.34, + "learning_rate": 2.5999089916272296e-06, + "loss": 0.4567, + "step": 7377 + }, + { + "epoch": 1.34, + "learning_rate": 2.5998543866035675e-06, + "loss": 0.0993, + "step": 7378 + }, + { + "epoch": 1.34, + "learning_rate": 2.5997997815799055e-06, + "loss": 1.1433, + "step": 7379 + }, + { + "epoch": 1.34, + "learning_rate": 2.599745176556243e-06, + "loss": 1.5576, + "step": 7380 + }, + { + "epoch": 1.34, + "learning_rate": 2.599690571532581e-06, + "loss": 0.5375, + "step": 7381 + }, + { + "epoch": 1.34, + "learning_rate": 2.599635966508919e-06, + "loss": 1.6761, + "step": 7382 + }, + { + "epoch": 1.34, + "learning_rate": 2.599581361485257e-06, + "loss": 0.0768, + "step": 7383 + }, + { + "epoch": 1.34, + "learning_rate": 2.5995267564615943e-06, + "loss": 0.8911, + "step": 7384 + }, + { + "epoch": 1.34, + "learning_rate": 2.5994721514379323e-06, + "loss": 2.2503, + "step": 7385 + }, + { + "epoch": 1.34, + "learning_rate": 2.5994175464142702e-06, + "loss": 0.8627, + "step": 7386 + }, + { + "epoch": 1.34, + "learning_rate": 2.599362941390608e-06, + "loss": 0.8063, + "step": 7387 + }, + { + "epoch": 1.34, + "learning_rate": 2.599308336366946e-06, + "loss": 3.0876, + "step": 7388 + }, + { + "epoch": 1.34, + "learning_rate": 2.5992537313432836e-06, + "loss": 0.6573, + "step": 7389 + }, + { + "epoch": 1.34, + "learning_rate": 2.5991991263196216e-06, + "loss": 0.7334, + "step": 7390 + }, + { + "epoch": 1.34, + "learning_rate": 2.599144521295959e-06, + "loss": 0.4858, + "step": 7391 + }, + { + "epoch": 1.34, + "learning_rate": 2.599089916272297e-06, + "loss": 1.8027, + "step": 7392 + }, + { + "epoch": 1.34, + "learning_rate": 2.599035311248635e-06, + "loss": 1.7588, + "step": 7393 + }, + { + "epoch": 1.34, + "learning_rate": 2.5989807062249725e-06, + "loss": 0.2474, + "step": 7394 + }, + { + "epoch": 1.34, + "learning_rate": 2.5989261012013104e-06, + "loss": 0.9832, + "step": 7395 + }, + { + "epoch": 1.34, + "learning_rate": 2.5988714961776484e-06, + "loss": 0.2183, + "step": 7396 + }, + { + "epoch": 1.35, + "learning_rate": 2.5988168911539863e-06, + "loss": 0.6514, + "step": 7397 + }, + { + "epoch": 1.35, + "learning_rate": 2.5987622861303243e-06, + "loss": 1.1558, + "step": 7398 + }, + { + "epoch": 1.35, + "learning_rate": 2.5987076811066622e-06, + "loss": 0.6043, + "step": 7399 + }, + { + "epoch": 1.35, + "learning_rate": 2.5986530760829997e-06, + "loss": 0.7249, + "step": 7400 + }, + { + "epoch": 1.35, + "learning_rate": 2.5985984710593377e-06, + "loss": 1.4227, + "step": 7401 + }, + { + "epoch": 1.35, + "learning_rate": 2.5985438660356756e-06, + "loss": 1.2697, + "step": 7402 + }, + { + "epoch": 1.35, + "learning_rate": 2.598489261012013e-06, + "loss": 0.5159, + "step": 7403 + }, + { + "epoch": 1.35, + "learning_rate": 2.598434655988351e-06, + "loss": 0.5684, + "step": 7404 + }, + { + "epoch": 1.35, + "learning_rate": 2.5983800509646886e-06, + "loss": 0.5436, + "step": 7405 + }, + { + "epoch": 1.35, + "learning_rate": 2.5983254459410265e-06, + "loss": 0.3633, + "step": 7406 + }, + { + "epoch": 1.35, + "learning_rate": 2.5982708409173645e-06, + "loss": 0.5352, + "step": 7407 + }, + { + "epoch": 1.35, + "learning_rate": 2.598216235893702e-06, + "loss": 0.6214, + "step": 7408 + }, + { + "epoch": 1.35, + "learning_rate": 2.59816163087004e-06, + "loss": 1.0031, + "step": 7409 + }, + { + "epoch": 1.35, + "learning_rate": 2.598107025846378e-06, + "loss": 0.5191, + "step": 7410 + }, + { + "epoch": 1.35, + "learning_rate": 2.598052420822716e-06, + "loss": 0.4698, + "step": 7411 + }, + { + "epoch": 1.35, + "learning_rate": 2.5979978157990538e-06, + "loss": 1.3809, + "step": 7412 + }, + { + "epoch": 1.35, + "learning_rate": 2.5979432107753917e-06, + "loss": 0.688, + "step": 7413 + }, + { + "epoch": 1.35, + "learning_rate": 2.5978886057517292e-06, + "loss": 1.2084, + "step": 7414 + }, + { + "epoch": 1.35, + "learning_rate": 2.597834000728067e-06, + "loss": 1.2135, + "step": 7415 + }, + { + "epoch": 1.35, + "learning_rate": 2.597779395704405e-06, + "loss": 2.1981, + "step": 7416 + }, + { + "epoch": 1.35, + "learning_rate": 2.5977247906807427e-06, + "loss": 0.0341, + "step": 7417 + }, + { + "epoch": 1.35, + "learning_rate": 2.5976701856570806e-06, + "loss": 0.5985, + "step": 7418 + }, + { + "epoch": 1.35, + "learning_rate": 2.597615580633418e-06, + "loss": 0.2878, + "step": 7419 + }, + { + "epoch": 1.35, + "learning_rate": 2.597560975609756e-06, + "loss": 2.4518, + "step": 7420 + }, + { + "epoch": 1.35, + "learning_rate": 2.597506370586094e-06, + "loss": 0.1259, + "step": 7421 + }, + { + "epoch": 1.35, + "learning_rate": 2.5974517655624315e-06, + "loss": 2.1346, + "step": 7422 + }, + { + "epoch": 1.35, + "learning_rate": 2.5973971605387695e-06, + "loss": 0.8279, + "step": 7423 + }, + { + "epoch": 1.35, + "learning_rate": 2.5973425555151074e-06, + "loss": 0.1845, + "step": 7424 + }, + { + "epoch": 1.35, + "learning_rate": 2.5972879504914454e-06, + "loss": 0.5828, + "step": 7425 + }, + { + "epoch": 1.35, + "learning_rate": 2.5972333454677833e-06, + "loss": 0.4881, + "step": 7426 + }, + { + "epoch": 1.35, + "learning_rate": 2.5971787404441212e-06, + "loss": 0.2929, + "step": 7427 + }, + { + "epoch": 1.35, + "learning_rate": 2.5971241354204588e-06, + "loss": 0.401, + "step": 7428 + }, + { + "epoch": 1.35, + "learning_rate": 2.5970695303967967e-06, + "loss": 2.0066, + "step": 7429 + }, + { + "epoch": 1.35, + "learning_rate": 2.5970149253731342e-06, + "loss": 0.172, + "step": 7430 + }, + { + "epoch": 1.35, + "learning_rate": 2.596960320349472e-06, + "loss": 0.0348, + "step": 7431 + }, + { + "epoch": 1.35, + "learning_rate": 2.59690571532581e-06, + "loss": 1.5798, + "step": 7432 + }, + { + "epoch": 1.35, + "learning_rate": 2.5968511103021476e-06, + "loss": 1.4729, + "step": 7433 + }, + { + "epoch": 1.35, + "learning_rate": 2.5967965052784856e-06, + "loss": 0.4073, + "step": 7434 + }, + { + "epoch": 1.35, + "learning_rate": 2.5967419002548235e-06, + "loss": 2.4678, + "step": 7435 + }, + { + "epoch": 1.35, + "learning_rate": 2.596687295231161e-06, + "loss": 1.0783, + "step": 7436 + }, + { + "epoch": 1.35, + "learning_rate": 2.596632690207499e-06, + "loss": 0.3884, + "step": 7437 + }, + { + "epoch": 1.35, + "learning_rate": 2.596578085183837e-06, + "loss": 1.766, + "step": 7438 + }, + { + "epoch": 1.35, + "learning_rate": 2.596523480160175e-06, + "loss": 1.8984, + "step": 7439 + }, + { + "epoch": 1.35, + "learning_rate": 2.596468875136513e-06, + "loss": 0.2053, + "step": 7440 + }, + { + "epoch": 1.35, + "learning_rate": 2.5964142701128508e-06, + "loss": 0.5122, + "step": 7441 + }, + { + "epoch": 1.35, + "learning_rate": 2.5963596650891883e-06, + "loss": 0.852, + "step": 7442 + }, + { + "epoch": 1.35, + "learning_rate": 2.5963050600655262e-06, + "loss": 0.2235, + "step": 7443 + }, + { + "epoch": 1.35, + "learning_rate": 2.5962504550418637e-06, + "loss": 0.4645, + "step": 7444 + }, + { + "epoch": 1.35, + "learning_rate": 2.5961958500182017e-06, + "loss": 0.9241, + "step": 7445 + }, + { + "epoch": 1.35, + "learning_rate": 2.5961412449945396e-06, + "loss": 2.3573, + "step": 7446 + }, + { + "epoch": 1.35, + "learning_rate": 2.596086639970877e-06, + "loss": 0.611, + "step": 7447 + }, + { + "epoch": 1.35, + "learning_rate": 2.596032034947215e-06, + "loss": 0.627, + "step": 7448 + }, + { + "epoch": 1.35, + "learning_rate": 2.595977429923553e-06, + "loss": 1.4818, + "step": 7449 + }, + { + "epoch": 1.35, + "learning_rate": 2.5959228248998906e-06, + "loss": 0.554, + "step": 7450 + }, + { + "epoch": 1.35, + "learning_rate": 2.5958682198762285e-06, + "loss": 0.5239, + "step": 7451 + }, + { + "epoch": 1.36, + "learning_rate": 2.5958136148525664e-06, + "loss": 0.337, + "step": 7452 + }, + { + "epoch": 1.36, + "learning_rate": 2.5957590098289044e-06, + "loss": 0.5751, + "step": 7453 + }, + { + "epoch": 1.36, + "learning_rate": 2.5957044048052423e-06, + "loss": 0.8232, + "step": 7454 + }, + { + "epoch": 1.36, + "learning_rate": 2.5956497997815803e-06, + "loss": 1.1528, + "step": 7455 + }, + { + "epoch": 1.36, + "learning_rate": 2.5955951947579178e-06, + "loss": 0.773, + "step": 7456 + }, + { + "epoch": 1.36, + "learning_rate": 2.5955405897342557e-06, + "loss": 0.461, + "step": 7457 + }, + { + "epoch": 1.36, + "learning_rate": 2.5954859847105933e-06, + "loss": 0.1823, + "step": 7458 + }, + { + "epoch": 1.36, + "learning_rate": 2.595431379686931e-06, + "loss": 0.3167, + "step": 7459 + }, + { + "epoch": 1.36, + "learning_rate": 2.595376774663269e-06, + "loss": 0.6393, + "step": 7460 + }, + { + "epoch": 1.36, + "learning_rate": 2.5953221696396067e-06, + "loss": 1.6678, + "step": 7461 + }, + { + "epoch": 1.36, + "learning_rate": 2.5952675646159446e-06, + "loss": 1.2693, + "step": 7462 + }, + { + "epoch": 1.36, + "learning_rate": 2.5952129595922825e-06, + "loss": 0.8018, + "step": 7463 + }, + { + "epoch": 1.36, + "learning_rate": 2.59515835456862e-06, + "loss": 0.6198, + "step": 7464 + }, + { + "epoch": 1.36, + "learning_rate": 2.595103749544958e-06, + "loss": 3.7176, + "step": 7465 + }, + { + "epoch": 1.36, + "learning_rate": 2.595049144521296e-06, + "loss": 0.031, + "step": 7466 + }, + { + "epoch": 1.36, + "learning_rate": 2.594994539497634e-06, + "loss": 1.0798, + "step": 7467 + }, + { + "epoch": 1.36, + "learning_rate": 2.594939934473972e-06, + "loss": 0.9704, + "step": 7468 + }, + { + "epoch": 1.36, + "learning_rate": 2.5948853294503098e-06, + "loss": 0.1163, + "step": 7469 + }, + { + "epoch": 1.36, + "learning_rate": 2.5948307244266473e-06, + "loss": 1.2093, + "step": 7470 + }, + { + "epoch": 1.36, + "learning_rate": 2.5947761194029852e-06, + "loss": 3.0444, + "step": 7471 + }, + { + "epoch": 1.36, + "learning_rate": 2.5947215143793228e-06, + "loss": 0.56, + "step": 7472 + }, + { + "epoch": 1.36, + "learning_rate": 2.5946669093556607e-06, + "loss": 1.1566, + "step": 7473 + }, + { + "epoch": 1.36, + "learning_rate": 2.5946123043319987e-06, + "loss": 0.2994, + "step": 7474 + }, + { + "epoch": 1.36, + "learning_rate": 2.594557699308336e-06, + "loss": 1.32, + "step": 7475 + }, + { + "epoch": 1.36, + "learning_rate": 2.594503094284674e-06, + "loss": 0.4984, + "step": 7476 + }, + { + "epoch": 1.36, + "learning_rate": 2.594448489261012e-06, + "loss": 0.9675, + "step": 7477 + }, + { + "epoch": 1.36, + "learning_rate": 2.59439388423735e-06, + "loss": 0.2374, + "step": 7478 + }, + { + "epoch": 1.36, + "learning_rate": 2.594339279213688e-06, + "loss": 0.4646, + "step": 7479 + }, + { + "epoch": 1.36, + "learning_rate": 2.594284674190026e-06, + "loss": 1.3321, + "step": 7480 + }, + { + "epoch": 1.36, + "learning_rate": 2.5942300691663634e-06, + "loss": 2.2068, + "step": 7481 + }, + { + "epoch": 1.36, + "learning_rate": 2.5941754641427014e-06, + "loss": 0.2868, + "step": 7482 + }, + { + "epoch": 1.36, + "learning_rate": 2.5941208591190393e-06, + "loss": 0.6519, + "step": 7483 + }, + { + "epoch": 1.36, + "learning_rate": 2.594066254095377e-06, + "loss": 2.0537, + "step": 7484 + }, + { + "epoch": 1.36, + "learning_rate": 2.5940116490717148e-06, + "loss": 0.1407, + "step": 7485 + }, + { + "epoch": 1.36, + "learning_rate": 2.5939570440480523e-06, + "loss": 0.2485, + "step": 7486 + }, + { + "epoch": 1.36, + "learning_rate": 2.5939024390243902e-06, + "loss": 0.3367, + "step": 7487 + }, + { + "epoch": 1.36, + "learning_rate": 2.593847834000728e-06, + "loss": 0.7506, + "step": 7488 + }, + { + "epoch": 1.36, + "learning_rate": 2.5937932289770657e-06, + "loss": 0.2968, + "step": 7489 + }, + { + "epoch": 1.36, + "learning_rate": 2.5937386239534036e-06, + "loss": 0.6082, + "step": 7490 + }, + { + "epoch": 1.36, + "learning_rate": 2.5936840189297416e-06, + "loss": 1.1091, + "step": 7491 + }, + { + "epoch": 1.36, + "learning_rate": 2.5936294139060795e-06, + "loss": 0.3621, + "step": 7492 + }, + { + "epoch": 1.36, + "learning_rate": 2.5935748088824175e-06, + "loss": 0.6797, + "step": 7493 + }, + { + "epoch": 1.36, + "learning_rate": 2.5935202038587554e-06, + "loss": 0.5559, + "step": 7494 + }, + { + "epoch": 1.36, + "learning_rate": 2.593465598835093e-06, + "loss": 0.1041, + "step": 7495 + }, + { + "epoch": 1.36, + "learning_rate": 2.593410993811431e-06, + "loss": 0.1651, + "step": 7496 + }, + { + "epoch": 1.36, + "learning_rate": 2.593356388787769e-06, + "loss": 0.4326, + "step": 7497 + }, + { + "epoch": 1.36, + "learning_rate": 2.5933017837641063e-06, + "loss": 0.096, + "step": 7498 + }, + { + "epoch": 1.36, + "learning_rate": 2.5932471787404443e-06, + "loss": 1.0056, + "step": 7499 + }, + { + "epoch": 1.36, + "learning_rate": 2.593192573716782e-06, + "loss": 0.0762, + "step": 7500 + }, + { + "epoch": 1.36, + "learning_rate": 2.5931379686931197e-06, + "loss": 0.6775, + "step": 7501 + }, + { + "epoch": 1.36, + "learning_rate": 2.5930833636694577e-06, + "loss": 0.4527, + "step": 7502 + }, + { + "epoch": 1.36, + "learning_rate": 2.593028758645795e-06, + "loss": 1.0015, + "step": 7503 + }, + { + "epoch": 1.36, + "learning_rate": 2.592974153622133e-06, + "loss": 1.5717, + "step": 7504 + }, + { + "epoch": 1.36, + "learning_rate": 2.592919548598471e-06, + "loss": 2.4966, + "step": 7505 + }, + { + "epoch": 1.36, + "learning_rate": 2.592864943574809e-06, + "loss": 1.023, + "step": 7506 + }, + { + "epoch": 1.37, + "learning_rate": 2.592810338551147e-06, + "loss": 0.8633, + "step": 7507 + }, + { + "epoch": 1.37, + "learning_rate": 2.592755733527485e-06, + "loss": 1.6798, + "step": 7508 + }, + { + "epoch": 1.37, + "learning_rate": 2.5927011285038224e-06, + "loss": 1.4803, + "step": 7509 + }, + { + "epoch": 1.37, + "learning_rate": 2.5926465234801604e-06, + "loss": 1.2769, + "step": 7510 + }, + { + "epoch": 1.37, + "learning_rate": 2.592591918456498e-06, + "loss": 0.5071, + "step": 7511 + }, + { + "epoch": 1.37, + "learning_rate": 2.592537313432836e-06, + "loss": 0.8333, + "step": 7512 + }, + { + "epoch": 1.37, + "learning_rate": 2.5924827084091738e-06, + "loss": 2.2054, + "step": 7513 + }, + { + "epoch": 1.37, + "learning_rate": 2.5924281033855113e-06, + "loss": 1.1348, + "step": 7514 + }, + { + "epoch": 1.37, + "learning_rate": 2.5923734983618492e-06, + "loss": 1.3683, + "step": 7515 + }, + { + "epoch": 1.37, + "learning_rate": 2.592318893338187e-06, + "loss": 0.9478, + "step": 7516 + }, + { + "epoch": 1.37, + "learning_rate": 2.5922642883145247e-06, + "loss": 0.1979, + "step": 7517 + }, + { + "epoch": 1.37, + "learning_rate": 2.5922096832908627e-06, + "loss": 1.5389, + "step": 7518 + }, + { + "epoch": 1.37, + "learning_rate": 2.5921550782672006e-06, + "loss": 1.7767, + "step": 7519 + }, + { + "epoch": 1.37, + "learning_rate": 2.5921004732435385e-06, + "loss": 1.1705, + "step": 7520 + }, + { + "epoch": 1.37, + "learning_rate": 2.5920458682198765e-06, + "loss": 2.3189, + "step": 7521 + }, + { + "epoch": 1.37, + "learning_rate": 2.5919912631962144e-06, + "loss": 2.848, + "step": 7522 + }, + { + "epoch": 1.37, + "learning_rate": 2.591936658172552e-06, + "loss": 0.9614, + "step": 7523 + }, + { + "epoch": 1.37, + "learning_rate": 2.59188205314889e-06, + "loss": 1.4393, + "step": 7524 + }, + { + "epoch": 1.37, + "learning_rate": 2.5918274481252274e-06, + "loss": 1.2969, + "step": 7525 + }, + { + "epoch": 1.37, + "learning_rate": 2.5917728431015654e-06, + "loss": 0.6603, + "step": 7526 + }, + { + "epoch": 1.37, + "learning_rate": 2.5917182380779033e-06, + "loss": 1.2794, + "step": 7527 + }, + { + "epoch": 1.37, + "learning_rate": 2.591663633054241e-06, + "loss": 0.7428, + "step": 7528 + }, + { + "epoch": 1.37, + "learning_rate": 2.5916090280305788e-06, + "loss": 0.7516, + "step": 7529 + }, + { + "epoch": 1.37, + "learning_rate": 2.5915544230069167e-06, + "loss": 1.0652, + "step": 7530 + }, + { + "epoch": 1.37, + "learning_rate": 2.5914998179832542e-06, + "loss": 1.6951, + "step": 7531 + }, + { + "epoch": 1.37, + "learning_rate": 2.591445212959592e-06, + "loss": 0.425, + "step": 7532 + }, + { + "epoch": 1.37, + "learning_rate": 2.59139060793593e-06, + "loss": 1.6338, + "step": 7533 + }, + { + "epoch": 1.37, + "learning_rate": 2.591336002912268e-06, + "loss": 0.66, + "step": 7534 + }, + { + "epoch": 1.37, + "learning_rate": 2.591281397888606e-06, + "loss": 0.0574, + "step": 7535 + }, + { + "epoch": 1.37, + "learning_rate": 2.591226792864944e-06, + "loss": 0.5426, + "step": 7536 + }, + { + "epoch": 1.37, + "learning_rate": 2.5911721878412815e-06, + "loss": 0.5533, + "step": 7537 + }, + { + "epoch": 1.37, + "learning_rate": 2.5911175828176194e-06, + "loss": 0.5352, + "step": 7538 + }, + { + "epoch": 1.37, + "learning_rate": 2.591062977793957e-06, + "loss": 0.0691, + "step": 7539 + }, + { + "epoch": 1.37, + "learning_rate": 2.591008372770295e-06, + "loss": 1.341, + "step": 7540 + }, + { + "epoch": 1.37, + "learning_rate": 2.590953767746633e-06, + "loss": 0.1618, + "step": 7541 + }, + { + "epoch": 1.37, + "learning_rate": 2.5908991627229703e-06, + "loss": 1.2903, + "step": 7542 + }, + { + "epoch": 1.37, + "learning_rate": 2.5908445576993083e-06, + "loss": 0.5099, + "step": 7543 + }, + { + "epoch": 1.37, + "learning_rate": 2.5907899526756462e-06, + "loss": 0.8356, + "step": 7544 + }, + { + "epoch": 1.37, + "learning_rate": 2.5907353476519837e-06, + "loss": 0.6772, + "step": 7545 + }, + { + "epoch": 1.37, + "learning_rate": 2.5906807426283217e-06, + "loss": 1.3399, + "step": 7546 + }, + { + "epoch": 1.37, + "learning_rate": 2.5906261376046596e-06, + "loss": 0.5418, + "step": 7547 + }, + { + "epoch": 1.37, + "learning_rate": 2.5905715325809976e-06, + "loss": 1.9663, + "step": 7548 + }, + { + "epoch": 1.37, + "learning_rate": 2.5905169275573355e-06, + "loss": 0.0389, + "step": 7549 + }, + { + "epoch": 1.37, + "learning_rate": 2.5904623225336735e-06, + "loss": 0.9295, + "step": 7550 + }, + { + "epoch": 1.37, + "learning_rate": 2.590407717510011e-06, + "loss": 0.7973, + "step": 7551 + }, + { + "epoch": 1.37, + "learning_rate": 2.590353112486349e-06, + "loss": 0.811, + "step": 7552 + }, + { + "epoch": 1.37, + "learning_rate": 2.5902985074626864e-06, + "loss": 0.6582, + "step": 7553 + }, + { + "epoch": 1.37, + "learning_rate": 2.5902439024390244e-06, + "loss": 1.8205, + "step": 7554 + }, + { + "epoch": 1.37, + "learning_rate": 2.5901892974153623e-06, + "loss": 0.204, + "step": 7555 + }, + { + "epoch": 1.37, + "learning_rate": 2.5901346923917e-06, + "loss": 1.0866, + "step": 7556 + }, + { + "epoch": 1.37, + "learning_rate": 2.5900800873680378e-06, + "loss": 1.5618, + "step": 7557 + }, + { + "epoch": 1.37, + "learning_rate": 2.5900254823443757e-06, + "loss": 0.8761, + "step": 7558 + }, + { + "epoch": 1.37, + "learning_rate": 2.5899708773207137e-06, + "loss": 0.1983, + "step": 7559 + }, + { + "epoch": 1.37, + "learning_rate": 2.5899162722970516e-06, + "loss": 2.6447, + "step": 7560 + }, + { + "epoch": 1.37, + "learning_rate": 2.5898616672733896e-06, + "loss": 0.2277, + "step": 7561 + }, + { + "epoch": 1.38, + "learning_rate": 2.589807062249727e-06, + "loss": 0.432, + "step": 7562 + }, + { + "epoch": 1.38, + "learning_rate": 2.589752457226065e-06, + "loss": 0.2697, + "step": 7563 + }, + { + "epoch": 1.38, + "learning_rate": 2.589697852202403e-06, + "loss": 0.6827, + "step": 7564 + }, + { + "epoch": 1.38, + "learning_rate": 2.5896432471787405e-06, + "loss": 2.0685, + "step": 7565 + }, + { + "epoch": 1.38, + "learning_rate": 2.5895886421550784e-06, + "loss": 0.9794, + "step": 7566 + }, + { + "epoch": 1.38, + "learning_rate": 2.589534037131416e-06, + "loss": 0.9417, + "step": 7567 + }, + { + "epoch": 1.38, + "learning_rate": 2.589479432107754e-06, + "loss": 0.6692, + "step": 7568 + }, + { + "epoch": 1.38, + "learning_rate": 2.589424827084092e-06, + "loss": 1.5427, + "step": 7569 + }, + { + "epoch": 1.38, + "learning_rate": 2.5893702220604294e-06, + "loss": 0.9316, + "step": 7570 + }, + { + "epoch": 1.38, + "learning_rate": 2.5893156170367673e-06, + "loss": 1.9406, + "step": 7571 + }, + { + "epoch": 1.38, + "learning_rate": 2.5892610120131052e-06, + "loss": 0.9244, + "step": 7572 + }, + { + "epoch": 1.38, + "learning_rate": 2.589206406989443e-06, + "loss": 0.9479, + "step": 7573 + }, + { + "epoch": 1.38, + "learning_rate": 2.589151801965781e-06, + "loss": 0.2832, + "step": 7574 + }, + { + "epoch": 1.38, + "learning_rate": 2.589097196942119e-06, + "loss": 0.0682, + "step": 7575 + }, + { + "epoch": 1.38, + "learning_rate": 2.5890425919184566e-06, + "loss": 0.8282, + "step": 7576 + }, + { + "epoch": 1.38, + "learning_rate": 2.5889879868947945e-06, + "loss": 0.8955, + "step": 7577 + }, + { + "epoch": 1.38, + "learning_rate": 2.588933381871132e-06, + "loss": 2.1928, + "step": 7578 + }, + { + "epoch": 1.38, + "learning_rate": 2.58887877684747e-06, + "loss": 0.1979, + "step": 7579 + }, + { + "epoch": 1.38, + "learning_rate": 2.588824171823808e-06, + "loss": 0.5538, + "step": 7580 + }, + { + "epoch": 1.38, + "learning_rate": 2.5887695668001455e-06, + "loss": 3.282, + "step": 7581 + }, + { + "epoch": 1.38, + "learning_rate": 2.5887149617764834e-06, + "loss": 0.6603, + "step": 7582 + }, + { + "epoch": 1.38, + "learning_rate": 2.5886603567528214e-06, + "loss": 0.9674, + "step": 7583 + }, + { + "epoch": 1.38, + "learning_rate": 2.588605751729159e-06, + "loss": 1.4793, + "step": 7584 + }, + { + "epoch": 1.38, + "learning_rate": 2.588551146705497e-06, + "loss": 0.0642, + "step": 7585 + }, + { + "epoch": 1.38, + "learning_rate": 2.5884965416818348e-06, + "loss": 0.6585, + "step": 7586 + }, + { + "epoch": 1.38, + "learning_rate": 2.5884419366581727e-06, + "loss": 0.3447, + "step": 7587 + }, + { + "epoch": 1.38, + "learning_rate": 2.5883873316345106e-06, + "loss": 3.1969, + "step": 7588 + }, + { + "epoch": 1.38, + "learning_rate": 2.5883327266108486e-06, + "loss": 1.3881, + "step": 7589 + }, + { + "epoch": 1.38, + "learning_rate": 2.588278121587186e-06, + "loss": 0.1464, + "step": 7590 + }, + { + "epoch": 1.38, + "learning_rate": 2.588223516563524e-06, + "loss": 0.5785, + "step": 7591 + }, + { + "epoch": 1.38, + "learning_rate": 2.5881689115398616e-06, + "loss": 0.1795, + "step": 7592 + }, + { + "epoch": 1.38, + "learning_rate": 2.5881143065161995e-06, + "loss": 1.136, + "step": 7593 + }, + { + "epoch": 1.38, + "learning_rate": 2.5880597014925375e-06, + "loss": 0.6596, + "step": 7594 + }, + { + "epoch": 1.38, + "learning_rate": 2.588005096468875e-06, + "loss": 0.0435, + "step": 7595 + }, + { + "epoch": 1.38, + "learning_rate": 2.587950491445213e-06, + "loss": 3.117, + "step": 7596 + }, + { + "epoch": 1.38, + "learning_rate": 2.587895886421551e-06, + "loss": 0.5978, + "step": 7597 + }, + { + "epoch": 1.38, + "learning_rate": 2.5878412813978884e-06, + "loss": 0.368, + "step": 7598 + }, + { + "epoch": 1.38, + "learning_rate": 2.5877866763742263e-06, + "loss": 0.3411, + "step": 7599 + }, + { + "epoch": 1.38, + "learning_rate": 2.5877320713505643e-06, + "loss": 0.4955, + "step": 7600 + }, + { + "epoch": 1.38, + "learning_rate": 2.5876774663269022e-06, + "loss": 0.0388, + "step": 7601 + }, + { + "epoch": 1.38, + "learning_rate": 2.58762286130324e-06, + "loss": 2.5188, + "step": 7602 + }, + { + "epoch": 1.38, + "learning_rate": 2.587568256279578e-06, + "loss": 0.058, + "step": 7603 + }, + { + "epoch": 1.38, + "learning_rate": 2.5875136512559156e-06, + "loss": 0.2949, + "step": 7604 + }, + { + "epoch": 1.38, + "learning_rate": 2.5874590462322536e-06, + "loss": 0.8866, + "step": 7605 + }, + { + "epoch": 1.38, + "learning_rate": 2.587404441208591e-06, + "loss": 0.084, + "step": 7606 + }, + { + "epoch": 1.38, + "learning_rate": 2.587349836184929e-06, + "loss": 0.6133, + "step": 7607 + }, + { + "epoch": 1.38, + "learning_rate": 2.587295231161267e-06, + "loss": 0.6659, + "step": 7608 + }, + { + "epoch": 1.38, + "learning_rate": 2.5872406261376045e-06, + "loss": 1.001, + "step": 7609 + }, + { + "epoch": 1.38, + "learning_rate": 2.5871860211139424e-06, + "loss": 0.4534, + "step": 7610 + }, + { + "epoch": 1.38, + "learning_rate": 2.5871314160902804e-06, + "loss": 0.6801, + "step": 7611 + }, + { + "epoch": 1.38, + "learning_rate": 2.587076811066618e-06, + "loss": 2.3874, + "step": 7612 + }, + { + "epoch": 1.38, + "learning_rate": 2.587022206042956e-06, + "loss": 1.3999, + "step": 7613 + }, + { + "epoch": 1.38, + "learning_rate": 2.5869676010192938e-06, + "loss": 0.3434, + "step": 7614 + }, + { + "epoch": 1.38, + "learning_rate": 2.5869129959956317e-06, + "loss": 0.9725, + "step": 7615 + }, + { + "epoch": 1.38, + "learning_rate": 2.5868583909719697e-06, + "loss": 1.4629, + "step": 7616 + }, + { + "epoch": 1.39, + "learning_rate": 2.5868037859483076e-06, + "loss": 1.5811, + "step": 7617 + }, + { + "epoch": 1.39, + "learning_rate": 2.586749180924645e-06, + "loss": 0.2802, + "step": 7618 + }, + { + "epoch": 1.39, + "learning_rate": 2.586694575900983e-06, + "loss": 0.9686, + "step": 7619 + }, + { + "epoch": 1.39, + "learning_rate": 2.5866399708773206e-06, + "loss": 0.8258, + "step": 7620 + }, + { + "epoch": 1.39, + "learning_rate": 2.5865853658536585e-06, + "loss": 0.5154, + "step": 7621 + }, + { + "epoch": 1.39, + "learning_rate": 2.5865307608299965e-06, + "loss": 0.2693, + "step": 7622 + }, + { + "epoch": 1.39, + "learning_rate": 2.586476155806334e-06, + "loss": 1.4003, + "step": 7623 + }, + { + "epoch": 1.39, + "learning_rate": 2.586421550782672e-06, + "loss": 1.0886, + "step": 7624 + }, + { + "epoch": 1.39, + "learning_rate": 2.58636694575901e-06, + "loss": 0.0372, + "step": 7625 + }, + { + "epoch": 1.39, + "learning_rate": 2.5863123407353474e-06, + "loss": 1.4878, + "step": 7626 + }, + { + "epoch": 1.39, + "learning_rate": 2.5862577357116854e-06, + "loss": 0.4173, + "step": 7627 + }, + { + "epoch": 1.39, + "learning_rate": 2.5862031306880233e-06, + "loss": 1.2344, + "step": 7628 + }, + { + "epoch": 1.39, + "learning_rate": 2.5861485256643612e-06, + "loss": 0.4504, + "step": 7629 + }, + { + "epoch": 1.39, + "learning_rate": 2.586093920640699e-06, + "loss": 1.3154, + "step": 7630 + }, + { + "epoch": 1.39, + "learning_rate": 2.586039315617037e-06, + "loss": 1.2598, + "step": 7631 + }, + { + "epoch": 1.39, + "learning_rate": 2.5859847105933746e-06, + "loss": 1.2674, + "step": 7632 + }, + { + "epoch": 1.39, + "learning_rate": 2.5859301055697126e-06, + "loss": 0.118, + "step": 7633 + }, + { + "epoch": 1.39, + "learning_rate": 2.58587550054605e-06, + "loss": 1.085, + "step": 7634 + }, + { + "epoch": 1.39, + "learning_rate": 2.585820895522388e-06, + "loss": 0.3148, + "step": 7635 + }, + { + "epoch": 1.39, + "learning_rate": 2.585766290498726e-06, + "loss": 1.2797, + "step": 7636 + }, + { + "epoch": 1.39, + "learning_rate": 2.5857116854750635e-06, + "loss": 0.9861, + "step": 7637 + }, + { + "epoch": 1.39, + "learning_rate": 2.5856570804514015e-06, + "loss": 0.349, + "step": 7638 + }, + { + "epoch": 1.39, + "learning_rate": 2.5856024754277394e-06, + "loss": 0.9481, + "step": 7639 + }, + { + "epoch": 1.39, + "learning_rate": 2.5855478704040773e-06, + "loss": 0.7545, + "step": 7640 + }, + { + "epoch": 1.39, + "learning_rate": 2.5854932653804153e-06, + "loss": 1.1713, + "step": 7641 + }, + { + "epoch": 1.39, + "learning_rate": 2.5854386603567532e-06, + "loss": 1.0772, + "step": 7642 + }, + { + "epoch": 1.39, + "learning_rate": 2.5853840553330908e-06, + "loss": 2.9694, + "step": 7643 + }, + { + "epoch": 1.39, + "learning_rate": 2.5853294503094287e-06, + "loss": 2.5651, + "step": 7644 + }, + { + "epoch": 1.39, + "learning_rate": 2.5852748452857666e-06, + "loss": 0.0312, + "step": 7645 + }, + { + "epoch": 1.39, + "learning_rate": 2.585220240262104e-06, + "loss": 1.7307, + "step": 7646 + }, + { + "epoch": 1.39, + "learning_rate": 2.585165635238442e-06, + "loss": 0.4578, + "step": 7647 + }, + { + "epoch": 1.39, + "learning_rate": 2.5851110302147796e-06, + "loss": 0.8689, + "step": 7648 + }, + { + "epoch": 1.39, + "learning_rate": 2.5850564251911176e-06, + "loss": 1.0763, + "step": 7649 + }, + { + "epoch": 1.39, + "learning_rate": 2.5850018201674555e-06, + "loss": 1.3353, + "step": 7650 + }, + { + "epoch": 1.39, + "learning_rate": 2.584947215143793e-06, + "loss": 0.054, + "step": 7651 + }, + { + "epoch": 1.39, + "learning_rate": 2.584892610120131e-06, + "loss": 1.3047, + "step": 7652 + }, + { + "epoch": 1.39, + "learning_rate": 2.584838005096469e-06, + "loss": 0.7876, + "step": 7653 + }, + { + "epoch": 1.39, + "learning_rate": 2.584783400072807e-06, + "loss": 1.6462, + "step": 7654 + }, + { + "epoch": 1.39, + "learning_rate": 2.584728795049145e-06, + "loss": 2.8877, + "step": 7655 + }, + { + "epoch": 1.39, + "learning_rate": 2.5846741900254827e-06, + "loss": 0.7686, + "step": 7656 + }, + { + "epoch": 1.39, + "learning_rate": 2.5846195850018203e-06, + "loss": 0.761, + "step": 7657 + }, + { + "epoch": 1.39, + "learning_rate": 2.5845649799781582e-06, + "loss": 0.9066, + "step": 7658 + }, + { + "epoch": 1.39, + "learning_rate": 2.5845103749544957e-06, + "loss": 0.0438, + "step": 7659 + }, + { + "epoch": 1.39, + "learning_rate": 2.5844557699308337e-06, + "loss": 1.0842, + "step": 7660 + }, + { + "epoch": 1.39, + "learning_rate": 2.5844011649071716e-06, + "loss": 0.3392, + "step": 7661 + }, + { + "epoch": 1.39, + "learning_rate": 2.584346559883509e-06, + "loss": 0.6942, + "step": 7662 + }, + { + "epoch": 1.39, + "learning_rate": 2.584291954859847e-06, + "loss": 0.6354, + "step": 7663 + }, + { + "epoch": 1.39, + "learning_rate": 2.584237349836185e-06, + "loss": 1.7051, + "step": 7664 + }, + { + "epoch": 1.39, + "learning_rate": 2.5841827448125225e-06, + "loss": 0.8172, + "step": 7665 + }, + { + "epoch": 1.39, + "learning_rate": 2.5841281397888605e-06, + "loss": 1.4234, + "step": 7666 + }, + { + "epoch": 1.39, + "learning_rate": 2.5840735347651984e-06, + "loss": 0.0324, + "step": 7667 + }, + { + "epoch": 1.39, + "learning_rate": 2.5840189297415364e-06, + "loss": 1.604, + "step": 7668 + }, + { + "epoch": 1.39, + "learning_rate": 2.5839643247178743e-06, + "loss": 1.6271, + "step": 7669 + }, + { + "epoch": 1.39, + "learning_rate": 2.5839097196942123e-06, + "loss": 0.7909, + "step": 7670 + }, + { + "epoch": 1.39, + "learning_rate": 2.5838551146705498e-06, + "loss": 1.3274, + "step": 7671 + }, + { + "epoch": 1.4, + "learning_rate": 2.5838005096468877e-06, + "loss": 0.0557, + "step": 7672 + }, + { + "epoch": 1.4, + "learning_rate": 2.5837459046232252e-06, + "loss": 1.4153, + "step": 7673 + }, + { + "epoch": 1.4, + "learning_rate": 2.583691299599563e-06, + "loss": 1.4142, + "step": 7674 + }, + { + "epoch": 1.4, + "learning_rate": 2.583636694575901e-06, + "loss": 2.8961, + "step": 7675 + }, + { + "epoch": 1.4, + "learning_rate": 2.5835820895522387e-06, + "loss": 0.3418, + "step": 7676 + }, + { + "epoch": 1.4, + "learning_rate": 2.5835274845285766e-06, + "loss": 0.3382, + "step": 7677 + }, + { + "epoch": 1.4, + "learning_rate": 2.5834728795049145e-06, + "loss": 1.2954, + "step": 7678 + }, + { + "epoch": 1.4, + "learning_rate": 2.583418274481252e-06, + "loss": 0.0631, + "step": 7679 + }, + { + "epoch": 1.4, + "learning_rate": 2.58336366945759e-06, + "loss": 2.6458, + "step": 7680 + }, + { + "epoch": 1.4, + "learning_rate": 2.583309064433928e-06, + "loss": 0.5811, + "step": 7681 + }, + { + "epoch": 1.4, + "learning_rate": 2.583254459410266e-06, + "loss": 1.4824, + "step": 7682 + }, + { + "epoch": 1.4, + "learning_rate": 2.583199854386604e-06, + "loss": 1.4462, + "step": 7683 + }, + { + "epoch": 1.4, + "learning_rate": 2.5831452493629418e-06, + "loss": 1.4859, + "step": 7684 + }, + { + "epoch": 1.4, + "learning_rate": 2.5830906443392793e-06, + "loss": 0.415, + "step": 7685 + }, + { + "epoch": 1.4, + "learning_rate": 2.5830360393156172e-06, + "loss": 0.0245, + "step": 7686 + }, + { + "epoch": 1.4, + "learning_rate": 2.5829814342919548e-06, + "loss": 0.8953, + "step": 7687 + }, + { + "epoch": 1.4, + "learning_rate": 2.5829268292682927e-06, + "loss": 0.4874, + "step": 7688 + }, + { + "epoch": 1.4, + "learning_rate": 2.5828722242446306e-06, + "loss": 1.4512, + "step": 7689 + }, + { + "epoch": 1.4, + "learning_rate": 2.582817619220968e-06, + "loss": 1.5364, + "step": 7690 + }, + { + "epoch": 1.4, + "learning_rate": 2.582763014197306e-06, + "loss": 1.7332, + "step": 7691 + }, + { + "epoch": 1.4, + "learning_rate": 2.582708409173644e-06, + "loss": 1.2338, + "step": 7692 + }, + { + "epoch": 1.4, + "learning_rate": 2.5826538041499816e-06, + "loss": 1.917, + "step": 7693 + }, + { + "epoch": 1.4, + "learning_rate": 2.5825991991263195e-06, + "loss": 1.2486, + "step": 7694 + }, + { + "epoch": 1.4, + "learning_rate": 2.5825445941026575e-06, + "loss": 0.3015, + "step": 7695 + }, + { + "epoch": 1.4, + "learning_rate": 2.5824899890789954e-06, + "loss": 0.8198, + "step": 7696 + }, + { + "epoch": 1.4, + "learning_rate": 2.5824353840553333e-06, + "loss": 0.2696, + "step": 7697 + }, + { + "epoch": 1.4, + "learning_rate": 2.5823807790316713e-06, + "loss": 0.5636, + "step": 7698 + }, + { + "epoch": 1.4, + "learning_rate": 2.582326174008009e-06, + "loss": 2.0271, + "step": 7699 + }, + { + "epoch": 1.4, + "learning_rate": 2.5822715689843468e-06, + "loss": 1.5827, + "step": 7700 + }, + { + "epoch": 1.4, + "learning_rate": 2.5822169639606843e-06, + "loss": 1.1875, + "step": 7701 + }, + { + "epoch": 1.4, + "learning_rate": 2.5821623589370222e-06, + "loss": 0.8489, + "step": 7702 + }, + { + "epoch": 1.4, + "learning_rate": 2.58210775391336e-06, + "loss": 3.2438, + "step": 7703 + }, + { + "epoch": 1.4, + "learning_rate": 2.5820531488896977e-06, + "loss": 0.4424, + "step": 7704 + }, + { + "epoch": 1.4, + "learning_rate": 2.5819985438660356e-06, + "loss": 0.0593, + "step": 7705 + }, + { + "epoch": 1.4, + "learning_rate": 2.5819439388423736e-06, + "loss": 2.2454, + "step": 7706 + }, + { + "epoch": 1.4, + "learning_rate": 2.581889333818711e-06, + "loss": 0.0781, + "step": 7707 + }, + { + "epoch": 1.4, + "learning_rate": 2.581834728795049e-06, + "loss": 0.6877, + "step": 7708 + }, + { + "epoch": 1.4, + "learning_rate": 2.581780123771387e-06, + "loss": 0.4798, + "step": 7709 + }, + { + "epoch": 1.4, + "learning_rate": 2.581725518747725e-06, + "loss": 0.4506, + "step": 7710 + }, + { + "epoch": 1.4, + "learning_rate": 2.581670913724063e-06, + "loss": 0.9357, + "step": 7711 + }, + { + "epoch": 1.4, + "learning_rate": 2.581616308700401e-06, + "loss": 0.2578, + "step": 7712 + }, + { + "epoch": 1.4, + "learning_rate": 2.5815617036767383e-06, + "loss": 0.6568, + "step": 7713 + }, + { + "epoch": 1.4, + "learning_rate": 2.5815070986530763e-06, + "loss": 0.1515, + "step": 7714 + }, + { + "epoch": 1.4, + "learning_rate": 2.5814524936294138e-06, + "loss": 0.5596, + "step": 7715 + }, + { + "epoch": 1.4, + "learning_rate": 2.5813978886057517e-06, + "loss": 1.6709, + "step": 7716 + }, + { + "epoch": 1.4, + "learning_rate": 2.5813432835820897e-06, + "loss": 0.0358, + "step": 7717 + }, + { + "epoch": 1.4, + "learning_rate": 2.581288678558427e-06, + "loss": 0.5835, + "step": 7718 + }, + { + "epoch": 1.4, + "learning_rate": 2.581234073534765e-06, + "loss": 0.3168, + "step": 7719 + }, + { + "epoch": 1.4, + "learning_rate": 2.581179468511103e-06, + "loss": 0.1574, + "step": 7720 + }, + { + "epoch": 1.4, + "learning_rate": 2.5811248634874406e-06, + "loss": 0.2973, + "step": 7721 + }, + { + "epoch": 1.4, + "learning_rate": 2.5810702584637785e-06, + "loss": 1.8216, + "step": 7722 + }, + { + "epoch": 1.4, + "learning_rate": 2.5810156534401165e-06, + "loss": 0.5241, + "step": 7723 + }, + { + "epoch": 1.4, + "learning_rate": 2.5809610484164544e-06, + "loss": 0.0768, + "step": 7724 + }, + { + "epoch": 1.4, + "learning_rate": 2.5809064433927924e-06, + "loss": 0.9288, + "step": 7725 + }, + { + "epoch": 1.4, + "learning_rate": 2.5808518383691303e-06, + "loss": 0.5691, + "step": 7726 + }, + { + "epoch": 1.41, + "learning_rate": 2.580797233345468e-06, + "loss": 0.4846, + "step": 7727 + }, + { + "epoch": 1.41, + "learning_rate": 2.5807426283218058e-06, + "loss": 1.7544, + "step": 7728 + }, + { + "epoch": 1.41, + "learning_rate": 2.5806880232981433e-06, + "loss": 0.3334, + "step": 7729 + }, + { + "epoch": 1.41, + "learning_rate": 2.5806334182744812e-06, + "loss": 0.3269, + "step": 7730 + }, + { + "epoch": 1.41, + "learning_rate": 2.580578813250819e-06, + "loss": 0.6493, + "step": 7731 + }, + { + "epoch": 1.41, + "learning_rate": 2.5805242082271567e-06, + "loss": 1.752, + "step": 7732 + }, + { + "epoch": 1.41, + "learning_rate": 2.5804696032034947e-06, + "loss": 0.7152, + "step": 7733 + }, + { + "epoch": 1.41, + "learning_rate": 2.5804149981798326e-06, + "loss": 0.751, + "step": 7734 + }, + { + "epoch": 1.41, + "learning_rate": 2.5803603931561705e-06, + "loss": 0.5541, + "step": 7735 + }, + { + "epoch": 1.41, + "learning_rate": 2.5803057881325085e-06, + "loss": 2.2162, + "step": 7736 + }, + { + "epoch": 1.41, + "learning_rate": 2.5802511831088464e-06, + "loss": 0.3726, + "step": 7737 + }, + { + "epoch": 1.41, + "learning_rate": 2.580196578085184e-06, + "loss": 0.7767, + "step": 7738 + }, + { + "epoch": 1.41, + "learning_rate": 2.580141973061522e-06, + "loss": 2.0316, + "step": 7739 + }, + { + "epoch": 1.41, + "learning_rate": 2.5800873680378594e-06, + "loss": 0.0316, + "step": 7740 + }, + { + "epoch": 1.41, + "learning_rate": 2.5800327630141974e-06, + "loss": 1.6363, + "step": 7741 + }, + { + "epoch": 1.41, + "learning_rate": 2.5799781579905353e-06, + "loss": 1.3881, + "step": 7742 + }, + { + "epoch": 1.41, + "learning_rate": 2.579923552966873e-06, + "loss": 0.9495, + "step": 7743 + }, + { + "epoch": 1.41, + "learning_rate": 2.5798689479432108e-06, + "loss": 0.3817, + "step": 7744 + }, + { + "epoch": 1.41, + "learning_rate": 2.5798143429195487e-06, + "loss": 0.0508, + "step": 7745 + }, + { + "epoch": 1.41, + "learning_rate": 2.5797597378958862e-06, + "loss": 0.15, + "step": 7746 + }, + { + "epoch": 1.41, + "learning_rate": 2.579705132872224e-06, + "loss": 0.8247, + "step": 7747 + }, + { + "epoch": 1.41, + "learning_rate": 2.579650527848562e-06, + "loss": 3.2345, + "step": 7748 + }, + { + "epoch": 1.41, + "learning_rate": 2.5795959228249e-06, + "loss": 1.9047, + "step": 7749 + }, + { + "epoch": 1.41, + "learning_rate": 2.579541317801238e-06, + "loss": 0.5724, + "step": 7750 + }, + { + "epoch": 1.41, + "learning_rate": 2.579486712777576e-06, + "loss": 0.822, + "step": 7751 + }, + { + "epoch": 1.41, + "learning_rate": 2.5794321077539135e-06, + "loss": 1.212, + "step": 7752 + }, + { + "epoch": 1.41, + "learning_rate": 2.5793775027302514e-06, + "loss": 1.4971, + "step": 7753 + }, + { + "epoch": 1.41, + "learning_rate": 2.579322897706589e-06, + "loss": 0.0995, + "step": 7754 + }, + { + "epoch": 1.41, + "learning_rate": 2.579268292682927e-06, + "loss": 1.1321, + "step": 7755 + }, + { + "epoch": 1.41, + "learning_rate": 2.579213687659265e-06, + "loss": 1.7482, + "step": 7756 + }, + { + "epoch": 1.41, + "learning_rate": 2.5791590826356023e-06, + "loss": 0.0623, + "step": 7757 + }, + { + "epoch": 1.41, + "learning_rate": 2.5791044776119403e-06, + "loss": 1.4759, + "step": 7758 + }, + { + "epoch": 1.41, + "learning_rate": 2.5790498725882782e-06, + "loss": 0.3561, + "step": 7759 + }, + { + "epoch": 1.41, + "learning_rate": 2.5789952675646157e-06, + "loss": 1.7452, + "step": 7760 + }, + { + "epoch": 1.41, + "learning_rate": 2.5789406625409537e-06, + "loss": 0.8702, + "step": 7761 + }, + { + "epoch": 1.41, + "learning_rate": 2.5788860575172916e-06, + "loss": 1.0285, + "step": 7762 + }, + { + "epoch": 1.41, + "learning_rate": 2.5788314524936296e-06, + "loss": 1.0639, + "step": 7763 + }, + { + "epoch": 1.41, + "learning_rate": 2.5787768474699675e-06, + "loss": 0.8846, + "step": 7764 + }, + { + "epoch": 1.41, + "learning_rate": 2.5787222424463055e-06, + "loss": 0.0782, + "step": 7765 + }, + { + "epoch": 1.41, + "learning_rate": 2.578667637422643e-06, + "loss": 1.044, + "step": 7766 + }, + { + "epoch": 1.41, + "learning_rate": 2.578613032398981e-06, + "loss": 0.0378, + "step": 7767 + }, + { + "epoch": 1.41, + "learning_rate": 2.5785584273753184e-06, + "loss": 0.0676, + "step": 7768 + }, + { + "epoch": 1.41, + "learning_rate": 2.5785038223516564e-06, + "loss": 1.506, + "step": 7769 + }, + { + "epoch": 1.41, + "learning_rate": 2.5784492173279943e-06, + "loss": 1.794, + "step": 7770 + }, + { + "epoch": 1.41, + "learning_rate": 2.578394612304332e-06, + "loss": 0.6102, + "step": 7771 + }, + { + "epoch": 1.41, + "learning_rate": 2.5783400072806698e-06, + "loss": 0.1257, + "step": 7772 + }, + { + "epoch": 1.41, + "learning_rate": 2.5782854022570077e-06, + "loss": 1.3238, + "step": 7773 + }, + { + "epoch": 1.41, + "learning_rate": 2.5782307972333452e-06, + "loss": 0.5419, + "step": 7774 + }, + { + "epoch": 1.41, + "learning_rate": 2.578176192209683e-06, + "loss": 1.247, + "step": 7775 + }, + { + "epoch": 1.41, + "learning_rate": 2.578121587186021e-06, + "loss": 1.5356, + "step": 7776 + }, + { + "epoch": 1.41, + "learning_rate": 2.578066982162359e-06, + "loss": 2.1581, + "step": 7777 + }, + { + "epoch": 1.41, + "learning_rate": 2.578012377138697e-06, + "loss": 1.4657, + "step": 7778 + }, + { + "epoch": 1.41, + "learning_rate": 2.577957772115035e-06, + "loss": 0.5137, + "step": 7779 + }, + { + "epoch": 1.41, + "learning_rate": 2.5779031670913725e-06, + "loss": 1.9304, + "step": 7780 + }, + { + "epoch": 1.41, + "learning_rate": 2.5778485620677104e-06, + "loss": 2.7952, + "step": 7781 + }, + { + "epoch": 1.42, + "learning_rate": 2.577793957044048e-06, + "loss": 0.5123, + "step": 7782 + }, + { + "epoch": 1.42, + "learning_rate": 2.577739352020386e-06, + "loss": 2.3079, + "step": 7783 + }, + { + "epoch": 1.42, + "learning_rate": 2.577684746996724e-06, + "loss": 0.7095, + "step": 7784 + }, + { + "epoch": 1.42, + "learning_rate": 2.5776301419730614e-06, + "loss": 2.3187, + "step": 7785 + }, + { + "epoch": 1.42, + "learning_rate": 2.5775755369493993e-06, + "loss": 0.4707, + "step": 7786 + }, + { + "epoch": 1.42, + "learning_rate": 2.5775209319257372e-06, + "loss": 1.479, + "step": 7787 + }, + { + "epoch": 1.42, + "learning_rate": 2.5774663269020748e-06, + "loss": 2.7544, + "step": 7788 + }, + { + "epoch": 1.42, + "learning_rate": 2.5774117218784127e-06, + "loss": 0.3358, + "step": 7789 + }, + { + "epoch": 1.42, + "learning_rate": 2.5773571168547506e-06, + "loss": 0.5505, + "step": 7790 + }, + { + "epoch": 1.42, + "learning_rate": 2.5773025118310886e-06, + "loss": 1.5857, + "step": 7791 + }, + { + "epoch": 1.42, + "learning_rate": 2.5772479068074265e-06, + "loss": 0.2048, + "step": 7792 + }, + { + "epoch": 1.42, + "learning_rate": 2.5771933017837645e-06, + "loss": 0.1772, + "step": 7793 + }, + { + "epoch": 1.42, + "learning_rate": 2.577138696760102e-06, + "loss": 2.7464, + "step": 7794 + }, + { + "epoch": 1.42, + "learning_rate": 2.57708409173644e-06, + "loss": 0.705, + "step": 7795 + }, + { + "epoch": 1.42, + "learning_rate": 2.5770294867127775e-06, + "loss": 1.2814, + "step": 7796 + }, + { + "epoch": 1.42, + "learning_rate": 2.5769748816891154e-06, + "loss": 0.1984, + "step": 7797 + }, + { + "epoch": 1.42, + "learning_rate": 2.5769202766654533e-06, + "loss": 0.3926, + "step": 7798 + }, + { + "epoch": 1.42, + "learning_rate": 2.576865671641791e-06, + "loss": 0.6835, + "step": 7799 + }, + { + "epoch": 1.42, + "learning_rate": 2.576811066618129e-06, + "loss": 0.1526, + "step": 7800 + }, + { + "epoch": 1.42, + "learning_rate": 2.5767564615944668e-06, + "loss": 0.3824, + "step": 7801 + }, + { + "epoch": 1.42, + "learning_rate": 2.5767018565708043e-06, + "loss": 0.9694, + "step": 7802 + }, + { + "epoch": 1.42, + "learning_rate": 2.5766472515471422e-06, + "loss": 0.4083, + "step": 7803 + }, + { + "epoch": 1.42, + "learning_rate": 2.57659264652348e-06, + "loss": 2.185, + "step": 7804 + }, + { + "epoch": 1.42, + "learning_rate": 2.576538041499818e-06, + "loss": 1.9245, + "step": 7805 + }, + { + "epoch": 1.42, + "learning_rate": 2.576483436476156e-06, + "loss": 0.871, + "step": 7806 + }, + { + "epoch": 1.42, + "learning_rate": 2.576428831452494e-06, + "loss": 0.5542, + "step": 7807 + }, + { + "epoch": 1.42, + "learning_rate": 2.5763742264288315e-06, + "loss": 0.5829, + "step": 7808 + }, + { + "epoch": 1.42, + "learning_rate": 2.5763196214051695e-06, + "loss": 1.922, + "step": 7809 + }, + { + "epoch": 1.42, + "learning_rate": 2.576265016381507e-06, + "loss": 0.8468, + "step": 7810 + }, + { + "epoch": 1.42, + "learning_rate": 2.576210411357845e-06, + "loss": 0.3197, + "step": 7811 + }, + { + "epoch": 1.42, + "learning_rate": 2.576155806334183e-06, + "loss": 1.7412, + "step": 7812 + }, + { + "epoch": 1.42, + "learning_rate": 2.5761012013105204e-06, + "loss": 1.2352, + "step": 7813 + }, + { + "epoch": 1.42, + "learning_rate": 2.5760465962868583e-06, + "loss": 2.266, + "step": 7814 + }, + { + "epoch": 1.42, + "learning_rate": 2.5759919912631963e-06, + "loss": 1.8647, + "step": 7815 + }, + { + "epoch": 1.42, + "learning_rate": 2.575937386239534e-06, + "loss": 1.4568, + "step": 7816 + }, + { + "epoch": 1.42, + "learning_rate": 2.575882781215872e-06, + "loss": 2.7937, + "step": 7817 + }, + { + "epoch": 1.42, + "learning_rate": 2.57582817619221e-06, + "loss": 0.998, + "step": 7818 + }, + { + "epoch": 1.42, + "learning_rate": 2.5757735711685476e-06, + "loss": 0.3261, + "step": 7819 + }, + { + "epoch": 1.42, + "learning_rate": 2.5757189661448856e-06, + "loss": 0.1087, + "step": 7820 + }, + { + "epoch": 1.42, + "learning_rate": 2.575664361121223e-06, + "loss": 0.071, + "step": 7821 + }, + { + "epoch": 1.42, + "learning_rate": 2.575609756097561e-06, + "loss": 1.1929, + "step": 7822 + }, + { + "epoch": 1.42, + "learning_rate": 2.575555151073899e-06, + "loss": 0.3357, + "step": 7823 + }, + { + "epoch": 1.42, + "learning_rate": 2.5755005460502365e-06, + "loss": 0.3492, + "step": 7824 + }, + { + "epoch": 1.42, + "learning_rate": 2.5754459410265744e-06, + "loss": 0.7995, + "step": 7825 + }, + { + "epoch": 1.42, + "learning_rate": 2.5753913360029124e-06, + "loss": 0.8241, + "step": 7826 + }, + { + "epoch": 1.42, + "learning_rate": 2.57533673097925e-06, + "loss": 0.5552, + "step": 7827 + }, + { + "epoch": 1.42, + "learning_rate": 2.575282125955588e-06, + "loss": 0.5585, + "step": 7828 + }, + { + "epoch": 1.42, + "learning_rate": 2.5752275209319258e-06, + "loss": 0.0429, + "step": 7829 + }, + { + "epoch": 1.42, + "learning_rate": 2.5751729159082637e-06, + "loss": 1.2808, + "step": 7830 + }, + { + "epoch": 1.42, + "learning_rate": 2.5751183108846017e-06, + "loss": 3.1202, + "step": 7831 + }, + { + "epoch": 1.42, + "learning_rate": 2.5750637058609396e-06, + "loss": 1.1485, + "step": 7832 + }, + { + "epoch": 1.42, + "learning_rate": 2.575009100837277e-06, + "loss": 0.0589, + "step": 7833 + }, + { + "epoch": 1.42, + "learning_rate": 2.574954495813615e-06, + "loss": 2.33, + "step": 7834 + }, + { + "epoch": 1.42, + "learning_rate": 2.5748998907899526e-06, + "loss": 0.0722, + "step": 7835 + }, + { + "epoch": 1.42, + "learning_rate": 2.5748452857662905e-06, + "loss": 0.4715, + "step": 7836 + }, + { + "epoch": 1.43, + "learning_rate": 2.5747906807426285e-06, + "loss": 1.7081, + "step": 7837 + }, + { + "epoch": 1.43, + "learning_rate": 2.574736075718966e-06, + "loss": 1.6244, + "step": 7838 + }, + { + "epoch": 1.43, + "learning_rate": 2.574681470695304e-06, + "loss": 0.2489, + "step": 7839 + }, + { + "epoch": 1.43, + "learning_rate": 2.574626865671642e-06, + "loss": 0.1039, + "step": 7840 + }, + { + "epoch": 1.43, + "learning_rate": 2.5745722606479794e-06, + "loss": 0.3515, + "step": 7841 + }, + { + "epoch": 1.43, + "learning_rate": 2.5745176556243174e-06, + "loss": 0.3035, + "step": 7842 + }, + { + "epoch": 1.43, + "learning_rate": 2.5744630506006553e-06, + "loss": 1.549, + "step": 7843 + }, + { + "epoch": 1.43, + "learning_rate": 2.5744084455769932e-06, + "loss": 1.6182, + "step": 7844 + }, + { + "epoch": 1.43, + "learning_rate": 2.574353840553331e-06, + "loss": 0.6149, + "step": 7845 + }, + { + "epoch": 1.43, + "learning_rate": 2.574299235529669e-06, + "loss": 0.9263, + "step": 7846 + }, + { + "epoch": 1.43, + "learning_rate": 2.5742446305060066e-06, + "loss": 0.2024, + "step": 7847 + }, + { + "epoch": 1.43, + "learning_rate": 2.5741900254823446e-06, + "loss": 1.9701, + "step": 7848 + }, + { + "epoch": 1.43, + "learning_rate": 2.574135420458682e-06, + "loss": 0.3712, + "step": 7849 + }, + { + "epoch": 1.43, + "learning_rate": 2.57408081543502e-06, + "loss": 1.4786, + "step": 7850 + }, + { + "epoch": 1.43, + "learning_rate": 2.574026210411358e-06, + "loss": 0.0414, + "step": 7851 + }, + { + "epoch": 1.43, + "learning_rate": 2.5739716053876955e-06, + "loss": 0.8321, + "step": 7852 + }, + { + "epoch": 1.43, + "learning_rate": 2.5739170003640335e-06, + "loss": 2.1163, + "step": 7853 + }, + { + "epoch": 1.43, + "learning_rate": 2.5738623953403714e-06, + "loss": 0.5628, + "step": 7854 + }, + { + "epoch": 1.43, + "learning_rate": 2.573807790316709e-06, + "loss": 1.7237, + "step": 7855 + }, + { + "epoch": 1.43, + "learning_rate": 2.573753185293047e-06, + "loss": 1.3586, + "step": 7856 + }, + { + "epoch": 1.43, + "learning_rate": 2.573698580269385e-06, + "loss": 0.4697, + "step": 7857 + }, + { + "epoch": 1.43, + "learning_rate": 2.5736439752457228e-06, + "loss": 0.8838, + "step": 7858 + }, + { + "epoch": 1.43, + "learning_rate": 2.5735893702220607e-06, + "loss": 0.1064, + "step": 7859 + }, + { + "epoch": 1.43, + "learning_rate": 2.5735347651983986e-06, + "loss": 1.1955, + "step": 7860 + }, + { + "epoch": 1.43, + "learning_rate": 2.573480160174736e-06, + "loss": 0.438, + "step": 7861 + }, + { + "epoch": 1.43, + "learning_rate": 2.573425555151074e-06, + "loss": 0.1638, + "step": 7862 + }, + { + "epoch": 1.43, + "learning_rate": 2.5733709501274116e-06, + "loss": 0.0722, + "step": 7863 + }, + { + "epoch": 1.43, + "learning_rate": 2.5733163451037496e-06, + "loss": 0.0656, + "step": 7864 + }, + { + "epoch": 1.43, + "learning_rate": 2.5732617400800875e-06, + "loss": 1.5531, + "step": 7865 + }, + { + "epoch": 1.43, + "learning_rate": 2.573207135056425e-06, + "loss": 1.2971, + "step": 7866 + }, + { + "epoch": 1.43, + "learning_rate": 2.573152530032763e-06, + "loss": 0.7685, + "step": 7867 + }, + { + "epoch": 1.43, + "learning_rate": 2.573097925009101e-06, + "loss": 0.3413, + "step": 7868 + }, + { + "epoch": 1.43, + "learning_rate": 2.5730433199854384e-06, + "loss": 0.2327, + "step": 7869 + }, + { + "epoch": 1.43, + "learning_rate": 2.5729887149617764e-06, + "loss": 1.7468, + "step": 7870 + }, + { + "epoch": 1.43, + "learning_rate": 2.5729341099381143e-06, + "loss": 2.6722, + "step": 7871 + }, + { + "epoch": 1.43, + "learning_rate": 2.5728795049144523e-06, + "loss": 0.2114, + "step": 7872 + }, + { + "epoch": 1.43, + "learning_rate": 2.57282489989079e-06, + "loss": 0.0617, + "step": 7873 + }, + { + "epoch": 1.43, + "learning_rate": 2.572770294867128e-06, + "loss": 0.6019, + "step": 7874 + }, + { + "epoch": 1.43, + "learning_rate": 2.5727156898434657e-06, + "loss": 2.4897, + "step": 7875 + }, + { + "epoch": 1.43, + "learning_rate": 2.5726610848198036e-06, + "loss": 0.605, + "step": 7876 + }, + { + "epoch": 1.43, + "learning_rate": 2.572606479796141e-06, + "loss": 0.449, + "step": 7877 + }, + { + "epoch": 1.43, + "learning_rate": 2.572551874772479e-06, + "loss": 0.7636, + "step": 7878 + }, + { + "epoch": 1.43, + "learning_rate": 2.572497269748817e-06, + "loss": 1.4528, + "step": 7879 + }, + { + "epoch": 1.43, + "learning_rate": 2.5724426647251545e-06, + "loss": 0.6816, + "step": 7880 + }, + { + "epoch": 1.43, + "learning_rate": 2.5723880597014925e-06, + "loss": 0.5188, + "step": 7881 + }, + { + "epoch": 1.43, + "learning_rate": 2.5723334546778304e-06, + "loss": 0.2629, + "step": 7882 + }, + { + "epoch": 1.43, + "learning_rate": 2.572278849654168e-06, + "loss": 0.454, + "step": 7883 + }, + { + "epoch": 1.43, + "learning_rate": 2.572224244630506e-06, + "loss": 1.1273, + "step": 7884 + }, + { + "epoch": 1.43, + "learning_rate": 2.572169639606844e-06, + "loss": 0.0909, + "step": 7885 + }, + { + "epoch": 1.43, + "learning_rate": 2.5721150345831818e-06, + "loss": 0.0276, + "step": 7886 + }, + { + "epoch": 1.43, + "learning_rate": 2.5720604295595197e-06, + "loss": 1.0845, + "step": 7887 + }, + { + "epoch": 1.43, + "learning_rate": 2.5720058245358572e-06, + "loss": 0.381, + "step": 7888 + }, + { + "epoch": 1.43, + "learning_rate": 2.571951219512195e-06, + "loss": 2.9235, + "step": 7889 + }, + { + "epoch": 1.43, + "learning_rate": 2.571896614488533e-06, + "loss": 1.0684, + "step": 7890 + }, + { + "epoch": 1.43, + "learning_rate": 2.5718420094648706e-06, + "loss": 0.4157, + "step": 7891 + }, + { + "epoch": 1.44, + "learning_rate": 2.5717874044412086e-06, + "loss": 2.7144, + "step": 7892 + }, + { + "epoch": 1.44, + "learning_rate": 2.5717327994175465e-06, + "loss": 0.8751, + "step": 7893 + }, + { + "epoch": 1.44, + "learning_rate": 2.571678194393884e-06, + "loss": 0.4443, + "step": 7894 + }, + { + "epoch": 1.44, + "learning_rate": 2.571623589370222e-06, + "loss": 0.508, + "step": 7895 + }, + { + "epoch": 1.44, + "learning_rate": 2.57156898434656e-06, + "loss": 0.4271, + "step": 7896 + }, + { + "epoch": 1.44, + "learning_rate": 2.571514379322898e-06, + "loss": 0.898, + "step": 7897 + }, + { + "epoch": 1.44, + "learning_rate": 2.571459774299236e-06, + "loss": 1.0572, + "step": 7898 + }, + { + "epoch": 1.44, + "learning_rate": 2.5714051692755738e-06, + "loss": 1.8441, + "step": 7899 + }, + { + "epoch": 1.44, + "learning_rate": 2.5713505642519113e-06, + "loss": 0.1973, + "step": 7900 + }, + { + "epoch": 1.44, + "learning_rate": 2.5712959592282492e-06, + "loss": 3.1995, + "step": 7901 + }, + { + "epoch": 1.44, + "learning_rate": 2.5712413542045868e-06, + "loss": 1.6176, + "step": 7902 + }, + { + "epoch": 1.44, + "learning_rate": 2.5711867491809247e-06, + "loss": 0.3384, + "step": 7903 + }, + { + "epoch": 1.44, + "learning_rate": 2.5711321441572626e-06, + "loss": 0.038, + "step": 7904 + }, + { + "epoch": 1.44, + "learning_rate": 2.5710775391336e-06, + "loss": 0.7329, + "step": 7905 + }, + { + "epoch": 1.44, + "learning_rate": 2.571022934109938e-06, + "loss": 1.8027, + "step": 7906 + }, + { + "epoch": 1.44, + "learning_rate": 2.570968329086276e-06, + "loss": 1.4851, + "step": 7907 + }, + { + "epoch": 1.44, + "learning_rate": 2.5709137240626136e-06, + "loss": 0.35, + "step": 7908 + }, + { + "epoch": 1.44, + "learning_rate": 2.5708591190389515e-06, + "loss": 0.1442, + "step": 7909 + }, + { + "epoch": 1.44, + "learning_rate": 2.5708045140152895e-06, + "loss": 1.2975, + "step": 7910 + }, + { + "epoch": 1.44, + "learning_rate": 2.5707499089916274e-06, + "loss": 2.195, + "step": 7911 + }, + { + "epoch": 1.44, + "learning_rate": 2.5706953039679653e-06, + "loss": 0.6958, + "step": 7912 + }, + { + "epoch": 1.44, + "learning_rate": 2.5706406989443033e-06, + "loss": 0.3199, + "step": 7913 + }, + { + "epoch": 1.44, + "learning_rate": 2.570586093920641e-06, + "loss": 2.3437, + "step": 7914 + }, + { + "epoch": 1.44, + "learning_rate": 2.5705314888969787e-06, + "loss": 1.7785, + "step": 7915 + }, + { + "epoch": 1.44, + "learning_rate": 2.5704768838733163e-06, + "loss": 0.8162, + "step": 7916 + }, + { + "epoch": 1.44, + "learning_rate": 2.5704222788496542e-06, + "loss": 0.8481, + "step": 7917 + }, + { + "epoch": 1.44, + "learning_rate": 2.570367673825992e-06, + "loss": 0.0454, + "step": 7918 + }, + { + "epoch": 1.44, + "learning_rate": 2.5703130688023297e-06, + "loss": 0.5203, + "step": 7919 + }, + { + "epoch": 1.44, + "learning_rate": 2.5702584637786676e-06, + "loss": 1.4108, + "step": 7920 + }, + { + "epoch": 1.44, + "learning_rate": 2.5702038587550056e-06, + "loss": 0.478, + "step": 7921 + }, + { + "epoch": 1.44, + "learning_rate": 2.570149253731343e-06, + "loss": 1.0449, + "step": 7922 + }, + { + "epoch": 1.44, + "learning_rate": 2.570094648707681e-06, + "loss": 0.7326, + "step": 7923 + }, + { + "epoch": 1.44, + "learning_rate": 2.570040043684019e-06, + "loss": 0.0305, + "step": 7924 + }, + { + "epoch": 1.44, + "learning_rate": 2.569985438660357e-06, + "loss": 2.0592, + "step": 7925 + }, + { + "epoch": 1.44, + "learning_rate": 2.569930833636695e-06, + "loss": 0.819, + "step": 7926 + }, + { + "epoch": 1.44, + "learning_rate": 2.569876228613033e-06, + "loss": 0.8559, + "step": 7927 + }, + { + "epoch": 1.44, + "learning_rate": 2.5698216235893703e-06, + "loss": 0.9119, + "step": 7928 + }, + { + "epoch": 1.44, + "learning_rate": 2.5697670185657083e-06, + "loss": 1.3988, + "step": 7929 + }, + { + "epoch": 1.44, + "learning_rate": 2.5697124135420458e-06, + "loss": 0.6888, + "step": 7930 + }, + { + "epoch": 1.44, + "learning_rate": 2.5696578085183837e-06, + "loss": 0.2149, + "step": 7931 + }, + { + "epoch": 1.44, + "learning_rate": 2.5696032034947217e-06, + "loss": 1.1889, + "step": 7932 + }, + { + "epoch": 1.44, + "learning_rate": 2.569548598471059e-06, + "loss": 0.177, + "step": 7933 + }, + { + "epoch": 1.44, + "learning_rate": 2.569493993447397e-06, + "loss": 1.4893, + "step": 7934 + }, + { + "epoch": 1.44, + "learning_rate": 2.569439388423735e-06, + "loss": 0.3581, + "step": 7935 + }, + { + "epoch": 1.44, + "learning_rate": 2.5693847834000726e-06, + "loss": 0.7001, + "step": 7936 + }, + { + "epoch": 1.44, + "learning_rate": 2.5693301783764105e-06, + "loss": 1.1403, + "step": 7937 + }, + { + "epoch": 1.44, + "learning_rate": 2.5692755733527485e-06, + "loss": 0.3768, + "step": 7938 + }, + { + "epoch": 1.44, + "learning_rate": 2.5692209683290864e-06, + "loss": 1.875, + "step": 7939 + }, + { + "epoch": 1.44, + "learning_rate": 2.5691663633054244e-06, + "loss": 0.9655, + "step": 7940 + }, + { + "epoch": 1.44, + "learning_rate": 2.5691117582817623e-06, + "loss": 1.5089, + "step": 7941 + }, + { + "epoch": 1.44, + "learning_rate": 2.5690571532581e-06, + "loss": 0.0729, + "step": 7942 + }, + { + "epoch": 1.44, + "learning_rate": 2.5690025482344378e-06, + "loss": 2.4911, + "step": 7943 + }, + { + "epoch": 1.44, + "learning_rate": 2.5689479432107753e-06, + "loss": 0.3015, + "step": 7944 + }, + { + "epoch": 1.44, + "learning_rate": 2.5688933381871132e-06, + "loss": 1.5075, + "step": 7945 + }, + { + "epoch": 1.44, + "learning_rate": 2.568838733163451e-06, + "loss": 2.1419, + "step": 7946 + }, + { + "epoch": 1.45, + "learning_rate": 2.5687841281397887e-06, + "loss": 0.7189, + "step": 7947 + }, + { + "epoch": 1.45, + "learning_rate": 2.5687295231161266e-06, + "loss": 0.1081, + "step": 7948 + }, + { + "epoch": 1.45, + "learning_rate": 2.5686749180924646e-06, + "loss": 1.2301, + "step": 7949 + }, + { + "epoch": 1.45, + "learning_rate": 2.568620313068802e-06, + "loss": 1.2309, + "step": 7950 + }, + { + "epoch": 1.45, + "learning_rate": 2.56856570804514e-06, + "loss": 0.7148, + "step": 7951 + }, + { + "epoch": 1.45, + "learning_rate": 2.568511103021478e-06, + "loss": 0.2446, + "step": 7952 + }, + { + "epoch": 1.45, + "learning_rate": 2.568456497997816e-06, + "loss": 0.5766, + "step": 7953 + }, + { + "epoch": 1.45, + "learning_rate": 2.568401892974154e-06, + "loss": 2.6031, + "step": 7954 + }, + { + "epoch": 1.45, + "learning_rate": 2.568347287950492e-06, + "loss": 1.515, + "step": 7955 + }, + { + "epoch": 1.45, + "learning_rate": 2.5682926829268293e-06, + "loss": 0.4278, + "step": 7956 + }, + { + "epoch": 1.45, + "learning_rate": 2.5682380779031673e-06, + "loss": 0.0919, + "step": 7957 + }, + { + "epoch": 1.45, + "learning_rate": 2.568183472879505e-06, + "loss": 3.0022, + "step": 7958 + }, + { + "epoch": 1.45, + "learning_rate": 2.5681288678558428e-06, + "loss": 0.0572, + "step": 7959 + }, + { + "epoch": 1.45, + "learning_rate": 2.5680742628321807e-06, + "loss": 0.0372, + "step": 7960 + }, + { + "epoch": 1.45, + "learning_rate": 2.5680196578085182e-06, + "loss": 0.26, + "step": 7961 + }, + { + "epoch": 1.45, + "learning_rate": 2.567965052784856e-06, + "loss": 0.037, + "step": 7962 + }, + { + "epoch": 1.45, + "learning_rate": 2.567910447761194e-06, + "loss": 0.8094, + "step": 7963 + }, + { + "epoch": 1.45, + "learning_rate": 2.5678558427375316e-06, + "loss": 1.6231, + "step": 7964 + }, + { + "epoch": 1.45, + "learning_rate": 2.5678012377138696e-06, + "loss": 0.4469, + "step": 7965 + }, + { + "epoch": 1.45, + "learning_rate": 2.5677466326902075e-06, + "loss": 1.8851, + "step": 7966 + }, + { + "epoch": 1.45, + "learning_rate": 2.5676920276665455e-06, + "loss": 0.075, + "step": 7967 + }, + { + "epoch": 1.45, + "learning_rate": 2.5676374226428834e-06, + "loss": 1.0202, + "step": 7968 + }, + { + "epoch": 1.45, + "learning_rate": 2.567582817619221e-06, + "loss": 0.1452, + "step": 7969 + }, + { + "epoch": 1.45, + "learning_rate": 2.567528212595559e-06, + "loss": 0.319, + "step": 7970 + }, + { + "epoch": 1.45, + "learning_rate": 2.567473607571897e-06, + "loss": 0.1043, + "step": 7971 + }, + { + "epoch": 1.45, + "learning_rate": 2.5674190025482343e-06, + "loss": 1.3259, + "step": 7972 + }, + { + "epoch": 1.45, + "learning_rate": 2.5673643975245723e-06, + "loss": 0.4247, + "step": 7973 + }, + { + "epoch": 1.45, + "learning_rate": 2.56730979250091e-06, + "loss": 1.2773, + "step": 7974 + }, + { + "epoch": 1.45, + "learning_rate": 2.5672551874772477e-06, + "loss": 1.8033, + "step": 7975 + }, + { + "epoch": 1.45, + "learning_rate": 2.5672005824535857e-06, + "loss": 2.8608, + "step": 7976 + }, + { + "epoch": 1.45, + "learning_rate": 2.5671459774299236e-06, + "loss": 1.1745, + "step": 7977 + }, + { + "epoch": 1.45, + "learning_rate": 2.5670913724062616e-06, + "loss": 1.0257, + "step": 7978 + }, + { + "epoch": 1.45, + "learning_rate": 2.5670367673825995e-06, + "loss": 1.4437, + "step": 7979 + }, + { + "epoch": 1.45, + "learning_rate": 2.5669821623589374e-06, + "loss": 0.0466, + "step": 7980 + }, + { + "epoch": 1.45, + "learning_rate": 2.566927557335275e-06, + "loss": 0.8196, + "step": 7981 + }, + { + "epoch": 1.45, + "learning_rate": 2.566872952311613e-06, + "loss": 0.6914, + "step": 7982 + }, + { + "epoch": 1.45, + "learning_rate": 2.5668183472879504e-06, + "loss": 0.8001, + "step": 7983 + }, + { + "epoch": 1.45, + "learning_rate": 2.5667637422642884e-06, + "loss": 0.0812, + "step": 7984 + }, + { + "epoch": 1.45, + "learning_rate": 2.5667091372406263e-06, + "loss": 1.3771, + "step": 7985 + }, + { + "epoch": 1.45, + "learning_rate": 2.566654532216964e-06, + "loss": 3.2743, + "step": 7986 + }, + { + "epoch": 1.45, + "learning_rate": 2.5665999271933018e-06, + "loss": 2.3084, + "step": 7987 + }, + { + "epoch": 1.45, + "learning_rate": 2.5665453221696397e-06, + "loss": 1.4872, + "step": 7988 + }, + { + "epoch": 1.45, + "learning_rate": 2.5664907171459772e-06, + "loss": 2.1245, + "step": 7989 + }, + { + "epoch": 1.45, + "learning_rate": 2.566436112122315e-06, + "loss": 0.6916, + "step": 7990 + }, + { + "epoch": 1.45, + "learning_rate": 2.566381507098653e-06, + "loss": 0.4449, + "step": 7991 + }, + { + "epoch": 1.45, + "learning_rate": 2.566326902074991e-06, + "loss": 2.0454, + "step": 7992 + }, + { + "epoch": 1.45, + "learning_rate": 2.566272297051329e-06, + "loss": 0.8257, + "step": 7993 + }, + { + "epoch": 1.45, + "learning_rate": 2.566217692027667e-06, + "loss": 0.3635, + "step": 7994 + }, + { + "epoch": 1.45, + "learning_rate": 2.5661630870040045e-06, + "loss": 0.5978, + "step": 7995 + }, + { + "epoch": 1.45, + "learning_rate": 2.5661084819803424e-06, + "loss": 1.1962, + "step": 7996 + }, + { + "epoch": 1.45, + "learning_rate": 2.56605387695668e-06, + "loss": 0.6265, + "step": 7997 + }, + { + "epoch": 1.45, + "learning_rate": 2.565999271933018e-06, + "loss": 1.9284, + "step": 7998 + }, + { + "epoch": 1.45, + "learning_rate": 2.565944666909356e-06, + "loss": 0.8259, + "step": 7999 + }, + { + "epoch": 1.45, + "learning_rate": 2.5658900618856933e-06, + "loss": 1.1353, + "step": 8000 + }, + { + "epoch": 1.45, + "learning_rate": 2.5658354568620313e-06, + "loss": 0.5202, + "step": 8001 + }, + { + "epoch": 1.46, + "learning_rate": 2.5657808518383692e-06, + "loss": 0.5126, + "step": 8002 + }, + { + "epoch": 1.46, + "learning_rate": 2.5657262468147068e-06, + "loss": 0.6821, + "step": 8003 + }, + { + "epoch": 1.46, + "learning_rate": 2.5656716417910447e-06, + "loss": 0.4502, + "step": 8004 + }, + { + "epoch": 1.46, + "learning_rate": 2.5656170367673826e-06, + "loss": 1.3396, + "step": 8005 + }, + { + "epoch": 1.46, + "learning_rate": 2.5655624317437206e-06, + "loss": 1.1315, + "step": 8006 + }, + { + "epoch": 1.46, + "learning_rate": 2.5655078267200585e-06, + "loss": 0.8089, + "step": 8007 + }, + { + "epoch": 1.46, + "learning_rate": 2.5654532216963965e-06, + "loss": 0.9416, + "step": 8008 + }, + { + "epoch": 1.46, + "learning_rate": 2.565398616672734e-06, + "loss": 1.426, + "step": 8009 + }, + { + "epoch": 1.46, + "learning_rate": 2.565344011649072e-06, + "loss": 0.0333, + "step": 8010 + }, + { + "epoch": 1.46, + "learning_rate": 2.5652894066254095e-06, + "loss": 1.8573, + "step": 8011 + }, + { + "epoch": 1.46, + "learning_rate": 2.5652348016017474e-06, + "loss": 0.5887, + "step": 8012 + }, + { + "epoch": 1.46, + "learning_rate": 2.5651801965780853e-06, + "loss": 1.5228, + "step": 8013 + }, + { + "epoch": 1.46, + "learning_rate": 2.565125591554423e-06, + "loss": 0.4524, + "step": 8014 + }, + { + "epoch": 1.46, + "learning_rate": 2.565070986530761e-06, + "loss": 1.7744, + "step": 8015 + }, + { + "epoch": 1.46, + "learning_rate": 2.5650163815070987e-06, + "loss": 0.6582, + "step": 8016 + }, + { + "epoch": 1.46, + "learning_rate": 2.5649617764834363e-06, + "loss": 2.2809, + "step": 8017 + }, + { + "epoch": 1.46, + "learning_rate": 2.5649071714597742e-06, + "loss": 0.6188, + "step": 8018 + }, + { + "epoch": 1.46, + "learning_rate": 2.564852566436112e-06, + "loss": 1.9495, + "step": 8019 + }, + { + "epoch": 1.46, + "learning_rate": 2.56479796141245e-06, + "loss": 1.5418, + "step": 8020 + }, + { + "epoch": 1.46, + "learning_rate": 2.564743356388788e-06, + "loss": 1.4166, + "step": 8021 + }, + { + "epoch": 1.46, + "learning_rate": 2.564688751365126e-06, + "loss": 1.2625, + "step": 8022 + }, + { + "epoch": 1.46, + "learning_rate": 2.5646341463414635e-06, + "loss": 2.1985, + "step": 8023 + }, + { + "epoch": 1.46, + "learning_rate": 2.5645795413178014e-06, + "loss": 0.9448, + "step": 8024 + }, + { + "epoch": 1.46, + "learning_rate": 2.564524936294139e-06, + "loss": 2.3922, + "step": 8025 + }, + { + "epoch": 1.46, + "learning_rate": 2.564470331270477e-06, + "loss": 1.8065, + "step": 8026 + }, + { + "epoch": 1.46, + "learning_rate": 2.564415726246815e-06, + "loss": 0.9453, + "step": 8027 + }, + { + "epoch": 1.46, + "learning_rate": 2.5643611212231524e-06, + "loss": 0.3411, + "step": 8028 + }, + { + "epoch": 1.46, + "learning_rate": 2.5643065161994903e-06, + "loss": 0.1312, + "step": 8029 + }, + { + "epoch": 1.46, + "learning_rate": 2.5642519111758283e-06, + "loss": 0.5997, + "step": 8030 + }, + { + "epoch": 1.46, + "learning_rate": 2.5641973061521658e-06, + "loss": 0.5336, + "step": 8031 + }, + { + "epoch": 1.46, + "learning_rate": 2.5641427011285037e-06, + "loss": 1.2843, + "step": 8032 + }, + { + "epoch": 1.46, + "learning_rate": 2.5640880961048417e-06, + "loss": 0.5127, + "step": 8033 + }, + { + "epoch": 1.46, + "learning_rate": 2.5640334910811796e-06, + "loss": 1.7011, + "step": 8034 + }, + { + "epoch": 1.46, + "learning_rate": 2.5639788860575176e-06, + "loss": 0.2361, + "step": 8035 + }, + { + "epoch": 1.46, + "learning_rate": 2.5639242810338555e-06, + "loss": 0.3572, + "step": 8036 + }, + { + "epoch": 1.46, + "learning_rate": 2.563869676010193e-06, + "loss": 1.3754, + "step": 8037 + }, + { + "epoch": 1.46, + "learning_rate": 2.563815070986531e-06, + "loss": 1.387, + "step": 8038 + }, + { + "epoch": 1.46, + "learning_rate": 2.5637604659628685e-06, + "loss": 1.5374, + "step": 8039 + }, + { + "epoch": 1.46, + "learning_rate": 2.5637058609392064e-06, + "loss": 0.7809, + "step": 8040 + }, + { + "epoch": 1.46, + "learning_rate": 2.5636512559155444e-06, + "loss": 0.6944, + "step": 8041 + }, + { + "epoch": 1.46, + "learning_rate": 2.563596650891882e-06, + "loss": 1.0063, + "step": 8042 + }, + { + "epoch": 1.46, + "learning_rate": 2.56354204586822e-06, + "loss": 0.0374, + "step": 8043 + }, + { + "epoch": 1.46, + "learning_rate": 2.5634874408445578e-06, + "loss": 0.0265, + "step": 8044 + }, + { + "epoch": 1.46, + "learning_rate": 2.5634328358208953e-06, + "loss": 0.4505, + "step": 8045 + }, + { + "epoch": 1.46, + "learning_rate": 2.5633782307972332e-06, + "loss": 0.0983, + "step": 8046 + }, + { + "epoch": 1.46, + "learning_rate": 2.563323625773571e-06, + "loss": 0.9125, + "step": 8047 + }, + { + "epoch": 1.46, + "learning_rate": 2.563269020749909e-06, + "loss": 0.2338, + "step": 8048 + }, + { + "epoch": 1.46, + "learning_rate": 2.563214415726247e-06, + "loss": 1.1687, + "step": 8049 + }, + { + "epoch": 1.46, + "learning_rate": 2.5631598107025846e-06, + "loss": 0.9079, + "step": 8050 + }, + { + "epoch": 1.46, + "learning_rate": 2.5631052056789225e-06, + "loss": 0.8574, + "step": 8051 + }, + { + "epoch": 1.46, + "learning_rate": 2.5630506006552605e-06, + "loss": 0.3426, + "step": 8052 + }, + { + "epoch": 1.46, + "learning_rate": 2.562995995631598e-06, + "loss": 1.553, + "step": 8053 + }, + { + "epoch": 1.46, + "learning_rate": 2.562941390607936e-06, + "loss": 0.5455, + "step": 8054 + }, + { + "epoch": 1.46, + "learning_rate": 2.562886785584274e-06, + "loss": 1.6008, + "step": 8055 + }, + { + "epoch": 1.46, + "learning_rate": 2.5628321805606114e-06, + "loss": 0.6982, + "step": 8056 + }, + { + "epoch": 1.47, + "learning_rate": 2.5627775755369493e-06, + "loss": 0.8211, + "step": 8057 + }, + { + "epoch": 1.47, + "learning_rate": 2.5627229705132873e-06, + "loss": 1.9602, + "step": 8058 + }, + { + "epoch": 1.47, + "learning_rate": 2.562668365489625e-06, + "loss": 0.981, + "step": 8059 + }, + { + "epoch": 1.47, + "learning_rate": 2.5626137604659628e-06, + "loss": 0.2086, + "step": 8060 + }, + { + "epoch": 1.47, + "learning_rate": 2.5625591554423007e-06, + "loss": 2.0157, + "step": 8061 + }, + { + "epoch": 1.47, + "learning_rate": 2.5625045504186386e-06, + "loss": 2.3498, + "step": 8062 + }, + { + "epoch": 1.47, + "learning_rate": 2.5624499453949766e-06, + "loss": 0.7312, + "step": 8063 + }, + { + "epoch": 1.47, + "learning_rate": 2.562395340371314e-06, + "loss": 1.5289, + "step": 8064 + }, + { + "epoch": 1.47, + "learning_rate": 2.562340735347652e-06, + "loss": 1.5834, + "step": 8065 + }, + { + "epoch": 1.47, + "learning_rate": 2.56228613032399e-06, + "loss": 2.4061, + "step": 8066 + }, + { + "epoch": 1.47, + "learning_rate": 2.5622315253003275e-06, + "loss": 0.6203, + "step": 8067 + }, + { + "epoch": 1.47, + "learning_rate": 2.5621769202766655e-06, + "loss": 1.0459, + "step": 8068 + }, + { + "epoch": 1.47, + "learning_rate": 2.5621223152530034e-06, + "loss": 0.4674, + "step": 8069 + }, + { + "epoch": 1.47, + "learning_rate": 2.562067710229341e-06, + "loss": 0.5111, + "step": 8070 + }, + { + "epoch": 1.47, + "learning_rate": 2.562013105205679e-06, + "loss": 0.4723, + "step": 8071 + }, + { + "epoch": 1.47, + "learning_rate": 2.561958500182017e-06, + "loss": 0.0582, + "step": 8072 + }, + { + "epoch": 1.47, + "learning_rate": 2.5619038951583547e-06, + "loss": 0.5287, + "step": 8073 + }, + { + "epoch": 1.47, + "learning_rate": 2.5618492901346927e-06, + "loss": 0.4218, + "step": 8074 + }, + { + "epoch": 1.47, + "learning_rate": 2.5617946851110306e-06, + "loss": 1.1628, + "step": 8075 + }, + { + "epoch": 1.47, + "learning_rate": 2.561740080087368e-06, + "loss": 1.5408, + "step": 8076 + }, + { + "epoch": 1.47, + "learning_rate": 2.561685475063706e-06, + "loss": 1.4604, + "step": 8077 + }, + { + "epoch": 1.47, + "learning_rate": 2.5616308700400436e-06, + "loss": 0.6639, + "step": 8078 + }, + { + "epoch": 1.47, + "learning_rate": 2.5615762650163816e-06, + "loss": 2.3381, + "step": 8079 + }, + { + "epoch": 1.47, + "learning_rate": 2.5615216599927195e-06, + "loss": 1.534, + "step": 8080 + }, + { + "epoch": 1.47, + "learning_rate": 2.561467054969057e-06, + "loss": 0.869, + "step": 8081 + }, + { + "epoch": 1.47, + "learning_rate": 2.561412449945395e-06, + "loss": 1.8166, + "step": 8082 + }, + { + "epoch": 1.47, + "learning_rate": 2.561357844921733e-06, + "loss": 1.8702, + "step": 8083 + }, + { + "epoch": 1.47, + "learning_rate": 2.5613032398980704e-06, + "loss": 1.8631, + "step": 8084 + }, + { + "epoch": 1.47, + "learning_rate": 2.5612486348744084e-06, + "loss": 2.296, + "step": 8085 + }, + { + "epoch": 1.47, + "learning_rate": 2.5611940298507463e-06, + "loss": 0.1407, + "step": 8086 + }, + { + "epoch": 1.47, + "learning_rate": 2.5611394248270843e-06, + "loss": 1.3613, + "step": 8087 + }, + { + "epoch": 1.47, + "learning_rate": 2.561084819803422e-06, + "loss": 0.825, + "step": 8088 + }, + { + "epoch": 1.47, + "learning_rate": 2.56103021477976e-06, + "loss": 2.0957, + "step": 8089 + }, + { + "epoch": 1.47, + "learning_rate": 2.5609756097560977e-06, + "loss": 1.4871, + "step": 8090 + }, + { + "epoch": 1.47, + "learning_rate": 2.5609210047324356e-06, + "loss": 0.4516, + "step": 8091 + }, + { + "epoch": 1.47, + "learning_rate": 2.560866399708773e-06, + "loss": 0.8378, + "step": 8092 + }, + { + "epoch": 1.47, + "learning_rate": 2.560811794685111e-06, + "loss": 1.9732, + "step": 8093 + }, + { + "epoch": 1.47, + "learning_rate": 2.560757189661449e-06, + "loss": 0.6487, + "step": 8094 + }, + { + "epoch": 1.47, + "learning_rate": 2.5607025846377865e-06, + "loss": 0.8765, + "step": 8095 + }, + { + "epoch": 1.47, + "learning_rate": 2.5606479796141245e-06, + "loss": 0.6921, + "step": 8096 + }, + { + "epoch": 1.47, + "learning_rate": 2.5605933745904624e-06, + "loss": 0.6085, + "step": 8097 + }, + { + "epoch": 1.47, + "learning_rate": 2.5605387695668e-06, + "loss": 0.026, + "step": 8098 + }, + { + "epoch": 1.47, + "learning_rate": 2.560484164543138e-06, + "loss": 1.3334, + "step": 8099 + }, + { + "epoch": 1.47, + "learning_rate": 2.560429559519476e-06, + "loss": 0.0741, + "step": 8100 + }, + { + "epoch": 1.47, + "learning_rate": 2.5603749544958138e-06, + "loss": 1.782, + "step": 8101 + }, + { + "epoch": 1.47, + "learning_rate": 2.5603203494721517e-06, + "loss": 3.4704, + "step": 8102 + }, + { + "epoch": 1.47, + "learning_rate": 2.5602657444484897e-06, + "loss": 1.6584, + "step": 8103 + }, + { + "epoch": 1.47, + "learning_rate": 2.560211139424827e-06, + "loss": 3.8466, + "step": 8104 + }, + { + "epoch": 1.47, + "learning_rate": 2.560156534401165e-06, + "loss": 0.3227, + "step": 8105 + }, + { + "epoch": 1.47, + "learning_rate": 2.5601019293775026e-06, + "loss": 0.0856, + "step": 8106 + }, + { + "epoch": 1.47, + "learning_rate": 2.5600473243538406e-06, + "loss": 3.2054, + "step": 8107 + }, + { + "epoch": 1.47, + "learning_rate": 2.5599927193301785e-06, + "loss": 0.909, + "step": 8108 + }, + { + "epoch": 1.47, + "learning_rate": 2.559938114306516e-06, + "loss": 0.0793, + "step": 8109 + }, + { + "epoch": 1.47, + "learning_rate": 2.559883509282854e-06, + "loss": 0.411, + "step": 8110 + }, + { + "epoch": 1.47, + "learning_rate": 2.559828904259192e-06, + "loss": 2.5798, + "step": 8111 + }, + { + "epoch": 1.48, + "learning_rate": 2.5597742992355295e-06, + "loss": 4.1848, + "step": 8112 + }, + { + "epoch": 1.48, + "learning_rate": 2.5597196942118674e-06, + "loss": 0.3682, + "step": 8113 + }, + { + "epoch": 1.48, + "learning_rate": 2.5596650891882053e-06, + "loss": 0.765, + "step": 8114 + }, + { + "epoch": 1.48, + "learning_rate": 2.5596104841645433e-06, + "loss": 1.0398, + "step": 8115 + }, + { + "epoch": 1.48, + "learning_rate": 2.5595558791408812e-06, + "loss": 0.148, + "step": 8116 + }, + { + "epoch": 1.48, + "learning_rate": 2.559501274117219e-06, + "loss": 0.8384, + "step": 8117 + }, + { + "epoch": 1.48, + "learning_rate": 2.5594466690935567e-06, + "loss": 1.4525, + "step": 8118 + }, + { + "epoch": 1.48, + "learning_rate": 2.5593920640698946e-06, + "loss": 0.7924, + "step": 8119 + }, + { + "epoch": 1.48, + "learning_rate": 2.559337459046232e-06, + "loss": 0.8337, + "step": 8120 + }, + { + "epoch": 1.48, + "learning_rate": 2.55928285402257e-06, + "loss": 0.0284, + "step": 8121 + }, + { + "epoch": 1.48, + "learning_rate": 2.559228248998908e-06, + "loss": 1.3923, + "step": 8122 + }, + { + "epoch": 1.48, + "learning_rate": 2.5591736439752456e-06, + "loss": 0.0972, + "step": 8123 + }, + { + "epoch": 1.48, + "learning_rate": 2.5591190389515835e-06, + "loss": 2.0931, + "step": 8124 + }, + { + "epoch": 1.48, + "learning_rate": 2.5590644339279215e-06, + "loss": 1.1269, + "step": 8125 + }, + { + "epoch": 1.48, + "learning_rate": 2.559009828904259e-06, + "loss": 2.1067, + "step": 8126 + }, + { + "epoch": 1.48, + "learning_rate": 2.558955223880597e-06, + "loss": 2.733, + "step": 8127 + }, + { + "epoch": 1.48, + "learning_rate": 2.558900618856935e-06, + "loss": 0.1804, + "step": 8128 + }, + { + "epoch": 1.48, + "learning_rate": 2.558846013833273e-06, + "loss": 0.3634, + "step": 8129 + }, + { + "epoch": 1.48, + "learning_rate": 2.5587914088096107e-06, + "loss": 1.8269, + "step": 8130 + }, + { + "epoch": 1.48, + "learning_rate": 2.5587368037859483e-06, + "loss": 0.31, + "step": 8131 + }, + { + "epoch": 1.48, + "learning_rate": 2.558682198762286e-06, + "loss": 0.6467, + "step": 8132 + }, + { + "epoch": 1.48, + "learning_rate": 2.558627593738624e-06, + "loss": 0.8539, + "step": 8133 + }, + { + "epoch": 1.48, + "learning_rate": 2.5585729887149617e-06, + "loss": 0.0553, + "step": 8134 + }, + { + "epoch": 1.48, + "learning_rate": 2.5585183836912996e-06, + "loss": 0.4952, + "step": 8135 + }, + { + "epoch": 1.48, + "learning_rate": 2.5584637786676376e-06, + "loss": 0.2501, + "step": 8136 + }, + { + "epoch": 1.48, + "learning_rate": 2.558409173643975e-06, + "loss": 0.0433, + "step": 8137 + }, + { + "epoch": 1.48, + "learning_rate": 2.558354568620313e-06, + "loss": 0.0603, + "step": 8138 + }, + { + "epoch": 1.48, + "learning_rate": 2.558299963596651e-06, + "loss": 0.4618, + "step": 8139 + }, + { + "epoch": 1.48, + "learning_rate": 2.5582453585729885e-06, + "loss": 0.3068, + "step": 8140 + }, + { + "epoch": 1.48, + "learning_rate": 2.5581907535493264e-06, + "loss": 0.9487, + "step": 8141 + }, + { + "epoch": 1.48, + "learning_rate": 2.5581361485256644e-06, + "loss": 0.2964, + "step": 8142 + }, + { + "epoch": 1.48, + "learning_rate": 2.5580815435020023e-06, + "loss": 0.962, + "step": 8143 + }, + { + "epoch": 1.48, + "learning_rate": 2.5580269384783403e-06, + "loss": 1.4641, + "step": 8144 + }, + { + "epoch": 1.48, + "learning_rate": 2.5579723334546778e-06, + "loss": 0.043, + "step": 8145 + }, + { + "epoch": 1.48, + "learning_rate": 2.5579177284310157e-06, + "loss": 1.8803, + "step": 8146 + }, + { + "epoch": 1.48, + "learning_rate": 2.5578631234073537e-06, + "loss": 2.1249, + "step": 8147 + }, + { + "epoch": 1.48, + "learning_rate": 2.557808518383691e-06, + "loss": 0.5967, + "step": 8148 + }, + { + "epoch": 1.48, + "learning_rate": 2.557753913360029e-06, + "loss": 0.7173, + "step": 8149 + }, + { + "epoch": 1.48, + "learning_rate": 2.557699308336367e-06, + "loss": 0.5818, + "step": 8150 + }, + { + "epoch": 1.48, + "learning_rate": 2.5576447033127046e-06, + "loss": 0.2502, + "step": 8151 + }, + { + "epoch": 1.48, + "learning_rate": 2.5575900982890425e-06, + "loss": 0.6588, + "step": 8152 + }, + { + "epoch": 1.48, + "learning_rate": 2.5575354932653805e-06, + "loss": 0.5285, + "step": 8153 + }, + { + "epoch": 1.48, + "learning_rate": 2.5574808882417184e-06, + "loss": 0.2974, + "step": 8154 + }, + { + "epoch": 1.48, + "learning_rate": 2.5574262832180564e-06, + "loss": 1.1043, + "step": 8155 + }, + { + "epoch": 1.48, + "learning_rate": 2.5573716781943943e-06, + "loss": 1.5999, + "step": 8156 + }, + { + "epoch": 1.48, + "learning_rate": 2.557317073170732e-06, + "loss": 0.8865, + "step": 8157 + }, + { + "epoch": 1.48, + "learning_rate": 2.5572624681470698e-06, + "loss": 1.2882, + "step": 8158 + }, + { + "epoch": 1.48, + "learning_rate": 2.5572078631234073e-06, + "loss": 1.2698, + "step": 8159 + }, + { + "epoch": 1.48, + "learning_rate": 2.5571532580997452e-06, + "loss": 1.6096, + "step": 8160 + }, + { + "epoch": 1.48, + "learning_rate": 2.557098653076083e-06, + "loss": 0.5155, + "step": 8161 + }, + { + "epoch": 1.48, + "learning_rate": 2.5570440480524207e-06, + "loss": 0.8637, + "step": 8162 + }, + { + "epoch": 1.48, + "learning_rate": 2.5569894430287586e-06, + "loss": 0.1764, + "step": 8163 + }, + { + "epoch": 1.48, + "learning_rate": 2.5569348380050966e-06, + "loss": 0.3698, + "step": 8164 + }, + { + "epoch": 1.48, + "learning_rate": 2.556880232981434e-06, + "loss": 0.5048, + "step": 8165 + }, + { + "epoch": 1.48, + "learning_rate": 2.556825627957772e-06, + "loss": 0.6613, + "step": 8166 + }, + { + "epoch": 1.49, + "learning_rate": 2.55677102293411e-06, + "loss": 0.9086, + "step": 8167 + }, + { + "epoch": 1.49, + "learning_rate": 2.556716417910448e-06, + "loss": 2.6571, + "step": 8168 + }, + { + "epoch": 1.49, + "learning_rate": 2.556661812886786e-06, + "loss": 0.4979, + "step": 8169 + }, + { + "epoch": 1.49, + "learning_rate": 2.556607207863124e-06, + "loss": 1.1078, + "step": 8170 + }, + { + "epoch": 1.49, + "learning_rate": 2.5565526028394613e-06, + "loss": 0.6336, + "step": 8171 + }, + { + "epoch": 1.49, + "learning_rate": 2.5564979978157993e-06, + "loss": 0.9165, + "step": 8172 + }, + { + "epoch": 1.49, + "learning_rate": 2.556443392792137e-06, + "loss": 0.7859, + "step": 8173 + }, + { + "epoch": 1.49, + "learning_rate": 2.5563887877684747e-06, + "loss": 0.1418, + "step": 8174 + }, + { + "epoch": 1.49, + "learning_rate": 2.5563341827448127e-06, + "loss": 0.3562, + "step": 8175 + }, + { + "epoch": 1.49, + "learning_rate": 2.55627957772115e-06, + "loss": 0.5611, + "step": 8176 + }, + { + "epoch": 1.49, + "learning_rate": 2.556224972697488e-06, + "loss": 1.317, + "step": 8177 + }, + { + "epoch": 1.49, + "learning_rate": 2.556170367673826e-06, + "loss": 0.6203, + "step": 8178 + }, + { + "epoch": 1.49, + "learning_rate": 2.5561157626501636e-06, + "loss": 0.3674, + "step": 8179 + }, + { + "epoch": 1.49, + "learning_rate": 2.5560611576265016e-06, + "loss": 0.4258, + "step": 8180 + }, + { + "epoch": 1.49, + "learning_rate": 2.5560065526028395e-06, + "loss": 0.8627, + "step": 8181 + }, + { + "epoch": 1.49, + "learning_rate": 2.5559519475791774e-06, + "loss": 2.1367, + "step": 8182 + }, + { + "epoch": 1.49, + "learning_rate": 2.5558973425555154e-06, + "loss": 0.8431, + "step": 8183 + }, + { + "epoch": 1.49, + "learning_rate": 2.5558427375318533e-06, + "loss": 1.6749, + "step": 8184 + }, + { + "epoch": 1.49, + "learning_rate": 2.555788132508191e-06, + "loss": 0.0801, + "step": 8185 + }, + { + "epoch": 1.49, + "learning_rate": 2.555733527484529e-06, + "loss": 0.0237, + "step": 8186 + }, + { + "epoch": 1.49, + "learning_rate": 2.5556789224608663e-06, + "loss": 0.5981, + "step": 8187 + }, + { + "epoch": 1.49, + "learning_rate": 2.5556243174372043e-06, + "loss": 0.9547, + "step": 8188 + }, + { + "epoch": 1.49, + "learning_rate": 2.555569712413542e-06, + "loss": 1.5483, + "step": 8189 + }, + { + "epoch": 1.49, + "learning_rate": 2.5555151073898797e-06, + "loss": 2.5211, + "step": 8190 + }, + { + "epoch": 1.49, + "learning_rate": 2.5554605023662177e-06, + "loss": 1.9415, + "step": 8191 + }, + { + "epoch": 1.49, + "learning_rate": 2.5554058973425556e-06, + "loss": 1.7824, + "step": 8192 + }, + { + "epoch": 1.49, + "learning_rate": 2.555351292318893e-06, + "loss": 0.6958, + "step": 8193 + }, + { + "epoch": 1.49, + "learning_rate": 2.555296687295231e-06, + "loss": 1.8714, + "step": 8194 + }, + { + "epoch": 1.49, + "learning_rate": 2.555242082271569e-06, + "loss": 2.1138, + "step": 8195 + }, + { + "epoch": 1.49, + "learning_rate": 2.555187477247907e-06, + "loss": 1.6579, + "step": 8196 + }, + { + "epoch": 1.49, + "learning_rate": 2.555132872224245e-06, + "loss": 1.833, + "step": 8197 + }, + { + "epoch": 1.49, + "learning_rate": 2.5550782672005824e-06, + "loss": 1.405, + "step": 8198 + }, + { + "epoch": 1.49, + "learning_rate": 2.5550236621769204e-06, + "loss": 0.6597, + "step": 8199 + }, + { + "epoch": 1.49, + "learning_rate": 2.5549690571532583e-06, + "loss": 0.5461, + "step": 8200 + }, + { + "epoch": 1.49, + "learning_rate": 2.554914452129596e-06, + "loss": 1.5484, + "step": 8201 + }, + { + "epoch": 1.49, + "learning_rate": 2.5548598471059338e-06, + "loss": 0.8656, + "step": 8202 + }, + { + "epoch": 1.49, + "learning_rate": 2.5548052420822717e-06, + "loss": 0.1317, + "step": 8203 + }, + { + "epoch": 1.49, + "learning_rate": 2.5547506370586092e-06, + "loss": 1.2362, + "step": 8204 + }, + { + "epoch": 1.49, + "learning_rate": 2.554696032034947e-06, + "loss": 0.7862, + "step": 8205 + }, + { + "epoch": 1.49, + "learning_rate": 2.554641427011285e-06, + "loss": 0.3321, + "step": 8206 + }, + { + "epoch": 1.49, + "learning_rate": 2.5545868219876226e-06, + "loss": 0.4132, + "step": 8207 + }, + { + "epoch": 1.49, + "learning_rate": 2.5545322169639606e-06, + "loss": 0.0489, + "step": 8208 + }, + { + "epoch": 1.49, + "learning_rate": 2.5544776119402985e-06, + "loss": 1.1205, + "step": 8209 + }, + { + "epoch": 1.49, + "learning_rate": 2.5544230069166365e-06, + "loss": 1.5978, + "step": 8210 + }, + { + "epoch": 1.49, + "learning_rate": 2.5543684018929744e-06, + "loss": 0.1396, + "step": 8211 + }, + { + "epoch": 1.49, + "learning_rate": 2.554313796869312e-06, + "loss": 0.1461, + "step": 8212 + }, + { + "epoch": 1.49, + "learning_rate": 2.55425919184565e-06, + "loss": 0.4291, + "step": 8213 + }, + { + "epoch": 1.49, + "learning_rate": 2.554204586821988e-06, + "loss": 0.4885, + "step": 8214 + }, + { + "epoch": 1.49, + "learning_rate": 2.5541499817983253e-06, + "loss": 0.0582, + "step": 8215 + }, + { + "epoch": 1.49, + "learning_rate": 2.5540953767746633e-06, + "loss": 1.353, + "step": 8216 + }, + { + "epoch": 1.49, + "learning_rate": 2.5540407717510012e-06, + "loss": 0.6942, + "step": 8217 + }, + { + "epoch": 1.49, + "learning_rate": 2.5539861667273388e-06, + "loss": 0.248, + "step": 8218 + }, + { + "epoch": 1.49, + "learning_rate": 2.5539315617036767e-06, + "loss": 2.2586, + "step": 8219 + }, + { + "epoch": 1.49, + "learning_rate": 2.5538769566800146e-06, + "loss": 1.4777, + "step": 8220 + }, + { + "epoch": 1.49, + "learning_rate": 2.553822351656352e-06, + "loss": 0.149, + "step": 8221 + }, + { + "epoch": 1.5, + "learning_rate": 2.55376774663269e-06, + "loss": 1.2409, + "step": 8222 + }, + { + "epoch": 1.5, + "learning_rate": 2.553713141609028e-06, + "loss": 1.0083, + "step": 8223 + }, + { + "epoch": 1.5, + "learning_rate": 2.553658536585366e-06, + "loss": 0.4933, + "step": 8224 + }, + { + "epoch": 1.5, + "learning_rate": 2.553603931561704e-06, + "loss": 0.5054, + "step": 8225 + }, + { + "epoch": 1.5, + "learning_rate": 2.5535493265380415e-06, + "loss": 1.9117, + "step": 8226 + }, + { + "epoch": 1.5, + "learning_rate": 2.5534947215143794e-06, + "loss": 1.0942, + "step": 8227 + }, + { + "epoch": 1.5, + "learning_rate": 2.5534401164907173e-06, + "loss": 0.9711, + "step": 8228 + }, + { + "epoch": 1.5, + "learning_rate": 2.553385511467055e-06, + "loss": 0.6775, + "step": 8229 + }, + { + "epoch": 1.5, + "learning_rate": 2.553330906443393e-06, + "loss": 1.694, + "step": 8230 + }, + { + "epoch": 1.5, + "learning_rate": 2.5532763014197307e-06, + "loss": 2.0523, + "step": 8231 + }, + { + "epoch": 1.5, + "learning_rate": 2.5532216963960683e-06, + "loss": 0.6106, + "step": 8232 + }, + { + "epoch": 1.5, + "learning_rate": 2.553167091372406e-06, + "loss": 0.7388, + "step": 8233 + }, + { + "epoch": 1.5, + "learning_rate": 2.553112486348744e-06, + "loss": 0.5506, + "step": 8234 + }, + { + "epoch": 1.5, + "learning_rate": 2.553057881325082e-06, + "loss": 0.8408, + "step": 8235 + }, + { + "epoch": 1.5, + "learning_rate": 2.55300327630142e-06, + "loss": 1.8971, + "step": 8236 + }, + { + "epoch": 1.5, + "learning_rate": 2.552948671277758e-06, + "loss": 3.1761, + "step": 8237 + }, + { + "epoch": 1.5, + "learning_rate": 2.5528940662540955e-06, + "loss": 0.1024, + "step": 8238 + }, + { + "epoch": 1.5, + "learning_rate": 2.5528394612304334e-06, + "loss": 0.6201, + "step": 8239 + }, + { + "epoch": 1.5, + "learning_rate": 2.552784856206771e-06, + "loss": 2.1872, + "step": 8240 + }, + { + "epoch": 1.5, + "learning_rate": 2.552730251183109e-06, + "loss": 0.4846, + "step": 8241 + }, + { + "epoch": 1.5, + "learning_rate": 2.552675646159447e-06, + "loss": 0.9359, + "step": 8242 + }, + { + "epoch": 1.5, + "learning_rate": 2.5526210411357844e-06, + "loss": 1.0425, + "step": 8243 + }, + { + "epoch": 1.5, + "learning_rate": 2.5525664361121223e-06, + "loss": 0.3488, + "step": 8244 + }, + { + "epoch": 1.5, + "learning_rate": 2.5525118310884603e-06, + "loss": 0.8196, + "step": 8245 + }, + { + "epoch": 1.5, + "learning_rate": 2.5524572260647978e-06, + "loss": 0.476, + "step": 8246 + }, + { + "epoch": 1.5, + "learning_rate": 2.5524026210411357e-06, + "loss": 0.518, + "step": 8247 + }, + { + "epoch": 1.5, + "learning_rate": 2.5523480160174737e-06, + "loss": 0.9778, + "step": 8248 + }, + { + "epoch": 1.5, + "learning_rate": 2.5522934109938116e-06, + "loss": 0.6209, + "step": 8249 + }, + { + "epoch": 1.5, + "learning_rate": 2.5522388059701496e-06, + "loss": 0.3312, + "step": 8250 + }, + { + "epoch": 1.5, + "learning_rate": 2.5521842009464875e-06, + "loss": 0.0653, + "step": 8251 + }, + { + "epoch": 1.5, + "learning_rate": 2.552129595922825e-06, + "loss": 0.0203, + "step": 8252 + }, + { + "epoch": 1.5, + "learning_rate": 2.552074990899163e-06, + "loss": 1.1646, + "step": 8253 + }, + { + "epoch": 1.5, + "learning_rate": 2.5520203858755005e-06, + "loss": 3.6216, + "step": 8254 + }, + { + "epoch": 1.5, + "learning_rate": 2.5519657808518384e-06, + "loss": 0.0929, + "step": 8255 + }, + { + "epoch": 1.5, + "learning_rate": 2.5519111758281764e-06, + "loss": 0.1398, + "step": 8256 + }, + { + "epoch": 1.5, + "learning_rate": 2.551856570804514e-06, + "loss": 0.8414, + "step": 8257 + }, + { + "epoch": 1.5, + "learning_rate": 2.551801965780852e-06, + "loss": 0.2008, + "step": 8258 + }, + { + "epoch": 1.5, + "learning_rate": 2.5517473607571898e-06, + "loss": 0.3577, + "step": 8259 + }, + { + "epoch": 1.5, + "learning_rate": 2.5516927557335273e-06, + "loss": 0.1302, + "step": 8260 + }, + { + "epoch": 1.5, + "learning_rate": 2.5516381507098652e-06, + "loss": 0.0596, + "step": 8261 + }, + { + "epoch": 1.5, + "learning_rate": 2.551583545686203e-06, + "loss": 2.9868, + "step": 8262 + }, + { + "epoch": 1.5, + "learning_rate": 2.551528940662541e-06, + "loss": 0.1666, + "step": 8263 + }, + { + "epoch": 1.5, + "learning_rate": 2.551474335638879e-06, + "loss": 2.6802, + "step": 8264 + }, + { + "epoch": 1.5, + "learning_rate": 2.551419730615217e-06, + "loss": 0.5668, + "step": 8265 + }, + { + "epoch": 1.5, + "learning_rate": 2.5513651255915545e-06, + "loss": 0.4262, + "step": 8266 + }, + { + "epoch": 1.5, + "learning_rate": 2.5513105205678925e-06, + "loss": 0.755, + "step": 8267 + }, + { + "epoch": 1.5, + "learning_rate": 2.55125591554423e-06, + "loss": 0.5858, + "step": 8268 + }, + { + "epoch": 1.5, + "learning_rate": 2.551201310520568e-06, + "loss": 0.5911, + "step": 8269 + }, + { + "epoch": 1.5, + "learning_rate": 2.551146705496906e-06, + "loss": 0.3644, + "step": 8270 + }, + { + "epoch": 1.5, + "learning_rate": 2.5510921004732434e-06, + "loss": 0.1703, + "step": 8271 + }, + { + "epoch": 1.5, + "learning_rate": 2.5510374954495813e-06, + "loss": 0.0425, + "step": 8272 + }, + { + "epoch": 1.5, + "learning_rate": 2.5509828904259193e-06, + "loss": 0.7167, + "step": 8273 + }, + { + "epoch": 1.5, + "learning_rate": 2.550928285402257e-06, + "loss": 0.0609, + "step": 8274 + }, + { + "epoch": 1.5, + "learning_rate": 2.5508736803785947e-06, + "loss": 1.7994, + "step": 8275 + }, + { + "epoch": 1.51, + "learning_rate": 2.5508190753549327e-06, + "loss": 0.8961, + "step": 8276 + }, + { + "epoch": 1.51, + "learning_rate": 2.5507644703312706e-06, + "loss": 0.8078, + "step": 8277 + }, + { + "epoch": 1.51, + "learning_rate": 2.5507098653076086e-06, + "loss": 1.115, + "step": 8278 + }, + { + "epoch": 1.51, + "learning_rate": 2.550655260283946e-06, + "loss": 0.8555, + "step": 8279 + }, + { + "epoch": 1.51, + "learning_rate": 2.550600655260284e-06, + "loss": 0.792, + "step": 8280 + }, + { + "epoch": 1.51, + "learning_rate": 2.550546050236622e-06, + "loss": 0.0474, + "step": 8281 + }, + { + "epoch": 1.51, + "learning_rate": 2.5504914452129595e-06, + "loss": 0.2994, + "step": 8282 + }, + { + "epoch": 1.51, + "learning_rate": 2.5504368401892974e-06, + "loss": 1.6772, + "step": 8283 + }, + { + "epoch": 1.51, + "learning_rate": 2.5503822351656354e-06, + "loss": 0.2159, + "step": 8284 + }, + { + "epoch": 1.51, + "learning_rate": 2.550327630141973e-06, + "loss": 0.4094, + "step": 8285 + }, + { + "epoch": 1.51, + "learning_rate": 2.550273025118311e-06, + "loss": 2.6132, + "step": 8286 + }, + { + "epoch": 1.51, + "learning_rate": 2.550218420094649e-06, + "loss": 0.8661, + "step": 8287 + }, + { + "epoch": 1.51, + "learning_rate": 2.5501638150709863e-06, + "loss": 0.6482, + "step": 8288 + }, + { + "epoch": 1.51, + "learning_rate": 2.5501092100473243e-06, + "loss": 1.5503, + "step": 8289 + }, + { + "epoch": 1.51, + "learning_rate": 2.550054605023662e-06, + "loss": 0.6817, + "step": 8290 + }, + { + "epoch": 1.51, + "learning_rate": 2.55e-06, + "loss": 0.8085, + "step": 8291 + }, + { + "epoch": 1.51, + "learning_rate": 2.549945394976338e-06, + "loss": 1.3927, + "step": 8292 + }, + { + "epoch": 1.51, + "learning_rate": 2.5498907899526756e-06, + "loss": 0.3618, + "step": 8293 + }, + { + "epoch": 1.51, + "learning_rate": 2.5498361849290136e-06, + "loss": 1.696, + "step": 8294 + }, + { + "epoch": 1.51, + "learning_rate": 2.5497815799053515e-06, + "loss": 0.8517, + "step": 8295 + }, + { + "epoch": 1.51, + "learning_rate": 2.549726974881689e-06, + "loss": 0.7863, + "step": 8296 + }, + { + "epoch": 1.51, + "learning_rate": 2.549672369858027e-06, + "loss": 0.3666, + "step": 8297 + }, + { + "epoch": 1.51, + "learning_rate": 2.549617764834365e-06, + "loss": 0.3816, + "step": 8298 + }, + { + "epoch": 1.51, + "learning_rate": 2.5495631598107024e-06, + "loss": 1.7381, + "step": 8299 + }, + { + "epoch": 1.51, + "learning_rate": 2.5495085547870404e-06, + "loss": 0.9023, + "step": 8300 + }, + { + "epoch": 1.51, + "learning_rate": 2.5494539497633783e-06, + "loss": 3.5134, + "step": 8301 + }, + { + "epoch": 1.51, + "learning_rate": 2.549399344739716e-06, + "loss": 0.7073, + "step": 8302 + }, + { + "epoch": 1.51, + "learning_rate": 2.5493447397160538e-06, + "loss": 0.4845, + "step": 8303 + }, + { + "epoch": 1.51, + "learning_rate": 2.5492901346923917e-06, + "loss": 1.1075, + "step": 8304 + }, + { + "epoch": 1.51, + "learning_rate": 2.5492355296687297e-06, + "loss": 0.4994, + "step": 8305 + }, + { + "epoch": 1.51, + "learning_rate": 2.5491809246450676e-06, + "loss": 2.0043, + "step": 8306 + }, + { + "epoch": 1.51, + "learning_rate": 2.549126319621405e-06, + "loss": 0.9147, + "step": 8307 + }, + { + "epoch": 1.51, + "learning_rate": 2.549071714597743e-06, + "loss": 1.4487, + "step": 8308 + }, + { + "epoch": 1.51, + "learning_rate": 2.549017109574081e-06, + "loss": 1.2674, + "step": 8309 + }, + { + "epoch": 1.51, + "learning_rate": 2.5489625045504185e-06, + "loss": 0.8111, + "step": 8310 + }, + { + "epoch": 1.51, + "learning_rate": 2.5489078995267565e-06, + "loss": 0.4353, + "step": 8311 + }, + { + "epoch": 1.51, + "learning_rate": 2.5488532945030944e-06, + "loss": 1.4886, + "step": 8312 + }, + { + "epoch": 1.51, + "learning_rate": 2.548798689479432e-06, + "loss": 0.7448, + "step": 8313 + }, + { + "epoch": 1.51, + "learning_rate": 2.54874408445577e-06, + "loss": 2.0154, + "step": 8314 + }, + { + "epoch": 1.51, + "learning_rate": 2.548689479432108e-06, + "loss": 0.1028, + "step": 8315 + }, + { + "epoch": 1.51, + "learning_rate": 2.5486348744084458e-06, + "loss": 1.1544, + "step": 8316 + }, + { + "epoch": 1.51, + "learning_rate": 2.5485802693847833e-06, + "loss": 0.653, + "step": 8317 + }, + { + "epoch": 1.51, + "learning_rate": 2.5485256643611212e-06, + "loss": 0.7138, + "step": 8318 + }, + { + "epoch": 1.51, + "learning_rate": 2.548471059337459e-06, + "loss": 0.7965, + "step": 8319 + }, + { + "epoch": 1.51, + "learning_rate": 2.548416454313797e-06, + "loss": 0.704, + "step": 8320 + }, + { + "epoch": 1.51, + "learning_rate": 2.5483618492901346e-06, + "loss": 0.6143, + "step": 8321 + }, + { + "epoch": 1.51, + "learning_rate": 2.5483072442664726e-06, + "loss": 1.324, + "step": 8322 + }, + { + "epoch": 1.51, + "learning_rate": 2.5482526392428105e-06, + "loss": 1.2211, + "step": 8323 + }, + { + "epoch": 1.51, + "learning_rate": 2.548198034219148e-06, + "loss": 0.5089, + "step": 8324 + }, + { + "epoch": 1.51, + "learning_rate": 2.548143429195486e-06, + "loss": 1.6509, + "step": 8325 + }, + { + "epoch": 1.51, + "learning_rate": 2.548088824171824e-06, + "loss": 0.6367, + "step": 8326 + }, + { + "epoch": 1.51, + "learning_rate": 2.5480342191481615e-06, + "loss": 0.742, + "step": 8327 + }, + { + "epoch": 1.51, + "learning_rate": 2.5479796141244994e-06, + "loss": 0.4036, + "step": 8328 + }, + { + "epoch": 1.51, + "learning_rate": 2.5479250091008373e-06, + "loss": 0.7146, + "step": 8329 + }, + { + "epoch": 1.51, + "learning_rate": 2.5478704040771753e-06, + "loss": 0.5464, + "step": 8330 + }, + { + "epoch": 1.52, + "learning_rate": 2.5478157990535132e-06, + "loss": 1.2184, + "step": 8331 + }, + { + "epoch": 1.52, + "learning_rate": 2.547761194029851e-06, + "loss": 0.5518, + "step": 8332 + }, + { + "epoch": 1.52, + "learning_rate": 2.5477065890061887e-06, + "loss": 0.6275, + "step": 8333 + }, + { + "epoch": 1.52, + "learning_rate": 2.5476519839825266e-06, + "loss": 1.1014, + "step": 8334 + }, + { + "epoch": 1.52, + "learning_rate": 2.547597378958864e-06, + "loss": 0.5874, + "step": 8335 + }, + { + "epoch": 1.52, + "learning_rate": 2.547542773935202e-06, + "loss": 0.5579, + "step": 8336 + }, + { + "epoch": 1.52, + "learning_rate": 2.54748816891154e-06, + "loss": 0.6359, + "step": 8337 + }, + { + "epoch": 1.52, + "learning_rate": 2.5474335638878776e-06, + "loss": 1.1547, + "step": 8338 + }, + { + "epoch": 1.52, + "learning_rate": 2.5473789588642155e-06, + "loss": 0.5211, + "step": 8339 + }, + { + "epoch": 1.52, + "learning_rate": 2.5473243538405534e-06, + "loss": 1.936, + "step": 8340 + }, + { + "epoch": 1.52, + "learning_rate": 2.547269748816891e-06, + "loss": 1.7786, + "step": 8341 + }, + { + "epoch": 1.52, + "learning_rate": 2.547215143793229e-06, + "loss": 1.6922, + "step": 8342 + }, + { + "epoch": 1.52, + "learning_rate": 2.547160538769567e-06, + "loss": 0.8384, + "step": 8343 + }, + { + "epoch": 1.52, + "learning_rate": 2.547105933745905e-06, + "loss": 0.0789, + "step": 8344 + }, + { + "epoch": 1.52, + "learning_rate": 2.5470513287222427e-06, + "loss": 2.0435, + "step": 8345 + }, + { + "epoch": 1.52, + "learning_rate": 2.5469967236985807e-06, + "loss": 0.1586, + "step": 8346 + }, + { + "epoch": 1.52, + "learning_rate": 2.546942118674918e-06, + "loss": 1.1657, + "step": 8347 + }, + { + "epoch": 1.52, + "learning_rate": 2.546887513651256e-06, + "loss": 0.806, + "step": 8348 + }, + { + "epoch": 1.52, + "learning_rate": 2.5468329086275937e-06, + "loss": 0.2968, + "step": 8349 + }, + { + "epoch": 1.52, + "learning_rate": 2.5467783036039316e-06, + "loss": 3.6155, + "step": 8350 + }, + { + "epoch": 1.52, + "learning_rate": 2.5467236985802696e-06, + "loss": 1.5438, + "step": 8351 + }, + { + "epoch": 1.52, + "learning_rate": 2.546669093556607e-06, + "loss": 0.8682, + "step": 8352 + }, + { + "epoch": 1.52, + "learning_rate": 2.546614488532945e-06, + "loss": 0.2783, + "step": 8353 + }, + { + "epoch": 1.52, + "learning_rate": 2.546559883509283e-06, + "loss": 0.0214, + "step": 8354 + }, + { + "epoch": 1.52, + "learning_rate": 2.5465052784856205e-06, + "loss": 0.0486, + "step": 8355 + }, + { + "epoch": 1.52, + "learning_rate": 2.5464506734619584e-06, + "loss": 1.1103, + "step": 8356 + }, + { + "epoch": 1.52, + "learning_rate": 2.5463960684382964e-06, + "loss": 0.7322, + "step": 8357 + }, + { + "epoch": 1.52, + "learning_rate": 2.5463414634146343e-06, + "loss": 0.029, + "step": 8358 + }, + { + "epoch": 1.52, + "learning_rate": 2.5462868583909723e-06, + "loss": 0.2167, + "step": 8359 + }, + { + "epoch": 1.52, + "learning_rate": 2.5462322533673098e-06, + "loss": 1.3756, + "step": 8360 + }, + { + "epoch": 1.52, + "learning_rate": 2.5461776483436477e-06, + "loss": 0.3562, + "step": 8361 + }, + { + "epoch": 1.52, + "learning_rate": 2.5461230433199857e-06, + "loss": 3.216, + "step": 8362 + }, + { + "epoch": 1.52, + "learning_rate": 2.546068438296323e-06, + "loss": 2.1664, + "step": 8363 + }, + { + "epoch": 1.52, + "learning_rate": 2.546013833272661e-06, + "loss": 0.9497, + "step": 8364 + }, + { + "epoch": 1.52, + "learning_rate": 2.545959228248999e-06, + "loss": 0.9225, + "step": 8365 + }, + { + "epoch": 1.52, + "learning_rate": 2.5459046232253366e-06, + "loss": 1.7784, + "step": 8366 + }, + { + "epoch": 1.52, + "learning_rate": 2.5458500182016745e-06, + "loss": 0.537, + "step": 8367 + }, + { + "epoch": 1.52, + "learning_rate": 2.5457954131780125e-06, + "loss": 1.0195, + "step": 8368 + }, + { + "epoch": 1.52, + "learning_rate": 2.54574080815435e-06, + "loss": 2.0291, + "step": 8369 + }, + { + "epoch": 1.52, + "learning_rate": 2.545686203130688e-06, + "loss": 0.3105, + "step": 8370 + }, + { + "epoch": 1.52, + "learning_rate": 2.545631598107026e-06, + "loss": 0.0604, + "step": 8371 + }, + { + "epoch": 1.52, + "learning_rate": 2.545576993083364e-06, + "loss": 0.7968, + "step": 8372 + }, + { + "epoch": 1.52, + "learning_rate": 2.5455223880597018e-06, + "loss": 0.7965, + "step": 8373 + }, + { + "epoch": 1.52, + "learning_rate": 2.5454677830360393e-06, + "loss": 0.5847, + "step": 8374 + }, + { + "epoch": 1.52, + "learning_rate": 2.5454131780123772e-06, + "loss": 1.5114, + "step": 8375 + }, + { + "epoch": 1.52, + "learning_rate": 2.545358572988715e-06, + "loss": 2.311, + "step": 8376 + }, + { + "epoch": 1.52, + "learning_rate": 2.5453039679650527e-06, + "loss": 1.0328, + "step": 8377 + }, + { + "epoch": 1.52, + "learning_rate": 2.5452493629413906e-06, + "loss": 1.8703, + "step": 8378 + }, + { + "epoch": 1.52, + "learning_rate": 2.5451947579177286e-06, + "loss": 0.3502, + "step": 8379 + }, + { + "epoch": 1.52, + "learning_rate": 2.545140152894066e-06, + "loss": 0.2967, + "step": 8380 + }, + { + "epoch": 1.52, + "learning_rate": 2.545085547870404e-06, + "loss": 0.4378, + "step": 8381 + }, + { + "epoch": 1.52, + "learning_rate": 2.545030942846742e-06, + "loss": 2.7817, + "step": 8382 + }, + { + "epoch": 1.52, + "learning_rate": 2.5449763378230795e-06, + "loss": 0.3007, + "step": 8383 + }, + { + "epoch": 1.52, + "learning_rate": 2.5449217327994175e-06, + "loss": 0.3437, + "step": 8384 + }, + { + "epoch": 1.52, + "learning_rate": 2.5448671277757554e-06, + "loss": 0.6076, + "step": 8385 + }, + { + "epoch": 1.53, + "learning_rate": 2.5448125227520933e-06, + "loss": 0.3679, + "step": 8386 + }, + { + "epoch": 1.53, + "learning_rate": 2.5447579177284313e-06, + "loss": 0.4447, + "step": 8387 + }, + { + "epoch": 1.53, + "learning_rate": 2.544703312704769e-06, + "loss": 1.8611, + "step": 8388 + }, + { + "epoch": 1.53, + "learning_rate": 2.5446487076811067e-06, + "loss": 3.7019, + "step": 8389 + }, + { + "epoch": 1.53, + "learning_rate": 2.5445941026574447e-06, + "loss": 0.2233, + "step": 8390 + }, + { + "epoch": 1.53, + "learning_rate": 2.544539497633782e-06, + "loss": 0.235, + "step": 8391 + }, + { + "epoch": 1.53, + "learning_rate": 2.54448489261012e-06, + "loss": 1.3909, + "step": 8392 + }, + { + "epoch": 1.53, + "learning_rate": 2.544430287586458e-06, + "loss": 0.6221, + "step": 8393 + }, + { + "epoch": 1.53, + "learning_rate": 2.5443756825627956e-06, + "loss": 1.3936, + "step": 8394 + }, + { + "epoch": 1.53, + "learning_rate": 2.5443210775391336e-06, + "loss": 0.7032, + "step": 8395 + }, + { + "epoch": 1.53, + "learning_rate": 2.5442664725154715e-06, + "loss": 1.1852, + "step": 8396 + }, + { + "epoch": 1.53, + "learning_rate": 2.544211867491809e-06, + "loss": 1.594, + "step": 8397 + }, + { + "epoch": 1.53, + "learning_rate": 2.544157262468147e-06, + "loss": 2.9346, + "step": 8398 + }, + { + "epoch": 1.53, + "learning_rate": 2.544102657444485e-06, + "loss": 0.4647, + "step": 8399 + }, + { + "epoch": 1.53, + "learning_rate": 2.544048052420823e-06, + "loss": 0.612, + "step": 8400 + }, + { + "epoch": 1.53, + "learning_rate": 2.543993447397161e-06, + "loss": 0.4239, + "step": 8401 + }, + { + "epoch": 1.53, + "learning_rate": 2.5439388423734983e-06, + "loss": 0.0508, + "step": 8402 + }, + { + "epoch": 1.53, + "learning_rate": 2.5438842373498363e-06, + "loss": 2.7059, + "step": 8403 + }, + { + "epoch": 1.53, + "learning_rate": 2.543829632326174e-06, + "loss": 1.1977, + "step": 8404 + }, + { + "epoch": 1.53, + "learning_rate": 2.5437750273025117e-06, + "loss": 4.8052, + "step": 8405 + }, + { + "epoch": 1.53, + "learning_rate": 2.5437204222788497e-06, + "loss": 0.0661, + "step": 8406 + }, + { + "epoch": 1.53, + "learning_rate": 2.5436658172551876e-06, + "loss": 0.0232, + "step": 8407 + }, + { + "epoch": 1.53, + "learning_rate": 2.543611212231525e-06, + "loss": 2.265, + "step": 8408 + }, + { + "epoch": 1.53, + "learning_rate": 2.543556607207863e-06, + "loss": 0.4683, + "step": 8409 + }, + { + "epoch": 1.53, + "learning_rate": 2.543502002184201e-06, + "loss": 0.7497, + "step": 8410 + }, + { + "epoch": 1.53, + "learning_rate": 2.543447397160539e-06, + "loss": 2.076, + "step": 8411 + }, + { + "epoch": 1.53, + "learning_rate": 2.543392792136877e-06, + "loss": 0.4484, + "step": 8412 + }, + { + "epoch": 1.53, + "learning_rate": 2.543338187113215e-06, + "loss": 0.3187, + "step": 8413 + }, + { + "epoch": 1.53, + "learning_rate": 2.5432835820895524e-06, + "loss": 1.6119, + "step": 8414 + }, + { + "epoch": 1.53, + "learning_rate": 2.5432289770658903e-06, + "loss": 0.5944, + "step": 8415 + }, + { + "epoch": 1.53, + "learning_rate": 2.543174372042228e-06, + "loss": 0.4983, + "step": 8416 + }, + { + "epoch": 1.53, + "learning_rate": 2.5431197670185658e-06, + "loss": 3.6325, + "step": 8417 + }, + { + "epoch": 1.53, + "learning_rate": 2.5430651619949037e-06, + "loss": 0.4313, + "step": 8418 + }, + { + "epoch": 1.53, + "learning_rate": 2.5430105569712412e-06, + "loss": 2.6456, + "step": 8419 + }, + { + "epoch": 1.53, + "learning_rate": 2.542955951947579e-06, + "loss": 0.1831, + "step": 8420 + }, + { + "epoch": 1.53, + "learning_rate": 2.542901346923917e-06, + "loss": 0.9757, + "step": 8421 + }, + { + "epoch": 1.53, + "learning_rate": 2.5428467419002546e-06, + "loss": 0.1246, + "step": 8422 + }, + { + "epoch": 1.53, + "learning_rate": 2.5427921368765926e-06, + "loss": 0.5865, + "step": 8423 + }, + { + "epoch": 1.53, + "learning_rate": 2.5427375318529305e-06, + "loss": 1.4543, + "step": 8424 + }, + { + "epoch": 1.53, + "learning_rate": 2.5426829268292685e-06, + "loss": 1.1927, + "step": 8425 + }, + { + "epoch": 1.53, + "learning_rate": 2.5426283218056064e-06, + "loss": 0.2907, + "step": 8426 + }, + { + "epoch": 1.53, + "learning_rate": 2.5425737167819444e-06, + "loss": 0.0367, + "step": 8427 + }, + { + "epoch": 1.53, + "learning_rate": 2.542519111758282e-06, + "loss": 0.8372, + "step": 8428 + }, + { + "epoch": 1.53, + "learning_rate": 2.54246450673462e-06, + "loss": 0.4055, + "step": 8429 + }, + { + "epoch": 1.53, + "learning_rate": 2.5424099017109573e-06, + "loss": 0.605, + "step": 8430 + }, + { + "epoch": 1.53, + "learning_rate": 2.5423552966872953e-06, + "loss": 0.304, + "step": 8431 + }, + { + "epoch": 1.53, + "learning_rate": 2.5423006916636332e-06, + "loss": 1.5651, + "step": 8432 + }, + { + "epoch": 1.53, + "learning_rate": 2.5422460866399707e-06, + "loss": 1.03, + "step": 8433 + }, + { + "epoch": 1.53, + "learning_rate": 2.5421914816163087e-06, + "loss": 1.2935, + "step": 8434 + }, + { + "epoch": 1.53, + "learning_rate": 2.5421368765926466e-06, + "loss": 0.174, + "step": 8435 + }, + { + "epoch": 1.53, + "learning_rate": 2.542082271568984e-06, + "loss": 0.7794, + "step": 8436 + }, + { + "epoch": 1.53, + "learning_rate": 2.542027666545322e-06, + "loss": 0.0298, + "step": 8437 + }, + { + "epoch": 1.53, + "learning_rate": 2.54197306152166e-06, + "loss": 0.3293, + "step": 8438 + }, + { + "epoch": 1.53, + "learning_rate": 2.541918456497998e-06, + "loss": 0.3039, + "step": 8439 + }, + { + "epoch": 1.53, + "learning_rate": 2.541863851474336e-06, + "loss": 1.7461, + "step": 8440 + }, + { + "epoch": 1.54, + "learning_rate": 2.5418092464506734e-06, + "loss": 0.5567, + "step": 8441 + }, + { + "epoch": 1.54, + "learning_rate": 2.5417546414270114e-06, + "loss": 0.2137, + "step": 8442 + }, + { + "epoch": 1.54, + "learning_rate": 2.5417000364033493e-06, + "loss": 1.1593, + "step": 8443 + }, + { + "epoch": 1.54, + "learning_rate": 2.541645431379687e-06, + "loss": 1.9307, + "step": 8444 + }, + { + "epoch": 1.54, + "learning_rate": 2.541590826356025e-06, + "loss": 0.1348, + "step": 8445 + }, + { + "epoch": 1.54, + "learning_rate": 2.5415362213323627e-06, + "loss": 0.3596, + "step": 8446 + }, + { + "epoch": 1.54, + "learning_rate": 2.5414816163087003e-06, + "loss": 1.2717, + "step": 8447 + }, + { + "epoch": 1.54, + "learning_rate": 2.541427011285038e-06, + "loss": 1.7535, + "step": 8448 + }, + { + "epoch": 1.54, + "learning_rate": 2.541372406261376e-06, + "loss": 1.7406, + "step": 8449 + }, + { + "epoch": 1.54, + "learning_rate": 2.5413178012377137e-06, + "loss": 0.691, + "step": 8450 + }, + { + "epoch": 1.54, + "learning_rate": 2.5412631962140516e-06, + "loss": 1.5307, + "step": 8451 + }, + { + "epoch": 1.54, + "learning_rate": 2.5412085911903896e-06, + "loss": 2.1832, + "step": 8452 + }, + { + "epoch": 1.54, + "learning_rate": 2.5411539861667275e-06, + "loss": 0.0657, + "step": 8453 + }, + { + "epoch": 1.54, + "learning_rate": 2.5410993811430654e-06, + "loss": 1.0429, + "step": 8454 + }, + { + "epoch": 1.54, + "learning_rate": 2.541044776119403e-06, + "loss": 0.926, + "step": 8455 + }, + { + "epoch": 1.54, + "learning_rate": 2.540990171095741e-06, + "loss": 0.2111, + "step": 8456 + }, + { + "epoch": 1.54, + "learning_rate": 2.540935566072079e-06, + "loss": 0.337, + "step": 8457 + }, + { + "epoch": 1.54, + "learning_rate": 2.5408809610484164e-06, + "loss": 1.2918, + "step": 8458 + }, + { + "epoch": 1.54, + "learning_rate": 2.5408263560247543e-06, + "loss": 2.1903, + "step": 8459 + }, + { + "epoch": 1.54, + "learning_rate": 2.5407717510010923e-06, + "loss": 1.4002, + "step": 8460 + }, + { + "epoch": 1.54, + "learning_rate": 2.5407171459774298e-06, + "loss": 1.0362, + "step": 8461 + }, + { + "epoch": 1.54, + "learning_rate": 2.5406625409537677e-06, + "loss": 2.6592, + "step": 8462 + }, + { + "epoch": 1.54, + "learning_rate": 2.5406079359301057e-06, + "loss": 1.3138, + "step": 8463 + }, + { + "epoch": 1.54, + "learning_rate": 2.540553330906443e-06, + "loss": 1.223, + "step": 8464 + }, + { + "epoch": 1.54, + "learning_rate": 2.540498725882781e-06, + "loss": 0.5202, + "step": 8465 + }, + { + "epoch": 1.54, + "learning_rate": 2.540444120859119e-06, + "loss": 0.9045, + "step": 8466 + }, + { + "epoch": 1.54, + "learning_rate": 2.540389515835457e-06, + "loss": 2.0888, + "step": 8467 + }, + { + "epoch": 1.54, + "learning_rate": 2.540334910811795e-06, + "loss": 0.0699, + "step": 8468 + }, + { + "epoch": 1.54, + "learning_rate": 2.5402803057881325e-06, + "loss": 2.7266, + "step": 8469 + }, + { + "epoch": 1.54, + "learning_rate": 2.5402257007644704e-06, + "loss": 0.1148, + "step": 8470 + }, + { + "epoch": 1.54, + "learning_rate": 2.5401710957408084e-06, + "loss": 0.1031, + "step": 8471 + }, + { + "epoch": 1.54, + "learning_rate": 2.540116490717146e-06, + "loss": 1.5465, + "step": 8472 + }, + { + "epoch": 1.54, + "learning_rate": 2.540061885693484e-06, + "loss": 0.3785, + "step": 8473 + }, + { + "epoch": 1.54, + "learning_rate": 2.5400072806698218e-06, + "loss": 2.7013, + "step": 8474 + }, + { + "epoch": 1.54, + "learning_rate": 2.5399526756461593e-06, + "loss": 0.3372, + "step": 8475 + }, + { + "epoch": 1.54, + "learning_rate": 2.5398980706224972e-06, + "loss": 0.9992, + "step": 8476 + }, + { + "epoch": 1.54, + "learning_rate": 2.539843465598835e-06, + "loss": 0.573, + "step": 8477 + }, + { + "epoch": 1.54, + "learning_rate": 2.5397888605751727e-06, + "loss": 0.0641, + "step": 8478 + }, + { + "epoch": 1.54, + "learning_rate": 2.5397342555515106e-06, + "loss": 0.6914, + "step": 8479 + }, + { + "epoch": 1.54, + "learning_rate": 2.5396796505278486e-06, + "loss": 0.2208, + "step": 8480 + }, + { + "epoch": 1.54, + "learning_rate": 2.5396250455041865e-06, + "loss": 0.6604, + "step": 8481 + }, + { + "epoch": 1.54, + "learning_rate": 2.5395704404805245e-06, + "loss": 0.9775, + "step": 8482 + }, + { + "epoch": 1.54, + "learning_rate": 2.539515835456862e-06, + "loss": 1.444, + "step": 8483 + }, + { + "epoch": 1.54, + "learning_rate": 2.5394612304332e-06, + "loss": 0.8088, + "step": 8484 + }, + { + "epoch": 1.54, + "learning_rate": 2.539406625409538e-06, + "loss": 0.6874, + "step": 8485 + }, + { + "epoch": 1.54, + "learning_rate": 2.5393520203858754e-06, + "loss": 0.7271, + "step": 8486 + }, + { + "epoch": 1.54, + "learning_rate": 2.5392974153622133e-06, + "loss": 1.8067, + "step": 8487 + }, + { + "epoch": 1.54, + "learning_rate": 2.5392428103385513e-06, + "loss": 0.7624, + "step": 8488 + }, + { + "epoch": 1.54, + "learning_rate": 2.539188205314889e-06, + "loss": 0.3396, + "step": 8489 + }, + { + "epoch": 1.54, + "learning_rate": 2.5391336002912267e-06, + "loss": 3.295, + "step": 8490 + }, + { + "epoch": 1.54, + "learning_rate": 2.5390789952675647e-06, + "loss": 2.1084, + "step": 8491 + }, + { + "epoch": 1.54, + "learning_rate": 2.5390243902439026e-06, + "loss": 1.8184, + "step": 8492 + }, + { + "epoch": 1.54, + "learning_rate": 2.5389697852202406e-06, + "loss": 0.4726, + "step": 8493 + }, + { + "epoch": 1.54, + "learning_rate": 2.5389151801965785e-06, + "loss": 1.1798, + "step": 8494 + }, + { + "epoch": 1.54, + "learning_rate": 2.538860575172916e-06, + "loss": 1.6022, + "step": 8495 + }, + { + "epoch": 1.55, + "learning_rate": 2.538805970149254e-06, + "loss": 0.7141, + "step": 8496 + }, + { + "epoch": 1.55, + "learning_rate": 2.5387513651255915e-06, + "loss": 0.9241, + "step": 8497 + }, + { + "epoch": 1.55, + "learning_rate": 2.5386967601019294e-06, + "loss": 0.5808, + "step": 8498 + }, + { + "epoch": 1.55, + "learning_rate": 2.5386421550782674e-06, + "loss": 1.7769, + "step": 8499 + }, + { + "epoch": 1.55, + "learning_rate": 2.538587550054605e-06, + "loss": 0.1642, + "step": 8500 + }, + { + "epoch": 1.55, + "learning_rate": 2.538532945030943e-06, + "loss": 0.5035, + "step": 8501 + }, + { + "epoch": 1.55, + "learning_rate": 2.538478340007281e-06, + "loss": 0.356, + "step": 8502 + }, + { + "epoch": 1.55, + "learning_rate": 2.5384237349836183e-06, + "loss": 1.4461, + "step": 8503 + }, + { + "epoch": 1.55, + "learning_rate": 2.5383691299599563e-06, + "loss": 0.7746, + "step": 8504 + }, + { + "epoch": 1.55, + "learning_rate": 2.538314524936294e-06, + "loss": 1.3767, + "step": 8505 + }, + { + "epoch": 1.55, + "learning_rate": 2.538259919912632e-06, + "loss": 0.3087, + "step": 8506 + }, + { + "epoch": 1.55, + "learning_rate": 2.53820531488897e-06, + "loss": 1.3439, + "step": 8507 + }, + { + "epoch": 1.55, + "learning_rate": 2.538150709865308e-06, + "loss": 0.0464, + "step": 8508 + }, + { + "epoch": 1.55, + "learning_rate": 2.5380961048416456e-06, + "loss": 0.6098, + "step": 8509 + }, + { + "epoch": 1.55, + "learning_rate": 2.5380414998179835e-06, + "loss": 1.7142, + "step": 8510 + }, + { + "epoch": 1.55, + "learning_rate": 2.537986894794321e-06, + "loss": 0.4095, + "step": 8511 + }, + { + "epoch": 1.55, + "learning_rate": 2.537932289770659e-06, + "loss": 1.8689, + "step": 8512 + }, + { + "epoch": 1.55, + "learning_rate": 2.537877684746997e-06, + "loss": 1.0417, + "step": 8513 + }, + { + "epoch": 1.55, + "learning_rate": 2.5378230797233344e-06, + "loss": 0.7258, + "step": 8514 + }, + { + "epoch": 1.55, + "learning_rate": 2.5377684746996724e-06, + "loss": 1.8095, + "step": 8515 + }, + { + "epoch": 1.55, + "learning_rate": 2.5377138696760103e-06, + "loss": 0.0953, + "step": 8516 + }, + { + "epoch": 1.55, + "learning_rate": 2.537659264652348e-06, + "loss": 2.0819, + "step": 8517 + }, + { + "epoch": 1.55, + "learning_rate": 2.5376046596286858e-06, + "loss": 0.1491, + "step": 8518 + }, + { + "epoch": 1.55, + "learning_rate": 2.5375500546050237e-06, + "loss": 1.7004, + "step": 8519 + }, + { + "epoch": 1.55, + "learning_rate": 2.5374954495813617e-06, + "loss": 0.6178, + "step": 8520 + }, + { + "epoch": 1.55, + "learning_rate": 2.5374408445576996e-06, + "loss": 1.3021, + "step": 8521 + }, + { + "epoch": 1.55, + "learning_rate": 2.537386239534037e-06, + "loss": 0.4459, + "step": 8522 + }, + { + "epoch": 1.55, + "learning_rate": 2.537331634510375e-06, + "loss": 1.7786, + "step": 8523 + }, + { + "epoch": 1.55, + "learning_rate": 2.537277029486713e-06, + "loss": 0.5128, + "step": 8524 + }, + { + "epoch": 1.55, + "learning_rate": 2.5372224244630505e-06, + "loss": 0.8698, + "step": 8525 + }, + { + "epoch": 1.55, + "learning_rate": 2.5371678194393885e-06, + "loss": 0.5924, + "step": 8526 + }, + { + "epoch": 1.55, + "learning_rate": 2.5371132144157264e-06, + "loss": 2.3153, + "step": 8527 + }, + { + "epoch": 1.55, + "learning_rate": 2.537058609392064e-06, + "loss": 1.6095, + "step": 8528 + }, + { + "epoch": 1.55, + "learning_rate": 2.537004004368402e-06, + "loss": 0.2646, + "step": 8529 + }, + { + "epoch": 1.55, + "learning_rate": 2.53694939934474e-06, + "loss": 1.719, + "step": 8530 + }, + { + "epoch": 1.55, + "learning_rate": 2.5368947943210773e-06, + "loss": 0.5816, + "step": 8531 + }, + { + "epoch": 1.55, + "learning_rate": 2.5368401892974153e-06, + "loss": 0.0767, + "step": 8532 + }, + { + "epoch": 1.55, + "learning_rate": 2.5367855842737532e-06, + "loss": 2.3207, + "step": 8533 + }, + { + "epoch": 1.55, + "learning_rate": 2.536730979250091e-06, + "loss": 1.2349, + "step": 8534 + }, + { + "epoch": 1.55, + "learning_rate": 2.536676374226429e-06, + "loss": 0.0644, + "step": 8535 + }, + { + "epoch": 1.55, + "learning_rate": 2.5366217692027666e-06, + "loss": 1.6788, + "step": 8536 + }, + { + "epoch": 1.55, + "learning_rate": 2.5365671641791046e-06, + "loss": 0.6146, + "step": 8537 + }, + { + "epoch": 1.55, + "learning_rate": 2.5365125591554425e-06, + "loss": 0.4698, + "step": 8538 + }, + { + "epoch": 1.55, + "learning_rate": 2.53645795413178e-06, + "loss": 1.041, + "step": 8539 + }, + { + "epoch": 1.55, + "learning_rate": 2.536403349108118e-06, + "loss": 1.3906, + "step": 8540 + }, + { + "epoch": 1.55, + "learning_rate": 2.536348744084456e-06, + "loss": 2.1962, + "step": 8541 + }, + { + "epoch": 1.55, + "learning_rate": 2.5362941390607934e-06, + "loss": 0.7637, + "step": 8542 + }, + { + "epoch": 1.55, + "learning_rate": 2.5362395340371314e-06, + "loss": 1.7155, + "step": 8543 + }, + { + "epoch": 1.55, + "learning_rate": 2.5361849290134693e-06, + "loss": 0.9517, + "step": 8544 + }, + { + "epoch": 1.55, + "learning_rate": 2.536130323989807e-06, + "loss": 1.3801, + "step": 8545 + }, + { + "epoch": 1.55, + "learning_rate": 2.536075718966145e-06, + "loss": 0.1242, + "step": 8546 + }, + { + "epoch": 1.55, + "learning_rate": 2.5360211139424827e-06, + "loss": 0.2129, + "step": 8547 + }, + { + "epoch": 1.55, + "learning_rate": 2.5359665089188207e-06, + "loss": 1.4907, + "step": 8548 + }, + { + "epoch": 1.55, + "learning_rate": 2.5359119038951586e-06, + "loss": 0.0194, + "step": 8549 + }, + { + "epoch": 1.55, + "learning_rate": 2.535857298871496e-06, + "loss": 0.3168, + "step": 8550 + }, + { + "epoch": 1.56, + "learning_rate": 2.535802693847834e-06, + "loss": 0.8066, + "step": 8551 + }, + { + "epoch": 1.56, + "learning_rate": 2.535748088824172e-06, + "loss": 0.451, + "step": 8552 + }, + { + "epoch": 1.56, + "learning_rate": 2.5356934838005096e-06, + "loss": 0.197, + "step": 8553 + }, + { + "epoch": 1.56, + "learning_rate": 2.5356388787768475e-06, + "loss": 1.0147, + "step": 8554 + }, + { + "epoch": 1.56, + "learning_rate": 2.5355842737531854e-06, + "loss": 0.4068, + "step": 8555 + }, + { + "epoch": 1.56, + "learning_rate": 2.535529668729523e-06, + "loss": 0.3011, + "step": 8556 + }, + { + "epoch": 1.56, + "learning_rate": 2.535475063705861e-06, + "loss": 0.0653, + "step": 8557 + }, + { + "epoch": 1.56, + "learning_rate": 2.535420458682199e-06, + "loss": 0.6979, + "step": 8558 + }, + { + "epoch": 1.56, + "learning_rate": 2.5353658536585364e-06, + "loss": 1.2389, + "step": 8559 + }, + { + "epoch": 1.56, + "learning_rate": 2.5353112486348743e-06, + "loss": 0.2016, + "step": 8560 + }, + { + "epoch": 1.56, + "learning_rate": 2.5352566436112123e-06, + "loss": 0.6449, + "step": 8561 + }, + { + "epoch": 1.56, + "learning_rate": 2.53520203858755e-06, + "loss": 0.757, + "step": 8562 + }, + { + "epoch": 1.56, + "learning_rate": 2.535147433563888e-06, + "loss": 2.1907, + "step": 8563 + }, + { + "epoch": 1.56, + "learning_rate": 2.5350928285402257e-06, + "loss": 0.7735, + "step": 8564 + }, + { + "epoch": 1.56, + "learning_rate": 2.5350382235165636e-06, + "loss": 1.0031, + "step": 8565 + }, + { + "epoch": 1.56, + "learning_rate": 2.5349836184929015e-06, + "loss": 0.9821, + "step": 8566 + }, + { + "epoch": 1.56, + "learning_rate": 2.534929013469239e-06, + "loss": 0.4748, + "step": 8567 + }, + { + "epoch": 1.56, + "learning_rate": 2.534874408445577e-06, + "loss": 0.3798, + "step": 8568 + }, + { + "epoch": 1.56, + "learning_rate": 2.534819803421915e-06, + "loss": 0.7075, + "step": 8569 + }, + { + "epoch": 1.56, + "learning_rate": 2.5347651983982525e-06, + "loss": 0.2838, + "step": 8570 + }, + { + "epoch": 1.56, + "learning_rate": 2.5347105933745904e-06, + "loss": 2.5866, + "step": 8571 + }, + { + "epoch": 1.56, + "learning_rate": 2.5346559883509284e-06, + "loss": 0.6422, + "step": 8572 + }, + { + "epoch": 1.56, + "learning_rate": 2.5346013833272663e-06, + "loss": 1.0729, + "step": 8573 + }, + { + "epoch": 1.56, + "learning_rate": 2.5345467783036042e-06, + "loss": 0.2714, + "step": 8574 + }, + { + "epoch": 1.56, + "learning_rate": 2.534492173279942e-06, + "loss": 0.4803, + "step": 8575 + }, + { + "epoch": 1.56, + "learning_rate": 2.5344375682562797e-06, + "loss": 0.9858, + "step": 8576 + }, + { + "epoch": 1.56, + "learning_rate": 2.5343829632326177e-06, + "loss": 0.373, + "step": 8577 + }, + { + "epoch": 1.56, + "learning_rate": 2.534328358208955e-06, + "loss": 1.2566, + "step": 8578 + }, + { + "epoch": 1.56, + "learning_rate": 2.534273753185293e-06, + "loss": 1.2809, + "step": 8579 + }, + { + "epoch": 1.56, + "learning_rate": 2.534219148161631e-06, + "loss": 3.1597, + "step": 8580 + }, + { + "epoch": 1.56, + "learning_rate": 2.5341645431379686e-06, + "loss": 0.545, + "step": 8581 + }, + { + "epoch": 1.56, + "learning_rate": 2.5341099381143065e-06, + "loss": 1.0218, + "step": 8582 + }, + { + "epoch": 1.56, + "learning_rate": 2.5340553330906445e-06, + "loss": 0.3523, + "step": 8583 + }, + { + "epoch": 1.56, + "learning_rate": 2.534000728066982e-06, + "loss": 0.1793, + "step": 8584 + }, + { + "epoch": 1.56, + "learning_rate": 2.53394612304332e-06, + "loss": 0.9583, + "step": 8585 + }, + { + "epoch": 1.56, + "learning_rate": 2.533891518019658e-06, + "loss": 1.7055, + "step": 8586 + }, + { + "epoch": 1.56, + "learning_rate": 2.533836912995996e-06, + "loss": 0.3051, + "step": 8587 + }, + { + "epoch": 1.56, + "learning_rate": 2.5337823079723338e-06, + "loss": 0.4797, + "step": 8588 + }, + { + "epoch": 1.56, + "learning_rate": 2.5337277029486713e-06, + "loss": 0.985, + "step": 8589 + }, + { + "epoch": 1.56, + "learning_rate": 2.5336730979250092e-06, + "loss": 1.0279, + "step": 8590 + }, + { + "epoch": 1.56, + "learning_rate": 2.533618492901347e-06, + "loss": 3.7629, + "step": 8591 + }, + { + "epoch": 1.56, + "learning_rate": 2.5335638878776847e-06, + "loss": 2.4314, + "step": 8592 + }, + { + "epoch": 1.56, + "learning_rate": 2.5335092828540226e-06, + "loss": 0.9103, + "step": 8593 + }, + { + "epoch": 1.56, + "learning_rate": 2.5334546778303606e-06, + "loss": 0.3183, + "step": 8594 + }, + { + "epoch": 1.56, + "learning_rate": 2.533400072806698e-06, + "loss": 0.0986, + "step": 8595 + }, + { + "epoch": 1.56, + "learning_rate": 2.533345467783036e-06, + "loss": 0.0395, + "step": 8596 + }, + { + "epoch": 1.56, + "learning_rate": 2.533290862759374e-06, + "loss": 0.7673, + "step": 8597 + }, + { + "epoch": 1.56, + "learning_rate": 2.5332362577357115e-06, + "loss": 0.3357, + "step": 8598 + }, + { + "epoch": 1.56, + "learning_rate": 2.5331816527120494e-06, + "loss": 0.0471, + "step": 8599 + }, + { + "epoch": 1.56, + "learning_rate": 2.5331270476883874e-06, + "loss": 0.4619, + "step": 8600 + }, + { + "epoch": 1.56, + "learning_rate": 2.5330724426647253e-06, + "loss": 0.5251, + "step": 8601 + }, + { + "epoch": 1.56, + "learning_rate": 2.5330178376410633e-06, + "loss": 1.2165, + "step": 8602 + }, + { + "epoch": 1.56, + "learning_rate": 2.532963232617401e-06, + "loss": 0.058, + "step": 8603 + }, + { + "epoch": 1.56, + "learning_rate": 2.5329086275937387e-06, + "loss": 0.5341, + "step": 8604 + }, + { + "epoch": 1.56, + "learning_rate": 2.5328540225700767e-06, + "loss": 0.8051, + "step": 8605 + }, + { + "epoch": 1.57, + "learning_rate": 2.532799417546414e-06, + "loss": 1.1166, + "step": 8606 + }, + { + "epoch": 1.57, + "learning_rate": 2.532744812522752e-06, + "loss": 1.8131, + "step": 8607 + }, + { + "epoch": 1.57, + "learning_rate": 2.53269020749909e-06, + "loss": 0.4892, + "step": 8608 + }, + { + "epoch": 1.57, + "learning_rate": 2.5326356024754276e-06, + "loss": 0.7047, + "step": 8609 + }, + { + "epoch": 1.57, + "learning_rate": 2.5325809974517656e-06, + "loss": 1.1875, + "step": 8610 + }, + { + "epoch": 1.57, + "learning_rate": 2.5325263924281035e-06, + "loss": 1.801, + "step": 8611 + }, + { + "epoch": 1.57, + "learning_rate": 2.532471787404441e-06, + "loss": 0.1273, + "step": 8612 + }, + { + "epoch": 1.57, + "learning_rate": 2.532417182380779e-06, + "loss": 0.0995, + "step": 8613 + }, + { + "epoch": 1.57, + "learning_rate": 2.532362577357117e-06, + "loss": 0.3736, + "step": 8614 + }, + { + "epoch": 1.57, + "learning_rate": 2.532307972333455e-06, + "loss": 1.39, + "step": 8615 + }, + { + "epoch": 1.57, + "learning_rate": 2.532253367309793e-06, + "loss": 0.9302, + "step": 8616 + }, + { + "epoch": 1.57, + "learning_rate": 2.5321987622861303e-06, + "loss": 0.6791, + "step": 8617 + }, + { + "epoch": 1.57, + "learning_rate": 2.5321441572624683e-06, + "loss": 0.1121, + "step": 8618 + }, + { + "epoch": 1.57, + "learning_rate": 2.532089552238806e-06, + "loss": 1.0229, + "step": 8619 + }, + { + "epoch": 1.57, + "learning_rate": 2.5320349472151437e-06, + "loss": 0.8555, + "step": 8620 + }, + { + "epoch": 1.57, + "learning_rate": 2.5319803421914817e-06, + "loss": 0.0802, + "step": 8621 + }, + { + "epoch": 1.57, + "learning_rate": 2.5319257371678196e-06, + "loss": 0.0329, + "step": 8622 + }, + { + "epoch": 1.57, + "learning_rate": 2.531871132144157e-06, + "loss": 0.5638, + "step": 8623 + }, + { + "epoch": 1.57, + "learning_rate": 2.531816527120495e-06, + "loss": 0.9784, + "step": 8624 + }, + { + "epoch": 1.57, + "learning_rate": 2.531761922096833e-06, + "loss": 0.189, + "step": 8625 + }, + { + "epoch": 1.57, + "learning_rate": 2.5317073170731705e-06, + "loss": 1.2015, + "step": 8626 + }, + { + "epoch": 1.57, + "learning_rate": 2.5316527120495085e-06, + "loss": 1.2759, + "step": 8627 + }, + { + "epoch": 1.57, + "learning_rate": 2.5315981070258464e-06, + "loss": 0.8475, + "step": 8628 + }, + { + "epoch": 1.57, + "learning_rate": 2.5315435020021844e-06, + "loss": 2.5233, + "step": 8629 + }, + { + "epoch": 1.57, + "learning_rate": 2.5314888969785223e-06, + "loss": 0.3882, + "step": 8630 + }, + { + "epoch": 1.57, + "learning_rate": 2.53143429195486e-06, + "loss": 0.0364, + "step": 8631 + }, + { + "epoch": 1.57, + "learning_rate": 2.5313796869311978e-06, + "loss": 1.18, + "step": 8632 + }, + { + "epoch": 1.57, + "learning_rate": 2.5313250819075357e-06, + "loss": 0.8158, + "step": 8633 + }, + { + "epoch": 1.57, + "learning_rate": 2.5312704768838732e-06, + "loss": 0.0759, + "step": 8634 + }, + { + "epoch": 1.57, + "learning_rate": 2.531215871860211e-06, + "loss": 1.5201, + "step": 8635 + }, + { + "epoch": 1.57, + "learning_rate": 2.531161266836549e-06, + "loss": 0.8922, + "step": 8636 + }, + { + "epoch": 1.57, + "learning_rate": 2.5311066618128866e-06, + "loss": 0.0607, + "step": 8637 + }, + { + "epoch": 1.57, + "learning_rate": 2.5310520567892246e-06, + "loss": 1.4448, + "step": 8638 + }, + { + "epoch": 1.57, + "learning_rate": 2.5309974517655625e-06, + "loss": 1.3176, + "step": 8639 + }, + { + "epoch": 1.57, + "learning_rate": 2.5309428467419e-06, + "loss": 0.3254, + "step": 8640 + }, + { + "epoch": 1.57, + "learning_rate": 2.530888241718238e-06, + "loss": 0.6246, + "step": 8641 + }, + { + "epoch": 1.57, + "learning_rate": 2.530833636694576e-06, + "loss": 1.2784, + "step": 8642 + }, + { + "epoch": 1.57, + "learning_rate": 2.530779031670914e-06, + "loss": 2.4859, + "step": 8643 + }, + { + "epoch": 1.57, + "learning_rate": 2.530724426647252e-06, + "loss": 0.6327, + "step": 8644 + }, + { + "epoch": 1.57, + "learning_rate": 2.5306698216235893e-06, + "loss": 0.6007, + "step": 8645 + }, + { + "epoch": 1.57, + "learning_rate": 2.5306152165999273e-06, + "loss": 0.6886, + "step": 8646 + }, + { + "epoch": 1.57, + "learning_rate": 2.5305606115762652e-06, + "loss": 2.7165, + "step": 8647 + }, + { + "epoch": 1.57, + "learning_rate": 2.5305060065526027e-06, + "loss": 0.935, + "step": 8648 + }, + { + "epoch": 1.57, + "learning_rate": 2.5304514015289407e-06, + "loss": 0.8439, + "step": 8649 + }, + { + "epoch": 1.57, + "learning_rate": 2.5303967965052786e-06, + "loss": 0.2315, + "step": 8650 + }, + { + "epoch": 1.57, + "learning_rate": 2.530342191481616e-06, + "loss": 1.9487, + "step": 8651 + }, + { + "epoch": 1.57, + "learning_rate": 2.530287586457954e-06, + "loss": 0.5305, + "step": 8652 + }, + { + "epoch": 1.57, + "learning_rate": 2.530232981434292e-06, + "loss": 0.7908, + "step": 8653 + }, + { + "epoch": 1.57, + "learning_rate": 2.5301783764106296e-06, + "loss": 2.5015, + "step": 8654 + }, + { + "epoch": 1.57, + "learning_rate": 2.5301237713869675e-06, + "loss": 1.1059, + "step": 8655 + }, + { + "epoch": 1.57, + "learning_rate": 2.5300691663633054e-06, + "loss": 1.2184, + "step": 8656 + }, + { + "epoch": 1.57, + "learning_rate": 2.5300145613396434e-06, + "loss": 1.2683, + "step": 8657 + }, + { + "epoch": 1.57, + "learning_rate": 2.5299599563159813e-06, + "loss": 2.7918, + "step": 8658 + }, + { + "epoch": 1.57, + "learning_rate": 2.529905351292319e-06, + "loss": 0.6492, + "step": 8659 + }, + { + "epoch": 1.57, + "learning_rate": 2.529850746268657e-06, + "loss": 1.047, + "step": 8660 + }, + { + "epoch": 1.58, + "learning_rate": 2.5297961412449947e-06, + "loss": 0.7297, + "step": 8661 + }, + { + "epoch": 1.58, + "learning_rate": 2.5297415362213323e-06, + "loss": 0.7464, + "step": 8662 + }, + { + "epoch": 1.58, + "learning_rate": 2.52968693119767e-06, + "loss": 1.456, + "step": 8663 + }, + { + "epoch": 1.58, + "learning_rate": 2.529632326174008e-06, + "loss": 1.1212, + "step": 8664 + }, + { + "epoch": 1.58, + "learning_rate": 2.5295777211503457e-06, + "loss": 0.0551, + "step": 8665 + }, + { + "epoch": 1.58, + "learning_rate": 2.5295231161266836e-06, + "loss": 0.1644, + "step": 8666 + }, + { + "epoch": 1.58, + "learning_rate": 2.5294685111030215e-06, + "loss": 1.273, + "step": 8667 + }, + { + "epoch": 1.58, + "learning_rate": 2.5294139060793595e-06, + "loss": 0.5944, + "step": 8668 + }, + { + "epoch": 1.58, + "learning_rate": 2.5293593010556974e-06, + "loss": 0.6114, + "step": 8669 + }, + { + "epoch": 1.58, + "learning_rate": 2.529304696032035e-06, + "loss": 0.5416, + "step": 8670 + }, + { + "epoch": 1.58, + "learning_rate": 2.529250091008373e-06, + "loss": 0.9888, + "step": 8671 + }, + { + "epoch": 1.58, + "learning_rate": 2.529195485984711e-06, + "loss": 0.9214, + "step": 8672 + }, + { + "epoch": 1.58, + "learning_rate": 2.5291408809610484e-06, + "loss": 0.163, + "step": 8673 + }, + { + "epoch": 1.58, + "learning_rate": 2.5290862759373863e-06, + "loss": 0.3047, + "step": 8674 + }, + { + "epoch": 1.58, + "learning_rate": 2.5290316709137243e-06, + "loss": 0.0306, + "step": 8675 + }, + { + "epoch": 1.58, + "learning_rate": 2.5289770658900618e-06, + "loss": 2.6796, + "step": 8676 + }, + { + "epoch": 1.58, + "learning_rate": 2.5289224608663997e-06, + "loss": 1.8224, + "step": 8677 + }, + { + "epoch": 1.58, + "learning_rate": 2.5288678558427377e-06, + "loss": 0.4674, + "step": 8678 + }, + { + "epoch": 1.58, + "learning_rate": 2.528813250819075e-06, + "loss": 0.8776, + "step": 8679 + }, + { + "epoch": 1.58, + "learning_rate": 2.528758645795413e-06, + "loss": 0.2428, + "step": 8680 + }, + { + "epoch": 1.58, + "learning_rate": 2.528704040771751e-06, + "loss": 0.1758, + "step": 8681 + }, + { + "epoch": 1.58, + "learning_rate": 2.528649435748089e-06, + "loss": 0.3246, + "step": 8682 + }, + { + "epoch": 1.58, + "learning_rate": 2.528594830724427e-06, + "loss": 0.4979, + "step": 8683 + }, + { + "epoch": 1.58, + "learning_rate": 2.5285402257007645e-06, + "loss": 1.8125, + "step": 8684 + }, + { + "epoch": 1.58, + "learning_rate": 2.5284856206771024e-06, + "loss": 2.7604, + "step": 8685 + }, + { + "epoch": 1.58, + "learning_rate": 2.5284310156534404e-06, + "loss": 1.5339, + "step": 8686 + }, + { + "epoch": 1.58, + "learning_rate": 2.528376410629778e-06, + "loss": 0.2686, + "step": 8687 + }, + { + "epoch": 1.58, + "learning_rate": 2.528321805606116e-06, + "loss": 2.1231, + "step": 8688 + }, + { + "epoch": 1.58, + "learning_rate": 2.5282672005824538e-06, + "loss": 0.0552, + "step": 8689 + }, + { + "epoch": 1.58, + "learning_rate": 2.5282125955587913e-06, + "loss": 0.0324, + "step": 8690 + }, + { + "epoch": 1.58, + "learning_rate": 2.5281579905351292e-06, + "loss": 1.7508, + "step": 8691 + }, + { + "epoch": 1.58, + "learning_rate": 2.528103385511467e-06, + "loss": 2.1309, + "step": 8692 + }, + { + "epoch": 1.58, + "learning_rate": 2.5280487804878047e-06, + "loss": 0.079, + "step": 8693 + }, + { + "epoch": 1.58, + "learning_rate": 2.5279941754641426e-06, + "loss": 0.0769, + "step": 8694 + }, + { + "epoch": 1.58, + "learning_rate": 2.5279395704404806e-06, + "loss": 0.7944, + "step": 8695 + }, + { + "epoch": 1.58, + "learning_rate": 2.5278849654168185e-06, + "loss": 3.1098, + "step": 8696 + }, + { + "epoch": 1.58, + "learning_rate": 2.5278303603931565e-06, + "loss": 2.1068, + "step": 8697 + }, + { + "epoch": 1.58, + "learning_rate": 2.527775755369494e-06, + "loss": 1.1296, + "step": 8698 + }, + { + "epoch": 1.58, + "learning_rate": 2.527721150345832e-06, + "loss": 1.347, + "step": 8699 + }, + { + "epoch": 1.58, + "learning_rate": 2.52766654532217e-06, + "loss": 0.4868, + "step": 8700 + }, + { + "epoch": 1.58, + "learning_rate": 2.5276119402985074e-06, + "loss": 1.4091, + "step": 8701 + }, + { + "epoch": 1.58, + "learning_rate": 2.5275573352748453e-06, + "loss": 2.9676, + "step": 8702 + }, + { + "epoch": 1.58, + "learning_rate": 2.5275027302511833e-06, + "loss": 0.7445, + "step": 8703 + }, + { + "epoch": 1.58, + "learning_rate": 2.527448125227521e-06, + "loss": 0.4956, + "step": 8704 + }, + { + "epoch": 1.58, + "learning_rate": 2.5273935202038587e-06, + "loss": 0.2931, + "step": 8705 + }, + { + "epoch": 1.58, + "learning_rate": 2.5273389151801967e-06, + "loss": 1.821, + "step": 8706 + }, + { + "epoch": 1.58, + "learning_rate": 2.527284310156534e-06, + "loss": 1.5588, + "step": 8707 + }, + { + "epoch": 1.58, + "learning_rate": 2.527229705132872e-06, + "loss": 0.7132, + "step": 8708 + }, + { + "epoch": 1.58, + "learning_rate": 2.52717510010921e-06, + "loss": 1.0402, + "step": 8709 + }, + { + "epoch": 1.58, + "learning_rate": 2.527120495085548e-06, + "loss": 0.5261, + "step": 8710 + }, + { + "epoch": 1.58, + "learning_rate": 2.527065890061886e-06, + "loss": 0.4295, + "step": 8711 + }, + { + "epoch": 1.58, + "learning_rate": 2.5270112850382235e-06, + "loss": 1.9301, + "step": 8712 + }, + { + "epoch": 1.58, + "learning_rate": 2.5269566800145614e-06, + "loss": 0.4413, + "step": 8713 + }, + { + "epoch": 1.58, + "learning_rate": 2.5269020749908994e-06, + "loss": 0.6993, + "step": 8714 + }, + { + "epoch": 1.58, + "learning_rate": 2.526847469967237e-06, + "loss": 0.4309, + "step": 8715 + }, + { + "epoch": 1.59, + "learning_rate": 2.526792864943575e-06, + "loss": 1.0286, + "step": 8716 + }, + { + "epoch": 1.59, + "learning_rate": 2.526738259919913e-06, + "loss": 0.0918, + "step": 8717 + }, + { + "epoch": 1.59, + "learning_rate": 2.5266836548962503e-06, + "loss": 0.8188, + "step": 8718 + }, + { + "epoch": 1.59, + "learning_rate": 2.5266290498725883e-06, + "loss": 0.6948, + "step": 8719 + }, + { + "epoch": 1.59, + "learning_rate": 2.526574444848926e-06, + "loss": 2.183, + "step": 8720 + }, + { + "epoch": 1.59, + "learning_rate": 2.5265198398252637e-06, + "loss": 0.1503, + "step": 8721 + }, + { + "epoch": 1.59, + "learning_rate": 2.5264652348016017e-06, + "loss": 1.6254, + "step": 8722 + }, + { + "epoch": 1.59, + "learning_rate": 2.5264106297779396e-06, + "loss": 0.8011, + "step": 8723 + }, + { + "epoch": 1.59, + "learning_rate": 2.5263560247542775e-06, + "loss": 1.5372, + "step": 8724 + }, + { + "epoch": 1.59, + "learning_rate": 2.5263014197306155e-06, + "loss": 1.2565, + "step": 8725 + }, + { + "epoch": 1.59, + "learning_rate": 2.526246814706953e-06, + "loss": 1.2015, + "step": 8726 + }, + { + "epoch": 1.59, + "learning_rate": 2.526192209683291e-06, + "loss": 1.5309, + "step": 8727 + }, + { + "epoch": 1.59, + "learning_rate": 2.526137604659629e-06, + "loss": 0.4081, + "step": 8728 + }, + { + "epoch": 1.59, + "learning_rate": 2.5260829996359664e-06, + "loss": 1.4784, + "step": 8729 + }, + { + "epoch": 1.59, + "learning_rate": 2.5260283946123044e-06, + "loss": 2.347, + "step": 8730 + }, + { + "epoch": 1.59, + "learning_rate": 2.5259737895886423e-06, + "loss": 1.5522, + "step": 8731 + }, + { + "epoch": 1.59, + "learning_rate": 2.52591918456498e-06, + "loss": 0.4647, + "step": 8732 + }, + { + "epoch": 1.59, + "learning_rate": 2.5258645795413178e-06, + "loss": 0.5189, + "step": 8733 + }, + { + "epoch": 1.59, + "learning_rate": 2.5258099745176557e-06, + "loss": 1.04, + "step": 8734 + }, + { + "epoch": 1.59, + "learning_rate": 2.5257553694939932e-06, + "loss": 0.3479, + "step": 8735 + }, + { + "epoch": 1.59, + "learning_rate": 2.525700764470331e-06, + "loss": 0.0258, + "step": 8736 + }, + { + "epoch": 1.59, + "learning_rate": 2.525646159446669e-06, + "loss": 0.4911, + "step": 8737 + }, + { + "epoch": 1.59, + "learning_rate": 2.525591554423007e-06, + "loss": 1.4836, + "step": 8738 + }, + { + "epoch": 1.59, + "learning_rate": 2.525536949399345e-06, + "loss": 1.2143, + "step": 8739 + }, + { + "epoch": 1.59, + "learning_rate": 2.5254823443756825e-06, + "loss": 2.0202, + "step": 8740 + }, + { + "epoch": 1.59, + "learning_rate": 2.5254277393520205e-06, + "loss": 0.2864, + "step": 8741 + }, + { + "epoch": 1.59, + "learning_rate": 2.5253731343283584e-06, + "loss": 0.7024, + "step": 8742 + }, + { + "epoch": 1.59, + "learning_rate": 2.525318529304696e-06, + "loss": 0.8553, + "step": 8743 + }, + { + "epoch": 1.59, + "learning_rate": 2.525263924281034e-06, + "loss": 1.1516, + "step": 8744 + }, + { + "epoch": 1.59, + "learning_rate": 2.525209319257372e-06, + "loss": 0.1814, + "step": 8745 + }, + { + "epoch": 1.59, + "learning_rate": 2.5251547142337093e-06, + "loss": 2.4522, + "step": 8746 + }, + { + "epoch": 1.59, + "learning_rate": 2.5251001092100473e-06, + "loss": 0.9453, + "step": 8747 + }, + { + "epoch": 1.59, + "learning_rate": 2.5250455041863852e-06, + "loss": 1.4394, + "step": 8748 + }, + { + "epoch": 1.59, + "learning_rate": 2.524990899162723e-06, + "loss": 0.8326, + "step": 8749 + }, + { + "epoch": 1.59, + "learning_rate": 2.524936294139061e-06, + "loss": 2.3065, + "step": 8750 + }, + { + "epoch": 1.59, + "learning_rate": 2.5248816891153986e-06, + "loss": 0.0557, + "step": 8751 + }, + { + "epoch": 1.59, + "learning_rate": 2.5248270840917366e-06, + "loss": 0.668, + "step": 8752 + }, + { + "epoch": 1.59, + "learning_rate": 2.5247724790680745e-06, + "loss": 1.1783, + "step": 8753 + }, + { + "epoch": 1.59, + "learning_rate": 2.524717874044412e-06, + "loss": 0.9096, + "step": 8754 + }, + { + "epoch": 1.59, + "learning_rate": 2.52466326902075e-06, + "loss": 0.5669, + "step": 8755 + }, + { + "epoch": 1.59, + "learning_rate": 2.524608663997088e-06, + "loss": 1.7285, + "step": 8756 + }, + { + "epoch": 1.59, + "learning_rate": 2.5245540589734254e-06, + "loss": 1.5804, + "step": 8757 + }, + { + "epoch": 1.59, + "learning_rate": 2.5244994539497634e-06, + "loss": 0.2839, + "step": 8758 + }, + { + "epoch": 1.59, + "learning_rate": 2.5244448489261013e-06, + "loss": 0.1871, + "step": 8759 + }, + { + "epoch": 1.59, + "learning_rate": 2.524390243902439e-06, + "loss": 0.4995, + "step": 8760 + }, + { + "epoch": 1.59, + "learning_rate": 2.524335638878777e-06, + "loss": 1.1817, + "step": 8761 + }, + { + "epoch": 1.59, + "learning_rate": 2.5242810338551147e-06, + "loss": 0.2099, + "step": 8762 + }, + { + "epoch": 1.59, + "learning_rate": 2.5242264288314527e-06, + "loss": 0.5629, + "step": 8763 + }, + { + "epoch": 1.59, + "learning_rate": 2.5241718238077906e-06, + "loss": 2.239, + "step": 8764 + }, + { + "epoch": 1.59, + "learning_rate": 2.524117218784128e-06, + "loss": 0.3157, + "step": 8765 + }, + { + "epoch": 1.59, + "learning_rate": 2.524062613760466e-06, + "loss": 0.7775, + "step": 8766 + }, + { + "epoch": 1.59, + "learning_rate": 2.524008008736804e-06, + "loss": 0.0307, + "step": 8767 + }, + { + "epoch": 1.59, + "learning_rate": 2.5239534037131416e-06, + "loss": 1.5099, + "step": 8768 + }, + { + "epoch": 1.59, + "learning_rate": 2.5238987986894795e-06, + "loss": 0.0251, + "step": 8769 + }, + { + "epoch": 1.59, + "learning_rate": 2.5238441936658174e-06, + "loss": 0.7514, + "step": 8770 + }, + { + "epoch": 1.6, + "learning_rate": 2.523789588642155e-06, + "loss": 0.404, + "step": 8771 + }, + { + "epoch": 1.6, + "learning_rate": 2.523734983618493e-06, + "loss": 0.0968, + "step": 8772 + }, + { + "epoch": 1.6, + "learning_rate": 2.523680378594831e-06, + "loss": 0.2938, + "step": 8773 + }, + { + "epoch": 1.6, + "learning_rate": 2.5236257735711684e-06, + "loss": 0.2573, + "step": 8774 + }, + { + "epoch": 1.6, + "learning_rate": 2.5235711685475063e-06, + "loss": 2.5178, + "step": 8775 + }, + { + "epoch": 1.6, + "learning_rate": 2.5235165635238443e-06, + "loss": 0.8058, + "step": 8776 + }, + { + "epoch": 1.6, + "learning_rate": 2.523461958500182e-06, + "loss": 0.7081, + "step": 8777 + }, + { + "epoch": 1.6, + "learning_rate": 2.52340735347652e-06, + "loss": 0.9303, + "step": 8778 + }, + { + "epoch": 1.6, + "learning_rate": 2.5233527484528577e-06, + "loss": 1.9558, + "step": 8779 + }, + { + "epoch": 1.6, + "learning_rate": 2.5232981434291956e-06, + "loss": 0.0775, + "step": 8780 + }, + { + "epoch": 1.6, + "learning_rate": 2.5232435384055335e-06, + "loss": 0.9502, + "step": 8781 + }, + { + "epoch": 1.6, + "learning_rate": 2.523188933381871e-06, + "loss": 0.9591, + "step": 8782 + }, + { + "epoch": 1.6, + "learning_rate": 2.523134328358209e-06, + "loss": 0.474, + "step": 8783 + }, + { + "epoch": 1.6, + "learning_rate": 2.523079723334547e-06, + "loss": 0.6682, + "step": 8784 + }, + { + "epoch": 1.6, + "learning_rate": 2.5230251183108845e-06, + "loss": 0.4148, + "step": 8785 + }, + { + "epoch": 1.6, + "learning_rate": 2.5229705132872224e-06, + "loss": 0.6436, + "step": 8786 + }, + { + "epoch": 1.6, + "learning_rate": 2.5229159082635604e-06, + "loss": 0.4079, + "step": 8787 + }, + { + "epoch": 1.6, + "learning_rate": 2.522861303239898e-06, + "loss": 0.5952, + "step": 8788 + }, + { + "epoch": 1.6, + "learning_rate": 2.522806698216236e-06, + "loss": 0.9132, + "step": 8789 + }, + { + "epoch": 1.6, + "learning_rate": 2.5227520931925738e-06, + "loss": 0.0618, + "step": 8790 + }, + { + "epoch": 1.6, + "learning_rate": 2.5226974881689117e-06, + "loss": 0.677, + "step": 8791 + }, + { + "epoch": 1.6, + "learning_rate": 2.5226428831452497e-06, + "loss": 0.2519, + "step": 8792 + }, + { + "epoch": 1.6, + "learning_rate": 2.522588278121587e-06, + "loss": 0.8113, + "step": 8793 + }, + { + "epoch": 1.6, + "learning_rate": 2.522533673097925e-06, + "loss": 0.8338, + "step": 8794 + }, + { + "epoch": 1.6, + "learning_rate": 2.522479068074263e-06, + "loss": 0.0688, + "step": 8795 + }, + { + "epoch": 1.6, + "learning_rate": 2.5224244630506006e-06, + "loss": 0.2876, + "step": 8796 + }, + { + "epoch": 1.6, + "learning_rate": 2.5223698580269385e-06, + "loss": 0.8447, + "step": 8797 + }, + { + "epoch": 1.6, + "learning_rate": 2.5223152530032765e-06, + "loss": 1.731, + "step": 8798 + }, + { + "epoch": 1.6, + "learning_rate": 2.522260647979614e-06, + "loss": 0.2924, + "step": 8799 + }, + { + "epoch": 1.6, + "learning_rate": 2.522206042955952e-06, + "loss": 2.1357, + "step": 8800 + }, + { + "epoch": 1.6, + "learning_rate": 2.52215143793229e-06, + "loss": 0.0169, + "step": 8801 + }, + { + "epoch": 1.6, + "learning_rate": 2.5220968329086274e-06, + "loss": 0.398, + "step": 8802 + }, + { + "epoch": 1.6, + "learning_rate": 2.5220422278849653e-06, + "loss": 1.546, + "step": 8803 + }, + { + "epoch": 1.6, + "learning_rate": 2.5219876228613033e-06, + "loss": 1.1582, + "step": 8804 + }, + { + "epoch": 1.6, + "learning_rate": 2.5219330178376412e-06, + "loss": 0.6628, + "step": 8805 + }, + { + "epoch": 1.6, + "learning_rate": 2.521878412813979e-06, + "loss": 1.2242, + "step": 8806 + }, + { + "epoch": 1.6, + "learning_rate": 2.5218238077903167e-06, + "loss": 0.3639, + "step": 8807 + }, + { + "epoch": 1.6, + "learning_rate": 2.5217692027666546e-06, + "loss": 0.034, + "step": 8808 + }, + { + "epoch": 1.6, + "learning_rate": 2.5217145977429926e-06, + "loss": 1.6227, + "step": 8809 + }, + { + "epoch": 1.6, + "learning_rate": 2.52165999271933e-06, + "loss": 1.5969, + "step": 8810 + }, + { + "epoch": 1.6, + "learning_rate": 2.521605387695668e-06, + "loss": 0.8318, + "step": 8811 + }, + { + "epoch": 1.6, + "learning_rate": 2.521550782672006e-06, + "loss": 0.1606, + "step": 8812 + }, + { + "epoch": 1.6, + "learning_rate": 2.5214961776483435e-06, + "loss": 1.4901, + "step": 8813 + }, + { + "epoch": 1.6, + "learning_rate": 2.5214415726246814e-06, + "loss": 0.7748, + "step": 8814 + }, + { + "epoch": 1.6, + "learning_rate": 2.5213869676010194e-06, + "loss": 1.9171, + "step": 8815 + }, + { + "epoch": 1.6, + "learning_rate": 2.521332362577357e-06, + "loss": 0.1529, + "step": 8816 + }, + { + "epoch": 1.6, + "learning_rate": 2.521277757553695e-06, + "loss": 2.1626, + "step": 8817 + }, + { + "epoch": 1.6, + "learning_rate": 2.521223152530033e-06, + "loss": 0.5119, + "step": 8818 + }, + { + "epoch": 1.6, + "learning_rate": 2.5211685475063707e-06, + "loss": 0.4325, + "step": 8819 + }, + { + "epoch": 1.6, + "learning_rate": 2.5211139424827087e-06, + "loss": 1.3946, + "step": 8820 + }, + { + "epoch": 1.6, + "learning_rate": 2.521059337459046e-06, + "loss": 0.4332, + "step": 8821 + }, + { + "epoch": 1.6, + "learning_rate": 2.521004732435384e-06, + "loss": 1.1124, + "step": 8822 + }, + { + "epoch": 1.6, + "learning_rate": 2.520950127411722e-06, + "loss": 0.0658, + "step": 8823 + }, + { + "epoch": 1.6, + "learning_rate": 2.5208955223880596e-06, + "loss": 1.1768, + "step": 8824 + }, + { + "epoch": 1.6, + "learning_rate": 2.5208409173643975e-06, + "loss": 0.6656, + "step": 8825 + }, + { + "epoch": 1.61, + "learning_rate": 2.5207863123407355e-06, + "loss": 0.5402, + "step": 8826 + }, + { + "epoch": 1.61, + "learning_rate": 2.520731707317073e-06, + "loss": 0.8589, + "step": 8827 + }, + { + "epoch": 1.61, + "learning_rate": 2.520677102293411e-06, + "loss": 2.0617, + "step": 8828 + }, + { + "epoch": 1.61, + "learning_rate": 2.520622497269749e-06, + "loss": 2.9198, + "step": 8829 + }, + { + "epoch": 1.61, + "learning_rate": 2.520567892246087e-06, + "loss": 2.3137, + "step": 8830 + }, + { + "epoch": 1.61, + "learning_rate": 2.5205132872224248e-06, + "loss": 0.1494, + "step": 8831 + }, + { + "epoch": 1.61, + "learning_rate": 2.5204586821987623e-06, + "loss": 0.1927, + "step": 8832 + }, + { + "epoch": 1.61, + "learning_rate": 2.5204040771751002e-06, + "loss": 0.6266, + "step": 8833 + }, + { + "epoch": 1.61, + "learning_rate": 2.520349472151438e-06, + "loss": 0.7062, + "step": 8834 + }, + { + "epoch": 1.61, + "learning_rate": 2.5202948671277757e-06, + "loss": 0.3675, + "step": 8835 + }, + { + "epoch": 1.61, + "learning_rate": 2.5202402621041137e-06, + "loss": 1.2821, + "step": 8836 + }, + { + "epoch": 1.61, + "learning_rate": 2.5201856570804516e-06, + "loss": 1.3099, + "step": 8837 + }, + { + "epoch": 1.61, + "learning_rate": 2.520131052056789e-06, + "loss": 1.1175, + "step": 8838 + }, + { + "epoch": 1.61, + "learning_rate": 2.520076447033127e-06, + "loss": 0.3707, + "step": 8839 + }, + { + "epoch": 1.61, + "learning_rate": 2.520021842009465e-06, + "loss": 0.3067, + "step": 8840 + }, + { + "epoch": 1.61, + "learning_rate": 2.5199672369858025e-06, + "loss": 1.8078, + "step": 8841 + }, + { + "epoch": 1.61, + "learning_rate": 2.5199126319621405e-06, + "loss": 1.2674, + "step": 8842 + }, + { + "epoch": 1.61, + "learning_rate": 2.5198580269384784e-06, + "loss": 1.6853, + "step": 8843 + }, + { + "epoch": 1.61, + "learning_rate": 2.5198034219148164e-06, + "loss": 0.3249, + "step": 8844 + }, + { + "epoch": 1.61, + "learning_rate": 2.5197488168911543e-06, + "loss": 0.8567, + "step": 8845 + }, + { + "epoch": 1.61, + "learning_rate": 2.519694211867492e-06, + "loss": 0.9729, + "step": 8846 + }, + { + "epoch": 1.61, + "learning_rate": 2.5196396068438298e-06, + "loss": 0.7518, + "step": 8847 + }, + { + "epoch": 1.61, + "learning_rate": 2.5195850018201677e-06, + "loss": 3.1247, + "step": 8848 + }, + { + "epoch": 1.61, + "learning_rate": 2.5195303967965052e-06, + "loss": 1.3122, + "step": 8849 + }, + { + "epoch": 1.61, + "learning_rate": 2.519475791772843e-06, + "loss": 0.5822, + "step": 8850 + }, + { + "epoch": 1.61, + "learning_rate": 2.519421186749181e-06, + "loss": 1.6498, + "step": 8851 + }, + { + "epoch": 1.61, + "learning_rate": 2.5193665817255186e-06, + "loss": 0.4576, + "step": 8852 + }, + { + "epoch": 1.61, + "learning_rate": 2.5193119767018566e-06, + "loss": 0.0275, + "step": 8853 + }, + { + "epoch": 1.61, + "learning_rate": 2.5192573716781945e-06, + "loss": 0.5102, + "step": 8854 + }, + { + "epoch": 1.61, + "learning_rate": 2.519202766654532e-06, + "loss": 0.5111, + "step": 8855 + }, + { + "epoch": 1.61, + "learning_rate": 2.51914816163087e-06, + "loss": 0.6294, + "step": 8856 + }, + { + "epoch": 1.61, + "learning_rate": 2.519093556607208e-06, + "loss": 0.0952, + "step": 8857 + }, + { + "epoch": 1.61, + "learning_rate": 2.519038951583546e-06, + "loss": 0.8391, + "step": 8858 + }, + { + "epoch": 1.61, + "learning_rate": 2.518984346559884e-06, + "loss": 0.8924, + "step": 8859 + }, + { + "epoch": 1.61, + "learning_rate": 2.5189297415362213e-06, + "loss": 0.9844, + "step": 8860 + }, + { + "epoch": 1.61, + "learning_rate": 2.5188751365125593e-06, + "loss": 1.6166, + "step": 8861 + }, + { + "epoch": 1.61, + "learning_rate": 2.5188205314888972e-06, + "loss": 0.0969, + "step": 8862 + }, + { + "epoch": 1.61, + "learning_rate": 2.5187659264652347e-06, + "loss": 0.4954, + "step": 8863 + }, + { + "epoch": 1.61, + "learning_rate": 2.5187113214415727e-06, + "loss": 0.0512, + "step": 8864 + }, + { + "epoch": 1.61, + "learning_rate": 2.5186567164179106e-06, + "loss": 0.578, + "step": 8865 + }, + { + "epoch": 1.61, + "learning_rate": 2.518602111394248e-06, + "loss": 0.1533, + "step": 8866 + }, + { + "epoch": 1.61, + "learning_rate": 2.518547506370586e-06, + "loss": 0.5459, + "step": 8867 + }, + { + "epoch": 1.61, + "learning_rate": 2.518492901346924e-06, + "loss": 0.6868, + "step": 8868 + }, + { + "epoch": 1.61, + "learning_rate": 2.5184382963232616e-06, + "loss": 0.2216, + "step": 8869 + }, + { + "epoch": 1.61, + "learning_rate": 2.5183836912995995e-06, + "loss": 0.7789, + "step": 8870 + }, + { + "epoch": 1.61, + "learning_rate": 2.5183290862759374e-06, + "loss": 2.0802, + "step": 8871 + }, + { + "epoch": 1.61, + "learning_rate": 2.5182744812522754e-06, + "loss": 0.955, + "step": 8872 + }, + { + "epoch": 1.61, + "learning_rate": 2.5182198762286133e-06, + "loss": 3.0148, + "step": 8873 + }, + { + "epoch": 1.61, + "learning_rate": 2.518165271204951e-06, + "loss": 0.2846, + "step": 8874 + }, + { + "epoch": 1.61, + "learning_rate": 2.518110666181289e-06, + "loss": 1.2145, + "step": 8875 + }, + { + "epoch": 1.61, + "learning_rate": 2.5180560611576267e-06, + "loss": 1.8489, + "step": 8876 + }, + { + "epoch": 1.61, + "learning_rate": 2.5180014561339643e-06, + "loss": 0.3544, + "step": 8877 + }, + { + "epoch": 1.61, + "learning_rate": 2.517946851110302e-06, + "loss": 0.2851, + "step": 8878 + }, + { + "epoch": 1.61, + "learning_rate": 2.51789224608664e-06, + "loss": 0.9955, + "step": 8879 + }, + { + "epoch": 1.61, + "learning_rate": 2.5178376410629777e-06, + "loss": 0.9008, + "step": 8880 + }, + { + "epoch": 1.62, + "learning_rate": 2.5177830360393156e-06, + "loss": 3.0119, + "step": 8881 + }, + { + "epoch": 1.62, + "learning_rate": 2.5177284310156535e-06, + "loss": 1.1054, + "step": 8882 + }, + { + "epoch": 1.62, + "learning_rate": 2.517673825991991e-06, + "loss": 0.5727, + "step": 8883 + }, + { + "epoch": 1.62, + "learning_rate": 2.517619220968329e-06, + "loss": 0.4545, + "step": 8884 + }, + { + "epoch": 1.62, + "learning_rate": 2.517564615944667e-06, + "loss": 0.1851, + "step": 8885 + }, + { + "epoch": 1.62, + "learning_rate": 2.517510010921005e-06, + "loss": 1.1221, + "step": 8886 + }, + { + "epoch": 1.62, + "learning_rate": 2.517455405897343e-06, + "loss": 1.4287, + "step": 8887 + }, + { + "epoch": 1.62, + "learning_rate": 2.5174008008736804e-06, + "loss": 0.9229, + "step": 8888 + }, + { + "epoch": 1.62, + "learning_rate": 2.5173461958500183e-06, + "loss": 1.2276, + "step": 8889 + }, + { + "epoch": 1.62, + "learning_rate": 2.5172915908263562e-06, + "loss": 0.3129, + "step": 8890 + }, + { + "epoch": 1.62, + "learning_rate": 2.5172369858026938e-06, + "loss": 0.4909, + "step": 8891 + }, + { + "epoch": 1.62, + "learning_rate": 2.5171823807790317e-06, + "loss": 0.0463, + "step": 8892 + }, + { + "epoch": 1.62, + "learning_rate": 2.5171277757553697e-06, + "loss": 0.4763, + "step": 8893 + }, + { + "epoch": 1.62, + "learning_rate": 2.517073170731707e-06, + "loss": 0.7524, + "step": 8894 + }, + { + "epoch": 1.62, + "learning_rate": 2.517018565708045e-06, + "loss": 1.3112, + "step": 8895 + }, + { + "epoch": 1.62, + "learning_rate": 2.516963960684383e-06, + "loss": 1.1639, + "step": 8896 + }, + { + "epoch": 1.62, + "learning_rate": 2.5169093556607206e-06, + "loss": 0.5653, + "step": 8897 + }, + { + "epoch": 1.62, + "learning_rate": 2.5168547506370585e-06, + "loss": 2.0485, + "step": 8898 + }, + { + "epoch": 1.62, + "learning_rate": 2.5168001456133965e-06, + "loss": 0.741, + "step": 8899 + }, + { + "epoch": 1.62, + "learning_rate": 2.5167455405897344e-06, + "loss": 0.6354, + "step": 8900 + }, + { + "epoch": 1.62, + "learning_rate": 2.5166909355660724e-06, + "loss": 1.1431, + "step": 8901 + }, + { + "epoch": 1.62, + "learning_rate": 2.51663633054241e-06, + "loss": 0.0419, + "step": 8902 + }, + { + "epoch": 1.62, + "learning_rate": 2.516581725518748e-06, + "loss": 0.287, + "step": 8903 + }, + { + "epoch": 1.62, + "learning_rate": 2.5165271204950858e-06, + "loss": 0.3723, + "step": 8904 + }, + { + "epoch": 1.62, + "learning_rate": 2.5164725154714233e-06, + "loss": 1.2, + "step": 8905 + }, + { + "epoch": 1.62, + "learning_rate": 2.5164179104477612e-06, + "loss": 0.7718, + "step": 8906 + }, + { + "epoch": 1.62, + "learning_rate": 2.516363305424099e-06, + "loss": 0.9011, + "step": 8907 + }, + { + "epoch": 1.62, + "learning_rate": 2.5163087004004367e-06, + "loss": 1.9425, + "step": 8908 + }, + { + "epoch": 1.62, + "learning_rate": 2.5162540953767746e-06, + "loss": 0.5691, + "step": 8909 + }, + { + "epoch": 1.62, + "learning_rate": 2.5161994903531126e-06, + "loss": 0.0252, + "step": 8910 + }, + { + "epoch": 1.62, + "learning_rate": 2.5161448853294505e-06, + "loss": 0.5992, + "step": 8911 + }, + { + "epoch": 1.62, + "learning_rate": 2.5160902803057885e-06, + "loss": 0.0417, + "step": 8912 + }, + { + "epoch": 1.62, + "learning_rate": 2.516035675282126e-06, + "loss": 0.7753, + "step": 8913 + }, + { + "epoch": 1.62, + "learning_rate": 2.515981070258464e-06, + "loss": 0.302, + "step": 8914 + }, + { + "epoch": 1.62, + "learning_rate": 2.515926465234802e-06, + "loss": 1.0026, + "step": 8915 + }, + { + "epoch": 1.62, + "learning_rate": 2.5158718602111394e-06, + "loss": 0.4499, + "step": 8916 + }, + { + "epoch": 1.62, + "learning_rate": 2.5158172551874773e-06, + "loss": 0.4945, + "step": 8917 + }, + { + "epoch": 1.62, + "learning_rate": 2.5157626501638153e-06, + "loss": 0.5931, + "step": 8918 + }, + { + "epoch": 1.62, + "learning_rate": 2.515708045140153e-06, + "loss": 1.5679, + "step": 8919 + }, + { + "epoch": 1.62, + "learning_rate": 2.5156534401164907e-06, + "loss": 0.7492, + "step": 8920 + }, + { + "epoch": 1.62, + "learning_rate": 2.5155988350928287e-06, + "loss": 0.4356, + "step": 8921 + }, + { + "epoch": 1.62, + "learning_rate": 2.515544230069166e-06, + "loss": 1.0806, + "step": 8922 + }, + { + "epoch": 1.62, + "learning_rate": 2.515489625045504e-06, + "loss": 3.0398, + "step": 8923 + }, + { + "epoch": 1.62, + "learning_rate": 2.515435020021842e-06, + "loss": 0.1947, + "step": 8924 + }, + { + "epoch": 1.62, + "learning_rate": 2.51538041499818e-06, + "loss": 1.1472, + "step": 8925 + }, + { + "epoch": 1.62, + "learning_rate": 2.515325809974518e-06, + "loss": 2.4853, + "step": 8926 + }, + { + "epoch": 1.62, + "learning_rate": 2.5152712049508555e-06, + "loss": 0.1506, + "step": 8927 + }, + { + "epoch": 1.62, + "learning_rate": 2.5152165999271934e-06, + "loss": 0.7597, + "step": 8928 + }, + { + "epoch": 1.62, + "learning_rate": 2.5151619949035314e-06, + "loss": 1.6685, + "step": 8929 + }, + { + "epoch": 1.62, + "learning_rate": 2.515107389879869e-06, + "loss": 0.9981, + "step": 8930 + }, + { + "epoch": 1.62, + "learning_rate": 2.515052784856207e-06, + "loss": 0.598, + "step": 8931 + }, + { + "epoch": 1.62, + "learning_rate": 2.5149981798325448e-06, + "loss": 0.2511, + "step": 8932 + }, + { + "epoch": 1.62, + "learning_rate": 2.5149435748088823e-06, + "loss": 0.1151, + "step": 8933 + }, + { + "epoch": 1.62, + "learning_rate": 2.5148889697852202e-06, + "loss": 3.1642, + "step": 8934 + }, + { + "epoch": 1.62, + "learning_rate": 2.514834364761558e-06, + "loss": 0.4816, + "step": 8935 + }, + { + "epoch": 1.63, + "learning_rate": 2.5147797597378957e-06, + "loss": 1.0426, + "step": 8936 + }, + { + "epoch": 1.63, + "learning_rate": 2.5147251547142337e-06, + "loss": 1.4736, + "step": 8937 + }, + { + "epoch": 1.63, + "learning_rate": 2.5146705496905716e-06, + "loss": 0.561, + "step": 8938 + }, + { + "epoch": 1.63, + "learning_rate": 2.5146159446669095e-06, + "loss": 1.7513, + "step": 8939 + }, + { + "epoch": 1.63, + "learning_rate": 2.5145613396432475e-06, + "loss": 0.6649, + "step": 8940 + }, + { + "epoch": 1.63, + "learning_rate": 2.514506734619585e-06, + "loss": 0.2427, + "step": 8941 + }, + { + "epoch": 1.63, + "learning_rate": 2.514452129595923e-06, + "loss": 0.0335, + "step": 8942 + }, + { + "epoch": 1.63, + "learning_rate": 2.514397524572261e-06, + "loss": 1.5857, + "step": 8943 + }, + { + "epoch": 1.63, + "learning_rate": 2.5143429195485984e-06, + "loss": 0.6024, + "step": 8944 + }, + { + "epoch": 1.63, + "learning_rate": 2.5142883145249364e-06, + "loss": 0.486, + "step": 8945 + }, + { + "epoch": 1.63, + "learning_rate": 2.5142337095012743e-06, + "loss": 0.5796, + "step": 8946 + }, + { + "epoch": 1.63, + "learning_rate": 2.514179104477612e-06, + "loss": 0.827, + "step": 8947 + }, + { + "epoch": 1.63, + "learning_rate": 2.5141244994539498e-06, + "loss": 1.8459, + "step": 8948 + }, + { + "epoch": 1.63, + "learning_rate": 2.5140698944302877e-06, + "loss": 0.437, + "step": 8949 + }, + { + "epoch": 1.63, + "learning_rate": 2.5140152894066252e-06, + "loss": 0.2459, + "step": 8950 + }, + { + "epoch": 1.63, + "learning_rate": 2.513960684382963e-06, + "loss": 0.0535, + "step": 8951 + }, + { + "epoch": 1.63, + "learning_rate": 2.513906079359301e-06, + "loss": 0.666, + "step": 8952 + }, + { + "epoch": 1.63, + "learning_rate": 2.513851474335639e-06, + "loss": 0.5111, + "step": 8953 + }, + { + "epoch": 1.63, + "learning_rate": 2.513796869311977e-06, + "loss": 0.9698, + "step": 8954 + }, + { + "epoch": 1.63, + "learning_rate": 2.5137422642883145e-06, + "loss": 0.0463, + "step": 8955 + }, + { + "epoch": 1.63, + "learning_rate": 2.5136876592646525e-06, + "loss": 1.1262, + "step": 8956 + }, + { + "epoch": 1.63, + "learning_rate": 2.5136330542409904e-06, + "loss": 1.3769, + "step": 8957 + }, + { + "epoch": 1.63, + "learning_rate": 2.513578449217328e-06, + "loss": 0.6046, + "step": 8958 + }, + { + "epoch": 1.63, + "learning_rate": 2.513523844193666e-06, + "loss": 0.0332, + "step": 8959 + }, + { + "epoch": 1.63, + "learning_rate": 2.513469239170004e-06, + "loss": 3.4562, + "step": 8960 + }, + { + "epoch": 1.63, + "learning_rate": 2.5134146341463413e-06, + "loss": 0.8929, + "step": 8961 + }, + { + "epoch": 1.63, + "learning_rate": 2.5133600291226793e-06, + "loss": 1.1285, + "step": 8962 + }, + { + "epoch": 1.63, + "learning_rate": 2.5133054240990172e-06, + "loss": 1.1584, + "step": 8963 + }, + { + "epoch": 1.63, + "learning_rate": 2.5132508190753547e-06, + "loss": 0.626, + "step": 8964 + }, + { + "epoch": 1.63, + "learning_rate": 2.5131962140516927e-06, + "loss": 2.0375, + "step": 8965 + }, + { + "epoch": 1.63, + "learning_rate": 2.5131416090280306e-06, + "loss": 2.4826, + "step": 8966 + }, + { + "epoch": 1.63, + "learning_rate": 2.5130870040043686e-06, + "loss": 0.5216, + "step": 8967 + }, + { + "epoch": 1.63, + "learning_rate": 2.5130323989807065e-06, + "loss": 0.3064, + "step": 8968 + }, + { + "epoch": 1.63, + "learning_rate": 2.512977793957044e-06, + "loss": 2.0222, + "step": 8969 + }, + { + "epoch": 1.63, + "learning_rate": 2.512923188933382e-06, + "loss": 1.3432, + "step": 8970 + }, + { + "epoch": 1.63, + "learning_rate": 2.51286858390972e-06, + "loss": 0.0739, + "step": 8971 + }, + { + "epoch": 1.63, + "learning_rate": 2.5128139788860574e-06, + "loss": 2.242, + "step": 8972 + }, + { + "epoch": 1.63, + "learning_rate": 2.5127593738623954e-06, + "loss": 0.7451, + "step": 8973 + }, + { + "epoch": 1.63, + "learning_rate": 2.5127047688387333e-06, + "loss": 0.4707, + "step": 8974 + }, + { + "epoch": 1.63, + "learning_rate": 2.512650163815071e-06, + "loss": 0.8943, + "step": 8975 + }, + { + "epoch": 1.63, + "learning_rate": 2.512595558791409e-06, + "loss": 1.869, + "step": 8976 + }, + { + "epoch": 1.63, + "learning_rate": 2.5125409537677467e-06, + "loss": 0.4935, + "step": 8977 + }, + { + "epoch": 1.63, + "learning_rate": 2.5124863487440843e-06, + "loss": 0.9417, + "step": 8978 + }, + { + "epoch": 1.63, + "learning_rate": 2.512431743720422e-06, + "loss": 0.1772, + "step": 8979 + }, + { + "epoch": 1.63, + "learning_rate": 2.51237713869676e-06, + "loss": 0.991, + "step": 8980 + }, + { + "epoch": 1.63, + "learning_rate": 2.512322533673098e-06, + "loss": 0.0346, + "step": 8981 + }, + { + "epoch": 1.63, + "learning_rate": 2.512267928649436e-06, + "loss": 1.3573, + "step": 8982 + }, + { + "epoch": 1.63, + "learning_rate": 2.5122133236257735e-06, + "loss": 0.246, + "step": 8983 + }, + { + "epoch": 1.63, + "learning_rate": 2.5121587186021115e-06, + "loss": 0.5494, + "step": 8984 + }, + { + "epoch": 1.63, + "learning_rate": 2.5121041135784494e-06, + "loss": 1.0836, + "step": 8985 + }, + { + "epoch": 1.63, + "learning_rate": 2.512049508554787e-06, + "loss": 0.7603, + "step": 8986 + }, + { + "epoch": 1.63, + "learning_rate": 2.511994903531125e-06, + "loss": 0.4738, + "step": 8987 + }, + { + "epoch": 1.63, + "learning_rate": 2.511940298507463e-06, + "loss": 0.3822, + "step": 8988 + }, + { + "epoch": 1.63, + "learning_rate": 2.5118856934838004e-06, + "loss": 0.658, + "step": 8989 + }, + { + "epoch": 1.63, + "learning_rate": 2.5118310884601383e-06, + "loss": 0.4653, + "step": 8990 + }, + { + "epoch": 1.64, + "learning_rate": 2.5117764834364762e-06, + "loss": 0.6227, + "step": 8991 + }, + { + "epoch": 1.64, + "learning_rate": 2.5117218784128138e-06, + "loss": 0.4653, + "step": 8992 + }, + { + "epoch": 1.64, + "learning_rate": 2.5116672733891517e-06, + "loss": 0.2116, + "step": 8993 + }, + { + "epoch": 1.64, + "learning_rate": 2.5116126683654897e-06, + "loss": 0.3842, + "step": 8994 + }, + { + "epoch": 1.64, + "learning_rate": 2.5115580633418276e-06, + "loss": 0.0433, + "step": 8995 + }, + { + "epoch": 1.64, + "learning_rate": 2.5115034583181655e-06, + "loss": 0.7946, + "step": 8996 + }, + { + "epoch": 1.64, + "learning_rate": 2.511448853294503e-06, + "loss": 0.6712, + "step": 8997 + }, + { + "epoch": 1.64, + "learning_rate": 2.511394248270841e-06, + "loss": 0.4968, + "step": 8998 + }, + { + "epoch": 1.64, + "learning_rate": 2.511339643247179e-06, + "loss": 0.1745, + "step": 8999 + }, + { + "epoch": 1.64, + "learning_rate": 2.5112850382235165e-06, + "loss": 0.1334, + "step": 9000 + }, + { + "epoch": 1.64, + "learning_rate": 2.5112304331998544e-06, + "loss": 0.2856, + "step": 9001 + }, + { + "epoch": 1.64, + "learning_rate": 2.5111758281761924e-06, + "loss": 0.491, + "step": 9002 + }, + { + "epoch": 1.64, + "learning_rate": 2.51112122315253e-06, + "loss": 0.4231, + "step": 9003 + }, + { + "epoch": 1.64, + "learning_rate": 2.511066618128868e-06, + "loss": 0.7254, + "step": 9004 + }, + { + "epoch": 1.64, + "learning_rate": 2.5110120131052058e-06, + "loss": 1.8144, + "step": 9005 + }, + { + "epoch": 1.64, + "learning_rate": 2.5109574080815437e-06, + "loss": 0.7354, + "step": 9006 + }, + { + "epoch": 1.64, + "learning_rate": 2.5109028030578816e-06, + "loss": 1.4714, + "step": 9007 + }, + { + "epoch": 1.64, + "learning_rate": 2.510848198034219e-06, + "loss": 0.1246, + "step": 9008 + }, + { + "epoch": 1.64, + "learning_rate": 2.510793593010557e-06, + "loss": 1.3588, + "step": 9009 + }, + { + "epoch": 1.64, + "learning_rate": 2.510738987986895e-06, + "loss": 0.0567, + "step": 9010 + }, + { + "epoch": 1.64, + "learning_rate": 2.5106843829632326e-06, + "loss": 0.1619, + "step": 9011 + }, + { + "epoch": 1.64, + "learning_rate": 2.5106297779395705e-06, + "loss": 0.6644, + "step": 9012 + }, + { + "epoch": 1.64, + "learning_rate": 2.5105751729159085e-06, + "loss": 0.3483, + "step": 9013 + }, + { + "epoch": 1.64, + "learning_rate": 2.510520567892246e-06, + "loss": 2.4144, + "step": 9014 + }, + { + "epoch": 1.64, + "learning_rate": 2.510465962868584e-06, + "loss": 1.7019, + "step": 9015 + }, + { + "epoch": 1.64, + "learning_rate": 2.510411357844922e-06, + "loss": 2.1627, + "step": 9016 + }, + { + "epoch": 1.64, + "learning_rate": 2.5103567528212594e-06, + "loss": 2.1823, + "step": 9017 + }, + { + "epoch": 1.64, + "learning_rate": 2.5103021477975973e-06, + "loss": 3.042, + "step": 9018 + }, + { + "epoch": 1.64, + "learning_rate": 2.5102475427739353e-06, + "loss": 0.995, + "step": 9019 + }, + { + "epoch": 1.64, + "learning_rate": 2.5101929377502732e-06, + "loss": 0.1497, + "step": 9020 + }, + { + "epoch": 1.64, + "learning_rate": 2.510138332726611e-06, + "loss": 0.4372, + "step": 9021 + }, + { + "epoch": 1.64, + "learning_rate": 2.5100837277029487e-06, + "loss": 0.6476, + "step": 9022 + }, + { + "epoch": 1.64, + "learning_rate": 2.5100291226792866e-06, + "loss": 1.8112, + "step": 9023 + }, + { + "epoch": 1.64, + "learning_rate": 2.5099745176556246e-06, + "loss": 0.673, + "step": 9024 + }, + { + "epoch": 1.64, + "learning_rate": 2.509919912631962e-06, + "loss": 1.1956, + "step": 9025 + }, + { + "epoch": 1.64, + "learning_rate": 2.5098653076083e-06, + "loss": 0.905, + "step": 9026 + }, + { + "epoch": 1.64, + "learning_rate": 2.509810702584638e-06, + "loss": 1.7397, + "step": 9027 + }, + { + "epoch": 1.64, + "learning_rate": 2.5097560975609755e-06, + "loss": 0.3404, + "step": 9028 + }, + { + "epoch": 1.64, + "learning_rate": 2.5097014925373134e-06, + "loss": 1.9376, + "step": 9029 + }, + { + "epoch": 1.64, + "learning_rate": 2.5096468875136514e-06, + "loss": 1.6051, + "step": 9030 + }, + { + "epoch": 1.64, + "learning_rate": 2.509592282489989e-06, + "loss": 2.3944, + "step": 9031 + }, + { + "epoch": 1.64, + "learning_rate": 2.509537677466327e-06, + "loss": 1.8739, + "step": 9032 + }, + { + "epoch": 1.64, + "learning_rate": 2.5094830724426648e-06, + "loss": 0.5046, + "step": 9033 + }, + { + "epoch": 1.64, + "learning_rate": 2.5094284674190027e-06, + "loss": 1.5458, + "step": 9034 + }, + { + "epoch": 1.64, + "learning_rate": 2.5093738623953407e-06, + "loss": 0.5775, + "step": 9035 + }, + { + "epoch": 1.64, + "learning_rate": 2.509319257371678e-06, + "loss": 1.0072, + "step": 9036 + }, + { + "epoch": 1.64, + "learning_rate": 2.509264652348016e-06, + "loss": 0.1331, + "step": 9037 + }, + { + "epoch": 1.64, + "learning_rate": 2.509210047324354e-06, + "loss": 1.5087, + "step": 9038 + }, + { + "epoch": 1.64, + "learning_rate": 2.5091554423006916e-06, + "loss": 2.3909, + "step": 9039 + }, + { + "epoch": 1.64, + "learning_rate": 2.5091008372770295e-06, + "loss": 0.3693, + "step": 9040 + }, + { + "epoch": 1.64, + "learning_rate": 2.5090462322533675e-06, + "loss": 0.7881, + "step": 9041 + }, + { + "epoch": 1.64, + "learning_rate": 2.508991627229705e-06, + "loss": 1.1005, + "step": 9042 + }, + { + "epoch": 1.64, + "learning_rate": 2.508937022206043e-06, + "loss": 0.7036, + "step": 9043 + }, + { + "epoch": 1.64, + "learning_rate": 2.508882417182381e-06, + "loss": 2.9927, + "step": 9044 + }, + { + "epoch": 1.64, + "learning_rate": 2.5088278121587184e-06, + "loss": 1.7611, + "step": 9045 + }, + { + "epoch": 1.65, + "learning_rate": 2.5087732071350564e-06, + "loss": 0.0679, + "step": 9046 + }, + { + "epoch": 1.65, + "learning_rate": 2.5087186021113943e-06, + "loss": 0.4664, + "step": 9047 + }, + { + "epoch": 1.65, + "learning_rate": 2.5086639970877322e-06, + "loss": 2.1285, + "step": 9048 + }, + { + "epoch": 1.65, + "learning_rate": 2.50860939206407e-06, + "loss": 0.1548, + "step": 9049 + }, + { + "epoch": 1.65, + "learning_rate": 2.5085547870404077e-06, + "loss": 0.3681, + "step": 9050 + }, + { + "epoch": 1.65, + "learning_rate": 2.5085001820167457e-06, + "loss": 0.5079, + "step": 9051 + }, + { + "epoch": 1.65, + "learning_rate": 2.5084455769930836e-06, + "loss": 0.5894, + "step": 9052 + }, + { + "epoch": 1.65, + "learning_rate": 2.508390971969421e-06, + "loss": 0.3846, + "step": 9053 + }, + { + "epoch": 1.65, + "learning_rate": 2.508336366945759e-06, + "loss": 0.3668, + "step": 9054 + }, + { + "epoch": 1.65, + "learning_rate": 2.508281761922097e-06, + "loss": 1.5289, + "step": 9055 + }, + { + "epoch": 1.65, + "learning_rate": 2.5082271568984345e-06, + "loss": 1.4282, + "step": 9056 + }, + { + "epoch": 1.65, + "learning_rate": 2.5081725518747725e-06, + "loss": 0.4026, + "step": 9057 + }, + { + "epoch": 1.65, + "learning_rate": 2.5081179468511104e-06, + "loss": 0.819, + "step": 9058 + }, + { + "epoch": 1.65, + "learning_rate": 2.508063341827448e-06, + "loss": 2.9916, + "step": 9059 + }, + { + "epoch": 1.65, + "learning_rate": 2.508008736803786e-06, + "loss": 1.8952, + "step": 9060 + }, + { + "epoch": 1.65, + "learning_rate": 2.507954131780124e-06, + "loss": 0.5142, + "step": 9061 + }, + { + "epoch": 1.65, + "learning_rate": 2.5078995267564618e-06, + "loss": 1.1409, + "step": 9062 + }, + { + "epoch": 1.65, + "learning_rate": 2.5078449217327997e-06, + "loss": 2.8622, + "step": 9063 + }, + { + "epoch": 1.65, + "learning_rate": 2.5077903167091372e-06, + "loss": 1.5033, + "step": 9064 + }, + { + "epoch": 1.65, + "learning_rate": 2.507735711685475e-06, + "loss": 0.5568, + "step": 9065 + }, + { + "epoch": 1.65, + "learning_rate": 2.507681106661813e-06, + "loss": 1.9227, + "step": 9066 + }, + { + "epoch": 1.65, + "learning_rate": 2.5076265016381506e-06, + "loss": 0.519, + "step": 9067 + }, + { + "epoch": 1.65, + "learning_rate": 2.5075718966144886e-06, + "loss": 0.7603, + "step": 9068 + }, + { + "epoch": 1.65, + "learning_rate": 2.5075172915908265e-06, + "loss": 1.1756, + "step": 9069 + }, + { + "epoch": 1.65, + "learning_rate": 2.507462686567164e-06, + "loss": 1.548, + "step": 9070 + }, + { + "epoch": 1.65, + "learning_rate": 2.507408081543502e-06, + "loss": 1.3809, + "step": 9071 + }, + { + "epoch": 1.65, + "learning_rate": 2.50735347651984e-06, + "loss": 1.0147, + "step": 9072 + }, + { + "epoch": 1.65, + "learning_rate": 2.5072988714961774e-06, + "loss": 0.0562, + "step": 9073 + }, + { + "epoch": 1.65, + "learning_rate": 2.5072442664725154e-06, + "loss": 0.2718, + "step": 9074 + }, + { + "epoch": 1.65, + "learning_rate": 2.5071896614488533e-06, + "loss": 0.4472, + "step": 9075 + }, + { + "epoch": 1.65, + "learning_rate": 2.5071350564251913e-06, + "loss": 0.2623, + "step": 9076 + }, + { + "epoch": 1.65, + "learning_rate": 2.5070804514015292e-06, + "loss": 0.1317, + "step": 9077 + }, + { + "epoch": 1.65, + "learning_rate": 2.5070258463778667e-06, + "loss": 1.1227, + "step": 9078 + }, + { + "epoch": 1.65, + "learning_rate": 2.5069712413542047e-06, + "loss": 1.3498, + "step": 9079 + }, + { + "epoch": 1.65, + "learning_rate": 2.5069166363305426e-06, + "loss": 3.7006, + "step": 9080 + }, + { + "epoch": 1.65, + "learning_rate": 2.50686203130688e-06, + "loss": 1.9586, + "step": 9081 + }, + { + "epoch": 1.65, + "learning_rate": 2.506807426283218e-06, + "loss": 0.7657, + "step": 9082 + }, + { + "epoch": 1.65, + "learning_rate": 2.506752821259556e-06, + "loss": 0.0537, + "step": 9083 + }, + { + "epoch": 1.65, + "learning_rate": 2.5066982162358935e-06, + "loss": 0.2507, + "step": 9084 + }, + { + "epoch": 1.65, + "learning_rate": 2.5066436112122315e-06, + "loss": 0.4845, + "step": 9085 + }, + { + "epoch": 1.65, + "learning_rate": 2.5065890061885694e-06, + "loss": 0.367, + "step": 9086 + }, + { + "epoch": 1.65, + "learning_rate": 2.5065344011649074e-06, + "loss": 0.7431, + "step": 9087 + }, + { + "epoch": 1.65, + "learning_rate": 2.5064797961412453e-06, + "loss": 1.476, + "step": 9088 + }, + { + "epoch": 1.65, + "learning_rate": 2.506425191117583e-06, + "loss": 0.8818, + "step": 9089 + }, + { + "epoch": 1.65, + "learning_rate": 2.5063705860939208e-06, + "loss": 1.429, + "step": 9090 + }, + { + "epoch": 1.65, + "learning_rate": 2.5063159810702587e-06, + "loss": 0.6414, + "step": 9091 + }, + { + "epoch": 1.65, + "learning_rate": 2.5062613760465962e-06, + "loss": 1.526, + "step": 9092 + }, + { + "epoch": 1.65, + "learning_rate": 2.506206771022934e-06, + "loss": 0.617, + "step": 9093 + }, + { + "epoch": 1.65, + "learning_rate": 2.506152165999272e-06, + "loss": 1.7952, + "step": 9094 + }, + { + "epoch": 1.65, + "learning_rate": 2.5060975609756097e-06, + "loss": 0.4985, + "step": 9095 + }, + { + "epoch": 1.65, + "learning_rate": 2.5060429559519476e-06, + "loss": 0.319, + "step": 9096 + }, + { + "epoch": 1.65, + "learning_rate": 2.5059883509282855e-06, + "loss": 0.9799, + "step": 9097 + }, + { + "epoch": 1.65, + "learning_rate": 2.505933745904623e-06, + "loss": 0.3071, + "step": 9098 + }, + { + "epoch": 1.65, + "learning_rate": 2.505879140880961e-06, + "loss": 0.1177, + "step": 9099 + }, + { + "epoch": 1.65, + "learning_rate": 2.505824535857299e-06, + "loss": 0.3161, + "step": 9100 + }, + { + "epoch": 1.66, + "learning_rate": 2.505769930833637e-06, + "loss": 2.0857, + "step": 9101 + }, + { + "epoch": 1.66, + "learning_rate": 2.505715325809975e-06, + "loss": 1.7045, + "step": 9102 + }, + { + "epoch": 1.66, + "learning_rate": 2.5056607207863124e-06, + "loss": 0.1518, + "step": 9103 + }, + { + "epoch": 1.66, + "learning_rate": 2.5056061157626503e-06, + "loss": 0.4677, + "step": 9104 + }, + { + "epoch": 1.66, + "learning_rate": 2.5055515107389882e-06, + "loss": 1.0998, + "step": 9105 + }, + { + "epoch": 1.66, + "learning_rate": 2.5054969057153258e-06, + "loss": 0.6301, + "step": 9106 + }, + { + "epoch": 1.66, + "learning_rate": 2.5054423006916637e-06, + "loss": 3.239, + "step": 9107 + }, + { + "epoch": 1.66, + "learning_rate": 2.5053876956680016e-06, + "loss": 0.3156, + "step": 9108 + }, + { + "epoch": 1.66, + "learning_rate": 2.505333090644339e-06, + "loss": 1.8311, + "step": 9109 + }, + { + "epoch": 1.66, + "learning_rate": 2.505278485620677e-06, + "loss": 0.3522, + "step": 9110 + }, + { + "epoch": 1.66, + "learning_rate": 2.505223880597015e-06, + "loss": 0.4623, + "step": 9111 + }, + { + "epoch": 1.66, + "learning_rate": 2.5051692755733526e-06, + "loss": 1.2687, + "step": 9112 + }, + { + "epoch": 1.66, + "learning_rate": 2.5051146705496905e-06, + "loss": 1.005, + "step": 9113 + }, + { + "epoch": 1.66, + "learning_rate": 2.5050600655260285e-06, + "loss": 0.7512, + "step": 9114 + }, + { + "epoch": 1.66, + "learning_rate": 2.5050054605023664e-06, + "loss": 1.5365, + "step": 9115 + }, + { + "epoch": 1.66, + "learning_rate": 2.5049508554787043e-06, + "loss": 2.3074, + "step": 9116 + }, + { + "epoch": 1.66, + "learning_rate": 2.504896250455042e-06, + "loss": 1.1687, + "step": 9117 + }, + { + "epoch": 1.66, + "learning_rate": 2.50484164543138e-06, + "loss": 1.556, + "step": 9118 + }, + { + "epoch": 1.66, + "learning_rate": 2.5047870404077178e-06, + "loss": 0.4054, + "step": 9119 + }, + { + "epoch": 1.66, + "learning_rate": 2.5047324353840553e-06, + "loss": 0.9779, + "step": 9120 + }, + { + "epoch": 1.66, + "learning_rate": 2.5046778303603932e-06, + "loss": 0.8174, + "step": 9121 + }, + { + "epoch": 1.66, + "learning_rate": 2.504623225336731e-06, + "loss": 1.4312, + "step": 9122 + }, + { + "epoch": 1.66, + "learning_rate": 2.5045686203130687e-06, + "loss": 0.6898, + "step": 9123 + }, + { + "epoch": 1.66, + "learning_rate": 2.5045140152894066e-06, + "loss": 2.0537, + "step": 9124 + }, + { + "epoch": 1.66, + "learning_rate": 2.5044594102657446e-06, + "loss": 0.3133, + "step": 9125 + }, + { + "epoch": 1.66, + "learning_rate": 2.504404805242082e-06, + "loss": 1.6716, + "step": 9126 + }, + { + "epoch": 1.66, + "learning_rate": 2.50435020021842e-06, + "loss": 1.2699, + "step": 9127 + }, + { + "epoch": 1.66, + "learning_rate": 2.504295595194758e-06, + "loss": 0.699, + "step": 9128 + }, + { + "epoch": 1.66, + "learning_rate": 2.504240990171096e-06, + "loss": 1.1232, + "step": 9129 + }, + { + "epoch": 1.66, + "learning_rate": 2.504186385147434e-06, + "loss": 0.9901, + "step": 9130 + }, + { + "epoch": 1.66, + "learning_rate": 2.5041317801237714e-06, + "loss": 1.2661, + "step": 9131 + }, + { + "epoch": 1.66, + "learning_rate": 2.5040771751001093e-06, + "loss": 0.0729, + "step": 9132 + }, + { + "epoch": 1.66, + "learning_rate": 2.5040225700764473e-06, + "loss": 0.3003, + "step": 9133 + }, + { + "epoch": 1.66, + "learning_rate": 2.5039679650527848e-06, + "loss": 0.5309, + "step": 9134 + }, + { + "epoch": 1.66, + "learning_rate": 2.5039133600291227e-06, + "loss": 0.4715, + "step": 9135 + }, + { + "epoch": 1.66, + "learning_rate": 2.5038587550054607e-06, + "loss": 1.4554, + "step": 9136 + }, + { + "epoch": 1.66, + "learning_rate": 2.503804149981798e-06, + "loss": 0.7849, + "step": 9137 + }, + { + "epoch": 1.66, + "learning_rate": 2.503749544958136e-06, + "loss": 1.6204, + "step": 9138 + }, + { + "epoch": 1.66, + "learning_rate": 2.503694939934474e-06, + "loss": 0.2755, + "step": 9139 + }, + { + "epoch": 1.66, + "learning_rate": 2.5036403349108116e-06, + "loss": 0.6495, + "step": 9140 + }, + { + "epoch": 1.66, + "learning_rate": 2.5035857298871495e-06, + "loss": 0.5883, + "step": 9141 + }, + { + "epoch": 1.66, + "learning_rate": 2.5035311248634875e-06, + "loss": 0.449, + "step": 9142 + }, + { + "epoch": 1.66, + "learning_rate": 2.5034765198398254e-06, + "loss": 2.6298, + "step": 9143 + }, + { + "epoch": 1.66, + "learning_rate": 2.5034219148161634e-06, + "loss": 0.5982, + "step": 9144 + }, + { + "epoch": 1.66, + "learning_rate": 2.503367309792501e-06, + "loss": 1.3923, + "step": 9145 + }, + { + "epoch": 1.66, + "learning_rate": 2.503312704768839e-06, + "loss": 0.7496, + "step": 9146 + }, + { + "epoch": 1.66, + "learning_rate": 2.5032580997451768e-06, + "loss": 1.0722, + "step": 9147 + }, + { + "epoch": 1.66, + "learning_rate": 2.5032034947215143e-06, + "loss": 1.2193, + "step": 9148 + }, + { + "epoch": 1.66, + "learning_rate": 2.5031488896978522e-06, + "loss": 2.1128, + "step": 9149 + }, + { + "epoch": 1.66, + "learning_rate": 2.50309428467419e-06, + "loss": 0.0386, + "step": 9150 + }, + { + "epoch": 1.66, + "learning_rate": 2.5030396796505277e-06, + "loss": 0.2403, + "step": 9151 + }, + { + "epoch": 1.66, + "learning_rate": 2.5029850746268657e-06, + "loss": 0.0526, + "step": 9152 + }, + { + "epoch": 1.66, + "learning_rate": 2.5029304696032036e-06, + "loss": 0.2705, + "step": 9153 + }, + { + "epoch": 1.66, + "learning_rate": 2.502875864579541e-06, + "loss": 1.0988, + "step": 9154 + }, + { + "epoch": 1.66, + "learning_rate": 2.502821259555879e-06, + "loss": 0.0386, + "step": 9155 + }, + { + "epoch": 1.67, + "learning_rate": 2.502766654532217e-06, + "loss": 1.0603, + "step": 9156 + }, + { + "epoch": 1.67, + "learning_rate": 2.502712049508555e-06, + "loss": 0.8293, + "step": 9157 + }, + { + "epoch": 1.67, + "learning_rate": 2.502657444484893e-06, + "loss": 0.0789, + "step": 9158 + }, + { + "epoch": 1.67, + "learning_rate": 2.5026028394612304e-06, + "loss": 0.504, + "step": 9159 + }, + { + "epoch": 1.67, + "learning_rate": 2.5025482344375684e-06, + "loss": 0.9386, + "step": 9160 + }, + { + "epoch": 1.67, + "learning_rate": 2.5024936294139063e-06, + "loss": 0.5123, + "step": 9161 + }, + { + "epoch": 1.67, + "learning_rate": 2.502439024390244e-06, + "loss": 0.5654, + "step": 9162 + }, + { + "epoch": 1.67, + "learning_rate": 2.5023844193665818e-06, + "loss": 1.5333, + "step": 9163 + }, + { + "epoch": 1.67, + "learning_rate": 2.5023298143429197e-06, + "loss": 2.3548, + "step": 9164 + }, + { + "epoch": 1.67, + "learning_rate": 2.5022752093192572e-06, + "loss": 0.0334, + "step": 9165 + }, + { + "epoch": 1.67, + "learning_rate": 2.502220604295595e-06, + "loss": 0.7277, + "step": 9166 + }, + { + "epoch": 1.67, + "learning_rate": 2.502165999271933e-06, + "loss": 0.1375, + "step": 9167 + }, + { + "epoch": 1.67, + "learning_rate": 2.502111394248271e-06, + "loss": 0.2769, + "step": 9168 + }, + { + "epoch": 1.67, + "learning_rate": 2.502056789224609e-06, + "loss": 0.0761, + "step": 9169 + }, + { + "epoch": 1.67, + "learning_rate": 2.5020021842009465e-06, + "loss": 0.1352, + "step": 9170 + }, + { + "epoch": 1.67, + "learning_rate": 2.5019475791772845e-06, + "loss": 0.6123, + "step": 9171 + }, + { + "epoch": 1.67, + "learning_rate": 2.5018929741536224e-06, + "loss": 2.0961, + "step": 9172 + }, + { + "epoch": 1.67, + "learning_rate": 2.50183836912996e-06, + "loss": 1.3993, + "step": 9173 + }, + { + "epoch": 1.67, + "learning_rate": 2.501783764106298e-06, + "loss": 0.0988, + "step": 9174 + }, + { + "epoch": 1.67, + "learning_rate": 2.501729159082636e-06, + "loss": 0.1149, + "step": 9175 + }, + { + "epoch": 1.67, + "learning_rate": 2.5016745540589733e-06, + "loss": 0.6868, + "step": 9176 + }, + { + "epoch": 1.67, + "learning_rate": 2.5016199490353113e-06, + "loss": 0.4955, + "step": 9177 + }, + { + "epoch": 1.67, + "learning_rate": 2.5015653440116492e-06, + "loss": 1.3932, + "step": 9178 + }, + { + "epoch": 1.67, + "learning_rate": 2.5015107389879867e-06, + "loss": 0.5217, + "step": 9179 + }, + { + "epoch": 1.67, + "learning_rate": 2.5014561339643247e-06, + "loss": 0.8555, + "step": 9180 + }, + { + "epoch": 1.67, + "learning_rate": 2.5014015289406626e-06, + "loss": 2.8035, + "step": 9181 + }, + { + "epoch": 1.67, + "learning_rate": 2.5013469239170006e-06, + "loss": 0.4804, + "step": 9182 + }, + { + "epoch": 1.67, + "learning_rate": 2.5012923188933385e-06, + "loss": 0.0899, + "step": 9183 + }, + { + "epoch": 1.67, + "learning_rate": 2.501237713869676e-06, + "loss": 0.1784, + "step": 9184 + }, + { + "epoch": 1.67, + "learning_rate": 2.501183108846014e-06, + "loss": 1.1803, + "step": 9185 + }, + { + "epoch": 1.67, + "learning_rate": 2.501128503822352e-06, + "loss": 1.7999, + "step": 9186 + }, + { + "epoch": 1.67, + "learning_rate": 2.5010738987986894e-06, + "loss": 0.3758, + "step": 9187 + }, + { + "epoch": 1.67, + "learning_rate": 2.5010192937750274e-06, + "loss": 1.3121, + "step": 9188 + }, + { + "epoch": 1.67, + "learning_rate": 2.5009646887513653e-06, + "loss": 0.6246, + "step": 9189 + }, + { + "epoch": 1.67, + "learning_rate": 2.500910083727703e-06, + "loss": 0.6251, + "step": 9190 + }, + { + "epoch": 1.67, + "learning_rate": 2.5008554787040408e-06, + "loss": 0.0635, + "step": 9191 + }, + { + "epoch": 1.67, + "learning_rate": 2.5008008736803787e-06, + "loss": 0.3382, + "step": 9192 + }, + { + "epoch": 1.67, + "learning_rate": 2.5007462686567162e-06, + "loss": 0.8216, + "step": 9193 + }, + { + "epoch": 1.67, + "learning_rate": 2.500691663633054e-06, + "loss": 1.6559, + "step": 9194 + }, + { + "epoch": 1.67, + "learning_rate": 2.500637058609392e-06, + "loss": 2.5527, + "step": 9195 + }, + { + "epoch": 1.67, + "learning_rate": 2.50058245358573e-06, + "loss": 0.8735, + "step": 9196 + }, + { + "epoch": 1.67, + "learning_rate": 2.500527848562068e-06, + "loss": 2.2515, + "step": 9197 + }, + { + "epoch": 1.67, + "learning_rate": 2.5004732435384055e-06, + "loss": 1.2406, + "step": 9198 + }, + { + "epoch": 1.67, + "learning_rate": 2.5004186385147435e-06, + "loss": 0.0296, + "step": 9199 + }, + { + "epoch": 1.67, + "learning_rate": 2.5003640334910814e-06, + "loss": 0.3024, + "step": 9200 + }, + { + "epoch": 1.67, + "learning_rate": 2.500309428467419e-06, + "loss": 0.0942, + "step": 9201 + }, + { + "epoch": 1.67, + "learning_rate": 2.500254823443757e-06, + "loss": 2.6434, + "step": 9202 + }, + { + "epoch": 1.67, + "learning_rate": 2.500200218420095e-06, + "loss": 0.5214, + "step": 9203 + }, + { + "epoch": 1.67, + "learning_rate": 2.5001456133964324e-06, + "loss": 2.7473, + "step": 9204 + }, + { + "epoch": 1.67, + "learning_rate": 2.5000910083727703e-06, + "loss": 0.4591, + "step": 9205 + }, + { + "epoch": 1.67, + "learning_rate": 2.5000364033491082e-06, + "loss": 0.5673, + "step": 9206 + }, + { + "epoch": 1.67, + "learning_rate": 2.4999817983254458e-06, + "loss": 1.4083, + "step": 9207 + }, + { + "epoch": 1.67, + "learning_rate": 2.4999271933017837e-06, + "loss": 1.5436, + "step": 9208 + }, + { + "epoch": 1.67, + "learning_rate": 2.4998725882781216e-06, + "loss": 1.2037, + "step": 9209 + }, + { + "epoch": 1.67, + "learning_rate": 2.4998179832544596e-06, + "loss": 0.8183, + "step": 9210 + }, + { + "epoch": 1.68, + "learning_rate": 2.4997633782307975e-06, + "loss": 1.6646, + "step": 9211 + }, + { + "epoch": 1.68, + "learning_rate": 2.499708773207135e-06, + "loss": 0.2529, + "step": 9212 + }, + { + "epoch": 1.68, + "learning_rate": 2.499654168183473e-06, + "loss": 0.5016, + "step": 9213 + }, + { + "epoch": 1.68, + "learning_rate": 2.499599563159811e-06, + "loss": 1.4839, + "step": 9214 + }, + { + "epoch": 1.68, + "learning_rate": 2.4995449581361485e-06, + "loss": 1.3329, + "step": 9215 + }, + { + "epoch": 1.68, + "learning_rate": 2.4994903531124864e-06, + "loss": 0.7746, + "step": 9216 + }, + { + "epoch": 1.68, + "learning_rate": 2.4994357480888243e-06, + "loss": 0.0269, + "step": 9217 + }, + { + "epoch": 1.68, + "learning_rate": 2.499381143065162e-06, + "loss": 0.497, + "step": 9218 + }, + { + "epoch": 1.68, + "learning_rate": 2.4993265380415e-06, + "loss": 1.6227, + "step": 9219 + }, + { + "epoch": 1.68, + "learning_rate": 2.4992719330178378e-06, + "loss": 0.804, + "step": 9220 + }, + { + "epoch": 1.68, + "learning_rate": 2.4992173279941753e-06, + "loss": 0.4505, + "step": 9221 + }, + { + "epoch": 1.68, + "learning_rate": 2.4991627229705132e-06, + "loss": 0.3544, + "step": 9222 + }, + { + "epoch": 1.68, + "learning_rate": 2.499108117946851e-06, + "loss": 1.0626, + "step": 9223 + }, + { + "epoch": 1.68, + "learning_rate": 2.499053512923189e-06, + "loss": 0.239, + "step": 9224 + }, + { + "epoch": 1.68, + "learning_rate": 2.498998907899527e-06, + "loss": 3.4804, + "step": 9225 + }, + { + "epoch": 1.68, + "learning_rate": 2.4989443028758646e-06, + "loss": 0.0653, + "step": 9226 + }, + { + "epoch": 1.68, + "learning_rate": 2.4988896978522025e-06, + "loss": 0.2614, + "step": 9227 + }, + { + "epoch": 1.68, + "learning_rate": 2.4988350928285405e-06, + "loss": 1.2369, + "step": 9228 + }, + { + "epoch": 1.68, + "learning_rate": 2.498780487804878e-06, + "loss": 0.9274, + "step": 9229 + }, + { + "epoch": 1.68, + "learning_rate": 2.498725882781216e-06, + "loss": 0.9384, + "step": 9230 + }, + { + "epoch": 1.68, + "learning_rate": 2.498671277757554e-06, + "loss": 0.1839, + "step": 9231 + }, + { + "epoch": 1.68, + "learning_rate": 2.4986166727338914e-06, + "loss": 0.4888, + "step": 9232 + }, + { + "epoch": 1.68, + "learning_rate": 2.4985620677102293e-06, + "loss": 0.0489, + "step": 9233 + }, + { + "epoch": 1.68, + "learning_rate": 2.4985074626865673e-06, + "loss": 0.1156, + "step": 9234 + }, + { + "epoch": 1.68, + "learning_rate": 2.498452857662905e-06, + "loss": 0.2834, + "step": 9235 + }, + { + "epoch": 1.68, + "learning_rate": 2.4983982526392427e-06, + "loss": 1.3539, + "step": 9236 + }, + { + "epoch": 1.68, + "learning_rate": 2.4983436476155807e-06, + "loss": 2.1753, + "step": 9237 + }, + { + "epoch": 1.68, + "learning_rate": 2.4982890425919186e-06, + "loss": 0.2482, + "step": 9238 + }, + { + "epoch": 1.68, + "learning_rate": 2.4982344375682566e-06, + "loss": 0.7803, + "step": 9239 + }, + { + "epoch": 1.68, + "learning_rate": 2.498179832544594e-06, + "loss": 2.3224, + "step": 9240 + }, + { + "epoch": 1.68, + "learning_rate": 2.498125227520932e-06, + "loss": 0.4231, + "step": 9241 + }, + { + "epoch": 1.68, + "learning_rate": 2.49807062249727e-06, + "loss": 1.8562, + "step": 9242 + }, + { + "epoch": 1.68, + "learning_rate": 2.4980160174736075e-06, + "loss": 0.8448, + "step": 9243 + }, + { + "epoch": 1.68, + "learning_rate": 2.4979614124499454e-06, + "loss": 0.9016, + "step": 9244 + }, + { + "epoch": 1.68, + "learning_rate": 2.4979068074262834e-06, + "loss": 1.3469, + "step": 9245 + }, + { + "epoch": 1.68, + "learning_rate": 2.497852202402621e-06, + "loss": 0.7723, + "step": 9246 + }, + { + "epoch": 1.68, + "learning_rate": 2.497797597378959e-06, + "loss": 0.8289, + "step": 9247 + }, + { + "epoch": 1.68, + "learning_rate": 2.4977429923552968e-06, + "loss": 1.0579, + "step": 9248 + }, + { + "epoch": 1.68, + "learning_rate": 2.4976883873316347e-06, + "loss": 0.393, + "step": 9249 + }, + { + "epoch": 1.68, + "learning_rate": 2.4976337823079722e-06, + "loss": 0.0506, + "step": 9250 + }, + { + "epoch": 1.68, + "learning_rate": 2.49757917728431e-06, + "loss": 1.6856, + "step": 9251 + }, + { + "epoch": 1.68, + "learning_rate": 2.497524572260648e-06, + "loss": 1.1101, + "step": 9252 + }, + { + "epoch": 1.68, + "learning_rate": 2.497469967236986e-06, + "loss": 1.5869, + "step": 9253 + }, + { + "epoch": 1.68, + "learning_rate": 2.4974153622133236e-06, + "loss": 0.4635, + "step": 9254 + }, + { + "epoch": 1.68, + "learning_rate": 2.4973607571896615e-06, + "loss": 0.7696, + "step": 9255 + }, + { + "epoch": 1.68, + "learning_rate": 2.4973061521659995e-06, + "loss": 1.0502, + "step": 9256 + }, + { + "epoch": 1.68, + "learning_rate": 2.497251547142337e-06, + "loss": 3.1643, + "step": 9257 + }, + { + "epoch": 1.68, + "learning_rate": 2.497196942118675e-06, + "loss": 0.1026, + "step": 9258 + }, + { + "epoch": 1.68, + "learning_rate": 2.497142337095013e-06, + "loss": 2.5052, + "step": 9259 + }, + { + "epoch": 1.68, + "learning_rate": 2.4970877320713504e-06, + "loss": 0.9524, + "step": 9260 + }, + { + "epoch": 1.68, + "learning_rate": 2.4970331270476884e-06, + "loss": 0.3648, + "step": 9261 + }, + { + "epoch": 1.68, + "learning_rate": 2.4969785220240263e-06, + "loss": 1.2015, + "step": 9262 + }, + { + "epoch": 1.68, + "learning_rate": 2.4969239170003642e-06, + "loss": 0.089, + "step": 9263 + }, + { + "epoch": 1.68, + "learning_rate": 2.496869311976702e-06, + "loss": 0.6204, + "step": 9264 + }, + { + "epoch": 1.68, + "learning_rate": 2.4968147069530397e-06, + "loss": 1.3653, + "step": 9265 + }, + { + "epoch": 1.69, + "learning_rate": 2.4967601019293776e-06, + "loss": 0.8916, + "step": 9266 + }, + { + "epoch": 1.69, + "learning_rate": 2.4967054969057156e-06, + "loss": 0.7634, + "step": 9267 + }, + { + "epoch": 1.69, + "learning_rate": 2.496650891882053e-06, + "loss": 1.137, + "step": 9268 + }, + { + "epoch": 1.69, + "learning_rate": 2.496596286858391e-06, + "loss": 0.6027, + "step": 9269 + }, + { + "epoch": 1.69, + "learning_rate": 2.496541681834729e-06, + "loss": 1.2289, + "step": 9270 + }, + { + "epoch": 1.69, + "learning_rate": 2.4964870768110665e-06, + "loss": 0.0641, + "step": 9271 + }, + { + "epoch": 1.69, + "learning_rate": 2.4964324717874045e-06, + "loss": 0.6472, + "step": 9272 + }, + { + "epoch": 1.69, + "learning_rate": 2.4963778667637424e-06, + "loss": 0.6876, + "step": 9273 + }, + { + "epoch": 1.69, + "learning_rate": 2.49632326174008e-06, + "loss": 2.316, + "step": 9274 + }, + { + "epoch": 1.69, + "learning_rate": 2.496268656716418e-06, + "loss": 0.6747, + "step": 9275 + }, + { + "epoch": 1.69, + "learning_rate": 2.496214051692756e-06, + "loss": 1.0283, + "step": 9276 + }, + { + "epoch": 1.69, + "learning_rate": 2.4961594466690938e-06, + "loss": 0.2644, + "step": 9277 + }, + { + "epoch": 1.69, + "learning_rate": 2.4961048416454317e-06, + "loss": 1.2152, + "step": 9278 + }, + { + "epoch": 1.69, + "learning_rate": 2.4960502366217692e-06, + "loss": 0.0687, + "step": 9279 + }, + { + "epoch": 1.69, + "learning_rate": 2.495995631598107e-06, + "loss": 1.8565, + "step": 9280 + }, + { + "epoch": 1.69, + "learning_rate": 2.495941026574445e-06, + "loss": 0.9616, + "step": 9281 + }, + { + "epoch": 1.69, + "learning_rate": 2.4958864215507826e-06, + "loss": 1.4169, + "step": 9282 + }, + { + "epoch": 1.69, + "learning_rate": 2.4958318165271206e-06, + "loss": 0.9175, + "step": 9283 + }, + { + "epoch": 1.69, + "learning_rate": 2.4957772115034585e-06, + "loss": 0.5842, + "step": 9284 + }, + { + "epoch": 1.69, + "learning_rate": 2.495722606479796e-06, + "loss": 0.3368, + "step": 9285 + }, + { + "epoch": 1.69, + "learning_rate": 2.495668001456134e-06, + "loss": 0.4276, + "step": 9286 + }, + { + "epoch": 1.69, + "learning_rate": 2.495613396432472e-06, + "loss": 1.4561, + "step": 9287 + }, + { + "epoch": 1.69, + "learning_rate": 2.4955587914088094e-06, + "loss": 0.284, + "step": 9288 + }, + { + "epoch": 1.69, + "learning_rate": 2.4955041863851474e-06, + "loss": 1.0552, + "step": 9289 + }, + { + "epoch": 1.69, + "learning_rate": 2.4954495813614853e-06, + "loss": 0.0164, + "step": 9290 + }, + { + "epoch": 1.69, + "learning_rate": 2.4953949763378233e-06, + "loss": 0.4778, + "step": 9291 + }, + { + "epoch": 1.69, + "learning_rate": 2.495340371314161e-06, + "loss": 1.1924, + "step": 9292 + }, + { + "epoch": 1.69, + "learning_rate": 2.4952857662904987e-06, + "loss": 0.3902, + "step": 9293 + }, + { + "epoch": 1.69, + "learning_rate": 2.4952311612668367e-06, + "loss": 0.5713, + "step": 9294 + }, + { + "epoch": 1.69, + "learning_rate": 2.4951765562431746e-06, + "loss": 0.2703, + "step": 9295 + }, + { + "epoch": 1.69, + "learning_rate": 2.495121951219512e-06, + "loss": 4.1097, + "step": 9296 + }, + { + "epoch": 1.69, + "learning_rate": 2.49506734619585e-06, + "loss": 0.3409, + "step": 9297 + }, + { + "epoch": 1.69, + "learning_rate": 2.495012741172188e-06, + "loss": 0.7225, + "step": 9298 + }, + { + "epoch": 1.69, + "learning_rate": 2.4949581361485255e-06, + "loss": 0.3408, + "step": 9299 + }, + { + "epoch": 1.69, + "learning_rate": 2.4949035311248635e-06, + "loss": 1.2346, + "step": 9300 + }, + { + "epoch": 1.69, + "learning_rate": 2.4948489261012014e-06, + "loss": 0.9429, + "step": 9301 + }, + { + "epoch": 1.69, + "learning_rate": 2.494794321077539e-06, + "loss": 0.0499, + "step": 9302 + }, + { + "epoch": 1.69, + "learning_rate": 2.494739716053877e-06, + "loss": 1.5202, + "step": 9303 + }, + { + "epoch": 1.69, + "learning_rate": 2.494685111030215e-06, + "loss": 1.4423, + "step": 9304 + }, + { + "epoch": 1.69, + "learning_rate": 2.4946305060065528e-06, + "loss": 2.2243, + "step": 9305 + }, + { + "epoch": 1.69, + "learning_rate": 2.4945759009828907e-06, + "loss": 0.4818, + "step": 9306 + }, + { + "epoch": 1.69, + "learning_rate": 2.4945212959592282e-06, + "loss": 3.0568, + "step": 9307 + }, + { + "epoch": 1.69, + "learning_rate": 2.494466690935566e-06, + "loss": 0.5616, + "step": 9308 + }, + { + "epoch": 1.69, + "learning_rate": 2.494412085911904e-06, + "loss": 2.3414, + "step": 9309 + }, + { + "epoch": 1.69, + "learning_rate": 2.4943574808882416e-06, + "loss": 0.5074, + "step": 9310 + }, + { + "epoch": 1.69, + "learning_rate": 2.4943028758645796e-06, + "loss": 0.5717, + "step": 9311 + }, + { + "epoch": 1.69, + "learning_rate": 2.4942482708409175e-06, + "loss": 0.5545, + "step": 9312 + }, + { + "epoch": 1.69, + "learning_rate": 2.494193665817255e-06, + "loss": 2.1275, + "step": 9313 + }, + { + "epoch": 1.69, + "learning_rate": 2.494139060793593e-06, + "loss": 0.4828, + "step": 9314 + }, + { + "epoch": 1.69, + "learning_rate": 2.494084455769931e-06, + "loss": 0.5745, + "step": 9315 + }, + { + "epoch": 1.69, + "learning_rate": 2.4940298507462685e-06, + "loss": 0.5547, + "step": 9316 + }, + { + "epoch": 1.69, + "learning_rate": 2.4939752457226064e-06, + "loss": 1.1, + "step": 9317 + }, + { + "epoch": 1.69, + "learning_rate": 2.4939206406989444e-06, + "loss": 0.5538, + "step": 9318 + }, + { + "epoch": 1.69, + "learning_rate": 2.4938660356752823e-06, + "loss": 1.8361, + "step": 9319 + }, + { + "epoch": 1.69, + "learning_rate": 2.4938114306516202e-06, + "loss": 0.0874, + "step": 9320 + }, + { + "epoch": 1.7, + "learning_rate": 2.4937568256279578e-06, + "loss": 0.4688, + "step": 9321 + }, + { + "epoch": 1.7, + "learning_rate": 2.4937022206042957e-06, + "loss": 0.4038, + "step": 9322 + }, + { + "epoch": 1.7, + "learning_rate": 2.4936476155806336e-06, + "loss": 0.1352, + "step": 9323 + }, + { + "epoch": 1.7, + "learning_rate": 2.493593010556971e-06, + "loss": 0.4739, + "step": 9324 + }, + { + "epoch": 1.7, + "learning_rate": 2.493538405533309e-06, + "loss": 0.1774, + "step": 9325 + }, + { + "epoch": 1.7, + "learning_rate": 2.493483800509647e-06, + "loss": 1.3808, + "step": 9326 + }, + { + "epoch": 1.7, + "learning_rate": 2.4934291954859846e-06, + "loss": 0.4107, + "step": 9327 + }, + { + "epoch": 1.7, + "learning_rate": 2.4933745904623225e-06, + "loss": 0.4453, + "step": 9328 + }, + { + "epoch": 1.7, + "learning_rate": 2.4933199854386605e-06, + "loss": 0.2235, + "step": 9329 + }, + { + "epoch": 1.7, + "learning_rate": 2.493265380414998e-06, + "loss": 0.7543, + "step": 9330 + }, + { + "epoch": 1.7, + "learning_rate": 2.493210775391336e-06, + "loss": 3.2222, + "step": 9331 + }, + { + "epoch": 1.7, + "learning_rate": 2.493156170367674e-06, + "loss": 0.0613, + "step": 9332 + }, + { + "epoch": 1.7, + "learning_rate": 2.493101565344012e-06, + "loss": 2.1273, + "step": 9333 + }, + { + "epoch": 1.7, + "learning_rate": 2.4930469603203498e-06, + "loss": 1.3314, + "step": 9334 + }, + { + "epoch": 1.7, + "learning_rate": 2.4929923552966873e-06, + "loss": 0.2766, + "step": 9335 + }, + { + "epoch": 1.7, + "learning_rate": 2.4929377502730252e-06, + "loss": 2.4271, + "step": 9336 + }, + { + "epoch": 1.7, + "learning_rate": 2.492883145249363e-06, + "loss": 0.8476, + "step": 9337 + }, + { + "epoch": 1.7, + "learning_rate": 2.4928285402257007e-06, + "loss": 1.017, + "step": 9338 + }, + { + "epoch": 1.7, + "learning_rate": 2.4927739352020386e-06, + "loss": 0.7247, + "step": 9339 + }, + { + "epoch": 1.7, + "learning_rate": 2.4927193301783766e-06, + "loss": 0.0999, + "step": 9340 + }, + { + "epoch": 1.7, + "learning_rate": 2.492664725154714e-06, + "loss": 1.1451, + "step": 9341 + }, + { + "epoch": 1.7, + "learning_rate": 2.492610120131052e-06, + "loss": 2.8279, + "step": 9342 + }, + { + "epoch": 1.7, + "learning_rate": 2.49255551510739e-06, + "loss": 0.8906, + "step": 9343 + }, + { + "epoch": 1.7, + "learning_rate": 2.492500910083728e-06, + "loss": 0.5588, + "step": 9344 + }, + { + "epoch": 1.7, + "learning_rate": 2.492446305060066e-06, + "loss": 0.4997, + "step": 9345 + }, + { + "epoch": 1.7, + "learning_rate": 2.4923917000364034e-06, + "loss": 1.0842, + "step": 9346 + }, + { + "epoch": 1.7, + "learning_rate": 2.4923370950127413e-06, + "loss": 0.1912, + "step": 9347 + }, + { + "epoch": 1.7, + "learning_rate": 2.4922824899890793e-06, + "loss": 0.9091, + "step": 9348 + }, + { + "epoch": 1.7, + "learning_rate": 2.4922278849654168e-06, + "loss": 0.2228, + "step": 9349 + }, + { + "epoch": 1.7, + "learning_rate": 2.4921732799417547e-06, + "loss": 1.0793, + "step": 9350 + }, + { + "epoch": 1.7, + "learning_rate": 2.4921186749180927e-06, + "loss": 0.9114, + "step": 9351 + }, + { + "epoch": 1.7, + "learning_rate": 2.49206406989443e-06, + "loss": 0.625, + "step": 9352 + }, + { + "epoch": 1.7, + "learning_rate": 2.492009464870768e-06, + "loss": 0.576, + "step": 9353 + }, + { + "epoch": 1.7, + "learning_rate": 2.491954859847106e-06, + "loss": 0.9707, + "step": 9354 + }, + { + "epoch": 1.7, + "learning_rate": 2.4919002548234436e-06, + "loss": 1.3699, + "step": 9355 + }, + { + "epoch": 1.7, + "learning_rate": 2.4918456497997815e-06, + "loss": 0.6933, + "step": 9356 + }, + { + "epoch": 1.7, + "learning_rate": 2.4917910447761195e-06, + "loss": 3.379, + "step": 9357 + }, + { + "epoch": 1.7, + "learning_rate": 2.4917364397524574e-06, + "loss": 1.766, + "step": 9358 + }, + { + "epoch": 1.7, + "learning_rate": 2.4916818347287954e-06, + "loss": 0.6895, + "step": 9359 + }, + { + "epoch": 1.7, + "learning_rate": 2.491627229705133e-06, + "loss": 0.1045, + "step": 9360 + }, + { + "epoch": 1.7, + "learning_rate": 2.491572624681471e-06, + "loss": 1.4672, + "step": 9361 + }, + { + "epoch": 1.7, + "learning_rate": 2.4915180196578088e-06, + "loss": 0.7284, + "step": 9362 + }, + { + "epoch": 1.7, + "learning_rate": 2.4914634146341463e-06, + "loss": 1.1151, + "step": 9363 + }, + { + "epoch": 1.7, + "learning_rate": 2.4914088096104842e-06, + "loss": 1.5399, + "step": 9364 + }, + { + "epoch": 1.7, + "learning_rate": 2.491354204586822e-06, + "loss": 0.9303, + "step": 9365 + }, + { + "epoch": 1.7, + "learning_rate": 2.4912995995631597e-06, + "loss": 1.6475, + "step": 9366 + }, + { + "epoch": 1.7, + "learning_rate": 2.4912449945394976e-06, + "loss": 1.3215, + "step": 9367 + }, + { + "epoch": 1.7, + "learning_rate": 2.4911903895158356e-06, + "loss": 1.1119, + "step": 9368 + }, + { + "epoch": 1.7, + "learning_rate": 2.491135784492173e-06, + "loss": 0.9607, + "step": 9369 + }, + { + "epoch": 1.7, + "learning_rate": 2.491081179468511e-06, + "loss": 0.3061, + "step": 9370 + }, + { + "epoch": 1.7, + "learning_rate": 2.491026574444849e-06, + "loss": 0.4655, + "step": 9371 + }, + { + "epoch": 1.7, + "learning_rate": 2.490971969421187e-06, + "loss": 0.9658, + "step": 9372 + }, + { + "epoch": 1.7, + "learning_rate": 2.490917364397525e-06, + "loss": 1.3463, + "step": 9373 + }, + { + "epoch": 1.7, + "learning_rate": 2.4908627593738624e-06, + "loss": 0.8763, + "step": 9374 + }, + { + "epoch": 1.7, + "learning_rate": 2.4908081543502003e-06, + "loss": 1.0496, + "step": 9375 + }, + { + "epoch": 1.71, + "learning_rate": 2.4907535493265383e-06, + "loss": 0.8284, + "step": 9376 + }, + { + "epoch": 1.71, + "learning_rate": 2.490698944302876e-06, + "loss": 3.1039, + "step": 9377 + }, + { + "epoch": 1.71, + "learning_rate": 2.4906443392792138e-06, + "loss": 1.113, + "step": 9378 + }, + { + "epoch": 1.71, + "learning_rate": 2.4905897342555517e-06, + "loss": 1.5892, + "step": 9379 + }, + { + "epoch": 1.71, + "learning_rate": 2.4905351292318892e-06, + "loss": 0.794, + "step": 9380 + }, + { + "epoch": 1.71, + "learning_rate": 2.490480524208227e-06, + "loss": 1.2686, + "step": 9381 + }, + { + "epoch": 1.71, + "learning_rate": 2.490425919184565e-06, + "loss": 0.7035, + "step": 9382 + }, + { + "epoch": 1.71, + "learning_rate": 2.4903713141609026e-06, + "loss": 0.5425, + "step": 9383 + }, + { + "epoch": 1.71, + "learning_rate": 2.4903167091372406e-06, + "loss": 2.0876, + "step": 9384 + }, + { + "epoch": 1.71, + "learning_rate": 2.4902621041135785e-06, + "loss": 1.8255, + "step": 9385 + }, + { + "epoch": 1.71, + "learning_rate": 2.4902074990899165e-06, + "loss": 2.4736, + "step": 9386 + }, + { + "epoch": 1.71, + "learning_rate": 2.4901528940662544e-06, + "loss": 0.5458, + "step": 9387 + }, + { + "epoch": 1.71, + "learning_rate": 2.490098289042592e-06, + "loss": 0.4607, + "step": 9388 + }, + { + "epoch": 1.71, + "learning_rate": 2.49004368401893e-06, + "loss": 0.214, + "step": 9389 + }, + { + "epoch": 1.71, + "learning_rate": 2.489989078995268e-06, + "loss": 0.2721, + "step": 9390 + }, + { + "epoch": 1.71, + "learning_rate": 2.4899344739716053e-06, + "loss": 0.1652, + "step": 9391 + }, + { + "epoch": 1.71, + "learning_rate": 2.4898798689479433e-06, + "loss": 0.394, + "step": 9392 + }, + { + "epoch": 1.71, + "learning_rate": 2.489825263924281e-06, + "loss": 0.4123, + "step": 9393 + }, + { + "epoch": 1.71, + "learning_rate": 2.4897706589006187e-06, + "loss": 0.0478, + "step": 9394 + }, + { + "epoch": 1.71, + "learning_rate": 2.4897160538769567e-06, + "loss": 0.0376, + "step": 9395 + }, + { + "epoch": 1.71, + "learning_rate": 2.4896614488532946e-06, + "loss": 0.7167, + "step": 9396 + }, + { + "epoch": 1.71, + "learning_rate": 2.489606843829632e-06, + "loss": 0.8043, + "step": 9397 + }, + { + "epoch": 1.71, + "learning_rate": 2.48955223880597e-06, + "loss": 0.1796, + "step": 9398 + }, + { + "epoch": 1.71, + "learning_rate": 2.489497633782308e-06, + "loss": 1.6367, + "step": 9399 + }, + { + "epoch": 1.71, + "learning_rate": 2.489443028758646e-06, + "loss": 0.3163, + "step": 9400 + }, + { + "epoch": 1.71, + "learning_rate": 2.489388423734984e-06, + "loss": 0.6769, + "step": 9401 + }, + { + "epoch": 1.71, + "learning_rate": 2.4893338187113214e-06, + "loss": 0.167, + "step": 9402 + }, + { + "epoch": 1.71, + "learning_rate": 2.4892792136876594e-06, + "loss": 0.0521, + "step": 9403 + }, + { + "epoch": 1.71, + "learning_rate": 2.4892246086639973e-06, + "loss": 3.2305, + "step": 9404 + }, + { + "epoch": 1.71, + "learning_rate": 2.489170003640335e-06, + "loss": 0.3399, + "step": 9405 + }, + { + "epoch": 1.71, + "learning_rate": 2.4891153986166728e-06, + "loss": 0.7268, + "step": 9406 + }, + { + "epoch": 1.71, + "learning_rate": 2.4890607935930107e-06, + "loss": 0.2959, + "step": 9407 + }, + { + "epoch": 1.71, + "learning_rate": 2.4890061885693482e-06, + "loss": 0.644, + "step": 9408 + }, + { + "epoch": 1.71, + "learning_rate": 2.488951583545686e-06, + "loss": 1.4602, + "step": 9409 + }, + { + "epoch": 1.71, + "learning_rate": 2.488896978522024e-06, + "loss": 0.1927, + "step": 9410 + }, + { + "epoch": 1.71, + "learning_rate": 2.4888423734983617e-06, + "loss": 1.0416, + "step": 9411 + }, + { + "epoch": 1.71, + "learning_rate": 2.4887877684746996e-06, + "loss": 0.5582, + "step": 9412 + }, + { + "epoch": 1.71, + "learning_rate": 2.4887331634510375e-06, + "loss": 0.5816, + "step": 9413 + }, + { + "epoch": 1.71, + "learning_rate": 2.4886785584273755e-06, + "loss": 0.7855, + "step": 9414 + }, + { + "epoch": 1.71, + "learning_rate": 2.4886239534037134e-06, + "loss": 2.0151, + "step": 9415 + }, + { + "epoch": 1.71, + "learning_rate": 2.488569348380051e-06, + "loss": 1.9521, + "step": 9416 + }, + { + "epoch": 1.71, + "learning_rate": 2.488514743356389e-06, + "loss": 5.7192, + "step": 9417 + }, + { + "epoch": 1.71, + "learning_rate": 2.488460138332727e-06, + "loss": 1.1954, + "step": 9418 + }, + { + "epoch": 1.71, + "learning_rate": 2.4884055333090644e-06, + "loss": 0.3049, + "step": 9419 + }, + { + "epoch": 1.71, + "learning_rate": 2.4883509282854023e-06, + "loss": 1.2298, + "step": 9420 + }, + { + "epoch": 1.71, + "learning_rate": 2.4882963232617402e-06, + "loss": 1.6516, + "step": 9421 + }, + { + "epoch": 1.71, + "learning_rate": 2.4882417182380778e-06, + "loss": 1.3545, + "step": 9422 + }, + { + "epoch": 1.71, + "learning_rate": 2.4881871132144157e-06, + "loss": 1.4559, + "step": 9423 + }, + { + "epoch": 1.71, + "learning_rate": 2.4881325081907536e-06, + "loss": 1.9427, + "step": 9424 + }, + { + "epoch": 1.71, + "learning_rate": 2.4880779031670916e-06, + "loss": 0.9788, + "step": 9425 + }, + { + "epoch": 1.71, + "learning_rate": 2.4880232981434295e-06, + "loss": 0.0873, + "step": 9426 + }, + { + "epoch": 1.71, + "learning_rate": 2.487968693119767e-06, + "loss": 0.0423, + "step": 9427 + }, + { + "epoch": 1.71, + "learning_rate": 2.487914088096105e-06, + "loss": 1.2387, + "step": 9428 + }, + { + "epoch": 1.71, + "learning_rate": 2.487859483072443e-06, + "loss": 1.0268, + "step": 9429 + }, + { + "epoch": 1.71, + "learning_rate": 2.4878048780487805e-06, + "loss": 0.9872, + "step": 9430 + }, + { + "epoch": 1.72, + "learning_rate": 2.4877502730251184e-06, + "loss": 0.5643, + "step": 9431 + }, + { + "epoch": 1.72, + "learning_rate": 2.4876956680014563e-06, + "loss": 0.291, + "step": 9432 + }, + { + "epoch": 1.72, + "learning_rate": 2.487641062977794e-06, + "loss": 0.2838, + "step": 9433 + }, + { + "epoch": 1.72, + "learning_rate": 2.487586457954132e-06, + "loss": 0.0452, + "step": 9434 + }, + { + "epoch": 1.72, + "learning_rate": 2.4875318529304698e-06, + "loss": 0.8728, + "step": 9435 + }, + { + "epoch": 1.72, + "learning_rate": 2.4874772479068073e-06, + "loss": 2.9334, + "step": 9436 + }, + { + "epoch": 1.72, + "learning_rate": 2.4874226428831452e-06, + "loss": 1.1363, + "step": 9437 + }, + { + "epoch": 1.72, + "learning_rate": 2.487368037859483e-06, + "loss": 0.9614, + "step": 9438 + }, + { + "epoch": 1.72, + "learning_rate": 2.487313432835821e-06, + "loss": 0.0178, + "step": 9439 + }, + { + "epoch": 1.72, + "learning_rate": 2.487258827812159e-06, + "loss": 0.0424, + "step": 9440 + }, + { + "epoch": 1.72, + "learning_rate": 2.4872042227884966e-06, + "loss": 1.4483, + "step": 9441 + }, + { + "epoch": 1.72, + "learning_rate": 2.4871496177648345e-06, + "loss": 2.7202, + "step": 9442 + }, + { + "epoch": 1.72, + "learning_rate": 2.4870950127411725e-06, + "loss": 1.7603, + "step": 9443 + }, + { + "epoch": 1.72, + "learning_rate": 2.48704040771751e-06, + "loss": 2.4199, + "step": 9444 + }, + { + "epoch": 1.72, + "learning_rate": 2.486985802693848e-06, + "loss": 0.5196, + "step": 9445 + }, + { + "epoch": 1.72, + "learning_rate": 2.486931197670186e-06, + "loss": 0.3511, + "step": 9446 + }, + { + "epoch": 1.72, + "learning_rate": 2.4868765926465234e-06, + "loss": 2.6034, + "step": 9447 + }, + { + "epoch": 1.72, + "learning_rate": 2.4868219876228613e-06, + "loss": 0.4685, + "step": 9448 + }, + { + "epoch": 1.72, + "learning_rate": 2.4867673825991993e-06, + "loss": 2.7709, + "step": 9449 + }, + { + "epoch": 1.72, + "learning_rate": 2.4867127775755368e-06, + "loss": 1.0833, + "step": 9450 + }, + { + "epoch": 1.72, + "learning_rate": 2.4866581725518747e-06, + "loss": 0.9972, + "step": 9451 + }, + { + "epoch": 1.72, + "learning_rate": 2.4866035675282127e-06, + "loss": 1.6962, + "step": 9452 + }, + { + "epoch": 1.72, + "learning_rate": 2.4865489625045506e-06, + "loss": 2.3442, + "step": 9453 + }, + { + "epoch": 1.72, + "learning_rate": 2.4864943574808886e-06, + "loss": 0.9188, + "step": 9454 + }, + { + "epoch": 1.72, + "learning_rate": 2.486439752457226e-06, + "loss": 1.38, + "step": 9455 + }, + { + "epoch": 1.72, + "learning_rate": 2.486385147433564e-06, + "loss": 0.9329, + "step": 9456 + }, + { + "epoch": 1.72, + "learning_rate": 2.486330542409902e-06, + "loss": 1.4073, + "step": 9457 + }, + { + "epoch": 1.72, + "learning_rate": 2.4862759373862395e-06, + "loss": 1.0042, + "step": 9458 + }, + { + "epoch": 1.72, + "learning_rate": 2.4862213323625774e-06, + "loss": 0.7347, + "step": 9459 + }, + { + "epoch": 1.72, + "learning_rate": 2.4861667273389154e-06, + "loss": 1.6883, + "step": 9460 + }, + { + "epoch": 1.72, + "learning_rate": 2.486112122315253e-06, + "loss": 0.8504, + "step": 9461 + }, + { + "epoch": 1.72, + "learning_rate": 2.486057517291591e-06, + "loss": 0.0652, + "step": 9462 + }, + { + "epoch": 1.72, + "learning_rate": 2.4860029122679288e-06, + "loss": 0.1097, + "step": 9463 + }, + { + "epoch": 1.72, + "learning_rate": 2.4859483072442663e-06, + "loss": 0.4649, + "step": 9464 + }, + { + "epoch": 1.72, + "learning_rate": 2.4858937022206042e-06, + "loss": 1.259, + "step": 9465 + }, + { + "epoch": 1.72, + "learning_rate": 2.485839097196942e-06, + "loss": 0.9408, + "step": 9466 + }, + { + "epoch": 1.72, + "learning_rate": 2.48578449217328e-06, + "loss": 0.8378, + "step": 9467 + }, + { + "epoch": 1.72, + "learning_rate": 2.485729887149618e-06, + "loss": 0.4337, + "step": 9468 + }, + { + "epoch": 1.72, + "learning_rate": 2.4856752821259556e-06, + "loss": 1.8434, + "step": 9469 + }, + { + "epoch": 1.72, + "learning_rate": 2.4856206771022935e-06, + "loss": 0.0379, + "step": 9470 + }, + { + "epoch": 1.72, + "learning_rate": 2.4855660720786315e-06, + "loss": 0.7955, + "step": 9471 + }, + { + "epoch": 1.72, + "learning_rate": 2.485511467054969e-06, + "loss": 0.7339, + "step": 9472 + }, + { + "epoch": 1.72, + "learning_rate": 2.485456862031307e-06, + "loss": 0.0883, + "step": 9473 + }, + { + "epoch": 1.72, + "learning_rate": 2.485402257007645e-06, + "loss": 0.2861, + "step": 9474 + }, + { + "epoch": 1.72, + "learning_rate": 2.4853476519839824e-06, + "loss": 0.9651, + "step": 9475 + }, + { + "epoch": 1.72, + "learning_rate": 2.4852930469603203e-06, + "loss": 0.1006, + "step": 9476 + }, + { + "epoch": 1.72, + "learning_rate": 2.4852384419366583e-06, + "loss": 1.6437, + "step": 9477 + }, + { + "epoch": 1.72, + "learning_rate": 2.485183836912996e-06, + "loss": 0.2818, + "step": 9478 + }, + { + "epoch": 1.72, + "learning_rate": 2.4851292318893338e-06, + "loss": 0.5956, + "step": 9479 + }, + { + "epoch": 1.72, + "learning_rate": 2.4850746268656717e-06, + "loss": 0.1857, + "step": 9480 + }, + { + "epoch": 1.72, + "learning_rate": 2.4850200218420096e-06, + "loss": 1.1284, + "step": 9481 + }, + { + "epoch": 1.72, + "learning_rate": 2.4849654168183476e-06, + "loss": 0.1517, + "step": 9482 + }, + { + "epoch": 1.72, + "learning_rate": 2.484910811794685e-06, + "loss": 0.1023, + "step": 9483 + }, + { + "epoch": 1.72, + "learning_rate": 2.484856206771023e-06, + "loss": 0.4267, + "step": 9484 + }, + { + "epoch": 1.72, + "learning_rate": 2.484801601747361e-06, + "loss": 2.737, + "step": 9485 + }, + { + "epoch": 1.73, + "learning_rate": 2.4847469967236985e-06, + "loss": 0.6808, + "step": 9486 + }, + { + "epoch": 1.73, + "learning_rate": 2.4846923917000365e-06, + "loss": 2.5887, + "step": 9487 + }, + { + "epoch": 1.73, + "learning_rate": 2.4846377866763744e-06, + "loss": 2.6789, + "step": 9488 + }, + { + "epoch": 1.73, + "learning_rate": 2.484583181652712e-06, + "loss": 0.8372, + "step": 9489 + }, + { + "epoch": 1.73, + "learning_rate": 2.48452857662905e-06, + "loss": 0.4895, + "step": 9490 + }, + { + "epoch": 1.73, + "learning_rate": 2.484473971605388e-06, + "loss": 1.3091, + "step": 9491 + }, + { + "epoch": 1.73, + "learning_rate": 2.4844193665817253e-06, + "loss": 2.5762, + "step": 9492 + }, + { + "epoch": 1.73, + "learning_rate": 2.4843647615580633e-06, + "loss": 0.0738, + "step": 9493 + }, + { + "epoch": 1.73, + "learning_rate": 2.4843101565344012e-06, + "loss": 1.5014, + "step": 9494 + }, + { + "epoch": 1.73, + "learning_rate": 2.484255551510739e-06, + "loss": 0.2809, + "step": 9495 + }, + { + "epoch": 1.73, + "learning_rate": 2.484200946487077e-06, + "loss": 0.7621, + "step": 9496 + }, + { + "epoch": 1.73, + "learning_rate": 2.4841463414634146e-06, + "loss": 0.2738, + "step": 9497 + }, + { + "epoch": 1.73, + "learning_rate": 2.4840917364397526e-06, + "loss": 2.1573, + "step": 9498 + }, + { + "epoch": 1.73, + "learning_rate": 2.4840371314160905e-06, + "loss": 0.2524, + "step": 9499 + }, + { + "epoch": 1.73, + "learning_rate": 2.483982526392428e-06, + "loss": 0.0898, + "step": 9500 + }, + { + "epoch": 1.73, + "learning_rate": 2.483927921368766e-06, + "loss": 0.6482, + "step": 9501 + }, + { + "epoch": 1.73, + "learning_rate": 2.483873316345104e-06, + "loss": 0.0347, + "step": 9502 + }, + { + "epoch": 1.73, + "learning_rate": 2.4838187113214414e-06, + "loss": 0.388, + "step": 9503 + }, + { + "epoch": 1.73, + "learning_rate": 2.4837641062977794e-06, + "loss": 0.4081, + "step": 9504 + }, + { + "epoch": 1.73, + "learning_rate": 2.4837095012741173e-06, + "loss": 0.3722, + "step": 9505 + }, + { + "epoch": 1.73, + "learning_rate": 2.4836548962504553e-06, + "loss": 1.1247, + "step": 9506 + }, + { + "epoch": 1.73, + "learning_rate": 2.483600291226793e-06, + "loss": 0.1051, + "step": 9507 + }, + { + "epoch": 1.73, + "learning_rate": 2.4835456862031307e-06, + "loss": 0.2835, + "step": 9508 + }, + { + "epoch": 1.73, + "learning_rate": 2.4834910811794687e-06, + "loss": 0.8117, + "step": 9509 + }, + { + "epoch": 1.73, + "learning_rate": 2.4834364761558066e-06, + "loss": 0.2702, + "step": 9510 + }, + { + "epoch": 1.73, + "learning_rate": 2.483381871132144e-06, + "loss": 1.0465, + "step": 9511 + }, + { + "epoch": 1.73, + "learning_rate": 2.483327266108482e-06, + "loss": 0.5215, + "step": 9512 + }, + { + "epoch": 1.73, + "learning_rate": 2.48327266108482e-06, + "loss": 0.0376, + "step": 9513 + }, + { + "epoch": 1.73, + "learning_rate": 2.4832180560611575e-06, + "loss": 1.0296, + "step": 9514 + }, + { + "epoch": 1.73, + "learning_rate": 2.4831634510374955e-06, + "loss": 0.5203, + "step": 9515 + }, + { + "epoch": 1.73, + "learning_rate": 2.4831088460138334e-06, + "loss": 3.0269, + "step": 9516 + }, + { + "epoch": 1.73, + "learning_rate": 2.483054240990171e-06, + "loss": 2.0879, + "step": 9517 + }, + { + "epoch": 1.73, + "learning_rate": 2.482999635966509e-06, + "loss": 0.5662, + "step": 9518 + }, + { + "epoch": 1.73, + "learning_rate": 2.482945030942847e-06, + "loss": 1.3417, + "step": 9519 + }, + { + "epoch": 1.73, + "learning_rate": 2.4828904259191848e-06, + "loss": 1.977, + "step": 9520 + }, + { + "epoch": 1.73, + "learning_rate": 2.4828358208955227e-06, + "loss": 2.4504, + "step": 9521 + }, + { + "epoch": 1.73, + "learning_rate": 2.4827812158718602e-06, + "loss": 1.6847, + "step": 9522 + }, + { + "epoch": 1.73, + "learning_rate": 2.482726610848198e-06, + "loss": 0.3816, + "step": 9523 + }, + { + "epoch": 1.73, + "learning_rate": 2.482672005824536e-06, + "loss": 0.5045, + "step": 9524 + }, + { + "epoch": 1.73, + "learning_rate": 2.4826174008008736e-06, + "loss": 2.9773, + "step": 9525 + }, + { + "epoch": 1.73, + "learning_rate": 2.4825627957772116e-06, + "loss": 0.8572, + "step": 9526 + }, + { + "epoch": 1.73, + "learning_rate": 2.4825081907535495e-06, + "loss": 2.0671, + "step": 9527 + }, + { + "epoch": 1.73, + "learning_rate": 2.482453585729887e-06, + "loss": 0.1472, + "step": 9528 + }, + { + "epoch": 1.73, + "learning_rate": 2.482398980706225e-06, + "loss": 0.3888, + "step": 9529 + }, + { + "epoch": 1.73, + "learning_rate": 2.482344375682563e-06, + "loss": 2.8595, + "step": 9530 + }, + { + "epoch": 1.73, + "learning_rate": 2.4822897706589005e-06, + "loss": 0.078, + "step": 9531 + }, + { + "epoch": 1.73, + "learning_rate": 2.4822351656352384e-06, + "loss": 1.1066, + "step": 9532 + }, + { + "epoch": 1.73, + "learning_rate": 2.4821805606115763e-06, + "loss": 1.0974, + "step": 9533 + }, + { + "epoch": 1.73, + "learning_rate": 2.4821259555879143e-06, + "loss": 0.9112, + "step": 9534 + }, + { + "epoch": 1.73, + "learning_rate": 2.4820713505642522e-06, + "loss": 2.0527, + "step": 9535 + }, + { + "epoch": 1.73, + "learning_rate": 2.4820167455405898e-06, + "loss": 0.1688, + "step": 9536 + }, + { + "epoch": 1.73, + "learning_rate": 2.4819621405169277e-06, + "loss": 1.2803, + "step": 9537 + }, + { + "epoch": 1.73, + "learning_rate": 2.4819075354932656e-06, + "loss": 3.4738, + "step": 9538 + }, + { + "epoch": 1.73, + "learning_rate": 2.481852930469603e-06, + "loss": 3.3071, + "step": 9539 + }, + { + "epoch": 1.73, + "learning_rate": 2.481798325445941e-06, + "loss": 0.7193, + "step": 9540 + }, + { + "epoch": 1.74, + "learning_rate": 2.481743720422279e-06, + "loss": 0.5635, + "step": 9541 + }, + { + "epoch": 1.74, + "learning_rate": 2.4816891153986166e-06, + "loss": 0.6165, + "step": 9542 + }, + { + "epoch": 1.74, + "learning_rate": 2.4816345103749545e-06, + "loss": 1.2055, + "step": 9543 + }, + { + "epoch": 1.74, + "learning_rate": 2.4815799053512925e-06, + "loss": 1.1015, + "step": 9544 + }, + { + "epoch": 1.74, + "learning_rate": 2.48152530032763e-06, + "loss": 0.6486, + "step": 9545 + }, + { + "epoch": 1.74, + "learning_rate": 2.481470695303968e-06, + "loss": 0.9776, + "step": 9546 + }, + { + "epoch": 1.74, + "learning_rate": 2.481416090280306e-06, + "loss": 0.4173, + "step": 9547 + }, + { + "epoch": 1.74, + "learning_rate": 2.481361485256644e-06, + "loss": 0.5309, + "step": 9548 + }, + { + "epoch": 1.74, + "learning_rate": 2.4813068802329817e-06, + "loss": 0.5315, + "step": 9549 + }, + { + "epoch": 1.74, + "learning_rate": 2.4812522752093193e-06, + "loss": 0.046, + "step": 9550 + }, + { + "epoch": 1.74, + "learning_rate": 2.481197670185657e-06, + "loss": 0.5401, + "step": 9551 + }, + { + "epoch": 1.74, + "learning_rate": 2.481143065161995e-06, + "loss": 1.1109, + "step": 9552 + }, + { + "epoch": 1.74, + "learning_rate": 2.4810884601383327e-06, + "loss": 0.2679, + "step": 9553 + }, + { + "epoch": 1.74, + "learning_rate": 2.4810338551146706e-06, + "loss": 1.7078, + "step": 9554 + }, + { + "epoch": 1.74, + "learning_rate": 2.4809792500910086e-06, + "loss": 1.5783, + "step": 9555 + }, + { + "epoch": 1.74, + "learning_rate": 2.480924645067346e-06, + "loss": 0.0756, + "step": 9556 + }, + { + "epoch": 1.74, + "learning_rate": 2.480870040043684e-06, + "loss": 0.4134, + "step": 9557 + }, + { + "epoch": 1.74, + "learning_rate": 2.480815435020022e-06, + "loss": 3.4068, + "step": 9558 + }, + { + "epoch": 1.74, + "learning_rate": 2.4807608299963595e-06, + "loss": 0.4747, + "step": 9559 + }, + { + "epoch": 1.74, + "learning_rate": 2.4807062249726974e-06, + "loss": 1.4198, + "step": 9560 + }, + { + "epoch": 1.74, + "learning_rate": 2.4806516199490354e-06, + "loss": 0.1943, + "step": 9561 + }, + { + "epoch": 1.74, + "learning_rate": 2.4805970149253733e-06, + "loss": 0.6266, + "step": 9562 + }, + { + "epoch": 1.74, + "learning_rate": 2.4805424099017113e-06, + "loss": 3.1773, + "step": 9563 + }, + { + "epoch": 1.74, + "learning_rate": 2.4804878048780488e-06, + "loss": 0.0336, + "step": 9564 + }, + { + "epoch": 1.74, + "learning_rate": 2.4804331998543867e-06, + "loss": 2.3077, + "step": 9565 + }, + { + "epoch": 1.74, + "learning_rate": 2.4803785948307247e-06, + "loss": 0.5821, + "step": 9566 + }, + { + "epoch": 1.74, + "learning_rate": 2.480323989807062e-06, + "loss": 0.9308, + "step": 9567 + }, + { + "epoch": 1.74, + "learning_rate": 2.4802693847834e-06, + "loss": 0.0199, + "step": 9568 + }, + { + "epoch": 1.74, + "learning_rate": 2.480214779759738e-06, + "loss": 0.1638, + "step": 9569 + }, + { + "epoch": 1.74, + "learning_rate": 2.4801601747360756e-06, + "loss": 0.9688, + "step": 9570 + }, + { + "epoch": 1.74, + "learning_rate": 2.4801055697124135e-06, + "loss": 1.2704, + "step": 9571 + }, + { + "epoch": 1.74, + "learning_rate": 2.4800509646887515e-06, + "loss": 1.5213, + "step": 9572 + }, + { + "epoch": 1.74, + "learning_rate": 2.479996359665089e-06, + "loss": 0.5933, + "step": 9573 + }, + { + "epoch": 1.74, + "learning_rate": 2.479941754641427e-06, + "loss": 0.3933, + "step": 9574 + }, + { + "epoch": 1.74, + "learning_rate": 2.479887149617765e-06, + "loss": 0.8962, + "step": 9575 + }, + { + "epoch": 1.74, + "learning_rate": 2.479832544594103e-06, + "loss": 0.569, + "step": 9576 + }, + { + "epoch": 1.74, + "learning_rate": 2.4797779395704408e-06, + "loss": 0.4884, + "step": 9577 + }, + { + "epoch": 1.74, + "learning_rate": 2.4797233345467783e-06, + "loss": 0.8995, + "step": 9578 + }, + { + "epoch": 1.74, + "learning_rate": 2.4796687295231162e-06, + "loss": 0.7773, + "step": 9579 + }, + { + "epoch": 1.74, + "learning_rate": 2.479614124499454e-06, + "loss": 1.0238, + "step": 9580 + }, + { + "epoch": 1.74, + "learning_rate": 2.4795595194757917e-06, + "loss": 1.398, + "step": 9581 + }, + { + "epoch": 1.74, + "learning_rate": 2.4795049144521296e-06, + "loss": 0.113, + "step": 9582 + }, + { + "epoch": 1.74, + "learning_rate": 2.4794503094284676e-06, + "loss": 2.8001, + "step": 9583 + }, + { + "epoch": 1.74, + "learning_rate": 2.479395704404805e-06, + "loss": 0.2711, + "step": 9584 + }, + { + "epoch": 1.74, + "learning_rate": 2.479341099381143e-06, + "loss": 1.4971, + "step": 9585 + }, + { + "epoch": 1.74, + "learning_rate": 2.479286494357481e-06, + "loss": 0.9237, + "step": 9586 + }, + { + "epoch": 1.74, + "learning_rate": 2.4792318893338185e-06, + "loss": 0.0658, + "step": 9587 + }, + { + "epoch": 1.74, + "learning_rate": 2.4791772843101565e-06, + "loss": 2.2304, + "step": 9588 + }, + { + "epoch": 1.74, + "learning_rate": 2.4791226792864944e-06, + "loss": 0.2548, + "step": 9589 + }, + { + "epoch": 1.74, + "learning_rate": 2.4790680742628323e-06, + "loss": 0.1039, + "step": 9590 + }, + { + "epoch": 1.74, + "learning_rate": 2.4790134692391703e-06, + "loss": 1.4501, + "step": 9591 + }, + { + "epoch": 1.74, + "learning_rate": 2.478958864215508e-06, + "loss": 0.7152, + "step": 9592 + }, + { + "epoch": 1.74, + "learning_rate": 2.4789042591918457e-06, + "loss": 1.6867, + "step": 9593 + }, + { + "epoch": 1.74, + "learning_rate": 2.4788496541681837e-06, + "loss": 1.7927, + "step": 9594 + }, + { + "epoch": 1.74, + "learning_rate": 2.4787950491445212e-06, + "loss": 0.5062, + "step": 9595 + }, + { + "epoch": 1.75, + "learning_rate": 2.478740444120859e-06, + "loss": 1.2164, + "step": 9596 + }, + { + "epoch": 1.75, + "learning_rate": 2.478685839097197e-06, + "loss": 1.3032, + "step": 9597 + }, + { + "epoch": 1.75, + "learning_rate": 2.4786312340735346e-06, + "loss": 1.0526, + "step": 9598 + }, + { + "epoch": 1.75, + "learning_rate": 2.4785766290498726e-06, + "loss": 1.8008, + "step": 9599 + }, + { + "epoch": 1.75, + "learning_rate": 2.4785220240262105e-06, + "loss": 0.1057, + "step": 9600 + }, + { + "epoch": 1.75, + "learning_rate": 2.4784674190025484e-06, + "loss": 1.0523, + "step": 9601 + }, + { + "epoch": 1.75, + "learning_rate": 2.4784128139788864e-06, + "loss": 1.3052, + "step": 9602 + }, + { + "epoch": 1.75, + "learning_rate": 2.478358208955224e-06, + "loss": 2.052, + "step": 9603 + }, + { + "epoch": 1.75, + "learning_rate": 2.478303603931562e-06, + "loss": 1.1634, + "step": 9604 + }, + { + "epoch": 1.75, + "learning_rate": 2.4782489989079e-06, + "loss": 0.4519, + "step": 9605 + }, + { + "epoch": 1.75, + "learning_rate": 2.4781943938842373e-06, + "loss": 0.892, + "step": 9606 + }, + { + "epoch": 1.75, + "learning_rate": 2.4781397888605753e-06, + "loss": 0.1734, + "step": 9607 + }, + { + "epoch": 1.75, + "learning_rate": 2.478085183836913e-06, + "loss": 0.4779, + "step": 9608 + }, + { + "epoch": 1.75, + "learning_rate": 2.4780305788132507e-06, + "loss": 1.2875, + "step": 9609 + }, + { + "epoch": 1.75, + "learning_rate": 2.4779759737895887e-06, + "loss": 0.4339, + "step": 9610 + }, + { + "epoch": 1.75, + "learning_rate": 2.4779213687659266e-06, + "loss": 0.0446, + "step": 9611 + }, + { + "epoch": 1.75, + "learning_rate": 2.477866763742264e-06, + "loss": 0.1761, + "step": 9612 + }, + { + "epoch": 1.75, + "learning_rate": 2.477812158718602e-06, + "loss": 0.9242, + "step": 9613 + }, + { + "epoch": 1.75, + "learning_rate": 2.47775755369494e-06, + "loss": 0.4888, + "step": 9614 + }, + { + "epoch": 1.75, + "learning_rate": 2.477702948671278e-06, + "loss": 2.2319, + "step": 9615 + }, + { + "epoch": 1.75, + "learning_rate": 2.477648343647616e-06, + "loss": 2.2498, + "step": 9616 + }, + { + "epoch": 1.75, + "learning_rate": 2.4775937386239534e-06, + "loss": 0.936, + "step": 9617 + }, + { + "epoch": 1.75, + "learning_rate": 2.4775391336002914e-06, + "loss": 1.5946, + "step": 9618 + }, + { + "epoch": 1.75, + "learning_rate": 2.4774845285766293e-06, + "loss": 0.6265, + "step": 9619 + }, + { + "epoch": 1.75, + "learning_rate": 2.477429923552967e-06, + "loss": 1.0567, + "step": 9620 + }, + { + "epoch": 1.75, + "learning_rate": 2.4773753185293048e-06, + "loss": 3.663, + "step": 9621 + }, + { + "epoch": 1.75, + "learning_rate": 2.4773207135056427e-06, + "loss": 0.272, + "step": 9622 + }, + { + "epoch": 1.75, + "learning_rate": 2.4772661084819802e-06, + "loss": 0.1932, + "step": 9623 + }, + { + "epoch": 1.75, + "learning_rate": 2.477211503458318e-06, + "loss": 0.0459, + "step": 9624 + }, + { + "epoch": 1.75, + "learning_rate": 2.477156898434656e-06, + "loss": 0.0493, + "step": 9625 + }, + { + "epoch": 1.75, + "learning_rate": 2.4771022934109936e-06, + "loss": 0.8996, + "step": 9626 + }, + { + "epoch": 1.75, + "learning_rate": 2.4770476883873316e-06, + "loss": 1.3576, + "step": 9627 + }, + { + "epoch": 1.75, + "learning_rate": 2.4769930833636695e-06, + "loss": 0.059, + "step": 9628 + }, + { + "epoch": 1.75, + "learning_rate": 2.4769384783400075e-06, + "loss": 0.6034, + "step": 9629 + }, + { + "epoch": 1.75, + "learning_rate": 2.4768838733163454e-06, + "loss": 2.7339, + "step": 9630 + }, + { + "epoch": 1.75, + "learning_rate": 2.476829268292683e-06, + "loss": 0.2002, + "step": 9631 + }, + { + "epoch": 1.75, + "learning_rate": 2.476774663269021e-06, + "loss": 1.0074, + "step": 9632 + }, + { + "epoch": 1.75, + "learning_rate": 2.476720058245359e-06, + "loss": 0.7813, + "step": 9633 + }, + { + "epoch": 1.75, + "learning_rate": 2.4766654532216963e-06, + "loss": 0.4679, + "step": 9634 + }, + { + "epoch": 1.75, + "learning_rate": 2.4766108481980343e-06, + "loss": 0.6957, + "step": 9635 + }, + { + "epoch": 1.75, + "learning_rate": 2.4765562431743722e-06, + "loss": 0.2244, + "step": 9636 + }, + { + "epoch": 1.75, + "learning_rate": 2.4765016381507098e-06, + "loss": 4.2667, + "step": 9637 + }, + { + "epoch": 1.75, + "learning_rate": 2.4764470331270477e-06, + "loss": 1.1405, + "step": 9638 + }, + { + "epoch": 1.75, + "learning_rate": 2.4763924281033856e-06, + "loss": 0.2001, + "step": 9639 + }, + { + "epoch": 1.75, + "learning_rate": 2.476337823079723e-06, + "loss": 0.662, + "step": 9640 + }, + { + "epoch": 1.75, + "learning_rate": 2.476283218056061e-06, + "loss": 1.6545, + "step": 9641 + }, + { + "epoch": 1.75, + "learning_rate": 2.476228613032399e-06, + "loss": 0.8127, + "step": 9642 + }, + { + "epoch": 1.75, + "learning_rate": 2.476174008008737e-06, + "loss": 1.0179, + "step": 9643 + }, + { + "epoch": 1.75, + "learning_rate": 2.476119402985075e-06, + "loss": 2.0904, + "step": 9644 + }, + { + "epoch": 1.75, + "learning_rate": 2.4760647979614125e-06, + "loss": 1.2048, + "step": 9645 + }, + { + "epoch": 1.75, + "learning_rate": 2.4760101929377504e-06, + "loss": 1.1083, + "step": 9646 + }, + { + "epoch": 1.75, + "learning_rate": 2.4759555879140883e-06, + "loss": 2.6041, + "step": 9647 + }, + { + "epoch": 1.75, + "learning_rate": 2.475900982890426e-06, + "loss": 1.5057, + "step": 9648 + }, + { + "epoch": 1.75, + "learning_rate": 2.475846377866764e-06, + "loss": 0.3247, + "step": 9649 + }, + { + "epoch": 1.75, + "learning_rate": 2.4757917728431017e-06, + "loss": 0.0347, + "step": 9650 + }, + { + "epoch": 1.76, + "learning_rate": 2.4757371678194393e-06, + "loss": 0.527, + "step": 9651 + }, + { + "epoch": 1.76, + "learning_rate": 2.475682562795777e-06, + "loss": 3.8162, + "step": 9652 + }, + { + "epoch": 1.76, + "learning_rate": 2.475627957772115e-06, + "loss": 3.0571, + "step": 9653 + }, + { + "epoch": 1.76, + "learning_rate": 2.4755733527484527e-06, + "loss": 0.3207, + "step": 9654 + }, + { + "epoch": 1.76, + "learning_rate": 2.4755187477247906e-06, + "loss": 0.3602, + "step": 9655 + }, + { + "epoch": 1.76, + "learning_rate": 2.4754641427011286e-06, + "loss": 0.4473, + "step": 9656 + }, + { + "epoch": 1.76, + "learning_rate": 2.4754095376774665e-06, + "loss": 1.1756, + "step": 9657 + }, + { + "epoch": 1.76, + "learning_rate": 2.4753549326538044e-06, + "loss": 2.5019, + "step": 9658 + }, + { + "epoch": 1.76, + "learning_rate": 2.475300327630142e-06, + "loss": 1.4379, + "step": 9659 + }, + { + "epoch": 1.76, + "learning_rate": 2.47524572260648e-06, + "loss": 1.3352, + "step": 9660 + }, + { + "epoch": 1.76, + "learning_rate": 2.475191117582818e-06, + "loss": 0.0504, + "step": 9661 + }, + { + "epoch": 1.76, + "learning_rate": 2.4751365125591554e-06, + "loss": 0.386, + "step": 9662 + }, + { + "epoch": 1.76, + "learning_rate": 2.4750819075354933e-06, + "loss": 1.2802, + "step": 9663 + }, + { + "epoch": 1.76, + "learning_rate": 2.4750273025118313e-06, + "loss": 1.2007, + "step": 9664 + }, + { + "epoch": 1.76, + "learning_rate": 2.4749726974881688e-06, + "loss": 0.0302, + "step": 9665 + }, + { + "epoch": 1.76, + "learning_rate": 2.4749180924645067e-06, + "loss": 2.7804, + "step": 9666 + }, + { + "epoch": 1.76, + "learning_rate": 2.4748634874408447e-06, + "loss": 0.189, + "step": 9667 + }, + { + "epoch": 1.76, + "learning_rate": 2.474808882417182e-06, + "loss": 1.5093, + "step": 9668 + }, + { + "epoch": 1.76, + "learning_rate": 2.47475427739352e-06, + "loss": 0.0131, + "step": 9669 + }, + { + "epoch": 1.76, + "learning_rate": 2.474699672369858e-06, + "loss": 0.1668, + "step": 9670 + }, + { + "epoch": 1.76, + "learning_rate": 2.474645067346196e-06, + "loss": 1.8885, + "step": 9671 + }, + { + "epoch": 1.76, + "learning_rate": 2.474590462322534e-06, + "loss": 3.219, + "step": 9672 + }, + { + "epoch": 1.76, + "learning_rate": 2.4745358572988715e-06, + "loss": 0.3446, + "step": 9673 + }, + { + "epoch": 1.76, + "learning_rate": 2.4744812522752094e-06, + "loss": 1.2286, + "step": 9674 + }, + { + "epoch": 1.76, + "learning_rate": 2.4744266472515474e-06, + "loss": 1.3522, + "step": 9675 + }, + { + "epoch": 1.76, + "learning_rate": 2.474372042227885e-06, + "loss": 0.0416, + "step": 9676 + }, + { + "epoch": 1.76, + "learning_rate": 2.474317437204223e-06, + "loss": 1.0699, + "step": 9677 + }, + { + "epoch": 1.76, + "learning_rate": 2.4742628321805608e-06, + "loss": 0.665, + "step": 9678 + }, + { + "epoch": 1.76, + "learning_rate": 2.4742082271568983e-06, + "loss": 0.4758, + "step": 9679 + }, + { + "epoch": 1.76, + "learning_rate": 2.4741536221332362e-06, + "loss": 0.5915, + "step": 9680 + }, + { + "epoch": 1.76, + "learning_rate": 2.474099017109574e-06, + "loss": 1.6451, + "step": 9681 + }, + { + "epoch": 1.76, + "learning_rate": 2.474044412085912e-06, + "loss": 0.2477, + "step": 9682 + }, + { + "epoch": 1.76, + "learning_rate": 2.47398980706225e-06, + "loss": 0.9638, + "step": 9683 + }, + { + "epoch": 1.76, + "learning_rate": 2.4739352020385876e-06, + "loss": 0.4704, + "step": 9684 + }, + { + "epoch": 1.76, + "learning_rate": 2.4738805970149255e-06, + "loss": 1.2498, + "step": 9685 + }, + { + "epoch": 1.76, + "learning_rate": 2.4738259919912635e-06, + "loss": 1.0239, + "step": 9686 + }, + { + "epoch": 1.76, + "learning_rate": 2.473771386967601e-06, + "loss": 0.1163, + "step": 9687 + }, + { + "epoch": 1.76, + "learning_rate": 2.473716781943939e-06, + "loss": 1.0557, + "step": 9688 + }, + { + "epoch": 1.76, + "learning_rate": 2.473662176920277e-06, + "loss": 2.3281, + "step": 9689 + }, + { + "epoch": 1.76, + "learning_rate": 2.4736075718966144e-06, + "loss": 1.4054, + "step": 9690 + }, + { + "epoch": 1.76, + "learning_rate": 2.4735529668729523e-06, + "loss": 1.1242, + "step": 9691 + }, + { + "epoch": 1.76, + "learning_rate": 2.4734983618492903e-06, + "loss": 0.4052, + "step": 9692 + }, + { + "epoch": 1.76, + "learning_rate": 2.473443756825628e-06, + "loss": 2.356, + "step": 9693 + }, + { + "epoch": 1.76, + "learning_rate": 2.4733891518019658e-06, + "loss": 0.1696, + "step": 9694 + }, + { + "epoch": 1.76, + "learning_rate": 2.4733345467783037e-06, + "loss": 1.3392, + "step": 9695 + }, + { + "epoch": 1.76, + "learning_rate": 2.4732799417546416e-06, + "loss": 1.3091, + "step": 9696 + }, + { + "epoch": 1.76, + "learning_rate": 2.4732253367309796e-06, + "loss": 0.1451, + "step": 9697 + }, + { + "epoch": 1.76, + "learning_rate": 2.473170731707317e-06, + "loss": 2.449, + "step": 9698 + }, + { + "epoch": 1.76, + "learning_rate": 2.473116126683655e-06, + "loss": 2.1525, + "step": 9699 + }, + { + "epoch": 1.76, + "learning_rate": 2.473061521659993e-06, + "loss": 2.1681, + "step": 9700 + }, + { + "epoch": 1.76, + "learning_rate": 2.4730069166363305e-06, + "loss": 0.3518, + "step": 9701 + }, + { + "epoch": 1.76, + "learning_rate": 2.4729523116126685e-06, + "loss": 0.333, + "step": 9702 + }, + { + "epoch": 1.76, + "learning_rate": 2.4728977065890064e-06, + "loss": 0.3818, + "step": 9703 + }, + { + "epoch": 1.76, + "learning_rate": 2.472843101565344e-06, + "loss": 0.5683, + "step": 9704 + }, + { + "epoch": 1.76, + "learning_rate": 2.472788496541682e-06, + "loss": 0.4406, + "step": 9705 + }, + { + "epoch": 1.77, + "learning_rate": 2.47273389151802e-06, + "loss": 2.5518, + "step": 9706 + }, + { + "epoch": 1.77, + "learning_rate": 2.4726792864943573e-06, + "loss": 0.7205, + "step": 9707 + }, + { + "epoch": 1.77, + "learning_rate": 2.4726246814706953e-06, + "loss": 0.3989, + "step": 9708 + }, + { + "epoch": 1.77, + "learning_rate": 2.472570076447033e-06, + "loss": 0.6795, + "step": 9709 + }, + { + "epoch": 1.77, + "learning_rate": 2.472515471423371e-06, + "loss": 0.1654, + "step": 9710 + }, + { + "epoch": 1.77, + "learning_rate": 2.472460866399709e-06, + "loss": 0.5105, + "step": 9711 + }, + { + "epoch": 1.77, + "learning_rate": 2.4724062613760466e-06, + "loss": 3.1646, + "step": 9712 + }, + { + "epoch": 1.77, + "learning_rate": 2.4723516563523846e-06, + "loss": 2.0216, + "step": 9713 + }, + { + "epoch": 1.77, + "learning_rate": 2.4722970513287225e-06, + "loss": 1.0806, + "step": 9714 + }, + { + "epoch": 1.77, + "learning_rate": 2.47224244630506e-06, + "loss": 1.0737, + "step": 9715 + }, + { + "epoch": 1.77, + "learning_rate": 2.472187841281398e-06, + "loss": 0.1589, + "step": 9716 + }, + { + "epoch": 1.77, + "learning_rate": 2.472133236257736e-06, + "loss": 0.4109, + "step": 9717 + }, + { + "epoch": 1.77, + "learning_rate": 2.4720786312340734e-06, + "loss": 1.0803, + "step": 9718 + }, + { + "epoch": 1.77, + "learning_rate": 2.4720240262104114e-06, + "loss": 0.5593, + "step": 9719 + }, + { + "epoch": 1.77, + "learning_rate": 2.4719694211867493e-06, + "loss": 1.9734, + "step": 9720 + }, + { + "epoch": 1.77, + "learning_rate": 2.471914816163087e-06, + "loss": 1.4973, + "step": 9721 + }, + { + "epoch": 1.77, + "learning_rate": 2.4718602111394248e-06, + "loss": 0.1301, + "step": 9722 + }, + { + "epoch": 1.77, + "learning_rate": 2.4718056061157627e-06, + "loss": 0.0977, + "step": 9723 + }, + { + "epoch": 1.77, + "learning_rate": 2.4717510010921007e-06, + "loss": 0.2074, + "step": 9724 + }, + { + "epoch": 1.77, + "learning_rate": 2.4716963960684386e-06, + "loss": 1.6706, + "step": 9725 + }, + { + "epoch": 1.77, + "learning_rate": 2.471641791044776e-06, + "loss": 2.6913, + "step": 9726 + }, + { + "epoch": 1.77, + "learning_rate": 2.471587186021114e-06, + "loss": 0.6694, + "step": 9727 + }, + { + "epoch": 1.77, + "learning_rate": 2.471532580997452e-06, + "loss": 1.2351, + "step": 9728 + }, + { + "epoch": 1.77, + "learning_rate": 2.4714779759737895e-06, + "loss": 0.092, + "step": 9729 + }, + { + "epoch": 1.77, + "learning_rate": 2.4714233709501275e-06, + "loss": 1.7999, + "step": 9730 + }, + { + "epoch": 1.77, + "learning_rate": 2.4713687659264654e-06, + "loss": 1.1035, + "step": 9731 + }, + { + "epoch": 1.77, + "learning_rate": 2.471314160902803e-06, + "loss": 0.5151, + "step": 9732 + }, + { + "epoch": 1.77, + "learning_rate": 2.471259555879141e-06, + "loss": 1.5879, + "step": 9733 + }, + { + "epoch": 1.77, + "learning_rate": 2.471204950855479e-06, + "loss": 1.5319, + "step": 9734 + }, + { + "epoch": 1.77, + "learning_rate": 2.4711503458318163e-06, + "loss": 0.0879, + "step": 9735 + }, + { + "epoch": 1.77, + "learning_rate": 2.4710957408081543e-06, + "loss": 2.0756, + "step": 9736 + }, + { + "epoch": 1.77, + "learning_rate": 2.4710411357844922e-06, + "loss": 1.5378, + "step": 9737 + }, + { + "epoch": 1.77, + "learning_rate": 2.47098653076083e-06, + "loss": 0.0717, + "step": 9738 + }, + { + "epoch": 1.77, + "learning_rate": 2.470931925737168e-06, + "loss": 0.8924, + "step": 9739 + }, + { + "epoch": 1.77, + "learning_rate": 2.4708773207135056e-06, + "loss": 1.2663, + "step": 9740 + }, + { + "epoch": 1.77, + "learning_rate": 2.4708227156898436e-06, + "loss": 1.4501, + "step": 9741 + }, + { + "epoch": 1.77, + "learning_rate": 2.4707681106661815e-06, + "loss": 0.3681, + "step": 9742 + }, + { + "epoch": 1.77, + "learning_rate": 2.470713505642519e-06, + "loss": 1.6201, + "step": 9743 + }, + { + "epoch": 1.77, + "learning_rate": 2.470658900618857e-06, + "loss": 1.4837, + "step": 9744 + }, + { + "epoch": 1.77, + "learning_rate": 2.470604295595195e-06, + "loss": 0.6156, + "step": 9745 + }, + { + "epoch": 1.77, + "learning_rate": 2.4705496905715325e-06, + "loss": 1.4773, + "step": 9746 + }, + { + "epoch": 1.77, + "learning_rate": 2.4704950855478704e-06, + "loss": 0.1105, + "step": 9747 + }, + { + "epoch": 1.77, + "learning_rate": 2.4704404805242083e-06, + "loss": 0.5821, + "step": 9748 + }, + { + "epoch": 1.77, + "learning_rate": 2.470385875500546e-06, + "loss": 0.4978, + "step": 9749 + }, + { + "epoch": 1.77, + "learning_rate": 2.470331270476884e-06, + "loss": 2.2763, + "step": 9750 + }, + { + "epoch": 1.77, + "learning_rate": 2.4702766654532217e-06, + "loss": 0.365, + "step": 9751 + }, + { + "epoch": 1.77, + "learning_rate": 2.4702220604295597e-06, + "loss": 0.1408, + "step": 9752 + }, + { + "epoch": 1.77, + "learning_rate": 2.4701674554058976e-06, + "loss": 0.0794, + "step": 9753 + }, + { + "epoch": 1.77, + "learning_rate": 2.470112850382235e-06, + "loss": 1.4883, + "step": 9754 + }, + { + "epoch": 1.77, + "learning_rate": 2.470058245358573e-06, + "loss": 2.721, + "step": 9755 + }, + { + "epoch": 1.77, + "learning_rate": 2.470003640334911e-06, + "loss": 1.202, + "step": 9756 + }, + { + "epoch": 1.77, + "learning_rate": 2.4699490353112486e-06, + "loss": 0.0493, + "step": 9757 + }, + { + "epoch": 1.77, + "learning_rate": 2.4698944302875865e-06, + "loss": 0.0351, + "step": 9758 + }, + { + "epoch": 1.77, + "learning_rate": 2.4698398252639244e-06, + "loss": 0.2352, + "step": 9759 + }, + { + "epoch": 1.77, + "learning_rate": 2.469785220240262e-06, + "loss": 1.7016, + "step": 9760 + }, + { + "epoch": 1.78, + "learning_rate": 2.4697306152166e-06, + "loss": 1.5624, + "step": 9761 + }, + { + "epoch": 1.78, + "learning_rate": 2.469676010192938e-06, + "loss": 0.0601, + "step": 9762 + }, + { + "epoch": 1.78, + "learning_rate": 2.469621405169276e-06, + "loss": 0.3502, + "step": 9763 + }, + { + "epoch": 1.78, + "learning_rate": 2.4695668001456137e-06, + "loss": 0.4654, + "step": 9764 + }, + { + "epoch": 1.78, + "learning_rate": 2.4695121951219513e-06, + "loss": 0.2839, + "step": 9765 + }, + { + "epoch": 1.78, + "learning_rate": 2.469457590098289e-06, + "loss": 0.6658, + "step": 9766 + }, + { + "epoch": 1.78, + "learning_rate": 2.469402985074627e-06, + "loss": 0.8506, + "step": 9767 + }, + { + "epoch": 1.78, + "learning_rate": 2.4693483800509647e-06, + "loss": 0.9385, + "step": 9768 + }, + { + "epoch": 1.78, + "learning_rate": 2.4692937750273026e-06, + "loss": 0.1473, + "step": 9769 + }, + { + "epoch": 1.78, + "learning_rate": 2.4692391700036406e-06, + "loss": 0.1309, + "step": 9770 + }, + { + "epoch": 1.78, + "learning_rate": 2.469184564979978e-06, + "loss": 1.8003, + "step": 9771 + }, + { + "epoch": 1.78, + "learning_rate": 2.469129959956316e-06, + "loss": 0.4206, + "step": 9772 + }, + { + "epoch": 1.78, + "learning_rate": 2.469075354932654e-06, + "loss": 0.4095, + "step": 9773 + }, + { + "epoch": 1.78, + "learning_rate": 2.4690207499089915e-06, + "loss": 1.545, + "step": 9774 + }, + { + "epoch": 1.78, + "learning_rate": 2.4689661448853294e-06, + "loss": 1.1704, + "step": 9775 + }, + { + "epoch": 1.78, + "learning_rate": 2.4689115398616674e-06, + "loss": 0.5552, + "step": 9776 + }, + { + "epoch": 1.78, + "learning_rate": 2.4688569348380053e-06, + "loss": 0.0879, + "step": 9777 + }, + { + "epoch": 1.78, + "learning_rate": 2.4688023298143433e-06, + "loss": 0.2929, + "step": 9778 + }, + { + "epoch": 1.78, + "learning_rate": 2.4687477247906808e-06, + "loss": 0.3327, + "step": 9779 + }, + { + "epoch": 1.78, + "learning_rate": 2.4686931197670187e-06, + "loss": 0.8227, + "step": 9780 + }, + { + "epoch": 1.78, + "learning_rate": 2.4686385147433567e-06, + "loss": 0.1758, + "step": 9781 + }, + { + "epoch": 1.78, + "learning_rate": 2.468583909719694e-06, + "loss": 0.9415, + "step": 9782 + }, + { + "epoch": 1.78, + "learning_rate": 2.468529304696032e-06, + "loss": 0.7165, + "step": 9783 + }, + { + "epoch": 1.78, + "learning_rate": 2.46847469967237e-06, + "loss": 0.5971, + "step": 9784 + }, + { + "epoch": 1.78, + "learning_rate": 2.4684200946487076e-06, + "loss": 1.9062, + "step": 9785 + }, + { + "epoch": 1.78, + "learning_rate": 2.4683654896250455e-06, + "loss": 0.5052, + "step": 9786 + }, + { + "epoch": 1.78, + "learning_rate": 2.4683108846013835e-06, + "loss": 0.2557, + "step": 9787 + }, + { + "epoch": 1.78, + "learning_rate": 2.468256279577721e-06, + "loss": 0.4515, + "step": 9788 + }, + { + "epoch": 1.78, + "learning_rate": 2.468201674554059e-06, + "loss": 0.1179, + "step": 9789 + }, + { + "epoch": 1.78, + "learning_rate": 2.468147069530397e-06, + "loss": 0.5287, + "step": 9790 + }, + { + "epoch": 1.78, + "learning_rate": 2.468092464506735e-06, + "loss": 0.8135, + "step": 9791 + }, + { + "epoch": 1.78, + "learning_rate": 2.4680378594830728e-06, + "loss": 0.7053, + "step": 9792 + }, + { + "epoch": 1.78, + "learning_rate": 2.4679832544594103e-06, + "loss": 0.0863, + "step": 9793 + }, + { + "epoch": 1.78, + "learning_rate": 2.4679286494357482e-06, + "loss": 0.2836, + "step": 9794 + }, + { + "epoch": 1.78, + "learning_rate": 2.467874044412086e-06, + "loss": 0.7833, + "step": 9795 + }, + { + "epoch": 1.78, + "learning_rate": 2.4678194393884237e-06, + "loss": 2.0195, + "step": 9796 + }, + { + "epoch": 1.78, + "learning_rate": 2.4677648343647616e-06, + "loss": 0.4213, + "step": 9797 + }, + { + "epoch": 1.78, + "learning_rate": 2.4677102293410996e-06, + "loss": 1.7482, + "step": 9798 + }, + { + "epoch": 1.78, + "learning_rate": 2.467655624317437e-06, + "loss": 1.338, + "step": 9799 + }, + { + "epoch": 1.78, + "learning_rate": 2.467601019293775e-06, + "loss": 1.3388, + "step": 9800 + }, + { + "epoch": 1.78, + "learning_rate": 2.467546414270113e-06, + "loss": 0.602, + "step": 9801 + }, + { + "epoch": 1.78, + "learning_rate": 2.4674918092464505e-06, + "loss": 0.9967, + "step": 9802 + }, + { + "epoch": 1.78, + "learning_rate": 2.4674372042227885e-06, + "loss": 1.5925, + "step": 9803 + }, + { + "epoch": 1.78, + "learning_rate": 2.4673825991991264e-06, + "loss": 1.8686, + "step": 9804 + }, + { + "epoch": 1.78, + "learning_rate": 2.4673279941754643e-06, + "loss": 0.6579, + "step": 9805 + }, + { + "epoch": 1.78, + "learning_rate": 2.4672733891518023e-06, + "loss": 1.5115, + "step": 9806 + }, + { + "epoch": 1.78, + "learning_rate": 2.46721878412814e-06, + "loss": 2.2621, + "step": 9807 + }, + { + "epoch": 1.78, + "learning_rate": 2.4671641791044777e-06, + "loss": 2.5331, + "step": 9808 + }, + { + "epoch": 1.78, + "learning_rate": 2.4671095740808157e-06, + "loss": 0.6812, + "step": 9809 + }, + { + "epoch": 1.78, + "learning_rate": 2.467054969057153e-06, + "loss": 0.6575, + "step": 9810 + }, + { + "epoch": 1.78, + "learning_rate": 2.467000364033491e-06, + "loss": 0.3515, + "step": 9811 + }, + { + "epoch": 1.78, + "learning_rate": 2.466945759009829e-06, + "loss": 0.3128, + "step": 9812 + }, + { + "epoch": 1.78, + "learning_rate": 2.4668911539861666e-06, + "loss": 0.0652, + "step": 9813 + }, + { + "epoch": 1.78, + "learning_rate": 2.4668365489625046e-06, + "loss": 1.024, + "step": 9814 + }, + { + "epoch": 1.78, + "learning_rate": 2.4667819439388425e-06, + "loss": 0.5471, + "step": 9815 + }, + { + "epoch": 1.79, + "learning_rate": 2.46672733891518e-06, + "loss": 0.8694, + "step": 9816 + }, + { + "epoch": 1.79, + "learning_rate": 2.466672733891518e-06, + "loss": 0.8628, + "step": 9817 + }, + { + "epoch": 1.79, + "learning_rate": 2.466618128867856e-06, + "loss": 0.4967, + "step": 9818 + }, + { + "epoch": 1.79, + "learning_rate": 2.466563523844194e-06, + "loss": 2.0561, + "step": 9819 + }, + { + "epoch": 1.79, + "learning_rate": 2.466508918820532e-06, + "loss": 3.0324, + "step": 9820 + }, + { + "epoch": 1.79, + "learning_rate": 2.4664543137968693e-06, + "loss": 0.6888, + "step": 9821 + }, + { + "epoch": 1.79, + "learning_rate": 2.4663997087732073e-06, + "loss": 1.2562, + "step": 9822 + }, + { + "epoch": 1.79, + "learning_rate": 2.466345103749545e-06, + "loss": 1.2048, + "step": 9823 + }, + { + "epoch": 1.79, + "learning_rate": 2.4662904987258827e-06, + "loss": 0.8631, + "step": 9824 + }, + { + "epoch": 1.79, + "learning_rate": 2.4662358937022207e-06, + "loss": 0.1589, + "step": 9825 + }, + { + "epoch": 1.79, + "learning_rate": 2.4661812886785586e-06, + "loss": 0.2797, + "step": 9826 + }, + { + "epoch": 1.79, + "learning_rate": 2.466126683654896e-06, + "loss": 0.7424, + "step": 9827 + }, + { + "epoch": 1.79, + "learning_rate": 2.466072078631234e-06, + "loss": 0.4185, + "step": 9828 + }, + { + "epoch": 1.79, + "learning_rate": 2.4660174736075716e-06, + "loss": 1.6659, + "step": 9829 + }, + { + "epoch": 1.79, + "learning_rate": 2.4659628685839095e-06, + "loss": 0.4039, + "step": 9830 + }, + { + "epoch": 1.79, + "learning_rate": 2.4659082635602475e-06, + "loss": 1.5196, + "step": 9831 + }, + { + "epoch": 1.79, + "learning_rate": 2.4658536585365854e-06, + "loss": 0.9437, + "step": 9832 + }, + { + "epoch": 1.79, + "learning_rate": 2.4657990535129234e-06, + "loss": 1.4607, + "step": 9833 + }, + { + "epoch": 1.79, + "learning_rate": 2.4657444484892613e-06, + "loss": 0.0373, + "step": 9834 + }, + { + "epoch": 1.79, + "learning_rate": 2.465689843465599e-06, + "loss": 0.9774, + "step": 9835 + }, + { + "epoch": 1.79, + "learning_rate": 2.4656352384419368e-06, + "loss": 1.9277, + "step": 9836 + }, + { + "epoch": 1.79, + "learning_rate": 2.4655806334182747e-06, + "loss": 0.2045, + "step": 9837 + }, + { + "epoch": 1.79, + "learning_rate": 2.4655260283946122e-06, + "loss": 0.3775, + "step": 9838 + }, + { + "epoch": 1.79, + "learning_rate": 2.46547142337095e-06, + "loss": 0.9622, + "step": 9839 + }, + { + "epoch": 1.79, + "learning_rate": 2.465416818347288e-06, + "loss": 1.3275, + "step": 9840 + }, + { + "epoch": 1.79, + "learning_rate": 2.4653622133236256e-06, + "loss": 0.2192, + "step": 9841 + }, + { + "epoch": 1.79, + "learning_rate": 2.4653076082999636e-06, + "loss": 0.9182, + "step": 9842 + }, + { + "epoch": 1.79, + "learning_rate": 2.4652530032763015e-06, + "loss": 0.8975, + "step": 9843 + }, + { + "epoch": 1.79, + "learning_rate": 2.4651983982526395e-06, + "loss": 1.4928, + "step": 9844 + }, + { + "epoch": 1.79, + "learning_rate": 2.4651437932289774e-06, + "loss": 0.377, + "step": 9845 + }, + { + "epoch": 1.79, + "learning_rate": 2.465089188205315e-06, + "loss": 0.7935, + "step": 9846 + }, + { + "epoch": 1.79, + "learning_rate": 2.465034583181653e-06, + "loss": 1.439, + "step": 9847 + }, + { + "epoch": 1.79, + "learning_rate": 2.464979978157991e-06, + "loss": 0.0608, + "step": 9848 + }, + { + "epoch": 1.79, + "learning_rate": 2.4649253731343283e-06, + "loss": 0.4477, + "step": 9849 + }, + { + "epoch": 1.79, + "learning_rate": 2.4648707681106663e-06, + "loss": 3.4948, + "step": 9850 + }, + { + "epoch": 1.79, + "learning_rate": 2.4648161630870042e-06, + "loss": 0.2925, + "step": 9851 + }, + { + "epoch": 1.79, + "learning_rate": 2.4647615580633417e-06, + "loss": 1.482, + "step": 9852 + }, + { + "epoch": 1.79, + "learning_rate": 2.4647069530396797e-06, + "loss": 1.6683, + "step": 9853 + }, + { + "epoch": 1.79, + "learning_rate": 2.4646523480160176e-06, + "loss": 1.493, + "step": 9854 + }, + { + "epoch": 1.79, + "learning_rate": 2.464597742992355e-06, + "loss": 0.0367, + "step": 9855 + }, + { + "epoch": 1.79, + "learning_rate": 2.464543137968693e-06, + "loss": 1.5012, + "step": 9856 + }, + { + "epoch": 1.79, + "learning_rate": 2.464488532945031e-06, + "loss": 4.0769, + "step": 9857 + }, + { + "epoch": 1.79, + "learning_rate": 2.464433927921369e-06, + "loss": 0.977, + "step": 9858 + }, + { + "epoch": 1.79, + "learning_rate": 2.464379322897707e-06, + "loss": 0.7128, + "step": 9859 + }, + { + "epoch": 1.79, + "learning_rate": 2.4643247178740444e-06, + "loss": 0.1688, + "step": 9860 + }, + { + "epoch": 1.79, + "learning_rate": 2.4642701128503824e-06, + "loss": 0.4768, + "step": 9861 + }, + { + "epoch": 1.79, + "learning_rate": 2.4642155078267203e-06, + "loss": 0.8215, + "step": 9862 + }, + { + "epoch": 1.79, + "learning_rate": 2.464160902803058e-06, + "loss": 1.193, + "step": 9863 + }, + { + "epoch": 1.79, + "learning_rate": 2.464106297779396e-06, + "loss": 0.4026, + "step": 9864 + }, + { + "epoch": 1.79, + "learning_rate": 2.4640516927557337e-06, + "loss": 1.3056, + "step": 9865 + }, + { + "epoch": 1.79, + "learning_rate": 2.4639970877320713e-06, + "loss": 2.3076, + "step": 9866 + }, + { + "epoch": 1.79, + "learning_rate": 2.463942482708409e-06, + "loss": 4.2806, + "step": 9867 + }, + { + "epoch": 1.79, + "learning_rate": 2.463887877684747e-06, + "loss": 0.7898, + "step": 9868 + }, + { + "epoch": 1.79, + "learning_rate": 2.4638332726610847e-06, + "loss": 0.9214, + "step": 9869 + }, + { + "epoch": 1.79, + "learning_rate": 2.4637786676374226e-06, + "loss": 1.0972, + "step": 9870 + }, + { + "epoch": 1.8, + "learning_rate": 2.4637240626137606e-06, + "loss": 0.7668, + "step": 9871 + }, + { + "epoch": 1.8, + "learning_rate": 2.4636694575900985e-06, + "loss": 0.5122, + "step": 9872 + }, + { + "epoch": 1.8, + "learning_rate": 2.4636148525664364e-06, + "loss": 0.6742, + "step": 9873 + }, + { + "epoch": 1.8, + "learning_rate": 2.463560247542774e-06, + "loss": 1.7058, + "step": 9874 + }, + { + "epoch": 1.8, + "learning_rate": 2.463505642519112e-06, + "loss": 0.4224, + "step": 9875 + }, + { + "epoch": 1.8, + "learning_rate": 2.46345103749545e-06, + "loss": 0.0506, + "step": 9876 + }, + { + "epoch": 1.8, + "learning_rate": 2.4633964324717874e-06, + "loss": 2.4641, + "step": 9877 + }, + { + "epoch": 1.8, + "learning_rate": 2.4633418274481253e-06, + "loss": 0.3605, + "step": 9878 + }, + { + "epoch": 1.8, + "learning_rate": 2.4632872224244633e-06, + "loss": 2.803, + "step": 9879 + }, + { + "epoch": 1.8, + "learning_rate": 2.4632326174008008e-06, + "loss": 0.9878, + "step": 9880 + }, + { + "epoch": 1.8, + "learning_rate": 2.4631780123771387e-06, + "loss": 0.8149, + "step": 9881 + }, + { + "epoch": 1.8, + "learning_rate": 2.4631234073534767e-06, + "loss": 2.1213, + "step": 9882 + }, + { + "epoch": 1.8, + "learning_rate": 2.463068802329814e-06, + "loss": 1.6596, + "step": 9883 + }, + { + "epoch": 1.8, + "learning_rate": 2.463014197306152e-06, + "loss": 1.1791, + "step": 9884 + }, + { + "epoch": 1.8, + "learning_rate": 2.46295959228249e-06, + "loss": 0.9381, + "step": 9885 + }, + { + "epoch": 1.8, + "learning_rate": 2.462904987258828e-06, + "loss": 0.0774, + "step": 9886 + }, + { + "epoch": 1.8, + "learning_rate": 2.462850382235166e-06, + "loss": 2.8354, + "step": 9887 + }, + { + "epoch": 1.8, + "learning_rate": 2.4627957772115035e-06, + "loss": 0.6533, + "step": 9888 + }, + { + "epoch": 1.8, + "learning_rate": 2.4627411721878414e-06, + "loss": 0.0647, + "step": 9889 + }, + { + "epoch": 1.8, + "learning_rate": 2.4626865671641794e-06, + "loss": 3.1132, + "step": 9890 + }, + { + "epoch": 1.8, + "learning_rate": 2.462631962140517e-06, + "loss": 0.8258, + "step": 9891 + }, + { + "epoch": 1.8, + "learning_rate": 2.462577357116855e-06, + "loss": 0.4456, + "step": 9892 + }, + { + "epoch": 1.8, + "learning_rate": 2.4625227520931928e-06, + "loss": 1.1411, + "step": 9893 + }, + { + "epoch": 1.8, + "learning_rate": 2.4624681470695303e-06, + "loss": 1.9785, + "step": 9894 + }, + { + "epoch": 1.8, + "learning_rate": 2.4624135420458682e-06, + "loss": 1.3924, + "step": 9895 + }, + { + "epoch": 1.8, + "learning_rate": 2.462358937022206e-06, + "loss": 0.2221, + "step": 9896 + }, + { + "epoch": 1.8, + "learning_rate": 2.4623043319985437e-06, + "loss": 0.9012, + "step": 9897 + }, + { + "epoch": 1.8, + "learning_rate": 2.4622497269748816e-06, + "loss": 0.0814, + "step": 9898 + }, + { + "epoch": 1.8, + "learning_rate": 2.4621951219512196e-06, + "loss": 0.65, + "step": 9899 + }, + { + "epoch": 1.8, + "learning_rate": 2.4621405169275575e-06, + "loss": 2.1815, + "step": 9900 + }, + { + "epoch": 1.8, + "learning_rate": 2.4620859119038955e-06, + "loss": 1.5087, + "step": 9901 + }, + { + "epoch": 1.8, + "learning_rate": 2.462031306880233e-06, + "loss": 0.5187, + "step": 9902 + }, + { + "epoch": 1.8, + "learning_rate": 2.461976701856571e-06, + "loss": 2.5048, + "step": 9903 + }, + { + "epoch": 1.8, + "learning_rate": 2.461922096832909e-06, + "loss": 0.0775, + "step": 9904 + }, + { + "epoch": 1.8, + "learning_rate": 2.4618674918092464e-06, + "loss": 2.6608, + "step": 9905 + }, + { + "epoch": 1.8, + "learning_rate": 2.4618128867855843e-06, + "loss": 1.9587, + "step": 9906 + }, + { + "epoch": 1.8, + "learning_rate": 2.4617582817619223e-06, + "loss": 1.5833, + "step": 9907 + }, + { + "epoch": 1.8, + "learning_rate": 2.46170367673826e-06, + "loss": 0.1217, + "step": 9908 + }, + { + "epoch": 1.8, + "learning_rate": 2.4616490717145977e-06, + "loss": 1.4085, + "step": 9909 + }, + { + "epoch": 1.8, + "learning_rate": 2.4615944666909353e-06, + "loss": 3.0296, + "step": 9910 + }, + { + "epoch": 1.8, + "learning_rate": 2.461539861667273e-06, + "loss": 0.6761, + "step": 9911 + }, + { + "epoch": 1.8, + "learning_rate": 2.461485256643611e-06, + "loss": 0.0346, + "step": 9912 + }, + { + "epoch": 1.8, + "learning_rate": 2.461430651619949e-06, + "loss": 0.0783, + "step": 9913 + }, + { + "epoch": 1.8, + "learning_rate": 2.461376046596287e-06, + "loss": 0.7647, + "step": 9914 + }, + { + "epoch": 1.8, + "learning_rate": 2.461321441572625e-06, + "loss": 0.8977, + "step": 9915 + }, + { + "epoch": 1.8, + "learning_rate": 2.4612668365489625e-06, + "loss": 0.3146, + "step": 9916 + }, + { + "epoch": 1.8, + "learning_rate": 2.4612122315253004e-06, + "loss": 0.7474, + "step": 9917 + }, + { + "epoch": 1.8, + "learning_rate": 2.4611576265016384e-06, + "loss": 0.7782, + "step": 9918 + }, + { + "epoch": 1.8, + "learning_rate": 2.461103021477976e-06, + "loss": 0.4941, + "step": 9919 + }, + { + "epoch": 1.8, + "learning_rate": 2.461048416454314e-06, + "loss": 0.0241, + "step": 9920 + }, + { + "epoch": 1.8, + "learning_rate": 2.460993811430652e-06, + "loss": 0.0606, + "step": 9921 + }, + { + "epoch": 1.8, + "learning_rate": 2.4609392064069893e-06, + "loss": 0.6773, + "step": 9922 + }, + { + "epoch": 1.8, + "learning_rate": 2.4608846013833273e-06, + "loss": 0.0129, + "step": 9923 + }, + { + "epoch": 1.8, + "learning_rate": 2.4608299963596648e-06, + "loss": 0.6624, + "step": 9924 + }, + { + "epoch": 1.8, + "learning_rate": 2.4607753913360027e-06, + "loss": 0.4676, + "step": 9925 + }, + { + "epoch": 1.81, + "learning_rate": 2.4607207863123407e-06, + "loss": 1.6675, + "step": 9926 + }, + { + "epoch": 1.81, + "learning_rate": 2.4606661812886786e-06, + "loss": 1.0702, + "step": 9927 + }, + { + "epoch": 1.81, + "learning_rate": 2.4606115762650166e-06, + "loss": 1.6268, + "step": 9928 + }, + { + "epoch": 1.81, + "learning_rate": 2.4605569712413545e-06, + "loss": 0.332, + "step": 9929 + }, + { + "epoch": 1.81, + "learning_rate": 2.460502366217692e-06, + "loss": 2.3615, + "step": 9930 + }, + { + "epoch": 1.81, + "learning_rate": 2.46044776119403e-06, + "loss": 1.382, + "step": 9931 + }, + { + "epoch": 1.81, + "learning_rate": 2.460393156170368e-06, + "loss": 1.3549, + "step": 9932 + }, + { + "epoch": 1.81, + "learning_rate": 2.4603385511467054e-06, + "loss": 0.5486, + "step": 9933 + }, + { + "epoch": 1.81, + "learning_rate": 2.4602839461230434e-06, + "loss": 1.7345, + "step": 9934 + }, + { + "epoch": 1.81, + "learning_rate": 2.4602293410993813e-06, + "loss": 0.1869, + "step": 9935 + }, + { + "epoch": 1.81, + "learning_rate": 2.460174736075719e-06, + "loss": 0.427, + "step": 9936 + }, + { + "epoch": 1.81, + "learning_rate": 2.4601201310520568e-06, + "loss": 2.0416, + "step": 9937 + }, + { + "epoch": 1.81, + "learning_rate": 2.4600655260283947e-06, + "loss": 0.6952, + "step": 9938 + }, + { + "epoch": 1.81, + "learning_rate": 2.4600109210047327e-06, + "loss": 0.765, + "step": 9939 + }, + { + "epoch": 1.81, + "learning_rate": 2.4599563159810706e-06, + "loss": 0.8296, + "step": 9940 + }, + { + "epoch": 1.81, + "learning_rate": 2.459901710957408e-06, + "loss": 0.0576, + "step": 9941 + }, + { + "epoch": 1.81, + "learning_rate": 2.459847105933746e-06, + "loss": 2.0484, + "step": 9942 + }, + { + "epoch": 1.81, + "learning_rate": 2.459792500910084e-06, + "loss": 0.1492, + "step": 9943 + }, + { + "epoch": 1.81, + "learning_rate": 2.4597378958864215e-06, + "loss": 5.7052, + "step": 9944 + }, + { + "epoch": 1.81, + "learning_rate": 2.4596832908627595e-06, + "loss": 2.0313, + "step": 9945 + }, + { + "epoch": 1.81, + "learning_rate": 2.4596286858390974e-06, + "loss": 0.9294, + "step": 9946 + }, + { + "epoch": 1.81, + "learning_rate": 2.459574080815435e-06, + "loss": 1.1859, + "step": 9947 + }, + { + "epoch": 1.81, + "learning_rate": 2.459519475791773e-06, + "loss": 0.8522, + "step": 9948 + }, + { + "epoch": 1.81, + "learning_rate": 2.459464870768111e-06, + "loss": 1.01, + "step": 9949 + }, + { + "epoch": 1.81, + "learning_rate": 2.4594102657444483e-06, + "loss": 0.0254, + "step": 9950 + }, + { + "epoch": 1.81, + "learning_rate": 2.4593556607207863e-06, + "loss": 0.5191, + "step": 9951 + }, + { + "epoch": 1.81, + "learning_rate": 2.4593010556971242e-06, + "loss": 1.1768, + "step": 9952 + }, + { + "epoch": 1.81, + "learning_rate": 2.459246450673462e-06, + "loss": 0.2612, + "step": 9953 + }, + { + "epoch": 1.81, + "learning_rate": 2.4591918456498e-06, + "loss": 0.4315, + "step": 9954 + }, + { + "epoch": 1.81, + "learning_rate": 2.4591372406261376e-06, + "loss": 1.1957, + "step": 9955 + }, + { + "epoch": 1.81, + "learning_rate": 2.4590826356024756e-06, + "loss": 0.5732, + "step": 9956 + }, + { + "epoch": 1.81, + "learning_rate": 2.4590280305788135e-06, + "loss": 2.8361, + "step": 9957 + }, + { + "epoch": 1.81, + "learning_rate": 2.458973425555151e-06, + "loss": 0.2891, + "step": 9958 + }, + { + "epoch": 1.81, + "learning_rate": 2.458918820531489e-06, + "loss": 2.7357, + "step": 9959 + }, + { + "epoch": 1.81, + "learning_rate": 2.458864215507827e-06, + "loss": 0.5324, + "step": 9960 + }, + { + "epoch": 1.81, + "learning_rate": 2.4588096104841645e-06, + "loss": 0.0646, + "step": 9961 + }, + { + "epoch": 1.81, + "learning_rate": 2.4587550054605024e-06, + "loss": 0.5824, + "step": 9962 + }, + { + "epoch": 1.81, + "learning_rate": 2.4587004004368403e-06, + "loss": 1.6922, + "step": 9963 + }, + { + "epoch": 1.81, + "learning_rate": 2.458645795413178e-06, + "loss": 1.2016, + "step": 9964 + }, + { + "epoch": 1.81, + "learning_rate": 2.458591190389516e-06, + "loss": 3.1101, + "step": 9965 + }, + { + "epoch": 1.81, + "learning_rate": 2.4585365853658537e-06, + "loss": 0.9121, + "step": 9966 + }, + { + "epoch": 1.81, + "learning_rate": 2.4584819803421917e-06, + "loss": 0.4801, + "step": 9967 + }, + { + "epoch": 1.81, + "learning_rate": 2.4584273753185296e-06, + "loss": 1.8642, + "step": 9968 + }, + { + "epoch": 1.81, + "learning_rate": 2.458372770294867e-06, + "loss": 0.5465, + "step": 9969 + }, + { + "epoch": 1.81, + "learning_rate": 2.458318165271205e-06, + "loss": 0.3703, + "step": 9970 + }, + { + "epoch": 1.81, + "learning_rate": 2.458263560247543e-06, + "loss": 1.2228, + "step": 9971 + }, + { + "epoch": 1.81, + "learning_rate": 2.4582089552238806e-06, + "loss": 1.4073, + "step": 9972 + }, + { + "epoch": 1.81, + "learning_rate": 2.4581543502002185e-06, + "loss": 1.2731, + "step": 9973 + }, + { + "epoch": 1.81, + "learning_rate": 2.4580997451765564e-06, + "loss": 1.8301, + "step": 9974 + }, + { + "epoch": 1.81, + "learning_rate": 2.458045140152894e-06, + "loss": 0.4853, + "step": 9975 + }, + { + "epoch": 1.81, + "learning_rate": 2.457990535129232e-06, + "loss": 0.4787, + "step": 9976 + }, + { + "epoch": 1.81, + "learning_rate": 2.45793593010557e-06, + "loss": 1.7523, + "step": 9977 + }, + { + "epoch": 1.81, + "learning_rate": 2.4578813250819074e-06, + "loss": 0.4558, + "step": 9978 + }, + { + "epoch": 1.81, + "learning_rate": 2.4578267200582453e-06, + "loss": 1.2597, + "step": 9979 + }, + { + "epoch": 1.81, + "learning_rate": 2.4577721150345833e-06, + "loss": 1.3593, + "step": 9980 + }, + { + "epoch": 1.82, + "learning_rate": 2.457717510010921e-06, + "loss": 0.0245, + "step": 9981 + }, + { + "epoch": 1.82, + "learning_rate": 2.457662904987259e-06, + "loss": 2.2545, + "step": 9982 + }, + { + "epoch": 1.82, + "learning_rate": 2.4576082999635967e-06, + "loss": 0.8635, + "step": 9983 + }, + { + "epoch": 1.82, + "learning_rate": 2.4575536949399346e-06, + "loss": 0.5927, + "step": 9984 + }, + { + "epoch": 1.82, + "learning_rate": 2.4574990899162726e-06, + "loss": 0.8936, + "step": 9985 + }, + { + "epoch": 1.82, + "learning_rate": 2.45744448489261e-06, + "loss": 0.4467, + "step": 9986 + }, + { + "epoch": 1.82, + "learning_rate": 2.457389879868948e-06, + "loss": 1.1774, + "step": 9987 + }, + { + "epoch": 1.82, + "learning_rate": 2.457335274845286e-06, + "loss": 0.8035, + "step": 9988 + }, + { + "epoch": 1.82, + "learning_rate": 2.4572806698216235e-06, + "loss": 0.5933, + "step": 9989 + }, + { + "epoch": 1.82, + "learning_rate": 2.4572260647979614e-06, + "loss": 0.3946, + "step": 9990 + }, + { + "epoch": 1.82, + "learning_rate": 2.457171459774299e-06, + "loss": 2.1947, + "step": 9991 + }, + { + "epoch": 1.82, + "learning_rate": 2.457116854750637e-06, + "loss": 0.2302, + "step": 9992 + }, + { + "epoch": 1.82, + "learning_rate": 2.457062249726975e-06, + "loss": 0.4736, + "step": 9993 + }, + { + "epoch": 1.82, + "learning_rate": 2.4570076447033128e-06, + "loss": 0.6281, + "step": 9994 + }, + { + "epoch": 1.82, + "learning_rate": 2.4569530396796507e-06, + "loss": 2.4683, + "step": 9995 + }, + { + "epoch": 1.82, + "learning_rate": 2.4568984346559887e-06, + "loss": 0.9443, + "step": 9996 + }, + { + "epoch": 1.82, + "learning_rate": 2.456843829632326e-06, + "loss": 0.7201, + "step": 9997 + }, + { + "epoch": 1.82, + "learning_rate": 2.456789224608664e-06, + "loss": 0.4598, + "step": 9998 + }, + { + "epoch": 1.82, + "learning_rate": 2.456734619585002e-06, + "loss": 0.6557, + "step": 9999 + }, + { + "epoch": 1.82, + "learning_rate": 2.4566800145613396e-06, + "loss": 0.5515, + "step": 10000 + }, + { + "epoch": 1.82, + "learning_rate": 2.4566254095376775e-06, + "loss": 1.7146, + "step": 10001 + }, + { + "epoch": 1.82, + "learning_rate": 2.4565708045140155e-06, + "loss": 1.0866, + "step": 10002 + }, + { + "epoch": 1.82, + "learning_rate": 2.456516199490353e-06, + "loss": 2.2251, + "step": 10003 + }, + { + "epoch": 1.82, + "learning_rate": 2.456461594466691e-06, + "loss": 0.6097, + "step": 10004 + }, + { + "epoch": 1.82, + "learning_rate": 2.4564069894430285e-06, + "loss": 1.1713, + "step": 10005 + }, + { + "epoch": 1.82, + "learning_rate": 2.4563523844193664e-06, + "loss": 1.3576, + "step": 10006 + }, + { + "epoch": 1.82, + "learning_rate": 2.4562977793957043e-06, + "loss": 0.3442, + "step": 10007 + }, + { + "epoch": 1.82, + "learning_rate": 2.4562431743720423e-06, + "loss": 0.9401, + "step": 10008 + }, + { + "epoch": 1.82, + "learning_rate": 2.4561885693483802e-06, + "loss": 0.3863, + "step": 10009 + }, + { + "epoch": 1.82, + "learning_rate": 2.456133964324718e-06, + "loss": 2.5967, + "step": 10010 + }, + { + "epoch": 1.82, + "learning_rate": 2.4560793593010557e-06, + "loss": 0.013, + "step": 10011 + }, + { + "epoch": 1.82, + "learning_rate": 2.4560247542773936e-06, + "loss": 0.8325, + "step": 10012 + }, + { + "epoch": 1.82, + "learning_rate": 2.4559701492537316e-06, + "loss": 1.1724, + "step": 10013 + }, + { + "epoch": 1.82, + "learning_rate": 2.455915544230069e-06, + "loss": 0.253, + "step": 10014 + }, + { + "epoch": 1.82, + "learning_rate": 2.455860939206407e-06, + "loss": 1.3557, + "step": 10015 + }, + { + "epoch": 1.82, + "learning_rate": 2.455806334182745e-06, + "loss": 0.1622, + "step": 10016 + }, + { + "epoch": 1.82, + "learning_rate": 2.4557517291590825e-06, + "loss": 2.2429, + "step": 10017 + }, + { + "epoch": 1.82, + "learning_rate": 2.4556971241354204e-06, + "loss": 1.5008, + "step": 10018 + }, + { + "epoch": 1.82, + "learning_rate": 2.4556425191117584e-06, + "loss": 0.0463, + "step": 10019 + }, + { + "epoch": 1.82, + "learning_rate": 2.4555879140880963e-06, + "loss": 0.0973, + "step": 10020 + }, + { + "epoch": 1.82, + "learning_rate": 2.4555333090644343e-06, + "loss": 1.0367, + "step": 10021 + }, + { + "epoch": 1.82, + "learning_rate": 2.455478704040772e-06, + "loss": 2.1314, + "step": 10022 + }, + { + "epoch": 1.82, + "learning_rate": 2.4554240990171097e-06, + "loss": 2.7237, + "step": 10023 + }, + { + "epoch": 1.82, + "learning_rate": 2.4553694939934477e-06, + "loss": 1.8006, + "step": 10024 + }, + { + "epoch": 1.82, + "learning_rate": 2.455314888969785e-06, + "loss": 1.3383, + "step": 10025 + }, + { + "epoch": 1.82, + "learning_rate": 2.455260283946123e-06, + "loss": 0.6054, + "step": 10026 + }, + { + "epoch": 1.82, + "learning_rate": 2.455205678922461e-06, + "loss": 1.3967, + "step": 10027 + }, + { + "epoch": 1.82, + "learning_rate": 2.4551510738987986e-06, + "loss": 0.4164, + "step": 10028 + }, + { + "epoch": 1.82, + "learning_rate": 2.4550964688751366e-06, + "loss": 0.4988, + "step": 10029 + }, + { + "epoch": 1.82, + "learning_rate": 2.4550418638514745e-06, + "loss": 0.0649, + "step": 10030 + }, + { + "epoch": 1.82, + "learning_rate": 2.454987258827812e-06, + "loss": 1.5821, + "step": 10031 + }, + { + "epoch": 1.82, + "learning_rate": 2.45493265380415e-06, + "loss": 0.3936, + "step": 10032 + }, + { + "epoch": 1.82, + "learning_rate": 2.454878048780488e-06, + "loss": 1.0466, + "step": 10033 + }, + { + "epoch": 1.82, + "learning_rate": 2.454823443756826e-06, + "loss": 0.0856, + "step": 10034 + }, + { + "epoch": 1.82, + "learning_rate": 2.454768838733164e-06, + "loss": 0.8885, + "step": 10035 + }, + { + "epoch": 1.83, + "learning_rate": 2.4547142337095013e-06, + "loss": 0.5329, + "step": 10036 + }, + { + "epoch": 1.83, + "learning_rate": 2.4546596286858393e-06, + "loss": 0.5571, + "step": 10037 + }, + { + "epoch": 1.83, + "learning_rate": 2.454605023662177e-06, + "loss": 0.5833, + "step": 10038 + }, + { + "epoch": 1.83, + "learning_rate": 2.4545504186385147e-06, + "loss": 0.3523, + "step": 10039 + }, + { + "epoch": 1.83, + "learning_rate": 2.4544958136148527e-06, + "loss": 0.9919, + "step": 10040 + }, + { + "epoch": 1.83, + "learning_rate": 2.4544412085911906e-06, + "loss": 1.2619, + "step": 10041 + }, + { + "epoch": 1.83, + "learning_rate": 2.454386603567528e-06, + "loss": 0.8563, + "step": 10042 + }, + { + "epoch": 1.83, + "learning_rate": 2.454331998543866e-06, + "loss": 1.8299, + "step": 10043 + }, + { + "epoch": 1.83, + "learning_rate": 2.454277393520204e-06, + "loss": 0.0633, + "step": 10044 + }, + { + "epoch": 1.83, + "learning_rate": 2.4542227884965415e-06, + "loss": 3.0918, + "step": 10045 + }, + { + "epoch": 1.83, + "learning_rate": 2.4541681834728795e-06, + "loss": 1.7929, + "step": 10046 + }, + { + "epoch": 1.83, + "learning_rate": 2.4541135784492174e-06, + "loss": 1.0948, + "step": 10047 + }, + { + "epoch": 1.83, + "learning_rate": 2.4540589734255554e-06, + "loss": 0.0502, + "step": 10048 + }, + { + "epoch": 1.83, + "learning_rate": 2.4540043684018933e-06, + "loss": 0.6953, + "step": 10049 + }, + { + "epoch": 1.83, + "learning_rate": 2.453949763378231e-06, + "loss": 0.4165, + "step": 10050 + }, + { + "epoch": 1.83, + "learning_rate": 2.4538951583545688e-06, + "loss": 1.8546, + "step": 10051 + }, + { + "epoch": 1.83, + "learning_rate": 2.4538405533309067e-06, + "loss": 1.322, + "step": 10052 + }, + { + "epoch": 1.83, + "learning_rate": 2.4537859483072442e-06, + "loss": 2.3711, + "step": 10053 + }, + { + "epoch": 1.83, + "learning_rate": 2.453731343283582e-06, + "loss": 0.5738, + "step": 10054 + }, + { + "epoch": 1.83, + "learning_rate": 2.45367673825992e-06, + "loss": 1.6017, + "step": 10055 + }, + { + "epoch": 1.83, + "learning_rate": 2.4536221332362576e-06, + "loss": 0.6247, + "step": 10056 + }, + { + "epoch": 1.83, + "learning_rate": 2.4535675282125956e-06, + "loss": 0.7106, + "step": 10057 + }, + { + "epoch": 1.83, + "learning_rate": 2.4535129231889335e-06, + "loss": 0.8191, + "step": 10058 + }, + { + "epoch": 1.83, + "learning_rate": 2.453458318165271e-06, + "loss": 1.5629, + "step": 10059 + }, + { + "epoch": 1.83, + "learning_rate": 2.453403713141609e-06, + "loss": 3.0731, + "step": 10060 + }, + { + "epoch": 1.83, + "learning_rate": 2.453349108117947e-06, + "loss": 0.0712, + "step": 10061 + }, + { + "epoch": 1.83, + "learning_rate": 2.453294503094285e-06, + "loss": 2.182, + "step": 10062 + }, + { + "epoch": 1.83, + "learning_rate": 2.453239898070623e-06, + "loss": 1.8756, + "step": 10063 + }, + { + "epoch": 1.83, + "learning_rate": 2.4531852930469603e-06, + "loss": 0.056, + "step": 10064 + }, + { + "epoch": 1.83, + "learning_rate": 2.4531306880232983e-06, + "loss": 1.8986, + "step": 10065 + }, + { + "epoch": 1.83, + "learning_rate": 2.4530760829996362e-06, + "loss": 1.8595, + "step": 10066 + }, + { + "epoch": 1.83, + "learning_rate": 2.4530214779759737e-06, + "loss": 0.2704, + "step": 10067 + }, + { + "epoch": 1.83, + "learning_rate": 2.4529668729523117e-06, + "loss": 2.4221, + "step": 10068 + }, + { + "epoch": 1.83, + "learning_rate": 2.4529122679286496e-06, + "loss": 0.0484, + "step": 10069 + }, + { + "epoch": 1.83, + "learning_rate": 2.452857662904987e-06, + "loss": 0.1864, + "step": 10070 + }, + { + "epoch": 1.83, + "learning_rate": 2.452803057881325e-06, + "loss": 0.1036, + "step": 10071 + }, + { + "epoch": 1.83, + "learning_rate": 2.4527484528576626e-06, + "loss": 1.9015, + "step": 10072 + }, + { + "epoch": 1.83, + "learning_rate": 2.4526938478340006e-06, + "loss": 0.4136, + "step": 10073 + }, + { + "epoch": 1.83, + "learning_rate": 2.4526392428103385e-06, + "loss": 0.5691, + "step": 10074 + }, + { + "epoch": 1.83, + "learning_rate": 2.4525846377866764e-06, + "loss": 1.0724, + "step": 10075 + }, + { + "epoch": 1.83, + "learning_rate": 2.4525300327630144e-06, + "loss": 0.5778, + "step": 10076 + }, + { + "epoch": 1.83, + "learning_rate": 2.4524754277393523e-06, + "loss": 2.5324, + "step": 10077 + }, + { + "epoch": 1.83, + "learning_rate": 2.45242082271569e-06, + "loss": 1.6572, + "step": 10078 + }, + { + "epoch": 1.83, + "learning_rate": 2.452366217692028e-06, + "loss": 0.0447, + "step": 10079 + }, + { + "epoch": 1.83, + "learning_rate": 2.4523116126683657e-06, + "loss": 0.8829, + "step": 10080 + }, + { + "epoch": 1.83, + "learning_rate": 2.4522570076447033e-06, + "loss": 0.7798, + "step": 10081 + }, + { + "epoch": 1.83, + "learning_rate": 2.452202402621041e-06, + "loss": 2.2849, + "step": 10082 + }, + { + "epoch": 1.83, + "learning_rate": 2.452147797597379e-06, + "loss": 0.4374, + "step": 10083 + }, + { + "epoch": 1.83, + "learning_rate": 2.4520931925737167e-06, + "loss": 0.6284, + "step": 10084 + }, + { + "epoch": 1.83, + "learning_rate": 2.4520385875500546e-06, + "loss": 0.345, + "step": 10085 + }, + { + "epoch": 1.83, + "learning_rate": 2.451983982526392e-06, + "loss": 1.1025, + "step": 10086 + }, + { + "epoch": 1.83, + "learning_rate": 2.45192937750273e-06, + "loss": 0.6506, + "step": 10087 + }, + { + "epoch": 1.83, + "learning_rate": 2.451874772479068e-06, + "loss": 1.3517, + "step": 10088 + }, + { + "epoch": 1.83, + "learning_rate": 2.451820167455406e-06, + "loss": 1.7474, + "step": 10089 + }, + { + "epoch": 1.83, + "learning_rate": 2.451765562431744e-06, + "loss": 0.3401, + "step": 10090 + }, + { + "epoch": 1.84, + "learning_rate": 2.451710957408082e-06, + "loss": 0.577, + "step": 10091 + }, + { + "epoch": 1.84, + "learning_rate": 2.4516563523844194e-06, + "loss": 1.7814, + "step": 10092 + }, + { + "epoch": 1.84, + "learning_rate": 2.4516017473607573e-06, + "loss": 0.6734, + "step": 10093 + }, + { + "epoch": 1.84, + "learning_rate": 2.4515471423370953e-06, + "loss": 0.8475, + "step": 10094 + }, + { + "epoch": 1.84, + "learning_rate": 2.4514925373134328e-06, + "loss": 2.6334, + "step": 10095 + }, + { + "epoch": 1.84, + "learning_rate": 2.4514379322897707e-06, + "loss": 1.7348, + "step": 10096 + }, + { + "epoch": 1.84, + "learning_rate": 2.4513833272661087e-06, + "loss": 0.0994, + "step": 10097 + }, + { + "epoch": 1.84, + "learning_rate": 2.451328722242446e-06, + "loss": 0.855, + "step": 10098 + }, + { + "epoch": 1.84, + "learning_rate": 2.451274117218784e-06, + "loss": 1.1299, + "step": 10099 + }, + { + "epoch": 1.84, + "learning_rate": 2.451219512195122e-06, + "loss": 0.3905, + "step": 10100 + }, + { + "epoch": 1.84, + "learning_rate": 2.45116490717146e-06, + "loss": 0.8263, + "step": 10101 + }, + { + "epoch": 1.84, + "learning_rate": 2.451110302147798e-06, + "loss": 0.3369, + "step": 10102 + }, + { + "epoch": 1.84, + "learning_rate": 2.4510556971241355e-06, + "loss": 0.6298, + "step": 10103 + }, + { + "epoch": 1.84, + "learning_rate": 2.4510010921004734e-06, + "loss": 1.1715, + "step": 10104 + }, + { + "epoch": 1.84, + "learning_rate": 2.4509464870768114e-06, + "loss": 0.6832, + "step": 10105 + }, + { + "epoch": 1.84, + "learning_rate": 2.450891882053149e-06, + "loss": 2.5255, + "step": 10106 + }, + { + "epoch": 1.84, + "learning_rate": 2.450837277029487e-06, + "loss": 2.2192, + "step": 10107 + }, + { + "epoch": 1.84, + "learning_rate": 2.4507826720058248e-06, + "loss": 0.1465, + "step": 10108 + }, + { + "epoch": 1.84, + "learning_rate": 2.4507280669821623e-06, + "loss": 0.4992, + "step": 10109 + }, + { + "epoch": 1.84, + "learning_rate": 2.4506734619585002e-06, + "loss": 1.6839, + "step": 10110 + }, + { + "epoch": 1.84, + "learning_rate": 2.450618856934838e-06, + "loss": 1.7257, + "step": 10111 + }, + { + "epoch": 1.84, + "learning_rate": 2.4505642519111757e-06, + "loss": 1.0911, + "step": 10112 + }, + { + "epoch": 1.84, + "learning_rate": 2.4505096468875136e-06, + "loss": 1.7897, + "step": 10113 + }, + { + "epoch": 1.84, + "learning_rate": 2.4504550418638516e-06, + "loss": 0.5805, + "step": 10114 + }, + { + "epoch": 1.84, + "learning_rate": 2.4504004368401895e-06, + "loss": 1.3017, + "step": 10115 + }, + { + "epoch": 1.84, + "learning_rate": 2.4503458318165275e-06, + "loss": 3.3635, + "step": 10116 + }, + { + "epoch": 1.84, + "learning_rate": 2.450291226792865e-06, + "loss": 1.6293, + "step": 10117 + }, + { + "epoch": 1.84, + "learning_rate": 2.450236621769203e-06, + "loss": 1.2209, + "step": 10118 + }, + { + "epoch": 1.84, + "learning_rate": 2.450182016745541e-06, + "loss": 1.7233, + "step": 10119 + }, + { + "epoch": 1.84, + "learning_rate": 2.4501274117218784e-06, + "loss": 1.4442, + "step": 10120 + }, + { + "epoch": 1.84, + "learning_rate": 2.4500728066982163e-06, + "loss": 2.363, + "step": 10121 + }, + { + "epoch": 1.84, + "learning_rate": 2.4500182016745543e-06, + "loss": 1.523, + "step": 10122 + }, + { + "epoch": 1.84, + "learning_rate": 2.449963596650892e-06, + "loss": 0.1918, + "step": 10123 + }, + { + "epoch": 1.84, + "learning_rate": 2.4499089916272297e-06, + "loss": 3.3931, + "step": 10124 + }, + { + "epoch": 1.84, + "learning_rate": 2.4498543866035677e-06, + "loss": 2.2805, + "step": 10125 + }, + { + "epoch": 1.84, + "learning_rate": 2.449799781579905e-06, + "loss": 1.2626, + "step": 10126 + }, + { + "epoch": 1.84, + "learning_rate": 2.449745176556243e-06, + "loss": 1.3976, + "step": 10127 + }, + { + "epoch": 1.84, + "learning_rate": 2.449690571532581e-06, + "loss": 1.9452, + "step": 10128 + }, + { + "epoch": 1.84, + "learning_rate": 2.449635966508919e-06, + "loss": 1.1646, + "step": 10129 + }, + { + "epoch": 1.84, + "learning_rate": 2.449581361485257e-06, + "loss": 0.0288, + "step": 10130 + }, + { + "epoch": 1.84, + "learning_rate": 2.4495267564615945e-06, + "loss": 0.4062, + "step": 10131 + }, + { + "epoch": 1.84, + "learning_rate": 2.4494721514379324e-06, + "loss": 0.3752, + "step": 10132 + }, + { + "epoch": 1.84, + "learning_rate": 2.4494175464142704e-06, + "loss": 0.226, + "step": 10133 + }, + { + "epoch": 1.84, + "learning_rate": 2.449362941390608e-06, + "loss": 1.1481, + "step": 10134 + }, + { + "epoch": 1.84, + "learning_rate": 2.449308336366946e-06, + "loss": 1.6385, + "step": 10135 + }, + { + "epoch": 1.84, + "learning_rate": 2.449253731343284e-06, + "loss": 1.9715, + "step": 10136 + }, + { + "epoch": 1.84, + "learning_rate": 2.4491991263196213e-06, + "loss": 1.7539, + "step": 10137 + }, + { + "epoch": 1.84, + "learning_rate": 2.4491445212959593e-06, + "loss": 0.5375, + "step": 10138 + }, + { + "epoch": 1.84, + "learning_rate": 2.4490899162722968e-06, + "loss": 0.8249, + "step": 10139 + }, + { + "epoch": 1.84, + "learning_rate": 2.4490353112486347e-06, + "loss": 0.0538, + "step": 10140 + }, + { + "epoch": 1.84, + "learning_rate": 2.4489807062249727e-06, + "loss": 2.4854, + "step": 10141 + }, + { + "epoch": 1.84, + "learning_rate": 2.4489261012013106e-06, + "loss": 0.707, + "step": 10142 + }, + { + "epoch": 1.84, + "learning_rate": 2.4488714961776485e-06, + "loss": 0.0466, + "step": 10143 + }, + { + "epoch": 1.84, + "learning_rate": 2.4488168911539865e-06, + "loss": 0.3746, + "step": 10144 + }, + { + "epoch": 1.84, + "learning_rate": 2.448762286130324e-06, + "loss": 1.3525, + "step": 10145 + }, + { + "epoch": 1.85, + "learning_rate": 2.448707681106662e-06, + "loss": 0.271, + "step": 10146 + }, + { + "epoch": 1.85, + "learning_rate": 2.448653076083e-06, + "loss": 0.9089, + "step": 10147 + }, + { + "epoch": 1.85, + "learning_rate": 2.4485984710593374e-06, + "loss": 0.5516, + "step": 10148 + }, + { + "epoch": 1.85, + "learning_rate": 2.4485438660356754e-06, + "loss": 2.0096, + "step": 10149 + }, + { + "epoch": 1.85, + "learning_rate": 2.4484892610120133e-06, + "loss": 1.5421, + "step": 10150 + }, + { + "epoch": 1.85, + "learning_rate": 2.448434655988351e-06, + "loss": 0.6832, + "step": 10151 + }, + { + "epoch": 1.85, + "learning_rate": 2.4483800509646888e-06, + "loss": 0.1836, + "step": 10152 + }, + { + "epoch": 1.85, + "learning_rate": 2.4483254459410263e-06, + "loss": 1.0856, + "step": 10153 + }, + { + "epoch": 1.85, + "learning_rate": 2.4482708409173642e-06, + "loss": 1.2229, + "step": 10154 + }, + { + "epoch": 1.85, + "learning_rate": 2.448216235893702e-06, + "loss": 0.3658, + "step": 10155 + }, + { + "epoch": 1.85, + "learning_rate": 2.44816163087004e-06, + "loss": 0.421, + "step": 10156 + }, + { + "epoch": 1.85, + "learning_rate": 2.448107025846378e-06, + "loss": 0.6112, + "step": 10157 + }, + { + "epoch": 1.85, + "learning_rate": 2.448052420822716e-06, + "loss": 0.6033, + "step": 10158 + }, + { + "epoch": 1.85, + "learning_rate": 2.4479978157990535e-06, + "loss": 2.3547, + "step": 10159 + }, + { + "epoch": 1.85, + "learning_rate": 2.4479432107753915e-06, + "loss": 1.3541, + "step": 10160 + }, + { + "epoch": 1.85, + "learning_rate": 2.4478886057517294e-06, + "loss": 0.0199, + "step": 10161 + }, + { + "epoch": 1.85, + "learning_rate": 2.447834000728067e-06, + "loss": 0.0986, + "step": 10162 + }, + { + "epoch": 1.85, + "learning_rate": 2.447779395704405e-06, + "loss": 0.0226, + "step": 10163 + }, + { + "epoch": 1.85, + "learning_rate": 2.447724790680743e-06, + "loss": 0.8847, + "step": 10164 + }, + { + "epoch": 1.85, + "learning_rate": 2.4476701856570803e-06, + "loss": 0.9689, + "step": 10165 + }, + { + "epoch": 1.85, + "learning_rate": 2.4476155806334183e-06, + "loss": 1.0191, + "step": 10166 + }, + { + "epoch": 1.85, + "learning_rate": 2.447560975609756e-06, + "loss": 1.2494, + "step": 10167 + }, + { + "epoch": 1.85, + "learning_rate": 2.4475063705860937e-06, + "loss": 0.372, + "step": 10168 + }, + { + "epoch": 1.85, + "learning_rate": 2.4474517655624317e-06, + "loss": 0.3827, + "step": 10169 + }, + { + "epoch": 1.85, + "learning_rate": 2.4473971605387696e-06, + "loss": 0.3293, + "step": 10170 + }, + { + "epoch": 1.85, + "learning_rate": 2.4473425555151076e-06, + "loss": 0.2784, + "step": 10171 + }, + { + "epoch": 1.85, + "learning_rate": 2.4472879504914455e-06, + "loss": 2.1203, + "step": 10172 + }, + { + "epoch": 1.85, + "learning_rate": 2.447233345467783e-06, + "loss": 0.0444, + "step": 10173 + }, + { + "epoch": 1.85, + "learning_rate": 2.447178740444121e-06, + "loss": 2.3455, + "step": 10174 + }, + { + "epoch": 1.85, + "learning_rate": 2.447124135420459e-06, + "loss": 1.8933, + "step": 10175 + }, + { + "epoch": 1.85, + "learning_rate": 2.4470695303967964e-06, + "loss": 0.549, + "step": 10176 + }, + { + "epoch": 1.85, + "learning_rate": 2.4470149253731344e-06, + "loss": 2.2367, + "step": 10177 + }, + { + "epoch": 1.85, + "learning_rate": 2.4469603203494723e-06, + "loss": 0.4392, + "step": 10178 + }, + { + "epoch": 1.85, + "learning_rate": 2.44690571532581e-06, + "loss": 0.6615, + "step": 10179 + }, + { + "epoch": 1.85, + "learning_rate": 2.446851110302148e-06, + "loss": 1.5511, + "step": 10180 + }, + { + "epoch": 1.85, + "learning_rate": 2.4467965052784857e-06, + "loss": 0.2183, + "step": 10181 + }, + { + "epoch": 1.85, + "learning_rate": 2.4467419002548237e-06, + "loss": 0.0437, + "step": 10182 + }, + { + "epoch": 1.85, + "learning_rate": 2.446687295231161e-06, + "loss": 1.1961, + "step": 10183 + }, + { + "epoch": 1.85, + "learning_rate": 2.446632690207499e-06, + "loss": 1.1426, + "step": 10184 + }, + { + "epoch": 1.85, + "learning_rate": 2.446578085183837e-06, + "loss": 1.6736, + "step": 10185 + }, + { + "epoch": 1.85, + "learning_rate": 2.446523480160175e-06, + "loss": 0.4869, + "step": 10186 + }, + { + "epoch": 1.85, + "learning_rate": 2.4464688751365126e-06, + "loss": 0.0568, + "step": 10187 + }, + { + "epoch": 1.85, + "learning_rate": 2.4464142701128505e-06, + "loss": 0.4801, + "step": 10188 + }, + { + "epoch": 1.85, + "learning_rate": 2.4463596650891884e-06, + "loss": 0.8956, + "step": 10189 + }, + { + "epoch": 1.85, + "learning_rate": 2.446305060065526e-06, + "loss": 0.6581, + "step": 10190 + }, + { + "epoch": 1.85, + "learning_rate": 2.446250455041864e-06, + "loss": 0.3174, + "step": 10191 + }, + { + "epoch": 1.85, + "learning_rate": 2.446195850018202e-06, + "loss": 1.6043, + "step": 10192 + }, + { + "epoch": 1.85, + "learning_rate": 2.4461412449945394e-06, + "loss": 1.552, + "step": 10193 + }, + { + "epoch": 1.85, + "learning_rate": 2.4460866399708773e-06, + "loss": 1.0088, + "step": 10194 + }, + { + "epoch": 1.85, + "learning_rate": 2.4460320349472153e-06, + "loss": 2.6326, + "step": 10195 + }, + { + "epoch": 1.85, + "learning_rate": 2.445977429923553e-06, + "loss": 0.1626, + "step": 10196 + }, + { + "epoch": 1.85, + "learning_rate": 2.445922824899891e-06, + "loss": 0.7439, + "step": 10197 + }, + { + "epoch": 1.85, + "learning_rate": 2.4458682198762287e-06, + "loss": 0.3089, + "step": 10198 + }, + { + "epoch": 1.85, + "learning_rate": 2.4458136148525666e-06, + "loss": 0.7124, + "step": 10199 + }, + { + "epoch": 1.85, + "learning_rate": 2.4457590098289045e-06, + "loss": 0.0418, + "step": 10200 + }, + { + "epoch": 1.86, + "learning_rate": 2.445704404805242e-06, + "loss": 0.419, + "step": 10201 + }, + { + "epoch": 1.86, + "learning_rate": 2.44564979978158e-06, + "loss": 0.8078, + "step": 10202 + }, + { + "epoch": 1.86, + "learning_rate": 2.445595194757918e-06, + "loss": 0.3902, + "step": 10203 + }, + { + "epoch": 1.86, + "learning_rate": 2.4455405897342555e-06, + "loss": 0.8797, + "step": 10204 + }, + { + "epoch": 1.86, + "learning_rate": 2.4454859847105934e-06, + "loss": 0.3863, + "step": 10205 + }, + { + "epoch": 1.86, + "learning_rate": 2.4454313796869314e-06, + "loss": 0.9087, + "step": 10206 + }, + { + "epoch": 1.86, + "learning_rate": 2.445376774663269e-06, + "loss": 1.7773, + "step": 10207 + }, + { + "epoch": 1.86, + "learning_rate": 2.445322169639607e-06, + "loss": 0.7832, + "step": 10208 + }, + { + "epoch": 1.86, + "learning_rate": 2.4452675646159448e-06, + "loss": 1.3548, + "step": 10209 + }, + { + "epoch": 1.86, + "learning_rate": 2.4452129595922827e-06, + "loss": 0.843, + "step": 10210 + }, + { + "epoch": 1.86, + "learning_rate": 2.4451583545686207e-06, + "loss": 0.7007, + "step": 10211 + }, + { + "epoch": 1.86, + "learning_rate": 2.445103749544958e-06, + "loss": 1.1526, + "step": 10212 + }, + { + "epoch": 1.86, + "learning_rate": 2.445049144521296e-06, + "loss": 0.0912, + "step": 10213 + }, + { + "epoch": 1.86, + "learning_rate": 2.444994539497634e-06, + "loss": 0.9014, + "step": 10214 + }, + { + "epoch": 1.86, + "learning_rate": 2.4449399344739716e-06, + "loss": 0.9514, + "step": 10215 + }, + { + "epoch": 1.86, + "learning_rate": 2.4448853294503095e-06, + "loss": 0.0245, + "step": 10216 + }, + { + "epoch": 1.86, + "learning_rate": 2.4448307244266475e-06, + "loss": 0.987, + "step": 10217 + }, + { + "epoch": 1.86, + "learning_rate": 2.444776119402985e-06, + "loss": 0.0284, + "step": 10218 + }, + { + "epoch": 1.86, + "learning_rate": 2.444721514379323e-06, + "loss": 1.5786, + "step": 10219 + }, + { + "epoch": 1.86, + "learning_rate": 2.4446669093556604e-06, + "loss": 0.1375, + "step": 10220 + }, + { + "epoch": 1.86, + "learning_rate": 2.4446123043319984e-06, + "loss": 1.8654, + "step": 10221 + }, + { + "epoch": 1.86, + "learning_rate": 2.4445576993083363e-06, + "loss": 0.9981, + "step": 10222 + }, + { + "epoch": 1.86, + "learning_rate": 2.4445030942846743e-06, + "loss": 0.771, + "step": 10223 + }, + { + "epoch": 1.86, + "learning_rate": 2.4444484892610122e-06, + "loss": 0.6253, + "step": 10224 + }, + { + "epoch": 1.86, + "learning_rate": 2.44439388423735e-06, + "loss": 0.1725, + "step": 10225 + }, + { + "epoch": 1.86, + "learning_rate": 2.4443392792136877e-06, + "loss": 0.5644, + "step": 10226 + }, + { + "epoch": 1.86, + "learning_rate": 2.4442846741900256e-06, + "loss": 1.7186, + "step": 10227 + }, + { + "epoch": 1.86, + "learning_rate": 2.4442300691663636e-06, + "loss": 0.4482, + "step": 10228 + }, + { + "epoch": 1.86, + "learning_rate": 2.444175464142701e-06, + "loss": 0.3084, + "step": 10229 + }, + { + "epoch": 1.86, + "learning_rate": 2.444120859119039e-06, + "loss": 0.7431, + "step": 10230 + }, + { + "epoch": 1.86, + "learning_rate": 2.444066254095377e-06, + "loss": 0.7343, + "step": 10231 + }, + { + "epoch": 1.86, + "learning_rate": 2.4440116490717145e-06, + "loss": 1.0313, + "step": 10232 + }, + { + "epoch": 1.86, + "learning_rate": 2.4439570440480524e-06, + "loss": 0.861, + "step": 10233 + }, + { + "epoch": 1.86, + "learning_rate": 2.44390243902439e-06, + "loss": 0.032, + "step": 10234 + }, + { + "epoch": 1.86, + "learning_rate": 2.443847834000728e-06, + "loss": 0.6542, + "step": 10235 + }, + { + "epoch": 1.86, + "learning_rate": 2.443793228977066e-06, + "loss": 1.5918, + "step": 10236 + }, + { + "epoch": 1.86, + "learning_rate": 2.443738623953404e-06, + "loss": 0.3572, + "step": 10237 + }, + { + "epoch": 1.86, + "learning_rate": 2.4436840189297417e-06, + "loss": 1.8112, + "step": 10238 + }, + { + "epoch": 1.86, + "learning_rate": 2.4436294139060797e-06, + "loss": 1.3512, + "step": 10239 + }, + { + "epoch": 1.86, + "learning_rate": 2.443574808882417e-06, + "loss": 1.1903, + "step": 10240 + }, + { + "epoch": 1.86, + "learning_rate": 2.443520203858755e-06, + "loss": 0.1441, + "step": 10241 + }, + { + "epoch": 1.86, + "learning_rate": 2.443465598835093e-06, + "loss": 0.6885, + "step": 10242 + }, + { + "epoch": 1.86, + "learning_rate": 2.4434109938114306e-06, + "loss": 0.6239, + "step": 10243 + }, + { + "epoch": 1.86, + "learning_rate": 2.4433563887877685e-06, + "loss": 1.7508, + "step": 10244 + }, + { + "epoch": 1.86, + "learning_rate": 2.4433017837641065e-06, + "loss": 0.1701, + "step": 10245 + }, + { + "epoch": 1.86, + "learning_rate": 2.443247178740444e-06, + "loss": 1.1654, + "step": 10246 + }, + { + "epoch": 1.86, + "learning_rate": 2.443192573716782e-06, + "loss": 0.5746, + "step": 10247 + }, + { + "epoch": 1.86, + "learning_rate": 2.4431379686931195e-06, + "loss": 0.5435, + "step": 10248 + }, + { + "epoch": 1.86, + "learning_rate": 2.4430833636694574e-06, + "loss": 0.0579, + "step": 10249 + }, + { + "epoch": 1.86, + "learning_rate": 2.4430287586457954e-06, + "loss": 2.3463, + "step": 10250 + }, + { + "epoch": 1.86, + "learning_rate": 2.4429741536221333e-06, + "loss": 2.4113, + "step": 10251 + }, + { + "epoch": 1.86, + "learning_rate": 2.4429195485984712e-06, + "loss": 2.7985, + "step": 10252 + }, + { + "epoch": 1.86, + "learning_rate": 2.442864943574809e-06, + "loss": 0.4232, + "step": 10253 + }, + { + "epoch": 1.86, + "learning_rate": 2.4428103385511467e-06, + "loss": 1.5652, + "step": 10254 + }, + { + "epoch": 1.86, + "learning_rate": 2.4427557335274847e-06, + "loss": 1.5115, + "step": 10255 + }, + { + "epoch": 1.87, + "learning_rate": 2.4427011285038226e-06, + "loss": 1.0154, + "step": 10256 + }, + { + "epoch": 1.87, + "learning_rate": 2.44264652348016e-06, + "loss": 0.2242, + "step": 10257 + }, + { + "epoch": 1.87, + "learning_rate": 2.442591918456498e-06, + "loss": 1.8745, + "step": 10258 + }, + { + "epoch": 1.87, + "learning_rate": 2.442537313432836e-06, + "loss": 1.9876, + "step": 10259 + }, + { + "epoch": 1.87, + "learning_rate": 2.4424827084091735e-06, + "loss": 1.0344, + "step": 10260 + }, + { + "epoch": 1.87, + "learning_rate": 2.4424281033855115e-06, + "loss": 0.3815, + "step": 10261 + }, + { + "epoch": 1.87, + "learning_rate": 2.442373498361849e-06, + "loss": 1.079, + "step": 10262 + }, + { + "epoch": 1.87, + "learning_rate": 2.442318893338187e-06, + "loss": 0.617, + "step": 10263 + }, + { + "epoch": 1.87, + "learning_rate": 2.442264288314525e-06, + "loss": 1.364, + "step": 10264 + }, + { + "epoch": 1.87, + "learning_rate": 2.442209683290863e-06, + "loss": 2.0099, + "step": 10265 + }, + { + "epoch": 1.87, + "learning_rate": 2.4421550782672008e-06, + "loss": 0.9058, + "step": 10266 + }, + { + "epoch": 1.87, + "learning_rate": 2.4421004732435387e-06, + "loss": 0.6151, + "step": 10267 + }, + { + "epoch": 1.87, + "learning_rate": 2.4420458682198762e-06, + "loss": 1.005, + "step": 10268 + }, + { + "epoch": 1.87, + "learning_rate": 2.441991263196214e-06, + "loss": 0.4679, + "step": 10269 + }, + { + "epoch": 1.87, + "learning_rate": 2.441936658172552e-06, + "loss": 3.537, + "step": 10270 + }, + { + "epoch": 1.87, + "learning_rate": 2.4418820531488896e-06, + "loss": 0.3312, + "step": 10271 + }, + { + "epoch": 1.87, + "learning_rate": 2.4418274481252276e-06, + "loss": 0.549, + "step": 10272 + }, + { + "epoch": 1.87, + "learning_rate": 2.4417728431015655e-06, + "loss": 0.2202, + "step": 10273 + }, + { + "epoch": 1.87, + "learning_rate": 2.441718238077903e-06, + "loss": 0.875, + "step": 10274 + }, + { + "epoch": 1.87, + "learning_rate": 2.441663633054241e-06, + "loss": 1.6067, + "step": 10275 + }, + { + "epoch": 1.87, + "learning_rate": 2.441609028030579e-06, + "loss": 0.1976, + "step": 10276 + }, + { + "epoch": 1.87, + "learning_rate": 2.441554423006917e-06, + "loss": 0.5935, + "step": 10277 + }, + { + "epoch": 1.87, + "learning_rate": 2.441499817983255e-06, + "loss": 0.0317, + "step": 10278 + }, + { + "epoch": 1.87, + "learning_rate": 2.4414452129595923e-06, + "loss": 0.6502, + "step": 10279 + }, + { + "epoch": 1.87, + "learning_rate": 2.4413906079359303e-06, + "loss": 0.0637, + "step": 10280 + }, + { + "epoch": 1.87, + "learning_rate": 2.4413360029122682e-06, + "loss": 1.7265, + "step": 10281 + }, + { + "epoch": 1.87, + "learning_rate": 2.4412813978886057e-06, + "loss": 0.6213, + "step": 10282 + }, + { + "epoch": 1.87, + "learning_rate": 2.4412267928649437e-06, + "loss": 0.5781, + "step": 10283 + }, + { + "epoch": 1.87, + "learning_rate": 2.4411721878412816e-06, + "loss": 1.5722, + "step": 10284 + }, + { + "epoch": 1.87, + "learning_rate": 2.441117582817619e-06, + "loss": 0.6369, + "step": 10285 + }, + { + "epoch": 1.87, + "learning_rate": 2.441062977793957e-06, + "loss": 1.4903, + "step": 10286 + }, + { + "epoch": 1.87, + "learning_rate": 2.441008372770295e-06, + "loss": 0.349, + "step": 10287 + }, + { + "epoch": 1.87, + "learning_rate": 2.4409537677466326e-06, + "loss": 0.3823, + "step": 10288 + }, + { + "epoch": 1.87, + "learning_rate": 2.4408991627229705e-06, + "loss": 0.0554, + "step": 10289 + }, + { + "epoch": 1.87, + "learning_rate": 2.4408445576993084e-06, + "loss": 1.3015, + "step": 10290 + }, + { + "epoch": 1.87, + "learning_rate": 2.4407899526756464e-06, + "loss": 2.8786, + "step": 10291 + }, + { + "epoch": 1.87, + "learning_rate": 2.4407353476519843e-06, + "loss": 1.8558, + "step": 10292 + }, + { + "epoch": 1.87, + "learning_rate": 2.440680742628322e-06, + "loss": 1.4954, + "step": 10293 + }, + { + "epoch": 1.87, + "learning_rate": 2.44062613760466e-06, + "loss": 1.6869, + "step": 10294 + }, + { + "epoch": 1.87, + "learning_rate": 2.4405715325809977e-06, + "loss": 0.6357, + "step": 10295 + }, + { + "epoch": 1.87, + "learning_rate": 2.4405169275573353e-06, + "loss": 1.1307, + "step": 10296 + }, + { + "epoch": 1.87, + "learning_rate": 2.440462322533673e-06, + "loss": 0.435, + "step": 10297 + }, + { + "epoch": 1.87, + "learning_rate": 2.440407717510011e-06, + "loss": 0.0519, + "step": 10298 + }, + { + "epoch": 1.87, + "learning_rate": 2.4403531124863487e-06, + "loss": 1.3506, + "step": 10299 + }, + { + "epoch": 1.87, + "learning_rate": 2.4402985074626866e-06, + "loss": 2.2262, + "step": 10300 + }, + { + "epoch": 1.87, + "learning_rate": 2.440243902439024e-06, + "loss": 1.1943, + "step": 10301 + }, + { + "epoch": 1.87, + "learning_rate": 2.440189297415362e-06, + "loss": 0.2746, + "step": 10302 + }, + { + "epoch": 1.87, + "learning_rate": 2.4401346923917e-06, + "loss": 0.7918, + "step": 10303 + }, + { + "epoch": 1.87, + "learning_rate": 2.440080087368038e-06, + "loss": 0.466, + "step": 10304 + }, + { + "epoch": 1.87, + "learning_rate": 2.440025482344376e-06, + "loss": 1.435, + "step": 10305 + }, + { + "epoch": 1.87, + "learning_rate": 2.439970877320714e-06, + "loss": 2.8909, + "step": 10306 + }, + { + "epoch": 1.87, + "learning_rate": 2.4399162722970514e-06, + "loss": 0.7272, + "step": 10307 + }, + { + "epoch": 1.87, + "learning_rate": 2.4398616672733893e-06, + "loss": 0.5737, + "step": 10308 + }, + { + "epoch": 1.87, + "learning_rate": 2.4398070622497272e-06, + "loss": 0.6895, + "step": 10309 + }, + { + "epoch": 1.87, + "learning_rate": 2.4397524572260648e-06, + "loss": 0.4839, + "step": 10310 + }, + { + "epoch": 1.88, + "learning_rate": 2.4396978522024027e-06, + "loss": 0.3554, + "step": 10311 + }, + { + "epoch": 1.88, + "learning_rate": 2.4396432471787407e-06, + "loss": 0.6332, + "step": 10312 + }, + { + "epoch": 1.88, + "learning_rate": 2.439588642155078e-06, + "loss": 0.2488, + "step": 10313 + }, + { + "epoch": 1.88, + "learning_rate": 2.439534037131416e-06, + "loss": 0.3476, + "step": 10314 + }, + { + "epoch": 1.88, + "learning_rate": 2.4394794321077536e-06, + "loss": 0.7662, + "step": 10315 + }, + { + "epoch": 1.88, + "learning_rate": 2.4394248270840916e-06, + "loss": 0.2118, + "step": 10316 + }, + { + "epoch": 1.88, + "learning_rate": 2.4393702220604295e-06, + "loss": 0.6937, + "step": 10317 + }, + { + "epoch": 1.88, + "learning_rate": 2.4393156170367675e-06, + "loss": 1.6058, + "step": 10318 + }, + { + "epoch": 1.88, + "learning_rate": 2.4392610120131054e-06, + "loss": 1.343, + "step": 10319 + }, + { + "epoch": 1.88, + "learning_rate": 2.4392064069894434e-06, + "loss": 2.0761, + "step": 10320 + }, + { + "epoch": 1.88, + "learning_rate": 2.439151801965781e-06, + "loss": 4.1277, + "step": 10321 + }, + { + "epoch": 1.88, + "learning_rate": 2.439097196942119e-06, + "loss": 1.0078, + "step": 10322 + }, + { + "epoch": 1.88, + "learning_rate": 2.4390425919184568e-06, + "loss": 1.0959, + "step": 10323 + }, + { + "epoch": 1.88, + "learning_rate": 2.4389879868947943e-06, + "loss": 1.6948, + "step": 10324 + }, + { + "epoch": 1.88, + "learning_rate": 2.4389333818711322e-06, + "loss": 0.9507, + "step": 10325 + }, + { + "epoch": 1.88, + "learning_rate": 2.43887877684747e-06, + "loss": 1.7105, + "step": 10326 + }, + { + "epoch": 1.88, + "learning_rate": 2.4388241718238077e-06, + "loss": 0.0389, + "step": 10327 + }, + { + "epoch": 1.88, + "learning_rate": 2.4387695668001456e-06, + "loss": 0.3834, + "step": 10328 + }, + { + "epoch": 1.88, + "learning_rate": 2.438714961776483e-06, + "loss": 1.3909, + "step": 10329 + }, + { + "epoch": 1.88, + "learning_rate": 2.438660356752821e-06, + "loss": 2.0966, + "step": 10330 + }, + { + "epoch": 1.88, + "learning_rate": 2.438605751729159e-06, + "loss": 0.022, + "step": 10331 + }, + { + "epoch": 1.88, + "learning_rate": 2.438551146705497e-06, + "loss": 1.1122, + "step": 10332 + }, + { + "epoch": 1.88, + "learning_rate": 2.438496541681835e-06, + "loss": 0.2696, + "step": 10333 + }, + { + "epoch": 1.88, + "learning_rate": 2.438441936658173e-06, + "loss": 0.1769, + "step": 10334 + }, + { + "epoch": 1.88, + "learning_rate": 2.4383873316345104e-06, + "loss": 0.9169, + "step": 10335 + }, + { + "epoch": 1.88, + "learning_rate": 2.4383327266108483e-06, + "loss": 1.6722, + "step": 10336 + }, + { + "epoch": 1.88, + "learning_rate": 2.4382781215871863e-06, + "loss": 0.8405, + "step": 10337 + }, + { + "epoch": 1.88, + "learning_rate": 2.438223516563524e-06, + "loss": 0.4622, + "step": 10338 + }, + { + "epoch": 1.88, + "learning_rate": 2.4381689115398617e-06, + "loss": 0.3256, + "step": 10339 + }, + { + "epoch": 1.88, + "learning_rate": 2.4381143065161997e-06, + "loss": 0.2183, + "step": 10340 + }, + { + "epoch": 1.88, + "learning_rate": 2.438059701492537e-06, + "loss": 0.9241, + "step": 10341 + }, + { + "epoch": 1.88, + "learning_rate": 2.438005096468875e-06, + "loss": 1.1469, + "step": 10342 + }, + { + "epoch": 1.88, + "learning_rate": 2.4379504914452127e-06, + "loss": 0.6799, + "step": 10343 + }, + { + "epoch": 1.88, + "learning_rate": 2.4378958864215506e-06, + "loss": 0.4848, + "step": 10344 + }, + { + "epoch": 1.88, + "learning_rate": 2.4378412813978886e-06, + "loss": 0.5423, + "step": 10345 + }, + { + "epoch": 1.88, + "learning_rate": 2.4377866763742265e-06, + "loss": 0.4127, + "step": 10346 + }, + { + "epoch": 1.88, + "learning_rate": 2.4377320713505644e-06, + "loss": 0.4355, + "step": 10347 + }, + { + "epoch": 1.88, + "learning_rate": 2.4376774663269024e-06, + "loss": 2.0732, + "step": 10348 + }, + { + "epoch": 1.88, + "learning_rate": 2.43762286130324e-06, + "loss": 0.0132, + "step": 10349 + }, + { + "epoch": 1.88, + "learning_rate": 2.437568256279578e-06, + "loss": 0.1695, + "step": 10350 + }, + { + "epoch": 1.88, + "learning_rate": 2.4375136512559158e-06, + "loss": 0.3688, + "step": 10351 + }, + { + "epoch": 1.88, + "learning_rate": 2.4374590462322533e-06, + "loss": 0.4492, + "step": 10352 + }, + { + "epoch": 1.88, + "learning_rate": 2.4374044412085913e-06, + "loss": 0.1557, + "step": 10353 + }, + { + "epoch": 1.88, + "learning_rate": 2.437349836184929e-06, + "loss": 1.1396, + "step": 10354 + }, + { + "epoch": 1.88, + "learning_rate": 2.4372952311612667e-06, + "loss": 1.2764, + "step": 10355 + }, + { + "epoch": 1.88, + "learning_rate": 2.4372406261376047e-06, + "loss": 2.0308, + "step": 10356 + }, + { + "epoch": 1.88, + "learning_rate": 2.4371860211139426e-06, + "loss": 1.1841, + "step": 10357 + }, + { + "epoch": 1.88, + "learning_rate": 2.4371314160902805e-06, + "loss": 0.1276, + "step": 10358 + }, + { + "epoch": 1.88, + "learning_rate": 2.4370768110666185e-06, + "loss": 0.3653, + "step": 10359 + }, + { + "epoch": 1.88, + "learning_rate": 2.437022206042956e-06, + "loss": 0.0544, + "step": 10360 + }, + { + "epoch": 1.88, + "learning_rate": 2.436967601019294e-06, + "loss": 0.8139, + "step": 10361 + }, + { + "epoch": 1.88, + "learning_rate": 2.436912995995632e-06, + "loss": 1.1241, + "step": 10362 + }, + { + "epoch": 1.88, + "learning_rate": 2.4368583909719694e-06, + "loss": 0.6064, + "step": 10363 + }, + { + "epoch": 1.88, + "learning_rate": 2.4368037859483074e-06, + "loss": 0.0504, + "step": 10364 + }, + { + "epoch": 1.88, + "learning_rate": 2.4367491809246453e-06, + "loss": 1.2408, + "step": 10365 + }, + { + "epoch": 1.89, + "learning_rate": 2.436694575900983e-06, + "loss": 1.2546, + "step": 10366 + }, + { + "epoch": 1.89, + "learning_rate": 2.4366399708773208e-06, + "loss": 0.6872, + "step": 10367 + }, + { + "epoch": 1.89, + "learning_rate": 2.4365853658536587e-06, + "loss": 0.0303, + "step": 10368 + }, + { + "epoch": 1.89, + "learning_rate": 2.4365307608299962e-06, + "loss": 0.9207, + "step": 10369 + }, + { + "epoch": 1.89, + "learning_rate": 2.436476155806334e-06, + "loss": 0.7898, + "step": 10370 + }, + { + "epoch": 1.89, + "learning_rate": 2.436421550782672e-06, + "loss": 0.8071, + "step": 10371 + }, + { + "epoch": 1.89, + "learning_rate": 2.43636694575901e-06, + "loss": 1.1087, + "step": 10372 + }, + { + "epoch": 1.89, + "learning_rate": 2.436312340735348e-06, + "loss": 1.2501, + "step": 10373 + }, + { + "epoch": 1.89, + "learning_rate": 2.4362577357116855e-06, + "loss": 1.673, + "step": 10374 + }, + { + "epoch": 1.89, + "learning_rate": 2.4362031306880235e-06, + "loss": 0.7698, + "step": 10375 + }, + { + "epoch": 1.89, + "learning_rate": 2.4361485256643614e-06, + "loss": 0.2194, + "step": 10376 + }, + { + "epoch": 1.89, + "learning_rate": 2.436093920640699e-06, + "loss": 1.8754, + "step": 10377 + }, + { + "epoch": 1.89, + "learning_rate": 2.436039315617037e-06, + "loss": 1.4979, + "step": 10378 + }, + { + "epoch": 1.89, + "learning_rate": 2.435984710593375e-06, + "loss": 0.5637, + "step": 10379 + }, + { + "epoch": 1.89, + "learning_rate": 2.4359301055697123e-06, + "loss": 0.0393, + "step": 10380 + }, + { + "epoch": 1.89, + "learning_rate": 2.4358755005460503e-06, + "loss": 2.4143, + "step": 10381 + }, + { + "epoch": 1.89, + "learning_rate": 2.435820895522388e-06, + "loss": 2.7662, + "step": 10382 + }, + { + "epoch": 1.89, + "learning_rate": 2.4357662904987257e-06, + "loss": 0.4341, + "step": 10383 + }, + { + "epoch": 1.89, + "learning_rate": 2.4357116854750637e-06, + "loss": 0.8511, + "step": 10384 + }, + { + "epoch": 1.89, + "learning_rate": 2.4356570804514016e-06, + "loss": 0.5616, + "step": 10385 + }, + { + "epoch": 1.89, + "learning_rate": 2.4356024754277396e-06, + "loss": 1.1502, + "step": 10386 + }, + { + "epoch": 1.89, + "learning_rate": 2.4355478704040775e-06, + "loss": 1.531, + "step": 10387 + }, + { + "epoch": 1.89, + "learning_rate": 2.435493265380415e-06, + "loss": 1.1895, + "step": 10388 + }, + { + "epoch": 1.89, + "learning_rate": 2.435438660356753e-06, + "loss": 0.3223, + "step": 10389 + }, + { + "epoch": 1.89, + "learning_rate": 2.435384055333091e-06, + "loss": 3.2945, + "step": 10390 + }, + { + "epoch": 1.89, + "learning_rate": 2.4353294503094284e-06, + "loss": 1.856, + "step": 10391 + }, + { + "epoch": 1.89, + "learning_rate": 2.4352748452857664e-06, + "loss": 1.2781, + "step": 10392 + }, + { + "epoch": 1.89, + "learning_rate": 2.4352202402621043e-06, + "loss": 0.4764, + "step": 10393 + }, + { + "epoch": 1.89, + "learning_rate": 2.435165635238442e-06, + "loss": 1.7675, + "step": 10394 + }, + { + "epoch": 1.89, + "learning_rate": 2.43511103021478e-06, + "loss": 1.4244, + "step": 10395 + }, + { + "epoch": 1.89, + "learning_rate": 2.4350564251911173e-06, + "loss": 0.602, + "step": 10396 + }, + { + "epoch": 1.89, + "learning_rate": 2.4350018201674553e-06, + "loss": 0.7636, + "step": 10397 + }, + { + "epoch": 1.89, + "learning_rate": 2.434947215143793e-06, + "loss": 0.9607, + "step": 10398 + }, + { + "epoch": 1.89, + "learning_rate": 2.434892610120131e-06, + "loss": 0.1357, + "step": 10399 + }, + { + "epoch": 1.89, + "learning_rate": 2.434838005096469e-06, + "loss": 0.46, + "step": 10400 + }, + { + "epoch": 1.89, + "learning_rate": 2.434783400072807e-06, + "loss": 1.0518, + "step": 10401 + }, + { + "epoch": 1.89, + "learning_rate": 2.4347287950491445e-06, + "loss": 1.021, + "step": 10402 + }, + { + "epoch": 1.89, + "learning_rate": 2.4346741900254825e-06, + "loss": 0.2259, + "step": 10403 + }, + { + "epoch": 1.89, + "learning_rate": 2.4346195850018204e-06, + "loss": 1.1273, + "step": 10404 + }, + { + "epoch": 1.89, + "learning_rate": 2.434564979978158e-06, + "loss": 0.9853, + "step": 10405 + }, + { + "epoch": 1.89, + "learning_rate": 2.434510374954496e-06, + "loss": 0.1847, + "step": 10406 + }, + { + "epoch": 1.89, + "learning_rate": 2.434455769930834e-06, + "loss": 0.8197, + "step": 10407 + }, + { + "epoch": 1.89, + "learning_rate": 2.4344011649071714e-06, + "loss": 2.4033, + "step": 10408 + }, + { + "epoch": 1.89, + "learning_rate": 2.4343465598835093e-06, + "loss": 1.0736, + "step": 10409 + }, + { + "epoch": 1.89, + "learning_rate": 2.434291954859847e-06, + "loss": 0.7924, + "step": 10410 + }, + { + "epoch": 1.89, + "learning_rate": 2.4342373498361848e-06, + "loss": 1.1775, + "step": 10411 + }, + { + "epoch": 1.89, + "learning_rate": 2.4341827448125227e-06, + "loss": 0.5903, + "step": 10412 + }, + { + "epoch": 1.89, + "learning_rate": 2.4341281397888607e-06, + "loss": 1.5725, + "step": 10413 + }, + { + "epoch": 1.89, + "learning_rate": 2.4340735347651986e-06, + "loss": 0.4296, + "step": 10414 + }, + { + "epoch": 1.89, + "learning_rate": 2.4340189297415365e-06, + "loss": 1.0109, + "step": 10415 + }, + { + "epoch": 1.89, + "learning_rate": 2.433964324717874e-06, + "loss": 0.4095, + "step": 10416 + }, + { + "epoch": 1.89, + "learning_rate": 2.433909719694212e-06, + "loss": 1.0943, + "step": 10417 + }, + { + "epoch": 1.89, + "learning_rate": 2.43385511467055e-06, + "loss": 0.6982, + "step": 10418 + }, + { + "epoch": 1.89, + "learning_rate": 2.4338005096468875e-06, + "loss": 1.1387, + "step": 10419 + }, + { + "epoch": 1.89, + "learning_rate": 2.4337459046232254e-06, + "loss": 0.0488, + "step": 10420 + }, + { + "epoch": 1.9, + "learning_rate": 2.4336912995995634e-06, + "loss": 0.9986, + "step": 10421 + }, + { + "epoch": 1.9, + "learning_rate": 2.433636694575901e-06, + "loss": 0.4812, + "step": 10422 + }, + { + "epoch": 1.9, + "learning_rate": 2.433582089552239e-06, + "loss": 1.2927, + "step": 10423 + }, + { + "epoch": 1.9, + "learning_rate": 2.4335274845285763e-06, + "loss": 0.7156, + "step": 10424 + }, + { + "epoch": 1.9, + "learning_rate": 2.4334728795049143e-06, + "loss": 1.07, + "step": 10425 + }, + { + "epoch": 1.9, + "learning_rate": 2.4334182744812522e-06, + "loss": 0.0518, + "step": 10426 + }, + { + "epoch": 1.9, + "learning_rate": 2.43336366945759e-06, + "loss": 0.4511, + "step": 10427 + }, + { + "epoch": 1.9, + "learning_rate": 2.433309064433928e-06, + "loss": 0.0275, + "step": 10428 + }, + { + "epoch": 1.9, + "learning_rate": 2.433254459410266e-06, + "loss": 0.0181, + "step": 10429 + }, + { + "epoch": 1.9, + "learning_rate": 2.4331998543866036e-06, + "loss": 0.6554, + "step": 10430 + }, + { + "epoch": 1.9, + "learning_rate": 2.4331452493629415e-06, + "loss": 2.0701, + "step": 10431 + }, + { + "epoch": 1.9, + "learning_rate": 2.4330906443392795e-06, + "loss": 1.1679, + "step": 10432 + }, + { + "epoch": 1.9, + "learning_rate": 2.433036039315617e-06, + "loss": 1.6418, + "step": 10433 + }, + { + "epoch": 1.9, + "learning_rate": 2.432981434291955e-06, + "loss": 1.4981, + "step": 10434 + }, + { + "epoch": 1.9, + "learning_rate": 2.432926829268293e-06, + "loss": 0.4389, + "step": 10435 + }, + { + "epoch": 1.9, + "learning_rate": 2.4328722242446304e-06, + "loss": 0.5878, + "step": 10436 + }, + { + "epoch": 1.9, + "learning_rate": 2.4328176192209683e-06, + "loss": 0.0917, + "step": 10437 + }, + { + "epoch": 1.9, + "learning_rate": 2.4327630141973063e-06, + "loss": 0.2673, + "step": 10438 + }, + { + "epoch": 1.9, + "learning_rate": 2.4327084091736442e-06, + "loss": 1.5039, + "step": 10439 + }, + { + "epoch": 1.9, + "learning_rate": 2.432653804149982e-06, + "loss": 0.5054, + "step": 10440 + }, + { + "epoch": 1.9, + "learning_rate": 2.4325991991263197e-06, + "loss": 2.2796, + "step": 10441 + }, + { + "epoch": 1.9, + "learning_rate": 2.4325445941026576e-06, + "loss": 0.1429, + "step": 10442 + }, + { + "epoch": 1.9, + "learning_rate": 2.4324899890789956e-06, + "loss": 0.289, + "step": 10443 + }, + { + "epoch": 1.9, + "learning_rate": 2.432435384055333e-06, + "loss": 1.1076, + "step": 10444 + }, + { + "epoch": 1.9, + "learning_rate": 2.432380779031671e-06, + "loss": 0.0472, + "step": 10445 + }, + { + "epoch": 1.9, + "learning_rate": 2.432326174008009e-06, + "loss": 0.8437, + "step": 10446 + }, + { + "epoch": 1.9, + "learning_rate": 2.4322715689843465e-06, + "loss": 0.0236, + "step": 10447 + }, + { + "epoch": 1.9, + "learning_rate": 2.4322169639606844e-06, + "loss": 0.5445, + "step": 10448 + }, + { + "epoch": 1.9, + "learning_rate": 2.432162358937022e-06, + "loss": 0.4678, + "step": 10449 + }, + { + "epoch": 1.9, + "learning_rate": 2.43210775391336e-06, + "loss": 1.0636, + "step": 10450 + }, + { + "epoch": 1.9, + "learning_rate": 2.432053148889698e-06, + "loss": 1.6917, + "step": 10451 + }, + { + "epoch": 1.9, + "learning_rate": 2.431998543866036e-06, + "loss": 0.877, + "step": 10452 + }, + { + "epoch": 1.9, + "learning_rate": 2.4319439388423737e-06, + "loss": 3.5679, + "step": 10453 + }, + { + "epoch": 1.9, + "learning_rate": 2.4318893338187117e-06, + "loss": 1.8943, + "step": 10454 + }, + { + "epoch": 1.9, + "learning_rate": 2.431834728795049e-06, + "loss": 0.9795, + "step": 10455 + }, + { + "epoch": 1.9, + "learning_rate": 2.431780123771387e-06, + "loss": 0.6913, + "step": 10456 + }, + { + "epoch": 1.9, + "learning_rate": 2.431725518747725e-06, + "loss": 0.1072, + "step": 10457 + }, + { + "epoch": 1.9, + "learning_rate": 2.4316709137240626e-06, + "loss": 4.1825, + "step": 10458 + }, + { + "epoch": 1.9, + "learning_rate": 2.4316163087004005e-06, + "loss": 1.3711, + "step": 10459 + }, + { + "epoch": 1.9, + "learning_rate": 2.4315617036767385e-06, + "loss": 0.2086, + "step": 10460 + }, + { + "epoch": 1.9, + "learning_rate": 2.431507098653076e-06, + "loss": 0.3824, + "step": 10461 + }, + { + "epoch": 1.9, + "learning_rate": 2.431452493629414e-06, + "loss": 0.1781, + "step": 10462 + }, + { + "epoch": 1.9, + "learning_rate": 2.4313978886057515e-06, + "loss": 2.2212, + "step": 10463 + }, + { + "epoch": 1.9, + "learning_rate": 2.4313432835820894e-06, + "loss": 3.1194, + "step": 10464 + }, + { + "epoch": 1.9, + "learning_rate": 2.4312886785584274e-06, + "loss": 0.0483, + "step": 10465 + }, + { + "epoch": 1.9, + "learning_rate": 2.4312340735347653e-06, + "loss": 0.636, + "step": 10466 + }, + { + "epoch": 1.9, + "learning_rate": 2.4311794685111032e-06, + "loss": 0.0645, + "step": 10467 + }, + { + "epoch": 1.9, + "learning_rate": 2.431124863487441e-06, + "loss": 0.0569, + "step": 10468 + }, + { + "epoch": 1.9, + "learning_rate": 2.4310702584637787e-06, + "loss": 0.4638, + "step": 10469 + }, + { + "epoch": 1.9, + "learning_rate": 2.4310156534401167e-06, + "loss": 0.866, + "step": 10470 + }, + { + "epoch": 1.9, + "learning_rate": 2.4309610484164546e-06, + "loss": 1.8762, + "step": 10471 + }, + { + "epoch": 1.9, + "learning_rate": 2.430906443392792e-06, + "loss": 4.3088, + "step": 10472 + }, + { + "epoch": 1.9, + "learning_rate": 2.43085183836913e-06, + "loss": 1.6102, + "step": 10473 + }, + { + "epoch": 1.9, + "learning_rate": 2.430797233345468e-06, + "loss": 0.9279, + "step": 10474 + }, + { + "epoch": 1.9, + "learning_rate": 2.4307426283218055e-06, + "loss": 0.348, + "step": 10475 + }, + { + "epoch": 1.91, + "learning_rate": 2.4306880232981435e-06, + "loss": 1.3368, + "step": 10476 + }, + { + "epoch": 1.91, + "learning_rate": 2.430633418274481e-06, + "loss": 0.4179, + "step": 10477 + }, + { + "epoch": 1.91, + "learning_rate": 2.430578813250819e-06, + "loss": 2.3352, + "step": 10478 + }, + { + "epoch": 1.91, + "learning_rate": 2.430524208227157e-06, + "loss": 0.9928, + "step": 10479 + }, + { + "epoch": 1.91, + "learning_rate": 2.430469603203495e-06, + "loss": 0.4752, + "step": 10480 + }, + { + "epoch": 1.91, + "learning_rate": 2.4304149981798328e-06, + "loss": 0.0305, + "step": 10481 + }, + { + "epoch": 1.91, + "learning_rate": 2.4303603931561707e-06, + "loss": 1.1026, + "step": 10482 + }, + { + "epoch": 1.91, + "learning_rate": 2.4303057881325082e-06, + "loss": 1.6729, + "step": 10483 + }, + { + "epoch": 1.91, + "learning_rate": 2.430251183108846e-06, + "loss": 0.0453, + "step": 10484 + }, + { + "epoch": 1.91, + "learning_rate": 2.430196578085184e-06, + "loss": 0.7725, + "step": 10485 + }, + { + "epoch": 1.91, + "learning_rate": 2.4301419730615216e-06, + "loss": 0.4434, + "step": 10486 + }, + { + "epoch": 1.91, + "learning_rate": 2.4300873680378596e-06, + "loss": 0.1901, + "step": 10487 + }, + { + "epoch": 1.91, + "learning_rate": 2.4300327630141975e-06, + "loss": 0.6403, + "step": 10488 + }, + { + "epoch": 1.91, + "learning_rate": 2.429978157990535e-06, + "loss": 0.5871, + "step": 10489 + }, + { + "epoch": 1.91, + "learning_rate": 2.429923552966873e-06, + "loss": 1.769, + "step": 10490 + }, + { + "epoch": 1.91, + "learning_rate": 2.4298689479432105e-06, + "loss": 0.1368, + "step": 10491 + }, + { + "epoch": 1.91, + "learning_rate": 2.4298143429195484e-06, + "loss": 0.8087, + "step": 10492 + }, + { + "epoch": 1.91, + "learning_rate": 2.4297597378958864e-06, + "loss": 0.1315, + "step": 10493 + }, + { + "epoch": 1.91, + "learning_rate": 2.4297051328722243e-06, + "loss": 2.1833, + "step": 10494 + }, + { + "epoch": 1.91, + "learning_rate": 2.4296505278485623e-06, + "loss": 0.5056, + "step": 10495 + }, + { + "epoch": 1.91, + "learning_rate": 2.4295959228249002e-06, + "loss": 0.7617, + "step": 10496 + }, + { + "epoch": 1.91, + "learning_rate": 2.4295413178012377e-06, + "loss": 1.599, + "step": 10497 + }, + { + "epoch": 1.91, + "learning_rate": 2.4294867127775757e-06, + "loss": 0.4642, + "step": 10498 + }, + { + "epoch": 1.91, + "learning_rate": 2.4294321077539136e-06, + "loss": 1.0492, + "step": 10499 + }, + { + "epoch": 1.91, + "learning_rate": 2.429377502730251e-06, + "loss": 0.5952, + "step": 10500 + }, + { + "epoch": 1.91, + "learning_rate": 2.429322897706589e-06, + "loss": 0.9558, + "step": 10501 + }, + { + "epoch": 1.91, + "learning_rate": 2.429268292682927e-06, + "loss": 0.9416, + "step": 10502 + }, + { + "epoch": 1.91, + "learning_rate": 2.4292136876592645e-06, + "loss": 0.6383, + "step": 10503 + }, + { + "epoch": 1.91, + "learning_rate": 2.4291590826356025e-06, + "loss": 0.0334, + "step": 10504 + }, + { + "epoch": 1.91, + "learning_rate": 2.42910447761194e-06, + "loss": 0.729, + "step": 10505 + }, + { + "epoch": 1.91, + "learning_rate": 2.429049872588278e-06, + "loss": 1.7911, + "step": 10506 + }, + { + "epoch": 1.91, + "learning_rate": 2.428995267564616e-06, + "loss": 0.8649, + "step": 10507 + }, + { + "epoch": 1.91, + "learning_rate": 2.428940662540954e-06, + "loss": 0.3618, + "step": 10508 + }, + { + "epoch": 1.91, + "learning_rate": 2.4288860575172918e-06, + "loss": 0.0975, + "step": 10509 + }, + { + "epoch": 1.91, + "learning_rate": 2.4288314524936297e-06, + "loss": 0.0332, + "step": 10510 + }, + { + "epoch": 1.91, + "learning_rate": 2.4287768474699672e-06, + "loss": 0.5451, + "step": 10511 + }, + { + "epoch": 1.91, + "learning_rate": 2.428722242446305e-06, + "loss": 0.1041, + "step": 10512 + }, + { + "epoch": 1.91, + "learning_rate": 2.428667637422643e-06, + "loss": 0.4186, + "step": 10513 + }, + { + "epoch": 1.91, + "learning_rate": 2.4286130323989807e-06, + "loss": 2.0686, + "step": 10514 + }, + { + "epoch": 1.91, + "learning_rate": 2.4285584273753186e-06, + "loss": 1.3063, + "step": 10515 + }, + { + "epoch": 1.91, + "learning_rate": 2.4285038223516565e-06, + "loss": 0.8476, + "step": 10516 + }, + { + "epoch": 1.91, + "learning_rate": 2.428449217327994e-06, + "loss": 0.0389, + "step": 10517 + }, + { + "epoch": 1.91, + "learning_rate": 2.428394612304332e-06, + "loss": 0.8293, + "step": 10518 + }, + { + "epoch": 1.91, + "learning_rate": 2.42834000728067e-06, + "loss": 0.1781, + "step": 10519 + }, + { + "epoch": 1.91, + "learning_rate": 2.4282854022570075e-06, + "loss": 0.6923, + "step": 10520 + }, + { + "epoch": 1.91, + "learning_rate": 2.4282307972333454e-06, + "loss": 0.0196, + "step": 10521 + }, + { + "epoch": 1.91, + "learning_rate": 2.4281761922096834e-06, + "loss": 1.3053, + "step": 10522 + }, + { + "epoch": 1.91, + "learning_rate": 2.4281215871860213e-06, + "loss": 0.4134, + "step": 10523 + }, + { + "epoch": 1.91, + "learning_rate": 2.4280669821623592e-06, + "loss": 1.4361, + "step": 10524 + }, + { + "epoch": 1.91, + "learning_rate": 2.4280123771386968e-06, + "loss": 0.4762, + "step": 10525 + }, + { + "epoch": 1.91, + "learning_rate": 2.4279577721150347e-06, + "loss": 0.7268, + "step": 10526 + }, + { + "epoch": 1.91, + "learning_rate": 2.4279031670913726e-06, + "loss": 0.4252, + "step": 10527 + }, + { + "epoch": 1.91, + "learning_rate": 2.42784856206771e-06, + "loss": 0.4656, + "step": 10528 + }, + { + "epoch": 1.91, + "learning_rate": 2.427793957044048e-06, + "loss": 1.2833, + "step": 10529 + }, + { + "epoch": 1.91, + "learning_rate": 2.4277393520203856e-06, + "loss": 0.9561, + "step": 10530 + }, + { + "epoch": 1.92, + "learning_rate": 2.4276847469967236e-06, + "loss": 1.3915, + "step": 10531 + }, + { + "epoch": 1.92, + "learning_rate": 2.4276301419730615e-06, + "loss": 0.9119, + "step": 10532 + }, + { + "epoch": 1.92, + "learning_rate": 2.4275755369493995e-06, + "loss": 0.2852, + "step": 10533 + }, + { + "epoch": 1.92, + "learning_rate": 2.4275209319257374e-06, + "loss": 0.7219, + "step": 10534 + }, + { + "epoch": 1.92, + "learning_rate": 2.4274663269020753e-06, + "loss": 0.1914, + "step": 10535 + }, + { + "epoch": 1.92, + "learning_rate": 2.427411721878413e-06, + "loss": 1.0497, + "step": 10536 + }, + { + "epoch": 1.92, + "learning_rate": 2.427357116854751e-06, + "loss": 0.0913, + "step": 10537 + }, + { + "epoch": 1.92, + "learning_rate": 2.4273025118310888e-06, + "loss": 3.5064, + "step": 10538 + }, + { + "epoch": 1.92, + "learning_rate": 2.4272479068074263e-06, + "loss": 0.7097, + "step": 10539 + }, + { + "epoch": 1.92, + "learning_rate": 2.4271933017837642e-06, + "loss": 0.5646, + "step": 10540 + }, + { + "epoch": 1.92, + "learning_rate": 2.427138696760102e-06, + "loss": 1.1713, + "step": 10541 + }, + { + "epoch": 1.92, + "learning_rate": 2.4270840917364397e-06, + "loss": 0.2516, + "step": 10542 + }, + { + "epoch": 1.92, + "learning_rate": 2.4270294867127776e-06, + "loss": 2.6232, + "step": 10543 + }, + { + "epoch": 1.92, + "learning_rate": 2.426974881689115e-06, + "loss": 3.0569, + "step": 10544 + }, + { + "epoch": 1.92, + "learning_rate": 2.426920276665453e-06, + "loss": 0.5798, + "step": 10545 + }, + { + "epoch": 1.92, + "learning_rate": 2.426865671641791e-06, + "loss": 0.8898, + "step": 10546 + }, + { + "epoch": 1.92, + "learning_rate": 2.426811066618129e-06, + "loss": 0.9744, + "step": 10547 + }, + { + "epoch": 1.92, + "learning_rate": 2.426756461594467e-06, + "loss": 0.3101, + "step": 10548 + }, + { + "epoch": 1.92, + "learning_rate": 2.426701856570805e-06, + "loss": 0.7096, + "step": 10549 + }, + { + "epoch": 1.92, + "learning_rate": 2.4266472515471424e-06, + "loss": 0.2939, + "step": 10550 + }, + { + "epoch": 1.92, + "learning_rate": 2.4265926465234803e-06, + "loss": 1.7602, + "step": 10551 + }, + { + "epoch": 1.92, + "learning_rate": 2.4265380414998183e-06, + "loss": 0.6122, + "step": 10552 + }, + { + "epoch": 1.92, + "learning_rate": 2.426483436476156e-06, + "loss": 1.8794, + "step": 10553 + }, + { + "epoch": 1.92, + "learning_rate": 2.4264288314524937e-06, + "loss": 0.8144, + "step": 10554 + }, + { + "epoch": 1.92, + "learning_rate": 2.4263742264288317e-06, + "loss": 0.8505, + "step": 10555 + }, + { + "epoch": 1.92, + "learning_rate": 2.426319621405169e-06, + "loss": 0.0538, + "step": 10556 + }, + { + "epoch": 1.92, + "learning_rate": 2.426265016381507e-06, + "loss": 0.1334, + "step": 10557 + }, + { + "epoch": 1.92, + "learning_rate": 2.4262104113578447e-06, + "loss": 1.2576, + "step": 10558 + }, + { + "epoch": 1.92, + "learning_rate": 2.4261558063341826e-06, + "loss": 1.7684, + "step": 10559 + }, + { + "epoch": 1.92, + "learning_rate": 2.4261012013105205e-06, + "loss": 0.0433, + "step": 10560 + }, + { + "epoch": 1.92, + "learning_rate": 2.4260465962868585e-06, + "loss": 0.4174, + "step": 10561 + }, + { + "epoch": 1.92, + "learning_rate": 2.4259919912631964e-06, + "loss": 2.5533, + "step": 10562 + }, + { + "epoch": 1.92, + "learning_rate": 2.4259373862395344e-06, + "loss": 0.5448, + "step": 10563 + }, + { + "epoch": 1.92, + "learning_rate": 2.425882781215872e-06, + "loss": 1.0221, + "step": 10564 + }, + { + "epoch": 1.92, + "learning_rate": 2.42582817619221e-06, + "loss": 1.5312, + "step": 10565 + }, + { + "epoch": 1.92, + "learning_rate": 2.4257735711685478e-06, + "loss": 3.0147, + "step": 10566 + }, + { + "epoch": 1.92, + "learning_rate": 2.4257189661448853e-06, + "loss": 2.0618, + "step": 10567 + }, + { + "epoch": 1.92, + "learning_rate": 2.4256643611212232e-06, + "loss": 1.4137, + "step": 10568 + }, + { + "epoch": 1.92, + "learning_rate": 2.425609756097561e-06, + "loss": 0.9574, + "step": 10569 + }, + { + "epoch": 1.92, + "learning_rate": 2.4255551510738987e-06, + "loss": 2.5199, + "step": 10570 + }, + { + "epoch": 1.92, + "learning_rate": 2.4255005460502367e-06, + "loss": 0.0424, + "step": 10571 + }, + { + "epoch": 1.92, + "learning_rate": 2.425445941026574e-06, + "loss": 2.0683, + "step": 10572 + }, + { + "epoch": 1.92, + "learning_rate": 2.425391336002912e-06, + "loss": 0.2274, + "step": 10573 + }, + { + "epoch": 1.92, + "learning_rate": 2.42533673097925e-06, + "loss": 0.0142, + "step": 10574 + }, + { + "epoch": 1.92, + "learning_rate": 2.425282125955588e-06, + "loss": 1.0409, + "step": 10575 + }, + { + "epoch": 1.92, + "learning_rate": 2.425227520931926e-06, + "loss": 0.3046, + "step": 10576 + }, + { + "epoch": 1.92, + "learning_rate": 2.425172915908264e-06, + "loss": 0.3277, + "step": 10577 + }, + { + "epoch": 1.92, + "learning_rate": 2.4251183108846014e-06, + "loss": 1.749, + "step": 10578 + }, + { + "epoch": 1.92, + "learning_rate": 2.4250637058609394e-06, + "loss": 0.171, + "step": 10579 + }, + { + "epoch": 1.92, + "learning_rate": 2.4250091008372773e-06, + "loss": 1.9803, + "step": 10580 + }, + { + "epoch": 1.92, + "learning_rate": 2.424954495813615e-06, + "loss": 0.9769, + "step": 10581 + }, + { + "epoch": 1.92, + "learning_rate": 2.4248998907899528e-06, + "loss": 1.3204, + "step": 10582 + }, + { + "epoch": 1.92, + "learning_rate": 2.4248452857662907e-06, + "loss": 0.7605, + "step": 10583 + }, + { + "epoch": 1.92, + "learning_rate": 2.4247906807426282e-06, + "loss": 1.3031, + "step": 10584 + }, + { + "epoch": 1.92, + "learning_rate": 2.424736075718966e-06, + "loss": 0.0297, + "step": 10585 + }, + { + "epoch": 1.93, + "learning_rate": 2.4246814706953037e-06, + "loss": 0.9247, + "step": 10586 + }, + { + "epoch": 1.93, + "learning_rate": 2.4246268656716416e-06, + "loss": 0.2879, + "step": 10587 + }, + { + "epoch": 1.93, + "learning_rate": 2.4245722606479796e-06, + "loss": 2.1849, + "step": 10588 + }, + { + "epoch": 1.93, + "learning_rate": 2.4245176556243175e-06, + "loss": 0.9932, + "step": 10589 + }, + { + "epoch": 1.93, + "learning_rate": 2.4244630506006555e-06, + "loss": 1.6337, + "step": 10590 + }, + { + "epoch": 1.93, + "learning_rate": 2.4244084455769934e-06, + "loss": 0.0442, + "step": 10591 + }, + { + "epoch": 1.93, + "learning_rate": 2.424353840553331e-06, + "loss": 0.5088, + "step": 10592 + }, + { + "epoch": 1.93, + "learning_rate": 2.424299235529669e-06, + "loss": 2.3841, + "step": 10593 + }, + { + "epoch": 1.93, + "learning_rate": 2.424244630506007e-06, + "loss": 0.6753, + "step": 10594 + }, + { + "epoch": 1.93, + "learning_rate": 2.4241900254823443e-06, + "loss": 0.747, + "step": 10595 + }, + { + "epoch": 1.93, + "learning_rate": 2.4241354204586823e-06, + "loss": 0.3912, + "step": 10596 + }, + { + "epoch": 1.93, + "learning_rate": 2.4240808154350202e-06, + "loss": 1.1372, + "step": 10597 + }, + { + "epoch": 1.93, + "learning_rate": 2.4240262104113577e-06, + "loss": 2.1686, + "step": 10598 + }, + { + "epoch": 1.93, + "learning_rate": 2.4239716053876957e-06, + "loss": 0.8768, + "step": 10599 + }, + { + "epoch": 1.93, + "learning_rate": 2.423917000364033e-06, + "loss": 0.4856, + "step": 10600 + }, + { + "epoch": 1.93, + "learning_rate": 2.423862395340371e-06, + "loss": 0.7009, + "step": 10601 + }, + { + "epoch": 1.93, + "learning_rate": 2.423807790316709e-06, + "loss": 1.5138, + "step": 10602 + }, + { + "epoch": 1.93, + "learning_rate": 2.423753185293047e-06, + "loss": 0.1611, + "step": 10603 + }, + { + "epoch": 1.93, + "learning_rate": 2.423698580269385e-06, + "loss": 1.2271, + "step": 10604 + }, + { + "epoch": 1.93, + "learning_rate": 2.423643975245723e-06, + "loss": 0.1133, + "step": 10605 + }, + { + "epoch": 1.93, + "learning_rate": 2.4235893702220604e-06, + "loss": 0.5939, + "step": 10606 + }, + { + "epoch": 1.93, + "learning_rate": 2.4235347651983984e-06, + "loss": 1.6228, + "step": 10607 + }, + { + "epoch": 1.93, + "learning_rate": 2.4234801601747363e-06, + "loss": 0.4851, + "step": 10608 + }, + { + "epoch": 1.93, + "learning_rate": 2.423425555151074e-06, + "loss": 0.5783, + "step": 10609 + }, + { + "epoch": 1.93, + "learning_rate": 2.4233709501274118e-06, + "loss": 0.0769, + "step": 10610 + }, + { + "epoch": 1.93, + "learning_rate": 2.4233163451037493e-06, + "loss": 0.0745, + "step": 10611 + }, + { + "epoch": 1.93, + "learning_rate": 2.4232617400800873e-06, + "loss": 0.489, + "step": 10612 + }, + { + "epoch": 1.93, + "learning_rate": 2.423207135056425e-06, + "loss": 2.1417, + "step": 10613 + }, + { + "epoch": 1.93, + "learning_rate": 2.423152530032763e-06, + "loss": 1.2922, + "step": 10614 + }, + { + "epoch": 1.93, + "learning_rate": 2.423097925009101e-06, + "loss": 1.4384, + "step": 10615 + }, + { + "epoch": 1.93, + "learning_rate": 2.423043319985439e-06, + "loss": 1.3057, + "step": 10616 + }, + { + "epoch": 1.93, + "learning_rate": 2.4229887149617765e-06, + "loss": 1.9115, + "step": 10617 + }, + { + "epoch": 1.93, + "learning_rate": 2.4229341099381145e-06, + "loss": 0.4166, + "step": 10618 + }, + { + "epoch": 1.93, + "learning_rate": 2.4228795049144524e-06, + "loss": 0.4018, + "step": 10619 + }, + { + "epoch": 1.93, + "learning_rate": 2.42282489989079e-06, + "loss": 0.8667, + "step": 10620 + }, + { + "epoch": 1.93, + "learning_rate": 2.422770294867128e-06, + "loss": 1.3336, + "step": 10621 + }, + { + "epoch": 1.93, + "learning_rate": 2.422715689843466e-06, + "loss": 1.7019, + "step": 10622 + }, + { + "epoch": 1.93, + "learning_rate": 2.4226610848198034e-06, + "loss": 2.4637, + "step": 10623 + }, + { + "epoch": 1.93, + "learning_rate": 2.4226064797961413e-06, + "loss": 1.143, + "step": 10624 + }, + { + "epoch": 1.93, + "learning_rate": 2.422551874772479e-06, + "loss": 0.0543, + "step": 10625 + }, + { + "epoch": 1.93, + "learning_rate": 2.4224972697488168e-06, + "loss": 0.1146, + "step": 10626 + }, + { + "epoch": 1.93, + "learning_rate": 2.4224426647251547e-06, + "loss": 2.65, + "step": 10627 + }, + { + "epoch": 1.93, + "learning_rate": 2.4223880597014927e-06, + "loss": 0.3321, + "step": 10628 + }, + { + "epoch": 1.93, + "learning_rate": 2.4223334546778306e-06, + "loss": 1.0805, + "step": 10629 + }, + { + "epoch": 1.93, + "learning_rate": 2.4222788496541685e-06, + "loss": 0.7452, + "step": 10630 + }, + { + "epoch": 1.93, + "learning_rate": 2.422224244630506e-06, + "loss": 3.2218, + "step": 10631 + }, + { + "epoch": 1.93, + "learning_rate": 2.422169639606844e-06, + "loss": 0.2427, + "step": 10632 + }, + { + "epoch": 1.93, + "learning_rate": 2.422115034583182e-06, + "loss": 1.2368, + "step": 10633 + }, + { + "epoch": 1.93, + "learning_rate": 2.4220604295595195e-06, + "loss": 1.0692, + "step": 10634 + }, + { + "epoch": 1.93, + "learning_rate": 2.4220058245358574e-06, + "loss": 0.0337, + "step": 10635 + }, + { + "epoch": 1.93, + "learning_rate": 2.4219512195121954e-06, + "loss": 0.5483, + "step": 10636 + }, + { + "epoch": 1.93, + "learning_rate": 2.421896614488533e-06, + "loss": 0.6288, + "step": 10637 + }, + { + "epoch": 1.93, + "learning_rate": 2.421842009464871e-06, + "loss": 0.7807, + "step": 10638 + }, + { + "epoch": 1.93, + "learning_rate": 2.4217874044412083e-06, + "loss": 0.0388, + "step": 10639 + }, + { + "epoch": 1.93, + "learning_rate": 2.4217327994175463e-06, + "loss": 0.82, + "step": 10640 + }, + { + "epoch": 1.94, + "learning_rate": 2.4216781943938842e-06, + "loss": 0.352, + "step": 10641 + }, + { + "epoch": 1.94, + "learning_rate": 2.421623589370222e-06, + "loss": 0.2285, + "step": 10642 + }, + { + "epoch": 1.94, + "learning_rate": 2.42156898434656e-06, + "loss": 0.5518, + "step": 10643 + }, + { + "epoch": 1.94, + "learning_rate": 2.421514379322898e-06, + "loss": 1.6928, + "step": 10644 + }, + { + "epoch": 1.94, + "learning_rate": 2.4214597742992356e-06, + "loss": 1.0249, + "step": 10645 + }, + { + "epoch": 1.94, + "learning_rate": 2.4214051692755735e-06, + "loss": 1.5285, + "step": 10646 + }, + { + "epoch": 1.94, + "learning_rate": 2.4213505642519115e-06, + "loss": 0.492, + "step": 10647 + }, + { + "epoch": 1.94, + "learning_rate": 2.421295959228249e-06, + "loss": 0.5197, + "step": 10648 + }, + { + "epoch": 1.94, + "learning_rate": 2.421241354204587e-06, + "loss": 0.9388, + "step": 10649 + }, + { + "epoch": 1.94, + "learning_rate": 2.421186749180925e-06, + "loss": 0.7772, + "step": 10650 + }, + { + "epoch": 1.94, + "learning_rate": 2.4211321441572624e-06, + "loss": 0.612, + "step": 10651 + }, + { + "epoch": 1.94, + "learning_rate": 2.4210775391336003e-06, + "loss": 0.6727, + "step": 10652 + }, + { + "epoch": 1.94, + "learning_rate": 2.421022934109938e-06, + "loss": 1.1816, + "step": 10653 + }, + { + "epoch": 1.94, + "learning_rate": 2.420968329086276e-06, + "loss": 0.1179, + "step": 10654 + }, + { + "epoch": 1.94, + "learning_rate": 2.4209137240626137e-06, + "loss": 1.5206, + "step": 10655 + }, + { + "epoch": 1.94, + "learning_rate": 2.4208591190389517e-06, + "loss": 0.9903, + "step": 10656 + }, + { + "epoch": 1.94, + "learning_rate": 2.4208045140152896e-06, + "loss": 1.2159, + "step": 10657 + }, + { + "epoch": 1.94, + "learning_rate": 2.4207499089916276e-06, + "loss": 0.4553, + "step": 10658 + }, + { + "epoch": 1.94, + "learning_rate": 2.420695303967965e-06, + "loss": 0.1835, + "step": 10659 + }, + { + "epoch": 1.94, + "learning_rate": 2.420640698944303e-06, + "loss": 0.5033, + "step": 10660 + }, + { + "epoch": 1.94, + "learning_rate": 2.420586093920641e-06, + "loss": 0.7985, + "step": 10661 + }, + { + "epoch": 1.94, + "learning_rate": 2.4205314888969785e-06, + "loss": 0.9535, + "step": 10662 + }, + { + "epoch": 1.94, + "learning_rate": 2.4204768838733164e-06, + "loss": 0.8882, + "step": 10663 + }, + { + "epoch": 1.94, + "learning_rate": 2.4204222788496544e-06, + "loss": 0.6812, + "step": 10664 + }, + { + "epoch": 1.94, + "learning_rate": 2.420367673825992e-06, + "loss": 0.062, + "step": 10665 + }, + { + "epoch": 1.94, + "learning_rate": 2.42031306880233e-06, + "loss": 0.2971, + "step": 10666 + }, + { + "epoch": 1.94, + "learning_rate": 2.4202584637786674e-06, + "loss": 1.4769, + "step": 10667 + }, + { + "epoch": 1.94, + "learning_rate": 2.4202038587550053e-06, + "loss": 0.4802, + "step": 10668 + }, + { + "epoch": 1.94, + "learning_rate": 2.4201492537313432e-06, + "loss": 0.5782, + "step": 10669 + }, + { + "epoch": 1.94, + "learning_rate": 2.420094648707681e-06, + "loss": 0.8844, + "step": 10670 + }, + { + "epoch": 1.94, + "learning_rate": 2.420040043684019e-06, + "loss": 0.6358, + "step": 10671 + }, + { + "epoch": 1.94, + "learning_rate": 2.419985438660357e-06, + "loss": 0.4066, + "step": 10672 + }, + { + "epoch": 1.94, + "learning_rate": 2.4199308336366946e-06, + "loss": 0.3656, + "step": 10673 + }, + { + "epoch": 1.94, + "learning_rate": 2.4198762286130325e-06, + "loss": 0.4354, + "step": 10674 + }, + { + "epoch": 1.94, + "learning_rate": 2.4198216235893705e-06, + "loss": 0.1602, + "step": 10675 + }, + { + "epoch": 1.94, + "learning_rate": 2.419767018565708e-06, + "loss": 0.6173, + "step": 10676 + }, + { + "epoch": 1.94, + "learning_rate": 2.419712413542046e-06, + "loss": 0.0202, + "step": 10677 + }, + { + "epoch": 1.94, + "learning_rate": 2.419657808518384e-06, + "loss": 0.4111, + "step": 10678 + }, + { + "epoch": 1.94, + "learning_rate": 2.4196032034947214e-06, + "loss": 1.9277, + "step": 10679 + }, + { + "epoch": 1.94, + "learning_rate": 2.4195485984710594e-06, + "loss": 0.6807, + "step": 10680 + }, + { + "epoch": 1.94, + "learning_rate": 2.419493993447397e-06, + "loss": 0.3835, + "step": 10681 + }, + { + "epoch": 1.94, + "learning_rate": 2.419439388423735e-06, + "loss": 0.8106, + "step": 10682 + }, + { + "epoch": 1.94, + "learning_rate": 2.4193847834000728e-06, + "loss": 2.4598, + "step": 10683 + }, + { + "epoch": 1.94, + "learning_rate": 2.4193301783764107e-06, + "loss": 1.4741, + "step": 10684 + }, + { + "epoch": 1.94, + "learning_rate": 2.4192755733527486e-06, + "loss": 0.5236, + "step": 10685 + }, + { + "epoch": 1.94, + "learning_rate": 2.4192209683290866e-06, + "loss": 1.0355, + "step": 10686 + }, + { + "epoch": 1.94, + "learning_rate": 2.419166363305424e-06, + "loss": 0.562, + "step": 10687 + }, + { + "epoch": 1.94, + "learning_rate": 2.419111758281762e-06, + "loss": 2.9576, + "step": 10688 + }, + { + "epoch": 1.94, + "learning_rate": 2.4190571532581e-06, + "loss": 0.1927, + "step": 10689 + }, + { + "epoch": 1.94, + "learning_rate": 2.4190025482344375e-06, + "loss": 0.1568, + "step": 10690 + }, + { + "epoch": 1.94, + "learning_rate": 2.4189479432107755e-06, + "loss": 0.5113, + "step": 10691 + }, + { + "epoch": 1.94, + "learning_rate": 2.418893338187113e-06, + "loss": 2.5813, + "step": 10692 + }, + { + "epoch": 1.94, + "learning_rate": 2.418838733163451e-06, + "loss": 0.9812, + "step": 10693 + }, + { + "epoch": 1.94, + "learning_rate": 2.418784128139789e-06, + "loss": 0.7959, + "step": 10694 + }, + { + "epoch": 1.94, + "learning_rate": 2.418729523116127e-06, + "loss": 0.344, + "step": 10695 + }, + { + "epoch": 1.95, + "learning_rate": 2.4186749180924648e-06, + "loss": 0.0579, + "step": 10696 + }, + { + "epoch": 1.95, + "learning_rate": 2.4186203130688027e-06, + "loss": 1.9386, + "step": 10697 + }, + { + "epoch": 1.95, + "learning_rate": 2.4185657080451402e-06, + "loss": 2.0602, + "step": 10698 + }, + { + "epoch": 1.95, + "learning_rate": 2.418511103021478e-06, + "loss": 0.8291, + "step": 10699 + }, + { + "epoch": 1.95, + "learning_rate": 2.418456497997816e-06, + "loss": 0.5484, + "step": 10700 + }, + { + "epoch": 1.95, + "learning_rate": 2.4184018929741536e-06, + "loss": 5.1037, + "step": 10701 + }, + { + "epoch": 1.95, + "learning_rate": 2.4183472879504916e-06, + "loss": 0.7478, + "step": 10702 + }, + { + "epoch": 1.95, + "learning_rate": 2.4182926829268295e-06, + "loss": 1.0893, + "step": 10703 + }, + { + "epoch": 1.95, + "learning_rate": 2.418238077903167e-06, + "loss": 0.5163, + "step": 10704 + }, + { + "epoch": 1.95, + "learning_rate": 2.418183472879505e-06, + "loss": 0.6818, + "step": 10705 + }, + { + "epoch": 1.95, + "learning_rate": 2.4181288678558425e-06, + "loss": 0.1665, + "step": 10706 + }, + { + "epoch": 1.95, + "learning_rate": 2.4180742628321804e-06, + "loss": 0.8121, + "step": 10707 + }, + { + "epoch": 1.95, + "learning_rate": 2.4180196578085184e-06, + "loss": 2.2755, + "step": 10708 + }, + { + "epoch": 1.95, + "learning_rate": 2.4179650527848563e-06, + "loss": 0.6143, + "step": 10709 + }, + { + "epoch": 1.95, + "learning_rate": 2.4179104477611943e-06, + "loss": 0.7032, + "step": 10710 + }, + { + "epoch": 1.95, + "learning_rate": 2.417855842737532e-06, + "loss": 0.2806, + "step": 10711 + }, + { + "epoch": 1.95, + "learning_rate": 2.4178012377138697e-06, + "loss": 0.9321, + "step": 10712 + }, + { + "epoch": 1.95, + "learning_rate": 2.4177466326902077e-06, + "loss": 0.4854, + "step": 10713 + }, + { + "epoch": 1.95, + "learning_rate": 2.4176920276665456e-06, + "loss": 0.4281, + "step": 10714 + }, + { + "epoch": 1.95, + "learning_rate": 2.417637422642883e-06, + "loss": 0.5853, + "step": 10715 + }, + { + "epoch": 1.95, + "learning_rate": 2.417582817619221e-06, + "loss": 0.0279, + "step": 10716 + }, + { + "epoch": 1.95, + "learning_rate": 2.417528212595559e-06, + "loss": 3.4859, + "step": 10717 + }, + { + "epoch": 1.95, + "learning_rate": 2.4174736075718965e-06, + "loss": 0.0587, + "step": 10718 + }, + { + "epoch": 1.95, + "learning_rate": 2.4174190025482345e-06, + "loss": 0.1788, + "step": 10719 + }, + { + "epoch": 1.95, + "learning_rate": 2.417364397524572e-06, + "loss": 0.6776, + "step": 10720 + }, + { + "epoch": 1.95, + "learning_rate": 2.41730979250091e-06, + "loss": 0.0291, + "step": 10721 + }, + { + "epoch": 1.95, + "learning_rate": 2.417255187477248e-06, + "loss": 4.7872, + "step": 10722 + }, + { + "epoch": 1.95, + "learning_rate": 2.417200582453586e-06, + "loss": 0.0396, + "step": 10723 + }, + { + "epoch": 1.95, + "learning_rate": 2.4171459774299238e-06, + "loss": 0.9617, + "step": 10724 + }, + { + "epoch": 1.95, + "learning_rate": 2.4170913724062617e-06, + "loss": 0.7162, + "step": 10725 + }, + { + "epoch": 1.95, + "learning_rate": 2.4170367673825992e-06, + "loss": 0.6625, + "step": 10726 + }, + { + "epoch": 1.95, + "learning_rate": 2.416982162358937e-06, + "loss": 0.8019, + "step": 10727 + }, + { + "epoch": 1.95, + "learning_rate": 2.416927557335275e-06, + "loss": 1.8951, + "step": 10728 + }, + { + "epoch": 1.95, + "learning_rate": 2.4168729523116127e-06, + "loss": 1.9689, + "step": 10729 + }, + { + "epoch": 1.95, + "learning_rate": 2.4168183472879506e-06, + "loss": 1.032, + "step": 10730 + }, + { + "epoch": 1.95, + "learning_rate": 2.4167637422642885e-06, + "loss": 1.0061, + "step": 10731 + }, + { + "epoch": 1.95, + "learning_rate": 2.416709137240626e-06, + "loss": 0.4179, + "step": 10732 + }, + { + "epoch": 1.95, + "learning_rate": 2.416654532216964e-06, + "loss": 1.2087, + "step": 10733 + }, + { + "epoch": 1.95, + "learning_rate": 2.4165999271933015e-06, + "loss": 1.7007, + "step": 10734 + }, + { + "epoch": 1.95, + "learning_rate": 2.4165453221696395e-06, + "loss": 0.3264, + "step": 10735 + }, + { + "epoch": 1.95, + "learning_rate": 2.4164907171459774e-06, + "loss": 0.3405, + "step": 10736 + }, + { + "epoch": 1.95, + "learning_rate": 2.4164361121223154e-06, + "loss": 0.4694, + "step": 10737 + }, + { + "epoch": 1.95, + "learning_rate": 2.4163815070986533e-06, + "loss": 1.355, + "step": 10738 + }, + { + "epoch": 1.95, + "learning_rate": 2.4163269020749912e-06, + "loss": 1.4752, + "step": 10739 + }, + { + "epoch": 1.95, + "learning_rate": 2.4162722970513288e-06, + "loss": 1.0554, + "step": 10740 + }, + { + "epoch": 1.95, + "learning_rate": 2.4162176920276667e-06, + "loss": 3.2872, + "step": 10741 + }, + { + "epoch": 1.95, + "learning_rate": 2.4161630870040046e-06, + "loss": 1.4204, + "step": 10742 + }, + { + "epoch": 1.95, + "learning_rate": 2.416108481980342e-06, + "loss": 2.5195, + "step": 10743 + }, + { + "epoch": 1.95, + "learning_rate": 2.41605387695668e-06, + "loss": 0.63, + "step": 10744 + }, + { + "epoch": 1.95, + "learning_rate": 2.415999271933018e-06, + "loss": 0.694, + "step": 10745 + }, + { + "epoch": 1.95, + "learning_rate": 2.4159446669093556e-06, + "loss": 1.0347, + "step": 10746 + }, + { + "epoch": 1.95, + "learning_rate": 2.4158900618856935e-06, + "loss": 0.3485, + "step": 10747 + }, + { + "epoch": 1.95, + "learning_rate": 2.415835456862031e-06, + "loss": 1.0185, + "step": 10748 + }, + { + "epoch": 1.95, + "learning_rate": 2.415780851838369e-06, + "loss": 2.1086, + "step": 10749 + }, + { + "epoch": 1.95, + "learning_rate": 2.415726246814707e-06, + "loss": 0.6839, + "step": 10750 + }, + { + "epoch": 1.96, + "learning_rate": 2.415671641791045e-06, + "loss": 1.5772, + "step": 10751 + }, + { + "epoch": 1.96, + "learning_rate": 2.415617036767383e-06, + "loss": 2.494, + "step": 10752 + }, + { + "epoch": 1.96, + "learning_rate": 2.4155624317437208e-06, + "loss": 1.9424, + "step": 10753 + }, + { + "epoch": 1.96, + "learning_rate": 2.4155078267200583e-06, + "loss": 0.1415, + "step": 10754 + }, + { + "epoch": 1.96, + "learning_rate": 2.4154532216963962e-06, + "loss": 1.3594, + "step": 10755 + }, + { + "epoch": 1.96, + "learning_rate": 2.415398616672734e-06, + "loss": 0.0351, + "step": 10756 + }, + { + "epoch": 1.96, + "learning_rate": 2.4153440116490717e-06, + "loss": 1.8109, + "step": 10757 + }, + { + "epoch": 1.96, + "learning_rate": 2.4152894066254096e-06, + "loss": 0.0912, + "step": 10758 + }, + { + "epoch": 1.96, + "learning_rate": 2.415234801601747e-06, + "loss": 0.7667, + "step": 10759 + }, + { + "epoch": 1.96, + "learning_rate": 2.415180196578085e-06, + "loss": 1.7404, + "step": 10760 + }, + { + "epoch": 1.96, + "learning_rate": 2.415125591554423e-06, + "loss": 0.1877, + "step": 10761 + }, + { + "epoch": 1.96, + "learning_rate": 2.4150709865307605e-06, + "loss": 1.0267, + "step": 10762 + }, + { + "epoch": 1.96, + "learning_rate": 2.4150163815070985e-06, + "loss": 0.3484, + "step": 10763 + }, + { + "epoch": 1.96, + "learning_rate": 2.4149617764834364e-06, + "loss": 2.4815, + "step": 10764 + }, + { + "epoch": 1.96, + "learning_rate": 2.4149071714597744e-06, + "loss": 0.9969, + "step": 10765 + }, + { + "epoch": 1.96, + "learning_rate": 2.4148525664361123e-06, + "loss": 2.0928, + "step": 10766 + }, + { + "epoch": 1.96, + "learning_rate": 2.4147979614124503e-06, + "loss": 1.2831, + "step": 10767 + }, + { + "epoch": 1.96, + "learning_rate": 2.4147433563887878e-06, + "loss": 0.3608, + "step": 10768 + }, + { + "epoch": 1.96, + "learning_rate": 2.4146887513651257e-06, + "loss": 0.3706, + "step": 10769 + }, + { + "epoch": 1.96, + "learning_rate": 2.4146341463414637e-06, + "loss": 0.4124, + "step": 10770 + }, + { + "epoch": 1.96, + "learning_rate": 2.414579541317801e-06, + "loss": 0.1942, + "step": 10771 + }, + { + "epoch": 1.96, + "learning_rate": 2.414524936294139e-06, + "loss": 0.0384, + "step": 10772 + }, + { + "epoch": 1.96, + "learning_rate": 2.4144703312704767e-06, + "loss": 1.9454, + "step": 10773 + }, + { + "epoch": 1.96, + "learning_rate": 2.4144157262468146e-06, + "loss": 0.9067, + "step": 10774 + }, + { + "epoch": 1.96, + "learning_rate": 2.4143611212231525e-06, + "loss": 1.367, + "step": 10775 + }, + { + "epoch": 1.96, + "learning_rate": 2.4143065161994905e-06, + "loss": 1.2242, + "step": 10776 + }, + { + "epoch": 1.96, + "learning_rate": 2.4142519111758284e-06, + "loss": 0.0713, + "step": 10777 + }, + { + "epoch": 1.96, + "learning_rate": 2.4141973061521664e-06, + "loss": 0.1397, + "step": 10778 + }, + { + "epoch": 1.96, + "learning_rate": 2.414142701128504e-06, + "loss": 1.5409, + "step": 10779 + }, + { + "epoch": 1.96, + "learning_rate": 2.414088096104842e-06, + "loss": 0.5678, + "step": 10780 + }, + { + "epoch": 1.96, + "learning_rate": 2.4140334910811798e-06, + "loss": 0.3432, + "step": 10781 + }, + { + "epoch": 1.96, + "learning_rate": 2.4139788860575173e-06, + "loss": 1.8829, + "step": 10782 + }, + { + "epoch": 1.96, + "learning_rate": 2.4139242810338552e-06, + "loss": 0.9285, + "step": 10783 + }, + { + "epoch": 1.96, + "learning_rate": 2.413869676010193e-06, + "loss": 0.6374, + "step": 10784 + }, + { + "epoch": 1.96, + "learning_rate": 2.4138150709865307e-06, + "loss": 1.9531, + "step": 10785 + }, + { + "epoch": 1.96, + "learning_rate": 2.4137604659628686e-06, + "loss": 0.7002, + "step": 10786 + }, + { + "epoch": 1.96, + "learning_rate": 2.413705860939206e-06, + "loss": 0.4568, + "step": 10787 + }, + { + "epoch": 1.96, + "learning_rate": 2.413651255915544e-06, + "loss": 0.9653, + "step": 10788 + }, + { + "epoch": 1.96, + "learning_rate": 2.413596650891882e-06, + "loss": 0.2611, + "step": 10789 + }, + { + "epoch": 1.96, + "learning_rate": 2.41354204586822e-06, + "loss": 1.7026, + "step": 10790 + }, + { + "epoch": 1.96, + "learning_rate": 2.413487440844558e-06, + "loss": 0.2144, + "step": 10791 + }, + { + "epoch": 1.96, + "learning_rate": 2.413432835820896e-06, + "loss": 1.6171, + "step": 10792 + }, + { + "epoch": 1.96, + "learning_rate": 2.4133782307972334e-06, + "loss": 0.8784, + "step": 10793 + }, + { + "epoch": 1.96, + "learning_rate": 2.4133236257735713e-06, + "loss": 0.3563, + "step": 10794 + }, + { + "epoch": 1.96, + "learning_rate": 2.4132690207499093e-06, + "loss": 0.5004, + "step": 10795 + }, + { + "epoch": 1.96, + "learning_rate": 2.413214415726247e-06, + "loss": 0.3745, + "step": 10796 + }, + { + "epoch": 1.96, + "learning_rate": 2.4131598107025848e-06, + "loss": 0.7876, + "step": 10797 + }, + { + "epoch": 1.96, + "learning_rate": 2.4131052056789227e-06, + "loss": 1.0551, + "step": 10798 + }, + { + "epoch": 1.96, + "learning_rate": 2.4130506006552602e-06, + "loss": 0.9162, + "step": 10799 + }, + { + "epoch": 1.96, + "learning_rate": 2.412995995631598e-06, + "loss": 0.9955, + "step": 10800 + }, + { + "epoch": 1.96, + "learning_rate": 2.4129413906079357e-06, + "loss": 0.076, + "step": 10801 + }, + { + "epoch": 1.96, + "learning_rate": 2.4128867855842736e-06, + "loss": 0.3837, + "step": 10802 + }, + { + "epoch": 1.96, + "learning_rate": 2.4128321805606116e-06, + "loss": 0.7498, + "step": 10803 + }, + { + "epoch": 1.96, + "learning_rate": 2.4127775755369495e-06, + "loss": 0.3058, + "step": 10804 + }, + { + "epoch": 1.96, + "learning_rate": 2.4127229705132875e-06, + "loss": 0.2167, + "step": 10805 + }, + { + "epoch": 1.97, + "learning_rate": 2.4126683654896254e-06, + "loss": 0.4738, + "step": 10806 + }, + { + "epoch": 1.97, + "learning_rate": 2.412613760465963e-06, + "loss": 0.1947, + "step": 10807 + }, + { + "epoch": 1.97, + "learning_rate": 2.412559155442301e-06, + "loss": 0.7808, + "step": 10808 + }, + { + "epoch": 1.97, + "learning_rate": 2.412504550418639e-06, + "loss": 1.1876, + "step": 10809 + }, + { + "epoch": 1.97, + "learning_rate": 2.4124499453949763e-06, + "loss": 2.1688, + "step": 10810 + }, + { + "epoch": 1.97, + "learning_rate": 2.4123953403713143e-06, + "loss": 1.3837, + "step": 10811 + }, + { + "epoch": 1.97, + "learning_rate": 2.4123407353476522e-06, + "loss": 0.4031, + "step": 10812 + }, + { + "epoch": 1.97, + "learning_rate": 2.4122861303239897e-06, + "loss": 0.3998, + "step": 10813 + }, + { + "epoch": 1.97, + "learning_rate": 2.4122315253003277e-06, + "loss": 0.7283, + "step": 10814 + }, + { + "epoch": 1.97, + "learning_rate": 2.412176920276665e-06, + "loss": 0.0369, + "step": 10815 + }, + { + "epoch": 1.97, + "learning_rate": 2.412122315253003e-06, + "loss": 0.0508, + "step": 10816 + }, + { + "epoch": 1.97, + "learning_rate": 2.412067710229341e-06, + "loss": 0.6522, + "step": 10817 + }, + { + "epoch": 1.97, + "learning_rate": 2.412013105205679e-06, + "loss": 1.683, + "step": 10818 + }, + { + "epoch": 1.97, + "learning_rate": 2.411958500182017e-06, + "loss": 0.537, + "step": 10819 + }, + { + "epoch": 1.97, + "learning_rate": 2.411903895158355e-06, + "loss": 0.1564, + "step": 10820 + }, + { + "epoch": 1.97, + "learning_rate": 2.4118492901346924e-06, + "loss": 1.3368, + "step": 10821 + }, + { + "epoch": 1.97, + "learning_rate": 2.4117946851110304e-06, + "loss": 0.3898, + "step": 10822 + }, + { + "epoch": 1.97, + "learning_rate": 2.4117400800873683e-06, + "loss": 0.2778, + "step": 10823 + }, + { + "epoch": 1.97, + "learning_rate": 2.411685475063706e-06, + "loss": 0.8073, + "step": 10824 + }, + { + "epoch": 1.97, + "learning_rate": 2.4116308700400438e-06, + "loss": 0.5451, + "step": 10825 + }, + { + "epoch": 1.97, + "learning_rate": 2.4115762650163817e-06, + "loss": 0.6883, + "step": 10826 + }, + { + "epoch": 1.97, + "learning_rate": 2.4115216599927192e-06, + "loss": 1.2198, + "step": 10827 + }, + { + "epoch": 1.97, + "learning_rate": 2.411467054969057e-06, + "loss": 2.0056, + "step": 10828 + }, + { + "epoch": 1.97, + "learning_rate": 2.4114124499453947e-06, + "loss": 0.6178, + "step": 10829 + }, + { + "epoch": 1.97, + "learning_rate": 2.4113578449217327e-06, + "loss": 0.5564, + "step": 10830 + }, + { + "epoch": 1.97, + "learning_rate": 2.4113032398980706e-06, + "loss": 0.2872, + "step": 10831 + }, + { + "epoch": 1.97, + "learning_rate": 2.4112486348744085e-06, + "loss": 0.4718, + "step": 10832 + }, + { + "epoch": 1.97, + "learning_rate": 2.4111940298507465e-06, + "loss": 2.4753, + "step": 10833 + }, + { + "epoch": 1.97, + "learning_rate": 2.4111394248270844e-06, + "loss": 0.1297, + "step": 10834 + }, + { + "epoch": 1.97, + "learning_rate": 2.411084819803422e-06, + "loss": 1.5838, + "step": 10835 + }, + { + "epoch": 1.97, + "learning_rate": 2.41103021477976e-06, + "loss": 0.0485, + "step": 10836 + }, + { + "epoch": 1.97, + "learning_rate": 2.410975609756098e-06, + "loss": 1.4311, + "step": 10837 + }, + { + "epoch": 1.97, + "learning_rate": 2.4109210047324354e-06, + "loss": 0.086, + "step": 10838 + }, + { + "epoch": 1.97, + "learning_rate": 2.4108663997087733e-06, + "loss": 0.0338, + "step": 10839 + }, + { + "epoch": 1.97, + "learning_rate": 2.410811794685111e-06, + "loss": 0.0522, + "step": 10840 + }, + { + "epoch": 1.97, + "learning_rate": 2.4107571896614488e-06, + "loss": 0.5698, + "step": 10841 + }, + { + "epoch": 1.97, + "learning_rate": 2.4107025846377867e-06, + "loss": 0.7151, + "step": 10842 + }, + { + "epoch": 1.97, + "learning_rate": 2.4106479796141242e-06, + "loss": 1.9643, + "step": 10843 + }, + { + "epoch": 1.97, + "learning_rate": 2.410593374590462e-06, + "loss": 1.234, + "step": 10844 + }, + { + "epoch": 1.97, + "learning_rate": 2.4105387695668e-06, + "loss": 1.8562, + "step": 10845 + }, + { + "epoch": 1.97, + "learning_rate": 2.410484164543138e-06, + "loss": 0.8425, + "step": 10846 + }, + { + "epoch": 1.97, + "learning_rate": 2.410429559519476e-06, + "loss": 0.3348, + "step": 10847 + }, + { + "epoch": 1.97, + "learning_rate": 2.410374954495814e-06, + "loss": 0.5785, + "step": 10848 + }, + { + "epoch": 1.97, + "learning_rate": 2.4103203494721515e-06, + "loss": 1.1496, + "step": 10849 + }, + { + "epoch": 1.97, + "learning_rate": 2.4102657444484894e-06, + "loss": 1.3352, + "step": 10850 + }, + { + "epoch": 1.97, + "learning_rate": 2.4102111394248273e-06, + "loss": 0.9172, + "step": 10851 + }, + { + "epoch": 1.97, + "learning_rate": 2.410156534401165e-06, + "loss": 1.0134, + "step": 10852 + }, + { + "epoch": 1.97, + "learning_rate": 2.410101929377503e-06, + "loss": 0.0214, + "step": 10853 + }, + { + "epoch": 1.97, + "learning_rate": 2.4100473243538403e-06, + "loss": 2.6636, + "step": 10854 + }, + { + "epoch": 1.97, + "learning_rate": 2.4099927193301783e-06, + "loss": 2.4588, + "step": 10855 + }, + { + "epoch": 1.97, + "learning_rate": 2.4099381143065162e-06, + "loss": 0.3046, + "step": 10856 + }, + { + "epoch": 1.97, + "learning_rate": 2.4098835092828537e-06, + "loss": 0.6075, + "step": 10857 + }, + { + "epoch": 1.97, + "learning_rate": 2.4098289042591917e-06, + "loss": 0.608, + "step": 10858 + }, + { + "epoch": 1.97, + "learning_rate": 2.4097742992355296e-06, + "loss": 0.7934, + "step": 10859 + }, + { + "epoch": 1.97, + "learning_rate": 2.4097196942118676e-06, + "loss": 1.5603, + "step": 10860 + }, + { + "epoch": 1.98, + "learning_rate": 2.4096650891882055e-06, + "loss": 0.2234, + "step": 10861 + }, + { + "epoch": 1.98, + "learning_rate": 2.4096104841645435e-06, + "loss": 1.1076, + "step": 10862 + }, + { + "epoch": 1.98, + "learning_rate": 2.409555879140881e-06, + "loss": 0.1153, + "step": 10863 + }, + { + "epoch": 1.98, + "learning_rate": 2.409501274117219e-06, + "loss": 0.407, + "step": 10864 + }, + { + "epoch": 1.98, + "learning_rate": 2.409446669093557e-06, + "loss": 1.3951, + "step": 10865 + }, + { + "epoch": 1.98, + "learning_rate": 2.4093920640698944e-06, + "loss": 0.7863, + "step": 10866 + }, + { + "epoch": 1.98, + "learning_rate": 2.4093374590462323e-06, + "loss": 0.8416, + "step": 10867 + }, + { + "epoch": 1.98, + "learning_rate": 2.40928285402257e-06, + "loss": 0.8691, + "step": 10868 + }, + { + "epoch": 1.98, + "learning_rate": 2.4092282489989078e-06, + "loss": 2.159, + "step": 10869 + }, + { + "epoch": 1.98, + "learning_rate": 2.4091736439752457e-06, + "loss": 0.1719, + "step": 10870 + }, + { + "epoch": 1.98, + "learning_rate": 2.4091190389515837e-06, + "loss": 0.2725, + "step": 10871 + }, + { + "epoch": 1.98, + "learning_rate": 2.4090644339279216e-06, + "loss": 0.4205, + "step": 10872 + }, + { + "epoch": 1.98, + "learning_rate": 2.4090098289042596e-06, + "loss": 1.994, + "step": 10873 + }, + { + "epoch": 1.98, + "learning_rate": 2.408955223880597e-06, + "loss": 0.3803, + "step": 10874 + }, + { + "epoch": 1.98, + "learning_rate": 2.408900618856935e-06, + "loss": 4.8162, + "step": 10875 + }, + { + "epoch": 1.98, + "learning_rate": 2.408846013833273e-06, + "loss": 3.4716, + "step": 10876 + }, + { + "epoch": 1.98, + "learning_rate": 2.4087914088096105e-06, + "loss": 0.2021, + "step": 10877 + }, + { + "epoch": 1.98, + "learning_rate": 2.4087368037859484e-06, + "loss": 0.1121, + "step": 10878 + }, + { + "epoch": 1.98, + "learning_rate": 2.4086821987622864e-06, + "loss": 0.2909, + "step": 10879 + }, + { + "epoch": 1.98, + "learning_rate": 2.408627593738624e-06, + "loss": 0.843, + "step": 10880 + }, + { + "epoch": 1.98, + "learning_rate": 2.408572988714962e-06, + "loss": 0.6711, + "step": 10881 + }, + { + "epoch": 1.98, + "learning_rate": 2.4085183836912994e-06, + "loss": 0.5929, + "step": 10882 + }, + { + "epoch": 1.98, + "learning_rate": 2.4084637786676373e-06, + "loss": 0.2568, + "step": 10883 + }, + { + "epoch": 1.98, + "learning_rate": 2.4084091736439752e-06, + "loss": 0.6975, + "step": 10884 + }, + { + "epoch": 1.98, + "learning_rate": 2.408354568620313e-06, + "loss": 0.0167, + "step": 10885 + }, + { + "epoch": 1.98, + "learning_rate": 2.408299963596651e-06, + "loss": 0.1722, + "step": 10886 + }, + { + "epoch": 1.98, + "learning_rate": 2.408245358572989e-06, + "loss": 0.5269, + "step": 10887 + }, + { + "epoch": 1.98, + "learning_rate": 2.4081907535493266e-06, + "loss": 1.2822, + "step": 10888 + }, + { + "epoch": 1.98, + "learning_rate": 2.4081361485256645e-06, + "loss": 1.9524, + "step": 10889 + }, + { + "epoch": 1.98, + "learning_rate": 2.4080815435020025e-06, + "loss": 0.3513, + "step": 10890 + }, + { + "epoch": 1.98, + "learning_rate": 2.40802693847834e-06, + "loss": 0.3874, + "step": 10891 + }, + { + "epoch": 1.98, + "learning_rate": 2.407972333454678e-06, + "loss": 0.633, + "step": 10892 + }, + { + "epoch": 1.98, + "learning_rate": 2.407917728431016e-06, + "loss": 2.7866, + "step": 10893 + }, + { + "epoch": 1.98, + "learning_rate": 2.4078631234073534e-06, + "loss": 0.3612, + "step": 10894 + }, + { + "epoch": 1.98, + "learning_rate": 2.4078085183836913e-06, + "loss": 2.1525, + "step": 10895 + }, + { + "epoch": 1.98, + "learning_rate": 2.407753913360029e-06, + "loss": 0.8197, + "step": 10896 + }, + { + "epoch": 1.98, + "learning_rate": 2.407699308336367e-06, + "loss": 0.4814, + "step": 10897 + }, + { + "epoch": 1.98, + "learning_rate": 2.4076447033127048e-06, + "loss": 0.8479, + "step": 10898 + }, + { + "epoch": 1.98, + "learning_rate": 2.4075900982890427e-06, + "loss": 1.8414, + "step": 10899 + }, + { + "epoch": 1.98, + "learning_rate": 2.4075354932653806e-06, + "loss": 1.3649, + "step": 10900 + }, + { + "epoch": 1.98, + "learning_rate": 2.4074808882417186e-06, + "loss": 2.4243, + "step": 10901 + }, + { + "epoch": 1.98, + "learning_rate": 2.407426283218056e-06, + "loss": 0.6904, + "step": 10902 + }, + { + "epoch": 1.98, + "learning_rate": 2.407371678194394e-06, + "loss": 2.2969, + "step": 10903 + }, + { + "epoch": 1.98, + "learning_rate": 2.407317073170732e-06, + "loss": 1.1809, + "step": 10904 + }, + { + "epoch": 1.98, + "learning_rate": 2.4072624681470695e-06, + "loss": 1.6666, + "step": 10905 + }, + { + "epoch": 1.98, + "learning_rate": 2.4072078631234075e-06, + "loss": 0.0496, + "step": 10906 + }, + { + "epoch": 1.98, + "learning_rate": 2.4071532580997454e-06, + "loss": 1.2467, + "step": 10907 + }, + { + "epoch": 1.98, + "learning_rate": 2.407098653076083e-06, + "loss": 2.969, + "step": 10908 + }, + { + "epoch": 1.98, + "learning_rate": 2.407044048052421e-06, + "loss": 1.8137, + "step": 10909 + }, + { + "epoch": 1.98, + "learning_rate": 2.4069894430287584e-06, + "loss": 1.4167, + "step": 10910 + }, + { + "epoch": 1.98, + "learning_rate": 2.4069348380050963e-06, + "loss": 0.4546, + "step": 10911 + }, + { + "epoch": 1.98, + "learning_rate": 2.4068802329814343e-06, + "loss": 0.3427, + "step": 10912 + }, + { + "epoch": 1.98, + "learning_rate": 2.4068256279577722e-06, + "loss": 1.5646, + "step": 10913 + }, + { + "epoch": 1.98, + "learning_rate": 2.40677102293411e-06, + "loss": 0.355, + "step": 10914 + }, + { + "epoch": 1.98, + "learning_rate": 2.406716417910448e-06, + "loss": 1.1141, + "step": 10915 + }, + { + "epoch": 1.99, + "learning_rate": 2.4066618128867856e-06, + "loss": 0.3293, + "step": 10916 + }, + { + "epoch": 1.99, + "learning_rate": 2.4066072078631236e-06, + "loss": 0.3311, + "step": 10917 + }, + { + "epoch": 1.99, + "learning_rate": 2.4065526028394615e-06, + "loss": 0.7932, + "step": 10918 + }, + { + "epoch": 1.99, + "learning_rate": 2.406497997815799e-06, + "loss": 1.1248, + "step": 10919 + }, + { + "epoch": 1.99, + "learning_rate": 2.406443392792137e-06, + "loss": 1.0284, + "step": 10920 + }, + { + "epoch": 1.99, + "learning_rate": 2.4063887877684745e-06, + "loss": 1.1891, + "step": 10921 + }, + { + "epoch": 1.99, + "learning_rate": 2.4063341827448124e-06, + "loss": 1.9085, + "step": 10922 + }, + { + "epoch": 1.99, + "learning_rate": 2.4062795777211504e-06, + "loss": 2.3712, + "step": 10923 + }, + { + "epoch": 1.99, + "learning_rate": 2.406224972697488e-06, + "loss": 0.0439, + "step": 10924 + }, + { + "epoch": 1.99, + "learning_rate": 2.406170367673826e-06, + "loss": 1.0454, + "step": 10925 + }, + { + "epoch": 1.99, + "learning_rate": 2.4061157626501638e-06, + "loss": 0.9151, + "step": 10926 + }, + { + "epoch": 1.99, + "learning_rate": 2.4060611576265017e-06, + "loss": 0.1723, + "step": 10927 + }, + { + "epoch": 1.99, + "learning_rate": 2.4060065526028397e-06, + "loss": 0.0469, + "step": 10928 + }, + { + "epoch": 1.99, + "learning_rate": 2.4059519475791776e-06, + "loss": 0.1716, + "step": 10929 + }, + { + "epoch": 1.99, + "learning_rate": 2.405897342555515e-06, + "loss": 0.4192, + "step": 10930 + }, + { + "epoch": 1.99, + "learning_rate": 2.405842737531853e-06, + "loss": 0.574, + "step": 10931 + }, + { + "epoch": 1.99, + "learning_rate": 2.405788132508191e-06, + "loss": 2.1439, + "step": 10932 + }, + { + "epoch": 1.99, + "learning_rate": 2.4057335274845285e-06, + "loss": 1.2196, + "step": 10933 + }, + { + "epoch": 1.99, + "learning_rate": 2.4056789224608665e-06, + "loss": 1.2643, + "step": 10934 + }, + { + "epoch": 1.99, + "learning_rate": 2.405624317437204e-06, + "loss": 0.3472, + "step": 10935 + }, + { + "epoch": 1.99, + "learning_rate": 2.405569712413542e-06, + "loss": 1.3839, + "step": 10936 + }, + { + "epoch": 1.99, + "learning_rate": 2.40551510738988e-06, + "loss": 0.9446, + "step": 10937 + }, + { + "epoch": 1.99, + "learning_rate": 2.4054605023662174e-06, + "loss": 1.2667, + "step": 10938 + }, + { + "epoch": 1.99, + "learning_rate": 2.4054058973425554e-06, + "loss": 0.2311, + "step": 10939 + }, + { + "epoch": 1.99, + "learning_rate": 2.4053512923188933e-06, + "loss": 1.1773, + "step": 10940 + }, + { + "epoch": 1.99, + "learning_rate": 2.4052966872952312e-06, + "loss": 0.7079, + "step": 10941 + }, + { + "epoch": 1.99, + "learning_rate": 2.405242082271569e-06, + "loss": 0.6161, + "step": 10942 + }, + { + "epoch": 1.99, + "learning_rate": 2.405187477247907e-06, + "loss": 1.0355, + "step": 10943 + }, + { + "epoch": 1.99, + "learning_rate": 2.4051328722242446e-06, + "loss": 0.7994, + "step": 10944 + }, + { + "epoch": 1.99, + "learning_rate": 2.4050782672005826e-06, + "loss": 1.1943, + "step": 10945 + }, + { + "epoch": 1.99, + "learning_rate": 2.4050236621769205e-06, + "loss": 1.5481, + "step": 10946 + }, + { + "epoch": 1.99, + "learning_rate": 2.404969057153258e-06, + "loss": 0.8644, + "step": 10947 + }, + { + "epoch": 1.99, + "learning_rate": 2.404914452129596e-06, + "loss": 0.9101, + "step": 10948 + }, + { + "epoch": 1.99, + "learning_rate": 2.4048598471059335e-06, + "loss": 0.1131, + "step": 10949 + }, + { + "epoch": 1.99, + "learning_rate": 2.4048052420822715e-06, + "loss": 0.4163, + "step": 10950 + }, + { + "epoch": 1.99, + "learning_rate": 2.4047506370586094e-06, + "loss": 1.8949, + "step": 10951 + }, + { + "epoch": 1.99, + "learning_rate": 2.4046960320349473e-06, + "loss": 0.0205, + "step": 10952 + }, + { + "epoch": 1.99, + "learning_rate": 2.4046414270112853e-06, + "loss": 0.4344, + "step": 10953 + }, + { + "epoch": 1.99, + "learning_rate": 2.4045868219876232e-06, + "loss": 0.9897, + "step": 10954 + }, + { + "epoch": 1.99, + "learning_rate": 2.4045322169639608e-06, + "loss": 1.1617, + "step": 10955 + }, + { + "epoch": 1.99, + "learning_rate": 2.4044776119402987e-06, + "loss": 0.0421, + "step": 10956 + }, + { + "epoch": 1.99, + "learning_rate": 2.4044230069166366e-06, + "loss": 0.1227, + "step": 10957 + }, + { + "epoch": 1.99, + "learning_rate": 2.404368401892974e-06, + "loss": 1.548, + "step": 10958 + }, + { + "epoch": 1.99, + "learning_rate": 2.404313796869312e-06, + "loss": 1.4042, + "step": 10959 + }, + { + "epoch": 1.99, + "learning_rate": 2.40425919184565e-06, + "loss": 0.7755, + "step": 10960 + }, + { + "epoch": 1.99, + "learning_rate": 2.4042045868219876e-06, + "loss": 0.8343, + "step": 10961 + }, + { + "epoch": 1.99, + "learning_rate": 2.4041499817983255e-06, + "loss": 0.531, + "step": 10962 + }, + { + "epoch": 1.99, + "learning_rate": 2.404095376774663e-06, + "loss": 0.6319, + "step": 10963 + }, + { + "epoch": 1.99, + "learning_rate": 2.404040771751001e-06, + "loss": 1.1066, + "step": 10964 + }, + { + "epoch": 1.99, + "learning_rate": 2.403986166727339e-06, + "loss": 2.6691, + "step": 10965 + }, + { + "epoch": 1.99, + "learning_rate": 2.403931561703677e-06, + "loss": 1.121, + "step": 10966 + }, + { + "epoch": 1.99, + "learning_rate": 2.403876956680015e-06, + "loss": 0.4884, + "step": 10967 + }, + { + "epoch": 1.99, + "learning_rate": 2.4038223516563527e-06, + "loss": 3.0643, + "step": 10968 + }, + { + "epoch": 1.99, + "learning_rate": 2.4037677466326903e-06, + "loss": 0.6066, + "step": 10969 + }, + { + "epoch": 1.99, + "learning_rate": 2.403713141609028e-06, + "loss": 1.1943, + "step": 10970 + }, + { + "epoch": 2.0, + "learning_rate": 2.403658536585366e-06, + "loss": 0.8302, + "step": 10971 + }, + { + "epoch": 2.0, + "learning_rate": 2.4036039315617037e-06, + "loss": 0.4166, + "step": 10972 + }, + { + "epoch": 2.0, + "learning_rate": 2.4035493265380416e-06, + "loss": 0.4821, + "step": 10973 + }, + { + "epoch": 2.0, + "learning_rate": 2.4034947215143796e-06, + "loss": 0.7027, + "step": 10974 + }, + { + "epoch": 2.0, + "learning_rate": 2.403440116490717e-06, + "loss": 2.0722, + "step": 10975 + }, + { + "epoch": 2.0, + "learning_rate": 2.403385511467055e-06, + "loss": 0.5774, + "step": 10976 + }, + { + "epoch": 2.0, + "learning_rate": 2.4033309064433925e-06, + "loss": 0.8537, + "step": 10977 + }, + { + "epoch": 2.0, + "learning_rate": 2.4032763014197305e-06, + "loss": 0.0715, + "step": 10978 + }, + { + "epoch": 2.0, + "learning_rate": 2.4032216963960684e-06, + "loss": 3.291, + "step": 10979 + }, + { + "epoch": 2.0, + "learning_rate": 2.4031670913724064e-06, + "loss": 1.5945, + "step": 10980 + }, + { + "epoch": 2.0, + "learning_rate": 2.4031124863487443e-06, + "loss": 0.8776, + "step": 10981 + }, + { + "epoch": 2.0, + "learning_rate": 2.4030578813250823e-06, + "loss": 0.3562, + "step": 10982 + }, + { + "epoch": 2.0, + "learning_rate": 2.4030032763014198e-06, + "loss": 0.0211, + "step": 10983 + }, + { + "epoch": 2.0, + "learning_rate": 2.4029486712777577e-06, + "loss": 0.3854, + "step": 10984 + }, + { + "epoch": 2.0, + "learning_rate": 2.4028940662540957e-06, + "loss": 0.3746, + "step": 10985 + }, + { + "epoch": 2.0, + "learning_rate": 2.402839461230433e-06, + "loss": 0.1689, + "step": 10986 + }, + { + "epoch": 2.0, + "learning_rate": 2.402784856206771e-06, + "loss": 1.7037, + "step": 10987 + }, + { + "epoch": 2.0, + "learning_rate": 2.402730251183109e-06, + "loss": 1.9694, + "step": 10988 + }, + { + "epoch": 2.0, + "learning_rate": 2.4026756461594466e-06, + "loss": 2.3252, + "step": 10989 + }, + { + "epoch": 2.0, + "learning_rate": 2.4026210411357845e-06, + "loss": 0.2769, + "step": 10990 + }, + { + "epoch": 2.0, + "learning_rate": 2.402566436112122e-06, + "loss": 0.3908, + "step": 10991 + }, + { + "epoch": 2.0, + "learning_rate": 2.40251183108846e-06, + "loss": 0.5651, + "step": 10992 + }, + { + "epoch": 2.0, + "learning_rate": 2.402457226064798e-06, + "loss": 0.0369, + "step": 10993 + }, + { + "epoch": 2.0, + "learning_rate": 2.402402621041136e-06, + "loss": 0.6401, + "step": 10994 + }, + { + "epoch": 2.0, + "learning_rate": 2.402348016017474e-06, + "loss": 0.8335, + "step": 10995 + }, + { + "epoch": 2.0, + "learning_rate": 2.4022934109938118e-06, + "loss": 1.0704, + "step": 10996 + }, + { + "epoch": 2.0, + "learning_rate": 2.4022388059701493e-06, + "loss": 0.4128, + "step": 10997 + }, + { + "epoch": 2.0, + "learning_rate": 2.4021842009464872e-06, + "loss": 3.611, + "step": 10998 + }, + { + "epoch": 2.0, + "eval_accuracy": 0.7909165302782324, + "eval_loss": 0.8391110301017761, + "eval_runtime": 135.7706, + "eval_samples_per_second": 18.001, + "eval_steps_per_second": 4.5, + "step": 10998 + }, + { + "epoch": 2.0, + "learning_rate": 2.402129595922825e-06, + "loss": 0.9678, + "step": 10999 + }, + { + "epoch": 2.0, + "learning_rate": 2.4020749908991627e-06, + "loss": 0.3228, + "step": 11000 + }, + { + "epoch": 2.0, + "learning_rate": 2.4020203858755006e-06, + "loss": 1.2542, + "step": 11001 + }, + { + "epoch": 2.0, + "learning_rate": 2.401965780851838e-06, + "loss": 0.0564, + "step": 11002 + }, + { + "epoch": 2.0, + "learning_rate": 2.401911175828176e-06, + "loss": 0.1598, + "step": 11003 + }, + { + "epoch": 2.0, + "learning_rate": 2.401856570804514e-06, + "loss": 0.7218, + "step": 11004 + }, + { + "epoch": 2.0, + "learning_rate": 2.4018019657808516e-06, + "loss": 0.0523, + "step": 11005 + }, + { + "epoch": 2.0, + "learning_rate": 2.4017473607571895e-06, + "loss": 2.1533, + "step": 11006 + }, + { + "epoch": 2.0, + "learning_rate": 2.4016927557335275e-06, + "loss": 0.3017, + "step": 11007 + }, + { + "epoch": 2.0, + "learning_rate": 2.4016381507098654e-06, + "loss": 0.2808, + "step": 11008 + }, + { + "epoch": 2.0, + "learning_rate": 2.4015835456862033e-06, + "loss": 0.1047, + "step": 11009 + }, + { + "epoch": 2.0, + "learning_rate": 2.4015289406625413e-06, + "loss": 0.3787, + "step": 11010 + }, + { + "epoch": 2.0, + "learning_rate": 2.401474335638879e-06, + "loss": 1.8033, + "step": 11011 + }, + { + "epoch": 2.0, + "learning_rate": 2.4014197306152168e-06, + "loss": 0.4972, + "step": 11012 + }, + { + "epoch": 2.0, + "learning_rate": 2.4013651255915547e-06, + "loss": 2.4865, + "step": 11013 + }, + { + "epoch": 2.0, + "learning_rate": 2.4013105205678922e-06, + "loss": 0.0125, + "step": 11014 + }, + { + "epoch": 2.0, + "learning_rate": 2.40125591554423e-06, + "loss": 0.4726, + "step": 11015 + }, + { + "epoch": 2.0, + "learning_rate": 2.4012013105205677e-06, + "loss": 1.2732, + "step": 11016 + }, + { + "epoch": 2.0, + "learning_rate": 2.4011467054969056e-06, + "loss": 0.5259, + "step": 11017 + }, + { + "epoch": 2.0, + "learning_rate": 2.4010921004732436e-06, + "loss": 0.0887, + "step": 11018 + }, + { + "epoch": 2.0, + "learning_rate": 2.401037495449581e-06, + "loss": 2.8273, + "step": 11019 + }, + { + "epoch": 2.0, + "learning_rate": 2.400982890425919e-06, + "loss": 0.4311, + "step": 11020 + }, + { + "epoch": 2.0, + "learning_rate": 2.400928285402257e-06, + "loss": 0.5683, + "step": 11021 + }, + { + "epoch": 2.0, + "learning_rate": 2.400873680378595e-06, + "loss": 2.7342, + "step": 11022 + }, + { + "epoch": 2.0, + "learning_rate": 2.400819075354933e-06, + "loss": 0.7746, + "step": 11023 + }, + { + "epoch": 2.0, + "learning_rate": 2.400764470331271e-06, + "loss": 0.9149, + "step": 11024 + }, + { + "epoch": 2.0, + "learning_rate": 2.4007098653076083e-06, + "loss": 0.0676, + "step": 11025 + }, + { + "epoch": 2.01, + "learning_rate": 2.4006552602839463e-06, + "loss": 0.4648, + "step": 11026 + }, + { + "epoch": 2.01, + "learning_rate": 2.400600655260284e-06, + "loss": 0.9481, + "step": 11027 + }, + { + "epoch": 2.01, + "learning_rate": 2.4005460502366217e-06, + "loss": 0.3911, + "step": 11028 + }, + { + "epoch": 2.01, + "learning_rate": 2.4004914452129597e-06, + "loss": 0.0273, + "step": 11029 + }, + { + "epoch": 2.01, + "learning_rate": 2.400436840189297e-06, + "loss": 0.9155, + "step": 11030 + }, + { + "epoch": 2.01, + "learning_rate": 2.400382235165635e-06, + "loss": 0.533, + "step": 11031 + }, + { + "epoch": 2.01, + "learning_rate": 2.400327630141973e-06, + "loss": 1.5023, + "step": 11032 + }, + { + "epoch": 2.01, + "learning_rate": 2.400273025118311e-06, + "loss": 0.4902, + "step": 11033 + }, + { + "epoch": 2.01, + "learning_rate": 2.400218420094649e-06, + "loss": 3.7083, + "step": 11034 + }, + { + "epoch": 2.01, + "learning_rate": 2.400163815070987e-06, + "loss": 2.0874, + "step": 11035 + }, + { + "epoch": 2.01, + "learning_rate": 2.4001092100473244e-06, + "loss": 1.7432, + "step": 11036 + }, + { + "epoch": 2.01, + "learning_rate": 2.4000546050236624e-06, + "loss": 3.9902, + "step": 11037 + }, + { + "epoch": 2.01, + "learning_rate": 2.4000000000000003e-06, + "loss": 1.847, + "step": 11038 + }, + { + "epoch": 2.01, + "learning_rate": 2.399945394976338e-06, + "loss": 0.0589, + "step": 11039 + }, + { + "epoch": 2.01, + "learning_rate": 2.3998907899526758e-06, + "loss": 0.8018, + "step": 11040 + }, + { + "epoch": 2.01, + "learning_rate": 2.3998361849290137e-06, + "loss": 0.7734, + "step": 11041 + }, + { + "epoch": 2.01, + "learning_rate": 2.3997815799053512e-06, + "loss": 0.1938, + "step": 11042 + }, + { + "epoch": 2.01, + "learning_rate": 2.399726974881689e-06, + "loss": 2.4553, + "step": 11043 + }, + { + "epoch": 2.01, + "learning_rate": 2.3996723698580267e-06, + "loss": 0.4564, + "step": 11044 + }, + { + "epoch": 2.01, + "learning_rate": 2.3996177648343646e-06, + "loss": 0.0341, + "step": 11045 + }, + { + "epoch": 2.01, + "learning_rate": 2.3995631598107026e-06, + "loss": 0.4173, + "step": 11046 + }, + { + "epoch": 2.01, + "learning_rate": 2.3995085547870405e-06, + "loss": 0.3443, + "step": 11047 + }, + { + "epoch": 2.01, + "learning_rate": 2.3994539497633785e-06, + "loss": 0.1382, + "step": 11048 + }, + { + "epoch": 2.01, + "learning_rate": 2.3993993447397164e-06, + "loss": 1.5762, + "step": 11049 + }, + { + "epoch": 2.01, + "learning_rate": 2.399344739716054e-06, + "loss": 0.8747, + "step": 11050 + }, + { + "epoch": 2.01, + "learning_rate": 2.399290134692392e-06, + "loss": 0.706, + "step": 11051 + }, + { + "epoch": 2.01, + "learning_rate": 2.39923552966873e-06, + "loss": 1.3244, + "step": 11052 + }, + { + "epoch": 2.01, + "learning_rate": 2.3991809246450673e-06, + "loss": 0.0365, + "step": 11053 + }, + { + "epoch": 2.01, + "learning_rate": 2.3991263196214053e-06, + "loss": 1.1803, + "step": 11054 + }, + { + "epoch": 2.01, + "learning_rate": 2.3990717145977432e-06, + "loss": 2.8612, + "step": 11055 + }, + { + "epoch": 2.01, + "learning_rate": 2.3990171095740808e-06, + "loss": 0.5421, + "step": 11056 + }, + { + "epoch": 2.01, + "learning_rate": 2.3989625045504187e-06, + "loss": 1.1142, + "step": 11057 + }, + { + "epoch": 2.01, + "learning_rate": 2.3989078995267562e-06, + "loss": 0.07, + "step": 11058 + }, + { + "epoch": 2.01, + "learning_rate": 2.398853294503094e-06, + "loss": 1.0071, + "step": 11059 + }, + { + "epoch": 2.01, + "learning_rate": 2.398798689479432e-06, + "loss": 0.4275, + "step": 11060 + }, + { + "epoch": 2.01, + "learning_rate": 2.39874408445577e-06, + "loss": 0.0121, + "step": 11061 + }, + { + "epoch": 2.01, + "learning_rate": 2.398689479432108e-06, + "loss": 0.0316, + "step": 11062 + }, + { + "epoch": 2.01, + "learning_rate": 2.398634874408446e-06, + "loss": 0.6224, + "step": 11063 + }, + { + "epoch": 2.01, + "learning_rate": 2.3985802693847835e-06, + "loss": 0.2731, + "step": 11064 + }, + { + "epoch": 2.01, + "learning_rate": 2.3985256643611214e-06, + "loss": 0.8201, + "step": 11065 + }, + { + "epoch": 2.01, + "learning_rate": 2.3984710593374593e-06, + "loss": 1.2376, + "step": 11066 + }, + { + "epoch": 2.01, + "learning_rate": 2.398416454313797e-06, + "loss": 3.1763, + "step": 11067 + }, + { + "epoch": 2.01, + "learning_rate": 2.398361849290135e-06, + "loss": 1.0768, + "step": 11068 + }, + { + "epoch": 2.01, + "learning_rate": 2.3983072442664727e-06, + "loss": 0.9509, + "step": 11069 + }, + { + "epoch": 2.01, + "learning_rate": 2.3982526392428103e-06, + "loss": 1.9347, + "step": 11070 + }, + { + "epoch": 2.01, + "learning_rate": 2.3981980342191482e-06, + "loss": 1.0254, + "step": 11071 + }, + { + "epoch": 2.01, + "learning_rate": 2.3981434291954857e-06, + "loss": 0.0794, + "step": 11072 + }, + { + "epoch": 2.01, + "learning_rate": 2.3980888241718237e-06, + "loss": 0.5757, + "step": 11073 + }, + { + "epoch": 2.01, + "learning_rate": 2.3980342191481616e-06, + "loss": 2.274, + "step": 11074 + }, + { + "epoch": 2.01, + "learning_rate": 2.3979796141244996e-06, + "loss": 1.3031, + "step": 11075 + }, + { + "epoch": 2.01, + "learning_rate": 2.3979250091008375e-06, + "loss": 0.2738, + "step": 11076 + }, + { + "epoch": 2.01, + "learning_rate": 2.3978704040771754e-06, + "loss": 1.813, + "step": 11077 + }, + { + "epoch": 2.01, + "learning_rate": 2.397815799053513e-06, + "loss": 1.1402, + "step": 11078 + }, + { + "epoch": 2.01, + "learning_rate": 2.397761194029851e-06, + "loss": 0.78, + "step": 11079 + }, + { + "epoch": 2.01, + "learning_rate": 2.397706589006189e-06, + "loss": 1.7361, + "step": 11080 + }, + { + "epoch": 2.02, + "learning_rate": 2.3976519839825264e-06, + "loss": 0.402, + "step": 11081 + }, + { + "epoch": 2.02, + "learning_rate": 2.3975973789588643e-06, + "loss": 0.2878, + "step": 11082 + }, + { + "epoch": 2.02, + "learning_rate": 2.397542773935202e-06, + "loss": 0.5581, + "step": 11083 + }, + { + "epoch": 2.02, + "learning_rate": 2.3974881689115398e-06, + "loss": 0.8141, + "step": 11084 + }, + { + "epoch": 2.02, + "learning_rate": 2.3974335638878777e-06, + "loss": 1.2693, + "step": 11085 + }, + { + "epoch": 2.02, + "learning_rate": 2.3973789588642152e-06, + "loss": 1.0727, + "step": 11086 + }, + { + "epoch": 2.02, + "learning_rate": 2.397324353840553e-06, + "loss": 0.7484, + "step": 11087 + }, + { + "epoch": 2.02, + "learning_rate": 2.397269748816891e-06, + "loss": 0.8283, + "step": 11088 + }, + { + "epoch": 2.02, + "learning_rate": 2.397215143793229e-06, + "loss": 1.5479, + "step": 11089 + }, + { + "epoch": 2.02, + "learning_rate": 2.397160538769567e-06, + "loss": 1.8412, + "step": 11090 + }, + { + "epoch": 2.02, + "learning_rate": 2.397105933745905e-06, + "loss": 1.1518, + "step": 11091 + }, + { + "epoch": 2.02, + "learning_rate": 2.3970513287222425e-06, + "loss": 0.3734, + "step": 11092 + }, + { + "epoch": 2.02, + "learning_rate": 2.3969967236985804e-06, + "loss": 0.506, + "step": 11093 + }, + { + "epoch": 2.02, + "learning_rate": 2.3969421186749184e-06, + "loss": 3.2504, + "step": 11094 + }, + { + "epoch": 2.02, + "learning_rate": 2.396887513651256e-06, + "loss": 0.2451, + "step": 11095 + }, + { + "epoch": 2.02, + "learning_rate": 2.396832908627594e-06, + "loss": 0.239, + "step": 11096 + }, + { + "epoch": 2.02, + "learning_rate": 2.3967783036039314e-06, + "loss": 2.1906, + "step": 11097 + }, + { + "epoch": 2.02, + "learning_rate": 2.3967236985802693e-06, + "loss": 0.9057, + "step": 11098 + }, + { + "epoch": 2.02, + "learning_rate": 2.3966690935566072e-06, + "loss": 2.1158, + "step": 11099 + }, + { + "epoch": 2.02, + "learning_rate": 2.3966144885329448e-06, + "loss": 0.5165, + "step": 11100 + }, + { + "epoch": 2.02, + "learning_rate": 2.3965598835092827e-06, + "loss": 1.0572, + "step": 11101 + }, + { + "epoch": 2.02, + "learning_rate": 2.3965052784856206e-06, + "loss": 1.6034, + "step": 11102 + }, + { + "epoch": 2.02, + "learning_rate": 2.3964506734619586e-06, + "loss": 0.6231, + "step": 11103 + }, + { + "epoch": 2.02, + "learning_rate": 2.3963960684382965e-06, + "loss": 0.0212, + "step": 11104 + }, + { + "epoch": 2.02, + "learning_rate": 2.3963414634146345e-06, + "loss": 0.4629, + "step": 11105 + }, + { + "epoch": 2.02, + "learning_rate": 2.396286858390972e-06, + "loss": 0.476, + "step": 11106 + }, + { + "epoch": 2.02, + "learning_rate": 2.39623225336731e-06, + "loss": 0.4008, + "step": 11107 + }, + { + "epoch": 2.02, + "learning_rate": 2.396177648343648e-06, + "loss": 0.4813, + "step": 11108 + }, + { + "epoch": 2.02, + "learning_rate": 2.3961230433199854e-06, + "loss": 1.4784, + "step": 11109 + }, + { + "epoch": 2.02, + "learning_rate": 2.3960684382963233e-06, + "loss": 1.9313, + "step": 11110 + }, + { + "epoch": 2.02, + "learning_rate": 2.396013833272661e-06, + "loss": 1.0414, + "step": 11111 + }, + { + "epoch": 2.02, + "learning_rate": 2.395959228248999e-06, + "loss": 1.1462, + "step": 11112 + }, + { + "epoch": 2.02, + "learning_rate": 2.3959046232253368e-06, + "loss": 0.6698, + "step": 11113 + }, + { + "epoch": 2.02, + "learning_rate": 2.3958500182016747e-06, + "loss": 0.7112, + "step": 11114 + }, + { + "epoch": 2.02, + "learning_rate": 2.3957954131780126e-06, + "loss": 0.4465, + "step": 11115 + }, + { + "epoch": 2.02, + "learning_rate": 2.39574080815435e-06, + "loss": 0.7315, + "step": 11116 + }, + { + "epoch": 2.02, + "learning_rate": 2.395686203130688e-06, + "loss": 0.2852, + "step": 11117 + }, + { + "epoch": 2.02, + "learning_rate": 2.395631598107026e-06, + "loss": 1.48, + "step": 11118 + }, + { + "epoch": 2.02, + "learning_rate": 2.395576993083364e-06, + "loss": 0.7252, + "step": 11119 + }, + { + "epoch": 2.02, + "learning_rate": 2.3955223880597015e-06, + "loss": 0.1089, + "step": 11120 + }, + { + "epoch": 2.02, + "learning_rate": 2.3954677830360395e-06, + "loss": 0.0174, + "step": 11121 + }, + { + "epoch": 2.02, + "learning_rate": 2.3954131780123774e-06, + "loss": 1.4575, + "step": 11122 + }, + { + "epoch": 2.02, + "learning_rate": 2.395358572988715e-06, + "loss": 0.7176, + "step": 11123 + }, + { + "epoch": 2.02, + "learning_rate": 2.395303967965053e-06, + "loss": 0.5328, + "step": 11124 + }, + { + "epoch": 2.02, + "learning_rate": 2.3952493629413904e-06, + "loss": 1.4078, + "step": 11125 + }, + { + "epoch": 2.02, + "learning_rate": 2.3951947579177283e-06, + "loss": 1.2804, + "step": 11126 + }, + { + "epoch": 2.02, + "learning_rate": 2.3951401528940663e-06, + "loss": 0.1542, + "step": 11127 + }, + { + "epoch": 2.02, + "learning_rate": 2.395085547870404e-06, + "loss": 1.5749, + "step": 11128 + }, + { + "epoch": 2.02, + "learning_rate": 2.395030942846742e-06, + "loss": 1.0955, + "step": 11129 + }, + { + "epoch": 2.02, + "learning_rate": 2.39497633782308e-06, + "loss": 0.3061, + "step": 11130 + }, + { + "epoch": 2.02, + "learning_rate": 2.3949217327994176e-06, + "loss": 0.4233, + "step": 11131 + }, + { + "epoch": 2.02, + "learning_rate": 2.3948671277757556e-06, + "loss": 0.1688, + "step": 11132 + }, + { + "epoch": 2.02, + "learning_rate": 2.3948125227520935e-06, + "loss": 1.4786, + "step": 11133 + }, + { + "epoch": 2.02, + "learning_rate": 2.394757917728431e-06, + "loss": 4.0656, + "step": 11134 + }, + { + "epoch": 2.02, + "learning_rate": 2.394703312704769e-06, + "loss": 1.7964, + "step": 11135 + }, + { + "epoch": 2.03, + "learning_rate": 2.394648707681107e-06, + "loss": 0.0265, + "step": 11136 + }, + { + "epoch": 2.03, + "learning_rate": 2.3945941026574444e-06, + "loss": 0.3714, + "step": 11137 + }, + { + "epoch": 2.03, + "learning_rate": 2.3945394976337824e-06, + "loss": 1.0962, + "step": 11138 + }, + { + "epoch": 2.03, + "learning_rate": 2.39448489261012e-06, + "loss": 0.9746, + "step": 11139 + }, + { + "epoch": 2.03, + "learning_rate": 2.394430287586458e-06, + "loss": 1.0195, + "step": 11140 + }, + { + "epoch": 2.03, + "learning_rate": 2.3943756825627958e-06, + "loss": 0.326, + "step": 11141 + }, + { + "epoch": 2.03, + "learning_rate": 2.3943210775391337e-06, + "loss": 0.4632, + "step": 11142 + }, + { + "epoch": 2.03, + "learning_rate": 2.3942664725154717e-06, + "loss": 1.1276, + "step": 11143 + }, + { + "epoch": 2.03, + "learning_rate": 2.3942118674918096e-06, + "loss": 0.4757, + "step": 11144 + }, + { + "epoch": 2.03, + "learning_rate": 2.394157262468147e-06, + "loss": 0.3101, + "step": 11145 + }, + { + "epoch": 2.03, + "learning_rate": 2.394102657444485e-06, + "loss": 0.3789, + "step": 11146 + }, + { + "epoch": 2.03, + "learning_rate": 2.394048052420823e-06, + "loss": 0.4132, + "step": 11147 + }, + { + "epoch": 2.03, + "learning_rate": 2.3939934473971605e-06, + "loss": 0.3675, + "step": 11148 + }, + { + "epoch": 2.03, + "learning_rate": 2.3939388423734985e-06, + "loss": 0.8124, + "step": 11149 + }, + { + "epoch": 2.03, + "learning_rate": 2.393884237349836e-06, + "loss": 1.9583, + "step": 11150 + }, + { + "epoch": 2.03, + "learning_rate": 2.393829632326174e-06, + "loss": 0.7066, + "step": 11151 + }, + { + "epoch": 2.03, + "learning_rate": 2.393775027302512e-06, + "loss": 1.0524, + "step": 11152 + }, + { + "epoch": 2.03, + "learning_rate": 2.3937204222788494e-06, + "loss": 1.7695, + "step": 11153 + }, + { + "epoch": 2.03, + "learning_rate": 2.3936658172551873e-06, + "loss": 0.4198, + "step": 11154 + }, + { + "epoch": 2.03, + "learning_rate": 2.3936112122315253e-06, + "loss": 2.5538, + "step": 11155 + }, + { + "epoch": 2.03, + "learning_rate": 2.3935566072078632e-06, + "loss": 0.1475, + "step": 11156 + }, + { + "epoch": 2.03, + "learning_rate": 2.393502002184201e-06, + "loss": 0.3841, + "step": 11157 + }, + { + "epoch": 2.03, + "learning_rate": 2.393447397160539e-06, + "loss": 0.3662, + "step": 11158 + }, + { + "epoch": 2.03, + "learning_rate": 2.3933927921368766e-06, + "loss": 0.6298, + "step": 11159 + }, + { + "epoch": 2.03, + "learning_rate": 2.3933381871132146e-06, + "loss": 2.107, + "step": 11160 + }, + { + "epoch": 2.03, + "learning_rate": 2.3932835820895525e-06, + "loss": 2.3055, + "step": 11161 + }, + { + "epoch": 2.03, + "learning_rate": 2.39322897706589e-06, + "loss": 0.8774, + "step": 11162 + }, + { + "epoch": 2.03, + "learning_rate": 2.393174372042228e-06, + "loss": 0.7326, + "step": 11163 + }, + { + "epoch": 2.03, + "learning_rate": 2.3931197670185655e-06, + "loss": 0.5649, + "step": 11164 + }, + { + "epoch": 2.03, + "learning_rate": 2.3930651619949035e-06, + "loss": 0.9128, + "step": 11165 + }, + { + "epoch": 2.03, + "learning_rate": 2.3930105569712414e-06, + "loss": 0.3791, + "step": 11166 + }, + { + "epoch": 2.03, + "learning_rate": 2.392955951947579e-06, + "loss": 1.2183, + "step": 11167 + }, + { + "epoch": 2.03, + "learning_rate": 2.392901346923917e-06, + "loss": 0.9525, + "step": 11168 + }, + { + "epoch": 2.03, + "learning_rate": 2.392846741900255e-06, + "loss": 0.4776, + "step": 11169 + }, + { + "epoch": 2.03, + "learning_rate": 2.3927921368765927e-06, + "loss": 0.0422, + "step": 11170 + }, + { + "epoch": 2.03, + "learning_rate": 2.3927375318529307e-06, + "loss": 1.3932, + "step": 11171 + }, + { + "epoch": 2.03, + "learning_rate": 2.3926829268292686e-06, + "loss": 1.0152, + "step": 11172 + }, + { + "epoch": 2.03, + "learning_rate": 2.392628321805606e-06, + "loss": 0.84, + "step": 11173 + }, + { + "epoch": 2.03, + "learning_rate": 2.392573716781944e-06, + "loss": 0.8907, + "step": 11174 + }, + { + "epoch": 2.03, + "learning_rate": 2.392519111758282e-06, + "loss": 0.3479, + "step": 11175 + }, + { + "epoch": 2.03, + "learning_rate": 2.3924645067346196e-06, + "loss": 0.3036, + "step": 11176 + }, + { + "epoch": 2.03, + "learning_rate": 2.3924099017109575e-06, + "loss": 0.2081, + "step": 11177 + }, + { + "epoch": 2.03, + "learning_rate": 2.392355296687295e-06, + "loss": 0.0463, + "step": 11178 + }, + { + "epoch": 2.03, + "learning_rate": 2.392300691663633e-06, + "loss": 2.801, + "step": 11179 + }, + { + "epoch": 2.03, + "learning_rate": 2.392246086639971e-06, + "loss": 0.0258, + "step": 11180 + }, + { + "epoch": 2.03, + "learning_rate": 2.3921914816163084e-06, + "loss": 1.3698, + "step": 11181 + }, + { + "epoch": 2.03, + "learning_rate": 2.3921368765926464e-06, + "loss": 1.2533, + "step": 11182 + }, + { + "epoch": 2.03, + "learning_rate": 2.3920822715689843e-06, + "loss": 0.0893, + "step": 11183 + }, + { + "epoch": 2.03, + "learning_rate": 2.3920276665453223e-06, + "loss": 2.4852, + "step": 11184 + }, + { + "epoch": 2.03, + "learning_rate": 2.39197306152166e-06, + "loss": 0.7156, + "step": 11185 + }, + { + "epoch": 2.03, + "learning_rate": 2.391918456497998e-06, + "loss": 0.6996, + "step": 11186 + }, + { + "epoch": 2.03, + "learning_rate": 2.3918638514743357e-06, + "loss": 0.6635, + "step": 11187 + }, + { + "epoch": 2.03, + "learning_rate": 2.3918092464506736e-06, + "loss": 1.8815, + "step": 11188 + }, + { + "epoch": 2.03, + "learning_rate": 2.3917546414270116e-06, + "loss": 0.7241, + "step": 11189 + }, + { + "epoch": 2.03, + "learning_rate": 2.391700036403349e-06, + "loss": 0.3655, + "step": 11190 + }, + { + "epoch": 2.04, + "learning_rate": 2.391645431379687e-06, + "loss": 1.3396, + "step": 11191 + }, + { + "epoch": 2.04, + "learning_rate": 2.3915908263560245e-06, + "loss": 2.6342, + "step": 11192 + }, + { + "epoch": 2.04, + "learning_rate": 2.3915362213323625e-06, + "loss": 0.6791, + "step": 11193 + }, + { + "epoch": 2.04, + "learning_rate": 2.3914816163087004e-06, + "loss": 0.0349, + "step": 11194 + }, + { + "epoch": 2.04, + "learning_rate": 2.391427011285038e-06, + "loss": 0.0864, + "step": 11195 + }, + { + "epoch": 2.04, + "learning_rate": 2.391372406261376e-06, + "loss": 0.4542, + "step": 11196 + }, + { + "epoch": 2.04, + "learning_rate": 2.391317801237714e-06, + "loss": 2.6588, + "step": 11197 + }, + { + "epoch": 2.04, + "learning_rate": 2.3912631962140518e-06, + "loss": 1.2639, + "step": 11198 + }, + { + "epoch": 2.04, + "learning_rate": 2.3912085911903897e-06, + "loss": 0.0437, + "step": 11199 + }, + { + "epoch": 2.04, + "learning_rate": 2.3911539861667277e-06, + "loss": 0.042, + "step": 11200 + }, + { + "epoch": 2.04, + "learning_rate": 2.391099381143065e-06, + "loss": 0.3472, + "step": 11201 + }, + { + "epoch": 2.04, + "learning_rate": 2.391044776119403e-06, + "loss": 1.015, + "step": 11202 + }, + { + "epoch": 2.04, + "learning_rate": 2.390990171095741e-06, + "loss": 0.15, + "step": 11203 + }, + { + "epoch": 2.04, + "learning_rate": 2.3909355660720786e-06, + "loss": 0.4174, + "step": 11204 + }, + { + "epoch": 2.04, + "learning_rate": 2.3908809610484165e-06, + "loss": 0.8948, + "step": 11205 + }, + { + "epoch": 2.04, + "learning_rate": 2.390826356024754e-06, + "loss": 1.4, + "step": 11206 + }, + { + "epoch": 2.04, + "learning_rate": 2.390771751001092e-06, + "loss": 0.7933, + "step": 11207 + }, + { + "epoch": 2.04, + "learning_rate": 2.39071714597743e-06, + "loss": 0.42, + "step": 11208 + }, + { + "epoch": 2.04, + "learning_rate": 2.390662540953768e-06, + "loss": 1.6577, + "step": 11209 + }, + { + "epoch": 2.04, + "learning_rate": 2.390607935930106e-06, + "loss": 0.2511, + "step": 11210 + }, + { + "epoch": 2.04, + "learning_rate": 2.3905533309064438e-06, + "loss": 1.191, + "step": 11211 + }, + { + "epoch": 2.04, + "learning_rate": 2.3904987258827813e-06, + "loss": 2.0215, + "step": 11212 + }, + { + "epoch": 2.04, + "learning_rate": 2.3904441208591192e-06, + "loss": 0.7319, + "step": 11213 + }, + { + "epoch": 2.04, + "learning_rate": 2.390389515835457e-06, + "loss": 0.7209, + "step": 11214 + }, + { + "epoch": 2.04, + "learning_rate": 2.3903349108117947e-06, + "loss": 2.3002, + "step": 11215 + }, + { + "epoch": 2.04, + "learning_rate": 2.3902803057881326e-06, + "loss": 1.6145, + "step": 11216 + }, + { + "epoch": 2.04, + "learning_rate": 2.3902257007644706e-06, + "loss": 1.9112, + "step": 11217 + }, + { + "epoch": 2.04, + "learning_rate": 2.390171095740808e-06, + "loss": 0.0184, + "step": 11218 + }, + { + "epoch": 2.04, + "learning_rate": 2.390116490717146e-06, + "loss": 0.0501, + "step": 11219 + }, + { + "epoch": 2.04, + "learning_rate": 2.3900618856934836e-06, + "loss": 0.3048, + "step": 11220 + }, + { + "epoch": 2.04, + "learning_rate": 2.3900072806698215e-06, + "loss": 0.035, + "step": 11221 + }, + { + "epoch": 2.04, + "learning_rate": 2.3899526756461595e-06, + "loss": 1.042, + "step": 11222 + }, + { + "epoch": 2.04, + "learning_rate": 2.3898980706224974e-06, + "loss": 0.0664, + "step": 11223 + }, + { + "epoch": 2.04, + "learning_rate": 2.3898434655988353e-06, + "loss": 1.328, + "step": 11224 + }, + { + "epoch": 2.04, + "learning_rate": 2.3897888605751733e-06, + "loss": 0.0916, + "step": 11225 + }, + { + "epoch": 2.04, + "learning_rate": 2.389734255551511e-06, + "loss": 0.4842, + "step": 11226 + }, + { + "epoch": 2.04, + "learning_rate": 2.3896796505278487e-06, + "loss": 0.3289, + "step": 11227 + }, + { + "epoch": 2.04, + "learning_rate": 2.3896250455041867e-06, + "loss": 0.7836, + "step": 11228 + }, + { + "epoch": 2.04, + "learning_rate": 2.389570440480524e-06, + "loss": 2.3127, + "step": 11229 + }, + { + "epoch": 2.04, + "learning_rate": 2.389515835456862e-06, + "loss": 0.7521, + "step": 11230 + }, + { + "epoch": 2.04, + "learning_rate": 2.3894612304331997e-06, + "loss": 0.6869, + "step": 11231 + }, + { + "epoch": 2.04, + "learning_rate": 2.3894066254095376e-06, + "loss": 0.048, + "step": 11232 + }, + { + "epoch": 2.04, + "learning_rate": 2.3893520203858756e-06, + "loss": 0.0427, + "step": 11233 + }, + { + "epoch": 2.04, + "learning_rate": 2.389297415362213e-06, + "loss": 0.5083, + "step": 11234 + }, + { + "epoch": 2.04, + "learning_rate": 2.389242810338551e-06, + "loss": 0.1832, + "step": 11235 + }, + { + "epoch": 2.04, + "learning_rate": 2.389188205314889e-06, + "loss": 2.3883, + "step": 11236 + }, + { + "epoch": 2.04, + "learning_rate": 2.389133600291227e-06, + "loss": 0.9514, + "step": 11237 + }, + { + "epoch": 2.04, + "learning_rate": 2.389078995267565e-06, + "loss": 0.1702, + "step": 11238 + }, + { + "epoch": 2.04, + "learning_rate": 2.389024390243903e-06, + "loss": 2.3281, + "step": 11239 + }, + { + "epoch": 2.04, + "learning_rate": 2.3889697852202403e-06, + "loss": 0.0762, + "step": 11240 + }, + { + "epoch": 2.04, + "learning_rate": 2.3889151801965783e-06, + "loss": 1.143, + "step": 11241 + }, + { + "epoch": 2.04, + "learning_rate": 2.388860575172916e-06, + "loss": 0.8489, + "step": 11242 + }, + { + "epoch": 2.04, + "learning_rate": 2.3888059701492537e-06, + "loss": 1.6727, + "step": 11243 + }, + { + "epoch": 2.04, + "learning_rate": 2.3887513651255917e-06, + "loss": 0.8775, + "step": 11244 + }, + { + "epoch": 2.04, + "learning_rate": 2.388696760101929e-06, + "loss": 2.1033, + "step": 11245 + }, + { + "epoch": 2.05, + "learning_rate": 2.388642155078267e-06, + "loss": 0.5564, + "step": 11246 + }, + { + "epoch": 2.05, + "learning_rate": 2.388587550054605e-06, + "loss": 0.3821, + "step": 11247 + }, + { + "epoch": 2.05, + "learning_rate": 2.3885329450309426e-06, + "loss": 0.5974, + "step": 11248 + }, + { + "epoch": 2.05, + "learning_rate": 2.3884783400072805e-06, + "loss": 2.2553, + "step": 11249 + }, + { + "epoch": 2.05, + "learning_rate": 2.3884237349836185e-06, + "loss": 0.5127, + "step": 11250 + }, + { + "epoch": 2.05, + "learning_rate": 2.3883691299599564e-06, + "loss": 0.5617, + "step": 11251 + }, + { + "epoch": 2.05, + "learning_rate": 2.3883145249362944e-06, + "loss": 0.0792, + "step": 11252 + }, + { + "epoch": 2.05, + "learning_rate": 2.3882599199126323e-06, + "loss": 2.754, + "step": 11253 + }, + { + "epoch": 2.05, + "learning_rate": 2.38820531488897e-06, + "loss": 1.5845, + "step": 11254 + }, + { + "epoch": 2.05, + "learning_rate": 2.3881507098653078e-06, + "loss": 0.5649, + "step": 11255 + }, + { + "epoch": 2.05, + "learning_rate": 2.3880961048416457e-06, + "loss": 0.5657, + "step": 11256 + }, + { + "epoch": 2.05, + "learning_rate": 2.3880414998179832e-06, + "loss": 0.7676, + "step": 11257 + }, + { + "epoch": 2.05, + "learning_rate": 2.387986894794321e-06, + "loss": 1.0056, + "step": 11258 + }, + { + "epoch": 2.05, + "learning_rate": 2.3879322897706587e-06, + "loss": 0.1816, + "step": 11259 + }, + { + "epoch": 2.05, + "learning_rate": 2.3878776847469966e-06, + "loss": 0.9301, + "step": 11260 + }, + { + "epoch": 2.05, + "learning_rate": 2.3878230797233346e-06, + "loss": 0.5501, + "step": 11261 + }, + { + "epoch": 2.05, + "learning_rate": 2.387768474699672e-06, + "loss": 2.028, + "step": 11262 + }, + { + "epoch": 2.05, + "learning_rate": 2.38771386967601e-06, + "loss": 0.1559, + "step": 11263 + }, + { + "epoch": 2.05, + "learning_rate": 2.387659264652348e-06, + "loss": 0.6079, + "step": 11264 + }, + { + "epoch": 2.05, + "learning_rate": 2.387604659628686e-06, + "loss": 0.26, + "step": 11265 + }, + { + "epoch": 2.05, + "learning_rate": 2.387550054605024e-06, + "loss": 1.5785, + "step": 11266 + }, + { + "epoch": 2.05, + "learning_rate": 2.387495449581362e-06, + "loss": 1.5769, + "step": 11267 + }, + { + "epoch": 2.05, + "learning_rate": 2.3874408445576993e-06, + "loss": 1.0012, + "step": 11268 + }, + { + "epoch": 2.05, + "learning_rate": 2.3873862395340373e-06, + "loss": 0.4253, + "step": 11269 + }, + { + "epoch": 2.05, + "learning_rate": 2.3873316345103752e-06, + "loss": 0.5799, + "step": 11270 + }, + { + "epoch": 2.05, + "learning_rate": 2.3872770294867128e-06, + "loss": 0.7195, + "step": 11271 + }, + { + "epoch": 2.05, + "learning_rate": 2.3872224244630507e-06, + "loss": 1.1716, + "step": 11272 + }, + { + "epoch": 2.05, + "learning_rate": 2.3871678194393882e-06, + "loss": 0.6979, + "step": 11273 + }, + { + "epoch": 2.05, + "learning_rate": 2.387113214415726e-06, + "loss": 0.1021, + "step": 11274 + }, + { + "epoch": 2.05, + "learning_rate": 2.387058609392064e-06, + "loss": 0.3743, + "step": 11275 + }, + { + "epoch": 2.05, + "learning_rate": 2.3870040043684016e-06, + "loss": 1.6064, + "step": 11276 + }, + { + "epoch": 2.05, + "learning_rate": 2.3869493993447396e-06, + "loss": 0.3093, + "step": 11277 + }, + { + "epoch": 2.05, + "learning_rate": 2.3868947943210775e-06, + "loss": 1.2319, + "step": 11278 + }, + { + "epoch": 2.05, + "learning_rate": 2.3868401892974155e-06, + "loss": 0.6528, + "step": 11279 + }, + { + "epoch": 2.05, + "learning_rate": 2.3867855842737534e-06, + "loss": 0.0636, + "step": 11280 + }, + { + "epoch": 2.05, + "learning_rate": 2.3867309792500913e-06, + "loss": 0.3902, + "step": 11281 + }, + { + "epoch": 2.05, + "learning_rate": 2.386676374226429e-06, + "loss": 1.6109, + "step": 11282 + }, + { + "epoch": 2.05, + "learning_rate": 2.386621769202767e-06, + "loss": 0.8498, + "step": 11283 + }, + { + "epoch": 2.05, + "learning_rate": 2.3865671641791047e-06, + "loss": 0.7877, + "step": 11284 + }, + { + "epoch": 2.05, + "learning_rate": 2.3865125591554423e-06, + "loss": 0.6542, + "step": 11285 + }, + { + "epoch": 2.05, + "learning_rate": 2.38645795413178e-06, + "loss": 0.0733, + "step": 11286 + }, + { + "epoch": 2.05, + "learning_rate": 2.3864033491081177e-06, + "loss": 0.0564, + "step": 11287 + }, + { + "epoch": 2.05, + "learning_rate": 2.3863487440844557e-06, + "loss": 0.0346, + "step": 11288 + }, + { + "epoch": 2.05, + "learning_rate": 2.3862941390607936e-06, + "loss": 0.2559, + "step": 11289 + }, + { + "epoch": 2.05, + "learning_rate": 2.3862395340371316e-06, + "loss": 0.5182, + "step": 11290 + }, + { + "epoch": 2.05, + "learning_rate": 2.3861849290134695e-06, + "loss": 0.6501, + "step": 11291 + }, + { + "epoch": 2.05, + "learning_rate": 2.3861303239898074e-06, + "loss": 1.9128, + "step": 11292 + }, + { + "epoch": 2.05, + "learning_rate": 2.386075718966145e-06, + "loss": 0.6419, + "step": 11293 + }, + { + "epoch": 2.05, + "learning_rate": 2.386021113942483e-06, + "loss": 0.3106, + "step": 11294 + }, + { + "epoch": 2.05, + "learning_rate": 2.385966508918821e-06, + "loss": 0.0827, + "step": 11295 + }, + { + "epoch": 2.05, + "learning_rate": 2.3859119038951584e-06, + "loss": 0.3451, + "step": 11296 + }, + { + "epoch": 2.05, + "learning_rate": 2.3858572988714963e-06, + "loss": 1.859, + "step": 11297 + }, + { + "epoch": 2.05, + "learning_rate": 2.3858026938478343e-06, + "loss": 0.9476, + "step": 11298 + }, + { + "epoch": 2.05, + "learning_rate": 2.3857480888241718e-06, + "loss": 0.8216, + "step": 11299 + }, + { + "epoch": 2.05, + "learning_rate": 2.3856934838005097e-06, + "loss": 1.622, + "step": 11300 + }, + { + "epoch": 2.06, + "learning_rate": 2.3856388787768472e-06, + "loss": 2.6052, + "step": 11301 + }, + { + "epoch": 2.06, + "learning_rate": 2.385584273753185e-06, + "loss": 0.0402, + "step": 11302 + }, + { + "epoch": 2.06, + "learning_rate": 2.385529668729523e-06, + "loss": 0.0346, + "step": 11303 + }, + { + "epoch": 2.06, + "learning_rate": 2.385475063705861e-06, + "loss": 0.5767, + "step": 11304 + }, + { + "epoch": 2.06, + "learning_rate": 2.385420458682199e-06, + "loss": 0.732, + "step": 11305 + }, + { + "epoch": 2.06, + "learning_rate": 2.385365853658537e-06, + "loss": 1.7316, + "step": 11306 + }, + { + "epoch": 2.06, + "learning_rate": 2.3853112486348745e-06, + "loss": 2.2951, + "step": 11307 + }, + { + "epoch": 2.06, + "learning_rate": 2.3852566436112124e-06, + "loss": 0.9554, + "step": 11308 + }, + { + "epoch": 2.06, + "learning_rate": 2.3852020385875504e-06, + "loss": 2.0285, + "step": 11309 + }, + { + "epoch": 2.06, + "learning_rate": 2.385147433563888e-06, + "loss": 1.1572, + "step": 11310 + }, + { + "epoch": 2.06, + "learning_rate": 2.385092828540226e-06, + "loss": 0.0501, + "step": 11311 + }, + { + "epoch": 2.06, + "learning_rate": 2.3850382235165633e-06, + "loss": 1.3608, + "step": 11312 + }, + { + "epoch": 2.06, + "learning_rate": 2.3849836184929013e-06, + "loss": 0.5426, + "step": 11313 + }, + { + "epoch": 2.06, + "learning_rate": 2.3849290134692392e-06, + "loss": 1.296, + "step": 11314 + }, + { + "epoch": 2.06, + "learning_rate": 2.3848744084455768e-06, + "loss": 1.6014, + "step": 11315 + }, + { + "epoch": 2.06, + "learning_rate": 2.3848198034219147e-06, + "loss": 0.2342, + "step": 11316 + }, + { + "epoch": 2.06, + "learning_rate": 2.3847651983982526e-06, + "loss": 1.7954, + "step": 11317 + }, + { + "epoch": 2.06, + "learning_rate": 2.3847105933745906e-06, + "loss": 2.3773, + "step": 11318 + }, + { + "epoch": 2.06, + "learning_rate": 2.3846559883509285e-06, + "loss": 0.0241, + "step": 11319 + }, + { + "epoch": 2.06, + "learning_rate": 2.3846013833272665e-06, + "loss": 0.2666, + "step": 11320 + }, + { + "epoch": 2.06, + "learning_rate": 2.384546778303604e-06, + "loss": 1.3948, + "step": 11321 + }, + { + "epoch": 2.06, + "learning_rate": 2.384492173279942e-06, + "loss": 1.1213, + "step": 11322 + }, + { + "epoch": 2.06, + "learning_rate": 2.38443756825628e-06, + "loss": 0.3193, + "step": 11323 + }, + { + "epoch": 2.06, + "learning_rate": 2.3843829632326174e-06, + "loss": 1.577, + "step": 11324 + }, + { + "epoch": 2.06, + "learning_rate": 2.3843283582089553e-06, + "loss": 0.22, + "step": 11325 + }, + { + "epoch": 2.06, + "learning_rate": 2.384273753185293e-06, + "loss": 0.4803, + "step": 11326 + }, + { + "epoch": 2.06, + "learning_rate": 2.384219148161631e-06, + "loss": 0.471, + "step": 11327 + }, + { + "epoch": 2.06, + "learning_rate": 2.3841645431379687e-06, + "loss": 0.5163, + "step": 11328 + }, + { + "epoch": 2.06, + "learning_rate": 2.3841099381143063e-06, + "loss": 0.1422, + "step": 11329 + }, + { + "epoch": 2.06, + "learning_rate": 2.384055333090644e-06, + "loss": 0.6159, + "step": 11330 + }, + { + "epoch": 2.06, + "learning_rate": 2.384000728066982e-06, + "loss": 0.5455, + "step": 11331 + }, + { + "epoch": 2.06, + "learning_rate": 2.38394612304332e-06, + "loss": 0.9825, + "step": 11332 + }, + { + "epoch": 2.06, + "learning_rate": 2.383891518019658e-06, + "loss": 1.4271, + "step": 11333 + }, + { + "epoch": 2.06, + "learning_rate": 2.383836912995996e-06, + "loss": 1.0691, + "step": 11334 + }, + { + "epoch": 2.06, + "learning_rate": 2.3837823079723335e-06, + "loss": 0.459, + "step": 11335 + }, + { + "epoch": 2.06, + "learning_rate": 2.3837277029486714e-06, + "loss": 0.6601, + "step": 11336 + }, + { + "epoch": 2.06, + "learning_rate": 2.3836730979250094e-06, + "loss": 0.0219, + "step": 11337 + }, + { + "epoch": 2.06, + "learning_rate": 2.383618492901347e-06, + "loss": 0.1123, + "step": 11338 + }, + { + "epoch": 2.06, + "learning_rate": 2.383563887877685e-06, + "loss": 2.0331, + "step": 11339 + }, + { + "epoch": 2.06, + "learning_rate": 2.3835092828540224e-06, + "loss": 0.3432, + "step": 11340 + }, + { + "epoch": 2.06, + "learning_rate": 2.3834546778303603e-06, + "loss": 0.5068, + "step": 11341 + }, + { + "epoch": 2.06, + "learning_rate": 2.3834000728066983e-06, + "loss": 3.5162, + "step": 11342 + }, + { + "epoch": 2.06, + "learning_rate": 2.3833454677830358e-06, + "loss": 0.7296, + "step": 11343 + }, + { + "epoch": 2.06, + "learning_rate": 2.3832908627593737e-06, + "loss": 1.1508, + "step": 11344 + }, + { + "epoch": 2.06, + "learning_rate": 2.3832362577357117e-06, + "loss": 0.0694, + "step": 11345 + }, + { + "epoch": 2.06, + "learning_rate": 2.3831816527120496e-06, + "loss": 0.4231, + "step": 11346 + }, + { + "epoch": 2.06, + "learning_rate": 2.3831270476883876e-06, + "loss": 0.4049, + "step": 11347 + }, + { + "epoch": 2.06, + "learning_rate": 2.3830724426647255e-06, + "loss": 0.4409, + "step": 11348 + }, + { + "epoch": 2.06, + "learning_rate": 2.383017837641063e-06, + "loss": 1.7619, + "step": 11349 + }, + { + "epoch": 2.06, + "learning_rate": 2.382963232617401e-06, + "loss": 0.9539, + "step": 11350 + }, + { + "epoch": 2.06, + "learning_rate": 2.382908627593739e-06, + "loss": 1.5568, + "step": 11351 + }, + { + "epoch": 2.06, + "learning_rate": 2.3828540225700764e-06, + "loss": 0.1992, + "step": 11352 + }, + { + "epoch": 2.06, + "learning_rate": 2.3827994175464144e-06, + "loss": 0.0785, + "step": 11353 + }, + { + "epoch": 2.06, + "learning_rate": 2.382744812522752e-06, + "loss": 0.018, + "step": 11354 + }, + { + "epoch": 2.06, + "learning_rate": 2.38269020749909e-06, + "loss": 1.1815, + "step": 11355 + }, + { + "epoch": 2.07, + "learning_rate": 2.3826356024754278e-06, + "loss": 1.2046, + "step": 11356 + }, + { + "epoch": 2.07, + "learning_rate": 2.3825809974517653e-06, + "loss": 0.3639, + "step": 11357 + }, + { + "epoch": 2.07, + "learning_rate": 2.3825263924281032e-06, + "loss": 1.3822, + "step": 11358 + }, + { + "epoch": 2.07, + "learning_rate": 2.382471787404441e-06, + "loss": 0.0895, + "step": 11359 + }, + { + "epoch": 2.07, + "learning_rate": 2.382417182380779e-06, + "loss": 0.5556, + "step": 11360 + }, + { + "epoch": 2.07, + "learning_rate": 2.382362577357117e-06, + "loss": 2.3881, + "step": 11361 + }, + { + "epoch": 2.07, + "learning_rate": 2.382307972333455e-06, + "loss": 1.3818, + "step": 11362 + }, + { + "epoch": 2.07, + "learning_rate": 2.3822533673097925e-06, + "loss": 0.4343, + "step": 11363 + }, + { + "epoch": 2.07, + "learning_rate": 2.3821987622861305e-06, + "loss": 0.3812, + "step": 11364 + }, + { + "epoch": 2.07, + "learning_rate": 2.3821441572624684e-06, + "loss": 0.9576, + "step": 11365 + }, + { + "epoch": 2.07, + "learning_rate": 2.382089552238806e-06, + "loss": 1.0587, + "step": 11366 + }, + { + "epoch": 2.07, + "learning_rate": 2.382034947215144e-06, + "loss": 0.4932, + "step": 11367 + }, + { + "epoch": 2.07, + "learning_rate": 2.3819803421914814e-06, + "loss": 0.9414, + "step": 11368 + }, + { + "epoch": 2.07, + "learning_rate": 2.3819257371678193e-06, + "loss": 0.3541, + "step": 11369 + }, + { + "epoch": 2.07, + "learning_rate": 2.3818711321441573e-06, + "loss": 0.0307, + "step": 11370 + }, + { + "epoch": 2.07, + "learning_rate": 2.3818165271204952e-06, + "loss": 0.0935, + "step": 11371 + }, + { + "epoch": 2.07, + "learning_rate": 2.381761922096833e-06, + "loss": 1.95, + "step": 11372 + }, + { + "epoch": 2.07, + "learning_rate": 2.381707317073171e-06, + "loss": 0.9755, + "step": 11373 + }, + { + "epoch": 2.07, + "learning_rate": 2.3816527120495086e-06, + "loss": 0.0425, + "step": 11374 + }, + { + "epoch": 2.07, + "learning_rate": 2.3815981070258466e-06, + "loss": 2.0594, + "step": 11375 + }, + { + "epoch": 2.07, + "learning_rate": 2.3815435020021845e-06, + "loss": 0.4202, + "step": 11376 + }, + { + "epoch": 2.07, + "learning_rate": 2.381488896978522e-06, + "loss": 0.1502, + "step": 11377 + }, + { + "epoch": 2.07, + "learning_rate": 2.38143429195486e-06, + "loss": 1.1413, + "step": 11378 + }, + { + "epoch": 2.07, + "learning_rate": 2.381379686931198e-06, + "loss": 0.5126, + "step": 11379 + }, + { + "epoch": 2.07, + "learning_rate": 2.3813250819075355e-06, + "loss": 3.3809, + "step": 11380 + }, + { + "epoch": 2.07, + "learning_rate": 2.3812704768838734e-06, + "loss": 0.604, + "step": 11381 + }, + { + "epoch": 2.07, + "learning_rate": 2.381215871860211e-06, + "loss": 0.7043, + "step": 11382 + }, + { + "epoch": 2.07, + "learning_rate": 2.381161266836549e-06, + "loss": 0.0905, + "step": 11383 + }, + { + "epoch": 2.07, + "learning_rate": 2.381106661812887e-06, + "loss": 0.8679, + "step": 11384 + }, + { + "epoch": 2.07, + "learning_rate": 2.3810520567892247e-06, + "loss": 0.8612, + "step": 11385 + }, + { + "epoch": 2.07, + "learning_rate": 2.3809974517655627e-06, + "loss": 1.3283, + "step": 11386 + }, + { + "epoch": 2.07, + "learning_rate": 2.3809428467419006e-06, + "loss": 0.1157, + "step": 11387 + }, + { + "epoch": 2.07, + "learning_rate": 2.380888241718238e-06, + "loss": 1.3568, + "step": 11388 + }, + { + "epoch": 2.07, + "learning_rate": 2.380833636694576e-06, + "loss": 1.2123, + "step": 11389 + }, + { + "epoch": 2.07, + "learning_rate": 2.380779031670914e-06, + "loss": 0.7888, + "step": 11390 + }, + { + "epoch": 2.07, + "learning_rate": 2.3807244266472516e-06, + "loss": 0.2931, + "step": 11391 + }, + { + "epoch": 2.07, + "learning_rate": 2.3806698216235895e-06, + "loss": 1.1024, + "step": 11392 + }, + { + "epoch": 2.07, + "learning_rate": 2.380615216599927e-06, + "loss": 0.666, + "step": 11393 + }, + { + "epoch": 2.07, + "learning_rate": 2.380560611576265e-06, + "loss": 0.0647, + "step": 11394 + }, + { + "epoch": 2.07, + "learning_rate": 2.380506006552603e-06, + "loss": 1.3342, + "step": 11395 + }, + { + "epoch": 2.07, + "learning_rate": 2.3804514015289404e-06, + "loss": 0.0223, + "step": 11396 + }, + { + "epoch": 2.07, + "learning_rate": 2.3803967965052784e-06, + "loss": 1.0841, + "step": 11397 + }, + { + "epoch": 2.07, + "learning_rate": 2.3803421914816163e-06, + "loss": 0.0439, + "step": 11398 + }, + { + "epoch": 2.07, + "learning_rate": 2.3802875864579543e-06, + "loss": 3.3799, + "step": 11399 + }, + { + "epoch": 2.07, + "learning_rate": 2.380232981434292e-06, + "loss": 0.3204, + "step": 11400 + }, + { + "epoch": 2.07, + "learning_rate": 2.38017837641063e-06, + "loss": 1.6695, + "step": 11401 + }, + { + "epoch": 2.07, + "learning_rate": 2.3801237713869677e-06, + "loss": 0.0146, + "step": 11402 + }, + { + "epoch": 2.07, + "learning_rate": 2.3800691663633056e-06, + "loss": 0.5868, + "step": 11403 + }, + { + "epoch": 2.07, + "learning_rate": 2.3800145613396436e-06, + "loss": 0.0519, + "step": 11404 + }, + { + "epoch": 2.07, + "learning_rate": 2.379959956315981e-06, + "loss": 0.4634, + "step": 11405 + }, + { + "epoch": 2.07, + "learning_rate": 2.379905351292319e-06, + "loss": 0.3559, + "step": 11406 + }, + { + "epoch": 2.07, + "learning_rate": 2.3798507462686565e-06, + "loss": 0.3117, + "step": 11407 + }, + { + "epoch": 2.07, + "learning_rate": 2.3797961412449945e-06, + "loss": 3.0574, + "step": 11408 + }, + { + "epoch": 2.07, + "learning_rate": 2.3797415362213324e-06, + "loss": 1.4705, + "step": 11409 + }, + { + "epoch": 2.07, + "learning_rate": 2.37968693119767e-06, + "loss": 0.8024, + "step": 11410 + }, + { + "epoch": 2.08, + "learning_rate": 2.379632326174008e-06, + "loss": 0.7911, + "step": 11411 + }, + { + "epoch": 2.08, + "learning_rate": 2.379577721150346e-06, + "loss": 0.4601, + "step": 11412 + }, + { + "epoch": 2.08, + "learning_rate": 2.3795231161266838e-06, + "loss": 0.3825, + "step": 11413 + }, + { + "epoch": 2.08, + "learning_rate": 2.3794685111030217e-06, + "loss": 2.53, + "step": 11414 + }, + { + "epoch": 2.08, + "learning_rate": 2.3794139060793597e-06, + "loss": 0.5868, + "step": 11415 + }, + { + "epoch": 2.08, + "learning_rate": 2.379359301055697e-06, + "loss": 1.5087, + "step": 11416 + }, + { + "epoch": 2.08, + "learning_rate": 2.379304696032035e-06, + "loss": 1.5335, + "step": 11417 + }, + { + "epoch": 2.08, + "learning_rate": 2.379250091008373e-06, + "loss": 0.0776, + "step": 11418 + }, + { + "epoch": 2.08, + "learning_rate": 2.3791954859847106e-06, + "loss": 0.5738, + "step": 11419 + }, + { + "epoch": 2.08, + "learning_rate": 2.3791408809610485e-06, + "loss": 0.2808, + "step": 11420 + }, + { + "epoch": 2.08, + "learning_rate": 2.379086275937386e-06, + "loss": 2.0539, + "step": 11421 + }, + { + "epoch": 2.08, + "learning_rate": 2.379031670913724e-06, + "loss": 0.747, + "step": 11422 + }, + { + "epoch": 2.08, + "learning_rate": 2.378977065890062e-06, + "loss": 1.5793, + "step": 11423 + }, + { + "epoch": 2.08, + "learning_rate": 2.3789224608663995e-06, + "loss": 0.8377, + "step": 11424 + }, + { + "epoch": 2.08, + "learning_rate": 2.3788678558427374e-06, + "loss": 0.6628, + "step": 11425 + }, + { + "epoch": 2.08, + "learning_rate": 2.3788132508190753e-06, + "loss": 0.4889, + "step": 11426 + }, + { + "epoch": 2.08, + "learning_rate": 2.3787586457954133e-06, + "loss": 0.8031, + "step": 11427 + }, + { + "epoch": 2.08, + "learning_rate": 2.3787040407717512e-06, + "loss": 1.0933, + "step": 11428 + }, + { + "epoch": 2.08, + "learning_rate": 2.378649435748089e-06, + "loss": 1.0567, + "step": 11429 + }, + { + "epoch": 2.08, + "learning_rate": 2.3785948307244267e-06, + "loss": 0.7076, + "step": 11430 + }, + { + "epoch": 2.08, + "learning_rate": 2.3785402257007646e-06, + "loss": 1.3659, + "step": 11431 + }, + { + "epoch": 2.08, + "learning_rate": 2.3784856206771026e-06, + "loss": 0.1377, + "step": 11432 + }, + { + "epoch": 2.08, + "learning_rate": 2.37843101565344e-06, + "loss": 0.4941, + "step": 11433 + }, + { + "epoch": 2.08, + "learning_rate": 2.378376410629778e-06, + "loss": 0.1755, + "step": 11434 + }, + { + "epoch": 2.08, + "learning_rate": 2.3783218056061156e-06, + "loss": 0.0732, + "step": 11435 + }, + { + "epoch": 2.08, + "learning_rate": 2.3782672005824535e-06, + "loss": 1.8747, + "step": 11436 + }, + { + "epoch": 2.08, + "learning_rate": 2.3782125955587914e-06, + "loss": 0.2479, + "step": 11437 + }, + { + "epoch": 2.08, + "learning_rate": 2.378157990535129e-06, + "loss": 0.8594, + "step": 11438 + }, + { + "epoch": 2.08, + "learning_rate": 2.378103385511467e-06, + "loss": 2.0906, + "step": 11439 + }, + { + "epoch": 2.08, + "learning_rate": 2.378048780487805e-06, + "loss": 0.7201, + "step": 11440 + }, + { + "epoch": 2.08, + "learning_rate": 2.377994175464143e-06, + "loss": 1.5259, + "step": 11441 + }, + { + "epoch": 2.08, + "learning_rate": 2.3779395704404807e-06, + "loss": 1.4839, + "step": 11442 + }, + { + "epoch": 2.08, + "learning_rate": 2.3778849654168187e-06, + "loss": 0.1546, + "step": 11443 + }, + { + "epoch": 2.08, + "learning_rate": 2.377830360393156e-06, + "loss": 0.4156, + "step": 11444 + }, + { + "epoch": 2.08, + "learning_rate": 2.377775755369494e-06, + "loss": 0.338, + "step": 11445 + }, + { + "epoch": 2.08, + "learning_rate": 2.377721150345832e-06, + "loss": 0.9091, + "step": 11446 + }, + { + "epoch": 2.08, + "learning_rate": 2.3776665453221696e-06, + "loss": 0.8353, + "step": 11447 + }, + { + "epoch": 2.08, + "learning_rate": 2.3776119402985076e-06, + "loss": 0.5295, + "step": 11448 + }, + { + "epoch": 2.08, + "learning_rate": 2.377557335274845e-06, + "loss": 0.6428, + "step": 11449 + }, + { + "epoch": 2.08, + "learning_rate": 2.377502730251183e-06, + "loss": 0.0676, + "step": 11450 + }, + { + "epoch": 2.08, + "learning_rate": 2.377448125227521e-06, + "loss": 1.1249, + "step": 11451 + }, + { + "epoch": 2.08, + "learning_rate": 2.377393520203859e-06, + "loss": 0.8609, + "step": 11452 + }, + { + "epoch": 2.08, + "learning_rate": 2.3773389151801964e-06, + "loss": 0.3027, + "step": 11453 + }, + { + "epoch": 2.08, + "learning_rate": 2.3772843101565344e-06, + "loss": 0.2295, + "step": 11454 + }, + { + "epoch": 2.08, + "learning_rate": 2.3772297051328723e-06, + "loss": 2.6803, + "step": 11455 + }, + { + "epoch": 2.08, + "learning_rate": 2.3771751001092103e-06, + "loss": 1.5321, + "step": 11456 + }, + { + "epoch": 2.08, + "learning_rate": 2.377120495085548e-06, + "loss": 0.5968, + "step": 11457 + }, + { + "epoch": 2.08, + "learning_rate": 2.3770658900618857e-06, + "loss": 0.0321, + "step": 11458 + }, + { + "epoch": 2.08, + "learning_rate": 2.3770112850382237e-06, + "loss": 0.2228, + "step": 11459 + }, + { + "epoch": 2.08, + "learning_rate": 2.376956680014561e-06, + "loss": 1.9244, + "step": 11460 + }, + { + "epoch": 2.08, + "learning_rate": 2.376902074990899e-06, + "loss": 0.3013, + "step": 11461 + }, + { + "epoch": 2.08, + "learning_rate": 2.376847469967237e-06, + "loss": 0.8277, + "step": 11462 + }, + { + "epoch": 2.08, + "learning_rate": 2.3767928649435746e-06, + "loss": 0.2306, + "step": 11463 + }, + { + "epoch": 2.08, + "learning_rate": 2.3767382599199125e-06, + "loss": 0.0659, + "step": 11464 + }, + { + "epoch": 2.08, + "learning_rate": 2.3766836548962505e-06, + "loss": 0.7489, + "step": 11465 + }, + { + "epoch": 2.09, + "learning_rate": 2.3766290498725884e-06, + "loss": 0.0472, + "step": 11466 + }, + { + "epoch": 2.09, + "learning_rate": 2.3765744448489264e-06, + "loss": 0.9757, + "step": 11467 + }, + { + "epoch": 2.09, + "learning_rate": 2.3765198398252643e-06, + "loss": 2.0166, + "step": 11468 + }, + { + "epoch": 2.09, + "learning_rate": 2.376465234801602e-06, + "loss": 0.8603, + "step": 11469 + }, + { + "epoch": 2.09, + "learning_rate": 2.3764106297779398e-06, + "loss": 1.167, + "step": 11470 + }, + { + "epoch": 2.09, + "learning_rate": 2.3763560247542777e-06, + "loss": 0.0266, + "step": 11471 + }, + { + "epoch": 2.09, + "learning_rate": 2.3763014197306152e-06, + "loss": 0.4258, + "step": 11472 + }, + { + "epoch": 2.09, + "learning_rate": 2.376246814706953e-06, + "loss": 0.6881, + "step": 11473 + }, + { + "epoch": 2.09, + "learning_rate": 2.3761922096832907e-06, + "loss": 0.0401, + "step": 11474 + }, + { + "epoch": 2.09, + "learning_rate": 2.3761376046596286e-06, + "loss": 0.6514, + "step": 11475 + }, + { + "epoch": 2.09, + "learning_rate": 2.3760829996359666e-06, + "loss": 1.8461, + "step": 11476 + }, + { + "epoch": 2.09, + "learning_rate": 2.376028394612304e-06, + "loss": 1.543, + "step": 11477 + }, + { + "epoch": 2.09, + "learning_rate": 2.375973789588642e-06, + "loss": 0.9043, + "step": 11478 + }, + { + "epoch": 2.09, + "learning_rate": 2.37591918456498e-06, + "loss": 1.2009, + "step": 11479 + }, + { + "epoch": 2.09, + "learning_rate": 2.375864579541318e-06, + "loss": 1.4853, + "step": 11480 + }, + { + "epoch": 2.09, + "learning_rate": 2.375809974517656e-06, + "loss": 0.7861, + "step": 11481 + }, + { + "epoch": 2.09, + "learning_rate": 2.375755369493994e-06, + "loss": 0.0198, + "step": 11482 + }, + { + "epoch": 2.09, + "learning_rate": 2.3757007644703313e-06, + "loss": 0.3643, + "step": 11483 + }, + { + "epoch": 2.09, + "learning_rate": 2.3756461594466693e-06, + "loss": 0.2396, + "step": 11484 + }, + { + "epoch": 2.09, + "learning_rate": 2.3755915544230072e-06, + "loss": 0.8096, + "step": 11485 + }, + { + "epoch": 2.09, + "learning_rate": 2.3755369493993447e-06, + "loss": 0.7962, + "step": 11486 + }, + { + "epoch": 2.09, + "learning_rate": 2.3754823443756827e-06, + "loss": 0.5513, + "step": 11487 + }, + { + "epoch": 2.09, + "learning_rate": 2.37542773935202e-06, + "loss": 0.0245, + "step": 11488 + }, + { + "epoch": 2.09, + "learning_rate": 2.375373134328358e-06, + "loss": 0.803, + "step": 11489 + }, + { + "epoch": 2.09, + "learning_rate": 2.375318529304696e-06, + "loss": 0.4564, + "step": 11490 + }, + { + "epoch": 2.09, + "learning_rate": 2.3752639242810336e-06, + "loss": 0.6644, + "step": 11491 + }, + { + "epoch": 2.09, + "learning_rate": 2.3752093192573716e-06, + "loss": 1.4747, + "step": 11492 + }, + { + "epoch": 2.09, + "learning_rate": 2.3751547142337095e-06, + "loss": 0.3284, + "step": 11493 + }, + { + "epoch": 2.09, + "learning_rate": 2.3751001092100474e-06, + "loss": 1.0807, + "step": 11494 + }, + { + "epoch": 2.09, + "learning_rate": 2.3750455041863854e-06, + "loss": 1.9049, + "step": 11495 + }, + { + "epoch": 2.09, + "learning_rate": 2.3749908991627233e-06, + "loss": 0.0889, + "step": 11496 + }, + { + "epoch": 2.09, + "learning_rate": 2.374936294139061e-06, + "loss": 0.4156, + "step": 11497 + }, + { + "epoch": 2.09, + "learning_rate": 2.374881689115399e-06, + "loss": 1.5152, + "step": 11498 + }, + { + "epoch": 2.09, + "learning_rate": 2.3748270840917367e-06, + "loss": 0.5438, + "step": 11499 + }, + { + "epoch": 2.09, + "learning_rate": 2.3747724790680743e-06, + "loss": 1.7257, + "step": 11500 + }, + { + "epoch": 2.09, + "learning_rate": 2.374717874044412e-06, + "loss": 1.6608, + "step": 11501 + }, + { + "epoch": 2.09, + "learning_rate": 2.3746632690207497e-06, + "loss": 0.4645, + "step": 11502 + }, + { + "epoch": 2.09, + "learning_rate": 2.3746086639970877e-06, + "loss": 1.084, + "step": 11503 + }, + { + "epoch": 2.09, + "learning_rate": 2.3745540589734256e-06, + "loss": 1.2225, + "step": 11504 + }, + { + "epoch": 2.09, + "learning_rate": 2.374499453949763e-06, + "loss": 0.2952, + "step": 11505 + }, + { + "epoch": 2.09, + "learning_rate": 2.374444848926101e-06, + "loss": 0.3669, + "step": 11506 + }, + { + "epoch": 2.09, + "learning_rate": 2.374390243902439e-06, + "loss": 0.841, + "step": 11507 + }, + { + "epoch": 2.09, + "learning_rate": 2.374335638878777e-06, + "loss": 1.1786, + "step": 11508 + }, + { + "epoch": 2.09, + "learning_rate": 2.374281033855115e-06, + "loss": 0.4502, + "step": 11509 + }, + { + "epoch": 2.09, + "learning_rate": 2.374226428831453e-06, + "loss": 0.5939, + "step": 11510 + }, + { + "epoch": 2.09, + "learning_rate": 2.3741718238077904e-06, + "loss": 0.3943, + "step": 11511 + }, + { + "epoch": 2.09, + "learning_rate": 2.3741172187841283e-06, + "loss": 0.0404, + "step": 11512 + }, + { + "epoch": 2.09, + "learning_rate": 2.3740626137604663e-06, + "loss": 2.5188, + "step": 11513 + }, + { + "epoch": 2.09, + "learning_rate": 2.3740080087368038e-06, + "loss": 1.2543, + "step": 11514 + }, + { + "epoch": 2.09, + "learning_rate": 2.3739534037131417e-06, + "loss": 1.814, + "step": 11515 + }, + { + "epoch": 2.09, + "learning_rate": 2.3738987986894792e-06, + "loss": 0.523, + "step": 11516 + }, + { + "epoch": 2.09, + "learning_rate": 2.373844193665817e-06, + "loss": 0.6508, + "step": 11517 + }, + { + "epoch": 2.09, + "learning_rate": 2.373789588642155e-06, + "loss": 0.0799, + "step": 11518 + }, + { + "epoch": 2.09, + "learning_rate": 2.3737349836184926e-06, + "loss": 0.2854, + "step": 11519 + }, + { + "epoch": 2.09, + "learning_rate": 2.3736803785948306e-06, + "loss": 0.7075, + "step": 11520 + }, + { + "epoch": 2.1, + "learning_rate": 2.3736257735711685e-06, + "loss": 0.0617, + "step": 11521 + }, + { + "epoch": 2.1, + "learning_rate": 2.3735711685475065e-06, + "loss": 1.7494, + "step": 11522 + }, + { + "epoch": 2.1, + "learning_rate": 2.3735165635238444e-06, + "loss": 0.6829, + "step": 11523 + }, + { + "epoch": 2.1, + "learning_rate": 2.3734619585001824e-06, + "loss": 0.772, + "step": 11524 + }, + { + "epoch": 2.1, + "learning_rate": 2.37340735347652e-06, + "loss": 0.7831, + "step": 11525 + }, + { + "epoch": 2.1, + "learning_rate": 2.373352748452858e-06, + "loss": 0.4379, + "step": 11526 + }, + { + "epoch": 2.1, + "learning_rate": 2.3732981434291958e-06, + "loss": 0.6017, + "step": 11527 + }, + { + "epoch": 2.1, + "learning_rate": 2.3732435384055333e-06, + "loss": 0.458, + "step": 11528 + }, + { + "epoch": 2.1, + "learning_rate": 2.3731889333818712e-06, + "loss": 0.0301, + "step": 11529 + }, + { + "epoch": 2.1, + "learning_rate": 2.3731343283582087e-06, + "loss": 1.2422, + "step": 11530 + }, + { + "epoch": 2.1, + "learning_rate": 2.3730797233345467e-06, + "loss": 0.4648, + "step": 11531 + }, + { + "epoch": 2.1, + "learning_rate": 2.3730251183108846e-06, + "loss": 0.7142, + "step": 11532 + }, + { + "epoch": 2.1, + "learning_rate": 2.372970513287222e-06, + "loss": 1.1707, + "step": 11533 + }, + { + "epoch": 2.1, + "learning_rate": 2.37291590826356e-06, + "loss": 1.3156, + "step": 11534 + }, + { + "epoch": 2.1, + "learning_rate": 2.372861303239898e-06, + "loss": 0.6597, + "step": 11535 + }, + { + "epoch": 2.1, + "learning_rate": 2.372806698216236e-06, + "loss": 0.0486, + "step": 11536 + }, + { + "epoch": 2.1, + "learning_rate": 2.372752093192574e-06, + "loss": 2.8033, + "step": 11537 + }, + { + "epoch": 2.1, + "learning_rate": 2.372697488168912e-06, + "loss": 0.8929, + "step": 11538 + }, + { + "epoch": 2.1, + "learning_rate": 2.3726428831452494e-06, + "loss": 0.0263, + "step": 11539 + }, + { + "epoch": 2.1, + "learning_rate": 2.3725882781215873e-06, + "loss": 2.2937, + "step": 11540 + }, + { + "epoch": 2.1, + "learning_rate": 2.372533673097925e-06, + "loss": 1.5018, + "step": 11541 + }, + { + "epoch": 2.1, + "learning_rate": 2.372479068074263e-06, + "loss": 1.0586, + "step": 11542 + }, + { + "epoch": 2.1, + "learning_rate": 2.3724244630506007e-06, + "loss": 0.0294, + "step": 11543 + }, + { + "epoch": 2.1, + "learning_rate": 2.3723698580269383e-06, + "loss": 0.4038, + "step": 11544 + }, + { + "epoch": 2.1, + "learning_rate": 2.372315253003276e-06, + "loss": 1.0013, + "step": 11545 + }, + { + "epoch": 2.1, + "learning_rate": 2.372260647979614e-06, + "loss": 2.2607, + "step": 11546 + }, + { + "epoch": 2.1, + "learning_rate": 2.372206042955952e-06, + "loss": 2.0973, + "step": 11547 + }, + { + "epoch": 2.1, + "learning_rate": 2.37215143793229e-06, + "loss": 2.2188, + "step": 11548 + }, + { + "epoch": 2.1, + "learning_rate": 2.372096832908628e-06, + "loss": 0.561, + "step": 11549 + }, + { + "epoch": 2.1, + "learning_rate": 2.3720422278849655e-06, + "loss": 3.0249, + "step": 11550 + }, + { + "epoch": 2.1, + "learning_rate": 2.3719876228613034e-06, + "loss": 0.0444, + "step": 11551 + }, + { + "epoch": 2.1, + "learning_rate": 2.3719330178376414e-06, + "loss": 0.3935, + "step": 11552 + }, + { + "epoch": 2.1, + "learning_rate": 2.371878412813979e-06, + "loss": 2.8738, + "step": 11553 + }, + { + "epoch": 2.1, + "learning_rate": 2.371823807790317e-06, + "loss": 0.6377, + "step": 11554 + }, + { + "epoch": 2.1, + "learning_rate": 2.3717692027666544e-06, + "loss": 0.1206, + "step": 11555 + }, + { + "epoch": 2.1, + "learning_rate": 2.3717145977429923e-06, + "loss": 0.6767, + "step": 11556 + }, + { + "epoch": 2.1, + "learning_rate": 2.3716599927193303e-06, + "loss": 0.6725, + "step": 11557 + }, + { + "epoch": 2.1, + "learning_rate": 2.3716053876956678e-06, + "loss": 0.5725, + "step": 11558 + }, + { + "epoch": 2.1, + "learning_rate": 2.3715507826720057e-06, + "loss": 0.9207, + "step": 11559 + }, + { + "epoch": 2.1, + "learning_rate": 2.3714961776483437e-06, + "loss": 0.9342, + "step": 11560 + }, + { + "epoch": 2.1, + "learning_rate": 2.3714415726246816e-06, + "loss": 0.8865, + "step": 11561 + }, + { + "epoch": 2.1, + "learning_rate": 2.3713869676010196e-06, + "loss": 0.5354, + "step": 11562 + }, + { + "epoch": 2.1, + "learning_rate": 2.3713323625773575e-06, + "loss": 0.7828, + "step": 11563 + }, + { + "epoch": 2.1, + "learning_rate": 2.371277757553695e-06, + "loss": 0.3726, + "step": 11564 + }, + { + "epoch": 2.1, + "learning_rate": 2.371223152530033e-06, + "loss": 1.3365, + "step": 11565 + }, + { + "epoch": 2.1, + "learning_rate": 2.371168547506371e-06, + "loss": 0.6159, + "step": 11566 + }, + { + "epoch": 2.1, + "learning_rate": 2.3711139424827084e-06, + "loss": 0.6948, + "step": 11567 + }, + { + "epoch": 2.1, + "learning_rate": 2.3710593374590464e-06, + "loss": 2.1799, + "step": 11568 + }, + { + "epoch": 2.1, + "learning_rate": 2.371004732435384e-06, + "loss": 0.7427, + "step": 11569 + }, + { + "epoch": 2.1, + "learning_rate": 2.370950127411722e-06, + "loss": 0.3374, + "step": 11570 + }, + { + "epoch": 2.1, + "learning_rate": 2.3708955223880598e-06, + "loss": 0.5871, + "step": 11571 + }, + { + "epoch": 2.1, + "learning_rate": 2.3708409173643973e-06, + "loss": 1.3981, + "step": 11572 + }, + { + "epoch": 2.1, + "learning_rate": 2.3707863123407352e-06, + "loss": 0.5462, + "step": 11573 + }, + { + "epoch": 2.1, + "learning_rate": 2.370731707317073e-06, + "loss": 0.4244, + "step": 11574 + }, + { + "epoch": 2.1, + "learning_rate": 2.370677102293411e-06, + "loss": 0.6103, + "step": 11575 + }, + { + "epoch": 2.11, + "learning_rate": 2.370622497269749e-06, + "loss": 0.174, + "step": 11576 + }, + { + "epoch": 2.11, + "learning_rate": 2.370567892246087e-06, + "loss": 0.7923, + "step": 11577 + }, + { + "epoch": 2.11, + "learning_rate": 2.3705132872224245e-06, + "loss": 0.4086, + "step": 11578 + }, + { + "epoch": 2.11, + "learning_rate": 2.3704586821987625e-06, + "loss": 0.361, + "step": 11579 + }, + { + "epoch": 2.11, + "learning_rate": 2.3704040771751004e-06, + "loss": 0.276, + "step": 11580 + }, + { + "epoch": 2.11, + "learning_rate": 2.370349472151438e-06, + "loss": 3.1497, + "step": 11581 + }, + { + "epoch": 2.11, + "learning_rate": 2.370294867127776e-06, + "loss": 0.8533, + "step": 11582 + }, + { + "epoch": 2.11, + "learning_rate": 2.3702402621041134e-06, + "loss": 1.0011, + "step": 11583 + }, + { + "epoch": 2.11, + "learning_rate": 2.3701856570804513e-06, + "loss": 0.6853, + "step": 11584 + }, + { + "epoch": 2.11, + "learning_rate": 2.3701310520567893e-06, + "loss": 0.4849, + "step": 11585 + }, + { + "epoch": 2.11, + "learning_rate": 2.370076447033127e-06, + "loss": 0.512, + "step": 11586 + }, + { + "epoch": 2.11, + "learning_rate": 2.3700218420094647e-06, + "loss": 0.0493, + "step": 11587 + }, + { + "epoch": 2.11, + "learning_rate": 2.3699672369858027e-06, + "loss": 1.4568, + "step": 11588 + }, + { + "epoch": 2.11, + "learning_rate": 2.3699126319621406e-06, + "loss": 1.1284, + "step": 11589 + }, + { + "epoch": 2.11, + "learning_rate": 2.3698580269384786e-06, + "loss": 2.6965, + "step": 11590 + }, + { + "epoch": 2.11, + "learning_rate": 2.3698034219148165e-06, + "loss": 0.3275, + "step": 11591 + }, + { + "epoch": 2.11, + "learning_rate": 2.369748816891154e-06, + "loss": 0.7403, + "step": 11592 + }, + { + "epoch": 2.11, + "learning_rate": 2.369694211867492e-06, + "loss": 0.0771, + "step": 11593 + }, + { + "epoch": 2.11, + "learning_rate": 2.36963960684383e-06, + "loss": 0.5328, + "step": 11594 + }, + { + "epoch": 2.11, + "learning_rate": 2.3695850018201674e-06, + "loss": 3.1731, + "step": 11595 + }, + { + "epoch": 2.11, + "learning_rate": 2.3695303967965054e-06, + "loss": 0.5132, + "step": 11596 + }, + { + "epoch": 2.11, + "learning_rate": 2.369475791772843e-06, + "loss": 0.6251, + "step": 11597 + }, + { + "epoch": 2.11, + "learning_rate": 2.369421186749181e-06, + "loss": 0.0191, + "step": 11598 + }, + { + "epoch": 2.11, + "learning_rate": 2.369366581725519e-06, + "loss": 3.5133, + "step": 11599 + }, + { + "epoch": 2.11, + "learning_rate": 2.3693119767018563e-06, + "loss": 1.1933, + "step": 11600 + }, + { + "epoch": 2.11, + "learning_rate": 2.3692573716781943e-06, + "loss": 0.4222, + "step": 11601 + }, + { + "epoch": 2.11, + "learning_rate": 2.369202766654532e-06, + "loss": 0.6928, + "step": 11602 + }, + { + "epoch": 2.11, + "learning_rate": 2.36914816163087e-06, + "loss": 1.3488, + "step": 11603 + }, + { + "epoch": 2.11, + "learning_rate": 2.369093556607208e-06, + "loss": 0.0468, + "step": 11604 + }, + { + "epoch": 2.11, + "learning_rate": 2.369038951583546e-06, + "loss": 0.1045, + "step": 11605 + }, + { + "epoch": 2.11, + "learning_rate": 2.3689843465598836e-06, + "loss": 1.0797, + "step": 11606 + }, + { + "epoch": 2.11, + "learning_rate": 2.3689297415362215e-06, + "loss": 0.7436, + "step": 11607 + }, + { + "epoch": 2.11, + "learning_rate": 2.3688751365125594e-06, + "loss": 0.8521, + "step": 11608 + }, + { + "epoch": 2.11, + "learning_rate": 2.368820531488897e-06, + "loss": 0.3937, + "step": 11609 + }, + { + "epoch": 2.11, + "learning_rate": 2.368765926465235e-06, + "loss": 1.2958, + "step": 11610 + }, + { + "epoch": 2.11, + "learning_rate": 2.3687113214415724e-06, + "loss": 1.3545, + "step": 11611 + }, + { + "epoch": 2.11, + "learning_rate": 2.3686567164179104e-06, + "loss": 0.0398, + "step": 11612 + }, + { + "epoch": 2.11, + "learning_rate": 2.3686021113942483e-06, + "loss": 1.5781, + "step": 11613 + }, + { + "epoch": 2.11, + "learning_rate": 2.368547506370586e-06, + "loss": 0.1324, + "step": 11614 + }, + { + "epoch": 2.11, + "learning_rate": 2.3684929013469238e-06, + "loss": 0.0364, + "step": 11615 + }, + { + "epoch": 2.11, + "learning_rate": 2.3684382963232617e-06, + "loss": 0.6261, + "step": 11616 + }, + { + "epoch": 2.11, + "learning_rate": 2.3683836912995997e-06, + "loss": 0.3395, + "step": 11617 + }, + { + "epoch": 2.11, + "learning_rate": 2.3683290862759376e-06, + "loss": 1.7736, + "step": 11618 + }, + { + "epoch": 2.11, + "learning_rate": 2.3682744812522755e-06, + "loss": 1.9189, + "step": 11619 + }, + { + "epoch": 2.11, + "learning_rate": 2.368219876228613e-06, + "loss": 0.2524, + "step": 11620 + }, + { + "epoch": 2.11, + "learning_rate": 2.368165271204951e-06, + "loss": 1.9317, + "step": 11621 + }, + { + "epoch": 2.11, + "learning_rate": 2.3681106661812885e-06, + "loss": 0.0387, + "step": 11622 + }, + { + "epoch": 2.11, + "learning_rate": 2.3680560611576265e-06, + "loss": 1.1839, + "step": 11623 + }, + { + "epoch": 2.11, + "learning_rate": 2.3680014561339644e-06, + "loss": 1.5238, + "step": 11624 + }, + { + "epoch": 2.11, + "learning_rate": 2.367946851110302e-06, + "loss": 0.8707, + "step": 11625 + }, + { + "epoch": 2.11, + "learning_rate": 2.36789224608664e-06, + "loss": 1.4577, + "step": 11626 + }, + { + "epoch": 2.11, + "learning_rate": 2.367837641062978e-06, + "loss": 2.4918, + "step": 11627 + }, + { + "epoch": 2.11, + "learning_rate": 2.3677830360393158e-06, + "loss": 1.8196, + "step": 11628 + }, + { + "epoch": 2.11, + "learning_rate": 2.3677284310156537e-06, + "loss": 0.2825, + "step": 11629 + }, + { + "epoch": 2.11, + "learning_rate": 2.3676738259919917e-06, + "loss": 0.5017, + "step": 11630 + }, + { + "epoch": 2.12, + "learning_rate": 2.367619220968329e-06, + "loss": 0.4372, + "step": 11631 + }, + { + "epoch": 2.12, + "learning_rate": 2.367564615944667e-06, + "loss": 2.1572, + "step": 11632 + }, + { + "epoch": 2.12, + "learning_rate": 2.367510010921005e-06, + "loss": 2.4076, + "step": 11633 + }, + { + "epoch": 2.12, + "learning_rate": 2.3674554058973426e-06, + "loss": 0.078, + "step": 11634 + }, + { + "epoch": 2.12, + "learning_rate": 2.3674008008736805e-06, + "loss": 1.0066, + "step": 11635 + }, + { + "epoch": 2.12, + "learning_rate": 2.367346195850018e-06, + "loss": 0.8576, + "step": 11636 + }, + { + "epoch": 2.12, + "learning_rate": 2.367291590826356e-06, + "loss": 1.0653, + "step": 11637 + }, + { + "epoch": 2.12, + "learning_rate": 2.367236985802694e-06, + "loss": 1.0868, + "step": 11638 + }, + { + "epoch": 2.12, + "learning_rate": 2.3671823807790315e-06, + "loss": 0.513, + "step": 11639 + }, + { + "epoch": 2.12, + "learning_rate": 2.3671277757553694e-06, + "loss": 0.4044, + "step": 11640 + }, + { + "epoch": 2.12, + "learning_rate": 2.3670731707317073e-06, + "loss": 0.027, + "step": 11641 + }, + { + "epoch": 2.12, + "learning_rate": 2.3670185657080453e-06, + "loss": 0.0265, + "step": 11642 + }, + { + "epoch": 2.12, + "learning_rate": 2.3669639606843832e-06, + "loss": 1.6057, + "step": 11643 + }, + { + "epoch": 2.12, + "learning_rate": 2.366909355660721e-06, + "loss": 0.0513, + "step": 11644 + }, + { + "epoch": 2.12, + "learning_rate": 2.3668547506370587e-06, + "loss": 0.6378, + "step": 11645 + }, + { + "epoch": 2.12, + "learning_rate": 2.3668001456133966e-06, + "loss": 0.8212, + "step": 11646 + }, + { + "epoch": 2.12, + "learning_rate": 2.3667455405897346e-06, + "loss": 0.2103, + "step": 11647 + }, + { + "epoch": 2.12, + "learning_rate": 2.366690935566072e-06, + "loss": 1.3311, + "step": 11648 + }, + { + "epoch": 2.12, + "learning_rate": 2.36663633054241e-06, + "loss": 3.0167, + "step": 11649 + }, + { + "epoch": 2.12, + "learning_rate": 2.3665817255187476e-06, + "loss": 0.5703, + "step": 11650 + }, + { + "epoch": 2.12, + "learning_rate": 2.3665271204950855e-06, + "loss": 0.753, + "step": 11651 + }, + { + "epoch": 2.12, + "learning_rate": 2.3664725154714234e-06, + "loss": 2.683, + "step": 11652 + }, + { + "epoch": 2.12, + "learning_rate": 2.366417910447761e-06, + "loss": 0.2783, + "step": 11653 + }, + { + "epoch": 2.12, + "learning_rate": 2.366363305424099e-06, + "loss": 1.084, + "step": 11654 + }, + { + "epoch": 2.12, + "learning_rate": 2.366308700400437e-06, + "loss": 0.4212, + "step": 11655 + }, + { + "epoch": 2.12, + "learning_rate": 2.366254095376775e-06, + "loss": 0.7571, + "step": 11656 + }, + { + "epoch": 2.12, + "learning_rate": 2.3661994903531127e-06, + "loss": 0.0413, + "step": 11657 + }, + { + "epoch": 2.12, + "learning_rate": 2.3661448853294507e-06, + "loss": 1.3525, + "step": 11658 + }, + { + "epoch": 2.12, + "learning_rate": 2.366090280305788e-06, + "loss": 0.2117, + "step": 11659 + }, + { + "epoch": 2.12, + "learning_rate": 2.366035675282126e-06, + "loss": 0.1059, + "step": 11660 + }, + { + "epoch": 2.12, + "learning_rate": 2.365981070258464e-06, + "loss": 1.1678, + "step": 11661 + }, + { + "epoch": 2.12, + "learning_rate": 2.3659264652348016e-06, + "loss": 1.8397, + "step": 11662 + }, + { + "epoch": 2.12, + "learning_rate": 2.3658718602111396e-06, + "loss": 0.0937, + "step": 11663 + }, + { + "epoch": 2.12, + "learning_rate": 2.365817255187477e-06, + "loss": 0.4242, + "step": 11664 + }, + { + "epoch": 2.12, + "learning_rate": 2.365762650163815e-06, + "loss": 0.3584, + "step": 11665 + }, + { + "epoch": 2.12, + "learning_rate": 2.365708045140153e-06, + "loss": 1.1521, + "step": 11666 + }, + { + "epoch": 2.12, + "learning_rate": 2.3656534401164905e-06, + "loss": 1.3247, + "step": 11667 + }, + { + "epoch": 2.12, + "learning_rate": 2.3655988350928284e-06, + "loss": 0.3742, + "step": 11668 + }, + { + "epoch": 2.12, + "learning_rate": 2.3655442300691664e-06, + "loss": 0.0337, + "step": 11669 + }, + { + "epoch": 2.12, + "learning_rate": 2.3654896250455043e-06, + "loss": 2.224, + "step": 11670 + }, + { + "epoch": 2.12, + "learning_rate": 2.3654350200218423e-06, + "loss": 0.4495, + "step": 11671 + }, + { + "epoch": 2.12, + "learning_rate": 2.36538041499818e-06, + "loss": 1.2172, + "step": 11672 + }, + { + "epoch": 2.12, + "learning_rate": 2.3653258099745177e-06, + "loss": 0.3581, + "step": 11673 + }, + { + "epoch": 2.12, + "learning_rate": 2.3652712049508557e-06, + "loss": 1.8756, + "step": 11674 + }, + { + "epoch": 2.12, + "learning_rate": 2.3652165999271936e-06, + "loss": 0.1108, + "step": 11675 + }, + { + "epoch": 2.12, + "learning_rate": 2.365161994903531e-06, + "loss": 0.0687, + "step": 11676 + }, + { + "epoch": 2.12, + "learning_rate": 2.365107389879869e-06, + "loss": 0.2659, + "step": 11677 + }, + { + "epoch": 2.12, + "learning_rate": 2.3650527848562066e-06, + "loss": 0.3446, + "step": 11678 + }, + { + "epoch": 2.12, + "learning_rate": 2.3649981798325445e-06, + "loss": 0.7206, + "step": 11679 + }, + { + "epoch": 2.12, + "learning_rate": 2.3649435748088825e-06, + "loss": 0.0165, + "step": 11680 + }, + { + "epoch": 2.12, + "learning_rate": 2.36488896978522e-06, + "loss": 0.764, + "step": 11681 + }, + { + "epoch": 2.12, + "learning_rate": 2.364834364761558e-06, + "loss": 1.2232, + "step": 11682 + }, + { + "epoch": 2.12, + "learning_rate": 2.364779759737896e-06, + "loss": 0.5895, + "step": 11683 + }, + { + "epoch": 2.12, + "learning_rate": 2.364725154714234e-06, + "loss": 0.5206, + "step": 11684 + }, + { + "epoch": 2.12, + "learning_rate": 2.3646705496905718e-06, + "loss": 0.0199, + "step": 11685 + }, + { + "epoch": 2.13, + "learning_rate": 2.3646159446669097e-06, + "loss": 0.2196, + "step": 11686 + }, + { + "epoch": 2.13, + "learning_rate": 2.3645613396432472e-06, + "loss": 1.2382, + "step": 11687 + }, + { + "epoch": 2.13, + "learning_rate": 2.364506734619585e-06, + "loss": 1.1302, + "step": 11688 + }, + { + "epoch": 2.13, + "learning_rate": 2.364452129595923e-06, + "loss": 1.3972, + "step": 11689 + }, + { + "epoch": 2.13, + "learning_rate": 2.3643975245722606e-06, + "loss": 0.3101, + "step": 11690 + }, + { + "epoch": 2.13, + "learning_rate": 2.3643429195485986e-06, + "loss": 0.6634, + "step": 11691 + }, + { + "epoch": 2.13, + "learning_rate": 2.364288314524936e-06, + "loss": 1.4682, + "step": 11692 + }, + { + "epoch": 2.13, + "learning_rate": 2.364233709501274e-06, + "loss": 1.5392, + "step": 11693 + }, + { + "epoch": 2.13, + "learning_rate": 2.364179104477612e-06, + "loss": 0.5079, + "step": 11694 + }, + { + "epoch": 2.13, + "learning_rate": 2.3641244994539495e-06, + "loss": 0.5556, + "step": 11695 + }, + { + "epoch": 2.13, + "learning_rate": 2.3640698944302874e-06, + "loss": 1.1155, + "step": 11696 + }, + { + "epoch": 2.13, + "learning_rate": 2.3640152894066254e-06, + "loss": 0.926, + "step": 11697 + }, + { + "epoch": 2.13, + "learning_rate": 2.3639606843829633e-06, + "loss": 0.5505, + "step": 11698 + }, + { + "epoch": 2.13, + "learning_rate": 2.3639060793593013e-06, + "loss": 0.0428, + "step": 11699 + }, + { + "epoch": 2.13, + "learning_rate": 2.3638514743356392e-06, + "loss": 1.843, + "step": 11700 + }, + { + "epoch": 2.13, + "learning_rate": 2.3637968693119767e-06, + "loss": 1.2188, + "step": 11701 + }, + { + "epoch": 2.13, + "learning_rate": 2.3637422642883147e-06, + "loss": 0.2652, + "step": 11702 + }, + { + "epoch": 2.13, + "learning_rate": 2.363687659264652e-06, + "loss": 0.4351, + "step": 11703 + }, + { + "epoch": 2.13, + "learning_rate": 2.36363305424099e-06, + "loss": 0.1735, + "step": 11704 + }, + { + "epoch": 2.13, + "learning_rate": 2.363578449217328e-06, + "loss": 0.5288, + "step": 11705 + }, + { + "epoch": 2.13, + "learning_rate": 2.3635238441936656e-06, + "loss": 1.1787, + "step": 11706 + }, + { + "epoch": 2.13, + "learning_rate": 2.3634692391700036e-06, + "loss": 0.4919, + "step": 11707 + }, + { + "epoch": 2.13, + "learning_rate": 2.3634146341463415e-06, + "loss": 0.019, + "step": 11708 + }, + { + "epoch": 2.13, + "learning_rate": 2.3633600291226794e-06, + "loss": 0.1629, + "step": 11709 + }, + { + "epoch": 2.13, + "learning_rate": 2.3633054240990174e-06, + "loss": 2.2166, + "step": 11710 + }, + { + "epoch": 2.13, + "learning_rate": 2.3632508190753553e-06, + "loss": 1.9066, + "step": 11711 + }, + { + "epoch": 2.13, + "learning_rate": 2.363196214051693e-06, + "loss": 0.1803, + "step": 11712 + }, + { + "epoch": 2.13, + "learning_rate": 2.363141609028031e-06, + "loss": 1.8987, + "step": 11713 + }, + { + "epoch": 2.13, + "learning_rate": 2.3630870040043687e-06, + "loss": 1.5383, + "step": 11714 + }, + { + "epoch": 2.13, + "learning_rate": 2.3630323989807063e-06, + "loss": 1.2508, + "step": 11715 + }, + { + "epoch": 2.13, + "learning_rate": 2.362977793957044e-06, + "loss": 0.3655, + "step": 11716 + }, + { + "epoch": 2.13, + "learning_rate": 2.3629231889333817e-06, + "loss": 0.4551, + "step": 11717 + }, + { + "epoch": 2.13, + "learning_rate": 2.3628685839097197e-06, + "loss": 2.0991, + "step": 11718 + }, + { + "epoch": 2.13, + "learning_rate": 2.3628139788860576e-06, + "loss": 0.8614, + "step": 11719 + }, + { + "epoch": 2.13, + "learning_rate": 2.362759373862395e-06, + "loss": 0.4853, + "step": 11720 + }, + { + "epoch": 2.13, + "learning_rate": 2.362704768838733e-06, + "loss": 0.8716, + "step": 11721 + }, + { + "epoch": 2.13, + "learning_rate": 2.362650163815071e-06, + "loss": 0.3206, + "step": 11722 + }, + { + "epoch": 2.13, + "learning_rate": 2.362595558791409e-06, + "loss": 0.5742, + "step": 11723 + }, + { + "epoch": 2.13, + "learning_rate": 2.362540953767747e-06, + "loss": 1.1682, + "step": 11724 + }, + { + "epoch": 2.13, + "learning_rate": 2.362486348744085e-06, + "loss": 0.2156, + "step": 11725 + }, + { + "epoch": 2.13, + "learning_rate": 2.3624317437204224e-06, + "loss": 2.2387, + "step": 11726 + }, + { + "epoch": 2.13, + "learning_rate": 2.3623771386967603e-06, + "loss": 0.4906, + "step": 11727 + }, + { + "epoch": 2.13, + "learning_rate": 2.3623225336730982e-06, + "loss": 1.5072, + "step": 11728 + }, + { + "epoch": 2.13, + "learning_rate": 2.3622679286494358e-06, + "loss": 0.3345, + "step": 11729 + }, + { + "epoch": 2.13, + "learning_rate": 2.3622133236257737e-06, + "loss": 2.3049, + "step": 11730 + }, + { + "epoch": 2.13, + "learning_rate": 2.3621587186021112e-06, + "loss": 1.2793, + "step": 11731 + }, + { + "epoch": 2.13, + "learning_rate": 2.362104113578449e-06, + "loss": 0.1533, + "step": 11732 + }, + { + "epoch": 2.13, + "learning_rate": 2.362049508554787e-06, + "loss": 0.3174, + "step": 11733 + }, + { + "epoch": 2.13, + "learning_rate": 2.3619949035311246e-06, + "loss": 1.0806, + "step": 11734 + }, + { + "epoch": 2.13, + "learning_rate": 2.3619402985074626e-06, + "loss": 0.3796, + "step": 11735 + }, + { + "epoch": 2.13, + "learning_rate": 2.3618856934838005e-06, + "loss": 1.0681, + "step": 11736 + }, + { + "epoch": 2.13, + "learning_rate": 2.3618310884601385e-06, + "loss": 0.3394, + "step": 11737 + }, + { + "epoch": 2.13, + "learning_rate": 2.3617764834364764e-06, + "loss": 0.9814, + "step": 11738 + }, + { + "epoch": 2.13, + "learning_rate": 2.3617218784128144e-06, + "loss": 0.1619, + "step": 11739 + }, + { + "epoch": 2.13, + "learning_rate": 2.361667273389152e-06, + "loss": 0.2484, + "step": 11740 + }, + { + "epoch": 2.14, + "learning_rate": 2.36161266836549e-06, + "loss": 0.5479, + "step": 11741 + }, + { + "epoch": 2.14, + "learning_rate": 2.3615580633418278e-06, + "loss": 0.031, + "step": 11742 + }, + { + "epoch": 2.14, + "learning_rate": 2.3615034583181653e-06, + "loss": 1.6877, + "step": 11743 + }, + { + "epoch": 2.14, + "learning_rate": 2.3614488532945032e-06, + "loss": 0.5951, + "step": 11744 + }, + { + "epoch": 2.14, + "learning_rate": 2.3613942482708407e-06, + "loss": 3.2129, + "step": 11745 + }, + { + "epoch": 2.14, + "learning_rate": 2.3613396432471787e-06, + "loss": 0.2751, + "step": 11746 + }, + { + "epoch": 2.14, + "learning_rate": 2.3612850382235166e-06, + "loss": 0.1001, + "step": 11747 + }, + { + "epoch": 2.14, + "learning_rate": 2.361230433199854e-06, + "loss": 1.9881, + "step": 11748 + }, + { + "epoch": 2.14, + "learning_rate": 2.361175828176192e-06, + "loss": 2.1902, + "step": 11749 + }, + { + "epoch": 2.14, + "learning_rate": 2.36112122315253e-06, + "loss": 0.3114, + "step": 11750 + }, + { + "epoch": 2.14, + "learning_rate": 2.361066618128868e-06, + "loss": 1.6449, + "step": 11751 + }, + { + "epoch": 2.14, + "learning_rate": 2.361012013105206e-06, + "loss": 0.3566, + "step": 11752 + }, + { + "epoch": 2.14, + "learning_rate": 2.360957408081544e-06, + "loss": 0.9705, + "step": 11753 + }, + { + "epoch": 2.14, + "learning_rate": 2.3609028030578814e-06, + "loss": 0.0602, + "step": 11754 + }, + { + "epoch": 2.14, + "learning_rate": 2.3608481980342193e-06, + "loss": 0.5269, + "step": 11755 + }, + { + "epoch": 2.14, + "learning_rate": 2.3607935930105573e-06, + "loss": 0.7026, + "step": 11756 + }, + { + "epoch": 2.14, + "learning_rate": 2.360738987986895e-06, + "loss": 0.6538, + "step": 11757 + }, + { + "epoch": 2.14, + "learning_rate": 2.3606843829632327e-06, + "loss": 0.5661, + "step": 11758 + }, + { + "epoch": 2.14, + "learning_rate": 2.3606297779395703e-06, + "loss": 4.2664, + "step": 11759 + }, + { + "epoch": 2.14, + "learning_rate": 2.360575172915908e-06, + "loss": 0.2307, + "step": 11760 + }, + { + "epoch": 2.14, + "learning_rate": 2.360520567892246e-06, + "loss": 0.266, + "step": 11761 + }, + { + "epoch": 2.14, + "learning_rate": 2.3604659628685837e-06, + "loss": 0.3343, + "step": 11762 + }, + { + "epoch": 2.14, + "learning_rate": 2.3604113578449216e-06, + "loss": 1.9815, + "step": 11763 + }, + { + "epoch": 2.14, + "learning_rate": 2.3603567528212596e-06, + "loss": 0.077, + "step": 11764 + }, + { + "epoch": 2.14, + "learning_rate": 2.3603021477975975e-06, + "loss": 0.7267, + "step": 11765 + }, + { + "epoch": 2.14, + "learning_rate": 2.3602475427739354e-06, + "loss": 1.8792, + "step": 11766 + }, + { + "epoch": 2.14, + "learning_rate": 2.3601929377502734e-06, + "loss": 2.7419, + "step": 11767 + }, + { + "epoch": 2.14, + "learning_rate": 2.360138332726611e-06, + "loss": 0.3941, + "step": 11768 + }, + { + "epoch": 2.14, + "learning_rate": 2.360083727702949e-06, + "loss": 0.3249, + "step": 11769 + }, + { + "epoch": 2.14, + "learning_rate": 2.3600291226792864e-06, + "loss": 0.8228, + "step": 11770 + }, + { + "epoch": 2.14, + "learning_rate": 2.3599745176556243e-06, + "loss": 0.6583, + "step": 11771 + }, + { + "epoch": 2.14, + "learning_rate": 2.3599199126319623e-06, + "loss": 1.0926, + "step": 11772 + }, + { + "epoch": 2.14, + "learning_rate": 2.3598653076082998e-06, + "loss": 2.4092, + "step": 11773 + }, + { + "epoch": 2.14, + "learning_rate": 2.3598107025846377e-06, + "loss": 3.0436, + "step": 11774 + }, + { + "epoch": 2.14, + "learning_rate": 2.3597560975609757e-06, + "loss": 1.074, + "step": 11775 + }, + { + "epoch": 2.14, + "learning_rate": 2.359701492537313e-06, + "loss": 0.1611, + "step": 11776 + }, + { + "epoch": 2.14, + "learning_rate": 2.359646887513651e-06, + "loss": 0.9816, + "step": 11777 + }, + { + "epoch": 2.14, + "learning_rate": 2.359592282489989e-06, + "loss": 3.0091, + "step": 11778 + }, + { + "epoch": 2.14, + "learning_rate": 2.359537677466327e-06, + "loss": 0.5234, + "step": 11779 + }, + { + "epoch": 2.14, + "learning_rate": 2.359483072442665e-06, + "loss": 1.0796, + "step": 11780 + }, + { + "epoch": 2.14, + "learning_rate": 2.359428467419003e-06, + "loss": 1.9625, + "step": 11781 + }, + { + "epoch": 2.14, + "learning_rate": 2.3593738623953404e-06, + "loss": 0.3627, + "step": 11782 + }, + { + "epoch": 2.14, + "learning_rate": 2.3593192573716784e-06, + "loss": 2.1899, + "step": 11783 + }, + { + "epoch": 2.14, + "learning_rate": 2.359264652348016e-06, + "loss": 1.4378, + "step": 11784 + }, + { + "epoch": 2.14, + "learning_rate": 2.359210047324354e-06, + "loss": 0.0539, + "step": 11785 + }, + { + "epoch": 2.14, + "learning_rate": 2.3591554423006918e-06, + "loss": 0.4807, + "step": 11786 + }, + { + "epoch": 2.14, + "learning_rate": 2.3591008372770293e-06, + "loss": 0.1663, + "step": 11787 + }, + { + "epoch": 2.14, + "learning_rate": 2.3590462322533672e-06, + "loss": 2.268, + "step": 11788 + }, + { + "epoch": 2.14, + "learning_rate": 2.358991627229705e-06, + "loss": 0.2855, + "step": 11789 + }, + { + "epoch": 2.14, + "learning_rate": 2.3589370222060427e-06, + "loss": 0.1739, + "step": 11790 + }, + { + "epoch": 2.14, + "learning_rate": 2.3588824171823806e-06, + "loss": 1.1094, + "step": 11791 + }, + { + "epoch": 2.14, + "learning_rate": 2.3588278121587186e-06, + "loss": 0.02, + "step": 11792 + }, + { + "epoch": 2.14, + "learning_rate": 2.3587732071350565e-06, + "loss": 5.1235, + "step": 11793 + }, + { + "epoch": 2.14, + "learning_rate": 2.3587186021113945e-06, + "loss": 0.4595, + "step": 11794 + }, + { + "epoch": 2.14, + "learning_rate": 2.3586639970877324e-06, + "loss": 0.1621, + "step": 11795 + }, + { + "epoch": 2.15, + "learning_rate": 2.35860939206407e-06, + "loss": 0.087, + "step": 11796 + }, + { + "epoch": 2.15, + "learning_rate": 2.358554787040408e-06, + "loss": 0.1597, + "step": 11797 + }, + { + "epoch": 2.15, + "learning_rate": 2.3585001820167454e-06, + "loss": 0.8061, + "step": 11798 + }, + { + "epoch": 2.15, + "learning_rate": 2.3584455769930833e-06, + "loss": 0.0785, + "step": 11799 + }, + { + "epoch": 2.15, + "learning_rate": 2.3583909719694213e-06, + "loss": 0.5506, + "step": 11800 + }, + { + "epoch": 2.15, + "learning_rate": 2.358336366945759e-06, + "loss": 1.232, + "step": 11801 + }, + { + "epoch": 2.15, + "learning_rate": 2.3582817619220967e-06, + "loss": 1.2978, + "step": 11802 + }, + { + "epoch": 2.15, + "learning_rate": 2.3582271568984347e-06, + "loss": 1.7779, + "step": 11803 + }, + { + "epoch": 2.15, + "learning_rate": 2.3581725518747726e-06, + "loss": 0.947, + "step": 11804 + }, + { + "epoch": 2.15, + "learning_rate": 2.3581179468511106e-06, + "loss": 0.6297, + "step": 11805 + }, + { + "epoch": 2.15, + "learning_rate": 2.3580633418274485e-06, + "loss": 0.4359, + "step": 11806 + }, + { + "epoch": 2.15, + "learning_rate": 2.358008736803786e-06, + "loss": 0.0348, + "step": 11807 + }, + { + "epoch": 2.15, + "learning_rate": 2.357954131780124e-06, + "loss": 0.3425, + "step": 11808 + }, + { + "epoch": 2.15, + "learning_rate": 2.357899526756462e-06, + "loss": 0.0699, + "step": 11809 + }, + { + "epoch": 2.15, + "learning_rate": 2.3578449217327994e-06, + "loss": 0.4131, + "step": 11810 + }, + { + "epoch": 2.15, + "learning_rate": 2.3577903167091374e-06, + "loss": 0.54, + "step": 11811 + }, + { + "epoch": 2.15, + "learning_rate": 2.357735711685475e-06, + "loss": 1.0222, + "step": 11812 + }, + { + "epoch": 2.15, + "learning_rate": 2.357681106661813e-06, + "loss": 1.7706, + "step": 11813 + }, + { + "epoch": 2.15, + "learning_rate": 2.357626501638151e-06, + "loss": 0.5025, + "step": 11814 + }, + { + "epoch": 2.15, + "learning_rate": 2.3575718966144883e-06, + "loss": 0.6098, + "step": 11815 + }, + { + "epoch": 2.15, + "learning_rate": 2.3575172915908263e-06, + "loss": 1.8718, + "step": 11816 + }, + { + "epoch": 2.15, + "learning_rate": 2.357462686567164e-06, + "loss": 0.8025, + "step": 11817 + }, + { + "epoch": 2.15, + "learning_rate": 2.357408081543502e-06, + "loss": 0.0151, + "step": 11818 + }, + { + "epoch": 2.15, + "learning_rate": 2.35735347651984e-06, + "loss": 0.9955, + "step": 11819 + }, + { + "epoch": 2.15, + "learning_rate": 2.357298871496178e-06, + "loss": 0.7477, + "step": 11820 + }, + { + "epoch": 2.15, + "learning_rate": 2.3572442664725155e-06, + "loss": 0.2841, + "step": 11821 + }, + { + "epoch": 2.15, + "learning_rate": 2.3571896614488535e-06, + "loss": 0.0221, + "step": 11822 + }, + { + "epoch": 2.15, + "learning_rate": 2.3571350564251914e-06, + "loss": 0.4446, + "step": 11823 + }, + { + "epoch": 2.15, + "learning_rate": 2.357080451401529e-06, + "loss": 1.5512, + "step": 11824 + }, + { + "epoch": 2.15, + "learning_rate": 2.357025846377867e-06, + "loss": 0.07, + "step": 11825 + }, + { + "epoch": 2.15, + "learning_rate": 2.3569712413542044e-06, + "loss": 0.1394, + "step": 11826 + }, + { + "epoch": 2.15, + "learning_rate": 2.3569166363305424e-06, + "loss": 0.8774, + "step": 11827 + }, + { + "epoch": 2.15, + "learning_rate": 2.3568620313068803e-06, + "loss": 0.0455, + "step": 11828 + }, + { + "epoch": 2.15, + "learning_rate": 2.356807426283218e-06, + "loss": 0.1314, + "step": 11829 + }, + { + "epoch": 2.15, + "learning_rate": 2.3567528212595558e-06, + "loss": 0.8459, + "step": 11830 + }, + { + "epoch": 2.15, + "learning_rate": 2.3566982162358937e-06, + "loss": 1.4322, + "step": 11831 + }, + { + "epoch": 2.15, + "learning_rate": 2.3566436112122317e-06, + "loss": 1.6524, + "step": 11832 + }, + { + "epoch": 2.15, + "learning_rate": 2.3565890061885696e-06, + "loss": 2.2686, + "step": 11833 + }, + { + "epoch": 2.15, + "learning_rate": 2.3565344011649075e-06, + "loss": 0.4708, + "step": 11834 + }, + { + "epoch": 2.15, + "learning_rate": 2.356479796141245e-06, + "loss": 0.6851, + "step": 11835 + }, + { + "epoch": 2.15, + "learning_rate": 2.356425191117583e-06, + "loss": 0.4563, + "step": 11836 + }, + { + "epoch": 2.15, + "learning_rate": 2.356370586093921e-06, + "loss": 0.2538, + "step": 11837 + }, + { + "epoch": 2.15, + "learning_rate": 2.3563159810702585e-06, + "loss": 1.53, + "step": 11838 + }, + { + "epoch": 2.15, + "learning_rate": 2.3562613760465964e-06, + "loss": 0.6996, + "step": 11839 + }, + { + "epoch": 2.15, + "learning_rate": 2.356206771022934e-06, + "loss": 2.0981, + "step": 11840 + }, + { + "epoch": 2.15, + "learning_rate": 2.356152165999272e-06, + "loss": 1.0333, + "step": 11841 + }, + { + "epoch": 2.15, + "learning_rate": 2.35609756097561e-06, + "loss": 0.1686, + "step": 11842 + }, + { + "epoch": 2.15, + "learning_rate": 2.3560429559519473e-06, + "loss": 0.8988, + "step": 11843 + }, + { + "epoch": 2.15, + "learning_rate": 2.3559883509282853e-06, + "loss": 0.7889, + "step": 11844 + }, + { + "epoch": 2.15, + "learning_rate": 2.3559337459046232e-06, + "loss": 1.45, + "step": 11845 + }, + { + "epoch": 2.15, + "learning_rate": 2.355879140880961e-06, + "loss": 0.6712, + "step": 11846 + }, + { + "epoch": 2.15, + "learning_rate": 2.355824535857299e-06, + "loss": 0.5015, + "step": 11847 + }, + { + "epoch": 2.15, + "learning_rate": 2.355769930833637e-06, + "loss": 0.5876, + "step": 11848 + }, + { + "epoch": 2.15, + "learning_rate": 2.3557153258099746e-06, + "loss": 0.3893, + "step": 11849 + }, + { + "epoch": 2.15, + "learning_rate": 2.3556607207863125e-06, + "loss": 1.2477, + "step": 11850 + }, + { + "epoch": 2.16, + "learning_rate": 2.35560611576265e-06, + "loss": 0.7578, + "step": 11851 + }, + { + "epoch": 2.16, + "learning_rate": 2.355551510738988e-06, + "loss": 0.6561, + "step": 11852 + }, + { + "epoch": 2.16, + "learning_rate": 2.355496905715326e-06, + "loss": 0.5045, + "step": 11853 + }, + { + "epoch": 2.16, + "learning_rate": 2.3554423006916634e-06, + "loss": 0.042, + "step": 11854 + }, + { + "epoch": 2.16, + "learning_rate": 2.3553876956680014e-06, + "loss": 0.541, + "step": 11855 + }, + { + "epoch": 2.16, + "learning_rate": 2.3553330906443393e-06, + "loss": 0.0786, + "step": 11856 + }, + { + "epoch": 2.16, + "learning_rate": 2.355278485620677e-06, + "loss": 0.9074, + "step": 11857 + }, + { + "epoch": 2.16, + "learning_rate": 2.355223880597015e-06, + "loss": 1.0032, + "step": 11858 + }, + { + "epoch": 2.16, + "learning_rate": 2.3551692755733527e-06, + "loss": 0.4213, + "step": 11859 + }, + { + "epoch": 2.16, + "learning_rate": 2.3551146705496907e-06, + "loss": 0.7517, + "step": 11860 + }, + { + "epoch": 2.16, + "learning_rate": 2.3550600655260286e-06, + "loss": 0.3717, + "step": 11861 + }, + { + "epoch": 2.16, + "learning_rate": 2.3550054605023666e-06, + "loss": 1.5998, + "step": 11862 + }, + { + "epoch": 2.16, + "learning_rate": 2.354950855478704e-06, + "loss": 0.3642, + "step": 11863 + }, + { + "epoch": 2.16, + "learning_rate": 2.354896250455042e-06, + "loss": 0.3969, + "step": 11864 + }, + { + "epoch": 2.16, + "learning_rate": 2.3548416454313796e-06, + "loss": 1.2487, + "step": 11865 + }, + { + "epoch": 2.16, + "learning_rate": 2.3547870404077175e-06, + "loss": 0.9232, + "step": 11866 + }, + { + "epoch": 2.16, + "learning_rate": 2.3547324353840554e-06, + "loss": 0.3304, + "step": 11867 + }, + { + "epoch": 2.16, + "learning_rate": 2.354677830360393e-06, + "loss": 0.457, + "step": 11868 + }, + { + "epoch": 2.16, + "learning_rate": 2.354623225336731e-06, + "loss": 0.2855, + "step": 11869 + }, + { + "epoch": 2.16, + "learning_rate": 2.354568620313069e-06, + "loss": 0.638, + "step": 11870 + }, + { + "epoch": 2.16, + "learning_rate": 2.3545140152894064e-06, + "loss": 0.4085, + "step": 11871 + }, + { + "epoch": 2.16, + "learning_rate": 2.3544594102657443e-06, + "loss": 2.2707, + "step": 11872 + }, + { + "epoch": 2.16, + "learning_rate": 2.3544048052420823e-06, + "loss": 0.0467, + "step": 11873 + }, + { + "epoch": 2.16, + "learning_rate": 2.35435020021842e-06, + "loss": 0.107, + "step": 11874 + }, + { + "epoch": 2.16, + "learning_rate": 2.354295595194758e-06, + "loss": 0.3384, + "step": 11875 + }, + { + "epoch": 2.16, + "learning_rate": 2.354240990171096e-06, + "loss": 1.863, + "step": 11876 + }, + { + "epoch": 2.16, + "learning_rate": 2.3541863851474336e-06, + "loss": 0.3258, + "step": 11877 + }, + { + "epoch": 2.16, + "learning_rate": 2.3541317801237715e-06, + "loss": 0.5106, + "step": 11878 + }, + { + "epoch": 2.16, + "learning_rate": 2.354077175100109e-06, + "loss": 0.8197, + "step": 11879 + }, + { + "epoch": 2.16, + "learning_rate": 2.354022570076447e-06, + "loss": 1.7413, + "step": 11880 + }, + { + "epoch": 2.16, + "learning_rate": 2.353967965052785e-06, + "loss": 0.0214, + "step": 11881 + }, + { + "epoch": 2.16, + "learning_rate": 2.3539133600291225e-06, + "loss": 0.515, + "step": 11882 + }, + { + "epoch": 2.16, + "learning_rate": 2.3538587550054604e-06, + "loss": 2.7347, + "step": 11883 + }, + { + "epoch": 2.16, + "learning_rate": 2.3538041499817984e-06, + "loss": 0.1975, + "step": 11884 + }, + { + "epoch": 2.16, + "learning_rate": 2.3537495449581363e-06, + "loss": 1.2334, + "step": 11885 + }, + { + "epoch": 2.16, + "learning_rate": 2.3536949399344742e-06, + "loss": 0.0333, + "step": 11886 + }, + { + "epoch": 2.16, + "learning_rate": 2.353640334910812e-06, + "loss": 1.2774, + "step": 11887 + }, + { + "epoch": 2.16, + "learning_rate": 2.3535857298871497e-06, + "loss": 0.3293, + "step": 11888 + }, + { + "epoch": 2.16, + "learning_rate": 2.3535311248634877e-06, + "loss": 1.5247, + "step": 11889 + }, + { + "epoch": 2.16, + "learning_rate": 2.3534765198398256e-06, + "loss": 0.6957, + "step": 11890 + }, + { + "epoch": 2.16, + "learning_rate": 2.353421914816163e-06, + "loss": 1.8888, + "step": 11891 + }, + { + "epoch": 2.16, + "learning_rate": 2.353367309792501e-06, + "loss": 1.2973, + "step": 11892 + }, + { + "epoch": 2.16, + "learning_rate": 2.3533127047688386e-06, + "loss": 0.125, + "step": 11893 + }, + { + "epoch": 2.16, + "learning_rate": 2.3532580997451765e-06, + "loss": 1.6802, + "step": 11894 + }, + { + "epoch": 2.16, + "learning_rate": 2.3532034947215145e-06, + "loss": 0.4691, + "step": 11895 + }, + { + "epoch": 2.16, + "learning_rate": 2.353148889697852e-06, + "loss": 1.4476, + "step": 11896 + }, + { + "epoch": 2.16, + "learning_rate": 2.35309428467419e-06, + "loss": 0.905, + "step": 11897 + }, + { + "epoch": 2.16, + "learning_rate": 2.353039679650528e-06, + "loss": 1.2457, + "step": 11898 + }, + { + "epoch": 2.16, + "learning_rate": 2.352985074626866e-06, + "loss": 0.0188, + "step": 11899 + }, + { + "epoch": 2.16, + "learning_rate": 2.3529304696032038e-06, + "loss": 0.3284, + "step": 11900 + }, + { + "epoch": 2.16, + "learning_rate": 2.3528758645795417e-06, + "loss": 0.2784, + "step": 11901 + }, + { + "epoch": 2.16, + "learning_rate": 2.3528212595558792e-06, + "loss": 2.1534, + "step": 11902 + }, + { + "epoch": 2.16, + "learning_rate": 2.352766654532217e-06, + "loss": 0.7065, + "step": 11903 + }, + { + "epoch": 2.16, + "learning_rate": 2.352712049508555e-06, + "loss": 0.5435, + "step": 11904 + }, + { + "epoch": 2.16, + "learning_rate": 2.3526574444848926e-06, + "loss": 0.0638, + "step": 11905 + }, + { + "epoch": 2.17, + "learning_rate": 2.3526028394612306e-06, + "loss": 1.5775, + "step": 11906 + }, + { + "epoch": 2.17, + "learning_rate": 2.352548234437568e-06, + "loss": 0.3209, + "step": 11907 + }, + { + "epoch": 2.17, + "learning_rate": 2.352493629413906e-06, + "loss": 0.6419, + "step": 11908 + }, + { + "epoch": 2.17, + "learning_rate": 2.352439024390244e-06, + "loss": 1.0489, + "step": 11909 + }, + { + "epoch": 2.17, + "learning_rate": 2.3523844193665815e-06, + "loss": 0.2156, + "step": 11910 + }, + { + "epoch": 2.17, + "learning_rate": 2.3523298143429194e-06, + "loss": 0.5477, + "step": 11911 + }, + { + "epoch": 2.17, + "learning_rate": 2.3522752093192574e-06, + "loss": 0.3612, + "step": 11912 + }, + { + "epoch": 2.17, + "learning_rate": 2.3522206042955953e-06, + "loss": 2.502, + "step": 11913 + }, + { + "epoch": 2.17, + "learning_rate": 2.3521659992719333e-06, + "loss": 0.4609, + "step": 11914 + }, + { + "epoch": 2.17, + "learning_rate": 2.3521113942482712e-06, + "loss": 0.2817, + "step": 11915 + }, + { + "epoch": 2.17, + "learning_rate": 2.3520567892246087e-06, + "loss": 0.3071, + "step": 11916 + }, + { + "epoch": 2.17, + "learning_rate": 2.3520021842009467e-06, + "loss": 2.0958, + "step": 11917 + }, + { + "epoch": 2.17, + "learning_rate": 2.3519475791772846e-06, + "loss": 0.1681, + "step": 11918 + }, + { + "epoch": 2.17, + "learning_rate": 2.351892974153622e-06, + "loss": 1.1446, + "step": 11919 + }, + { + "epoch": 2.17, + "learning_rate": 2.35183836912996e-06, + "loss": 0.0731, + "step": 11920 + }, + { + "epoch": 2.17, + "learning_rate": 2.3517837641062976e-06, + "loss": 0.3798, + "step": 11921 + }, + { + "epoch": 2.17, + "learning_rate": 2.3517291590826356e-06, + "loss": 0.7457, + "step": 11922 + }, + { + "epoch": 2.17, + "learning_rate": 2.3516745540589735e-06, + "loss": 1.0982, + "step": 11923 + }, + { + "epoch": 2.17, + "learning_rate": 2.351619949035311e-06, + "loss": 0.1225, + "step": 11924 + }, + { + "epoch": 2.17, + "learning_rate": 2.351565344011649e-06, + "loss": 0.6719, + "step": 11925 + }, + { + "epoch": 2.17, + "learning_rate": 2.351510738987987e-06, + "loss": 0.1531, + "step": 11926 + }, + { + "epoch": 2.17, + "learning_rate": 2.351456133964325e-06, + "loss": 0.4214, + "step": 11927 + }, + { + "epoch": 2.17, + "learning_rate": 2.3514015289406628e-06, + "loss": 3.1068, + "step": 11928 + }, + { + "epoch": 2.17, + "learning_rate": 2.3513469239170007e-06, + "loss": 0.527, + "step": 11929 + }, + { + "epoch": 2.17, + "learning_rate": 2.3512923188933383e-06, + "loss": 0.468, + "step": 11930 + }, + { + "epoch": 2.17, + "learning_rate": 2.351237713869676e-06, + "loss": 1.7971, + "step": 11931 + }, + { + "epoch": 2.17, + "learning_rate": 2.3511831088460137e-06, + "loss": 2.8937, + "step": 11932 + }, + { + "epoch": 2.17, + "learning_rate": 2.3511285038223517e-06, + "loss": 1.4733, + "step": 11933 + }, + { + "epoch": 2.17, + "learning_rate": 2.3510738987986896e-06, + "loss": 0.3989, + "step": 11934 + }, + { + "epoch": 2.17, + "learning_rate": 2.351019293775027e-06, + "loss": 1.0587, + "step": 11935 + }, + { + "epoch": 2.17, + "learning_rate": 2.350964688751365e-06, + "loss": 2.0493, + "step": 11936 + }, + { + "epoch": 2.17, + "learning_rate": 2.350910083727703e-06, + "loss": 0.2421, + "step": 11937 + }, + { + "epoch": 2.17, + "learning_rate": 2.3508554787040405e-06, + "loss": 2.7857, + "step": 11938 + }, + { + "epoch": 2.17, + "learning_rate": 2.3508008736803785e-06, + "loss": 0.3645, + "step": 11939 + }, + { + "epoch": 2.17, + "learning_rate": 2.3507462686567164e-06, + "loss": 0.9404, + "step": 11940 + }, + { + "epoch": 2.17, + "learning_rate": 2.3506916636330544e-06, + "loss": 2.3488, + "step": 11941 + }, + { + "epoch": 2.17, + "learning_rate": 2.3506370586093923e-06, + "loss": 0.3705, + "step": 11942 + }, + { + "epoch": 2.17, + "learning_rate": 2.3505824535857302e-06, + "loss": 0.2464, + "step": 11943 + }, + { + "epoch": 2.17, + "learning_rate": 2.3505278485620678e-06, + "loss": 1.3194, + "step": 11944 + }, + { + "epoch": 2.17, + "learning_rate": 2.3504732435384057e-06, + "loss": 0.7143, + "step": 11945 + }, + { + "epoch": 2.17, + "learning_rate": 2.3504186385147432e-06, + "loss": 1.0072, + "step": 11946 + }, + { + "epoch": 2.17, + "learning_rate": 2.350364033491081e-06, + "loss": 0.1346, + "step": 11947 + }, + { + "epoch": 2.17, + "learning_rate": 2.350309428467419e-06, + "loss": 0.391, + "step": 11948 + }, + { + "epoch": 2.17, + "learning_rate": 2.3502548234437566e-06, + "loss": 2.3604, + "step": 11949 + }, + { + "epoch": 2.17, + "learning_rate": 2.3502002184200946e-06, + "loss": 0.2807, + "step": 11950 + }, + { + "epoch": 2.17, + "learning_rate": 2.3501456133964325e-06, + "loss": 0.9712, + "step": 11951 + }, + { + "epoch": 2.17, + "learning_rate": 2.35009100837277e-06, + "loss": 0.3892, + "step": 11952 + }, + { + "epoch": 2.17, + "learning_rate": 2.350036403349108e-06, + "loss": 1.9789, + "step": 11953 + }, + { + "epoch": 2.17, + "learning_rate": 2.349981798325446e-06, + "loss": 1.105, + "step": 11954 + }, + { + "epoch": 2.17, + "learning_rate": 2.349927193301784e-06, + "loss": 0.5126, + "step": 11955 + }, + { + "epoch": 2.17, + "learning_rate": 2.349872588278122e-06, + "loss": 2.1262, + "step": 11956 + }, + { + "epoch": 2.17, + "learning_rate": 2.3498179832544598e-06, + "loss": 1.2029, + "step": 11957 + }, + { + "epoch": 2.17, + "learning_rate": 2.3497633782307973e-06, + "loss": 0.6906, + "step": 11958 + }, + { + "epoch": 2.17, + "learning_rate": 2.3497087732071352e-06, + "loss": 0.6845, + "step": 11959 + }, + { + "epoch": 2.17, + "learning_rate": 2.3496541681834727e-06, + "loss": 0.7998, + "step": 11960 + }, + { + "epoch": 2.18, + "learning_rate": 2.3495995631598107e-06, + "loss": 1.8, + "step": 11961 + }, + { + "epoch": 2.18, + "learning_rate": 2.3495449581361486e-06, + "loss": 1.4614, + "step": 11962 + }, + { + "epoch": 2.18, + "learning_rate": 2.349490353112486e-06, + "loss": 0.0508, + "step": 11963 + }, + { + "epoch": 2.18, + "learning_rate": 2.349435748088824e-06, + "loss": 0.6244, + "step": 11964 + }, + { + "epoch": 2.18, + "learning_rate": 2.349381143065162e-06, + "loss": 0.6223, + "step": 11965 + }, + { + "epoch": 2.18, + "learning_rate": 2.3493265380415e-06, + "loss": 1.3602, + "step": 11966 + }, + { + "epoch": 2.18, + "learning_rate": 2.349271933017838e-06, + "loss": 0.0222, + "step": 11967 + }, + { + "epoch": 2.18, + "learning_rate": 2.349217327994176e-06, + "loss": 0.833, + "step": 11968 + }, + { + "epoch": 2.18, + "learning_rate": 2.3491627229705134e-06, + "loss": 0.5923, + "step": 11969 + }, + { + "epoch": 2.18, + "learning_rate": 2.3491081179468513e-06, + "loss": 0.4043, + "step": 11970 + }, + { + "epoch": 2.18, + "learning_rate": 2.3490535129231893e-06, + "loss": 0.4463, + "step": 11971 + }, + { + "epoch": 2.18, + "learning_rate": 2.348998907899527e-06, + "loss": 0.5395, + "step": 11972 + }, + { + "epoch": 2.18, + "learning_rate": 2.3489443028758647e-06, + "loss": 0.1582, + "step": 11973 + }, + { + "epoch": 2.18, + "learning_rate": 2.3488896978522023e-06, + "loss": 0.2012, + "step": 11974 + }, + { + "epoch": 2.18, + "learning_rate": 2.34883509282854e-06, + "loss": 3.3044, + "step": 11975 + }, + { + "epoch": 2.18, + "learning_rate": 2.348780487804878e-06, + "loss": 1.3526, + "step": 11976 + }, + { + "epoch": 2.18, + "learning_rate": 2.3487258827812157e-06, + "loss": 0.0821, + "step": 11977 + }, + { + "epoch": 2.18, + "learning_rate": 2.3486712777575536e-06, + "loss": 0.6304, + "step": 11978 + }, + { + "epoch": 2.18, + "learning_rate": 2.3486166727338915e-06, + "loss": 0.2735, + "step": 11979 + }, + { + "epoch": 2.18, + "learning_rate": 2.3485620677102295e-06, + "loss": 0.8059, + "step": 11980 + }, + { + "epoch": 2.18, + "learning_rate": 2.3485074626865674e-06, + "loss": 0.1677, + "step": 11981 + }, + { + "epoch": 2.18, + "learning_rate": 2.3484528576629054e-06, + "loss": 0.0574, + "step": 11982 + }, + { + "epoch": 2.18, + "learning_rate": 2.348398252639243e-06, + "loss": 0.7185, + "step": 11983 + }, + { + "epoch": 2.18, + "learning_rate": 2.348343647615581e-06, + "loss": 0.9276, + "step": 11984 + }, + { + "epoch": 2.18, + "learning_rate": 2.3482890425919188e-06, + "loss": 0.8268, + "step": 11985 + }, + { + "epoch": 2.18, + "learning_rate": 2.3482344375682563e-06, + "loss": 0.2381, + "step": 11986 + }, + { + "epoch": 2.18, + "learning_rate": 2.3481798325445942e-06, + "loss": 0.02, + "step": 11987 + }, + { + "epoch": 2.18, + "learning_rate": 2.3481252275209318e-06, + "loss": 0.2427, + "step": 11988 + }, + { + "epoch": 2.18, + "learning_rate": 2.3480706224972697e-06, + "loss": 1.9895, + "step": 11989 + }, + { + "epoch": 2.18, + "learning_rate": 2.3480160174736077e-06, + "loss": 0.4334, + "step": 11990 + }, + { + "epoch": 2.18, + "learning_rate": 2.347961412449945e-06, + "loss": 0.8732, + "step": 11991 + }, + { + "epoch": 2.18, + "learning_rate": 2.347906807426283e-06, + "loss": 0.3596, + "step": 11992 + }, + { + "epoch": 2.18, + "learning_rate": 2.347852202402621e-06, + "loss": 0.1223, + "step": 11993 + }, + { + "epoch": 2.18, + "learning_rate": 2.347797597378959e-06, + "loss": 1.223, + "step": 11994 + }, + { + "epoch": 2.18, + "learning_rate": 2.347742992355297e-06, + "loss": 0.1439, + "step": 11995 + }, + { + "epoch": 2.18, + "learning_rate": 2.347688387331635e-06, + "loss": 1.0394, + "step": 11996 + }, + { + "epoch": 2.18, + "learning_rate": 2.3476337823079724e-06, + "loss": 1.4677, + "step": 11997 + }, + { + "epoch": 2.18, + "learning_rate": 2.3475791772843104e-06, + "loss": 1.8223, + "step": 11998 + }, + { + "epoch": 2.18, + "learning_rate": 2.3475245722606483e-06, + "loss": 0.3501, + "step": 11999 + }, + { + "epoch": 2.18, + "learning_rate": 2.347469967236986e-06, + "loss": 0.4447, + "step": 12000 + }, + { + "epoch": 2.18, + "learning_rate": 2.3474153622133238e-06, + "loss": 0.7342, + "step": 12001 + }, + { + "epoch": 2.18, + "learning_rate": 2.3473607571896613e-06, + "loss": 1.2993, + "step": 12002 + }, + { + "epoch": 2.18, + "learning_rate": 2.3473061521659992e-06, + "loss": 0.3978, + "step": 12003 + }, + { + "epoch": 2.18, + "learning_rate": 2.347251547142337e-06, + "loss": 0.6759, + "step": 12004 + }, + { + "epoch": 2.18, + "learning_rate": 2.3471969421186747e-06, + "loss": 0.4503, + "step": 12005 + }, + { + "epoch": 2.18, + "learning_rate": 2.3471423370950126e-06, + "loss": 0.0679, + "step": 12006 + }, + { + "epoch": 2.18, + "learning_rate": 2.3470877320713506e-06, + "loss": 1.7493, + "step": 12007 + }, + { + "epoch": 2.18, + "learning_rate": 2.3470331270476885e-06, + "loss": 1.0149, + "step": 12008 + }, + { + "epoch": 2.18, + "learning_rate": 2.3469785220240265e-06, + "loss": 0.1726, + "step": 12009 + }, + { + "epoch": 2.18, + "learning_rate": 2.3469239170003644e-06, + "loss": 1.6217, + "step": 12010 + }, + { + "epoch": 2.18, + "learning_rate": 2.346869311976702e-06, + "loss": 0.0589, + "step": 12011 + }, + { + "epoch": 2.18, + "learning_rate": 2.34681470695304e-06, + "loss": 1.164, + "step": 12012 + }, + { + "epoch": 2.18, + "learning_rate": 2.3467601019293774e-06, + "loss": 2.0395, + "step": 12013 + }, + { + "epoch": 2.18, + "learning_rate": 2.3467054969057153e-06, + "loss": 0.0721, + "step": 12014 + }, + { + "epoch": 2.18, + "learning_rate": 2.3466508918820533e-06, + "loss": 1.0129, + "step": 12015 + }, + { + "epoch": 2.19, + "learning_rate": 2.346596286858391e-06, + "loss": 1.2638, + "step": 12016 + }, + { + "epoch": 2.19, + "learning_rate": 2.3465416818347287e-06, + "loss": 0.2424, + "step": 12017 + }, + { + "epoch": 2.19, + "learning_rate": 2.3464870768110667e-06, + "loss": 0.9687, + "step": 12018 + }, + { + "epoch": 2.19, + "learning_rate": 2.346432471787404e-06, + "loss": 0.207, + "step": 12019 + }, + { + "epoch": 2.19, + "learning_rate": 2.346377866763742e-06, + "loss": 0.5086, + "step": 12020 + }, + { + "epoch": 2.19, + "learning_rate": 2.34632326174008e-06, + "loss": 0.5971, + "step": 12021 + }, + { + "epoch": 2.19, + "learning_rate": 2.346268656716418e-06, + "loss": 0.0247, + "step": 12022 + }, + { + "epoch": 2.19, + "learning_rate": 2.346214051692756e-06, + "loss": 0.3325, + "step": 12023 + }, + { + "epoch": 2.19, + "learning_rate": 2.346159446669094e-06, + "loss": 0.4776, + "step": 12024 + }, + { + "epoch": 2.19, + "learning_rate": 2.3461048416454314e-06, + "loss": 1.0016, + "step": 12025 + }, + { + "epoch": 2.19, + "learning_rate": 2.3460502366217694e-06, + "loss": 1.4345, + "step": 12026 + }, + { + "epoch": 2.19, + "learning_rate": 2.345995631598107e-06, + "loss": 0.1071, + "step": 12027 + }, + { + "epoch": 2.19, + "learning_rate": 2.345941026574445e-06, + "loss": 0.3256, + "step": 12028 + }, + { + "epoch": 2.19, + "learning_rate": 2.345886421550783e-06, + "loss": 1.4409, + "step": 12029 + }, + { + "epoch": 2.19, + "learning_rate": 2.3458318165271203e-06, + "loss": 0.3575, + "step": 12030 + }, + { + "epoch": 2.19, + "learning_rate": 2.3457772115034583e-06, + "loss": 1.4672, + "step": 12031 + }, + { + "epoch": 2.19, + "learning_rate": 2.345722606479796e-06, + "loss": 0.6719, + "step": 12032 + }, + { + "epoch": 2.19, + "learning_rate": 2.3456680014561337e-06, + "loss": 0.1754, + "step": 12033 + }, + { + "epoch": 2.19, + "learning_rate": 2.3456133964324717e-06, + "loss": 0.0724, + "step": 12034 + }, + { + "epoch": 2.19, + "learning_rate": 2.3455587914088096e-06, + "loss": 0.8859, + "step": 12035 + }, + { + "epoch": 2.19, + "learning_rate": 2.3455041863851475e-06, + "loss": 0.0265, + "step": 12036 + }, + { + "epoch": 2.19, + "learning_rate": 2.3454495813614855e-06, + "loss": 0.7746, + "step": 12037 + }, + { + "epoch": 2.19, + "learning_rate": 2.3453949763378234e-06, + "loss": 1.0175, + "step": 12038 + }, + { + "epoch": 2.19, + "learning_rate": 2.345340371314161e-06, + "loss": 0.392, + "step": 12039 + }, + { + "epoch": 2.19, + "learning_rate": 2.345285766290499e-06, + "loss": 0.5592, + "step": 12040 + }, + { + "epoch": 2.19, + "learning_rate": 2.3452311612668364e-06, + "loss": 0.2718, + "step": 12041 + }, + { + "epoch": 2.19, + "learning_rate": 2.3451765562431744e-06, + "loss": 0.989, + "step": 12042 + }, + { + "epoch": 2.19, + "learning_rate": 2.3451219512195123e-06, + "loss": 0.2823, + "step": 12043 + }, + { + "epoch": 2.19, + "learning_rate": 2.34506734619585e-06, + "loss": 0.0473, + "step": 12044 + }, + { + "epoch": 2.19, + "learning_rate": 2.3450127411721878e-06, + "loss": 1.0704, + "step": 12045 + }, + { + "epoch": 2.19, + "learning_rate": 2.3449581361485257e-06, + "loss": 1.2606, + "step": 12046 + }, + { + "epoch": 2.19, + "learning_rate": 2.3449035311248637e-06, + "loss": 0.1738, + "step": 12047 + }, + { + "epoch": 2.19, + "learning_rate": 2.3448489261012016e-06, + "loss": 0.6288, + "step": 12048 + }, + { + "epoch": 2.19, + "learning_rate": 2.344794321077539e-06, + "loss": 0.7203, + "step": 12049 + }, + { + "epoch": 2.19, + "learning_rate": 2.344739716053877e-06, + "loss": 0.6704, + "step": 12050 + }, + { + "epoch": 2.19, + "learning_rate": 2.344685111030215e-06, + "loss": 1.3623, + "step": 12051 + }, + { + "epoch": 2.19, + "learning_rate": 2.344630506006553e-06, + "loss": 0.4212, + "step": 12052 + }, + { + "epoch": 2.19, + "learning_rate": 2.3445759009828905e-06, + "loss": 1.1902, + "step": 12053 + }, + { + "epoch": 2.19, + "learning_rate": 2.3445212959592284e-06, + "loss": 0.5212, + "step": 12054 + }, + { + "epoch": 2.19, + "learning_rate": 2.344466690935566e-06, + "loss": 0.9711, + "step": 12055 + }, + { + "epoch": 2.19, + "learning_rate": 2.344412085911904e-06, + "loss": 0.5572, + "step": 12056 + }, + { + "epoch": 2.19, + "learning_rate": 2.344357480888242e-06, + "loss": 0.2549, + "step": 12057 + }, + { + "epoch": 2.19, + "learning_rate": 2.3443028758645793e-06, + "loss": 1.1357, + "step": 12058 + }, + { + "epoch": 2.19, + "learning_rate": 2.3442482708409173e-06, + "loss": 0.7736, + "step": 12059 + }, + { + "epoch": 2.19, + "learning_rate": 2.3441936658172552e-06, + "loss": 2.1841, + "step": 12060 + }, + { + "epoch": 2.19, + "learning_rate": 2.344139060793593e-06, + "loss": 0.7151, + "step": 12061 + }, + { + "epoch": 2.19, + "learning_rate": 2.344084455769931e-06, + "loss": 0.9084, + "step": 12062 + }, + { + "epoch": 2.19, + "learning_rate": 2.344029850746269e-06, + "loss": 0.4298, + "step": 12063 + }, + { + "epoch": 2.19, + "learning_rate": 2.3439752457226066e-06, + "loss": 0.0618, + "step": 12064 + }, + { + "epoch": 2.19, + "learning_rate": 2.3439206406989445e-06, + "loss": 0.0561, + "step": 12065 + }, + { + "epoch": 2.19, + "learning_rate": 2.3438660356752825e-06, + "loss": 0.3839, + "step": 12066 + }, + { + "epoch": 2.19, + "learning_rate": 2.34381143065162e-06, + "loss": 1.5098, + "step": 12067 + }, + { + "epoch": 2.19, + "learning_rate": 2.343756825627958e-06, + "loss": 0.0519, + "step": 12068 + }, + { + "epoch": 2.19, + "learning_rate": 2.3437022206042954e-06, + "loss": 0.0771, + "step": 12069 + }, + { + "epoch": 2.19, + "learning_rate": 2.3436476155806334e-06, + "loss": 0.1724, + "step": 12070 + }, + { + "epoch": 2.2, + "learning_rate": 2.3435930105569713e-06, + "loss": 1.0438, + "step": 12071 + }, + { + "epoch": 2.2, + "learning_rate": 2.343538405533309e-06, + "loss": 0.7848, + "step": 12072 + }, + { + "epoch": 2.2, + "learning_rate": 2.343483800509647e-06, + "loss": 1.5762, + "step": 12073 + }, + { + "epoch": 2.2, + "learning_rate": 2.3434291954859847e-06, + "loss": 0.689, + "step": 12074 + }, + { + "epoch": 2.2, + "learning_rate": 2.3433745904623227e-06, + "loss": 1.465, + "step": 12075 + }, + { + "epoch": 2.2, + "learning_rate": 2.3433199854386606e-06, + "loss": 1.1675, + "step": 12076 + }, + { + "epoch": 2.2, + "learning_rate": 2.3432653804149986e-06, + "loss": 0.4259, + "step": 12077 + }, + { + "epoch": 2.2, + "learning_rate": 2.343210775391336e-06, + "loss": 1.1469, + "step": 12078 + }, + { + "epoch": 2.2, + "learning_rate": 2.343156170367674e-06, + "loss": 0.9468, + "step": 12079 + }, + { + "epoch": 2.2, + "learning_rate": 2.3431015653440115e-06, + "loss": 0.0554, + "step": 12080 + }, + { + "epoch": 2.2, + "learning_rate": 2.3430469603203495e-06, + "loss": 1.6293, + "step": 12081 + }, + { + "epoch": 2.2, + "learning_rate": 2.3429923552966874e-06, + "loss": 0.3771, + "step": 12082 + }, + { + "epoch": 2.2, + "learning_rate": 2.342937750273025e-06, + "loss": 0.0848, + "step": 12083 + }, + { + "epoch": 2.2, + "learning_rate": 2.342883145249363e-06, + "loss": 0.7944, + "step": 12084 + }, + { + "epoch": 2.2, + "learning_rate": 2.342828540225701e-06, + "loss": 1.2772, + "step": 12085 + }, + { + "epoch": 2.2, + "learning_rate": 2.3427739352020384e-06, + "loss": 2.3298, + "step": 12086 + }, + { + "epoch": 2.2, + "learning_rate": 2.3427193301783763e-06, + "loss": 0.535, + "step": 12087 + }, + { + "epoch": 2.2, + "learning_rate": 2.3426647251547142e-06, + "loss": 1.1805, + "step": 12088 + }, + { + "epoch": 2.2, + "learning_rate": 2.342610120131052e-06, + "loss": 1.3274, + "step": 12089 + }, + { + "epoch": 2.2, + "learning_rate": 2.34255551510739e-06, + "loss": 0.9024, + "step": 12090 + }, + { + "epoch": 2.2, + "learning_rate": 2.342500910083728e-06, + "loss": 0.759, + "step": 12091 + }, + { + "epoch": 2.2, + "learning_rate": 2.3424463050600656e-06, + "loss": 0.0391, + "step": 12092 + }, + { + "epoch": 2.2, + "learning_rate": 2.3423917000364035e-06, + "loss": 0.9331, + "step": 12093 + }, + { + "epoch": 2.2, + "learning_rate": 2.342337095012741e-06, + "loss": 0.791, + "step": 12094 + }, + { + "epoch": 2.2, + "learning_rate": 2.342282489989079e-06, + "loss": 0.159, + "step": 12095 + }, + { + "epoch": 2.2, + "learning_rate": 2.342227884965417e-06, + "loss": 0.0399, + "step": 12096 + }, + { + "epoch": 2.2, + "learning_rate": 2.3421732799417545e-06, + "loss": 0.1615, + "step": 12097 + }, + { + "epoch": 2.2, + "learning_rate": 2.3421186749180924e-06, + "loss": 0.0598, + "step": 12098 + }, + { + "epoch": 2.2, + "learning_rate": 2.3420640698944304e-06, + "loss": 0.3697, + "step": 12099 + }, + { + "epoch": 2.2, + "learning_rate": 2.342009464870768e-06, + "loss": 0.0404, + "step": 12100 + }, + { + "epoch": 2.2, + "learning_rate": 2.341954859847106e-06, + "loss": 0.127, + "step": 12101 + }, + { + "epoch": 2.2, + "learning_rate": 2.3419002548234438e-06, + "loss": 1.4704, + "step": 12102 + }, + { + "epoch": 2.2, + "learning_rate": 2.3418456497997817e-06, + "loss": 0.2302, + "step": 12103 + }, + { + "epoch": 2.2, + "learning_rate": 2.3417910447761196e-06, + "loss": 0.6845, + "step": 12104 + }, + { + "epoch": 2.2, + "learning_rate": 2.3417364397524576e-06, + "loss": 0.5255, + "step": 12105 + }, + { + "epoch": 2.2, + "learning_rate": 2.341681834728795e-06, + "loss": 1.6533, + "step": 12106 + }, + { + "epoch": 2.2, + "learning_rate": 2.341627229705133e-06, + "loss": 1.0072, + "step": 12107 + }, + { + "epoch": 2.2, + "learning_rate": 2.3415726246814706e-06, + "loss": 0.6061, + "step": 12108 + }, + { + "epoch": 2.2, + "learning_rate": 2.3415180196578085e-06, + "loss": 0.7424, + "step": 12109 + }, + { + "epoch": 2.2, + "learning_rate": 2.3414634146341465e-06, + "loss": 0.7086, + "step": 12110 + }, + { + "epoch": 2.2, + "learning_rate": 2.341408809610484e-06, + "loss": 0.8516, + "step": 12111 + }, + { + "epoch": 2.2, + "learning_rate": 2.341354204586822e-06, + "loss": 0.3357, + "step": 12112 + }, + { + "epoch": 2.2, + "learning_rate": 2.34129959956316e-06, + "loss": 0.4578, + "step": 12113 + }, + { + "epoch": 2.2, + "learning_rate": 2.3412449945394974e-06, + "loss": 0.5352, + "step": 12114 + }, + { + "epoch": 2.2, + "learning_rate": 2.3411903895158353e-06, + "loss": 0.6378, + "step": 12115 + }, + { + "epoch": 2.2, + "learning_rate": 2.3411357844921733e-06, + "loss": 0.6663, + "step": 12116 + }, + { + "epoch": 2.2, + "learning_rate": 2.3410811794685112e-06, + "loss": 0.776, + "step": 12117 + }, + { + "epoch": 2.2, + "learning_rate": 2.341026574444849e-06, + "loss": 0.9996, + "step": 12118 + }, + { + "epoch": 2.2, + "learning_rate": 2.340971969421187e-06, + "loss": 0.3077, + "step": 12119 + }, + { + "epoch": 2.2, + "learning_rate": 2.3409173643975246e-06, + "loss": 2.6622, + "step": 12120 + }, + { + "epoch": 2.2, + "learning_rate": 2.3408627593738626e-06, + "loss": 1.2569, + "step": 12121 + }, + { + "epoch": 2.2, + "learning_rate": 2.3408081543502e-06, + "loss": 0.7862, + "step": 12122 + }, + { + "epoch": 2.2, + "learning_rate": 2.340753549326538e-06, + "loss": 0.6924, + "step": 12123 + }, + { + "epoch": 2.2, + "learning_rate": 2.340698944302876e-06, + "loss": 0.4972, + "step": 12124 + }, + { + "epoch": 2.2, + "learning_rate": 2.3406443392792135e-06, + "loss": 2.2981, + "step": 12125 + }, + { + "epoch": 2.21, + "learning_rate": 2.3405897342555514e-06, + "loss": 0.6954, + "step": 12126 + }, + { + "epoch": 2.21, + "learning_rate": 2.3405351292318894e-06, + "loss": 1.5664, + "step": 12127 + }, + { + "epoch": 2.21, + "learning_rate": 2.340480524208227e-06, + "loss": 1.4422, + "step": 12128 + }, + { + "epoch": 2.21, + "learning_rate": 2.340425919184565e-06, + "loss": 1.0181, + "step": 12129 + }, + { + "epoch": 2.21, + "learning_rate": 2.340371314160903e-06, + "loss": 1.1665, + "step": 12130 + }, + { + "epoch": 2.21, + "learning_rate": 2.3403167091372407e-06, + "loss": 0.0218, + "step": 12131 + }, + { + "epoch": 2.21, + "learning_rate": 2.3402621041135787e-06, + "loss": 0.2546, + "step": 12132 + }, + { + "epoch": 2.21, + "learning_rate": 2.3402074990899166e-06, + "loss": 1.6843, + "step": 12133 + }, + { + "epoch": 2.21, + "learning_rate": 2.340152894066254e-06, + "loss": 0.8885, + "step": 12134 + }, + { + "epoch": 2.21, + "learning_rate": 2.340098289042592e-06, + "loss": 0.0508, + "step": 12135 + }, + { + "epoch": 2.21, + "learning_rate": 2.3400436840189296e-06, + "loss": 0.0896, + "step": 12136 + }, + { + "epoch": 2.21, + "learning_rate": 2.3399890789952675e-06, + "loss": 0.0301, + "step": 12137 + }, + { + "epoch": 2.21, + "learning_rate": 2.3399344739716055e-06, + "loss": 0.1363, + "step": 12138 + }, + { + "epoch": 2.21, + "learning_rate": 2.339879868947943e-06, + "loss": 0.8553, + "step": 12139 + }, + { + "epoch": 2.21, + "learning_rate": 2.339825263924281e-06, + "loss": 0.6158, + "step": 12140 + }, + { + "epoch": 2.21, + "learning_rate": 2.339770658900619e-06, + "loss": 0.5951, + "step": 12141 + }, + { + "epoch": 2.21, + "learning_rate": 2.339716053876957e-06, + "loss": 0.0239, + "step": 12142 + }, + { + "epoch": 2.21, + "learning_rate": 2.3396614488532948e-06, + "loss": 1.0102, + "step": 12143 + }, + { + "epoch": 2.21, + "learning_rate": 2.3396068438296327e-06, + "loss": 0.1165, + "step": 12144 + }, + { + "epoch": 2.21, + "learning_rate": 2.3395522388059702e-06, + "loss": 2.4293, + "step": 12145 + }, + { + "epoch": 2.21, + "learning_rate": 2.339497633782308e-06, + "loss": 0.7771, + "step": 12146 + }, + { + "epoch": 2.21, + "learning_rate": 2.339443028758646e-06, + "loss": 1.1924, + "step": 12147 + }, + { + "epoch": 2.21, + "learning_rate": 2.3393884237349837e-06, + "loss": 1.6794, + "step": 12148 + }, + { + "epoch": 2.21, + "learning_rate": 2.3393338187113216e-06, + "loss": 1.8736, + "step": 12149 + }, + { + "epoch": 2.21, + "learning_rate": 2.339279213687659e-06, + "loss": 0.7716, + "step": 12150 + }, + { + "epoch": 2.21, + "learning_rate": 2.339224608663997e-06, + "loss": 3.9537, + "step": 12151 + }, + { + "epoch": 2.21, + "learning_rate": 2.339170003640335e-06, + "loss": 1.7191, + "step": 12152 + }, + { + "epoch": 2.21, + "learning_rate": 2.3391153986166725e-06, + "loss": 0.1658, + "step": 12153 + }, + { + "epoch": 2.21, + "learning_rate": 2.3390607935930105e-06, + "loss": 0.9607, + "step": 12154 + }, + { + "epoch": 2.21, + "learning_rate": 2.3390061885693484e-06, + "loss": 0.0254, + "step": 12155 + }, + { + "epoch": 2.21, + "learning_rate": 2.3389515835456864e-06, + "loss": 0.7629, + "step": 12156 + }, + { + "epoch": 2.21, + "learning_rate": 2.3388969785220243e-06, + "loss": 1.9393, + "step": 12157 + }, + { + "epoch": 2.21, + "learning_rate": 2.3388423734983622e-06, + "loss": 0.9145, + "step": 12158 + }, + { + "epoch": 2.21, + "learning_rate": 2.3387877684746998e-06, + "loss": 0.9112, + "step": 12159 + }, + { + "epoch": 2.21, + "learning_rate": 2.3387331634510377e-06, + "loss": 0.0412, + "step": 12160 + }, + { + "epoch": 2.21, + "learning_rate": 2.3386785584273752e-06, + "loss": 1.4233, + "step": 12161 + }, + { + "epoch": 2.21, + "learning_rate": 2.338623953403713e-06, + "loss": 0.6251, + "step": 12162 + }, + { + "epoch": 2.21, + "learning_rate": 2.338569348380051e-06, + "loss": 1.2509, + "step": 12163 + }, + { + "epoch": 2.21, + "learning_rate": 2.3385147433563886e-06, + "loss": 1.6375, + "step": 12164 + }, + { + "epoch": 2.21, + "learning_rate": 2.3384601383327266e-06, + "loss": 1.6945, + "step": 12165 + }, + { + "epoch": 2.21, + "learning_rate": 2.3384055333090645e-06, + "loss": 0.3138, + "step": 12166 + }, + { + "epoch": 2.21, + "learning_rate": 2.338350928285402e-06, + "loss": 0.7598, + "step": 12167 + }, + { + "epoch": 2.21, + "learning_rate": 2.33829632326174e-06, + "loss": 0.5263, + "step": 12168 + }, + { + "epoch": 2.21, + "learning_rate": 2.338241718238078e-06, + "loss": 1.0245, + "step": 12169 + }, + { + "epoch": 2.21, + "learning_rate": 2.338187113214416e-06, + "loss": 0.8128, + "step": 12170 + }, + { + "epoch": 2.21, + "learning_rate": 2.338132508190754e-06, + "loss": 0.9413, + "step": 12171 + }, + { + "epoch": 2.21, + "learning_rate": 2.3380779031670918e-06, + "loss": 0.8836, + "step": 12172 + }, + { + "epoch": 2.21, + "learning_rate": 2.3380232981434293e-06, + "loss": 0.3833, + "step": 12173 + }, + { + "epoch": 2.21, + "learning_rate": 2.3379686931197672e-06, + "loss": 0.9894, + "step": 12174 + }, + { + "epoch": 2.21, + "learning_rate": 2.3379140880961047e-06, + "loss": 0.0823, + "step": 12175 + }, + { + "epoch": 2.21, + "learning_rate": 2.3378594830724427e-06, + "loss": 0.8615, + "step": 12176 + }, + { + "epoch": 2.21, + "learning_rate": 2.3378048780487806e-06, + "loss": 0.465, + "step": 12177 + }, + { + "epoch": 2.21, + "learning_rate": 2.337750273025118e-06, + "loss": 0.9783, + "step": 12178 + }, + { + "epoch": 2.21, + "learning_rate": 2.337695668001456e-06, + "loss": 1.5244, + "step": 12179 + }, + { + "epoch": 2.21, + "learning_rate": 2.337641062977794e-06, + "loss": 0.0474, + "step": 12180 + }, + { + "epoch": 2.22, + "learning_rate": 2.3375864579541316e-06, + "loss": 1.3695, + "step": 12181 + }, + { + "epoch": 2.22, + "learning_rate": 2.3375318529304695e-06, + "loss": 1.7856, + "step": 12182 + }, + { + "epoch": 2.22, + "learning_rate": 2.3374772479068074e-06, + "loss": 0.8589, + "step": 12183 + }, + { + "epoch": 2.22, + "learning_rate": 2.3374226428831454e-06, + "loss": 1.5398, + "step": 12184 + }, + { + "epoch": 2.22, + "learning_rate": 2.3373680378594833e-06, + "loss": 0.8562, + "step": 12185 + }, + { + "epoch": 2.22, + "learning_rate": 2.3373134328358213e-06, + "loss": 0.0666, + "step": 12186 + }, + { + "epoch": 2.22, + "learning_rate": 2.3372588278121588e-06, + "loss": 0.5133, + "step": 12187 + }, + { + "epoch": 2.22, + "learning_rate": 2.3372042227884967e-06, + "loss": 0.3858, + "step": 12188 + }, + { + "epoch": 2.22, + "learning_rate": 2.3371496177648343e-06, + "loss": 0.3779, + "step": 12189 + }, + { + "epoch": 2.22, + "learning_rate": 2.337095012741172e-06, + "loss": 1.0689, + "step": 12190 + }, + { + "epoch": 2.22, + "learning_rate": 2.33704040771751e-06, + "loss": 0.082, + "step": 12191 + }, + { + "epoch": 2.22, + "learning_rate": 2.3369858026938477e-06, + "loss": 1.8545, + "step": 12192 + }, + { + "epoch": 2.22, + "learning_rate": 2.3369311976701856e-06, + "loss": 0.9282, + "step": 12193 + }, + { + "epoch": 2.22, + "learning_rate": 2.3368765926465235e-06, + "loss": 0.492, + "step": 12194 + }, + { + "epoch": 2.22, + "learning_rate": 2.336821987622861e-06, + "loss": 0.3859, + "step": 12195 + }, + { + "epoch": 2.22, + "learning_rate": 2.336767382599199e-06, + "loss": 0.7424, + "step": 12196 + }, + { + "epoch": 2.22, + "learning_rate": 2.336712777575537e-06, + "loss": 0.0378, + "step": 12197 + }, + { + "epoch": 2.22, + "learning_rate": 2.336658172551875e-06, + "loss": 0.0504, + "step": 12198 + }, + { + "epoch": 2.22, + "learning_rate": 2.336603567528213e-06, + "loss": 1.692, + "step": 12199 + }, + { + "epoch": 2.22, + "learning_rate": 2.3365489625045508e-06, + "loss": 0.8354, + "step": 12200 + }, + { + "epoch": 2.22, + "learning_rate": 2.3364943574808883e-06, + "loss": 0.4795, + "step": 12201 + }, + { + "epoch": 2.22, + "learning_rate": 2.3364397524572262e-06, + "loss": 1.5712, + "step": 12202 + }, + { + "epoch": 2.22, + "learning_rate": 2.3363851474335638e-06, + "loss": 0.6029, + "step": 12203 + }, + { + "epoch": 2.22, + "learning_rate": 2.3363305424099017e-06, + "loss": 1.4927, + "step": 12204 + }, + { + "epoch": 2.22, + "learning_rate": 2.3362759373862397e-06, + "loss": 1.176, + "step": 12205 + }, + { + "epoch": 2.22, + "learning_rate": 2.336221332362577e-06, + "loss": 1.2269, + "step": 12206 + }, + { + "epoch": 2.22, + "learning_rate": 2.336166727338915e-06, + "loss": 0.3566, + "step": 12207 + }, + { + "epoch": 2.22, + "learning_rate": 2.336112122315253e-06, + "loss": 1.169, + "step": 12208 + }, + { + "epoch": 2.22, + "learning_rate": 2.3360575172915906e-06, + "loss": 0.0591, + "step": 12209 + }, + { + "epoch": 2.22, + "learning_rate": 2.3360029122679285e-06, + "loss": 0.6694, + "step": 12210 + }, + { + "epoch": 2.22, + "learning_rate": 2.3359483072442665e-06, + "loss": 1.0877, + "step": 12211 + }, + { + "epoch": 2.22, + "learning_rate": 2.3358937022206044e-06, + "loss": 0.3007, + "step": 12212 + }, + { + "epoch": 2.22, + "learning_rate": 2.3358390971969424e-06, + "loss": 0.8717, + "step": 12213 + }, + { + "epoch": 2.22, + "learning_rate": 2.3357844921732803e-06, + "loss": 0.5308, + "step": 12214 + }, + { + "epoch": 2.22, + "learning_rate": 2.335729887149618e-06, + "loss": 2.4747, + "step": 12215 + }, + { + "epoch": 2.22, + "learning_rate": 2.3356752821259558e-06, + "loss": 1.5776, + "step": 12216 + }, + { + "epoch": 2.22, + "learning_rate": 2.3356206771022933e-06, + "loss": 0.2198, + "step": 12217 + }, + { + "epoch": 2.22, + "learning_rate": 2.3355660720786312e-06, + "loss": 1.7218, + "step": 12218 + }, + { + "epoch": 2.22, + "learning_rate": 2.335511467054969e-06, + "loss": 0.0797, + "step": 12219 + }, + { + "epoch": 2.22, + "learning_rate": 2.3354568620313067e-06, + "loss": 0.5138, + "step": 12220 + }, + { + "epoch": 2.22, + "learning_rate": 2.3354022570076446e-06, + "loss": 1.8559, + "step": 12221 + }, + { + "epoch": 2.22, + "learning_rate": 2.3353476519839826e-06, + "loss": 0.8731, + "step": 12222 + }, + { + "epoch": 2.22, + "learning_rate": 2.3352930469603205e-06, + "loss": 1.351, + "step": 12223 + }, + { + "epoch": 2.22, + "learning_rate": 2.3352384419366585e-06, + "loss": 1.1129, + "step": 12224 + }, + { + "epoch": 2.22, + "learning_rate": 2.3351838369129964e-06, + "loss": 0.0386, + "step": 12225 + }, + { + "epoch": 2.22, + "learning_rate": 2.335129231889334e-06, + "loss": 2.5145, + "step": 12226 + }, + { + "epoch": 2.22, + "learning_rate": 2.335074626865672e-06, + "loss": 0.938, + "step": 12227 + }, + { + "epoch": 2.22, + "learning_rate": 2.33502002184201e-06, + "loss": 0.7092, + "step": 12228 + }, + { + "epoch": 2.22, + "learning_rate": 2.3349654168183473e-06, + "loss": 1.5669, + "step": 12229 + }, + { + "epoch": 2.22, + "learning_rate": 2.3349108117946853e-06, + "loss": 0.0257, + "step": 12230 + }, + { + "epoch": 2.22, + "learning_rate": 2.334856206771023e-06, + "loss": 0.4024, + "step": 12231 + }, + { + "epoch": 2.22, + "learning_rate": 2.3348016017473607e-06, + "loss": 0.4661, + "step": 12232 + }, + { + "epoch": 2.22, + "learning_rate": 2.3347469967236987e-06, + "loss": 1.3674, + "step": 12233 + }, + { + "epoch": 2.22, + "learning_rate": 2.334692391700036e-06, + "loss": 0.355, + "step": 12234 + }, + { + "epoch": 2.22, + "learning_rate": 2.334637786676374e-06, + "loss": 1.4078, + "step": 12235 + }, + { + "epoch": 2.23, + "learning_rate": 2.334583181652712e-06, + "loss": 0.0898, + "step": 12236 + }, + { + "epoch": 2.23, + "learning_rate": 2.33452857662905e-06, + "loss": 2.7375, + "step": 12237 + }, + { + "epoch": 2.23, + "learning_rate": 2.334473971605388e-06, + "loss": 1.0441, + "step": 12238 + }, + { + "epoch": 2.23, + "learning_rate": 2.334419366581726e-06, + "loss": 0.5812, + "step": 12239 + }, + { + "epoch": 2.23, + "learning_rate": 2.3343647615580634e-06, + "loss": 0.978, + "step": 12240 + }, + { + "epoch": 2.23, + "learning_rate": 2.3343101565344014e-06, + "loss": 2.2289, + "step": 12241 + }, + { + "epoch": 2.23, + "learning_rate": 2.334255551510739e-06, + "loss": 0.6817, + "step": 12242 + }, + { + "epoch": 2.23, + "learning_rate": 2.334200946487077e-06, + "loss": 1.8106, + "step": 12243 + }, + { + "epoch": 2.23, + "learning_rate": 2.3341463414634148e-06, + "loss": 0.3105, + "step": 12244 + }, + { + "epoch": 2.23, + "learning_rate": 2.3340917364397523e-06, + "loss": 0.0484, + "step": 12245 + }, + { + "epoch": 2.23, + "learning_rate": 2.3340371314160902e-06, + "loss": 0.0232, + "step": 12246 + }, + { + "epoch": 2.23, + "learning_rate": 2.333982526392428e-06, + "loss": 0.7394, + "step": 12247 + }, + { + "epoch": 2.23, + "learning_rate": 2.3339279213687657e-06, + "loss": 1.7759, + "step": 12248 + }, + { + "epoch": 2.23, + "learning_rate": 2.3338733163451037e-06, + "loss": 0.3179, + "step": 12249 + }, + { + "epoch": 2.23, + "learning_rate": 2.3338187113214416e-06, + "loss": 0.5976, + "step": 12250 + }, + { + "epoch": 2.23, + "learning_rate": 2.3337641062977795e-06, + "loss": 0.1262, + "step": 12251 + }, + { + "epoch": 2.23, + "learning_rate": 2.3337095012741175e-06, + "loss": 1.3209, + "step": 12252 + }, + { + "epoch": 2.23, + "learning_rate": 2.3336548962504554e-06, + "loss": 0.5754, + "step": 12253 + }, + { + "epoch": 2.23, + "learning_rate": 2.333600291226793e-06, + "loss": 0.9613, + "step": 12254 + }, + { + "epoch": 2.23, + "learning_rate": 2.333545686203131e-06, + "loss": 1.3523, + "step": 12255 + }, + { + "epoch": 2.23, + "learning_rate": 2.3334910811794684e-06, + "loss": 1.494, + "step": 12256 + }, + { + "epoch": 2.23, + "learning_rate": 2.3334364761558064e-06, + "loss": 1.0568, + "step": 12257 + }, + { + "epoch": 2.23, + "learning_rate": 2.3333818711321443e-06, + "loss": 4.3157, + "step": 12258 + }, + { + "epoch": 2.23, + "learning_rate": 2.333327266108482e-06, + "loss": 1.5378, + "step": 12259 + }, + { + "epoch": 2.23, + "learning_rate": 2.3332726610848198e-06, + "loss": 0.2698, + "step": 12260 + }, + { + "epoch": 2.23, + "learning_rate": 2.3332180560611577e-06, + "loss": 0.9448, + "step": 12261 + }, + { + "epoch": 2.23, + "learning_rate": 2.3331634510374952e-06, + "loss": 2.3771, + "step": 12262 + }, + { + "epoch": 2.23, + "learning_rate": 2.333108846013833e-06, + "loss": 1.2799, + "step": 12263 + }, + { + "epoch": 2.23, + "learning_rate": 2.333054240990171e-06, + "loss": 0.9414, + "step": 12264 + }, + { + "epoch": 2.23, + "learning_rate": 2.332999635966509e-06, + "loss": 1.2686, + "step": 12265 + }, + { + "epoch": 2.23, + "learning_rate": 2.332945030942847e-06, + "loss": 1.018, + "step": 12266 + }, + { + "epoch": 2.23, + "learning_rate": 2.332890425919185e-06, + "loss": 0.926, + "step": 12267 + }, + { + "epoch": 2.23, + "learning_rate": 2.3328358208955225e-06, + "loss": 1.4914, + "step": 12268 + }, + { + "epoch": 2.23, + "learning_rate": 2.3327812158718604e-06, + "loss": 0.9762, + "step": 12269 + }, + { + "epoch": 2.23, + "learning_rate": 2.332726610848198e-06, + "loss": 1.8083, + "step": 12270 + }, + { + "epoch": 2.23, + "learning_rate": 2.332672005824536e-06, + "loss": 0.0665, + "step": 12271 + }, + { + "epoch": 2.23, + "learning_rate": 2.332617400800874e-06, + "loss": 0.4082, + "step": 12272 + }, + { + "epoch": 2.23, + "learning_rate": 2.3325627957772113e-06, + "loss": 0.3714, + "step": 12273 + }, + { + "epoch": 2.23, + "learning_rate": 2.3325081907535493e-06, + "loss": 1.6824, + "step": 12274 + }, + { + "epoch": 2.23, + "learning_rate": 2.3324535857298872e-06, + "loss": 0.3602, + "step": 12275 + }, + { + "epoch": 2.23, + "learning_rate": 2.3323989807062247e-06, + "loss": 0.5116, + "step": 12276 + }, + { + "epoch": 2.23, + "learning_rate": 2.3323443756825627e-06, + "loss": 0.8001, + "step": 12277 + }, + { + "epoch": 2.23, + "learning_rate": 2.3322897706589006e-06, + "loss": 0.0199, + "step": 12278 + }, + { + "epoch": 2.23, + "learning_rate": 2.3322351656352386e-06, + "loss": 1.0237, + "step": 12279 + }, + { + "epoch": 2.23, + "learning_rate": 2.3321805606115765e-06, + "loss": 0.9665, + "step": 12280 + }, + { + "epoch": 2.23, + "learning_rate": 2.3321259555879145e-06, + "loss": 1.5766, + "step": 12281 + }, + { + "epoch": 2.23, + "learning_rate": 2.332071350564252e-06, + "loss": 0.6043, + "step": 12282 + }, + { + "epoch": 2.23, + "learning_rate": 2.33201674554059e-06, + "loss": 0.4811, + "step": 12283 + }, + { + "epoch": 2.23, + "learning_rate": 2.3319621405169274e-06, + "loss": 0.122, + "step": 12284 + }, + { + "epoch": 2.23, + "learning_rate": 2.3319075354932654e-06, + "loss": 2.0921, + "step": 12285 + }, + { + "epoch": 2.23, + "learning_rate": 2.3318529304696033e-06, + "loss": 1.3833, + "step": 12286 + }, + { + "epoch": 2.23, + "learning_rate": 2.331798325445941e-06, + "loss": 0.0939, + "step": 12287 + }, + { + "epoch": 2.23, + "learning_rate": 2.3317437204222788e-06, + "loss": 0.5145, + "step": 12288 + }, + { + "epoch": 2.23, + "learning_rate": 2.3316891153986167e-06, + "loss": 1.0107, + "step": 12289 + }, + { + "epoch": 2.23, + "learning_rate": 2.3316345103749543e-06, + "loss": 1.9338, + "step": 12290 + }, + { + "epoch": 2.24, + "learning_rate": 2.331579905351292e-06, + "loss": 2.3965, + "step": 12291 + }, + { + "epoch": 2.24, + "learning_rate": 2.33152530032763e-06, + "loss": 0.4437, + "step": 12292 + }, + { + "epoch": 2.24, + "learning_rate": 2.331470695303968e-06, + "loss": 0.794, + "step": 12293 + }, + { + "epoch": 2.24, + "learning_rate": 2.331416090280306e-06, + "loss": 0.5124, + "step": 12294 + }, + { + "epoch": 2.24, + "learning_rate": 2.331361485256644e-06, + "loss": 1.4169, + "step": 12295 + }, + { + "epoch": 2.24, + "learning_rate": 2.3313068802329815e-06, + "loss": 0.4153, + "step": 12296 + }, + { + "epoch": 2.24, + "learning_rate": 2.3312522752093194e-06, + "loss": 0.0145, + "step": 12297 + }, + { + "epoch": 2.24, + "learning_rate": 2.331197670185657e-06, + "loss": 0.4329, + "step": 12298 + }, + { + "epoch": 2.24, + "learning_rate": 2.331143065161995e-06, + "loss": 0.3479, + "step": 12299 + }, + { + "epoch": 2.24, + "learning_rate": 2.331088460138333e-06, + "loss": 0.7358, + "step": 12300 + }, + { + "epoch": 2.24, + "learning_rate": 2.3310338551146704e-06, + "loss": 0.0267, + "step": 12301 + }, + { + "epoch": 2.24, + "learning_rate": 2.3309792500910083e-06, + "loss": 1.2657, + "step": 12302 + }, + { + "epoch": 2.24, + "learning_rate": 2.3309246450673462e-06, + "loss": 1.9521, + "step": 12303 + }, + { + "epoch": 2.24, + "learning_rate": 2.330870040043684e-06, + "loss": 0.4639, + "step": 12304 + }, + { + "epoch": 2.24, + "learning_rate": 2.330815435020022e-06, + "loss": 0.0891, + "step": 12305 + }, + { + "epoch": 2.24, + "learning_rate": 2.33076082999636e-06, + "loss": 0.6239, + "step": 12306 + }, + { + "epoch": 2.24, + "learning_rate": 2.3307062249726976e-06, + "loss": 0.4143, + "step": 12307 + }, + { + "epoch": 2.24, + "learning_rate": 2.3306516199490355e-06, + "loss": 0.8512, + "step": 12308 + }, + { + "epoch": 2.24, + "learning_rate": 2.3305970149253735e-06, + "loss": 0.7839, + "step": 12309 + }, + { + "epoch": 2.24, + "learning_rate": 2.330542409901711e-06, + "loss": 1.4732, + "step": 12310 + }, + { + "epoch": 2.24, + "learning_rate": 2.330487804878049e-06, + "loss": 0.095, + "step": 12311 + }, + { + "epoch": 2.24, + "learning_rate": 2.3304331998543865e-06, + "loss": 0.1373, + "step": 12312 + }, + { + "epoch": 2.24, + "learning_rate": 2.3303785948307244e-06, + "loss": 1.2392, + "step": 12313 + }, + { + "epoch": 2.24, + "learning_rate": 2.3303239898070624e-06, + "loss": 0.0355, + "step": 12314 + }, + { + "epoch": 2.24, + "learning_rate": 2.3302693847834e-06, + "loss": 0.4453, + "step": 12315 + }, + { + "epoch": 2.24, + "learning_rate": 2.330214779759738e-06, + "loss": 1.0344, + "step": 12316 + }, + { + "epoch": 2.24, + "learning_rate": 2.3301601747360758e-06, + "loss": 0.8508, + "step": 12317 + }, + { + "epoch": 2.24, + "learning_rate": 2.3301055697124137e-06, + "loss": 0.1492, + "step": 12318 + }, + { + "epoch": 2.24, + "learning_rate": 2.3300509646887516e-06, + "loss": 0.3899, + "step": 12319 + }, + { + "epoch": 2.24, + "learning_rate": 2.3299963596650896e-06, + "loss": 0.9906, + "step": 12320 + }, + { + "epoch": 2.24, + "learning_rate": 2.329941754641427e-06, + "loss": 0.8782, + "step": 12321 + }, + { + "epoch": 2.24, + "learning_rate": 2.329887149617765e-06, + "loss": 1.0377, + "step": 12322 + }, + { + "epoch": 2.24, + "learning_rate": 2.3298325445941026e-06, + "loss": 0.4474, + "step": 12323 + }, + { + "epoch": 2.24, + "learning_rate": 2.3297779395704405e-06, + "loss": 1.7881, + "step": 12324 + }, + { + "epoch": 2.24, + "learning_rate": 2.3297233345467785e-06, + "loss": 0.7353, + "step": 12325 + }, + { + "epoch": 2.24, + "learning_rate": 2.329668729523116e-06, + "loss": 0.418, + "step": 12326 + }, + { + "epoch": 2.24, + "learning_rate": 2.329614124499454e-06, + "loss": 0.464, + "step": 12327 + }, + { + "epoch": 2.24, + "learning_rate": 2.329559519475792e-06, + "loss": 0.9689, + "step": 12328 + }, + { + "epoch": 2.24, + "learning_rate": 2.3295049144521294e-06, + "loss": 0.3627, + "step": 12329 + }, + { + "epoch": 2.24, + "learning_rate": 2.3294503094284673e-06, + "loss": 1.6836, + "step": 12330 + }, + { + "epoch": 2.24, + "learning_rate": 2.3293957044048053e-06, + "loss": 0.6451, + "step": 12331 + }, + { + "epoch": 2.24, + "learning_rate": 2.3293410993811432e-06, + "loss": 0.3815, + "step": 12332 + }, + { + "epoch": 2.24, + "learning_rate": 2.329286494357481e-06, + "loss": 1.728, + "step": 12333 + }, + { + "epoch": 2.24, + "learning_rate": 2.329231889333819e-06, + "loss": 0.3452, + "step": 12334 + }, + { + "epoch": 2.24, + "learning_rate": 2.3291772843101566e-06, + "loss": 1.6163, + "step": 12335 + }, + { + "epoch": 2.24, + "learning_rate": 2.3291226792864946e-06, + "loss": 0.0277, + "step": 12336 + }, + { + "epoch": 2.24, + "learning_rate": 2.329068074262832e-06, + "loss": 0.3981, + "step": 12337 + }, + { + "epoch": 2.24, + "learning_rate": 2.32901346923917e-06, + "loss": 1.8638, + "step": 12338 + }, + { + "epoch": 2.24, + "learning_rate": 2.328958864215508e-06, + "loss": 1.4509, + "step": 12339 + }, + { + "epoch": 2.24, + "learning_rate": 2.3289042591918455e-06, + "loss": 0.6567, + "step": 12340 + }, + { + "epoch": 2.24, + "learning_rate": 2.3288496541681834e-06, + "loss": 0.4643, + "step": 12341 + }, + { + "epoch": 2.24, + "learning_rate": 2.3287950491445214e-06, + "loss": 0.8139, + "step": 12342 + }, + { + "epoch": 2.24, + "learning_rate": 2.328740444120859e-06, + "loss": 3.3795, + "step": 12343 + }, + { + "epoch": 2.24, + "learning_rate": 2.328685839097197e-06, + "loss": 0.8196, + "step": 12344 + }, + { + "epoch": 2.24, + "learning_rate": 2.3286312340735348e-06, + "loss": 0.9919, + "step": 12345 + }, + { + "epoch": 2.25, + "learning_rate": 2.3285766290498727e-06, + "loss": 1.1443, + "step": 12346 + }, + { + "epoch": 2.25, + "learning_rate": 2.3285220240262107e-06, + "loss": 0.6871, + "step": 12347 + }, + { + "epoch": 2.25, + "learning_rate": 2.3284674190025486e-06, + "loss": 1.4061, + "step": 12348 + }, + { + "epoch": 2.25, + "learning_rate": 2.328412813978886e-06, + "loss": 1.9405, + "step": 12349 + }, + { + "epoch": 2.25, + "learning_rate": 2.328358208955224e-06, + "loss": 3.1109, + "step": 12350 + }, + { + "epoch": 2.25, + "learning_rate": 2.3283036039315616e-06, + "loss": 0.426, + "step": 12351 + }, + { + "epoch": 2.25, + "learning_rate": 2.3282489989078995e-06, + "loss": 0.0898, + "step": 12352 + }, + { + "epoch": 2.25, + "learning_rate": 2.3281943938842375e-06, + "loss": 0.1746, + "step": 12353 + }, + { + "epoch": 2.25, + "learning_rate": 2.328139788860575e-06, + "loss": 0.8246, + "step": 12354 + }, + { + "epoch": 2.25, + "learning_rate": 2.328085183836913e-06, + "loss": 0.4931, + "step": 12355 + }, + { + "epoch": 2.25, + "learning_rate": 2.328030578813251e-06, + "loss": 0.1369, + "step": 12356 + }, + { + "epoch": 2.25, + "learning_rate": 2.3279759737895884e-06, + "loss": 0.1863, + "step": 12357 + }, + { + "epoch": 2.25, + "learning_rate": 2.3279213687659264e-06, + "loss": 0.4887, + "step": 12358 + }, + { + "epoch": 2.25, + "learning_rate": 2.3278667637422643e-06, + "loss": 0.6674, + "step": 12359 + }, + { + "epoch": 2.25, + "learning_rate": 2.3278121587186022e-06, + "loss": 1.3657, + "step": 12360 + }, + { + "epoch": 2.25, + "learning_rate": 2.32775755369494e-06, + "loss": 1.5058, + "step": 12361 + }, + { + "epoch": 2.25, + "learning_rate": 2.327702948671278e-06, + "loss": 1.7286, + "step": 12362 + }, + { + "epoch": 2.25, + "learning_rate": 2.3276483436476156e-06, + "loss": 2.7707, + "step": 12363 + }, + { + "epoch": 2.25, + "learning_rate": 2.3275937386239536e-06, + "loss": 1.1072, + "step": 12364 + }, + { + "epoch": 2.25, + "learning_rate": 2.327539133600291e-06, + "loss": 0.4035, + "step": 12365 + }, + { + "epoch": 2.25, + "learning_rate": 2.327484528576629e-06, + "loss": 3.1328, + "step": 12366 + }, + { + "epoch": 2.25, + "learning_rate": 2.327429923552967e-06, + "loss": 0.5006, + "step": 12367 + }, + { + "epoch": 2.25, + "learning_rate": 2.3273753185293045e-06, + "loss": 0.0618, + "step": 12368 + }, + { + "epoch": 2.25, + "learning_rate": 2.3273207135056425e-06, + "loss": 0.0191, + "step": 12369 + }, + { + "epoch": 2.25, + "learning_rate": 2.3272661084819804e-06, + "loss": 0.2593, + "step": 12370 + }, + { + "epoch": 2.25, + "learning_rate": 2.327211503458318e-06, + "loss": 0.0233, + "step": 12371 + }, + { + "epoch": 2.25, + "learning_rate": 2.327156898434656e-06, + "loss": 0.5782, + "step": 12372 + }, + { + "epoch": 2.25, + "learning_rate": 2.327102293410994e-06, + "loss": 1.6443, + "step": 12373 + }, + { + "epoch": 2.25, + "learning_rate": 2.3270476883873318e-06, + "loss": 1.7003, + "step": 12374 + }, + { + "epoch": 2.25, + "learning_rate": 2.3269930833636697e-06, + "loss": 1.2725, + "step": 12375 + }, + { + "epoch": 2.25, + "learning_rate": 2.3269384783400076e-06, + "loss": 2.7715, + "step": 12376 + }, + { + "epoch": 2.25, + "learning_rate": 2.326883873316345e-06, + "loss": 1.4379, + "step": 12377 + }, + { + "epoch": 2.25, + "learning_rate": 2.326829268292683e-06, + "loss": 0.2986, + "step": 12378 + }, + { + "epoch": 2.25, + "learning_rate": 2.3267746632690206e-06, + "loss": 1.5569, + "step": 12379 + }, + { + "epoch": 2.25, + "learning_rate": 2.3267200582453586e-06, + "loss": 0.6139, + "step": 12380 + }, + { + "epoch": 2.25, + "learning_rate": 2.3266654532216965e-06, + "loss": 2.5612, + "step": 12381 + }, + { + "epoch": 2.25, + "learning_rate": 2.326610848198034e-06, + "loss": 0.0919, + "step": 12382 + }, + { + "epoch": 2.25, + "learning_rate": 2.326556243174372e-06, + "loss": 1.2578, + "step": 12383 + }, + { + "epoch": 2.25, + "learning_rate": 2.32650163815071e-06, + "loss": 0.9347, + "step": 12384 + }, + { + "epoch": 2.25, + "learning_rate": 2.326447033127048e-06, + "loss": 0.7547, + "step": 12385 + }, + { + "epoch": 2.25, + "learning_rate": 2.3263924281033854e-06, + "loss": 0.7223, + "step": 12386 + }, + { + "epoch": 2.25, + "learning_rate": 2.3263378230797233e-06, + "loss": 0.2612, + "step": 12387 + }, + { + "epoch": 2.25, + "learning_rate": 2.3262832180560613e-06, + "loss": 0.3551, + "step": 12388 + }, + { + "epoch": 2.25, + "learning_rate": 2.3262286130323992e-06, + "loss": 0.0554, + "step": 12389 + }, + { + "epoch": 2.25, + "learning_rate": 2.3261740080087367e-06, + "loss": 1.8483, + "step": 12390 + }, + { + "epoch": 2.25, + "learning_rate": 2.3261194029850747e-06, + "loss": 0.9317, + "step": 12391 + }, + { + "epoch": 2.25, + "learning_rate": 2.3260647979614126e-06, + "loss": 0.6965, + "step": 12392 + }, + { + "epoch": 2.25, + "learning_rate": 2.32601019293775e-06, + "loss": 0.1417, + "step": 12393 + }, + { + "epoch": 2.25, + "learning_rate": 2.325955587914088e-06, + "loss": 2.5128, + "step": 12394 + }, + { + "epoch": 2.25, + "learning_rate": 2.325900982890426e-06, + "loss": 0.133, + "step": 12395 + }, + { + "epoch": 2.25, + "learning_rate": 2.3258463778667635e-06, + "loss": 0.1356, + "step": 12396 + }, + { + "epoch": 2.25, + "learning_rate": 2.3257917728431015e-06, + "loss": 0.4277, + "step": 12397 + }, + { + "epoch": 2.25, + "learning_rate": 2.3257371678194394e-06, + "loss": 1.5461, + "step": 12398 + }, + { + "epoch": 2.25, + "learning_rate": 2.3256825627957774e-06, + "loss": 0.068, + "step": 12399 + }, + { + "epoch": 2.25, + "learning_rate": 2.3256279577721153e-06, + "loss": 0.9492, + "step": 12400 + }, + { + "epoch": 2.26, + "learning_rate": 2.3255733527484533e-06, + "loss": 1.2105, + "step": 12401 + }, + { + "epoch": 2.26, + "learning_rate": 2.3255187477247908e-06, + "loss": 1.3055, + "step": 12402 + }, + { + "epoch": 2.26, + "learning_rate": 2.3254641427011287e-06, + "loss": 3.5704, + "step": 12403 + }, + { + "epoch": 2.26, + "learning_rate": 2.3254095376774662e-06, + "loss": 0.1398, + "step": 12404 + }, + { + "epoch": 2.26, + "learning_rate": 2.325354932653804e-06, + "loss": 0.3647, + "step": 12405 + }, + { + "epoch": 2.26, + "learning_rate": 2.325300327630142e-06, + "loss": 0.4374, + "step": 12406 + }, + { + "epoch": 2.26, + "learning_rate": 2.3252457226064797e-06, + "loss": 0.4812, + "step": 12407 + }, + { + "epoch": 2.26, + "learning_rate": 2.3251911175828176e-06, + "loss": 1.9925, + "step": 12408 + }, + { + "epoch": 2.26, + "learning_rate": 2.3251365125591555e-06, + "loss": 0.3866, + "step": 12409 + }, + { + "epoch": 2.26, + "learning_rate": 2.325081907535493e-06, + "loss": 1.2719, + "step": 12410 + }, + { + "epoch": 2.26, + "learning_rate": 2.325027302511831e-06, + "loss": 0.3049, + "step": 12411 + }, + { + "epoch": 2.26, + "learning_rate": 2.324972697488169e-06, + "loss": 0.2362, + "step": 12412 + }, + { + "epoch": 2.26, + "learning_rate": 2.324918092464507e-06, + "loss": 0.0923, + "step": 12413 + }, + { + "epoch": 2.26, + "learning_rate": 2.324863487440845e-06, + "loss": 0.484, + "step": 12414 + }, + { + "epoch": 2.26, + "learning_rate": 2.3248088824171828e-06, + "loss": 0.276, + "step": 12415 + }, + { + "epoch": 2.26, + "learning_rate": 2.3247542773935203e-06, + "loss": 0.0651, + "step": 12416 + }, + { + "epoch": 2.26, + "learning_rate": 2.3246996723698582e-06, + "loss": 0.0157, + "step": 12417 + }, + { + "epoch": 2.26, + "learning_rate": 2.3246450673461958e-06, + "loss": 0.4345, + "step": 12418 + }, + { + "epoch": 2.26, + "learning_rate": 2.3245904623225337e-06, + "loss": 0.4884, + "step": 12419 + }, + { + "epoch": 2.26, + "learning_rate": 2.3245358572988716e-06, + "loss": 0.7513, + "step": 12420 + }, + { + "epoch": 2.26, + "learning_rate": 2.324481252275209e-06, + "loss": 0.7662, + "step": 12421 + }, + { + "epoch": 2.26, + "learning_rate": 2.324426647251547e-06, + "loss": 2.6513, + "step": 12422 + }, + { + "epoch": 2.26, + "learning_rate": 2.324372042227885e-06, + "loss": 0.0352, + "step": 12423 + }, + { + "epoch": 2.26, + "learning_rate": 2.3243174372042226e-06, + "loss": 0.6756, + "step": 12424 + }, + { + "epoch": 2.26, + "learning_rate": 2.3242628321805605e-06, + "loss": 3.0681, + "step": 12425 + }, + { + "epoch": 2.26, + "learning_rate": 2.3242082271568985e-06, + "loss": 2.2323, + "step": 12426 + }, + { + "epoch": 2.26, + "learning_rate": 2.3241536221332364e-06, + "loss": 0.2868, + "step": 12427 + }, + { + "epoch": 2.26, + "learning_rate": 2.3240990171095743e-06, + "loss": 0.941, + "step": 12428 + }, + { + "epoch": 2.26, + "learning_rate": 2.3240444120859123e-06, + "loss": 0.0808, + "step": 12429 + }, + { + "epoch": 2.26, + "learning_rate": 2.32398980706225e-06, + "loss": 1.4737, + "step": 12430 + }, + { + "epoch": 2.26, + "learning_rate": 2.3239352020385878e-06, + "loss": 0.3296, + "step": 12431 + }, + { + "epoch": 2.26, + "learning_rate": 2.3238805970149253e-06, + "loss": 2.027, + "step": 12432 + }, + { + "epoch": 2.26, + "learning_rate": 2.3238259919912632e-06, + "loss": 0.0117, + "step": 12433 + }, + { + "epoch": 2.26, + "learning_rate": 2.323771386967601e-06, + "loss": 0.6323, + "step": 12434 + }, + { + "epoch": 2.26, + "learning_rate": 2.3237167819439387e-06, + "loss": 0.5647, + "step": 12435 + }, + { + "epoch": 2.26, + "learning_rate": 2.3236621769202766e-06, + "loss": 0.127, + "step": 12436 + }, + { + "epoch": 2.26, + "learning_rate": 2.3236075718966146e-06, + "loss": 0.4321, + "step": 12437 + }, + { + "epoch": 2.26, + "learning_rate": 2.323552966872952e-06, + "loss": 0.1084, + "step": 12438 + }, + { + "epoch": 2.26, + "learning_rate": 2.32349836184929e-06, + "loss": 1.0582, + "step": 12439 + }, + { + "epoch": 2.26, + "learning_rate": 2.323443756825628e-06, + "loss": 0.4584, + "step": 12440 + }, + { + "epoch": 2.26, + "learning_rate": 2.323389151801966e-06, + "loss": 0.2826, + "step": 12441 + }, + { + "epoch": 2.26, + "learning_rate": 2.323334546778304e-06, + "loss": 1.3874, + "step": 12442 + }, + { + "epoch": 2.26, + "learning_rate": 2.323279941754642e-06, + "loss": 0.4242, + "step": 12443 + }, + { + "epoch": 2.26, + "learning_rate": 2.3232253367309793e-06, + "loss": 0.2307, + "step": 12444 + }, + { + "epoch": 2.26, + "learning_rate": 2.3231707317073173e-06, + "loss": 1.1224, + "step": 12445 + }, + { + "epoch": 2.26, + "learning_rate": 2.3231161266836548e-06, + "loss": 0.4515, + "step": 12446 + }, + { + "epoch": 2.26, + "learning_rate": 2.3230615216599927e-06, + "loss": 1.9602, + "step": 12447 + }, + { + "epoch": 2.26, + "learning_rate": 2.3230069166363307e-06, + "loss": 1.4356, + "step": 12448 + }, + { + "epoch": 2.26, + "learning_rate": 2.322952311612668e-06, + "loss": 0.4007, + "step": 12449 + }, + { + "epoch": 2.26, + "learning_rate": 2.322897706589006e-06, + "loss": 0.6737, + "step": 12450 + }, + { + "epoch": 2.26, + "learning_rate": 2.322843101565344e-06, + "loss": 0.6239, + "step": 12451 + }, + { + "epoch": 2.26, + "learning_rate": 2.3227884965416816e-06, + "loss": 0.2308, + "step": 12452 + }, + { + "epoch": 2.26, + "learning_rate": 2.3227338915180195e-06, + "loss": 0.6979, + "step": 12453 + }, + { + "epoch": 2.26, + "learning_rate": 2.3226792864943575e-06, + "loss": 1.1574, + "step": 12454 + }, + { + "epoch": 2.26, + "learning_rate": 2.3226246814706954e-06, + "loss": 0.8533, + "step": 12455 + }, + { + "epoch": 2.27, + "learning_rate": 2.3225700764470334e-06, + "loss": 0.427, + "step": 12456 + }, + { + "epoch": 2.27, + "learning_rate": 2.3225154714233713e-06, + "loss": 0.0466, + "step": 12457 + }, + { + "epoch": 2.27, + "learning_rate": 2.322460866399709e-06, + "loss": 1.2925, + "step": 12458 + }, + { + "epoch": 2.27, + "learning_rate": 2.3224062613760468e-06, + "loss": 0.3873, + "step": 12459 + }, + { + "epoch": 2.27, + "learning_rate": 2.3223516563523843e-06, + "loss": 0.3482, + "step": 12460 + }, + { + "epoch": 2.27, + "learning_rate": 2.3222970513287222e-06, + "loss": 1.0088, + "step": 12461 + }, + { + "epoch": 2.27, + "learning_rate": 2.32224244630506e-06, + "loss": 0.5617, + "step": 12462 + }, + { + "epoch": 2.27, + "learning_rate": 2.3221878412813977e-06, + "loss": 0.0354, + "step": 12463 + }, + { + "epoch": 2.27, + "learning_rate": 2.3221332362577356e-06, + "loss": 0.0162, + "step": 12464 + }, + { + "epoch": 2.27, + "learning_rate": 2.3220786312340736e-06, + "loss": 1.0185, + "step": 12465 + }, + { + "epoch": 2.27, + "learning_rate": 2.322024026210411e-06, + "loss": 3.3791, + "step": 12466 + }, + { + "epoch": 2.27, + "learning_rate": 2.321969421186749e-06, + "loss": 0.3538, + "step": 12467 + }, + { + "epoch": 2.27, + "learning_rate": 2.321914816163087e-06, + "loss": 0.0652, + "step": 12468 + }, + { + "epoch": 2.27, + "learning_rate": 2.321860211139425e-06, + "loss": 0.2606, + "step": 12469 + }, + { + "epoch": 2.27, + "learning_rate": 2.321805606115763e-06, + "loss": 0.1405, + "step": 12470 + }, + { + "epoch": 2.27, + "learning_rate": 2.3217510010921004e-06, + "loss": 0.0924, + "step": 12471 + }, + { + "epoch": 2.27, + "learning_rate": 2.3216963960684383e-06, + "loss": 1.5367, + "step": 12472 + }, + { + "epoch": 2.27, + "learning_rate": 2.3216417910447763e-06, + "loss": 2.0026, + "step": 12473 + }, + { + "epoch": 2.27, + "learning_rate": 2.321587186021114e-06, + "loss": 0.2927, + "step": 12474 + }, + { + "epoch": 2.27, + "learning_rate": 2.3215325809974518e-06, + "loss": 0.2934, + "step": 12475 + }, + { + "epoch": 2.27, + "learning_rate": 2.3214779759737897e-06, + "loss": 0.9343, + "step": 12476 + }, + { + "epoch": 2.27, + "learning_rate": 2.3214233709501272e-06, + "loss": 0.2603, + "step": 12477 + }, + { + "epoch": 2.27, + "learning_rate": 2.321368765926465e-06, + "loss": 0.2843, + "step": 12478 + }, + { + "epoch": 2.27, + "learning_rate": 2.321314160902803e-06, + "loss": 1.1186, + "step": 12479 + }, + { + "epoch": 2.27, + "learning_rate": 2.321259555879141e-06, + "loss": 0.9412, + "step": 12480 + }, + { + "epoch": 2.27, + "learning_rate": 2.321204950855479e-06, + "loss": 1.9847, + "step": 12481 + }, + { + "epoch": 2.27, + "learning_rate": 2.321150345831817e-06, + "loss": 2.6362, + "step": 12482 + }, + { + "epoch": 2.27, + "learning_rate": 2.3210957408081545e-06, + "loss": 0.881, + "step": 12483 + }, + { + "epoch": 2.27, + "learning_rate": 2.3210411357844924e-06, + "loss": 1.1988, + "step": 12484 + }, + { + "epoch": 2.27, + "learning_rate": 2.32098653076083e-06, + "loss": 0.3328, + "step": 12485 + }, + { + "epoch": 2.27, + "learning_rate": 2.320931925737168e-06, + "loss": 0.1122, + "step": 12486 + }, + { + "epoch": 2.27, + "learning_rate": 2.320877320713506e-06, + "loss": 0.3935, + "step": 12487 + }, + { + "epoch": 2.27, + "learning_rate": 2.3208227156898433e-06, + "loss": 0.6807, + "step": 12488 + }, + { + "epoch": 2.27, + "learning_rate": 2.3207681106661813e-06, + "loss": 1.0798, + "step": 12489 + }, + { + "epoch": 2.27, + "learning_rate": 2.3207135056425192e-06, + "loss": 0.3779, + "step": 12490 + }, + { + "epoch": 2.27, + "learning_rate": 2.3206589006188567e-06, + "loss": 0.0404, + "step": 12491 + }, + { + "epoch": 2.27, + "learning_rate": 2.3206042955951947e-06, + "loss": 0.6876, + "step": 12492 + }, + { + "epoch": 2.27, + "learning_rate": 2.3205496905715326e-06, + "loss": 0.6408, + "step": 12493 + }, + { + "epoch": 2.27, + "learning_rate": 2.3204950855478706e-06, + "loss": 0.7955, + "step": 12494 + }, + { + "epoch": 2.27, + "learning_rate": 2.3204404805242085e-06, + "loss": 0.2502, + "step": 12495 + }, + { + "epoch": 2.27, + "learning_rate": 2.3203858755005465e-06, + "loss": 1.3902, + "step": 12496 + }, + { + "epoch": 2.27, + "learning_rate": 2.320331270476884e-06, + "loss": 0.3859, + "step": 12497 + }, + { + "epoch": 2.27, + "learning_rate": 2.320276665453222e-06, + "loss": 3.3968, + "step": 12498 + }, + { + "epoch": 2.27, + "learning_rate": 2.3202220604295594e-06, + "loss": 1.6184, + "step": 12499 + }, + { + "epoch": 2.27, + "learning_rate": 2.3201674554058974e-06, + "loss": 0.9597, + "step": 12500 + }, + { + "epoch": 2.27, + "learning_rate": 2.3201128503822353e-06, + "loss": 2.3208, + "step": 12501 + }, + { + "epoch": 2.27, + "learning_rate": 2.320058245358573e-06, + "loss": 0.4161, + "step": 12502 + }, + { + "epoch": 2.27, + "learning_rate": 2.3200036403349108e-06, + "loss": 0.6032, + "step": 12503 + }, + { + "epoch": 2.27, + "learning_rate": 2.3199490353112487e-06, + "loss": 1.0846, + "step": 12504 + }, + { + "epoch": 2.27, + "learning_rate": 2.3198944302875862e-06, + "loss": 0.0242, + "step": 12505 + }, + { + "epoch": 2.27, + "learning_rate": 2.319839825263924e-06, + "loss": 0.0343, + "step": 12506 + }, + { + "epoch": 2.27, + "learning_rate": 2.319785220240262e-06, + "loss": 0.4828, + "step": 12507 + }, + { + "epoch": 2.27, + "learning_rate": 2.3197306152166e-06, + "loss": 1.377, + "step": 12508 + }, + { + "epoch": 2.27, + "learning_rate": 2.319676010192938e-06, + "loss": 0.9016, + "step": 12509 + }, + { + "epoch": 2.27, + "learning_rate": 2.319621405169276e-06, + "loss": 0.5377, + "step": 12510 + }, + { + "epoch": 2.28, + "learning_rate": 2.3195668001456135e-06, + "loss": 0.8624, + "step": 12511 + }, + { + "epoch": 2.28, + "learning_rate": 2.3195121951219514e-06, + "loss": 0.0357, + "step": 12512 + }, + { + "epoch": 2.28, + "learning_rate": 2.319457590098289e-06, + "loss": 1.5313, + "step": 12513 + }, + { + "epoch": 2.28, + "learning_rate": 2.319402985074627e-06, + "loss": 1.4035, + "step": 12514 + }, + { + "epoch": 2.28, + "learning_rate": 2.319348380050965e-06, + "loss": 0.2013, + "step": 12515 + }, + { + "epoch": 2.28, + "learning_rate": 2.3192937750273024e-06, + "loss": 0.8386, + "step": 12516 + }, + { + "epoch": 2.28, + "learning_rate": 2.3192391700036403e-06, + "loss": 0.7781, + "step": 12517 + }, + { + "epoch": 2.28, + "learning_rate": 2.3191845649799782e-06, + "loss": 2.5039, + "step": 12518 + }, + { + "epoch": 2.28, + "learning_rate": 2.3191299599563158e-06, + "loss": 0.7298, + "step": 12519 + }, + { + "epoch": 2.28, + "learning_rate": 2.3190753549326537e-06, + "loss": 0.287, + "step": 12520 + }, + { + "epoch": 2.28, + "learning_rate": 2.3190207499089916e-06, + "loss": 2.4173, + "step": 12521 + }, + { + "epoch": 2.28, + "learning_rate": 2.3189661448853296e-06, + "loss": 0.143, + "step": 12522 + }, + { + "epoch": 2.28, + "learning_rate": 2.3189115398616675e-06, + "loss": 0.5087, + "step": 12523 + }, + { + "epoch": 2.28, + "learning_rate": 2.3188569348380055e-06, + "loss": 1.2046, + "step": 12524 + }, + { + "epoch": 2.28, + "learning_rate": 2.318802329814343e-06, + "loss": 2.4158, + "step": 12525 + }, + { + "epoch": 2.28, + "learning_rate": 2.318747724790681e-06, + "loss": 3.5674, + "step": 12526 + }, + { + "epoch": 2.28, + "learning_rate": 2.3186931197670185e-06, + "loss": 0.5569, + "step": 12527 + }, + { + "epoch": 2.28, + "learning_rate": 2.3186385147433564e-06, + "loss": 1.9561, + "step": 12528 + }, + { + "epoch": 2.28, + "learning_rate": 2.3185839097196943e-06, + "loss": 1.3257, + "step": 12529 + }, + { + "epoch": 2.28, + "learning_rate": 2.318529304696032e-06, + "loss": 0.296, + "step": 12530 + }, + { + "epoch": 2.28, + "learning_rate": 2.31847469967237e-06, + "loss": 0.3921, + "step": 12531 + }, + { + "epoch": 2.28, + "learning_rate": 2.3184200946487078e-06, + "loss": 1.0793, + "step": 12532 + }, + { + "epoch": 2.28, + "learning_rate": 2.3183654896250453e-06, + "loss": 0.0428, + "step": 12533 + }, + { + "epoch": 2.28, + "learning_rate": 2.3183108846013832e-06, + "loss": 1.3651, + "step": 12534 + }, + { + "epoch": 2.28, + "learning_rate": 2.318256279577721e-06, + "loss": 0.1808, + "step": 12535 + }, + { + "epoch": 2.28, + "learning_rate": 2.318201674554059e-06, + "loss": 0.7451, + "step": 12536 + }, + { + "epoch": 2.28, + "learning_rate": 2.318147069530397e-06, + "loss": 0.4359, + "step": 12537 + }, + { + "epoch": 2.28, + "learning_rate": 2.318092464506735e-06, + "loss": 1.1144, + "step": 12538 + }, + { + "epoch": 2.28, + "learning_rate": 2.3180378594830725e-06, + "loss": 0.4856, + "step": 12539 + }, + { + "epoch": 2.28, + "learning_rate": 2.3179832544594105e-06, + "loss": 0.2601, + "step": 12540 + }, + { + "epoch": 2.28, + "learning_rate": 2.317928649435748e-06, + "loss": 0.3988, + "step": 12541 + }, + { + "epoch": 2.28, + "learning_rate": 2.317874044412086e-06, + "loss": 0.786, + "step": 12542 + }, + { + "epoch": 2.28, + "learning_rate": 2.317819439388424e-06, + "loss": 0.8273, + "step": 12543 + }, + { + "epoch": 2.28, + "learning_rate": 2.3177648343647614e-06, + "loss": 0.3421, + "step": 12544 + }, + { + "epoch": 2.28, + "learning_rate": 2.3177102293410993e-06, + "loss": 0.296, + "step": 12545 + }, + { + "epoch": 2.28, + "learning_rate": 2.3176556243174373e-06, + "loss": 1.6484, + "step": 12546 + }, + { + "epoch": 2.28, + "learning_rate": 2.3176010192937748e-06, + "loss": 0.2018, + "step": 12547 + }, + { + "epoch": 2.28, + "learning_rate": 2.3175464142701127e-06, + "loss": 1.6121, + "step": 12548 + }, + { + "epoch": 2.28, + "learning_rate": 2.3174918092464507e-06, + "loss": 3.5917, + "step": 12549 + }, + { + "epoch": 2.28, + "learning_rate": 2.3174372042227886e-06, + "loss": 1.5042, + "step": 12550 + }, + { + "epoch": 2.28, + "learning_rate": 2.3173825991991266e-06, + "loss": 1.8519, + "step": 12551 + }, + { + "epoch": 2.28, + "learning_rate": 2.317327994175464e-06, + "loss": 1.4465, + "step": 12552 + }, + { + "epoch": 2.28, + "learning_rate": 2.317273389151802e-06, + "loss": 0.3138, + "step": 12553 + }, + { + "epoch": 2.28, + "learning_rate": 2.31721878412814e-06, + "loss": 1.6447, + "step": 12554 + }, + { + "epoch": 2.28, + "learning_rate": 2.3171641791044775e-06, + "loss": 1.1179, + "step": 12555 + }, + { + "epoch": 2.28, + "learning_rate": 2.3171095740808154e-06, + "loss": 1.5218, + "step": 12556 + }, + { + "epoch": 2.28, + "learning_rate": 2.3170549690571534e-06, + "loss": 1.5089, + "step": 12557 + }, + { + "epoch": 2.28, + "learning_rate": 2.317000364033491e-06, + "loss": 0.7238, + "step": 12558 + }, + { + "epoch": 2.28, + "learning_rate": 2.316945759009829e-06, + "loss": 0.3964, + "step": 12559 + }, + { + "epoch": 2.28, + "learning_rate": 2.3168911539861668e-06, + "loss": 0.703, + "step": 12560 + }, + { + "epoch": 2.28, + "learning_rate": 2.3168365489625047e-06, + "loss": 0.1334, + "step": 12561 + }, + { + "epoch": 2.28, + "learning_rate": 2.3167819439388427e-06, + "loss": 0.5241, + "step": 12562 + }, + { + "epoch": 2.28, + "learning_rate": 2.3167273389151806e-06, + "loss": 0.5318, + "step": 12563 + }, + { + "epoch": 2.28, + "learning_rate": 2.316672733891518e-06, + "loss": 0.4431, + "step": 12564 + }, + { + "epoch": 2.28, + "learning_rate": 2.316618128867856e-06, + "loss": 0.3979, + "step": 12565 + }, + { + "epoch": 2.29, + "learning_rate": 2.3165635238441936e-06, + "loss": 0.5805, + "step": 12566 + }, + { + "epoch": 2.29, + "learning_rate": 2.3165089188205315e-06, + "loss": 0.8728, + "step": 12567 + }, + { + "epoch": 2.29, + "learning_rate": 2.3164543137968695e-06, + "loss": 2.0663, + "step": 12568 + }, + { + "epoch": 2.29, + "learning_rate": 2.316399708773207e-06, + "loss": 1.293, + "step": 12569 + }, + { + "epoch": 2.29, + "learning_rate": 2.316345103749545e-06, + "loss": 1.4762, + "step": 12570 + }, + { + "epoch": 2.29, + "learning_rate": 2.316290498725883e-06, + "loss": 0.0406, + "step": 12571 + }, + { + "epoch": 2.29, + "learning_rate": 2.3162358937022204e-06, + "loss": 0.3724, + "step": 12572 + }, + { + "epoch": 2.29, + "learning_rate": 2.3161812886785584e-06, + "loss": 0.1557, + "step": 12573 + }, + { + "epoch": 2.29, + "learning_rate": 2.3161266836548963e-06, + "loss": 1.3001, + "step": 12574 + }, + { + "epoch": 2.29, + "learning_rate": 2.3160720786312342e-06, + "loss": 0.6968, + "step": 12575 + }, + { + "epoch": 2.29, + "learning_rate": 2.316017473607572e-06, + "loss": 0.0458, + "step": 12576 + }, + { + "epoch": 2.29, + "learning_rate": 2.31596286858391e-06, + "loss": 0.6743, + "step": 12577 + }, + { + "epoch": 2.29, + "learning_rate": 2.3159082635602476e-06, + "loss": 0.6353, + "step": 12578 + }, + { + "epoch": 2.29, + "learning_rate": 2.3158536585365856e-06, + "loss": 0.455, + "step": 12579 + }, + { + "epoch": 2.29, + "learning_rate": 2.315799053512923e-06, + "loss": 1.0454, + "step": 12580 + }, + { + "epoch": 2.29, + "learning_rate": 2.315744448489261e-06, + "loss": 1.976, + "step": 12581 + }, + { + "epoch": 2.29, + "learning_rate": 2.315689843465599e-06, + "loss": 1.1365, + "step": 12582 + }, + { + "epoch": 2.29, + "learning_rate": 2.3156352384419365e-06, + "loss": 1.2384, + "step": 12583 + }, + { + "epoch": 2.29, + "learning_rate": 2.3155806334182745e-06, + "loss": 0.0454, + "step": 12584 + }, + { + "epoch": 2.29, + "learning_rate": 2.3155260283946124e-06, + "loss": 0.023, + "step": 12585 + }, + { + "epoch": 2.29, + "learning_rate": 2.31547142337095e-06, + "loss": 1.1674, + "step": 12586 + }, + { + "epoch": 2.29, + "learning_rate": 2.315416818347288e-06, + "loss": 2.795, + "step": 12587 + }, + { + "epoch": 2.29, + "learning_rate": 2.315362213323626e-06, + "loss": 0.4253, + "step": 12588 + }, + { + "epoch": 2.29, + "learning_rate": 2.3153076082999638e-06, + "loss": 0.4356, + "step": 12589 + }, + { + "epoch": 2.29, + "learning_rate": 2.3152530032763017e-06, + "loss": 1.7391, + "step": 12590 + }, + { + "epoch": 2.29, + "learning_rate": 2.3151983982526396e-06, + "loss": 1.661, + "step": 12591 + }, + { + "epoch": 2.29, + "learning_rate": 2.315143793228977e-06, + "loss": 0.4956, + "step": 12592 + }, + { + "epoch": 2.29, + "learning_rate": 2.315089188205315e-06, + "loss": 1.1429, + "step": 12593 + }, + { + "epoch": 2.29, + "learning_rate": 2.3150345831816526e-06, + "loss": 2.0743, + "step": 12594 + }, + { + "epoch": 2.29, + "learning_rate": 2.3149799781579906e-06, + "loss": 1.4938, + "step": 12595 + }, + { + "epoch": 2.29, + "learning_rate": 2.3149253731343285e-06, + "loss": 0.2731, + "step": 12596 + }, + { + "epoch": 2.29, + "learning_rate": 2.314870768110666e-06, + "loss": 2.282, + "step": 12597 + }, + { + "epoch": 2.29, + "learning_rate": 2.314816163087004e-06, + "loss": 0.0986, + "step": 12598 + }, + { + "epoch": 2.29, + "learning_rate": 2.314761558063342e-06, + "loss": 2.0452, + "step": 12599 + }, + { + "epoch": 2.29, + "learning_rate": 2.3147069530396794e-06, + "loss": 0.2937, + "step": 12600 + }, + { + "epoch": 2.29, + "learning_rate": 2.3146523480160174e-06, + "loss": 0.8089, + "step": 12601 + }, + { + "epoch": 2.29, + "learning_rate": 2.3145977429923553e-06, + "loss": 3.0362, + "step": 12602 + }, + { + "epoch": 2.29, + "learning_rate": 2.3145431379686933e-06, + "loss": 0.6195, + "step": 12603 + }, + { + "epoch": 2.29, + "learning_rate": 2.314488532945031e-06, + "loss": 1.0312, + "step": 12604 + }, + { + "epoch": 2.29, + "learning_rate": 2.314433927921369e-06, + "loss": 0.6879, + "step": 12605 + }, + { + "epoch": 2.29, + "learning_rate": 2.3143793228977067e-06, + "loss": 0.9483, + "step": 12606 + }, + { + "epoch": 2.29, + "learning_rate": 2.3143247178740446e-06, + "loss": 0.1253, + "step": 12607 + }, + { + "epoch": 2.29, + "learning_rate": 2.314270112850382e-06, + "loss": 0.5023, + "step": 12608 + }, + { + "epoch": 2.29, + "learning_rate": 2.31421550782672e-06, + "loss": 0.855, + "step": 12609 + }, + { + "epoch": 2.29, + "learning_rate": 2.314160902803058e-06, + "loss": 1.863, + "step": 12610 + }, + { + "epoch": 2.29, + "learning_rate": 2.3141062977793955e-06, + "loss": 0.3429, + "step": 12611 + }, + { + "epoch": 2.29, + "learning_rate": 2.3140516927557335e-06, + "loss": 0.3646, + "step": 12612 + }, + { + "epoch": 2.29, + "learning_rate": 2.3139970877320714e-06, + "loss": 0.3976, + "step": 12613 + }, + { + "epoch": 2.29, + "learning_rate": 2.313942482708409e-06, + "loss": 0.0393, + "step": 12614 + }, + { + "epoch": 2.29, + "learning_rate": 2.313887877684747e-06, + "loss": 0.6937, + "step": 12615 + }, + { + "epoch": 2.29, + "learning_rate": 2.313833272661085e-06, + "loss": 0.4585, + "step": 12616 + }, + { + "epoch": 2.29, + "learning_rate": 2.3137786676374228e-06, + "loss": 0.5792, + "step": 12617 + }, + { + "epoch": 2.29, + "learning_rate": 2.3137240626137607e-06, + "loss": 0.2518, + "step": 12618 + }, + { + "epoch": 2.29, + "learning_rate": 2.3136694575900987e-06, + "loss": 0.8016, + "step": 12619 + }, + { + "epoch": 2.29, + "learning_rate": 2.313614852566436e-06, + "loss": 0.0812, + "step": 12620 + }, + { + "epoch": 2.3, + "learning_rate": 2.313560247542774e-06, + "loss": 0.2882, + "step": 12621 + }, + { + "epoch": 2.3, + "learning_rate": 2.3135056425191116e-06, + "loss": 0.4588, + "step": 12622 + }, + { + "epoch": 2.3, + "learning_rate": 2.3134510374954496e-06, + "loss": 1.4036, + "step": 12623 + }, + { + "epoch": 2.3, + "learning_rate": 2.3133964324717875e-06, + "loss": 0.3356, + "step": 12624 + }, + { + "epoch": 2.3, + "learning_rate": 2.313341827448125e-06, + "loss": 0.06, + "step": 12625 + }, + { + "epoch": 2.3, + "learning_rate": 2.313287222424463e-06, + "loss": 0.4987, + "step": 12626 + }, + { + "epoch": 2.3, + "learning_rate": 2.313232617400801e-06, + "loss": 0.1082, + "step": 12627 + }, + { + "epoch": 2.3, + "learning_rate": 2.3131780123771385e-06, + "loss": 0.5187, + "step": 12628 + }, + { + "epoch": 2.3, + "learning_rate": 2.3131234073534764e-06, + "loss": 0.0167, + "step": 12629 + }, + { + "epoch": 2.3, + "learning_rate": 2.3130688023298143e-06, + "loss": 0.0217, + "step": 12630 + }, + { + "epoch": 2.3, + "learning_rate": 2.3130141973061523e-06, + "loss": 0.4058, + "step": 12631 + }, + { + "epoch": 2.3, + "learning_rate": 2.3129595922824902e-06, + "loss": 0.375, + "step": 12632 + }, + { + "epoch": 2.3, + "learning_rate": 2.3129049872588278e-06, + "loss": 1.0926, + "step": 12633 + }, + { + "epoch": 2.3, + "learning_rate": 2.3128503822351657e-06, + "loss": 0.4431, + "step": 12634 + }, + { + "epoch": 2.3, + "learning_rate": 2.3127957772115036e-06, + "loss": 0.6228, + "step": 12635 + }, + { + "epoch": 2.3, + "learning_rate": 2.312741172187841e-06, + "loss": 0.2911, + "step": 12636 + }, + { + "epoch": 2.3, + "learning_rate": 2.312686567164179e-06, + "loss": 0.3336, + "step": 12637 + }, + { + "epoch": 2.3, + "learning_rate": 2.312631962140517e-06, + "loss": 0.1177, + "step": 12638 + }, + { + "epoch": 2.3, + "learning_rate": 2.3125773571168546e-06, + "loss": 0.2542, + "step": 12639 + }, + { + "epoch": 2.3, + "learning_rate": 2.3125227520931925e-06, + "loss": 1.0776, + "step": 12640 + }, + { + "epoch": 2.3, + "learning_rate": 2.3124681470695305e-06, + "loss": 0.6912, + "step": 12641 + }, + { + "epoch": 2.3, + "learning_rate": 2.3124135420458684e-06, + "loss": 0.5322, + "step": 12642 + }, + { + "epoch": 2.3, + "learning_rate": 2.3123589370222063e-06, + "loss": 1.3182, + "step": 12643 + }, + { + "epoch": 2.3, + "learning_rate": 2.3123043319985443e-06, + "loss": 0.2904, + "step": 12644 + }, + { + "epoch": 2.3, + "learning_rate": 2.312249726974882e-06, + "loss": 1.4398, + "step": 12645 + }, + { + "epoch": 2.3, + "learning_rate": 2.3121951219512197e-06, + "loss": 0.4012, + "step": 12646 + }, + { + "epoch": 2.3, + "learning_rate": 2.3121405169275573e-06, + "loss": 0.2027, + "step": 12647 + }, + { + "epoch": 2.3, + "learning_rate": 2.312085911903895e-06, + "loss": 0.8271, + "step": 12648 + }, + { + "epoch": 2.3, + "learning_rate": 2.312031306880233e-06, + "loss": 0.3193, + "step": 12649 + }, + { + "epoch": 2.3, + "learning_rate": 2.3119767018565707e-06, + "loss": 1.9541, + "step": 12650 + }, + { + "epoch": 2.3, + "learning_rate": 2.3119220968329086e-06, + "loss": 1.0142, + "step": 12651 + }, + { + "epoch": 2.3, + "learning_rate": 2.3118674918092466e-06, + "loss": 1.0529, + "step": 12652 + }, + { + "epoch": 2.3, + "learning_rate": 2.311812886785584e-06, + "loss": 1.3169, + "step": 12653 + }, + { + "epoch": 2.3, + "learning_rate": 2.311758281761922e-06, + "loss": 2.2248, + "step": 12654 + }, + { + "epoch": 2.3, + "learning_rate": 2.31170367673826e-06, + "loss": 0.5355, + "step": 12655 + }, + { + "epoch": 2.3, + "learning_rate": 2.311649071714598e-06, + "loss": 1.0067, + "step": 12656 + }, + { + "epoch": 2.3, + "learning_rate": 2.311594466690936e-06, + "loss": 0.0133, + "step": 12657 + }, + { + "epoch": 2.3, + "learning_rate": 2.311539861667274e-06, + "loss": 2.1786, + "step": 12658 + }, + { + "epoch": 2.3, + "learning_rate": 2.3114852566436113e-06, + "loss": 0.2663, + "step": 12659 + }, + { + "epoch": 2.3, + "learning_rate": 2.3114306516199493e-06, + "loss": 0.5555, + "step": 12660 + }, + { + "epoch": 2.3, + "learning_rate": 2.3113760465962868e-06, + "loss": 0.1913, + "step": 12661 + }, + { + "epoch": 2.3, + "learning_rate": 2.3113214415726247e-06, + "loss": 4.3389, + "step": 12662 + }, + { + "epoch": 2.3, + "learning_rate": 2.3112668365489627e-06, + "loss": 0.909, + "step": 12663 + }, + { + "epoch": 2.3, + "learning_rate": 2.3112122315253e-06, + "loss": 0.6144, + "step": 12664 + }, + { + "epoch": 2.3, + "learning_rate": 2.311157626501638e-06, + "loss": 0.751, + "step": 12665 + }, + { + "epoch": 2.3, + "learning_rate": 2.311103021477976e-06, + "loss": 1.9079, + "step": 12666 + }, + { + "epoch": 2.3, + "learning_rate": 2.3110484164543136e-06, + "loss": 1.9852, + "step": 12667 + }, + { + "epoch": 2.3, + "learning_rate": 2.3109938114306515e-06, + "loss": 3.0911, + "step": 12668 + }, + { + "epoch": 2.3, + "learning_rate": 2.3109392064069895e-06, + "loss": 0.7704, + "step": 12669 + }, + { + "epoch": 2.3, + "learning_rate": 2.3108846013833274e-06, + "loss": 1.1664, + "step": 12670 + }, + { + "epoch": 2.3, + "learning_rate": 2.3108299963596654e-06, + "loss": 0.575, + "step": 12671 + }, + { + "epoch": 2.3, + "learning_rate": 2.3107753913360033e-06, + "loss": 1.7357, + "step": 12672 + }, + { + "epoch": 2.3, + "learning_rate": 2.310720786312341e-06, + "loss": 1.1773, + "step": 12673 + }, + { + "epoch": 2.3, + "learning_rate": 2.3106661812886788e-06, + "loss": 0.7366, + "step": 12674 + }, + { + "epoch": 2.3, + "learning_rate": 2.3106115762650163e-06, + "loss": 0.6099, + "step": 12675 + }, + { + "epoch": 2.31, + "learning_rate": 2.3105569712413542e-06, + "loss": 0.0317, + "step": 12676 + }, + { + "epoch": 2.31, + "learning_rate": 2.310502366217692e-06, + "loss": 1.917, + "step": 12677 + }, + { + "epoch": 2.31, + "learning_rate": 2.3104477611940297e-06, + "loss": 1.2937, + "step": 12678 + }, + { + "epoch": 2.31, + "learning_rate": 2.3103931561703676e-06, + "loss": 0.0265, + "step": 12679 + }, + { + "epoch": 2.31, + "learning_rate": 2.3103385511467056e-06, + "loss": 0.6214, + "step": 12680 + }, + { + "epoch": 2.31, + "learning_rate": 2.310283946123043e-06, + "loss": 1.1434, + "step": 12681 + }, + { + "epoch": 2.31, + "learning_rate": 2.310229341099381e-06, + "loss": 0.0372, + "step": 12682 + }, + { + "epoch": 2.31, + "learning_rate": 2.310174736075719e-06, + "loss": 0.3653, + "step": 12683 + }, + { + "epoch": 2.31, + "learning_rate": 2.310120131052057e-06, + "loss": 0.5073, + "step": 12684 + }, + { + "epoch": 2.31, + "learning_rate": 2.310065526028395e-06, + "loss": 1.1899, + "step": 12685 + }, + { + "epoch": 2.31, + "learning_rate": 2.310010921004733e-06, + "loss": 1.5358, + "step": 12686 + }, + { + "epoch": 2.31, + "learning_rate": 2.3099563159810703e-06, + "loss": 1.1222, + "step": 12687 + }, + { + "epoch": 2.31, + "learning_rate": 2.3099017109574083e-06, + "loss": 0.0325, + "step": 12688 + }, + { + "epoch": 2.31, + "learning_rate": 2.309847105933746e-06, + "loss": 0.5701, + "step": 12689 + }, + { + "epoch": 2.31, + "learning_rate": 2.3097925009100838e-06, + "loss": 0.2672, + "step": 12690 + }, + { + "epoch": 2.31, + "learning_rate": 2.3097378958864217e-06, + "loss": 0.6959, + "step": 12691 + }, + { + "epoch": 2.31, + "learning_rate": 2.3096832908627592e-06, + "loss": 0.4061, + "step": 12692 + }, + { + "epoch": 2.31, + "learning_rate": 2.309628685839097e-06, + "loss": 1.3394, + "step": 12693 + }, + { + "epoch": 2.31, + "learning_rate": 2.309574080815435e-06, + "loss": 0.884, + "step": 12694 + }, + { + "epoch": 2.31, + "learning_rate": 2.3095194757917726e-06, + "loss": 1.7066, + "step": 12695 + }, + { + "epoch": 2.31, + "learning_rate": 2.3094648707681106e-06, + "loss": 0.0499, + "step": 12696 + }, + { + "epoch": 2.31, + "learning_rate": 2.3094102657444485e-06, + "loss": 0.3438, + "step": 12697 + }, + { + "epoch": 2.31, + "learning_rate": 2.3093556607207865e-06, + "loss": 0.0309, + "step": 12698 + }, + { + "epoch": 2.31, + "learning_rate": 2.3093010556971244e-06, + "loss": 0.3171, + "step": 12699 + }, + { + "epoch": 2.31, + "learning_rate": 2.309246450673462e-06, + "loss": 0.8171, + "step": 12700 + }, + { + "epoch": 2.31, + "learning_rate": 2.3091918456498e-06, + "loss": 0.4603, + "step": 12701 + }, + { + "epoch": 2.31, + "learning_rate": 2.309137240626138e-06, + "loss": 2.2632, + "step": 12702 + }, + { + "epoch": 2.31, + "learning_rate": 2.3090826356024753e-06, + "loss": 0.4233, + "step": 12703 + }, + { + "epoch": 2.31, + "learning_rate": 2.3090280305788133e-06, + "loss": 3.4888, + "step": 12704 + }, + { + "epoch": 2.31, + "learning_rate": 2.308973425555151e-06, + "loss": 0.6104, + "step": 12705 + }, + { + "epoch": 2.31, + "learning_rate": 2.3089188205314887e-06, + "loss": 0.3033, + "step": 12706 + }, + { + "epoch": 2.31, + "learning_rate": 2.3088642155078267e-06, + "loss": 1.3739, + "step": 12707 + }, + { + "epoch": 2.31, + "learning_rate": 2.3088096104841646e-06, + "loss": 0.2522, + "step": 12708 + }, + { + "epoch": 2.31, + "learning_rate": 2.308755005460502e-06, + "loss": 2.4146, + "step": 12709 + }, + { + "epoch": 2.31, + "learning_rate": 2.30870040043684e-06, + "loss": 0.8393, + "step": 12710 + }, + { + "epoch": 2.31, + "learning_rate": 2.308645795413178e-06, + "loss": 0.6705, + "step": 12711 + }, + { + "epoch": 2.31, + "learning_rate": 2.308591190389516e-06, + "loss": 0.8087, + "step": 12712 + }, + { + "epoch": 2.31, + "learning_rate": 2.308536585365854e-06, + "loss": 0.074, + "step": 12713 + }, + { + "epoch": 2.31, + "learning_rate": 2.3084819803421914e-06, + "loss": 0.4202, + "step": 12714 + }, + { + "epoch": 2.31, + "learning_rate": 2.3084273753185294e-06, + "loss": 0.6151, + "step": 12715 + }, + { + "epoch": 2.31, + "learning_rate": 2.3083727702948673e-06, + "loss": 0.7529, + "step": 12716 + }, + { + "epoch": 2.31, + "learning_rate": 2.308318165271205e-06, + "loss": 0.469, + "step": 12717 + }, + { + "epoch": 2.31, + "learning_rate": 2.3082635602475428e-06, + "loss": 1.274, + "step": 12718 + }, + { + "epoch": 2.31, + "learning_rate": 2.3082089552238807e-06, + "loss": 1.2705, + "step": 12719 + }, + { + "epoch": 2.31, + "learning_rate": 2.3081543502002182e-06, + "loss": 0.1396, + "step": 12720 + }, + { + "epoch": 2.31, + "learning_rate": 2.308099745176556e-06, + "loss": 1.276, + "step": 12721 + }, + { + "epoch": 2.31, + "learning_rate": 2.308045140152894e-06, + "loss": 0.0358, + "step": 12722 + }, + { + "epoch": 2.31, + "learning_rate": 2.3079905351292316e-06, + "loss": 2.2441, + "step": 12723 + }, + { + "epoch": 2.31, + "learning_rate": 2.3079359301055696e-06, + "loss": 0.0213, + "step": 12724 + }, + { + "epoch": 2.31, + "learning_rate": 2.3078813250819075e-06, + "loss": 0.4089, + "step": 12725 + }, + { + "epoch": 2.31, + "learning_rate": 2.3078267200582455e-06, + "loss": 0.8492, + "step": 12726 + }, + { + "epoch": 2.31, + "learning_rate": 2.3077721150345834e-06, + "loss": 2.7536, + "step": 12727 + }, + { + "epoch": 2.31, + "learning_rate": 2.307717510010921e-06, + "loss": 1.8996, + "step": 12728 + }, + { + "epoch": 2.31, + "learning_rate": 2.307662904987259e-06, + "loss": 0.557, + "step": 12729 + }, + { + "epoch": 2.31, + "learning_rate": 2.307608299963597e-06, + "loss": 0.7366, + "step": 12730 + }, + { + "epoch": 2.32, + "learning_rate": 2.3075536949399343e-06, + "loss": 3.0515, + "step": 12731 + }, + { + "epoch": 2.32, + "learning_rate": 2.3074990899162723e-06, + "loss": 1.1774, + "step": 12732 + }, + { + "epoch": 2.32, + "learning_rate": 2.3074444848926102e-06, + "loss": 2.0093, + "step": 12733 + }, + { + "epoch": 2.32, + "learning_rate": 2.3073898798689478e-06, + "loss": 1.0636, + "step": 12734 + }, + { + "epoch": 2.32, + "learning_rate": 2.3073352748452857e-06, + "loss": 0.2455, + "step": 12735 + }, + { + "epoch": 2.32, + "learning_rate": 2.3072806698216236e-06, + "loss": 0.4287, + "step": 12736 + }, + { + "epoch": 2.32, + "learning_rate": 2.3072260647979616e-06, + "loss": 0.995, + "step": 12737 + }, + { + "epoch": 2.32, + "learning_rate": 2.3071714597742995e-06, + "loss": 0.2007, + "step": 12738 + }, + { + "epoch": 2.32, + "learning_rate": 2.3071168547506375e-06, + "loss": 1.1277, + "step": 12739 + }, + { + "epoch": 2.32, + "learning_rate": 2.307062249726975e-06, + "loss": 1.8426, + "step": 12740 + }, + { + "epoch": 2.32, + "learning_rate": 2.307007644703313e-06, + "loss": 0.7487, + "step": 12741 + }, + { + "epoch": 2.32, + "learning_rate": 2.3069530396796505e-06, + "loss": 0.242, + "step": 12742 + }, + { + "epoch": 2.32, + "learning_rate": 2.3068984346559884e-06, + "loss": 0.1045, + "step": 12743 + }, + { + "epoch": 2.32, + "learning_rate": 2.3068438296323263e-06, + "loss": 0.0607, + "step": 12744 + }, + { + "epoch": 2.32, + "learning_rate": 2.306789224608664e-06, + "loss": 1.13, + "step": 12745 + }, + { + "epoch": 2.32, + "learning_rate": 2.306734619585002e-06, + "loss": 2.0529, + "step": 12746 + }, + { + "epoch": 2.32, + "learning_rate": 2.3066800145613397e-06, + "loss": 0.7758, + "step": 12747 + }, + { + "epoch": 2.32, + "learning_rate": 2.3066254095376773e-06, + "loss": 1.8414, + "step": 12748 + }, + { + "epoch": 2.32, + "learning_rate": 2.3065708045140152e-06, + "loss": 1.1992, + "step": 12749 + }, + { + "epoch": 2.32, + "learning_rate": 2.306516199490353e-06, + "loss": 0.2778, + "step": 12750 + }, + { + "epoch": 2.32, + "learning_rate": 2.306461594466691e-06, + "loss": 0.072, + "step": 12751 + }, + { + "epoch": 2.32, + "learning_rate": 2.306406989443029e-06, + "loss": 1.6026, + "step": 12752 + }, + { + "epoch": 2.32, + "learning_rate": 2.306352384419367e-06, + "loss": 0.5031, + "step": 12753 + }, + { + "epoch": 2.32, + "learning_rate": 2.3062977793957045e-06, + "loss": 1.0291, + "step": 12754 + }, + { + "epoch": 2.32, + "learning_rate": 2.3062431743720424e-06, + "loss": 0.2259, + "step": 12755 + }, + { + "epoch": 2.32, + "learning_rate": 2.30618856934838e-06, + "loss": 0.4256, + "step": 12756 + }, + { + "epoch": 2.32, + "learning_rate": 2.306133964324718e-06, + "loss": 0.8922, + "step": 12757 + }, + { + "epoch": 2.32, + "learning_rate": 2.306079359301056e-06, + "loss": 1.8529, + "step": 12758 + }, + { + "epoch": 2.32, + "learning_rate": 2.3060247542773934e-06, + "loss": 0.7627, + "step": 12759 + }, + { + "epoch": 2.32, + "learning_rate": 2.3059701492537313e-06, + "loss": 0.5706, + "step": 12760 + }, + { + "epoch": 2.32, + "learning_rate": 2.3059155442300693e-06, + "loss": 1.7754, + "step": 12761 + }, + { + "epoch": 2.32, + "learning_rate": 2.3058609392064068e-06, + "loss": 0.2303, + "step": 12762 + }, + { + "epoch": 2.32, + "learning_rate": 2.3058063341827447e-06, + "loss": 1.1892, + "step": 12763 + }, + { + "epoch": 2.32, + "learning_rate": 2.3057517291590827e-06, + "loss": 1.4454, + "step": 12764 + }, + { + "epoch": 2.32, + "learning_rate": 2.3056971241354206e-06, + "loss": 0.0264, + "step": 12765 + }, + { + "epoch": 2.32, + "learning_rate": 2.3056425191117586e-06, + "loss": 1.8538, + "step": 12766 + }, + { + "epoch": 2.32, + "learning_rate": 2.3055879140880965e-06, + "loss": 1.263, + "step": 12767 + }, + { + "epoch": 2.32, + "learning_rate": 2.305533309064434e-06, + "loss": 3.9018, + "step": 12768 + }, + { + "epoch": 2.32, + "learning_rate": 2.305478704040772e-06, + "loss": 1.205, + "step": 12769 + }, + { + "epoch": 2.32, + "learning_rate": 2.3054240990171095e-06, + "loss": 0.1398, + "step": 12770 + }, + { + "epoch": 2.32, + "learning_rate": 2.3053694939934474e-06, + "loss": 0.4787, + "step": 12771 + }, + { + "epoch": 2.32, + "learning_rate": 2.3053148889697854e-06, + "loss": 0.8189, + "step": 12772 + }, + { + "epoch": 2.32, + "learning_rate": 2.305260283946123e-06, + "loss": 0.2146, + "step": 12773 + }, + { + "epoch": 2.32, + "learning_rate": 2.305205678922461e-06, + "loss": 0.5028, + "step": 12774 + }, + { + "epoch": 2.32, + "learning_rate": 2.3051510738987988e-06, + "loss": 3.1634, + "step": 12775 + }, + { + "epoch": 2.32, + "learning_rate": 2.3050964688751363e-06, + "loss": 0.5396, + "step": 12776 + }, + { + "epoch": 2.32, + "learning_rate": 2.3050418638514742e-06, + "loss": 0.7441, + "step": 12777 + }, + { + "epoch": 2.32, + "learning_rate": 2.304987258827812e-06, + "loss": 0.8726, + "step": 12778 + }, + { + "epoch": 2.32, + "learning_rate": 2.30493265380415e-06, + "loss": 0.4362, + "step": 12779 + }, + { + "epoch": 2.32, + "learning_rate": 2.304878048780488e-06, + "loss": 1.084, + "step": 12780 + }, + { + "epoch": 2.32, + "learning_rate": 2.3048234437568256e-06, + "loss": 0.4007, + "step": 12781 + }, + { + "epoch": 2.32, + "learning_rate": 2.3047688387331635e-06, + "loss": 1.2146, + "step": 12782 + }, + { + "epoch": 2.32, + "learning_rate": 2.3047142337095015e-06, + "loss": 0.0751, + "step": 12783 + }, + { + "epoch": 2.32, + "learning_rate": 2.304659628685839e-06, + "loss": 1.9744, + "step": 12784 + }, + { + "epoch": 2.32, + "learning_rate": 2.304605023662177e-06, + "loss": 1.1437, + "step": 12785 + }, + { + "epoch": 2.33, + "learning_rate": 2.304550418638515e-06, + "loss": 0.8555, + "step": 12786 + }, + { + "epoch": 2.33, + "learning_rate": 2.3044958136148524e-06, + "loss": 0.0302, + "step": 12787 + }, + { + "epoch": 2.33, + "learning_rate": 2.3044412085911903e-06, + "loss": 0.2902, + "step": 12788 + }, + { + "epoch": 2.33, + "learning_rate": 2.3043866035675283e-06, + "loss": 0.3532, + "step": 12789 + }, + { + "epoch": 2.33, + "learning_rate": 2.304331998543866e-06, + "loss": 1.7895, + "step": 12790 + }, + { + "epoch": 2.33, + "learning_rate": 2.3042773935202038e-06, + "loss": 1.4596, + "step": 12791 + }, + { + "epoch": 2.33, + "learning_rate": 2.3042227884965417e-06, + "loss": 0.1306, + "step": 12792 + }, + { + "epoch": 2.33, + "learning_rate": 2.3041681834728796e-06, + "loss": 0.3527, + "step": 12793 + }, + { + "epoch": 2.33, + "learning_rate": 2.3041135784492176e-06, + "loss": 0.2598, + "step": 12794 + }, + { + "epoch": 2.33, + "learning_rate": 2.304058973425555e-06, + "loss": 1.2708, + "step": 12795 + }, + { + "epoch": 2.33, + "learning_rate": 2.304004368401893e-06, + "loss": 0.7026, + "step": 12796 + }, + { + "epoch": 2.33, + "learning_rate": 2.303949763378231e-06, + "loss": 0.49, + "step": 12797 + }, + { + "epoch": 2.33, + "learning_rate": 2.3038951583545685e-06, + "loss": 0.0833, + "step": 12798 + }, + { + "epoch": 2.33, + "learning_rate": 2.3038405533309065e-06, + "loss": 3.9756, + "step": 12799 + }, + { + "epoch": 2.33, + "learning_rate": 2.3037859483072444e-06, + "loss": 0.1373, + "step": 12800 + }, + { + "epoch": 2.33, + "learning_rate": 2.303731343283582e-06, + "loss": 1.4022, + "step": 12801 + }, + { + "epoch": 2.33, + "learning_rate": 2.30367673825992e-06, + "loss": 0.5873, + "step": 12802 + }, + { + "epoch": 2.33, + "learning_rate": 2.303622133236258e-06, + "loss": 2.4922, + "step": 12803 + }, + { + "epoch": 2.33, + "learning_rate": 2.3035675282125953e-06, + "loss": 1.3585, + "step": 12804 + }, + { + "epoch": 2.33, + "learning_rate": 2.3035129231889333e-06, + "loss": 1.5923, + "step": 12805 + }, + { + "epoch": 2.33, + "learning_rate": 2.303458318165271e-06, + "loss": 1.4727, + "step": 12806 + }, + { + "epoch": 2.33, + "learning_rate": 2.303403713141609e-06, + "loss": 0.4865, + "step": 12807 + }, + { + "epoch": 2.33, + "learning_rate": 2.303349108117947e-06, + "loss": 0.4094, + "step": 12808 + }, + { + "epoch": 2.33, + "learning_rate": 2.3032945030942846e-06, + "loss": 1.2328, + "step": 12809 + }, + { + "epoch": 2.33, + "learning_rate": 2.3032398980706226e-06, + "loss": 0.0271, + "step": 12810 + }, + { + "epoch": 2.33, + "learning_rate": 2.3031852930469605e-06, + "loss": 0.4823, + "step": 12811 + }, + { + "epoch": 2.33, + "learning_rate": 2.303130688023298e-06, + "loss": 0.0634, + "step": 12812 + }, + { + "epoch": 2.33, + "learning_rate": 2.303076082999636e-06, + "loss": 0.0147, + "step": 12813 + }, + { + "epoch": 2.33, + "learning_rate": 2.303021477975974e-06, + "loss": 0.3054, + "step": 12814 + }, + { + "epoch": 2.33, + "learning_rate": 2.3029668729523114e-06, + "loss": 0.2443, + "step": 12815 + }, + { + "epoch": 2.33, + "learning_rate": 2.3029122679286494e-06, + "loss": 0.886, + "step": 12816 + }, + { + "epoch": 2.33, + "learning_rate": 2.3028576629049873e-06, + "loss": 0.3793, + "step": 12817 + }, + { + "epoch": 2.33, + "learning_rate": 2.3028030578813253e-06, + "loss": 0.5066, + "step": 12818 + }, + { + "epoch": 2.33, + "learning_rate": 2.302748452857663e-06, + "loss": 0.8311, + "step": 12819 + }, + { + "epoch": 2.33, + "learning_rate": 2.302693847834001e-06, + "loss": 0.3666, + "step": 12820 + }, + { + "epoch": 2.33, + "learning_rate": 2.3026392428103387e-06, + "loss": 3.3324, + "step": 12821 + }, + { + "epoch": 2.33, + "learning_rate": 2.3025846377866766e-06, + "loss": 1.7074, + "step": 12822 + }, + { + "epoch": 2.33, + "learning_rate": 2.302530032763014e-06, + "loss": 0.365, + "step": 12823 + }, + { + "epoch": 2.33, + "learning_rate": 2.302475427739352e-06, + "loss": 0.6485, + "step": 12824 + }, + { + "epoch": 2.33, + "learning_rate": 2.30242082271569e-06, + "loss": 0.4427, + "step": 12825 + }, + { + "epoch": 2.33, + "learning_rate": 2.3023662176920275e-06, + "loss": 0.4944, + "step": 12826 + }, + { + "epoch": 2.33, + "learning_rate": 2.3023116126683655e-06, + "loss": 2.0001, + "step": 12827 + }, + { + "epoch": 2.33, + "learning_rate": 2.3022570076447034e-06, + "loss": 0.7758, + "step": 12828 + }, + { + "epoch": 2.33, + "learning_rate": 2.302202402621041e-06, + "loss": 0.0584, + "step": 12829 + }, + { + "epoch": 2.33, + "learning_rate": 2.302147797597379e-06, + "loss": 1.3769, + "step": 12830 + }, + { + "epoch": 2.33, + "learning_rate": 2.302093192573717e-06, + "loss": 0.6489, + "step": 12831 + }, + { + "epoch": 2.33, + "learning_rate": 2.3020385875500548e-06, + "loss": 0.4835, + "step": 12832 + }, + { + "epoch": 2.33, + "learning_rate": 2.3019839825263927e-06, + "loss": 0.7288, + "step": 12833 + }, + { + "epoch": 2.33, + "learning_rate": 2.3019293775027307e-06, + "loss": 2.7185, + "step": 12834 + }, + { + "epoch": 2.33, + "learning_rate": 2.301874772479068e-06, + "loss": 0.5395, + "step": 12835 + }, + { + "epoch": 2.33, + "learning_rate": 2.301820167455406e-06, + "loss": 2.4243, + "step": 12836 + }, + { + "epoch": 2.33, + "learning_rate": 2.3017655624317436e-06, + "loss": 0.3937, + "step": 12837 + }, + { + "epoch": 2.33, + "learning_rate": 2.3017109574080816e-06, + "loss": 0.154, + "step": 12838 + }, + { + "epoch": 2.33, + "learning_rate": 2.3016563523844195e-06, + "loss": 0.5912, + "step": 12839 + }, + { + "epoch": 2.33, + "learning_rate": 2.301601747360757e-06, + "loss": 1.8574, + "step": 12840 + }, + { + "epoch": 2.34, + "learning_rate": 2.301547142337095e-06, + "loss": 1.2208, + "step": 12841 + }, + { + "epoch": 2.34, + "learning_rate": 2.301492537313433e-06, + "loss": 0.0527, + "step": 12842 + }, + { + "epoch": 2.34, + "learning_rate": 2.3014379322897705e-06, + "loss": 0.0338, + "step": 12843 + }, + { + "epoch": 2.34, + "learning_rate": 2.3013833272661084e-06, + "loss": 1.3485, + "step": 12844 + }, + { + "epoch": 2.34, + "learning_rate": 2.3013287222424463e-06, + "loss": 1.3947, + "step": 12845 + }, + { + "epoch": 2.34, + "learning_rate": 2.3012741172187843e-06, + "loss": 0.6407, + "step": 12846 + }, + { + "epoch": 2.34, + "learning_rate": 2.3012195121951222e-06, + "loss": 1.2805, + "step": 12847 + }, + { + "epoch": 2.34, + "learning_rate": 2.30116490717146e-06, + "loss": 1.8994, + "step": 12848 + }, + { + "epoch": 2.34, + "learning_rate": 2.3011103021477977e-06, + "loss": 0.2258, + "step": 12849 + }, + { + "epoch": 2.34, + "learning_rate": 2.3010556971241356e-06, + "loss": 0.6128, + "step": 12850 + }, + { + "epoch": 2.34, + "learning_rate": 2.301001092100473e-06, + "loss": 0.5898, + "step": 12851 + }, + { + "epoch": 2.34, + "learning_rate": 2.300946487076811e-06, + "loss": 0.4203, + "step": 12852 + }, + { + "epoch": 2.34, + "learning_rate": 2.300891882053149e-06, + "loss": 1.5554, + "step": 12853 + }, + { + "epoch": 2.34, + "learning_rate": 2.3008372770294866e-06, + "loss": 0.6234, + "step": 12854 + }, + { + "epoch": 2.34, + "learning_rate": 2.3007826720058245e-06, + "loss": 0.1982, + "step": 12855 + }, + { + "epoch": 2.34, + "learning_rate": 2.3007280669821625e-06, + "loss": 0.622, + "step": 12856 + }, + { + "epoch": 2.34, + "learning_rate": 2.3006734619585e-06, + "loss": 0.7871, + "step": 12857 + }, + { + "epoch": 2.34, + "learning_rate": 2.300618856934838e-06, + "loss": 0.2531, + "step": 12858 + }, + { + "epoch": 2.34, + "learning_rate": 2.300564251911176e-06, + "loss": 0.8002, + "step": 12859 + }, + { + "epoch": 2.34, + "learning_rate": 2.300509646887514e-06, + "loss": 0.4427, + "step": 12860 + }, + { + "epoch": 2.34, + "learning_rate": 2.3004550418638517e-06, + "loss": 0.6383, + "step": 12861 + }, + { + "epoch": 2.34, + "learning_rate": 2.3004004368401893e-06, + "loss": 2.6341, + "step": 12862 + }, + { + "epoch": 2.34, + "learning_rate": 2.300345831816527e-06, + "loss": 1.9802, + "step": 12863 + }, + { + "epoch": 2.34, + "learning_rate": 2.300291226792865e-06, + "loss": 1.2958, + "step": 12864 + }, + { + "epoch": 2.34, + "learning_rate": 2.3002366217692027e-06, + "loss": 0.7395, + "step": 12865 + }, + { + "epoch": 2.34, + "learning_rate": 2.3001820167455406e-06, + "loss": 1.5056, + "step": 12866 + }, + { + "epoch": 2.34, + "learning_rate": 2.3001274117218786e-06, + "loss": 0.6817, + "step": 12867 + }, + { + "epoch": 2.34, + "learning_rate": 2.300072806698216e-06, + "loss": 0.8229, + "step": 12868 + }, + { + "epoch": 2.34, + "learning_rate": 2.300018201674554e-06, + "loss": 0.7612, + "step": 12869 + }, + { + "epoch": 2.34, + "learning_rate": 2.299963596650892e-06, + "loss": 1.0228, + "step": 12870 + }, + { + "epoch": 2.34, + "learning_rate": 2.2999089916272295e-06, + "loss": 2.8813, + "step": 12871 + }, + { + "epoch": 2.34, + "learning_rate": 2.2998543866035674e-06, + "loss": 1.0223, + "step": 12872 + }, + { + "epoch": 2.34, + "learning_rate": 2.2997997815799054e-06, + "loss": 0.31, + "step": 12873 + }, + { + "epoch": 2.34, + "learning_rate": 2.2997451765562433e-06, + "loss": 1.5683, + "step": 12874 + }, + { + "epoch": 2.34, + "learning_rate": 2.2996905715325813e-06, + "loss": 2.2055, + "step": 12875 + }, + { + "epoch": 2.34, + "learning_rate": 2.2996359665089188e-06, + "loss": 0.7135, + "step": 12876 + }, + { + "epoch": 2.34, + "learning_rate": 2.2995813614852567e-06, + "loss": 0.9933, + "step": 12877 + }, + { + "epoch": 2.34, + "learning_rate": 2.2995267564615947e-06, + "loss": 1.7747, + "step": 12878 + }, + { + "epoch": 2.34, + "learning_rate": 2.299472151437932e-06, + "loss": 0.8, + "step": 12879 + }, + { + "epoch": 2.34, + "learning_rate": 2.29941754641427e-06, + "loss": 1.6038, + "step": 12880 + }, + { + "epoch": 2.34, + "learning_rate": 2.299362941390608e-06, + "loss": 0.6691, + "step": 12881 + }, + { + "epoch": 2.34, + "learning_rate": 2.2993083363669456e-06, + "loss": 0.5502, + "step": 12882 + }, + { + "epoch": 2.34, + "learning_rate": 2.2992537313432835e-06, + "loss": 1.6223, + "step": 12883 + }, + { + "epoch": 2.34, + "learning_rate": 2.2991991263196215e-06, + "loss": 0.7459, + "step": 12884 + }, + { + "epoch": 2.34, + "learning_rate": 2.299144521295959e-06, + "loss": 0.9108, + "step": 12885 + }, + { + "epoch": 2.34, + "learning_rate": 2.299089916272297e-06, + "loss": 0.8666, + "step": 12886 + }, + { + "epoch": 2.34, + "learning_rate": 2.299035311248635e-06, + "loss": 0.9668, + "step": 12887 + }, + { + "epoch": 2.34, + "learning_rate": 2.298980706224973e-06, + "loss": 0.6515, + "step": 12888 + }, + { + "epoch": 2.34, + "learning_rate": 2.2989261012013108e-06, + "loss": 1.1431, + "step": 12889 + }, + { + "epoch": 2.34, + "learning_rate": 2.2988714961776483e-06, + "loss": 0.4851, + "step": 12890 + }, + { + "epoch": 2.34, + "learning_rate": 2.2988168911539862e-06, + "loss": 0.4594, + "step": 12891 + }, + { + "epoch": 2.34, + "learning_rate": 2.298762286130324e-06, + "loss": 0.0608, + "step": 12892 + }, + { + "epoch": 2.34, + "learning_rate": 2.2987076811066617e-06, + "loss": 1.2341, + "step": 12893 + }, + { + "epoch": 2.34, + "learning_rate": 2.2986530760829996e-06, + "loss": 0.2233, + "step": 12894 + }, + { + "epoch": 2.34, + "learning_rate": 2.2985984710593376e-06, + "loss": 1.188, + "step": 12895 + }, + { + "epoch": 2.35, + "learning_rate": 2.298543866035675e-06, + "loss": 1.0739, + "step": 12896 + }, + { + "epoch": 2.35, + "learning_rate": 2.298489261012013e-06, + "loss": 2.6151, + "step": 12897 + }, + { + "epoch": 2.35, + "learning_rate": 2.298434655988351e-06, + "loss": 0.8143, + "step": 12898 + }, + { + "epoch": 2.35, + "learning_rate": 2.298380050964689e-06, + "loss": 0.7026, + "step": 12899 + }, + { + "epoch": 2.35, + "learning_rate": 2.298325445941027e-06, + "loss": 0.3661, + "step": 12900 + }, + { + "epoch": 2.35, + "learning_rate": 2.298270840917365e-06, + "loss": 2.3243, + "step": 12901 + }, + { + "epoch": 2.35, + "learning_rate": 2.2982162358937023e-06, + "loss": 1.0154, + "step": 12902 + }, + { + "epoch": 2.35, + "learning_rate": 2.2981616308700403e-06, + "loss": 4.3868, + "step": 12903 + }, + { + "epoch": 2.35, + "learning_rate": 2.298107025846378e-06, + "loss": 0.7147, + "step": 12904 + }, + { + "epoch": 2.35, + "learning_rate": 2.2980524208227157e-06, + "loss": 0.6698, + "step": 12905 + }, + { + "epoch": 2.35, + "learning_rate": 2.2979978157990537e-06, + "loss": 0.0114, + "step": 12906 + }, + { + "epoch": 2.35, + "learning_rate": 2.297943210775391e-06, + "loss": 0.2869, + "step": 12907 + }, + { + "epoch": 2.35, + "learning_rate": 2.297888605751729e-06, + "loss": 1.7172, + "step": 12908 + }, + { + "epoch": 2.35, + "learning_rate": 2.297834000728067e-06, + "loss": 0.0313, + "step": 12909 + }, + { + "epoch": 2.35, + "learning_rate": 2.2977793957044046e-06, + "loss": 1.9873, + "step": 12910 + }, + { + "epoch": 2.35, + "learning_rate": 2.2977247906807426e-06, + "loss": 2.3083, + "step": 12911 + }, + { + "epoch": 2.35, + "learning_rate": 2.2976701856570805e-06, + "loss": 0.843, + "step": 12912 + }, + { + "epoch": 2.35, + "learning_rate": 2.2976155806334184e-06, + "loss": 0.7396, + "step": 12913 + }, + { + "epoch": 2.35, + "learning_rate": 2.2975609756097564e-06, + "loss": 0.2413, + "step": 12914 + }, + { + "epoch": 2.35, + "learning_rate": 2.2975063705860943e-06, + "loss": 0.4245, + "step": 12915 + }, + { + "epoch": 2.35, + "learning_rate": 2.297451765562432e-06, + "loss": 0.8649, + "step": 12916 + }, + { + "epoch": 2.35, + "learning_rate": 2.29739716053877e-06, + "loss": 0.0126, + "step": 12917 + }, + { + "epoch": 2.35, + "learning_rate": 2.2973425555151073e-06, + "loss": 0.9464, + "step": 12918 + }, + { + "epoch": 2.35, + "learning_rate": 2.2972879504914453e-06, + "loss": 0.8182, + "step": 12919 + }, + { + "epoch": 2.35, + "learning_rate": 2.297233345467783e-06, + "loss": 2.1999, + "step": 12920 + }, + { + "epoch": 2.35, + "learning_rate": 2.2971787404441207e-06, + "loss": 0.1718, + "step": 12921 + }, + { + "epoch": 2.35, + "learning_rate": 2.2971241354204587e-06, + "loss": 0.4914, + "step": 12922 + }, + { + "epoch": 2.35, + "learning_rate": 2.2970695303967966e-06, + "loss": 0.067, + "step": 12923 + }, + { + "epoch": 2.35, + "learning_rate": 2.297014925373134e-06, + "loss": 0.981, + "step": 12924 + }, + { + "epoch": 2.35, + "learning_rate": 2.296960320349472e-06, + "loss": 0.3378, + "step": 12925 + }, + { + "epoch": 2.35, + "learning_rate": 2.29690571532581e-06, + "loss": 0.0358, + "step": 12926 + }, + { + "epoch": 2.35, + "learning_rate": 2.296851110302148e-06, + "loss": 0.8808, + "step": 12927 + }, + { + "epoch": 2.35, + "learning_rate": 2.296796505278486e-06, + "loss": 0.8588, + "step": 12928 + }, + { + "epoch": 2.35, + "learning_rate": 2.296741900254824e-06, + "loss": 0.6192, + "step": 12929 + }, + { + "epoch": 2.35, + "learning_rate": 2.2966872952311614e-06, + "loss": 0.4484, + "step": 12930 + }, + { + "epoch": 2.35, + "learning_rate": 2.2966326902074993e-06, + "loss": 0.5184, + "step": 12931 + }, + { + "epoch": 2.35, + "learning_rate": 2.296578085183837e-06, + "loss": 1.1445, + "step": 12932 + }, + { + "epoch": 2.35, + "learning_rate": 2.2965234801601748e-06, + "loss": 0.627, + "step": 12933 + }, + { + "epoch": 2.35, + "learning_rate": 2.2964688751365127e-06, + "loss": 1.273, + "step": 12934 + }, + { + "epoch": 2.35, + "learning_rate": 2.2964142701128502e-06, + "loss": 0.1229, + "step": 12935 + }, + { + "epoch": 2.35, + "learning_rate": 2.296359665089188e-06, + "loss": 1.7374, + "step": 12936 + }, + { + "epoch": 2.35, + "learning_rate": 2.296305060065526e-06, + "loss": 0.2891, + "step": 12937 + }, + { + "epoch": 2.35, + "learning_rate": 2.2962504550418636e-06, + "loss": 0.5083, + "step": 12938 + }, + { + "epoch": 2.35, + "learning_rate": 2.2961958500182016e-06, + "loss": 0.9004, + "step": 12939 + }, + { + "epoch": 2.35, + "learning_rate": 2.2961412449945395e-06, + "loss": 0.439, + "step": 12940 + }, + { + "epoch": 2.35, + "learning_rate": 2.2960866399708775e-06, + "loss": 0.3817, + "step": 12941 + }, + { + "epoch": 2.35, + "learning_rate": 2.2960320349472154e-06, + "loss": 1.0574, + "step": 12942 + }, + { + "epoch": 2.35, + "learning_rate": 2.295977429923553e-06, + "loss": 0.9824, + "step": 12943 + }, + { + "epoch": 2.35, + "learning_rate": 2.295922824899891e-06, + "loss": 1.3842, + "step": 12944 + }, + { + "epoch": 2.35, + "learning_rate": 2.295868219876229e-06, + "loss": 0.6546, + "step": 12945 + }, + { + "epoch": 2.35, + "learning_rate": 2.2958136148525663e-06, + "loss": 4.7656, + "step": 12946 + }, + { + "epoch": 2.35, + "learning_rate": 2.2957590098289043e-06, + "loss": 0.1129, + "step": 12947 + }, + { + "epoch": 2.35, + "learning_rate": 2.2957044048052422e-06, + "loss": 0.233, + "step": 12948 + }, + { + "epoch": 2.35, + "learning_rate": 2.2956497997815798e-06, + "loss": 1.2227, + "step": 12949 + }, + { + "epoch": 2.35, + "learning_rate": 2.2955951947579177e-06, + "loss": 0.3093, + "step": 12950 + }, + { + "epoch": 2.36, + "learning_rate": 2.2955405897342556e-06, + "loss": 0.225, + "step": 12951 + }, + { + "epoch": 2.36, + "learning_rate": 2.295485984710593e-06, + "loss": 0.5707, + "step": 12952 + }, + { + "epoch": 2.36, + "learning_rate": 2.295431379686931e-06, + "loss": 0.982, + "step": 12953 + }, + { + "epoch": 2.36, + "learning_rate": 2.295376774663269e-06, + "loss": 1.117, + "step": 12954 + }, + { + "epoch": 2.36, + "learning_rate": 2.295322169639607e-06, + "loss": 0.0625, + "step": 12955 + }, + { + "epoch": 2.36, + "learning_rate": 2.295267564615945e-06, + "loss": 0.7328, + "step": 12956 + }, + { + "epoch": 2.36, + "learning_rate": 2.2952129595922825e-06, + "loss": 0.1622, + "step": 12957 + }, + { + "epoch": 2.36, + "learning_rate": 2.2951583545686204e-06, + "loss": 3.4945, + "step": 12958 + }, + { + "epoch": 2.36, + "learning_rate": 2.2951037495449583e-06, + "loss": 1.4302, + "step": 12959 + }, + { + "epoch": 2.36, + "learning_rate": 2.295049144521296e-06, + "loss": 0.0241, + "step": 12960 + }, + { + "epoch": 2.36, + "learning_rate": 2.294994539497634e-06, + "loss": 0.258, + "step": 12961 + }, + { + "epoch": 2.36, + "learning_rate": 2.2949399344739717e-06, + "loss": 0.1459, + "step": 12962 + }, + { + "epoch": 2.36, + "learning_rate": 2.2948853294503093e-06, + "loss": 0.8534, + "step": 12963 + }, + { + "epoch": 2.36, + "learning_rate": 2.294830724426647e-06, + "loss": 0.5653, + "step": 12964 + }, + { + "epoch": 2.36, + "learning_rate": 2.294776119402985e-06, + "loss": 0.5659, + "step": 12965 + }, + { + "epoch": 2.36, + "learning_rate": 2.2947215143793227e-06, + "loss": 2.27, + "step": 12966 + }, + { + "epoch": 2.36, + "learning_rate": 2.2946669093556606e-06, + "loss": 0.9985, + "step": 12967 + }, + { + "epoch": 2.36, + "learning_rate": 2.2946123043319986e-06, + "loss": 0.8792, + "step": 12968 + }, + { + "epoch": 2.36, + "learning_rate": 2.2945576993083365e-06, + "loss": 0.54, + "step": 12969 + }, + { + "epoch": 2.36, + "learning_rate": 2.2945030942846744e-06, + "loss": 0.8878, + "step": 12970 + }, + { + "epoch": 2.36, + "learning_rate": 2.294448489261012e-06, + "loss": 0.2364, + "step": 12971 + }, + { + "epoch": 2.36, + "learning_rate": 2.29439388423735e-06, + "loss": 0.3691, + "step": 12972 + }, + { + "epoch": 2.36, + "learning_rate": 2.294339279213688e-06, + "loss": 0.2558, + "step": 12973 + }, + { + "epoch": 2.36, + "learning_rate": 2.2942846741900254e-06, + "loss": 0.5833, + "step": 12974 + }, + { + "epoch": 2.36, + "learning_rate": 2.2942300691663633e-06, + "loss": 0.0251, + "step": 12975 + }, + { + "epoch": 2.36, + "learning_rate": 2.2941754641427013e-06, + "loss": 1.9579, + "step": 12976 + }, + { + "epoch": 2.36, + "learning_rate": 2.2941208591190388e-06, + "loss": 1.9849, + "step": 12977 + }, + { + "epoch": 2.36, + "learning_rate": 2.2940662540953767e-06, + "loss": 1.2205, + "step": 12978 + }, + { + "epoch": 2.36, + "learning_rate": 2.2940116490717147e-06, + "loss": 2.3518, + "step": 12979 + }, + { + "epoch": 2.36, + "learning_rate": 2.2939570440480526e-06, + "loss": 0.7856, + "step": 12980 + }, + { + "epoch": 2.36, + "learning_rate": 2.2939024390243906e-06, + "loss": 2.3499, + "step": 12981 + }, + { + "epoch": 2.36, + "learning_rate": 2.293847834000728e-06, + "loss": 0.143, + "step": 12982 + }, + { + "epoch": 2.36, + "learning_rate": 2.293793228977066e-06, + "loss": 0.6217, + "step": 12983 + }, + { + "epoch": 2.36, + "learning_rate": 2.293738623953404e-06, + "loss": 0.036, + "step": 12984 + }, + { + "epoch": 2.36, + "learning_rate": 2.2936840189297415e-06, + "loss": 1.3562, + "step": 12985 + }, + { + "epoch": 2.36, + "learning_rate": 2.2936294139060794e-06, + "loss": 0.5417, + "step": 12986 + }, + { + "epoch": 2.36, + "learning_rate": 2.2935748088824174e-06, + "loss": 0.2586, + "step": 12987 + }, + { + "epoch": 2.36, + "learning_rate": 2.293520203858755e-06, + "loss": 1.0131, + "step": 12988 + }, + { + "epoch": 2.36, + "learning_rate": 2.293465598835093e-06, + "loss": 1.0592, + "step": 12989 + }, + { + "epoch": 2.36, + "learning_rate": 2.2934109938114308e-06, + "loss": 1.9716, + "step": 12990 + }, + { + "epoch": 2.36, + "learning_rate": 2.2933563887877683e-06, + "loss": 0.6798, + "step": 12991 + }, + { + "epoch": 2.36, + "learning_rate": 2.2933017837641062e-06, + "loss": 0.6172, + "step": 12992 + }, + { + "epoch": 2.36, + "learning_rate": 2.293247178740444e-06, + "loss": 1.1723, + "step": 12993 + }, + { + "epoch": 2.36, + "learning_rate": 2.293192573716782e-06, + "loss": 0.6548, + "step": 12994 + }, + { + "epoch": 2.36, + "learning_rate": 2.29313796869312e-06, + "loss": 0.027, + "step": 12995 + }, + { + "epoch": 2.36, + "learning_rate": 2.293083363669458e-06, + "loss": 0.1153, + "step": 12996 + }, + { + "epoch": 2.36, + "learning_rate": 2.2930287586457955e-06, + "loss": 0.8729, + "step": 12997 + }, + { + "epoch": 2.36, + "learning_rate": 2.2929741536221335e-06, + "loss": 1.4801, + "step": 12998 + }, + { + "epoch": 2.36, + "learning_rate": 2.292919548598471e-06, + "loss": 1.2976, + "step": 12999 + }, + { + "epoch": 2.36, + "learning_rate": 2.292864943574809e-06, + "loss": 2.1387, + "step": 13000 + }, + { + "epoch": 2.36, + "learning_rate": 2.292810338551147e-06, + "loss": 0.7539, + "step": 13001 + }, + { + "epoch": 2.36, + "learning_rate": 2.2927557335274844e-06, + "loss": 1.7605, + "step": 13002 + }, + { + "epoch": 2.36, + "learning_rate": 2.2927011285038223e-06, + "loss": 0.2517, + "step": 13003 + }, + { + "epoch": 2.36, + "learning_rate": 2.2926465234801603e-06, + "loss": 1.9701, + "step": 13004 + }, + { + "epoch": 2.36, + "learning_rate": 2.292591918456498e-06, + "loss": 0.5201, + "step": 13005 + }, + { + "epoch": 2.37, + "learning_rate": 2.2925373134328357e-06, + "loss": 1.634, + "step": 13006 + }, + { + "epoch": 2.37, + "learning_rate": 2.2924827084091737e-06, + "loss": 0.5319, + "step": 13007 + }, + { + "epoch": 2.37, + "learning_rate": 2.2924281033855116e-06, + "loss": 1.4177, + "step": 13008 + }, + { + "epoch": 2.37, + "learning_rate": 2.2923734983618496e-06, + "loss": 1.2918, + "step": 13009 + }, + { + "epoch": 2.37, + "learning_rate": 2.292318893338187e-06, + "loss": 2.1059, + "step": 13010 + }, + { + "epoch": 2.37, + "learning_rate": 2.292264288314525e-06, + "loss": 1.2236, + "step": 13011 + }, + { + "epoch": 2.37, + "learning_rate": 2.292209683290863e-06, + "loss": 0.5306, + "step": 13012 + }, + { + "epoch": 2.37, + "learning_rate": 2.2921550782672005e-06, + "loss": 1.9555, + "step": 13013 + }, + { + "epoch": 2.37, + "learning_rate": 2.2921004732435384e-06, + "loss": 0.248, + "step": 13014 + }, + { + "epoch": 2.37, + "learning_rate": 2.2920458682198764e-06, + "loss": 1.5819, + "step": 13015 + }, + { + "epoch": 2.37, + "learning_rate": 2.291991263196214e-06, + "loss": 1.0518, + "step": 13016 + }, + { + "epoch": 2.37, + "learning_rate": 2.291936658172552e-06, + "loss": 2.001, + "step": 13017 + }, + { + "epoch": 2.37, + "learning_rate": 2.29188205314889e-06, + "loss": 0.8241, + "step": 13018 + }, + { + "epoch": 2.37, + "learning_rate": 2.2918274481252273e-06, + "loss": 0.3036, + "step": 13019 + }, + { + "epoch": 2.37, + "learning_rate": 2.2917728431015653e-06, + "loss": 1.2132, + "step": 13020 + }, + { + "epoch": 2.37, + "learning_rate": 2.291718238077903e-06, + "loss": 1.6508, + "step": 13021 + }, + { + "epoch": 2.37, + "learning_rate": 2.291663633054241e-06, + "loss": 2.4899, + "step": 13022 + }, + { + "epoch": 2.37, + "learning_rate": 2.291609028030579e-06, + "loss": 0.886, + "step": 13023 + }, + { + "epoch": 2.37, + "learning_rate": 2.2915544230069166e-06, + "loss": 1.6117, + "step": 13024 + }, + { + "epoch": 2.37, + "learning_rate": 2.2914998179832546e-06, + "loss": 0.2469, + "step": 13025 + }, + { + "epoch": 2.37, + "learning_rate": 2.2914452129595925e-06, + "loss": 0.0243, + "step": 13026 + }, + { + "epoch": 2.37, + "learning_rate": 2.29139060793593e-06, + "loss": 0.4401, + "step": 13027 + }, + { + "epoch": 2.37, + "learning_rate": 2.291336002912268e-06, + "loss": 0.803, + "step": 13028 + }, + { + "epoch": 2.37, + "learning_rate": 2.291281397888606e-06, + "loss": 0.5831, + "step": 13029 + }, + { + "epoch": 2.37, + "learning_rate": 2.2912267928649434e-06, + "loss": 0.6543, + "step": 13030 + }, + { + "epoch": 2.37, + "learning_rate": 2.2911721878412814e-06, + "loss": 0.7639, + "step": 13031 + }, + { + "epoch": 2.37, + "learning_rate": 2.2911175828176193e-06, + "loss": 0.4778, + "step": 13032 + }, + { + "epoch": 2.37, + "learning_rate": 2.291062977793957e-06, + "loss": 0.46, + "step": 13033 + }, + { + "epoch": 2.37, + "learning_rate": 2.2910083727702948e-06, + "loss": 0.2488, + "step": 13034 + }, + { + "epoch": 2.37, + "learning_rate": 2.2909537677466327e-06, + "loss": 0.7065, + "step": 13035 + }, + { + "epoch": 2.37, + "learning_rate": 2.2908991627229707e-06, + "loss": 0.4579, + "step": 13036 + }, + { + "epoch": 2.37, + "learning_rate": 2.2908445576993086e-06, + "loss": 0.4223, + "step": 13037 + }, + { + "epoch": 2.37, + "learning_rate": 2.290789952675646e-06, + "loss": 0.037, + "step": 13038 + }, + { + "epoch": 2.37, + "learning_rate": 2.290735347651984e-06, + "loss": 0.9907, + "step": 13039 + }, + { + "epoch": 2.37, + "learning_rate": 2.290680742628322e-06, + "loss": 0.2249, + "step": 13040 + }, + { + "epoch": 2.37, + "learning_rate": 2.2906261376046595e-06, + "loss": 0.3842, + "step": 13041 + }, + { + "epoch": 2.37, + "learning_rate": 2.2905715325809975e-06, + "loss": 2.5876, + "step": 13042 + }, + { + "epoch": 2.37, + "learning_rate": 2.2905169275573354e-06, + "loss": 1.1247, + "step": 13043 + }, + { + "epoch": 2.37, + "learning_rate": 2.290462322533673e-06, + "loss": 0.6566, + "step": 13044 + }, + { + "epoch": 2.37, + "learning_rate": 2.290407717510011e-06, + "loss": 1.571, + "step": 13045 + }, + { + "epoch": 2.37, + "learning_rate": 2.290353112486349e-06, + "loss": 0.0326, + "step": 13046 + }, + { + "epoch": 2.37, + "learning_rate": 2.2902985074626863e-06, + "loss": 0.1675, + "step": 13047 + }, + { + "epoch": 2.37, + "learning_rate": 2.2902439024390243e-06, + "loss": 0.0212, + "step": 13048 + }, + { + "epoch": 2.37, + "learning_rate": 2.2901892974153622e-06, + "loss": 3.0512, + "step": 13049 + }, + { + "epoch": 2.37, + "learning_rate": 2.2901346923917e-06, + "loss": 0.3729, + "step": 13050 + }, + { + "epoch": 2.37, + "learning_rate": 2.290080087368038e-06, + "loss": 0.9843, + "step": 13051 + }, + { + "epoch": 2.37, + "learning_rate": 2.2900254823443756e-06, + "loss": 0.6328, + "step": 13052 + }, + { + "epoch": 2.37, + "learning_rate": 2.2899708773207136e-06, + "loss": 0.4412, + "step": 13053 + }, + { + "epoch": 2.37, + "learning_rate": 2.2899162722970515e-06, + "loss": 0.1773, + "step": 13054 + }, + { + "epoch": 2.37, + "learning_rate": 2.289861667273389e-06, + "loss": 0.4229, + "step": 13055 + }, + { + "epoch": 2.37, + "learning_rate": 2.289807062249727e-06, + "loss": 0.2643, + "step": 13056 + }, + { + "epoch": 2.37, + "learning_rate": 2.289752457226065e-06, + "loss": 0.6781, + "step": 13057 + }, + { + "epoch": 2.37, + "learning_rate": 2.2896978522024025e-06, + "loss": 0.2913, + "step": 13058 + }, + { + "epoch": 2.37, + "learning_rate": 2.2896432471787404e-06, + "loss": 0.1997, + "step": 13059 + }, + { + "epoch": 2.37, + "learning_rate": 2.2895886421550783e-06, + "loss": 0.489, + "step": 13060 + }, + { + "epoch": 2.38, + "learning_rate": 2.289534037131416e-06, + "loss": 0.1324, + "step": 13061 + }, + { + "epoch": 2.38, + "learning_rate": 2.289479432107754e-06, + "loss": 1.8698, + "step": 13062 + }, + { + "epoch": 2.38, + "learning_rate": 2.2894248270840917e-06, + "loss": 0.9275, + "step": 13063 + }, + { + "epoch": 2.38, + "learning_rate": 2.2893702220604297e-06, + "loss": 0.0219, + "step": 13064 + }, + { + "epoch": 2.38, + "learning_rate": 2.2893156170367676e-06, + "loss": 0.0277, + "step": 13065 + }, + { + "epoch": 2.38, + "learning_rate": 2.289261012013105e-06, + "loss": 1.2536, + "step": 13066 + }, + { + "epoch": 2.38, + "learning_rate": 2.289206406989443e-06, + "loss": 1.9665, + "step": 13067 + }, + { + "epoch": 2.38, + "learning_rate": 2.289151801965781e-06, + "loss": 1.5436, + "step": 13068 + }, + { + "epoch": 2.38, + "learning_rate": 2.2890971969421186e-06, + "loss": 1.0278, + "step": 13069 + }, + { + "epoch": 2.38, + "learning_rate": 2.2890425919184565e-06, + "loss": 0.671, + "step": 13070 + }, + { + "epoch": 2.38, + "learning_rate": 2.2889879868947944e-06, + "loss": 0.5526, + "step": 13071 + }, + { + "epoch": 2.38, + "learning_rate": 2.288933381871132e-06, + "loss": 2.5419, + "step": 13072 + }, + { + "epoch": 2.38, + "learning_rate": 2.28887877684747e-06, + "loss": 4.0178, + "step": 13073 + }, + { + "epoch": 2.38, + "learning_rate": 2.288824171823808e-06, + "loss": 0.3874, + "step": 13074 + }, + { + "epoch": 2.38, + "learning_rate": 2.288769566800146e-06, + "loss": 1.0677, + "step": 13075 + }, + { + "epoch": 2.38, + "learning_rate": 2.2887149617764837e-06, + "loss": 1.7519, + "step": 13076 + }, + { + "epoch": 2.38, + "learning_rate": 2.2886603567528217e-06, + "loss": 0.0304, + "step": 13077 + }, + { + "epoch": 2.38, + "learning_rate": 2.288605751729159e-06, + "loss": 1.3733, + "step": 13078 + }, + { + "epoch": 2.38, + "learning_rate": 2.288551146705497e-06, + "loss": 0.3611, + "step": 13079 + }, + { + "epoch": 2.38, + "learning_rate": 2.2884965416818347e-06, + "loss": 1.1145, + "step": 13080 + }, + { + "epoch": 2.38, + "learning_rate": 2.2884419366581726e-06, + "loss": 0.2075, + "step": 13081 + }, + { + "epoch": 2.38, + "learning_rate": 2.2883873316345106e-06, + "loss": 0.8505, + "step": 13082 + }, + { + "epoch": 2.38, + "learning_rate": 2.288332726610848e-06, + "loss": 0.0657, + "step": 13083 + }, + { + "epoch": 2.38, + "learning_rate": 2.288278121587186e-06, + "loss": 0.6429, + "step": 13084 + }, + { + "epoch": 2.38, + "learning_rate": 2.288223516563524e-06, + "loss": 1.3478, + "step": 13085 + }, + { + "epoch": 2.38, + "learning_rate": 2.2881689115398615e-06, + "loss": 0.3715, + "step": 13086 + }, + { + "epoch": 2.38, + "learning_rate": 2.2881143065161994e-06, + "loss": 3.6211, + "step": 13087 + }, + { + "epoch": 2.38, + "learning_rate": 2.2880597014925374e-06, + "loss": 0.0438, + "step": 13088 + }, + { + "epoch": 2.38, + "learning_rate": 2.2880050964688753e-06, + "loss": 0.0369, + "step": 13089 + }, + { + "epoch": 2.38, + "learning_rate": 2.2879504914452133e-06, + "loss": 0.2762, + "step": 13090 + }, + { + "epoch": 2.38, + "learning_rate": 2.2878958864215508e-06, + "loss": 0.2398, + "step": 13091 + }, + { + "epoch": 2.38, + "learning_rate": 2.2878412813978887e-06, + "loss": 0.9891, + "step": 13092 + }, + { + "epoch": 2.38, + "learning_rate": 2.2877866763742267e-06, + "loss": 0.6377, + "step": 13093 + }, + { + "epoch": 2.38, + "learning_rate": 2.287732071350564e-06, + "loss": 0.4543, + "step": 13094 + }, + { + "epoch": 2.38, + "learning_rate": 2.287677466326902e-06, + "loss": 0.9249, + "step": 13095 + }, + { + "epoch": 2.38, + "learning_rate": 2.28762286130324e-06, + "loss": 0.8667, + "step": 13096 + }, + { + "epoch": 2.38, + "learning_rate": 2.2875682562795776e-06, + "loss": 0.0522, + "step": 13097 + }, + { + "epoch": 2.38, + "learning_rate": 2.2875136512559155e-06, + "loss": 0.0408, + "step": 13098 + }, + { + "epoch": 2.38, + "learning_rate": 2.2874590462322535e-06, + "loss": 2.3279, + "step": 13099 + }, + { + "epoch": 2.38, + "learning_rate": 2.287404441208591e-06, + "loss": 0.27, + "step": 13100 + }, + { + "epoch": 2.38, + "learning_rate": 2.287349836184929e-06, + "loss": 1.9093, + "step": 13101 + }, + { + "epoch": 2.38, + "learning_rate": 2.287295231161267e-06, + "loss": 0.8908, + "step": 13102 + }, + { + "epoch": 2.38, + "learning_rate": 2.287240626137605e-06, + "loss": 0.0199, + "step": 13103 + }, + { + "epoch": 2.38, + "learning_rate": 2.2871860211139428e-06, + "loss": 1.863, + "step": 13104 + }, + { + "epoch": 2.38, + "learning_rate": 2.2871314160902803e-06, + "loss": 2.3847, + "step": 13105 + }, + { + "epoch": 2.38, + "learning_rate": 2.2870768110666182e-06, + "loss": 1.4127, + "step": 13106 + }, + { + "epoch": 2.38, + "learning_rate": 2.287022206042956e-06, + "loss": 0.4761, + "step": 13107 + }, + { + "epoch": 2.38, + "learning_rate": 2.2869676010192937e-06, + "loss": 1.4352, + "step": 13108 + }, + { + "epoch": 2.38, + "learning_rate": 2.2869129959956316e-06, + "loss": 2.268, + "step": 13109 + }, + { + "epoch": 2.38, + "learning_rate": 2.2868583909719696e-06, + "loss": 1.0823, + "step": 13110 + }, + { + "epoch": 2.38, + "learning_rate": 2.286803785948307e-06, + "loss": 3.1625, + "step": 13111 + }, + { + "epoch": 2.38, + "learning_rate": 2.286749180924645e-06, + "loss": 0.6189, + "step": 13112 + }, + { + "epoch": 2.38, + "learning_rate": 2.286694575900983e-06, + "loss": 2.6601, + "step": 13113 + }, + { + "epoch": 2.38, + "learning_rate": 2.2866399708773205e-06, + "loss": 1.1854, + "step": 13114 + }, + { + "epoch": 2.38, + "learning_rate": 2.2865853658536584e-06, + "loss": 1.8338, + "step": 13115 + }, + { + "epoch": 2.39, + "learning_rate": 2.2865307608299964e-06, + "loss": 0.3463, + "step": 13116 + }, + { + "epoch": 2.39, + "learning_rate": 2.2864761558063343e-06, + "loss": 0.9111, + "step": 13117 + }, + { + "epoch": 2.39, + "learning_rate": 2.2864215507826723e-06, + "loss": 1.4841, + "step": 13118 + }, + { + "epoch": 2.39, + "learning_rate": 2.28636694575901e-06, + "loss": 0.523, + "step": 13119 + }, + { + "epoch": 2.39, + "learning_rate": 2.2863123407353477e-06, + "loss": 0.305, + "step": 13120 + }, + { + "epoch": 2.39, + "learning_rate": 2.2862577357116857e-06, + "loss": 2.1147, + "step": 13121 + }, + { + "epoch": 2.39, + "learning_rate": 2.286203130688023e-06, + "loss": 1.7421, + "step": 13122 + }, + { + "epoch": 2.39, + "learning_rate": 2.286148525664361e-06, + "loss": 0.0488, + "step": 13123 + }, + { + "epoch": 2.39, + "learning_rate": 2.286093920640699e-06, + "loss": 0.0751, + "step": 13124 + }, + { + "epoch": 2.39, + "learning_rate": 2.2860393156170366e-06, + "loss": 0.8705, + "step": 13125 + }, + { + "epoch": 2.39, + "learning_rate": 2.2859847105933746e-06, + "loss": 0.2732, + "step": 13126 + }, + { + "epoch": 2.39, + "learning_rate": 2.2859301055697125e-06, + "loss": 2.6021, + "step": 13127 + }, + { + "epoch": 2.39, + "learning_rate": 2.28587550054605e-06, + "loss": 1.0907, + "step": 13128 + }, + { + "epoch": 2.39, + "learning_rate": 2.285820895522388e-06, + "loss": 0.895, + "step": 13129 + }, + { + "epoch": 2.39, + "learning_rate": 2.285766290498726e-06, + "loss": 0.1926, + "step": 13130 + }, + { + "epoch": 2.39, + "learning_rate": 2.285711685475064e-06, + "loss": 0.9608, + "step": 13131 + }, + { + "epoch": 2.39, + "learning_rate": 2.285657080451402e-06, + "loss": 0.4244, + "step": 13132 + }, + { + "epoch": 2.39, + "learning_rate": 2.2856024754277393e-06, + "loss": 0.3955, + "step": 13133 + }, + { + "epoch": 2.39, + "learning_rate": 2.2855478704040773e-06, + "loss": 0.0714, + "step": 13134 + }, + { + "epoch": 2.39, + "learning_rate": 2.285493265380415e-06, + "loss": 0.9474, + "step": 13135 + }, + { + "epoch": 2.39, + "learning_rate": 2.2854386603567527e-06, + "loss": 1.6401, + "step": 13136 + }, + { + "epoch": 2.39, + "learning_rate": 2.2853840553330907e-06, + "loss": 0.7202, + "step": 13137 + }, + { + "epoch": 2.39, + "learning_rate": 2.2853294503094286e-06, + "loss": 0.8397, + "step": 13138 + }, + { + "epoch": 2.39, + "learning_rate": 2.285274845285766e-06, + "loss": 2.7092, + "step": 13139 + }, + { + "epoch": 2.39, + "learning_rate": 2.285220240262104e-06, + "loss": 1.8494, + "step": 13140 + }, + { + "epoch": 2.39, + "learning_rate": 2.285165635238442e-06, + "loss": 1.109, + "step": 13141 + }, + { + "epoch": 2.39, + "learning_rate": 2.2851110302147795e-06, + "loss": 1.1464, + "step": 13142 + }, + { + "epoch": 2.39, + "learning_rate": 2.2850564251911175e-06, + "loss": 1.5268, + "step": 13143 + }, + { + "epoch": 2.39, + "learning_rate": 2.2850018201674554e-06, + "loss": 0.5856, + "step": 13144 + }, + { + "epoch": 2.39, + "learning_rate": 2.2849472151437934e-06, + "loss": 1.131, + "step": 13145 + }, + { + "epoch": 2.39, + "learning_rate": 2.2848926101201313e-06, + "loss": 0.2519, + "step": 13146 + }, + { + "epoch": 2.39, + "learning_rate": 2.284838005096469e-06, + "loss": 0.391, + "step": 13147 + }, + { + "epoch": 2.39, + "learning_rate": 2.2847834000728068e-06, + "loss": 0.279, + "step": 13148 + }, + { + "epoch": 2.39, + "learning_rate": 2.2847287950491447e-06, + "loss": 2.4563, + "step": 13149 + }, + { + "epoch": 2.39, + "learning_rate": 2.2846741900254822e-06, + "loss": 0.6587, + "step": 13150 + }, + { + "epoch": 2.39, + "learning_rate": 2.28461958500182e-06, + "loss": 0.659, + "step": 13151 + }, + { + "epoch": 2.39, + "learning_rate": 2.284564979978158e-06, + "loss": 0.8817, + "step": 13152 + }, + { + "epoch": 2.39, + "learning_rate": 2.2845103749544956e-06, + "loss": 0.4342, + "step": 13153 + }, + { + "epoch": 2.39, + "learning_rate": 2.2844557699308336e-06, + "loss": 0.7809, + "step": 13154 + }, + { + "epoch": 2.39, + "learning_rate": 2.2844011649071715e-06, + "loss": 0.7514, + "step": 13155 + }, + { + "epoch": 2.39, + "learning_rate": 2.2843465598835095e-06, + "loss": 1.3358, + "step": 13156 + }, + { + "epoch": 2.39, + "learning_rate": 2.2842919548598474e-06, + "loss": 0.4224, + "step": 13157 + }, + { + "epoch": 2.39, + "learning_rate": 2.2842373498361854e-06, + "loss": 0.4747, + "step": 13158 + }, + { + "epoch": 2.39, + "learning_rate": 2.284182744812523e-06, + "loss": 0.0142, + "step": 13159 + }, + { + "epoch": 2.39, + "learning_rate": 2.284128139788861e-06, + "loss": 0.476, + "step": 13160 + }, + { + "epoch": 2.39, + "learning_rate": 2.2840735347651983e-06, + "loss": 0.3839, + "step": 13161 + }, + { + "epoch": 2.39, + "learning_rate": 2.2840189297415363e-06, + "loss": 0.392, + "step": 13162 + }, + { + "epoch": 2.39, + "learning_rate": 2.2839643247178742e-06, + "loss": 0.4589, + "step": 13163 + }, + { + "epoch": 2.39, + "learning_rate": 2.2839097196942117e-06, + "loss": 0.2613, + "step": 13164 + }, + { + "epoch": 2.39, + "learning_rate": 2.2838551146705497e-06, + "loss": 0.9233, + "step": 13165 + }, + { + "epoch": 2.39, + "learning_rate": 2.2838005096468876e-06, + "loss": 0.0881, + "step": 13166 + }, + { + "epoch": 2.39, + "learning_rate": 2.283745904623225e-06, + "loss": 2.9552, + "step": 13167 + }, + { + "epoch": 2.39, + "learning_rate": 2.283691299599563e-06, + "loss": 1.9541, + "step": 13168 + }, + { + "epoch": 2.39, + "learning_rate": 2.283636694575901e-06, + "loss": 0.0422, + "step": 13169 + }, + { + "epoch": 2.39, + "learning_rate": 2.283582089552239e-06, + "loss": 0.2833, + "step": 13170 + }, + { + "epoch": 2.4, + "learning_rate": 2.283527484528577e-06, + "loss": 0.6188, + "step": 13171 + }, + { + "epoch": 2.4, + "learning_rate": 2.2834728795049144e-06, + "loss": 1.0212, + "step": 13172 + }, + { + "epoch": 2.4, + "learning_rate": 2.2834182744812524e-06, + "loss": 0.1304, + "step": 13173 + }, + { + "epoch": 2.4, + "learning_rate": 2.2833636694575903e-06, + "loss": 1.1304, + "step": 13174 + }, + { + "epoch": 2.4, + "learning_rate": 2.283309064433928e-06, + "loss": 2.386, + "step": 13175 + }, + { + "epoch": 2.4, + "learning_rate": 2.283254459410266e-06, + "loss": 1.632, + "step": 13176 + }, + { + "epoch": 2.4, + "learning_rate": 2.2831998543866037e-06, + "loss": 1.4791, + "step": 13177 + }, + { + "epoch": 2.4, + "learning_rate": 2.2831452493629413e-06, + "loss": 0.2677, + "step": 13178 + }, + { + "epoch": 2.4, + "learning_rate": 2.283090644339279e-06, + "loss": 3.0772, + "step": 13179 + }, + { + "epoch": 2.4, + "learning_rate": 2.283036039315617e-06, + "loss": 2.2945, + "step": 13180 + }, + { + "epoch": 2.4, + "learning_rate": 2.2829814342919547e-06, + "loss": 0.1476, + "step": 13181 + }, + { + "epoch": 2.4, + "learning_rate": 2.2829268292682926e-06, + "loss": 0.3534, + "step": 13182 + }, + { + "epoch": 2.4, + "learning_rate": 2.2828722242446306e-06, + "loss": 1.6397, + "step": 13183 + }, + { + "epoch": 2.4, + "learning_rate": 2.2828176192209685e-06, + "loss": 1.1119, + "step": 13184 + }, + { + "epoch": 2.4, + "learning_rate": 2.2827630141973064e-06, + "loss": 2.2245, + "step": 13185 + }, + { + "epoch": 2.4, + "learning_rate": 2.282708409173644e-06, + "loss": 0.7584, + "step": 13186 + }, + { + "epoch": 2.4, + "learning_rate": 2.282653804149982e-06, + "loss": 0.4221, + "step": 13187 + }, + { + "epoch": 2.4, + "learning_rate": 2.28259919912632e-06, + "loss": 0.5508, + "step": 13188 + }, + { + "epoch": 2.4, + "learning_rate": 2.2825445941026574e-06, + "loss": 0.0441, + "step": 13189 + }, + { + "epoch": 2.4, + "learning_rate": 2.2824899890789953e-06, + "loss": 0.2917, + "step": 13190 + }, + { + "epoch": 2.4, + "learning_rate": 2.2824353840553333e-06, + "loss": 1.3351, + "step": 13191 + }, + { + "epoch": 2.4, + "learning_rate": 2.2823807790316708e-06, + "loss": 1.7485, + "step": 13192 + }, + { + "epoch": 2.4, + "learning_rate": 2.2823261740080087e-06, + "loss": 1.1084, + "step": 13193 + }, + { + "epoch": 2.4, + "learning_rate": 2.2822715689843467e-06, + "loss": 0.7317, + "step": 13194 + }, + { + "epoch": 2.4, + "learning_rate": 2.282216963960684e-06, + "loss": 0.3912, + "step": 13195 + }, + { + "epoch": 2.4, + "learning_rate": 2.282162358937022e-06, + "loss": 0.6451, + "step": 13196 + }, + { + "epoch": 2.4, + "learning_rate": 2.28210775391336e-06, + "loss": 0.2431, + "step": 13197 + }, + { + "epoch": 2.4, + "learning_rate": 2.282053148889698e-06, + "loss": 0.648, + "step": 13198 + }, + { + "epoch": 2.4, + "learning_rate": 2.281998543866036e-06, + "loss": 1.463, + "step": 13199 + }, + { + "epoch": 2.4, + "learning_rate": 2.2819439388423735e-06, + "loss": 1.7929, + "step": 13200 + }, + { + "epoch": 2.4, + "learning_rate": 2.2818893338187114e-06, + "loss": 0.1778, + "step": 13201 + }, + { + "epoch": 2.4, + "learning_rate": 2.2818347287950494e-06, + "loss": 0.7253, + "step": 13202 + }, + { + "epoch": 2.4, + "learning_rate": 2.281780123771387e-06, + "loss": 0.1401, + "step": 13203 + }, + { + "epoch": 2.4, + "learning_rate": 2.281725518747725e-06, + "loss": 0.6984, + "step": 13204 + }, + { + "epoch": 2.4, + "learning_rate": 2.2816709137240628e-06, + "loss": 2.1166, + "step": 13205 + }, + { + "epoch": 2.4, + "learning_rate": 2.2816163087004003e-06, + "loss": 0.3465, + "step": 13206 + }, + { + "epoch": 2.4, + "learning_rate": 2.2815617036767382e-06, + "loss": 1.7169, + "step": 13207 + }, + { + "epoch": 2.4, + "learning_rate": 2.281507098653076e-06, + "loss": 1.8375, + "step": 13208 + }, + { + "epoch": 2.4, + "learning_rate": 2.2814524936294137e-06, + "loss": 0.0213, + "step": 13209 + }, + { + "epoch": 2.4, + "learning_rate": 2.2813978886057516e-06, + "loss": 0.0818, + "step": 13210 + }, + { + "epoch": 2.4, + "learning_rate": 2.2813432835820896e-06, + "loss": 0.4072, + "step": 13211 + }, + { + "epoch": 2.4, + "learning_rate": 2.2812886785584275e-06, + "loss": 0.4367, + "step": 13212 + }, + { + "epoch": 2.4, + "learning_rate": 2.2812340735347655e-06, + "loss": 1.7333, + "step": 13213 + }, + { + "epoch": 2.4, + "learning_rate": 2.281179468511103e-06, + "loss": 0.0486, + "step": 13214 + }, + { + "epoch": 2.4, + "learning_rate": 2.281124863487441e-06, + "loss": 0.2436, + "step": 13215 + }, + { + "epoch": 2.4, + "learning_rate": 2.281070258463779e-06, + "loss": 0.8938, + "step": 13216 + }, + { + "epoch": 2.4, + "learning_rate": 2.2810156534401164e-06, + "loss": 2.5119, + "step": 13217 + }, + { + "epoch": 2.4, + "learning_rate": 2.2809610484164543e-06, + "loss": 0.7204, + "step": 13218 + }, + { + "epoch": 2.4, + "learning_rate": 2.2809064433927923e-06, + "loss": 1.0222, + "step": 13219 + }, + { + "epoch": 2.4, + "learning_rate": 2.28085183836913e-06, + "loss": 1.1076, + "step": 13220 + }, + { + "epoch": 2.4, + "learning_rate": 2.2807972333454677e-06, + "loss": 2.087, + "step": 13221 + }, + { + "epoch": 2.4, + "learning_rate": 2.2807426283218057e-06, + "loss": 1.1515, + "step": 13222 + }, + { + "epoch": 2.4, + "learning_rate": 2.280688023298143e-06, + "loss": 0.6293, + "step": 13223 + }, + { + "epoch": 2.4, + "learning_rate": 2.280633418274481e-06, + "loss": 1.29, + "step": 13224 + }, + { + "epoch": 2.4, + "learning_rate": 2.280578813250819e-06, + "loss": 0.4837, + "step": 13225 + }, + { + "epoch": 2.41, + "learning_rate": 2.280524208227157e-06, + "loss": 1.5046, + "step": 13226 + }, + { + "epoch": 2.41, + "learning_rate": 2.280469603203495e-06, + "loss": 0.9234, + "step": 13227 + }, + { + "epoch": 2.41, + "learning_rate": 2.2804149981798325e-06, + "loss": 1.7516, + "step": 13228 + }, + { + "epoch": 2.41, + "learning_rate": 2.2803603931561704e-06, + "loss": 0.5631, + "step": 13229 + }, + { + "epoch": 2.41, + "learning_rate": 2.2803057881325084e-06, + "loss": 0.3909, + "step": 13230 + }, + { + "epoch": 2.41, + "learning_rate": 2.280251183108846e-06, + "loss": 0.1834, + "step": 13231 + }, + { + "epoch": 2.41, + "learning_rate": 2.280196578085184e-06, + "loss": 0.6397, + "step": 13232 + }, + { + "epoch": 2.41, + "learning_rate": 2.280141973061522e-06, + "loss": 0.3566, + "step": 13233 + }, + { + "epoch": 2.41, + "learning_rate": 2.2800873680378593e-06, + "loss": 1.484, + "step": 13234 + }, + { + "epoch": 2.41, + "learning_rate": 2.2800327630141973e-06, + "loss": 0.1684, + "step": 13235 + }, + { + "epoch": 2.41, + "learning_rate": 2.279978157990535e-06, + "loss": 0.0996, + "step": 13236 + }, + { + "epoch": 2.41, + "learning_rate": 2.279923552966873e-06, + "loss": 3.4235, + "step": 13237 + }, + { + "epoch": 2.41, + "learning_rate": 2.279868947943211e-06, + "loss": 2.2747, + "step": 13238 + }, + { + "epoch": 2.41, + "learning_rate": 2.279814342919549e-06, + "loss": 0.4543, + "step": 13239 + }, + { + "epoch": 2.41, + "learning_rate": 2.2797597378958866e-06, + "loss": 0.5655, + "step": 13240 + }, + { + "epoch": 2.41, + "learning_rate": 2.2797051328722245e-06, + "loss": 0.0226, + "step": 13241 + }, + { + "epoch": 2.41, + "learning_rate": 2.279650527848562e-06, + "loss": 0.9955, + "step": 13242 + }, + { + "epoch": 2.41, + "learning_rate": 2.2795959228249e-06, + "loss": 1.0808, + "step": 13243 + }, + { + "epoch": 2.41, + "learning_rate": 2.279541317801238e-06, + "loss": 0.5318, + "step": 13244 + }, + { + "epoch": 2.41, + "learning_rate": 2.2794867127775754e-06, + "loss": 0.6715, + "step": 13245 + }, + { + "epoch": 2.41, + "learning_rate": 2.2794321077539134e-06, + "loss": 2.2396, + "step": 13246 + }, + { + "epoch": 2.41, + "learning_rate": 2.2793775027302513e-06, + "loss": 0.2521, + "step": 13247 + }, + { + "epoch": 2.41, + "learning_rate": 2.279322897706589e-06, + "loss": 1.0017, + "step": 13248 + }, + { + "epoch": 2.41, + "learning_rate": 2.2792682926829268e-06, + "loss": 1.7506, + "step": 13249 + }, + { + "epoch": 2.41, + "learning_rate": 2.2792136876592647e-06, + "loss": 0.1272, + "step": 13250 + }, + { + "epoch": 2.41, + "learning_rate": 2.2791590826356027e-06, + "loss": 0.4802, + "step": 13251 + }, + { + "epoch": 2.41, + "learning_rate": 2.2791044776119406e-06, + "loss": 0.3831, + "step": 13252 + }, + { + "epoch": 2.41, + "learning_rate": 2.279049872588278e-06, + "loss": 0.1175, + "step": 13253 + }, + { + "epoch": 2.41, + "learning_rate": 2.278995267564616e-06, + "loss": 0.6554, + "step": 13254 + }, + { + "epoch": 2.41, + "learning_rate": 2.278940662540954e-06, + "loss": 1.0069, + "step": 13255 + }, + { + "epoch": 2.41, + "learning_rate": 2.2788860575172915e-06, + "loss": 0.478, + "step": 13256 + }, + { + "epoch": 2.41, + "learning_rate": 2.2788314524936295e-06, + "loss": 0.9368, + "step": 13257 + }, + { + "epoch": 2.41, + "learning_rate": 2.2787768474699674e-06, + "loss": 1.0511, + "step": 13258 + }, + { + "epoch": 2.41, + "learning_rate": 2.278722242446305e-06, + "loss": 1.9683, + "step": 13259 + }, + { + "epoch": 2.41, + "learning_rate": 2.278667637422643e-06, + "loss": 2.6948, + "step": 13260 + }, + { + "epoch": 2.41, + "learning_rate": 2.278613032398981e-06, + "loss": 0.6434, + "step": 13261 + }, + { + "epoch": 2.41, + "learning_rate": 2.2785584273753183e-06, + "loss": 0.0158, + "step": 13262 + }, + { + "epoch": 2.41, + "learning_rate": 2.2785038223516563e-06, + "loss": 1.3598, + "step": 13263 + }, + { + "epoch": 2.41, + "learning_rate": 2.2784492173279942e-06, + "loss": 1.8153, + "step": 13264 + }, + { + "epoch": 2.41, + "learning_rate": 2.278394612304332e-06, + "loss": 0.705, + "step": 13265 + }, + { + "epoch": 2.41, + "learning_rate": 2.27834000728067e-06, + "loss": 0.4106, + "step": 13266 + }, + { + "epoch": 2.41, + "learning_rate": 2.2782854022570076e-06, + "loss": 1.6081, + "step": 13267 + }, + { + "epoch": 2.41, + "learning_rate": 2.2782307972333456e-06, + "loss": 1.1123, + "step": 13268 + }, + { + "epoch": 2.41, + "learning_rate": 2.2781761922096835e-06, + "loss": 0.9989, + "step": 13269 + }, + { + "epoch": 2.41, + "learning_rate": 2.278121587186021e-06, + "loss": 1.6978, + "step": 13270 + }, + { + "epoch": 2.41, + "learning_rate": 2.278066982162359e-06, + "loss": 0.3708, + "step": 13271 + }, + { + "epoch": 2.41, + "learning_rate": 2.278012377138697e-06, + "loss": 0.0842, + "step": 13272 + }, + { + "epoch": 2.41, + "learning_rate": 2.2779577721150344e-06, + "loss": 0.8565, + "step": 13273 + }, + { + "epoch": 2.41, + "learning_rate": 2.2779031670913724e-06, + "loss": 0.9877, + "step": 13274 + }, + { + "epoch": 2.41, + "learning_rate": 2.2778485620677103e-06, + "loss": 2.0594, + "step": 13275 + }, + { + "epoch": 2.41, + "learning_rate": 2.277793957044048e-06, + "loss": 0.8042, + "step": 13276 + }, + { + "epoch": 2.41, + "learning_rate": 2.277739352020386e-06, + "loss": 1.7195, + "step": 13277 + }, + { + "epoch": 2.41, + "learning_rate": 2.2776847469967237e-06, + "loss": 3.4926, + "step": 13278 + }, + { + "epoch": 2.41, + "learning_rate": 2.2776301419730617e-06, + "loss": 0.654, + "step": 13279 + }, + { + "epoch": 2.41, + "learning_rate": 2.2775755369493996e-06, + "loss": 0.3151, + "step": 13280 + }, + { + "epoch": 2.42, + "learning_rate": 2.277520931925737e-06, + "loss": 0.6646, + "step": 13281 + }, + { + "epoch": 2.42, + "learning_rate": 2.277466326902075e-06, + "loss": 1.3842, + "step": 13282 + }, + { + "epoch": 2.42, + "learning_rate": 2.277411721878413e-06, + "loss": 1.2296, + "step": 13283 + }, + { + "epoch": 2.42, + "learning_rate": 2.2773571168547506e-06, + "loss": 0.7953, + "step": 13284 + }, + { + "epoch": 2.42, + "learning_rate": 2.2773025118310885e-06, + "loss": 3.9704, + "step": 13285 + }, + { + "epoch": 2.42, + "learning_rate": 2.2772479068074264e-06, + "loss": 0.6433, + "step": 13286 + }, + { + "epoch": 2.42, + "learning_rate": 2.277193301783764e-06, + "loss": 0.6551, + "step": 13287 + }, + { + "epoch": 2.42, + "learning_rate": 2.277138696760102e-06, + "loss": 0.8464, + "step": 13288 + }, + { + "epoch": 2.42, + "learning_rate": 2.27708409173644e-06, + "loss": 0.3446, + "step": 13289 + }, + { + "epoch": 2.42, + "learning_rate": 2.2770294867127774e-06, + "loss": 2.027, + "step": 13290 + }, + { + "epoch": 2.42, + "learning_rate": 2.2769748816891153e-06, + "loss": 0.1883, + "step": 13291 + }, + { + "epoch": 2.42, + "learning_rate": 2.2769202766654533e-06, + "loss": 0.2749, + "step": 13292 + }, + { + "epoch": 2.42, + "learning_rate": 2.276865671641791e-06, + "loss": 0.3313, + "step": 13293 + }, + { + "epoch": 2.42, + "learning_rate": 2.276811066618129e-06, + "loss": 0.0265, + "step": 13294 + }, + { + "epoch": 2.42, + "learning_rate": 2.2767564615944667e-06, + "loss": 0.7852, + "step": 13295 + }, + { + "epoch": 2.42, + "learning_rate": 2.2767018565708046e-06, + "loss": 0.0169, + "step": 13296 + }, + { + "epoch": 2.42, + "learning_rate": 2.2766472515471425e-06, + "loss": 0.5201, + "step": 13297 + }, + { + "epoch": 2.42, + "learning_rate": 2.27659264652348e-06, + "loss": 0.3707, + "step": 13298 + }, + { + "epoch": 2.42, + "learning_rate": 2.276538041499818e-06, + "loss": 0.0426, + "step": 13299 + }, + { + "epoch": 2.42, + "learning_rate": 2.276483436476156e-06, + "loss": 0.8806, + "step": 13300 + }, + { + "epoch": 2.42, + "learning_rate": 2.2764288314524935e-06, + "loss": 0.3412, + "step": 13301 + }, + { + "epoch": 2.42, + "learning_rate": 2.2763742264288314e-06, + "loss": 1.5142, + "step": 13302 + }, + { + "epoch": 2.42, + "learning_rate": 2.2763196214051694e-06, + "loss": 1.256, + "step": 13303 + }, + { + "epoch": 2.42, + "learning_rate": 2.276265016381507e-06, + "loss": 2.1908, + "step": 13304 + }, + { + "epoch": 2.42, + "learning_rate": 2.276210411357845e-06, + "loss": 1.6098, + "step": 13305 + }, + { + "epoch": 2.42, + "learning_rate": 2.2761558063341828e-06, + "loss": 0.4064, + "step": 13306 + }, + { + "epoch": 2.42, + "learning_rate": 2.2761012013105207e-06, + "loss": 0.5594, + "step": 13307 + }, + { + "epoch": 2.42, + "learning_rate": 2.2760465962868587e-06, + "loss": 0.0578, + "step": 13308 + }, + { + "epoch": 2.42, + "learning_rate": 2.275991991263196e-06, + "loss": 0.291, + "step": 13309 + }, + { + "epoch": 2.42, + "learning_rate": 2.275937386239534e-06, + "loss": 0.4606, + "step": 13310 + }, + { + "epoch": 2.42, + "learning_rate": 2.275882781215872e-06, + "loss": 1.0285, + "step": 13311 + }, + { + "epoch": 2.42, + "learning_rate": 2.2758281761922096e-06, + "loss": 0.4423, + "step": 13312 + }, + { + "epoch": 2.42, + "learning_rate": 2.2757735711685475e-06, + "loss": 0.2334, + "step": 13313 + }, + { + "epoch": 2.42, + "learning_rate": 2.2757189661448855e-06, + "loss": 0.0795, + "step": 13314 + }, + { + "epoch": 2.42, + "learning_rate": 2.275664361121223e-06, + "loss": 0.3883, + "step": 13315 + }, + { + "epoch": 2.42, + "learning_rate": 2.275609756097561e-06, + "loss": 0.3353, + "step": 13316 + }, + { + "epoch": 2.42, + "learning_rate": 2.275555151073899e-06, + "loss": 0.6513, + "step": 13317 + }, + { + "epoch": 2.42, + "learning_rate": 2.275500546050237e-06, + "loss": 1.5346, + "step": 13318 + }, + { + "epoch": 2.42, + "learning_rate": 2.2754459410265743e-06, + "loss": 1.4, + "step": 13319 + }, + { + "epoch": 2.42, + "learning_rate": 2.2753913360029123e-06, + "loss": 1.126, + "step": 13320 + }, + { + "epoch": 2.42, + "learning_rate": 2.2753367309792502e-06, + "loss": 0.0304, + "step": 13321 + }, + { + "epoch": 2.42, + "learning_rate": 2.275282125955588e-06, + "loss": 1.0637, + "step": 13322 + }, + { + "epoch": 2.42, + "learning_rate": 2.2752275209319257e-06, + "loss": 0.9975, + "step": 13323 + }, + { + "epoch": 2.42, + "learning_rate": 2.2751729159082636e-06, + "loss": 2.8728, + "step": 13324 + }, + { + "epoch": 2.42, + "learning_rate": 2.2751183108846016e-06, + "loss": 0.4266, + "step": 13325 + }, + { + "epoch": 2.42, + "learning_rate": 2.275063705860939e-06, + "loss": 1.8628, + "step": 13326 + }, + { + "epoch": 2.42, + "learning_rate": 2.275009100837277e-06, + "loss": 0.0285, + "step": 13327 + }, + { + "epoch": 2.42, + "learning_rate": 2.274954495813615e-06, + "loss": 1.1864, + "step": 13328 + }, + { + "epoch": 2.42, + "learning_rate": 2.2748998907899525e-06, + "loss": 0.2011, + "step": 13329 + }, + { + "epoch": 2.42, + "learning_rate": 2.2748452857662904e-06, + "loss": 0.558, + "step": 13330 + }, + { + "epoch": 2.42, + "learning_rate": 2.2747906807426284e-06, + "loss": 0.9354, + "step": 13331 + }, + { + "epoch": 2.42, + "learning_rate": 2.2747360757189663e-06, + "loss": 1.7714, + "step": 13332 + }, + { + "epoch": 2.42, + "learning_rate": 2.2746814706953043e-06, + "loss": 1.3063, + "step": 13333 + }, + { + "epoch": 2.42, + "learning_rate": 2.274626865671642e-06, + "loss": 3.4469, + "step": 13334 + }, + { + "epoch": 2.42, + "learning_rate": 2.2745722606479797e-06, + "loss": 2.1325, + "step": 13335 + }, + { + "epoch": 2.43, + "learning_rate": 2.2745176556243177e-06, + "loss": 0.5424, + "step": 13336 + }, + { + "epoch": 2.43, + "learning_rate": 2.274463050600655e-06, + "loss": 1.7042, + "step": 13337 + }, + { + "epoch": 2.43, + "learning_rate": 2.274408445576993e-06, + "loss": 0.852, + "step": 13338 + }, + { + "epoch": 2.43, + "learning_rate": 2.274353840553331e-06, + "loss": 0.1297, + "step": 13339 + }, + { + "epoch": 2.43, + "learning_rate": 2.2742992355296686e-06, + "loss": 0.0145, + "step": 13340 + }, + { + "epoch": 2.43, + "learning_rate": 2.2742446305060066e-06, + "loss": 0.7633, + "step": 13341 + }, + { + "epoch": 2.43, + "learning_rate": 2.2741900254823445e-06, + "loss": 0.3539, + "step": 13342 + }, + { + "epoch": 2.43, + "learning_rate": 2.274135420458682e-06, + "loss": 0.5172, + "step": 13343 + }, + { + "epoch": 2.43, + "learning_rate": 2.27408081543502e-06, + "loss": 1.709, + "step": 13344 + }, + { + "epoch": 2.43, + "learning_rate": 2.274026210411358e-06, + "loss": 0.5351, + "step": 13345 + }, + { + "epoch": 2.43, + "learning_rate": 2.273971605387696e-06, + "loss": 0.9235, + "step": 13346 + }, + { + "epoch": 2.43, + "learning_rate": 2.273917000364034e-06, + "loss": 1.1059, + "step": 13347 + }, + { + "epoch": 2.43, + "learning_rate": 2.2738623953403713e-06, + "loss": 1.3415, + "step": 13348 + }, + { + "epoch": 2.43, + "learning_rate": 2.2738077903167093e-06, + "loss": 0.2859, + "step": 13349 + }, + { + "epoch": 2.43, + "learning_rate": 2.273753185293047e-06, + "loss": 2.2645, + "step": 13350 + }, + { + "epoch": 2.43, + "learning_rate": 2.2736985802693847e-06, + "loss": 0.9749, + "step": 13351 + }, + { + "epoch": 2.43, + "learning_rate": 2.2736439752457227e-06, + "loss": 1.18, + "step": 13352 + }, + { + "epoch": 2.43, + "learning_rate": 2.2735893702220606e-06, + "loss": 0.2638, + "step": 13353 + }, + { + "epoch": 2.43, + "learning_rate": 2.273534765198398e-06, + "loss": 0.0274, + "step": 13354 + }, + { + "epoch": 2.43, + "learning_rate": 2.273480160174736e-06, + "loss": 0.2126, + "step": 13355 + }, + { + "epoch": 2.43, + "learning_rate": 2.273425555151074e-06, + "loss": 0.1169, + "step": 13356 + }, + { + "epoch": 2.43, + "learning_rate": 2.2733709501274115e-06, + "loss": 0.2805, + "step": 13357 + }, + { + "epoch": 2.43, + "learning_rate": 2.2733163451037495e-06, + "loss": 0.5594, + "step": 13358 + }, + { + "epoch": 2.43, + "learning_rate": 2.2732617400800874e-06, + "loss": 0.3801, + "step": 13359 + }, + { + "epoch": 2.43, + "learning_rate": 2.2732071350564254e-06, + "loss": 1.1767, + "step": 13360 + }, + { + "epoch": 2.43, + "learning_rate": 2.2731525300327633e-06, + "loss": 0.5148, + "step": 13361 + }, + { + "epoch": 2.43, + "learning_rate": 2.273097925009101e-06, + "loss": 1.7481, + "step": 13362 + }, + { + "epoch": 2.43, + "learning_rate": 2.2730433199854388e-06, + "loss": 0.4965, + "step": 13363 + }, + { + "epoch": 2.43, + "learning_rate": 2.2729887149617767e-06, + "loss": 0.0414, + "step": 13364 + }, + { + "epoch": 2.43, + "learning_rate": 2.2729341099381142e-06, + "loss": 2.0057, + "step": 13365 + }, + { + "epoch": 2.43, + "learning_rate": 2.272879504914452e-06, + "loss": 0.7108, + "step": 13366 + }, + { + "epoch": 2.43, + "learning_rate": 2.27282489989079e-06, + "loss": 0.2814, + "step": 13367 + }, + { + "epoch": 2.43, + "learning_rate": 2.2727702948671276e-06, + "loss": 1.3809, + "step": 13368 + }, + { + "epoch": 2.43, + "learning_rate": 2.2727156898434656e-06, + "loss": 1.0208, + "step": 13369 + }, + { + "epoch": 2.43, + "learning_rate": 2.2726610848198035e-06, + "loss": 0.4749, + "step": 13370 + }, + { + "epoch": 2.43, + "learning_rate": 2.272606479796141e-06, + "loss": 0.1899, + "step": 13371 + }, + { + "epoch": 2.43, + "learning_rate": 2.272551874772479e-06, + "loss": 0.3053, + "step": 13372 + }, + { + "epoch": 2.43, + "learning_rate": 2.272497269748817e-06, + "loss": 0.3752, + "step": 13373 + }, + { + "epoch": 2.43, + "learning_rate": 2.272442664725155e-06, + "loss": 0.4132, + "step": 13374 + }, + { + "epoch": 2.43, + "learning_rate": 2.272388059701493e-06, + "loss": 0.0682, + "step": 13375 + }, + { + "epoch": 2.43, + "learning_rate": 2.2723334546778303e-06, + "loss": 0.2247, + "step": 13376 + }, + { + "epoch": 2.43, + "learning_rate": 2.2722788496541683e-06, + "loss": 0.749, + "step": 13377 + }, + { + "epoch": 2.43, + "learning_rate": 2.2722242446305062e-06, + "loss": 0.38, + "step": 13378 + }, + { + "epoch": 2.43, + "learning_rate": 2.2721696396068437e-06, + "loss": 0.9011, + "step": 13379 + }, + { + "epoch": 2.43, + "learning_rate": 2.2721150345831817e-06, + "loss": 1.419, + "step": 13380 + }, + { + "epoch": 2.43, + "learning_rate": 2.2720604295595196e-06, + "loss": 0.4659, + "step": 13381 + }, + { + "epoch": 2.43, + "learning_rate": 2.272005824535857e-06, + "loss": 2.4056, + "step": 13382 + }, + { + "epoch": 2.43, + "learning_rate": 2.271951219512195e-06, + "loss": 0.1623, + "step": 13383 + }, + { + "epoch": 2.43, + "learning_rate": 2.271896614488533e-06, + "loss": 0.8184, + "step": 13384 + }, + { + "epoch": 2.43, + "learning_rate": 2.2718420094648706e-06, + "loss": 1.1128, + "step": 13385 + }, + { + "epoch": 2.43, + "learning_rate": 2.2717874044412085e-06, + "loss": 0.7109, + "step": 13386 + }, + { + "epoch": 2.43, + "learning_rate": 2.2717327994175464e-06, + "loss": 0.326, + "step": 13387 + }, + { + "epoch": 2.43, + "learning_rate": 2.2716781943938844e-06, + "loss": 1.1108, + "step": 13388 + }, + { + "epoch": 2.43, + "learning_rate": 2.2716235893702223e-06, + "loss": 0.6715, + "step": 13389 + }, + { + "epoch": 2.43, + "learning_rate": 2.27156898434656e-06, + "loss": 0.0392, + "step": 13390 + }, + { + "epoch": 2.44, + "learning_rate": 2.271514379322898e-06, + "loss": 0.6456, + "step": 13391 + }, + { + "epoch": 2.44, + "learning_rate": 2.2714597742992357e-06, + "loss": 0.4963, + "step": 13392 + }, + { + "epoch": 2.44, + "learning_rate": 2.2714051692755733e-06, + "loss": 0.4918, + "step": 13393 + }, + { + "epoch": 2.44, + "learning_rate": 2.271350564251911e-06, + "loss": 1.332, + "step": 13394 + }, + { + "epoch": 2.44, + "learning_rate": 2.271295959228249e-06, + "loss": 0.4511, + "step": 13395 + }, + { + "epoch": 2.44, + "learning_rate": 2.2712413542045867e-06, + "loss": 0.2388, + "step": 13396 + }, + { + "epoch": 2.44, + "learning_rate": 2.2711867491809246e-06, + "loss": 2.7189, + "step": 13397 + }, + { + "epoch": 2.44, + "learning_rate": 2.2711321441572625e-06, + "loss": 1.322, + "step": 13398 + }, + { + "epoch": 2.44, + "learning_rate": 2.2710775391336e-06, + "loss": 1.3529, + "step": 13399 + }, + { + "epoch": 2.44, + "learning_rate": 2.271022934109938e-06, + "loss": 1.1326, + "step": 13400 + }, + { + "epoch": 2.44, + "learning_rate": 2.270968329086276e-06, + "loss": 3.1923, + "step": 13401 + }, + { + "epoch": 2.44, + "learning_rate": 2.270913724062614e-06, + "loss": 2.1249, + "step": 13402 + }, + { + "epoch": 2.44, + "learning_rate": 2.270859119038952e-06, + "loss": 0.1073, + "step": 13403 + }, + { + "epoch": 2.44, + "learning_rate": 2.2708045140152894e-06, + "loss": 2.007, + "step": 13404 + }, + { + "epoch": 2.44, + "learning_rate": 2.2707499089916273e-06, + "loss": 1.5454, + "step": 13405 + }, + { + "epoch": 2.44, + "learning_rate": 2.2706953039679652e-06, + "loss": 0.3656, + "step": 13406 + }, + { + "epoch": 2.44, + "learning_rate": 2.2706406989443028e-06, + "loss": 1.0105, + "step": 13407 + }, + { + "epoch": 2.44, + "learning_rate": 2.2705860939206407e-06, + "loss": 0.6958, + "step": 13408 + }, + { + "epoch": 2.44, + "learning_rate": 2.2705314888969787e-06, + "loss": 0.2206, + "step": 13409 + }, + { + "epoch": 2.44, + "learning_rate": 2.270476883873316e-06, + "loss": 0.0168, + "step": 13410 + }, + { + "epoch": 2.44, + "learning_rate": 2.270422278849654e-06, + "loss": 0.6233, + "step": 13411 + }, + { + "epoch": 2.44, + "learning_rate": 2.270367673825992e-06, + "loss": 1.019, + "step": 13412 + }, + { + "epoch": 2.44, + "learning_rate": 2.27031306880233e-06, + "loss": 0.0453, + "step": 13413 + }, + { + "epoch": 2.44, + "learning_rate": 2.270258463778668e-06, + "loss": 0.0218, + "step": 13414 + }, + { + "epoch": 2.44, + "learning_rate": 2.2702038587550055e-06, + "loss": 0.3153, + "step": 13415 + }, + { + "epoch": 2.44, + "learning_rate": 2.2701492537313434e-06, + "loss": 0.8288, + "step": 13416 + }, + { + "epoch": 2.44, + "learning_rate": 2.2700946487076814e-06, + "loss": 0.4575, + "step": 13417 + }, + { + "epoch": 2.44, + "learning_rate": 2.270040043684019e-06, + "loss": 1.3959, + "step": 13418 + }, + { + "epoch": 2.44, + "learning_rate": 2.269985438660357e-06, + "loss": 0.8117, + "step": 13419 + }, + { + "epoch": 2.44, + "learning_rate": 2.2699308336366948e-06, + "loss": 0.3333, + "step": 13420 + }, + { + "epoch": 2.44, + "learning_rate": 2.2698762286130323e-06, + "loss": 1.1549, + "step": 13421 + }, + { + "epoch": 2.44, + "learning_rate": 2.2698216235893702e-06, + "loss": 0.0203, + "step": 13422 + }, + { + "epoch": 2.44, + "learning_rate": 2.269767018565708e-06, + "loss": 0.9909, + "step": 13423 + }, + { + "epoch": 2.44, + "learning_rate": 2.2697124135420457e-06, + "loss": 0.909, + "step": 13424 + }, + { + "epoch": 2.44, + "learning_rate": 2.2696578085183836e-06, + "loss": 0.534, + "step": 13425 + }, + { + "epoch": 2.44, + "learning_rate": 2.2696032034947216e-06, + "loss": 0.6083, + "step": 13426 + }, + { + "epoch": 2.44, + "learning_rate": 2.2695485984710595e-06, + "loss": 1.9124, + "step": 13427 + }, + { + "epoch": 2.44, + "learning_rate": 2.2694939934473975e-06, + "loss": 1.496, + "step": 13428 + }, + { + "epoch": 2.44, + "learning_rate": 2.269439388423735e-06, + "loss": 0.2125, + "step": 13429 + }, + { + "epoch": 2.44, + "learning_rate": 2.269384783400073e-06, + "loss": 0.0329, + "step": 13430 + }, + { + "epoch": 2.44, + "learning_rate": 2.269330178376411e-06, + "loss": 0.0158, + "step": 13431 + }, + { + "epoch": 2.44, + "learning_rate": 2.2692755733527484e-06, + "loss": 0.0442, + "step": 13432 + }, + { + "epoch": 2.44, + "learning_rate": 2.2692209683290863e-06, + "loss": 0.0657, + "step": 13433 + }, + { + "epoch": 2.44, + "learning_rate": 2.2691663633054243e-06, + "loss": 1.2455, + "step": 13434 + }, + { + "epoch": 2.44, + "learning_rate": 2.269111758281762e-06, + "loss": 0.8805, + "step": 13435 + }, + { + "epoch": 2.44, + "learning_rate": 2.2690571532580997e-06, + "loss": 0.3307, + "step": 13436 + }, + { + "epoch": 2.44, + "learning_rate": 2.2690025482344377e-06, + "loss": 0.41, + "step": 13437 + }, + { + "epoch": 2.44, + "learning_rate": 2.268947943210775e-06, + "loss": 0.0478, + "step": 13438 + }, + { + "epoch": 2.44, + "learning_rate": 2.268893338187113e-06, + "loss": 0.3637, + "step": 13439 + }, + { + "epoch": 2.44, + "learning_rate": 2.268838733163451e-06, + "loss": 0.6909, + "step": 13440 + }, + { + "epoch": 2.44, + "learning_rate": 2.268784128139789e-06, + "loss": 0.3708, + "step": 13441 + }, + { + "epoch": 2.44, + "learning_rate": 2.268729523116127e-06, + "loss": 0.413, + "step": 13442 + }, + { + "epoch": 2.44, + "learning_rate": 2.2686749180924645e-06, + "loss": 0.4811, + "step": 13443 + }, + { + "epoch": 2.44, + "learning_rate": 2.2686203130688024e-06, + "loss": 0.1881, + "step": 13444 + }, + { + "epoch": 2.44, + "learning_rate": 2.2685657080451404e-06, + "loss": 0.6868, + "step": 13445 + }, + { + "epoch": 2.45, + "learning_rate": 2.268511103021478e-06, + "loss": 0.0402, + "step": 13446 + }, + { + "epoch": 2.45, + "learning_rate": 2.268456497997816e-06, + "loss": 1.9103, + "step": 13447 + }, + { + "epoch": 2.45, + "learning_rate": 2.268401892974154e-06, + "loss": 0.8443, + "step": 13448 + }, + { + "epoch": 2.45, + "learning_rate": 2.2683472879504913e-06, + "loss": 2.3091, + "step": 13449 + }, + { + "epoch": 2.45, + "learning_rate": 2.2682926829268293e-06, + "loss": 0.6818, + "step": 13450 + }, + { + "epoch": 2.45, + "learning_rate": 2.268238077903167e-06, + "loss": 2.6343, + "step": 13451 + }, + { + "epoch": 2.45, + "learning_rate": 2.2681834728795047e-06, + "loss": 0.1319, + "step": 13452 + }, + { + "epoch": 2.45, + "learning_rate": 2.2681288678558427e-06, + "loss": 0.5152, + "step": 13453 + }, + { + "epoch": 2.45, + "learning_rate": 2.2680742628321806e-06, + "loss": 0.0169, + "step": 13454 + }, + { + "epoch": 2.45, + "learning_rate": 2.2680196578085185e-06, + "loss": 0.3968, + "step": 13455 + }, + { + "epoch": 2.45, + "learning_rate": 2.2679650527848565e-06, + "loss": 0.7044, + "step": 13456 + }, + { + "epoch": 2.45, + "learning_rate": 2.267910447761194e-06, + "loss": 0.1116, + "step": 13457 + }, + { + "epoch": 2.45, + "learning_rate": 2.267855842737532e-06, + "loss": 0.8305, + "step": 13458 + }, + { + "epoch": 2.45, + "learning_rate": 2.26780123771387e-06, + "loss": 0.0759, + "step": 13459 + }, + { + "epoch": 2.45, + "learning_rate": 2.2677466326902074e-06, + "loss": 0.0259, + "step": 13460 + }, + { + "epoch": 2.45, + "learning_rate": 2.2676920276665454e-06, + "loss": 0.0288, + "step": 13461 + }, + { + "epoch": 2.45, + "learning_rate": 2.2676374226428833e-06, + "loss": 0.238, + "step": 13462 + }, + { + "epoch": 2.45, + "learning_rate": 2.267582817619221e-06, + "loss": 1.6631, + "step": 13463 + }, + { + "epoch": 2.45, + "learning_rate": 2.2675282125955588e-06, + "loss": 1.1982, + "step": 13464 + }, + { + "epoch": 2.45, + "learning_rate": 2.2674736075718967e-06, + "loss": 1.0969, + "step": 13465 + }, + { + "epoch": 2.45, + "learning_rate": 2.2674190025482342e-06, + "loss": 0.1916, + "step": 13466 + }, + { + "epoch": 2.45, + "learning_rate": 2.267364397524572e-06, + "loss": 2.884, + "step": 13467 + }, + { + "epoch": 2.45, + "learning_rate": 2.26730979250091e-06, + "loss": 0.8782, + "step": 13468 + }, + { + "epoch": 2.45, + "learning_rate": 2.267255187477248e-06, + "loss": 0.4085, + "step": 13469 + }, + { + "epoch": 2.45, + "learning_rate": 2.267200582453586e-06, + "loss": 0.3091, + "step": 13470 + }, + { + "epoch": 2.45, + "learning_rate": 2.2671459774299235e-06, + "loss": 0.1283, + "step": 13471 + }, + { + "epoch": 2.45, + "learning_rate": 2.2670913724062615e-06, + "loss": 1.6253, + "step": 13472 + }, + { + "epoch": 2.45, + "learning_rate": 2.2670367673825994e-06, + "loss": 1.4334, + "step": 13473 + }, + { + "epoch": 2.45, + "learning_rate": 2.266982162358937e-06, + "loss": 1.0254, + "step": 13474 + }, + { + "epoch": 2.45, + "learning_rate": 2.266927557335275e-06, + "loss": 0.6062, + "step": 13475 + }, + { + "epoch": 2.45, + "learning_rate": 2.266872952311613e-06, + "loss": 1.0506, + "step": 13476 + }, + { + "epoch": 2.45, + "learning_rate": 2.2668183472879503e-06, + "loss": 0.0606, + "step": 13477 + }, + { + "epoch": 2.45, + "learning_rate": 2.2667637422642883e-06, + "loss": 1.3977, + "step": 13478 + }, + { + "epoch": 2.45, + "learning_rate": 2.2667091372406262e-06, + "loss": 0.0814, + "step": 13479 + }, + { + "epoch": 2.45, + "learning_rate": 2.2666545322169637e-06, + "loss": 1.6828, + "step": 13480 + }, + { + "epoch": 2.45, + "learning_rate": 2.2665999271933017e-06, + "loss": 1.0199, + "step": 13481 + }, + { + "epoch": 2.45, + "learning_rate": 2.2665453221696396e-06, + "loss": 1.2166, + "step": 13482 + }, + { + "epoch": 2.45, + "learning_rate": 2.2664907171459776e-06, + "loss": 0.7685, + "step": 13483 + }, + { + "epoch": 2.45, + "learning_rate": 2.2664361121223155e-06, + "loss": 1.2092, + "step": 13484 + }, + { + "epoch": 2.45, + "learning_rate": 2.266381507098653e-06, + "loss": 0.9993, + "step": 13485 + }, + { + "epoch": 2.45, + "learning_rate": 2.266326902074991e-06, + "loss": 1.5185, + "step": 13486 + }, + { + "epoch": 2.45, + "learning_rate": 2.266272297051329e-06, + "loss": 0.4349, + "step": 13487 + }, + { + "epoch": 2.45, + "learning_rate": 2.2662176920276664e-06, + "loss": 0.5313, + "step": 13488 + }, + { + "epoch": 2.45, + "learning_rate": 2.2661630870040044e-06, + "loss": 0.8329, + "step": 13489 + }, + { + "epoch": 2.45, + "learning_rate": 2.2661084819803423e-06, + "loss": 0.5546, + "step": 13490 + }, + { + "epoch": 2.45, + "learning_rate": 2.26605387695668e-06, + "loss": 0.4175, + "step": 13491 + }, + { + "epoch": 2.45, + "learning_rate": 2.265999271933018e-06, + "loss": 2.4339, + "step": 13492 + }, + { + "epoch": 2.45, + "learning_rate": 2.2659446669093557e-06, + "loss": 1.0672, + "step": 13493 + }, + { + "epoch": 2.45, + "learning_rate": 2.2658900618856937e-06, + "loss": 0.3177, + "step": 13494 + }, + { + "epoch": 2.45, + "learning_rate": 2.2658354568620316e-06, + "loss": 0.583, + "step": 13495 + }, + { + "epoch": 2.45, + "learning_rate": 2.265780851838369e-06, + "loss": 0.0227, + "step": 13496 + }, + { + "epoch": 2.45, + "learning_rate": 2.265726246814707e-06, + "loss": 1.3377, + "step": 13497 + }, + { + "epoch": 2.45, + "learning_rate": 2.265671641791045e-06, + "loss": 0.0407, + "step": 13498 + }, + { + "epoch": 2.45, + "learning_rate": 2.2656170367673826e-06, + "loss": 1.3276, + "step": 13499 + }, + { + "epoch": 2.45, + "learning_rate": 2.2655624317437205e-06, + "loss": 2.6633, + "step": 13500 + }, + { + "epoch": 2.46, + "learning_rate": 2.2655078267200584e-06, + "loss": 0.8137, + "step": 13501 + }, + { + "epoch": 2.46, + "learning_rate": 2.265453221696396e-06, + "loss": 0.4639, + "step": 13502 + }, + { + "epoch": 2.46, + "learning_rate": 2.265398616672734e-06, + "loss": 0.5683, + "step": 13503 + }, + { + "epoch": 2.46, + "learning_rate": 2.265344011649072e-06, + "loss": 0.9049, + "step": 13504 + }, + { + "epoch": 2.46, + "learning_rate": 2.2652894066254094e-06, + "loss": 1.2066, + "step": 13505 + }, + { + "epoch": 2.46, + "learning_rate": 2.2652348016017473e-06, + "loss": 0.3264, + "step": 13506 + }, + { + "epoch": 2.46, + "learning_rate": 2.2651801965780853e-06, + "loss": 0.5808, + "step": 13507 + }, + { + "epoch": 2.46, + "learning_rate": 2.265125591554423e-06, + "loss": 0.6024, + "step": 13508 + }, + { + "epoch": 2.46, + "learning_rate": 2.265070986530761e-06, + "loss": 0.4264, + "step": 13509 + }, + { + "epoch": 2.46, + "learning_rate": 2.2650163815070987e-06, + "loss": 0.2955, + "step": 13510 + }, + { + "epoch": 2.46, + "learning_rate": 2.2649617764834366e-06, + "loss": 0.1491, + "step": 13511 + }, + { + "epoch": 2.46, + "learning_rate": 2.2649071714597745e-06, + "loss": 2.023, + "step": 13512 + }, + { + "epoch": 2.46, + "learning_rate": 2.264852566436112e-06, + "loss": 1.0669, + "step": 13513 + }, + { + "epoch": 2.46, + "learning_rate": 2.26479796141245e-06, + "loss": 3.9402, + "step": 13514 + }, + { + "epoch": 2.46, + "learning_rate": 2.264743356388788e-06, + "loss": 0.2848, + "step": 13515 + }, + { + "epoch": 2.46, + "learning_rate": 2.2646887513651255e-06, + "loss": 0.7162, + "step": 13516 + }, + { + "epoch": 2.46, + "learning_rate": 2.2646341463414634e-06, + "loss": 0.7027, + "step": 13517 + }, + { + "epoch": 2.46, + "learning_rate": 2.2645795413178014e-06, + "loss": 3.2309, + "step": 13518 + }, + { + "epoch": 2.46, + "learning_rate": 2.264524936294139e-06, + "loss": 0.1126, + "step": 13519 + }, + { + "epoch": 2.46, + "learning_rate": 2.264470331270477e-06, + "loss": 0.0802, + "step": 13520 + }, + { + "epoch": 2.46, + "learning_rate": 2.2644157262468148e-06, + "loss": 0.5888, + "step": 13521 + }, + { + "epoch": 2.46, + "learning_rate": 2.2643611212231527e-06, + "loss": 1.3361, + "step": 13522 + }, + { + "epoch": 2.46, + "learning_rate": 2.2643065161994907e-06, + "loss": 1.5786, + "step": 13523 + }, + { + "epoch": 2.46, + "learning_rate": 2.264251911175828e-06, + "loss": 0.8138, + "step": 13524 + }, + { + "epoch": 2.46, + "learning_rate": 2.264197306152166e-06, + "loss": 0.6137, + "step": 13525 + }, + { + "epoch": 2.46, + "learning_rate": 2.264142701128504e-06, + "loss": 1.1527, + "step": 13526 + }, + { + "epoch": 2.46, + "learning_rate": 2.2640880961048416e-06, + "loss": 0.4638, + "step": 13527 + }, + { + "epoch": 2.46, + "learning_rate": 2.2640334910811795e-06, + "loss": 0.7022, + "step": 13528 + }, + { + "epoch": 2.46, + "learning_rate": 2.2639788860575175e-06, + "loss": 0.2517, + "step": 13529 + }, + { + "epoch": 2.46, + "learning_rate": 2.263924281033855e-06, + "loss": 1.1891, + "step": 13530 + }, + { + "epoch": 2.46, + "learning_rate": 2.263869676010193e-06, + "loss": 1.3117, + "step": 13531 + }, + { + "epoch": 2.46, + "learning_rate": 2.263815070986531e-06, + "loss": 0.6667, + "step": 13532 + }, + { + "epoch": 2.46, + "learning_rate": 2.2637604659628684e-06, + "loss": 0.7719, + "step": 13533 + }, + { + "epoch": 2.46, + "learning_rate": 2.2637058609392063e-06, + "loss": 0.7419, + "step": 13534 + }, + { + "epoch": 2.46, + "learning_rate": 2.2636512559155443e-06, + "loss": 1.5206, + "step": 13535 + }, + { + "epoch": 2.46, + "learning_rate": 2.2635966508918822e-06, + "loss": 0.1561, + "step": 13536 + }, + { + "epoch": 2.46, + "learning_rate": 2.26354204586822e-06, + "loss": 1.6912, + "step": 13537 + }, + { + "epoch": 2.46, + "learning_rate": 2.2634874408445577e-06, + "loss": 0.25, + "step": 13538 + }, + { + "epoch": 2.46, + "learning_rate": 2.2634328358208956e-06, + "loss": 1.1859, + "step": 13539 + }, + { + "epoch": 2.46, + "learning_rate": 2.2633782307972336e-06, + "loss": 1.3415, + "step": 13540 + }, + { + "epoch": 2.46, + "learning_rate": 2.263323625773571e-06, + "loss": 0.1091, + "step": 13541 + }, + { + "epoch": 2.46, + "learning_rate": 2.263269020749909e-06, + "loss": 0.9076, + "step": 13542 + }, + { + "epoch": 2.46, + "learning_rate": 2.263214415726247e-06, + "loss": 1.6595, + "step": 13543 + }, + { + "epoch": 2.46, + "learning_rate": 2.2631598107025845e-06, + "loss": 0.1705, + "step": 13544 + }, + { + "epoch": 2.46, + "learning_rate": 2.2631052056789224e-06, + "loss": 2.3823, + "step": 13545 + }, + { + "epoch": 2.46, + "learning_rate": 2.2630506006552604e-06, + "loss": 0.4478, + "step": 13546 + }, + { + "epoch": 2.46, + "learning_rate": 2.262995995631598e-06, + "loss": 1.0142, + "step": 13547 + }, + { + "epoch": 2.46, + "learning_rate": 2.262941390607936e-06, + "loss": 0.3442, + "step": 13548 + }, + { + "epoch": 2.46, + "learning_rate": 2.262886785584274e-06, + "loss": 1.8322, + "step": 13549 + }, + { + "epoch": 2.46, + "learning_rate": 2.2628321805606117e-06, + "loss": 1.2737, + "step": 13550 + }, + { + "epoch": 2.46, + "learning_rate": 2.2627775755369497e-06, + "loss": 1.6697, + "step": 13551 + }, + { + "epoch": 2.46, + "learning_rate": 2.262722970513287e-06, + "loss": 1.3554, + "step": 13552 + }, + { + "epoch": 2.46, + "learning_rate": 2.262668365489625e-06, + "loss": 1.1609, + "step": 13553 + }, + { + "epoch": 2.46, + "learning_rate": 2.262613760465963e-06, + "loss": 0.6592, + "step": 13554 + }, + { + "epoch": 2.46, + "learning_rate": 2.2625591554423006e-06, + "loss": 0.9792, + "step": 13555 + }, + { + "epoch": 2.47, + "learning_rate": 2.2625045504186385e-06, + "loss": 1.1721, + "step": 13556 + }, + { + "epoch": 2.47, + "learning_rate": 2.2624499453949765e-06, + "loss": 1.3224, + "step": 13557 + }, + { + "epoch": 2.47, + "learning_rate": 2.262395340371314e-06, + "loss": 0.171, + "step": 13558 + }, + { + "epoch": 2.47, + "learning_rate": 2.262340735347652e-06, + "loss": 2.4198, + "step": 13559 + }, + { + "epoch": 2.47, + "learning_rate": 2.26228613032399e-06, + "loss": 0.0644, + "step": 13560 + }, + { + "epoch": 2.47, + "learning_rate": 2.2622315253003274e-06, + "loss": 1.0986, + "step": 13561 + }, + { + "epoch": 2.47, + "learning_rate": 2.2621769202766654e-06, + "loss": 0.7005, + "step": 13562 + }, + { + "epoch": 2.47, + "learning_rate": 2.2621223152530033e-06, + "loss": 0.0419, + "step": 13563 + }, + { + "epoch": 2.47, + "learning_rate": 2.2620677102293412e-06, + "loss": 1.5996, + "step": 13564 + }, + { + "epoch": 2.47, + "learning_rate": 2.262013105205679e-06, + "loss": 0.0285, + "step": 13565 + }, + { + "epoch": 2.47, + "learning_rate": 2.2619585001820167e-06, + "loss": 0.036, + "step": 13566 + }, + { + "epoch": 2.47, + "learning_rate": 2.2619038951583547e-06, + "loss": 2.5256, + "step": 13567 + }, + { + "epoch": 2.47, + "learning_rate": 2.2618492901346926e-06, + "loss": 0.6417, + "step": 13568 + }, + { + "epoch": 2.47, + "learning_rate": 2.26179468511103e-06, + "loss": 0.7835, + "step": 13569 + }, + { + "epoch": 2.47, + "learning_rate": 2.261740080087368e-06, + "loss": 1.711, + "step": 13570 + }, + { + "epoch": 2.47, + "learning_rate": 2.261685475063706e-06, + "loss": 0.5663, + "step": 13571 + }, + { + "epoch": 2.47, + "learning_rate": 2.2616308700400435e-06, + "loss": 1.3475, + "step": 13572 + }, + { + "epoch": 2.47, + "learning_rate": 2.2615762650163815e-06, + "loss": 0.7557, + "step": 13573 + }, + { + "epoch": 2.47, + "learning_rate": 2.2615216599927194e-06, + "loss": 1.7977, + "step": 13574 + }, + { + "epoch": 2.47, + "learning_rate": 2.2614670549690574e-06, + "loss": 0.5935, + "step": 13575 + }, + { + "epoch": 2.47, + "learning_rate": 2.2614124499453953e-06, + "loss": 0.6461, + "step": 13576 + }, + { + "epoch": 2.47, + "learning_rate": 2.261357844921733e-06, + "loss": 2.2244, + "step": 13577 + }, + { + "epoch": 2.47, + "learning_rate": 2.2613032398980708e-06, + "loss": 0.5363, + "step": 13578 + }, + { + "epoch": 2.47, + "learning_rate": 2.2612486348744087e-06, + "loss": 0.2935, + "step": 13579 + }, + { + "epoch": 2.47, + "learning_rate": 2.2611940298507462e-06, + "loss": 0.0887, + "step": 13580 + }, + { + "epoch": 2.47, + "learning_rate": 2.261139424827084e-06, + "loss": 0.1427, + "step": 13581 + }, + { + "epoch": 2.47, + "learning_rate": 2.261084819803422e-06, + "loss": 0.2792, + "step": 13582 + }, + { + "epoch": 2.47, + "learning_rate": 2.2610302147797596e-06, + "loss": 2.1954, + "step": 13583 + }, + { + "epoch": 2.47, + "learning_rate": 2.2609756097560976e-06, + "loss": 0.6536, + "step": 13584 + }, + { + "epoch": 2.47, + "learning_rate": 2.2609210047324355e-06, + "loss": 0.482, + "step": 13585 + }, + { + "epoch": 2.47, + "learning_rate": 2.260866399708773e-06, + "loss": 0.087, + "step": 13586 + }, + { + "epoch": 2.47, + "learning_rate": 2.260811794685111e-06, + "loss": 0.0883, + "step": 13587 + }, + { + "epoch": 2.47, + "learning_rate": 2.260757189661449e-06, + "loss": 1.3011, + "step": 13588 + }, + { + "epoch": 2.47, + "learning_rate": 2.260702584637787e-06, + "loss": 0.2249, + "step": 13589 + }, + { + "epoch": 2.47, + "learning_rate": 2.260647979614125e-06, + "loss": 0.8873, + "step": 13590 + }, + { + "epoch": 2.47, + "learning_rate": 2.2605933745904623e-06, + "loss": 2.2098, + "step": 13591 + }, + { + "epoch": 2.47, + "learning_rate": 2.2605387695668003e-06, + "loss": 0.9118, + "step": 13592 + }, + { + "epoch": 2.47, + "learning_rate": 2.2604841645431382e-06, + "loss": 0.5483, + "step": 13593 + }, + { + "epoch": 2.47, + "learning_rate": 2.2604295595194757e-06, + "loss": 1.344, + "step": 13594 + }, + { + "epoch": 2.47, + "learning_rate": 2.2603749544958137e-06, + "loss": 0.0123, + "step": 13595 + }, + { + "epoch": 2.47, + "learning_rate": 2.2603203494721516e-06, + "loss": 0.0556, + "step": 13596 + }, + { + "epoch": 2.47, + "learning_rate": 2.260265744448489e-06, + "loss": 0.6536, + "step": 13597 + }, + { + "epoch": 2.47, + "learning_rate": 2.260211139424827e-06, + "loss": 0.4398, + "step": 13598 + }, + { + "epoch": 2.47, + "learning_rate": 2.260156534401165e-06, + "loss": 2.6941, + "step": 13599 + }, + { + "epoch": 2.47, + "learning_rate": 2.2601019293775026e-06, + "loss": 0.6003, + "step": 13600 + }, + { + "epoch": 2.47, + "learning_rate": 2.2600473243538405e-06, + "loss": 0.8549, + "step": 13601 + }, + { + "epoch": 2.47, + "learning_rate": 2.2599927193301784e-06, + "loss": 2.0015, + "step": 13602 + }, + { + "epoch": 2.47, + "learning_rate": 2.2599381143065164e-06, + "loss": 0.8717, + "step": 13603 + }, + { + "epoch": 2.47, + "learning_rate": 2.2598835092828543e-06, + "loss": 0.6469, + "step": 13604 + }, + { + "epoch": 2.47, + "learning_rate": 2.259828904259192e-06, + "loss": 1.8542, + "step": 13605 + }, + { + "epoch": 2.47, + "learning_rate": 2.25977429923553e-06, + "loss": 0.9125, + "step": 13606 + }, + { + "epoch": 2.47, + "learning_rate": 2.2597196942118677e-06, + "loss": 0.686, + "step": 13607 + }, + { + "epoch": 2.47, + "learning_rate": 2.2596650891882053e-06, + "loss": 1.7424, + "step": 13608 + }, + { + "epoch": 2.47, + "learning_rate": 2.259610484164543e-06, + "loss": 0.7094, + "step": 13609 + }, + { + "epoch": 2.47, + "learning_rate": 2.259555879140881e-06, + "loss": 1.8992, + "step": 13610 + }, + { + "epoch": 2.48, + "learning_rate": 2.2595012741172187e-06, + "loss": 0.5896, + "step": 13611 + }, + { + "epoch": 2.48, + "learning_rate": 2.2594466690935566e-06, + "loss": 0.3392, + "step": 13612 + }, + { + "epoch": 2.48, + "learning_rate": 2.2593920640698945e-06, + "loss": 0.418, + "step": 13613 + }, + { + "epoch": 2.48, + "learning_rate": 2.259337459046232e-06, + "loss": 0.7063, + "step": 13614 + }, + { + "epoch": 2.48, + "learning_rate": 2.25928285402257e-06, + "loss": 1.7087, + "step": 13615 + }, + { + "epoch": 2.48, + "learning_rate": 2.259228248998908e-06, + "loss": 0.2828, + "step": 13616 + }, + { + "epoch": 2.48, + "learning_rate": 2.259173643975246e-06, + "loss": 0.6933, + "step": 13617 + }, + { + "epoch": 2.48, + "learning_rate": 2.259119038951584e-06, + "loss": 0.1242, + "step": 13618 + }, + { + "epoch": 2.48, + "learning_rate": 2.2590644339279214e-06, + "loss": 0.7036, + "step": 13619 + }, + { + "epoch": 2.48, + "learning_rate": 2.2590098289042593e-06, + "loss": 0.0646, + "step": 13620 + }, + { + "epoch": 2.48, + "learning_rate": 2.2589552238805972e-06, + "loss": 0.0928, + "step": 13621 + }, + { + "epoch": 2.48, + "learning_rate": 2.2589006188569348e-06, + "loss": 0.3936, + "step": 13622 + }, + { + "epoch": 2.48, + "learning_rate": 2.2588460138332727e-06, + "loss": 0.4133, + "step": 13623 + }, + { + "epoch": 2.48, + "learning_rate": 2.2587914088096107e-06, + "loss": 0.3091, + "step": 13624 + }, + { + "epoch": 2.48, + "learning_rate": 2.258736803785948e-06, + "loss": 0.3865, + "step": 13625 + }, + { + "epoch": 2.48, + "learning_rate": 2.258682198762286e-06, + "loss": 0.7734, + "step": 13626 + }, + { + "epoch": 2.48, + "learning_rate": 2.258627593738624e-06, + "loss": 0.0239, + "step": 13627 + }, + { + "epoch": 2.48, + "learning_rate": 2.2585729887149616e-06, + "loss": 1.1908, + "step": 13628 + }, + { + "epoch": 2.48, + "learning_rate": 2.2585183836912995e-06, + "loss": 0.2232, + "step": 13629 + }, + { + "epoch": 2.48, + "learning_rate": 2.2584637786676375e-06, + "loss": 1.9873, + "step": 13630 + }, + { + "epoch": 2.48, + "learning_rate": 2.2584091736439754e-06, + "loss": 0.7014, + "step": 13631 + }, + { + "epoch": 2.48, + "learning_rate": 2.2583545686203134e-06, + "loss": 0.8637, + "step": 13632 + }, + { + "epoch": 2.48, + "learning_rate": 2.258299963596651e-06, + "loss": 3.1235, + "step": 13633 + }, + { + "epoch": 2.48, + "learning_rate": 2.258245358572989e-06, + "loss": 1.4322, + "step": 13634 + }, + { + "epoch": 2.48, + "learning_rate": 2.2581907535493268e-06, + "loss": 0.7245, + "step": 13635 + }, + { + "epoch": 2.48, + "learning_rate": 2.2581361485256643e-06, + "loss": 0.2322, + "step": 13636 + }, + { + "epoch": 2.48, + "learning_rate": 2.2580815435020022e-06, + "loss": 1.49, + "step": 13637 + }, + { + "epoch": 2.48, + "learning_rate": 2.25802693847834e-06, + "loss": 2.3194, + "step": 13638 + }, + { + "epoch": 2.48, + "learning_rate": 2.2579723334546777e-06, + "loss": 0.2099, + "step": 13639 + }, + { + "epoch": 2.48, + "learning_rate": 2.2579177284310156e-06, + "loss": 1.0378, + "step": 13640 + }, + { + "epoch": 2.48, + "learning_rate": 2.2578631234073536e-06, + "loss": 1.4264, + "step": 13641 + }, + { + "epoch": 2.48, + "learning_rate": 2.257808518383691e-06, + "loss": 0.1217, + "step": 13642 + }, + { + "epoch": 2.48, + "learning_rate": 2.257753913360029e-06, + "loss": 2.4434, + "step": 13643 + }, + { + "epoch": 2.48, + "learning_rate": 2.257699308336367e-06, + "loss": 2.0536, + "step": 13644 + }, + { + "epoch": 2.48, + "learning_rate": 2.257644703312705e-06, + "loss": 0.0602, + "step": 13645 + }, + { + "epoch": 2.48, + "learning_rate": 2.257590098289043e-06, + "loss": 0.5151, + "step": 13646 + }, + { + "epoch": 2.48, + "learning_rate": 2.2575354932653804e-06, + "loss": 1.5733, + "step": 13647 + }, + { + "epoch": 2.48, + "learning_rate": 2.2574808882417183e-06, + "loss": 0.3443, + "step": 13648 + }, + { + "epoch": 2.48, + "learning_rate": 2.2574262832180563e-06, + "loss": 1.5671, + "step": 13649 + }, + { + "epoch": 2.48, + "learning_rate": 2.257371678194394e-06, + "loss": 1.5314, + "step": 13650 + }, + { + "epoch": 2.48, + "learning_rate": 2.2573170731707317e-06, + "loss": 0.8939, + "step": 13651 + }, + { + "epoch": 2.48, + "learning_rate": 2.2572624681470697e-06, + "loss": 0.0291, + "step": 13652 + }, + { + "epoch": 2.48, + "learning_rate": 2.257207863123407e-06, + "loss": 0.9037, + "step": 13653 + }, + { + "epoch": 2.48, + "learning_rate": 2.257153258099745e-06, + "loss": 1.8751, + "step": 13654 + }, + { + "epoch": 2.48, + "learning_rate": 2.257098653076083e-06, + "loss": 0.6074, + "step": 13655 + }, + { + "epoch": 2.48, + "learning_rate": 2.2570440480524206e-06, + "loss": 0.3016, + "step": 13656 + }, + { + "epoch": 2.48, + "learning_rate": 2.2569894430287585e-06, + "loss": 0.2881, + "step": 13657 + }, + { + "epoch": 2.48, + "learning_rate": 2.2569348380050965e-06, + "loss": 0.2847, + "step": 13658 + }, + { + "epoch": 2.48, + "learning_rate": 2.2568802329814344e-06, + "loss": 1.0353, + "step": 13659 + }, + { + "epoch": 2.48, + "learning_rate": 2.2568256279577724e-06, + "loss": 1.5366, + "step": 13660 + }, + { + "epoch": 2.48, + "learning_rate": 2.25677102293411e-06, + "loss": 0.0207, + "step": 13661 + }, + { + "epoch": 2.48, + "learning_rate": 2.256716417910448e-06, + "loss": 2.1702, + "step": 13662 + }, + { + "epoch": 2.48, + "learning_rate": 2.2566618128867858e-06, + "loss": 0.4131, + "step": 13663 + }, + { + "epoch": 2.48, + "learning_rate": 2.2566072078631233e-06, + "loss": 1.1572, + "step": 13664 + }, + { + "epoch": 2.48, + "learning_rate": 2.2565526028394612e-06, + "loss": 3.7175, + "step": 13665 + }, + { + "epoch": 2.49, + "learning_rate": 2.256497997815799e-06, + "loss": 0.3305, + "step": 13666 + }, + { + "epoch": 2.49, + "learning_rate": 2.2564433927921367e-06, + "loss": 0.0365, + "step": 13667 + }, + { + "epoch": 2.49, + "learning_rate": 2.2563887877684747e-06, + "loss": 0.3098, + "step": 13668 + }, + { + "epoch": 2.49, + "learning_rate": 2.2563341827448126e-06, + "loss": 1.5878, + "step": 13669 + }, + { + "epoch": 2.49, + "learning_rate": 2.2562795777211505e-06, + "loss": 3.9889, + "step": 13670 + }, + { + "epoch": 2.49, + "learning_rate": 2.2562249726974885e-06, + "loss": 0.8106, + "step": 13671 + }, + { + "epoch": 2.49, + "learning_rate": 2.256170367673826e-06, + "loss": 0.2809, + "step": 13672 + }, + { + "epoch": 2.49, + "learning_rate": 2.256115762650164e-06, + "loss": 0.2482, + "step": 13673 + }, + { + "epoch": 2.49, + "learning_rate": 2.256061157626502e-06, + "loss": 0.2785, + "step": 13674 + }, + { + "epoch": 2.49, + "learning_rate": 2.2560065526028394e-06, + "loss": 1.3031, + "step": 13675 + }, + { + "epoch": 2.49, + "learning_rate": 2.2559519475791774e-06, + "loss": 1.6932, + "step": 13676 + }, + { + "epoch": 2.49, + "learning_rate": 2.2558973425555153e-06, + "loss": 0.5896, + "step": 13677 + }, + { + "epoch": 2.49, + "learning_rate": 2.255842737531853e-06, + "loss": 0.1634, + "step": 13678 + }, + { + "epoch": 2.49, + "learning_rate": 2.2557881325081908e-06, + "loss": 1.4623, + "step": 13679 + }, + { + "epoch": 2.49, + "learning_rate": 2.2557335274845287e-06, + "loss": 1.6944, + "step": 13680 + }, + { + "epoch": 2.49, + "learning_rate": 2.2556789224608662e-06, + "loss": 1.4739, + "step": 13681 + }, + { + "epoch": 2.49, + "learning_rate": 2.255624317437204e-06, + "loss": 0.442, + "step": 13682 + }, + { + "epoch": 2.49, + "learning_rate": 2.255569712413542e-06, + "loss": 1.7097, + "step": 13683 + }, + { + "epoch": 2.49, + "learning_rate": 2.25551510738988e-06, + "loss": 1.6114, + "step": 13684 + }, + { + "epoch": 2.49, + "learning_rate": 2.255460502366218e-06, + "loss": 2.7014, + "step": 13685 + }, + { + "epoch": 2.49, + "learning_rate": 2.2554058973425555e-06, + "loss": 1.6559, + "step": 13686 + }, + { + "epoch": 2.49, + "learning_rate": 2.2553512923188935e-06, + "loss": 0.5765, + "step": 13687 + }, + { + "epoch": 2.49, + "learning_rate": 2.2552966872952314e-06, + "loss": 0.0365, + "step": 13688 + }, + { + "epoch": 2.49, + "learning_rate": 2.255242082271569e-06, + "loss": 0.0341, + "step": 13689 + }, + { + "epoch": 2.49, + "learning_rate": 2.255187477247907e-06, + "loss": 1.2887, + "step": 13690 + }, + { + "epoch": 2.49, + "learning_rate": 2.255132872224245e-06, + "loss": 0.024, + "step": 13691 + }, + { + "epoch": 2.49, + "learning_rate": 2.2550782672005823e-06, + "loss": 3.6429, + "step": 13692 + }, + { + "epoch": 2.49, + "learning_rate": 2.2550236621769203e-06, + "loss": 0.8064, + "step": 13693 + }, + { + "epoch": 2.49, + "learning_rate": 2.2549690571532582e-06, + "loss": 0.5794, + "step": 13694 + }, + { + "epoch": 2.49, + "learning_rate": 2.2549144521295957e-06, + "loss": 1.19, + "step": 13695 + }, + { + "epoch": 2.49, + "learning_rate": 2.2548598471059337e-06, + "loss": 2.7302, + "step": 13696 + }, + { + "epoch": 2.49, + "learning_rate": 2.2548052420822716e-06, + "loss": 2.0255, + "step": 13697 + }, + { + "epoch": 2.49, + "learning_rate": 2.2547506370586096e-06, + "loss": 0.3808, + "step": 13698 + }, + { + "epoch": 2.49, + "learning_rate": 2.2546960320349475e-06, + "loss": 0.3628, + "step": 13699 + }, + { + "epoch": 2.49, + "learning_rate": 2.254641427011285e-06, + "loss": 0.0191, + "step": 13700 + }, + { + "epoch": 2.49, + "learning_rate": 2.254586821987623e-06, + "loss": 3.1559, + "step": 13701 + }, + { + "epoch": 2.49, + "learning_rate": 2.254532216963961e-06, + "loss": 0.7155, + "step": 13702 + }, + { + "epoch": 2.49, + "learning_rate": 2.2544776119402984e-06, + "loss": 0.6691, + "step": 13703 + }, + { + "epoch": 2.49, + "learning_rate": 2.2544230069166364e-06, + "loss": 1.1305, + "step": 13704 + }, + { + "epoch": 2.49, + "learning_rate": 2.2543684018929743e-06, + "loss": 2.0237, + "step": 13705 + }, + { + "epoch": 2.49, + "learning_rate": 2.254313796869312e-06, + "loss": 1.9963, + "step": 13706 + }, + { + "epoch": 2.49, + "learning_rate": 2.25425919184565e-06, + "loss": 0.1219, + "step": 13707 + }, + { + "epoch": 2.49, + "learning_rate": 2.2542045868219877e-06, + "loss": 1.1405, + "step": 13708 + }, + { + "epoch": 2.49, + "learning_rate": 2.2541499817983253e-06, + "loss": 0.5658, + "step": 13709 + }, + { + "epoch": 2.49, + "learning_rate": 2.254095376774663e-06, + "loss": 2.2041, + "step": 13710 + }, + { + "epoch": 2.49, + "learning_rate": 2.254040771751001e-06, + "loss": 1.2221, + "step": 13711 + }, + { + "epoch": 2.49, + "learning_rate": 2.253986166727339e-06, + "loss": 1.465, + "step": 13712 + }, + { + "epoch": 2.49, + "learning_rate": 2.253931561703677e-06, + "loss": 1.0174, + "step": 13713 + }, + { + "epoch": 2.49, + "learning_rate": 2.2538769566800145e-06, + "loss": 0.4009, + "step": 13714 + }, + { + "epoch": 2.49, + "learning_rate": 2.2538223516563525e-06, + "loss": 0.9689, + "step": 13715 + }, + { + "epoch": 2.49, + "learning_rate": 2.2537677466326904e-06, + "loss": 1.5314, + "step": 13716 + }, + { + "epoch": 2.49, + "learning_rate": 2.253713141609028e-06, + "loss": 0.2689, + "step": 13717 + }, + { + "epoch": 2.49, + "learning_rate": 2.253658536585366e-06, + "loss": 0.7572, + "step": 13718 + }, + { + "epoch": 2.49, + "learning_rate": 2.253603931561704e-06, + "loss": 0.5296, + "step": 13719 + }, + { + "epoch": 2.49, + "learning_rate": 2.2535493265380414e-06, + "loss": 0.0147, + "step": 13720 + }, + { + "epoch": 2.5, + "learning_rate": 2.2534947215143793e-06, + "loss": 0.5918, + "step": 13721 + }, + { + "epoch": 2.5, + "learning_rate": 2.2534401164907172e-06, + "loss": 0.0541, + "step": 13722 + }, + { + "epoch": 2.5, + "learning_rate": 2.2533855114670548e-06, + "loss": 0.0192, + "step": 13723 + }, + { + "epoch": 2.5, + "learning_rate": 2.2533309064433927e-06, + "loss": 0.797, + "step": 13724 + }, + { + "epoch": 2.5, + "learning_rate": 2.2532763014197307e-06, + "loss": 2.0012, + "step": 13725 + }, + { + "epoch": 2.5, + "learning_rate": 2.2532216963960686e-06, + "loss": 0.162, + "step": 13726 + }, + { + "epoch": 2.5, + "learning_rate": 2.2531670913724065e-06, + "loss": 1.3522, + "step": 13727 + }, + { + "epoch": 2.5, + "learning_rate": 2.253112486348744e-06, + "loss": 0.3709, + "step": 13728 + }, + { + "epoch": 2.5, + "learning_rate": 2.253057881325082e-06, + "loss": 0.8456, + "step": 13729 + }, + { + "epoch": 2.5, + "learning_rate": 2.25300327630142e-06, + "loss": 0.5507, + "step": 13730 + }, + { + "epoch": 2.5, + "learning_rate": 2.2529486712777575e-06, + "loss": 3.6399, + "step": 13731 + }, + { + "epoch": 2.5, + "learning_rate": 2.2528940662540954e-06, + "loss": 1.6207, + "step": 13732 + }, + { + "epoch": 2.5, + "learning_rate": 2.2528394612304334e-06, + "loss": 1.1171, + "step": 13733 + }, + { + "epoch": 2.5, + "learning_rate": 2.252784856206771e-06, + "loss": 1.812, + "step": 13734 + }, + { + "epoch": 2.5, + "learning_rate": 2.252730251183109e-06, + "loss": 0.0892, + "step": 13735 + }, + { + "epoch": 2.5, + "learning_rate": 2.2526756461594468e-06, + "loss": 0.3724, + "step": 13736 + }, + { + "epoch": 2.5, + "learning_rate": 2.2526210411357843e-06, + "loss": 0.3681, + "step": 13737 + }, + { + "epoch": 2.5, + "learning_rate": 2.2525664361121222e-06, + "loss": 0.0483, + "step": 13738 + }, + { + "epoch": 2.5, + "learning_rate": 2.25251183108846e-06, + "loss": 1.1774, + "step": 13739 + }, + { + "epoch": 2.5, + "learning_rate": 2.252457226064798e-06, + "loss": 0.2222, + "step": 13740 + }, + { + "epoch": 2.5, + "learning_rate": 2.252402621041136e-06, + "loss": 2.385, + "step": 13741 + }, + { + "epoch": 2.5, + "learning_rate": 2.2523480160174736e-06, + "loss": 1.8909, + "step": 13742 + }, + { + "epoch": 2.5, + "learning_rate": 2.2522934109938115e-06, + "loss": 0.0673, + "step": 13743 + }, + { + "epoch": 2.5, + "learning_rate": 2.2522388059701495e-06, + "loss": 1.9333, + "step": 13744 + }, + { + "epoch": 2.5, + "learning_rate": 2.252184200946487e-06, + "loss": 0.0427, + "step": 13745 + }, + { + "epoch": 2.5, + "learning_rate": 2.252129595922825e-06, + "loss": 0.483, + "step": 13746 + }, + { + "epoch": 2.5, + "learning_rate": 2.252074990899163e-06, + "loss": 0.2591, + "step": 13747 + }, + { + "epoch": 2.5, + "learning_rate": 2.2520203858755004e-06, + "loss": 0.0388, + "step": 13748 + }, + { + "epoch": 2.5, + "learning_rate": 2.2519657808518383e-06, + "loss": 1.6507, + "step": 13749 + }, + { + "epoch": 2.5, + "learning_rate": 2.2519111758281763e-06, + "loss": 0.9064, + "step": 13750 + }, + { + "epoch": 2.5, + "learning_rate": 2.2518565708045142e-06, + "loss": 0.2941, + "step": 13751 + }, + { + "epoch": 2.5, + "learning_rate": 2.251801965780852e-06, + "loss": 0.9096, + "step": 13752 + }, + { + "epoch": 2.5, + "learning_rate": 2.2517473607571897e-06, + "loss": 1.0576, + "step": 13753 + }, + { + "epoch": 2.5, + "learning_rate": 2.2516927557335276e-06, + "loss": 0.9241, + "step": 13754 + }, + { + "epoch": 2.5, + "learning_rate": 2.2516381507098656e-06, + "loss": 0.3825, + "step": 13755 + }, + { + "epoch": 2.5, + "learning_rate": 2.251583545686203e-06, + "loss": 0.0264, + "step": 13756 + }, + { + "epoch": 2.5, + "learning_rate": 2.251528940662541e-06, + "loss": 0.606, + "step": 13757 + }, + { + "epoch": 2.5, + "learning_rate": 2.251474335638879e-06, + "loss": 2.9945, + "step": 13758 + }, + { + "epoch": 2.5, + "learning_rate": 2.2514197306152165e-06, + "loss": 0.0192, + "step": 13759 + }, + { + "epoch": 2.5, + "learning_rate": 2.2513651255915544e-06, + "loss": 0.511, + "step": 13760 + }, + { + "epoch": 2.5, + "learning_rate": 2.2513105205678924e-06, + "loss": 0.4044, + "step": 13761 + }, + { + "epoch": 2.5, + "learning_rate": 2.25125591554423e-06, + "loss": 1.0091, + "step": 13762 + }, + { + "epoch": 2.5, + "learning_rate": 2.251201310520568e-06, + "loss": 0.5059, + "step": 13763 + }, + { + "epoch": 2.5, + "learning_rate": 2.2511467054969058e-06, + "loss": 1.1243, + "step": 13764 + }, + { + "epoch": 2.5, + "learning_rate": 2.2510921004732437e-06, + "loss": 0.0961, + "step": 13765 + }, + { + "epoch": 2.5, + "learning_rate": 2.2510374954495817e-06, + "loss": 0.5182, + "step": 13766 + }, + { + "epoch": 2.5, + "learning_rate": 2.250982890425919e-06, + "loss": 0.0393, + "step": 13767 + }, + { + "epoch": 2.5, + "learning_rate": 2.250928285402257e-06, + "loss": 1.9375, + "step": 13768 + }, + { + "epoch": 2.5, + "learning_rate": 2.250873680378595e-06, + "loss": 0.02, + "step": 13769 + }, + { + "epoch": 2.5, + "learning_rate": 2.2508190753549326e-06, + "loss": 2.7261, + "step": 13770 + }, + { + "epoch": 2.5, + "learning_rate": 2.2507644703312705e-06, + "loss": 2.1779, + "step": 13771 + }, + { + "epoch": 2.5, + "learning_rate": 2.2507098653076085e-06, + "loss": 0.2683, + "step": 13772 + }, + { + "epoch": 2.5, + "learning_rate": 2.250655260283946e-06, + "loss": 2.519, + "step": 13773 + }, + { + "epoch": 2.5, + "learning_rate": 2.250600655260284e-06, + "loss": 0.5664, + "step": 13774 + }, + { + "epoch": 2.51, + "learning_rate": 2.250546050236622e-06, + "loss": 1.112, + "step": 13775 + }, + { + "epoch": 2.51, + "learning_rate": 2.2504914452129594e-06, + "loss": 1.0766, + "step": 13776 + }, + { + "epoch": 2.51, + "learning_rate": 2.2504368401892974e-06, + "loss": 0.2586, + "step": 13777 + }, + { + "epoch": 2.51, + "learning_rate": 2.2503822351656353e-06, + "loss": 1.7218, + "step": 13778 + }, + { + "epoch": 2.51, + "learning_rate": 2.2503276301419732e-06, + "loss": 0.7438, + "step": 13779 + }, + { + "epoch": 2.51, + "learning_rate": 2.250273025118311e-06, + "loss": 2.2217, + "step": 13780 + }, + { + "epoch": 2.51, + "learning_rate": 2.2502184200946487e-06, + "loss": 0.0571, + "step": 13781 + }, + { + "epoch": 2.51, + "learning_rate": 2.2501638150709867e-06, + "loss": 3.1208, + "step": 13782 + }, + { + "epoch": 2.51, + "learning_rate": 2.2501092100473246e-06, + "loss": 0.6736, + "step": 13783 + }, + { + "epoch": 2.51, + "learning_rate": 2.250054605023662e-06, + "loss": 0.1441, + "step": 13784 + }, + { + "epoch": 2.51, + "learning_rate": 2.25e-06, + "loss": 0.745, + "step": 13785 + }, + { + "epoch": 2.51, + "learning_rate": 2.249945394976338e-06, + "loss": 0.2708, + "step": 13786 + }, + { + "epoch": 2.51, + "learning_rate": 2.2498907899526755e-06, + "loss": 1.4572, + "step": 13787 + }, + { + "epoch": 2.51, + "learning_rate": 2.2498361849290135e-06, + "loss": 0.0313, + "step": 13788 + }, + { + "epoch": 2.51, + "learning_rate": 2.2497815799053514e-06, + "loss": 2.1854, + "step": 13789 + }, + { + "epoch": 2.51, + "learning_rate": 2.249726974881689e-06, + "loss": 0.6777, + "step": 13790 + }, + { + "epoch": 2.51, + "learning_rate": 2.249672369858027e-06, + "loss": 0.7984, + "step": 13791 + }, + { + "epoch": 2.51, + "learning_rate": 2.249617764834365e-06, + "loss": 0.008, + "step": 13792 + }, + { + "epoch": 2.51, + "learning_rate": 2.2495631598107028e-06, + "loss": 1.2958, + "step": 13793 + }, + { + "epoch": 2.51, + "learning_rate": 2.2495085547870407e-06, + "loss": 0.1793, + "step": 13794 + }, + { + "epoch": 2.51, + "learning_rate": 2.2494539497633782e-06, + "loss": 2.0078, + "step": 13795 + }, + { + "epoch": 2.51, + "learning_rate": 2.249399344739716e-06, + "loss": 0.4506, + "step": 13796 + }, + { + "epoch": 2.51, + "learning_rate": 2.249344739716054e-06, + "loss": 0.0203, + "step": 13797 + }, + { + "epoch": 2.51, + "learning_rate": 2.2492901346923916e-06, + "loss": 2.2686, + "step": 13798 + }, + { + "epoch": 2.51, + "learning_rate": 2.2492355296687296e-06, + "loss": 2.7051, + "step": 13799 + }, + { + "epoch": 2.51, + "learning_rate": 2.2491809246450675e-06, + "loss": 0.6637, + "step": 13800 + }, + { + "epoch": 2.51, + "learning_rate": 2.249126319621405e-06, + "loss": 1.0662, + "step": 13801 + }, + { + "epoch": 2.51, + "learning_rate": 2.249071714597743e-06, + "loss": 0.4843, + "step": 13802 + }, + { + "epoch": 2.51, + "learning_rate": 2.249017109574081e-06, + "loss": 1.3667, + "step": 13803 + }, + { + "epoch": 2.51, + "learning_rate": 2.2489625045504184e-06, + "loss": 0.036, + "step": 13804 + }, + { + "epoch": 2.51, + "learning_rate": 2.2489078995267564e-06, + "loss": 0.4444, + "step": 13805 + }, + { + "epoch": 2.51, + "learning_rate": 2.2488532945030943e-06, + "loss": 0.8058, + "step": 13806 + }, + { + "epoch": 2.51, + "learning_rate": 2.2487986894794323e-06, + "loss": 0.5068, + "step": 13807 + }, + { + "epoch": 2.51, + "learning_rate": 2.2487440844557702e-06, + "loss": 0.046, + "step": 13808 + }, + { + "epoch": 2.51, + "learning_rate": 2.2486894794321077e-06, + "loss": 0.1398, + "step": 13809 + }, + { + "epoch": 2.51, + "learning_rate": 2.2486348744084457e-06, + "loss": 0.5134, + "step": 13810 + }, + { + "epoch": 2.51, + "learning_rate": 2.2485802693847836e-06, + "loss": 0.8496, + "step": 13811 + }, + { + "epoch": 2.51, + "learning_rate": 2.248525664361121e-06, + "loss": 0.34, + "step": 13812 + }, + { + "epoch": 2.51, + "learning_rate": 2.248471059337459e-06, + "loss": 0.6107, + "step": 13813 + }, + { + "epoch": 2.51, + "learning_rate": 2.248416454313797e-06, + "loss": 0.4317, + "step": 13814 + }, + { + "epoch": 2.51, + "learning_rate": 2.2483618492901345e-06, + "loss": 2.2849, + "step": 13815 + }, + { + "epoch": 2.51, + "learning_rate": 2.2483072442664725e-06, + "loss": 1.9987, + "step": 13816 + }, + { + "epoch": 2.51, + "learning_rate": 2.2482526392428104e-06, + "loss": 0.5206, + "step": 13817 + }, + { + "epoch": 2.51, + "learning_rate": 2.248198034219148e-06, + "loss": 1.878, + "step": 13818 + }, + { + "epoch": 2.51, + "learning_rate": 2.248143429195486e-06, + "loss": 0.1671, + "step": 13819 + }, + { + "epoch": 2.51, + "learning_rate": 2.248088824171824e-06, + "loss": 0.8685, + "step": 13820 + }, + { + "epoch": 2.51, + "learning_rate": 2.2480342191481618e-06, + "loss": 0.4785, + "step": 13821 + }, + { + "epoch": 2.51, + "learning_rate": 2.2479796141244997e-06, + "loss": 0.5659, + "step": 13822 + }, + { + "epoch": 2.51, + "learning_rate": 2.2479250091008372e-06, + "loss": 3.5838, + "step": 13823 + }, + { + "epoch": 2.51, + "learning_rate": 2.247870404077175e-06, + "loss": 0.062, + "step": 13824 + }, + { + "epoch": 2.51, + "learning_rate": 2.247815799053513e-06, + "loss": 0.5954, + "step": 13825 + }, + { + "epoch": 2.51, + "learning_rate": 2.2477611940298507e-06, + "loss": 0.0942, + "step": 13826 + }, + { + "epoch": 2.51, + "learning_rate": 2.2477065890061886e-06, + "loss": 2.344, + "step": 13827 + }, + { + "epoch": 2.51, + "learning_rate": 2.2476519839825265e-06, + "loss": 0.0258, + "step": 13828 + }, + { + "epoch": 2.51, + "learning_rate": 2.247597378958864e-06, + "loss": 0.1541, + "step": 13829 + }, + { + "epoch": 2.52, + "learning_rate": 2.247542773935202e-06, + "loss": 3.1264, + "step": 13830 + }, + { + "epoch": 2.52, + "learning_rate": 2.24748816891154e-06, + "loss": 0.5218, + "step": 13831 + }, + { + "epoch": 2.52, + "learning_rate": 2.247433563887878e-06, + "loss": 0.3134, + "step": 13832 + }, + { + "epoch": 2.52, + "learning_rate": 2.247378958864216e-06, + "loss": 0.0435, + "step": 13833 + }, + { + "epoch": 2.52, + "learning_rate": 2.2473243538405534e-06, + "loss": 0.1741, + "step": 13834 + }, + { + "epoch": 2.52, + "learning_rate": 2.2472697488168913e-06, + "loss": 1.0304, + "step": 13835 + }, + { + "epoch": 2.52, + "learning_rate": 2.2472151437932292e-06, + "loss": 1.1609, + "step": 13836 + }, + { + "epoch": 2.52, + "learning_rate": 2.2471605387695668e-06, + "loss": 0.7444, + "step": 13837 + }, + { + "epoch": 2.52, + "learning_rate": 2.2471059337459047e-06, + "loss": 0.5921, + "step": 13838 + }, + { + "epoch": 2.52, + "learning_rate": 2.2470513287222426e-06, + "loss": 0.4626, + "step": 13839 + }, + { + "epoch": 2.52, + "learning_rate": 2.24699672369858e-06, + "loss": 0.2645, + "step": 13840 + }, + { + "epoch": 2.52, + "learning_rate": 2.246942118674918e-06, + "loss": 1.1398, + "step": 13841 + }, + { + "epoch": 2.52, + "learning_rate": 2.246887513651256e-06, + "loss": 1.5214, + "step": 13842 + }, + { + "epoch": 2.52, + "learning_rate": 2.2468329086275936e-06, + "loss": 0.5986, + "step": 13843 + }, + { + "epoch": 2.52, + "learning_rate": 2.2467783036039315e-06, + "loss": 0.0205, + "step": 13844 + }, + { + "epoch": 2.52, + "learning_rate": 2.2467236985802695e-06, + "loss": 0.22, + "step": 13845 + }, + { + "epoch": 2.52, + "learning_rate": 2.2466690935566074e-06, + "loss": 0.9636, + "step": 13846 + }, + { + "epoch": 2.52, + "learning_rate": 2.2466144885329453e-06, + "loss": 0.4591, + "step": 13847 + }, + { + "epoch": 2.52, + "learning_rate": 2.246559883509283e-06, + "loss": 1.6748, + "step": 13848 + }, + { + "epoch": 2.52, + "learning_rate": 2.246505278485621e-06, + "loss": 0.0616, + "step": 13849 + }, + { + "epoch": 2.52, + "learning_rate": 2.2464506734619588e-06, + "loss": 1.1876, + "step": 13850 + }, + { + "epoch": 2.52, + "learning_rate": 2.2463960684382963e-06, + "loss": 0.0279, + "step": 13851 + }, + { + "epoch": 2.52, + "learning_rate": 2.2463414634146342e-06, + "loss": 0.5325, + "step": 13852 + }, + { + "epoch": 2.52, + "learning_rate": 2.246286858390972e-06, + "loss": 0.4712, + "step": 13853 + }, + { + "epoch": 2.52, + "learning_rate": 2.2462322533673097e-06, + "loss": 0.5796, + "step": 13854 + }, + { + "epoch": 2.52, + "learning_rate": 2.2461776483436476e-06, + "loss": 0.1771, + "step": 13855 + }, + { + "epoch": 2.52, + "learning_rate": 2.2461230433199856e-06, + "loss": 0.8603, + "step": 13856 + }, + { + "epoch": 2.52, + "learning_rate": 2.246068438296323e-06, + "loss": 1.4014, + "step": 13857 + }, + { + "epoch": 2.52, + "learning_rate": 2.246013833272661e-06, + "loss": 0.0215, + "step": 13858 + }, + { + "epoch": 2.52, + "learning_rate": 2.245959228248999e-06, + "loss": 0.8428, + "step": 13859 + }, + { + "epoch": 2.52, + "learning_rate": 2.245904623225337e-06, + "loss": 0.5911, + "step": 13860 + }, + { + "epoch": 2.52, + "learning_rate": 2.245850018201675e-06, + "loss": 0.7004, + "step": 13861 + }, + { + "epoch": 2.52, + "learning_rate": 2.2457954131780124e-06, + "loss": 0.499, + "step": 13862 + }, + { + "epoch": 2.52, + "learning_rate": 2.2457408081543503e-06, + "loss": 0.817, + "step": 13863 + }, + { + "epoch": 2.52, + "learning_rate": 2.2456862031306883e-06, + "loss": 2.5409, + "step": 13864 + }, + { + "epoch": 2.52, + "learning_rate": 2.2456315981070258e-06, + "loss": 0.1311, + "step": 13865 + }, + { + "epoch": 2.52, + "learning_rate": 2.2455769930833637e-06, + "loss": 1.1557, + "step": 13866 + }, + { + "epoch": 2.52, + "learning_rate": 2.2455223880597017e-06, + "loss": 0.43, + "step": 13867 + }, + { + "epoch": 2.52, + "learning_rate": 2.245467783036039e-06, + "loss": 1.9874, + "step": 13868 + }, + { + "epoch": 2.52, + "learning_rate": 2.245413178012377e-06, + "loss": 0.2399, + "step": 13869 + }, + { + "epoch": 2.52, + "learning_rate": 2.245358572988715e-06, + "loss": 2.2143, + "step": 13870 + }, + { + "epoch": 2.52, + "learning_rate": 2.2453039679650526e-06, + "loss": 0.0305, + "step": 13871 + }, + { + "epoch": 2.52, + "learning_rate": 2.2452493629413905e-06, + "loss": 1.2478, + "step": 13872 + }, + { + "epoch": 2.52, + "learning_rate": 2.2451947579177285e-06, + "loss": 0.6594, + "step": 13873 + }, + { + "epoch": 2.52, + "learning_rate": 2.2451401528940664e-06, + "loss": 0.456, + "step": 13874 + }, + { + "epoch": 2.52, + "learning_rate": 2.2450855478704044e-06, + "loss": 2.9839, + "step": 13875 + }, + { + "epoch": 2.52, + "learning_rate": 2.245030942846742e-06, + "loss": 0.2715, + "step": 13876 + }, + { + "epoch": 2.52, + "learning_rate": 2.24497633782308e-06, + "loss": 0.3742, + "step": 13877 + }, + { + "epoch": 2.52, + "learning_rate": 2.2449217327994178e-06, + "loss": 0.021, + "step": 13878 + }, + { + "epoch": 2.52, + "learning_rate": 2.2448671277757553e-06, + "loss": 1.673, + "step": 13879 + }, + { + "epoch": 2.52, + "learning_rate": 2.2448125227520932e-06, + "loss": 0.762, + "step": 13880 + }, + { + "epoch": 2.52, + "learning_rate": 2.244757917728431e-06, + "loss": 0.4586, + "step": 13881 + }, + { + "epoch": 2.52, + "learning_rate": 2.2447033127047687e-06, + "loss": 0.8516, + "step": 13882 + }, + { + "epoch": 2.52, + "learning_rate": 2.2446487076811067e-06, + "loss": 1.9557, + "step": 13883 + }, + { + "epoch": 2.52, + "learning_rate": 2.2445941026574446e-06, + "loss": 2.2834, + "step": 13884 + }, + { + "epoch": 2.53, + "learning_rate": 2.244539497633782e-06, + "loss": 1.3964, + "step": 13885 + }, + { + "epoch": 2.53, + "learning_rate": 2.24448489261012e-06, + "loss": 0.6178, + "step": 13886 + }, + { + "epoch": 2.53, + "learning_rate": 2.244430287586458e-06, + "loss": 1.925, + "step": 13887 + }, + { + "epoch": 2.53, + "learning_rate": 2.244375682562796e-06, + "loss": 0.2837, + "step": 13888 + }, + { + "epoch": 2.53, + "learning_rate": 2.244321077539134e-06, + "loss": 1.1468, + "step": 13889 + }, + { + "epoch": 2.53, + "learning_rate": 2.2442664725154714e-06, + "loss": 1.0138, + "step": 13890 + }, + { + "epoch": 2.53, + "learning_rate": 2.2442118674918094e-06, + "loss": 0.0166, + "step": 13891 + }, + { + "epoch": 2.53, + "learning_rate": 2.2441572624681473e-06, + "loss": 0.5357, + "step": 13892 + }, + { + "epoch": 2.53, + "learning_rate": 2.244102657444485e-06, + "loss": 0.315, + "step": 13893 + }, + { + "epoch": 2.53, + "learning_rate": 2.2440480524208228e-06, + "loss": 0.5447, + "step": 13894 + }, + { + "epoch": 2.53, + "learning_rate": 2.2439934473971607e-06, + "loss": 0.7594, + "step": 13895 + }, + { + "epoch": 2.53, + "learning_rate": 2.2439388423734982e-06, + "loss": 0.4299, + "step": 13896 + }, + { + "epoch": 2.53, + "learning_rate": 2.243884237349836e-06, + "loss": 0.9186, + "step": 13897 + }, + { + "epoch": 2.53, + "learning_rate": 2.243829632326174e-06, + "loss": 0.0261, + "step": 13898 + }, + { + "epoch": 2.53, + "learning_rate": 2.2437750273025116e-06, + "loss": 0.6388, + "step": 13899 + }, + { + "epoch": 2.53, + "learning_rate": 2.2437204222788496e-06, + "loss": 0.2949, + "step": 13900 + }, + { + "epoch": 2.53, + "learning_rate": 2.2436658172551875e-06, + "loss": 1.3737, + "step": 13901 + }, + { + "epoch": 2.53, + "learning_rate": 2.2436112122315255e-06, + "loss": 0.2494, + "step": 13902 + }, + { + "epoch": 2.53, + "learning_rate": 2.2435566072078634e-06, + "loss": 1.1973, + "step": 13903 + }, + { + "epoch": 2.53, + "learning_rate": 2.243502002184201e-06, + "loss": 0.3312, + "step": 13904 + }, + { + "epoch": 2.53, + "learning_rate": 2.243447397160539e-06, + "loss": 2.3609, + "step": 13905 + }, + { + "epoch": 2.53, + "learning_rate": 2.243392792136877e-06, + "loss": 0.023, + "step": 13906 + }, + { + "epoch": 2.53, + "learning_rate": 2.2433381871132143e-06, + "loss": 0.3735, + "step": 13907 + }, + { + "epoch": 2.53, + "learning_rate": 2.2432835820895523e-06, + "loss": 0.0239, + "step": 13908 + }, + { + "epoch": 2.53, + "learning_rate": 2.2432289770658902e-06, + "loss": 1.7925, + "step": 13909 + }, + { + "epoch": 2.53, + "learning_rate": 2.2431743720422277e-06, + "loss": 0.9476, + "step": 13910 + }, + { + "epoch": 2.53, + "learning_rate": 2.2431197670185657e-06, + "loss": 0.1106, + "step": 13911 + }, + { + "epoch": 2.53, + "learning_rate": 2.2430651619949036e-06, + "loss": 0.7642, + "step": 13912 + }, + { + "epoch": 2.53, + "learning_rate": 2.2430105569712416e-06, + "loss": 0.5154, + "step": 13913 + }, + { + "epoch": 2.53, + "learning_rate": 2.2429559519475795e-06, + "loss": 0.9242, + "step": 13914 + }, + { + "epoch": 2.53, + "learning_rate": 2.242901346923917e-06, + "loss": 1.6083, + "step": 13915 + }, + { + "epoch": 2.53, + "learning_rate": 2.242846741900255e-06, + "loss": 0.5237, + "step": 13916 + }, + { + "epoch": 2.53, + "learning_rate": 2.242792136876593e-06, + "loss": 1.0832, + "step": 13917 + }, + { + "epoch": 2.53, + "learning_rate": 2.2427375318529304e-06, + "loss": 0.3457, + "step": 13918 + }, + { + "epoch": 2.53, + "learning_rate": 2.2426829268292684e-06, + "loss": 0.4377, + "step": 13919 + }, + { + "epoch": 2.53, + "learning_rate": 2.2426283218056063e-06, + "loss": 0.0519, + "step": 13920 + }, + { + "epoch": 2.53, + "learning_rate": 2.242573716781944e-06, + "loss": 1.2984, + "step": 13921 + }, + { + "epoch": 2.53, + "learning_rate": 2.2425191117582818e-06, + "loss": 0.1023, + "step": 13922 + }, + { + "epoch": 2.53, + "learning_rate": 2.2424645067346197e-06, + "loss": 0.0172, + "step": 13923 + }, + { + "epoch": 2.53, + "learning_rate": 2.2424099017109572e-06, + "loss": 2.8455, + "step": 13924 + }, + { + "epoch": 2.53, + "learning_rate": 2.242355296687295e-06, + "loss": 1.9767, + "step": 13925 + }, + { + "epoch": 2.53, + "learning_rate": 2.242300691663633e-06, + "loss": 0.1594, + "step": 13926 + }, + { + "epoch": 2.53, + "learning_rate": 2.242246086639971e-06, + "loss": 0.6908, + "step": 13927 + }, + { + "epoch": 2.53, + "learning_rate": 2.242191481616309e-06, + "loss": 0.4349, + "step": 13928 + }, + { + "epoch": 2.53, + "learning_rate": 2.2421368765926465e-06, + "loss": 0.8562, + "step": 13929 + }, + { + "epoch": 2.53, + "learning_rate": 2.2420822715689845e-06, + "loss": 2.485, + "step": 13930 + }, + { + "epoch": 2.53, + "learning_rate": 2.2420276665453224e-06, + "loss": 0.1666, + "step": 13931 + }, + { + "epoch": 2.53, + "learning_rate": 2.24197306152166e-06, + "loss": 1.1321, + "step": 13932 + }, + { + "epoch": 2.53, + "learning_rate": 2.241918456497998e-06, + "loss": 0.1938, + "step": 13933 + }, + { + "epoch": 2.53, + "learning_rate": 2.241863851474336e-06, + "loss": 0.6024, + "step": 13934 + }, + { + "epoch": 2.53, + "learning_rate": 2.2418092464506734e-06, + "loss": 2.4042, + "step": 13935 + }, + { + "epoch": 2.53, + "learning_rate": 2.2417546414270113e-06, + "loss": 0.3278, + "step": 13936 + }, + { + "epoch": 2.53, + "learning_rate": 2.2417000364033492e-06, + "loss": 0.9742, + "step": 13937 + }, + { + "epoch": 2.53, + "learning_rate": 2.2416454313796868e-06, + "loss": 1.5131, + "step": 13938 + }, + { + "epoch": 2.53, + "learning_rate": 2.2415908263560247e-06, + "loss": 0.2136, + "step": 13939 + }, + { + "epoch": 2.54, + "learning_rate": 2.2415362213323626e-06, + "loss": 0.9068, + "step": 13940 + }, + { + "epoch": 2.54, + "learning_rate": 2.2414816163087006e-06, + "loss": 0.4801, + "step": 13941 + }, + { + "epoch": 2.54, + "learning_rate": 2.2414270112850385e-06, + "loss": 2.3805, + "step": 13942 + }, + { + "epoch": 2.54, + "learning_rate": 2.241372406261376e-06, + "loss": 0.2901, + "step": 13943 + }, + { + "epoch": 2.54, + "learning_rate": 2.241317801237714e-06, + "loss": 0.8187, + "step": 13944 + }, + { + "epoch": 2.54, + "learning_rate": 2.241263196214052e-06, + "loss": 0.2959, + "step": 13945 + }, + { + "epoch": 2.54, + "learning_rate": 2.2412085911903895e-06, + "loss": 2.3428, + "step": 13946 + }, + { + "epoch": 2.54, + "learning_rate": 2.2411539861667274e-06, + "loss": 0.274, + "step": 13947 + }, + { + "epoch": 2.54, + "learning_rate": 2.2410993811430653e-06, + "loss": 0.9761, + "step": 13948 + }, + { + "epoch": 2.54, + "learning_rate": 2.241044776119403e-06, + "loss": 0.3223, + "step": 13949 + }, + { + "epoch": 2.54, + "learning_rate": 2.240990171095741e-06, + "loss": 0.0184, + "step": 13950 + }, + { + "epoch": 2.54, + "learning_rate": 2.2409355660720788e-06, + "loss": 1.662, + "step": 13951 + }, + { + "epoch": 2.54, + "learning_rate": 2.2408809610484163e-06, + "loss": 1.5383, + "step": 13952 + }, + { + "epoch": 2.54, + "learning_rate": 2.2408263560247542e-06, + "loss": 0.4669, + "step": 13953 + }, + { + "epoch": 2.54, + "learning_rate": 2.240771751001092e-06, + "loss": 4.1997, + "step": 13954 + }, + { + "epoch": 2.54, + "learning_rate": 2.24071714597743e-06, + "loss": 0.1298, + "step": 13955 + }, + { + "epoch": 2.54, + "learning_rate": 2.240662540953768e-06, + "loss": 2.8858, + "step": 13956 + }, + { + "epoch": 2.54, + "learning_rate": 2.2406079359301056e-06, + "loss": 1.8509, + "step": 13957 + }, + { + "epoch": 2.54, + "learning_rate": 2.2405533309064435e-06, + "loss": 0.4983, + "step": 13958 + }, + { + "epoch": 2.54, + "learning_rate": 2.2404987258827815e-06, + "loss": 0.9808, + "step": 13959 + }, + { + "epoch": 2.54, + "learning_rate": 2.240444120859119e-06, + "loss": 0.3434, + "step": 13960 + }, + { + "epoch": 2.54, + "learning_rate": 2.240389515835457e-06, + "loss": 0.8121, + "step": 13961 + }, + { + "epoch": 2.54, + "learning_rate": 2.240334910811795e-06, + "loss": 1.0549, + "step": 13962 + }, + { + "epoch": 2.54, + "learning_rate": 2.2402803057881324e-06, + "loss": 0.0657, + "step": 13963 + }, + { + "epoch": 2.54, + "learning_rate": 2.2402257007644703e-06, + "loss": 0.0983, + "step": 13964 + }, + { + "epoch": 2.54, + "learning_rate": 2.2401710957408083e-06, + "loss": 0.9575, + "step": 13965 + }, + { + "epoch": 2.54, + "learning_rate": 2.240116490717146e-06, + "loss": 0.2734, + "step": 13966 + }, + { + "epoch": 2.54, + "learning_rate": 2.2400618856934837e-06, + "loss": 0.6507, + "step": 13967 + }, + { + "epoch": 2.54, + "learning_rate": 2.2400072806698217e-06, + "loss": 0.0498, + "step": 13968 + }, + { + "epoch": 2.54, + "learning_rate": 2.2399526756461596e-06, + "loss": 0.1163, + "step": 13969 + }, + { + "epoch": 2.54, + "learning_rate": 2.2398980706224976e-06, + "loss": 0.1473, + "step": 13970 + }, + { + "epoch": 2.54, + "learning_rate": 2.239843465598835e-06, + "loss": 0.3939, + "step": 13971 + }, + { + "epoch": 2.54, + "learning_rate": 2.239788860575173e-06, + "loss": 1.1106, + "step": 13972 + }, + { + "epoch": 2.54, + "learning_rate": 2.239734255551511e-06, + "loss": 5.4265, + "step": 13973 + }, + { + "epoch": 2.54, + "learning_rate": 2.2396796505278485e-06, + "loss": 0.8853, + "step": 13974 + }, + { + "epoch": 2.54, + "learning_rate": 2.2396250455041864e-06, + "loss": 0.7157, + "step": 13975 + }, + { + "epoch": 2.54, + "learning_rate": 2.2395704404805244e-06, + "loss": 0.223, + "step": 13976 + }, + { + "epoch": 2.54, + "learning_rate": 2.239515835456862e-06, + "loss": 0.528, + "step": 13977 + }, + { + "epoch": 2.54, + "learning_rate": 2.2394612304332e-06, + "loss": 0.288, + "step": 13978 + }, + { + "epoch": 2.54, + "learning_rate": 2.2394066254095378e-06, + "loss": 0.0733, + "step": 13979 + }, + { + "epoch": 2.54, + "learning_rate": 2.2393520203858753e-06, + "loss": 0.6542, + "step": 13980 + }, + { + "epoch": 2.54, + "learning_rate": 2.2392974153622132e-06, + "loss": 0.9322, + "step": 13981 + }, + { + "epoch": 2.54, + "learning_rate": 2.239242810338551e-06, + "loss": 1.1096, + "step": 13982 + }, + { + "epoch": 2.54, + "learning_rate": 2.239188205314889e-06, + "loss": 0.2216, + "step": 13983 + }, + { + "epoch": 2.54, + "learning_rate": 2.239133600291227e-06, + "loss": 2.3014, + "step": 13984 + }, + { + "epoch": 2.54, + "learning_rate": 2.2390789952675646e-06, + "loss": 0.8294, + "step": 13985 + }, + { + "epoch": 2.54, + "learning_rate": 2.2390243902439025e-06, + "loss": 0.3971, + "step": 13986 + }, + { + "epoch": 2.54, + "learning_rate": 2.2389697852202405e-06, + "loss": 1.1594, + "step": 13987 + }, + { + "epoch": 2.54, + "learning_rate": 2.238915180196578e-06, + "loss": 0.6271, + "step": 13988 + }, + { + "epoch": 2.54, + "learning_rate": 2.238860575172916e-06, + "loss": 0.3929, + "step": 13989 + }, + { + "epoch": 2.54, + "learning_rate": 2.238805970149254e-06, + "loss": 0.8698, + "step": 13990 + }, + { + "epoch": 2.54, + "learning_rate": 2.2387513651255914e-06, + "loss": 0.039, + "step": 13991 + }, + { + "epoch": 2.54, + "learning_rate": 2.2386967601019294e-06, + "loss": 0.5231, + "step": 13992 + }, + { + "epoch": 2.54, + "learning_rate": 2.2386421550782673e-06, + "loss": 0.8254, + "step": 13993 + }, + { + "epoch": 2.54, + "learning_rate": 2.238587550054605e-06, + "loss": 1.8054, + "step": 13994 + }, + { + "epoch": 2.55, + "learning_rate": 2.2385329450309428e-06, + "loss": 0.3392, + "step": 13995 + }, + { + "epoch": 2.55, + "learning_rate": 2.2384783400072807e-06, + "loss": 0.928, + "step": 13996 + }, + { + "epoch": 2.55, + "learning_rate": 2.2384237349836186e-06, + "loss": 3.4184, + "step": 13997 + }, + { + "epoch": 2.55, + "learning_rate": 2.2383691299599566e-06, + "loss": 0.3703, + "step": 13998 + }, + { + "epoch": 2.55, + "learning_rate": 2.238314524936294e-06, + "loss": 0.6697, + "step": 13999 + }, + { + "epoch": 2.55, + "learning_rate": 2.238259919912632e-06, + "loss": 0.0113, + "step": 14000 + }, + { + "epoch": 2.55, + "learning_rate": 2.23820531488897e-06, + "loss": 0.809, + "step": 14001 + }, + { + "epoch": 2.55, + "learning_rate": 2.2381507098653075e-06, + "loss": 0.7725, + "step": 14002 + }, + { + "epoch": 2.55, + "learning_rate": 2.2380961048416455e-06, + "loss": 0.5936, + "step": 14003 + }, + { + "epoch": 2.55, + "learning_rate": 2.2380414998179834e-06, + "loss": 0.5857, + "step": 14004 + }, + { + "epoch": 2.55, + "learning_rate": 2.237986894794321e-06, + "loss": 0.5408, + "step": 14005 + }, + { + "epoch": 2.55, + "learning_rate": 2.237932289770659e-06, + "loss": 1.2355, + "step": 14006 + }, + { + "epoch": 2.55, + "learning_rate": 2.237877684746997e-06, + "loss": 2.4824, + "step": 14007 + }, + { + "epoch": 2.55, + "learning_rate": 2.2378230797233348e-06, + "loss": 0.9531, + "step": 14008 + }, + { + "epoch": 2.55, + "learning_rate": 2.2377684746996727e-06, + "loss": 0.606, + "step": 14009 + }, + { + "epoch": 2.55, + "learning_rate": 2.2377138696760102e-06, + "loss": 1.9438, + "step": 14010 + }, + { + "epoch": 2.55, + "learning_rate": 2.237659264652348e-06, + "loss": 1.5267, + "step": 14011 + }, + { + "epoch": 2.55, + "learning_rate": 2.237604659628686e-06, + "loss": 0.5768, + "step": 14012 + }, + { + "epoch": 2.55, + "learning_rate": 2.2375500546050236e-06, + "loss": 0.0636, + "step": 14013 + }, + { + "epoch": 2.55, + "learning_rate": 2.2374954495813616e-06, + "loss": 0.4226, + "step": 14014 + }, + { + "epoch": 2.55, + "learning_rate": 2.2374408445576995e-06, + "loss": 0.9316, + "step": 14015 + }, + { + "epoch": 2.55, + "learning_rate": 2.237386239534037e-06, + "loss": 0.8174, + "step": 14016 + }, + { + "epoch": 2.55, + "learning_rate": 2.237331634510375e-06, + "loss": 2.4232, + "step": 14017 + }, + { + "epoch": 2.55, + "learning_rate": 2.237277029486713e-06, + "loss": 0.0592, + "step": 14018 + }, + { + "epoch": 2.55, + "learning_rate": 2.2372224244630504e-06, + "loss": 1.31, + "step": 14019 + }, + { + "epoch": 2.55, + "learning_rate": 2.2371678194393884e-06, + "loss": 0.5725, + "step": 14020 + }, + { + "epoch": 2.55, + "learning_rate": 2.2371132144157263e-06, + "loss": 1.8038, + "step": 14021 + }, + { + "epoch": 2.55, + "learning_rate": 2.2370586093920643e-06, + "loss": 0.2342, + "step": 14022 + }, + { + "epoch": 2.55, + "learning_rate": 2.237004004368402e-06, + "loss": 1.0021, + "step": 14023 + }, + { + "epoch": 2.55, + "learning_rate": 2.2369493993447397e-06, + "loss": 1.0247, + "step": 14024 + }, + { + "epoch": 2.55, + "learning_rate": 2.2368947943210777e-06, + "loss": 0.51, + "step": 14025 + }, + { + "epoch": 2.55, + "learning_rate": 2.2368401892974156e-06, + "loss": 1.442, + "step": 14026 + }, + { + "epoch": 2.55, + "learning_rate": 2.236785584273753e-06, + "loss": 0.3659, + "step": 14027 + }, + { + "epoch": 2.55, + "learning_rate": 2.236730979250091e-06, + "loss": 1.1358, + "step": 14028 + }, + { + "epoch": 2.55, + "learning_rate": 2.236676374226429e-06, + "loss": 0.0553, + "step": 14029 + }, + { + "epoch": 2.55, + "learning_rate": 2.2366217692027665e-06, + "loss": 0.0884, + "step": 14030 + }, + { + "epoch": 2.55, + "learning_rate": 2.2365671641791045e-06, + "loss": 2.2203, + "step": 14031 + }, + { + "epoch": 2.55, + "learning_rate": 2.2365125591554424e-06, + "loss": 0.2608, + "step": 14032 + }, + { + "epoch": 2.55, + "learning_rate": 2.23645795413178e-06, + "loss": 0.5779, + "step": 14033 + }, + { + "epoch": 2.55, + "learning_rate": 2.236403349108118e-06, + "loss": 2.5193, + "step": 14034 + }, + { + "epoch": 2.55, + "learning_rate": 2.236348744084456e-06, + "loss": 1.2425, + "step": 14035 + }, + { + "epoch": 2.55, + "learning_rate": 2.2362941390607938e-06, + "loss": 0.6906, + "step": 14036 + }, + { + "epoch": 2.55, + "learning_rate": 2.2362395340371317e-06, + "loss": 0.3948, + "step": 14037 + }, + { + "epoch": 2.55, + "learning_rate": 2.2361849290134692e-06, + "loss": 0.0522, + "step": 14038 + }, + { + "epoch": 2.55, + "learning_rate": 2.236130323989807e-06, + "loss": 0.7752, + "step": 14039 + }, + { + "epoch": 2.55, + "learning_rate": 2.236075718966145e-06, + "loss": 0.0315, + "step": 14040 + }, + { + "epoch": 2.55, + "learning_rate": 2.2360211139424826e-06, + "loss": 2.2188, + "step": 14041 + }, + { + "epoch": 2.55, + "learning_rate": 2.2359665089188206e-06, + "loss": 0.3168, + "step": 14042 + }, + { + "epoch": 2.55, + "learning_rate": 2.2359119038951585e-06, + "loss": 0.0418, + "step": 14043 + }, + { + "epoch": 2.55, + "learning_rate": 2.235857298871496e-06, + "loss": 1.548, + "step": 14044 + }, + { + "epoch": 2.55, + "learning_rate": 2.235802693847834e-06, + "loss": 0.9169, + "step": 14045 + }, + { + "epoch": 2.55, + "learning_rate": 2.235748088824172e-06, + "loss": 0.6158, + "step": 14046 + }, + { + "epoch": 2.55, + "learning_rate": 2.2356934838005095e-06, + "loss": 0.966, + "step": 14047 + }, + { + "epoch": 2.55, + "learning_rate": 2.2356388787768474e-06, + "loss": 1.3197, + "step": 14048 + }, + { + "epoch": 2.55, + "learning_rate": 2.2355842737531853e-06, + "loss": 0.2944, + "step": 14049 + }, + { + "epoch": 2.56, + "learning_rate": 2.2355296687295233e-06, + "loss": 0.0185, + "step": 14050 + }, + { + "epoch": 2.56, + "learning_rate": 2.2354750637058612e-06, + "loss": 0.5177, + "step": 14051 + }, + { + "epoch": 2.56, + "learning_rate": 2.2354204586821988e-06, + "loss": 0.7614, + "step": 14052 + }, + { + "epoch": 2.56, + "learning_rate": 2.2353658536585367e-06, + "loss": 0.9248, + "step": 14053 + }, + { + "epoch": 2.56, + "learning_rate": 2.2353112486348746e-06, + "loss": 0.5331, + "step": 14054 + }, + { + "epoch": 2.56, + "learning_rate": 2.235256643611212e-06, + "loss": 1.0446, + "step": 14055 + }, + { + "epoch": 2.56, + "learning_rate": 2.23520203858755e-06, + "loss": 0.0644, + "step": 14056 + }, + { + "epoch": 2.56, + "learning_rate": 2.235147433563888e-06, + "loss": 0.0075, + "step": 14057 + }, + { + "epoch": 2.56, + "learning_rate": 2.2350928285402256e-06, + "loss": 0.4821, + "step": 14058 + }, + { + "epoch": 2.56, + "learning_rate": 2.2350382235165635e-06, + "loss": 1.4174, + "step": 14059 + }, + { + "epoch": 2.56, + "learning_rate": 2.2349836184929015e-06, + "loss": 1.5364, + "step": 14060 + }, + { + "epoch": 2.56, + "learning_rate": 2.234929013469239e-06, + "loss": 2.2728, + "step": 14061 + }, + { + "epoch": 2.56, + "learning_rate": 2.234874408445577e-06, + "loss": 1.6783, + "step": 14062 + }, + { + "epoch": 2.56, + "learning_rate": 2.234819803421915e-06, + "loss": 0.3104, + "step": 14063 + }, + { + "epoch": 2.56, + "learning_rate": 2.234765198398253e-06, + "loss": 1.3466, + "step": 14064 + }, + { + "epoch": 2.56, + "learning_rate": 2.2347105933745908e-06, + "loss": 1.0992, + "step": 14065 + }, + { + "epoch": 2.56, + "learning_rate": 2.2346559883509283e-06, + "loss": 0.0684, + "step": 14066 + }, + { + "epoch": 2.56, + "learning_rate": 2.2346013833272662e-06, + "loss": 0.4387, + "step": 14067 + }, + { + "epoch": 2.56, + "learning_rate": 2.234546778303604e-06, + "loss": 0.5585, + "step": 14068 + }, + { + "epoch": 2.56, + "learning_rate": 2.2344921732799417e-06, + "loss": 0.0399, + "step": 14069 + }, + { + "epoch": 2.56, + "learning_rate": 2.2344375682562796e-06, + "loss": 0.5185, + "step": 14070 + }, + { + "epoch": 2.56, + "learning_rate": 2.2343829632326176e-06, + "loss": 0.23, + "step": 14071 + }, + { + "epoch": 2.56, + "learning_rate": 2.234328358208955e-06, + "loss": 1.4461, + "step": 14072 + }, + { + "epoch": 2.56, + "learning_rate": 2.234273753185293e-06, + "loss": 2.3045, + "step": 14073 + }, + { + "epoch": 2.56, + "learning_rate": 2.234219148161631e-06, + "loss": 0.5141, + "step": 14074 + }, + { + "epoch": 2.56, + "learning_rate": 2.2341645431379685e-06, + "loss": 0.0342, + "step": 14075 + }, + { + "epoch": 2.56, + "learning_rate": 2.2341099381143064e-06, + "loss": 0.3349, + "step": 14076 + }, + { + "epoch": 2.56, + "learning_rate": 2.2340553330906444e-06, + "loss": 0.9404, + "step": 14077 + }, + { + "epoch": 2.56, + "learning_rate": 2.2340007280669823e-06, + "loss": 1.3859, + "step": 14078 + }, + { + "epoch": 2.56, + "learning_rate": 2.2339461230433203e-06, + "loss": 1.2594, + "step": 14079 + }, + { + "epoch": 2.56, + "learning_rate": 2.2338915180196578e-06, + "loss": 0.1943, + "step": 14080 + }, + { + "epoch": 2.56, + "learning_rate": 2.2338369129959957e-06, + "loss": 0.2616, + "step": 14081 + }, + { + "epoch": 2.56, + "learning_rate": 2.2337823079723337e-06, + "loss": 1.4379, + "step": 14082 + }, + { + "epoch": 2.56, + "learning_rate": 2.233727702948671e-06, + "loss": 1.356, + "step": 14083 + }, + { + "epoch": 2.56, + "learning_rate": 2.233673097925009e-06, + "loss": 0.5546, + "step": 14084 + }, + { + "epoch": 2.56, + "learning_rate": 2.233618492901347e-06, + "loss": 2.8196, + "step": 14085 + }, + { + "epoch": 2.56, + "learning_rate": 2.2335638878776846e-06, + "loss": 0.6818, + "step": 14086 + }, + { + "epoch": 2.56, + "learning_rate": 2.2335092828540225e-06, + "loss": 2.3837, + "step": 14087 + }, + { + "epoch": 2.56, + "learning_rate": 2.2334546778303605e-06, + "loss": 0.436, + "step": 14088 + }, + { + "epoch": 2.56, + "learning_rate": 2.2334000728066984e-06, + "loss": 0.9476, + "step": 14089 + }, + { + "epoch": 2.56, + "learning_rate": 2.2333454677830364e-06, + "loss": 1.0508, + "step": 14090 + }, + { + "epoch": 2.56, + "learning_rate": 2.233290862759374e-06, + "loss": 0.021, + "step": 14091 + }, + { + "epoch": 2.56, + "learning_rate": 2.233236257735712e-06, + "loss": 0.0924, + "step": 14092 + }, + { + "epoch": 2.56, + "learning_rate": 2.2331816527120498e-06, + "loss": 0.5117, + "step": 14093 + }, + { + "epoch": 2.56, + "learning_rate": 2.2331270476883873e-06, + "loss": 1.0196, + "step": 14094 + }, + { + "epoch": 2.56, + "learning_rate": 2.2330724426647252e-06, + "loss": 0.6646, + "step": 14095 + }, + { + "epoch": 2.56, + "learning_rate": 2.233017837641063e-06, + "loss": 0.2375, + "step": 14096 + }, + { + "epoch": 2.56, + "learning_rate": 2.2329632326174007e-06, + "loss": 0.2394, + "step": 14097 + }, + { + "epoch": 2.56, + "learning_rate": 2.2329086275937386e-06, + "loss": 0.6454, + "step": 14098 + }, + { + "epoch": 2.56, + "learning_rate": 2.2328540225700766e-06, + "loss": 0.4762, + "step": 14099 + }, + { + "epoch": 2.56, + "learning_rate": 2.232799417546414e-06, + "loss": 0.8032, + "step": 14100 + }, + { + "epoch": 2.56, + "learning_rate": 2.232744812522752e-06, + "loss": 0.478, + "step": 14101 + }, + { + "epoch": 2.56, + "learning_rate": 2.23269020749909e-06, + "loss": 0.6569, + "step": 14102 + }, + { + "epoch": 2.56, + "learning_rate": 2.232635602475428e-06, + "loss": 1.4251, + "step": 14103 + }, + { + "epoch": 2.56, + "learning_rate": 2.232580997451766e-06, + "loss": 0.0956, + "step": 14104 + }, + { + "epoch": 2.57, + "learning_rate": 2.2325263924281034e-06, + "loss": 0.8661, + "step": 14105 + }, + { + "epoch": 2.57, + "learning_rate": 2.2324717874044413e-06, + "loss": 0.8541, + "step": 14106 + }, + { + "epoch": 2.57, + "learning_rate": 2.2324171823807793e-06, + "loss": 1.0555, + "step": 14107 + }, + { + "epoch": 2.57, + "learning_rate": 2.232362577357117e-06, + "loss": 0.4038, + "step": 14108 + }, + { + "epoch": 2.57, + "learning_rate": 2.2323079723334548e-06, + "loss": 0.4432, + "step": 14109 + }, + { + "epoch": 2.57, + "learning_rate": 2.2322533673097927e-06, + "loss": 2.2275, + "step": 14110 + }, + { + "epoch": 2.57, + "learning_rate": 2.2321987622861302e-06, + "loss": 0.0338, + "step": 14111 + }, + { + "epoch": 2.57, + "learning_rate": 2.232144157262468e-06, + "loss": 1.2763, + "step": 14112 + }, + { + "epoch": 2.57, + "learning_rate": 2.232089552238806e-06, + "loss": 0.0861, + "step": 14113 + }, + { + "epoch": 2.57, + "learning_rate": 2.2320349472151436e-06, + "loss": 1.9599, + "step": 14114 + }, + { + "epoch": 2.57, + "learning_rate": 2.2319803421914816e-06, + "loss": 0.441, + "step": 14115 + }, + { + "epoch": 2.57, + "learning_rate": 2.2319257371678195e-06, + "loss": 0.7135, + "step": 14116 + }, + { + "epoch": 2.57, + "learning_rate": 2.2318711321441575e-06, + "loss": 0.1327, + "step": 14117 + }, + { + "epoch": 2.57, + "learning_rate": 2.2318165271204954e-06, + "loss": 0.0239, + "step": 14118 + }, + { + "epoch": 2.57, + "learning_rate": 2.231761922096833e-06, + "loss": 1.315, + "step": 14119 + }, + { + "epoch": 2.57, + "learning_rate": 2.231707317073171e-06, + "loss": 0.9936, + "step": 14120 + }, + { + "epoch": 2.57, + "learning_rate": 2.231652712049509e-06, + "loss": 2.1155, + "step": 14121 + }, + { + "epoch": 2.57, + "learning_rate": 2.2315981070258463e-06, + "loss": 0.0288, + "step": 14122 + }, + { + "epoch": 2.57, + "learning_rate": 2.2315435020021843e-06, + "loss": 0.5067, + "step": 14123 + }, + { + "epoch": 2.57, + "learning_rate": 2.231488896978522e-06, + "loss": 0.6412, + "step": 14124 + }, + { + "epoch": 2.57, + "learning_rate": 2.2314342919548597e-06, + "loss": 0.3414, + "step": 14125 + }, + { + "epoch": 2.57, + "learning_rate": 2.2313796869311977e-06, + "loss": 2.5526, + "step": 14126 + }, + { + "epoch": 2.57, + "learning_rate": 2.2313250819075356e-06, + "loss": 1.1171, + "step": 14127 + }, + { + "epoch": 2.57, + "learning_rate": 2.231270476883873e-06, + "loss": 1.9823, + "step": 14128 + }, + { + "epoch": 2.57, + "learning_rate": 2.231215871860211e-06, + "loss": 2.0017, + "step": 14129 + }, + { + "epoch": 2.57, + "learning_rate": 2.231161266836549e-06, + "loss": 0.151, + "step": 14130 + }, + { + "epoch": 2.57, + "learning_rate": 2.231106661812887e-06, + "loss": 2.4982, + "step": 14131 + }, + { + "epoch": 2.57, + "learning_rate": 2.231052056789225e-06, + "loss": 0.926, + "step": 14132 + }, + { + "epoch": 2.57, + "learning_rate": 2.2309974517655624e-06, + "loss": 1.6568, + "step": 14133 + }, + { + "epoch": 2.57, + "learning_rate": 2.2309428467419004e-06, + "loss": 0.9742, + "step": 14134 + }, + { + "epoch": 2.57, + "learning_rate": 2.2308882417182383e-06, + "loss": 0.4077, + "step": 14135 + }, + { + "epoch": 2.57, + "learning_rate": 2.230833636694576e-06, + "loss": 0.3009, + "step": 14136 + }, + { + "epoch": 2.57, + "learning_rate": 2.2307790316709138e-06, + "loss": 1.3856, + "step": 14137 + }, + { + "epoch": 2.57, + "learning_rate": 2.2307244266472517e-06, + "loss": 1.1333, + "step": 14138 + }, + { + "epoch": 2.57, + "learning_rate": 2.2306698216235892e-06, + "loss": 0.7871, + "step": 14139 + }, + { + "epoch": 2.57, + "learning_rate": 2.230615216599927e-06, + "loss": 0.5066, + "step": 14140 + }, + { + "epoch": 2.57, + "learning_rate": 2.230560611576265e-06, + "loss": 1.2785, + "step": 14141 + }, + { + "epoch": 2.57, + "learning_rate": 2.2305060065526027e-06, + "loss": 0.4118, + "step": 14142 + }, + { + "epoch": 2.57, + "learning_rate": 2.2304514015289406e-06, + "loss": 2.5887, + "step": 14143 + }, + { + "epoch": 2.57, + "learning_rate": 2.2303967965052785e-06, + "loss": 1.6373, + "step": 14144 + }, + { + "epoch": 2.57, + "learning_rate": 2.2303421914816165e-06, + "loss": 1.0185, + "step": 14145 + }, + { + "epoch": 2.57, + "learning_rate": 2.2302875864579544e-06, + "loss": 0.5627, + "step": 14146 + }, + { + "epoch": 2.57, + "learning_rate": 2.230232981434292e-06, + "loss": 1.3397, + "step": 14147 + }, + { + "epoch": 2.57, + "learning_rate": 2.23017837641063e-06, + "loss": 1.1686, + "step": 14148 + }, + { + "epoch": 2.57, + "learning_rate": 2.230123771386968e-06, + "loss": 0.4033, + "step": 14149 + }, + { + "epoch": 2.57, + "learning_rate": 2.2300691663633054e-06, + "loss": 0.6297, + "step": 14150 + }, + { + "epoch": 2.57, + "learning_rate": 2.2300145613396433e-06, + "loss": 0.4249, + "step": 14151 + }, + { + "epoch": 2.57, + "learning_rate": 2.2299599563159812e-06, + "loss": 1.1736, + "step": 14152 + }, + { + "epoch": 2.57, + "learning_rate": 2.2299053512923188e-06, + "loss": 0.0251, + "step": 14153 + }, + { + "epoch": 2.57, + "learning_rate": 2.2298507462686567e-06, + "loss": 1.1696, + "step": 14154 + }, + { + "epoch": 2.57, + "learning_rate": 2.2297961412449946e-06, + "loss": 0.3147, + "step": 14155 + }, + { + "epoch": 2.57, + "learning_rate": 2.229741536221332e-06, + "loss": 2.3679, + "step": 14156 + }, + { + "epoch": 2.57, + "learning_rate": 2.22968693119767e-06, + "loss": 0.8453, + "step": 14157 + }, + { + "epoch": 2.57, + "learning_rate": 2.229632326174008e-06, + "loss": 0.5461, + "step": 14158 + }, + { + "epoch": 2.57, + "learning_rate": 2.229577721150346e-06, + "loss": 2.5496, + "step": 14159 + }, + { + "epoch": 2.58, + "learning_rate": 2.229523116126684e-06, + "loss": 1.9249, + "step": 14160 + }, + { + "epoch": 2.58, + "learning_rate": 2.2294685111030215e-06, + "loss": 0.0444, + "step": 14161 + }, + { + "epoch": 2.58, + "learning_rate": 2.2294139060793594e-06, + "loss": 0.2266, + "step": 14162 + }, + { + "epoch": 2.58, + "learning_rate": 2.2293593010556973e-06, + "loss": 1.0924, + "step": 14163 + }, + { + "epoch": 2.58, + "learning_rate": 2.229304696032035e-06, + "loss": 0.2815, + "step": 14164 + }, + { + "epoch": 2.58, + "learning_rate": 2.229250091008373e-06, + "loss": 0.6294, + "step": 14165 + }, + { + "epoch": 2.58, + "learning_rate": 2.2291954859847108e-06, + "loss": 0.1148, + "step": 14166 + }, + { + "epoch": 2.58, + "learning_rate": 2.2291408809610483e-06, + "loss": 0.036, + "step": 14167 + }, + { + "epoch": 2.58, + "learning_rate": 2.2290862759373862e-06, + "loss": 0.5224, + "step": 14168 + }, + { + "epoch": 2.58, + "learning_rate": 2.229031670913724e-06, + "loss": 0.0608, + "step": 14169 + }, + { + "epoch": 2.58, + "learning_rate": 2.228977065890062e-06, + "loss": 2.6024, + "step": 14170 + }, + { + "epoch": 2.58, + "learning_rate": 2.2289224608664e-06, + "loss": 1.6776, + "step": 14171 + }, + { + "epoch": 2.58, + "learning_rate": 2.2288678558427376e-06, + "loss": 0.7629, + "step": 14172 + }, + { + "epoch": 2.58, + "learning_rate": 2.2288132508190755e-06, + "loss": 2.5672, + "step": 14173 + }, + { + "epoch": 2.58, + "learning_rate": 2.2287586457954135e-06, + "loss": 0.4272, + "step": 14174 + }, + { + "epoch": 2.58, + "learning_rate": 2.228704040771751e-06, + "loss": 1.1442, + "step": 14175 + }, + { + "epoch": 2.58, + "learning_rate": 2.228649435748089e-06, + "loss": 0.7164, + "step": 14176 + }, + { + "epoch": 2.58, + "learning_rate": 2.228594830724427e-06, + "loss": 1.5596, + "step": 14177 + }, + { + "epoch": 2.58, + "learning_rate": 2.2285402257007644e-06, + "loss": 0.4052, + "step": 14178 + }, + { + "epoch": 2.58, + "learning_rate": 2.2284856206771023e-06, + "loss": 0.1593, + "step": 14179 + }, + { + "epoch": 2.58, + "learning_rate": 2.2284310156534403e-06, + "loss": 0.8326, + "step": 14180 + }, + { + "epoch": 2.58, + "learning_rate": 2.2283764106297778e-06, + "loss": 0.8412, + "step": 14181 + }, + { + "epoch": 2.58, + "learning_rate": 2.2283218056061157e-06, + "loss": 0.8541, + "step": 14182 + }, + { + "epoch": 2.58, + "learning_rate": 2.2282672005824537e-06, + "loss": 2.6617, + "step": 14183 + }, + { + "epoch": 2.58, + "learning_rate": 2.2282125955587916e-06, + "loss": 0.683, + "step": 14184 + }, + { + "epoch": 2.58, + "learning_rate": 2.2281579905351296e-06, + "loss": 0.2026, + "step": 14185 + }, + { + "epoch": 2.58, + "learning_rate": 2.228103385511467e-06, + "loss": 0.5845, + "step": 14186 + }, + { + "epoch": 2.58, + "learning_rate": 2.228048780487805e-06, + "loss": 0.545, + "step": 14187 + }, + { + "epoch": 2.58, + "learning_rate": 2.227994175464143e-06, + "loss": 1.8174, + "step": 14188 + }, + { + "epoch": 2.58, + "learning_rate": 2.2279395704404805e-06, + "loss": 0.4397, + "step": 14189 + }, + { + "epoch": 2.58, + "learning_rate": 2.2278849654168184e-06, + "loss": 0.2664, + "step": 14190 + }, + { + "epoch": 2.58, + "learning_rate": 2.2278303603931564e-06, + "loss": 0.2694, + "step": 14191 + }, + { + "epoch": 2.58, + "learning_rate": 2.227775755369494e-06, + "loss": 0.153, + "step": 14192 + }, + { + "epoch": 2.58, + "learning_rate": 2.227721150345832e-06, + "loss": 1.9766, + "step": 14193 + }, + { + "epoch": 2.58, + "learning_rate": 2.2276665453221698e-06, + "loss": 0.0132, + "step": 14194 + }, + { + "epoch": 2.58, + "learning_rate": 2.2276119402985073e-06, + "loss": 0.8126, + "step": 14195 + }, + { + "epoch": 2.58, + "learning_rate": 2.2275573352748452e-06, + "loss": 0.848, + "step": 14196 + }, + { + "epoch": 2.58, + "learning_rate": 2.227502730251183e-06, + "loss": 0.9195, + "step": 14197 + }, + { + "epoch": 2.58, + "learning_rate": 2.227448125227521e-06, + "loss": 0.8987, + "step": 14198 + }, + { + "epoch": 2.58, + "learning_rate": 2.227393520203859e-06, + "loss": 1.2417, + "step": 14199 + }, + { + "epoch": 2.58, + "learning_rate": 2.2273389151801966e-06, + "loss": 1.0729, + "step": 14200 + }, + { + "epoch": 2.58, + "learning_rate": 2.2272843101565345e-06, + "loss": 0.0192, + "step": 14201 + }, + { + "epoch": 2.58, + "learning_rate": 2.2272297051328725e-06, + "loss": 0.6197, + "step": 14202 + }, + { + "epoch": 2.58, + "learning_rate": 2.22717510010921e-06, + "loss": 1.1048, + "step": 14203 + }, + { + "epoch": 2.58, + "learning_rate": 2.227120495085548e-06, + "loss": 0.0823, + "step": 14204 + }, + { + "epoch": 2.58, + "learning_rate": 2.227065890061886e-06, + "loss": 1.4341, + "step": 14205 + }, + { + "epoch": 2.58, + "learning_rate": 2.2270112850382234e-06, + "loss": 0.4165, + "step": 14206 + }, + { + "epoch": 2.58, + "learning_rate": 2.2269566800145613e-06, + "loss": 0.6335, + "step": 14207 + }, + { + "epoch": 2.58, + "learning_rate": 2.2269020749908993e-06, + "loss": 1.8612, + "step": 14208 + }, + { + "epoch": 2.58, + "learning_rate": 2.226847469967237e-06, + "loss": 1.4407, + "step": 14209 + }, + { + "epoch": 2.58, + "learning_rate": 2.2267928649435748e-06, + "loss": 0.5401, + "step": 14210 + }, + { + "epoch": 2.58, + "learning_rate": 2.2267382599199127e-06, + "loss": 2.0024, + "step": 14211 + }, + { + "epoch": 2.58, + "learning_rate": 2.2266836548962506e-06, + "loss": 0.3453, + "step": 14212 + }, + { + "epoch": 2.58, + "learning_rate": 2.2266290498725886e-06, + "loss": 1.4362, + "step": 14213 + }, + { + "epoch": 2.58, + "learning_rate": 2.226574444848926e-06, + "loss": 0.0101, + "step": 14214 + }, + { + "epoch": 2.59, + "learning_rate": 2.226519839825264e-06, + "loss": 0.2335, + "step": 14215 + }, + { + "epoch": 2.59, + "learning_rate": 2.226465234801602e-06, + "loss": 0.7918, + "step": 14216 + }, + { + "epoch": 2.59, + "learning_rate": 2.2264106297779395e-06, + "loss": 0.9055, + "step": 14217 + }, + { + "epoch": 2.59, + "learning_rate": 2.2263560247542775e-06, + "loss": 2.5488, + "step": 14218 + }, + { + "epoch": 2.59, + "learning_rate": 2.2263014197306154e-06, + "loss": 0.447, + "step": 14219 + }, + { + "epoch": 2.59, + "learning_rate": 2.226246814706953e-06, + "loss": 0.0135, + "step": 14220 + }, + { + "epoch": 2.59, + "learning_rate": 2.226192209683291e-06, + "loss": 1.6492, + "step": 14221 + }, + { + "epoch": 2.59, + "learning_rate": 2.226137604659629e-06, + "loss": 0.5652, + "step": 14222 + }, + { + "epoch": 2.59, + "learning_rate": 2.2260829996359663e-06, + "loss": 0.7123, + "step": 14223 + }, + { + "epoch": 2.59, + "learning_rate": 2.2260283946123043e-06, + "loss": 0.449, + "step": 14224 + }, + { + "epoch": 2.59, + "learning_rate": 2.225973789588642e-06, + "loss": 0.9046, + "step": 14225 + }, + { + "epoch": 2.59, + "learning_rate": 2.22591918456498e-06, + "loss": 0.3821, + "step": 14226 + }, + { + "epoch": 2.59, + "learning_rate": 2.225864579541318e-06, + "loss": 0.0494, + "step": 14227 + }, + { + "epoch": 2.59, + "learning_rate": 2.2258099745176556e-06, + "loss": 1.5843, + "step": 14228 + }, + { + "epoch": 2.59, + "learning_rate": 2.2257553694939936e-06, + "loss": 0.6533, + "step": 14229 + }, + { + "epoch": 2.59, + "learning_rate": 2.2257007644703315e-06, + "loss": 0.6349, + "step": 14230 + }, + { + "epoch": 2.59, + "learning_rate": 2.225646159446669e-06, + "loss": 2.156, + "step": 14231 + }, + { + "epoch": 2.59, + "learning_rate": 2.225591554423007e-06, + "loss": 1.2514, + "step": 14232 + }, + { + "epoch": 2.59, + "learning_rate": 2.225536949399345e-06, + "loss": 0.3922, + "step": 14233 + }, + { + "epoch": 2.59, + "learning_rate": 2.2254823443756824e-06, + "loss": 0.7212, + "step": 14234 + }, + { + "epoch": 2.59, + "learning_rate": 2.2254277393520204e-06, + "loss": 0.0551, + "step": 14235 + }, + { + "epoch": 2.59, + "learning_rate": 2.2253731343283583e-06, + "loss": 0.8556, + "step": 14236 + }, + { + "epoch": 2.59, + "learning_rate": 2.225318529304696e-06, + "loss": 1.9585, + "step": 14237 + }, + { + "epoch": 2.59, + "learning_rate": 2.2252639242810338e-06, + "loss": 0.4487, + "step": 14238 + }, + { + "epoch": 2.59, + "learning_rate": 2.2252093192573717e-06, + "loss": 3.4487, + "step": 14239 + }, + { + "epoch": 2.59, + "learning_rate": 2.2251547142337097e-06, + "loss": 0.6111, + "step": 14240 + }, + { + "epoch": 2.59, + "learning_rate": 2.2251001092100476e-06, + "loss": 0.8714, + "step": 14241 + }, + { + "epoch": 2.59, + "learning_rate": 2.225045504186385e-06, + "loss": 1.8632, + "step": 14242 + }, + { + "epoch": 2.59, + "learning_rate": 2.224990899162723e-06, + "loss": 0.6292, + "step": 14243 + }, + { + "epoch": 2.59, + "learning_rate": 2.224936294139061e-06, + "loss": 1.3021, + "step": 14244 + }, + { + "epoch": 2.59, + "learning_rate": 2.2248816891153985e-06, + "loss": 2.5954, + "step": 14245 + }, + { + "epoch": 2.59, + "learning_rate": 2.2248270840917365e-06, + "loss": 0.0265, + "step": 14246 + }, + { + "epoch": 2.59, + "learning_rate": 2.2247724790680744e-06, + "loss": 1.2248, + "step": 14247 + }, + { + "epoch": 2.59, + "learning_rate": 2.224717874044412e-06, + "loss": 0.0427, + "step": 14248 + }, + { + "epoch": 2.59, + "learning_rate": 2.22466326902075e-06, + "loss": 0.5946, + "step": 14249 + }, + { + "epoch": 2.59, + "learning_rate": 2.224608663997088e-06, + "loss": 1.3543, + "step": 14250 + }, + { + "epoch": 2.59, + "learning_rate": 2.2245540589734258e-06, + "loss": 0.2891, + "step": 14251 + }, + { + "epoch": 2.59, + "learning_rate": 2.2244994539497633e-06, + "loss": 0.6066, + "step": 14252 + }, + { + "epoch": 2.59, + "learning_rate": 2.2244448489261012e-06, + "loss": 1.203, + "step": 14253 + }, + { + "epoch": 2.59, + "learning_rate": 2.224390243902439e-06, + "loss": 0.0446, + "step": 14254 + }, + { + "epoch": 2.59, + "learning_rate": 2.224335638878777e-06, + "loss": 0.4632, + "step": 14255 + }, + { + "epoch": 2.59, + "learning_rate": 2.2242810338551146e-06, + "loss": 2.3565, + "step": 14256 + }, + { + "epoch": 2.59, + "learning_rate": 2.2242264288314526e-06, + "loss": 0.6952, + "step": 14257 + }, + { + "epoch": 2.59, + "learning_rate": 2.2241718238077905e-06, + "loss": 3.8097, + "step": 14258 + }, + { + "epoch": 2.59, + "learning_rate": 2.224117218784128e-06, + "loss": 0.0295, + "step": 14259 + }, + { + "epoch": 2.59, + "learning_rate": 2.224062613760466e-06, + "loss": 0.0684, + "step": 14260 + }, + { + "epoch": 2.59, + "learning_rate": 2.224008008736804e-06, + "loss": 0.426, + "step": 14261 + }, + { + "epoch": 2.59, + "learning_rate": 2.2239534037131415e-06, + "loss": 0.3313, + "step": 14262 + }, + { + "epoch": 2.59, + "learning_rate": 2.2238987986894794e-06, + "loss": 0.893, + "step": 14263 + }, + { + "epoch": 2.59, + "learning_rate": 2.2238441936658173e-06, + "loss": 0.0958, + "step": 14264 + }, + { + "epoch": 2.59, + "learning_rate": 2.2237895886421553e-06, + "loss": 0.5041, + "step": 14265 + }, + { + "epoch": 2.59, + "learning_rate": 2.2237349836184932e-06, + "loss": 0.1248, + "step": 14266 + }, + { + "epoch": 2.59, + "learning_rate": 2.2236803785948308e-06, + "loss": 0.2879, + "step": 14267 + }, + { + "epoch": 2.59, + "learning_rate": 2.2236257735711687e-06, + "loss": 0.0142, + "step": 14268 + }, + { + "epoch": 2.59, + "learning_rate": 2.2235711685475066e-06, + "loss": 0.6181, + "step": 14269 + }, + { + "epoch": 2.6, + "learning_rate": 2.223516563523844e-06, + "loss": 0.1553, + "step": 14270 + }, + { + "epoch": 2.6, + "learning_rate": 2.223461958500182e-06, + "loss": 0.1084, + "step": 14271 + }, + { + "epoch": 2.6, + "learning_rate": 2.22340735347652e-06, + "loss": 1.9116, + "step": 14272 + }, + { + "epoch": 2.6, + "learning_rate": 2.2233527484528576e-06, + "loss": 0.0235, + "step": 14273 + }, + { + "epoch": 2.6, + "learning_rate": 2.2232981434291955e-06, + "loss": 1.5307, + "step": 14274 + }, + { + "epoch": 2.6, + "learning_rate": 2.2232435384055335e-06, + "loss": 0.6375, + "step": 14275 + }, + { + "epoch": 2.6, + "learning_rate": 2.223188933381871e-06, + "loss": 0.042, + "step": 14276 + }, + { + "epoch": 2.6, + "learning_rate": 2.223134328358209e-06, + "loss": 0.3991, + "step": 14277 + }, + { + "epoch": 2.6, + "learning_rate": 2.223079723334547e-06, + "loss": 0.555, + "step": 14278 + }, + { + "epoch": 2.6, + "learning_rate": 2.223025118310885e-06, + "loss": 0.7469, + "step": 14279 + }, + { + "epoch": 2.6, + "learning_rate": 2.2229705132872227e-06, + "loss": 0.7688, + "step": 14280 + }, + { + "epoch": 2.6, + "learning_rate": 2.2229159082635603e-06, + "loss": 0.0655, + "step": 14281 + }, + { + "epoch": 2.6, + "learning_rate": 2.222861303239898e-06, + "loss": 1.0386, + "step": 14282 + }, + { + "epoch": 2.6, + "learning_rate": 2.222806698216236e-06, + "loss": 0.7132, + "step": 14283 + }, + { + "epoch": 2.6, + "learning_rate": 2.2227520931925737e-06, + "loss": 0.083, + "step": 14284 + }, + { + "epoch": 2.6, + "learning_rate": 2.2226974881689116e-06, + "loss": 0.5543, + "step": 14285 + }, + { + "epoch": 2.6, + "learning_rate": 2.2226428831452496e-06, + "loss": 0.566, + "step": 14286 + }, + { + "epoch": 2.6, + "learning_rate": 2.222588278121587e-06, + "loss": 0.5489, + "step": 14287 + }, + { + "epoch": 2.6, + "learning_rate": 2.222533673097925e-06, + "loss": 0.5623, + "step": 14288 + }, + { + "epoch": 2.6, + "learning_rate": 2.222479068074263e-06, + "loss": 0.0633, + "step": 14289 + }, + { + "epoch": 2.6, + "learning_rate": 2.2224244630506005e-06, + "loss": 0.0683, + "step": 14290 + }, + { + "epoch": 2.6, + "learning_rate": 2.2223698580269384e-06, + "loss": 0.8027, + "step": 14291 + }, + { + "epoch": 2.6, + "learning_rate": 2.2223152530032764e-06, + "loss": 0.7827, + "step": 14292 + }, + { + "epoch": 2.6, + "learning_rate": 2.2222606479796143e-06, + "loss": 1.6809, + "step": 14293 + }, + { + "epoch": 2.6, + "learning_rate": 2.2222060429559523e-06, + "loss": 0.4452, + "step": 14294 + }, + { + "epoch": 2.6, + "learning_rate": 2.2221514379322898e-06, + "loss": 2.9649, + "step": 14295 + }, + { + "epoch": 2.6, + "learning_rate": 2.2220968329086277e-06, + "loss": 3.003, + "step": 14296 + }, + { + "epoch": 2.6, + "learning_rate": 2.2220422278849657e-06, + "loss": 0.931, + "step": 14297 + }, + { + "epoch": 2.6, + "learning_rate": 2.221987622861303e-06, + "loss": 1.4107, + "step": 14298 + }, + { + "epoch": 2.6, + "learning_rate": 2.221933017837641e-06, + "loss": 1.2509, + "step": 14299 + }, + { + "epoch": 2.6, + "learning_rate": 2.221878412813979e-06, + "loss": 1.2222, + "step": 14300 + }, + { + "epoch": 2.6, + "learning_rate": 2.2218238077903166e-06, + "loss": 0.5804, + "step": 14301 + }, + { + "epoch": 2.6, + "learning_rate": 2.2217692027666545e-06, + "loss": 0.0196, + "step": 14302 + }, + { + "epoch": 2.6, + "learning_rate": 2.2217145977429925e-06, + "loss": 0.0321, + "step": 14303 + }, + { + "epoch": 2.6, + "learning_rate": 2.22165999271933e-06, + "loss": 1.2639, + "step": 14304 + }, + { + "epoch": 2.6, + "learning_rate": 2.221605387695668e-06, + "loss": 0.7032, + "step": 14305 + }, + { + "epoch": 2.6, + "learning_rate": 2.221550782672006e-06, + "loss": 0.8806, + "step": 14306 + }, + { + "epoch": 2.6, + "learning_rate": 2.221496177648344e-06, + "loss": 1.722, + "step": 14307 + }, + { + "epoch": 2.6, + "learning_rate": 2.2214415726246818e-06, + "loss": 0.4258, + "step": 14308 + }, + { + "epoch": 2.6, + "learning_rate": 2.2213869676010193e-06, + "loss": 0.4822, + "step": 14309 + }, + { + "epoch": 2.6, + "learning_rate": 2.2213323625773572e-06, + "loss": 0.1927, + "step": 14310 + }, + { + "epoch": 2.6, + "learning_rate": 2.221277757553695e-06, + "loss": 0.9539, + "step": 14311 + }, + { + "epoch": 2.6, + "learning_rate": 2.2212231525300327e-06, + "loss": 0.4377, + "step": 14312 + }, + { + "epoch": 2.6, + "learning_rate": 2.2211685475063706e-06, + "loss": 0.3601, + "step": 14313 + }, + { + "epoch": 2.6, + "learning_rate": 2.2211139424827086e-06, + "loss": 0.5961, + "step": 14314 + }, + { + "epoch": 2.6, + "learning_rate": 2.221059337459046e-06, + "loss": 0.2739, + "step": 14315 + }, + { + "epoch": 2.6, + "learning_rate": 2.221004732435384e-06, + "loss": 0.2478, + "step": 14316 + }, + { + "epoch": 2.6, + "learning_rate": 2.220950127411722e-06, + "loss": 0.6823, + "step": 14317 + }, + { + "epoch": 2.6, + "learning_rate": 2.2208955223880595e-06, + "loss": 0.5814, + "step": 14318 + }, + { + "epoch": 2.6, + "learning_rate": 2.2208409173643975e-06, + "loss": 0.919, + "step": 14319 + }, + { + "epoch": 2.6, + "learning_rate": 2.2207863123407354e-06, + "loss": 0.6333, + "step": 14320 + }, + { + "epoch": 2.6, + "learning_rate": 2.2207317073170733e-06, + "loss": 0.0262, + "step": 14321 + }, + { + "epoch": 2.6, + "learning_rate": 2.2206771022934113e-06, + "loss": 0.5674, + "step": 14322 + }, + { + "epoch": 2.6, + "learning_rate": 2.220622497269749e-06, + "loss": 1.5656, + "step": 14323 + }, + { + "epoch": 2.6, + "learning_rate": 2.2205678922460867e-06, + "loss": 1.8419, + "step": 14324 + }, + { + "epoch": 2.61, + "learning_rate": 2.2205132872224247e-06, + "loss": 0.4864, + "step": 14325 + }, + { + "epoch": 2.61, + "learning_rate": 2.2204586821987622e-06, + "loss": 0.0432, + "step": 14326 + }, + { + "epoch": 2.61, + "learning_rate": 2.2204040771751e-06, + "loss": 0.0117, + "step": 14327 + }, + { + "epoch": 2.61, + "learning_rate": 2.220349472151438e-06, + "loss": 0.6558, + "step": 14328 + }, + { + "epoch": 2.61, + "learning_rate": 2.2202948671277756e-06, + "loss": 0.1599, + "step": 14329 + }, + { + "epoch": 2.61, + "learning_rate": 2.2202402621041136e-06, + "loss": 3.4785, + "step": 14330 + }, + { + "epoch": 2.61, + "learning_rate": 2.220185657080451e-06, + "loss": 0.6381, + "step": 14331 + }, + { + "epoch": 2.61, + "learning_rate": 2.220131052056789e-06, + "loss": 0.6053, + "step": 14332 + }, + { + "epoch": 2.61, + "learning_rate": 2.220076447033127e-06, + "loss": 0.0752, + "step": 14333 + }, + { + "epoch": 2.61, + "learning_rate": 2.220021842009465e-06, + "loss": 0.3437, + "step": 14334 + }, + { + "epoch": 2.61, + "learning_rate": 2.219967236985803e-06, + "loss": 0.1435, + "step": 14335 + }, + { + "epoch": 2.61, + "learning_rate": 2.219912631962141e-06, + "loss": 1.5726, + "step": 14336 + }, + { + "epoch": 2.61, + "learning_rate": 2.2198580269384783e-06, + "loss": 0.5763, + "step": 14337 + }, + { + "epoch": 2.61, + "learning_rate": 2.2198034219148163e-06, + "loss": 0.2162, + "step": 14338 + }, + { + "epoch": 2.61, + "learning_rate": 2.219748816891154e-06, + "loss": 0.0617, + "step": 14339 + }, + { + "epoch": 2.61, + "learning_rate": 2.2196942118674917e-06, + "loss": 0.0294, + "step": 14340 + }, + { + "epoch": 2.61, + "learning_rate": 2.2196396068438297e-06, + "loss": 1.5254, + "step": 14341 + }, + { + "epoch": 2.61, + "learning_rate": 2.2195850018201676e-06, + "loss": 0.0229, + "step": 14342 + }, + { + "epoch": 2.61, + "learning_rate": 2.219530396796505e-06, + "loss": 0.0416, + "step": 14343 + }, + { + "epoch": 2.61, + "learning_rate": 2.219475791772843e-06, + "loss": 0.1088, + "step": 14344 + }, + { + "epoch": 2.61, + "learning_rate": 2.219421186749181e-06, + "loss": 0.0556, + "step": 14345 + }, + { + "epoch": 2.61, + "learning_rate": 2.219366581725519e-06, + "loss": 1.2407, + "step": 14346 + }, + { + "epoch": 2.61, + "learning_rate": 2.219311976701857e-06, + "loss": 2.7756, + "step": 14347 + }, + { + "epoch": 2.61, + "learning_rate": 2.2192573716781944e-06, + "loss": 0.8159, + "step": 14348 + }, + { + "epoch": 2.61, + "learning_rate": 2.2192027666545324e-06, + "loss": 1.6392, + "step": 14349 + }, + { + "epoch": 2.61, + "learning_rate": 2.2191481616308703e-06, + "loss": 1.627, + "step": 14350 + }, + { + "epoch": 2.61, + "learning_rate": 2.219093556607208e-06, + "loss": 0.9872, + "step": 14351 + }, + { + "epoch": 2.61, + "learning_rate": 2.2190389515835458e-06, + "loss": 1.2433, + "step": 14352 + }, + { + "epoch": 2.61, + "learning_rate": 2.2189843465598837e-06, + "loss": 5.1595, + "step": 14353 + }, + { + "epoch": 2.61, + "learning_rate": 2.2189297415362212e-06, + "loss": 0.3144, + "step": 14354 + }, + { + "epoch": 2.61, + "learning_rate": 2.218875136512559e-06, + "loss": 0.319, + "step": 14355 + }, + { + "epoch": 2.61, + "learning_rate": 2.218820531488897e-06, + "loss": 0.5895, + "step": 14356 + }, + { + "epoch": 2.61, + "learning_rate": 2.2187659264652346e-06, + "loss": 0.777, + "step": 14357 + }, + { + "epoch": 2.61, + "learning_rate": 2.2187113214415726e-06, + "loss": 0.4827, + "step": 14358 + }, + { + "epoch": 2.61, + "learning_rate": 2.2186567164179105e-06, + "loss": 0.333, + "step": 14359 + }, + { + "epoch": 2.61, + "learning_rate": 2.2186021113942485e-06, + "loss": 0.401, + "step": 14360 + }, + { + "epoch": 2.61, + "learning_rate": 2.2185475063705864e-06, + "loss": 0.2313, + "step": 14361 + }, + { + "epoch": 2.61, + "learning_rate": 2.218492901346924e-06, + "loss": 0.2676, + "step": 14362 + }, + { + "epoch": 2.61, + "learning_rate": 2.218438296323262e-06, + "loss": 0.3177, + "step": 14363 + }, + { + "epoch": 2.61, + "learning_rate": 2.2183836912996e-06, + "loss": 0.2968, + "step": 14364 + }, + { + "epoch": 2.61, + "learning_rate": 2.2183290862759373e-06, + "loss": 0.4803, + "step": 14365 + }, + { + "epoch": 2.61, + "learning_rate": 2.2182744812522753e-06, + "loss": 1.4201, + "step": 14366 + }, + { + "epoch": 2.61, + "learning_rate": 2.2182198762286132e-06, + "loss": 2.4554, + "step": 14367 + }, + { + "epoch": 2.61, + "learning_rate": 2.2181652712049508e-06, + "loss": 2.0678, + "step": 14368 + }, + { + "epoch": 2.61, + "learning_rate": 2.2181106661812887e-06, + "loss": 0.5298, + "step": 14369 + }, + { + "epoch": 2.61, + "learning_rate": 2.2180560611576266e-06, + "loss": 0.1872, + "step": 14370 + }, + { + "epoch": 2.61, + "learning_rate": 2.218001456133964e-06, + "loss": 1.154, + "step": 14371 + }, + { + "epoch": 2.61, + "learning_rate": 2.217946851110302e-06, + "loss": 0.0442, + "step": 14372 + }, + { + "epoch": 2.61, + "learning_rate": 2.21789224608664e-06, + "loss": 1.2278, + "step": 14373 + }, + { + "epoch": 2.61, + "learning_rate": 2.217837641062978e-06, + "loss": 0.0554, + "step": 14374 + }, + { + "epoch": 2.61, + "learning_rate": 2.217783036039316e-06, + "loss": 0.0623, + "step": 14375 + }, + { + "epoch": 2.61, + "learning_rate": 2.2177284310156535e-06, + "loss": 0.7614, + "step": 14376 + }, + { + "epoch": 2.61, + "learning_rate": 2.2176738259919914e-06, + "loss": 0.82, + "step": 14377 + }, + { + "epoch": 2.61, + "learning_rate": 2.2176192209683293e-06, + "loss": 1.3705, + "step": 14378 + }, + { + "epoch": 2.61, + "learning_rate": 2.217564615944667e-06, + "loss": 1.7164, + "step": 14379 + }, + { + "epoch": 2.62, + "learning_rate": 2.217510010921005e-06, + "loss": 0.5793, + "step": 14380 + }, + { + "epoch": 2.62, + "learning_rate": 2.2174554058973427e-06, + "loss": 1.1224, + "step": 14381 + }, + { + "epoch": 2.62, + "learning_rate": 2.2174008008736803e-06, + "loss": 0.031, + "step": 14382 + }, + { + "epoch": 2.62, + "learning_rate": 2.217346195850018e-06, + "loss": 1.1915, + "step": 14383 + }, + { + "epoch": 2.62, + "learning_rate": 2.217291590826356e-06, + "loss": 1.8767, + "step": 14384 + }, + { + "epoch": 2.62, + "learning_rate": 2.2172369858026937e-06, + "loss": 1.1781, + "step": 14385 + }, + { + "epoch": 2.62, + "learning_rate": 2.2171823807790316e-06, + "loss": 0.0304, + "step": 14386 + }, + { + "epoch": 2.62, + "learning_rate": 2.2171277757553696e-06, + "loss": 0.3757, + "step": 14387 + }, + { + "epoch": 2.62, + "learning_rate": 2.2170731707317075e-06, + "loss": 2.5082, + "step": 14388 + }, + { + "epoch": 2.62, + "learning_rate": 2.2170185657080454e-06, + "loss": 0.4251, + "step": 14389 + }, + { + "epoch": 2.62, + "learning_rate": 2.216963960684383e-06, + "loss": 0.9859, + "step": 14390 + }, + { + "epoch": 2.62, + "learning_rate": 2.216909355660721e-06, + "loss": 0.3694, + "step": 14391 + }, + { + "epoch": 2.62, + "learning_rate": 2.216854750637059e-06, + "loss": 0.4395, + "step": 14392 + }, + { + "epoch": 2.62, + "learning_rate": 2.2168001456133964e-06, + "loss": 1.068, + "step": 14393 + }, + { + "epoch": 2.62, + "learning_rate": 2.2167455405897343e-06, + "loss": 0.9596, + "step": 14394 + }, + { + "epoch": 2.62, + "learning_rate": 2.2166909355660723e-06, + "loss": 0.3983, + "step": 14395 + }, + { + "epoch": 2.62, + "learning_rate": 2.2166363305424098e-06, + "loss": 1.291, + "step": 14396 + }, + { + "epoch": 2.62, + "learning_rate": 2.2165817255187477e-06, + "loss": 1.5178, + "step": 14397 + }, + { + "epoch": 2.62, + "learning_rate": 2.2165271204950857e-06, + "loss": 0.6815, + "step": 14398 + }, + { + "epoch": 2.62, + "learning_rate": 2.216472515471423e-06, + "loss": 0.947, + "step": 14399 + }, + { + "epoch": 2.62, + "learning_rate": 2.216417910447761e-06, + "loss": 0.3023, + "step": 14400 + }, + { + "epoch": 2.62, + "learning_rate": 2.216363305424099e-06, + "loss": 0.5373, + "step": 14401 + }, + { + "epoch": 2.62, + "learning_rate": 2.216308700400437e-06, + "loss": 2.79, + "step": 14402 + }, + { + "epoch": 2.62, + "learning_rate": 2.216254095376775e-06, + "loss": 0.4557, + "step": 14403 + }, + { + "epoch": 2.62, + "learning_rate": 2.2161994903531125e-06, + "loss": 0.3137, + "step": 14404 + }, + { + "epoch": 2.62, + "learning_rate": 2.2161448853294504e-06, + "loss": 0.1611, + "step": 14405 + }, + { + "epoch": 2.62, + "learning_rate": 2.2160902803057884e-06, + "loss": 1.673, + "step": 14406 + }, + { + "epoch": 2.62, + "learning_rate": 2.216035675282126e-06, + "loss": 1.2387, + "step": 14407 + }, + { + "epoch": 2.62, + "learning_rate": 2.215981070258464e-06, + "loss": 0.0145, + "step": 14408 + }, + { + "epoch": 2.62, + "learning_rate": 2.2159264652348018e-06, + "loss": 1.1984, + "step": 14409 + }, + { + "epoch": 2.62, + "learning_rate": 2.2158718602111393e-06, + "loss": 0.5607, + "step": 14410 + }, + { + "epoch": 2.62, + "learning_rate": 2.2158172551874772e-06, + "loss": 1.1122, + "step": 14411 + }, + { + "epoch": 2.62, + "learning_rate": 2.2157626501638148e-06, + "loss": 0.0198, + "step": 14412 + }, + { + "epoch": 2.62, + "learning_rate": 2.2157080451401527e-06, + "loss": 0.6007, + "step": 14413 + }, + { + "epoch": 2.62, + "learning_rate": 2.2156534401164906e-06, + "loss": 0.2142, + "step": 14414 + }, + { + "epoch": 2.62, + "learning_rate": 2.2155988350928286e-06, + "loss": 0.0899, + "step": 14415 + }, + { + "epoch": 2.62, + "learning_rate": 2.2155442300691665e-06, + "loss": 0.6484, + "step": 14416 + }, + { + "epoch": 2.62, + "learning_rate": 2.2154896250455045e-06, + "loss": 0.0358, + "step": 14417 + }, + { + "epoch": 2.62, + "learning_rate": 2.215435020021842e-06, + "loss": 1.6257, + "step": 14418 + }, + { + "epoch": 2.62, + "learning_rate": 2.21538041499818e-06, + "loss": 0.9348, + "step": 14419 + }, + { + "epoch": 2.62, + "learning_rate": 2.215325809974518e-06, + "loss": 0.8004, + "step": 14420 + }, + { + "epoch": 2.62, + "learning_rate": 2.2152712049508554e-06, + "loss": 0.0234, + "step": 14421 + }, + { + "epoch": 2.62, + "learning_rate": 2.2152165999271933e-06, + "loss": 0.0981, + "step": 14422 + }, + { + "epoch": 2.62, + "learning_rate": 2.2151619949035313e-06, + "loss": 1.7036, + "step": 14423 + }, + { + "epoch": 2.62, + "learning_rate": 2.215107389879869e-06, + "loss": 0.8396, + "step": 14424 + }, + { + "epoch": 2.62, + "learning_rate": 2.2150527848562068e-06, + "loss": 0.0465, + "step": 14425 + }, + { + "epoch": 2.62, + "learning_rate": 2.2149981798325447e-06, + "loss": 1.2097, + "step": 14426 + }, + { + "epoch": 2.62, + "learning_rate": 2.2149435748088826e-06, + "loss": 0.1627, + "step": 14427 + }, + { + "epoch": 2.62, + "learning_rate": 2.2148889697852206e-06, + "loss": 1.0326, + "step": 14428 + }, + { + "epoch": 2.62, + "learning_rate": 2.214834364761558e-06, + "loss": 0.0287, + "step": 14429 + }, + { + "epoch": 2.62, + "learning_rate": 2.214779759737896e-06, + "loss": 0.0194, + "step": 14430 + }, + { + "epoch": 2.62, + "learning_rate": 2.214725154714234e-06, + "loss": 2.4482, + "step": 14431 + }, + { + "epoch": 2.62, + "learning_rate": 2.2146705496905715e-06, + "loss": 1.4455, + "step": 14432 + }, + { + "epoch": 2.62, + "learning_rate": 2.2146159446669095e-06, + "loss": 1.1354, + "step": 14433 + }, + { + "epoch": 2.62, + "learning_rate": 2.2145613396432474e-06, + "loss": 1.8052, + "step": 14434 + }, + { + "epoch": 2.63, + "learning_rate": 2.214506734619585e-06, + "loss": 0.8922, + "step": 14435 + }, + { + "epoch": 2.63, + "learning_rate": 2.214452129595923e-06, + "loss": 0.8241, + "step": 14436 + }, + { + "epoch": 2.63, + "learning_rate": 2.214397524572261e-06, + "loss": 0.1317, + "step": 14437 + }, + { + "epoch": 2.63, + "learning_rate": 2.2143429195485983e-06, + "loss": 0.6129, + "step": 14438 + }, + { + "epoch": 2.63, + "learning_rate": 2.2142883145249363e-06, + "loss": 0.0184, + "step": 14439 + }, + { + "epoch": 2.63, + "learning_rate": 2.214233709501274e-06, + "loss": 0.4822, + "step": 14440 + }, + { + "epoch": 2.63, + "learning_rate": 2.214179104477612e-06, + "loss": 1.3233, + "step": 14441 + }, + { + "epoch": 2.63, + "learning_rate": 2.21412449945395e-06, + "loss": 0.8476, + "step": 14442 + }, + { + "epoch": 2.63, + "learning_rate": 2.2140698944302876e-06, + "loss": 2.3384, + "step": 14443 + }, + { + "epoch": 2.63, + "learning_rate": 2.2140152894066256e-06, + "loss": 0.83, + "step": 14444 + }, + { + "epoch": 2.63, + "learning_rate": 2.2139606843829635e-06, + "loss": 0.7302, + "step": 14445 + }, + { + "epoch": 2.63, + "learning_rate": 2.213906079359301e-06, + "loss": 0.5024, + "step": 14446 + }, + { + "epoch": 2.63, + "learning_rate": 2.213851474335639e-06, + "loss": 0.8088, + "step": 14447 + }, + { + "epoch": 2.63, + "learning_rate": 2.213796869311977e-06, + "loss": 2.6252, + "step": 14448 + }, + { + "epoch": 2.63, + "learning_rate": 2.2137422642883144e-06, + "loss": 1.3811, + "step": 14449 + }, + { + "epoch": 2.63, + "learning_rate": 2.2136876592646524e-06, + "loss": 0.7113, + "step": 14450 + }, + { + "epoch": 2.63, + "learning_rate": 2.2136330542409903e-06, + "loss": 2.3986, + "step": 14451 + }, + { + "epoch": 2.63, + "learning_rate": 2.213578449217328e-06, + "loss": 0.9559, + "step": 14452 + }, + { + "epoch": 2.63, + "learning_rate": 2.2135238441936658e-06, + "loss": 1.4775, + "step": 14453 + }, + { + "epoch": 2.63, + "learning_rate": 2.2134692391700037e-06, + "loss": 0.7054, + "step": 14454 + }, + { + "epoch": 2.63, + "learning_rate": 2.2134146341463417e-06, + "loss": 0.3314, + "step": 14455 + }, + { + "epoch": 2.63, + "learning_rate": 2.2133600291226796e-06, + "loss": 0.6165, + "step": 14456 + }, + { + "epoch": 2.63, + "learning_rate": 2.213305424099017e-06, + "loss": 1.0608, + "step": 14457 + }, + { + "epoch": 2.63, + "learning_rate": 2.213250819075355e-06, + "loss": 0.0369, + "step": 14458 + }, + { + "epoch": 2.63, + "learning_rate": 2.213196214051693e-06, + "loss": 0.6028, + "step": 14459 + }, + { + "epoch": 2.63, + "learning_rate": 2.2131416090280305e-06, + "loss": 0.4235, + "step": 14460 + }, + { + "epoch": 2.63, + "learning_rate": 2.2130870040043685e-06, + "loss": 0.5665, + "step": 14461 + }, + { + "epoch": 2.63, + "learning_rate": 2.2130323989807064e-06, + "loss": 1.9365, + "step": 14462 + }, + { + "epoch": 2.63, + "learning_rate": 2.212977793957044e-06, + "loss": 0.0141, + "step": 14463 + }, + { + "epoch": 2.63, + "learning_rate": 2.212923188933382e-06, + "loss": 0.0461, + "step": 14464 + }, + { + "epoch": 2.63, + "learning_rate": 2.21286858390972e-06, + "loss": 0.6233, + "step": 14465 + }, + { + "epoch": 2.63, + "learning_rate": 2.2128139788860573e-06, + "loss": 0.8269, + "step": 14466 + }, + { + "epoch": 2.63, + "learning_rate": 2.2127593738623953e-06, + "loss": 1.3349, + "step": 14467 + }, + { + "epoch": 2.63, + "learning_rate": 2.2127047688387332e-06, + "loss": 0.2254, + "step": 14468 + }, + { + "epoch": 2.63, + "learning_rate": 2.212650163815071e-06, + "loss": 1.3198, + "step": 14469 + }, + { + "epoch": 2.63, + "learning_rate": 2.212595558791409e-06, + "loss": 0.7766, + "step": 14470 + }, + { + "epoch": 2.63, + "learning_rate": 2.2125409537677466e-06, + "loss": 0.6826, + "step": 14471 + }, + { + "epoch": 2.63, + "learning_rate": 2.2124863487440846e-06, + "loss": 0.2189, + "step": 14472 + }, + { + "epoch": 2.63, + "learning_rate": 2.2124317437204225e-06, + "loss": 0.1357, + "step": 14473 + }, + { + "epoch": 2.63, + "learning_rate": 2.21237713869676e-06, + "loss": 0.2831, + "step": 14474 + }, + { + "epoch": 2.63, + "learning_rate": 2.212322533673098e-06, + "loss": 1.2376, + "step": 14475 + }, + { + "epoch": 2.63, + "learning_rate": 2.212267928649436e-06, + "loss": 0.2788, + "step": 14476 + }, + { + "epoch": 2.63, + "learning_rate": 2.2122133236257735e-06, + "loss": 0.7929, + "step": 14477 + }, + { + "epoch": 2.63, + "learning_rate": 2.2121587186021114e-06, + "loss": 0.4428, + "step": 14478 + }, + { + "epoch": 2.63, + "learning_rate": 2.2121041135784493e-06, + "loss": 1.0956, + "step": 14479 + }, + { + "epoch": 2.63, + "learning_rate": 2.212049508554787e-06, + "loss": 1.2884, + "step": 14480 + }, + { + "epoch": 2.63, + "learning_rate": 2.211994903531125e-06, + "loss": 0.9053, + "step": 14481 + }, + { + "epoch": 2.63, + "learning_rate": 2.2119402985074627e-06, + "loss": 1.2722, + "step": 14482 + }, + { + "epoch": 2.63, + "learning_rate": 2.2118856934838007e-06, + "loss": 0.813, + "step": 14483 + }, + { + "epoch": 2.63, + "learning_rate": 2.2118310884601386e-06, + "loss": 1.4426, + "step": 14484 + }, + { + "epoch": 2.63, + "learning_rate": 2.211776483436476e-06, + "loss": 0.154, + "step": 14485 + }, + { + "epoch": 2.63, + "learning_rate": 2.211721878412814e-06, + "loss": 1.9288, + "step": 14486 + }, + { + "epoch": 2.63, + "learning_rate": 2.211667273389152e-06, + "loss": 0.6174, + "step": 14487 + }, + { + "epoch": 2.63, + "learning_rate": 2.2116126683654896e-06, + "loss": 0.7406, + "step": 14488 + }, + { + "epoch": 2.63, + "learning_rate": 2.2115580633418275e-06, + "loss": 0.1004, + "step": 14489 + }, + { + "epoch": 2.64, + "learning_rate": 2.2115034583181654e-06, + "loss": 0.0542, + "step": 14490 + }, + { + "epoch": 2.64, + "learning_rate": 2.211448853294503e-06, + "loss": 1.6579, + "step": 14491 + }, + { + "epoch": 2.64, + "learning_rate": 2.211394248270841e-06, + "loss": 0.2911, + "step": 14492 + }, + { + "epoch": 2.64, + "learning_rate": 2.2113396432471784e-06, + "loss": 0.7905, + "step": 14493 + }, + { + "epoch": 2.64, + "learning_rate": 2.2112850382235164e-06, + "loss": 0.9262, + "step": 14494 + }, + { + "epoch": 2.64, + "learning_rate": 2.2112304331998543e-06, + "loss": 0.0527, + "step": 14495 + }, + { + "epoch": 2.64, + "learning_rate": 2.2111758281761923e-06, + "loss": 1.1428, + "step": 14496 + }, + { + "epoch": 2.64, + "learning_rate": 2.21112122315253e-06, + "loss": 1.5639, + "step": 14497 + }, + { + "epoch": 2.64, + "learning_rate": 2.211066618128868e-06, + "loss": 1.5289, + "step": 14498 + }, + { + "epoch": 2.64, + "learning_rate": 2.2110120131052057e-06, + "loss": 0.5531, + "step": 14499 + }, + { + "epoch": 2.64, + "learning_rate": 2.2109574080815436e-06, + "loss": 1.3991, + "step": 14500 + }, + { + "epoch": 2.64, + "learning_rate": 2.2109028030578816e-06, + "loss": 0.5118, + "step": 14501 + }, + { + "epoch": 2.64, + "learning_rate": 2.210848198034219e-06, + "loss": 1.0668, + "step": 14502 + }, + { + "epoch": 2.64, + "learning_rate": 2.210793593010557e-06, + "loss": 0.7526, + "step": 14503 + }, + { + "epoch": 2.64, + "learning_rate": 2.210738987986895e-06, + "loss": 2.3276, + "step": 14504 + }, + { + "epoch": 2.64, + "learning_rate": 2.2106843829632325e-06, + "loss": 2.6468, + "step": 14505 + }, + { + "epoch": 2.64, + "learning_rate": 2.2106297779395704e-06, + "loss": 0.9482, + "step": 14506 + }, + { + "epoch": 2.64, + "learning_rate": 2.2105751729159084e-06, + "loss": 1.3478, + "step": 14507 + }, + { + "epoch": 2.64, + "learning_rate": 2.2105205678922463e-06, + "loss": 0.105, + "step": 14508 + }, + { + "epoch": 2.64, + "learning_rate": 2.2104659628685843e-06, + "loss": 1.8246, + "step": 14509 + }, + { + "epoch": 2.64, + "learning_rate": 2.2104113578449218e-06, + "loss": 0.7971, + "step": 14510 + }, + { + "epoch": 2.64, + "learning_rate": 2.2103567528212597e-06, + "loss": 0.5393, + "step": 14511 + }, + { + "epoch": 2.64, + "learning_rate": 2.2103021477975977e-06, + "loss": 0.7033, + "step": 14512 + }, + { + "epoch": 2.64, + "learning_rate": 2.210247542773935e-06, + "loss": 0.4388, + "step": 14513 + }, + { + "epoch": 2.64, + "learning_rate": 2.210192937750273e-06, + "loss": 0.1934, + "step": 14514 + }, + { + "epoch": 2.64, + "learning_rate": 2.210138332726611e-06, + "loss": 0.6391, + "step": 14515 + }, + { + "epoch": 2.64, + "learning_rate": 2.2100837277029486e-06, + "loss": 1.1224, + "step": 14516 + }, + { + "epoch": 2.64, + "learning_rate": 2.2100291226792865e-06, + "loss": 0.65, + "step": 14517 + }, + { + "epoch": 2.64, + "learning_rate": 2.2099745176556245e-06, + "loss": 1.5777, + "step": 14518 + }, + { + "epoch": 2.64, + "learning_rate": 2.209919912631962e-06, + "loss": 1.9939, + "step": 14519 + }, + { + "epoch": 2.64, + "learning_rate": 2.2098653076083e-06, + "loss": 0.4439, + "step": 14520 + }, + { + "epoch": 2.64, + "learning_rate": 2.209810702584638e-06, + "loss": 2.9252, + "step": 14521 + }, + { + "epoch": 2.64, + "learning_rate": 2.209756097560976e-06, + "loss": 0.1269, + "step": 14522 + }, + { + "epoch": 2.64, + "learning_rate": 2.2097014925373138e-06, + "loss": 0.7439, + "step": 14523 + }, + { + "epoch": 2.64, + "learning_rate": 2.2096468875136513e-06, + "loss": 1.6283, + "step": 14524 + }, + { + "epoch": 2.64, + "learning_rate": 2.2095922824899892e-06, + "loss": 0.024, + "step": 14525 + }, + { + "epoch": 2.64, + "learning_rate": 2.209537677466327e-06, + "loss": 0.7252, + "step": 14526 + }, + { + "epoch": 2.64, + "learning_rate": 2.2094830724426647e-06, + "loss": 0.2659, + "step": 14527 + }, + { + "epoch": 2.64, + "learning_rate": 2.2094284674190026e-06, + "loss": 2.2117, + "step": 14528 + }, + { + "epoch": 2.64, + "learning_rate": 2.2093738623953406e-06, + "loss": 0.3215, + "step": 14529 + }, + { + "epoch": 2.64, + "learning_rate": 2.209319257371678e-06, + "loss": 0.0513, + "step": 14530 + }, + { + "epoch": 2.64, + "learning_rate": 2.209264652348016e-06, + "loss": 0.3521, + "step": 14531 + }, + { + "epoch": 2.64, + "learning_rate": 2.209210047324354e-06, + "loss": 0.4721, + "step": 14532 + }, + { + "epoch": 2.64, + "learning_rate": 2.2091554423006915e-06, + "loss": 2.6607, + "step": 14533 + }, + { + "epoch": 2.64, + "learning_rate": 2.2091008372770295e-06, + "loss": 1.3003, + "step": 14534 + }, + { + "epoch": 2.64, + "learning_rate": 2.2090462322533674e-06, + "loss": 0.7731, + "step": 14535 + }, + { + "epoch": 2.64, + "learning_rate": 2.2089916272297053e-06, + "loss": 2.5686, + "step": 14536 + }, + { + "epoch": 2.64, + "learning_rate": 2.2089370222060433e-06, + "loss": 0.6032, + "step": 14537 + }, + { + "epoch": 2.64, + "learning_rate": 2.208882417182381e-06, + "loss": 0.9295, + "step": 14538 + }, + { + "epoch": 2.64, + "learning_rate": 2.2088278121587187e-06, + "loss": 0.7817, + "step": 14539 + }, + { + "epoch": 2.64, + "learning_rate": 2.2087732071350567e-06, + "loss": 0.1468, + "step": 14540 + }, + { + "epoch": 2.64, + "learning_rate": 2.208718602111394e-06, + "loss": 1.0243, + "step": 14541 + }, + { + "epoch": 2.64, + "learning_rate": 2.208663997087732e-06, + "loss": 1.0811, + "step": 14542 + }, + { + "epoch": 2.64, + "learning_rate": 2.20860939206407e-06, + "loss": 0.4556, + "step": 14543 + }, + { + "epoch": 2.64, + "learning_rate": 2.2085547870404076e-06, + "loss": 0.4586, + "step": 14544 + }, + { + "epoch": 2.65, + "learning_rate": 2.2085001820167456e-06, + "loss": 0.2733, + "step": 14545 + }, + { + "epoch": 2.65, + "learning_rate": 2.2084455769930835e-06, + "loss": 0.0146, + "step": 14546 + }, + { + "epoch": 2.65, + "learning_rate": 2.208390971969421e-06, + "loss": 1.0705, + "step": 14547 + }, + { + "epoch": 2.65, + "learning_rate": 2.208336366945759e-06, + "loss": 1.2961, + "step": 14548 + }, + { + "epoch": 2.65, + "learning_rate": 2.208281761922097e-06, + "loss": 0.0181, + "step": 14549 + }, + { + "epoch": 2.65, + "learning_rate": 2.208227156898435e-06, + "loss": 0.4397, + "step": 14550 + }, + { + "epoch": 2.65, + "learning_rate": 2.208172551874773e-06, + "loss": 1.2149, + "step": 14551 + }, + { + "epoch": 2.65, + "learning_rate": 2.2081179468511103e-06, + "loss": 1.9716, + "step": 14552 + }, + { + "epoch": 2.65, + "learning_rate": 2.2080633418274483e-06, + "loss": 0.3015, + "step": 14553 + }, + { + "epoch": 2.65, + "learning_rate": 2.208008736803786e-06, + "loss": 0.6775, + "step": 14554 + }, + { + "epoch": 2.65, + "learning_rate": 2.2079541317801237e-06, + "loss": 0.5586, + "step": 14555 + }, + { + "epoch": 2.65, + "learning_rate": 2.2078995267564617e-06, + "loss": 0.1703, + "step": 14556 + }, + { + "epoch": 2.65, + "learning_rate": 2.2078449217327996e-06, + "loss": 0.3062, + "step": 14557 + }, + { + "epoch": 2.65, + "learning_rate": 2.207790316709137e-06, + "loss": 0.6533, + "step": 14558 + }, + { + "epoch": 2.65, + "learning_rate": 2.207735711685475e-06, + "loss": 0.0125, + "step": 14559 + }, + { + "epoch": 2.65, + "learning_rate": 2.207681106661813e-06, + "loss": 1.0992, + "step": 14560 + }, + { + "epoch": 2.65, + "learning_rate": 2.2076265016381505e-06, + "loss": 1.8657, + "step": 14561 + }, + { + "epoch": 2.65, + "learning_rate": 2.2075718966144885e-06, + "loss": 0.3441, + "step": 14562 + }, + { + "epoch": 2.65, + "learning_rate": 2.2075172915908264e-06, + "loss": 2.2102, + "step": 14563 + }, + { + "epoch": 2.65, + "learning_rate": 2.2074626865671644e-06, + "loss": 1.0484, + "step": 14564 + }, + { + "epoch": 2.65, + "learning_rate": 2.2074080815435023e-06, + "loss": 0.0389, + "step": 14565 + }, + { + "epoch": 2.65, + "learning_rate": 2.20735347651984e-06, + "loss": 0.2526, + "step": 14566 + }, + { + "epoch": 2.65, + "learning_rate": 2.2072988714961778e-06, + "loss": 0.1587, + "step": 14567 + }, + { + "epoch": 2.65, + "learning_rate": 2.2072442664725157e-06, + "loss": 0.7025, + "step": 14568 + }, + { + "epoch": 2.65, + "learning_rate": 2.2071896614488532e-06, + "loss": 1.38, + "step": 14569 + }, + { + "epoch": 2.65, + "learning_rate": 2.207135056425191e-06, + "loss": 0.2512, + "step": 14570 + }, + { + "epoch": 2.65, + "learning_rate": 2.207080451401529e-06, + "loss": 0.5038, + "step": 14571 + }, + { + "epoch": 2.65, + "learning_rate": 2.2070258463778666e-06, + "loss": 0.4652, + "step": 14572 + }, + { + "epoch": 2.65, + "learning_rate": 2.2069712413542046e-06, + "loss": 0.1063, + "step": 14573 + }, + { + "epoch": 2.65, + "learning_rate": 2.206916636330542e-06, + "loss": 0.2899, + "step": 14574 + }, + { + "epoch": 2.65, + "learning_rate": 2.20686203130688e-06, + "loss": 0.4399, + "step": 14575 + }, + { + "epoch": 2.65, + "learning_rate": 2.206807426283218e-06, + "loss": 1.9817, + "step": 14576 + }, + { + "epoch": 2.65, + "learning_rate": 2.206752821259556e-06, + "loss": 0.014, + "step": 14577 + }, + { + "epoch": 2.65, + "learning_rate": 2.206698216235894e-06, + "loss": 0.5257, + "step": 14578 + }, + { + "epoch": 2.65, + "learning_rate": 2.206643611212232e-06, + "loss": 0.4838, + "step": 14579 + }, + { + "epoch": 2.65, + "learning_rate": 2.2065890061885693e-06, + "loss": 0.0669, + "step": 14580 + }, + { + "epoch": 2.65, + "learning_rate": 2.2065344011649073e-06, + "loss": 2.3693, + "step": 14581 + }, + { + "epoch": 2.65, + "learning_rate": 2.2064797961412452e-06, + "loss": 0.6492, + "step": 14582 + }, + { + "epoch": 2.65, + "learning_rate": 2.2064251911175827e-06, + "loss": 1.3111, + "step": 14583 + }, + { + "epoch": 2.65, + "learning_rate": 2.2063705860939207e-06, + "loss": 2.3726, + "step": 14584 + }, + { + "epoch": 2.65, + "learning_rate": 2.2063159810702586e-06, + "loss": 0.4575, + "step": 14585 + }, + { + "epoch": 2.65, + "learning_rate": 2.206261376046596e-06, + "loss": 0.837, + "step": 14586 + }, + { + "epoch": 2.65, + "learning_rate": 2.206206771022934e-06, + "loss": 0.0194, + "step": 14587 + }, + { + "epoch": 2.65, + "learning_rate": 2.206152165999272e-06, + "loss": 0.0585, + "step": 14588 + }, + { + "epoch": 2.65, + "learning_rate": 2.2060975609756096e-06, + "loss": 1.078, + "step": 14589 + }, + { + "epoch": 2.65, + "learning_rate": 2.2060429559519475e-06, + "loss": 0.318, + "step": 14590 + }, + { + "epoch": 2.65, + "learning_rate": 2.2059883509282854e-06, + "loss": 0.581, + "step": 14591 + }, + { + "epoch": 2.65, + "learning_rate": 2.2059337459046234e-06, + "loss": 0.7909, + "step": 14592 + }, + { + "epoch": 2.65, + "learning_rate": 2.2058791408809613e-06, + "loss": 1.1326, + "step": 14593 + }, + { + "epoch": 2.65, + "learning_rate": 2.205824535857299e-06, + "loss": 0.7691, + "step": 14594 + }, + { + "epoch": 2.65, + "learning_rate": 2.205769930833637e-06, + "loss": 0.0546, + "step": 14595 + }, + { + "epoch": 2.65, + "learning_rate": 2.2057153258099747e-06, + "loss": 1.2412, + "step": 14596 + }, + { + "epoch": 2.65, + "learning_rate": 2.2056607207863123e-06, + "loss": 0.9357, + "step": 14597 + }, + { + "epoch": 2.65, + "learning_rate": 2.20560611576265e-06, + "loss": 0.5426, + "step": 14598 + }, + { + "epoch": 2.65, + "learning_rate": 2.205551510738988e-06, + "loss": 0.8067, + "step": 14599 + }, + { + "epoch": 2.66, + "learning_rate": 2.2054969057153257e-06, + "loss": 0.3375, + "step": 14600 + }, + { + "epoch": 2.66, + "learning_rate": 2.2054423006916636e-06, + "loss": 1.5639, + "step": 14601 + }, + { + "epoch": 2.66, + "learning_rate": 2.2053876956680016e-06, + "loss": 0.753, + "step": 14602 + }, + { + "epoch": 2.66, + "learning_rate": 2.2053330906443395e-06, + "loss": 0.3949, + "step": 14603 + }, + { + "epoch": 2.66, + "learning_rate": 2.2052784856206774e-06, + "loss": 0.9479, + "step": 14604 + }, + { + "epoch": 2.66, + "learning_rate": 2.205223880597015e-06, + "loss": 1.5287, + "step": 14605 + }, + { + "epoch": 2.66, + "learning_rate": 2.205169275573353e-06, + "loss": 2.519, + "step": 14606 + }, + { + "epoch": 2.66, + "learning_rate": 2.205114670549691e-06, + "loss": 0.9818, + "step": 14607 + }, + { + "epoch": 2.66, + "learning_rate": 2.2050600655260284e-06, + "loss": 0.7968, + "step": 14608 + }, + { + "epoch": 2.66, + "learning_rate": 2.2050054605023663e-06, + "loss": 0.3704, + "step": 14609 + }, + { + "epoch": 2.66, + "learning_rate": 2.2049508554787043e-06, + "loss": 1.2262, + "step": 14610 + }, + { + "epoch": 2.66, + "learning_rate": 2.2048962504550418e-06, + "loss": 0.3589, + "step": 14611 + }, + { + "epoch": 2.66, + "learning_rate": 2.2048416454313797e-06, + "loss": 1.986, + "step": 14612 + }, + { + "epoch": 2.66, + "learning_rate": 2.2047870404077177e-06, + "loss": 0.2934, + "step": 14613 + }, + { + "epoch": 2.66, + "learning_rate": 2.204732435384055e-06, + "loss": 0.6212, + "step": 14614 + }, + { + "epoch": 2.66, + "learning_rate": 2.204677830360393e-06, + "loss": 0.0795, + "step": 14615 + }, + { + "epoch": 2.66, + "learning_rate": 2.204623225336731e-06, + "loss": 0.9836, + "step": 14616 + }, + { + "epoch": 2.66, + "learning_rate": 2.204568620313069e-06, + "loss": 1.7666, + "step": 14617 + }, + { + "epoch": 2.66, + "learning_rate": 2.204514015289407e-06, + "loss": 0.2799, + "step": 14618 + }, + { + "epoch": 2.66, + "learning_rate": 2.2044594102657445e-06, + "loss": 1.5033, + "step": 14619 + }, + { + "epoch": 2.66, + "learning_rate": 2.2044048052420824e-06, + "loss": 0.927, + "step": 14620 + }, + { + "epoch": 2.66, + "learning_rate": 2.2043502002184204e-06, + "loss": 1.0288, + "step": 14621 + }, + { + "epoch": 2.66, + "learning_rate": 2.204295595194758e-06, + "loss": 0.3018, + "step": 14622 + }, + { + "epoch": 2.66, + "learning_rate": 2.204240990171096e-06, + "loss": 0.5182, + "step": 14623 + }, + { + "epoch": 2.66, + "learning_rate": 2.2041863851474338e-06, + "loss": 0.1046, + "step": 14624 + }, + { + "epoch": 2.66, + "learning_rate": 2.2041317801237713e-06, + "loss": 0.5404, + "step": 14625 + }, + { + "epoch": 2.66, + "learning_rate": 2.2040771751001092e-06, + "loss": 1.0099, + "step": 14626 + }, + { + "epoch": 2.66, + "learning_rate": 2.204022570076447e-06, + "loss": 0.8203, + "step": 14627 + }, + { + "epoch": 2.66, + "learning_rate": 2.2039679650527847e-06, + "loss": 0.7993, + "step": 14628 + }, + { + "epoch": 2.66, + "learning_rate": 2.2039133600291226e-06, + "loss": 0.0625, + "step": 14629 + }, + { + "epoch": 2.66, + "learning_rate": 2.2038587550054606e-06, + "loss": 0.5544, + "step": 14630 + }, + { + "epoch": 2.66, + "learning_rate": 2.2038041499817985e-06, + "loss": 0.2641, + "step": 14631 + }, + { + "epoch": 2.66, + "learning_rate": 2.2037495449581365e-06, + "loss": 0.5994, + "step": 14632 + }, + { + "epoch": 2.66, + "learning_rate": 2.203694939934474e-06, + "loss": 1.3538, + "step": 14633 + }, + { + "epoch": 2.66, + "learning_rate": 2.203640334910812e-06, + "loss": 0.3525, + "step": 14634 + }, + { + "epoch": 2.66, + "learning_rate": 2.20358572988715e-06, + "loss": 0.7105, + "step": 14635 + }, + { + "epoch": 2.66, + "learning_rate": 2.2035311248634874e-06, + "loss": 0.1119, + "step": 14636 + }, + { + "epoch": 2.66, + "learning_rate": 2.2034765198398253e-06, + "loss": 0.7541, + "step": 14637 + }, + { + "epoch": 2.66, + "learning_rate": 2.2034219148161633e-06, + "loss": 0.0753, + "step": 14638 + }, + { + "epoch": 2.66, + "learning_rate": 2.203367309792501e-06, + "loss": 0.568, + "step": 14639 + }, + { + "epoch": 2.66, + "learning_rate": 2.2033127047688387e-06, + "loss": 0.415, + "step": 14640 + }, + { + "epoch": 2.66, + "learning_rate": 2.2032580997451763e-06, + "loss": 1.0135, + "step": 14641 + }, + { + "epoch": 2.66, + "learning_rate": 2.203203494721514e-06, + "loss": 0.4505, + "step": 14642 + }, + { + "epoch": 2.66, + "learning_rate": 2.203148889697852e-06, + "loss": 1.5572, + "step": 14643 + }, + { + "epoch": 2.66, + "learning_rate": 2.20309428467419e-06, + "loss": 0.0159, + "step": 14644 + }, + { + "epoch": 2.66, + "learning_rate": 2.203039679650528e-06, + "loss": 0.3157, + "step": 14645 + }, + { + "epoch": 2.66, + "learning_rate": 2.202985074626866e-06, + "loss": 0.5143, + "step": 14646 + }, + { + "epoch": 2.66, + "learning_rate": 2.2029304696032035e-06, + "loss": 0.471, + "step": 14647 + }, + { + "epoch": 2.66, + "learning_rate": 2.2028758645795414e-06, + "loss": 0.4521, + "step": 14648 + }, + { + "epoch": 2.66, + "learning_rate": 2.2028212595558794e-06, + "loss": 0.3969, + "step": 14649 + }, + { + "epoch": 2.66, + "learning_rate": 2.202766654532217e-06, + "loss": 1.0479, + "step": 14650 + }, + { + "epoch": 2.66, + "learning_rate": 2.202712049508555e-06, + "loss": 0.691, + "step": 14651 + }, + { + "epoch": 2.66, + "learning_rate": 2.202657444484893e-06, + "loss": 1.21, + "step": 14652 + }, + { + "epoch": 2.66, + "learning_rate": 2.2026028394612303e-06, + "loss": 1.9546, + "step": 14653 + }, + { + "epoch": 2.66, + "learning_rate": 2.2025482344375683e-06, + "loss": 1.7449, + "step": 14654 + }, + { + "epoch": 2.67, + "learning_rate": 2.2024936294139058e-06, + "loss": 1.1118, + "step": 14655 + }, + { + "epoch": 2.67, + "learning_rate": 2.2024390243902437e-06, + "loss": 0.9238, + "step": 14656 + }, + { + "epoch": 2.67, + "learning_rate": 2.2023844193665817e-06, + "loss": 1.3786, + "step": 14657 + }, + { + "epoch": 2.67, + "learning_rate": 2.2023298143429196e-06, + "loss": 0.7756, + "step": 14658 + }, + { + "epoch": 2.67, + "learning_rate": 2.2022752093192576e-06, + "loss": 1.3214, + "step": 14659 + }, + { + "epoch": 2.67, + "learning_rate": 2.2022206042955955e-06, + "loss": 0.5761, + "step": 14660 + }, + { + "epoch": 2.67, + "learning_rate": 2.202165999271933e-06, + "loss": 2.2291, + "step": 14661 + }, + { + "epoch": 2.67, + "learning_rate": 2.202111394248271e-06, + "loss": 0.156, + "step": 14662 + }, + { + "epoch": 2.67, + "learning_rate": 2.202056789224609e-06, + "loss": 0.178, + "step": 14663 + }, + { + "epoch": 2.67, + "learning_rate": 2.2020021842009464e-06, + "loss": 1.1782, + "step": 14664 + }, + { + "epoch": 2.67, + "learning_rate": 2.2019475791772844e-06, + "loss": 0.7465, + "step": 14665 + }, + { + "epoch": 2.67, + "learning_rate": 2.2018929741536223e-06, + "loss": 0.0177, + "step": 14666 + }, + { + "epoch": 2.67, + "learning_rate": 2.20183836912996e-06, + "loss": 1.8339, + "step": 14667 + }, + { + "epoch": 2.67, + "learning_rate": 2.2017837641062978e-06, + "loss": 2.0655, + "step": 14668 + }, + { + "epoch": 2.67, + "learning_rate": 2.2017291590826353e-06, + "loss": 1.2107, + "step": 14669 + }, + { + "epoch": 2.67, + "learning_rate": 2.2016745540589732e-06, + "loss": 0.2378, + "step": 14670 + }, + { + "epoch": 2.67, + "learning_rate": 2.201619949035311e-06, + "loss": 0.4881, + "step": 14671 + }, + { + "epoch": 2.67, + "learning_rate": 2.201565344011649e-06, + "loss": 2.8286, + "step": 14672 + }, + { + "epoch": 2.67, + "learning_rate": 2.201510738987987e-06, + "loss": 2.2309, + "step": 14673 + }, + { + "epoch": 2.67, + "learning_rate": 2.201456133964325e-06, + "loss": 0.0213, + "step": 14674 + }, + { + "epoch": 2.67, + "learning_rate": 2.2014015289406625e-06, + "loss": 0.3601, + "step": 14675 + }, + { + "epoch": 2.67, + "learning_rate": 2.2013469239170005e-06, + "loss": 1.4529, + "step": 14676 + }, + { + "epoch": 2.67, + "learning_rate": 2.2012923188933384e-06, + "loss": 0.5904, + "step": 14677 + }, + { + "epoch": 2.67, + "learning_rate": 2.201237713869676e-06, + "loss": 0.2444, + "step": 14678 + }, + { + "epoch": 2.67, + "learning_rate": 2.201183108846014e-06, + "loss": 1.0722, + "step": 14679 + }, + { + "epoch": 2.67, + "learning_rate": 2.201128503822352e-06, + "loss": 1.5552, + "step": 14680 + }, + { + "epoch": 2.67, + "learning_rate": 2.2010738987986893e-06, + "loss": 0.514, + "step": 14681 + }, + { + "epoch": 2.67, + "learning_rate": 2.2010192937750273e-06, + "loss": 0.4589, + "step": 14682 + }, + { + "epoch": 2.67, + "learning_rate": 2.2009646887513652e-06, + "loss": 1.6873, + "step": 14683 + }, + { + "epoch": 2.67, + "learning_rate": 2.200910083727703e-06, + "loss": 1.6471, + "step": 14684 + }, + { + "epoch": 2.67, + "learning_rate": 2.200855478704041e-06, + "loss": 1.5323, + "step": 14685 + }, + { + "epoch": 2.67, + "learning_rate": 2.2008008736803786e-06, + "loss": 0.8791, + "step": 14686 + }, + { + "epoch": 2.67, + "learning_rate": 2.2007462686567166e-06, + "loss": 1.6848, + "step": 14687 + }, + { + "epoch": 2.67, + "learning_rate": 2.2006916636330545e-06, + "loss": 0.0759, + "step": 14688 + }, + { + "epoch": 2.67, + "learning_rate": 2.200637058609392e-06, + "loss": 0.0802, + "step": 14689 + }, + { + "epoch": 2.67, + "learning_rate": 2.20058245358573e-06, + "loss": 0.0953, + "step": 14690 + }, + { + "epoch": 2.67, + "learning_rate": 2.200527848562068e-06, + "loss": 0.5521, + "step": 14691 + }, + { + "epoch": 2.67, + "learning_rate": 2.2004732435384054e-06, + "loss": 0.1861, + "step": 14692 + }, + { + "epoch": 2.67, + "learning_rate": 2.2004186385147434e-06, + "loss": 0.0152, + "step": 14693 + }, + { + "epoch": 2.67, + "learning_rate": 2.2003640334910813e-06, + "loss": 1.5158, + "step": 14694 + }, + { + "epoch": 2.67, + "learning_rate": 2.200309428467419e-06, + "loss": 0.8876, + "step": 14695 + }, + { + "epoch": 2.67, + "learning_rate": 2.200254823443757e-06, + "loss": 1.495, + "step": 14696 + }, + { + "epoch": 2.67, + "learning_rate": 2.2002002184200947e-06, + "loss": 0.0123, + "step": 14697 + }, + { + "epoch": 2.67, + "learning_rate": 2.2001456133964327e-06, + "loss": 0.5111, + "step": 14698 + }, + { + "epoch": 2.67, + "learning_rate": 2.2000910083727706e-06, + "loss": 0.0302, + "step": 14699 + }, + { + "epoch": 2.67, + "learning_rate": 2.200036403349108e-06, + "loss": 2.9726, + "step": 14700 + }, + { + "epoch": 2.67, + "learning_rate": 2.199981798325446e-06, + "loss": 0.6289, + "step": 14701 + }, + { + "epoch": 2.67, + "learning_rate": 2.199927193301784e-06, + "loss": 1.334, + "step": 14702 + }, + { + "epoch": 2.67, + "learning_rate": 2.1998725882781216e-06, + "loss": 0.5994, + "step": 14703 + }, + { + "epoch": 2.67, + "learning_rate": 2.1998179832544595e-06, + "loss": 0.385, + "step": 14704 + }, + { + "epoch": 2.67, + "learning_rate": 2.1997633782307974e-06, + "loss": 1.3385, + "step": 14705 + }, + { + "epoch": 2.67, + "learning_rate": 2.199708773207135e-06, + "loss": 0.3278, + "step": 14706 + }, + { + "epoch": 2.67, + "learning_rate": 2.199654168183473e-06, + "loss": 1.2683, + "step": 14707 + }, + { + "epoch": 2.67, + "learning_rate": 2.199599563159811e-06, + "loss": 1.1092, + "step": 14708 + }, + { + "epoch": 2.67, + "learning_rate": 2.1995449581361484e-06, + "loss": 0.4504, + "step": 14709 + }, + { + "epoch": 2.68, + "learning_rate": 2.1994903531124863e-06, + "loss": 0.1825, + "step": 14710 + }, + { + "epoch": 2.68, + "learning_rate": 2.1994357480888243e-06, + "loss": 1.941, + "step": 14711 + }, + { + "epoch": 2.68, + "learning_rate": 2.199381143065162e-06, + "loss": 2.8088, + "step": 14712 + }, + { + "epoch": 2.68, + "learning_rate": 2.1993265380415e-06, + "loss": 0.675, + "step": 14713 + }, + { + "epoch": 2.68, + "learning_rate": 2.1992719330178377e-06, + "loss": 1.0144, + "step": 14714 + }, + { + "epoch": 2.68, + "learning_rate": 2.1992173279941756e-06, + "loss": 0.4435, + "step": 14715 + }, + { + "epoch": 2.68, + "learning_rate": 2.1991627229705136e-06, + "loss": 1.0594, + "step": 14716 + }, + { + "epoch": 2.68, + "learning_rate": 2.199108117946851e-06, + "loss": 1.1381, + "step": 14717 + }, + { + "epoch": 2.68, + "learning_rate": 2.199053512923189e-06, + "loss": 1.348, + "step": 14718 + }, + { + "epoch": 2.68, + "learning_rate": 2.198998907899527e-06, + "loss": 3.4013, + "step": 14719 + }, + { + "epoch": 2.68, + "learning_rate": 2.1989443028758645e-06, + "loss": 2.0482, + "step": 14720 + }, + { + "epoch": 2.68, + "learning_rate": 2.1988896978522024e-06, + "loss": 0.2727, + "step": 14721 + }, + { + "epoch": 2.68, + "learning_rate": 2.19883509282854e-06, + "loss": 0.6404, + "step": 14722 + }, + { + "epoch": 2.68, + "learning_rate": 2.198780487804878e-06, + "loss": 0.6436, + "step": 14723 + }, + { + "epoch": 2.68, + "learning_rate": 2.198725882781216e-06, + "loss": 0.0156, + "step": 14724 + }, + { + "epoch": 2.68, + "learning_rate": 2.1986712777575538e-06, + "loss": 0.3727, + "step": 14725 + }, + { + "epoch": 2.68, + "learning_rate": 2.1986166727338917e-06, + "loss": 1.6384, + "step": 14726 + }, + { + "epoch": 2.68, + "learning_rate": 2.1985620677102297e-06, + "loss": 0.8324, + "step": 14727 + }, + { + "epoch": 2.68, + "learning_rate": 2.198507462686567e-06, + "loss": 0.8528, + "step": 14728 + }, + { + "epoch": 2.68, + "learning_rate": 2.198452857662905e-06, + "loss": 0.5113, + "step": 14729 + }, + { + "epoch": 2.68, + "learning_rate": 2.198398252639243e-06, + "loss": 0.4123, + "step": 14730 + }, + { + "epoch": 2.68, + "learning_rate": 2.1983436476155806e-06, + "loss": 0.6187, + "step": 14731 + }, + { + "epoch": 2.68, + "learning_rate": 2.1982890425919185e-06, + "loss": 0.0345, + "step": 14732 + }, + { + "epoch": 2.68, + "learning_rate": 2.1982344375682565e-06, + "loss": 0.6331, + "step": 14733 + }, + { + "epoch": 2.68, + "learning_rate": 2.198179832544594e-06, + "loss": 0.4396, + "step": 14734 + }, + { + "epoch": 2.68, + "learning_rate": 2.198125227520932e-06, + "loss": 2.0205, + "step": 14735 + }, + { + "epoch": 2.68, + "learning_rate": 2.1980706224972695e-06, + "loss": 0.3739, + "step": 14736 + }, + { + "epoch": 2.68, + "learning_rate": 2.1980160174736074e-06, + "loss": 0.3458, + "step": 14737 + }, + { + "epoch": 2.68, + "learning_rate": 2.1979614124499453e-06, + "loss": 0.1268, + "step": 14738 + }, + { + "epoch": 2.68, + "learning_rate": 2.1979068074262833e-06, + "loss": 0.3117, + "step": 14739 + }, + { + "epoch": 2.68, + "learning_rate": 2.1978522024026212e-06, + "loss": 1.6084, + "step": 14740 + }, + { + "epoch": 2.68, + "learning_rate": 2.197797597378959e-06, + "loss": 0.5069, + "step": 14741 + }, + { + "epoch": 2.68, + "learning_rate": 2.1977429923552967e-06, + "loss": 0.0542, + "step": 14742 + }, + { + "epoch": 2.68, + "learning_rate": 2.1976883873316346e-06, + "loss": 0.8851, + "step": 14743 + }, + { + "epoch": 2.68, + "learning_rate": 2.1976337823079726e-06, + "loss": 1.0933, + "step": 14744 + }, + { + "epoch": 2.68, + "learning_rate": 2.19757917728431e-06, + "loss": 2.3386, + "step": 14745 + }, + { + "epoch": 2.68, + "learning_rate": 2.197524572260648e-06, + "loss": 1.1178, + "step": 14746 + }, + { + "epoch": 2.68, + "learning_rate": 2.197469967236986e-06, + "loss": 2.7826, + "step": 14747 + }, + { + "epoch": 2.68, + "learning_rate": 2.1974153622133235e-06, + "loss": 1.6734, + "step": 14748 + }, + { + "epoch": 2.68, + "learning_rate": 2.1973607571896614e-06, + "loss": 0.0579, + "step": 14749 + }, + { + "epoch": 2.68, + "learning_rate": 2.197306152165999e-06, + "loss": 0.2632, + "step": 14750 + }, + { + "epoch": 2.68, + "learning_rate": 2.197251547142337e-06, + "loss": 0.3051, + "step": 14751 + }, + { + "epoch": 2.68, + "learning_rate": 2.197196942118675e-06, + "loss": 0.2868, + "step": 14752 + }, + { + "epoch": 2.68, + "learning_rate": 2.197142337095013e-06, + "loss": 0.1471, + "step": 14753 + }, + { + "epoch": 2.68, + "learning_rate": 2.1970877320713507e-06, + "loss": 0.5424, + "step": 14754 + }, + { + "epoch": 2.68, + "learning_rate": 2.1970331270476887e-06, + "loss": 0.9194, + "step": 14755 + }, + { + "epoch": 2.68, + "learning_rate": 2.196978522024026e-06, + "loss": 0.4663, + "step": 14756 + }, + { + "epoch": 2.68, + "learning_rate": 2.196923917000364e-06, + "loss": 1.4646, + "step": 14757 + }, + { + "epoch": 2.68, + "learning_rate": 2.196869311976702e-06, + "loss": 0.5209, + "step": 14758 + }, + { + "epoch": 2.68, + "learning_rate": 2.1968147069530396e-06, + "loss": 1.3898, + "step": 14759 + }, + { + "epoch": 2.68, + "learning_rate": 2.1967601019293776e-06, + "loss": 0.0506, + "step": 14760 + }, + { + "epoch": 2.68, + "learning_rate": 2.1967054969057155e-06, + "loss": 0.5398, + "step": 14761 + }, + { + "epoch": 2.68, + "learning_rate": 2.196650891882053e-06, + "loss": 0.4195, + "step": 14762 + }, + { + "epoch": 2.68, + "learning_rate": 2.196596286858391e-06, + "loss": 0.9907, + "step": 14763 + }, + { + "epoch": 2.68, + "learning_rate": 2.196541681834729e-06, + "loss": 1.1366, + "step": 14764 + }, + { + "epoch": 2.69, + "learning_rate": 2.196487076811067e-06, + "loss": 0.788, + "step": 14765 + }, + { + "epoch": 2.69, + "learning_rate": 2.196432471787405e-06, + "loss": 0.1537, + "step": 14766 + }, + { + "epoch": 2.69, + "learning_rate": 2.1963778667637423e-06, + "loss": 0.4315, + "step": 14767 + }, + { + "epoch": 2.69, + "learning_rate": 2.1963232617400803e-06, + "loss": 4.3229, + "step": 14768 + }, + { + "epoch": 2.69, + "learning_rate": 2.196268656716418e-06, + "loss": 0.4865, + "step": 14769 + }, + { + "epoch": 2.69, + "learning_rate": 2.1962140516927557e-06, + "loss": 1.4228, + "step": 14770 + }, + { + "epoch": 2.69, + "learning_rate": 2.1961594466690937e-06, + "loss": 0.5662, + "step": 14771 + }, + { + "epoch": 2.69, + "learning_rate": 2.1961048416454316e-06, + "loss": 0.7738, + "step": 14772 + }, + { + "epoch": 2.69, + "learning_rate": 2.196050236621769e-06, + "loss": 0.3533, + "step": 14773 + }, + { + "epoch": 2.69, + "learning_rate": 2.195995631598107e-06, + "loss": 0.1091, + "step": 14774 + }, + { + "epoch": 2.69, + "learning_rate": 2.195941026574445e-06, + "loss": 1.8908, + "step": 14775 + }, + { + "epoch": 2.69, + "learning_rate": 2.1958864215507825e-06, + "loss": 0.1543, + "step": 14776 + }, + { + "epoch": 2.69, + "learning_rate": 2.1958318165271205e-06, + "loss": 0.561, + "step": 14777 + }, + { + "epoch": 2.69, + "learning_rate": 2.1957772115034584e-06, + "loss": 0.5496, + "step": 14778 + }, + { + "epoch": 2.69, + "learning_rate": 2.1957226064797964e-06, + "loss": 1.5127, + "step": 14779 + }, + { + "epoch": 2.69, + "learning_rate": 2.1956680014561343e-06, + "loss": 0.3339, + "step": 14780 + }, + { + "epoch": 2.69, + "learning_rate": 2.195613396432472e-06, + "loss": 0.3684, + "step": 14781 + }, + { + "epoch": 2.69, + "learning_rate": 2.1955587914088098e-06, + "loss": 0.5743, + "step": 14782 + }, + { + "epoch": 2.69, + "learning_rate": 2.1955041863851477e-06, + "loss": 2.5209, + "step": 14783 + }, + { + "epoch": 2.69, + "learning_rate": 2.1954495813614852e-06, + "loss": 0.9402, + "step": 14784 + }, + { + "epoch": 2.69, + "learning_rate": 2.195394976337823e-06, + "loss": 0.0504, + "step": 14785 + }, + { + "epoch": 2.69, + "learning_rate": 2.195340371314161e-06, + "loss": 0.7418, + "step": 14786 + }, + { + "epoch": 2.69, + "learning_rate": 2.1952857662904986e-06, + "loss": 0.0286, + "step": 14787 + }, + { + "epoch": 2.69, + "learning_rate": 2.1952311612668366e-06, + "loss": 0.1019, + "step": 14788 + }, + { + "epoch": 2.69, + "learning_rate": 2.1951765562431745e-06, + "loss": 0.4704, + "step": 14789 + }, + { + "epoch": 2.69, + "learning_rate": 2.195121951219512e-06, + "loss": 1.5348, + "step": 14790 + }, + { + "epoch": 2.69, + "learning_rate": 2.19506734619585e-06, + "loss": 2.2507, + "step": 14791 + }, + { + "epoch": 2.69, + "learning_rate": 2.195012741172188e-06, + "loss": 0.4672, + "step": 14792 + }, + { + "epoch": 2.69, + "learning_rate": 2.194958136148526e-06, + "loss": 0.4028, + "step": 14793 + }, + { + "epoch": 2.69, + "learning_rate": 2.194903531124864e-06, + "loss": 1.5979, + "step": 14794 + }, + { + "epoch": 2.69, + "learning_rate": 2.1948489261012013e-06, + "loss": 0.4019, + "step": 14795 + }, + { + "epoch": 2.69, + "learning_rate": 2.1947943210775393e-06, + "loss": 1.7637, + "step": 14796 + }, + { + "epoch": 2.69, + "learning_rate": 2.1947397160538772e-06, + "loss": 1.1866, + "step": 14797 + }, + { + "epoch": 2.69, + "learning_rate": 2.1946851110302147e-06, + "loss": 0.5161, + "step": 14798 + }, + { + "epoch": 2.69, + "learning_rate": 2.1946305060065527e-06, + "loss": 1.4224, + "step": 14799 + }, + { + "epoch": 2.69, + "learning_rate": 2.1945759009828906e-06, + "loss": 0.0079, + "step": 14800 + }, + { + "epoch": 2.69, + "learning_rate": 2.194521295959228e-06, + "loss": 0.977, + "step": 14801 + }, + { + "epoch": 2.69, + "learning_rate": 2.194466690935566e-06, + "loss": 0.0307, + "step": 14802 + }, + { + "epoch": 2.69, + "learning_rate": 2.1944120859119036e-06, + "loss": 0.3804, + "step": 14803 + }, + { + "epoch": 2.69, + "learning_rate": 2.1943574808882416e-06, + "loss": 0.9316, + "step": 14804 + }, + { + "epoch": 2.69, + "learning_rate": 2.1943028758645795e-06, + "loss": 1.2722, + "step": 14805 + }, + { + "epoch": 2.69, + "learning_rate": 2.1942482708409174e-06, + "loss": 1.1057, + "step": 14806 + }, + { + "epoch": 2.69, + "learning_rate": 2.1941936658172554e-06, + "loss": 1.264, + "step": 14807 + }, + { + "epoch": 2.69, + "learning_rate": 2.1941390607935933e-06, + "loss": 0.2274, + "step": 14808 + }, + { + "epoch": 2.69, + "learning_rate": 2.194084455769931e-06, + "loss": 0.0094, + "step": 14809 + }, + { + "epoch": 2.69, + "learning_rate": 2.194029850746269e-06, + "loss": 1.268, + "step": 14810 + }, + { + "epoch": 2.69, + "learning_rate": 2.1939752457226067e-06, + "loss": 0.8132, + "step": 14811 + }, + { + "epoch": 2.69, + "learning_rate": 2.1939206406989443e-06, + "loss": 0.5028, + "step": 14812 + }, + { + "epoch": 2.69, + "learning_rate": 2.193866035675282e-06, + "loss": 0.6051, + "step": 14813 + }, + { + "epoch": 2.69, + "learning_rate": 2.19381143065162e-06, + "loss": 0.027, + "step": 14814 + }, + { + "epoch": 2.69, + "learning_rate": 2.1937568256279577e-06, + "loss": 0.3203, + "step": 14815 + }, + { + "epoch": 2.69, + "learning_rate": 2.1937022206042956e-06, + "loss": 2.2659, + "step": 14816 + }, + { + "epoch": 2.69, + "learning_rate": 2.193647615580633e-06, + "loss": 0.8183, + "step": 14817 + }, + { + "epoch": 2.69, + "learning_rate": 2.193593010556971e-06, + "loss": 0.4443, + "step": 14818 + }, + { + "epoch": 2.69, + "learning_rate": 2.193538405533309e-06, + "loss": 1.5526, + "step": 14819 + }, + { + "epoch": 2.7, + "learning_rate": 2.193483800509647e-06, + "loss": 0.0555, + "step": 14820 + }, + { + "epoch": 2.7, + "learning_rate": 2.193429195485985e-06, + "loss": 0.451, + "step": 14821 + }, + { + "epoch": 2.7, + "learning_rate": 2.193374590462323e-06, + "loss": 0.3715, + "step": 14822 + }, + { + "epoch": 2.7, + "learning_rate": 2.1933199854386604e-06, + "loss": 1.3874, + "step": 14823 + }, + { + "epoch": 2.7, + "learning_rate": 2.1932653804149983e-06, + "loss": 0.661, + "step": 14824 + }, + { + "epoch": 2.7, + "learning_rate": 2.1932107753913363e-06, + "loss": 3.7537, + "step": 14825 + }, + { + "epoch": 2.7, + "learning_rate": 2.1931561703676738e-06, + "loss": 0.2273, + "step": 14826 + }, + { + "epoch": 2.7, + "learning_rate": 2.1931015653440117e-06, + "loss": 0.3445, + "step": 14827 + }, + { + "epoch": 2.7, + "learning_rate": 2.1930469603203497e-06, + "loss": 0.0346, + "step": 14828 + }, + { + "epoch": 2.7, + "learning_rate": 2.192992355296687e-06, + "loss": 1.4629, + "step": 14829 + }, + { + "epoch": 2.7, + "learning_rate": 2.192937750273025e-06, + "loss": 1.742, + "step": 14830 + }, + { + "epoch": 2.7, + "learning_rate": 2.1928831452493626e-06, + "loss": 0.5616, + "step": 14831 + }, + { + "epoch": 2.7, + "learning_rate": 2.1928285402257006e-06, + "loss": 0.3832, + "step": 14832 + }, + { + "epoch": 2.7, + "learning_rate": 2.1927739352020385e-06, + "loss": 1.9215, + "step": 14833 + }, + { + "epoch": 2.7, + "learning_rate": 2.1927193301783765e-06, + "loss": 0.8763, + "step": 14834 + }, + { + "epoch": 2.7, + "learning_rate": 2.1926647251547144e-06, + "loss": 0.2384, + "step": 14835 + }, + { + "epoch": 2.7, + "learning_rate": 2.1926101201310524e-06, + "loss": 0.257, + "step": 14836 + }, + { + "epoch": 2.7, + "learning_rate": 2.19255551510739e-06, + "loss": 1.7276, + "step": 14837 + }, + { + "epoch": 2.7, + "learning_rate": 2.192500910083728e-06, + "loss": 1.7372, + "step": 14838 + }, + { + "epoch": 2.7, + "learning_rate": 2.1924463050600658e-06, + "loss": 0.0682, + "step": 14839 + }, + { + "epoch": 2.7, + "learning_rate": 2.1923917000364033e-06, + "loss": 3.1014, + "step": 14840 + }, + { + "epoch": 2.7, + "learning_rate": 2.1923370950127412e-06, + "loss": 0.2993, + "step": 14841 + }, + { + "epoch": 2.7, + "learning_rate": 2.192282489989079e-06, + "loss": 0.2434, + "step": 14842 + }, + { + "epoch": 2.7, + "learning_rate": 2.1922278849654167e-06, + "loss": 0.1623, + "step": 14843 + }, + { + "epoch": 2.7, + "learning_rate": 2.1921732799417546e-06, + "loss": 0.5256, + "step": 14844 + }, + { + "epoch": 2.7, + "learning_rate": 2.1921186749180926e-06, + "loss": 1.0643, + "step": 14845 + }, + { + "epoch": 2.7, + "learning_rate": 2.1920640698944305e-06, + "loss": 0.4028, + "step": 14846 + }, + { + "epoch": 2.7, + "learning_rate": 2.1920094648707685e-06, + "loss": 1.4505, + "step": 14847 + }, + { + "epoch": 2.7, + "learning_rate": 2.191954859847106e-06, + "loss": 0.5172, + "step": 14848 + }, + { + "epoch": 2.7, + "learning_rate": 2.191900254823444e-06, + "loss": 1.7361, + "step": 14849 + }, + { + "epoch": 2.7, + "learning_rate": 2.191845649799782e-06, + "loss": 0.1038, + "step": 14850 + }, + { + "epoch": 2.7, + "learning_rate": 2.1917910447761194e-06, + "loss": 1.3005, + "step": 14851 + }, + { + "epoch": 2.7, + "learning_rate": 2.1917364397524573e-06, + "loss": 0.8891, + "step": 14852 + }, + { + "epoch": 2.7, + "learning_rate": 2.1916818347287953e-06, + "loss": 0.6021, + "step": 14853 + }, + { + "epoch": 2.7, + "learning_rate": 2.191627229705133e-06, + "loss": 0.2512, + "step": 14854 + }, + { + "epoch": 2.7, + "learning_rate": 2.1915726246814707e-06, + "loss": 0.3601, + "step": 14855 + }, + { + "epoch": 2.7, + "learning_rate": 2.1915180196578087e-06, + "loss": 0.2112, + "step": 14856 + }, + { + "epoch": 2.7, + "learning_rate": 2.191463414634146e-06, + "loss": 0.0771, + "step": 14857 + }, + { + "epoch": 2.7, + "learning_rate": 2.191408809610484e-06, + "loss": 0.7192, + "step": 14858 + }, + { + "epoch": 2.7, + "learning_rate": 2.191354204586822e-06, + "loss": 1.4414, + "step": 14859 + }, + { + "epoch": 2.7, + "learning_rate": 2.19129959956316e-06, + "loss": 2.2725, + "step": 14860 + }, + { + "epoch": 2.7, + "learning_rate": 2.191244994539498e-06, + "loss": 1.7438, + "step": 14861 + }, + { + "epoch": 2.7, + "learning_rate": 2.1911903895158355e-06, + "loss": 0.0289, + "step": 14862 + }, + { + "epoch": 2.7, + "learning_rate": 2.1911357844921734e-06, + "loss": 0.7695, + "step": 14863 + }, + { + "epoch": 2.7, + "learning_rate": 2.1910811794685114e-06, + "loss": 1.0344, + "step": 14864 + }, + { + "epoch": 2.7, + "learning_rate": 2.191026574444849e-06, + "loss": 1.3843, + "step": 14865 + }, + { + "epoch": 2.7, + "learning_rate": 2.190971969421187e-06, + "loss": 0.5542, + "step": 14866 + }, + { + "epoch": 2.7, + "learning_rate": 2.190917364397525e-06, + "loss": 0.5338, + "step": 14867 + }, + { + "epoch": 2.7, + "learning_rate": 2.1908627593738623e-06, + "loss": 0.344, + "step": 14868 + }, + { + "epoch": 2.7, + "learning_rate": 2.1908081543502003e-06, + "loss": 2.0352, + "step": 14869 + }, + { + "epoch": 2.7, + "learning_rate": 2.190753549326538e-06, + "loss": 0.1322, + "step": 14870 + }, + { + "epoch": 2.7, + "learning_rate": 2.1906989443028757e-06, + "loss": 0.3485, + "step": 14871 + }, + { + "epoch": 2.7, + "learning_rate": 2.1906443392792137e-06, + "loss": 1.4725, + "step": 14872 + }, + { + "epoch": 2.7, + "learning_rate": 2.1905897342555516e-06, + "loss": 0.0644, + "step": 14873 + }, + { + "epoch": 2.7, + "learning_rate": 2.1905351292318895e-06, + "loss": 0.4831, + "step": 14874 + }, + { + "epoch": 2.71, + "learning_rate": 2.1904805242082275e-06, + "loss": 0.9918, + "step": 14875 + }, + { + "epoch": 2.71, + "learning_rate": 2.190425919184565e-06, + "loss": 1.0616, + "step": 14876 + }, + { + "epoch": 2.71, + "learning_rate": 2.190371314160903e-06, + "loss": 0.3336, + "step": 14877 + }, + { + "epoch": 2.71, + "learning_rate": 2.190316709137241e-06, + "loss": 0.6239, + "step": 14878 + }, + { + "epoch": 2.71, + "learning_rate": 2.1902621041135784e-06, + "loss": 0.9655, + "step": 14879 + }, + { + "epoch": 2.71, + "learning_rate": 2.1902074990899164e-06, + "loss": 0.5006, + "step": 14880 + }, + { + "epoch": 2.71, + "learning_rate": 2.1901528940662543e-06, + "loss": 0.3378, + "step": 14881 + }, + { + "epoch": 2.71, + "learning_rate": 2.190098289042592e-06, + "loss": 0.4086, + "step": 14882 + }, + { + "epoch": 2.71, + "learning_rate": 2.1900436840189298e-06, + "loss": 4.396, + "step": 14883 + }, + { + "epoch": 2.71, + "learning_rate": 2.1899890789952673e-06, + "loss": 1.4429, + "step": 14884 + }, + { + "epoch": 2.71, + "learning_rate": 2.1899344739716052e-06, + "loss": 0.0097, + "step": 14885 + }, + { + "epoch": 2.71, + "learning_rate": 2.189879868947943e-06, + "loss": 0.9052, + "step": 14886 + }, + { + "epoch": 2.71, + "learning_rate": 2.189825263924281e-06, + "loss": 0.6857, + "step": 14887 + }, + { + "epoch": 2.71, + "learning_rate": 2.189770658900619e-06, + "loss": 0.7119, + "step": 14888 + }, + { + "epoch": 2.71, + "learning_rate": 2.189716053876957e-06, + "loss": 0.5528, + "step": 14889 + }, + { + "epoch": 2.71, + "learning_rate": 2.1896614488532945e-06, + "loss": 0.2369, + "step": 14890 + }, + { + "epoch": 2.71, + "learning_rate": 2.1896068438296325e-06, + "loss": 0.3836, + "step": 14891 + }, + { + "epoch": 2.71, + "learning_rate": 2.1895522388059704e-06, + "loss": 0.0383, + "step": 14892 + }, + { + "epoch": 2.71, + "learning_rate": 2.189497633782308e-06, + "loss": 0.5989, + "step": 14893 + }, + { + "epoch": 2.71, + "learning_rate": 2.189443028758646e-06, + "loss": 0.7935, + "step": 14894 + }, + { + "epoch": 2.71, + "learning_rate": 2.189388423734984e-06, + "loss": 0.0359, + "step": 14895 + }, + { + "epoch": 2.71, + "learning_rate": 2.1893338187113213e-06, + "loss": 0.6459, + "step": 14896 + }, + { + "epoch": 2.71, + "learning_rate": 2.1892792136876593e-06, + "loss": 0.4956, + "step": 14897 + }, + { + "epoch": 2.71, + "learning_rate": 2.189224608663997e-06, + "loss": 1.1004, + "step": 14898 + }, + { + "epoch": 2.71, + "learning_rate": 2.1891700036403347e-06, + "loss": 0.0197, + "step": 14899 + }, + { + "epoch": 2.71, + "learning_rate": 2.1891153986166727e-06, + "loss": 1.4234, + "step": 14900 + }, + { + "epoch": 2.71, + "learning_rate": 2.1890607935930106e-06, + "loss": 0.7153, + "step": 14901 + }, + { + "epoch": 2.71, + "learning_rate": 2.1890061885693486e-06, + "loss": 0.3641, + "step": 14902 + }, + { + "epoch": 2.71, + "learning_rate": 2.1889515835456865e-06, + "loss": 1.2925, + "step": 14903 + }, + { + "epoch": 2.71, + "learning_rate": 2.188896978522024e-06, + "loss": 1.3222, + "step": 14904 + }, + { + "epoch": 2.71, + "learning_rate": 2.188842373498362e-06, + "loss": 0.1264, + "step": 14905 + }, + { + "epoch": 2.71, + "learning_rate": 2.1887877684747e-06, + "loss": 0.0276, + "step": 14906 + }, + { + "epoch": 2.71, + "learning_rate": 2.1887331634510374e-06, + "loss": 0.7973, + "step": 14907 + }, + { + "epoch": 2.71, + "learning_rate": 2.1886785584273754e-06, + "loss": 0.1938, + "step": 14908 + }, + { + "epoch": 2.71, + "learning_rate": 2.1886239534037133e-06, + "loss": 0.506, + "step": 14909 + }, + { + "epoch": 2.71, + "learning_rate": 2.188569348380051e-06, + "loss": 0.1886, + "step": 14910 + }, + { + "epoch": 2.71, + "learning_rate": 2.188514743356389e-06, + "loss": 0.7516, + "step": 14911 + }, + { + "epoch": 2.71, + "learning_rate": 2.1884601383327263e-06, + "loss": 0.063, + "step": 14912 + }, + { + "epoch": 2.71, + "learning_rate": 2.1884055333090643e-06, + "loss": 0.6072, + "step": 14913 + }, + { + "epoch": 2.71, + "learning_rate": 2.188350928285402e-06, + "loss": 0.7755, + "step": 14914 + }, + { + "epoch": 2.71, + "learning_rate": 2.18829632326174e-06, + "loss": 1.6309, + "step": 14915 + }, + { + "epoch": 2.71, + "learning_rate": 2.188241718238078e-06, + "loss": 1.5241, + "step": 14916 + }, + { + "epoch": 2.71, + "learning_rate": 2.188187113214416e-06, + "loss": 0.85, + "step": 14917 + }, + { + "epoch": 2.71, + "learning_rate": 2.1881325081907536e-06, + "loss": 0.511, + "step": 14918 + }, + { + "epoch": 2.71, + "learning_rate": 2.1880779031670915e-06, + "loss": 1.2063, + "step": 14919 + }, + { + "epoch": 2.71, + "learning_rate": 2.1880232981434294e-06, + "loss": 0.4909, + "step": 14920 + }, + { + "epoch": 2.71, + "learning_rate": 2.187968693119767e-06, + "loss": 1.1734, + "step": 14921 + }, + { + "epoch": 2.71, + "learning_rate": 2.187914088096105e-06, + "loss": 1.3306, + "step": 14922 + }, + { + "epoch": 2.71, + "learning_rate": 2.187859483072443e-06, + "loss": 0.9218, + "step": 14923 + }, + { + "epoch": 2.71, + "learning_rate": 2.1878048780487804e-06, + "loss": 0.3227, + "step": 14924 + }, + { + "epoch": 2.71, + "learning_rate": 2.1877502730251183e-06, + "loss": 0.1525, + "step": 14925 + }, + { + "epoch": 2.71, + "learning_rate": 2.187695668001456e-06, + "loss": 0.5, + "step": 14926 + }, + { + "epoch": 2.71, + "learning_rate": 2.1876410629777938e-06, + "loss": 0.4936, + "step": 14927 + }, + { + "epoch": 2.71, + "learning_rate": 2.1875864579541317e-06, + "loss": 0.6087, + "step": 14928 + }, + { + "epoch": 2.71, + "learning_rate": 2.1875318529304697e-06, + "loss": 1.0301, + "step": 14929 + }, + { + "epoch": 2.72, + "learning_rate": 2.1874772479068076e-06, + "loss": 1.4991, + "step": 14930 + }, + { + "epoch": 2.72, + "learning_rate": 2.1874226428831455e-06, + "loss": 0.8948, + "step": 14931 + }, + { + "epoch": 2.72, + "learning_rate": 2.187368037859483e-06, + "loss": 2.1458, + "step": 14932 + }, + { + "epoch": 2.72, + "learning_rate": 2.187313432835821e-06, + "loss": 1.4121, + "step": 14933 + }, + { + "epoch": 2.72, + "learning_rate": 2.187258827812159e-06, + "loss": 0.8647, + "step": 14934 + }, + { + "epoch": 2.72, + "learning_rate": 2.1872042227884965e-06, + "loss": 0.3258, + "step": 14935 + }, + { + "epoch": 2.72, + "learning_rate": 2.1871496177648344e-06, + "loss": 1.0463, + "step": 14936 + }, + { + "epoch": 2.72, + "learning_rate": 2.1870950127411724e-06, + "loss": 1.2799, + "step": 14937 + }, + { + "epoch": 2.72, + "learning_rate": 2.18704040771751e-06, + "loss": 0.0495, + "step": 14938 + }, + { + "epoch": 2.72, + "learning_rate": 2.186985802693848e-06, + "loss": 0.0195, + "step": 14939 + }, + { + "epoch": 2.72, + "learning_rate": 2.1869311976701858e-06, + "loss": 0.3598, + "step": 14940 + }, + { + "epoch": 2.72, + "learning_rate": 2.1868765926465237e-06, + "loss": 0.1672, + "step": 14941 + }, + { + "epoch": 2.72, + "learning_rate": 2.1868219876228617e-06, + "loss": 1.7059, + "step": 14942 + }, + { + "epoch": 2.72, + "learning_rate": 2.186767382599199e-06, + "loss": 0.3286, + "step": 14943 + }, + { + "epoch": 2.72, + "learning_rate": 2.186712777575537e-06, + "loss": 1.34, + "step": 14944 + }, + { + "epoch": 2.72, + "learning_rate": 2.186658172551875e-06, + "loss": 0.6244, + "step": 14945 + }, + { + "epoch": 2.72, + "learning_rate": 2.1866035675282126e-06, + "loss": 1.6267, + "step": 14946 + }, + { + "epoch": 2.72, + "learning_rate": 2.1865489625045505e-06, + "loss": 0.9665, + "step": 14947 + }, + { + "epoch": 2.72, + "learning_rate": 2.1864943574808885e-06, + "loss": 2.403, + "step": 14948 + }, + { + "epoch": 2.72, + "learning_rate": 2.186439752457226e-06, + "loss": 2.1667, + "step": 14949 + }, + { + "epoch": 2.72, + "learning_rate": 2.186385147433564e-06, + "loss": 2.0251, + "step": 14950 + }, + { + "epoch": 2.72, + "learning_rate": 2.1863305424099014e-06, + "loss": 0.0179, + "step": 14951 + }, + { + "epoch": 2.72, + "learning_rate": 2.1862759373862394e-06, + "loss": 0.4904, + "step": 14952 + }, + { + "epoch": 2.72, + "learning_rate": 2.1862213323625773e-06, + "loss": 0.1048, + "step": 14953 + }, + { + "epoch": 2.72, + "learning_rate": 2.1861667273389153e-06, + "loss": 0.7902, + "step": 14954 + }, + { + "epoch": 2.72, + "learning_rate": 2.1861121223152532e-06, + "loss": 0.6781, + "step": 14955 + }, + { + "epoch": 2.72, + "learning_rate": 2.186057517291591e-06, + "loss": 2.2679, + "step": 14956 + }, + { + "epoch": 2.72, + "learning_rate": 2.1860029122679287e-06, + "loss": 0.2154, + "step": 14957 + }, + { + "epoch": 2.72, + "learning_rate": 2.1859483072442666e-06, + "loss": 0.4326, + "step": 14958 + }, + { + "epoch": 2.72, + "learning_rate": 2.1858937022206046e-06, + "loss": 0.3038, + "step": 14959 + }, + { + "epoch": 2.72, + "learning_rate": 2.185839097196942e-06, + "loss": 3.1944, + "step": 14960 + }, + { + "epoch": 2.72, + "learning_rate": 2.18578449217328e-06, + "loss": 0.0173, + "step": 14961 + }, + { + "epoch": 2.72, + "learning_rate": 2.185729887149618e-06, + "loss": 1.4984, + "step": 14962 + }, + { + "epoch": 2.72, + "learning_rate": 2.1856752821259555e-06, + "loss": 3.0855, + "step": 14963 + }, + { + "epoch": 2.72, + "learning_rate": 2.1856206771022934e-06, + "loss": 0.0257, + "step": 14964 + }, + { + "epoch": 2.72, + "learning_rate": 2.185566072078631e-06, + "loss": 0.039, + "step": 14965 + }, + { + "epoch": 2.72, + "learning_rate": 2.185511467054969e-06, + "loss": 0.0521, + "step": 14966 + }, + { + "epoch": 2.72, + "learning_rate": 2.185456862031307e-06, + "loss": 0.6966, + "step": 14967 + }, + { + "epoch": 2.72, + "learning_rate": 2.185402257007645e-06, + "loss": 0.0383, + "step": 14968 + }, + { + "epoch": 2.72, + "learning_rate": 2.1853476519839827e-06, + "loss": 1.5157, + "step": 14969 + }, + { + "epoch": 2.72, + "learning_rate": 2.1852930469603207e-06, + "loss": 0.1898, + "step": 14970 + }, + { + "epoch": 2.72, + "learning_rate": 2.185238441936658e-06, + "loss": 0.7196, + "step": 14971 + }, + { + "epoch": 2.72, + "learning_rate": 2.185183836912996e-06, + "loss": 0.0276, + "step": 14972 + }, + { + "epoch": 2.72, + "learning_rate": 2.185129231889334e-06, + "loss": 0.0106, + "step": 14973 + }, + { + "epoch": 2.72, + "learning_rate": 2.1850746268656716e-06, + "loss": 0.7226, + "step": 14974 + }, + { + "epoch": 2.72, + "learning_rate": 2.1850200218420095e-06, + "loss": 0.2335, + "step": 14975 + }, + { + "epoch": 2.72, + "learning_rate": 2.1849654168183475e-06, + "loss": 0.023, + "step": 14976 + }, + { + "epoch": 2.72, + "learning_rate": 2.184910811794685e-06, + "loss": 0.8166, + "step": 14977 + }, + { + "epoch": 2.72, + "learning_rate": 2.184856206771023e-06, + "loss": 1.6099, + "step": 14978 + }, + { + "epoch": 2.72, + "learning_rate": 2.1848016017473605e-06, + "loss": 2.1694, + "step": 14979 + }, + { + "epoch": 2.72, + "learning_rate": 2.1847469967236984e-06, + "loss": 1.8927, + "step": 14980 + }, + { + "epoch": 2.72, + "learning_rate": 2.1846923917000364e-06, + "loss": 2.3576, + "step": 14981 + }, + { + "epoch": 2.72, + "learning_rate": 2.1846377866763743e-06, + "loss": 0.528, + "step": 14982 + }, + { + "epoch": 2.72, + "learning_rate": 2.1845831816527122e-06, + "loss": 1.3559, + "step": 14983 + }, + { + "epoch": 2.72, + "learning_rate": 2.18452857662905e-06, + "loss": 0.9168, + "step": 14984 + }, + { + "epoch": 2.73, + "learning_rate": 2.1844739716053877e-06, + "loss": 0.9222, + "step": 14985 + }, + { + "epoch": 2.73, + "learning_rate": 2.1844193665817257e-06, + "loss": 2.2896, + "step": 14986 + }, + { + "epoch": 2.73, + "learning_rate": 2.1843647615580636e-06, + "loss": 2.4356, + "step": 14987 + }, + { + "epoch": 2.73, + "learning_rate": 2.184310156534401e-06, + "loss": 0.3258, + "step": 14988 + }, + { + "epoch": 2.73, + "learning_rate": 2.184255551510739e-06, + "loss": 0.2579, + "step": 14989 + }, + { + "epoch": 2.73, + "learning_rate": 2.184200946487077e-06, + "loss": 1.7048, + "step": 14990 + }, + { + "epoch": 2.73, + "learning_rate": 2.1841463414634145e-06, + "loss": 0.4085, + "step": 14991 + }, + { + "epoch": 2.73, + "learning_rate": 2.1840917364397525e-06, + "loss": 0.1442, + "step": 14992 + }, + { + "epoch": 2.73, + "learning_rate": 2.18403713141609e-06, + "loss": 1.8043, + "step": 14993 + }, + { + "epoch": 2.73, + "learning_rate": 2.183982526392428e-06, + "loss": 1.1424, + "step": 14994 + }, + { + "epoch": 2.73, + "learning_rate": 2.183927921368766e-06, + "loss": 0.1459, + "step": 14995 + }, + { + "epoch": 2.73, + "learning_rate": 2.183873316345104e-06, + "loss": 0.6034, + "step": 14996 + }, + { + "epoch": 2.73, + "learning_rate": 2.1838187113214418e-06, + "loss": 1.0485, + "step": 14997 + }, + { + "epoch": 2.73, + "learning_rate": 2.1837641062977797e-06, + "loss": 1.9856, + "step": 14998 + }, + { + "epoch": 2.73, + "learning_rate": 2.1837095012741172e-06, + "loss": 0.5962, + "step": 14999 + }, + { + "epoch": 2.73, + "learning_rate": 2.183654896250455e-06, + "loss": 0.556, + "step": 15000 + }, + { + "epoch": 2.73, + "learning_rate": 2.183600291226793e-06, + "loss": 0.1408, + "step": 15001 + }, + { + "epoch": 2.73, + "learning_rate": 2.1835456862031306e-06, + "loss": 3.9085, + "step": 15002 + }, + { + "epoch": 2.73, + "learning_rate": 2.1834910811794686e-06, + "loss": 2.1687, + "step": 15003 + }, + { + "epoch": 2.73, + "learning_rate": 2.1834364761558065e-06, + "loss": 0.4102, + "step": 15004 + }, + { + "epoch": 2.73, + "learning_rate": 2.183381871132144e-06, + "loss": 1.2984, + "step": 15005 + }, + { + "epoch": 2.73, + "learning_rate": 2.183327266108482e-06, + "loss": 1.5803, + "step": 15006 + }, + { + "epoch": 2.73, + "learning_rate": 2.1832726610848195e-06, + "loss": 0.6406, + "step": 15007 + }, + { + "epoch": 2.73, + "learning_rate": 2.1832180560611574e-06, + "loss": 0.2106, + "step": 15008 + }, + { + "epoch": 2.73, + "learning_rate": 2.1831634510374954e-06, + "loss": 0.4995, + "step": 15009 + }, + { + "epoch": 2.73, + "learning_rate": 2.1831088460138333e-06, + "loss": 0.0705, + "step": 15010 + }, + { + "epoch": 2.73, + "learning_rate": 2.1830542409901713e-06, + "loss": 0.3247, + "step": 15011 + }, + { + "epoch": 2.73, + "learning_rate": 2.1829996359665092e-06, + "loss": 0.6382, + "step": 15012 + }, + { + "epoch": 2.73, + "learning_rate": 2.1829450309428467e-06, + "loss": 0.6854, + "step": 15013 + }, + { + "epoch": 2.73, + "learning_rate": 2.1828904259191847e-06, + "loss": 0.3676, + "step": 15014 + }, + { + "epoch": 2.73, + "learning_rate": 2.1828358208955226e-06, + "loss": 0.3381, + "step": 15015 + }, + { + "epoch": 2.73, + "learning_rate": 2.18278121587186e-06, + "loss": 0.1621, + "step": 15016 + }, + { + "epoch": 2.73, + "learning_rate": 2.182726610848198e-06, + "loss": 0.1535, + "step": 15017 + }, + { + "epoch": 2.73, + "learning_rate": 2.182672005824536e-06, + "loss": 0.7681, + "step": 15018 + }, + { + "epoch": 2.73, + "learning_rate": 2.1826174008008736e-06, + "loss": 0.0681, + "step": 15019 + }, + { + "epoch": 2.73, + "learning_rate": 2.1825627957772115e-06, + "loss": 0.1974, + "step": 15020 + }, + { + "epoch": 2.73, + "learning_rate": 2.1825081907535494e-06, + "loss": 0.4956, + "step": 15021 + }, + { + "epoch": 2.73, + "learning_rate": 2.1824535857298874e-06, + "loss": 1.1161, + "step": 15022 + }, + { + "epoch": 2.73, + "learning_rate": 2.1823989807062253e-06, + "loss": 1.5473, + "step": 15023 + }, + { + "epoch": 2.73, + "learning_rate": 2.182344375682563e-06, + "loss": 0.4745, + "step": 15024 + }, + { + "epoch": 2.73, + "learning_rate": 2.182289770658901e-06, + "loss": 1.0681, + "step": 15025 + }, + { + "epoch": 2.73, + "learning_rate": 2.1822351656352387e-06, + "loss": 1.4693, + "step": 15026 + }, + { + "epoch": 2.73, + "learning_rate": 2.1821805606115763e-06, + "loss": 1.0382, + "step": 15027 + }, + { + "epoch": 2.73, + "learning_rate": 2.182125955587914e-06, + "loss": 0.0363, + "step": 15028 + }, + { + "epoch": 2.73, + "learning_rate": 2.182071350564252e-06, + "loss": 0.0262, + "step": 15029 + }, + { + "epoch": 2.73, + "learning_rate": 2.1820167455405897e-06, + "loss": 0.4088, + "step": 15030 + }, + { + "epoch": 2.73, + "learning_rate": 2.1819621405169276e-06, + "loss": 2.0681, + "step": 15031 + }, + { + "epoch": 2.73, + "learning_rate": 2.181907535493265e-06, + "loss": 0.6335, + "step": 15032 + }, + { + "epoch": 2.73, + "learning_rate": 2.181852930469603e-06, + "loss": 0.0199, + "step": 15033 + }, + { + "epoch": 2.73, + "learning_rate": 2.181798325445941e-06, + "loss": 2.5187, + "step": 15034 + }, + { + "epoch": 2.73, + "learning_rate": 2.181743720422279e-06, + "loss": 0.3476, + "step": 15035 + }, + { + "epoch": 2.73, + "learning_rate": 2.181689115398617e-06, + "loss": 0.6761, + "step": 15036 + }, + { + "epoch": 2.73, + "learning_rate": 2.181634510374955e-06, + "loss": 1.126, + "step": 15037 + }, + { + "epoch": 2.73, + "learning_rate": 2.1815799053512924e-06, + "loss": 0.8866, + "step": 15038 + }, + { + "epoch": 2.73, + "learning_rate": 2.1815253003276303e-06, + "loss": 0.8931, + "step": 15039 + }, + { + "epoch": 2.74, + "learning_rate": 2.1814706953039682e-06, + "loss": 1.1242, + "step": 15040 + }, + { + "epoch": 2.74, + "learning_rate": 2.1814160902803058e-06, + "loss": 0.3899, + "step": 15041 + }, + { + "epoch": 2.74, + "learning_rate": 2.1813614852566437e-06, + "loss": 1.5731, + "step": 15042 + }, + { + "epoch": 2.74, + "learning_rate": 2.1813068802329817e-06, + "loss": 1.4589, + "step": 15043 + }, + { + "epoch": 2.74, + "learning_rate": 2.181252275209319e-06, + "loss": 0.0399, + "step": 15044 + }, + { + "epoch": 2.74, + "learning_rate": 2.181197670185657e-06, + "loss": 0.327, + "step": 15045 + }, + { + "epoch": 2.74, + "learning_rate": 2.1811430651619946e-06, + "loss": 0.2455, + "step": 15046 + }, + { + "epoch": 2.74, + "learning_rate": 2.1810884601383326e-06, + "loss": 1.4371, + "step": 15047 + }, + { + "epoch": 2.74, + "learning_rate": 2.1810338551146705e-06, + "loss": 0.34, + "step": 15048 + }, + { + "epoch": 2.74, + "learning_rate": 2.1809792500910085e-06, + "loss": 2.1224, + "step": 15049 + }, + { + "epoch": 2.74, + "learning_rate": 2.1809246450673464e-06, + "loss": 0.7382, + "step": 15050 + }, + { + "epoch": 2.74, + "learning_rate": 2.1808700400436844e-06, + "loss": 0.5215, + "step": 15051 + }, + { + "epoch": 2.74, + "learning_rate": 2.180815435020022e-06, + "loss": 0.9076, + "step": 15052 + }, + { + "epoch": 2.74, + "learning_rate": 2.18076082999636e-06, + "loss": 1.8449, + "step": 15053 + }, + { + "epoch": 2.74, + "learning_rate": 2.1807062249726978e-06, + "loss": 0.6622, + "step": 15054 + }, + { + "epoch": 2.74, + "learning_rate": 2.1806516199490353e-06, + "loss": 0.2814, + "step": 15055 + }, + { + "epoch": 2.74, + "learning_rate": 2.1805970149253732e-06, + "loss": 0.9693, + "step": 15056 + }, + { + "epoch": 2.74, + "learning_rate": 2.180542409901711e-06, + "loss": 0.0623, + "step": 15057 + }, + { + "epoch": 2.74, + "learning_rate": 2.1804878048780487e-06, + "loss": 2.7196, + "step": 15058 + }, + { + "epoch": 2.74, + "learning_rate": 2.1804331998543866e-06, + "loss": 1.0847, + "step": 15059 + }, + { + "epoch": 2.74, + "learning_rate": 2.180378594830724e-06, + "loss": 2.0146, + "step": 15060 + }, + { + "epoch": 2.74, + "learning_rate": 2.180323989807062e-06, + "loss": 0.6437, + "step": 15061 + }, + { + "epoch": 2.74, + "learning_rate": 2.1802693847834e-06, + "loss": 1.3915, + "step": 15062 + }, + { + "epoch": 2.74, + "learning_rate": 2.180214779759738e-06, + "loss": 0.3046, + "step": 15063 + }, + { + "epoch": 2.74, + "learning_rate": 2.180160174736076e-06, + "loss": 0.2847, + "step": 15064 + }, + { + "epoch": 2.74, + "learning_rate": 2.180105569712414e-06, + "loss": 1.0418, + "step": 15065 + }, + { + "epoch": 2.74, + "learning_rate": 2.1800509646887514e-06, + "loss": 0.2032, + "step": 15066 + }, + { + "epoch": 2.74, + "learning_rate": 2.1799963596650893e-06, + "loss": 0.2959, + "step": 15067 + }, + { + "epoch": 2.74, + "learning_rate": 2.1799417546414273e-06, + "loss": 0.0339, + "step": 15068 + }, + { + "epoch": 2.74, + "learning_rate": 2.179887149617765e-06, + "loss": 2.1385, + "step": 15069 + }, + { + "epoch": 2.74, + "learning_rate": 2.1798325445941027e-06, + "loss": 0.9622, + "step": 15070 + }, + { + "epoch": 2.74, + "learning_rate": 2.1797779395704407e-06, + "loss": 1.0206, + "step": 15071 + }, + { + "epoch": 2.74, + "learning_rate": 2.179723334546778e-06, + "loss": 0.3123, + "step": 15072 + }, + { + "epoch": 2.74, + "learning_rate": 2.179668729523116e-06, + "loss": 0.4283, + "step": 15073 + }, + { + "epoch": 2.74, + "learning_rate": 2.1796141244994537e-06, + "loss": 0.7721, + "step": 15074 + }, + { + "epoch": 2.74, + "learning_rate": 2.1795595194757916e-06, + "loss": 2.3969, + "step": 15075 + }, + { + "epoch": 2.74, + "learning_rate": 2.1795049144521296e-06, + "loss": 0.419, + "step": 15076 + }, + { + "epoch": 2.74, + "learning_rate": 2.1794503094284675e-06, + "loss": 0.98, + "step": 15077 + }, + { + "epoch": 2.74, + "learning_rate": 2.1793957044048054e-06, + "loss": 0.667, + "step": 15078 + }, + { + "epoch": 2.74, + "learning_rate": 2.1793410993811434e-06, + "loss": 1.6102, + "step": 15079 + }, + { + "epoch": 2.74, + "learning_rate": 2.179286494357481e-06, + "loss": 0.3061, + "step": 15080 + }, + { + "epoch": 2.74, + "learning_rate": 2.179231889333819e-06, + "loss": 1.817, + "step": 15081 + }, + { + "epoch": 2.74, + "learning_rate": 2.1791772843101568e-06, + "loss": 0.7283, + "step": 15082 + }, + { + "epoch": 2.74, + "learning_rate": 2.1791226792864943e-06, + "loss": 0.3117, + "step": 15083 + }, + { + "epoch": 2.74, + "learning_rate": 2.1790680742628323e-06, + "loss": 2.4481, + "step": 15084 + }, + { + "epoch": 2.74, + "learning_rate": 2.17901346923917e-06, + "loss": 0.4551, + "step": 15085 + }, + { + "epoch": 2.74, + "learning_rate": 2.1789588642155077e-06, + "loss": 0.118, + "step": 15086 + }, + { + "epoch": 2.74, + "learning_rate": 2.1789042591918457e-06, + "loss": 1.3786, + "step": 15087 + }, + { + "epoch": 2.74, + "learning_rate": 2.178849654168183e-06, + "loss": 0.0924, + "step": 15088 + }, + { + "epoch": 2.74, + "learning_rate": 2.178795049144521e-06, + "loss": 1.0318, + "step": 15089 + }, + { + "epoch": 2.74, + "learning_rate": 2.178740444120859e-06, + "loss": 0.1405, + "step": 15090 + }, + { + "epoch": 2.74, + "learning_rate": 2.178685839097197e-06, + "loss": 0.4167, + "step": 15091 + }, + { + "epoch": 2.74, + "learning_rate": 2.178631234073535e-06, + "loss": 0.3283, + "step": 15092 + }, + { + "epoch": 2.74, + "learning_rate": 2.178576629049873e-06, + "loss": 2.6549, + "step": 15093 + }, + { + "epoch": 2.74, + "learning_rate": 2.1785220240262104e-06, + "loss": 1.835, + "step": 15094 + }, + { + "epoch": 2.75, + "learning_rate": 2.1784674190025484e-06, + "loss": 0.5823, + "step": 15095 + }, + { + "epoch": 2.75, + "learning_rate": 2.1784128139788863e-06, + "loss": 2.0091, + "step": 15096 + }, + { + "epoch": 2.75, + "learning_rate": 2.178358208955224e-06, + "loss": 0.9692, + "step": 15097 + }, + { + "epoch": 2.75, + "learning_rate": 2.1783036039315618e-06, + "loss": 0.1322, + "step": 15098 + }, + { + "epoch": 2.75, + "learning_rate": 2.1782489989078997e-06, + "loss": 2.2154, + "step": 15099 + }, + { + "epoch": 2.75, + "learning_rate": 2.1781943938842372e-06, + "loss": 1.0811, + "step": 15100 + }, + { + "epoch": 2.75, + "learning_rate": 2.178139788860575e-06, + "loss": 0.63, + "step": 15101 + }, + { + "epoch": 2.75, + "learning_rate": 2.178085183836913e-06, + "loss": 2.1106, + "step": 15102 + }, + { + "epoch": 2.75, + "learning_rate": 2.178030578813251e-06, + "loss": 1.6447, + "step": 15103 + }, + { + "epoch": 2.75, + "learning_rate": 2.177975973789589e-06, + "loss": 0.6941, + "step": 15104 + }, + { + "epoch": 2.75, + "learning_rate": 2.1779213687659265e-06, + "loss": 0.6937, + "step": 15105 + }, + { + "epoch": 2.75, + "learning_rate": 2.1778667637422645e-06, + "loss": 0.2872, + "step": 15106 + }, + { + "epoch": 2.75, + "learning_rate": 2.1778121587186024e-06, + "loss": 0.9922, + "step": 15107 + }, + { + "epoch": 2.75, + "learning_rate": 2.17775755369494e-06, + "loss": 0.4514, + "step": 15108 + }, + { + "epoch": 2.75, + "learning_rate": 2.177702948671278e-06, + "loss": 1.3629, + "step": 15109 + }, + { + "epoch": 2.75, + "learning_rate": 2.177648343647616e-06, + "loss": 0.0634, + "step": 15110 + }, + { + "epoch": 2.75, + "learning_rate": 2.1775937386239533e-06, + "loss": 0.6774, + "step": 15111 + }, + { + "epoch": 2.75, + "learning_rate": 2.1775391336002913e-06, + "loss": 0.6777, + "step": 15112 + }, + { + "epoch": 2.75, + "learning_rate": 2.177484528576629e-06, + "loss": 0.2769, + "step": 15113 + }, + { + "epoch": 2.75, + "learning_rate": 2.1774299235529667e-06, + "loss": 0.8459, + "step": 15114 + }, + { + "epoch": 2.75, + "learning_rate": 2.1773753185293047e-06, + "loss": 0.7693, + "step": 15115 + }, + { + "epoch": 2.75, + "learning_rate": 2.1773207135056426e-06, + "loss": 1.8905, + "step": 15116 + }, + { + "epoch": 2.75, + "learning_rate": 2.1772661084819806e-06, + "loss": 0.059, + "step": 15117 + }, + { + "epoch": 2.75, + "learning_rate": 2.1772115034583185e-06, + "loss": 0.7104, + "step": 15118 + }, + { + "epoch": 2.75, + "learning_rate": 2.177156898434656e-06, + "loss": 2.6654, + "step": 15119 + }, + { + "epoch": 2.75, + "learning_rate": 2.177102293410994e-06, + "loss": 0.3765, + "step": 15120 + }, + { + "epoch": 2.75, + "learning_rate": 2.177047688387332e-06, + "loss": 0.7044, + "step": 15121 + }, + { + "epoch": 2.75, + "learning_rate": 2.1769930833636694e-06, + "loss": 2.6923, + "step": 15122 + }, + { + "epoch": 2.75, + "learning_rate": 2.1769384783400074e-06, + "loss": 0.9838, + "step": 15123 + }, + { + "epoch": 2.75, + "learning_rate": 2.1768838733163453e-06, + "loss": 1.0701, + "step": 15124 + }, + { + "epoch": 2.75, + "learning_rate": 2.176829268292683e-06, + "loss": 0.779, + "step": 15125 + }, + { + "epoch": 2.75, + "learning_rate": 2.176774663269021e-06, + "loss": 0.7345, + "step": 15126 + }, + { + "epoch": 2.75, + "learning_rate": 2.1767200582453583e-06, + "loss": 0.0492, + "step": 15127 + }, + { + "epoch": 2.75, + "learning_rate": 2.1766654532216963e-06, + "loss": 2.7133, + "step": 15128 + }, + { + "epoch": 2.75, + "learning_rate": 2.176610848198034e-06, + "loss": 0.558, + "step": 15129 + }, + { + "epoch": 2.75, + "learning_rate": 2.176556243174372e-06, + "loss": 0.7873, + "step": 15130 + }, + { + "epoch": 2.75, + "learning_rate": 2.17650163815071e-06, + "loss": 1.6094, + "step": 15131 + }, + { + "epoch": 2.75, + "learning_rate": 2.176447033127048e-06, + "loss": 0.0798, + "step": 15132 + }, + { + "epoch": 2.75, + "learning_rate": 2.1763924281033855e-06, + "loss": 0.1457, + "step": 15133 + }, + { + "epoch": 2.75, + "learning_rate": 2.1763378230797235e-06, + "loss": 0.7311, + "step": 15134 + }, + { + "epoch": 2.75, + "learning_rate": 2.1762832180560614e-06, + "loss": 0.3232, + "step": 15135 + }, + { + "epoch": 2.75, + "learning_rate": 2.176228613032399e-06, + "loss": 0.0538, + "step": 15136 + }, + { + "epoch": 2.75, + "learning_rate": 2.176174008008737e-06, + "loss": 0.9438, + "step": 15137 + }, + { + "epoch": 2.75, + "learning_rate": 2.176119402985075e-06, + "loss": 0.5222, + "step": 15138 + }, + { + "epoch": 2.75, + "learning_rate": 2.1760647979614124e-06, + "loss": 0.0655, + "step": 15139 + }, + { + "epoch": 2.75, + "learning_rate": 2.1760101929377503e-06, + "loss": 0.3192, + "step": 15140 + }, + { + "epoch": 2.75, + "learning_rate": 2.175955587914088e-06, + "loss": 0.0131, + "step": 15141 + }, + { + "epoch": 2.75, + "learning_rate": 2.1759009828904258e-06, + "loss": 0.2883, + "step": 15142 + }, + { + "epoch": 2.75, + "learning_rate": 2.1758463778667637e-06, + "loss": 0.0935, + "step": 15143 + }, + { + "epoch": 2.75, + "learning_rate": 2.1757917728431017e-06, + "loss": 0.1057, + "step": 15144 + }, + { + "epoch": 2.75, + "learning_rate": 2.1757371678194396e-06, + "loss": 1.5504, + "step": 15145 + }, + { + "epoch": 2.75, + "learning_rate": 2.1756825627957775e-06, + "loss": 0.2793, + "step": 15146 + }, + { + "epoch": 2.75, + "learning_rate": 2.175627957772115e-06, + "loss": 0.1399, + "step": 15147 + }, + { + "epoch": 2.75, + "learning_rate": 2.175573352748453e-06, + "loss": 0.0321, + "step": 15148 + }, + { + "epoch": 2.75, + "learning_rate": 2.175518747724791e-06, + "loss": 2.6599, + "step": 15149 + }, + { + "epoch": 2.76, + "learning_rate": 2.1754641427011285e-06, + "loss": 1.3975, + "step": 15150 + }, + { + "epoch": 2.76, + "learning_rate": 2.1754095376774664e-06, + "loss": 1.1938, + "step": 15151 + }, + { + "epoch": 2.76, + "learning_rate": 2.1753549326538044e-06, + "loss": 3.0033, + "step": 15152 + }, + { + "epoch": 2.76, + "learning_rate": 2.175300327630142e-06, + "loss": 0.2276, + "step": 15153 + }, + { + "epoch": 2.76, + "learning_rate": 2.17524572260648e-06, + "loss": 0.1256, + "step": 15154 + }, + { + "epoch": 2.76, + "learning_rate": 2.1751911175828173e-06, + "loss": 1.2015, + "step": 15155 + }, + { + "epoch": 2.76, + "learning_rate": 2.1751365125591553e-06, + "loss": 0.5056, + "step": 15156 + }, + { + "epoch": 2.76, + "learning_rate": 2.1750819075354932e-06, + "loss": 0.5908, + "step": 15157 + }, + { + "epoch": 2.76, + "learning_rate": 2.175027302511831e-06, + "loss": 0.5878, + "step": 15158 + }, + { + "epoch": 2.76, + "learning_rate": 2.174972697488169e-06, + "loss": 0.3254, + "step": 15159 + }, + { + "epoch": 2.76, + "learning_rate": 2.174918092464507e-06, + "loss": 0.4906, + "step": 15160 + }, + { + "epoch": 2.76, + "learning_rate": 2.1748634874408446e-06, + "loss": 0.747, + "step": 15161 + }, + { + "epoch": 2.76, + "learning_rate": 2.1748088824171825e-06, + "loss": 0.0233, + "step": 15162 + }, + { + "epoch": 2.76, + "learning_rate": 2.1747542773935205e-06, + "loss": 0.1596, + "step": 15163 + }, + { + "epoch": 2.76, + "learning_rate": 2.174699672369858e-06, + "loss": 1.9452, + "step": 15164 + }, + { + "epoch": 2.76, + "learning_rate": 2.174645067346196e-06, + "loss": 1.1866, + "step": 15165 + }, + { + "epoch": 2.76, + "learning_rate": 2.174590462322534e-06, + "loss": 0.8152, + "step": 15166 + }, + { + "epoch": 2.76, + "learning_rate": 2.1745358572988714e-06, + "loss": 1.2683, + "step": 15167 + }, + { + "epoch": 2.76, + "learning_rate": 2.1744812522752093e-06, + "loss": 1.2115, + "step": 15168 + }, + { + "epoch": 2.76, + "learning_rate": 2.174426647251547e-06, + "loss": 0.1894, + "step": 15169 + }, + { + "epoch": 2.76, + "learning_rate": 2.174372042227885e-06, + "loss": 0.3788, + "step": 15170 + }, + { + "epoch": 2.76, + "learning_rate": 2.1743174372042227e-06, + "loss": 0.3408, + "step": 15171 + }, + { + "epoch": 2.76, + "learning_rate": 2.1742628321805607e-06, + "loss": 0.5966, + "step": 15172 + }, + { + "epoch": 2.76, + "learning_rate": 2.1742082271568986e-06, + "loss": 0.8573, + "step": 15173 + }, + { + "epoch": 2.76, + "learning_rate": 2.1741536221332366e-06, + "loss": 0.6447, + "step": 15174 + }, + { + "epoch": 2.76, + "learning_rate": 2.174099017109574e-06, + "loss": 0.9172, + "step": 15175 + }, + { + "epoch": 2.76, + "learning_rate": 2.174044412085912e-06, + "loss": 1.4423, + "step": 15176 + }, + { + "epoch": 2.76, + "learning_rate": 2.17398980706225e-06, + "loss": 0.0251, + "step": 15177 + }, + { + "epoch": 2.76, + "learning_rate": 2.1739352020385875e-06, + "loss": 0.2857, + "step": 15178 + }, + { + "epoch": 2.76, + "learning_rate": 2.1738805970149254e-06, + "loss": 0.9375, + "step": 15179 + }, + { + "epoch": 2.76, + "learning_rate": 2.1738259919912634e-06, + "loss": 0.0185, + "step": 15180 + }, + { + "epoch": 2.76, + "learning_rate": 2.173771386967601e-06, + "loss": 0.0809, + "step": 15181 + }, + { + "epoch": 2.76, + "learning_rate": 2.173716781943939e-06, + "loss": 1.0638, + "step": 15182 + }, + { + "epoch": 2.76, + "learning_rate": 2.173662176920277e-06, + "loss": 0.4617, + "step": 15183 + }, + { + "epoch": 2.76, + "learning_rate": 2.1736075718966147e-06, + "loss": 0.8099, + "step": 15184 + }, + { + "epoch": 2.76, + "learning_rate": 2.1735529668729523e-06, + "loss": 0.4144, + "step": 15185 + }, + { + "epoch": 2.76, + "learning_rate": 2.17349836184929e-06, + "loss": 1.3602, + "step": 15186 + }, + { + "epoch": 2.76, + "learning_rate": 2.173443756825628e-06, + "loss": 0.0134, + "step": 15187 + }, + { + "epoch": 2.76, + "learning_rate": 2.173389151801966e-06, + "loss": 0.2881, + "step": 15188 + }, + { + "epoch": 2.76, + "learning_rate": 2.1733345467783036e-06, + "loss": 0.2763, + "step": 15189 + }, + { + "epoch": 2.76, + "learning_rate": 2.1732799417546415e-06, + "loss": 0.5585, + "step": 15190 + }, + { + "epoch": 2.76, + "learning_rate": 2.1732253367309795e-06, + "loss": 0.1144, + "step": 15191 + }, + { + "epoch": 2.76, + "learning_rate": 2.173170731707317e-06, + "loss": 1.019, + "step": 15192 + }, + { + "epoch": 2.76, + "learning_rate": 2.173116126683655e-06, + "loss": 0.7103, + "step": 15193 + }, + { + "epoch": 2.76, + "learning_rate": 2.1730615216599925e-06, + "loss": 0.2666, + "step": 15194 + }, + { + "epoch": 2.76, + "learning_rate": 2.1730069166363304e-06, + "loss": 0.6052, + "step": 15195 + }, + { + "epoch": 2.76, + "learning_rate": 2.1729523116126684e-06, + "loss": 0.3726, + "step": 15196 + }, + { + "epoch": 2.76, + "learning_rate": 2.1728977065890063e-06, + "loss": 0.4294, + "step": 15197 + }, + { + "epoch": 2.76, + "learning_rate": 2.1728431015653442e-06, + "loss": 1.4815, + "step": 15198 + }, + { + "epoch": 2.76, + "learning_rate": 2.172788496541682e-06, + "loss": 0.0439, + "step": 15199 + }, + { + "epoch": 2.76, + "learning_rate": 2.1727338915180197e-06, + "loss": 0.9483, + "step": 15200 + }, + { + "epoch": 2.76, + "learning_rate": 2.1726792864943577e-06, + "loss": 0.0179, + "step": 15201 + }, + { + "epoch": 2.76, + "learning_rate": 2.1726246814706956e-06, + "loss": 1.9443, + "step": 15202 + }, + { + "epoch": 2.76, + "learning_rate": 2.172570076447033e-06, + "loss": 3.0785, + "step": 15203 + }, + { + "epoch": 2.76, + "learning_rate": 2.172515471423371e-06, + "loss": 0.6497, + "step": 15204 + }, + { + "epoch": 2.77, + "learning_rate": 2.172460866399709e-06, + "loss": 1.5081, + "step": 15205 + }, + { + "epoch": 2.77, + "learning_rate": 2.1724062613760465e-06, + "loss": 1.5544, + "step": 15206 + }, + { + "epoch": 2.77, + "learning_rate": 2.1723516563523845e-06, + "loss": 0.3147, + "step": 15207 + }, + { + "epoch": 2.77, + "learning_rate": 2.172297051328722e-06, + "loss": 0.3576, + "step": 15208 + }, + { + "epoch": 2.77, + "learning_rate": 2.17224244630506e-06, + "loss": 0.525, + "step": 15209 + }, + { + "epoch": 2.77, + "learning_rate": 2.172187841281398e-06, + "loss": 0.1203, + "step": 15210 + }, + { + "epoch": 2.77, + "learning_rate": 2.172133236257736e-06, + "loss": 1.6133, + "step": 15211 + }, + { + "epoch": 2.77, + "learning_rate": 2.1720786312340738e-06, + "loss": 0.5526, + "step": 15212 + }, + { + "epoch": 2.77, + "learning_rate": 2.1720240262104117e-06, + "loss": 0.2983, + "step": 15213 + }, + { + "epoch": 2.77, + "learning_rate": 2.1719694211867492e-06, + "loss": 0.0863, + "step": 15214 + }, + { + "epoch": 2.77, + "learning_rate": 2.171914816163087e-06, + "loss": 0.2485, + "step": 15215 + }, + { + "epoch": 2.77, + "learning_rate": 2.171860211139425e-06, + "loss": 0.4028, + "step": 15216 + }, + { + "epoch": 2.77, + "learning_rate": 2.1718056061157626e-06, + "loss": 0.4807, + "step": 15217 + }, + { + "epoch": 2.77, + "learning_rate": 2.1717510010921006e-06, + "loss": 2.0262, + "step": 15218 + }, + { + "epoch": 2.77, + "learning_rate": 2.1716963960684385e-06, + "loss": 0.4539, + "step": 15219 + }, + { + "epoch": 2.77, + "learning_rate": 2.171641791044776e-06, + "loss": 0.7108, + "step": 15220 + }, + { + "epoch": 2.77, + "learning_rate": 2.171587186021114e-06, + "loss": 1.0924, + "step": 15221 + }, + { + "epoch": 2.77, + "learning_rate": 2.1715325809974515e-06, + "loss": 0.8946, + "step": 15222 + }, + { + "epoch": 2.77, + "learning_rate": 2.1714779759737894e-06, + "loss": 2.4847, + "step": 15223 + }, + { + "epoch": 2.77, + "learning_rate": 2.1714233709501274e-06, + "loss": 0.1484, + "step": 15224 + }, + { + "epoch": 2.77, + "learning_rate": 2.1713687659264653e-06, + "loss": 3.7045, + "step": 15225 + }, + { + "epoch": 2.77, + "learning_rate": 2.1713141609028033e-06, + "loss": 0.1206, + "step": 15226 + }, + { + "epoch": 2.77, + "learning_rate": 2.1712595558791412e-06, + "loss": 1.8401, + "step": 15227 + }, + { + "epoch": 2.77, + "learning_rate": 2.1712049508554787e-06, + "loss": 0.6687, + "step": 15228 + }, + { + "epoch": 2.77, + "learning_rate": 2.1711503458318167e-06, + "loss": 0.8733, + "step": 15229 + }, + { + "epoch": 2.77, + "learning_rate": 2.1710957408081546e-06, + "loss": 0.1075, + "step": 15230 + }, + { + "epoch": 2.77, + "learning_rate": 2.171041135784492e-06, + "loss": 0.7109, + "step": 15231 + }, + { + "epoch": 2.77, + "learning_rate": 2.17098653076083e-06, + "loss": 0.3568, + "step": 15232 + }, + { + "epoch": 2.77, + "learning_rate": 2.170931925737168e-06, + "loss": 0.3922, + "step": 15233 + }, + { + "epoch": 2.77, + "learning_rate": 2.1708773207135055e-06, + "loss": 0.5447, + "step": 15234 + }, + { + "epoch": 2.77, + "learning_rate": 2.1708227156898435e-06, + "loss": 1.6318, + "step": 15235 + }, + { + "epoch": 2.77, + "learning_rate": 2.170768110666181e-06, + "loss": 0.4385, + "step": 15236 + }, + { + "epoch": 2.77, + "learning_rate": 2.170713505642519e-06, + "loss": 5.2434, + "step": 15237 + }, + { + "epoch": 2.77, + "learning_rate": 2.170658900618857e-06, + "loss": 0.541, + "step": 15238 + }, + { + "epoch": 2.77, + "learning_rate": 2.170604295595195e-06, + "loss": 0.4044, + "step": 15239 + }, + { + "epoch": 2.77, + "learning_rate": 2.1705496905715328e-06, + "loss": 0.3248, + "step": 15240 + }, + { + "epoch": 2.77, + "learning_rate": 2.1704950855478707e-06, + "loss": 0.1591, + "step": 15241 + }, + { + "epoch": 2.77, + "learning_rate": 2.1704404805242082e-06, + "loss": 0.1205, + "step": 15242 + }, + { + "epoch": 2.77, + "learning_rate": 2.170385875500546e-06, + "loss": 0.9123, + "step": 15243 + }, + { + "epoch": 2.77, + "learning_rate": 2.170331270476884e-06, + "loss": 2.5352, + "step": 15244 + }, + { + "epoch": 2.77, + "learning_rate": 2.1702766654532217e-06, + "loss": 0.1722, + "step": 15245 + }, + { + "epoch": 2.77, + "learning_rate": 2.1702220604295596e-06, + "loss": 0.0433, + "step": 15246 + }, + { + "epoch": 2.77, + "learning_rate": 2.1701674554058975e-06, + "loss": 1.4325, + "step": 15247 + }, + { + "epoch": 2.77, + "learning_rate": 2.170112850382235e-06, + "loss": 1.3646, + "step": 15248 + }, + { + "epoch": 2.77, + "learning_rate": 2.170058245358573e-06, + "loss": 2.9799, + "step": 15249 + }, + { + "epoch": 2.77, + "learning_rate": 2.1700036403349105e-06, + "loss": 0.8333, + "step": 15250 + }, + { + "epoch": 2.77, + "learning_rate": 2.1699490353112485e-06, + "loss": 0.7224, + "step": 15251 + }, + { + "epoch": 2.77, + "learning_rate": 2.1698944302875864e-06, + "loss": 0.3539, + "step": 15252 + }, + { + "epoch": 2.77, + "learning_rate": 2.1698398252639244e-06, + "loss": 1.1492, + "step": 15253 + }, + { + "epoch": 2.77, + "learning_rate": 2.1697852202402623e-06, + "loss": 2.1844, + "step": 15254 + }, + { + "epoch": 2.77, + "learning_rate": 2.1697306152166002e-06, + "loss": 0.6927, + "step": 15255 + }, + { + "epoch": 2.77, + "learning_rate": 2.1696760101929378e-06, + "loss": 4.8246, + "step": 15256 + }, + { + "epoch": 2.77, + "learning_rate": 2.1696214051692757e-06, + "loss": 0.2367, + "step": 15257 + }, + { + "epoch": 2.77, + "learning_rate": 2.1695668001456136e-06, + "loss": 0.258, + "step": 15258 + }, + { + "epoch": 2.77, + "learning_rate": 2.169512195121951e-06, + "loss": 1.7767, + "step": 15259 + }, + { + "epoch": 2.78, + "learning_rate": 2.169457590098289e-06, + "loss": 0.2386, + "step": 15260 + }, + { + "epoch": 2.78, + "learning_rate": 2.1694029850746266e-06, + "loss": 0.3036, + "step": 15261 + }, + { + "epoch": 2.78, + "learning_rate": 2.1693483800509646e-06, + "loss": 0.0408, + "step": 15262 + }, + { + "epoch": 2.78, + "learning_rate": 2.1692937750273025e-06, + "loss": 0.4297, + "step": 15263 + }, + { + "epoch": 2.78, + "learning_rate": 2.16923917000364e-06, + "loss": 0.0351, + "step": 15264 + }, + { + "epoch": 2.78, + "learning_rate": 2.169184564979978e-06, + "loss": 1.6287, + "step": 15265 + }, + { + "epoch": 2.78, + "learning_rate": 2.169129959956316e-06, + "loss": 2.1081, + "step": 15266 + }, + { + "epoch": 2.78, + "learning_rate": 2.169075354932654e-06, + "loss": 0.2401, + "step": 15267 + }, + { + "epoch": 2.78, + "learning_rate": 2.169020749908992e-06, + "loss": 0.1511, + "step": 15268 + }, + { + "epoch": 2.78, + "learning_rate": 2.1689661448853298e-06, + "loss": 3.3587, + "step": 15269 + }, + { + "epoch": 2.78, + "learning_rate": 2.1689115398616673e-06, + "loss": 0.2214, + "step": 15270 + }, + { + "epoch": 2.78, + "learning_rate": 2.1688569348380052e-06, + "loss": 2.0972, + "step": 15271 + }, + { + "epoch": 2.78, + "learning_rate": 2.168802329814343e-06, + "loss": 0.3588, + "step": 15272 + }, + { + "epoch": 2.78, + "learning_rate": 2.1687477247906807e-06, + "loss": 0.0483, + "step": 15273 + }, + { + "epoch": 2.78, + "learning_rate": 2.1686931197670186e-06, + "loss": 0.2098, + "step": 15274 + }, + { + "epoch": 2.78, + "learning_rate": 2.168638514743356e-06, + "loss": 1.0053, + "step": 15275 + }, + { + "epoch": 2.78, + "learning_rate": 2.168583909719694e-06, + "loss": 2.2544, + "step": 15276 + }, + { + "epoch": 2.78, + "learning_rate": 2.168529304696032e-06, + "loss": 2.5297, + "step": 15277 + }, + { + "epoch": 2.78, + "learning_rate": 2.16847469967237e-06, + "loss": 2.0978, + "step": 15278 + }, + { + "epoch": 2.78, + "learning_rate": 2.168420094648708e-06, + "loss": 0.5199, + "step": 15279 + }, + { + "epoch": 2.78, + "learning_rate": 2.168365489625046e-06, + "loss": 0.0596, + "step": 15280 + }, + { + "epoch": 2.78, + "learning_rate": 2.1683108846013834e-06, + "loss": 0.5469, + "step": 15281 + }, + { + "epoch": 2.78, + "learning_rate": 2.1682562795777213e-06, + "loss": 0.0175, + "step": 15282 + }, + { + "epoch": 2.78, + "learning_rate": 2.1682016745540593e-06, + "loss": 0.6192, + "step": 15283 + }, + { + "epoch": 2.78, + "learning_rate": 2.168147069530397e-06, + "loss": 1.2424, + "step": 15284 + }, + { + "epoch": 2.78, + "learning_rate": 2.1680924645067347e-06, + "loss": 0.3111, + "step": 15285 + }, + { + "epoch": 2.78, + "learning_rate": 2.1680378594830727e-06, + "loss": 2.0877, + "step": 15286 + }, + { + "epoch": 2.78, + "learning_rate": 2.16798325445941e-06, + "loss": 1.3836, + "step": 15287 + }, + { + "epoch": 2.78, + "learning_rate": 2.167928649435748e-06, + "loss": 1.2029, + "step": 15288 + }, + { + "epoch": 2.78, + "learning_rate": 2.1678740444120857e-06, + "loss": 0.7475, + "step": 15289 + }, + { + "epoch": 2.78, + "learning_rate": 2.1678194393884236e-06, + "loss": 0.6884, + "step": 15290 + }, + { + "epoch": 2.78, + "learning_rate": 2.1677648343647615e-06, + "loss": 0.0253, + "step": 15291 + }, + { + "epoch": 2.78, + "learning_rate": 2.1677102293410995e-06, + "loss": 0.4524, + "step": 15292 + }, + { + "epoch": 2.78, + "learning_rate": 2.1676556243174374e-06, + "loss": 0.5567, + "step": 15293 + }, + { + "epoch": 2.78, + "learning_rate": 2.1676010192937754e-06, + "loss": 1.3798, + "step": 15294 + }, + { + "epoch": 2.78, + "learning_rate": 2.167546414270113e-06, + "loss": 0.7747, + "step": 15295 + }, + { + "epoch": 2.78, + "learning_rate": 2.167491809246451e-06, + "loss": 0.919, + "step": 15296 + }, + { + "epoch": 2.78, + "learning_rate": 2.1674372042227888e-06, + "loss": 0.9938, + "step": 15297 + }, + { + "epoch": 2.78, + "learning_rate": 2.1673825991991263e-06, + "loss": 0.8238, + "step": 15298 + }, + { + "epoch": 2.78, + "learning_rate": 2.1673279941754642e-06, + "loss": 0.9425, + "step": 15299 + }, + { + "epoch": 2.78, + "learning_rate": 2.167273389151802e-06, + "loss": 3.5104, + "step": 15300 + }, + { + "epoch": 2.78, + "learning_rate": 2.1672187841281397e-06, + "loss": 0.2198, + "step": 15301 + }, + { + "epoch": 2.78, + "learning_rate": 2.1671641791044777e-06, + "loss": 2.2805, + "step": 15302 + }, + { + "epoch": 2.78, + "learning_rate": 2.167109574080815e-06, + "loss": 1.1149, + "step": 15303 + }, + { + "epoch": 2.78, + "learning_rate": 2.167054969057153e-06, + "loss": 0.878, + "step": 15304 + }, + { + "epoch": 2.78, + "learning_rate": 2.167000364033491e-06, + "loss": 1.7908, + "step": 15305 + }, + { + "epoch": 2.78, + "learning_rate": 2.166945759009829e-06, + "loss": 0.0483, + "step": 15306 + }, + { + "epoch": 2.78, + "learning_rate": 2.166891153986167e-06, + "loss": 0.2554, + "step": 15307 + }, + { + "epoch": 2.78, + "learning_rate": 2.166836548962505e-06, + "loss": 1.0181, + "step": 15308 + }, + { + "epoch": 2.78, + "learning_rate": 2.1667819439388424e-06, + "loss": 0.8417, + "step": 15309 + }, + { + "epoch": 2.78, + "learning_rate": 2.1667273389151804e-06, + "loss": 1.3671, + "step": 15310 + }, + { + "epoch": 2.78, + "learning_rate": 2.1666727338915183e-06, + "loss": 0.6843, + "step": 15311 + }, + { + "epoch": 2.78, + "learning_rate": 2.166618128867856e-06, + "loss": 0.6127, + "step": 15312 + }, + { + "epoch": 2.78, + "learning_rate": 2.1665635238441938e-06, + "loss": 1.0426, + "step": 15313 + }, + { + "epoch": 2.78, + "learning_rate": 2.1665089188205317e-06, + "loss": 0.595, + "step": 15314 + }, + { + "epoch": 2.79, + "learning_rate": 2.1664543137968692e-06, + "loss": 0.7825, + "step": 15315 + }, + { + "epoch": 2.79, + "learning_rate": 2.166399708773207e-06, + "loss": 0.2833, + "step": 15316 + }, + { + "epoch": 2.79, + "learning_rate": 2.1663451037495447e-06, + "loss": 1.7909, + "step": 15317 + }, + { + "epoch": 2.79, + "learning_rate": 2.1662904987258826e-06, + "loss": 0.5809, + "step": 15318 + }, + { + "epoch": 2.79, + "learning_rate": 2.1662358937022206e-06, + "loss": 0.4572, + "step": 15319 + }, + { + "epoch": 2.79, + "learning_rate": 2.1661812886785585e-06, + "loss": 0.0266, + "step": 15320 + }, + { + "epoch": 2.79, + "learning_rate": 2.1661266836548965e-06, + "loss": 1.7116, + "step": 15321 + }, + { + "epoch": 2.79, + "learning_rate": 2.1660720786312344e-06, + "loss": 0.3881, + "step": 15322 + }, + { + "epoch": 2.79, + "learning_rate": 2.166017473607572e-06, + "loss": 0.0428, + "step": 15323 + }, + { + "epoch": 2.79, + "learning_rate": 2.16596286858391e-06, + "loss": 2.4993, + "step": 15324 + }, + { + "epoch": 2.79, + "learning_rate": 2.165908263560248e-06, + "loss": 0.1403, + "step": 15325 + }, + { + "epoch": 2.79, + "learning_rate": 2.1658536585365853e-06, + "loss": 0.7256, + "step": 15326 + }, + { + "epoch": 2.79, + "learning_rate": 2.1657990535129233e-06, + "loss": 0.2872, + "step": 15327 + }, + { + "epoch": 2.79, + "learning_rate": 2.1657444484892612e-06, + "loss": 1.8217, + "step": 15328 + }, + { + "epoch": 2.79, + "learning_rate": 2.1656898434655987e-06, + "loss": 0.3269, + "step": 15329 + }, + { + "epoch": 2.79, + "learning_rate": 2.1656352384419367e-06, + "loss": 1.7053, + "step": 15330 + }, + { + "epoch": 2.79, + "learning_rate": 2.165580633418274e-06, + "loss": 1.4664, + "step": 15331 + }, + { + "epoch": 2.79, + "learning_rate": 2.165526028394612e-06, + "loss": 0.711, + "step": 15332 + }, + { + "epoch": 2.79, + "learning_rate": 2.16547142337095e-06, + "loss": 0.5721, + "step": 15333 + }, + { + "epoch": 2.79, + "learning_rate": 2.165416818347288e-06, + "loss": 0.8873, + "step": 15334 + }, + { + "epoch": 2.79, + "learning_rate": 2.165362213323626e-06, + "loss": 3.2437, + "step": 15335 + }, + { + "epoch": 2.79, + "learning_rate": 2.165307608299964e-06, + "loss": 0.5298, + "step": 15336 + }, + { + "epoch": 2.79, + "learning_rate": 2.1652530032763014e-06, + "loss": 0.916, + "step": 15337 + }, + { + "epoch": 2.79, + "learning_rate": 2.1651983982526394e-06, + "loss": 0.031, + "step": 15338 + }, + { + "epoch": 2.79, + "learning_rate": 2.1651437932289773e-06, + "loss": 2.3832, + "step": 15339 + }, + { + "epoch": 2.79, + "learning_rate": 2.165089188205315e-06, + "loss": 1.0545, + "step": 15340 + }, + { + "epoch": 2.79, + "learning_rate": 2.1650345831816528e-06, + "loss": 0.0555, + "step": 15341 + }, + { + "epoch": 2.79, + "learning_rate": 2.1649799781579903e-06, + "loss": 0.0188, + "step": 15342 + }, + { + "epoch": 2.79, + "learning_rate": 2.1649253731343282e-06, + "loss": 0.2214, + "step": 15343 + }, + { + "epoch": 2.79, + "learning_rate": 2.164870768110666e-06, + "loss": 0.3842, + "step": 15344 + }, + { + "epoch": 2.79, + "learning_rate": 2.1648161630870037e-06, + "loss": 0.5299, + "step": 15345 + }, + { + "epoch": 2.79, + "learning_rate": 2.1647615580633417e-06, + "loss": 1.0598, + "step": 15346 + }, + { + "epoch": 2.79, + "learning_rate": 2.1647069530396796e-06, + "loss": 2.0159, + "step": 15347 + }, + { + "epoch": 2.79, + "learning_rate": 2.1646523480160175e-06, + "loss": 0.7651, + "step": 15348 + }, + { + "epoch": 2.79, + "learning_rate": 2.1645977429923555e-06, + "loss": 1.2465, + "step": 15349 + }, + { + "epoch": 2.79, + "learning_rate": 2.1645431379686934e-06, + "loss": 1.1332, + "step": 15350 + }, + { + "epoch": 2.79, + "learning_rate": 2.164488532945031e-06, + "loss": 0.0394, + "step": 15351 + }, + { + "epoch": 2.79, + "learning_rate": 2.164433927921369e-06, + "loss": 0.2752, + "step": 15352 + }, + { + "epoch": 2.79, + "learning_rate": 2.164379322897707e-06, + "loss": 0.6941, + "step": 15353 + }, + { + "epoch": 2.79, + "learning_rate": 2.1643247178740444e-06, + "loss": 1.4302, + "step": 15354 + }, + { + "epoch": 2.79, + "learning_rate": 2.1642701128503823e-06, + "loss": 0.4181, + "step": 15355 + }, + { + "epoch": 2.79, + "learning_rate": 2.16421550782672e-06, + "loss": 1.7806, + "step": 15356 + }, + { + "epoch": 2.79, + "learning_rate": 2.1641609028030578e-06, + "loss": 0.5774, + "step": 15357 + }, + { + "epoch": 2.79, + "learning_rate": 2.1641062977793957e-06, + "loss": 1.0325, + "step": 15358 + }, + { + "epoch": 2.79, + "learning_rate": 2.1640516927557337e-06, + "loss": 0.1529, + "step": 15359 + }, + { + "epoch": 2.79, + "learning_rate": 2.1639970877320716e-06, + "loss": 0.0141, + "step": 15360 + }, + { + "epoch": 2.79, + "learning_rate": 2.1639424827084095e-06, + "loss": 0.4164, + "step": 15361 + }, + { + "epoch": 2.79, + "learning_rate": 2.163887877684747e-06, + "loss": 0.3696, + "step": 15362 + }, + { + "epoch": 2.79, + "learning_rate": 2.163833272661085e-06, + "loss": 0.7828, + "step": 15363 + }, + { + "epoch": 2.79, + "learning_rate": 2.163778667637423e-06, + "loss": 1.1975, + "step": 15364 + }, + { + "epoch": 2.79, + "learning_rate": 2.1637240626137605e-06, + "loss": 0.0345, + "step": 15365 + }, + { + "epoch": 2.79, + "learning_rate": 2.1636694575900984e-06, + "loss": 0.9238, + "step": 15366 + }, + { + "epoch": 2.79, + "learning_rate": 2.1636148525664364e-06, + "loss": 0.7481, + "step": 15367 + }, + { + "epoch": 2.79, + "learning_rate": 2.163560247542774e-06, + "loss": 2.1891, + "step": 15368 + }, + { + "epoch": 2.79, + "learning_rate": 2.163505642519112e-06, + "loss": 0.2434, + "step": 15369 + }, + { + "epoch": 2.8, + "learning_rate": 2.1634510374954493e-06, + "loss": 0.2257, + "step": 15370 + }, + { + "epoch": 2.8, + "learning_rate": 2.1633964324717873e-06, + "loss": 0.4499, + "step": 15371 + }, + { + "epoch": 2.8, + "learning_rate": 2.1633418274481252e-06, + "loss": 1.2024, + "step": 15372 + }, + { + "epoch": 2.8, + "learning_rate": 2.163287222424463e-06, + "loss": 0.5884, + "step": 15373 + }, + { + "epoch": 2.8, + "learning_rate": 2.163232617400801e-06, + "loss": 0.5184, + "step": 15374 + }, + { + "epoch": 2.8, + "learning_rate": 2.163178012377139e-06, + "loss": 0.406, + "step": 15375 + }, + { + "epoch": 2.8, + "learning_rate": 2.1631234073534766e-06, + "loss": 0.5613, + "step": 15376 + }, + { + "epoch": 2.8, + "learning_rate": 2.1630688023298145e-06, + "loss": 0.0363, + "step": 15377 + }, + { + "epoch": 2.8, + "learning_rate": 2.1630141973061525e-06, + "loss": 0.3449, + "step": 15378 + }, + { + "epoch": 2.8, + "learning_rate": 2.16295959228249e-06, + "loss": 0.3047, + "step": 15379 + }, + { + "epoch": 2.8, + "learning_rate": 2.162904987258828e-06, + "loss": 1.4519, + "step": 15380 + }, + { + "epoch": 2.8, + "learning_rate": 2.162850382235166e-06, + "loss": 1.7651, + "step": 15381 + }, + { + "epoch": 2.8, + "learning_rate": 2.1627957772115034e-06, + "loss": 0.6666, + "step": 15382 + }, + { + "epoch": 2.8, + "learning_rate": 2.1627411721878413e-06, + "loss": 1.3298, + "step": 15383 + }, + { + "epoch": 2.8, + "learning_rate": 2.162686567164179e-06, + "loss": 1.8405, + "step": 15384 + }, + { + "epoch": 2.8, + "learning_rate": 2.162631962140517e-06, + "loss": 0.2722, + "step": 15385 + }, + { + "epoch": 2.8, + "learning_rate": 2.1625773571168547e-06, + "loss": 0.4642, + "step": 15386 + }, + { + "epoch": 2.8, + "learning_rate": 2.1625227520931927e-06, + "loss": 0.0522, + "step": 15387 + }, + { + "epoch": 2.8, + "learning_rate": 2.1624681470695306e-06, + "loss": 1.413, + "step": 15388 + }, + { + "epoch": 2.8, + "learning_rate": 2.1624135420458686e-06, + "loss": 1.0834, + "step": 15389 + }, + { + "epoch": 2.8, + "learning_rate": 2.162358937022206e-06, + "loss": 0.4775, + "step": 15390 + }, + { + "epoch": 2.8, + "learning_rate": 2.162304331998544e-06, + "loss": 1.6393, + "step": 15391 + }, + { + "epoch": 2.8, + "learning_rate": 2.162249726974882e-06, + "loss": 0.0335, + "step": 15392 + }, + { + "epoch": 2.8, + "learning_rate": 2.1621951219512195e-06, + "loss": 0.3017, + "step": 15393 + }, + { + "epoch": 2.8, + "learning_rate": 2.1621405169275574e-06, + "loss": 1.3833, + "step": 15394 + }, + { + "epoch": 2.8, + "learning_rate": 2.1620859119038954e-06, + "loss": 0.02, + "step": 15395 + }, + { + "epoch": 2.8, + "learning_rate": 2.162031306880233e-06, + "loss": 1.3548, + "step": 15396 + }, + { + "epoch": 2.8, + "learning_rate": 2.161976701856571e-06, + "loss": 0.5188, + "step": 15397 + }, + { + "epoch": 2.8, + "learning_rate": 2.1619220968329084e-06, + "loss": 0.7655, + "step": 15398 + }, + { + "epoch": 2.8, + "learning_rate": 2.1618674918092463e-06, + "loss": 0.5069, + "step": 15399 + }, + { + "epoch": 2.8, + "learning_rate": 2.1618128867855842e-06, + "loss": 4.4015, + "step": 15400 + }, + { + "epoch": 2.8, + "learning_rate": 2.161758281761922e-06, + "loss": 2.1196, + "step": 15401 + }, + { + "epoch": 2.8, + "learning_rate": 2.16170367673826e-06, + "loss": 0.5216, + "step": 15402 + }, + { + "epoch": 2.8, + "learning_rate": 2.161649071714598e-06, + "loss": 0.6602, + "step": 15403 + }, + { + "epoch": 2.8, + "learning_rate": 2.1615944666909356e-06, + "loss": 0.6862, + "step": 15404 + }, + { + "epoch": 2.8, + "learning_rate": 2.1615398616672735e-06, + "loss": 0.7333, + "step": 15405 + }, + { + "epoch": 2.8, + "learning_rate": 2.1614852566436115e-06, + "loss": 0.4408, + "step": 15406 + }, + { + "epoch": 2.8, + "learning_rate": 2.161430651619949e-06, + "loss": 0.4192, + "step": 15407 + }, + { + "epoch": 2.8, + "learning_rate": 2.161376046596287e-06, + "loss": 0.2262, + "step": 15408 + }, + { + "epoch": 2.8, + "learning_rate": 2.161321441572625e-06, + "loss": 0.0651, + "step": 15409 + }, + { + "epoch": 2.8, + "learning_rate": 2.1612668365489624e-06, + "loss": 0.3584, + "step": 15410 + }, + { + "epoch": 2.8, + "learning_rate": 2.1612122315253004e-06, + "loss": 0.6435, + "step": 15411 + }, + { + "epoch": 2.8, + "learning_rate": 2.161157626501638e-06, + "loss": 1.3206, + "step": 15412 + }, + { + "epoch": 2.8, + "learning_rate": 2.161103021477976e-06, + "loss": 0.1217, + "step": 15413 + }, + { + "epoch": 2.8, + "learning_rate": 2.1610484164543138e-06, + "loss": 0.5663, + "step": 15414 + }, + { + "epoch": 2.8, + "learning_rate": 2.1609938114306517e-06, + "loss": 0.7601, + "step": 15415 + }, + { + "epoch": 2.8, + "learning_rate": 2.1609392064069896e-06, + "loss": 0.9864, + "step": 15416 + }, + { + "epoch": 2.8, + "learning_rate": 2.1608846013833276e-06, + "loss": 1.2652, + "step": 15417 + }, + { + "epoch": 2.8, + "learning_rate": 2.160829996359665e-06, + "loss": 0.0321, + "step": 15418 + }, + { + "epoch": 2.8, + "learning_rate": 2.160775391336003e-06, + "loss": 0.2887, + "step": 15419 + }, + { + "epoch": 2.8, + "learning_rate": 2.160720786312341e-06, + "loss": 1.327, + "step": 15420 + }, + { + "epoch": 2.8, + "learning_rate": 2.1606661812886785e-06, + "loss": 0.171, + "step": 15421 + }, + { + "epoch": 2.8, + "learning_rate": 2.1606115762650165e-06, + "loss": 0.9481, + "step": 15422 + }, + { + "epoch": 2.8, + "learning_rate": 2.160556971241354e-06, + "loss": 0.2194, + "step": 15423 + }, + { + "epoch": 2.8, + "learning_rate": 2.160502366217692e-06, + "loss": 0.7223, + "step": 15424 + }, + { + "epoch": 2.81, + "learning_rate": 2.16044776119403e-06, + "loss": 0.1619, + "step": 15425 + }, + { + "epoch": 2.81, + "learning_rate": 2.1603931561703674e-06, + "loss": 0.6999, + "step": 15426 + }, + { + "epoch": 2.81, + "learning_rate": 2.1603385511467053e-06, + "loss": 0.6483, + "step": 15427 + }, + { + "epoch": 2.81, + "learning_rate": 2.1602839461230433e-06, + "loss": 0.0083, + "step": 15428 + }, + { + "epoch": 2.81, + "learning_rate": 2.1602293410993812e-06, + "loss": 2.8315, + "step": 15429 + }, + { + "epoch": 2.81, + "learning_rate": 2.160174736075719e-06, + "loss": 0.0073, + "step": 15430 + }, + { + "epoch": 2.81, + "learning_rate": 2.160120131052057e-06, + "loss": 0.0373, + "step": 15431 + }, + { + "epoch": 2.81, + "learning_rate": 2.1600655260283946e-06, + "loss": 0.4244, + "step": 15432 + }, + { + "epoch": 2.81, + "learning_rate": 2.1600109210047326e-06, + "loss": 0.807, + "step": 15433 + }, + { + "epoch": 2.81, + "learning_rate": 2.1599563159810705e-06, + "loss": 1.7869, + "step": 15434 + }, + { + "epoch": 2.81, + "learning_rate": 2.159901710957408e-06, + "loss": 1.2938, + "step": 15435 + }, + { + "epoch": 2.81, + "learning_rate": 2.159847105933746e-06, + "loss": 0.3542, + "step": 15436 + }, + { + "epoch": 2.81, + "learning_rate": 2.1597925009100835e-06, + "loss": 0.3021, + "step": 15437 + }, + { + "epoch": 2.81, + "learning_rate": 2.1597378958864214e-06, + "loss": 0.147, + "step": 15438 + }, + { + "epoch": 2.81, + "learning_rate": 2.1596832908627594e-06, + "loss": 2.717, + "step": 15439 + }, + { + "epoch": 2.81, + "learning_rate": 2.1596286858390973e-06, + "loss": 0.7114, + "step": 15440 + }, + { + "epoch": 2.81, + "learning_rate": 2.1595740808154353e-06, + "loss": 0.4141, + "step": 15441 + }, + { + "epoch": 2.81, + "learning_rate": 2.159519475791773e-06, + "loss": 0.0232, + "step": 15442 + }, + { + "epoch": 2.81, + "learning_rate": 2.1594648707681107e-06, + "loss": 0.6309, + "step": 15443 + }, + { + "epoch": 2.81, + "learning_rate": 2.1594102657444487e-06, + "loss": 0.9117, + "step": 15444 + }, + { + "epoch": 2.81, + "learning_rate": 2.1593556607207866e-06, + "loss": 0.4928, + "step": 15445 + }, + { + "epoch": 2.81, + "learning_rate": 2.159301055697124e-06, + "loss": 1.0337, + "step": 15446 + }, + { + "epoch": 2.81, + "learning_rate": 2.159246450673462e-06, + "loss": 0.5609, + "step": 15447 + }, + { + "epoch": 2.81, + "learning_rate": 2.1591918456498e-06, + "loss": 0.285, + "step": 15448 + }, + { + "epoch": 2.81, + "learning_rate": 2.1591372406261375e-06, + "loss": 0.5528, + "step": 15449 + }, + { + "epoch": 2.81, + "learning_rate": 2.1590826356024755e-06, + "loss": 0.0155, + "step": 15450 + }, + { + "epoch": 2.81, + "learning_rate": 2.159028030578813e-06, + "loss": 0.0185, + "step": 15451 + }, + { + "epoch": 2.81, + "learning_rate": 2.158973425555151e-06, + "loss": 0.2134, + "step": 15452 + }, + { + "epoch": 2.81, + "learning_rate": 2.158918820531489e-06, + "loss": 0.1638, + "step": 15453 + }, + { + "epoch": 2.81, + "learning_rate": 2.158864215507827e-06, + "loss": 2.3041, + "step": 15454 + }, + { + "epoch": 2.81, + "learning_rate": 2.1588096104841648e-06, + "loss": 0.1813, + "step": 15455 + }, + { + "epoch": 2.81, + "learning_rate": 2.1587550054605027e-06, + "loss": 0.3712, + "step": 15456 + }, + { + "epoch": 2.81, + "learning_rate": 2.1587004004368402e-06, + "loss": 0.0211, + "step": 15457 + }, + { + "epoch": 2.81, + "learning_rate": 2.158645795413178e-06, + "loss": 0.569, + "step": 15458 + }, + { + "epoch": 2.81, + "learning_rate": 2.158591190389516e-06, + "loss": 0.704, + "step": 15459 + }, + { + "epoch": 2.81, + "learning_rate": 2.1585365853658537e-06, + "loss": 0.102, + "step": 15460 + }, + { + "epoch": 2.81, + "learning_rate": 2.1584819803421916e-06, + "loss": 0.1086, + "step": 15461 + }, + { + "epoch": 2.81, + "learning_rate": 2.1584273753185295e-06, + "loss": 1.9288, + "step": 15462 + }, + { + "epoch": 2.81, + "learning_rate": 2.158372770294867e-06, + "loss": 0.1165, + "step": 15463 + }, + { + "epoch": 2.81, + "learning_rate": 2.158318165271205e-06, + "loss": 0.1261, + "step": 15464 + }, + { + "epoch": 2.81, + "learning_rate": 2.1582635602475425e-06, + "loss": 0.3685, + "step": 15465 + }, + { + "epoch": 2.81, + "learning_rate": 2.1582089552238805e-06, + "loss": 2.7815, + "step": 15466 + }, + { + "epoch": 2.81, + "learning_rate": 2.1581543502002184e-06, + "loss": 0.6133, + "step": 15467 + }, + { + "epoch": 2.81, + "learning_rate": 2.1580997451765564e-06, + "loss": 0.0472, + "step": 15468 + }, + { + "epoch": 2.81, + "learning_rate": 2.1580451401528943e-06, + "loss": 1.0478, + "step": 15469 + }, + { + "epoch": 2.81, + "learning_rate": 2.1579905351292322e-06, + "loss": 1.2721, + "step": 15470 + }, + { + "epoch": 2.81, + "learning_rate": 2.1579359301055698e-06, + "loss": 0.7756, + "step": 15471 + }, + { + "epoch": 2.81, + "learning_rate": 2.1578813250819077e-06, + "loss": 1.4476, + "step": 15472 + }, + { + "epoch": 2.81, + "learning_rate": 2.1578267200582456e-06, + "loss": 0.4274, + "step": 15473 + }, + { + "epoch": 2.81, + "learning_rate": 2.157772115034583e-06, + "loss": 1.6524, + "step": 15474 + }, + { + "epoch": 2.81, + "learning_rate": 2.157717510010921e-06, + "loss": 1.5481, + "step": 15475 + }, + { + "epoch": 2.81, + "learning_rate": 2.157662904987259e-06, + "loss": 1.6311, + "step": 15476 + }, + { + "epoch": 2.81, + "learning_rate": 2.1576082999635966e-06, + "loss": 0.2041, + "step": 15477 + }, + { + "epoch": 2.81, + "learning_rate": 2.1575536949399345e-06, + "loss": 0.8924, + "step": 15478 + }, + { + "epoch": 2.81, + "learning_rate": 2.157499089916272e-06, + "loss": 2.9381, + "step": 15479 + }, + { + "epoch": 2.82, + "learning_rate": 2.15744448489261e-06, + "loss": 1.0269, + "step": 15480 + }, + { + "epoch": 2.82, + "learning_rate": 2.157389879868948e-06, + "loss": 0.3277, + "step": 15481 + }, + { + "epoch": 2.82, + "learning_rate": 2.157335274845286e-06, + "loss": 0.083, + "step": 15482 + }, + { + "epoch": 2.82, + "learning_rate": 2.157280669821624e-06, + "loss": 0.4273, + "step": 15483 + }, + { + "epoch": 2.82, + "learning_rate": 2.1572260647979618e-06, + "loss": 0.3055, + "step": 15484 + }, + { + "epoch": 2.82, + "learning_rate": 2.1571714597742993e-06, + "loss": 0.0262, + "step": 15485 + }, + { + "epoch": 2.82, + "learning_rate": 2.1571168547506372e-06, + "loss": 2.3126, + "step": 15486 + }, + { + "epoch": 2.82, + "learning_rate": 2.157062249726975e-06, + "loss": 0.6439, + "step": 15487 + }, + { + "epoch": 2.82, + "learning_rate": 2.1570076447033127e-06, + "loss": 0.022, + "step": 15488 + }, + { + "epoch": 2.82, + "learning_rate": 2.1569530396796506e-06, + "loss": 0.4583, + "step": 15489 + }, + { + "epoch": 2.82, + "learning_rate": 2.1568984346559886e-06, + "loss": 0.3595, + "step": 15490 + }, + { + "epoch": 2.82, + "learning_rate": 2.156843829632326e-06, + "loss": 1.788, + "step": 15491 + }, + { + "epoch": 2.82, + "learning_rate": 2.156789224608664e-06, + "loss": 0.8191, + "step": 15492 + }, + { + "epoch": 2.82, + "learning_rate": 2.1567346195850015e-06, + "loss": 1.6243, + "step": 15493 + }, + { + "epoch": 2.82, + "learning_rate": 2.1566800145613395e-06, + "loss": 0.1377, + "step": 15494 + }, + { + "epoch": 2.82, + "learning_rate": 2.1566254095376774e-06, + "loss": 0.2927, + "step": 15495 + }, + { + "epoch": 2.82, + "learning_rate": 2.1565708045140154e-06, + "loss": 1.3276, + "step": 15496 + }, + { + "epoch": 2.82, + "learning_rate": 2.1565161994903533e-06, + "loss": 3.5893, + "step": 15497 + }, + { + "epoch": 2.82, + "learning_rate": 2.1564615944666913e-06, + "loss": 0.021, + "step": 15498 + }, + { + "epoch": 2.82, + "learning_rate": 2.1564069894430288e-06, + "loss": 2.2154, + "step": 15499 + }, + { + "epoch": 2.82, + "learning_rate": 2.1563523844193667e-06, + "loss": 1.2302, + "step": 15500 + }, + { + "epoch": 2.82, + "learning_rate": 2.1562977793957047e-06, + "loss": 2.4107, + "step": 15501 + }, + { + "epoch": 2.82, + "learning_rate": 2.156243174372042e-06, + "loss": 0.5782, + "step": 15502 + }, + { + "epoch": 2.82, + "learning_rate": 2.15618856934838e-06, + "loss": 0.0243, + "step": 15503 + }, + { + "epoch": 2.82, + "learning_rate": 2.1561339643247177e-06, + "loss": 1.4142, + "step": 15504 + }, + { + "epoch": 2.82, + "learning_rate": 2.1560793593010556e-06, + "loss": 1.951, + "step": 15505 + }, + { + "epoch": 2.82, + "learning_rate": 2.1560247542773935e-06, + "loss": 1.0683, + "step": 15506 + }, + { + "epoch": 2.82, + "learning_rate": 2.155970149253731e-06, + "loss": 0.5532, + "step": 15507 + }, + { + "epoch": 2.82, + "learning_rate": 2.155915544230069e-06, + "loss": 0.3951, + "step": 15508 + }, + { + "epoch": 2.82, + "learning_rate": 2.155860939206407e-06, + "loss": 0.1731, + "step": 15509 + }, + { + "epoch": 2.82, + "learning_rate": 2.155806334182745e-06, + "loss": 1.5023, + "step": 15510 + }, + { + "epoch": 2.82, + "learning_rate": 2.155751729159083e-06, + "loss": 0.3873, + "step": 15511 + }, + { + "epoch": 2.82, + "learning_rate": 2.1556971241354208e-06, + "loss": 3.4019, + "step": 15512 + }, + { + "epoch": 2.82, + "learning_rate": 2.1556425191117583e-06, + "loss": 2.9355, + "step": 15513 + }, + { + "epoch": 2.82, + "learning_rate": 2.1555879140880962e-06, + "loss": 0.362, + "step": 15514 + }, + { + "epoch": 2.82, + "learning_rate": 2.155533309064434e-06, + "loss": 0.4961, + "step": 15515 + }, + { + "epoch": 2.82, + "learning_rate": 2.1554787040407717e-06, + "loss": 0.2319, + "step": 15516 + }, + { + "epoch": 2.82, + "learning_rate": 2.1554240990171096e-06, + "loss": 0.5221, + "step": 15517 + }, + { + "epoch": 2.82, + "learning_rate": 2.155369493993447e-06, + "loss": 2.4108, + "step": 15518 + }, + { + "epoch": 2.82, + "learning_rate": 2.155314888969785e-06, + "loss": 0.095, + "step": 15519 + }, + { + "epoch": 2.82, + "learning_rate": 2.155260283946123e-06, + "loss": 0.1821, + "step": 15520 + }, + { + "epoch": 2.82, + "learning_rate": 2.155205678922461e-06, + "loss": 0.0141, + "step": 15521 + }, + { + "epoch": 2.82, + "learning_rate": 2.1551510738987985e-06, + "loss": 1.1606, + "step": 15522 + }, + { + "epoch": 2.82, + "learning_rate": 2.1550964688751365e-06, + "loss": 2.0631, + "step": 15523 + }, + { + "epoch": 2.82, + "learning_rate": 2.1550418638514744e-06, + "loss": 1.53, + "step": 15524 + }, + { + "epoch": 2.82, + "learning_rate": 2.1549872588278123e-06, + "loss": 0.0858, + "step": 15525 + }, + { + "epoch": 2.82, + "learning_rate": 2.1549326538041503e-06, + "loss": 0.1305, + "step": 15526 + }, + { + "epoch": 2.82, + "learning_rate": 2.154878048780488e-06, + "loss": 0.412, + "step": 15527 + }, + { + "epoch": 2.82, + "learning_rate": 2.1548234437568258e-06, + "loss": 0.1332, + "step": 15528 + }, + { + "epoch": 2.82, + "learning_rate": 2.1547688387331637e-06, + "loss": 0.52, + "step": 15529 + }, + { + "epoch": 2.82, + "learning_rate": 2.1547142337095012e-06, + "loss": 0.0856, + "step": 15530 + }, + { + "epoch": 2.82, + "learning_rate": 2.154659628685839e-06, + "loss": 0.8957, + "step": 15531 + }, + { + "epoch": 2.82, + "learning_rate": 2.1546050236621767e-06, + "loss": 0.1743, + "step": 15532 + }, + { + "epoch": 2.82, + "learning_rate": 2.1545504186385146e-06, + "loss": 0.8585, + "step": 15533 + }, + { + "epoch": 2.82, + "learning_rate": 2.1544958136148526e-06, + "loss": 0.3955, + "step": 15534 + }, + { + "epoch": 2.83, + "learning_rate": 2.1544412085911905e-06, + "loss": 0.7944, + "step": 15535 + }, + { + "epoch": 2.83, + "learning_rate": 2.1543866035675285e-06, + "loss": 0.3065, + "step": 15536 + }, + { + "epoch": 2.83, + "learning_rate": 2.1543319985438664e-06, + "loss": 0.6172, + "step": 15537 + }, + { + "epoch": 2.83, + "learning_rate": 2.154277393520204e-06, + "loss": 1.2783, + "step": 15538 + }, + { + "epoch": 2.83, + "learning_rate": 2.154222788496542e-06, + "loss": 0.6262, + "step": 15539 + }, + { + "epoch": 2.83, + "learning_rate": 2.15416818347288e-06, + "loss": 1.8064, + "step": 15540 + }, + { + "epoch": 2.83, + "learning_rate": 2.1541135784492173e-06, + "loss": 0.4366, + "step": 15541 + }, + { + "epoch": 2.83, + "learning_rate": 2.1540589734255553e-06, + "loss": 1.1285, + "step": 15542 + }, + { + "epoch": 2.83, + "learning_rate": 2.1540043684018932e-06, + "loss": 0.0471, + "step": 15543 + }, + { + "epoch": 2.83, + "learning_rate": 2.1539497633782307e-06, + "loss": 1.1484, + "step": 15544 + }, + { + "epoch": 2.83, + "learning_rate": 2.1538951583545687e-06, + "loss": 0.2538, + "step": 15545 + }, + { + "epoch": 2.83, + "learning_rate": 2.153840553330906e-06, + "loss": 1.5907, + "step": 15546 + }, + { + "epoch": 2.83, + "learning_rate": 2.153785948307244e-06, + "loss": 0.0444, + "step": 15547 + }, + { + "epoch": 2.83, + "learning_rate": 2.153731343283582e-06, + "loss": 0.8839, + "step": 15548 + }, + { + "epoch": 2.83, + "learning_rate": 2.15367673825992e-06, + "loss": 0.5941, + "step": 15549 + }, + { + "epoch": 2.83, + "learning_rate": 2.153622133236258e-06, + "loss": 0.786, + "step": 15550 + }, + { + "epoch": 2.83, + "learning_rate": 2.153567528212596e-06, + "loss": 0.6895, + "step": 15551 + }, + { + "epoch": 2.83, + "learning_rate": 2.1535129231889334e-06, + "loss": 0.5878, + "step": 15552 + }, + { + "epoch": 2.83, + "learning_rate": 2.1534583181652714e-06, + "loss": 0.2887, + "step": 15553 + }, + { + "epoch": 2.83, + "learning_rate": 2.1534037131416093e-06, + "loss": 0.3417, + "step": 15554 + }, + { + "epoch": 2.83, + "learning_rate": 2.153349108117947e-06, + "loss": 2.6932, + "step": 15555 + }, + { + "epoch": 2.83, + "learning_rate": 2.1532945030942848e-06, + "loss": 0.6672, + "step": 15556 + }, + { + "epoch": 2.83, + "learning_rate": 2.1532398980706227e-06, + "loss": 1.3256, + "step": 15557 + }, + { + "epoch": 2.83, + "learning_rate": 2.1531852930469602e-06, + "loss": 0.1762, + "step": 15558 + }, + { + "epoch": 2.83, + "learning_rate": 2.153130688023298e-06, + "loss": 0.3714, + "step": 15559 + }, + { + "epoch": 2.83, + "learning_rate": 2.1530760829996357e-06, + "loss": 1.8525, + "step": 15560 + }, + { + "epoch": 2.83, + "learning_rate": 2.1530214779759737e-06, + "loss": 0.4427, + "step": 15561 + }, + { + "epoch": 2.83, + "learning_rate": 2.1529668729523116e-06, + "loss": 0.0475, + "step": 15562 + }, + { + "epoch": 2.83, + "learning_rate": 2.1529122679286495e-06, + "loss": 0.6996, + "step": 15563 + }, + { + "epoch": 2.83, + "learning_rate": 2.1528576629049875e-06, + "loss": 0.1243, + "step": 15564 + }, + { + "epoch": 2.83, + "learning_rate": 2.1528030578813254e-06, + "loss": 0.7831, + "step": 15565 + }, + { + "epoch": 2.83, + "learning_rate": 2.152748452857663e-06, + "loss": 0.6981, + "step": 15566 + }, + { + "epoch": 2.83, + "learning_rate": 2.152693847834001e-06, + "loss": 0.6519, + "step": 15567 + }, + { + "epoch": 2.83, + "learning_rate": 2.152639242810339e-06, + "loss": 0.2188, + "step": 15568 + }, + { + "epoch": 2.83, + "learning_rate": 2.1525846377866764e-06, + "loss": 0.914, + "step": 15569 + }, + { + "epoch": 2.83, + "learning_rate": 2.1525300327630143e-06, + "loss": 1.0933, + "step": 15570 + }, + { + "epoch": 2.83, + "learning_rate": 2.152475427739352e-06, + "loss": 0.7256, + "step": 15571 + }, + { + "epoch": 2.83, + "learning_rate": 2.1524208227156898e-06, + "loss": 2.0767, + "step": 15572 + }, + { + "epoch": 2.83, + "learning_rate": 2.1523662176920277e-06, + "loss": 0.0857, + "step": 15573 + }, + { + "epoch": 2.83, + "learning_rate": 2.1523116126683652e-06, + "loss": 3.934, + "step": 15574 + }, + { + "epoch": 2.83, + "learning_rate": 2.152257007644703e-06, + "loss": 0.503, + "step": 15575 + }, + { + "epoch": 2.83, + "learning_rate": 2.152202402621041e-06, + "loss": 0.016, + "step": 15576 + }, + { + "epoch": 2.83, + "learning_rate": 2.152147797597379e-06, + "loss": 0.4986, + "step": 15577 + }, + { + "epoch": 2.83, + "learning_rate": 2.152093192573717e-06, + "loss": 2.5333, + "step": 15578 + }, + { + "epoch": 2.83, + "learning_rate": 2.152038587550055e-06, + "loss": 0.8638, + "step": 15579 + }, + { + "epoch": 2.83, + "learning_rate": 2.1519839825263925e-06, + "loss": 0.3947, + "step": 15580 + }, + { + "epoch": 2.83, + "learning_rate": 2.1519293775027304e-06, + "loss": 0.1776, + "step": 15581 + }, + { + "epoch": 2.83, + "learning_rate": 2.1518747724790683e-06, + "loss": 0.2598, + "step": 15582 + }, + { + "epoch": 2.83, + "learning_rate": 2.151820167455406e-06, + "loss": 0.1464, + "step": 15583 + }, + { + "epoch": 2.83, + "learning_rate": 2.151765562431744e-06, + "loss": 1.6924, + "step": 15584 + }, + { + "epoch": 2.83, + "learning_rate": 2.1517109574080813e-06, + "loss": 0.3433, + "step": 15585 + }, + { + "epoch": 2.83, + "learning_rate": 2.1516563523844193e-06, + "loss": 0.0591, + "step": 15586 + }, + { + "epoch": 2.83, + "learning_rate": 2.1516017473607572e-06, + "loss": 2.6493, + "step": 15587 + }, + { + "epoch": 2.83, + "learning_rate": 2.1515471423370947e-06, + "loss": 0.5246, + "step": 15588 + }, + { + "epoch": 2.83, + "learning_rate": 2.1514925373134327e-06, + "loss": 0.0296, + "step": 15589 + }, + { + "epoch": 2.84, + "learning_rate": 2.1514379322897706e-06, + "loss": 1.3506, + "step": 15590 + }, + { + "epoch": 2.84, + "learning_rate": 2.1513833272661086e-06, + "loss": 0.4968, + "step": 15591 + }, + { + "epoch": 2.84, + "learning_rate": 2.1513287222424465e-06, + "loss": 0.1341, + "step": 15592 + }, + { + "epoch": 2.84, + "learning_rate": 2.1512741172187845e-06, + "loss": 0.5051, + "step": 15593 + }, + { + "epoch": 2.84, + "learning_rate": 2.151219512195122e-06, + "loss": 1.0628, + "step": 15594 + }, + { + "epoch": 2.84, + "learning_rate": 2.15116490717146e-06, + "loss": 0.9356, + "step": 15595 + }, + { + "epoch": 2.84, + "learning_rate": 2.151110302147798e-06, + "loss": 3.6674, + "step": 15596 + }, + { + "epoch": 2.84, + "learning_rate": 2.1510556971241354e-06, + "loss": 0.3075, + "step": 15597 + }, + { + "epoch": 2.84, + "learning_rate": 2.1510010921004733e-06, + "loss": 0.9732, + "step": 15598 + }, + { + "epoch": 2.84, + "learning_rate": 2.150946487076811e-06, + "loss": 0.3882, + "step": 15599 + }, + { + "epoch": 2.84, + "learning_rate": 2.1508918820531488e-06, + "loss": 0.4717, + "step": 15600 + }, + { + "epoch": 2.84, + "learning_rate": 2.1508372770294867e-06, + "loss": 0.0651, + "step": 15601 + }, + { + "epoch": 2.84, + "learning_rate": 2.1507826720058242e-06, + "loss": 0.0261, + "step": 15602 + }, + { + "epoch": 2.84, + "learning_rate": 2.150728066982162e-06, + "loss": 0.5215, + "step": 15603 + }, + { + "epoch": 2.84, + "learning_rate": 2.1506734619585e-06, + "loss": 1.8635, + "step": 15604 + }, + { + "epoch": 2.84, + "learning_rate": 2.150618856934838e-06, + "loss": 0.6662, + "step": 15605 + }, + { + "epoch": 2.84, + "learning_rate": 2.150564251911176e-06, + "loss": 0.0175, + "step": 15606 + }, + { + "epoch": 2.84, + "learning_rate": 2.150509646887514e-06, + "loss": 0.3989, + "step": 15607 + }, + { + "epoch": 2.84, + "learning_rate": 2.1504550418638515e-06, + "loss": 0.7324, + "step": 15608 + }, + { + "epoch": 2.84, + "learning_rate": 2.1504004368401894e-06, + "loss": 0.9375, + "step": 15609 + }, + { + "epoch": 2.84, + "learning_rate": 2.1503458318165274e-06, + "loss": 1.937, + "step": 15610 + }, + { + "epoch": 2.84, + "learning_rate": 2.150291226792865e-06, + "loss": 0.8862, + "step": 15611 + }, + { + "epoch": 2.84, + "learning_rate": 2.150236621769203e-06, + "loss": 1.7827, + "step": 15612 + }, + { + "epoch": 2.84, + "learning_rate": 2.1501820167455404e-06, + "loss": 2.2586, + "step": 15613 + }, + { + "epoch": 2.84, + "learning_rate": 2.1501274117218783e-06, + "loss": 1.5737, + "step": 15614 + }, + { + "epoch": 2.84, + "learning_rate": 2.1500728066982162e-06, + "loss": 0.5408, + "step": 15615 + }, + { + "epoch": 2.84, + "learning_rate": 2.150018201674554e-06, + "loss": 1.5802, + "step": 15616 + }, + { + "epoch": 2.84, + "learning_rate": 2.149963596650892e-06, + "loss": 0.2563, + "step": 15617 + }, + { + "epoch": 2.84, + "learning_rate": 2.14990899162723e-06, + "loss": 5.3912, + "step": 15618 + }, + { + "epoch": 2.84, + "learning_rate": 2.1498543866035676e-06, + "loss": 0.3623, + "step": 15619 + }, + { + "epoch": 2.84, + "learning_rate": 2.1497997815799055e-06, + "loss": 0.0146, + "step": 15620 + }, + { + "epoch": 2.84, + "learning_rate": 2.1497451765562435e-06, + "loss": 0.2375, + "step": 15621 + }, + { + "epoch": 2.84, + "learning_rate": 2.149690571532581e-06, + "loss": 1.2319, + "step": 15622 + }, + { + "epoch": 2.84, + "learning_rate": 2.149635966508919e-06, + "loss": 2.3729, + "step": 15623 + }, + { + "epoch": 2.84, + "learning_rate": 2.149581361485257e-06, + "loss": 0.0248, + "step": 15624 + }, + { + "epoch": 2.84, + "learning_rate": 2.1495267564615944e-06, + "loss": 0.7837, + "step": 15625 + }, + { + "epoch": 2.84, + "learning_rate": 2.1494721514379323e-06, + "loss": 0.5693, + "step": 15626 + }, + { + "epoch": 2.84, + "learning_rate": 2.14941754641427e-06, + "loss": 0.0207, + "step": 15627 + }, + { + "epoch": 2.84, + "learning_rate": 2.149362941390608e-06, + "loss": 1.2997, + "step": 15628 + }, + { + "epoch": 2.84, + "learning_rate": 2.1493083363669458e-06, + "loss": 2.3622, + "step": 15629 + }, + { + "epoch": 2.84, + "learning_rate": 2.1492537313432837e-06, + "loss": 0.6976, + "step": 15630 + }, + { + "epoch": 2.84, + "learning_rate": 2.1491991263196216e-06, + "loss": 0.8382, + "step": 15631 + }, + { + "epoch": 2.84, + "learning_rate": 2.1491445212959596e-06, + "loss": 0.0177, + "step": 15632 + }, + { + "epoch": 2.84, + "learning_rate": 2.149089916272297e-06, + "loss": 2.7729, + "step": 15633 + }, + { + "epoch": 2.84, + "learning_rate": 2.149035311248635e-06, + "loss": 1.4618, + "step": 15634 + }, + { + "epoch": 2.84, + "learning_rate": 2.148980706224973e-06, + "loss": 0.1353, + "step": 15635 + }, + { + "epoch": 2.84, + "learning_rate": 2.1489261012013105e-06, + "loss": 0.549, + "step": 15636 + }, + { + "epoch": 2.84, + "learning_rate": 2.1488714961776485e-06, + "loss": 1.6153, + "step": 15637 + }, + { + "epoch": 2.84, + "learning_rate": 2.1488168911539864e-06, + "loss": 1.0021, + "step": 15638 + }, + { + "epoch": 2.84, + "learning_rate": 2.148762286130324e-06, + "loss": 1.8244, + "step": 15639 + }, + { + "epoch": 2.84, + "learning_rate": 2.148707681106662e-06, + "loss": 1.1414, + "step": 15640 + }, + { + "epoch": 2.84, + "learning_rate": 2.1486530760829994e-06, + "loss": 0.1563, + "step": 15641 + }, + { + "epoch": 2.84, + "learning_rate": 2.1485984710593373e-06, + "loss": 1.39, + "step": 15642 + }, + { + "epoch": 2.84, + "learning_rate": 2.1485438660356753e-06, + "loss": 0.5037, + "step": 15643 + }, + { + "epoch": 2.84, + "learning_rate": 2.1484892610120132e-06, + "loss": 0.4951, + "step": 15644 + }, + { + "epoch": 2.85, + "learning_rate": 2.148434655988351e-06, + "loss": 0.2249, + "step": 15645 + }, + { + "epoch": 2.85, + "learning_rate": 2.148380050964689e-06, + "loss": 0.423, + "step": 15646 + }, + { + "epoch": 2.85, + "learning_rate": 2.1483254459410266e-06, + "loss": 0.1091, + "step": 15647 + }, + { + "epoch": 2.85, + "learning_rate": 2.1482708409173646e-06, + "loss": 0.4134, + "step": 15648 + }, + { + "epoch": 2.85, + "learning_rate": 2.1482162358937025e-06, + "loss": 1.1282, + "step": 15649 + }, + { + "epoch": 2.85, + "learning_rate": 2.14816163087004e-06, + "loss": 0.8891, + "step": 15650 + }, + { + "epoch": 2.85, + "learning_rate": 2.148107025846378e-06, + "loss": 0.3053, + "step": 15651 + }, + { + "epoch": 2.85, + "learning_rate": 2.1480524208227155e-06, + "loss": 2.1151, + "step": 15652 + }, + { + "epoch": 2.85, + "learning_rate": 2.1479978157990534e-06, + "loss": 0.5678, + "step": 15653 + }, + { + "epoch": 2.85, + "learning_rate": 2.1479432107753914e-06, + "loss": 0.0919, + "step": 15654 + }, + { + "epoch": 2.85, + "learning_rate": 2.147888605751729e-06, + "loss": 1.3803, + "step": 15655 + }, + { + "epoch": 2.85, + "learning_rate": 2.147834000728067e-06, + "loss": 0.2815, + "step": 15656 + }, + { + "epoch": 2.85, + "learning_rate": 2.1477793957044048e-06, + "loss": 0.8759, + "step": 15657 + }, + { + "epoch": 2.85, + "learning_rate": 2.1477247906807427e-06, + "loss": 1.1959, + "step": 15658 + }, + { + "epoch": 2.85, + "learning_rate": 2.1476701856570807e-06, + "loss": 0.4742, + "step": 15659 + }, + { + "epoch": 2.85, + "learning_rate": 2.1476155806334186e-06, + "loss": 1.3434, + "step": 15660 + }, + { + "epoch": 2.85, + "learning_rate": 2.147560975609756e-06, + "loss": 0.8427, + "step": 15661 + }, + { + "epoch": 2.85, + "learning_rate": 2.147506370586094e-06, + "loss": 0.5484, + "step": 15662 + }, + { + "epoch": 2.85, + "learning_rate": 2.147451765562432e-06, + "loss": 0.5583, + "step": 15663 + }, + { + "epoch": 2.85, + "learning_rate": 2.1473971605387695e-06, + "loss": 1.8205, + "step": 15664 + }, + { + "epoch": 2.85, + "learning_rate": 2.1473425555151075e-06, + "loss": 0.1314, + "step": 15665 + }, + { + "epoch": 2.85, + "learning_rate": 2.147287950491445e-06, + "loss": 1.2579, + "step": 15666 + }, + { + "epoch": 2.85, + "learning_rate": 2.147233345467783e-06, + "loss": 0.2797, + "step": 15667 + }, + { + "epoch": 2.85, + "learning_rate": 2.147178740444121e-06, + "loss": 2.0863, + "step": 15668 + }, + { + "epoch": 2.85, + "learning_rate": 2.1471241354204584e-06, + "loss": 0.5064, + "step": 15669 + }, + { + "epoch": 2.85, + "learning_rate": 2.1470695303967964e-06, + "loss": 3.2665, + "step": 15670 + }, + { + "epoch": 2.85, + "learning_rate": 2.1470149253731343e-06, + "loss": 0.1719, + "step": 15671 + }, + { + "epoch": 2.85, + "learning_rate": 2.1469603203494722e-06, + "loss": 0.3063, + "step": 15672 + }, + { + "epoch": 2.85, + "learning_rate": 2.14690571532581e-06, + "loss": 0.545, + "step": 15673 + }, + { + "epoch": 2.85, + "learning_rate": 2.146851110302148e-06, + "loss": 0.0286, + "step": 15674 + }, + { + "epoch": 2.85, + "learning_rate": 2.1467965052784856e-06, + "loss": 0.5843, + "step": 15675 + }, + { + "epoch": 2.85, + "learning_rate": 2.1467419002548236e-06, + "loss": 0.9392, + "step": 15676 + }, + { + "epoch": 2.85, + "learning_rate": 2.1466872952311615e-06, + "loss": 4.9574, + "step": 15677 + }, + { + "epoch": 2.85, + "learning_rate": 2.146632690207499e-06, + "loss": 0.3937, + "step": 15678 + }, + { + "epoch": 2.85, + "learning_rate": 2.146578085183837e-06, + "loss": 0.0142, + "step": 15679 + }, + { + "epoch": 2.85, + "learning_rate": 2.1465234801601745e-06, + "loss": 2.0189, + "step": 15680 + }, + { + "epoch": 2.85, + "learning_rate": 2.1464688751365125e-06, + "loss": 0.2731, + "step": 15681 + }, + { + "epoch": 2.85, + "learning_rate": 2.1464142701128504e-06, + "loss": 1.5867, + "step": 15682 + }, + { + "epoch": 2.85, + "learning_rate": 2.146359665089188e-06, + "loss": 0.5936, + "step": 15683 + }, + { + "epoch": 2.85, + "learning_rate": 2.146305060065526e-06, + "loss": 0.1677, + "step": 15684 + }, + { + "epoch": 2.85, + "learning_rate": 2.146250455041864e-06, + "loss": 0.1187, + "step": 15685 + }, + { + "epoch": 2.85, + "learning_rate": 2.1461958500182018e-06, + "loss": 0.6796, + "step": 15686 + }, + { + "epoch": 2.85, + "learning_rate": 2.1461412449945397e-06, + "loss": 3.4521, + "step": 15687 + }, + { + "epoch": 2.85, + "learning_rate": 2.1460866399708776e-06, + "loss": 0.4235, + "step": 15688 + }, + { + "epoch": 2.85, + "learning_rate": 2.146032034947215e-06, + "loss": 1.0919, + "step": 15689 + }, + { + "epoch": 2.85, + "learning_rate": 2.145977429923553e-06, + "loss": 0.0391, + "step": 15690 + }, + { + "epoch": 2.85, + "learning_rate": 2.145922824899891e-06, + "loss": 0.0858, + "step": 15691 + }, + { + "epoch": 2.85, + "learning_rate": 2.1458682198762286e-06, + "loss": 0.906, + "step": 15692 + }, + { + "epoch": 2.85, + "learning_rate": 2.1458136148525665e-06, + "loss": 1.0323, + "step": 15693 + }, + { + "epoch": 2.85, + "learning_rate": 2.145759009828904e-06, + "loss": 1.0331, + "step": 15694 + }, + { + "epoch": 2.85, + "learning_rate": 2.145704404805242e-06, + "loss": 1.5171, + "step": 15695 + }, + { + "epoch": 2.85, + "learning_rate": 2.14564979978158e-06, + "loss": 0.8438, + "step": 15696 + }, + { + "epoch": 2.85, + "learning_rate": 2.145595194757918e-06, + "loss": 0.0199, + "step": 15697 + }, + { + "epoch": 2.85, + "learning_rate": 2.145540589734256e-06, + "loss": 0.2962, + "step": 15698 + }, + { + "epoch": 2.85, + "learning_rate": 2.1454859847105937e-06, + "loss": 1.2571, + "step": 15699 + }, + { + "epoch": 2.86, + "learning_rate": 2.1454313796869313e-06, + "loss": 0.0435, + "step": 15700 + }, + { + "epoch": 2.86, + "learning_rate": 2.145376774663269e-06, + "loss": 2.3625, + "step": 15701 + }, + { + "epoch": 2.86, + "learning_rate": 2.145322169639607e-06, + "loss": 1.186, + "step": 15702 + }, + { + "epoch": 2.86, + "learning_rate": 2.1452675646159447e-06, + "loss": 0.8561, + "step": 15703 + }, + { + "epoch": 2.86, + "learning_rate": 2.1452129595922826e-06, + "loss": 0.118, + "step": 15704 + }, + { + "epoch": 2.86, + "learning_rate": 2.1451583545686206e-06, + "loss": 0.5987, + "step": 15705 + }, + { + "epoch": 2.86, + "learning_rate": 2.145103749544958e-06, + "loss": 0.936, + "step": 15706 + }, + { + "epoch": 2.86, + "learning_rate": 2.145049144521296e-06, + "loss": 0.9493, + "step": 15707 + }, + { + "epoch": 2.86, + "learning_rate": 2.1449945394976335e-06, + "loss": 0.9909, + "step": 15708 + }, + { + "epoch": 2.86, + "learning_rate": 2.1449399344739715e-06, + "loss": 0.1655, + "step": 15709 + }, + { + "epoch": 2.86, + "learning_rate": 2.1448853294503094e-06, + "loss": 1.0172, + "step": 15710 + }, + { + "epoch": 2.86, + "learning_rate": 2.1448307244266474e-06, + "loss": 0.0114, + "step": 15711 + }, + { + "epoch": 2.86, + "learning_rate": 2.1447761194029853e-06, + "loss": 1.8027, + "step": 15712 + }, + { + "epoch": 2.86, + "learning_rate": 2.1447215143793233e-06, + "loss": 1.9346, + "step": 15713 + }, + { + "epoch": 2.86, + "learning_rate": 2.1446669093556608e-06, + "loss": 0.7825, + "step": 15714 + }, + { + "epoch": 2.86, + "learning_rate": 2.1446123043319987e-06, + "loss": 1.4886, + "step": 15715 + }, + { + "epoch": 2.86, + "learning_rate": 2.1445576993083367e-06, + "loss": 0.3138, + "step": 15716 + }, + { + "epoch": 2.86, + "learning_rate": 2.144503094284674e-06, + "loss": 0.5495, + "step": 15717 + }, + { + "epoch": 2.86, + "learning_rate": 2.144448489261012e-06, + "loss": 1.0651, + "step": 15718 + }, + { + "epoch": 2.86, + "learning_rate": 2.14439388423735e-06, + "loss": 0.9663, + "step": 15719 + }, + { + "epoch": 2.86, + "learning_rate": 2.1443392792136876e-06, + "loss": 0.5316, + "step": 15720 + }, + { + "epoch": 2.86, + "learning_rate": 2.1442846741900255e-06, + "loss": 1.1005, + "step": 15721 + }, + { + "epoch": 2.86, + "learning_rate": 2.144230069166363e-06, + "loss": 0.2464, + "step": 15722 + }, + { + "epoch": 2.86, + "learning_rate": 2.144175464142701e-06, + "loss": 0.4291, + "step": 15723 + }, + { + "epoch": 2.86, + "learning_rate": 2.144120859119039e-06, + "loss": 0.3574, + "step": 15724 + }, + { + "epoch": 2.86, + "learning_rate": 2.144066254095377e-06, + "loss": 1.2206, + "step": 15725 + }, + { + "epoch": 2.86, + "learning_rate": 2.144011649071715e-06, + "loss": 0.4405, + "step": 15726 + }, + { + "epoch": 2.86, + "learning_rate": 2.1439570440480528e-06, + "loss": 1.6561, + "step": 15727 + }, + { + "epoch": 2.86, + "learning_rate": 2.1439024390243903e-06, + "loss": 0.4176, + "step": 15728 + }, + { + "epoch": 2.86, + "learning_rate": 2.1438478340007282e-06, + "loss": 0.2339, + "step": 15729 + }, + { + "epoch": 2.86, + "learning_rate": 2.143793228977066e-06, + "loss": 1.3432, + "step": 15730 + }, + { + "epoch": 2.86, + "learning_rate": 2.1437386239534037e-06, + "loss": 0.4032, + "step": 15731 + }, + { + "epoch": 2.86, + "learning_rate": 2.1436840189297416e-06, + "loss": 0.6202, + "step": 15732 + }, + { + "epoch": 2.86, + "learning_rate": 2.143629413906079e-06, + "loss": 0.2266, + "step": 15733 + }, + { + "epoch": 2.86, + "learning_rate": 2.143574808882417e-06, + "loss": 0.9923, + "step": 15734 + }, + { + "epoch": 2.86, + "learning_rate": 2.143520203858755e-06, + "loss": 1.9037, + "step": 15735 + }, + { + "epoch": 2.86, + "learning_rate": 2.1434655988350926e-06, + "loss": 0.8135, + "step": 15736 + }, + { + "epoch": 2.86, + "learning_rate": 2.1434109938114305e-06, + "loss": 0.304, + "step": 15737 + }, + { + "epoch": 2.86, + "learning_rate": 2.1433563887877685e-06, + "loss": 0.5358, + "step": 15738 + }, + { + "epoch": 2.86, + "learning_rate": 2.1433017837641064e-06, + "loss": 0.0201, + "step": 15739 + }, + { + "epoch": 2.86, + "learning_rate": 2.1432471787404443e-06, + "loss": 0.139, + "step": 15740 + }, + { + "epoch": 2.86, + "learning_rate": 2.1431925737167823e-06, + "loss": 0.0218, + "step": 15741 + }, + { + "epoch": 2.86, + "learning_rate": 2.14313796869312e-06, + "loss": 0.2387, + "step": 15742 + }, + { + "epoch": 2.86, + "learning_rate": 2.1430833636694578e-06, + "loss": 0.2388, + "step": 15743 + }, + { + "epoch": 2.86, + "learning_rate": 2.1430287586457957e-06, + "loss": 0.8258, + "step": 15744 + }, + { + "epoch": 2.86, + "learning_rate": 2.1429741536221332e-06, + "loss": 1.1614, + "step": 15745 + }, + { + "epoch": 2.86, + "learning_rate": 2.142919548598471e-06, + "loss": 0.4326, + "step": 15746 + }, + { + "epoch": 2.86, + "learning_rate": 2.1428649435748087e-06, + "loss": 0.3821, + "step": 15747 + }, + { + "epoch": 2.86, + "learning_rate": 2.1428103385511466e-06, + "loss": 0.1216, + "step": 15748 + }, + { + "epoch": 2.86, + "learning_rate": 2.1427557335274846e-06, + "loss": 0.8156, + "step": 15749 + }, + { + "epoch": 2.86, + "learning_rate": 2.142701128503822e-06, + "loss": 2.2422, + "step": 15750 + }, + { + "epoch": 2.86, + "learning_rate": 2.14264652348016e-06, + "loss": 0.5671, + "step": 15751 + }, + { + "epoch": 2.86, + "learning_rate": 2.142591918456498e-06, + "loss": 0.3354, + "step": 15752 + }, + { + "epoch": 2.86, + "learning_rate": 2.142537313432836e-06, + "loss": 0.4081, + "step": 15753 + }, + { + "epoch": 2.86, + "learning_rate": 2.142482708409174e-06, + "loss": 0.9764, + "step": 15754 + }, + { + "epoch": 2.87, + "learning_rate": 2.142428103385512e-06, + "loss": 0.4886, + "step": 15755 + }, + { + "epoch": 2.87, + "learning_rate": 2.1423734983618493e-06, + "loss": 2.4077, + "step": 15756 + }, + { + "epoch": 2.87, + "learning_rate": 2.1423188933381873e-06, + "loss": 0.6794, + "step": 15757 + }, + { + "epoch": 2.87, + "learning_rate": 2.142264288314525e-06, + "loss": 0.4341, + "step": 15758 + }, + { + "epoch": 2.87, + "learning_rate": 2.1422096832908627e-06, + "loss": 1.3177, + "step": 15759 + }, + { + "epoch": 2.87, + "learning_rate": 2.1421550782672007e-06, + "loss": 0.5233, + "step": 15760 + }, + { + "epoch": 2.87, + "learning_rate": 2.142100473243538e-06, + "loss": 0.9719, + "step": 15761 + }, + { + "epoch": 2.87, + "learning_rate": 2.142045868219876e-06, + "loss": 1.3341, + "step": 15762 + }, + { + "epoch": 2.87, + "learning_rate": 2.141991263196214e-06, + "loss": 0.3824, + "step": 15763 + }, + { + "epoch": 2.87, + "learning_rate": 2.1419366581725516e-06, + "loss": 0.1534, + "step": 15764 + }, + { + "epoch": 2.87, + "learning_rate": 2.1418820531488895e-06, + "loss": 1.82, + "step": 15765 + }, + { + "epoch": 2.87, + "learning_rate": 2.1418274481252275e-06, + "loss": 1.3058, + "step": 15766 + }, + { + "epoch": 2.87, + "learning_rate": 2.1417728431015654e-06, + "loss": 1.3436, + "step": 15767 + }, + { + "epoch": 2.87, + "learning_rate": 2.1417182380779034e-06, + "loss": 0.2806, + "step": 15768 + }, + { + "epoch": 2.87, + "learning_rate": 2.1416636330542413e-06, + "loss": 1.5376, + "step": 15769 + }, + { + "epoch": 2.87, + "learning_rate": 2.141609028030579e-06, + "loss": 0.0247, + "step": 15770 + }, + { + "epoch": 2.87, + "learning_rate": 2.1415544230069168e-06, + "loss": 0.1195, + "step": 15771 + }, + { + "epoch": 2.87, + "learning_rate": 2.1414998179832547e-06, + "loss": 0.3407, + "step": 15772 + }, + { + "epoch": 2.87, + "learning_rate": 2.1414452129595922e-06, + "loss": 0.1837, + "step": 15773 + }, + { + "epoch": 2.87, + "learning_rate": 2.14139060793593e-06, + "loss": 2.4449, + "step": 15774 + }, + { + "epoch": 2.87, + "learning_rate": 2.1413360029122677e-06, + "loss": 0.9367, + "step": 15775 + }, + { + "epoch": 2.87, + "learning_rate": 2.1412813978886056e-06, + "loss": 1.2469, + "step": 15776 + }, + { + "epoch": 2.87, + "learning_rate": 2.1412267928649436e-06, + "loss": 0.5199, + "step": 15777 + }, + { + "epoch": 2.87, + "learning_rate": 2.1411721878412815e-06, + "loss": 1.5186, + "step": 15778 + }, + { + "epoch": 2.87, + "learning_rate": 2.1411175828176195e-06, + "loss": 1.6063, + "step": 15779 + }, + { + "epoch": 2.87, + "learning_rate": 2.1410629777939574e-06, + "loss": 1.6144, + "step": 15780 + }, + { + "epoch": 2.87, + "learning_rate": 2.141008372770295e-06, + "loss": 0.6261, + "step": 15781 + }, + { + "epoch": 2.87, + "learning_rate": 2.140953767746633e-06, + "loss": 0.6571, + "step": 15782 + }, + { + "epoch": 2.87, + "learning_rate": 2.140899162722971e-06, + "loss": 0.4449, + "step": 15783 + }, + { + "epoch": 2.87, + "learning_rate": 2.1408445576993083e-06, + "loss": 0.3114, + "step": 15784 + }, + { + "epoch": 2.87, + "learning_rate": 2.1407899526756463e-06, + "loss": 0.8756, + "step": 15785 + }, + { + "epoch": 2.87, + "learning_rate": 2.1407353476519842e-06, + "loss": 3.0168, + "step": 15786 + }, + { + "epoch": 2.87, + "learning_rate": 2.1406807426283218e-06, + "loss": 0.0658, + "step": 15787 + }, + { + "epoch": 2.87, + "learning_rate": 2.1406261376046597e-06, + "loss": 0.3872, + "step": 15788 + }, + { + "epoch": 2.87, + "learning_rate": 2.1405715325809972e-06, + "loss": 1.8312, + "step": 15789 + }, + { + "epoch": 2.87, + "learning_rate": 2.140516927557335e-06, + "loss": 1.1689, + "step": 15790 + }, + { + "epoch": 2.87, + "learning_rate": 2.140462322533673e-06, + "loss": 1.2293, + "step": 15791 + }, + { + "epoch": 2.87, + "learning_rate": 2.140407717510011e-06, + "loss": 0.3865, + "step": 15792 + }, + { + "epoch": 2.87, + "learning_rate": 2.140353112486349e-06, + "loss": 0.9021, + "step": 15793 + }, + { + "epoch": 2.87, + "learning_rate": 2.140298507462687e-06, + "loss": 0.0273, + "step": 15794 + }, + { + "epoch": 2.87, + "learning_rate": 2.1402439024390245e-06, + "loss": 2.3665, + "step": 15795 + }, + { + "epoch": 2.87, + "learning_rate": 2.1401892974153624e-06, + "loss": 0.7677, + "step": 15796 + }, + { + "epoch": 2.87, + "learning_rate": 2.1401346923917003e-06, + "loss": 2.3903, + "step": 15797 + }, + { + "epoch": 2.87, + "learning_rate": 2.140080087368038e-06, + "loss": 1.344, + "step": 15798 + }, + { + "epoch": 2.87, + "learning_rate": 2.140025482344376e-06, + "loss": 1.2615, + "step": 15799 + }, + { + "epoch": 2.87, + "learning_rate": 2.1399708773207137e-06, + "loss": 0.685, + "step": 15800 + }, + { + "epoch": 2.87, + "learning_rate": 2.1399162722970513e-06, + "loss": 0.4689, + "step": 15801 + }, + { + "epoch": 2.87, + "learning_rate": 2.139861667273389e-06, + "loss": 0.5713, + "step": 15802 + }, + { + "epoch": 2.87, + "learning_rate": 2.1398070622497267e-06, + "loss": 0.2717, + "step": 15803 + }, + { + "epoch": 2.87, + "learning_rate": 2.1397524572260647e-06, + "loss": 0.3738, + "step": 15804 + }, + { + "epoch": 2.87, + "learning_rate": 2.1396978522024026e-06, + "loss": 1.3078, + "step": 15805 + }, + { + "epoch": 2.87, + "learning_rate": 2.1396432471787406e-06, + "loss": 1.8483, + "step": 15806 + }, + { + "epoch": 2.87, + "learning_rate": 2.1395886421550785e-06, + "loss": 2.613, + "step": 15807 + }, + { + "epoch": 2.87, + "learning_rate": 2.1395340371314164e-06, + "loss": 2.3929, + "step": 15808 + }, + { + "epoch": 2.87, + "learning_rate": 2.139479432107754e-06, + "loss": 1.1951, + "step": 15809 + }, + { + "epoch": 2.88, + "learning_rate": 2.139424827084092e-06, + "loss": 0.05, + "step": 15810 + }, + { + "epoch": 2.88, + "learning_rate": 2.13937022206043e-06, + "loss": 0.2486, + "step": 15811 + }, + { + "epoch": 2.88, + "learning_rate": 2.1393156170367674e-06, + "loss": 0.2187, + "step": 15812 + }, + { + "epoch": 2.88, + "learning_rate": 2.1392610120131053e-06, + "loss": 0.2839, + "step": 15813 + }, + { + "epoch": 2.88, + "learning_rate": 2.139206406989443e-06, + "loss": 1.0237, + "step": 15814 + }, + { + "epoch": 2.88, + "learning_rate": 2.1391518019657808e-06, + "loss": 1.077, + "step": 15815 + }, + { + "epoch": 2.88, + "learning_rate": 2.1390971969421187e-06, + "loss": 0.6113, + "step": 15816 + }, + { + "epoch": 2.88, + "learning_rate": 2.1390425919184562e-06, + "loss": 1.0126, + "step": 15817 + }, + { + "epoch": 2.88, + "learning_rate": 2.138987986894794e-06, + "loss": 0.0152, + "step": 15818 + }, + { + "epoch": 2.88, + "learning_rate": 2.138933381871132e-06, + "loss": 1.2545, + "step": 15819 + }, + { + "epoch": 2.88, + "learning_rate": 2.13887877684747e-06, + "loss": 0.421, + "step": 15820 + }, + { + "epoch": 2.88, + "learning_rate": 2.138824171823808e-06, + "loss": 0.3679, + "step": 15821 + }, + { + "epoch": 2.88, + "learning_rate": 2.138769566800146e-06, + "loss": 1.7806, + "step": 15822 + }, + { + "epoch": 2.88, + "learning_rate": 2.1387149617764835e-06, + "loss": 1.8794, + "step": 15823 + }, + { + "epoch": 2.88, + "learning_rate": 2.1386603567528214e-06, + "loss": 1.0348, + "step": 15824 + }, + { + "epoch": 2.88, + "learning_rate": 2.1386057517291594e-06, + "loss": 0.2825, + "step": 15825 + }, + { + "epoch": 2.88, + "learning_rate": 2.138551146705497e-06, + "loss": 0.7348, + "step": 15826 + }, + { + "epoch": 2.88, + "learning_rate": 2.138496541681835e-06, + "loss": 0.6744, + "step": 15827 + }, + { + "epoch": 2.88, + "learning_rate": 2.1384419366581724e-06, + "loss": 0.5234, + "step": 15828 + }, + { + "epoch": 2.88, + "learning_rate": 2.1383873316345103e-06, + "loss": 1.4606, + "step": 15829 + }, + { + "epoch": 2.88, + "learning_rate": 2.1383327266108482e-06, + "loss": 0.5447, + "step": 15830 + }, + { + "epoch": 2.88, + "learning_rate": 2.1382781215871858e-06, + "loss": 0.3841, + "step": 15831 + }, + { + "epoch": 2.88, + "learning_rate": 2.1382235165635237e-06, + "loss": 0.021, + "step": 15832 + }, + { + "epoch": 2.88, + "learning_rate": 2.1381689115398616e-06, + "loss": 1.0152, + "step": 15833 + }, + { + "epoch": 2.88, + "learning_rate": 2.1381143065161996e-06, + "loss": 0.5085, + "step": 15834 + }, + { + "epoch": 2.88, + "learning_rate": 2.1380597014925375e-06, + "loss": 0.5517, + "step": 15835 + }, + { + "epoch": 2.88, + "learning_rate": 2.1380050964688755e-06, + "loss": 1.0197, + "step": 15836 + }, + { + "epoch": 2.88, + "learning_rate": 2.137950491445213e-06, + "loss": 1.5239, + "step": 15837 + }, + { + "epoch": 2.88, + "learning_rate": 2.137895886421551e-06, + "loss": 1.3237, + "step": 15838 + }, + { + "epoch": 2.88, + "learning_rate": 2.137841281397889e-06, + "loss": 0.1092, + "step": 15839 + }, + { + "epoch": 2.88, + "learning_rate": 2.1377866763742264e-06, + "loss": 3.3432, + "step": 15840 + }, + { + "epoch": 2.88, + "learning_rate": 2.1377320713505643e-06, + "loss": 0.9132, + "step": 15841 + }, + { + "epoch": 2.88, + "learning_rate": 2.137677466326902e-06, + "loss": 1.7124, + "step": 15842 + }, + { + "epoch": 2.88, + "learning_rate": 2.13762286130324e-06, + "loss": 1.9784, + "step": 15843 + }, + { + "epoch": 2.88, + "learning_rate": 2.1375682562795778e-06, + "loss": 1.0395, + "step": 15844 + }, + { + "epoch": 2.88, + "learning_rate": 2.1375136512559153e-06, + "loss": 3.0013, + "step": 15845 + }, + { + "epoch": 2.88, + "learning_rate": 2.1374590462322532e-06, + "loss": 0.3085, + "step": 15846 + }, + { + "epoch": 2.88, + "learning_rate": 2.137404441208591e-06, + "loss": 0.5884, + "step": 15847 + }, + { + "epoch": 2.88, + "learning_rate": 2.137349836184929e-06, + "loss": 0.0417, + "step": 15848 + }, + { + "epoch": 2.88, + "learning_rate": 2.137295231161267e-06, + "loss": 3.9993, + "step": 15849 + }, + { + "epoch": 2.88, + "learning_rate": 2.137240626137605e-06, + "loss": 0.4447, + "step": 15850 + }, + { + "epoch": 2.88, + "learning_rate": 2.1371860211139425e-06, + "loss": 0.5684, + "step": 15851 + }, + { + "epoch": 2.88, + "learning_rate": 2.1371314160902805e-06, + "loss": 0.563, + "step": 15852 + }, + { + "epoch": 2.88, + "learning_rate": 2.1370768110666184e-06, + "loss": 0.5394, + "step": 15853 + }, + { + "epoch": 2.88, + "learning_rate": 2.137022206042956e-06, + "loss": 0.2495, + "step": 15854 + }, + { + "epoch": 2.88, + "learning_rate": 2.136967601019294e-06, + "loss": 0.4506, + "step": 15855 + }, + { + "epoch": 2.88, + "learning_rate": 2.1369129959956314e-06, + "loss": 0.5472, + "step": 15856 + }, + { + "epoch": 2.88, + "learning_rate": 2.1368583909719693e-06, + "loss": 0.059, + "step": 15857 + }, + { + "epoch": 2.88, + "learning_rate": 2.1368037859483073e-06, + "loss": 0.2979, + "step": 15858 + }, + { + "epoch": 2.88, + "learning_rate": 2.1367491809246448e-06, + "loss": 0.4053, + "step": 15859 + }, + { + "epoch": 2.88, + "learning_rate": 2.1366945759009827e-06, + "loss": 1.7932, + "step": 15860 + }, + { + "epoch": 2.88, + "learning_rate": 2.1366399708773207e-06, + "loss": 0.4841, + "step": 15861 + }, + { + "epoch": 2.88, + "learning_rate": 2.1365853658536586e-06, + "loss": 0.0276, + "step": 15862 + }, + { + "epoch": 2.88, + "learning_rate": 2.1365307608299966e-06, + "loss": 0.8205, + "step": 15863 + }, + { + "epoch": 2.88, + "learning_rate": 2.1364761558063345e-06, + "loss": 1.0533, + "step": 15864 + }, + { + "epoch": 2.89, + "learning_rate": 2.136421550782672e-06, + "loss": 0.721, + "step": 15865 + }, + { + "epoch": 2.89, + "learning_rate": 2.13636694575901e-06, + "loss": 0.1056, + "step": 15866 + }, + { + "epoch": 2.89, + "learning_rate": 2.136312340735348e-06, + "loss": 0.0602, + "step": 15867 + }, + { + "epoch": 2.89, + "learning_rate": 2.1362577357116854e-06, + "loss": 3.9111, + "step": 15868 + }, + { + "epoch": 2.89, + "learning_rate": 2.1362031306880234e-06, + "loss": 0.0159, + "step": 15869 + }, + { + "epoch": 2.89, + "learning_rate": 2.136148525664361e-06, + "loss": 0.1179, + "step": 15870 + }, + { + "epoch": 2.89, + "learning_rate": 2.136093920640699e-06, + "loss": 0.7829, + "step": 15871 + }, + { + "epoch": 2.89, + "learning_rate": 2.1360393156170368e-06, + "loss": 1.6757, + "step": 15872 + }, + { + "epoch": 2.89, + "learning_rate": 2.1359847105933747e-06, + "loss": 1.0047, + "step": 15873 + }, + { + "epoch": 2.89, + "learning_rate": 2.1359301055697127e-06, + "loss": 0.8446, + "step": 15874 + }, + { + "epoch": 2.89, + "learning_rate": 2.1358755005460506e-06, + "loss": 0.976, + "step": 15875 + }, + { + "epoch": 2.89, + "learning_rate": 2.135820895522388e-06, + "loss": 0.7998, + "step": 15876 + }, + { + "epoch": 2.89, + "learning_rate": 2.135766290498726e-06, + "loss": 1.7697, + "step": 15877 + }, + { + "epoch": 2.89, + "learning_rate": 2.135711685475064e-06, + "loss": 2.7476, + "step": 15878 + }, + { + "epoch": 2.89, + "learning_rate": 2.1356570804514015e-06, + "loss": 1.32, + "step": 15879 + }, + { + "epoch": 2.89, + "learning_rate": 2.1356024754277395e-06, + "loss": 0.578, + "step": 15880 + }, + { + "epoch": 2.89, + "learning_rate": 2.135547870404077e-06, + "loss": 0.6541, + "step": 15881 + }, + { + "epoch": 2.89, + "learning_rate": 2.135493265380415e-06, + "loss": 1.3951, + "step": 15882 + }, + { + "epoch": 2.89, + "learning_rate": 2.135438660356753e-06, + "loss": 1.0356, + "step": 15883 + }, + { + "epoch": 2.89, + "learning_rate": 2.1353840553330904e-06, + "loss": 1.6576, + "step": 15884 + }, + { + "epoch": 2.89, + "learning_rate": 2.1353294503094283e-06, + "loss": 1.5063, + "step": 15885 + }, + { + "epoch": 2.89, + "learning_rate": 2.1352748452857663e-06, + "loss": 1.6771, + "step": 15886 + }, + { + "epoch": 2.89, + "learning_rate": 2.1352202402621042e-06, + "loss": 0.2152, + "step": 15887 + }, + { + "epoch": 2.89, + "learning_rate": 2.135165635238442e-06, + "loss": 0.3318, + "step": 15888 + }, + { + "epoch": 2.89, + "learning_rate": 2.13511103021478e-06, + "loss": 0.8867, + "step": 15889 + }, + { + "epoch": 2.89, + "learning_rate": 2.1350564251911176e-06, + "loss": 0.4738, + "step": 15890 + }, + { + "epoch": 2.89, + "learning_rate": 2.1350018201674556e-06, + "loss": 1.429, + "step": 15891 + }, + { + "epoch": 2.89, + "learning_rate": 2.1349472151437935e-06, + "loss": 0.1185, + "step": 15892 + }, + { + "epoch": 2.89, + "learning_rate": 2.134892610120131e-06, + "loss": 0.6624, + "step": 15893 + }, + { + "epoch": 2.89, + "learning_rate": 2.134838005096469e-06, + "loss": 0.2661, + "step": 15894 + }, + { + "epoch": 2.89, + "learning_rate": 2.1347834000728065e-06, + "loss": 0.3075, + "step": 15895 + }, + { + "epoch": 2.89, + "learning_rate": 2.1347287950491445e-06, + "loss": 0.2531, + "step": 15896 + }, + { + "epoch": 2.89, + "learning_rate": 2.1346741900254824e-06, + "loss": 0.0121, + "step": 15897 + }, + { + "epoch": 2.89, + "learning_rate": 2.13461958500182e-06, + "loss": 0.1441, + "step": 15898 + }, + { + "epoch": 2.89, + "learning_rate": 2.134564979978158e-06, + "loss": 0.5004, + "step": 15899 + }, + { + "epoch": 2.89, + "learning_rate": 2.134510374954496e-06, + "loss": 2.7175, + "step": 15900 + }, + { + "epoch": 2.89, + "learning_rate": 2.1344557699308337e-06, + "loss": 1.5024, + "step": 15901 + }, + { + "epoch": 2.89, + "learning_rate": 2.1344011649071717e-06, + "loss": 1.216, + "step": 15902 + }, + { + "epoch": 2.89, + "learning_rate": 2.1343465598835096e-06, + "loss": 0.0182, + "step": 15903 + }, + { + "epoch": 2.89, + "learning_rate": 2.134291954859847e-06, + "loss": 2.1989, + "step": 15904 + }, + { + "epoch": 2.89, + "learning_rate": 2.134237349836185e-06, + "loss": 0.358, + "step": 15905 + }, + { + "epoch": 2.89, + "learning_rate": 2.134182744812523e-06, + "loss": 0.1361, + "step": 15906 + }, + { + "epoch": 2.89, + "learning_rate": 2.1341281397888606e-06, + "loss": 0.6086, + "step": 15907 + }, + { + "epoch": 2.89, + "learning_rate": 2.1340735347651985e-06, + "loss": 0.4669, + "step": 15908 + }, + { + "epoch": 2.89, + "learning_rate": 2.134018929741536e-06, + "loss": 3.5251, + "step": 15909 + }, + { + "epoch": 2.89, + "learning_rate": 2.133964324717874e-06, + "loss": 1.1879, + "step": 15910 + }, + { + "epoch": 2.89, + "learning_rate": 2.133909719694212e-06, + "loss": 0.0351, + "step": 15911 + }, + { + "epoch": 2.89, + "learning_rate": 2.1338551146705494e-06, + "loss": 1.8661, + "step": 15912 + }, + { + "epoch": 2.89, + "learning_rate": 2.1338005096468874e-06, + "loss": 0.5559, + "step": 15913 + }, + { + "epoch": 2.89, + "learning_rate": 2.1337459046232253e-06, + "loss": 0.0324, + "step": 15914 + }, + { + "epoch": 2.89, + "learning_rate": 2.1336912995995633e-06, + "loss": 1.1476, + "step": 15915 + }, + { + "epoch": 2.89, + "learning_rate": 2.133636694575901e-06, + "loss": 1.1481, + "step": 15916 + }, + { + "epoch": 2.89, + "learning_rate": 2.133582089552239e-06, + "loss": 0.9047, + "step": 15917 + }, + { + "epoch": 2.89, + "learning_rate": 2.1335274845285767e-06, + "loss": 1.0489, + "step": 15918 + }, + { + "epoch": 2.89, + "learning_rate": 2.1334728795049146e-06, + "loss": 1.312, + "step": 15919 + }, + { + "epoch": 2.9, + "learning_rate": 2.1334182744812526e-06, + "loss": 0.1002, + "step": 15920 + }, + { + "epoch": 2.9, + "learning_rate": 2.13336366945759e-06, + "loss": 2.4536, + "step": 15921 + }, + { + "epoch": 2.9, + "learning_rate": 2.133309064433928e-06, + "loss": 0.2544, + "step": 15922 + }, + { + "epoch": 2.9, + "learning_rate": 2.1332544594102655e-06, + "loss": 1.3611, + "step": 15923 + }, + { + "epoch": 2.9, + "learning_rate": 2.1331998543866035e-06, + "loss": 0.7194, + "step": 15924 + }, + { + "epoch": 2.9, + "learning_rate": 2.1331452493629414e-06, + "loss": 0.6156, + "step": 15925 + }, + { + "epoch": 2.9, + "learning_rate": 2.133090644339279e-06, + "loss": 0.7847, + "step": 15926 + }, + { + "epoch": 2.9, + "learning_rate": 2.133036039315617e-06, + "loss": 0.0911, + "step": 15927 + }, + { + "epoch": 2.9, + "learning_rate": 2.132981434291955e-06, + "loss": 0.0327, + "step": 15928 + }, + { + "epoch": 2.9, + "learning_rate": 2.1329268292682928e-06, + "loss": 0.3981, + "step": 15929 + }, + { + "epoch": 2.9, + "learning_rate": 2.1328722242446307e-06, + "loss": 0.2019, + "step": 15930 + }, + { + "epoch": 2.9, + "learning_rate": 2.1328176192209687e-06, + "loss": 0.2857, + "step": 15931 + }, + { + "epoch": 2.9, + "learning_rate": 2.132763014197306e-06, + "loss": 1.1715, + "step": 15932 + }, + { + "epoch": 2.9, + "learning_rate": 2.132708409173644e-06, + "loss": 0.2753, + "step": 15933 + }, + { + "epoch": 2.9, + "learning_rate": 2.132653804149982e-06, + "loss": 0.7993, + "step": 15934 + }, + { + "epoch": 2.9, + "learning_rate": 2.1325991991263196e-06, + "loss": 0.1729, + "step": 15935 + }, + { + "epoch": 2.9, + "learning_rate": 2.1325445941026575e-06, + "loss": 0.7786, + "step": 15936 + }, + { + "epoch": 2.9, + "learning_rate": 2.132489989078995e-06, + "loss": 0.9971, + "step": 15937 + }, + { + "epoch": 2.9, + "learning_rate": 2.132435384055333e-06, + "loss": 0.2338, + "step": 15938 + }, + { + "epoch": 2.9, + "learning_rate": 2.132380779031671e-06, + "loss": 0.4136, + "step": 15939 + }, + { + "epoch": 2.9, + "learning_rate": 2.1323261740080085e-06, + "loss": 0.619, + "step": 15940 + }, + { + "epoch": 2.9, + "learning_rate": 2.1322715689843464e-06, + "loss": 0.5976, + "step": 15941 + }, + { + "epoch": 2.9, + "learning_rate": 2.1322169639606843e-06, + "loss": 2.3291, + "step": 15942 + }, + { + "epoch": 2.9, + "learning_rate": 2.1321623589370223e-06, + "loss": 1.4951, + "step": 15943 + }, + { + "epoch": 2.9, + "learning_rate": 2.1321077539133602e-06, + "loss": 0.401, + "step": 15944 + }, + { + "epoch": 2.9, + "learning_rate": 2.132053148889698e-06, + "loss": 0.0994, + "step": 15945 + }, + { + "epoch": 2.9, + "learning_rate": 2.1319985438660357e-06, + "loss": 1.8855, + "step": 15946 + }, + { + "epoch": 2.9, + "learning_rate": 2.1319439388423736e-06, + "loss": 0.2932, + "step": 15947 + }, + { + "epoch": 2.9, + "learning_rate": 2.1318893338187116e-06, + "loss": 1.4558, + "step": 15948 + }, + { + "epoch": 2.9, + "learning_rate": 2.131834728795049e-06, + "loss": 0.0116, + "step": 15949 + }, + { + "epoch": 2.9, + "learning_rate": 2.131780123771387e-06, + "loss": 0.5106, + "step": 15950 + }, + { + "epoch": 2.9, + "learning_rate": 2.1317255187477246e-06, + "loss": 0.6568, + "step": 15951 + }, + { + "epoch": 2.9, + "learning_rate": 2.1316709137240625e-06, + "loss": 0.5459, + "step": 15952 + }, + { + "epoch": 2.9, + "learning_rate": 2.1316163087004005e-06, + "loss": 0.9826, + "step": 15953 + }, + { + "epoch": 2.9, + "learning_rate": 2.1315617036767384e-06, + "loss": 0.9277, + "step": 15954 + }, + { + "epoch": 2.9, + "learning_rate": 2.1315070986530763e-06, + "loss": 0.1369, + "step": 15955 + }, + { + "epoch": 2.9, + "learning_rate": 2.1314524936294143e-06, + "loss": 0.5169, + "step": 15956 + }, + { + "epoch": 2.9, + "learning_rate": 2.131397888605752e-06, + "loss": 0.3667, + "step": 15957 + }, + { + "epoch": 2.9, + "learning_rate": 2.1313432835820897e-06, + "loss": 0.4536, + "step": 15958 + }, + { + "epoch": 2.9, + "learning_rate": 2.1312886785584277e-06, + "loss": 0.026, + "step": 15959 + }, + { + "epoch": 2.9, + "learning_rate": 2.131234073534765e-06, + "loss": 1.7525, + "step": 15960 + }, + { + "epoch": 2.9, + "learning_rate": 2.131179468511103e-06, + "loss": 0.9512, + "step": 15961 + }, + { + "epoch": 2.9, + "learning_rate": 2.1311248634874407e-06, + "loss": 1.5262, + "step": 15962 + }, + { + "epoch": 2.9, + "learning_rate": 2.1310702584637786e-06, + "loss": 0.352, + "step": 15963 + }, + { + "epoch": 2.9, + "learning_rate": 2.1310156534401166e-06, + "loss": 2.0147, + "step": 15964 + }, + { + "epoch": 2.9, + "learning_rate": 2.130961048416454e-06, + "loss": 0.3523, + "step": 15965 + }, + { + "epoch": 2.9, + "learning_rate": 2.130906443392792e-06, + "loss": 0.9285, + "step": 15966 + }, + { + "epoch": 2.9, + "learning_rate": 2.13085183836913e-06, + "loss": 4.0443, + "step": 15967 + }, + { + "epoch": 2.9, + "learning_rate": 2.130797233345468e-06, + "loss": 0.6564, + "step": 15968 + }, + { + "epoch": 2.9, + "learning_rate": 2.130742628321806e-06, + "loss": 0.3651, + "step": 15969 + }, + { + "epoch": 2.9, + "learning_rate": 2.130688023298144e-06, + "loss": 0.3298, + "step": 15970 + }, + { + "epoch": 2.9, + "learning_rate": 2.1306334182744813e-06, + "loss": 0.5223, + "step": 15971 + }, + { + "epoch": 2.9, + "learning_rate": 2.1305788132508193e-06, + "loss": 0.4782, + "step": 15972 + }, + { + "epoch": 2.9, + "learning_rate": 2.130524208227157e-06, + "loss": 0.2356, + "step": 15973 + }, + { + "epoch": 2.9, + "learning_rate": 2.1304696032034947e-06, + "loss": 0.6141, + "step": 15974 + }, + { + "epoch": 2.91, + "learning_rate": 2.1304149981798327e-06, + "loss": 0.0341, + "step": 15975 + }, + { + "epoch": 2.91, + "learning_rate": 2.13036039315617e-06, + "loss": 0.2743, + "step": 15976 + }, + { + "epoch": 2.91, + "learning_rate": 2.130305788132508e-06, + "loss": 1.7429, + "step": 15977 + }, + { + "epoch": 2.91, + "learning_rate": 2.130251183108846e-06, + "loss": 0.2103, + "step": 15978 + }, + { + "epoch": 2.91, + "learning_rate": 2.1301965780851836e-06, + "loss": 0.7987, + "step": 15979 + }, + { + "epoch": 2.91, + "learning_rate": 2.1301419730615215e-06, + "loss": 1.0649, + "step": 15980 + }, + { + "epoch": 2.91, + "learning_rate": 2.1300873680378595e-06, + "loss": 1.1398, + "step": 15981 + }, + { + "epoch": 2.91, + "learning_rate": 2.1300327630141974e-06, + "loss": 0.287, + "step": 15982 + }, + { + "epoch": 2.91, + "learning_rate": 2.1299781579905354e-06, + "loss": 0.3404, + "step": 15983 + }, + { + "epoch": 2.91, + "learning_rate": 2.1299235529668733e-06, + "loss": 0.3932, + "step": 15984 + }, + { + "epoch": 2.91, + "learning_rate": 2.129868947943211e-06, + "loss": 0.5129, + "step": 15985 + }, + { + "epoch": 2.91, + "learning_rate": 2.1298143429195488e-06, + "loss": 1.3248, + "step": 15986 + }, + { + "epoch": 2.91, + "learning_rate": 2.1297597378958867e-06, + "loss": 3.3864, + "step": 15987 + }, + { + "epoch": 2.91, + "learning_rate": 2.1297051328722242e-06, + "loss": 0.324, + "step": 15988 + }, + { + "epoch": 2.91, + "learning_rate": 2.129650527848562e-06, + "loss": 0.6809, + "step": 15989 + }, + { + "epoch": 2.91, + "learning_rate": 2.1295959228248997e-06, + "loss": 0.65, + "step": 15990 + }, + { + "epoch": 2.91, + "learning_rate": 2.1295413178012376e-06, + "loss": 0.6874, + "step": 15991 + }, + { + "epoch": 2.91, + "learning_rate": 2.1294867127775756e-06, + "loss": 1.8038, + "step": 15992 + }, + { + "epoch": 2.91, + "learning_rate": 2.129432107753913e-06, + "loss": 0.954, + "step": 15993 + }, + { + "epoch": 2.91, + "learning_rate": 2.129377502730251e-06, + "loss": 1.4339, + "step": 15994 + }, + { + "epoch": 2.91, + "learning_rate": 2.129322897706589e-06, + "loss": 0.0295, + "step": 15995 + }, + { + "epoch": 2.91, + "learning_rate": 2.129268292682927e-06, + "loss": 0.3821, + "step": 15996 + }, + { + "epoch": 2.91, + "learning_rate": 2.129213687659265e-06, + "loss": 0.0752, + "step": 15997 + }, + { + "epoch": 2.91, + "learning_rate": 2.129159082635603e-06, + "loss": 0.0463, + "step": 15998 + }, + { + "epoch": 2.91, + "learning_rate": 2.1291044776119403e-06, + "loss": 0.7087, + "step": 15999 + }, + { + "epoch": 2.91, + "learning_rate": 2.1290498725882783e-06, + "loss": 0.0079, + "step": 16000 + }, + { + "epoch": 2.91, + "learning_rate": 2.1289952675646162e-06, + "loss": 0.374, + "step": 16001 + }, + { + "epoch": 2.91, + "learning_rate": 2.1289406625409538e-06, + "loss": 0.1431, + "step": 16002 + }, + { + "epoch": 2.91, + "learning_rate": 2.1288860575172917e-06, + "loss": 0.0201, + "step": 16003 + }, + { + "epoch": 2.91, + "learning_rate": 2.1288314524936292e-06, + "loss": 0.3271, + "step": 16004 + }, + { + "epoch": 2.91, + "learning_rate": 2.128776847469967e-06, + "loss": 0.2512, + "step": 16005 + }, + { + "epoch": 2.91, + "learning_rate": 2.128722242446305e-06, + "loss": 1.0031, + "step": 16006 + }, + { + "epoch": 2.91, + "learning_rate": 2.1286676374226426e-06, + "loss": 0.6354, + "step": 16007 + }, + { + "epoch": 2.91, + "learning_rate": 2.1286130323989806e-06, + "loss": 1.3435, + "step": 16008 + }, + { + "epoch": 2.91, + "learning_rate": 2.1285584273753185e-06, + "loss": 1.2997, + "step": 16009 + }, + { + "epoch": 2.91, + "learning_rate": 2.1285038223516565e-06, + "loss": 0.8349, + "step": 16010 + }, + { + "epoch": 2.91, + "learning_rate": 2.1284492173279944e-06, + "loss": 1.8937, + "step": 16011 + }, + { + "epoch": 2.91, + "learning_rate": 2.1283946123043323e-06, + "loss": 0.8409, + "step": 16012 + }, + { + "epoch": 2.91, + "learning_rate": 2.12834000728067e-06, + "loss": 2.1154, + "step": 16013 + }, + { + "epoch": 2.91, + "learning_rate": 2.128285402257008e-06, + "loss": 0.7566, + "step": 16014 + }, + { + "epoch": 2.91, + "learning_rate": 2.1282307972333457e-06, + "loss": 0.4316, + "step": 16015 + }, + { + "epoch": 2.91, + "learning_rate": 2.1281761922096833e-06, + "loss": 0.0357, + "step": 16016 + }, + { + "epoch": 2.91, + "learning_rate": 2.128121587186021e-06, + "loss": 0.0227, + "step": 16017 + }, + { + "epoch": 2.91, + "learning_rate": 2.1280669821623587e-06, + "loss": 0.1381, + "step": 16018 + }, + { + "epoch": 2.91, + "learning_rate": 2.1280123771386967e-06, + "loss": 1.3234, + "step": 16019 + }, + { + "epoch": 2.91, + "learning_rate": 2.1279577721150346e-06, + "loss": 0.3128, + "step": 16020 + }, + { + "epoch": 2.91, + "learning_rate": 2.127903167091372e-06, + "loss": 1.2734, + "step": 16021 + }, + { + "epoch": 2.91, + "learning_rate": 2.12784856206771e-06, + "loss": 0.521, + "step": 16022 + }, + { + "epoch": 2.91, + "learning_rate": 2.127793957044048e-06, + "loss": 0.2193, + "step": 16023 + }, + { + "epoch": 2.91, + "learning_rate": 2.127739352020386e-06, + "loss": 0.6323, + "step": 16024 + }, + { + "epoch": 2.91, + "learning_rate": 2.127684746996724e-06, + "loss": 1.5643, + "step": 16025 + }, + { + "epoch": 2.91, + "learning_rate": 2.127630141973062e-06, + "loss": 0.9724, + "step": 16026 + }, + { + "epoch": 2.91, + "learning_rate": 2.1275755369493994e-06, + "loss": 1.1874, + "step": 16027 + }, + { + "epoch": 2.91, + "learning_rate": 2.1275209319257373e-06, + "loss": 0.6626, + "step": 16028 + }, + { + "epoch": 2.91, + "learning_rate": 2.1274663269020753e-06, + "loss": 1.852, + "step": 16029 + }, + { + "epoch": 2.92, + "learning_rate": 2.1274117218784128e-06, + "loss": 0.2255, + "step": 16030 + }, + { + "epoch": 2.92, + "learning_rate": 2.1273571168547507e-06, + "loss": 1.3101, + "step": 16031 + }, + { + "epoch": 2.92, + "learning_rate": 2.1273025118310882e-06, + "loss": 0.3114, + "step": 16032 + }, + { + "epoch": 2.92, + "learning_rate": 2.127247906807426e-06, + "loss": 0.762, + "step": 16033 + }, + { + "epoch": 2.92, + "learning_rate": 2.127193301783764e-06, + "loss": 0.0214, + "step": 16034 + }, + { + "epoch": 2.92, + "learning_rate": 2.127138696760102e-06, + "loss": 0.0236, + "step": 16035 + }, + { + "epoch": 2.92, + "learning_rate": 2.12708409173644e-06, + "loss": 0.2616, + "step": 16036 + }, + { + "epoch": 2.92, + "learning_rate": 2.127029486712778e-06, + "loss": 0.0193, + "step": 16037 + }, + { + "epoch": 2.92, + "learning_rate": 2.1269748816891155e-06, + "loss": 1.7063, + "step": 16038 + }, + { + "epoch": 2.92, + "learning_rate": 2.1269202766654534e-06, + "loss": 0.5912, + "step": 16039 + }, + { + "epoch": 2.92, + "learning_rate": 2.1268656716417914e-06, + "loss": 0.5773, + "step": 16040 + }, + { + "epoch": 2.92, + "learning_rate": 2.126811066618129e-06, + "loss": 1.8257, + "step": 16041 + }, + { + "epoch": 2.92, + "learning_rate": 2.126756461594467e-06, + "loss": 1.4559, + "step": 16042 + }, + { + "epoch": 2.92, + "learning_rate": 2.1267018565708043e-06, + "loss": 1.8035, + "step": 16043 + }, + { + "epoch": 2.92, + "learning_rate": 2.1266472515471423e-06, + "loss": 0.3451, + "step": 16044 + }, + { + "epoch": 2.92, + "learning_rate": 2.1265926465234802e-06, + "loss": 0.9562, + "step": 16045 + }, + { + "epoch": 2.92, + "learning_rate": 2.1265380414998178e-06, + "loss": 1.3741, + "step": 16046 + }, + { + "epoch": 2.92, + "learning_rate": 2.1264834364761557e-06, + "loss": 2.793, + "step": 16047 + }, + { + "epoch": 2.92, + "learning_rate": 2.1264288314524936e-06, + "loss": 1.4513, + "step": 16048 + }, + { + "epoch": 2.92, + "learning_rate": 2.1263742264288316e-06, + "loss": 1.1681, + "step": 16049 + }, + { + "epoch": 2.92, + "learning_rate": 2.1263196214051695e-06, + "loss": 3.2887, + "step": 16050 + }, + { + "epoch": 2.92, + "learning_rate": 2.1262650163815075e-06, + "loss": 0.2401, + "step": 16051 + }, + { + "epoch": 2.92, + "learning_rate": 2.126210411357845e-06, + "loss": 2.0077, + "step": 16052 + }, + { + "epoch": 2.92, + "learning_rate": 2.126155806334183e-06, + "loss": 0.3925, + "step": 16053 + }, + { + "epoch": 2.92, + "learning_rate": 2.126101201310521e-06, + "loss": 0.5219, + "step": 16054 + }, + { + "epoch": 2.92, + "learning_rate": 2.1260465962868584e-06, + "loss": 1.3515, + "step": 16055 + }, + { + "epoch": 2.92, + "learning_rate": 2.1259919912631963e-06, + "loss": 0.6089, + "step": 16056 + }, + { + "epoch": 2.92, + "learning_rate": 2.125937386239534e-06, + "loss": 0.3957, + "step": 16057 + }, + { + "epoch": 2.92, + "learning_rate": 2.125882781215872e-06, + "loss": 0.286, + "step": 16058 + }, + { + "epoch": 2.92, + "learning_rate": 2.1258281761922097e-06, + "loss": 0.4791, + "step": 16059 + }, + { + "epoch": 2.92, + "learning_rate": 2.1257735711685473e-06, + "loss": 0.4282, + "step": 16060 + }, + { + "epoch": 2.92, + "learning_rate": 2.125718966144885e-06, + "loss": 0.0291, + "step": 16061 + }, + { + "epoch": 2.92, + "learning_rate": 2.125664361121223e-06, + "loss": 0.9105, + "step": 16062 + }, + { + "epoch": 2.92, + "learning_rate": 2.125609756097561e-06, + "loss": 0.5796, + "step": 16063 + }, + { + "epoch": 2.92, + "learning_rate": 2.125555151073899e-06, + "loss": 0.6941, + "step": 16064 + }, + { + "epoch": 2.92, + "learning_rate": 2.125500546050237e-06, + "loss": 0.1348, + "step": 16065 + }, + { + "epoch": 2.92, + "learning_rate": 2.1254459410265745e-06, + "loss": 0.6702, + "step": 16066 + }, + { + "epoch": 2.92, + "learning_rate": 2.1253913360029124e-06, + "loss": 2.7565, + "step": 16067 + }, + { + "epoch": 2.92, + "learning_rate": 2.1253367309792504e-06, + "loss": 0.2031, + "step": 16068 + }, + { + "epoch": 2.92, + "learning_rate": 2.125282125955588e-06, + "loss": 0.8585, + "step": 16069 + }, + { + "epoch": 2.92, + "learning_rate": 2.125227520931926e-06, + "loss": 2.2137, + "step": 16070 + }, + { + "epoch": 2.92, + "learning_rate": 2.1251729159082634e-06, + "loss": 1.8372, + "step": 16071 + }, + { + "epoch": 2.92, + "learning_rate": 2.1251183108846013e-06, + "loss": 0.2521, + "step": 16072 + }, + { + "epoch": 2.92, + "learning_rate": 2.1250637058609393e-06, + "loss": 0.5378, + "step": 16073 + }, + { + "epoch": 2.92, + "learning_rate": 2.1250091008372768e-06, + "loss": 0.2655, + "step": 16074 + }, + { + "epoch": 2.92, + "learning_rate": 2.1249544958136147e-06, + "loss": 1.9467, + "step": 16075 + }, + { + "epoch": 2.92, + "learning_rate": 2.1248998907899527e-06, + "loss": 0.0645, + "step": 16076 + }, + { + "epoch": 2.92, + "learning_rate": 2.1248452857662906e-06, + "loss": 0.8542, + "step": 16077 + }, + { + "epoch": 2.92, + "learning_rate": 2.1247906807426286e-06, + "loss": 1.3611, + "step": 16078 + }, + { + "epoch": 2.92, + "learning_rate": 2.1247360757189665e-06, + "loss": 2.3206, + "step": 16079 + }, + { + "epoch": 2.92, + "learning_rate": 2.124681470695304e-06, + "loss": 0.3912, + "step": 16080 + }, + { + "epoch": 2.92, + "learning_rate": 2.124626865671642e-06, + "loss": 1.5107, + "step": 16081 + }, + { + "epoch": 2.92, + "learning_rate": 2.12457226064798e-06, + "loss": 1.111, + "step": 16082 + }, + { + "epoch": 2.92, + "learning_rate": 2.1245176556243174e-06, + "loss": 0.0252, + "step": 16083 + }, + { + "epoch": 2.92, + "learning_rate": 2.1244630506006554e-06, + "loss": 0.5676, + "step": 16084 + }, + { + "epoch": 2.93, + "learning_rate": 2.124408445576993e-06, + "loss": 0.0117, + "step": 16085 + }, + { + "epoch": 2.93, + "learning_rate": 2.124353840553331e-06, + "loss": 0.3489, + "step": 16086 + }, + { + "epoch": 2.93, + "learning_rate": 2.1242992355296688e-06, + "loss": 1.686, + "step": 16087 + }, + { + "epoch": 2.93, + "learning_rate": 2.1242446305060063e-06, + "loss": 0.3261, + "step": 16088 + }, + { + "epoch": 2.93, + "learning_rate": 2.1241900254823442e-06, + "loss": 2.6577, + "step": 16089 + }, + { + "epoch": 2.93, + "learning_rate": 2.124135420458682e-06, + "loss": 0.3652, + "step": 16090 + }, + { + "epoch": 2.93, + "learning_rate": 2.12408081543502e-06, + "loss": 1.1751, + "step": 16091 + }, + { + "epoch": 2.93, + "learning_rate": 2.124026210411358e-06, + "loss": 0.0845, + "step": 16092 + }, + { + "epoch": 2.93, + "learning_rate": 2.123971605387696e-06, + "loss": 0.4614, + "step": 16093 + }, + { + "epoch": 2.93, + "learning_rate": 2.1239170003640335e-06, + "loss": 0.0339, + "step": 16094 + }, + { + "epoch": 2.93, + "learning_rate": 2.1238623953403715e-06, + "loss": 0.9004, + "step": 16095 + }, + { + "epoch": 2.93, + "learning_rate": 2.1238077903167094e-06, + "loss": 0.9966, + "step": 16096 + }, + { + "epoch": 2.93, + "learning_rate": 2.123753185293047e-06, + "loss": 0.0357, + "step": 16097 + }, + { + "epoch": 2.93, + "learning_rate": 2.123698580269385e-06, + "loss": 0.6081, + "step": 16098 + }, + { + "epoch": 2.93, + "learning_rate": 2.1236439752457224e-06, + "loss": 1.684, + "step": 16099 + }, + { + "epoch": 2.93, + "learning_rate": 2.1235893702220603e-06, + "loss": 1.0006, + "step": 16100 + }, + { + "epoch": 2.93, + "learning_rate": 2.1235347651983983e-06, + "loss": 0.3827, + "step": 16101 + }, + { + "epoch": 2.93, + "learning_rate": 2.123480160174736e-06, + "loss": 0.1466, + "step": 16102 + }, + { + "epoch": 2.93, + "learning_rate": 2.1234255551510738e-06, + "loss": 0.0165, + "step": 16103 + }, + { + "epoch": 2.93, + "learning_rate": 2.1233709501274117e-06, + "loss": 0.5801, + "step": 16104 + }, + { + "epoch": 2.93, + "learning_rate": 2.1233163451037496e-06, + "loss": 1.8877, + "step": 16105 + }, + { + "epoch": 2.93, + "learning_rate": 2.1232617400800876e-06, + "loss": 1.551, + "step": 16106 + }, + { + "epoch": 2.93, + "learning_rate": 2.1232071350564255e-06, + "loss": 1.6813, + "step": 16107 + }, + { + "epoch": 2.93, + "learning_rate": 2.123152530032763e-06, + "loss": 1.5905, + "step": 16108 + }, + { + "epoch": 2.93, + "learning_rate": 2.123097925009101e-06, + "loss": 0.0116, + "step": 16109 + }, + { + "epoch": 2.93, + "learning_rate": 2.123043319985439e-06, + "loss": 0.3587, + "step": 16110 + }, + { + "epoch": 2.93, + "learning_rate": 2.1229887149617765e-06, + "loss": 0.7394, + "step": 16111 + }, + { + "epoch": 2.93, + "learning_rate": 2.1229341099381144e-06, + "loss": 0.6574, + "step": 16112 + }, + { + "epoch": 2.93, + "learning_rate": 2.122879504914452e-06, + "loss": 0.0185, + "step": 16113 + }, + { + "epoch": 2.93, + "learning_rate": 2.12282489989079e-06, + "loss": 0.5909, + "step": 16114 + }, + { + "epoch": 2.93, + "learning_rate": 2.122770294867128e-06, + "loss": 0.0397, + "step": 16115 + }, + { + "epoch": 2.93, + "learning_rate": 2.1227156898434657e-06, + "loss": 1.7466, + "step": 16116 + }, + { + "epoch": 2.93, + "learning_rate": 2.1226610848198037e-06, + "loss": 0.9818, + "step": 16117 + }, + { + "epoch": 2.93, + "learning_rate": 2.122606479796141e-06, + "loss": 1.0149, + "step": 16118 + }, + { + "epoch": 2.93, + "learning_rate": 2.122551874772479e-06, + "loss": 0.6587, + "step": 16119 + }, + { + "epoch": 2.93, + "learning_rate": 2.122497269748817e-06, + "loss": 1.152, + "step": 16120 + }, + { + "epoch": 2.93, + "learning_rate": 2.122442664725155e-06, + "loss": 0.2733, + "step": 16121 + }, + { + "epoch": 2.93, + "learning_rate": 2.1223880597014926e-06, + "loss": 1.0009, + "step": 16122 + }, + { + "epoch": 2.93, + "learning_rate": 2.1223334546778305e-06, + "loss": 0.9531, + "step": 16123 + }, + { + "epoch": 2.93, + "learning_rate": 2.122278849654168e-06, + "loss": 0.9985, + "step": 16124 + }, + { + "epoch": 2.93, + "learning_rate": 2.122224244630506e-06, + "loss": 2.977, + "step": 16125 + }, + { + "epoch": 2.93, + "learning_rate": 2.122169639606844e-06, + "loss": 1.5821, + "step": 16126 + }, + { + "epoch": 2.93, + "learning_rate": 2.1221150345831814e-06, + "loss": 0.2806, + "step": 16127 + }, + { + "epoch": 2.93, + "learning_rate": 2.1220604295595194e-06, + "loss": 0.3799, + "step": 16128 + }, + { + "epoch": 2.93, + "learning_rate": 2.1220058245358573e-06, + "loss": 0.614, + "step": 16129 + }, + { + "epoch": 2.93, + "learning_rate": 2.1219512195121953e-06, + "loss": 0.3326, + "step": 16130 + }, + { + "epoch": 2.93, + "learning_rate": 2.121896614488533e-06, + "loss": 0.4227, + "step": 16131 + }, + { + "epoch": 2.93, + "learning_rate": 2.121842009464871e-06, + "loss": 1.3563, + "step": 16132 + }, + { + "epoch": 2.93, + "learning_rate": 2.1217874044412087e-06, + "loss": 0.3473, + "step": 16133 + }, + { + "epoch": 2.93, + "learning_rate": 2.1217327994175466e-06, + "loss": 0.0682, + "step": 16134 + }, + { + "epoch": 2.93, + "learning_rate": 2.1216781943938846e-06, + "loss": 1.6544, + "step": 16135 + }, + { + "epoch": 2.93, + "learning_rate": 2.121623589370222e-06, + "loss": 0.0272, + "step": 16136 + }, + { + "epoch": 2.93, + "learning_rate": 2.12156898434656e-06, + "loss": 1.3394, + "step": 16137 + }, + { + "epoch": 2.93, + "learning_rate": 2.1215143793228975e-06, + "loss": 0.1646, + "step": 16138 + }, + { + "epoch": 2.93, + "learning_rate": 2.1214597742992355e-06, + "loss": 0.1332, + "step": 16139 + }, + { + "epoch": 2.94, + "learning_rate": 2.1214051692755734e-06, + "loss": 0.0536, + "step": 16140 + }, + { + "epoch": 2.94, + "learning_rate": 2.121350564251911e-06, + "loss": 1.7428, + "step": 16141 + }, + { + "epoch": 2.94, + "learning_rate": 2.121295959228249e-06, + "loss": 0.9846, + "step": 16142 + }, + { + "epoch": 2.94, + "learning_rate": 2.121241354204587e-06, + "loss": 0.1617, + "step": 16143 + }, + { + "epoch": 2.94, + "learning_rate": 2.1211867491809248e-06, + "loss": 0.2636, + "step": 16144 + }, + { + "epoch": 2.94, + "learning_rate": 2.1211321441572627e-06, + "loss": 0.118, + "step": 16145 + }, + { + "epoch": 2.94, + "learning_rate": 2.1210775391336007e-06, + "loss": 0.394, + "step": 16146 + }, + { + "epoch": 2.94, + "learning_rate": 2.121022934109938e-06, + "loss": 2.9799, + "step": 16147 + }, + { + "epoch": 2.94, + "learning_rate": 2.120968329086276e-06, + "loss": 1.3818, + "step": 16148 + }, + { + "epoch": 2.94, + "learning_rate": 2.120913724062614e-06, + "loss": 0.2488, + "step": 16149 + }, + { + "epoch": 2.94, + "learning_rate": 2.1208591190389516e-06, + "loss": 0.5094, + "step": 16150 + }, + { + "epoch": 2.94, + "learning_rate": 2.1208045140152895e-06, + "loss": 0.8396, + "step": 16151 + }, + { + "epoch": 2.94, + "learning_rate": 2.120749908991627e-06, + "loss": 1.4791, + "step": 16152 + }, + { + "epoch": 2.94, + "learning_rate": 2.120695303967965e-06, + "loss": 0.0433, + "step": 16153 + }, + { + "epoch": 2.94, + "learning_rate": 2.120640698944303e-06, + "loss": 0.2516, + "step": 16154 + }, + { + "epoch": 2.94, + "learning_rate": 2.1205860939206405e-06, + "loss": 0.6431, + "step": 16155 + }, + { + "epoch": 2.94, + "learning_rate": 2.1205314888969784e-06, + "loss": 1.6002, + "step": 16156 + }, + { + "epoch": 2.94, + "learning_rate": 2.1204768838733163e-06, + "loss": 1.8589, + "step": 16157 + }, + { + "epoch": 2.94, + "learning_rate": 2.1204222788496543e-06, + "loss": 1.5808, + "step": 16158 + }, + { + "epoch": 2.94, + "learning_rate": 2.1203676738259922e-06, + "loss": 0.2238, + "step": 16159 + }, + { + "epoch": 2.94, + "learning_rate": 2.12031306880233e-06, + "loss": 1.6026, + "step": 16160 + }, + { + "epoch": 2.94, + "learning_rate": 2.1202584637786677e-06, + "loss": 0.3137, + "step": 16161 + }, + { + "epoch": 2.94, + "learning_rate": 2.1202038587550056e-06, + "loss": 0.1046, + "step": 16162 + }, + { + "epoch": 2.94, + "learning_rate": 2.1201492537313436e-06, + "loss": 2.6841, + "step": 16163 + }, + { + "epoch": 2.94, + "learning_rate": 2.120094648707681e-06, + "loss": 0.0491, + "step": 16164 + }, + { + "epoch": 2.94, + "learning_rate": 2.120040043684019e-06, + "loss": 1.1127, + "step": 16165 + }, + { + "epoch": 2.94, + "learning_rate": 2.1199854386603566e-06, + "loss": 0.0179, + "step": 16166 + }, + { + "epoch": 2.94, + "learning_rate": 2.1199308336366945e-06, + "loss": 0.2081, + "step": 16167 + }, + { + "epoch": 2.94, + "learning_rate": 2.1198762286130324e-06, + "loss": 0.1731, + "step": 16168 + }, + { + "epoch": 2.94, + "learning_rate": 2.11982162358937e-06, + "loss": 2.4674, + "step": 16169 + }, + { + "epoch": 2.94, + "learning_rate": 2.119767018565708e-06, + "loss": 0.6865, + "step": 16170 + }, + { + "epoch": 2.94, + "learning_rate": 2.119712413542046e-06, + "loss": 0.8004, + "step": 16171 + }, + { + "epoch": 2.94, + "learning_rate": 2.119657808518384e-06, + "loss": 0.2283, + "step": 16172 + }, + { + "epoch": 2.94, + "learning_rate": 2.1196032034947217e-06, + "loss": 0.1204, + "step": 16173 + }, + { + "epoch": 2.94, + "learning_rate": 2.1195485984710597e-06, + "loss": 0.9696, + "step": 16174 + }, + { + "epoch": 2.94, + "learning_rate": 2.119493993447397e-06, + "loss": 1.413, + "step": 16175 + }, + { + "epoch": 2.94, + "learning_rate": 2.119439388423735e-06, + "loss": 0.5657, + "step": 16176 + }, + { + "epoch": 2.94, + "learning_rate": 2.119384783400073e-06, + "loss": 1.3658, + "step": 16177 + }, + { + "epoch": 2.94, + "learning_rate": 2.1193301783764106e-06, + "loss": 0.6554, + "step": 16178 + }, + { + "epoch": 2.94, + "learning_rate": 2.1192755733527486e-06, + "loss": 0.3803, + "step": 16179 + }, + { + "epoch": 2.94, + "learning_rate": 2.119220968329086e-06, + "loss": 0.9945, + "step": 16180 + }, + { + "epoch": 2.94, + "learning_rate": 2.119166363305424e-06, + "loss": 3.0804, + "step": 16181 + }, + { + "epoch": 2.94, + "learning_rate": 2.119111758281762e-06, + "loss": 0.8949, + "step": 16182 + }, + { + "epoch": 2.94, + "learning_rate": 2.1190571532580995e-06, + "loss": 1.9129, + "step": 16183 + }, + { + "epoch": 2.94, + "learning_rate": 2.1190025482344374e-06, + "loss": 0.014, + "step": 16184 + }, + { + "epoch": 2.94, + "learning_rate": 2.1189479432107754e-06, + "loss": 0.0722, + "step": 16185 + }, + { + "epoch": 2.94, + "learning_rate": 2.1188933381871133e-06, + "loss": 0.6526, + "step": 16186 + }, + { + "epoch": 2.94, + "learning_rate": 2.1188387331634513e-06, + "loss": 1.9844, + "step": 16187 + }, + { + "epoch": 2.94, + "learning_rate": 2.118784128139789e-06, + "loss": 0.0135, + "step": 16188 + }, + { + "epoch": 2.94, + "learning_rate": 2.1187295231161267e-06, + "loss": 0.9219, + "step": 16189 + }, + { + "epoch": 2.94, + "learning_rate": 2.1186749180924647e-06, + "loss": 0.9245, + "step": 16190 + }, + { + "epoch": 2.94, + "learning_rate": 2.118620313068802e-06, + "loss": 1.4704, + "step": 16191 + }, + { + "epoch": 2.94, + "learning_rate": 2.11856570804514e-06, + "loss": 0.304, + "step": 16192 + }, + { + "epoch": 2.94, + "learning_rate": 2.118511103021478e-06, + "loss": 0.4175, + "step": 16193 + }, + { + "epoch": 2.94, + "learning_rate": 2.1184564979978156e-06, + "loss": 1.0278, + "step": 16194 + }, + { + "epoch": 2.95, + "learning_rate": 2.1184018929741535e-06, + "loss": 0.0856, + "step": 16195 + }, + { + "epoch": 2.95, + "learning_rate": 2.1183472879504915e-06, + "loss": 0.9315, + "step": 16196 + }, + { + "epoch": 2.95, + "learning_rate": 2.118292682926829e-06, + "loss": 0.017, + "step": 16197 + }, + { + "epoch": 2.95, + "learning_rate": 2.118238077903167e-06, + "loss": 1.3201, + "step": 16198 + }, + { + "epoch": 2.95, + "learning_rate": 2.118183472879505e-06, + "loss": 1.224, + "step": 16199 + }, + { + "epoch": 2.95, + "learning_rate": 2.118128867855843e-06, + "loss": 0.5765, + "step": 16200 + }, + { + "epoch": 2.95, + "learning_rate": 2.1180742628321808e-06, + "loss": 1.2187, + "step": 16201 + }, + { + "epoch": 2.95, + "learning_rate": 2.1180196578085187e-06, + "loss": 3.5834, + "step": 16202 + }, + { + "epoch": 2.95, + "learning_rate": 2.1179650527848562e-06, + "loss": 1.4184, + "step": 16203 + }, + { + "epoch": 2.95, + "learning_rate": 2.117910447761194e-06, + "loss": 1.423, + "step": 16204 + }, + { + "epoch": 2.95, + "learning_rate": 2.1178558427375317e-06, + "loss": 2.2931, + "step": 16205 + }, + { + "epoch": 2.95, + "learning_rate": 2.1178012377138696e-06, + "loss": 0.2361, + "step": 16206 + }, + { + "epoch": 2.95, + "learning_rate": 2.1177466326902076e-06, + "loss": 0.9531, + "step": 16207 + }, + { + "epoch": 2.95, + "learning_rate": 2.117692027666545e-06, + "loss": 0.6456, + "step": 16208 + }, + { + "epoch": 2.95, + "learning_rate": 2.117637422642883e-06, + "loss": 0.7068, + "step": 16209 + }, + { + "epoch": 2.95, + "learning_rate": 2.117582817619221e-06, + "loss": 0.9045, + "step": 16210 + }, + { + "epoch": 2.95, + "learning_rate": 2.117528212595559e-06, + "loss": 1.0601, + "step": 16211 + }, + { + "epoch": 2.95, + "learning_rate": 2.117473607571897e-06, + "loss": 0.7402, + "step": 16212 + }, + { + "epoch": 2.95, + "learning_rate": 2.117419002548235e-06, + "loss": 0.0343, + "step": 16213 + }, + { + "epoch": 2.95, + "learning_rate": 2.1173643975245723e-06, + "loss": 0.3478, + "step": 16214 + }, + { + "epoch": 2.95, + "learning_rate": 2.1173097925009103e-06, + "loss": 0.9427, + "step": 16215 + }, + { + "epoch": 2.95, + "learning_rate": 2.1172551874772482e-06, + "loss": 0.851, + "step": 16216 + }, + { + "epoch": 2.95, + "learning_rate": 2.1172005824535857e-06, + "loss": 0.6699, + "step": 16217 + }, + { + "epoch": 2.95, + "learning_rate": 2.1171459774299237e-06, + "loss": 0.3592, + "step": 16218 + }, + { + "epoch": 2.95, + "learning_rate": 2.117091372406261e-06, + "loss": 0.5029, + "step": 16219 + }, + { + "epoch": 2.95, + "learning_rate": 2.117036767382599e-06, + "loss": 0.0515, + "step": 16220 + }, + { + "epoch": 2.95, + "learning_rate": 2.116982162358937e-06, + "loss": 0.2558, + "step": 16221 + }, + { + "epoch": 2.95, + "learning_rate": 2.1169275573352746e-06, + "loss": 0.1489, + "step": 16222 + }, + { + "epoch": 2.95, + "learning_rate": 2.1168729523116126e-06, + "loss": 0.7698, + "step": 16223 + }, + { + "epoch": 2.95, + "learning_rate": 2.1168183472879505e-06, + "loss": 0.3521, + "step": 16224 + }, + { + "epoch": 2.95, + "learning_rate": 2.1167637422642884e-06, + "loss": 0.0563, + "step": 16225 + }, + { + "epoch": 2.95, + "learning_rate": 2.1167091372406264e-06, + "loss": 0.4516, + "step": 16226 + }, + { + "epoch": 2.95, + "learning_rate": 2.1166545322169643e-06, + "loss": 0.3477, + "step": 16227 + }, + { + "epoch": 2.95, + "learning_rate": 2.116599927193302e-06, + "loss": 0.1741, + "step": 16228 + }, + { + "epoch": 2.95, + "learning_rate": 2.11654532216964e-06, + "loss": 0.6232, + "step": 16229 + }, + { + "epoch": 2.95, + "learning_rate": 2.1164907171459777e-06, + "loss": 0.1758, + "step": 16230 + }, + { + "epoch": 2.95, + "learning_rate": 2.1164361121223153e-06, + "loss": 0.619, + "step": 16231 + }, + { + "epoch": 2.95, + "learning_rate": 2.116381507098653e-06, + "loss": 0.4617, + "step": 16232 + }, + { + "epoch": 2.95, + "learning_rate": 2.1163269020749907e-06, + "loss": 0.0198, + "step": 16233 + }, + { + "epoch": 2.95, + "learning_rate": 2.1162722970513287e-06, + "loss": 1.2904, + "step": 16234 + }, + { + "epoch": 2.95, + "learning_rate": 2.1162176920276666e-06, + "loss": 1.0748, + "step": 16235 + }, + { + "epoch": 2.95, + "learning_rate": 2.116163087004004e-06, + "loss": 1.8118, + "step": 16236 + }, + { + "epoch": 2.95, + "learning_rate": 2.116108481980342e-06, + "loss": 1.5475, + "step": 16237 + }, + { + "epoch": 2.95, + "learning_rate": 2.11605387695668e-06, + "loss": 1.7081, + "step": 16238 + }, + { + "epoch": 2.95, + "learning_rate": 2.115999271933018e-06, + "loss": 2.4875, + "step": 16239 + }, + { + "epoch": 2.95, + "learning_rate": 2.115944666909356e-06, + "loss": 1.0209, + "step": 16240 + }, + { + "epoch": 2.95, + "learning_rate": 2.115890061885694e-06, + "loss": 0.7337, + "step": 16241 + }, + { + "epoch": 2.95, + "learning_rate": 2.1158354568620314e-06, + "loss": 0.1747, + "step": 16242 + }, + { + "epoch": 2.95, + "learning_rate": 2.1157808518383693e-06, + "loss": 0.4683, + "step": 16243 + }, + { + "epoch": 2.95, + "learning_rate": 2.1157262468147073e-06, + "loss": 0.5323, + "step": 16244 + }, + { + "epoch": 2.95, + "learning_rate": 2.1156716417910448e-06, + "loss": 0.1448, + "step": 16245 + }, + { + "epoch": 2.95, + "learning_rate": 2.1156170367673827e-06, + "loss": 0.5304, + "step": 16246 + }, + { + "epoch": 2.95, + "learning_rate": 2.1155624317437202e-06, + "loss": 0.3441, + "step": 16247 + }, + { + "epoch": 2.95, + "learning_rate": 2.115507826720058e-06, + "loss": 0.1433, + "step": 16248 + }, + { + "epoch": 2.95, + "learning_rate": 2.115453221696396e-06, + "loss": 0.4266, + "step": 16249 + }, + { + "epoch": 2.96, + "learning_rate": 2.1153986166727336e-06, + "loss": 0.105, + "step": 16250 + }, + { + "epoch": 2.96, + "learning_rate": 2.1153440116490716e-06, + "loss": 1.0309, + "step": 16251 + }, + { + "epoch": 2.96, + "learning_rate": 2.1152894066254095e-06, + "loss": 0.3627, + "step": 16252 + }, + { + "epoch": 2.96, + "learning_rate": 2.1152348016017475e-06, + "loss": 2.6863, + "step": 16253 + }, + { + "epoch": 2.96, + "learning_rate": 2.1151801965780854e-06, + "loss": 2.3131, + "step": 16254 + }, + { + "epoch": 2.96, + "learning_rate": 2.1151255915544234e-06, + "loss": 0.2631, + "step": 16255 + }, + { + "epoch": 2.96, + "learning_rate": 2.115070986530761e-06, + "loss": 0.4857, + "step": 16256 + }, + { + "epoch": 2.96, + "learning_rate": 2.115016381507099e-06, + "loss": 0.5883, + "step": 16257 + }, + { + "epoch": 2.96, + "learning_rate": 2.1149617764834368e-06, + "loss": 0.014, + "step": 16258 + }, + { + "epoch": 2.96, + "learning_rate": 2.1149071714597743e-06, + "loss": 0.6522, + "step": 16259 + }, + { + "epoch": 2.96, + "learning_rate": 2.1148525664361122e-06, + "loss": 0.5249, + "step": 16260 + }, + { + "epoch": 2.96, + "learning_rate": 2.1147979614124497e-06, + "loss": 0.0253, + "step": 16261 + }, + { + "epoch": 2.96, + "learning_rate": 2.1147433563887877e-06, + "loss": 1.0271, + "step": 16262 + }, + { + "epoch": 2.96, + "learning_rate": 2.1146887513651256e-06, + "loss": 0.4121, + "step": 16263 + }, + { + "epoch": 2.96, + "learning_rate": 2.114634146341463e-06, + "loss": 1.3646, + "step": 16264 + }, + { + "epoch": 2.96, + "learning_rate": 2.114579541317801e-06, + "loss": 0.3101, + "step": 16265 + }, + { + "epoch": 2.96, + "learning_rate": 2.114524936294139e-06, + "loss": 0.481, + "step": 16266 + }, + { + "epoch": 2.96, + "learning_rate": 2.114470331270477e-06, + "loss": 2.7775, + "step": 16267 + }, + { + "epoch": 2.96, + "learning_rate": 2.114415726246815e-06, + "loss": 2.6004, + "step": 16268 + }, + { + "epoch": 2.96, + "learning_rate": 2.114361121223153e-06, + "loss": 0.7763, + "step": 16269 + }, + { + "epoch": 2.96, + "learning_rate": 2.1143065161994904e-06, + "loss": 0.162, + "step": 16270 + }, + { + "epoch": 2.96, + "learning_rate": 2.1142519111758283e-06, + "loss": 0.4838, + "step": 16271 + }, + { + "epoch": 2.96, + "learning_rate": 2.114197306152166e-06, + "loss": 0.5893, + "step": 16272 + }, + { + "epoch": 2.96, + "learning_rate": 2.114142701128504e-06, + "loss": 1.8431, + "step": 16273 + }, + { + "epoch": 2.96, + "learning_rate": 2.1140880961048417e-06, + "loss": 0.0576, + "step": 16274 + }, + { + "epoch": 2.96, + "learning_rate": 2.1140334910811793e-06, + "loss": 0.3769, + "step": 16275 + }, + { + "epoch": 2.96, + "learning_rate": 2.113978886057517e-06, + "loss": 0.1457, + "step": 16276 + }, + { + "epoch": 2.96, + "learning_rate": 2.113924281033855e-06, + "loss": 0.3774, + "step": 16277 + }, + { + "epoch": 2.96, + "learning_rate": 2.1138696760101927e-06, + "loss": 0.0257, + "step": 16278 + }, + { + "epoch": 2.96, + "learning_rate": 2.1138150709865306e-06, + "loss": 0.0124, + "step": 16279 + }, + { + "epoch": 2.96, + "learning_rate": 2.1137604659628686e-06, + "loss": 0.0493, + "step": 16280 + }, + { + "epoch": 2.96, + "learning_rate": 2.1137058609392065e-06, + "loss": 1.0391, + "step": 16281 + }, + { + "epoch": 2.96, + "learning_rate": 2.1136512559155444e-06, + "loss": 0.2397, + "step": 16282 + }, + { + "epoch": 2.96, + "learning_rate": 2.1135966508918824e-06, + "loss": 0.2207, + "step": 16283 + }, + { + "epoch": 2.96, + "learning_rate": 2.11354204586822e-06, + "loss": 2.0207, + "step": 16284 + }, + { + "epoch": 2.96, + "learning_rate": 2.113487440844558e-06, + "loss": 0.0852, + "step": 16285 + }, + { + "epoch": 2.96, + "learning_rate": 2.1134328358208954e-06, + "loss": 0.195, + "step": 16286 + }, + { + "epoch": 2.96, + "learning_rate": 2.1133782307972333e-06, + "loss": 0.0941, + "step": 16287 + }, + { + "epoch": 2.96, + "learning_rate": 2.1133236257735713e-06, + "loss": 0.4361, + "step": 16288 + }, + { + "epoch": 2.96, + "learning_rate": 2.1132690207499088e-06, + "loss": 1.0602, + "step": 16289 + }, + { + "epoch": 2.96, + "learning_rate": 2.1132144157262467e-06, + "loss": 2.7622, + "step": 16290 + }, + { + "epoch": 2.96, + "learning_rate": 2.1131598107025847e-06, + "loss": 0.4784, + "step": 16291 + }, + { + "epoch": 2.96, + "learning_rate": 2.1131052056789226e-06, + "loss": 0.5765, + "step": 16292 + }, + { + "epoch": 2.96, + "learning_rate": 2.1130506006552606e-06, + "loss": 1.1295, + "step": 16293 + }, + { + "epoch": 2.96, + "learning_rate": 2.1129959956315985e-06, + "loss": 0.0596, + "step": 16294 + }, + { + "epoch": 2.96, + "learning_rate": 2.112941390607936e-06, + "loss": 1.109, + "step": 16295 + }, + { + "epoch": 2.96, + "learning_rate": 2.112886785584274e-06, + "loss": 0.9503, + "step": 16296 + }, + { + "epoch": 2.96, + "learning_rate": 2.112832180560612e-06, + "loss": 0.968, + "step": 16297 + }, + { + "epoch": 2.96, + "learning_rate": 2.1127775755369494e-06, + "loss": 0.928, + "step": 16298 + }, + { + "epoch": 2.96, + "learning_rate": 2.1127229705132874e-06, + "loss": 0.2651, + "step": 16299 + }, + { + "epoch": 2.96, + "learning_rate": 2.112668365489625e-06, + "loss": 0.9007, + "step": 16300 + }, + { + "epoch": 2.96, + "learning_rate": 2.112613760465963e-06, + "loss": 2.4521, + "step": 16301 + }, + { + "epoch": 2.96, + "learning_rate": 2.1125591554423008e-06, + "loss": 0.1015, + "step": 16302 + }, + { + "epoch": 2.96, + "learning_rate": 2.1125045504186383e-06, + "loss": 1.2333, + "step": 16303 + }, + { + "epoch": 2.96, + "learning_rate": 2.1124499453949762e-06, + "loss": 1.3793, + "step": 16304 + }, + { + "epoch": 2.97, + "learning_rate": 2.112395340371314e-06, + "loss": 0.1108, + "step": 16305 + }, + { + "epoch": 2.97, + "learning_rate": 2.112340735347652e-06, + "loss": 0.3405, + "step": 16306 + }, + { + "epoch": 2.97, + "learning_rate": 2.11228613032399e-06, + "loss": 0.0373, + "step": 16307 + }, + { + "epoch": 2.97, + "learning_rate": 2.112231525300328e-06, + "loss": 1.2049, + "step": 16308 + }, + { + "epoch": 2.97, + "learning_rate": 2.1121769202766655e-06, + "loss": 1.636, + "step": 16309 + }, + { + "epoch": 2.97, + "learning_rate": 2.1121223152530035e-06, + "loss": 1.0538, + "step": 16310 + }, + { + "epoch": 2.97, + "learning_rate": 2.1120677102293414e-06, + "loss": 0.8739, + "step": 16311 + }, + { + "epoch": 2.97, + "learning_rate": 2.112013105205679e-06, + "loss": 0.0582, + "step": 16312 + }, + { + "epoch": 2.97, + "learning_rate": 2.111958500182017e-06, + "loss": 1.1721, + "step": 16313 + }, + { + "epoch": 2.97, + "learning_rate": 2.1119038951583544e-06, + "loss": 1.069, + "step": 16314 + }, + { + "epoch": 2.97, + "learning_rate": 2.1118492901346923e-06, + "loss": 0.4593, + "step": 16315 + }, + { + "epoch": 2.97, + "learning_rate": 2.1117946851110303e-06, + "loss": 1.3562, + "step": 16316 + }, + { + "epoch": 2.97, + "learning_rate": 2.111740080087368e-06, + "loss": 1.8122, + "step": 16317 + }, + { + "epoch": 2.97, + "learning_rate": 2.1116854750637057e-06, + "loss": 1.3074, + "step": 16318 + }, + { + "epoch": 2.97, + "learning_rate": 2.1116308700400437e-06, + "loss": 0.4591, + "step": 16319 + }, + { + "epoch": 2.97, + "learning_rate": 2.1115762650163816e-06, + "loss": 0.0064, + "step": 16320 + }, + { + "epoch": 2.97, + "learning_rate": 2.1115216599927196e-06, + "loss": 2.0729, + "step": 16321 + }, + { + "epoch": 2.97, + "learning_rate": 2.1114670549690575e-06, + "loss": 0.1255, + "step": 16322 + }, + { + "epoch": 2.97, + "learning_rate": 2.111412449945395e-06, + "loss": 0.0333, + "step": 16323 + }, + { + "epoch": 2.97, + "learning_rate": 2.111357844921733e-06, + "loss": 1.1599, + "step": 16324 + }, + { + "epoch": 2.97, + "learning_rate": 2.111303239898071e-06, + "loss": 1.6448, + "step": 16325 + }, + { + "epoch": 2.97, + "learning_rate": 2.1112486348744084e-06, + "loss": 3.7061, + "step": 16326 + }, + { + "epoch": 2.97, + "learning_rate": 2.1111940298507464e-06, + "loss": 0.0218, + "step": 16327 + }, + { + "epoch": 2.97, + "learning_rate": 2.111139424827084e-06, + "loss": 1.8863, + "step": 16328 + }, + { + "epoch": 2.97, + "learning_rate": 2.111084819803422e-06, + "loss": 0.6459, + "step": 16329 + }, + { + "epoch": 2.97, + "learning_rate": 2.11103021477976e-06, + "loss": 1.3647, + "step": 16330 + }, + { + "epoch": 2.97, + "learning_rate": 2.1109756097560973e-06, + "loss": 0.0201, + "step": 16331 + }, + { + "epoch": 2.97, + "learning_rate": 2.1109210047324353e-06, + "loss": 3.3393, + "step": 16332 + }, + { + "epoch": 2.97, + "learning_rate": 2.110866399708773e-06, + "loss": 0.1612, + "step": 16333 + }, + { + "epoch": 2.97, + "learning_rate": 2.110811794685111e-06, + "loss": 1.1509, + "step": 16334 + }, + { + "epoch": 2.97, + "learning_rate": 2.110757189661449e-06, + "loss": 2.0674, + "step": 16335 + }, + { + "epoch": 2.97, + "learning_rate": 2.110702584637787e-06, + "loss": 1.4595, + "step": 16336 + }, + { + "epoch": 2.97, + "learning_rate": 2.1106479796141246e-06, + "loss": 0.348, + "step": 16337 + }, + { + "epoch": 2.97, + "learning_rate": 2.1105933745904625e-06, + "loss": 0.7283, + "step": 16338 + }, + { + "epoch": 2.97, + "learning_rate": 2.1105387695668004e-06, + "loss": 0.0666, + "step": 16339 + }, + { + "epoch": 2.97, + "learning_rate": 2.110484164543138e-06, + "loss": 0.3432, + "step": 16340 + }, + { + "epoch": 2.97, + "learning_rate": 2.110429559519476e-06, + "loss": 0.0247, + "step": 16341 + }, + { + "epoch": 2.97, + "learning_rate": 2.1103749544958134e-06, + "loss": 1.1306, + "step": 16342 + }, + { + "epoch": 2.97, + "learning_rate": 2.1103203494721514e-06, + "loss": 0.8553, + "step": 16343 + }, + { + "epoch": 2.97, + "learning_rate": 2.1102657444484893e-06, + "loss": 0.8738, + "step": 16344 + }, + { + "epoch": 2.97, + "learning_rate": 2.110211139424827e-06, + "loss": 1.6688, + "step": 16345 + }, + { + "epoch": 2.97, + "learning_rate": 2.1101565344011648e-06, + "loss": 0.1034, + "step": 16346 + }, + { + "epoch": 2.97, + "learning_rate": 2.1101019293775027e-06, + "loss": 0.5637, + "step": 16347 + }, + { + "epoch": 2.97, + "learning_rate": 2.1100473243538407e-06, + "loss": 0.6346, + "step": 16348 + }, + { + "epoch": 2.97, + "learning_rate": 2.1099927193301786e-06, + "loss": 1.2332, + "step": 16349 + }, + { + "epoch": 2.97, + "learning_rate": 2.1099381143065165e-06, + "loss": 0.1709, + "step": 16350 + }, + { + "epoch": 2.97, + "learning_rate": 2.109883509282854e-06, + "loss": 0.5163, + "step": 16351 + }, + { + "epoch": 2.97, + "learning_rate": 2.109828904259192e-06, + "loss": 0.4444, + "step": 16352 + }, + { + "epoch": 2.97, + "learning_rate": 2.1097742992355295e-06, + "loss": 2.5178, + "step": 16353 + }, + { + "epoch": 2.97, + "learning_rate": 2.1097196942118675e-06, + "loss": 1.4233, + "step": 16354 + }, + { + "epoch": 2.97, + "learning_rate": 2.1096650891882054e-06, + "loss": 0.4199, + "step": 16355 + }, + { + "epoch": 2.97, + "learning_rate": 2.109610484164543e-06, + "loss": 1.1312, + "step": 16356 + }, + { + "epoch": 2.97, + "learning_rate": 2.109555879140881e-06, + "loss": 2.4155, + "step": 16357 + }, + { + "epoch": 2.97, + "learning_rate": 2.109501274117219e-06, + "loss": 1.151, + "step": 16358 + }, + { + "epoch": 2.97, + "learning_rate": 2.1094466690935563e-06, + "loss": 2.7968, + "step": 16359 + }, + { + "epoch": 2.98, + "learning_rate": 2.1093920640698943e-06, + "loss": 2.0319, + "step": 16360 + }, + { + "epoch": 2.98, + "learning_rate": 2.1093374590462322e-06, + "loss": 0.5766, + "step": 16361 + }, + { + "epoch": 2.98, + "learning_rate": 2.10928285402257e-06, + "loss": 0.9419, + "step": 16362 + }, + { + "epoch": 2.98, + "learning_rate": 2.109228248998908e-06, + "loss": 0.7112, + "step": 16363 + }, + { + "epoch": 2.98, + "learning_rate": 2.109173643975246e-06, + "loss": 0.5174, + "step": 16364 + }, + { + "epoch": 2.98, + "learning_rate": 2.1091190389515836e-06, + "loss": 0.0385, + "step": 16365 + }, + { + "epoch": 2.98, + "learning_rate": 2.1090644339279215e-06, + "loss": 0.6629, + "step": 16366 + }, + { + "epoch": 2.98, + "learning_rate": 2.109009828904259e-06, + "loss": 1.0901, + "step": 16367 + }, + { + "epoch": 2.98, + "learning_rate": 2.108955223880597e-06, + "loss": 0.2225, + "step": 16368 + }, + { + "epoch": 2.98, + "learning_rate": 2.108900618856935e-06, + "loss": 0.2158, + "step": 16369 + }, + { + "epoch": 2.98, + "learning_rate": 2.1088460138332725e-06, + "loss": 0.0262, + "step": 16370 + }, + { + "epoch": 2.98, + "learning_rate": 2.1087914088096104e-06, + "loss": 0.3161, + "step": 16371 + }, + { + "epoch": 2.98, + "learning_rate": 2.1087368037859483e-06, + "loss": 0.6571, + "step": 16372 + }, + { + "epoch": 2.98, + "learning_rate": 2.1086821987622863e-06, + "loss": 0.0163, + "step": 16373 + }, + { + "epoch": 2.98, + "learning_rate": 2.1086275937386242e-06, + "loss": 0.5442, + "step": 16374 + }, + { + "epoch": 2.98, + "learning_rate": 2.108572988714962e-06, + "loss": 1.4891, + "step": 16375 + }, + { + "epoch": 2.98, + "learning_rate": 2.1085183836912997e-06, + "loss": 0.2732, + "step": 16376 + }, + { + "epoch": 2.98, + "learning_rate": 2.1084637786676376e-06, + "loss": 0.2927, + "step": 16377 + }, + { + "epoch": 2.98, + "learning_rate": 2.1084091736439756e-06, + "loss": 1.4907, + "step": 16378 + }, + { + "epoch": 2.98, + "learning_rate": 2.108354568620313e-06, + "loss": 0.4761, + "step": 16379 + }, + { + "epoch": 2.98, + "learning_rate": 2.108299963596651e-06, + "loss": 0.7056, + "step": 16380 + }, + { + "epoch": 2.98, + "learning_rate": 2.1082453585729886e-06, + "loss": 2.2448, + "step": 16381 + }, + { + "epoch": 2.98, + "learning_rate": 2.1081907535493265e-06, + "loss": 0.2496, + "step": 16382 + }, + { + "epoch": 2.98, + "learning_rate": 2.1081361485256644e-06, + "loss": 0.6006, + "step": 16383 + }, + { + "epoch": 2.98, + "learning_rate": 2.108081543502002e-06, + "loss": 0.2071, + "step": 16384 + }, + { + "epoch": 2.98, + "learning_rate": 2.10802693847834e-06, + "loss": 0.1682, + "step": 16385 + }, + { + "epoch": 2.98, + "learning_rate": 2.107972333454678e-06, + "loss": 0.3389, + "step": 16386 + }, + { + "epoch": 2.98, + "learning_rate": 2.107917728431016e-06, + "loss": 0.0412, + "step": 16387 + }, + { + "epoch": 2.98, + "learning_rate": 2.1078631234073537e-06, + "loss": 0.3876, + "step": 16388 + }, + { + "epoch": 2.98, + "learning_rate": 2.1078085183836917e-06, + "loss": 0.1128, + "step": 16389 + }, + { + "epoch": 2.98, + "learning_rate": 2.107753913360029e-06, + "loss": 0.0197, + "step": 16390 + }, + { + "epoch": 2.98, + "learning_rate": 2.107699308336367e-06, + "loss": 1.2149, + "step": 16391 + }, + { + "epoch": 2.98, + "learning_rate": 2.107644703312705e-06, + "loss": 1.9722, + "step": 16392 + }, + { + "epoch": 2.98, + "learning_rate": 2.1075900982890426e-06, + "loss": 1.8634, + "step": 16393 + }, + { + "epoch": 2.98, + "learning_rate": 2.1075354932653806e-06, + "loss": 0.0164, + "step": 16394 + }, + { + "epoch": 2.98, + "learning_rate": 2.107480888241718e-06, + "loss": 1.0175, + "step": 16395 + }, + { + "epoch": 2.98, + "learning_rate": 2.107426283218056e-06, + "loss": 0.2068, + "step": 16396 + }, + { + "epoch": 2.98, + "learning_rate": 2.107371678194394e-06, + "loss": 0.0694, + "step": 16397 + }, + { + "epoch": 2.98, + "learning_rate": 2.1073170731707315e-06, + "loss": 2.3998, + "step": 16398 + }, + { + "epoch": 2.98, + "learning_rate": 2.1072624681470694e-06, + "loss": 2.3785, + "step": 16399 + }, + { + "epoch": 2.98, + "learning_rate": 2.1072078631234074e-06, + "loss": 1.8186, + "step": 16400 + }, + { + "epoch": 2.98, + "learning_rate": 2.1071532580997453e-06, + "loss": 1.3361, + "step": 16401 + }, + { + "epoch": 2.98, + "learning_rate": 2.1070986530760833e-06, + "loss": 0.0219, + "step": 16402 + }, + { + "epoch": 2.98, + "learning_rate": 2.107044048052421e-06, + "loss": 1.7943, + "step": 16403 + }, + { + "epoch": 2.98, + "learning_rate": 2.1069894430287587e-06, + "loss": 1.9516, + "step": 16404 + }, + { + "epoch": 2.98, + "learning_rate": 2.1069348380050967e-06, + "loss": 0.6941, + "step": 16405 + }, + { + "epoch": 2.98, + "learning_rate": 2.1068802329814346e-06, + "loss": 0.5788, + "step": 16406 + }, + { + "epoch": 2.98, + "learning_rate": 2.106825627957772e-06, + "loss": 0.309, + "step": 16407 + }, + { + "epoch": 2.98, + "learning_rate": 2.10677102293411e-06, + "loss": 0.3287, + "step": 16408 + }, + { + "epoch": 2.98, + "learning_rate": 2.1067164179104476e-06, + "loss": 0.522, + "step": 16409 + }, + { + "epoch": 2.98, + "learning_rate": 2.1066618128867855e-06, + "loss": 1.7751, + "step": 16410 + }, + { + "epoch": 2.98, + "learning_rate": 2.1066072078631235e-06, + "loss": 0.7615, + "step": 16411 + }, + { + "epoch": 2.98, + "learning_rate": 2.106552602839461e-06, + "loss": 0.4826, + "step": 16412 + }, + { + "epoch": 2.98, + "learning_rate": 2.106497997815799e-06, + "loss": 2.4582, + "step": 16413 + }, + { + "epoch": 2.98, + "learning_rate": 2.106443392792137e-06, + "loss": 1.032, + "step": 16414 + }, + { + "epoch": 2.99, + "learning_rate": 2.106388787768475e-06, + "loss": 0.3244, + "step": 16415 + }, + { + "epoch": 2.99, + "learning_rate": 2.1063341827448128e-06, + "loss": 2.263, + "step": 16416 + }, + { + "epoch": 2.99, + "learning_rate": 2.1062795777211507e-06, + "loss": 1.2644, + "step": 16417 + }, + { + "epoch": 2.99, + "learning_rate": 2.1062249726974882e-06, + "loss": 0.1877, + "step": 16418 + }, + { + "epoch": 2.99, + "learning_rate": 2.106170367673826e-06, + "loss": 0.0241, + "step": 16419 + }, + { + "epoch": 2.99, + "learning_rate": 2.106115762650164e-06, + "loss": 1.3984, + "step": 16420 + }, + { + "epoch": 2.99, + "learning_rate": 2.1060611576265016e-06, + "loss": 0.4988, + "step": 16421 + }, + { + "epoch": 2.99, + "learning_rate": 2.1060065526028396e-06, + "loss": 1.4142, + "step": 16422 + }, + { + "epoch": 2.99, + "learning_rate": 2.105951947579177e-06, + "loss": 0.895, + "step": 16423 + }, + { + "epoch": 2.99, + "learning_rate": 2.105897342555515e-06, + "loss": 0.3322, + "step": 16424 + }, + { + "epoch": 2.99, + "learning_rate": 2.105842737531853e-06, + "loss": 1.9288, + "step": 16425 + }, + { + "epoch": 2.99, + "learning_rate": 2.1057881325081905e-06, + "loss": 0.0694, + "step": 16426 + }, + { + "epoch": 2.99, + "learning_rate": 2.1057335274845284e-06, + "loss": 0.8213, + "step": 16427 + }, + { + "epoch": 2.99, + "learning_rate": 2.1056789224608664e-06, + "loss": 2.5622, + "step": 16428 + }, + { + "epoch": 2.99, + "learning_rate": 2.1056243174372043e-06, + "loss": 1.914, + "step": 16429 + }, + { + "epoch": 2.99, + "learning_rate": 2.1055697124135423e-06, + "loss": 0.7883, + "step": 16430 + }, + { + "epoch": 2.99, + "learning_rate": 2.1055151073898802e-06, + "loss": 0.803, + "step": 16431 + }, + { + "epoch": 2.99, + "learning_rate": 2.1054605023662177e-06, + "loss": 0.0764, + "step": 16432 + }, + { + "epoch": 2.99, + "learning_rate": 2.1054058973425557e-06, + "loss": 2.5858, + "step": 16433 + }, + { + "epoch": 2.99, + "learning_rate": 2.105351292318893e-06, + "loss": 2.4231, + "step": 16434 + }, + { + "epoch": 2.99, + "learning_rate": 2.105296687295231e-06, + "loss": 1.1911, + "step": 16435 + }, + { + "epoch": 2.99, + "learning_rate": 2.105242082271569e-06, + "loss": 0.5295, + "step": 16436 + }, + { + "epoch": 2.99, + "learning_rate": 2.1051874772479066e-06, + "loss": 0.7335, + "step": 16437 + }, + { + "epoch": 2.99, + "learning_rate": 2.1051328722242446e-06, + "loss": 0.2969, + "step": 16438 + }, + { + "epoch": 2.99, + "learning_rate": 2.1050782672005825e-06, + "loss": 1.5275, + "step": 16439 + }, + { + "epoch": 2.99, + "learning_rate": 2.10502366217692e-06, + "loss": 0.2914, + "step": 16440 + }, + { + "epoch": 2.99, + "learning_rate": 2.104969057153258e-06, + "loss": 0.0762, + "step": 16441 + }, + { + "epoch": 2.99, + "learning_rate": 2.104914452129596e-06, + "loss": 0.9464, + "step": 16442 + }, + { + "epoch": 2.99, + "learning_rate": 2.104859847105934e-06, + "loss": 0.4278, + "step": 16443 + }, + { + "epoch": 2.99, + "learning_rate": 2.104805242082272e-06, + "loss": 0.2588, + "step": 16444 + }, + { + "epoch": 2.99, + "learning_rate": 2.1047506370586097e-06, + "loss": 1.1665, + "step": 16445 + }, + { + "epoch": 2.99, + "learning_rate": 2.1046960320349473e-06, + "loss": 2.6032, + "step": 16446 + }, + { + "epoch": 2.99, + "learning_rate": 2.104641427011285e-06, + "loss": 0.6278, + "step": 16447 + }, + { + "epoch": 2.99, + "learning_rate": 2.1045868219876227e-06, + "loss": 0.4696, + "step": 16448 + }, + { + "epoch": 2.99, + "learning_rate": 2.1045322169639607e-06, + "loss": 1.4036, + "step": 16449 + }, + { + "epoch": 2.99, + "learning_rate": 2.1044776119402986e-06, + "loss": 2.6447, + "step": 16450 + }, + { + "epoch": 2.99, + "learning_rate": 2.104423006916636e-06, + "loss": 1.7108, + "step": 16451 + }, + { + "epoch": 2.99, + "learning_rate": 2.104368401892974e-06, + "loss": 1.353, + "step": 16452 + }, + { + "epoch": 2.99, + "learning_rate": 2.104313796869312e-06, + "loss": 0.6121, + "step": 16453 + }, + { + "epoch": 2.99, + "learning_rate": 2.10425919184565e-06, + "loss": 0.1188, + "step": 16454 + }, + { + "epoch": 2.99, + "learning_rate": 2.1042045868219875e-06, + "loss": 0.7394, + "step": 16455 + }, + { + "epoch": 2.99, + "learning_rate": 2.1041499817983254e-06, + "loss": 0.2205, + "step": 16456 + }, + { + "epoch": 2.99, + "learning_rate": 2.1040953767746634e-06, + "loss": 1.1479, + "step": 16457 + }, + { + "epoch": 2.99, + "learning_rate": 2.1040407717510013e-06, + "loss": 1.8913, + "step": 16458 + }, + { + "epoch": 2.99, + "learning_rate": 2.1039861667273392e-06, + "loss": 0.3195, + "step": 16459 + }, + { + "epoch": 2.99, + "learning_rate": 2.1039315617036768e-06, + "loss": 0.8847, + "step": 16460 + }, + { + "epoch": 2.99, + "learning_rate": 2.1038769566800147e-06, + "loss": 0.4014, + "step": 16461 + }, + { + "epoch": 2.99, + "learning_rate": 2.1038223516563522e-06, + "loss": 0.4492, + "step": 16462 + }, + { + "epoch": 2.99, + "learning_rate": 2.10376774663269e-06, + "loss": 0.3457, + "step": 16463 + }, + { + "epoch": 2.99, + "learning_rate": 2.103713141609028e-06, + "loss": 1.1969, + "step": 16464 + }, + { + "epoch": 2.99, + "learning_rate": 2.1036585365853656e-06, + "loss": 1.2821, + "step": 16465 + }, + { + "epoch": 2.99, + "learning_rate": 2.1036039315617036e-06, + "loss": 0.2982, + "step": 16466 + }, + { + "epoch": 2.99, + "learning_rate": 2.1035493265380415e-06, + "loss": 3.4846, + "step": 16467 + }, + { + "epoch": 2.99, + "learning_rate": 2.1034947215143795e-06, + "loss": 0.0363, + "step": 16468 + }, + { + "epoch": 2.99, + "learning_rate": 2.1034401164907174e-06, + "loss": 2.1249, + "step": 16469 + }, + { + "epoch": 3.0, + "learning_rate": 2.1033855114670554e-06, + "loss": 2.1357, + "step": 16470 + }, + { + "epoch": 3.0, + "learning_rate": 2.103330906443393e-06, + "loss": 1.2878, + "step": 16471 + }, + { + "epoch": 3.0, + "learning_rate": 2.103276301419731e-06, + "loss": 0.6228, + "step": 16472 + }, + { + "epoch": 3.0, + "learning_rate": 2.1032216963960688e-06, + "loss": 2.6587, + "step": 16473 + }, + { + "epoch": 3.0, + "learning_rate": 2.1031670913724063e-06, + "loss": 2.2334, + "step": 16474 + }, + { + "epoch": 3.0, + "learning_rate": 2.1031124863487442e-06, + "loss": 0.4582, + "step": 16475 + }, + { + "epoch": 3.0, + "learning_rate": 2.1030578813250817e-06, + "loss": 2.0579, + "step": 16476 + }, + { + "epoch": 3.0, + "learning_rate": 2.1030032763014197e-06, + "loss": 0.4925, + "step": 16477 + }, + { + "epoch": 3.0, + "learning_rate": 2.1029486712777576e-06, + "loss": 3.6242, + "step": 16478 + }, + { + "epoch": 3.0, + "learning_rate": 2.102894066254095e-06, + "loss": 0.468, + "step": 16479 + }, + { + "epoch": 3.0, + "learning_rate": 2.102839461230433e-06, + "loss": 0.031, + "step": 16480 + }, + { + "epoch": 3.0, + "learning_rate": 2.102784856206771e-06, + "loss": 3.2927, + "step": 16481 + }, + { + "epoch": 3.0, + "learning_rate": 2.102730251183109e-06, + "loss": 0.0556, + "step": 16482 + }, + { + "epoch": 3.0, + "learning_rate": 2.102675646159447e-06, + "loss": 0.5822, + "step": 16483 + }, + { + "epoch": 3.0, + "learning_rate": 2.102621041135785e-06, + "loss": 0.1508, + "step": 16484 + }, + { + "epoch": 3.0, + "learning_rate": 2.1025664361121224e-06, + "loss": 0.3856, + "step": 16485 + }, + { + "epoch": 3.0, + "learning_rate": 2.1025118310884603e-06, + "loss": 1.3462, + "step": 16486 + }, + { + "epoch": 3.0, + "learning_rate": 2.1024572260647983e-06, + "loss": 0.1736, + "step": 16487 + }, + { + "epoch": 3.0, + "learning_rate": 2.102402621041136e-06, + "loss": 0.6954, + "step": 16488 + }, + { + "epoch": 3.0, + "learning_rate": 2.1023480160174737e-06, + "loss": 2.601, + "step": 16489 + }, + { + "epoch": 3.0, + "learning_rate": 2.1022934109938113e-06, + "loss": 0.795, + "step": 16490 + }, + { + "epoch": 3.0, + "learning_rate": 2.102238805970149e-06, + "loss": 0.5235, + "step": 16491 + }, + { + "epoch": 3.0, + "learning_rate": 2.102184200946487e-06, + "loss": 0.8267, + "step": 16492 + }, + { + "epoch": 3.0, + "learning_rate": 2.1021295959228247e-06, + "loss": 1.2721, + "step": 16493 + }, + { + "epoch": 3.0, + "learning_rate": 2.1020749908991626e-06, + "loss": 0.1266, + "step": 16494 + }, + { + "epoch": 3.0, + "learning_rate": 2.1020203858755006e-06, + "loss": 0.0731, + "step": 16495 + }, + { + "epoch": 3.0, + "learning_rate": 2.1019657808518385e-06, + "loss": 0.8549, + "step": 16496 + }, + { + "epoch": 3.0, + "learning_rate": 2.1019111758281764e-06, + "loss": 4.829, + "step": 16497 + }, + { + "epoch": 3.0, + "eval_accuracy": 0.8162847790507365, + "eval_loss": 0.8254396915435791, + "eval_runtime": 135.7327, + "eval_samples_per_second": 18.006, + "eval_steps_per_second": 4.501, + "step": 16497 + }, + { + "epoch": 3.0, + "learning_rate": 2.1018565708045144e-06, + "loss": 1.1787, + "step": 16498 + }, + { + "epoch": 3.0, + "learning_rate": 2.101801965780852e-06, + "loss": 0.8355, + "step": 16499 + }, + { + "epoch": 3.0, + "learning_rate": 2.10174736075719e-06, + "loss": 1.779, + "step": 16500 + }, + { + "epoch": 3.0, + "learning_rate": 2.1016927557335274e-06, + "loss": 0.3479, + "step": 16501 + }, + { + "epoch": 3.0, + "learning_rate": 2.1016381507098653e-06, + "loss": 1.2439, + "step": 16502 + }, + { + "epoch": 3.0, + "learning_rate": 2.1015835456862033e-06, + "loss": 1.8522, + "step": 16503 + }, + { + "epoch": 3.0, + "learning_rate": 2.1015289406625408e-06, + "loss": 0.8022, + "step": 16504 + }, + { + "epoch": 3.0, + "learning_rate": 2.1014743356388787e-06, + "loss": 0.2499, + "step": 16505 + }, + { + "epoch": 3.0, + "learning_rate": 2.1014197306152167e-06, + "loss": 0.3455, + "step": 16506 + }, + { + "epoch": 3.0, + "learning_rate": 2.101365125591554e-06, + "loss": 1.773, + "step": 16507 + }, + { + "epoch": 3.0, + "learning_rate": 2.101310520567892e-06, + "loss": 0.3065, + "step": 16508 + }, + { + "epoch": 3.0, + "learning_rate": 2.10125591554423e-06, + "loss": 1.1231, + "step": 16509 + }, + { + "epoch": 3.0, + "learning_rate": 2.101201310520568e-06, + "loss": 1.2203, + "step": 16510 + }, + { + "epoch": 3.0, + "learning_rate": 2.101146705496906e-06, + "loss": 0.4794, + "step": 16511 + }, + { + "epoch": 3.0, + "learning_rate": 2.101092100473244e-06, + "loss": 0.2871, + "step": 16512 + }, + { + "epoch": 3.0, + "learning_rate": 2.1010374954495814e-06, + "loss": 2.2642, + "step": 16513 + }, + { + "epoch": 3.0, + "learning_rate": 2.1009828904259194e-06, + "loss": 0.6773, + "step": 16514 + }, + { + "epoch": 3.0, + "learning_rate": 2.100928285402257e-06, + "loss": 2.2152, + "step": 16515 + }, + { + "epoch": 3.0, + "learning_rate": 2.100873680378595e-06, + "loss": 0.7727, + "step": 16516 + }, + { + "epoch": 3.0, + "learning_rate": 2.1008190753549328e-06, + "loss": 0.7487, + "step": 16517 + }, + { + "epoch": 3.0, + "learning_rate": 2.1007644703312703e-06, + "loss": 0.0165, + "step": 16518 + }, + { + "epoch": 3.0, + "learning_rate": 2.1007098653076082e-06, + "loss": 0.385, + "step": 16519 + }, + { + "epoch": 3.0, + "learning_rate": 2.100655260283946e-06, + "loss": 0.5593, + "step": 16520 + }, + { + "epoch": 3.0, + "learning_rate": 2.1006006552602837e-06, + "loss": 0.6582, + "step": 16521 + }, + { + "epoch": 3.0, + "learning_rate": 2.1005460502366216e-06, + "loss": 1.2233, + "step": 16522 + }, + { + "epoch": 3.0, + "learning_rate": 2.1004914452129596e-06, + "loss": 1.9015, + "step": 16523 + }, + { + "epoch": 3.0, + "learning_rate": 2.1004368401892975e-06, + "loss": 0.7193, + "step": 16524 + }, + { + "epoch": 3.01, + "learning_rate": 2.1003822351656355e-06, + "loss": 1.3349, + "step": 16525 + }, + { + "epoch": 3.01, + "learning_rate": 2.1003276301419734e-06, + "loss": 0.302, + "step": 16526 + }, + { + "epoch": 3.01, + "learning_rate": 2.100273025118311e-06, + "loss": 0.8163, + "step": 16527 + }, + { + "epoch": 3.01, + "learning_rate": 2.100218420094649e-06, + "loss": 0.2761, + "step": 16528 + }, + { + "epoch": 3.01, + "learning_rate": 2.1001638150709864e-06, + "loss": 0.1495, + "step": 16529 + }, + { + "epoch": 3.01, + "learning_rate": 2.1001092100473243e-06, + "loss": 0.2472, + "step": 16530 + }, + { + "epoch": 3.01, + "learning_rate": 2.1000546050236623e-06, + "loss": 0.3108, + "step": 16531 + }, + { + "epoch": 3.01, + "learning_rate": 2.1e-06, + "loss": 1.072, + "step": 16532 + }, + { + "epoch": 3.01, + "learning_rate": 2.0999453949763377e-06, + "loss": 0.6646, + "step": 16533 + }, + { + "epoch": 3.01, + "learning_rate": 2.0998907899526757e-06, + "loss": 0.3511, + "step": 16534 + }, + { + "epoch": 3.01, + "learning_rate": 2.099836184929013e-06, + "loss": 0.8165, + "step": 16535 + }, + { + "epoch": 3.01, + "learning_rate": 2.099781579905351e-06, + "loss": 0.9975, + "step": 16536 + }, + { + "epoch": 3.01, + "learning_rate": 2.099726974881689e-06, + "loss": 0.9094, + "step": 16537 + }, + { + "epoch": 3.01, + "learning_rate": 2.099672369858027e-06, + "loss": 0.5345, + "step": 16538 + }, + { + "epoch": 3.01, + "learning_rate": 2.099617764834365e-06, + "loss": 1.1324, + "step": 16539 + }, + { + "epoch": 3.01, + "learning_rate": 2.099563159810703e-06, + "loss": 0.7193, + "step": 16540 + }, + { + "epoch": 3.01, + "learning_rate": 2.0995085547870404e-06, + "loss": 0.3748, + "step": 16541 + }, + { + "epoch": 3.01, + "learning_rate": 2.0994539497633784e-06, + "loss": 0.5256, + "step": 16542 + }, + { + "epoch": 3.01, + "learning_rate": 2.099399344739716e-06, + "loss": 0.1658, + "step": 16543 + }, + { + "epoch": 3.01, + "learning_rate": 2.099344739716054e-06, + "loss": 0.6094, + "step": 16544 + }, + { + "epoch": 3.01, + "learning_rate": 2.099290134692392e-06, + "loss": 1.3393, + "step": 16545 + }, + { + "epoch": 3.01, + "learning_rate": 2.0992355296687293e-06, + "loss": 0.7665, + "step": 16546 + }, + { + "epoch": 3.01, + "learning_rate": 2.0991809246450673e-06, + "loss": 1.9239, + "step": 16547 + }, + { + "epoch": 3.01, + "learning_rate": 2.099126319621405e-06, + "loss": 0.186, + "step": 16548 + }, + { + "epoch": 3.01, + "learning_rate": 2.099071714597743e-06, + "loss": 0.2686, + "step": 16549 + }, + { + "epoch": 3.01, + "learning_rate": 2.099017109574081e-06, + "loss": 1.1276, + "step": 16550 + }, + { + "epoch": 3.01, + "learning_rate": 2.098962504550419e-06, + "loss": 1.7843, + "step": 16551 + }, + { + "epoch": 3.01, + "learning_rate": 2.0989078995267565e-06, + "loss": 0.5612, + "step": 16552 + }, + { + "epoch": 3.01, + "learning_rate": 2.0988532945030945e-06, + "loss": 0.494, + "step": 16553 + }, + { + "epoch": 3.01, + "learning_rate": 2.0987986894794324e-06, + "loss": 1.0631, + "step": 16554 + }, + { + "epoch": 3.01, + "learning_rate": 2.09874408445577e-06, + "loss": 0.6273, + "step": 16555 + }, + { + "epoch": 3.01, + "learning_rate": 2.098689479432108e-06, + "loss": 0.6122, + "step": 16556 + }, + { + "epoch": 3.01, + "learning_rate": 2.0986348744084454e-06, + "loss": 0.7738, + "step": 16557 + }, + { + "epoch": 3.01, + "learning_rate": 2.0985802693847834e-06, + "loss": 1.4476, + "step": 16558 + }, + { + "epoch": 3.01, + "learning_rate": 2.0985256643611213e-06, + "loss": 0.4805, + "step": 16559 + }, + { + "epoch": 3.01, + "learning_rate": 2.098471059337459e-06, + "loss": 2.2209, + "step": 16560 + }, + { + "epoch": 3.01, + "learning_rate": 2.0984164543137968e-06, + "loss": 0.4583, + "step": 16561 + }, + { + "epoch": 3.01, + "learning_rate": 2.0983618492901347e-06, + "loss": 2.6884, + "step": 16562 + }, + { + "epoch": 3.01, + "learning_rate": 2.0983072442664727e-06, + "loss": 3.4735, + "step": 16563 + }, + { + "epoch": 3.01, + "learning_rate": 2.0982526392428106e-06, + "loss": 0.666, + "step": 16564 + }, + { + "epoch": 3.01, + "learning_rate": 2.0981980342191485e-06, + "loss": 1.558, + "step": 16565 + }, + { + "epoch": 3.01, + "learning_rate": 2.098143429195486e-06, + "loss": 1.0851, + "step": 16566 + }, + { + "epoch": 3.01, + "learning_rate": 2.098088824171824e-06, + "loss": 0.0069, + "step": 16567 + }, + { + "epoch": 3.01, + "learning_rate": 2.098034219148162e-06, + "loss": 0.3196, + "step": 16568 + }, + { + "epoch": 3.01, + "learning_rate": 2.0979796141244995e-06, + "loss": 2.6158, + "step": 16569 + }, + { + "epoch": 3.01, + "learning_rate": 2.0979250091008374e-06, + "loss": 0.4817, + "step": 16570 + }, + { + "epoch": 3.01, + "learning_rate": 2.097870404077175e-06, + "loss": 0.37, + "step": 16571 + }, + { + "epoch": 3.01, + "learning_rate": 2.097815799053513e-06, + "loss": 0.3807, + "step": 16572 + }, + { + "epoch": 3.01, + "learning_rate": 2.097761194029851e-06, + "loss": 0.019, + "step": 16573 + }, + { + "epoch": 3.01, + "learning_rate": 2.0977065890061883e-06, + "loss": 0.9588, + "step": 16574 + }, + { + "epoch": 3.01, + "learning_rate": 2.0976519839825263e-06, + "loss": 0.2473, + "step": 16575 + }, + { + "epoch": 3.01, + "learning_rate": 2.0975973789588642e-06, + "loss": 0.0361, + "step": 16576 + }, + { + "epoch": 3.01, + "learning_rate": 2.097542773935202e-06, + "loss": 0.2012, + "step": 16577 + }, + { + "epoch": 3.01, + "learning_rate": 2.09748816891154e-06, + "loss": 1.1846, + "step": 16578 + }, + { + "epoch": 3.01, + "learning_rate": 2.097433563887878e-06, + "loss": 0.3432, + "step": 16579 + }, + { + "epoch": 3.02, + "learning_rate": 2.0973789588642156e-06, + "loss": 0.0244, + "step": 16580 + }, + { + "epoch": 3.02, + "learning_rate": 2.0973243538405535e-06, + "loss": 0.2135, + "step": 16581 + }, + { + "epoch": 3.02, + "learning_rate": 2.097269748816891e-06, + "loss": 1.8807, + "step": 16582 + }, + { + "epoch": 3.02, + "learning_rate": 2.097215143793229e-06, + "loss": 0.2517, + "step": 16583 + }, + { + "epoch": 3.02, + "learning_rate": 2.097160538769567e-06, + "loss": 2.028, + "step": 16584 + }, + { + "epoch": 3.02, + "learning_rate": 2.0971059337459044e-06, + "loss": 0.4785, + "step": 16585 + }, + { + "epoch": 3.02, + "learning_rate": 2.0970513287222424e-06, + "loss": 0.6336, + "step": 16586 + }, + { + "epoch": 3.02, + "learning_rate": 2.0969967236985803e-06, + "loss": 1.5869, + "step": 16587 + }, + { + "epoch": 3.02, + "learning_rate": 2.096942118674918e-06, + "loss": 0.0817, + "step": 16588 + }, + { + "epoch": 3.02, + "learning_rate": 2.096887513651256e-06, + "loss": 0.5319, + "step": 16589 + }, + { + "epoch": 3.02, + "learning_rate": 2.0968329086275937e-06, + "loss": 0.9928, + "step": 16590 + }, + { + "epoch": 3.02, + "learning_rate": 2.0967783036039317e-06, + "loss": 0.096, + "step": 16591 + }, + { + "epoch": 3.02, + "learning_rate": 2.0967236985802696e-06, + "loss": 0.4476, + "step": 16592 + }, + { + "epoch": 3.02, + "learning_rate": 2.0966690935566076e-06, + "loss": 0.3427, + "step": 16593 + }, + { + "epoch": 3.02, + "learning_rate": 2.096614488532945e-06, + "loss": 0.2995, + "step": 16594 + }, + { + "epoch": 3.02, + "learning_rate": 2.096559883509283e-06, + "loss": 0.2275, + "step": 16595 + }, + { + "epoch": 3.02, + "learning_rate": 2.0965052784856206e-06, + "loss": 1.3082, + "step": 16596 + }, + { + "epoch": 3.02, + "learning_rate": 2.0964506734619585e-06, + "loss": 1.8846, + "step": 16597 + }, + { + "epoch": 3.02, + "learning_rate": 2.0963960684382964e-06, + "loss": 0.3256, + "step": 16598 + }, + { + "epoch": 3.02, + "learning_rate": 2.096341463414634e-06, + "loss": 1.0525, + "step": 16599 + }, + { + "epoch": 3.02, + "learning_rate": 2.096286858390972e-06, + "loss": 0.3478, + "step": 16600 + }, + { + "epoch": 3.02, + "learning_rate": 2.09623225336731e-06, + "loss": 1.704, + "step": 16601 + }, + { + "epoch": 3.02, + "learning_rate": 2.0961776483436474e-06, + "loss": 0.5897, + "step": 16602 + }, + { + "epoch": 3.02, + "learning_rate": 2.0961230433199853e-06, + "loss": 0.0248, + "step": 16603 + }, + { + "epoch": 3.02, + "learning_rate": 2.0960684382963233e-06, + "loss": 0.2883, + "step": 16604 + }, + { + "epoch": 3.02, + "learning_rate": 2.096013833272661e-06, + "loss": 1.6425, + "step": 16605 + }, + { + "epoch": 3.02, + "learning_rate": 2.095959228248999e-06, + "loss": 1.7023, + "step": 16606 + }, + { + "epoch": 3.02, + "learning_rate": 2.095904623225337e-06, + "loss": 0.4368, + "step": 16607 + }, + { + "epoch": 3.02, + "learning_rate": 2.0958500182016746e-06, + "loss": 0.7656, + "step": 16608 + }, + { + "epoch": 3.02, + "learning_rate": 2.0957954131780125e-06, + "loss": 2.3715, + "step": 16609 + }, + { + "epoch": 3.02, + "learning_rate": 2.09574080815435e-06, + "loss": 1.3124, + "step": 16610 + }, + { + "epoch": 3.02, + "learning_rate": 2.095686203130688e-06, + "loss": 0.7981, + "step": 16611 + }, + { + "epoch": 3.02, + "learning_rate": 2.095631598107026e-06, + "loss": 0.9177, + "step": 16612 + }, + { + "epoch": 3.02, + "learning_rate": 2.0955769930833635e-06, + "loss": 0.8022, + "step": 16613 + }, + { + "epoch": 3.02, + "learning_rate": 2.0955223880597014e-06, + "loss": 0.5706, + "step": 16614 + }, + { + "epoch": 3.02, + "learning_rate": 2.0954677830360394e-06, + "loss": 0.0442, + "step": 16615 + }, + { + "epoch": 3.02, + "learning_rate": 2.095413178012377e-06, + "loss": 0.0121, + "step": 16616 + }, + { + "epoch": 3.02, + "learning_rate": 2.095358572988715e-06, + "loss": 0.7864, + "step": 16617 + }, + { + "epoch": 3.02, + "learning_rate": 2.0953039679650528e-06, + "loss": 0.0201, + "step": 16618 + }, + { + "epoch": 3.02, + "learning_rate": 2.0952493629413907e-06, + "loss": 0.1045, + "step": 16619 + }, + { + "epoch": 3.02, + "learning_rate": 2.0951947579177287e-06, + "loss": 0.2464, + "step": 16620 + }, + { + "epoch": 3.02, + "learning_rate": 2.0951401528940666e-06, + "loss": 2.1339, + "step": 16621 + }, + { + "epoch": 3.02, + "learning_rate": 2.095085547870404e-06, + "loss": 1.2986, + "step": 16622 + }, + { + "epoch": 3.02, + "learning_rate": 2.095030942846742e-06, + "loss": 1.348, + "step": 16623 + }, + { + "epoch": 3.02, + "learning_rate": 2.0949763378230796e-06, + "loss": 0.4142, + "step": 16624 + }, + { + "epoch": 3.02, + "learning_rate": 2.0949217327994175e-06, + "loss": 1.6249, + "step": 16625 + }, + { + "epoch": 3.02, + "learning_rate": 2.0948671277757555e-06, + "loss": 0.3945, + "step": 16626 + }, + { + "epoch": 3.02, + "learning_rate": 2.094812522752093e-06, + "loss": 0.8324, + "step": 16627 + }, + { + "epoch": 3.02, + "learning_rate": 2.094757917728431e-06, + "loss": 1.6066, + "step": 16628 + }, + { + "epoch": 3.02, + "learning_rate": 2.094703312704769e-06, + "loss": 0.4423, + "step": 16629 + }, + { + "epoch": 3.02, + "learning_rate": 2.094648707681107e-06, + "loss": 2.4712, + "step": 16630 + }, + { + "epoch": 3.02, + "learning_rate": 2.0945941026574448e-06, + "loss": 0.0204, + "step": 16631 + }, + { + "epoch": 3.02, + "learning_rate": 2.0945394976337827e-06, + "loss": 1.2769, + "step": 16632 + }, + { + "epoch": 3.02, + "learning_rate": 2.0944848926101202e-06, + "loss": 0.2936, + "step": 16633 + }, + { + "epoch": 3.02, + "learning_rate": 2.094430287586458e-06, + "loss": 1.1296, + "step": 16634 + }, + { + "epoch": 3.03, + "learning_rate": 2.094375682562796e-06, + "loss": 0.2246, + "step": 16635 + }, + { + "epoch": 3.03, + "learning_rate": 2.0943210775391336e-06, + "loss": 0.0232, + "step": 16636 + }, + { + "epoch": 3.03, + "learning_rate": 2.0942664725154716e-06, + "loss": 0.2288, + "step": 16637 + }, + { + "epoch": 3.03, + "learning_rate": 2.094211867491809e-06, + "loss": 0.4267, + "step": 16638 + }, + { + "epoch": 3.03, + "learning_rate": 2.094157262468147e-06, + "loss": 0.9236, + "step": 16639 + }, + { + "epoch": 3.03, + "learning_rate": 2.094102657444485e-06, + "loss": 0.902, + "step": 16640 + }, + { + "epoch": 3.03, + "learning_rate": 2.0940480524208225e-06, + "loss": 2.3077, + "step": 16641 + }, + { + "epoch": 3.03, + "learning_rate": 2.0939934473971604e-06, + "loss": 0.3332, + "step": 16642 + }, + { + "epoch": 3.03, + "learning_rate": 2.0939388423734984e-06, + "loss": 0.0253, + "step": 16643 + }, + { + "epoch": 3.03, + "learning_rate": 2.0938842373498363e-06, + "loss": 1.1091, + "step": 16644 + }, + { + "epoch": 3.03, + "learning_rate": 2.0938296323261743e-06, + "loss": 1.7962, + "step": 16645 + }, + { + "epoch": 3.03, + "learning_rate": 2.0937750273025122e-06, + "loss": 1.2032, + "step": 16646 + }, + { + "epoch": 3.03, + "learning_rate": 2.0937204222788497e-06, + "loss": 0.0844, + "step": 16647 + }, + { + "epoch": 3.03, + "learning_rate": 2.0936658172551877e-06, + "loss": 0.5261, + "step": 16648 + }, + { + "epoch": 3.03, + "learning_rate": 2.0936112122315256e-06, + "loss": 2.5747, + "step": 16649 + }, + { + "epoch": 3.03, + "learning_rate": 2.093556607207863e-06, + "loss": 0.4024, + "step": 16650 + }, + { + "epoch": 3.03, + "learning_rate": 2.093502002184201e-06, + "loss": 0.9897, + "step": 16651 + }, + { + "epoch": 3.03, + "learning_rate": 2.0934473971605386e-06, + "loss": 0.3268, + "step": 16652 + }, + { + "epoch": 3.03, + "learning_rate": 2.0933927921368766e-06, + "loss": 0.8026, + "step": 16653 + }, + { + "epoch": 3.03, + "learning_rate": 2.0933381871132145e-06, + "loss": 1.118, + "step": 16654 + }, + { + "epoch": 3.03, + "learning_rate": 2.093283582089552e-06, + "loss": 1.4258, + "step": 16655 + }, + { + "epoch": 3.03, + "learning_rate": 2.09322897706589e-06, + "loss": 0.0239, + "step": 16656 + }, + { + "epoch": 3.03, + "learning_rate": 2.093174372042228e-06, + "loss": 1.091, + "step": 16657 + }, + { + "epoch": 3.03, + "learning_rate": 2.093119767018566e-06, + "loss": 0.8803, + "step": 16658 + }, + { + "epoch": 3.03, + "learning_rate": 2.0930651619949038e-06, + "loss": 0.0209, + "step": 16659 + }, + { + "epoch": 3.03, + "learning_rate": 2.0930105569712417e-06, + "loss": 1.6019, + "step": 16660 + }, + { + "epoch": 3.03, + "learning_rate": 2.0929559519475793e-06, + "loss": 0.7945, + "step": 16661 + }, + { + "epoch": 3.03, + "learning_rate": 2.092901346923917e-06, + "loss": 0.1462, + "step": 16662 + }, + { + "epoch": 3.03, + "learning_rate": 2.0928467419002547e-06, + "loss": 0.4058, + "step": 16663 + }, + { + "epoch": 3.03, + "learning_rate": 2.0927921368765927e-06, + "loss": 0.8471, + "step": 16664 + }, + { + "epoch": 3.03, + "learning_rate": 2.0927375318529306e-06, + "loss": 0.4339, + "step": 16665 + }, + { + "epoch": 3.03, + "learning_rate": 2.092682926829268e-06, + "loss": 0.5823, + "step": 16666 + }, + { + "epoch": 3.03, + "learning_rate": 2.092628321805606e-06, + "loss": 1.0749, + "step": 16667 + }, + { + "epoch": 3.03, + "learning_rate": 2.092573716781944e-06, + "loss": 1.0316, + "step": 16668 + }, + { + "epoch": 3.03, + "learning_rate": 2.0925191117582815e-06, + "loss": 0.0118, + "step": 16669 + }, + { + "epoch": 3.03, + "learning_rate": 2.0924645067346195e-06, + "loss": 0.3835, + "step": 16670 + }, + { + "epoch": 3.03, + "learning_rate": 2.0924099017109574e-06, + "loss": 0.464, + "step": 16671 + }, + { + "epoch": 3.03, + "learning_rate": 2.0923552966872954e-06, + "loss": 1.456, + "step": 16672 + }, + { + "epoch": 3.03, + "learning_rate": 2.0923006916636333e-06, + "loss": 0.2703, + "step": 16673 + }, + { + "epoch": 3.03, + "learning_rate": 2.0922460866399712e-06, + "loss": 1.364, + "step": 16674 + }, + { + "epoch": 3.03, + "learning_rate": 2.0921914816163088e-06, + "loss": 0.2552, + "step": 16675 + }, + { + "epoch": 3.03, + "learning_rate": 2.0921368765926467e-06, + "loss": 0.7999, + "step": 16676 + }, + { + "epoch": 3.03, + "learning_rate": 2.0920822715689842e-06, + "loss": 2.7499, + "step": 16677 + }, + { + "epoch": 3.03, + "learning_rate": 2.092027666545322e-06, + "loss": 0.0442, + "step": 16678 + }, + { + "epoch": 3.03, + "learning_rate": 2.09197306152166e-06, + "loss": 0.256, + "step": 16679 + }, + { + "epoch": 3.03, + "learning_rate": 2.0919184564979976e-06, + "loss": 1.5817, + "step": 16680 + }, + { + "epoch": 3.03, + "learning_rate": 2.0918638514743356e-06, + "loss": 0.8631, + "step": 16681 + }, + { + "epoch": 3.03, + "learning_rate": 2.0918092464506735e-06, + "loss": 0.0189, + "step": 16682 + }, + { + "epoch": 3.03, + "learning_rate": 2.091754641427011e-06, + "loss": 0.6392, + "step": 16683 + }, + { + "epoch": 3.03, + "learning_rate": 2.091700036403349e-06, + "loss": 0.6171, + "step": 16684 + }, + { + "epoch": 3.03, + "learning_rate": 2.091645431379687e-06, + "loss": 0.0146, + "step": 16685 + }, + { + "epoch": 3.03, + "learning_rate": 2.091590826356025e-06, + "loss": 0.745, + "step": 16686 + }, + { + "epoch": 3.03, + "learning_rate": 2.091536221332363e-06, + "loss": 1.407, + "step": 16687 + }, + { + "epoch": 3.03, + "learning_rate": 2.0914816163087008e-06, + "loss": 0.8593, + "step": 16688 + }, + { + "epoch": 3.03, + "learning_rate": 2.0914270112850383e-06, + "loss": 1.8187, + "step": 16689 + }, + { + "epoch": 3.04, + "learning_rate": 2.0913724062613762e-06, + "loss": 1.9223, + "step": 16690 + }, + { + "epoch": 3.04, + "learning_rate": 2.0913178012377137e-06, + "loss": 4.0181, + "step": 16691 + }, + { + "epoch": 3.04, + "learning_rate": 2.0912631962140517e-06, + "loss": 0.0808, + "step": 16692 + }, + { + "epoch": 3.04, + "learning_rate": 2.0912085911903896e-06, + "loss": 0.3407, + "step": 16693 + }, + { + "epoch": 3.04, + "learning_rate": 2.091153986166727e-06, + "loss": 0.4734, + "step": 16694 + }, + { + "epoch": 3.04, + "learning_rate": 2.091099381143065e-06, + "loss": 0.5237, + "step": 16695 + }, + { + "epoch": 3.04, + "learning_rate": 2.091044776119403e-06, + "loss": 0.1028, + "step": 16696 + }, + { + "epoch": 3.04, + "learning_rate": 2.0909901710957406e-06, + "loss": 0.2213, + "step": 16697 + }, + { + "epoch": 3.04, + "learning_rate": 2.0909355660720785e-06, + "loss": 0.7848, + "step": 16698 + }, + { + "epoch": 3.04, + "learning_rate": 2.0908809610484164e-06, + "loss": 0.5948, + "step": 16699 + }, + { + "epoch": 3.04, + "learning_rate": 2.0908263560247544e-06, + "loss": 0.3482, + "step": 16700 + }, + { + "epoch": 3.04, + "learning_rate": 2.0907717510010923e-06, + "loss": 0.6672, + "step": 16701 + }, + { + "epoch": 3.04, + "learning_rate": 2.0907171459774303e-06, + "loss": 0.0115, + "step": 16702 + }, + { + "epoch": 3.04, + "learning_rate": 2.090662540953768e-06, + "loss": 0.463, + "step": 16703 + }, + { + "epoch": 3.04, + "learning_rate": 2.0906079359301057e-06, + "loss": 0.6149, + "step": 16704 + }, + { + "epoch": 3.04, + "learning_rate": 2.0905533309064433e-06, + "loss": 0.044, + "step": 16705 + }, + { + "epoch": 3.04, + "learning_rate": 2.090498725882781e-06, + "loss": 0.2448, + "step": 16706 + }, + { + "epoch": 3.04, + "learning_rate": 2.090444120859119e-06, + "loss": 0.0388, + "step": 16707 + }, + { + "epoch": 3.04, + "learning_rate": 2.0903895158354567e-06, + "loss": 0.0763, + "step": 16708 + }, + { + "epoch": 3.04, + "learning_rate": 2.0903349108117946e-06, + "loss": 1.1868, + "step": 16709 + }, + { + "epoch": 3.04, + "learning_rate": 2.0902803057881325e-06, + "loss": 0.032, + "step": 16710 + }, + { + "epoch": 3.04, + "learning_rate": 2.0902257007644705e-06, + "loss": 0.4907, + "step": 16711 + }, + { + "epoch": 3.04, + "learning_rate": 2.0901710957408084e-06, + "loss": 2.3435, + "step": 16712 + }, + { + "epoch": 3.04, + "learning_rate": 2.0901164907171464e-06, + "loss": 0.033, + "step": 16713 + }, + { + "epoch": 3.04, + "learning_rate": 2.090061885693484e-06, + "loss": 0.6425, + "step": 16714 + }, + { + "epoch": 3.04, + "learning_rate": 2.090007280669822e-06, + "loss": 1.4589, + "step": 16715 + }, + { + "epoch": 3.04, + "learning_rate": 2.0899526756461598e-06, + "loss": 1.3176, + "step": 16716 + }, + { + "epoch": 3.04, + "learning_rate": 2.0898980706224973e-06, + "loss": 1.1802, + "step": 16717 + }, + { + "epoch": 3.04, + "learning_rate": 2.0898434655988352e-06, + "loss": 0.9044, + "step": 16718 + }, + { + "epoch": 3.04, + "learning_rate": 2.0897888605751728e-06, + "loss": 0.0344, + "step": 16719 + }, + { + "epoch": 3.04, + "learning_rate": 2.0897342555515107e-06, + "loss": 0.5394, + "step": 16720 + }, + { + "epoch": 3.04, + "learning_rate": 2.0896796505278487e-06, + "loss": 0.3955, + "step": 16721 + }, + { + "epoch": 3.04, + "learning_rate": 2.089625045504186e-06, + "loss": 0.7385, + "step": 16722 + }, + { + "epoch": 3.04, + "learning_rate": 2.089570440480524e-06, + "loss": 1.1136, + "step": 16723 + }, + { + "epoch": 3.04, + "learning_rate": 2.089515835456862e-06, + "loss": 0.3793, + "step": 16724 + }, + { + "epoch": 3.04, + "learning_rate": 2.0894612304332e-06, + "loss": 0.547, + "step": 16725 + }, + { + "epoch": 3.04, + "learning_rate": 2.089406625409538e-06, + "loss": 0.47, + "step": 16726 + }, + { + "epoch": 3.04, + "learning_rate": 2.089352020385876e-06, + "loss": 2.1306, + "step": 16727 + }, + { + "epoch": 3.04, + "learning_rate": 2.0892974153622134e-06, + "loss": 0.4639, + "step": 16728 + }, + { + "epoch": 3.04, + "learning_rate": 2.0892428103385514e-06, + "loss": 0.4681, + "step": 16729 + }, + { + "epoch": 3.04, + "learning_rate": 2.0891882053148893e-06, + "loss": 0.7735, + "step": 16730 + }, + { + "epoch": 3.04, + "learning_rate": 2.089133600291227e-06, + "loss": 0.9359, + "step": 16731 + }, + { + "epoch": 3.04, + "learning_rate": 2.0890789952675648e-06, + "loss": 1.4653, + "step": 16732 + }, + { + "epoch": 3.04, + "learning_rate": 2.0890243902439023e-06, + "loss": 0.3146, + "step": 16733 + }, + { + "epoch": 3.04, + "learning_rate": 2.0889697852202402e-06, + "loss": 0.6986, + "step": 16734 + }, + { + "epoch": 3.04, + "learning_rate": 2.088915180196578e-06, + "loss": 1.7831, + "step": 16735 + }, + { + "epoch": 3.04, + "learning_rate": 2.0888605751729157e-06, + "loss": 0.1477, + "step": 16736 + }, + { + "epoch": 3.04, + "learning_rate": 2.0888059701492536e-06, + "loss": 1.986, + "step": 16737 + }, + { + "epoch": 3.04, + "learning_rate": 2.0887513651255916e-06, + "loss": 1.1566, + "step": 16738 + }, + { + "epoch": 3.04, + "learning_rate": 2.0886967601019295e-06, + "loss": 0.6929, + "step": 16739 + }, + { + "epoch": 3.04, + "learning_rate": 2.0886421550782675e-06, + "loss": 1.094, + "step": 16740 + }, + { + "epoch": 3.04, + "learning_rate": 2.0885875500546054e-06, + "loss": 2.9236, + "step": 16741 + }, + { + "epoch": 3.04, + "learning_rate": 2.088532945030943e-06, + "loss": 0.1447, + "step": 16742 + }, + { + "epoch": 3.04, + "learning_rate": 2.088478340007281e-06, + "loss": 0.4697, + "step": 16743 + }, + { + "epoch": 3.04, + "learning_rate": 2.0884237349836184e-06, + "loss": 0.4064, + "step": 16744 + }, + { + "epoch": 3.05, + "learning_rate": 2.0883691299599563e-06, + "loss": 0.0178, + "step": 16745 + }, + { + "epoch": 3.05, + "learning_rate": 2.0883145249362943e-06, + "loss": 0.5526, + "step": 16746 + }, + { + "epoch": 3.05, + "learning_rate": 2.088259919912632e-06, + "loss": 0.329, + "step": 16747 + }, + { + "epoch": 3.05, + "learning_rate": 2.0882053148889697e-06, + "loss": 2.2081, + "step": 16748 + }, + { + "epoch": 3.05, + "learning_rate": 2.0881507098653077e-06, + "loss": 1.4493, + "step": 16749 + }, + { + "epoch": 3.05, + "learning_rate": 2.088096104841645e-06, + "loss": 0.3425, + "step": 16750 + }, + { + "epoch": 3.05, + "learning_rate": 2.088041499817983e-06, + "loss": 0.4268, + "step": 16751 + }, + { + "epoch": 3.05, + "learning_rate": 2.087986894794321e-06, + "loss": 1.336, + "step": 16752 + }, + { + "epoch": 3.05, + "learning_rate": 2.087932289770659e-06, + "loss": 0.206, + "step": 16753 + }, + { + "epoch": 3.05, + "learning_rate": 2.087877684746997e-06, + "loss": 0.9779, + "step": 16754 + }, + { + "epoch": 3.05, + "learning_rate": 2.087823079723335e-06, + "loss": 2.7465, + "step": 16755 + }, + { + "epoch": 3.05, + "learning_rate": 2.0877684746996724e-06, + "loss": 0.5976, + "step": 16756 + }, + { + "epoch": 3.05, + "learning_rate": 2.0877138696760104e-06, + "loss": 0.0948, + "step": 16757 + }, + { + "epoch": 3.05, + "learning_rate": 2.087659264652348e-06, + "loss": 0.9871, + "step": 16758 + }, + { + "epoch": 3.05, + "learning_rate": 2.087604659628686e-06, + "loss": 0.1368, + "step": 16759 + }, + { + "epoch": 3.05, + "learning_rate": 2.087550054605024e-06, + "loss": 0.0323, + "step": 16760 + }, + { + "epoch": 3.05, + "learning_rate": 2.0874954495813613e-06, + "loss": 0.9329, + "step": 16761 + }, + { + "epoch": 3.05, + "learning_rate": 2.0874408445576993e-06, + "loss": 0.2417, + "step": 16762 + }, + { + "epoch": 3.05, + "learning_rate": 2.087386239534037e-06, + "loss": 0.1138, + "step": 16763 + }, + { + "epoch": 3.05, + "learning_rate": 2.0873316345103747e-06, + "loss": 1.8064, + "step": 16764 + }, + { + "epoch": 3.05, + "learning_rate": 2.0872770294867127e-06, + "loss": 1.0712, + "step": 16765 + }, + { + "epoch": 3.05, + "learning_rate": 2.0872224244630506e-06, + "loss": 0.4018, + "step": 16766 + }, + { + "epoch": 3.05, + "learning_rate": 2.0871678194393885e-06, + "loss": 1.424, + "step": 16767 + }, + { + "epoch": 3.05, + "learning_rate": 2.0871132144157265e-06, + "loss": 1.6536, + "step": 16768 + }, + { + "epoch": 3.05, + "learning_rate": 2.0870586093920644e-06, + "loss": 1.4356, + "step": 16769 + }, + { + "epoch": 3.05, + "learning_rate": 2.087004004368402e-06, + "loss": 0.76, + "step": 16770 + }, + { + "epoch": 3.05, + "learning_rate": 2.08694939934474e-06, + "loss": 1.4289, + "step": 16771 + }, + { + "epoch": 3.05, + "learning_rate": 2.0868947943210774e-06, + "loss": 0.8364, + "step": 16772 + }, + { + "epoch": 3.05, + "learning_rate": 2.0868401892974154e-06, + "loss": 0.0334, + "step": 16773 + }, + { + "epoch": 3.05, + "learning_rate": 2.0867855842737533e-06, + "loss": 0.6939, + "step": 16774 + }, + { + "epoch": 3.05, + "learning_rate": 2.086730979250091e-06, + "loss": 0.5126, + "step": 16775 + }, + { + "epoch": 3.05, + "learning_rate": 2.0866763742264288e-06, + "loss": 0.9967, + "step": 16776 + }, + { + "epoch": 3.05, + "learning_rate": 2.0866217692027667e-06, + "loss": 1.2674, + "step": 16777 + }, + { + "epoch": 3.05, + "learning_rate": 2.0865671641791042e-06, + "loss": 4.9112, + "step": 16778 + }, + { + "epoch": 3.05, + "learning_rate": 2.086512559155442e-06, + "loss": 0.6727, + "step": 16779 + }, + { + "epoch": 3.05, + "learning_rate": 2.08645795413178e-06, + "loss": 1.9715, + "step": 16780 + }, + { + "epoch": 3.05, + "learning_rate": 2.086403349108118e-06, + "loss": 0.9381, + "step": 16781 + }, + { + "epoch": 3.05, + "learning_rate": 2.086348744084456e-06, + "loss": 2.3019, + "step": 16782 + }, + { + "epoch": 3.05, + "learning_rate": 2.086294139060794e-06, + "loss": 0.5192, + "step": 16783 + }, + { + "epoch": 3.05, + "learning_rate": 2.0862395340371315e-06, + "loss": 1.245, + "step": 16784 + }, + { + "epoch": 3.05, + "learning_rate": 2.0861849290134694e-06, + "loss": 0.8548, + "step": 16785 + }, + { + "epoch": 3.05, + "learning_rate": 2.086130323989807e-06, + "loss": 1.6536, + "step": 16786 + }, + { + "epoch": 3.05, + "learning_rate": 2.086075718966145e-06, + "loss": 1.5854, + "step": 16787 + }, + { + "epoch": 3.05, + "learning_rate": 2.086021113942483e-06, + "loss": 1.1242, + "step": 16788 + }, + { + "epoch": 3.05, + "learning_rate": 2.0859665089188203e-06, + "loss": 0.0157, + "step": 16789 + }, + { + "epoch": 3.05, + "learning_rate": 2.0859119038951583e-06, + "loss": 0.7453, + "step": 16790 + }, + { + "epoch": 3.05, + "learning_rate": 2.0858572988714962e-06, + "loss": 1.0106, + "step": 16791 + }, + { + "epoch": 3.05, + "learning_rate": 2.0858026938478337e-06, + "loss": 0.0173, + "step": 16792 + }, + { + "epoch": 3.05, + "learning_rate": 2.0857480888241717e-06, + "loss": 0.0234, + "step": 16793 + }, + { + "epoch": 3.05, + "learning_rate": 2.0856934838005096e-06, + "loss": 0.0949, + "step": 16794 + }, + { + "epoch": 3.05, + "learning_rate": 2.0856388787768476e-06, + "loss": 1.6082, + "step": 16795 + }, + { + "epoch": 3.05, + "learning_rate": 2.0855842737531855e-06, + "loss": 0.4412, + "step": 16796 + }, + { + "epoch": 3.05, + "learning_rate": 2.0855296687295235e-06, + "loss": 0.7967, + "step": 16797 + }, + { + "epoch": 3.05, + "learning_rate": 2.085475063705861e-06, + "loss": 0.1245, + "step": 16798 + }, + { + "epoch": 3.05, + "learning_rate": 2.085420458682199e-06, + "loss": 1.7814, + "step": 16799 + }, + { + "epoch": 3.06, + "learning_rate": 2.0853658536585364e-06, + "loss": 0.2455, + "step": 16800 + }, + { + "epoch": 3.06, + "learning_rate": 2.0853112486348744e-06, + "loss": 0.4202, + "step": 16801 + }, + { + "epoch": 3.06, + "learning_rate": 2.0852566436112123e-06, + "loss": 0.1719, + "step": 16802 + }, + { + "epoch": 3.06, + "learning_rate": 2.08520203858755e-06, + "loss": 1.6524, + "step": 16803 + }, + { + "epoch": 3.06, + "learning_rate": 2.085147433563888e-06, + "loss": 0.9941, + "step": 16804 + }, + { + "epoch": 3.06, + "learning_rate": 2.0850928285402257e-06, + "loss": 0.592, + "step": 16805 + }, + { + "epoch": 3.06, + "learning_rate": 2.0850382235165637e-06, + "loss": 1.6514, + "step": 16806 + }, + { + "epoch": 3.06, + "learning_rate": 2.0849836184929016e-06, + "loss": 0.3716, + "step": 16807 + }, + { + "epoch": 3.06, + "learning_rate": 2.0849290134692396e-06, + "loss": 0.7567, + "step": 16808 + }, + { + "epoch": 3.06, + "learning_rate": 2.084874408445577e-06, + "loss": 0.626, + "step": 16809 + }, + { + "epoch": 3.06, + "learning_rate": 2.084819803421915e-06, + "loss": 0.0566, + "step": 16810 + }, + { + "epoch": 3.06, + "learning_rate": 2.084765198398253e-06, + "loss": 0.7786, + "step": 16811 + }, + { + "epoch": 3.06, + "learning_rate": 2.0847105933745905e-06, + "loss": 0.4709, + "step": 16812 + }, + { + "epoch": 3.06, + "learning_rate": 2.0846559883509284e-06, + "loss": 1.4617, + "step": 16813 + }, + { + "epoch": 3.06, + "learning_rate": 2.084601383327266e-06, + "loss": 1.3134, + "step": 16814 + }, + { + "epoch": 3.06, + "learning_rate": 2.084546778303604e-06, + "loss": 0.7404, + "step": 16815 + }, + { + "epoch": 3.06, + "learning_rate": 2.084492173279942e-06, + "loss": 1.3642, + "step": 16816 + }, + { + "epoch": 3.06, + "learning_rate": 2.0844375682562794e-06, + "loss": 1.7856, + "step": 16817 + }, + { + "epoch": 3.06, + "learning_rate": 2.0843829632326173e-06, + "loss": 0.6456, + "step": 16818 + }, + { + "epoch": 3.06, + "learning_rate": 2.0843283582089552e-06, + "loss": 0.4654, + "step": 16819 + }, + { + "epoch": 3.06, + "learning_rate": 2.084273753185293e-06, + "loss": 0.9832, + "step": 16820 + }, + { + "epoch": 3.06, + "learning_rate": 2.084219148161631e-06, + "loss": 1.4877, + "step": 16821 + }, + { + "epoch": 3.06, + "learning_rate": 2.084164543137969e-06, + "loss": 1.7756, + "step": 16822 + }, + { + "epoch": 3.06, + "learning_rate": 2.0841099381143066e-06, + "loss": 0.557, + "step": 16823 + }, + { + "epoch": 3.06, + "learning_rate": 2.0840553330906445e-06, + "loss": 0.188, + "step": 16824 + }, + { + "epoch": 3.06, + "learning_rate": 2.084000728066982e-06, + "loss": 1.6601, + "step": 16825 + }, + { + "epoch": 3.06, + "learning_rate": 2.08394612304332e-06, + "loss": 0.4599, + "step": 16826 + }, + { + "epoch": 3.06, + "learning_rate": 2.083891518019658e-06, + "loss": 0.4941, + "step": 16827 + }, + { + "epoch": 3.06, + "learning_rate": 2.0838369129959955e-06, + "loss": 0.8225, + "step": 16828 + }, + { + "epoch": 3.06, + "learning_rate": 2.0837823079723334e-06, + "loss": 1.0697, + "step": 16829 + }, + { + "epoch": 3.06, + "learning_rate": 2.0837277029486714e-06, + "loss": 0.1972, + "step": 16830 + }, + { + "epoch": 3.06, + "learning_rate": 2.083673097925009e-06, + "loss": 0.828, + "step": 16831 + }, + { + "epoch": 3.06, + "learning_rate": 2.083618492901347e-06, + "loss": 1.5918, + "step": 16832 + }, + { + "epoch": 3.06, + "learning_rate": 2.0835638878776848e-06, + "loss": 0.1329, + "step": 16833 + }, + { + "epoch": 3.06, + "learning_rate": 2.0835092828540227e-06, + "loss": 0.3702, + "step": 16834 + }, + { + "epoch": 3.06, + "learning_rate": 2.0834546778303606e-06, + "loss": 0.7398, + "step": 16835 + }, + { + "epoch": 3.06, + "learning_rate": 2.0834000728066986e-06, + "loss": 1.7113, + "step": 16836 + }, + { + "epoch": 3.06, + "learning_rate": 2.083345467783036e-06, + "loss": 1.4353, + "step": 16837 + }, + { + "epoch": 3.06, + "learning_rate": 2.083290862759374e-06, + "loss": 0.3311, + "step": 16838 + }, + { + "epoch": 3.06, + "learning_rate": 2.0832362577357116e-06, + "loss": 0.2773, + "step": 16839 + }, + { + "epoch": 3.06, + "learning_rate": 2.0831816527120495e-06, + "loss": 1.2368, + "step": 16840 + }, + { + "epoch": 3.06, + "learning_rate": 2.0831270476883875e-06, + "loss": 0.337, + "step": 16841 + }, + { + "epoch": 3.06, + "learning_rate": 2.083072442664725e-06, + "loss": 0.0799, + "step": 16842 + }, + { + "epoch": 3.06, + "learning_rate": 2.083017837641063e-06, + "loss": 1.1382, + "step": 16843 + }, + { + "epoch": 3.06, + "learning_rate": 2.082963232617401e-06, + "loss": 0.0474, + "step": 16844 + }, + { + "epoch": 3.06, + "learning_rate": 2.0829086275937384e-06, + "loss": 0.0985, + "step": 16845 + }, + { + "epoch": 3.06, + "learning_rate": 2.0828540225700763e-06, + "loss": 0.479, + "step": 16846 + }, + { + "epoch": 3.06, + "learning_rate": 2.0827994175464143e-06, + "loss": 0.1427, + "step": 16847 + }, + { + "epoch": 3.06, + "learning_rate": 2.0827448125227522e-06, + "loss": 1.7535, + "step": 16848 + }, + { + "epoch": 3.06, + "learning_rate": 2.08269020749909e-06, + "loss": 2.1272, + "step": 16849 + }, + { + "epoch": 3.06, + "learning_rate": 2.082635602475428e-06, + "loss": 0.3376, + "step": 16850 + }, + { + "epoch": 3.06, + "learning_rate": 2.0825809974517656e-06, + "loss": 0.0276, + "step": 16851 + }, + { + "epoch": 3.06, + "learning_rate": 2.0825263924281036e-06, + "loss": 3.7002, + "step": 16852 + }, + { + "epoch": 3.06, + "learning_rate": 2.082471787404441e-06, + "loss": 1.4319, + "step": 16853 + }, + { + "epoch": 3.06, + "learning_rate": 2.082417182380779e-06, + "loss": 0.3109, + "step": 16854 + }, + { + "epoch": 3.07, + "learning_rate": 2.082362577357117e-06, + "loss": 0.5759, + "step": 16855 + }, + { + "epoch": 3.07, + "learning_rate": 2.0823079723334545e-06, + "loss": 0.5847, + "step": 16856 + }, + { + "epoch": 3.07, + "learning_rate": 2.0822533673097924e-06, + "loss": 0.8488, + "step": 16857 + }, + { + "epoch": 3.07, + "learning_rate": 2.0821987622861304e-06, + "loss": 1.2205, + "step": 16858 + }, + { + "epoch": 3.07, + "learning_rate": 2.082144157262468e-06, + "loss": 0.4599, + "step": 16859 + }, + { + "epoch": 3.07, + "learning_rate": 2.082089552238806e-06, + "loss": 1.5409, + "step": 16860 + }, + { + "epoch": 3.07, + "learning_rate": 2.082034947215144e-06, + "loss": 0.4134, + "step": 16861 + }, + { + "epoch": 3.07, + "learning_rate": 2.0819803421914817e-06, + "loss": 0.1339, + "step": 16862 + }, + { + "epoch": 3.07, + "learning_rate": 2.0819257371678197e-06, + "loss": 0.3978, + "step": 16863 + }, + { + "epoch": 3.07, + "learning_rate": 2.0818711321441576e-06, + "loss": 0.2868, + "step": 16864 + }, + { + "epoch": 3.07, + "learning_rate": 2.081816527120495e-06, + "loss": 2.469, + "step": 16865 + }, + { + "epoch": 3.07, + "learning_rate": 2.081761922096833e-06, + "loss": 0.6083, + "step": 16866 + }, + { + "epoch": 3.07, + "learning_rate": 2.0817073170731706e-06, + "loss": 2.921, + "step": 16867 + }, + { + "epoch": 3.07, + "learning_rate": 2.0816527120495085e-06, + "loss": 0.8393, + "step": 16868 + }, + { + "epoch": 3.07, + "learning_rate": 2.0815981070258465e-06, + "loss": 2.1629, + "step": 16869 + }, + { + "epoch": 3.07, + "learning_rate": 2.081543502002184e-06, + "loss": 0.0671, + "step": 16870 + }, + { + "epoch": 3.07, + "learning_rate": 2.081488896978522e-06, + "loss": 0.761, + "step": 16871 + }, + { + "epoch": 3.07, + "learning_rate": 2.08143429195486e-06, + "loss": 0.5361, + "step": 16872 + }, + { + "epoch": 3.07, + "learning_rate": 2.0813796869311974e-06, + "loss": 1.7169, + "step": 16873 + }, + { + "epoch": 3.07, + "learning_rate": 2.0813250819075354e-06, + "loss": 1.8654, + "step": 16874 + }, + { + "epoch": 3.07, + "learning_rate": 2.0812704768838733e-06, + "loss": 0.425, + "step": 16875 + }, + { + "epoch": 3.07, + "learning_rate": 2.0812158718602112e-06, + "loss": 0.0165, + "step": 16876 + }, + { + "epoch": 3.07, + "learning_rate": 2.081161266836549e-06, + "loss": 0.8432, + "step": 16877 + }, + { + "epoch": 3.07, + "learning_rate": 2.081106661812887e-06, + "loss": 0.3954, + "step": 16878 + }, + { + "epoch": 3.07, + "learning_rate": 2.0810520567892247e-06, + "loss": 0.3784, + "step": 16879 + }, + { + "epoch": 3.07, + "learning_rate": 2.0809974517655626e-06, + "loss": 0.9931, + "step": 16880 + }, + { + "epoch": 3.07, + "learning_rate": 2.0809428467419e-06, + "loss": 0.1541, + "step": 16881 + }, + { + "epoch": 3.07, + "learning_rate": 2.080888241718238e-06, + "loss": 3.2201, + "step": 16882 + }, + { + "epoch": 3.07, + "learning_rate": 2.080833636694576e-06, + "loss": 0.9719, + "step": 16883 + }, + { + "epoch": 3.07, + "learning_rate": 2.0807790316709135e-06, + "loss": 0.0624, + "step": 16884 + }, + { + "epoch": 3.07, + "learning_rate": 2.0807244266472515e-06, + "loss": 0.6796, + "step": 16885 + }, + { + "epoch": 3.07, + "learning_rate": 2.0806698216235894e-06, + "loss": 1.4752, + "step": 16886 + }, + { + "epoch": 3.07, + "learning_rate": 2.0806152165999274e-06, + "loss": 1.5226, + "step": 16887 + }, + { + "epoch": 3.07, + "learning_rate": 2.0805606115762653e-06, + "loss": 1.1347, + "step": 16888 + }, + { + "epoch": 3.07, + "learning_rate": 2.0805060065526032e-06, + "loss": 1.5489, + "step": 16889 + }, + { + "epoch": 3.07, + "learning_rate": 2.0804514015289408e-06, + "loss": 1.6697, + "step": 16890 + }, + { + "epoch": 3.07, + "learning_rate": 2.0803967965052787e-06, + "loss": 1.598, + "step": 16891 + }, + { + "epoch": 3.07, + "learning_rate": 2.0803421914816162e-06, + "loss": 2.0633, + "step": 16892 + }, + { + "epoch": 3.07, + "learning_rate": 2.080287586457954e-06, + "loss": 0.8886, + "step": 16893 + }, + { + "epoch": 3.07, + "learning_rate": 2.080232981434292e-06, + "loss": 1.897, + "step": 16894 + }, + { + "epoch": 3.07, + "learning_rate": 2.0801783764106296e-06, + "loss": 0.0408, + "step": 16895 + }, + { + "epoch": 3.07, + "learning_rate": 2.0801237713869676e-06, + "loss": 2.0189, + "step": 16896 + }, + { + "epoch": 3.07, + "learning_rate": 2.0800691663633055e-06, + "loss": 1.3909, + "step": 16897 + }, + { + "epoch": 3.07, + "learning_rate": 2.080014561339643e-06, + "loss": 0.1819, + "step": 16898 + }, + { + "epoch": 3.07, + "learning_rate": 2.079959956315981e-06, + "loss": 0.3611, + "step": 16899 + }, + { + "epoch": 3.07, + "learning_rate": 2.079905351292319e-06, + "loss": 2.7076, + "step": 16900 + }, + { + "epoch": 3.07, + "learning_rate": 2.079850746268657e-06, + "loss": 0.2617, + "step": 16901 + }, + { + "epoch": 3.07, + "learning_rate": 2.079796141244995e-06, + "loss": 0.5104, + "step": 16902 + }, + { + "epoch": 3.07, + "learning_rate": 2.0797415362213328e-06, + "loss": 0.0509, + "step": 16903 + }, + { + "epoch": 3.07, + "learning_rate": 2.0796869311976703e-06, + "loss": 1.1973, + "step": 16904 + }, + { + "epoch": 3.07, + "learning_rate": 2.0796323261740082e-06, + "loss": 2.2421, + "step": 16905 + }, + { + "epoch": 3.07, + "learning_rate": 2.0795777211503457e-06, + "loss": 0.0317, + "step": 16906 + }, + { + "epoch": 3.07, + "learning_rate": 2.0795231161266837e-06, + "loss": 0.7388, + "step": 16907 + }, + { + "epoch": 3.07, + "learning_rate": 2.0794685111030216e-06, + "loss": 0.1773, + "step": 16908 + }, + { + "epoch": 3.07, + "learning_rate": 2.079413906079359e-06, + "loss": 0.1548, + "step": 16909 + }, + { + "epoch": 3.08, + "learning_rate": 2.079359301055697e-06, + "loss": 0.2292, + "step": 16910 + }, + { + "epoch": 3.08, + "learning_rate": 2.079304696032035e-06, + "loss": 2.5109, + "step": 16911 + }, + { + "epoch": 3.08, + "learning_rate": 2.0792500910083725e-06, + "loss": 0.2529, + "step": 16912 + }, + { + "epoch": 3.08, + "learning_rate": 2.0791954859847105e-06, + "loss": 1.0598, + "step": 16913 + }, + { + "epoch": 3.08, + "learning_rate": 2.0791408809610484e-06, + "loss": 0.3605, + "step": 16914 + }, + { + "epoch": 3.08, + "learning_rate": 2.0790862759373864e-06, + "loss": 0.0766, + "step": 16915 + }, + { + "epoch": 3.08, + "learning_rate": 2.0790316709137243e-06, + "loss": 0.3422, + "step": 16916 + }, + { + "epoch": 3.08, + "learning_rate": 2.0789770658900623e-06, + "loss": 0.3534, + "step": 16917 + }, + { + "epoch": 3.08, + "learning_rate": 2.0789224608663998e-06, + "loss": 0.7524, + "step": 16918 + }, + { + "epoch": 3.08, + "learning_rate": 2.0788678558427377e-06, + "loss": 0.7463, + "step": 16919 + }, + { + "epoch": 3.08, + "learning_rate": 2.0788132508190752e-06, + "loss": 0.858, + "step": 16920 + }, + { + "epoch": 3.08, + "learning_rate": 2.078758645795413e-06, + "loss": 0.3309, + "step": 16921 + }, + { + "epoch": 3.08, + "learning_rate": 2.078704040771751e-06, + "loss": 0.2151, + "step": 16922 + }, + { + "epoch": 3.08, + "learning_rate": 2.0786494357480887e-06, + "loss": 0.3823, + "step": 16923 + }, + { + "epoch": 3.08, + "learning_rate": 2.0785948307244266e-06, + "loss": 0.5294, + "step": 16924 + }, + { + "epoch": 3.08, + "learning_rate": 2.0785402257007645e-06, + "loss": 0.3561, + "step": 16925 + }, + { + "epoch": 3.08, + "learning_rate": 2.078485620677102e-06, + "loss": 0.5737, + "step": 16926 + }, + { + "epoch": 3.08, + "learning_rate": 2.07843101565344e-06, + "loss": 0.0866, + "step": 16927 + }, + { + "epoch": 3.08, + "learning_rate": 2.078376410629778e-06, + "loss": 1.762, + "step": 16928 + }, + { + "epoch": 3.08, + "learning_rate": 2.078321805606116e-06, + "loss": 2.2435, + "step": 16929 + }, + { + "epoch": 3.08, + "learning_rate": 2.078267200582454e-06, + "loss": 0.5146, + "step": 16930 + }, + { + "epoch": 3.08, + "learning_rate": 2.0782125955587918e-06, + "loss": 0.5532, + "step": 16931 + }, + { + "epoch": 3.08, + "learning_rate": 2.0781579905351293e-06, + "loss": 0.0434, + "step": 16932 + }, + { + "epoch": 3.08, + "learning_rate": 2.0781033855114672e-06, + "loss": 0.7618, + "step": 16933 + }, + { + "epoch": 3.08, + "learning_rate": 2.0780487804878048e-06, + "loss": 0.5779, + "step": 16934 + }, + { + "epoch": 3.08, + "learning_rate": 2.0779941754641427e-06, + "loss": 0.4486, + "step": 16935 + }, + { + "epoch": 3.08, + "learning_rate": 2.0779395704404807e-06, + "loss": 0.0303, + "step": 16936 + }, + { + "epoch": 3.08, + "learning_rate": 2.077884965416818e-06, + "loss": 0.9349, + "step": 16937 + }, + { + "epoch": 3.08, + "learning_rate": 2.077830360393156e-06, + "loss": 0.3703, + "step": 16938 + }, + { + "epoch": 3.08, + "learning_rate": 2.077775755369494e-06, + "loss": 0.4499, + "step": 16939 + }, + { + "epoch": 3.08, + "learning_rate": 2.0777211503458316e-06, + "loss": 0.0236, + "step": 16940 + }, + { + "epoch": 3.08, + "learning_rate": 2.0776665453221695e-06, + "loss": 1.7088, + "step": 16941 + }, + { + "epoch": 3.08, + "learning_rate": 2.0776119402985075e-06, + "loss": 1.9751, + "step": 16942 + }, + { + "epoch": 3.08, + "learning_rate": 2.0775573352748454e-06, + "loss": 0.6051, + "step": 16943 + }, + { + "epoch": 3.08, + "learning_rate": 2.0775027302511834e-06, + "loss": 1.1501, + "step": 16944 + }, + { + "epoch": 3.08, + "learning_rate": 2.0774481252275213e-06, + "loss": 0.9747, + "step": 16945 + }, + { + "epoch": 3.08, + "learning_rate": 2.077393520203859e-06, + "loss": 0.3558, + "step": 16946 + }, + { + "epoch": 3.08, + "learning_rate": 2.0773389151801968e-06, + "loss": 0.2346, + "step": 16947 + }, + { + "epoch": 3.08, + "learning_rate": 2.0772843101565343e-06, + "loss": 0.7814, + "step": 16948 + }, + { + "epoch": 3.08, + "learning_rate": 2.0772297051328722e-06, + "loss": 0.3212, + "step": 16949 + }, + { + "epoch": 3.08, + "learning_rate": 2.07717510010921e-06, + "loss": 0.8083, + "step": 16950 + }, + { + "epoch": 3.08, + "learning_rate": 2.0771204950855477e-06, + "loss": 1.3712, + "step": 16951 + }, + { + "epoch": 3.08, + "learning_rate": 2.0770658900618856e-06, + "loss": 0.0862, + "step": 16952 + }, + { + "epoch": 3.08, + "learning_rate": 2.0770112850382236e-06, + "loss": 0.0244, + "step": 16953 + }, + { + "epoch": 3.08, + "learning_rate": 2.076956680014561e-06, + "loss": 2.4437, + "step": 16954 + }, + { + "epoch": 3.08, + "learning_rate": 2.076902074990899e-06, + "loss": 0.2668, + "step": 16955 + }, + { + "epoch": 3.08, + "learning_rate": 2.076847469967237e-06, + "loss": 0.0501, + "step": 16956 + }, + { + "epoch": 3.08, + "learning_rate": 2.076792864943575e-06, + "loss": 0.7108, + "step": 16957 + }, + { + "epoch": 3.08, + "learning_rate": 2.076738259919913e-06, + "loss": 0.2216, + "step": 16958 + }, + { + "epoch": 3.08, + "learning_rate": 2.076683654896251e-06, + "loss": 0.8726, + "step": 16959 + }, + { + "epoch": 3.08, + "learning_rate": 2.0766290498725883e-06, + "loss": 0.6759, + "step": 16960 + }, + { + "epoch": 3.08, + "learning_rate": 2.0765744448489263e-06, + "loss": 0.548, + "step": 16961 + }, + { + "epoch": 3.08, + "learning_rate": 2.076519839825264e-06, + "loss": 0.2917, + "step": 16962 + }, + { + "epoch": 3.08, + "learning_rate": 2.0764652348016017e-06, + "loss": 2.5316, + "step": 16963 + }, + { + "epoch": 3.08, + "learning_rate": 2.0764106297779397e-06, + "loss": 0.3528, + "step": 16964 + }, + { + "epoch": 3.09, + "learning_rate": 2.076356024754277e-06, + "loss": 0.7656, + "step": 16965 + }, + { + "epoch": 3.09, + "learning_rate": 2.076301419730615e-06, + "loss": 1.4484, + "step": 16966 + }, + { + "epoch": 3.09, + "learning_rate": 2.076246814706953e-06, + "loss": 0.3678, + "step": 16967 + }, + { + "epoch": 3.09, + "learning_rate": 2.076192209683291e-06, + "loss": 1.331, + "step": 16968 + }, + { + "epoch": 3.09, + "learning_rate": 2.076137604659629e-06, + "loss": 0.5978, + "step": 16969 + }, + { + "epoch": 3.09, + "learning_rate": 2.076082999635967e-06, + "loss": 0.4233, + "step": 16970 + }, + { + "epoch": 3.09, + "learning_rate": 2.0760283946123044e-06, + "loss": 1.1404, + "step": 16971 + }, + { + "epoch": 3.09, + "learning_rate": 2.0759737895886424e-06, + "loss": 0.1382, + "step": 16972 + }, + { + "epoch": 3.09, + "learning_rate": 2.07591918456498e-06, + "loss": 0.6638, + "step": 16973 + }, + { + "epoch": 3.09, + "learning_rate": 2.075864579541318e-06, + "loss": 1.1254, + "step": 16974 + }, + { + "epoch": 3.09, + "learning_rate": 2.0758099745176558e-06, + "loss": 0.87, + "step": 16975 + }, + { + "epoch": 3.09, + "learning_rate": 2.0757553694939933e-06, + "loss": 0.9379, + "step": 16976 + }, + { + "epoch": 3.09, + "learning_rate": 2.0757007644703312e-06, + "loss": 1.5626, + "step": 16977 + }, + { + "epoch": 3.09, + "learning_rate": 2.075646159446669e-06, + "loss": 0.0199, + "step": 16978 + }, + { + "epoch": 3.09, + "learning_rate": 2.0755915544230067e-06, + "loss": 0.0152, + "step": 16979 + }, + { + "epoch": 3.09, + "learning_rate": 2.0755369493993447e-06, + "loss": 0.4599, + "step": 16980 + }, + { + "epoch": 3.09, + "learning_rate": 2.0754823443756826e-06, + "loss": 0.0465, + "step": 16981 + }, + { + "epoch": 3.09, + "learning_rate": 2.0754277393520205e-06, + "loss": 1.4466, + "step": 16982 + }, + { + "epoch": 3.09, + "learning_rate": 2.0753731343283585e-06, + "loss": 0.1958, + "step": 16983 + }, + { + "epoch": 3.09, + "learning_rate": 2.0753185293046964e-06, + "loss": 0.5815, + "step": 16984 + }, + { + "epoch": 3.09, + "learning_rate": 2.075263924281034e-06, + "loss": 7.0344, + "step": 16985 + }, + { + "epoch": 3.09, + "learning_rate": 2.075209319257372e-06, + "loss": 0.3588, + "step": 16986 + }, + { + "epoch": 3.09, + "learning_rate": 2.0751547142337094e-06, + "loss": 0.7187, + "step": 16987 + }, + { + "epoch": 3.09, + "learning_rate": 2.0751001092100474e-06, + "loss": 0.3682, + "step": 16988 + }, + { + "epoch": 3.09, + "learning_rate": 2.0750455041863853e-06, + "loss": 0.1313, + "step": 16989 + }, + { + "epoch": 3.09, + "learning_rate": 2.074990899162723e-06, + "loss": 1.9298, + "step": 16990 + }, + { + "epoch": 3.09, + "learning_rate": 2.0749362941390608e-06, + "loss": 0.0299, + "step": 16991 + }, + { + "epoch": 3.09, + "learning_rate": 2.0748816891153987e-06, + "loss": 0.8419, + "step": 16992 + }, + { + "epoch": 3.09, + "learning_rate": 2.0748270840917362e-06, + "loss": 1.104, + "step": 16993 + }, + { + "epoch": 3.09, + "learning_rate": 2.074772479068074e-06, + "loss": 0.6255, + "step": 16994 + }, + { + "epoch": 3.09, + "learning_rate": 2.074717874044412e-06, + "loss": 0.2456, + "step": 16995 + }, + { + "epoch": 3.09, + "learning_rate": 2.07466326902075e-06, + "loss": 0.1086, + "step": 16996 + }, + { + "epoch": 3.09, + "learning_rate": 2.074608663997088e-06, + "loss": 1.161, + "step": 16997 + }, + { + "epoch": 3.09, + "learning_rate": 2.074554058973426e-06, + "loss": 0.9017, + "step": 16998 + }, + { + "epoch": 3.09, + "learning_rate": 2.0744994539497635e-06, + "loss": 0.2111, + "step": 16999 + }, + { + "epoch": 3.09, + "learning_rate": 2.0744448489261014e-06, + "loss": 0.1269, + "step": 17000 + }, + { + "epoch": 3.09, + "learning_rate": 2.074390243902439e-06, + "loss": 1.3337, + "step": 17001 + }, + { + "epoch": 3.09, + "learning_rate": 2.074335638878777e-06, + "loss": 0.5217, + "step": 17002 + }, + { + "epoch": 3.09, + "learning_rate": 2.074281033855115e-06, + "loss": 0.6384, + "step": 17003 + }, + { + "epoch": 3.09, + "learning_rate": 2.0742264288314523e-06, + "loss": 1.3401, + "step": 17004 + }, + { + "epoch": 3.09, + "learning_rate": 2.0741718238077903e-06, + "loss": 0.1639, + "step": 17005 + }, + { + "epoch": 3.09, + "learning_rate": 2.0741172187841282e-06, + "loss": 0.9144, + "step": 17006 + }, + { + "epoch": 3.09, + "learning_rate": 2.0740626137604657e-06, + "loss": 1.3967, + "step": 17007 + }, + { + "epoch": 3.09, + "learning_rate": 2.0740080087368037e-06, + "loss": 0.4148, + "step": 17008 + }, + { + "epoch": 3.09, + "learning_rate": 2.0739534037131416e-06, + "loss": 0.908, + "step": 17009 + }, + { + "epoch": 3.09, + "learning_rate": 2.0738987986894796e-06, + "loss": 0.3203, + "step": 17010 + }, + { + "epoch": 3.09, + "learning_rate": 2.0738441936658175e-06, + "loss": 2.7483, + "step": 17011 + }, + { + "epoch": 3.09, + "learning_rate": 2.0737895886421555e-06, + "loss": 0.2724, + "step": 17012 + }, + { + "epoch": 3.09, + "learning_rate": 2.073734983618493e-06, + "loss": 0.2683, + "step": 17013 + }, + { + "epoch": 3.09, + "learning_rate": 2.073680378594831e-06, + "loss": 1.1688, + "step": 17014 + }, + { + "epoch": 3.09, + "learning_rate": 2.0736257735711684e-06, + "loss": 1.1157, + "step": 17015 + }, + { + "epoch": 3.09, + "learning_rate": 2.0735711685475064e-06, + "loss": 0.8333, + "step": 17016 + }, + { + "epoch": 3.09, + "learning_rate": 2.0735165635238443e-06, + "loss": 0.7506, + "step": 17017 + }, + { + "epoch": 3.09, + "learning_rate": 2.073461958500182e-06, + "loss": 1.7224, + "step": 17018 + }, + { + "epoch": 3.09, + "learning_rate": 2.0734073534765198e-06, + "loss": 1.253, + "step": 17019 + }, + { + "epoch": 3.1, + "learning_rate": 2.0733527484528577e-06, + "loss": 0.9829, + "step": 17020 + }, + { + "epoch": 3.1, + "learning_rate": 2.0732981434291953e-06, + "loss": 0.2346, + "step": 17021 + }, + { + "epoch": 3.1, + "learning_rate": 2.073243538405533e-06, + "loss": 0.031, + "step": 17022 + }, + { + "epoch": 3.1, + "learning_rate": 2.073188933381871e-06, + "loss": 0.5772, + "step": 17023 + }, + { + "epoch": 3.1, + "learning_rate": 2.073134328358209e-06, + "loss": 0.8669, + "step": 17024 + }, + { + "epoch": 3.1, + "learning_rate": 2.073079723334547e-06, + "loss": 0.2122, + "step": 17025 + }, + { + "epoch": 3.1, + "learning_rate": 2.073025118310885e-06, + "loss": 0.7086, + "step": 17026 + }, + { + "epoch": 3.1, + "learning_rate": 2.0729705132872225e-06, + "loss": 0.31, + "step": 17027 + }, + { + "epoch": 3.1, + "learning_rate": 2.0729159082635604e-06, + "loss": 0.6601, + "step": 17028 + }, + { + "epoch": 3.1, + "learning_rate": 2.072861303239898e-06, + "loss": 0.9187, + "step": 17029 + }, + { + "epoch": 3.1, + "learning_rate": 2.072806698216236e-06, + "loss": 0.0248, + "step": 17030 + }, + { + "epoch": 3.1, + "learning_rate": 2.072752093192574e-06, + "loss": 0.5983, + "step": 17031 + }, + { + "epoch": 3.1, + "learning_rate": 2.0726974881689114e-06, + "loss": 0.691, + "step": 17032 + }, + { + "epoch": 3.1, + "learning_rate": 2.0726428831452493e-06, + "loss": 0.4441, + "step": 17033 + }, + { + "epoch": 3.1, + "learning_rate": 2.0725882781215872e-06, + "loss": 0.8123, + "step": 17034 + }, + { + "epoch": 3.1, + "learning_rate": 2.0725336730979248e-06, + "loss": 0.4239, + "step": 17035 + }, + { + "epoch": 3.1, + "learning_rate": 2.0724790680742627e-06, + "loss": 1.8156, + "step": 17036 + }, + { + "epoch": 3.1, + "learning_rate": 2.0724244630506007e-06, + "loss": 0.7108, + "step": 17037 + }, + { + "epoch": 3.1, + "learning_rate": 2.0723698580269386e-06, + "loss": 0.2449, + "step": 17038 + }, + { + "epoch": 3.1, + "learning_rate": 2.0723152530032765e-06, + "loss": 0.1365, + "step": 17039 + }, + { + "epoch": 3.1, + "learning_rate": 2.0722606479796145e-06, + "loss": 0.4553, + "step": 17040 + }, + { + "epoch": 3.1, + "learning_rate": 2.072206042955952e-06, + "loss": 1.3655, + "step": 17041 + }, + { + "epoch": 3.1, + "learning_rate": 2.07215143793229e-06, + "loss": 0.1644, + "step": 17042 + }, + { + "epoch": 3.1, + "learning_rate": 2.0720968329086275e-06, + "loss": 0.7295, + "step": 17043 + }, + { + "epoch": 3.1, + "learning_rate": 2.0720422278849654e-06, + "loss": 0.026, + "step": 17044 + }, + { + "epoch": 3.1, + "learning_rate": 2.0719876228613034e-06, + "loss": 3.1695, + "step": 17045 + }, + { + "epoch": 3.1, + "learning_rate": 2.071933017837641e-06, + "loss": 0.9132, + "step": 17046 + }, + { + "epoch": 3.1, + "learning_rate": 2.071878412813979e-06, + "loss": 3.1117, + "step": 17047 + }, + { + "epoch": 3.1, + "learning_rate": 2.0718238077903168e-06, + "loss": 1.4026, + "step": 17048 + }, + { + "epoch": 3.1, + "learning_rate": 2.0717692027666547e-06, + "loss": 3.23, + "step": 17049 + }, + { + "epoch": 3.1, + "learning_rate": 2.0717145977429926e-06, + "loss": 0.6573, + "step": 17050 + }, + { + "epoch": 3.1, + "learning_rate": 2.07165999271933e-06, + "loss": 0.4223, + "step": 17051 + }, + { + "epoch": 3.1, + "learning_rate": 2.071605387695668e-06, + "loss": 0.9958, + "step": 17052 + }, + { + "epoch": 3.1, + "learning_rate": 2.071550782672006e-06, + "loss": 0.5487, + "step": 17053 + }, + { + "epoch": 3.1, + "learning_rate": 2.0714961776483436e-06, + "loss": 0.2912, + "step": 17054 + }, + { + "epoch": 3.1, + "learning_rate": 2.0714415726246815e-06, + "loss": 0.0251, + "step": 17055 + }, + { + "epoch": 3.1, + "learning_rate": 2.0713869676010195e-06, + "loss": 1.0491, + "step": 17056 + }, + { + "epoch": 3.1, + "learning_rate": 2.071332362577357e-06, + "loss": 0.5451, + "step": 17057 + }, + { + "epoch": 3.1, + "learning_rate": 2.071277757553695e-06, + "loss": 1.5213, + "step": 17058 + }, + { + "epoch": 3.1, + "learning_rate": 2.071223152530033e-06, + "loss": 0.0814, + "step": 17059 + }, + { + "epoch": 3.1, + "learning_rate": 2.0711685475063704e-06, + "loss": 0.8169, + "step": 17060 + }, + { + "epoch": 3.1, + "learning_rate": 2.0711139424827083e-06, + "loss": 1.9838, + "step": 17061 + }, + { + "epoch": 3.1, + "learning_rate": 2.0710593374590463e-06, + "loss": 0.0068, + "step": 17062 + }, + { + "epoch": 3.1, + "learning_rate": 2.0710047324353842e-06, + "loss": 0.6821, + "step": 17063 + }, + { + "epoch": 3.1, + "learning_rate": 2.070950127411722e-06, + "loss": 0.9277, + "step": 17064 + }, + { + "epoch": 3.1, + "learning_rate": 2.07089552238806e-06, + "loss": 0.1125, + "step": 17065 + }, + { + "epoch": 3.1, + "learning_rate": 2.0708409173643976e-06, + "loss": 3.9075, + "step": 17066 + }, + { + "epoch": 3.1, + "learning_rate": 2.0707863123407356e-06, + "loss": 0.7948, + "step": 17067 + }, + { + "epoch": 3.1, + "learning_rate": 2.070731707317073e-06, + "loss": 0.2102, + "step": 17068 + }, + { + "epoch": 3.1, + "learning_rate": 2.070677102293411e-06, + "loss": 0.2345, + "step": 17069 + }, + { + "epoch": 3.1, + "learning_rate": 2.070622497269749e-06, + "loss": 1.867, + "step": 17070 + }, + { + "epoch": 3.1, + "learning_rate": 2.0705678922460865e-06, + "loss": 0.3166, + "step": 17071 + }, + { + "epoch": 3.1, + "learning_rate": 2.0705132872224244e-06, + "loss": 1.6181, + "step": 17072 + }, + { + "epoch": 3.1, + "learning_rate": 2.0704586821987624e-06, + "loss": 1.6446, + "step": 17073 + }, + { + "epoch": 3.1, + "learning_rate": 2.0704040771751e-06, + "loss": 0.652, + "step": 17074 + }, + { + "epoch": 3.11, + "learning_rate": 2.070349472151438e-06, + "loss": 0.2527, + "step": 17075 + }, + { + "epoch": 3.11, + "learning_rate": 2.0702948671277758e-06, + "loss": 1.0808, + "step": 17076 + }, + { + "epoch": 3.11, + "learning_rate": 2.0702402621041137e-06, + "loss": 0.1663, + "step": 17077 + }, + { + "epoch": 3.11, + "learning_rate": 2.0701856570804517e-06, + "loss": 1.0245, + "step": 17078 + }, + { + "epoch": 3.11, + "learning_rate": 2.0701310520567896e-06, + "loss": 0.0267, + "step": 17079 + }, + { + "epoch": 3.11, + "learning_rate": 2.070076447033127e-06, + "loss": 0.6925, + "step": 17080 + }, + { + "epoch": 3.11, + "learning_rate": 2.070021842009465e-06, + "loss": 0.2216, + "step": 17081 + }, + { + "epoch": 3.11, + "learning_rate": 2.0699672369858026e-06, + "loss": 1.3856, + "step": 17082 + }, + { + "epoch": 3.11, + "learning_rate": 2.0699126319621405e-06, + "loss": 0.5255, + "step": 17083 + }, + { + "epoch": 3.11, + "learning_rate": 2.0698580269384785e-06, + "loss": 0.3089, + "step": 17084 + }, + { + "epoch": 3.11, + "learning_rate": 2.069803421914816e-06, + "loss": 0.1018, + "step": 17085 + }, + { + "epoch": 3.11, + "learning_rate": 2.069748816891154e-06, + "loss": 0.7933, + "step": 17086 + }, + { + "epoch": 3.11, + "learning_rate": 2.069694211867492e-06, + "loss": 0.0314, + "step": 17087 + }, + { + "epoch": 3.11, + "learning_rate": 2.0696396068438294e-06, + "loss": 0.0419, + "step": 17088 + }, + { + "epoch": 3.11, + "learning_rate": 2.0695850018201674e-06, + "loss": 0.0516, + "step": 17089 + }, + { + "epoch": 3.11, + "learning_rate": 2.0695303967965053e-06, + "loss": 0.3617, + "step": 17090 + }, + { + "epoch": 3.11, + "learning_rate": 2.0694757917728432e-06, + "loss": 0.0264, + "step": 17091 + }, + { + "epoch": 3.11, + "learning_rate": 2.069421186749181e-06, + "loss": 0.4019, + "step": 17092 + }, + { + "epoch": 3.11, + "learning_rate": 2.069366581725519e-06, + "loss": 0.8511, + "step": 17093 + }, + { + "epoch": 3.11, + "learning_rate": 2.0693119767018566e-06, + "loss": 1.9893, + "step": 17094 + }, + { + "epoch": 3.11, + "learning_rate": 2.0692573716781946e-06, + "loss": 1.0268, + "step": 17095 + }, + { + "epoch": 3.11, + "learning_rate": 2.069202766654532e-06, + "loss": 0.642, + "step": 17096 + }, + { + "epoch": 3.11, + "learning_rate": 2.06914816163087e-06, + "loss": 1.5265, + "step": 17097 + }, + { + "epoch": 3.11, + "learning_rate": 2.069093556607208e-06, + "loss": 1.0119, + "step": 17098 + }, + { + "epoch": 3.11, + "learning_rate": 2.0690389515835455e-06, + "loss": 1.1507, + "step": 17099 + }, + { + "epoch": 3.11, + "learning_rate": 2.0689843465598835e-06, + "loss": 1.5826, + "step": 17100 + }, + { + "epoch": 3.11, + "learning_rate": 2.0689297415362214e-06, + "loss": 1.3994, + "step": 17101 + }, + { + "epoch": 3.11, + "learning_rate": 2.068875136512559e-06, + "loss": 0.0669, + "step": 17102 + }, + { + "epoch": 3.11, + "learning_rate": 2.068820531488897e-06, + "loss": 0.0273, + "step": 17103 + }, + { + "epoch": 3.11, + "learning_rate": 2.068765926465235e-06, + "loss": 0.5145, + "step": 17104 + }, + { + "epoch": 3.11, + "learning_rate": 2.0687113214415728e-06, + "loss": 2.939, + "step": 17105 + }, + { + "epoch": 3.11, + "learning_rate": 2.0686567164179107e-06, + "loss": 0.4739, + "step": 17106 + }, + { + "epoch": 3.11, + "learning_rate": 2.0686021113942486e-06, + "loss": 0.264, + "step": 17107 + }, + { + "epoch": 3.11, + "learning_rate": 2.068547506370586e-06, + "loss": 2.0535, + "step": 17108 + }, + { + "epoch": 3.11, + "learning_rate": 2.068492901346924e-06, + "loss": 1.1992, + "step": 17109 + }, + { + "epoch": 3.11, + "learning_rate": 2.0684382963232616e-06, + "loss": 0.3848, + "step": 17110 + }, + { + "epoch": 3.11, + "learning_rate": 2.0683836912995996e-06, + "loss": 1.8491, + "step": 17111 + }, + { + "epoch": 3.11, + "learning_rate": 2.0683290862759375e-06, + "loss": 0.059, + "step": 17112 + }, + { + "epoch": 3.11, + "learning_rate": 2.068274481252275e-06, + "loss": 2.2116, + "step": 17113 + }, + { + "epoch": 3.11, + "learning_rate": 2.068219876228613e-06, + "loss": 0.9611, + "step": 17114 + }, + { + "epoch": 3.11, + "learning_rate": 2.068165271204951e-06, + "loss": 1.4301, + "step": 17115 + }, + { + "epoch": 3.11, + "learning_rate": 2.0681106661812884e-06, + "loss": 0.5961, + "step": 17116 + }, + { + "epoch": 3.11, + "learning_rate": 2.0680560611576264e-06, + "loss": 1.6763, + "step": 17117 + }, + { + "epoch": 3.11, + "learning_rate": 2.0680014561339643e-06, + "loss": 0.1142, + "step": 17118 + }, + { + "epoch": 3.11, + "learning_rate": 2.0679468511103023e-06, + "loss": 1.9096, + "step": 17119 + }, + { + "epoch": 3.11, + "learning_rate": 2.0678922460866402e-06, + "loss": 1.9237, + "step": 17120 + }, + { + "epoch": 3.11, + "learning_rate": 2.067837641062978e-06, + "loss": 0.9782, + "step": 17121 + }, + { + "epoch": 3.11, + "learning_rate": 2.0677830360393157e-06, + "loss": 1.1752, + "step": 17122 + }, + { + "epoch": 3.11, + "learning_rate": 2.0677284310156536e-06, + "loss": 1.0352, + "step": 17123 + }, + { + "epoch": 3.11, + "learning_rate": 2.067673825991991e-06, + "loss": 0.5336, + "step": 17124 + }, + { + "epoch": 3.11, + "learning_rate": 2.067619220968329e-06, + "loss": 2.2218, + "step": 17125 + }, + { + "epoch": 3.11, + "learning_rate": 2.067564615944667e-06, + "loss": 0.4142, + "step": 17126 + }, + { + "epoch": 3.11, + "learning_rate": 2.0675100109210045e-06, + "loss": 0.3022, + "step": 17127 + }, + { + "epoch": 3.11, + "learning_rate": 2.0674554058973425e-06, + "loss": 0.3941, + "step": 17128 + }, + { + "epoch": 3.11, + "learning_rate": 2.0674008008736804e-06, + "loss": 1.7236, + "step": 17129 + }, + { + "epoch": 3.12, + "learning_rate": 2.067346195850018e-06, + "loss": 0.3087, + "step": 17130 + }, + { + "epoch": 3.12, + "learning_rate": 2.067291590826356e-06, + "loss": 1.002, + "step": 17131 + }, + { + "epoch": 3.12, + "learning_rate": 2.067236985802694e-06, + "loss": 0.0573, + "step": 17132 + }, + { + "epoch": 3.12, + "learning_rate": 2.0671823807790318e-06, + "loss": 1.9017, + "step": 17133 + }, + { + "epoch": 3.12, + "learning_rate": 2.0671277757553697e-06, + "loss": 1.8109, + "step": 17134 + }, + { + "epoch": 3.12, + "learning_rate": 2.0670731707317072e-06, + "loss": 1.8649, + "step": 17135 + }, + { + "epoch": 3.12, + "learning_rate": 2.067018565708045e-06, + "loss": 0.398, + "step": 17136 + }, + { + "epoch": 3.12, + "learning_rate": 2.066963960684383e-06, + "loss": 0.9653, + "step": 17137 + }, + { + "epoch": 3.12, + "learning_rate": 2.0669093556607207e-06, + "loss": 0.9358, + "step": 17138 + }, + { + "epoch": 3.12, + "learning_rate": 2.0668547506370586e-06, + "loss": 0.6271, + "step": 17139 + }, + { + "epoch": 3.12, + "learning_rate": 2.0668001456133965e-06, + "loss": 0.0534, + "step": 17140 + }, + { + "epoch": 3.12, + "learning_rate": 2.066745540589734e-06, + "loss": 1.6176, + "step": 17141 + }, + { + "epoch": 3.12, + "learning_rate": 2.066690935566072e-06, + "loss": 0.8364, + "step": 17142 + }, + { + "epoch": 3.12, + "learning_rate": 2.06663633054241e-06, + "loss": 0.287, + "step": 17143 + }, + { + "epoch": 3.12, + "learning_rate": 2.066581725518748e-06, + "loss": 0.0729, + "step": 17144 + }, + { + "epoch": 3.12, + "learning_rate": 2.066527120495086e-06, + "loss": 0.9843, + "step": 17145 + }, + { + "epoch": 3.12, + "learning_rate": 2.0664725154714238e-06, + "loss": 0.8001, + "step": 17146 + }, + { + "epoch": 3.12, + "learning_rate": 2.0664179104477613e-06, + "loss": 0.013, + "step": 17147 + }, + { + "epoch": 3.12, + "learning_rate": 2.0663633054240992e-06, + "loss": 0.5327, + "step": 17148 + }, + { + "epoch": 3.12, + "learning_rate": 2.0663087004004368e-06, + "loss": 0.0178, + "step": 17149 + }, + { + "epoch": 3.12, + "learning_rate": 2.0662540953767747e-06, + "loss": 0.1189, + "step": 17150 + }, + { + "epoch": 3.12, + "learning_rate": 2.0661994903531126e-06, + "loss": 1.8562, + "step": 17151 + }, + { + "epoch": 3.12, + "learning_rate": 2.06614488532945e-06, + "loss": 0.5083, + "step": 17152 + }, + { + "epoch": 3.12, + "learning_rate": 2.066090280305788e-06, + "loss": 0.5457, + "step": 17153 + }, + { + "epoch": 3.12, + "learning_rate": 2.066035675282126e-06, + "loss": 1.4696, + "step": 17154 + }, + { + "epoch": 3.12, + "learning_rate": 2.0659810702584636e-06, + "loss": 0.6257, + "step": 17155 + }, + { + "epoch": 3.12, + "learning_rate": 2.0659264652348015e-06, + "loss": 0.5223, + "step": 17156 + }, + { + "epoch": 3.12, + "learning_rate": 2.0658718602111395e-06, + "loss": 1.0618, + "step": 17157 + }, + { + "epoch": 3.12, + "learning_rate": 2.0658172551874774e-06, + "loss": 0.5327, + "step": 17158 + }, + { + "epoch": 3.12, + "learning_rate": 2.0657626501638153e-06, + "loss": 0.4015, + "step": 17159 + }, + { + "epoch": 3.12, + "learning_rate": 2.0657080451401533e-06, + "loss": 0.5143, + "step": 17160 + }, + { + "epoch": 3.12, + "learning_rate": 2.065653440116491e-06, + "loss": 0.8714, + "step": 17161 + }, + { + "epoch": 3.12, + "learning_rate": 2.0655988350928288e-06, + "loss": 0.1741, + "step": 17162 + }, + { + "epoch": 3.12, + "learning_rate": 2.0655442300691663e-06, + "loss": 0.0525, + "step": 17163 + }, + { + "epoch": 3.12, + "learning_rate": 2.0654896250455042e-06, + "loss": 0.3728, + "step": 17164 + }, + { + "epoch": 3.12, + "learning_rate": 2.065435020021842e-06, + "loss": 0.0357, + "step": 17165 + }, + { + "epoch": 3.12, + "learning_rate": 2.0653804149981797e-06, + "loss": 0.4568, + "step": 17166 + }, + { + "epoch": 3.12, + "learning_rate": 2.0653258099745176e-06, + "loss": 0.2005, + "step": 17167 + }, + { + "epoch": 3.12, + "learning_rate": 2.0652712049508556e-06, + "loss": 0.5639, + "step": 17168 + }, + { + "epoch": 3.12, + "learning_rate": 2.065216599927193e-06, + "loss": 1.0947, + "step": 17169 + }, + { + "epoch": 3.12, + "learning_rate": 2.065161994903531e-06, + "loss": 0.2136, + "step": 17170 + }, + { + "epoch": 3.12, + "learning_rate": 2.065107389879869e-06, + "loss": 0.4926, + "step": 17171 + }, + { + "epoch": 3.12, + "learning_rate": 2.065052784856207e-06, + "loss": 0.1185, + "step": 17172 + }, + { + "epoch": 3.12, + "learning_rate": 2.064998179832545e-06, + "loss": 0.3793, + "step": 17173 + }, + { + "epoch": 3.12, + "learning_rate": 2.064943574808883e-06, + "loss": 1.1945, + "step": 17174 + }, + { + "epoch": 3.12, + "learning_rate": 2.0648889697852203e-06, + "loss": 0.3747, + "step": 17175 + }, + { + "epoch": 3.12, + "learning_rate": 2.0648343647615583e-06, + "loss": 0.6847, + "step": 17176 + }, + { + "epoch": 3.12, + "learning_rate": 2.0647797597378958e-06, + "loss": 0.0176, + "step": 17177 + }, + { + "epoch": 3.12, + "learning_rate": 2.0647251547142337e-06, + "loss": 0.7261, + "step": 17178 + }, + { + "epoch": 3.12, + "learning_rate": 2.0646705496905717e-06, + "loss": 2.2131, + "step": 17179 + }, + { + "epoch": 3.12, + "learning_rate": 2.064615944666909e-06, + "loss": 0.9084, + "step": 17180 + }, + { + "epoch": 3.12, + "learning_rate": 2.064561339643247e-06, + "loss": 0.5654, + "step": 17181 + }, + { + "epoch": 3.12, + "learning_rate": 2.064506734619585e-06, + "loss": 1.3744, + "step": 17182 + }, + { + "epoch": 3.12, + "learning_rate": 2.0644521295959226e-06, + "loss": 0.3525, + "step": 17183 + }, + { + "epoch": 3.12, + "learning_rate": 2.0643975245722605e-06, + "loss": 0.0099, + "step": 17184 + }, + { + "epoch": 3.13, + "learning_rate": 2.0643429195485985e-06, + "loss": 0.9201, + "step": 17185 + }, + { + "epoch": 3.13, + "learning_rate": 2.0642883145249364e-06, + "loss": 3.2809, + "step": 17186 + }, + { + "epoch": 3.13, + "learning_rate": 2.0642337095012744e-06, + "loss": 1.3722, + "step": 17187 + }, + { + "epoch": 3.13, + "learning_rate": 2.0641791044776123e-06, + "loss": 0.2173, + "step": 17188 + }, + { + "epoch": 3.13, + "learning_rate": 2.06412449945395e-06, + "loss": 0.3827, + "step": 17189 + }, + { + "epoch": 3.13, + "learning_rate": 2.0640698944302878e-06, + "loss": 1.3804, + "step": 17190 + }, + { + "epoch": 3.13, + "learning_rate": 2.0640152894066253e-06, + "loss": 0.7259, + "step": 17191 + }, + { + "epoch": 3.13, + "learning_rate": 2.0639606843829632e-06, + "loss": 0.6509, + "step": 17192 + }, + { + "epoch": 3.13, + "learning_rate": 2.063906079359301e-06, + "loss": 0.4728, + "step": 17193 + }, + { + "epoch": 3.13, + "learning_rate": 2.0638514743356387e-06, + "loss": 0.7491, + "step": 17194 + }, + { + "epoch": 3.13, + "learning_rate": 2.0637968693119766e-06, + "loss": 0.0449, + "step": 17195 + }, + { + "epoch": 3.13, + "learning_rate": 2.0637422642883146e-06, + "loss": 1.6446, + "step": 17196 + }, + { + "epoch": 3.13, + "learning_rate": 2.063687659264652e-06, + "loss": 1.1668, + "step": 17197 + }, + { + "epoch": 3.13, + "learning_rate": 2.06363305424099e-06, + "loss": 1.2627, + "step": 17198 + }, + { + "epoch": 3.13, + "learning_rate": 2.063578449217328e-06, + "loss": 0.0268, + "step": 17199 + }, + { + "epoch": 3.13, + "learning_rate": 2.063523844193666e-06, + "loss": 0.3334, + "step": 17200 + }, + { + "epoch": 3.13, + "learning_rate": 2.063469239170004e-06, + "loss": 0.0183, + "step": 17201 + }, + { + "epoch": 3.13, + "learning_rate": 2.0634146341463414e-06, + "loss": 0.168, + "step": 17202 + }, + { + "epoch": 3.13, + "learning_rate": 2.0633600291226793e-06, + "loss": 0.6304, + "step": 17203 + }, + { + "epoch": 3.13, + "learning_rate": 2.0633054240990173e-06, + "loss": 0.3954, + "step": 17204 + }, + { + "epoch": 3.13, + "learning_rate": 2.063250819075355e-06, + "loss": 0.2413, + "step": 17205 + }, + { + "epoch": 3.13, + "learning_rate": 2.0631962140516928e-06, + "loss": 0.4859, + "step": 17206 + }, + { + "epoch": 3.13, + "learning_rate": 2.0631416090280307e-06, + "loss": 1.6513, + "step": 17207 + }, + { + "epoch": 3.13, + "learning_rate": 2.0630870040043682e-06, + "loss": 1.1296, + "step": 17208 + }, + { + "epoch": 3.13, + "learning_rate": 2.063032398980706e-06, + "loss": 0.2691, + "step": 17209 + }, + { + "epoch": 3.13, + "learning_rate": 2.062977793957044e-06, + "loss": 1.6609, + "step": 17210 + }, + { + "epoch": 3.13, + "learning_rate": 2.0629231889333816e-06, + "loss": 0.7397, + "step": 17211 + }, + { + "epoch": 3.13, + "learning_rate": 2.0628685839097196e-06, + "loss": 0.3232, + "step": 17212 + }, + { + "epoch": 3.13, + "learning_rate": 2.0628139788860575e-06, + "loss": 0.2915, + "step": 17213 + }, + { + "epoch": 3.13, + "learning_rate": 2.0627593738623955e-06, + "loss": 0.9449, + "step": 17214 + }, + { + "epoch": 3.13, + "learning_rate": 2.0627047688387334e-06, + "loss": 0.5144, + "step": 17215 + }, + { + "epoch": 3.13, + "learning_rate": 2.062650163815071e-06, + "loss": 1.1596, + "step": 17216 + }, + { + "epoch": 3.13, + "learning_rate": 2.062595558791409e-06, + "loss": 2.4436, + "step": 17217 + }, + { + "epoch": 3.13, + "learning_rate": 2.062540953767747e-06, + "loss": 0.0102, + "step": 17218 + }, + { + "epoch": 3.13, + "learning_rate": 2.0624863487440843e-06, + "loss": 0.024, + "step": 17219 + }, + { + "epoch": 3.13, + "learning_rate": 2.0624317437204223e-06, + "loss": 0.2952, + "step": 17220 + }, + { + "epoch": 3.13, + "learning_rate": 2.0623771386967602e-06, + "loss": 2.1054, + "step": 17221 + }, + { + "epoch": 3.13, + "learning_rate": 2.0623225336730977e-06, + "loss": 0.2558, + "step": 17222 + }, + { + "epoch": 3.13, + "learning_rate": 2.0622679286494357e-06, + "loss": 0.8069, + "step": 17223 + }, + { + "epoch": 3.13, + "learning_rate": 2.0622133236257736e-06, + "loss": 0.2057, + "step": 17224 + }, + { + "epoch": 3.13, + "learning_rate": 2.0621587186021116e-06, + "loss": 0.2949, + "step": 17225 + }, + { + "epoch": 3.13, + "learning_rate": 2.0621041135784495e-06, + "loss": 0.3689, + "step": 17226 + }, + { + "epoch": 3.13, + "learning_rate": 2.0620495085547874e-06, + "loss": 1.6005, + "step": 17227 + }, + { + "epoch": 3.13, + "learning_rate": 2.061994903531125e-06, + "loss": 0.4026, + "step": 17228 + }, + { + "epoch": 3.13, + "learning_rate": 2.061940298507463e-06, + "loss": 1.434, + "step": 17229 + }, + { + "epoch": 3.13, + "learning_rate": 2.0618856934838004e-06, + "loss": 0.2935, + "step": 17230 + }, + { + "epoch": 3.13, + "learning_rate": 2.0618310884601384e-06, + "loss": 0.7243, + "step": 17231 + }, + { + "epoch": 3.13, + "learning_rate": 2.0617764834364763e-06, + "loss": 0.0263, + "step": 17232 + }, + { + "epoch": 3.13, + "learning_rate": 2.061721878412814e-06, + "loss": 0.1314, + "step": 17233 + }, + { + "epoch": 3.13, + "learning_rate": 2.0616672733891518e-06, + "loss": 0.2785, + "step": 17234 + }, + { + "epoch": 3.13, + "learning_rate": 2.0616126683654897e-06, + "loss": 0.9298, + "step": 17235 + }, + { + "epoch": 3.13, + "learning_rate": 2.0615580633418272e-06, + "loss": 1.8226, + "step": 17236 + }, + { + "epoch": 3.13, + "learning_rate": 2.061503458318165e-06, + "loss": 1.1196, + "step": 17237 + }, + { + "epoch": 3.13, + "learning_rate": 2.061448853294503e-06, + "loss": 0.6718, + "step": 17238 + }, + { + "epoch": 3.13, + "learning_rate": 2.061394248270841e-06, + "loss": 0.1177, + "step": 17239 + }, + { + "epoch": 3.14, + "learning_rate": 2.061339643247179e-06, + "loss": 0.6141, + "step": 17240 + }, + { + "epoch": 3.14, + "learning_rate": 2.061285038223517e-06, + "loss": 0.4456, + "step": 17241 + }, + { + "epoch": 3.14, + "learning_rate": 2.0612304331998545e-06, + "loss": 3.0866, + "step": 17242 + }, + { + "epoch": 3.14, + "learning_rate": 2.0611758281761924e-06, + "loss": 2.0176, + "step": 17243 + }, + { + "epoch": 3.14, + "learning_rate": 2.06112122315253e-06, + "loss": 0.2601, + "step": 17244 + }, + { + "epoch": 3.14, + "learning_rate": 2.061066618128868e-06, + "loss": 0.0255, + "step": 17245 + }, + { + "epoch": 3.14, + "learning_rate": 2.061012013105206e-06, + "loss": 0.2498, + "step": 17246 + }, + { + "epoch": 3.14, + "learning_rate": 2.0609574080815434e-06, + "loss": 0.5398, + "step": 17247 + }, + { + "epoch": 3.14, + "learning_rate": 2.0609028030578813e-06, + "loss": 0.3513, + "step": 17248 + }, + { + "epoch": 3.14, + "learning_rate": 2.0608481980342192e-06, + "loss": 0.5464, + "step": 17249 + }, + { + "epoch": 3.14, + "learning_rate": 2.0607935930105568e-06, + "loss": 0.4731, + "step": 17250 + }, + { + "epoch": 3.14, + "learning_rate": 2.0607389879868947e-06, + "loss": 2.5101, + "step": 17251 + }, + { + "epoch": 3.14, + "learning_rate": 2.0606843829632326e-06, + "loss": 0.3712, + "step": 17252 + }, + { + "epoch": 3.14, + "learning_rate": 2.0606297779395706e-06, + "loss": 0.9958, + "step": 17253 + }, + { + "epoch": 3.14, + "learning_rate": 2.0605751729159085e-06, + "loss": 0.2401, + "step": 17254 + }, + { + "epoch": 3.14, + "learning_rate": 2.0605205678922465e-06, + "loss": 0.0156, + "step": 17255 + }, + { + "epoch": 3.14, + "learning_rate": 2.060465962868584e-06, + "loss": 2.2575, + "step": 17256 + }, + { + "epoch": 3.14, + "learning_rate": 2.060411357844922e-06, + "loss": 0.9152, + "step": 17257 + }, + { + "epoch": 3.14, + "learning_rate": 2.0603567528212595e-06, + "loss": 1.7974, + "step": 17258 + }, + { + "epoch": 3.14, + "learning_rate": 2.0603021477975974e-06, + "loss": 0.2817, + "step": 17259 + }, + { + "epoch": 3.14, + "learning_rate": 2.0602475427739353e-06, + "loss": 0.021, + "step": 17260 + }, + { + "epoch": 3.14, + "learning_rate": 2.060192937750273e-06, + "loss": 1.2761, + "step": 17261 + }, + { + "epoch": 3.14, + "learning_rate": 2.060138332726611e-06, + "loss": 0.4712, + "step": 17262 + }, + { + "epoch": 3.14, + "learning_rate": 2.0600837277029488e-06, + "loss": 0.8132, + "step": 17263 + }, + { + "epoch": 3.14, + "learning_rate": 2.0600291226792863e-06, + "loss": 0.5256, + "step": 17264 + }, + { + "epoch": 3.14, + "learning_rate": 2.0599745176556242e-06, + "loss": 0.0952, + "step": 17265 + }, + { + "epoch": 3.14, + "learning_rate": 2.059919912631962e-06, + "loss": 0.0362, + "step": 17266 + }, + { + "epoch": 3.14, + "learning_rate": 2.0598653076083e-06, + "loss": 0.5896, + "step": 17267 + }, + { + "epoch": 3.14, + "learning_rate": 2.059810702584638e-06, + "loss": 1.9335, + "step": 17268 + }, + { + "epoch": 3.14, + "learning_rate": 2.059756097560976e-06, + "loss": 0.2636, + "step": 17269 + }, + { + "epoch": 3.14, + "learning_rate": 2.0597014925373135e-06, + "loss": 0.0129, + "step": 17270 + }, + { + "epoch": 3.14, + "learning_rate": 2.0596468875136515e-06, + "loss": 2.5278, + "step": 17271 + }, + { + "epoch": 3.14, + "learning_rate": 2.059592282489989e-06, + "loss": 0.1184, + "step": 17272 + }, + { + "epoch": 3.14, + "learning_rate": 2.059537677466327e-06, + "loss": 2.1707, + "step": 17273 + }, + { + "epoch": 3.14, + "learning_rate": 2.059483072442665e-06, + "loss": 0.3315, + "step": 17274 + }, + { + "epoch": 3.14, + "learning_rate": 2.0594284674190024e-06, + "loss": 1.5866, + "step": 17275 + }, + { + "epoch": 3.14, + "learning_rate": 2.0593738623953403e-06, + "loss": 0.0344, + "step": 17276 + }, + { + "epoch": 3.14, + "learning_rate": 2.0593192573716783e-06, + "loss": 0.0464, + "step": 17277 + }, + { + "epoch": 3.14, + "learning_rate": 2.0592646523480158e-06, + "loss": 0.0167, + "step": 17278 + }, + { + "epoch": 3.14, + "learning_rate": 2.0592100473243537e-06, + "loss": 0.5794, + "step": 17279 + }, + { + "epoch": 3.14, + "learning_rate": 2.0591554423006917e-06, + "loss": 0.7207, + "step": 17280 + }, + { + "epoch": 3.14, + "learning_rate": 2.0591008372770296e-06, + "loss": 0.7251, + "step": 17281 + }, + { + "epoch": 3.14, + "learning_rate": 2.0590462322533676e-06, + "loss": 0.0233, + "step": 17282 + }, + { + "epoch": 3.14, + "learning_rate": 2.058991627229705e-06, + "loss": 0.2507, + "step": 17283 + }, + { + "epoch": 3.14, + "learning_rate": 2.058937022206043e-06, + "loss": 0.0149, + "step": 17284 + }, + { + "epoch": 3.14, + "learning_rate": 2.058882417182381e-06, + "loss": 0.5479, + "step": 17285 + }, + { + "epoch": 3.14, + "learning_rate": 2.0588278121587185e-06, + "loss": 0.1825, + "step": 17286 + }, + { + "epoch": 3.14, + "learning_rate": 2.0587732071350564e-06, + "loss": 0.1467, + "step": 17287 + }, + { + "epoch": 3.14, + "learning_rate": 2.0587186021113944e-06, + "loss": 1.614, + "step": 17288 + }, + { + "epoch": 3.14, + "learning_rate": 2.058663997087732e-06, + "loss": 0.2775, + "step": 17289 + }, + { + "epoch": 3.14, + "learning_rate": 2.05860939206407e-06, + "loss": 1.5996, + "step": 17290 + }, + { + "epoch": 3.14, + "learning_rate": 2.0585547870404078e-06, + "loss": 1.0245, + "step": 17291 + }, + { + "epoch": 3.14, + "learning_rate": 2.0585001820167453e-06, + "loss": 1.5964, + "step": 17292 + }, + { + "epoch": 3.14, + "learning_rate": 2.0584455769930832e-06, + "loss": 0.5743, + "step": 17293 + }, + { + "epoch": 3.14, + "learning_rate": 2.058390971969421e-06, + "loss": 0.9815, + "step": 17294 + }, + { + "epoch": 3.15, + "learning_rate": 2.058336366945759e-06, + "loss": 0.1076, + "step": 17295 + }, + { + "epoch": 3.15, + "learning_rate": 2.058281761922097e-06, + "loss": 0.9478, + "step": 17296 + }, + { + "epoch": 3.15, + "learning_rate": 2.0582271568984346e-06, + "loss": 0.2785, + "step": 17297 + }, + { + "epoch": 3.15, + "learning_rate": 2.0581725518747725e-06, + "loss": 0.4688, + "step": 17298 + }, + { + "epoch": 3.15, + "learning_rate": 2.0581179468511105e-06, + "loss": 0.3394, + "step": 17299 + }, + { + "epoch": 3.15, + "learning_rate": 2.058063341827448e-06, + "loss": 2.0944, + "step": 17300 + }, + { + "epoch": 3.15, + "learning_rate": 2.058008736803786e-06, + "loss": 2.3606, + "step": 17301 + }, + { + "epoch": 3.15, + "learning_rate": 2.057954131780124e-06, + "loss": 1.3085, + "step": 17302 + }, + { + "epoch": 3.15, + "learning_rate": 2.0578995267564614e-06, + "loss": 0.7921, + "step": 17303 + }, + { + "epoch": 3.15, + "learning_rate": 2.0578449217327994e-06, + "loss": 0.7806, + "step": 17304 + }, + { + "epoch": 3.15, + "learning_rate": 2.0577903167091373e-06, + "loss": 0.3485, + "step": 17305 + }, + { + "epoch": 3.15, + "learning_rate": 2.0577357116854752e-06, + "loss": 0.7215, + "step": 17306 + }, + { + "epoch": 3.15, + "learning_rate": 2.057681106661813e-06, + "loss": 0.2731, + "step": 17307 + }, + { + "epoch": 3.15, + "learning_rate": 2.057626501638151e-06, + "loss": 0.033, + "step": 17308 + }, + { + "epoch": 3.15, + "learning_rate": 2.0575718966144886e-06, + "loss": 1.0793, + "step": 17309 + }, + { + "epoch": 3.15, + "learning_rate": 2.0575172915908266e-06, + "loss": 0.9675, + "step": 17310 + }, + { + "epoch": 3.15, + "learning_rate": 2.057462686567164e-06, + "loss": 0.3044, + "step": 17311 + }, + { + "epoch": 3.15, + "learning_rate": 2.057408081543502e-06, + "loss": 0.9242, + "step": 17312 + }, + { + "epoch": 3.15, + "learning_rate": 2.05735347651984e-06, + "loss": 0.0915, + "step": 17313 + }, + { + "epoch": 3.15, + "learning_rate": 2.0572988714961775e-06, + "loss": 0.6741, + "step": 17314 + }, + { + "epoch": 3.15, + "learning_rate": 2.0572442664725155e-06, + "loss": 0.835, + "step": 17315 + }, + { + "epoch": 3.15, + "learning_rate": 2.0571896614488534e-06, + "loss": 0.5209, + "step": 17316 + }, + { + "epoch": 3.15, + "learning_rate": 2.057135056425191e-06, + "loss": 0.7542, + "step": 17317 + }, + { + "epoch": 3.15, + "learning_rate": 2.057080451401529e-06, + "loss": 0.4734, + "step": 17318 + }, + { + "epoch": 3.15, + "learning_rate": 2.057025846377867e-06, + "loss": 1.171, + "step": 17319 + }, + { + "epoch": 3.15, + "learning_rate": 2.0569712413542048e-06, + "loss": 0.3769, + "step": 17320 + }, + { + "epoch": 3.15, + "learning_rate": 2.0569166363305427e-06, + "loss": 0.931, + "step": 17321 + }, + { + "epoch": 3.15, + "learning_rate": 2.0568620313068806e-06, + "loss": 0.3171, + "step": 17322 + }, + { + "epoch": 3.15, + "learning_rate": 2.056807426283218e-06, + "loss": 1.0919, + "step": 17323 + }, + { + "epoch": 3.15, + "learning_rate": 2.056752821259556e-06, + "loss": 0.461, + "step": 17324 + }, + { + "epoch": 3.15, + "learning_rate": 2.0566982162358936e-06, + "loss": 0.6611, + "step": 17325 + }, + { + "epoch": 3.15, + "learning_rate": 2.0566436112122316e-06, + "loss": 0.0296, + "step": 17326 + }, + { + "epoch": 3.15, + "learning_rate": 2.0565890061885695e-06, + "loss": 0.3235, + "step": 17327 + }, + { + "epoch": 3.15, + "learning_rate": 2.056534401164907e-06, + "loss": 0.5236, + "step": 17328 + }, + { + "epoch": 3.15, + "learning_rate": 2.056479796141245e-06, + "loss": 0.8738, + "step": 17329 + }, + { + "epoch": 3.15, + "learning_rate": 2.056425191117583e-06, + "loss": 1.0355, + "step": 17330 + }, + { + "epoch": 3.15, + "learning_rate": 2.0563705860939204e-06, + "loss": 0.3675, + "step": 17331 + }, + { + "epoch": 3.15, + "learning_rate": 2.0563159810702584e-06, + "loss": 0.4557, + "step": 17332 + }, + { + "epoch": 3.15, + "learning_rate": 2.0562613760465963e-06, + "loss": 0.4446, + "step": 17333 + }, + { + "epoch": 3.15, + "learning_rate": 2.0562067710229343e-06, + "loss": 0.1116, + "step": 17334 + }, + { + "epoch": 3.15, + "learning_rate": 2.056152165999272e-06, + "loss": 0.3648, + "step": 17335 + }, + { + "epoch": 3.15, + "learning_rate": 2.05609756097561e-06, + "loss": 1.9544, + "step": 17336 + }, + { + "epoch": 3.15, + "learning_rate": 2.0560429559519477e-06, + "loss": 0.1268, + "step": 17337 + }, + { + "epoch": 3.15, + "learning_rate": 2.0559883509282856e-06, + "loss": 0.6215, + "step": 17338 + }, + { + "epoch": 3.15, + "learning_rate": 2.055933745904623e-06, + "loss": 0.2567, + "step": 17339 + }, + { + "epoch": 3.15, + "learning_rate": 2.055879140880961e-06, + "loss": 0.7114, + "step": 17340 + }, + { + "epoch": 3.15, + "learning_rate": 2.055824535857299e-06, + "loss": 0.0304, + "step": 17341 + }, + { + "epoch": 3.15, + "learning_rate": 2.0557699308336365e-06, + "loss": 0.0218, + "step": 17342 + }, + { + "epoch": 3.15, + "learning_rate": 2.0557153258099745e-06, + "loss": 0.0159, + "step": 17343 + }, + { + "epoch": 3.15, + "learning_rate": 2.0556607207863124e-06, + "loss": 0.0374, + "step": 17344 + }, + { + "epoch": 3.15, + "learning_rate": 2.05560611576265e-06, + "loss": 0.0159, + "step": 17345 + }, + { + "epoch": 3.15, + "learning_rate": 2.055551510738988e-06, + "loss": 0.156, + "step": 17346 + }, + { + "epoch": 3.15, + "learning_rate": 2.055496905715326e-06, + "loss": 0.8881, + "step": 17347 + }, + { + "epoch": 3.15, + "learning_rate": 2.0554423006916638e-06, + "loss": 0.5054, + "step": 17348 + }, + { + "epoch": 3.15, + "learning_rate": 2.0553876956680017e-06, + "loss": 2.3468, + "step": 17349 + }, + { + "epoch": 3.16, + "learning_rate": 2.0553330906443397e-06, + "loss": 0.1209, + "step": 17350 + }, + { + "epoch": 3.16, + "learning_rate": 2.055278485620677e-06, + "loss": 0.243, + "step": 17351 + }, + { + "epoch": 3.16, + "learning_rate": 2.055223880597015e-06, + "loss": 0.2573, + "step": 17352 + }, + { + "epoch": 3.16, + "learning_rate": 2.0551692755733526e-06, + "loss": 0.2799, + "step": 17353 + }, + { + "epoch": 3.16, + "learning_rate": 2.0551146705496906e-06, + "loss": 1.0912, + "step": 17354 + }, + { + "epoch": 3.16, + "learning_rate": 2.0550600655260285e-06, + "loss": 1.2121, + "step": 17355 + }, + { + "epoch": 3.16, + "learning_rate": 2.055005460502366e-06, + "loss": 0.1016, + "step": 17356 + }, + { + "epoch": 3.16, + "learning_rate": 2.054950855478704e-06, + "loss": 0.8599, + "step": 17357 + }, + { + "epoch": 3.16, + "learning_rate": 2.054896250455042e-06, + "loss": 1.5823, + "step": 17358 + }, + { + "epoch": 3.16, + "learning_rate": 2.0548416454313795e-06, + "loss": 0.6506, + "step": 17359 + }, + { + "epoch": 3.16, + "learning_rate": 2.0547870404077174e-06, + "loss": 2.196, + "step": 17360 + }, + { + "epoch": 3.16, + "learning_rate": 2.0547324353840553e-06, + "loss": 3.7747, + "step": 17361 + }, + { + "epoch": 3.16, + "learning_rate": 2.0546778303603933e-06, + "loss": 2.5429, + "step": 17362 + }, + { + "epoch": 3.16, + "learning_rate": 2.0546232253367312e-06, + "loss": 0.4296, + "step": 17363 + }, + { + "epoch": 3.16, + "learning_rate": 2.0545686203130688e-06, + "loss": 1.7623, + "step": 17364 + }, + { + "epoch": 3.16, + "learning_rate": 2.0545140152894067e-06, + "loss": 0.1325, + "step": 17365 + }, + { + "epoch": 3.16, + "learning_rate": 2.0544594102657446e-06, + "loss": 0.6661, + "step": 17366 + }, + { + "epoch": 3.16, + "learning_rate": 2.054404805242082e-06, + "loss": 0.2859, + "step": 17367 + }, + { + "epoch": 3.16, + "learning_rate": 2.05435020021842e-06, + "loss": 0.4615, + "step": 17368 + }, + { + "epoch": 3.16, + "learning_rate": 2.054295595194758e-06, + "loss": 2.3572, + "step": 17369 + }, + { + "epoch": 3.16, + "learning_rate": 2.0542409901710956e-06, + "loss": 1.8788, + "step": 17370 + }, + { + "epoch": 3.16, + "learning_rate": 2.0541863851474335e-06, + "loss": 0.3099, + "step": 17371 + }, + { + "epoch": 3.16, + "learning_rate": 2.0541317801237715e-06, + "loss": 0.5927, + "step": 17372 + }, + { + "epoch": 3.16, + "learning_rate": 2.054077175100109e-06, + "loss": 0.0528, + "step": 17373 + }, + { + "epoch": 3.16, + "learning_rate": 2.054022570076447e-06, + "loss": 0.054, + "step": 17374 + }, + { + "epoch": 3.16, + "learning_rate": 2.053967965052785e-06, + "loss": 0.8758, + "step": 17375 + }, + { + "epoch": 3.16, + "learning_rate": 2.053913360029123e-06, + "loss": 0.0245, + "step": 17376 + }, + { + "epoch": 3.16, + "learning_rate": 2.0538587550054607e-06, + "loss": 0.5429, + "step": 17377 + }, + { + "epoch": 3.16, + "learning_rate": 2.0538041499817983e-06, + "loss": 0.12, + "step": 17378 + }, + { + "epoch": 3.16, + "learning_rate": 2.053749544958136e-06, + "loss": 0.4393, + "step": 17379 + }, + { + "epoch": 3.16, + "learning_rate": 2.053694939934474e-06, + "loss": 0.2592, + "step": 17380 + }, + { + "epoch": 3.16, + "learning_rate": 2.0536403349108117e-06, + "loss": 0.901, + "step": 17381 + }, + { + "epoch": 3.16, + "learning_rate": 2.0535857298871496e-06, + "loss": 0.3109, + "step": 17382 + }, + { + "epoch": 3.16, + "learning_rate": 2.0535311248634876e-06, + "loss": 0.013, + "step": 17383 + }, + { + "epoch": 3.16, + "learning_rate": 2.053476519839825e-06, + "loss": 0.051, + "step": 17384 + }, + { + "epoch": 3.16, + "learning_rate": 2.053421914816163e-06, + "loss": 2.3178, + "step": 17385 + }, + { + "epoch": 3.16, + "learning_rate": 2.053367309792501e-06, + "loss": 0.3833, + "step": 17386 + }, + { + "epoch": 3.16, + "learning_rate": 2.053312704768839e-06, + "loss": 0.4354, + "step": 17387 + }, + { + "epoch": 3.16, + "learning_rate": 2.0532580997451764e-06, + "loss": 0.722, + "step": 17388 + }, + { + "epoch": 3.16, + "learning_rate": 2.0532034947215144e-06, + "loss": 1.0942, + "step": 17389 + }, + { + "epoch": 3.16, + "learning_rate": 2.0531488896978523e-06, + "loss": 0.921, + "step": 17390 + }, + { + "epoch": 3.16, + "learning_rate": 2.0530942846741903e-06, + "loss": 0.7738, + "step": 17391 + }, + { + "epoch": 3.16, + "learning_rate": 2.0530396796505278e-06, + "loss": 2.1335, + "step": 17392 + }, + { + "epoch": 3.16, + "learning_rate": 2.0529850746268657e-06, + "loss": 1.9123, + "step": 17393 + }, + { + "epoch": 3.16, + "learning_rate": 2.0529304696032037e-06, + "loss": 1.5844, + "step": 17394 + }, + { + "epoch": 3.16, + "learning_rate": 2.052875864579541e-06, + "loss": 0.4014, + "step": 17395 + }, + { + "epoch": 3.16, + "learning_rate": 2.052821259555879e-06, + "loss": 0.4139, + "step": 17396 + }, + { + "epoch": 3.16, + "learning_rate": 2.052766654532217e-06, + "loss": 0.8413, + "step": 17397 + }, + { + "epoch": 3.16, + "learning_rate": 2.0527120495085546e-06, + "loss": 0.1513, + "step": 17398 + }, + { + "epoch": 3.16, + "learning_rate": 2.0526574444848925e-06, + "loss": 0.0593, + "step": 17399 + }, + { + "epoch": 3.16, + "learning_rate": 2.0526028394612305e-06, + "loss": 1.8481, + "step": 17400 + }, + { + "epoch": 3.16, + "learning_rate": 2.0525482344375684e-06, + "loss": 0.4119, + "step": 17401 + }, + { + "epoch": 3.16, + "learning_rate": 2.0524936294139064e-06, + "loss": 1.4203, + "step": 17402 + }, + { + "epoch": 3.16, + "learning_rate": 2.0524390243902443e-06, + "loss": 0.5223, + "step": 17403 + }, + { + "epoch": 3.16, + "learning_rate": 2.052384419366582e-06, + "loss": 0.0602, + "step": 17404 + }, + { + "epoch": 3.17, + "learning_rate": 2.0523298143429198e-06, + "loss": 1.5927, + "step": 17405 + }, + { + "epoch": 3.17, + "learning_rate": 2.0522752093192573e-06, + "loss": 0.7335, + "step": 17406 + }, + { + "epoch": 3.17, + "learning_rate": 2.0522206042955952e-06, + "loss": 1.0611, + "step": 17407 + }, + { + "epoch": 3.17, + "learning_rate": 2.052165999271933e-06, + "loss": 1.5748, + "step": 17408 + }, + { + "epoch": 3.17, + "learning_rate": 2.0521113942482707e-06, + "loss": 0.6887, + "step": 17409 + }, + { + "epoch": 3.17, + "learning_rate": 2.0520567892246086e-06, + "loss": 0.4175, + "step": 17410 + }, + { + "epoch": 3.17, + "learning_rate": 2.0520021842009466e-06, + "loss": 0.2838, + "step": 17411 + }, + { + "epoch": 3.17, + "learning_rate": 2.051947579177284e-06, + "loss": 0.2475, + "step": 17412 + }, + { + "epoch": 3.17, + "learning_rate": 2.051892974153622e-06, + "loss": 1.0464, + "step": 17413 + }, + { + "epoch": 3.17, + "learning_rate": 2.05183836912996e-06, + "loss": 1.2579, + "step": 17414 + }, + { + "epoch": 3.17, + "learning_rate": 2.051783764106298e-06, + "loss": 0.5512, + "step": 17415 + }, + { + "epoch": 3.17, + "learning_rate": 2.051729159082636e-06, + "loss": 0.3379, + "step": 17416 + }, + { + "epoch": 3.17, + "learning_rate": 2.051674554058974e-06, + "loss": 0.8051, + "step": 17417 + }, + { + "epoch": 3.17, + "learning_rate": 2.0516199490353113e-06, + "loss": 0.3903, + "step": 17418 + }, + { + "epoch": 3.17, + "learning_rate": 2.0515653440116493e-06, + "loss": 0.0164, + "step": 17419 + }, + { + "epoch": 3.17, + "learning_rate": 2.051510738987987e-06, + "loss": 0.76, + "step": 17420 + }, + { + "epoch": 3.17, + "learning_rate": 2.0514561339643248e-06, + "loss": 1.6589, + "step": 17421 + }, + { + "epoch": 3.17, + "learning_rate": 2.0514015289406627e-06, + "loss": 0.4181, + "step": 17422 + }, + { + "epoch": 3.17, + "learning_rate": 2.0513469239170002e-06, + "loss": 0.5131, + "step": 17423 + }, + { + "epoch": 3.17, + "learning_rate": 2.051292318893338e-06, + "loss": 0.0209, + "step": 17424 + }, + { + "epoch": 3.17, + "learning_rate": 2.051237713869676e-06, + "loss": 1.3412, + "step": 17425 + }, + { + "epoch": 3.17, + "learning_rate": 2.0511831088460136e-06, + "loss": 0.8955, + "step": 17426 + }, + { + "epoch": 3.17, + "learning_rate": 2.0511285038223516e-06, + "loss": 0.0559, + "step": 17427 + }, + { + "epoch": 3.17, + "learning_rate": 2.0510738987986895e-06, + "loss": 0.0237, + "step": 17428 + }, + { + "epoch": 3.17, + "learning_rate": 2.0510192937750275e-06, + "loss": 1.0988, + "step": 17429 + }, + { + "epoch": 3.17, + "learning_rate": 2.0509646887513654e-06, + "loss": 0.4305, + "step": 17430 + }, + { + "epoch": 3.17, + "learning_rate": 2.0509100837277033e-06, + "loss": 0.0811, + "step": 17431 + }, + { + "epoch": 3.17, + "learning_rate": 2.050855478704041e-06, + "loss": 0.6724, + "step": 17432 + }, + { + "epoch": 3.17, + "learning_rate": 2.050800873680379e-06, + "loss": 2.1756, + "step": 17433 + }, + { + "epoch": 3.17, + "learning_rate": 2.0507462686567163e-06, + "loss": 0.6749, + "step": 17434 + }, + { + "epoch": 3.17, + "learning_rate": 2.0506916636330543e-06, + "loss": 1.1587, + "step": 17435 + }, + { + "epoch": 3.17, + "learning_rate": 2.050637058609392e-06, + "loss": 0.811, + "step": 17436 + }, + { + "epoch": 3.17, + "learning_rate": 2.0505824535857297e-06, + "loss": 0.4161, + "step": 17437 + }, + { + "epoch": 3.17, + "learning_rate": 2.0505278485620677e-06, + "loss": 0.7793, + "step": 17438 + }, + { + "epoch": 3.17, + "learning_rate": 2.0504732435384056e-06, + "loss": 0.339, + "step": 17439 + }, + { + "epoch": 3.17, + "learning_rate": 2.050418638514743e-06, + "loss": 1.4341, + "step": 17440 + }, + { + "epoch": 3.17, + "learning_rate": 2.050364033491081e-06, + "loss": 1.1081, + "step": 17441 + }, + { + "epoch": 3.17, + "learning_rate": 2.050309428467419e-06, + "loss": 2.5556, + "step": 17442 + }, + { + "epoch": 3.17, + "learning_rate": 2.050254823443757e-06, + "loss": 0.2773, + "step": 17443 + }, + { + "epoch": 3.17, + "learning_rate": 2.050200218420095e-06, + "loss": 0.2266, + "step": 17444 + }, + { + "epoch": 3.17, + "learning_rate": 2.0501456133964324e-06, + "loss": 1.4291, + "step": 17445 + }, + { + "epoch": 3.17, + "learning_rate": 2.0500910083727704e-06, + "loss": 0.3739, + "step": 17446 + }, + { + "epoch": 3.17, + "learning_rate": 2.0500364033491083e-06, + "loss": 0.015, + "step": 17447 + }, + { + "epoch": 3.17, + "learning_rate": 2.049981798325446e-06, + "loss": 0.1058, + "step": 17448 + }, + { + "epoch": 3.17, + "learning_rate": 2.0499271933017838e-06, + "loss": 0.108, + "step": 17449 + }, + { + "epoch": 3.17, + "learning_rate": 2.0498725882781217e-06, + "loss": 0.861, + "step": 17450 + }, + { + "epoch": 3.17, + "learning_rate": 2.0498179832544592e-06, + "loss": 0.3411, + "step": 17451 + }, + { + "epoch": 3.17, + "learning_rate": 2.049763378230797e-06, + "loss": 1.5848, + "step": 17452 + }, + { + "epoch": 3.17, + "learning_rate": 2.049708773207135e-06, + "loss": 0.7929, + "step": 17453 + }, + { + "epoch": 3.17, + "learning_rate": 2.0496541681834726e-06, + "loss": 1.5683, + "step": 17454 + }, + { + "epoch": 3.17, + "learning_rate": 2.0495995631598106e-06, + "loss": 0.8184, + "step": 17455 + }, + { + "epoch": 3.17, + "learning_rate": 2.0495449581361485e-06, + "loss": 0.2475, + "step": 17456 + }, + { + "epoch": 3.17, + "learning_rate": 2.0494903531124865e-06, + "loss": 0.6707, + "step": 17457 + }, + { + "epoch": 3.17, + "learning_rate": 2.0494357480888244e-06, + "loss": 1.3424, + "step": 17458 + }, + { + "epoch": 3.17, + "learning_rate": 2.049381143065162e-06, + "loss": 1.5382, + "step": 17459 + }, + { + "epoch": 3.18, + "learning_rate": 2.0493265380415e-06, + "loss": 0.033, + "step": 17460 + }, + { + "epoch": 3.18, + "learning_rate": 2.049271933017838e-06, + "loss": 0.4358, + "step": 17461 + }, + { + "epoch": 3.18, + "learning_rate": 2.0492173279941753e-06, + "loss": 1.3824, + "step": 17462 + }, + { + "epoch": 3.18, + "learning_rate": 2.0491627229705133e-06, + "loss": 0.3898, + "step": 17463 + }, + { + "epoch": 3.18, + "learning_rate": 2.0491081179468512e-06, + "loss": 0.7565, + "step": 17464 + }, + { + "epoch": 3.18, + "learning_rate": 2.0490535129231888e-06, + "loss": 0.9481, + "step": 17465 + }, + { + "epoch": 3.18, + "learning_rate": 2.0489989078995267e-06, + "loss": 1.1147, + "step": 17466 + }, + { + "epoch": 3.18, + "learning_rate": 2.0489443028758646e-06, + "loss": 0.5588, + "step": 17467 + }, + { + "epoch": 3.18, + "learning_rate": 2.048889697852202e-06, + "loss": 0.6812, + "step": 17468 + }, + { + "epoch": 3.18, + "learning_rate": 2.04883509282854e-06, + "loss": 1.1367, + "step": 17469 + }, + { + "epoch": 3.18, + "learning_rate": 2.048780487804878e-06, + "loss": 0.3721, + "step": 17470 + }, + { + "epoch": 3.18, + "learning_rate": 2.048725882781216e-06, + "loss": 1.1086, + "step": 17471 + }, + { + "epoch": 3.18, + "learning_rate": 2.048671277757554e-06, + "loss": 0.0312, + "step": 17472 + }, + { + "epoch": 3.18, + "learning_rate": 2.0486166727338915e-06, + "loss": 1.8409, + "step": 17473 + }, + { + "epoch": 3.18, + "learning_rate": 2.0485620677102294e-06, + "loss": 0.0182, + "step": 17474 + }, + { + "epoch": 3.18, + "learning_rate": 2.0485074626865673e-06, + "loss": 1.4471, + "step": 17475 + }, + { + "epoch": 3.18, + "learning_rate": 2.048452857662905e-06, + "loss": 0.3369, + "step": 17476 + }, + { + "epoch": 3.18, + "learning_rate": 2.048398252639243e-06, + "loss": 0.8499, + "step": 17477 + }, + { + "epoch": 3.18, + "learning_rate": 2.0483436476155807e-06, + "loss": 0.0216, + "step": 17478 + }, + { + "epoch": 3.18, + "learning_rate": 2.0482890425919183e-06, + "loss": 1.3194, + "step": 17479 + }, + { + "epoch": 3.18, + "learning_rate": 2.0482344375682562e-06, + "loss": 2.6905, + "step": 17480 + }, + { + "epoch": 3.18, + "learning_rate": 2.048179832544594e-06, + "loss": 0.5236, + "step": 17481 + }, + { + "epoch": 3.18, + "learning_rate": 2.048125227520932e-06, + "loss": 0.0458, + "step": 17482 + }, + { + "epoch": 3.18, + "learning_rate": 2.04807062249727e-06, + "loss": 0.2717, + "step": 17483 + }, + { + "epoch": 3.18, + "learning_rate": 2.048016017473608e-06, + "loss": 0.8599, + "step": 17484 + }, + { + "epoch": 3.18, + "learning_rate": 2.0479614124499455e-06, + "loss": 0.5401, + "step": 17485 + }, + { + "epoch": 3.18, + "learning_rate": 2.0479068074262834e-06, + "loss": 0.2576, + "step": 17486 + }, + { + "epoch": 3.18, + "learning_rate": 2.047852202402621e-06, + "loss": 0.7127, + "step": 17487 + }, + { + "epoch": 3.18, + "learning_rate": 2.047797597378959e-06, + "loss": 1.2448, + "step": 17488 + }, + { + "epoch": 3.18, + "learning_rate": 2.047742992355297e-06, + "loss": 0.0238, + "step": 17489 + }, + { + "epoch": 3.18, + "learning_rate": 2.0476883873316344e-06, + "loss": 0.3065, + "step": 17490 + }, + { + "epoch": 3.18, + "learning_rate": 2.0476337823079723e-06, + "loss": 1.3742, + "step": 17491 + }, + { + "epoch": 3.18, + "learning_rate": 2.0475791772843103e-06, + "loss": 0.0558, + "step": 17492 + }, + { + "epoch": 3.18, + "learning_rate": 2.0475245722606478e-06, + "loss": 0.1183, + "step": 17493 + }, + { + "epoch": 3.18, + "learning_rate": 2.0474699672369857e-06, + "loss": 2.5513, + "step": 17494 + }, + { + "epoch": 3.18, + "learning_rate": 2.0474153622133237e-06, + "loss": 2.1986, + "step": 17495 + }, + { + "epoch": 3.18, + "learning_rate": 2.0473607571896616e-06, + "loss": 0.7106, + "step": 17496 + }, + { + "epoch": 3.18, + "learning_rate": 2.0473061521659996e-06, + "loss": 1.472, + "step": 17497 + }, + { + "epoch": 3.18, + "learning_rate": 2.0472515471423375e-06, + "loss": 0.405, + "step": 17498 + }, + { + "epoch": 3.18, + "learning_rate": 2.047196942118675e-06, + "loss": 0.6061, + "step": 17499 + }, + { + "epoch": 3.18, + "learning_rate": 2.047142337095013e-06, + "loss": 0.3476, + "step": 17500 + }, + { + "epoch": 3.18, + "learning_rate": 2.0470877320713505e-06, + "loss": 1.4051, + "step": 17501 + }, + { + "epoch": 3.18, + "learning_rate": 2.0470331270476884e-06, + "loss": 0.5101, + "step": 17502 + }, + { + "epoch": 3.18, + "learning_rate": 2.0469785220240264e-06, + "loss": 0.2416, + "step": 17503 + }, + { + "epoch": 3.18, + "learning_rate": 2.046923917000364e-06, + "loss": 0.9535, + "step": 17504 + }, + { + "epoch": 3.18, + "learning_rate": 2.046869311976702e-06, + "loss": 1.4811, + "step": 17505 + }, + { + "epoch": 3.18, + "learning_rate": 2.0468147069530398e-06, + "loss": 0.626, + "step": 17506 + }, + { + "epoch": 3.18, + "learning_rate": 2.0467601019293773e-06, + "loss": 1.0117, + "step": 17507 + }, + { + "epoch": 3.18, + "learning_rate": 2.0467054969057152e-06, + "loss": 0.3584, + "step": 17508 + }, + { + "epoch": 3.18, + "learning_rate": 2.046650891882053e-06, + "loss": 0.1657, + "step": 17509 + }, + { + "epoch": 3.18, + "learning_rate": 2.046596286858391e-06, + "loss": 0.6144, + "step": 17510 + }, + { + "epoch": 3.18, + "learning_rate": 2.046541681834729e-06, + "loss": 0.3501, + "step": 17511 + }, + { + "epoch": 3.18, + "learning_rate": 2.0464870768110666e-06, + "loss": 0.4964, + "step": 17512 + }, + { + "epoch": 3.18, + "learning_rate": 2.0464324717874045e-06, + "loss": 1.3533, + "step": 17513 + }, + { + "epoch": 3.18, + "learning_rate": 2.0463778667637425e-06, + "loss": 0.6687, + "step": 17514 + }, + { + "epoch": 3.19, + "learning_rate": 2.04632326174008e-06, + "loss": 0.5942, + "step": 17515 + }, + { + "epoch": 3.19, + "learning_rate": 2.046268656716418e-06, + "loss": 0.2093, + "step": 17516 + }, + { + "epoch": 3.19, + "learning_rate": 2.046214051692756e-06, + "loss": 0.0625, + "step": 17517 + }, + { + "epoch": 3.19, + "learning_rate": 2.0461594466690934e-06, + "loss": 1.7193, + "step": 17518 + }, + { + "epoch": 3.19, + "learning_rate": 2.0461048416454313e-06, + "loss": 1.6506, + "step": 17519 + }, + { + "epoch": 3.19, + "learning_rate": 2.0460502366217693e-06, + "loss": 0.2154, + "step": 17520 + }, + { + "epoch": 3.19, + "learning_rate": 2.045995631598107e-06, + "loss": 0.4669, + "step": 17521 + }, + { + "epoch": 3.19, + "learning_rate": 2.0459410265744448e-06, + "loss": 1.468, + "step": 17522 + }, + { + "epoch": 3.19, + "learning_rate": 2.0458864215507827e-06, + "loss": 0.3769, + "step": 17523 + }, + { + "epoch": 3.19, + "learning_rate": 2.0458318165271206e-06, + "loss": 0.2849, + "step": 17524 + }, + { + "epoch": 3.19, + "learning_rate": 2.0457772115034586e-06, + "loss": 1.2456, + "step": 17525 + }, + { + "epoch": 3.19, + "learning_rate": 2.045722606479796e-06, + "loss": 0.8115, + "step": 17526 + }, + { + "epoch": 3.19, + "learning_rate": 2.045668001456134e-06, + "loss": 1.7273, + "step": 17527 + }, + { + "epoch": 3.19, + "learning_rate": 2.045613396432472e-06, + "loss": 0.1392, + "step": 17528 + }, + { + "epoch": 3.19, + "learning_rate": 2.0455587914088095e-06, + "loss": 0.1872, + "step": 17529 + }, + { + "epoch": 3.19, + "learning_rate": 2.0455041863851475e-06, + "loss": 0.0113, + "step": 17530 + }, + { + "epoch": 3.19, + "learning_rate": 2.0454495813614854e-06, + "loss": 0.0841, + "step": 17531 + }, + { + "epoch": 3.19, + "learning_rate": 2.045394976337823e-06, + "loss": 0.4006, + "step": 17532 + }, + { + "epoch": 3.19, + "learning_rate": 2.045340371314161e-06, + "loss": 0.121, + "step": 17533 + }, + { + "epoch": 3.19, + "learning_rate": 2.045285766290499e-06, + "loss": 1.4189, + "step": 17534 + }, + { + "epoch": 3.19, + "learning_rate": 2.0452311612668363e-06, + "loss": 2.1167, + "step": 17535 + }, + { + "epoch": 3.19, + "learning_rate": 2.0451765562431743e-06, + "loss": 1.4829, + "step": 17536 + }, + { + "epoch": 3.19, + "learning_rate": 2.045121951219512e-06, + "loss": 0.2847, + "step": 17537 + }, + { + "epoch": 3.19, + "learning_rate": 2.04506734619585e-06, + "loss": 0.068, + "step": 17538 + }, + { + "epoch": 3.19, + "learning_rate": 2.045012741172188e-06, + "loss": 0.0528, + "step": 17539 + }, + { + "epoch": 3.19, + "learning_rate": 2.0449581361485256e-06, + "loss": 2.3688, + "step": 17540 + }, + { + "epoch": 3.19, + "learning_rate": 2.0449035311248636e-06, + "loss": 0.8259, + "step": 17541 + }, + { + "epoch": 3.19, + "learning_rate": 2.0448489261012015e-06, + "loss": 1.5873, + "step": 17542 + }, + { + "epoch": 3.19, + "learning_rate": 2.044794321077539e-06, + "loss": 0.46, + "step": 17543 + }, + { + "epoch": 3.19, + "learning_rate": 2.044739716053877e-06, + "loss": 0.5495, + "step": 17544 + }, + { + "epoch": 3.19, + "learning_rate": 2.044685111030215e-06, + "loss": 1.0686, + "step": 17545 + }, + { + "epoch": 3.19, + "learning_rate": 2.0446305060065524e-06, + "loss": 0.6819, + "step": 17546 + }, + { + "epoch": 3.19, + "learning_rate": 2.0445759009828904e-06, + "loss": 0.4127, + "step": 17547 + }, + { + "epoch": 3.19, + "learning_rate": 2.0445212959592283e-06, + "loss": 0.3419, + "step": 17548 + }, + { + "epoch": 3.19, + "learning_rate": 2.044466690935566e-06, + "loss": 0.714, + "step": 17549 + }, + { + "epoch": 3.19, + "learning_rate": 2.0444120859119038e-06, + "loss": 0.0266, + "step": 17550 + }, + { + "epoch": 3.19, + "learning_rate": 2.0443574808882417e-06, + "loss": 0.042, + "step": 17551 + }, + { + "epoch": 3.19, + "learning_rate": 2.0443028758645797e-06, + "loss": 2.1485, + "step": 17552 + }, + { + "epoch": 3.19, + "learning_rate": 2.0442482708409176e-06, + "loss": 4.156, + "step": 17553 + }, + { + "epoch": 3.19, + "learning_rate": 2.044193665817255e-06, + "loss": 1.1039, + "step": 17554 + }, + { + "epoch": 3.19, + "learning_rate": 2.044139060793593e-06, + "loss": 1.7838, + "step": 17555 + }, + { + "epoch": 3.19, + "learning_rate": 2.044084455769931e-06, + "loss": 0.4627, + "step": 17556 + }, + { + "epoch": 3.19, + "learning_rate": 2.0440298507462685e-06, + "loss": 0.0998, + "step": 17557 + }, + { + "epoch": 3.19, + "learning_rate": 2.0439752457226065e-06, + "loss": 0.746, + "step": 17558 + }, + { + "epoch": 3.19, + "learning_rate": 2.0439206406989444e-06, + "loss": 0.2265, + "step": 17559 + }, + { + "epoch": 3.19, + "learning_rate": 2.043866035675282e-06, + "loss": 0.4544, + "step": 17560 + }, + { + "epoch": 3.19, + "learning_rate": 2.04381143065162e-06, + "loss": 1.9921, + "step": 17561 + }, + { + "epoch": 3.19, + "learning_rate": 2.043756825627958e-06, + "loss": 1.4397, + "step": 17562 + }, + { + "epoch": 3.19, + "learning_rate": 2.0437022206042958e-06, + "loss": 0.0817, + "step": 17563 + }, + { + "epoch": 3.19, + "learning_rate": 2.0436476155806337e-06, + "loss": 0.6461, + "step": 17564 + }, + { + "epoch": 3.19, + "learning_rate": 2.0435930105569717e-06, + "loss": 0.7735, + "step": 17565 + }, + { + "epoch": 3.19, + "learning_rate": 2.043538405533309e-06, + "loss": 0.4542, + "step": 17566 + }, + { + "epoch": 3.19, + "learning_rate": 2.043483800509647e-06, + "loss": 0.3222, + "step": 17567 + }, + { + "epoch": 3.19, + "learning_rate": 2.0434291954859846e-06, + "loss": 1.1793, + "step": 17568 + }, + { + "epoch": 3.19, + "learning_rate": 2.0433745904623226e-06, + "loss": 1.3559, + "step": 17569 + }, + { + "epoch": 3.2, + "learning_rate": 2.0433199854386605e-06, + "loss": 1.2772, + "step": 17570 + }, + { + "epoch": 3.2, + "learning_rate": 2.043265380414998e-06, + "loss": 0.8638, + "step": 17571 + }, + { + "epoch": 3.2, + "learning_rate": 2.043210775391336e-06, + "loss": 0.4324, + "step": 17572 + }, + { + "epoch": 3.2, + "learning_rate": 2.043156170367674e-06, + "loss": 0.3556, + "step": 17573 + }, + { + "epoch": 3.2, + "learning_rate": 2.0431015653440115e-06, + "loss": 0.0647, + "step": 17574 + }, + { + "epoch": 3.2, + "learning_rate": 2.0430469603203494e-06, + "loss": 0.7298, + "step": 17575 + }, + { + "epoch": 3.2, + "learning_rate": 2.0429923552966873e-06, + "loss": 1.2932, + "step": 17576 + }, + { + "epoch": 3.2, + "learning_rate": 2.0429377502730253e-06, + "loss": 0.2719, + "step": 17577 + }, + { + "epoch": 3.2, + "learning_rate": 2.0428831452493632e-06, + "loss": 0.0994, + "step": 17578 + }, + { + "epoch": 3.2, + "learning_rate": 2.042828540225701e-06, + "loss": 0.7576, + "step": 17579 + }, + { + "epoch": 3.2, + "learning_rate": 2.0427739352020387e-06, + "loss": 0.6563, + "step": 17580 + }, + { + "epoch": 3.2, + "learning_rate": 2.0427193301783766e-06, + "loss": 0.6284, + "step": 17581 + }, + { + "epoch": 3.2, + "learning_rate": 2.042664725154714e-06, + "loss": 1.1645, + "step": 17582 + }, + { + "epoch": 3.2, + "learning_rate": 2.042610120131052e-06, + "loss": 0.0155, + "step": 17583 + }, + { + "epoch": 3.2, + "learning_rate": 2.04255551510739e-06, + "loss": 1.2684, + "step": 17584 + }, + { + "epoch": 3.2, + "learning_rate": 2.0425009100837276e-06, + "loss": 0.2351, + "step": 17585 + }, + { + "epoch": 3.2, + "learning_rate": 2.0424463050600655e-06, + "loss": 0.3266, + "step": 17586 + }, + { + "epoch": 3.2, + "learning_rate": 2.0423917000364035e-06, + "loss": 0.3542, + "step": 17587 + }, + { + "epoch": 3.2, + "learning_rate": 2.042337095012741e-06, + "loss": 1.0938, + "step": 17588 + }, + { + "epoch": 3.2, + "learning_rate": 2.042282489989079e-06, + "loss": 0.0216, + "step": 17589 + }, + { + "epoch": 3.2, + "learning_rate": 2.042227884965417e-06, + "loss": 0.6862, + "step": 17590 + }, + { + "epoch": 3.2, + "learning_rate": 2.042173279941755e-06, + "loss": 0.2485, + "step": 17591 + }, + { + "epoch": 3.2, + "learning_rate": 2.0421186749180927e-06, + "loss": 1.3472, + "step": 17592 + }, + { + "epoch": 3.2, + "learning_rate": 2.0420640698944303e-06, + "loss": 0.4374, + "step": 17593 + }, + { + "epoch": 3.2, + "learning_rate": 2.042009464870768e-06, + "loss": 0.8165, + "step": 17594 + }, + { + "epoch": 3.2, + "learning_rate": 2.041954859847106e-06, + "loss": 0.0226, + "step": 17595 + }, + { + "epoch": 3.2, + "learning_rate": 2.0419002548234437e-06, + "loss": 1.4273, + "step": 17596 + }, + { + "epoch": 3.2, + "learning_rate": 2.0418456497997816e-06, + "loss": 0.3125, + "step": 17597 + }, + { + "epoch": 3.2, + "learning_rate": 2.0417910447761196e-06, + "loss": 0.6024, + "step": 17598 + }, + { + "epoch": 3.2, + "learning_rate": 2.041736439752457e-06, + "loss": 2.3707, + "step": 17599 + }, + { + "epoch": 3.2, + "learning_rate": 2.041681834728795e-06, + "loss": 0.6301, + "step": 17600 + }, + { + "epoch": 3.2, + "learning_rate": 2.041627229705133e-06, + "loss": 2.0847, + "step": 17601 + }, + { + "epoch": 3.2, + "learning_rate": 2.0415726246814705e-06, + "loss": 1.3467, + "step": 17602 + }, + { + "epoch": 3.2, + "learning_rate": 2.0415180196578084e-06, + "loss": 2.0063, + "step": 17603 + }, + { + "epoch": 3.2, + "learning_rate": 2.0414634146341464e-06, + "loss": 0.4463, + "step": 17604 + }, + { + "epoch": 3.2, + "learning_rate": 2.0414088096104843e-06, + "loss": 0.0305, + "step": 17605 + }, + { + "epoch": 3.2, + "learning_rate": 2.0413542045868223e-06, + "loss": 0.0449, + "step": 17606 + }, + { + "epoch": 3.2, + "learning_rate": 2.0412995995631598e-06, + "loss": 3.6915, + "step": 17607 + }, + { + "epoch": 3.2, + "learning_rate": 2.0412449945394977e-06, + "loss": 0.7138, + "step": 17608 + }, + { + "epoch": 3.2, + "learning_rate": 2.0411903895158357e-06, + "loss": 0.4704, + "step": 17609 + }, + { + "epoch": 3.2, + "learning_rate": 2.041135784492173e-06, + "loss": 2.3634, + "step": 17610 + }, + { + "epoch": 3.2, + "learning_rate": 2.041081179468511e-06, + "loss": 0.1639, + "step": 17611 + }, + { + "epoch": 3.2, + "learning_rate": 2.041026574444849e-06, + "loss": 0.4564, + "step": 17612 + }, + { + "epoch": 3.2, + "learning_rate": 2.0409719694211866e-06, + "loss": 0.6177, + "step": 17613 + }, + { + "epoch": 3.2, + "learning_rate": 2.0409173643975245e-06, + "loss": 0.0363, + "step": 17614 + }, + { + "epoch": 3.2, + "learning_rate": 2.0408627593738625e-06, + "loss": 0.2252, + "step": 17615 + }, + { + "epoch": 3.2, + "learning_rate": 2.0408081543502e-06, + "loss": 0.6382, + "step": 17616 + }, + { + "epoch": 3.2, + "learning_rate": 2.040753549326538e-06, + "loss": 2.0642, + "step": 17617 + }, + { + "epoch": 3.2, + "learning_rate": 2.040698944302876e-06, + "loss": 1.4004, + "step": 17618 + }, + { + "epoch": 3.2, + "learning_rate": 2.040644339279214e-06, + "loss": 1.3352, + "step": 17619 + }, + { + "epoch": 3.2, + "learning_rate": 2.0405897342555518e-06, + "loss": 0.8447, + "step": 17620 + }, + { + "epoch": 3.2, + "learning_rate": 2.0405351292318893e-06, + "loss": 0.13, + "step": 17621 + }, + { + "epoch": 3.2, + "learning_rate": 2.0404805242082272e-06, + "loss": 0.5067, + "step": 17622 + }, + { + "epoch": 3.2, + "learning_rate": 2.040425919184565e-06, + "loss": 1.9098, + "step": 17623 + }, + { + "epoch": 3.2, + "learning_rate": 2.0403713141609027e-06, + "loss": 1.3531, + "step": 17624 + }, + { + "epoch": 3.21, + "learning_rate": 2.0403167091372406e-06, + "loss": 0.0365, + "step": 17625 + }, + { + "epoch": 3.21, + "learning_rate": 2.0402621041135786e-06, + "loss": 1.4224, + "step": 17626 + }, + { + "epoch": 3.21, + "learning_rate": 2.040207499089916e-06, + "loss": 0.0198, + "step": 17627 + }, + { + "epoch": 3.21, + "learning_rate": 2.040152894066254e-06, + "loss": 0.0302, + "step": 17628 + }, + { + "epoch": 3.21, + "learning_rate": 2.040098289042592e-06, + "loss": 2.6448, + "step": 17629 + }, + { + "epoch": 3.21, + "learning_rate": 2.0400436840189295e-06, + "loss": 0.0484, + "step": 17630 + }, + { + "epoch": 3.21, + "learning_rate": 2.0399890789952675e-06, + "loss": 0.4024, + "step": 17631 + }, + { + "epoch": 3.21, + "learning_rate": 2.0399344739716054e-06, + "loss": 1.1771, + "step": 17632 + }, + { + "epoch": 3.21, + "learning_rate": 2.0398798689479433e-06, + "loss": 0.7739, + "step": 17633 + }, + { + "epoch": 3.21, + "learning_rate": 2.0398252639242813e-06, + "loss": 0.8452, + "step": 17634 + }, + { + "epoch": 3.21, + "learning_rate": 2.039770658900619e-06, + "loss": 0.103, + "step": 17635 + }, + { + "epoch": 3.21, + "learning_rate": 2.0397160538769567e-06, + "loss": 0.8735, + "step": 17636 + }, + { + "epoch": 3.21, + "learning_rate": 2.0396614488532947e-06, + "loss": 0.1534, + "step": 17637 + }, + { + "epoch": 3.21, + "learning_rate": 2.039606843829632e-06, + "loss": 0.5692, + "step": 17638 + }, + { + "epoch": 3.21, + "learning_rate": 2.03955223880597e-06, + "loss": 1.6731, + "step": 17639 + }, + { + "epoch": 3.21, + "learning_rate": 2.039497633782308e-06, + "loss": 0.2298, + "step": 17640 + }, + { + "epoch": 3.21, + "learning_rate": 2.0394430287586456e-06, + "loss": 0.4074, + "step": 17641 + }, + { + "epoch": 3.21, + "learning_rate": 2.0393884237349836e-06, + "loss": 0.2055, + "step": 17642 + }, + { + "epoch": 3.21, + "learning_rate": 2.0393338187113215e-06, + "loss": 0.4031, + "step": 17643 + }, + { + "epoch": 3.21, + "learning_rate": 2.0392792136876594e-06, + "loss": 0.5573, + "step": 17644 + }, + { + "epoch": 3.21, + "learning_rate": 2.0392246086639974e-06, + "loss": 0.02, + "step": 17645 + }, + { + "epoch": 3.21, + "learning_rate": 2.0391700036403353e-06, + "loss": 1.0935, + "step": 17646 + }, + { + "epoch": 3.21, + "learning_rate": 2.039115398616673e-06, + "loss": 1.0002, + "step": 17647 + }, + { + "epoch": 3.21, + "learning_rate": 2.039060793593011e-06, + "loss": 1.1129, + "step": 17648 + }, + { + "epoch": 3.21, + "learning_rate": 2.0390061885693483e-06, + "loss": 0.6524, + "step": 17649 + }, + { + "epoch": 3.21, + "learning_rate": 2.0389515835456863e-06, + "loss": 0.4177, + "step": 17650 + }, + { + "epoch": 3.21, + "learning_rate": 2.038896978522024e-06, + "loss": 0.0372, + "step": 17651 + }, + { + "epoch": 3.21, + "learning_rate": 2.0388423734983617e-06, + "loss": 0.0453, + "step": 17652 + }, + { + "epoch": 3.21, + "learning_rate": 2.0387877684746997e-06, + "loss": 0.4979, + "step": 17653 + }, + { + "epoch": 3.21, + "learning_rate": 2.0387331634510376e-06, + "loss": 0.8862, + "step": 17654 + }, + { + "epoch": 3.21, + "learning_rate": 2.038678558427375e-06, + "loss": 2.0749, + "step": 17655 + }, + { + "epoch": 3.21, + "learning_rate": 2.038623953403713e-06, + "loss": 1.4852, + "step": 17656 + }, + { + "epoch": 3.21, + "learning_rate": 2.038569348380051e-06, + "loss": 1.2797, + "step": 17657 + }, + { + "epoch": 3.21, + "learning_rate": 2.038514743356389e-06, + "loss": 0.0488, + "step": 17658 + }, + { + "epoch": 3.21, + "learning_rate": 2.038460138332727e-06, + "loss": 0.102, + "step": 17659 + }, + { + "epoch": 3.21, + "learning_rate": 2.038405533309065e-06, + "loss": 0.0299, + "step": 17660 + }, + { + "epoch": 3.21, + "learning_rate": 2.0383509282854024e-06, + "loss": 0.9086, + "step": 17661 + }, + { + "epoch": 3.21, + "learning_rate": 2.0382963232617403e-06, + "loss": 0.2542, + "step": 17662 + }, + { + "epoch": 3.21, + "learning_rate": 2.038241718238078e-06, + "loss": 0.4715, + "step": 17663 + }, + { + "epoch": 3.21, + "learning_rate": 2.0381871132144158e-06, + "loss": 0.4129, + "step": 17664 + }, + { + "epoch": 3.21, + "learning_rate": 2.0381325081907537e-06, + "loss": 0.5797, + "step": 17665 + }, + { + "epoch": 3.21, + "learning_rate": 2.0380779031670912e-06, + "loss": 0.6674, + "step": 17666 + }, + { + "epoch": 3.21, + "learning_rate": 2.038023298143429e-06, + "loss": 0.1002, + "step": 17667 + }, + { + "epoch": 3.21, + "learning_rate": 2.037968693119767e-06, + "loss": 0.2536, + "step": 17668 + }, + { + "epoch": 3.21, + "learning_rate": 2.0379140880961046e-06, + "loss": 0.6239, + "step": 17669 + }, + { + "epoch": 3.21, + "learning_rate": 2.0378594830724426e-06, + "loss": 0.8709, + "step": 17670 + }, + { + "epoch": 3.21, + "learning_rate": 2.0378048780487805e-06, + "loss": 0.3307, + "step": 17671 + }, + { + "epoch": 3.21, + "learning_rate": 2.0377502730251185e-06, + "loss": 1.1968, + "step": 17672 + }, + { + "epoch": 3.21, + "learning_rate": 2.0376956680014564e-06, + "loss": 0.0124, + "step": 17673 + }, + { + "epoch": 3.21, + "learning_rate": 2.037641062977794e-06, + "loss": 0.0619, + "step": 17674 + }, + { + "epoch": 3.21, + "learning_rate": 2.037586457954132e-06, + "loss": 1.6224, + "step": 17675 + }, + { + "epoch": 3.21, + "learning_rate": 2.03753185293047e-06, + "loss": 0.0407, + "step": 17676 + }, + { + "epoch": 3.21, + "learning_rate": 2.0374772479068073e-06, + "loss": 1.7177, + "step": 17677 + }, + { + "epoch": 3.21, + "learning_rate": 2.0374226428831453e-06, + "loss": 1.3333, + "step": 17678 + }, + { + "epoch": 3.21, + "learning_rate": 2.0373680378594832e-06, + "loss": 0.3418, + "step": 17679 + }, + { + "epoch": 3.22, + "learning_rate": 2.0373134328358208e-06, + "loss": 1.2677, + "step": 17680 + }, + { + "epoch": 3.22, + "learning_rate": 2.0372588278121587e-06, + "loss": 0.5544, + "step": 17681 + }, + { + "epoch": 3.22, + "learning_rate": 2.0372042227884966e-06, + "loss": 2.5046, + "step": 17682 + }, + { + "epoch": 3.22, + "learning_rate": 2.037149617764834e-06, + "loss": 1.2366, + "step": 17683 + }, + { + "epoch": 3.22, + "learning_rate": 2.037095012741172e-06, + "loss": 0.3634, + "step": 17684 + }, + { + "epoch": 3.22, + "learning_rate": 2.03704040771751e-06, + "loss": 0.0163, + "step": 17685 + }, + { + "epoch": 3.22, + "learning_rate": 2.036985802693848e-06, + "loss": 0.0205, + "step": 17686 + }, + { + "epoch": 3.22, + "learning_rate": 2.036931197670186e-06, + "loss": 0.7101, + "step": 17687 + }, + { + "epoch": 3.22, + "learning_rate": 2.0368765926465235e-06, + "loss": 0.3966, + "step": 17688 + }, + { + "epoch": 3.22, + "learning_rate": 2.0368219876228614e-06, + "loss": 0.8551, + "step": 17689 + }, + { + "epoch": 3.22, + "learning_rate": 2.0367673825991993e-06, + "loss": 1.2992, + "step": 17690 + }, + { + "epoch": 3.22, + "learning_rate": 2.036712777575537e-06, + "loss": 2.2905, + "step": 17691 + }, + { + "epoch": 3.22, + "learning_rate": 2.036658172551875e-06, + "loss": 0.6354, + "step": 17692 + }, + { + "epoch": 3.22, + "learning_rate": 2.0366035675282127e-06, + "loss": 1.8398, + "step": 17693 + }, + { + "epoch": 3.22, + "learning_rate": 2.0365489625045503e-06, + "loss": 0.5003, + "step": 17694 + }, + { + "epoch": 3.22, + "learning_rate": 2.036494357480888e-06, + "loss": 1.4244, + "step": 17695 + }, + { + "epoch": 3.22, + "learning_rate": 2.036439752457226e-06, + "loss": 1.1296, + "step": 17696 + }, + { + "epoch": 3.22, + "learning_rate": 2.0363851474335637e-06, + "loss": 0.1024, + "step": 17697 + }, + { + "epoch": 3.22, + "learning_rate": 2.0363305424099016e-06, + "loss": 1.4081, + "step": 17698 + }, + { + "epoch": 3.22, + "learning_rate": 2.0362759373862396e-06, + "loss": 0.3783, + "step": 17699 + }, + { + "epoch": 3.22, + "learning_rate": 2.0362213323625775e-06, + "loss": 0.6128, + "step": 17700 + }, + { + "epoch": 3.22, + "learning_rate": 2.0361667273389154e-06, + "loss": 0.2236, + "step": 17701 + }, + { + "epoch": 3.22, + "learning_rate": 2.036112122315253e-06, + "loss": 0.4114, + "step": 17702 + }, + { + "epoch": 3.22, + "learning_rate": 2.036057517291591e-06, + "loss": 0.1228, + "step": 17703 + }, + { + "epoch": 3.22, + "learning_rate": 2.036002912267929e-06, + "loss": 0.0478, + "step": 17704 + }, + { + "epoch": 3.22, + "learning_rate": 2.0359483072442664e-06, + "loss": 0.5398, + "step": 17705 + }, + { + "epoch": 3.22, + "learning_rate": 2.0358937022206043e-06, + "loss": 0.4804, + "step": 17706 + }, + { + "epoch": 3.22, + "learning_rate": 2.0358390971969423e-06, + "loss": 0.0315, + "step": 17707 + }, + { + "epoch": 3.22, + "learning_rate": 2.0357844921732798e-06, + "loss": 1.1678, + "step": 17708 + }, + { + "epoch": 3.22, + "learning_rate": 2.0357298871496177e-06, + "loss": 2.0687, + "step": 17709 + }, + { + "epoch": 3.22, + "learning_rate": 2.0356752821259557e-06, + "loss": 0.025, + "step": 17710 + }, + { + "epoch": 3.22, + "learning_rate": 2.035620677102293e-06, + "loss": 0.4573, + "step": 17711 + }, + { + "epoch": 3.22, + "learning_rate": 2.035566072078631e-06, + "loss": 0.2537, + "step": 17712 + }, + { + "epoch": 3.22, + "learning_rate": 2.035511467054969e-06, + "loss": 1.2665, + "step": 17713 + }, + { + "epoch": 3.22, + "learning_rate": 2.035456862031307e-06, + "loss": 0.9874, + "step": 17714 + }, + { + "epoch": 3.22, + "learning_rate": 2.035402257007645e-06, + "loss": 1.3894, + "step": 17715 + }, + { + "epoch": 3.22, + "learning_rate": 2.0353476519839825e-06, + "loss": 0.5609, + "step": 17716 + }, + { + "epoch": 3.22, + "learning_rate": 2.0352930469603204e-06, + "loss": 0.0117, + "step": 17717 + }, + { + "epoch": 3.22, + "learning_rate": 2.0352384419366584e-06, + "loss": 1.659, + "step": 17718 + }, + { + "epoch": 3.22, + "learning_rate": 2.035183836912996e-06, + "loss": 0.5634, + "step": 17719 + }, + { + "epoch": 3.22, + "learning_rate": 2.035129231889334e-06, + "loss": 1.1602, + "step": 17720 + }, + { + "epoch": 3.22, + "learning_rate": 2.0350746268656718e-06, + "loss": 0.7951, + "step": 17721 + }, + { + "epoch": 3.22, + "learning_rate": 2.0350200218420093e-06, + "loss": 0.1107, + "step": 17722 + }, + { + "epoch": 3.22, + "learning_rate": 2.0349654168183472e-06, + "loss": 1.5532, + "step": 17723 + }, + { + "epoch": 3.22, + "learning_rate": 2.034910811794685e-06, + "loss": 0.046, + "step": 17724 + }, + { + "epoch": 3.22, + "learning_rate": 2.0348562067710227e-06, + "loss": 0.0218, + "step": 17725 + }, + { + "epoch": 3.22, + "learning_rate": 2.0348016017473606e-06, + "loss": 0.898, + "step": 17726 + }, + { + "epoch": 3.22, + "learning_rate": 2.0347469967236986e-06, + "loss": 0.2195, + "step": 17727 + }, + { + "epoch": 3.22, + "learning_rate": 2.0346923917000365e-06, + "loss": 0.7022, + "step": 17728 + }, + { + "epoch": 3.22, + "learning_rate": 2.0346377866763745e-06, + "loss": 3.2811, + "step": 17729 + }, + { + "epoch": 3.22, + "learning_rate": 2.034583181652712e-06, + "loss": 2.7749, + "step": 17730 + }, + { + "epoch": 3.22, + "learning_rate": 2.03452857662905e-06, + "loss": 0.824, + "step": 17731 + }, + { + "epoch": 3.22, + "learning_rate": 2.034473971605388e-06, + "loss": 1.0482, + "step": 17732 + }, + { + "epoch": 3.22, + "learning_rate": 2.0344193665817254e-06, + "loss": 0.7652, + "step": 17733 + }, + { + "epoch": 3.22, + "learning_rate": 2.0343647615580633e-06, + "loss": 2.1789, + "step": 17734 + }, + { + "epoch": 3.23, + "learning_rate": 2.0343101565344013e-06, + "loss": 0.0594, + "step": 17735 + }, + { + "epoch": 3.23, + "learning_rate": 2.034255551510739e-06, + "loss": 0.0253, + "step": 17736 + }, + { + "epoch": 3.23, + "learning_rate": 2.0342009464870767e-06, + "loss": 0.5317, + "step": 17737 + }, + { + "epoch": 3.23, + "learning_rate": 2.0341463414634147e-06, + "loss": 1.1545, + "step": 17738 + }, + { + "epoch": 3.23, + "learning_rate": 2.0340917364397526e-06, + "loss": 0.5471, + "step": 17739 + }, + { + "epoch": 3.23, + "learning_rate": 2.0340371314160906e-06, + "loss": 0.1277, + "step": 17740 + }, + { + "epoch": 3.23, + "learning_rate": 2.0339825263924285e-06, + "loss": 0.4221, + "step": 17741 + }, + { + "epoch": 3.23, + "learning_rate": 2.033927921368766e-06, + "loss": 0.2123, + "step": 17742 + }, + { + "epoch": 3.23, + "learning_rate": 2.033873316345104e-06, + "loss": 0.5851, + "step": 17743 + }, + { + "epoch": 3.23, + "learning_rate": 2.0338187113214415e-06, + "loss": 0.8685, + "step": 17744 + }, + { + "epoch": 3.23, + "learning_rate": 2.0337641062977794e-06, + "loss": 1.195, + "step": 17745 + }, + { + "epoch": 3.23, + "learning_rate": 2.0337095012741174e-06, + "loss": 1.8478, + "step": 17746 + }, + { + "epoch": 3.23, + "learning_rate": 2.033654896250455e-06, + "loss": 0.2198, + "step": 17747 + }, + { + "epoch": 3.23, + "learning_rate": 2.033600291226793e-06, + "loss": 1.0563, + "step": 17748 + }, + { + "epoch": 3.23, + "learning_rate": 2.033545686203131e-06, + "loss": 0.0191, + "step": 17749 + }, + { + "epoch": 3.23, + "learning_rate": 2.0334910811794683e-06, + "loss": 1.0508, + "step": 17750 + }, + { + "epoch": 3.23, + "learning_rate": 2.0334364761558063e-06, + "loss": 0.8078, + "step": 17751 + }, + { + "epoch": 3.23, + "learning_rate": 2.033381871132144e-06, + "loss": 0.2383, + "step": 17752 + }, + { + "epoch": 3.23, + "learning_rate": 2.033327266108482e-06, + "loss": 0.0272, + "step": 17753 + }, + { + "epoch": 3.23, + "learning_rate": 2.03327266108482e-06, + "loss": 2.3074, + "step": 17754 + }, + { + "epoch": 3.23, + "learning_rate": 2.0332180560611576e-06, + "loss": 1.1581, + "step": 17755 + }, + { + "epoch": 3.23, + "learning_rate": 2.0331634510374956e-06, + "loss": 0.0191, + "step": 17756 + }, + { + "epoch": 3.23, + "learning_rate": 2.0331088460138335e-06, + "loss": 0.5302, + "step": 17757 + }, + { + "epoch": 3.23, + "learning_rate": 2.033054240990171e-06, + "loss": 0.1747, + "step": 17758 + }, + { + "epoch": 3.23, + "learning_rate": 2.032999635966509e-06, + "loss": 2.581, + "step": 17759 + }, + { + "epoch": 3.23, + "learning_rate": 2.032945030942847e-06, + "loss": 1.2817, + "step": 17760 + }, + { + "epoch": 3.23, + "learning_rate": 2.0328904259191844e-06, + "loss": 0.5037, + "step": 17761 + }, + { + "epoch": 3.23, + "learning_rate": 2.0328358208955224e-06, + "loss": 0.5261, + "step": 17762 + }, + { + "epoch": 3.23, + "learning_rate": 2.0327812158718603e-06, + "loss": 0.3402, + "step": 17763 + }, + { + "epoch": 3.23, + "learning_rate": 2.032726610848198e-06, + "loss": 2.8006, + "step": 17764 + }, + { + "epoch": 3.23, + "learning_rate": 2.0326720058245358e-06, + "loss": 0.072, + "step": 17765 + }, + { + "epoch": 3.23, + "learning_rate": 2.0326174008008737e-06, + "loss": 0.0166, + "step": 17766 + }, + { + "epoch": 3.23, + "learning_rate": 2.0325627957772117e-06, + "loss": 1.088, + "step": 17767 + }, + { + "epoch": 3.23, + "learning_rate": 2.0325081907535496e-06, + "loss": 0.8384, + "step": 17768 + }, + { + "epoch": 3.23, + "learning_rate": 2.032453585729887e-06, + "loss": 0.9067, + "step": 17769 + }, + { + "epoch": 3.23, + "learning_rate": 2.032398980706225e-06, + "loss": 1.0308, + "step": 17770 + }, + { + "epoch": 3.23, + "learning_rate": 2.032344375682563e-06, + "loss": 0.3837, + "step": 17771 + }, + { + "epoch": 3.23, + "learning_rate": 2.0322897706589005e-06, + "loss": 0.631, + "step": 17772 + }, + { + "epoch": 3.23, + "learning_rate": 2.0322351656352385e-06, + "loss": 0.341, + "step": 17773 + }, + { + "epoch": 3.23, + "learning_rate": 2.0321805606115764e-06, + "loss": 0.2209, + "step": 17774 + }, + { + "epoch": 3.23, + "learning_rate": 2.032125955587914e-06, + "loss": 3.3353, + "step": 17775 + }, + { + "epoch": 3.23, + "learning_rate": 2.032071350564252e-06, + "loss": 0.2292, + "step": 17776 + }, + { + "epoch": 3.23, + "learning_rate": 2.03201674554059e-06, + "loss": 1.7168, + "step": 17777 + }, + { + "epoch": 3.23, + "learning_rate": 2.0319621405169273e-06, + "loss": 0.8994, + "step": 17778 + }, + { + "epoch": 3.23, + "learning_rate": 2.0319075354932653e-06, + "loss": 0.0769, + "step": 17779 + }, + { + "epoch": 3.23, + "learning_rate": 2.0318529304696032e-06, + "loss": 2.4512, + "step": 17780 + }, + { + "epoch": 3.23, + "learning_rate": 2.031798325445941e-06, + "loss": 0.4247, + "step": 17781 + }, + { + "epoch": 3.23, + "learning_rate": 2.031743720422279e-06, + "loss": 1.656, + "step": 17782 + }, + { + "epoch": 3.23, + "learning_rate": 2.0316891153986166e-06, + "loss": 1.6044, + "step": 17783 + }, + { + "epoch": 3.23, + "learning_rate": 2.0316345103749546e-06, + "loss": 0.2859, + "step": 17784 + }, + { + "epoch": 3.23, + "learning_rate": 2.0315799053512925e-06, + "loss": 0.0198, + "step": 17785 + }, + { + "epoch": 3.23, + "learning_rate": 2.03152530032763e-06, + "loss": 3.8052, + "step": 17786 + }, + { + "epoch": 3.23, + "learning_rate": 2.031470695303968e-06, + "loss": 0.1449, + "step": 17787 + }, + { + "epoch": 3.23, + "learning_rate": 2.031416090280306e-06, + "loss": 0.6841, + "step": 17788 + }, + { + "epoch": 3.23, + "learning_rate": 2.0313614852566435e-06, + "loss": 0.0103, + "step": 17789 + }, + { + "epoch": 3.24, + "learning_rate": 2.0313068802329814e-06, + "loss": 0.1233, + "step": 17790 + }, + { + "epoch": 3.24, + "learning_rate": 2.0312522752093193e-06, + "loss": 0.5534, + "step": 17791 + }, + { + "epoch": 3.24, + "learning_rate": 2.031197670185657e-06, + "loss": 0.027, + "step": 17792 + }, + { + "epoch": 3.24, + "learning_rate": 2.031143065161995e-06, + "loss": 0.307, + "step": 17793 + }, + { + "epoch": 3.24, + "learning_rate": 2.0310884601383327e-06, + "loss": 1.2694, + "step": 17794 + }, + { + "epoch": 3.24, + "learning_rate": 2.0310338551146707e-06, + "loss": 0.9675, + "step": 17795 + }, + { + "epoch": 3.24, + "learning_rate": 2.0309792500910086e-06, + "loss": 0.7806, + "step": 17796 + }, + { + "epoch": 3.24, + "learning_rate": 2.030924645067346e-06, + "loss": 0.07, + "step": 17797 + }, + { + "epoch": 3.24, + "learning_rate": 2.030870040043684e-06, + "loss": 1.2856, + "step": 17798 + }, + { + "epoch": 3.24, + "learning_rate": 2.030815435020022e-06, + "loss": 0.2688, + "step": 17799 + }, + { + "epoch": 3.24, + "learning_rate": 2.0307608299963596e-06, + "loss": 1.0138, + "step": 17800 + }, + { + "epoch": 3.24, + "learning_rate": 2.0307062249726975e-06, + "loss": 0.8499, + "step": 17801 + }, + { + "epoch": 3.24, + "learning_rate": 2.0306516199490354e-06, + "loss": 0.3956, + "step": 17802 + }, + { + "epoch": 3.24, + "learning_rate": 2.030597014925373e-06, + "loss": 0.1881, + "step": 17803 + }, + { + "epoch": 3.24, + "learning_rate": 2.030542409901711e-06, + "loss": 0.0296, + "step": 17804 + }, + { + "epoch": 3.24, + "learning_rate": 2.030487804878049e-06, + "loss": 0.7191, + "step": 17805 + }, + { + "epoch": 3.24, + "learning_rate": 2.0304331998543864e-06, + "loss": 0.2995, + "step": 17806 + }, + { + "epoch": 3.24, + "learning_rate": 2.0303785948307243e-06, + "loss": 0.3344, + "step": 17807 + }, + { + "epoch": 3.24, + "learning_rate": 2.0303239898070623e-06, + "loss": 1.2398, + "step": 17808 + }, + { + "epoch": 3.24, + "learning_rate": 2.0302693847834e-06, + "loss": 0.4628, + "step": 17809 + }, + { + "epoch": 3.24, + "learning_rate": 2.030214779759738e-06, + "loss": 0.4629, + "step": 17810 + }, + { + "epoch": 3.24, + "learning_rate": 2.0301601747360757e-06, + "loss": 0.8572, + "step": 17811 + }, + { + "epoch": 3.24, + "learning_rate": 2.0301055697124136e-06, + "loss": 0.5525, + "step": 17812 + }, + { + "epoch": 3.24, + "learning_rate": 2.0300509646887516e-06, + "loss": 1.2039, + "step": 17813 + }, + { + "epoch": 3.24, + "learning_rate": 2.029996359665089e-06, + "loss": 0.1768, + "step": 17814 + }, + { + "epoch": 3.24, + "learning_rate": 2.029941754641427e-06, + "loss": 0.3517, + "step": 17815 + }, + { + "epoch": 3.24, + "learning_rate": 2.029887149617765e-06, + "loss": 0.5025, + "step": 17816 + }, + { + "epoch": 3.24, + "learning_rate": 2.0298325445941025e-06, + "loss": 0.0569, + "step": 17817 + }, + { + "epoch": 3.24, + "learning_rate": 2.0297779395704404e-06, + "loss": 0.0458, + "step": 17818 + }, + { + "epoch": 3.24, + "learning_rate": 2.0297233345467784e-06, + "loss": 1.0449, + "step": 17819 + }, + { + "epoch": 3.24, + "learning_rate": 2.0296687295231163e-06, + "loss": 0.2739, + "step": 17820 + }, + { + "epoch": 3.24, + "learning_rate": 2.0296141244994543e-06, + "loss": 1.2471, + "step": 17821 + }, + { + "epoch": 3.24, + "learning_rate": 2.0295595194757918e-06, + "loss": 0.8818, + "step": 17822 + }, + { + "epoch": 3.24, + "learning_rate": 2.0295049144521297e-06, + "loss": 0.0525, + "step": 17823 + }, + { + "epoch": 3.24, + "learning_rate": 2.0294503094284677e-06, + "loss": 0.2951, + "step": 17824 + }, + { + "epoch": 3.24, + "learning_rate": 2.029395704404805e-06, + "loss": 0.6594, + "step": 17825 + }, + { + "epoch": 3.24, + "learning_rate": 2.029341099381143e-06, + "loss": 0.7809, + "step": 17826 + }, + { + "epoch": 3.24, + "learning_rate": 2.029286494357481e-06, + "loss": 0.9362, + "step": 17827 + }, + { + "epoch": 3.24, + "learning_rate": 2.0292318893338186e-06, + "loss": 1.8698, + "step": 17828 + }, + { + "epoch": 3.24, + "learning_rate": 2.0291772843101565e-06, + "loss": 0.4476, + "step": 17829 + }, + { + "epoch": 3.24, + "learning_rate": 2.0291226792864945e-06, + "loss": 0.2957, + "step": 17830 + }, + { + "epoch": 3.24, + "learning_rate": 2.029068074262832e-06, + "loss": 0.2994, + "step": 17831 + }, + { + "epoch": 3.24, + "learning_rate": 2.02901346923917e-06, + "loss": 0.7521, + "step": 17832 + }, + { + "epoch": 3.24, + "learning_rate": 2.028958864215508e-06, + "loss": 0.012, + "step": 17833 + }, + { + "epoch": 3.24, + "learning_rate": 2.028904259191846e-06, + "loss": 0.159, + "step": 17834 + }, + { + "epoch": 3.24, + "learning_rate": 2.0288496541681838e-06, + "loss": 0.0533, + "step": 17835 + }, + { + "epoch": 3.24, + "learning_rate": 2.0287950491445213e-06, + "loss": 0.2622, + "step": 17836 + }, + { + "epoch": 3.24, + "learning_rate": 2.0287404441208592e-06, + "loss": 0.0204, + "step": 17837 + }, + { + "epoch": 3.24, + "learning_rate": 2.028685839097197e-06, + "loss": 0.5672, + "step": 17838 + }, + { + "epoch": 3.24, + "learning_rate": 2.0286312340735347e-06, + "loss": 2.5504, + "step": 17839 + }, + { + "epoch": 3.24, + "learning_rate": 2.0285766290498726e-06, + "loss": 1.9275, + "step": 17840 + }, + { + "epoch": 3.24, + "learning_rate": 2.0285220240262106e-06, + "loss": 0.2318, + "step": 17841 + }, + { + "epoch": 3.24, + "learning_rate": 2.028467419002548e-06, + "loss": 2.0313, + "step": 17842 + }, + { + "epoch": 3.24, + "learning_rate": 2.028412813978886e-06, + "loss": 0.0477, + "step": 17843 + }, + { + "epoch": 3.24, + "learning_rate": 2.028358208955224e-06, + "loss": 0.5579, + "step": 17844 + }, + { + "epoch": 3.25, + "learning_rate": 2.0283036039315615e-06, + "loss": 0.5601, + "step": 17845 + }, + { + "epoch": 3.25, + "learning_rate": 2.0282489989078994e-06, + "loss": 1.4681, + "step": 17846 + }, + { + "epoch": 3.25, + "learning_rate": 2.0281943938842374e-06, + "loss": 0.488, + "step": 17847 + }, + { + "epoch": 3.25, + "learning_rate": 2.0281397888605753e-06, + "loss": 0.1483, + "step": 17848 + }, + { + "epoch": 3.25, + "learning_rate": 2.0280851838369133e-06, + "loss": 1.1934, + "step": 17849 + }, + { + "epoch": 3.25, + "learning_rate": 2.028030578813251e-06, + "loss": 0.642, + "step": 17850 + }, + { + "epoch": 3.25, + "learning_rate": 2.0279759737895887e-06, + "loss": 0.5001, + "step": 17851 + }, + { + "epoch": 3.25, + "learning_rate": 2.0279213687659267e-06, + "loss": 0.2729, + "step": 17852 + }, + { + "epoch": 3.25, + "learning_rate": 2.027866763742264e-06, + "loss": 0.5481, + "step": 17853 + }, + { + "epoch": 3.25, + "learning_rate": 2.027812158718602e-06, + "loss": 0.324, + "step": 17854 + }, + { + "epoch": 3.25, + "learning_rate": 2.02775755369494e-06, + "loss": 0.225, + "step": 17855 + }, + { + "epoch": 3.25, + "learning_rate": 2.0277029486712776e-06, + "loss": 1.288, + "step": 17856 + }, + { + "epoch": 3.25, + "learning_rate": 2.0276483436476156e-06, + "loss": 0.0651, + "step": 17857 + }, + { + "epoch": 3.25, + "learning_rate": 2.0275937386239535e-06, + "loss": 0.2555, + "step": 17858 + }, + { + "epoch": 3.25, + "learning_rate": 2.027539133600291e-06, + "loss": 0.1943, + "step": 17859 + }, + { + "epoch": 3.25, + "learning_rate": 2.027484528576629e-06, + "loss": 0.3054, + "step": 17860 + }, + { + "epoch": 3.25, + "learning_rate": 2.027429923552967e-06, + "loss": 1.0452, + "step": 17861 + }, + { + "epoch": 3.25, + "learning_rate": 2.027375318529305e-06, + "loss": 1.2152, + "step": 17862 + }, + { + "epoch": 3.25, + "learning_rate": 2.027320713505643e-06, + "loss": 0.0126, + "step": 17863 + }, + { + "epoch": 3.25, + "learning_rate": 2.0272661084819803e-06, + "loss": 0.0291, + "step": 17864 + }, + { + "epoch": 3.25, + "learning_rate": 2.0272115034583183e-06, + "loss": 1.3098, + "step": 17865 + }, + { + "epoch": 3.25, + "learning_rate": 2.027156898434656e-06, + "loss": 0.6157, + "step": 17866 + }, + { + "epoch": 3.25, + "learning_rate": 2.0271022934109937e-06, + "loss": 0.2978, + "step": 17867 + }, + { + "epoch": 3.25, + "learning_rate": 2.0270476883873317e-06, + "loss": 1.7805, + "step": 17868 + }, + { + "epoch": 3.25, + "learning_rate": 2.0269930833636696e-06, + "loss": 0.1362, + "step": 17869 + }, + { + "epoch": 3.25, + "learning_rate": 2.026938478340007e-06, + "loss": 0.5668, + "step": 17870 + }, + { + "epoch": 3.25, + "learning_rate": 2.026883873316345e-06, + "loss": 1.5282, + "step": 17871 + }, + { + "epoch": 3.25, + "learning_rate": 2.026829268292683e-06, + "loss": 0.4657, + "step": 17872 + }, + { + "epoch": 3.25, + "learning_rate": 2.0267746632690205e-06, + "loss": 0.6232, + "step": 17873 + }, + { + "epoch": 3.25, + "learning_rate": 2.0267200582453585e-06, + "loss": 1.2232, + "step": 17874 + }, + { + "epoch": 3.25, + "learning_rate": 2.0266654532216964e-06, + "loss": 0.0782, + "step": 17875 + }, + { + "epoch": 3.25, + "learning_rate": 2.0266108481980344e-06, + "loss": 0.2688, + "step": 17876 + }, + { + "epoch": 3.25, + "learning_rate": 2.0265562431743723e-06, + "loss": 1.809, + "step": 17877 + }, + { + "epoch": 3.25, + "learning_rate": 2.02650163815071e-06, + "loss": 0.3268, + "step": 17878 + }, + { + "epoch": 3.25, + "learning_rate": 2.0264470331270478e-06, + "loss": 1.1361, + "step": 17879 + }, + { + "epoch": 3.25, + "learning_rate": 2.0263924281033857e-06, + "loss": 1.2226, + "step": 17880 + }, + { + "epoch": 3.25, + "learning_rate": 2.0263378230797232e-06, + "loss": 0.1236, + "step": 17881 + }, + { + "epoch": 3.25, + "learning_rate": 2.026283218056061e-06, + "loss": 0.5759, + "step": 17882 + }, + { + "epoch": 3.25, + "learning_rate": 2.026228613032399e-06, + "loss": 1.7824, + "step": 17883 + }, + { + "epoch": 3.25, + "learning_rate": 2.0261740080087366e-06, + "loss": 0.1118, + "step": 17884 + }, + { + "epoch": 3.25, + "learning_rate": 2.0261194029850746e-06, + "loss": 0.0206, + "step": 17885 + }, + { + "epoch": 3.25, + "learning_rate": 2.0260647979614125e-06, + "loss": 0.1913, + "step": 17886 + }, + { + "epoch": 3.25, + "learning_rate": 2.02601019293775e-06, + "loss": 0.3057, + "step": 17887 + }, + { + "epoch": 3.25, + "learning_rate": 2.025955587914088e-06, + "loss": 0.0161, + "step": 17888 + }, + { + "epoch": 3.25, + "learning_rate": 2.025900982890426e-06, + "loss": 1.7017, + "step": 17889 + }, + { + "epoch": 3.25, + "learning_rate": 2.025846377866764e-06, + "loss": 1.3871, + "step": 17890 + }, + { + "epoch": 3.25, + "learning_rate": 2.025791772843102e-06, + "loss": 0.0699, + "step": 17891 + }, + { + "epoch": 3.25, + "learning_rate": 2.0257371678194393e-06, + "loss": 1.5681, + "step": 17892 + }, + { + "epoch": 3.25, + "learning_rate": 2.0256825627957773e-06, + "loss": 1.4235, + "step": 17893 + }, + { + "epoch": 3.25, + "learning_rate": 2.0256279577721152e-06, + "loss": 0.194, + "step": 17894 + }, + { + "epoch": 3.25, + "learning_rate": 2.0255733527484527e-06, + "loss": 3.2407, + "step": 17895 + }, + { + "epoch": 3.25, + "learning_rate": 2.0255187477247907e-06, + "loss": 2.3218, + "step": 17896 + }, + { + "epoch": 3.25, + "learning_rate": 2.0254641427011286e-06, + "loss": 0.5715, + "step": 17897 + }, + { + "epoch": 3.25, + "learning_rate": 2.025409537677466e-06, + "loss": 0.153, + "step": 17898 + }, + { + "epoch": 3.25, + "learning_rate": 2.025354932653804e-06, + "loss": 0.0461, + "step": 17899 + }, + { + "epoch": 3.26, + "learning_rate": 2.025300327630142e-06, + "loss": 0.7086, + "step": 17900 + }, + { + "epoch": 3.26, + "learning_rate": 2.02524572260648e-06, + "loss": 0.4186, + "step": 17901 + }, + { + "epoch": 3.26, + "learning_rate": 2.025191117582818e-06, + "loss": 0.5863, + "step": 17902 + }, + { + "epoch": 3.26, + "learning_rate": 2.0251365125591554e-06, + "loss": 0.0537, + "step": 17903 + }, + { + "epoch": 3.26, + "learning_rate": 2.0250819075354934e-06, + "loss": 2.7837, + "step": 17904 + }, + { + "epoch": 3.26, + "learning_rate": 2.0250273025118313e-06, + "loss": 0.8554, + "step": 17905 + }, + { + "epoch": 3.26, + "learning_rate": 2.024972697488169e-06, + "loss": 1.124, + "step": 17906 + }, + { + "epoch": 3.26, + "learning_rate": 2.024918092464507e-06, + "loss": 1.1018, + "step": 17907 + }, + { + "epoch": 3.26, + "learning_rate": 2.0248634874408447e-06, + "loss": 0.3171, + "step": 17908 + }, + { + "epoch": 3.26, + "learning_rate": 2.0248088824171823e-06, + "loss": 0.6461, + "step": 17909 + }, + { + "epoch": 3.26, + "learning_rate": 2.02475427739352e-06, + "loss": 0.8444, + "step": 17910 + }, + { + "epoch": 3.26, + "learning_rate": 2.024699672369858e-06, + "loss": 1.3976, + "step": 17911 + }, + { + "epoch": 3.26, + "learning_rate": 2.0246450673461957e-06, + "loss": 0.8614, + "step": 17912 + }, + { + "epoch": 3.26, + "learning_rate": 2.0245904623225336e-06, + "loss": 2.873, + "step": 17913 + }, + { + "epoch": 3.26, + "learning_rate": 2.0245358572988716e-06, + "loss": 0.0879, + "step": 17914 + }, + { + "epoch": 3.26, + "learning_rate": 2.0244812522752095e-06, + "loss": 0.8137, + "step": 17915 + }, + { + "epoch": 3.26, + "learning_rate": 2.0244266472515474e-06, + "loss": 0.4153, + "step": 17916 + }, + { + "epoch": 3.26, + "learning_rate": 2.024372042227885e-06, + "loss": 2.0187, + "step": 17917 + }, + { + "epoch": 3.26, + "learning_rate": 2.024317437204223e-06, + "loss": 1.1268, + "step": 17918 + }, + { + "epoch": 3.26, + "learning_rate": 2.024262832180561e-06, + "loss": 0.5214, + "step": 17919 + }, + { + "epoch": 3.26, + "learning_rate": 2.0242082271568984e-06, + "loss": 0.4065, + "step": 17920 + }, + { + "epoch": 3.26, + "learning_rate": 2.0241536221332363e-06, + "loss": 0.4576, + "step": 17921 + }, + { + "epoch": 3.26, + "learning_rate": 2.0240990171095743e-06, + "loss": 0.9901, + "step": 17922 + }, + { + "epoch": 3.26, + "learning_rate": 2.0240444120859118e-06, + "loss": 0.0415, + "step": 17923 + }, + { + "epoch": 3.26, + "learning_rate": 2.0239898070622497e-06, + "loss": 0.0204, + "step": 17924 + }, + { + "epoch": 3.26, + "learning_rate": 2.0239352020385877e-06, + "loss": 0.2732, + "step": 17925 + }, + { + "epoch": 3.26, + "learning_rate": 2.023880597014925e-06, + "loss": 2.3295, + "step": 17926 + }, + { + "epoch": 3.26, + "learning_rate": 2.023825991991263e-06, + "loss": 0.4516, + "step": 17927 + }, + { + "epoch": 3.26, + "learning_rate": 2.023771386967601e-06, + "loss": 1.9963, + "step": 17928 + }, + { + "epoch": 3.26, + "learning_rate": 2.023716781943939e-06, + "loss": 0.4146, + "step": 17929 + }, + { + "epoch": 3.26, + "learning_rate": 2.023662176920277e-06, + "loss": 1.8991, + "step": 17930 + }, + { + "epoch": 3.26, + "learning_rate": 2.0236075718966145e-06, + "loss": 1.9011, + "step": 17931 + }, + { + "epoch": 3.26, + "learning_rate": 2.0235529668729524e-06, + "loss": 0.2192, + "step": 17932 + }, + { + "epoch": 3.26, + "learning_rate": 2.0234983618492904e-06, + "loss": 0.3442, + "step": 17933 + }, + { + "epoch": 3.26, + "learning_rate": 2.023443756825628e-06, + "loss": 1.0799, + "step": 17934 + }, + { + "epoch": 3.26, + "learning_rate": 2.023389151801966e-06, + "loss": 0.4016, + "step": 17935 + }, + { + "epoch": 3.26, + "learning_rate": 2.0233345467783038e-06, + "loss": 0.4772, + "step": 17936 + }, + { + "epoch": 3.26, + "learning_rate": 2.0232799417546413e-06, + "loss": 0.1569, + "step": 17937 + }, + { + "epoch": 3.26, + "learning_rate": 2.0232253367309792e-06, + "loss": 0.1506, + "step": 17938 + }, + { + "epoch": 3.26, + "learning_rate": 2.023170731707317e-06, + "loss": 0.4478, + "step": 17939 + }, + { + "epoch": 3.26, + "learning_rate": 2.0231161266836547e-06, + "loss": 0.0366, + "step": 17940 + }, + { + "epoch": 3.26, + "learning_rate": 2.0230615216599926e-06, + "loss": 0.6256, + "step": 17941 + }, + { + "epoch": 3.26, + "learning_rate": 2.0230069166363306e-06, + "loss": 1.5894, + "step": 17942 + }, + { + "epoch": 3.26, + "learning_rate": 2.0229523116126685e-06, + "loss": 1.0201, + "step": 17943 + }, + { + "epoch": 3.26, + "learning_rate": 2.0228977065890065e-06, + "loss": 0.1551, + "step": 17944 + }, + { + "epoch": 3.26, + "learning_rate": 2.022843101565344e-06, + "loss": 1.0416, + "step": 17945 + }, + { + "epoch": 3.26, + "learning_rate": 2.022788496541682e-06, + "loss": 0.5459, + "step": 17946 + }, + { + "epoch": 3.26, + "learning_rate": 2.02273389151802e-06, + "loss": 1.2523, + "step": 17947 + }, + { + "epoch": 3.26, + "learning_rate": 2.0226792864943574e-06, + "loss": 0.2737, + "step": 17948 + }, + { + "epoch": 3.26, + "learning_rate": 2.0226246814706953e-06, + "loss": 0.8893, + "step": 17949 + }, + { + "epoch": 3.26, + "learning_rate": 2.0225700764470333e-06, + "loss": 0.2762, + "step": 17950 + }, + { + "epoch": 3.26, + "learning_rate": 2.022515471423371e-06, + "loss": 0.7716, + "step": 17951 + }, + { + "epoch": 3.26, + "learning_rate": 2.0224608663997087e-06, + "loss": 1.6554, + "step": 17952 + }, + { + "epoch": 3.26, + "learning_rate": 2.0224062613760467e-06, + "loss": 0.0138, + "step": 17953 + }, + { + "epoch": 3.26, + "learning_rate": 2.022351656352384e-06, + "loss": 2.4646, + "step": 17954 + }, + { + "epoch": 3.27, + "learning_rate": 2.022297051328722e-06, + "loss": 1.8834, + "step": 17955 + }, + { + "epoch": 3.27, + "learning_rate": 2.02224244630506e-06, + "loss": 0.0124, + "step": 17956 + }, + { + "epoch": 3.27, + "learning_rate": 2.022187841281398e-06, + "loss": 1.6586, + "step": 17957 + }, + { + "epoch": 3.27, + "learning_rate": 2.022133236257736e-06, + "loss": 0.563, + "step": 17958 + }, + { + "epoch": 3.27, + "learning_rate": 2.0220786312340735e-06, + "loss": 1.5567, + "step": 17959 + }, + { + "epoch": 3.27, + "learning_rate": 2.0220240262104114e-06, + "loss": 1.3202, + "step": 17960 + }, + { + "epoch": 3.27, + "learning_rate": 2.0219694211867494e-06, + "loss": 0.0199, + "step": 17961 + }, + { + "epoch": 3.27, + "learning_rate": 2.021914816163087e-06, + "loss": 0.9583, + "step": 17962 + }, + { + "epoch": 3.27, + "learning_rate": 2.021860211139425e-06, + "loss": 1.0449, + "step": 17963 + }, + { + "epoch": 3.27, + "learning_rate": 2.021805606115763e-06, + "loss": 1.0352, + "step": 17964 + }, + { + "epoch": 3.27, + "learning_rate": 2.0217510010921003e-06, + "loss": 0.2335, + "step": 17965 + }, + { + "epoch": 3.27, + "learning_rate": 2.0216963960684383e-06, + "loss": 2.4539, + "step": 17966 + }, + { + "epoch": 3.27, + "learning_rate": 2.021641791044776e-06, + "loss": 0.2915, + "step": 17967 + }, + { + "epoch": 3.27, + "learning_rate": 2.0215871860211137e-06, + "loss": 2.7042, + "step": 17968 + }, + { + "epoch": 3.27, + "learning_rate": 2.0215325809974517e-06, + "loss": 0.6253, + "step": 17969 + }, + { + "epoch": 3.27, + "learning_rate": 2.0214779759737896e-06, + "loss": 0.1713, + "step": 17970 + }, + { + "epoch": 3.27, + "learning_rate": 2.0214233709501276e-06, + "loss": 0.9544, + "step": 17971 + }, + { + "epoch": 3.27, + "learning_rate": 2.0213687659264655e-06, + "loss": 0.6425, + "step": 17972 + }, + { + "epoch": 3.27, + "learning_rate": 2.021314160902803e-06, + "loss": 0.6718, + "step": 17973 + }, + { + "epoch": 3.27, + "learning_rate": 2.021259555879141e-06, + "loss": 0.3097, + "step": 17974 + }, + { + "epoch": 3.27, + "learning_rate": 2.021204950855479e-06, + "loss": 0.0322, + "step": 17975 + }, + { + "epoch": 3.27, + "learning_rate": 2.0211503458318164e-06, + "loss": 0.4027, + "step": 17976 + }, + { + "epoch": 3.27, + "learning_rate": 2.0210957408081544e-06, + "loss": 0.226, + "step": 17977 + }, + { + "epoch": 3.27, + "learning_rate": 2.0210411357844923e-06, + "loss": 1.0565, + "step": 17978 + }, + { + "epoch": 3.27, + "learning_rate": 2.02098653076083e-06, + "loss": 0.279, + "step": 17979 + }, + { + "epoch": 3.27, + "learning_rate": 2.0209319257371678e-06, + "loss": 1.8385, + "step": 17980 + }, + { + "epoch": 3.27, + "learning_rate": 2.0208773207135057e-06, + "loss": 1.251, + "step": 17981 + }, + { + "epoch": 3.27, + "learning_rate": 2.0208227156898437e-06, + "loss": 0.2817, + "step": 17982 + }, + { + "epoch": 3.27, + "learning_rate": 2.0207681106661816e-06, + "loss": 1.5602, + "step": 17983 + }, + { + "epoch": 3.27, + "learning_rate": 2.020713505642519e-06, + "loss": 0.4779, + "step": 17984 + }, + { + "epoch": 3.27, + "learning_rate": 2.020658900618857e-06, + "loss": 0.0499, + "step": 17985 + }, + { + "epoch": 3.27, + "learning_rate": 2.020604295595195e-06, + "loss": 0.0601, + "step": 17986 + }, + { + "epoch": 3.27, + "learning_rate": 2.0205496905715325e-06, + "loss": 0.9268, + "step": 17987 + }, + { + "epoch": 3.27, + "learning_rate": 2.0204950855478705e-06, + "loss": 2.6143, + "step": 17988 + }, + { + "epoch": 3.27, + "learning_rate": 2.0204404805242084e-06, + "loss": 0.0287, + "step": 17989 + }, + { + "epoch": 3.27, + "learning_rate": 2.020385875500546e-06, + "loss": 1.6284, + "step": 17990 + }, + { + "epoch": 3.27, + "learning_rate": 2.020331270476884e-06, + "loss": 1.0446, + "step": 17991 + }, + { + "epoch": 3.27, + "learning_rate": 2.020276665453222e-06, + "loss": 0.9134, + "step": 17992 + }, + { + "epoch": 3.27, + "learning_rate": 2.0202220604295593e-06, + "loss": 0.0419, + "step": 17993 + }, + { + "epoch": 3.27, + "learning_rate": 2.0201674554058973e-06, + "loss": 1.9876, + "step": 17994 + }, + { + "epoch": 3.27, + "learning_rate": 2.0201128503822352e-06, + "loss": 0.0292, + "step": 17995 + }, + { + "epoch": 3.27, + "learning_rate": 2.020058245358573e-06, + "loss": 0.1656, + "step": 17996 + }, + { + "epoch": 3.27, + "learning_rate": 2.020003640334911e-06, + "loss": 0.2847, + "step": 17997 + }, + { + "epoch": 3.27, + "learning_rate": 2.0199490353112486e-06, + "loss": 0.5068, + "step": 17998 + }, + { + "epoch": 3.27, + "learning_rate": 2.0198944302875866e-06, + "loss": 0.5879, + "step": 17999 + }, + { + "epoch": 3.27, + "learning_rate": 2.0198398252639245e-06, + "loss": 1.1183, + "step": 18000 + }, + { + "epoch": 3.27, + "learning_rate": 2.019785220240262e-06, + "loss": 0.6155, + "step": 18001 + }, + { + "epoch": 3.27, + "learning_rate": 2.0197306152166e-06, + "loss": 1.39, + "step": 18002 + }, + { + "epoch": 3.27, + "learning_rate": 2.019676010192938e-06, + "loss": 0.6793, + "step": 18003 + }, + { + "epoch": 3.27, + "learning_rate": 2.0196214051692754e-06, + "loss": 0.3773, + "step": 18004 + }, + { + "epoch": 3.27, + "learning_rate": 2.0195668001456134e-06, + "loss": 0.2134, + "step": 18005 + }, + { + "epoch": 3.27, + "learning_rate": 2.0195121951219513e-06, + "loss": 2.76, + "step": 18006 + }, + { + "epoch": 3.27, + "learning_rate": 2.019457590098289e-06, + "loss": 1.9686, + "step": 18007 + }, + { + "epoch": 3.27, + "learning_rate": 2.019402985074627e-06, + "loss": 0.4446, + "step": 18008 + }, + { + "epoch": 3.27, + "learning_rate": 2.0193483800509647e-06, + "loss": 0.3954, + "step": 18009 + }, + { + "epoch": 3.28, + "learning_rate": 2.0192937750273027e-06, + "loss": 1.5705, + "step": 18010 + }, + { + "epoch": 3.28, + "learning_rate": 2.0192391700036406e-06, + "loss": 0.5092, + "step": 18011 + }, + { + "epoch": 3.28, + "learning_rate": 2.019184564979978e-06, + "loss": 0.1407, + "step": 18012 + }, + { + "epoch": 3.28, + "learning_rate": 2.019129959956316e-06, + "loss": 1.7123, + "step": 18013 + }, + { + "epoch": 3.28, + "learning_rate": 2.019075354932654e-06, + "loss": 1.6831, + "step": 18014 + }, + { + "epoch": 3.28, + "learning_rate": 2.0190207499089916e-06, + "loss": 1.5926, + "step": 18015 + }, + { + "epoch": 3.28, + "learning_rate": 2.0189661448853295e-06, + "loss": 0.443, + "step": 18016 + }, + { + "epoch": 3.28, + "learning_rate": 2.0189115398616674e-06, + "loss": 0.5628, + "step": 18017 + }, + { + "epoch": 3.28, + "learning_rate": 2.018856934838005e-06, + "loss": 0.2233, + "step": 18018 + }, + { + "epoch": 3.28, + "learning_rate": 2.018802329814343e-06, + "loss": 0.0199, + "step": 18019 + }, + { + "epoch": 3.28, + "learning_rate": 2.018747724790681e-06, + "loss": 0.4487, + "step": 18020 + }, + { + "epoch": 3.28, + "learning_rate": 2.0186931197670184e-06, + "loss": 0.5391, + "step": 18021 + }, + { + "epoch": 3.28, + "learning_rate": 2.0186385147433563e-06, + "loss": 0.9592, + "step": 18022 + }, + { + "epoch": 3.28, + "learning_rate": 2.0185839097196943e-06, + "loss": 0.4579, + "step": 18023 + }, + { + "epoch": 3.28, + "learning_rate": 2.018529304696032e-06, + "loss": 0.0179, + "step": 18024 + }, + { + "epoch": 3.28, + "learning_rate": 2.01847469967237e-06, + "loss": 0.2944, + "step": 18025 + }, + { + "epoch": 3.28, + "learning_rate": 2.0184200946487077e-06, + "loss": 2.4584, + "step": 18026 + }, + { + "epoch": 3.28, + "learning_rate": 2.0183654896250456e-06, + "loss": 0.041, + "step": 18027 + }, + { + "epoch": 3.28, + "learning_rate": 2.0183108846013835e-06, + "loss": 1.6569, + "step": 18028 + }, + { + "epoch": 3.28, + "learning_rate": 2.018256279577721e-06, + "loss": 0.8183, + "step": 18029 + }, + { + "epoch": 3.28, + "learning_rate": 2.018201674554059e-06, + "loss": 1.0294, + "step": 18030 + }, + { + "epoch": 3.28, + "learning_rate": 2.018147069530397e-06, + "loss": 0.8037, + "step": 18031 + }, + { + "epoch": 3.28, + "learning_rate": 2.0180924645067345e-06, + "loss": 1.7752, + "step": 18032 + }, + { + "epoch": 3.28, + "learning_rate": 2.0180378594830724e-06, + "loss": 0.4897, + "step": 18033 + }, + { + "epoch": 3.28, + "learning_rate": 2.0179832544594104e-06, + "loss": 0.0103, + "step": 18034 + }, + { + "epoch": 3.28, + "learning_rate": 2.017928649435748e-06, + "loss": 0.706, + "step": 18035 + }, + { + "epoch": 3.28, + "learning_rate": 2.017874044412086e-06, + "loss": 1.36, + "step": 18036 + }, + { + "epoch": 3.28, + "learning_rate": 2.0178194393884238e-06, + "loss": 0.434, + "step": 18037 + }, + { + "epoch": 3.28, + "learning_rate": 2.0177648343647617e-06, + "loss": 0.7425, + "step": 18038 + }, + { + "epoch": 3.28, + "learning_rate": 2.0177102293410997e-06, + "loss": 0.1764, + "step": 18039 + }, + { + "epoch": 3.28, + "learning_rate": 2.017655624317437e-06, + "loss": 1.3885, + "step": 18040 + }, + { + "epoch": 3.28, + "learning_rate": 2.017601019293775e-06, + "loss": 0.7363, + "step": 18041 + }, + { + "epoch": 3.28, + "learning_rate": 2.017546414270113e-06, + "loss": 0.2046, + "step": 18042 + }, + { + "epoch": 3.28, + "learning_rate": 2.0174918092464506e-06, + "loss": 0.2321, + "step": 18043 + }, + { + "epoch": 3.28, + "learning_rate": 2.0174372042227885e-06, + "loss": 0.2547, + "step": 18044 + }, + { + "epoch": 3.28, + "learning_rate": 2.0173825991991265e-06, + "loss": 0.0432, + "step": 18045 + }, + { + "epoch": 3.28, + "learning_rate": 2.017327994175464e-06, + "loss": 0.1865, + "step": 18046 + }, + { + "epoch": 3.28, + "learning_rate": 2.017273389151802e-06, + "loss": 1.1249, + "step": 18047 + }, + { + "epoch": 3.28, + "learning_rate": 2.01721878412814e-06, + "loss": 0.5931, + "step": 18048 + }, + { + "epoch": 3.28, + "learning_rate": 2.0171641791044774e-06, + "loss": 0.9615, + "step": 18049 + }, + { + "epoch": 3.28, + "learning_rate": 2.0171095740808153e-06, + "loss": 0.6585, + "step": 18050 + }, + { + "epoch": 3.28, + "learning_rate": 2.0170549690571533e-06, + "loss": 0.2969, + "step": 18051 + }, + { + "epoch": 3.28, + "learning_rate": 2.0170003640334912e-06, + "loss": 1.3393, + "step": 18052 + }, + { + "epoch": 3.28, + "learning_rate": 2.016945759009829e-06, + "loss": 0.0206, + "step": 18053 + }, + { + "epoch": 3.28, + "learning_rate": 2.0168911539861667e-06, + "loss": 0.8995, + "step": 18054 + }, + { + "epoch": 3.28, + "learning_rate": 2.0168365489625046e-06, + "loss": 1.5009, + "step": 18055 + }, + { + "epoch": 3.28, + "learning_rate": 2.0167819439388426e-06, + "loss": 0.0415, + "step": 18056 + }, + { + "epoch": 3.28, + "learning_rate": 2.01672733891518e-06, + "loss": 0.6639, + "step": 18057 + }, + { + "epoch": 3.28, + "learning_rate": 2.016672733891518e-06, + "loss": 0.0102, + "step": 18058 + }, + { + "epoch": 3.28, + "learning_rate": 2.016618128867856e-06, + "loss": 0.4431, + "step": 18059 + }, + { + "epoch": 3.28, + "learning_rate": 2.0165635238441935e-06, + "loss": 0.0185, + "step": 18060 + }, + { + "epoch": 3.28, + "learning_rate": 2.0165089188205314e-06, + "loss": 4.6449, + "step": 18061 + }, + { + "epoch": 3.28, + "learning_rate": 2.0164543137968694e-06, + "loss": 1.0776, + "step": 18062 + }, + { + "epoch": 3.28, + "learning_rate": 2.016399708773207e-06, + "loss": 0.4874, + "step": 18063 + }, + { + "epoch": 3.28, + "learning_rate": 2.016345103749545e-06, + "loss": 1.3889, + "step": 18064 + }, + { + "epoch": 3.29, + "learning_rate": 2.016290498725883e-06, + "loss": 0.0354, + "step": 18065 + }, + { + "epoch": 3.29, + "learning_rate": 2.0162358937022207e-06, + "loss": 1.0938, + "step": 18066 + }, + { + "epoch": 3.29, + "learning_rate": 2.0161812886785587e-06, + "loss": 2.0102, + "step": 18067 + }, + { + "epoch": 3.29, + "learning_rate": 2.016126683654896e-06, + "loss": 0.0866, + "step": 18068 + }, + { + "epoch": 3.29, + "learning_rate": 2.016072078631234e-06, + "loss": 0.5668, + "step": 18069 + }, + { + "epoch": 3.29, + "learning_rate": 2.016017473607572e-06, + "loss": 1.9906, + "step": 18070 + }, + { + "epoch": 3.29, + "learning_rate": 2.0159628685839096e-06, + "loss": 0.2767, + "step": 18071 + }, + { + "epoch": 3.29, + "learning_rate": 2.0159082635602476e-06, + "loss": 0.0183, + "step": 18072 + }, + { + "epoch": 3.29, + "learning_rate": 2.0158536585365855e-06, + "loss": 0.3541, + "step": 18073 + }, + { + "epoch": 3.29, + "learning_rate": 2.015799053512923e-06, + "loss": 1.6961, + "step": 18074 + }, + { + "epoch": 3.29, + "learning_rate": 2.015744448489261e-06, + "loss": 0.8177, + "step": 18075 + }, + { + "epoch": 3.29, + "learning_rate": 2.015689843465599e-06, + "loss": 0.6189, + "step": 18076 + }, + { + "epoch": 3.29, + "learning_rate": 2.015635238441937e-06, + "loss": 1.2544, + "step": 18077 + }, + { + "epoch": 3.29, + "learning_rate": 2.015580633418275e-06, + "loss": 0.5855, + "step": 18078 + }, + { + "epoch": 3.29, + "learning_rate": 2.0155260283946123e-06, + "loss": 0.9884, + "step": 18079 + }, + { + "epoch": 3.29, + "learning_rate": 2.0154714233709503e-06, + "loss": 0.2052, + "step": 18080 + }, + { + "epoch": 3.29, + "learning_rate": 2.015416818347288e-06, + "loss": 0.917, + "step": 18081 + }, + { + "epoch": 3.29, + "learning_rate": 2.0153622133236257e-06, + "loss": 0.1976, + "step": 18082 + }, + { + "epoch": 3.29, + "learning_rate": 2.0153076082999637e-06, + "loss": 0.1595, + "step": 18083 + }, + { + "epoch": 3.29, + "learning_rate": 2.0152530032763016e-06, + "loss": 1.6905, + "step": 18084 + }, + { + "epoch": 3.29, + "learning_rate": 2.015198398252639e-06, + "loss": 0.3265, + "step": 18085 + }, + { + "epoch": 3.29, + "learning_rate": 2.015143793228977e-06, + "loss": 0.0356, + "step": 18086 + }, + { + "epoch": 3.29, + "learning_rate": 2.015089188205315e-06, + "loss": 0.8251, + "step": 18087 + }, + { + "epoch": 3.29, + "learning_rate": 2.0150345831816525e-06, + "loss": 1.8725, + "step": 18088 + }, + { + "epoch": 3.29, + "learning_rate": 2.0149799781579905e-06, + "loss": 0.7581, + "step": 18089 + }, + { + "epoch": 3.29, + "learning_rate": 2.0149253731343284e-06, + "loss": 0.3101, + "step": 18090 + }, + { + "epoch": 3.29, + "learning_rate": 2.0148707681106664e-06, + "loss": 0.0424, + "step": 18091 + }, + { + "epoch": 3.29, + "learning_rate": 2.0148161630870043e-06, + "loss": 0.1348, + "step": 18092 + }, + { + "epoch": 3.29, + "learning_rate": 2.014761558063342e-06, + "loss": 0.2678, + "step": 18093 + }, + { + "epoch": 3.29, + "learning_rate": 2.0147069530396798e-06, + "loss": 1.0838, + "step": 18094 + }, + { + "epoch": 3.29, + "learning_rate": 2.0146523480160177e-06, + "loss": 1.9167, + "step": 18095 + }, + { + "epoch": 3.29, + "learning_rate": 2.0145977429923552e-06, + "loss": 0.7468, + "step": 18096 + }, + { + "epoch": 3.29, + "learning_rate": 2.014543137968693e-06, + "loss": 0.1324, + "step": 18097 + }, + { + "epoch": 3.29, + "learning_rate": 2.014488532945031e-06, + "loss": 0.0339, + "step": 18098 + }, + { + "epoch": 3.29, + "learning_rate": 2.0144339279213686e-06, + "loss": 0.8617, + "step": 18099 + }, + { + "epoch": 3.29, + "learning_rate": 2.0143793228977066e-06, + "loss": 0.0101, + "step": 18100 + }, + { + "epoch": 3.29, + "learning_rate": 2.0143247178740445e-06, + "loss": 1.401, + "step": 18101 + }, + { + "epoch": 3.29, + "learning_rate": 2.014270112850382e-06, + "loss": 0.509, + "step": 18102 + }, + { + "epoch": 3.29, + "learning_rate": 2.01421550782672e-06, + "loss": 1.8677, + "step": 18103 + }, + { + "epoch": 3.29, + "learning_rate": 2.014160902803058e-06, + "loss": 0.9617, + "step": 18104 + }, + { + "epoch": 3.29, + "learning_rate": 2.014106297779396e-06, + "loss": 0.4692, + "step": 18105 + }, + { + "epoch": 3.29, + "learning_rate": 2.014051692755734e-06, + "loss": 0.2646, + "step": 18106 + }, + { + "epoch": 3.29, + "learning_rate": 2.0139970877320713e-06, + "loss": 0.2677, + "step": 18107 + }, + { + "epoch": 3.29, + "learning_rate": 2.0139424827084093e-06, + "loss": 0.7224, + "step": 18108 + }, + { + "epoch": 3.29, + "learning_rate": 2.0138878776847472e-06, + "loss": 1.138, + "step": 18109 + }, + { + "epoch": 3.29, + "learning_rate": 2.0138332726610847e-06, + "loss": 1.3136, + "step": 18110 + }, + { + "epoch": 3.29, + "learning_rate": 2.0137786676374227e-06, + "loss": 2.1273, + "step": 18111 + }, + { + "epoch": 3.29, + "learning_rate": 2.0137240626137606e-06, + "loss": 0.4902, + "step": 18112 + }, + { + "epoch": 3.29, + "learning_rate": 2.013669457590098e-06, + "loss": 1.3718, + "step": 18113 + }, + { + "epoch": 3.29, + "learning_rate": 2.013614852566436e-06, + "loss": 1.9838, + "step": 18114 + }, + { + "epoch": 3.29, + "learning_rate": 2.013560247542774e-06, + "loss": 1.0962, + "step": 18115 + }, + { + "epoch": 3.29, + "learning_rate": 2.0135056425191116e-06, + "loss": 0.4548, + "step": 18116 + }, + { + "epoch": 3.29, + "learning_rate": 2.0134510374954495e-06, + "loss": 0.908, + "step": 18117 + }, + { + "epoch": 3.29, + "learning_rate": 2.0133964324717874e-06, + "loss": 0.4414, + "step": 18118 + }, + { + "epoch": 3.29, + "learning_rate": 2.0133418274481254e-06, + "loss": 0.734, + "step": 18119 + }, + { + "epoch": 3.3, + "learning_rate": 2.0132872224244633e-06, + "loss": 0.4965, + "step": 18120 + }, + { + "epoch": 3.3, + "learning_rate": 2.013232617400801e-06, + "loss": 0.3529, + "step": 18121 + }, + { + "epoch": 3.3, + "learning_rate": 2.013178012377139e-06, + "loss": 1.4505, + "step": 18122 + }, + { + "epoch": 3.3, + "learning_rate": 2.0131234073534767e-06, + "loss": 0.4806, + "step": 18123 + }, + { + "epoch": 3.3, + "learning_rate": 2.0130688023298143e-06, + "loss": 0.7631, + "step": 18124 + }, + { + "epoch": 3.3, + "learning_rate": 2.013014197306152e-06, + "loss": 0.0134, + "step": 18125 + }, + { + "epoch": 3.3, + "learning_rate": 2.01295959228249e-06, + "loss": 0.8251, + "step": 18126 + }, + { + "epoch": 3.3, + "learning_rate": 2.0129049872588277e-06, + "loss": 0.2758, + "step": 18127 + }, + { + "epoch": 3.3, + "learning_rate": 2.0128503822351656e-06, + "loss": 0.3141, + "step": 18128 + }, + { + "epoch": 3.3, + "learning_rate": 2.0127957772115035e-06, + "loss": 1.3159, + "step": 18129 + }, + { + "epoch": 3.3, + "learning_rate": 2.012741172187841e-06, + "loss": 0.1335, + "step": 18130 + }, + { + "epoch": 3.3, + "learning_rate": 2.012686567164179e-06, + "loss": 1.1059, + "step": 18131 + }, + { + "epoch": 3.3, + "learning_rate": 2.012631962140517e-06, + "loss": 0.0187, + "step": 18132 + }, + { + "epoch": 3.3, + "learning_rate": 2.012577357116855e-06, + "loss": 0.9426, + "step": 18133 + }, + { + "epoch": 3.3, + "learning_rate": 2.012522752093193e-06, + "loss": 0.7854, + "step": 18134 + }, + { + "epoch": 3.3, + "learning_rate": 2.0124681470695304e-06, + "loss": 0.6786, + "step": 18135 + }, + { + "epoch": 3.3, + "learning_rate": 2.0124135420458683e-06, + "loss": 0.2352, + "step": 18136 + }, + { + "epoch": 3.3, + "learning_rate": 2.0123589370222062e-06, + "loss": 0.5817, + "step": 18137 + }, + { + "epoch": 3.3, + "learning_rate": 2.0123043319985438e-06, + "loss": 1.4596, + "step": 18138 + }, + { + "epoch": 3.3, + "learning_rate": 2.0122497269748817e-06, + "loss": 0.8664, + "step": 18139 + }, + { + "epoch": 3.3, + "learning_rate": 2.0121951219512197e-06, + "loss": 1.986, + "step": 18140 + }, + { + "epoch": 3.3, + "learning_rate": 2.012140516927557e-06, + "loss": 0.0331, + "step": 18141 + }, + { + "epoch": 3.3, + "learning_rate": 2.012085911903895e-06, + "loss": 0.7318, + "step": 18142 + }, + { + "epoch": 3.3, + "learning_rate": 2.012031306880233e-06, + "loss": 0.0406, + "step": 18143 + }, + { + "epoch": 3.3, + "learning_rate": 2.0119767018565706e-06, + "loss": 0.2426, + "step": 18144 + }, + { + "epoch": 3.3, + "learning_rate": 2.0119220968329085e-06, + "loss": 2.4965, + "step": 18145 + }, + { + "epoch": 3.3, + "learning_rate": 2.0118674918092465e-06, + "loss": 0.7289, + "step": 18146 + }, + { + "epoch": 3.3, + "learning_rate": 2.0118128867855844e-06, + "loss": 2.4344, + "step": 18147 + }, + { + "epoch": 3.3, + "learning_rate": 2.0117582817619224e-06, + "loss": 0.5364, + "step": 18148 + }, + { + "epoch": 3.3, + "learning_rate": 2.01170367673826e-06, + "loss": 0.4565, + "step": 18149 + }, + { + "epoch": 3.3, + "learning_rate": 2.011649071714598e-06, + "loss": 1.1228, + "step": 18150 + }, + { + "epoch": 3.3, + "learning_rate": 2.0115944666909358e-06, + "loss": 0.8895, + "step": 18151 + }, + { + "epoch": 3.3, + "learning_rate": 2.0115398616672733e-06, + "loss": 1.0171, + "step": 18152 + }, + { + "epoch": 3.3, + "learning_rate": 2.0114852566436112e-06, + "loss": 0.0483, + "step": 18153 + }, + { + "epoch": 3.3, + "learning_rate": 2.011430651619949e-06, + "loss": 0.2191, + "step": 18154 + }, + { + "epoch": 3.3, + "learning_rate": 2.0113760465962867e-06, + "loss": 0.813, + "step": 18155 + }, + { + "epoch": 3.3, + "learning_rate": 2.0113214415726246e-06, + "loss": 1.5022, + "step": 18156 + }, + { + "epoch": 3.3, + "learning_rate": 2.0112668365489626e-06, + "loss": 1.3407, + "step": 18157 + }, + { + "epoch": 3.3, + "learning_rate": 2.0112122315253005e-06, + "loss": 3.353, + "step": 18158 + }, + { + "epoch": 3.3, + "learning_rate": 2.0111576265016385e-06, + "loss": 0.3232, + "step": 18159 + }, + { + "epoch": 3.3, + "learning_rate": 2.011103021477976e-06, + "loss": 0.897, + "step": 18160 + }, + { + "epoch": 3.3, + "learning_rate": 2.011048416454314e-06, + "loss": 2.5112, + "step": 18161 + }, + { + "epoch": 3.3, + "learning_rate": 2.010993811430652e-06, + "loss": 1.752, + "step": 18162 + }, + { + "epoch": 3.3, + "learning_rate": 2.0109392064069894e-06, + "loss": 0.2383, + "step": 18163 + }, + { + "epoch": 3.3, + "learning_rate": 2.0108846013833273e-06, + "loss": 0.4957, + "step": 18164 + }, + { + "epoch": 3.3, + "learning_rate": 2.0108299963596653e-06, + "loss": 0.5031, + "step": 18165 + }, + { + "epoch": 3.3, + "learning_rate": 2.010775391336003e-06, + "loss": 0.1345, + "step": 18166 + }, + { + "epoch": 3.3, + "learning_rate": 2.0107207863123407e-06, + "loss": 0.4108, + "step": 18167 + }, + { + "epoch": 3.3, + "learning_rate": 2.0106661812886787e-06, + "loss": 1.6615, + "step": 18168 + }, + { + "epoch": 3.3, + "learning_rate": 2.010611576265016e-06, + "loss": 1.5089, + "step": 18169 + }, + { + "epoch": 3.3, + "learning_rate": 2.010556971241354e-06, + "loss": 0.3948, + "step": 18170 + }, + { + "epoch": 3.3, + "learning_rate": 2.010502366217692e-06, + "loss": 0.6031, + "step": 18171 + }, + { + "epoch": 3.3, + "learning_rate": 2.01044776119403e-06, + "loss": 1.1787, + "step": 18172 + }, + { + "epoch": 3.3, + "learning_rate": 2.010393156170368e-06, + "loss": 0.3276, + "step": 18173 + }, + { + "epoch": 3.3, + "learning_rate": 2.0103385511467055e-06, + "loss": 0.3921, + "step": 18174 + }, + { + "epoch": 3.31, + "learning_rate": 2.0102839461230434e-06, + "loss": 1.1348, + "step": 18175 + }, + { + "epoch": 3.31, + "learning_rate": 2.0102293410993814e-06, + "loss": 1.5658, + "step": 18176 + }, + { + "epoch": 3.31, + "learning_rate": 2.010174736075719e-06, + "loss": 0.2831, + "step": 18177 + }, + { + "epoch": 3.31, + "learning_rate": 2.010120131052057e-06, + "loss": 1.1124, + "step": 18178 + }, + { + "epoch": 3.31, + "learning_rate": 2.010065526028395e-06, + "loss": 1.7922, + "step": 18179 + }, + { + "epoch": 3.31, + "learning_rate": 2.0100109210047323e-06, + "loss": 0.2949, + "step": 18180 + }, + { + "epoch": 3.31, + "learning_rate": 2.0099563159810703e-06, + "loss": 0.08, + "step": 18181 + }, + { + "epoch": 3.31, + "learning_rate": 2.009901710957408e-06, + "loss": 0.5641, + "step": 18182 + }, + { + "epoch": 3.31, + "learning_rate": 2.0098471059337457e-06, + "loss": 3.4223, + "step": 18183 + }, + { + "epoch": 3.31, + "learning_rate": 2.0097925009100837e-06, + "loss": 2.9497, + "step": 18184 + }, + { + "epoch": 3.31, + "learning_rate": 2.0097378958864216e-06, + "loss": 0.4875, + "step": 18185 + }, + { + "epoch": 3.31, + "learning_rate": 2.0096832908627595e-06, + "loss": 0.0088, + "step": 18186 + }, + { + "epoch": 3.31, + "learning_rate": 2.0096286858390975e-06, + "loss": 2.3796, + "step": 18187 + }, + { + "epoch": 3.31, + "learning_rate": 2.009574080815435e-06, + "loss": 0.4309, + "step": 18188 + }, + { + "epoch": 3.31, + "learning_rate": 2.009519475791773e-06, + "loss": 0.5026, + "step": 18189 + }, + { + "epoch": 3.31, + "learning_rate": 2.009464870768111e-06, + "loss": 3.4841, + "step": 18190 + }, + { + "epoch": 3.31, + "learning_rate": 2.0094102657444484e-06, + "loss": 0.3574, + "step": 18191 + }, + { + "epoch": 3.31, + "learning_rate": 2.0093556607207864e-06, + "loss": 0.5411, + "step": 18192 + }, + { + "epoch": 3.31, + "learning_rate": 2.0093010556971243e-06, + "loss": 1.6196, + "step": 18193 + }, + { + "epoch": 3.31, + "learning_rate": 2.009246450673462e-06, + "loss": 0.5747, + "step": 18194 + }, + { + "epoch": 3.31, + "learning_rate": 2.0091918456497998e-06, + "loss": 0.9563, + "step": 18195 + }, + { + "epoch": 3.31, + "learning_rate": 2.0091372406261377e-06, + "loss": 0.6178, + "step": 18196 + }, + { + "epoch": 3.31, + "learning_rate": 2.0090826356024752e-06, + "loss": 2.1092, + "step": 18197 + }, + { + "epoch": 3.31, + "learning_rate": 2.009028030578813e-06, + "loss": 1.4624, + "step": 18198 + }, + { + "epoch": 3.31, + "learning_rate": 2.008973425555151e-06, + "loss": 0.0656, + "step": 18199 + }, + { + "epoch": 3.31, + "learning_rate": 2.008918820531489e-06, + "loss": 0.4693, + "step": 18200 + }, + { + "epoch": 3.31, + "learning_rate": 2.008864215507827e-06, + "loss": 0.1644, + "step": 18201 + }, + { + "epoch": 3.31, + "learning_rate": 2.0088096104841645e-06, + "loss": 1.3737, + "step": 18202 + }, + { + "epoch": 3.31, + "learning_rate": 2.0087550054605025e-06, + "loss": 0.6391, + "step": 18203 + }, + { + "epoch": 3.31, + "learning_rate": 2.0087004004368404e-06, + "loss": 0.2521, + "step": 18204 + }, + { + "epoch": 3.31, + "learning_rate": 2.008645795413178e-06, + "loss": 1.7179, + "step": 18205 + }, + { + "epoch": 3.31, + "learning_rate": 2.008591190389516e-06, + "loss": 0.1234, + "step": 18206 + }, + { + "epoch": 3.31, + "learning_rate": 2.008536585365854e-06, + "loss": 0.2774, + "step": 18207 + }, + { + "epoch": 3.31, + "learning_rate": 2.0084819803421913e-06, + "loss": 4.3104, + "step": 18208 + }, + { + "epoch": 3.31, + "learning_rate": 2.0084273753185293e-06, + "loss": 0.6892, + "step": 18209 + }, + { + "epoch": 3.31, + "learning_rate": 2.0083727702948672e-06, + "loss": 0.9899, + "step": 18210 + }, + { + "epoch": 3.31, + "learning_rate": 2.0083181652712047e-06, + "loss": 0.0141, + "step": 18211 + }, + { + "epoch": 3.31, + "learning_rate": 2.0082635602475427e-06, + "loss": 1.6916, + "step": 18212 + }, + { + "epoch": 3.31, + "learning_rate": 2.0082089552238806e-06, + "loss": 0.0109, + "step": 18213 + }, + { + "epoch": 3.31, + "learning_rate": 2.0081543502002186e-06, + "loss": 1.0219, + "step": 18214 + }, + { + "epoch": 3.31, + "learning_rate": 2.0080997451765565e-06, + "loss": 0.5001, + "step": 18215 + }, + { + "epoch": 3.31, + "learning_rate": 2.008045140152894e-06, + "loss": 0.4019, + "step": 18216 + }, + { + "epoch": 3.31, + "learning_rate": 2.007990535129232e-06, + "loss": 1.0442, + "step": 18217 + }, + { + "epoch": 3.31, + "learning_rate": 2.00793593010557e-06, + "loss": 2.9347, + "step": 18218 + }, + { + "epoch": 3.31, + "learning_rate": 2.0078813250819074e-06, + "loss": 0.4645, + "step": 18219 + }, + { + "epoch": 3.31, + "learning_rate": 2.0078267200582454e-06, + "loss": 1.9865, + "step": 18220 + }, + { + "epoch": 3.31, + "learning_rate": 2.0077721150345833e-06, + "loss": 0.0261, + "step": 18221 + }, + { + "epoch": 3.31, + "learning_rate": 2.007717510010921e-06, + "loss": 2.0005, + "step": 18222 + }, + { + "epoch": 3.31, + "learning_rate": 2.007662904987259e-06, + "loss": 0.9291, + "step": 18223 + }, + { + "epoch": 3.31, + "learning_rate": 2.0076082999635967e-06, + "loss": 0.2105, + "step": 18224 + }, + { + "epoch": 3.31, + "learning_rate": 2.0075536949399343e-06, + "loss": 0.4175, + "step": 18225 + }, + { + "epoch": 3.31, + "learning_rate": 2.007499089916272e-06, + "loss": 0.0154, + "step": 18226 + }, + { + "epoch": 3.31, + "learning_rate": 2.00744448489261e-06, + "loss": 0.5419, + "step": 18227 + }, + { + "epoch": 3.31, + "learning_rate": 2.007389879868948e-06, + "loss": 0.3291, + "step": 18228 + }, + { + "epoch": 3.31, + "learning_rate": 2.007335274845286e-06, + "loss": 0.2308, + "step": 18229 + }, + { + "epoch": 3.32, + "learning_rate": 2.0072806698216236e-06, + "loss": 0.0998, + "step": 18230 + }, + { + "epoch": 3.32, + "learning_rate": 2.0072260647979615e-06, + "loss": 0.2217, + "step": 18231 + }, + { + "epoch": 3.32, + "learning_rate": 2.0071714597742994e-06, + "loss": 0.2609, + "step": 18232 + }, + { + "epoch": 3.32, + "learning_rate": 2.007116854750637e-06, + "loss": 0.6309, + "step": 18233 + }, + { + "epoch": 3.32, + "learning_rate": 2.007062249726975e-06, + "loss": 0.2202, + "step": 18234 + }, + { + "epoch": 3.32, + "learning_rate": 2.007007644703313e-06, + "loss": 1.3119, + "step": 18235 + }, + { + "epoch": 3.32, + "learning_rate": 2.0069530396796504e-06, + "loss": 1.2167, + "step": 18236 + }, + { + "epoch": 3.32, + "learning_rate": 2.0068984346559883e-06, + "loss": 1.07, + "step": 18237 + }, + { + "epoch": 3.32, + "learning_rate": 2.0068438296323263e-06, + "loss": 2.2097, + "step": 18238 + }, + { + "epoch": 3.32, + "learning_rate": 2.006789224608664e-06, + "loss": 1.1215, + "step": 18239 + }, + { + "epoch": 3.32, + "learning_rate": 2.006734619585002e-06, + "loss": 0.0209, + "step": 18240 + }, + { + "epoch": 3.32, + "learning_rate": 2.0066800145613397e-06, + "loss": 0.8014, + "step": 18241 + }, + { + "epoch": 3.32, + "learning_rate": 2.0066254095376776e-06, + "loss": 0.7043, + "step": 18242 + }, + { + "epoch": 3.32, + "learning_rate": 2.0065708045140155e-06, + "loss": 1.3852, + "step": 18243 + }, + { + "epoch": 3.32, + "learning_rate": 2.006516199490353e-06, + "loss": 0.4235, + "step": 18244 + }, + { + "epoch": 3.32, + "learning_rate": 2.006461594466691e-06, + "loss": 2.1568, + "step": 18245 + }, + { + "epoch": 3.32, + "learning_rate": 2.006406989443029e-06, + "loss": 0.7296, + "step": 18246 + }, + { + "epoch": 3.32, + "learning_rate": 2.0063523844193665e-06, + "loss": 0.6595, + "step": 18247 + }, + { + "epoch": 3.32, + "learning_rate": 2.0062977793957044e-06, + "loss": 0.8943, + "step": 18248 + }, + { + "epoch": 3.32, + "learning_rate": 2.0062431743720424e-06, + "loss": 0.0818, + "step": 18249 + }, + { + "epoch": 3.32, + "learning_rate": 2.00618856934838e-06, + "loss": 0.529, + "step": 18250 + }, + { + "epoch": 3.32, + "learning_rate": 2.006133964324718e-06, + "loss": 0.2187, + "step": 18251 + }, + { + "epoch": 3.32, + "learning_rate": 2.0060793593010558e-06, + "loss": 0.4534, + "step": 18252 + }, + { + "epoch": 3.32, + "learning_rate": 2.0060247542773937e-06, + "loss": 1.7362, + "step": 18253 + }, + { + "epoch": 3.32, + "learning_rate": 2.0059701492537317e-06, + "loss": 0.7524, + "step": 18254 + }, + { + "epoch": 3.32, + "learning_rate": 2.005915544230069e-06, + "loss": 1.5546, + "step": 18255 + }, + { + "epoch": 3.32, + "learning_rate": 2.005860939206407e-06, + "loss": 0.8772, + "step": 18256 + }, + { + "epoch": 3.32, + "learning_rate": 2.005806334182745e-06, + "loss": 0.4367, + "step": 18257 + }, + { + "epoch": 3.32, + "learning_rate": 2.0057517291590826e-06, + "loss": 0.0215, + "step": 18258 + }, + { + "epoch": 3.32, + "learning_rate": 2.0056971241354205e-06, + "loss": 0.8904, + "step": 18259 + }, + { + "epoch": 3.32, + "learning_rate": 2.0056425191117585e-06, + "loss": 0.3046, + "step": 18260 + }, + { + "epoch": 3.32, + "learning_rate": 2.005587914088096e-06, + "loss": 0.1802, + "step": 18261 + }, + { + "epoch": 3.32, + "learning_rate": 2.005533309064434e-06, + "loss": 1.136, + "step": 18262 + }, + { + "epoch": 3.32, + "learning_rate": 2.005478704040772e-06, + "loss": 0.5199, + "step": 18263 + }, + { + "epoch": 3.32, + "learning_rate": 2.0054240990171094e-06, + "loss": 1.9451, + "step": 18264 + }, + { + "epoch": 3.32, + "learning_rate": 2.0053694939934473e-06, + "loss": 0.3334, + "step": 18265 + }, + { + "epoch": 3.32, + "learning_rate": 2.0053148889697853e-06, + "loss": 0.871, + "step": 18266 + }, + { + "epoch": 3.32, + "learning_rate": 2.0052602839461232e-06, + "loss": 0.0186, + "step": 18267 + }, + { + "epoch": 3.32, + "learning_rate": 2.005205678922461e-06, + "loss": 0.3633, + "step": 18268 + }, + { + "epoch": 3.32, + "learning_rate": 2.0051510738987987e-06, + "loss": 0.3774, + "step": 18269 + }, + { + "epoch": 3.32, + "learning_rate": 2.0050964688751366e-06, + "loss": 0.3397, + "step": 18270 + }, + { + "epoch": 3.32, + "learning_rate": 2.0050418638514746e-06, + "loss": 0.0259, + "step": 18271 + }, + { + "epoch": 3.32, + "learning_rate": 2.004987258827812e-06, + "loss": 2.2099, + "step": 18272 + }, + { + "epoch": 3.32, + "learning_rate": 2.00493265380415e-06, + "loss": 1.0782, + "step": 18273 + }, + { + "epoch": 3.32, + "learning_rate": 2.004878048780488e-06, + "loss": 0.4529, + "step": 18274 + }, + { + "epoch": 3.32, + "learning_rate": 2.0048234437568255e-06, + "loss": 0.0205, + "step": 18275 + }, + { + "epoch": 3.32, + "learning_rate": 2.0047688387331634e-06, + "loss": 0.856, + "step": 18276 + }, + { + "epoch": 3.32, + "learning_rate": 2.0047142337095014e-06, + "loss": 1.0726, + "step": 18277 + }, + { + "epoch": 3.32, + "learning_rate": 2.004659628685839e-06, + "loss": 1.2059, + "step": 18278 + }, + { + "epoch": 3.32, + "learning_rate": 2.004605023662177e-06, + "loss": 0.0843, + "step": 18279 + }, + { + "epoch": 3.32, + "learning_rate": 2.004550418638515e-06, + "loss": 0.6949, + "step": 18280 + }, + { + "epoch": 3.32, + "learning_rate": 2.0044958136148527e-06, + "loss": 3.5807, + "step": 18281 + }, + { + "epoch": 3.32, + "learning_rate": 2.0044412085911907e-06, + "loss": 0.5153, + "step": 18282 + }, + { + "epoch": 3.32, + "learning_rate": 2.004386603567528e-06, + "loss": 0.4154, + "step": 18283 + }, + { + "epoch": 3.32, + "learning_rate": 2.004331998543866e-06, + "loss": 1.0114, + "step": 18284 + }, + { + "epoch": 3.33, + "learning_rate": 2.004277393520204e-06, + "loss": 0.4606, + "step": 18285 + }, + { + "epoch": 3.33, + "learning_rate": 2.0042227884965416e-06, + "loss": 0.5253, + "step": 18286 + }, + { + "epoch": 3.33, + "learning_rate": 2.0041681834728795e-06, + "loss": 2.0492, + "step": 18287 + }, + { + "epoch": 3.33, + "learning_rate": 2.0041135784492175e-06, + "loss": 0.2187, + "step": 18288 + }, + { + "epoch": 3.33, + "learning_rate": 2.004058973425555e-06, + "loss": 4.3063, + "step": 18289 + }, + { + "epoch": 3.33, + "learning_rate": 2.004004368401893e-06, + "loss": 0.5582, + "step": 18290 + }, + { + "epoch": 3.33, + "learning_rate": 2.003949763378231e-06, + "loss": 0.2197, + "step": 18291 + }, + { + "epoch": 3.33, + "learning_rate": 2.0038951583545684e-06, + "loss": 0.3776, + "step": 18292 + }, + { + "epoch": 3.33, + "learning_rate": 2.0038405533309064e-06, + "loss": 1.0614, + "step": 18293 + }, + { + "epoch": 3.33, + "learning_rate": 2.0037859483072443e-06, + "loss": 0.2032, + "step": 18294 + }, + { + "epoch": 3.33, + "learning_rate": 2.0037313432835822e-06, + "loss": 0.1389, + "step": 18295 + }, + { + "epoch": 3.33, + "learning_rate": 2.00367673825992e-06, + "loss": 0.8516, + "step": 18296 + }, + { + "epoch": 3.33, + "learning_rate": 2.0036221332362577e-06, + "loss": 1.9164, + "step": 18297 + }, + { + "epoch": 3.33, + "learning_rate": 2.0035675282125957e-06, + "loss": 0.4867, + "step": 18298 + }, + { + "epoch": 3.33, + "learning_rate": 2.0035129231889336e-06, + "loss": 0.0149, + "step": 18299 + }, + { + "epoch": 3.33, + "learning_rate": 2.003458318165271e-06, + "loss": 1.4143, + "step": 18300 + }, + { + "epoch": 3.33, + "learning_rate": 2.003403713141609e-06, + "loss": 1.0547, + "step": 18301 + }, + { + "epoch": 3.33, + "learning_rate": 2.003349108117947e-06, + "loss": 0.1337, + "step": 18302 + }, + { + "epoch": 3.33, + "learning_rate": 2.0032945030942845e-06, + "loss": 1.9599, + "step": 18303 + }, + { + "epoch": 3.33, + "learning_rate": 2.0032398980706225e-06, + "loss": 0.4704, + "step": 18304 + }, + { + "epoch": 3.33, + "learning_rate": 2.0031852930469604e-06, + "loss": 0.4836, + "step": 18305 + }, + { + "epoch": 3.33, + "learning_rate": 2.003130688023298e-06, + "loss": 2.0355, + "step": 18306 + }, + { + "epoch": 3.33, + "learning_rate": 2.003076082999636e-06, + "loss": 0.6032, + "step": 18307 + }, + { + "epoch": 3.33, + "learning_rate": 2.003021477975974e-06, + "loss": 0.7455, + "step": 18308 + }, + { + "epoch": 3.33, + "learning_rate": 2.0029668729523118e-06, + "loss": 1.3223, + "step": 18309 + }, + { + "epoch": 3.33, + "learning_rate": 2.0029122679286497e-06, + "loss": 0.4861, + "step": 18310 + }, + { + "epoch": 3.33, + "learning_rate": 2.0028576629049872e-06, + "loss": 0.2636, + "step": 18311 + }, + { + "epoch": 3.33, + "learning_rate": 2.002803057881325e-06, + "loss": 0.3713, + "step": 18312 + }, + { + "epoch": 3.33, + "learning_rate": 2.002748452857663e-06, + "loss": 0.0482, + "step": 18313 + }, + { + "epoch": 3.33, + "learning_rate": 2.0026938478340006e-06, + "loss": 0.6337, + "step": 18314 + }, + { + "epoch": 3.33, + "learning_rate": 2.0026392428103386e-06, + "loss": 1.8471, + "step": 18315 + }, + { + "epoch": 3.33, + "learning_rate": 2.0025846377866765e-06, + "loss": 1.6177, + "step": 18316 + }, + { + "epoch": 3.33, + "learning_rate": 2.002530032763014e-06, + "loss": 0.7583, + "step": 18317 + }, + { + "epoch": 3.33, + "learning_rate": 2.002475427739352e-06, + "loss": 0.33, + "step": 18318 + }, + { + "epoch": 3.33, + "learning_rate": 2.00242082271569e-06, + "loss": 0.1124, + "step": 18319 + }, + { + "epoch": 3.33, + "learning_rate": 2.002366217692028e-06, + "loss": 0.0094, + "step": 18320 + }, + { + "epoch": 3.33, + "learning_rate": 2.0023116126683654e-06, + "loss": 1.4581, + "step": 18321 + }, + { + "epoch": 3.33, + "learning_rate": 2.0022570076447033e-06, + "loss": 0.0475, + "step": 18322 + }, + { + "epoch": 3.33, + "learning_rate": 2.0022024026210413e-06, + "loss": 0.5854, + "step": 18323 + }, + { + "epoch": 3.33, + "learning_rate": 2.0021477975973792e-06, + "loss": 0.9089, + "step": 18324 + }, + { + "epoch": 3.33, + "learning_rate": 2.0020931925737167e-06, + "loss": 0.2961, + "step": 18325 + }, + { + "epoch": 3.33, + "learning_rate": 2.0020385875500547e-06, + "loss": 0.7038, + "step": 18326 + }, + { + "epoch": 3.33, + "learning_rate": 2.0019839825263926e-06, + "loss": 0.0241, + "step": 18327 + }, + { + "epoch": 3.33, + "learning_rate": 2.00192937750273e-06, + "loss": 1.059, + "step": 18328 + }, + { + "epoch": 3.33, + "learning_rate": 2.001874772479068e-06, + "loss": 0.906, + "step": 18329 + }, + { + "epoch": 3.33, + "learning_rate": 2.001820167455406e-06, + "loss": 0.4311, + "step": 18330 + }, + { + "epoch": 3.33, + "learning_rate": 2.0017655624317436e-06, + "loss": 0.0122, + "step": 18331 + }, + { + "epoch": 3.33, + "learning_rate": 2.0017109574080815e-06, + "loss": 1.1171, + "step": 18332 + }, + { + "epoch": 3.33, + "learning_rate": 2.0016563523844194e-06, + "loss": 0.2244, + "step": 18333 + }, + { + "epoch": 3.33, + "learning_rate": 2.0016017473607574e-06, + "loss": 0.2735, + "step": 18334 + }, + { + "epoch": 3.33, + "learning_rate": 2.0015471423370953e-06, + "loss": 0.9204, + "step": 18335 + }, + { + "epoch": 3.33, + "learning_rate": 2.001492537313433e-06, + "loss": 1.0089, + "step": 18336 + }, + { + "epoch": 3.33, + "learning_rate": 2.0014379322897708e-06, + "loss": 0.8163, + "step": 18337 + }, + { + "epoch": 3.33, + "learning_rate": 2.0013833272661087e-06, + "loss": 0.3483, + "step": 18338 + }, + { + "epoch": 3.33, + "learning_rate": 2.0013287222424463e-06, + "loss": 1.827, + "step": 18339 + }, + { + "epoch": 3.34, + "learning_rate": 2.001274117218784e-06, + "loss": 0.4392, + "step": 18340 + }, + { + "epoch": 3.34, + "learning_rate": 2.001219512195122e-06, + "loss": 0.1632, + "step": 18341 + }, + { + "epoch": 3.34, + "learning_rate": 2.0011649071714597e-06, + "loss": 1.2624, + "step": 18342 + }, + { + "epoch": 3.34, + "learning_rate": 2.0011103021477976e-06, + "loss": 0.3944, + "step": 18343 + }, + { + "epoch": 3.34, + "learning_rate": 2.0010556971241355e-06, + "loss": 1.6178, + "step": 18344 + }, + { + "epoch": 3.34, + "learning_rate": 2.001001092100473e-06, + "loss": 0.8453, + "step": 18345 + }, + { + "epoch": 3.34, + "learning_rate": 2.000946487076811e-06, + "loss": 0.3315, + "step": 18346 + }, + { + "epoch": 3.34, + "learning_rate": 2.000891882053149e-06, + "loss": 0.1854, + "step": 18347 + }, + { + "epoch": 3.34, + "learning_rate": 2.000837277029487e-06, + "loss": 1.087, + "step": 18348 + }, + { + "epoch": 3.34, + "learning_rate": 2.000782672005825e-06, + "loss": 0.7931, + "step": 18349 + }, + { + "epoch": 3.34, + "learning_rate": 2.0007280669821624e-06, + "loss": 1.0985, + "step": 18350 + }, + { + "epoch": 3.34, + "learning_rate": 2.0006734619585003e-06, + "loss": 0.3408, + "step": 18351 + }, + { + "epoch": 3.34, + "learning_rate": 2.0006188569348382e-06, + "loss": 0.2227, + "step": 18352 + }, + { + "epoch": 3.34, + "learning_rate": 2.0005642519111758e-06, + "loss": 0.1088, + "step": 18353 + }, + { + "epoch": 3.34, + "learning_rate": 2.0005096468875137e-06, + "loss": 0.252, + "step": 18354 + }, + { + "epoch": 3.34, + "learning_rate": 2.0004550418638517e-06, + "loss": 0.5157, + "step": 18355 + }, + { + "epoch": 3.34, + "learning_rate": 2.000400436840189e-06, + "loss": 3.2021, + "step": 18356 + }, + { + "epoch": 3.34, + "learning_rate": 2.000345831816527e-06, + "loss": 0.3585, + "step": 18357 + }, + { + "epoch": 3.34, + "learning_rate": 2.000291226792865e-06, + "loss": 0.0321, + "step": 18358 + }, + { + "epoch": 3.34, + "learning_rate": 2.0002366217692026e-06, + "loss": 0.4242, + "step": 18359 + }, + { + "epoch": 3.34, + "learning_rate": 2.0001820167455405e-06, + "loss": 1.3118, + "step": 18360 + }, + { + "epoch": 3.34, + "learning_rate": 2.0001274117218785e-06, + "loss": 1.2219, + "step": 18361 + }, + { + "epoch": 3.34, + "learning_rate": 2.0000728066982164e-06, + "loss": 2.1053, + "step": 18362 + }, + { + "epoch": 3.34, + "learning_rate": 2.0000182016745544e-06, + "loss": 0.5142, + "step": 18363 + }, + { + "epoch": 3.34, + "learning_rate": 1.999963596650892e-06, + "loss": 1.2832, + "step": 18364 + }, + { + "epoch": 3.34, + "learning_rate": 1.99990899162723e-06, + "loss": 0.186, + "step": 18365 + }, + { + "epoch": 3.34, + "learning_rate": 1.9998543866035678e-06, + "loss": 0.7654, + "step": 18366 + }, + { + "epoch": 3.34, + "learning_rate": 1.9997997815799053e-06, + "loss": 0.1344, + "step": 18367 + }, + { + "epoch": 3.34, + "learning_rate": 1.9997451765562432e-06, + "loss": 0.0379, + "step": 18368 + }, + { + "epoch": 3.34, + "learning_rate": 1.999690571532581e-06, + "loss": 0.5526, + "step": 18369 + }, + { + "epoch": 3.34, + "learning_rate": 1.9996359665089187e-06, + "loss": 1.8466, + "step": 18370 + }, + { + "epoch": 3.34, + "learning_rate": 1.9995813614852566e-06, + "loss": 0.822, + "step": 18371 + }, + { + "epoch": 3.34, + "learning_rate": 1.9995267564615946e-06, + "loss": 2.6083, + "step": 18372 + }, + { + "epoch": 3.34, + "learning_rate": 1.999472151437932e-06, + "loss": 0.6397, + "step": 18373 + }, + { + "epoch": 3.34, + "learning_rate": 1.99941754641427e-06, + "loss": 1.2278, + "step": 18374 + }, + { + "epoch": 3.34, + "learning_rate": 1.999362941390608e-06, + "loss": 0.2437, + "step": 18375 + }, + { + "epoch": 3.34, + "learning_rate": 1.999308336366946e-06, + "loss": 0.337, + "step": 18376 + }, + { + "epoch": 3.34, + "learning_rate": 1.999253731343284e-06, + "loss": 0.2609, + "step": 18377 + }, + { + "epoch": 3.34, + "learning_rate": 1.9991991263196214e-06, + "loss": 0.9768, + "step": 18378 + }, + { + "epoch": 3.34, + "learning_rate": 1.9991445212959593e-06, + "loss": 2.3414, + "step": 18379 + }, + { + "epoch": 3.34, + "learning_rate": 1.9990899162722973e-06, + "loss": 2.3147, + "step": 18380 + }, + { + "epoch": 3.34, + "learning_rate": 1.999035311248635e-06, + "loss": 0.5914, + "step": 18381 + }, + { + "epoch": 3.34, + "learning_rate": 1.9989807062249727e-06, + "loss": 0.1085, + "step": 18382 + }, + { + "epoch": 3.34, + "learning_rate": 1.9989261012013107e-06, + "loss": 0.074, + "step": 18383 + }, + { + "epoch": 3.34, + "learning_rate": 1.998871496177648e-06, + "loss": 1.5161, + "step": 18384 + }, + { + "epoch": 3.34, + "learning_rate": 1.998816891153986e-06, + "loss": 0.1437, + "step": 18385 + }, + { + "epoch": 3.34, + "learning_rate": 1.998762286130324e-06, + "loss": 0.1627, + "step": 18386 + }, + { + "epoch": 3.34, + "learning_rate": 1.9987076811066616e-06, + "loss": 0.7422, + "step": 18387 + }, + { + "epoch": 3.34, + "learning_rate": 1.9986530760829995e-06, + "loss": 0.5015, + "step": 18388 + }, + { + "epoch": 3.34, + "learning_rate": 1.9985984710593375e-06, + "loss": 0.2527, + "step": 18389 + }, + { + "epoch": 3.34, + "learning_rate": 1.9985438660356754e-06, + "loss": 0.4801, + "step": 18390 + }, + { + "epoch": 3.34, + "learning_rate": 1.9984892610120134e-06, + "loss": 1.9276, + "step": 18391 + }, + { + "epoch": 3.34, + "learning_rate": 1.998434655988351e-06, + "loss": 2.4142, + "step": 18392 + }, + { + "epoch": 3.34, + "learning_rate": 1.998380050964689e-06, + "loss": 1.0293, + "step": 18393 + }, + { + "epoch": 3.34, + "learning_rate": 1.9983254459410268e-06, + "loss": 1.0184, + "step": 18394 + }, + { + "epoch": 3.35, + "learning_rate": 1.9982708409173643e-06, + "loss": 0.3859, + "step": 18395 + }, + { + "epoch": 3.35, + "learning_rate": 1.9982162358937022e-06, + "loss": 0.2213, + "step": 18396 + }, + { + "epoch": 3.35, + "learning_rate": 1.99816163087004e-06, + "loss": 0.4609, + "step": 18397 + }, + { + "epoch": 3.35, + "learning_rate": 1.9981070258463777e-06, + "loss": 1.5485, + "step": 18398 + }, + { + "epoch": 3.35, + "learning_rate": 1.9980524208227157e-06, + "loss": 0.8572, + "step": 18399 + }, + { + "epoch": 3.35, + "learning_rate": 1.9979978157990536e-06, + "loss": 0.5534, + "step": 18400 + }, + { + "epoch": 3.35, + "learning_rate": 1.997943210775391e-06, + "loss": 0.4298, + "step": 18401 + }, + { + "epoch": 3.35, + "learning_rate": 1.997888605751729e-06, + "loss": 1.5558, + "step": 18402 + }, + { + "epoch": 3.35, + "learning_rate": 1.997834000728067e-06, + "loss": 0.9135, + "step": 18403 + }, + { + "epoch": 3.35, + "learning_rate": 1.997779395704405e-06, + "loss": 0.3311, + "step": 18404 + }, + { + "epoch": 3.35, + "learning_rate": 1.997724790680743e-06, + "loss": 1.2872, + "step": 18405 + }, + { + "epoch": 3.35, + "learning_rate": 1.9976701856570804e-06, + "loss": 0.6567, + "step": 18406 + }, + { + "epoch": 3.35, + "learning_rate": 1.9976155806334184e-06, + "loss": 0.8082, + "step": 18407 + }, + { + "epoch": 3.35, + "learning_rate": 1.9975609756097563e-06, + "loss": 1.5633, + "step": 18408 + }, + { + "epoch": 3.35, + "learning_rate": 1.997506370586094e-06, + "loss": 2.0676, + "step": 18409 + }, + { + "epoch": 3.35, + "learning_rate": 1.9974517655624318e-06, + "loss": 0.0693, + "step": 18410 + }, + { + "epoch": 3.35, + "learning_rate": 1.9973971605387697e-06, + "loss": 0.3263, + "step": 18411 + }, + { + "epoch": 3.35, + "learning_rate": 1.9973425555151072e-06, + "loss": 0.0094, + "step": 18412 + }, + { + "epoch": 3.35, + "learning_rate": 1.997287950491445e-06, + "loss": 0.5138, + "step": 18413 + }, + { + "epoch": 3.35, + "learning_rate": 1.997233345467783e-06, + "loss": 0.2756, + "step": 18414 + }, + { + "epoch": 3.35, + "learning_rate": 1.997178740444121e-06, + "loss": 0.1876, + "step": 18415 + }, + { + "epoch": 3.35, + "learning_rate": 1.997124135420459e-06, + "loss": 1.5221, + "step": 18416 + }, + { + "epoch": 3.35, + "learning_rate": 1.9970695303967965e-06, + "loss": 2.4152, + "step": 18417 + }, + { + "epoch": 3.35, + "learning_rate": 1.9970149253731345e-06, + "loss": 0.2841, + "step": 18418 + }, + { + "epoch": 3.35, + "learning_rate": 1.9969603203494724e-06, + "loss": 0.6079, + "step": 18419 + }, + { + "epoch": 3.35, + "learning_rate": 1.99690571532581e-06, + "loss": 0.2175, + "step": 18420 + }, + { + "epoch": 3.35, + "learning_rate": 1.996851110302148e-06, + "loss": 1.1796, + "step": 18421 + }, + { + "epoch": 3.35, + "learning_rate": 1.996796505278486e-06, + "loss": 0.0188, + "step": 18422 + }, + { + "epoch": 3.35, + "learning_rate": 1.9967419002548233e-06, + "loss": 0.5554, + "step": 18423 + }, + { + "epoch": 3.35, + "learning_rate": 1.9966872952311613e-06, + "loss": 1.1248, + "step": 18424 + }, + { + "epoch": 3.35, + "learning_rate": 1.9966326902074992e-06, + "loss": 0.5772, + "step": 18425 + }, + { + "epoch": 3.35, + "learning_rate": 1.9965780851838367e-06, + "loss": 0.1693, + "step": 18426 + }, + { + "epoch": 3.35, + "learning_rate": 1.9965234801601747e-06, + "loss": 0.4668, + "step": 18427 + }, + { + "epoch": 3.35, + "learning_rate": 1.9964688751365126e-06, + "loss": 0.5542, + "step": 18428 + }, + { + "epoch": 3.35, + "learning_rate": 1.9964142701128506e-06, + "loss": 0.2426, + "step": 18429 + }, + { + "epoch": 3.35, + "learning_rate": 1.9963596650891885e-06, + "loss": 0.3542, + "step": 18430 + }, + { + "epoch": 3.35, + "learning_rate": 1.996305060065526e-06, + "loss": 0.0113, + "step": 18431 + }, + { + "epoch": 3.35, + "learning_rate": 1.996250455041864e-06, + "loss": 1.3144, + "step": 18432 + }, + { + "epoch": 3.35, + "learning_rate": 1.996195850018202e-06, + "loss": 0.1139, + "step": 18433 + }, + { + "epoch": 3.35, + "learning_rate": 1.9961412449945394e-06, + "loss": 0.1606, + "step": 18434 + }, + { + "epoch": 3.35, + "learning_rate": 1.9960866399708774e-06, + "loss": 1.0199, + "step": 18435 + }, + { + "epoch": 3.35, + "learning_rate": 1.9960320349472153e-06, + "loss": 0.7523, + "step": 18436 + }, + { + "epoch": 3.35, + "learning_rate": 1.995977429923553e-06, + "loss": 0.5493, + "step": 18437 + }, + { + "epoch": 3.35, + "learning_rate": 1.995922824899891e-06, + "loss": 0.5314, + "step": 18438 + }, + { + "epoch": 3.35, + "learning_rate": 1.9958682198762287e-06, + "loss": 0.0129, + "step": 18439 + }, + { + "epoch": 3.35, + "learning_rate": 1.9958136148525663e-06, + "loss": 0.1646, + "step": 18440 + }, + { + "epoch": 3.35, + "learning_rate": 1.995759009828904e-06, + "loss": 0.0142, + "step": 18441 + }, + { + "epoch": 3.35, + "learning_rate": 1.995704404805242e-06, + "loss": 1.6134, + "step": 18442 + }, + { + "epoch": 3.35, + "learning_rate": 1.99564979978158e-06, + "loss": 0.6809, + "step": 18443 + }, + { + "epoch": 3.35, + "learning_rate": 1.995595194757918e-06, + "loss": 0.4965, + "step": 18444 + }, + { + "epoch": 3.35, + "learning_rate": 1.9955405897342555e-06, + "loss": 1.0055, + "step": 18445 + }, + { + "epoch": 3.35, + "learning_rate": 1.9954859847105935e-06, + "loss": 0.4603, + "step": 18446 + }, + { + "epoch": 3.35, + "learning_rate": 1.9954313796869314e-06, + "loss": 0.2677, + "step": 18447 + }, + { + "epoch": 3.35, + "learning_rate": 1.995376774663269e-06, + "loss": 0.3526, + "step": 18448 + }, + { + "epoch": 3.35, + "learning_rate": 1.995322169639607e-06, + "loss": 0.7603, + "step": 18449 + }, + { + "epoch": 3.36, + "learning_rate": 1.995267564615945e-06, + "loss": 0.5837, + "step": 18450 + }, + { + "epoch": 3.36, + "learning_rate": 1.9952129595922824e-06, + "loss": 0.6561, + "step": 18451 + }, + { + "epoch": 3.36, + "learning_rate": 1.9951583545686203e-06, + "loss": 0.7611, + "step": 18452 + }, + { + "epoch": 3.36, + "learning_rate": 1.9951037495449582e-06, + "loss": 0.8491, + "step": 18453 + }, + { + "epoch": 3.36, + "learning_rate": 1.9950491445212958e-06, + "loss": 0.45, + "step": 18454 + }, + { + "epoch": 3.36, + "learning_rate": 1.9949945394976337e-06, + "loss": 1.2623, + "step": 18455 + }, + { + "epoch": 3.36, + "learning_rate": 1.9949399344739717e-06, + "loss": 0.4237, + "step": 18456 + }, + { + "epoch": 3.36, + "learning_rate": 1.9948853294503096e-06, + "loss": 0.3684, + "step": 18457 + }, + { + "epoch": 3.36, + "learning_rate": 1.9948307244266475e-06, + "loss": 2.3535, + "step": 18458 + }, + { + "epoch": 3.36, + "learning_rate": 1.994776119402985e-06, + "loss": 0.8265, + "step": 18459 + }, + { + "epoch": 3.36, + "learning_rate": 1.994721514379323e-06, + "loss": 0.0126, + "step": 18460 + }, + { + "epoch": 3.36, + "learning_rate": 1.994666909355661e-06, + "loss": 0.5151, + "step": 18461 + }, + { + "epoch": 3.36, + "learning_rate": 1.9946123043319985e-06, + "loss": 1.6578, + "step": 18462 + }, + { + "epoch": 3.36, + "learning_rate": 1.9945576993083364e-06, + "loss": 0.8726, + "step": 18463 + }, + { + "epoch": 3.36, + "learning_rate": 1.9945030942846744e-06, + "loss": 1.7645, + "step": 18464 + }, + { + "epoch": 3.36, + "learning_rate": 1.994448489261012e-06, + "loss": 2.4662, + "step": 18465 + }, + { + "epoch": 3.36, + "learning_rate": 1.99439388423735e-06, + "loss": 0.3702, + "step": 18466 + }, + { + "epoch": 3.36, + "learning_rate": 1.9943392792136878e-06, + "loss": 0.3251, + "step": 18467 + }, + { + "epoch": 3.36, + "learning_rate": 1.9942846741900253e-06, + "loss": 0.0482, + "step": 18468 + }, + { + "epoch": 3.36, + "learning_rate": 1.9942300691663632e-06, + "loss": 0.7941, + "step": 18469 + }, + { + "epoch": 3.36, + "learning_rate": 1.994175464142701e-06, + "loss": 0.2532, + "step": 18470 + }, + { + "epoch": 3.36, + "learning_rate": 1.994120859119039e-06, + "loss": 0.2912, + "step": 18471 + }, + { + "epoch": 3.36, + "learning_rate": 1.994066254095377e-06, + "loss": 0.7781, + "step": 18472 + }, + { + "epoch": 3.36, + "learning_rate": 1.9940116490717146e-06, + "loss": 1.0034, + "step": 18473 + }, + { + "epoch": 3.36, + "learning_rate": 1.9939570440480525e-06, + "loss": 0.7578, + "step": 18474 + }, + { + "epoch": 3.36, + "learning_rate": 1.9939024390243905e-06, + "loss": 0.0971, + "step": 18475 + }, + { + "epoch": 3.36, + "learning_rate": 1.993847834000728e-06, + "loss": 0.5392, + "step": 18476 + }, + { + "epoch": 3.36, + "learning_rate": 1.993793228977066e-06, + "loss": 1.058, + "step": 18477 + }, + { + "epoch": 3.36, + "learning_rate": 1.993738623953404e-06, + "loss": 0.0971, + "step": 18478 + }, + { + "epoch": 3.36, + "learning_rate": 1.9936840189297414e-06, + "loss": 0.9858, + "step": 18479 + }, + { + "epoch": 3.36, + "learning_rate": 1.9936294139060793e-06, + "loss": 0.0256, + "step": 18480 + }, + { + "epoch": 3.36, + "learning_rate": 1.9935748088824173e-06, + "loss": 0.2416, + "step": 18481 + }, + { + "epoch": 3.36, + "learning_rate": 1.993520203858755e-06, + "loss": 0.0237, + "step": 18482 + }, + { + "epoch": 3.36, + "learning_rate": 1.9934655988350927e-06, + "loss": 2.0157, + "step": 18483 + }, + { + "epoch": 3.36, + "learning_rate": 1.9934109938114307e-06, + "loss": 0.48, + "step": 18484 + }, + { + "epoch": 3.36, + "learning_rate": 1.9933563887877686e-06, + "loss": 0.713, + "step": 18485 + }, + { + "epoch": 3.36, + "learning_rate": 1.9933017837641066e-06, + "loss": 0.5943, + "step": 18486 + }, + { + "epoch": 3.36, + "learning_rate": 1.993247178740444e-06, + "loss": 0.5336, + "step": 18487 + }, + { + "epoch": 3.36, + "learning_rate": 1.993192573716782e-06, + "loss": 1.7185, + "step": 18488 + }, + { + "epoch": 3.36, + "learning_rate": 1.99313796869312e-06, + "loss": 0.0727, + "step": 18489 + }, + { + "epoch": 3.36, + "learning_rate": 1.9930833636694575e-06, + "loss": 0.9006, + "step": 18490 + }, + { + "epoch": 3.36, + "learning_rate": 1.9930287586457954e-06, + "loss": 1.1392, + "step": 18491 + }, + { + "epoch": 3.36, + "learning_rate": 1.9929741536221334e-06, + "loss": 0.4273, + "step": 18492 + }, + { + "epoch": 3.36, + "learning_rate": 1.992919548598471e-06, + "loss": 0.7919, + "step": 18493 + }, + { + "epoch": 3.36, + "learning_rate": 1.992864943574809e-06, + "loss": 0.0986, + "step": 18494 + }, + { + "epoch": 3.36, + "learning_rate": 1.9928103385511468e-06, + "loss": 2.7778, + "step": 18495 + }, + { + "epoch": 3.36, + "learning_rate": 1.9927557335274847e-06, + "loss": 1.103, + "step": 18496 + }, + { + "epoch": 3.36, + "learning_rate": 1.9927011285038227e-06, + "loss": 0.2172, + "step": 18497 + }, + { + "epoch": 3.36, + "learning_rate": 1.99264652348016e-06, + "loss": 1.6497, + "step": 18498 + }, + { + "epoch": 3.36, + "learning_rate": 1.992591918456498e-06, + "loss": 0.2052, + "step": 18499 + }, + { + "epoch": 3.36, + "learning_rate": 1.992537313432836e-06, + "loss": 0.1308, + "step": 18500 + }, + { + "epoch": 3.36, + "learning_rate": 1.9924827084091736e-06, + "loss": 0.2918, + "step": 18501 + }, + { + "epoch": 3.36, + "learning_rate": 1.9924281033855115e-06, + "loss": 1.4178, + "step": 18502 + }, + { + "epoch": 3.36, + "learning_rate": 1.9923734983618495e-06, + "loss": 0.1585, + "step": 18503 + }, + { + "epoch": 3.36, + "learning_rate": 1.992318893338187e-06, + "loss": 2.6878, + "step": 18504 + }, + { + "epoch": 3.37, + "learning_rate": 1.992264288314525e-06, + "loss": 0.0466, + "step": 18505 + }, + { + "epoch": 3.37, + "learning_rate": 1.992209683290863e-06, + "loss": 0.3814, + "step": 18506 + }, + { + "epoch": 3.37, + "learning_rate": 1.9921550782672004e-06, + "loss": 0.6837, + "step": 18507 + }, + { + "epoch": 3.37, + "learning_rate": 1.9921004732435384e-06, + "loss": 0.0455, + "step": 18508 + }, + { + "epoch": 3.37, + "learning_rate": 1.9920458682198763e-06, + "loss": 0.0159, + "step": 18509 + }, + { + "epoch": 3.37, + "learning_rate": 1.9919912631962142e-06, + "loss": 0.5871, + "step": 18510 + }, + { + "epoch": 3.37, + "learning_rate": 1.991936658172552e-06, + "loss": 0.778, + "step": 18511 + }, + { + "epoch": 3.37, + "learning_rate": 1.9918820531488897e-06, + "loss": 0.8004, + "step": 18512 + }, + { + "epoch": 3.37, + "learning_rate": 1.9918274481252277e-06, + "loss": 0.2513, + "step": 18513 + }, + { + "epoch": 3.37, + "learning_rate": 1.9917728431015656e-06, + "loss": 0.2058, + "step": 18514 + }, + { + "epoch": 3.37, + "learning_rate": 1.991718238077903e-06, + "loss": 2.9077, + "step": 18515 + }, + { + "epoch": 3.37, + "learning_rate": 1.991663633054241e-06, + "loss": 1.7353, + "step": 18516 + }, + { + "epoch": 3.37, + "learning_rate": 1.991609028030579e-06, + "loss": 1.1331, + "step": 18517 + }, + { + "epoch": 3.37, + "learning_rate": 1.9915544230069165e-06, + "loss": 0.7088, + "step": 18518 + }, + { + "epoch": 3.37, + "learning_rate": 1.9914998179832545e-06, + "loss": 1.4729, + "step": 18519 + }, + { + "epoch": 3.37, + "learning_rate": 1.9914452129595924e-06, + "loss": 0.1367, + "step": 18520 + }, + { + "epoch": 3.37, + "learning_rate": 1.99139060793593e-06, + "loss": 0.3842, + "step": 18521 + }, + { + "epoch": 3.37, + "learning_rate": 1.991336002912268e-06, + "loss": 1.8097, + "step": 18522 + }, + { + "epoch": 3.37, + "learning_rate": 1.991281397888606e-06, + "loss": 0.3173, + "step": 18523 + }, + { + "epoch": 3.37, + "learning_rate": 1.9912267928649438e-06, + "loss": 0.5316, + "step": 18524 + }, + { + "epoch": 3.37, + "learning_rate": 1.9911721878412817e-06, + "loss": 0.6246, + "step": 18525 + }, + { + "epoch": 3.37, + "learning_rate": 1.9911175828176192e-06, + "loss": 1.5912, + "step": 18526 + }, + { + "epoch": 3.37, + "learning_rate": 1.991062977793957e-06, + "loss": 0.2137, + "step": 18527 + }, + { + "epoch": 3.37, + "learning_rate": 1.991008372770295e-06, + "loss": 1.7612, + "step": 18528 + }, + { + "epoch": 3.37, + "learning_rate": 1.9909537677466326e-06, + "loss": 0.1632, + "step": 18529 + }, + { + "epoch": 3.37, + "learning_rate": 1.9908991627229706e-06, + "loss": 1.1159, + "step": 18530 + }, + { + "epoch": 3.37, + "learning_rate": 1.9908445576993085e-06, + "loss": 3.1253, + "step": 18531 + }, + { + "epoch": 3.37, + "learning_rate": 1.990789952675646e-06, + "loss": 0.1801, + "step": 18532 + }, + { + "epoch": 3.37, + "learning_rate": 1.990735347651984e-06, + "loss": 0.4944, + "step": 18533 + }, + { + "epoch": 3.37, + "learning_rate": 1.990680742628322e-06, + "loss": 0.75, + "step": 18534 + }, + { + "epoch": 3.37, + "learning_rate": 1.9906261376046594e-06, + "loss": 0.8692, + "step": 18535 + }, + { + "epoch": 3.37, + "learning_rate": 1.9905715325809974e-06, + "loss": 0.7602, + "step": 18536 + }, + { + "epoch": 3.37, + "learning_rate": 1.9905169275573353e-06, + "loss": 0.7506, + "step": 18537 + }, + { + "epoch": 3.37, + "learning_rate": 1.9904623225336733e-06, + "loss": 0.8164, + "step": 18538 + }, + { + "epoch": 3.37, + "learning_rate": 1.9904077175100112e-06, + "loss": 0.2141, + "step": 18539 + }, + { + "epoch": 3.37, + "learning_rate": 1.9903531124863487e-06, + "loss": 0.8905, + "step": 18540 + }, + { + "epoch": 3.37, + "learning_rate": 1.9902985074626867e-06, + "loss": 0.4116, + "step": 18541 + }, + { + "epoch": 3.37, + "learning_rate": 1.9902439024390246e-06, + "loss": 0.7684, + "step": 18542 + }, + { + "epoch": 3.37, + "learning_rate": 1.990189297415362e-06, + "loss": 0.834, + "step": 18543 + }, + { + "epoch": 3.37, + "learning_rate": 1.9901346923917e-06, + "loss": 0.0628, + "step": 18544 + }, + { + "epoch": 3.37, + "learning_rate": 1.990080087368038e-06, + "loss": 1.2204, + "step": 18545 + }, + { + "epoch": 3.37, + "learning_rate": 1.9900254823443755e-06, + "loss": 1.0985, + "step": 18546 + }, + { + "epoch": 3.37, + "learning_rate": 1.9899708773207135e-06, + "loss": 0.6274, + "step": 18547 + }, + { + "epoch": 3.37, + "learning_rate": 1.9899162722970514e-06, + "loss": 0.4179, + "step": 18548 + }, + { + "epoch": 3.37, + "learning_rate": 1.989861667273389e-06, + "loss": 1.4998, + "step": 18549 + }, + { + "epoch": 3.37, + "learning_rate": 1.989807062249727e-06, + "loss": 0.6775, + "step": 18550 + }, + { + "epoch": 3.37, + "learning_rate": 1.989752457226065e-06, + "loss": 0.2156, + "step": 18551 + }, + { + "epoch": 3.37, + "learning_rate": 1.9896978522024028e-06, + "loss": 0.1544, + "step": 18552 + }, + { + "epoch": 3.37, + "learning_rate": 1.9896432471787407e-06, + "loss": 0.4003, + "step": 18553 + }, + { + "epoch": 3.37, + "learning_rate": 1.9895886421550782e-06, + "loss": 0.8969, + "step": 18554 + }, + { + "epoch": 3.37, + "learning_rate": 1.989534037131416e-06, + "loss": 0.3469, + "step": 18555 + }, + { + "epoch": 3.37, + "learning_rate": 1.989479432107754e-06, + "loss": 0.0355, + "step": 18556 + }, + { + "epoch": 3.37, + "learning_rate": 1.9894248270840917e-06, + "loss": 3.1543, + "step": 18557 + }, + { + "epoch": 3.37, + "learning_rate": 1.9893702220604296e-06, + "loss": 0.0214, + "step": 18558 + }, + { + "epoch": 3.37, + "learning_rate": 1.9893156170367675e-06, + "loss": 2.3624, + "step": 18559 + }, + { + "epoch": 3.38, + "learning_rate": 1.989261012013105e-06, + "loss": 1.6899, + "step": 18560 + }, + { + "epoch": 3.38, + "learning_rate": 1.989206406989443e-06, + "loss": 2.0365, + "step": 18561 + }, + { + "epoch": 3.38, + "learning_rate": 1.989151801965781e-06, + "loss": 0.2139, + "step": 18562 + }, + { + "epoch": 3.38, + "learning_rate": 1.9890971969421185e-06, + "loss": 4.3427, + "step": 18563 + }, + { + "epoch": 3.38, + "learning_rate": 1.9890425919184564e-06, + "loss": 1.2689, + "step": 18564 + }, + { + "epoch": 3.38, + "learning_rate": 1.9889879868947944e-06, + "loss": 0.3366, + "step": 18565 + }, + { + "epoch": 3.38, + "learning_rate": 1.9889333818711323e-06, + "loss": 2.3996, + "step": 18566 + }, + { + "epoch": 3.38, + "learning_rate": 1.9888787768474702e-06, + "loss": 0.231, + "step": 18567 + }, + { + "epoch": 3.38, + "learning_rate": 1.9888241718238078e-06, + "loss": 0.7997, + "step": 18568 + }, + { + "epoch": 3.38, + "learning_rate": 1.9887695668001457e-06, + "loss": 1.6791, + "step": 18569 + }, + { + "epoch": 3.38, + "learning_rate": 1.9887149617764836e-06, + "loss": 0.8638, + "step": 18570 + }, + { + "epoch": 3.38, + "learning_rate": 1.988660356752821e-06, + "loss": 1.0763, + "step": 18571 + }, + { + "epoch": 3.38, + "learning_rate": 1.988605751729159e-06, + "loss": 1.3869, + "step": 18572 + }, + { + "epoch": 3.38, + "learning_rate": 1.988551146705497e-06, + "loss": 0.3079, + "step": 18573 + }, + { + "epoch": 3.38, + "learning_rate": 1.9884965416818346e-06, + "loss": 1.6505, + "step": 18574 + }, + { + "epoch": 3.38, + "learning_rate": 1.9884419366581725e-06, + "loss": 0.6301, + "step": 18575 + }, + { + "epoch": 3.38, + "learning_rate": 1.9883873316345105e-06, + "loss": 2.1989, + "step": 18576 + }, + { + "epoch": 3.38, + "learning_rate": 1.9883327266108484e-06, + "loss": 0.2346, + "step": 18577 + }, + { + "epoch": 3.38, + "learning_rate": 1.9882781215871863e-06, + "loss": 0.8514, + "step": 18578 + }, + { + "epoch": 3.38, + "learning_rate": 1.988223516563524e-06, + "loss": 0.7789, + "step": 18579 + }, + { + "epoch": 3.38, + "learning_rate": 1.988168911539862e-06, + "loss": 3.6699, + "step": 18580 + }, + { + "epoch": 3.38, + "learning_rate": 1.9881143065161998e-06, + "loss": 1.7982, + "step": 18581 + }, + { + "epoch": 3.38, + "learning_rate": 1.9880597014925373e-06, + "loss": 0.4226, + "step": 18582 + }, + { + "epoch": 3.38, + "learning_rate": 1.9880050964688752e-06, + "loss": 2.2578, + "step": 18583 + }, + { + "epoch": 3.38, + "learning_rate": 1.987950491445213e-06, + "loss": 0.2885, + "step": 18584 + }, + { + "epoch": 3.38, + "learning_rate": 1.9878958864215507e-06, + "loss": 1.3228, + "step": 18585 + }, + { + "epoch": 3.38, + "learning_rate": 1.9878412813978886e-06, + "loss": 1.308, + "step": 18586 + }, + { + "epoch": 3.38, + "learning_rate": 1.9877866763742266e-06, + "loss": 1.8269, + "step": 18587 + }, + { + "epoch": 3.38, + "learning_rate": 1.987732071350564e-06, + "loss": 1.2563, + "step": 18588 + }, + { + "epoch": 3.38, + "learning_rate": 1.987677466326902e-06, + "loss": 0.275, + "step": 18589 + }, + { + "epoch": 3.38, + "learning_rate": 1.98762286130324e-06, + "loss": 0.0653, + "step": 18590 + }, + { + "epoch": 3.38, + "learning_rate": 1.987568256279578e-06, + "loss": 0.1143, + "step": 18591 + }, + { + "epoch": 3.38, + "learning_rate": 1.987513651255916e-06, + "loss": 0.5002, + "step": 18592 + }, + { + "epoch": 3.38, + "learning_rate": 1.9874590462322534e-06, + "loss": 2.0039, + "step": 18593 + }, + { + "epoch": 3.38, + "learning_rate": 1.9874044412085913e-06, + "loss": 0.677, + "step": 18594 + }, + { + "epoch": 3.38, + "learning_rate": 1.9873498361849293e-06, + "loss": 0.9646, + "step": 18595 + }, + { + "epoch": 3.38, + "learning_rate": 1.9872952311612668e-06, + "loss": 0.0219, + "step": 18596 + }, + { + "epoch": 3.38, + "learning_rate": 1.9872406261376047e-06, + "loss": 0.4594, + "step": 18597 + }, + { + "epoch": 3.38, + "learning_rate": 1.9871860211139427e-06, + "loss": 1.3488, + "step": 18598 + }, + { + "epoch": 3.38, + "learning_rate": 1.98713141609028e-06, + "loss": 0.3814, + "step": 18599 + }, + { + "epoch": 3.38, + "learning_rate": 1.987076811066618e-06, + "loss": 0.0186, + "step": 18600 + }, + { + "epoch": 3.38, + "learning_rate": 1.987022206042956e-06, + "loss": 2.7516, + "step": 18601 + }, + { + "epoch": 3.38, + "learning_rate": 1.9869676010192936e-06, + "loss": 0.7082, + "step": 18602 + }, + { + "epoch": 3.38, + "learning_rate": 1.9869129959956315e-06, + "loss": 0.7406, + "step": 18603 + }, + { + "epoch": 3.38, + "learning_rate": 1.9868583909719695e-06, + "loss": 0.3279, + "step": 18604 + }, + { + "epoch": 3.38, + "learning_rate": 1.9868037859483074e-06, + "loss": 4.4313, + "step": 18605 + }, + { + "epoch": 3.38, + "learning_rate": 1.9867491809246454e-06, + "loss": 1.2857, + "step": 18606 + }, + { + "epoch": 3.38, + "learning_rate": 1.986694575900983e-06, + "loss": 0.0097, + "step": 18607 + }, + { + "epoch": 3.38, + "learning_rate": 1.986639970877321e-06, + "loss": 0.9457, + "step": 18608 + }, + { + "epoch": 3.38, + "learning_rate": 1.9865853658536588e-06, + "loss": 0.8253, + "step": 18609 + }, + { + "epoch": 3.38, + "learning_rate": 1.9865307608299963e-06, + "loss": 0.3071, + "step": 18610 + }, + { + "epoch": 3.38, + "learning_rate": 1.9864761558063342e-06, + "loss": 0.8839, + "step": 18611 + }, + { + "epoch": 3.38, + "learning_rate": 1.986421550782672e-06, + "loss": 0.0484, + "step": 18612 + }, + { + "epoch": 3.38, + "learning_rate": 1.9863669457590097e-06, + "loss": 0.8449, + "step": 18613 + }, + { + "epoch": 3.38, + "learning_rate": 1.9863123407353477e-06, + "loss": 0.6636, + "step": 18614 + }, + { + "epoch": 3.39, + "learning_rate": 1.9862577357116856e-06, + "loss": 0.7456, + "step": 18615 + }, + { + "epoch": 3.39, + "learning_rate": 1.986203130688023e-06, + "loss": 0.0806, + "step": 18616 + }, + { + "epoch": 3.39, + "learning_rate": 1.986148525664361e-06, + "loss": 0.1427, + "step": 18617 + }, + { + "epoch": 3.39, + "learning_rate": 1.986093920640699e-06, + "loss": 0.5719, + "step": 18618 + }, + { + "epoch": 3.39, + "learning_rate": 1.986039315617037e-06, + "loss": 0.3027, + "step": 18619 + }, + { + "epoch": 3.39, + "learning_rate": 1.985984710593375e-06, + "loss": 0.4136, + "step": 18620 + }, + { + "epoch": 3.39, + "learning_rate": 1.9859301055697124e-06, + "loss": 1.7513, + "step": 18621 + }, + { + "epoch": 3.39, + "learning_rate": 1.9858755005460504e-06, + "loss": 0.4557, + "step": 18622 + }, + { + "epoch": 3.39, + "learning_rate": 1.9858208955223883e-06, + "loss": 0.1928, + "step": 18623 + }, + { + "epoch": 3.39, + "learning_rate": 1.985766290498726e-06, + "loss": 0.0472, + "step": 18624 + }, + { + "epoch": 3.39, + "learning_rate": 1.9857116854750638e-06, + "loss": 1.0867, + "step": 18625 + }, + { + "epoch": 3.39, + "learning_rate": 1.9856570804514017e-06, + "loss": 3.2845, + "step": 18626 + }, + { + "epoch": 3.39, + "learning_rate": 1.9856024754277392e-06, + "loss": 4.5034, + "step": 18627 + }, + { + "epoch": 3.39, + "learning_rate": 1.985547870404077e-06, + "loss": 0.6002, + "step": 18628 + }, + { + "epoch": 3.39, + "learning_rate": 1.985493265380415e-06, + "loss": 1.8262, + "step": 18629 + }, + { + "epoch": 3.39, + "learning_rate": 1.9854386603567526e-06, + "loss": 0.6061, + "step": 18630 + }, + { + "epoch": 3.39, + "learning_rate": 1.9853840553330906e-06, + "loss": 0.6549, + "step": 18631 + }, + { + "epoch": 3.39, + "learning_rate": 1.9853294503094285e-06, + "loss": 0.7173, + "step": 18632 + }, + { + "epoch": 3.39, + "learning_rate": 1.9852748452857665e-06, + "loss": 2.3938, + "step": 18633 + }, + { + "epoch": 3.39, + "learning_rate": 1.9852202402621044e-06, + "loss": 2.6643, + "step": 18634 + }, + { + "epoch": 3.39, + "learning_rate": 1.985165635238442e-06, + "loss": 0.1965, + "step": 18635 + }, + { + "epoch": 3.39, + "learning_rate": 1.98511103021478e-06, + "loss": 0.1072, + "step": 18636 + }, + { + "epoch": 3.39, + "learning_rate": 1.985056425191118e-06, + "loss": 0.0154, + "step": 18637 + }, + { + "epoch": 3.39, + "learning_rate": 1.9850018201674553e-06, + "loss": 1.3035, + "step": 18638 + }, + { + "epoch": 3.39, + "learning_rate": 1.9849472151437933e-06, + "loss": 0.0361, + "step": 18639 + }, + { + "epoch": 3.39, + "learning_rate": 1.9848926101201312e-06, + "loss": 1.1671, + "step": 18640 + }, + { + "epoch": 3.39, + "learning_rate": 1.9848380050964687e-06, + "loss": 0.554, + "step": 18641 + }, + { + "epoch": 3.39, + "learning_rate": 1.9847834000728067e-06, + "loss": 1.0154, + "step": 18642 + }, + { + "epoch": 3.39, + "learning_rate": 1.9847287950491446e-06, + "loss": 1.512, + "step": 18643 + }, + { + "epoch": 3.39, + "learning_rate": 1.984674190025482e-06, + "loss": 3.6542, + "step": 18644 + }, + { + "epoch": 3.39, + "learning_rate": 1.98461958500182e-06, + "loss": 0.0968, + "step": 18645 + }, + { + "epoch": 3.39, + "learning_rate": 1.984564979978158e-06, + "loss": 0.0134, + "step": 18646 + }, + { + "epoch": 3.39, + "learning_rate": 1.984510374954496e-06, + "loss": 0.0165, + "step": 18647 + }, + { + "epoch": 3.39, + "learning_rate": 1.984455769930834e-06, + "loss": 0.8988, + "step": 18648 + }, + { + "epoch": 3.39, + "learning_rate": 1.9844011649071714e-06, + "loss": 0.0979, + "step": 18649 + }, + { + "epoch": 3.39, + "learning_rate": 1.9843465598835094e-06, + "loss": 0.26, + "step": 18650 + }, + { + "epoch": 3.39, + "learning_rate": 1.9842919548598473e-06, + "loss": 0.4379, + "step": 18651 + }, + { + "epoch": 3.39, + "learning_rate": 1.984237349836185e-06, + "loss": 0.1415, + "step": 18652 + }, + { + "epoch": 3.39, + "learning_rate": 1.9841827448125228e-06, + "loss": 0.8863, + "step": 18653 + }, + { + "epoch": 3.39, + "learning_rate": 1.9841281397888607e-06, + "loss": 0.1597, + "step": 18654 + }, + { + "epoch": 3.39, + "learning_rate": 1.9840735347651982e-06, + "loss": 1.1471, + "step": 18655 + }, + { + "epoch": 3.39, + "learning_rate": 1.984018929741536e-06, + "loss": 3.3713, + "step": 18656 + }, + { + "epoch": 3.39, + "learning_rate": 1.983964324717874e-06, + "loss": 0.206, + "step": 18657 + }, + { + "epoch": 3.39, + "learning_rate": 1.9839097196942117e-06, + "loss": 0.7963, + "step": 18658 + }, + { + "epoch": 3.39, + "learning_rate": 1.9838551146705496e-06, + "loss": 0.0351, + "step": 18659 + }, + { + "epoch": 3.39, + "learning_rate": 1.9838005096468875e-06, + "loss": 0.5471, + "step": 18660 + }, + { + "epoch": 3.39, + "learning_rate": 1.9837459046232255e-06, + "loss": 0.7493, + "step": 18661 + }, + { + "epoch": 3.39, + "learning_rate": 1.9836912995995634e-06, + "loss": 1.062, + "step": 18662 + }, + { + "epoch": 3.39, + "learning_rate": 1.983636694575901e-06, + "loss": 0.0214, + "step": 18663 + }, + { + "epoch": 3.39, + "learning_rate": 1.983582089552239e-06, + "loss": 0.1337, + "step": 18664 + }, + { + "epoch": 3.39, + "learning_rate": 1.983527484528577e-06, + "loss": 1.2898, + "step": 18665 + }, + { + "epoch": 3.39, + "learning_rate": 1.9834728795049144e-06, + "loss": 0.3809, + "step": 18666 + }, + { + "epoch": 3.39, + "learning_rate": 1.9834182744812523e-06, + "loss": 0.015, + "step": 18667 + }, + { + "epoch": 3.39, + "learning_rate": 1.9833636694575902e-06, + "loss": 0.3737, + "step": 18668 + }, + { + "epoch": 3.39, + "learning_rate": 1.9833090644339278e-06, + "loss": 0.05, + "step": 18669 + }, + { + "epoch": 3.4, + "learning_rate": 1.9832544594102657e-06, + "loss": 0.1087, + "step": 18670 + }, + { + "epoch": 3.4, + "learning_rate": 1.9831998543866036e-06, + "loss": 0.8405, + "step": 18671 + }, + { + "epoch": 3.4, + "learning_rate": 1.9831452493629416e-06, + "loss": 0.712, + "step": 18672 + }, + { + "epoch": 3.4, + "learning_rate": 1.9830906443392795e-06, + "loss": 1.145, + "step": 18673 + }, + { + "epoch": 3.4, + "learning_rate": 1.983036039315617e-06, + "loss": 0.0686, + "step": 18674 + }, + { + "epoch": 3.4, + "learning_rate": 1.982981434291955e-06, + "loss": 1.3427, + "step": 18675 + }, + { + "epoch": 3.4, + "learning_rate": 1.982926829268293e-06, + "loss": 0.7888, + "step": 18676 + }, + { + "epoch": 3.4, + "learning_rate": 1.9828722242446305e-06, + "loss": 0.6609, + "step": 18677 + }, + { + "epoch": 3.4, + "learning_rate": 1.9828176192209684e-06, + "loss": 0.4622, + "step": 18678 + }, + { + "epoch": 3.4, + "learning_rate": 1.9827630141973063e-06, + "loss": 1.2053, + "step": 18679 + }, + { + "epoch": 3.4, + "learning_rate": 1.982708409173644e-06, + "loss": 1.1287, + "step": 18680 + }, + { + "epoch": 3.4, + "learning_rate": 1.982653804149982e-06, + "loss": 1.4365, + "step": 18681 + }, + { + "epoch": 3.4, + "learning_rate": 1.9825991991263198e-06, + "loss": 0.3727, + "step": 18682 + }, + { + "epoch": 3.4, + "learning_rate": 1.9825445941026573e-06, + "loss": 0.6326, + "step": 18683 + }, + { + "epoch": 3.4, + "learning_rate": 1.9824899890789952e-06, + "loss": 1.2082, + "step": 18684 + }, + { + "epoch": 3.4, + "learning_rate": 1.982435384055333e-06, + "loss": 0.2347, + "step": 18685 + }, + { + "epoch": 3.4, + "learning_rate": 1.982380779031671e-06, + "loss": 0.0219, + "step": 18686 + }, + { + "epoch": 3.4, + "learning_rate": 1.982326174008009e-06, + "loss": 0.3595, + "step": 18687 + }, + { + "epoch": 3.4, + "learning_rate": 1.9822715689843466e-06, + "loss": 0.3448, + "step": 18688 + }, + { + "epoch": 3.4, + "learning_rate": 1.9822169639606845e-06, + "loss": 0.2348, + "step": 18689 + }, + { + "epoch": 3.4, + "learning_rate": 1.9821623589370225e-06, + "loss": 0.0352, + "step": 18690 + }, + { + "epoch": 3.4, + "learning_rate": 1.98210775391336e-06, + "loss": 0.3264, + "step": 18691 + }, + { + "epoch": 3.4, + "learning_rate": 1.982053148889698e-06, + "loss": 1.2566, + "step": 18692 + }, + { + "epoch": 3.4, + "learning_rate": 1.981998543866036e-06, + "loss": 0.3104, + "step": 18693 + }, + { + "epoch": 3.4, + "learning_rate": 1.9819439388423734e-06, + "loss": 0.3921, + "step": 18694 + }, + { + "epoch": 3.4, + "learning_rate": 1.9818893338187113e-06, + "loss": 0.1937, + "step": 18695 + }, + { + "epoch": 3.4, + "learning_rate": 1.9818347287950493e-06, + "loss": 0.019, + "step": 18696 + }, + { + "epoch": 3.4, + "learning_rate": 1.981780123771387e-06, + "loss": 1.5723, + "step": 18697 + }, + { + "epoch": 3.4, + "learning_rate": 1.9817255187477247e-06, + "loss": 0.3829, + "step": 18698 + }, + { + "epoch": 3.4, + "learning_rate": 1.9816709137240627e-06, + "loss": 0.9769, + "step": 18699 + }, + { + "epoch": 3.4, + "learning_rate": 1.9816163087004006e-06, + "loss": 1.9558, + "step": 18700 + }, + { + "epoch": 3.4, + "learning_rate": 1.9815617036767386e-06, + "loss": 1.2924, + "step": 18701 + }, + { + "epoch": 3.4, + "learning_rate": 1.981507098653076e-06, + "loss": 0.1696, + "step": 18702 + }, + { + "epoch": 3.4, + "learning_rate": 1.981452493629414e-06, + "loss": 0.0196, + "step": 18703 + }, + { + "epoch": 3.4, + "learning_rate": 1.981397888605752e-06, + "loss": 0.0058, + "step": 18704 + }, + { + "epoch": 3.4, + "learning_rate": 1.9813432835820895e-06, + "loss": 1.9192, + "step": 18705 + }, + { + "epoch": 3.4, + "learning_rate": 1.9812886785584274e-06, + "loss": 0.0343, + "step": 18706 + }, + { + "epoch": 3.4, + "learning_rate": 1.9812340735347654e-06, + "loss": 0.5783, + "step": 18707 + }, + { + "epoch": 3.4, + "learning_rate": 1.981179468511103e-06, + "loss": 1.3289, + "step": 18708 + }, + { + "epoch": 3.4, + "learning_rate": 1.981124863487441e-06, + "loss": 0.5231, + "step": 18709 + }, + { + "epoch": 3.4, + "learning_rate": 1.9810702584637788e-06, + "loss": 0.2299, + "step": 18710 + }, + { + "epoch": 3.4, + "learning_rate": 1.9810156534401163e-06, + "loss": 0.2154, + "step": 18711 + }, + { + "epoch": 3.4, + "learning_rate": 1.9809610484164542e-06, + "loss": 0.0295, + "step": 18712 + }, + { + "epoch": 3.4, + "learning_rate": 1.980906443392792e-06, + "loss": 2.1056, + "step": 18713 + }, + { + "epoch": 3.4, + "learning_rate": 1.98085183836913e-06, + "loss": 0.016, + "step": 18714 + }, + { + "epoch": 3.4, + "learning_rate": 1.980797233345468e-06, + "loss": 0.0227, + "step": 18715 + }, + { + "epoch": 3.4, + "learning_rate": 1.9807426283218056e-06, + "loss": 0.2153, + "step": 18716 + }, + { + "epoch": 3.4, + "learning_rate": 1.9806880232981435e-06, + "loss": 0.1205, + "step": 18717 + }, + { + "epoch": 3.4, + "learning_rate": 1.9806334182744815e-06, + "loss": 1.0692, + "step": 18718 + }, + { + "epoch": 3.4, + "learning_rate": 1.980578813250819e-06, + "loss": 1.1591, + "step": 18719 + }, + { + "epoch": 3.4, + "learning_rate": 1.980524208227157e-06, + "loss": 0.433, + "step": 18720 + }, + { + "epoch": 3.4, + "learning_rate": 1.980469603203495e-06, + "loss": 1.261, + "step": 18721 + }, + { + "epoch": 3.4, + "learning_rate": 1.9804149981798324e-06, + "loss": 0.4755, + "step": 18722 + }, + { + "epoch": 3.4, + "learning_rate": 1.9803603931561704e-06, + "loss": 0.371, + "step": 18723 + }, + { + "epoch": 3.4, + "learning_rate": 1.9803057881325083e-06, + "loss": 1.8766, + "step": 18724 + }, + { + "epoch": 3.41, + "learning_rate": 1.980251183108846e-06, + "loss": 0.2122, + "step": 18725 + }, + { + "epoch": 3.41, + "learning_rate": 1.9801965780851838e-06, + "loss": 1.5699, + "step": 18726 + }, + { + "epoch": 3.41, + "learning_rate": 1.9801419730615217e-06, + "loss": 0.2118, + "step": 18727 + }, + { + "epoch": 3.41, + "learning_rate": 1.9800873680378596e-06, + "loss": 0.7968, + "step": 18728 + }, + { + "epoch": 3.41, + "learning_rate": 1.9800327630141976e-06, + "loss": 0.6117, + "step": 18729 + }, + { + "epoch": 3.41, + "learning_rate": 1.979978157990535e-06, + "loss": 0.3165, + "step": 18730 + }, + { + "epoch": 3.41, + "learning_rate": 1.979923552966873e-06, + "loss": 0.216, + "step": 18731 + }, + { + "epoch": 3.41, + "learning_rate": 1.979868947943211e-06, + "loss": 1.3099, + "step": 18732 + }, + { + "epoch": 3.41, + "learning_rate": 1.9798143429195485e-06, + "loss": 0.75, + "step": 18733 + }, + { + "epoch": 3.41, + "learning_rate": 1.9797597378958865e-06, + "loss": 0.8395, + "step": 18734 + }, + { + "epoch": 3.41, + "learning_rate": 1.9797051328722244e-06, + "loss": 1.8368, + "step": 18735 + }, + { + "epoch": 3.41, + "learning_rate": 1.979650527848562e-06, + "loss": 1.1165, + "step": 18736 + }, + { + "epoch": 3.41, + "learning_rate": 1.9795959228249e-06, + "loss": 0.2871, + "step": 18737 + }, + { + "epoch": 3.41, + "learning_rate": 1.979541317801238e-06, + "loss": 0.1359, + "step": 18738 + }, + { + "epoch": 3.41, + "learning_rate": 1.9794867127775753e-06, + "loss": 0.4957, + "step": 18739 + }, + { + "epoch": 3.41, + "learning_rate": 1.9794321077539133e-06, + "loss": 1.3102, + "step": 18740 + }, + { + "epoch": 3.41, + "learning_rate": 1.9793775027302512e-06, + "loss": 0.6959, + "step": 18741 + }, + { + "epoch": 3.41, + "learning_rate": 1.979322897706589e-06, + "loss": 1.2284, + "step": 18742 + }, + { + "epoch": 3.41, + "learning_rate": 1.979268292682927e-06, + "loss": 0.2356, + "step": 18743 + }, + { + "epoch": 3.41, + "learning_rate": 1.9792136876592646e-06, + "loss": 0.4067, + "step": 18744 + }, + { + "epoch": 3.41, + "learning_rate": 1.9791590826356026e-06, + "loss": 0.6296, + "step": 18745 + }, + { + "epoch": 3.41, + "learning_rate": 1.9791044776119405e-06, + "loss": 0.1987, + "step": 18746 + }, + { + "epoch": 3.41, + "learning_rate": 1.979049872588278e-06, + "loss": 1.4486, + "step": 18747 + }, + { + "epoch": 3.41, + "learning_rate": 1.978995267564616e-06, + "loss": 1.4397, + "step": 18748 + }, + { + "epoch": 3.41, + "learning_rate": 1.978940662540954e-06, + "loss": 0.8165, + "step": 18749 + }, + { + "epoch": 3.41, + "learning_rate": 1.9788860575172914e-06, + "loss": 1.6939, + "step": 18750 + }, + { + "epoch": 3.41, + "learning_rate": 1.9788314524936294e-06, + "loss": 1.8074, + "step": 18751 + }, + { + "epoch": 3.41, + "learning_rate": 1.9787768474699673e-06, + "loss": 1.1631, + "step": 18752 + }, + { + "epoch": 3.41, + "learning_rate": 1.9787222424463053e-06, + "loss": 1.6994, + "step": 18753 + }, + { + "epoch": 3.41, + "learning_rate": 1.978667637422643e-06, + "loss": 0.3147, + "step": 18754 + }, + { + "epoch": 3.41, + "learning_rate": 1.9786130323989807e-06, + "loss": 0.5834, + "step": 18755 + }, + { + "epoch": 3.41, + "learning_rate": 1.9785584273753187e-06, + "loss": 1.6305, + "step": 18756 + }, + { + "epoch": 3.41, + "learning_rate": 1.9785038223516566e-06, + "loss": 0.2016, + "step": 18757 + }, + { + "epoch": 3.41, + "learning_rate": 1.978449217327994e-06, + "loss": 0.2957, + "step": 18758 + }, + { + "epoch": 3.41, + "learning_rate": 1.978394612304332e-06, + "loss": 2.5434, + "step": 18759 + }, + { + "epoch": 3.41, + "learning_rate": 1.97834000728067e-06, + "loss": 1.1183, + "step": 18760 + }, + { + "epoch": 3.41, + "learning_rate": 1.9782854022570075e-06, + "loss": 0.0147, + "step": 18761 + }, + { + "epoch": 3.41, + "learning_rate": 1.9782307972333455e-06, + "loss": 0.0516, + "step": 18762 + }, + { + "epoch": 3.41, + "learning_rate": 1.9781761922096834e-06, + "loss": 0.5558, + "step": 18763 + }, + { + "epoch": 3.41, + "learning_rate": 1.978121587186021e-06, + "loss": 1.4114, + "step": 18764 + }, + { + "epoch": 3.41, + "learning_rate": 1.978066982162359e-06, + "loss": 0.0537, + "step": 18765 + }, + { + "epoch": 3.41, + "learning_rate": 1.978012377138697e-06, + "loss": 0.5091, + "step": 18766 + }, + { + "epoch": 3.41, + "learning_rate": 1.9779577721150348e-06, + "loss": 0.1313, + "step": 18767 + }, + { + "epoch": 3.41, + "learning_rate": 1.9779031670913727e-06, + "loss": 0.8914, + "step": 18768 + }, + { + "epoch": 3.41, + "learning_rate": 1.9778485620677102e-06, + "loss": 0.5168, + "step": 18769 + }, + { + "epoch": 3.41, + "learning_rate": 1.977793957044048e-06, + "loss": 0.0856, + "step": 18770 + }, + { + "epoch": 3.41, + "learning_rate": 1.977739352020386e-06, + "loss": 0.7917, + "step": 18771 + }, + { + "epoch": 3.41, + "learning_rate": 1.9776847469967236e-06, + "loss": 2.7443, + "step": 18772 + }, + { + "epoch": 3.41, + "learning_rate": 1.9776301419730616e-06, + "loss": 0.7149, + "step": 18773 + }, + { + "epoch": 3.41, + "learning_rate": 1.9775755369493995e-06, + "loss": 0.698, + "step": 18774 + }, + { + "epoch": 3.41, + "learning_rate": 1.977520931925737e-06, + "loss": 0.0086, + "step": 18775 + }, + { + "epoch": 3.41, + "learning_rate": 1.977466326902075e-06, + "loss": 0.6568, + "step": 18776 + }, + { + "epoch": 3.41, + "learning_rate": 1.977411721878413e-06, + "loss": 0.4113, + "step": 18777 + }, + { + "epoch": 3.41, + "learning_rate": 1.9773571168547505e-06, + "loss": 0.1023, + "step": 18778 + }, + { + "epoch": 3.41, + "learning_rate": 1.9773025118310884e-06, + "loss": 0.4842, + "step": 18779 + }, + { + "epoch": 3.42, + "learning_rate": 1.9772479068074263e-06, + "loss": 2.8957, + "step": 18780 + }, + { + "epoch": 3.42, + "learning_rate": 1.9771933017837643e-06, + "loss": 1.0582, + "step": 18781 + }, + { + "epoch": 3.42, + "learning_rate": 1.9771386967601022e-06, + "loss": 0.2092, + "step": 18782 + }, + { + "epoch": 3.42, + "learning_rate": 1.9770840917364398e-06, + "loss": 1.7385, + "step": 18783 + }, + { + "epoch": 3.42, + "learning_rate": 1.9770294867127777e-06, + "loss": 0.0873, + "step": 18784 + }, + { + "epoch": 3.42, + "learning_rate": 1.9769748816891156e-06, + "loss": 3.0349, + "step": 18785 + }, + { + "epoch": 3.42, + "learning_rate": 1.976920276665453e-06, + "loss": 0.8198, + "step": 18786 + }, + { + "epoch": 3.42, + "learning_rate": 1.976865671641791e-06, + "loss": 0.0522, + "step": 18787 + }, + { + "epoch": 3.42, + "learning_rate": 1.976811066618129e-06, + "loss": 0.0118, + "step": 18788 + }, + { + "epoch": 3.42, + "learning_rate": 1.9767564615944666e-06, + "loss": 2.412, + "step": 18789 + }, + { + "epoch": 3.42, + "learning_rate": 1.9767018565708045e-06, + "loss": 0.5166, + "step": 18790 + }, + { + "epoch": 3.42, + "learning_rate": 1.9766472515471425e-06, + "loss": 1.4096, + "step": 18791 + }, + { + "epoch": 3.42, + "learning_rate": 1.97659264652348e-06, + "loss": 0.254, + "step": 18792 + }, + { + "epoch": 3.42, + "learning_rate": 1.976538041499818e-06, + "loss": 0.3617, + "step": 18793 + }, + { + "epoch": 3.42, + "learning_rate": 1.976483436476156e-06, + "loss": 0.0612, + "step": 18794 + }, + { + "epoch": 3.42, + "learning_rate": 1.976428831452494e-06, + "loss": 2.8587, + "step": 18795 + }, + { + "epoch": 3.42, + "learning_rate": 1.9763742264288317e-06, + "loss": 2.1842, + "step": 18796 + }, + { + "epoch": 3.42, + "learning_rate": 1.9763196214051693e-06, + "loss": 1.5154, + "step": 18797 + }, + { + "epoch": 3.42, + "learning_rate": 1.9762650163815072e-06, + "loss": 1.1015, + "step": 18798 + }, + { + "epoch": 3.42, + "learning_rate": 1.976210411357845e-06, + "loss": 0.2495, + "step": 18799 + }, + { + "epoch": 3.42, + "learning_rate": 1.9761558063341827e-06, + "loss": 0.3761, + "step": 18800 + }, + { + "epoch": 3.42, + "learning_rate": 1.9761012013105206e-06, + "loss": 3.1605, + "step": 18801 + }, + { + "epoch": 3.42, + "learning_rate": 1.9760465962868586e-06, + "loss": 1.5516, + "step": 18802 + }, + { + "epoch": 3.42, + "learning_rate": 1.975991991263196e-06, + "loss": 1.2097, + "step": 18803 + }, + { + "epoch": 3.42, + "learning_rate": 1.975937386239534e-06, + "loss": 1.6939, + "step": 18804 + }, + { + "epoch": 3.42, + "learning_rate": 1.975882781215872e-06, + "loss": 0.4545, + "step": 18805 + }, + { + "epoch": 3.42, + "learning_rate": 1.9758281761922095e-06, + "loss": 0.9428, + "step": 18806 + }, + { + "epoch": 3.42, + "learning_rate": 1.9757735711685474e-06, + "loss": 0.8371, + "step": 18807 + }, + { + "epoch": 3.42, + "learning_rate": 1.9757189661448854e-06, + "loss": 1.4755, + "step": 18808 + }, + { + "epoch": 3.42, + "learning_rate": 1.9756643611212233e-06, + "loss": 0.4066, + "step": 18809 + }, + { + "epoch": 3.42, + "learning_rate": 1.9756097560975613e-06, + "loss": 1.6659, + "step": 18810 + }, + { + "epoch": 3.42, + "learning_rate": 1.9755551510738988e-06, + "loss": 2.5162, + "step": 18811 + }, + { + "epoch": 3.42, + "learning_rate": 1.9755005460502367e-06, + "loss": 1.159, + "step": 18812 + }, + { + "epoch": 3.42, + "learning_rate": 1.9754459410265747e-06, + "loss": 0.8294, + "step": 18813 + }, + { + "epoch": 3.42, + "learning_rate": 1.975391336002912e-06, + "loss": 1.7664, + "step": 18814 + }, + { + "epoch": 3.42, + "learning_rate": 1.97533673097925e-06, + "loss": 0.5384, + "step": 18815 + }, + { + "epoch": 3.42, + "learning_rate": 1.975282125955588e-06, + "loss": 2.1847, + "step": 18816 + }, + { + "epoch": 3.42, + "learning_rate": 1.9752275209319256e-06, + "loss": 1.1574, + "step": 18817 + }, + { + "epoch": 3.42, + "learning_rate": 1.9751729159082635e-06, + "loss": 1.1223, + "step": 18818 + }, + { + "epoch": 3.42, + "learning_rate": 1.9751183108846015e-06, + "loss": 0.3906, + "step": 18819 + }, + { + "epoch": 3.42, + "learning_rate": 1.975063705860939e-06, + "loss": 0.0571, + "step": 18820 + }, + { + "epoch": 3.42, + "learning_rate": 1.975009100837277e-06, + "loss": 0.7067, + "step": 18821 + }, + { + "epoch": 3.42, + "learning_rate": 1.974954495813615e-06, + "loss": 0.9488, + "step": 18822 + }, + { + "epoch": 3.42, + "learning_rate": 1.974899890789953e-06, + "loss": 0.1641, + "step": 18823 + }, + { + "epoch": 3.42, + "learning_rate": 1.9748452857662908e-06, + "loss": 1.295, + "step": 18824 + }, + { + "epoch": 3.42, + "learning_rate": 1.9747906807426283e-06, + "loss": 0.2858, + "step": 18825 + }, + { + "epoch": 3.42, + "learning_rate": 1.9747360757189662e-06, + "loss": 0.7873, + "step": 18826 + }, + { + "epoch": 3.42, + "learning_rate": 1.974681470695304e-06, + "loss": 1.0248, + "step": 18827 + }, + { + "epoch": 3.42, + "learning_rate": 1.9746268656716417e-06, + "loss": 0.3336, + "step": 18828 + }, + { + "epoch": 3.42, + "learning_rate": 1.9745722606479796e-06, + "loss": 0.5722, + "step": 18829 + }, + { + "epoch": 3.42, + "learning_rate": 1.9745176556243176e-06, + "loss": 2.1124, + "step": 18830 + }, + { + "epoch": 3.42, + "learning_rate": 1.974463050600655e-06, + "loss": 0.016, + "step": 18831 + }, + { + "epoch": 3.42, + "learning_rate": 1.974408445576993e-06, + "loss": 0.3374, + "step": 18832 + }, + { + "epoch": 3.42, + "learning_rate": 1.974353840553331e-06, + "loss": 2.4321, + "step": 18833 + }, + { + "epoch": 3.42, + "learning_rate": 1.974299235529669e-06, + "loss": 1.8424, + "step": 18834 + }, + { + "epoch": 3.43, + "learning_rate": 1.974244630506007e-06, + "loss": 0.0235, + "step": 18835 + }, + { + "epoch": 3.43, + "learning_rate": 1.9741900254823444e-06, + "loss": 1.4643, + "step": 18836 + }, + { + "epoch": 3.43, + "learning_rate": 1.9741354204586823e-06, + "loss": 0.0843, + "step": 18837 + }, + { + "epoch": 3.43, + "learning_rate": 1.9740808154350203e-06, + "loss": 0.0693, + "step": 18838 + }, + { + "epoch": 3.43, + "learning_rate": 1.974026210411358e-06, + "loss": 0.3048, + "step": 18839 + }, + { + "epoch": 3.43, + "learning_rate": 1.9739716053876958e-06, + "loss": 0.3267, + "step": 18840 + }, + { + "epoch": 3.43, + "learning_rate": 1.9739170003640337e-06, + "loss": 0.0288, + "step": 18841 + }, + { + "epoch": 3.43, + "learning_rate": 1.9738623953403712e-06, + "loss": 0.3555, + "step": 18842 + }, + { + "epoch": 3.43, + "learning_rate": 1.973807790316709e-06, + "loss": 1.4363, + "step": 18843 + }, + { + "epoch": 3.43, + "learning_rate": 1.973753185293047e-06, + "loss": 0.3096, + "step": 18844 + }, + { + "epoch": 3.43, + "learning_rate": 1.9736985802693846e-06, + "loss": 0.5909, + "step": 18845 + }, + { + "epoch": 3.43, + "learning_rate": 1.9736439752457226e-06, + "loss": 0.0802, + "step": 18846 + }, + { + "epoch": 3.43, + "learning_rate": 1.9735893702220605e-06, + "loss": 1.9356, + "step": 18847 + }, + { + "epoch": 3.43, + "learning_rate": 1.9735347651983985e-06, + "loss": 1.0333, + "step": 18848 + }, + { + "epoch": 3.43, + "learning_rate": 1.9734801601747364e-06, + "loss": 2.1459, + "step": 18849 + }, + { + "epoch": 3.43, + "learning_rate": 1.973425555151074e-06, + "loss": 0.0574, + "step": 18850 + }, + { + "epoch": 3.43, + "learning_rate": 1.973370950127412e-06, + "loss": 0.1644, + "step": 18851 + }, + { + "epoch": 3.43, + "learning_rate": 1.97331634510375e-06, + "loss": 0.6036, + "step": 18852 + }, + { + "epoch": 3.43, + "learning_rate": 1.9732617400800873e-06, + "loss": 0.4535, + "step": 18853 + }, + { + "epoch": 3.43, + "learning_rate": 1.9732071350564253e-06, + "loss": 0.4456, + "step": 18854 + }, + { + "epoch": 3.43, + "learning_rate": 1.973152530032763e-06, + "loss": 0.749, + "step": 18855 + }, + { + "epoch": 3.43, + "learning_rate": 1.9730979250091007e-06, + "loss": 1.0764, + "step": 18856 + }, + { + "epoch": 3.43, + "learning_rate": 1.9730433199854387e-06, + "loss": 0.0305, + "step": 18857 + }, + { + "epoch": 3.43, + "learning_rate": 1.9729887149617766e-06, + "loss": 0.0473, + "step": 18858 + }, + { + "epoch": 3.43, + "learning_rate": 1.972934109938114e-06, + "loss": 0.2183, + "step": 18859 + }, + { + "epoch": 3.43, + "learning_rate": 1.972879504914452e-06, + "loss": 0.5996, + "step": 18860 + }, + { + "epoch": 3.43, + "learning_rate": 1.97282489989079e-06, + "loss": 0.6936, + "step": 18861 + }, + { + "epoch": 3.43, + "learning_rate": 1.972770294867128e-06, + "loss": 0.0127, + "step": 18862 + }, + { + "epoch": 3.43, + "learning_rate": 1.972715689843466e-06, + "loss": 1.2171, + "step": 18863 + }, + { + "epoch": 3.43, + "learning_rate": 1.9726610848198034e-06, + "loss": 0.686, + "step": 18864 + }, + { + "epoch": 3.43, + "learning_rate": 1.9726064797961414e-06, + "loss": 0.0193, + "step": 18865 + }, + { + "epoch": 3.43, + "learning_rate": 1.9725518747724793e-06, + "loss": 1.1258, + "step": 18866 + }, + { + "epoch": 3.43, + "learning_rate": 1.972497269748817e-06, + "loss": 0.0096, + "step": 18867 + }, + { + "epoch": 3.43, + "learning_rate": 1.9724426647251548e-06, + "loss": 0.66, + "step": 18868 + }, + { + "epoch": 3.43, + "learning_rate": 1.9723880597014927e-06, + "loss": 0.3705, + "step": 18869 + }, + { + "epoch": 3.43, + "learning_rate": 1.9723334546778302e-06, + "loss": 0.014, + "step": 18870 + }, + { + "epoch": 3.43, + "learning_rate": 1.972278849654168e-06, + "loss": 0.7791, + "step": 18871 + }, + { + "epoch": 3.43, + "learning_rate": 1.972224244630506e-06, + "loss": 1.4276, + "step": 18872 + }, + { + "epoch": 3.43, + "learning_rate": 1.9721696396068437e-06, + "loss": 3.2214, + "step": 18873 + }, + { + "epoch": 3.43, + "learning_rate": 1.9721150345831816e-06, + "loss": 2.462, + "step": 18874 + }, + { + "epoch": 3.43, + "learning_rate": 1.9720604295595195e-06, + "loss": 0.7707, + "step": 18875 + }, + { + "epoch": 3.43, + "learning_rate": 1.9720058245358575e-06, + "loss": 0.0443, + "step": 18876 + }, + { + "epoch": 3.43, + "learning_rate": 1.9719512195121954e-06, + "loss": 1.1349, + "step": 18877 + }, + { + "epoch": 3.43, + "learning_rate": 1.971896614488533e-06, + "loss": 0.1029, + "step": 18878 + }, + { + "epoch": 3.43, + "learning_rate": 1.971842009464871e-06, + "loss": 0.3712, + "step": 18879 + }, + { + "epoch": 3.43, + "learning_rate": 1.971787404441209e-06, + "loss": 1.0703, + "step": 18880 + }, + { + "epoch": 3.43, + "learning_rate": 1.9717327994175464e-06, + "loss": 0.0096, + "step": 18881 + }, + { + "epoch": 3.43, + "learning_rate": 1.9716781943938843e-06, + "loss": 1.1285, + "step": 18882 + }, + { + "epoch": 3.43, + "learning_rate": 1.9716235893702222e-06, + "loss": 0.4699, + "step": 18883 + }, + { + "epoch": 3.43, + "learning_rate": 1.9715689843465598e-06, + "loss": 0.0383, + "step": 18884 + }, + { + "epoch": 3.43, + "learning_rate": 1.9715143793228977e-06, + "loss": 0.244, + "step": 18885 + }, + { + "epoch": 3.43, + "learning_rate": 1.9714597742992356e-06, + "loss": 2.1889, + "step": 18886 + }, + { + "epoch": 3.43, + "learning_rate": 1.971405169275573e-06, + "loss": 0.0304, + "step": 18887 + }, + { + "epoch": 3.43, + "learning_rate": 1.971350564251911e-06, + "loss": 1.352, + "step": 18888 + }, + { + "epoch": 3.43, + "learning_rate": 1.971295959228249e-06, + "loss": 0.4115, + "step": 18889 + }, + { + "epoch": 3.44, + "learning_rate": 1.971241354204587e-06, + "loss": 0.8883, + "step": 18890 + }, + { + "epoch": 3.44, + "learning_rate": 1.971186749180925e-06, + "loss": 0.3892, + "step": 18891 + }, + { + "epoch": 3.44, + "learning_rate": 1.9711321441572625e-06, + "loss": 0.683, + "step": 18892 + }, + { + "epoch": 3.44, + "learning_rate": 1.9710775391336004e-06, + "loss": 1.1458, + "step": 18893 + }, + { + "epoch": 3.44, + "learning_rate": 1.9710229341099383e-06, + "loss": 0.0287, + "step": 18894 + }, + { + "epoch": 3.44, + "learning_rate": 1.970968329086276e-06, + "loss": 0.9611, + "step": 18895 + }, + { + "epoch": 3.44, + "learning_rate": 1.970913724062614e-06, + "loss": 1.056, + "step": 18896 + }, + { + "epoch": 3.44, + "learning_rate": 1.9708591190389518e-06, + "loss": 0.2523, + "step": 18897 + }, + { + "epoch": 3.44, + "learning_rate": 1.9708045140152893e-06, + "loss": 0.9712, + "step": 18898 + }, + { + "epoch": 3.44, + "learning_rate": 1.9707499089916272e-06, + "loss": 0.6769, + "step": 18899 + }, + { + "epoch": 3.44, + "learning_rate": 1.970695303967965e-06, + "loss": 1.1339, + "step": 18900 + }, + { + "epoch": 3.44, + "learning_rate": 1.9706406989443027e-06, + "loss": 0.1663, + "step": 18901 + }, + { + "epoch": 3.44, + "learning_rate": 1.9705860939206406e-06, + "loss": 1.0104, + "step": 18902 + }, + { + "epoch": 3.44, + "learning_rate": 1.9705314888969786e-06, + "loss": 0.076, + "step": 18903 + }, + { + "epoch": 3.44, + "learning_rate": 1.9704768838733165e-06, + "loss": 0.035, + "step": 18904 + }, + { + "epoch": 3.44, + "learning_rate": 1.9704222788496545e-06, + "loss": 2.2684, + "step": 18905 + }, + { + "epoch": 3.44, + "learning_rate": 1.970367673825992e-06, + "loss": 0.3748, + "step": 18906 + }, + { + "epoch": 3.44, + "learning_rate": 1.97031306880233e-06, + "loss": 0.967, + "step": 18907 + }, + { + "epoch": 3.44, + "learning_rate": 1.970258463778668e-06, + "loss": 0.2642, + "step": 18908 + }, + { + "epoch": 3.44, + "learning_rate": 1.9702038587550054e-06, + "loss": 1.1026, + "step": 18909 + }, + { + "epoch": 3.44, + "learning_rate": 1.9701492537313433e-06, + "loss": 0.7793, + "step": 18910 + }, + { + "epoch": 3.44, + "learning_rate": 1.9700946487076813e-06, + "loss": 0.2345, + "step": 18911 + }, + { + "epoch": 3.44, + "learning_rate": 1.9700400436840188e-06, + "loss": 0.0391, + "step": 18912 + }, + { + "epoch": 3.44, + "learning_rate": 1.9699854386603567e-06, + "loss": 0.3742, + "step": 18913 + }, + { + "epoch": 3.44, + "learning_rate": 1.9699308336366947e-06, + "loss": 0.3142, + "step": 18914 + }, + { + "epoch": 3.44, + "learning_rate": 1.9698762286130326e-06, + "loss": 1.6209, + "step": 18915 + }, + { + "epoch": 3.44, + "learning_rate": 1.9698216235893706e-06, + "loss": 1.9581, + "step": 18916 + }, + { + "epoch": 3.44, + "learning_rate": 1.969767018565708e-06, + "loss": 0.173, + "step": 18917 + }, + { + "epoch": 3.44, + "learning_rate": 1.969712413542046e-06, + "loss": 1.5319, + "step": 18918 + }, + { + "epoch": 3.44, + "learning_rate": 1.969657808518384e-06, + "loss": 0.8574, + "step": 18919 + }, + { + "epoch": 3.44, + "learning_rate": 1.9696032034947215e-06, + "loss": 0.8615, + "step": 18920 + }, + { + "epoch": 3.44, + "learning_rate": 1.9695485984710594e-06, + "loss": 0.2174, + "step": 18921 + }, + { + "epoch": 3.44, + "learning_rate": 1.9694939934473974e-06, + "loss": 0.0145, + "step": 18922 + }, + { + "epoch": 3.44, + "learning_rate": 1.969439388423735e-06, + "loss": 0.7436, + "step": 18923 + }, + { + "epoch": 3.44, + "learning_rate": 1.969384783400073e-06, + "loss": 0.4534, + "step": 18924 + }, + { + "epoch": 3.44, + "learning_rate": 1.9693301783764108e-06, + "loss": 0.585, + "step": 18925 + }, + { + "epoch": 3.44, + "learning_rate": 1.9692755733527483e-06, + "loss": 0.0628, + "step": 18926 + }, + { + "epoch": 3.44, + "learning_rate": 1.9692209683290862e-06, + "loss": 0.9435, + "step": 18927 + }, + { + "epoch": 3.44, + "learning_rate": 1.969166363305424e-06, + "loss": 0.2143, + "step": 18928 + }, + { + "epoch": 3.44, + "learning_rate": 1.969111758281762e-06, + "loss": 3.5092, + "step": 18929 + }, + { + "epoch": 3.44, + "learning_rate": 1.9690571532581e-06, + "loss": 0.4098, + "step": 18930 + }, + { + "epoch": 3.44, + "learning_rate": 1.9690025482344376e-06, + "loss": 0.7184, + "step": 18931 + }, + { + "epoch": 3.44, + "learning_rate": 1.9689479432107755e-06, + "loss": 2.3039, + "step": 18932 + }, + { + "epoch": 3.44, + "learning_rate": 1.9688933381871135e-06, + "loss": 0.1912, + "step": 18933 + }, + { + "epoch": 3.44, + "learning_rate": 1.968838733163451e-06, + "loss": 0.12, + "step": 18934 + }, + { + "epoch": 3.44, + "learning_rate": 1.968784128139789e-06, + "loss": 0.022, + "step": 18935 + }, + { + "epoch": 3.44, + "learning_rate": 1.968729523116127e-06, + "loss": 2.2791, + "step": 18936 + }, + { + "epoch": 3.44, + "learning_rate": 1.9686749180924644e-06, + "loss": 0.0227, + "step": 18937 + }, + { + "epoch": 3.44, + "learning_rate": 1.9686203130688023e-06, + "loss": 1.1884, + "step": 18938 + }, + { + "epoch": 3.44, + "learning_rate": 1.9685657080451403e-06, + "loss": 0.3674, + "step": 18939 + }, + { + "epoch": 3.44, + "learning_rate": 1.968511103021478e-06, + "loss": 0.7075, + "step": 18940 + }, + { + "epoch": 3.44, + "learning_rate": 1.9684564979978158e-06, + "loss": 1.3168, + "step": 18941 + }, + { + "epoch": 3.44, + "learning_rate": 1.9684018929741537e-06, + "loss": 2.6043, + "step": 18942 + }, + { + "epoch": 3.44, + "learning_rate": 1.9683472879504916e-06, + "loss": 0.3376, + "step": 18943 + }, + { + "epoch": 3.44, + "learning_rate": 1.9682926829268296e-06, + "loss": 1.0726, + "step": 18944 + }, + { + "epoch": 3.45, + "learning_rate": 1.968238077903167e-06, + "loss": 0.0954, + "step": 18945 + }, + { + "epoch": 3.45, + "learning_rate": 1.968183472879505e-06, + "loss": 0.8743, + "step": 18946 + }, + { + "epoch": 3.45, + "learning_rate": 1.968128867855843e-06, + "loss": 0.6767, + "step": 18947 + }, + { + "epoch": 3.45, + "learning_rate": 1.9680742628321805e-06, + "loss": 0.6541, + "step": 18948 + }, + { + "epoch": 3.45, + "learning_rate": 1.9680196578085185e-06, + "loss": 0.0603, + "step": 18949 + }, + { + "epoch": 3.45, + "learning_rate": 1.9679650527848564e-06, + "loss": 0.1037, + "step": 18950 + }, + { + "epoch": 3.45, + "learning_rate": 1.967910447761194e-06, + "loss": 3.9133, + "step": 18951 + }, + { + "epoch": 3.45, + "learning_rate": 1.967855842737532e-06, + "loss": 0.3463, + "step": 18952 + }, + { + "epoch": 3.45, + "learning_rate": 1.96780123771387e-06, + "loss": 0.4621, + "step": 18953 + }, + { + "epoch": 3.45, + "learning_rate": 1.9677466326902073e-06, + "loss": 0.3461, + "step": 18954 + }, + { + "epoch": 3.45, + "learning_rate": 1.9676920276665453e-06, + "loss": 1.3135, + "step": 18955 + }, + { + "epoch": 3.45, + "learning_rate": 1.967637422642883e-06, + "loss": 1.3847, + "step": 18956 + }, + { + "epoch": 3.45, + "learning_rate": 1.967582817619221e-06, + "loss": 2.8266, + "step": 18957 + }, + { + "epoch": 3.45, + "learning_rate": 1.967528212595559e-06, + "loss": 0.0102, + "step": 18958 + }, + { + "epoch": 3.45, + "learning_rate": 1.9674736075718966e-06, + "loss": 0.0188, + "step": 18959 + }, + { + "epoch": 3.45, + "learning_rate": 1.9674190025482346e-06, + "loss": 2.4183, + "step": 18960 + }, + { + "epoch": 3.45, + "learning_rate": 1.9673643975245725e-06, + "loss": 1.1308, + "step": 18961 + }, + { + "epoch": 3.45, + "learning_rate": 1.96730979250091e-06, + "loss": 1.0604, + "step": 18962 + }, + { + "epoch": 3.45, + "learning_rate": 1.967255187477248e-06, + "loss": 0.2209, + "step": 18963 + }, + { + "epoch": 3.45, + "learning_rate": 1.967200582453586e-06, + "loss": 0.9459, + "step": 18964 + }, + { + "epoch": 3.45, + "learning_rate": 1.9671459774299234e-06, + "loss": 3.0314, + "step": 18965 + }, + { + "epoch": 3.45, + "learning_rate": 1.9670913724062614e-06, + "loss": 0.7018, + "step": 18966 + }, + { + "epoch": 3.45, + "learning_rate": 1.9670367673825993e-06, + "loss": 0.0143, + "step": 18967 + }, + { + "epoch": 3.45, + "learning_rate": 1.966982162358937e-06, + "loss": 0.3257, + "step": 18968 + }, + { + "epoch": 3.45, + "learning_rate": 1.9669275573352748e-06, + "loss": 0.0075, + "step": 18969 + }, + { + "epoch": 3.45, + "learning_rate": 1.9668729523116127e-06, + "loss": 3.1827, + "step": 18970 + }, + { + "epoch": 3.45, + "learning_rate": 1.9668183472879507e-06, + "loss": 0.0295, + "step": 18971 + }, + { + "epoch": 3.45, + "learning_rate": 1.9667637422642886e-06, + "loss": 0.3681, + "step": 18972 + }, + { + "epoch": 3.45, + "learning_rate": 1.966709137240626e-06, + "loss": 0.2241, + "step": 18973 + }, + { + "epoch": 3.45, + "learning_rate": 1.966654532216964e-06, + "loss": 0.0082, + "step": 18974 + }, + { + "epoch": 3.45, + "learning_rate": 1.966599927193302e-06, + "loss": 0.197, + "step": 18975 + }, + { + "epoch": 3.45, + "learning_rate": 1.9665453221696395e-06, + "loss": 0.8174, + "step": 18976 + }, + { + "epoch": 3.45, + "learning_rate": 1.9664907171459775e-06, + "loss": 0.2165, + "step": 18977 + }, + { + "epoch": 3.45, + "learning_rate": 1.9664361121223154e-06, + "loss": 0.2925, + "step": 18978 + }, + { + "epoch": 3.45, + "learning_rate": 1.966381507098653e-06, + "loss": 0.0237, + "step": 18979 + }, + { + "epoch": 3.45, + "learning_rate": 1.966326902074991e-06, + "loss": 0.1272, + "step": 18980 + }, + { + "epoch": 3.45, + "learning_rate": 1.966272297051329e-06, + "loss": 0.335, + "step": 18981 + }, + { + "epoch": 3.45, + "learning_rate": 1.9662176920276664e-06, + "loss": 0.7405, + "step": 18982 + }, + { + "epoch": 3.45, + "learning_rate": 1.9661630870040043e-06, + "loss": 1.2868, + "step": 18983 + }, + { + "epoch": 3.45, + "learning_rate": 1.9661084819803422e-06, + "loss": 0.5727, + "step": 18984 + }, + { + "epoch": 3.45, + "learning_rate": 1.96605387695668e-06, + "loss": 0.7864, + "step": 18985 + }, + { + "epoch": 3.45, + "learning_rate": 1.965999271933018e-06, + "loss": 0.52, + "step": 18986 + }, + { + "epoch": 3.45, + "learning_rate": 1.9659446669093556e-06, + "loss": 0.1469, + "step": 18987 + }, + { + "epoch": 3.45, + "learning_rate": 1.9658900618856936e-06, + "loss": 0.2777, + "step": 18988 + }, + { + "epoch": 3.45, + "learning_rate": 1.9658354568620315e-06, + "loss": 0.1117, + "step": 18989 + }, + { + "epoch": 3.45, + "learning_rate": 1.965780851838369e-06, + "loss": 0.8107, + "step": 18990 + }, + { + "epoch": 3.45, + "learning_rate": 1.965726246814707e-06, + "loss": 0.4794, + "step": 18991 + }, + { + "epoch": 3.45, + "learning_rate": 1.965671641791045e-06, + "loss": 0.6328, + "step": 18992 + }, + { + "epoch": 3.45, + "learning_rate": 1.9656170367673825e-06, + "loss": 0.1378, + "step": 18993 + }, + { + "epoch": 3.45, + "learning_rate": 1.9655624317437204e-06, + "loss": 0.0399, + "step": 18994 + }, + { + "epoch": 3.45, + "learning_rate": 1.965507826720058e-06, + "loss": 1.4724, + "step": 18995 + }, + { + "epoch": 3.45, + "learning_rate": 1.965453221696396e-06, + "loss": 0.1751, + "step": 18996 + }, + { + "epoch": 3.45, + "learning_rate": 1.965398616672734e-06, + "loss": 1.1611, + "step": 18997 + }, + { + "epoch": 3.45, + "learning_rate": 1.9653440116490718e-06, + "loss": 0.5489, + "step": 18998 + }, + { + "epoch": 3.45, + "learning_rate": 1.9652894066254097e-06, + "loss": 0.0289, + "step": 18999 + }, + { + "epoch": 3.46, + "learning_rate": 1.9652348016017476e-06, + "loss": 1.8726, + "step": 19000 + }, + { + "epoch": 3.46, + "learning_rate": 1.965180196578085e-06, + "loss": 0.849, + "step": 19001 + }, + { + "epoch": 3.46, + "learning_rate": 1.965125591554423e-06, + "loss": 0.1001, + "step": 19002 + }, + { + "epoch": 3.46, + "learning_rate": 1.965070986530761e-06, + "loss": 0.6029, + "step": 19003 + }, + { + "epoch": 3.46, + "learning_rate": 1.9650163815070986e-06, + "loss": 0.9601, + "step": 19004 + }, + { + "epoch": 3.46, + "learning_rate": 1.9649617764834365e-06, + "loss": 0.3971, + "step": 19005 + }, + { + "epoch": 3.46, + "learning_rate": 1.9649071714597745e-06, + "loss": 0.0425, + "step": 19006 + }, + { + "epoch": 3.46, + "learning_rate": 1.964852566436112e-06, + "loss": 0.4362, + "step": 19007 + }, + { + "epoch": 3.46, + "learning_rate": 1.96479796141245e-06, + "loss": 0.2064, + "step": 19008 + }, + { + "epoch": 3.46, + "learning_rate": 1.964743356388788e-06, + "loss": 0.927, + "step": 19009 + }, + { + "epoch": 3.46, + "learning_rate": 1.964688751365126e-06, + "loss": 1.4878, + "step": 19010 + }, + { + "epoch": 3.46, + "learning_rate": 1.9646341463414637e-06, + "loss": 0.6689, + "step": 19011 + }, + { + "epoch": 3.46, + "learning_rate": 1.9645795413178013e-06, + "loss": 0.0274, + "step": 19012 + }, + { + "epoch": 3.46, + "learning_rate": 1.964524936294139e-06, + "loss": 0.2443, + "step": 19013 + }, + { + "epoch": 3.46, + "learning_rate": 1.964470331270477e-06, + "loss": 1.028, + "step": 19014 + }, + { + "epoch": 3.46, + "learning_rate": 1.9644157262468147e-06, + "loss": 0.2616, + "step": 19015 + }, + { + "epoch": 3.46, + "learning_rate": 1.9643611212231526e-06, + "loss": 0.0174, + "step": 19016 + }, + { + "epoch": 3.46, + "learning_rate": 1.9643065161994906e-06, + "loss": 0.2901, + "step": 19017 + }, + { + "epoch": 3.46, + "learning_rate": 1.964251911175828e-06, + "loss": 0.1511, + "step": 19018 + }, + { + "epoch": 3.46, + "learning_rate": 1.964197306152166e-06, + "loss": 0.4023, + "step": 19019 + }, + { + "epoch": 3.46, + "learning_rate": 1.964142701128504e-06, + "loss": 1.6174, + "step": 19020 + }, + { + "epoch": 3.46, + "learning_rate": 1.9640880961048415e-06, + "loss": 1.1336, + "step": 19021 + }, + { + "epoch": 3.46, + "learning_rate": 1.9640334910811794e-06, + "loss": 0.7721, + "step": 19022 + }, + { + "epoch": 3.46, + "learning_rate": 1.9639788860575174e-06, + "loss": 1.6071, + "step": 19023 + }, + { + "epoch": 3.46, + "learning_rate": 1.9639242810338553e-06, + "loss": 2.043, + "step": 19024 + }, + { + "epoch": 3.46, + "learning_rate": 1.9638696760101933e-06, + "loss": 0.5048, + "step": 19025 + }, + { + "epoch": 3.46, + "learning_rate": 1.9638150709865308e-06, + "loss": 0.7579, + "step": 19026 + }, + { + "epoch": 3.46, + "learning_rate": 1.9637604659628687e-06, + "loss": 0.8975, + "step": 19027 + }, + { + "epoch": 3.46, + "learning_rate": 1.9637058609392067e-06, + "loss": 3.0397, + "step": 19028 + }, + { + "epoch": 3.46, + "learning_rate": 1.963651255915544e-06, + "loss": 0.0377, + "step": 19029 + }, + { + "epoch": 3.46, + "learning_rate": 1.963596650891882e-06, + "loss": 0.2183, + "step": 19030 + }, + { + "epoch": 3.46, + "learning_rate": 1.96354204586822e-06, + "loss": 0.0139, + "step": 19031 + }, + { + "epoch": 3.46, + "learning_rate": 1.9634874408445576e-06, + "loss": 1.2174, + "step": 19032 + }, + { + "epoch": 3.46, + "learning_rate": 1.9634328358208955e-06, + "loss": 0.9214, + "step": 19033 + }, + { + "epoch": 3.46, + "learning_rate": 1.9633782307972335e-06, + "loss": 0.4744, + "step": 19034 + }, + { + "epoch": 3.46, + "learning_rate": 1.963323625773571e-06, + "loss": 0.3704, + "step": 19035 + }, + { + "epoch": 3.46, + "learning_rate": 1.963269020749909e-06, + "loss": 0.5072, + "step": 19036 + }, + { + "epoch": 3.46, + "learning_rate": 1.963214415726247e-06, + "loss": 0.2324, + "step": 19037 + }, + { + "epoch": 3.46, + "learning_rate": 1.963159810702585e-06, + "loss": 0.8233, + "step": 19038 + }, + { + "epoch": 3.46, + "learning_rate": 1.9631052056789228e-06, + "loss": 0.1231, + "step": 19039 + }, + { + "epoch": 3.46, + "learning_rate": 1.9630506006552603e-06, + "loss": 0.598, + "step": 19040 + }, + { + "epoch": 3.46, + "learning_rate": 1.9629959956315982e-06, + "loss": 1.2882, + "step": 19041 + }, + { + "epoch": 3.46, + "learning_rate": 1.962941390607936e-06, + "loss": 0.5724, + "step": 19042 + }, + { + "epoch": 3.46, + "learning_rate": 1.9628867855842737e-06, + "loss": 0.0385, + "step": 19043 + }, + { + "epoch": 3.46, + "learning_rate": 1.9628321805606116e-06, + "loss": 0.3037, + "step": 19044 + }, + { + "epoch": 3.46, + "learning_rate": 1.9627775755369496e-06, + "loss": 0.0251, + "step": 19045 + }, + { + "epoch": 3.46, + "learning_rate": 1.962722970513287e-06, + "loss": 0.7705, + "step": 19046 + }, + { + "epoch": 3.46, + "learning_rate": 1.962668365489625e-06, + "loss": 0.0195, + "step": 19047 + }, + { + "epoch": 3.46, + "learning_rate": 1.962613760465963e-06, + "loss": 0.1724, + "step": 19048 + }, + { + "epoch": 3.46, + "learning_rate": 1.9625591554423005e-06, + "loss": 0.2677, + "step": 19049 + }, + { + "epoch": 3.46, + "learning_rate": 1.9625045504186385e-06, + "loss": 0.8166, + "step": 19050 + }, + { + "epoch": 3.46, + "learning_rate": 1.9624499453949764e-06, + "loss": 0.2053, + "step": 19051 + }, + { + "epoch": 3.46, + "learning_rate": 1.9623953403713143e-06, + "loss": 1.8762, + "step": 19052 + }, + { + "epoch": 3.46, + "learning_rate": 1.9623407353476523e-06, + "loss": 0.6793, + "step": 19053 + }, + { + "epoch": 3.46, + "learning_rate": 1.96228613032399e-06, + "loss": 0.6397, + "step": 19054 + }, + { + "epoch": 3.47, + "learning_rate": 1.9622315253003277e-06, + "loss": 0.6453, + "step": 19055 + }, + { + "epoch": 3.47, + "learning_rate": 1.9621769202766657e-06, + "loss": 1.0229, + "step": 19056 + }, + { + "epoch": 3.47, + "learning_rate": 1.9621223152530032e-06, + "loss": 1.279, + "step": 19057 + }, + { + "epoch": 3.47, + "learning_rate": 1.962067710229341e-06, + "loss": 0.6418, + "step": 19058 + }, + { + "epoch": 3.47, + "learning_rate": 1.962013105205679e-06, + "loss": 0.1475, + "step": 19059 + }, + { + "epoch": 3.47, + "learning_rate": 1.9619585001820166e-06, + "loss": 1.8407, + "step": 19060 + }, + { + "epoch": 3.47, + "learning_rate": 1.9619038951583546e-06, + "loss": 0.6933, + "step": 19061 + }, + { + "epoch": 3.47, + "learning_rate": 1.961849290134692e-06, + "loss": 1.988, + "step": 19062 + }, + { + "epoch": 3.47, + "learning_rate": 1.96179468511103e-06, + "loss": 0.4376, + "step": 19063 + }, + { + "epoch": 3.47, + "learning_rate": 1.961740080087368e-06, + "loss": 0.0945, + "step": 19064 + }, + { + "epoch": 3.47, + "learning_rate": 1.961685475063706e-06, + "loss": 0.607, + "step": 19065 + }, + { + "epoch": 3.47, + "learning_rate": 1.961630870040044e-06, + "loss": 1.006, + "step": 19066 + }, + { + "epoch": 3.47, + "learning_rate": 1.961576265016382e-06, + "loss": 0.3324, + "step": 19067 + }, + { + "epoch": 3.47, + "learning_rate": 1.9615216599927193e-06, + "loss": 1.8821, + "step": 19068 + }, + { + "epoch": 3.47, + "learning_rate": 1.9614670549690573e-06, + "loss": 0.0242, + "step": 19069 + }, + { + "epoch": 3.47, + "learning_rate": 1.961412449945395e-06, + "loss": 0.6271, + "step": 19070 + }, + { + "epoch": 3.47, + "learning_rate": 1.9613578449217327e-06, + "loss": 0.0227, + "step": 19071 + }, + { + "epoch": 3.47, + "learning_rate": 1.9613032398980707e-06, + "loss": 2.1058, + "step": 19072 + }, + { + "epoch": 3.47, + "learning_rate": 1.9612486348744086e-06, + "loss": 0.5661, + "step": 19073 + }, + { + "epoch": 3.47, + "learning_rate": 1.961194029850746e-06, + "loss": 0.6988, + "step": 19074 + }, + { + "epoch": 3.47, + "learning_rate": 1.961139424827084e-06, + "loss": 0.0057, + "step": 19075 + }, + { + "epoch": 3.47, + "learning_rate": 1.9610848198034216e-06, + "loss": 0.6478, + "step": 19076 + }, + { + "epoch": 3.47, + "learning_rate": 1.9610302147797595e-06, + "loss": 0.3804, + "step": 19077 + }, + { + "epoch": 3.47, + "learning_rate": 1.9609756097560975e-06, + "loss": 1.411, + "step": 19078 + }, + { + "epoch": 3.47, + "learning_rate": 1.9609210047324354e-06, + "loss": 1.4006, + "step": 19079 + }, + { + "epoch": 3.47, + "learning_rate": 1.9608663997087734e-06, + "loss": 0.2596, + "step": 19080 + }, + { + "epoch": 3.47, + "learning_rate": 1.9608117946851113e-06, + "loss": 0.0709, + "step": 19081 + }, + { + "epoch": 3.47, + "learning_rate": 1.960757189661449e-06, + "loss": 2.2046, + "step": 19082 + }, + { + "epoch": 3.47, + "learning_rate": 1.9607025846377868e-06, + "loss": 1.7809, + "step": 19083 + }, + { + "epoch": 3.47, + "learning_rate": 1.9606479796141247e-06, + "loss": 0.9614, + "step": 19084 + }, + { + "epoch": 3.47, + "learning_rate": 1.9605933745904622e-06, + "loss": 1.0701, + "step": 19085 + }, + { + "epoch": 3.47, + "learning_rate": 1.9605387695668e-06, + "loss": 1.0945, + "step": 19086 + }, + { + "epoch": 3.47, + "learning_rate": 1.960484164543138e-06, + "loss": 0.5264, + "step": 19087 + }, + { + "epoch": 3.47, + "learning_rate": 1.9604295595194756e-06, + "loss": 0.5883, + "step": 19088 + }, + { + "epoch": 3.47, + "learning_rate": 1.9603749544958136e-06, + "loss": 0.2637, + "step": 19089 + }, + { + "epoch": 3.47, + "learning_rate": 1.9603203494721515e-06, + "loss": 0.5322, + "step": 19090 + }, + { + "epoch": 3.47, + "learning_rate": 1.9602657444484895e-06, + "loss": 0.1106, + "step": 19091 + }, + { + "epoch": 3.47, + "learning_rate": 1.9602111394248274e-06, + "loss": 1.0104, + "step": 19092 + }, + { + "epoch": 3.47, + "learning_rate": 1.960156534401165e-06, + "loss": 0.2575, + "step": 19093 + }, + { + "epoch": 3.47, + "learning_rate": 1.960101929377503e-06, + "loss": 0.51, + "step": 19094 + }, + { + "epoch": 3.47, + "learning_rate": 1.960047324353841e-06, + "loss": 0.3365, + "step": 19095 + }, + { + "epoch": 3.47, + "learning_rate": 1.9599927193301783e-06, + "loss": 0.9387, + "step": 19096 + }, + { + "epoch": 3.47, + "learning_rate": 1.9599381143065163e-06, + "loss": 0.7513, + "step": 19097 + }, + { + "epoch": 3.47, + "learning_rate": 1.9598835092828542e-06, + "loss": 0.2196, + "step": 19098 + }, + { + "epoch": 3.47, + "learning_rate": 1.9598289042591918e-06, + "loss": 0.0138, + "step": 19099 + }, + { + "epoch": 3.47, + "learning_rate": 1.9597742992355297e-06, + "loss": 2.0327, + "step": 19100 + }, + { + "epoch": 3.47, + "learning_rate": 1.9597196942118676e-06, + "loss": 1.7727, + "step": 19101 + }, + { + "epoch": 3.47, + "learning_rate": 1.959665089188205e-06, + "loss": 0.2933, + "step": 19102 + }, + { + "epoch": 3.47, + "learning_rate": 1.959610484164543e-06, + "loss": 0.5352, + "step": 19103 + }, + { + "epoch": 3.47, + "learning_rate": 1.959555879140881e-06, + "loss": 1.3769, + "step": 19104 + }, + { + "epoch": 3.47, + "learning_rate": 1.959501274117219e-06, + "loss": 0.2314, + "step": 19105 + }, + { + "epoch": 3.47, + "learning_rate": 1.959446669093557e-06, + "loss": 0.512, + "step": 19106 + }, + { + "epoch": 3.47, + "learning_rate": 1.9593920640698945e-06, + "loss": 0.4656, + "step": 19107 + }, + { + "epoch": 3.47, + "learning_rate": 1.9593374590462324e-06, + "loss": 0.8474, + "step": 19108 + }, + { + "epoch": 3.47, + "learning_rate": 1.9592828540225703e-06, + "loss": 1.4431, + "step": 19109 + }, + { + "epoch": 3.48, + "learning_rate": 1.959228248998908e-06, + "loss": 0.3575, + "step": 19110 + }, + { + "epoch": 3.48, + "learning_rate": 1.959173643975246e-06, + "loss": 1.7694, + "step": 19111 + }, + { + "epoch": 3.48, + "learning_rate": 1.9591190389515837e-06, + "loss": 0.0109, + "step": 19112 + }, + { + "epoch": 3.48, + "learning_rate": 1.9590644339279213e-06, + "loss": 2.0762, + "step": 19113 + }, + { + "epoch": 3.48, + "learning_rate": 1.959009828904259e-06, + "loss": 2.3546, + "step": 19114 + }, + { + "epoch": 3.48, + "learning_rate": 1.958955223880597e-06, + "loss": 0.02, + "step": 19115 + }, + { + "epoch": 3.48, + "learning_rate": 1.9589006188569347e-06, + "loss": 1.7187, + "step": 19116 + }, + { + "epoch": 3.48, + "learning_rate": 1.9588460138332726e-06, + "loss": 0.0115, + "step": 19117 + }, + { + "epoch": 3.48, + "learning_rate": 1.9587914088096106e-06, + "loss": 0.0083, + "step": 19118 + }, + { + "epoch": 3.48, + "learning_rate": 1.9587368037859485e-06, + "loss": 0.2167, + "step": 19119 + }, + { + "epoch": 3.48, + "learning_rate": 1.9586821987622864e-06, + "loss": 0.1299, + "step": 19120 + }, + { + "epoch": 3.48, + "learning_rate": 1.958627593738624e-06, + "loss": 0.5817, + "step": 19121 + }, + { + "epoch": 3.48, + "learning_rate": 1.958572988714962e-06, + "loss": 2.253, + "step": 19122 + }, + { + "epoch": 3.48, + "learning_rate": 1.9585183836913e-06, + "loss": 2.7306, + "step": 19123 + }, + { + "epoch": 3.48, + "learning_rate": 1.9584637786676374e-06, + "loss": 0.1238, + "step": 19124 + }, + { + "epoch": 3.48, + "learning_rate": 1.9584091736439753e-06, + "loss": 0.3495, + "step": 19125 + }, + { + "epoch": 3.48, + "learning_rate": 1.9583545686203133e-06, + "loss": 0.5405, + "step": 19126 + }, + { + "epoch": 3.48, + "learning_rate": 1.9582999635966508e-06, + "loss": 0.3822, + "step": 19127 + }, + { + "epoch": 3.48, + "learning_rate": 1.9582453585729887e-06, + "loss": 2.599, + "step": 19128 + }, + { + "epoch": 3.48, + "learning_rate": 1.9581907535493267e-06, + "loss": 0.7588, + "step": 19129 + }, + { + "epoch": 3.48, + "learning_rate": 1.958136148525664e-06, + "loss": 2.223, + "step": 19130 + }, + { + "epoch": 3.48, + "learning_rate": 1.958081543502002e-06, + "loss": 1.0716, + "step": 19131 + }, + { + "epoch": 3.48, + "learning_rate": 1.95802693847834e-06, + "loss": 0.9063, + "step": 19132 + }, + { + "epoch": 3.48, + "learning_rate": 1.957972333454678e-06, + "loss": 1.4669, + "step": 19133 + }, + { + "epoch": 3.48, + "learning_rate": 1.957917728431016e-06, + "loss": 0.0523, + "step": 19134 + }, + { + "epoch": 3.48, + "learning_rate": 1.9578631234073535e-06, + "loss": 0.2903, + "step": 19135 + }, + { + "epoch": 3.48, + "learning_rate": 1.9578085183836914e-06, + "loss": 0.8267, + "step": 19136 + }, + { + "epoch": 3.48, + "learning_rate": 1.9577539133600294e-06, + "loss": 0.62, + "step": 19137 + }, + { + "epoch": 3.48, + "learning_rate": 1.957699308336367e-06, + "loss": 0.0763, + "step": 19138 + }, + { + "epoch": 3.48, + "learning_rate": 1.957644703312705e-06, + "loss": 0.059, + "step": 19139 + }, + { + "epoch": 3.48, + "learning_rate": 1.9575900982890428e-06, + "loss": 1.1159, + "step": 19140 + }, + { + "epoch": 3.48, + "learning_rate": 1.9575354932653803e-06, + "loss": 0.2145, + "step": 19141 + }, + { + "epoch": 3.48, + "learning_rate": 1.9574808882417182e-06, + "loss": 2.337, + "step": 19142 + }, + { + "epoch": 3.48, + "learning_rate": 1.9574262832180558e-06, + "loss": 0.7152, + "step": 19143 + }, + { + "epoch": 3.48, + "learning_rate": 1.9573716781943937e-06, + "loss": 0.1955, + "step": 19144 + }, + { + "epoch": 3.48, + "learning_rate": 1.9573170731707316e-06, + "loss": 0.0518, + "step": 19145 + }, + { + "epoch": 3.48, + "learning_rate": 1.9572624681470696e-06, + "loss": 0.9322, + "step": 19146 + }, + { + "epoch": 3.48, + "learning_rate": 1.9572078631234075e-06, + "loss": 1.0306, + "step": 19147 + }, + { + "epoch": 3.48, + "learning_rate": 1.9571532580997455e-06, + "loss": 0.5279, + "step": 19148 + }, + { + "epoch": 3.48, + "learning_rate": 1.957098653076083e-06, + "loss": 3.6272, + "step": 19149 + }, + { + "epoch": 3.48, + "learning_rate": 1.957044048052421e-06, + "loss": 0.456, + "step": 19150 + }, + { + "epoch": 3.48, + "learning_rate": 1.956989443028759e-06, + "loss": 0.4303, + "step": 19151 + }, + { + "epoch": 3.48, + "learning_rate": 1.9569348380050964e-06, + "loss": 0.952, + "step": 19152 + }, + { + "epoch": 3.48, + "learning_rate": 1.9568802329814343e-06, + "loss": 0.2034, + "step": 19153 + }, + { + "epoch": 3.48, + "learning_rate": 1.9568256279577723e-06, + "loss": 0.2212, + "step": 19154 + }, + { + "epoch": 3.48, + "learning_rate": 1.95677102293411e-06, + "loss": 3.383, + "step": 19155 + }, + { + "epoch": 3.48, + "learning_rate": 1.9567164179104478e-06, + "loss": 0.7486, + "step": 19156 + }, + { + "epoch": 3.48, + "learning_rate": 1.9566618128867853e-06, + "loss": 1.453, + "step": 19157 + }, + { + "epoch": 3.48, + "learning_rate": 1.9566072078631232e-06, + "loss": 0.7921, + "step": 19158 + }, + { + "epoch": 3.48, + "learning_rate": 1.956552602839461e-06, + "loss": 0.8451, + "step": 19159 + }, + { + "epoch": 3.48, + "learning_rate": 1.956497997815799e-06, + "loss": 0.3477, + "step": 19160 + }, + { + "epoch": 3.48, + "learning_rate": 1.956443392792137e-06, + "loss": 2.1921, + "step": 19161 + }, + { + "epoch": 3.48, + "learning_rate": 1.956388787768475e-06, + "loss": 1.8956, + "step": 19162 + }, + { + "epoch": 3.48, + "learning_rate": 1.9563341827448125e-06, + "loss": 0.0779, + "step": 19163 + }, + { + "epoch": 3.48, + "learning_rate": 1.9562795777211505e-06, + "loss": 0.2485, + "step": 19164 + }, + { + "epoch": 3.49, + "learning_rate": 1.9562249726974884e-06, + "loss": 0.2469, + "step": 19165 + }, + { + "epoch": 3.49, + "learning_rate": 1.956170367673826e-06, + "loss": 3.0083, + "step": 19166 + }, + { + "epoch": 3.49, + "learning_rate": 1.956115762650164e-06, + "loss": 0.0562, + "step": 19167 + }, + { + "epoch": 3.49, + "learning_rate": 1.956061157626502e-06, + "loss": 0.058, + "step": 19168 + }, + { + "epoch": 3.49, + "learning_rate": 1.9560065526028393e-06, + "loss": 0.0862, + "step": 19169 + }, + { + "epoch": 3.49, + "learning_rate": 1.9559519475791773e-06, + "loss": 0.0176, + "step": 19170 + }, + { + "epoch": 3.49, + "learning_rate": 1.955897342555515e-06, + "loss": 0.0396, + "step": 19171 + }, + { + "epoch": 3.49, + "learning_rate": 1.955842737531853e-06, + "loss": 0.3981, + "step": 19172 + }, + { + "epoch": 3.49, + "learning_rate": 1.955788132508191e-06, + "loss": 0.3189, + "step": 19173 + }, + { + "epoch": 3.49, + "learning_rate": 1.9557335274845286e-06, + "loss": 2.8502, + "step": 19174 + }, + { + "epoch": 3.49, + "learning_rate": 1.9556789224608666e-06, + "loss": 0.0695, + "step": 19175 + }, + { + "epoch": 3.49, + "learning_rate": 1.9556243174372045e-06, + "loss": 2.7795, + "step": 19176 + }, + { + "epoch": 3.49, + "learning_rate": 1.955569712413542e-06, + "loss": 0.1503, + "step": 19177 + }, + { + "epoch": 3.49, + "learning_rate": 1.95551510738988e-06, + "loss": 0.0136, + "step": 19178 + }, + { + "epoch": 3.49, + "learning_rate": 1.955460502366218e-06, + "loss": 0.0305, + "step": 19179 + }, + { + "epoch": 3.49, + "learning_rate": 1.9554058973425554e-06, + "loss": 0.7508, + "step": 19180 + }, + { + "epoch": 3.49, + "learning_rate": 1.9553512923188934e-06, + "loss": 0.6614, + "step": 19181 + }, + { + "epoch": 3.49, + "learning_rate": 1.9552966872952313e-06, + "loss": 0.0261, + "step": 19182 + }, + { + "epoch": 3.49, + "learning_rate": 1.955242082271569e-06, + "loss": 0.0837, + "step": 19183 + }, + { + "epoch": 3.49, + "learning_rate": 1.9551874772479068e-06, + "loss": 0.0185, + "step": 19184 + }, + { + "epoch": 3.49, + "learning_rate": 1.9551328722242447e-06, + "loss": 2.0773, + "step": 19185 + }, + { + "epoch": 3.49, + "learning_rate": 1.9550782672005827e-06, + "loss": 0.0198, + "step": 19186 + }, + { + "epoch": 3.49, + "learning_rate": 1.9550236621769206e-06, + "loss": 0.2361, + "step": 19187 + }, + { + "epoch": 3.49, + "learning_rate": 1.954969057153258e-06, + "loss": 0.3551, + "step": 19188 + }, + { + "epoch": 3.49, + "learning_rate": 1.954914452129596e-06, + "loss": 0.0211, + "step": 19189 + }, + { + "epoch": 3.49, + "learning_rate": 1.954859847105934e-06, + "loss": 0.8902, + "step": 19190 + }, + { + "epoch": 3.49, + "learning_rate": 1.9548052420822715e-06, + "loss": 1.1823, + "step": 19191 + }, + { + "epoch": 3.49, + "learning_rate": 1.9547506370586095e-06, + "loss": 0.0236, + "step": 19192 + }, + { + "epoch": 3.49, + "learning_rate": 1.9546960320349474e-06, + "loss": 1.0734, + "step": 19193 + }, + { + "epoch": 3.49, + "learning_rate": 1.954641427011285e-06, + "loss": 0.0668, + "step": 19194 + }, + { + "epoch": 3.49, + "learning_rate": 1.954586821987623e-06, + "loss": 1.0755, + "step": 19195 + }, + { + "epoch": 3.49, + "learning_rate": 1.954532216963961e-06, + "loss": 1.1334, + "step": 19196 + }, + { + "epoch": 3.49, + "learning_rate": 1.9544776119402983e-06, + "loss": 1.6377, + "step": 19197 + }, + { + "epoch": 3.49, + "learning_rate": 1.9544230069166363e-06, + "loss": 0.1032, + "step": 19198 + }, + { + "epoch": 3.49, + "learning_rate": 1.9543684018929742e-06, + "loss": 0.1012, + "step": 19199 + }, + { + "epoch": 3.49, + "learning_rate": 1.954313796869312e-06, + "loss": 2.1713, + "step": 19200 + }, + { + "epoch": 3.49, + "learning_rate": 1.95425919184565e-06, + "loss": 0.1906, + "step": 19201 + }, + { + "epoch": 3.49, + "learning_rate": 1.9542045868219876e-06, + "loss": 0.5942, + "step": 19202 + }, + { + "epoch": 3.49, + "learning_rate": 1.9541499817983256e-06, + "loss": 0.0192, + "step": 19203 + }, + { + "epoch": 3.49, + "learning_rate": 1.9540953767746635e-06, + "loss": 0.2118, + "step": 19204 + }, + { + "epoch": 3.49, + "learning_rate": 1.954040771751001e-06, + "loss": 0.039, + "step": 19205 + }, + { + "epoch": 3.49, + "learning_rate": 1.953986166727339e-06, + "loss": 0.7292, + "step": 19206 + }, + { + "epoch": 3.49, + "learning_rate": 1.953931561703677e-06, + "loss": 2.6166, + "step": 19207 + }, + { + "epoch": 3.49, + "learning_rate": 1.9538769566800145e-06, + "loss": 0.0719, + "step": 19208 + }, + { + "epoch": 3.49, + "learning_rate": 1.9538223516563524e-06, + "loss": 0.0126, + "step": 19209 + }, + { + "epoch": 3.49, + "learning_rate": 1.9537677466326903e-06, + "loss": 0.0229, + "step": 19210 + }, + { + "epoch": 3.49, + "learning_rate": 1.953713141609028e-06, + "loss": 0.024, + "step": 19211 + }, + { + "epoch": 3.49, + "learning_rate": 1.953658536585366e-06, + "loss": 1.3037, + "step": 19212 + }, + { + "epoch": 3.49, + "learning_rate": 1.9536039315617037e-06, + "loss": 2.9021, + "step": 19213 + }, + { + "epoch": 3.49, + "learning_rate": 1.9535493265380417e-06, + "loss": 0.1913, + "step": 19214 + }, + { + "epoch": 3.49, + "learning_rate": 1.9534947215143796e-06, + "loss": 1.0804, + "step": 19215 + }, + { + "epoch": 3.49, + "learning_rate": 1.953440116490717e-06, + "loss": 1.1038, + "step": 19216 + }, + { + "epoch": 3.49, + "learning_rate": 1.953385511467055e-06, + "loss": 0.6306, + "step": 19217 + }, + { + "epoch": 3.49, + "learning_rate": 1.953330906443393e-06, + "loss": 1.5748, + "step": 19218 + }, + { + "epoch": 3.49, + "learning_rate": 1.9532763014197306e-06, + "loss": 0.0883, + "step": 19219 + }, + { + "epoch": 3.5, + "learning_rate": 1.9532216963960685e-06, + "loss": 0.031, + "step": 19220 + }, + { + "epoch": 3.5, + "learning_rate": 1.9531670913724064e-06, + "loss": 0.554, + "step": 19221 + }, + { + "epoch": 3.5, + "learning_rate": 1.953112486348744e-06, + "loss": 0.269, + "step": 19222 + }, + { + "epoch": 3.5, + "learning_rate": 1.953057881325082e-06, + "loss": 0.5448, + "step": 19223 + }, + { + "epoch": 3.5, + "learning_rate": 1.9530032763014194e-06, + "loss": 0.3993, + "step": 19224 + }, + { + "epoch": 3.5, + "learning_rate": 1.9529486712777574e-06, + "loss": 1.2148, + "step": 19225 + }, + { + "epoch": 3.5, + "learning_rate": 1.9528940662540953e-06, + "loss": 0.7583, + "step": 19226 + }, + { + "epoch": 3.5, + "learning_rate": 1.9528394612304333e-06, + "loss": 1.1474, + "step": 19227 + }, + { + "epoch": 3.5, + "learning_rate": 1.952784856206771e-06, + "loss": 0.0427, + "step": 19228 + }, + { + "epoch": 3.5, + "learning_rate": 1.952730251183109e-06, + "loss": 1.7647, + "step": 19229 + }, + { + "epoch": 3.5, + "learning_rate": 1.9526756461594467e-06, + "loss": 0.1179, + "step": 19230 + }, + { + "epoch": 3.5, + "learning_rate": 1.9526210411357846e-06, + "loss": 2.7737, + "step": 19231 + }, + { + "epoch": 3.5, + "learning_rate": 1.9525664361121226e-06, + "loss": 0.0161, + "step": 19232 + }, + { + "epoch": 3.5, + "learning_rate": 1.95251183108846e-06, + "loss": 0.9831, + "step": 19233 + }, + { + "epoch": 3.5, + "learning_rate": 1.952457226064798e-06, + "loss": 2.2272, + "step": 19234 + }, + { + "epoch": 3.5, + "learning_rate": 1.952402621041136e-06, + "loss": 0.0236, + "step": 19235 + }, + { + "epoch": 3.5, + "learning_rate": 1.9523480160174735e-06, + "loss": 3.2848, + "step": 19236 + }, + { + "epoch": 3.5, + "learning_rate": 1.9522934109938114e-06, + "loss": 0.5747, + "step": 19237 + }, + { + "epoch": 3.5, + "learning_rate": 1.952238805970149e-06, + "loss": 0.3387, + "step": 19238 + }, + { + "epoch": 3.5, + "learning_rate": 1.952184200946487e-06, + "loss": 0.9153, + "step": 19239 + }, + { + "epoch": 3.5, + "learning_rate": 1.952129595922825e-06, + "loss": 2.1596, + "step": 19240 + }, + { + "epoch": 3.5, + "learning_rate": 1.9520749908991628e-06, + "loss": 1.1571, + "step": 19241 + }, + { + "epoch": 3.5, + "learning_rate": 1.9520203858755007e-06, + "loss": 0.5659, + "step": 19242 + }, + { + "epoch": 3.5, + "learning_rate": 1.9519657808518387e-06, + "loss": 1.0446, + "step": 19243 + }, + { + "epoch": 3.5, + "learning_rate": 1.951911175828176e-06, + "loss": 0.1858, + "step": 19244 + }, + { + "epoch": 3.5, + "learning_rate": 1.951856570804514e-06, + "loss": 0.5419, + "step": 19245 + }, + { + "epoch": 3.5, + "learning_rate": 1.951801965780852e-06, + "loss": 0.3309, + "step": 19246 + }, + { + "epoch": 3.5, + "learning_rate": 1.9517473607571896e-06, + "loss": 0.0139, + "step": 19247 + }, + { + "epoch": 3.5, + "learning_rate": 1.9516927557335275e-06, + "loss": 0.3243, + "step": 19248 + }, + { + "epoch": 3.5, + "learning_rate": 1.9516381507098655e-06, + "loss": 0.2363, + "step": 19249 + }, + { + "epoch": 3.5, + "learning_rate": 1.951583545686203e-06, + "loss": 1.1026, + "step": 19250 + }, + { + "epoch": 3.5, + "learning_rate": 1.951528940662541e-06, + "loss": 0.1815, + "step": 19251 + }, + { + "epoch": 3.5, + "learning_rate": 1.951474335638879e-06, + "loss": 0.0958, + "step": 19252 + }, + { + "epoch": 3.5, + "learning_rate": 1.951419730615217e-06, + "loss": 0.8149, + "step": 19253 + }, + { + "epoch": 3.5, + "learning_rate": 1.9513651255915543e-06, + "loss": 0.0283, + "step": 19254 + }, + { + "epoch": 3.5, + "learning_rate": 1.9513105205678923e-06, + "loss": 1.167, + "step": 19255 + }, + { + "epoch": 3.5, + "learning_rate": 1.9512559155442302e-06, + "loss": 0.0127, + "step": 19256 + }, + { + "epoch": 3.5, + "learning_rate": 1.951201310520568e-06, + "loss": 0.9127, + "step": 19257 + }, + { + "epoch": 3.5, + "learning_rate": 1.9511467054969057e-06, + "loss": 0.5298, + "step": 19258 + }, + { + "epoch": 3.5, + "learning_rate": 1.9510921004732436e-06, + "loss": 3.6045, + "step": 19259 + }, + { + "epoch": 3.5, + "learning_rate": 1.9510374954495816e-06, + "loss": 0.4494, + "step": 19260 + }, + { + "epoch": 3.5, + "learning_rate": 1.950982890425919e-06, + "loss": 0.3555, + "step": 19261 + }, + { + "epoch": 3.5, + "learning_rate": 1.950928285402257e-06, + "loss": 1.4594, + "step": 19262 + }, + { + "epoch": 3.5, + "learning_rate": 1.950873680378595e-06, + "loss": 0.8035, + "step": 19263 + }, + { + "epoch": 3.5, + "learning_rate": 1.9508190753549325e-06, + "loss": 0.3726, + "step": 19264 + }, + { + "epoch": 3.5, + "learning_rate": 1.9507644703312705e-06, + "loss": 0.0446, + "step": 19265 + }, + { + "epoch": 3.5, + "learning_rate": 1.9507098653076084e-06, + "loss": 0.0538, + "step": 19266 + }, + { + "epoch": 3.5, + "learning_rate": 1.9506552602839463e-06, + "loss": 0.1707, + "step": 19267 + }, + { + "epoch": 3.5, + "learning_rate": 1.9506006552602843e-06, + "loss": 1.3542, + "step": 19268 + }, + { + "epoch": 3.5, + "learning_rate": 1.950546050236622e-06, + "loss": 4.1557, + "step": 19269 + }, + { + "epoch": 3.5, + "learning_rate": 1.9504914452129597e-06, + "loss": 0.0138, + "step": 19270 + }, + { + "epoch": 3.5, + "learning_rate": 1.9504368401892977e-06, + "loss": 0.3395, + "step": 19271 + }, + { + "epoch": 3.5, + "learning_rate": 1.950382235165635e-06, + "loss": 0.0195, + "step": 19272 + }, + { + "epoch": 3.5, + "learning_rate": 1.950327630141973e-06, + "loss": 1.112, + "step": 19273 + }, + { + "epoch": 3.51, + "learning_rate": 1.950273025118311e-06, + "loss": 0.4438, + "step": 19274 + }, + { + "epoch": 3.51, + "learning_rate": 1.9502184200946486e-06, + "loss": 0.5194, + "step": 19275 + }, + { + "epoch": 3.51, + "learning_rate": 1.9501638150709866e-06, + "loss": 0.3558, + "step": 19276 + }, + { + "epoch": 3.51, + "learning_rate": 1.9501092100473245e-06, + "loss": 0.2451, + "step": 19277 + }, + { + "epoch": 3.51, + "learning_rate": 1.950054605023662e-06, + "loss": 2.1193, + "step": 19278 + }, + { + "epoch": 3.51, + "learning_rate": 1.95e-06, + "loss": 2.5803, + "step": 19279 + }, + { + "epoch": 3.51, + "learning_rate": 1.949945394976338e-06, + "loss": 1.0878, + "step": 19280 + }, + { + "epoch": 3.51, + "learning_rate": 1.949890789952676e-06, + "loss": 0.1752, + "step": 19281 + }, + { + "epoch": 3.51, + "learning_rate": 1.949836184929014e-06, + "loss": 2.0181, + "step": 19282 + }, + { + "epoch": 3.51, + "learning_rate": 1.9497815799053513e-06, + "loss": 0.6005, + "step": 19283 + }, + { + "epoch": 3.51, + "learning_rate": 1.9497269748816893e-06, + "loss": 0.0178, + "step": 19284 + }, + { + "epoch": 3.51, + "learning_rate": 1.949672369858027e-06, + "loss": 0.4285, + "step": 19285 + }, + { + "epoch": 3.51, + "learning_rate": 1.9496177648343647e-06, + "loss": 2.4063, + "step": 19286 + }, + { + "epoch": 3.51, + "learning_rate": 1.9495631598107027e-06, + "loss": 1.9097, + "step": 19287 + }, + { + "epoch": 3.51, + "learning_rate": 1.9495085547870406e-06, + "loss": 0.0207, + "step": 19288 + }, + { + "epoch": 3.51, + "learning_rate": 1.949453949763378e-06, + "loss": 0.5941, + "step": 19289 + }, + { + "epoch": 3.51, + "learning_rate": 1.949399344739716e-06, + "loss": 0.0126, + "step": 19290 + }, + { + "epoch": 3.51, + "learning_rate": 1.949344739716054e-06, + "loss": 0.9824, + "step": 19291 + }, + { + "epoch": 3.51, + "learning_rate": 1.9492901346923915e-06, + "loss": 1.8234, + "step": 19292 + }, + { + "epoch": 3.51, + "learning_rate": 1.9492355296687295e-06, + "loss": 0.0963, + "step": 19293 + }, + { + "epoch": 3.51, + "learning_rate": 1.9491809246450674e-06, + "loss": 1.6436, + "step": 19294 + }, + { + "epoch": 3.51, + "learning_rate": 1.9491263196214054e-06, + "loss": 3.5182, + "step": 19295 + }, + { + "epoch": 3.51, + "learning_rate": 1.9490717145977433e-06, + "loss": 0.0242, + "step": 19296 + }, + { + "epoch": 3.51, + "learning_rate": 1.949017109574081e-06, + "loss": 1.3489, + "step": 19297 + }, + { + "epoch": 3.51, + "learning_rate": 1.9489625045504188e-06, + "loss": 0.8483, + "step": 19298 + }, + { + "epoch": 3.51, + "learning_rate": 1.9489078995267567e-06, + "loss": 1.9315, + "step": 19299 + }, + { + "epoch": 3.51, + "learning_rate": 1.9488532945030942e-06, + "loss": 0.4018, + "step": 19300 + }, + { + "epoch": 3.51, + "learning_rate": 1.948798689479432e-06, + "loss": 0.2317, + "step": 19301 + }, + { + "epoch": 3.51, + "learning_rate": 1.94874408445577e-06, + "loss": 1.2228, + "step": 19302 + }, + { + "epoch": 3.51, + "learning_rate": 1.9486894794321076e-06, + "loss": 0.2137, + "step": 19303 + }, + { + "epoch": 3.51, + "learning_rate": 1.9486348744084456e-06, + "loss": 3.2976, + "step": 19304 + }, + { + "epoch": 3.51, + "learning_rate": 1.948580269384783e-06, + "loss": 0.4603, + "step": 19305 + }, + { + "epoch": 3.51, + "learning_rate": 1.948525664361121e-06, + "loss": 1.1416, + "step": 19306 + }, + { + "epoch": 3.51, + "learning_rate": 1.948471059337459e-06, + "loss": 0.2919, + "step": 19307 + }, + { + "epoch": 3.51, + "learning_rate": 1.948416454313797e-06, + "loss": 0.0103, + "step": 19308 + }, + { + "epoch": 3.51, + "learning_rate": 1.948361849290135e-06, + "loss": 0.1589, + "step": 19309 + }, + { + "epoch": 3.51, + "learning_rate": 1.948307244266473e-06, + "loss": 0.3691, + "step": 19310 + }, + { + "epoch": 3.51, + "learning_rate": 1.9482526392428103e-06, + "loss": 0.5864, + "step": 19311 + }, + { + "epoch": 3.51, + "learning_rate": 1.9481980342191483e-06, + "loss": 0.0649, + "step": 19312 + }, + { + "epoch": 3.51, + "learning_rate": 1.9481434291954862e-06, + "loss": 0.5236, + "step": 19313 + }, + { + "epoch": 3.51, + "learning_rate": 1.9480888241718237e-06, + "loss": 0.0447, + "step": 19314 + }, + { + "epoch": 3.51, + "learning_rate": 1.9480342191481617e-06, + "loss": 0.3499, + "step": 19315 + }, + { + "epoch": 3.51, + "learning_rate": 1.9479796141244996e-06, + "loss": 1.7382, + "step": 19316 + }, + { + "epoch": 3.51, + "learning_rate": 1.947925009100837e-06, + "loss": 0.5027, + "step": 19317 + }, + { + "epoch": 3.51, + "learning_rate": 1.947870404077175e-06, + "loss": 0.3655, + "step": 19318 + }, + { + "epoch": 3.51, + "learning_rate": 1.9478157990535126e-06, + "loss": 0.4004, + "step": 19319 + }, + { + "epoch": 3.51, + "learning_rate": 1.9477611940298506e-06, + "loss": 0.6642, + "step": 19320 + }, + { + "epoch": 3.51, + "learning_rate": 1.9477065890061885e-06, + "loss": 0.6034, + "step": 19321 + }, + { + "epoch": 3.51, + "learning_rate": 1.9476519839825264e-06, + "loss": 2.589, + "step": 19322 + }, + { + "epoch": 3.51, + "learning_rate": 1.9475973789588644e-06, + "loss": 0.0464, + "step": 19323 + }, + { + "epoch": 3.51, + "learning_rate": 1.9475427739352023e-06, + "loss": 0.6339, + "step": 19324 + }, + { + "epoch": 3.51, + "learning_rate": 1.94748816891154e-06, + "loss": 0.0157, + "step": 19325 + }, + { + "epoch": 3.51, + "learning_rate": 1.947433563887878e-06, + "loss": 1.7056, + "step": 19326 + }, + { + "epoch": 3.51, + "learning_rate": 1.9473789588642157e-06, + "loss": 0.3627, + "step": 19327 + }, + { + "epoch": 3.51, + "learning_rate": 1.9473243538405533e-06, + "loss": 2.0177, + "step": 19328 + }, + { + "epoch": 3.52, + "learning_rate": 1.947269748816891e-06, + "loss": 0.649, + "step": 19329 + }, + { + "epoch": 3.52, + "learning_rate": 1.947215143793229e-06, + "loss": 0.6783, + "step": 19330 + }, + { + "epoch": 3.52, + "learning_rate": 1.9471605387695667e-06, + "loss": 0.0947, + "step": 19331 + }, + { + "epoch": 3.52, + "learning_rate": 1.9471059337459046e-06, + "loss": 0.2777, + "step": 19332 + }, + { + "epoch": 3.52, + "learning_rate": 1.947051328722242e-06, + "loss": 1.0131, + "step": 19333 + }, + { + "epoch": 3.52, + "learning_rate": 1.94699672369858e-06, + "loss": 0.8929, + "step": 19334 + }, + { + "epoch": 3.52, + "learning_rate": 1.946942118674918e-06, + "loss": 0.5022, + "step": 19335 + }, + { + "epoch": 3.52, + "learning_rate": 1.946887513651256e-06, + "loss": 0.0142, + "step": 19336 + }, + { + "epoch": 3.52, + "learning_rate": 1.946832908627594e-06, + "loss": 1.198, + "step": 19337 + }, + { + "epoch": 3.52, + "learning_rate": 1.946778303603932e-06, + "loss": 0.1533, + "step": 19338 + }, + { + "epoch": 3.52, + "learning_rate": 1.9467236985802694e-06, + "loss": 0.8074, + "step": 19339 + }, + { + "epoch": 3.52, + "learning_rate": 1.9466690935566073e-06, + "loss": 0.8055, + "step": 19340 + }, + { + "epoch": 3.52, + "learning_rate": 1.9466144885329453e-06, + "loss": 5.3911, + "step": 19341 + }, + { + "epoch": 3.52, + "learning_rate": 1.9465598835092828e-06, + "loss": 0.8582, + "step": 19342 + }, + { + "epoch": 3.52, + "learning_rate": 1.9465052784856207e-06, + "loss": 0.3844, + "step": 19343 + }, + { + "epoch": 3.52, + "learning_rate": 1.9464506734619587e-06, + "loss": 0.2744, + "step": 19344 + }, + { + "epoch": 3.52, + "learning_rate": 1.946396068438296e-06, + "loss": 0.4257, + "step": 19345 + }, + { + "epoch": 3.52, + "learning_rate": 1.946341463414634e-06, + "loss": 0.699, + "step": 19346 + }, + { + "epoch": 3.52, + "learning_rate": 1.946286858390972e-06, + "loss": 0.2663, + "step": 19347 + }, + { + "epoch": 3.52, + "learning_rate": 1.94623225336731e-06, + "loss": 0.0126, + "step": 19348 + }, + { + "epoch": 3.52, + "learning_rate": 1.946177648343648e-06, + "loss": 0.2908, + "step": 19349 + }, + { + "epoch": 3.52, + "learning_rate": 1.9461230433199855e-06, + "loss": 0.5497, + "step": 19350 + }, + { + "epoch": 3.52, + "learning_rate": 1.9460684382963234e-06, + "loss": 0.3347, + "step": 19351 + }, + { + "epoch": 3.52, + "learning_rate": 1.9460138332726614e-06, + "loss": 0.2245, + "step": 19352 + }, + { + "epoch": 3.52, + "learning_rate": 1.945959228248999e-06, + "loss": 0.901, + "step": 19353 + }, + { + "epoch": 3.52, + "learning_rate": 1.945904623225337e-06, + "loss": 0.2555, + "step": 19354 + }, + { + "epoch": 3.52, + "learning_rate": 1.9458500182016748e-06, + "loss": 1.6803, + "step": 19355 + }, + { + "epoch": 3.52, + "learning_rate": 1.9457954131780123e-06, + "loss": 0.5819, + "step": 19356 + }, + { + "epoch": 3.52, + "learning_rate": 1.9457408081543502e-06, + "loss": 1.2118, + "step": 19357 + }, + { + "epoch": 3.52, + "learning_rate": 1.945686203130688e-06, + "loss": 0.2147, + "step": 19358 + }, + { + "epoch": 3.52, + "learning_rate": 1.9456315981070257e-06, + "loss": 0.7644, + "step": 19359 + }, + { + "epoch": 3.52, + "learning_rate": 1.9455769930833636e-06, + "loss": 0.0201, + "step": 19360 + }, + { + "epoch": 3.52, + "learning_rate": 1.9455223880597016e-06, + "loss": 0.688, + "step": 19361 + }, + { + "epoch": 3.52, + "learning_rate": 1.9454677830360395e-06, + "loss": 2.2437, + "step": 19362 + }, + { + "epoch": 3.52, + "learning_rate": 1.9454131780123775e-06, + "loss": 1.1095, + "step": 19363 + }, + { + "epoch": 3.52, + "learning_rate": 1.945358572988715e-06, + "loss": 0.7812, + "step": 19364 + }, + { + "epoch": 3.52, + "learning_rate": 1.945303967965053e-06, + "loss": 1.2273, + "step": 19365 + }, + { + "epoch": 3.52, + "learning_rate": 1.945249362941391e-06, + "loss": 2.194, + "step": 19366 + }, + { + "epoch": 3.52, + "learning_rate": 1.9451947579177284e-06, + "loss": 1.8022, + "step": 19367 + }, + { + "epoch": 3.52, + "learning_rate": 1.9451401528940663e-06, + "loss": 1.5784, + "step": 19368 + }, + { + "epoch": 3.52, + "learning_rate": 1.9450855478704043e-06, + "loss": 0.4676, + "step": 19369 + }, + { + "epoch": 3.52, + "learning_rate": 1.945030942846742e-06, + "loss": 0.305, + "step": 19370 + }, + { + "epoch": 3.52, + "learning_rate": 1.9449763378230797e-06, + "loss": 1.1269, + "step": 19371 + }, + { + "epoch": 3.52, + "learning_rate": 1.9449217327994177e-06, + "loss": 0.426, + "step": 19372 + }, + { + "epoch": 3.52, + "learning_rate": 1.944867127775755e-06, + "loss": 0.7897, + "step": 19373 + }, + { + "epoch": 3.52, + "learning_rate": 1.944812522752093e-06, + "loss": 0.0588, + "step": 19374 + }, + { + "epoch": 3.52, + "learning_rate": 1.944757917728431e-06, + "loss": 0.0352, + "step": 19375 + }, + { + "epoch": 3.52, + "learning_rate": 1.944703312704769e-06, + "loss": 0.3679, + "step": 19376 + }, + { + "epoch": 3.52, + "learning_rate": 1.944648707681107e-06, + "loss": 0.933, + "step": 19377 + }, + { + "epoch": 3.52, + "learning_rate": 1.9445941026574445e-06, + "loss": 1.9161, + "step": 19378 + }, + { + "epoch": 3.52, + "learning_rate": 1.9445394976337824e-06, + "loss": 1.1547, + "step": 19379 + }, + { + "epoch": 3.52, + "learning_rate": 1.9444848926101204e-06, + "loss": 0.0366, + "step": 19380 + }, + { + "epoch": 3.52, + "learning_rate": 1.944430287586458e-06, + "loss": 0.0238, + "step": 19381 + }, + { + "epoch": 3.52, + "learning_rate": 1.944375682562796e-06, + "loss": 2.3838, + "step": 19382 + }, + { + "epoch": 3.52, + "learning_rate": 1.944321077539134e-06, + "loss": 1.1257, + "step": 19383 + }, + { + "epoch": 3.53, + "learning_rate": 1.9442664725154713e-06, + "loss": 0.9169, + "step": 19384 + }, + { + "epoch": 3.53, + "learning_rate": 1.9442118674918093e-06, + "loss": 0.5278, + "step": 19385 + }, + { + "epoch": 3.53, + "learning_rate": 1.9441572624681468e-06, + "loss": 1.2224, + "step": 19386 + }, + { + "epoch": 3.53, + "learning_rate": 1.9441026574444847e-06, + "loss": 0.1751, + "step": 19387 + }, + { + "epoch": 3.53, + "learning_rate": 1.9440480524208227e-06, + "loss": 2.2063, + "step": 19388 + }, + { + "epoch": 3.53, + "learning_rate": 1.9439934473971606e-06, + "loss": 0.4699, + "step": 19389 + }, + { + "epoch": 3.53, + "learning_rate": 1.9439388423734986e-06, + "loss": 0.5286, + "step": 19390 + }, + { + "epoch": 3.53, + "learning_rate": 1.9438842373498365e-06, + "loss": 0.1916, + "step": 19391 + }, + { + "epoch": 3.53, + "learning_rate": 1.943829632326174e-06, + "loss": 0.128, + "step": 19392 + }, + { + "epoch": 3.53, + "learning_rate": 1.943775027302512e-06, + "loss": 0.0909, + "step": 19393 + }, + { + "epoch": 3.53, + "learning_rate": 1.94372042227885e-06, + "loss": 0.6775, + "step": 19394 + }, + { + "epoch": 3.53, + "learning_rate": 1.9436658172551874e-06, + "loss": 0.7478, + "step": 19395 + }, + { + "epoch": 3.53, + "learning_rate": 1.9436112122315254e-06, + "loss": 1.9729, + "step": 19396 + }, + { + "epoch": 3.53, + "learning_rate": 1.9435566072078633e-06, + "loss": 0.5073, + "step": 19397 + }, + { + "epoch": 3.53, + "learning_rate": 1.943502002184201e-06, + "loss": 1.9954, + "step": 19398 + }, + { + "epoch": 3.53, + "learning_rate": 1.9434473971605388e-06, + "loss": 0.2658, + "step": 19399 + }, + { + "epoch": 3.53, + "learning_rate": 1.9433927921368763e-06, + "loss": 0.7241, + "step": 19400 + }, + { + "epoch": 3.53, + "learning_rate": 1.9433381871132142e-06, + "loss": 0.6802, + "step": 19401 + }, + { + "epoch": 3.53, + "learning_rate": 1.943283582089552e-06, + "loss": 0.3041, + "step": 19402 + }, + { + "epoch": 3.53, + "learning_rate": 1.94322897706589e-06, + "loss": 0.491, + "step": 19403 + }, + { + "epoch": 3.53, + "learning_rate": 1.943174372042228e-06, + "loss": 1.0514, + "step": 19404 + }, + { + "epoch": 3.53, + "learning_rate": 1.943119767018566e-06, + "loss": 0.2077, + "step": 19405 + }, + { + "epoch": 3.53, + "learning_rate": 1.9430651619949035e-06, + "loss": 0.1667, + "step": 19406 + }, + { + "epoch": 3.53, + "learning_rate": 1.9430105569712415e-06, + "loss": 0.9698, + "step": 19407 + }, + { + "epoch": 3.53, + "learning_rate": 1.9429559519475794e-06, + "loss": 0.2991, + "step": 19408 + }, + { + "epoch": 3.53, + "learning_rate": 1.942901346923917e-06, + "loss": 0.6452, + "step": 19409 + }, + { + "epoch": 3.53, + "learning_rate": 1.942846741900255e-06, + "loss": 1.4289, + "step": 19410 + }, + { + "epoch": 3.53, + "learning_rate": 1.942792136876593e-06, + "loss": 1.9292, + "step": 19411 + }, + { + "epoch": 3.53, + "learning_rate": 1.9427375318529303e-06, + "loss": 0.924, + "step": 19412 + }, + { + "epoch": 3.53, + "learning_rate": 1.9426829268292683e-06, + "loss": 0.6577, + "step": 19413 + }, + { + "epoch": 3.53, + "learning_rate": 1.942628321805606e-06, + "loss": 0.0142, + "step": 19414 + }, + { + "epoch": 3.53, + "learning_rate": 1.9425737167819437e-06, + "loss": 1.4539, + "step": 19415 + }, + { + "epoch": 3.53, + "learning_rate": 1.9425191117582817e-06, + "loss": 1.7049, + "step": 19416 + }, + { + "epoch": 3.53, + "learning_rate": 1.9424645067346196e-06, + "loss": 0.6619, + "step": 19417 + }, + { + "epoch": 3.53, + "learning_rate": 1.9424099017109576e-06, + "loss": 1.6143, + "step": 19418 + }, + { + "epoch": 3.53, + "learning_rate": 1.9423552966872955e-06, + "loss": 0.1764, + "step": 19419 + }, + { + "epoch": 3.53, + "learning_rate": 1.942300691663633e-06, + "loss": 0.1446, + "step": 19420 + }, + { + "epoch": 3.53, + "learning_rate": 1.942246086639971e-06, + "loss": 0.5428, + "step": 19421 + }, + { + "epoch": 3.53, + "learning_rate": 1.942191481616309e-06, + "loss": 0.7644, + "step": 19422 + }, + { + "epoch": 3.53, + "learning_rate": 1.9421368765926464e-06, + "loss": 0.557, + "step": 19423 + }, + { + "epoch": 3.53, + "learning_rate": 1.9420822715689844e-06, + "loss": 0.5526, + "step": 19424 + }, + { + "epoch": 3.53, + "learning_rate": 1.9420276665453223e-06, + "loss": 0.214, + "step": 19425 + }, + { + "epoch": 3.53, + "learning_rate": 1.94197306152166e-06, + "loss": 0.4017, + "step": 19426 + }, + { + "epoch": 3.53, + "learning_rate": 1.941918456497998e-06, + "loss": 0.4618, + "step": 19427 + }, + { + "epoch": 3.53, + "learning_rate": 1.9418638514743357e-06, + "loss": 2.1269, + "step": 19428 + }, + { + "epoch": 3.53, + "learning_rate": 1.9418092464506737e-06, + "loss": 0.2002, + "step": 19429 + }, + { + "epoch": 3.53, + "learning_rate": 1.9417546414270116e-06, + "loss": 0.0202, + "step": 19430 + }, + { + "epoch": 3.53, + "learning_rate": 1.941700036403349e-06, + "loss": 1.0436, + "step": 19431 + }, + { + "epoch": 3.53, + "learning_rate": 1.941645431379687e-06, + "loss": 0.0589, + "step": 19432 + }, + { + "epoch": 3.53, + "learning_rate": 1.941590826356025e-06, + "loss": 0.1487, + "step": 19433 + }, + { + "epoch": 3.53, + "learning_rate": 1.9415362213323626e-06, + "loss": 1.9013, + "step": 19434 + }, + { + "epoch": 3.53, + "learning_rate": 1.9414816163087005e-06, + "loss": 0.5052, + "step": 19435 + }, + { + "epoch": 3.53, + "learning_rate": 1.9414270112850384e-06, + "loss": 0.018, + "step": 19436 + }, + { + "epoch": 3.53, + "learning_rate": 1.941372406261376e-06, + "loss": 1.1117, + "step": 19437 + }, + { + "epoch": 3.53, + "learning_rate": 1.941317801237714e-06, + "loss": 0.0165, + "step": 19438 + }, + { + "epoch": 3.54, + "learning_rate": 1.941263196214052e-06, + "loss": 2.2304, + "step": 19439 + }, + { + "epoch": 3.54, + "learning_rate": 1.9412085911903894e-06, + "loss": 0.1113, + "step": 19440 + }, + { + "epoch": 3.54, + "learning_rate": 1.9411539861667273e-06, + "loss": 0.6644, + "step": 19441 + }, + { + "epoch": 3.54, + "learning_rate": 1.9410993811430653e-06, + "loss": 3.8302, + "step": 19442 + }, + { + "epoch": 3.54, + "learning_rate": 1.941044776119403e-06, + "loss": 0.2852, + "step": 19443 + }, + { + "epoch": 3.54, + "learning_rate": 1.940990171095741e-06, + "loss": 0.8381, + "step": 19444 + }, + { + "epoch": 3.54, + "learning_rate": 1.9409355660720787e-06, + "loss": 2.3535, + "step": 19445 + }, + { + "epoch": 3.54, + "learning_rate": 1.9408809610484166e-06, + "loss": 1.0914, + "step": 19446 + }, + { + "epoch": 3.54, + "learning_rate": 1.9408263560247545e-06, + "loss": 0.097, + "step": 19447 + }, + { + "epoch": 3.54, + "learning_rate": 1.940771751001092e-06, + "loss": 0.631, + "step": 19448 + }, + { + "epoch": 3.54, + "learning_rate": 1.94071714597743e-06, + "loss": 0.2973, + "step": 19449 + }, + { + "epoch": 3.54, + "learning_rate": 1.940662540953768e-06, + "loss": 0.5541, + "step": 19450 + }, + { + "epoch": 3.54, + "learning_rate": 1.9406079359301055e-06, + "loss": 0.0363, + "step": 19451 + }, + { + "epoch": 3.54, + "learning_rate": 1.9405533309064434e-06, + "loss": 0.2917, + "step": 19452 + }, + { + "epoch": 3.54, + "learning_rate": 1.940498725882781e-06, + "loss": 0.4036, + "step": 19453 + }, + { + "epoch": 3.54, + "learning_rate": 1.940444120859119e-06, + "loss": 0.3906, + "step": 19454 + }, + { + "epoch": 3.54, + "learning_rate": 1.940389515835457e-06, + "loss": 0.0179, + "step": 19455 + }, + { + "epoch": 3.54, + "learning_rate": 1.9403349108117948e-06, + "loss": 0.1523, + "step": 19456 + }, + { + "epoch": 3.54, + "learning_rate": 1.9402803057881327e-06, + "loss": 0.0096, + "step": 19457 + }, + { + "epoch": 3.54, + "learning_rate": 1.9402257007644707e-06, + "loss": 0.7749, + "step": 19458 + }, + { + "epoch": 3.54, + "learning_rate": 1.940171095740808e-06, + "loss": 0.6813, + "step": 19459 + }, + { + "epoch": 3.54, + "learning_rate": 1.940116490717146e-06, + "loss": 0.6229, + "step": 19460 + }, + { + "epoch": 3.54, + "learning_rate": 1.940061885693484e-06, + "loss": 0.0306, + "step": 19461 + }, + { + "epoch": 3.54, + "learning_rate": 1.9400072806698216e-06, + "loss": 1.8764, + "step": 19462 + }, + { + "epoch": 3.54, + "learning_rate": 1.9399526756461595e-06, + "loss": 1.4477, + "step": 19463 + }, + { + "epoch": 3.54, + "learning_rate": 1.9398980706224975e-06, + "loss": 0.9916, + "step": 19464 + }, + { + "epoch": 3.54, + "learning_rate": 1.939843465598835e-06, + "loss": 0.0753, + "step": 19465 + }, + { + "epoch": 3.54, + "learning_rate": 1.939788860575173e-06, + "loss": 0.5244, + "step": 19466 + }, + { + "epoch": 3.54, + "learning_rate": 1.9397342555515105e-06, + "loss": 0.187, + "step": 19467 + }, + { + "epoch": 3.54, + "learning_rate": 1.9396796505278484e-06, + "loss": 0.5128, + "step": 19468 + }, + { + "epoch": 3.54, + "learning_rate": 1.9396250455041863e-06, + "loss": 1.1828, + "step": 19469 + }, + { + "epoch": 3.54, + "learning_rate": 1.9395704404805243e-06, + "loss": 0.4615, + "step": 19470 + }, + { + "epoch": 3.54, + "learning_rate": 1.9395158354568622e-06, + "loss": 0.6782, + "step": 19471 + }, + { + "epoch": 3.54, + "learning_rate": 1.9394612304332e-06, + "loss": 0.7131, + "step": 19472 + }, + { + "epoch": 3.54, + "learning_rate": 1.9394066254095377e-06, + "loss": 0.2475, + "step": 19473 + }, + { + "epoch": 3.54, + "learning_rate": 1.9393520203858756e-06, + "loss": 1.3224, + "step": 19474 + }, + { + "epoch": 3.54, + "learning_rate": 1.9392974153622136e-06, + "loss": 0.4439, + "step": 19475 + }, + { + "epoch": 3.54, + "learning_rate": 1.939242810338551e-06, + "loss": 1.8657, + "step": 19476 + }, + { + "epoch": 3.54, + "learning_rate": 1.939188205314889e-06, + "loss": 0.568, + "step": 19477 + }, + { + "epoch": 3.54, + "learning_rate": 1.939133600291227e-06, + "loss": 1.1146, + "step": 19478 + }, + { + "epoch": 3.54, + "learning_rate": 1.9390789952675645e-06, + "loss": 0.0184, + "step": 19479 + }, + { + "epoch": 3.54, + "learning_rate": 1.9390243902439024e-06, + "loss": 0.1749, + "step": 19480 + }, + { + "epoch": 3.54, + "learning_rate": 1.93896978522024e-06, + "loss": 0.0554, + "step": 19481 + }, + { + "epoch": 3.54, + "learning_rate": 1.938915180196578e-06, + "loss": 0.8361, + "step": 19482 + }, + { + "epoch": 3.54, + "learning_rate": 1.938860575172916e-06, + "loss": 0.0305, + "step": 19483 + }, + { + "epoch": 3.54, + "learning_rate": 1.938805970149254e-06, + "loss": 0.2699, + "step": 19484 + }, + { + "epoch": 3.54, + "learning_rate": 1.9387513651255917e-06, + "loss": 0.2947, + "step": 19485 + }, + { + "epoch": 3.54, + "learning_rate": 1.9386967601019297e-06, + "loss": 0.3058, + "step": 19486 + }, + { + "epoch": 3.54, + "learning_rate": 1.938642155078267e-06, + "loss": 1.2293, + "step": 19487 + }, + { + "epoch": 3.54, + "learning_rate": 1.938587550054605e-06, + "loss": 2.1068, + "step": 19488 + }, + { + "epoch": 3.54, + "learning_rate": 1.938532945030943e-06, + "loss": 0.7533, + "step": 19489 + }, + { + "epoch": 3.54, + "learning_rate": 1.9384783400072806e-06, + "loss": 0.0487, + "step": 19490 + }, + { + "epoch": 3.54, + "learning_rate": 1.9384237349836186e-06, + "loss": 0.2338, + "step": 19491 + }, + { + "epoch": 3.54, + "learning_rate": 1.9383691299599565e-06, + "loss": 2.0098, + "step": 19492 + }, + { + "epoch": 3.54, + "learning_rate": 1.938314524936294e-06, + "loss": 0.0143, + "step": 19493 + }, + { + "epoch": 3.55, + "learning_rate": 1.938259919912632e-06, + "loss": 0.2754, + "step": 19494 + }, + { + "epoch": 3.55, + "learning_rate": 1.9382053148889695e-06, + "loss": 0.271, + "step": 19495 + }, + { + "epoch": 3.55, + "learning_rate": 1.9381507098653074e-06, + "loss": 0.2491, + "step": 19496 + }, + { + "epoch": 3.55, + "learning_rate": 1.9380961048416454e-06, + "loss": 0.4997, + "step": 19497 + }, + { + "epoch": 3.55, + "learning_rate": 1.9380414998179833e-06, + "loss": 1.4103, + "step": 19498 + }, + { + "epoch": 3.55, + "learning_rate": 1.9379868947943213e-06, + "loss": 2.1175, + "step": 19499 + }, + { + "epoch": 3.55, + "learning_rate": 1.937932289770659e-06, + "loss": 0.2355, + "step": 19500 + }, + { + "epoch": 3.55, + "learning_rate": 1.9378776847469967e-06, + "loss": 0.4933, + "step": 19501 + }, + { + "epoch": 3.55, + "learning_rate": 1.9378230797233347e-06, + "loss": 0.9274, + "step": 19502 + }, + { + "epoch": 3.55, + "learning_rate": 1.9377684746996726e-06, + "loss": 0.4113, + "step": 19503 + }, + { + "epoch": 3.55, + "learning_rate": 1.93771386967601e-06, + "loss": 1.888, + "step": 19504 + }, + { + "epoch": 3.55, + "learning_rate": 1.937659264652348e-06, + "loss": 0.8798, + "step": 19505 + }, + { + "epoch": 3.55, + "learning_rate": 1.937604659628686e-06, + "loss": 0.8749, + "step": 19506 + }, + { + "epoch": 3.55, + "learning_rate": 1.9375500546050235e-06, + "loss": 0.6596, + "step": 19507 + }, + { + "epoch": 3.55, + "learning_rate": 1.9374954495813615e-06, + "loss": 1.4086, + "step": 19508 + }, + { + "epoch": 3.55, + "learning_rate": 1.9374408445576994e-06, + "loss": 0.4717, + "step": 19509 + }, + { + "epoch": 3.55, + "learning_rate": 1.9373862395340374e-06, + "loss": 1.33, + "step": 19510 + }, + { + "epoch": 3.55, + "learning_rate": 1.9373316345103753e-06, + "loss": 1.6913, + "step": 19511 + }, + { + "epoch": 3.55, + "learning_rate": 1.937277029486713e-06, + "loss": 0.0093, + "step": 19512 + }, + { + "epoch": 3.55, + "learning_rate": 1.9372224244630508e-06, + "loss": 0.0131, + "step": 19513 + }, + { + "epoch": 3.55, + "learning_rate": 1.9371678194393887e-06, + "loss": 1.31, + "step": 19514 + }, + { + "epoch": 3.55, + "learning_rate": 1.9371132144157262e-06, + "loss": 1.3447, + "step": 19515 + }, + { + "epoch": 3.55, + "learning_rate": 1.937058609392064e-06, + "loss": 0.7339, + "step": 19516 + }, + { + "epoch": 3.55, + "learning_rate": 1.937004004368402e-06, + "loss": 2.1191, + "step": 19517 + }, + { + "epoch": 3.55, + "learning_rate": 1.9369493993447396e-06, + "loss": 4.2093, + "step": 19518 + }, + { + "epoch": 3.55, + "learning_rate": 1.9368947943210776e-06, + "loss": 0.6775, + "step": 19519 + }, + { + "epoch": 3.55, + "learning_rate": 1.9368401892974155e-06, + "loss": 0.1828, + "step": 19520 + }, + { + "epoch": 3.55, + "learning_rate": 1.936785584273753e-06, + "loss": 0.4515, + "step": 19521 + }, + { + "epoch": 3.55, + "learning_rate": 1.936730979250091e-06, + "loss": 0.1842, + "step": 19522 + }, + { + "epoch": 3.55, + "learning_rate": 1.936676374226429e-06, + "loss": 3.0918, + "step": 19523 + }, + { + "epoch": 3.55, + "learning_rate": 1.936621769202767e-06, + "loss": 0.3878, + "step": 19524 + }, + { + "epoch": 3.55, + "learning_rate": 1.936567164179105e-06, + "loss": 0.0287, + "step": 19525 + }, + { + "epoch": 3.55, + "learning_rate": 1.9365125591554423e-06, + "loss": 0.4015, + "step": 19526 + }, + { + "epoch": 3.55, + "learning_rate": 1.9364579541317803e-06, + "loss": 0.0242, + "step": 19527 + }, + { + "epoch": 3.55, + "learning_rate": 1.9364033491081182e-06, + "loss": 0.0576, + "step": 19528 + }, + { + "epoch": 3.55, + "learning_rate": 1.9363487440844557e-06, + "loss": 2.5676, + "step": 19529 + }, + { + "epoch": 3.55, + "learning_rate": 1.9362941390607937e-06, + "loss": 0.1474, + "step": 19530 + }, + { + "epoch": 3.55, + "learning_rate": 1.9362395340371316e-06, + "loss": 0.2728, + "step": 19531 + }, + { + "epoch": 3.55, + "learning_rate": 1.936184929013469e-06, + "loss": 0.2878, + "step": 19532 + }, + { + "epoch": 3.55, + "learning_rate": 1.936130323989807e-06, + "loss": 0.6286, + "step": 19533 + }, + { + "epoch": 3.55, + "learning_rate": 1.9360757189661446e-06, + "loss": 2.8186, + "step": 19534 + }, + { + "epoch": 3.55, + "learning_rate": 1.9360211139424826e-06, + "loss": 1.9453, + "step": 19535 + }, + { + "epoch": 3.55, + "learning_rate": 1.9359665089188205e-06, + "loss": 1.2825, + "step": 19536 + }, + { + "epoch": 3.55, + "learning_rate": 1.9359119038951584e-06, + "loss": 1.4886, + "step": 19537 + }, + { + "epoch": 3.55, + "learning_rate": 1.9358572988714964e-06, + "loss": 0.3243, + "step": 19538 + }, + { + "epoch": 3.55, + "learning_rate": 1.9358026938478343e-06, + "loss": 0.2228, + "step": 19539 + }, + { + "epoch": 3.55, + "learning_rate": 1.935748088824172e-06, + "loss": 0.7603, + "step": 19540 + }, + { + "epoch": 3.55, + "learning_rate": 1.93569348380051e-06, + "loss": 0.2366, + "step": 19541 + }, + { + "epoch": 3.55, + "learning_rate": 1.9356388787768477e-06, + "loss": 2.7436, + "step": 19542 + }, + { + "epoch": 3.55, + "learning_rate": 1.9355842737531853e-06, + "loss": 0.738, + "step": 19543 + }, + { + "epoch": 3.55, + "learning_rate": 1.935529668729523e-06, + "loss": 0.0641, + "step": 19544 + }, + { + "epoch": 3.55, + "learning_rate": 1.935475063705861e-06, + "loss": 0.01, + "step": 19545 + }, + { + "epoch": 3.55, + "learning_rate": 1.9354204586821987e-06, + "loss": 0.4368, + "step": 19546 + }, + { + "epoch": 3.55, + "learning_rate": 1.9353658536585366e-06, + "loss": 0.8518, + "step": 19547 + }, + { + "epoch": 3.55, + "learning_rate": 1.935311248634874e-06, + "loss": 1.4133, + "step": 19548 + }, + { + "epoch": 3.56, + "learning_rate": 1.935256643611212e-06, + "loss": 1.1855, + "step": 19549 + }, + { + "epoch": 3.56, + "learning_rate": 1.93520203858755e-06, + "loss": 0.4058, + "step": 19550 + }, + { + "epoch": 3.56, + "learning_rate": 1.935147433563888e-06, + "loss": 0.3923, + "step": 19551 + }, + { + "epoch": 3.56, + "learning_rate": 1.935092828540226e-06, + "loss": 0.1235, + "step": 19552 + }, + { + "epoch": 3.56, + "learning_rate": 1.935038223516564e-06, + "loss": 0.1917, + "step": 19553 + }, + { + "epoch": 3.56, + "learning_rate": 1.9349836184929014e-06, + "loss": 0.9241, + "step": 19554 + }, + { + "epoch": 3.56, + "learning_rate": 1.9349290134692393e-06, + "loss": 0.0167, + "step": 19555 + }, + { + "epoch": 3.56, + "learning_rate": 1.9348744084455773e-06, + "loss": 0.2058, + "step": 19556 + }, + { + "epoch": 3.56, + "learning_rate": 1.9348198034219148e-06, + "loss": 0.6148, + "step": 19557 + }, + { + "epoch": 3.56, + "learning_rate": 1.9347651983982527e-06, + "loss": 0.2417, + "step": 19558 + }, + { + "epoch": 3.56, + "learning_rate": 1.9347105933745907e-06, + "loss": 0.6747, + "step": 19559 + }, + { + "epoch": 3.56, + "learning_rate": 1.934655988350928e-06, + "loss": 1.2347, + "step": 19560 + }, + { + "epoch": 3.56, + "learning_rate": 1.934601383327266e-06, + "loss": 0.291, + "step": 19561 + }, + { + "epoch": 3.56, + "learning_rate": 1.9345467783036036e-06, + "loss": 0.2922, + "step": 19562 + }, + { + "epoch": 3.56, + "learning_rate": 1.9344921732799416e-06, + "loss": 0.9073, + "step": 19563 + }, + { + "epoch": 3.56, + "learning_rate": 1.9344375682562795e-06, + "loss": 1.5286, + "step": 19564 + }, + { + "epoch": 3.56, + "learning_rate": 1.9343829632326175e-06, + "loss": 0.6858, + "step": 19565 + }, + { + "epoch": 3.56, + "learning_rate": 1.9343283582089554e-06, + "loss": 1.4517, + "step": 19566 + }, + { + "epoch": 3.56, + "learning_rate": 1.9342737531852934e-06, + "loss": 1.0255, + "step": 19567 + }, + { + "epoch": 3.56, + "learning_rate": 1.934219148161631e-06, + "loss": 1.6706, + "step": 19568 + }, + { + "epoch": 3.56, + "learning_rate": 1.934164543137969e-06, + "loss": 2.1179, + "step": 19569 + }, + { + "epoch": 3.56, + "learning_rate": 1.9341099381143068e-06, + "loss": 1.5081, + "step": 19570 + }, + { + "epoch": 3.56, + "learning_rate": 1.9340553330906443e-06, + "loss": 2.4863, + "step": 19571 + }, + { + "epoch": 3.56, + "learning_rate": 1.9340007280669822e-06, + "loss": 3.3585, + "step": 19572 + }, + { + "epoch": 3.56, + "learning_rate": 1.93394612304332e-06, + "loss": 0.0434, + "step": 19573 + }, + { + "epoch": 3.56, + "learning_rate": 1.9338915180196577e-06, + "loss": 0.1202, + "step": 19574 + }, + { + "epoch": 3.56, + "learning_rate": 1.9338369129959956e-06, + "loss": 0.373, + "step": 19575 + }, + { + "epoch": 3.56, + "learning_rate": 1.933782307972333e-06, + "loss": 0.8945, + "step": 19576 + }, + { + "epoch": 3.56, + "learning_rate": 1.933727702948671e-06, + "loss": 2.7979, + "step": 19577 + }, + { + "epoch": 3.56, + "learning_rate": 1.933673097925009e-06, + "loss": 1.829, + "step": 19578 + }, + { + "epoch": 3.56, + "learning_rate": 1.933618492901347e-06, + "loss": 0.1802, + "step": 19579 + }, + { + "epoch": 3.56, + "learning_rate": 1.933563887877685e-06, + "loss": 2.2225, + "step": 19580 + }, + { + "epoch": 3.56, + "learning_rate": 1.933509282854023e-06, + "loss": 0.252, + "step": 19581 + }, + { + "epoch": 3.56, + "learning_rate": 1.9334546778303604e-06, + "loss": 0.0402, + "step": 19582 + }, + { + "epoch": 3.56, + "learning_rate": 1.9334000728066983e-06, + "loss": 1.2176, + "step": 19583 + }, + { + "epoch": 3.56, + "learning_rate": 1.9333454677830363e-06, + "loss": 0.1998, + "step": 19584 + }, + { + "epoch": 3.56, + "learning_rate": 1.933290862759374e-06, + "loss": 1.3123, + "step": 19585 + }, + { + "epoch": 3.56, + "learning_rate": 1.9332362577357117e-06, + "loss": 0.9728, + "step": 19586 + }, + { + "epoch": 3.56, + "learning_rate": 1.9331816527120497e-06, + "loss": 1.5572, + "step": 19587 + }, + { + "epoch": 3.56, + "learning_rate": 1.933127047688387e-06, + "loss": 0.4545, + "step": 19588 + }, + { + "epoch": 3.56, + "learning_rate": 1.933072442664725e-06, + "loss": 0.2417, + "step": 19589 + }, + { + "epoch": 3.56, + "learning_rate": 1.9330178376410627e-06, + "loss": 0.2294, + "step": 19590 + }, + { + "epoch": 3.56, + "learning_rate": 1.9329632326174006e-06, + "loss": 1.0044, + "step": 19591 + }, + { + "epoch": 3.56, + "learning_rate": 1.9329086275937386e-06, + "loss": 0.494, + "step": 19592 + }, + { + "epoch": 3.56, + "learning_rate": 1.9328540225700765e-06, + "loss": 0.0105, + "step": 19593 + }, + { + "epoch": 3.56, + "learning_rate": 1.9327994175464144e-06, + "loss": 0.3958, + "step": 19594 + }, + { + "epoch": 3.56, + "learning_rate": 1.9327448125227524e-06, + "loss": 1.0916, + "step": 19595 + }, + { + "epoch": 3.56, + "learning_rate": 1.93269020749909e-06, + "loss": 0.735, + "step": 19596 + }, + { + "epoch": 3.56, + "learning_rate": 1.932635602475428e-06, + "loss": 2.7571, + "step": 19597 + }, + { + "epoch": 3.56, + "learning_rate": 1.932580997451766e-06, + "loss": 0.6662, + "step": 19598 + }, + { + "epoch": 3.56, + "learning_rate": 1.9325263924281033e-06, + "loss": 0.677, + "step": 19599 + }, + { + "epoch": 3.56, + "learning_rate": 1.9324717874044413e-06, + "loss": 0.3104, + "step": 19600 + }, + { + "epoch": 3.56, + "learning_rate": 1.932417182380779e-06, + "loss": 0.7696, + "step": 19601 + }, + { + "epoch": 3.56, + "learning_rate": 1.9323625773571167e-06, + "loss": 0.6599, + "step": 19602 + }, + { + "epoch": 3.56, + "learning_rate": 1.9323079723334547e-06, + "loss": 3.7076, + "step": 19603 + }, + { + "epoch": 3.57, + "learning_rate": 1.9322533673097926e-06, + "loss": 0.4345, + "step": 19604 + }, + { + "epoch": 3.57, + "learning_rate": 1.9321987622861305e-06, + "loss": 0.1008, + "step": 19605 + }, + { + "epoch": 3.57, + "learning_rate": 1.9321441572624685e-06, + "loss": 0.684, + "step": 19606 + }, + { + "epoch": 3.57, + "learning_rate": 1.932089552238806e-06, + "loss": 1.2449, + "step": 19607 + }, + { + "epoch": 3.57, + "learning_rate": 1.932034947215144e-06, + "loss": 0.0366, + "step": 19608 + }, + { + "epoch": 3.57, + "learning_rate": 1.931980342191482e-06, + "loss": 0.0192, + "step": 19609 + }, + { + "epoch": 3.57, + "learning_rate": 1.9319257371678194e-06, + "loss": 3.0358, + "step": 19610 + }, + { + "epoch": 3.57, + "learning_rate": 1.9318711321441574e-06, + "loss": 0.3301, + "step": 19611 + }, + { + "epoch": 3.57, + "learning_rate": 1.9318165271204953e-06, + "loss": 2.1593, + "step": 19612 + }, + { + "epoch": 3.57, + "learning_rate": 1.931761922096833e-06, + "loss": 0.0293, + "step": 19613 + }, + { + "epoch": 3.57, + "learning_rate": 1.9317073170731708e-06, + "loss": 0.031, + "step": 19614 + }, + { + "epoch": 3.57, + "learning_rate": 1.9316527120495083e-06, + "loss": 1.9146, + "step": 19615 + }, + { + "epoch": 3.57, + "learning_rate": 1.9315981070258462e-06, + "loss": 0.0588, + "step": 19616 + }, + { + "epoch": 3.57, + "learning_rate": 1.931543502002184e-06, + "loss": 0.4787, + "step": 19617 + }, + { + "epoch": 3.57, + "learning_rate": 1.931488896978522e-06, + "loss": 0.6705, + "step": 19618 + }, + { + "epoch": 3.57, + "learning_rate": 1.93143429195486e-06, + "loss": 0.0681, + "step": 19619 + }, + { + "epoch": 3.57, + "learning_rate": 1.931379686931198e-06, + "loss": 0.0529, + "step": 19620 + }, + { + "epoch": 3.57, + "learning_rate": 1.9313250819075355e-06, + "loss": 1.1476, + "step": 19621 + }, + { + "epoch": 3.57, + "learning_rate": 1.9312704768838735e-06, + "loss": 0.5055, + "step": 19622 + }, + { + "epoch": 3.57, + "learning_rate": 1.9312158718602114e-06, + "loss": 0.4506, + "step": 19623 + }, + { + "epoch": 3.57, + "learning_rate": 1.931161266836549e-06, + "loss": 0.4817, + "step": 19624 + }, + { + "epoch": 3.57, + "learning_rate": 1.931106661812887e-06, + "loss": 0.5885, + "step": 19625 + }, + { + "epoch": 3.57, + "learning_rate": 1.931052056789225e-06, + "loss": 1.9919, + "step": 19626 + }, + { + "epoch": 3.57, + "learning_rate": 1.9309974517655623e-06, + "loss": 0.4327, + "step": 19627 + }, + { + "epoch": 3.57, + "learning_rate": 1.9309428467419003e-06, + "loss": 0.5253, + "step": 19628 + }, + { + "epoch": 3.57, + "learning_rate": 1.930888241718238e-06, + "loss": 0.4409, + "step": 19629 + }, + { + "epoch": 3.57, + "learning_rate": 1.9308336366945757e-06, + "loss": 0.3248, + "step": 19630 + }, + { + "epoch": 3.57, + "learning_rate": 1.9307790316709137e-06, + "loss": 0.027, + "step": 19631 + }, + { + "epoch": 3.57, + "learning_rate": 1.9307244266472516e-06, + "loss": 0.2417, + "step": 19632 + }, + { + "epoch": 3.57, + "learning_rate": 1.9306698216235896e-06, + "loss": 0.7178, + "step": 19633 + }, + { + "epoch": 3.57, + "learning_rate": 1.9306152165999275e-06, + "loss": 0.7214, + "step": 19634 + }, + { + "epoch": 3.57, + "learning_rate": 1.930560611576265e-06, + "loss": 1.9344, + "step": 19635 + }, + { + "epoch": 3.57, + "learning_rate": 1.930506006552603e-06, + "loss": 1.5231, + "step": 19636 + }, + { + "epoch": 3.57, + "learning_rate": 1.930451401528941e-06, + "loss": 0.0717, + "step": 19637 + }, + { + "epoch": 3.57, + "learning_rate": 1.9303967965052784e-06, + "loss": 0.5689, + "step": 19638 + }, + { + "epoch": 3.57, + "learning_rate": 1.9303421914816164e-06, + "loss": 1.3572, + "step": 19639 + }, + { + "epoch": 3.57, + "learning_rate": 1.9302875864579543e-06, + "loss": 1.9211, + "step": 19640 + }, + { + "epoch": 3.57, + "learning_rate": 1.930232981434292e-06, + "loss": 0.6107, + "step": 19641 + }, + { + "epoch": 3.57, + "learning_rate": 1.93017837641063e-06, + "loss": 0.0202, + "step": 19642 + }, + { + "epoch": 3.57, + "learning_rate": 1.9301237713869673e-06, + "loss": 0.1053, + "step": 19643 + }, + { + "epoch": 3.57, + "learning_rate": 1.9300691663633053e-06, + "loss": 0.0113, + "step": 19644 + }, + { + "epoch": 3.57, + "learning_rate": 1.930014561339643e-06, + "loss": 0.8593, + "step": 19645 + }, + { + "epoch": 3.57, + "learning_rate": 1.929959956315981e-06, + "loss": 0.2562, + "step": 19646 + }, + { + "epoch": 3.57, + "learning_rate": 1.929905351292319e-06, + "loss": 1.4388, + "step": 19647 + }, + { + "epoch": 3.57, + "learning_rate": 1.929850746268657e-06, + "loss": 2.0519, + "step": 19648 + }, + { + "epoch": 3.57, + "learning_rate": 1.9297961412449946e-06, + "loss": 0.1078, + "step": 19649 + }, + { + "epoch": 3.57, + "learning_rate": 1.9297415362213325e-06, + "loss": 0.2395, + "step": 19650 + }, + { + "epoch": 3.57, + "learning_rate": 1.9296869311976704e-06, + "loss": 0.5033, + "step": 19651 + }, + { + "epoch": 3.57, + "learning_rate": 1.929632326174008e-06, + "loss": 0.6194, + "step": 19652 + }, + { + "epoch": 3.57, + "learning_rate": 1.929577721150346e-06, + "loss": 0.292, + "step": 19653 + }, + { + "epoch": 3.57, + "learning_rate": 1.929523116126684e-06, + "loss": 0.2887, + "step": 19654 + }, + { + "epoch": 3.57, + "learning_rate": 1.9294685111030214e-06, + "loss": 1.871, + "step": 19655 + }, + { + "epoch": 3.57, + "learning_rate": 1.9294139060793593e-06, + "loss": 0.0104, + "step": 19656 + }, + { + "epoch": 3.57, + "learning_rate": 1.929359301055697e-06, + "loss": 0.0162, + "step": 19657 + }, + { + "epoch": 3.57, + "learning_rate": 1.9293046960320348e-06, + "loss": 0.6745, + "step": 19658 + }, + { + "epoch": 3.58, + "learning_rate": 1.9292500910083727e-06, + "loss": 0.0157, + "step": 19659 + }, + { + "epoch": 3.58, + "learning_rate": 1.9291954859847107e-06, + "loss": 1.1114, + "step": 19660 + }, + { + "epoch": 3.58, + "learning_rate": 1.9291408809610486e-06, + "loss": 0.0379, + "step": 19661 + }, + { + "epoch": 3.58, + "learning_rate": 1.9290862759373865e-06, + "loss": 0.5497, + "step": 19662 + }, + { + "epoch": 3.58, + "learning_rate": 1.929031670913724e-06, + "loss": 0.0098, + "step": 19663 + }, + { + "epoch": 3.58, + "learning_rate": 1.928977065890062e-06, + "loss": 0.1742, + "step": 19664 + }, + { + "epoch": 3.58, + "learning_rate": 1.9289224608664e-06, + "loss": 0.5117, + "step": 19665 + }, + { + "epoch": 3.58, + "learning_rate": 1.9288678558427375e-06, + "loss": 0.1736, + "step": 19666 + }, + { + "epoch": 3.58, + "learning_rate": 1.9288132508190754e-06, + "loss": 1.7842, + "step": 19667 + }, + { + "epoch": 3.58, + "learning_rate": 1.9287586457954134e-06, + "loss": 0.0752, + "step": 19668 + }, + { + "epoch": 3.58, + "learning_rate": 1.928704040771751e-06, + "loss": 1.4518, + "step": 19669 + }, + { + "epoch": 3.58, + "learning_rate": 1.928649435748089e-06, + "loss": 0.0205, + "step": 19670 + }, + { + "epoch": 3.58, + "learning_rate": 1.9285948307244263e-06, + "loss": 0.2911, + "step": 19671 + }, + { + "epoch": 3.58, + "learning_rate": 1.9285402257007643e-06, + "loss": 0.7367, + "step": 19672 + }, + { + "epoch": 3.58, + "learning_rate": 1.9284856206771022e-06, + "loss": 0.9215, + "step": 19673 + }, + { + "epoch": 3.58, + "learning_rate": 1.92843101565344e-06, + "loss": 0.2485, + "step": 19674 + }, + { + "epoch": 3.58, + "learning_rate": 1.928376410629778e-06, + "loss": 1.487, + "step": 19675 + }, + { + "epoch": 3.58, + "learning_rate": 1.928321805606116e-06, + "loss": 2.1376, + "step": 19676 + }, + { + "epoch": 3.58, + "learning_rate": 1.9282672005824536e-06, + "loss": 0.0176, + "step": 19677 + }, + { + "epoch": 3.58, + "learning_rate": 1.9282125955587915e-06, + "loss": 0.9755, + "step": 19678 + }, + { + "epoch": 3.58, + "learning_rate": 1.9281579905351295e-06, + "loss": 0.359, + "step": 19679 + }, + { + "epoch": 3.58, + "learning_rate": 1.928103385511467e-06, + "loss": 4.2475, + "step": 19680 + }, + { + "epoch": 3.58, + "learning_rate": 1.928048780487805e-06, + "loss": 0.821, + "step": 19681 + }, + { + "epoch": 3.58, + "learning_rate": 1.927994175464143e-06, + "loss": 0.0841, + "step": 19682 + }, + { + "epoch": 3.58, + "learning_rate": 1.9279395704404804e-06, + "loss": 0.1342, + "step": 19683 + }, + { + "epoch": 3.58, + "learning_rate": 1.9278849654168183e-06, + "loss": 1.0919, + "step": 19684 + }, + { + "epoch": 3.58, + "learning_rate": 1.9278303603931563e-06, + "loss": 0.2417, + "step": 19685 + }, + { + "epoch": 3.58, + "learning_rate": 1.9277757553694942e-06, + "loss": 0.1889, + "step": 19686 + }, + { + "epoch": 3.58, + "learning_rate": 1.927721150345832e-06, + "loss": 0.0207, + "step": 19687 + }, + { + "epoch": 3.58, + "learning_rate": 1.9276665453221697e-06, + "loss": 1.646, + "step": 19688 + }, + { + "epoch": 3.58, + "learning_rate": 1.9276119402985076e-06, + "loss": 0.5533, + "step": 19689 + }, + { + "epoch": 3.58, + "learning_rate": 1.9275573352748456e-06, + "loss": 0.7809, + "step": 19690 + }, + { + "epoch": 3.58, + "learning_rate": 1.927502730251183e-06, + "loss": 0.6424, + "step": 19691 + }, + { + "epoch": 3.58, + "learning_rate": 1.927448125227521e-06, + "loss": 0.6203, + "step": 19692 + }, + { + "epoch": 3.58, + "learning_rate": 1.927393520203859e-06, + "loss": 0.9625, + "step": 19693 + }, + { + "epoch": 3.58, + "learning_rate": 1.9273389151801965e-06, + "loss": 0.0918, + "step": 19694 + }, + { + "epoch": 3.58, + "learning_rate": 1.9272843101565344e-06, + "loss": 3.2461, + "step": 19695 + }, + { + "epoch": 3.58, + "learning_rate": 1.927229705132872e-06, + "loss": 1.087, + "step": 19696 + }, + { + "epoch": 3.58, + "learning_rate": 1.92717510010921e-06, + "loss": 0.0128, + "step": 19697 + }, + { + "epoch": 3.58, + "learning_rate": 1.927120495085548e-06, + "loss": 0.2027, + "step": 19698 + }, + { + "epoch": 3.58, + "learning_rate": 1.927065890061886e-06, + "loss": 1.0187, + "step": 19699 + }, + { + "epoch": 3.58, + "learning_rate": 1.9270112850382237e-06, + "loss": 0.3554, + "step": 19700 + }, + { + "epoch": 3.58, + "learning_rate": 1.9269566800145617e-06, + "loss": 1.0111, + "step": 19701 + }, + { + "epoch": 3.58, + "learning_rate": 1.926902074990899e-06, + "loss": 0.0909, + "step": 19702 + }, + { + "epoch": 3.58, + "learning_rate": 1.926847469967237e-06, + "loss": 0.29, + "step": 19703 + }, + { + "epoch": 3.58, + "learning_rate": 1.926792864943575e-06, + "loss": 0.2001, + "step": 19704 + }, + { + "epoch": 3.58, + "learning_rate": 1.9267382599199126e-06, + "loss": 1.0369, + "step": 19705 + }, + { + "epoch": 3.58, + "learning_rate": 1.9266836548962505e-06, + "loss": 0.2732, + "step": 19706 + }, + { + "epoch": 3.58, + "learning_rate": 1.9266290498725885e-06, + "loss": 0.594, + "step": 19707 + }, + { + "epoch": 3.58, + "learning_rate": 1.926574444848926e-06, + "loss": 0.0114, + "step": 19708 + }, + { + "epoch": 3.58, + "learning_rate": 1.926519839825264e-06, + "loss": 0.6737, + "step": 19709 + }, + { + "epoch": 3.58, + "learning_rate": 1.9264652348016015e-06, + "loss": 0.6873, + "step": 19710 + }, + { + "epoch": 3.58, + "learning_rate": 1.9264106297779394e-06, + "loss": 0.5914, + "step": 19711 + }, + { + "epoch": 3.58, + "learning_rate": 1.9263560247542774e-06, + "loss": 0.0132, + "step": 19712 + }, + { + "epoch": 3.58, + "learning_rate": 1.9263014197306153e-06, + "loss": 0.5582, + "step": 19713 + }, + { + "epoch": 3.59, + "learning_rate": 1.9262468147069532e-06, + "loss": 1.6439, + "step": 19714 + }, + { + "epoch": 3.59, + "learning_rate": 1.926192209683291e-06, + "loss": 0.1984, + "step": 19715 + }, + { + "epoch": 3.59, + "learning_rate": 1.9261376046596287e-06, + "loss": 0.2084, + "step": 19716 + }, + { + "epoch": 3.59, + "learning_rate": 1.9260829996359667e-06, + "loss": 0.5481, + "step": 19717 + }, + { + "epoch": 3.59, + "learning_rate": 1.9260283946123046e-06, + "loss": 1.7514, + "step": 19718 + }, + { + "epoch": 3.59, + "learning_rate": 1.925973789588642e-06, + "loss": 0.0473, + "step": 19719 + }, + { + "epoch": 3.59, + "learning_rate": 1.92591918456498e-06, + "loss": 1.1276, + "step": 19720 + }, + { + "epoch": 3.59, + "learning_rate": 1.925864579541318e-06, + "loss": 0.7629, + "step": 19721 + }, + { + "epoch": 3.59, + "learning_rate": 1.9258099745176555e-06, + "loss": 0.0229, + "step": 19722 + }, + { + "epoch": 3.59, + "learning_rate": 1.9257553694939935e-06, + "loss": 0.3325, + "step": 19723 + }, + { + "epoch": 3.59, + "learning_rate": 1.925700764470331e-06, + "loss": 0.9972, + "step": 19724 + }, + { + "epoch": 3.59, + "learning_rate": 1.925646159446669e-06, + "loss": 0.4557, + "step": 19725 + }, + { + "epoch": 3.59, + "learning_rate": 1.925591554423007e-06, + "loss": 1.8698, + "step": 19726 + }, + { + "epoch": 3.59, + "learning_rate": 1.925536949399345e-06, + "loss": 1.4832, + "step": 19727 + }, + { + "epoch": 3.59, + "learning_rate": 1.9254823443756828e-06, + "loss": 1.4508, + "step": 19728 + }, + { + "epoch": 3.59, + "learning_rate": 1.9254277393520207e-06, + "loss": 0.3589, + "step": 19729 + }, + { + "epoch": 3.59, + "learning_rate": 1.9253731343283582e-06, + "loss": 1.0029, + "step": 19730 + }, + { + "epoch": 3.59, + "learning_rate": 1.925318529304696e-06, + "loss": 0.0185, + "step": 19731 + }, + { + "epoch": 3.59, + "learning_rate": 1.925263924281034e-06, + "loss": 1.624, + "step": 19732 + }, + { + "epoch": 3.59, + "learning_rate": 1.9252093192573716e-06, + "loss": 1.1352, + "step": 19733 + }, + { + "epoch": 3.59, + "learning_rate": 1.9251547142337096e-06, + "loss": 1.1827, + "step": 19734 + }, + { + "epoch": 3.59, + "learning_rate": 1.9251001092100475e-06, + "loss": 0.0242, + "step": 19735 + }, + { + "epoch": 3.59, + "learning_rate": 1.925045504186385e-06, + "loss": 0.5424, + "step": 19736 + }, + { + "epoch": 3.59, + "learning_rate": 1.924990899162723e-06, + "loss": 0.01, + "step": 19737 + }, + { + "epoch": 3.59, + "learning_rate": 1.9249362941390605e-06, + "loss": 0.0127, + "step": 19738 + }, + { + "epoch": 3.59, + "learning_rate": 1.9248816891153984e-06, + "loss": 0.4306, + "step": 19739 + }, + { + "epoch": 3.59, + "learning_rate": 1.9248270840917364e-06, + "loss": 0.4136, + "step": 19740 + }, + { + "epoch": 3.59, + "learning_rate": 1.9247724790680743e-06, + "loss": 0.2236, + "step": 19741 + }, + { + "epoch": 3.59, + "learning_rate": 1.9247178740444123e-06, + "loss": 0.0604, + "step": 19742 + }, + { + "epoch": 3.59, + "learning_rate": 1.9246632690207502e-06, + "loss": 0.118, + "step": 19743 + }, + { + "epoch": 3.59, + "learning_rate": 1.9246086639970877e-06, + "loss": 0.0276, + "step": 19744 + }, + { + "epoch": 3.59, + "learning_rate": 1.9245540589734257e-06, + "loss": 1.0279, + "step": 19745 + }, + { + "epoch": 3.59, + "learning_rate": 1.9244994539497636e-06, + "loss": 0.0171, + "step": 19746 + }, + { + "epoch": 3.59, + "learning_rate": 1.924444848926101e-06, + "loss": 0.3045, + "step": 19747 + }, + { + "epoch": 3.59, + "learning_rate": 1.924390243902439e-06, + "loss": 0.0273, + "step": 19748 + }, + { + "epoch": 3.59, + "learning_rate": 1.924335638878777e-06, + "loss": 0.8208, + "step": 19749 + }, + { + "epoch": 3.59, + "learning_rate": 1.9242810338551146e-06, + "loss": 1.3357, + "step": 19750 + }, + { + "epoch": 3.59, + "learning_rate": 1.9242264288314525e-06, + "loss": 1.3738, + "step": 19751 + }, + { + "epoch": 3.59, + "learning_rate": 1.92417182380779e-06, + "loss": 0.797, + "step": 19752 + }, + { + "epoch": 3.59, + "learning_rate": 1.924117218784128e-06, + "loss": 0.2152, + "step": 19753 + }, + { + "epoch": 3.59, + "learning_rate": 1.924062613760466e-06, + "loss": 0.3732, + "step": 19754 + }, + { + "epoch": 3.59, + "learning_rate": 1.924008008736804e-06, + "loss": 0.1802, + "step": 19755 + }, + { + "epoch": 3.59, + "learning_rate": 1.923953403713142e-06, + "loss": 0.0125, + "step": 19756 + }, + { + "epoch": 3.59, + "learning_rate": 1.9238987986894797e-06, + "loss": 0.1306, + "step": 19757 + }, + { + "epoch": 3.59, + "learning_rate": 1.9238441936658173e-06, + "loss": 0.7645, + "step": 19758 + }, + { + "epoch": 3.59, + "learning_rate": 1.923789588642155e-06, + "loss": 0.0408, + "step": 19759 + }, + { + "epoch": 3.59, + "learning_rate": 1.923734983618493e-06, + "loss": 1.5247, + "step": 19760 + }, + { + "epoch": 3.59, + "learning_rate": 1.9236803785948307e-06, + "loss": 0.5992, + "step": 19761 + }, + { + "epoch": 3.59, + "learning_rate": 1.9236257735711686e-06, + "loss": 0.3818, + "step": 19762 + }, + { + "epoch": 3.59, + "learning_rate": 1.923571168547506e-06, + "loss": 0.349, + "step": 19763 + }, + { + "epoch": 3.59, + "learning_rate": 1.923516563523844e-06, + "loss": 0.9408, + "step": 19764 + }, + { + "epoch": 3.59, + "learning_rate": 1.923461958500182e-06, + "loss": 1.0256, + "step": 19765 + }, + { + "epoch": 3.59, + "learning_rate": 1.92340735347652e-06, + "loss": 0.1583, + "step": 19766 + }, + { + "epoch": 3.59, + "learning_rate": 1.923352748452858e-06, + "loss": 0.4462, + "step": 19767 + }, + { + "epoch": 3.59, + "learning_rate": 1.923298143429196e-06, + "loss": 1.1961, + "step": 19768 + }, + { + "epoch": 3.6, + "learning_rate": 1.9232435384055334e-06, + "loss": 0.4177, + "step": 19769 + }, + { + "epoch": 3.6, + "learning_rate": 1.9231889333818713e-06, + "loss": 0.9469, + "step": 19770 + }, + { + "epoch": 3.6, + "learning_rate": 1.9231343283582092e-06, + "loss": 1.0471, + "step": 19771 + }, + { + "epoch": 3.6, + "learning_rate": 1.9230797233345468e-06, + "loss": 0.3324, + "step": 19772 + }, + { + "epoch": 3.6, + "learning_rate": 1.9230251183108847e-06, + "loss": 0.5211, + "step": 19773 + }, + { + "epoch": 3.6, + "learning_rate": 1.9229705132872227e-06, + "loss": 0.1832, + "step": 19774 + }, + { + "epoch": 3.6, + "learning_rate": 1.92291590826356e-06, + "loss": 4.8846, + "step": 19775 + }, + { + "epoch": 3.6, + "learning_rate": 1.922861303239898e-06, + "loss": 0.0407, + "step": 19776 + }, + { + "epoch": 3.6, + "learning_rate": 1.9228066982162356e-06, + "loss": 1.1265, + "step": 19777 + }, + { + "epoch": 3.6, + "learning_rate": 1.9227520931925736e-06, + "loss": 0.1344, + "step": 19778 + }, + { + "epoch": 3.6, + "learning_rate": 1.9226974881689115e-06, + "loss": 4.2593, + "step": 19779 + }, + { + "epoch": 3.6, + "learning_rate": 1.9226428831452495e-06, + "loss": 0.0535, + "step": 19780 + }, + { + "epoch": 3.6, + "learning_rate": 1.9225882781215874e-06, + "loss": 0.3936, + "step": 19781 + }, + { + "epoch": 3.6, + "learning_rate": 1.9225336730979254e-06, + "loss": 1.6849, + "step": 19782 + }, + { + "epoch": 3.6, + "learning_rate": 1.922479068074263e-06, + "loss": 2.499, + "step": 19783 + }, + { + "epoch": 3.6, + "learning_rate": 1.922424463050601e-06, + "loss": 0.3994, + "step": 19784 + }, + { + "epoch": 3.6, + "learning_rate": 1.9223698580269388e-06, + "loss": 0.4687, + "step": 19785 + }, + { + "epoch": 3.6, + "learning_rate": 1.9223152530032763e-06, + "loss": 0.6362, + "step": 19786 + }, + { + "epoch": 3.6, + "learning_rate": 1.9222606479796142e-06, + "loss": 1.1348, + "step": 19787 + }, + { + "epoch": 3.6, + "learning_rate": 1.922206042955952e-06, + "loss": 0.0133, + "step": 19788 + }, + { + "epoch": 3.6, + "learning_rate": 1.9221514379322897e-06, + "loss": 0.6205, + "step": 19789 + }, + { + "epoch": 3.6, + "learning_rate": 1.9220968329086276e-06, + "loss": 1.7856, + "step": 19790 + }, + { + "epoch": 3.6, + "learning_rate": 1.922042227884965e-06, + "loss": 0.2653, + "step": 19791 + }, + { + "epoch": 3.6, + "learning_rate": 1.921987622861303e-06, + "loss": 1.0023, + "step": 19792 + }, + { + "epoch": 3.6, + "learning_rate": 1.921933017837641e-06, + "loss": 1.6289, + "step": 19793 + }, + { + "epoch": 3.6, + "learning_rate": 1.921878412813979e-06, + "loss": 0.1551, + "step": 19794 + }, + { + "epoch": 3.6, + "learning_rate": 1.921823807790317e-06, + "loss": 0.395, + "step": 19795 + }, + { + "epoch": 3.6, + "learning_rate": 1.921769202766655e-06, + "loss": 1.0786, + "step": 19796 + }, + { + "epoch": 3.6, + "learning_rate": 1.9217145977429924e-06, + "loss": 1.0711, + "step": 19797 + }, + { + "epoch": 3.6, + "learning_rate": 1.9216599927193303e-06, + "loss": 0.6717, + "step": 19798 + }, + { + "epoch": 3.6, + "learning_rate": 1.9216053876956683e-06, + "loss": 0.391, + "step": 19799 + }, + { + "epoch": 3.6, + "learning_rate": 1.921550782672006e-06, + "loss": 0.0113, + "step": 19800 + }, + { + "epoch": 3.6, + "learning_rate": 1.9214961776483437e-06, + "loss": 1.083, + "step": 19801 + }, + { + "epoch": 3.6, + "learning_rate": 1.9214415726246817e-06, + "loss": 2.5798, + "step": 19802 + }, + { + "epoch": 3.6, + "learning_rate": 1.921386967601019e-06, + "loss": 0.7009, + "step": 19803 + }, + { + "epoch": 3.6, + "learning_rate": 1.921332362577357e-06, + "loss": 3.5402, + "step": 19804 + }, + { + "epoch": 3.6, + "learning_rate": 1.9212777575536947e-06, + "loss": 0.0349, + "step": 19805 + }, + { + "epoch": 3.6, + "learning_rate": 1.9212231525300326e-06, + "loss": 1.2312, + "step": 19806 + }, + { + "epoch": 3.6, + "learning_rate": 1.9211685475063706e-06, + "loss": 0.0287, + "step": 19807 + }, + { + "epoch": 3.6, + "learning_rate": 1.9211139424827085e-06, + "loss": 0.3165, + "step": 19808 + }, + { + "epoch": 3.6, + "learning_rate": 1.9210593374590464e-06, + "loss": 1.0056, + "step": 19809 + }, + { + "epoch": 3.6, + "learning_rate": 1.9210047324353844e-06, + "loss": 1.5808, + "step": 19810 + }, + { + "epoch": 3.6, + "learning_rate": 1.920950127411722e-06, + "loss": 0.3209, + "step": 19811 + }, + { + "epoch": 3.6, + "learning_rate": 1.92089552238806e-06, + "loss": 0.5725, + "step": 19812 + }, + { + "epoch": 3.6, + "learning_rate": 1.9208409173643978e-06, + "loss": 1.5622, + "step": 19813 + }, + { + "epoch": 3.6, + "learning_rate": 1.9207863123407353e-06, + "loss": 0.0238, + "step": 19814 + }, + { + "epoch": 3.6, + "learning_rate": 1.9207317073170733e-06, + "loss": 0.5105, + "step": 19815 + }, + { + "epoch": 3.6, + "learning_rate": 1.920677102293411e-06, + "loss": 0.019, + "step": 19816 + }, + { + "epoch": 3.6, + "learning_rate": 1.9206224972697487e-06, + "loss": 1.3754, + "step": 19817 + }, + { + "epoch": 3.6, + "learning_rate": 1.9205678922460867e-06, + "loss": 1.6078, + "step": 19818 + }, + { + "epoch": 3.6, + "learning_rate": 1.920513287222424e-06, + "loss": 0.7408, + "step": 19819 + }, + { + "epoch": 3.6, + "learning_rate": 1.920458682198762e-06, + "loss": 0.2427, + "step": 19820 + }, + { + "epoch": 3.6, + "learning_rate": 1.9204040771751e-06, + "loss": 0.6059, + "step": 19821 + }, + { + "epoch": 3.6, + "learning_rate": 1.920349472151438e-06, + "loss": 2.1727, + "step": 19822 + }, + { + "epoch": 3.6, + "learning_rate": 1.920294867127776e-06, + "loss": 0.3227, + "step": 19823 + }, + { + "epoch": 3.61, + "learning_rate": 1.920240262104114e-06, + "loss": 0.3332, + "step": 19824 + }, + { + "epoch": 3.61, + "learning_rate": 1.9201856570804514e-06, + "loss": 0.8387, + "step": 19825 + }, + { + "epoch": 3.61, + "learning_rate": 1.9201310520567894e-06, + "loss": 0.2076, + "step": 19826 + }, + { + "epoch": 3.61, + "learning_rate": 1.9200764470331273e-06, + "loss": 2.2862, + "step": 19827 + }, + { + "epoch": 3.61, + "learning_rate": 1.920021842009465e-06, + "loss": 0.2449, + "step": 19828 + }, + { + "epoch": 3.61, + "learning_rate": 1.9199672369858028e-06, + "loss": 0.4551, + "step": 19829 + }, + { + "epoch": 3.61, + "learning_rate": 1.9199126319621407e-06, + "loss": 0.6032, + "step": 19830 + }, + { + "epoch": 3.61, + "learning_rate": 1.9198580269384782e-06, + "loss": 0.2711, + "step": 19831 + }, + { + "epoch": 3.61, + "learning_rate": 1.919803421914816e-06, + "loss": 0.1155, + "step": 19832 + }, + { + "epoch": 3.61, + "learning_rate": 1.9197488168911537e-06, + "loss": 0.4366, + "step": 19833 + }, + { + "epoch": 3.61, + "learning_rate": 1.9196942118674916e-06, + "loss": 0.0143, + "step": 19834 + }, + { + "epoch": 3.61, + "learning_rate": 1.9196396068438296e-06, + "loss": 2.9289, + "step": 19835 + }, + { + "epoch": 3.61, + "learning_rate": 1.9195850018201675e-06, + "loss": 2.7541, + "step": 19836 + }, + { + "epoch": 3.61, + "learning_rate": 1.9195303967965055e-06, + "loss": 0.3284, + "step": 19837 + }, + { + "epoch": 3.61, + "learning_rate": 1.9194757917728434e-06, + "loss": 0.7071, + "step": 19838 + }, + { + "epoch": 3.61, + "learning_rate": 1.919421186749181e-06, + "loss": 0.0695, + "step": 19839 + }, + { + "epoch": 3.61, + "learning_rate": 1.919366581725519e-06, + "loss": 0.0219, + "step": 19840 + }, + { + "epoch": 3.61, + "learning_rate": 1.919311976701857e-06, + "loss": 0.0287, + "step": 19841 + }, + { + "epoch": 3.61, + "learning_rate": 1.9192573716781943e-06, + "loss": 0.3213, + "step": 19842 + }, + { + "epoch": 3.61, + "learning_rate": 1.9192027666545323e-06, + "loss": 0.7878, + "step": 19843 + }, + { + "epoch": 3.61, + "learning_rate": 1.91914816163087e-06, + "loss": 0.0168, + "step": 19844 + }, + { + "epoch": 3.61, + "learning_rate": 1.9190935566072077e-06, + "loss": 0.1366, + "step": 19845 + }, + { + "epoch": 3.61, + "learning_rate": 1.9190389515835457e-06, + "loss": 1.8883, + "step": 19846 + }, + { + "epoch": 3.61, + "learning_rate": 1.9189843465598836e-06, + "loss": 0.4511, + "step": 19847 + }, + { + "epoch": 3.61, + "learning_rate": 1.9189297415362216e-06, + "loss": 0.2369, + "step": 19848 + }, + { + "epoch": 3.61, + "learning_rate": 1.9188751365125595e-06, + "loss": 0.0361, + "step": 19849 + }, + { + "epoch": 3.61, + "learning_rate": 1.918820531488897e-06, + "loss": 0.0371, + "step": 19850 + }, + { + "epoch": 3.61, + "learning_rate": 1.918765926465235e-06, + "loss": 0.0573, + "step": 19851 + }, + { + "epoch": 3.61, + "learning_rate": 1.918711321441573e-06, + "loss": 0.8202, + "step": 19852 + }, + { + "epoch": 3.61, + "learning_rate": 1.9186567164179104e-06, + "loss": 0.552, + "step": 19853 + }, + { + "epoch": 3.61, + "learning_rate": 1.9186021113942484e-06, + "loss": 0.8112, + "step": 19854 + }, + { + "epoch": 3.61, + "learning_rate": 1.9185475063705863e-06, + "loss": 0.0358, + "step": 19855 + }, + { + "epoch": 3.61, + "learning_rate": 1.918492901346924e-06, + "loss": 0.4243, + "step": 19856 + }, + { + "epoch": 3.61, + "learning_rate": 1.918438296323262e-06, + "loss": 0.2213, + "step": 19857 + }, + { + "epoch": 3.61, + "learning_rate": 1.9183836912995993e-06, + "loss": 0.8247, + "step": 19858 + }, + { + "epoch": 3.61, + "learning_rate": 1.9183290862759373e-06, + "loss": 0.3005, + "step": 19859 + }, + { + "epoch": 3.61, + "learning_rate": 1.918274481252275e-06, + "loss": 0.1544, + "step": 19860 + }, + { + "epoch": 3.61, + "learning_rate": 1.918219876228613e-06, + "loss": 2.3509, + "step": 19861 + }, + { + "epoch": 3.61, + "learning_rate": 1.918165271204951e-06, + "loss": 1.6075, + "step": 19862 + }, + { + "epoch": 3.61, + "learning_rate": 1.918110666181289e-06, + "loss": 0.1702, + "step": 19863 + }, + { + "epoch": 3.61, + "learning_rate": 1.9180560611576265e-06, + "loss": 0.5244, + "step": 19864 + }, + { + "epoch": 3.61, + "learning_rate": 1.9180014561339645e-06, + "loss": 0.7013, + "step": 19865 + }, + { + "epoch": 3.61, + "learning_rate": 1.9179468511103024e-06, + "loss": 0.2379, + "step": 19866 + }, + { + "epoch": 3.61, + "learning_rate": 1.91789224608664e-06, + "loss": 1.2294, + "step": 19867 + }, + { + "epoch": 3.61, + "learning_rate": 1.917837641062978e-06, + "loss": 1.4533, + "step": 19868 + }, + { + "epoch": 3.61, + "learning_rate": 1.917783036039316e-06, + "loss": 1.6644, + "step": 19869 + }, + { + "epoch": 3.61, + "learning_rate": 1.9177284310156534e-06, + "loss": 0.3623, + "step": 19870 + }, + { + "epoch": 3.61, + "learning_rate": 1.9176738259919913e-06, + "loss": 0.5995, + "step": 19871 + }, + { + "epoch": 3.61, + "learning_rate": 1.917619220968329e-06, + "loss": 0.3298, + "step": 19872 + }, + { + "epoch": 3.61, + "learning_rate": 1.9175646159446668e-06, + "loss": 0.8416, + "step": 19873 + }, + { + "epoch": 3.61, + "learning_rate": 1.9175100109210047e-06, + "loss": 0.0116, + "step": 19874 + }, + { + "epoch": 3.61, + "learning_rate": 1.9174554058973427e-06, + "loss": 1.3178, + "step": 19875 + }, + { + "epoch": 3.61, + "learning_rate": 1.9174008008736806e-06, + "loss": 0.6226, + "step": 19876 + }, + { + "epoch": 3.61, + "learning_rate": 1.9173461958500185e-06, + "loss": 0.4103, + "step": 19877 + }, + { + "epoch": 3.61, + "learning_rate": 1.917291590826356e-06, + "loss": 5.772, + "step": 19878 + }, + { + "epoch": 3.62, + "learning_rate": 1.917236985802694e-06, + "loss": 0.0229, + "step": 19879 + }, + { + "epoch": 3.62, + "learning_rate": 1.917182380779032e-06, + "loss": 0.2639, + "step": 19880 + }, + { + "epoch": 3.62, + "learning_rate": 1.9171277757553695e-06, + "loss": 0.8501, + "step": 19881 + }, + { + "epoch": 3.62, + "learning_rate": 1.9170731707317074e-06, + "loss": 1.4568, + "step": 19882 + }, + { + "epoch": 3.62, + "learning_rate": 1.9170185657080454e-06, + "loss": 0.7225, + "step": 19883 + }, + { + "epoch": 3.62, + "learning_rate": 1.916963960684383e-06, + "loss": 0.0321, + "step": 19884 + }, + { + "epoch": 3.62, + "learning_rate": 1.916909355660721e-06, + "loss": 1.2885, + "step": 19885 + }, + { + "epoch": 3.62, + "learning_rate": 1.9168547506370583e-06, + "loss": 0.7151, + "step": 19886 + }, + { + "epoch": 3.62, + "learning_rate": 1.9168001456133963e-06, + "loss": 1.3449, + "step": 19887 + }, + { + "epoch": 3.62, + "learning_rate": 1.9167455405897342e-06, + "loss": 0.0933, + "step": 19888 + }, + { + "epoch": 3.62, + "learning_rate": 1.916690935566072e-06, + "loss": 0.0209, + "step": 19889 + }, + { + "epoch": 3.62, + "learning_rate": 1.91663633054241e-06, + "loss": 0.1214, + "step": 19890 + }, + { + "epoch": 3.62, + "learning_rate": 1.916581725518748e-06, + "loss": 0.4893, + "step": 19891 + }, + { + "epoch": 3.62, + "learning_rate": 1.9165271204950856e-06, + "loss": 0.3924, + "step": 19892 + }, + { + "epoch": 3.62, + "learning_rate": 1.9164725154714235e-06, + "loss": 0.3443, + "step": 19893 + }, + { + "epoch": 3.62, + "learning_rate": 1.9164179104477615e-06, + "loss": 0.7606, + "step": 19894 + }, + { + "epoch": 3.62, + "learning_rate": 1.916363305424099e-06, + "loss": 0.0218, + "step": 19895 + }, + { + "epoch": 3.62, + "learning_rate": 1.916308700400437e-06, + "loss": 0.0124, + "step": 19896 + }, + { + "epoch": 3.62, + "learning_rate": 1.916254095376775e-06, + "loss": 0.8766, + "step": 19897 + }, + { + "epoch": 3.62, + "learning_rate": 1.9161994903531124e-06, + "loss": 0.6315, + "step": 19898 + }, + { + "epoch": 3.62, + "learning_rate": 1.9161448853294503e-06, + "loss": 0.0343, + "step": 19899 + }, + { + "epoch": 3.62, + "learning_rate": 1.916090280305788e-06, + "loss": 0.5642, + "step": 19900 + }, + { + "epoch": 3.62, + "learning_rate": 1.916035675282126e-06, + "loss": 1.4548, + "step": 19901 + }, + { + "epoch": 3.62, + "learning_rate": 1.9159810702584637e-06, + "loss": 0.7996, + "step": 19902 + }, + { + "epoch": 3.62, + "learning_rate": 1.9159264652348017e-06, + "loss": 0.2733, + "step": 19903 + }, + { + "epoch": 3.62, + "learning_rate": 1.9158718602111396e-06, + "loss": 1.4746, + "step": 19904 + }, + { + "epoch": 3.62, + "learning_rate": 1.9158172551874776e-06, + "loss": 1.7521, + "step": 19905 + }, + { + "epoch": 3.62, + "learning_rate": 1.915762650163815e-06, + "loss": 0.8728, + "step": 19906 + }, + { + "epoch": 3.62, + "learning_rate": 1.915708045140153e-06, + "loss": 0.7374, + "step": 19907 + }, + { + "epoch": 3.62, + "learning_rate": 1.915653440116491e-06, + "loss": 1.3112, + "step": 19908 + }, + { + "epoch": 3.62, + "learning_rate": 1.9155988350928285e-06, + "loss": 1.0814, + "step": 19909 + }, + { + "epoch": 3.62, + "learning_rate": 1.9155442300691664e-06, + "loss": 0.9331, + "step": 19910 + }, + { + "epoch": 3.62, + "learning_rate": 1.9154896250455044e-06, + "loss": 1.3769, + "step": 19911 + }, + { + "epoch": 3.62, + "learning_rate": 1.915435020021842e-06, + "loss": 0.5528, + "step": 19912 + }, + { + "epoch": 3.62, + "learning_rate": 1.91538041499818e-06, + "loss": 0.6884, + "step": 19913 + }, + { + "epoch": 3.62, + "learning_rate": 1.9153258099745174e-06, + "loss": 0.653, + "step": 19914 + }, + { + "epoch": 3.62, + "learning_rate": 1.9152712049508553e-06, + "loss": 0.0773, + "step": 19915 + }, + { + "epoch": 3.62, + "learning_rate": 1.9152165999271933e-06, + "loss": 0.5028, + "step": 19916 + }, + { + "epoch": 3.62, + "learning_rate": 1.915161994903531e-06, + "loss": 0.6792, + "step": 19917 + }, + { + "epoch": 3.62, + "learning_rate": 1.915107389879869e-06, + "loss": 0.2574, + "step": 19918 + }, + { + "epoch": 3.62, + "learning_rate": 1.915052784856207e-06, + "loss": 0.8843, + "step": 19919 + }, + { + "epoch": 3.62, + "learning_rate": 1.9149981798325446e-06, + "loss": 0.0449, + "step": 19920 + }, + { + "epoch": 3.62, + "learning_rate": 1.9149435748088825e-06, + "loss": 1.5712, + "step": 19921 + }, + { + "epoch": 3.62, + "learning_rate": 1.9148889697852205e-06, + "loss": 0.1061, + "step": 19922 + }, + { + "epoch": 3.62, + "learning_rate": 1.914834364761558e-06, + "loss": 0.7811, + "step": 19923 + }, + { + "epoch": 3.62, + "learning_rate": 1.914779759737896e-06, + "loss": 1.1065, + "step": 19924 + }, + { + "epoch": 3.62, + "learning_rate": 1.9147251547142335e-06, + "loss": 3.3137, + "step": 19925 + }, + { + "epoch": 3.62, + "learning_rate": 1.9146705496905714e-06, + "loss": 0.5691, + "step": 19926 + }, + { + "epoch": 3.62, + "learning_rate": 1.9146159446669094e-06, + "loss": 0.5137, + "step": 19927 + }, + { + "epoch": 3.62, + "learning_rate": 1.914561339643247e-06, + "loss": 0.2601, + "step": 19928 + }, + { + "epoch": 3.62, + "learning_rate": 1.914506734619585e-06, + "loss": 0.3911, + "step": 19929 + }, + { + "epoch": 3.62, + "learning_rate": 1.9144521295959228e-06, + "loss": 0.0294, + "step": 19930 + }, + { + "epoch": 3.62, + "learning_rate": 1.9143975245722607e-06, + "loss": 0.8993, + "step": 19931 + }, + { + "epoch": 3.62, + "learning_rate": 1.9143429195485987e-06, + "loss": 1.9499, + "step": 19932 + }, + { + "epoch": 3.62, + "learning_rate": 1.9142883145249366e-06, + "loss": 0.0963, + "step": 19933 + }, + { + "epoch": 3.63, + "learning_rate": 1.914233709501274e-06, + "loss": 0.4372, + "step": 19934 + }, + { + "epoch": 3.63, + "learning_rate": 1.914179104477612e-06, + "loss": 1.2702, + "step": 19935 + }, + { + "epoch": 3.63, + "learning_rate": 1.91412449945395e-06, + "loss": 0.8267, + "step": 19936 + }, + { + "epoch": 3.63, + "learning_rate": 1.9140698944302875e-06, + "loss": 0.2312, + "step": 19937 + }, + { + "epoch": 3.63, + "learning_rate": 1.9140152894066255e-06, + "loss": 0.4934, + "step": 19938 + }, + { + "epoch": 3.63, + "learning_rate": 1.913960684382963e-06, + "loss": 2.056, + "step": 19939 + }, + { + "epoch": 3.63, + "learning_rate": 1.913906079359301e-06, + "loss": 0.9576, + "step": 19940 + }, + { + "epoch": 3.63, + "learning_rate": 1.913851474335639e-06, + "loss": 1.9684, + "step": 19941 + }, + { + "epoch": 3.63, + "learning_rate": 1.913796869311977e-06, + "loss": 1.563, + "step": 19942 + }, + { + "epoch": 3.63, + "learning_rate": 1.9137422642883148e-06, + "loss": 0.1111, + "step": 19943 + }, + { + "epoch": 3.63, + "learning_rate": 1.9136876592646527e-06, + "loss": 1.9443, + "step": 19944 + }, + { + "epoch": 3.63, + "learning_rate": 1.9136330542409902e-06, + "loss": 0.5243, + "step": 19945 + }, + { + "epoch": 3.63, + "learning_rate": 1.913578449217328e-06, + "loss": 0.0734, + "step": 19946 + }, + { + "epoch": 3.63, + "learning_rate": 1.913523844193666e-06, + "loss": 0.4109, + "step": 19947 + }, + { + "epoch": 3.63, + "learning_rate": 1.9134692391700036e-06, + "loss": 0.0586, + "step": 19948 + }, + { + "epoch": 3.63, + "learning_rate": 1.9134146341463416e-06, + "loss": 0.3699, + "step": 19949 + }, + { + "epoch": 3.63, + "learning_rate": 1.9133600291226795e-06, + "loss": 0.9749, + "step": 19950 + }, + { + "epoch": 3.63, + "learning_rate": 1.913305424099017e-06, + "loss": 0.8928, + "step": 19951 + }, + { + "epoch": 3.63, + "learning_rate": 1.913250819075355e-06, + "loss": 1.7018, + "step": 19952 + }, + { + "epoch": 3.63, + "learning_rate": 1.9131962140516925e-06, + "loss": 1.5545, + "step": 19953 + }, + { + "epoch": 3.63, + "learning_rate": 1.9131416090280304e-06, + "loss": 0.4077, + "step": 19954 + }, + { + "epoch": 3.63, + "learning_rate": 1.9130870040043684e-06, + "loss": 0.5719, + "step": 19955 + }, + { + "epoch": 3.63, + "learning_rate": 1.9130323989807063e-06, + "loss": 1.1072, + "step": 19956 + }, + { + "epoch": 3.63, + "learning_rate": 1.9129777939570443e-06, + "loss": 0.3517, + "step": 19957 + }, + { + "epoch": 3.63, + "learning_rate": 1.9129231889333822e-06, + "loss": 0.7801, + "step": 19958 + }, + { + "epoch": 3.63, + "learning_rate": 1.9128685839097197e-06, + "loss": 0.1726, + "step": 19959 + }, + { + "epoch": 3.63, + "learning_rate": 1.9128139788860577e-06, + "loss": 0.854, + "step": 19960 + }, + { + "epoch": 3.63, + "learning_rate": 1.9127593738623956e-06, + "loss": 1.6508, + "step": 19961 + }, + { + "epoch": 3.63, + "learning_rate": 1.912704768838733e-06, + "loss": 0.201, + "step": 19962 + }, + { + "epoch": 3.63, + "learning_rate": 1.912650163815071e-06, + "loss": 0.2246, + "step": 19963 + }, + { + "epoch": 3.63, + "learning_rate": 1.912595558791409e-06, + "loss": 0.8202, + "step": 19964 + }, + { + "epoch": 3.63, + "learning_rate": 1.9125409537677465e-06, + "loss": 0.2757, + "step": 19965 + }, + { + "epoch": 3.63, + "learning_rate": 1.9124863487440845e-06, + "loss": 2.5652, + "step": 19966 + }, + { + "epoch": 3.63, + "learning_rate": 1.912431743720422e-06, + "loss": 1.7566, + "step": 19967 + }, + { + "epoch": 3.63, + "learning_rate": 1.91237713869676e-06, + "loss": 0.0249, + "step": 19968 + }, + { + "epoch": 3.63, + "learning_rate": 1.912322533673098e-06, + "loss": 0.3266, + "step": 19969 + }, + { + "epoch": 3.63, + "learning_rate": 1.912267928649436e-06, + "loss": 0.2824, + "step": 19970 + }, + { + "epoch": 3.63, + "learning_rate": 1.9122133236257738e-06, + "loss": 0.3621, + "step": 19971 + }, + { + "epoch": 3.63, + "learning_rate": 1.9121587186021117e-06, + "loss": 2.7213, + "step": 19972 + }, + { + "epoch": 3.63, + "learning_rate": 1.9121041135784492e-06, + "loss": 0.0396, + "step": 19973 + }, + { + "epoch": 3.63, + "learning_rate": 1.912049508554787e-06, + "loss": 1.6082, + "step": 19974 + }, + { + "epoch": 3.63, + "learning_rate": 1.911994903531125e-06, + "loss": 1.8149, + "step": 19975 + }, + { + "epoch": 3.63, + "learning_rate": 1.9119402985074627e-06, + "loss": 1.7347, + "step": 19976 + }, + { + "epoch": 3.63, + "learning_rate": 1.9118856934838006e-06, + "loss": 0.4209, + "step": 19977 + }, + { + "epoch": 3.63, + "learning_rate": 1.9118310884601385e-06, + "loss": 0.3685, + "step": 19978 + }, + { + "epoch": 3.63, + "learning_rate": 1.911776483436476e-06, + "loss": 1.0584, + "step": 19979 + }, + { + "epoch": 3.63, + "learning_rate": 1.911721878412814e-06, + "loss": 0.1135, + "step": 19980 + }, + { + "epoch": 3.63, + "learning_rate": 1.9116672733891515e-06, + "loss": 0.1251, + "step": 19981 + }, + { + "epoch": 3.63, + "learning_rate": 1.9116126683654895e-06, + "loss": 0.7651, + "step": 19982 + }, + { + "epoch": 3.63, + "learning_rate": 1.9115580633418274e-06, + "loss": 0.6282, + "step": 19983 + }, + { + "epoch": 3.63, + "learning_rate": 1.9115034583181654e-06, + "loss": 0.2405, + "step": 19984 + }, + { + "epoch": 3.63, + "learning_rate": 1.9114488532945033e-06, + "loss": 0.0322, + "step": 19985 + }, + { + "epoch": 3.63, + "learning_rate": 1.9113942482708412e-06, + "loss": 0.6527, + "step": 19986 + }, + { + "epoch": 3.63, + "learning_rate": 1.9113396432471788e-06, + "loss": 3.7896, + "step": 19987 + }, + { + "epoch": 3.63, + "learning_rate": 1.9112850382235167e-06, + "loss": 0.7659, + "step": 19988 + }, + { + "epoch": 3.64, + "learning_rate": 1.9112304331998546e-06, + "loss": 1.5667, + "step": 19989 + }, + { + "epoch": 3.64, + "learning_rate": 1.911175828176192e-06, + "loss": 0.5391, + "step": 19990 + }, + { + "epoch": 3.64, + "learning_rate": 1.91112122315253e-06, + "loss": 0.0689, + "step": 19991 + }, + { + "epoch": 3.64, + "learning_rate": 1.911066618128868e-06, + "loss": 0.8211, + "step": 19992 + }, + { + "epoch": 3.64, + "learning_rate": 1.9110120131052056e-06, + "loss": 0.7724, + "step": 19993 + }, + { + "epoch": 3.64, + "learning_rate": 1.9109574080815435e-06, + "loss": 1.6918, + "step": 19994 + }, + { + "epoch": 3.64, + "learning_rate": 1.910902803057881e-06, + "loss": 1.0636, + "step": 19995 + }, + { + "epoch": 3.64, + "learning_rate": 1.910848198034219e-06, + "loss": 0.2968, + "step": 19996 + }, + { + "epoch": 3.64, + "learning_rate": 1.910793593010557e-06, + "loss": 0.691, + "step": 19997 + }, + { + "epoch": 3.64, + "learning_rate": 1.910738987986895e-06, + "loss": 4.1051, + "step": 19998 + }, + { + "epoch": 3.64, + "learning_rate": 1.910684382963233e-06, + "loss": 0.4433, + "step": 19999 + }, + { + "epoch": 3.64, + "learning_rate": 1.9106297779395708e-06, + "loss": 0.6309, + "step": 20000 + }, + { + "epoch": 3.64, + "learning_rate": 1.9105751729159083e-06, + "loss": 0.4233, + "step": 20001 + }, + { + "epoch": 3.64, + "learning_rate": 1.9105205678922462e-06, + "loss": 0.5333, + "step": 20002 + }, + { + "epoch": 3.64, + "learning_rate": 1.910465962868584e-06, + "loss": 0.9573, + "step": 20003 + }, + { + "epoch": 3.64, + "learning_rate": 1.9104113578449217e-06, + "loss": 2.543, + "step": 20004 + }, + { + "epoch": 3.64, + "learning_rate": 1.9103567528212596e-06, + "loss": 1.1895, + "step": 20005 + }, + { + "epoch": 3.64, + "learning_rate": 1.910302147797597e-06, + "loss": 0.0933, + "step": 20006 + }, + { + "epoch": 3.64, + "learning_rate": 1.910247542773935e-06, + "loss": 0.3067, + "step": 20007 + }, + { + "epoch": 3.64, + "learning_rate": 1.910192937750273e-06, + "loss": 0.1173, + "step": 20008 + }, + { + "epoch": 3.64, + "learning_rate": 1.9101383327266106e-06, + "loss": 0.3434, + "step": 20009 + }, + { + "epoch": 3.64, + "learning_rate": 1.9100837277029485e-06, + "loss": 0.5257, + "step": 20010 + }, + { + "epoch": 3.64, + "learning_rate": 1.9100291226792864e-06, + "loss": 0.7013, + "step": 20011 + }, + { + "epoch": 3.64, + "learning_rate": 1.9099745176556244e-06, + "loss": 1.2275, + "step": 20012 + }, + { + "epoch": 3.64, + "learning_rate": 1.9099199126319623e-06, + "loss": 0.9078, + "step": 20013 + }, + { + "epoch": 3.64, + "learning_rate": 1.9098653076083003e-06, + "loss": 0.7747, + "step": 20014 + }, + { + "epoch": 3.64, + "learning_rate": 1.909810702584638e-06, + "loss": 0.3023, + "step": 20015 + }, + { + "epoch": 3.64, + "learning_rate": 1.9097560975609757e-06, + "loss": 0.4697, + "step": 20016 + }, + { + "epoch": 3.64, + "learning_rate": 1.9097014925373137e-06, + "loss": 0.2321, + "step": 20017 + }, + { + "epoch": 3.64, + "learning_rate": 1.909646887513651e-06, + "loss": 0.5643, + "step": 20018 + }, + { + "epoch": 3.64, + "learning_rate": 1.909592282489989e-06, + "loss": 1.378, + "step": 20019 + }, + { + "epoch": 3.64, + "learning_rate": 1.9095376774663267e-06, + "loss": 0.0132, + "step": 20020 + }, + { + "epoch": 3.64, + "learning_rate": 1.9094830724426646e-06, + "loss": 0.0476, + "step": 20021 + }, + { + "epoch": 3.64, + "learning_rate": 1.9094284674190025e-06, + "loss": 1.2737, + "step": 20022 + }, + { + "epoch": 3.64, + "learning_rate": 1.9093738623953405e-06, + "loss": 2.4109, + "step": 20023 + }, + { + "epoch": 3.64, + "learning_rate": 1.9093192573716784e-06, + "loss": 0.5686, + "step": 20024 + }, + { + "epoch": 3.64, + "learning_rate": 1.9092646523480164e-06, + "loss": 1.9678, + "step": 20025 + }, + { + "epoch": 3.64, + "learning_rate": 1.909210047324354e-06, + "loss": 0.0334, + "step": 20026 + }, + { + "epoch": 3.64, + "learning_rate": 1.909155442300692e-06, + "loss": 0.532, + "step": 20027 + }, + { + "epoch": 3.64, + "learning_rate": 1.9091008372770298e-06, + "loss": 0.0086, + "step": 20028 + }, + { + "epoch": 3.64, + "learning_rate": 1.9090462322533673e-06, + "loss": 4.3916, + "step": 20029 + }, + { + "epoch": 3.64, + "learning_rate": 1.9089916272297052e-06, + "loss": 0.0416, + "step": 20030 + }, + { + "epoch": 3.64, + "learning_rate": 1.908937022206043e-06, + "loss": 3.2841, + "step": 20031 + }, + { + "epoch": 3.64, + "learning_rate": 1.9088824171823807e-06, + "loss": 0.7164, + "step": 20032 + }, + { + "epoch": 3.64, + "learning_rate": 1.9088278121587187e-06, + "loss": 1.0478, + "step": 20033 + }, + { + "epoch": 3.64, + "learning_rate": 1.908773207135056e-06, + "loss": 0.3203, + "step": 20034 + }, + { + "epoch": 3.64, + "learning_rate": 1.908718602111394e-06, + "loss": 0.012, + "step": 20035 + }, + { + "epoch": 3.64, + "learning_rate": 1.908663997087732e-06, + "loss": 1.0498, + "step": 20036 + }, + { + "epoch": 3.64, + "learning_rate": 1.90860939206407e-06, + "loss": 0.7666, + "step": 20037 + }, + { + "epoch": 3.64, + "learning_rate": 1.908554787040408e-06, + "loss": 0.5323, + "step": 20038 + }, + { + "epoch": 3.64, + "learning_rate": 1.908500182016746e-06, + "loss": 1.1854, + "step": 20039 + }, + { + "epoch": 3.64, + "learning_rate": 1.9084455769930834e-06, + "loss": 0.5291, + "step": 20040 + }, + { + "epoch": 3.64, + "learning_rate": 1.9083909719694214e-06, + "loss": 2.6431, + "step": 20041 + }, + { + "epoch": 3.64, + "learning_rate": 1.9083363669457593e-06, + "loss": 0.0455, + "step": 20042 + }, + { + "epoch": 3.64, + "learning_rate": 1.908281761922097e-06, + "loss": 1.0422, + "step": 20043 + }, + { + "epoch": 3.65, + "learning_rate": 1.9082271568984348e-06, + "loss": 1.0898, + "step": 20044 + }, + { + "epoch": 3.65, + "learning_rate": 1.9081725518747727e-06, + "loss": 0.212, + "step": 20045 + }, + { + "epoch": 3.65, + "learning_rate": 1.9081179468511102e-06, + "loss": 0.2392, + "step": 20046 + }, + { + "epoch": 3.65, + "learning_rate": 1.908063341827448e-06, + "loss": 0.8489, + "step": 20047 + }, + { + "epoch": 3.65, + "learning_rate": 1.9080087368037857e-06, + "loss": 0.4596, + "step": 20048 + }, + { + "epoch": 3.65, + "learning_rate": 1.9079541317801236e-06, + "loss": 0.3008, + "step": 20049 + }, + { + "epoch": 3.65, + "learning_rate": 1.9078995267564616e-06, + "loss": 0.7905, + "step": 20050 + }, + { + "epoch": 3.65, + "learning_rate": 1.9078449217327995e-06, + "loss": 2.3215, + "step": 20051 + }, + { + "epoch": 3.65, + "learning_rate": 1.9077903167091375e-06, + "loss": 0.0416, + "step": 20052 + }, + { + "epoch": 3.65, + "learning_rate": 1.9077357116854754e-06, + "loss": 0.2883, + "step": 20053 + }, + { + "epoch": 3.65, + "learning_rate": 1.907681106661813e-06, + "loss": 0.9634, + "step": 20054 + }, + { + "epoch": 3.65, + "learning_rate": 1.907626501638151e-06, + "loss": 0.3435, + "step": 20055 + }, + { + "epoch": 3.65, + "learning_rate": 1.907571896614489e-06, + "loss": 0.555, + "step": 20056 + }, + { + "epoch": 3.65, + "learning_rate": 1.9075172915908263e-06, + "loss": 1.7284, + "step": 20057 + }, + { + "epoch": 3.65, + "learning_rate": 1.9074626865671643e-06, + "loss": 0.2941, + "step": 20058 + }, + { + "epoch": 3.65, + "learning_rate": 1.9074080815435022e-06, + "loss": 0.2842, + "step": 20059 + }, + { + "epoch": 3.65, + "learning_rate": 1.9073534765198397e-06, + "loss": 0.4086, + "step": 20060 + }, + { + "epoch": 3.65, + "learning_rate": 1.9072988714961777e-06, + "loss": 0.0191, + "step": 20061 + }, + { + "epoch": 3.65, + "learning_rate": 1.9072442664725154e-06, + "loss": 1.1977, + "step": 20062 + }, + { + "epoch": 3.65, + "learning_rate": 1.9071896614488534e-06, + "loss": 0.0125, + "step": 20063 + }, + { + "epoch": 3.65, + "learning_rate": 1.9071350564251913e-06, + "loss": 1.667, + "step": 20064 + }, + { + "epoch": 3.65, + "learning_rate": 1.9070804514015288e-06, + "loss": 0.0146, + "step": 20065 + }, + { + "epoch": 3.65, + "learning_rate": 1.9070258463778668e-06, + "loss": 0.1093, + "step": 20066 + }, + { + "epoch": 3.65, + "learning_rate": 1.9069712413542047e-06, + "loss": 3.5936, + "step": 20067 + }, + { + "epoch": 3.65, + "learning_rate": 1.9069166363305424e-06, + "loss": 0.9138, + "step": 20068 + }, + { + "epoch": 3.65, + "learning_rate": 1.9068620313068804e-06, + "loss": 0.6301, + "step": 20069 + }, + { + "epoch": 3.65, + "learning_rate": 1.9068074262832183e-06, + "loss": 1.3234, + "step": 20070 + }, + { + "epoch": 3.65, + "learning_rate": 1.9067528212595558e-06, + "loss": 0.0532, + "step": 20071 + }, + { + "epoch": 3.65, + "learning_rate": 1.9066982162358938e-06, + "loss": 0.3541, + "step": 20072 + }, + { + "epoch": 3.65, + "learning_rate": 1.9066436112122313e-06, + "loss": 2.3867, + "step": 20073 + }, + { + "epoch": 3.65, + "learning_rate": 1.9065890061885692e-06, + "loss": 0.0236, + "step": 20074 + }, + { + "epoch": 3.65, + "learning_rate": 1.9065344011649072e-06, + "loss": 1.2314, + "step": 20075 + }, + { + "epoch": 3.65, + "learning_rate": 1.906479796141245e-06, + "loss": 1.5665, + "step": 20076 + }, + { + "epoch": 3.65, + "learning_rate": 1.9064251911175829e-06, + "loss": 0.2061, + "step": 20077 + }, + { + "epoch": 3.65, + "learning_rate": 1.9063705860939208e-06, + "loss": 0.0311, + "step": 20078 + }, + { + "epoch": 3.65, + "learning_rate": 1.9063159810702583e-06, + "loss": 0.1948, + "step": 20079 + }, + { + "epoch": 3.65, + "learning_rate": 1.9062613760465963e-06, + "loss": 0.1342, + "step": 20080 + }, + { + "epoch": 3.65, + "learning_rate": 1.9062067710229342e-06, + "loss": 0.7888, + "step": 20081 + }, + { + "epoch": 3.65, + "learning_rate": 1.906152165999272e-06, + "loss": 0.537, + "step": 20082 + }, + { + "epoch": 3.65, + "learning_rate": 1.9060975609756099e-06, + "loss": 1.3147, + "step": 20083 + }, + { + "epoch": 3.65, + "learning_rate": 1.9060429559519478e-06, + "loss": 0.2236, + "step": 20084 + }, + { + "epoch": 3.65, + "learning_rate": 1.9059883509282854e-06, + "loss": 1.023, + "step": 20085 + }, + { + "epoch": 3.65, + "learning_rate": 1.9059337459046233e-06, + "loss": 0.202, + "step": 20086 + }, + { + "epoch": 3.65, + "learning_rate": 1.905879140880961e-06, + "loss": 0.3062, + "step": 20087 + }, + { + "epoch": 3.65, + "learning_rate": 1.905824535857299e-06, + "loss": 0.5334, + "step": 20088 + }, + { + "epoch": 3.65, + "learning_rate": 1.905769930833637e-06, + "loss": 1.0078, + "step": 20089 + }, + { + "epoch": 3.65, + "learning_rate": 1.9057153258099744e-06, + "loss": 1.1535, + "step": 20090 + }, + { + "epoch": 3.65, + "learning_rate": 1.9056607207863124e-06, + "loss": 0.539, + "step": 20091 + }, + { + "epoch": 3.65, + "learning_rate": 1.9056061157626503e-06, + "loss": 2.5576, + "step": 20092 + }, + { + "epoch": 3.65, + "learning_rate": 1.9055515107389878e-06, + "loss": 1.207, + "step": 20093 + }, + { + "epoch": 3.65, + "learning_rate": 1.9054969057153258e-06, + "loss": 1.5504, + "step": 20094 + }, + { + "epoch": 3.65, + "learning_rate": 1.9054423006916637e-06, + "loss": 1.0614, + "step": 20095 + }, + { + "epoch": 3.65, + "learning_rate": 1.9053876956680015e-06, + "loss": 0.8941, + "step": 20096 + }, + { + "epoch": 3.65, + "learning_rate": 1.9053330906443394e-06, + "loss": 0.2145, + "step": 20097 + }, + { + "epoch": 3.65, + "learning_rate": 1.9052784856206774e-06, + "loss": 0.7094, + "step": 20098 + }, + { + "epoch": 3.66, + "learning_rate": 1.9052238805970149e-06, + "loss": 0.2596, + "step": 20099 + }, + { + "epoch": 3.66, + "learning_rate": 1.9051692755733528e-06, + "loss": 0.1118, + "step": 20100 + }, + { + "epoch": 3.66, + "learning_rate": 1.9051146705496905e-06, + "loss": 2.2857, + "step": 20101 + }, + { + "epoch": 3.66, + "learning_rate": 1.9050600655260285e-06, + "loss": 1.8287, + "step": 20102 + }, + { + "epoch": 3.66, + "learning_rate": 1.9050054605023664e-06, + "loss": 2.0889, + "step": 20103 + }, + { + "epoch": 3.66, + "learning_rate": 1.904950855478704e-06, + "loss": 0.0721, + "step": 20104 + }, + { + "epoch": 3.66, + "learning_rate": 1.904896250455042e-06, + "loss": 0.4237, + "step": 20105 + }, + { + "epoch": 3.66, + "learning_rate": 1.9048416454313798e-06, + "loss": 0.206, + "step": 20106 + }, + { + "epoch": 3.66, + "learning_rate": 1.9047870404077174e-06, + "loss": 1.2858, + "step": 20107 + }, + { + "epoch": 3.66, + "learning_rate": 1.9047324353840553e-06, + "loss": 0.5766, + "step": 20108 + }, + { + "epoch": 3.66, + "learning_rate": 1.9046778303603932e-06, + "loss": 0.9885, + "step": 20109 + }, + { + "epoch": 3.66, + "learning_rate": 1.904623225336731e-06, + "loss": 0.4763, + "step": 20110 + }, + { + "epoch": 3.66, + "learning_rate": 1.904568620313069e-06, + "loss": 2.3359, + "step": 20111 + }, + { + "epoch": 3.66, + "learning_rate": 1.9045140152894069e-06, + "loss": 0.2355, + "step": 20112 + }, + { + "epoch": 3.66, + "learning_rate": 1.9044594102657444e-06, + "loss": 2.8724, + "step": 20113 + }, + { + "epoch": 3.66, + "learning_rate": 1.9044048052420823e-06, + "loss": 0.2383, + "step": 20114 + }, + { + "epoch": 3.66, + "learning_rate": 1.90435020021842e-06, + "loss": 1.3322, + "step": 20115 + }, + { + "epoch": 3.66, + "learning_rate": 1.904295595194758e-06, + "loss": 0.014, + "step": 20116 + }, + { + "epoch": 3.66, + "learning_rate": 1.904240990171096e-06, + "loss": 0.3914, + "step": 20117 + }, + { + "epoch": 3.66, + "learning_rate": 1.9041863851474335e-06, + "loss": 0.6794, + "step": 20118 + }, + { + "epoch": 3.66, + "learning_rate": 1.9041317801237714e-06, + "loss": 0.611, + "step": 20119 + }, + { + "epoch": 3.66, + "learning_rate": 1.9040771751001094e-06, + "loss": 1.0947, + "step": 20120 + }, + { + "epoch": 3.66, + "learning_rate": 1.904022570076447e-06, + "loss": 0.2043, + "step": 20121 + }, + { + "epoch": 3.66, + "learning_rate": 1.9039679650527848e-06, + "loss": 0.2319, + "step": 20122 + }, + { + "epoch": 3.66, + "learning_rate": 1.9039133600291228e-06, + "loss": 1.1926, + "step": 20123 + }, + { + "epoch": 3.66, + "learning_rate": 1.9038587550054605e-06, + "loss": 1.2604, + "step": 20124 + }, + { + "epoch": 3.66, + "learning_rate": 1.9038041499817984e-06, + "loss": 0.5323, + "step": 20125 + }, + { + "epoch": 3.66, + "learning_rate": 1.9037495449581364e-06, + "loss": 0.0267, + "step": 20126 + }, + { + "epoch": 3.66, + "learning_rate": 1.903694939934474e-06, + "loss": 0.1957, + "step": 20127 + }, + { + "epoch": 3.66, + "learning_rate": 1.9036403349108118e-06, + "loss": 0.3971, + "step": 20128 + }, + { + "epoch": 3.66, + "learning_rate": 1.9035857298871496e-06, + "loss": 0.4312, + "step": 20129 + }, + { + "epoch": 3.66, + "learning_rate": 1.9035311248634875e-06, + "loss": 0.6251, + "step": 20130 + }, + { + "epoch": 3.66, + "learning_rate": 1.9034765198398255e-06, + "loss": 0.0058, + "step": 20131 + }, + { + "epoch": 3.66, + "learning_rate": 1.903421914816163e-06, + "loss": 0.0222, + "step": 20132 + }, + { + "epoch": 3.66, + "learning_rate": 1.903367309792501e-06, + "loss": 0.0245, + "step": 20133 + }, + { + "epoch": 3.66, + "learning_rate": 1.9033127047688389e-06, + "loss": 0.2147, + "step": 20134 + }, + { + "epoch": 3.66, + "learning_rate": 1.9032580997451766e-06, + "loss": 1.6394, + "step": 20135 + }, + { + "epoch": 3.66, + "learning_rate": 1.9032034947215145e-06, + "loss": 2.6286, + "step": 20136 + }, + { + "epoch": 3.66, + "learning_rate": 1.9031488896978525e-06, + "loss": 0.3932, + "step": 20137 + }, + { + "epoch": 3.66, + "learning_rate": 1.90309428467419e-06, + "loss": 0.1343, + "step": 20138 + }, + { + "epoch": 3.66, + "learning_rate": 1.903039679650528e-06, + "loss": 3.2302, + "step": 20139 + }, + { + "epoch": 3.66, + "learning_rate": 1.9029850746268659e-06, + "loss": 0.6537, + "step": 20140 + }, + { + "epoch": 3.66, + "learning_rate": 1.9029304696032034e-06, + "loss": 0.314, + "step": 20141 + }, + { + "epoch": 3.66, + "learning_rate": 1.9028758645795414e-06, + "loss": 2.1593, + "step": 20142 + }, + { + "epoch": 3.66, + "learning_rate": 1.902821259555879e-06, + "loss": 0.2505, + "step": 20143 + }, + { + "epoch": 3.66, + "learning_rate": 1.902766654532217e-06, + "loss": 2.4472, + "step": 20144 + }, + { + "epoch": 3.66, + "learning_rate": 1.902712049508555e-06, + "loss": 0.0499, + "step": 20145 + }, + { + "epoch": 3.66, + "learning_rate": 1.9026574444848925e-06, + "loss": 0.7861, + "step": 20146 + }, + { + "epoch": 3.66, + "learning_rate": 1.9026028394612304e-06, + "loss": 0.8467, + "step": 20147 + }, + { + "epoch": 3.66, + "learning_rate": 1.9025482344375684e-06, + "loss": 0.3734, + "step": 20148 + }, + { + "epoch": 3.66, + "learning_rate": 1.9024936294139061e-06, + "loss": 1.2382, + "step": 20149 + }, + { + "epoch": 3.66, + "learning_rate": 1.902439024390244e-06, + "loss": 2.4868, + "step": 20150 + }, + { + "epoch": 3.66, + "learning_rate": 1.902384419366582e-06, + "loss": 0.3477, + "step": 20151 + }, + { + "epoch": 3.66, + "learning_rate": 1.9023298143429195e-06, + "loss": 0.1977, + "step": 20152 + }, + { + "epoch": 3.66, + "learning_rate": 1.9022752093192575e-06, + "loss": 1.8444, + "step": 20153 + }, + { + "epoch": 3.67, + "learning_rate": 1.902220604295595e-06, + "loss": 0.8488, + "step": 20154 + }, + { + "epoch": 3.67, + "learning_rate": 1.902165999271933e-06, + "loss": 1.5767, + "step": 20155 + }, + { + "epoch": 3.67, + "learning_rate": 1.9021113942482709e-06, + "loss": 1.017, + "step": 20156 + }, + { + "epoch": 3.67, + "learning_rate": 1.9020567892246086e-06, + "loss": 0.018, + "step": 20157 + }, + { + "epoch": 3.67, + "learning_rate": 1.9020021842009465e-06, + "loss": 0.104, + "step": 20158 + }, + { + "epoch": 3.67, + "learning_rate": 1.9019475791772845e-06, + "loss": 0.6486, + "step": 20159 + }, + { + "epoch": 3.67, + "learning_rate": 1.901892974153622e-06, + "loss": 0.7454, + "step": 20160 + }, + { + "epoch": 3.67, + "learning_rate": 1.90183836912996e-06, + "loss": 0.9491, + "step": 20161 + }, + { + "epoch": 3.67, + "learning_rate": 1.9017837641062979e-06, + "loss": 0.3457, + "step": 20162 + }, + { + "epoch": 3.67, + "learning_rate": 1.9017291590826356e-06, + "loss": 1.3942, + "step": 20163 + }, + { + "epoch": 3.67, + "learning_rate": 1.9016745540589736e-06, + "loss": 0.0196, + "step": 20164 + }, + { + "epoch": 3.67, + "learning_rate": 1.9016199490353115e-06, + "loss": 3.5428, + "step": 20165 + }, + { + "epoch": 3.67, + "learning_rate": 1.901565344011649e-06, + "loss": 0.1847, + "step": 20166 + }, + { + "epoch": 3.67, + "learning_rate": 1.901510738987987e-06, + "loss": 1.2136, + "step": 20167 + }, + { + "epoch": 3.67, + "learning_rate": 1.9014561339643247e-06, + "loss": 0.2793, + "step": 20168 + }, + { + "epoch": 3.67, + "learning_rate": 1.9014015289406626e-06, + "loss": 0.0239, + "step": 20169 + }, + { + "epoch": 3.67, + "learning_rate": 1.9013469239170006e-06, + "loss": 3.3646, + "step": 20170 + }, + { + "epoch": 3.67, + "learning_rate": 1.9012923188933381e-06, + "loss": 1.1073, + "step": 20171 + }, + { + "epoch": 3.67, + "learning_rate": 1.901237713869676e-06, + "loss": 0.0281, + "step": 20172 + }, + { + "epoch": 3.67, + "learning_rate": 1.901183108846014e-06, + "loss": 0.8431, + "step": 20173 + }, + { + "epoch": 3.67, + "learning_rate": 1.9011285038223515e-06, + "loss": 0.2575, + "step": 20174 + }, + { + "epoch": 3.67, + "learning_rate": 1.9010738987986895e-06, + "loss": 0.3052, + "step": 20175 + }, + { + "epoch": 3.67, + "learning_rate": 1.9010192937750274e-06, + "loss": 0.678, + "step": 20176 + }, + { + "epoch": 3.67, + "learning_rate": 1.9009646887513651e-06, + "loss": 0.2305, + "step": 20177 + }, + { + "epoch": 3.67, + "learning_rate": 1.900910083727703e-06, + "loss": 0.3215, + "step": 20178 + }, + { + "epoch": 3.67, + "learning_rate": 1.900855478704041e-06, + "loss": 0.2831, + "step": 20179 + }, + { + "epoch": 3.67, + "learning_rate": 1.9008008736803785e-06, + "loss": 2.6219, + "step": 20180 + }, + { + "epoch": 3.67, + "learning_rate": 1.9007462686567165e-06, + "loss": 0.4975, + "step": 20181 + }, + { + "epoch": 3.67, + "learning_rate": 1.9006916636330542e-06, + "loss": 0.8312, + "step": 20182 + }, + { + "epoch": 3.67, + "learning_rate": 1.9006370586093922e-06, + "loss": 1.1456, + "step": 20183 + }, + { + "epoch": 3.67, + "learning_rate": 1.9005824535857301e-06, + "loss": 0.0354, + "step": 20184 + }, + { + "epoch": 3.67, + "learning_rate": 1.9005278485620676e-06, + "loss": 0.2833, + "step": 20185 + }, + { + "epoch": 3.67, + "learning_rate": 1.9004732435384056e-06, + "loss": 0.8424, + "step": 20186 + }, + { + "epoch": 3.67, + "learning_rate": 1.9004186385147435e-06, + "loss": 0.0287, + "step": 20187 + }, + { + "epoch": 3.67, + "learning_rate": 1.900364033491081e-06, + "loss": 0.9664, + "step": 20188 + }, + { + "epoch": 3.67, + "learning_rate": 1.900309428467419e-06, + "loss": 1.4735, + "step": 20189 + }, + { + "epoch": 3.67, + "learning_rate": 1.900254823443757e-06, + "loss": 0.3355, + "step": 20190 + }, + { + "epoch": 3.67, + "learning_rate": 1.9002002184200947e-06, + "loss": 0.9389, + "step": 20191 + }, + { + "epoch": 3.67, + "learning_rate": 1.9001456133964326e-06, + "loss": 0.4478, + "step": 20192 + }, + { + "epoch": 3.67, + "learning_rate": 1.9000910083727705e-06, + "loss": 0.3226, + "step": 20193 + }, + { + "epoch": 3.67, + "learning_rate": 1.900036403349108e-06, + "loss": 1.0923, + "step": 20194 + }, + { + "epoch": 3.67, + "learning_rate": 1.899981798325446e-06, + "loss": 2.5061, + "step": 20195 + }, + { + "epoch": 3.67, + "learning_rate": 1.8999271933017837e-06, + "loss": 0.278, + "step": 20196 + }, + { + "epoch": 3.67, + "learning_rate": 1.8998725882781217e-06, + "loss": 1.1064, + "step": 20197 + }, + { + "epoch": 3.67, + "learning_rate": 1.8998179832544596e-06, + "loss": 0.1888, + "step": 20198 + }, + { + "epoch": 3.67, + "learning_rate": 1.8997633782307971e-06, + "loss": 0.0324, + "step": 20199 + }, + { + "epoch": 3.67, + "learning_rate": 1.899708773207135e-06, + "loss": 2.5152, + "step": 20200 + }, + { + "epoch": 3.67, + "learning_rate": 1.899654168183473e-06, + "loss": 1.4342, + "step": 20201 + }, + { + "epoch": 3.67, + "learning_rate": 1.8995995631598105e-06, + "loss": 1.2372, + "step": 20202 + }, + { + "epoch": 3.67, + "learning_rate": 1.8995449581361485e-06, + "loss": 4.2791, + "step": 20203 + }, + { + "epoch": 3.67, + "learning_rate": 1.8994903531124864e-06, + "loss": 0.0383, + "step": 20204 + }, + { + "epoch": 3.67, + "learning_rate": 1.8994357480888242e-06, + "loss": 0.4794, + "step": 20205 + }, + { + "epoch": 3.67, + "learning_rate": 1.8993811430651621e-06, + "loss": 0.0155, + "step": 20206 + }, + { + "epoch": 3.67, + "learning_rate": 1.8993265380415e-06, + "loss": 1.1355, + "step": 20207 + }, + { + "epoch": 3.67, + "learning_rate": 1.8992719330178376e-06, + "loss": 0.2358, + "step": 20208 + }, + { + "epoch": 3.68, + "learning_rate": 1.8992173279941755e-06, + "loss": 0.1793, + "step": 20209 + }, + { + "epoch": 3.68, + "learning_rate": 1.8991627229705132e-06, + "loss": 0.4185, + "step": 20210 + }, + { + "epoch": 3.68, + "learning_rate": 1.8991081179468512e-06, + "loss": 0.2679, + "step": 20211 + }, + { + "epoch": 3.68, + "learning_rate": 1.8990535129231891e-06, + "loss": 0.3532, + "step": 20212 + }, + { + "epoch": 3.68, + "learning_rate": 1.8989989078995267e-06, + "loss": 0.625, + "step": 20213 + }, + { + "epoch": 3.68, + "learning_rate": 1.8989443028758646e-06, + "loss": 0.2315, + "step": 20214 + }, + { + "epoch": 3.68, + "learning_rate": 1.8988896978522025e-06, + "loss": 0.3035, + "step": 20215 + }, + { + "epoch": 3.68, + "learning_rate": 1.8988350928285403e-06, + "loss": 0.4181, + "step": 20216 + }, + { + "epoch": 3.68, + "learning_rate": 1.8987804878048782e-06, + "loss": 1.1333, + "step": 20217 + }, + { + "epoch": 3.68, + "learning_rate": 1.8987258827812162e-06, + "loss": 0.0165, + "step": 20218 + }, + { + "epoch": 3.68, + "learning_rate": 1.8986712777575537e-06, + "loss": 0.0717, + "step": 20219 + }, + { + "epoch": 3.68, + "learning_rate": 1.8986166727338916e-06, + "loss": 1.7858, + "step": 20220 + }, + { + "epoch": 3.68, + "learning_rate": 1.8985620677102296e-06, + "loss": 1.0611, + "step": 20221 + }, + { + "epoch": 3.68, + "learning_rate": 1.898507462686567e-06, + "loss": 0.0088, + "step": 20222 + }, + { + "epoch": 3.68, + "learning_rate": 1.898452857662905e-06, + "loss": 1.4119, + "step": 20223 + }, + { + "epoch": 3.68, + "learning_rate": 1.8983982526392428e-06, + "loss": 0.2022, + "step": 20224 + }, + { + "epoch": 3.68, + "learning_rate": 1.8983436476155807e-06, + "loss": 0.4007, + "step": 20225 + }, + { + "epoch": 3.68, + "learning_rate": 1.8982890425919186e-06, + "loss": 0.8019, + "step": 20226 + }, + { + "epoch": 3.68, + "learning_rate": 1.8982344375682562e-06, + "loss": 0.7236, + "step": 20227 + }, + { + "epoch": 3.68, + "learning_rate": 1.8981798325445941e-06, + "loss": 1.5706, + "step": 20228 + }, + { + "epoch": 3.68, + "learning_rate": 1.898125227520932e-06, + "loss": 2.0893, + "step": 20229 + }, + { + "epoch": 3.68, + "learning_rate": 1.8980706224972698e-06, + "loss": 0.0128, + "step": 20230 + }, + { + "epoch": 3.68, + "learning_rate": 1.8980160174736077e-06, + "loss": 0.2877, + "step": 20231 + }, + { + "epoch": 3.68, + "learning_rate": 1.8979614124499457e-06, + "loss": 0.0109, + "step": 20232 + }, + { + "epoch": 3.68, + "learning_rate": 1.8979068074262832e-06, + "loss": 0.1216, + "step": 20233 + }, + { + "epoch": 3.68, + "learning_rate": 1.8978522024026211e-06, + "loss": 1.5751, + "step": 20234 + }, + { + "epoch": 3.68, + "learning_rate": 1.8977975973789587e-06, + "loss": 0.0278, + "step": 20235 + }, + { + "epoch": 3.68, + "learning_rate": 1.8977429923552966e-06, + "loss": 0.2959, + "step": 20236 + }, + { + "epoch": 3.68, + "learning_rate": 1.8976883873316345e-06, + "loss": 1.8503, + "step": 20237 + }, + { + "epoch": 3.68, + "learning_rate": 1.8976337823079723e-06, + "loss": 0.013, + "step": 20238 + }, + { + "epoch": 3.68, + "learning_rate": 1.8975791772843102e-06, + "loss": 1.2083, + "step": 20239 + }, + { + "epoch": 3.68, + "learning_rate": 1.8975245722606482e-06, + "loss": 0.2681, + "step": 20240 + }, + { + "epoch": 3.68, + "learning_rate": 1.8974699672369857e-06, + "loss": 0.8065, + "step": 20241 + }, + { + "epoch": 3.68, + "learning_rate": 1.8974153622133236e-06, + "loss": 0.0355, + "step": 20242 + }, + { + "epoch": 3.68, + "learning_rate": 1.8973607571896616e-06, + "loss": 1.6214, + "step": 20243 + }, + { + "epoch": 3.68, + "learning_rate": 1.8973061521659993e-06, + "loss": 0.0178, + "step": 20244 + }, + { + "epoch": 3.68, + "learning_rate": 1.8972515471423372e-06, + "loss": 0.0248, + "step": 20245 + }, + { + "epoch": 3.68, + "learning_rate": 1.8971969421186752e-06, + "loss": 1.3925, + "step": 20246 + }, + { + "epoch": 3.68, + "learning_rate": 1.8971423370950127e-06, + "loss": 0.7234, + "step": 20247 + }, + { + "epoch": 3.68, + "learning_rate": 1.8970877320713506e-06, + "loss": 0.9295, + "step": 20248 + }, + { + "epoch": 3.68, + "learning_rate": 1.8970331270476884e-06, + "loss": 1.2333, + "step": 20249 + }, + { + "epoch": 3.68, + "learning_rate": 1.8969785220240263e-06, + "loss": 0.9225, + "step": 20250 + }, + { + "epoch": 3.68, + "learning_rate": 1.896923917000364e-06, + "loss": 0.5416, + "step": 20251 + }, + { + "epoch": 3.68, + "learning_rate": 1.8968693119767018e-06, + "loss": 2.3994, + "step": 20252 + }, + { + "epoch": 3.68, + "learning_rate": 1.8968147069530397e-06, + "loss": 1.4849, + "step": 20253 + }, + { + "epoch": 3.68, + "learning_rate": 1.8967601019293777e-06, + "loss": 0.6095, + "step": 20254 + }, + { + "epoch": 3.68, + "learning_rate": 1.8967054969057152e-06, + "loss": 1.4207, + "step": 20255 + }, + { + "epoch": 3.68, + "learning_rate": 1.8966508918820531e-06, + "loss": 0.3061, + "step": 20256 + }, + { + "epoch": 3.68, + "learning_rate": 1.896596286858391e-06, + "loss": 1.5497, + "step": 20257 + }, + { + "epoch": 3.68, + "learning_rate": 1.8965416818347288e-06, + "loss": 0.0287, + "step": 20258 + }, + { + "epoch": 3.68, + "learning_rate": 1.8964870768110668e-06, + "loss": 0.5606, + "step": 20259 + }, + { + "epoch": 3.68, + "learning_rate": 1.8964324717874047e-06, + "loss": 0.4837, + "step": 20260 + }, + { + "epoch": 3.68, + "learning_rate": 1.8963778667637422e-06, + "loss": 0.7629, + "step": 20261 + }, + { + "epoch": 3.68, + "learning_rate": 1.8963232617400802e-06, + "loss": 0.007, + "step": 20262 + }, + { + "epoch": 3.68, + "learning_rate": 1.8962686567164179e-06, + "loss": 0.362, + "step": 20263 + }, + { + "epoch": 3.69, + "learning_rate": 1.8962140516927558e-06, + "loss": 0.1946, + "step": 20264 + }, + { + "epoch": 3.69, + "learning_rate": 1.8961594466690938e-06, + "loss": 0.0419, + "step": 20265 + }, + { + "epoch": 3.69, + "learning_rate": 1.8961048416454313e-06, + "loss": 0.0445, + "step": 20266 + }, + { + "epoch": 3.69, + "learning_rate": 1.8960502366217692e-06, + "loss": 0.0365, + "step": 20267 + }, + { + "epoch": 3.69, + "learning_rate": 1.8959956315981072e-06, + "loss": 0.017, + "step": 20268 + }, + { + "epoch": 3.69, + "learning_rate": 1.8959410265744447e-06, + "loss": 0.3131, + "step": 20269 + }, + { + "epoch": 3.69, + "learning_rate": 1.8958864215507827e-06, + "loss": 1.1644, + "step": 20270 + }, + { + "epoch": 3.69, + "learning_rate": 1.8958318165271206e-06, + "loss": 0.1899, + "step": 20271 + }, + { + "epoch": 3.69, + "learning_rate": 1.8957772115034583e-06, + "loss": 0.1771, + "step": 20272 + }, + { + "epoch": 3.69, + "learning_rate": 1.8957226064797963e-06, + "loss": 0.7042, + "step": 20273 + }, + { + "epoch": 3.69, + "learning_rate": 1.8956680014561342e-06, + "loss": 0.8692, + "step": 20274 + }, + { + "epoch": 3.69, + "learning_rate": 1.8956133964324717e-06, + "loss": 0.4092, + "step": 20275 + }, + { + "epoch": 3.69, + "learning_rate": 1.8955587914088097e-06, + "loss": 2.3574, + "step": 20276 + }, + { + "epoch": 3.69, + "learning_rate": 1.8955041863851474e-06, + "loss": 0.4162, + "step": 20277 + }, + { + "epoch": 3.69, + "learning_rate": 1.8954495813614854e-06, + "loss": 3.1285, + "step": 20278 + }, + { + "epoch": 3.69, + "learning_rate": 1.8953949763378233e-06, + "loss": 0.0169, + "step": 20279 + }, + { + "epoch": 3.69, + "learning_rate": 1.8953403713141608e-06, + "loss": 1.413, + "step": 20280 + }, + { + "epoch": 3.69, + "learning_rate": 1.8952857662904988e-06, + "loss": 0.2152, + "step": 20281 + }, + { + "epoch": 3.69, + "learning_rate": 1.8952311612668367e-06, + "loss": 1.3185, + "step": 20282 + }, + { + "epoch": 3.69, + "learning_rate": 1.8951765562431742e-06, + "loss": 0.2998, + "step": 20283 + }, + { + "epoch": 3.69, + "learning_rate": 1.8951219512195122e-06, + "loss": 0.1093, + "step": 20284 + }, + { + "epoch": 3.69, + "learning_rate": 1.8950673461958501e-06, + "loss": 0.679, + "step": 20285 + }, + { + "epoch": 3.69, + "learning_rate": 1.8950127411721878e-06, + "loss": 2.0927, + "step": 20286 + }, + { + "epoch": 3.69, + "learning_rate": 1.8949581361485258e-06, + "loss": 0.0685, + "step": 20287 + }, + { + "epoch": 3.69, + "learning_rate": 1.8949035311248637e-06, + "loss": 0.6558, + "step": 20288 + }, + { + "epoch": 3.69, + "learning_rate": 1.8948489261012012e-06, + "loss": 1.588, + "step": 20289 + }, + { + "epoch": 3.69, + "learning_rate": 1.8947943210775392e-06, + "loss": 1.2484, + "step": 20290 + }, + { + "epoch": 3.69, + "learning_rate": 1.894739716053877e-06, + "loss": 0.3548, + "step": 20291 + }, + { + "epoch": 3.69, + "learning_rate": 1.8946851110302149e-06, + "loss": 0.7411, + "step": 20292 + }, + { + "epoch": 3.69, + "learning_rate": 1.8946305060065528e-06, + "loss": 0.3947, + "step": 20293 + }, + { + "epoch": 3.69, + "learning_rate": 1.8945759009828903e-06, + "loss": 0.2937, + "step": 20294 + }, + { + "epoch": 3.69, + "learning_rate": 1.8945212959592283e-06, + "loss": 0.853, + "step": 20295 + }, + { + "epoch": 3.69, + "learning_rate": 1.8944666909355662e-06, + "loss": 3.7683, + "step": 20296 + }, + { + "epoch": 3.69, + "learning_rate": 1.894412085911904e-06, + "loss": 0.5564, + "step": 20297 + }, + { + "epoch": 3.69, + "learning_rate": 1.8943574808882419e-06, + "loss": 0.0495, + "step": 20298 + }, + { + "epoch": 3.69, + "learning_rate": 1.8943028758645798e-06, + "loss": 0.0728, + "step": 20299 + }, + { + "epoch": 3.69, + "learning_rate": 1.8942482708409174e-06, + "loss": 0.1765, + "step": 20300 + }, + { + "epoch": 3.69, + "learning_rate": 1.8941936658172553e-06, + "loss": 0.0184, + "step": 20301 + }, + { + "epoch": 3.69, + "learning_rate": 1.8941390607935932e-06, + "loss": 0.8854, + "step": 20302 + }, + { + "epoch": 3.69, + "learning_rate": 1.8940844557699308e-06, + "loss": 0.3142, + "step": 20303 + }, + { + "epoch": 3.69, + "learning_rate": 1.8940298507462687e-06, + "loss": 0.9477, + "step": 20304 + }, + { + "epoch": 3.69, + "learning_rate": 1.8939752457226064e-06, + "loss": 2.0238, + "step": 20305 + }, + { + "epoch": 3.69, + "learning_rate": 1.8939206406989444e-06, + "loss": 1.6866, + "step": 20306 + }, + { + "epoch": 3.69, + "learning_rate": 1.8938660356752823e-06, + "loss": 1.2559, + "step": 20307 + }, + { + "epoch": 3.69, + "learning_rate": 1.8938114306516198e-06, + "loss": 0.8278, + "step": 20308 + }, + { + "epoch": 3.69, + "learning_rate": 1.8937568256279578e-06, + "loss": 0.3101, + "step": 20309 + }, + { + "epoch": 3.69, + "learning_rate": 1.8937022206042957e-06, + "loss": 0.2971, + "step": 20310 + }, + { + "epoch": 3.69, + "learning_rate": 1.8936476155806335e-06, + "loss": 0.1292, + "step": 20311 + }, + { + "epoch": 3.69, + "learning_rate": 1.8935930105569714e-06, + "loss": 0.4922, + "step": 20312 + }, + { + "epoch": 3.69, + "learning_rate": 1.8935384055333093e-06, + "loss": 0.0779, + "step": 20313 + }, + { + "epoch": 3.69, + "learning_rate": 1.8934838005096469e-06, + "loss": 2.1857, + "step": 20314 + }, + { + "epoch": 3.69, + "learning_rate": 1.8934291954859848e-06, + "loss": 0.8658, + "step": 20315 + }, + { + "epoch": 3.69, + "learning_rate": 1.8933745904623223e-06, + "loss": 0.5268, + "step": 20316 + }, + { + "epoch": 3.69, + "learning_rate": 1.8933199854386603e-06, + "loss": 0.2405, + "step": 20317 + }, + { + "epoch": 3.69, + "learning_rate": 1.8932653804149982e-06, + "loss": 0.8629, + "step": 20318 + }, + { + "epoch": 3.7, + "learning_rate": 1.893210775391336e-06, + "loss": 0.8668, + "step": 20319 + }, + { + "epoch": 3.7, + "learning_rate": 1.8931561703676739e-06, + "loss": 0.8219, + "step": 20320 + }, + { + "epoch": 3.7, + "learning_rate": 1.8931015653440118e-06, + "loss": 0.2642, + "step": 20321 + }, + { + "epoch": 3.7, + "learning_rate": 1.8930469603203494e-06, + "loss": 1.0305, + "step": 20322 + }, + { + "epoch": 3.7, + "learning_rate": 1.8929923552966873e-06, + "loss": 0.2648, + "step": 20323 + }, + { + "epoch": 3.7, + "learning_rate": 1.8929377502730252e-06, + "loss": 1.0841, + "step": 20324 + }, + { + "epoch": 3.7, + "learning_rate": 1.892883145249363e-06, + "loss": 0.4979, + "step": 20325 + }, + { + "epoch": 3.7, + "learning_rate": 1.892828540225701e-06, + "loss": 0.008, + "step": 20326 + }, + { + "epoch": 3.7, + "learning_rate": 1.8927739352020389e-06, + "loss": 0.0547, + "step": 20327 + }, + { + "epoch": 3.7, + "learning_rate": 1.8927193301783764e-06, + "loss": 0.2429, + "step": 20328 + }, + { + "epoch": 3.7, + "learning_rate": 1.8926647251547143e-06, + "loss": 0.1101, + "step": 20329 + }, + { + "epoch": 3.7, + "learning_rate": 1.8926101201310518e-06, + "loss": 2.0804, + "step": 20330 + }, + { + "epoch": 3.7, + "learning_rate": 1.8925555151073898e-06, + "loss": 0.4644, + "step": 20331 + }, + { + "epoch": 3.7, + "learning_rate": 1.8925009100837277e-06, + "loss": 0.1434, + "step": 20332 + }, + { + "epoch": 3.7, + "learning_rate": 1.8924463050600655e-06, + "loss": 0.4227, + "step": 20333 + }, + { + "epoch": 3.7, + "learning_rate": 1.8923917000364034e-06, + "loss": 0.3693, + "step": 20334 + }, + { + "epoch": 3.7, + "learning_rate": 1.8923370950127413e-06, + "loss": 0.0528, + "step": 20335 + }, + { + "epoch": 3.7, + "learning_rate": 1.8922824899890789e-06, + "loss": 0.4771, + "step": 20336 + }, + { + "epoch": 3.7, + "learning_rate": 1.8922278849654168e-06, + "loss": 3.3304, + "step": 20337 + }, + { + "epoch": 3.7, + "learning_rate": 1.8921732799417548e-06, + "loss": 0.9928, + "step": 20338 + }, + { + "epoch": 3.7, + "learning_rate": 1.8921186749180925e-06, + "loss": 0.0911, + "step": 20339 + }, + { + "epoch": 3.7, + "learning_rate": 1.8920640698944304e-06, + "loss": 0.3237, + "step": 20340 + }, + { + "epoch": 3.7, + "learning_rate": 1.8920094648707684e-06, + "loss": 0.9382, + "step": 20341 + }, + { + "epoch": 3.7, + "learning_rate": 1.8919548598471059e-06, + "loss": 0.1409, + "step": 20342 + }, + { + "epoch": 3.7, + "learning_rate": 1.8919002548234438e-06, + "loss": 0.0805, + "step": 20343 + }, + { + "epoch": 3.7, + "learning_rate": 1.8918456497997816e-06, + "loss": 0.2894, + "step": 20344 + }, + { + "epoch": 3.7, + "learning_rate": 1.8917910447761195e-06, + "loss": 1.0147, + "step": 20345 + }, + { + "epoch": 3.7, + "learning_rate": 1.8917364397524575e-06, + "loss": 1.9494, + "step": 20346 + }, + { + "epoch": 3.7, + "learning_rate": 1.891681834728795e-06, + "loss": 0.0096, + "step": 20347 + }, + { + "epoch": 3.7, + "learning_rate": 1.891627229705133e-06, + "loss": 1.2608, + "step": 20348 + }, + { + "epoch": 3.7, + "learning_rate": 1.8915726246814709e-06, + "loss": 1.2466, + "step": 20349 + }, + { + "epoch": 3.7, + "learning_rate": 1.8915180196578084e-06, + "loss": 1.9382, + "step": 20350 + }, + { + "epoch": 3.7, + "learning_rate": 1.8914634146341463e-06, + "loss": 0.0796, + "step": 20351 + }, + { + "epoch": 3.7, + "learning_rate": 1.8914088096104843e-06, + "loss": 0.467, + "step": 20352 + }, + { + "epoch": 3.7, + "learning_rate": 1.891354204586822e-06, + "loss": 0.5253, + "step": 20353 + }, + { + "epoch": 3.7, + "learning_rate": 1.89129959956316e-06, + "loss": 1.3411, + "step": 20354 + }, + { + "epoch": 3.7, + "learning_rate": 1.8912449945394979e-06, + "loss": 1.3425, + "step": 20355 + }, + { + "epoch": 3.7, + "learning_rate": 1.8911903895158354e-06, + "loss": 0.1128, + "step": 20356 + }, + { + "epoch": 3.7, + "learning_rate": 1.8911357844921733e-06, + "loss": 0.4575, + "step": 20357 + }, + { + "epoch": 3.7, + "learning_rate": 1.891081179468511e-06, + "loss": 0.0183, + "step": 20358 + }, + { + "epoch": 3.7, + "learning_rate": 1.891026574444849e-06, + "loss": 0.0228, + "step": 20359 + }, + { + "epoch": 3.7, + "learning_rate": 1.890971969421187e-06, + "loss": 2.0426, + "step": 20360 + }, + { + "epoch": 3.7, + "learning_rate": 1.8909173643975245e-06, + "loss": 0.0971, + "step": 20361 + }, + { + "epoch": 3.7, + "learning_rate": 1.8908627593738624e-06, + "loss": 0.8537, + "step": 20362 + }, + { + "epoch": 3.7, + "learning_rate": 1.8908081543502004e-06, + "loss": 0.269, + "step": 20363 + }, + { + "epoch": 3.7, + "learning_rate": 1.890753549326538e-06, + "loss": 1.8223, + "step": 20364 + }, + { + "epoch": 3.7, + "learning_rate": 1.8906989443028758e-06, + "loss": 2.144, + "step": 20365 + }, + { + "epoch": 3.7, + "learning_rate": 1.8906443392792138e-06, + "loss": 1.2661, + "step": 20366 + }, + { + "epoch": 3.7, + "learning_rate": 1.8905897342555515e-06, + "loss": 0.1138, + "step": 20367 + }, + { + "epoch": 3.7, + "learning_rate": 1.8905351292318895e-06, + "loss": 0.9339, + "step": 20368 + }, + { + "epoch": 3.7, + "learning_rate": 1.8904805242082274e-06, + "loss": 0.0882, + "step": 20369 + }, + { + "epoch": 3.7, + "learning_rate": 1.890425919184565e-06, + "loss": 0.0174, + "step": 20370 + }, + { + "epoch": 3.7, + "learning_rate": 1.8903713141609029e-06, + "loss": 0.6795, + "step": 20371 + }, + { + "epoch": 3.7, + "learning_rate": 1.8903167091372406e-06, + "loss": 1.1592, + "step": 20372 + }, + { + "epoch": 3.7, + "learning_rate": 1.8902621041135785e-06, + "loss": 0.8231, + "step": 20373 + }, + { + "epoch": 3.71, + "learning_rate": 1.8902074990899165e-06, + "loss": 0.7682, + "step": 20374 + }, + { + "epoch": 3.71, + "learning_rate": 1.890152894066254e-06, + "loss": 0.879, + "step": 20375 + }, + { + "epoch": 3.71, + "learning_rate": 1.890098289042592e-06, + "loss": 0.7527, + "step": 20376 + }, + { + "epoch": 3.71, + "learning_rate": 1.8900436840189299e-06, + "loss": 0.6028, + "step": 20377 + }, + { + "epoch": 3.71, + "learning_rate": 1.8899890789952676e-06, + "loss": 0.7065, + "step": 20378 + }, + { + "epoch": 3.71, + "learning_rate": 1.8899344739716056e-06, + "loss": 0.3053, + "step": 20379 + }, + { + "epoch": 3.71, + "learning_rate": 1.8898798689479435e-06, + "loss": 0.4889, + "step": 20380 + }, + { + "epoch": 3.71, + "learning_rate": 1.889825263924281e-06, + "loss": 0.2161, + "step": 20381 + }, + { + "epoch": 3.71, + "learning_rate": 1.889770658900619e-06, + "loss": 0.6321, + "step": 20382 + }, + { + "epoch": 3.71, + "learning_rate": 1.8897160538769565e-06, + "loss": 0.3044, + "step": 20383 + }, + { + "epoch": 3.71, + "learning_rate": 1.8896614488532944e-06, + "loss": 0.9981, + "step": 20384 + }, + { + "epoch": 3.71, + "learning_rate": 1.8896068438296324e-06, + "loss": 0.7522, + "step": 20385 + }, + { + "epoch": 3.71, + "learning_rate": 1.8895522388059701e-06, + "loss": 0.0169, + "step": 20386 + }, + { + "epoch": 3.71, + "learning_rate": 1.889497633782308e-06, + "loss": 1.5922, + "step": 20387 + }, + { + "epoch": 3.71, + "learning_rate": 1.889443028758646e-06, + "loss": 0.5536, + "step": 20388 + }, + { + "epoch": 3.71, + "learning_rate": 1.8893884237349835e-06, + "loss": 0.2, + "step": 20389 + }, + { + "epoch": 3.71, + "learning_rate": 1.8893338187113215e-06, + "loss": 4.2654, + "step": 20390 + }, + { + "epoch": 3.71, + "learning_rate": 1.8892792136876594e-06, + "loss": 1.47, + "step": 20391 + }, + { + "epoch": 3.71, + "learning_rate": 1.8892246086639971e-06, + "loss": 1.2344, + "step": 20392 + }, + { + "epoch": 3.71, + "learning_rate": 1.889170003640335e-06, + "loss": 0.5426, + "step": 20393 + }, + { + "epoch": 3.71, + "learning_rate": 1.889115398616673e-06, + "loss": 0.4265, + "step": 20394 + }, + { + "epoch": 3.71, + "learning_rate": 1.8890607935930105e-06, + "loss": 1.1128, + "step": 20395 + }, + { + "epoch": 3.71, + "learning_rate": 1.8890061885693485e-06, + "loss": 0.4905, + "step": 20396 + }, + { + "epoch": 3.71, + "learning_rate": 1.888951583545686e-06, + "loss": 0.0106, + "step": 20397 + }, + { + "epoch": 3.71, + "learning_rate": 1.888896978522024e-06, + "loss": 2.1635, + "step": 20398 + }, + { + "epoch": 3.71, + "learning_rate": 1.8888423734983619e-06, + "loss": 0.9893, + "step": 20399 + }, + { + "epoch": 3.71, + "learning_rate": 1.8887877684746996e-06, + "loss": 0.0611, + "step": 20400 + }, + { + "epoch": 3.71, + "learning_rate": 1.8887331634510376e-06, + "loss": 0.0898, + "step": 20401 + }, + { + "epoch": 3.71, + "learning_rate": 1.8886785584273755e-06, + "loss": 0.1989, + "step": 20402 + }, + { + "epoch": 3.71, + "learning_rate": 1.888623953403713e-06, + "loss": 0.1363, + "step": 20403 + }, + { + "epoch": 3.71, + "learning_rate": 1.888569348380051e-06, + "loss": 0.8092, + "step": 20404 + }, + { + "epoch": 3.71, + "learning_rate": 1.888514743356389e-06, + "loss": 0.6594, + "step": 20405 + }, + { + "epoch": 3.71, + "learning_rate": 1.8884601383327266e-06, + "loss": 0.1723, + "step": 20406 + }, + { + "epoch": 3.71, + "learning_rate": 1.8884055333090646e-06, + "loss": 0.1737, + "step": 20407 + }, + { + "epoch": 3.71, + "learning_rate": 1.8883509282854025e-06, + "loss": 2.126, + "step": 20408 + }, + { + "epoch": 3.71, + "learning_rate": 1.88829632326174e-06, + "loss": 0.0296, + "step": 20409 + }, + { + "epoch": 3.71, + "learning_rate": 1.888241718238078e-06, + "loss": 2.2274, + "step": 20410 + }, + { + "epoch": 3.71, + "learning_rate": 1.8881871132144155e-06, + "loss": 0.7347, + "step": 20411 + }, + { + "epoch": 3.71, + "learning_rate": 1.8881325081907535e-06, + "loss": 0.2139, + "step": 20412 + }, + { + "epoch": 3.71, + "learning_rate": 1.8880779031670914e-06, + "loss": 1.9154, + "step": 20413 + }, + { + "epoch": 3.71, + "learning_rate": 1.8880232981434291e-06, + "loss": 0.329, + "step": 20414 + }, + { + "epoch": 3.71, + "learning_rate": 1.887968693119767e-06, + "loss": 0.7206, + "step": 20415 + }, + { + "epoch": 3.71, + "learning_rate": 1.887914088096105e-06, + "loss": 0.2479, + "step": 20416 + }, + { + "epoch": 3.71, + "learning_rate": 1.8878594830724425e-06, + "loss": 0.0253, + "step": 20417 + }, + { + "epoch": 3.71, + "learning_rate": 1.8878048780487805e-06, + "loss": 1.4193, + "step": 20418 + }, + { + "epoch": 3.71, + "learning_rate": 1.8877502730251184e-06, + "loss": 1.8719, + "step": 20419 + }, + { + "epoch": 3.71, + "learning_rate": 1.8876956680014562e-06, + "loss": 0.2571, + "step": 20420 + }, + { + "epoch": 3.71, + "learning_rate": 1.887641062977794e-06, + "loss": 0.5891, + "step": 20421 + }, + { + "epoch": 3.71, + "learning_rate": 1.887586457954132e-06, + "loss": 2.0082, + "step": 20422 + }, + { + "epoch": 3.71, + "learning_rate": 1.8875318529304696e-06, + "loss": 0.7445, + "step": 20423 + }, + { + "epoch": 3.71, + "learning_rate": 1.8874772479068075e-06, + "loss": 1.1909, + "step": 20424 + }, + { + "epoch": 3.71, + "learning_rate": 1.8874226428831452e-06, + "loss": 0.6458, + "step": 20425 + }, + { + "epoch": 3.71, + "learning_rate": 1.8873680378594832e-06, + "loss": 0.4325, + "step": 20426 + }, + { + "epoch": 3.71, + "learning_rate": 1.8873134328358211e-06, + "loss": 0.2193, + "step": 20427 + }, + { + "epoch": 3.71, + "learning_rate": 1.8872588278121586e-06, + "loss": 0.1943, + "step": 20428 + }, + { + "epoch": 3.72, + "learning_rate": 1.8872042227884966e-06, + "loss": 1.7514, + "step": 20429 + }, + { + "epoch": 3.72, + "learning_rate": 1.8871496177648345e-06, + "loss": 0.4664, + "step": 20430 + }, + { + "epoch": 3.72, + "learning_rate": 1.887095012741172e-06, + "loss": 1.0735, + "step": 20431 + }, + { + "epoch": 3.72, + "learning_rate": 1.88704040771751e-06, + "loss": 0.9862, + "step": 20432 + }, + { + "epoch": 3.72, + "learning_rate": 1.886985802693848e-06, + "loss": 0.2319, + "step": 20433 + }, + { + "epoch": 3.72, + "learning_rate": 1.8869311976701857e-06, + "loss": 0.395, + "step": 20434 + }, + { + "epoch": 3.72, + "learning_rate": 1.8868765926465236e-06, + "loss": 0.3072, + "step": 20435 + }, + { + "epoch": 3.72, + "learning_rate": 1.8868219876228616e-06, + "loss": 0.0934, + "step": 20436 + }, + { + "epoch": 3.72, + "learning_rate": 1.886767382599199e-06, + "loss": 0.0853, + "step": 20437 + }, + { + "epoch": 3.72, + "learning_rate": 1.886712777575537e-06, + "loss": 0.2008, + "step": 20438 + }, + { + "epoch": 3.72, + "learning_rate": 1.8866581725518748e-06, + "loss": 0.4396, + "step": 20439 + }, + { + "epoch": 3.72, + "learning_rate": 1.8866035675282127e-06, + "loss": 0.9609, + "step": 20440 + }, + { + "epoch": 3.72, + "learning_rate": 1.8865489625045506e-06, + "loss": 0.1278, + "step": 20441 + }, + { + "epoch": 3.72, + "learning_rate": 1.8864943574808882e-06, + "loss": 0.2109, + "step": 20442 + }, + { + "epoch": 3.72, + "learning_rate": 1.886439752457226e-06, + "loss": 1.4874, + "step": 20443 + }, + { + "epoch": 3.72, + "learning_rate": 1.886385147433564e-06, + "loss": 0.3301, + "step": 20444 + }, + { + "epoch": 3.72, + "learning_rate": 1.8863305424099016e-06, + "loss": 2.8588, + "step": 20445 + }, + { + "epoch": 3.72, + "learning_rate": 1.8862759373862395e-06, + "loss": 0.1733, + "step": 20446 + }, + { + "epoch": 3.72, + "learning_rate": 1.8862213323625775e-06, + "loss": 0.043, + "step": 20447 + }, + { + "epoch": 3.72, + "learning_rate": 1.8861667273389152e-06, + "loss": 0.1224, + "step": 20448 + }, + { + "epoch": 3.72, + "learning_rate": 1.8861121223152531e-06, + "loss": 3.4733, + "step": 20449 + }, + { + "epoch": 3.72, + "learning_rate": 1.886057517291591e-06, + "loss": 0.388, + "step": 20450 + }, + { + "epoch": 3.72, + "learning_rate": 1.8860029122679286e-06, + "loss": 0.4892, + "step": 20451 + }, + { + "epoch": 3.72, + "learning_rate": 1.8859483072442665e-06, + "loss": 0.3586, + "step": 20452 + }, + { + "epoch": 3.72, + "learning_rate": 1.8858937022206043e-06, + "loss": 0.4506, + "step": 20453 + }, + { + "epoch": 3.72, + "learning_rate": 1.8858390971969422e-06, + "loss": 0.3967, + "step": 20454 + }, + { + "epoch": 3.72, + "learning_rate": 1.8857844921732802e-06, + "loss": 0.0066, + "step": 20455 + }, + { + "epoch": 3.72, + "learning_rate": 1.8857298871496177e-06, + "loss": 0.6965, + "step": 20456 + }, + { + "epoch": 3.72, + "learning_rate": 1.8856752821259556e-06, + "loss": 1.4151, + "step": 20457 + }, + { + "epoch": 3.72, + "learning_rate": 1.8856206771022936e-06, + "loss": 0.8521, + "step": 20458 + }, + { + "epoch": 3.72, + "learning_rate": 1.885566072078631e-06, + "loss": 1.2678, + "step": 20459 + }, + { + "epoch": 3.72, + "learning_rate": 1.885511467054969e-06, + "loss": 1.4649, + "step": 20460 + }, + { + "epoch": 3.72, + "learning_rate": 1.885456862031307e-06, + "loss": 0.2209, + "step": 20461 + }, + { + "epoch": 3.72, + "learning_rate": 1.8854022570076447e-06, + "loss": 0.3311, + "step": 20462 + }, + { + "epoch": 3.72, + "learning_rate": 1.8853476519839826e-06, + "loss": 0.7818, + "step": 20463 + }, + { + "epoch": 3.72, + "learning_rate": 1.8852930469603202e-06, + "loss": 0.4543, + "step": 20464 + }, + { + "epoch": 3.72, + "learning_rate": 1.8852384419366581e-06, + "loss": 1.1242, + "step": 20465 + }, + { + "epoch": 3.72, + "learning_rate": 1.885183836912996e-06, + "loss": 0.8062, + "step": 20466 + }, + { + "epoch": 3.72, + "learning_rate": 1.8851292318893338e-06, + "loss": 0.0896, + "step": 20467 + }, + { + "epoch": 3.72, + "learning_rate": 1.8850746268656717e-06, + "loss": 1.1431, + "step": 20468 + }, + { + "epoch": 3.72, + "learning_rate": 1.8850200218420097e-06, + "loss": 0.2347, + "step": 20469 + }, + { + "epoch": 3.72, + "learning_rate": 1.8849654168183472e-06, + "loss": 1.0655, + "step": 20470 + }, + { + "epoch": 3.72, + "learning_rate": 1.8849108117946851e-06, + "loss": 2.5305, + "step": 20471 + }, + { + "epoch": 3.72, + "learning_rate": 1.884856206771023e-06, + "loss": 1.032, + "step": 20472 + }, + { + "epoch": 3.72, + "learning_rate": 1.8848016017473608e-06, + "loss": 0.0954, + "step": 20473 + }, + { + "epoch": 3.72, + "learning_rate": 1.8847469967236988e-06, + "loss": 0.0823, + "step": 20474 + }, + { + "epoch": 3.72, + "learning_rate": 1.8846923917000367e-06, + "loss": 1.4213, + "step": 20475 + }, + { + "epoch": 3.72, + "learning_rate": 1.8846377866763742e-06, + "loss": 0.2828, + "step": 20476 + }, + { + "epoch": 3.72, + "learning_rate": 1.8845831816527122e-06, + "loss": 0.8098, + "step": 20477 + }, + { + "epoch": 3.72, + "learning_rate": 1.8845285766290497e-06, + "loss": 2.0346, + "step": 20478 + }, + { + "epoch": 3.72, + "learning_rate": 1.8844739716053876e-06, + "loss": 1.8387, + "step": 20479 + }, + { + "epoch": 3.72, + "learning_rate": 1.8844193665817256e-06, + "loss": 0.3157, + "step": 20480 + }, + { + "epoch": 3.72, + "learning_rate": 1.8843647615580633e-06, + "loss": 1.9244, + "step": 20481 + }, + { + "epoch": 3.72, + "learning_rate": 1.8843101565344012e-06, + "loss": 1.3981, + "step": 20482 + }, + { + "epoch": 3.72, + "learning_rate": 1.8842555515107392e-06, + "loss": 0.8553, + "step": 20483 + }, + { + "epoch": 3.73, + "learning_rate": 1.8842009464870767e-06, + "loss": 0.7575, + "step": 20484 + }, + { + "epoch": 3.73, + "learning_rate": 1.8841463414634146e-06, + "loss": 0.8566, + "step": 20485 + }, + { + "epoch": 3.73, + "learning_rate": 1.8840917364397526e-06, + "loss": 0.2063, + "step": 20486 + }, + { + "epoch": 3.73, + "learning_rate": 1.8840371314160903e-06, + "loss": 0.6662, + "step": 20487 + }, + { + "epoch": 3.73, + "learning_rate": 1.8839825263924283e-06, + "loss": 0.0224, + "step": 20488 + }, + { + "epoch": 3.73, + "learning_rate": 1.8839279213687662e-06, + "loss": 0.7464, + "step": 20489 + }, + { + "epoch": 3.73, + "learning_rate": 1.8838733163451037e-06, + "loss": 0.7182, + "step": 20490 + }, + { + "epoch": 3.73, + "learning_rate": 1.8838187113214417e-06, + "loss": 0.018, + "step": 20491 + }, + { + "epoch": 3.73, + "learning_rate": 1.8837641062977792e-06, + "loss": 1.5237, + "step": 20492 + }, + { + "epoch": 3.73, + "learning_rate": 1.8837095012741171e-06, + "loss": 0.893, + "step": 20493 + }, + { + "epoch": 3.73, + "learning_rate": 1.883654896250455e-06, + "loss": 3.4539, + "step": 20494 + }, + { + "epoch": 3.73, + "learning_rate": 1.8836002912267928e-06, + "loss": 0.0182, + "step": 20495 + }, + { + "epoch": 3.73, + "learning_rate": 1.8835456862031308e-06, + "loss": 0.292, + "step": 20496 + }, + { + "epoch": 3.73, + "learning_rate": 1.8834910811794687e-06, + "loss": 2.289, + "step": 20497 + }, + { + "epoch": 3.73, + "learning_rate": 1.8834364761558062e-06, + "loss": 0.9407, + "step": 20498 + }, + { + "epoch": 3.73, + "learning_rate": 1.8833818711321442e-06, + "loss": 0.3152, + "step": 20499 + }, + { + "epoch": 3.73, + "learning_rate": 1.883327266108482e-06, + "loss": 1.344, + "step": 20500 + }, + { + "epoch": 3.73, + "learning_rate": 1.8832726610848198e-06, + "loss": 2.4006, + "step": 20501 + }, + { + "epoch": 3.73, + "learning_rate": 1.8832180560611578e-06, + "loss": 0.8082, + "step": 20502 + }, + { + "epoch": 3.73, + "learning_rate": 1.8831634510374957e-06, + "loss": 0.0588, + "step": 20503 + }, + { + "epoch": 3.73, + "learning_rate": 1.8831088460138332e-06, + "loss": 0.6074, + "step": 20504 + }, + { + "epoch": 3.73, + "learning_rate": 1.8830542409901712e-06, + "loss": 0.299, + "step": 20505 + }, + { + "epoch": 3.73, + "learning_rate": 1.882999635966509e-06, + "loss": 0.4601, + "step": 20506 + }, + { + "epoch": 3.73, + "learning_rate": 1.8829450309428469e-06, + "loss": 1.2381, + "step": 20507 + }, + { + "epoch": 3.73, + "learning_rate": 1.8828904259191848e-06, + "loss": 0.232, + "step": 20508 + }, + { + "epoch": 3.73, + "learning_rate": 1.8828358208955223e-06, + "loss": 0.3014, + "step": 20509 + }, + { + "epoch": 3.73, + "learning_rate": 1.8827812158718603e-06, + "loss": 1.4854, + "step": 20510 + }, + { + "epoch": 3.73, + "learning_rate": 1.8827266108481982e-06, + "loss": 0.8938, + "step": 20511 + }, + { + "epoch": 3.73, + "learning_rate": 1.8826720058245357e-06, + "loss": 0.6781, + "step": 20512 + }, + { + "epoch": 3.73, + "learning_rate": 1.8826174008008737e-06, + "loss": 1.2344, + "step": 20513 + }, + { + "epoch": 3.73, + "learning_rate": 1.8825627957772116e-06, + "loss": 0.5152, + "step": 20514 + }, + { + "epoch": 3.73, + "learning_rate": 1.8825081907535493e-06, + "loss": 0.2241, + "step": 20515 + }, + { + "epoch": 3.73, + "learning_rate": 1.8824535857298873e-06, + "loss": 0.1219, + "step": 20516 + }, + { + "epoch": 3.73, + "learning_rate": 1.8823989807062252e-06, + "loss": 0.4443, + "step": 20517 + }, + { + "epoch": 3.73, + "learning_rate": 1.8823443756825628e-06, + "loss": 0.2646, + "step": 20518 + }, + { + "epoch": 3.73, + "learning_rate": 1.8822897706589007e-06, + "loss": 0.3207, + "step": 20519 + }, + { + "epoch": 3.73, + "learning_rate": 1.8822351656352384e-06, + "loss": 1.7599, + "step": 20520 + }, + { + "epoch": 3.73, + "learning_rate": 1.8821805606115764e-06, + "loss": 1.4955, + "step": 20521 + }, + { + "epoch": 3.73, + "learning_rate": 1.8821259555879143e-06, + "loss": 0.3508, + "step": 20522 + }, + { + "epoch": 3.73, + "learning_rate": 1.8820713505642518e-06, + "loss": 2.0347, + "step": 20523 + }, + { + "epoch": 3.73, + "learning_rate": 1.8820167455405898e-06, + "loss": 0.7173, + "step": 20524 + }, + { + "epoch": 3.73, + "learning_rate": 1.8819621405169277e-06, + "loss": 0.0519, + "step": 20525 + }, + { + "epoch": 3.73, + "learning_rate": 1.8819075354932652e-06, + "loss": 0.3709, + "step": 20526 + }, + { + "epoch": 3.73, + "learning_rate": 1.8818529304696032e-06, + "loss": 2.0905, + "step": 20527 + }, + { + "epoch": 3.73, + "learning_rate": 1.8817983254459411e-06, + "loss": 0.0111, + "step": 20528 + }, + { + "epoch": 3.73, + "learning_rate": 1.8817437204222789e-06, + "loss": 0.0795, + "step": 20529 + }, + { + "epoch": 3.73, + "learning_rate": 1.8816891153986168e-06, + "loss": 1.4387, + "step": 20530 + }, + { + "epoch": 3.73, + "learning_rate": 1.8816345103749547e-06, + "loss": 1.5074, + "step": 20531 + }, + { + "epoch": 3.73, + "learning_rate": 1.8815799053512923e-06, + "loss": 0.01, + "step": 20532 + }, + { + "epoch": 3.73, + "learning_rate": 1.8815253003276302e-06, + "loss": 0.0249, + "step": 20533 + }, + { + "epoch": 3.73, + "learning_rate": 1.881470695303968e-06, + "loss": 0.0098, + "step": 20534 + }, + { + "epoch": 3.73, + "learning_rate": 1.8814160902803059e-06, + "loss": 0.1791, + "step": 20535 + }, + { + "epoch": 3.73, + "learning_rate": 1.8813614852566438e-06, + "loss": 0.0119, + "step": 20536 + }, + { + "epoch": 3.73, + "learning_rate": 1.8813068802329813e-06, + "loss": 1.3468, + "step": 20537 + }, + { + "epoch": 3.73, + "learning_rate": 1.8812522752093193e-06, + "loss": 0.612, + "step": 20538 + }, + { + "epoch": 3.74, + "learning_rate": 1.8811976701856572e-06, + "loss": 1.553, + "step": 20539 + }, + { + "epoch": 3.74, + "learning_rate": 1.8811430651619948e-06, + "loss": 0.2312, + "step": 20540 + }, + { + "epoch": 3.74, + "learning_rate": 1.8810884601383327e-06, + "loss": 0.0798, + "step": 20541 + }, + { + "epoch": 3.74, + "learning_rate": 1.8810338551146706e-06, + "loss": 0.6544, + "step": 20542 + }, + { + "epoch": 3.74, + "learning_rate": 1.8809792500910084e-06, + "loss": 1.7794, + "step": 20543 + }, + { + "epoch": 3.74, + "learning_rate": 1.8809246450673463e-06, + "loss": 1.5985, + "step": 20544 + }, + { + "epoch": 3.74, + "learning_rate": 1.8808700400436838e-06, + "loss": 0.646, + "step": 20545 + }, + { + "epoch": 3.74, + "learning_rate": 1.8808154350200218e-06, + "loss": 0.343, + "step": 20546 + }, + { + "epoch": 3.74, + "learning_rate": 1.8807608299963597e-06, + "loss": 0.0079, + "step": 20547 + }, + { + "epoch": 3.74, + "learning_rate": 1.8807062249726975e-06, + "loss": 0.3353, + "step": 20548 + }, + { + "epoch": 3.74, + "learning_rate": 1.8806516199490354e-06, + "loss": 0.1835, + "step": 20549 + }, + { + "epoch": 3.74, + "learning_rate": 1.8805970149253733e-06, + "loss": 1.1285, + "step": 20550 + }, + { + "epoch": 3.74, + "learning_rate": 1.8805424099017109e-06, + "loss": 0.012, + "step": 20551 + }, + { + "epoch": 3.74, + "learning_rate": 1.8804878048780488e-06, + "loss": 0.2993, + "step": 20552 + }, + { + "epoch": 3.74, + "learning_rate": 1.8804331998543867e-06, + "loss": 1.2504, + "step": 20553 + }, + { + "epoch": 3.74, + "learning_rate": 1.8803785948307245e-06, + "loss": 0.0268, + "step": 20554 + }, + { + "epoch": 3.74, + "learning_rate": 1.8803239898070624e-06, + "loss": 2.5563, + "step": 20555 + }, + { + "epoch": 3.74, + "learning_rate": 1.8802693847834004e-06, + "loss": 0.4847, + "step": 20556 + }, + { + "epoch": 3.74, + "learning_rate": 1.8802147797597379e-06, + "loss": 0.8178, + "step": 20557 + }, + { + "epoch": 3.74, + "learning_rate": 1.8801601747360758e-06, + "loss": 0.2815, + "step": 20558 + }, + { + "epoch": 3.74, + "learning_rate": 1.8801055697124134e-06, + "loss": 0.0092, + "step": 20559 + }, + { + "epoch": 3.74, + "learning_rate": 1.8800509646887513e-06, + "loss": 0.9479, + "step": 20560 + }, + { + "epoch": 3.74, + "learning_rate": 1.8799963596650892e-06, + "loss": 1.1061, + "step": 20561 + }, + { + "epoch": 3.74, + "learning_rate": 1.879941754641427e-06, + "loss": 0.1933, + "step": 20562 + }, + { + "epoch": 3.74, + "learning_rate": 1.879887149617765e-06, + "loss": 2.3897, + "step": 20563 + }, + { + "epoch": 3.74, + "learning_rate": 1.8798325445941029e-06, + "loss": 1.5558, + "step": 20564 + }, + { + "epoch": 3.74, + "learning_rate": 1.8797779395704404e-06, + "loss": 0.5944, + "step": 20565 + }, + { + "epoch": 3.74, + "learning_rate": 1.8797233345467783e-06, + "loss": 0.8549, + "step": 20566 + }, + { + "epoch": 3.74, + "learning_rate": 1.8796687295231163e-06, + "loss": 0.7953, + "step": 20567 + }, + { + "epoch": 3.74, + "learning_rate": 1.879614124499454e-06, + "loss": 0.2107, + "step": 20568 + }, + { + "epoch": 3.74, + "learning_rate": 1.879559519475792e-06, + "loss": 0.7451, + "step": 20569 + }, + { + "epoch": 3.74, + "learning_rate": 1.8795049144521299e-06, + "loss": 1.1668, + "step": 20570 + }, + { + "epoch": 3.74, + "learning_rate": 1.8794503094284674e-06, + "loss": 1.5152, + "step": 20571 + }, + { + "epoch": 3.74, + "learning_rate": 1.8793957044048053e-06, + "loss": 0.437, + "step": 20572 + }, + { + "epoch": 3.74, + "learning_rate": 1.8793410993811429e-06, + "loss": 1.2066, + "step": 20573 + }, + { + "epoch": 3.74, + "learning_rate": 1.8792864943574808e-06, + "loss": 0.8416, + "step": 20574 + }, + { + "epoch": 3.74, + "learning_rate": 1.8792318893338188e-06, + "loss": 0.4536, + "step": 20575 + }, + { + "epoch": 3.74, + "learning_rate": 1.8791772843101565e-06, + "loss": 0.6247, + "step": 20576 + }, + { + "epoch": 3.74, + "learning_rate": 1.8791226792864944e-06, + "loss": 1.523, + "step": 20577 + }, + { + "epoch": 3.74, + "learning_rate": 1.8790680742628324e-06, + "loss": 1.0945, + "step": 20578 + }, + { + "epoch": 3.74, + "learning_rate": 1.8790134692391699e-06, + "loss": 1.7, + "step": 20579 + }, + { + "epoch": 3.74, + "learning_rate": 1.8789588642155078e-06, + "loss": 1.9397, + "step": 20580 + }, + { + "epoch": 3.74, + "learning_rate": 1.8789042591918458e-06, + "loss": 0.0318, + "step": 20581 + }, + { + "epoch": 3.74, + "learning_rate": 1.8788496541681835e-06, + "loss": 2.8582, + "step": 20582 + }, + { + "epoch": 3.74, + "learning_rate": 1.8787950491445215e-06, + "loss": 0.8982, + "step": 20583 + }, + { + "epoch": 3.74, + "learning_rate": 1.8787404441208594e-06, + "loss": 0.2381, + "step": 20584 + }, + { + "epoch": 3.74, + "learning_rate": 1.878685839097197e-06, + "loss": 1.1007, + "step": 20585 + }, + { + "epoch": 3.74, + "learning_rate": 1.8786312340735349e-06, + "loss": 1.9291, + "step": 20586 + }, + { + "epoch": 3.74, + "learning_rate": 1.8785766290498726e-06, + "loss": 0.2436, + "step": 20587 + }, + { + "epoch": 3.74, + "learning_rate": 1.8785220240262103e-06, + "loss": 0.922, + "step": 20588 + }, + { + "epoch": 3.74, + "learning_rate": 1.8784674190025483e-06, + "loss": 0.2271, + "step": 20589 + }, + { + "epoch": 3.74, + "learning_rate": 1.878412813978886e-06, + "loss": 0.0165, + "step": 20590 + }, + { + "epoch": 3.74, + "learning_rate": 1.878358208955224e-06, + "loss": 1.4372, + "step": 20591 + }, + { + "epoch": 3.74, + "learning_rate": 1.8783036039315619e-06, + "loss": 1.5057, + "step": 20592 + }, + { + "epoch": 3.74, + "learning_rate": 1.8782489989078994e-06, + "loss": 0.3125, + "step": 20593 + }, + { + "epoch": 3.75, + "learning_rate": 1.8781943938842373e-06, + "loss": 0.21, + "step": 20594 + }, + { + "epoch": 3.75, + "learning_rate": 1.8781397888605753e-06, + "loss": 0.0947, + "step": 20595 + }, + { + "epoch": 3.75, + "learning_rate": 1.878085183836913e-06, + "loss": 0.5691, + "step": 20596 + }, + { + "epoch": 3.75, + "learning_rate": 1.878030578813251e-06, + "loss": 0.0125, + "step": 20597 + }, + { + "epoch": 3.75, + "learning_rate": 1.877975973789589e-06, + "loss": 2.784, + "step": 20598 + }, + { + "epoch": 3.75, + "learning_rate": 1.8779213687659264e-06, + "loss": 1.3756, + "step": 20599 + }, + { + "epoch": 3.75, + "learning_rate": 1.8778667637422644e-06, + "loss": 0.5915, + "step": 20600 + }, + { + "epoch": 3.75, + "learning_rate": 1.877812158718602e-06, + "loss": 0.8137, + "step": 20601 + }, + { + "epoch": 3.75, + "learning_rate": 1.87775755369494e-06, + "loss": 1.7264, + "step": 20602 + }, + { + "epoch": 3.75, + "learning_rate": 1.877702948671278e-06, + "loss": 0.0986, + "step": 20603 + }, + { + "epoch": 3.75, + "learning_rate": 1.8776483436476155e-06, + "loss": 0.5957, + "step": 20604 + }, + { + "epoch": 3.75, + "learning_rate": 1.8775937386239535e-06, + "loss": 2.2078, + "step": 20605 + }, + { + "epoch": 3.75, + "learning_rate": 1.8775391336002914e-06, + "loss": 0.0144, + "step": 20606 + }, + { + "epoch": 3.75, + "learning_rate": 1.877484528576629e-06, + "loss": 0.4457, + "step": 20607 + }, + { + "epoch": 3.75, + "learning_rate": 1.8774299235529669e-06, + "loss": 0.0059, + "step": 20608 + }, + { + "epoch": 3.75, + "learning_rate": 1.8773753185293048e-06, + "loss": 0.3559, + "step": 20609 + }, + { + "epoch": 3.75, + "learning_rate": 1.8773207135056425e-06, + "loss": 2.6166, + "step": 20610 + }, + { + "epoch": 3.75, + "learning_rate": 1.8772661084819805e-06, + "loss": 1.7608, + "step": 20611 + }, + { + "epoch": 3.75, + "learning_rate": 1.8772115034583184e-06, + "loss": 0.7085, + "step": 20612 + }, + { + "epoch": 3.75, + "learning_rate": 1.877156898434656e-06, + "loss": 3.8594, + "step": 20613 + }, + { + "epoch": 3.75, + "learning_rate": 1.8771022934109939e-06, + "loss": 7.4296, + "step": 20614 + }, + { + "epoch": 3.75, + "learning_rate": 1.8770476883873316e-06, + "loss": 0.3602, + "step": 20615 + }, + { + "epoch": 3.75, + "learning_rate": 1.8769930833636696e-06, + "loss": 0.0626, + "step": 20616 + }, + { + "epoch": 3.75, + "learning_rate": 1.8769384783400075e-06, + "loss": 0.338, + "step": 20617 + }, + { + "epoch": 3.75, + "learning_rate": 1.876883873316345e-06, + "loss": 1.2167, + "step": 20618 + }, + { + "epoch": 3.75, + "learning_rate": 1.876829268292683e-06, + "loss": 1.5706, + "step": 20619 + }, + { + "epoch": 3.75, + "learning_rate": 1.876774663269021e-06, + "loss": 2.3953, + "step": 20620 + }, + { + "epoch": 3.75, + "learning_rate": 1.8767200582453584e-06, + "loss": 1.0265, + "step": 20621 + }, + { + "epoch": 3.75, + "learning_rate": 1.8766654532216964e-06, + "loss": 0.0414, + "step": 20622 + }, + { + "epoch": 3.75, + "learning_rate": 1.8766108481980343e-06, + "loss": 2.1221, + "step": 20623 + }, + { + "epoch": 3.75, + "learning_rate": 1.876556243174372e-06, + "loss": 0.4517, + "step": 20624 + }, + { + "epoch": 3.75, + "learning_rate": 1.87650163815071e-06, + "loss": 0.2296, + "step": 20625 + }, + { + "epoch": 3.75, + "learning_rate": 1.8764470331270475e-06, + "loss": 0.5348, + "step": 20626 + }, + { + "epoch": 3.75, + "learning_rate": 1.8763924281033855e-06, + "loss": 0.3225, + "step": 20627 + }, + { + "epoch": 3.75, + "learning_rate": 1.8763378230797234e-06, + "loss": 0.9563, + "step": 20628 + }, + { + "epoch": 3.75, + "learning_rate": 1.8762832180560611e-06, + "loss": 0.5396, + "step": 20629 + }, + { + "epoch": 3.75, + "learning_rate": 1.876228613032399e-06, + "loss": 0.3921, + "step": 20630 + }, + { + "epoch": 3.75, + "learning_rate": 1.876174008008737e-06, + "loss": 1.0743, + "step": 20631 + }, + { + "epoch": 3.75, + "learning_rate": 1.8761194029850745e-06, + "loss": 0.1069, + "step": 20632 + }, + { + "epoch": 3.75, + "learning_rate": 1.8760647979614125e-06, + "loss": 1.0699, + "step": 20633 + }, + { + "epoch": 3.75, + "learning_rate": 1.8760101929377504e-06, + "loss": 1.5586, + "step": 20634 + }, + { + "epoch": 3.75, + "learning_rate": 1.8759555879140882e-06, + "loss": 1.7376, + "step": 20635 + }, + { + "epoch": 3.75, + "learning_rate": 1.875900982890426e-06, + "loss": 0.4903, + "step": 20636 + }, + { + "epoch": 3.75, + "learning_rate": 1.875846377866764e-06, + "loss": 0.7869, + "step": 20637 + }, + { + "epoch": 3.75, + "learning_rate": 1.8757917728431016e-06, + "loss": 0.7593, + "step": 20638 + }, + { + "epoch": 3.75, + "learning_rate": 1.8757371678194395e-06, + "loss": 0.0174, + "step": 20639 + }, + { + "epoch": 3.75, + "learning_rate": 1.875682562795777e-06, + "loss": 0.5712, + "step": 20640 + }, + { + "epoch": 3.75, + "learning_rate": 1.875627957772115e-06, + "loss": 0.0861, + "step": 20641 + }, + { + "epoch": 3.75, + "learning_rate": 1.875573352748453e-06, + "loss": 1.6712, + "step": 20642 + }, + { + "epoch": 3.75, + "learning_rate": 1.8755187477247906e-06, + "loss": 2.0557, + "step": 20643 + }, + { + "epoch": 3.75, + "learning_rate": 1.8754641427011286e-06, + "loss": 0.4851, + "step": 20644 + }, + { + "epoch": 3.75, + "learning_rate": 1.8754095376774665e-06, + "loss": 1.2616, + "step": 20645 + }, + { + "epoch": 3.75, + "learning_rate": 1.875354932653804e-06, + "loss": 0.3522, + "step": 20646 + }, + { + "epoch": 3.75, + "learning_rate": 1.875300327630142e-06, + "loss": 1.3007, + "step": 20647 + }, + { + "epoch": 3.75, + "learning_rate": 1.87524572260648e-06, + "loss": 0.8148, + "step": 20648 + }, + { + "epoch": 3.76, + "learning_rate": 1.8751911175828177e-06, + "loss": 1.5492, + "step": 20649 + }, + { + "epoch": 3.76, + "learning_rate": 1.8751365125591556e-06, + "loss": 0.019, + "step": 20650 + }, + { + "epoch": 3.76, + "learning_rate": 1.8750819075354936e-06, + "loss": 0.8788, + "step": 20651 + }, + { + "epoch": 3.76, + "learning_rate": 1.875027302511831e-06, + "loss": 0.6227, + "step": 20652 + }, + { + "epoch": 3.76, + "learning_rate": 1.874972697488169e-06, + "loss": 0.2379, + "step": 20653 + }, + { + "epoch": 3.76, + "learning_rate": 1.8749180924645065e-06, + "loss": 2.7525, + "step": 20654 + }, + { + "epoch": 3.76, + "learning_rate": 1.8748634874408445e-06, + "loss": 0.1053, + "step": 20655 + }, + { + "epoch": 3.76, + "learning_rate": 1.8748088824171824e-06, + "loss": 0.4825, + "step": 20656 + }, + { + "epoch": 3.76, + "learning_rate": 1.8747542773935202e-06, + "loss": 0.7018, + "step": 20657 + }, + { + "epoch": 3.76, + "learning_rate": 1.874699672369858e-06, + "loss": 1.6866, + "step": 20658 + }, + { + "epoch": 3.76, + "learning_rate": 1.874645067346196e-06, + "loss": 0.3835, + "step": 20659 + }, + { + "epoch": 3.76, + "learning_rate": 1.8745904623225336e-06, + "loss": 0.5093, + "step": 20660 + }, + { + "epoch": 3.76, + "learning_rate": 1.8745358572988715e-06, + "loss": 0.5725, + "step": 20661 + }, + { + "epoch": 3.76, + "learning_rate": 1.8744812522752095e-06, + "loss": 0.6194, + "step": 20662 + }, + { + "epoch": 3.76, + "learning_rate": 1.8744266472515472e-06, + "loss": 1.5114, + "step": 20663 + }, + { + "epoch": 3.76, + "learning_rate": 1.8743720422278851e-06, + "loss": 0.3702, + "step": 20664 + }, + { + "epoch": 3.76, + "learning_rate": 1.874317437204223e-06, + "loss": 0.6414, + "step": 20665 + }, + { + "epoch": 3.76, + "learning_rate": 1.8742628321805606e-06, + "loss": 1.3625, + "step": 20666 + }, + { + "epoch": 3.76, + "learning_rate": 1.8742082271568985e-06, + "loss": 0.6106, + "step": 20667 + }, + { + "epoch": 3.76, + "learning_rate": 1.874153622133236e-06, + "loss": 0.7278, + "step": 20668 + }, + { + "epoch": 3.76, + "learning_rate": 1.874099017109574e-06, + "loss": 0.0156, + "step": 20669 + }, + { + "epoch": 3.76, + "learning_rate": 1.874044412085912e-06, + "loss": 0.2097, + "step": 20670 + }, + { + "epoch": 3.76, + "learning_rate": 1.8739898070622497e-06, + "loss": 0.2373, + "step": 20671 + }, + { + "epoch": 3.76, + "learning_rate": 1.8739352020385876e-06, + "loss": 1.123, + "step": 20672 + }, + { + "epoch": 3.76, + "learning_rate": 1.8738805970149256e-06, + "loss": 0.3307, + "step": 20673 + }, + { + "epoch": 3.76, + "learning_rate": 1.873825991991263e-06, + "loss": 0.0209, + "step": 20674 + }, + { + "epoch": 3.76, + "learning_rate": 1.873771386967601e-06, + "loss": 0.0161, + "step": 20675 + }, + { + "epoch": 3.76, + "learning_rate": 1.873716781943939e-06, + "loss": 0.2747, + "step": 20676 + }, + { + "epoch": 3.76, + "learning_rate": 1.8736621769202767e-06, + "loss": 0.4172, + "step": 20677 + }, + { + "epoch": 3.76, + "learning_rate": 1.8736075718966146e-06, + "loss": 1.0752, + "step": 20678 + }, + { + "epoch": 3.76, + "learning_rate": 1.8735529668729526e-06, + "loss": 0.1933, + "step": 20679 + }, + { + "epoch": 3.76, + "learning_rate": 1.87349836184929e-06, + "loss": 0.0423, + "step": 20680 + }, + { + "epoch": 3.76, + "learning_rate": 1.873443756825628e-06, + "loss": 3.1088, + "step": 20681 + }, + { + "epoch": 3.76, + "learning_rate": 1.8733891518019658e-06, + "loss": 0.3954, + "step": 20682 + }, + { + "epoch": 3.76, + "learning_rate": 1.8733345467783037e-06, + "loss": 1.7656, + "step": 20683 + }, + { + "epoch": 3.76, + "learning_rate": 1.8732799417546417e-06, + "loss": 2.8744, + "step": 20684 + }, + { + "epoch": 3.76, + "learning_rate": 1.8732253367309792e-06, + "loss": 0.4202, + "step": 20685 + }, + { + "epoch": 3.76, + "learning_rate": 1.8731707317073171e-06, + "loss": 0.7486, + "step": 20686 + }, + { + "epoch": 3.76, + "learning_rate": 1.873116126683655e-06, + "loss": 1.3934, + "step": 20687 + }, + { + "epoch": 3.76, + "learning_rate": 1.8730615216599926e-06, + "loss": 0.4943, + "step": 20688 + }, + { + "epoch": 3.76, + "learning_rate": 1.8730069166363305e-06, + "loss": 0.6828, + "step": 20689 + }, + { + "epoch": 3.76, + "learning_rate": 1.8729523116126685e-06, + "loss": 0.8218, + "step": 20690 + }, + { + "epoch": 3.76, + "learning_rate": 1.8728977065890062e-06, + "loss": 0.3351, + "step": 20691 + }, + { + "epoch": 3.76, + "learning_rate": 1.8728431015653442e-06, + "loss": 0.1373, + "step": 20692 + }, + { + "epoch": 3.76, + "learning_rate": 1.8727884965416817e-06, + "loss": 1.204, + "step": 20693 + }, + { + "epoch": 3.76, + "learning_rate": 1.8727338915180196e-06, + "loss": 0.2943, + "step": 20694 + }, + { + "epoch": 3.76, + "learning_rate": 1.8726792864943576e-06, + "loss": 0.3823, + "step": 20695 + }, + { + "epoch": 3.76, + "learning_rate": 1.8726246814706953e-06, + "loss": 0.4125, + "step": 20696 + }, + { + "epoch": 3.76, + "learning_rate": 1.8725700764470332e-06, + "loss": 1.0124, + "step": 20697 + }, + { + "epoch": 3.76, + "learning_rate": 1.8725154714233712e-06, + "loss": 0.0073, + "step": 20698 + }, + { + "epoch": 3.76, + "learning_rate": 1.8724608663997087e-06, + "loss": 3.8195, + "step": 20699 + }, + { + "epoch": 3.76, + "learning_rate": 1.8724062613760466e-06, + "loss": 0.2025, + "step": 20700 + }, + { + "epoch": 3.76, + "learning_rate": 1.8723516563523846e-06, + "loss": 0.1956, + "step": 20701 + }, + { + "epoch": 3.76, + "learning_rate": 1.872297051328722e-06, + "loss": 1.4811, + "step": 20702 + }, + { + "epoch": 3.76, + "learning_rate": 1.87224244630506e-06, + "loss": 0.7464, + "step": 20703 + }, + { + "epoch": 3.77, + "learning_rate": 1.872187841281398e-06, + "loss": 0.0222, + "step": 20704 + }, + { + "epoch": 3.77, + "learning_rate": 1.8721332362577357e-06, + "loss": 0.2813, + "step": 20705 + }, + { + "epoch": 3.77, + "learning_rate": 1.8720786312340737e-06, + "loss": 0.5431, + "step": 20706 + }, + { + "epoch": 3.77, + "learning_rate": 1.8720240262104112e-06, + "loss": 1.9579, + "step": 20707 + }, + { + "epoch": 3.77, + "learning_rate": 1.8719694211867491e-06, + "loss": 0.35, + "step": 20708 + }, + { + "epoch": 3.77, + "learning_rate": 1.871914816163087e-06, + "loss": 2.1247, + "step": 20709 + }, + { + "epoch": 3.77, + "learning_rate": 1.8718602111394248e-06, + "loss": 1.154, + "step": 20710 + }, + { + "epoch": 3.77, + "learning_rate": 1.8718056061157627e-06, + "loss": 1.6235, + "step": 20711 + }, + { + "epoch": 3.77, + "learning_rate": 1.8717510010921007e-06, + "loss": 0.5349, + "step": 20712 + }, + { + "epoch": 3.77, + "learning_rate": 1.8716963960684382e-06, + "loss": 3.1529, + "step": 20713 + }, + { + "epoch": 3.77, + "learning_rate": 1.8716417910447762e-06, + "loss": 0.0287, + "step": 20714 + }, + { + "epoch": 3.77, + "learning_rate": 1.871587186021114e-06, + "loss": 0.7142, + "step": 20715 + }, + { + "epoch": 3.77, + "learning_rate": 1.8715325809974518e-06, + "loss": 1.4182, + "step": 20716 + }, + { + "epoch": 3.77, + "learning_rate": 1.8714779759737898e-06, + "loss": 0.5525, + "step": 20717 + }, + { + "epoch": 3.77, + "learning_rate": 1.8714233709501275e-06, + "loss": 0.0147, + "step": 20718 + }, + { + "epoch": 3.77, + "learning_rate": 1.8713687659264652e-06, + "loss": 0.3785, + "step": 20719 + }, + { + "epoch": 3.77, + "learning_rate": 1.8713141609028032e-06, + "loss": 0.425, + "step": 20720 + }, + { + "epoch": 3.77, + "learning_rate": 1.8712595558791407e-06, + "loss": 0.8617, + "step": 20721 + }, + { + "epoch": 3.77, + "learning_rate": 1.8712049508554786e-06, + "loss": 2.3035, + "step": 20722 + }, + { + "epoch": 3.77, + "learning_rate": 1.8711503458318166e-06, + "loss": 1.6777, + "step": 20723 + }, + { + "epoch": 3.77, + "learning_rate": 1.8710957408081543e-06, + "loss": 0.0428, + "step": 20724 + }, + { + "epoch": 3.77, + "learning_rate": 1.8710411357844923e-06, + "loss": 0.8553, + "step": 20725 + }, + { + "epoch": 3.77, + "learning_rate": 1.8709865307608302e-06, + "loss": 0.3193, + "step": 20726 + }, + { + "epoch": 3.77, + "learning_rate": 1.8709319257371677e-06, + "loss": 1.677, + "step": 20727 + }, + { + "epoch": 3.77, + "learning_rate": 1.8708773207135057e-06, + "loss": 0.2405, + "step": 20728 + }, + { + "epoch": 3.77, + "learning_rate": 1.8708227156898436e-06, + "loss": 0.8643, + "step": 20729 + }, + { + "epoch": 3.77, + "learning_rate": 1.8707681106661813e-06, + "loss": 1.1789, + "step": 20730 + }, + { + "epoch": 3.77, + "learning_rate": 1.8707135056425193e-06, + "loss": 0.1174, + "step": 20731 + }, + { + "epoch": 3.77, + "learning_rate": 1.8706589006188572e-06, + "loss": 0.6544, + "step": 20732 + }, + { + "epoch": 3.77, + "learning_rate": 1.8706042955951947e-06, + "loss": 0.7876, + "step": 20733 + }, + { + "epoch": 3.77, + "learning_rate": 1.8705496905715327e-06, + "loss": 2.6474, + "step": 20734 + }, + { + "epoch": 3.77, + "learning_rate": 1.8704950855478702e-06, + "loss": 0.0732, + "step": 20735 + }, + { + "epoch": 3.77, + "learning_rate": 1.8704404805242082e-06, + "loss": 0.7272, + "step": 20736 + }, + { + "epoch": 3.77, + "learning_rate": 1.870385875500546e-06, + "loss": 0.9973, + "step": 20737 + }, + { + "epoch": 3.77, + "learning_rate": 1.8703312704768838e-06, + "loss": 0.0077, + "step": 20738 + }, + { + "epoch": 3.77, + "learning_rate": 1.8702766654532218e-06, + "loss": 0.0103, + "step": 20739 + }, + { + "epoch": 3.77, + "learning_rate": 1.8702220604295597e-06, + "loss": 1.7012, + "step": 20740 + }, + { + "epoch": 3.77, + "learning_rate": 1.8701674554058972e-06, + "loss": 1.6603, + "step": 20741 + }, + { + "epoch": 3.77, + "learning_rate": 1.8701128503822352e-06, + "loss": 0.2301, + "step": 20742 + }, + { + "epoch": 3.77, + "learning_rate": 1.8700582453585731e-06, + "loss": 1.1297, + "step": 20743 + }, + { + "epoch": 3.77, + "learning_rate": 1.8700036403349109e-06, + "loss": 2.1292, + "step": 20744 + }, + { + "epoch": 3.77, + "learning_rate": 1.8699490353112488e-06, + "loss": 0.0628, + "step": 20745 + }, + { + "epoch": 3.77, + "learning_rate": 1.8698944302875867e-06, + "loss": 0.7368, + "step": 20746 + }, + { + "epoch": 3.77, + "learning_rate": 1.8698398252639243e-06, + "loss": 0.901, + "step": 20747 + }, + { + "epoch": 3.77, + "learning_rate": 1.8697852202402622e-06, + "loss": 0.0179, + "step": 20748 + }, + { + "epoch": 3.77, + "learning_rate": 1.8697306152165997e-06, + "loss": 0.6638, + "step": 20749 + }, + { + "epoch": 3.77, + "learning_rate": 1.8696760101929377e-06, + "loss": 0.0104, + "step": 20750 + }, + { + "epoch": 3.77, + "learning_rate": 1.8696214051692756e-06, + "loss": 0.0571, + "step": 20751 + }, + { + "epoch": 3.77, + "learning_rate": 1.8695668001456133e-06, + "loss": 1.0205, + "step": 20752 + }, + { + "epoch": 3.77, + "learning_rate": 1.8695121951219513e-06, + "loss": 2.5921, + "step": 20753 + }, + { + "epoch": 3.77, + "learning_rate": 1.8694575900982892e-06, + "loss": 0.1691, + "step": 20754 + }, + { + "epoch": 3.77, + "learning_rate": 1.8694029850746268e-06, + "loss": 0.0285, + "step": 20755 + }, + { + "epoch": 3.77, + "learning_rate": 1.8693483800509647e-06, + "loss": 0.8707, + "step": 20756 + }, + { + "epoch": 3.77, + "learning_rate": 1.8692937750273026e-06, + "loss": 0.9164, + "step": 20757 + }, + { + "epoch": 3.77, + "learning_rate": 1.8692391700036404e-06, + "loss": 0.3638, + "step": 20758 + }, + { + "epoch": 3.78, + "learning_rate": 1.8691845649799783e-06, + "loss": 0.3852, + "step": 20759 + }, + { + "epoch": 3.78, + "learning_rate": 1.8691299599563163e-06, + "loss": 0.0601, + "step": 20760 + }, + { + "epoch": 3.78, + "learning_rate": 1.8690753549326538e-06, + "loss": 0.299, + "step": 20761 + }, + { + "epoch": 3.78, + "learning_rate": 1.8690207499089917e-06, + "loss": 2.0313, + "step": 20762 + }, + { + "epoch": 3.78, + "learning_rate": 1.8689661448853295e-06, + "loss": 1.0235, + "step": 20763 + }, + { + "epoch": 3.78, + "learning_rate": 1.8689115398616674e-06, + "loss": 1.4505, + "step": 20764 + }, + { + "epoch": 3.78, + "learning_rate": 1.8688569348380053e-06, + "loss": 1.9331, + "step": 20765 + }, + { + "epoch": 3.78, + "learning_rate": 1.8688023298143429e-06, + "loss": 0.0627, + "step": 20766 + }, + { + "epoch": 3.78, + "learning_rate": 1.8687477247906808e-06, + "loss": 0.0101, + "step": 20767 + }, + { + "epoch": 3.78, + "learning_rate": 1.8686931197670187e-06, + "loss": 0.5097, + "step": 20768 + }, + { + "epoch": 3.78, + "learning_rate": 1.8686385147433563e-06, + "loss": 1.3501, + "step": 20769 + }, + { + "epoch": 3.78, + "learning_rate": 1.8685839097196942e-06, + "loss": 0.0519, + "step": 20770 + }, + { + "epoch": 3.78, + "learning_rate": 1.8685293046960322e-06, + "loss": 1.1191, + "step": 20771 + }, + { + "epoch": 3.78, + "learning_rate": 1.8684746996723699e-06, + "loss": 2.1152, + "step": 20772 + }, + { + "epoch": 3.78, + "learning_rate": 1.8684200946487078e-06, + "loss": 0.2789, + "step": 20773 + }, + { + "epoch": 3.78, + "learning_rate": 1.8683654896250453e-06, + "loss": 0.2553, + "step": 20774 + }, + { + "epoch": 3.78, + "learning_rate": 1.8683108846013833e-06, + "loss": 0.5031, + "step": 20775 + }, + { + "epoch": 3.78, + "learning_rate": 1.8682562795777212e-06, + "loss": 0.1194, + "step": 20776 + }, + { + "epoch": 3.78, + "learning_rate": 1.868201674554059e-06, + "loss": 0.445, + "step": 20777 + }, + { + "epoch": 3.78, + "learning_rate": 1.868147069530397e-06, + "loss": 0.5972, + "step": 20778 + }, + { + "epoch": 3.78, + "learning_rate": 1.8680924645067349e-06, + "loss": 0.0241, + "step": 20779 + }, + { + "epoch": 3.78, + "learning_rate": 1.8680378594830724e-06, + "loss": 0.8558, + "step": 20780 + }, + { + "epoch": 3.78, + "learning_rate": 1.8679832544594103e-06, + "loss": 0.9338, + "step": 20781 + }, + { + "epoch": 3.78, + "learning_rate": 1.8679286494357483e-06, + "loss": 1.1491, + "step": 20782 + }, + { + "epoch": 3.78, + "learning_rate": 1.8678740444120858e-06, + "loss": 0.088, + "step": 20783 + }, + { + "epoch": 3.78, + "learning_rate": 1.8678194393884237e-06, + "loss": 0.7631, + "step": 20784 + }, + { + "epoch": 3.78, + "learning_rate": 1.8677648343647617e-06, + "loss": 0.5296, + "step": 20785 + }, + { + "epoch": 3.78, + "learning_rate": 1.8677102293410994e-06, + "loss": 0.0932, + "step": 20786 + }, + { + "epoch": 3.78, + "learning_rate": 1.8676556243174373e-06, + "loss": 0.3187, + "step": 20787 + }, + { + "epoch": 3.78, + "learning_rate": 1.8676010192937749e-06, + "loss": 0.5349, + "step": 20788 + }, + { + "epoch": 3.78, + "learning_rate": 1.8675464142701128e-06, + "loss": 2.26, + "step": 20789 + }, + { + "epoch": 3.78, + "learning_rate": 1.8674918092464507e-06, + "loss": 0.657, + "step": 20790 + }, + { + "epoch": 3.78, + "learning_rate": 1.8674372042227885e-06, + "loss": 0.0276, + "step": 20791 + }, + { + "epoch": 3.78, + "learning_rate": 1.8673825991991264e-06, + "loss": 3.5136, + "step": 20792 + }, + { + "epoch": 3.78, + "learning_rate": 1.8673279941754644e-06, + "loss": 1.6894, + "step": 20793 + }, + { + "epoch": 3.78, + "learning_rate": 1.8672733891518019e-06, + "loss": 1.1894, + "step": 20794 + }, + { + "epoch": 3.78, + "learning_rate": 1.8672187841281398e-06, + "loss": 0.7389, + "step": 20795 + }, + { + "epoch": 3.78, + "learning_rate": 1.8671641791044778e-06, + "loss": 0.2998, + "step": 20796 + }, + { + "epoch": 3.78, + "learning_rate": 1.8671095740808153e-06, + "loss": 2.1917, + "step": 20797 + }, + { + "epoch": 3.78, + "learning_rate": 1.8670549690571532e-06, + "loss": 1.1413, + "step": 20798 + }, + { + "epoch": 3.78, + "learning_rate": 1.8670003640334912e-06, + "loss": 0.2817, + "step": 20799 + }, + { + "epoch": 3.78, + "learning_rate": 1.866945759009829e-06, + "loss": 2.9374, + "step": 20800 + }, + { + "epoch": 3.78, + "learning_rate": 1.8668911539861669e-06, + "loss": 0.2113, + "step": 20801 + }, + { + "epoch": 3.78, + "learning_rate": 1.8668365489625044e-06, + "loss": 0.121, + "step": 20802 + }, + { + "epoch": 3.78, + "learning_rate": 1.8667819439388423e-06, + "loss": 0.6925, + "step": 20803 + }, + { + "epoch": 3.78, + "learning_rate": 1.8667273389151803e-06, + "loss": 0.0162, + "step": 20804 + }, + { + "epoch": 3.78, + "learning_rate": 1.866672733891518e-06, + "loss": 0.1626, + "step": 20805 + }, + { + "epoch": 3.78, + "learning_rate": 1.866618128867856e-06, + "loss": 0.3539, + "step": 20806 + }, + { + "epoch": 3.78, + "learning_rate": 1.8665635238441939e-06, + "loss": 0.6052, + "step": 20807 + }, + { + "epoch": 3.78, + "learning_rate": 1.8665089188205314e-06, + "loss": 1.7411, + "step": 20808 + }, + { + "epoch": 3.78, + "learning_rate": 1.8664543137968693e-06, + "loss": 0.1205, + "step": 20809 + }, + { + "epoch": 3.78, + "learning_rate": 1.8663997087732073e-06, + "loss": 0.3815, + "step": 20810 + }, + { + "epoch": 3.78, + "learning_rate": 1.866345103749545e-06, + "loss": 0.8993, + "step": 20811 + }, + { + "epoch": 3.78, + "learning_rate": 1.866290498725883e-06, + "loss": 3.8718, + "step": 20812 + }, + { + "epoch": 3.78, + "learning_rate": 1.866235893702221e-06, + "loss": 1.2105, + "step": 20813 + }, + { + "epoch": 3.79, + "learning_rate": 1.8661812886785584e-06, + "loss": 0.8618, + "step": 20814 + }, + { + "epoch": 3.79, + "learning_rate": 1.8661266836548964e-06, + "loss": 0.7232, + "step": 20815 + }, + { + "epoch": 3.79, + "learning_rate": 1.8660720786312339e-06, + "loss": 0.067, + "step": 20816 + }, + { + "epoch": 3.79, + "learning_rate": 1.8660174736075718e-06, + "loss": 0.2175, + "step": 20817 + }, + { + "epoch": 3.79, + "learning_rate": 1.8659628685839098e-06, + "loss": 0.9324, + "step": 20818 + }, + { + "epoch": 3.79, + "learning_rate": 1.8659082635602475e-06, + "loss": 0.2871, + "step": 20819 + }, + { + "epoch": 3.79, + "learning_rate": 1.8658536585365854e-06, + "loss": 1.9023, + "step": 20820 + }, + { + "epoch": 3.79, + "learning_rate": 1.8657990535129234e-06, + "loss": 0.4045, + "step": 20821 + }, + { + "epoch": 3.79, + "learning_rate": 1.865744448489261e-06, + "loss": 1.3655, + "step": 20822 + }, + { + "epoch": 3.79, + "learning_rate": 1.8656898434655989e-06, + "loss": 1.0373, + "step": 20823 + }, + { + "epoch": 3.79, + "learning_rate": 1.8656352384419368e-06, + "loss": 0.3877, + "step": 20824 + }, + { + "epoch": 3.79, + "learning_rate": 1.8655806334182745e-06, + "loss": 1.7198, + "step": 20825 + }, + { + "epoch": 3.79, + "learning_rate": 1.8655260283946125e-06, + "loss": 0.2877, + "step": 20826 + }, + { + "epoch": 3.79, + "learning_rate": 1.8654714233709504e-06, + "loss": 0.7116, + "step": 20827 + }, + { + "epoch": 3.79, + "learning_rate": 1.865416818347288e-06, + "loss": 0.8366, + "step": 20828 + }, + { + "epoch": 3.79, + "learning_rate": 1.8653622133236259e-06, + "loss": 0.5176, + "step": 20829 + }, + { + "epoch": 3.79, + "learning_rate": 1.8653076082999634e-06, + "loss": 0.1416, + "step": 20830 + }, + { + "epoch": 3.79, + "learning_rate": 1.8652530032763013e-06, + "loss": 0.388, + "step": 20831 + }, + { + "epoch": 3.79, + "learning_rate": 1.8651983982526393e-06, + "loss": 1.8271, + "step": 20832 + }, + { + "epoch": 3.79, + "learning_rate": 1.865143793228977e-06, + "loss": 0.8331, + "step": 20833 + }, + { + "epoch": 3.79, + "learning_rate": 1.865089188205315e-06, + "loss": 0.204, + "step": 20834 + }, + { + "epoch": 3.79, + "learning_rate": 1.865034583181653e-06, + "loss": 0.0219, + "step": 20835 + }, + { + "epoch": 3.79, + "learning_rate": 1.8649799781579904e-06, + "loss": 0.4602, + "step": 20836 + }, + { + "epoch": 3.79, + "learning_rate": 1.8649253731343284e-06, + "loss": 0.4523, + "step": 20837 + }, + { + "epoch": 3.79, + "learning_rate": 1.8648707681106663e-06, + "loss": 0.0413, + "step": 20838 + }, + { + "epoch": 3.79, + "learning_rate": 1.864816163087004e-06, + "loss": 1.6437, + "step": 20839 + }, + { + "epoch": 3.79, + "learning_rate": 1.864761558063342e-06, + "loss": 0.6629, + "step": 20840 + }, + { + "epoch": 3.79, + "learning_rate": 1.86470695303968e-06, + "loss": 1.4937, + "step": 20841 + }, + { + "epoch": 3.79, + "learning_rate": 1.8646523480160175e-06, + "loss": 0.2227, + "step": 20842 + }, + { + "epoch": 3.79, + "learning_rate": 1.8645977429923554e-06, + "loss": 0.4543, + "step": 20843 + }, + { + "epoch": 3.79, + "learning_rate": 1.8645431379686931e-06, + "loss": 0.1971, + "step": 20844 + }, + { + "epoch": 3.79, + "learning_rate": 1.864488532945031e-06, + "loss": 0.5212, + "step": 20845 + }, + { + "epoch": 3.79, + "learning_rate": 1.864433927921369e-06, + "loss": 0.4619, + "step": 20846 + }, + { + "epoch": 3.79, + "learning_rate": 1.8643793228977065e-06, + "loss": 0.7996, + "step": 20847 + }, + { + "epoch": 3.79, + "learning_rate": 1.8643247178740445e-06, + "loss": 0.0066, + "step": 20848 + }, + { + "epoch": 3.79, + "learning_rate": 1.8642701128503824e-06, + "loss": 0.0145, + "step": 20849 + }, + { + "epoch": 3.79, + "learning_rate": 1.86421550782672e-06, + "loss": 1.6371, + "step": 20850 + }, + { + "epoch": 3.79, + "learning_rate": 1.8641609028030579e-06, + "loss": 0.5949, + "step": 20851 + }, + { + "epoch": 3.79, + "learning_rate": 1.8641062977793958e-06, + "loss": 0.0178, + "step": 20852 + }, + { + "epoch": 3.79, + "learning_rate": 1.8640516927557336e-06, + "loss": 0.6172, + "step": 20853 + }, + { + "epoch": 3.79, + "learning_rate": 1.8639970877320715e-06, + "loss": 0.7054, + "step": 20854 + }, + { + "epoch": 3.79, + "learning_rate": 1.863942482708409e-06, + "loss": 0.0901, + "step": 20855 + }, + { + "epoch": 3.79, + "learning_rate": 1.863887877684747e-06, + "loss": 1.5523, + "step": 20856 + }, + { + "epoch": 3.79, + "learning_rate": 1.863833272661085e-06, + "loss": 0.1919, + "step": 20857 + }, + { + "epoch": 3.79, + "learning_rate": 1.8637786676374226e-06, + "loss": 0.5592, + "step": 20858 + }, + { + "epoch": 3.79, + "learning_rate": 1.8637240626137606e-06, + "loss": 1.7587, + "step": 20859 + }, + { + "epoch": 3.79, + "learning_rate": 1.8636694575900985e-06, + "loss": 0.4257, + "step": 20860 + }, + { + "epoch": 3.79, + "learning_rate": 1.863614852566436e-06, + "loss": 1.0036, + "step": 20861 + }, + { + "epoch": 3.79, + "learning_rate": 1.863560247542774e-06, + "loss": 1.5671, + "step": 20862 + }, + { + "epoch": 3.79, + "learning_rate": 1.863505642519112e-06, + "loss": 0.2116, + "step": 20863 + }, + { + "epoch": 3.79, + "learning_rate": 1.8634510374954495e-06, + "loss": 0.8653, + "step": 20864 + }, + { + "epoch": 3.79, + "learning_rate": 1.8633964324717874e-06, + "loss": 0.0178, + "step": 20865 + }, + { + "epoch": 3.79, + "learning_rate": 1.8633418274481253e-06, + "loss": 0.0479, + "step": 20866 + }, + { + "epoch": 3.79, + "learning_rate": 1.863287222424463e-06, + "loss": 0.0859, + "step": 20867 + }, + { + "epoch": 3.79, + "learning_rate": 1.863232617400801e-06, + "loss": 1.8526, + "step": 20868 + }, + { + "epoch": 3.8, + "learning_rate": 1.8631780123771385e-06, + "loss": 0.0438, + "step": 20869 + }, + { + "epoch": 3.8, + "learning_rate": 1.8631234073534765e-06, + "loss": 0.4783, + "step": 20870 + }, + { + "epoch": 3.8, + "learning_rate": 1.8630688023298144e-06, + "loss": 0.536, + "step": 20871 + }, + { + "epoch": 3.8, + "learning_rate": 1.8630141973061522e-06, + "loss": 0.1726, + "step": 20872 + }, + { + "epoch": 3.8, + "learning_rate": 1.86295959228249e-06, + "loss": 0.1141, + "step": 20873 + }, + { + "epoch": 3.8, + "learning_rate": 1.862904987258828e-06, + "loss": 0.4203, + "step": 20874 + }, + { + "epoch": 3.8, + "learning_rate": 1.8628503822351656e-06, + "loss": 0.0384, + "step": 20875 + }, + { + "epoch": 3.8, + "learning_rate": 1.8627957772115035e-06, + "loss": 0.7656, + "step": 20876 + }, + { + "epoch": 3.8, + "learning_rate": 1.8627411721878414e-06, + "loss": 3.2652, + "step": 20877 + }, + { + "epoch": 3.8, + "learning_rate": 1.862686567164179e-06, + "loss": 0.323, + "step": 20878 + }, + { + "epoch": 3.8, + "learning_rate": 1.862631962140517e-06, + "loss": 0.3766, + "step": 20879 + }, + { + "epoch": 3.8, + "learning_rate": 1.8625773571168549e-06, + "loss": 0.116, + "step": 20880 + }, + { + "epoch": 3.8, + "learning_rate": 1.8625227520931926e-06, + "loss": 1.1721, + "step": 20881 + }, + { + "epoch": 3.8, + "learning_rate": 1.8624681470695305e-06, + "loss": 0.0283, + "step": 20882 + }, + { + "epoch": 3.8, + "learning_rate": 1.862413542045868e-06, + "loss": 0.0712, + "step": 20883 + }, + { + "epoch": 3.8, + "learning_rate": 1.862358937022206e-06, + "loss": 0.5264, + "step": 20884 + }, + { + "epoch": 3.8, + "learning_rate": 1.862304331998544e-06, + "loss": 0.5378, + "step": 20885 + }, + { + "epoch": 3.8, + "learning_rate": 1.8622497269748817e-06, + "loss": 0.4332, + "step": 20886 + }, + { + "epoch": 3.8, + "learning_rate": 1.8621951219512196e-06, + "loss": 0.0149, + "step": 20887 + }, + { + "epoch": 3.8, + "learning_rate": 1.8621405169275576e-06, + "loss": 0.9852, + "step": 20888 + }, + { + "epoch": 3.8, + "learning_rate": 1.862085911903895e-06, + "loss": 0.2344, + "step": 20889 + }, + { + "epoch": 3.8, + "learning_rate": 1.862031306880233e-06, + "loss": 0.6504, + "step": 20890 + }, + { + "epoch": 3.8, + "learning_rate": 1.861976701856571e-06, + "loss": 0.3443, + "step": 20891 + }, + { + "epoch": 3.8, + "learning_rate": 1.8619220968329087e-06, + "loss": 1.634, + "step": 20892 + }, + { + "epoch": 3.8, + "learning_rate": 1.8618674918092466e-06, + "loss": 0.2382, + "step": 20893 + }, + { + "epoch": 3.8, + "learning_rate": 1.8618128867855846e-06, + "loss": 0.6822, + "step": 20894 + }, + { + "epoch": 3.8, + "learning_rate": 1.861758281761922e-06, + "loss": 0.2436, + "step": 20895 + }, + { + "epoch": 3.8, + "learning_rate": 1.86170367673826e-06, + "loss": 0.0325, + "step": 20896 + }, + { + "epoch": 3.8, + "learning_rate": 1.8616490717145976e-06, + "loss": 0.0303, + "step": 20897 + }, + { + "epoch": 3.8, + "learning_rate": 1.8615944666909355e-06, + "loss": 0.0127, + "step": 20898 + }, + { + "epoch": 3.8, + "learning_rate": 1.8615398616672734e-06, + "loss": 1.2353, + "step": 20899 + }, + { + "epoch": 3.8, + "learning_rate": 1.8614852566436112e-06, + "loss": 0.6919, + "step": 20900 + }, + { + "epoch": 3.8, + "learning_rate": 1.8614306516199491e-06, + "loss": 0.0143, + "step": 20901 + }, + { + "epoch": 3.8, + "learning_rate": 1.861376046596287e-06, + "loss": 0.3901, + "step": 20902 + }, + { + "epoch": 3.8, + "learning_rate": 1.8613214415726246e-06, + "loss": 0.7928, + "step": 20903 + }, + { + "epoch": 3.8, + "learning_rate": 1.8612668365489625e-06, + "loss": 1.0023, + "step": 20904 + }, + { + "epoch": 3.8, + "learning_rate": 1.8612122315253005e-06, + "loss": 0.3541, + "step": 20905 + }, + { + "epoch": 3.8, + "learning_rate": 1.8611576265016382e-06, + "loss": 0.2114, + "step": 20906 + }, + { + "epoch": 3.8, + "learning_rate": 1.8611030214779761e-06, + "loss": 0.1471, + "step": 20907 + }, + { + "epoch": 3.8, + "learning_rate": 1.861048416454314e-06, + "loss": 0.3503, + "step": 20908 + }, + { + "epoch": 3.8, + "learning_rate": 1.8609938114306516e-06, + "loss": 0.0443, + "step": 20909 + }, + { + "epoch": 3.8, + "learning_rate": 1.8609392064069896e-06, + "loss": 0.4578, + "step": 20910 + }, + { + "epoch": 3.8, + "learning_rate": 1.860884601383327e-06, + "loss": 0.0196, + "step": 20911 + }, + { + "epoch": 3.8, + "learning_rate": 1.860829996359665e-06, + "loss": 0.6877, + "step": 20912 + }, + { + "epoch": 3.8, + "learning_rate": 1.860775391336003e-06, + "loss": 0.9337, + "step": 20913 + }, + { + "epoch": 3.8, + "learning_rate": 1.8607207863123407e-06, + "loss": 1.3964, + "step": 20914 + }, + { + "epoch": 3.8, + "learning_rate": 1.8606661812886786e-06, + "loss": 1.3926, + "step": 20915 + }, + { + "epoch": 3.8, + "learning_rate": 1.8606115762650166e-06, + "loss": 0.4135, + "step": 20916 + }, + { + "epoch": 3.8, + "learning_rate": 1.860556971241354e-06, + "loss": 0.2004, + "step": 20917 + }, + { + "epoch": 3.8, + "learning_rate": 1.860502366217692e-06, + "loss": 0.7824, + "step": 20918 + }, + { + "epoch": 3.8, + "learning_rate": 1.86044776119403e-06, + "loss": 4.0413, + "step": 20919 + }, + { + "epoch": 3.8, + "learning_rate": 1.8603931561703677e-06, + "loss": 1.581, + "step": 20920 + }, + { + "epoch": 3.8, + "learning_rate": 1.8603385511467057e-06, + "loss": 0.2927, + "step": 20921 + }, + { + "epoch": 3.8, + "learning_rate": 1.8602839461230436e-06, + "loss": 1.9465, + "step": 20922 + }, + { + "epoch": 3.8, + "learning_rate": 1.8602293410993811e-06, + "loss": 0.1889, + "step": 20923 + }, + { + "epoch": 3.81, + "learning_rate": 1.860174736075719e-06, + "loss": 0.7573, + "step": 20924 + }, + { + "epoch": 3.81, + "learning_rate": 1.8601201310520566e-06, + "loss": 0.0315, + "step": 20925 + }, + { + "epoch": 3.81, + "learning_rate": 1.8600655260283945e-06, + "loss": 0.6196, + "step": 20926 + }, + { + "epoch": 3.81, + "learning_rate": 1.8600109210047325e-06, + "loss": 0.1973, + "step": 20927 + }, + { + "epoch": 3.81, + "learning_rate": 1.8599563159810702e-06, + "loss": 1.2969, + "step": 20928 + }, + { + "epoch": 3.81, + "learning_rate": 1.8599017109574082e-06, + "loss": 2.342, + "step": 20929 + }, + { + "epoch": 3.81, + "learning_rate": 1.859847105933746e-06, + "loss": 2.1892, + "step": 20930 + }, + { + "epoch": 3.81, + "learning_rate": 1.8597925009100836e-06, + "loss": 1.5786, + "step": 20931 + }, + { + "epoch": 3.81, + "learning_rate": 1.8597378958864216e-06, + "loss": 0.4862, + "step": 20932 + }, + { + "epoch": 3.81, + "learning_rate": 1.8596832908627595e-06, + "loss": 0.0237, + "step": 20933 + }, + { + "epoch": 3.81, + "learning_rate": 1.8596286858390972e-06, + "loss": 0.224, + "step": 20934 + }, + { + "epoch": 3.81, + "learning_rate": 1.8595740808154352e-06, + "loss": 0.7168, + "step": 20935 + }, + { + "epoch": 3.81, + "learning_rate": 1.8595194757917727e-06, + "loss": 0.933, + "step": 20936 + }, + { + "epoch": 3.81, + "learning_rate": 1.8594648707681106e-06, + "loss": 0.0093, + "step": 20937 + }, + { + "epoch": 3.81, + "learning_rate": 1.8594102657444486e-06, + "loss": 0.0086, + "step": 20938 + }, + { + "epoch": 3.81, + "learning_rate": 1.8593556607207863e-06, + "loss": 0.3326, + "step": 20939 + }, + { + "epoch": 3.81, + "learning_rate": 1.8593010556971243e-06, + "loss": 0.9489, + "step": 20940 + }, + { + "epoch": 3.81, + "learning_rate": 1.8592464506734622e-06, + "loss": 0.2534, + "step": 20941 + }, + { + "epoch": 3.81, + "learning_rate": 1.8591918456497997e-06, + "loss": 0.457, + "step": 20942 + }, + { + "epoch": 3.81, + "learning_rate": 1.8591372406261377e-06, + "loss": 0.096, + "step": 20943 + }, + { + "epoch": 3.81, + "learning_rate": 1.8590826356024756e-06, + "loss": 1.1395, + "step": 20944 + }, + { + "epoch": 3.81, + "learning_rate": 1.8590280305788131e-06, + "loss": 1.1799, + "step": 20945 + }, + { + "epoch": 3.81, + "learning_rate": 1.858973425555151e-06, + "loss": 0.1679, + "step": 20946 + }, + { + "epoch": 3.81, + "learning_rate": 1.858918820531489e-06, + "loss": 0.0633, + "step": 20947 + }, + { + "epoch": 3.81, + "learning_rate": 1.8588642155078267e-06, + "loss": 0.2143, + "step": 20948 + }, + { + "epoch": 3.81, + "learning_rate": 1.8588096104841647e-06, + "loss": 0.028, + "step": 20949 + }, + { + "epoch": 3.81, + "learning_rate": 1.8587550054605022e-06, + "loss": 1.5142, + "step": 20950 + }, + { + "epoch": 3.81, + "learning_rate": 1.8587004004368402e-06, + "loss": 1.159, + "step": 20951 + }, + { + "epoch": 3.81, + "learning_rate": 1.858645795413178e-06, + "loss": 0.4255, + "step": 20952 + }, + { + "epoch": 3.81, + "learning_rate": 1.8585911903895158e-06, + "loss": 0.5621, + "step": 20953 + }, + { + "epoch": 3.81, + "learning_rate": 1.8585365853658538e-06, + "loss": 0.2578, + "step": 20954 + }, + { + "epoch": 3.81, + "learning_rate": 1.8584819803421917e-06, + "loss": 0.3037, + "step": 20955 + }, + { + "epoch": 3.81, + "learning_rate": 1.8584273753185292e-06, + "loss": 0.4745, + "step": 20956 + }, + { + "epoch": 3.81, + "learning_rate": 1.8583727702948672e-06, + "loss": 0.3143, + "step": 20957 + }, + { + "epoch": 3.81, + "learning_rate": 1.8583181652712051e-06, + "loss": 0.2762, + "step": 20958 + }, + { + "epoch": 3.81, + "learning_rate": 1.8582635602475426e-06, + "loss": 0.3664, + "step": 20959 + }, + { + "epoch": 3.81, + "learning_rate": 1.8582089552238806e-06, + "loss": 0.608, + "step": 20960 + }, + { + "epoch": 3.81, + "learning_rate": 1.8581543502002185e-06, + "loss": 0.0643, + "step": 20961 + }, + { + "epoch": 3.81, + "learning_rate": 1.8580997451765563e-06, + "loss": 1.1141, + "step": 20962 + }, + { + "epoch": 3.81, + "learning_rate": 1.8580451401528942e-06, + "loss": 0.0428, + "step": 20963 + }, + { + "epoch": 3.81, + "learning_rate": 1.8579905351292317e-06, + "loss": 0.046, + "step": 20964 + }, + { + "epoch": 3.81, + "learning_rate": 1.8579359301055697e-06, + "loss": 1.0488, + "step": 20965 + }, + { + "epoch": 3.81, + "learning_rate": 1.8578813250819076e-06, + "loss": 0.1579, + "step": 20966 + }, + { + "epoch": 3.81, + "learning_rate": 1.8578267200582453e-06, + "loss": 0.3494, + "step": 20967 + }, + { + "epoch": 3.81, + "learning_rate": 1.8577721150345833e-06, + "loss": 0.104, + "step": 20968 + }, + { + "epoch": 3.81, + "learning_rate": 1.8577175100109212e-06, + "loss": 1.1231, + "step": 20969 + }, + { + "epoch": 3.81, + "learning_rate": 1.8576629049872587e-06, + "loss": 0.4004, + "step": 20970 + }, + { + "epoch": 3.81, + "learning_rate": 1.8576082999635967e-06, + "loss": 0.321, + "step": 20971 + }, + { + "epoch": 3.81, + "learning_rate": 1.8575536949399346e-06, + "loss": 0.9394, + "step": 20972 + }, + { + "epoch": 3.81, + "learning_rate": 1.8574990899162724e-06, + "loss": 0.0604, + "step": 20973 + }, + { + "epoch": 3.81, + "learning_rate": 1.8574444848926103e-06, + "loss": 0.0103, + "step": 20974 + }, + { + "epoch": 3.81, + "learning_rate": 1.8573898798689483e-06, + "loss": 0.2506, + "step": 20975 + }, + { + "epoch": 3.81, + "learning_rate": 1.8573352748452858e-06, + "loss": 0.0488, + "step": 20976 + }, + { + "epoch": 3.81, + "learning_rate": 1.8572806698216237e-06, + "loss": 0.2754, + "step": 20977 + }, + { + "epoch": 3.81, + "learning_rate": 1.8572260647979612e-06, + "loss": 0.0453, + "step": 20978 + }, + { + "epoch": 3.82, + "learning_rate": 1.8571714597742992e-06, + "loss": 0.4826, + "step": 20979 + }, + { + "epoch": 3.82, + "learning_rate": 1.8571168547506371e-06, + "loss": 1.8389, + "step": 20980 + }, + { + "epoch": 3.82, + "learning_rate": 1.8570622497269749e-06, + "loss": 1.0325, + "step": 20981 + }, + { + "epoch": 3.82, + "learning_rate": 1.8570076447033128e-06, + "loss": 0.9471, + "step": 20982 + }, + { + "epoch": 3.82, + "learning_rate": 1.8569530396796507e-06, + "loss": 1.3032, + "step": 20983 + }, + { + "epoch": 3.82, + "learning_rate": 1.8568984346559883e-06, + "loss": 0.3566, + "step": 20984 + }, + { + "epoch": 3.82, + "learning_rate": 1.8568438296323262e-06, + "loss": 0.2348, + "step": 20985 + }, + { + "epoch": 3.82, + "learning_rate": 1.8567892246086641e-06, + "loss": 2.5231, + "step": 20986 + }, + { + "epoch": 3.82, + "learning_rate": 1.8567346195850019e-06, + "loss": 0.9753, + "step": 20987 + }, + { + "epoch": 3.82, + "learning_rate": 1.8566800145613398e-06, + "loss": 0.3939, + "step": 20988 + }, + { + "epoch": 3.82, + "learning_rate": 1.8566254095376778e-06, + "loss": 0.5521, + "step": 20989 + }, + { + "epoch": 3.82, + "learning_rate": 1.8565708045140153e-06, + "loss": 0.0449, + "step": 20990 + }, + { + "epoch": 3.82, + "learning_rate": 1.8565161994903532e-06, + "loss": 0.7423, + "step": 20991 + }, + { + "epoch": 3.82, + "learning_rate": 1.8564615944666907e-06, + "loss": 0.1875, + "step": 20992 + }, + { + "epoch": 3.82, + "learning_rate": 1.8564069894430287e-06, + "loss": 2.0286, + "step": 20993 + }, + { + "epoch": 3.82, + "learning_rate": 1.8563523844193666e-06, + "loss": 0.0169, + "step": 20994 + }, + { + "epoch": 3.82, + "learning_rate": 1.8562977793957044e-06, + "loss": 0.0278, + "step": 20995 + }, + { + "epoch": 3.82, + "learning_rate": 1.8562431743720423e-06, + "loss": 0.8312, + "step": 20996 + }, + { + "epoch": 3.82, + "learning_rate": 1.8561885693483803e-06, + "loss": 1.9845, + "step": 20997 + }, + { + "epoch": 3.82, + "learning_rate": 1.8561339643247178e-06, + "loss": 1.8971, + "step": 20998 + }, + { + "epoch": 3.82, + "learning_rate": 1.8560793593010557e-06, + "loss": 0.3462, + "step": 20999 + }, + { + "epoch": 3.82, + "learning_rate": 1.8560247542773937e-06, + "loss": 0.0316, + "step": 21000 + }, + { + "epoch": 3.82, + "learning_rate": 1.8559701492537314e-06, + "loss": 0.038, + "step": 21001 + }, + { + "epoch": 3.82, + "learning_rate": 1.8559155442300693e-06, + "loss": 0.5791, + "step": 21002 + }, + { + "epoch": 3.82, + "learning_rate": 1.8558609392064069e-06, + "loss": 0.0359, + "step": 21003 + }, + { + "epoch": 3.82, + "learning_rate": 1.8558063341827448e-06, + "loss": 0.47, + "step": 21004 + }, + { + "epoch": 3.82, + "learning_rate": 1.8557517291590827e-06, + "loss": 0.1709, + "step": 21005 + }, + { + "epoch": 3.82, + "learning_rate": 1.8556971241354203e-06, + "loss": 0.4169, + "step": 21006 + }, + { + "epoch": 3.82, + "learning_rate": 1.8556425191117582e-06, + "loss": 0.0074, + "step": 21007 + }, + { + "epoch": 3.82, + "learning_rate": 1.8555879140880961e-06, + "loss": 0.9777, + "step": 21008 + }, + { + "epoch": 3.82, + "learning_rate": 1.8555333090644339e-06, + "loss": 0.4614, + "step": 21009 + }, + { + "epoch": 3.82, + "learning_rate": 1.8554787040407718e-06, + "loss": 0.1865, + "step": 21010 + }, + { + "epoch": 3.82, + "learning_rate": 1.8554240990171098e-06, + "loss": 0.2574, + "step": 21011 + }, + { + "epoch": 3.82, + "learning_rate": 1.8553694939934473e-06, + "loss": 0.3437, + "step": 21012 + }, + { + "epoch": 3.82, + "learning_rate": 1.8553148889697852e-06, + "loss": 1.4986, + "step": 21013 + }, + { + "epoch": 3.82, + "learning_rate": 1.8552602839461232e-06, + "loss": 0.1076, + "step": 21014 + }, + { + "epoch": 3.82, + "learning_rate": 1.855205678922461e-06, + "loss": 1.5452, + "step": 21015 + }, + { + "epoch": 3.82, + "learning_rate": 1.8551510738987988e-06, + "loss": 4.1913, + "step": 21016 + }, + { + "epoch": 3.82, + "learning_rate": 1.8550964688751364e-06, + "loss": 1.3562, + "step": 21017 + }, + { + "epoch": 3.82, + "learning_rate": 1.8550418638514743e-06, + "loss": 0.2986, + "step": 21018 + }, + { + "epoch": 3.82, + "learning_rate": 1.8549872588278123e-06, + "loss": 0.3413, + "step": 21019 + }, + { + "epoch": 3.82, + "learning_rate": 1.85493265380415e-06, + "loss": 0.0305, + "step": 21020 + }, + { + "epoch": 3.82, + "learning_rate": 1.854878048780488e-06, + "loss": 0.0176, + "step": 21021 + }, + { + "epoch": 3.82, + "learning_rate": 1.8548234437568259e-06, + "loss": 0.3598, + "step": 21022 + }, + { + "epoch": 3.82, + "learning_rate": 1.8547688387331634e-06, + "loss": 1.1784, + "step": 21023 + }, + { + "epoch": 3.82, + "learning_rate": 1.8547142337095013e-06, + "loss": 0.4332, + "step": 21024 + }, + { + "epoch": 3.82, + "learning_rate": 1.8546596286858393e-06, + "loss": 0.5497, + "step": 21025 + }, + { + "epoch": 3.82, + "learning_rate": 1.8546050236621768e-06, + "loss": 1.0559, + "step": 21026 + }, + { + "epoch": 3.82, + "learning_rate": 1.8545504186385147e-06, + "loss": 1.1678, + "step": 21027 + }, + { + "epoch": 3.82, + "learning_rate": 1.8544958136148527e-06, + "loss": 0.8634, + "step": 21028 + }, + { + "epoch": 3.82, + "learning_rate": 1.8544412085911904e-06, + "loss": 0.5438, + "step": 21029 + }, + { + "epoch": 3.82, + "learning_rate": 1.8543866035675284e-06, + "loss": 0.1442, + "step": 21030 + }, + { + "epoch": 3.82, + "learning_rate": 1.8543319985438659e-06, + "loss": 3.248, + "step": 21031 + }, + { + "epoch": 3.82, + "learning_rate": 1.8542773935202038e-06, + "loss": 0.5694, + "step": 21032 + }, + { + "epoch": 3.82, + "learning_rate": 1.8542227884965418e-06, + "loss": 0.0996, + "step": 21033 + }, + { + "epoch": 3.83, + "learning_rate": 1.8541681834728795e-06, + "loss": 0.2188, + "step": 21034 + }, + { + "epoch": 3.83, + "learning_rate": 1.8541135784492174e-06, + "loss": 2.3978, + "step": 21035 + }, + { + "epoch": 3.83, + "learning_rate": 1.8540589734255554e-06, + "loss": 0.0512, + "step": 21036 + }, + { + "epoch": 3.83, + "learning_rate": 1.854004368401893e-06, + "loss": 0.1589, + "step": 21037 + }, + { + "epoch": 3.83, + "learning_rate": 1.8539497633782309e-06, + "loss": 1.3984, + "step": 21038 + }, + { + "epoch": 3.83, + "learning_rate": 1.8538951583545688e-06, + "loss": 0.0326, + "step": 21039 + }, + { + "epoch": 3.83, + "learning_rate": 1.8538405533309063e-06, + "loss": 0.6933, + "step": 21040 + }, + { + "epoch": 3.83, + "learning_rate": 1.8537859483072443e-06, + "loss": 0.1576, + "step": 21041 + }, + { + "epoch": 3.83, + "learning_rate": 1.8537313432835822e-06, + "loss": 0.1067, + "step": 21042 + }, + { + "epoch": 3.83, + "learning_rate": 1.85367673825992e-06, + "loss": 0.4872, + "step": 21043 + }, + { + "epoch": 3.83, + "learning_rate": 1.8536221332362579e-06, + "loss": 1.2264, + "step": 21044 + }, + { + "epoch": 3.83, + "learning_rate": 1.8535675282125954e-06, + "loss": 0.0193, + "step": 21045 + }, + { + "epoch": 3.83, + "learning_rate": 1.8535129231889333e-06, + "loss": 3.2368, + "step": 21046 + }, + { + "epoch": 3.83, + "learning_rate": 1.8534583181652713e-06, + "loss": 0.01, + "step": 21047 + }, + { + "epoch": 3.83, + "learning_rate": 1.853403713141609e-06, + "loss": 1.7048, + "step": 21048 + }, + { + "epoch": 3.83, + "learning_rate": 1.853349108117947e-06, + "loss": 0.7369, + "step": 21049 + }, + { + "epoch": 3.83, + "learning_rate": 1.853294503094285e-06, + "loss": 0.5186, + "step": 21050 + }, + { + "epoch": 3.83, + "learning_rate": 1.8532398980706224e-06, + "loss": 0.2842, + "step": 21051 + }, + { + "epoch": 3.83, + "learning_rate": 1.8531852930469604e-06, + "loss": 1.638, + "step": 21052 + }, + { + "epoch": 3.83, + "learning_rate": 1.8531306880232983e-06, + "loss": 1.6746, + "step": 21053 + }, + { + "epoch": 3.83, + "learning_rate": 1.853076082999636e-06, + "loss": 0.2848, + "step": 21054 + }, + { + "epoch": 3.83, + "learning_rate": 1.8530214779759738e-06, + "loss": 0.7382, + "step": 21055 + }, + { + "epoch": 3.83, + "learning_rate": 1.8529668729523117e-06, + "loss": 0.0475, + "step": 21056 + }, + { + "epoch": 3.83, + "learning_rate": 1.8529122679286494e-06, + "loss": 0.8048, + "step": 21057 + }, + { + "epoch": 3.83, + "learning_rate": 1.8528576629049874e-06, + "loss": 1.6449, + "step": 21058 + }, + { + "epoch": 3.83, + "learning_rate": 1.852803057881325e-06, + "loss": 0.916, + "step": 21059 + }, + { + "epoch": 3.83, + "learning_rate": 1.8527484528576629e-06, + "loss": 0.3385, + "step": 21060 + }, + { + "epoch": 3.83, + "learning_rate": 1.8526938478340008e-06, + "loss": 0.4176, + "step": 21061 + }, + { + "epoch": 3.83, + "learning_rate": 1.8526392428103385e-06, + "loss": 0.4161, + "step": 21062 + }, + { + "epoch": 3.83, + "learning_rate": 1.8525846377866765e-06, + "loss": 0.0182, + "step": 21063 + }, + { + "epoch": 3.83, + "learning_rate": 1.8525300327630144e-06, + "loss": 2.3326, + "step": 21064 + }, + { + "epoch": 3.83, + "learning_rate": 1.852475427739352e-06, + "loss": 0.0106, + "step": 21065 + }, + { + "epoch": 3.83, + "learning_rate": 1.8524208227156899e-06, + "loss": 1.349, + "step": 21066 + }, + { + "epoch": 3.83, + "learning_rate": 1.8523662176920278e-06, + "loss": 1.1489, + "step": 21067 + }, + { + "epoch": 3.83, + "learning_rate": 1.8523116126683656e-06, + "loss": 0.8379, + "step": 21068 + }, + { + "epoch": 3.83, + "learning_rate": 1.8522570076447035e-06, + "loss": 0.0158, + "step": 21069 + }, + { + "epoch": 3.83, + "learning_rate": 1.8522024026210414e-06, + "loss": 1.2478, + "step": 21070 + }, + { + "epoch": 3.83, + "learning_rate": 1.852147797597379e-06, + "loss": 2.1988, + "step": 21071 + }, + { + "epoch": 3.83, + "learning_rate": 1.852093192573717e-06, + "loss": 0.0331, + "step": 21072 + }, + { + "epoch": 3.83, + "learning_rate": 1.8520385875500544e-06, + "loss": 1.1331, + "step": 21073 + }, + { + "epoch": 3.83, + "learning_rate": 1.8519839825263924e-06, + "loss": 0.2239, + "step": 21074 + }, + { + "epoch": 3.83, + "learning_rate": 1.8519293775027303e-06, + "loss": 0.3116, + "step": 21075 + }, + { + "epoch": 3.83, + "learning_rate": 1.851874772479068e-06, + "loss": 0.3397, + "step": 21076 + }, + { + "epoch": 3.83, + "learning_rate": 1.851820167455406e-06, + "loss": 0.3169, + "step": 21077 + }, + { + "epoch": 3.83, + "learning_rate": 1.851765562431744e-06, + "loss": 0.0173, + "step": 21078 + }, + { + "epoch": 3.83, + "learning_rate": 1.8517109574080814e-06, + "loss": 2.697, + "step": 21079 + }, + { + "epoch": 3.83, + "learning_rate": 1.8516563523844194e-06, + "loss": 0.5721, + "step": 21080 + }, + { + "epoch": 3.83, + "learning_rate": 1.8516017473607573e-06, + "loss": 0.0241, + "step": 21081 + }, + { + "epoch": 3.83, + "learning_rate": 1.851547142337095e-06, + "loss": 0.9502, + "step": 21082 + }, + { + "epoch": 3.83, + "learning_rate": 1.851492537313433e-06, + "loss": 0.0301, + "step": 21083 + }, + { + "epoch": 3.83, + "learning_rate": 1.8514379322897705e-06, + "loss": 0.3278, + "step": 21084 + }, + { + "epoch": 3.83, + "learning_rate": 1.8513833272661085e-06, + "loss": 2.0772, + "step": 21085 + }, + { + "epoch": 3.83, + "learning_rate": 1.8513287222424464e-06, + "loss": 1.1879, + "step": 21086 + }, + { + "epoch": 3.83, + "learning_rate": 1.851274117218784e-06, + "loss": 1.0367, + "step": 21087 + }, + { + "epoch": 3.83, + "learning_rate": 1.8512195121951219e-06, + "loss": 0.909, + "step": 21088 + }, + { + "epoch": 3.84, + "learning_rate": 1.8511649071714598e-06, + "loss": 2.5298, + "step": 21089 + }, + { + "epoch": 3.84, + "learning_rate": 1.8511103021477976e-06, + "loss": 0.2531, + "step": 21090 + }, + { + "epoch": 3.84, + "learning_rate": 1.8510556971241355e-06, + "loss": 0.1628, + "step": 21091 + }, + { + "epoch": 3.84, + "learning_rate": 1.8510010921004734e-06, + "loss": 0.6784, + "step": 21092 + }, + { + "epoch": 3.84, + "learning_rate": 1.850946487076811e-06, + "loss": 1.1216, + "step": 21093 + }, + { + "epoch": 3.84, + "learning_rate": 1.850891882053149e-06, + "loss": 1.1512, + "step": 21094 + }, + { + "epoch": 3.84, + "learning_rate": 1.8508372770294868e-06, + "loss": 0.3809, + "step": 21095 + }, + { + "epoch": 3.84, + "learning_rate": 1.8507826720058246e-06, + "loss": 0.9535, + "step": 21096 + }, + { + "epoch": 3.84, + "learning_rate": 1.8507280669821625e-06, + "loss": 1.0828, + "step": 21097 + }, + { + "epoch": 3.84, + "learning_rate": 1.8506734619585e-06, + "loss": 0.6457, + "step": 21098 + }, + { + "epoch": 3.84, + "learning_rate": 1.850618856934838e-06, + "loss": 0.3686, + "step": 21099 + }, + { + "epoch": 3.84, + "learning_rate": 1.850564251911176e-06, + "loss": 0.4263, + "step": 21100 + }, + { + "epoch": 3.84, + "learning_rate": 1.8505096468875137e-06, + "loss": 1.1765, + "step": 21101 + }, + { + "epoch": 3.84, + "learning_rate": 1.8504550418638516e-06, + "loss": 1.1735, + "step": 21102 + }, + { + "epoch": 3.84, + "learning_rate": 1.8504004368401895e-06, + "loss": 0.1328, + "step": 21103 + }, + { + "epoch": 3.84, + "learning_rate": 1.850345831816527e-06, + "loss": 0.0125, + "step": 21104 + }, + { + "epoch": 3.84, + "learning_rate": 1.850291226792865e-06, + "loss": 0.1931, + "step": 21105 + }, + { + "epoch": 3.84, + "learning_rate": 1.850236621769203e-06, + "loss": 0.4526, + "step": 21106 + }, + { + "epoch": 3.84, + "learning_rate": 1.8501820167455405e-06, + "loss": 0.1711, + "step": 21107 + }, + { + "epoch": 3.84, + "learning_rate": 1.8501274117218784e-06, + "loss": 1.6409, + "step": 21108 + }, + { + "epoch": 3.84, + "learning_rate": 1.8500728066982164e-06, + "loss": 0.3999, + "step": 21109 + }, + { + "epoch": 3.84, + "learning_rate": 1.850018201674554e-06, + "loss": 1.0489, + "step": 21110 + }, + { + "epoch": 3.84, + "learning_rate": 1.849963596650892e-06, + "loss": 1.4417, + "step": 21111 + }, + { + "epoch": 3.84, + "learning_rate": 1.8499089916272296e-06, + "loss": 0.024, + "step": 21112 + }, + { + "epoch": 3.84, + "learning_rate": 1.8498543866035675e-06, + "loss": 1.3704, + "step": 21113 + }, + { + "epoch": 3.84, + "learning_rate": 1.8497997815799054e-06, + "loss": 0.0057, + "step": 21114 + }, + { + "epoch": 3.84, + "learning_rate": 1.8497451765562432e-06, + "loss": 0.0325, + "step": 21115 + }, + { + "epoch": 3.84, + "learning_rate": 1.8496905715325811e-06, + "loss": 0.8253, + "step": 21116 + }, + { + "epoch": 3.84, + "learning_rate": 1.849635966508919e-06, + "loss": 0.1905, + "step": 21117 + }, + { + "epoch": 3.84, + "learning_rate": 1.8495813614852566e-06, + "loss": 0.0148, + "step": 21118 + }, + { + "epoch": 3.84, + "learning_rate": 1.8495267564615945e-06, + "loss": 0.3624, + "step": 21119 + }, + { + "epoch": 3.84, + "learning_rate": 1.8494721514379325e-06, + "loss": 0.706, + "step": 21120 + }, + { + "epoch": 3.84, + "learning_rate": 1.84941754641427e-06, + "loss": 3.2742, + "step": 21121 + }, + { + "epoch": 3.84, + "learning_rate": 1.849362941390608e-06, + "loss": 0.5291, + "step": 21122 + }, + { + "epoch": 3.84, + "learning_rate": 1.8493083363669459e-06, + "loss": 1.6478, + "step": 21123 + }, + { + "epoch": 3.84, + "learning_rate": 1.8492537313432836e-06, + "loss": 0.5143, + "step": 21124 + }, + { + "epoch": 3.84, + "learning_rate": 1.8491991263196216e-06, + "loss": 0.7016, + "step": 21125 + }, + { + "epoch": 3.84, + "learning_rate": 1.849144521295959e-06, + "loss": 0.3654, + "step": 21126 + }, + { + "epoch": 3.84, + "learning_rate": 1.849089916272297e-06, + "loss": 0.0137, + "step": 21127 + }, + { + "epoch": 3.84, + "learning_rate": 1.849035311248635e-06, + "loss": 1.3927, + "step": 21128 + }, + { + "epoch": 3.84, + "learning_rate": 1.8489807062249727e-06, + "loss": 1.0481, + "step": 21129 + }, + { + "epoch": 3.84, + "learning_rate": 1.8489261012013106e-06, + "loss": 0.9401, + "step": 21130 + }, + { + "epoch": 3.84, + "learning_rate": 1.8488714961776486e-06, + "loss": 0.5183, + "step": 21131 + }, + { + "epoch": 3.84, + "learning_rate": 1.848816891153986e-06, + "loss": 4.2679, + "step": 21132 + }, + { + "epoch": 3.84, + "learning_rate": 1.848762286130324e-06, + "loss": 0.1565, + "step": 21133 + }, + { + "epoch": 3.84, + "learning_rate": 1.848707681106662e-06, + "loss": 0.3103, + "step": 21134 + }, + { + "epoch": 3.84, + "learning_rate": 1.8486530760829995e-06, + "loss": 0.9867, + "step": 21135 + }, + { + "epoch": 3.84, + "learning_rate": 1.8485984710593374e-06, + "loss": 0.3617, + "step": 21136 + }, + { + "epoch": 3.84, + "learning_rate": 1.8485438660356754e-06, + "loss": 1.1736, + "step": 21137 + }, + { + "epoch": 3.84, + "learning_rate": 1.8484892610120131e-06, + "loss": 0.0137, + "step": 21138 + }, + { + "epoch": 3.84, + "learning_rate": 1.848434655988351e-06, + "loss": 2.1452, + "step": 21139 + }, + { + "epoch": 3.84, + "learning_rate": 1.8483800509646886e-06, + "loss": 0.0092, + "step": 21140 + }, + { + "epoch": 3.84, + "learning_rate": 1.8483254459410265e-06, + "loss": 1.6829, + "step": 21141 + }, + { + "epoch": 3.84, + "learning_rate": 1.8482708409173645e-06, + "loss": 1.5384, + "step": 21142 + }, + { + "epoch": 3.84, + "learning_rate": 1.8482162358937022e-06, + "loss": 0.271, + "step": 21143 + }, + { + "epoch": 3.85, + "learning_rate": 1.8481616308700401e-06, + "loss": 0.4965, + "step": 21144 + }, + { + "epoch": 3.85, + "learning_rate": 1.848107025846378e-06, + "loss": 2.6742, + "step": 21145 + }, + { + "epoch": 3.85, + "learning_rate": 1.8480524208227156e-06, + "loss": 0.7146, + "step": 21146 + }, + { + "epoch": 3.85, + "learning_rate": 1.8479978157990536e-06, + "loss": 0.009, + "step": 21147 + }, + { + "epoch": 3.85, + "learning_rate": 1.8479432107753915e-06, + "loss": 0.1556, + "step": 21148 + }, + { + "epoch": 3.85, + "learning_rate": 1.8478886057517292e-06, + "loss": 0.4812, + "step": 21149 + }, + { + "epoch": 3.85, + "learning_rate": 1.8478340007280672e-06, + "loss": 0.4226, + "step": 21150 + }, + { + "epoch": 3.85, + "learning_rate": 1.8477793957044051e-06, + "loss": 1.6912, + "step": 21151 + }, + { + "epoch": 3.85, + "learning_rate": 1.8477247906807426e-06, + "loss": 0.1849, + "step": 21152 + }, + { + "epoch": 3.85, + "learning_rate": 1.8476701856570806e-06, + "loss": 2.9138, + "step": 21153 + }, + { + "epoch": 3.85, + "learning_rate": 1.847615580633418e-06, + "loss": 1.8904, + "step": 21154 + }, + { + "epoch": 3.85, + "learning_rate": 1.847560975609756e-06, + "loss": 0.7853, + "step": 21155 + }, + { + "epoch": 3.85, + "learning_rate": 1.847506370586094e-06, + "loss": 0.1781, + "step": 21156 + }, + { + "epoch": 3.85, + "learning_rate": 1.8474517655624317e-06, + "loss": 0.8468, + "step": 21157 + }, + { + "epoch": 3.85, + "learning_rate": 1.8473971605387697e-06, + "loss": 1.1132, + "step": 21158 + }, + { + "epoch": 3.85, + "learning_rate": 1.8473425555151076e-06, + "loss": 0.3536, + "step": 21159 + }, + { + "epoch": 3.85, + "learning_rate": 1.8472879504914451e-06, + "loss": 1.906, + "step": 21160 + }, + { + "epoch": 3.85, + "learning_rate": 1.847233345467783e-06, + "loss": 0.3312, + "step": 21161 + }, + { + "epoch": 3.85, + "learning_rate": 1.847178740444121e-06, + "loss": 0.1027, + "step": 21162 + }, + { + "epoch": 3.85, + "learning_rate": 1.8471241354204587e-06, + "loss": 0.2991, + "step": 21163 + }, + { + "epoch": 3.85, + "learning_rate": 1.8470695303967967e-06, + "loss": 1.8704, + "step": 21164 + }, + { + "epoch": 3.85, + "learning_rate": 1.8470149253731342e-06, + "loss": 0.0404, + "step": 21165 + }, + { + "epoch": 3.85, + "learning_rate": 1.8469603203494721e-06, + "loss": 2.6919, + "step": 21166 + }, + { + "epoch": 3.85, + "learning_rate": 1.84690571532581e-06, + "loss": 0.6843, + "step": 21167 + }, + { + "epoch": 3.85, + "learning_rate": 1.8468511103021476e-06, + "loss": 0.7129, + "step": 21168 + }, + { + "epoch": 3.85, + "learning_rate": 1.8467965052784856e-06, + "loss": 0.0205, + "step": 21169 + }, + { + "epoch": 3.85, + "learning_rate": 1.8467419002548235e-06, + "loss": 0.5795, + "step": 21170 + }, + { + "epoch": 3.85, + "learning_rate": 1.8466872952311612e-06, + "loss": 0.0179, + "step": 21171 + }, + { + "epoch": 3.85, + "learning_rate": 1.8466326902074992e-06, + "loss": 1.0469, + "step": 21172 + }, + { + "epoch": 3.85, + "learning_rate": 1.8465780851838371e-06, + "loss": 0.0163, + "step": 21173 + }, + { + "epoch": 3.85, + "learning_rate": 1.8465234801601746e-06, + "loss": 1.5139, + "step": 21174 + }, + { + "epoch": 3.85, + "learning_rate": 1.8464688751365126e-06, + "loss": 0.7004, + "step": 21175 + }, + { + "epoch": 3.85, + "learning_rate": 1.8464142701128505e-06, + "loss": 0.2886, + "step": 21176 + }, + { + "epoch": 3.85, + "learning_rate": 1.8463596650891883e-06, + "loss": 0.481, + "step": 21177 + }, + { + "epoch": 3.85, + "learning_rate": 1.8463050600655262e-06, + "loss": 0.0416, + "step": 21178 + }, + { + "epoch": 3.85, + "learning_rate": 1.8462504550418637e-06, + "loss": 0.2805, + "step": 21179 + }, + { + "epoch": 3.85, + "learning_rate": 1.8461958500182017e-06, + "loss": 0.0168, + "step": 21180 + }, + { + "epoch": 3.85, + "learning_rate": 1.8461412449945396e-06, + "loss": 0.0151, + "step": 21181 + }, + { + "epoch": 3.85, + "learning_rate": 1.8460866399708773e-06, + "loss": 0.2672, + "step": 21182 + }, + { + "epoch": 3.85, + "learning_rate": 1.8460320349472153e-06, + "loss": 0.0379, + "step": 21183 + }, + { + "epoch": 3.85, + "learning_rate": 1.845977429923553e-06, + "loss": 2.1158, + "step": 21184 + }, + { + "epoch": 3.85, + "learning_rate": 1.8459228248998907e-06, + "loss": 0.4942, + "step": 21185 + }, + { + "epoch": 3.85, + "learning_rate": 1.8458682198762287e-06, + "loss": 2.4546, + "step": 21186 + }, + { + "epoch": 3.85, + "learning_rate": 1.8458136148525666e-06, + "loss": 1.4146, + "step": 21187 + }, + { + "epoch": 3.85, + "learning_rate": 1.8457590098289041e-06, + "loss": 1.0575, + "step": 21188 + }, + { + "epoch": 3.85, + "learning_rate": 1.845704404805242e-06, + "loss": 0.011, + "step": 21189 + }, + { + "epoch": 3.85, + "learning_rate": 1.84564979978158e-06, + "loss": 3.2722, + "step": 21190 + }, + { + "epoch": 3.85, + "learning_rate": 1.8455951947579178e-06, + "loss": 0.7475, + "step": 21191 + }, + { + "epoch": 3.85, + "learning_rate": 1.8455405897342557e-06, + "loss": 0.1514, + "step": 21192 + }, + { + "epoch": 3.85, + "learning_rate": 1.8454859847105932e-06, + "loss": 0.0215, + "step": 21193 + }, + { + "epoch": 3.85, + "learning_rate": 1.8454313796869312e-06, + "loss": 0.0675, + "step": 21194 + }, + { + "epoch": 3.85, + "learning_rate": 1.8453767746632691e-06, + "loss": 1.7354, + "step": 21195 + }, + { + "epoch": 3.85, + "learning_rate": 1.8453221696396068e-06, + "loss": 1.2269, + "step": 21196 + }, + { + "epoch": 3.85, + "learning_rate": 1.8452675646159448e-06, + "loss": 0.6147, + "step": 21197 + }, + { + "epoch": 3.85, + "learning_rate": 1.8452129595922827e-06, + "loss": 0.4845, + "step": 21198 + }, + { + "epoch": 3.86, + "learning_rate": 1.8451583545686203e-06, + "loss": 0.2876, + "step": 21199 + }, + { + "epoch": 3.86, + "learning_rate": 1.8451037495449582e-06, + "loss": 1.6167, + "step": 21200 + }, + { + "epoch": 3.86, + "learning_rate": 1.8450491445212961e-06, + "loss": 0.21, + "step": 21201 + }, + { + "epoch": 3.86, + "learning_rate": 1.8449945394976337e-06, + "loss": 0.3983, + "step": 21202 + }, + { + "epoch": 3.86, + "learning_rate": 1.8449399344739716e-06, + "loss": 0.8098, + "step": 21203 + }, + { + "epoch": 3.86, + "learning_rate": 1.8448853294503095e-06, + "loss": 0.6936, + "step": 21204 + }, + { + "epoch": 3.86, + "learning_rate": 1.8448307244266473e-06, + "loss": 0.5305, + "step": 21205 + }, + { + "epoch": 3.86, + "learning_rate": 1.8447761194029852e-06, + "loss": 0.1086, + "step": 21206 + }, + { + "epoch": 3.86, + "learning_rate": 1.8447215143793227e-06, + "loss": 0.5599, + "step": 21207 + }, + { + "epoch": 3.86, + "learning_rate": 1.8446669093556607e-06, + "loss": 0.4821, + "step": 21208 + }, + { + "epoch": 3.86, + "learning_rate": 1.8446123043319986e-06, + "loss": 1.2104, + "step": 21209 + }, + { + "epoch": 3.86, + "learning_rate": 1.8445576993083364e-06, + "loss": 0.7212, + "step": 21210 + }, + { + "epoch": 3.86, + "learning_rate": 1.8445030942846743e-06, + "loss": 0.4615, + "step": 21211 + }, + { + "epoch": 3.86, + "learning_rate": 1.8444484892610123e-06, + "loss": 1.5496, + "step": 21212 + }, + { + "epoch": 3.86, + "learning_rate": 1.8443938842373498e-06, + "loss": 0.397, + "step": 21213 + }, + { + "epoch": 3.86, + "learning_rate": 1.8443392792136877e-06, + "loss": 0.0129, + "step": 21214 + }, + { + "epoch": 3.86, + "learning_rate": 1.8442846741900257e-06, + "loss": 0.203, + "step": 21215 + }, + { + "epoch": 3.86, + "learning_rate": 1.8442300691663632e-06, + "loss": 2.45, + "step": 21216 + }, + { + "epoch": 3.86, + "learning_rate": 1.8441754641427011e-06, + "loss": 0.1187, + "step": 21217 + }, + { + "epoch": 3.86, + "learning_rate": 1.844120859119039e-06, + "loss": 0.029, + "step": 21218 + }, + { + "epoch": 3.86, + "learning_rate": 1.8440662540953768e-06, + "loss": 0.2656, + "step": 21219 + }, + { + "epoch": 3.86, + "learning_rate": 1.8440116490717147e-06, + "loss": 0.4497, + "step": 21220 + }, + { + "epoch": 3.86, + "learning_rate": 1.8439570440480523e-06, + "loss": 0.1634, + "step": 21221 + }, + { + "epoch": 3.86, + "learning_rate": 1.8439024390243902e-06, + "loss": 0.0134, + "step": 21222 + }, + { + "epoch": 3.86, + "learning_rate": 1.8438478340007281e-06, + "loss": 1.5809, + "step": 21223 + }, + { + "epoch": 3.86, + "learning_rate": 1.8437932289770659e-06, + "loss": 0.7761, + "step": 21224 + }, + { + "epoch": 3.86, + "learning_rate": 1.8437386239534038e-06, + "loss": 0.9839, + "step": 21225 + }, + { + "epoch": 3.86, + "learning_rate": 1.8436840189297418e-06, + "loss": 1.1495, + "step": 21226 + }, + { + "epoch": 3.86, + "learning_rate": 1.8436294139060793e-06, + "loss": 2.1396, + "step": 21227 + }, + { + "epoch": 3.86, + "learning_rate": 1.8435748088824172e-06, + "loss": 1.3931, + "step": 21228 + }, + { + "epoch": 3.86, + "learning_rate": 1.8435202038587552e-06, + "loss": 0.8168, + "step": 21229 + }, + { + "epoch": 3.86, + "learning_rate": 1.843465598835093e-06, + "loss": 1.4895, + "step": 21230 + }, + { + "epoch": 3.86, + "learning_rate": 1.8434109938114308e-06, + "loss": 0.2272, + "step": 21231 + }, + { + "epoch": 3.86, + "learning_rate": 1.8433563887877688e-06, + "loss": 1.1384, + "step": 21232 + }, + { + "epoch": 3.86, + "learning_rate": 1.8433017837641063e-06, + "loss": 0.5561, + "step": 21233 + }, + { + "epoch": 3.86, + "learning_rate": 1.8432471787404443e-06, + "loss": 0.2946, + "step": 21234 + }, + { + "epoch": 3.86, + "learning_rate": 1.8431925737167818e-06, + "loss": 0.5091, + "step": 21235 + }, + { + "epoch": 3.86, + "learning_rate": 1.8431379686931197e-06, + "loss": 1.7885, + "step": 21236 + }, + { + "epoch": 3.86, + "learning_rate": 1.8430833636694577e-06, + "loss": 1.2428, + "step": 21237 + }, + { + "epoch": 3.86, + "learning_rate": 1.8430287586457954e-06, + "loss": 0.3413, + "step": 21238 + }, + { + "epoch": 3.86, + "learning_rate": 1.8429741536221333e-06, + "loss": 0.2955, + "step": 21239 + }, + { + "epoch": 3.86, + "learning_rate": 1.8429195485984713e-06, + "loss": 1.5305, + "step": 21240 + }, + { + "epoch": 3.86, + "learning_rate": 1.8428649435748088e-06, + "loss": 0.7233, + "step": 21241 + }, + { + "epoch": 3.86, + "learning_rate": 1.8428103385511467e-06, + "loss": 0.5778, + "step": 21242 + }, + { + "epoch": 3.86, + "learning_rate": 1.8427557335274847e-06, + "loss": 1.7294, + "step": 21243 + }, + { + "epoch": 3.86, + "learning_rate": 1.8427011285038224e-06, + "loss": 1.6914, + "step": 21244 + }, + { + "epoch": 3.86, + "learning_rate": 1.8426465234801604e-06, + "loss": 0.937, + "step": 21245 + }, + { + "epoch": 3.86, + "learning_rate": 1.8425919184564979e-06, + "loss": 2.8687, + "step": 21246 + }, + { + "epoch": 3.86, + "learning_rate": 1.8425373134328358e-06, + "loss": 1.4014, + "step": 21247 + }, + { + "epoch": 3.86, + "learning_rate": 1.8424827084091738e-06, + "loss": 2.25, + "step": 21248 + }, + { + "epoch": 3.86, + "learning_rate": 1.8424281033855113e-06, + "loss": 0.8568, + "step": 21249 + }, + { + "epoch": 3.86, + "learning_rate": 1.8423734983618492e-06, + "loss": 0.3054, + "step": 21250 + }, + { + "epoch": 3.86, + "learning_rate": 1.8423188933381872e-06, + "loss": 0.8464, + "step": 21251 + }, + { + "epoch": 3.86, + "learning_rate": 1.842264288314525e-06, + "loss": 0.1508, + "step": 21252 + }, + { + "epoch": 3.86, + "learning_rate": 1.8422096832908628e-06, + "loss": 0.3685, + "step": 21253 + }, + { + "epoch": 3.87, + "learning_rate": 1.8421550782672008e-06, + "loss": 0.1864, + "step": 21254 + }, + { + "epoch": 3.87, + "learning_rate": 1.8421004732435383e-06, + "loss": 0.8878, + "step": 21255 + }, + { + "epoch": 3.87, + "learning_rate": 1.8420458682198763e-06, + "loss": 0.2137, + "step": 21256 + }, + { + "epoch": 3.87, + "learning_rate": 1.8419912631962142e-06, + "loss": 0.6438, + "step": 21257 + }, + { + "epoch": 3.87, + "learning_rate": 1.841936658172552e-06, + "loss": 0.0121, + "step": 21258 + }, + { + "epoch": 3.87, + "learning_rate": 1.8418820531488899e-06, + "loss": 0.0197, + "step": 21259 + }, + { + "epoch": 3.87, + "learning_rate": 1.8418274481252274e-06, + "loss": 0.5833, + "step": 21260 + }, + { + "epoch": 3.87, + "learning_rate": 1.8417728431015653e-06, + "loss": 0.1012, + "step": 21261 + }, + { + "epoch": 3.87, + "learning_rate": 1.8417182380779033e-06, + "loss": 0.6, + "step": 21262 + }, + { + "epoch": 3.87, + "learning_rate": 1.8416636330542408e-06, + "loss": 0.0903, + "step": 21263 + }, + { + "epoch": 3.87, + "learning_rate": 1.8416090280305787e-06, + "loss": 0.4194, + "step": 21264 + }, + { + "epoch": 3.87, + "learning_rate": 1.8415544230069167e-06, + "loss": 0.3727, + "step": 21265 + }, + { + "epoch": 3.87, + "learning_rate": 1.8414998179832544e-06, + "loss": 0.3379, + "step": 21266 + }, + { + "epoch": 3.87, + "learning_rate": 1.8414452129595924e-06, + "loss": 0.7501, + "step": 21267 + }, + { + "epoch": 3.87, + "learning_rate": 1.8413906079359303e-06, + "loss": 0.1828, + "step": 21268 + }, + { + "epoch": 3.87, + "learning_rate": 1.8413360029122678e-06, + "loss": 2.002, + "step": 21269 + }, + { + "epoch": 3.87, + "learning_rate": 1.8412813978886058e-06, + "loss": 1.5992, + "step": 21270 + }, + { + "epoch": 3.87, + "learning_rate": 1.8412267928649437e-06, + "loss": 0.1812, + "step": 21271 + }, + { + "epoch": 3.87, + "learning_rate": 1.8411721878412814e-06, + "loss": 0.453, + "step": 21272 + }, + { + "epoch": 3.87, + "learning_rate": 1.8411175828176194e-06, + "loss": 0.249, + "step": 21273 + }, + { + "epoch": 3.87, + "learning_rate": 1.841062977793957e-06, + "loss": 0.4693, + "step": 21274 + }, + { + "epoch": 3.87, + "learning_rate": 1.8410083727702948e-06, + "loss": 0.4922, + "step": 21275 + }, + { + "epoch": 3.87, + "learning_rate": 1.8409537677466328e-06, + "loss": 0.7, + "step": 21276 + }, + { + "epoch": 3.87, + "learning_rate": 1.8408991627229705e-06, + "loss": 1.2676, + "step": 21277 + }, + { + "epoch": 3.87, + "learning_rate": 1.8408445576993085e-06, + "loss": 0.0237, + "step": 21278 + }, + { + "epoch": 3.87, + "learning_rate": 1.8407899526756464e-06, + "loss": 0.8565, + "step": 21279 + }, + { + "epoch": 3.87, + "learning_rate": 1.840735347651984e-06, + "loss": 0.0731, + "step": 21280 + }, + { + "epoch": 3.87, + "learning_rate": 1.8406807426283219e-06, + "loss": 0.0135, + "step": 21281 + }, + { + "epoch": 3.87, + "learning_rate": 1.8406261376046598e-06, + "loss": 2.1126, + "step": 21282 + }, + { + "epoch": 3.87, + "learning_rate": 1.8405715325809973e-06, + "loss": 0.7907, + "step": 21283 + }, + { + "epoch": 3.87, + "learning_rate": 1.8405169275573353e-06, + "loss": 0.3149, + "step": 21284 + }, + { + "epoch": 3.87, + "learning_rate": 1.8404623225336732e-06, + "loss": 0.2736, + "step": 21285 + }, + { + "epoch": 3.87, + "learning_rate": 1.840407717510011e-06, + "loss": 1.3176, + "step": 21286 + }, + { + "epoch": 3.87, + "learning_rate": 1.840353112486349e-06, + "loss": 0.282, + "step": 21287 + }, + { + "epoch": 3.87, + "learning_rate": 1.8402985074626864e-06, + "loss": 0.6951, + "step": 21288 + }, + { + "epoch": 3.87, + "learning_rate": 1.8402439024390244e-06, + "loss": 1.39, + "step": 21289 + }, + { + "epoch": 3.87, + "learning_rate": 1.8401892974153623e-06, + "loss": 0.179, + "step": 21290 + }, + { + "epoch": 3.87, + "learning_rate": 1.8401346923917e-06, + "loss": 0.0325, + "step": 21291 + }, + { + "epoch": 3.87, + "learning_rate": 1.840080087368038e-06, + "loss": 2.0543, + "step": 21292 + }, + { + "epoch": 3.87, + "learning_rate": 1.840025482344376e-06, + "loss": 0.94, + "step": 21293 + }, + { + "epoch": 3.87, + "learning_rate": 1.8399708773207134e-06, + "loss": 1.0561, + "step": 21294 + }, + { + "epoch": 3.87, + "learning_rate": 1.8399162722970514e-06, + "loss": 0.698, + "step": 21295 + }, + { + "epoch": 3.87, + "learning_rate": 1.8398616672733893e-06, + "loss": 1.1211, + "step": 21296 + }, + { + "epoch": 3.87, + "learning_rate": 1.8398070622497269e-06, + "loss": 0.1059, + "step": 21297 + }, + { + "epoch": 3.87, + "learning_rate": 1.8397524572260648e-06, + "loss": 3.1359, + "step": 21298 + }, + { + "epoch": 3.87, + "learning_rate": 1.8396978522024027e-06, + "loss": 0.6995, + "step": 21299 + }, + { + "epoch": 3.87, + "learning_rate": 1.8396432471787405e-06, + "loss": 0.0609, + "step": 21300 + }, + { + "epoch": 3.87, + "learning_rate": 1.8395886421550784e-06, + "loss": 0.0334, + "step": 21301 + }, + { + "epoch": 3.87, + "learning_rate": 1.839534037131416e-06, + "loss": 0.0111, + "step": 21302 + }, + { + "epoch": 3.87, + "learning_rate": 1.8394794321077539e-06, + "loss": 0.0174, + "step": 21303 + }, + { + "epoch": 3.87, + "learning_rate": 1.8394248270840918e-06, + "loss": 0.1846, + "step": 21304 + }, + { + "epoch": 3.87, + "learning_rate": 1.8393702220604296e-06, + "loss": 0.126, + "step": 21305 + }, + { + "epoch": 3.87, + "learning_rate": 1.8393156170367675e-06, + "loss": 2.0981, + "step": 21306 + }, + { + "epoch": 3.87, + "learning_rate": 1.8392610120131054e-06, + "loss": 0.0359, + "step": 21307 + }, + { + "epoch": 3.87, + "learning_rate": 1.839206406989443e-06, + "loss": 1.1213, + "step": 21308 + }, + { + "epoch": 3.88, + "learning_rate": 1.839151801965781e-06, + "loss": 0.2931, + "step": 21309 + }, + { + "epoch": 3.88, + "learning_rate": 1.8390971969421188e-06, + "loss": 0.6428, + "step": 21310 + }, + { + "epoch": 3.88, + "learning_rate": 1.8390425919184566e-06, + "loss": 0.4058, + "step": 21311 + }, + { + "epoch": 3.88, + "learning_rate": 1.8389879868947945e-06, + "loss": 3.5805, + "step": 21312 + }, + { + "epoch": 3.88, + "learning_rate": 1.838933381871132e-06, + "loss": 0.3612, + "step": 21313 + }, + { + "epoch": 3.88, + "learning_rate": 1.83887877684747e-06, + "loss": 0.0232, + "step": 21314 + }, + { + "epoch": 3.88, + "learning_rate": 1.838824171823808e-06, + "loss": 1.0927, + "step": 21315 + }, + { + "epoch": 3.88, + "learning_rate": 1.8387695668001454e-06, + "loss": 0.4817, + "step": 21316 + }, + { + "epoch": 3.88, + "learning_rate": 1.8387149617764834e-06, + "loss": 0.0937, + "step": 21317 + }, + { + "epoch": 3.88, + "learning_rate": 1.8386603567528213e-06, + "loss": 0.0787, + "step": 21318 + }, + { + "epoch": 3.88, + "learning_rate": 1.838605751729159e-06, + "loss": 0.048, + "step": 21319 + }, + { + "epoch": 3.88, + "learning_rate": 1.838551146705497e-06, + "loss": 1.6202, + "step": 21320 + }, + { + "epoch": 3.88, + "learning_rate": 1.838496541681835e-06, + "loss": 1.0139, + "step": 21321 + }, + { + "epoch": 3.88, + "learning_rate": 1.8384419366581725e-06, + "loss": 0.0172, + "step": 21322 + }, + { + "epoch": 3.88, + "learning_rate": 1.8383873316345104e-06, + "loss": 1.5166, + "step": 21323 + }, + { + "epoch": 3.88, + "learning_rate": 1.8383327266108484e-06, + "loss": 0.2332, + "step": 21324 + }, + { + "epoch": 3.88, + "learning_rate": 1.838278121587186e-06, + "loss": 0.8806, + "step": 21325 + }, + { + "epoch": 3.88, + "learning_rate": 1.838223516563524e-06, + "loss": 0.548, + "step": 21326 + }, + { + "epoch": 3.88, + "learning_rate": 1.8381689115398616e-06, + "loss": 0.0805, + "step": 21327 + }, + { + "epoch": 3.88, + "learning_rate": 1.8381143065161995e-06, + "loss": 0.4429, + "step": 21328 + }, + { + "epoch": 3.88, + "learning_rate": 1.8380597014925374e-06, + "loss": 0.2449, + "step": 21329 + }, + { + "epoch": 3.88, + "learning_rate": 1.838005096468875e-06, + "loss": 1.0166, + "step": 21330 + }, + { + "epoch": 3.88, + "learning_rate": 1.837950491445213e-06, + "loss": 0.5073, + "step": 21331 + }, + { + "epoch": 3.88, + "learning_rate": 1.8378958864215508e-06, + "loss": 1.1056, + "step": 21332 + }, + { + "epoch": 3.88, + "learning_rate": 1.8378412813978886e-06, + "loss": 0.1884, + "step": 21333 + }, + { + "epoch": 3.88, + "learning_rate": 1.8377866763742265e-06, + "loss": 0.2867, + "step": 21334 + }, + { + "epoch": 3.88, + "learning_rate": 1.8377320713505645e-06, + "loss": 1.1329, + "step": 21335 + }, + { + "epoch": 3.88, + "learning_rate": 1.837677466326902e-06, + "loss": 0.099, + "step": 21336 + }, + { + "epoch": 3.88, + "learning_rate": 1.83762286130324e-06, + "loss": 0.2552, + "step": 21337 + }, + { + "epoch": 3.88, + "learning_rate": 1.8375682562795779e-06, + "loss": 2.0888, + "step": 21338 + }, + { + "epoch": 3.88, + "learning_rate": 1.8375136512559156e-06, + "loss": 0.3268, + "step": 21339 + }, + { + "epoch": 3.88, + "learning_rate": 1.8374590462322535e-06, + "loss": 0.2634, + "step": 21340 + }, + { + "epoch": 3.88, + "learning_rate": 1.837404441208591e-06, + "loss": 0.3294, + "step": 21341 + }, + { + "epoch": 3.88, + "learning_rate": 1.837349836184929e-06, + "loss": 1.2217, + "step": 21342 + }, + { + "epoch": 3.88, + "learning_rate": 1.837295231161267e-06, + "loss": 0.0152, + "step": 21343 + }, + { + "epoch": 3.88, + "learning_rate": 1.8372406261376045e-06, + "loss": 1.2022, + "step": 21344 + }, + { + "epoch": 3.88, + "learning_rate": 1.8371860211139424e-06, + "loss": 1.2667, + "step": 21345 + }, + { + "epoch": 3.88, + "learning_rate": 1.8371314160902804e-06, + "loss": 2.053, + "step": 21346 + }, + { + "epoch": 3.88, + "learning_rate": 1.837076811066618e-06, + "loss": 1.545, + "step": 21347 + }, + { + "epoch": 3.88, + "learning_rate": 1.837022206042956e-06, + "loss": 0.2857, + "step": 21348 + }, + { + "epoch": 3.88, + "learning_rate": 1.836967601019294e-06, + "loss": 0.4028, + "step": 21349 + }, + { + "epoch": 3.88, + "learning_rate": 1.8369129959956315e-06, + "loss": 0.794, + "step": 21350 + }, + { + "epoch": 3.88, + "learning_rate": 1.8368583909719694e-06, + "loss": 1.1436, + "step": 21351 + }, + { + "epoch": 3.88, + "learning_rate": 1.8368037859483074e-06, + "loss": 1.1696, + "step": 21352 + }, + { + "epoch": 3.88, + "learning_rate": 1.8367491809246451e-06, + "loss": 0.5553, + "step": 21353 + }, + { + "epoch": 3.88, + "learning_rate": 1.836694575900983e-06, + "loss": 0.2892, + "step": 21354 + }, + { + "epoch": 3.88, + "learning_rate": 1.8366399708773206e-06, + "loss": 0.9313, + "step": 21355 + }, + { + "epoch": 3.88, + "learning_rate": 1.8365853658536585e-06, + "loss": 0.0218, + "step": 21356 + }, + { + "epoch": 3.88, + "learning_rate": 1.8365307608299965e-06, + "loss": 0.2925, + "step": 21357 + }, + { + "epoch": 3.88, + "learning_rate": 1.8364761558063342e-06, + "loss": 1.7224, + "step": 21358 + }, + { + "epoch": 3.88, + "learning_rate": 1.8364215507826721e-06, + "loss": 1.1288, + "step": 21359 + }, + { + "epoch": 3.88, + "learning_rate": 1.83636694575901e-06, + "loss": 0.2103, + "step": 21360 + }, + { + "epoch": 3.88, + "learning_rate": 1.8363123407353476e-06, + "loss": 0.4309, + "step": 21361 + }, + { + "epoch": 3.88, + "learning_rate": 1.8362577357116855e-06, + "loss": 0.3541, + "step": 21362 + }, + { + "epoch": 3.88, + "learning_rate": 1.8362031306880235e-06, + "loss": 0.6378, + "step": 21363 + }, + { + "epoch": 3.89, + "learning_rate": 1.836148525664361e-06, + "loss": 1.3226, + "step": 21364 + }, + { + "epoch": 3.89, + "learning_rate": 1.836093920640699e-06, + "loss": 0.4533, + "step": 21365 + }, + { + "epoch": 3.89, + "learning_rate": 1.836039315617037e-06, + "loss": 0.0732, + "step": 21366 + }, + { + "epoch": 3.89, + "learning_rate": 1.8359847105933746e-06, + "loss": 0.0271, + "step": 21367 + }, + { + "epoch": 3.89, + "learning_rate": 1.8359301055697126e-06, + "loss": 0.7961, + "step": 21368 + }, + { + "epoch": 3.89, + "learning_rate": 1.83587550054605e-06, + "loss": 1.823, + "step": 21369 + }, + { + "epoch": 3.89, + "learning_rate": 1.835820895522388e-06, + "loss": 0.5962, + "step": 21370 + }, + { + "epoch": 3.89, + "learning_rate": 1.835766290498726e-06, + "loss": 0.2295, + "step": 21371 + }, + { + "epoch": 3.89, + "learning_rate": 1.8357116854750637e-06, + "loss": 0.0345, + "step": 21372 + }, + { + "epoch": 3.89, + "learning_rate": 1.8356570804514017e-06, + "loss": 0.6347, + "step": 21373 + }, + { + "epoch": 3.89, + "learning_rate": 1.8356024754277396e-06, + "loss": 0.0085, + "step": 21374 + }, + { + "epoch": 3.89, + "learning_rate": 1.8355478704040771e-06, + "loss": 0.0511, + "step": 21375 + }, + { + "epoch": 3.89, + "learning_rate": 1.835493265380415e-06, + "loss": 0.1933, + "step": 21376 + }, + { + "epoch": 3.89, + "learning_rate": 1.835438660356753e-06, + "loss": 0.4523, + "step": 21377 + }, + { + "epoch": 3.89, + "learning_rate": 1.8353840553330905e-06, + "loss": 2.4212, + "step": 21378 + }, + { + "epoch": 3.89, + "learning_rate": 1.8353294503094285e-06, + "loss": 0.0167, + "step": 21379 + }, + { + "epoch": 3.89, + "learning_rate": 1.8352748452857664e-06, + "loss": 0.3353, + "step": 21380 + }, + { + "epoch": 3.89, + "learning_rate": 1.8352202402621041e-06, + "loss": 0.0175, + "step": 21381 + }, + { + "epoch": 3.89, + "learning_rate": 1.835165635238442e-06, + "loss": 0.7484, + "step": 21382 + }, + { + "epoch": 3.89, + "learning_rate": 1.8351110302147796e-06, + "loss": 1.0556, + "step": 21383 + }, + { + "epoch": 3.89, + "learning_rate": 1.8350564251911176e-06, + "loss": 0.2683, + "step": 21384 + }, + { + "epoch": 3.89, + "learning_rate": 1.8350018201674555e-06, + "loss": 0.5977, + "step": 21385 + }, + { + "epoch": 3.89, + "learning_rate": 1.8349472151437932e-06, + "loss": 1.8222, + "step": 21386 + }, + { + "epoch": 3.89, + "learning_rate": 1.8348926101201312e-06, + "loss": 1.7365, + "step": 21387 + }, + { + "epoch": 3.89, + "learning_rate": 1.8348380050964691e-06, + "loss": 0.2855, + "step": 21388 + }, + { + "epoch": 3.89, + "learning_rate": 1.8347834000728066e-06, + "loss": 3.4993, + "step": 21389 + }, + { + "epoch": 3.89, + "learning_rate": 1.8347287950491446e-06, + "loss": 0.3774, + "step": 21390 + }, + { + "epoch": 3.89, + "learning_rate": 1.8346741900254825e-06, + "loss": 0.8445, + "step": 21391 + }, + { + "epoch": 3.89, + "learning_rate": 1.83461958500182e-06, + "loss": 1.2523, + "step": 21392 + }, + { + "epoch": 3.89, + "learning_rate": 1.834564979978158e-06, + "loss": 1.6209, + "step": 21393 + }, + { + "epoch": 3.89, + "learning_rate": 1.8345103749544957e-06, + "loss": 0.0114, + "step": 21394 + }, + { + "epoch": 3.89, + "learning_rate": 1.8344557699308337e-06, + "loss": 0.514, + "step": 21395 + }, + { + "epoch": 3.89, + "learning_rate": 1.8344011649071716e-06, + "loss": 2.4098, + "step": 21396 + }, + { + "epoch": 3.89, + "learning_rate": 1.8343465598835091e-06, + "loss": 0.4832, + "step": 21397 + }, + { + "epoch": 3.89, + "learning_rate": 1.834291954859847e-06, + "loss": 0.2262, + "step": 21398 + }, + { + "epoch": 3.89, + "learning_rate": 1.834237349836185e-06, + "loss": 1.4571, + "step": 21399 + }, + { + "epoch": 3.89, + "learning_rate": 1.8341827448125227e-06, + "loss": 1.2761, + "step": 21400 + }, + { + "epoch": 3.89, + "learning_rate": 1.8341281397888607e-06, + "loss": 0.1058, + "step": 21401 + }, + { + "epoch": 3.89, + "learning_rate": 1.8340735347651986e-06, + "loss": 2.1755, + "step": 21402 + }, + { + "epoch": 3.89, + "learning_rate": 1.8340189297415361e-06, + "loss": 0.1234, + "step": 21403 + }, + { + "epoch": 3.89, + "learning_rate": 1.833964324717874e-06, + "loss": 0.9088, + "step": 21404 + }, + { + "epoch": 3.89, + "learning_rate": 1.833909719694212e-06, + "loss": 0.4756, + "step": 21405 + }, + { + "epoch": 3.89, + "learning_rate": 1.8338551146705498e-06, + "loss": 0.0907, + "step": 21406 + }, + { + "epoch": 3.89, + "learning_rate": 1.8338005096468877e-06, + "loss": 0.2213, + "step": 21407 + }, + { + "epoch": 3.89, + "learning_rate": 1.8337459046232252e-06, + "loss": 0.9054, + "step": 21408 + }, + { + "epoch": 3.89, + "learning_rate": 1.8336912995995632e-06, + "loss": 0.0073, + "step": 21409 + }, + { + "epoch": 3.89, + "learning_rate": 1.8336366945759011e-06, + "loss": 0.4451, + "step": 21410 + }, + { + "epoch": 3.89, + "learning_rate": 1.8335820895522386e-06, + "loss": 1.6648, + "step": 21411 + }, + { + "epoch": 3.89, + "learning_rate": 1.8335274845285766e-06, + "loss": 0.2882, + "step": 21412 + }, + { + "epoch": 3.89, + "learning_rate": 1.8334728795049145e-06, + "loss": 0.1481, + "step": 21413 + }, + { + "epoch": 3.89, + "learning_rate": 1.8334182744812523e-06, + "loss": 1.9521, + "step": 21414 + }, + { + "epoch": 3.89, + "learning_rate": 1.8333636694575902e-06, + "loss": 0.7839, + "step": 21415 + }, + { + "epoch": 3.89, + "learning_rate": 1.8333090644339281e-06, + "loss": 0.2468, + "step": 21416 + }, + { + "epoch": 3.89, + "learning_rate": 1.8332544594102657e-06, + "loss": 0.1555, + "step": 21417 + }, + { + "epoch": 3.89, + "learning_rate": 1.8331998543866036e-06, + "loss": 0.5169, + "step": 21418 + }, + { + "epoch": 3.9, + "learning_rate": 1.8331452493629415e-06, + "loss": 1.1189, + "step": 21419 + }, + { + "epoch": 3.9, + "learning_rate": 1.8330906443392793e-06, + "loss": 0.2007, + "step": 21420 + }, + { + "epoch": 3.9, + "learning_rate": 1.8330360393156172e-06, + "loss": 0.0115, + "step": 21421 + }, + { + "epoch": 3.9, + "learning_rate": 1.8329814342919547e-06, + "loss": 0.6145, + "step": 21422 + }, + { + "epoch": 3.9, + "learning_rate": 1.8329268292682927e-06, + "loss": 0.1318, + "step": 21423 + }, + { + "epoch": 3.9, + "learning_rate": 1.8328722242446306e-06, + "loss": 0.2628, + "step": 21424 + }, + { + "epoch": 3.9, + "learning_rate": 1.8328176192209681e-06, + "loss": 0.9671, + "step": 21425 + }, + { + "epoch": 3.9, + "learning_rate": 1.832763014197306e-06, + "loss": 1.6028, + "step": 21426 + }, + { + "epoch": 3.9, + "learning_rate": 1.832708409173644e-06, + "loss": 0.1209, + "step": 21427 + }, + { + "epoch": 3.9, + "learning_rate": 1.8326538041499818e-06, + "loss": 0.1767, + "step": 21428 + }, + { + "epoch": 3.9, + "learning_rate": 1.8325991991263197e-06, + "loss": 0.5752, + "step": 21429 + }, + { + "epoch": 3.9, + "learning_rate": 1.8325445941026577e-06, + "loss": 2.0439, + "step": 21430 + }, + { + "epoch": 3.9, + "learning_rate": 1.8324899890789952e-06, + "loss": 0.7846, + "step": 21431 + }, + { + "epoch": 3.9, + "learning_rate": 1.8324353840553331e-06, + "loss": 0.0106, + "step": 21432 + }, + { + "epoch": 3.9, + "learning_rate": 1.832380779031671e-06, + "loss": 0.8592, + "step": 21433 + }, + { + "epoch": 3.9, + "learning_rate": 1.8323261740080088e-06, + "loss": 0.0401, + "step": 21434 + }, + { + "epoch": 3.9, + "learning_rate": 1.8322715689843467e-06, + "loss": 1.5776, + "step": 21435 + }, + { + "epoch": 3.9, + "learning_rate": 1.8322169639606843e-06, + "loss": 0.044, + "step": 21436 + }, + { + "epoch": 3.9, + "learning_rate": 1.8321623589370222e-06, + "loss": 0.9189, + "step": 21437 + }, + { + "epoch": 3.9, + "learning_rate": 1.8321077539133601e-06, + "loss": 1.7053, + "step": 21438 + }, + { + "epoch": 3.9, + "learning_rate": 1.8320531488896979e-06, + "loss": 0.369, + "step": 21439 + }, + { + "epoch": 3.9, + "learning_rate": 1.8319985438660358e-06, + "loss": 0.2065, + "step": 21440 + }, + { + "epoch": 3.9, + "learning_rate": 1.8319439388423738e-06, + "loss": 0.2554, + "step": 21441 + }, + { + "epoch": 3.9, + "learning_rate": 1.8318893338187113e-06, + "loss": 1.1328, + "step": 21442 + }, + { + "epoch": 3.9, + "learning_rate": 1.8318347287950492e-06, + "loss": 0.8198, + "step": 21443 + }, + { + "epoch": 3.9, + "learning_rate": 1.8317801237713872e-06, + "loss": 1.5092, + "step": 21444 + }, + { + "epoch": 3.9, + "learning_rate": 1.8317255187477247e-06, + "loss": 0.1003, + "step": 21445 + }, + { + "epoch": 3.9, + "learning_rate": 1.8316709137240626e-06, + "loss": 0.0188, + "step": 21446 + }, + { + "epoch": 3.9, + "learning_rate": 1.8316163087004006e-06, + "loss": 0.7994, + "step": 21447 + }, + { + "epoch": 3.9, + "learning_rate": 1.8315617036767383e-06, + "loss": 0.0214, + "step": 21448 + }, + { + "epoch": 3.9, + "learning_rate": 1.8315070986530762e-06, + "loss": 1.3589, + "step": 21449 + }, + { + "epoch": 3.9, + "learning_rate": 1.8314524936294138e-06, + "loss": 0.0346, + "step": 21450 + }, + { + "epoch": 3.9, + "learning_rate": 1.8313978886057517e-06, + "loss": 0.9157, + "step": 21451 + }, + { + "epoch": 3.9, + "learning_rate": 1.8313432835820897e-06, + "loss": 0.2564, + "step": 21452 + }, + { + "epoch": 3.9, + "learning_rate": 1.8312886785584274e-06, + "loss": 0.3639, + "step": 21453 + }, + { + "epoch": 3.9, + "learning_rate": 1.8312340735347653e-06, + "loss": 0.9086, + "step": 21454 + }, + { + "epoch": 3.9, + "learning_rate": 1.8311794685111033e-06, + "loss": 0.5624, + "step": 21455 + }, + { + "epoch": 3.9, + "learning_rate": 1.8311248634874408e-06, + "loss": 0.8268, + "step": 21456 + }, + { + "epoch": 3.9, + "learning_rate": 1.8310702584637787e-06, + "loss": 1.123, + "step": 21457 + }, + { + "epoch": 3.9, + "learning_rate": 1.8310156534401167e-06, + "loss": 1.3241, + "step": 21458 + }, + { + "epoch": 3.9, + "learning_rate": 1.8309610484164542e-06, + "loss": 0.8487, + "step": 21459 + }, + { + "epoch": 3.9, + "learning_rate": 1.8309064433927921e-06, + "loss": 2.3568, + "step": 21460 + }, + { + "epoch": 3.9, + "learning_rate": 1.83085183836913e-06, + "loss": 1.7285, + "step": 21461 + }, + { + "epoch": 3.9, + "learning_rate": 1.8307972333454678e-06, + "loss": 0.3103, + "step": 21462 + }, + { + "epoch": 3.9, + "learning_rate": 1.8307426283218058e-06, + "loss": 0.5314, + "step": 21463 + }, + { + "epoch": 3.9, + "learning_rate": 1.8306880232981433e-06, + "loss": 0.6166, + "step": 21464 + }, + { + "epoch": 3.9, + "learning_rate": 1.8306334182744812e-06, + "loss": 0.4048, + "step": 21465 + }, + { + "epoch": 3.9, + "learning_rate": 1.8305788132508192e-06, + "loss": 0.9888, + "step": 21466 + }, + { + "epoch": 3.9, + "learning_rate": 1.830524208227157e-06, + "loss": 0.5008, + "step": 21467 + }, + { + "epoch": 3.9, + "learning_rate": 1.8304696032034948e-06, + "loss": 1.6743, + "step": 21468 + }, + { + "epoch": 3.9, + "learning_rate": 1.8304149981798328e-06, + "loss": 0.1478, + "step": 21469 + }, + { + "epoch": 3.9, + "learning_rate": 1.8303603931561703e-06, + "loss": 0.3124, + "step": 21470 + }, + { + "epoch": 3.9, + "learning_rate": 1.8303057881325082e-06, + "loss": 0.9485, + "step": 21471 + }, + { + "epoch": 3.9, + "learning_rate": 1.8302511831088462e-06, + "loss": 2.0782, + "step": 21472 + }, + { + "epoch": 3.9, + "learning_rate": 1.8301965780851837e-06, + "loss": 0.0802, + "step": 21473 + }, + { + "epoch": 3.91, + "learning_rate": 1.8301419730615217e-06, + "loss": 0.2094, + "step": 21474 + }, + { + "epoch": 3.91, + "learning_rate": 1.8300873680378594e-06, + "loss": 0.1781, + "step": 21475 + }, + { + "epoch": 3.91, + "learning_rate": 1.8300327630141973e-06, + "loss": 1.6645, + "step": 21476 + }, + { + "epoch": 3.91, + "learning_rate": 1.8299781579905353e-06, + "loss": 1.9278, + "step": 21477 + }, + { + "epoch": 3.91, + "learning_rate": 1.8299235529668728e-06, + "loss": 0.9814, + "step": 21478 + }, + { + "epoch": 3.91, + "learning_rate": 1.8298689479432107e-06, + "loss": 0.5694, + "step": 21479 + }, + { + "epoch": 3.91, + "learning_rate": 1.8298143429195487e-06, + "loss": 1.5456, + "step": 21480 + }, + { + "epoch": 3.91, + "learning_rate": 1.8297597378958864e-06, + "loss": 0.5115, + "step": 21481 + }, + { + "epoch": 3.91, + "learning_rate": 1.8297051328722244e-06, + "loss": 0.0233, + "step": 21482 + }, + { + "epoch": 3.91, + "learning_rate": 1.8296505278485623e-06, + "loss": 0.844, + "step": 21483 + }, + { + "epoch": 3.91, + "learning_rate": 1.8295959228248998e-06, + "loss": 0.0107, + "step": 21484 + }, + { + "epoch": 3.91, + "learning_rate": 1.8295413178012378e-06, + "loss": 0.4746, + "step": 21485 + }, + { + "epoch": 3.91, + "learning_rate": 1.8294867127775757e-06, + "loss": 1.5054, + "step": 21486 + }, + { + "epoch": 3.91, + "learning_rate": 1.8294321077539134e-06, + "loss": 0.257, + "step": 21487 + }, + { + "epoch": 3.91, + "learning_rate": 1.8293775027302514e-06, + "loss": 0.6762, + "step": 21488 + }, + { + "epoch": 3.91, + "learning_rate": 1.829322897706589e-06, + "loss": 0.1482, + "step": 21489 + }, + { + "epoch": 3.91, + "learning_rate": 1.8292682926829268e-06, + "loss": 2.8816, + "step": 21490 + }, + { + "epoch": 3.91, + "learning_rate": 1.8292136876592648e-06, + "loss": 0.1917, + "step": 21491 + }, + { + "epoch": 3.91, + "learning_rate": 1.8291590826356023e-06, + "loss": 0.5356, + "step": 21492 + }, + { + "epoch": 3.91, + "learning_rate": 1.8291044776119403e-06, + "loss": 0.188, + "step": 21493 + }, + { + "epoch": 3.91, + "learning_rate": 1.8290498725882782e-06, + "loss": 0.4027, + "step": 21494 + }, + { + "epoch": 3.91, + "learning_rate": 1.828995267564616e-06, + "loss": 3.76, + "step": 21495 + }, + { + "epoch": 3.91, + "learning_rate": 1.8289406625409539e-06, + "loss": 0.8457, + "step": 21496 + }, + { + "epoch": 3.91, + "learning_rate": 1.8288860575172918e-06, + "loss": 0.9481, + "step": 21497 + }, + { + "epoch": 3.91, + "learning_rate": 1.8288314524936293e-06, + "loss": 0.622, + "step": 21498 + }, + { + "epoch": 3.91, + "learning_rate": 1.8287768474699673e-06, + "loss": 0.0252, + "step": 21499 + }, + { + "epoch": 3.91, + "learning_rate": 1.8287222424463052e-06, + "loss": 0.043, + "step": 21500 + }, + { + "epoch": 3.91, + "learning_rate": 1.828667637422643e-06, + "loss": 2.8265, + "step": 21501 + }, + { + "epoch": 3.91, + "learning_rate": 1.828613032398981e-06, + "loss": 1.9468, + "step": 21502 + }, + { + "epoch": 3.91, + "learning_rate": 1.8285584273753184e-06, + "loss": 0.9329, + "step": 21503 + }, + { + "epoch": 3.91, + "learning_rate": 1.8285038223516564e-06, + "loss": 0.3429, + "step": 21504 + }, + { + "epoch": 3.91, + "learning_rate": 1.8284492173279943e-06, + "loss": 0.0256, + "step": 21505 + }, + { + "epoch": 3.91, + "learning_rate": 1.8283946123043318e-06, + "loss": 1.2051, + "step": 21506 + }, + { + "epoch": 3.91, + "learning_rate": 1.8283400072806698e-06, + "loss": 0.3094, + "step": 21507 + }, + { + "epoch": 3.91, + "learning_rate": 1.8282854022570077e-06, + "loss": 0.0229, + "step": 21508 + }, + { + "epoch": 3.91, + "learning_rate": 1.8282307972333454e-06, + "loss": 0.4053, + "step": 21509 + }, + { + "epoch": 3.91, + "learning_rate": 1.8281761922096834e-06, + "loss": 0.1227, + "step": 21510 + }, + { + "epoch": 3.91, + "learning_rate": 1.8281215871860213e-06, + "loss": 0.0169, + "step": 21511 + }, + { + "epoch": 3.91, + "learning_rate": 1.8280669821623588e-06, + "loss": 0.4914, + "step": 21512 + }, + { + "epoch": 3.91, + "learning_rate": 1.8280123771386968e-06, + "loss": 1.2737, + "step": 21513 + }, + { + "epoch": 3.91, + "learning_rate": 1.8279577721150347e-06, + "loss": 0.1274, + "step": 21514 + }, + { + "epoch": 3.91, + "learning_rate": 1.8279031670913725e-06, + "loss": 0.5253, + "step": 21515 + }, + { + "epoch": 3.91, + "learning_rate": 1.8278485620677104e-06, + "loss": 0.0428, + "step": 21516 + }, + { + "epoch": 3.91, + "learning_rate": 1.827793957044048e-06, + "loss": 0.4292, + "step": 21517 + }, + { + "epoch": 3.91, + "learning_rate": 1.8277393520203859e-06, + "loss": 0.1523, + "step": 21518 + }, + { + "epoch": 3.91, + "learning_rate": 1.8276847469967238e-06, + "loss": 0.1934, + "step": 21519 + }, + { + "epoch": 3.91, + "learning_rate": 1.8276301419730615e-06, + "loss": 0.3983, + "step": 21520 + }, + { + "epoch": 3.91, + "learning_rate": 1.8275755369493993e-06, + "loss": 0.0155, + "step": 21521 + }, + { + "epoch": 3.91, + "learning_rate": 1.8275209319257372e-06, + "loss": 2.3078, + "step": 21522 + }, + { + "epoch": 3.91, + "learning_rate": 1.827466326902075e-06, + "loss": 0.0502, + "step": 21523 + }, + { + "epoch": 3.91, + "learning_rate": 1.827411721878413e-06, + "loss": 0.8224, + "step": 21524 + }, + { + "epoch": 3.91, + "learning_rate": 1.8273571168547508e-06, + "loss": 0.2789, + "step": 21525 + }, + { + "epoch": 3.91, + "learning_rate": 1.8273025118310884e-06, + "loss": 0.7119, + "step": 21526 + }, + { + "epoch": 3.91, + "learning_rate": 1.8272479068074263e-06, + "loss": 2.0483, + "step": 21527 + }, + { + "epoch": 3.91, + "learning_rate": 1.8271933017837642e-06, + "loss": 1.0043, + "step": 21528 + }, + { + "epoch": 3.92, + "learning_rate": 1.827138696760102e-06, + "loss": 0.6473, + "step": 21529 + }, + { + "epoch": 3.92, + "learning_rate": 1.82708409173644e-06, + "loss": 1.8733, + "step": 21530 + }, + { + "epoch": 3.92, + "learning_rate": 1.8270294867127774e-06, + "loss": 0.1149, + "step": 21531 + }, + { + "epoch": 3.92, + "learning_rate": 1.8269748816891154e-06, + "loss": 0.0165, + "step": 21532 + }, + { + "epoch": 3.92, + "learning_rate": 1.8269202766654533e-06, + "loss": 0.2295, + "step": 21533 + }, + { + "epoch": 3.92, + "learning_rate": 1.826865671641791e-06, + "loss": 0.1251, + "step": 21534 + }, + { + "epoch": 3.92, + "learning_rate": 1.826811066618129e-06, + "loss": 1.0922, + "step": 21535 + }, + { + "epoch": 3.92, + "learning_rate": 1.826756461594467e-06, + "loss": 0.6463, + "step": 21536 + }, + { + "epoch": 3.92, + "learning_rate": 1.8267018565708045e-06, + "loss": 0.0738, + "step": 21537 + }, + { + "epoch": 3.92, + "learning_rate": 1.8266472515471424e-06, + "loss": 0.5931, + "step": 21538 + }, + { + "epoch": 3.92, + "learning_rate": 1.8265926465234804e-06, + "loss": 0.033, + "step": 21539 + }, + { + "epoch": 3.92, + "learning_rate": 1.8265380414998179e-06, + "loss": 1.5184, + "step": 21540 + }, + { + "epoch": 3.92, + "learning_rate": 1.8264834364761558e-06, + "loss": 0.7207, + "step": 21541 + }, + { + "epoch": 3.92, + "learning_rate": 1.8264288314524938e-06, + "loss": 0.4664, + "step": 21542 + }, + { + "epoch": 3.92, + "learning_rate": 1.8263742264288315e-06, + "loss": 2.4919, + "step": 21543 + }, + { + "epoch": 3.92, + "learning_rate": 1.8263196214051694e-06, + "loss": 2.2583, + "step": 21544 + }, + { + "epoch": 3.92, + "learning_rate": 1.826265016381507e-06, + "loss": 0.1223, + "step": 21545 + }, + { + "epoch": 3.92, + "learning_rate": 1.826210411357845e-06, + "loss": 0.4109, + "step": 21546 + }, + { + "epoch": 3.92, + "learning_rate": 1.8261558063341828e-06, + "loss": 0.6445, + "step": 21547 + }, + { + "epoch": 3.92, + "learning_rate": 1.8261012013105206e-06, + "loss": 0.2416, + "step": 21548 + }, + { + "epoch": 3.92, + "learning_rate": 1.8260465962868585e-06, + "loss": 0.2177, + "step": 21549 + }, + { + "epoch": 3.92, + "learning_rate": 1.8259919912631965e-06, + "loss": 0.2106, + "step": 21550 + }, + { + "epoch": 3.92, + "learning_rate": 1.825937386239534e-06, + "loss": 0.0251, + "step": 21551 + }, + { + "epoch": 3.92, + "learning_rate": 1.825882781215872e-06, + "loss": 0.4169, + "step": 21552 + }, + { + "epoch": 3.92, + "learning_rate": 1.8258281761922099e-06, + "loss": 0.0299, + "step": 21553 + }, + { + "epoch": 3.92, + "learning_rate": 1.8257735711685474e-06, + "loss": 0.6398, + "step": 21554 + }, + { + "epoch": 3.92, + "learning_rate": 1.8257189661448853e-06, + "loss": 0.9948, + "step": 21555 + }, + { + "epoch": 3.92, + "learning_rate": 1.825664361121223e-06, + "loss": 0.4406, + "step": 21556 + }, + { + "epoch": 3.92, + "learning_rate": 1.825609756097561e-06, + "loss": 0.0742, + "step": 21557 + }, + { + "epoch": 3.92, + "learning_rate": 1.825555151073899e-06, + "loss": 1.9581, + "step": 21558 + }, + { + "epoch": 3.92, + "learning_rate": 1.8255005460502365e-06, + "loss": 0.4276, + "step": 21559 + }, + { + "epoch": 3.92, + "learning_rate": 1.8254459410265744e-06, + "loss": 0.6413, + "step": 21560 + }, + { + "epoch": 3.92, + "learning_rate": 1.8253913360029124e-06, + "loss": 0.6383, + "step": 21561 + }, + { + "epoch": 3.92, + "learning_rate": 1.82533673097925e-06, + "loss": 0.239, + "step": 21562 + }, + { + "epoch": 3.92, + "learning_rate": 1.825282125955588e-06, + "loss": 1.8552, + "step": 21563 + }, + { + "epoch": 3.92, + "learning_rate": 1.825227520931926e-06, + "loss": 0.4494, + "step": 21564 + }, + { + "epoch": 3.92, + "learning_rate": 1.8251729159082635e-06, + "loss": 1.0842, + "step": 21565 + }, + { + "epoch": 3.92, + "learning_rate": 1.8251183108846014e-06, + "loss": 0.0307, + "step": 21566 + }, + { + "epoch": 3.92, + "learning_rate": 1.8250637058609394e-06, + "loss": 2.1555, + "step": 21567 + }, + { + "epoch": 3.92, + "learning_rate": 1.8250091008372771e-06, + "loss": 0.2724, + "step": 21568 + }, + { + "epoch": 3.92, + "learning_rate": 1.824954495813615e-06, + "loss": 0.6018, + "step": 21569 + }, + { + "epoch": 3.92, + "learning_rate": 1.8248998907899526e-06, + "loss": 0.8475, + "step": 21570 + }, + { + "epoch": 3.92, + "learning_rate": 1.8248452857662905e-06, + "loss": 3.5083, + "step": 21571 + }, + { + "epoch": 3.92, + "learning_rate": 1.8247906807426285e-06, + "loss": 0.3029, + "step": 21572 + }, + { + "epoch": 3.92, + "learning_rate": 1.824736075718966e-06, + "loss": 3.4219, + "step": 21573 + }, + { + "epoch": 3.92, + "learning_rate": 1.824681470695304e-06, + "loss": 0.0057, + "step": 21574 + }, + { + "epoch": 3.92, + "learning_rate": 1.8246268656716419e-06, + "loss": 0.0389, + "step": 21575 + }, + { + "epoch": 3.92, + "learning_rate": 1.8245722606479796e-06, + "loss": 0.3822, + "step": 21576 + }, + { + "epoch": 3.92, + "learning_rate": 1.8245176556243175e-06, + "loss": 0.7384, + "step": 21577 + }, + { + "epoch": 3.92, + "learning_rate": 1.8244630506006555e-06, + "loss": 0.8159, + "step": 21578 + }, + { + "epoch": 3.92, + "learning_rate": 1.824408445576993e-06, + "loss": 0.735, + "step": 21579 + }, + { + "epoch": 3.92, + "learning_rate": 1.824353840553331e-06, + "loss": 0.2729, + "step": 21580 + }, + { + "epoch": 3.92, + "learning_rate": 1.8242992355296689e-06, + "loss": 2.4598, + "step": 21581 + }, + { + "epoch": 3.92, + "learning_rate": 1.8242446305060066e-06, + "loss": 0.2961, + "step": 21582 + }, + { + "epoch": 3.92, + "learning_rate": 1.8241900254823446e-06, + "loss": 0.112, + "step": 21583 + }, + { + "epoch": 3.93, + "learning_rate": 1.824135420458682e-06, + "loss": 0.2001, + "step": 21584 + }, + { + "epoch": 3.93, + "learning_rate": 1.82408081543502e-06, + "loss": 1.0059, + "step": 21585 + }, + { + "epoch": 3.93, + "learning_rate": 1.824026210411358e-06, + "loss": 0.9413, + "step": 21586 + }, + { + "epoch": 3.93, + "learning_rate": 1.8239716053876955e-06, + "loss": 0.1225, + "step": 21587 + }, + { + "epoch": 3.93, + "learning_rate": 1.8239170003640334e-06, + "loss": 0.0216, + "step": 21588 + }, + { + "epoch": 3.93, + "learning_rate": 1.8238623953403714e-06, + "loss": 0.63, + "step": 21589 + }, + { + "epoch": 3.93, + "learning_rate": 1.8238077903167091e-06, + "loss": 0.0772, + "step": 21590 + }, + { + "epoch": 3.93, + "learning_rate": 1.823753185293047e-06, + "loss": 1.6306, + "step": 21591 + }, + { + "epoch": 3.93, + "learning_rate": 1.823698580269385e-06, + "loss": 0.3112, + "step": 21592 + }, + { + "epoch": 3.93, + "learning_rate": 1.8236439752457225e-06, + "loss": 0.3948, + "step": 21593 + }, + { + "epoch": 3.93, + "learning_rate": 1.8235893702220605e-06, + "loss": 0.2414, + "step": 21594 + }, + { + "epoch": 3.93, + "learning_rate": 1.8235347651983984e-06, + "loss": 0.6285, + "step": 21595 + }, + { + "epoch": 3.93, + "learning_rate": 1.8234801601747361e-06, + "loss": 0.583, + "step": 21596 + }, + { + "epoch": 3.93, + "learning_rate": 1.823425555151074e-06, + "loss": 0.213, + "step": 21597 + }, + { + "epoch": 3.93, + "learning_rate": 1.8233709501274116e-06, + "loss": 0.8, + "step": 21598 + }, + { + "epoch": 3.93, + "learning_rate": 1.8233163451037495e-06, + "loss": 2.472, + "step": 21599 + }, + { + "epoch": 3.93, + "learning_rate": 1.8232617400800875e-06, + "loss": 1.5155, + "step": 21600 + }, + { + "epoch": 3.93, + "learning_rate": 1.823207135056425e-06, + "loss": 0.7822, + "step": 21601 + }, + { + "epoch": 3.93, + "learning_rate": 1.823152530032763e-06, + "loss": 0.4829, + "step": 21602 + }, + { + "epoch": 3.93, + "learning_rate": 1.823097925009101e-06, + "loss": 0.8204, + "step": 21603 + }, + { + "epoch": 3.93, + "learning_rate": 1.8230433199854386e-06, + "loss": 0.7464, + "step": 21604 + }, + { + "epoch": 3.93, + "learning_rate": 1.8229887149617766e-06, + "loss": 3.0174, + "step": 21605 + }, + { + "epoch": 3.93, + "learning_rate": 1.8229341099381145e-06, + "loss": 0.7919, + "step": 21606 + }, + { + "epoch": 3.93, + "learning_rate": 1.822879504914452e-06, + "loss": 1.4018, + "step": 21607 + }, + { + "epoch": 3.93, + "learning_rate": 1.82282489989079e-06, + "loss": 0.2152, + "step": 21608 + }, + { + "epoch": 3.93, + "learning_rate": 1.822770294867128e-06, + "loss": 1.5133, + "step": 21609 + }, + { + "epoch": 3.93, + "learning_rate": 1.8227156898434657e-06, + "loss": 1.6929, + "step": 21610 + }, + { + "epoch": 3.93, + "learning_rate": 1.8226610848198036e-06, + "loss": 0.2984, + "step": 21611 + }, + { + "epoch": 3.93, + "learning_rate": 1.8226064797961411e-06, + "loss": 1.3056, + "step": 21612 + }, + { + "epoch": 3.93, + "learning_rate": 1.822551874772479e-06, + "loss": 0.9844, + "step": 21613 + }, + { + "epoch": 3.93, + "learning_rate": 1.822497269748817e-06, + "loss": 1.5567, + "step": 21614 + }, + { + "epoch": 3.93, + "learning_rate": 1.8224426647251547e-06, + "loss": 0.3256, + "step": 21615 + }, + { + "epoch": 3.93, + "learning_rate": 1.8223880597014927e-06, + "loss": 1.5598, + "step": 21616 + }, + { + "epoch": 3.93, + "learning_rate": 1.8223334546778306e-06, + "loss": 0.2338, + "step": 21617 + }, + { + "epoch": 3.93, + "learning_rate": 1.8222788496541681e-06, + "loss": 0.1796, + "step": 21618 + }, + { + "epoch": 3.93, + "learning_rate": 1.822224244630506e-06, + "loss": 1.0597, + "step": 21619 + }, + { + "epoch": 3.93, + "learning_rate": 1.822169639606844e-06, + "loss": 2.5487, + "step": 21620 + }, + { + "epoch": 3.93, + "learning_rate": 1.8221150345831815e-06, + "loss": 0.41, + "step": 21621 + }, + { + "epoch": 3.93, + "learning_rate": 1.8220604295595195e-06, + "loss": 0.774, + "step": 21622 + }, + { + "epoch": 3.93, + "learning_rate": 1.8220058245358572e-06, + "loss": 0.5077, + "step": 21623 + }, + { + "epoch": 3.93, + "learning_rate": 1.8219512195121952e-06, + "loss": 0.0367, + "step": 21624 + }, + { + "epoch": 3.93, + "learning_rate": 1.8218966144885331e-06, + "loss": 1.0498, + "step": 21625 + }, + { + "epoch": 3.93, + "learning_rate": 1.8218420094648706e-06, + "loss": 0.4541, + "step": 21626 + }, + { + "epoch": 3.93, + "learning_rate": 1.8217874044412086e-06, + "loss": 3.2823, + "step": 21627 + }, + { + "epoch": 3.93, + "learning_rate": 1.8217327994175465e-06, + "loss": 0.7128, + "step": 21628 + }, + { + "epoch": 3.93, + "learning_rate": 1.8216781943938842e-06, + "loss": 0.2674, + "step": 21629 + }, + { + "epoch": 3.93, + "learning_rate": 1.8216235893702222e-06, + "loss": 0.0285, + "step": 21630 + }, + { + "epoch": 3.93, + "learning_rate": 1.8215689843465601e-06, + "loss": 1.2816, + "step": 21631 + }, + { + "epoch": 3.93, + "learning_rate": 1.8215143793228977e-06, + "loss": 0.957, + "step": 21632 + }, + { + "epoch": 3.93, + "learning_rate": 1.8214597742992356e-06, + "loss": 1.0276, + "step": 21633 + }, + { + "epoch": 3.93, + "learning_rate": 1.8214051692755735e-06, + "loss": 0.0179, + "step": 21634 + }, + { + "epoch": 3.93, + "learning_rate": 1.821350564251911e-06, + "loss": 0.8438, + "step": 21635 + }, + { + "epoch": 3.93, + "learning_rate": 1.821295959228249e-06, + "loss": 1.4989, + "step": 21636 + }, + { + "epoch": 3.93, + "learning_rate": 1.8212413542045867e-06, + "loss": 0.0286, + "step": 21637 + }, + { + "epoch": 3.93, + "learning_rate": 1.8211867491809247e-06, + "loss": 0.017, + "step": 21638 + }, + { + "epoch": 3.94, + "learning_rate": 1.8211321441572626e-06, + "loss": 0.5944, + "step": 21639 + }, + { + "epoch": 3.94, + "learning_rate": 1.8210775391336001e-06, + "loss": 0.1542, + "step": 21640 + }, + { + "epoch": 3.94, + "learning_rate": 1.821022934109938e-06, + "loss": 0.5086, + "step": 21641 + }, + { + "epoch": 3.94, + "learning_rate": 1.820968329086276e-06, + "loss": 0.2403, + "step": 21642 + }, + { + "epoch": 3.94, + "learning_rate": 1.8209137240626138e-06, + "loss": 0.0099, + "step": 21643 + }, + { + "epoch": 3.94, + "learning_rate": 1.8208591190389517e-06, + "loss": 0.1234, + "step": 21644 + }, + { + "epoch": 3.94, + "learning_rate": 1.8208045140152896e-06, + "loss": 0.9527, + "step": 21645 + }, + { + "epoch": 3.94, + "learning_rate": 1.8207499089916272e-06, + "loss": 0.8903, + "step": 21646 + }, + { + "epoch": 3.94, + "learning_rate": 1.8206953039679651e-06, + "loss": 0.0412, + "step": 21647 + }, + { + "epoch": 3.94, + "learning_rate": 1.820640698944303e-06, + "loss": 0.2145, + "step": 21648 + }, + { + "epoch": 3.94, + "learning_rate": 1.8205860939206408e-06, + "loss": 0.0251, + "step": 21649 + }, + { + "epoch": 3.94, + "learning_rate": 1.8205314888969787e-06, + "loss": 0.3073, + "step": 21650 + }, + { + "epoch": 3.94, + "learning_rate": 1.8204768838733162e-06, + "loss": 1.4526, + "step": 21651 + }, + { + "epoch": 3.94, + "learning_rate": 1.8204222788496542e-06, + "loss": 0.3989, + "step": 21652 + }, + { + "epoch": 3.94, + "learning_rate": 1.8203676738259921e-06, + "loss": 1.9455, + "step": 21653 + }, + { + "epoch": 3.94, + "learning_rate": 1.8203130688023297e-06, + "loss": 0.0264, + "step": 21654 + }, + { + "epoch": 3.94, + "learning_rate": 1.8202584637786676e-06, + "loss": 2.5432, + "step": 21655 + }, + { + "epoch": 3.94, + "learning_rate": 1.8202038587550055e-06, + "loss": 1.3888, + "step": 21656 + }, + { + "epoch": 3.94, + "learning_rate": 1.8201492537313433e-06, + "loss": 0.0486, + "step": 21657 + }, + { + "epoch": 3.94, + "learning_rate": 1.8200946487076812e-06, + "loss": 0.758, + "step": 21658 + }, + { + "epoch": 3.94, + "learning_rate": 1.8200400436840192e-06, + "loss": 0.1934, + "step": 21659 + }, + { + "epoch": 3.94, + "learning_rate": 1.8199854386603567e-06, + "loss": 0.2294, + "step": 21660 + }, + { + "epoch": 3.94, + "learning_rate": 1.8199308336366946e-06, + "loss": 0.3743, + "step": 21661 + }, + { + "epoch": 3.94, + "learning_rate": 1.8198762286130326e-06, + "loss": 0.489, + "step": 21662 + }, + { + "epoch": 3.94, + "learning_rate": 1.8198216235893703e-06, + "loss": 0.143, + "step": 21663 + }, + { + "epoch": 3.94, + "learning_rate": 1.8197670185657082e-06, + "loss": 0.4564, + "step": 21664 + }, + { + "epoch": 3.94, + "learning_rate": 1.8197124135420458e-06, + "loss": 2.904, + "step": 21665 + }, + { + "epoch": 3.94, + "learning_rate": 1.8196578085183837e-06, + "loss": 0.1215, + "step": 21666 + }, + { + "epoch": 3.94, + "learning_rate": 1.8196032034947216e-06, + "loss": 0.6741, + "step": 21667 + }, + { + "epoch": 3.94, + "learning_rate": 1.8195485984710592e-06, + "loss": 0.317, + "step": 21668 + }, + { + "epoch": 3.94, + "learning_rate": 1.8194939934473971e-06, + "loss": 0.193, + "step": 21669 + }, + { + "epoch": 3.94, + "learning_rate": 1.819439388423735e-06, + "loss": 0.1706, + "step": 21670 + }, + { + "epoch": 3.94, + "learning_rate": 1.8193847834000728e-06, + "loss": 0.2, + "step": 21671 + }, + { + "epoch": 3.94, + "learning_rate": 1.8193301783764107e-06, + "loss": 0.0178, + "step": 21672 + }, + { + "epoch": 3.94, + "learning_rate": 1.8192755733527487e-06, + "loss": 1.8206, + "step": 21673 + }, + { + "epoch": 3.94, + "learning_rate": 1.8192209683290862e-06, + "loss": 2.4157, + "step": 21674 + }, + { + "epoch": 3.94, + "learning_rate": 1.8191663633054241e-06, + "loss": 0.9191, + "step": 21675 + }, + { + "epoch": 3.94, + "learning_rate": 1.819111758281762e-06, + "loss": 0.1369, + "step": 21676 + }, + { + "epoch": 3.94, + "learning_rate": 1.8190571532580998e-06, + "loss": 0.1033, + "step": 21677 + }, + { + "epoch": 3.94, + "learning_rate": 1.8190025482344378e-06, + "loss": 0.8883, + "step": 21678 + }, + { + "epoch": 3.94, + "learning_rate": 1.8189479432107753e-06, + "loss": 0.6252, + "step": 21679 + }, + { + "epoch": 3.94, + "learning_rate": 1.8188933381871132e-06, + "loss": 0.1712, + "step": 21680 + }, + { + "epoch": 3.94, + "learning_rate": 1.8188387331634512e-06, + "loss": 0.8693, + "step": 21681 + }, + { + "epoch": 3.94, + "learning_rate": 1.8187841281397887e-06, + "loss": 0.8562, + "step": 21682 + }, + { + "epoch": 3.94, + "learning_rate": 1.8187295231161266e-06, + "loss": 1.1134, + "step": 21683 + }, + { + "epoch": 3.94, + "learning_rate": 1.8186749180924646e-06, + "loss": 0.6244, + "step": 21684 + }, + { + "epoch": 3.94, + "learning_rate": 1.8186203130688023e-06, + "loss": 0.3207, + "step": 21685 + }, + { + "epoch": 3.94, + "learning_rate": 1.8185657080451402e-06, + "loss": 0.2627, + "step": 21686 + }, + { + "epoch": 3.94, + "learning_rate": 1.8185111030214782e-06, + "loss": 0.8432, + "step": 21687 + }, + { + "epoch": 3.94, + "learning_rate": 1.8184564979978157e-06, + "loss": 0.2242, + "step": 21688 + }, + { + "epoch": 3.94, + "learning_rate": 1.8184018929741537e-06, + "loss": 1.3534, + "step": 21689 + }, + { + "epoch": 3.94, + "learning_rate": 1.8183472879504916e-06, + "loss": 0.1266, + "step": 21690 + }, + { + "epoch": 3.94, + "learning_rate": 1.8182926829268293e-06, + "loss": 1.1185, + "step": 21691 + }, + { + "epoch": 3.94, + "learning_rate": 1.8182380779031673e-06, + "loss": 1.1571, + "step": 21692 + }, + { + "epoch": 3.94, + "learning_rate": 1.8181834728795048e-06, + "loss": 0.0413, + "step": 21693 + }, + { + "epoch": 3.95, + "learning_rate": 1.8181288678558427e-06, + "loss": 0.3729, + "step": 21694 + }, + { + "epoch": 3.95, + "learning_rate": 1.8180742628321807e-06, + "loss": 0.2903, + "step": 21695 + }, + { + "epoch": 3.95, + "learning_rate": 1.8180196578085184e-06, + "loss": 0.0112, + "step": 21696 + }, + { + "epoch": 3.95, + "learning_rate": 1.8179650527848564e-06, + "loss": 0.0329, + "step": 21697 + }, + { + "epoch": 3.95, + "learning_rate": 1.8179104477611943e-06, + "loss": 0.0822, + "step": 21698 + }, + { + "epoch": 3.95, + "learning_rate": 1.8178558427375318e-06, + "loss": 1.6352, + "step": 21699 + }, + { + "epoch": 3.95, + "learning_rate": 1.8178012377138698e-06, + "loss": 0.7668, + "step": 21700 + }, + { + "epoch": 3.95, + "learning_rate": 1.8177466326902077e-06, + "loss": 1.4223, + "step": 21701 + }, + { + "epoch": 3.95, + "learning_rate": 1.8176920276665452e-06, + "loss": 0.1102, + "step": 21702 + }, + { + "epoch": 3.95, + "learning_rate": 1.8176374226428832e-06, + "loss": 0.5424, + "step": 21703 + }, + { + "epoch": 3.95, + "learning_rate": 1.817582817619221e-06, + "loss": 2.0989, + "step": 21704 + }, + { + "epoch": 3.95, + "learning_rate": 1.8175282125955588e-06, + "loss": 1.4518, + "step": 21705 + }, + { + "epoch": 3.95, + "learning_rate": 1.8174736075718968e-06, + "loss": 0.4204, + "step": 21706 + }, + { + "epoch": 3.95, + "learning_rate": 1.8174190025482343e-06, + "loss": 0.407, + "step": 21707 + }, + { + "epoch": 3.95, + "learning_rate": 1.8173643975245722e-06, + "loss": 1.849, + "step": 21708 + }, + { + "epoch": 3.95, + "learning_rate": 1.8173097925009102e-06, + "loss": 0.0152, + "step": 21709 + }, + { + "epoch": 3.95, + "learning_rate": 1.817255187477248e-06, + "loss": 0.4774, + "step": 21710 + }, + { + "epoch": 3.95, + "learning_rate": 1.8172005824535859e-06, + "loss": 0.4939, + "step": 21711 + }, + { + "epoch": 3.95, + "learning_rate": 1.8171459774299238e-06, + "loss": 1.335, + "step": 21712 + }, + { + "epoch": 3.95, + "learning_rate": 1.8170913724062613e-06, + "loss": 0.6788, + "step": 21713 + }, + { + "epoch": 3.95, + "learning_rate": 1.8170367673825993e-06, + "loss": 0.1859, + "step": 21714 + }, + { + "epoch": 3.95, + "learning_rate": 1.8169821623589372e-06, + "loss": 0.9211, + "step": 21715 + }, + { + "epoch": 3.95, + "learning_rate": 1.8169275573352747e-06, + "loss": 1.612, + "step": 21716 + }, + { + "epoch": 3.95, + "learning_rate": 1.8168729523116127e-06, + "loss": 0.0157, + "step": 21717 + }, + { + "epoch": 3.95, + "learning_rate": 1.8168183472879504e-06, + "loss": 0.0166, + "step": 21718 + }, + { + "epoch": 3.95, + "learning_rate": 1.8167637422642884e-06, + "loss": 1.7064, + "step": 21719 + }, + { + "epoch": 3.95, + "learning_rate": 1.8167091372406263e-06, + "loss": 2.1048, + "step": 21720 + }, + { + "epoch": 3.95, + "learning_rate": 1.8166545322169638e-06, + "loss": 0.5852, + "step": 21721 + }, + { + "epoch": 3.95, + "learning_rate": 1.8165999271933018e-06, + "loss": 0.0185, + "step": 21722 + }, + { + "epoch": 3.95, + "learning_rate": 1.8165453221696397e-06, + "loss": 0.9098, + "step": 21723 + }, + { + "epoch": 3.95, + "learning_rate": 1.8164907171459774e-06, + "loss": 0.025, + "step": 21724 + }, + { + "epoch": 3.95, + "learning_rate": 1.8164361121223154e-06, + "loss": 0.3849, + "step": 21725 + }, + { + "epoch": 3.95, + "learning_rate": 1.8163815070986533e-06, + "loss": 2.4608, + "step": 21726 + }, + { + "epoch": 3.95, + "learning_rate": 1.8163269020749908e-06, + "loss": 0.282, + "step": 21727 + }, + { + "epoch": 3.95, + "learning_rate": 1.8162722970513288e-06, + "loss": 0.4741, + "step": 21728 + }, + { + "epoch": 3.95, + "learning_rate": 1.8162176920276667e-06, + "loss": 0.2017, + "step": 21729 + }, + { + "epoch": 3.95, + "learning_rate": 1.8161630870040042e-06, + "loss": 1.3045, + "step": 21730 + }, + { + "epoch": 3.95, + "learning_rate": 1.8161084819803422e-06, + "loss": 0.5819, + "step": 21731 + }, + { + "epoch": 3.95, + "learning_rate": 1.81605387695668e-06, + "loss": 1.1756, + "step": 21732 + }, + { + "epoch": 3.95, + "learning_rate": 1.8159992719330179e-06, + "loss": 3.0926, + "step": 21733 + }, + { + "epoch": 3.95, + "learning_rate": 1.8159446669093558e-06, + "loss": 0.1627, + "step": 21734 + }, + { + "epoch": 3.95, + "learning_rate": 1.8158900618856933e-06, + "loss": 0.2998, + "step": 21735 + }, + { + "epoch": 3.95, + "learning_rate": 1.8158354568620313e-06, + "loss": 0.0537, + "step": 21736 + }, + { + "epoch": 3.95, + "learning_rate": 1.8157808518383692e-06, + "loss": 0.3212, + "step": 21737 + }, + { + "epoch": 3.95, + "learning_rate": 1.815726246814707e-06, + "loss": 0.871, + "step": 21738 + }, + { + "epoch": 3.95, + "learning_rate": 1.8156716417910449e-06, + "loss": 0.7265, + "step": 21739 + }, + { + "epoch": 3.95, + "learning_rate": 1.8156170367673828e-06, + "loss": 0.5498, + "step": 21740 + }, + { + "epoch": 3.95, + "learning_rate": 1.8155624317437204e-06, + "loss": 0.4129, + "step": 21741 + }, + { + "epoch": 3.95, + "learning_rate": 1.8155078267200583e-06, + "loss": 0.6932, + "step": 21742 + }, + { + "epoch": 3.95, + "learning_rate": 1.8154532216963962e-06, + "loss": 1.5633, + "step": 21743 + }, + { + "epoch": 3.95, + "learning_rate": 1.815398616672734e-06, + "loss": 0.3472, + "step": 21744 + }, + { + "epoch": 3.95, + "learning_rate": 1.815344011649072e-06, + "loss": 0.7892, + "step": 21745 + }, + { + "epoch": 3.95, + "learning_rate": 1.8152894066254094e-06, + "loss": 0.645, + "step": 21746 + }, + { + "epoch": 3.95, + "learning_rate": 1.8152348016017474e-06, + "loss": 0.0131, + "step": 21747 + }, + { + "epoch": 3.95, + "learning_rate": 1.8151801965780853e-06, + "loss": 0.0856, + "step": 21748 + }, + { + "epoch": 3.96, + "learning_rate": 1.8151255915544228e-06, + "loss": 0.796, + "step": 21749 + }, + { + "epoch": 3.96, + "learning_rate": 1.8150709865307608e-06, + "loss": 0.1757, + "step": 21750 + }, + { + "epoch": 3.96, + "learning_rate": 1.8150163815070987e-06, + "loss": 1.2438, + "step": 21751 + }, + { + "epoch": 3.96, + "learning_rate": 1.8149617764834365e-06, + "loss": 0.1249, + "step": 21752 + }, + { + "epoch": 3.96, + "learning_rate": 1.8149071714597744e-06, + "loss": 1.816, + "step": 21753 + }, + { + "epoch": 3.96, + "learning_rate": 1.8148525664361123e-06, + "loss": 1.1118, + "step": 21754 + }, + { + "epoch": 3.96, + "learning_rate": 1.8147979614124499e-06, + "loss": 1.6588, + "step": 21755 + }, + { + "epoch": 3.96, + "learning_rate": 1.8147433563887878e-06, + "loss": 0.4235, + "step": 21756 + }, + { + "epoch": 3.96, + "learning_rate": 1.8146887513651258e-06, + "loss": 0.4912, + "step": 21757 + }, + { + "epoch": 3.96, + "learning_rate": 1.8146341463414635e-06, + "loss": 1.2677, + "step": 21758 + }, + { + "epoch": 3.96, + "learning_rate": 1.8145795413178014e-06, + "loss": 2.808, + "step": 21759 + }, + { + "epoch": 3.96, + "learning_rate": 1.814524936294139e-06, + "loss": 0.0051, + "step": 21760 + }, + { + "epoch": 3.96, + "learning_rate": 1.814470331270477e-06, + "loss": 2.9358, + "step": 21761 + }, + { + "epoch": 3.96, + "learning_rate": 1.8144157262468148e-06, + "loss": 0.6055, + "step": 21762 + }, + { + "epoch": 3.96, + "learning_rate": 1.8143611212231524e-06, + "loss": 0.3534, + "step": 21763 + }, + { + "epoch": 3.96, + "learning_rate": 1.8143065161994903e-06, + "loss": 0.0325, + "step": 21764 + }, + { + "epoch": 3.96, + "learning_rate": 1.8142519111758282e-06, + "loss": 1.0127, + "step": 21765 + }, + { + "epoch": 3.96, + "learning_rate": 1.814197306152166e-06, + "loss": 0.3831, + "step": 21766 + }, + { + "epoch": 3.96, + "learning_rate": 1.814142701128504e-06, + "loss": 0.1686, + "step": 21767 + }, + { + "epoch": 3.96, + "learning_rate": 1.8140880961048419e-06, + "loss": 3.2208, + "step": 21768 + }, + { + "epoch": 3.96, + "learning_rate": 1.8140334910811794e-06, + "loss": 0.8825, + "step": 21769 + }, + { + "epoch": 3.96, + "learning_rate": 1.8139788860575173e-06, + "loss": 0.0555, + "step": 21770 + }, + { + "epoch": 3.96, + "learning_rate": 1.8139242810338553e-06, + "loss": 2.2737, + "step": 21771 + }, + { + "epoch": 3.96, + "learning_rate": 1.813869676010193e-06, + "loss": 2.5866, + "step": 21772 + }, + { + "epoch": 3.96, + "learning_rate": 1.813815070986531e-06, + "loss": 0.3991, + "step": 21773 + }, + { + "epoch": 3.96, + "learning_rate": 1.8137604659628685e-06, + "loss": 1.3299, + "step": 21774 + }, + { + "epoch": 3.96, + "learning_rate": 1.8137058609392064e-06, + "loss": 2.5802, + "step": 21775 + }, + { + "epoch": 3.96, + "learning_rate": 1.8136512559155444e-06, + "loss": 1.6585, + "step": 21776 + }, + { + "epoch": 3.96, + "learning_rate": 1.813596650891882e-06, + "loss": 0.6336, + "step": 21777 + }, + { + "epoch": 3.96, + "learning_rate": 1.81354204586822e-06, + "loss": 0.324, + "step": 21778 + }, + { + "epoch": 3.96, + "learning_rate": 1.813487440844558e-06, + "loss": 1.1948, + "step": 21779 + }, + { + "epoch": 3.96, + "learning_rate": 1.8134328358208955e-06, + "loss": 0.4856, + "step": 21780 + }, + { + "epoch": 3.96, + "learning_rate": 1.8133782307972334e-06, + "loss": 1.003, + "step": 21781 + }, + { + "epoch": 3.96, + "learning_rate": 1.8133236257735714e-06, + "loss": 0.0683, + "step": 21782 + }, + { + "epoch": 3.96, + "learning_rate": 1.813269020749909e-06, + "loss": 0.022, + "step": 21783 + }, + { + "epoch": 3.96, + "learning_rate": 1.8132144157262468e-06, + "loss": 1.3811, + "step": 21784 + }, + { + "epoch": 3.96, + "learning_rate": 1.8131598107025846e-06, + "loss": 2.2391, + "step": 21785 + }, + { + "epoch": 3.96, + "learning_rate": 1.8131052056789225e-06, + "loss": 2.321, + "step": 21786 + }, + { + "epoch": 3.96, + "learning_rate": 1.8130506006552605e-06, + "loss": 2.3365, + "step": 21787 + }, + { + "epoch": 3.96, + "learning_rate": 1.812995995631598e-06, + "loss": 0.2449, + "step": 21788 + }, + { + "epoch": 3.96, + "learning_rate": 1.812941390607936e-06, + "loss": 0.1834, + "step": 21789 + }, + { + "epoch": 3.96, + "learning_rate": 1.8128867855842739e-06, + "loss": 0.0296, + "step": 21790 + }, + { + "epoch": 3.96, + "learning_rate": 1.8128321805606116e-06, + "loss": 0.4115, + "step": 21791 + }, + { + "epoch": 3.96, + "learning_rate": 1.8127775755369495e-06, + "loss": 0.0411, + "step": 21792 + }, + { + "epoch": 3.96, + "learning_rate": 1.8127229705132875e-06, + "loss": 0.7704, + "step": 21793 + }, + { + "epoch": 3.96, + "learning_rate": 1.812668365489625e-06, + "loss": 1.5437, + "step": 21794 + }, + { + "epoch": 3.96, + "learning_rate": 1.812613760465963e-06, + "loss": 0.505, + "step": 21795 + }, + { + "epoch": 3.96, + "learning_rate": 1.8125591554423009e-06, + "loss": 1.6196, + "step": 21796 + }, + { + "epoch": 3.96, + "learning_rate": 1.8125045504186384e-06, + "loss": 2.3254, + "step": 21797 + }, + { + "epoch": 3.96, + "learning_rate": 1.8124499453949764e-06, + "loss": 0.1144, + "step": 21798 + }, + { + "epoch": 3.96, + "learning_rate": 1.812395340371314e-06, + "loss": 0.5843, + "step": 21799 + }, + { + "epoch": 3.96, + "learning_rate": 1.812340735347652e-06, + "loss": 3.5039, + "step": 21800 + }, + { + "epoch": 3.96, + "learning_rate": 1.81228613032399e-06, + "loss": 0.4909, + "step": 21801 + }, + { + "epoch": 3.96, + "learning_rate": 1.8122315253003275e-06, + "loss": 0.1514, + "step": 21802 + }, + { + "epoch": 3.96, + "learning_rate": 1.8121769202766654e-06, + "loss": 0.0117, + "step": 21803 + }, + { + "epoch": 3.97, + "learning_rate": 1.8121223152530034e-06, + "loss": 1.2683, + "step": 21804 + }, + { + "epoch": 3.97, + "learning_rate": 1.8120677102293411e-06, + "loss": 0.7859, + "step": 21805 + }, + { + "epoch": 3.97, + "learning_rate": 1.812013105205679e-06, + "loss": 0.3641, + "step": 21806 + }, + { + "epoch": 3.97, + "learning_rate": 1.811958500182017e-06, + "loss": 0.4616, + "step": 21807 + }, + { + "epoch": 3.97, + "learning_rate": 1.8119038951583545e-06, + "loss": 0.4561, + "step": 21808 + }, + { + "epoch": 3.97, + "learning_rate": 1.8118492901346925e-06, + "loss": 2.331, + "step": 21809 + }, + { + "epoch": 3.97, + "learning_rate": 1.8117946851110304e-06, + "loss": 2.4784, + "step": 21810 + }, + { + "epoch": 3.97, + "learning_rate": 1.811740080087368e-06, + "loss": 0.1866, + "step": 21811 + }, + { + "epoch": 3.97, + "learning_rate": 1.8116854750637059e-06, + "loss": 0.0139, + "step": 21812 + }, + { + "epoch": 3.97, + "learning_rate": 1.8116308700400436e-06, + "loss": 1.036, + "step": 21813 + }, + { + "epoch": 3.97, + "learning_rate": 1.8115762650163815e-06, + "loss": 0.8847, + "step": 21814 + }, + { + "epoch": 3.97, + "learning_rate": 1.8115216599927195e-06, + "loss": 1.6188, + "step": 21815 + }, + { + "epoch": 3.97, + "learning_rate": 1.811467054969057e-06, + "loss": 0.5958, + "step": 21816 + }, + { + "epoch": 3.97, + "learning_rate": 1.811412449945395e-06, + "loss": 0.3005, + "step": 21817 + }, + { + "epoch": 3.97, + "learning_rate": 1.8113578449217329e-06, + "loss": 0.1928, + "step": 21818 + }, + { + "epoch": 3.97, + "learning_rate": 1.8113032398980706e-06, + "loss": 2.5569, + "step": 21819 + }, + { + "epoch": 3.97, + "learning_rate": 1.8112486348744086e-06, + "loss": 0.333, + "step": 21820 + }, + { + "epoch": 3.97, + "learning_rate": 1.8111940298507465e-06, + "loss": 1.076, + "step": 21821 + }, + { + "epoch": 3.97, + "learning_rate": 1.811139424827084e-06, + "loss": 1.5417, + "step": 21822 + }, + { + "epoch": 3.97, + "learning_rate": 1.811084819803422e-06, + "loss": 0.3479, + "step": 21823 + }, + { + "epoch": 3.97, + "learning_rate": 1.81103021477976e-06, + "loss": 2.5492, + "step": 21824 + }, + { + "epoch": 3.97, + "learning_rate": 1.8109756097560976e-06, + "loss": 0.5177, + "step": 21825 + }, + { + "epoch": 3.97, + "learning_rate": 1.8109210047324356e-06, + "loss": 0.1422, + "step": 21826 + }, + { + "epoch": 3.97, + "learning_rate": 1.8108663997087731e-06, + "loss": 0.3198, + "step": 21827 + }, + { + "epoch": 3.97, + "learning_rate": 1.810811794685111e-06, + "loss": 0.1495, + "step": 21828 + }, + { + "epoch": 3.97, + "learning_rate": 1.810757189661449e-06, + "loss": 0.8809, + "step": 21829 + }, + { + "epoch": 3.97, + "learning_rate": 1.8107025846377865e-06, + "loss": 0.1186, + "step": 21830 + }, + { + "epoch": 3.97, + "learning_rate": 1.8106479796141245e-06, + "loss": 0.0296, + "step": 21831 + }, + { + "epoch": 3.97, + "learning_rate": 1.8105933745904624e-06, + "loss": 0.1019, + "step": 21832 + }, + { + "epoch": 3.97, + "learning_rate": 1.8105387695668001e-06, + "loss": 0.2903, + "step": 21833 + }, + { + "epoch": 3.97, + "learning_rate": 1.810484164543138e-06, + "loss": 0.1118, + "step": 21834 + }, + { + "epoch": 3.97, + "learning_rate": 1.810429559519476e-06, + "loss": 0.0665, + "step": 21835 + }, + { + "epoch": 3.97, + "learning_rate": 1.8103749544958135e-06, + "loss": 1.4242, + "step": 21836 + }, + { + "epoch": 3.97, + "learning_rate": 1.8103203494721515e-06, + "loss": 0.1104, + "step": 21837 + }, + { + "epoch": 3.97, + "learning_rate": 1.8102657444484894e-06, + "loss": 0.5752, + "step": 21838 + }, + { + "epoch": 3.97, + "learning_rate": 1.8102111394248272e-06, + "loss": 0.0056, + "step": 21839 + }, + { + "epoch": 3.97, + "learning_rate": 1.810156534401165e-06, + "loss": 0.0323, + "step": 21840 + }, + { + "epoch": 3.97, + "learning_rate": 1.8101019293775026e-06, + "loss": 1.2627, + "step": 21841 + }, + { + "epoch": 3.97, + "learning_rate": 1.8100473243538406e-06, + "loss": 0.3285, + "step": 21842 + }, + { + "epoch": 3.97, + "learning_rate": 1.8099927193301785e-06, + "loss": 0.8091, + "step": 21843 + }, + { + "epoch": 3.97, + "learning_rate": 1.809938114306516e-06, + "loss": 0.0251, + "step": 21844 + }, + { + "epoch": 3.97, + "learning_rate": 1.809883509282854e-06, + "loss": 0.7336, + "step": 21845 + }, + { + "epoch": 3.97, + "learning_rate": 1.809828904259192e-06, + "loss": 0.0853, + "step": 21846 + }, + { + "epoch": 3.97, + "learning_rate": 1.8097742992355297e-06, + "loss": 0.0174, + "step": 21847 + }, + { + "epoch": 3.97, + "learning_rate": 1.8097196942118676e-06, + "loss": 1.9735, + "step": 21848 + }, + { + "epoch": 3.97, + "learning_rate": 1.8096650891882055e-06, + "loss": 0.3064, + "step": 21849 + }, + { + "epoch": 3.97, + "learning_rate": 1.809610484164543e-06, + "loss": 0.0142, + "step": 21850 + }, + { + "epoch": 3.97, + "learning_rate": 1.809555879140881e-06, + "loss": 1.1144, + "step": 21851 + }, + { + "epoch": 3.97, + "learning_rate": 1.809501274117219e-06, + "loss": 0.1332, + "step": 21852 + }, + { + "epoch": 3.97, + "learning_rate": 1.8094466690935567e-06, + "loss": 0.9306, + "step": 21853 + }, + { + "epoch": 3.97, + "learning_rate": 1.8093920640698946e-06, + "loss": 0.7735, + "step": 21854 + }, + { + "epoch": 3.97, + "learning_rate": 1.8093374590462321e-06, + "loss": 0.0185, + "step": 21855 + }, + { + "epoch": 3.97, + "learning_rate": 1.80928285402257e-06, + "loss": 0.7812, + "step": 21856 + }, + { + "epoch": 3.97, + "learning_rate": 1.809228248998908e-06, + "loss": 0.5528, + "step": 21857 + }, + { + "epoch": 3.97, + "learning_rate": 1.8091736439752455e-06, + "loss": 0.4603, + "step": 21858 + }, + { + "epoch": 3.98, + "learning_rate": 1.8091190389515835e-06, + "loss": 0.6702, + "step": 21859 + }, + { + "epoch": 3.98, + "learning_rate": 1.8090644339279214e-06, + "loss": 0.1688, + "step": 21860 + }, + { + "epoch": 3.98, + "learning_rate": 1.8090098289042592e-06, + "loss": 1.3898, + "step": 21861 + }, + { + "epoch": 3.98, + "learning_rate": 1.8089552238805971e-06, + "loss": 0.0984, + "step": 21862 + }, + { + "epoch": 3.98, + "learning_rate": 1.808900618856935e-06, + "loss": 0.4679, + "step": 21863 + }, + { + "epoch": 3.98, + "learning_rate": 1.8088460138332726e-06, + "loss": 0.1343, + "step": 21864 + }, + { + "epoch": 3.98, + "learning_rate": 1.8087914088096105e-06, + "loss": 0.6025, + "step": 21865 + }, + { + "epoch": 3.98, + "learning_rate": 1.8087368037859482e-06, + "loss": 0.2136, + "step": 21866 + }, + { + "epoch": 3.98, + "learning_rate": 1.8086821987622862e-06, + "loss": 0.6747, + "step": 21867 + }, + { + "epoch": 3.98, + "learning_rate": 1.8086275937386241e-06, + "loss": 0.753, + "step": 21868 + }, + { + "epoch": 3.98, + "learning_rate": 1.8085729887149617e-06, + "loss": 1.0328, + "step": 21869 + }, + { + "epoch": 3.98, + "learning_rate": 1.8085183836912996e-06, + "loss": 1.0986, + "step": 21870 + }, + { + "epoch": 3.98, + "learning_rate": 1.8084637786676375e-06, + "loss": 0.6393, + "step": 21871 + }, + { + "epoch": 3.98, + "learning_rate": 1.8084091736439753e-06, + "loss": 0.4302, + "step": 21872 + }, + { + "epoch": 3.98, + "learning_rate": 1.8083545686203132e-06, + "loss": 0.876, + "step": 21873 + }, + { + "epoch": 3.98, + "learning_rate": 1.8082999635966512e-06, + "loss": 0.016, + "step": 21874 + }, + { + "epoch": 3.98, + "learning_rate": 1.8082453585729887e-06, + "loss": 0.0779, + "step": 21875 + }, + { + "epoch": 3.98, + "learning_rate": 1.8081907535493266e-06, + "loss": 0.0177, + "step": 21876 + }, + { + "epoch": 3.98, + "learning_rate": 1.8081361485256646e-06, + "loss": 2.1574, + "step": 21877 + }, + { + "epoch": 3.98, + "learning_rate": 1.808081543502002e-06, + "loss": 1.9746, + "step": 21878 + }, + { + "epoch": 3.98, + "learning_rate": 1.80802693847834e-06, + "loss": 0.0373, + "step": 21879 + }, + { + "epoch": 3.98, + "learning_rate": 1.8079723334546778e-06, + "loss": 2.2637, + "step": 21880 + }, + { + "epoch": 3.98, + "learning_rate": 1.8079177284310157e-06, + "loss": 2.0552, + "step": 21881 + }, + { + "epoch": 3.98, + "learning_rate": 1.8078631234073536e-06, + "loss": 0.9835, + "step": 21882 + }, + { + "epoch": 3.98, + "learning_rate": 1.8078085183836912e-06, + "loss": 0.4204, + "step": 21883 + }, + { + "epoch": 3.98, + "learning_rate": 1.8077539133600291e-06, + "loss": 0.6228, + "step": 21884 + }, + { + "epoch": 3.98, + "learning_rate": 1.807699308336367e-06, + "loss": 0.0343, + "step": 21885 + }, + { + "epoch": 3.98, + "learning_rate": 1.8076447033127048e-06, + "loss": 0.5649, + "step": 21886 + }, + { + "epoch": 3.98, + "learning_rate": 1.8075900982890427e-06, + "loss": 0.9722, + "step": 21887 + }, + { + "epoch": 3.98, + "learning_rate": 1.8075354932653807e-06, + "loss": 0.0162, + "step": 21888 + }, + { + "epoch": 3.98, + "learning_rate": 1.8074808882417182e-06, + "loss": 0.1696, + "step": 21889 + }, + { + "epoch": 3.98, + "learning_rate": 1.8074262832180561e-06, + "loss": 0.3561, + "step": 21890 + }, + { + "epoch": 3.98, + "learning_rate": 1.807371678194394e-06, + "loss": 0.8113, + "step": 21891 + }, + { + "epoch": 3.98, + "learning_rate": 1.8073170731707316e-06, + "loss": 0.1025, + "step": 21892 + }, + { + "epoch": 3.98, + "learning_rate": 1.8072624681470695e-06, + "loss": 0.3918, + "step": 21893 + }, + { + "epoch": 3.98, + "learning_rate": 1.8072078631234073e-06, + "loss": 0.5996, + "step": 21894 + }, + { + "epoch": 3.98, + "learning_rate": 1.8071532580997452e-06, + "loss": 3.0018, + "step": 21895 + }, + { + "epoch": 3.98, + "learning_rate": 1.8070986530760832e-06, + "loss": 0.0286, + "step": 21896 + }, + { + "epoch": 3.98, + "learning_rate": 1.8070440480524207e-06, + "loss": 0.2567, + "step": 21897 + }, + { + "epoch": 3.98, + "learning_rate": 1.8069894430287586e-06, + "loss": 0.0252, + "step": 21898 + }, + { + "epoch": 3.98, + "learning_rate": 1.8069348380050966e-06, + "loss": 0.0712, + "step": 21899 + }, + { + "epoch": 3.98, + "learning_rate": 1.8068802329814343e-06, + "loss": 0.0087, + "step": 21900 + }, + { + "epoch": 3.98, + "learning_rate": 1.8068256279577722e-06, + "loss": 0.2384, + "step": 21901 + }, + { + "epoch": 3.98, + "learning_rate": 1.8067710229341102e-06, + "loss": 0.1184, + "step": 21902 + }, + { + "epoch": 3.98, + "learning_rate": 1.8067164179104477e-06, + "loss": 0.0349, + "step": 21903 + }, + { + "epoch": 3.98, + "learning_rate": 1.8066618128867856e-06, + "loss": 0.0576, + "step": 21904 + }, + { + "epoch": 3.98, + "learning_rate": 1.8066072078631236e-06, + "loss": 0.0435, + "step": 21905 + }, + { + "epoch": 3.98, + "learning_rate": 1.8065526028394613e-06, + "loss": 1.0735, + "step": 21906 + }, + { + "epoch": 3.98, + "learning_rate": 1.8064979978157993e-06, + "loss": 0.6552, + "step": 21907 + }, + { + "epoch": 3.98, + "learning_rate": 1.8064433927921368e-06, + "loss": 0.1106, + "step": 21908 + }, + { + "epoch": 3.98, + "learning_rate": 1.8063887877684747e-06, + "loss": 0.0314, + "step": 21909 + }, + { + "epoch": 3.98, + "learning_rate": 1.8063341827448127e-06, + "loss": 1.2483, + "step": 21910 + }, + { + "epoch": 3.98, + "learning_rate": 1.8062795777211502e-06, + "loss": 0.1501, + "step": 21911 + }, + { + "epoch": 3.98, + "learning_rate": 1.8062249726974881e-06, + "loss": 1.1902, + "step": 21912 + }, + { + "epoch": 3.98, + "learning_rate": 1.806170367673826e-06, + "loss": 0.1933, + "step": 21913 + }, + { + "epoch": 3.99, + "learning_rate": 1.8061157626501638e-06, + "loss": 0.3951, + "step": 21914 + }, + { + "epoch": 3.99, + "learning_rate": 1.8060611576265018e-06, + "loss": 0.0225, + "step": 21915 + }, + { + "epoch": 3.99, + "learning_rate": 1.8060065526028397e-06, + "loss": 0.8398, + "step": 21916 + }, + { + "epoch": 3.99, + "learning_rate": 1.8059519475791772e-06, + "loss": 0.7779, + "step": 21917 + }, + { + "epoch": 3.99, + "learning_rate": 1.8058973425555152e-06, + "loss": 0.0414, + "step": 21918 + }, + { + "epoch": 3.99, + "learning_rate": 1.805842737531853e-06, + "loss": 0.1344, + "step": 21919 + }, + { + "epoch": 3.99, + "learning_rate": 1.8057881325081908e-06, + "loss": 0.4809, + "step": 21920 + }, + { + "epoch": 3.99, + "learning_rate": 1.8057335274845288e-06, + "loss": 0.0313, + "step": 21921 + }, + { + "epoch": 3.99, + "learning_rate": 1.8056789224608663e-06, + "loss": 1.8069, + "step": 21922 + }, + { + "epoch": 3.99, + "learning_rate": 1.8056243174372042e-06, + "loss": 0.0515, + "step": 21923 + }, + { + "epoch": 3.99, + "learning_rate": 1.8055697124135422e-06, + "loss": 3.2218, + "step": 21924 + }, + { + "epoch": 3.99, + "learning_rate": 1.8055151073898797e-06, + "loss": 0.4664, + "step": 21925 + }, + { + "epoch": 3.99, + "learning_rate": 1.8054605023662176e-06, + "loss": 1.8765, + "step": 21926 + }, + { + "epoch": 3.99, + "learning_rate": 1.8054058973425556e-06, + "loss": 3.0759, + "step": 21927 + }, + { + "epoch": 3.99, + "learning_rate": 1.8053512923188933e-06, + "loss": 0.2005, + "step": 21928 + }, + { + "epoch": 3.99, + "learning_rate": 1.8052966872952313e-06, + "loss": 0.0892, + "step": 21929 + }, + { + "epoch": 3.99, + "learning_rate": 1.8052420822715692e-06, + "loss": 0.1085, + "step": 21930 + }, + { + "epoch": 3.99, + "learning_rate": 1.8051874772479067e-06, + "loss": 0.5577, + "step": 21931 + }, + { + "epoch": 3.99, + "learning_rate": 1.8051328722242447e-06, + "loss": 0.5705, + "step": 21932 + }, + { + "epoch": 3.99, + "learning_rate": 1.8050782672005824e-06, + "loss": 0.5561, + "step": 21933 + }, + { + "epoch": 3.99, + "learning_rate": 1.8050236621769203e-06, + "loss": 2.1566, + "step": 21934 + }, + { + "epoch": 3.99, + "learning_rate": 1.8049690571532583e-06, + "loss": 3.5136, + "step": 21935 + }, + { + "epoch": 3.99, + "learning_rate": 1.8049144521295958e-06, + "loss": 2.2624, + "step": 21936 + }, + { + "epoch": 3.99, + "learning_rate": 1.8048598471059338e-06, + "loss": 0.1838, + "step": 21937 + }, + { + "epoch": 3.99, + "learning_rate": 1.8048052420822717e-06, + "loss": 0.0818, + "step": 21938 + }, + { + "epoch": 3.99, + "learning_rate": 1.8047506370586092e-06, + "loss": 1.0687, + "step": 21939 + }, + { + "epoch": 3.99, + "learning_rate": 1.8046960320349472e-06, + "loss": 1.6339, + "step": 21940 + }, + { + "epoch": 3.99, + "learning_rate": 1.8046414270112851e-06, + "loss": 0.0383, + "step": 21941 + }, + { + "epoch": 3.99, + "learning_rate": 1.8045868219876228e-06, + "loss": 0.434, + "step": 21942 + }, + { + "epoch": 3.99, + "learning_rate": 1.8045322169639608e-06, + "loss": 0.2367, + "step": 21943 + }, + { + "epoch": 3.99, + "learning_rate": 1.8044776119402987e-06, + "loss": 0.0064, + "step": 21944 + }, + { + "epoch": 3.99, + "learning_rate": 1.8044230069166362e-06, + "loss": 1.8027, + "step": 21945 + }, + { + "epoch": 3.99, + "learning_rate": 1.8043684018929742e-06, + "loss": 0.5728, + "step": 21946 + }, + { + "epoch": 3.99, + "learning_rate": 1.804313796869312e-06, + "loss": 0.2346, + "step": 21947 + }, + { + "epoch": 3.99, + "learning_rate": 1.8042591918456499e-06, + "loss": 1.132, + "step": 21948 + }, + { + "epoch": 3.99, + "learning_rate": 1.8042045868219878e-06, + "loss": 0.5828, + "step": 21949 + }, + { + "epoch": 3.99, + "learning_rate": 1.8041499817983253e-06, + "loss": 0.3516, + "step": 21950 + }, + { + "epoch": 3.99, + "learning_rate": 1.8040953767746633e-06, + "loss": 0.5328, + "step": 21951 + }, + { + "epoch": 3.99, + "learning_rate": 1.8040407717510012e-06, + "loss": 0.3472, + "step": 21952 + }, + { + "epoch": 3.99, + "learning_rate": 1.803986166727339e-06, + "loss": 0.2098, + "step": 21953 + }, + { + "epoch": 3.99, + "learning_rate": 1.8039315617036769e-06, + "loss": 2.0093, + "step": 21954 + }, + { + "epoch": 3.99, + "learning_rate": 1.8038769566800148e-06, + "loss": 0.0112, + "step": 21955 + }, + { + "epoch": 3.99, + "learning_rate": 1.8038223516563524e-06, + "loss": 0.3902, + "step": 21956 + }, + { + "epoch": 3.99, + "learning_rate": 1.8037677466326903e-06, + "loss": 3.224, + "step": 21957 + }, + { + "epoch": 3.99, + "learning_rate": 1.8037131416090282e-06, + "loss": 0.4424, + "step": 21958 + }, + { + "epoch": 3.99, + "learning_rate": 1.8036585365853658e-06, + "loss": 3.1811, + "step": 21959 + }, + { + "epoch": 3.99, + "learning_rate": 1.8036039315617037e-06, + "loss": 0.4883, + "step": 21960 + }, + { + "epoch": 3.99, + "learning_rate": 1.8035493265380414e-06, + "loss": 0.0152, + "step": 21961 + }, + { + "epoch": 3.99, + "learning_rate": 1.8034947215143794e-06, + "loss": 0.0794, + "step": 21962 + }, + { + "epoch": 3.99, + "learning_rate": 1.8034401164907173e-06, + "loss": 3.3572, + "step": 21963 + }, + { + "epoch": 3.99, + "learning_rate": 1.8033855114670548e-06, + "loss": 1.4698, + "step": 21964 + }, + { + "epoch": 3.99, + "learning_rate": 1.8033309064433928e-06, + "loss": 2.8316, + "step": 21965 + }, + { + "epoch": 3.99, + "learning_rate": 1.8032763014197307e-06, + "loss": 2.1241, + "step": 21966 + }, + { + "epoch": 3.99, + "learning_rate": 1.8032216963960685e-06, + "loss": 0.1346, + "step": 21967 + }, + { + "epoch": 3.99, + "learning_rate": 1.8031670913724064e-06, + "loss": 0.2566, + "step": 21968 + }, + { + "epoch": 4.0, + "learning_rate": 1.8031124863487443e-06, + "loss": 1.7147, + "step": 21969 + }, + { + "epoch": 4.0, + "learning_rate": 1.8030578813250819e-06, + "loss": 0.58, + "step": 21970 + }, + { + "epoch": 4.0, + "learning_rate": 1.8030032763014198e-06, + "loss": 0.7379, + "step": 21971 + }, + { + "epoch": 4.0, + "learning_rate": 1.8029486712777578e-06, + "loss": 1.2507, + "step": 21972 + }, + { + "epoch": 4.0, + "learning_rate": 1.8028940662540953e-06, + "loss": 0.5174, + "step": 21973 + }, + { + "epoch": 4.0, + "learning_rate": 1.8028394612304332e-06, + "loss": 0.2883, + "step": 21974 + }, + { + "epoch": 4.0, + "learning_rate": 1.802784856206771e-06, + "loss": 0.7157, + "step": 21975 + }, + { + "epoch": 4.0, + "learning_rate": 1.8027302511831089e-06, + "loss": 0.2452, + "step": 21976 + }, + { + "epoch": 4.0, + "learning_rate": 1.8026756461594468e-06, + "loss": 0.3889, + "step": 21977 + }, + { + "epoch": 4.0, + "learning_rate": 1.8026210411357844e-06, + "loss": 1.8757, + "step": 21978 + }, + { + "epoch": 4.0, + "learning_rate": 1.8025664361121223e-06, + "loss": 0.9325, + "step": 21979 + }, + { + "epoch": 4.0, + "learning_rate": 1.8025118310884602e-06, + "loss": 0.6313, + "step": 21980 + }, + { + "epoch": 4.0, + "learning_rate": 1.802457226064798e-06, + "loss": 0.3153, + "step": 21981 + }, + { + "epoch": 4.0, + "learning_rate": 1.802402621041136e-06, + "loss": 0.2761, + "step": 21982 + }, + { + "epoch": 4.0, + "learning_rate": 1.8023480160174739e-06, + "loss": 0.4497, + "step": 21983 + }, + { + "epoch": 4.0, + "learning_rate": 1.8022934109938114e-06, + "loss": 0.411, + "step": 21984 + }, + { + "epoch": 4.0, + "learning_rate": 1.8022388059701493e-06, + "loss": 0.6506, + "step": 21985 + }, + { + "epoch": 4.0, + "learning_rate": 1.8021842009464873e-06, + "loss": 0.3859, + "step": 21986 + }, + { + "epoch": 4.0, + "learning_rate": 1.802129595922825e-06, + "loss": 0.133, + "step": 21987 + }, + { + "epoch": 4.0, + "learning_rate": 1.8020749908991627e-06, + "loss": 1.3468, + "step": 21988 + }, + { + "epoch": 4.0, + "learning_rate": 1.8020203858755005e-06, + "loss": 0.4875, + "step": 21989 + }, + { + "epoch": 4.0, + "learning_rate": 1.8019657808518384e-06, + "loss": 0.3324, + "step": 21990 + }, + { + "epoch": 4.0, + "learning_rate": 1.8019111758281763e-06, + "loss": 3.0316, + "step": 21991 + }, + { + "epoch": 4.0, + "learning_rate": 1.8018565708045139e-06, + "loss": 0.274, + "step": 21992 + }, + { + "epoch": 4.0, + "learning_rate": 1.8018019657808518e-06, + "loss": 0.6516, + "step": 21993 + }, + { + "epoch": 4.0, + "learning_rate": 1.8017473607571898e-06, + "loss": 0.2737, + "step": 21994 + }, + { + "epoch": 4.0, + "learning_rate": 1.8016927557335275e-06, + "loss": 1.0978, + "step": 21995 + }, + { + "epoch": 4.0, + "learning_rate": 1.8016381507098654e-06, + "loss": 0.0214, + "step": 21996 + }, + { + "epoch": 4.0, + "eval_accuracy": 0.8211947626841244, + "eval_loss": 0.7773312926292419, + "eval_runtime": 135.6764, + "eval_samples_per_second": 18.013, + "eval_steps_per_second": 4.503, + "step": 21996 + }, + { + "epoch": 4.0, + "learning_rate": 1.8015835456862034e-06, + "loss": 0.9143, + "step": 21997 + }, + { + "epoch": 4.0, + "learning_rate": 1.8015289406625409e-06, + "loss": 3.0245, + "step": 21998 + }, + { + "epoch": 4.0, + "learning_rate": 1.8014743356388788e-06, + "loss": 0.0298, + "step": 21999 + }, + { + "epoch": 4.0, + "learning_rate": 1.8014197306152168e-06, + "loss": 0.3814, + "step": 22000 + }, + { + "epoch": 4.0, + "learning_rate": 1.8013651255915545e-06, + "loss": 1.0471, + "step": 22001 + }, + { + "epoch": 4.0, + "learning_rate": 1.8013105205678925e-06, + "loss": 0.0201, + "step": 22002 + }, + { + "epoch": 4.0, + "learning_rate": 1.80125591554423e-06, + "loss": 0.0125, + "step": 22003 + }, + { + "epoch": 4.0, + "learning_rate": 1.801201310520568e-06, + "loss": 0.0769, + "step": 22004 + }, + { + "epoch": 4.0, + "learning_rate": 1.8011467054969059e-06, + "loss": 0.6008, + "step": 22005 + }, + { + "epoch": 4.0, + "learning_rate": 1.8010921004732434e-06, + "loss": 0.1739, + "step": 22006 + }, + { + "epoch": 4.0, + "learning_rate": 1.8010374954495813e-06, + "loss": 0.0247, + "step": 22007 + }, + { + "epoch": 4.0, + "learning_rate": 1.8009828904259193e-06, + "loss": 0.4626, + "step": 22008 + }, + { + "epoch": 4.0, + "learning_rate": 1.800928285402257e-06, + "loss": 0.9763, + "step": 22009 + }, + { + "epoch": 4.0, + "learning_rate": 1.800873680378595e-06, + "loss": 0.2804, + "step": 22010 + }, + { + "epoch": 4.0, + "learning_rate": 1.8008190753549329e-06, + "loss": 0.054, + "step": 22011 + }, + { + "epoch": 4.0, + "learning_rate": 1.8007644703312704e-06, + "loss": 0.4435, + "step": 22012 + }, + { + "epoch": 4.0, + "learning_rate": 1.8007098653076083e-06, + "loss": 0.7084, + "step": 22013 + }, + { + "epoch": 4.0, + "learning_rate": 1.800655260283946e-06, + "loss": 0.4062, + "step": 22014 + }, + { + "epoch": 4.0, + "learning_rate": 1.800600655260284e-06, + "loss": 1.0651, + "step": 22015 + }, + { + "epoch": 4.0, + "learning_rate": 1.800546050236622e-06, + "loss": 0.0142, + "step": 22016 + }, + { + "epoch": 4.0, + "learning_rate": 1.8004914452129595e-06, + "loss": 0.3138, + "step": 22017 + }, + { + "epoch": 4.0, + "learning_rate": 1.8004368401892974e-06, + "loss": 0.0581, + "step": 22018 + }, + { + "epoch": 4.0, + "learning_rate": 1.8003822351656354e-06, + "loss": 0.1985, + "step": 22019 + }, + { + "epoch": 4.0, + "learning_rate": 1.8003276301419729e-06, + "loss": 0.1035, + "step": 22020 + }, + { + "epoch": 4.0, + "learning_rate": 1.8002730251183108e-06, + "loss": 1.6444, + "step": 22021 + }, + { + "epoch": 4.0, + "learning_rate": 1.8002184200946488e-06, + "loss": 0.06, + "step": 22022 + }, + { + "epoch": 4.0, + "learning_rate": 1.8001638150709865e-06, + "loss": 0.7676, + "step": 22023 + }, + { + "epoch": 4.01, + "learning_rate": 1.8001092100473245e-06, + "loss": 0.0448, + "step": 22024 + }, + { + "epoch": 4.01, + "learning_rate": 1.8000546050236624e-06, + "loss": 0.6316, + "step": 22025 + }, + { + "epoch": 4.01, + "learning_rate": 1.8e-06, + "loss": 0.375, + "step": 22026 + }, + { + "epoch": 4.01, + "learning_rate": 1.7999453949763379e-06, + "loss": 0.1162, + "step": 22027 + }, + { + "epoch": 4.01, + "learning_rate": 1.7998907899526756e-06, + "loss": 0.5852, + "step": 22028 + }, + { + "epoch": 4.01, + "learning_rate": 1.7998361849290135e-06, + "loss": 1.0012, + "step": 22029 + }, + { + "epoch": 4.01, + "learning_rate": 1.7997815799053515e-06, + "loss": 0.8135, + "step": 22030 + }, + { + "epoch": 4.01, + "learning_rate": 1.799726974881689e-06, + "loss": 1.8336, + "step": 22031 + }, + { + "epoch": 4.01, + "learning_rate": 1.799672369858027e-06, + "loss": 1.7401, + "step": 22032 + }, + { + "epoch": 4.01, + "learning_rate": 1.7996177648343649e-06, + "loss": 0.1777, + "step": 22033 + }, + { + "epoch": 4.01, + "learning_rate": 1.7995631598107026e-06, + "loss": 1.7492, + "step": 22034 + }, + { + "epoch": 4.01, + "learning_rate": 1.7995085547870406e-06, + "loss": 0.3992, + "step": 22035 + }, + { + "epoch": 4.01, + "learning_rate": 1.7994539497633785e-06, + "loss": 0.8695, + "step": 22036 + }, + { + "epoch": 4.01, + "learning_rate": 1.799399344739716e-06, + "loss": 0.3311, + "step": 22037 + }, + { + "epoch": 4.01, + "learning_rate": 1.799344739716054e-06, + "loss": 0.8741, + "step": 22038 + }, + { + "epoch": 4.01, + "learning_rate": 1.799290134692392e-06, + "loss": 1.5765, + "step": 22039 + }, + { + "epoch": 4.01, + "learning_rate": 1.7992355296687294e-06, + "loss": 0.0395, + "step": 22040 + }, + { + "epoch": 4.01, + "learning_rate": 1.7991809246450674e-06, + "loss": 0.6599, + "step": 22041 + }, + { + "epoch": 4.01, + "learning_rate": 1.7991263196214051e-06, + "loss": 0.4222, + "step": 22042 + }, + { + "epoch": 4.01, + "learning_rate": 1.799071714597743e-06, + "loss": 1.9515, + "step": 22043 + }, + { + "epoch": 4.01, + "learning_rate": 1.799017109574081e-06, + "loss": 0.0235, + "step": 22044 + }, + { + "epoch": 4.01, + "learning_rate": 1.7989625045504185e-06, + "loss": 0.6159, + "step": 22045 + }, + { + "epoch": 4.01, + "learning_rate": 1.7989078995267565e-06, + "loss": 0.0552, + "step": 22046 + }, + { + "epoch": 4.01, + "learning_rate": 1.7988532945030944e-06, + "loss": 0.0297, + "step": 22047 + }, + { + "epoch": 4.01, + "learning_rate": 1.7987986894794321e-06, + "loss": 0.8302, + "step": 22048 + }, + { + "epoch": 4.01, + "learning_rate": 1.79874408445577e-06, + "loss": 0.6249, + "step": 22049 + }, + { + "epoch": 4.01, + "learning_rate": 1.798689479432108e-06, + "loss": 0.4884, + "step": 22050 + }, + { + "epoch": 4.01, + "learning_rate": 1.7986348744084455e-06, + "loss": 0.3834, + "step": 22051 + }, + { + "epoch": 4.01, + "learning_rate": 1.7985802693847835e-06, + "loss": 0.4754, + "step": 22052 + }, + { + "epoch": 4.01, + "learning_rate": 1.7985256643611214e-06, + "loss": 1.1442, + "step": 22053 + }, + { + "epoch": 4.01, + "learning_rate": 1.798471059337459e-06, + "loss": 1.3431, + "step": 22054 + }, + { + "epoch": 4.01, + "learning_rate": 1.7984164543137969e-06, + "loss": 1.6505, + "step": 22055 + }, + { + "epoch": 4.01, + "learning_rate": 1.7983618492901346e-06, + "loss": 0.0526, + "step": 22056 + }, + { + "epoch": 4.01, + "learning_rate": 1.7983072442664726e-06, + "loss": 0.8854, + "step": 22057 + }, + { + "epoch": 4.01, + "learning_rate": 1.7982526392428105e-06, + "loss": 0.1781, + "step": 22058 + }, + { + "epoch": 4.01, + "learning_rate": 1.798198034219148e-06, + "loss": 1.3531, + "step": 22059 + }, + { + "epoch": 4.01, + "learning_rate": 1.798143429195486e-06, + "loss": 0.6035, + "step": 22060 + }, + { + "epoch": 4.01, + "learning_rate": 1.798088824171824e-06, + "loss": 0.5235, + "step": 22061 + }, + { + "epoch": 4.01, + "learning_rate": 1.7980342191481616e-06, + "loss": 0.6463, + "step": 22062 + }, + { + "epoch": 4.01, + "learning_rate": 1.7979796141244996e-06, + "loss": 0.5464, + "step": 22063 + }, + { + "epoch": 4.01, + "learning_rate": 1.7979250091008375e-06, + "loss": 0.8444, + "step": 22064 + }, + { + "epoch": 4.01, + "learning_rate": 1.797870404077175e-06, + "loss": 0.3202, + "step": 22065 + }, + { + "epoch": 4.01, + "learning_rate": 1.797815799053513e-06, + "loss": 1.4267, + "step": 22066 + }, + { + "epoch": 4.01, + "learning_rate": 1.797761194029851e-06, + "loss": 0.261, + "step": 22067 + }, + { + "epoch": 4.01, + "learning_rate": 1.7977065890061885e-06, + "loss": 1.1627, + "step": 22068 + }, + { + "epoch": 4.01, + "learning_rate": 1.7976519839825264e-06, + "loss": 0.0391, + "step": 22069 + }, + { + "epoch": 4.01, + "learning_rate": 1.7975973789588641e-06, + "loss": 0.6983, + "step": 22070 + }, + { + "epoch": 4.01, + "learning_rate": 1.797542773935202e-06, + "loss": 0.3047, + "step": 22071 + }, + { + "epoch": 4.01, + "learning_rate": 1.79748816891154e-06, + "loss": 1.4014, + "step": 22072 + }, + { + "epoch": 4.01, + "learning_rate": 1.7974335638878775e-06, + "loss": 0.3122, + "step": 22073 + }, + { + "epoch": 4.01, + "learning_rate": 1.7973789588642155e-06, + "loss": 0.3258, + "step": 22074 + }, + { + "epoch": 4.01, + "learning_rate": 1.7973243538405534e-06, + "loss": 0.1082, + "step": 22075 + }, + { + "epoch": 4.01, + "learning_rate": 1.7972697488168912e-06, + "loss": 0.1535, + "step": 22076 + }, + { + "epoch": 4.01, + "learning_rate": 1.797215143793229e-06, + "loss": 0.1161, + "step": 22077 + }, + { + "epoch": 4.01, + "learning_rate": 1.797160538769567e-06, + "loss": 0.7464, + "step": 22078 + }, + { + "epoch": 4.02, + "learning_rate": 1.7971059337459046e-06, + "loss": 0.0167, + "step": 22079 + }, + { + "epoch": 4.02, + "learning_rate": 1.7970513287222425e-06, + "loss": 0.3625, + "step": 22080 + }, + { + "epoch": 4.02, + "learning_rate": 1.7969967236985805e-06, + "loss": 0.9613, + "step": 22081 + }, + { + "epoch": 4.02, + "learning_rate": 1.7969421186749182e-06, + "loss": 0.8721, + "step": 22082 + }, + { + "epoch": 4.02, + "learning_rate": 1.7968875136512561e-06, + "loss": 0.3629, + "step": 22083 + }, + { + "epoch": 4.02, + "learning_rate": 1.7968329086275936e-06, + "loss": 1.1799, + "step": 22084 + }, + { + "epoch": 4.02, + "learning_rate": 1.7967783036039316e-06, + "loss": 0.181, + "step": 22085 + }, + { + "epoch": 4.02, + "learning_rate": 1.7967236985802695e-06, + "loss": 0.3396, + "step": 22086 + }, + { + "epoch": 4.02, + "learning_rate": 1.796669093556607e-06, + "loss": 0.0111, + "step": 22087 + }, + { + "epoch": 4.02, + "learning_rate": 1.796614488532945e-06, + "loss": 1.8564, + "step": 22088 + }, + { + "epoch": 4.02, + "learning_rate": 1.796559883509283e-06, + "loss": 0.0877, + "step": 22089 + }, + { + "epoch": 4.02, + "learning_rate": 1.7965052784856207e-06, + "loss": 0.1147, + "step": 22090 + }, + { + "epoch": 4.02, + "learning_rate": 1.7964506734619586e-06, + "loss": 0.8696, + "step": 22091 + }, + { + "epoch": 4.02, + "learning_rate": 1.7963960684382966e-06, + "loss": 0.0942, + "step": 22092 + }, + { + "epoch": 4.02, + "learning_rate": 1.796341463414634e-06, + "loss": 0.0162, + "step": 22093 + }, + { + "epoch": 4.02, + "learning_rate": 1.796286858390972e-06, + "loss": 0.024, + "step": 22094 + }, + { + "epoch": 4.02, + "learning_rate": 1.7962322533673098e-06, + "loss": 0.0587, + "step": 22095 + }, + { + "epoch": 4.02, + "learning_rate": 1.7961776483436477e-06, + "loss": 1.5185, + "step": 22096 + }, + { + "epoch": 4.02, + "learning_rate": 1.7961230433199856e-06, + "loss": 0.7078, + "step": 22097 + }, + { + "epoch": 4.02, + "learning_rate": 1.7960684382963232e-06, + "loss": 0.7116, + "step": 22098 + }, + { + "epoch": 4.02, + "learning_rate": 1.796013833272661e-06, + "loss": 0.2724, + "step": 22099 + }, + { + "epoch": 4.02, + "learning_rate": 1.795959228248999e-06, + "loss": 0.8652, + "step": 22100 + }, + { + "epoch": 4.02, + "learning_rate": 1.7959046232253366e-06, + "loss": 0.8731, + "step": 22101 + }, + { + "epoch": 4.02, + "learning_rate": 1.7958500182016745e-06, + "loss": 0.1873, + "step": 22102 + }, + { + "epoch": 4.02, + "learning_rate": 1.7957954131780125e-06, + "loss": 1.3323, + "step": 22103 + }, + { + "epoch": 4.02, + "learning_rate": 1.7957408081543502e-06, + "loss": 1.5771, + "step": 22104 + }, + { + "epoch": 4.02, + "learning_rate": 1.7956862031306881e-06, + "loss": 0.4811, + "step": 22105 + }, + { + "epoch": 4.02, + "learning_rate": 1.795631598107026e-06, + "loss": 0.7839, + "step": 22106 + }, + { + "epoch": 4.02, + "learning_rate": 1.7955769930833636e-06, + "loss": 0.396, + "step": 22107 + }, + { + "epoch": 4.02, + "learning_rate": 1.7955223880597015e-06, + "loss": 2.1145, + "step": 22108 + }, + { + "epoch": 4.02, + "learning_rate": 1.7954677830360393e-06, + "loss": 0.2, + "step": 22109 + }, + { + "epoch": 4.02, + "learning_rate": 1.7954131780123772e-06, + "loss": 0.5199, + "step": 22110 + }, + { + "epoch": 4.02, + "learning_rate": 1.7953585729887152e-06, + "loss": 0.0993, + "step": 22111 + }, + { + "epoch": 4.02, + "learning_rate": 1.7953039679650527e-06, + "loss": 0.6074, + "step": 22112 + }, + { + "epoch": 4.02, + "learning_rate": 1.7952493629413906e-06, + "loss": 0.0381, + "step": 22113 + }, + { + "epoch": 4.02, + "learning_rate": 1.7951947579177286e-06, + "loss": 0.3237, + "step": 22114 + }, + { + "epoch": 4.02, + "learning_rate": 1.7951401528940663e-06, + "loss": 0.3897, + "step": 22115 + }, + { + "epoch": 4.02, + "learning_rate": 1.7950855478704042e-06, + "loss": 2.6812, + "step": 22116 + }, + { + "epoch": 4.02, + "learning_rate": 1.795030942846742e-06, + "loss": 1.1805, + "step": 22117 + }, + { + "epoch": 4.02, + "learning_rate": 1.7949763378230797e-06, + "loss": 0.4837, + "step": 22118 + }, + { + "epoch": 4.02, + "learning_rate": 1.7949217327994176e-06, + "loss": 0.0392, + "step": 22119 + }, + { + "epoch": 4.02, + "learning_rate": 1.7948671277757556e-06, + "loss": 1.2782, + "step": 22120 + }, + { + "epoch": 4.02, + "learning_rate": 1.7948125227520931e-06, + "loss": 0.9867, + "step": 22121 + }, + { + "epoch": 4.02, + "learning_rate": 1.794757917728431e-06, + "loss": 1.7554, + "step": 22122 + }, + { + "epoch": 4.02, + "learning_rate": 1.7947033127047688e-06, + "loss": 1.1004, + "step": 22123 + }, + { + "epoch": 4.02, + "learning_rate": 1.7946487076811067e-06, + "loss": 0.1447, + "step": 22124 + }, + { + "epoch": 4.02, + "learning_rate": 1.7945941026574447e-06, + "loss": 0.5828, + "step": 22125 + }, + { + "epoch": 4.02, + "learning_rate": 1.7945394976337822e-06, + "loss": 0.0325, + "step": 22126 + }, + { + "epoch": 4.02, + "learning_rate": 1.7944848926101201e-06, + "loss": 0.3491, + "step": 22127 + }, + { + "epoch": 4.02, + "learning_rate": 1.794430287586458e-06, + "loss": 2.4535, + "step": 22128 + }, + { + "epoch": 4.02, + "learning_rate": 1.7943756825627958e-06, + "loss": 0.6795, + "step": 22129 + }, + { + "epoch": 4.02, + "learning_rate": 1.7943210775391337e-06, + "loss": 1.7885, + "step": 22130 + }, + { + "epoch": 4.02, + "learning_rate": 1.7942664725154717e-06, + "loss": 0.2872, + "step": 22131 + }, + { + "epoch": 4.02, + "learning_rate": 1.7942118674918092e-06, + "loss": 1.7881, + "step": 22132 + }, + { + "epoch": 4.02, + "learning_rate": 1.7941572624681472e-06, + "loss": 0.5415, + "step": 22133 + }, + { + "epoch": 4.03, + "learning_rate": 1.794102657444485e-06, + "loss": 0.0165, + "step": 22134 + }, + { + "epoch": 4.03, + "learning_rate": 1.7940480524208226e-06, + "loss": 0.5135, + "step": 22135 + }, + { + "epoch": 4.03, + "learning_rate": 1.7939934473971606e-06, + "loss": 0.4632, + "step": 22136 + }, + { + "epoch": 4.03, + "learning_rate": 1.7939388423734983e-06, + "loss": 0.4009, + "step": 22137 + }, + { + "epoch": 4.03, + "learning_rate": 1.7938842373498362e-06, + "loss": 0.7336, + "step": 22138 + }, + { + "epoch": 4.03, + "learning_rate": 1.7938296323261742e-06, + "loss": 0.0234, + "step": 22139 + }, + { + "epoch": 4.03, + "learning_rate": 1.7937750273025117e-06, + "loss": 0.0231, + "step": 22140 + }, + { + "epoch": 4.03, + "learning_rate": 1.7937204222788496e-06, + "loss": 0.0111, + "step": 22141 + }, + { + "epoch": 4.03, + "learning_rate": 1.7936658172551876e-06, + "loss": 1.4743, + "step": 22142 + }, + { + "epoch": 4.03, + "learning_rate": 1.7936112122315253e-06, + "loss": 0.0087, + "step": 22143 + }, + { + "epoch": 4.03, + "learning_rate": 1.7935566072078633e-06, + "loss": 0.0206, + "step": 22144 + }, + { + "epoch": 4.03, + "learning_rate": 1.7935020021842012e-06, + "loss": 0.2785, + "step": 22145 + }, + { + "epoch": 4.03, + "learning_rate": 1.7934473971605387e-06, + "loss": 1.5625, + "step": 22146 + }, + { + "epoch": 4.03, + "learning_rate": 1.7933927921368767e-06, + "loss": 0.5395, + "step": 22147 + }, + { + "epoch": 4.03, + "learning_rate": 1.7933381871132146e-06, + "loss": 2.0348, + "step": 22148 + }, + { + "epoch": 4.03, + "learning_rate": 1.7932835820895521e-06, + "loss": 0.4201, + "step": 22149 + }, + { + "epoch": 4.03, + "learning_rate": 1.79322897706589e-06, + "loss": 0.2912, + "step": 22150 + }, + { + "epoch": 4.03, + "learning_rate": 1.7931743720422278e-06, + "loss": 0.2354, + "step": 22151 + }, + { + "epoch": 4.03, + "learning_rate": 1.7931197670185658e-06, + "loss": 1.3274, + "step": 22152 + }, + { + "epoch": 4.03, + "learning_rate": 1.7930651619949037e-06, + "loss": 0.097, + "step": 22153 + }, + { + "epoch": 4.03, + "learning_rate": 1.7930105569712412e-06, + "loss": 2.8077, + "step": 22154 + }, + { + "epoch": 4.03, + "learning_rate": 1.7929559519475792e-06, + "loss": 0.0085, + "step": 22155 + }, + { + "epoch": 4.03, + "learning_rate": 1.792901346923917e-06, + "loss": 0.12, + "step": 22156 + }, + { + "epoch": 4.03, + "learning_rate": 1.7928467419002548e-06, + "loss": 0.0713, + "step": 22157 + }, + { + "epoch": 4.03, + "learning_rate": 1.7927921368765928e-06, + "loss": 0.9194, + "step": 22158 + }, + { + "epoch": 4.03, + "learning_rate": 1.7927375318529307e-06, + "loss": 0.702, + "step": 22159 + }, + { + "epoch": 4.03, + "learning_rate": 1.7926829268292682e-06, + "loss": 0.2861, + "step": 22160 + }, + { + "epoch": 4.03, + "learning_rate": 1.7926283218056062e-06, + "loss": 0.9394, + "step": 22161 + }, + { + "epoch": 4.03, + "learning_rate": 1.7925737167819441e-06, + "loss": 1.6667, + "step": 22162 + }, + { + "epoch": 4.03, + "learning_rate": 1.7925191117582819e-06, + "loss": 0.9338, + "step": 22163 + }, + { + "epoch": 4.03, + "learning_rate": 1.7924645067346198e-06, + "loss": 0.8619, + "step": 22164 + }, + { + "epoch": 4.03, + "learning_rate": 1.7924099017109573e-06, + "loss": 1.3477, + "step": 22165 + }, + { + "epoch": 4.03, + "learning_rate": 1.7923552966872953e-06, + "loss": 0.015, + "step": 22166 + }, + { + "epoch": 4.03, + "learning_rate": 1.7923006916636332e-06, + "loss": 0.0126, + "step": 22167 + }, + { + "epoch": 4.03, + "learning_rate": 1.7922460866399707e-06, + "loss": 0.4886, + "step": 22168 + }, + { + "epoch": 4.03, + "learning_rate": 1.7921914816163087e-06, + "loss": 0.2043, + "step": 22169 + }, + { + "epoch": 4.03, + "learning_rate": 1.7921368765926466e-06, + "loss": 0.2877, + "step": 22170 + }, + { + "epoch": 4.03, + "learning_rate": 1.7920822715689843e-06, + "loss": 0.3417, + "step": 22171 + }, + { + "epoch": 4.03, + "learning_rate": 1.7920276665453223e-06, + "loss": 0.0164, + "step": 22172 + }, + { + "epoch": 4.03, + "learning_rate": 1.7919730615216602e-06, + "loss": 2.8029, + "step": 22173 + }, + { + "epoch": 4.03, + "learning_rate": 1.7919184564979978e-06, + "loss": 0.0854, + "step": 22174 + }, + { + "epoch": 4.03, + "learning_rate": 1.7918638514743357e-06, + "loss": 0.9696, + "step": 22175 + }, + { + "epoch": 4.03, + "learning_rate": 1.7918092464506734e-06, + "loss": 0.0108, + "step": 22176 + }, + { + "epoch": 4.03, + "learning_rate": 1.7917546414270114e-06, + "loss": 0.0182, + "step": 22177 + }, + { + "epoch": 4.03, + "learning_rate": 1.7917000364033493e-06, + "loss": 0.531, + "step": 22178 + }, + { + "epoch": 4.03, + "learning_rate": 1.7916454313796868e-06, + "loss": 0.3408, + "step": 22179 + }, + { + "epoch": 4.03, + "learning_rate": 1.7915908263560248e-06, + "loss": 0.6666, + "step": 22180 + }, + { + "epoch": 4.03, + "learning_rate": 1.7915362213323627e-06, + "loss": 0.6487, + "step": 22181 + }, + { + "epoch": 4.03, + "learning_rate": 1.7914816163087002e-06, + "loss": 0.621, + "step": 22182 + }, + { + "epoch": 4.03, + "learning_rate": 1.7914270112850382e-06, + "loss": 0.3777, + "step": 22183 + }, + { + "epoch": 4.03, + "learning_rate": 1.7913724062613761e-06, + "loss": 0.1001, + "step": 22184 + }, + { + "epoch": 4.03, + "learning_rate": 1.7913178012377139e-06, + "loss": 0.2395, + "step": 22185 + }, + { + "epoch": 4.03, + "learning_rate": 1.7912631962140518e-06, + "loss": 2.0275, + "step": 22186 + }, + { + "epoch": 4.03, + "learning_rate": 1.7912085911903897e-06, + "loss": 0.5599, + "step": 22187 + }, + { + "epoch": 4.03, + "learning_rate": 1.7911539861667273e-06, + "loss": 0.6572, + "step": 22188 + }, + { + "epoch": 4.04, + "learning_rate": 1.7910993811430652e-06, + "loss": 1.0447, + "step": 22189 + }, + { + "epoch": 4.04, + "learning_rate": 1.791044776119403e-06, + "loss": 2.0856, + "step": 22190 + }, + { + "epoch": 4.04, + "learning_rate": 1.7909901710957409e-06, + "loss": 0.2999, + "step": 22191 + }, + { + "epoch": 4.04, + "learning_rate": 1.7909355660720788e-06, + "loss": 0.6314, + "step": 22192 + }, + { + "epoch": 4.04, + "learning_rate": 1.7908809610484163e-06, + "loss": 0.5001, + "step": 22193 + }, + { + "epoch": 4.04, + "learning_rate": 1.7908263560247543e-06, + "loss": 1.9156, + "step": 22194 + }, + { + "epoch": 4.04, + "learning_rate": 1.7907717510010922e-06, + "loss": 0.9891, + "step": 22195 + }, + { + "epoch": 4.04, + "learning_rate": 1.7907171459774298e-06, + "loss": 0.0862, + "step": 22196 + }, + { + "epoch": 4.04, + "learning_rate": 1.7906625409537677e-06, + "loss": 0.8175, + "step": 22197 + }, + { + "epoch": 4.04, + "learning_rate": 1.7906079359301056e-06, + "loss": 0.3341, + "step": 22198 + }, + { + "epoch": 4.04, + "learning_rate": 1.7905533309064434e-06, + "loss": 1.114, + "step": 22199 + }, + { + "epoch": 4.04, + "learning_rate": 1.7904987258827813e-06, + "loss": 1.5217, + "step": 22200 + }, + { + "epoch": 4.04, + "learning_rate": 1.7904441208591193e-06, + "loss": 0.029, + "step": 22201 + }, + { + "epoch": 4.04, + "learning_rate": 1.7903895158354568e-06, + "loss": 0.503, + "step": 22202 + }, + { + "epoch": 4.04, + "learning_rate": 1.7903349108117947e-06, + "loss": 1.6269, + "step": 22203 + }, + { + "epoch": 4.04, + "learning_rate": 1.7902803057881325e-06, + "loss": 1.4103, + "step": 22204 + }, + { + "epoch": 4.04, + "learning_rate": 1.7902257007644704e-06, + "loss": 0.2564, + "step": 22205 + }, + { + "epoch": 4.04, + "learning_rate": 1.7901710957408083e-06, + "loss": 0.6713, + "step": 22206 + }, + { + "epoch": 4.04, + "learning_rate": 1.7901164907171459e-06, + "loss": 0.2659, + "step": 22207 + }, + { + "epoch": 4.04, + "learning_rate": 1.7900618856934838e-06, + "loss": 1.2808, + "step": 22208 + }, + { + "epoch": 4.04, + "learning_rate": 1.7900072806698217e-06, + "loss": 0.9825, + "step": 22209 + }, + { + "epoch": 4.04, + "learning_rate": 1.7899526756461595e-06, + "loss": 0.13, + "step": 22210 + }, + { + "epoch": 4.04, + "learning_rate": 1.7898980706224974e-06, + "loss": 2.0169, + "step": 22211 + }, + { + "epoch": 4.04, + "learning_rate": 1.7898434655988354e-06, + "loss": 0.4284, + "step": 22212 + }, + { + "epoch": 4.04, + "learning_rate": 1.7897888605751729e-06, + "loss": 0.0138, + "step": 22213 + }, + { + "epoch": 4.04, + "learning_rate": 1.7897342555515108e-06, + "loss": 0.0913, + "step": 22214 + }, + { + "epoch": 4.04, + "learning_rate": 1.7896796505278488e-06, + "loss": 0.2277, + "step": 22215 + }, + { + "epoch": 4.04, + "learning_rate": 1.7896250455041863e-06, + "loss": 0.5307, + "step": 22216 + }, + { + "epoch": 4.04, + "learning_rate": 1.7895704404805242e-06, + "loss": 0.5077, + "step": 22217 + }, + { + "epoch": 4.04, + "learning_rate": 1.789515835456862e-06, + "loss": 0.3938, + "step": 22218 + }, + { + "epoch": 4.04, + "learning_rate": 1.7894612304332e-06, + "loss": 0.2261, + "step": 22219 + }, + { + "epoch": 4.04, + "learning_rate": 1.7894066254095379e-06, + "loss": 2.0943, + "step": 22220 + }, + { + "epoch": 4.04, + "learning_rate": 1.7893520203858754e-06, + "loss": 3.3908, + "step": 22221 + }, + { + "epoch": 4.04, + "learning_rate": 1.7892974153622133e-06, + "loss": 1.6461, + "step": 22222 + }, + { + "epoch": 4.04, + "learning_rate": 1.7892428103385513e-06, + "loss": 1.7292, + "step": 22223 + }, + { + "epoch": 4.04, + "learning_rate": 1.789188205314889e-06, + "loss": 3.372, + "step": 22224 + }, + { + "epoch": 4.04, + "learning_rate": 1.789133600291227e-06, + "loss": 0.3655, + "step": 22225 + }, + { + "epoch": 4.04, + "learning_rate": 1.7890789952675649e-06, + "loss": 0.0875, + "step": 22226 + }, + { + "epoch": 4.04, + "learning_rate": 1.7890243902439024e-06, + "loss": 0.994, + "step": 22227 + }, + { + "epoch": 4.04, + "learning_rate": 1.7889697852202403e-06, + "loss": 0.2846, + "step": 22228 + }, + { + "epoch": 4.04, + "learning_rate": 1.7889151801965783e-06, + "loss": 0.1863, + "step": 22229 + }, + { + "epoch": 4.04, + "learning_rate": 1.7888605751729158e-06, + "loss": 0.3372, + "step": 22230 + }, + { + "epoch": 4.04, + "learning_rate": 1.7888059701492538e-06, + "loss": 0.0158, + "step": 22231 + }, + { + "epoch": 4.04, + "learning_rate": 1.7887513651255915e-06, + "loss": 4.0846, + "step": 22232 + }, + { + "epoch": 4.04, + "learning_rate": 1.7886967601019294e-06, + "loss": 0.804, + "step": 22233 + }, + { + "epoch": 4.04, + "learning_rate": 1.7886421550782674e-06, + "loss": 1.4385, + "step": 22234 + }, + { + "epoch": 4.04, + "learning_rate": 1.7885875500546049e-06, + "loss": 0.6132, + "step": 22235 + }, + { + "epoch": 4.04, + "learning_rate": 1.7885329450309428e-06, + "loss": 0.8736, + "step": 22236 + }, + { + "epoch": 4.04, + "learning_rate": 1.7884783400072808e-06, + "loss": 0.2497, + "step": 22237 + }, + { + "epoch": 4.04, + "learning_rate": 1.7884237349836185e-06, + "loss": 0.202, + "step": 22238 + }, + { + "epoch": 4.04, + "learning_rate": 1.7883691299599565e-06, + "loss": 0.7853, + "step": 22239 + }, + { + "epoch": 4.04, + "learning_rate": 1.7883145249362944e-06, + "loss": 1.8304, + "step": 22240 + }, + { + "epoch": 4.04, + "learning_rate": 1.788259919912632e-06, + "loss": 2.0943, + "step": 22241 + }, + { + "epoch": 4.04, + "learning_rate": 1.7882053148889699e-06, + "loss": 0.6504, + "step": 22242 + }, + { + "epoch": 4.04, + "learning_rate": 1.7881507098653078e-06, + "loss": 1.5038, + "step": 22243 + }, + { + "epoch": 4.05, + "learning_rate": 1.7880961048416455e-06, + "loss": 0.3431, + "step": 22244 + }, + { + "epoch": 4.05, + "learning_rate": 1.7880414998179835e-06, + "loss": 0.1659, + "step": 22245 + }, + { + "epoch": 4.05, + "learning_rate": 1.787986894794321e-06, + "loss": 0.3571, + "step": 22246 + }, + { + "epoch": 4.05, + "learning_rate": 1.787932289770659e-06, + "loss": 0.089, + "step": 22247 + }, + { + "epoch": 4.05, + "learning_rate": 1.7878776847469969e-06, + "loss": 0.8413, + "step": 22248 + }, + { + "epoch": 4.05, + "learning_rate": 1.7878230797233344e-06, + "loss": 0.0161, + "step": 22249 + }, + { + "epoch": 4.05, + "learning_rate": 1.7877684746996723e-06, + "loss": 0.4677, + "step": 22250 + }, + { + "epoch": 4.05, + "learning_rate": 1.7877138696760103e-06, + "loss": 0.0201, + "step": 22251 + }, + { + "epoch": 4.05, + "learning_rate": 1.787659264652348e-06, + "loss": 0.3814, + "step": 22252 + }, + { + "epoch": 4.05, + "learning_rate": 1.787604659628686e-06, + "loss": 0.7526, + "step": 22253 + }, + { + "epoch": 4.05, + "learning_rate": 1.787550054605024e-06, + "loss": 0.1146, + "step": 22254 + }, + { + "epoch": 4.05, + "learning_rate": 1.7874954495813614e-06, + "loss": 2.1593, + "step": 22255 + }, + { + "epoch": 4.05, + "learning_rate": 1.7874408445576994e-06, + "loss": 0.7268, + "step": 22256 + }, + { + "epoch": 4.05, + "learning_rate": 1.787386239534037e-06, + "loss": 0.4862, + "step": 22257 + }, + { + "epoch": 4.05, + "learning_rate": 1.787331634510375e-06, + "loss": 0.6584, + "step": 22258 + }, + { + "epoch": 4.05, + "learning_rate": 1.787277029486713e-06, + "loss": 0.0261, + "step": 22259 + }, + { + "epoch": 4.05, + "learning_rate": 1.7872224244630505e-06, + "loss": 0.0793, + "step": 22260 + }, + { + "epoch": 4.05, + "learning_rate": 1.7871678194393885e-06, + "loss": 1.1063, + "step": 22261 + }, + { + "epoch": 4.05, + "learning_rate": 1.7871132144157264e-06, + "loss": 1.0832, + "step": 22262 + }, + { + "epoch": 4.05, + "learning_rate": 1.787058609392064e-06, + "loss": 1.5341, + "step": 22263 + }, + { + "epoch": 4.05, + "learning_rate": 1.7870040043684019e-06, + "loss": 0.0164, + "step": 22264 + }, + { + "epoch": 4.05, + "learning_rate": 1.7869493993447398e-06, + "loss": 0.2688, + "step": 22265 + }, + { + "epoch": 4.05, + "learning_rate": 1.7868947943210775e-06, + "loss": 0.4762, + "step": 22266 + }, + { + "epoch": 4.05, + "learning_rate": 1.7868401892974155e-06, + "loss": 1.4885, + "step": 22267 + }, + { + "epoch": 4.05, + "learning_rate": 1.7867855842737534e-06, + "loss": 0.1715, + "step": 22268 + }, + { + "epoch": 4.05, + "learning_rate": 1.786730979250091e-06, + "loss": 0.9456, + "step": 22269 + }, + { + "epoch": 4.05, + "learning_rate": 1.7866763742264289e-06, + "loss": 1.3543, + "step": 22270 + }, + { + "epoch": 4.05, + "learning_rate": 1.7866217692027666e-06, + "loss": 0.1893, + "step": 22271 + }, + { + "epoch": 4.05, + "learning_rate": 1.7865671641791046e-06, + "loss": 0.7153, + "step": 22272 + }, + { + "epoch": 4.05, + "learning_rate": 1.7865125591554425e-06, + "loss": 0.1122, + "step": 22273 + }, + { + "epoch": 4.05, + "learning_rate": 1.78645795413178e-06, + "loss": 0.4566, + "step": 22274 + }, + { + "epoch": 4.05, + "learning_rate": 1.786403349108118e-06, + "loss": 0.5709, + "step": 22275 + }, + { + "epoch": 4.05, + "learning_rate": 1.786348744084456e-06, + "loss": 0.4047, + "step": 22276 + }, + { + "epoch": 4.05, + "learning_rate": 1.7862941390607934e-06, + "loss": 0.5499, + "step": 22277 + }, + { + "epoch": 4.05, + "learning_rate": 1.7862395340371314e-06, + "loss": 1.3208, + "step": 22278 + }, + { + "epoch": 4.05, + "learning_rate": 1.7861849290134693e-06, + "loss": 0.3378, + "step": 22279 + }, + { + "epoch": 4.05, + "learning_rate": 1.786130323989807e-06, + "loss": 0.0929, + "step": 22280 + }, + { + "epoch": 4.05, + "learning_rate": 1.786075718966145e-06, + "loss": 0.2097, + "step": 22281 + }, + { + "epoch": 4.05, + "learning_rate": 1.786021113942483e-06, + "loss": 0.0133, + "step": 22282 + }, + { + "epoch": 4.05, + "learning_rate": 1.7859665089188205e-06, + "loss": 0.9155, + "step": 22283 + }, + { + "epoch": 4.05, + "learning_rate": 1.7859119038951584e-06, + "loss": 0.9958, + "step": 22284 + }, + { + "epoch": 4.05, + "learning_rate": 1.7858572988714961e-06, + "loss": 0.4213, + "step": 22285 + }, + { + "epoch": 4.05, + "learning_rate": 1.785802693847834e-06, + "loss": 0.7518, + "step": 22286 + }, + { + "epoch": 4.05, + "learning_rate": 1.785748088824172e-06, + "loss": 0.2374, + "step": 22287 + }, + { + "epoch": 4.05, + "learning_rate": 1.7856934838005095e-06, + "loss": 2.3953, + "step": 22288 + }, + { + "epoch": 4.05, + "learning_rate": 1.7856388787768475e-06, + "loss": 2.2299, + "step": 22289 + }, + { + "epoch": 4.05, + "learning_rate": 1.7855842737531854e-06, + "loss": 0.5239, + "step": 22290 + }, + { + "epoch": 4.05, + "learning_rate": 1.7855296687295232e-06, + "loss": 0.2127, + "step": 22291 + }, + { + "epoch": 4.05, + "learning_rate": 1.785475063705861e-06, + "loss": 0.0599, + "step": 22292 + }, + { + "epoch": 4.05, + "learning_rate": 1.785420458682199e-06, + "loss": 0.3232, + "step": 22293 + }, + { + "epoch": 4.05, + "learning_rate": 1.7853658536585366e-06, + "loss": 0.853, + "step": 22294 + }, + { + "epoch": 4.05, + "learning_rate": 1.7853112486348745e-06, + "loss": 0.2827, + "step": 22295 + }, + { + "epoch": 4.05, + "learning_rate": 1.7852566436112124e-06, + "loss": 0.5778, + "step": 22296 + }, + { + "epoch": 4.05, + "learning_rate": 1.78520203858755e-06, + "loss": 0.3334, + "step": 22297 + }, + { + "epoch": 4.05, + "learning_rate": 1.785147433563888e-06, + "loss": 0.3996, + "step": 22298 + }, + { + "epoch": 4.06, + "learning_rate": 1.7850928285402256e-06, + "loss": 0.2663, + "step": 22299 + }, + { + "epoch": 4.06, + "learning_rate": 1.7850382235165636e-06, + "loss": 3.1615, + "step": 22300 + }, + { + "epoch": 4.06, + "learning_rate": 1.7849836184929015e-06, + "loss": 0.1834, + "step": 22301 + }, + { + "epoch": 4.06, + "learning_rate": 1.784929013469239e-06, + "loss": 0.2243, + "step": 22302 + }, + { + "epoch": 4.06, + "learning_rate": 1.784874408445577e-06, + "loss": 1.0269, + "step": 22303 + }, + { + "epoch": 4.06, + "learning_rate": 1.784819803421915e-06, + "loss": 0.0351, + "step": 22304 + }, + { + "epoch": 4.06, + "learning_rate": 1.7847651983982527e-06, + "loss": 0.356, + "step": 22305 + }, + { + "epoch": 4.06, + "learning_rate": 1.7847105933745906e-06, + "loss": 1.025, + "step": 22306 + }, + { + "epoch": 4.06, + "learning_rate": 1.7846559883509286e-06, + "loss": 0.0956, + "step": 22307 + }, + { + "epoch": 4.06, + "learning_rate": 1.784601383327266e-06, + "loss": 0.0861, + "step": 22308 + }, + { + "epoch": 4.06, + "learning_rate": 1.784546778303604e-06, + "loss": 0.4735, + "step": 22309 + }, + { + "epoch": 4.06, + "learning_rate": 1.784492173279942e-06, + "loss": 0.186, + "step": 22310 + }, + { + "epoch": 4.06, + "learning_rate": 1.7844375682562795e-06, + "loss": 0.4341, + "step": 22311 + }, + { + "epoch": 4.06, + "learning_rate": 1.7843829632326174e-06, + "loss": 1.9486, + "step": 22312 + }, + { + "epoch": 4.06, + "learning_rate": 1.7843283582089552e-06, + "loss": 0.2127, + "step": 22313 + }, + { + "epoch": 4.06, + "learning_rate": 1.784273753185293e-06, + "loss": 1.0931, + "step": 22314 + }, + { + "epoch": 4.06, + "learning_rate": 1.784219148161631e-06, + "loss": 0.6184, + "step": 22315 + }, + { + "epoch": 4.06, + "learning_rate": 1.7841645431379686e-06, + "loss": 0.1638, + "step": 22316 + }, + { + "epoch": 4.06, + "learning_rate": 1.7841099381143065e-06, + "loss": 0.0211, + "step": 22317 + }, + { + "epoch": 4.06, + "learning_rate": 1.7840553330906445e-06, + "loss": 0.463, + "step": 22318 + }, + { + "epoch": 4.06, + "learning_rate": 1.7840007280669822e-06, + "loss": 1.5461, + "step": 22319 + }, + { + "epoch": 4.06, + "learning_rate": 1.7839461230433201e-06, + "loss": 0.0132, + "step": 22320 + }, + { + "epoch": 4.06, + "learning_rate": 1.783891518019658e-06, + "loss": 2.4338, + "step": 22321 + }, + { + "epoch": 4.06, + "learning_rate": 1.7838369129959956e-06, + "loss": 0.1086, + "step": 22322 + }, + { + "epoch": 4.06, + "learning_rate": 1.7837823079723335e-06, + "loss": 0.2552, + "step": 22323 + }, + { + "epoch": 4.06, + "learning_rate": 1.783727702948671e-06, + "loss": 0.3382, + "step": 22324 + }, + { + "epoch": 4.06, + "learning_rate": 1.783673097925009e-06, + "loss": 0.0156, + "step": 22325 + }, + { + "epoch": 4.06, + "learning_rate": 1.783618492901347e-06, + "loss": 0.0089, + "step": 22326 + }, + { + "epoch": 4.06, + "learning_rate": 1.7835638878776847e-06, + "loss": 0.2873, + "step": 22327 + }, + { + "epoch": 4.06, + "learning_rate": 1.7835092828540226e-06, + "loss": 0.3177, + "step": 22328 + }, + { + "epoch": 4.06, + "learning_rate": 1.7834546778303606e-06, + "loss": 0.1836, + "step": 22329 + }, + { + "epoch": 4.06, + "learning_rate": 1.783400072806698e-06, + "loss": 0.0855, + "step": 22330 + }, + { + "epoch": 4.06, + "learning_rate": 1.783345467783036e-06, + "loss": 0.023, + "step": 22331 + }, + { + "epoch": 4.06, + "learning_rate": 1.783290862759374e-06, + "loss": 0.8642, + "step": 22332 + }, + { + "epoch": 4.06, + "learning_rate": 1.7832362577357117e-06, + "loss": 1.7685, + "step": 22333 + }, + { + "epoch": 4.06, + "learning_rate": 1.7831816527120496e-06, + "loss": 0.3766, + "step": 22334 + }, + { + "epoch": 4.06, + "learning_rate": 1.7831270476883876e-06, + "loss": 0.6145, + "step": 22335 + }, + { + "epoch": 4.06, + "learning_rate": 1.783072442664725e-06, + "loss": 0.2363, + "step": 22336 + }, + { + "epoch": 4.06, + "learning_rate": 1.783017837641063e-06, + "loss": 1.1893, + "step": 22337 + }, + { + "epoch": 4.06, + "learning_rate": 1.7829632326174008e-06, + "loss": 0.4665, + "step": 22338 + }, + { + "epoch": 4.06, + "learning_rate": 1.7829086275937387e-06, + "loss": 1.5301, + "step": 22339 + }, + { + "epoch": 4.06, + "learning_rate": 1.7828540225700767e-06, + "loss": 0.8548, + "step": 22340 + }, + { + "epoch": 4.06, + "learning_rate": 1.7827994175464142e-06, + "loss": 0.8203, + "step": 22341 + }, + { + "epoch": 4.06, + "learning_rate": 1.7827448125227521e-06, + "loss": 0.2136, + "step": 22342 + }, + { + "epoch": 4.06, + "learning_rate": 1.78269020749909e-06, + "loss": 0.0216, + "step": 22343 + }, + { + "epoch": 4.06, + "learning_rate": 1.7826356024754276e-06, + "loss": 0.1741, + "step": 22344 + }, + { + "epoch": 4.06, + "learning_rate": 1.7825809974517655e-06, + "loss": 0.4036, + "step": 22345 + }, + { + "epoch": 4.06, + "learning_rate": 1.7825263924281035e-06, + "loss": 1.2701, + "step": 22346 + }, + { + "epoch": 4.06, + "learning_rate": 1.7824717874044412e-06, + "loss": 0.4605, + "step": 22347 + }, + { + "epoch": 4.06, + "learning_rate": 1.7824171823807792e-06, + "loss": 0.6137, + "step": 22348 + }, + { + "epoch": 4.06, + "learning_rate": 1.782362577357117e-06, + "loss": 0.0336, + "step": 22349 + }, + { + "epoch": 4.06, + "learning_rate": 1.7823079723334546e-06, + "loss": 0.0118, + "step": 22350 + }, + { + "epoch": 4.06, + "learning_rate": 1.7822533673097926e-06, + "loss": 0.9057, + "step": 22351 + }, + { + "epoch": 4.06, + "learning_rate": 1.7821987622861303e-06, + "loss": 0.3825, + "step": 22352 + }, + { + "epoch": 4.06, + "learning_rate": 1.7821441572624682e-06, + "loss": 0.3031, + "step": 22353 + }, + { + "epoch": 4.07, + "learning_rate": 1.7820895522388062e-06, + "loss": 1.6687, + "step": 22354 + }, + { + "epoch": 4.07, + "learning_rate": 1.7820349472151437e-06, + "loss": 2.3635, + "step": 22355 + }, + { + "epoch": 4.07, + "learning_rate": 1.7819803421914816e-06, + "loss": 0.2992, + "step": 22356 + }, + { + "epoch": 4.07, + "learning_rate": 1.7819257371678196e-06, + "loss": 0.0777, + "step": 22357 + }, + { + "epoch": 4.07, + "learning_rate": 1.781871132144157e-06, + "loss": 0.028, + "step": 22358 + }, + { + "epoch": 4.07, + "learning_rate": 1.781816527120495e-06, + "loss": 0.1804, + "step": 22359 + }, + { + "epoch": 4.07, + "learning_rate": 1.781761922096833e-06, + "loss": 1.0399, + "step": 22360 + }, + { + "epoch": 4.07, + "learning_rate": 1.7817073170731707e-06, + "loss": 2.24, + "step": 22361 + }, + { + "epoch": 4.07, + "learning_rate": 1.7816527120495087e-06, + "loss": 1.7604, + "step": 22362 + }, + { + "epoch": 4.07, + "learning_rate": 1.7815981070258466e-06, + "loss": 2.7234, + "step": 22363 + }, + { + "epoch": 4.07, + "learning_rate": 1.7815435020021841e-06, + "loss": 0.34, + "step": 22364 + }, + { + "epoch": 4.07, + "learning_rate": 1.781488896978522e-06, + "loss": 0.6812, + "step": 22365 + }, + { + "epoch": 4.07, + "learning_rate": 1.7814342919548598e-06, + "loss": 0.0127, + "step": 22366 + }, + { + "epoch": 4.07, + "learning_rate": 1.7813796869311977e-06, + "loss": 0.019, + "step": 22367 + }, + { + "epoch": 4.07, + "learning_rate": 1.7813250819075357e-06, + "loss": 0.3619, + "step": 22368 + }, + { + "epoch": 4.07, + "learning_rate": 1.7812704768838732e-06, + "loss": 1.1906, + "step": 22369 + }, + { + "epoch": 4.07, + "learning_rate": 1.7812158718602112e-06, + "loss": 0.3697, + "step": 22370 + }, + { + "epoch": 4.07, + "learning_rate": 1.781161266836549e-06, + "loss": 0.0187, + "step": 22371 + }, + { + "epoch": 4.07, + "learning_rate": 1.7811066618128868e-06, + "loss": 0.2486, + "step": 22372 + }, + { + "epoch": 4.07, + "learning_rate": 1.7810520567892248e-06, + "loss": 0.9289, + "step": 22373 + }, + { + "epoch": 4.07, + "learning_rate": 1.7809974517655627e-06, + "loss": 1.067, + "step": 22374 + }, + { + "epoch": 4.07, + "learning_rate": 1.7809428467419002e-06, + "loss": 1.3971, + "step": 22375 + }, + { + "epoch": 4.07, + "learning_rate": 1.7808882417182382e-06, + "loss": 2.4244, + "step": 22376 + }, + { + "epoch": 4.07, + "learning_rate": 1.7808336366945761e-06, + "loss": 1.6035, + "step": 22377 + }, + { + "epoch": 4.07, + "learning_rate": 1.7807790316709136e-06, + "loss": 0.5086, + "step": 22378 + }, + { + "epoch": 4.07, + "learning_rate": 1.7807244266472516e-06, + "loss": 0.2454, + "step": 22379 + }, + { + "epoch": 4.07, + "learning_rate": 1.7806698216235893e-06, + "loss": 0.3245, + "step": 22380 + }, + { + "epoch": 4.07, + "learning_rate": 1.7806152165999273e-06, + "loss": 0.1811, + "step": 22381 + }, + { + "epoch": 4.07, + "learning_rate": 1.7805606115762652e-06, + "loss": 0.0156, + "step": 22382 + }, + { + "epoch": 4.07, + "learning_rate": 1.7805060065526027e-06, + "loss": 0.5077, + "step": 22383 + }, + { + "epoch": 4.07, + "learning_rate": 1.7804514015289407e-06, + "loss": 0.0171, + "step": 22384 + }, + { + "epoch": 4.07, + "learning_rate": 1.7803967965052786e-06, + "loss": 1.5176, + "step": 22385 + }, + { + "epoch": 4.07, + "learning_rate": 1.7803421914816163e-06, + "loss": 0.088, + "step": 22386 + }, + { + "epoch": 4.07, + "learning_rate": 1.7802875864579543e-06, + "loss": 0.5207, + "step": 22387 + }, + { + "epoch": 4.07, + "learning_rate": 1.7802329814342922e-06, + "loss": 1.3879, + "step": 22388 + }, + { + "epoch": 4.07, + "learning_rate": 1.7801783764106297e-06, + "loss": 1.3492, + "step": 22389 + }, + { + "epoch": 4.07, + "learning_rate": 1.7801237713869677e-06, + "loss": 1.3658, + "step": 22390 + }, + { + "epoch": 4.07, + "learning_rate": 1.7800691663633056e-06, + "loss": 0.4646, + "step": 22391 + }, + { + "epoch": 4.07, + "learning_rate": 1.7800145613396432e-06, + "loss": 0.177, + "step": 22392 + }, + { + "epoch": 4.07, + "learning_rate": 1.779959956315981e-06, + "loss": 0.8945, + "step": 22393 + }, + { + "epoch": 4.07, + "learning_rate": 1.7799053512923188e-06, + "loss": 0.5193, + "step": 22394 + }, + { + "epoch": 4.07, + "learning_rate": 1.7798507462686568e-06, + "loss": 0.9362, + "step": 22395 + }, + { + "epoch": 4.07, + "learning_rate": 1.7797961412449947e-06, + "loss": 1.0654, + "step": 22396 + }, + { + "epoch": 4.07, + "learning_rate": 1.7797415362213322e-06, + "loss": 1.2571, + "step": 22397 + }, + { + "epoch": 4.07, + "learning_rate": 1.7796869311976702e-06, + "loss": 0.8739, + "step": 22398 + }, + { + "epoch": 4.07, + "learning_rate": 1.7796323261740081e-06, + "loss": 0.6808, + "step": 22399 + }, + { + "epoch": 4.07, + "learning_rate": 1.7795777211503459e-06, + "loss": 0.4384, + "step": 22400 + }, + { + "epoch": 4.07, + "learning_rate": 1.7795231161266838e-06, + "loss": 2.1146, + "step": 22401 + }, + { + "epoch": 4.07, + "learning_rate": 1.7794685111030217e-06, + "loss": 0.3291, + "step": 22402 + }, + { + "epoch": 4.07, + "learning_rate": 1.7794139060793593e-06, + "loss": 0.9705, + "step": 22403 + }, + { + "epoch": 4.07, + "learning_rate": 1.7793593010556972e-06, + "loss": 0.1629, + "step": 22404 + }, + { + "epoch": 4.07, + "learning_rate": 1.7793046960320347e-06, + "loss": 0.0149, + "step": 22405 + }, + { + "epoch": 4.07, + "learning_rate": 1.7792500910083727e-06, + "loss": 0.701, + "step": 22406 + }, + { + "epoch": 4.07, + "learning_rate": 1.7791954859847106e-06, + "loss": 0.6339, + "step": 22407 + }, + { + "epoch": 4.07, + "learning_rate": 1.7791408809610483e-06, + "loss": 0.2637, + "step": 22408 + }, + { + "epoch": 4.08, + "learning_rate": 1.7790862759373863e-06, + "loss": 0.1506, + "step": 22409 + }, + { + "epoch": 4.08, + "learning_rate": 1.7790316709137242e-06, + "loss": 0.272, + "step": 22410 + }, + { + "epoch": 4.08, + "learning_rate": 1.7789770658900618e-06, + "loss": 0.1806, + "step": 22411 + }, + { + "epoch": 4.08, + "learning_rate": 1.7789224608663997e-06, + "loss": 0.2485, + "step": 22412 + }, + { + "epoch": 4.08, + "learning_rate": 1.7788678558427376e-06, + "loss": 2.1225, + "step": 22413 + }, + { + "epoch": 4.08, + "learning_rate": 1.7788132508190754e-06, + "loss": 0.0614, + "step": 22414 + }, + { + "epoch": 4.08, + "learning_rate": 1.7787586457954133e-06, + "loss": 0.2674, + "step": 22415 + }, + { + "epoch": 4.08, + "learning_rate": 1.7787040407717513e-06, + "loss": 0.9978, + "step": 22416 + }, + { + "epoch": 4.08, + "learning_rate": 1.7786494357480888e-06, + "loss": 0.442, + "step": 22417 + }, + { + "epoch": 4.08, + "learning_rate": 1.7785948307244267e-06, + "loss": 0.6165, + "step": 22418 + }, + { + "epoch": 4.08, + "learning_rate": 1.7785402257007645e-06, + "loss": 0.8599, + "step": 22419 + }, + { + "epoch": 4.08, + "learning_rate": 1.7784856206771024e-06, + "loss": 0.2551, + "step": 22420 + }, + { + "epoch": 4.08, + "learning_rate": 1.7784310156534403e-06, + "loss": 0.0152, + "step": 22421 + }, + { + "epoch": 4.08, + "learning_rate": 1.7783764106297779e-06, + "loss": 0.0849, + "step": 22422 + }, + { + "epoch": 4.08, + "learning_rate": 1.7783218056061158e-06, + "loss": 0.4655, + "step": 22423 + }, + { + "epoch": 4.08, + "learning_rate": 1.7782672005824537e-06, + "loss": 1.1746, + "step": 22424 + }, + { + "epoch": 4.08, + "learning_rate": 1.7782125955587913e-06, + "loss": 0.1681, + "step": 22425 + }, + { + "epoch": 4.08, + "learning_rate": 1.7781579905351292e-06, + "loss": 1.7113, + "step": 22426 + }, + { + "epoch": 4.08, + "learning_rate": 1.7781033855114672e-06, + "loss": 0.4277, + "step": 22427 + }, + { + "epoch": 4.08, + "learning_rate": 1.7780487804878049e-06, + "loss": 0.6726, + "step": 22428 + }, + { + "epoch": 4.08, + "learning_rate": 1.7779941754641428e-06, + "loss": 0.008, + "step": 22429 + }, + { + "epoch": 4.08, + "learning_rate": 1.7779395704404808e-06, + "loss": 0.5391, + "step": 22430 + }, + { + "epoch": 4.08, + "learning_rate": 1.7778849654168183e-06, + "loss": 1.3743, + "step": 22431 + }, + { + "epoch": 4.08, + "learning_rate": 1.7778303603931562e-06, + "loss": 0.1683, + "step": 22432 + }, + { + "epoch": 4.08, + "learning_rate": 1.777775755369494e-06, + "loss": 0.0173, + "step": 22433 + }, + { + "epoch": 4.08, + "learning_rate": 1.777721150345832e-06, + "loss": 0.2514, + "step": 22434 + }, + { + "epoch": 4.08, + "learning_rate": 1.7776665453221699e-06, + "loss": 0.5276, + "step": 22435 + }, + { + "epoch": 4.08, + "learning_rate": 1.7776119402985074e-06, + "loss": 0.2561, + "step": 22436 + }, + { + "epoch": 4.08, + "learning_rate": 1.7775573352748453e-06, + "loss": 0.8927, + "step": 22437 + }, + { + "epoch": 4.08, + "learning_rate": 1.7775027302511833e-06, + "loss": 0.3064, + "step": 22438 + }, + { + "epoch": 4.08, + "learning_rate": 1.7774481252275208e-06, + "loss": 1.5217, + "step": 22439 + }, + { + "epoch": 4.08, + "learning_rate": 1.7773935202038587e-06, + "loss": 0.0376, + "step": 22440 + }, + { + "epoch": 4.08, + "learning_rate": 1.7773389151801967e-06, + "loss": 0.1402, + "step": 22441 + }, + { + "epoch": 4.08, + "learning_rate": 1.7772843101565344e-06, + "loss": 0.279, + "step": 22442 + }, + { + "epoch": 4.08, + "learning_rate": 1.7772297051328723e-06, + "loss": 2.0533, + "step": 22443 + }, + { + "epoch": 4.08, + "learning_rate": 1.7771751001092103e-06, + "loss": 0.0169, + "step": 22444 + }, + { + "epoch": 4.08, + "learning_rate": 1.7771204950855478e-06, + "loss": 0.2934, + "step": 22445 + }, + { + "epoch": 4.08, + "learning_rate": 1.7770658900618857e-06, + "loss": 1.6808, + "step": 22446 + }, + { + "epoch": 4.08, + "learning_rate": 1.7770112850382235e-06, + "loss": 1.5044, + "step": 22447 + }, + { + "epoch": 4.08, + "learning_rate": 1.7769566800145614e-06, + "loss": 0.3854, + "step": 22448 + }, + { + "epoch": 4.08, + "learning_rate": 1.7769020749908994e-06, + "loss": 0.8034, + "step": 22449 + }, + { + "epoch": 4.08, + "learning_rate": 1.7768474699672369e-06, + "loss": 0.1474, + "step": 22450 + }, + { + "epoch": 4.08, + "learning_rate": 1.7767928649435748e-06, + "loss": 0.57, + "step": 22451 + }, + { + "epoch": 4.08, + "learning_rate": 1.7767382599199128e-06, + "loss": 0.0632, + "step": 22452 + }, + { + "epoch": 4.08, + "learning_rate": 1.7766836548962505e-06, + "loss": 0.0728, + "step": 22453 + }, + { + "epoch": 4.08, + "learning_rate": 1.7766290498725882e-06, + "loss": 0.4124, + "step": 22454 + }, + { + "epoch": 4.08, + "learning_rate": 1.7765744448489262e-06, + "loss": 0.1745, + "step": 22455 + }, + { + "epoch": 4.08, + "learning_rate": 1.776519839825264e-06, + "loss": 0.0063, + "step": 22456 + }, + { + "epoch": 4.08, + "learning_rate": 1.7764652348016019e-06, + "loss": 0.0804, + "step": 22457 + }, + { + "epoch": 4.08, + "learning_rate": 1.7764106297779398e-06, + "loss": 0.0067, + "step": 22458 + }, + { + "epoch": 4.08, + "learning_rate": 1.7763560247542773e-06, + "loss": 0.0312, + "step": 22459 + }, + { + "epoch": 4.08, + "learning_rate": 1.7763014197306153e-06, + "loss": 1.6083, + "step": 22460 + }, + { + "epoch": 4.08, + "learning_rate": 1.776246814706953e-06, + "loss": 0.0751, + "step": 22461 + }, + { + "epoch": 4.08, + "learning_rate": 1.776192209683291e-06, + "loss": 1.8404, + "step": 22462 + }, + { + "epoch": 4.08, + "learning_rate": 1.7761376046596289e-06, + "loss": 1.295, + "step": 22463 + }, + { + "epoch": 4.09, + "learning_rate": 1.7760829996359664e-06, + "loss": 1.7037, + "step": 22464 + }, + { + "epoch": 4.09, + "learning_rate": 1.7760283946123043e-06, + "loss": 0.0139, + "step": 22465 + }, + { + "epoch": 4.09, + "learning_rate": 1.7759737895886423e-06, + "loss": 0.9845, + "step": 22466 + }, + { + "epoch": 4.09, + "learning_rate": 1.77591918456498e-06, + "loss": 0.0275, + "step": 22467 + }, + { + "epoch": 4.09, + "learning_rate": 1.775864579541318e-06, + "loss": 1.7897, + "step": 22468 + }, + { + "epoch": 4.09, + "learning_rate": 1.775809974517656e-06, + "loss": 2.0238, + "step": 22469 + }, + { + "epoch": 4.09, + "learning_rate": 1.7757553694939934e-06, + "loss": 0.2269, + "step": 22470 + }, + { + "epoch": 4.09, + "learning_rate": 1.7757007644703314e-06, + "loss": 1.5532, + "step": 22471 + }, + { + "epoch": 4.09, + "learning_rate": 1.7756461594466693e-06, + "loss": 0.3373, + "step": 22472 + }, + { + "epoch": 4.09, + "learning_rate": 1.7755915544230068e-06, + "loss": 1.5796, + "step": 22473 + }, + { + "epoch": 4.09, + "learning_rate": 1.7755369493993448e-06, + "loss": 1.375, + "step": 22474 + }, + { + "epoch": 4.09, + "learning_rate": 1.7754823443756825e-06, + "loss": 0.5888, + "step": 22475 + }, + { + "epoch": 4.09, + "learning_rate": 1.7754277393520204e-06, + "loss": 0.7224, + "step": 22476 + }, + { + "epoch": 4.09, + "learning_rate": 1.7753731343283584e-06, + "loss": 0.6325, + "step": 22477 + }, + { + "epoch": 4.09, + "learning_rate": 1.775318529304696e-06, + "loss": 0.0735, + "step": 22478 + }, + { + "epoch": 4.09, + "learning_rate": 1.7752639242810339e-06, + "loss": 1.2914, + "step": 22479 + }, + { + "epoch": 4.09, + "learning_rate": 1.7752093192573718e-06, + "loss": 1.4982, + "step": 22480 + }, + { + "epoch": 4.09, + "learning_rate": 1.7751547142337095e-06, + "loss": 0.2964, + "step": 22481 + }, + { + "epoch": 4.09, + "learning_rate": 1.7751001092100475e-06, + "loss": 1.0554, + "step": 22482 + }, + { + "epoch": 4.09, + "learning_rate": 1.7750455041863854e-06, + "loss": 0.2047, + "step": 22483 + }, + { + "epoch": 4.09, + "learning_rate": 1.774990899162723e-06, + "loss": 1.4382, + "step": 22484 + }, + { + "epoch": 4.09, + "learning_rate": 1.7749362941390609e-06, + "loss": 0.0148, + "step": 22485 + }, + { + "epoch": 4.09, + "learning_rate": 1.7748816891153984e-06, + "loss": 1.7592, + "step": 22486 + }, + { + "epoch": 4.09, + "learning_rate": 1.7748270840917363e-06, + "loss": 0.4352, + "step": 22487 + }, + { + "epoch": 4.09, + "learning_rate": 1.7747724790680743e-06, + "loss": 1.0162, + "step": 22488 + }, + { + "epoch": 4.09, + "learning_rate": 1.774717874044412e-06, + "loss": 1.291, + "step": 22489 + }, + { + "epoch": 4.09, + "learning_rate": 1.77466326902075e-06, + "loss": 1.9668, + "step": 22490 + }, + { + "epoch": 4.09, + "learning_rate": 1.774608663997088e-06, + "loss": 0.7189, + "step": 22491 + }, + { + "epoch": 4.09, + "learning_rate": 1.7745540589734254e-06, + "loss": 0.7019, + "step": 22492 + }, + { + "epoch": 4.09, + "learning_rate": 1.7744994539497634e-06, + "loss": 0.8429, + "step": 22493 + }, + { + "epoch": 4.09, + "learning_rate": 1.7744448489261013e-06, + "loss": 0.4579, + "step": 22494 + }, + { + "epoch": 4.09, + "learning_rate": 1.774390243902439e-06, + "loss": 0.2851, + "step": 22495 + }, + { + "epoch": 4.09, + "learning_rate": 1.774335638878777e-06, + "loss": 0.5709, + "step": 22496 + }, + { + "epoch": 4.09, + "learning_rate": 1.774281033855115e-06, + "loss": 1.8483, + "step": 22497 + }, + { + "epoch": 4.09, + "learning_rate": 1.7742264288314525e-06, + "loss": 0.0739, + "step": 22498 + }, + { + "epoch": 4.09, + "learning_rate": 1.7741718238077904e-06, + "loss": 1.3818, + "step": 22499 + }, + { + "epoch": 4.09, + "learning_rate": 1.7741172187841281e-06, + "loss": 0.0194, + "step": 22500 + }, + { + "epoch": 4.09, + "learning_rate": 1.774062613760466e-06, + "loss": 0.022, + "step": 22501 + }, + { + "epoch": 4.09, + "learning_rate": 1.774008008736804e-06, + "loss": 0.0087, + "step": 22502 + }, + { + "epoch": 4.09, + "learning_rate": 1.7739534037131415e-06, + "loss": 1.986, + "step": 22503 + }, + { + "epoch": 4.09, + "learning_rate": 1.7738987986894795e-06, + "loss": 0.0053, + "step": 22504 + }, + { + "epoch": 4.09, + "learning_rate": 1.7738441936658174e-06, + "loss": 0.7072, + "step": 22505 + }, + { + "epoch": 4.09, + "learning_rate": 1.773789588642155e-06, + "loss": 2.2099, + "step": 22506 + }, + { + "epoch": 4.09, + "learning_rate": 1.7737349836184929e-06, + "loss": 1.4236, + "step": 22507 + }, + { + "epoch": 4.09, + "learning_rate": 1.7736803785948308e-06, + "loss": 1.1186, + "step": 22508 + }, + { + "epoch": 4.09, + "learning_rate": 1.7736257735711686e-06, + "loss": 1.8407, + "step": 22509 + }, + { + "epoch": 4.09, + "learning_rate": 1.7735711685475065e-06, + "loss": 1.3187, + "step": 22510 + }, + { + "epoch": 4.09, + "learning_rate": 1.7735165635238444e-06, + "loss": 1.879, + "step": 22511 + }, + { + "epoch": 4.09, + "learning_rate": 1.773461958500182e-06, + "loss": 0.0428, + "step": 22512 + }, + { + "epoch": 4.09, + "learning_rate": 1.77340735347652e-06, + "loss": 0.6434, + "step": 22513 + }, + { + "epoch": 4.09, + "learning_rate": 1.7733527484528576e-06, + "loss": 0.8901, + "step": 22514 + }, + { + "epoch": 4.09, + "learning_rate": 1.7732981434291956e-06, + "loss": 0.2406, + "step": 22515 + }, + { + "epoch": 4.09, + "learning_rate": 1.7732435384055335e-06, + "loss": 0.7269, + "step": 22516 + }, + { + "epoch": 4.09, + "learning_rate": 1.773188933381871e-06, + "loss": 0.1617, + "step": 22517 + }, + { + "epoch": 4.09, + "learning_rate": 1.773134328358209e-06, + "loss": 0.0497, + "step": 22518 + }, + { + "epoch": 4.1, + "learning_rate": 1.773079723334547e-06, + "loss": 0.9311, + "step": 22519 + }, + { + "epoch": 4.1, + "learning_rate": 1.7730251183108845e-06, + "loss": 1.2117, + "step": 22520 + }, + { + "epoch": 4.1, + "learning_rate": 1.7729705132872224e-06, + "loss": 0.2441, + "step": 22521 + }, + { + "epoch": 4.1, + "learning_rate": 1.7729159082635603e-06, + "loss": 0.1089, + "step": 22522 + }, + { + "epoch": 4.1, + "learning_rate": 1.772861303239898e-06, + "loss": 1.2205, + "step": 22523 + }, + { + "epoch": 4.1, + "learning_rate": 1.772806698216236e-06, + "loss": 1.6613, + "step": 22524 + }, + { + "epoch": 4.1, + "learning_rate": 1.772752093192574e-06, + "loss": 0.4927, + "step": 22525 + }, + { + "epoch": 4.1, + "learning_rate": 1.7726974881689115e-06, + "loss": 0.8141, + "step": 22526 + }, + { + "epoch": 4.1, + "learning_rate": 1.7726428831452494e-06, + "loss": 0.2078, + "step": 22527 + }, + { + "epoch": 4.1, + "learning_rate": 1.7725882781215872e-06, + "loss": 0.4516, + "step": 22528 + }, + { + "epoch": 4.1, + "learning_rate": 1.772533673097925e-06, + "loss": 0.4246, + "step": 22529 + }, + { + "epoch": 4.1, + "learning_rate": 1.772479068074263e-06, + "loss": 0.8631, + "step": 22530 + }, + { + "epoch": 4.1, + "learning_rate": 1.7724244630506006e-06, + "loss": 2.2866, + "step": 22531 + }, + { + "epoch": 4.1, + "learning_rate": 1.7723698580269385e-06, + "loss": 0.9093, + "step": 22532 + }, + { + "epoch": 4.1, + "learning_rate": 1.7723152530032764e-06, + "loss": 0.5508, + "step": 22533 + }, + { + "epoch": 4.1, + "learning_rate": 1.772260647979614e-06, + "loss": 0.1656, + "step": 22534 + }, + { + "epoch": 4.1, + "learning_rate": 1.772206042955952e-06, + "loss": 0.9147, + "step": 22535 + }, + { + "epoch": 4.1, + "learning_rate": 1.7721514379322899e-06, + "loss": 0.0931, + "step": 22536 + }, + { + "epoch": 4.1, + "learning_rate": 1.7720968329086276e-06, + "loss": 0.4229, + "step": 22537 + }, + { + "epoch": 4.1, + "learning_rate": 1.7720422278849655e-06, + "loss": 2.7446, + "step": 22538 + }, + { + "epoch": 4.1, + "learning_rate": 1.7719876228613035e-06, + "loss": 1.4707, + "step": 22539 + }, + { + "epoch": 4.1, + "learning_rate": 1.771933017837641e-06, + "loss": 0.0076, + "step": 22540 + }, + { + "epoch": 4.1, + "learning_rate": 1.771878412813979e-06, + "loss": 0.3499, + "step": 22541 + }, + { + "epoch": 4.1, + "learning_rate": 1.7718238077903167e-06, + "loss": 0.0571, + "step": 22542 + }, + { + "epoch": 4.1, + "learning_rate": 1.7717692027666546e-06, + "loss": 0.4947, + "step": 22543 + }, + { + "epoch": 4.1, + "learning_rate": 1.7717145977429926e-06, + "loss": 1.5134, + "step": 22544 + }, + { + "epoch": 4.1, + "learning_rate": 1.77165999271933e-06, + "loss": 0.388, + "step": 22545 + }, + { + "epoch": 4.1, + "learning_rate": 1.771605387695668e-06, + "loss": 0.0213, + "step": 22546 + }, + { + "epoch": 4.1, + "learning_rate": 1.771550782672006e-06, + "loss": 0.2237, + "step": 22547 + }, + { + "epoch": 4.1, + "learning_rate": 1.7714961776483437e-06, + "loss": 0.8538, + "step": 22548 + }, + { + "epoch": 4.1, + "learning_rate": 1.7714415726246816e-06, + "loss": 0.7995, + "step": 22549 + }, + { + "epoch": 4.1, + "learning_rate": 1.7713869676010196e-06, + "loss": 3.2798, + "step": 22550 + }, + { + "epoch": 4.1, + "learning_rate": 1.771332362577357e-06, + "loss": 0.1336, + "step": 22551 + }, + { + "epoch": 4.1, + "learning_rate": 1.771277757553695e-06, + "loss": 1.4864, + "step": 22552 + }, + { + "epoch": 4.1, + "learning_rate": 1.771223152530033e-06, + "loss": 0.0102, + "step": 22553 + }, + { + "epoch": 4.1, + "learning_rate": 1.7711685475063705e-06, + "loss": 1.3194, + "step": 22554 + }, + { + "epoch": 4.1, + "learning_rate": 1.7711139424827084e-06, + "loss": 0.9587, + "step": 22555 + }, + { + "epoch": 4.1, + "learning_rate": 1.7710593374590462e-06, + "loss": 0.3346, + "step": 22556 + }, + { + "epoch": 4.1, + "learning_rate": 1.7710047324353841e-06, + "loss": 0.2035, + "step": 22557 + }, + { + "epoch": 4.1, + "learning_rate": 1.770950127411722e-06, + "loss": 0.5022, + "step": 22558 + }, + { + "epoch": 4.1, + "learning_rate": 1.7708955223880596e-06, + "loss": 0.9385, + "step": 22559 + }, + { + "epoch": 4.1, + "learning_rate": 1.7708409173643975e-06, + "loss": 0.4649, + "step": 22560 + }, + { + "epoch": 4.1, + "learning_rate": 1.7707863123407355e-06, + "loss": 0.0503, + "step": 22561 + }, + { + "epoch": 4.1, + "learning_rate": 1.7707317073170732e-06, + "loss": 1.7109, + "step": 22562 + }, + { + "epoch": 4.1, + "learning_rate": 1.7706771022934111e-06, + "loss": 1.6798, + "step": 22563 + }, + { + "epoch": 4.1, + "learning_rate": 1.770622497269749e-06, + "loss": 0.0146, + "step": 22564 + }, + { + "epoch": 4.1, + "learning_rate": 1.7705678922460866e-06, + "loss": 0.0182, + "step": 22565 + }, + { + "epoch": 4.1, + "learning_rate": 1.7705132872224246e-06, + "loss": 1.654, + "step": 22566 + }, + { + "epoch": 4.1, + "learning_rate": 1.770458682198762e-06, + "loss": 0.0148, + "step": 22567 + }, + { + "epoch": 4.1, + "learning_rate": 1.7704040771751e-06, + "loss": 0.6911, + "step": 22568 + }, + { + "epoch": 4.1, + "learning_rate": 1.770349472151438e-06, + "loss": 2.8601, + "step": 22569 + }, + { + "epoch": 4.1, + "learning_rate": 1.7702948671277757e-06, + "loss": 0.3249, + "step": 22570 + }, + { + "epoch": 4.1, + "learning_rate": 1.7702402621041136e-06, + "loss": 0.4088, + "step": 22571 + }, + { + "epoch": 4.1, + "learning_rate": 1.7701856570804516e-06, + "loss": 0.2453, + "step": 22572 + }, + { + "epoch": 4.1, + "learning_rate": 1.770131052056789e-06, + "loss": 0.2838, + "step": 22573 + }, + { + "epoch": 4.11, + "learning_rate": 1.770076447033127e-06, + "loss": 0.133, + "step": 22574 + }, + { + "epoch": 4.11, + "learning_rate": 1.770021842009465e-06, + "loss": 1.1499, + "step": 22575 + }, + { + "epoch": 4.11, + "learning_rate": 1.7699672369858027e-06, + "loss": 0.3964, + "step": 22576 + }, + { + "epoch": 4.11, + "learning_rate": 1.7699126319621407e-06, + "loss": 1.7591, + "step": 22577 + }, + { + "epoch": 4.11, + "learning_rate": 1.7698580269384786e-06, + "loss": 0.3256, + "step": 22578 + }, + { + "epoch": 4.11, + "learning_rate": 1.7698034219148161e-06, + "loss": 0.5829, + "step": 22579 + }, + { + "epoch": 4.11, + "learning_rate": 1.769748816891154e-06, + "loss": 0.4192, + "step": 22580 + }, + { + "epoch": 4.11, + "learning_rate": 1.7696942118674918e-06, + "loss": 0.7917, + "step": 22581 + }, + { + "epoch": 4.11, + "learning_rate": 1.7696396068438297e-06, + "loss": 0.0465, + "step": 22582 + }, + { + "epoch": 4.11, + "learning_rate": 1.7695850018201677e-06, + "loss": 0.0347, + "step": 22583 + }, + { + "epoch": 4.11, + "learning_rate": 1.7695303967965052e-06, + "loss": 1.1249, + "step": 22584 + }, + { + "epoch": 4.11, + "learning_rate": 1.7694757917728431e-06, + "loss": 1.3411, + "step": 22585 + }, + { + "epoch": 4.11, + "learning_rate": 1.769421186749181e-06, + "loss": 0.8149, + "step": 22586 + }, + { + "epoch": 4.11, + "learning_rate": 1.7693665817255186e-06, + "loss": 0.2114, + "step": 22587 + }, + { + "epoch": 4.11, + "learning_rate": 1.7693119767018566e-06, + "loss": 0.4998, + "step": 22588 + }, + { + "epoch": 4.11, + "learning_rate": 1.7692573716781945e-06, + "loss": 1.2625, + "step": 22589 + }, + { + "epoch": 4.11, + "learning_rate": 1.7692027666545322e-06, + "loss": 0.3707, + "step": 22590 + }, + { + "epoch": 4.11, + "learning_rate": 1.7691481616308702e-06, + "loss": 0.5085, + "step": 22591 + }, + { + "epoch": 4.11, + "learning_rate": 1.7690935566072081e-06, + "loss": 0.3049, + "step": 22592 + }, + { + "epoch": 4.11, + "learning_rate": 1.7690389515835456e-06, + "loss": 0.1788, + "step": 22593 + }, + { + "epoch": 4.11, + "learning_rate": 1.7689843465598836e-06, + "loss": 0.6003, + "step": 22594 + }, + { + "epoch": 4.11, + "learning_rate": 1.7689297415362213e-06, + "loss": 0.8845, + "step": 22595 + }, + { + "epoch": 4.11, + "learning_rate": 1.7688751365125593e-06, + "loss": 1.4528, + "step": 22596 + }, + { + "epoch": 4.11, + "learning_rate": 1.7688205314888972e-06, + "loss": 0.0405, + "step": 22597 + }, + { + "epoch": 4.11, + "learning_rate": 1.7687659264652347e-06, + "loss": 0.9847, + "step": 22598 + }, + { + "epoch": 4.11, + "learning_rate": 1.7687113214415727e-06, + "loss": 0.2867, + "step": 22599 + }, + { + "epoch": 4.11, + "learning_rate": 1.7686567164179106e-06, + "loss": 0.3639, + "step": 22600 + }, + { + "epoch": 4.11, + "learning_rate": 1.7686021113942481e-06, + "loss": 0.9968, + "step": 22601 + }, + { + "epoch": 4.11, + "learning_rate": 1.768547506370586e-06, + "loss": 0.7745, + "step": 22602 + }, + { + "epoch": 4.11, + "learning_rate": 1.768492901346924e-06, + "loss": 0.8947, + "step": 22603 + }, + { + "epoch": 4.11, + "learning_rate": 1.7684382963232617e-06, + "loss": 0.0202, + "step": 22604 + }, + { + "epoch": 4.11, + "learning_rate": 1.7683836912995997e-06, + "loss": 0.4386, + "step": 22605 + }, + { + "epoch": 4.11, + "learning_rate": 1.7683290862759376e-06, + "loss": 0.6684, + "step": 22606 + }, + { + "epoch": 4.11, + "learning_rate": 1.7682744812522752e-06, + "loss": 1.9236, + "step": 22607 + }, + { + "epoch": 4.11, + "learning_rate": 1.768219876228613e-06, + "loss": 1.4492, + "step": 22608 + }, + { + "epoch": 4.11, + "learning_rate": 1.7681652712049508e-06, + "loss": 0.033, + "step": 22609 + }, + { + "epoch": 4.11, + "learning_rate": 1.7681106661812888e-06, + "loss": 1.4552, + "step": 22610 + }, + { + "epoch": 4.11, + "learning_rate": 1.7680560611576267e-06, + "loss": 1.2394, + "step": 22611 + }, + { + "epoch": 4.11, + "learning_rate": 1.7680014561339642e-06, + "loss": 0.5665, + "step": 22612 + }, + { + "epoch": 4.11, + "learning_rate": 1.7679468511103022e-06, + "loss": 0.1621, + "step": 22613 + }, + { + "epoch": 4.11, + "learning_rate": 1.7678922460866401e-06, + "loss": 0.01, + "step": 22614 + }, + { + "epoch": 4.11, + "learning_rate": 1.7678376410629776e-06, + "loss": 0.968, + "step": 22615 + }, + { + "epoch": 4.11, + "learning_rate": 1.7677830360393156e-06, + "loss": 0.6679, + "step": 22616 + }, + { + "epoch": 4.11, + "learning_rate": 1.7677284310156535e-06, + "loss": 0.2206, + "step": 22617 + }, + { + "epoch": 4.11, + "learning_rate": 1.7676738259919913e-06, + "loss": 0.0652, + "step": 22618 + }, + { + "epoch": 4.11, + "learning_rate": 1.7676192209683292e-06, + "loss": 0.7021, + "step": 22619 + }, + { + "epoch": 4.11, + "learning_rate": 1.7675646159446671e-06, + "loss": 1.2192, + "step": 22620 + }, + { + "epoch": 4.11, + "learning_rate": 1.7675100109210047e-06, + "loss": 0.1262, + "step": 22621 + }, + { + "epoch": 4.11, + "learning_rate": 1.7674554058973426e-06, + "loss": 2.9719, + "step": 22622 + }, + { + "epoch": 4.11, + "learning_rate": 1.7674008008736803e-06, + "loss": 0.2724, + "step": 22623 + }, + { + "epoch": 4.11, + "learning_rate": 1.7673461958500183e-06, + "loss": 0.1683, + "step": 22624 + }, + { + "epoch": 4.11, + "learning_rate": 1.7672915908263562e-06, + "loss": 0.6631, + "step": 22625 + }, + { + "epoch": 4.11, + "learning_rate": 1.7672369858026937e-06, + "loss": 0.3595, + "step": 22626 + }, + { + "epoch": 4.11, + "learning_rate": 1.7671823807790317e-06, + "loss": 0.0455, + "step": 22627 + }, + { + "epoch": 4.11, + "learning_rate": 1.7671277757553696e-06, + "loss": 0.5446, + "step": 22628 + }, + { + "epoch": 4.12, + "learning_rate": 1.7670731707317074e-06, + "loss": 0.7607, + "step": 22629 + }, + { + "epoch": 4.12, + "learning_rate": 1.7670185657080453e-06, + "loss": 1.0575, + "step": 22630 + }, + { + "epoch": 4.12, + "learning_rate": 1.7669639606843833e-06, + "loss": 1.3003, + "step": 22631 + }, + { + "epoch": 4.12, + "learning_rate": 1.7669093556607208e-06, + "loss": 0.3389, + "step": 22632 + }, + { + "epoch": 4.12, + "learning_rate": 1.7668547506370587e-06, + "loss": 1.6638, + "step": 22633 + }, + { + "epoch": 4.12, + "learning_rate": 1.7668001456133962e-06, + "loss": 0.3701, + "step": 22634 + }, + { + "epoch": 4.12, + "learning_rate": 1.7667455405897342e-06, + "loss": 2.1448, + "step": 22635 + }, + { + "epoch": 4.12, + "learning_rate": 1.7666909355660721e-06, + "loss": 0.1139, + "step": 22636 + }, + { + "epoch": 4.12, + "learning_rate": 1.7666363305424099e-06, + "loss": 0.5459, + "step": 22637 + }, + { + "epoch": 4.12, + "learning_rate": 1.7665817255187478e-06, + "loss": 0.2905, + "step": 22638 + }, + { + "epoch": 4.12, + "learning_rate": 1.7665271204950857e-06, + "loss": 0.0202, + "step": 22639 + }, + { + "epoch": 4.12, + "learning_rate": 1.7664725154714233e-06, + "loss": 0.867, + "step": 22640 + }, + { + "epoch": 4.12, + "learning_rate": 1.7664179104477612e-06, + "loss": 0.1701, + "step": 22641 + }, + { + "epoch": 4.12, + "learning_rate": 1.7663633054240991e-06, + "loss": 0.1125, + "step": 22642 + }, + { + "epoch": 4.12, + "learning_rate": 1.7663087004004369e-06, + "loss": 0.0737, + "step": 22643 + }, + { + "epoch": 4.12, + "learning_rate": 1.7662540953767748e-06, + "loss": 0.6755, + "step": 22644 + }, + { + "epoch": 4.12, + "learning_rate": 1.7661994903531128e-06, + "loss": 0.1615, + "step": 22645 + }, + { + "epoch": 4.12, + "learning_rate": 1.7661448853294503e-06, + "loss": 0.1201, + "step": 22646 + }, + { + "epoch": 4.12, + "learning_rate": 1.7660902803057882e-06, + "loss": 0.0149, + "step": 22647 + }, + { + "epoch": 4.12, + "learning_rate": 1.7660356752821257e-06, + "loss": 1.2666, + "step": 22648 + }, + { + "epoch": 4.12, + "learning_rate": 1.7659810702584637e-06, + "loss": 0.6522, + "step": 22649 + }, + { + "epoch": 4.12, + "learning_rate": 1.7659264652348016e-06, + "loss": 0.8021, + "step": 22650 + }, + { + "epoch": 4.12, + "learning_rate": 1.7658718602111394e-06, + "loss": 0.0199, + "step": 22651 + }, + { + "epoch": 4.12, + "learning_rate": 1.7658172551874773e-06, + "loss": 0.026, + "step": 22652 + }, + { + "epoch": 4.12, + "learning_rate": 1.7657626501638153e-06, + "loss": 0.5593, + "step": 22653 + }, + { + "epoch": 4.12, + "learning_rate": 1.7657080451401528e-06, + "loss": 0.5216, + "step": 22654 + }, + { + "epoch": 4.12, + "learning_rate": 1.7656534401164907e-06, + "loss": 0.4351, + "step": 22655 + }, + { + "epoch": 4.12, + "learning_rate": 1.7655988350928287e-06, + "loss": 0.1339, + "step": 22656 + }, + { + "epoch": 4.12, + "learning_rate": 1.7655442300691664e-06, + "loss": 0.8576, + "step": 22657 + }, + { + "epoch": 4.12, + "learning_rate": 1.7654896250455043e-06, + "loss": 0.3743, + "step": 22658 + }, + { + "epoch": 4.12, + "learning_rate": 1.7654350200218423e-06, + "loss": 0.9721, + "step": 22659 + }, + { + "epoch": 4.12, + "learning_rate": 1.7653804149981798e-06, + "loss": 0.5061, + "step": 22660 + }, + { + "epoch": 4.12, + "learning_rate": 1.7653258099745177e-06, + "loss": 0.1106, + "step": 22661 + }, + { + "epoch": 4.12, + "learning_rate": 1.7652712049508553e-06, + "loss": 0.0693, + "step": 22662 + }, + { + "epoch": 4.12, + "learning_rate": 1.7652165999271932e-06, + "loss": 2.7045, + "step": 22663 + }, + { + "epoch": 4.12, + "learning_rate": 1.7651619949035311e-06, + "loss": 0.4076, + "step": 22664 + }, + { + "epoch": 4.12, + "learning_rate": 1.7651073898798689e-06, + "loss": 0.4291, + "step": 22665 + }, + { + "epoch": 4.12, + "learning_rate": 1.7650527848562068e-06, + "loss": 0.6778, + "step": 22666 + }, + { + "epoch": 4.12, + "learning_rate": 1.7649981798325448e-06, + "loss": 1.0185, + "step": 22667 + }, + { + "epoch": 4.12, + "learning_rate": 1.7649435748088823e-06, + "loss": 0.0371, + "step": 22668 + }, + { + "epoch": 4.12, + "learning_rate": 1.7648889697852202e-06, + "loss": 0.4078, + "step": 22669 + }, + { + "epoch": 4.12, + "learning_rate": 1.7648343647615582e-06, + "loss": 0.2903, + "step": 22670 + }, + { + "epoch": 4.12, + "learning_rate": 1.764779759737896e-06, + "loss": 0.857, + "step": 22671 + }, + { + "epoch": 4.12, + "learning_rate": 1.7647251547142338e-06, + "loss": 0.0093, + "step": 22672 + }, + { + "epoch": 4.12, + "learning_rate": 1.7646705496905718e-06, + "loss": 0.3867, + "step": 22673 + }, + { + "epoch": 4.12, + "learning_rate": 1.7646159446669093e-06, + "loss": 0.6016, + "step": 22674 + }, + { + "epoch": 4.12, + "learning_rate": 1.7645613396432473e-06, + "loss": 2.3667, + "step": 22675 + }, + { + "epoch": 4.12, + "learning_rate": 1.764506734619585e-06, + "loss": 0.1422, + "step": 22676 + }, + { + "epoch": 4.12, + "learning_rate": 1.764452129595923e-06, + "loss": 1.6844, + "step": 22677 + }, + { + "epoch": 4.12, + "learning_rate": 1.7643975245722609e-06, + "loss": 0.3808, + "step": 22678 + }, + { + "epoch": 4.12, + "learning_rate": 1.7643429195485984e-06, + "loss": 1.2584, + "step": 22679 + }, + { + "epoch": 4.12, + "learning_rate": 1.7642883145249363e-06, + "loss": 0.0775, + "step": 22680 + }, + { + "epoch": 4.12, + "learning_rate": 1.7642337095012743e-06, + "loss": 2.3597, + "step": 22681 + }, + { + "epoch": 4.12, + "learning_rate": 1.7641791044776118e-06, + "loss": 0.0868, + "step": 22682 + }, + { + "epoch": 4.12, + "learning_rate": 1.7641244994539497e-06, + "loss": 0.3989, + "step": 22683 + }, + { + "epoch": 4.13, + "learning_rate": 1.7640698944302877e-06, + "loss": 0.2549, + "step": 22684 + }, + { + "epoch": 4.13, + "learning_rate": 1.7640152894066254e-06, + "loss": 0.2899, + "step": 22685 + }, + { + "epoch": 4.13, + "learning_rate": 1.7639606843829634e-06, + "loss": 1.7551, + "step": 22686 + }, + { + "epoch": 4.13, + "learning_rate": 1.7639060793593013e-06, + "loss": 1.763, + "step": 22687 + }, + { + "epoch": 4.13, + "learning_rate": 1.7638514743356388e-06, + "loss": 0.7914, + "step": 22688 + }, + { + "epoch": 4.13, + "learning_rate": 1.7637968693119768e-06, + "loss": 1.0457, + "step": 22689 + }, + { + "epoch": 4.13, + "learning_rate": 1.7637422642883145e-06, + "loss": 0.3513, + "step": 22690 + }, + { + "epoch": 4.13, + "learning_rate": 1.7636876592646524e-06, + "loss": 0.0764, + "step": 22691 + }, + { + "epoch": 4.13, + "learning_rate": 1.7636330542409904e-06, + "loss": 0.173, + "step": 22692 + }, + { + "epoch": 4.13, + "learning_rate": 1.763578449217328e-06, + "loss": 1.1731, + "step": 22693 + }, + { + "epoch": 4.13, + "learning_rate": 1.7635238441936659e-06, + "loss": 0.3295, + "step": 22694 + }, + { + "epoch": 4.13, + "learning_rate": 1.7634692391700038e-06, + "loss": 0.2012, + "step": 22695 + }, + { + "epoch": 4.13, + "learning_rate": 1.7634146341463413e-06, + "loss": 0.6592, + "step": 22696 + }, + { + "epoch": 4.13, + "learning_rate": 1.7633600291226793e-06, + "loss": 0.0353, + "step": 22697 + }, + { + "epoch": 4.13, + "learning_rate": 1.7633054240990172e-06, + "loss": 0.7207, + "step": 22698 + }, + { + "epoch": 4.13, + "learning_rate": 1.763250819075355e-06, + "loss": 0.8018, + "step": 22699 + }, + { + "epoch": 4.13, + "learning_rate": 1.7631962140516929e-06, + "loss": 0.6627, + "step": 22700 + }, + { + "epoch": 4.13, + "learning_rate": 1.7631416090280308e-06, + "loss": 1.1446, + "step": 22701 + }, + { + "epoch": 4.13, + "learning_rate": 1.7630870040043683e-06, + "loss": 1.9291, + "step": 22702 + }, + { + "epoch": 4.13, + "learning_rate": 1.7630323989807063e-06, + "loss": 0.843, + "step": 22703 + }, + { + "epoch": 4.13, + "learning_rate": 1.762977793957044e-06, + "loss": 0.8367, + "step": 22704 + }, + { + "epoch": 4.13, + "learning_rate": 1.762923188933382e-06, + "loss": 1.6149, + "step": 22705 + }, + { + "epoch": 4.13, + "learning_rate": 1.76286858390972e-06, + "loss": 0.9938, + "step": 22706 + }, + { + "epoch": 4.13, + "learning_rate": 1.7628139788860574e-06, + "loss": 0.0507, + "step": 22707 + }, + { + "epoch": 4.13, + "learning_rate": 1.7627593738623954e-06, + "loss": 0.0574, + "step": 22708 + }, + { + "epoch": 4.13, + "learning_rate": 1.7627047688387333e-06, + "loss": 1.2139, + "step": 22709 + }, + { + "epoch": 4.13, + "learning_rate": 1.762650163815071e-06, + "loss": 0.7267, + "step": 22710 + }, + { + "epoch": 4.13, + "learning_rate": 1.762595558791409e-06, + "loss": 0.9112, + "step": 22711 + }, + { + "epoch": 4.13, + "learning_rate": 1.762540953767747e-06, + "loss": 0.6499, + "step": 22712 + }, + { + "epoch": 4.13, + "learning_rate": 1.7624863487440844e-06, + "loss": 0.259, + "step": 22713 + }, + { + "epoch": 4.13, + "learning_rate": 1.7624317437204224e-06, + "loss": 1.9272, + "step": 22714 + }, + { + "epoch": 4.13, + "learning_rate": 1.76237713869676e-06, + "loss": 0.2355, + "step": 22715 + }, + { + "epoch": 4.13, + "learning_rate": 1.7623225336730979e-06, + "loss": 2.8389, + "step": 22716 + }, + { + "epoch": 4.13, + "learning_rate": 1.7622679286494358e-06, + "loss": 0.4545, + "step": 22717 + }, + { + "epoch": 4.13, + "learning_rate": 1.7622133236257735e-06, + "loss": 0.109, + "step": 22718 + }, + { + "epoch": 4.13, + "learning_rate": 1.7621587186021115e-06, + "loss": 2.3424, + "step": 22719 + }, + { + "epoch": 4.13, + "learning_rate": 1.7621041135784494e-06, + "loss": 1.5955, + "step": 22720 + }, + { + "epoch": 4.13, + "learning_rate": 1.762049508554787e-06, + "loss": 0.0078, + "step": 22721 + }, + { + "epoch": 4.13, + "learning_rate": 1.7619949035311249e-06, + "loss": 0.0991, + "step": 22722 + }, + { + "epoch": 4.13, + "learning_rate": 1.7619402985074628e-06, + "loss": 0.0925, + "step": 22723 + }, + { + "epoch": 4.13, + "learning_rate": 1.7618856934838006e-06, + "loss": 1.8231, + "step": 22724 + }, + { + "epoch": 4.13, + "learning_rate": 1.7618310884601385e-06, + "loss": 1.0042, + "step": 22725 + }, + { + "epoch": 4.13, + "learning_rate": 1.7617764834364764e-06, + "loss": 0.1548, + "step": 22726 + }, + { + "epoch": 4.13, + "learning_rate": 1.761721878412814e-06, + "loss": 1.0357, + "step": 22727 + }, + { + "epoch": 4.13, + "learning_rate": 1.761667273389152e-06, + "loss": 2.8963, + "step": 22728 + }, + { + "epoch": 4.13, + "learning_rate": 1.7616126683654894e-06, + "loss": 2.4799, + "step": 22729 + }, + { + "epoch": 4.13, + "learning_rate": 1.7615580633418274e-06, + "loss": 0.9437, + "step": 22730 + }, + { + "epoch": 4.13, + "learning_rate": 1.7615034583181653e-06, + "loss": 0.1611, + "step": 22731 + }, + { + "epoch": 4.13, + "learning_rate": 1.761448853294503e-06, + "loss": 0.5011, + "step": 22732 + }, + { + "epoch": 4.13, + "learning_rate": 1.761394248270841e-06, + "loss": 0.5088, + "step": 22733 + }, + { + "epoch": 4.13, + "learning_rate": 1.761339643247179e-06, + "loss": 0.0401, + "step": 22734 + }, + { + "epoch": 4.13, + "learning_rate": 1.7612850382235164e-06, + "loss": 0.9409, + "step": 22735 + }, + { + "epoch": 4.13, + "learning_rate": 1.7612304331998544e-06, + "loss": 0.2273, + "step": 22736 + }, + { + "epoch": 4.13, + "learning_rate": 1.7611758281761923e-06, + "loss": 0.0513, + "step": 22737 + }, + { + "epoch": 4.13, + "learning_rate": 1.76112122315253e-06, + "loss": 0.1286, + "step": 22738 + }, + { + "epoch": 4.14, + "learning_rate": 1.761066618128868e-06, + "loss": 0.1857, + "step": 22739 + }, + { + "epoch": 4.14, + "learning_rate": 1.761012013105206e-06, + "loss": 0.1745, + "step": 22740 + }, + { + "epoch": 4.14, + "learning_rate": 1.7609574080815435e-06, + "loss": 0.0179, + "step": 22741 + }, + { + "epoch": 4.14, + "learning_rate": 1.7609028030578814e-06, + "loss": 0.269, + "step": 22742 + }, + { + "epoch": 4.14, + "learning_rate": 1.760848198034219e-06, + "loss": 0.6021, + "step": 22743 + }, + { + "epoch": 4.14, + "learning_rate": 1.7607935930105569e-06, + "loss": 0.1322, + "step": 22744 + }, + { + "epoch": 4.14, + "learning_rate": 1.7607389879868948e-06, + "loss": 0.2078, + "step": 22745 + }, + { + "epoch": 4.14, + "learning_rate": 1.7606843829632326e-06, + "loss": 0.1935, + "step": 22746 + }, + { + "epoch": 4.14, + "learning_rate": 1.7606297779395705e-06, + "loss": 0.2676, + "step": 22747 + }, + { + "epoch": 4.14, + "learning_rate": 1.7605751729159084e-06, + "loss": 1.1797, + "step": 22748 + }, + { + "epoch": 4.14, + "learning_rate": 1.760520567892246e-06, + "loss": 0.1436, + "step": 22749 + }, + { + "epoch": 4.14, + "learning_rate": 1.760465962868584e-06, + "loss": 1.7208, + "step": 22750 + }, + { + "epoch": 4.14, + "learning_rate": 1.7604113578449218e-06, + "loss": 0.9797, + "step": 22751 + }, + { + "epoch": 4.14, + "learning_rate": 1.7603567528212596e-06, + "loss": 2.5241, + "step": 22752 + }, + { + "epoch": 4.14, + "learning_rate": 1.7603021477975975e-06, + "loss": 2.1953, + "step": 22753 + }, + { + "epoch": 4.14, + "learning_rate": 1.7602475427739355e-06, + "loss": 0.1538, + "step": 22754 + }, + { + "epoch": 4.14, + "learning_rate": 1.760192937750273e-06, + "loss": 0.582, + "step": 22755 + }, + { + "epoch": 4.14, + "learning_rate": 1.760138332726611e-06, + "loss": 1.6846, + "step": 22756 + }, + { + "epoch": 4.14, + "learning_rate": 1.7600837277029487e-06, + "loss": 0.956, + "step": 22757 + }, + { + "epoch": 4.14, + "learning_rate": 1.7600291226792866e-06, + "loss": 0.0285, + "step": 22758 + }, + { + "epoch": 4.14, + "learning_rate": 1.7599745176556245e-06, + "loss": 1.5456, + "step": 22759 + }, + { + "epoch": 4.14, + "learning_rate": 1.759919912631962e-06, + "loss": 0.2829, + "step": 22760 + }, + { + "epoch": 4.14, + "learning_rate": 1.7598653076083e-06, + "loss": 0.5442, + "step": 22761 + }, + { + "epoch": 4.14, + "learning_rate": 1.759810702584638e-06, + "loss": 0.9361, + "step": 22762 + }, + { + "epoch": 4.14, + "learning_rate": 1.7597560975609755e-06, + "loss": 0.7692, + "step": 22763 + }, + { + "epoch": 4.14, + "learning_rate": 1.7597014925373134e-06, + "loss": 1.3176, + "step": 22764 + }, + { + "epoch": 4.14, + "learning_rate": 1.7596468875136514e-06, + "loss": 0.0146, + "step": 22765 + }, + { + "epoch": 4.14, + "learning_rate": 1.759592282489989e-06, + "loss": 4.4781, + "step": 22766 + }, + { + "epoch": 4.14, + "learning_rate": 1.759537677466327e-06, + "loss": 1.2969, + "step": 22767 + }, + { + "epoch": 4.14, + "learning_rate": 1.759483072442665e-06, + "loss": 0.3662, + "step": 22768 + }, + { + "epoch": 4.14, + "learning_rate": 1.7594284674190025e-06, + "loss": 0.0654, + "step": 22769 + }, + { + "epoch": 4.14, + "learning_rate": 1.7593738623953404e-06, + "loss": 0.4329, + "step": 22770 + }, + { + "epoch": 4.14, + "learning_rate": 1.7593192573716782e-06, + "loss": 0.4793, + "step": 22771 + }, + { + "epoch": 4.14, + "learning_rate": 1.7592646523480161e-06, + "loss": 2.8777, + "step": 22772 + }, + { + "epoch": 4.14, + "learning_rate": 1.759210047324354e-06, + "loss": 0.1992, + "step": 22773 + }, + { + "epoch": 4.14, + "learning_rate": 1.7591554423006916e-06, + "loss": 0.0065, + "step": 22774 + }, + { + "epoch": 4.14, + "learning_rate": 1.7591008372770295e-06, + "loss": 0.2163, + "step": 22775 + }, + { + "epoch": 4.14, + "learning_rate": 1.7590462322533675e-06, + "loss": 0.086, + "step": 22776 + }, + { + "epoch": 4.14, + "learning_rate": 1.758991627229705e-06, + "loss": 0.2623, + "step": 22777 + }, + { + "epoch": 4.14, + "learning_rate": 1.758937022206043e-06, + "loss": 0.5249, + "step": 22778 + }, + { + "epoch": 4.14, + "learning_rate": 1.7588824171823809e-06, + "loss": 0.2332, + "step": 22779 + }, + { + "epoch": 4.14, + "learning_rate": 1.7588278121587186e-06, + "loss": 2.1748, + "step": 22780 + }, + { + "epoch": 4.14, + "learning_rate": 1.7587732071350565e-06, + "loss": 1.7979, + "step": 22781 + }, + { + "epoch": 4.14, + "learning_rate": 1.7587186021113945e-06, + "loss": 0.4011, + "step": 22782 + }, + { + "epoch": 4.14, + "learning_rate": 1.758663997087732e-06, + "loss": 0.2744, + "step": 22783 + }, + { + "epoch": 4.14, + "learning_rate": 1.75860939206407e-06, + "loss": 0.3094, + "step": 22784 + }, + { + "epoch": 4.14, + "learning_rate": 1.7585547870404077e-06, + "loss": 2.4407, + "step": 22785 + }, + { + "epoch": 4.14, + "learning_rate": 1.7585001820167456e-06, + "loss": 1.9094, + "step": 22786 + }, + { + "epoch": 4.14, + "learning_rate": 1.7584455769930836e-06, + "loss": 0.0949, + "step": 22787 + }, + { + "epoch": 4.14, + "learning_rate": 1.758390971969421e-06, + "loss": 1.8665, + "step": 22788 + }, + { + "epoch": 4.14, + "learning_rate": 1.758336366945759e-06, + "loss": 0.3809, + "step": 22789 + }, + { + "epoch": 4.14, + "learning_rate": 1.758281761922097e-06, + "loss": 0.0133, + "step": 22790 + }, + { + "epoch": 4.14, + "learning_rate": 1.7582271568984345e-06, + "loss": 0.3415, + "step": 22791 + }, + { + "epoch": 4.14, + "learning_rate": 1.7581725518747724e-06, + "loss": 0.0146, + "step": 22792 + }, + { + "epoch": 4.14, + "learning_rate": 1.7581179468511104e-06, + "loss": 0.1054, + "step": 22793 + }, + { + "epoch": 4.15, + "learning_rate": 1.7580633418274481e-06, + "loss": 1.191, + "step": 22794 + }, + { + "epoch": 4.15, + "learning_rate": 1.758008736803786e-06, + "loss": 1.0574, + "step": 22795 + }, + { + "epoch": 4.15, + "learning_rate": 1.7579541317801236e-06, + "loss": 1.8641, + "step": 22796 + }, + { + "epoch": 4.15, + "learning_rate": 1.7578995267564615e-06, + "loss": 0.0143, + "step": 22797 + }, + { + "epoch": 4.15, + "learning_rate": 1.7578449217327995e-06, + "loss": 1.6342, + "step": 22798 + }, + { + "epoch": 4.15, + "learning_rate": 1.7577903167091372e-06, + "loss": 2.9498, + "step": 22799 + }, + { + "epoch": 4.15, + "learning_rate": 1.7577357116854751e-06, + "loss": 0.0255, + "step": 22800 + }, + { + "epoch": 4.15, + "learning_rate": 1.757681106661813e-06, + "loss": 0.0362, + "step": 22801 + }, + { + "epoch": 4.15, + "learning_rate": 1.7576265016381506e-06, + "loss": 1.0044, + "step": 22802 + }, + { + "epoch": 4.15, + "learning_rate": 1.7575718966144886e-06, + "loss": 0.7524, + "step": 22803 + }, + { + "epoch": 4.15, + "learning_rate": 1.7575172915908265e-06, + "loss": 0.5309, + "step": 22804 + }, + { + "epoch": 4.15, + "learning_rate": 1.7574626865671642e-06, + "loss": 0.8381, + "step": 22805 + }, + { + "epoch": 4.15, + "learning_rate": 1.7574080815435022e-06, + "loss": 0.3399, + "step": 22806 + }, + { + "epoch": 4.15, + "learning_rate": 1.7573534765198401e-06, + "loss": 0.1998, + "step": 22807 + }, + { + "epoch": 4.15, + "learning_rate": 1.7572988714961776e-06, + "loss": 0.1139, + "step": 22808 + }, + { + "epoch": 4.15, + "learning_rate": 1.7572442664725156e-06, + "loss": 0.0393, + "step": 22809 + }, + { + "epoch": 4.15, + "learning_rate": 1.757189661448853e-06, + "loss": 1.1128, + "step": 22810 + }, + { + "epoch": 4.15, + "learning_rate": 1.757135056425191e-06, + "loss": 0.2559, + "step": 22811 + }, + { + "epoch": 4.15, + "learning_rate": 1.757080451401529e-06, + "loss": 1.4383, + "step": 22812 + }, + { + "epoch": 4.15, + "learning_rate": 1.7570258463778667e-06, + "loss": 0.0188, + "step": 22813 + }, + { + "epoch": 4.15, + "learning_rate": 1.7569712413542047e-06, + "loss": 0.4306, + "step": 22814 + }, + { + "epoch": 4.15, + "learning_rate": 1.7569166363305426e-06, + "loss": 0.3342, + "step": 22815 + }, + { + "epoch": 4.15, + "learning_rate": 1.7568620313068801e-06, + "loss": 0.0342, + "step": 22816 + }, + { + "epoch": 4.15, + "learning_rate": 1.756807426283218e-06, + "loss": 0.2058, + "step": 22817 + }, + { + "epoch": 4.15, + "learning_rate": 1.756752821259556e-06, + "loss": 0.2723, + "step": 22818 + }, + { + "epoch": 4.15, + "learning_rate": 1.7566982162358937e-06, + "loss": 0.2404, + "step": 22819 + }, + { + "epoch": 4.15, + "learning_rate": 1.7566436112122317e-06, + "loss": 1.3678, + "step": 22820 + }, + { + "epoch": 4.15, + "learning_rate": 1.7565890061885696e-06, + "loss": 1.8978, + "step": 22821 + }, + { + "epoch": 4.15, + "learning_rate": 1.7565344011649071e-06, + "loss": 0.0255, + "step": 22822 + }, + { + "epoch": 4.15, + "learning_rate": 1.756479796141245e-06, + "loss": 1.724, + "step": 22823 + }, + { + "epoch": 4.15, + "learning_rate": 1.7564251911175826e-06, + "loss": 0.0795, + "step": 22824 + }, + { + "epoch": 4.15, + "learning_rate": 1.7563705860939206e-06, + "loss": 2.0207, + "step": 22825 + }, + { + "epoch": 4.15, + "learning_rate": 1.7563159810702585e-06, + "loss": 0.0135, + "step": 22826 + }, + { + "epoch": 4.15, + "learning_rate": 1.7562613760465962e-06, + "loss": 0.0172, + "step": 22827 + }, + { + "epoch": 4.15, + "learning_rate": 1.7562067710229342e-06, + "loss": 0.2855, + "step": 22828 + }, + { + "epoch": 4.15, + "learning_rate": 1.7561521659992721e-06, + "loss": 0.0563, + "step": 22829 + }, + { + "epoch": 4.15, + "learning_rate": 1.7560975609756096e-06, + "loss": 0.5052, + "step": 22830 + }, + { + "epoch": 4.15, + "learning_rate": 1.7560429559519476e-06, + "loss": 0.8628, + "step": 22831 + }, + { + "epoch": 4.15, + "learning_rate": 1.7559883509282855e-06, + "loss": 0.2164, + "step": 22832 + }, + { + "epoch": 4.15, + "learning_rate": 1.7559337459046233e-06, + "loss": 0.5436, + "step": 22833 + }, + { + "epoch": 4.15, + "learning_rate": 1.7558791408809612e-06, + "loss": 0.4663, + "step": 22834 + }, + { + "epoch": 4.15, + "learning_rate": 1.7558245358572991e-06, + "loss": 0.2834, + "step": 22835 + }, + { + "epoch": 4.15, + "learning_rate": 1.7557699308336367e-06, + "loss": 0.1626, + "step": 22836 + }, + { + "epoch": 4.15, + "learning_rate": 1.7557153258099746e-06, + "loss": 0.1866, + "step": 22837 + }, + { + "epoch": 4.15, + "learning_rate": 1.7556607207863123e-06, + "loss": 0.7653, + "step": 22838 + }, + { + "epoch": 4.15, + "learning_rate": 1.7556061157626503e-06, + "loss": 0.4696, + "step": 22839 + }, + { + "epoch": 4.15, + "learning_rate": 1.7555515107389882e-06, + "loss": 0.7829, + "step": 22840 + }, + { + "epoch": 4.15, + "learning_rate": 1.7554969057153257e-06, + "loss": 0.3356, + "step": 22841 + }, + { + "epoch": 4.15, + "learning_rate": 1.7554423006916637e-06, + "loss": 0.7482, + "step": 22842 + }, + { + "epoch": 4.15, + "learning_rate": 1.7553876956680016e-06, + "loss": 0.4976, + "step": 22843 + }, + { + "epoch": 4.15, + "learning_rate": 1.7553330906443391e-06, + "loss": 0.9923, + "step": 22844 + }, + { + "epoch": 4.15, + "learning_rate": 1.755278485620677e-06, + "loss": 0.5963, + "step": 22845 + }, + { + "epoch": 4.15, + "learning_rate": 1.755223880597015e-06, + "loss": 0.4421, + "step": 22846 + }, + { + "epoch": 4.15, + "learning_rate": 1.7551692755733528e-06, + "loss": 1.1365, + "step": 22847 + }, + { + "epoch": 4.15, + "learning_rate": 1.7551146705496907e-06, + "loss": 0.159, + "step": 22848 + }, + { + "epoch": 4.16, + "learning_rate": 1.7550600655260287e-06, + "loss": 0.2421, + "step": 22849 + }, + { + "epoch": 4.16, + "learning_rate": 1.7550054605023662e-06, + "loss": 0.4524, + "step": 22850 + }, + { + "epoch": 4.16, + "learning_rate": 1.7549508554787041e-06, + "loss": 0.0528, + "step": 22851 + }, + { + "epoch": 4.16, + "learning_rate": 1.7548962504550418e-06, + "loss": 0.2395, + "step": 22852 + }, + { + "epoch": 4.16, + "learning_rate": 1.7548416454313798e-06, + "loss": 0.4338, + "step": 22853 + }, + { + "epoch": 4.16, + "learning_rate": 1.7547870404077177e-06, + "loss": 1.3878, + "step": 22854 + }, + { + "epoch": 4.16, + "learning_rate": 1.7547324353840553e-06, + "loss": 0.0182, + "step": 22855 + }, + { + "epoch": 4.16, + "learning_rate": 1.7546778303603932e-06, + "loss": 0.0121, + "step": 22856 + }, + { + "epoch": 4.16, + "learning_rate": 1.7546232253367311e-06, + "loss": 0.5912, + "step": 22857 + }, + { + "epoch": 4.16, + "learning_rate": 1.7545686203130687e-06, + "loss": 0.1311, + "step": 22858 + }, + { + "epoch": 4.16, + "learning_rate": 1.7545140152894066e-06, + "loss": 0.1595, + "step": 22859 + }, + { + "epoch": 4.16, + "learning_rate": 1.7544594102657445e-06, + "loss": 0.7427, + "step": 22860 + }, + { + "epoch": 4.16, + "learning_rate": 1.7544048052420823e-06, + "loss": 0.9312, + "step": 22861 + }, + { + "epoch": 4.16, + "learning_rate": 1.7543502002184202e-06, + "loss": 0.554, + "step": 22862 + }, + { + "epoch": 4.16, + "learning_rate": 1.7542955951947582e-06, + "loss": 0.0821, + "step": 22863 + }, + { + "epoch": 4.16, + "learning_rate": 1.7542409901710957e-06, + "loss": 0.3284, + "step": 22864 + }, + { + "epoch": 4.16, + "learning_rate": 1.7541863851474336e-06, + "loss": 0.431, + "step": 22865 + }, + { + "epoch": 4.16, + "learning_rate": 1.7541317801237714e-06, + "loss": 0.4589, + "step": 22866 + }, + { + "epoch": 4.16, + "learning_rate": 1.7540771751001093e-06, + "loss": 2.4552, + "step": 22867 + }, + { + "epoch": 4.16, + "learning_rate": 1.7540225700764472e-06, + "loss": 1.105, + "step": 22868 + }, + { + "epoch": 4.16, + "learning_rate": 1.7539679650527848e-06, + "loss": 0.2172, + "step": 22869 + }, + { + "epoch": 4.16, + "learning_rate": 1.7539133600291227e-06, + "loss": 0.3349, + "step": 22870 + }, + { + "epoch": 4.16, + "learning_rate": 1.7538587550054607e-06, + "loss": 0.0291, + "step": 22871 + }, + { + "epoch": 4.16, + "learning_rate": 1.7538041499817982e-06, + "loss": 0.7016, + "step": 22872 + }, + { + "epoch": 4.16, + "learning_rate": 1.7537495449581361e-06, + "loss": 1.5429, + "step": 22873 + }, + { + "epoch": 4.16, + "learning_rate": 1.753694939934474e-06, + "loss": 1.7488, + "step": 22874 + }, + { + "epoch": 4.16, + "learning_rate": 1.7536403349108118e-06, + "loss": 0.3258, + "step": 22875 + }, + { + "epoch": 4.16, + "learning_rate": 1.7535857298871497e-06, + "loss": 0.8254, + "step": 22876 + }, + { + "epoch": 4.16, + "learning_rate": 1.7535311248634873e-06, + "loss": 0.8229, + "step": 22877 + }, + { + "epoch": 4.16, + "learning_rate": 1.7534765198398252e-06, + "loss": 0.6878, + "step": 22878 + }, + { + "epoch": 4.16, + "learning_rate": 1.7534219148161631e-06, + "loss": 1.0409, + "step": 22879 + }, + { + "epoch": 4.16, + "learning_rate": 1.7533673097925009e-06, + "loss": 0.3154, + "step": 22880 + }, + { + "epoch": 4.16, + "learning_rate": 1.7533127047688388e-06, + "loss": 0.7708, + "step": 22881 + }, + { + "epoch": 4.16, + "learning_rate": 1.7532580997451768e-06, + "loss": 0.5491, + "step": 22882 + }, + { + "epoch": 4.16, + "learning_rate": 1.7532034947215143e-06, + "loss": 0.2777, + "step": 22883 + }, + { + "epoch": 4.16, + "learning_rate": 1.7531488896978522e-06, + "loss": 0.5402, + "step": 22884 + }, + { + "epoch": 4.16, + "learning_rate": 1.7530942846741902e-06, + "loss": 1.8397, + "step": 22885 + }, + { + "epoch": 4.16, + "learning_rate": 1.753039679650528e-06, + "loss": 0.4699, + "step": 22886 + }, + { + "epoch": 4.16, + "learning_rate": 1.7529850746268658e-06, + "loss": 1.3983, + "step": 22887 + }, + { + "epoch": 4.16, + "learning_rate": 1.7529304696032038e-06, + "loss": 2.0415, + "step": 22888 + }, + { + "epoch": 4.16, + "learning_rate": 1.7528758645795413e-06, + "loss": 0.8209, + "step": 22889 + }, + { + "epoch": 4.16, + "learning_rate": 1.7528212595558793e-06, + "loss": 0.5621, + "step": 22890 + }, + { + "epoch": 4.16, + "learning_rate": 1.7527666545322168e-06, + "loss": 0.0176, + "step": 22891 + }, + { + "epoch": 4.16, + "learning_rate": 1.7527120495085547e-06, + "loss": 0.4812, + "step": 22892 + }, + { + "epoch": 4.16, + "learning_rate": 1.7526574444848927e-06, + "loss": 0.0343, + "step": 22893 + }, + { + "epoch": 4.16, + "learning_rate": 1.7526028394612304e-06, + "loss": 0.2112, + "step": 22894 + }, + { + "epoch": 4.16, + "learning_rate": 1.7525482344375683e-06, + "loss": 0.1781, + "step": 22895 + }, + { + "epoch": 4.16, + "learning_rate": 1.7524936294139063e-06, + "loss": 0.2322, + "step": 22896 + }, + { + "epoch": 4.16, + "learning_rate": 1.7524390243902438e-06, + "loss": 0.1221, + "step": 22897 + }, + { + "epoch": 4.16, + "learning_rate": 1.7523844193665817e-06, + "loss": 1.2062, + "step": 22898 + }, + { + "epoch": 4.16, + "learning_rate": 1.7523298143429197e-06, + "loss": 0.0886, + "step": 22899 + }, + { + "epoch": 4.16, + "learning_rate": 1.7522752093192574e-06, + "loss": 0.0343, + "step": 22900 + }, + { + "epoch": 4.16, + "learning_rate": 1.7522206042955954e-06, + "loss": 0.9125, + "step": 22901 + }, + { + "epoch": 4.16, + "learning_rate": 1.7521659992719333e-06, + "loss": 1.2132, + "step": 22902 + }, + { + "epoch": 4.16, + "learning_rate": 1.7521113942482708e-06, + "loss": 0.5298, + "step": 22903 + }, + { + "epoch": 4.17, + "learning_rate": 1.7520567892246088e-06, + "loss": 1.8902, + "step": 22904 + }, + { + "epoch": 4.17, + "learning_rate": 1.7520021842009463e-06, + "loss": 1.2013, + "step": 22905 + }, + { + "epoch": 4.17, + "learning_rate": 1.7519475791772842e-06, + "loss": 2.0491, + "step": 22906 + }, + { + "epoch": 4.17, + "learning_rate": 1.7518929741536222e-06, + "loss": 1.3909, + "step": 22907 + }, + { + "epoch": 4.17, + "learning_rate": 1.75183836912996e-06, + "loss": 1.9684, + "step": 22908 + }, + { + "epoch": 4.17, + "learning_rate": 1.7517837641062978e-06, + "loss": 0.5962, + "step": 22909 + }, + { + "epoch": 4.17, + "learning_rate": 1.7517291590826358e-06, + "loss": 2.0425, + "step": 22910 + }, + { + "epoch": 4.17, + "learning_rate": 1.7516745540589733e-06, + "loss": 0.828, + "step": 22911 + }, + { + "epoch": 4.17, + "learning_rate": 1.7516199490353113e-06, + "loss": 0.3601, + "step": 22912 + }, + { + "epoch": 4.17, + "learning_rate": 1.7515653440116492e-06, + "loss": 0.1981, + "step": 22913 + }, + { + "epoch": 4.17, + "learning_rate": 1.751510738987987e-06, + "loss": 0.7829, + "step": 22914 + }, + { + "epoch": 4.17, + "learning_rate": 1.7514561339643249e-06, + "loss": 0.1152, + "step": 22915 + }, + { + "epoch": 4.17, + "learning_rate": 1.7514015289406628e-06, + "loss": 0.5587, + "step": 22916 + }, + { + "epoch": 4.17, + "learning_rate": 1.7513469239170003e-06, + "loss": 0.2554, + "step": 22917 + }, + { + "epoch": 4.17, + "learning_rate": 1.7512923188933383e-06, + "loss": 0.957, + "step": 22918 + }, + { + "epoch": 4.17, + "learning_rate": 1.751237713869676e-06, + "loss": 0.1197, + "step": 22919 + }, + { + "epoch": 4.17, + "learning_rate": 1.751183108846014e-06, + "loss": 1.1402, + "step": 22920 + }, + { + "epoch": 4.17, + "learning_rate": 1.7511285038223517e-06, + "loss": 1.0401, + "step": 22921 + }, + { + "epoch": 4.17, + "learning_rate": 1.7510738987986894e-06, + "loss": 1.4164, + "step": 22922 + }, + { + "epoch": 4.17, + "learning_rate": 1.7510192937750274e-06, + "loss": 0.3962, + "step": 22923 + }, + { + "epoch": 4.17, + "learning_rate": 1.7509646887513653e-06, + "loss": 0.0441, + "step": 22924 + }, + { + "epoch": 4.17, + "learning_rate": 1.7509100837277028e-06, + "loss": 0.8561, + "step": 22925 + }, + { + "epoch": 4.17, + "learning_rate": 1.7508554787040408e-06, + "loss": 0.0575, + "step": 22926 + }, + { + "epoch": 4.17, + "learning_rate": 1.7508008736803787e-06, + "loss": 0.1785, + "step": 22927 + }, + { + "epoch": 4.17, + "learning_rate": 1.7507462686567164e-06, + "loss": 0.3085, + "step": 22928 + }, + { + "epoch": 4.17, + "learning_rate": 1.7506916636330544e-06, + "loss": 0.3334, + "step": 22929 + }, + { + "epoch": 4.17, + "learning_rate": 1.7506370586093923e-06, + "loss": 0.6219, + "step": 22930 + }, + { + "epoch": 4.17, + "learning_rate": 1.7505824535857298e-06, + "loss": 0.1578, + "step": 22931 + }, + { + "epoch": 4.17, + "learning_rate": 1.7505278485620678e-06, + "loss": 1.0772, + "step": 22932 + }, + { + "epoch": 4.17, + "learning_rate": 1.7504732435384055e-06, + "loss": 0.3376, + "step": 22933 + }, + { + "epoch": 4.17, + "learning_rate": 1.7504186385147435e-06, + "loss": 0.4337, + "step": 22934 + }, + { + "epoch": 4.17, + "learning_rate": 1.7503640334910814e-06, + "loss": 0.5195, + "step": 22935 + }, + { + "epoch": 4.17, + "learning_rate": 1.750309428467419e-06, + "loss": 1.4897, + "step": 22936 + }, + { + "epoch": 4.17, + "learning_rate": 1.7502548234437569e-06, + "loss": 0.189, + "step": 22937 + }, + { + "epoch": 4.17, + "learning_rate": 1.7502002184200948e-06, + "loss": 0.5469, + "step": 22938 + }, + { + "epoch": 4.17, + "learning_rate": 1.7501456133964323e-06, + "loss": 0.5475, + "step": 22939 + }, + { + "epoch": 4.17, + "learning_rate": 1.7500910083727703e-06, + "loss": 1.6256, + "step": 22940 + }, + { + "epoch": 4.17, + "learning_rate": 1.7500364033491082e-06, + "loss": 0.3223, + "step": 22941 + }, + { + "epoch": 4.17, + "learning_rate": 1.749981798325446e-06, + "loss": 1.0916, + "step": 22942 + }, + { + "epoch": 4.17, + "learning_rate": 1.749927193301784e-06, + "loss": 0.4109, + "step": 22943 + }, + { + "epoch": 4.17, + "learning_rate": 1.7498725882781214e-06, + "loss": 0.5954, + "step": 22944 + }, + { + "epoch": 4.17, + "learning_rate": 1.7498179832544594e-06, + "loss": 0.0116, + "step": 22945 + }, + { + "epoch": 4.17, + "learning_rate": 1.7497633782307973e-06, + "loss": 1.7087, + "step": 22946 + }, + { + "epoch": 4.17, + "learning_rate": 1.749708773207135e-06, + "loss": 0.1181, + "step": 22947 + }, + { + "epoch": 4.17, + "learning_rate": 1.749654168183473e-06, + "loss": 0.2578, + "step": 22948 + }, + { + "epoch": 4.17, + "learning_rate": 1.749599563159811e-06, + "loss": 0.148, + "step": 22949 + }, + { + "epoch": 4.17, + "learning_rate": 1.7495449581361484e-06, + "loss": 0.7677, + "step": 22950 + }, + { + "epoch": 4.17, + "learning_rate": 1.7494903531124864e-06, + "loss": 0.1387, + "step": 22951 + }, + { + "epoch": 4.17, + "learning_rate": 1.7494357480888243e-06, + "loss": 0.2179, + "step": 22952 + }, + { + "epoch": 4.17, + "learning_rate": 1.7493811430651618e-06, + "loss": 0.558, + "step": 22953 + }, + { + "epoch": 4.17, + "learning_rate": 1.7493265380414998e-06, + "loss": 0.5399, + "step": 22954 + }, + { + "epoch": 4.17, + "learning_rate": 1.7492719330178377e-06, + "loss": 0.6801, + "step": 22955 + }, + { + "epoch": 4.17, + "learning_rate": 1.7492173279941755e-06, + "loss": 0.8991, + "step": 22956 + }, + { + "epoch": 4.17, + "learning_rate": 1.7491627229705134e-06, + "loss": 0.6686, + "step": 22957 + }, + { + "epoch": 4.17, + "learning_rate": 1.749108117946851e-06, + "loss": 0.3803, + "step": 22958 + }, + { + "epoch": 4.18, + "learning_rate": 1.7490535129231889e-06, + "loss": 0.0143, + "step": 22959 + }, + { + "epoch": 4.18, + "learning_rate": 1.7489989078995268e-06, + "loss": 0.9094, + "step": 22960 + }, + { + "epoch": 4.18, + "learning_rate": 1.7489443028758646e-06, + "loss": 3.0141, + "step": 22961 + }, + { + "epoch": 4.18, + "learning_rate": 1.7488896978522025e-06, + "loss": 0.7045, + "step": 22962 + }, + { + "epoch": 4.18, + "learning_rate": 1.7488350928285404e-06, + "loss": 0.4921, + "step": 22963 + }, + { + "epoch": 4.18, + "learning_rate": 1.748780487804878e-06, + "loss": 0.4859, + "step": 22964 + }, + { + "epoch": 4.18, + "learning_rate": 1.748725882781216e-06, + "loss": 0.9229, + "step": 22965 + }, + { + "epoch": 4.18, + "learning_rate": 1.7486712777575538e-06, + "loss": 0.2271, + "step": 22966 + }, + { + "epoch": 4.18, + "learning_rate": 1.7486166727338916e-06, + "loss": 0.0111, + "step": 22967 + }, + { + "epoch": 4.18, + "learning_rate": 1.7485620677102295e-06, + "loss": 3.2952, + "step": 22968 + }, + { + "epoch": 4.18, + "learning_rate": 1.7485074626865675e-06, + "loss": 0.8056, + "step": 22969 + }, + { + "epoch": 4.18, + "learning_rate": 1.748452857662905e-06, + "loss": 0.9637, + "step": 22970 + }, + { + "epoch": 4.18, + "learning_rate": 1.748398252639243e-06, + "loss": 1.4712, + "step": 22971 + }, + { + "epoch": 4.18, + "learning_rate": 1.7483436476155804e-06, + "loss": 0.1427, + "step": 22972 + }, + { + "epoch": 4.18, + "learning_rate": 1.7482890425919184e-06, + "loss": 0.2566, + "step": 22973 + }, + { + "epoch": 4.18, + "learning_rate": 1.7482344375682563e-06, + "loss": 2.5277, + "step": 22974 + }, + { + "epoch": 4.18, + "learning_rate": 1.748179832544594e-06, + "loss": 0.0164, + "step": 22975 + }, + { + "epoch": 4.18, + "learning_rate": 1.748125227520932e-06, + "loss": 0.0077, + "step": 22976 + }, + { + "epoch": 4.18, + "learning_rate": 1.74807062249727e-06, + "loss": 0.4711, + "step": 22977 + }, + { + "epoch": 4.18, + "learning_rate": 1.7480160174736075e-06, + "loss": 0.2505, + "step": 22978 + }, + { + "epoch": 4.18, + "learning_rate": 1.7479614124499454e-06, + "loss": 0.3877, + "step": 22979 + }, + { + "epoch": 4.18, + "learning_rate": 1.7479068074262834e-06, + "loss": 0.5968, + "step": 22980 + }, + { + "epoch": 4.18, + "learning_rate": 1.747852202402621e-06, + "loss": 0.4803, + "step": 22981 + }, + { + "epoch": 4.18, + "learning_rate": 1.747797597378959e-06, + "loss": 1.2953, + "step": 22982 + }, + { + "epoch": 4.18, + "learning_rate": 1.747742992355297e-06, + "loss": 0.0092, + "step": 22983 + }, + { + "epoch": 4.18, + "learning_rate": 1.7476883873316345e-06, + "loss": 0.1437, + "step": 22984 + }, + { + "epoch": 4.18, + "learning_rate": 1.7476337823079724e-06, + "loss": 1.1626, + "step": 22985 + }, + { + "epoch": 4.18, + "learning_rate": 1.74757917728431e-06, + "loss": 0.5155, + "step": 22986 + }, + { + "epoch": 4.18, + "learning_rate": 1.747524572260648e-06, + "loss": 0.5126, + "step": 22987 + }, + { + "epoch": 4.18, + "learning_rate": 1.7474699672369858e-06, + "loss": 0.3532, + "step": 22988 + }, + { + "epoch": 4.18, + "learning_rate": 1.7474153622133236e-06, + "loss": 0.0186, + "step": 22989 + }, + { + "epoch": 4.18, + "learning_rate": 1.7473607571896615e-06, + "loss": 0.242, + "step": 22990 + }, + { + "epoch": 4.18, + "learning_rate": 1.7473061521659995e-06, + "loss": 0.2918, + "step": 22991 + }, + { + "epoch": 4.18, + "learning_rate": 1.747251547142337e-06, + "loss": 2.1753, + "step": 22992 + }, + { + "epoch": 4.18, + "learning_rate": 1.747196942118675e-06, + "loss": 0.8943, + "step": 22993 + }, + { + "epoch": 4.18, + "learning_rate": 1.7471423370950129e-06, + "loss": 0.0138, + "step": 22994 + }, + { + "epoch": 4.18, + "learning_rate": 1.7470877320713506e-06, + "loss": 0.7726, + "step": 22995 + }, + { + "epoch": 4.18, + "learning_rate": 1.7470331270476885e-06, + "loss": 0.3116, + "step": 22996 + }, + { + "epoch": 4.18, + "learning_rate": 1.7469785220240265e-06, + "loss": 2.0603, + "step": 22997 + }, + { + "epoch": 4.18, + "learning_rate": 1.746923917000364e-06, + "loss": 0.2041, + "step": 22998 + }, + { + "epoch": 4.18, + "learning_rate": 1.746869311976702e-06, + "loss": 1.7206, + "step": 22999 + }, + { + "epoch": 4.18, + "learning_rate": 1.7468147069530395e-06, + "loss": 1.5885, + "step": 23000 + }, + { + "epoch": 4.18, + "learning_rate": 1.7467601019293774e-06, + "loss": 0.9071, + "step": 23001 + }, + { + "epoch": 4.18, + "learning_rate": 1.7467054969057154e-06, + "loss": 1.63, + "step": 23002 + }, + { + "epoch": 4.18, + "learning_rate": 1.746650891882053e-06, + "loss": 1.6933, + "step": 23003 + }, + { + "epoch": 4.18, + "learning_rate": 1.746596286858391e-06, + "loss": 0.997, + "step": 23004 + }, + { + "epoch": 4.18, + "learning_rate": 1.746541681834729e-06, + "loss": 0.6443, + "step": 23005 + }, + { + "epoch": 4.18, + "learning_rate": 1.7464870768110665e-06, + "loss": 0.6164, + "step": 23006 + }, + { + "epoch": 4.18, + "learning_rate": 1.7464324717874044e-06, + "loss": 0.0339, + "step": 23007 + }, + { + "epoch": 4.18, + "learning_rate": 1.7463778667637424e-06, + "loss": 1.3996, + "step": 23008 + }, + { + "epoch": 4.18, + "learning_rate": 1.7463232617400801e-06, + "loss": 0.0215, + "step": 23009 + }, + { + "epoch": 4.18, + "learning_rate": 1.746268656716418e-06, + "loss": 0.023, + "step": 23010 + }, + { + "epoch": 4.18, + "learning_rate": 1.746214051692756e-06, + "loss": 0.5172, + "step": 23011 + }, + { + "epoch": 4.18, + "learning_rate": 1.7461594466690935e-06, + "loss": 0.9626, + "step": 23012 + }, + { + "epoch": 4.18, + "learning_rate": 1.7461048416454315e-06, + "loss": 1.1969, + "step": 23013 + }, + { + "epoch": 4.19, + "learning_rate": 1.7460502366217692e-06, + "loss": 2.2715, + "step": 23014 + }, + { + "epoch": 4.19, + "learning_rate": 1.7459956315981071e-06, + "loss": 0.4635, + "step": 23015 + }, + { + "epoch": 4.19, + "learning_rate": 1.745941026574445e-06, + "loss": 2.6392, + "step": 23016 + }, + { + "epoch": 4.19, + "learning_rate": 1.7458864215507826e-06, + "loss": 0.5132, + "step": 23017 + }, + { + "epoch": 4.19, + "learning_rate": 1.7458318165271205e-06, + "loss": 0.109, + "step": 23018 + }, + { + "epoch": 4.19, + "learning_rate": 1.7457772115034585e-06, + "loss": 0.0223, + "step": 23019 + }, + { + "epoch": 4.19, + "learning_rate": 1.745722606479796e-06, + "loss": 0.7986, + "step": 23020 + }, + { + "epoch": 4.19, + "learning_rate": 1.745668001456134e-06, + "loss": 0.8655, + "step": 23021 + }, + { + "epoch": 4.19, + "learning_rate": 1.745613396432472e-06, + "loss": 0.0207, + "step": 23022 + }, + { + "epoch": 4.19, + "learning_rate": 1.7455587914088096e-06, + "loss": 1.0262, + "step": 23023 + }, + { + "epoch": 4.19, + "learning_rate": 1.7455041863851476e-06, + "loss": 0.0154, + "step": 23024 + }, + { + "epoch": 4.19, + "learning_rate": 1.745449581361485e-06, + "loss": 0.3722, + "step": 23025 + }, + { + "epoch": 4.19, + "learning_rate": 1.745394976337823e-06, + "loss": 0.2597, + "step": 23026 + }, + { + "epoch": 4.19, + "learning_rate": 1.745340371314161e-06, + "loss": 0.1877, + "step": 23027 + }, + { + "epoch": 4.19, + "learning_rate": 1.7452857662904987e-06, + "loss": 0.5795, + "step": 23028 + }, + { + "epoch": 4.19, + "learning_rate": 1.7452311612668367e-06, + "loss": 0.3014, + "step": 23029 + }, + { + "epoch": 4.19, + "learning_rate": 1.7451765562431746e-06, + "loss": 1.7229, + "step": 23030 + }, + { + "epoch": 4.19, + "learning_rate": 1.7451219512195121e-06, + "loss": 1.1715, + "step": 23031 + }, + { + "epoch": 4.19, + "learning_rate": 1.74506734619585e-06, + "loss": 1.2825, + "step": 23032 + }, + { + "epoch": 4.19, + "learning_rate": 1.745012741172188e-06, + "loss": 0.17, + "step": 23033 + }, + { + "epoch": 4.19, + "learning_rate": 1.7449581361485255e-06, + "loss": 0.0232, + "step": 23034 + }, + { + "epoch": 4.19, + "learning_rate": 1.7449035311248635e-06, + "loss": 2.5214, + "step": 23035 + }, + { + "epoch": 4.19, + "learning_rate": 1.7448489261012014e-06, + "loss": 0.024, + "step": 23036 + }, + { + "epoch": 4.19, + "learning_rate": 1.7447943210775391e-06, + "loss": 0.102, + "step": 23037 + }, + { + "epoch": 4.19, + "learning_rate": 1.744739716053877e-06, + "loss": 0.0576, + "step": 23038 + }, + { + "epoch": 4.19, + "learning_rate": 1.7446851110302146e-06, + "loss": 0.8729, + "step": 23039 + }, + { + "epoch": 4.19, + "learning_rate": 1.7446305060065525e-06, + "loss": 0.686, + "step": 23040 + }, + { + "epoch": 4.19, + "learning_rate": 1.7445759009828905e-06, + "loss": 0.2128, + "step": 23041 + }, + { + "epoch": 4.19, + "learning_rate": 1.7445212959592282e-06, + "loss": 0.4826, + "step": 23042 + }, + { + "epoch": 4.19, + "learning_rate": 1.7444666909355662e-06, + "loss": 1.1001, + "step": 23043 + }, + { + "epoch": 4.19, + "learning_rate": 1.7444120859119041e-06, + "loss": 0.0225, + "step": 23044 + }, + { + "epoch": 4.19, + "learning_rate": 1.7443574808882416e-06, + "loss": 0.3009, + "step": 23045 + }, + { + "epoch": 4.19, + "learning_rate": 1.7443028758645796e-06, + "loss": 0.6794, + "step": 23046 + }, + { + "epoch": 4.19, + "learning_rate": 1.7442482708409175e-06, + "loss": 0.0986, + "step": 23047 + }, + { + "epoch": 4.19, + "learning_rate": 1.7441936658172552e-06, + "loss": 0.2213, + "step": 23048 + }, + { + "epoch": 4.19, + "learning_rate": 1.7441390607935932e-06, + "loss": 0.2515, + "step": 23049 + }, + { + "epoch": 4.19, + "learning_rate": 1.744084455769931e-06, + "loss": 1.841, + "step": 23050 + }, + { + "epoch": 4.19, + "learning_rate": 1.7440298507462687e-06, + "loss": 0.0101, + "step": 23051 + }, + { + "epoch": 4.19, + "learning_rate": 1.7439752457226066e-06, + "loss": 0.3774, + "step": 23052 + }, + { + "epoch": 4.19, + "learning_rate": 1.7439206406989441e-06, + "loss": 0.0105, + "step": 23053 + }, + { + "epoch": 4.19, + "learning_rate": 1.743866035675282e-06, + "loss": 0.1366, + "step": 23054 + }, + { + "epoch": 4.19, + "learning_rate": 1.74381143065162e-06, + "loss": 0.8172, + "step": 23055 + }, + { + "epoch": 4.19, + "learning_rate": 1.7437568256279577e-06, + "loss": 1.5311, + "step": 23056 + }, + { + "epoch": 4.19, + "learning_rate": 1.7437022206042957e-06, + "loss": 0.2531, + "step": 23057 + }, + { + "epoch": 4.19, + "learning_rate": 1.7436476155806336e-06, + "loss": 1.8919, + "step": 23058 + }, + { + "epoch": 4.19, + "learning_rate": 1.7435930105569711e-06, + "loss": 0.3979, + "step": 23059 + }, + { + "epoch": 4.19, + "learning_rate": 1.743538405533309e-06, + "loss": 0.0721, + "step": 23060 + }, + { + "epoch": 4.19, + "learning_rate": 1.743483800509647e-06, + "loss": 6.8703, + "step": 23061 + }, + { + "epoch": 4.19, + "learning_rate": 1.7434291954859848e-06, + "loss": 0.4445, + "step": 23062 + }, + { + "epoch": 4.19, + "learning_rate": 1.7433745904623227e-06, + "loss": 1.0246, + "step": 23063 + }, + { + "epoch": 4.19, + "learning_rate": 1.7433199854386606e-06, + "loss": 0.9176, + "step": 23064 + }, + { + "epoch": 4.19, + "learning_rate": 1.7432653804149982e-06, + "loss": 0.0599, + "step": 23065 + }, + { + "epoch": 4.19, + "learning_rate": 1.7432107753913361e-06, + "loss": 0.007, + "step": 23066 + }, + { + "epoch": 4.19, + "learning_rate": 1.7431561703676736e-06, + "loss": 1.4189, + "step": 23067 + }, + { + "epoch": 4.19, + "learning_rate": 1.7431015653440116e-06, + "loss": 0.0618, + "step": 23068 + }, + { + "epoch": 4.2, + "learning_rate": 1.7430469603203495e-06, + "loss": 0.7065, + "step": 23069 + }, + { + "epoch": 4.2, + "learning_rate": 1.7429923552966873e-06, + "loss": 0.3466, + "step": 23070 + }, + { + "epoch": 4.2, + "learning_rate": 1.7429377502730252e-06, + "loss": 0.0103, + "step": 23071 + }, + { + "epoch": 4.2, + "learning_rate": 1.7428831452493631e-06, + "loss": 1.5122, + "step": 23072 + }, + { + "epoch": 4.2, + "learning_rate": 1.7428285402257007e-06, + "loss": 2.3149, + "step": 23073 + }, + { + "epoch": 4.2, + "learning_rate": 1.7427739352020386e-06, + "loss": 0.0238, + "step": 23074 + }, + { + "epoch": 4.2, + "learning_rate": 1.7427193301783765e-06, + "loss": 1.0185, + "step": 23075 + }, + { + "epoch": 4.2, + "learning_rate": 1.7426647251547143e-06, + "loss": 2.1364, + "step": 23076 + }, + { + "epoch": 4.2, + "learning_rate": 1.7426101201310522e-06, + "loss": 2.7684, + "step": 23077 + }, + { + "epoch": 4.2, + "learning_rate": 1.7425555151073902e-06, + "loss": 2.351, + "step": 23078 + }, + { + "epoch": 4.2, + "learning_rate": 1.7425009100837277e-06, + "loss": 0.4963, + "step": 23079 + }, + { + "epoch": 4.2, + "learning_rate": 1.7424463050600656e-06, + "loss": 0.3818, + "step": 23080 + }, + { + "epoch": 4.2, + "learning_rate": 1.7423917000364031e-06, + "loss": 0.3167, + "step": 23081 + }, + { + "epoch": 4.2, + "learning_rate": 1.742337095012741e-06, + "loss": 0.2094, + "step": 23082 + }, + { + "epoch": 4.2, + "learning_rate": 1.742282489989079e-06, + "loss": 0.0194, + "step": 23083 + }, + { + "epoch": 4.2, + "learning_rate": 1.7422278849654168e-06, + "loss": 0.0171, + "step": 23084 + }, + { + "epoch": 4.2, + "learning_rate": 1.7421732799417547e-06, + "loss": 2.1563, + "step": 23085 + }, + { + "epoch": 4.2, + "learning_rate": 1.7421186749180927e-06, + "loss": 0.8332, + "step": 23086 + }, + { + "epoch": 4.2, + "learning_rate": 1.7420640698944302e-06, + "loss": 0.2275, + "step": 23087 + }, + { + "epoch": 4.2, + "learning_rate": 1.7420094648707681e-06, + "loss": 0.0519, + "step": 23088 + }, + { + "epoch": 4.2, + "learning_rate": 1.741954859847106e-06, + "loss": 0.5917, + "step": 23089 + }, + { + "epoch": 4.2, + "learning_rate": 1.7419002548234438e-06, + "loss": 1.2015, + "step": 23090 + }, + { + "epoch": 4.2, + "learning_rate": 1.7418456497997817e-06, + "loss": 1.0969, + "step": 23091 + }, + { + "epoch": 4.2, + "learning_rate": 1.7417910447761197e-06, + "loss": 0.0316, + "step": 23092 + }, + { + "epoch": 4.2, + "learning_rate": 1.7417364397524572e-06, + "loss": 0.0112, + "step": 23093 + }, + { + "epoch": 4.2, + "learning_rate": 1.7416818347287951e-06, + "loss": 0.1071, + "step": 23094 + }, + { + "epoch": 4.2, + "learning_rate": 1.7416272297051329e-06, + "loss": 0.4983, + "step": 23095 + }, + { + "epoch": 4.2, + "learning_rate": 1.7415726246814708e-06, + "loss": 0.6976, + "step": 23096 + }, + { + "epoch": 4.2, + "learning_rate": 1.7415180196578088e-06, + "loss": 0.5053, + "step": 23097 + }, + { + "epoch": 4.2, + "learning_rate": 1.7414634146341463e-06, + "loss": 0.0331, + "step": 23098 + }, + { + "epoch": 4.2, + "learning_rate": 1.7414088096104842e-06, + "loss": 0.8466, + "step": 23099 + }, + { + "epoch": 4.2, + "learning_rate": 1.7413542045868222e-06, + "loss": 1.4751, + "step": 23100 + }, + { + "epoch": 4.2, + "learning_rate": 1.7412995995631597e-06, + "loss": 0.2317, + "step": 23101 + }, + { + "epoch": 4.2, + "learning_rate": 1.7412449945394976e-06, + "loss": 0.2616, + "step": 23102 + }, + { + "epoch": 4.2, + "learning_rate": 1.7411903895158356e-06, + "loss": 0.0153, + "step": 23103 + }, + { + "epoch": 4.2, + "learning_rate": 1.7411357844921733e-06, + "loss": 0.4193, + "step": 23104 + }, + { + "epoch": 4.2, + "learning_rate": 1.7410811794685112e-06, + "loss": 0.1718, + "step": 23105 + }, + { + "epoch": 4.2, + "learning_rate": 1.7410265744448488e-06, + "loss": 0.879, + "step": 23106 + }, + { + "epoch": 4.2, + "learning_rate": 1.7409719694211867e-06, + "loss": 1.5777, + "step": 23107 + }, + { + "epoch": 4.2, + "learning_rate": 1.7409173643975247e-06, + "loss": 0.3052, + "step": 23108 + }, + { + "epoch": 4.2, + "learning_rate": 1.7408627593738624e-06, + "loss": 0.008, + "step": 23109 + }, + { + "epoch": 4.2, + "learning_rate": 1.7408081543502003e-06, + "loss": 1.1619, + "step": 23110 + }, + { + "epoch": 4.2, + "learning_rate": 1.7407535493265383e-06, + "loss": 1.6729, + "step": 23111 + }, + { + "epoch": 4.2, + "learning_rate": 1.7406989443028758e-06, + "loss": 0.0111, + "step": 23112 + }, + { + "epoch": 4.2, + "learning_rate": 1.7406443392792137e-06, + "loss": 0.158, + "step": 23113 + }, + { + "epoch": 4.2, + "learning_rate": 1.7405897342555517e-06, + "loss": 1.7712, + "step": 23114 + }, + { + "epoch": 4.2, + "learning_rate": 1.7405351292318892e-06, + "loss": 0.8093, + "step": 23115 + }, + { + "epoch": 4.2, + "learning_rate": 1.7404805242082271e-06, + "loss": 2.0267, + "step": 23116 + }, + { + "epoch": 4.2, + "learning_rate": 1.740425919184565e-06, + "loss": 0.2698, + "step": 23117 + }, + { + "epoch": 4.2, + "learning_rate": 1.7403713141609028e-06, + "loss": 0.3473, + "step": 23118 + }, + { + "epoch": 4.2, + "learning_rate": 1.7403167091372408e-06, + "loss": 0.2455, + "step": 23119 + }, + { + "epoch": 4.2, + "learning_rate": 1.7402621041135783e-06, + "loss": 1.0445, + "step": 23120 + }, + { + "epoch": 4.2, + "learning_rate": 1.7402074990899162e-06, + "loss": 1.2232, + "step": 23121 + }, + { + "epoch": 4.2, + "learning_rate": 1.7401528940662542e-06, + "loss": 0.6254, + "step": 23122 + }, + { + "epoch": 4.2, + "learning_rate": 1.740098289042592e-06, + "loss": 2.7732, + "step": 23123 + }, + { + "epoch": 4.21, + "learning_rate": 1.7400436840189298e-06, + "loss": 0.2065, + "step": 23124 + }, + { + "epoch": 4.21, + "learning_rate": 1.7399890789952678e-06, + "loss": 0.462, + "step": 23125 + }, + { + "epoch": 4.21, + "learning_rate": 1.7399344739716053e-06, + "loss": 0.0089, + "step": 23126 + }, + { + "epoch": 4.21, + "learning_rate": 1.7398798689479432e-06, + "loss": 0.1981, + "step": 23127 + }, + { + "epoch": 4.21, + "learning_rate": 1.7398252639242812e-06, + "loss": 0.0065, + "step": 23128 + }, + { + "epoch": 4.21, + "learning_rate": 1.7397706589006187e-06, + "loss": 0.0144, + "step": 23129 + }, + { + "epoch": 4.21, + "learning_rate": 1.7397160538769567e-06, + "loss": 0.0123, + "step": 23130 + }, + { + "epoch": 4.21, + "learning_rate": 1.7396614488532946e-06, + "loss": 0.7073, + "step": 23131 + }, + { + "epoch": 4.21, + "learning_rate": 1.7396068438296323e-06, + "loss": 0.9307, + "step": 23132 + }, + { + "epoch": 4.21, + "learning_rate": 1.7395522388059703e-06, + "loss": 0.1605, + "step": 23133 + }, + { + "epoch": 4.21, + "learning_rate": 1.7394976337823078e-06, + "loss": 0.9728, + "step": 23134 + }, + { + "epoch": 4.21, + "learning_rate": 1.7394430287586457e-06, + "loss": 2.5152, + "step": 23135 + }, + { + "epoch": 4.21, + "learning_rate": 1.7393884237349837e-06, + "loss": 1.2475, + "step": 23136 + }, + { + "epoch": 4.21, + "learning_rate": 1.7393338187113214e-06, + "loss": 2.148, + "step": 23137 + }, + { + "epoch": 4.21, + "learning_rate": 1.7392792136876594e-06, + "loss": 0.4325, + "step": 23138 + }, + { + "epoch": 4.21, + "learning_rate": 1.7392246086639973e-06, + "loss": 0.3606, + "step": 23139 + }, + { + "epoch": 4.21, + "learning_rate": 1.7391700036403348e-06, + "loss": 1.4294, + "step": 23140 + }, + { + "epoch": 4.21, + "learning_rate": 1.7391153986166728e-06, + "loss": 0.3102, + "step": 23141 + }, + { + "epoch": 4.21, + "learning_rate": 1.7390607935930107e-06, + "loss": 1.2055, + "step": 23142 + }, + { + "epoch": 4.21, + "learning_rate": 1.7390061885693484e-06, + "loss": 0.2885, + "step": 23143 + }, + { + "epoch": 4.21, + "learning_rate": 1.7389515835456864e-06, + "loss": 1.6703, + "step": 23144 + }, + { + "epoch": 4.21, + "learning_rate": 1.7388969785220243e-06, + "loss": 0.0706, + "step": 23145 + }, + { + "epoch": 4.21, + "learning_rate": 1.7388423734983618e-06, + "loss": 1.3681, + "step": 23146 + }, + { + "epoch": 4.21, + "learning_rate": 1.7387877684746998e-06, + "loss": 0.0141, + "step": 23147 + }, + { + "epoch": 4.21, + "learning_rate": 1.7387331634510373e-06, + "loss": 0.2938, + "step": 23148 + }, + { + "epoch": 4.21, + "learning_rate": 1.7386785584273753e-06, + "loss": 2.78, + "step": 23149 + }, + { + "epoch": 4.21, + "learning_rate": 1.7386239534037132e-06, + "loss": 0.4845, + "step": 23150 + }, + { + "epoch": 4.21, + "learning_rate": 1.738569348380051e-06, + "loss": 0.3058, + "step": 23151 + }, + { + "epoch": 4.21, + "learning_rate": 1.7385147433563889e-06, + "loss": 0.4306, + "step": 23152 + }, + { + "epoch": 4.21, + "learning_rate": 1.7384601383327268e-06, + "loss": 0.1994, + "step": 23153 + }, + { + "epoch": 4.21, + "learning_rate": 1.7384055333090643e-06, + "loss": 1.1181, + "step": 23154 + }, + { + "epoch": 4.21, + "learning_rate": 1.7383509282854023e-06, + "loss": 0.381, + "step": 23155 + }, + { + "epoch": 4.21, + "learning_rate": 1.7382963232617402e-06, + "loss": 0.0157, + "step": 23156 + }, + { + "epoch": 4.21, + "learning_rate": 1.738241718238078e-06, + "loss": 0.5035, + "step": 23157 + }, + { + "epoch": 4.21, + "learning_rate": 1.7381871132144159e-06, + "loss": 0.9737, + "step": 23158 + }, + { + "epoch": 4.21, + "learning_rate": 1.7381325081907538e-06, + "loss": 2.004, + "step": 23159 + }, + { + "epoch": 4.21, + "learning_rate": 1.7380779031670914e-06, + "loss": 2.3044, + "step": 23160 + }, + { + "epoch": 4.21, + "learning_rate": 1.7380232981434293e-06, + "loss": 0.6434, + "step": 23161 + }, + { + "epoch": 4.21, + "learning_rate": 1.7379686931197668e-06, + "loss": 0.0166, + "step": 23162 + }, + { + "epoch": 4.21, + "learning_rate": 1.7379140880961048e-06, + "loss": 0.0435, + "step": 23163 + }, + { + "epoch": 4.21, + "learning_rate": 1.7378594830724427e-06, + "loss": 0.009, + "step": 23164 + }, + { + "epoch": 4.21, + "learning_rate": 1.7378048780487804e-06, + "loss": 0.4575, + "step": 23165 + }, + { + "epoch": 4.21, + "learning_rate": 1.7377502730251184e-06, + "loss": 0.2068, + "step": 23166 + }, + { + "epoch": 4.21, + "learning_rate": 1.7376956680014563e-06, + "loss": 1.4574, + "step": 23167 + }, + { + "epoch": 4.21, + "learning_rate": 1.7376410629777938e-06, + "loss": 0.6073, + "step": 23168 + }, + { + "epoch": 4.21, + "learning_rate": 1.7375864579541318e-06, + "loss": 0.3616, + "step": 23169 + }, + { + "epoch": 4.21, + "learning_rate": 1.7375318529304697e-06, + "loss": 1.0045, + "step": 23170 + }, + { + "epoch": 4.21, + "learning_rate": 1.7374772479068075e-06, + "loss": 0.2852, + "step": 23171 + }, + { + "epoch": 4.21, + "learning_rate": 1.7374226428831454e-06, + "loss": 1.1757, + "step": 23172 + }, + { + "epoch": 4.21, + "learning_rate": 1.7373680378594834e-06, + "loss": 0.0144, + "step": 23173 + }, + { + "epoch": 4.21, + "learning_rate": 1.7373134328358209e-06, + "loss": 1.7502, + "step": 23174 + }, + { + "epoch": 4.21, + "learning_rate": 1.7372588278121588e-06, + "loss": 0.2762, + "step": 23175 + }, + { + "epoch": 4.21, + "learning_rate": 1.7372042227884965e-06, + "loss": 0.1511, + "step": 23176 + }, + { + "epoch": 4.21, + "learning_rate": 1.7371496177648345e-06, + "loss": 0.2735, + "step": 23177 + }, + { + "epoch": 4.21, + "learning_rate": 1.7370950127411724e-06, + "loss": 0.7552, + "step": 23178 + }, + { + "epoch": 4.22, + "learning_rate": 1.73704040771751e-06, + "loss": 1.2914, + "step": 23179 + }, + { + "epoch": 4.22, + "learning_rate": 1.736985802693848e-06, + "loss": 1.3246, + "step": 23180 + }, + { + "epoch": 4.22, + "learning_rate": 1.7369311976701858e-06, + "loss": 0.2523, + "step": 23181 + }, + { + "epoch": 4.22, + "learning_rate": 1.7368765926465234e-06, + "loss": 0.739, + "step": 23182 + }, + { + "epoch": 4.22, + "learning_rate": 1.7368219876228613e-06, + "loss": 1.3607, + "step": 23183 + }, + { + "epoch": 4.22, + "learning_rate": 1.7367673825991992e-06, + "loss": 0.3519, + "step": 23184 + }, + { + "epoch": 4.22, + "learning_rate": 1.736712777575537e-06, + "loss": 0.463, + "step": 23185 + }, + { + "epoch": 4.22, + "learning_rate": 1.736658172551875e-06, + "loss": 0.4836, + "step": 23186 + }, + { + "epoch": 4.22, + "learning_rate": 1.7366035675282124e-06, + "loss": 0.2086, + "step": 23187 + }, + { + "epoch": 4.22, + "learning_rate": 1.7365489625045504e-06, + "loss": 0.0815, + "step": 23188 + }, + { + "epoch": 4.22, + "learning_rate": 1.7364943574808883e-06, + "loss": 0.22, + "step": 23189 + }, + { + "epoch": 4.22, + "learning_rate": 1.736439752457226e-06, + "loss": 0.5067, + "step": 23190 + }, + { + "epoch": 4.22, + "learning_rate": 1.736385147433564e-06, + "loss": 0.8752, + "step": 23191 + }, + { + "epoch": 4.22, + "learning_rate": 1.736330542409902e-06, + "loss": 0.0152, + "step": 23192 + }, + { + "epoch": 4.22, + "learning_rate": 1.7362759373862395e-06, + "loss": 1.7954, + "step": 23193 + }, + { + "epoch": 4.22, + "learning_rate": 1.7362213323625774e-06, + "loss": 0.6789, + "step": 23194 + }, + { + "epoch": 4.22, + "learning_rate": 1.7361667273389154e-06, + "loss": 0.1537, + "step": 23195 + }, + { + "epoch": 4.22, + "learning_rate": 1.7361121223152529e-06, + "loss": 2.7014, + "step": 23196 + }, + { + "epoch": 4.22, + "learning_rate": 1.7360575172915908e-06, + "loss": 0.1182, + "step": 23197 + }, + { + "epoch": 4.22, + "learning_rate": 1.7360029122679288e-06, + "loss": 0.4371, + "step": 23198 + }, + { + "epoch": 4.22, + "learning_rate": 1.7359483072442665e-06, + "loss": 0.0575, + "step": 23199 + }, + { + "epoch": 4.22, + "learning_rate": 1.7358937022206044e-06, + "loss": 1.1188, + "step": 23200 + }, + { + "epoch": 4.22, + "learning_rate": 1.735839097196942e-06, + "loss": 0.7435, + "step": 23201 + }, + { + "epoch": 4.22, + "learning_rate": 1.73578449217328e-06, + "loss": 0.6159, + "step": 23202 + }, + { + "epoch": 4.22, + "learning_rate": 1.7357298871496178e-06, + "loss": 0.2543, + "step": 23203 + }, + { + "epoch": 4.22, + "learning_rate": 1.7356752821259556e-06, + "loss": 0.678, + "step": 23204 + }, + { + "epoch": 4.22, + "learning_rate": 1.7356206771022935e-06, + "loss": 0.0309, + "step": 23205 + }, + { + "epoch": 4.22, + "learning_rate": 1.7355660720786315e-06, + "loss": 0.6124, + "step": 23206 + }, + { + "epoch": 4.22, + "learning_rate": 1.735511467054969e-06, + "loss": 0.5166, + "step": 23207 + }, + { + "epoch": 4.22, + "learning_rate": 1.735456862031307e-06, + "loss": 0.062, + "step": 23208 + }, + { + "epoch": 4.22, + "learning_rate": 1.7354022570076449e-06, + "loss": 0.7424, + "step": 23209 + }, + { + "epoch": 4.22, + "learning_rate": 1.7353476519839824e-06, + "loss": 0.1028, + "step": 23210 + }, + { + "epoch": 4.22, + "learning_rate": 1.7352930469603203e-06, + "loss": 0.3209, + "step": 23211 + }, + { + "epoch": 4.22, + "learning_rate": 1.7352384419366583e-06, + "loss": 0.4546, + "step": 23212 + }, + { + "epoch": 4.22, + "learning_rate": 1.735183836912996e-06, + "loss": 0.5123, + "step": 23213 + }, + { + "epoch": 4.22, + "learning_rate": 1.735129231889334e-06, + "loss": 0.0254, + "step": 23214 + }, + { + "epoch": 4.22, + "learning_rate": 1.7350746268656715e-06, + "loss": 0.251, + "step": 23215 + }, + { + "epoch": 4.22, + "learning_rate": 1.7350200218420094e-06, + "loss": 0.2376, + "step": 23216 + }, + { + "epoch": 4.22, + "learning_rate": 1.7349654168183474e-06, + "loss": 0.3044, + "step": 23217 + }, + { + "epoch": 4.22, + "learning_rate": 1.734910811794685e-06, + "loss": 0.8925, + "step": 23218 + }, + { + "epoch": 4.22, + "learning_rate": 1.734856206771023e-06, + "loss": 0.5189, + "step": 23219 + }, + { + "epoch": 4.22, + "learning_rate": 1.734801601747361e-06, + "loss": 0.8138, + "step": 23220 + }, + { + "epoch": 4.22, + "learning_rate": 1.7347469967236985e-06, + "loss": 0.0154, + "step": 23221 + }, + { + "epoch": 4.22, + "learning_rate": 1.7346923917000364e-06, + "loss": 0.9945, + "step": 23222 + }, + { + "epoch": 4.22, + "learning_rate": 1.7346377866763744e-06, + "loss": 1.0885, + "step": 23223 + }, + { + "epoch": 4.22, + "learning_rate": 1.7345831816527121e-06, + "loss": 0.0224, + "step": 23224 + }, + { + "epoch": 4.22, + "learning_rate": 1.73452857662905e-06, + "loss": 1.0844, + "step": 23225 + }, + { + "epoch": 4.22, + "learning_rate": 1.734473971605388e-06, + "loss": 0.091, + "step": 23226 + }, + { + "epoch": 4.22, + "learning_rate": 1.7344193665817255e-06, + "loss": 0.815, + "step": 23227 + }, + { + "epoch": 4.22, + "learning_rate": 1.7343647615580635e-06, + "loss": 1.5585, + "step": 23228 + }, + { + "epoch": 4.22, + "learning_rate": 1.734310156534401e-06, + "loss": 0.2105, + "step": 23229 + }, + { + "epoch": 4.22, + "learning_rate": 1.734255551510739e-06, + "loss": 0.2798, + "step": 23230 + }, + { + "epoch": 4.22, + "learning_rate": 1.7342009464870769e-06, + "loss": 0.4265, + "step": 23231 + }, + { + "epoch": 4.22, + "learning_rate": 1.7341463414634146e-06, + "loss": 2.2547, + "step": 23232 + }, + { + "epoch": 4.22, + "learning_rate": 1.7340917364397525e-06, + "loss": 0.0231, + "step": 23233 + }, + { + "epoch": 4.23, + "learning_rate": 1.7340371314160905e-06, + "loss": 1.1928, + "step": 23234 + }, + { + "epoch": 4.23, + "learning_rate": 1.733982526392428e-06, + "loss": 0.6762, + "step": 23235 + }, + { + "epoch": 4.23, + "learning_rate": 1.733927921368766e-06, + "loss": 0.0276, + "step": 23236 + }, + { + "epoch": 4.23, + "learning_rate": 1.7338733163451039e-06, + "loss": 0.0586, + "step": 23237 + }, + { + "epoch": 4.23, + "learning_rate": 1.7338187113214416e-06, + "loss": 2.6597, + "step": 23238 + }, + { + "epoch": 4.23, + "learning_rate": 1.7337641062977796e-06, + "loss": 0.0441, + "step": 23239 + }, + { + "epoch": 4.23, + "learning_rate": 1.7337095012741175e-06, + "loss": 0.3558, + "step": 23240 + }, + { + "epoch": 4.23, + "learning_rate": 1.733654896250455e-06, + "loss": 0.6009, + "step": 23241 + }, + { + "epoch": 4.23, + "learning_rate": 1.733600291226793e-06, + "loss": 3.497, + "step": 23242 + }, + { + "epoch": 4.23, + "learning_rate": 1.7335456862031305e-06, + "loss": 0.0967, + "step": 23243 + }, + { + "epoch": 4.23, + "learning_rate": 1.7334910811794684e-06, + "loss": 0.0177, + "step": 23244 + }, + { + "epoch": 4.23, + "learning_rate": 1.7334364761558064e-06, + "loss": 0.0471, + "step": 23245 + }, + { + "epoch": 4.23, + "learning_rate": 1.7333818711321441e-06, + "loss": 0.6178, + "step": 23246 + }, + { + "epoch": 4.23, + "learning_rate": 1.733327266108482e-06, + "loss": 0.1615, + "step": 23247 + }, + { + "epoch": 4.23, + "learning_rate": 1.73327266108482e-06, + "loss": 1.988, + "step": 23248 + }, + { + "epoch": 4.23, + "learning_rate": 1.7332180560611575e-06, + "loss": 1.6229, + "step": 23249 + }, + { + "epoch": 4.23, + "learning_rate": 1.7331634510374955e-06, + "loss": 0.3603, + "step": 23250 + }, + { + "epoch": 4.23, + "learning_rate": 1.7331088460138334e-06, + "loss": 0.0594, + "step": 23251 + }, + { + "epoch": 4.23, + "learning_rate": 1.7330542409901711e-06, + "loss": 0.4821, + "step": 23252 + }, + { + "epoch": 4.23, + "learning_rate": 1.732999635966509e-06, + "loss": 0.4005, + "step": 23253 + }, + { + "epoch": 4.23, + "learning_rate": 1.7329450309428466e-06, + "loss": 0.2876, + "step": 23254 + }, + { + "epoch": 4.23, + "learning_rate": 1.7328904259191845e-06, + "loss": 2.4375, + "step": 23255 + }, + { + "epoch": 4.23, + "learning_rate": 1.7328358208955225e-06, + "loss": 0.0131, + "step": 23256 + }, + { + "epoch": 4.23, + "learning_rate": 1.73278121587186e-06, + "loss": 0.2929, + "step": 23257 + }, + { + "epoch": 4.23, + "learning_rate": 1.732726610848198e-06, + "loss": 1.2707, + "step": 23258 + }, + { + "epoch": 4.23, + "learning_rate": 1.732672005824536e-06, + "loss": 0.5307, + "step": 23259 + }, + { + "epoch": 4.23, + "learning_rate": 1.7326174008008736e-06, + "loss": 0.4927, + "step": 23260 + }, + { + "epoch": 4.23, + "learning_rate": 1.7325627957772116e-06, + "loss": 0.611, + "step": 23261 + }, + { + "epoch": 4.23, + "learning_rate": 1.7325081907535495e-06, + "loss": 0.4659, + "step": 23262 + }, + { + "epoch": 4.23, + "learning_rate": 1.732453585729887e-06, + "loss": 0.7377, + "step": 23263 + }, + { + "epoch": 4.23, + "learning_rate": 1.732398980706225e-06, + "loss": 0.0226, + "step": 23264 + }, + { + "epoch": 4.23, + "learning_rate": 1.732344375682563e-06, + "loss": 1.6468, + "step": 23265 + }, + { + "epoch": 4.23, + "learning_rate": 1.7322897706589007e-06, + "loss": 0.0302, + "step": 23266 + }, + { + "epoch": 4.23, + "learning_rate": 1.7322351656352386e-06, + "loss": 0.0957, + "step": 23267 + }, + { + "epoch": 4.23, + "learning_rate": 1.7321805606115761e-06, + "loss": 0.0459, + "step": 23268 + }, + { + "epoch": 4.23, + "learning_rate": 1.732125955587914e-06, + "loss": 0.5324, + "step": 23269 + }, + { + "epoch": 4.23, + "learning_rate": 1.732071350564252e-06, + "loss": 2.6098, + "step": 23270 + }, + { + "epoch": 4.23, + "learning_rate": 1.7320167455405897e-06, + "loss": 2.0143, + "step": 23271 + }, + { + "epoch": 4.23, + "learning_rate": 1.7319621405169277e-06, + "loss": 0.1776, + "step": 23272 + }, + { + "epoch": 4.23, + "learning_rate": 1.7319075354932656e-06, + "loss": 0.251, + "step": 23273 + }, + { + "epoch": 4.23, + "learning_rate": 1.7318529304696031e-06, + "loss": 0.0092, + "step": 23274 + }, + { + "epoch": 4.23, + "learning_rate": 1.731798325445941e-06, + "loss": 0.0293, + "step": 23275 + }, + { + "epoch": 4.23, + "learning_rate": 1.731743720422279e-06, + "loss": 1.4482, + "step": 23276 + }, + { + "epoch": 4.23, + "learning_rate": 1.7316891153986165e-06, + "loss": 0.3832, + "step": 23277 + }, + { + "epoch": 4.23, + "learning_rate": 1.7316345103749545e-06, + "loss": 3.3729, + "step": 23278 + }, + { + "epoch": 4.23, + "learning_rate": 1.7315799053512924e-06, + "loss": 0.0155, + "step": 23279 + }, + { + "epoch": 4.23, + "learning_rate": 1.7315253003276302e-06, + "loss": 1.5196, + "step": 23280 + }, + { + "epoch": 4.23, + "learning_rate": 1.7314706953039681e-06, + "loss": 1.1496, + "step": 23281 + }, + { + "epoch": 4.23, + "learning_rate": 1.7314160902803056e-06, + "loss": 0.2599, + "step": 23282 + }, + { + "epoch": 4.23, + "learning_rate": 1.7313614852566436e-06, + "loss": 2.4624, + "step": 23283 + }, + { + "epoch": 4.23, + "learning_rate": 1.7313068802329815e-06, + "loss": 0.5969, + "step": 23284 + }, + { + "epoch": 4.23, + "learning_rate": 1.7312522752093192e-06, + "loss": 0.5264, + "step": 23285 + }, + { + "epoch": 4.23, + "learning_rate": 1.7311976701856572e-06, + "loss": 0.0783, + "step": 23286 + }, + { + "epoch": 4.23, + "learning_rate": 1.7311430651619951e-06, + "loss": 0.8658, + "step": 23287 + }, + { + "epoch": 4.23, + "learning_rate": 1.7310884601383327e-06, + "loss": 1.4961, + "step": 23288 + }, + { + "epoch": 4.24, + "learning_rate": 1.7310338551146706e-06, + "loss": 0.2862, + "step": 23289 + }, + { + "epoch": 4.24, + "learning_rate": 1.7309792500910085e-06, + "loss": 0.5399, + "step": 23290 + }, + { + "epoch": 4.24, + "learning_rate": 1.730924645067346e-06, + "loss": 0.882, + "step": 23291 + }, + { + "epoch": 4.24, + "learning_rate": 1.730870040043684e-06, + "loss": 2.5444, + "step": 23292 + }, + { + "epoch": 4.24, + "learning_rate": 1.730815435020022e-06, + "loss": 0.5453, + "step": 23293 + }, + { + "epoch": 4.24, + "learning_rate": 1.7307608299963597e-06, + "loss": 2.3922, + "step": 23294 + }, + { + "epoch": 4.24, + "learning_rate": 1.7307062249726976e-06, + "loss": 0.9473, + "step": 23295 + }, + { + "epoch": 4.24, + "learning_rate": 1.7306516199490351e-06, + "loss": 2.8699, + "step": 23296 + }, + { + "epoch": 4.24, + "learning_rate": 1.730597014925373e-06, + "loss": 0.0246, + "step": 23297 + }, + { + "epoch": 4.24, + "learning_rate": 1.730542409901711e-06, + "loss": 0.0107, + "step": 23298 + }, + { + "epoch": 4.24, + "learning_rate": 1.7304878048780488e-06, + "loss": 0.6644, + "step": 23299 + }, + { + "epoch": 4.24, + "learning_rate": 1.7304331998543867e-06, + "loss": 0.0696, + "step": 23300 + }, + { + "epoch": 4.24, + "learning_rate": 1.7303785948307246e-06, + "loss": 0.7989, + "step": 23301 + }, + { + "epoch": 4.24, + "learning_rate": 1.7303239898070622e-06, + "loss": 0.7302, + "step": 23302 + }, + { + "epoch": 4.24, + "learning_rate": 1.7302693847834001e-06, + "loss": 0.1777, + "step": 23303 + }, + { + "epoch": 4.24, + "learning_rate": 1.730214779759738e-06, + "loss": 0.3826, + "step": 23304 + }, + { + "epoch": 4.24, + "learning_rate": 1.7301601747360758e-06, + "loss": 0.7805, + "step": 23305 + }, + { + "epoch": 4.24, + "learning_rate": 1.7301055697124137e-06, + "loss": 1.6546, + "step": 23306 + }, + { + "epoch": 4.24, + "learning_rate": 1.7300509646887517e-06, + "loss": 0.043, + "step": 23307 + }, + { + "epoch": 4.24, + "learning_rate": 1.7299963596650892e-06, + "loss": 0.1577, + "step": 23308 + }, + { + "epoch": 4.24, + "learning_rate": 1.7299417546414271e-06, + "loss": 0.3023, + "step": 23309 + }, + { + "epoch": 4.24, + "learning_rate": 1.7298871496177647e-06, + "loss": 0.0869, + "step": 23310 + }, + { + "epoch": 4.24, + "learning_rate": 1.7298325445941026e-06, + "loss": 1.0181, + "step": 23311 + }, + { + "epoch": 4.24, + "learning_rate": 1.7297779395704405e-06, + "loss": 0.015, + "step": 23312 + }, + { + "epoch": 4.24, + "learning_rate": 1.7297233345467783e-06, + "loss": 1.0989, + "step": 23313 + }, + { + "epoch": 4.24, + "learning_rate": 1.7296687295231162e-06, + "loss": 2.604, + "step": 23314 + }, + { + "epoch": 4.24, + "learning_rate": 1.7296141244994542e-06, + "loss": 0.307, + "step": 23315 + }, + { + "epoch": 4.24, + "learning_rate": 1.7295595194757917e-06, + "loss": 1.6072, + "step": 23316 + }, + { + "epoch": 4.24, + "learning_rate": 1.7295049144521296e-06, + "loss": 0.0212, + "step": 23317 + }, + { + "epoch": 4.24, + "learning_rate": 1.7294503094284676e-06, + "loss": 0.027, + "step": 23318 + }, + { + "epoch": 4.24, + "learning_rate": 1.7293957044048053e-06, + "loss": 2.1757, + "step": 23319 + }, + { + "epoch": 4.24, + "learning_rate": 1.7293410993811432e-06, + "loss": 0.7055, + "step": 23320 + }, + { + "epoch": 4.24, + "learning_rate": 1.7292864943574812e-06, + "loss": 0.2313, + "step": 23321 + }, + { + "epoch": 4.24, + "learning_rate": 1.7292318893338187e-06, + "loss": 1.2845, + "step": 23322 + }, + { + "epoch": 4.24, + "learning_rate": 1.7291772843101566e-06, + "loss": 0.5671, + "step": 23323 + }, + { + "epoch": 4.24, + "learning_rate": 1.7291226792864942e-06, + "loss": 1.2307, + "step": 23324 + }, + { + "epoch": 4.24, + "learning_rate": 1.7290680742628321e-06, + "loss": 1.0293, + "step": 23325 + }, + { + "epoch": 4.24, + "learning_rate": 1.72901346923917e-06, + "loss": 0.9432, + "step": 23326 + }, + { + "epoch": 4.24, + "learning_rate": 1.7289588642155078e-06, + "loss": 0.4221, + "step": 23327 + }, + { + "epoch": 4.24, + "learning_rate": 1.7289042591918457e-06, + "loss": 0.0468, + "step": 23328 + }, + { + "epoch": 4.24, + "learning_rate": 1.7288496541681837e-06, + "loss": 1.3957, + "step": 23329 + }, + { + "epoch": 4.24, + "learning_rate": 1.7287950491445212e-06, + "loss": 1.858, + "step": 23330 + }, + { + "epoch": 4.24, + "learning_rate": 1.7287404441208591e-06, + "loss": 1.2816, + "step": 23331 + }, + { + "epoch": 4.24, + "learning_rate": 1.728685839097197e-06, + "loss": 0.0093, + "step": 23332 + }, + { + "epoch": 4.24, + "learning_rate": 1.7286312340735348e-06, + "loss": 0.0329, + "step": 23333 + }, + { + "epoch": 4.24, + "learning_rate": 1.7285766290498728e-06, + "loss": 1.3, + "step": 23334 + }, + { + "epoch": 4.24, + "learning_rate": 1.7285220240262103e-06, + "loss": 0.307, + "step": 23335 + }, + { + "epoch": 4.24, + "learning_rate": 1.7284674190025482e-06, + "loss": 0.0839, + "step": 23336 + }, + { + "epoch": 4.24, + "learning_rate": 1.7284128139788862e-06, + "loss": 1.0756, + "step": 23337 + }, + { + "epoch": 4.24, + "learning_rate": 1.7283582089552237e-06, + "loss": 1.1567, + "step": 23338 + }, + { + "epoch": 4.24, + "learning_rate": 1.7283036039315616e-06, + "loss": 0.2181, + "step": 23339 + }, + { + "epoch": 4.24, + "learning_rate": 1.7282489989078996e-06, + "loss": 0.2958, + "step": 23340 + }, + { + "epoch": 4.24, + "learning_rate": 1.7281943938842373e-06, + "loss": 0.0776, + "step": 23341 + }, + { + "epoch": 4.24, + "learning_rate": 1.7281397888605752e-06, + "loss": 0.4572, + "step": 23342 + }, + { + "epoch": 4.24, + "learning_rate": 1.7280851838369132e-06, + "loss": 0.0137, + "step": 23343 + }, + { + "epoch": 4.25, + "learning_rate": 1.7280305788132507e-06, + "loss": 0.6812, + "step": 23344 + }, + { + "epoch": 4.25, + "learning_rate": 1.7279759737895887e-06, + "loss": 0.1615, + "step": 23345 + }, + { + "epoch": 4.25, + "learning_rate": 1.7279213687659266e-06, + "loss": 0.4057, + "step": 23346 + }, + { + "epoch": 4.25, + "learning_rate": 1.7278667637422643e-06, + "loss": 0.8502, + "step": 23347 + }, + { + "epoch": 4.25, + "learning_rate": 1.7278121587186023e-06, + "loss": 0.3383, + "step": 23348 + }, + { + "epoch": 4.25, + "learning_rate": 1.7277575536949398e-06, + "loss": 0.1624, + "step": 23349 + }, + { + "epoch": 4.25, + "learning_rate": 1.7277029486712777e-06, + "loss": 0.4802, + "step": 23350 + }, + { + "epoch": 4.25, + "learning_rate": 1.7276483436476157e-06, + "loss": 0.445, + "step": 23351 + }, + { + "epoch": 4.25, + "learning_rate": 1.7275937386239534e-06, + "loss": 0.2241, + "step": 23352 + }, + { + "epoch": 4.25, + "learning_rate": 1.7275391336002914e-06, + "loss": 0.4696, + "step": 23353 + }, + { + "epoch": 4.25, + "learning_rate": 1.7274845285766293e-06, + "loss": 0.0222, + "step": 23354 + }, + { + "epoch": 4.25, + "learning_rate": 1.7274299235529668e-06, + "loss": 2.6089, + "step": 23355 + }, + { + "epoch": 4.25, + "learning_rate": 1.7273753185293048e-06, + "loss": 0.6032, + "step": 23356 + }, + { + "epoch": 4.25, + "learning_rate": 1.7273207135056427e-06, + "loss": 0.0539, + "step": 23357 + }, + { + "epoch": 4.25, + "learning_rate": 1.7272661084819802e-06, + "loss": 0.7154, + "step": 23358 + }, + { + "epoch": 4.25, + "learning_rate": 1.7272115034583182e-06, + "loss": 1.9177, + "step": 23359 + }, + { + "epoch": 4.25, + "learning_rate": 1.7271568984346561e-06, + "loss": 1.2459, + "step": 23360 + }, + { + "epoch": 4.25, + "learning_rate": 1.7271022934109938e-06, + "loss": 0.393, + "step": 23361 + }, + { + "epoch": 4.25, + "learning_rate": 1.7270476883873318e-06, + "loss": 0.3542, + "step": 23362 + }, + { + "epoch": 4.25, + "learning_rate": 1.7269930833636693e-06, + "loss": 0.2348, + "step": 23363 + }, + { + "epoch": 4.25, + "learning_rate": 1.7269384783400072e-06, + "loss": 0.083, + "step": 23364 + }, + { + "epoch": 4.25, + "learning_rate": 1.7268838733163452e-06, + "loss": 0.3823, + "step": 23365 + }, + { + "epoch": 4.25, + "learning_rate": 1.726829268292683e-06, + "loss": 0.8095, + "step": 23366 + }, + { + "epoch": 4.25, + "learning_rate": 1.7267746632690209e-06, + "loss": 0.0747, + "step": 23367 + }, + { + "epoch": 4.25, + "learning_rate": 1.7267200582453588e-06, + "loss": 0.0633, + "step": 23368 + }, + { + "epoch": 4.25, + "learning_rate": 1.7266654532216963e-06, + "loss": 0.3603, + "step": 23369 + }, + { + "epoch": 4.25, + "learning_rate": 1.7266108481980343e-06, + "loss": 0.4163, + "step": 23370 + }, + { + "epoch": 4.25, + "learning_rate": 1.7265562431743722e-06, + "loss": 0.1758, + "step": 23371 + }, + { + "epoch": 4.25, + "learning_rate": 1.7265016381507097e-06, + "loss": 0.5168, + "step": 23372 + }, + { + "epoch": 4.25, + "learning_rate": 1.7264470331270477e-06, + "loss": 0.947, + "step": 23373 + }, + { + "epoch": 4.25, + "learning_rate": 1.7263924281033856e-06, + "loss": 0.5063, + "step": 23374 + }, + { + "epoch": 4.25, + "learning_rate": 1.7263378230797234e-06, + "loss": 0.0099, + "step": 23375 + }, + { + "epoch": 4.25, + "learning_rate": 1.7262832180560613e-06, + "loss": 1.4659, + "step": 23376 + }, + { + "epoch": 4.25, + "learning_rate": 1.7262286130323988e-06, + "loss": 0.2539, + "step": 23377 + }, + { + "epoch": 4.25, + "learning_rate": 1.7261740080087368e-06, + "loss": 0.5383, + "step": 23378 + }, + { + "epoch": 4.25, + "learning_rate": 1.7261194029850747e-06, + "loss": 0.4325, + "step": 23379 + }, + { + "epoch": 4.25, + "learning_rate": 1.7260647979614124e-06, + "loss": 2.3764, + "step": 23380 + }, + { + "epoch": 4.25, + "learning_rate": 1.7260101929377504e-06, + "loss": 0.0355, + "step": 23381 + }, + { + "epoch": 4.25, + "learning_rate": 1.7259555879140883e-06, + "loss": 0.1104, + "step": 23382 + }, + { + "epoch": 4.25, + "learning_rate": 1.7259009828904258e-06, + "loss": 0.0142, + "step": 23383 + }, + { + "epoch": 4.25, + "learning_rate": 1.7258463778667638e-06, + "loss": 0.0092, + "step": 23384 + }, + { + "epoch": 4.25, + "learning_rate": 1.7257917728431017e-06, + "loss": 1.0373, + "step": 23385 + }, + { + "epoch": 4.25, + "learning_rate": 1.7257371678194395e-06, + "loss": 0.7857, + "step": 23386 + }, + { + "epoch": 4.25, + "learning_rate": 1.7256825627957772e-06, + "loss": 0.4388, + "step": 23387 + }, + { + "epoch": 4.25, + "learning_rate": 1.7256279577721151e-06, + "loss": 1.2394, + "step": 23388 + }, + { + "epoch": 4.25, + "learning_rate": 1.7255733527484529e-06, + "loss": 0.7213, + "step": 23389 + }, + { + "epoch": 4.25, + "learning_rate": 1.7255187477247908e-06, + "loss": 0.403, + "step": 23390 + }, + { + "epoch": 4.25, + "learning_rate": 1.7254641427011283e-06, + "loss": 0.3392, + "step": 23391 + }, + { + "epoch": 4.25, + "learning_rate": 1.7254095376774663e-06, + "loss": 0.8134, + "step": 23392 + }, + { + "epoch": 4.25, + "learning_rate": 1.7253549326538042e-06, + "loss": 0.0318, + "step": 23393 + }, + { + "epoch": 4.25, + "learning_rate": 1.725300327630142e-06, + "loss": 0.0979, + "step": 23394 + }, + { + "epoch": 4.25, + "learning_rate": 1.7252457226064799e-06, + "loss": 1.5728, + "step": 23395 + }, + { + "epoch": 4.25, + "learning_rate": 1.7251911175828178e-06, + "loss": 0.4107, + "step": 23396 + }, + { + "epoch": 4.25, + "learning_rate": 1.7251365125591554e-06, + "loss": 0.1924, + "step": 23397 + }, + { + "epoch": 4.25, + "learning_rate": 1.7250819075354933e-06, + "loss": 0.2758, + "step": 23398 + }, + { + "epoch": 4.26, + "learning_rate": 1.7250273025118312e-06, + "loss": 1.9959, + "step": 23399 + }, + { + "epoch": 4.26, + "learning_rate": 1.724972697488169e-06, + "loss": 0.2886, + "step": 23400 + }, + { + "epoch": 4.26, + "learning_rate": 1.724918092464507e-06, + "loss": 0.0158, + "step": 23401 + }, + { + "epoch": 4.26, + "learning_rate": 1.7248634874408449e-06, + "loss": 2.3384, + "step": 23402 + }, + { + "epoch": 4.26, + "learning_rate": 1.7248088824171824e-06, + "loss": 1.6549, + "step": 23403 + }, + { + "epoch": 4.26, + "learning_rate": 1.7247542773935203e-06, + "loss": 1.2941, + "step": 23404 + }, + { + "epoch": 4.26, + "learning_rate": 1.7246996723698578e-06, + "loss": 0.5656, + "step": 23405 + }, + { + "epoch": 4.26, + "learning_rate": 1.7246450673461958e-06, + "loss": 0.648, + "step": 23406 + }, + { + "epoch": 4.26, + "learning_rate": 1.7245904623225337e-06, + "loss": 0.3155, + "step": 23407 + }, + { + "epoch": 4.26, + "learning_rate": 1.7245358572988715e-06, + "loss": 0.1036, + "step": 23408 + }, + { + "epoch": 4.26, + "learning_rate": 1.7244812522752094e-06, + "loss": 0.2833, + "step": 23409 + }, + { + "epoch": 4.26, + "learning_rate": 1.7244266472515473e-06, + "loss": 0.8081, + "step": 23410 + }, + { + "epoch": 4.26, + "learning_rate": 1.7243720422278849e-06, + "loss": 1.4626, + "step": 23411 + }, + { + "epoch": 4.26, + "learning_rate": 1.7243174372042228e-06, + "loss": 0.7843, + "step": 23412 + }, + { + "epoch": 4.26, + "learning_rate": 1.7242628321805608e-06, + "loss": 0.0097, + "step": 23413 + }, + { + "epoch": 4.26, + "learning_rate": 1.7242082271568985e-06, + "loss": 0.4665, + "step": 23414 + }, + { + "epoch": 4.26, + "learning_rate": 1.7241536221332364e-06, + "loss": 0.5811, + "step": 23415 + }, + { + "epoch": 4.26, + "learning_rate": 1.724099017109574e-06, + "loss": 0.7861, + "step": 23416 + }, + { + "epoch": 4.26, + "learning_rate": 1.7240444120859119e-06, + "loss": 0.1582, + "step": 23417 + }, + { + "epoch": 4.26, + "learning_rate": 1.7239898070622498e-06, + "loss": 0.5768, + "step": 23418 + }, + { + "epoch": 4.26, + "learning_rate": 1.7239352020385874e-06, + "loss": 1.2966, + "step": 23419 + }, + { + "epoch": 4.26, + "learning_rate": 1.7238805970149253e-06, + "loss": 0.1888, + "step": 23420 + }, + { + "epoch": 4.26, + "learning_rate": 1.7238259919912632e-06, + "loss": 0.0239, + "step": 23421 + }, + { + "epoch": 4.26, + "learning_rate": 1.723771386967601e-06, + "loss": 1.6085, + "step": 23422 + }, + { + "epoch": 4.26, + "learning_rate": 1.723716781943939e-06, + "loss": 1.7713, + "step": 23423 + }, + { + "epoch": 4.26, + "learning_rate": 1.7236621769202769e-06, + "loss": 0.8236, + "step": 23424 + }, + { + "epoch": 4.26, + "learning_rate": 1.7236075718966144e-06, + "loss": 1.2595, + "step": 23425 + }, + { + "epoch": 4.26, + "learning_rate": 1.7235529668729523e-06, + "loss": 0.8542, + "step": 23426 + }, + { + "epoch": 4.26, + "learning_rate": 1.7234983618492903e-06, + "loss": 0.0153, + "step": 23427 + }, + { + "epoch": 4.26, + "learning_rate": 1.723443756825628e-06, + "loss": 0.0187, + "step": 23428 + }, + { + "epoch": 4.26, + "learning_rate": 1.723389151801966e-06, + "loss": 1.3129, + "step": 23429 + }, + { + "epoch": 4.26, + "learning_rate": 1.7233345467783035e-06, + "loss": 0.0138, + "step": 23430 + }, + { + "epoch": 4.26, + "learning_rate": 1.7232799417546414e-06, + "loss": 0.2895, + "step": 23431 + }, + { + "epoch": 4.26, + "learning_rate": 1.7232253367309794e-06, + "loss": 2.8084, + "step": 23432 + }, + { + "epoch": 4.26, + "learning_rate": 1.723170731707317e-06, + "loss": 1.1657, + "step": 23433 + }, + { + "epoch": 4.26, + "learning_rate": 1.723116126683655e-06, + "loss": 1.7735, + "step": 23434 + }, + { + "epoch": 4.26, + "learning_rate": 1.723061521659993e-06, + "loss": 0.0137, + "step": 23435 + }, + { + "epoch": 4.26, + "learning_rate": 1.7230069166363305e-06, + "loss": 3.1387, + "step": 23436 + }, + { + "epoch": 4.26, + "learning_rate": 1.7229523116126684e-06, + "loss": 0.2161, + "step": 23437 + }, + { + "epoch": 4.26, + "learning_rate": 1.7228977065890064e-06, + "loss": 0.7095, + "step": 23438 + }, + { + "epoch": 4.26, + "learning_rate": 1.722843101565344e-06, + "loss": 1.3703, + "step": 23439 + }, + { + "epoch": 4.26, + "learning_rate": 1.7227884965416818e-06, + "loss": 1.6636, + "step": 23440 + }, + { + "epoch": 4.26, + "learning_rate": 1.7227338915180198e-06, + "loss": 0.2966, + "step": 23441 + }, + { + "epoch": 4.26, + "learning_rate": 1.7226792864943575e-06, + "loss": 0.0994, + "step": 23442 + }, + { + "epoch": 4.26, + "learning_rate": 1.7226246814706955e-06, + "loss": 0.1541, + "step": 23443 + }, + { + "epoch": 4.26, + "learning_rate": 1.722570076447033e-06, + "loss": 2.0689, + "step": 23444 + }, + { + "epoch": 4.26, + "learning_rate": 1.722515471423371e-06, + "loss": 0.0468, + "step": 23445 + }, + { + "epoch": 4.26, + "learning_rate": 1.7224608663997089e-06, + "loss": 2.0628, + "step": 23446 + }, + { + "epoch": 4.26, + "learning_rate": 1.7224062613760466e-06, + "loss": 0.1622, + "step": 23447 + }, + { + "epoch": 4.26, + "learning_rate": 1.7223516563523845e-06, + "loss": 0.0114, + "step": 23448 + }, + { + "epoch": 4.26, + "learning_rate": 1.7222970513287225e-06, + "loss": 0.3839, + "step": 23449 + }, + { + "epoch": 4.26, + "learning_rate": 1.72224244630506e-06, + "loss": 1.8079, + "step": 23450 + }, + { + "epoch": 4.26, + "learning_rate": 1.722187841281398e-06, + "loss": 0.722, + "step": 23451 + }, + { + "epoch": 4.26, + "learning_rate": 1.7221332362577359e-06, + "loss": 1.54, + "step": 23452 + }, + { + "epoch": 4.26, + "learning_rate": 1.7220786312340734e-06, + "loss": 0.0369, + "step": 23453 + }, + { + "epoch": 4.27, + "learning_rate": 1.7220240262104114e-06, + "loss": 0.5489, + "step": 23454 + }, + { + "epoch": 4.27, + "learning_rate": 1.7219694211867493e-06, + "loss": 0.0079, + "step": 23455 + }, + { + "epoch": 4.27, + "learning_rate": 1.721914816163087e-06, + "loss": 0.6717, + "step": 23456 + }, + { + "epoch": 4.27, + "learning_rate": 1.721860211139425e-06, + "loss": 0.357, + "step": 23457 + }, + { + "epoch": 4.27, + "learning_rate": 1.7218056061157625e-06, + "loss": 0.261, + "step": 23458 + }, + { + "epoch": 4.27, + "learning_rate": 1.7217510010921004e-06, + "loss": 0.2054, + "step": 23459 + }, + { + "epoch": 4.27, + "learning_rate": 1.7216963960684384e-06, + "loss": 0.3803, + "step": 23460 + }, + { + "epoch": 4.27, + "learning_rate": 1.7216417910447761e-06, + "loss": 0.7937, + "step": 23461 + }, + { + "epoch": 4.27, + "learning_rate": 1.721587186021114e-06, + "loss": 0.3188, + "step": 23462 + }, + { + "epoch": 4.27, + "learning_rate": 1.721532580997452e-06, + "loss": 0.2527, + "step": 23463 + }, + { + "epoch": 4.27, + "learning_rate": 1.7214779759737895e-06, + "loss": 0.0371, + "step": 23464 + }, + { + "epoch": 4.27, + "learning_rate": 1.7214233709501275e-06, + "loss": 0.1331, + "step": 23465 + }, + { + "epoch": 4.27, + "learning_rate": 1.7213687659264654e-06, + "loss": 1.1296, + "step": 23466 + }, + { + "epoch": 4.27, + "learning_rate": 1.721314160902803e-06, + "loss": 0.7209, + "step": 23467 + }, + { + "epoch": 4.27, + "learning_rate": 1.7212595558791409e-06, + "loss": 0.9425, + "step": 23468 + }, + { + "epoch": 4.27, + "learning_rate": 1.7212049508554788e-06, + "loss": 2.7394, + "step": 23469 + }, + { + "epoch": 4.27, + "learning_rate": 1.7211503458318165e-06, + "loss": 0.5664, + "step": 23470 + }, + { + "epoch": 4.27, + "learning_rate": 1.7210957408081545e-06, + "loss": 1.477, + "step": 23471 + }, + { + "epoch": 4.27, + "learning_rate": 1.721041135784492e-06, + "loss": 0.0172, + "step": 23472 + }, + { + "epoch": 4.27, + "learning_rate": 1.72098653076083e-06, + "loss": 1.5501, + "step": 23473 + }, + { + "epoch": 4.27, + "learning_rate": 1.7209319257371679e-06, + "loss": 2.0039, + "step": 23474 + }, + { + "epoch": 4.27, + "learning_rate": 1.7208773207135056e-06, + "loss": 2.8378, + "step": 23475 + }, + { + "epoch": 4.27, + "learning_rate": 1.7208227156898436e-06, + "loss": 0.2314, + "step": 23476 + }, + { + "epoch": 4.27, + "learning_rate": 1.7207681106661815e-06, + "loss": 0.6036, + "step": 23477 + }, + { + "epoch": 4.27, + "learning_rate": 1.720713505642519e-06, + "loss": 0.0107, + "step": 23478 + }, + { + "epoch": 4.27, + "learning_rate": 1.720658900618857e-06, + "loss": 0.9088, + "step": 23479 + }, + { + "epoch": 4.27, + "learning_rate": 1.720604295595195e-06, + "loss": 0.6754, + "step": 23480 + }, + { + "epoch": 4.27, + "learning_rate": 1.7205496905715326e-06, + "loss": 2.0453, + "step": 23481 + }, + { + "epoch": 4.27, + "learning_rate": 1.7204950855478706e-06, + "loss": 0.9058, + "step": 23482 + }, + { + "epoch": 4.27, + "learning_rate": 1.7204404805242085e-06, + "loss": 0.0157, + "step": 23483 + }, + { + "epoch": 4.27, + "learning_rate": 1.720385875500546e-06, + "loss": 0.1564, + "step": 23484 + }, + { + "epoch": 4.27, + "learning_rate": 1.720331270476884e-06, + "loss": 1.3135, + "step": 23485 + }, + { + "epoch": 4.27, + "learning_rate": 1.7202766654532215e-06, + "loss": 1.4238, + "step": 23486 + }, + { + "epoch": 4.27, + "learning_rate": 1.7202220604295595e-06, + "loss": 0.611, + "step": 23487 + }, + { + "epoch": 4.27, + "learning_rate": 1.7201674554058974e-06, + "loss": 0.3844, + "step": 23488 + }, + { + "epoch": 4.27, + "learning_rate": 1.7201128503822351e-06, + "loss": 0.2997, + "step": 23489 + }, + { + "epoch": 4.27, + "learning_rate": 1.720058245358573e-06, + "loss": 0.2713, + "step": 23490 + }, + { + "epoch": 4.27, + "learning_rate": 1.720003640334911e-06, + "loss": 0.4231, + "step": 23491 + }, + { + "epoch": 4.27, + "learning_rate": 1.7199490353112485e-06, + "loss": 0.0207, + "step": 23492 + }, + { + "epoch": 4.27, + "learning_rate": 1.7198944302875865e-06, + "loss": 0.5963, + "step": 23493 + }, + { + "epoch": 4.27, + "learning_rate": 1.7198398252639244e-06, + "loss": 0.0717, + "step": 23494 + }, + { + "epoch": 4.27, + "learning_rate": 1.7197852202402622e-06, + "loss": 0.0118, + "step": 23495 + }, + { + "epoch": 4.27, + "learning_rate": 1.7197306152166e-06, + "loss": 1.1855, + "step": 23496 + }, + { + "epoch": 4.27, + "learning_rate": 1.7196760101929376e-06, + "loss": 1.3008, + "step": 23497 + }, + { + "epoch": 4.27, + "learning_rate": 1.7196214051692756e-06, + "loss": 0.9215, + "step": 23498 + }, + { + "epoch": 4.27, + "learning_rate": 1.7195668001456135e-06, + "loss": 0.0574, + "step": 23499 + }, + { + "epoch": 4.27, + "learning_rate": 1.719512195121951e-06, + "loss": 0.7527, + "step": 23500 + }, + { + "epoch": 4.27, + "learning_rate": 1.719457590098289e-06, + "loss": 0.782, + "step": 23501 + }, + { + "epoch": 4.27, + "learning_rate": 1.719402985074627e-06, + "loss": 0.2251, + "step": 23502 + }, + { + "epoch": 4.27, + "learning_rate": 1.7193483800509646e-06, + "loss": 1.0501, + "step": 23503 + }, + { + "epoch": 4.27, + "learning_rate": 1.7192937750273026e-06, + "loss": 0.0253, + "step": 23504 + }, + { + "epoch": 4.27, + "learning_rate": 1.7192391700036405e-06, + "loss": 0.0196, + "step": 23505 + }, + { + "epoch": 4.27, + "learning_rate": 1.719184564979978e-06, + "loss": 0.0121, + "step": 23506 + }, + { + "epoch": 4.27, + "learning_rate": 1.719129959956316e-06, + "loss": 0.3068, + "step": 23507 + }, + { + "epoch": 4.27, + "learning_rate": 1.719075354932654e-06, + "loss": 0.7226, + "step": 23508 + }, + { + "epoch": 4.28, + "learning_rate": 1.7190207499089917e-06, + "loss": 1.2361, + "step": 23509 + }, + { + "epoch": 4.28, + "learning_rate": 1.7189661448853296e-06, + "loss": 0.3679, + "step": 23510 + }, + { + "epoch": 4.28, + "learning_rate": 1.7189115398616671e-06, + "loss": 0.25, + "step": 23511 + }, + { + "epoch": 4.28, + "learning_rate": 1.718856934838005e-06, + "loss": 0.2382, + "step": 23512 + }, + { + "epoch": 4.28, + "learning_rate": 1.718802329814343e-06, + "loss": 0.2838, + "step": 23513 + }, + { + "epoch": 4.28, + "learning_rate": 1.7187477247906808e-06, + "loss": 0.7073, + "step": 23514 + }, + { + "epoch": 4.28, + "learning_rate": 1.7186931197670187e-06, + "loss": 0.2331, + "step": 23515 + }, + { + "epoch": 4.28, + "learning_rate": 1.7186385147433566e-06, + "loss": 0.1346, + "step": 23516 + }, + { + "epoch": 4.28, + "learning_rate": 1.7185839097196942e-06, + "loss": 0.0152, + "step": 23517 + }, + { + "epoch": 4.28, + "learning_rate": 1.7185293046960321e-06, + "loss": 0.5177, + "step": 23518 + }, + { + "epoch": 4.28, + "learning_rate": 1.71847469967237e-06, + "loss": 1.9759, + "step": 23519 + }, + { + "epoch": 4.28, + "learning_rate": 1.7184200946487076e-06, + "loss": 1.3187, + "step": 23520 + }, + { + "epoch": 4.28, + "learning_rate": 1.7183654896250455e-06, + "loss": 0.792, + "step": 23521 + }, + { + "epoch": 4.28, + "learning_rate": 1.7183108846013835e-06, + "loss": 2.9135, + "step": 23522 + }, + { + "epoch": 4.28, + "learning_rate": 1.7182562795777212e-06, + "loss": 0.2134, + "step": 23523 + }, + { + "epoch": 4.28, + "learning_rate": 1.7182016745540591e-06, + "loss": 0.55, + "step": 23524 + }, + { + "epoch": 4.28, + "learning_rate": 1.7181470695303967e-06, + "loss": 0.1852, + "step": 23525 + }, + { + "epoch": 4.28, + "learning_rate": 1.7180924645067346e-06, + "loss": 0.5195, + "step": 23526 + }, + { + "epoch": 4.28, + "learning_rate": 1.7180378594830725e-06, + "loss": 0.3515, + "step": 23527 + }, + { + "epoch": 4.28, + "learning_rate": 1.7179832544594103e-06, + "loss": 0.1822, + "step": 23528 + }, + { + "epoch": 4.28, + "learning_rate": 1.7179286494357482e-06, + "loss": 0.5016, + "step": 23529 + }, + { + "epoch": 4.28, + "learning_rate": 1.7178740444120862e-06, + "loss": 0.0106, + "step": 23530 + }, + { + "epoch": 4.28, + "learning_rate": 1.7178194393884237e-06, + "loss": 0.0663, + "step": 23531 + }, + { + "epoch": 4.28, + "learning_rate": 1.7177648343647616e-06, + "loss": 1.0933, + "step": 23532 + }, + { + "epoch": 4.28, + "learning_rate": 1.7177102293410996e-06, + "loss": 0.0209, + "step": 23533 + }, + { + "epoch": 4.28, + "learning_rate": 1.717655624317437e-06, + "loss": 0.5156, + "step": 23534 + }, + { + "epoch": 4.28, + "learning_rate": 1.717601019293775e-06, + "loss": 1.5371, + "step": 23535 + }, + { + "epoch": 4.28, + "learning_rate": 1.717546414270113e-06, + "loss": 1.888, + "step": 23536 + }, + { + "epoch": 4.28, + "learning_rate": 1.7174918092464507e-06, + "loss": 0.688, + "step": 23537 + }, + { + "epoch": 4.28, + "learning_rate": 1.7174372042227886e-06, + "loss": 0.8768, + "step": 23538 + }, + { + "epoch": 4.28, + "learning_rate": 1.7173825991991262e-06, + "loss": 2.1792, + "step": 23539 + }, + { + "epoch": 4.28, + "learning_rate": 1.7173279941754641e-06, + "loss": 0.0873, + "step": 23540 + }, + { + "epoch": 4.28, + "learning_rate": 1.717273389151802e-06, + "loss": 0.1109, + "step": 23541 + }, + { + "epoch": 4.28, + "learning_rate": 1.7172187841281398e-06, + "loss": 0.7028, + "step": 23542 + }, + { + "epoch": 4.28, + "learning_rate": 1.7171641791044777e-06, + "loss": 4.2248, + "step": 23543 + }, + { + "epoch": 4.28, + "learning_rate": 1.7171095740808157e-06, + "loss": 0.0848, + "step": 23544 + }, + { + "epoch": 4.28, + "learning_rate": 1.7170549690571532e-06, + "loss": 0.5272, + "step": 23545 + }, + { + "epoch": 4.28, + "learning_rate": 1.7170003640334911e-06, + "loss": 0.5539, + "step": 23546 + }, + { + "epoch": 4.28, + "learning_rate": 1.716945759009829e-06, + "loss": 0.5114, + "step": 23547 + }, + { + "epoch": 4.28, + "learning_rate": 1.7168911539861666e-06, + "loss": 0.2775, + "step": 23548 + }, + { + "epoch": 4.28, + "learning_rate": 1.7168365489625045e-06, + "loss": 0.1192, + "step": 23549 + }, + { + "epoch": 4.28, + "learning_rate": 1.7167819439388425e-06, + "loss": 1.6255, + "step": 23550 + }, + { + "epoch": 4.28, + "learning_rate": 1.7167273389151802e-06, + "loss": 0.0251, + "step": 23551 + }, + { + "epoch": 4.28, + "learning_rate": 1.7166727338915182e-06, + "loss": 0.1132, + "step": 23552 + }, + { + "epoch": 4.28, + "learning_rate": 1.7166181288678557e-06, + "loss": 1.4065, + "step": 23553 + }, + { + "epoch": 4.28, + "learning_rate": 1.7165635238441936e-06, + "loss": 0.1759, + "step": 23554 + }, + { + "epoch": 4.28, + "learning_rate": 1.7165089188205316e-06, + "loss": 0.2082, + "step": 23555 + }, + { + "epoch": 4.28, + "learning_rate": 1.7164543137968693e-06, + "loss": 2.4201, + "step": 23556 + }, + { + "epoch": 4.28, + "learning_rate": 1.7163997087732072e-06, + "loss": 0.3693, + "step": 23557 + }, + { + "epoch": 4.28, + "learning_rate": 1.7163451037495452e-06, + "loss": 1.3181, + "step": 23558 + }, + { + "epoch": 4.28, + "learning_rate": 1.7162904987258827e-06, + "loss": 1.6085, + "step": 23559 + }, + { + "epoch": 4.28, + "learning_rate": 1.7162358937022206e-06, + "loss": 0.0144, + "step": 23560 + }, + { + "epoch": 4.28, + "learning_rate": 1.7161812886785586e-06, + "loss": 1.2241, + "step": 23561 + }, + { + "epoch": 4.28, + "learning_rate": 1.7161266836548963e-06, + "loss": 1.0738, + "step": 23562 + }, + { + "epoch": 4.28, + "learning_rate": 1.7160720786312343e-06, + "loss": 0.8614, + "step": 23563 + }, + { + "epoch": 4.29, + "learning_rate": 1.7160174736075718e-06, + "loss": 0.7012, + "step": 23564 + }, + { + "epoch": 4.29, + "learning_rate": 1.7159628685839097e-06, + "loss": 0.0099, + "step": 23565 + }, + { + "epoch": 4.29, + "learning_rate": 1.7159082635602477e-06, + "loss": 0.1601, + "step": 23566 + }, + { + "epoch": 4.29, + "learning_rate": 1.7158536585365852e-06, + "loss": 1.0372, + "step": 23567 + }, + { + "epoch": 4.29, + "learning_rate": 1.7157990535129231e-06, + "loss": 0.6365, + "step": 23568 + }, + { + "epoch": 4.29, + "learning_rate": 1.715744448489261e-06, + "loss": 0.1996, + "step": 23569 + }, + { + "epoch": 4.29, + "learning_rate": 1.7156898434655988e-06, + "loss": 0.0137, + "step": 23570 + }, + { + "epoch": 4.29, + "learning_rate": 1.7156352384419368e-06, + "loss": 5.4604, + "step": 23571 + }, + { + "epoch": 4.29, + "learning_rate": 1.7155806334182747e-06, + "loss": 0.0136, + "step": 23572 + }, + { + "epoch": 4.29, + "learning_rate": 1.7155260283946122e-06, + "loss": 0.5123, + "step": 23573 + }, + { + "epoch": 4.29, + "learning_rate": 1.7154714233709502e-06, + "loss": 0.311, + "step": 23574 + }, + { + "epoch": 4.29, + "learning_rate": 1.715416818347288e-06, + "loss": 0.3252, + "step": 23575 + }, + { + "epoch": 4.29, + "learning_rate": 1.7153622133236258e-06, + "loss": 0.0195, + "step": 23576 + }, + { + "epoch": 4.29, + "learning_rate": 1.7153076082999638e-06, + "loss": 0.7304, + "step": 23577 + }, + { + "epoch": 4.29, + "learning_rate": 1.7152530032763013e-06, + "loss": 0.26, + "step": 23578 + }, + { + "epoch": 4.29, + "learning_rate": 1.7151983982526392e-06, + "loss": 1.9507, + "step": 23579 + }, + { + "epoch": 4.29, + "learning_rate": 1.7151437932289772e-06, + "loss": 0.4997, + "step": 23580 + }, + { + "epoch": 4.29, + "learning_rate": 1.7150891882053147e-06, + "loss": 1.0475, + "step": 23581 + }, + { + "epoch": 4.29, + "learning_rate": 1.7150345831816526e-06, + "loss": 1.3589, + "step": 23582 + }, + { + "epoch": 4.29, + "learning_rate": 1.7149799781579906e-06, + "loss": 0.168, + "step": 23583 + }, + { + "epoch": 4.29, + "learning_rate": 1.7149253731343283e-06, + "loss": 1.2169, + "step": 23584 + }, + { + "epoch": 4.29, + "learning_rate": 1.7148707681106663e-06, + "loss": 1.3024, + "step": 23585 + }, + { + "epoch": 4.29, + "learning_rate": 1.7148161630870042e-06, + "loss": 0.3655, + "step": 23586 + }, + { + "epoch": 4.29, + "learning_rate": 1.7147615580633417e-06, + "loss": 0.1409, + "step": 23587 + }, + { + "epoch": 4.29, + "learning_rate": 1.7147069530396797e-06, + "loss": 0.2202, + "step": 23588 + }, + { + "epoch": 4.29, + "learning_rate": 1.7146523480160176e-06, + "loss": 0.9256, + "step": 23589 + }, + { + "epoch": 4.29, + "learning_rate": 1.7145977429923553e-06, + "loss": 0.8149, + "step": 23590 + }, + { + "epoch": 4.29, + "learning_rate": 1.7145431379686933e-06, + "loss": 1.3811, + "step": 23591 + }, + { + "epoch": 4.29, + "learning_rate": 1.7144885329450308e-06, + "loss": 0.0073, + "step": 23592 + }, + { + "epoch": 4.29, + "learning_rate": 1.7144339279213688e-06, + "loss": 1.6056, + "step": 23593 + }, + { + "epoch": 4.29, + "learning_rate": 1.7143793228977067e-06, + "loss": 0.1054, + "step": 23594 + }, + { + "epoch": 4.29, + "learning_rate": 1.7143247178740442e-06, + "loss": 0.2096, + "step": 23595 + }, + { + "epoch": 4.29, + "learning_rate": 1.7142701128503822e-06, + "loss": 2.3119, + "step": 23596 + }, + { + "epoch": 4.29, + "learning_rate": 1.71421550782672e-06, + "loss": 0.2179, + "step": 23597 + }, + { + "epoch": 4.29, + "learning_rate": 1.7141609028030578e-06, + "loss": 0.4632, + "step": 23598 + }, + { + "epoch": 4.29, + "learning_rate": 1.7141062977793958e-06, + "loss": 0.4357, + "step": 23599 + }, + { + "epoch": 4.29, + "learning_rate": 1.7140516927557337e-06, + "loss": 0.0195, + "step": 23600 + }, + { + "epoch": 4.29, + "learning_rate": 1.7139970877320712e-06, + "loss": 0.8983, + "step": 23601 + }, + { + "epoch": 4.29, + "learning_rate": 1.7139424827084092e-06, + "loss": 0.0702, + "step": 23602 + }, + { + "epoch": 4.29, + "learning_rate": 1.7138878776847471e-06, + "loss": 1.0062, + "step": 23603 + }, + { + "epoch": 4.29, + "learning_rate": 1.7138332726610849e-06, + "loss": 0.4946, + "step": 23604 + }, + { + "epoch": 4.29, + "learning_rate": 1.7137786676374228e-06, + "loss": 2.6847, + "step": 23605 + }, + { + "epoch": 4.29, + "learning_rate": 1.7137240626137603e-06, + "loss": 0.0439, + "step": 23606 + }, + { + "epoch": 4.29, + "learning_rate": 1.7136694575900983e-06, + "loss": 0.5245, + "step": 23607 + }, + { + "epoch": 4.29, + "learning_rate": 1.7136148525664362e-06, + "loss": 0.4446, + "step": 23608 + }, + { + "epoch": 4.29, + "learning_rate": 1.713560247542774e-06, + "loss": 0.5977, + "step": 23609 + }, + { + "epoch": 4.29, + "learning_rate": 1.7135056425191119e-06, + "loss": 1.5718, + "step": 23610 + }, + { + "epoch": 4.29, + "learning_rate": 1.7134510374954498e-06, + "loss": 0.9034, + "step": 23611 + }, + { + "epoch": 4.29, + "learning_rate": 1.7133964324717874e-06, + "loss": 0.3675, + "step": 23612 + }, + { + "epoch": 4.29, + "learning_rate": 1.7133418274481253e-06, + "loss": 0.2606, + "step": 23613 + }, + { + "epoch": 4.29, + "learning_rate": 1.7132872224244632e-06, + "loss": 1.5046, + "step": 23614 + }, + { + "epoch": 4.29, + "learning_rate": 1.7132326174008008e-06, + "loss": 0.797, + "step": 23615 + }, + { + "epoch": 4.29, + "learning_rate": 1.7131780123771387e-06, + "loss": 0.0514, + "step": 23616 + }, + { + "epoch": 4.29, + "learning_rate": 1.7131234073534766e-06, + "loss": 0.1949, + "step": 23617 + }, + { + "epoch": 4.29, + "learning_rate": 1.7130688023298144e-06, + "loss": 0.0638, + "step": 23618 + }, + { + "epoch": 4.3, + "learning_rate": 1.7130141973061523e-06, + "loss": 0.2978, + "step": 23619 + }, + { + "epoch": 4.3, + "learning_rate": 1.7129595922824898e-06, + "loss": 1.5701, + "step": 23620 + }, + { + "epoch": 4.3, + "learning_rate": 1.7129049872588278e-06, + "loss": 0.1168, + "step": 23621 + }, + { + "epoch": 4.3, + "learning_rate": 1.7128503822351657e-06, + "loss": 0.2035, + "step": 23622 + }, + { + "epoch": 4.3, + "learning_rate": 1.7127957772115035e-06, + "loss": 1.3277, + "step": 23623 + }, + { + "epoch": 4.3, + "learning_rate": 1.7127411721878414e-06, + "loss": 0.9162, + "step": 23624 + }, + { + "epoch": 4.3, + "learning_rate": 1.7126865671641793e-06, + "loss": 2.7286, + "step": 23625 + }, + { + "epoch": 4.3, + "learning_rate": 1.7126319621405169e-06, + "loss": 0.2122, + "step": 23626 + }, + { + "epoch": 4.3, + "learning_rate": 1.7125773571168548e-06, + "loss": 0.5983, + "step": 23627 + }, + { + "epoch": 4.3, + "learning_rate": 1.7125227520931928e-06, + "loss": 0.4406, + "step": 23628 + }, + { + "epoch": 4.3, + "learning_rate": 1.7124681470695303e-06, + "loss": 1.4331, + "step": 23629 + }, + { + "epoch": 4.3, + "learning_rate": 1.7124135420458682e-06, + "loss": 0.2694, + "step": 23630 + }, + { + "epoch": 4.3, + "learning_rate": 1.7123589370222062e-06, + "loss": 0.3952, + "step": 23631 + }, + { + "epoch": 4.3, + "learning_rate": 1.7123043319985439e-06, + "loss": 2.4817, + "step": 23632 + }, + { + "epoch": 4.3, + "learning_rate": 1.7122497269748818e-06, + "loss": 2.025, + "step": 23633 + }, + { + "epoch": 4.3, + "learning_rate": 1.7121951219512194e-06, + "loss": 0.466, + "step": 23634 + }, + { + "epoch": 4.3, + "learning_rate": 1.7121405169275573e-06, + "loss": 1.1904, + "step": 23635 + }, + { + "epoch": 4.3, + "learning_rate": 1.7120859119038952e-06, + "loss": 0.8114, + "step": 23636 + }, + { + "epoch": 4.3, + "learning_rate": 1.712031306880233e-06, + "loss": 1.6403, + "step": 23637 + }, + { + "epoch": 4.3, + "learning_rate": 1.711976701856571e-06, + "loss": 0.7614, + "step": 23638 + }, + { + "epoch": 4.3, + "learning_rate": 1.7119220968329089e-06, + "loss": 0.5939, + "step": 23639 + }, + { + "epoch": 4.3, + "learning_rate": 1.7118674918092464e-06, + "loss": 0.0095, + "step": 23640 + }, + { + "epoch": 4.3, + "learning_rate": 1.7118128867855843e-06, + "loss": 0.0425, + "step": 23641 + }, + { + "epoch": 4.3, + "learning_rate": 1.7117582817619223e-06, + "loss": 0.1268, + "step": 23642 + }, + { + "epoch": 4.3, + "learning_rate": 1.71170367673826e-06, + "loss": 2.881, + "step": 23643 + }, + { + "epoch": 4.3, + "learning_rate": 1.711649071714598e-06, + "loss": 0.3114, + "step": 23644 + }, + { + "epoch": 4.3, + "learning_rate": 1.7115944666909355e-06, + "loss": 0.3689, + "step": 23645 + }, + { + "epoch": 4.3, + "learning_rate": 1.7115398616672734e-06, + "loss": 0.3187, + "step": 23646 + }, + { + "epoch": 4.3, + "learning_rate": 1.7114852566436113e-06, + "loss": 0.0184, + "step": 23647 + }, + { + "epoch": 4.3, + "learning_rate": 1.7114306516199489e-06, + "loss": 0.3602, + "step": 23648 + }, + { + "epoch": 4.3, + "learning_rate": 1.7113760465962868e-06, + "loss": 0.0151, + "step": 23649 + }, + { + "epoch": 4.3, + "learning_rate": 1.7113214415726248e-06, + "loss": 0.0138, + "step": 23650 + }, + { + "epoch": 4.3, + "learning_rate": 1.7112668365489625e-06, + "loss": 0.2457, + "step": 23651 + }, + { + "epoch": 4.3, + "learning_rate": 1.7112122315253004e-06, + "loss": 0.0485, + "step": 23652 + }, + { + "epoch": 4.3, + "learning_rate": 1.7111576265016384e-06, + "loss": 3.333, + "step": 23653 + }, + { + "epoch": 4.3, + "learning_rate": 1.7111030214779759e-06, + "loss": 0.0215, + "step": 23654 + }, + { + "epoch": 4.3, + "learning_rate": 1.7110484164543138e-06, + "loss": 0.4401, + "step": 23655 + }, + { + "epoch": 4.3, + "learning_rate": 1.7109938114306518e-06, + "loss": 1.3687, + "step": 23656 + }, + { + "epoch": 4.3, + "learning_rate": 1.7109392064069895e-06, + "loss": 0.7707, + "step": 23657 + }, + { + "epoch": 4.3, + "learning_rate": 1.7108846013833275e-06, + "loss": 0.343, + "step": 23658 + }, + { + "epoch": 4.3, + "learning_rate": 1.710829996359665e-06, + "loss": 1.2304, + "step": 23659 + }, + { + "epoch": 4.3, + "learning_rate": 1.710775391336003e-06, + "loss": 0.1251, + "step": 23660 + }, + { + "epoch": 4.3, + "learning_rate": 1.7107207863123409e-06, + "loss": 0.1949, + "step": 23661 + }, + { + "epoch": 4.3, + "learning_rate": 1.7106661812886784e-06, + "loss": 0.1504, + "step": 23662 + }, + { + "epoch": 4.3, + "learning_rate": 1.7106115762650163e-06, + "loss": 0.5883, + "step": 23663 + }, + { + "epoch": 4.3, + "learning_rate": 1.7105569712413543e-06, + "loss": 0.2634, + "step": 23664 + }, + { + "epoch": 4.3, + "learning_rate": 1.710502366217692e-06, + "loss": 1.5568, + "step": 23665 + }, + { + "epoch": 4.3, + "learning_rate": 1.71044776119403e-06, + "loss": 0.4032, + "step": 23666 + }, + { + "epoch": 4.3, + "learning_rate": 1.7103931561703679e-06, + "loss": 0.0093, + "step": 23667 + }, + { + "epoch": 4.3, + "learning_rate": 1.7103385511467054e-06, + "loss": 1.2471, + "step": 23668 + }, + { + "epoch": 4.3, + "learning_rate": 1.7102839461230433e-06, + "loss": 0.7641, + "step": 23669 + }, + { + "epoch": 4.3, + "learning_rate": 1.7102293410993813e-06, + "loss": 2.2936, + "step": 23670 + }, + { + "epoch": 4.3, + "learning_rate": 1.710174736075719e-06, + "loss": 0.2865, + "step": 23671 + }, + { + "epoch": 4.3, + "learning_rate": 1.710120131052057e-06, + "loss": 0.6422, + "step": 23672 + }, + { + "epoch": 4.3, + "learning_rate": 1.7100655260283945e-06, + "loss": 0.1581, + "step": 23673 + }, + { + "epoch": 4.31, + "learning_rate": 1.7100109210047324e-06, + "loss": 0.9217, + "step": 23674 + }, + { + "epoch": 4.31, + "learning_rate": 1.7099563159810704e-06, + "loss": 0.3903, + "step": 23675 + }, + { + "epoch": 4.31, + "learning_rate": 1.7099017109574079e-06, + "loss": 0.0115, + "step": 23676 + }, + { + "epoch": 4.31, + "learning_rate": 1.7098471059337458e-06, + "loss": 0.6956, + "step": 23677 + }, + { + "epoch": 4.31, + "learning_rate": 1.7097925009100838e-06, + "loss": 0.5058, + "step": 23678 + }, + { + "epoch": 4.31, + "learning_rate": 1.7097378958864215e-06, + "loss": 0.7439, + "step": 23679 + }, + { + "epoch": 4.31, + "learning_rate": 1.7096832908627595e-06, + "loss": 0.2099, + "step": 23680 + }, + { + "epoch": 4.31, + "learning_rate": 1.7096286858390974e-06, + "loss": 1.8583, + "step": 23681 + }, + { + "epoch": 4.31, + "learning_rate": 1.709574080815435e-06, + "loss": 0.8659, + "step": 23682 + }, + { + "epoch": 4.31, + "learning_rate": 1.7095194757917729e-06, + "loss": 0.8207, + "step": 23683 + }, + { + "epoch": 4.31, + "learning_rate": 1.7094648707681108e-06, + "loss": 0.3051, + "step": 23684 + }, + { + "epoch": 4.31, + "learning_rate": 1.7094102657444485e-06, + "loss": 0.0136, + "step": 23685 + }, + { + "epoch": 4.31, + "learning_rate": 1.7093556607207865e-06, + "loss": 1.3077, + "step": 23686 + }, + { + "epoch": 4.31, + "learning_rate": 1.709301055697124e-06, + "loss": 0.3875, + "step": 23687 + }, + { + "epoch": 4.31, + "learning_rate": 1.709246450673462e-06, + "loss": 0.3151, + "step": 23688 + }, + { + "epoch": 4.31, + "learning_rate": 1.7091918456497999e-06, + "loss": 0.391, + "step": 23689 + }, + { + "epoch": 4.31, + "learning_rate": 1.7091372406261376e-06, + "loss": 1.0806, + "step": 23690 + }, + { + "epoch": 4.31, + "learning_rate": 1.7090826356024756e-06, + "loss": 0.2444, + "step": 23691 + }, + { + "epoch": 4.31, + "learning_rate": 1.7090280305788135e-06, + "loss": 0.9041, + "step": 23692 + }, + { + "epoch": 4.31, + "learning_rate": 1.708973425555151e-06, + "loss": 0.0556, + "step": 23693 + }, + { + "epoch": 4.31, + "learning_rate": 1.708918820531489e-06, + "loss": 0.4092, + "step": 23694 + }, + { + "epoch": 4.31, + "learning_rate": 1.708864215507827e-06, + "loss": 1.4157, + "step": 23695 + }, + { + "epoch": 4.31, + "learning_rate": 1.7088096104841644e-06, + "loss": 0.3425, + "step": 23696 + }, + { + "epoch": 4.31, + "learning_rate": 1.7087550054605024e-06, + "loss": 2.0071, + "step": 23697 + }, + { + "epoch": 4.31, + "learning_rate": 1.7087004004368403e-06, + "loss": 1.151, + "step": 23698 + }, + { + "epoch": 4.31, + "learning_rate": 1.708645795413178e-06, + "loss": 0.4793, + "step": 23699 + }, + { + "epoch": 4.31, + "learning_rate": 1.708591190389516e-06, + "loss": 0.3821, + "step": 23700 + }, + { + "epoch": 4.31, + "learning_rate": 1.7085365853658535e-06, + "loss": 1.2812, + "step": 23701 + }, + { + "epoch": 4.31, + "learning_rate": 1.7084819803421915e-06, + "loss": 0.2644, + "step": 23702 + }, + { + "epoch": 4.31, + "learning_rate": 1.7084273753185294e-06, + "loss": 0.1793, + "step": 23703 + }, + { + "epoch": 4.31, + "learning_rate": 1.7083727702948671e-06, + "loss": 0.5431, + "step": 23704 + }, + { + "epoch": 4.31, + "learning_rate": 1.708318165271205e-06, + "loss": 0.8452, + "step": 23705 + }, + { + "epoch": 4.31, + "learning_rate": 1.708263560247543e-06, + "loss": 0.4139, + "step": 23706 + }, + { + "epoch": 4.31, + "learning_rate": 1.7082089552238805e-06, + "loss": 0.709, + "step": 23707 + }, + { + "epoch": 4.31, + "learning_rate": 1.7081543502002185e-06, + "loss": 0.8833, + "step": 23708 + }, + { + "epoch": 4.31, + "learning_rate": 1.7080997451765564e-06, + "loss": 0.0293, + "step": 23709 + }, + { + "epoch": 4.31, + "learning_rate": 1.708045140152894e-06, + "loss": 0.5912, + "step": 23710 + }, + { + "epoch": 4.31, + "learning_rate": 1.7079905351292319e-06, + "loss": 3.6248, + "step": 23711 + }, + { + "epoch": 4.31, + "learning_rate": 1.7079359301055698e-06, + "loss": 1.3084, + "step": 23712 + }, + { + "epoch": 4.31, + "learning_rate": 1.7078813250819076e-06, + "loss": 0.0462, + "step": 23713 + }, + { + "epoch": 4.31, + "learning_rate": 1.7078267200582455e-06, + "loss": 0.0532, + "step": 23714 + }, + { + "epoch": 4.31, + "learning_rate": 1.707772115034583e-06, + "loss": 0.0185, + "step": 23715 + }, + { + "epoch": 4.31, + "learning_rate": 1.707717510010921e-06, + "loss": 0.996, + "step": 23716 + }, + { + "epoch": 4.31, + "learning_rate": 1.707662904987259e-06, + "loss": 0.114, + "step": 23717 + }, + { + "epoch": 4.31, + "learning_rate": 1.7076082999635966e-06, + "loss": 0.0074, + "step": 23718 + }, + { + "epoch": 4.31, + "learning_rate": 1.7075536949399346e-06, + "loss": 0.5465, + "step": 23719 + }, + { + "epoch": 4.31, + "learning_rate": 1.7074990899162725e-06, + "loss": 0.1157, + "step": 23720 + }, + { + "epoch": 4.31, + "learning_rate": 1.70744448489261e-06, + "loss": 0.0539, + "step": 23721 + }, + { + "epoch": 4.31, + "learning_rate": 1.707389879868948e-06, + "loss": 0.3091, + "step": 23722 + }, + { + "epoch": 4.31, + "learning_rate": 1.707335274845286e-06, + "loss": 0.0086, + "step": 23723 + }, + { + "epoch": 4.31, + "learning_rate": 1.7072806698216235e-06, + "loss": 0.7597, + "step": 23724 + }, + { + "epoch": 4.31, + "learning_rate": 1.7072260647979614e-06, + "loss": 0.4722, + "step": 23725 + }, + { + "epoch": 4.31, + "learning_rate": 1.7071714597742991e-06, + "loss": 0.506, + "step": 23726 + }, + { + "epoch": 4.31, + "learning_rate": 1.707116854750637e-06, + "loss": 0.0185, + "step": 23727 + }, + { + "epoch": 4.31, + "learning_rate": 1.707062249726975e-06, + "loss": 0.0136, + "step": 23728 + }, + { + "epoch": 4.32, + "learning_rate": 1.7070076447033125e-06, + "loss": 1.1404, + "step": 23729 + }, + { + "epoch": 4.32, + "learning_rate": 1.7069530396796505e-06, + "loss": 0.3541, + "step": 23730 + }, + { + "epoch": 4.32, + "learning_rate": 1.7068984346559884e-06, + "loss": 1.5002, + "step": 23731 + }, + { + "epoch": 4.32, + "learning_rate": 1.7068438296323262e-06, + "loss": 0.0085, + "step": 23732 + }, + { + "epoch": 4.32, + "learning_rate": 1.706789224608664e-06, + "loss": 2.7753, + "step": 23733 + }, + { + "epoch": 4.32, + "learning_rate": 1.706734619585002e-06, + "loss": 0.4997, + "step": 23734 + }, + { + "epoch": 4.32, + "learning_rate": 1.7066800145613396e-06, + "loss": 0.4058, + "step": 23735 + }, + { + "epoch": 4.32, + "learning_rate": 1.7066254095376775e-06, + "loss": 0.0407, + "step": 23736 + }, + { + "epoch": 4.32, + "learning_rate": 1.7065708045140155e-06, + "loss": 0.0105, + "step": 23737 + }, + { + "epoch": 4.32, + "learning_rate": 1.7065161994903532e-06, + "loss": 0.0577, + "step": 23738 + }, + { + "epoch": 4.32, + "learning_rate": 1.7064615944666911e-06, + "loss": 0.1967, + "step": 23739 + }, + { + "epoch": 4.32, + "learning_rate": 1.7064069894430286e-06, + "loss": 4.2202, + "step": 23740 + }, + { + "epoch": 4.32, + "learning_rate": 1.7063523844193666e-06, + "loss": 0.347, + "step": 23741 + }, + { + "epoch": 4.32, + "learning_rate": 1.7062977793957045e-06, + "loss": 0.2315, + "step": 23742 + }, + { + "epoch": 4.32, + "learning_rate": 1.706243174372042e-06, + "loss": 0.3771, + "step": 23743 + }, + { + "epoch": 4.32, + "learning_rate": 1.70618856934838e-06, + "loss": 0.3698, + "step": 23744 + }, + { + "epoch": 4.32, + "learning_rate": 1.706133964324718e-06, + "loss": 1.3586, + "step": 23745 + }, + { + "epoch": 4.32, + "learning_rate": 1.7060793593010557e-06, + "loss": 0.1304, + "step": 23746 + }, + { + "epoch": 4.32, + "learning_rate": 1.7060247542773936e-06, + "loss": 1.5259, + "step": 23747 + }, + { + "epoch": 4.32, + "learning_rate": 1.7059701492537316e-06, + "loss": 0.8755, + "step": 23748 + }, + { + "epoch": 4.32, + "learning_rate": 1.705915544230069e-06, + "loss": 0.3208, + "step": 23749 + }, + { + "epoch": 4.32, + "learning_rate": 1.705860939206407e-06, + "loss": 0.2087, + "step": 23750 + }, + { + "epoch": 4.32, + "learning_rate": 1.705806334182745e-06, + "loss": 0.1888, + "step": 23751 + }, + { + "epoch": 4.32, + "learning_rate": 1.7057517291590827e-06, + "loss": 0.3251, + "step": 23752 + }, + { + "epoch": 4.32, + "learning_rate": 1.7056971241354206e-06, + "loss": 0.0715, + "step": 23753 + }, + { + "epoch": 4.32, + "learning_rate": 1.7056425191117582e-06, + "loss": 0.5211, + "step": 23754 + }, + { + "epoch": 4.32, + "learning_rate": 1.705587914088096e-06, + "loss": 0.7001, + "step": 23755 + }, + { + "epoch": 4.32, + "learning_rate": 1.705533309064434e-06, + "loss": 1.2173, + "step": 23756 + }, + { + "epoch": 4.32, + "learning_rate": 1.7054787040407716e-06, + "loss": 1.964, + "step": 23757 + }, + { + "epoch": 4.32, + "learning_rate": 1.7054240990171095e-06, + "loss": 0.5608, + "step": 23758 + }, + { + "epoch": 4.32, + "learning_rate": 1.7053694939934475e-06, + "loss": 0.2211, + "step": 23759 + }, + { + "epoch": 4.32, + "learning_rate": 1.7053148889697852e-06, + "loss": 0.416, + "step": 23760 + }, + { + "epoch": 4.32, + "learning_rate": 1.7052602839461231e-06, + "loss": 0.0158, + "step": 23761 + }, + { + "epoch": 4.32, + "learning_rate": 1.705205678922461e-06, + "loss": 0.0374, + "step": 23762 + }, + { + "epoch": 4.32, + "learning_rate": 1.7051510738987986e-06, + "loss": 0.8347, + "step": 23763 + }, + { + "epoch": 4.32, + "learning_rate": 1.7050964688751365e-06, + "loss": 0.7607, + "step": 23764 + }, + { + "epoch": 4.32, + "learning_rate": 1.7050418638514745e-06, + "loss": 0.1945, + "step": 23765 + }, + { + "epoch": 4.32, + "learning_rate": 1.7049872588278122e-06, + "loss": 0.2753, + "step": 23766 + }, + { + "epoch": 4.32, + "learning_rate": 1.7049326538041502e-06, + "loss": 0.2257, + "step": 23767 + }, + { + "epoch": 4.32, + "learning_rate": 1.7048780487804877e-06, + "loss": 1.331, + "step": 23768 + }, + { + "epoch": 4.32, + "learning_rate": 1.7048234437568256e-06, + "loss": 0.0336, + "step": 23769 + }, + { + "epoch": 4.32, + "learning_rate": 1.7047688387331636e-06, + "loss": 0.5901, + "step": 23770 + }, + { + "epoch": 4.32, + "learning_rate": 1.7047142337095013e-06, + "loss": 0.2442, + "step": 23771 + }, + { + "epoch": 4.32, + "learning_rate": 1.7046596286858392e-06, + "loss": 1.4115, + "step": 23772 + }, + { + "epoch": 4.32, + "learning_rate": 1.7046050236621772e-06, + "loss": 0.0107, + "step": 23773 + }, + { + "epoch": 4.32, + "learning_rate": 1.7045504186385147e-06, + "loss": 0.2065, + "step": 23774 + }, + { + "epoch": 4.32, + "learning_rate": 1.7044958136148526e-06, + "loss": 0.4845, + "step": 23775 + }, + { + "epoch": 4.32, + "learning_rate": 1.7044412085911906e-06, + "loss": 0.7271, + "step": 23776 + }, + { + "epoch": 4.32, + "learning_rate": 1.704386603567528e-06, + "loss": 0.0286, + "step": 23777 + }, + { + "epoch": 4.32, + "learning_rate": 1.704331998543866e-06, + "loss": 0.7152, + "step": 23778 + }, + { + "epoch": 4.32, + "learning_rate": 1.704277393520204e-06, + "loss": 0.0992, + "step": 23779 + }, + { + "epoch": 4.32, + "learning_rate": 1.7042227884965417e-06, + "loss": 0.3378, + "step": 23780 + }, + { + "epoch": 4.32, + "learning_rate": 1.7041681834728797e-06, + "loss": 1.2347, + "step": 23781 + }, + { + "epoch": 4.32, + "learning_rate": 1.7041135784492172e-06, + "loss": 2.8557, + "step": 23782 + }, + { + "epoch": 4.32, + "learning_rate": 1.7040589734255551e-06, + "loss": 0.3427, + "step": 23783 + }, + { + "epoch": 4.33, + "learning_rate": 1.704004368401893e-06, + "loss": 0.4535, + "step": 23784 + }, + { + "epoch": 4.33, + "learning_rate": 1.7039497633782308e-06, + "loss": 3.1764, + "step": 23785 + }, + { + "epoch": 4.33, + "learning_rate": 1.7038951583545687e-06, + "loss": 0.209, + "step": 23786 + }, + { + "epoch": 4.33, + "learning_rate": 1.7038405533309067e-06, + "loss": 2.051, + "step": 23787 + }, + { + "epoch": 4.33, + "learning_rate": 1.7037859483072442e-06, + "loss": 0.2861, + "step": 23788 + }, + { + "epoch": 4.33, + "learning_rate": 1.7037313432835822e-06, + "loss": 0.8933, + "step": 23789 + }, + { + "epoch": 4.33, + "learning_rate": 1.70367673825992e-06, + "loss": 0.4583, + "step": 23790 + }, + { + "epoch": 4.33, + "learning_rate": 1.7036221332362576e-06, + "loss": 0.1626, + "step": 23791 + }, + { + "epoch": 4.33, + "learning_rate": 1.7035675282125956e-06, + "loss": 0.0067, + "step": 23792 + }, + { + "epoch": 4.33, + "learning_rate": 1.7035129231889335e-06, + "loss": 0.8759, + "step": 23793 + }, + { + "epoch": 4.33, + "learning_rate": 1.7034583181652712e-06, + "loss": 1.5602, + "step": 23794 + }, + { + "epoch": 4.33, + "learning_rate": 1.7034037131416092e-06, + "loss": 0.7488, + "step": 23795 + }, + { + "epoch": 4.33, + "learning_rate": 1.7033491081179467e-06, + "loss": 0.9855, + "step": 23796 + }, + { + "epoch": 4.33, + "learning_rate": 1.7032945030942846e-06, + "loss": 0.8467, + "step": 23797 + }, + { + "epoch": 4.33, + "learning_rate": 1.7032398980706226e-06, + "loss": 0.3601, + "step": 23798 + }, + { + "epoch": 4.33, + "learning_rate": 1.7031852930469603e-06, + "loss": 0.2617, + "step": 23799 + }, + { + "epoch": 4.33, + "learning_rate": 1.7031306880232983e-06, + "loss": 0.1422, + "step": 23800 + }, + { + "epoch": 4.33, + "learning_rate": 1.7030760829996362e-06, + "loss": 3.088, + "step": 23801 + }, + { + "epoch": 4.33, + "learning_rate": 1.7030214779759737e-06, + "loss": 0.8759, + "step": 23802 + }, + { + "epoch": 4.33, + "learning_rate": 1.7029668729523117e-06, + "loss": 1.1906, + "step": 23803 + }, + { + "epoch": 4.33, + "learning_rate": 1.7029122679286496e-06, + "loss": 0.1824, + "step": 23804 + }, + { + "epoch": 4.33, + "learning_rate": 1.7028576629049871e-06, + "loss": 0.5979, + "step": 23805 + }, + { + "epoch": 4.33, + "learning_rate": 1.702803057881325e-06, + "loss": 2.5575, + "step": 23806 + }, + { + "epoch": 4.33, + "learning_rate": 1.7027484528576628e-06, + "loss": 1.9278, + "step": 23807 + }, + { + "epoch": 4.33, + "learning_rate": 1.7026938478340008e-06, + "loss": 0.0095, + "step": 23808 + }, + { + "epoch": 4.33, + "learning_rate": 1.7026392428103387e-06, + "loss": 0.9952, + "step": 23809 + }, + { + "epoch": 4.33, + "learning_rate": 1.7025846377866762e-06, + "loss": 0.4354, + "step": 23810 + }, + { + "epoch": 4.33, + "learning_rate": 1.7025300327630142e-06, + "loss": 0.3394, + "step": 23811 + }, + { + "epoch": 4.33, + "learning_rate": 1.702475427739352e-06, + "loss": 0.008, + "step": 23812 + }, + { + "epoch": 4.33, + "learning_rate": 1.7024208227156898e-06, + "loss": 0.2912, + "step": 23813 + }, + { + "epoch": 4.33, + "learning_rate": 1.7023662176920278e-06, + "loss": 1.3608, + "step": 23814 + }, + { + "epoch": 4.33, + "learning_rate": 1.7023116126683657e-06, + "loss": 0.9954, + "step": 23815 + }, + { + "epoch": 4.33, + "learning_rate": 1.7022570076447032e-06, + "loss": 1.1697, + "step": 23816 + }, + { + "epoch": 4.33, + "learning_rate": 1.7022024026210412e-06, + "loss": 0.8491, + "step": 23817 + }, + { + "epoch": 4.33, + "learning_rate": 1.7021477975973791e-06, + "loss": 0.4316, + "step": 23818 + }, + { + "epoch": 4.33, + "learning_rate": 1.7020931925737169e-06, + "loss": 1.385, + "step": 23819 + }, + { + "epoch": 4.33, + "learning_rate": 1.7020385875500548e-06, + "loss": 0.0963, + "step": 23820 + }, + { + "epoch": 4.33, + "learning_rate": 1.7019839825263923e-06, + "loss": 0.6736, + "step": 23821 + }, + { + "epoch": 4.33, + "learning_rate": 1.7019293775027303e-06, + "loss": 0.652, + "step": 23822 + }, + { + "epoch": 4.33, + "learning_rate": 1.7018747724790682e-06, + "loss": 1.0366, + "step": 23823 + }, + { + "epoch": 4.33, + "learning_rate": 1.7018201674554057e-06, + "loss": 1.1916, + "step": 23824 + }, + { + "epoch": 4.33, + "learning_rate": 1.7017655624317437e-06, + "loss": 1.4495, + "step": 23825 + }, + { + "epoch": 4.33, + "learning_rate": 1.7017109574080816e-06, + "loss": 0.9592, + "step": 23826 + }, + { + "epoch": 4.33, + "learning_rate": 1.7016563523844193e-06, + "loss": 0.4319, + "step": 23827 + }, + { + "epoch": 4.33, + "learning_rate": 1.7016017473607573e-06, + "loss": 0.9827, + "step": 23828 + }, + { + "epoch": 4.33, + "learning_rate": 1.7015471423370952e-06, + "loss": 0.8794, + "step": 23829 + }, + { + "epoch": 4.33, + "learning_rate": 1.7014925373134328e-06, + "loss": 1.7579, + "step": 23830 + }, + { + "epoch": 4.33, + "learning_rate": 1.7014379322897707e-06, + "loss": 0.8862, + "step": 23831 + }, + { + "epoch": 4.33, + "learning_rate": 1.7013833272661086e-06, + "loss": 1.1576, + "step": 23832 + }, + { + "epoch": 4.33, + "learning_rate": 1.7013287222424464e-06, + "loss": 0.3652, + "step": 23833 + }, + { + "epoch": 4.33, + "learning_rate": 1.7012741172187843e-06, + "loss": 1.9911, + "step": 23834 + }, + { + "epoch": 4.33, + "learning_rate": 1.7012195121951218e-06, + "loss": 0.4872, + "step": 23835 + }, + { + "epoch": 4.33, + "learning_rate": 1.7011649071714598e-06, + "loss": 1.4096, + "step": 23836 + }, + { + "epoch": 4.33, + "learning_rate": 1.7011103021477977e-06, + "loss": 1.8647, + "step": 23837 + }, + { + "epoch": 4.33, + "learning_rate": 1.7010556971241352e-06, + "loss": 0.7793, + "step": 23838 + }, + { + "epoch": 4.34, + "learning_rate": 1.7010010921004732e-06, + "loss": 0.1697, + "step": 23839 + }, + { + "epoch": 4.34, + "learning_rate": 1.7009464870768111e-06, + "loss": 1.057, + "step": 23840 + }, + { + "epoch": 4.34, + "learning_rate": 1.7008918820531489e-06, + "loss": 1.6835, + "step": 23841 + }, + { + "epoch": 4.34, + "learning_rate": 1.7008372770294868e-06, + "loss": 0.2933, + "step": 23842 + }, + { + "epoch": 4.34, + "learning_rate": 1.7007826720058247e-06, + "loss": 0.4228, + "step": 23843 + }, + { + "epoch": 4.34, + "learning_rate": 1.7007280669821623e-06, + "loss": 0.3175, + "step": 23844 + }, + { + "epoch": 4.34, + "learning_rate": 1.7006734619585002e-06, + "loss": 0.5048, + "step": 23845 + }, + { + "epoch": 4.34, + "learning_rate": 1.7006188569348382e-06, + "loss": 0.6982, + "step": 23846 + }, + { + "epoch": 4.34, + "learning_rate": 1.7005642519111759e-06, + "loss": 0.2219, + "step": 23847 + }, + { + "epoch": 4.34, + "learning_rate": 1.7005096468875138e-06, + "loss": 0.0255, + "step": 23848 + }, + { + "epoch": 4.34, + "learning_rate": 1.7004550418638513e-06, + "loss": 0.2351, + "step": 23849 + }, + { + "epoch": 4.34, + "learning_rate": 1.7004004368401893e-06, + "loss": 0.3067, + "step": 23850 + }, + { + "epoch": 4.34, + "learning_rate": 1.7003458318165272e-06, + "loss": 0.9624, + "step": 23851 + }, + { + "epoch": 4.34, + "learning_rate": 1.700291226792865e-06, + "loss": 0.0327, + "step": 23852 + }, + { + "epoch": 4.34, + "learning_rate": 1.700236621769203e-06, + "loss": 1.9644, + "step": 23853 + }, + { + "epoch": 4.34, + "learning_rate": 1.7001820167455406e-06, + "loss": 1.4259, + "step": 23854 + }, + { + "epoch": 4.34, + "learning_rate": 1.7001274117218784e-06, + "loss": 0.5093, + "step": 23855 + }, + { + "epoch": 4.34, + "learning_rate": 1.7000728066982163e-06, + "loss": 1.1123, + "step": 23856 + }, + { + "epoch": 4.34, + "learning_rate": 1.7000182016745543e-06, + "loss": 0.6891, + "step": 23857 + }, + { + "epoch": 4.34, + "learning_rate": 1.6999635966508918e-06, + "loss": 1.2815, + "step": 23858 + }, + { + "epoch": 4.34, + "learning_rate": 1.6999089916272297e-06, + "loss": 0.0856, + "step": 23859 + }, + { + "epoch": 4.34, + "learning_rate": 1.6998543866035677e-06, + "loss": 0.0211, + "step": 23860 + }, + { + "epoch": 4.34, + "learning_rate": 1.6997997815799054e-06, + "loss": 1.002, + "step": 23861 + }, + { + "epoch": 4.34, + "learning_rate": 1.6997451765562433e-06, + "loss": 1.4735, + "step": 23862 + }, + { + "epoch": 4.34, + "learning_rate": 1.6996905715325809e-06, + "loss": 0.5475, + "step": 23863 + }, + { + "epoch": 4.34, + "learning_rate": 1.6996359665089188e-06, + "loss": 0.0764, + "step": 23864 + }, + { + "epoch": 4.34, + "learning_rate": 1.6995813614852567e-06, + "loss": 0.9415, + "step": 23865 + }, + { + "epoch": 4.34, + "learning_rate": 1.6995267564615945e-06, + "loss": 0.7615, + "step": 23866 + }, + { + "epoch": 4.34, + "learning_rate": 1.6994721514379324e-06, + "loss": 0.1928, + "step": 23867 + }, + { + "epoch": 4.34, + "learning_rate": 1.6994175464142704e-06, + "loss": 0.1603, + "step": 23868 + }, + { + "epoch": 4.34, + "learning_rate": 1.6993629413906079e-06, + "loss": 0.2157, + "step": 23869 + }, + { + "epoch": 4.34, + "learning_rate": 1.6993083363669458e-06, + "loss": 0.0695, + "step": 23870 + }, + { + "epoch": 4.34, + "learning_rate": 1.6992537313432838e-06, + "loss": 0.0091, + "step": 23871 + }, + { + "epoch": 4.34, + "learning_rate": 1.6991991263196213e-06, + "loss": 0.476, + "step": 23872 + }, + { + "epoch": 4.34, + "learning_rate": 1.6991445212959592e-06, + "loss": 1.0383, + "step": 23873 + }, + { + "epoch": 4.34, + "learning_rate": 1.699089916272297e-06, + "loss": 0.6942, + "step": 23874 + }, + { + "epoch": 4.34, + "learning_rate": 1.699035311248635e-06, + "loss": 1.6348, + "step": 23875 + }, + { + "epoch": 4.34, + "learning_rate": 1.6989807062249729e-06, + "loss": 0.0096, + "step": 23876 + }, + { + "epoch": 4.34, + "learning_rate": 1.6989261012013104e-06, + "loss": 1.1298, + "step": 23877 + }, + { + "epoch": 4.34, + "learning_rate": 1.6988714961776483e-06, + "loss": 0.3658, + "step": 23878 + }, + { + "epoch": 4.34, + "learning_rate": 1.6988168911539863e-06, + "loss": 1.2139, + "step": 23879 + }, + { + "epoch": 4.34, + "learning_rate": 1.698762286130324e-06, + "loss": 0.5292, + "step": 23880 + }, + { + "epoch": 4.34, + "learning_rate": 1.698707681106662e-06, + "loss": 0.4695, + "step": 23881 + }, + { + "epoch": 4.34, + "learning_rate": 1.6986530760829999e-06, + "loss": 0.8214, + "step": 23882 + }, + { + "epoch": 4.34, + "learning_rate": 1.6985984710593374e-06, + "loss": 0.3429, + "step": 23883 + }, + { + "epoch": 4.34, + "learning_rate": 1.6985438660356753e-06, + "loss": 0.0679, + "step": 23884 + }, + { + "epoch": 4.34, + "learning_rate": 1.6984892610120133e-06, + "loss": 0.526, + "step": 23885 + }, + { + "epoch": 4.34, + "learning_rate": 1.6984346559883508e-06, + "loss": 0.9248, + "step": 23886 + }, + { + "epoch": 4.34, + "learning_rate": 1.6983800509646887e-06, + "loss": 0.3974, + "step": 23887 + }, + { + "epoch": 4.34, + "learning_rate": 1.6983254459410265e-06, + "loss": 0.6466, + "step": 23888 + }, + { + "epoch": 4.34, + "learning_rate": 1.6982708409173644e-06, + "loss": 0.1879, + "step": 23889 + }, + { + "epoch": 4.34, + "learning_rate": 1.6982162358937024e-06, + "loss": 0.8557, + "step": 23890 + }, + { + "epoch": 4.34, + "learning_rate": 1.6981616308700399e-06, + "loss": 0.187, + "step": 23891 + }, + { + "epoch": 4.34, + "learning_rate": 1.6981070258463778e-06, + "loss": 1.8967, + "step": 23892 + }, + { + "epoch": 4.34, + "learning_rate": 1.6980524208227158e-06, + "loss": 1.1497, + "step": 23893 + }, + { + "epoch": 4.35, + "learning_rate": 1.6979978157990535e-06, + "loss": 0.3568, + "step": 23894 + }, + { + "epoch": 4.35, + "learning_rate": 1.6979432107753914e-06, + "loss": 0.0277, + "step": 23895 + }, + { + "epoch": 4.35, + "learning_rate": 1.6978886057517294e-06, + "loss": 0.4536, + "step": 23896 + }, + { + "epoch": 4.35, + "learning_rate": 1.697834000728067e-06, + "loss": 2.7284, + "step": 23897 + }, + { + "epoch": 4.35, + "learning_rate": 1.6977793957044049e-06, + "loss": 3.2345, + "step": 23898 + }, + { + "epoch": 4.35, + "learning_rate": 1.6977247906807428e-06, + "loss": 0.2409, + "step": 23899 + }, + { + "epoch": 4.35, + "learning_rate": 1.6976701856570805e-06, + "loss": 0.631, + "step": 23900 + }, + { + "epoch": 4.35, + "learning_rate": 1.6976155806334185e-06, + "loss": 0.2461, + "step": 23901 + }, + { + "epoch": 4.35, + "learning_rate": 1.697560975609756e-06, + "loss": 0.0248, + "step": 23902 + }, + { + "epoch": 4.35, + "learning_rate": 1.697506370586094e-06, + "loss": 2.5896, + "step": 23903 + }, + { + "epoch": 4.35, + "learning_rate": 1.6974517655624319e-06, + "loss": 0.0127, + "step": 23904 + }, + { + "epoch": 4.35, + "learning_rate": 1.6973971605387694e-06, + "loss": 0.3439, + "step": 23905 + }, + { + "epoch": 4.35, + "learning_rate": 1.6973425555151073e-06, + "loss": 1.6884, + "step": 23906 + }, + { + "epoch": 4.35, + "learning_rate": 1.6972879504914453e-06, + "loss": 2.066, + "step": 23907 + }, + { + "epoch": 4.35, + "learning_rate": 1.697233345467783e-06, + "loss": 0.4903, + "step": 23908 + }, + { + "epoch": 4.35, + "learning_rate": 1.697178740444121e-06, + "loss": 0.2218, + "step": 23909 + }, + { + "epoch": 4.35, + "learning_rate": 1.697124135420459e-06, + "loss": 0.278, + "step": 23910 + }, + { + "epoch": 4.35, + "learning_rate": 1.6970695303967964e-06, + "loss": 0.1104, + "step": 23911 + }, + { + "epoch": 4.35, + "learning_rate": 1.6970149253731344e-06, + "loss": 0.3998, + "step": 23912 + }, + { + "epoch": 4.35, + "learning_rate": 1.6969603203494723e-06, + "loss": 0.058, + "step": 23913 + }, + { + "epoch": 4.35, + "learning_rate": 1.69690571532581e-06, + "loss": 0.5784, + "step": 23914 + }, + { + "epoch": 4.35, + "learning_rate": 1.696851110302148e-06, + "loss": 0.0117, + "step": 23915 + }, + { + "epoch": 4.35, + "learning_rate": 1.6967965052784855e-06, + "loss": 0.541, + "step": 23916 + }, + { + "epoch": 4.35, + "learning_rate": 1.6967419002548235e-06, + "loss": 0.2479, + "step": 23917 + }, + { + "epoch": 4.35, + "learning_rate": 1.6966872952311614e-06, + "loss": 0.3848, + "step": 23918 + }, + { + "epoch": 4.35, + "learning_rate": 1.696632690207499e-06, + "loss": 0.2564, + "step": 23919 + }, + { + "epoch": 4.35, + "learning_rate": 1.6965780851838369e-06, + "loss": 0.709, + "step": 23920 + }, + { + "epoch": 4.35, + "learning_rate": 1.6965234801601748e-06, + "loss": 0.5924, + "step": 23921 + }, + { + "epoch": 4.35, + "learning_rate": 1.6964688751365125e-06, + "loss": 0.1121, + "step": 23922 + }, + { + "epoch": 4.35, + "learning_rate": 1.6964142701128505e-06, + "loss": 0.827, + "step": 23923 + }, + { + "epoch": 4.35, + "learning_rate": 1.6963596650891884e-06, + "loss": 0.3167, + "step": 23924 + }, + { + "epoch": 4.35, + "learning_rate": 1.696305060065526e-06, + "loss": 1.0774, + "step": 23925 + }, + { + "epoch": 4.35, + "learning_rate": 1.6962504550418639e-06, + "loss": 0.5763, + "step": 23926 + }, + { + "epoch": 4.35, + "learning_rate": 1.6961958500182018e-06, + "loss": 3.0688, + "step": 23927 + }, + { + "epoch": 4.35, + "learning_rate": 1.6961412449945396e-06, + "loss": 0.0071, + "step": 23928 + }, + { + "epoch": 4.35, + "learning_rate": 1.6960866399708775e-06, + "loss": 0.7713, + "step": 23929 + }, + { + "epoch": 4.35, + "learning_rate": 1.696032034947215e-06, + "loss": 0.2172, + "step": 23930 + }, + { + "epoch": 4.35, + "learning_rate": 1.695977429923553e-06, + "loss": 1.4453, + "step": 23931 + }, + { + "epoch": 4.35, + "learning_rate": 1.695922824899891e-06, + "loss": 0.0383, + "step": 23932 + }, + { + "epoch": 4.35, + "learning_rate": 1.6958682198762284e-06, + "loss": 1.5253, + "step": 23933 + }, + { + "epoch": 4.35, + "learning_rate": 1.6958136148525664e-06, + "loss": 0.8061, + "step": 23934 + }, + { + "epoch": 4.35, + "learning_rate": 1.6957590098289043e-06, + "loss": 0.8321, + "step": 23935 + }, + { + "epoch": 4.35, + "learning_rate": 1.695704404805242e-06, + "loss": 0.0092, + "step": 23936 + }, + { + "epoch": 4.35, + "learning_rate": 1.69564979978158e-06, + "loss": 0.1439, + "step": 23937 + }, + { + "epoch": 4.35, + "learning_rate": 1.695595194757918e-06, + "loss": 0.4537, + "step": 23938 + }, + { + "epoch": 4.35, + "learning_rate": 1.6955405897342555e-06, + "loss": 2.2694, + "step": 23939 + }, + { + "epoch": 4.35, + "learning_rate": 1.6954859847105934e-06, + "loss": 0.2652, + "step": 23940 + }, + { + "epoch": 4.35, + "learning_rate": 1.6954313796869313e-06, + "loss": 1.2278, + "step": 23941 + }, + { + "epoch": 4.35, + "learning_rate": 1.695376774663269e-06, + "loss": 0.0084, + "step": 23942 + }, + { + "epoch": 4.35, + "learning_rate": 1.695322169639607e-06, + "loss": 0.0715, + "step": 23943 + }, + { + "epoch": 4.35, + "learning_rate": 1.6952675646159445e-06, + "loss": 0.3691, + "step": 23944 + }, + { + "epoch": 4.35, + "learning_rate": 1.6952129595922825e-06, + "loss": 1.2569, + "step": 23945 + }, + { + "epoch": 4.35, + "learning_rate": 1.6951583545686204e-06, + "loss": 0.3562, + "step": 23946 + }, + { + "epoch": 4.35, + "learning_rate": 1.6951037495449582e-06, + "loss": 0.0935, + "step": 23947 + }, + { + "epoch": 4.35, + "learning_rate": 1.695049144521296e-06, + "loss": 0.0825, + "step": 23948 + }, + { + "epoch": 4.36, + "learning_rate": 1.694994539497634e-06, + "loss": 2.4194, + "step": 23949 + }, + { + "epoch": 4.36, + "learning_rate": 1.6949399344739716e-06, + "loss": 0.1771, + "step": 23950 + }, + { + "epoch": 4.36, + "learning_rate": 1.6948853294503095e-06, + "loss": 1.686, + "step": 23951 + }, + { + "epoch": 4.36, + "learning_rate": 1.6948307244266474e-06, + "loss": 0.0962, + "step": 23952 + }, + { + "epoch": 4.36, + "learning_rate": 1.694776119402985e-06, + "loss": 0.9794, + "step": 23953 + }, + { + "epoch": 4.36, + "learning_rate": 1.694721514379323e-06, + "loss": 1.3288, + "step": 23954 + }, + { + "epoch": 4.36, + "learning_rate": 1.6946669093556606e-06, + "loss": 0.3243, + "step": 23955 + }, + { + "epoch": 4.36, + "learning_rate": 1.6946123043319986e-06, + "loss": 0.4157, + "step": 23956 + }, + { + "epoch": 4.36, + "learning_rate": 1.6945576993083365e-06, + "loss": 0.2095, + "step": 23957 + }, + { + "epoch": 4.36, + "learning_rate": 1.694503094284674e-06, + "loss": 0.3699, + "step": 23958 + }, + { + "epoch": 4.36, + "learning_rate": 1.694448489261012e-06, + "loss": 0.2792, + "step": 23959 + }, + { + "epoch": 4.36, + "learning_rate": 1.69439388423735e-06, + "loss": 0.0304, + "step": 23960 + }, + { + "epoch": 4.36, + "learning_rate": 1.6943392792136877e-06, + "loss": 0.2567, + "step": 23961 + }, + { + "epoch": 4.36, + "learning_rate": 1.6942846741900256e-06, + "loss": 0.761, + "step": 23962 + }, + { + "epoch": 4.36, + "learning_rate": 1.6942300691663636e-06, + "loss": 1.6059, + "step": 23963 + }, + { + "epoch": 4.36, + "learning_rate": 1.694175464142701e-06, + "loss": 0.4924, + "step": 23964 + }, + { + "epoch": 4.36, + "learning_rate": 1.694120859119039e-06, + "loss": 1.4823, + "step": 23965 + }, + { + "epoch": 4.36, + "learning_rate": 1.694066254095377e-06, + "loss": 0.0138, + "step": 23966 + }, + { + "epoch": 4.36, + "learning_rate": 1.6940116490717145e-06, + "loss": 0.0224, + "step": 23967 + }, + { + "epoch": 4.36, + "learning_rate": 1.6939570440480524e-06, + "loss": 1.3849, + "step": 23968 + }, + { + "epoch": 4.36, + "learning_rate": 1.6939024390243902e-06, + "loss": 0.0498, + "step": 23969 + }, + { + "epoch": 4.36, + "learning_rate": 1.693847834000728e-06, + "loss": 1.0123, + "step": 23970 + }, + { + "epoch": 4.36, + "learning_rate": 1.693793228977066e-06, + "loss": 1.0643, + "step": 23971 + }, + { + "epoch": 4.36, + "learning_rate": 1.6937386239534036e-06, + "loss": 0.9456, + "step": 23972 + }, + { + "epoch": 4.36, + "learning_rate": 1.6936840189297415e-06, + "loss": 0.0178, + "step": 23973 + }, + { + "epoch": 4.36, + "learning_rate": 1.6936294139060794e-06, + "loss": 0.2241, + "step": 23974 + }, + { + "epoch": 4.36, + "learning_rate": 1.6935748088824172e-06, + "loss": 0.5286, + "step": 23975 + }, + { + "epoch": 4.36, + "learning_rate": 1.6935202038587551e-06, + "loss": 1.046, + "step": 23976 + }, + { + "epoch": 4.36, + "learning_rate": 1.693465598835093e-06, + "loss": 0.2906, + "step": 23977 + }, + { + "epoch": 4.36, + "learning_rate": 1.6934109938114306e-06, + "loss": 0.8498, + "step": 23978 + }, + { + "epoch": 4.36, + "learning_rate": 1.6933563887877685e-06, + "loss": 0.102, + "step": 23979 + }, + { + "epoch": 4.36, + "learning_rate": 1.6933017837641065e-06, + "loss": 0.0271, + "step": 23980 + }, + { + "epoch": 4.36, + "learning_rate": 1.6932471787404442e-06, + "loss": 1.8624, + "step": 23981 + }, + { + "epoch": 4.36, + "learning_rate": 1.6931925737167821e-06, + "loss": 0.9208, + "step": 23982 + }, + { + "epoch": 4.36, + "learning_rate": 1.6931379686931197e-06, + "loss": 0.5108, + "step": 23983 + }, + { + "epoch": 4.36, + "learning_rate": 1.6930833636694576e-06, + "loss": 0.3562, + "step": 23984 + }, + { + "epoch": 4.36, + "learning_rate": 1.6930287586457956e-06, + "loss": 0.6528, + "step": 23985 + }, + { + "epoch": 4.36, + "learning_rate": 1.692974153622133e-06, + "loss": 1.7299, + "step": 23986 + }, + { + "epoch": 4.36, + "learning_rate": 1.692919548598471e-06, + "loss": 2.0612, + "step": 23987 + }, + { + "epoch": 4.36, + "learning_rate": 1.692864943574809e-06, + "loss": 1.4182, + "step": 23988 + }, + { + "epoch": 4.36, + "learning_rate": 1.6928103385511467e-06, + "loss": 0.185, + "step": 23989 + }, + { + "epoch": 4.36, + "learning_rate": 1.6927557335274846e-06, + "loss": 1.3527, + "step": 23990 + }, + { + "epoch": 4.36, + "learning_rate": 1.6927011285038226e-06, + "loss": 0.6619, + "step": 23991 + }, + { + "epoch": 4.36, + "learning_rate": 1.69264652348016e-06, + "loss": 0.229, + "step": 23992 + }, + { + "epoch": 4.36, + "learning_rate": 1.692591918456498e-06, + "loss": 0.1375, + "step": 23993 + }, + { + "epoch": 4.36, + "learning_rate": 1.692537313432836e-06, + "loss": 0.9145, + "step": 23994 + }, + { + "epoch": 4.36, + "learning_rate": 1.6924827084091737e-06, + "loss": 0.1833, + "step": 23995 + }, + { + "epoch": 4.36, + "learning_rate": 1.6924281033855117e-06, + "loss": 0.5318, + "step": 23996 + }, + { + "epoch": 4.36, + "learning_rate": 1.6923734983618492e-06, + "loss": 3.1355, + "step": 23997 + }, + { + "epoch": 4.36, + "learning_rate": 1.6923188933381871e-06, + "loss": 0.0108, + "step": 23998 + }, + { + "epoch": 4.36, + "learning_rate": 1.692264288314525e-06, + "loss": 0.0195, + "step": 23999 + }, + { + "epoch": 4.36, + "learning_rate": 1.6922096832908626e-06, + "loss": 0.4673, + "step": 24000 + }, + { + "epoch": 4.36, + "learning_rate": 1.6921550782672005e-06, + "loss": 1.7049, + "step": 24001 + }, + { + "epoch": 4.36, + "learning_rate": 1.6921004732435385e-06, + "loss": 0.0203, + "step": 24002 + }, + { + "epoch": 4.36, + "learning_rate": 1.6920458682198762e-06, + "loss": 0.4687, + "step": 24003 + }, + { + "epoch": 4.37, + "learning_rate": 1.6919912631962142e-06, + "loss": 0.1055, + "step": 24004 + }, + { + "epoch": 4.37, + "learning_rate": 1.691936658172552e-06, + "loss": 0.0109, + "step": 24005 + }, + { + "epoch": 4.37, + "learning_rate": 1.6918820531488896e-06, + "loss": 0.1649, + "step": 24006 + }, + { + "epoch": 4.37, + "learning_rate": 1.6918274481252276e-06, + "loss": 0.0653, + "step": 24007 + }, + { + "epoch": 4.37, + "learning_rate": 1.6917728431015655e-06, + "loss": 0.3775, + "step": 24008 + }, + { + "epoch": 4.37, + "learning_rate": 1.6917182380779032e-06, + "loss": 2.8382, + "step": 24009 + }, + { + "epoch": 4.37, + "learning_rate": 1.6916636330542412e-06, + "loss": 0.0668, + "step": 24010 + }, + { + "epoch": 4.37, + "learning_rate": 1.6916090280305787e-06, + "loss": 0.0072, + "step": 24011 + }, + { + "epoch": 4.37, + "learning_rate": 1.6915544230069166e-06, + "loss": 0.8794, + "step": 24012 + }, + { + "epoch": 4.37, + "learning_rate": 1.6914998179832546e-06, + "loss": 1.8815, + "step": 24013 + }, + { + "epoch": 4.37, + "learning_rate": 1.691445212959592e-06, + "loss": 0.0403, + "step": 24014 + }, + { + "epoch": 4.37, + "learning_rate": 1.69139060793593e-06, + "loss": 0.9129, + "step": 24015 + }, + { + "epoch": 4.37, + "learning_rate": 1.691336002912268e-06, + "loss": 2.5885, + "step": 24016 + }, + { + "epoch": 4.37, + "learning_rate": 1.6912813978886057e-06, + "loss": 1.1257, + "step": 24017 + }, + { + "epoch": 4.37, + "learning_rate": 1.6912267928649437e-06, + "loss": 0.0965, + "step": 24018 + }, + { + "epoch": 4.37, + "learning_rate": 1.6911721878412816e-06, + "loss": 0.009, + "step": 24019 + }, + { + "epoch": 4.37, + "learning_rate": 1.6911175828176191e-06, + "loss": 1.1536, + "step": 24020 + }, + { + "epoch": 4.37, + "learning_rate": 1.691062977793957e-06, + "loss": 1.1057, + "step": 24021 + }, + { + "epoch": 4.37, + "learning_rate": 1.691008372770295e-06, + "loss": 0.5344, + "step": 24022 + }, + { + "epoch": 4.37, + "learning_rate": 1.6909537677466327e-06, + "loss": 2.0354, + "step": 24023 + }, + { + "epoch": 4.37, + "learning_rate": 1.6908991627229707e-06, + "loss": 0.0084, + "step": 24024 + }, + { + "epoch": 4.37, + "learning_rate": 1.6908445576993082e-06, + "loss": 0.1982, + "step": 24025 + }, + { + "epoch": 4.37, + "learning_rate": 1.6907899526756462e-06, + "loss": 2.2587, + "step": 24026 + }, + { + "epoch": 4.37, + "learning_rate": 1.690735347651984e-06, + "loss": 0.1692, + "step": 24027 + }, + { + "epoch": 4.37, + "learning_rate": 1.6906807426283218e-06, + "loss": 0.0191, + "step": 24028 + }, + { + "epoch": 4.37, + "learning_rate": 1.6906261376046598e-06, + "loss": 0.0073, + "step": 24029 + }, + { + "epoch": 4.37, + "learning_rate": 1.6905715325809977e-06, + "loss": 0.233, + "step": 24030 + }, + { + "epoch": 4.37, + "learning_rate": 1.6905169275573352e-06, + "loss": 0.3454, + "step": 24031 + }, + { + "epoch": 4.37, + "learning_rate": 1.6904623225336732e-06, + "loss": 1.382, + "step": 24032 + }, + { + "epoch": 4.37, + "learning_rate": 1.6904077175100111e-06, + "loss": 0.2679, + "step": 24033 + }, + { + "epoch": 4.37, + "learning_rate": 1.6903531124863486e-06, + "loss": 0.2868, + "step": 24034 + }, + { + "epoch": 4.37, + "learning_rate": 1.6902985074626866e-06, + "loss": 1.3381, + "step": 24035 + }, + { + "epoch": 4.37, + "learning_rate": 1.6902439024390243e-06, + "loss": 1.2219, + "step": 24036 + }, + { + "epoch": 4.37, + "learning_rate": 1.6901892974153623e-06, + "loss": 1.8938, + "step": 24037 + }, + { + "epoch": 4.37, + "learning_rate": 1.6901346923917002e-06, + "loss": 0.6215, + "step": 24038 + }, + { + "epoch": 4.37, + "learning_rate": 1.6900800873680377e-06, + "loss": 1.6701, + "step": 24039 + }, + { + "epoch": 4.37, + "learning_rate": 1.6900254823443757e-06, + "loss": 0.015, + "step": 24040 + }, + { + "epoch": 4.37, + "learning_rate": 1.6899708773207136e-06, + "loss": 1.7166, + "step": 24041 + }, + { + "epoch": 4.37, + "learning_rate": 1.6899162722970513e-06, + "loss": 1.3602, + "step": 24042 + }, + { + "epoch": 4.37, + "learning_rate": 1.6898616672733893e-06, + "loss": 4.6885, + "step": 24043 + }, + { + "epoch": 4.37, + "learning_rate": 1.6898070622497272e-06, + "loss": 0.742, + "step": 24044 + }, + { + "epoch": 4.37, + "learning_rate": 1.6897524572260647e-06, + "loss": 0.333, + "step": 24045 + }, + { + "epoch": 4.37, + "learning_rate": 1.6896978522024027e-06, + "loss": 0.4868, + "step": 24046 + }, + { + "epoch": 4.37, + "learning_rate": 1.6896432471787406e-06, + "loss": 0.6896, + "step": 24047 + }, + { + "epoch": 4.37, + "learning_rate": 1.6895886421550782e-06, + "loss": 0.6446, + "step": 24048 + }, + { + "epoch": 4.37, + "learning_rate": 1.689534037131416e-06, + "loss": 1.3328, + "step": 24049 + }, + { + "epoch": 4.37, + "learning_rate": 1.6894794321077538e-06, + "loss": 0.3093, + "step": 24050 + }, + { + "epoch": 4.37, + "learning_rate": 1.6894248270840918e-06, + "loss": 1.6398, + "step": 24051 + }, + { + "epoch": 4.37, + "learning_rate": 1.6893702220604297e-06, + "loss": 0.0176, + "step": 24052 + }, + { + "epoch": 4.37, + "learning_rate": 1.6893156170367672e-06, + "loss": 0.013, + "step": 24053 + }, + { + "epoch": 4.37, + "learning_rate": 1.6892610120131052e-06, + "loss": 0.0192, + "step": 24054 + }, + { + "epoch": 4.37, + "learning_rate": 1.6892064069894431e-06, + "loss": 0.0103, + "step": 24055 + }, + { + "epoch": 4.37, + "learning_rate": 1.6891518019657809e-06, + "loss": 1.1383, + "step": 24056 + }, + { + "epoch": 4.37, + "learning_rate": 1.6890971969421188e-06, + "loss": 1.0829, + "step": 24057 + }, + { + "epoch": 4.37, + "learning_rate": 1.6890425919184567e-06, + "loss": 1.4383, + "step": 24058 + }, + { + "epoch": 4.38, + "learning_rate": 1.6889879868947943e-06, + "loss": 0.4061, + "step": 24059 + }, + { + "epoch": 4.38, + "learning_rate": 1.6889333818711322e-06, + "loss": 1.7543, + "step": 24060 + }, + { + "epoch": 4.38, + "learning_rate": 1.6888787768474701e-06, + "loss": 0.3737, + "step": 24061 + }, + { + "epoch": 4.38, + "learning_rate": 1.6888241718238077e-06, + "loss": 1.8446, + "step": 24062 + }, + { + "epoch": 4.38, + "learning_rate": 1.6887695668001456e-06, + "loss": 0.3644, + "step": 24063 + }, + { + "epoch": 4.38, + "learning_rate": 1.6887149617764833e-06, + "loss": 0.7275, + "step": 24064 + }, + { + "epoch": 4.38, + "learning_rate": 1.6886603567528213e-06, + "loss": 1.6956, + "step": 24065 + }, + { + "epoch": 4.38, + "learning_rate": 1.6886057517291592e-06, + "loss": 0.3215, + "step": 24066 + }, + { + "epoch": 4.38, + "learning_rate": 1.6885511467054967e-06, + "loss": 0.0757, + "step": 24067 + }, + { + "epoch": 4.38, + "learning_rate": 1.6884965416818347e-06, + "loss": 0.6344, + "step": 24068 + }, + { + "epoch": 4.38, + "learning_rate": 1.6884419366581726e-06, + "loss": 0.4901, + "step": 24069 + }, + { + "epoch": 4.38, + "learning_rate": 1.6883873316345104e-06, + "loss": 1.3132, + "step": 24070 + }, + { + "epoch": 4.38, + "learning_rate": 1.6883327266108483e-06, + "loss": 1.0907, + "step": 24071 + }, + { + "epoch": 4.38, + "learning_rate": 1.6882781215871863e-06, + "loss": 0.9713, + "step": 24072 + }, + { + "epoch": 4.38, + "learning_rate": 1.6882235165635238e-06, + "loss": 0.2819, + "step": 24073 + }, + { + "epoch": 4.38, + "learning_rate": 1.6881689115398617e-06, + "loss": 0.123, + "step": 24074 + }, + { + "epoch": 4.38, + "learning_rate": 1.6881143065161997e-06, + "loss": 0.2545, + "step": 24075 + }, + { + "epoch": 4.38, + "learning_rate": 1.6880597014925374e-06, + "loss": 2.807, + "step": 24076 + }, + { + "epoch": 4.38, + "learning_rate": 1.6880050964688753e-06, + "loss": 0.6585, + "step": 24077 + }, + { + "epoch": 4.38, + "learning_rate": 1.6879504914452129e-06, + "loss": 1.3262, + "step": 24078 + }, + { + "epoch": 4.38, + "learning_rate": 1.6878958864215508e-06, + "loss": 0.0154, + "step": 24079 + }, + { + "epoch": 4.38, + "learning_rate": 1.6878412813978887e-06, + "loss": 0.3439, + "step": 24080 + }, + { + "epoch": 4.38, + "learning_rate": 1.6877866763742263e-06, + "loss": 0.2593, + "step": 24081 + }, + { + "epoch": 4.38, + "learning_rate": 1.6877320713505642e-06, + "loss": 1.0137, + "step": 24082 + }, + { + "epoch": 4.38, + "learning_rate": 1.6876774663269022e-06, + "loss": 0.5039, + "step": 24083 + }, + { + "epoch": 4.38, + "learning_rate": 1.6876228613032399e-06, + "loss": 0.1609, + "step": 24084 + }, + { + "epoch": 4.38, + "learning_rate": 1.6875682562795778e-06, + "loss": 1.1175, + "step": 24085 + }, + { + "epoch": 4.38, + "learning_rate": 1.6875136512559158e-06, + "loss": 0.2342, + "step": 24086 + }, + { + "epoch": 4.38, + "learning_rate": 1.6874590462322533e-06, + "loss": 0.7895, + "step": 24087 + }, + { + "epoch": 4.38, + "learning_rate": 1.6874044412085912e-06, + "loss": 0.6503, + "step": 24088 + }, + { + "epoch": 4.38, + "learning_rate": 1.6873498361849292e-06, + "loss": 0.1245, + "step": 24089 + }, + { + "epoch": 4.38, + "learning_rate": 1.687295231161267e-06, + "loss": 0.8226, + "step": 24090 + }, + { + "epoch": 4.38, + "learning_rate": 1.6872406261376049e-06, + "loss": 0.4206, + "step": 24091 + }, + { + "epoch": 4.38, + "learning_rate": 1.6871860211139424e-06, + "loss": 0.0371, + "step": 24092 + }, + { + "epoch": 4.38, + "learning_rate": 1.6871314160902803e-06, + "loss": 1.0271, + "step": 24093 + }, + { + "epoch": 4.38, + "learning_rate": 1.6870768110666183e-06, + "loss": 0.2952, + "step": 24094 + }, + { + "epoch": 4.38, + "learning_rate": 1.6870222060429558e-06, + "loss": 4.0894, + "step": 24095 + }, + { + "epoch": 4.38, + "learning_rate": 1.6869676010192937e-06, + "loss": 0.1538, + "step": 24096 + }, + { + "epoch": 4.38, + "learning_rate": 1.6869129959956317e-06, + "loss": 0.2492, + "step": 24097 + }, + { + "epoch": 4.38, + "learning_rate": 1.6868583909719694e-06, + "loss": 1.5518, + "step": 24098 + }, + { + "epoch": 4.38, + "learning_rate": 1.6868037859483073e-06, + "loss": 1.0073, + "step": 24099 + }, + { + "epoch": 4.38, + "learning_rate": 1.6867491809246453e-06, + "loss": 0.3112, + "step": 24100 + }, + { + "epoch": 4.38, + "learning_rate": 1.6866945759009828e-06, + "loss": 0.4762, + "step": 24101 + }, + { + "epoch": 4.38, + "learning_rate": 1.6866399708773207e-06, + "loss": 0.0242, + "step": 24102 + }, + { + "epoch": 4.38, + "learning_rate": 1.6865853658536587e-06, + "loss": 0.5533, + "step": 24103 + }, + { + "epoch": 4.38, + "learning_rate": 1.6865307608299964e-06, + "loss": 1.6036, + "step": 24104 + }, + { + "epoch": 4.38, + "learning_rate": 1.6864761558063344e-06, + "loss": 0.4564, + "step": 24105 + }, + { + "epoch": 4.38, + "learning_rate": 1.6864215507826719e-06, + "loss": 0.5189, + "step": 24106 + }, + { + "epoch": 4.38, + "learning_rate": 1.6863669457590098e-06, + "loss": 0.0693, + "step": 24107 + }, + { + "epoch": 4.38, + "learning_rate": 1.6863123407353478e-06, + "loss": 0.9696, + "step": 24108 + }, + { + "epoch": 4.38, + "learning_rate": 1.6862577357116855e-06, + "loss": 0.3057, + "step": 24109 + }, + { + "epoch": 4.38, + "learning_rate": 1.6862031306880234e-06, + "loss": 0.26, + "step": 24110 + }, + { + "epoch": 4.38, + "learning_rate": 1.6861485256643614e-06, + "loss": 2.5066, + "step": 24111 + }, + { + "epoch": 4.38, + "learning_rate": 1.686093920640699e-06, + "loss": 2.1644, + "step": 24112 + }, + { + "epoch": 4.38, + "learning_rate": 1.6860393156170369e-06, + "loss": 1.6447, + "step": 24113 + }, + { + "epoch": 4.39, + "learning_rate": 1.6859847105933748e-06, + "loss": 2.2075, + "step": 24114 + }, + { + "epoch": 4.39, + "learning_rate": 1.6859301055697123e-06, + "loss": 0.4473, + "step": 24115 + }, + { + "epoch": 4.39, + "learning_rate": 1.6858755005460503e-06, + "loss": 0.0821, + "step": 24116 + }, + { + "epoch": 4.39, + "learning_rate": 1.685820895522388e-06, + "loss": 1.5894, + "step": 24117 + }, + { + "epoch": 4.39, + "learning_rate": 1.685766290498726e-06, + "loss": 0.3563, + "step": 24118 + }, + { + "epoch": 4.39, + "learning_rate": 1.6857116854750639e-06, + "loss": 1.4446, + "step": 24119 + }, + { + "epoch": 4.39, + "learning_rate": 1.6856570804514014e-06, + "loss": 0.743, + "step": 24120 + }, + { + "epoch": 4.39, + "learning_rate": 1.6856024754277393e-06, + "loss": 0.5682, + "step": 24121 + }, + { + "epoch": 4.39, + "learning_rate": 1.6855478704040773e-06, + "loss": 0.2106, + "step": 24122 + }, + { + "epoch": 4.39, + "learning_rate": 1.685493265380415e-06, + "loss": 0.9981, + "step": 24123 + }, + { + "epoch": 4.39, + "learning_rate": 1.685438660356753e-06, + "loss": 1.4939, + "step": 24124 + }, + { + "epoch": 4.39, + "learning_rate": 1.685384055333091e-06, + "loss": 0.2748, + "step": 24125 + }, + { + "epoch": 4.39, + "learning_rate": 1.6853294503094284e-06, + "loss": 0.7021, + "step": 24126 + }, + { + "epoch": 4.39, + "learning_rate": 1.6852748452857664e-06, + "loss": 0.2531, + "step": 24127 + }, + { + "epoch": 4.39, + "learning_rate": 1.6852202402621043e-06, + "loss": 0.4246, + "step": 24128 + }, + { + "epoch": 4.39, + "learning_rate": 1.6851656352384418e-06, + "loss": 1.2166, + "step": 24129 + }, + { + "epoch": 4.39, + "learning_rate": 1.6851110302147798e-06, + "loss": 0.3558, + "step": 24130 + }, + { + "epoch": 4.39, + "learning_rate": 1.6850564251911175e-06, + "loss": 0.03, + "step": 24131 + }, + { + "epoch": 4.39, + "learning_rate": 1.6850018201674554e-06, + "loss": 1.4871, + "step": 24132 + }, + { + "epoch": 4.39, + "learning_rate": 1.6849472151437934e-06, + "loss": 0.1609, + "step": 24133 + }, + { + "epoch": 4.39, + "learning_rate": 1.684892610120131e-06, + "loss": 0.4905, + "step": 24134 + }, + { + "epoch": 4.39, + "learning_rate": 1.6848380050964689e-06, + "loss": 0.9393, + "step": 24135 + }, + { + "epoch": 4.39, + "learning_rate": 1.6847834000728068e-06, + "loss": 1.3876, + "step": 24136 + }, + { + "epoch": 4.39, + "learning_rate": 1.6847287950491445e-06, + "loss": 0.0083, + "step": 24137 + }, + { + "epoch": 4.39, + "learning_rate": 1.6846741900254825e-06, + "loss": 0.0129, + "step": 24138 + }, + { + "epoch": 4.39, + "learning_rate": 1.6846195850018204e-06, + "loss": 0.0253, + "step": 24139 + }, + { + "epoch": 4.39, + "learning_rate": 1.684564979978158e-06, + "loss": 0.3623, + "step": 24140 + }, + { + "epoch": 4.39, + "learning_rate": 1.6845103749544959e-06, + "loss": 0.5027, + "step": 24141 + }, + { + "epoch": 4.39, + "learning_rate": 1.6844557699308338e-06, + "loss": 0.6424, + "step": 24142 + }, + { + "epoch": 4.39, + "learning_rate": 1.6844011649071713e-06, + "loss": 0.5355, + "step": 24143 + }, + { + "epoch": 4.39, + "learning_rate": 1.6843465598835093e-06, + "loss": 0.8676, + "step": 24144 + }, + { + "epoch": 4.39, + "learning_rate": 1.684291954859847e-06, + "loss": 0.1891, + "step": 24145 + }, + { + "epoch": 4.39, + "learning_rate": 1.684237349836185e-06, + "loss": 0.2309, + "step": 24146 + }, + { + "epoch": 4.39, + "learning_rate": 1.684182744812523e-06, + "loss": 0.1709, + "step": 24147 + }, + { + "epoch": 4.39, + "learning_rate": 1.6841281397888604e-06, + "loss": 0.2028, + "step": 24148 + }, + { + "epoch": 4.39, + "learning_rate": 1.6840735347651984e-06, + "loss": 1.3522, + "step": 24149 + }, + { + "epoch": 4.39, + "learning_rate": 1.6840189297415363e-06, + "loss": 0.2786, + "step": 24150 + }, + { + "epoch": 4.39, + "learning_rate": 1.683964324717874e-06, + "loss": 2.7685, + "step": 24151 + }, + { + "epoch": 4.39, + "learning_rate": 1.683909719694212e-06, + "loss": 0.4654, + "step": 24152 + }, + { + "epoch": 4.39, + "learning_rate": 1.68385511467055e-06, + "loss": 0.9355, + "step": 24153 + }, + { + "epoch": 4.39, + "learning_rate": 1.6838005096468874e-06, + "loss": 0.7178, + "step": 24154 + }, + { + "epoch": 4.39, + "learning_rate": 1.6837459046232254e-06, + "loss": 0.4641, + "step": 24155 + }, + { + "epoch": 4.39, + "learning_rate": 1.6836912995995633e-06, + "loss": 1.419, + "step": 24156 + }, + { + "epoch": 4.39, + "learning_rate": 1.683636694575901e-06, + "loss": 2.1715, + "step": 24157 + }, + { + "epoch": 4.39, + "learning_rate": 1.683582089552239e-06, + "loss": 0.363, + "step": 24158 + }, + { + "epoch": 4.39, + "learning_rate": 1.6835274845285765e-06, + "loss": 0.0062, + "step": 24159 + }, + { + "epoch": 4.39, + "learning_rate": 1.6834728795049145e-06, + "loss": 0.8026, + "step": 24160 + }, + { + "epoch": 4.39, + "learning_rate": 1.6834182744812524e-06, + "loss": 1.3707, + "step": 24161 + }, + { + "epoch": 4.39, + "learning_rate": 1.68336366945759e-06, + "loss": 1.516, + "step": 24162 + }, + { + "epoch": 4.39, + "learning_rate": 1.6833090644339279e-06, + "loss": 0.3969, + "step": 24163 + }, + { + "epoch": 4.39, + "learning_rate": 1.6832544594102658e-06, + "loss": 0.5824, + "step": 24164 + }, + { + "epoch": 4.39, + "learning_rate": 1.6831998543866036e-06, + "loss": 0.0089, + "step": 24165 + }, + { + "epoch": 4.39, + "learning_rate": 1.6831452493629415e-06, + "loss": 0.7199, + "step": 24166 + }, + { + "epoch": 4.39, + "learning_rate": 1.6830906443392794e-06, + "loss": 0.0092, + "step": 24167 + }, + { + "epoch": 4.39, + "learning_rate": 1.683036039315617e-06, + "loss": 0.2885, + "step": 24168 + }, + { + "epoch": 4.4, + "learning_rate": 1.682981434291955e-06, + "loss": 0.0455, + "step": 24169 + }, + { + "epoch": 4.4, + "learning_rate": 1.6829268292682928e-06, + "loss": 0.8664, + "step": 24170 + }, + { + "epoch": 4.4, + "learning_rate": 1.6828722242446306e-06, + "loss": 0.2087, + "step": 24171 + }, + { + "epoch": 4.4, + "learning_rate": 1.6828176192209685e-06, + "loss": 0.4142, + "step": 24172 + }, + { + "epoch": 4.4, + "learning_rate": 1.682763014197306e-06, + "loss": 2.1272, + "step": 24173 + }, + { + "epoch": 4.4, + "learning_rate": 1.682708409173644e-06, + "loss": 0.0841, + "step": 24174 + }, + { + "epoch": 4.4, + "learning_rate": 1.682653804149982e-06, + "loss": 0.0933, + "step": 24175 + }, + { + "epoch": 4.4, + "learning_rate": 1.6825991991263195e-06, + "loss": 0.0223, + "step": 24176 + }, + { + "epoch": 4.4, + "learning_rate": 1.6825445941026574e-06, + "loss": 0.8312, + "step": 24177 + }, + { + "epoch": 4.4, + "learning_rate": 1.6824899890789953e-06, + "loss": 0.9507, + "step": 24178 + }, + { + "epoch": 4.4, + "learning_rate": 1.682435384055333e-06, + "loss": 0.6574, + "step": 24179 + }, + { + "epoch": 4.4, + "learning_rate": 1.682380779031671e-06, + "loss": 0.3069, + "step": 24180 + }, + { + "epoch": 4.4, + "learning_rate": 1.682326174008009e-06, + "loss": 0.5383, + "step": 24181 + }, + { + "epoch": 4.4, + "learning_rate": 1.6822715689843465e-06, + "loss": 0.0945, + "step": 24182 + }, + { + "epoch": 4.4, + "learning_rate": 1.6822169639606844e-06, + "loss": 0.439, + "step": 24183 + }, + { + "epoch": 4.4, + "learning_rate": 1.6821623589370222e-06, + "loss": 1.0602, + "step": 24184 + }, + { + "epoch": 4.4, + "learning_rate": 1.68210775391336e-06, + "loss": 0.0883, + "step": 24185 + }, + { + "epoch": 4.4, + "learning_rate": 1.682053148889698e-06, + "loss": 1.3014, + "step": 24186 + }, + { + "epoch": 4.4, + "learning_rate": 1.6819985438660356e-06, + "loss": 0.1955, + "step": 24187 + }, + { + "epoch": 4.4, + "learning_rate": 1.6819439388423735e-06, + "loss": 1.5889, + "step": 24188 + }, + { + "epoch": 4.4, + "learning_rate": 1.6818893338187114e-06, + "loss": 1.972, + "step": 24189 + }, + { + "epoch": 4.4, + "learning_rate": 1.681834728795049e-06, + "loss": 1.7076, + "step": 24190 + }, + { + "epoch": 4.4, + "learning_rate": 1.681780123771387e-06, + "loss": 0.5984, + "step": 24191 + }, + { + "epoch": 4.4, + "learning_rate": 1.6817255187477249e-06, + "loss": 0.2123, + "step": 24192 + }, + { + "epoch": 4.4, + "learning_rate": 1.6816709137240626e-06, + "loss": 1.8539, + "step": 24193 + }, + { + "epoch": 4.4, + "learning_rate": 1.6816163087004005e-06, + "loss": 0.1739, + "step": 24194 + }, + { + "epoch": 4.4, + "learning_rate": 1.6815617036767385e-06, + "loss": 0.0179, + "step": 24195 + }, + { + "epoch": 4.4, + "learning_rate": 1.681507098653076e-06, + "loss": 0.8017, + "step": 24196 + }, + { + "epoch": 4.4, + "learning_rate": 1.681452493629414e-06, + "loss": 0.7968, + "step": 24197 + }, + { + "epoch": 4.4, + "learning_rate": 1.6813978886057517e-06, + "loss": 1.2905, + "step": 24198 + }, + { + "epoch": 4.4, + "learning_rate": 1.6813432835820896e-06, + "loss": 0.5936, + "step": 24199 + }, + { + "epoch": 4.4, + "learning_rate": 1.6812886785584276e-06, + "loss": 0.4639, + "step": 24200 + }, + { + "epoch": 4.4, + "learning_rate": 1.681234073534765e-06, + "loss": 0.0088, + "step": 24201 + }, + { + "epoch": 4.4, + "learning_rate": 1.681179468511103e-06, + "loss": 0.0154, + "step": 24202 + }, + { + "epoch": 4.4, + "learning_rate": 1.681124863487441e-06, + "loss": 0.743, + "step": 24203 + }, + { + "epoch": 4.4, + "learning_rate": 1.6810702584637787e-06, + "loss": 0.9088, + "step": 24204 + }, + { + "epoch": 4.4, + "learning_rate": 1.6810156534401166e-06, + "loss": 0.2408, + "step": 24205 + }, + { + "epoch": 4.4, + "learning_rate": 1.6809610484164546e-06, + "loss": 1.1588, + "step": 24206 + }, + { + "epoch": 4.4, + "learning_rate": 1.680906443392792e-06, + "loss": 0.1326, + "step": 24207 + }, + { + "epoch": 4.4, + "learning_rate": 1.68085183836913e-06, + "loss": 0.0097, + "step": 24208 + }, + { + "epoch": 4.4, + "learning_rate": 1.680797233345468e-06, + "loss": 0.7562, + "step": 24209 + }, + { + "epoch": 4.4, + "learning_rate": 1.6807426283218055e-06, + "loss": 0.8389, + "step": 24210 + }, + { + "epoch": 4.4, + "learning_rate": 1.6806880232981434e-06, + "loss": 0.6729, + "step": 24211 + }, + { + "epoch": 4.4, + "learning_rate": 1.6806334182744812e-06, + "loss": 0.0869, + "step": 24212 + }, + { + "epoch": 4.4, + "learning_rate": 1.6805788132508191e-06, + "loss": 1.4611, + "step": 24213 + }, + { + "epoch": 4.4, + "learning_rate": 1.680524208227157e-06, + "loss": 4.7752, + "step": 24214 + }, + { + "epoch": 4.4, + "learning_rate": 1.6804696032034946e-06, + "loss": 2.5287, + "step": 24215 + }, + { + "epoch": 4.4, + "learning_rate": 1.6804149981798325e-06, + "loss": 0.7553, + "step": 24216 + }, + { + "epoch": 4.4, + "learning_rate": 1.6803603931561705e-06, + "loss": 0.133, + "step": 24217 + }, + { + "epoch": 4.4, + "learning_rate": 1.6803057881325082e-06, + "loss": 0.1511, + "step": 24218 + }, + { + "epoch": 4.4, + "learning_rate": 1.6802511831088461e-06, + "loss": 0.6218, + "step": 24219 + }, + { + "epoch": 4.4, + "learning_rate": 1.680196578085184e-06, + "loss": 0.1659, + "step": 24220 + }, + { + "epoch": 4.4, + "learning_rate": 1.6801419730615216e-06, + "loss": 0.3714, + "step": 24221 + }, + { + "epoch": 4.4, + "learning_rate": 1.6800873680378596e-06, + "loss": 0.398, + "step": 24222 + }, + { + "epoch": 4.4, + "learning_rate": 1.6800327630141975e-06, + "loss": 0.0185, + "step": 24223 + }, + { + "epoch": 4.41, + "learning_rate": 1.679978157990535e-06, + "loss": 0.9706, + "step": 24224 + }, + { + "epoch": 4.41, + "learning_rate": 1.679923552966873e-06, + "loss": 3.1735, + "step": 24225 + }, + { + "epoch": 4.41, + "learning_rate": 1.6798689479432107e-06, + "loss": 0.0119, + "step": 24226 + }, + { + "epoch": 4.41, + "learning_rate": 1.6798143429195486e-06, + "loss": 0.9146, + "step": 24227 + }, + { + "epoch": 4.41, + "learning_rate": 1.6797597378958866e-06, + "loss": 0.0452, + "step": 24228 + }, + { + "epoch": 4.41, + "learning_rate": 1.679705132872224e-06, + "loss": 1.245, + "step": 24229 + }, + { + "epoch": 4.41, + "learning_rate": 1.679650527848562e-06, + "loss": 0.0045, + "step": 24230 + }, + { + "epoch": 4.41, + "learning_rate": 1.6795959228249e-06, + "loss": 0.1038, + "step": 24231 + }, + { + "epoch": 4.41, + "learning_rate": 1.6795413178012377e-06, + "loss": 1.7208, + "step": 24232 + }, + { + "epoch": 4.41, + "learning_rate": 1.6794867127775757e-06, + "loss": 2.0651, + "step": 24233 + }, + { + "epoch": 4.41, + "learning_rate": 1.6794321077539136e-06, + "loss": 0.2552, + "step": 24234 + }, + { + "epoch": 4.41, + "learning_rate": 1.6793775027302511e-06, + "loss": 1.2729, + "step": 24235 + }, + { + "epoch": 4.41, + "learning_rate": 1.679322897706589e-06, + "loss": 0.5302, + "step": 24236 + }, + { + "epoch": 4.41, + "learning_rate": 1.679268292682927e-06, + "loss": 1.193, + "step": 24237 + }, + { + "epoch": 4.41, + "learning_rate": 1.6792136876592647e-06, + "loss": 1.7739, + "step": 24238 + }, + { + "epoch": 4.41, + "learning_rate": 1.6791590826356027e-06, + "loss": 1.2686, + "step": 24239 + }, + { + "epoch": 4.41, + "learning_rate": 1.6791044776119402e-06, + "loss": 0.3738, + "step": 24240 + }, + { + "epoch": 4.41, + "learning_rate": 1.6790498725882781e-06, + "loss": 0.3024, + "step": 24241 + }, + { + "epoch": 4.41, + "learning_rate": 1.678995267564616e-06, + "loss": 1.5167, + "step": 24242 + }, + { + "epoch": 4.41, + "learning_rate": 1.6789406625409536e-06, + "loss": 1.0422, + "step": 24243 + }, + { + "epoch": 4.41, + "learning_rate": 1.6788860575172916e-06, + "loss": 0.8353, + "step": 24244 + }, + { + "epoch": 4.41, + "learning_rate": 1.6788314524936295e-06, + "loss": 0.6189, + "step": 24245 + }, + { + "epoch": 4.41, + "learning_rate": 1.6787768474699672e-06, + "loss": 0.0295, + "step": 24246 + }, + { + "epoch": 4.41, + "learning_rate": 1.6787222424463052e-06, + "loss": 0.5, + "step": 24247 + }, + { + "epoch": 4.41, + "learning_rate": 1.6786676374226431e-06, + "loss": 0.8603, + "step": 24248 + }, + { + "epoch": 4.41, + "learning_rate": 1.6786130323989806e-06, + "loss": 0.7102, + "step": 24249 + }, + { + "epoch": 4.41, + "learning_rate": 1.6785584273753186e-06, + "loss": 0.0933, + "step": 24250 + }, + { + "epoch": 4.41, + "learning_rate": 1.6785038223516565e-06, + "loss": 0.4196, + "step": 24251 + }, + { + "epoch": 4.41, + "learning_rate": 1.6784492173279943e-06, + "loss": 0.5682, + "step": 24252 + }, + { + "epoch": 4.41, + "learning_rate": 1.6783946123043322e-06, + "loss": 0.3072, + "step": 24253 + }, + { + "epoch": 4.41, + "learning_rate": 1.6783400072806697e-06, + "loss": 0.6812, + "step": 24254 + }, + { + "epoch": 4.41, + "learning_rate": 1.6782854022570077e-06, + "loss": 0.0066, + "step": 24255 + }, + { + "epoch": 4.41, + "learning_rate": 1.6782307972333456e-06, + "loss": 0.8284, + "step": 24256 + }, + { + "epoch": 4.41, + "learning_rate": 1.6781761922096831e-06, + "loss": 0.0049, + "step": 24257 + }, + { + "epoch": 4.41, + "learning_rate": 1.678121587186021e-06, + "loss": 0.2514, + "step": 24258 + }, + { + "epoch": 4.41, + "learning_rate": 1.678066982162359e-06, + "loss": 0.0199, + "step": 24259 + }, + { + "epoch": 4.41, + "learning_rate": 1.6780123771386967e-06, + "loss": 0.0258, + "step": 24260 + }, + { + "epoch": 4.41, + "learning_rate": 1.6779577721150347e-06, + "loss": 2.1219, + "step": 24261 + }, + { + "epoch": 4.41, + "learning_rate": 1.6779031670913726e-06, + "loss": 0.1075, + "step": 24262 + }, + { + "epoch": 4.41, + "learning_rate": 1.6778485620677102e-06, + "loss": 1.2947, + "step": 24263 + }, + { + "epoch": 4.41, + "learning_rate": 1.677793957044048e-06, + "loss": 2.6141, + "step": 24264 + }, + { + "epoch": 4.41, + "learning_rate": 1.6777393520203858e-06, + "loss": 0.1576, + "step": 24265 + }, + { + "epoch": 4.41, + "learning_rate": 1.6776847469967238e-06, + "loss": 0.1273, + "step": 24266 + }, + { + "epoch": 4.41, + "learning_rate": 1.6776301419730617e-06, + "loss": 0.3514, + "step": 24267 + }, + { + "epoch": 4.41, + "learning_rate": 1.6775755369493992e-06, + "loss": 0.1109, + "step": 24268 + }, + { + "epoch": 4.41, + "learning_rate": 1.6775209319257372e-06, + "loss": 0.2281, + "step": 24269 + }, + { + "epoch": 4.41, + "learning_rate": 1.6774663269020751e-06, + "loss": 1.5771, + "step": 24270 + }, + { + "epoch": 4.41, + "learning_rate": 1.6774117218784126e-06, + "loss": 3.3758, + "step": 24271 + }, + { + "epoch": 4.41, + "learning_rate": 1.6773571168547506e-06, + "loss": 0.3227, + "step": 24272 + }, + { + "epoch": 4.41, + "learning_rate": 1.6773025118310885e-06, + "loss": 0.1072, + "step": 24273 + }, + { + "epoch": 4.41, + "learning_rate": 1.6772479068074263e-06, + "loss": 1.6543, + "step": 24274 + }, + { + "epoch": 4.41, + "learning_rate": 1.6771933017837642e-06, + "loss": 0.4679, + "step": 24275 + }, + { + "epoch": 4.41, + "learning_rate": 1.6771386967601021e-06, + "loss": 0.7018, + "step": 24276 + }, + { + "epoch": 4.41, + "learning_rate": 1.6770840917364397e-06, + "loss": 0.1813, + "step": 24277 + }, + { + "epoch": 4.41, + "learning_rate": 1.6770294867127776e-06, + "loss": 1.4372, + "step": 24278 + }, + { + "epoch": 4.42, + "learning_rate": 1.6769748816891153e-06, + "loss": 0.4516, + "step": 24279 + }, + { + "epoch": 4.42, + "learning_rate": 1.6769202766654533e-06, + "loss": 0.0511, + "step": 24280 + }, + { + "epoch": 4.42, + "learning_rate": 1.6768656716417912e-06, + "loss": 1.2703, + "step": 24281 + }, + { + "epoch": 4.42, + "learning_rate": 1.6768110666181287e-06, + "loss": 1.2827, + "step": 24282 + }, + { + "epoch": 4.42, + "learning_rate": 1.6767564615944667e-06, + "loss": 0.0248, + "step": 24283 + }, + { + "epoch": 4.42, + "learning_rate": 1.6767018565708046e-06, + "loss": 0.3623, + "step": 24284 + }, + { + "epoch": 4.42, + "learning_rate": 1.6766472515471424e-06, + "loss": 0.9654, + "step": 24285 + }, + { + "epoch": 4.42, + "learning_rate": 1.6765926465234803e-06, + "loss": 0.0171, + "step": 24286 + }, + { + "epoch": 4.42, + "learning_rate": 1.6765380414998183e-06, + "loss": 0.0399, + "step": 24287 + }, + { + "epoch": 4.42, + "learning_rate": 1.6764834364761558e-06, + "loss": 0.1586, + "step": 24288 + }, + { + "epoch": 4.42, + "learning_rate": 1.6764288314524937e-06, + "loss": 0.0944, + "step": 24289 + }, + { + "epoch": 4.42, + "learning_rate": 1.6763742264288317e-06, + "loss": 0.0819, + "step": 24290 + }, + { + "epoch": 4.42, + "learning_rate": 1.6763196214051692e-06, + "loss": 0.0376, + "step": 24291 + }, + { + "epoch": 4.42, + "learning_rate": 1.6762650163815071e-06, + "loss": 0.1615, + "step": 24292 + }, + { + "epoch": 4.42, + "learning_rate": 1.6762104113578449e-06, + "loss": 0.0394, + "step": 24293 + }, + { + "epoch": 4.42, + "learning_rate": 1.6761558063341828e-06, + "loss": 0.8725, + "step": 24294 + }, + { + "epoch": 4.42, + "learning_rate": 1.6761012013105207e-06, + "loss": 0.2929, + "step": 24295 + }, + { + "epoch": 4.42, + "learning_rate": 1.6760465962868583e-06, + "loss": 0.5299, + "step": 24296 + }, + { + "epoch": 4.42, + "learning_rate": 1.6759919912631962e-06, + "loss": 2.9634, + "step": 24297 + }, + { + "epoch": 4.42, + "learning_rate": 1.6759373862395341e-06, + "loss": 0.1691, + "step": 24298 + }, + { + "epoch": 4.42, + "learning_rate": 1.6758827812158719e-06, + "loss": 0.4064, + "step": 24299 + }, + { + "epoch": 4.42, + "learning_rate": 1.6758281761922098e-06, + "loss": 0.296, + "step": 24300 + }, + { + "epoch": 4.42, + "learning_rate": 1.6757735711685478e-06, + "loss": 1.243, + "step": 24301 + }, + { + "epoch": 4.42, + "learning_rate": 1.6757189661448853e-06, + "loss": 0.3226, + "step": 24302 + }, + { + "epoch": 4.42, + "learning_rate": 1.6756643611212232e-06, + "loss": 0.4734, + "step": 24303 + }, + { + "epoch": 4.42, + "learning_rate": 1.6756097560975612e-06, + "loss": 0.2869, + "step": 24304 + }, + { + "epoch": 4.42, + "learning_rate": 1.6755551510738987e-06, + "loss": 0.0177, + "step": 24305 + }, + { + "epoch": 4.42, + "learning_rate": 1.6755005460502366e-06, + "loss": 2.3047, + "step": 24306 + }, + { + "epoch": 4.42, + "learning_rate": 1.6754459410265744e-06, + "loss": 3.3696, + "step": 24307 + }, + { + "epoch": 4.42, + "learning_rate": 1.6753913360029123e-06, + "loss": 1.9047, + "step": 24308 + }, + { + "epoch": 4.42, + "learning_rate": 1.6753367309792503e-06, + "loss": 1.1296, + "step": 24309 + }, + { + "epoch": 4.42, + "learning_rate": 1.6752821259555878e-06, + "loss": 1.0086, + "step": 24310 + }, + { + "epoch": 4.42, + "learning_rate": 1.6752275209319257e-06, + "loss": 0.0224, + "step": 24311 + }, + { + "epoch": 4.42, + "learning_rate": 1.6751729159082637e-06, + "loss": 0.2009, + "step": 24312 + }, + { + "epoch": 4.42, + "learning_rate": 1.6751183108846014e-06, + "loss": 1.1811, + "step": 24313 + }, + { + "epoch": 4.42, + "learning_rate": 1.6750637058609393e-06, + "loss": 1.8009, + "step": 24314 + }, + { + "epoch": 4.42, + "learning_rate": 1.6750091008372773e-06, + "loss": 0.8527, + "step": 24315 + }, + { + "epoch": 4.42, + "learning_rate": 1.6749544958136148e-06, + "loss": 1.4614, + "step": 24316 + }, + { + "epoch": 4.42, + "learning_rate": 1.6748998907899527e-06, + "loss": 0.0275, + "step": 24317 + }, + { + "epoch": 4.42, + "learning_rate": 1.6748452857662907e-06, + "loss": 0.028, + "step": 24318 + }, + { + "epoch": 4.42, + "learning_rate": 1.6747906807426284e-06, + "loss": 0.6104, + "step": 24319 + }, + { + "epoch": 4.42, + "learning_rate": 1.6747360757189661e-06, + "loss": 0.8081, + "step": 24320 + }, + { + "epoch": 4.42, + "learning_rate": 1.6746814706953039e-06, + "loss": 0.963, + "step": 24321 + }, + { + "epoch": 4.42, + "learning_rate": 1.6746268656716418e-06, + "loss": 1.7233, + "step": 24322 + }, + { + "epoch": 4.42, + "learning_rate": 1.6745722606479798e-06, + "loss": 1.158, + "step": 24323 + }, + { + "epoch": 4.42, + "learning_rate": 1.6745176556243173e-06, + "loss": 0.0115, + "step": 24324 + }, + { + "epoch": 4.42, + "learning_rate": 1.6744630506006552e-06, + "loss": 0.7809, + "step": 24325 + }, + { + "epoch": 4.42, + "learning_rate": 1.6744084455769932e-06, + "loss": 0.5106, + "step": 24326 + }, + { + "epoch": 4.42, + "learning_rate": 1.674353840553331e-06, + "loss": 0.0085, + "step": 24327 + }, + { + "epoch": 4.42, + "learning_rate": 1.6742992355296688e-06, + "loss": 1.8647, + "step": 24328 + }, + { + "epoch": 4.42, + "learning_rate": 1.6742446305060068e-06, + "loss": 0.6135, + "step": 24329 + }, + { + "epoch": 4.42, + "learning_rate": 1.6741900254823443e-06, + "loss": 1.071, + "step": 24330 + }, + { + "epoch": 4.42, + "learning_rate": 1.6741354204586823e-06, + "loss": 2.481, + "step": 24331 + }, + { + "epoch": 4.42, + "learning_rate": 1.6740808154350202e-06, + "loss": 0.2428, + "step": 24332 + }, + { + "epoch": 4.42, + "learning_rate": 1.674026210411358e-06, + "loss": 0.2093, + "step": 24333 + }, + { + "epoch": 4.43, + "learning_rate": 1.6739716053876959e-06, + "loss": 0.3488, + "step": 24334 + }, + { + "epoch": 4.43, + "learning_rate": 1.6739170003640334e-06, + "loss": 0.6796, + "step": 24335 + }, + { + "epoch": 4.43, + "learning_rate": 1.6738623953403713e-06, + "loss": 2.8063, + "step": 24336 + }, + { + "epoch": 4.43, + "learning_rate": 1.6738077903167093e-06, + "loss": 0.1771, + "step": 24337 + }, + { + "epoch": 4.43, + "learning_rate": 1.6737531852930468e-06, + "loss": 0.2, + "step": 24338 + }, + { + "epoch": 4.43, + "learning_rate": 1.6736985802693847e-06, + "loss": 0.1032, + "step": 24339 + }, + { + "epoch": 4.43, + "learning_rate": 1.6736439752457227e-06, + "loss": 0.1596, + "step": 24340 + }, + { + "epoch": 4.43, + "learning_rate": 1.6735893702220604e-06, + "loss": 0.2509, + "step": 24341 + }, + { + "epoch": 4.43, + "learning_rate": 1.6735347651983984e-06, + "loss": 0.7166, + "step": 24342 + }, + { + "epoch": 4.43, + "learning_rate": 1.6734801601747363e-06, + "loss": 1.0627, + "step": 24343 + }, + { + "epoch": 4.43, + "learning_rate": 1.6734255551510738e-06, + "loss": 0.1505, + "step": 24344 + }, + { + "epoch": 4.43, + "learning_rate": 1.6733709501274118e-06, + "loss": 1.0287, + "step": 24345 + }, + { + "epoch": 4.43, + "learning_rate": 1.6733163451037495e-06, + "loss": 0.1551, + "step": 24346 + }, + { + "epoch": 4.43, + "learning_rate": 1.6732617400800874e-06, + "loss": 0.4131, + "step": 24347 + }, + { + "epoch": 4.43, + "learning_rate": 1.6732071350564254e-06, + "loss": 1.0263, + "step": 24348 + }, + { + "epoch": 4.43, + "learning_rate": 1.673152530032763e-06, + "loss": 0.0291, + "step": 24349 + }, + { + "epoch": 4.43, + "learning_rate": 1.6730979250091008e-06, + "loss": 0.0158, + "step": 24350 + }, + { + "epoch": 4.43, + "learning_rate": 1.6730433199854388e-06, + "loss": 0.9855, + "step": 24351 + }, + { + "epoch": 4.43, + "learning_rate": 1.6729887149617763e-06, + "loss": 1.662, + "step": 24352 + }, + { + "epoch": 4.43, + "learning_rate": 1.6729341099381143e-06, + "loss": 0.0234, + "step": 24353 + }, + { + "epoch": 4.43, + "learning_rate": 1.6728795049144522e-06, + "loss": 0.0214, + "step": 24354 + }, + { + "epoch": 4.43, + "learning_rate": 1.67282489989079e-06, + "loss": 0.7019, + "step": 24355 + }, + { + "epoch": 4.43, + "learning_rate": 1.6727702948671279e-06, + "loss": 0.4386, + "step": 24356 + }, + { + "epoch": 4.43, + "learning_rate": 1.6727156898434658e-06, + "loss": 0.288, + "step": 24357 + }, + { + "epoch": 4.43, + "learning_rate": 1.6726610848198033e-06, + "loss": 0.2032, + "step": 24358 + }, + { + "epoch": 4.43, + "learning_rate": 1.6726064797961413e-06, + "loss": 0.036, + "step": 24359 + }, + { + "epoch": 4.43, + "learning_rate": 1.672551874772479e-06, + "loss": 0.3694, + "step": 24360 + }, + { + "epoch": 4.43, + "learning_rate": 1.672497269748817e-06, + "loss": 0.8455, + "step": 24361 + }, + { + "epoch": 4.43, + "learning_rate": 1.672442664725155e-06, + "loss": 0.1465, + "step": 24362 + }, + { + "epoch": 4.43, + "learning_rate": 1.6723880597014924e-06, + "loss": 0.3291, + "step": 24363 + }, + { + "epoch": 4.43, + "learning_rate": 1.6723334546778304e-06, + "loss": 0.7078, + "step": 24364 + }, + { + "epoch": 4.43, + "learning_rate": 1.6722788496541683e-06, + "loss": 0.2147, + "step": 24365 + }, + { + "epoch": 4.43, + "learning_rate": 1.672224244630506e-06, + "loss": 0.0395, + "step": 24366 + }, + { + "epoch": 4.43, + "learning_rate": 1.672169639606844e-06, + "loss": 0.2837, + "step": 24367 + }, + { + "epoch": 4.43, + "learning_rate": 1.672115034583182e-06, + "loss": 0.4113, + "step": 24368 + }, + { + "epoch": 4.43, + "learning_rate": 1.6720604295595194e-06, + "loss": 0.4305, + "step": 24369 + }, + { + "epoch": 4.43, + "learning_rate": 1.6720058245358574e-06, + "loss": 0.1206, + "step": 24370 + }, + { + "epoch": 4.43, + "learning_rate": 1.6719512195121953e-06, + "loss": 1.506, + "step": 24371 + }, + { + "epoch": 4.43, + "learning_rate": 1.6718966144885329e-06, + "loss": 1.0331, + "step": 24372 + }, + { + "epoch": 4.43, + "learning_rate": 1.6718420094648708e-06, + "loss": 0.5163, + "step": 24373 + }, + { + "epoch": 4.43, + "learning_rate": 1.6717874044412085e-06, + "loss": 0.3648, + "step": 24374 + }, + { + "epoch": 4.43, + "learning_rate": 1.6717327994175465e-06, + "loss": 2.7462, + "step": 24375 + }, + { + "epoch": 4.43, + "learning_rate": 1.6716781943938844e-06, + "loss": 0.2674, + "step": 24376 + }, + { + "epoch": 4.43, + "learning_rate": 1.671623589370222e-06, + "loss": 0.0186, + "step": 24377 + }, + { + "epoch": 4.43, + "learning_rate": 1.6715689843465599e-06, + "loss": 0.4563, + "step": 24378 + }, + { + "epoch": 4.43, + "learning_rate": 1.6715143793228978e-06, + "loss": 0.3918, + "step": 24379 + }, + { + "epoch": 4.43, + "learning_rate": 1.6714597742992356e-06, + "loss": 0.1012, + "step": 24380 + }, + { + "epoch": 4.43, + "learning_rate": 1.6714051692755735e-06, + "loss": 0.035, + "step": 24381 + }, + { + "epoch": 4.43, + "learning_rate": 1.6713505642519114e-06, + "loss": 1.1897, + "step": 24382 + }, + { + "epoch": 4.43, + "learning_rate": 1.671295959228249e-06, + "loss": 0.2033, + "step": 24383 + }, + { + "epoch": 4.43, + "learning_rate": 1.671241354204587e-06, + "loss": 0.4996, + "step": 24384 + }, + { + "epoch": 4.43, + "learning_rate": 1.6711867491809248e-06, + "loss": 1.0021, + "step": 24385 + }, + { + "epoch": 4.43, + "learning_rate": 1.6711321441572624e-06, + "loss": 0.1321, + "step": 24386 + }, + { + "epoch": 4.43, + "learning_rate": 1.6710775391336003e-06, + "loss": 0.0498, + "step": 24387 + }, + { + "epoch": 4.43, + "learning_rate": 1.671022934109938e-06, + "loss": 1.4519, + "step": 24388 + }, + { + "epoch": 4.44, + "learning_rate": 1.670968329086276e-06, + "loss": 1.532, + "step": 24389 + }, + { + "epoch": 4.44, + "learning_rate": 1.670913724062614e-06, + "loss": 0.1658, + "step": 24390 + }, + { + "epoch": 4.44, + "learning_rate": 1.6708591190389514e-06, + "loss": 0.011, + "step": 24391 + }, + { + "epoch": 4.44, + "learning_rate": 1.6708045140152894e-06, + "loss": 2.1383, + "step": 24392 + }, + { + "epoch": 4.44, + "learning_rate": 1.6707499089916273e-06, + "loss": 1.4317, + "step": 24393 + }, + { + "epoch": 4.44, + "learning_rate": 1.670695303967965e-06, + "loss": 0.7361, + "step": 24394 + }, + { + "epoch": 4.44, + "learning_rate": 1.670640698944303e-06, + "loss": 0.3521, + "step": 24395 + }, + { + "epoch": 4.44, + "learning_rate": 1.670586093920641e-06, + "loss": 0.3566, + "step": 24396 + }, + { + "epoch": 4.44, + "learning_rate": 1.6705314888969785e-06, + "loss": 0.0075, + "step": 24397 + }, + { + "epoch": 4.44, + "learning_rate": 1.6704768838733164e-06, + "loss": 1.3184, + "step": 24398 + }, + { + "epoch": 4.44, + "learning_rate": 1.6704222788496544e-06, + "loss": 1.3814, + "step": 24399 + }, + { + "epoch": 4.44, + "learning_rate": 1.6703676738259919e-06, + "loss": 0.0812, + "step": 24400 + }, + { + "epoch": 4.44, + "learning_rate": 1.6703130688023298e-06, + "loss": 0.0269, + "step": 24401 + }, + { + "epoch": 4.44, + "learning_rate": 1.6702584637786676e-06, + "loss": 3.5995, + "step": 24402 + }, + { + "epoch": 4.44, + "learning_rate": 1.6702038587550055e-06, + "loss": 1.4584, + "step": 24403 + }, + { + "epoch": 4.44, + "learning_rate": 1.6701492537313434e-06, + "loss": 0.1063, + "step": 24404 + }, + { + "epoch": 4.44, + "learning_rate": 1.670094648707681e-06, + "loss": 0.1279, + "step": 24405 + }, + { + "epoch": 4.44, + "learning_rate": 1.670040043684019e-06, + "loss": 0.5668, + "step": 24406 + }, + { + "epoch": 4.44, + "learning_rate": 1.6699854386603568e-06, + "loss": 1.386, + "step": 24407 + }, + { + "epoch": 4.44, + "learning_rate": 1.6699308336366946e-06, + "loss": 0.2062, + "step": 24408 + }, + { + "epoch": 4.44, + "learning_rate": 1.6698762286130325e-06, + "loss": 0.3381, + "step": 24409 + }, + { + "epoch": 4.44, + "learning_rate": 1.6698216235893705e-06, + "loss": 0.012, + "step": 24410 + }, + { + "epoch": 4.44, + "learning_rate": 1.669767018565708e-06, + "loss": 0.0325, + "step": 24411 + }, + { + "epoch": 4.44, + "learning_rate": 1.669712413542046e-06, + "loss": 0.8208, + "step": 24412 + }, + { + "epoch": 4.44, + "learning_rate": 1.6696578085183839e-06, + "loss": 0.7249, + "step": 24413 + }, + { + "epoch": 4.44, + "learning_rate": 1.6696032034947216e-06, + "loss": 0.2729, + "step": 24414 + }, + { + "epoch": 4.44, + "learning_rate": 1.6695485984710595e-06, + "loss": 0.2111, + "step": 24415 + }, + { + "epoch": 4.44, + "learning_rate": 1.669493993447397e-06, + "loss": 0.9614, + "step": 24416 + }, + { + "epoch": 4.44, + "learning_rate": 1.669439388423735e-06, + "loss": 0.1699, + "step": 24417 + }, + { + "epoch": 4.44, + "learning_rate": 1.669384783400073e-06, + "loss": 1.4729, + "step": 24418 + }, + { + "epoch": 4.44, + "learning_rate": 1.6693301783764105e-06, + "loss": 0.4253, + "step": 24419 + }, + { + "epoch": 4.44, + "learning_rate": 1.6692755733527484e-06, + "loss": 0.2494, + "step": 24420 + }, + { + "epoch": 4.44, + "learning_rate": 1.6692209683290864e-06, + "loss": 0.8477, + "step": 24421 + }, + { + "epoch": 4.44, + "learning_rate": 1.669166363305424e-06, + "loss": 0.2743, + "step": 24422 + }, + { + "epoch": 4.44, + "learning_rate": 1.669111758281762e-06, + "loss": 0.41, + "step": 24423 + }, + { + "epoch": 4.44, + "learning_rate": 1.6690571532581e-06, + "loss": 0.2538, + "step": 24424 + }, + { + "epoch": 4.44, + "learning_rate": 1.6690025482344375e-06, + "loss": 2.127, + "step": 24425 + }, + { + "epoch": 4.44, + "learning_rate": 1.6689479432107754e-06, + "loss": 0.0825, + "step": 24426 + }, + { + "epoch": 4.44, + "learning_rate": 1.6688933381871132e-06, + "loss": 2.7526, + "step": 24427 + }, + { + "epoch": 4.44, + "learning_rate": 1.6688387331634511e-06, + "loss": 0.7807, + "step": 24428 + }, + { + "epoch": 4.44, + "learning_rate": 1.668784128139789e-06, + "loss": 0.4702, + "step": 24429 + }, + { + "epoch": 4.44, + "learning_rate": 1.6687295231161266e-06, + "loss": 0.0115, + "step": 24430 + }, + { + "epoch": 4.44, + "learning_rate": 1.6686749180924645e-06, + "loss": 0.1669, + "step": 24431 + }, + { + "epoch": 4.44, + "learning_rate": 1.6686203130688025e-06, + "loss": 2.1265, + "step": 24432 + }, + { + "epoch": 4.44, + "learning_rate": 1.66856570804514e-06, + "loss": 0.1018, + "step": 24433 + }, + { + "epoch": 4.44, + "learning_rate": 1.668511103021478e-06, + "loss": 0.2999, + "step": 24434 + }, + { + "epoch": 4.44, + "learning_rate": 1.6684564979978159e-06, + "loss": 0.6921, + "step": 24435 + }, + { + "epoch": 4.44, + "learning_rate": 1.6684018929741536e-06, + "loss": 0.2609, + "step": 24436 + }, + { + "epoch": 4.44, + "learning_rate": 1.6683472879504915e-06, + "loss": 0.6632, + "step": 24437 + }, + { + "epoch": 4.44, + "learning_rate": 1.6682926829268295e-06, + "loss": 0.2819, + "step": 24438 + }, + { + "epoch": 4.44, + "learning_rate": 1.668238077903167e-06, + "loss": 0.0134, + "step": 24439 + }, + { + "epoch": 4.44, + "learning_rate": 1.668183472879505e-06, + "loss": 0.3447, + "step": 24440 + }, + { + "epoch": 4.44, + "learning_rate": 1.6681288678558427e-06, + "loss": 1.1337, + "step": 24441 + }, + { + "epoch": 4.44, + "learning_rate": 1.6680742628321806e-06, + "loss": 0.7274, + "step": 24442 + }, + { + "epoch": 4.44, + "learning_rate": 1.6680196578085186e-06, + "loss": 0.0193, + "step": 24443 + }, + { + "epoch": 4.45, + "learning_rate": 1.667965052784856e-06, + "loss": 0.0883, + "step": 24444 + }, + { + "epoch": 4.45, + "learning_rate": 1.667910447761194e-06, + "loss": 0.3519, + "step": 24445 + }, + { + "epoch": 4.45, + "learning_rate": 1.667855842737532e-06, + "loss": 2.2552, + "step": 24446 + }, + { + "epoch": 4.45, + "learning_rate": 1.6678012377138697e-06, + "loss": 0.4464, + "step": 24447 + }, + { + "epoch": 4.45, + "learning_rate": 1.6677466326902077e-06, + "loss": 0.1381, + "step": 24448 + }, + { + "epoch": 4.45, + "learning_rate": 1.6676920276665456e-06, + "loss": 0.0076, + "step": 24449 + }, + { + "epoch": 4.45, + "learning_rate": 1.6676374226428831e-06, + "loss": 0.6897, + "step": 24450 + }, + { + "epoch": 4.45, + "learning_rate": 1.667582817619221e-06, + "loss": 4.5679, + "step": 24451 + }, + { + "epoch": 4.45, + "learning_rate": 1.667528212595559e-06, + "loss": 0.0074, + "step": 24452 + }, + { + "epoch": 4.45, + "learning_rate": 1.6674736075718965e-06, + "loss": 0.6866, + "step": 24453 + }, + { + "epoch": 4.45, + "learning_rate": 1.6674190025482345e-06, + "loss": 0.7919, + "step": 24454 + }, + { + "epoch": 4.45, + "learning_rate": 1.6673643975245722e-06, + "loss": 0.218, + "step": 24455 + }, + { + "epoch": 4.45, + "learning_rate": 1.6673097925009101e-06, + "loss": 1.1556, + "step": 24456 + }, + { + "epoch": 4.45, + "learning_rate": 1.667255187477248e-06, + "loss": 0.9825, + "step": 24457 + }, + { + "epoch": 4.45, + "learning_rate": 1.6672005824535856e-06, + "loss": 2.8453, + "step": 24458 + }, + { + "epoch": 4.45, + "learning_rate": 1.6671459774299236e-06, + "loss": 0.5556, + "step": 24459 + }, + { + "epoch": 4.45, + "learning_rate": 1.6670913724062615e-06, + "loss": 0.0474, + "step": 24460 + }, + { + "epoch": 4.45, + "learning_rate": 1.6670367673825992e-06, + "loss": 0.1876, + "step": 24461 + }, + { + "epoch": 4.45, + "learning_rate": 1.6669821623589372e-06, + "loss": 0.4438, + "step": 24462 + }, + { + "epoch": 4.45, + "learning_rate": 1.6669275573352751e-06, + "loss": 0.5643, + "step": 24463 + }, + { + "epoch": 4.45, + "learning_rate": 1.6668729523116126e-06, + "loss": 0.4402, + "step": 24464 + }, + { + "epoch": 4.45, + "learning_rate": 1.6668183472879506e-06, + "loss": 0.5802, + "step": 24465 + }, + { + "epoch": 4.45, + "learning_rate": 1.6667637422642885e-06, + "loss": 0.639, + "step": 24466 + }, + { + "epoch": 4.45, + "learning_rate": 1.666709137240626e-06, + "loss": 0.5281, + "step": 24467 + }, + { + "epoch": 4.45, + "learning_rate": 1.666654532216964e-06, + "loss": 0.0837, + "step": 24468 + }, + { + "epoch": 4.45, + "learning_rate": 1.6665999271933017e-06, + "loss": 0.0131, + "step": 24469 + }, + { + "epoch": 4.45, + "learning_rate": 1.6665453221696397e-06, + "loss": 0.428, + "step": 24470 + }, + { + "epoch": 4.45, + "learning_rate": 1.6664907171459776e-06, + "loss": 0.7212, + "step": 24471 + }, + { + "epoch": 4.45, + "learning_rate": 1.6664361121223151e-06, + "loss": 0.8732, + "step": 24472 + }, + { + "epoch": 4.45, + "learning_rate": 1.666381507098653e-06, + "loss": 0.1891, + "step": 24473 + }, + { + "epoch": 4.45, + "learning_rate": 1.666326902074991e-06, + "loss": 0.4141, + "step": 24474 + }, + { + "epoch": 4.45, + "learning_rate": 1.6662722970513287e-06, + "loss": 0.1031, + "step": 24475 + }, + { + "epoch": 4.45, + "learning_rate": 1.6662176920276667e-06, + "loss": 0.4008, + "step": 24476 + }, + { + "epoch": 4.45, + "learning_rate": 1.6661630870040046e-06, + "loss": 0.809, + "step": 24477 + }, + { + "epoch": 4.45, + "learning_rate": 1.6661084819803421e-06, + "loss": 0.4312, + "step": 24478 + }, + { + "epoch": 4.45, + "learning_rate": 1.66605387695668e-06, + "loss": 1.3919, + "step": 24479 + }, + { + "epoch": 4.45, + "learning_rate": 1.665999271933018e-06, + "loss": 0.9153, + "step": 24480 + }, + { + "epoch": 4.45, + "learning_rate": 1.6659446669093556e-06, + "loss": 1.5696, + "step": 24481 + }, + { + "epoch": 4.45, + "learning_rate": 1.6658900618856935e-06, + "loss": 0.4733, + "step": 24482 + }, + { + "epoch": 4.45, + "learning_rate": 1.6658354568620312e-06, + "loss": 0.2855, + "step": 24483 + }, + { + "epoch": 4.45, + "learning_rate": 1.6657808518383692e-06, + "loss": 0.4074, + "step": 24484 + }, + { + "epoch": 4.45, + "learning_rate": 1.6657262468147071e-06, + "loss": 1.4413, + "step": 24485 + }, + { + "epoch": 4.45, + "learning_rate": 1.6656716417910446e-06, + "loss": 1.4997, + "step": 24486 + }, + { + "epoch": 4.45, + "learning_rate": 1.6656170367673826e-06, + "loss": 0.1613, + "step": 24487 + }, + { + "epoch": 4.45, + "learning_rate": 1.6655624317437205e-06, + "loss": 0.784, + "step": 24488 + }, + { + "epoch": 4.45, + "learning_rate": 1.6655078267200583e-06, + "loss": 0.0293, + "step": 24489 + }, + { + "epoch": 4.45, + "learning_rate": 1.6654532216963962e-06, + "loss": 2.4949, + "step": 24490 + }, + { + "epoch": 4.45, + "learning_rate": 1.6653986166727341e-06, + "loss": 6.014, + "step": 24491 + }, + { + "epoch": 4.45, + "learning_rate": 1.6653440116490717e-06, + "loss": 0.0179, + "step": 24492 + }, + { + "epoch": 4.45, + "learning_rate": 1.6652894066254096e-06, + "loss": 1.9624, + "step": 24493 + }, + { + "epoch": 4.45, + "learning_rate": 1.6652348016017473e-06, + "loss": 0.3066, + "step": 24494 + }, + { + "epoch": 4.45, + "learning_rate": 1.6651801965780853e-06, + "loss": 0.0921, + "step": 24495 + }, + { + "epoch": 4.45, + "learning_rate": 1.6651255915544232e-06, + "loss": 0.3205, + "step": 24496 + }, + { + "epoch": 4.45, + "learning_rate": 1.6650709865307607e-06, + "loss": 1.1807, + "step": 24497 + }, + { + "epoch": 4.45, + "learning_rate": 1.6650163815070987e-06, + "loss": 0.4159, + "step": 24498 + }, + { + "epoch": 4.46, + "learning_rate": 1.6649617764834366e-06, + "loss": 0.1828, + "step": 24499 + }, + { + "epoch": 4.46, + "learning_rate": 1.6649071714597741e-06, + "loss": 0.0119, + "step": 24500 + }, + { + "epoch": 4.46, + "learning_rate": 1.664852566436112e-06, + "loss": 1.5555, + "step": 24501 + }, + { + "epoch": 4.46, + "learning_rate": 1.66479796141245e-06, + "loss": 0.804, + "step": 24502 + }, + { + "epoch": 4.46, + "learning_rate": 1.6647433563887878e-06, + "loss": 0.019, + "step": 24503 + }, + { + "epoch": 4.46, + "learning_rate": 1.6646887513651257e-06, + "loss": 0.014, + "step": 24504 + }, + { + "epoch": 4.46, + "learning_rate": 1.6646341463414637e-06, + "loss": 0.9136, + "step": 24505 + }, + { + "epoch": 4.46, + "learning_rate": 1.6645795413178012e-06, + "loss": 0.8682, + "step": 24506 + }, + { + "epoch": 4.46, + "learning_rate": 1.6645249362941391e-06, + "loss": 0.5614, + "step": 24507 + }, + { + "epoch": 4.46, + "learning_rate": 1.6644703312704768e-06, + "loss": 0.0105, + "step": 24508 + }, + { + "epoch": 4.46, + "learning_rate": 1.6644157262468148e-06, + "loss": 0.0843, + "step": 24509 + }, + { + "epoch": 4.46, + "learning_rate": 1.6643611212231527e-06, + "loss": 0.5544, + "step": 24510 + }, + { + "epoch": 4.46, + "learning_rate": 1.6643065161994903e-06, + "loss": 0.4082, + "step": 24511 + }, + { + "epoch": 4.46, + "learning_rate": 1.6642519111758282e-06, + "loss": 1.5043, + "step": 24512 + }, + { + "epoch": 4.46, + "learning_rate": 1.6641973061521661e-06, + "loss": 0.0163, + "step": 24513 + }, + { + "epoch": 4.46, + "learning_rate": 1.6641427011285037e-06, + "loss": 0.6437, + "step": 24514 + }, + { + "epoch": 4.46, + "learning_rate": 1.6640880961048416e-06, + "loss": 0.3521, + "step": 24515 + }, + { + "epoch": 4.46, + "learning_rate": 1.6640334910811795e-06, + "loss": 2.972, + "step": 24516 + }, + { + "epoch": 4.46, + "learning_rate": 1.6639788860575173e-06, + "loss": 0.0094, + "step": 24517 + }, + { + "epoch": 4.46, + "learning_rate": 1.6639242810338552e-06, + "loss": 0.1733, + "step": 24518 + }, + { + "epoch": 4.46, + "learning_rate": 1.6638696760101932e-06, + "loss": 0.8316, + "step": 24519 + }, + { + "epoch": 4.46, + "learning_rate": 1.6638150709865307e-06, + "loss": 2.0218, + "step": 24520 + }, + { + "epoch": 4.46, + "learning_rate": 1.6637604659628686e-06, + "loss": 0.0431, + "step": 24521 + }, + { + "epoch": 4.46, + "learning_rate": 1.6637058609392064e-06, + "loss": 0.0089, + "step": 24522 + }, + { + "epoch": 4.46, + "learning_rate": 1.6636512559155443e-06, + "loss": 0.3143, + "step": 24523 + }, + { + "epoch": 4.46, + "learning_rate": 1.6635966508918822e-06, + "loss": 2.0136, + "step": 24524 + }, + { + "epoch": 4.46, + "learning_rate": 1.6635420458682198e-06, + "loss": 0.4515, + "step": 24525 + }, + { + "epoch": 4.46, + "learning_rate": 1.6634874408445577e-06, + "loss": 0.6314, + "step": 24526 + }, + { + "epoch": 4.46, + "learning_rate": 1.6634328358208957e-06, + "loss": 0.6679, + "step": 24527 + }, + { + "epoch": 4.46, + "learning_rate": 1.6633782307972332e-06, + "loss": 0.0744, + "step": 24528 + }, + { + "epoch": 4.46, + "learning_rate": 1.6633236257735711e-06, + "loss": 0.0166, + "step": 24529 + }, + { + "epoch": 4.46, + "learning_rate": 1.663269020749909e-06, + "loss": 1.0581, + "step": 24530 + }, + { + "epoch": 4.46, + "learning_rate": 1.6632144157262468e-06, + "loss": 1.1319, + "step": 24531 + }, + { + "epoch": 4.46, + "learning_rate": 1.6631598107025847e-06, + "loss": 0.517, + "step": 24532 + }, + { + "epoch": 4.46, + "learning_rate": 1.6631052056789227e-06, + "loss": 0.2427, + "step": 24533 + }, + { + "epoch": 4.46, + "learning_rate": 1.6630506006552602e-06, + "loss": 0.2185, + "step": 24534 + }, + { + "epoch": 4.46, + "learning_rate": 1.6629959956315981e-06, + "loss": 0.3429, + "step": 24535 + }, + { + "epoch": 4.46, + "learning_rate": 1.6629413906079359e-06, + "loss": 0.2585, + "step": 24536 + }, + { + "epoch": 4.46, + "learning_rate": 1.6628867855842738e-06, + "loss": 0.0385, + "step": 24537 + }, + { + "epoch": 4.46, + "learning_rate": 1.6628321805606118e-06, + "loss": 0.0287, + "step": 24538 + }, + { + "epoch": 4.46, + "learning_rate": 1.6627775755369493e-06, + "loss": 1.2386, + "step": 24539 + }, + { + "epoch": 4.46, + "learning_rate": 1.6627229705132872e-06, + "loss": 0.5978, + "step": 24540 + }, + { + "epoch": 4.46, + "learning_rate": 1.6626683654896252e-06, + "loss": 0.394, + "step": 24541 + }, + { + "epoch": 4.46, + "learning_rate": 1.662613760465963e-06, + "loss": 1.2169, + "step": 24542 + }, + { + "epoch": 4.46, + "learning_rate": 1.6625591554423008e-06, + "loss": 1.3709, + "step": 24543 + }, + { + "epoch": 4.46, + "learning_rate": 1.6625045504186388e-06, + "loss": 1.6172, + "step": 24544 + }, + { + "epoch": 4.46, + "learning_rate": 1.6624499453949763e-06, + "loss": 1.6329, + "step": 24545 + }, + { + "epoch": 4.46, + "learning_rate": 1.6623953403713143e-06, + "loss": 0.4843, + "step": 24546 + }, + { + "epoch": 4.46, + "learning_rate": 1.6623407353476522e-06, + "loss": 1.0784, + "step": 24547 + }, + { + "epoch": 4.46, + "learning_rate": 1.6622861303239897e-06, + "loss": 0.0161, + "step": 24548 + }, + { + "epoch": 4.46, + "learning_rate": 1.6622315253003277e-06, + "loss": 3.5286, + "step": 24549 + }, + { + "epoch": 4.46, + "learning_rate": 1.6621769202766654e-06, + "loss": 0.8058, + "step": 24550 + }, + { + "epoch": 4.46, + "learning_rate": 1.6621223152530033e-06, + "loss": 0.1527, + "step": 24551 + }, + { + "epoch": 4.46, + "learning_rate": 1.6620677102293413e-06, + "loss": 0.0199, + "step": 24552 + }, + { + "epoch": 4.46, + "learning_rate": 1.6620131052056788e-06, + "loss": 0.8882, + "step": 24553 + }, + { + "epoch": 4.47, + "learning_rate": 1.6619585001820167e-06, + "loss": 0.8963, + "step": 24554 + }, + { + "epoch": 4.47, + "learning_rate": 1.6619038951583547e-06, + "loss": 0.0149, + "step": 24555 + }, + { + "epoch": 4.47, + "learning_rate": 1.6618492901346924e-06, + "loss": 0.4426, + "step": 24556 + }, + { + "epoch": 4.47, + "learning_rate": 1.6617946851110304e-06, + "loss": 0.2045, + "step": 24557 + }, + { + "epoch": 4.47, + "learning_rate": 1.6617400800873683e-06, + "loss": 0.635, + "step": 24558 + }, + { + "epoch": 4.47, + "learning_rate": 1.6616854750637058e-06, + "loss": 0.0142, + "step": 24559 + }, + { + "epoch": 4.47, + "learning_rate": 1.6616308700400438e-06, + "loss": 1.0248, + "step": 24560 + }, + { + "epoch": 4.47, + "learning_rate": 1.6615762650163817e-06, + "loss": 1.2607, + "step": 24561 + }, + { + "epoch": 4.47, + "learning_rate": 1.6615216599927192e-06, + "loss": 0.6809, + "step": 24562 + }, + { + "epoch": 4.47, + "learning_rate": 1.6614670549690572e-06, + "loss": 0.1181, + "step": 24563 + }, + { + "epoch": 4.47, + "learning_rate": 1.661412449945395e-06, + "loss": 1.3377, + "step": 24564 + }, + { + "epoch": 4.47, + "learning_rate": 1.6613578449217328e-06, + "loss": 0.5445, + "step": 24565 + }, + { + "epoch": 4.47, + "learning_rate": 1.6613032398980708e-06, + "loss": 0.2939, + "step": 24566 + }, + { + "epoch": 4.47, + "learning_rate": 1.6612486348744083e-06, + "loss": 0.2813, + "step": 24567 + }, + { + "epoch": 4.47, + "learning_rate": 1.6611940298507463e-06, + "loss": 1.1684, + "step": 24568 + }, + { + "epoch": 4.47, + "learning_rate": 1.6611394248270842e-06, + "loss": 0.4674, + "step": 24569 + }, + { + "epoch": 4.47, + "learning_rate": 1.661084819803422e-06, + "loss": 0.3826, + "step": 24570 + }, + { + "epoch": 4.47, + "learning_rate": 1.6610302147797599e-06, + "loss": 0.7563, + "step": 24571 + }, + { + "epoch": 4.47, + "learning_rate": 1.6609756097560978e-06, + "loss": 0.2574, + "step": 24572 + }, + { + "epoch": 4.47, + "learning_rate": 1.6609210047324353e-06, + "loss": 0.0084, + "step": 24573 + }, + { + "epoch": 4.47, + "learning_rate": 1.6608663997087733e-06, + "loss": 0.7437, + "step": 24574 + }, + { + "epoch": 4.47, + "learning_rate": 1.660811794685111e-06, + "loss": 0.7282, + "step": 24575 + }, + { + "epoch": 4.47, + "learning_rate": 1.660757189661449e-06, + "loss": 0.0718, + "step": 24576 + }, + { + "epoch": 4.47, + "learning_rate": 1.660702584637787e-06, + "loss": 0.0257, + "step": 24577 + }, + { + "epoch": 4.47, + "learning_rate": 1.6606479796141244e-06, + "loss": 0.3943, + "step": 24578 + }, + { + "epoch": 4.47, + "learning_rate": 1.6605933745904624e-06, + "loss": 1.8175, + "step": 24579 + }, + { + "epoch": 4.47, + "learning_rate": 1.6605387695668003e-06, + "loss": 1.827, + "step": 24580 + }, + { + "epoch": 4.47, + "learning_rate": 1.6604841645431378e-06, + "loss": 0.2218, + "step": 24581 + }, + { + "epoch": 4.47, + "learning_rate": 1.6604295595194758e-06, + "loss": 1.1686, + "step": 24582 + }, + { + "epoch": 4.47, + "learning_rate": 1.6603749544958137e-06, + "loss": 0.4261, + "step": 24583 + }, + { + "epoch": 4.47, + "learning_rate": 1.6603203494721514e-06, + "loss": 1.6868, + "step": 24584 + }, + { + "epoch": 4.47, + "learning_rate": 1.6602657444484894e-06, + "loss": 1.1859, + "step": 24585 + }, + { + "epoch": 4.47, + "learning_rate": 1.6602111394248273e-06, + "loss": 0.32, + "step": 24586 + }, + { + "epoch": 4.47, + "learning_rate": 1.6601565344011648e-06, + "loss": 0.0273, + "step": 24587 + }, + { + "epoch": 4.47, + "learning_rate": 1.6601019293775028e-06, + "loss": 0.1913, + "step": 24588 + }, + { + "epoch": 4.47, + "learning_rate": 1.6600473243538405e-06, + "loss": 0.4453, + "step": 24589 + }, + { + "epoch": 4.47, + "learning_rate": 1.6599927193301785e-06, + "loss": 0.0191, + "step": 24590 + }, + { + "epoch": 4.47, + "learning_rate": 1.6599381143065164e-06, + "loss": 0.1241, + "step": 24591 + }, + { + "epoch": 4.47, + "learning_rate": 1.659883509282854e-06, + "loss": 1.2615, + "step": 24592 + }, + { + "epoch": 4.47, + "learning_rate": 1.6598289042591919e-06, + "loss": 0.5103, + "step": 24593 + }, + { + "epoch": 4.47, + "learning_rate": 1.6597742992355298e-06, + "loss": 0.0139, + "step": 24594 + }, + { + "epoch": 4.47, + "learning_rate": 1.6597196942118673e-06, + "loss": 1.5779, + "step": 24595 + }, + { + "epoch": 4.47, + "learning_rate": 1.6596650891882053e-06, + "loss": 1.5128, + "step": 24596 + }, + { + "epoch": 4.47, + "learning_rate": 1.6596104841645432e-06, + "loss": 0.0122, + "step": 24597 + }, + { + "epoch": 4.47, + "learning_rate": 1.659555879140881e-06, + "loss": 1.075, + "step": 24598 + }, + { + "epoch": 4.47, + "learning_rate": 1.659501274117219e-06, + "loss": 0.3998, + "step": 24599 + }, + { + "epoch": 4.47, + "learning_rate": 1.6594466690935568e-06, + "loss": 0.0053, + "step": 24600 + }, + { + "epoch": 4.47, + "learning_rate": 1.6593920640698944e-06, + "loss": 0.369, + "step": 24601 + }, + { + "epoch": 4.47, + "learning_rate": 1.6593374590462323e-06, + "loss": 0.5475, + "step": 24602 + }, + { + "epoch": 4.47, + "learning_rate": 1.65928285402257e-06, + "loss": 1.862, + "step": 24603 + }, + { + "epoch": 4.47, + "learning_rate": 1.659228248998908e-06, + "loss": 0.2035, + "step": 24604 + }, + { + "epoch": 4.47, + "learning_rate": 1.659173643975246e-06, + "loss": 0.3192, + "step": 24605 + }, + { + "epoch": 4.47, + "learning_rate": 1.6591190389515834e-06, + "loss": 0.7739, + "step": 24606 + }, + { + "epoch": 4.47, + "learning_rate": 1.6590644339279214e-06, + "loss": 0.2411, + "step": 24607 + }, + { + "epoch": 4.47, + "learning_rate": 1.6590098289042593e-06, + "loss": 0.3505, + "step": 24608 + }, + { + "epoch": 4.48, + "learning_rate": 1.6589552238805968e-06, + "loss": 0.0487, + "step": 24609 + }, + { + "epoch": 4.48, + "learning_rate": 1.6589006188569348e-06, + "loss": 0.7946, + "step": 24610 + }, + { + "epoch": 4.48, + "learning_rate": 1.6588460138332727e-06, + "loss": 0.0424, + "step": 24611 + }, + { + "epoch": 4.48, + "learning_rate": 1.6587914088096105e-06, + "loss": 1.4894, + "step": 24612 + }, + { + "epoch": 4.48, + "learning_rate": 1.6587368037859484e-06, + "loss": 0.0713, + "step": 24613 + }, + { + "epoch": 4.48, + "learning_rate": 1.6586821987622864e-06, + "loss": 1.0923, + "step": 24614 + }, + { + "epoch": 4.48, + "learning_rate": 1.6586275937386239e-06, + "loss": 1.6054, + "step": 24615 + }, + { + "epoch": 4.48, + "learning_rate": 1.6585729887149618e-06, + "loss": 0.5509, + "step": 24616 + }, + { + "epoch": 4.48, + "learning_rate": 1.6585183836912995e-06, + "loss": 0.0947, + "step": 24617 + }, + { + "epoch": 4.48, + "learning_rate": 1.6584637786676375e-06, + "loss": 2.5072, + "step": 24618 + }, + { + "epoch": 4.48, + "learning_rate": 1.6584091736439754e-06, + "loss": 0.0379, + "step": 24619 + }, + { + "epoch": 4.48, + "learning_rate": 1.658354568620313e-06, + "loss": 0.9685, + "step": 24620 + }, + { + "epoch": 4.48, + "learning_rate": 1.658299963596651e-06, + "loss": 1.3404, + "step": 24621 + }, + { + "epoch": 4.48, + "learning_rate": 1.6582453585729888e-06, + "loss": 1.9156, + "step": 24622 + }, + { + "epoch": 4.48, + "learning_rate": 1.6581907535493266e-06, + "loss": 1.4872, + "step": 24623 + }, + { + "epoch": 4.48, + "learning_rate": 1.6581361485256645e-06, + "loss": 1.5707, + "step": 24624 + }, + { + "epoch": 4.48, + "learning_rate": 1.6580815435020025e-06, + "loss": 1.1073, + "step": 24625 + }, + { + "epoch": 4.48, + "learning_rate": 1.65802693847834e-06, + "loss": 1.1241, + "step": 24626 + }, + { + "epoch": 4.48, + "learning_rate": 1.657972333454678e-06, + "loss": 0.6209, + "step": 24627 + }, + { + "epoch": 4.48, + "learning_rate": 1.6579177284310159e-06, + "loss": 0.0506, + "step": 24628 + }, + { + "epoch": 4.48, + "learning_rate": 1.6578631234073534e-06, + "loss": 0.0341, + "step": 24629 + }, + { + "epoch": 4.48, + "learning_rate": 1.6578085183836913e-06, + "loss": 1.7035, + "step": 24630 + }, + { + "epoch": 4.48, + "learning_rate": 1.657753913360029e-06, + "loss": 0.7437, + "step": 24631 + }, + { + "epoch": 4.48, + "learning_rate": 1.657699308336367e-06, + "loss": 0.7696, + "step": 24632 + }, + { + "epoch": 4.48, + "learning_rate": 1.657644703312705e-06, + "loss": 0.0147, + "step": 24633 + }, + { + "epoch": 4.48, + "learning_rate": 1.6575900982890425e-06, + "loss": 0.3226, + "step": 24634 + }, + { + "epoch": 4.48, + "learning_rate": 1.6575354932653804e-06, + "loss": 0.5543, + "step": 24635 + }, + { + "epoch": 4.48, + "learning_rate": 1.6574808882417184e-06, + "loss": 0.2671, + "step": 24636 + }, + { + "epoch": 4.48, + "learning_rate": 1.657426283218056e-06, + "loss": 0.0253, + "step": 24637 + }, + { + "epoch": 4.48, + "learning_rate": 1.657371678194394e-06, + "loss": 0.0412, + "step": 24638 + }, + { + "epoch": 4.48, + "learning_rate": 1.657317073170732e-06, + "loss": 0.6751, + "step": 24639 + }, + { + "epoch": 4.48, + "learning_rate": 1.6572624681470695e-06, + "loss": 0.1993, + "step": 24640 + }, + { + "epoch": 4.48, + "learning_rate": 1.6572078631234074e-06, + "loss": 0.3836, + "step": 24641 + }, + { + "epoch": 4.48, + "learning_rate": 1.6571532580997454e-06, + "loss": 0.2528, + "step": 24642 + }, + { + "epoch": 4.48, + "learning_rate": 1.657098653076083e-06, + "loss": 1.5987, + "step": 24643 + }, + { + "epoch": 4.48, + "learning_rate": 1.6570440480524208e-06, + "loss": 0.0042, + "step": 24644 + }, + { + "epoch": 4.48, + "learning_rate": 1.6569894430287586e-06, + "loss": 0.432, + "step": 24645 + }, + { + "epoch": 4.48, + "learning_rate": 1.6569348380050965e-06, + "loss": 0.0237, + "step": 24646 + }, + { + "epoch": 4.48, + "learning_rate": 1.6568802329814345e-06, + "loss": 1.5262, + "step": 24647 + }, + { + "epoch": 4.48, + "learning_rate": 1.656825627957772e-06, + "loss": 0.0876, + "step": 24648 + }, + { + "epoch": 4.48, + "learning_rate": 1.65677102293411e-06, + "loss": 0.1009, + "step": 24649 + }, + { + "epoch": 4.48, + "learning_rate": 1.6567164179104479e-06, + "loss": 0.1524, + "step": 24650 + }, + { + "epoch": 4.48, + "learning_rate": 1.6566618128867856e-06, + "loss": 0.463, + "step": 24651 + }, + { + "epoch": 4.48, + "learning_rate": 1.6566072078631235e-06, + "loss": 1.5176, + "step": 24652 + }, + { + "epoch": 4.48, + "learning_rate": 1.6565526028394615e-06, + "loss": 0.3959, + "step": 24653 + }, + { + "epoch": 4.48, + "learning_rate": 1.656497997815799e-06, + "loss": 0.5754, + "step": 24654 + }, + { + "epoch": 4.48, + "learning_rate": 1.656443392792137e-06, + "loss": 3.2313, + "step": 24655 + }, + { + "epoch": 4.48, + "learning_rate": 1.6563887877684747e-06, + "loss": 0.013, + "step": 24656 + }, + { + "epoch": 4.48, + "learning_rate": 1.6563341827448124e-06, + "loss": 1.3819, + "step": 24657 + }, + { + "epoch": 4.48, + "learning_rate": 1.6562795777211504e-06, + "loss": 0.0909, + "step": 24658 + }, + { + "epoch": 4.48, + "learning_rate": 1.656224972697488e-06, + "loss": 1.3419, + "step": 24659 + }, + { + "epoch": 4.48, + "learning_rate": 1.656170367673826e-06, + "loss": 1.2055, + "step": 24660 + }, + { + "epoch": 4.48, + "learning_rate": 1.656115762650164e-06, + "loss": 2.1065, + "step": 24661 + }, + { + "epoch": 4.48, + "learning_rate": 1.6560611576265015e-06, + "loss": 0.5872, + "step": 24662 + }, + { + "epoch": 4.48, + "learning_rate": 1.6560065526028394e-06, + "loss": 0.0704, + "step": 24663 + }, + { + "epoch": 4.49, + "learning_rate": 1.6559519475791774e-06, + "loss": 0.4526, + "step": 24664 + }, + { + "epoch": 4.49, + "learning_rate": 1.6558973425555151e-06, + "loss": 2.552, + "step": 24665 + }, + { + "epoch": 4.49, + "learning_rate": 1.655842737531853e-06, + "loss": 0.3453, + "step": 24666 + }, + { + "epoch": 4.49, + "learning_rate": 1.655788132508191e-06, + "loss": 0.0089, + "step": 24667 + }, + { + "epoch": 4.49, + "learning_rate": 1.6557335274845285e-06, + "loss": 0.1529, + "step": 24668 + }, + { + "epoch": 4.49, + "learning_rate": 1.6556789224608665e-06, + "loss": 1.3831, + "step": 24669 + }, + { + "epoch": 4.49, + "learning_rate": 1.6556243174372042e-06, + "loss": 0.2132, + "step": 24670 + }, + { + "epoch": 4.49, + "learning_rate": 1.6555697124135421e-06, + "loss": 0.2125, + "step": 24671 + }, + { + "epoch": 4.49, + "learning_rate": 1.65551510738988e-06, + "loss": 0.5723, + "step": 24672 + }, + { + "epoch": 4.49, + "learning_rate": 1.6554605023662176e-06, + "loss": 0.4477, + "step": 24673 + }, + { + "epoch": 4.49, + "learning_rate": 1.6554058973425555e-06, + "loss": 0.7161, + "step": 24674 + }, + { + "epoch": 4.49, + "learning_rate": 1.6553512923188935e-06, + "loss": 1.122, + "step": 24675 + }, + { + "epoch": 4.49, + "learning_rate": 1.655296687295231e-06, + "loss": 0.7782, + "step": 24676 + }, + { + "epoch": 4.49, + "learning_rate": 1.655242082271569e-06, + "loss": 0.1591, + "step": 24677 + }, + { + "epoch": 4.49, + "learning_rate": 1.655187477247907e-06, + "loss": 0.1763, + "step": 24678 + }, + { + "epoch": 4.49, + "learning_rate": 1.6551328722242446e-06, + "loss": 0.3983, + "step": 24679 + }, + { + "epoch": 4.49, + "learning_rate": 1.6550782672005826e-06, + "loss": 0.0103, + "step": 24680 + }, + { + "epoch": 4.49, + "learning_rate": 1.6550236621769205e-06, + "loss": 1.9697, + "step": 24681 + }, + { + "epoch": 4.49, + "learning_rate": 1.654969057153258e-06, + "loss": 2.2425, + "step": 24682 + }, + { + "epoch": 4.49, + "learning_rate": 1.654914452129596e-06, + "loss": 0.0156, + "step": 24683 + }, + { + "epoch": 4.49, + "learning_rate": 1.6548598471059337e-06, + "loss": 0.6087, + "step": 24684 + }, + { + "epoch": 4.49, + "learning_rate": 1.6548052420822717e-06, + "loss": 0.0369, + "step": 24685 + }, + { + "epoch": 4.49, + "learning_rate": 1.6547506370586096e-06, + "loss": 1.3644, + "step": 24686 + }, + { + "epoch": 4.49, + "learning_rate": 1.6546960320349471e-06, + "loss": 0.4339, + "step": 24687 + }, + { + "epoch": 4.49, + "learning_rate": 1.654641427011285e-06, + "loss": 0.1995, + "step": 24688 + }, + { + "epoch": 4.49, + "learning_rate": 1.654586821987623e-06, + "loss": 0.137, + "step": 24689 + }, + { + "epoch": 4.49, + "learning_rate": 1.6545322169639605e-06, + "loss": 0.8293, + "step": 24690 + }, + { + "epoch": 4.49, + "learning_rate": 1.6544776119402985e-06, + "loss": 0.5219, + "step": 24691 + }, + { + "epoch": 4.49, + "learning_rate": 1.6544230069166364e-06, + "loss": 2.0181, + "step": 24692 + }, + { + "epoch": 4.49, + "learning_rate": 1.6543684018929741e-06, + "loss": 0.2512, + "step": 24693 + }, + { + "epoch": 4.49, + "learning_rate": 1.654313796869312e-06, + "loss": 0.5565, + "step": 24694 + }, + { + "epoch": 4.49, + "learning_rate": 1.65425919184565e-06, + "loss": 0.3724, + "step": 24695 + }, + { + "epoch": 4.49, + "learning_rate": 1.6542045868219875e-06, + "loss": 0.272, + "step": 24696 + }, + { + "epoch": 4.49, + "learning_rate": 1.6541499817983255e-06, + "loss": 0.3874, + "step": 24697 + }, + { + "epoch": 4.49, + "learning_rate": 1.6540953767746632e-06, + "loss": 0.6163, + "step": 24698 + }, + { + "epoch": 4.49, + "learning_rate": 1.6540407717510012e-06, + "loss": 0.2225, + "step": 24699 + }, + { + "epoch": 4.49, + "learning_rate": 1.6539861667273391e-06, + "loss": 0.2817, + "step": 24700 + }, + { + "epoch": 4.49, + "learning_rate": 1.6539315617036766e-06, + "loss": 1.0222, + "step": 24701 + }, + { + "epoch": 4.49, + "learning_rate": 1.6538769566800146e-06, + "loss": 0.3206, + "step": 24702 + }, + { + "epoch": 4.49, + "learning_rate": 1.6538223516563525e-06, + "loss": 1.4443, + "step": 24703 + }, + { + "epoch": 4.49, + "learning_rate": 1.6537677466326902e-06, + "loss": 0.0373, + "step": 24704 + }, + { + "epoch": 4.49, + "learning_rate": 1.6537131416090282e-06, + "loss": 0.4674, + "step": 24705 + }, + { + "epoch": 4.49, + "learning_rate": 1.6536585365853661e-06, + "loss": 0.1192, + "step": 24706 + }, + { + "epoch": 4.49, + "learning_rate": 1.6536039315617037e-06, + "loss": 0.4785, + "step": 24707 + }, + { + "epoch": 4.49, + "learning_rate": 1.6535493265380416e-06, + "loss": 0.0418, + "step": 24708 + }, + { + "epoch": 4.49, + "learning_rate": 1.6534947215143795e-06, + "loss": 0.5483, + "step": 24709 + }, + { + "epoch": 4.49, + "learning_rate": 1.653440116490717e-06, + "loss": 4.9378, + "step": 24710 + }, + { + "epoch": 4.49, + "learning_rate": 1.653385511467055e-06, + "loss": 0.5173, + "step": 24711 + }, + { + "epoch": 4.49, + "learning_rate": 1.6533309064433927e-06, + "loss": 0.1369, + "step": 24712 + }, + { + "epoch": 4.49, + "learning_rate": 1.6532763014197307e-06, + "loss": 0.0484, + "step": 24713 + }, + { + "epoch": 4.49, + "learning_rate": 1.6532216963960686e-06, + "loss": 0.5186, + "step": 24714 + }, + { + "epoch": 4.49, + "learning_rate": 1.6531670913724061e-06, + "loss": 0.0133, + "step": 24715 + }, + { + "epoch": 4.49, + "learning_rate": 1.653112486348744e-06, + "loss": 2.2313, + "step": 24716 + }, + { + "epoch": 4.49, + "learning_rate": 1.653057881325082e-06, + "loss": 0.4247, + "step": 24717 + }, + { + "epoch": 4.49, + "learning_rate": 1.6530032763014198e-06, + "loss": 1.5265, + "step": 24718 + }, + { + "epoch": 4.5, + "learning_rate": 1.6529486712777577e-06, + "loss": 0.7317, + "step": 24719 + }, + { + "epoch": 4.5, + "learning_rate": 1.6528940662540956e-06, + "loss": 0.0915, + "step": 24720 + }, + { + "epoch": 4.5, + "learning_rate": 1.6528394612304332e-06, + "loss": 5.4419, + "step": 24721 + }, + { + "epoch": 4.5, + "learning_rate": 1.6527848562067711e-06, + "loss": 0.0188, + "step": 24722 + }, + { + "epoch": 4.5, + "learning_rate": 1.652730251183109e-06, + "loss": 0.0239, + "step": 24723 + }, + { + "epoch": 4.5, + "learning_rate": 1.6526756461594466e-06, + "loss": 0.0371, + "step": 24724 + }, + { + "epoch": 4.5, + "learning_rate": 1.6526210411357845e-06, + "loss": 0.0168, + "step": 24725 + }, + { + "epoch": 4.5, + "learning_rate": 1.6525664361121223e-06, + "loss": 0.6704, + "step": 24726 + }, + { + "epoch": 4.5, + "learning_rate": 1.6525118310884602e-06, + "loss": 2.0503, + "step": 24727 + }, + { + "epoch": 4.5, + "learning_rate": 1.6524572260647981e-06, + "loss": 0.0085, + "step": 24728 + }, + { + "epoch": 4.5, + "learning_rate": 1.6524026210411357e-06, + "loss": 0.949, + "step": 24729 + }, + { + "epoch": 4.5, + "learning_rate": 1.6523480160174736e-06, + "loss": 1.2311, + "step": 24730 + }, + { + "epoch": 4.5, + "learning_rate": 1.6522934109938115e-06, + "loss": 0.5546, + "step": 24731 + }, + { + "epoch": 4.5, + "learning_rate": 1.6522388059701493e-06, + "loss": 0.3004, + "step": 24732 + }, + { + "epoch": 4.5, + "learning_rate": 1.6521842009464872e-06, + "loss": 1.7059, + "step": 24733 + }, + { + "epoch": 4.5, + "learning_rate": 1.6521295959228252e-06, + "loss": 0.1837, + "step": 24734 + }, + { + "epoch": 4.5, + "learning_rate": 1.6520749908991627e-06, + "loss": 0.6059, + "step": 24735 + }, + { + "epoch": 4.5, + "learning_rate": 1.6520203858755006e-06, + "loss": 1.0385, + "step": 24736 + }, + { + "epoch": 4.5, + "learning_rate": 1.6519657808518381e-06, + "loss": 0.1907, + "step": 24737 + }, + { + "epoch": 4.5, + "learning_rate": 1.651911175828176e-06, + "loss": 0.0201, + "step": 24738 + }, + { + "epoch": 4.5, + "learning_rate": 1.651856570804514e-06, + "loss": 1.1095, + "step": 24739 + }, + { + "epoch": 4.5, + "learning_rate": 1.6518019657808518e-06, + "loss": 0.3115, + "step": 24740 + }, + { + "epoch": 4.5, + "learning_rate": 1.6517473607571897e-06, + "loss": 0.0218, + "step": 24741 + }, + { + "epoch": 4.5, + "learning_rate": 1.6516927557335277e-06, + "loss": 0.4563, + "step": 24742 + }, + { + "epoch": 4.5, + "learning_rate": 1.6516381507098652e-06, + "loss": 0.308, + "step": 24743 + }, + { + "epoch": 4.5, + "learning_rate": 1.6515835456862031e-06, + "loss": 1.5933, + "step": 24744 + }, + { + "epoch": 4.5, + "learning_rate": 1.651528940662541e-06, + "loss": 0.2877, + "step": 24745 + }, + { + "epoch": 4.5, + "learning_rate": 1.6514743356388788e-06, + "loss": 0.2637, + "step": 24746 + }, + { + "epoch": 4.5, + "learning_rate": 1.6514197306152167e-06, + "loss": 3.9602, + "step": 24747 + }, + { + "epoch": 4.5, + "learning_rate": 1.6513651255915547e-06, + "loss": 1.0019, + "step": 24748 + }, + { + "epoch": 4.5, + "learning_rate": 1.6513105205678922e-06, + "loss": 1.0291, + "step": 24749 + }, + { + "epoch": 4.5, + "learning_rate": 1.6512559155442301e-06, + "loss": 0.0266, + "step": 24750 + }, + { + "epoch": 4.5, + "learning_rate": 1.6512013105205679e-06, + "loss": 0.0471, + "step": 24751 + }, + { + "epoch": 4.5, + "learning_rate": 1.6511467054969058e-06, + "loss": 0.2966, + "step": 24752 + }, + { + "epoch": 4.5, + "learning_rate": 1.6510921004732438e-06, + "loss": 0.1152, + "step": 24753 + }, + { + "epoch": 4.5, + "learning_rate": 1.6510374954495813e-06, + "loss": 0.0979, + "step": 24754 + }, + { + "epoch": 4.5, + "learning_rate": 1.6509828904259192e-06, + "loss": 0.1311, + "step": 24755 + }, + { + "epoch": 4.5, + "learning_rate": 1.6509282854022572e-06, + "loss": 1.4528, + "step": 24756 + }, + { + "epoch": 4.5, + "learning_rate": 1.6508736803785947e-06, + "loss": 0.1888, + "step": 24757 + }, + { + "epoch": 4.5, + "learning_rate": 1.6508190753549326e-06, + "loss": 0.3803, + "step": 24758 + }, + { + "epoch": 4.5, + "learning_rate": 1.6507644703312706e-06, + "loss": 0.5413, + "step": 24759 + }, + { + "epoch": 4.5, + "learning_rate": 1.6507098653076083e-06, + "loss": 0.5077, + "step": 24760 + }, + { + "epoch": 4.5, + "learning_rate": 1.6506552602839462e-06, + "loss": 0.0566, + "step": 24761 + }, + { + "epoch": 4.5, + "learning_rate": 1.6506006552602842e-06, + "loss": 0.0122, + "step": 24762 + }, + { + "epoch": 4.5, + "learning_rate": 1.6505460502366217e-06, + "loss": 0.9261, + "step": 24763 + }, + { + "epoch": 4.5, + "learning_rate": 1.6504914452129597e-06, + "loss": 0.2589, + "step": 24764 + }, + { + "epoch": 4.5, + "learning_rate": 1.6504368401892974e-06, + "loss": 0.4848, + "step": 24765 + }, + { + "epoch": 4.5, + "learning_rate": 1.6503822351656353e-06, + "loss": 1.5357, + "step": 24766 + }, + { + "epoch": 4.5, + "learning_rate": 1.6503276301419733e-06, + "loss": 0.2563, + "step": 24767 + }, + { + "epoch": 4.5, + "learning_rate": 1.6502730251183108e-06, + "loss": 1.5381, + "step": 24768 + }, + { + "epoch": 4.5, + "learning_rate": 1.6502184200946487e-06, + "loss": 1.5886, + "step": 24769 + }, + { + "epoch": 4.5, + "learning_rate": 1.6501638150709867e-06, + "loss": 1.0668, + "step": 24770 + }, + { + "epoch": 4.5, + "learning_rate": 1.6501092100473242e-06, + "loss": 1.193, + "step": 24771 + }, + { + "epoch": 4.5, + "learning_rate": 1.6500546050236621e-06, + "loss": 1.1838, + "step": 24772 + }, + { + "epoch": 4.51, + "learning_rate": 1.65e-06, + "loss": 3.1815, + "step": 24773 + }, + { + "epoch": 4.51, + "learning_rate": 1.6499453949763378e-06, + "loss": 0.3377, + "step": 24774 + }, + { + "epoch": 4.51, + "learning_rate": 1.6498907899526758e-06, + "loss": 0.1894, + "step": 24775 + }, + { + "epoch": 4.51, + "learning_rate": 1.6498361849290137e-06, + "loss": 0.0332, + "step": 24776 + }, + { + "epoch": 4.51, + "learning_rate": 1.6497815799053512e-06, + "loss": 0.5138, + "step": 24777 + }, + { + "epoch": 4.51, + "learning_rate": 1.6497269748816892e-06, + "loss": 0.398, + "step": 24778 + }, + { + "epoch": 4.51, + "learning_rate": 1.649672369858027e-06, + "loss": 0.1356, + "step": 24779 + }, + { + "epoch": 4.51, + "learning_rate": 1.6496177648343648e-06, + "loss": 0.0232, + "step": 24780 + }, + { + "epoch": 4.51, + "learning_rate": 1.6495631598107028e-06, + "loss": 0.1226, + "step": 24781 + }, + { + "epoch": 4.51, + "learning_rate": 1.6495085547870403e-06, + "loss": 0.0097, + "step": 24782 + }, + { + "epoch": 4.51, + "learning_rate": 1.6494539497633782e-06, + "loss": 1.8483, + "step": 24783 + }, + { + "epoch": 4.51, + "learning_rate": 1.6493993447397162e-06, + "loss": 0.0509, + "step": 24784 + }, + { + "epoch": 4.51, + "learning_rate": 1.649344739716054e-06, + "loss": 0.561, + "step": 24785 + }, + { + "epoch": 4.51, + "learning_rate": 1.6492901346923917e-06, + "loss": 1.7895, + "step": 24786 + }, + { + "epoch": 4.51, + "learning_rate": 1.6492355296687296e-06, + "loss": 0.6048, + "step": 24787 + }, + { + "epoch": 4.51, + "learning_rate": 1.6491809246450673e-06, + "loss": 1.7914, + "step": 24788 + }, + { + "epoch": 4.51, + "learning_rate": 1.6491263196214053e-06, + "loss": 3.1433, + "step": 24789 + }, + { + "epoch": 4.51, + "learning_rate": 1.6490717145977432e-06, + "loss": 0.0341, + "step": 24790 + }, + { + "epoch": 4.51, + "learning_rate": 1.6490171095740807e-06, + "loss": 1.7491, + "step": 24791 + }, + { + "epoch": 4.51, + "learning_rate": 1.6489625045504187e-06, + "loss": 2.476, + "step": 24792 + }, + { + "epoch": 4.51, + "learning_rate": 1.6489078995267564e-06, + "loss": 0.0279, + "step": 24793 + }, + { + "epoch": 4.51, + "learning_rate": 1.6488532945030944e-06, + "loss": 0.3977, + "step": 24794 + }, + { + "epoch": 4.51, + "learning_rate": 1.6487986894794323e-06, + "loss": 0.3316, + "step": 24795 + }, + { + "epoch": 4.51, + "learning_rate": 1.6487440844557698e-06, + "loss": 0.0129, + "step": 24796 + }, + { + "epoch": 4.51, + "learning_rate": 1.6486894794321078e-06, + "loss": 0.5764, + "step": 24797 + }, + { + "epoch": 4.51, + "learning_rate": 1.6486348744084457e-06, + "loss": 0.3445, + "step": 24798 + }, + { + "epoch": 4.51, + "learning_rate": 1.6485802693847834e-06, + "loss": 0.1971, + "step": 24799 + }, + { + "epoch": 4.51, + "learning_rate": 1.6485256643611214e-06, + "loss": 0.7674, + "step": 24800 + }, + { + "epoch": 4.51, + "learning_rate": 1.6484710593374593e-06, + "loss": 0.4635, + "step": 24801 + }, + { + "epoch": 4.51, + "learning_rate": 1.6484164543137968e-06, + "loss": 0.9644, + "step": 24802 + }, + { + "epoch": 4.51, + "learning_rate": 1.6483618492901348e-06, + "loss": 1.0457, + "step": 24803 + }, + { + "epoch": 4.51, + "learning_rate": 1.6483072442664727e-06, + "loss": 0.5066, + "step": 24804 + }, + { + "epoch": 4.51, + "learning_rate": 1.6482526392428102e-06, + "loss": 0.0133, + "step": 24805 + }, + { + "epoch": 4.51, + "learning_rate": 1.6481980342191482e-06, + "loss": 1.4773, + "step": 24806 + }, + { + "epoch": 4.51, + "learning_rate": 1.648143429195486e-06, + "loss": 0.6969, + "step": 24807 + }, + { + "epoch": 4.51, + "learning_rate": 1.6480888241718239e-06, + "loss": 0.3966, + "step": 24808 + }, + { + "epoch": 4.51, + "learning_rate": 1.6480342191481618e-06, + "loss": 0.5488, + "step": 24809 + }, + { + "epoch": 4.51, + "learning_rate": 1.6479796141244993e-06, + "loss": 0.2047, + "step": 24810 + }, + { + "epoch": 4.51, + "learning_rate": 1.6479250091008373e-06, + "loss": 1.6879, + "step": 24811 + }, + { + "epoch": 4.51, + "learning_rate": 1.6478704040771752e-06, + "loss": 1.5554, + "step": 24812 + }, + { + "epoch": 4.51, + "learning_rate": 1.647815799053513e-06, + "loss": 1.7577, + "step": 24813 + }, + { + "epoch": 4.51, + "learning_rate": 1.6477611940298509e-06, + "loss": 0.8401, + "step": 24814 + }, + { + "epoch": 4.51, + "learning_rate": 1.6477065890061888e-06, + "loss": 1.8369, + "step": 24815 + }, + { + "epoch": 4.51, + "learning_rate": 1.6476519839825264e-06, + "loss": 0.0783, + "step": 24816 + }, + { + "epoch": 4.51, + "learning_rate": 1.6475973789588643e-06, + "loss": 0.0792, + "step": 24817 + }, + { + "epoch": 4.51, + "learning_rate": 1.6475427739352018e-06, + "loss": 0.5257, + "step": 24818 + }, + { + "epoch": 4.51, + "learning_rate": 1.6474881689115398e-06, + "loss": 1.1766, + "step": 24819 + }, + { + "epoch": 4.51, + "learning_rate": 1.6474335638878777e-06, + "loss": 0.1355, + "step": 24820 + }, + { + "epoch": 4.51, + "learning_rate": 1.6473789588642154e-06, + "loss": 0.1979, + "step": 24821 + }, + { + "epoch": 4.51, + "learning_rate": 1.6473243538405534e-06, + "loss": 0.1858, + "step": 24822 + }, + { + "epoch": 4.51, + "learning_rate": 1.6472697488168913e-06, + "loss": 0.3704, + "step": 24823 + }, + { + "epoch": 4.51, + "learning_rate": 1.6472151437932288e-06, + "loss": 0.3394, + "step": 24824 + }, + { + "epoch": 4.51, + "learning_rate": 1.6471605387695668e-06, + "loss": 0.251, + "step": 24825 + }, + { + "epoch": 4.51, + "learning_rate": 1.6471059337459047e-06, + "loss": 0.212, + "step": 24826 + }, + { + "epoch": 4.51, + "learning_rate": 1.6470513287222425e-06, + "loss": 0.9637, + "step": 24827 + }, + { + "epoch": 4.52, + "learning_rate": 1.6469967236985804e-06, + "loss": 0.1026, + "step": 24828 + }, + { + "epoch": 4.52, + "learning_rate": 1.6469421186749183e-06, + "loss": 0.3311, + "step": 24829 + }, + { + "epoch": 4.52, + "learning_rate": 1.6468875136512559e-06, + "loss": 3.0906, + "step": 24830 + }, + { + "epoch": 4.52, + "learning_rate": 1.6468329086275938e-06, + "loss": 0.2724, + "step": 24831 + }, + { + "epoch": 4.52, + "learning_rate": 1.6467783036039315e-06, + "loss": 0.468, + "step": 24832 + }, + { + "epoch": 4.52, + "learning_rate": 1.6467236985802695e-06, + "loss": 1.5245, + "step": 24833 + }, + { + "epoch": 4.52, + "learning_rate": 1.6466690935566074e-06, + "loss": 2.7362, + "step": 24834 + }, + { + "epoch": 4.52, + "learning_rate": 1.646614488532945e-06, + "loss": 0.0242, + "step": 24835 + }, + { + "epoch": 4.52, + "learning_rate": 1.646559883509283e-06, + "loss": 0.0037, + "step": 24836 + }, + { + "epoch": 4.52, + "learning_rate": 1.6465052784856208e-06, + "loss": 0.3984, + "step": 24837 + }, + { + "epoch": 4.52, + "learning_rate": 1.6464506734619584e-06, + "loss": 3.1777, + "step": 24838 + }, + { + "epoch": 4.52, + "learning_rate": 1.6463960684382963e-06, + "loss": 0.5794, + "step": 24839 + }, + { + "epoch": 4.52, + "learning_rate": 1.6463414634146342e-06, + "loss": 1.4661, + "step": 24840 + }, + { + "epoch": 4.52, + "learning_rate": 1.646286858390972e-06, + "loss": 0.4006, + "step": 24841 + }, + { + "epoch": 4.52, + "learning_rate": 1.64623225336731e-06, + "loss": 0.0116, + "step": 24842 + }, + { + "epoch": 4.52, + "learning_rate": 1.6461776483436479e-06, + "loss": 2.4789, + "step": 24843 + }, + { + "epoch": 4.52, + "learning_rate": 1.6461230433199854e-06, + "loss": 0.285, + "step": 24844 + }, + { + "epoch": 4.52, + "learning_rate": 1.6460684382963233e-06, + "loss": 2.0489, + "step": 24845 + }, + { + "epoch": 4.52, + "learning_rate": 1.646013833272661e-06, + "loss": 0.3337, + "step": 24846 + }, + { + "epoch": 4.52, + "learning_rate": 1.645959228248999e-06, + "loss": 1.915, + "step": 24847 + }, + { + "epoch": 4.52, + "learning_rate": 1.645904623225337e-06, + "loss": 2.1905, + "step": 24848 + }, + { + "epoch": 4.52, + "learning_rate": 1.6458500182016745e-06, + "loss": 0.0072, + "step": 24849 + }, + { + "epoch": 4.52, + "learning_rate": 1.6457954131780124e-06, + "loss": 0.0373, + "step": 24850 + }, + { + "epoch": 4.52, + "learning_rate": 1.6457408081543504e-06, + "loss": 2.5029, + "step": 24851 + }, + { + "epoch": 4.52, + "learning_rate": 1.6456862031306879e-06, + "loss": 1.2961, + "step": 24852 + }, + { + "epoch": 4.52, + "learning_rate": 1.6456315981070258e-06, + "loss": 1.3726, + "step": 24853 + }, + { + "epoch": 4.52, + "learning_rate": 1.6455769930833638e-06, + "loss": 0.2296, + "step": 24854 + }, + { + "epoch": 4.52, + "learning_rate": 1.6455223880597015e-06, + "loss": 0.3091, + "step": 24855 + }, + { + "epoch": 4.52, + "learning_rate": 1.6454677830360394e-06, + "loss": 1.254, + "step": 24856 + }, + { + "epoch": 4.52, + "learning_rate": 1.6454131780123774e-06, + "loss": 1.1966, + "step": 24857 + }, + { + "epoch": 4.52, + "learning_rate": 1.645358572988715e-06, + "loss": 0.1923, + "step": 24858 + }, + { + "epoch": 4.52, + "learning_rate": 1.6453039679650528e-06, + "loss": 2.1324, + "step": 24859 + }, + { + "epoch": 4.52, + "learning_rate": 1.6452493629413906e-06, + "loss": 2.6628, + "step": 24860 + }, + { + "epoch": 4.52, + "learning_rate": 1.6451947579177285e-06, + "loss": 0.7693, + "step": 24861 + }, + { + "epoch": 4.52, + "learning_rate": 1.6451401528940665e-06, + "loss": 0.2257, + "step": 24862 + }, + { + "epoch": 4.52, + "learning_rate": 1.645085547870404e-06, + "loss": 2.9681, + "step": 24863 + }, + { + "epoch": 4.52, + "learning_rate": 1.645030942846742e-06, + "loss": 2.4355, + "step": 24864 + }, + { + "epoch": 4.52, + "learning_rate": 1.6449763378230799e-06, + "loss": 0.8255, + "step": 24865 + }, + { + "epoch": 4.52, + "learning_rate": 1.6449217327994174e-06, + "loss": 0.1703, + "step": 24866 + }, + { + "epoch": 4.52, + "learning_rate": 1.6448671277757553e-06, + "loss": 0.2251, + "step": 24867 + }, + { + "epoch": 4.52, + "learning_rate": 1.6448125227520933e-06, + "loss": 1.5193, + "step": 24868 + }, + { + "epoch": 4.52, + "learning_rate": 1.644757917728431e-06, + "loss": 2.4681, + "step": 24869 + }, + { + "epoch": 4.52, + "learning_rate": 1.644703312704769e-06, + "loss": 0.376, + "step": 24870 + }, + { + "epoch": 4.52, + "learning_rate": 1.6446487076811069e-06, + "loss": 0.0825, + "step": 24871 + }, + { + "epoch": 4.52, + "learning_rate": 1.6445941026574444e-06, + "loss": 1.3464, + "step": 24872 + }, + { + "epoch": 4.52, + "learning_rate": 1.6445394976337824e-06, + "loss": 0.6203, + "step": 24873 + }, + { + "epoch": 4.52, + "learning_rate": 1.64448489261012e-06, + "loss": 0.1863, + "step": 24874 + }, + { + "epoch": 4.52, + "learning_rate": 1.644430287586458e-06, + "loss": 0.4314, + "step": 24875 + }, + { + "epoch": 4.52, + "learning_rate": 1.644375682562796e-06, + "loss": 0.0317, + "step": 24876 + }, + { + "epoch": 4.52, + "learning_rate": 1.6443210775391335e-06, + "loss": 0.501, + "step": 24877 + }, + { + "epoch": 4.52, + "learning_rate": 1.6442664725154714e-06, + "loss": 1.0548, + "step": 24878 + }, + { + "epoch": 4.52, + "learning_rate": 1.6442118674918094e-06, + "loss": 0.0725, + "step": 24879 + }, + { + "epoch": 4.52, + "learning_rate": 1.6441572624681471e-06, + "loss": 0.9191, + "step": 24880 + }, + { + "epoch": 4.52, + "learning_rate": 1.644102657444485e-06, + "loss": 0.0097, + "step": 24881 + }, + { + "epoch": 4.52, + "learning_rate": 1.644048052420823e-06, + "loss": 0.0081, + "step": 24882 + }, + { + "epoch": 4.53, + "learning_rate": 1.6439934473971605e-06, + "loss": 1.0291, + "step": 24883 + }, + { + "epoch": 4.53, + "learning_rate": 1.6439388423734985e-06, + "loss": 2.8799, + "step": 24884 + }, + { + "epoch": 4.53, + "learning_rate": 1.643884237349836e-06, + "loss": 1.463, + "step": 24885 + }, + { + "epoch": 4.53, + "learning_rate": 1.643829632326174e-06, + "loss": 1.1564, + "step": 24886 + }, + { + "epoch": 4.53, + "learning_rate": 1.6437750273025119e-06, + "loss": 0.3092, + "step": 24887 + }, + { + "epoch": 4.53, + "learning_rate": 1.6437204222788496e-06, + "loss": 0.0116, + "step": 24888 + }, + { + "epoch": 4.53, + "learning_rate": 1.6436658172551875e-06, + "loss": 0.8508, + "step": 24889 + }, + { + "epoch": 4.53, + "learning_rate": 1.6436112122315255e-06, + "loss": 0.2398, + "step": 24890 + }, + { + "epoch": 4.53, + "learning_rate": 1.643556607207863e-06, + "loss": 0.019, + "step": 24891 + }, + { + "epoch": 4.53, + "learning_rate": 1.643502002184201e-06, + "loss": 0.1706, + "step": 24892 + }, + { + "epoch": 4.53, + "learning_rate": 1.6434473971605389e-06, + "loss": 0.6561, + "step": 24893 + }, + { + "epoch": 4.53, + "learning_rate": 1.6433927921368766e-06, + "loss": 0.1724, + "step": 24894 + }, + { + "epoch": 4.53, + "learning_rate": 1.6433381871132146e-06, + "loss": 0.2633, + "step": 24895 + }, + { + "epoch": 4.53, + "learning_rate": 1.6432835820895525e-06, + "loss": 0.4705, + "step": 24896 + }, + { + "epoch": 4.53, + "learning_rate": 1.64322897706589e-06, + "loss": 0.1112, + "step": 24897 + }, + { + "epoch": 4.53, + "learning_rate": 1.643174372042228e-06, + "loss": 0.3164, + "step": 24898 + }, + { + "epoch": 4.53, + "learning_rate": 1.6431197670185655e-06, + "loss": 0.2344, + "step": 24899 + }, + { + "epoch": 4.53, + "learning_rate": 1.6430651619949034e-06, + "loss": 0.1839, + "step": 24900 + }, + { + "epoch": 4.53, + "learning_rate": 1.6430105569712414e-06, + "loss": 0.0495, + "step": 24901 + }, + { + "epoch": 4.53, + "learning_rate": 1.6429559519475791e-06, + "loss": 0.3428, + "step": 24902 + }, + { + "epoch": 4.53, + "learning_rate": 1.642901346923917e-06, + "loss": 0.1026, + "step": 24903 + }, + { + "epoch": 4.53, + "learning_rate": 1.642846741900255e-06, + "loss": 0.7485, + "step": 24904 + }, + { + "epoch": 4.53, + "learning_rate": 1.6427921368765925e-06, + "loss": 0.4862, + "step": 24905 + }, + { + "epoch": 4.53, + "learning_rate": 1.6427375318529305e-06, + "loss": 2.398, + "step": 24906 + }, + { + "epoch": 4.53, + "learning_rate": 1.6426829268292684e-06, + "loss": 0.0481, + "step": 24907 + }, + { + "epoch": 4.53, + "learning_rate": 1.6426283218056061e-06, + "loss": 0.2126, + "step": 24908 + }, + { + "epoch": 4.53, + "learning_rate": 1.642573716781944e-06, + "loss": 5.0694, + "step": 24909 + }, + { + "epoch": 4.53, + "learning_rate": 1.642519111758282e-06, + "loss": 0.4189, + "step": 24910 + }, + { + "epoch": 4.53, + "learning_rate": 1.6424645067346195e-06, + "loss": 0.7237, + "step": 24911 + }, + { + "epoch": 4.53, + "learning_rate": 1.6424099017109575e-06, + "loss": 2.0616, + "step": 24912 + }, + { + "epoch": 4.53, + "learning_rate": 1.6423552966872952e-06, + "loss": 1.5162, + "step": 24913 + }, + { + "epoch": 4.53, + "learning_rate": 1.6423006916636332e-06, + "loss": 0.7706, + "step": 24914 + }, + { + "epoch": 4.53, + "learning_rate": 1.642246086639971e-06, + "loss": 0.3737, + "step": 24915 + }, + { + "epoch": 4.53, + "learning_rate": 1.6421914816163086e-06, + "loss": 0.2495, + "step": 24916 + }, + { + "epoch": 4.53, + "learning_rate": 1.6421368765926466e-06, + "loss": 0.6718, + "step": 24917 + }, + { + "epoch": 4.53, + "learning_rate": 1.6420822715689845e-06, + "loss": 0.2205, + "step": 24918 + }, + { + "epoch": 4.53, + "learning_rate": 1.642027666545322e-06, + "loss": 2.5062, + "step": 24919 + }, + { + "epoch": 4.53, + "learning_rate": 1.64197306152166e-06, + "loss": 0.7421, + "step": 24920 + }, + { + "epoch": 4.53, + "learning_rate": 1.641918456497998e-06, + "loss": 1.1862, + "step": 24921 + }, + { + "epoch": 4.53, + "learning_rate": 1.6418638514743357e-06, + "loss": 3.495, + "step": 24922 + }, + { + "epoch": 4.53, + "learning_rate": 1.6418092464506736e-06, + "loss": 0.7436, + "step": 24923 + }, + { + "epoch": 4.53, + "learning_rate": 1.6417546414270115e-06, + "loss": 1.1999, + "step": 24924 + }, + { + "epoch": 4.53, + "learning_rate": 1.641700036403349e-06, + "loss": 0.0247, + "step": 24925 + }, + { + "epoch": 4.53, + "learning_rate": 1.641645431379687e-06, + "loss": 0.0385, + "step": 24926 + }, + { + "epoch": 4.53, + "learning_rate": 1.6415908263560247e-06, + "loss": 0.1504, + "step": 24927 + }, + { + "epoch": 4.53, + "learning_rate": 1.6415362213323627e-06, + "loss": 0.5111, + "step": 24928 + }, + { + "epoch": 4.53, + "learning_rate": 1.6414816163087006e-06, + "loss": 0.4556, + "step": 24929 + }, + { + "epoch": 4.53, + "learning_rate": 1.6414270112850381e-06, + "loss": 1.1715, + "step": 24930 + }, + { + "epoch": 4.53, + "learning_rate": 1.641372406261376e-06, + "loss": 0.5673, + "step": 24931 + }, + { + "epoch": 4.53, + "learning_rate": 1.641317801237714e-06, + "loss": 1.1049, + "step": 24932 + }, + { + "epoch": 4.53, + "learning_rate": 1.6412631962140515e-06, + "loss": 0.4451, + "step": 24933 + }, + { + "epoch": 4.53, + "learning_rate": 1.6412085911903895e-06, + "loss": 0.6824, + "step": 24934 + }, + { + "epoch": 4.53, + "learning_rate": 1.6411539861667274e-06, + "loss": 0.1411, + "step": 24935 + }, + { + "epoch": 4.53, + "learning_rate": 1.6410993811430652e-06, + "loss": 0.3439, + "step": 24936 + }, + { + "epoch": 4.53, + "learning_rate": 1.6410447761194031e-06, + "loss": 0.5552, + "step": 24937 + }, + { + "epoch": 4.54, + "learning_rate": 1.640990171095741e-06, + "loss": 0.7496, + "step": 24938 + }, + { + "epoch": 4.54, + "learning_rate": 1.6409355660720786e-06, + "loss": 0.0197, + "step": 24939 + }, + { + "epoch": 4.54, + "learning_rate": 1.6408809610484165e-06, + "loss": 0.3885, + "step": 24940 + }, + { + "epoch": 4.54, + "learning_rate": 1.6408263560247542e-06, + "loss": 0.4549, + "step": 24941 + }, + { + "epoch": 4.54, + "learning_rate": 1.6407717510010922e-06, + "loss": 0.5148, + "step": 24942 + }, + { + "epoch": 4.54, + "learning_rate": 1.6407171459774301e-06, + "loss": 0.0321, + "step": 24943 + }, + { + "epoch": 4.54, + "learning_rate": 1.6406625409537677e-06, + "loss": 1.1643, + "step": 24944 + }, + { + "epoch": 4.54, + "learning_rate": 1.6406079359301056e-06, + "loss": 0.0151, + "step": 24945 + }, + { + "epoch": 4.54, + "learning_rate": 1.6405533309064435e-06, + "loss": 3.025, + "step": 24946 + }, + { + "epoch": 4.54, + "learning_rate": 1.640498725882781e-06, + "loss": 0.34, + "step": 24947 + }, + { + "epoch": 4.54, + "learning_rate": 1.640444120859119e-06, + "loss": 0.0869, + "step": 24948 + }, + { + "epoch": 4.54, + "learning_rate": 1.640389515835457e-06, + "loss": 0.3711, + "step": 24949 + }, + { + "epoch": 4.54, + "learning_rate": 1.6403349108117947e-06, + "loss": 0.0258, + "step": 24950 + }, + { + "epoch": 4.54, + "learning_rate": 1.6402803057881326e-06, + "loss": 1.0144, + "step": 24951 + }, + { + "epoch": 4.54, + "learning_rate": 1.6402257007644706e-06, + "loss": 0.7982, + "step": 24952 + }, + { + "epoch": 4.54, + "learning_rate": 1.640171095740808e-06, + "loss": 0.1653, + "step": 24953 + }, + { + "epoch": 4.54, + "learning_rate": 1.640116490717146e-06, + "loss": 0.2106, + "step": 24954 + }, + { + "epoch": 4.54, + "learning_rate": 1.6400618856934838e-06, + "loss": 0.449, + "step": 24955 + }, + { + "epoch": 4.54, + "learning_rate": 1.6400072806698217e-06, + "loss": 0.2294, + "step": 24956 + }, + { + "epoch": 4.54, + "learning_rate": 1.6399526756461596e-06, + "loss": 2.0788, + "step": 24957 + }, + { + "epoch": 4.54, + "learning_rate": 1.6398980706224972e-06, + "loss": 0.6948, + "step": 24958 + }, + { + "epoch": 4.54, + "learning_rate": 1.6398434655988351e-06, + "loss": 1.7417, + "step": 24959 + }, + { + "epoch": 4.54, + "learning_rate": 1.639788860575173e-06, + "loss": 0.4856, + "step": 24960 + }, + { + "epoch": 4.54, + "learning_rate": 1.6397342555515108e-06, + "loss": 0.0186, + "step": 24961 + }, + { + "epoch": 4.54, + "learning_rate": 1.6396796505278487e-06, + "loss": 0.9938, + "step": 24962 + }, + { + "epoch": 4.54, + "learning_rate": 1.6396250455041867e-06, + "loss": 0.14, + "step": 24963 + }, + { + "epoch": 4.54, + "learning_rate": 1.6395704404805242e-06, + "loss": 0.787, + "step": 24964 + }, + { + "epoch": 4.54, + "learning_rate": 1.6395158354568621e-06, + "loss": 2.1813, + "step": 24965 + }, + { + "epoch": 4.54, + "learning_rate": 1.6394612304331997e-06, + "loss": 0.1163, + "step": 24966 + }, + { + "epoch": 4.54, + "learning_rate": 1.6394066254095376e-06, + "loss": 0.5217, + "step": 24967 + }, + { + "epoch": 4.54, + "learning_rate": 1.6393520203858755e-06, + "loss": 0.5022, + "step": 24968 + }, + { + "epoch": 4.54, + "learning_rate": 1.6392974153622133e-06, + "loss": 0.5692, + "step": 24969 + }, + { + "epoch": 4.54, + "learning_rate": 1.6392428103385512e-06, + "loss": 1.6992, + "step": 24970 + }, + { + "epoch": 4.54, + "learning_rate": 1.6391882053148892e-06, + "loss": 0.3832, + "step": 24971 + }, + { + "epoch": 4.54, + "learning_rate": 1.6391336002912267e-06, + "loss": 0.7999, + "step": 24972 + }, + { + "epoch": 4.54, + "learning_rate": 1.6390789952675646e-06, + "loss": 0.7879, + "step": 24973 + }, + { + "epoch": 4.54, + "learning_rate": 1.6390243902439026e-06, + "loss": 0.1278, + "step": 24974 + }, + { + "epoch": 4.54, + "learning_rate": 1.6389697852202403e-06, + "loss": 0.163, + "step": 24975 + }, + { + "epoch": 4.54, + "learning_rate": 1.6389151801965782e-06, + "loss": 0.1966, + "step": 24976 + }, + { + "epoch": 4.54, + "learning_rate": 1.6388605751729162e-06, + "loss": 0.2948, + "step": 24977 + }, + { + "epoch": 4.54, + "learning_rate": 1.6388059701492537e-06, + "loss": 0.0426, + "step": 24978 + }, + { + "epoch": 4.54, + "learning_rate": 1.6387513651255916e-06, + "loss": 2.206, + "step": 24979 + }, + { + "epoch": 4.54, + "learning_rate": 1.6386967601019292e-06, + "loss": 0.2587, + "step": 24980 + }, + { + "epoch": 4.54, + "learning_rate": 1.6386421550782671e-06, + "loss": 0.3771, + "step": 24981 + }, + { + "epoch": 4.54, + "learning_rate": 1.638587550054605e-06, + "loss": 0.0091, + "step": 24982 + }, + { + "epoch": 4.54, + "learning_rate": 1.6385329450309428e-06, + "loss": 0.4577, + "step": 24983 + }, + { + "epoch": 4.54, + "learning_rate": 1.6384783400072807e-06, + "loss": 0.8031, + "step": 24984 + }, + { + "epoch": 4.54, + "learning_rate": 1.6384237349836187e-06, + "loss": 0.0793, + "step": 24985 + }, + { + "epoch": 4.54, + "learning_rate": 1.6383691299599562e-06, + "loss": 0.8249, + "step": 24986 + }, + { + "epoch": 4.54, + "learning_rate": 1.6383145249362941e-06, + "loss": 1.2187, + "step": 24987 + }, + { + "epoch": 4.54, + "learning_rate": 1.638259919912632e-06, + "loss": 0.026, + "step": 24988 + }, + { + "epoch": 4.54, + "learning_rate": 1.6382053148889698e-06, + "loss": 3.0481, + "step": 24989 + }, + { + "epoch": 4.54, + "learning_rate": 1.6381507098653078e-06, + "loss": 0.5266, + "step": 24990 + }, + { + "epoch": 4.54, + "learning_rate": 1.6380961048416457e-06, + "loss": 0.0154, + "step": 24991 + }, + { + "epoch": 4.54, + "learning_rate": 1.6380414998179832e-06, + "loss": 0.2596, + "step": 24992 + }, + { + "epoch": 4.55, + "learning_rate": 1.6379868947943212e-06, + "loss": 0.6301, + "step": 24993 + }, + { + "epoch": 4.55, + "learning_rate": 1.6379322897706587e-06, + "loss": 0.0101, + "step": 24994 + }, + { + "epoch": 4.55, + "learning_rate": 1.6378776847469966e-06, + "loss": 0.1466, + "step": 24995 + }, + { + "epoch": 4.55, + "learning_rate": 1.6378230797233346e-06, + "loss": 1.3504, + "step": 24996 + }, + { + "epoch": 4.55, + "learning_rate": 1.6377684746996723e-06, + "loss": 1.3583, + "step": 24997 + }, + { + "epoch": 4.55, + "learning_rate": 1.6377138696760102e-06, + "loss": 0.9097, + "step": 24998 + }, + { + "epoch": 4.55, + "learning_rate": 1.6376592646523482e-06, + "loss": 0.1261, + "step": 24999 + }, + { + "epoch": 4.55, + "learning_rate": 1.6376046596286857e-06, + "loss": 0.0407, + "step": 25000 + }, + { + "epoch": 4.55, + "learning_rate": 1.6375500546050236e-06, + "loss": 1.1075, + "step": 25001 + }, + { + "epoch": 4.55, + "learning_rate": 1.6374954495813616e-06, + "loss": 0.8072, + "step": 25002 + }, + { + "epoch": 4.55, + "learning_rate": 1.6374408445576993e-06, + "loss": 0.64, + "step": 25003 + }, + { + "epoch": 4.55, + "learning_rate": 1.6373862395340373e-06, + "loss": 0.0197, + "step": 25004 + }, + { + "epoch": 4.55, + "learning_rate": 1.6373316345103752e-06, + "loss": 2.4967, + "step": 25005 + }, + { + "epoch": 4.55, + "learning_rate": 1.6372770294867127e-06, + "loss": 0.5728, + "step": 25006 + }, + { + "epoch": 4.55, + "learning_rate": 1.6372224244630507e-06, + "loss": 0.2879, + "step": 25007 + }, + { + "epoch": 4.55, + "learning_rate": 1.6371678194393884e-06, + "loss": 0.0775, + "step": 25008 + }, + { + "epoch": 4.55, + "learning_rate": 1.6371132144157263e-06, + "loss": 0.3381, + "step": 25009 + }, + { + "epoch": 4.55, + "learning_rate": 1.6370586093920643e-06, + "loss": 2.2794, + "step": 25010 + }, + { + "epoch": 4.55, + "learning_rate": 1.6370040043684018e-06, + "loss": 0.8412, + "step": 25011 + }, + { + "epoch": 4.55, + "learning_rate": 1.6369493993447398e-06, + "loss": 0.3657, + "step": 25012 + }, + { + "epoch": 4.55, + "learning_rate": 1.6368947943210777e-06, + "loss": 1.4529, + "step": 25013 + }, + { + "epoch": 4.55, + "learning_rate": 1.6368401892974152e-06, + "loss": 1.8507, + "step": 25014 + }, + { + "epoch": 4.55, + "learning_rate": 1.6367855842737532e-06, + "loss": 0.6889, + "step": 25015 + }, + { + "epoch": 4.55, + "learning_rate": 1.6367309792500911e-06, + "loss": 1.08, + "step": 25016 + }, + { + "epoch": 4.55, + "learning_rate": 1.6366763742264288e-06, + "loss": 0.451, + "step": 25017 + }, + { + "epoch": 4.55, + "learning_rate": 1.6366217692027668e-06, + "loss": 0.049, + "step": 25018 + }, + { + "epoch": 4.55, + "learning_rate": 1.6365671641791047e-06, + "loss": 0.6229, + "step": 25019 + }, + { + "epoch": 4.55, + "learning_rate": 1.6365125591554422e-06, + "loss": 1.6135, + "step": 25020 + }, + { + "epoch": 4.55, + "learning_rate": 1.6364579541317802e-06, + "loss": 1.2311, + "step": 25021 + }, + { + "epoch": 4.55, + "learning_rate": 1.636403349108118e-06, + "loss": 0.6914, + "step": 25022 + }, + { + "epoch": 4.55, + "learning_rate": 1.6363487440844559e-06, + "loss": 0.0457, + "step": 25023 + }, + { + "epoch": 4.55, + "learning_rate": 1.6362941390607938e-06, + "loss": 1.65, + "step": 25024 + }, + { + "epoch": 4.55, + "learning_rate": 1.6362395340371313e-06, + "loss": 0.581, + "step": 25025 + }, + { + "epoch": 4.55, + "learning_rate": 1.6361849290134693e-06, + "loss": 0.152, + "step": 25026 + }, + { + "epoch": 4.55, + "learning_rate": 1.6361303239898072e-06, + "loss": 1.0959, + "step": 25027 + }, + { + "epoch": 4.55, + "learning_rate": 1.6360757189661447e-06, + "loss": 0.9779, + "step": 25028 + }, + { + "epoch": 4.55, + "learning_rate": 1.6360211139424827e-06, + "loss": 0.0243, + "step": 25029 + }, + { + "epoch": 4.55, + "learning_rate": 1.6359665089188206e-06, + "loss": 0.0708, + "step": 25030 + }, + { + "epoch": 4.55, + "learning_rate": 1.6359119038951584e-06, + "loss": 3.0984, + "step": 25031 + }, + { + "epoch": 4.55, + "learning_rate": 1.6358572988714963e-06, + "loss": 0.8846, + "step": 25032 + }, + { + "epoch": 4.55, + "learning_rate": 1.6358026938478342e-06, + "loss": 0.0301, + "step": 25033 + }, + { + "epoch": 4.55, + "learning_rate": 1.6357480888241718e-06, + "loss": 1.7069, + "step": 25034 + }, + { + "epoch": 4.55, + "learning_rate": 1.6356934838005097e-06, + "loss": 0.6155, + "step": 25035 + }, + { + "epoch": 4.55, + "learning_rate": 1.6356388787768474e-06, + "loss": 0.2804, + "step": 25036 + }, + { + "epoch": 4.55, + "learning_rate": 1.6355842737531854e-06, + "loss": 1.7759, + "step": 25037 + }, + { + "epoch": 4.55, + "learning_rate": 1.6355296687295233e-06, + "loss": 0.1381, + "step": 25038 + }, + { + "epoch": 4.55, + "learning_rate": 1.6354750637058608e-06, + "loss": 1.1878, + "step": 25039 + }, + { + "epoch": 4.55, + "learning_rate": 1.6354204586821988e-06, + "loss": 0.463, + "step": 25040 + }, + { + "epoch": 4.55, + "learning_rate": 1.6353658536585367e-06, + "loss": 0.6403, + "step": 25041 + }, + { + "epoch": 4.55, + "learning_rate": 1.6353112486348745e-06, + "loss": 0.0159, + "step": 25042 + }, + { + "epoch": 4.55, + "learning_rate": 1.6352566436112124e-06, + "loss": 0.0169, + "step": 25043 + }, + { + "epoch": 4.55, + "learning_rate": 1.6352020385875503e-06, + "loss": 1.1104, + "step": 25044 + }, + { + "epoch": 4.55, + "learning_rate": 1.6351474335638879e-06, + "loss": 0.5403, + "step": 25045 + }, + { + "epoch": 4.55, + "learning_rate": 1.6350928285402258e-06, + "loss": 1.0303, + "step": 25046 + }, + { + "epoch": 4.55, + "learning_rate": 1.6350382235165633e-06, + "loss": 0.8111, + "step": 25047 + }, + { + "epoch": 4.56, + "learning_rate": 1.6349836184929013e-06, + "loss": 1.6314, + "step": 25048 + }, + { + "epoch": 4.56, + "learning_rate": 1.6349290134692392e-06, + "loss": 1.5482, + "step": 25049 + }, + { + "epoch": 4.56, + "learning_rate": 1.634874408445577e-06, + "loss": 0.0845, + "step": 25050 + }, + { + "epoch": 4.56, + "learning_rate": 1.6348198034219149e-06, + "loss": 0.4722, + "step": 25051 + }, + { + "epoch": 4.56, + "learning_rate": 1.6347651983982528e-06, + "loss": 0.3575, + "step": 25052 + }, + { + "epoch": 4.56, + "learning_rate": 1.6347105933745904e-06, + "loss": 0.2269, + "step": 25053 + }, + { + "epoch": 4.56, + "learning_rate": 1.6346559883509283e-06, + "loss": 1.2714, + "step": 25054 + }, + { + "epoch": 4.56, + "learning_rate": 1.6346013833272662e-06, + "loss": 0.2334, + "step": 25055 + }, + { + "epoch": 4.56, + "learning_rate": 1.634546778303604e-06, + "loss": 1.6252, + "step": 25056 + }, + { + "epoch": 4.56, + "learning_rate": 1.634492173279942e-06, + "loss": 1.5104, + "step": 25057 + }, + { + "epoch": 4.56, + "learning_rate": 1.6344375682562799e-06, + "loss": 0.0281, + "step": 25058 + }, + { + "epoch": 4.56, + "learning_rate": 1.6343829632326174e-06, + "loss": 0.5266, + "step": 25059 + }, + { + "epoch": 4.56, + "learning_rate": 1.6343283582089553e-06, + "loss": 0.0323, + "step": 25060 + }, + { + "epoch": 4.56, + "learning_rate": 1.6342737531852928e-06, + "loss": 2.0896, + "step": 25061 + }, + { + "epoch": 4.56, + "learning_rate": 1.6342191481616308e-06, + "loss": 0.0132, + "step": 25062 + }, + { + "epoch": 4.56, + "learning_rate": 1.6341645431379687e-06, + "loss": 0.2706, + "step": 25063 + }, + { + "epoch": 4.56, + "learning_rate": 1.6341099381143065e-06, + "loss": 2.6521, + "step": 25064 + }, + { + "epoch": 4.56, + "learning_rate": 1.6340553330906444e-06, + "loss": 0.5086, + "step": 25065 + }, + { + "epoch": 4.56, + "learning_rate": 1.6340007280669823e-06, + "loss": 2.6816, + "step": 25066 + }, + { + "epoch": 4.56, + "learning_rate": 1.6339461230433199e-06, + "loss": 0.8073, + "step": 25067 + }, + { + "epoch": 4.56, + "learning_rate": 1.6338915180196578e-06, + "loss": 0.8715, + "step": 25068 + }, + { + "epoch": 4.56, + "learning_rate": 1.6338369129959958e-06, + "loss": 0.483, + "step": 25069 + }, + { + "epoch": 4.56, + "learning_rate": 1.6337823079723335e-06, + "loss": 0.0425, + "step": 25070 + }, + { + "epoch": 4.56, + "learning_rate": 1.6337277029486714e-06, + "loss": 0.2845, + "step": 25071 + }, + { + "epoch": 4.56, + "learning_rate": 1.6336730979250094e-06, + "loss": 0.4292, + "step": 25072 + }, + { + "epoch": 4.56, + "learning_rate": 1.6336184929013469e-06, + "loss": 0.9386, + "step": 25073 + }, + { + "epoch": 4.56, + "learning_rate": 1.6335638878776848e-06, + "loss": 0.1531, + "step": 25074 + }, + { + "epoch": 4.56, + "learning_rate": 1.6335092828540224e-06, + "loss": 1.1249, + "step": 25075 + }, + { + "epoch": 4.56, + "learning_rate": 1.6334546778303603e-06, + "loss": 0.2619, + "step": 25076 + }, + { + "epoch": 4.56, + "learning_rate": 1.6334000728066982e-06, + "loss": 0.3243, + "step": 25077 + }, + { + "epoch": 4.56, + "learning_rate": 1.633345467783036e-06, + "loss": 1.3674, + "step": 25078 + }, + { + "epoch": 4.56, + "learning_rate": 1.633290862759374e-06, + "loss": 3.6837, + "step": 25079 + }, + { + "epoch": 4.56, + "learning_rate": 1.6332362577357119e-06, + "loss": 1.6581, + "step": 25080 + }, + { + "epoch": 4.56, + "learning_rate": 1.6331816527120494e-06, + "loss": 0.5258, + "step": 25081 + }, + { + "epoch": 4.56, + "learning_rate": 1.6331270476883873e-06, + "loss": 0.1884, + "step": 25082 + }, + { + "epoch": 4.56, + "learning_rate": 1.6330724426647253e-06, + "loss": 0.4036, + "step": 25083 + }, + { + "epoch": 4.56, + "learning_rate": 1.633017837641063e-06, + "loss": 1.0893, + "step": 25084 + }, + { + "epoch": 4.56, + "learning_rate": 1.632963232617401e-06, + "loss": 0.1743, + "step": 25085 + }, + { + "epoch": 4.56, + "learning_rate": 1.6329086275937389e-06, + "loss": 2.2184, + "step": 25086 + }, + { + "epoch": 4.56, + "learning_rate": 1.6328540225700764e-06, + "loss": 0.439, + "step": 25087 + }, + { + "epoch": 4.56, + "learning_rate": 1.6327994175464143e-06, + "loss": 1.5461, + "step": 25088 + }, + { + "epoch": 4.56, + "learning_rate": 1.632744812522752e-06, + "loss": 1.0012, + "step": 25089 + }, + { + "epoch": 4.56, + "learning_rate": 1.63269020749909e-06, + "loss": 0.7697, + "step": 25090 + }, + { + "epoch": 4.56, + "learning_rate": 1.632635602475428e-06, + "loss": 0.4031, + "step": 25091 + }, + { + "epoch": 4.56, + "learning_rate": 1.6325809974517655e-06, + "loss": 0.5082, + "step": 25092 + }, + { + "epoch": 4.56, + "learning_rate": 1.6325263924281034e-06, + "loss": 2.1769, + "step": 25093 + }, + { + "epoch": 4.56, + "learning_rate": 1.6324717874044414e-06, + "loss": 0.4115, + "step": 25094 + }, + { + "epoch": 4.56, + "learning_rate": 1.6324171823807789e-06, + "loss": 0.1343, + "step": 25095 + }, + { + "epoch": 4.56, + "learning_rate": 1.6323625773571168e-06, + "loss": 0.2195, + "step": 25096 + }, + { + "epoch": 4.56, + "learning_rate": 1.6323079723334548e-06, + "loss": 0.3476, + "step": 25097 + }, + { + "epoch": 4.56, + "learning_rate": 1.6322533673097925e-06, + "loss": 0.223, + "step": 25098 + }, + { + "epoch": 4.56, + "learning_rate": 1.6321987622861305e-06, + "loss": 1.4695, + "step": 25099 + }, + { + "epoch": 4.56, + "learning_rate": 1.6321441572624684e-06, + "loss": 0.4238, + "step": 25100 + }, + { + "epoch": 4.56, + "learning_rate": 1.632089552238806e-06, + "loss": 1.2064, + "step": 25101 + }, + { + "epoch": 4.56, + "learning_rate": 1.6320349472151439e-06, + "loss": 0.5688, + "step": 25102 + }, + { + "epoch": 4.57, + "learning_rate": 1.6319803421914816e-06, + "loss": 1.8616, + "step": 25103 + }, + { + "epoch": 4.57, + "learning_rate": 1.6319257371678195e-06, + "loss": 1.2662, + "step": 25104 + }, + { + "epoch": 4.57, + "learning_rate": 1.6318711321441575e-06, + "loss": 1.8712, + "step": 25105 + }, + { + "epoch": 4.57, + "learning_rate": 1.631816527120495e-06, + "loss": 0.1374, + "step": 25106 + }, + { + "epoch": 4.57, + "learning_rate": 1.631761922096833e-06, + "loss": 0.4395, + "step": 25107 + }, + { + "epoch": 4.57, + "learning_rate": 1.6317073170731709e-06, + "loss": 1.2719, + "step": 25108 + }, + { + "epoch": 4.57, + "learning_rate": 1.6316527120495084e-06, + "loss": 1.6142, + "step": 25109 + }, + { + "epoch": 4.57, + "learning_rate": 1.6315981070258464e-06, + "loss": 0.2749, + "step": 25110 + }, + { + "epoch": 4.57, + "learning_rate": 1.6315435020021843e-06, + "loss": 0.0076, + "step": 25111 + }, + { + "epoch": 4.57, + "learning_rate": 1.631488896978522e-06, + "loss": 0.73, + "step": 25112 + }, + { + "epoch": 4.57, + "learning_rate": 1.63143429195486e-06, + "loss": 1.1456, + "step": 25113 + }, + { + "epoch": 4.57, + "learning_rate": 1.631379686931198e-06, + "loss": 2.4479, + "step": 25114 + }, + { + "epoch": 4.57, + "learning_rate": 1.6313250819075354e-06, + "loss": 0.25, + "step": 25115 + }, + { + "epoch": 4.57, + "learning_rate": 1.6312704768838734e-06, + "loss": 0.5312, + "step": 25116 + }, + { + "epoch": 4.57, + "learning_rate": 1.6312158718602111e-06, + "loss": 0.5265, + "step": 25117 + }, + { + "epoch": 4.57, + "learning_rate": 1.631161266836549e-06, + "loss": 0.1861, + "step": 25118 + }, + { + "epoch": 4.57, + "learning_rate": 1.631106661812887e-06, + "loss": 0.3613, + "step": 25119 + }, + { + "epoch": 4.57, + "learning_rate": 1.6310520567892245e-06, + "loss": 0.4135, + "step": 25120 + }, + { + "epoch": 4.57, + "learning_rate": 1.6309974517655625e-06, + "loss": 0.6947, + "step": 25121 + }, + { + "epoch": 4.57, + "learning_rate": 1.6309428467419004e-06, + "loss": 0.8478, + "step": 25122 + }, + { + "epoch": 4.57, + "learning_rate": 1.630888241718238e-06, + "loss": 0.0482, + "step": 25123 + }, + { + "epoch": 4.57, + "learning_rate": 1.6308336366945759e-06, + "loss": 0.2223, + "step": 25124 + }, + { + "epoch": 4.57, + "learning_rate": 1.6307790316709138e-06, + "loss": 0.0367, + "step": 25125 + }, + { + "epoch": 4.57, + "learning_rate": 1.6307244266472515e-06, + "loss": 0.0236, + "step": 25126 + }, + { + "epoch": 4.57, + "learning_rate": 1.6306698216235895e-06, + "loss": 2.3405, + "step": 25127 + }, + { + "epoch": 4.57, + "learning_rate": 1.630615216599927e-06, + "loss": 0.133, + "step": 25128 + }, + { + "epoch": 4.57, + "learning_rate": 1.630560611576265e-06, + "loss": 0.9085, + "step": 25129 + }, + { + "epoch": 4.57, + "learning_rate": 1.6305060065526029e-06, + "loss": 0.9041, + "step": 25130 + }, + { + "epoch": 4.57, + "learning_rate": 1.6304514015289406e-06, + "loss": 0.2121, + "step": 25131 + }, + { + "epoch": 4.57, + "learning_rate": 1.6303967965052786e-06, + "loss": 0.0076, + "step": 25132 + }, + { + "epoch": 4.57, + "learning_rate": 1.6303421914816165e-06, + "loss": 0.7163, + "step": 25133 + }, + { + "epoch": 4.57, + "learning_rate": 1.630287586457954e-06, + "loss": 0.0575, + "step": 25134 + }, + { + "epoch": 4.57, + "learning_rate": 1.630232981434292e-06, + "loss": 0.2712, + "step": 25135 + }, + { + "epoch": 4.57, + "learning_rate": 1.63017837641063e-06, + "loss": 0.1054, + "step": 25136 + }, + { + "epoch": 4.57, + "learning_rate": 1.6301237713869676e-06, + "loss": 1.0522, + "step": 25137 + }, + { + "epoch": 4.57, + "learning_rate": 1.6300691663633056e-06, + "loss": 0.9955, + "step": 25138 + }, + { + "epoch": 4.57, + "learning_rate": 1.6300145613396435e-06, + "loss": 0.0114, + "step": 25139 + }, + { + "epoch": 4.57, + "learning_rate": 1.629959956315981e-06, + "loss": 2.5909, + "step": 25140 + }, + { + "epoch": 4.57, + "learning_rate": 1.629905351292319e-06, + "loss": 0.0096, + "step": 25141 + }, + { + "epoch": 4.57, + "learning_rate": 1.6298507462686565e-06, + "loss": 0.9985, + "step": 25142 + }, + { + "epoch": 4.57, + "learning_rate": 1.6297961412449945e-06, + "loss": 0.9665, + "step": 25143 + }, + { + "epoch": 4.57, + "learning_rate": 1.6297415362213324e-06, + "loss": 0.7495, + "step": 25144 + }, + { + "epoch": 4.57, + "learning_rate": 1.6296869311976701e-06, + "loss": 0.0388, + "step": 25145 + }, + { + "epoch": 4.57, + "learning_rate": 1.629632326174008e-06, + "loss": 1.473, + "step": 25146 + }, + { + "epoch": 4.57, + "learning_rate": 1.629577721150346e-06, + "loss": 0.0147, + "step": 25147 + }, + { + "epoch": 4.57, + "learning_rate": 1.6295231161266835e-06, + "loss": 0.214, + "step": 25148 + }, + { + "epoch": 4.57, + "learning_rate": 1.6294685111030215e-06, + "loss": 0.9776, + "step": 25149 + }, + { + "epoch": 4.57, + "learning_rate": 1.6294139060793594e-06, + "loss": 0.6346, + "step": 25150 + }, + { + "epoch": 4.57, + "learning_rate": 1.6293593010556972e-06, + "loss": 0.9608, + "step": 25151 + }, + { + "epoch": 4.57, + "learning_rate": 1.629304696032035e-06, + "loss": 0.5937, + "step": 25152 + }, + { + "epoch": 4.57, + "learning_rate": 1.629250091008373e-06, + "loss": 0.6067, + "step": 25153 + }, + { + "epoch": 4.57, + "learning_rate": 1.6291954859847106e-06, + "loss": 2.501, + "step": 25154 + }, + { + "epoch": 4.57, + "learning_rate": 1.6291408809610485e-06, + "loss": 0.7328, + "step": 25155 + }, + { + "epoch": 4.57, + "learning_rate": 1.629086275937386e-06, + "loss": 0.4244, + "step": 25156 + }, + { + "epoch": 4.57, + "learning_rate": 1.629031670913724e-06, + "loss": 0.8965, + "step": 25157 + }, + { + "epoch": 4.58, + "learning_rate": 1.628977065890062e-06, + "loss": 0.5641, + "step": 25158 + }, + { + "epoch": 4.58, + "learning_rate": 1.6289224608663996e-06, + "loss": 0.6132, + "step": 25159 + }, + { + "epoch": 4.58, + "learning_rate": 1.6288678558427376e-06, + "loss": 2.4466, + "step": 25160 + }, + { + "epoch": 4.58, + "learning_rate": 1.6288132508190755e-06, + "loss": 0.9793, + "step": 25161 + }, + { + "epoch": 4.58, + "learning_rate": 1.628758645795413e-06, + "loss": 0.5491, + "step": 25162 + }, + { + "epoch": 4.58, + "learning_rate": 1.628704040771751e-06, + "loss": 0.2218, + "step": 25163 + }, + { + "epoch": 4.58, + "learning_rate": 1.628649435748089e-06, + "loss": 2.1938, + "step": 25164 + }, + { + "epoch": 4.58, + "learning_rate": 1.6285948307244267e-06, + "loss": 1.1727, + "step": 25165 + }, + { + "epoch": 4.58, + "learning_rate": 1.6285402257007646e-06, + "loss": 0.1242, + "step": 25166 + }, + { + "epoch": 4.58, + "learning_rate": 1.6284856206771026e-06, + "loss": 1.0916, + "step": 25167 + }, + { + "epoch": 4.58, + "learning_rate": 1.62843101565344e-06, + "loss": 0.0826, + "step": 25168 + }, + { + "epoch": 4.58, + "learning_rate": 1.628376410629778e-06, + "loss": 0.4575, + "step": 25169 + }, + { + "epoch": 4.58, + "learning_rate": 1.6283218056061158e-06, + "loss": 0.2661, + "step": 25170 + }, + { + "epoch": 4.58, + "learning_rate": 1.6282672005824537e-06, + "loss": 0.6568, + "step": 25171 + }, + { + "epoch": 4.58, + "learning_rate": 1.6282125955587916e-06, + "loss": 1.3735, + "step": 25172 + }, + { + "epoch": 4.58, + "learning_rate": 1.6281579905351292e-06, + "loss": 1.3215, + "step": 25173 + }, + { + "epoch": 4.58, + "learning_rate": 1.628103385511467e-06, + "loss": 0.0117, + "step": 25174 + }, + { + "epoch": 4.58, + "learning_rate": 1.628048780487805e-06, + "loss": 0.1771, + "step": 25175 + }, + { + "epoch": 4.58, + "learning_rate": 1.6279941754641426e-06, + "loss": 2.0325, + "step": 25176 + }, + { + "epoch": 4.58, + "learning_rate": 1.6279395704404805e-06, + "loss": 0.4627, + "step": 25177 + }, + { + "epoch": 4.58, + "learning_rate": 1.6278849654168185e-06, + "loss": 0.2477, + "step": 25178 + }, + { + "epoch": 4.58, + "learning_rate": 1.6278303603931562e-06, + "loss": 0.3496, + "step": 25179 + }, + { + "epoch": 4.58, + "learning_rate": 1.6277757553694941e-06, + "loss": 0.0054, + "step": 25180 + }, + { + "epoch": 4.58, + "learning_rate": 1.627721150345832e-06, + "loss": 0.7669, + "step": 25181 + }, + { + "epoch": 4.58, + "learning_rate": 1.6276665453221696e-06, + "loss": 0.0337, + "step": 25182 + }, + { + "epoch": 4.58, + "learning_rate": 1.6276119402985075e-06, + "loss": 0.2617, + "step": 25183 + }, + { + "epoch": 4.58, + "learning_rate": 1.6275573352748453e-06, + "loss": 0.1603, + "step": 25184 + }, + { + "epoch": 4.58, + "learning_rate": 1.6275027302511832e-06, + "loss": 0.2664, + "step": 25185 + }, + { + "epoch": 4.58, + "learning_rate": 1.6274481252275212e-06, + "loss": 0.0474, + "step": 25186 + }, + { + "epoch": 4.58, + "learning_rate": 1.6273935202038587e-06, + "loss": 0.1322, + "step": 25187 + }, + { + "epoch": 4.58, + "learning_rate": 1.6273389151801966e-06, + "loss": 1.6247, + "step": 25188 + }, + { + "epoch": 4.58, + "learning_rate": 1.6272843101565346e-06, + "loss": 0.3829, + "step": 25189 + }, + { + "epoch": 4.58, + "learning_rate": 1.627229705132872e-06, + "loss": 0.5064, + "step": 25190 + }, + { + "epoch": 4.58, + "learning_rate": 1.62717510010921e-06, + "loss": 0.5515, + "step": 25191 + }, + { + "epoch": 4.58, + "learning_rate": 1.627120495085548e-06, + "loss": 0.0069, + "step": 25192 + }, + { + "epoch": 4.58, + "learning_rate": 1.6270658900618857e-06, + "loss": 0.1263, + "step": 25193 + }, + { + "epoch": 4.58, + "learning_rate": 1.6270112850382236e-06, + "loss": 2.0842, + "step": 25194 + }, + { + "epoch": 4.58, + "learning_rate": 1.6269566800145612e-06, + "loss": 0.1979, + "step": 25195 + }, + { + "epoch": 4.58, + "learning_rate": 1.6269020749908991e-06, + "loss": 1.2852, + "step": 25196 + }, + { + "epoch": 4.58, + "learning_rate": 1.626847469967237e-06, + "loss": 0.455, + "step": 25197 + }, + { + "epoch": 4.58, + "learning_rate": 1.6267928649435748e-06, + "loss": 1.0668, + "step": 25198 + }, + { + "epoch": 4.58, + "learning_rate": 1.6267382599199127e-06, + "loss": 2.5929, + "step": 25199 + }, + { + "epoch": 4.58, + "learning_rate": 1.6266836548962507e-06, + "loss": 2.0251, + "step": 25200 + }, + { + "epoch": 4.58, + "learning_rate": 1.6266290498725882e-06, + "loss": 0.0733, + "step": 25201 + }, + { + "epoch": 4.58, + "learning_rate": 1.6265744448489261e-06, + "loss": 0.447, + "step": 25202 + }, + { + "epoch": 4.58, + "learning_rate": 1.626519839825264e-06, + "loss": 0.2798, + "step": 25203 + }, + { + "epoch": 4.58, + "learning_rate": 1.6264652348016016e-06, + "loss": 1.6591, + "step": 25204 + }, + { + "epoch": 4.58, + "learning_rate": 1.6264106297779395e-06, + "loss": 1.2793, + "step": 25205 + }, + { + "epoch": 4.58, + "learning_rate": 1.6263560247542775e-06, + "loss": 0.0587, + "step": 25206 + }, + { + "epoch": 4.58, + "learning_rate": 1.6263014197306152e-06, + "loss": 0.0087, + "step": 25207 + }, + { + "epoch": 4.58, + "learning_rate": 1.6262468147069532e-06, + "loss": 4.6864, + "step": 25208 + }, + { + "epoch": 4.58, + "learning_rate": 1.6261922096832907e-06, + "loss": 0.773, + "step": 25209 + }, + { + "epoch": 4.58, + "learning_rate": 1.6261376046596286e-06, + "loss": 0.2178, + "step": 25210 + }, + { + "epoch": 4.58, + "learning_rate": 1.6260829996359666e-06, + "loss": 0.0067, + "step": 25211 + }, + { + "epoch": 4.58, + "learning_rate": 1.6260283946123043e-06, + "loss": 1.1035, + "step": 25212 + }, + { + "epoch": 4.59, + "learning_rate": 1.6259737895886422e-06, + "loss": 1.579, + "step": 25213 + }, + { + "epoch": 4.59, + "learning_rate": 1.6259191845649802e-06, + "loss": 0.0177, + "step": 25214 + }, + { + "epoch": 4.59, + "learning_rate": 1.6258645795413177e-06, + "loss": 0.5705, + "step": 25215 + }, + { + "epoch": 4.59, + "learning_rate": 1.6258099745176556e-06, + "loss": 0.165, + "step": 25216 + }, + { + "epoch": 4.59, + "learning_rate": 1.6257553694939936e-06, + "loss": 0.0121, + "step": 25217 + }, + { + "epoch": 4.59, + "learning_rate": 1.6257007644703313e-06, + "loss": 0.4373, + "step": 25218 + }, + { + "epoch": 4.59, + "learning_rate": 1.6256461594466693e-06, + "loss": 0.1266, + "step": 25219 + }, + { + "epoch": 4.59, + "learning_rate": 1.6255915544230072e-06, + "loss": 1.4772, + "step": 25220 + }, + { + "epoch": 4.59, + "learning_rate": 1.6255369493993447e-06, + "loss": 0.8179, + "step": 25221 + }, + { + "epoch": 4.59, + "learning_rate": 1.6254823443756827e-06, + "loss": 0.5414, + "step": 25222 + }, + { + "epoch": 4.59, + "learning_rate": 1.6254277393520202e-06, + "loss": 1.8444, + "step": 25223 + }, + { + "epoch": 4.59, + "learning_rate": 1.6253731343283581e-06, + "loss": 0.0888, + "step": 25224 + }, + { + "epoch": 4.59, + "learning_rate": 1.625318529304696e-06, + "loss": 0.0079, + "step": 25225 + }, + { + "epoch": 4.59, + "learning_rate": 1.6252639242810338e-06, + "loss": 0.0143, + "step": 25226 + }, + { + "epoch": 4.59, + "learning_rate": 1.6252093192573718e-06, + "loss": 0.0119, + "step": 25227 + }, + { + "epoch": 4.59, + "learning_rate": 1.6251547142337097e-06, + "loss": 0.1146, + "step": 25228 + }, + { + "epoch": 4.59, + "learning_rate": 1.6251001092100472e-06, + "loss": 0.2407, + "step": 25229 + }, + { + "epoch": 4.59, + "learning_rate": 1.6250455041863852e-06, + "loss": 0.0869, + "step": 25230 + }, + { + "epoch": 4.59, + "learning_rate": 1.624990899162723e-06, + "loss": 0.7084, + "step": 25231 + }, + { + "epoch": 4.59, + "learning_rate": 1.6249362941390608e-06, + "loss": 2.1818, + "step": 25232 + }, + { + "epoch": 4.59, + "learning_rate": 1.6248816891153988e-06, + "loss": 0.1819, + "step": 25233 + }, + { + "epoch": 4.59, + "learning_rate": 1.6248270840917367e-06, + "loss": 0.4566, + "step": 25234 + }, + { + "epoch": 4.59, + "learning_rate": 1.6247724790680742e-06, + "loss": 0.4365, + "step": 25235 + }, + { + "epoch": 4.59, + "learning_rate": 1.6247178740444122e-06, + "loss": 0.1164, + "step": 25236 + }, + { + "epoch": 4.59, + "learning_rate": 1.6246632690207497e-06, + "loss": 0.0343, + "step": 25237 + }, + { + "epoch": 4.59, + "learning_rate": 1.6246086639970876e-06, + "loss": 0.1461, + "step": 25238 + }, + { + "epoch": 4.59, + "learning_rate": 1.6245540589734256e-06, + "loss": 0.4778, + "step": 25239 + }, + { + "epoch": 4.59, + "learning_rate": 1.6244994539497633e-06, + "loss": 0.6674, + "step": 25240 + }, + { + "epoch": 4.59, + "learning_rate": 1.6244448489261013e-06, + "loss": 0.2955, + "step": 25241 + }, + { + "epoch": 4.59, + "learning_rate": 1.6243902439024392e-06, + "loss": 1.7316, + "step": 25242 + }, + { + "epoch": 4.59, + "learning_rate": 1.6243356388787767e-06, + "loss": 0.7901, + "step": 25243 + }, + { + "epoch": 4.59, + "learning_rate": 1.6242810338551147e-06, + "loss": 1.059, + "step": 25244 + }, + { + "epoch": 4.59, + "learning_rate": 1.6242264288314526e-06, + "loss": 0.5144, + "step": 25245 + }, + { + "epoch": 4.59, + "learning_rate": 1.6241718238077903e-06, + "loss": 0.6584, + "step": 25246 + }, + { + "epoch": 4.59, + "learning_rate": 1.6241172187841283e-06, + "loss": 1.4824, + "step": 25247 + }, + { + "epoch": 4.59, + "learning_rate": 1.6240626137604662e-06, + "loss": 1.2733, + "step": 25248 + }, + { + "epoch": 4.59, + "learning_rate": 1.6240080087368038e-06, + "loss": 1.3106, + "step": 25249 + }, + { + "epoch": 4.59, + "learning_rate": 1.6239534037131417e-06, + "loss": 1.5851, + "step": 25250 + }, + { + "epoch": 4.59, + "learning_rate": 1.6238987986894794e-06, + "loss": 0.0192, + "step": 25251 + }, + { + "epoch": 4.59, + "learning_rate": 1.6238441936658174e-06, + "loss": 0.4227, + "step": 25252 + }, + { + "epoch": 4.59, + "learning_rate": 1.623789588642155e-06, + "loss": 0.4688, + "step": 25253 + }, + { + "epoch": 4.59, + "learning_rate": 1.6237349836184928e-06, + "loss": 0.1949, + "step": 25254 + }, + { + "epoch": 4.59, + "learning_rate": 1.6236803785948308e-06, + "loss": 1.1992, + "step": 25255 + }, + { + "epoch": 4.59, + "learning_rate": 1.6236257735711687e-06, + "loss": 2.2043, + "step": 25256 + }, + { + "epoch": 4.59, + "learning_rate": 1.6235711685475062e-06, + "loss": 0.0413, + "step": 25257 + }, + { + "epoch": 4.59, + "learning_rate": 1.6235165635238442e-06, + "loss": 0.0126, + "step": 25258 + }, + { + "epoch": 4.59, + "learning_rate": 1.6234619585001821e-06, + "loss": 0.1229, + "step": 25259 + }, + { + "epoch": 4.59, + "learning_rate": 1.6234073534765199e-06, + "loss": 0.3468, + "step": 25260 + }, + { + "epoch": 4.59, + "learning_rate": 1.6233527484528578e-06, + "loss": 0.1507, + "step": 25261 + }, + { + "epoch": 4.59, + "learning_rate": 1.6232981434291957e-06, + "loss": 1.1618, + "step": 25262 + }, + { + "epoch": 4.59, + "learning_rate": 1.6232435384055333e-06, + "loss": 0.0452, + "step": 25263 + }, + { + "epoch": 4.59, + "learning_rate": 1.6231889333818712e-06, + "loss": 0.1381, + "step": 25264 + }, + { + "epoch": 4.59, + "learning_rate": 1.623134328358209e-06, + "loss": 0.2673, + "step": 25265 + }, + { + "epoch": 4.59, + "learning_rate": 1.6230797233345469e-06, + "loss": 0.4625, + "step": 25266 + }, + { + "epoch": 4.59, + "learning_rate": 1.6230251183108848e-06, + "loss": 1.7872, + "step": 25267 + }, + { + "epoch": 4.6, + "learning_rate": 1.6229705132872223e-06, + "loss": 0.0916, + "step": 25268 + }, + { + "epoch": 4.6, + "learning_rate": 1.6229159082635603e-06, + "loss": 1.5213, + "step": 25269 + }, + { + "epoch": 4.6, + "learning_rate": 1.6228613032398982e-06, + "loss": 0.3803, + "step": 25270 + }, + { + "epoch": 4.6, + "learning_rate": 1.6228066982162358e-06, + "loss": 0.1276, + "step": 25271 + }, + { + "epoch": 4.6, + "learning_rate": 1.6227520931925737e-06, + "loss": 0.1591, + "step": 25272 + }, + { + "epoch": 4.6, + "learning_rate": 1.6226974881689116e-06, + "loss": 0.4726, + "step": 25273 + }, + { + "epoch": 4.6, + "learning_rate": 1.6226428831452494e-06, + "loss": 4.2986, + "step": 25274 + }, + { + "epoch": 4.6, + "learning_rate": 1.6225882781215873e-06, + "loss": 1.6246, + "step": 25275 + }, + { + "epoch": 4.6, + "learning_rate": 1.6225336730979248e-06, + "loss": 0.3357, + "step": 25276 + }, + { + "epoch": 4.6, + "learning_rate": 1.6224790680742628e-06, + "loss": 0.0109, + "step": 25277 + }, + { + "epoch": 4.6, + "learning_rate": 1.6224244630506007e-06, + "loss": 0.0072, + "step": 25278 + }, + { + "epoch": 4.6, + "learning_rate": 1.6223698580269385e-06, + "loss": 1.3381, + "step": 25279 + }, + { + "epoch": 4.6, + "learning_rate": 1.6223152530032764e-06, + "loss": 0.5739, + "step": 25280 + }, + { + "epoch": 4.6, + "learning_rate": 1.6222606479796143e-06, + "loss": 0.3746, + "step": 25281 + }, + { + "epoch": 4.6, + "learning_rate": 1.6222060429559519e-06, + "loss": 1.1824, + "step": 25282 + }, + { + "epoch": 4.6, + "learning_rate": 1.6221514379322898e-06, + "loss": 0.6654, + "step": 25283 + }, + { + "epoch": 4.6, + "learning_rate": 1.6220968329086277e-06, + "loss": 2.5283, + "step": 25284 + }, + { + "epoch": 4.6, + "learning_rate": 1.6220422278849653e-06, + "loss": 0.9828, + "step": 25285 + }, + { + "epoch": 4.6, + "learning_rate": 1.6219876228613032e-06, + "loss": 1.8935, + "step": 25286 + }, + { + "epoch": 4.6, + "learning_rate": 1.6219330178376412e-06, + "loss": 0.4609, + "step": 25287 + }, + { + "epoch": 4.6, + "learning_rate": 1.6218784128139789e-06, + "loss": 0.4175, + "step": 25288 + }, + { + "epoch": 4.6, + "learning_rate": 1.6218238077903168e-06, + "loss": 2.3739, + "step": 25289 + }, + { + "epoch": 4.6, + "learning_rate": 1.6217692027666544e-06, + "loss": 0.0122, + "step": 25290 + }, + { + "epoch": 4.6, + "learning_rate": 1.6217145977429923e-06, + "loss": 0.749, + "step": 25291 + }, + { + "epoch": 4.6, + "learning_rate": 1.6216599927193302e-06, + "loss": 0.1117, + "step": 25292 + }, + { + "epoch": 4.6, + "learning_rate": 1.621605387695668e-06, + "loss": 0.0498, + "step": 25293 + }, + { + "epoch": 4.6, + "learning_rate": 1.621550782672006e-06, + "loss": 2.7646, + "step": 25294 + }, + { + "epoch": 4.6, + "learning_rate": 1.6214961776483439e-06, + "loss": 0.0323, + "step": 25295 + }, + { + "epoch": 4.6, + "learning_rate": 1.6214415726246814e-06, + "loss": 0.0683, + "step": 25296 + }, + { + "epoch": 4.6, + "learning_rate": 1.6213869676010193e-06, + "loss": 0.5106, + "step": 25297 + }, + { + "epoch": 4.6, + "learning_rate": 1.6213323625773573e-06, + "loss": 0.0751, + "step": 25298 + }, + { + "epoch": 4.6, + "learning_rate": 1.621277757553695e-06, + "loss": 0.2611, + "step": 25299 + }, + { + "epoch": 4.6, + "learning_rate": 1.621223152530033e-06, + "loss": 1.8632, + "step": 25300 + }, + { + "epoch": 4.6, + "learning_rate": 1.6211685475063709e-06, + "loss": 1.3131, + "step": 25301 + }, + { + "epoch": 4.6, + "learning_rate": 1.6211139424827084e-06, + "loss": 0.6101, + "step": 25302 + }, + { + "epoch": 4.6, + "learning_rate": 1.6210593374590463e-06, + "loss": 1.0805, + "step": 25303 + }, + { + "epoch": 4.6, + "learning_rate": 1.6210047324353839e-06, + "loss": 0.0134, + "step": 25304 + }, + { + "epoch": 4.6, + "learning_rate": 1.6209501274117218e-06, + "loss": 0.0118, + "step": 25305 + }, + { + "epoch": 4.6, + "learning_rate": 1.6208955223880598e-06, + "loss": 0.5095, + "step": 25306 + }, + { + "epoch": 4.6, + "learning_rate": 1.6208409173643975e-06, + "loss": 0.0073, + "step": 25307 + }, + { + "epoch": 4.6, + "learning_rate": 1.6207863123407354e-06, + "loss": 0.9575, + "step": 25308 + }, + { + "epoch": 4.6, + "learning_rate": 1.6207317073170734e-06, + "loss": 0.337, + "step": 25309 + }, + { + "epoch": 4.6, + "learning_rate": 1.6206771022934109e-06, + "loss": 0.2586, + "step": 25310 + }, + { + "epoch": 4.6, + "learning_rate": 1.6206224972697488e-06, + "loss": 1.2909, + "step": 25311 + }, + { + "epoch": 4.6, + "learning_rate": 1.6205678922460868e-06, + "loss": 0.0179, + "step": 25312 + }, + { + "epoch": 4.6, + "learning_rate": 1.6205132872224245e-06, + "loss": 0.2973, + "step": 25313 + }, + { + "epoch": 4.6, + "learning_rate": 1.6204586821987625e-06, + "loss": 0.294, + "step": 25314 + }, + { + "epoch": 4.6, + "learning_rate": 1.6204040771751004e-06, + "loss": 0.0725, + "step": 25315 + }, + { + "epoch": 4.6, + "learning_rate": 1.620349472151438e-06, + "loss": 0.0065, + "step": 25316 + }, + { + "epoch": 4.6, + "learning_rate": 1.6202948671277759e-06, + "loss": 1.9055, + "step": 25317 + }, + { + "epoch": 4.6, + "learning_rate": 1.6202402621041134e-06, + "loss": 0.2159, + "step": 25318 + }, + { + "epoch": 4.6, + "learning_rate": 1.6201856570804513e-06, + "loss": 0.1774, + "step": 25319 + }, + { + "epoch": 4.6, + "learning_rate": 1.6201310520567893e-06, + "loss": 0.7076, + "step": 25320 + }, + { + "epoch": 4.6, + "learning_rate": 1.620076447033127e-06, + "loss": 0.1842, + "step": 25321 + }, + { + "epoch": 4.6, + "learning_rate": 1.620021842009465e-06, + "loss": 0.5685, + "step": 25322 + }, + { + "epoch": 4.61, + "learning_rate": 1.6199672369858029e-06, + "loss": 1.007, + "step": 25323 + }, + { + "epoch": 4.61, + "learning_rate": 1.6199126319621404e-06, + "loss": 0.2661, + "step": 25324 + }, + { + "epoch": 4.61, + "learning_rate": 1.6198580269384783e-06, + "loss": 0.7384, + "step": 25325 + }, + { + "epoch": 4.61, + "learning_rate": 1.6198034219148163e-06, + "loss": 0.1464, + "step": 25326 + }, + { + "epoch": 4.61, + "learning_rate": 1.619748816891154e-06, + "loss": 1.12, + "step": 25327 + }, + { + "epoch": 4.61, + "learning_rate": 1.619694211867492e-06, + "loss": 2.0086, + "step": 25328 + }, + { + "epoch": 4.61, + "learning_rate": 1.61963960684383e-06, + "loss": 0.0183, + "step": 25329 + }, + { + "epoch": 4.61, + "learning_rate": 1.6195850018201674e-06, + "loss": 0.1606, + "step": 25330 + }, + { + "epoch": 4.61, + "learning_rate": 1.6195303967965054e-06, + "loss": 1.7451, + "step": 25331 + }, + { + "epoch": 4.61, + "learning_rate": 1.6194757917728429e-06, + "loss": 1.6363, + "step": 25332 + }, + { + "epoch": 4.61, + "learning_rate": 1.6194211867491808e-06, + "loss": 0.8729, + "step": 25333 + }, + { + "epoch": 4.61, + "learning_rate": 1.6193665817255188e-06, + "loss": 0.6479, + "step": 25334 + }, + { + "epoch": 4.61, + "learning_rate": 1.6193119767018565e-06, + "loss": 0.0132, + "step": 25335 + }, + { + "epoch": 4.61, + "learning_rate": 1.6192573716781945e-06, + "loss": 0.0219, + "step": 25336 + }, + { + "epoch": 4.61, + "learning_rate": 1.6192027666545324e-06, + "loss": 1.8564, + "step": 25337 + }, + { + "epoch": 4.61, + "learning_rate": 1.61914816163087e-06, + "loss": 0.1976, + "step": 25338 + }, + { + "epoch": 4.61, + "learning_rate": 1.6190935566072079e-06, + "loss": 0.251, + "step": 25339 + }, + { + "epoch": 4.61, + "learning_rate": 1.6190389515835458e-06, + "loss": 0.215, + "step": 25340 + }, + { + "epoch": 4.61, + "learning_rate": 1.6189843465598835e-06, + "loss": 0.0199, + "step": 25341 + }, + { + "epoch": 4.61, + "learning_rate": 1.6189297415362215e-06, + "loss": 0.0338, + "step": 25342 + }, + { + "epoch": 4.61, + "learning_rate": 1.6188751365125594e-06, + "loss": 0.2498, + "step": 25343 + }, + { + "epoch": 4.61, + "learning_rate": 1.618820531488897e-06, + "loss": 0.6579, + "step": 25344 + }, + { + "epoch": 4.61, + "learning_rate": 1.6187659264652349e-06, + "loss": 0.1794, + "step": 25345 + }, + { + "epoch": 4.61, + "learning_rate": 1.6187113214415726e-06, + "loss": 0.9404, + "step": 25346 + }, + { + "epoch": 4.61, + "learning_rate": 1.6186567164179106e-06, + "loss": 0.0603, + "step": 25347 + }, + { + "epoch": 4.61, + "learning_rate": 1.6186021113942485e-06, + "loss": 0.0069, + "step": 25348 + }, + { + "epoch": 4.61, + "learning_rate": 1.618547506370586e-06, + "loss": 1.2153, + "step": 25349 + }, + { + "epoch": 4.61, + "learning_rate": 1.618492901346924e-06, + "loss": 1.8583, + "step": 25350 + }, + { + "epoch": 4.61, + "learning_rate": 1.618438296323262e-06, + "loss": 0.0455, + "step": 25351 + }, + { + "epoch": 4.61, + "learning_rate": 1.6183836912995994e-06, + "loss": 0.0091, + "step": 25352 + }, + { + "epoch": 4.61, + "learning_rate": 1.6183290862759374e-06, + "loss": 0.6405, + "step": 25353 + }, + { + "epoch": 4.61, + "learning_rate": 1.6182744812522753e-06, + "loss": 0.2775, + "step": 25354 + }, + { + "epoch": 4.61, + "learning_rate": 1.618219876228613e-06, + "loss": 0.0088, + "step": 25355 + }, + { + "epoch": 4.61, + "learning_rate": 1.618165271204951e-06, + "loss": 0.0468, + "step": 25356 + }, + { + "epoch": 4.61, + "learning_rate": 1.6181106661812885e-06, + "loss": 0.0116, + "step": 25357 + }, + { + "epoch": 4.61, + "learning_rate": 1.6180560611576265e-06, + "loss": 0.3559, + "step": 25358 + }, + { + "epoch": 4.61, + "learning_rate": 1.6180014561339644e-06, + "loss": 0.0079, + "step": 25359 + }, + { + "epoch": 4.61, + "learning_rate": 1.6179468511103021e-06, + "loss": 0.4149, + "step": 25360 + }, + { + "epoch": 4.61, + "learning_rate": 1.61789224608664e-06, + "loss": 0.405, + "step": 25361 + }, + { + "epoch": 4.61, + "learning_rate": 1.617837641062978e-06, + "loss": 1.4899, + "step": 25362 + }, + { + "epoch": 4.61, + "learning_rate": 1.6177830360393155e-06, + "loss": 0.301, + "step": 25363 + }, + { + "epoch": 4.61, + "learning_rate": 1.6177284310156535e-06, + "loss": 0.5217, + "step": 25364 + }, + { + "epoch": 4.61, + "learning_rate": 1.6176738259919914e-06, + "loss": 0.296, + "step": 25365 + }, + { + "epoch": 4.61, + "learning_rate": 1.617619220968329e-06, + "loss": 1.7816, + "step": 25366 + }, + { + "epoch": 4.61, + "learning_rate": 1.6175646159446669e-06, + "loss": 0.2177, + "step": 25367 + }, + { + "epoch": 4.61, + "learning_rate": 1.6175100109210048e-06, + "loss": 0.3958, + "step": 25368 + }, + { + "epoch": 4.61, + "learning_rate": 1.6174554058973426e-06, + "loss": 0.5858, + "step": 25369 + }, + { + "epoch": 4.61, + "learning_rate": 1.6174008008736805e-06, + "loss": 1.0331, + "step": 25370 + }, + { + "epoch": 4.61, + "learning_rate": 1.617346195850018e-06, + "loss": 2.2707, + "step": 25371 + }, + { + "epoch": 4.61, + "learning_rate": 1.617291590826356e-06, + "loss": 0.1865, + "step": 25372 + }, + { + "epoch": 4.61, + "learning_rate": 1.617236985802694e-06, + "loss": 0.68, + "step": 25373 + }, + { + "epoch": 4.61, + "learning_rate": 1.6171823807790316e-06, + "loss": 0.2, + "step": 25374 + }, + { + "epoch": 4.61, + "learning_rate": 1.6171277757553696e-06, + "loss": 3.6261, + "step": 25375 + }, + { + "epoch": 4.61, + "learning_rate": 1.6170731707317075e-06, + "loss": 1.863, + "step": 25376 + }, + { + "epoch": 4.61, + "learning_rate": 1.617018565708045e-06, + "loss": 0.1518, + "step": 25377 + }, + { + "epoch": 4.62, + "learning_rate": 1.616963960684383e-06, + "loss": 2.2304, + "step": 25378 + }, + { + "epoch": 4.62, + "learning_rate": 1.616909355660721e-06, + "loss": 1.1543, + "step": 25379 + }, + { + "epoch": 4.62, + "learning_rate": 1.6168547506370587e-06, + "loss": 0.1161, + "step": 25380 + }, + { + "epoch": 4.62, + "learning_rate": 1.6168001456133966e-06, + "loss": 0.6864, + "step": 25381 + }, + { + "epoch": 4.62, + "learning_rate": 1.6167455405897346e-06, + "loss": 0.4455, + "step": 25382 + }, + { + "epoch": 4.62, + "learning_rate": 1.616690935566072e-06, + "loss": 0.0127, + "step": 25383 + }, + { + "epoch": 4.62, + "learning_rate": 1.61663633054241e-06, + "loss": 0.151, + "step": 25384 + }, + { + "epoch": 4.62, + "learning_rate": 1.6165817255187475e-06, + "loss": 0.1188, + "step": 25385 + }, + { + "epoch": 4.62, + "learning_rate": 1.6165271204950855e-06, + "loss": 0.2691, + "step": 25386 + }, + { + "epoch": 4.62, + "learning_rate": 1.6164725154714234e-06, + "loss": 1.0522, + "step": 25387 + }, + { + "epoch": 4.62, + "learning_rate": 1.6164179104477612e-06, + "loss": 0.0098, + "step": 25388 + }, + { + "epoch": 4.62, + "learning_rate": 1.616363305424099e-06, + "loss": 0.8773, + "step": 25389 + }, + { + "epoch": 4.62, + "learning_rate": 1.616308700400437e-06, + "loss": 0.6379, + "step": 25390 + }, + { + "epoch": 4.62, + "learning_rate": 1.6162540953767746e-06, + "loss": 0.0071, + "step": 25391 + }, + { + "epoch": 4.62, + "learning_rate": 1.6161994903531125e-06, + "loss": 0.9403, + "step": 25392 + }, + { + "epoch": 4.62, + "learning_rate": 1.6161448853294505e-06, + "loss": 0.4696, + "step": 25393 + }, + { + "epoch": 4.62, + "learning_rate": 1.6160902803057882e-06, + "loss": 0.0169, + "step": 25394 + }, + { + "epoch": 4.62, + "learning_rate": 1.6160356752821261e-06, + "loss": 0.0787, + "step": 25395 + }, + { + "epoch": 4.62, + "learning_rate": 1.615981070258464e-06, + "loss": 0.927, + "step": 25396 + }, + { + "epoch": 4.62, + "learning_rate": 1.6159264652348016e-06, + "loss": 0.5876, + "step": 25397 + }, + { + "epoch": 4.62, + "learning_rate": 1.6158718602111395e-06, + "loss": 1.6301, + "step": 25398 + }, + { + "epoch": 4.62, + "learning_rate": 1.615817255187477e-06, + "loss": 0.149, + "step": 25399 + }, + { + "epoch": 4.62, + "learning_rate": 1.615762650163815e-06, + "loss": 0.563, + "step": 25400 + }, + { + "epoch": 4.62, + "learning_rate": 1.615708045140153e-06, + "loss": 0.2379, + "step": 25401 + }, + { + "epoch": 4.62, + "learning_rate": 1.6156534401164907e-06, + "loss": 1.2343, + "step": 25402 + }, + { + "epoch": 4.62, + "learning_rate": 1.6155988350928286e-06, + "loss": 2.236, + "step": 25403 + }, + { + "epoch": 4.62, + "learning_rate": 1.6155442300691666e-06, + "loss": 0.0191, + "step": 25404 + }, + { + "epoch": 4.62, + "learning_rate": 1.615489625045504e-06, + "loss": 0.0097, + "step": 25405 + }, + { + "epoch": 4.62, + "learning_rate": 1.615435020021842e-06, + "loss": 1.1475, + "step": 25406 + }, + { + "epoch": 4.62, + "learning_rate": 1.61538041499818e-06, + "loss": 0.2119, + "step": 25407 + }, + { + "epoch": 4.62, + "learning_rate": 1.6153258099745177e-06, + "loss": 0.3528, + "step": 25408 + }, + { + "epoch": 4.62, + "learning_rate": 1.6152712049508556e-06, + "loss": 0.0543, + "step": 25409 + }, + { + "epoch": 4.62, + "learning_rate": 1.6152165999271936e-06, + "loss": 0.4211, + "step": 25410 + }, + { + "epoch": 4.62, + "learning_rate": 1.615161994903531e-06, + "loss": 0.2891, + "step": 25411 + }, + { + "epoch": 4.62, + "learning_rate": 1.615107389879869e-06, + "loss": 2.9779, + "step": 25412 + }, + { + "epoch": 4.62, + "learning_rate": 1.6150527848562066e-06, + "loss": 0.5296, + "step": 25413 + }, + { + "epoch": 4.62, + "learning_rate": 1.6149981798325445e-06, + "loss": 2.0569, + "step": 25414 + }, + { + "epoch": 4.62, + "learning_rate": 1.6149435748088825e-06, + "loss": 0.7393, + "step": 25415 + }, + { + "epoch": 4.62, + "learning_rate": 1.6148889697852202e-06, + "loss": 0.1839, + "step": 25416 + }, + { + "epoch": 4.62, + "learning_rate": 1.6148343647615581e-06, + "loss": 0.0784, + "step": 25417 + }, + { + "epoch": 4.62, + "learning_rate": 1.614779759737896e-06, + "loss": 2.3056, + "step": 25418 + }, + { + "epoch": 4.62, + "learning_rate": 1.6147251547142336e-06, + "loss": 0.9584, + "step": 25419 + }, + { + "epoch": 4.62, + "learning_rate": 1.6146705496905715e-06, + "loss": 0.3732, + "step": 25420 + }, + { + "epoch": 4.62, + "learning_rate": 1.6146159446669095e-06, + "loss": 0.15, + "step": 25421 + }, + { + "epoch": 4.62, + "learning_rate": 1.6145613396432472e-06, + "loss": 2.8094, + "step": 25422 + }, + { + "epoch": 4.62, + "learning_rate": 1.6145067346195852e-06, + "loss": 0.2939, + "step": 25423 + }, + { + "epoch": 4.62, + "learning_rate": 1.614452129595923e-06, + "loss": 0.555, + "step": 25424 + }, + { + "epoch": 4.62, + "learning_rate": 1.6143975245722606e-06, + "loss": 0.7319, + "step": 25425 + }, + { + "epoch": 4.62, + "learning_rate": 1.6143429195485986e-06, + "loss": 0.2623, + "step": 25426 + }, + { + "epoch": 4.62, + "learning_rate": 1.6142883145249363e-06, + "loss": 0.181, + "step": 25427 + }, + { + "epoch": 4.62, + "learning_rate": 1.6142337095012742e-06, + "loss": 0.9259, + "step": 25428 + }, + { + "epoch": 4.62, + "learning_rate": 1.6141791044776122e-06, + "loss": 0.2859, + "step": 25429 + }, + { + "epoch": 4.62, + "learning_rate": 1.6141244994539497e-06, + "loss": 0.9882, + "step": 25430 + }, + { + "epoch": 4.62, + "learning_rate": 1.6140698944302876e-06, + "loss": 0.1157, + "step": 25431 + }, + { + "epoch": 4.62, + "learning_rate": 1.6140152894066256e-06, + "loss": 0.4493, + "step": 25432 + }, + { + "epoch": 4.63, + "learning_rate": 1.613960684382963e-06, + "loss": 0.0641, + "step": 25433 + }, + { + "epoch": 4.63, + "learning_rate": 1.613906079359301e-06, + "loss": 0.0068, + "step": 25434 + }, + { + "epoch": 4.63, + "learning_rate": 1.613851474335639e-06, + "loss": 1.821, + "step": 25435 + }, + { + "epoch": 4.63, + "learning_rate": 1.6137968693119767e-06, + "loss": 0.1028, + "step": 25436 + }, + { + "epoch": 4.63, + "learning_rate": 1.6137422642883147e-06, + "loss": 0.0204, + "step": 25437 + }, + { + "epoch": 4.63, + "learning_rate": 1.6136876592646522e-06, + "loss": 0.188, + "step": 25438 + }, + { + "epoch": 4.63, + "learning_rate": 1.6136330542409901e-06, + "loss": 0.5013, + "step": 25439 + }, + { + "epoch": 4.63, + "learning_rate": 1.613578449217328e-06, + "loss": 0.707, + "step": 25440 + }, + { + "epoch": 4.63, + "learning_rate": 1.6135238441936658e-06, + "loss": 0.5457, + "step": 25441 + }, + { + "epoch": 4.63, + "learning_rate": 1.6134692391700037e-06, + "loss": 0.0271, + "step": 25442 + }, + { + "epoch": 4.63, + "learning_rate": 1.6134146341463417e-06, + "loss": 2.5275, + "step": 25443 + }, + { + "epoch": 4.63, + "learning_rate": 1.6133600291226792e-06, + "loss": 0.7626, + "step": 25444 + }, + { + "epoch": 4.63, + "learning_rate": 1.6133054240990172e-06, + "loss": 0.0621, + "step": 25445 + }, + { + "epoch": 4.63, + "learning_rate": 1.613250819075355e-06, + "loss": 2.9372, + "step": 25446 + }, + { + "epoch": 4.63, + "learning_rate": 1.6131962140516926e-06, + "loss": 0.1569, + "step": 25447 + }, + { + "epoch": 4.63, + "learning_rate": 1.6131416090280306e-06, + "loss": 1.6085, + "step": 25448 + }, + { + "epoch": 4.63, + "learning_rate": 1.6130870040043685e-06, + "loss": 1.2475, + "step": 25449 + }, + { + "epoch": 4.63, + "learning_rate": 1.6130323989807062e-06, + "loss": 1.3471, + "step": 25450 + }, + { + "epoch": 4.63, + "learning_rate": 1.6129777939570442e-06, + "loss": 0.2315, + "step": 25451 + }, + { + "epoch": 4.63, + "learning_rate": 1.6129231889333817e-06, + "loss": 0.5403, + "step": 25452 + }, + { + "epoch": 4.63, + "learning_rate": 1.6128685839097196e-06, + "loss": 2.4146, + "step": 25453 + }, + { + "epoch": 4.63, + "learning_rate": 1.6128139788860576e-06, + "loss": 0.1199, + "step": 25454 + }, + { + "epoch": 4.63, + "learning_rate": 1.6127593738623953e-06, + "loss": 1.0477, + "step": 25455 + }, + { + "epoch": 4.63, + "learning_rate": 1.6127047688387333e-06, + "loss": 0.0423, + "step": 25456 + }, + { + "epoch": 4.63, + "learning_rate": 1.6126501638150712e-06, + "loss": 0.3336, + "step": 25457 + }, + { + "epoch": 4.63, + "learning_rate": 1.6125955587914087e-06, + "loss": 0.4849, + "step": 25458 + }, + { + "epoch": 4.63, + "learning_rate": 1.6125409537677467e-06, + "loss": 0.004, + "step": 25459 + }, + { + "epoch": 4.63, + "learning_rate": 1.6124863487440846e-06, + "loss": 1.1505, + "step": 25460 + }, + { + "epoch": 4.63, + "learning_rate": 1.6124317437204221e-06, + "loss": 0.5836, + "step": 25461 + }, + { + "epoch": 4.63, + "learning_rate": 1.61237713869676e-06, + "loss": 0.0151, + "step": 25462 + }, + { + "epoch": 4.63, + "learning_rate": 1.612322533673098e-06, + "loss": 3.4888, + "step": 25463 + }, + { + "epoch": 4.63, + "learning_rate": 1.6122679286494357e-06, + "loss": 1.7743, + "step": 25464 + }, + { + "epoch": 4.63, + "learning_rate": 1.6122133236257737e-06, + "loss": 0.2602, + "step": 25465 + }, + { + "epoch": 4.63, + "learning_rate": 1.6121587186021112e-06, + "loss": 1.3362, + "step": 25466 + }, + { + "epoch": 4.63, + "learning_rate": 1.6121041135784492e-06, + "loss": 1.7679, + "step": 25467 + }, + { + "epoch": 4.63, + "learning_rate": 1.612049508554787e-06, + "loss": 0.4042, + "step": 25468 + }, + { + "epoch": 4.63, + "learning_rate": 1.6119949035311248e-06, + "loss": 0.4706, + "step": 25469 + }, + { + "epoch": 4.63, + "learning_rate": 1.6119402985074628e-06, + "loss": 0.2665, + "step": 25470 + }, + { + "epoch": 4.63, + "learning_rate": 1.6118856934838007e-06, + "loss": 2.3301, + "step": 25471 + }, + { + "epoch": 4.63, + "learning_rate": 1.6118310884601382e-06, + "loss": 0.2272, + "step": 25472 + }, + { + "epoch": 4.63, + "learning_rate": 1.6117764834364762e-06, + "loss": 1.7389, + "step": 25473 + }, + { + "epoch": 4.63, + "learning_rate": 1.6117218784128141e-06, + "loss": 2.4571, + "step": 25474 + }, + { + "epoch": 4.63, + "learning_rate": 1.6116672733891519e-06, + "loss": 1.5852, + "step": 25475 + }, + { + "epoch": 4.63, + "learning_rate": 1.6116126683654898e-06, + "loss": 0.0657, + "step": 25476 + }, + { + "epoch": 4.63, + "learning_rate": 1.6115580633418277e-06, + "loss": 0.1124, + "step": 25477 + }, + { + "epoch": 4.63, + "learning_rate": 1.6115034583181653e-06, + "loss": 0.3338, + "step": 25478 + }, + { + "epoch": 4.63, + "learning_rate": 1.6114488532945032e-06, + "loss": 1.1955, + "step": 25479 + }, + { + "epoch": 4.63, + "learning_rate": 1.6113942482708407e-06, + "loss": 1.1528, + "step": 25480 + }, + { + "epoch": 4.63, + "learning_rate": 1.6113396432471787e-06, + "loss": 1.4539, + "step": 25481 + }, + { + "epoch": 4.63, + "learning_rate": 1.6112850382235166e-06, + "loss": 0.2795, + "step": 25482 + }, + { + "epoch": 4.63, + "learning_rate": 1.6112304331998543e-06, + "loss": 1.35, + "step": 25483 + }, + { + "epoch": 4.63, + "learning_rate": 1.6111758281761923e-06, + "loss": 0.8821, + "step": 25484 + }, + { + "epoch": 4.63, + "learning_rate": 1.6111212231525302e-06, + "loss": 0.0288, + "step": 25485 + }, + { + "epoch": 4.63, + "learning_rate": 1.6110666181288678e-06, + "loss": 1.9908, + "step": 25486 + }, + { + "epoch": 4.63, + "learning_rate": 1.6110120131052057e-06, + "loss": 0.2251, + "step": 25487 + }, + { + "epoch": 4.64, + "learning_rate": 1.6109574080815436e-06, + "loss": 0.0101, + "step": 25488 + }, + { + "epoch": 4.64, + "learning_rate": 1.6109028030578814e-06, + "loss": 0.3527, + "step": 25489 + }, + { + "epoch": 4.64, + "learning_rate": 1.6108481980342193e-06, + "loss": 0.0812, + "step": 25490 + }, + { + "epoch": 4.64, + "learning_rate": 1.6107935930105573e-06, + "loss": 0.0356, + "step": 25491 + }, + { + "epoch": 4.64, + "learning_rate": 1.6107389879868948e-06, + "loss": 1.2363, + "step": 25492 + }, + { + "epoch": 4.64, + "learning_rate": 1.6106843829632327e-06, + "loss": 0.9201, + "step": 25493 + }, + { + "epoch": 4.64, + "learning_rate": 1.6106297779395702e-06, + "loss": 0.0969, + "step": 25494 + }, + { + "epoch": 4.64, + "learning_rate": 1.6105751729159082e-06, + "loss": 2.4024, + "step": 25495 + }, + { + "epoch": 4.64, + "learning_rate": 1.6105205678922461e-06, + "loss": 0.1715, + "step": 25496 + }, + { + "epoch": 4.64, + "learning_rate": 1.6104659628685839e-06, + "loss": 0.1491, + "step": 25497 + }, + { + "epoch": 4.64, + "learning_rate": 1.6104113578449218e-06, + "loss": 1.3742, + "step": 25498 + }, + { + "epoch": 4.64, + "learning_rate": 1.6103567528212597e-06, + "loss": 0.0162, + "step": 25499 + }, + { + "epoch": 4.64, + "learning_rate": 1.6103021477975973e-06, + "loss": 0.3893, + "step": 25500 + }, + { + "epoch": 4.64, + "learning_rate": 1.6102475427739352e-06, + "loss": 0.0146, + "step": 25501 + }, + { + "epoch": 4.64, + "learning_rate": 1.6101929377502732e-06, + "loss": 0.1097, + "step": 25502 + }, + { + "epoch": 4.64, + "learning_rate": 1.6101383327266109e-06, + "loss": 1.6347, + "step": 25503 + }, + { + "epoch": 4.64, + "learning_rate": 1.6100837277029488e-06, + "loss": 0.0048, + "step": 25504 + }, + { + "epoch": 4.64, + "learning_rate": 1.6100291226792863e-06, + "loss": 0.6497, + "step": 25505 + }, + { + "epoch": 4.64, + "learning_rate": 1.6099745176556243e-06, + "loss": 0.5142, + "step": 25506 + }, + { + "epoch": 4.64, + "learning_rate": 1.6099199126319622e-06, + "loss": 0.5665, + "step": 25507 + }, + { + "epoch": 4.64, + "learning_rate": 1.6098653076083e-06, + "loss": 1.071, + "step": 25508 + }, + { + "epoch": 4.64, + "learning_rate": 1.609810702584638e-06, + "loss": 1.5938, + "step": 25509 + }, + { + "epoch": 4.64, + "learning_rate": 1.6097560975609759e-06, + "loss": 0.6189, + "step": 25510 + }, + { + "epoch": 4.64, + "learning_rate": 1.6097014925373134e-06, + "loss": 0.6238, + "step": 25511 + }, + { + "epoch": 4.64, + "learning_rate": 1.6096468875136513e-06, + "loss": 0.0105, + "step": 25512 + }, + { + "epoch": 4.64, + "learning_rate": 1.6095922824899893e-06, + "loss": 3.3405, + "step": 25513 + }, + { + "epoch": 4.64, + "learning_rate": 1.6095376774663268e-06, + "loss": 0.018, + "step": 25514 + }, + { + "epoch": 4.64, + "learning_rate": 1.6094830724426647e-06, + "loss": 1.1206, + "step": 25515 + }, + { + "epoch": 4.64, + "learning_rate": 1.6094284674190027e-06, + "loss": 1.7594, + "step": 25516 + }, + { + "epoch": 4.64, + "learning_rate": 1.6093738623953404e-06, + "loss": 0.1691, + "step": 25517 + }, + { + "epoch": 4.64, + "learning_rate": 1.6093192573716783e-06, + "loss": 0.2454, + "step": 25518 + }, + { + "epoch": 4.64, + "learning_rate": 1.6092646523480159e-06, + "loss": 0.1928, + "step": 25519 + }, + { + "epoch": 4.64, + "learning_rate": 1.6092100473243538e-06, + "loss": 0.7567, + "step": 25520 + }, + { + "epoch": 4.64, + "learning_rate": 1.6091554423006917e-06, + "loss": 0.8574, + "step": 25521 + }, + { + "epoch": 4.64, + "learning_rate": 1.6091008372770295e-06, + "loss": 0.1382, + "step": 25522 + }, + { + "epoch": 4.64, + "learning_rate": 1.6090462322533674e-06, + "loss": 0.2729, + "step": 25523 + }, + { + "epoch": 4.64, + "learning_rate": 1.6089916272297054e-06, + "loss": 1.2258, + "step": 25524 + }, + { + "epoch": 4.64, + "learning_rate": 1.6089370222060429e-06, + "loss": 0.5349, + "step": 25525 + }, + { + "epoch": 4.64, + "learning_rate": 1.6088824171823808e-06, + "loss": 0.3503, + "step": 25526 + }, + { + "epoch": 4.64, + "learning_rate": 1.6088278121587188e-06, + "loss": 0.3846, + "step": 25527 + }, + { + "epoch": 4.64, + "learning_rate": 1.6087732071350563e-06, + "loss": 2.5928, + "step": 25528 + }, + { + "epoch": 4.64, + "learning_rate": 1.6087186021113942e-06, + "loss": 0.0095, + "step": 25529 + }, + { + "epoch": 4.64, + "learning_rate": 1.6086639970877322e-06, + "loss": 0.7971, + "step": 25530 + }, + { + "epoch": 4.64, + "learning_rate": 1.60860939206407e-06, + "loss": 0.4944, + "step": 25531 + }, + { + "epoch": 4.64, + "learning_rate": 1.6085547870404079e-06, + "loss": 0.6199, + "step": 25532 + }, + { + "epoch": 4.64, + "learning_rate": 1.6085001820167454e-06, + "loss": 1.1095, + "step": 25533 + }, + { + "epoch": 4.64, + "learning_rate": 1.6084455769930833e-06, + "loss": 1.8843, + "step": 25534 + }, + { + "epoch": 4.64, + "learning_rate": 1.6083909719694213e-06, + "loss": 0.0111, + "step": 25535 + }, + { + "epoch": 4.64, + "learning_rate": 1.608336366945759e-06, + "loss": 0.2416, + "step": 25536 + }, + { + "epoch": 4.64, + "learning_rate": 1.608281761922097e-06, + "loss": 0.1081, + "step": 25537 + }, + { + "epoch": 4.64, + "learning_rate": 1.6082271568984349e-06, + "loss": 0.0329, + "step": 25538 + }, + { + "epoch": 4.64, + "learning_rate": 1.6081725518747724e-06, + "loss": 0.2984, + "step": 25539 + }, + { + "epoch": 4.64, + "learning_rate": 1.6081179468511103e-06, + "loss": 0.4347, + "step": 25540 + }, + { + "epoch": 4.64, + "learning_rate": 1.6080633418274483e-06, + "loss": 1.0999, + "step": 25541 + }, + { + "epoch": 4.64, + "learning_rate": 1.6080087368037858e-06, + "loss": 0.054, + "step": 25542 + }, + { + "epoch": 4.65, + "learning_rate": 1.6079541317801237e-06, + "loss": 0.3343, + "step": 25543 + }, + { + "epoch": 4.65, + "learning_rate": 1.6078995267564617e-06, + "loss": 0.0293, + "step": 25544 + }, + { + "epoch": 4.65, + "learning_rate": 1.6078449217327994e-06, + "loss": 0.0857, + "step": 25545 + }, + { + "epoch": 4.65, + "learning_rate": 1.6077903167091374e-06, + "loss": 0.7892, + "step": 25546 + }, + { + "epoch": 4.65, + "learning_rate": 1.6077357116854749e-06, + "loss": 0.4802, + "step": 25547 + }, + { + "epoch": 4.65, + "learning_rate": 1.6076811066618128e-06, + "loss": 0.0121, + "step": 25548 + }, + { + "epoch": 4.65, + "learning_rate": 1.6076265016381508e-06, + "loss": 1.0915, + "step": 25549 + }, + { + "epoch": 4.65, + "learning_rate": 1.6075718966144885e-06, + "loss": 0.7438, + "step": 25550 + }, + { + "epoch": 4.65, + "learning_rate": 1.6075172915908264e-06, + "loss": 0.022, + "step": 25551 + }, + { + "epoch": 4.65, + "learning_rate": 1.6074626865671644e-06, + "loss": 2.8041, + "step": 25552 + }, + { + "epoch": 4.65, + "learning_rate": 1.607408081543502e-06, + "loss": 0.7302, + "step": 25553 + }, + { + "epoch": 4.65, + "learning_rate": 1.6073534765198399e-06, + "loss": 0.1655, + "step": 25554 + }, + { + "epoch": 4.65, + "learning_rate": 1.6072988714961778e-06, + "loss": 0.7265, + "step": 25555 + }, + { + "epoch": 4.65, + "learning_rate": 1.6072442664725155e-06, + "loss": 0.02, + "step": 25556 + }, + { + "epoch": 4.65, + "learning_rate": 1.6071896614488535e-06, + "loss": 1.5045, + "step": 25557 + }, + { + "epoch": 4.65, + "learning_rate": 1.6071350564251914e-06, + "loss": 1.0581, + "step": 25558 + }, + { + "epoch": 4.65, + "learning_rate": 1.607080451401529e-06, + "loss": 1.329, + "step": 25559 + }, + { + "epoch": 4.65, + "learning_rate": 1.6070258463778669e-06, + "loss": 0.3199, + "step": 25560 + }, + { + "epoch": 4.65, + "learning_rate": 1.6069712413542044e-06, + "loss": 2.9272, + "step": 25561 + }, + { + "epoch": 4.65, + "learning_rate": 1.6069166363305423e-06, + "loss": 0.3812, + "step": 25562 + }, + { + "epoch": 4.65, + "learning_rate": 1.6068620313068803e-06, + "loss": 1.0462, + "step": 25563 + }, + { + "epoch": 4.65, + "learning_rate": 1.606807426283218e-06, + "loss": 0.5047, + "step": 25564 + }, + { + "epoch": 4.65, + "learning_rate": 1.606752821259556e-06, + "loss": 0.2463, + "step": 25565 + }, + { + "epoch": 4.65, + "learning_rate": 1.606698216235894e-06, + "loss": 0.1698, + "step": 25566 + }, + { + "epoch": 4.65, + "learning_rate": 1.6066436112122314e-06, + "loss": 0.3603, + "step": 25567 + }, + { + "epoch": 4.65, + "learning_rate": 1.6065890061885694e-06, + "loss": 1.0434, + "step": 25568 + }, + { + "epoch": 4.65, + "learning_rate": 1.6065344011649073e-06, + "loss": 1.2974, + "step": 25569 + }, + { + "epoch": 4.65, + "learning_rate": 1.606479796141245e-06, + "loss": 1.6152, + "step": 25570 + }, + { + "epoch": 4.65, + "learning_rate": 1.606425191117583e-06, + "loss": 2.2436, + "step": 25571 + }, + { + "epoch": 4.65, + "learning_rate": 1.606370586093921e-06, + "loss": 0.1735, + "step": 25572 + }, + { + "epoch": 4.65, + "learning_rate": 1.6063159810702585e-06, + "loss": 0.448, + "step": 25573 + }, + { + "epoch": 4.65, + "learning_rate": 1.6062613760465964e-06, + "loss": 0.0858, + "step": 25574 + }, + { + "epoch": 4.65, + "learning_rate": 1.606206771022934e-06, + "loss": 1.164, + "step": 25575 + }, + { + "epoch": 4.65, + "learning_rate": 1.6061521659992719e-06, + "loss": 0.9785, + "step": 25576 + }, + { + "epoch": 4.65, + "learning_rate": 1.6060975609756098e-06, + "loss": 1.4099, + "step": 25577 + }, + { + "epoch": 4.65, + "learning_rate": 1.6060429559519475e-06, + "loss": 0.3952, + "step": 25578 + }, + { + "epoch": 4.65, + "learning_rate": 1.6059883509282855e-06, + "loss": 0.3007, + "step": 25579 + }, + { + "epoch": 4.65, + "learning_rate": 1.6059337459046234e-06, + "loss": 0.9337, + "step": 25580 + }, + { + "epoch": 4.65, + "learning_rate": 1.605879140880961e-06, + "loss": 0.3108, + "step": 25581 + }, + { + "epoch": 4.65, + "learning_rate": 1.6058245358572989e-06, + "loss": 1.6831, + "step": 25582 + }, + { + "epoch": 4.65, + "learning_rate": 1.6057699308336368e-06, + "loss": 0.1522, + "step": 25583 + }, + { + "epoch": 4.65, + "learning_rate": 1.6057153258099746e-06, + "loss": 1.3102, + "step": 25584 + }, + { + "epoch": 4.65, + "learning_rate": 1.6056607207863125e-06, + "loss": 0.3104, + "step": 25585 + }, + { + "epoch": 4.65, + "learning_rate": 1.60560611576265e-06, + "loss": 0.7118, + "step": 25586 + }, + { + "epoch": 4.65, + "learning_rate": 1.605551510738988e-06, + "loss": 1.2885, + "step": 25587 + }, + { + "epoch": 4.65, + "learning_rate": 1.605496905715326e-06, + "loss": 0.3828, + "step": 25588 + }, + { + "epoch": 4.65, + "learning_rate": 1.6054423006916636e-06, + "loss": 1.5265, + "step": 25589 + }, + { + "epoch": 4.65, + "learning_rate": 1.6053876956680014e-06, + "loss": 1.4079, + "step": 25590 + }, + { + "epoch": 4.65, + "learning_rate": 1.6053330906443393e-06, + "loss": 0.0145, + "step": 25591 + }, + { + "epoch": 4.65, + "learning_rate": 1.605278485620677e-06, + "loss": 0.0931, + "step": 25592 + }, + { + "epoch": 4.65, + "learning_rate": 1.605223880597015e-06, + "loss": 0.4025, + "step": 25593 + }, + { + "epoch": 4.65, + "learning_rate": 1.605169275573353e-06, + "loss": 1.9597, + "step": 25594 + }, + { + "epoch": 4.65, + "learning_rate": 1.6051146705496905e-06, + "loss": 0.3514, + "step": 25595 + }, + { + "epoch": 4.65, + "learning_rate": 1.6050600655260284e-06, + "loss": 0.1887, + "step": 25596 + }, + { + "epoch": 4.65, + "learning_rate": 1.6050054605023663e-06, + "loss": 0.4858, + "step": 25597 + }, + { + "epoch": 4.66, + "learning_rate": 1.604950855478704e-06, + "loss": 0.9461, + "step": 25598 + }, + { + "epoch": 4.66, + "learning_rate": 1.604896250455042e-06, + "loss": 1.268, + "step": 25599 + }, + { + "epoch": 4.66, + "learning_rate": 1.6048416454313795e-06, + "loss": 0.0268, + "step": 25600 + }, + { + "epoch": 4.66, + "learning_rate": 1.6047870404077175e-06, + "loss": 0.1069, + "step": 25601 + }, + { + "epoch": 4.66, + "learning_rate": 1.6047324353840554e-06, + "loss": 0.8392, + "step": 25602 + }, + { + "epoch": 4.66, + "learning_rate": 1.6046778303603932e-06, + "loss": 1.2273, + "step": 25603 + }, + { + "epoch": 4.66, + "learning_rate": 1.604623225336731e-06, + "loss": 0.5004, + "step": 25604 + }, + { + "epoch": 4.66, + "learning_rate": 1.604568620313069e-06, + "loss": 0.0167, + "step": 25605 + }, + { + "epoch": 4.66, + "learning_rate": 1.6045140152894066e-06, + "loss": 1.1125, + "step": 25606 + }, + { + "epoch": 4.66, + "learning_rate": 1.6044594102657445e-06, + "loss": 0.0256, + "step": 25607 + }, + { + "epoch": 4.66, + "learning_rate": 1.6044048052420824e-06, + "loss": 0.1893, + "step": 25608 + }, + { + "epoch": 4.66, + "learning_rate": 1.60435020021842e-06, + "loss": 2.9534, + "step": 25609 + }, + { + "epoch": 4.66, + "learning_rate": 1.604295595194758e-06, + "loss": 0.1826, + "step": 25610 + }, + { + "epoch": 4.66, + "learning_rate": 1.6042409901710959e-06, + "loss": 1.4005, + "step": 25611 + }, + { + "epoch": 4.66, + "learning_rate": 1.6041863851474336e-06, + "loss": 0.249, + "step": 25612 + }, + { + "epoch": 4.66, + "learning_rate": 1.6041317801237715e-06, + "loss": 0.3159, + "step": 25613 + }, + { + "epoch": 4.66, + "learning_rate": 1.604077175100109e-06, + "loss": 0.1438, + "step": 25614 + }, + { + "epoch": 4.66, + "learning_rate": 1.604022570076447e-06, + "loss": 0.2722, + "step": 25615 + }, + { + "epoch": 4.66, + "learning_rate": 1.603967965052785e-06, + "loss": 0.5167, + "step": 25616 + }, + { + "epoch": 4.66, + "learning_rate": 1.6039133600291227e-06, + "loss": 2.1899, + "step": 25617 + }, + { + "epoch": 4.66, + "learning_rate": 1.6038587550054606e-06, + "loss": 0.013, + "step": 25618 + }, + { + "epoch": 4.66, + "learning_rate": 1.6038041499817986e-06, + "loss": 0.4298, + "step": 25619 + }, + { + "epoch": 4.66, + "learning_rate": 1.603749544958136e-06, + "loss": 0.0214, + "step": 25620 + }, + { + "epoch": 4.66, + "learning_rate": 1.603694939934474e-06, + "loss": 0.0377, + "step": 25621 + }, + { + "epoch": 4.66, + "learning_rate": 1.603640334910812e-06, + "loss": 0.0353, + "step": 25622 + }, + { + "epoch": 4.66, + "learning_rate": 1.6035857298871495e-06, + "loss": 0.7235, + "step": 25623 + }, + { + "epoch": 4.66, + "learning_rate": 1.6035311248634874e-06, + "loss": 1.3221, + "step": 25624 + }, + { + "epoch": 4.66, + "learning_rate": 1.6034765198398254e-06, + "loss": 0.2684, + "step": 25625 + }, + { + "epoch": 4.66, + "learning_rate": 1.603421914816163e-06, + "loss": 0.0121, + "step": 25626 + }, + { + "epoch": 4.66, + "learning_rate": 1.603367309792501e-06, + "loss": 1.668, + "step": 25627 + }, + { + "epoch": 4.66, + "learning_rate": 1.6033127047688386e-06, + "loss": 0.0078, + "step": 25628 + }, + { + "epoch": 4.66, + "learning_rate": 1.6032580997451765e-06, + "loss": 0.0946, + "step": 25629 + }, + { + "epoch": 4.66, + "learning_rate": 1.6032034947215144e-06, + "loss": 0.1324, + "step": 25630 + }, + { + "epoch": 4.66, + "learning_rate": 1.6031488896978522e-06, + "loss": 0.9421, + "step": 25631 + }, + { + "epoch": 4.66, + "learning_rate": 1.6030942846741901e-06, + "loss": 1.2076, + "step": 25632 + }, + { + "epoch": 4.66, + "learning_rate": 1.603039679650528e-06, + "loss": 0.0741, + "step": 25633 + }, + { + "epoch": 4.66, + "learning_rate": 1.6029850746268656e-06, + "loss": 0.3303, + "step": 25634 + }, + { + "epoch": 4.66, + "learning_rate": 1.6029304696032035e-06, + "loss": 1.9605, + "step": 25635 + }, + { + "epoch": 4.66, + "learning_rate": 1.6028758645795415e-06, + "loss": 2.2588, + "step": 25636 + }, + { + "epoch": 4.66, + "learning_rate": 1.6028212595558792e-06, + "loss": 1.9897, + "step": 25637 + }, + { + "epoch": 4.66, + "learning_rate": 1.6027666545322171e-06, + "loss": 1.4809, + "step": 25638 + }, + { + "epoch": 4.66, + "learning_rate": 1.602712049508555e-06, + "loss": 1.0456, + "step": 25639 + }, + { + "epoch": 4.66, + "learning_rate": 1.6026574444848926e-06, + "loss": 2.0982, + "step": 25640 + }, + { + "epoch": 4.66, + "learning_rate": 1.6026028394612306e-06, + "loss": 0.0137, + "step": 25641 + }, + { + "epoch": 4.66, + "learning_rate": 1.602548234437568e-06, + "loss": 0.0112, + "step": 25642 + }, + { + "epoch": 4.66, + "learning_rate": 1.602493629413906e-06, + "loss": 0.2916, + "step": 25643 + }, + { + "epoch": 4.66, + "learning_rate": 1.602439024390244e-06, + "loss": 0.1388, + "step": 25644 + }, + { + "epoch": 4.66, + "learning_rate": 1.6023844193665817e-06, + "loss": 0.1502, + "step": 25645 + }, + { + "epoch": 4.66, + "learning_rate": 1.6023298143429196e-06, + "loss": 0.3577, + "step": 25646 + }, + { + "epoch": 4.66, + "learning_rate": 1.6022752093192576e-06, + "loss": 1.0217, + "step": 25647 + }, + { + "epoch": 4.66, + "learning_rate": 1.602220604295595e-06, + "loss": 0.9896, + "step": 25648 + }, + { + "epoch": 4.66, + "learning_rate": 1.602165999271933e-06, + "loss": 1.1926, + "step": 25649 + }, + { + "epoch": 4.66, + "learning_rate": 1.602111394248271e-06, + "loss": 0.007, + "step": 25650 + }, + { + "epoch": 4.66, + "learning_rate": 1.6020567892246087e-06, + "loss": 0.7049, + "step": 25651 + }, + { + "epoch": 4.66, + "learning_rate": 1.6020021842009467e-06, + "loss": 0.0103, + "step": 25652 + }, + { + "epoch": 4.67, + "learning_rate": 1.6019475791772846e-06, + "loss": 0.0099, + "step": 25653 + }, + { + "epoch": 4.67, + "learning_rate": 1.6018929741536221e-06, + "loss": 2.1156, + "step": 25654 + }, + { + "epoch": 4.67, + "learning_rate": 1.60183836912996e-06, + "loss": 0.137, + "step": 25655 + }, + { + "epoch": 4.67, + "learning_rate": 1.6017837641062976e-06, + "loss": 0.4844, + "step": 25656 + }, + { + "epoch": 4.67, + "learning_rate": 1.6017291590826355e-06, + "loss": 2.6579, + "step": 25657 + }, + { + "epoch": 4.67, + "learning_rate": 1.6016745540589735e-06, + "loss": 0.081, + "step": 25658 + }, + { + "epoch": 4.67, + "learning_rate": 1.6016199490353112e-06, + "loss": 0.2373, + "step": 25659 + }, + { + "epoch": 4.67, + "learning_rate": 1.6015653440116492e-06, + "loss": 1.1251, + "step": 25660 + }, + { + "epoch": 4.67, + "learning_rate": 1.601510738987987e-06, + "loss": 1.1723, + "step": 25661 + }, + { + "epoch": 4.67, + "learning_rate": 1.6014561339643246e-06, + "loss": 2.6095, + "step": 25662 + }, + { + "epoch": 4.67, + "learning_rate": 1.6014015289406626e-06, + "loss": 0.0847, + "step": 25663 + }, + { + "epoch": 4.67, + "learning_rate": 1.6013469239170005e-06, + "loss": 1.145, + "step": 25664 + }, + { + "epoch": 4.67, + "learning_rate": 1.6012923188933382e-06, + "loss": 0.615, + "step": 25665 + }, + { + "epoch": 4.67, + "learning_rate": 1.6012377138696762e-06, + "loss": 0.1429, + "step": 25666 + }, + { + "epoch": 4.67, + "learning_rate": 1.6011831088460137e-06, + "loss": 0.0442, + "step": 25667 + }, + { + "epoch": 4.67, + "learning_rate": 1.6011285038223516e-06, + "loss": 0.6253, + "step": 25668 + }, + { + "epoch": 4.67, + "learning_rate": 1.6010738987986896e-06, + "loss": 0.3692, + "step": 25669 + }, + { + "epoch": 4.67, + "learning_rate": 1.601019293775027e-06, + "loss": 0.8663, + "step": 25670 + }, + { + "epoch": 4.67, + "learning_rate": 1.600964688751365e-06, + "loss": 0.9035, + "step": 25671 + }, + { + "epoch": 4.67, + "learning_rate": 1.600910083727703e-06, + "loss": 0.2611, + "step": 25672 + }, + { + "epoch": 4.67, + "learning_rate": 1.6008554787040407e-06, + "loss": 0.0088, + "step": 25673 + }, + { + "epoch": 4.67, + "learning_rate": 1.6008008736803787e-06, + "loss": 0.0168, + "step": 25674 + }, + { + "epoch": 4.67, + "learning_rate": 1.6007462686567166e-06, + "loss": 1.581, + "step": 25675 + }, + { + "epoch": 4.67, + "learning_rate": 1.6006916636330541e-06, + "loss": 0.0378, + "step": 25676 + }, + { + "epoch": 4.67, + "learning_rate": 1.600637058609392e-06, + "loss": 0.6001, + "step": 25677 + }, + { + "epoch": 4.67, + "learning_rate": 1.60058245358573e-06, + "loss": 1.4221, + "step": 25678 + }, + { + "epoch": 4.67, + "learning_rate": 1.6005278485620677e-06, + "loss": 2.0412, + "step": 25679 + }, + { + "epoch": 4.67, + "learning_rate": 1.6004732435384057e-06, + "loss": 0.1509, + "step": 25680 + }, + { + "epoch": 4.67, + "learning_rate": 1.6004186385147432e-06, + "loss": 1.9617, + "step": 25681 + }, + { + "epoch": 4.67, + "learning_rate": 1.6003640334910812e-06, + "loss": 0.0198, + "step": 25682 + }, + { + "epoch": 4.67, + "learning_rate": 1.600309428467419e-06, + "loss": 0.5883, + "step": 25683 + }, + { + "epoch": 4.67, + "learning_rate": 1.6002548234437568e-06, + "loss": 0.0495, + "step": 25684 + }, + { + "epoch": 4.67, + "learning_rate": 1.6002002184200948e-06, + "loss": 0.5092, + "step": 25685 + }, + { + "epoch": 4.67, + "learning_rate": 1.6001456133964327e-06, + "loss": 1.2269, + "step": 25686 + }, + { + "epoch": 4.67, + "learning_rate": 1.6000910083727702e-06, + "loss": 0.1369, + "step": 25687 + }, + { + "epoch": 4.67, + "learning_rate": 1.6000364033491082e-06, + "loss": 0.3863, + "step": 25688 + }, + { + "epoch": 4.67, + "learning_rate": 1.5999817983254461e-06, + "loss": 1.3054, + "step": 25689 + }, + { + "epoch": 4.67, + "learning_rate": 1.5999271933017836e-06, + "loss": 0.8451, + "step": 25690 + }, + { + "epoch": 4.67, + "learning_rate": 1.5998725882781216e-06, + "loss": 2.4182, + "step": 25691 + }, + { + "epoch": 4.67, + "learning_rate": 1.5998179832544595e-06, + "loss": 0.1993, + "step": 25692 + }, + { + "epoch": 4.67, + "learning_rate": 1.5997633782307973e-06, + "loss": 2.358, + "step": 25693 + }, + { + "epoch": 4.67, + "learning_rate": 1.5997087732071352e-06, + "loss": 0.3216, + "step": 25694 + }, + { + "epoch": 4.67, + "learning_rate": 1.5996541681834727e-06, + "loss": 0.7999, + "step": 25695 + }, + { + "epoch": 4.67, + "learning_rate": 1.5995995631598107e-06, + "loss": 0.0059, + "step": 25696 + }, + { + "epoch": 4.67, + "learning_rate": 1.5995449581361486e-06, + "loss": 0.1858, + "step": 25697 + }, + { + "epoch": 4.67, + "learning_rate": 1.5994903531124863e-06, + "loss": 0.289, + "step": 25698 + }, + { + "epoch": 4.67, + "learning_rate": 1.5994357480888243e-06, + "loss": 0.0447, + "step": 25699 + }, + { + "epoch": 4.67, + "learning_rate": 1.5993811430651622e-06, + "loss": 0.0633, + "step": 25700 + }, + { + "epoch": 4.67, + "learning_rate": 1.5993265380414997e-06, + "loss": 0.027, + "step": 25701 + }, + { + "epoch": 4.67, + "learning_rate": 1.5992719330178377e-06, + "loss": 0.4842, + "step": 25702 + }, + { + "epoch": 4.67, + "learning_rate": 1.5992173279941756e-06, + "loss": 1.0602, + "step": 25703 + }, + { + "epoch": 4.67, + "learning_rate": 1.5991627229705132e-06, + "loss": 1.2043, + "step": 25704 + }, + { + "epoch": 4.67, + "learning_rate": 1.599108117946851e-06, + "loss": 2.0339, + "step": 25705 + }, + { + "epoch": 4.67, + "learning_rate": 1.599053512923189e-06, + "loss": 0.2702, + "step": 25706 + }, + { + "epoch": 4.67, + "learning_rate": 1.5989989078995268e-06, + "loss": 0.4561, + "step": 25707 + }, + { + "epoch": 4.68, + "learning_rate": 1.5989443028758647e-06, + "loss": 1.0837, + "step": 25708 + }, + { + "epoch": 4.68, + "learning_rate": 1.5988896978522022e-06, + "loss": 0.1481, + "step": 25709 + }, + { + "epoch": 4.68, + "learning_rate": 1.5988350928285402e-06, + "loss": 1.2063, + "step": 25710 + }, + { + "epoch": 4.68, + "learning_rate": 1.5987804878048781e-06, + "loss": 0.1741, + "step": 25711 + }, + { + "epoch": 4.68, + "learning_rate": 1.5987258827812159e-06, + "loss": 0.3893, + "step": 25712 + }, + { + "epoch": 4.68, + "learning_rate": 1.5986712777575538e-06, + "loss": 0.899, + "step": 25713 + }, + { + "epoch": 4.68, + "learning_rate": 1.5986166727338917e-06, + "loss": 1.6482, + "step": 25714 + }, + { + "epoch": 4.68, + "learning_rate": 1.5985620677102293e-06, + "loss": 0.2209, + "step": 25715 + }, + { + "epoch": 4.68, + "learning_rate": 1.5985074626865672e-06, + "loss": 1.5684, + "step": 25716 + }, + { + "epoch": 4.68, + "learning_rate": 1.5984528576629051e-06, + "loss": 0.8585, + "step": 25717 + }, + { + "epoch": 4.68, + "learning_rate": 1.5983982526392429e-06, + "loss": 0.7461, + "step": 25718 + }, + { + "epoch": 4.68, + "learning_rate": 1.5983436476155806e-06, + "loss": 0.237, + "step": 25719 + }, + { + "epoch": 4.68, + "learning_rate": 1.5982890425919186e-06, + "loss": 0.008, + "step": 25720 + }, + { + "epoch": 4.68, + "learning_rate": 1.5982344375682563e-06, + "loss": 0.4007, + "step": 25721 + }, + { + "epoch": 4.68, + "learning_rate": 1.5981798325445942e-06, + "loss": 0.3796, + "step": 25722 + }, + { + "epoch": 4.68, + "learning_rate": 1.5981252275209317e-06, + "loss": 0.2674, + "step": 25723 + }, + { + "epoch": 4.68, + "learning_rate": 1.5980706224972697e-06, + "loss": 0.2632, + "step": 25724 + }, + { + "epoch": 4.68, + "learning_rate": 1.5980160174736076e-06, + "loss": 1.0063, + "step": 25725 + }, + { + "epoch": 4.68, + "learning_rate": 1.5979614124499454e-06, + "loss": 0.373, + "step": 25726 + }, + { + "epoch": 4.68, + "learning_rate": 1.5979068074262833e-06, + "loss": 0.5977, + "step": 25727 + }, + { + "epoch": 4.68, + "learning_rate": 1.5978522024026213e-06, + "loss": 0.3119, + "step": 25728 + }, + { + "epoch": 4.68, + "learning_rate": 1.5977975973789588e-06, + "loss": 1.4686, + "step": 25729 + }, + { + "epoch": 4.68, + "learning_rate": 1.5977429923552967e-06, + "loss": 0.4716, + "step": 25730 + }, + { + "epoch": 4.68, + "learning_rate": 1.5976883873316347e-06, + "loss": 0.2114, + "step": 25731 + }, + { + "epoch": 4.68, + "learning_rate": 1.5976337823079724e-06, + "loss": 1.2686, + "step": 25732 + }, + { + "epoch": 4.68, + "learning_rate": 1.5975791772843103e-06, + "loss": 0.0255, + "step": 25733 + }, + { + "epoch": 4.68, + "learning_rate": 1.5975245722606483e-06, + "loss": 1.6154, + "step": 25734 + }, + { + "epoch": 4.68, + "learning_rate": 1.5974699672369858e-06, + "loss": 0.8247, + "step": 25735 + }, + { + "epoch": 4.68, + "learning_rate": 1.5974153622133237e-06, + "loss": 0.326, + "step": 25736 + }, + { + "epoch": 4.68, + "learning_rate": 1.5973607571896613e-06, + "loss": 0.0307, + "step": 25737 + }, + { + "epoch": 4.68, + "learning_rate": 1.5973061521659992e-06, + "loss": 0.1616, + "step": 25738 + }, + { + "epoch": 4.68, + "learning_rate": 1.5972515471423371e-06, + "loss": 1.4789, + "step": 25739 + }, + { + "epoch": 4.68, + "learning_rate": 1.5971969421186749e-06, + "loss": 0.7467, + "step": 25740 + }, + { + "epoch": 4.68, + "learning_rate": 1.5971423370950128e-06, + "loss": 0.0935, + "step": 25741 + }, + { + "epoch": 4.68, + "learning_rate": 1.5970877320713508e-06, + "loss": 0.5341, + "step": 25742 + }, + { + "epoch": 4.68, + "learning_rate": 1.5970331270476883e-06, + "loss": 1.8508, + "step": 25743 + }, + { + "epoch": 4.68, + "learning_rate": 1.5969785220240262e-06, + "loss": 0.6239, + "step": 25744 + }, + { + "epoch": 4.68, + "learning_rate": 1.5969239170003642e-06, + "loss": 1.5979, + "step": 25745 + }, + { + "epoch": 4.68, + "learning_rate": 1.596869311976702e-06, + "loss": 0.687, + "step": 25746 + }, + { + "epoch": 4.68, + "learning_rate": 1.5968147069530398e-06, + "loss": 1.125, + "step": 25747 + }, + { + "epoch": 4.68, + "learning_rate": 1.5967601019293774e-06, + "loss": 0.1452, + "step": 25748 + }, + { + "epoch": 4.68, + "learning_rate": 1.5967054969057153e-06, + "loss": 0.0168, + "step": 25749 + }, + { + "epoch": 4.68, + "learning_rate": 1.5966508918820533e-06, + "loss": 0.0377, + "step": 25750 + }, + { + "epoch": 4.68, + "learning_rate": 1.5965962868583908e-06, + "loss": 0.2593, + "step": 25751 + }, + { + "epoch": 4.68, + "learning_rate": 1.5965416818347287e-06, + "loss": 0.6455, + "step": 25752 + }, + { + "epoch": 4.68, + "learning_rate": 1.5964870768110667e-06, + "loss": 0.7951, + "step": 25753 + }, + { + "epoch": 4.68, + "learning_rate": 1.5964324717874044e-06, + "loss": 0.0128, + "step": 25754 + }, + { + "epoch": 4.68, + "learning_rate": 1.5963778667637423e-06, + "loss": 0.7191, + "step": 25755 + }, + { + "epoch": 4.68, + "learning_rate": 1.5963232617400803e-06, + "loss": 0.3677, + "step": 25756 + }, + { + "epoch": 4.68, + "learning_rate": 1.5962686567164178e-06, + "loss": 0.4157, + "step": 25757 + }, + { + "epoch": 4.68, + "learning_rate": 1.5962140516927557e-06, + "loss": 1.745, + "step": 25758 + }, + { + "epoch": 4.68, + "learning_rate": 1.5961594466690937e-06, + "loss": 0.2188, + "step": 25759 + }, + { + "epoch": 4.68, + "learning_rate": 1.5961048416454314e-06, + "loss": 0.2267, + "step": 25760 + }, + { + "epoch": 4.68, + "learning_rate": 1.5960502366217694e-06, + "loss": 0.1656, + "step": 25761 + }, + { + "epoch": 4.68, + "learning_rate": 1.5959956315981069e-06, + "loss": 0.9048, + "step": 25762 + }, + { + "epoch": 4.69, + "learning_rate": 1.5959410265744448e-06, + "loss": 0.0299, + "step": 25763 + }, + { + "epoch": 4.69, + "learning_rate": 1.5958864215507828e-06, + "loss": 0.531, + "step": 25764 + }, + { + "epoch": 4.69, + "learning_rate": 1.5958318165271205e-06, + "loss": 2.1181, + "step": 25765 + }, + { + "epoch": 4.69, + "learning_rate": 1.5957772115034584e-06, + "loss": 0.0346, + "step": 25766 + }, + { + "epoch": 4.69, + "learning_rate": 1.5957226064797964e-06, + "loss": 0.0091, + "step": 25767 + }, + { + "epoch": 4.69, + "learning_rate": 1.595668001456134e-06, + "loss": 0.2123, + "step": 25768 + }, + { + "epoch": 4.69, + "learning_rate": 1.5956133964324719e-06, + "loss": 1.746, + "step": 25769 + }, + { + "epoch": 4.69, + "learning_rate": 1.5955587914088098e-06, + "loss": 0.0693, + "step": 25770 + }, + { + "epoch": 4.69, + "learning_rate": 1.5955041863851473e-06, + "loss": 0.0636, + "step": 25771 + }, + { + "epoch": 4.69, + "learning_rate": 1.5954495813614853e-06, + "loss": 0.4675, + "step": 25772 + }, + { + "epoch": 4.69, + "learning_rate": 1.5953949763378232e-06, + "loss": 0.1653, + "step": 25773 + }, + { + "epoch": 4.69, + "learning_rate": 1.595340371314161e-06, + "loss": 0.0678, + "step": 25774 + }, + { + "epoch": 4.69, + "learning_rate": 1.5952857662904989e-06, + "loss": 0.7888, + "step": 25775 + }, + { + "epoch": 4.69, + "learning_rate": 1.5952311612668364e-06, + "loss": 1.3294, + "step": 25776 + }, + { + "epoch": 4.69, + "learning_rate": 1.5951765562431743e-06, + "loss": 0.1065, + "step": 25777 + }, + { + "epoch": 4.69, + "learning_rate": 1.5951219512195123e-06, + "loss": 0.392, + "step": 25778 + }, + { + "epoch": 4.69, + "learning_rate": 1.59506734619585e-06, + "loss": 0.0482, + "step": 25779 + }, + { + "epoch": 4.69, + "learning_rate": 1.595012741172188e-06, + "loss": 0.7307, + "step": 25780 + }, + { + "epoch": 4.69, + "learning_rate": 1.594958136148526e-06, + "loss": 0.0106, + "step": 25781 + }, + { + "epoch": 4.69, + "learning_rate": 1.5949035311248634e-06, + "loss": 0.1947, + "step": 25782 + }, + { + "epoch": 4.69, + "learning_rate": 1.5948489261012014e-06, + "loss": 0.5924, + "step": 25783 + }, + { + "epoch": 4.69, + "learning_rate": 1.5947943210775393e-06, + "loss": 0.9428, + "step": 25784 + }, + { + "epoch": 4.69, + "learning_rate": 1.5947397160538768e-06, + "loss": 0.1554, + "step": 25785 + }, + { + "epoch": 4.69, + "learning_rate": 1.5946851110302148e-06, + "loss": 0.0097, + "step": 25786 + }, + { + "epoch": 4.69, + "learning_rate": 1.5946305060065527e-06, + "loss": 1.9199, + "step": 25787 + }, + { + "epoch": 4.69, + "learning_rate": 1.5945759009828904e-06, + "loss": 0.0243, + "step": 25788 + }, + { + "epoch": 4.69, + "learning_rate": 1.5945212959592284e-06, + "loss": 0.3196, + "step": 25789 + }, + { + "epoch": 4.69, + "learning_rate": 1.594466690935566e-06, + "loss": 0.029, + "step": 25790 + }, + { + "epoch": 4.69, + "learning_rate": 1.5944120859119039e-06, + "loss": 0.1705, + "step": 25791 + }, + { + "epoch": 4.69, + "learning_rate": 1.5943574808882418e-06, + "loss": 1.4685, + "step": 25792 + }, + { + "epoch": 4.69, + "learning_rate": 1.5943028758645795e-06, + "loss": 0.2076, + "step": 25793 + }, + { + "epoch": 4.69, + "learning_rate": 1.5942482708409175e-06, + "loss": 0.9883, + "step": 25794 + }, + { + "epoch": 4.69, + "learning_rate": 1.5941936658172554e-06, + "loss": 0.0508, + "step": 25795 + }, + { + "epoch": 4.69, + "learning_rate": 1.594139060793593e-06, + "loss": 0.1841, + "step": 25796 + }, + { + "epoch": 4.69, + "learning_rate": 1.5940844557699309e-06, + "loss": 0.6157, + "step": 25797 + }, + { + "epoch": 4.69, + "learning_rate": 1.5940298507462688e-06, + "loss": 0.0999, + "step": 25798 + }, + { + "epoch": 4.69, + "learning_rate": 1.5939752457226063e-06, + "loss": 0.2735, + "step": 25799 + }, + { + "epoch": 4.69, + "learning_rate": 1.5939206406989443e-06, + "loss": 0.6377, + "step": 25800 + }, + { + "epoch": 4.69, + "learning_rate": 1.5938660356752822e-06, + "loss": 0.3559, + "step": 25801 + }, + { + "epoch": 4.69, + "learning_rate": 1.59381143065162e-06, + "loss": 0.3569, + "step": 25802 + }, + { + "epoch": 4.69, + "learning_rate": 1.593756825627958e-06, + "loss": 0.1519, + "step": 25803 + }, + { + "epoch": 4.69, + "learning_rate": 1.5937022206042954e-06, + "loss": 0.4686, + "step": 25804 + }, + { + "epoch": 4.69, + "learning_rate": 1.5936476155806334e-06, + "loss": 0.2876, + "step": 25805 + }, + { + "epoch": 4.69, + "learning_rate": 1.5935930105569713e-06, + "loss": 0.6592, + "step": 25806 + }, + { + "epoch": 4.69, + "learning_rate": 1.593538405533309e-06, + "loss": 0.0534, + "step": 25807 + }, + { + "epoch": 4.69, + "learning_rate": 1.593483800509647e-06, + "loss": 0.0248, + "step": 25808 + }, + { + "epoch": 4.69, + "learning_rate": 1.593429195485985e-06, + "loss": 0.013, + "step": 25809 + }, + { + "epoch": 4.69, + "learning_rate": 1.5933745904623224e-06, + "loss": 0.986, + "step": 25810 + }, + { + "epoch": 4.69, + "learning_rate": 1.5933199854386604e-06, + "loss": 1.8344, + "step": 25811 + }, + { + "epoch": 4.69, + "learning_rate": 1.5932653804149983e-06, + "loss": 0.5653, + "step": 25812 + }, + { + "epoch": 4.69, + "learning_rate": 1.593210775391336e-06, + "loss": 3.5749, + "step": 25813 + }, + { + "epoch": 4.69, + "learning_rate": 1.593156170367674e-06, + "loss": 0.8043, + "step": 25814 + }, + { + "epoch": 4.69, + "learning_rate": 1.5931015653440115e-06, + "loss": 1.9695, + "step": 25815 + }, + { + "epoch": 4.69, + "learning_rate": 1.5930469603203495e-06, + "loss": 0.1094, + "step": 25816 + }, + { + "epoch": 4.69, + "learning_rate": 1.5929923552966874e-06, + "loss": 0.5027, + "step": 25817 + }, + { + "epoch": 4.7, + "learning_rate": 1.592937750273025e-06, + "loss": 1.7661, + "step": 25818 + }, + { + "epoch": 4.7, + "learning_rate": 1.5928831452493629e-06, + "loss": 0.2697, + "step": 25819 + }, + { + "epoch": 4.7, + "learning_rate": 1.5928285402257008e-06, + "loss": 0.3867, + "step": 25820 + }, + { + "epoch": 4.7, + "learning_rate": 1.5927739352020386e-06, + "loss": 0.0218, + "step": 25821 + }, + { + "epoch": 4.7, + "learning_rate": 1.5927193301783765e-06, + "loss": 0.8148, + "step": 25822 + }, + { + "epoch": 4.7, + "learning_rate": 1.5926647251547144e-06, + "loss": 1.0058, + "step": 25823 + }, + { + "epoch": 4.7, + "learning_rate": 1.592610120131052e-06, + "loss": 0.4565, + "step": 25824 + }, + { + "epoch": 4.7, + "learning_rate": 1.59255551510739e-06, + "loss": 0.1288, + "step": 25825 + }, + { + "epoch": 4.7, + "learning_rate": 1.5925009100837278e-06, + "loss": 0.4385, + "step": 25826 + }, + { + "epoch": 4.7, + "learning_rate": 1.5924463050600656e-06, + "loss": 0.2707, + "step": 25827 + }, + { + "epoch": 4.7, + "learning_rate": 1.5923917000364035e-06, + "loss": 0.9622, + "step": 25828 + }, + { + "epoch": 4.7, + "learning_rate": 1.592337095012741e-06, + "loss": 1.7902, + "step": 25829 + }, + { + "epoch": 4.7, + "learning_rate": 1.592282489989079e-06, + "loss": 1.0007, + "step": 25830 + }, + { + "epoch": 4.7, + "learning_rate": 1.592227884965417e-06, + "loss": 0.2606, + "step": 25831 + }, + { + "epoch": 4.7, + "learning_rate": 1.5921732799417545e-06, + "loss": 1.4819, + "step": 25832 + }, + { + "epoch": 4.7, + "learning_rate": 1.5921186749180924e-06, + "loss": 0.9614, + "step": 25833 + }, + { + "epoch": 4.7, + "learning_rate": 1.5920640698944303e-06, + "loss": 1.1522, + "step": 25834 + }, + { + "epoch": 4.7, + "learning_rate": 1.592009464870768e-06, + "loss": 0.3614, + "step": 25835 + }, + { + "epoch": 4.7, + "learning_rate": 1.591954859847106e-06, + "loss": 0.3148, + "step": 25836 + }, + { + "epoch": 4.7, + "learning_rate": 1.591900254823444e-06, + "loss": 0.665, + "step": 25837 + }, + { + "epoch": 4.7, + "learning_rate": 1.5918456497997815e-06, + "loss": 1.8216, + "step": 25838 + }, + { + "epoch": 4.7, + "learning_rate": 1.5917910447761194e-06, + "loss": 0.1835, + "step": 25839 + }, + { + "epoch": 4.7, + "learning_rate": 1.5917364397524574e-06, + "loss": 0.312, + "step": 25840 + }, + { + "epoch": 4.7, + "learning_rate": 1.591681834728795e-06, + "loss": 0.0654, + "step": 25841 + }, + { + "epoch": 4.7, + "learning_rate": 1.591627229705133e-06, + "loss": 0.056, + "step": 25842 + }, + { + "epoch": 4.7, + "learning_rate": 1.5915726246814706e-06, + "loss": 0.5219, + "step": 25843 + }, + { + "epoch": 4.7, + "learning_rate": 1.5915180196578085e-06, + "loss": 0.0159, + "step": 25844 + }, + { + "epoch": 4.7, + "learning_rate": 1.5914634146341464e-06, + "loss": 0.2677, + "step": 25845 + }, + { + "epoch": 4.7, + "learning_rate": 1.5914088096104842e-06, + "loss": 0.4388, + "step": 25846 + }, + { + "epoch": 4.7, + "learning_rate": 1.5913542045868221e-06, + "loss": 0.2698, + "step": 25847 + }, + { + "epoch": 4.7, + "learning_rate": 1.59129959956316e-06, + "loss": 0.0157, + "step": 25848 + }, + { + "epoch": 4.7, + "learning_rate": 1.5912449945394976e-06, + "loss": 0.291, + "step": 25849 + }, + { + "epoch": 4.7, + "learning_rate": 1.5911903895158355e-06, + "loss": 1.0585, + "step": 25850 + }, + { + "epoch": 4.7, + "learning_rate": 1.5911357844921735e-06, + "loss": 0.0219, + "step": 25851 + }, + { + "epoch": 4.7, + "learning_rate": 1.591081179468511e-06, + "loss": 0.9326, + "step": 25852 + }, + { + "epoch": 4.7, + "learning_rate": 1.591026574444849e-06, + "loss": 0.0125, + "step": 25853 + }, + { + "epoch": 4.7, + "learning_rate": 1.5909719694211869e-06, + "loss": 0.8261, + "step": 25854 + }, + { + "epoch": 4.7, + "learning_rate": 1.5909173643975246e-06, + "loss": 1.6035, + "step": 25855 + }, + { + "epoch": 4.7, + "learning_rate": 1.5908627593738626e-06, + "loss": 0.1406, + "step": 25856 + }, + { + "epoch": 4.7, + "learning_rate": 1.5908081543502e-06, + "loss": 2.1722, + "step": 25857 + }, + { + "epoch": 4.7, + "learning_rate": 1.590753549326538e-06, + "loss": 0.1962, + "step": 25858 + }, + { + "epoch": 4.7, + "learning_rate": 1.590698944302876e-06, + "loss": 0.364, + "step": 25859 + }, + { + "epoch": 4.7, + "learning_rate": 1.5906443392792137e-06, + "loss": 0.0528, + "step": 25860 + }, + { + "epoch": 4.7, + "learning_rate": 1.5905897342555516e-06, + "loss": 0.0193, + "step": 25861 + }, + { + "epoch": 4.7, + "learning_rate": 1.5905351292318896e-06, + "loss": 1.2993, + "step": 25862 + }, + { + "epoch": 4.7, + "learning_rate": 1.590480524208227e-06, + "loss": 1.8513, + "step": 25863 + }, + { + "epoch": 4.7, + "learning_rate": 1.590425919184565e-06, + "loss": 0.8927, + "step": 25864 + }, + { + "epoch": 4.7, + "learning_rate": 1.590371314160903e-06, + "loss": 0.602, + "step": 25865 + }, + { + "epoch": 4.7, + "learning_rate": 1.5903167091372405e-06, + "loss": 0.1865, + "step": 25866 + }, + { + "epoch": 4.7, + "learning_rate": 1.5902621041135784e-06, + "loss": 0.9539, + "step": 25867 + }, + { + "epoch": 4.7, + "learning_rate": 1.5902074990899164e-06, + "loss": 0.3759, + "step": 25868 + }, + { + "epoch": 4.7, + "learning_rate": 1.5901528940662541e-06, + "loss": 1.1427, + "step": 25869 + }, + { + "epoch": 4.7, + "learning_rate": 1.590098289042592e-06, + "loss": 0.3068, + "step": 25870 + }, + { + "epoch": 4.7, + "learning_rate": 1.5900436840189296e-06, + "loss": 1.4564, + "step": 25871 + }, + { + "epoch": 4.7, + "learning_rate": 1.5899890789952675e-06, + "loss": 0.1949, + "step": 25872 + }, + { + "epoch": 4.71, + "learning_rate": 1.5899344739716055e-06, + "loss": 0.4032, + "step": 25873 + }, + { + "epoch": 4.71, + "learning_rate": 1.5898798689479432e-06, + "loss": 0.145, + "step": 25874 + }, + { + "epoch": 4.71, + "learning_rate": 1.5898252639242811e-06, + "loss": 0.063, + "step": 25875 + }, + { + "epoch": 4.71, + "learning_rate": 1.589770658900619e-06, + "loss": 0.3229, + "step": 25876 + }, + { + "epoch": 4.71, + "learning_rate": 1.5897160538769566e-06, + "loss": 0.0396, + "step": 25877 + }, + { + "epoch": 4.71, + "learning_rate": 1.5896614488532946e-06, + "loss": 0.336, + "step": 25878 + }, + { + "epoch": 4.71, + "learning_rate": 1.5896068438296325e-06, + "loss": 0.0116, + "step": 25879 + }, + { + "epoch": 4.71, + "learning_rate": 1.58955223880597e-06, + "loss": 0.1669, + "step": 25880 + }, + { + "epoch": 4.71, + "learning_rate": 1.589497633782308e-06, + "loss": 2.0485, + "step": 25881 + }, + { + "epoch": 4.71, + "learning_rate": 1.589443028758646e-06, + "loss": 0.3535, + "step": 25882 + }, + { + "epoch": 4.71, + "learning_rate": 1.5893884237349836e-06, + "loss": 0.7946, + "step": 25883 + }, + { + "epoch": 4.71, + "learning_rate": 1.5893338187113216e-06, + "loss": 0.9655, + "step": 25884 + }, + { + "epoch": 4.71, + "learning_rate": 1.589279213687659e-06, + "loss": 2.1355, + "step": 25885 + }, + { + "epoch": 4.71, + "learning_rate": 1.589224608663997e-06, + "loss": 1.9836, + "step": 25886 + }, + { + "epoch": 4.71, + "learning_rate": 1.589170003640335e-06, + "loss": 0.5265, + "step": 25887 + }, + { + "epoch": 4.71, + "learning_rate": 1.5891153986166727e-06, + "loss": 0.6772, + "step": 25888 + }, + { + "epoch": 4.71, + "learning_rate": 1.5890607935930107e-06, + "loss": 0.4717, + "step": 25889 + }, + { + "epoch": 4.71, + "learning_rate": 1.5890061885693486e-06, + "loss": 0.0536, + "step": 25890 + }, + { + "epoch": 4.71, + "learning_rate": 1.5889515835456861e-06, + "loss": 0.1334, + "step": 25891 + }, + { + "epoch": 4.71, + "learning_rate": 1.588896978522024e-06, + "loss": 0.5628, + "step": 25892 + }, + { + "epoch": 4.71, + "learning_rate": 1.588842373498362e-06, + "loss": 0.5985, + "step": 25893 + }, + { + "epoch": 4.71, + "learning_rate": 1.5887877684746997e-06, + "loss": 1.9758, + "step": 25894 + }, + { + "epoch": 4.71, + "learning_rate": 1.5887331634510377e-06, + "loss": 0.3106, + "step": 25895 + }, + { + "epoch": 4.71, + "learning_rate": 1.5886785584273752e-06, + "loss": 3.6898, + "step": 25896 + }, + { + "epoch": 4.71, + "learning_rate": 1.5886239534037131e-06, + "loss": 1.544, + "step": 25897 + }, + { + "epoch": 4.71, + "learning_rate": 1.588569348380051e-06, + "loss": 0.6319, + "step": 25898 + }, + { + "epoch": 4.71, + "learning_rate": 1.5885147433563886e-06, + "loss": 0.3232, + "step": 25899 + }, + { + "epoch": 4.71, + "learning_rate": 1.5884601383327266e-06, + "loss": 0.0142, + "step": 25900 + }, + { + "epoch": 4.71, + "learning_rate": 1.5884055333090645e-06, + "loss": 0.4552, + "step": 25901 + }, + { + "epoch": 4.71, + "learning_rate": 1.5883509282854022e-06, + "loss": 0.0153, + "step": 25902 + }, + { + "epoch": 4.71, + "learning_rate": 1.5882963232617402e-06, + "loss": 0.0171, + "step": 25903 + }, + { + "epoch": 4.71, + "learning_rate": 1.5882417182380781e-06, + "loss": 1.7419, + "step": 25904 + }, + { + "epoch": 4.71, + "learning_rate": 1.5881871132144156e-06, + "loss": 0.0111, + "step": 25905 + }, + { + "epoch": 4.71, + "learning_rate": 1.5881325081907536e-06, + "loss": 0.52, + "step": 25906 + }, + { + "epoch": 4.71, + "learning_rate": 1.5880779031670915e-06, + "loss": 0.2337, + "step": 25907 + }, + { + "epoch": 4.71, + "learning_rate": 1.5880232981434293e-06, + "loss": 0.7658, + "step": 25908 + }, + { + "epoch": 4.71, + "learning_rate": 1.5879686931197672e-06, + "loss": 0.2729, + "step": 25909 + }, + { + "epoch": 4.71, + "learning_rate": 1.5879140880961047e-06, + "loss": 0.6791, + "step": 25910 + }, + { + "epoch": 4.71, + "learning_rate": 1.5878594830724427e-06, + "loss": 0.9437, + "step": 25911 + }, + { + "epoch": 4.71, + "learning_rate": 1.5878048780487806e-06, + "loss": 0.0262, + "step": 25912 + }, + { + "epoch": 4.71, + "learning_rate": 1.5877502730251181e-06, + "loss": 0.3486, + "step": 25913 + }, + { + "epoch": 4.71, + "learning_rate": 1.587695668001456e-06, + "loss": 0.2337, + "step": 25914 + }, + { + "epoch": 4.71, + "learning_rate": 1.587641062977794e-06, + "loss": 0.3487, + "step": 25915 + }, + { + "epoch": 4.71, + "learning_rate": 1.5875864579541317e-06, + "loss": 0.4647, + "step": 25916 + }, + { + "epoch": 4.71, + "learning_rate": 1.5875318529304697e-06, + "loss": 0.0447, + "step": 25917 + }, + { + "epoch": 4.71, + "learning_rate": 1.5874772479068076e-06, + "loss": 0.8339, + "step": 25918 + }, + { + "epoch": 4.71, + "learning_rate": 1.5874226428831451e-06, + "loss": 0.0097, + "step": 25919 + }, + { + "epoch": 4.71, + "learning_rate": 1.587368037859483e-06, + "loss": 0.0173, + "step": 25920 + }, + { + "epoch": 4.71, + "learning_rate": 1.587313432835821e-06, + "loss": 0.6991, + "step": 25921 + }, + { + "epoch": 4.71, + "learning_rate": 1.5872588278121588e-06, + "loss": 1.2628, + "step": 25922 + }, + { + "epoch": 4.71, + "learning_rate": 1.5872042227884967e-06, + "loss": 1.523, + "step": 25923 + }, + { + "epoch": 4.71, + "learning_rate": 1.5871496177648342e-06, + "loss": 1.8395, + "step": 25924 + }, + { + "epoch": 4.71, + "learning_rate": 1.5870950127411722e-06, + "loss": 0.2016, + "step": 25925 + }, + { + "epoch": 4.71, + "learning_rate": 1.5870404077175101e-06, + "loss": 1.4006, + "step": 25926 + }, + { + "epoch": 4.71, + "learning_rate": 1.5869858026938476e-06, + "loss": 0.0262, + "step": 25927 + }, + { + "epoch": 4.72, + "learning_rate": 1.5869311976701856e-06, + "loss": 0.109, + "step": 25928 + }, + { + "epoch": 4.72, + "learning_rate": 1.5868765926465235e-06, + "loss": 0.0111, + "step": 25929 + }, + { + "epoch": 4.72, + "learning_rate": 1.5868219876228613e-06, + "loss": 0.6479, + "step": 25930 + }, + { + "epoch": 4.72, + "learning_rate": 1.5867673825991992e-06, + "loss": 0.722, + "step": 25931 + }, + { + "epoch": 4.72, + "learning_rate": 1.5867127775755371e-06, + "loss": 0.0604, + "step": 25932 + }, + { + "epoch": 4.72, + "learning_rate": 1.5866581725518747e-06, + "loss": 0.0848, + "step": 25933 + }, + { + "epoch": 4.72, + "learning_rate": 1.5866035675282126e-06, + "loss": 0.3995, + "step": 25934 + }, + { + "epoch": 4.72, + "learning_rate": 1.5865489625045505e-06, + "loss": 0.0288, + "step": 25935 + }, + { + "epoch": 4.72, + "learning_rate": 1.5864943574808883e-06, + "loss": 0.023, + "step": 25936 + }, + { + "epoch": 4.72, + "learning_rate": 1.5864397524572262e-06, + "loss": 0.2618, + "step": 25937 + }, + { + "epoch": 4.72, + "learning_rate": 1.5863851474335637e-06, + "loss": 0.1375, + "step": 25938 + }, + { + "epoch": 4.72, + "learning_rate": 1.5863305424099017e-06, + "loss": 0.2841, + "step": 25939 + }, + { + "epoch": 4.72, + "learning_rate": 1.5862759373862396e-06, + "loss": 1.6755, + "step": 25940 + }, + { + "epoch": 4.72, + "learning_rate": 1.5862213323625774e-06, + "loss": 0.6559, + "step": 25941 + }, + { + "epoch": 4.72, + "learning_rate": 1.5861667273389153e-06, + "loss": 0.007, + "step": 25942 + }, + { + "epoch": 4.72, + "learning_rate": 1.5861121223152532e-06, + "loss": 0.3716, + "step": 25943 + }, + { + "epoch": 4.72, + "learning_rate": 1.5860575172915908e-06, + "loss": 1.2697, + "step": 25944 + }, + { + "epoch": 4.72, + "learning_rate": 1.5860029122679287e-06, + "loss": 0.2497, + "step": 25945 + }, + { + "epoch": 4.72, + "learning_rate": 1.5859483072442667e-06, + "loss": 0.2173, + "step": 25946 + }, + { + "epoch": 4.72, + "learning_rate": 1.5858937022206042e-06, + "loss": 0.9362, + "step": 25947 + }, + { + "epoch": 4.72, + "learning_rate": 1.5858390971969421e-06, + "loss": 1.6817, + "step": 25948 + }, + { + "epoch": 4.72, + "learning_rate": 1.58578449217328e-06, + "loss": 0.0105, + "step": 25949 + }, + { + "epoch": 4.72, + "learning_rate": 1.5857298871496178e-06, + "loss": 1.8379, + "step": 25950 + }, + { + "epoch": 4.72, + "learning_rate": 1.5856752821259557e-06, + "loss": 2.8774, + "step": 25951 + }, + { + "epoch": 4.72, + "learning_rate": 1.5856206771022933e-06, + "loss": 1.2908, + "step": 25952 + }, + { + "epoch": 4.72, + "learning_rate": 1.5855660720786312e-06, + "loss": 2.1285, + "step": 25953 + }, + { + "epoch": 4.72, + "learning_rate": 1.5855114670549691e-06, + "loss": 0.7516, + "step": 25954 + }, + { + "epoch": 4.72, + "learning_rate": 1.5854568620313069e-06, + "loss": 0.2701, + "step": 25955 + }, + { + "epoch": 4.72, + "learning_rate": 1.5854022570076448e-06, + "loss": 0.7773, + "step": 25956 + }, + { + "epoch": 4.72, + "learning_rate": 1.5853476519839828e-06, + "loss": 0.7183, + "step": 25957 + }, + { + "epoch": 4.72, + "learning_rate": 1.5852930469603203e-06, + "loss": 0.0156, + "step": 25958 + }, + { + "epoch": 4.72, + "learning_rate": 1.5852384419366582e-06, + "loss": 1.249, + "step": 25959 + }, + { + "epoch": 4.72, + "learning_rate": 1.5851838369129962e-06, + "loss": 1.0252, + "step": 25960 + }, + { + "epoch": 4.72, + "learning_rate": 1.5851292318893337e-06, + "loss": 0.6476, + "step": 25961 + }, + { + "epoch": 4.72, + "learning_rate": 1.5850746268656716e-06, + "loss": 1.285, + "step": 25962 + }, + { + "epoch": 4.72, + "learning_rate": 1.5850200218420096e-06, + "loss": 1.5219, + "step": 25963 + }, + { + "epoch": 4.72, + "learning_rate": 1.5849654168183473e-06, + "loss": 0.2787, + "step": 25964 + }, + { + "epoch": 4.72, + "learning_rate": 1.5849108117946853e-06, + "loss": 0.5413, + "step": 25965 + }, + { + "epoch": 4.72, + "learning_rate": 1.5848562067710228e-06, + "loss": 1.4958, + "step": 25966 + }, + { + "epoch": 4.72, + "learning_rate": 1.5848016017473607e-06, + "loss": 1.7709, + "step": 25967 + }, + { + "epoch": 4.72, + "learning_rate": 1.5847469967236987e-06, + "loss": 0.2036, + "step": 25968 + }, + { + "epoch": 4.72, + "learning_rate": 1.5846923917000364e-06, + "loss": 2.2383, + "step": 25969 + }, + { + "epoch": 4.72, + "learning_rate": 1.5846377866763743e-06, + "loss": 0.4989, + "step": 25970 + }, + { + "epoch": 4.72, + "learning_rate": 1.5845831816527123e-06, + "loss": 0.7123, + "step": 25971 + }, + { + "epoch": 4.72, + "learning_rate": 1.5845285766290498e-06, + "loss": 0.0251, + "step": 25972 + }, + { + "epoch": 4.72, + "learning_rate": 1.5844739716053877e-06, + "loss": 2.4152, + "step": 25973 + }, + { + "epoch": 4.72, + "learning_rate": 1.5844193665817257e-06, + "loss": 0.5434, + "step": 25974 + }, + { + "epoch": 4.72, + "learning_rate": 1.5843647615580634e-06, + "loss": 2.8268, + "step": 25975 + }, + { + "epoch": 4.72, + "learning_rate": 1.5843101565344014e-06, + "loss": 0.0081, + "step": 25976 + }, + { + "epoch": 4.72, + "learning_rate": 1.5842555515107389e-06, + "loss": 0.0168, + "step": 25977 + }, + { + "epoch": 4.72, + "learning_rate": 1.5842009464870768e-06, + "loss": 0.0113, + "step": 25978 + }, + { + "epoch": 4.72, + "learning_rate": 1.5841463414634148e-06, + "loss": 1.9368, + "step": 25979 + }, + { + "epoch": 4.72, + "learning_rate": 1.5840917364397523e-06, + "loss": 1.0073, + "step": 25980 + }, + { + "epoch": 4.72, + "learning_rate": 1.5840371314160902e-06, + "loss": 0.6345, + "step": 25981 + }, + { + "epoch": 4.72, + "learning_rate": 1.5839825263924282e-06, + "loss": 1.7659, + "step": 25982 + }, + { + "epoch": 4.73, + "learning_rate": 1.583927921368766e-06, + "loss": 1.0074, + "step": 25983 + }, + { + "epoch": 4.73, + "learning_rate": 1.5838733163451038e-06, + "loss": 0.7339, + "step": 25984 + }, + { + "epoch": 4.73, + "learning_rate": 1.5838187113214418e-06, + "loss": 0.3741, + "step": 25985 + }, + { + "epoch": 4.73, + "learning_rate": 1.5837641062977793e-06, + "loss": 0.0072, + "step": 25986 + }, + { + "epoch": 4.73, + "learning_rate": 1.5837095012741173e-06, + "loss": 0.1097, + "step": 25987 + }, + { + "epoch": 4.73, + "learning_rate": 1.5836548962504552e-06, + "loss": 0.1586, + "step": 25988 + }, + { + "epoch": 4.73, + "learning_rate": 1.583600291226793e-06, + "loss": 0.9755, + "step": 25989 + }, + { + "epoch": 4.73, + "learning_rate": 1.5835456862031309e-06, + "loss": 0.3316, + "step": 25990 + }, + { + "epoch": 4.73, + "learning_rate": 1.5834910811794684e-06, + "loss": 1.1109, + "step": 25991 + }, + { + "epoch": 4.73, + "learning_rate": 1.5834364761558063e-06, + "loss": 0.1289, + "step": 25992 + }, + { + "epoch": 4.73, + "learning_rate": 1.5833818711321443e-06, + "loss": 0.5351, + "step": 25993 + }, + { + "epoch": 4.73, + "learning_rate": 1.5833272661084818e-06, + "loss": 0.0349, + "step": 25994 + }, + { + "epoch": 4.73, + "learning_rate": 1.5832726610848197e-06, + "loss": 0.3094, + "step": 25995 + }, + { + "epoch": 4.73, + "learning_rate": 1.5832180560611577e-06, + "loss": 1.3555, + "step": 25996 + }, + { + "epoch": 4.73, + "learning_rate": 1.5831634510374954e-06, + "loss": 2.9624, + "step": 25997 + }, + { + "epoch": 4.73, + "learning_rate": 1.5831088460138334e-06, + "loss": 3.364, + "step": 25998 + }, + { + "epoch": 4.73, + "learning_rate": 1.5830542409901713e-06, + "loss": 1.1954, + "step": 25999 + }, + { + "epoch": 4.73, + "learning_rate": 1.5829996359665088e-06, + "loss": 0.1714, + "step": 26000 + }, + { + "epoch": 4.73, + "learning_rate": 1.5829450309428468e-06, + "loss": 0.6945, + "step": 26001 + }, + { + "epoch": 4.73, + "learning_rate": 1.5828904259191847e-06, + "loss": 0.7398, + "step": 26002 + }, + { + "epoch": 4.73, + "learning_rate": 1.5828358208955224e-06, + "loss": 0.8626, + "step": 26003 + }, + { + "epoch": 4.73, + "learning_rate": 1.5827812158718604e-06, + "loss": 1.5642, + "step": 26004 + }, + { + "epoch": 4.73, + "learning_rate": 1.582726610848198e-06, + "loss": 0.098, + "step": 26005 + }, + { + "epoch": 4.73, + "learning_rate": 1.5826720058245358e-06, + "loss": 0.0088, + "step": 26006 + }, + { + "epoch": 4.73, + "learning_rate": 1.5826174008008738e-06, + "loss": 0.0719, + "step": 26007 + }, + { + "epoch": 4.73, + "learning_rate": 1.5825627957772113e-06, + "loss": 0.693, + "step": 26008 + }, + { + "epoch": 4.73, + "learning_rate": 1.5825081907535493e-06, + "loss": 0.8013, + "step": 26009 + }, + { + "epoch": 4.73, + "learning_rate": 1.5824535857298872e-06, + "loss": 0.9169, + "step": 26010 + }, + { + "epoch": 4.73, + "learning_rate": 1.582398980706225e-06, + "loss": 0.4925, + "step": 26011 + }, + { + "epoch": 4.73, + "learning_rate": 1.5823443756825629e-06, + "loss": 0.1232, + "step": 26012 + }, + { + "epoch": 4.73, + "learning_rate": 1.5822897706589008e-06, + "loss": 1.5143, + "step": 26013 + }, + { + "epoch": 4.73, + "learning_rate": 1.5822351656352383e-06, + "loss": 0.281, + "step": 26014 + }, + { + "epoch": 4.73, + "learning_rate": 1.5821805606115763e-06, + "loss": 0.1488, + "step": 26015 + }, + { + "epoch": 4.73, + "learning_rate": 1.5821259555879142e-06, + "loss": 0.0151, + "step": 26016 + }, + { + "epoch": 4.73, + "learning_rate": 1.582071350564252e-06, + "loss": 0.1234, + "step": 26017 + }, + { + "epoch": 4.73, + "learning_rate": 1.58201674554059e-06, + "loss": 0.5275, + "step": 26018 + }, + { + "epoch": 4.73, + "learning_rate": 1.5819621405169274e-06, + "loss": 0.8719, + "step": 26019 + }, + { + "epoch": 4.73, + "learning_rate": 1.5819075354932654e-06, + "loss": 0.0774, + "step": 26020 + }, + { + "epoch": 4.73, + "learning_rate": 1.5818529304696033e-06, + "loss": 0.6349, + "step": 26021 + }, + { + "epoch": 4.73, + "learning_rate": 1.581798325445941e-06, + "loss": 0.877, + "step": 26022 + }, + { + "epoch": 4.73, + "learning_rate": 1.581743720422279e-06, + "loss": 1.6428, + "step": 26023 + }, + { + "epoch": 4.73, + "learning_rate": 1.581689115398617e-06, + "loss": 1.4158, + "step": 26024 + }, + { + "epoch": 4.73, + "learning_rate": 1.5816345103749544e-06, + "loss": 0.5154, + "step": 26025 + }, + { + "epoch": 4.73, + "learning_rate": 1.5815799053512924e-06, + "loss": 0.0114, + "step": 26026 + }, + { + "epoch": 4.73, + "learning_rate": 1.5815253003276303e-06, + "loss": 1.4521, + "step": 26027 + }, + { + "epoch": 4.73, + "learning_rate": 1.5814706953039679e-06, + "loss": 0.0568, + "step": 26028 + }, + { + "epoch": 4.73, + "learning_rate": 1.5814160902803058e-06, + "loss": 0.1235, + "step": 26029 + }, + { + "epoch": 4.73, + "learning_rate": 1.5813614852566437e-06, + "loss": 1.2185, + "step": 26030 + }, + { + "epoch": 4.73, + "learning_rate": 1.5813068802329815e-06, + "loss": 0.2506, + "step": 26031 + }, + { + "epoch": 4.73, + "learning_rate": 1.5812522752093194e-06, + "loss": 0.0347, + "step": 26032 + }, + { + "epoch": 4.73, + "learning_rate": 1.581197670185657e-06, + "loss": 0.0392, + "step": 26033 + }, + { + "epoch": 4.73, + "learning_rate": 1.5811430651619949e-06, + "loss": 0.4484, + "step": 26034 + }, + { + "epoch": 4.73, + "learning_rate": 1.5810884601383328e-06, + "loss": 0.0092, + "step": 26035 + }, + { + "epoch": 4.73, + "learning_rate": 1.5810338551146706e-06, + "loss": 0.2933, + "step": 26036 + }, + { + "epoch": 4.73, + "learning_rate": 1.5809792500910085e-06, + "loss": 0.2766, + "step": 26037 + }, + { + "epoch": 4.74, + "learning_rate": 1.5809246450673464e-06, + "loss": 0.245, + "step": 26038 + }, + { + "epoch": 4.74, + "learning_rate": 1.580870040043684e-06, + "loss": 0.2632, + "step": 26039 + }, + { + "epoch": 4.74, + "learning_rate": 1.580815435020022e-06, + "loss": 0.2065, + "step": 26040 + }, + { + "epoch": 4.74, + "learning_rate": 1.5807608299963598e-06, + "loss": 2.1785, + "step": 26041 + }, + { + "epoch": 4.74, + "learning_rate": 1.5807062249726974e-06, + "loss": 0.0936, + "step": 26042 + }, + { + "epoch": 4.74, + "learning_rate": 1.5806516199490353e-06, + "loss": 0.2793, + "step": 26043 + }, + { + "epoch": 4.74, + "learning_rate": 1.5805970149253733e-06, + "loss": 1.051, + "step": 26044 + }, + { + "epoch": 4.74, + "learning_rate": 1.580542409901711e-06, + "loss": 0.6991, + "step": 26045 + }, + { + "epoch": 4.74, + "learning_rate": 1.580487804878049e-06, + "loss": 0.1855, + "step": 26046 + }, + { + "epoch": 4.74, + "learning_rate": 1.5804331998543864e-06, + "loss": 3.4124, + "step": 26047 + }, + { + "epoch": 4.74, + "learning_rate": 1.5803785948307244e-06, + "loss": 0.7091, + "step": 26048 + }, + { + "epoch": 4.74, + "learning_rate": 1.5803239898070623e-06, + "loss": 0.8673, + "step": 26049 + }, + { + "epoch": 4.74, + "learning_rate": 1.5802693847834e-06, + "loss": 0.0152, + "step": 26050 + }, + { + "epoch": 4.74, + "learning_rate": 1.580214779759738e-06, + "loss": 1.5129, + "step": 26051 + }, + { + "epoch": 4.74, + "learning_rate": 1.580160174736076e-06, + "loss": 0.9265, + "step": 26052 + }, + { + "epoch": 4.74, + "learning_rate": 1.5801055697124135e-06, + "loss": 0.2397, + "step": 26053 + }, + { + "epoch": 4.74, + "learning_rate": 1.5800509646887514e-06, + "loss": 2.8783, + "step": 26054 + }, + { + "epoch": 4.74, + "learning_rate": 1.5799963596650894e-06, + "loss": 0.2539, + "step": 26055 + }, + { + "epoch": 4.74, + "learning_rate": 1.5799417546414269e-06, + "loss": 0.4381, + "step": 26056 + }, + { + "epoch": 4.74, + "learning_rate": 1.5798871496177648e-06, + "loss": 0.2477, + "step": 26057 + }, + { + "epoch": 4.74, + "learning_rate": 1.5798325445941026e-06, + "loss": 1.8258, + "step": 26058 + }, + { + "epoch": 4.74, + "learning_rate": 1.5797779395704405e-06, + "loss": 0.5923, + "step": 26059 + }, + { + "epoch": 4.74, + "learning_rate": 1.5797233345467784e-06, + "loss": 1.1002, + "step": 26060 + }, + { + "epoch": 4.74, + "learning_rate": 1.579668729523116e-06, + "loss": 0.1294, + "step": 26061 + }, + { + "epoch": 4.74, + "learning_rate": 1.579614124499454e-06, + "loss": 0.0117, + "step": 26062 + }, + { + "epoch": 4.74, + "learning_rate": 1.5795595194757918e-06, + "loss": 0.4808, + "step": 26063 + }, + { + "epoch": 4.74, + "learning_rate": 1.5795049144521296e-06, + "loss": 0.2871, + "step": 26064 + }, + { + "epoch": 4.74, + "learning_rate": 1.5794503094284675e-06, + "loss": 0.8337, + "step": 26065 + }, + { + "epoch": 4.74, + "learning_rate": 1.5793957044048055e-06, + "loss": 0.2371, + "step": 26066 + }, + { + "epoch": 4.74, + "learning_rate": 1.579341099381143e-06, + "loss": 1.6353, + "step": 26067 + }, + { + "epoch": 4.74, + "learning_rate": 1.579286494357481e-06, + "loss": 0.1848, + "step": 26068 + }, + { + "epoch": 4.74, + "learning_rate": 1.5792318893338189e-06, + "loss": 0.4811, + "step": 26069 + }, + { + "epoch": 4.74, + "learning_rate": 1.5791772843101566e-06, + "loss": 0.1718, + "step": 26070 + }, + { + "epoch": 4.74, + "learning_rate": 1.5791226792864945e-06, + "loss": 3.0658, + "step": 26071 + }, + { + "epoch": 4.74, + "learning_rate": 1.579068074262832e-06, + "loss": 0.1723, + "step": 26072 + }, + { + "epoch": 4.74, + "learning_rate": 1.57901346923917e-06, + "loss": 0.3875, + "step": 26073 + }, + { + "epoch": 4.74, + "learning_rate": 1.578958864215508e-06, + "loss": 0.2325, + "step": 26074 + }, + { + "epoch": 4.74, + "learning_rate": 1.5789042591918455e-06, + "loss": 0.0822, + "step": 26075 + }, + { + "epoch": 4.74, + "learning_rate": 1.5788496541681834e-06, + "loss": 0.0336, + "step": 26076 + }, + { + "epoch": 4.74, + "learning_rate": 1.5787950491445214e-06, + "loss": 1.3237, + "step": 26077 + }, + { + "epoch": 4.74, + "learning_rate": 1.578740444120859e-06, + "loss": 1.9447, + "step": 26078 + }, + { + "epoch": 4.74, + "learning_rate": 1.578685839097197e-06, + "loss": 0.0088, + "step": 26079 + }, + { + "epoch": 4.74, + "learning_rate": 1.578631234073535e-06, + "loss": 0.9489, + "step": 26080 + }, + { + "epoch": 4.74, + "learning_rate": 1.5785766290498725e-06, + "loss": 0.1828, + "step": 26081 + }, + { + "epoch": 4.74, + "learning_rate": 1.5785220240262104e-06, + "loss": 0.8024, + "step": 26082 + }, + { + "epoch": 4.74, + "learning_rate": 1.5784674190025484e-06, + "loss": 0.437, + "step": 26083 + }, + { + "epoch": 4.74, + "learning_rate": 1.5784128139788861e-06, + "loss": 0.7794, + "step": 26084 + }, + { + "epoch": 4.74, + "learning_rate": 1.578358208955224e-06, + "loss": 0.0149, + "step": 26085 + }, + { + "epoch": 4.74, + "learning_rate": 1.5783036039315616e-06, + "loss": 1.3909, + "step": 26086 + }, + { + "epoch": 4.74, + "learning_rate": 1.5782489989078995e-06, + "loss": 0.9058, + "step": 26087 + }, + { + "epoch": 4.74, + "learning_rate": 1.5781943938842375e-06, + "loss": 0.1208, + "step": 26088 + }, + { + "epoch": 4.74, + "learning_rate": 1.578139788860575e-06, + "loss": 0.2889, + "step": 26089 + }, + { + "epoch": 4.74, + "learning_rate": 1.578085183836913e-06, + "loss": 0.4672, + "step": 26090 + }, + { + "epoch": 4.74, + "learning_rate": 1.5780305788132509e-06, + "loss": 1.5251, + "step": 26091 + }, + { + "epoch": 4.74, + "learning_rate": 1.5779759737895886e-06, + "loss": 0.0117, + "step": 26092 + }, + { + "epoch": 4.75, + "learning_rate": 1.5779213687659265e-06, + "loss": 1.0527, + "step": 26093 + }, + { + "epoch": 4.75, + "learning_rate": 1.5778667637422645e-06, + "loss": 0.0923, + "step": 26094 + }, + { + "epoch": 4.75, + "learning_rate": 1.577812158718602e-06, + "loss": 0.4845, + "step": 26095 + }, + { + "epoch": 4.75, + "learning_rate": 1.57775755369494e-06, + "loss": 0.8419, + "step": 26096 + }, + { + "epoch": 4.75, + "learning_rate": 1.577702948671278e-06, + "loss": 1.3277, + "step": 26097 + }, + { + "epoch": 4.75, + "learning_rate": 1.5776483436476156e-06, + "loss": 0.494, + "step": 26098 + }, + { + "epoch": 4.75, + "learning_rate": 1.5775937386239536e-06, + "loss": 1.1217, + "step": 26099 + }, + { + "epoch": 4.75, + "learning_rate": 1.577539133600291e-06, + "loss": 0.6882, + "step": 26100 + }, + { + "epoch": 4.75, + "learning_rate": 1.577484528576629e-06, + "loss": 0.3037, + "step": 26101 + }, + { + "epoch": 4.75, + "learning_rate": 1.577429923552967e-06, + "loss": 1.0684, + "step": 26102 + }, + { + "epoch": 4.75, + "learning_rate": 1.5773753185293047e-06, + "loss": 0.0104, + "step": 26103 + }, + { + "epoch": 4.75, + "learning_rate": 1.5773207135056427e-06, + "loss": 0.9068, + "step": 26104 + }, + { + "epoch": 4.75, + "learning_rate": 1.5772661084819806e-06, + "loss": 0.3682, + "step": 26105 + }, + { + "epoch": 4.75, + "learning_rate": 1.5772115034583181e-06, + "loss": 1.6189, + "step": 26106 + }, + { + "epoch": 4.75, + "learning_rate": 1.577156898434656e-06, + "loss": 0.2046, + "step": 26107 + }, + { + "epoch": 4.75, + "learning_rate": 1.577102293410994e-06, + "loss": 0.1564, + "step": 26108 + }, + { + "epoch": 4.75, + "learning_rate": 1.5770476883873315e-06, + "loss": 1.6034, + "step": 26109 + }, + { + "epoch": 4.75, + "learning_rate": 1.5769930833636695e-06, + "loss": 0.1645, + "step": 26110 + }, + { + "epoch": 4.75, + "learning_rate": 1.5769384783400074e-06, + "loss": 0.2146, + "step": 26111 + }, + { + "epoch": 4.75, + "learning_rate": 1.5768838733163451e-06, + "loss": 0.7873, + "step": 26112 + }, + { + "epoch": 4.75, + "learning_rate": 1.576829268292683e-06, + "loss": 0.8709, + "step": 26113 + }, + { + "epoch": 4.75, + "learning_rate": 1.5767746632690206e-06, + "loss": 1.0111, + "step": 26114 + }, + { + "epoch": 4.75, + "learning_rate": 1.5767200582453585e-06, + "loss": 0.0184, + "step": 26115 + }, + { + "epoch": 4.75, + "learning_rate": 1.5766654532216965e-06, + "loss": 0.711, + "step": 26116 + }, + { + "epoch": 4.75, + "learning_rate": 1.5766108481980342e-06, + "loss": 0.4287, + "step": 26117 + }, + { + "epoch": 4.75, + "learning_rate": 1.5765562431743722e-06, + "loss": 0.8969, + "step": 26118 + }, + { + "epoch": 4.75, + "learning_rate": 1.5765016381507101e-06, + "loss": 1.1807, + "step": 26119 + }, + { + "epoch": 4.75, + "learning_rate": 1.5764470331270476e-06, + "loss": 0.0798, + "step": 26120 + }, + { + "epoch": 4.75, + "learning_rate": 1.5763924281033856e-06, + "loss": 0.4616, + "step": 26121 + }, + { + "epoch": 4.75, + "learning_rate": 1.5763378230797235e-06, + "loss": 1.6564, + "step": 26122 + }, + { + "epoch": 4.75, + "learning_rate": 1.576283218056061e-06, + "loss": 0.233, + "step": 26123 + }, + { + "epoch": 4.75, + "learning_rate": 1.576228613032399e-06, + "loss": 0.0195, + "step": 26124 + }, + { + "epoch": 4.75, + "learning_rate": 1.5761740080087367e-06, + "loss": 0.0403, + "step": 26125 + }, + { + "epoch": 4.75, + "learning_rate": 1.5761194029850747e-06, + "loss": 1.7358, + "step": 26126 + }, + { + "epoch": 4.75, + "learning_rate": 1.5760647979614126e-06, + "loss": 0.0103, + "step": 26127 + }, + { + "epoch": 4.75, + "learning_rate": 1.5760101929377501e-06, + "loss": 1.0762, + "step": 26128 + }, + { + "epoch": 4.75, + "learning_rate": 1.575955587914088e-06, + "loss": 1.5795, + "step": 26129 + }, + { + "epoch": 4.75, + "learning_rate": 1.575900982890426e-06, + "loss": 0.1341, + "step": 26130 + }, + { + "epoch": 4.75, + "learning_rate": 1.5758463778667637e-06, + "loss": 1.0928, + "step": 26131 + }, + { + "epoch": 4.75, + "learning_rate": 1.5757917728431017e-06, + "loss": 0.4228, + "step": 26132 + }, + { + "epoch": 4.75, + "learning_rate": 1.5757371678194396e-06, + "loss": 0.5913, + "step": 26133 + }, + { + "epoch": 4.75, + "learning_rate": 1.5756825627957771e-06, + "loss": 1.3034, + "step": 26134 + }, + { + "epoch": 4.75, + "learning_rate": 1.575627957772115e-06, + "loss": 0.6377, + "step": 26135 + }, + { + "epoch": 4.75, + "learning_rate": 1.575573352748453e-06, + "loss": 0.0141, + "step": 26136 + }, + { + "epoch": 4.75, + "learning_rate": 1.5755187477247906e-06, + "loss": 1.0962, + "step": 26137 + }, + { + "epoch": 4.75, + "learning_rate": 1.5754641427011285e-06, + "loss": 0.1693, + "step": 26138 + }, + { + "epoch": 4.75, + "learning_rate": 1.5754095376774662e-06, + "loss": 0.0191, + "step": 26139 + }, + { + "epoch": 4.75, + "learning_rate": 1.5753549326538042e-06, + "loss": 0.015, + "step": 26140 + }, + { + "epoch": 4.75, + "learning_rate": 1.5753003276301421e-06, + "loss": 0.5318, + "step": 26141 + }, + { + "epoch": 4.75, + "learning_rate": 1.5752457226064796e-06, + "loss": 1.248, + "step": 26142 + }, + { + "epoch": 4.75, + "learning_rate": 1.5751911175828176e-06, + "loss": 0.1057, + "step": 26143 + }, + { + "epoch": 4.75, + "learning_rate": 1.5751365125591555e-06, + "loss": 0.9022, + "step": 26144 + }, + { + "epoch": 4.75, + "learning_rate": 1.5750819075354933e-06, + "loss": 0.3019, + "step": 26145 + }, + { + "epoch": 4.75, + "learning_rate": 1.5750273025118312e-06, + "loss": 0.3983, + "step": 26146 + }, + { + "epoch": 4.75, + "learning_rate": 1.5749726974881691e-06, + "loss": 0.3254, + "step": 26147 + }, + { + "epoch": 4.76, + "learning_rate": 1.5749180924645067e-06, + "loss": 0.0174, + "step": 26148 + }, + { + "epoch": 4.76, + "learning_rate": 1.5748634874408446e-06, + "loss": 1.2421, + "step": 26149 + }, + { + "epoch": 4.76, + "learning_rate": 1.5748088824171825e-06, + "loss": 0.1172, + "step": 26150 + }, + { + "epoch": 4.76, + "learning_rate": 1.5747542773935203e-06, + "loss": 1.9999, + "step": 26151 + }, + { + "epoch": 4.76, + "learning_rate": 1.5746996723698582e-06, + "loss": 0.4524, + "step": 26152 + }, + { + "epoch": 4.76, + "learning_rate": 1.5746450673461957e-06, + "loss": 0.0231, + "step": 26153 + }, + { + "epoch": 4.76, + "learning_rate": 1.5745904623225337e-06, + "loss": 0.7436, + "step": 26154 + }, + { + "epoch": 4.76, + "learning_rate": 1.5745358572988716e-06, + "loss": 0.5565, + "step": 26155 + }, + { + "epoch": 4.76, + "learning_rate": 1.5744812522752091e-06, + "loss": 2.8784, + "step": 26156 + }, + { + "epoch": 4.76, + "learning_rate": 1.574426647251547e-06, + "loss": 0.1739, + "step": 26157 + }, + { + "epoch": 4.76, + "learning_rate": 1.574372042227885e-06, + "loss": 0.5802, + "step": 26158 + }, + { + "epoch": 4.76, + "learning_rate": 1.5743174372042228e-06, + "loss": 0.0199, + "step": 26159 + }, + { + "epoch": 4.76, + "learning_rate": 1.5742628321805607e-06, + "loss": 0.0347, + "step": 26160 + }, + { + "epoch": 4.76, + "learning_rate": 1.5742082271568987e-06, + "loss": 0.0354, + "step": 26161 + }, + { + "epoch": 4.76, + "learning_rate": 1.5741536221332362e-06, + "loss": 0.6938, + "step": 26162 + }, + { + "epoch": 4.76, + "learning_rate": 1.5740990171095741e-06, + "loss": 0.1054, + "step": 26163 + }, + { + "epoch": 4.76, + "learning_rate": 1.574044412085912e-06, + "loss": 0.3864, + "step": 26164 + }, + { + "epoch": 4.76, + "learning_rate": 1.5739898070622498e-06, + "loss": 0.8738, + "step": 26165 + }, + { + "epoch": 4.76, + "learning_rate": 1.5739352020385877e-06, + "loss": 0.6203, + "step": 26166 + }, + { + "epoch": 4.76, + "learning_rate": 1.5738805970149253e-06, + "loss": 0.3011, + "step": 26167 + }, + { + "epoch": 4.76, + "learning_rate": 1.5738259919912632e-06, + "loss": 0.4324, + "step": 26168 + }, + { + "epoch": 4.76, + "learning_rate": 1.5737713869676011e-06, + "loss": 0.5796, + "step": 26169 + }, + { + "epoch": 4.76, + "learning_rate": 1.5737167819439387e-06, + "loss": 0.015, + "step": 26170 + }, + { + "epoch": 4.76, + "learning_rate": 1.5736621769202766e-06, + "loss": 0.9209, + "step": 26171 + }, + { + "epoch": 4.76, + "learning_rate": 1.5736075718966145e-06, + "loss": 0.9075, + "step": 26172 + }, + { + "epoch": 4.76, + "learning_rate": 1.5735529668729523e-06, + "loss": 4.0457, + "step": 26173 + }, + { + "epoch": 4.76, + "learning_rate": 1.5734983618492902e-06, + "loss": 0.6895, + "step": 26174 + }, + { + "epoch": 4.76, + "learning_rate": 1.5734437568256282e-06, + "loss": 0.0156, + "step": 26175 + }, + { + "epoch": 4.76, + "learning_rate": 1.5733891518019657e-06, + "loss": 1.449, + "step": 26176 + }, + { + "epoch": 4.76, + "learning_rate": 1.5733345467783036e-06, + "loss": 0.7634, + "step": 26177 + }, + { + "epoch": 4.76, + "learning_rate": 1.5732799417546416e-06, + "loss": 0.0473, + "step": 26178 + }, + { + "epoch": 4.76, + "learning_rate": 1.5732253367309793e-06, + "loss": 3.497, + "step": 26179 + }, + { + "epoch": 4.76, + "learning_rate": 1.5731707317073172e-06, + "loss": 0.0171, + "step": 26180 + }, + { + "epoch": 4.76, + "learning_rate": 1.5731161266836548e-06, + "loss": 0.0135, + "step": 26181 + }, + { + "epoch": 4.76, + "learning_rate": 1.5730615216599927e-06, + "loss": 0.2052, + "step": 26182 + }, + { + "epoch": 4.76, + "learning_rate": 1.5730069166363307e-06, + "loss": 0.0107, + "step": 26183 + }, + { + "epoch": 4.76, + "learning_rate": 1.5729523116126684e-06, + "loss": 1.0618, + "step": 26184 + }, + { + "epoch": 4.76, + "learning_rate": 1.5728977065890063e-06, + "loss": 0.1763, + "step": 26185 + }, + { + "epoch": 4.76, + "learning_rate": 1.572843101565344e-06, + "loss": 0.1714, + "step": 26186 + }, + { + "epoch": 4.76, + "learning_rate": 1.5727884965416818e-06, + "loss": 1.0562, + "step": 26187 + }, + { + "epoch": 4.76, + "learning_rate": 1.5727338915180197e-06, + "loss": 0.3983, + "step": 26188 + }, + { + "epoch": 4.76, + "learning_rate": 1.5726792864943577e-06, + "loss": 0.8851, + "step": 26189 + }, + { + "epoch": 4.76, + "learning_rate": 1.5726246814706952e-06, + "loss": 0.0101, + "step": 26190 + }, + { + "epoch": 4.76, + "learning_rate": 1.5725700764470331e-06, + "loss": 0.7506, + "step": 26191 + }, + { + "epoch": 4.76, + "learning_rate": 1.572515471423371e-06, + "loss": 1.6565, + "step": 26192 + }, + { + "epoch": 4.76, + "learning_rate": 1.5724608663997088e-06, + "loss": 0.4866, + "step": 26193 + }, + { + "epoch": 4.76, + "learning_rate": 1.5724062613760468e-06, + "loss": 0.0095, + "step": 26194 + }, + { + "epoch": 4.76, + "learning_rate": 1.5723516563523843e-06, + "loss": 0.2391, + "step": 26195 + }, + { + "epoch": 4.76, + "learning_rate": 1.5722970513287222e-06, + "loss": 0.2457, + "step": 26196 + }, + { + "epoch": 4.76, + "learning_rate": 1.5722424463050602e-06, + "loss": 0.3186, + "step": 26197 + }, + { + "epoch": 4.76, + "learning_rate": 1.572187841281398e-06, + "loss": 0.7458, + "step": 26198 + }, + { + "epoch": 4.76, + "learning_rate": 1.5721332362577358e-06, + "loss": 0.1989, + "step": 26199 + }, + { + "epoch": 4.76, + "learning_rate": 1.5720786312340738e-06, + "loss": 0.0169, + "step": 26200 + }, + { + "epoch": 4.76, + "learning_rate": 1.5720240262104113e-06, + "loss": 1.4709, + "step": 26201 + }, + { + "epoch": 4.76, + "learning_rate": 1.5719694211867492e-06, + "loss": 2.1625, + "step": 26202 + }, + { + "epoch": 4.77, + "learning_rate": 1.5719148161630872e-06, + "loss": 1.2367, + "step": 26203 + }, + { + "epoch": 4.77, + "learning_rate": 1.5718602111394247e-06, + "loss": 0.0548, + "step": 26204 + }, + { + "epoch": 4.77, + "learning_rate": 1.5718056061157627e-06, + "loss": 0.8399, + "step": 26205 + }, + { + "epoch": 4.77, + "learning_rate": 1.5717510010921004e-06, + "loss": 0.5911, + "step": 26206 + }, + { + "epoch": 4.77, + "learning_rate": 1.5716963960684383e-06, + "loss": 0.2063, + "step": 26207 + }, + { + "epoch": 4.77, + "learning_rate": 1.5716417910447763e-06, + "loss": 1.2381, + "step": 26208 + }, + { + "epoch": 4.77, + "learning_rate": 1.5715871860211138e-06, + "loss": 0.0536, + "step": 26209 + }, + { + "epoch": 4.77, + "learning_rate": 1.5715325809974517e-06, + "loss": 0.3026, + "step": 26210 + }, + { + "epoch": 4.77, + "learning_rate": 1.5714779759737897e-06, + "loss": 0.3605, + "step": 26211 + }, + { + "epoch": 4.77, + "learning_rate": 1.5714233709501274e-06, + "loss": 0.5229, + "step": 26212 + }, + { + "epoch": 4.77, + "learning_rate": 1.5713687659264654e-06, + "loss": 0.8076, + "step": 26213 + }, + { + "epoch": 4.77, + "learning_rate": 1.5713141609028033e-06, + "loss": 0.5685, + "step": 26214 + }, + { + "epoch": 4.77, + "learning_rate": 1.5712595558791408e-06, + "loss": 0.0207, + "step": 26215 + }, + { + "epoch": 4.77, + "learning_rate": 1.5712049508554788e-06, + "loss": 0.7499, + "step": 26216 + }, + { + "epoch": 4.77, + "learning_rate": 1.5711503458318167e-06, + "loss": 0.2439, + "step": 26217 + }, + { + "epoch": 4.77, + "learning_rate": 1.5710957408081542e-06, + "loss": 0.2654, + "step": 26218 + }, + { + "epoch": 4.77, + "learning_rate": 1.5710411357844922e-06, + "loss": 1.3531, + "step": 26219 + }, + { + "epoch": 4.77, + "learning_rate": 1.57098653076083e-06, + "loss": 0.5968, + "step": 26220 + }, + { + "epoch": 4.77, + "learning_rate": 1.5709319257371678e-06, + "loss": 0.7556, + "step": 26221 + }, + { + "epoch": 4.77, + "learning_rate": 1.5708773207135058e-06, + "loss": 0.3582, + "step": 26222 + }, + { + "epoch": 4.77, + "learning_rate": 1.5708227156898433e-06, + "loss": 0.0276, + "step": 26223 + }, + { + "epoch": 4.77, + "learning_rate": 1.5707681106661813e-06, + "loss": 0.3473, + "step": 26224 + }, + { + "epoch": 4.77, + "learning_rate": 1.5707135056425192e-06, + "loss": 0.0795, + "step": 26225 + }, + { + "epoch": 4.77, + "learning_rate": 1.570658900618857e-06, + "loss": 0.9985, + "step": 26226 + }, + { + "epoch": 4.77, + "learning_rate": 1.5706042955951949e-06, + "loss": 0.2863, + "step": 26227 + }, + { + "epoch": 4.77, + "learning_rate": 1.5705496905715328e-06, + "loss": 1.1049, + "step": 26228 + }, + { + "epoch": 4.77, + "learning_rate": 1.5704950855478703e-06, + "loss": 0.2796, + "step": 26229 + }, + { + "epoch": 4.77, + "learning_rate": 1.5704404805242083e-06, + "loss": 0.1573, + "step": 26230 + }, + { + "epoch": 4.77, + "learning_rate": 1.5703858755005462e-06, + "loss": 0.1193, + "step": 26231 + }, + { + "epoch": 4.77, + "learning_rate": 1.570331270476884e-06, + "loss": 0.4109, + "step": 26232 + }, + { + "epoch": 4.77, + "learning_rate": 1.570276665453222e-06, + "loss": 2.3533, + "step": 26233 + }, + { + "epoch": 4.77, + "learning_rate": 1.5702220604295594e-06, + "loss": 0.6421, + "step": 26234 + }, + { + "epoch": 4.77, + "learning_rate": 1.5701674554058974e-06, + "loss": 1.03, + "step": 26235 + }, + { + "epoch": 4.77, + "learning_rate": 1.5701128503822353e-06, + "loss": 0.1479, + "step": 26236 + }, + { + "epoch": 4.77, + "learning_rate": 1.5700582453585728e-06, + "loss": 0.1547, + "step": 26237 + }, + { + "epoch": 4.77, + "learning_rate": 1.5700036403349108e-06, + "loss": 0.0511, + "step": 26238 + }, + { + "epoch": 4.77, + "learning_rate": 1.5699490353112487e-06, + "loss": 0.5308, + "step": 26239 + }, + { + "epoch": 4.77, + "learning_rate": 1.5698944302875864e-06, + "loss": 0.0135, + "step": 26240 + }, + { + "epoch": 4.77, + "learning_rate": 1.5698398252639244e-06, + "loss": 0.5848, + "step": 26241 + }, + { + "epoch": 4.77, + "learning_rate": 1.5697852202402623e-06, + "loss": 0.7025, + "step": 26242 + }, + { + "epoch": 4.77, + "learning_rate": 1.5697306152165998e-06, + "loss": 0.8566, + "step": 26243 + }, + { + "epoch": 4.77, + "learning_rate": 1.5696760101929378e-06, + "loss": 0.5219, + "step": 26244 + }, + { + "epoch": 4.77, + "learning_rate": 1.5696214051692757e-06, + "loss": 0.2857, + "step": 26245 + }, + { + "epoch": 4.77, + "learning_rate": 1.5695668001456135e-06, + "loss": 0.3736, + "step": 26246 + }, + { + "epoch": 4.77, + "learning_rate": 1.5695121951219514e-06, + "loss": 0.895, + "step": 26247 + }, + { + "epoch": 4.77, + "learning_rate": 1.569457590098289e-06, + "loss": 3.6227, + "step": 26248 + }, + { + "epoch": 4.77, + "learning_rate": 1.5694029850746269e-06, + "loss": 0.1744, + "step": 26249 + }, + { + "epoch": 4.77, + "learning_rate": 1.5693483800509648e-06, + "loss": 0.0691, + "step": 26250 + }, + { + "epoch": 4.77, + "learning_rate": 1.5692937750273023e-06, + "loss": 4.2613, + "step": 26251 + }, + { + "epoch": 4.77, + "learning_rate": 1.5692391700036403e-06, + "loss": 2.021, + "step": 26252 + }, + { + "epoch": 4.77, + "learning_rate": 1.5691845649799782e-06, + "loss": 0.1187, + "step": 26253 + }, + { + "epoch": 4.77, + "learning_rate": 1.569129959956316e-06, + "loss": 1.321, + "step": 26254 + }, + { + "epoch": 4.77, + "learning_rate": 1.569075354932654e-06, + "loss": 0.3531, + "step": 26255 + }, + { + "epoch": 4.77, + "learning_rate": 1.5690207499089918e-06, + "loss": 0.6934, + "step": 26256 + }, + { + "epoch": 4.77, + "learning_rate": 1.5689661448853294e-06, + "loss": 2.8431, + "step": 26257 + }, + { + "epoch": 4.78, + "learning_rate": 1.5689115398616673e-06, + "loss": 0.9001, + "step": 26258 + }, + { + "epoch": 4.78, + "learning_rate": 1.5688569348380052e-06, + "loss": 0.3874, + "step": 26259 + }, + { + "epoch": 4.78, + "learning_rate": 1.568802329814343e-06, + "loss": 0.343, + "step": 26260 + }, + { + "epoch": 4.78, + "learning_rate": 1.568747724790681e-06, + "loss": 0.0404, + "step": 26261 + }, + { + "epoch": 4.78, + "learning_rate": 1.5686931197670184e-06, + "loss": 0.2032, + "step": 26262 + }, + { + "epoch": 4.78, + "learning_rate": 1.5686385147433564e-06, + "loss": 2.0619, + "step": 26263 + }, + { + "epoch": 4.78, + "learning_rate": 1.5685839097196943e-06, + "loss": 1.5964, + "step": 26264 + }, + { + "epoch": 4.78, + "learning_rate": 1.5685293046960318e-06, + "loss": 0.7709, + "step": 26265 + }, + { + "epoch": 4.78, + "learning_rate": 1.5684746996723698e-06, + "loss": 2.5047, + "step": 26266 + }, + { + "epoch": 4.78, + "learning_rate": 1.5684200946487077e-06, + "loss": 1.5966, + "step": 26267 + }, + { + "epoch": 4.78, + "learning_rate": 1.5683654896250455e-06, + "loss": 0.6249, + "step": 26268 + }, + { + "epoch": 4.78, + "learning_rate": 1.5683108846013834e-06, + "loss": 0.093, + "step": 26269 + }, + { + "epoch": 4.78, + "learning_rate": 1.5682562795777214e-06, + "loss": 2.2445, + "step": 26270 + }, + { + "epoch": 4.78, + "learning_rate": 1.5682016745540589e-06, + "loss": 0.7056, + "step": 26271 + }, + { + "epoch": 4.78, + "learning_rate": 1.5681470695303968e-06, + "loss": 0.5552, + "step": 26272 + }, + { + "epoch": 4.78, + "learning_rate": 1.5680924645067348e-06, + "loss": 0.6768, + "step": 26273 + }, + { + "epoch": 4.78, + "learning_rate": 1.5680378594830725e-06, + "loss": 0.5601, + "step": 26274 + }, + { + "epoch": 4.78, + "learning_rate": 1.5679832544594104e-06, + "loss": 0.0122, + "step": 26275 + }, + { + "epoch": 4.78, + "learning_rate": 1.567928649435748e-06, + "loss": 0.0643, + "step": 26276 + }, + { + "epoch": 4.78, + "learning_rate": 1.567874044412086e-06, + "loss": 0.1068, + "step": 26277 + }, + { + "epoch": 4.78, + "learning_rate": 1.5678194393884238e-06, + "loss": 0.0533, + "step": 26278 + }, + { + "epoch": 4.78, + "learning_rate": 1.5677648343647616e-06, + "loss": 0.7703, + "step": 26279 + }, + { + "epoch": 4.78, + "learning_rate": 1.5677102293410995e-06, + "loss": 0.0235, + "step": 26280 + }, + { + "epoch": 4.78, + "learning_rate": 1.5676556243174375e-06, + "loss": 0.3456, + "step": 26281 + }, + { + "epoch": 4.78, + "learning_rate": 1.567601019293775e-06, + "loss": 0.2406, + "step": 26282 + }, + { + "epoch": 4.78, + "learning_rate": 1.567546414270113e-06, + "loss": 0.574, + "step": 26283 + }, + { + "epoch": 4.78, + "learning_rate": 1.5674918092464509e-06, + "loss": 0.0068, + "step": 26284 + }, + { + "epoch": 4.78, + "learning_rate": 1.5674372042227884e-06, + "loss": 0.776, + "step": 26285 + }, + { + "epoch": 4.78, + "learning_rate": 1.5673825991991263e-06, + "loss": 0.8829, + "step": 26286 + }, + { + "epoch": 4.78, + "learning_rate": 1.567327994175464e-06, + "loss": 0.7538, + "step": 26287 + }, + { + "epoch": 4.78, + "learning_rate": 1.567273389151802e-06, + "loss": 0.5214, + "step": 26288 + }, + { + "epoch": 4.78, + "learning_rate": 1.56721878412814e-06, + "loss": 0.0185, + "step": 26289 + }, + { + "epoch": 4.78, + "learning_rate": 1.5671641791044775e-06, + "loss": 2.2045, + "step": 26290 + }, + { + "epoch": 4.78, + "learning_rate": 1.5671095740808154e-06, + "loss": 1.0318, + "step": 26291 + }, + { + "epoch": 4.78, + "learning_rate": 1.5670549690571534e-06, + "loss": 0.4154, + "step": 26292 + }, + { + "epoch": 4.78, + "learning_rate": 1.567000364033491e-06, + "loss": 4.7092, + "step": 26293 + }, + { + "epoch": 4.78, + "learning_rate": 1.566945759009829e-06, + "loss": 0.0639, + "step": 26294 + }, + { + "epoch": 4.78, + "learning_rate": 1.566891153986167e-06, + "loss": 0.618, + "step": 26295 + }, + { + "epoch": 4.78, + "learning_rate": 1.5668365489625045e-06, + "loss": 0.6688, + "step": 26296 + }, + { + "epoch": 4.78, + "learning_rate": 1.5667819439388424e-06, + "loss": 0.0381, + "step": 26297 + }, + { + "epoch": 4.78, + "learning_rate": 1.5667273389151804e-06, + "loss": 0.3532, + "step": 26298 + }, + { + "epoch": 4.78, + "learning_rate": 1.566672733891518e-06, + "loss": 0.1709, + "step": 26299 + }, + { + "epoch": 4.78, + "learning_rate": 1.5666181288678558e-06, + "loss": 0.0114, + "step": 26300 + }, + { + "epoch": 4.78, + "learning_rate": 1.5665635238441936e-06, + "loss": 0.7185, + "step": 26301 + }, + { + "epoch": 4.78, + "learning_rate": 1.5665089188205315e-06, + "loss": 1.3441, + "step": 26302 + }, + { + "epoch": 4.78, + "learning_rate": 1.5664543137968695e-06, + "loss": 0.556, + "step": 26303 + }, + { + "epoch": 4.78, + "learning_rate": 1.566399708773207e-06, + "loss": 0.5341, + "step": 26304 + }, + { + "epoch": 4.78, + "learning_rate": 1.566345103749545e-06, + "loss": 2.9513, + "step": 26305 + }, + { + "epoch": 4.78, + "learning_rate": 1.5662904987258829e-06, + "loss": 0.0086, + "step": 26306 + }, + { + "epoch": 4.78, + "learning_rate": 1.5662358937022206e-06, + "loss": 0.1667, + "step": 26307 + }, + { + "epoch": 4.78, + "learning_rate": 1.5661812886785585e-06, + "loss": 0.244, + "step": 26308 + }, + { + "epoch": 4.78, + "learning_rate": 1.5661266836548965e-06, + "loss": 1.5014, + "step": 26309 + }, + { + "epoch": 4.78, + "learning_rate": 1.566072078631234e-06, + "loss": 1.0406, + "step": 26310 + }, + { + "epoch": 4.78, + "learning_rate": 1.566017473607572e-06, + "loss": 0.1943, + "step": 26311 + }, + { + "epoch": 4.78, + "learning_rate": 1.5659628685839099e-06, + "loss": 0.7811, + "step": 26312 + }, + { + "epoch": 4.79, + "learning_rate": 1.5659082635602476e-06, + "loss": 0.3187, + "step": 26313 + }, + { + "epoch": 4.79, + "learning_rate": 1.5658536585365856e-06, + "loss": 1.0162, + "step": 26314 + }, + { + "epoch": 4.79, + "learning_rate": 1.565799053512923e-06, + "loss": 1.1664, + "step": 26315 + }, + { + "epoch": 4.79, + "learning_rate": 1.565744448489261e-06, + "loss": 0.0268, + "step": 26316 + }, + { + "epoch": 4.79, + "learning_rate": 1.565689843465599e-06, + "loss": 1.2386, + "step": 26317 + }, + { + "epoch": 4.79, + "learning_rate": 1.5656352384419365e-06, + "loss": 1.5401, + "step": 26318 + }, + { + "epoch": 4.79, + "learning_rate": 1.5655806334182744e-06, + "loss": 0.3518, + "step": 26319 + }, + { + "epoch": 4.79, + "learning_rate": 1.5655260283946124e-06, + "loss": 2.3612, + "step": 26320 + }, + { + "epoch": 4.79, + "learning_rate": 1.5654714233709501e-06, + "loss": 0.7911, + "step": 26321 + }, + { + "epoch": 4.79, + "learning_rate": 1.565416818347288e-06, + "loss": 0.6241, + "step": 26322 + }, + { + "epoch": 4.79, + "learning_rate": 1.565362213323626e-06, + "loss": 0.1166, + "step": 26323 + }, + { + "epoch": 4.79, + "learning_rate": 1.5653076082999635e-06, + "loss": 0.0185, + "step": 26324 + }, + { + "epoch": 4.79, + "learning_rate": 1.5652530032763015e-06, + "loss": 1.9079, + "step": 26325 + }, + { + "epoch": 4.79, + "learning_rate": 1.5651983982526394e-06, + "loss": 0.4633, + "step": 26326 + }, + { + "epoch": 4.79, + "learning_rate": 1.5651437932289771e-06, + "loss": 2.4201, + "step": 26327 + }, + { + "epoch": 4.79, + "learning_rate": 1.565089188205315e-06, + "loss": 0.3851, + "step": 26328 + }, + { + "epoch": 4.79, + "learning_rate": 1.5650345831816526e-06, + "loss": 0.0947, + "step": 26329 + }, + { + "epoch": 4.79, + "learning_rate": 1.5649799781579905e-06, + "loss": 0.564, + "step": 26330 + }, + { + "epoch": 4.79, + "learning_rate": 1.5649253731343285e-06, + "loss": 0.5741, + "step": 26331 + }, + { + "epoch": 4.79, + "learning_rate": 1.564870768110666e-06, + "loss": 2.8395, + "step": 26332 + }, + { + "epoch": 4.79, + "learning_rate": 1.564816163087004e-06, + "loss": 0.8908, + "step": 26333 + }, + { + "epoch": 4.79, + "learning_rate": 1.564761558063342e-06, + "loss": 0.1764, + "step": 26334 + }, + { + "epoch": 4.79, + "learning_rate": 1.5647069530396796e-06, + "loss": 0.8638, + "step": 26335 + }, + { + "epoch": 4.79, + "learning_rate": 1.5646523480160176e-06, + "loss": 0.0333, + "step": 26336 + }, + { + "epoch": 4.79, + "learning_rate": 1.5645977429923555e-06, + "loss": 0.1708, + "step": 26337 + }, + { + "epoch": 4.79, + "learning_rate": 1.564543137968693e-06, + "loss": 0.4357, + "step": 26338 + }, + { + "epoch": 4.79, + "learning_rate": 1.564488532945031e-06, + "loss": 0.015, + "step": 26339 + }, + { + "epoch": 4.79, + "learning_rate": 1.564433927921369e-06, + "loss": 0.0111, + "step": 26340 + }, + { + "epoch": 4.79, + "learning_rate": 1.5643793228977067e-06, + "loss": 0.5152, + "step": 26341 + }, + { + "epoch": 4.79, + "learning_rate": 1.5643247178740446e-06, + "loss": 2.1983, + "step": 26342 + }, + { + "epoch": 4.79, + "learning_rate": 1.5642701128503821e-06, + "loss": 0.3276, + "step": 26343 + }, + { + "epoch": 4.79, + "learning_rate": 1.56421550782672e-06, + "loss": 0.9806, + "step": 26344 + }, + { + "epoch": 4.79, + "learning_rate": 1.564160902803058e-06, + "loss": 0.4323, + "step": 26345 + }, + { + "epoch": 4.79, + "learning_rate": 1.5641062977793955e-06, + "loss": 0.3425, + "step": 26346 + }, + { + "epoch": 4.79, + "learning_rate": 1.5640516927557335e-06, + "loss": 0.4355, + "step": 26347 + }, + { + "epoch": 4.79, + "learning_rate": 1.5639970877320714e-06, + "loss": 0.071, + "step": 26348 + }, + { + "epoch": 4.79, + "learning_rate": 1.5639424827084091e-06, + "loss": 0.5846, + "step": 26349 + }, + { + "epoch": 4.79, + "learning_rate": 1.563887877684747e-06, + "loss": 0.0395, + "step": 26350 + }, + { + "epoch": 4.79, + "learning_rate": 1.563833272661085e-06, + "loss": 0.2945, + "step": 26351 + }, + { + "epoch": 4.79, + "learning_rate": 1.5637786676374225e-06, + "loss": 3.9608, + "step": 26352 + }, + { + "epoch": 4.79, + "learning_rate": 1.5637240626137605e-06, + "loss": 0.5331, + "step": 26353 + }, + { + "epoch": 4.79, + "learning_rate": 1.5636694575900984e-06, + "loss": 0.0139, + "step": 26354 + }, + { + "epoch": 4.79, + "learning_rate": 1.5636148525664362e-06, + "loss": 0.9875, + "step": 26355 + }, + { + "epoch": 4.79, + "learning_rate": 1.5635602475427741e-06, + "loss": 1.857, + "step": 26356 + }, + { + "epoch": 4.79, + "learning_rate": 1.5635056425191116e-06, + "loss": 0.6227, + "step": 26357 + }, + { + "epoch": 4.79, + "learning_rate": 1.5634510374954496e-06, + "loss": 1.9309, + "step": 26358 + }, + { + "epoch": 4.79, + "learning_rate": 1.5633964324717875e-06, + "loss": 0.0357, + "step": 26359 + }, + { + "epoch": 4.79, + "learning_rate": 1.5633418274481252e-06, + "loss": 1.8225, + "step": 26360 + }, + { + "epoch": 4.79, + "learning_rate": 1.5632872224244632e-06, + "loss": 0.0664, + "step": 26361 + }, + { + "epoch": 4.79, + "learning_rate": 1.5632326174008011e-06, + "loss": 1.3641, + "step": 26362 + }, + { + "epoch": 4.79, + "learning_rate": 1.5631780123771387e-06, + "loss": 0.1889, + "step": 26363 + }, + { + "epoch": 4.79, + "learning_rate": 1.5631234073534766e-06, + "loss": 1.1428, + "step": 26364 + }, + { + "epoch": 4.79, + "learning_rate": 1.5630688023298145e-06, + "loss": 0.4855, + "step": 26365 + }, + { + "epoch": 4.79, + "learning_rate": 1.563014197306152e-06, + "loss": 0.072, + "step": 26366 + }, + { + "epoch": 4.79, + "learning_rate": 1.56295959228249e-06, + "loss": 0.5581, + "step": 26367 + }, + { + "epoch": 4.8, + "learning_rate": 1.5629049872588277e-06, + "loss": 0.4276, + "step": 26368 + }, + { + "epoch": 4.8, + "learning_rate": 1.5628503822351657e-06, + "loss": 1.4377, + "step": 26369 + }, + { + "epoch": 4.8, + "learning_rate": 1.5627957772115036e-06, + "loss": 1.662, + "step": 26370 + }, + { + "epoch": 4.8, + "learning_rate": 1.5627411721878411e-06, + "loss": 0.6886, + "step": 26371 + }, + { + "epoch": 4.8, + "learning_rate": 1.562686567164179e-06, + "loss": 0.7967, + "step": 26372 + }, + { + "epoch": 4.8, + "learning_rate": 1.562631962140517e-06, + "loss": 0.0054, + "step": 26373 + }, + { + "epoch": 4.8, + "learning_rate": 1.5625773571168548e-06, + "loss": 0.8447, + "step": 26374 + }, + { + "epoch": 4.8, + "learning_rate": 1.5625227520931927e-06, + "loss": 0.0098, + "step": 26375 + }, + { + "epoch": 4.8, + "learning_rate": 1.5624681470695306e-06, + "loss": 0.5666, + "step": 26376 + }, + { + "epoch": 4.8, + "learning_rate": 1.5624135420458682e-06, + "loss": 0.9422, + "step": 26377 + }, + { + "epoch": 4.8, + "learning_rate": 1.5623589370222061e-06, + "loss": 1.821, + "step": 26378 + }, + { + "epoch": 4.8, + "learning_rate": 1.562304331998544e-06, + "loss": 1.7084, + "step": 26379 + }, + { + "epoch": 4.8, + "learning_rate": 1.5622497269748816e-06, + "loss": 1.0934, + "step": 26380 + }, + { + "epoch": 4.8, + "learning_rate": 1.5621951219512195e-06, + "loss": 2.2092, + "step": 26381 + }, + { + "epoch": 4.8, + "learning_rate": 1.5621405169275572e-06, + "loss": 0.2017, + "step": 26382 + }, + { + "epoch": 4.8, + "learning_rate": 1.5620859119038952e-06, + "loss": 0.4195, + "step": 26383 + }, + { + "epoch": 4.8, + "learning_rate": 1.5620313068802331e-06, + "loss": 0.0456, + "step": 26384 + }, + { + "epoch": 4.8, + "learning_rate": 1.5619767018565707e-06, + "loss": 0.0229, + "step": 26385 + }, + { + "epoch": 4.8, + "learning_rate": 1.5619220968329086e-06, + "loss": 0.0559, + "step": 26386 + }, + { + "epoch": 4.8, + "learning_rate": 1.5618674918092465e-06, + "loss": 0.4107, + "step": 26387 + }, + { + "epoch": 4.8, + "learning_rate": 1.5618128867855843e-06, + "loss": 0.5416, + "step": 26388 + }, + { + "epoch": 4.8, + "learning_rate": 1.5617582817619222e-06, + "loss": 0.0211, + "step": 26389 + }, + { + "epoch": 4.8, + "learning_rate": 1.5617036767382602e-06, + "loss": 0.3444, + "step": 26390 + }, + { + "epoch": 4.8, + "learning_rate": 1.5616490717145977e-06, + "loss": 0.027, + "step": 26391 + }, + { + "epoch": 4.8, + "learning_rate": 1.5615944666909356e-06, + "loss": 0.217, + "step": 26392 + }, + { + "epoch": 4.8, + "learning_rate": 1.5615398616672736e-06, + "loss": 0.829, + "step": 26393 + }, + { + "epoch": 4.8, + "learning_rate": 1.561485256643611e-06, + "loss": 1.0592, + "step": 26394 + }, + { + "epoch": 4.8, + "learning_rate": 1.561430651619949e-06, + "loss": 0.3849, + "step": 26395 + }, + { + "epoch": 4.8, + "learning_rate": 1.5613760465962868e-06, + "loss": 0.4056, + "step": 26396 + }, + { + "epoch": 4.8, + "learning_rate": 1.5613214415726247e-06, + "loss": 0.0083, + "step": 26397 + }, + { + "epoch": 4.8, + "learning_rate": 1.5612668365489626e-06, + "loss": 0.2006, + "step": 26398 + }, + { + "epoch": 4.8, + "learning_rate": 1.5612122315253002e-06, + "loss": 0.0353, + "step": 26399 + }, + { + "epoch": 4.8, + "learning_rate": 1.5611576265016381e-06, + "loss": 0.2441, + "step": 26400 + }, + { + "epoch": 4.8, + "learning_rate": 1.561103021477976e-06, + "loss": 0.0544, + "step": 26401 + }, + { + "epoch": 4.8, + "learning_rate": 1.5610484164543138e-06, + "loss": 1.3149, + "step": 26402 + }, + { + "epoch": 4.8, + "learning_rate": 1.5609938114306517e-06, + "loss": 0.0715, + "step": 26403 + }, + { + "epoch": 4.8, + "learning_rate": 1.5609392064069897e-06, + "loss": 0.5452, + "step": 26404 + }, + { + "epoch": 4.8, + "learning_rate": 1.5608846013833272e-06, + "loss": 0.0054, + "step": 26405 + }, + { + "epoch": 4.8, + "learning_rate": 1.5608299963596651e-06, + "loss": 1.0451, + "step": 26406 + }, + { + "epoch": 4.8, + "learning_rate": 1.560775391336003e-06, + "loss": 0.0755, + "step": 26407 + }, + { + "epoch": 4.8, + "learning_rate": 1.5607207863123408e-06, + "loss": 1.1962, + "step": 26408 + }, + { + "epoch": 4.8, + "learning_rate": 1.5606661812886788e-06, + "loss": 0.0629, + "step": 26409 + }, + { + "epoch": 4.8, + "learning_rate": 1.5606115762650163e-06, + "loss": 0.2908, + "step": 26410 + }, + { + "epoch": 4.8, + "learning_rate": 1.5605569712413542e-06, + "loss": 1.2704, + "step": 26411 + }, + { + "epoch": 4.8, + "learning_rate": 1.5605023662176922e-06, + "loss": 0.0097, + "step": 26412 + }, + { + "epoch": 4.8, + "learning_rate": 1.5604477611940297e-06, + "loss": 0.5397, + "step": 26413 + }, + { + "epoch": 4.8, + "learning_rate": 1.5603931561703676e-06, + "loss": 2.3981, + "step": 26414 + }, + { + "epoch": 4.8, + "learning_rate": 1.5603385511467056e-06, + "loss": 0.4922, + "step": 26415 + }, + { + "epoch": 4.8, + "learning_rate": 1.5602839461230433e-06, + "loss": 0.0771, + "step": 26416 + }, + { + "epoch": 4.8, + "learning_rate": 1.5602293410993812e-06, + "loss": 1.3785, + "step": 26417 + }, + { + "epoch": 4.8, + "learning_rate": 1.5601747360757192e-06, + "loss": 1.7282, + "step": 26418 + }, + { + "epoch": 4.8, + "learning_rate": 1.5601201310520567e-06, + "loss": 0.7593, + "step": 26419 + }, + { + "epoch": 4.8, + "learning_rate": 1.5600655260283947e-06, + "loss": 1.261, + "step": 26420 + }, + { + "epoch": 4.8, + "learning_rate": 1.5600109210047326e-06, + "loss": 0.8518, + "step": 26421 + }, + { + "epoch": 4.8, + "learning_rate": 1.5599563159810703e-06, + "loss": 0.1434, + "step": 26422 + }, + { + "epoch": 4.81, + "learning_rate": 1.5599017109574083e-06, + "loss": 0.5383, + "step": 26423 + }, + { + "epoch": 4.81, + "learning_rate": 1.5598471059337458e-06, + "loss": 1.2334, + "step": 26424 + }, + { + "epoch": 4.81, + "learning_rate": 1.5597925009100837e-06, + "loss": 0.5437, + "step": 26425 + }, + { + "epoch": 4.81, + "learning_rate": 1.5597378958864217e-06, + "loss": 0.5817, + "step": 26426 + }, + { + "epoch": 4.81, + "learning_rate": 1.5596832908627592e-06, + "loss": 0.2666, + "step": 26427 + }, + { + "epoch": 4.81, + "learning_rate": 1.5596286858390971e-06, + "loss": 0.0109, + "step": 26428 + }, + { + "epoch": 4.81, + "learning_rate": 1.559574080815435e-06, + "loss": 0.3863, + "step": 26429 + }, + { + "epoch": 4.81, + "learning_rate": 1.5595194757917728e-06, + "loss": 0.2771, + "step": 26430 + }, + { + "epoch": 4.81, + "learning_rate": 1.5594648707681108e-06, + "loss": 1.3503, + "step": 26431 + }, + { + "epoch": 4.81, + "learning_rate": 1.5594102657444487e-06, + "loss": 0.0709, + "step": 26432 + }, + { + "epoch": 4.81, + "learning_rate": 1.5593556607207862e-06, + "loss": 0.7047, + "step": 26433 + }, + { + "epoch": 4.81, + "learning_rate": 1.5593010556971242e-06, + "loss": 0.005, + "step": 26434 + }, + { + "epoch": 4.81, + "learning_rate": 1.559246450673462e-06, + "loss": 0.6157, + "step": 26435 + }, + { + "epoch": 4.81, + "learning_rate": 1.5591918456497998e-06, + "loss": 2.2389, + "step": 26436 + }, + { + "epoch": 4.81, + "learning_rate": 1.5591372406261378e-06, + "loss": 0.0151, + "step": 26437 + }, + { + "epoch": 4.81, + "learning_rate": 1.5590826356024753e-06, + "loss": 0.2338, + "step": 26438 + }, + { + "epoch": 4.81, + "learning_rate": 1.5590280305788132e-06, + "loss": 1.6749, + "step": 26439 + }, + { + "epoch": 4.81, + "learning_rate": 1.5589734255551512e-06, + "loss": 0.0967, + "step": 26440 + }, + { + "epoch": 4.81, + "learning_rate": 1.558918820531489e-06, + "loss": 0.4716, + "step": 26441 + }, + { + "epoch": 4.81, + "learning_rate": 1.5588642155078269e-06, + "loss": 0.1722, + "step": 26442 + }, + { + "epoch": 4.81, + "learning_rate": 1.5588096104841648e-06, + "loss": 0.4665, + "step": 26443 + }, + { + "epoch": 4.81, + "learning_rate": 1.5587550054605023e-06, + "loss": 1.1395, + "step": 26444 + }, + { + "epoch": 4.81, + "learning_rate": 1.5587004004368403e-06, + "loss": 0.0166, + "step": 26445 + }, + { + "epoch": 4.81, + "learning_rate": 1.5586457954131782e-06, + "loss": 0.6723, + "step": 26446 + }, + { + "epoch": 4.81, + "learning_rate": 1.5585911903895157e-06, + "loss": 1.4992, + "step": 26447 + }, + { + "epoch": 4.81, + "learning_rate": 1.5585365853658537e-06, + "loss": 0.9882, + "step": 26448 + }, + { + "epoch": 4.81, + "learning_rate": 1.5584819803421914e-06, + "loss": 0.2334, + "step": 26449 + }, + { + "epoch": 4.81, + "learning_rate": 1.5584273753185294e-06, + "loss": 0.0616, + "step": 26450 + }, + { + "epoch": 4.81, + "learning_rate": 1.5583727702948673e-06, + "loss": 1.2146, + "step": 26451 + }, + { + "epoch": 4.81, + "learning_rate": 1.5583181652712048e-06, + "loss": 0.0044, + "step": 26452 + }, + { + "epoch": 4.81, + "learning_rate": 1.5582635602475428e-06, + "loss": 0.9935, + "step": 26453 + }, + { + "epoch": 4.81, + "learning_rate": 1.5582089552238807e-06, + "loss": 0.4762, + "step": 26454 + }, + { + "epoch": 4.81, + "learning_rate": 1.5581543502002184e-06, + "loss": 0.8702, + "step": 26455 + }, + { + "epoch": 4.81, + "learning_rate": 1.5580997451765564e-06, + "loss": 0.13, + "step": 26456 + }, + { + "epoch": 4.81, + "learning_rate": 1.5580451401528943e-06, + "loss": 0.1554, + "step": 26457 + }, + { + "epoch": 4.81, + "learning_rate": 1.5579905351292318e-06, + "loss": 0.2639, + "step": 26458 + }, + { + "epoch": 4.81, + "learning_rate": 1.5579359301055698e-06, + "loss": 0.8078, + "step": 26459 + }, + { + "epoch": 4.81, + "learning_rate": 1.5578813250819077e-06, + "loss": 1.1854, + "step": 26460 + }, + { + "epoch": 4.81, + "learning_rate": 1.5578267200582452e-06, + "loss": 0.5161, + "step": 26461 + }, + { + "epoch": 4.81, + "learning_rate": 1.5577721150345832e-06, + "loss": 0.9008, + "step": 26462 + }, + { + "epoch": 4.81, + "learning_rate": 1.557717510010921e-06, + "loss": 0.0157, + "step": 26463 + }, + { + "epoch": 4.81, + "learning_rate": 1.5576629049872589e-06, + "loss": 0.1461, + "step": 26464 + }, + { + "epoch": 4.81, + "learning_rate": 1.5576082999635968e-06, + "loss": 1.4756, + "step": 26465 + }, + { + "epoch": 4.81, + "learning_rate": 1.5575536949399343e-06, + "loss": 0.9624, + "step": 26466 + }, + { + "epoch": 4.81, + "learning_rate": 1.5574990899162723e-06, + "loss": 0.011, + "step": 26467 + }, + { + "epoch": 4.81, + "learning_rate": 1.5574444848926102e-06, + "loss": 0.4081, + "step": 26468 + }, + { + "epoch": 4.81, + "learning_rate": 1.557389879868948e-06, + "loss": 1.6542, + "step": 26469 + }, + { + "epoch": 4.81, + "learning_rate": 1.5573352748452859e-06, + "loss": 0.5882, + "step": 26470 + }, + { + "epoch": 4.81, + "learning_rate": 1.5572806698216238e-06, + "loss": 1.9105, + "step": 26471 + }, + { + "epoch": 4.81, + "learning_rate": 1.5572260647979614e-06, + "loss": 1.7603, + "step": 26472 + }, + { + "epoch": 4.81, + "learning_rate": 1.5571714597742993e-06, + "loss": 0.6496, + "step": 26473 + }, + { + "epoch": 4.81, + "learning_rate": 1.5571168547506372e-06, + "loss": 0.3688, + "step": 26474 + }, + { + "epoch": 4.81, + "learning_rate": 1.5570622497269748e-06, + "loss": 0.0907, + "step": 26475 + }, + { + "epoch": 4.81, + "learning_rate": 1.5570076447033127e-06, + "loss": 0.162, + "step": 26476 + }, + { + "epoch": 4.81, + "learning_rate": 1.5569530396796504e-06, + "loss": 0.2436, + "step": 26477 + }, + { + "epoch": 4.82, + "learning_rate": 1.5568984346559884e-06, + "loss": 0.0195, + "step": 26478 + }, + { + "epoch": 4.82, + "learning_rate": 1.5568438296323263e-06, + "loss": 1.6188, + "step": 26479 + }, + { + "epoch": 4.82, + "learning_rate": 1.5567892246086638e-06, + "loss": 0.0396, + "step": 26480 + }, + { + "epoch": 4.82, + "learning_rate": 1.5567346195850018e-06, + "loss": 2.5201, + "step": 26481 + }, + { + "epoch": 4.82, + "learning_rate": 1.5566800145613397e-06, + "loss": 0.161, + "step": 26482 + }, + { + "epoch": 4.82, + "learning_rate": 1.5566254095376775e-06, + "loss": 0.9964, + "step": 26483 + }, + { + "epoch": 4.82, + "learning_rate": 1.5565708045140154e-06, + "loss": 0.8984, + "step": 26484 + }, + { + "epoch": 4.82, + "learning_rate": 1.5565161994903533e-06, + "loss": 0.011, + "step": 26485 + }, + { + "epoch": 4.82, + "learning_rate": 1.5564615944666909e-06, + "loss": 0.7063, + "step": 26486 + }, + { + "epoch": 4.82, + "learning_rate": 1.5564069894430288e-06, + "loss": 0.3776, + "step": 26487 + }, + { + "epoch": 4.82, + "learning_rate": 1.5563523844193668e-06, + "loss": 2.0124, + "step": 26488 + }, + { + "epoch": 4.82, + "learning_rate": 1.5562977793957045e-06, + "loss": 0.8746, + "step": 26489 + }, + { + "epoch": 4.82, + "learning_rate": 1.5562431743720424e-06, + "loss": 1.0262, + "step": 26490 + }, + { + "epoch": 4.82, + "learning_rate": 1.55618856934838e-06, + "loss": 1.6206, + "step": 26491 + }, + { + "epoch": 4.82, + "learning_rate": 1.5561339643247179e-06, + "loss": 0.156, + "step": 26492 + }, + { + "epoch": 4.82, + "learning_rate": 1.5560793593010558e-06, + "loss": 0.1625, + "step": 26493 + }, + { + "epoch": 4.82, + "learning_rate": 1.5560247542773934e-06, + "loss": 0.5543, + "step": 26494 + }, + { + "epoch": 4.82, + "learning_rate": 1.5559701492537313e-06, + "loss": 0.0123, + "step": 26495 + }, + { + "epoch": 4.82, + "learning_rate": 1.5559155442300692e-06, + "loss": 2.2095, + "step": 26496 + }, + { + "epoch": 4.82, + "learning_rate": 1.555860939206407e-06, + "loss": 2.06, + "step": 26497 + }, + { + "epoch": 4.82, + "learning_rate": 1.555806334182745e-06, + "loss": 0.2563, + "step": 26498 + }, + { + "epoch": 4.82, + "learning_rate": 1.5557517291590829e-06, + "loss": 0.3802, + "step": 26499 + }, + { + "epoch": 4.82, + "learning_rate": 1.5556971241354204e-06, + "loss": 1.5873, + "step": 26500 + }, + { + "epoch": 4.82, + "learning_rate": 1.5556425191117583e-06, + "loss": 0.4271, + "step": 26501 + }, + { + "epoch": 4.82, + "learning_rate": 1.5555879140880963e-06, + "loss": 0.4343, + "step": 26502 + }, + { + "epoch": 4.82, + "learning_rate": 1.555533309064434e-06, + "loss": 0.2795, + "step": 26503 + }, + { + "epoch": 4.82, + "learning_rate": 1.555478704040772e-06, + "loss": 2.6646, + "step": 26504 + }, + { + "epoch": 4.82, + "learning_rate": 1.5554240990171095e-06, + "loss": 1.0972, + "step": 26505 + }, + { + "epoch": 4.82, + "learning_rate": 1.5553694939934474e-06, + "loss": 1.8619, + "step": 26506 + }, + { + "epoch": 4.82, + "learning_rate": 1.5553148889697854e-06, + "loss": 0.1201, + "step": 26507 + }, + { + "epoch": 4.82, + "learning_rate": 1.5552602839461229e-06, + "loss": 1.2546, + "step": 26508 + }, + { + "epoch": 4.82, + "learning_rate": 1.5552056789224608e-06, + "loss": 0.2011, + "step": 26509 + }, + { + "epoch": 4.82, + "learning_rate": 1.5551510738987988e-06, + "loss": 0.4539, + "step": 26510 + }, + { + "epoch": 4.82, + "learning_rate": 1.5550964688751365e-06, + "loss": 0.1397, + "step": 26511 + }, + { + "epoch": 4.82, + "learning_rate": 1.5550418638514744e-06, + "loss": 1.1459, + "step": 26512 + }, + { + "epoch": 4.82, + "learning_rate": 1.5549872588278124e-06, + "loss": 0.3597, + "step": 26513 + }, + { + "epoch": 4.82, + "learning_rate": 1.55493265380415e-06, + "loss": 0.2818, + "step": 26514 + }, + { + "epoch": 4.82, + "learning_rate": 1.5548780487804878e-06, + "loss": 5.9267, + "step": 26515 + }, + { + "epoch": 4.82, + "learning_rate": 1.5548234437568256e-06, + "loss": 0.1058, + "step": 26516 + }, + { + "epoch": 4.82, + "learning_rate": 1.5547688387331635e-06, + "loss": 0.0077, + "step": 26517 + }, + { + "epoch": 4.82, + "learning_rate": 1.5547142337095015e-06, + "loss": 0.4504, + "step": 26518 + }, + { + "epoch": 4.82, + "learning_rate": 1.554659628685839e-06, + "loss": 0.817, + "step": 26519 + }, + { + "epoch": 4.82, + "learning_rate": 1.554605023662177e-06, + "loss": 0.3622, + "step": 26520 + }, + { + "epoch": 4.82, + "learning_rate": 1.5545504186385149e-06, + "loss": 0.3628, + "step": 26521 + }, + { + "epoch": 4.82, + "learning_rate": 1.5544958136148526e-06, + "loss": 1.1587, + "step": 26522 + }, + { + "epoch": 4.82, + "learning_rate": 1.5544412085911903e-06, + "loss": 0.5015, + "step": 26523 + }, + { + "epoch": 4.82, + "learning_rate": 1.5543866035675283e-06, + "loss": 0.8913, + "step": 26524 + }, + { + "epoch": 4.82, + "learning_rate": 1.554331998543866e-06, + "loss": 0.4259, + "step": 26525 + }, + { + "epoch": 4.82, + "learning_rate": 1.554277393520204e-06, + "loss": 0.3427, + "step": 26526 + }, + { + "epoch": 4.82, + "learning_rate": 1.5542227884965419e-06, + "loss": 0.0524, + "step": 26527 + }, + { + "epoch": 4.82, + "learning_rate": 1.5541681834728794e-06, + "loss": 0.0097, + "step": 26528 + }, + { + "epoch": 4.82, + "learning_rate": 1.5541135784492174e-06, + "loss": 0.4317, + "step": 26529 + }, + { + "epoch": 4.82, + "learning_rate": 1.554058973425555e-06, + "loss": 1.0179, + "step": 26530 + }, + { + "epoch": 4.82, + "learning_rate": 1.554004368401893e-06, + "loss": 2.2241, + "step": 26531 + }, + { + "epoch": 4.82, + "learning_rate": 1.553949763378231e-06, + "loss": 0.0099, + "step": 26532 + }, + { + "epoch": 4.83, + "learning_rate": 1.5538951583545685e-06, + "loss": 0.9115, + "step": 26533 + }, + { + "epoch": 4.83, + "learning_rate": 1.5538405533309064e-06, + "loss": 0.3567, + "step": 26534 + }, + { + "epoch": 4.83, + "learning_rate": 1.5537859483072444e-06, + "loss": 0.4231, + "step": 26535 + }, + { + "epoch": 4.83, + "learning_rate": 1.5537313432835821e-06, + "loss": 0.177, + "step": 26536 + }, + { + "epoch": 4.83, + "learning_rate": 1.55367673825992e-06, + "loss": 0.579, + "step": 26537 + }, + { + "epoch": 4.83, + "learning_rate": 1.553622133236258e-06, + "loss": 0.3271, + "step": 26538 + }, + { + "epoch": 4.83, + "learning_rate": 1.5535675282125955e-06, + "loss": 0.4378, + "step": 26539 + }, + { + "epoch": 4.83, + "learning_rate": 1.5535129231889335e-06, + "loss": 0.9407, + "step": 26540 + }, + { + "epoch": 4.83, + "learning_rate": 1.5534583181652714e-06, + "loss": 2.3179, + "step": 26541 + }, + { + "epoch": 4.83, + "learning_rate": 1.553403713141609e-06, + "loss": 0.5394, + "step": 26542 + }, + { + "epoch": 4.83, + "learning_rate": 1.5533491081179469e-06, + "loss": 1.1168, + "step": 26543 + }, + { + "epoch": 4.83, + "learning_rate": 1.5532945030942846e-06, + "loss": 0.6434, + "step": 26544 + }, + { + "epoch": 4.83, + "learning_rate": 1.5532398980706225e-06, + "loss": 0.4949, + "step": 26545 + }, + { + "epoch": 4.83, + "learning_rate": 1.5531852930469605e-06, + "loss": 0.0582, + "step": 26546 + }, + { + "epoch": 4.83, + "learning_rate": 1.553130688023298e-06, + "loss": 2.5121, + "step": 26547 + }, + { + "epoch": 4.83, + "learning_rate": 1.553076082999636e-06, + "loss": 0.3158, + "step": 26548 + }, + { + "epoch": 4.83, + "learning_rate": 1.5530214779759739e-06, + "loss": 0.0581, + "step": 26549 + }, + { + "epoch": 4.83, + "learning_rate": 1.5529668729523116e-06, + "loss": 1.4639, + "step": 26550 + }, + { + "epoch": 4.83, + "learning_rate": 1.5529122679286496e-06, + "loss": 0.2977, + "step": 26551 + }, + { + "epoch": 4.83, + "learning_rate": 1.5528576629049875e-06, + "loss": 2.7991, + "step": 26552 + }, + { + "epoch": 4.83, + "learning_rate": 1.552803057881325e-06, + "loss": 1.5451, + "step": 26553 + }, + { + "epoch": 4.83, + "learning_rate": 1.552748452857663e-06, + "loss": 0.2069, + "step": 26554 + }, + { + "epoch": 4.83, + "learning_rate": 1.552693847834001e-06, + "loss": 0.8111, + "step": 26555 + }, + { + "epoch": 4.83, + "learning_rate": 1.5526392428103384e-06, + "loss": 0.4957, + "step": 26556 + }, + { + "epoch": 4.83, + "learning_rate": 1.5525846377866764e-06, + "loss": 2.48, + "step": 26557 + }, + { + "epoch": 4.83, + "learning_rate": 1.5525300327630141e-06, + "loss": 0.3724, + "step": 26558 + }, + { + "epoch": 4.83, + "learning_rate": 1.552475427739352e-06, + "loss": 0.0646, + "step": 26559 + }, + { + "epoch": 4.83, + "learning_rate": 1.55242082271569e-06, + "loss": 0.5648, + "step": 26560 + }, + { + "epoch": 4.83, + "learning_rate": 1.5523662176920275e-06, + "loss": 0.0256, + "step": 26561 + }, + { + "epoch": 4.83, + "learning_rate": 1.5523116126683655e-06, + "loss": 1.0124, + "step": 26562 + }, + { + "epoch": 4.83, + "learning_rate": 1.5522570076447034e-06, + "loss": 0.1244, + "step": 26563 + }, + { + "epoch": 4.83, + "learning_rate": 1.5522024026210411e-06, + "loss": 0.2796, + "step": 26564 + }, + { + "epoch": 4.83, + "learning_rate": 1.552147797597379e-06, + "loss": 0.9535, + "step": 26565 + }, + { + "epoch": 4.83, + "learning_rate": 1.552093192573717e-06, + "loss": 2.4257, + "step": 26566 + }, + { + "epoch": 4.83, + "learning_rate": 1.5520385875500545e-06, + "loss": 0.6756, + "step": 26567 + }, + { + "epoch": 4.83, + "learning_rate": 1.5519839825263925e-06, + "loss": 1.7737, + "step": 26568 + }, + { + "epoch": 4.83, + "learning_rate": 1.5519293775027304e-06, + "loss": 0.6186, + "step": 26569 + }, + { + "epoch": 4.83, + "learning_rate": 1.5518747724790682e-06, + "loss": 0.7829, + "step": 26570 + }, + { + "epoch": 4.83, + "learning_rate": 1.551820167455406e-06, + "loss": 0.1707, + "step": 26571 + }, + { + "epoch": 4.83, + "learning_rate": 1.5517655624317436e-06, + "loss": 0.0079, + "step": 26572 + }, + { + "epoch": 4.83, + "learning_rate": 1.5517109574080816e-06, + "loss": 0.025, + "step": 26573 + }, + { + "epoch": 4.83, + "learning_rate": 1.5516563523844195e-06, + "loss": 0.0214, + "step": 26574 + }, + { + "epoch": 4.83, + "learning_rate": 1.551601747360757e-06, + "loss": 0.053, + "step": 26575 + }, + { + "epoch": 4.83, + "learning_rate": 1.551547142337095e-06, + "loss": 2.4127, + "step": 26576 + }, + { + "epoch": 4.83, + "learning_rate": 1.551492537313433e-06, + "loss": 2.2756, + "step": 26577 + }, + { + "epoch": 4.83, + "learning_rate": 1.5514379322897706e-06, + "loss": 0.1145, + "step": 26578 + }, + { + "epoch": 4.83, + "learning_rate": 1.5513833272661086e-06, + "loss": 0.8076, + "step": 26579 + }, + { + "epoch": 4.83, + "learning_rate": 1.5513287222424465e-06, + "loss": 1.0434, + "step": 26580 + }, + { + "epoch": 4.83, + "learning_rate": 1.551274117218784e-06, + "loss": 1.6679, + "step": 26581 + }, + { + "epoch": 4.83, + "learning_rate": 1.551219512195122e-06, + "loss": 1.4955, + "step": 26582 + }, + { + "epoch": 4.83, + "learning_rate": 1.55116490717146e-06, + "loss": 1.7208, + "step": 26583 + }, + { + "epoch": 4.83, + "learning_rate": 1.5511103021477977e-06, + "loss": 0.5214, + "step": 26584 + }, + { + "epoch": 4.83, + "learning_rate": 1.5510556971241356e-06, + "loss": 0.1953, + "step": 26585 + }, + { + "epoch": 4.83, + "learning_rate": 1.5510010921004731e-06, + "loss": 0.1145, + "step": 26586 + }, + { + "epoch": 4.83, + "learning_rate": 1.550946487076811e-06, + "loss": 0.8531, + "step": 26587 + }, + { + "epoch": 4.84, + "learning_rate": 1.550891882053149e-06, + "loss": 1.1532, + "step": 26588 + }, + { + "epoch": 4.84, + "learning_rate": 1.5508372770294865e-06, + "loss": 0.2656, + "step": 26589 + }, + { + "epoch": 4.84, + "learning_rate": 1.5507826720058245e-06, + "loss": 0.0288, + "step": 26590 + }, + { + "epoch": 4.84, + "learning_rate": 1.5507280669821624e-06, + "loss": 1.2577, + "step": 26591 + }, + { + "epoch": 4.84, + "learning_rate": 1.5506734619585002e-06, + "loss": 0.0114, + "step": 26592 + }, + { + "epoch": 4.84, + "learning_rate": 1.5506188569348381e-06, + "loss": 0.9813, + "step": 26593 + }, + { + "epoch": 4.84, + "learning_rate": 1.550564251911176e-06, + "loss": 0.0223, + "step": 26594 + }, + { + "epoch": 4.84, + "learning_rate": 1.5505096468875136e-06, + "loss": 2.8671, + "step": 26595 + }, + { + "epoch": 4.84, + "learning_rate": 1.5504550418638515e-06, + "loss": 0.4806, + "step": 26596 + }, + { + "epoch": 4.84, + "learning_rate": 1.5504004368401892e-06, + "loss": 0.3884, + "step": 26597 + }, + { + "epoch": 4.84, + "learning_rate": 1.5503458318165272e-06, + "loss": 1.9363, + "step": 26598 + }, + { + "epoch": 4.84, + "learning_rate": 1.5502912267928651e-06, + "loss": 0.0401, + "step": 26599 + }, + { + "epoch": 4.84, + "learning_rate": 1.5502366217692027e-06, + "loss": 3.9249, + "step": 26600 + }, + { + "epoch": 4.84, + "learning_rate": 1.5501820167455406e-06, + "loss": 0.1617, + "step": 26601 + }, + { + "epoch": 4.84, + "learning_rate": 1.5501274117218785e-06, + "loss": 0.899, + "step": 26602 + }, + { + "epoch": 4.84, + "learning_rate": 1.550072806698216e-06, + "loss": 1.5284, + "step": 26603 + }, + { + "epoch": 4.84, + "learning_rate": 1.550018201674554e-06, + "loss": 0.377, + "step": 26604 + }, + { + "epoch": 4.84, + "learning_rate": 1.549963596650892e-06, + "loss": 1.0403, + "step": 26605 + }, + { + "epoch": 4.84, + "learning_rate": 1.5499089916272297e-06, + "loss": 2.0642, + "step": 26606 + }, + { + "epoch": 4.84, + "learning_rate": 1.5498543866035676e-06, + "loss": 0.4425, + "step": 26607 + }, + { + "epoch": 4.84, + "learning_rate": 1.5497997815799056e-06, + "loss": 0.2064, + "step": 26608 + }, + { + "epoch": 4.84, + "learning_rate": 1.549745176556243e-06, + "loss": 0.0873, + "step": 26609 + }, + { + "epoch": 4.84, + "learning_rate": 1.549690571532581e-06, + "loss": 1.0231, + "step": 26610 + }, + { + "epoch": 4.84, + "learning_rate": 1.5496359665089188e-06, + "loss": 0.2659, + "step": 26611 + }, + { + "epoch": 4.84, + "learning_rate": 1.5495813614852567e-06, + "loss": 0.474, + "step": 26612 + }, + { + "epoch": 4.84, + "learning_rate": 1.5495267564615946e-06, + "loss": 0.5874, + "step": 26613 + }, + { + "epoch": 4.84, + "learning_rate": 1.5494721514379322e-06, + "loss": 1.836, + "step": 26614 + }, + { + "epoch": 4.84, + "learning_rate": 1.5494175464142701e-06, + "loss": 2.0476, + "step": 26615 + }, + { + "epoch": 4.84, + "learning_rate": 1.549362941390608e-06, + "loss": 2.0343, + "step": 26616 + }, + { + "epoch": 4.84, + "learning_rate": 1.5493083363669458e-06, + "loss": 0.2352, + "step": 26617 + }, + { + "epoch": 4.84, + "learning_rate": 1.5492537313432837e-06, + "loss": 0.1291, + "step": 26618 + }, + { + "epoch": 4.84, + "learning_rate": 1.5491991263196217e-06, + "loss": 0.98, + "step": 26619 + }, + { + "epoch": 4.84, + "learning_rate": 1.5491445212959592e-06, + "loss": 1.9161, + "step": 26620 + }, + { + "epoch": 4.84, + "learning_rate": 1.5490899162722971e-06, + "loss": 1.1992, + "step": 26621 + }, + { + "epoch": 4.84, + "learning_rate": 1.549035311248635e-06, + "loss": 1.0814, + "step": 26622 + }, + { + "epoch": 4.84, + "learning_rate": 1.5489807062249726e-06, + "loss": 2.5741, + "step": 26623 + }, + { + "epoch": 4.84, + "learning_rate": 1.5489261012013105e-06, + "loss": 0.0256, + "step": 26624 + }, + { + "epoch": 4.84, + "learning_rate": 1.5488714961776483e-06, + "loss": 1.11, + "step": 26625 + }, + { + "epoch": 4.84, + "learning_rate": 1.5488168911539862e-06, + "loss": 0.9153, + "step": 26626 + }, + { + "epoch": 4.84, + "learning_rate": 1.5487622861303242e-06, + "loss": 0.743, + "step": 26627 + }, + { + "epoch": 4.84, + "learning_rate": 1.5487076811066617e-06, + "loss": 0.0537, + "step": 26628 + }, + { + "epoch": 4.84, + "learning_rate": 1.5486530760829996e-06, + "loss": 0.2692, + "step": 26629 + }, + { + "epoch": 4.84, + "learning_rate": 1.5485984710593376e-06, + "loss": 0.4628, + "step": 26630 + }, + { + "epoch": 4.84, + "learning_rate": 1.5485438660356753e-06, + "loss": 0.0102, + "step": 26631 + }, + { + "epoch": 4.84, + "learning_rate": 1.5484892610120132e-06, + "loss": 0.8028, + "step": 26632 + }, + { + "epoch": 4.84, + "learning_rate": 1.5484346559883512e-06, + "loss": 0.7983, + "step": 26633 + }, + { + "epoch": 4.84, + "learning_rate": 1.5483800509646887e-06, + "loss": 0.008, + "step": 26634 + }, + { + "epoch": 4.84, + "learning_rate": 1.5483254459410266e-06, + "loss": 0.935, + "step": 26635 + }, + { + "epoch": 4.84, + "learning_rate": 1.5482708409173646e-06, + "loss": 0.3408, + "step": 26636 + }, + { + "epoch": 4.84, + "learning_rate": 1.5482162358937021e-06, + "loss": 1.7748, + "step": 26637 + }, + { + "epoch": 4.84, + "learning_rate": 1.54816163087004e-06, + "loss": 0.1303, + "step": 26638 + }, + { + "epoch": 4.84, + "learning_rate": 1.5481070258463778e-06, + "loss": 0.2303, + "step": 26639 + }, + { + "epoch": 4.84, + "learning_rate": 1.5480524208227157e-06, + "loss": 1.5625, + "step": 26640 + }, + { + "epoch": 4.84, + "learning_rate": 1.5479978157990537e-06, + "loss": 0.0348, + "step": 26641 + }, + { + "epoch": 4.84, + "learning_rate": 1.5479432107753912e-06, + "loss": 0.7793, + "step": 26642 + }, + { + "epoch": 4.85, + "learning_rate": 1.5478886057517291e-06, + "loss": 0.0589, + "step": 26643 + }, + { + "epoch": 4.85, + "learning_rate": 1.547834000728067e-06, + "loss": 0.1211, + "step": 26644 + }, + { + "epoch": 4.85, + "learning_rate": 1.5477793957044048e-06, + "loss": 1.7018, + "step": 26645 + }, + { + "epoch": 4.85, + "learning_rate": 1.5477247906807428e-06, + "loss": 0.3078, + "step": 26646 + }, + { + "epoch": 4.85, + "learning_rate": 1.5476701856570807e-06, + "loss": 0.4095, + "step": 26647 + }, + { + "epoch": 4.85, + "learning_rate": 1.5476155806334182e-06, + "loss": 1.0686, + "step": 26648 + }, + { + "epoch": 4.85, + "learning_rate": 1.5475609756097562e-06, + "loss": 1.0002, + "step": 26649 + }, + { + "epoch": 4.85, + "learning_rate": 1.547506370586094e-06, + "loss": 0.0262, + "step": 26650 + }, + { + "epoch": 4.85, + "learning_rate": 1.5474517655624318e-06, + "loss": 0.5919, + "step": 26651 + }, + { + "epoch": 4.85, + "learning_rate": 1.5473971605387696e-06, + "loss": 2.0771, + "step": 26652 + }, + { + "epoch": 4.85, + "learning_rate": 1.5473425555151073e-06, + "loss": 0.7149, + "step": 26653 + }, + { + "epoch": 4.85, + "learning_rate": 1.5472879504914452e-06, + "loss": 0.9033, + "step": 26654 + }, + { + "epoch": 4.85, + "learning_rate": 1.5472333454677832e-06, + "loss": 0.0118, + "step": 26655 + }, + { + "epoch": 4.85, + "learning_rate": 1.5471787404441207e-06, + "loss": 0.3045, + "step": 26656 + }, + { + "epoch": 4.85, + "learning_rate": 1.5471241354204586e-06, + "loss": 0.2136, + "step": 26657 + }, + { + "epoch": 4.85, + "learning_rate": 1.5470695303967966e-06, + "loss": 0.1358, + "step": 26658 + }, + { + "epoch": 4.85, + "learning_rate": 1.5470149253731343e-06, + "loss": 0.2287, + "step": 26659 + }, + { + "epoch": 4.85, + "learning_rate": 1.5469603203494723e-06, + "loss": 0.6899, + "step": 26660 + }, + { + "epoch": 4.85, + "learning_rate": 1.5469057153258102e-06, + "loss": 2.0947, + "step": 26661 + }, + { + "epoch": 4.85, + "learning_rate": 1.5468511103021477e-06, + "loss": 0.7459, + "step": 26662 + }, + { + "epoch": 4.85, + "learning_rate": 1.5467965052784857e-06, + "loss": 2.5877, + "step": 26663 + }, + { + "epoch": 4.85, + "learning_rate": 1.5467419002548236e-06, + "loss": 0.0972, + "step": 26664 + }, + { + "epoch": 4.85, + "learning_rate": 1.5466872952311613e-06, + "loss": 1.3161, + "step": 26665 + }, + { + "epoch": 4.85, + "learning_rate": 1.5466326902074993e-06, + "loss": 0.5949, + "step": 26666 + }, + { + "epoch": 4.85, + "learning_rate": 1.5465780851838368e-06, + "loss": 1.1887, + "step": 26667 + }, + { + "epoch": 4.85, + "learning_rate": 1.5465234801601748e-06, + "loss": 0.2465, + "step": 26668 + }, + { + "epoch": 4.85, + "learning_rate": 1.5464688751365127e-06, + "loss": 0.3438, + "step": 26669 + }, + { + "epoch": 4.85, + "learning_rate": 1.5464142701128502e-06, + "loss": 0.3536, + "step": 26670 + }, + { + "epoch": 4.85, + "learning_rate": 1.5463596650891882e-06, + "loss": 0.0536, + "step": 26671 + }, + { + "epoch": 4.85, + "learning_rate": 1.546305060065526e-06, + "loss": 0.0949, + "step": 26672 + }, + { + "epoch": 4.85, + "learning_rate": 1.5462504550418638e-06, + "loss": 0.4938, + "step": 26673 + }, + { + "epoch": 4.85, + "learning_rate": 1.5461958500182018e-06, + "loss": 0.1605, + "step": 26674 + }, + { + "epoch": 4.85, + "learning_rate": 1.5461412449945397e-06, + "loss": 1.2692, + "step": 26675 + }, + { + "epoch": 4.85, + "learning_rate": 1.5460866399708772e-06, + "loss": 0.3112, + "step": 26676 + }, + { + "epoch": 4.85, + "learning_rate": 1.5460320349472152e-06, + "loss": 0.8258, + "step": 26677 + }, + { + "epoch": 4.85, + "learning_rate": 1.545977429923553e-06, + "loss": 2.0106, + "step": 26678 + }, + { + "epoch": 4.85, + "learning_rate": 1.5459228248998909e-06, + "loss": 0.7573, + "step": 26679 + }, + { + "epoch": 4.85, + "learning_rate": 1.5458682198762288e-06, + "loss": 0.6091, + "step": 26680 + }, + { + "epoch": 4.85, + "learning_rate": 1.5458136148525663e-06, + "loss": 0.0169, + "step": 26681 + }, + { + "epoch": 4.85, + "learning_rate": 1.5457590098289043e-06, + "loss": 1.2515, + "step": 26682 + }, + { + "epoch": 4.85, + "learning_rate": 1.5457044048052422e-06, + "loss": 1.6776, + "step": 26683 + }, + { + "epoch": 4.85, + "learning_rate": 1.5456497997815797e-06, + "loss": 2.2395, + "step": 26684 + }, + { + "epoch": 4.85, + "learning_rate": 1.5455951947579177e-06, + "loss": 1.6475, + "step": 26685 + }, + { + "epoch": 4.85, + "learning_rate": 1.5455405897342556e-06, + "loss": 0.6775, + "step": 26686 + }, + { + "epoch": 4.85, + "learning_rate": 1.5454859847105934e-06, + "loss": 0.357, + "step": 26687 + }, + { + "epoch": 4.85, + "learning_rate": 1.5454313796869313e-06, + "loss": 1.7257, + "step": 26688 + }, + { + "epoch": 4.85, + "learning_rate": 1.5453767746632692e-06, + "loss": 0.1472, + "step": 26689 + }, + { + "epoch": 4.85, + "learning_rate": 1.5453221696396068e-06, + "loss": 2.6163, + "step": 26690 + }, + { + "epoch": 4.85, + "learning_rate": 1.5452675646159447e-06, + "loss": 0.0115, + "step": 26691 + }, + { + "epoch": 4.85, + "learning_rate": 1.5452129595922824e-06, + "loss": 0.4105, + "step": 26692 + }, + { + "epoch": 4.85, + "learning_rate": 1.5451583545686204e-06, + "loss": 3.1114, + "step": 26693 + }, + { + "epoch": 4.85, + "learning_rate": 1.5451037495449583e-06, + "loss": 2.0553, + "step": 26694 + }, + { + "epoch": 4.85, + "learning_rate": 1.5450491445212958e-06, + "loss": 0.2623, + "step": 26695 + }, + { + "epoch": 4.85, + "learning_rate": 1.5449945394976338e-06, + "loss": 1.0475, + "step": 26696 + }, + { + "epoch": 4.85, + "learning_rate": 1.5449399344739717e-06, + "loss": 0.2259, + "step": 26697 + }, + { + "epoch": 4.86, + "learning_rate": 1.5448853294503095e-06, + "loss": 1.4003, + "step": 26698 + }, + { + "epoch": 4.86, + "learning_rate": 1.5448307244266474e-06, + "loss": 0.1278, + "step": 26699 + }, + { + "epoch": 4.86, + "learning_rate": 1.5447761194029853e-06, + "loss": 0.8691, + "step": 26700 + }, + { + "epoch": 4.86, + "learning_rate": 1.5447215143793229e-06, + "loss": 0.1354, + "step": 26701 + }, + { + "epoch": 4.86, + "learning_rate": 1.5446669093556608e-06, + "loss": 0.3553, + "step": 26702 + }, + { + "epoch": 4.86, + "learning_rate": 1.5446123043319988e-06, + "loss": 0.2526, + "step": 26703 + }, + { + "epoch": 4.86, + "learning_rate": 1.5445576993083363e-06, + "loss": 0.3886, + "step": 26704 + }, + { + "epoch": 4.86, + "learning_rate": 1.5445030942846742e-06, + "loss": 0.3855, + "step": 26705 + }, + { + "epoch": 4.86, + "learning_rate": 1.544448489261012e-06, + "loss": 0.2995, + "step": 26706 + }, + { + "epoch": 4.86, + "learning_rate": 1.5443938842373499e-06, + "loss": 0.0765, + "step": 26707 + }, + { + "epoch": 4.86, + "learning_rate": 1.5443392792136878e-06, + "loss": 0.6736, + "step": 26708 + }, + { + "epoch": 4.86, + "learning_rate": 1.5442846741900254e-06, + "loss": 0.076, + "step": 26709 + }, + { + "epoch": 4.86, + "learning_rate": 1.5442300691663633e-06, + "loss": 1.0381, + "step": 26710 + }, + { + "epoch": 4.86, + "learning_rate": 1.5441754641427012e-06, + "loss": 0.3677, + "step": 26711 + }, + { + "epoch": 4.86, + "learning_rate": 1.544120859119039e-06, + "loss": 0.1734, + "step": 26712 + }, + { + "epoch": 4.86, + "learning_rate": 1.544066254095377e-06, + "loss": 0.429, + "step": 26713 + }, + { + "epoch": 4.86, + "learning_rate": 1.5440116490717149e-06, + "loss": 1.3493, + "step": 26714 + }, + { + "epoch": 4.86, + "learning_rate": 1.5439570440480524e-06, + "loss": 2.434, + "step": 26715 + }, + { + "epoch": 4.86, + "learning_rate": 1.5439024390243903e-06, + "loss": 0.399, + "step": 26716 + }, + { + "epoch": 4.86, + "learning_rate": 1.5438478340007283e-06, + "loss": 0.526, + "step": 26717 + }, + { + "epoch": 4.86, + "learning_rate": 1.5437932289770658e-06, + "loss": 0.158, + "step": 26718 + }, + { + "epoch": 4.86, + "learning_rate": 1.5437386239534037e-06, + "loss": 1.2931, + "step": 26719 + }, + { + "epoch": 4.86, + "learning_rate": 1.5436840189297415e-06, + "loss": 0.019, + "step": 26720 + }, + { + "epoch": 4.86, + "learning_rate": 1.5436294139060794e-06, + "loss": 0.3807, + "step": 26721 + }, + { + "epoch": 4.86, + "learning_rate": 1.5435748088824173e-06, + "loss": 0.4175, + "step": 26722 + }, + { + "epoch": 4.86, + "learning_rate": 1.5435202038587549e-06, + "loss": 0.5713, + "step": 26723 + }, + { + "epoch": 4.86, + "learning_rate": 1.5434655988350928e-06, + "loss": 0.2235, + "step": 26724 + }, + { + "epoch": 4.86, + "learning_rate": 1.5434109938114308e-06, + "loss": 0.0156, + "step": 26725 + }, + { + "epoch": 4.86, + "learning_rate": 1.5433563887877685e-06, + "loss": 1.9429, + "step": 26726 + }, + { + "epoch": 4.86, + "learning_rate": 1.5433017837641064e-06, + "loss": 0.1124, + "step": 26727 + }, + { + "epoch": 4.86, + "learning_rate": 1.5432471787404444e-06, + "loss": 0.1278, + "step": 26728 + }, + { + "epoch": 4.86, + "learning_rate": 1.5431925737167819e-06, + "loss": 0.0596, + "step": 26729 + }, + { + "epoch": 4.86, + "learning_rate": 1.5431379686931198e-06, + "loss": 0.1934, + "step": 26730 + }, + { + "epoch": 4.86, + "learning_rate": 1.5430833636694578e-06, + "loss": 1.3042, + "step": 26731 + }, + { + "epoch": 4.86, + "learning_rate": 1.5430287586457953e-06, + "loss": 0.3391, + "step": 26732 + }, + { + "epoch": 4.86, + "learning_rate": 1.5429741536221332e-06, + "loss": 0.6037, + "step": 26733 + }, + { + "epoch": 4.86, + "learning_rate": 1.542919548598471e-06, + "loss": 0.1744, + "step": 26734 + }, + { + "epoch": 4.86, + "learning_rate": 1.542864943574809e-06, + "loss": 0.3109, + "step": 26735 + }, + { + "epoch": 4.86, + "learning_rate": 1.5428103385511469e-06, + "loss": 0.36, + "step": 26736 + }, + { + "epoch": 4.86, + "learning_rate": 1.5427557335274844e-06, + "loss": 1.5406, + "step": 26737 + }, + { + "epoch": 4.86, + "learning_rate": 1.5427011285038223e-06, + "loss": 0.1417, + "step": 26738 + }, + { + "epoch": 4.86, + "learning_rate": 1.5426465234801603e-06, + "loss": 3.1036, + "step": 26739 + }, + { + "epoch": 4.86, + "learning_rate": 1.542591918456498e-06, + "loss": 0.0315, + "step": 26740 + }, + { + "epoch": 4.86, + "learning_rate": 1.542537313432836e-06, + "loss": 0.0325, + "step": 26741 + }, + { + "epoch": 4.86, + "learning_rate": 1.5424827084091739e-06, + "loss": 0.0538, + "step": 26742 + }, + { + "epoch": 4.86, + "learning_rate": 1.5424281033855114e-06, + "loss": 1.7848, + "step": 26743 + }, + { + "epoch": 4.86, + "learning_rate": 1.5423734983618493e-06, + "loss": 0.0544, + "step": 26744 + }, + { + "epoch": 4.86, + "learning_rate": 1.542318893338187e-06, + "loss": 1.7563, + "step": 26745 + }, + { + "epoch": 4.86, + "learning_rate": 1.542264288314525e-06, + "loss": 1.4367, + "step": 26746 + }, + { + "epoch": 4.86, + "learning_rate": 1.542209683290863e-06, + "loss": 0.0087, + "step": 26747 + }, + { + "epoch": 4.86, + "learning_rate": 1.5421550782672005e-06, + "loss": 0.0552, + "step": 26748 + }, + { + "epoch": 4.86, + "learning_rate": 1.5421004732435384e-06, + "loss": 1.6721, + "step": 26749 + }, + { + "epoch": 4.86, + "learning_rate": 1.5420458682198764e-06, + "loss": 0.2867, + "step": 26750 + }, + { + "epoch": 4.86, + "learning_rate": 1.5419912631962139e-06, + "loss": 0.5603, + "step": 26751 + }, + { + "epoch": 4.86, + "learning_rate": 1.5419366581725518e-06, + "loss": 0.0187, + "step": 26752 + }, + { + "epoch": 4.87, + "learning_rate": 1.5418820531488898e-06, + "loss": 0.3272, + "step": 26753 + }, + { + "epoch": 4.87, + "learning_rate": 1.5418274481252275e-06, + "loss": 0.299, + "step": 26754 + }, + { + "epoch": 4.87, + "learning_rate": 1.5417728431015655e-06, + "loss": 0.09, + "step": 26755 + }, + { + "epoch": 4.87, + "learning_rate": 1.5417182380779034e-06, + "loss": 1.639, + "step": 26756 + }, + { + "epoch": 4.87, + "learning_rate": 1.541663633054241e-06, + "loss": 0.5647, + "step": 26757 + }, + { + "epoch": 4.87, + "learning_rate": 1.5416090280305789e-06, + "loss": 2.0137, + "step": 26758 + }, + { + "epoch": 4.87, + "learning_rate": 1.5415544230069166e-06, + "loss": 0.1533, + "step": 26759 + }, + { + "epoch": 4.87, + "learning_rate": 1.5414998179832545e-06, + "loss": 0.4235, + "step": 26760 + }, + { + "epoch": 4.87, + "learning_rate": 1.5414452129595925e-06, + "loss": 1.8543, + "step": 26761 + }, + { + "epoch": 4.87, + "learning_rate": 1.54139060793593e-06, + "loss": 2.5656, + "step": 26762 + }, + { + "epoch": 4.87, + "learning_rate": 1.541336002912268e-06, + "loss": 0.3853, + "step": 26763 + }, + { + "epoch": 4.87, + "learning_rate": 1.5412813978886059e-06, + "loss": 1.7724, + "step": 26764 + }, + { + "epoch": 4.87, + "learning_rate": 1.5412267928649434e-06, + "loss": 0.0655, + "step": 26765 + }, + { + "epoch": 4.87, + "learning_rate": 1.5411721878412814e-06, + "loss": 1.9667, + "step": 26766 + }, + { + "epoch": 4.87, + "learning_rate": 1.5411175828176193e-06, + "loss": 0.0234, + "step": 26767 + }, + { + "epoch": 4.87, + "learning_rate": 1.541062977793957e-06, + "loss": 0.1989, + "step": 26768 + }, + { + "epoch": 4.87, + "learning_rate": 1.541008372770295e-06, + "loss": 0.4717, + "step": 26769 + }, + { + "epoch": 4.87, + "learning_rate": 1.540953767746633e-06, + "loss": 0.1434, + "step": 26770 + }, + { + "epoch": 4.87, + "learning_rate": 1.5408991627229704e-06, + "loss": 2.0521, + "step": 26771 + }, + { + "epoch": 4.87, + "learning_rate": 1.5408445576993084e-06, + "loss": 0.2474, + "step": 26772 + }, + { + "epoch": 4.87, + "learning_rate": 1.5407899526756461e-06, + "loss": 0.5597, + "step": 26773 + }, + { + "epoch": 4.87, + "learning_rate": 1.540735347651984e-06, + "loss": 1.0028, + "step": 26774 + }, + { + "epoch": 4.87, + "learning_rate": 1.540680742628322e-06, + "loss": 0.0059, + "step": 26775 + }, + { + "epoch": 4.87, + "learning_rate": 1.5406261376046595e-06, + "loss": 1.1103, + "step": 26776 + }, + { + "epoch": 4.87, + "learning_rate": 1.5405715325809975e-06, + "loss": 0.5124, + "step": 26777 + }, + { + "epoch": 4.87, + "learning_rate": 1.5405169275573354e-06, + "loss": 0.2889, + "step": 26778 + }, + { + "epoch": 4.87, + "learning_rate": 1.5404623225336731e-06, + "loss": 0.3088, + "step": 26779 + }, + { + "epoch": 4.87, + "learning_rate": 1.540407717510011e-06, + "loss": 1.0583, + "step": 26780 + }, + { + "epoch": 4.87, + "learning_rate": 1.540353112486349e-06, + "loss": 0.4399, + "step": 26781 + }, + { + "epoch": 4.87, + "learning_rate": 1.5402985074626865e-06, + "loss": 2.288, + "step": 26782 + }, + { + "epoch": 4.87, + "learning_rate": 1.5402439024390245e-06, + "loss": 0.012, + "step": 26783 + }, + { + "epoch": 4.87, + "learning_rate": 1.5401892974153624e-06, + "loss": 0.6804, + "step": 26784 + }, + { + "epoch": 4.87, + "learning_rate": 1.5401346923917e-06, + "loss": 0.1654, + "step": 26785 + }, + { + "epoch": 4.87, + "learning_rate": 1.5400800873680379e-06, + "loss": 2.0729, + "step": 26786 + }, + { + "epoch": 4.87, + "learning_rate": 1.5400254823443756e-06, + "loss": 1.1683, + "step": 26787 + }, + { + "epoch": 4.87, + "learning_rate": 1.5399708773207136e-06, + "loss": 0.8831, + "step": 26788 + }, + { + "epoch": 4.87, + "learning_rate": 1.5399162722970515e-06, + "loss": 0.7964, + "step": 26789 + }, + { + "epoch": 4.87, + "learning_rate": 1.539861667273389e-06, + "loss": 0.4452, + "step": 26790 + }, + { + "epoch": 4.87, + "learning_rate": 1.539807062249727e-06, + "loss": 2.0548, + "step": 26791 + }, + { + "epoch": 4.87, + "learning_rate": 1.539752457226065e-06, + "loss": 0.4276, + "step": 26792 + }, + { + "epoch": 4.87, + "learning_rate": 1.5396978522024026e-06, + "loss": 0.2591, + "step": 26793 + }, + { + "epoch": 4.87, + "learning_rate": 1.5396432471787406e-06, + "loss": 0.3267, + "step": 26794 + }, + { + "epoch": 4.87, + "learning_rate": 1.5395886421550785e-06, + "loss": 0.2533, + "step": 26795 + }, + { + "epoch": 4.87, + "learning_rate": 1.539534037131416e-06, + "loss": 0.7355, + "step": 26796 + }, + { + "epoch": 4.87, + "learning_rate": 1.539479432107754e-06, + "loss": 0.2379, + "step": 26797 + }, + { + "epoch": 4.87, + "learning_rate": 1.539424827084092e-06, + "loss": 0.0579, + "step": 26798 + }, + { + "epoch": 4.87, + "learning_rate": 1.5393702220604295e-06, + "loss": 0.0113, + "step": 26799 + }, + { + "epoch": 4.87, + "learning_rate": 1.5393156170367674e-06, + "loss": 0.0109, + "step": 26800 + }, + { + "epoch": 4.87, + "learning_rate": 1.5392610120131051e-06, + "loss": 0.0585, + "step": 26801 + }, + { + "epoch": 4.87, + "learning_rate": 1.539206406989443e-06, + "loss": 0.2939, + "step": 26802 + }, + { + "epoch": 4.87, + "learning_rate": 1.539151801965781e-06, + "loss": 0.0238, + "step": 26803 + }, + { + "epoch": 4.87, + "learning_rate": 1.5390971969421185e-06, + "loss": 0.021, + "step": 26804 + }, + { + "epoch": 4.87, + "learning_rate": 1.5390425919184565e-06, + "loss": 0.0627, + "step": 26805 + }, + { + "epoch": 4.87, + "learning_rate": 1.5389879868947944e-06, + "loss": 0.2471, + "step": 26806 + }, + { + "epoch": 4.87, + "learning_rate": 1.5389333818711322e-06, + "loss": 0.2661, + "step": 26807 + }, + { + "epoch": 4.88, + "learning_rate": 1.53887877684747e-06, + "loss": 0.6615, + "step": 26808 + }, + { + "epoch": 4.88, + "learning_rate": 1.538824171823808e-06, + "loss": 0.2671, + "step": 26809 + }, + { + "epoch": 4.88, + "learning_rate": 1.5387695668001456e-06, + "loss": 0.2409, + "step": 26810 + }, + { + "epoch": 4.88, + "learning_rate": 1.5387149617764835e-06, + "loss": 2.1045, + "step": 26811 + }, + { + "epoch": 4.88, + "learning_rate": 1.5386603567528215e-06, + "loss": 0.1542, + "step": 26812 + }, + { + "epoch": 4.88, + "learning_rate": 1.538605751729159e-06, + "loss": 0.7794, + "step": 26813 + }, + { + "epoch": 4.88, + "learning_rate": 1.538551146705497e-06, + "loss": 0.1454, + "step": 26814 + }, + { + "epoch": 4.88, + "learning_rate": 1.5384965416818346e-06, + "loss": 0.2834, + "step": 26815 + }, + { + "epoch": 4.88, + "learning_rate": 1.5384419366581726e-06, + "loss": 0.0757, + "step": 26816 + }, + { + "epoch": 4.88, + "learning_rate": 1.5383873316345105e-06, + "loss": 0.0585, + "step": 26817 + }, + { + "epoch": 4.88, + "learning_rate": 1.538332726610848e-06, + "loss": 0.0259, + "step": 26818 + }, + { + "epoch": 4.88, + "learning_rate": 1.538278121587186e-06, + "loss": 0.3796, + "step": 26819 + }, + { + "epoch": 4.88, + "learning_rate": 1.538223516563524e-06, + "loss": 0.2443, + "step": 26820 + }, + { + "epoch": 4.88, + "learning_rate": 1.5381689115398617e-06, + "loss": 1.7407, + "step": 26821 + }, + { + "epoch": 4.88, + "learning_rate": 1.5381143065161996e-06, + "loss": 2.0766, + "step": 26822 + }, + { + "epoch": 4.88, + "learning_rate": 1.5380597014925376e-06, + "loss": 0.2621, + "step": 26823 + }, + { + "epoch": 4.88, + "learning_rate": 1.538005096468875e-06, + "loss": 0.0553, + "step": 26824 + }, + { + "epoch": 4.88, + "learning_rate": 1.537950491445213e-06, + "loss": 0.6711, + "step": 26825 + }, + { + "epoch": 4.88, + "learning_rate": 1.5378958864215508e-06, + "loss": 0.3296, + "step": 26826 + }, + { + "epoch": 4.88, + "learning_rate": 1.5378412813978887e-06, + "loss": 0.1727, + "step": 26827 + }, + { + "epoch": 4.88, + "learning_rate": 1.5377866763742266e-06, + "loss": 0.5928, + "step": 26828 + }, + { + "epoch": 4.88, + "learning_rate": 1.5377320713505642e-06, + "loss": 3.0525, + "step": 26829 + }, + { + "epoch": 4.88, + "learning_rate": 1.537677466326902e-06, + "loss": 0.7835, + "step": 26830 + }, + { + "epoch": 4.88, + "learning_rate": 1.53762286130324e-06, + "loss": 0.0258, + "step": 26831 + }, + { + "epoch": 4.88, + "learning_rate": 1.5375682562795776e-06, + "loss": 1.7174, + "step": 26832 + }, + { + "epoch": 4.88, + "learning_rate": 1.5375136512559155e-06, + "loss": 1.0095, + "step": 26833 + }, + { + "epoch": 4.88, + "learning_rate": 1.5374590462322535e-06, + "loss": 0.2267, + "step": 26834 + }, + { + "epoch": 4.88, + "learning_rate": 1.5374044412085912e-06, + "loss": 1.2999, + "step": 26835 + }, + { + "epoch": 4.88, + "learning_rate": 1.5373498361849291e-06, + "loss": 1.5475, + "step": 26836 + }, + { + "epoch": 4.88, + "learning_rate": 1.537295231161267e-06, + "loss": 1.2884, + "step": 26837 + }, + { + "epoch": 4.88, + "learning_rate": 1.5372406261376046e-06, + "loss": 0.3013, + "step": 26838 + }, + { + "epoch": 4.88, + "learning_rate": 1.5371860211139425e-06, + "loss": 0.923, + "step": 26839 + }, + { + "epoch": 4.88, + "learning_rate": 1.5371314160902803e-06, + "loss": 0.9256, + "step": 26840 + }, + { + "epoch": 4.88, + "learning_rate": 1.5370768110666182e-06, + "loss": 0.1455, + "step": 26841 + }, + { + "epoch": 4.88, + "learning_rate": 1.5370222060429562e-06, + "loss": 1.4533, + "step": 26842 + }, + { + "epoch": 4.88, + "learning_rate": 1.5369676010192937e-06, + "loss": 0.1324, + "step": 26843 + }, + { + "epoch": 4.88, + "learning_rate": 1.5369129959956316e-06, + "loss": 0.1828, + "step": 26844 + }, + { + "epoch": 4.88, + "learning_rate": 1.5368583909719696e-06, + "loss": 0.0081, + "step": 26845 + }, + { + "epoch": 4.88, + "learning_rate": 1.536803785948307e-06, + "loss": 2.229, + "step": 26846 + }, + { + "epoch": 4.88, + "learning_rate": 1.536749180924645e-06, + "loss": 0.2197, + "step": 26847 + }, + { + "epoch": 4.88, + "learning_rate": 1.536694575900983e-06, + "loss": 0.3613, + "step": 26848 + }, + { + "epoch": 4.88, + "learning_rate": 1.5366399708773207e-06, + "loss": 0.0189, + "step": 26849 + }, + { + "epoch": 4.88, + "learning_rate": 1.5365853658536586e-06, + "loss": 0.8874, + "step": 26850 + }, + { + "epoch": 4.88, + "learning_rate": 1.5365307608299966e-06, + "loss": 1.0155, + "step": 26851 + }, + { + "epoch": 4.88, + "learning_rate": 1.5364761558063341e-06, + "loss": 0.1241, + "step": 26852 + }, + { + "epoch": 4.88, + "learning_rate": 1.536421550782672e-06, + "loss": 0.5171, + "step": 26853 + }, + { + "epoch": 4.88, + "learning_rate": 1.5363669457590098e-06, + "loss": 0.5322, + "step": 26854 + }, + { + "epoch": 4.88, + "learning_rate": 1.5363123407353477e-06, + "loss": 0.737, + "step": 26855 + }, + { + "epoch": 4.88, + "learning_rate": 1.5362577357116857e-06, + "loss": 0.2248, + "step": 26856 + }, + { + "epoch": 4.88, + "learning_rate": 1.5362031306880232e-06, + "loss": 0.0991, + "step": 26857 + }, + { + "epoch": 4.88, + "learning_rate": 1.5361485256643611e-06, + "loss": 0.6103, + "step": 26858 + }, + { + "epoch": 4.88, + "learning_rate": 1.536093920640699e-06, + "loss": 0.97, + "step": 26859 + }, + { + "epoch": 4.88, + "learning_rate": 1.5360393156170366e-06, + "loss": 0.0371, + "step": 26860 + }, + { + "epoch": 4.88, + "learning_rate": 1.5359847105933745e-06, + "loss": 2.6226, + "step": 26861 + }, + { + "epoch": 4.88, + "learning_rate": 1.5359301055697125e-06, + "loss": 0.4905, + "step": 26862 + }, + { + "epoch": 4.89, + "learning_rate": 1.5358755005460502e-06, + "loss": 0.6886, + "step": 26863 + }, + { + "epoch": 4.89, + "learning_rate": 1.5358208955223882e-06, + "loss": 0.359, + "step": 26864 + }, + { + "epoch": 4.89, + "learning_rate": 1.535766290498726e-06, + "loss": 0.2505, + "step": 26865 + }, + { + "epoch": 4.89, + "learning_rate": 1.5357116854750636e-06, + "loss": 0.9689, + "step": 26866 + }, + { + "epoch": 4.89, + "learning_rate": 1.5356570804514016e-06, + "loss": 0.3878, + "step": 26867 + }, + { + "epoch": 4.89, + "learning_rate": 1.5356024754277393e-06, + "loss": 1.4094, + "step": 26868 + }, + { + "epoch": 4.89, + "learning_rate": 1.5355478704040772e-06, + "loss": 1.1318, + "step": 26869 + }, + { + "epoch": 4.89, + "learning_rate": 1.5354932653804152e-06, + "loss": 1.3089, + "step": 26870 + }, + { + "epoch": 4.89, + "learning_rate": 1.5354386603567527e-06, + "loss": 1.2672, + "step": 26871 + }, + { + "epoch": 4.89, + "learning_rate": 1.5353840553330906e-06, + "loss": 0.3783, + "step": 26872 + }, + { + "epoch": 4.89, + "learning_rate": 1.5353294503094286e-06, + "loss": 0.1284, + "step": 26873 + }, + { + "epoch": 4.89, + "learning_rate": 1.5352748452857663e-06, + "loss": 0.2256, + "step": 26874 + }, + { + "epoch": 4.89, + "learning_rate": 1.5352202402621043e-06, + "loss": 0.1881, + "step": 26875 + }, + { + "epoch": 4.89, + "learning_rate": 1.5351656352384422e-06, + "loss": 1.7638, + "step": 26876 + }, + { + "epoch": 4.89, + "learning_rate": 1.5351110302147797e-06, + "loss": 3.9972, + "step": 26877 + }, + { + "epoch": 4.89, + "learning_rate": 1.5350564251911177e-06, + "loss": 1.791, + "step": 26878 + }, + { + "epoch": 4.89, + "learning_rate": 1.5350018201674556e-06, + "loss": 2.2435, + "step": 26879 + }, + { + "epoch": 4.89, + "learning_rate": 1.5349472151437931e-06, + "loss": 2.3722, + "step": 26880 + }, + { + "epoch": 4.89, + "learning_rate": 1.534892610120131e-06, + "loss": 0.2662, + "step": 26881 + }, + { + "epoch": 4.89, + "learning_rate": 1.5348380050964688e-06, + "loss": 1.2314, + "step": 26882 + }, + { + "epoch": 4.89, + "learning_rate": 1.5347834000728068e-06, + "loss": 0.0694, + "step": 26883 + }, + { + "epoch": 4.89, + "learning_rate": 1.5347287950491447e-06, + "loss": 0.9363, + "step": 26884 + }, + { + "epoch": 4.89, + "learning_rate": 1.5346741900254822e-06, + "loss": 0.0301, + "step": 26885 + }, + { + "epoch": 4.89, + "learning_rate": 1.5346195850018202e-06, + "loss": 0.0814, + "step": 26886 + }, + { + "epoch": 4.89, + "learning_rate": 1.534564979978158e-06, + "loss": 0.7373, + "step": 26887 + }, + { + "epoch": 4.89, + "learning_rate": 1.5345103749544958e-06, + "loss": 0.7855, + "step": 26888 + }, + { + "epoch": 4.89, + "learning_rate": 1.5344557699308338e-06, + "loss": 0.293, + "step": 26889 + }, + { + "epoch": 4.89, + "learning_rate": 1.5344011649071717e-06, + "loss": 0.927, + "step": 26890 + }, + { + "epoch": 4.89, + "learning_rate": 1.5343465598835092e-06, + "loss": 0.2241, + "step": 26891 + }, + { + "epoch": 4.89, + "learning_rate": 1.5342919548598472e-06, + "loss": 0.9523, + "step": 26892 + }, + { + "epoch": 4.89, + "learning_rate": 1.5342373498361851e-06, + "loss": 0.1418, + "step": 26893 + }, + { + "epoch": 4.89, + "learning_rate": 1.5341827448125226e-06, + "loss": 0.7759, + "step": 26894 + }, + { + "epoch": 4.89, + "learning_rate": 1.5341281397888606e-06, + "loss": 0.2963, + "step": 26895 + }, + { + "epoch": 4.89, + "learning_rate": 1.5340735347651983e-06, + "loss": 1.1869, + "step": 26896 + }, + { + "epoch": 4.89, + "learning_rate": 1.5340189297415363e-06, + "loss": 0.0573, + "step": 26897 + }, + { + "epoch": 4.89, + "learning_rate": 1.5339643247178742e-06, + "loss": 0.1156, + "step": 26898 + }, + { + "epoch": 4.89, + "learning_rate": 1.5339097196942117e-06, + "loss": 0.1106, + "step": 26899 + }, + { + "epoch": 4.89, + "learning_rate": 1.5338551146705497e-06, + "loss": 0.1496, + "step": 26900 + }, + { + "epoch": 4.89, + "learning_rate": 1.5338005096468876e-06, + "loss": 3.0396, + "step": 26901 + }, + { + "epoch": 4.89, + "learning_rate": 1.5337459046232253e-06, + "loss": 0.8016, + "step": 26902 + }, + { + "epoch": 4.89, + "learning_rate": 1.5336912995995633e-06, + "loss": 0.1606, + "step": 26903 + }, + { + "epoch": 4.89, + "learning_rate": 1.5336366945759012e-06, + "loss": 0.2426, + "step": 26904 + }, + { + "epoch": 4.89, + "learning_rate": 1.5335820895522388e-06, + "loss": 0.3095, + "step": 26905 + }, + { + "epoch": 4.89, + "learning_rate": 1.5335274845285767e-06, + "loss": 4.945, + "step": 26906 + }, + { + "epoch": 4.89, + "learning_rate": 1.5334728795049144e-06, + "loss": 0.4523, + "step": 26907 + }, + { + "epoch": 4.89, + "learning_rate": 1.5334182744812524e-06, + "loss": 0.0419, + "step": 26908 + }, + { + "epoch": 4.89, + "learning_rate": 1.5333636694575903e-06, + "loss": 0.3382, + "step": 26909 + }, + { + "epoch": 4.89, + "learning_rate": 1.5333090644339278e-06, + "loss": 0.7788, + "step": 26910 + }, + { + "epoch": 4.89, + "learning_rate": 1.5332544594102658e-06, + "loss": 0.019, + "step": 26911 + }, + { + "epoch": 4.89, + "learning_rate": 1.5331998543866037e-06, + "loss": 0.944, + "step": 26912 + }, + { + "epoch": 4.89, + "learning_rate": 1.5331452493629412e-06, + "loss": 2.9762, + "step": 26913 + }, + { + "epoch": 4.89, + "learning_rate": 1.5330906443392792e-06, + "loss": 0.6515, + "step": 26914 + }, + { + "epoch": 4.89, + "learning_rate": 1.5330360393156171e-06, + "loss": 0.0099, + "step": 26915 + }, + { + "epoch": 4.89, + "learning_rate": 1.5329814342919549e-06, + "loss": 0.8346, + "step": 26916 + }, + { + "epoch": 4.89, + "learning_rate": 1.5329268292682928e-06, + "loss": 0.0344, + "step": 26917 + }, + { + "epoch": 4.9, + "learning_rate": 1.5328722242446307e-06, + "loss": 0.2628, + "step": 26918 + }, + { + "epoch": 4.9, + "learning_rate": 1.5328176192209683e-06, + "loss": 0.1325, + "step": 26919 + }, + { + "epoch": 4.9, + "learning_rate": 1.5327630141973062e-06, + "loss": 0.3463, + "step": 26920 + }, + { + "epoch": 4.9, + "learning_rate": 1.532708409173644e-06, + "loss": 0.6365, + "step": 26921 + }, + { + "epoch": 4.9, + "learning_rate": 1.5326538041499819e-06, + "loss": 1.3913, + "step": 26922 + }, + { + "epoch": 4.9, + "learning_rate": 1.5325991991263198e-06, + "loss": 0.4777, + "step": 26923 + }, + { + "epoch": 4.9, + "learning_rate": 1.5325445941026573e-06, + "loss": 1.3409, + "step": 26924 + }, + { + "epoch": 4.9, + "learning_rate": 1.5324899890789953e-06, + "loss": 0.0112, + "step": 26925 + }, + { + "epoch": 4.9, + "learning_rate": 1.5324353840553332e-06, + "loss": 0.6463, + "step": 26926 + }, + { + "epoch": 4.9, + "learning_rate": 1.5323807790316708e-06, + "loss": 0.1948, + "step": 26927 + }, + { + "epoch": 4.9, + "learning_rate": 1.5323261740080087e-06, + "loss": 1.2203, + "step": 26928 + }, + { + "epoch": 4.9, + "learning_rate": 1.5322715689843466e-06, + "loss": 1.4979, + "step": 26929 + }, + { + "epoch": 4.9, + "learning_rate": 1.5322169639606844e-06, + "loss": 0.4825, + "step": 26930 + }, + { + "epoch": 4.9, + "learning_rate": 1.5321623589370223e-06, + "loss": 0.8397, + "step": 26931 + }, + { + "epoch": 4.9, + "learning_rate": 1.5321077539133603e-06, + "loss": 0.0804, + "step": 26932 + }, + { + "epoch": 4.9, + "learning_rate": 1.5320531488896978e-06, + "loss": 0.0701, + "step": 26933 + }, + { + "epoch": 4.9, + "learning_rate": 1.5319985438660357e-06, + "loss": 0.2105, + "step": 26934 + }, + { + "epoch": 4.9, + "learning_rate": 1.5319439388423735e-06, + "loss": 0.1939, + "step": 26935 + }, + { + "epoch": 4.9, + "learning_rate": 1.5318893338187114e-06, + "loss": 0.0743, + "step": 26936 + }, + { + "epoch": 4.9, + "learning_rate": 1.5318347287950493e-06, + "loss": 0.2445, + "step": 26937 + }, + { + "epoch": 4.9, + "learning_rate": 1.5317801237713869e-06, + "loss": 0.0171, + "step": 26938 + }, + { + "epoch": 4.9, + "learning_rate": 1.5317255187477248e-06, + "loss": 0.1268, + "step": 26939 + }, + { + "epoch": 4.9, + "learning_rate": 1.5316709137240627e-06, + "loss": 1.6401, + "step": 26940 + }, + { + "epoch": 4.9, + "learning_rate": 1.5316163087004003e-06, + "loss": 2.641, + "step": 26941 + }, + { + "epoch": 4.9, + "learning_rate": 1.5315617036767382e-06, + "loss": 0.838, + "step": 26942 + }, + { + "epoch": 4.9, + "learning_rate": 1.5315070986530762e-06, + "loss": 0.0175, + "step": 26943 + }, + { + "epoch": 4.9, + "learning_rate": 1.5314524936294139e-06, + "loss": 0.9717, + "step": 26944 + }, + { + "epoch": 4.9, + "learning_rate": 1.5313978886057518e-06, + "loss": 1.3092, + "step": 26945 + }, + { + "epoch": 4.9, + "learning_rate": 1.5313432835820898e-06, + "loss": 0.0165, + "step": 26946 + }, + { + "epoch": 4.9, + "learning_rate": 1.5312886785584273e-06, + "loss": 1.1003, + "step": 26947 + }, + { + "epoch": 4.9, + "learning_rate": 1.5312340735347652e-06, + "loss": 0.0165, + "step": 26948 + }, + { + "epoch": 4.9, + "learning_rate": 1.531179468511103e-06, + "loss": 0.2869, + "step": 26949 + }, + { + "epoch": 4.9, + "learning_rate": 1.531124863487441e-06, + "loss": 0.3302, + "step": 26950 + }, + { + "epoch": 4.9, + "learning_rate": 1.5310702584637789e-06, + "loss": 0.0482, + "step": 26951 + }, + { + "epoch": 4.9, + "learning_rate": 1.5310156534401164e-06, + "loss": 0.0296, + "step": 26952 + }, + { + "epoch": 4.9, + "learning_rate": 1.5309610484164543e-06, + "loss": 0.2874, + "step": 26953 + }, + { + "epoch": 4.9, + "learning_rate": 1.5309064433927923e-06, + "loss": 0.1323, + "step": 26954 + }, + { + "epoch": 4.9, + "learning_rate": 1.53085183836913e-06, + "loss": 1.6245, + "step": 26955 + }, + { + "epoch": 4.9, + "learning_rate": 1.530797233345468e-06, + "loss": 0.2306, + "step": 26956 + }, + { + "epoch": 4.9, + "learning_rate": 1.5307426283218059e-06, + "loss": 1.0701, + "step": 26957 + }, + { + "epoch": 4.9, + "learning_rate": 1.5306880232981434e-06, + "loss": 0.0086, + "step": 26958 + }, + { + "epoch": 4.9, + "learning_rate": 1.5306334182744813e-06, + "loss": 0.1864, + "step": 26959 + }, + { + "epoch": 4.9, + "learning_rate": 1.5305788132508193e-06, + "loss": 1.2997, + "step": 26960 + }, + { + "epoch": 4.9, + "learning_rate": 1.5305242082271568e-06, + "loss": 0.2801, + "step": 26961 + }, + { + "epoch": 4.9, + "learning_rate": 1.5304696032034948e-06, + "loss": 0.0154, + "step": 26962 + }, + { + "epoch": 4.9, + "learning_rate": 1.5304149981798325e-06, + "loss": 1.6195, + "step": 26963 + }, + { + "epoch": 4.9, + "learning_rate": 1.5303603931561704e-06, + "loss": 0.6437, + "step": 26964 + }, + { + "epoch": 4.9, + "learning_rate": 1.5303057881325084e-06, + "loss": 0.4602, + "step": 26965 + }, + { + "epoch": 4.9, + "learning_rate": 1.5302511831088459e-06, + "loss": 0.5975, + "step": 26966 + }, + { + "epoch": 4.9, + "learning_rate": 1.5301965780851838e-06, + "loss": 0.6839, + "step": 26967 + }, + { + "epoch": 4.9, + "learning_rate": 1.5301419730615218e-06, + "loss": 1.7876, + "step": 26968 + }, + { + "epoch": 4.9, + "learning_rate": 1.5300873680378595e-06, + "loss": 1.6932, + "step": 26969 + }, + { + "epoch": 4.9, + "learning_rate": 1.5300327630141975e-06, + "loss": 0.7303, + "step": 26970 + }, + { + "epoch": 4.9, + "learning_rate": 1.5299781579905354e-06, + "loss": 1.4354, + "step": 26971 + }, + { + "epoch": 4.9, + "learning_rate": 1.529923552966873e-06, + "loss": 0.2189, + "step": 26972 + }, + { + "epoch": 4.91, + "learning_rate": 1.5298689479432109e-06, + "loss": 0.0136, + "step": 26973 + }, + { + "epoch": 4.91, + "learning_rate": 1.5298143429195488e-06, + "loss": 1.0211, + "step": 26974 + }, + { + "epoch": 4.91, + "learning_rate": 1.5297597378958863e-06, + "loss": 0.0848, + "step": 26975 + }, + { + "epoch": 4.91, + "learning_rate": 1.5297051328722243e-06, + "loss": 1.612, + "step": 26976 + }, + { + "epoch": 4.91, + "learning_rate": 1.529650527848562e-06, + "loss": 0.0075, + "step": 26977 + }, + { + "epoch": 4.91, + "learning_rate": 1.5295959228249e-06, + "loss": 0.1296, + "step": 26978 + }, + { + "epoch": 4.91, + "learning_rate": 1.5295413178012379e-06, + "loss": 0.7599, + "step": 26979 + }, + { + "epoch": 4.91, + "learning_rate": 1.5294867127775754e-06, + "loss": 0.3599, + "step": 26980 + }, + { + "epoch": 4.91, + "learning_rate": 1.5294321077539133e-06, + "loss": 1.2688, + "step": 26981 + }, + { + "epoch": 4.91, + "learning_rate": 1.5293775027302513e-06, + "loss": 0.4778, + "step": 26982 + }, + { + "epoch": 4.91, + "learning_rate": 1.529322897706589e-06, + "loss": 0.4597, + "step": 26983 + }, + { + "epoch": 4.91, + "learning_rate": 1.529268292682927e-06, + "loss": 0.5308, + "step": 26984 + }, + { + "epoch": 4.91, + "learning_rate": 1.529213687659265e-06, + "loss": 0.5008, + "step": 26985 + }, + { + "epoch": 4.91, + "learning_rate": 1.5291590826356024e-06, + "loss": 0.0703, + "step": 26986 + }, + { + "epoch": 4.91, + "learning_rate": 1.5291044776119404e-06, + "loss": 0.3163, + "step": 26987 + }, + { + "epoch": 4.91, + "learning_rate": 1.529049872588278e-06, + "loss": 0.1263, + "step": 26988 + }, + { + "epoch": 4.91, + "learning_rate": 1.5289952675646158e-06, + "loss": 0.2632, + "step": 26989 + }, + { + "epoch": 4.91, + "learning_rate": 1.5289406625409538e-06, + "loss": 0.2792, + "step": 26990 + }, + { + "epoch": 4.91, + "learning_rate": 1.5288860575172915e-06, + "loss": 1.1749, + "step": 26991 + }, + { + "epoch": 4.91, + "learning_rate": 1.5288314524936295e-06, + "loss": 0.0115, + "step": 26992 + }, + { + "epoch": 4.91, + "learning_rate": 1.5287768474699674e-06, + "loss": 1.2604, + "step": 26993 + }, + { + "epoch": 4.91, + "learning_rate": 1.528722242446305e-06, + "loss": 0.0135, + "step": 26994 + }, + { + "epoch": 4.91, + "learning_rate": 1.5286676374226429e-06, + "loss": 0.7632, + "step": 26995 + }, + { + "epoch": 4.91, + "learning_rate": 1.5286130323989808e-06, + "loss": 0.6278, + "step": 26996 + }, + { + "epoch": 4.91, + "learning_rate": 1.5285584273753185e-06, + "loss": 0.6104, + "step": 26997 + }, + { + "epoch": 4.91, + "learning_rate": 1.5285038223516565e-06, + "loss": 0.2794, + "step": 26998 + }, + { + "epoch": 4.91, + "learning_rate": 1.5284492173279944e-06, + "loss": 0.0754, + "step": 26999 + }, + { + "epoch": 4.91, + "learning_rate": 1.528394612304332e-06, + "loss": 0.9054, + "step": 27000 + }, + { + "epoch": 4.91, + "learning_rate": 1.5283400072806699e-06, + "loss": 2.3548, + "step": 27001 + }, + { + "epoch": 4.91, + "learning_rate": 1.5282854022570076e-06, + "loss": 0.8442, + "step": 27002 + }, + { + "epoch": 4.91, + "learning_rate": 1.5282307972333456e-06, + "loss": 0.7578, + "step": 27003 + }, + { + "epoch": 4.91, + "learning_rate": 1.5281761922096835e-06, + "loss": 0.9916, + "step": 27004 + }, + { + "epoch": 4.91, + "learning_rate": 1.528121587186021e-06, + "loss": 0.0351, + "step": 27005 + }, + { + "epoch": 4.91, + "learning_rate": 1.528066982162359e-06, + "loss": 1.4956, + "step": 27006 + }, + { + "epoch": 4.91, + "learning_rate": 1.528012377138697e-06, + "loss": 0.4885, + "step": 27007 + }, + { + "epoch": 4.91, + "learning_rate": 1.5279577721150344e-06, + "loss": 1.1109, + "step": 27008 + }, + { + "epoch": 4.91, + "learning_rate": 1.5279031670913724e-06, + "loss": 0.0323, + "step": 27009 + }, + { + "epoch": 4.91, + "learning_rate": 1.5278485620677103e-06, + "loss": 0.1373, + "step": 27010 + }, + { + "epoch": 4.91, + "learning_rate": 1.527793957044048e-06, + "loss": 0.4628, + "step": 27011 + }, + { + "epoch": 4.91, + "learning_rate": 1.527739352020386e-06, + "loss": 0.254, + "step": 27012 + }, + { + "epoch": 4.91, + "learning_rate": 1.527684746996724e-06, + "loss": 0.179, + "step": 27013 + }, + { + "epoch": 4.91, + "learning_rate": 1.5276301419730615e-06, + "loss": 1.6706, + "step": 27014 + }, + { + "epoch": 4.91, + "learning_rate": 1.5275755369493994e-06, + "loss": 1.2922, + "step": 27015 + }, + { + "epoch": 4.91, + "learning_rate": 1.5275209319257371e-06, + "loss": 0.163, + "step": 27016 + }, + { + "epoch": 4.91, + "learning_rate": 1.527466326902075e-06, + "loss": 0.0291, + "step": 27017 + }, + { + "epoch": 4.91, + "learning_rate": 1.527411721878413e-06, + "loss": 0.5449, + "step": 27018 + }, + { + "epoch": 4.91, + "learning_rate": 1.5273571168547505e-06, + "loss": 0.4081, + "step": 27019 + }, + { + "epoch": 4.91, + "learning_rate": 1.5273025118310885e-06, + "loss": 0.837, + "step": 27020 + }, + { + "epoch": 4.91, + "learning_rate": 1.5272479068074264e-06, + "loss": 1.4818, + "step": 27021 + }, + { + "epoch": 4.91, + "learning_rate": 1.527193301783764e-06, + "loss": 1.8066, + "step": 27022 + }, + { + "epoch": 4.91, + "learning_rate": 1.5271386967601019e-06, + "loss": 0.3679, + "step": 27023 + }, + { + "epoch": 4.91, + "learning_rate": 1.5270840917364398e-06, + "loss": 0.4301, + "step": 27024 + }, + { + "epoch": 4.91, + "learning_rate": 1.5270294867127776e-06, + "loss": 0.6419, + "step": 27025 + }, + { + "epoch": 4.91, + "learning_rate": 1.5269748816891155e-06, + "loss": 1.8407, + "step": 27026 + }, + { + "epoch": 4.91, + "learning_rate": 1.5269202766654534e-06, + "loss": 0.063, + "step": 27027 + }, + { + "epoch": 4.92, + "learning_rate": 1.526865671641791e-06, + "loss": 0.2743, + "step": 27028 + }, + { + "epoch": 4.92, + "learning_rate": 1.526811066618129e-06, + "loss": 1.7286, + "step": 27029 + }, + { + "epoch": 4.92, + "learning_rate": 1.5267564615944666e-06, + "loss": 0.3905, + "step": 27030 + }, + { + "epoch": 4.92, + "learning_rate": 1.5267018565708046e-06, + "loss": 1.8943, + "step": 27031 + }, + { + "epoch": 4.92, + "learning_rate": 1.5266472515471425e-06, + "loss": 1.4444, + "step": 27032 + }, + { + "epoch": 4.92, + "learning_rate": 1.52659264652348e-06, + "loss": 0.0229, + "step": 27033 + }, + { + "epoch": 4.92, + "learning_rate": 1.526538041499818e-06, + "loss": 2.2259, + "step": 27034 + }, + { + "epoch": 4.92, + "learning_rate": 1.526483436476156e-06, + "loss": 0.6351, + "step": 27035 + }, + { + "epoch": 4.92, + "learning_rate": 1.5264288314524937e-06, + "loss": 0.441, + "step": 27036 + }, + { + "epoch": 4.92, + "learning_rate": 1.5263742264288316e-06, + "loss": 0.0615, + "step": 27037 + }, + { + "epoch": 4.92, + "learning_rate": 1.5263196214051696e-06, + "loss": 0.8723, + "step": 27038 + }, + { + "epoch": 4.92, + "learning_rate": 1.526265016381507e-06, + "loss": 0.0101, + "step": 27039 + }, + { + "epoch": 4.92, + "learning_rate": 1.526210411357845e-06, + "loss": 0.0109, + "step": 27040 + }, + { + "epoch": 4.92, + "learning_rate": 1.526155806334183e-06, + "loss": 0.0174, + "step": 27041 + }, + { + "epoch": 4.92, + "learning_rate": 1.5261012013105205e-06, + "loss": 0.2679, + "step": 27042 + }, + { + "epoch": 4.92, + "learning_rate": 1.5260465962868584e-06, + "loss": 1.7288, + "step": 27043 + }, + { + "epoch": 4.92, + "learning_rate": 1.5259919912631962e-06, + "loss": 0.3371, + "step": 27044 + }, + { + "epoch": 4.92, + "learning_rate": 1.525937386239534e-06, + "loss": 0.1751, + "step": 27045 + }, + { + "epoch": 4.92, + "learning_rate": 1.525882781215872e-06, + "loss": 0.8097, + "step": 27046 + }, + { + "epoch": 4.92, + "learning_rate": 1.5258281761922096e-06, + "loss": 1.1769, + "step": 27047 + }, + { + "epoch": 4.92, + "learning_rate": 1.5257735711685475e-06, + "loss": 0.82, + "step": 27048 + }, + { + "epoch": 4.92, + "learning_rate": 1.5257189661448854e-06, + "loss": 0.0791, + "step": 27049 + }, + { + "epoch": 4.92, + "learning_rate": 1.5256643611212232e-06, + "loss": 0.7488, + "step": 27050 + }, + { + "epoch": 4.92, + "learning_rate": 1.5256097560975611e-06, + "loss": 0.2435, + "step": 27051 + }, + { + "epoch": 4.92, + "learning_rate": 1.525555151073899e-06, + "loss": 0.254, + "step": 27052 + }, + { + "epoch": 4.92, + "learning_rate": 1.5255005460502366e-06, + "loss": 0.0245, + "step": 27053 + }, + { + "epoch": 4.92, + "learning_rate": 1.5254459410265745e-06, + "loss": 0.8687, + "step": 27054 + }, + { + "epoch": 4.92, + "learning_rate": 1.525391336002912e-06, + "loss": 0.6396, + "step": 27055 + }, + { + "epoch": 4.92, + "learning_rate": 1.52533673097925e-06, + "loss": 0.3605, + "step": 27056 + }, + { + "epoch": 4.92, + "learning_rate": 1.525282125955588e-06, + "loss": 0.6314, + "step": 27057 + }, + { + "epoch": 4.92, + "learning_rate": 1.5252275209319257e-06, + "loss": 0.6168, + "step": 27058 + }, + { + "epoch": 4.92, + "learning_rate": 1.5251729159082636e-06, + "loss": 0.8895, + "step": 27059 + }, + { + "epoch": 4.92, + "learning_rate": 1.5251183108846016e-06, + "loss": 1.087, + "step": 27060 + }, + { + "epoch": 4.92, + "learning_rate": 1.525063705860939e-06, + "loss": 0.1276, + "step": 27061 + }, + { + "epoch": 4.92, + "learning_rate": 1.525009100837277e-06, + "loss": 0.7833, + "step": 27062 + }, + { + "epoch": 4.92, + "learning_rate": 1.524954495813615e-06, + "loss": 1.2845, + "step": 27063 + }, + { + "epoch": 4.92, + "learning_rate": 1.5248998907899527e-06, + "loss": 1.1166, + "step": 27064 + }, + { + "epoch": 4.92, + "learning_rate": 1.5248452857662906e-06, + "loss": 0.4943, + "step": 27065 + }, + { + "epoch": 4.92, + "learning_rate": 1.5247906807426286e-06, + "loss": 3.1249, + "step": 27066 + }, + { + "epoch": 4.92, + "learning_rate": 1.524736075718966e-06, + "loss": 0.6367, + "step": 27067 + }, + { + "epoch": 4.92, + "learning_rate": 1.524681470695304e-06, + "loss": 0.7969, + "step": 27068 + }, + { + "epoch": 4.92, + "learning_rate": 1.5246268656716416e-06, + "loss": 2.8966, + "step": 27069 + }, + { + "epoch": 4.92, + "learning_rate": 1.5245722606479795e-06, + "loss": 0.1467, + "step": 27070 + }, + { + "epoch": 4.92, + "learning_rate": 1.5245176556243175e-06, + "loss": 0.0184, + "step": 27071 + }, + { + "epoch": 4.92, + "learning_rate": 1.5244630506006552e-06, + "loss": 0.4699, + "step": 27072 + }, + { + "epoch": 4.92, + "learning_rate": 1.5244084455769931e-06, + "loss": 0.0312, + "step": 27073 + }, + { + "epoch": 4.92, + "learning_rate": 1.524353840553331e-06, + "loss": 1.678, + "step": 27074 + }, + { + "epoch": 4.92, + "learning_rate": 1.5242992355296686e-06, + "loss": 1.0624, + "step": 27075 + }, + { + "epoch": 4.92, + "learning_rate": 1.5242446305060065e-06, + "loss": 1.6309, + "step": 27076 + }, + { + "epoch": 4.92, + "learning_rate": 1.5241900254823445e-06, + "loss": 2.3116, + "step": 27077 + }, + { + "epoch": 4.92, + "learning_rate": 1.5241354204586822e-06, + "loss": 0.1299, + "step": 27078 + }, + { + "epoch": 4.92, + "learning_rate": 1.5240808154350202e-06, + "loss": 0.0338, + "step": 27079 + }, + { + "epoch": 4.92, + "learning_rate": 1.524026210411358e-06, + "loss": 0.0372, + "step": 27080 + }, + { + "epoch": 4.92, + "learning_rate": 1.5239716053876956e-06, + "loss": 4.227, + "step": 27081 + }, + { + "epoch": 4.92, + "learning_rate": 1.5239170003640336e-06, + "loss": 1.1867, + "step": 27082 + }, + { + "epoch": 4.93, + "learning_rate": 1.5238623953403713e-06, + "loss": 0.2588, + "step": 27083 + }, + { + "epoch": 4.93, + "learning_rate": 1.5238077903167092e-06, + "loss": 0.8091, + "step": 27084 + }, + { + "epoch": 4.93, + "learning_rate": 1.5237531852930472e-06, + "loss": 0.0084, + "step": 27085 + }, + { + "epoch": 4.93, + "learning_rate": 1.5236985802693847e-06, + "loss": 1.2104, + "step": 27086 + }, + { + "epoch": 4.93, + "learning_rate": 1.5236439752457226e-06, + "loss": 1.2376, + "step": 27087 + }, + { + "epoch": 4.93, + "learning_rate": 1.5235893702220606e-06, + "loss": 0.216, + "step": 27088 + }, + { + "epoch": 4.93, + "learning_rate": 1.523534765198398e-06, + "loss": 1.7956, + "step": 27089 + }, + { + "epoch": 4.93, + "learning_rate": 1.523480160174736e-06, + "loss": 0.1727, + "step": 27090 + }, + { + "epoch": 4.93, + "learning_rate": 1.523425555151074e-06, + "loss": 0.0161, + "step": 27091 + }, + { + "epoch": 4.93, + "learning_rate": 1.5233709501274117e-06, + "loss": 0.2691, + "step": 27092 + }, + { + "epoch": 4.93, + "learning_rate": 1.5233163451037497e-06, + "loss": 0.8363, + "step": 27093 + }, + { + "epoch": 4.93, + "learning_rate": 1.5232617400800876e-06, + "loss": 0.5674, + "step": 27094 + }, + { + "epoch": 4.93, + "learning_rate": 1.5232071350564251e-06, + "loss": 0.0348, + "step": 27095 + }, + { + "epoch": 4.93, + "learning_rate": 1.523152530032763e-06, + "loss": 0.2448, + "step": 27096 + }, + { + "epoch": 4.93, + "learning_rate": 1.5230979250091008e-06, + "loss": 0.0078, + "step": 27097 + }, + { + "epoch": 4.93, + "learning_rate": 1.5230433199854387e-06, + "loss": 1.0494, + "step": 27098 + }, + { + "epoch": 4.93, + "learning_rate": 1.5229887149617767e-06, + "loss": 0.3126, + "step": 27099 + }, + { + "epoch": 4.93, + "learning_rate": 1.5229341099381142e-06, + "loss": 0.2743, + "step": 27100 + }, + { + "epoch": 4.93, + "learning_rate": 1.5228795049144522e-06, + "loss": 1.2119, + "step": 27101 + }, + { + "epoch": 4.93, + "learning_rate": 1.52282489989079e-06, + "loss": 0.315, + "step": 27102 + }, + { + "epoch": 4.93, + "learning_rate": 1.5227702948671276e-06, + "loss": 1.1566, + "step": 27103 + }, + { + "epoch": 4.93, + "learning_rate": 1.5227156898434656e-06, + "loss": 0.0667, + "step": 27104 + }, + { + "epoch": 4.93, + "learning_rate": 1.5226610848198035e-06, + "loss": 0.4263, + "step": 27105 + }, + { + "epoch": 4.93, + "learning_rate": 1.5226064797961412e-06, + "loss": 0.2934, + "step": 27106 + }, + { + "epoch": 4.93, + "learning_rate": 1.5225518747724792e-06, + "loss": 0.012, + "step": 27107 + }, + { + "epoch": 4.93, + "learning_rate": 1.5224972697488171e-06, + "loss": 1.2253, + "step": 27108 + }, + { + "epoch": 4.93, + "learning_rate": 1.5224426647251546e-06, + "loss": 0.4508, + "step": 27109 + }, + { + "epoch": 4.93, + "learning_rate": 1.5223880597014926e-06, + "loss": 0.4943, + "step": 27110 + }, + { + "epoch": 4.93, + "learning_rate": 1.5223334546778303e-06, + "loss": 2.2919, + "step": 27111 + }, + { + "epoch": 4.93, + "learning_rate": 1.5222788496541683e-06, + "loss": 1.2648, + "step": 27112 + }, + { + "epoch": 4.93, + "learning_rate": 1.5222242446305062e-06, + "loss": 0.0207, + "step": 27113 + }, + { + "epoch": 4.93, + "learning_rate": 1.5221696396068437e-06, + "loss": 0.0096, + "step": 27114 + }, + { + "epoch": 4.93, + "learning_rate": 1.5221150345831817e-06, + "loss": 0.2833, + "step": 27115 + }, + { + "epoch": 4.93, + "learning_rate": 1.5220604295595196e-06, + "loss": 1.0142, + "step": 27116 + }, + { + "epoch": 4.93, + "learning_rate": 1.5220058245358573e-06, + "loss": 1.2923, + "step": 27117 + }, + { + "epoch": 4.93, + "learning_rate": 1.5219512195121953e-06, + "loss": 0.5075, + "step": 27118 + }, + { + "epoch": 4.93, + "learning_rate": 1.521896614488533e-06, + "loss": 1.1418, + "step": 27119 + }, + { + "epoch": 4.93, + "learning_rate": 1.5218420094648707e-06, + "loss": 0.1697, + "step": 27120 + }, + { + "epoch": 4.93, + "learning_rate": 1.5217874044412087e-06, + "loss": 0.2333, + "step": 27121 + }, + { + "epoch": 4.93, + "learning_rate": 1.5217327994175466e-06, + "loss": 1.531, + "step": 27122 + }, + { + "epoch": 4.93, + "learning_rate": 1.5216781943938842e-06, + "loss": 0.0103, + "step": 27123 + }, + { + "epoch": 4.93, + "learning_rate": 1.521623589370222e-06, + "loss": 2.6111, + "step": 27124 + }, + { + "epoch": 4.93, + "learning_rate": 1.5215689843465598e-06, + "loss": 0.1867, + "step": 27125 + }, + { + "epoch": 4.93, + "learning_rate": 1.5215143793228978e-06, + "loss": 0.4538, + "step": 27126 + }, + { + "epoch": 4.93, + "learning_rate": 1.5214597742992357e-06, + "loss": 0.3324, + "step": 27127 + }, + { + "epoch": 4.93, + "learning_rate": 1.5214051692755732e-06, + "loss": 0.3221, + "step": 27128 + }, + { + "epoch": 4.93, + "learning_rate": 1.5213505642519112e-06, + "loss": 2.1844, + "step": 27129 + }, + { + "epoch": 4.93, + "learning_rate": 1.5212959592282491e-06, + "loss": 0.5748, + "step": 27130 + }, + { + "epoch": 4.93, + "learning_rate": 1.5212413542045869e-06, + "loss": 0.2152, + "step": 27131 + }, + { + "epoch": 4.93, + "learning_rate": 1.5211867491809248e-06, + "loss": 1.2202, + "step": 27132 + }, + { + "epoch": 4.93, + "learning_rate": 1.5211321441572627e-06, + "loss": 1.5268, + "step": 27133 + }, + { + "epoch": 4.93, + "learning_rate": 1.5210775391336003e-06, + "loss": 0.6708, + "step": 27134 + }, + { + "epoch": 4.93, + "learning_rate": 1.5210229341099382e-06, + "loss": 0.1235, + "step": 27135 + }, + { + "epoch": 4.93, + "learning_rate": 1.5209683290862757e-06, + "loss": 2.0976, + "step": 27136 + }, + { + "epoch": 4.93, + "learning_rate": 1.5209137240626137e-06, + "loss": 1.084, + "step": 27137 + }, + { + "epoch": 4.94, + "learning_rate": 1.5208591190389516e-06, + "loss": 2.264, + "step": 27138 + }, + { + "epoch": 4.94, + "learning_rate": 1.5208045140152893e-06, + "loss": 0.1354, + "step": 27139 + }, + { + "epoch": 4.94, + "learning_rate": 1.5207499089916273e-06, + "loss": 0.5953, + "step": 27140 + }, + { + "epoch": 4.94, + "learning_rate": 1.5206953039679652e-06, + "loss": 0.501, + "step": 27141 + }, + { + "epoch": 4.94, + "learning_rate": 1.5206406989443028e-06, + "loss": 1.1943, + "step": 27142 + }, + { + "epoch": 4.94, + "learning_rate": 1.5205860939206407e-06, + "loss": 0.4985, + "step": 27143 + }, + { + "epoch": 4.94, + "learning_rate": 1.5205314888969786e-06, + "loss": 0.01, + "step": 27144 + }, + { + "epoch": 4.94, + "learning_rate": 1.5204768838733164e-06, + "loss": 0.6422, + "step": 27145 + }, + { + "epoch": 4.94, + "learning_rate": 1.5204222788496543e-06, + "loss": 0.3764, + "step": 27146 + }, + { + "epoch": 4.94, + "learning_rate": 1.5203676738259923e-06, + "loss": 0.7977, + "step": 27147 + }, + { + "epoch": 4.94, + "learning_rate": 1.5203130688023298e-06, + "loss": 1.1966, + "step": 27148 + }, + { + "epoch": 4.94, + "learning_rate": 1.5202584637786677e-06, + "loss": 4.6696, + "step": 27149 + }, + { + "epoch": 4.94, + "learning_rate": 1.5202038587550052e-06, + "loss": 0.3533, + "step": 27150 + }, + { + "epoch": 4.94, + "learning_rate": 1.5201492537313432e-06, + "loss": 0.2414, + "step": 27151 + }, + { + "epoch": 4.94, + "learning_rate": 1.5200946487076811e-06, + "loss": 1.1363, + "step": 27152 + }, + { + "epoch": 4.94, + "learning_rate": 1.5200400436840189e-06, + "loss": 3.017, + "step": 27153 + }, + { + "epoch": 4.94, + "learning_rate": 1.5199854386603568e-06, + "loss": 0.1893, + "step": 27154 + }, + { + "epoch": 4.94, + "learning_rate": 1.5199308336366947e-06, + "loss": 1.426, + "step": 27155 + }, + { + "epoch": 4.94, + "learning_rate": 1.5198762286130323e-06, + "loss": 0.5294, + "step": 27156 + }, + { + "epoch": 4.94, + "learning_rate": 1.5198216235893702e-06, + "loss": 0.708, + "step": 27157 + }, + { + "epoch": 4.94, + "learning_rate": 1.5197670185657082e-06, + "loss": 0.7106, + "step": 27158 + }, + { + "epoch": 4.94, + "learning_rate": 1.5197124135420459e-06, + "loss": 0.0053, + "step": 27159 + }, + { + "epoch": 4.94, + "learning_rate": 1.5196578085183838e-06, + "loss": 0.4199, + "step": 27160 + }, + { + "epoch": 4.94, + "learning_rate": 1.5196032034947218e-06, + "loss": 2.3709, + "step": 27161 + }, + { + "epoch": 4.94, + "learning_rate": 1.5195485984710593e-06, + "loss": 0.2726, + "step": 27162 + }, + { + "epoch": 4.94, + "learning_rate": 1.5194939934473972e-06, + "loss": 0.318, + "step": 27163 + }, + { + "epoch": 4.94, + "learning_rate": 1.519439388423735e-06, + "loss": 1.3565, + "step": 27164 + }, + { + "epoch": 4.94, + "learning_rate": 1.519384783400073e-06, + "loss": 0.0106, + "step": 27165 + }, + { + "epoch": 4.94, + "learning_rate": 1.5193301783764109e-06, + "loss": 2.274, + "step": 27166 + }, + { + "epoch": 4.94, + "learning_rate": 1.5192755733527484e-06, + "loss": 0.0128, + "step": 27167 + }, + { + "epoch": 4.94, + "learning_rate": 1.5192209683290863e-06, + "loss": 0.7508, + "step": 27168 + }, + { + "epoch": 4.94, + "learning_rate": 1.5191663633054243e-06, + "loss": 1.1092, + "step": 27169 + }, + { + "epoch": 4.94, + "learning_rate": 1.5191117582817618e-06, + "loss": 0.6376, + "step": 27170 + }, + { + "epoch": 4.94, + "learning_rate": 1.5190571532580997e-06, + "loss": 0.2663, + "step": 27171 + }, + { + "epoch": 4.94, + "learning_rate": 1.5190025482344377e-06, + "loss": 0.0079, + "step": 27172 + }, + { + "epoch": 4.94, + "learning_rate": 1.5189479432107754e-06, + "loss": 1.215, + "step": 27173 + }, + { + "epoch": 4.94, + "learning_rate": 1.5188933381871133e-06, + "loss": 3.0763, + "step": 27174 + }, + { + "epoch": 4.94, + "learning_rate": 1.5188387331634513e-06, + "loss": 2.0104, + "step": 27175 + }, + { + "epoch": 4.94, + "learning_rate": 1.5187841281397888e-06, + "loss": 0.0084, + "step": 27176 + }, + { + "epoch": 4.94, + "learning_rate": 1.5187295231161267e-06, + "loss": 0.1721, + "step": 27177 + }, + { + "epoch": 4.94, + "learning_rate": 1.5186749180924645e-06, + "loss": 1.0174, + "step": 27178 + }, + { + "epoch": 4.94, + "learning_rate": 1.5186203130688024e-06, + "loss": 0.7709, + "step": 27179 + }, + { + "epoch": 4.94, + "learning_rate": 1.5185657080451404e-06, + "loss": 1.0011, + "step": 27180 + }, + { + "epoch": 4.94, + "learning_rate": 1.5185111030214779e-06, + "loss": 0.0147, + "step": 27181 + }, + { + "epoch": 4.94, + "learning_rate": 1.5184564979978158e-06, + "loss": 0.8796, + "step": 27182 + }, + { + "epoch": 4.94, + "learning_rate": 1.5184018929741538e-06, + "loss": 0.2193, + "step": 27183 + }, + { + "epoch": 4.94, + "learning_rate": 1.5183472879504913e-06, + "loss": 0.2638, + "step": 27184 + }, + { + "epoch": 4.94, + "learning_rate": 1.5182926829268292e-06, + "loss": 0.138, + "step": 27185 + }, + { + "epoch": 4.94, + "learning_rate": 1.5182380779031672e-06, + "loss": 0.2396, + "step": 27186 + }, + { + "epoch": 4.94, + "learning_rate": 1.518183472879505e-06, + "loss": 1.2254, + "step": 27187 + }, + { + "epoch": 4.94, + "learning_rate": 1.5181288678558429e-06, + "loss": 0.0723, + "step": 27188 + }, + { + "epoch": 4.94, + "learning_rate": 1.5180742628321808e-06, + "loss": 0.682, + "step": 27189 + }, + { + "epoch": 4.94, + "learning_rate": 1.5180196578085183e-06, + "loss": 0.1157, + "step": 27190 + }, + { + "epoch": 4.94, + "learning_rate": 1.5179650527848563e-06, + "loss": 0.2731, + "step": 27191 + }, + { + "epoch": 4.94, + "learning_rate": 1.517910447761194e-06, + "loss": 0.0108, + "step": 27192 + }, + { + "epoch": 4.95, + "learning_rate": 1.517855842737532e-06, + "loss": 0.8517, + "step": 27193 + }, + { + "epoch": 4.95, + "learning_rate": 1.5178012377138699e-06, + "loss": 3.0353, + "step": 27194 + }, + { + "epoch": 4.95, + "learning_rate": 1.5177466326902074e-06, + "loss": 0.0118, + "step": 27195 + }, + { + "epoch": 4.95, + "learning_rate": 1.5176920276665453e-06, + "loss": 1.0022, + "step": 27196 + }, + { + "epoch": 4.95, + "learning_rate": 1.5176374226428833e-06, + "loss": 0.1164, + "step": 27197 + }, + { + "epoch": 4.95, + "learning_rate": 1.5175828176192208e-06, + "loss": 0.373, + "step": 27198 + }, + { + "epoch": 4.95, + "learning_rate": 1.5175282125955587e-06, + "loss": 0.1671, + "step": 27199 + }, + { + "epoch": 4.95, + "learning_rate": 1.5174736075718967e-06, + "loss": 0.1128, + "step": 27200 + }, + { + "epoch": 4.95, + "learning_rate": 1.5174190025482344e-06, + "loss": 0.0073, + "step": 27201 + }, + { + "epoch": 4.95, + "learning_rate": 1.5173643975245724e-06, + "loss": 1.2748, + "step": 27202 + }, + { + "epoch": 4.95, + "learning_rate": 1.5173097925009103e-06, + "loss": 0.0097, + "step": 27203 + }, + { + "epoch": 4.95, + "learning_rate": 1.5172551874772478e-06, + "loss": 0.531, + "step": 27204 + }, + { + "epoch": 4.95, + "learning_rate": 1.5172005824535858e-06, + "loss": 0.1401, + "step": 27205 + }, + { + "epoch": 4.95, + "learning_rate": 1.5171459774299235e-06, + "loss": 0.2207, + "step": 27206 + }, + { + "epoch": 4.95, + "learning_rate": 1.5170913724062614e-06, + "loss": 3.5741, + "step": 27207 + }, + { + "epoch": 4.95, + "learning_rate": 1.5170367673825994e-06, + "loss": 0.609, + "step": 27208 + }, + { + "epoch": 4.95, + "learning_rate": 1.516982162358937e-06, + "loss": 0.3359, + "step": 27209 + }, + { + "epoch": 4.95, + "learning_rate": 1.5169275573352749e-06, + "loss": 0.2963, + "step": 27210 + }, + { + "epoch": 4.95, + "learning_rate": 1.5168729523116128e-06, + "loss": 0.078, + "step": 27211 + }, + { + "epoch": 4.95, + "learning_rate": 1.5168183472879505e-06, + "loss": 0.8757, + "step": 27212 + }, + { + "epoch": 4.95, + "learning_rate": 1.5167637422642885e-06, + "loss": 0.6577, + "step": 27213 + }, + { + "epoch": 4.95, + "learning_rate": 1.5167091372406264e-06, + "loss": 1.1061, + "step": 27214 + }, + { + "epoch": 4.95, + "learning_rate": 1.516654532216964e-06, + "loss": 0.4305, + "step": 27215 + }, + { + "epoch": 4.95, + "learning_rate": 1.5165999271933019e-06, + "loss": 0.5171, + "step": 27216 + }, + { + "epoch": 4.95, + "learning_rate": 1.5165453221696394e-06, + "loss": 1.0037, + "step": 27217 + }, + { + "epoch": 4.95, + "learning_rate": 1.5164907171459773e-06, + "loss": 0.3033, + "step": 27218 + }, + { + "epoch": 4.95, + "learning_rate": 1.5164361121223153e-06, + "loss": 0.1524, + "step": 27219 + }, + { + "epoch": 4.95, + "learning_rate": 1.516381507098653e-06, + "loss": 0.9525, + "step": 27220 + }, + { + "epoch": 4.95, + "learning_rate": 1.516326902074991e-06, + "loss": 0.3968, + "step": 27221 + }, + { + "epoch": 4.95, + "learning_rate": 1.516272297051329e-06, + "loss": 0.1835, + "step": 27222 + }, + { + "epoch": 4.95, + "learning_rate": 1.5162176920276664e-06, + "loss": 0.2077, + "step": 27223 + }, + { + "epoch": 4.95, + "learning_rate": 1.5161630870040044e-06, + "loss": 0.1372, + "step": 27224 + }, + { + "epoch": 4.95, + "learning_rate": 1.5161084819803423e-06, + "loss": 0.3237, + "step": 27225 + }, + { + "epoch": 4.95, + "learning_rate": 1.51605387695668e-06, + "loss": 0.5435, + "step": 27226 + }, + { + "epoch": 4.95, + "learning_rate": 1.515999271933018e-06, + "loss": 0.646, + "step": 27227 + }, + { + "epoch": 4.95, + "learning_rate": 1.515944666909356e-06, + "loss": 1.1997, + "step": 27228 + }, + { + "epoch": 4.95, + "learning_rate": 1.5158900618856934e-06, + "loss": 0.5891, + "step": 27229 + }, + { + "epoch": 4.95, + "learning_rate": 1.5158354568620314e-06, + "loss": 0.583, + "step": 27230 + }, + { + "epoch": 4.95, + "learning_rate": 1.515780851838369e-06, + "loss": 0.0134, + "step": 27231 + }, + { + "epoch": 4.95, + "learning_rate": 1.5157262468147069e-06, + "loss": 0.0546, + "step": 27232 + }, + { + "epoch": 4.95, + "learning_rate": 1.5156716417910448e-06, + "loss": 1.0981, + "step": 27233 + }, + { + "epoch": 4.95, + "learning_rate": 1.5156170367673825e-06, + "loss": 0.0297, + "step": 27234 + }, + { + "epoch": 4.95, + "learning_rate": 1.5155624317437205e-06, + "loss": 2.0252, + "step": 27235 + }, + { + "epoch": 4.95, + "learning_rate": 1.5155078267200584e-06, + "loss": 0.0291, + "step": 27236 + }, + { + "epoch": 4.95, + "learning_rate": 1.515453221696396e-06, + "loss": 0.4203, + "step": 27237 + }, + { + "epoch": 4.95, + "learning_rate": 1.5153986166727339e-06, + "loss": 0.0074, + "step": 27238 + }, + { + "epoch": 4.95, + "learning_rate": 1.5153440116490718e-06, + "loss": 0.0197, + "step": 27239 + }, + { + "epoch": 4.95, + "learning_rate": 1.5152894066254096e-06, + "loss": 0.0699, + "step": 27240 + }, + { + "epoch": 4.95, + "learning_rate": 1.5152348016017475e-06, + "loss": 2.587, + "step": 27241 + }, + { + "epoch": 4.95, + "learning_rate": 1.5151801965780854e-06, + "loss": 0.0409, + "step": 27242 + }, + { + "epoch": 4.95, + "learning_rate": 1.515125591554423e-06, + "loss": 1.4682, + "step": 27243 + }, + { + "epoch": 4.95, + "learning_rate": 1.515070986530761e-06, + "loss": 0.7203, + "step": 27244 + }, + { + "epoch": 4.95, + "learning_rate": 1.5150163815070986e-06, + "loss": 0.0449, + "step": 27245 + }, + { + "epoch": 4.95, + "learning_rate": 1.5149617764834366e-06, + "loss": 0.5154, + "step": 27246 + }, + { + "epoch": 4.95, + "learning_rate": 1.5149071714597745e-06, + "loss": 0.2315, + "step": 27247 + }, + { + "epoch": 4.96, + "learning_rate": 1.514852566436112e-06, + "loss": 1.0895, + "step": 27248 + }, + { + "epoch": 4.96, + "learning_rate": 1.51479796141245e-06, + "loss": 1.104, + "step": 27249 + }, + { + "epoch": 4.96, + "learning_rate": 1.514743356388788e-06, + "loss": 0.7265, + "step": 27250 + }, + { + "epoch": 4.96, + "learning_rate": 1.5146887513651255e-06, + "loss": 0.6183, + "step": 27251 + }, + { + "epoch": 4.96, + "learning_rate": 1.5146341463414634e-06, + "loss": 1.1902, + "step": 27252 + }, + { + "epoch": 4.96, + "learning_rate": 1.5145795413178013e-06, + "loss": 0.2542, + "step": 27253 + }, + { + "epoch": 4.96, + "learning_rate": 1.514524936294139e-06, + "loss": 0.3271, + "step": 27254 + }, + { + "epoch": 4.96, + "learning_rate": 1.514470331270477e-06, + "loss": 0.1114, + "step": 27255 + }, + { + "epoch": 4.96, + "learning_rate": 1.514415726246815e-06, + "loss": 0.0101, + "step": 27256 + }, + { + "epoch": 4.96, + "learning_rate": 1.5143611212231525e-06, + "loss": 0.2074, + "step": 27257 + }, + { + "epoch": 4.96, + "learning_rate": 1.5143065161994904e-06, + "loss": 0.004, + "step": 27258 + }, + { + "epoch": 4.96, + "learning_rate": 1.5142519111758282e-06, + "loss": 0.0762, + "step": 27259 + }, + { + "epoch": 4.96, + "learning_rate": 1.514197306152166e-06, + "loss": 0.242, + "step": 27260 + }, + { + "epoch": 4.96, + "learning_rate": 1.514142701128504e-06, + "loss": 0.1647, + "step": 27261 + }, + { + "epoch": 4.96, + "learning_rate": 1.5140880961048416e-06, + "loss": 1.2023, + "step": 27262 + }, + { + "epoch": 4.96, + "learning_rate": 1.5140334910811795e-06, + "loss": 0.0947, + "step": 27263 + }, + { + "epoch": 4.96, + "learning_rate": 1.5139788860575174e-06, + "loss": 0.2616, + "step": 27264 + }, + { + "epoch": 4.96, + "learning_rate": 1.513924281033855e-06, + "loss": 0.1126, + "step": 27265 + }, + { + "epoch": 4.96, + "learning_rate": 1.513869676010193e-06, + "loss": 1.4746, + "step": 27266 + }, + { + "epoch": 4.96, + "learning_rate": 1.5138150709865309e-06, + "loss": 1.6287, + "step": 27267 + }, + { + "epoch": 4.96, + "learning_rate": 1.5137604659628686e-06, + "loss": 1.1207, + "step": 27268 + }, + { + "epoch": 4.96, + "learning_rate": 1.5137058609392065e-06, + "loss": 0.1255, + "step": 27269 + }, + { + "epoch": 4.96, + "learning_rate": 1.5136512559155445e-06, + "loss": 0.8848, + "step": 27270 + }, + { + "epoch": 4.96, + "learning_rate": 1.513596650891882e-06, + "loss": 0.5503, + "step": 27271 + }, + { + "epoch": 4.96, + "learning_rate": 1.51354204586822e-06, + "loss": 0.0299, + "step": 27272 + }, + { + "epoch": 4.96, + "learning_rate": 1.5134874408445577e-06, + "loss": 1.322, + "step": 27273 + }, + { + "epoch": 4.96, + "learning_rate": 1.5134328358208956e-06, + "loss": 1.9847, + "step": 27274 + }, + { + "epoch": 4.96, + "learning_rate": 1.5133782307972336e-06, + "loss": 2.0772, + "step": 27275 + }, + { + "epoch": 4.96, + "learning_rate": 1.513323625773571e-06, + "loss": 0.016, + "step": 27276 + }, + { + "epoch": 4.96, + "learning_rate": 1.513269020749909e-06, + "loss": 0.2157, + "step": 27277 + }, + { + "epoch": 4.96, + "learning_rate": 1.513214415726247e-06, + "loss": 0.7371, + "step": 27278 + }, + { + "epoch": 4.96, + "learning_rate": 1.5131598107025845e-06, + "loss": 2.5938, + "step": 27279 + }, + { + "epoch": 4.96, + "learning_rate": 1.5131052056789224e-06, + "loss": 0.7829, + "step": 27280 + }, + { + "epoch": 4.96, + "learning_rate": 1.5130506006552604e-06, + "loss": 0.2072, + "step": 27281 + }, + { + "epoch": 4.96, + "learning_rate": 1.512995995631598e-06, + "loss": 0.0556, + "step": 27282 + }, + { + "epoch": 4.96, + "learning_rate": 1.512941390607936e-06, + "loss": 2.5293, + "step": 27283 + }, + { + "epoch": 4.96, + "learning_rate": 1.512886785584274e-06, + "loss": 0.0074, + "step": 27284 + }, + { + "epoch": 4.96, + "learning_rate": 1.5128321805606115e-06, + "loss": 0.0276, + "step": 27285 + }, + { + "epoch": 4.96, + "learning_rate": 1.5127775755369494e-06, + "loss": 0.5794, + "step": 27286 + }, + { + "epoch": 4.96, + "learning_rate": 1.5127229705132872e-06, + "loss": 1.1346, + "step": 27287 + }, + { + "epoch": 4.96, + "learning_rate": 1.5126683654896251e-06, + "loss": 0.4983, + "step": 27288 + }, + { + "epoch": 4.96, + "learning_rate": 1.512613760465963e-06, + "loss": 0.0739, + "step": 27289 + }, + { + "epoch": 4.96, + "learning_rate": 1.5125591554423006e-06, + "loss": 0.4476, + "step": 27290 + }, + { + "epoch": 4.96, + "learning_rate": 1.5125045504186385e-06, + "loss": 1.757, + "step": 27291 + }, + { + "epoch": 4.96, + "learning_rate": 1.5124499453949765e-06, + "loss": 0.0278, + "step": 27292 + }, + { + "epoch": 4.96, + "learning_rate": 1.5123953403713142e-06, + "loss": 1.082, + "step": 27293 + }, + { + "epoch": 4.96, + "learning_rate": 1.5123407353476521e-06, + "loss": 0.1739, + "step": 27294 + }, + { + "epoch": 4.96, + "learning_rate": 1.51228613032399e-06, + "loss": 0.286, + "step": 27295 + }, + { + "epoch": 4.96, + "learning_rate": 1.5122315253003276e-06, + "loss": 0.0076, + "step": 27296 + }, + { + "epoch": 4.96, + "learning_rate": 1.5121769202766656e-06, + "loss": 1.7025, + "step": 27297 + }, + { + "epoch": 4.96, + "learning_rate": 1.512122315253003e-06, + "loss": 0.8017, + "step": 27298 + }, + { + "epoch": 4.96, + "learning_rate": 1.512067710229341e-06, + "loss": 0.4582, + "step": 27299 + }, + { + "epoch": 4.96, + "learning_rate": 1.512013105205679e-06, + "loss": 0.1413, + "step": 27300 + }, + { + "epoch": 4.96, + "learning_rate": 1.5119585001820167e-06, + "loss": 0.3547, + "step": 27301 + }, + { + "epoch": 4.96, + "learning_rate": 1.5119038951583546e-06, + "loss": 0.0091, + "step": 27302 + }, + { + "epoch": 4.97, + "learning_rate": 1.5118492901346926e-06, + "loss": 0.5143, + "step": 27303 + }, + { + "epoch": 4.97, + "learning_rate": 1.51179468511103e-06, + "loss": 0.8922, + "step": 27304 + }, + { + "epoch": 4.97, + "learning_rate": 1.511740080087368e-06, + "loss": 0.1364, + "step": 27305 + }, + { + "epoch": 4.97, + "learning_rate": 1.511685475063706e-06, + "loss": 0.4571, + "step": 27306 + }, + { + "epoch": 4.97, + "learning_rate": 1.5116308700400437e-06, + "loss": 0.1319, + "step": 27307 + }, + { + "epoch": 4.97, + "learning_rate": 1.5115762650163817e-06, + "loss": 0.7795, + "step": 27308 + }, + { + "epoch": 4.97, + "learning_rate": 1.5115216599927196e-06, + "loss": 0.3102, + "step": 27309 + }, + { + "epoch": 4.97, + "learning_rate": 1.5114670549690571e-06, + "loss": 1.66, + "step": 27310 + }, + { + "epoch": 4.97, + "learning_rate": 1.511412449945395e-06, + "loss": 1.1615, + "step": 27311 + }, + { + "epoch": 4.97, + "learning_rate": 1.5113578449217326e-06, + "loss": 2.7004, + "step": 27312 + }, + { + "epoch": 4.97, + "learning_rate": 1.5113032398980705e-06, + "loss": 0.5693, + "step": 27313 + }, + { + "epoch": 4.97, + "learning_rate": 1.5112486348744085e-06, + "loss": 1.4554, + "step": 27314 + }, + { + "epoch": 4.97, + "learning_rate": 1.5111940298507462e-06, + "loss": 0.1394, + "step": 27315 + }, + { + "epoch": 4.97, + "learning_rate": 1.5111394248270841e-06, + "loss": 1.8322, + "step": 27316 + }, + { + "epoch": 4.97, + "learning_rate": 1.511084819803422e-06, + "loss": 0.5239, + "step": 27317 + }, + { + "epoch": 4.97, + "learning_rate": 1.5110302147797596e-06, + "loss": 0.1782, + "step": 27318 + }, + { + "epoch": 4.97, + "learning_rate": 1.5109756097560976e-06, + "loss": 0.6654, + "step": 27319 + }, + { + "epoch": 4.97, + "learning_rate": 1.5109210047324355e-06, + "loss": 0.0221, + "step": 27320 + }, + { + "epoch": 4.97, + "learning_rate": 1.5108663997087732e-06, + "loss": 3.1739, + "step": 27321 + }, + { + "epoch": 4.97, + "learning_rate": 1.5108117946851112e-06, + "loss": 0.7985, + "step": 27322 + }, + { + "epoch": 4.97, + "learning_rate": 1.5107571896614491e-06, + "loss": 0.023, + "step": 27323 + }, + { + "epoch": 4.97, + "learning_rate": 1.5107025846377866e-06, + "loss": 0.062, + "step": 27324 + }, + { + "epoch": 4.97, + "learning_rate": 1.5106479796141246e-06, + "loss": 2.3814, + "step": 27325 + }, + { + "epoch": 4.97, + "learning_rate": 1.510593374590462e-06, + "loss": 0.7888, + "step": 27326 + }, + { + "epoch": 4.97, + "learning_rate": 1.5105387695668e-06, + "loss": 0.2282, + "step": 27327 + }, + { + "epoch": 4.97, + "learning_rate": 1.510484164543138e-06, + "loss": 0.2289, + "step": 27328 + }, + { + "epoch": 4.97, + "learning_rate": 1.5104295595194757e-06, + "loss": 0.5043, + "step": 27329 + }, + { + "epoch": 4.97, + "learning_rate": 1.5103749544958137e-06, + "loss": 1.1449, + "step": 27330 + }, + { + "epoch": 4.97, + "learning_rate": 1.5103203494721516e-06, + "loss": 0.2291, + "step": 27331 + }, + { + "epoch": 4.97, + "learning_rate": 1.5102657444484891e-06, + "loss": 0.7027, + "step": 27332 + }, + { + "epoch": 4.97, + "learning_rate": 1.510211139424827e-06, + "loss": 0.1004, + "step": 27333 + }, + { + "epoch": 4.97, + "learning_rate": 1.510156534401165e-06, + "loss": 0.7116, + "step": 27334 + }, + { + "epoch": 4.97, + "learning_rate": 1.5101019293775027e-06, + "loss": 0.9293, + "step": 27335 + }, + { + "epoch": 4.97, + "learning_rate": 1.5100473243538407e-06, + "loss": 0.4513, + "step": 27336 + }, + { + "epoch": 4.97, + "learning_rate": 1.5099927193301786e-06, + "loss": 1.8559, + "step": 27337 + }, + { + "epoch": 4.97, + "learning_rate": 1.5099381143065162e-06, + "loss": 0.0547, + "step": 27338 + }, + { + "epoch": 4.97, + "learning_rate": 1.509883509282854e-06, + "loss": 0.105, + "step": 27339 + }, + { + "epoch": 4.97, + "learning_rate": 1.5098289042591918e-06, + "loss": 1.4893, + "step": 27340 + }, + { + "epoch": 4.97, + "learning_rate": 1.5097742992355298e-06, + "loss": 0.5032, + "step": 27341 + }, + { + "epoch": 4.97, + "learning_rate": 1.5097196942118677e-06, + "loss": 1.3177, + "step": 27342 + }, + { + "epoch": 4.97, + "learning_rate": 1.5096650891882052e-06, + "loss": 0.3593, + "step": 27343 + }, + { + "epoch": 4.97, + "learning_rate": 1.5096104841645432e-06, + "loss": 0.0092, + "step": 27344 + }, + { + "epoch": 4.97, + "learning_rate": 1.5095558791408811e-06, + "loss": 0.2368, + "step": 27345 + }, + { + "epoch": 4.97, + "learning_rate": 1.5095012741172186e-06, + "loss": 0.0169, + "step": 27346 + }, + { + "epoch": 4.97, + "learning_rate": 1.5094466690935566e-06, + "loss": 1.5309, + "step": 27347 + }, + { + "epoch": 4.97, + "learning_rate": 1.5093920640698945e-06, + "loss": 0.0534, + "step": 27348 + }, + { + "epoch": 4.97, + "learning_rate": 1.5093374590462323e-06, + "loss": 1.3003, + "step": 27349 + }, + { + "epoch": 4.97, + "learning_rate": 1.5092828540225702e-06, + "loss": 1.4789, + "step": 27350 + }, + { + "epoch": 4.97, + "learning_rate": 1.5092282489989081e-06, + "loss": 0.0079, + "step": 27351 + }, + { + "epoch": 4.97, + "learning_rate": 1.5091736439752457e-06, + "loss": 0.3823, + "step": 27352 + }, + { + "epoch": 4.97, + "learning_rate": 1.5091190389515836e-06, + "loss": 1.6886, + "step": 27353 + }, + { + "epoch": 4.97, + "learning_rate": 1.5090644339279213e-06, + "loss": 0.6229, + "step": 27354 + }, + { + "epoch": 4.97, + "learning_rate": 1.5090098289042593e-06, + "loss": 1.2063, + "step": 27355 + }, + { + "epoch": 4.97, + "learning_rate": 1.5089552238805972e-06, + "loss": 0.7867, + "step": 27356 + }, + { + "epoch": 4.97, + "learning_rate": 1.5089006188569347e-06, + "loss": 0.2264, + "step": 27357 + }, + { + "epoch": 4.98, + "learning_rate": 1.5088460138332727e-06, + "loss": 0.7704, + "step": 27358 + }, + { + "epoch": 4.98, + "learning_rate": 1.5087914088096106e-06, + "loss": 0.4428, + "step": 27359 + }, + { + "epoch": 4.98, + "learning_rate": 1.5087368037859482e-06, + "loss": 0.6004, + "step": 27360 + }, + { + "epoch": 4.98, + "learning_rate": 1.508682198762286e-06, + "loss": 0.9559, + "step": 27361 + }, + { + "epoch": 4.98, + "learning_rate": 1.508627593738624e-06, + "loss": 2.2922, + "step": 27362 + }, + { + "epoch": 4.98, + "learning_rate": 1.5085729887149618e-06, + "loss": 1.9426, + "step": 27363 + }, + { + "epoch": 4.98, + "learning_rate": 1.5085183836912997e-06, + "loss": 0.0706, + "step": 27364 + }, + { + "epoch": 4.98, + "learning_rate": 1.5084637786676372e-06, + "loss": 1.015, + "step": 27365 + }, + { + "epoch": 4.98, + "learning_rate": 1.5084091736439752e-06, + "loss": 0.6709, + "step": 27366 + }, + { + "epoch": 4.98, + "learning_rate": 1.5083545686203131e-06, + "loss": 0.3596, + "step": 27367 + }, + { + "epoch": 4.98, + "learning_rate": 1.5082999635966509e-06, + "loss": 0.3075, + "step": 27368 + }, + { + "epoch": 4.98, + "learning_rate": 1.5082453585729888e-06, + "loss": 0.4422, + "step": 27369 + }, + { + "epoch": 4.98, + "learning_rate": 1.5081907535493267e-06, + "loss": 0.3508, + "step": 27370 + }, + { + "epoch": 4.98, + "learning_rate": 1.5081361485256643e-06, + "loss": 0.8657, + "step": 27371 + }, + { + "epoch": 4.98, + "learning_rate": 1.5080815435020022e-06, + "loss": 0.1417, + "step": 27372 + }, + { + "epoch": 4.98, + "learning_rate": 1.5080269384783401e-06, + "loss": 0.9665, + "step": 27373 + }, + { + "epoch": 4.98, + "learning_rate": 1.5079723334546779e-06, + "loss": 0.0576, + "step": 27374 + }, + { + "epoch": 4.98, + "learning_rate": 1.5079177284310158e-06, + "loss": 3.2677, + "step": 27375 + }, + { + "epoch": 4.98, + "learning_rate": 1.5078631234073538e-06, + "loss": 0.717, + "step": 27376 + }, + { + "epoch": 4.98, + "learning_rate": 1.5078085183836913e-06, + "loss": 0.2716, + "step": 27377 + }, + { + "epoch": 4.98, + "learning_rate": 1.5077539133600292e-06, + "loss": 0.1963, + "step": 27378 + }, + { + "epoch": 4.98, + "learning_rate": 1.5076993083363667e-06, + "loss": 0.3431, + "step": 27379 + }, + { + "epoch": 4.98, + "learning_rate": 1.5076447033127047e-06, + "loss": 0.4883, + "step": 27380 + }, + { + "epoch": 4.98, + "learning_rate": 1.5075900982890426e-06, + "loss": 0.0581, + "step": 27381 + }, + { + "epoch": 4.98, + "learning_rate": 1.5075354932653804e-06, + "loss": 1.6458, + "step": 27382 + }, + { + "epoch": 4.98, + "learning_rate": 1.5074808882417183e-06, + "loss": 0.4688, + "step": 27383 + }, + { + "epoch": 4.98, + "learning_rate": 1.5074262832180563e-06, + "loss": 0.0909, + "step": 27384 + }, + { + "epoch": 4.98, + "learning_rate": 1.5073716781943938e-06, + "loss": 0.3624, + "step": 27385 + }, + { + "epoch": 4.98, + "learning_rate": 1.5073170731707317e-06, + "loss": 0.2909, + "step": 27386 + }, + { + "epoch": 4.98, + "learning_rate": 1.5072624681470697e-06, + "loss": 0.1214, + "step": 27387 + }, + { + "epoch": 4.98, + "learning_rate": 1.5072078631234074e-06, + "loss": 0.2741, + "step": 27388 + }, + { + "epoch": 4.98, + "learning_rate": 1.5071532580997453e-06, + "loss": 0.0161, + "step": 27389 + }, + { + "epoch": 4.98, + "learning_rate": 1.5070986530760833e-06, + "loss": 1.3871, + "step": 27390 + }, + { + "epoch": 4.98, + "learning_rate": 1.5070440480524208e-06, + "loss": 0.8262, + "step": 27391 + }, + { + "epoch": 4.98, + "learning_rate": 1.5069894430287587e-06, + "loss": 0.0129, + "step": 27392 + }, + { + "epoch": 4.98, + "learning_rate": 1.5069348380050963e-06, + "loss": 3.3008, + "step": 27393 + }, + { + "epoch": 4.98, + "learning_rate": 1.5068802329814342e-06, + "loss": 0.1256, + "step": 27394 + }, + { + "epoch": 4.98, + "learning_rate": 1.5068256279577721e-06, + "loss": 0.2358, + "step": 27395 + }, + { + "epoch": 4.98, + "learning_rate": 1.5067710229341099e-06, + "loss": 0.366, + "step": 27396 + }, + { + "epoch": 4.98, + "learning_rate": 1.5067164179104478e-06, + "loss": 0.4688, + "step": 27397 + }, + { + "epoch": 4.98, + "learning_rate": 1.5066618128867858e-06, + "loss": 0.9112, + "step": 27398 + }, + { + "epoch": 4.98, + "learning_rate": 1.5066072078631233e-06, + "loss": 1.5156, + "step": 27399 + }, + { + "epoch": 4.98, + "learning_rate": 1.5065526028394612e-06, + "loss": 1.0695, + "step": 27400 + }, + { + "epoch": 4.98, + "learning_rate": 1.5064979978157992e-06, + "loss": 0.8426, + "step": 27401 + }, + { + "epoch": 4.98, + "learning_rate": 1.506443392792137e-06, + "loss": 0.5531, + "step": 27402 + }, + { + "epoch": 4.98, + "learning_rate": 1.5063887877684748e-06, + "loss": 0.4317, + "step": 27403 + }, + { + "epoch": 4.98, + "learning_rate": 1.5063341827448128e-06, + "loss": 0.1067, + "step": 27404 + }, + { + "epoch": 4.98, + "learning_rate": 1.5062795777211503e-06, + "loss": 0.4234, + "step": 27405 + }, + { + "epoch": 4.98, + "learning_rate": 1.5062249726974883e-06, + "loss": 0.3062, + "step": 27406 + }, + { + "epoch": 4.98, + "learning_rate": 1.5061703676738258e-06, + "loss": 1.0767, + "step": 27407 + }, + { + "epoch": 4.98, + "learning_rate": 1.5061157626501637e-06, + "loss": 0.7797, + "step": 27408 + }, + { + "epoch": 4.98, + "learning_rate": 1.5060611576265017e-06, + "loss": 0.5986, + "step": 27409 + }, + { + "epoch": 4.98, + "learning_rate": 1.5060065526028394e-06, + "loss": 0.3587, + "step": 27410 + }, + { + "epoch": 4.98, + "learning_rate": 1.5059519475791773e-06, + "loss": 0.3456, + "step": 27411 + }, + { + "epoch": 4.98, + "learning_rate": 1.5058973425555153e-06, + "loss": 0.3118, + "step": 27412 + }, + { + "epoch": 4.99, + "learning_rate": 1.5058427375318528e-06, + "loss": 0.1696, + "step": 27413 + }, + { + "epoch": 4.99, + "learning_rate": 1.5057881325081907e-06, + "loss": 1.2829, + "step": 27414 + }, + { + "epoch": 4.99, + "learning_rate": 1.5057335274845287e-06, + "loss": 0.2663, + "step": 27415 + }, + { + "epoch": 4.99, + "learning_rate": 1.5056789224608664e-06, + "loss": 1.1474, + "step": 27416 + }, + { + "epoch": 4.99, + "learning_rate": 1.5056243174372044e-06, + "loss": 0.7598, + "step": 27417 + }, + { + "epoch": 4.99, + "learning_rate": 1.5055697124135423e-06, + "loss": 0.8175, + "step": 27418 + }, + { + "epoch": 4.99, + "learning_rate": 1.5055151073898798e-06, + "loss": 0.0234, + "step": 27419 + }, + { + "epoch": 4.99, + "learning_rate": 1.5054605023662178e-06, + "loss": 0.0892, + "step": 27420 + }, + { + "epoch": 4.99, + "learning_rate": 1.5054058973425555e-06, + "loss": 3.4066, + "step": 27421 + }, + { + "epoch": 4.99, + "learning_rate": 1.5053512923188934e-06, + "loss": 0.5709, + "step": 27422 + }, + { + "epoch": 4.99, + "learning_rate": 1.5052966872952314e-06, + "loss": 1.4929, + "step": 27423 + }, + { + "epoch": 4.99, + "learning_rate": 1.505242082271569e-06, + "loss": 0.011, + "step": 27424 + }, + { + "epoch": 4.99, + "learning_rate": 1.5051874772479069e-06, + "loss": 0.0971, + "step": 27425 + }, + { + "epoch": 4.99, + "learning_rate": 1.5051328722242448e-06, + "loss": 1.6836, + "step": 27426 + }, + { + "epoch": 4.99, + "learning_rate": 1.5050782672005823e-06, + "loss": 0.8095, + "step": 27427 + }, + { + "epoch": 4.99, + "learning_rate": 1.5050236621769203e-06, + "loss": 0.3111, + "step": 27428 + }, + { + "epoch": 4.99, + "learning_rate": 1.5049690571532582e-06, + "loss": 1.2241, + "step": 27429 + }, + { + "epoch": 4.99, + "learning_rate": 1.504914452129596e-06, + "loss": 0.6368, + "step": 27430 + }, + { + "epoch": 4.99, + "learning_rate": 1.5048598471059339e-06, + "loss": 1.4688, + "step": 27431 + }, + { + "epoch": 4.99, + "learning_rate": 1.5048052420822718e-06, + "loss": 0.1241, + "step": 27432 + }, + { + "epoch": 4.99, + "learning_rate": 1.5047506370586093e-06, + "loss": 0.0336, + "step": 27433 + }, + { + "epoch": 4.99, + "learning_rate": 1.5046960320349473e-06, + "loss": 0.676, + "step": 27434 + }, + { + "epoch": 4.99, + "learning_rate": 1.504641427011285e-06, + "loss": 2.0752, + "step": 27435 + }, + { + "epoch": 4.99, + "learning_rate": 1.504586821987623e-06, + "loss": 0.2985, + "step": 27436 + }, + { + "epoch": 4.99, + "learning_rate": 1.504532216963961e-06, + "loss": 0.056, + "step": 27437 + }, + { + "epoch": 4.99, + "learning_rate": 1.5044776119402984e-06, + "loss": 1.7565, + "step": 27438 + }, + { + "epoch": 4.99, + "learning_rate": 1.5044230069166364e-06, + "loss": 1.2902, + "step": 27439 + }, + { + "epoch": 4.99, + "learning_rate": 1.5043684018929743e-06, + "loss": 0.2639, + "step": 27440 + }, + { + "epoch": 4.99, + "learning_rate": 1.5043137968693118e-06, + "loss": 0.2551, + "step": 27441 + }, + { + "epoch": 4.99, + "learning_rate": 1.5042591918456498e-06, + "loss": 1.6611, + "step": 27442 + }, + { + "epoch": 4.99, + "learning_rate": 1.5042045868219877e-06, + "loss": 0.0368, + "step": 27443 + }, + { + "epoch": 4.99, + "learning_rate": 1.5041499817983254e-06, + "loss": 0.3426, + "step": 27444 + }, + { + "epoch": 4.99, + "learning_rate": 1.5040953767746634e-06, + "loss": 1.3075, + "step": 27445 + }, + { + "epoch": 4.99, + "learning_rate": 1.504040771751001e-06, + "loss": 0.017, + "step": 27446 + }, + { + "epoch": 4.99, + "learning_rate": 1.5039861667273389e-06, + "loss": 0.3164, + "step": 27447 + }, + { + "epoch": 4.99, + "learning_rate": 1.5039315617036768e-06, + "loss": 0.9625, + "step": 27448 + }, + { + "epoch": 4.99, + "learning_rate": 1.5038769566800145e-06, + "loss": 0.0086, + "step": 27449 + }, + { + "epoch": 4.99, + "learning_rate": 1.5038223516563525e-06, + "loss": 0.0979, + "step": 27450 + }, + { + "epoch": 4.99, + "learning_rate": 1.5037677466326904e-06, + "loss": 0.3654, + "step": 27451 + }, + { + "epoch": 4.99, + "learning_rate": 1.503713141609028e-06, + "loss": 2.768, + "step": 27452 + }, + { + "epoch": 4.99, + "learning_rate": 1.5036585365853659e-06, + "loss": 1.9207, + "step": 27453 + }, + { + "epoch": 4.99, + "learning_rate": 1.5036039315617038e-06, + "loss": 1.0594, + "step": 27454 + }, + { + "epoch": 4.99, + "learning_rate": 1.5035493265380416e-06, + "loss": 0.1838, + "step": 27455 + }, + { + "epoch": 4.99, + "learning_rate": 1.5034947215143793e-06, + "loss": 1.4725, + "step": 27456 + }, + { + "epoch": 4.99, + "learning_rate": 1.5034401164907172e-06, + "loss": 0.0244, + "step": 27457 + }, + { + "epoch": 4.99, + "learning_rate": 1.503385511467055e-06, + "loss": 0.6888, + "step": 27458 + }, + { + "epoch": 4.99, + "learning_rate": 1.503330906443393e-06, + "loss": 0.6955, + "step": 27459 + }, + { + "epoch": 4.99, + "learning_rate": 1.5032763014197304e-06, + "loss": 0.1444, + "step": 27460 + }, + { + "epoch": 4.99, + "learning_rate": 1.5032216963960684e-06, + "loss": 0.3581, + "step": 27461 + }, + { + "epoch": 4.99, + "learning_rate": 1.5031670913724063e-06, + "loss": 0.0788, + "step": 27462 + }, + { + "epoch": 4.99, + "learning_rate": 1.503112486348744e-06, + "loss": 0.7352, + "step": 27463 + }, + { + "epoch": 4.99, + "learning_rate": 1.503057881325082e-06, + "loss": 1.1671, + "step": 27464 + }, + { + "epoch": 4.99, + "learning_rate": 1.50300327630142e-06, + "loss": 0.0116, + "step": 27465 + }, + { + "epoch": 4.99, + "learning_rate": 1.5029486712777574e-06, + "loss": 0.674, + "step": 27466 + }, + { + "epoch": 4.99, + "learning_rate": 1.5028940662540954e-06, + "loss": 5.7626, + "step": 27467 + }, + { + "epoch": 5.0, + "learning_rate": 1.5028394612304333e-06, + "loss": 0.1143, + "step": 27468 + }, + { + "epoch": 5.0, + "learning_rate": 1.502784856206771e-06, + "loss": 1.6423, + "step": 27469 + }, + { + "epoch": 5.0, + "learning_rate": 1.502730251183109e-06, + "loss": 0.7244, + "step": 27470 + }, + { + "epoch": 5.0, + "learning_rate": 1.502675646159447e-06, + "loss": 0.3269, + "step": 27471 + }, + { + "epoch": 5.0, + "learning_rate": 1.5026210411357845e-06, + "loss": 0.8901, + "step": 27472 + }, + { + "epoch": 5.0, + "learning_rate": 1.5025664361121224e-06, + "loss": 0.1471, + "step": 27473 + }, + { + "epoch": 5.0, + "learning_rate": 1.50251183108846e-06, + "loss": 0.6073, + "step": 27474 + }, + { + "epoch": 5.0, + "learning_rate": 1.5024572260647979e-06, + "loss": 0.0162, + "step": 27475 + }, + { + "epoch": 5.0, + "learning_rate": 1.5024026210411358e-06, + "loss": 0.453, + "step": 27476 + }, + { + "epoch": 5.0, + "learning_rate": 1.5023480160174736e-06, + "loss": 0.1888, + "step": 27477 + }, + { + "epoch": 5.0, + "learning_rate": 1.5022934109938115e-06, + "loss": 1.0932, + "step": 27478 + }, + { + "epoch": 5.0, + "learning_rate": 1.5022388059701494e-06, + "loss": 0.5523, + "step": 27479 + }, + { + "epoch": 5.0, + "learning_rate": 1.502184200946487e-06, + "loss": 0.142, + "step": 27480 + }, + { + "epoch": 5.0, + "learning_rate": 1.502129595922825e-06, + "loss": 0.8778, + "step": 27481 + }, + { + "epoch": 5.0, + "learning_rate": 1.5020749908991628e-06, + "loss": 0.2486, + "step": 27482 + }, + { + "epoch": 5.0, + "learning_rate": 1.5020203858755006e-06, + "loss": 1.8421, + "step": 27483 + }, + { + "epoch": 5.0, + "learning_rate": 1.5019657808518385e-06, + "loss": 0.444, + "step": 27484 + }, + { + "epoch": 5.0, + "learning_rate": 1.5019111758281765e-06, + "loss": 0.7937, + "step": 27485 + }, + { + "epoch": 5.0, + "learning_rate": 1.501856570804514e-06, + "loss": 0.3285, + "step": 27486 + }, + { + "epoch": 5.0, + "learning_rate": 1.501801965780852e-06, + "loss": 1.1454, + "step": 27487 + }, + { + "epoch": 5.0, + "learning_rate": 1.5017473607571894e-06, + "loss": 1.1958, + "step": 27488 + }, + { + "epoch": 5.0, + "learning_rate": 1.5016927557335274e-06, + "loss": 0.2777, + "step": 27489 + }, + { + "epoch": 5.0, + "learning_rate": 1.5016381507098653e-06, + "loss": 0.4654, + "step": 27490 + }, + { + "epoch": 5.0, + "learning_rate": 1.501583545686203e-06, + "loss": 0.0716, + "step": 27491 + }, + { + "epoch": 5.0, + "learning_rate": 1.501528940662541e-06, + "loss": 1.0547, + "step": 27492 + }, + { + "epoch": 5.0, + "learning_rate": 1.501474335638879e-06, + "loss": 0.1055, + "step": 27493 + }, + { + "epoch": 5.0, + "learning_rate": 1.5014197306152165e-06, + "loss": 0.3387, + "step": 27494 + }, + { + "epoch": 5.0, + "learning_rate": 1.5013651255915544e-06, + "loss": 2.1828, + "step": 27495 + }, + { + "epoch": 5.0, + "eval_accuracy": 0.8539279869067103, + "eval_loss": 0.6641533374786377, + "eval_runtime": 140.9317, + "eval_samples_per_second": 17.342, + "eval_steps_per_second": 4.335, + "step": 27495 + }, + { + "epoch": 5.0, + "learning_rate": 1.5013105205678924e-06, + "loss": 1.8916, + "step": 27496 + }, + { + "epoch": 5.0, + "learning_rate": 1.50125591554423e-06, + "loss": 0.8972, + "step": 27497 + }, + { + "epoch": 5.0, + "learning_rate": 1.501201310520568e-06, + "loss": 0.4183, + "step": 27498 + }, + { + "epoch": 5.0, + "learning_rate": 1.501146705496906e-06, + "loss": 1.4, + "step": 27499 + }, + { + "epoch": 5.0, + "learning_rate": 1.5010921004732435e-06, + "loss": 0.0481, + "step": 27500 + }, + { + "epoch": 5.0, + "learning_rate": 1.5010374954495814e-06, + "loss": 2.3774, + "step": 27501 + }, + { + "epoch": 5.0, + "learning_rate": 1.5009828904259192e-06, + "loss": 0.0183, + "step": 27502 + }, + { + "epoch": 5.0, + "learning_rate": 1.5009282854022571e-06, + "loss": 0.0103, + "step": 27503 + }, + { + "epoch": 5.0, + "learning_rate": 1.500873680378595e-06, + "loss": 0.0091, + "step": 27504 + }, + { + "epoch": 5.0, + "learning_rate": 1.5008190753549326e-06, + "loss": 0.2208, + "step": 27505 + }, + { + "epoch": 5.0, + "learning_rate": 1.5007644703312705e-06, + "loss": 1.1024, + "step": 27506 + }, + { + "epoch": 5.0, + "learning_rate": 1.5007098653076085e-06, + "loss": 0.1717, + "step": 27507 + }, + { + "epoch": 5.0, + "learning_rate": 1.500655260283946e-06, + "loss": 1.0973, + "step": 27508 + }, + { + "epoch": 5.0, + "learning_rate": 1.500600655260284e-06, + "loss": 0.1407, + "step": 27509 + }, + { + "epoch": 5.0, + "learning_rate": 1.5005460502366219e-06, + "loss": 0.0076, + "step": 27510 + }, + { + "epoch": 5.0, + "learning_rate": 1.5004914452129596e-06, + "loss": 0.5926, + "step": 27511 + }, + { + "epoch": 5.0, + "learning_rate": 1.5004368401892975e-06, + "loss": 0.0886, + "step": 27512 + }, + { + "epoch": 5.0, + "learning_rate": 1.5003822351656355e-06, + "loss": 0.0383, + "step": 27513 + }, + { + "epoch": 5.0, + "learning_rate": 1.500327630141973e-06, + "loss": 0.3611, + "step": 27514 + }, + { + "epoch": 5.0, + "learning_rate": 1.500273025118311e-06, + "loss": 0.0041, + "step": 27515 + }, + { + "epoch": 5.0, + "learning_rate": 1.5002184200946487e-06, + "loss": 0.1934, + "step": 27516 + }, + { + "epoch": 5.0, + "learning_rate": 1.5001638150709866e-06, + "loss": 0.0122, + "step": 27517 + }, + { + "epoch": 5.0, + "learning_rate": 1.5001092100473246e-06, + "loss": 0.8206, + "step": 27518 + }, + { + "epoch": 5.0, + "learning_rate": 1.500054605023662e-06, + "loss": 0.0213, + "step": 27519 + }, + { + "epoch": 5.0, + "learning_rate": 1.5e-06, + "loss": 0.3634, + "step": 27520 + }, + { + "epoch": 5.0, + "learning_rate": 1.4999453949763378e-06, + "loss": 1.0232, + "step": 27521 + }, + { + "epoch": 5.0, + "learning_rate": 1.4998907899526757e-06, + "loss": 0.157, + "step": 27522 + }, + { + "epoch": 5.01, + "learning_rate": 1.4998361849290134e-06, + "loss": 0.3309, + "step": 27523 + }, + { + "epoch": 5.01, + "learning_rate": 1.4997815799053514e-06, + "loss": 0.0552, + "step": 27524 + }, + { + "epoch": 5.01, + "learning_rate": 1.4997269748816891e-06, + "loss": 0.0737, + "step": 27525 + }, + { + "epoch": 5.01, + "learning_rate": 1.499672369858027e-06, + "loss": 2.0249, + "step": 27526 + }, + { + "epoch": 5.01, + "learning_rate": 1.4996177648343648e-06, + "loss": 0.5949, + "step": 27527 + }, + { + "epoch": 5.01, + "learning_rate": 1.4995631598107025e-06, + "loss": 0.3749, + "step": 27528 + }, + { + "epoch": 5.01, + "learning_rate": 1.4995085547870405e-06, + "loss": 0.2491, + "step": 27529 + }, + { + "epoch": 5.01, + "learning_rate": 1.4994539497633782e-06, + "loss": 1.3352, + "step": 27530 + }, + { + "epoch": 5.01, + "learning_rate": 1.4993993447397161e-06, + "loss": 0.0127, + "step": 27531 + }, + { + "epoch": 5.01, + "learning_rate": 1.4993447397160539e-06, + "loss": 0.5828, + "step": 27532 + }, + { + "epoch": 5.01, + "learning_rate": 1.4992901346923918e-06, + "loss": 0.7476, + "step": 27533 + }, + { + "epoch": 5.01, + "learning_rate": 1.4992355296687296e-06, + "loss": 1.3665, + "step": 27534 + }, + { + "epoch": 5.01, + "learning_rate": 1.4991809246450673e-06, + "loss": 0.017, + "step": 27535 + }, + { + "epoch": 5.01, + "learning_rate": 1.4991263196214052e-06, + "loss": 0.3708, + "step": 27536 + }, + { + "epoch": 5.01, + "learning_rate": 1.499071714597743e-06, + "loss": 0.2279, + "step": 27537 + }, + { + "epoch": 5.01, + "learning_rate": 1.499017109574081e-06, + "loss": 2.5602, + "step": 27538 + }, + { + "epoch": 5.01, + "learning_rate": 1.4989625045504186e-06, + "loss": 1.7419, + "step": 27539 + }, + { + "epoch": 5.01, + "learning_rate": 1.4989078995267566e-06, + "loss": 0.2162, + "step": 27540 + }, + { + "epoch": 5.01, + "learning_rate": 1.4988532945030943e-06, + "loss": 0.4285, + "step": 27541 + }, + { + "epoch": 5.01, + "learning_rate": 1.498798689479432e-06, + "loss": 0.4275, + "step": 27542 + }, + { + "epoch": 5.01, + "learning_rate": 1.49874408445577e-06, + "loss": 0.0079, + "step": 27543 + }, + { + "epoch": 5.01, + "learning_rate": 1.498689479432108e-06, + "loss": 0.7663, + "step": 27544 + }, + { + "epoch": 5.01, + "learning_rate": 1.4986348744084457e-06, + "loss": 2.6584, + "step": 27545 + }, + { + "epoch": 5.01, + "learning_rate": 1.4985802693847834e-06, + "loss": 0.5746, + "step": 27546 + }, + { + "epoch": 5.01, + "learning_rate": 1.4985256643611213e-06, + "loss": 0.0454, + "step": 27547 + }, + { + "epoch": 5.01, + "learning_rate": 1.498471059337459e-06, + "loss": 0.3853, + "step": 27548 + }, + { + "epoch": 5.01, + "learning_rate": 1.4984164543137968e-06, + "loss": 0.2521, + "step": 27549 + }, + { + "epoch": 5.01, + "learning_rate": 1.4983618492901347e-06, + "loss": 1.499, + "step": 27550 + }, + { + "epoch": 5.01, + "learning_rate": 1.4983072442664727e-06, + "loss": 0.2456, + "step": 27551 + }, + { + "epoch": 5.01, + "learning_rate": 1.4982526392428104e-06, + "loss": 0.1615, + "step": 27552 + }, + { + "epoch": 5.01, + "learning_rate": 1.4981980342191481e-06, + "loss": 1.1176, + "step": 27553 + }, + { + "epoch": 5.01, + "learning_rate": 1.498143429195486e-06, + "loss": 0.736, + "step": 27554 + }, + { + "epoch": 5.01, + "learning_rate": 1.4980888241718238e-06, + "loss": 1.3167, + "step": 27555 + }, + { + "epoch": 5.01, + "learning_rate": 1.4980342191481616e-06, + "loss": 0.4011, + "step": 27556 + }, + { + "epoch": 5.01, + "learning_rate": 1.4979796141244995e-06, + "loss": 0.2847, + "step": 27557 + }, + { + "epoch": 5.01, + "learning_rate": 1.4979250091008374e-06, + "loss": 0.0139, + "step": 27558 + }, + { + "epoch": 5.01, + "learning_rate": 1.4978704040771752e-06, + "loss": 0.132, + "step": 27559 + }, + { + "epoch": 5.01, + "learning_rate": 1.497815799053513e-06, + "loss": 0.233, + "step": 27560 + }, + { + "epoch": 5.01, + "learning_rate": 1.4977611940298508e-06, + "loss": 1.2055, + "step": 27561 + }, + { + "epoch": 5.01, + "learning_rate": 1.4977065890061886e-06, + "loss": 0.6074, + "step": 27562 + }, + { + "epoch": 5.01, + "learning_rate": 1.4976519839825263e-06, + "loss": 0.6238, + "step": 27563 + }, + { + "epoch": 5.01, + "learning_rate": 1.4975973789588643e-06, + "loss": 0.4446, + "step": 27564 + }, + { + "epoch": 5.01, + "learning_rate": 1.4975427739352022e-06, + "loss": 0.0093, + "step": 27565 + }, + { + "epoch": 5.01, + "learning_rate": 1.49748816891154e-06, + "loss": 0.844, + "step": 27566 + }, + { + "epoch": 5.01, + "learning_rate": 1.4974335638878777e-06, + "loss": 0.2813, + "step": 27567 + }, + { + "epoch": 5.01, + "learning_rate": 1.4973789588642156e-06, + "loss": 0.4689, + "step": 27568 + }, + { + "epoch": 5.01, + "learning_rate": 1.4973243538405533e-06, + "loss": 0.0136, + "step": 27569 + }, + { + "epoch": 5.01, + "learning_rate": 1.497269748816891e-06, + "loss": 2.5125, + "step": 27570 + }, + { + "epoch": 5.01, + "learning_rate": 1.497215143793229e-06, + "loss": 0.0901, + "step": 27571 + }, + { + "epoch": 5.01, + "learning_rate": 1.497160538769567e-06, + "loss": 0.1614, + "step": 27572 + }, + { + "epoch": 5.01, + "learning_rate": 1.4971059337459047e-06, + "loss": 0.9297, + "step": 27573 + }, + { + "epoch": 5.01, + "learning_rate": 1.4970513287222424e-06, + "loss": 0.5767, + "step": 27574 + }, + { + "epoch": 5.01, + "learning_rate": 1.4969967236985804e-06, + "loss": 0.8269, + "step": 27575 + }, + { + "epoch": 5.01, + "learning_rate": 1.496942118674918e-06, + "loss": 0.971, + "step": 27576 + }, + { + "epoch": 5.01, + "learning_rate": 1.4968875136512558e-06, + "loss": 0.029, + "step": 27577 + }, + { + "epoch": 5.02, + "learning_rate": 1.4968329086275938e-06, + "loss": 2.5725, + "step": 27578 + }, + { + "epoch": 5.02, + "learning_rate": 1.4967783036039317e-06, + "loss": 1.0347, + "step": 27579 + }, + { + "epoch": 5.02, + "learning_rate": 1.4967236985802694e-06, + "loss": 2.8026, + "step": 27580 + }, + { + "epoch": 5.02, + "learning_rate": 1.4966690935566072e-06, + "loss": 0.1204, + "step": 27581 + }, + { + "epoch": 5.02, + "learning_rate": 1.4966144885329451e-06, + "loss": 1.6492, + "step": 27582 + }, + { + "epoch": 5.02, + "learning_rate": 1.4965598835092828e-06, + "loss": 0.7573, + "step": 27583 + }, + { + "epoch": 5.02, + "learning_rate": 1.4965052784856208e-06, + "loss": 0.0108, + "step": 27584 + }, + { + "epoch": 5.02, + "learning_rate": 1.4964506734619585e-06, + "loss": 0.6213, + "step": 27585 + }, + { + "epoch": 5.02, + "learning_rate": 1.4963960684382965e-06, + "loss": 1.3512, + "step": 27586 + }, + { + "epoch": 5.02, + "learning_rate": 1.4963414634146342e-06, + "loss": 0.0991, + "step": 27587 + }, + { + "epoch": 5.02, + "learning_rate": 1.496286858390972e-06, + "loss": 0.258, + "step": 27588 + }, + { + "epoch": 5.02, + "learning_rate": 1.4962322533673099e-06, + "loss": 0.3684, + "step": 27589 + }, + { + "epoch": 5.02, + "learning_rate": 1.4961776483436476e-06, + "loss": 0.4266, + "step": 27590 + }, + { + "epoch": 5.02, + "learning_rate": 1.4961230433199855e-06, + "loss": 0.3835, + "step": 27591 + }, + { + "epoch": 5.02, + "learning_rate": 1.4960684382963233e-06, + "loss": 0.0858, + "step": 27592 + }, + { + "epoch": 5.02, + "learning_rate": 1.4960138332726612e-06, + "loss": 2.1957, + "step": 27593 + }, + { + "epoch": 5.02, + "learning_rate": 1.495959228248999e-06, + "loss": 2.0817, + "step": 27594 + }, + { + "epoch": 5.02, + "learning_rate": 1.4959046232253367e-06, + "loss": 0.2493, + "step": 27595 + }, + { + "epoch": 5.02, + "learning_rate": 1.4958500182016746e-06, + "loss": 0.037, + "step": 27596 + }, + { + "epoch": 5.02, + "learning_rate": 1.4957954131780124e-06, + "loss": 2.5139, + "step": 27597 + }, + { + "epoch": 5.02, + "learning_rate": 1.4957408081543503e-06, + "loss": 2.3428, + "step": 27598 + }, + { + "epoch": 5.02, + "learning_rate": 1.495686203130688e-06, + "loss": 0.7033, + "step": 27599 + }, + { + "epoch": 5.02, + "learning_rate": 1.495631598107026e-06, + "loss": 0.1114, + "step": 27600 + }, + { + "epoch": 5.02, + "learning_rate": 1.4955769930833637e-06, + "loss": 0.4561, + "step": 27601 + }, + { + "epoch": 5.02, + "learning_rate": 1.4955223880597014e-06, + "loss": 0.3316, + "step": 27602 + }, + { + "epoch": 5.02, + "learning_rate": 1.4954677830360394e-06, + "loss": 0.1883, + "step": 27603 + }, + { + "epoch": 5.02, + "learning_rate": 1.4954131780123771e-06, + "loss": 0.883, + "step": 27604 + }, + { + "epoch": 5.02, + "learning_rate": 1.495358572988715e-06, + "loss": 0.4524, + "step": 27605 + }, + { + "epoch": 5.02, + "learning_rate": 1.4953039679650528e-06, + "loss": 0.6289, + "step": 27606 + }, + { + "epoch": 5.02, + "learning_rate": 1.4952493629413907e-06, + "loss": 0.0229, + "step": 27607 + }, + { + "epoch": 5.02, + "learning_rate": 1.4951947579177285e-06, + "loss": 0.0155, + "step": 27608 + }, + { + "epoch": 5.02, + "learning_rate": 1.4951401528940662e-06, + "loss": 0.0236, + "step": 27609 + }, + { + "epoch": 5.02, + "learning_rate": 1.4950855478704041e-06, + "loss": 0.238, + "step": 27610 + }, + { + "epoch": 5.02, + "learning_rate": 1.4950309428467419e-06, + "loss": 1.8894, + "step": 27611 + }, + { + "epoch": 5.02, + "learning_rate": 1.4949763378230798e-06, + "loss": 0.2333, + "step": 27612 + }, + { + "epoch": 5.02, + "learning_rate": 1.4949217327994176e-06, + "loss": 0.2782, + "step": 27613 + }, + { + "epoch": 5.02, + "learning_rate": 1.4948671277757555e-06, + "loss": 0.3005, + "step": 27614 + }, + { + "epoch": 5.02, + "learning_rate": 1.4948125227520932e-06, + "loss": 0.0116, + "step": 27615 + }, + { + "epoch": 5.02, + "learning_rate": 1.494757917728431e-06, + "loss": 0.5223, + "step": 27616 + }, + { + "epoch": 5.02, + "learning_rate": 1.494703312704769e-06, + "loss": 2.8515, + "step": 27617 + }, + { + "epoch": 5.02, + "learning_rate": 1.4946487076811066e-06, + "loss": 0.8032, + "step": 27618 + }, + { + "epoch": 5.02, + "learning_rate": 1.4945941026574446e-06, + "loss": 1.1744, + "step": 27619 + }, + { + "epoch": 5.02, + "learning_rate": 1.4945394976337823e-06, + "loss": 0.2036, + "step": 27620 + }, + { + "epoch": 5.02, + "learning_rate": 1.4944848926101203e-06, + "loss": 0.2834, + "step": 27621 + }, + { + "epoch": 5.02, + "learning_rate": 1.494430287586458e-06, + "loss": 0.0366, + "step": 27622 + }, + { + "epoch": 5.02, + "learning_rate": 1.4943756825627957e-06, + "loss": 0.3225, + "step": 27623 + }, + { + "epoch": 5.02, + "learning_rate": 1.4943210775391337e-06, + "loss": 1.2032, + "step": 27624 + }, + { + "epoch": 5.02, + "learning_rate": 1.4942664725154714e-06, + "loss": 0.022, + "step": 27625 + }, + { + "epoch": 5.02, + "learning_rate": 1.4942118674918093e-06, + "loss": 0.0086, + "step": 27626 + }, + { + "epoch": 5.02, + "learning_rate": 1.494157262468147e-06, + "loss": 1.2228, + "step": 27627 + }, + { + "epoch": 5.02, + "learning_rate": 1.494102657444485e-06, + "loss": 0.4954, + "step": 27628 + }, + { + "epoch": 5.02, + "learning_rate": 1.4940480524208227e-06, + "loss": 0.4291, + "step": 27629 + }, + { + "epoch": 5.02, + "learning_rate": 1.4939934473971605e-06, + "loss": 0.2262, + "step": 27630 + }, + { + "epoch": 5.02, + "learning_rate": 1.4939388423734984e-06, + "loss": 1.5431, + "step": 27631 + }, + { + "epoch": 5.02, + "learning_rate": 1.4938842373498364e-06, + "loss": 0.0101, + "step": 27632 + }, + { + "epoch": 5.03, + "learning_rate": 1.493829632326174e-06, + "loss": 0.393, + "step": 27633 + }, + { + "epoch": 5.03, + "learning_rate": 1.4937750273025118e-06, + "loss": 0.8276, + "step": 27634 + }, + { + "epoch": 5.03, + "learning_rate": 1.4937204222788498e-06, + "loss": 0.2061, + "step": 27635 + }, + { + "epoch": 5.03, + "learning_rate": 1.4936658172551875e-06, + "loss": 0.9184, + "step": 27636 + }, + { + "epoch": 5.03, + "learning_rate": 1.4936112122315252e-06, + "loss": 0.8394, + "step": 27637 + }, + { + "epoch": 5.03, + "learning_rate": 1.4935566072078632e-06, + "loss": 0.0232, + "step": 27638 + }, + { + "epoch": 5.03, + "learning_rate": 1.4935020021842011e-06, + "loss": 0.0224, + "step": 27639 + }, + { + "epoch": 5.03, + "learning_rate": 1.4934473971605388e-06, + "loss": 0.3844, + "step": 27640 + }, + { + "epoch": 5.03, + "learning_rate": 1.4933927921368766e-06, + "loss": 0.0348, + "step": 27641 + }, + { + "epoch": 5.03, + "learning_rate": 1.4933381871132145e-06, + "loss": 0.9913, + "step": 27642 + }, + { + "epoch": 5.03, + "learning_rate": 1.4932835820895523e-06, + "loss": 0.0136, + "step": 27643 + }, + { + "epoch": 5.03, + "learning_rate": 1.49322897706589e-06, + "loss": 3.3003, + "step": 27644 + }, + { + "epoch": 5.03, + "learning_rate": 1.493174372042228e-06, + "loss": 0.2961, + "step": 27645 + }, + { + "epoch": 5.03, + "learning_rate": 1.4931197670185659e-06, + "loss": 1.1918, + "step": 27646 + }, + { + "epoch": 5.03, + "learning_rate": 1.4930651619949036e-06, + "loss": 0.4907, + "step": 27647 + }, + { + "epoch": 5.03, + "learning_rate": 1.4930105569712413e-06, + "loss": 0.0221, + "step": 27648 + }, + { + "epoch": 5.03, + "learning_rate": 1.4929559519475793e-06, + "loss": 0.0946, + "step": 27649 + }, + { + "epoch": 5.03, + "learning_rate": 1.492901346923917e-06, + "loss": 0.4628, + "step": 27650 + }, + { + "epoch": 5.03, + "learning_rate": 1.4928467419002547e-06, + "loss": 1.0155, + "step": 27651 + }, + { + "epoch": 5.03, + "learning_rate": 1.4927921368765927e-06, + "loss": 0.268, + "step": 27652 + }, + { + "epoch": 5.03, + "learning_rate": 1.4927375318529306e-06, + "loss": 0.6647, + "step": 27653 + }, + { + "epoch": 5.03, + "learning_rate": 1.4926829268292684e-06, + "loss": 0.3273, + "step": 27654 + }, + { + "epoch": 5.03, + "learning_rate": 1.492628321805606e-06, + "loss": 0.6184, + "step": 27655 + }, + { + "epoch": 5.03, + "learning_rate": 1.492573716781944e-06, + "loss": 0.2143, + "step": 27656 + }, + { + "epoch": 5.03, + "learning_rate": 1.4925191117582818e-06, + "loss": 2.1108, + "step": 27657 + }, + { + "epoch": 5.03, + "learning_rate": 1.4924645067346195e-06, + "loss": 0.4525, + "step": 27658 + }, + { + "epoch": 5.03, + "learning_rate": 1.4924099017109574e-06, + "loss": 0.9715, + "step": 27659 + }, + { + "epoch": 5.03, + "learning_rate": 1.4923552966872954e-06, + "loss": 0.0108, + "step": 27660 + }, + { + "epoch": 5.03, + "learning_rate": 1.4923006916636331e-06, + "loss": 2.8433, + "step": 27661 + }, + { + "epoch": 5.03, + "learning_rate": 1.4922460866399708e-06, + "loss": 0.1415, + "step": 27662 + }, + { + "epoch": 5.03, + "learning_rate": 1.4921914816163088e-06, + "loss": 0.4948, + "step": 27663 + }, + { + "epoch": 5.03, + "learning_rate": 1.4921368765926465e-06, + "loss": 1.7499, + "step": 27664 + }, + { + "epoch": 5.03, + "learning_rate": 1.4920822715689843e-06, + "loss": 0.9276, + "step": 27665 + }, + { + "epoch": 5.03, + "learning_rate": 1.4920276665453222e-06, + "loss": 1.1404, + "step": 27666 + }, + { + "epoch": 5.03, + "learning_rate": 1.4919730615216601e-06, + "loss": 1.8975, + "step": 27667 + }, + { + "epoch": 5.03, + "learning_rate": 1.4919184564979979e-06, + "loss": 2.6955, + "step": 27668 + }, + { + "epoch": 5.03, + "learning_rate": 1.4918638514743356e-06, + "loss": 0.0963, + "step": 27669 + }, + { + "epoch": 5.03, + "learning_rate": 1.4918092464506735e-06, + "loss": 1.6299, + "step": 27670 + }, + { + "epoch": 5.03, + "learning_rate": 1.4917546414270113e-06, + "loss": 0.0254, + "step": 27671 + }, + { + "epoch": 5.03, + "learning_rate": 1.4917000364033492e-06, + "loss": 0.5079, + "step": 27672 + }, + { + "epoch": 5.03, + "learning_rate": 1.491645431379687e-06, + "loss": 0.0122, + "step": 27673 + }, + { + "epoch": 5.03, + "learning_rate": 1.491590826356025e-06, + "loss": 0.1163, + "step": 27674 + }, + { + "epoch": 5.03, + "learning_rate": 1.4915362213323626e-06, + "loss": 0.3776, + "step": 27675 + }, + { + "epoch": 5.03, + "learning_rate": 1.4914816163087004e-06, + "loss": 0.0439, + "step": 27676 + }, + { + "epoch": 5.03, + "learning_rate": 1.4914270112850383e-06, + "loss": 0.1514, + "step": 27677 + }, + { + "epoch": 5.03, + "learning_rate": 1.491372406261376e-06, + "loss": 0.6004, + "step": 27678 + }, + { + "epoch": 5.03, + "learning_rate": 1.491317801237714e-06, + "loss": 0.6013, + "step": 27679 + }, + { + "epoch": 5.03, + "learning_rate": 1.4912631962140517e-06, + "loss": 1.3839, + "step": 27680 + }, + { + "epoch": 5.03, + "learning_rate": 1.4912085911903897e-06, + "loss": 0.5786, + "step": 27681 + }, + { + "epoch": 5.03, + "learning_rate": 1.4911539861667274e-06, + "loss": 0.0345, + "step": 27682 + }, + { + "epoch": 5.03, + "learning_rate": 1.4910993811430651e-06, + "loss": 0.0106, + "step": 27683 + }, + { + "epoch": 5.03, + "learning_rate": 1.491044776119403e-06, + "loss": 0.1667, + "step": 27684 + }, + { + "epoch": 5.03, + "learning_rate": 1.4909901710957408e-06, + "loss": 0.4363, + "step": 27685 + }, + { + "epoch": 5.03, + "learning_rate": 1.4909355660720787e-06, + "loss": 0.1621, + "step": 27686 + }, + { + "epoch": 5.03, + "learning_rate": 1.4908809610484165e-06, + "loss": 0.0292, + "step": 27687 + }, + { + "epoch": 5.04, + "learning_rate": 1.4908263560247544e-06, + "loss": 0.6808, + "step": 27688 + }, + { + "epoch": 5.04, + "learning_rate": 1.4907717510010921e-06, + "loss": 2.2527, + "step": 27689 + }, + { + "epoch": 5.04, + "learning_rate": 1.4907171459774299e-06, + "loss": 0.0061, + "step": 27690 + }, + { + "epoch": 5.04, + "learning_rate": 1.4906625409537678e-06, + "loss": 0.0133, + "step": 27691 + }, + { + "epoch": 5.04, + "learning_rate": 1.4906079359301055e-06, + "loss": 1.0807, + "step": 27692 + }, + { + "epoch": 5.04, + "learning_rate": 1.4905533309064435e-06, + "loss": 0.1224, + "step": 27693 + }, + { + "epoch": 5.04, + "learning_rate": 1.4904987258827812e-06, + "loss": 0.2573, + "step": 27694 + }, + { + "epoch": 5.04, + "learning_rate": 1.4904441208591192e-06, + "loss": 0.0137, + "step": 27695 + }, + { + "epoch": 5.04, + "learning_rate": 1.490389515835457e-06, + "loss": 0.2075, + "step": 27696 + }, + { + "epoch": 5.04, + "learning_rate": 1.4903349108117946e-06, + "loss": 0.1196, + "step": 27697 + }, + { + "epoch": 5.04, + "learning_rate": 1.4902803057881326e-06, + "loss": 0.1379, + "step": 27698 + }, + { + "epoch": 5.04, + "learning_rate": 1.4902257007644703e-06, + "loss": 1.4339, + "step": 27699 + }, + { + "epoch": 5.04, + "learning_rate": 1.4901710957408082e-06, + "loss": 0.5407, + "step": 27700 + }, + { + "epoch": 5.04, + "learning_rate": 1.490116490717146e-06, + "loss": 2.0132, + "step": 27701 + }, + { + "epoch": 5.04, + "learning_rate": 1.490061885693484e-06, + "loss": 1.0083, + "step": 27702 + }, + { + "epoch": 5.04, + "learning_rate": 1.4900072806698217e-06, + "loss": 1.2819, + "step": 27703 + }, + { + "epoch": 5.04, + "learning_rate": 1.4899526756461594e-06, + "loss": 0.298, + "step": 27704 + }, + { + "epoch": 5.04, + "learning_rate": 1.4898980706224973e-06, + "loss": 1.4082, + "step": 27705 + }, + { + "epoch": 5.04, + "learning_rate": 1.489843465598835e-06, + "loss": 2.1473, + "step": 27706 + }, + { + "epoch": 5.04, + "learning_rate": 1.489788860575173e-06, + "loss": 0.0216, + "step": 27707 + }, + { + "epoch": 5.04, + "learning_rate": 1.4897342555515107e-06, + "loss": 0.0111, + "step": 27708 + }, + { + "epoch": 5.04, + "learning_rate": 1.4896796505278487e-06, + "loss": 0.411, + "step": 27709 + }, + { + "epoch": 5.04, + "learning_rate": 1.4896250455041864e-06, + "loss": 0.9644, + "step": 27710 + }, + { + "epoch": 5.04, + "learning_rate": 1.4895704404805241e-06, + "loss": 0.4334, + "step": 27711 + }, + { + "epoch": 5.04, + "learning_rate": 1.489515835456862e-06, + "loss": 0.9195, + "step": 27712 + }, + { + "epoch": 5.04, + "learning_rate": 1.4894612304332e-06, + "loss": 1.5928, + "step": 27713 + }, + { + "epoch": 5.04, + "learning_rate": 1.4894066254095378e-06, + "loss": 1.1639, + "step": 27714 + }, + { + "epoch": 5.04, + "learning_rate": 1.4893520203858755e-06, + "loss": 0.8328, + "step": 27715 + }, + { + "epoch": 5.04, + "learning_rate": 1.4892974153622134e-06, + "loss": 0.0089, + "step": 27716 + }, + { + "epoch": 5.04, + "learning_rate": 1.4892428103385512e-06, + "loss": 2.2998, + "step": 27717 + }, + { + "epoch": 5.04, + "learning_rate": 1.489188205314889e-06, + "loss": 1.5095, + "step": 27718 + }, + { + "epoch": 5.04, + "learning_rate": 1.4891336002912268e-06, + "loss": 3.8831, + "step": 27719 + }, + { + "epoch": 5.04, + "learning_rate": 1.4890789952675648e-06, + "loss": 0.3435, + "step": 27720 + }, + { + "epoch": 5.04, + "learning_rate": 1.4890243902439025e-06, + "loss": 0.3845, + "step": 27721 + }, + { + "epoch": 5.04, + "learning_rate": 1.4889697852202403e-06, + "loss": 2.0106, + "step": 27722 + }, + { + "epoch": 5.04, + "learning_rate": 1.4889151801965782e-06, + "loss": 0.0366, + "step": 27723 + }, + { + "epoch": 5.04, + "learning_rate": 1.488860575172916e-06, + "loss": 0.5957, + "step": 27724 + }, + { + "epoch": 5.04, + "learning_rate": 1.4888059701492537e-06, + "loss": 0.7553, + "step": 27725 + }, + { + "epoch": 5.04, + "learning_rate": 1.4887513651255916e-06, + "loss": 0.4763, + "step": 27726 + }, + { + "epoch": 5.04, + "learning_rate": 1.4886967601019295e-06, + "loss": 0.706, + "step": 27727 + }, + { + "epoch": 5.04, + "learning_rate": 1.4886421550782673e-06, + "loss": 1.8944, + "step": 27728 + }, + { + "epoch": 5.04, + "learning_rate": 1.488587550054605e-06, + "loss": 0.2302, + "step": 27729 + }, + { + "epoch": 5.04, + "learning_rate": 1.488532945030943e-06, + "loss": 0.1805, + "step": 27730 + }, + { + "epoch": 5.04, + "learning_rate": 1.4884783400072807e-06, + "loss": 0.007, + "step": 27731 + }, + { + "epoch": 5.04, + "learning_rate": 1.4884237349836184e-06, + "loss": 0.6357, + "step": 27732 + }, + { + "epoch": 5.04, + "learning_rate": 1.4883691299599564e-06, + "loss": 0.013, + "step": 27733 + }, + { + "epoch": 5.04, + "learning_rate": 1.4883145249362943e-06, + "loss": 0.0203, + "step": 27734 + }, + { + "epoch": 5.04, + "learning_rate": 1.488259919912632e-06, + "loss": 1.0323, + "step": 27735 + }, + { + "epoch": 5.04, + "learning_rate": 1.4882053148889698e-06, + "loss": 0.0493, + "step": 27736 + }, + { + "epoch": 5.04, + "learning_rate": 1.4881507098653077e-06, + "loss": 0.0756, + "step": 27737 + }, + { + "epoch": 5.04, + "learning_rate": 1.4880961048416454e-06, + "loss": 0.1326, + "step": 27738 + }, + { + "epoch": 5.04, + "learning_rate": 1.4880414998179832e-06, + "loss": 2.6851, + "step": 27739 + }, + { + "epoch": 5.04, + "learning_rate": 1.4879868947943211e-06, + "loss": 1.2756, + "step": 27740 + }, + { + "epoch": 5.04, + "learning_rate": 1.487932289770659e-06, + "loss": 0.0077, + "step": 27741 + }, + { + "epoch": 5.04, + "learning_rate": 1.4878776847469968e-06, + "loss": 1.1626, + "step": 27742 + }, + { + "epoch": 5.05, + "learning_rate": 1.4878230797233345e-06, + "loss": 0.014, + "step": 27743 + }, + { + "epoch": 5.05, + "learning_rate": 1.4877684746996725e-06, + "loss": 2.7939, + "step": 27744 + }, + { + "epoch": 5.05, + "learning_rate": 1.4877138696760102e-06, + "loss": 0.0272, + "step": 27745 + }, + { + "epoch": 5.05, + "learning_rate": 1.487659264652348e-06, + "loss": 2.3282, + "step": 27746 + }, + { + "epoch": 5.05, + "learning_rate": 1.4876046596286859e-06, + "loss": 0.1824, + "step": 27747 + }, + { + "epoch": 5.05, + "learning_rate": 1.4875500546050238e-06, + "loss": 1.2397, + "step": 27748 + }, + { + "epoch": 5.05, + "learning_rate": 1.4874954495813615e-06, + "loss": 0.5399, + "step": 27749 + }, + { + "epoch": 5.05, + "learning_rate": 1.4874408445576993e-06, + "loss": 1.9438, + "step": 27750 + }, + { + "epoch": 5.05, + "learning_rate": 1.4873862395340372e-06, + "loss": 1.8168, + "step": 27751 + }, + { + "epoch": 5.05, + "learning_rate": 1.487331634510375e-06, + "loss": 1.5614, + "step": 27752 + }, + { + "epoch": 5.05, + "learning_rate": 1.487277029486713e-06, + "loss": 0.0318, + "step": 27753 + }, + { + "epoch": 5.05, + "learning_rate": 1.4872224244630506e-06, + "loss": 1.0053, + "step": 27754 + }, + { + "epoch": 5.05, + "learning_rate": 1.4871678194393886e-06, + "loss": 0.034, + "step": 27755 + }, + { + "epoch": 5.05, + "learning_rate": 1.4871132144157263e-06, + "loss": 0.8911, + "step": 27756 + }, + { + "epoch": 5.05, + "learning_rate": 1.487058609392064e-06, + "loss": 0.1959, + "step": 27757 + }, + { + "epoch": 5.05, + "learning_rate": 1.487004004368402e-06, + "loss": 0.1408, + "step": 27758 + }, + { + "epoch": 5.05, + "learning_rate": 1.4869493993447397e-06, + "loss": 0.4768, + "step": 27759 + }, + { + "epoch": 5.05, + "learning_rate": 1.4868947943210777e-06, + "loss": 0.1781, + "step": 27760 + }, + { + "epoch": 5.05, + "learning_rate": 1.4868401892974154e-06, + "loss": 0.023, + "step": 27761 + }, + { + "epoch": 5.05, + "learning_rate": 1.4867855842737533e-06, + "loss": 0.3394, + "step": 27762 + }, + { + "epoch": 5.05, + "learning_rate": 1.486730979250091e-06, + "loss": 0.2123, + "step": 27763 + }, + { + "epoch": 5.05, + "learning_rate": 1.4866763742264288e-06, + "loss": 0.6109, + "step": 27764 + }, + { + "epoch": 5.05, + "learning_rate": 1.4866217692027667e-06, + "loss": 0.2402, + "step": 27765 + }, + { + "epoch": 5.05, + "learning_rate": 1.4865671641791045e-06, + "loss": 0.153, + "step": 27766 + }, + { + "epoch": 5.05, + "learning_rate": 1.4865125591554424e-06, + "loss": 0.0301, + "step": 27767 + }, + { + "epoch": 5.05, + "learning_rate": 1.4864579541317801e-06, + "loss": 2.3939, + "step": 27768 + }, + { + "epoch": 5.05, + "learning_rate": 1.486403349108118e-06, + "loss": 0.4579, + "step": 27769 + }, + { + "epoch": 5.05, + "learning_rate": 1.4863487440844558e-06, + "loss": 2.6045, + "step": 27770 + }, + { + "epoch": 5.05, + "learning_rate": 1.4862941390607935e-06, + "loss": 0.0158, + "step": 27771 + }, + { + "epoch": 5.05, + "learning_rate": 1.4862395340371315e-06, + "loss": 0.292, + "step": 27772 + }, + { + "epoch": 5.05, + "learning_rate": 1.4861849290134692e-06, + "loss": 0.007, + "step": 27773 + }, + { + "epoch": 5.05, + "learning_rate": 1.4861303239898072e-06, + "loss": 0.6472, + "step": 27774 + }, + { + "epoch": 5.05, + "learning_rate": 1.486075718966145e-06, + "loss": 1.3284, + "step": 27775 + }, + { + "epoch": 5.05, + "learning_rate": 1.4860211139424828e-06, + "loss": 0.1084, + "step": 27776 + }, + { + "epoch": 5.05, + "learning_rate": 1.4859665089188206e-06, + "loss": 0.5756, + "step": 27777 + }, + { + "epoch": 5.05, + "learning_rate": 1.4859119038951583e-06, + "loss": 0.0443, + "step": 27778 + }, + { + "epoch": 5.05, + "learning_rate": 1.4858572988714962e-06, + "loss": 1.1967, + "step": 27779 + }, + { + "epoch": 5.05, + "learning_rate": 1.485802693847834e-06, + "loss": 0.214, + "step": 27780 + }, + { + "epoch": 5.05, + "learning_rate": 1.485748088824172e-06, + "loss": 0.3681, + "step": 27781 + }, + { + "epoch": 5.05, + "learning_rate": 1.4856934838005097e-06, + "loss": 0.0279, + "step": 27782 + }, + { + "epoch": 5.05, + "learning_rate": 1.4856388787768476e-06, + "loss": 0.3599, + "step": 27783 + }, + { + "epoch": 5.05, + "learning_rate": 1.4855842737531853e-06, + "loss": 0.3053, + "step": 27784 + }, + { + "epoch": 5.05, + "learning_rate": 1.485529668729523e-06, + "loss": 0.5695, + "step": 27785 + }, + { + "epoch": 5.05, + "learning_rate": 1.485475063705861e-06, + "loss": 0.0142, + "step": 27786 + }, + { + "epoch": 5.05, + "learning_rate": 1.4854204586821987e-06, + "loss": 0.0124, + "step": 27787 + }, + { + "epoch": 5.05, + "learning_rate": 1.4853658536585367e-06, + "loss": 2.3383, + "step": 27788 + }, + { + "epoch": 5.05, + "learning_rate": 1.4853112486348744e-06, + "loss": 0.0322, + "step": 27789 + }, + { + "epoch": 5.05, + "learning_rate": 1.4852566436112124e-06, + "loss": 0.7984, + "step": 27790 + }, + { + "epoch": 5.05, + "learning_rate": 1.48520203858755e-06, + "loss": 0.1147, + "step": 27791 + }, + { + "epoch": 5.05, + "learning_rate": 1.4851474335638878e-06, + "loss": 0.2439, + "step": 27792 + }, + { + "epoch": 5.05, + "learning_rate": 1.4850928285402256e-06, + "loss": 0.1472, + "step": 27793 + }, + { + "epoch": 5.05, + "learning_rate": 1.4850382235165635e-06, + "loss": 0.384, + "step": 27794 + }, + { + "epoch": 5.05, + "learning_rate": 1.4849836184929014e-06, + "loss": 0.0079, + "step": 27795 + }, + { + "epoch": 5.05, + "learning_rate": 1.4849290134692392e-06, + "loss": 1.1506, + "step": 27796 + }, + { + "epoch": 5.05, + "learning_rate": 1.4848744084455771e-06, + "loss": 2.1354, + "step": 27797 + }, + { + "epoch": 5.06, + "learning_rate": 1.4848198034219148e-06, + "loss": 0.5003, + "step": 27798 + }, + { + "epoch": 5.06, + "learning_rate": 1.4847651983982526e-06, + "loss": 0.2385, + "step": 27799 + }, + { + "epoch": 5.06, + "learning_rate": 1.4847105933745905e-06, + "loss": 0.197, + "step": 27800 + }, + { + "epoch": 5.06, + "learning_rate": 1.4846559883509285e-06, + "loss": 0.1193, + "step": 27801 + }, + { + "epoch": 5.06, + "learning_rate": 1.4846013833272662e-06, + "loss": 0.3716, + "step": 27802 + }, + { + "epoch": 5.06, + "learning_rate": 1.484546778303604e-06, + "loss": 0.72, + "step": 27803 + }, + { + "epoch": 5.06, + "learning_rate": 1.4844921732799419e-06, + "loss": 0.244, + "step": 27804 + }, + { + "epoch": 5.06, + "learning_rate": 1.4844375682562796e-06, + "loss": 0.786, + "step": 27805 + }, + { + "epoch": 5.06, + "learning_rate": 1.4843829632326173e-06, + "loss": 0.5682, + "step": 27806 + }, + { + "epoch": 5.06, + "learning_rate": 1.4843283582089553e-06, + "loss": 0.9823, + "step": 27807 + }, + { + "epoch": 5.06, + "learning_rate": 1.4842737531852932e-06, + "loss": 0.8644, + "step": 27808 + }, + { + "epoch": 5.06, + "learning_rate": 1.484219148161631e-06, + "loss": 0.0347, + "step": 27809 + }, + { + "epoch": 5.06, + "learning_rate": 1.4841645431379687e-06, + "loss": 1.3058, + "step": 27810 + }, + { + "epoch": 5.06, + "learning_rate": 1.4841099381143066e-06, + "loss": 0.0168, + "step": 27811 + }, + { + "epoch": 5.06, + "learning_rate": 1.4840553330906444e-06, + "loss": 1.7097, + "step": 27812 + }, + { + "epoch": 5.06, + "learning_rate": 1.484000728066982e-06, + "loss": 0.0137, + "step": 27813 + }, + { + "epoch": 5.06, + "learning_rate": 1.48394612304332e-06, + "loss": 0.0435, + "step": 27814 + }, + { + "epoch": 5.06, + "learning_rate": 1.483891518019658e-06, + "loss": 0.3336, + "step": 27815 + }, + { + "epoch": 5.06, + "learning_rate": 1.4838369129959957e-06, + "loss": 1.9138, + "step": 27816 + }, + { + "epoch": 5.06, + "learning_rate": 1.4837823079723334e-06, + "loss": 0.5005, + "step": 27817 + }, + { + "epoch": 5.06, + "learning_rate": 1.4837277029486714e-06, + "loss": 1.2915, + "step": 27818 + }, + { + "epoch": 5.06, + "learning_rate": 1.4836730979250091e-06, + "loss": 0.2289, + "step": 27819 + }, + { + "epoch": 5.06, + "learning_rate": 1.4836184929013468e-06, + "loss": 1.5353, + "step": 27820 + }, + { + "epoch": 5.06, + "learning_rate": 1.4835638878776848e-06, + "loss": 0.0693, + "step": 27821 + }, + { + "epoch": 5.06, + "learning_rate": 1.4835092828540227e-06, + "loss": 1.0731, + "step": 27822 + }, + { + "epoch": 5.06, + "learning_rate": 1.4834546778303605e-06, + "loss": 0.6592, + "step": 27823 + }, + { + "epoch": 5.06, + "learning_rate": 1.4834000728066982e-06, + "loss": 0.1565, + "step": 27824 + }, + { + "epoch": 5.06, + "learning_rate": 1.4833454677830361e-06, + "loss": 0.0107, + "step": 27825 + }, + { + "epoch": 5.06, + "learning_rate": 1.4832908627593739e-06, + "loss": 1.4063, + "step": 27826 + }, + { + "epoch": 5.06, + "learning_rate": 1.4832362577357116e-06, + "loss": 1.1914, + "step": 27827 + }, + { + "epoch": 5.06, + "learning_rate": 1.4831816527120495e-06, + "loss": 0.1373, + "step": 27828 + }, + { + "epoch": 5.06, + "learning_rate": 1.4831270476883875e-06, + "loss": 0.0557, + "step": 27829 + }, + { + "epoch": 5.06, + "learning_rate": 1.4830724426647252e-06, + "loss": 0.0569, + "step": 27830 + }, + { + "epoch": 5.06, + "learning_rate": 1.483017837641063e-06, + "loss": 0.0572, + "step": 27831 + }, + { + "epoch": 5.06, + "learning_rate": 1.482963232617401e-06, + "loss": 0.01, + "step": 27832 + }, + { + "epoch": 5.06, + "learning_rate": 1.4829086275937386e-06, + "loss": 0.1763, + "step": 27833 + }, + { + "epoch": 5.06, + "learning_rate": 1.4828540225700764e-06, + "loss": 0.0443, + "step": 27834 + }, + { + "epoch": 5.06, + "learning_rate": 1.4827994175464143e-06, + "loss": 1.5977, + "step": 27835 + }, + { + "epoch": 5.06, + "learning_rate": 1.4827448125227522e-06, + "loss": 0.2806, + "step": 27836 + }, + { + "epoch": 5.06, + "learning_rate": 1.48269020749909e-06, + "loss": 0.5185, + "step": 27837 + }, + { + "epoch": 5.06, + "learning_rate": 1.4826356024754277e-06, + "loss": 0.013, + "step": 27838 + }, + { + "epoch": 5.06, + "learning_rate": 1.4825809974517657e-06, + "loss": 0.0188, + "step": 27839 + }, + { + "epoch": 5.06, + "learning_rate": 1.4825263924281034e-06, + "loss": 0.418, + "step": 27840 + }, + { + "epoch": 5.06, + "learning_rate": 1.4824717874044413e-06, + "loss": 1.0855, + "step": 27841 + }, + { + "epoch": 5.06, + "learning_rate": 1.482417182380779e-06, + "loss": 0.025, + "step": 27842 + }, + { + "epoch": 5.06, + "learning_rate": 1.482362577357117e-06, + "loss": 2.5679, + "step": 27843 + }, + { + "epoch": 5.06, + "learning_rate": 1.4823079723334547e-06, + "loss": 0.6812, + "step": 27844 + }, + { + "epoch": 5.06, + "learning_rate": 1.4822533673097925e-06, + "loss": 0.0165, + "step": 27845 + }, + { + "epoch": 5.06, + "learning_rate": 1.4821987622861304e-06, + "loss": 1.4531, + "step": 27846 + }, + { + "epoch": 5.06, + "learning_rate": 1.4821441572624681e-06, + "loss": 1.0171, + "step": 27847 + }, + { + "epoch": 5.06, + "learning_rate": 1.482089552238806e-06, + "loss": 0.2132, + "step": 27848 + }, + { + "epoch": 5.06, + "learning_rate": 1.4820349472151438e-06, + "loss": 1.2009, + "step": 27849 + }, + { + "epoch": 5.06, + "learning_rate": 1.4819803421914818e-06, + "loss": 0.4778, + "step": 27850 + }, + { + "epoch": 5.06, + "learning_rate": 1.4819257371678195e-06, + "loss": 0.6089, + "step": 27851 + }, + { + "epoch": 5.06, + "learning_rate": 1.4818711321441572e-06, + "loss": 0.7591, + "step": 27852 + }, + { + "epoch": 5.07, + "learning_rate": 1.4818165271204952e-06, + "loss": 0.6774, + "step": 27853 + }, + { + "epoch": 5.07, + "learning_rate": 1.481761922096833e-06, + "loss": 1.2411, + "step": 27854 + }, + { + "epoch": 5.07, + "learning_rate": 1.4817073170731708e-06, + "loss": 0.0945, + "step": 27855 + }, + { + "epoch": 5.07, + "learning_rate": 1.4816527120495086e-06, + "loss": 0.7079, + "step": 27856 + }, + { + "epoch": 5.07, + "learning_rate": 1.4815981070258465e-06, + "loss": 0.0068, + "step": 27857 + }, + { + "epoch": 5.07, + "learning_rate": 1.4815435020021842e-06, + "loss": 0.1124, + "step": 27858 + }, + { + "epoch": 5.07, + "learning_rate": 1.481488896978522e-06, + "loss": 1.4251, + "step": 27859 + }, + { + "epoch": 5.07, + "learning_rate": 1.48143429195486e-06, + "loss": 0.6904, + "step": 27860 + }, + { + "epoch": 5.07, + "learning_rate": 1.4813796869311977e-06, + "loss": 0.3369, + "step": 27861 + }, + { + "epoch": 5.07, + "learning_rate": 1.4813250819075356e-06, + "loss": 0.1118, + "step": 27862 + }, + { + "epoch": 5.07, + "learning_rate": 1.4812704768838733e-06, + "loss": 0.2683, + "step": 27863 + }, + { + "epoch": 5.07, + "learning_rate": 1.4812158718602113e-06, + "loss": 1.0167, + "step": 27864 + }, + { + "epoch": 5.07, + "learning_rate": 1.481161266836549e-06, + "loss": 0.6751, + "step": 27865 + }, + { + "epoch": 5.07, + "learning_rate": 1.4811066618128867e-06, + "loss": 0.2386, + "step": 27866 + }, + { + "epoch": 5.07, + "learning_rate": 1.4810520567892247e-06, + "loss": 0.1921, + "step": 27867 + }, + { + "epoch": 5.07, + "learning_rate": 1.4809974517655624e-06, + "loss": 0.0731, + "step": 27868 + }, + { + "epoch": 5.07, + "learning_rate": 1.4809428467419004e-06, + "loss": 0.0199, + "step": 27869 + }, + { + "epoch": 5.07, + "learning_rate": 1.480888241718238e-06, + "loss": 0.052, + "step": 27870 + }, + { + "epoch": 5.07, + "learning_rate": 1.480833636694576e-06, + "loss": 0.5, + "step": 27871 + }, + { + "epoch": 5.07, + "learning_rate": 1.4807790316709138e-06, + "loss": 0.3065, + "step": 27872 + }, + { + "epoch": 5.07, + "learning_rate": 1.4807244266472515e-06, + "loss": 0.5118, + "step": 27873 + }, + { + "epoch": 5.07, + "learning_rate": 1.4806698216235892e-06, + "loss": 1.1095, + "step": 27874 + }, + { + "epoch": 5.07, + "learning_rate": 1.4806152165999272e-06, + "loss": 0.0178, + "step": 27875 + }, + { + "epoch": 5.07, + "learning_rate": 1.4805606115762651e-06, + "loss": 0.0983, + "step": 27876 + }, + { + "epoch": 5.07, + "learning_rate": 1.4805060065526028e-06, + "loss": 0.82, + "step": 27877 + }, + { + "epoch": 5.07, + "learning_rate": 1.4804514015289408e-06, + "loss": 0.2114, + "step": 27878 + }, + { + "epoch": 5.07, + "learning_rate": 1.4803967965052785e-06, + "loss": 0.4501, + "step": 27879 + }, + { + "epoch": 5.07, + "learning_rate": 1.4803421914816163e-06, + "loss": 1.2444, + "step": 27880 + }, + { + "epoch": 5.07, + "learning_rate": 1.4802875864579542e-06, + "loss": 0.7812, + "step": 27881 + }, + { + "epoch": 5.07, + "learning_rate": 1.4802329814342921e-06, + "loss": 0.8182, + "step": 27882 + }, + { + "epoch": 5.07, + "learning_rate": 1.4801783764106299e-06, + "loss": 3.0618, + "step": 27883 + }, + { + "epoch": 5.07, + "learning_rate": 1.4801237713869676e-06, + "loss": 0.6728, + "step": 27884 + }, + { + "epoch": 5.07, + "learning_rate": 1.4800691663633055e-06, + "loss": 0.0068, + "step": 27885 + }, + { + "epoch": 5.07, + "learning_rate": 1.4800145613396433e-06, + "loss": 0.0217, + "step": 27886 + }, + { + "epoch": 5.07, + "learning_rate": 1.479959956315981e-06, + "loss": 1.9264, + "step": 27887 + }, + { + "epoch": 5.07, + "learning_rate": 1.479905351292319e-06, + "loss": 0.1131, + "step": 27888 + }, + { + "epoch": 5.07, + "learning_rate": 1.4798507462686569e-06, + "loss": 0.987, + "step": 27889 + }, + { + "epoch": 5.07, + "learning_rate": 1.4797961412449946e-06, + "loss": 2.1466, + "step": 27890 + }, + { + "epoch": 5.07, + "learning_rate": 1.4797415362213324e-06, + "loss": 0.9794, + "step": 27891 + }, + { + "epoch": 5.07, + "learning_rate": 1.4796869311976703e-06, + "loss": 0.0229, + "step": 27892 + }, + { + "epoch": 5.07, + "learning_rate": 1.479632326174008e-06, + "loss": 1.4181, + "step": 27893 + }, + { + "epoch": 5.07, + "learning_rate": 1.4795777211503458e-06, + "loss": 1.3071, + "step": 27894 + }, + { + "epoch": 5.07, + "learning_rate": 1.4795231161266837e-06, + "loss": 0.36, + "step": 27895 + }, + { + "epoch": 5.07, + "learning_rate": 1.4794685111030217e-06, + "loss": 1.3419, + "step": 27896 + }, + { + "epoch": 5.07, + "learning_rate": 1.4794139060793594e-06, + "loss": 0.594, + "step": 27897 + }, + { + "epoch": 5.07, + "learning_rate": 1.4793593010556971e-06, + "loss": 0.1545, + "step": 27898 + }, + { + "epoch": 5.07, + "learning_rate": 1.479304696032035e-06, + "loss": 0.3566, + "step": 27899 + }, + { + "epoch": 5.07, + "learning_rate": 1.4792500910083728e-06, + "loss": 0.4171, + "step": 27900 + }, + { + "epoch": 5.07, + "learning_rate": 1.4791954859847105e-06, + "loss": 2.0085, + "step": 27901 + }, + { + "epoch": 5.07, + "learning_rate": 1.4791408809610485e-06, + "loss": 1.4067, + "step": 27902 + }, + { + "epoch": 5.07, + "learning_rate": 1.4790862759373864e-06, + "loss": 4.1306, + "step": 27903 + }, + { + "epoch": 5.07, + "learning_rate": 1.4790316709137241e-06, + "loss": 0.0108, + "step": 27904 + }, + { + "epoch": 5.07, + "learning_rate": 1.4789770658900619e-06, + "loss": 0.0145, + "step": 27905 + }, + { + "epoch": 5.07, + "learning_rate": 1.4789224608663998e-06, + "loss": 1.4581, + "step": 27906 + }, + { + "epoch": 5.07, + "learning_rate": 1.4788678558427375e-06, + "loss": 0.0595, + "step": 27907 + }, + { + "epoch": 5.08, + "learning_rate": 1.4788132508190753e-06, + "loss": 0.4445, + "step": 27908 + }, + { + "epoch": 5.08, + "learning_rate": 1.4787586457954132e-06, + "loss": 0.2026, + "step": 27909 + }, + { + "epoch": 5.08, + "learning_rate": 1.4787040407717512e-06, + "loss": 0.5185, + "step": 27910 + }, + { + "epoch": 5.08, + "learning_rate": 1.478649435748089e-06, + "loss": 0.2211, + "step": 27911 + }, + { + "epoch": 5.08, + "learning_rate": 1.4785948307244266e-06, + "loss": 0.7094, + "step": 27912 + }, + { + "epoch": 5.08, + "learning_rate": 1.4785402257007646e-06, + "loss": 0.0071, + "step": 27913 + }, + { + "epoch": 5.08, + "learning_rate": 1.4784856206771023e-06, + "loss": 1.292, + "step": 27914 + }, + { + "epoch": 5.08, + "learning_rate": 1.47843101565344e-06, + "loss": 0.1007, + "step": 27915 + }, + { + "epoch": 5.08, + "learning_rate": 1.478376410629778e-06, + "loss": 0.0416, + "step": 27916 + }, + { + "epoch": 5.08, + "learning_rate": 1.478321805606116e-06, + "loss": 0.4544, + "step": 27917 + }, + { + "epoch": 5.08, + "learning_rate": 1.4782672005824537e-06, + "loss": 0.0922, + "step": 27918 + }, + { + "epoch": 5.08, + "learning_rate": 1.4782125955587914e-06, + "loss": 0.3184, + "step": 27919 + }, + { + "epoch": 5.08, + "learning_rate": 1.4781579905351293e-06, + "loss": 0.1696, + "step": 27920 + }, + { + "epoch": 5.08, + "learning_rate": 1.478103385511467e-06, + "loss": 0.0185, + "step": 27921 + }, + { + "epoch": 5.08, + "learning_rate": 1.4780487804878048e-06, + "loss": 0.6154, + "step": 27922 + }, + { + "epoch": 5.08, + "learning_rate": 1.4779941754641427e-06, + "loss": 0.0228, + "step": 27923 + }, + { + "epoch": 5.08, + "learning_rate": 1.4779395704404807e-06, + "loss": 0.2246, + "step": 27924 + }, + { + "epoch": 5.08, + "learning_rate": 1.4778849654168184e-06, + "loss": 1.4091, + "step": 27925 + }, + { + "epoch": 5.08, + "learning_rate": 1.4778303603931561e-06, + "loss": 0.3152, + "step": 27926 + }, + { + "epoch": 5.08, + "learning_rate": 1.477775755369494e-06, + "loss": 0.1406, + "step": 27927 + }, + { + "epoch": 5.08, + "learning_rate": 1.4777211503458318e-06, + "loss": 1.0446, + "step": 27928 + }, + { + "epoch": 5.08, + "learning_rate": 1.4776665453221698e-06, + "loss": 0.0238, + "step": 27929 + }, + { + "epoch": 5.08, + "learning_rate": 1.4776119402985075e-06, + "loss": 0.3968, + "step": 27930 + }, + { + "epoch": 5.08, + "learning_rate": 1.4775573352748454e-06, + "loss": 2.2961, + "step": 27931 + }, + { + "epoch": 5.08, + "learning_rate": 1.4775027302511832e-06, + "loss": 0.3248, + "step": 27932 + }, + { + "epoch": 5.08, + "learning_rate": 1.477448125227521e-06, + "loss": 0.0182, + "step": 27933 + }, + { + "epoch": 5.08, + "learning_rate": 1.4773935202038588e-06, + "loss": 0.0211, + "step": 27934 + }, + { + "epoch": 5.08, + "learning_rate": 1.4773389151801966e-06, + "loss": 0.193, + "step": 27935 + }, + { + "epoch": 5.08, + "learning_rate": 1.4772843101565345e-06, + "loss": 0.0099, + "step": 27936 + }, + { + "epoch": 5.08, + "learning_rate": 1.4772297051328722e-06, + "loss": 0.0558, + "step": 27937 + }, + { + "epoch": 5.08, + "learning_rate": 1.4771751001092102e-06, + "loss": 1.9252, + "step": 27938 + }, + { + "epoch": 5.08, + "learning_rate": 1.477120495085548e-06, + "loss": 0.0173, + "step": 27939 + }, + { + "epoch": 5.08, + "learning_rate": 1.4770658900618857e-06, + "loss": 0.4434, + "step": 27940 + }, + { + "epoch": 5.08, + "learning_rate": 1.4770112850382236e-06, + "loss": 0.0138, + "step": 27941 + }, + { + "epoch": 5.08, + "learning_rate": 1.4769566800145613e-06, + "loss": 0.2293, + "step": 27942 + }, + { + "epoch": 5.08, + "learning_rate": 1.4769020749908993e-06, + "loss": 0.1055, + "step": 27943 + }, + { + "epoch": 5.08, + "learning_rate": 1.476847469967237e-06, + "loss": 0.4354, + "step": 27944 + }, + { + "epoch": 5.08, + "learning_rate": 1.476792864943575e-06, + "loss": 0.5324, + "step": 27945 + }, + { + "epoch": 5.08, + "learning_rate": 1.4767382599199127e-06, + "loss": 1.2643, + "step": 27946 + }, + { + "epoch": 5.08, + "learning_rate": 1.4766836548962504e-06, + "loss": 0.6673, + "step": 27947 + }, + { + "epoch": 5.08, + "learning_rate": 1.4766290498725881e-06, + "loss": 0.0378, + "step": 27948 + }, + { + "epoch": 5.08, + "learning_rate": 1.476574444848926e-06, + "loss": 0.5271, + "step": 27949 + }, + { + "epoch": 5.08, + "learning_rate": 1.476519839825264e-06, + "loss": 0.8135, + "step": 27950 + }, + { + "epoch": 5.08, + "learning_rate": 1.4764652348016018e-06, + "loss": 0.4262, + "step": 27951 + }, + { + "epoch": 5.08, + "learning_rate": 1.4764106297779397e-06, + "loss": 1.7223, + "step": 27952 + }, + { + "epoch": 5.08, + "learning_rate": 1.4763560247542774e-06, + "loss": 1.7824, + "step": 27953 + }, + { + "epoch": 5.08, + "learning_rate": 1.4763014197306152e-06, + "loss": 0.4255, + "step": 27954 + }, + { + "epoch": 5.08, + "learning_rate": 1.476246814706953e-06, + "loss": 1.1082, + "step": 27955 + }, + { + "epoch": 5.08, + "learning_rate": 1.4761922096832908e-06, + "loss": 0.1273, + "step": 27956 + }, + { + "epoch": 5.08, + "learning_rate": 1.4761376046596288e-06, + "loss": 0.0115, + "step": 27957 + }, + { + "epoch": 5.08, + "learning_rate": 1.4760829996359665e-06, + "loss": 0.1065, + "step": 27958 + }, + { + "epoch": 5.08, + "learning_rate": 1.4760283946123045e-06, + "loss": 0.2566, + "step": 27959 + }, + { + "epoch": 5.08, + "learning_rate": 1.4759737895886422e-06, + "loss": 0.1558, + "step": 27960 + }, + { + "epoch": 5.08, + "learning_rate": 1.47591918456498e-06, + "loss": 0.8656, + "step": 27961 + }, + { + "epoch": 5.08, + "learning_rate": 1.4758645795413177e-06, + "loss": 0.874, + "step": 27962 + }, + { + "epoch": 5.09, + "learning_rate": 1.4758099745176556e-06, + "loss": 2.3089, + "step": 27963 + }, + { + "epoch": 5.09, + "learning_rate": 1.4757553694939935e-06, + "loss": 2.4942, + "step": 27964 + }, + { + "epoch": 5.09, + "learning_rate": 1.4757007644703313e-06, + "loss": 0.7205, + "step": 27965 + }, + { + "epoch": 5.09, + "learning_rate": 1.4756461594466692e-06, + "loss": 0.2628, + "step": 27966 + }, + { + "epoch": 5.09, + "learning_rate": 1.475591554423007e-06, + "loss": 0.1594, + "step": 27967 + }, + { + "epoch": 5.09, + "learning_rate": 1.4755369493993447e-06, + "loss": 1.2796, + "step": 27968 + }, + { + "epoch": 5.09, + "learning_rate": 1.4754823443756826e-06, + "loss": 0.0535, + "step": 27969 + }, + { + "epoch": 5.09, + "learning_rate": 1.4754277393520206e-06, + "loss": 1.534, + "step": 27970 + }, + { + "epoch": 5.09, + "learning_rate": 1.4753731343283583e-06, + "loss": 0.0714, + "step": 27971 + }, + { + "epoch": 5.09, + "learning_rate": 1.475318529304696e-06, + "loss": 0.0724, + "step": 27972 + }, + { + "epoch": 5.09, + "learning_rate": 1.475263924281034e-06, + "loss": 0.3983, + "step": 27973 + }, + { + "epoch": 5.09, + "learning_rate": 1.4752093192573717e-06, + "loss": 1.0757, + "step": 27974 + }, + { + "epoch": 5.09, + "learning_rate": 1.4751547142337094e-06, + "loss": 0.1864, + "step": 27975 + }, + { + "epoch": 5.09, + "learning_rate": 1.4751001092100474e-06, + "loss": 0.0089, + "step": 27976 + }, + { + "epoch": 5.09, + "learning_rate": 1.4750455041863853e-06, + "loss": 0.4876, + "step": 27977 + }, + { + "epoch": 5.09, + "learning_rate": 1.474990899162723e-06, + "loss": 0.5782, + "step": 27978 + }, + { + "epoch": 5.09, + "learning_rate": 1.4749362941390608e-06, + "loss": 0.641, + "step": 27979 + }, + { + "epoch": 5.09, + "learning_rate": 1.4748816891153987e-06, + "loss": 1.3936, + "step": 27980 + }, + { + "epoch": 5.09, + "learning_rate": 1.4748270840917365e-06, + "loss": 0.3491, + "step": 27981 + }, + { + "epoch": 5.09, + "learning_rate": 1.4747724790680742e-06, + "loss": 0.3903, + "step": 27982 + }, + { + "epoch": 5.09, + "learning_rate": 1.4747178740444121e-06, + "loss": 0.3522, + "step": 27983 + }, + { + "epoch": 5.09, + "learning_rate": 1.47466326902075e-06, + "loss": 0.8604, + "step": 27984 + }, + { + "epoch": 5.09, + "learning_rate": 1.4746086639970878e-06, + "loss": 0.0996, + "step": 27985 + }, + { + "epoch": 5.09, + "learning_rate": 1.4745540589734255e-06, + "loss": 0.0402, + "step": 27986 + }, + { + "epoch": 5.09, + "learning_rate": 1.4744994539497635e-06, + "loss": 0.593, + "step": 27987 + }, + { + "epoch": 5.09, + "learning_rate": 1.4744448489261012e-06, + "loss": 0.5275, + "step": 27988 + }, + { + "epoch": 5.09, + "learning_rate": 1.474390243902439e-06, + "loss": 0.2035, + "step": 27989 + }, + { + "epoch": 5.09, + "learning_rate": 1.474335638878777e-06, + "loss": 0.0301, + "step": 27990 + }, + { + "epoch": 5.09, + "learning_rate": 1.4742810338551148e-06, + "loss": 0.0471, + "step": 27991 + }, + { + "epoch": 5.09, + "learning_rate": 1.4742264288314526e-06, + "loss": 1.6443, + "step": 27992 + }, + { + "epoch": 5.09, + "learning_rate": 1.4741718238077903e-06, + "loss": 3.0691, + "step": 27993 + }, + { + "epoch": 5.09, + "learning_rate": 1.4741172187841282e-06, + "loss": 0.7992, + "step": 27994 + }, + { + "epoch": 5.09, + "learning_rate": 1.474062613760466e-06, + "loss": 1.5989, + "step": 27995 + }, + { + "epoch": 5.09, + "learning_rate": 1.4740080087368037e-06, + "loss": 0.0108, + "step": 27996 + }, + { + "epoch": 5.09, + "learning_rate": 1.4739534037131417e-06, + "loss": 0.2945, + "step": 27997 + }, + { + "epoch": 5.09, + "learning_rate": 1.4738987986894796e-06, + "loss": 0.4026, + "step": 27998 + }, + { + "epoch": 5.09, + "learning_rate": 1.4738441936658173e-06, + "loss": 0.1108, + "step": 27999 + }, + { + "epoch": 5.09, + "learning_rate": 1.473789588642155e-06, + "loss": 0.5919, + "step": 28000 + }, + { + "epoch": 5.09, + "learning_rate": 1.473734983618493e-06, + "loss": 1.281, + "step": 28001 + }, + { + "epoch": 5.09, + "learning_rate": 1.4736803785948307e-06, + "loss": 1.267, + "step": 28002 + }, + { + "epoch": 5.09, + "learning_rate": 1.4736257735711685e-06, + "loss": 0.3472, + "step": 28003 + }, + { + "epoch": 5.09, + "learning_rate": 1.4735711685475064e-06, + "loss": 0.3617, + "step": 28004 + }, + { + "epoch": 5.09, + "learning_rate": 1.4735165635238444e-06, + "loss": 0.9854, + "step": 28005 + }, + { + "epoch": 5.09, + "learning_rate": 1.473461958500182e-06, + "loss": 0.2486, + "step": 28006 + }, + { + "epoch": 5.09, + "learning_rate": 1.4734073534765198e-06, + "loss": 0.1624, + "step": 28007 + }, + { + "epoch": 5.09, + "learning_rate": 1.4733527484528578e-06, + "loss": 0.285, + "step": 28008 + }, + { + "epoch": 5.09, + "learning_rate": 1.4732981434291955e-06, + "loss": 0.1214, + "step": 28009 + }, + { + "epoch": 5.09, + "learning_rate": 1.4732435384055334e-06, + "loss": 0.1352, + "step": 28010 + }, + { + "epoch": 5.09, + "learning_rate": 1.4731889333818712e-06, + "loss": 2.3102, + "step": 28011 + }, + { + "epoch": 5.09, + "learning_rate": 1.4731343283582091e-06, + "loss": 0.0045, + "step": 28012 + }, + { + "epoch": 5.09, + "learning_rate": 1.4730797233345468e-06, + "loss": 0.777, + "step": 28013 + }, + { + "epoch": 5.09, + "learning_rate": 1.4730251183108846e-06, + "loss": 1.4837, + "step": 28014 + }, + { + "epoch": 5.09, + "learning_rate": 1.4729705132872225e-06, + "loss": 1.4668, + "step": 28015 + }, + { + "epoch": 5.09, + "learning_rate": 1.4729159082635602e-06, + "loss": 0.9661, + "step": 28016 + }, + { + "epoch": 5.09, + "learning_rate": 1.4728613032398982e-06, + "loss": 0.5581, + "step": 28017 + }, + { + "epoch": 5.1, + "learning_rate": 1.472806698216236e-06, + "loss": 0.259, + "step": 28018 + }, + { + "epoch": 5.1, + "learning_rate": 1.4727520931925739e-06, + "loss": 1.6663, + "step": 28019 + }, + { + "epoch": 5.1, + "learning_rate": 1.4726974881689116e-06, + "loss": 0.6156, + "step": 28020 + }, + { + "epoch": 5.1, + "learning_rate": 1.4726428831452493e-06, + "loss": 0.7062, + "step": 28021 + }, + { + "epoch": 5.1, + "learning_rate": 1.4725882781215873e-06, + "loss": 0.1617, + "step": 28022 + }, + { + "epoch": 5.1, + "learning_rate": 1.472533673097925e-06, + "loss": 0.1825, + "step": 28023 + }, + { + "epoch": 5.1, + "learning_rate": 1.472479068074263e-06, + "loss": 0.324, + "step": 28024 + }, + { + "epoch": 5.1, + "learning_rate": 1.4724244630506007e-06, + "loss": 2.4087, + "step": 28025 + }, + { + "epoch": 5.1, + "learning_rate": 1.4723698580269386e-06, + "loss": 1.7217, + "step": 28026 + }, + { + "epoch": 5.1, + "learning_rate": 1.4723152530032764e-06, + "loss": 0.8149, + "step": 28027 + }, + { + "epoch": 5.1, + "learning_rate": 1.472260647979614e-06, + "loss": 0.3933, + "step": 28028 + }, + { + "epoch": 5.1, + "learning_rate": 1.4722060429559518e-06, + "loss": 1.0463, + "step": 28029 + }, + { + "epoch": 5.1, + "learning_rate": 1.4721514379322898e-06, + "loss": 0.3677, + "step": 28030 + }, + { + "epoch": 5.1, + "learning_rate": 1.4720968329086277e-06, + "loss": 0.0269, + "step": 28031 + }, + { + "epoch": 5.1, + "learning_rate": 1.4720422278849654e-06, + "loss": 0.1282, + "step": 28032 + }, + { + "epoch": 5.1, + "learning_rate": 1.4719876228613034e-06, + "loss": 0.8768, + "step": 28033 + }, + { + "epoch": 5.1, + "learning_rate": 1.4719330178376411e-06, + "loss": 0.014, + "step": 28034 + }, + { + "epoch": 5.1, + "learning_rate": 1.4718784128139788e-06, + "loss": 0.3434, + "step": 28035 + }, + { + "epoch": 5.1, + "learning_rate": 1.4718238077903166e-06, + "loss": 0.0054, + "step": 28036 + }, + { + "epoch": 5.1, + "learning_rate": 1.4717692027666545e-06, + "loss": 0.4552, + "step": 28037 + }, + { + "epoch": 5.1, + "learning_rate": 1.4717145977429925e-06, + "loss": 0.108, + "step": 28038 + }, + { + "epoch": 5.1, + "learning_rate": 1.4716599927193302e-06, + "loss": 0.1921, + "step": 28039 + }, + { + "epoch": 5.1, + "learning_rate": 1.4716053876956681e-06, + "loss": 0.0151, + "step": 28040 + }, + { + "epoch": 5.1, + "learning_rate": 1.4715507826720059e-06, + "loss": 0.1574, + "step": 28041 + }, + { + "epoch": 5.1, + "learning_rate": 1.4714961776483436e-06, + "loss": 0.1862, + "step": 28042 + }, + { + "epoch": 5.1, + "learning_rate": 1.4714415726246813e-06, + "loss": 1.8449, + "step": 28043 + }, + { + "epoch": 5.1, + "learning_rate": 1.4713869676010193e-06, + "loss": 0.5038, + "step": 28044 + }, + { + "epoch": 5.1, + "learning_rate": 1.4713323625773572e-06, + "loss": 0.371, + "step": 28045 + }, + { + "epoch": 5.1, + "learning_rate": 1.471277757553695e-06, + "loss": 1.1429, + "step": 28046 + }, + { + "epoch": 5.1, + "learning_rate": 1.4712231525300329e-06, + "loss": 0.6023, + "step": 28047 + }, + { + "epoch": 5.1, + "learning_rate": 1.4711685475063706e-06, + "loss": 0.2726, + "step": 28048 + }, + { + "epoch": 5.1, + "learning_rate": 1.4711139424827084e-06, + "loss": 0.2928, + "step": 28049 + }, + { + "epoch": 5.1, + "learning_rate": 1.4710593374590463e-06, + "loss": 1.1261, + "step": 28050 + }, + { + "epoch": 5.1, + "learning_rate": 1.4710047324353842e-06, + "loss": 0.0592, + "step": 28051 + }, + { + "epoch": 5.1, + "learning_rate": 1.470950127411722e-06, + "loss": 0.2773, + "step": 28052 + }, + { + "epoch": 5.1, + "learning_rate": 1.4708955223880597e-06, + "loss": 0.713, + "step": 28053 + }, + { + "epoch": 5.1, + "learning_rate": 1.4708409173643976e-06, + "loss": 0.2503, + "step": 28054 + }, + { + "epoch": 5.1, + "learning_rate": 1.4707863123407354e-06, + "loss": 0.2688, + "step": 28055 + }, + { + "epoch": 5.1, + "learning_rate": 1.4707317073170731e-06, + "loss": 0.1613, + "step": 28056 + }, + { + "epoch": 5.1, + "learning_rate": 1.470677102293411e-06, + "loss": 1.4619, + "step": 28057 + }, + { + "epoch": 5.1, + "learning_rate": 1.470622497269749e-06, + "loss": 1.4621, + "step": 28058 + }, + { + "epoch": 5.1, + "learning_rate": 1.4705678922460867e-06, + "loss": 1.3293, + "step": 28059 + }, + { + "epoch": 5.1, + "learning_rate": 1.4705132872224245e-06, + "loss": 1.2037, + "step": 28060 + }, + { + "epoch": 5.1, + "learning_rate": 1.4704586821987624e-06, + "loss": 0.1999, + "step": 28061 + }, + { + "epoch": 5.1, + "learning_rate": 1.4704040771751001e-06, + "loss": 2.6044, + "step": 28062 + }, + { + "epoch": 5.1, + "learning_rate": 1.4703494721514379e-06, + "loss": 0.3218, + "step": 28063 + }, + { + "epoch": 5.1, + "learning_rate": 1.4702948671277758e-06, + "loss": 0.8019, + "step": 28064 + }, + { + "epoch": 5.1, + "learning_rate": 1.4702402621041138e-06, + "loss": 0.0999, + "step": 28065 + }, + { + "epoch": 5.1, + "learning_rate": 1.4701856570804515e-06, + "loss": 0.3072, + "step": 28066 + }, + { + "epoch": 5.1, + "learning_rate": 1.4701310520567892e-06, + "loss": 1.3674, + "step": 28067 + }, + { + "epoch": 5.1, + "learning_rate": 1.4700764470331272e-06, + "loss": 0.0094, + "step": 28068 + }, + { + "epoch": 5.1, + "learning_rate": 1.4700218420094649e-06, + "loss": 0.9805, + "step": 28069 + }, + { + "epoch": 5.1, + "learning_rate": 1.4699672369858026e-06, + "loss": 1.1666, + "step": 28070 + }, + { + "epoch": 5.1, + "learning_rate": 1.4699126319621406e-06, + "loss": 3.9385, + "step": 28071 + }, + { + "epoch": 5.1, + "learning_rate": 1.4698580269384785e-06, + "loss": 1.1459, + "step": 28072 + }, + { + "epoch": 5.11, + "learning_rate": 1.4698034219148162e-06, + "loss": 0.1761, + "step": 28073 + }, + { + "epoch": 5.11, + "learning_rate": 1.469748816891154e-06, + "loss": 0.7739, + "step": 28074 + }, + { + "epoch": 5.11, + "learning_rate": 1.469694211867492e-06, + "loss": 0.8022, + "step": 28075 + }, + { + "epoch": 5.11, + "learning_rate": 1.4696396068438297e-06, + "loss": 0.1804, + "step": 28076 + }, + { + "epoch": 5.11, + "learning_rate": 1.4695850018201674e-06, + "loss": 1.2677, + "step": 28077 + }, + { + "epoch": 5.11, + "learning_rate": 1.4695303967965053e-06, + "loss": 1.31, + "step": 28078 + }, + { + "epoch": 5.11, + "learning_rate": 1.4694757917728433e-06, + "loss": 0.3272, + "step": 28079 + }, + { + "epoch": 5.11, + "learning_rate": 1.469421186749181e-06, + "loss": 0.2994, + "step": 28080 + }, + { + "epoch": 5.11, + "learning_rate": 1.4693665817255187e-06, + "loss": 0.3122, + "step": 28081 + }, + { + "epoch": 5.11, + "learning_rate": 1.4693119767018567e-06, + "loss": 0.2592, + "step": 28082 + }, + { + "epoch": 5.11, + "learning_rate": 1.4692573716781944e-06, + "loss": 0.47, + "step": 28083 + }, + { + "epoch": 5.11, + "learning_rate": 1.4692027666545321e-06, + "loss": 0.2181, + "step": 28084 + }, + { + "epoch": 5.11, + "learning_rate": 1.46914816163087e-06, + "loss": 0.5909, + "step": 28085 + }, + { + "epoch": 5.11, + "learning_rate": 1.469093556607208e-06, + "loss": 0.1297, + "step": 28086 + }, + { + "epoch": 5.11, + "learning_rate": 1.4690389515835458e-06, + "loss": 1.91, + "step": 28087 + }, + { + "epoch": 5.11, + "learning_rate": 1.4689843465598835e-06, + "loss": 2.9456, + "step": 28088 + }, + { + "epoch": 5.11, + "learning_rate": 1.4689297415362214e-06, + "loss": 0.7692, + "step": 28089 + }, + { + "epoch": 5.11, + "learning_rate": 1.4688751365125592e-06, + "loss": 0.9777, + "step": 28090 + }, + { + "epoch": 5.11, + "learning_rate": 1.468820531488897e-06, + "loss": 0.0202, + "step": 28091 + }, + { + "epoch": 5.11, + "learning_rate": 1.4687659264652348e-06, + "loss": 0.2933, + "step": 28092 + }, + { + "epoch": 5.11, + "learning_rate": 1.4687113214415728e-06, + "loss": 0.2904, + "step": 28093 + }, + { + "epoch": 5.11, + "learning_rate": 1.4686567164179105e-06, + "loss": 0.5014, + "step": 28094 + }, + { + "epoch": 5.11, + "learning_rate": 1.4686021113942482e-06, + "loss": 0.3191, + "step": 28095 + }, + { + "epoch": 5.11, + "learning_rate": 1.4685475063705862e-06, + "loss": 1.928, + "step": 28096 + }, + { + "epoch": 5.11, + "learning_rate": 1.468492901346924e-06, + "loss": 1.4336, + "step": 28097 + }, + { + "epoch": 5.11, + "learning_rate": 1.4684382963232619e-06, + "loss": 0.0117, + "step": 28098 + }, + { + "epoch": 5.11, + "learning_rate": 1.4683836912995996e-06, + "loss": 0.1532, + "step": 28099 + }, + { + "epoch": 5.11, + "learning_rate": 1.4683290862759375e-06, + "loss": 0.8429, + "step": 28100 + }, + { + "epoch": 5.11, + "learning_rate": 1.4682744812522753e-06, + "loss": 0.9028, + "step": 28101 + }, + { + "epoch": 5.11, + "learning_rate": 1.468219876228613e-06, + "loss": 1.2043, + "step": 28102 + }, + { + "epoch": 5.11, + "learning_rate": 1.4681652712049507e-06, + "loss": 1.0344, + "step": 28103 + }, + { + "epoch": 5.11, + "learning_rate": 1.4681106661812887e-06, + "loss": 0.8577, + "step": 28104 + }, + { + "epoch": 5.11, + "learning_rate": 1.4680560611576266e-06, + "loss": 0.0554, + "step": 28105 + }, + { + "epoch": 5.11, + "learning_rate": 1.4680014561339644e-06, + "loss": 0.6083, + "step": 28106 + }, + { + "epoch": 5.11, + "learning_rate": 1.4679468511103023e-06, + "loss": 0.7057, + "step": 28107 + }, + { + "epoch": 5.11, + "learning_rate": 1.46789224608664e-06, + "loss": 2.6556, + "step": 28108 + }, + { + "epoch": 5.11, + "learning_rate": 1.4678376410629778e-06, + "loss": 0.8241, + "step": 28109 + }, + { + "epoch": 5.11, + "learning_rate": 1.4677830360393155e-06, + "loss": 0.0085, + "step": 28110 + }, + { + "epoch": 5.11, + "learning_rate": 1.4677284310156534e-06, + "loss": 0.0124, + "step": 28111 + }, + { + "epoch": 5.11, + "learning_rate": 1.4676738259919914e-06, + "loss": 0.2317, + "step": 28112 + }, + { + "epoch": 5.11, + "learning_rate": 1.4676192209683291e-06, + "loss": 0.1657, + "step": 28113 + }, + { + "epoch": 5.11, + "learning_rate": 1.467564615944667e-06, + "loss": 0.0237, + "step": 28114 + }, + { + "epoch": 5.11, + "learning_rate": 1.4675100109210048e-06, + "loss": 0.523, + "step": 28115 + }, + { + "epoch": 5.11, + "learning_rate": 1.4674554058973425e-06, + "loss": 1.321, + "step": 28116 + }, + { + "epoch": 5.11, + "learning_rate": 1.4674008008736802e-06, + "loss": 0.7831, + "step": 28117 + }, + { + "epoch": 5.11, + "learning_rate": 1.4673461958500182e-06, + "loss": 0.0649, + "step": 28118 + }, + { + "epoch": 5.11, + "learning_rate": 1.4672915908263561e-06, + "loss": 0.7419, + "step": 28119 + }, + { + "epoch": 5.11, + "learning_rate": 1.4672369858026939e-06, + "loss": 0.2571, + "step": 28120 + }, + { + "epoch": 5.11, + "learning_rate": 1.4671823807790318e-06, + "loss": 1.3336, + "step": 28121 + }, + { + "epoch": 5.11, + "learning_rate": 1.4671277757553695e-06, + "loss": 0.7483, + "step": 28122 + }, + { + "epoch": 5.11, + "learning_rate": 1.4670731707317073e-06, + "loss": 0.0118, + "step": 28123 + }, + { + "epoch": 5.11, + "learning_rate": 1.467018565708045e-06, + "loss": 2.0595, + "step": 28124 + }, + { + "epoch": 5.11, + "learning_rate": 1.466963960684383e-06, + "loss": 0.0662, + "step": 28125 + }, + { + "epoch": 5.11, + "learning_rate": 1.4669093556607209e-06, + "loss": 0.0252, + "step": 28126 + }, + { + "epoch": 5.11, + "learning_rate": 1.4668547506370586e-06, + "loss": 1.5999, + "step": 28127 + }, + { + "epoch": 5.12, + "learning_rate": 1.4668001456133966e-06, + "loss": 1.425, + "step": 28128 + }, + { + "epoch": 5.12, + "learning_rate": 1.4667455405897343e-06, + "loss": 0.6061, + "step": 28129 + }, + { + "epoch": 5.12, + "learning_rate": 1.466690935566072e-06, + "loss": 0.2385, + "step": 28130 + }, + { + "epoch": 5.12, + "learning_rate": 1.4666363305424098e-06, + "loss": 1.0131, + "step": 28131 + }, + { + "epoch": 5.12, + "learning_rate": 1.4665817255187477e-06, + "loss": 0.0102, + "step": 28132 + }, + { + "epoch": 5.12, + "learning_rate": 1.4665271204950856e-06, + "loss": 0.0087, + "step": 28133 + }, + { + "epoch": 5.12, + "learning_rate": 1.4664725154714234e-06, + "loss": 0.0137, + "step": 28134 + }, + { + "epoch": 5.12, + "learning_rate": 1.4664179104477613e-06, + "loss": 1.5501, + "step": 28135 + }, + { + "epoch": 5.12, + "learning_rate": 1.466363305424099e-06, + "loss": 2.0377, + "step": 28136 + }, + { + "epoch": 5.12, + "learning_rate": 1.4663087004004368e-06, + "loss": 0.0747, + "step": 28137 + }, + { + "epoch": 5.12, + "learning_rate": 1.4662540953767747e-06, + "loss": 0.0111, + "step": 28138 + }, + { + "epoch": 5.12, + "learning_rate": 1.4661994903531127e-06, + "loss": 0.2281, + "step": 28139 + }, + { + "epoch": 5.12, + "learning_rate": 1.4661448853294504e-06, + "loss": 0.3079, + "step": 28140 + }, + { + "epoch": 5.12, + "learning_rate": 1.4660902803057881e-06, + "loss": 0.0522, + "step": 28141 + }, + { + "epoch": 5.12, + "learning_rate": 1.466035675282126e-06, + "loss": 0.0094, + "step": 28142 + }, + { + "epoch": 5.12, + "learning_rate": 1.4659810702584638e-06, + "loss": 0.1299, + "step": 28143 + }, + { + "epoch": 5.12, + "learning_rate": 1.4659264652348015e-06, + "loss": 0.0105, + "step": 28144 + }, + { + "epoch": 5.12, + "learning_rate": 1.4658718602111395e-06, + "loss": 0.0557, + "step": 28145 + }, + { + "epoch": 5.12, + "learning_rate": 1.4658172551874774e-06, + "loss": 1.8704, + "step": 28146 + }, + { + "epoch": 5.12, + "learning_rate": 1.4657626501638152e-06, + "loss": 0.0074, + "step": 28147 + }, + { + "epoch": 5.12, + "learning_rate": 1.4657080451401529e-06, + "loss": 0.7814, + "step": 28148 + }, + { + "epoch": 5.12, + "learning_rate": 1.4656534401164908e-06, + "loss": 0.2605, + "step": 28149 + }, + { + "epoch": 5.12, + "learning_rate": 1.4655988350928286e-06, + "loss": 0.0281, + "step": 28150 + }, + { + "epoch": 5.12, + "learning_rate": 1.4655442300691663e-06, + "loss": 2.5047, + "step": 28151 + }, + { + "epoch": 5.12, + "learning_rate": 1.4654896250455042e-06, + "loss": 0.7016, + "step": 28152 + }, + { + "epoch": 5.12, + "learning_rate": 1.4654350200218422e-06, + "loss": 1.5706, + "step": 28153 + }, + { + "epoch": 5.12, + "learning_rate": 1.46538041499818e-06, + "loss": 0.3779, + "step": 28154 + }, + { + "epoch": 5.12, + "learning_rate": 1.4653258099745176e-06, + "loss": 0.578, + "step": 28155 + }, + { + "epoch": 5.12, + "learning_rate": 1.4652712049508556e-06, + "loss": 0.2369, + "step": 28156 + }, + { + "epoch": 5.12, + "learning_rate": 1.4652165999271933e-06, + "loss": 0.7066, + "step": 28157 + }, + { + "epoch": 5.12, + "learning_rate": 1.465161994903531e-06, + "loss": 0.0175, + "step": 28158 + }, + { + "epoch": 5.12, + "learning_rate": 1.465107389879869e-06, + "loss": 0.0144, + "step": 28159 + }, + { + "epoch": 5.12, + "learning_rate": 1.465052784856207e-06, + "loss": 0.1974, + "step": 28160 + }, + { + "epoch": 5.12, + "learning_rate": 1.4649981798325447e-06, + "loss": 0.0135, + "step": 28161 + }, + { + "epoch": 5.12, + "learning_rate": 1.4649435748088824e-06, + "loss": 0.5364, + "step": 28162 + }, + { + "epoch": 5.12, + "learning_rate": 1.4648889697852203e-06, + "loss": 0.2719, + "step": 28163 + }, + { + "epoch": 5.12, + "learning_rate": 1.464834364761558e-06, + "loss": 0.0095, + "step": 28164 + }, + { + "epoch": 5.12, + "learning_rate": 1.4647797597378958e-06, + "loss": 0.2082, + "step": 28165 + }, + { + "epoch": 5.12, + "learning_rate": 1.4647251547142338e-06, + "loss": 0.0205, + "step": 28166 + }, + { + "epoch": 5.12, + "learning_rate": 1.4646705496905717e-06, + "loss": 0.0644, + "step": 28167 + }, + { + "epoch": 5.12, + "learning_rate": 1.4646159446669094e-06, + "loss": 0.1696, + "step": 28168 + }, + { + "epoch": 5.12, + "learning_rate": 1.4645613396432472e-06, + "loss": 0.1911, + "step": 28169 + }, + { + "epoch": 5.12, + "learning_rate": 1.4645067346195851e-06, + "loss": 1.184, + "step": 28170 + }, + { + "epoch": 5.12, + "learning_rate": 1.4644521295959228e-06, + "loss": 2.0978, + "step": 28171 + }, + { + "epoch": 5.12, + "learning_rate": 1.4643975245722606e-06, + "loss": 0.0085, + "step": 28172 + }, + { + "epoch": 5.12, + "learning_rate": 1.4643429195485985e-06, + "loss": 1.2334, + "step": 28173 + }, + { + "epoch": 5.12, + "learning_rate": 1.4642883145249365e-06, + "loss": 2.7771, + "step": 28174 + }, + { + "epoch": 5.12, + "learning_rate": 1.4642337095012742e-06, + "loss": 1.2576, + "step": 28175 + }, + { + "epoch": 5.12, + "learning_rate": 1.464179104477612e-06, + "loss": 1.7242, + "step": 28176 + }, + { + "epoch": 5.12, + "learning_rate": 1.4641244994539499e-06, + "loss": 1.1749, + "step": 28177 + }, + { + "epoch": 5.12, + "learning_rate": 1.4640698944302876e-06, + "loss": 0.3653, + "step": 28178 + }, + { + "epoch": 5.12, + "learning_rate": 1.4640152894066255e-06, + "loss": 0.2081, + "step": 28179 + }, + { + "epoch": 5.12, + "learning_rate": 1.4639606843829633e-06, + "loss": 1.7425, + "step": 28180 + }, + { + "epoch": 5.12, + "learning_rate": 1.4639060793593012e-06, + "loss": 5.2795, + "step": 28181 + }, + { + "epoch": 5.12, + "learning_rate": 1.463851474335639e-06, + "loss": 0.5981, + "step": 28182 + }, + { + "epoch": 5.13, + "learning_rate": 1.4637968693119767e-06, + "loss": 0.0075, + "step": 28183 + }, + { + "epoch": 5.13, + "learning_rate": 1.4637422642883144e-06, + "loss": 0.325, + "step": 28184 + }, + { + "epoch": 5.13, + "learning_rate": 1.4636876592646524e-06, + "loss": 3.9257, + "step": 28185 + }, + { + "epoch": 5.13, + "learning_rate": 1.4636330542409903e-06, + "loss": 0.0598, + "step": 28186 + }, + { + "epoch": 5.13, + "learning_rate": 1.463578449217328e-06, + "loss": 0.7082, + "step": 28187 + }, + { + "epoch": 5.13, + "learning_rate": 1.463523844193666e-06, + "loss": 0.563, + "step": 28188 + }, + { + "epoch": 5.13, + "learning_rate": 1.4634692391700037e-06, + "loss": 0.0581, + "step": 28189 + }, + { + "epoch": 5.13, + "learning_rate": 1.4634146341463414e-06, + "loss": 0.2486, + "step": 28190 + }, + { + "epoch": 5.13, + "learning_rate": 1.4633600291226792e-06, + "loss": 0.3305, + "step": 28191 + }, + { + "epoch": 5.13, + "learning_rate": 1.4633054240990171e-06, + "loss": 0.0335, + "step": 28192 + }, + { + "epoch": 5.13, + "learning_rate": 1.463250819075355e-06, + "loss": 0.8318, + "step": 28193 + }, + { + "epoch": 5.13, + "learning_rate": 1.4631962140516928e-06, + "loss": 2.0971, + "step": 28194 + }, + { + "epoch": 5.13, + "learning_rate": 1.4631416090280307e-06, + "loss": 0.0786, + "step": 28195 + }, + { + "epoch": 5.13, + "learning_rate": 1.4630870040043685e-06, + "loss": 0.4556, + "step": 28196 + }, + { + "epoch": 5.13, + "learning_rate": 1.4630323989807062e-06, + "loss": 0.1305, + "step": 28197 + }, + { + "epoch": 5.13, + "learning_rate": 1.462977793957044e-06, + "loss": 0.3262, + "step": 28198 + }, + { + "epoch": 5.13, + "learning_rate": 1.4629231889333819e-06, + "loss": 0.5883, + "step": 28199 + }, + { + "epoch": 5.13, + "learning_rate": 1.4628685839097198e-06, + "loss": 0.225, + "step": 28200 + }, + { + "epoch": 5.13, + "learning_rate": 1.4628139788860575e-06, + "loss": 1.0274, + "step": 28201 + }, + { + "epoch": 5.13, + "learning_rate": 1.4627593738623955e-06, + "loss": 0.1956, + "step": 28202 + }, + { + "epoch": 5.13, + "learning_rate": 1.4627047688387332e-06, + "loss": 0.3741, + "step": 28203 + }, + { + "epoch": 5.13, + "learning_rate": 1.462650163815071e-06, + "loss": 1.1204, + "step": 28204 + }, + { + "epoch": 5.13, + "learning_rate": 1.4625955587914087e-06, + "loss": 0.014, + "step": 28205 + }, + { + "epoch": 5.13, + "learning_rate": 1.4625409537677466e-06, + "loss": 0.0502, + "step": 28206 + }, + { + "epoch": 5.13, + "learning_rate": 1.4624863487440846e-06, + "loss": 0.083, + "step": 28207 + }, + { + "epoch": 5.13, + "learning_rate": 1.4624317437204223e-06, + "loss": 1.8199, + "step": 28208 + }, + { + "epoch": 5.13, + "learning_rate": 1.4623771386967602e-06, + "loss": 0.8684, + "step": 28209 + }, + { + "epoch": 5.13, + "learning_rate": 1.462322533673098e-06, + "loss": 1.0514, + "step": 28210 + }, + { + "epoch": 5.13, + "learning_rate": 1.4622679286494357e-06, + "loss": 0.0646, + "step": 28211 + }, + { + "epoch": 5.13, + "learning_rate": 1.4622133236257734e-06, + "loss": 0.0154, + "step": 28212 + }, + { + "epoch": 5.13, + "learning_rate": 1.4621587186021114e-06, + "loss": 0.5276, + "step": 28213 + }, + { + "epoch": 5.13, + "learning_rate": 1.4621041135784493e-06, + "loss": 0.291, + "step": 28214 + }, + { + "epoch": 5.13, + "learning_rate": 1.462049508554787e-06, + "loss": 0.068, + "step": 28215 + }, + { + "epoch": 5.13, + "learning_rate": 1.461994903531125e-06, + "loss": 0.2136, + "step": 28216 + }, + { + "epoch": 5.13, + "learning_rate": 1.4619402985074627e-06, + "loss": 1.7517, + "step": 28217 + }, + { + "epoch": 5.13, + "learning_rate": 1.4618856934838005e-06, + "loss": 0.3607, + "step": 28218 + }, + { + "epoch": 5.13, + "learning_rate": 1.4618310884601384e-06, + "loss": 0.0813, + "step": 28219 + }, + { + "epoch": 5.13, + "learning_rate": 1.4617764834364761e-06, + "loss": 0.1984, + "step": 28220 + }, + { + "epoch": 5.13, + "learning_rate": 1.461721878412814e-06, + "loss": 1.5669, + "step": 28221 + }, + { + "epoch": 5.13, + "learning_rate": 1.4616672733891518e-06, + "loss": 0.0058, + "step": 28222 + }, + { + "epoch": 5.13, + "learning_rate": 1.4616126683654898e-06, + "loss": 0.0078, + "step": 28223 + }, + { + "epoch": 5.13, + "learning_rate": 1.4615580633418275e-06, + "loss": 0.5687, + "step": 28224 + }, + { + "epoch": 5.13, + "learning_rate": 1.4615034583181652e-06, + "loss": 0.1647, + "step": 28225 + }, + { + "epoch": 5.13, + "learning_rate": 1.4614488532945032e-06, + "loss": 2.0496, + "step": 28226 + }, + { + "epoch": 5.13, + "learning_rate": 1.461394248270841e-06, + "loss": 0.2748, + "step": 28227 + }, + { + "epoch": 5.13, + "learning_rate": 1.4613396432471788e-06, + "loss": 1.4559, + "step": 28228 + }, + { + "epoch": 5.13, + "learning_rate": 1.4612850382235166e-06, + "loss": 1.1349, + "step": 28229 + }, + { + "epoch": 5.13, + "learning_rate": 1.4612304331998545e-06, + "loss": 1.9396, + "step": 28230 + }, + { + "epoch": 5.13, + "learning_rate": 1.4611758281761922e-06, + "loss": 0.012, + "step": 28231 + }, + { + "epoch": 5.13, + "learning_rate": 1.46112122315253e-06, + "loss": 1.4378, + "step": 28232 + }, + { + "epoch": 5.13, + "learning_rate": 1.461066618128868e-06, + "loss": 3.3913, + "step": 28233 + }, + { + "epoch": 5.13, + "learning_rate": 1.4610120131052059e-06, + "loss": 0.0094, + "step": 28234 + }, + { + "epoch": 5.13, + "learning_rate": 1.4609574080815436e-06, + "loss": 0.0187, + "step": 28235 + }, + { + "epoch": 5.13, + "learning_rate": 1.4609028030578813e-06, + "loss": 0.254, + "step": 28236 + }, + { + "epoch": 5.13, + "learning_rate": 1.4608481980342193e-06, + "loss": 0.213, + "step": 28237 + }, + { + "epoch": 5.14, + "learning_rate": 1.460793593010557e-06, + "loss": 1.514, + "step": 28238 + }, + { + "epoch": 5.14, + "learning_rate": 1.4607389879868947e-06, + "loss": 0.0782, + "step": 28239 + }, + { + "epoch": 5.14, + "learning_rate": 1.4606843829632327e-06, + "loss": 0.3595, + "step": 28240 + }, + { + "epoch": 5.14, + "learning_rate": 1.4606297779395706e-06, + "loss": 0.3477, + "step": 28241 + }, + { + "epoch": 5.14, + "learning_rate": 1.4605751729159083e-06, + "loss": 0.8329, + "step": 28242 + }, + { + "epoch": 5.14, + "learning_rate": 1.460520567892246e-06, + "loss": 0.0634, + "step": 28243 + }, + { + "epoch": 5.14, + "learning_rate": 1.460465962868584e-06, + "loss": 0.2024, + "step": 28244 + }, + { + "epoch": 5.14, + "learning_rate": 1.4604113578449218e-06, + "loss": 0.5003, + "step": 28245 + }, + { + "epoch": 5.14, + "learning_rate": 1.4603567528212595e-06, + "loss": 0.4418, + "step": 28246 + }, + { + "epoch": 5.14, + "learning_rate": 1.4603021477975974e-06, + "loss": 1.1506, + "step": 28247 + }, + { + "epoch": 5.14, + "learning_rate": 1.4602475427739354e-06, + "loss": 0.9566, + "step": 28248 + }, + { + "epoch": 5.14, + "learning_rate": 1.460192937750273e-06, + "loss": 1.5849, + "step": 28249 + }, + { + "epoch": 5.14, + "learning_rate": 1.4601383327266108e-06, + "loss": 2.1116, + "step": 28250 + }, + { + "epoch": 5.14, + "learning_rate": 1.4600837277029488e-06, + "loss": 0.0193, + "step": 28251 + }, + { + "epoch": 5.14, + "learning_rate": 1.4600291226792865e-06, + "loss": 0.9517, + "step": 28252 + }, + { + "epoch": 5.14, + "learning_rate": 1.4599745176556242e-06, + "loss": 1.6252, + "step": 28253 + }, + { + "epoch": 5.14, + "learning_rate": 1.4599199126319622e-06, + "loss": 1.4343, + "step": 28254 + }, + { + "epoch": 5.14, + "learning_rate": 1.4598653076083001e-06, + "loss": 1.5528, + "step": 28255 + }, + { + "epoch": 5.14, + "learning_rate": 1.4598107025846379e-06, + "loss": 0.8336, + "step": 28256 + }, + { + "epoch": 5.14, + "learning_rate": 1.4597560975609756e-06, + "loss": 0.0908, + "step": 28257 + }, + { + "epoch": 5.14, + "learning_rate": 1.4597014925373133e-06, + "loss": 1.5929, + "step": 28258 + }, + { + "epoch": 5.14, + "learning_rate": 1.4596468875136513e-06, + "loss": 0.7505, + "step": 28259 + }, + { + "epoch": 5.14, + "learning_rate": 1.459592282489989e-06, + "loss": 2.2258, + "step": 28260 + }, + { + "epoch": 5.14, + "learning_rate": 1.459537677466327e-06, + "loss": 2.3766, + "step": 28261 + }, + { + "epoch": 5.14, + "learning_rate": 1.4594830724426649e-06, + "loss": 0.101, + "step": 28262 + }, + { + "epoch": 5.14, + "learning_rate": 1.4594284674190026e-06, + "loss": 0.0299, + "step": 28263 + }, + { + "epoch": 5.14, + "learning_rate": 1.4593738623953404e-06, + "loss": 1.8742, + "step": 28264 + }, + { + "epoch": 5.14, + "learning_rate": 1.459319257371678e-06, + "loss": 0.3325, + "step": 28265 + }, + { + "epoch": 5.14, + "learning_rate": 1.459264652348016e-06, + "loss": 0.2691, + "step": 28266 + }, + { + "epoch": 5.14, + "learning_rate": 1.459210047324354e-06, + "loss": 2.4849, + "step": 28267 + }, + { + "epoch": 5.14, + "learning_rate": 1.4591554423006917e-06, + "loss": 0.0458, + "step": 28268 + }, + { + "epoch": 5.14, + "learning_rate": 1.4591008372770296e-06, + "loss": 1.5262, + "step": 28269 + }, + { + "epoch": 5.14, + "learning_rate": 1.4590462322533674e-06, + "loss": 0.5723, + "step": 28270 + }, + { + "epoch": 5.14, + "learning_rate": 1.4589916272297051e-06, + "loss": 0.8273, + "step": 28271 + }, + { + "epoch": 5.14, + "learning_rate": 1.4589370222060428e-06, + "loss": 0.6387, + "step": 28272 + }, + { + "epoch": 5.14, + "learning_rate": 1.4588824171823808e-06, + "loss": 0.8335, + "step": 28273 + }, + { + "epoch": 5.14, + "learning_rate": 1.4588278121587187e-06, + "loss": 0.0797, + "step": 28274 + }, + { + "epoch": 5.14, + "learning_rate": 1.4587732071350565e-06, + "loss": 0.2812, + "step": 28275 + }, + { + "epoch": 5.14, + "learning_rate": 1.4587186021113944e-06, + "loss": 0.4678, + "step": 28276 + }, + { + "epoch": 5.14, + "learning_rate": 1.4586639970877321e-06, + "loss": 0.1285, + "step": 28277 + }, + { + "epoch": 5.14, + "learning_rate": 1.4586093920640699e-06, + "loss": 0.7355, + "step": 28278 + }, + { + "epoch": 5.14, + "learning_rate": 1.4585547870404076e-06, + "loss": 0.5896, + "step": 28279 + }, + { + "epoch": 5.14, + "learning_rate": 1.4585001820167455e-06, + "loss": 0.3811, + "step": 28280 + }, + { + "epoch": 5.14, + "learning_rate": 1.4584455769930835e-06, + "loss": 1.2521, + "step": 28281 + }, + { + "epoch": 5.14, + "learning_rate": 1.4583909719694212e-06, + "loss": 0.0146, + "step": 28282 + }, + { + "epoch": 5.14, + "learning_rate": 1.4583363669457592e-06, + "loss": 1.1557, + "step": 28283 + }, + { + "epoch": 5.14, + "learning_rate": 1.4582817619220969e-06, + "loss": 0.7504, + "step": 28284 + }, + { + "epoch": 5.14, + "learning_rate": 1.4582271568984346e-06, + "loss": 0.1138, + "step": 28285 + }, + { + "epoch": 5.14, + "learning_rate": 1.4581725518747724e-06, + "loss": 0.6552, + "step": 28286 + }, + { + "epoch": 5.14, + "learning_rate": 1.4581179468511103e-06, + "loss": 0.7167, + "step": 28287 + }, + { + "epoch": 5.14, + "learning_rate": 1.4580633418274482e-06, + "loss": 0.442, + "step": 28288 + }, + { + "epoch": 5.14, + "learning_rate": 1.458008736803786e-06, + "loss": 0.0528, + "step": 28289 + }, + { + "epoch": 5.14, + "learning_rate": 1.457954131780124e-06, + "loss": 2.1351, + "step": 28290 + }, + { + "epoch": 5.14, + "learning_rate": 1.4578995267564616e-06, + "loss": 0.0087, + "step": 28291 + }, + { + "epoch": 5.14, + "learning_rate": 1.4578449217327994e-06, + "loss": 0.3305, + "step": 28292 + }, + { + "epoch": 5.15, + "learning_rate": 1.4577903167091371e-06, + "loss": 0.1922, + "step": 28293 + }, + { + "epoch": 5.15, + "learning_rate": 1.457735711685475e-06, + "loss": 1.3665, + "step": 28294 + }, + { + "epoch": 5.15, + "learning_rate": 1.457681106661813e-06, + "loss": 1.9068, + "step": 28295 + }, + { + "epoch": 5.15, + "learning_rate": 1.4576265016381507e-06, + "loss": 0.0625, + "step": 28296 + }, + { + "epoch": 5.15, + "learning_rate": 1.4575718966144887e-06, + "loss": 0.2007, + "step": 28297 + }, + { + "epoch": 5.15, + "learning_rate": 1.4575172915908264e-06, + "loss": 0.0136, + "step": 28298 + }, + { + "epoch": 5.15, + "learning_rate": 1.4574626865671641e-06, + "loss": 0.0099, + "step": 28299 + }, + { + "epoch": 5.15, + "learning_rate": 1.4574080815435019e-06, + "loss": 2.4834, + "step": 28300 + }, + { + "epoch": 5.15, + "learning_rate": 1.4573534765198398e-06, + "loss": 0.6848, + "step": 28301 + }, + { + "epoch": 5.15, + "learning_rate": 1.4572988714961778e-06, + "loss": 0.3404, + "step": 28302 + }, + { + "epoch": 5.15, + "learning_rate": 1.4572442664725155e-06, + "loss": 0.1792, + "step": 28303 + }, + { + "epoch": 5.15, + "learning_rate": 1.4571896614488534e-06, + "loss": 0.1319, + "step": 28304 + }, + { + "epoch": 5.15, + "learning_rate": 1.4571350564251912e-06, + "loss": 0.3955, + "step": 28305 + }, + { + "epoch": 5.15, + "learning_rate": 1.4570804514015289e-06, + "loss": 0.8513, + "step": 28306 + }, + { + "epoch": 5.15, + "learning_rate": 1.4570258463778668e-06, + "loss": 0.508, + "step": 28307 + }, + { + "epoch": 5.15, + "learning_rate": 1.4569712413542048e-06, + "loss": 0.1451, + "step": 28308 + }, + { + "epoch": 5.15, + "learning_rate": 1.4569166363305425e-06, + "loss": 0.4541, + "step": 28309 + }, + { + "epoch": 5.15, + "learning_rate": 1.4568620313068802e-06, + "loss": 2.0259, + "step": 28310 + }, + { + "epoch": 5.15, + "learning_rate": 1.4568074262832182e-06, + "loss": 1.0923, + "step": 28311 + }, + { + "epoch": 5.15, + "learning_rate": 1.456752821259556e-06, + "loss": 0.0667, + "step": 28312 + }, + { + "epoch": 5.15, + "learning_rate": 1.4566982162358936e-06, + "loss": 1.1348, + "step": 28313 + }, + { + "epoch": 5.15, + "learning_rate": 1.4566436112122316e-06, + "loss": 0.149, + "step": 28314 + }, + { + "epoch": 5.15, + "learning_rate": 1.4565890061885695e-06, + "loss": 0.4824, + "step": 28315 + }, + { + "epoch": 5.15, + "learning_rate": 1.4565344011649073e-06, + "loss": 1.7095, + "step": 28316 + }, + { + "epoch": 5.15, + "learning_rate": 1.456479796141245e-06, + "loss": 3.3821, + "step": 28317 + }, + { + "epoch": 5.15, + "learning_rate": 1.456425191117583e-06, + "loss": 0.1212, + "step": 28318 + }, + { + "epoch": 5.15, + "learning_rate": 1.4563705860939207e-06, + "loss": 0.2655, + "step": 28319 + }, + { + "epoch": 5.15, + "learning_rate": 1.4563159810702584e-06, + "loss": 1.3447, + "step": 28320 + }, + { + "epoch": 5.15, + "learning_rate": 1.4562613760465963e-06, + "loss": 2.7058, + "step": 28321 + }, + { + "epoch": 5.15, + "learning_rate": 1.4562067710229343e-06, + "loss": 0.0114, + "step": 28322 + }, + { + "epoch": 5.15, + "learning_rate": 1.456152165999272e-06, + "loss": 0.8819, + "step": 28323 + }, + { + "epoch": 5.15, + "learning_rate": 1.4560975609756098e-06, + "loss": 1.2776, + "step": 28324 + }, + { + "epoch": 5.15, + "learning_rate": 1.4560429559519477e-06, + "loss": 1.0503, + "step": 28325 + }, + { + "epoch": 5.15, + "learning_rate": 1.4559883509282854e-06, + "loss": 0.0547, + "step": 28326 + }, + { + "epoch": 5.15, + "learning_rate": 1.4559337459046232e-06, + "loss": 0.4096, + "step": 28327 + }, + { + "epoch": 5.15, + "learning_rate": 1.455879140880961e-06, + "loss": 1.051, + "step": 28328 + }, + { + "epoch": 5.15, + "learning_rate": 1.455824535857299e-06, + "loss": 0.0819, + "step": 28329 + }, + { + "epoch": 5.15, + "learning_rate": 1.4557699308336368e-06, + "loss": 0.7388, + "step": 28330 + }, + { + "epoch": 5.15, + "learning_rate": 1.4557153258099745e-06, + "loss": 0.2467, + "step": 28331 + }, + { + "epoch": 5.15, + "learning_rate": 1.4556607207863125e-06, + "loss": 0.5218, + "step": 28332 + }, + { + "epoch": 5.15, + "learning_rate": 1.4556061157626502e-06, + "loss": 0.1546, + "step": 28333 + }, + { + "epoch": 5.15, + "learning_rate": 1.455551510738988e-06, + "loss": 0.4316, + "step": 28334 + }, + { + "epoch": 5.15, + "learning_rate": 1.4554969057153259e-06, + "loss": 0.5094, + "step": 28335 + }, + { + "epoch": 5.15, + "learning_rate": 1.4554423006916638e-06, + "loss": 0.9249, + "step": 28336 + }, + { + "epoch": 5.15, + "learning_rate": 1.4553876956680015e-06, + "loss": 0.6189, + "step": 28337 + }, + { + "epoch": 5.15, + "learning_rate": 1.4553330906443393e-06, + "loss": 0.162, + "step": 28338 + }, + { + "epoch": 5.15, + "learning_rate": 1.455278485620677e-06, + "loss": 1.4307, + "step": 28339 + }, + { + "epoch": 5.15, + "learning_rate": 1.455223880597015e-06, + "loss": 0.009, + "step": 28340 + }, + { + "epoch": 5.15, + "learning_rate": 1.4551692755733527e-06, + "loss": 0.2664, + "step": 28341 + }, + { + "epoch": 5.15, + "learning_rate": 1.4551146705496906e-06, + "loss": 0.2119, + "step": 28342 + }, + { + "epoch": 5.15, + "learning_rate": 1.4550600655260286e-06, + "loss": 2.867, + "step": 28343 + }, + { + "epoch": 5.15, + "learning_rate": 1.4550054605023663e-06, + "loss": 0.3358, + "step": 28344 + }, + { + "epoch": 5.15, + "learning_rate": 1.454950855478704e-06, + "loss": 0.5668, + "step": 28345 + }, + { + "epoch": 5.15, + "learning_rate": 1.4548962504550418e-06, + "loss": 0.1886, + "step": 28346 + }, + { + "epoch": 5.15, + "learning_rate": 1.4548416454313797e-06, + "loss": 0.1839, + "step": 28347 + }, + { + "epoch": 5.16, + "learning_rate": 1.4547870404077176e-06, + "loss": 0.0076, + "step": 28348 + }, + { + "epoch": 5.16, + "learning_rate": 1.4547324353840554e-06, + "loss": 0.0149, + "step": 28349 + }, + { + "epoch": 5.16, + "learning_rate": 1.4546778303603933e-06, + "loss": 1.9327, + "step": 28350 + }, + { + "epoch": 5.16, + "learning_rate": 1.454623225336731e-06, + "loss": 0.3469, + "step": 28351 + }, + { + "epoch": 5.16, + "learning_rate": 1.4545686203130688e-06, + "loss": 0.1769, + "step": 28352 + }, + { + "epoch": 5.16, + "learning_rate": 1.4545140152894065e-06, + "loss": 0.422, + "step": 28353 + }, + { + "epoch": 5.16, + "learning_rate": 1.4544594102657445e-06, + "loss": 0.2714, + "step": 28354 + }, + { + "epoch": 5.16, + "learning_rate": 1.4544048052420824e-06, + "loss": 0.2522, + "step": 28355 + }, + { + "epoch": 5.16, + "learning_rate": 1.4543502002184201e-06, + "loss": 0.5427, + "step": 28356 + }, + { + "epoch": 5.16, + "learning_rate": 1.454295595194758e-06, + "loss": 2.2447, + "step": 28357 + }, + { + "epoch": 5.16, + "learning_rate": 1.4542409901710958e-06, + "loss": 0.3721, + "step": 28358 + }, + { + "epoch": 5.16, + "learning_rate": 1.4541863851474335e-06, + "loss": 0.5001, + "step": 28359 + }, + { + "epoch": 5.16, + "learning_rate": 1.4541317801237713e-06, + "loss": 0.1314, + "step": 28360 + }, + { + "epoch": 5.16, + "learning_rate": 1.4540771751001092e-06, + "loss": 0.5154, + "step": 28361 + }, + { + "epoch": 5.16, + "learning_rate": 1.4540225700764472e-06, + "loss": 0.4243, + "step": 28362 + }, + { + "epoch": 5.16, + "learning_rate": 1.4539679650527849e-06, + "loss": 0.6527, + "step": 28363 + }, + { + "epoch": 5.16, + "learning_rate": 1.4539133600291228e-06, + "loss": 0.5663, + "step": 28364 + }, + { + "epoch": 5.16, + "learning_rate": 1.4538587550054606e-06, + "loss": 0.621, + "step": 28365 + }, + { + "epoch": 5.16, + "learning_rate": 1.4538041499817983e-06, + "loss": 0.5225, + "step": 28366 + }, + { + "epoch": 5.16, + "learning_rate": 1.453749544958136e-06, + "loss": 0.3162, + "step": 28367 + }, + { + "epoch": 5.16, + "learning_rate": 1.453694939934474e-06, + "loss": 0.0284, + "step": 28368 + }, + { + "epoch": 5.16, + "learning_rate": 1.453640334910812e-06, + "loss": 0.0187, + "step": 28369 + }, + { + "epoch": 5.16, + "learning_rate": 1.4535857298871496e-06, + "loss": 0.0163, + "step": 28370 + }, + { + "epoch": 5.16, + "learning_rate": 1.4535311248634876e-06, + "loss": 0.0654, + "step": 28371 + }, + { + "epoch": 5.16, + "learning_rate": 1.4534765198398253e-06, + "loss": 0.4506, + "step": 28372 + }, + { + "epoch": 5.16, + "learning_rate": 1.453421914816163e-06, + "loss": 0.1498, + "step": 28373 + }, + { + "epoch": 5.16, + "learning_rate": 1.4533673097925008e-06, + "loss": 0.7912, + "step": 28374 + }, + { + "epoch": 5.16, + "learning_rate": 1.4533127047688387e-06, + "loss": 0.2213, + "step": 28375 + }, + { + "epoch": 5.16, + "learning_rate": 1.4532580997451767e-06, + "loss": 1.7282, + "step": 28376 + }, + { + "epoch": 5.16, + "learning_rate": 1.4532034947215144e-06, + "loss": 0.1332, + "step": 28377 + }, + { + "epoch": 5.16, + "learning_rate": 1.4531488896978523e-06, + "loss": 0.5679, + "step": 28378 + }, + { + "epoch": 5.16, + "learning_rate": 1.45309428467419e-06, + "loss": 0.8056, + "step": 28379 + }, + { + "epoch": 5.16, + "learning_rate": 1.4530396796505278e-06, + "loss": 1.4068, + "step": 28380 + }, + { + "epoch": 5.16, + "learning_rate": 1.4529850746268655e-06, + "loss": 0.1486, + "step": 28381 + }, + { + "epoch": 5.16, + "learning_rate": 1.4529304696032035e-06, + "loss": 0.777, + "step": 28382 + }, + { + "epoch": 5.16, + "learning_rate": 1.4528758645795414e-06, + "loss": 3.5729, + "step": 28383 + }, + { + "epoch": 5.16, + "learning_rate": 1.4528212595558792e-06, + "loss": 1.8249, + "step": 28384 + }, + { + "epoch": 5.16, + "learning_rate": 1.452766654532217e-06, + "loss": 0.0088, + "step": 28385 + }, + { + "epoch": 5.16, + "learning_rate": 1.4527120495085548e-06, + "loss": 0.3095, + "step": 28386 + }, + { + "epoch": 5.16, + "learning_rate": 1.4526574444848926e-06, + "loss": 0.3032, + "step": 28387 + }, + { + "epoch": 5.16, + "learning_rate": 1.4526028394612305e-06, + "loss": 0.3077, + "step": 28388 + }, + { + "epoch": 5.16, + "learning_rate": 1.4525482344375682e-06, + "loss": 0.8879, + "step": 28389 + }, + { + "epoch": 5.16, + "learning_rate": 1.4524936294139062e-06, + "loss": 0.011, + "step": 28390 + }, + { + "epoch": 5.16, + "learning_rate": 1.452439024390244e-06, + "loss": 2.5389, + "step": 28391 + }, + { + "epoch": 5.16, + "learning_rate": 1.4523844193665819e-06, + "loss": 0.0244, + "step": 28392 + }, + { + "epoch": 5.16, + "learning_rate": 1.4523298143429196e-06, + "loss": 3.2219, + "step": 28393 + }, + { + "epoch": 5.16, + "learning_rate": 1.4522752093192573e-06, + "loss": 0.2222, + "step": 28394 + }, + { + "epoch": 5.16, + "learning_rate": 1.4522206042955953e-06, + "loss": 1.4314, + "step": 28395 + }, + { + "epoch": 5.16, + "learning_rate": 1.4521659992719332e-06, + "loss": 1.538, + "step": 28396 + }, + { + "epoch": 5.16, + "learning_rate": 1.452111394248271e-06, + "loss": 1.2464, + "step": 28397 + }, + { + "epoch": 5.16, + "learning_rate": 1.4520567892246087e-06, + "loss": 0.2834, + "step": 28398 + }, + { + "epoch": 5.16, + "learning_rate": 1.4520021842009466e-06, + "loss": 1.5907, + "step": 28399 + }, + { + "epoch": 5.16, + "learning_rate": 1.4519475791772843e-06, + "loss": 0.0303, + "step": 28400 + }, + { + "epoch": 5.16, + "learning_rate": 1.451892974153622e-06, + "loss": 0.4939, + "step": 28401 + }, + { + "epoch": 5.16, + "learning_rate": 1.45183836912996e-06, + "loss": 1.4045, + "step": 28402 + }, + { + "epoch": 5.17, + "learning_rate": 1.451783764106298e-06, + "loss": 1.3238, + "step": 28403 + }, + { + "epoch": 5.17, + "learning_rate": 1.4517291590826357e-06, + "loss": 2.2814, + "step": 28404 + }, + { + "epoch": 5.17, + "learning_rate": 1.4516745540589734e-06, + "loss": 0.604, + "step": 28405 + }, + { + "epoch": 5.17, + "learning_rate": 1.4516199490353114e-06, + "loss": 2.2076, + "step": 28406 + }, + { + "epoch": 5.17, + "learning_rate": 1.451565344011649e-06, + "loss": 0.0042, + "step": 28407 + }, + { + "epoch": 5.17, + "learning_rate": 1.4515107389879868e-06, + "loss": 0.129, + "step": 28408 + }, + { + "epoch": 5.17, + "learning_rate": 1.4514561339643248e-06, + "loss": 0.5902, + "step": 28409 + }, + { + "epoch": 5.17, + "learning_rate": 1.4514015289406627e-06, + "loss": 5.908, + "step": 28410 + }, + { + "epoch": 5.17, + "learning_rate": 1.4513469239170005e-06, + "loss": 2.6859, + "step": 28411 + }, + { + "epoch": 5.17, + "learning_rate": 1.4512923188933382e-06, + "loss": 0.0197, + "step": 28412 + }, + { + "epoch": 5.17, + "learning_rate": 1.451237713869676e-06, + "loss": 1.9826, + "step": 28413 + }, + { + "epoch": 5.17, + "learning_rate": 1.4511831088460139e-06, + "loss": 0.1749, + "step": 28414 + }, + { + "epoch": 5.17, + "learning_rate": 1.4511285038223516e-06, + "loss": 0.2086, + "step": 28415 + }, + { + "epoch": 5.17, + "learning_rate": 1.4510738987986895e-06, + "loss": 0.0096, + "step": 28416 + }, + { + "epoch": 5.17, + "learning_rate": 1.4510192937750275e-06, + "loss": 2.8439, + "step": 28417 + }, + { + "epoch": 5.17, + "learning_rate": 1.4509646887513652e-06, + "loss": 0.1499, + "step": 28418 + }, + { + "epoch": 5.17, + "learning_rate": 1.450910083727703e-06, + "loss": 0.9272, + "step": 28419 + }, + { + "epoch": 5.17, + "learning_rate": 1.4508554787040407e-06, + "loss": 0.6792, + "step": 28420 + }, + { + "epoch": 5.17, + "learning_rate": 1.4508008736803786e-06, + "loss": 2.8004, + "step": 28421 + }, + { + "epoch": 5.17, + "learning_rate": 1.4507462686567163e-06, + "loss": 1.3155, + "step": 28422 + }, + { + "epoch": 5.17, + "learning_rate": 1.4506916636330543e-06, + "loss": 1.0787, + "step": 28423 + }, + { + "epoch": 5.17, + "learning_rate": 1.4506370586093922e-06, + "loss": 0.335, + "step": 28424 + }, + { + "epoch": 5.17, + "learning_rate": 1.45058245358573e-06, + "loss": 0.265, + "step": 28425 + }, + { + "epoch": 5.17, + "learning_rate": 1.4505278485620677e-06, + "loss": 0.0051, + "step": 28426 + }, + { + "epoch": 5.17, + "learning_rate": 1.4504732435384054e-06, + "loss": 0.3518, + "step": 28427 + }, + { + "epoch": 5.17, + "learning_rate": 1.4504186385147434e-06, + "loss": 0.1965, + "step": 28428 + }, + { + "epoch": 5.17, + "learning_rate": 1.450364033491081e-06, + "loss": 0.7621, + "step": 28429 + }, + { + "epoch": 5.17, + "learning_rate": 1.450309428467419e-06, + "loss": 0.0274, + "step": 28430 + }, + { + "epoch": 5.17, + "learning_rate": 1.450254823443757e-06, + "loss": 0.8069, + "step": 28431 + }, + { + "epoch": 5.17, + "learning_rate": 1.4502002184200947e-06, + "loss": 2.8984, + "step": 28432 + }, + { + "epoch": 5.17, + "learning_rate": 1.4501456133964325e-06, + "loss": 0.5795, + "step": 28433 + }, + { + "epoch": 5.17, + "learning_rate": 1.4500910083727702e-06, + "loss": 0.1682, + "step": 28434 + }, + { + "epoch": 5.17, + "learning_rate": 1.4500364033491081e-06, + "loss": 0.2579, + "step": 28435 + }, + { + "epoch": 5.17, + "learning_rate": 1.449981798325446e-06, + "loss": 0.2506, + "step": 28436 + }, + { + "epoch": 5.17, + "learning_rate": 1.4499271933017838e-06, + "loss": 1.0837, + "step": 28437 + }, + { + "epoch": 5.17, + "learning_rate": 1.4498725882781217e-06, + "loss": 2.0229, + "step": 28438 + }, + { + "epoch": 5.17, + "learning_rate": 1.4498179832544595e-06, + "loss": 0.0369, + "step": 28439 + }, + { + "epoch": 5.17, + "learning_rate": 1.4497633782307972e-06, + "loss": 0.0093, + "step": 28440 + }, + { + "epoch": 5.17, + "learning_rate": 1.449708773207135e-06, + "loss": 0.0067, + "step": 28441 + }, + { + "epoch": 5.17, + "learning_rate": 1.4496541681834729e-06, + "loss": 0.6734, + "step": 28442 + }, + { + "epoch": 5.17, + "learning_rate": 1.4495995631598108e-06, + "loss": 0.3509, + "step": 28443 + }, + { + "epoch": 5.17, + "learning_rate": 1.4495449581361486e-06, + "loss": 0.107, + "step": 28444 + }, + { + "epoch": 5.17, + "learning_rate": 1.4494903531124865e-06, + "loss": 0.5896, + "step": 28445 + }, + { + "epoch": 5.17, + "learning_rate": 1.4494357480888242e-06, + "loss": 0.4021, + "step": 28446 + }, + { + "epoch": 5.17, + "learning_rate": 1.449381143065162e-06, + "loss": 0.7386, + "step": 28447 + }, + { + "epoch": 5.17, + "learning_rate": 1.4493265380414997e-06, + "loss": 0.0075, + "step": 28448 + }, + { + "epoch": 5.17, + "learning_rate": 1.4492719330178376e-06, + "loss": 0.0116, + "step": 28449 + }, + { + "epoch": 5.17, + "learning_rate": 1.4492173279941756e-06, + "loss": 0.1469, + "step": 28450 + }, + { + "epoch": 5.17, + "learning_rate": 1.4491627229705133e-06, + "loss": 0.0213, + "step": 28451 + }, + { + "epoch": 5.17, + "learning_rate": 1.4491081179468513e-06, + "loss": 0.5248, + "step": 28452 + }, + { + "epoch": 5.17, + "learning_rate": 1.449053512923189e-06, + "loss": 0.092, + "step": 28453 + }, + { + "epoch": 5.17, + "learning_rate": 1.4489989078995267e-06, + "loss": 0.7812, + "step": 28454 + }, + { + "epoch": 5.17, + "learning_rate": 1.4489443028758645e-06, + "loss": 0.0111, + "step": 28455 + }, + { + "epoch": 5.17, + "learning_rate": 1.4488896978522024e-06, + "loss": 0.01, + "step": 28456 + }, + { + "epoch": 5.17, + "learning_rate": 1.4488350928285403e-06, + "loss": 1.2733, + "step": 28457 + }, + { + "epoch": 5.18, + "learning_rate": 1.448780487804878e-06, + "loss": 0.0046, + "step": 28458 + }, + { + "epoch": 5.18, + "learning_rate": 1.448725882781216e-06, + "loss": 0.1845, + "step": 28459 + }, + { + "epoch": 5.18, + "learning_rate": 1.4486712777575538e-06, + "loss": 0.4993, + "step": 28460 + }, + { + "epoch": 5.18, + "learning_rate": 1.4486166727338915e-06, + "loss": 0.0219, + "step": 28461 + }, + { + "epoch": 5.18, + "learning_rate": 1.4485620677102292e-06, + "loss": 0.1369, + "step": 28462 + }, + { + "epoch": 5.18, + "learning_rate": 1.4485074626865672e-06, + "loss": 0.314, + "step": 28463 + }, + { + "epoch": 5.18, + "learning_rate": 1.448452857662905e-06, + "loss": 0.0112, + "step": 28464 + }, + { + "epoch": 5.18, + "learning_rate": 1.4483982526392428e-06, + "loss": 0.0677, + "step": 28465 + }, + { + "epoch": 5.18, + "learning_rate": 1.4483436476155808e-06, + "loss": 2.2573, + "step": 28466 + }, + { + "epoch": 5.18, + "learning_rate": 1.4482890425919185e-06, + "loss": 1.4533, + "step": 28467 + }, + { + "epoch": 5.18, + "learning_rate": 1.4482344375682562e-06, + "loss": 1.1846, + "step": 28468 + }, + { + "epoch": 5.18, + "learning_rate": 1.448179832544594e-06, + "loss": 0.3661, + "step": 28469 + }, + { + "epoch": 5.18, + "learning_rate": 1.448125227520932e-06, + "loss": 0.633, + "step": 28470 + }, + { + "epoch": 5.18, + "learning_rate": 1.4480706224972699e-06, + "loss": 4.0564, + "step": 28471 + }, + { + "epoch": 5.18, + "learning_rate": 1.4480160174736076e-06, + "loss": 0.8751, + "step": 28472 + }, + { + "epoch": 5.18, + "learning_rate": 1.4479614124499455e-06, + "loss": 0.6741, + "step": 28473 + }, + { + "epoch": 5.18, + "learning_rate": 1.4479068074262833e-06, + "loss": 0.3804, + "step": 28474 + }, + { + "epoch": 5.18, + "learning_rate": 1.447852202402621e-06, + "loss": 0.272, + "step": 28475 + }, + { + "epoch": 5.18, + "learning_rate": 1.447797597378959e-06, + "loss": 0.0145, + "step": 28476 + }, + { + "epoch": 5.18, + "learning_rate": 1.4477429923552969e-06, + "loss": 1.4951, + "step": 28477 + }, + { + "epoch": 5.18, + "learning_rate": 1.4476883873316346e-06, + "loss": 1.8435, + "step": 28478 + }, + { + "epoch": 5.18, + "learning_rate": 1.4476337823079723e-06, + "loss": 0.4484, + "step": 28479 + }, + { + "epoch": 5.18, + "learning_rate": 1.4475791772843103e-06, + "loss": 0.1224, + "step": 28480 + }, + { + "epoch": 5.18, + "learning_rate": 1.447524572260648e-06, + "loss": 0.123, + "step": 28481 + }, + { + "epoch": 5.18, + "learning_rate": 1.4474699672369858e-06, + "loss": 0.3894, + "step": 28482 + }, + { + "epoch": 5.18, + "learning_rate": 1.4474153622133237e-06, + "loss": 1.413, + "step": 28483 + }, + { + "epoch": 5.18, + "learning_rate": 1.4473607571896616e-06, + "loss": 3.3595, + "step": 28484 + }, + { + "epoch": 5.18, + "learning_rate": 1.4473061521659994e-06, + "loss": 0.9436, + "step": 28485 + }, + { + "epoch": 5.18, + "learning_rate": 1.447251547142337e-06, + "loss": 0.0578, + "step": 28486 + }, + { + "epoch": 5.18, + "learning_rate": 1.447196942118675e-06, + "loss": 0.1728, + "step": 28487 + }, + { + "epoch": 5.18, + "learning_rate": 1.4471423370950128e-06, + "loss": 0.8916, + "step": 28488 + }, + { + "epoch": 5.18, + "learning_rate": 1.4470877320713505e-06, + "loss": 0.0127, + "step": 28489 + }, + { + "epoch": 5.18, + "learning_rate": 1.4470331270476885e-06, + "loss": 0.1693, + "step": 28490 + }, + { + "epoch": 5.18, + "learning_rate": 1.4469785220240264e-06, + "loss": 0.0122, + "step": 28491 + }, + { + "epoch": 5.18, + "learning_rate": 1.4469239170003641e-06, + "loss": 0.0098, + "step": 28492 + }, + { + "epoch": 5.18, + "learning_rate": 1.4468693119767019e-06, + "loss": 0.0489, + "step": 28493 + }, + { + "epoch": 5.18, + "learning_rate": 1.4468147069530396e-06, + "loss": 0.359, + "step": 28494 + }, + { + "epoch": 5.18, + "learning_rate": 1.4467601019293775e-06, + "loss": 0.1191, + "step": 28495 + }, + { + "epoch": 5.18, + "learning_rate": 1.4467054969057153e-06, + "loss": 0.3268, + "step": 28496 + }, + { + "epoch": 5.18, + "learning_rate": 1.4466508918820532e-06, + "loss": 0.3066, + "step": 28497 + }, + { + "epoch": 5.18, + "learning_rate": 1.4465962868583912e-06, + "loss": 0.2002, + "step": 28498 + }, + { + "epoch": 5.18, + "learning_rate": 1.4465416818347289e-06, + "loss": 0.0055, + "step": 28499 + }, + { + "epoch": 5.18, + "learning_rate": 1.4464870768110666e-06, + "loss": 0.0498, + "step": 28500 + }, + { + "epoch": 5.18, + "learning_rate": 1.4464324717874043e-06, + "loss": 2.3091, + "step": 28501 + }, + { + "epoch": 5.18, + "learning_rate": 1.4463778667637423e-06, + "loss": 1.2251, + "step": 28502 + }, + { + "epoch": 5.18, + "learning_rate": 1.44632326174008e-06, + "loss": 0.4926, + "step": 28503 + }, + { + "epoch": 5.18, + "learning_rate": 1.446268656716418e-06, + "loss": 0.8274, + "step": 28504 + }, + { + "epoch": 5.18, + "learning_rate": 1.446214051692756e-06, + "loss": 2.2966, + "step": 28505 + }, + { + "epoch": 5.18, + "learning_rate": 1.4461594466690936e-06, + "loss": 0.4752, + "step": 28506 + }, + { + "epoch": 5.18, + "learning_rate": 1.4461048416454314e-06, + "loss": 0.1563, + "step": 28507 + }, + { + "epoch": 5.18, + "learning_rate": 1.446050236621769e-06, + "loss": 0.9664, + "step": 28508 + }, + { + "epoch": 5.18, + "learning_rate": 1.445995631598107e-06, + "loss": 0.2169, + "step": 28509 + }, + { + "epoch": 5.18, + "learning_rate": 1.4459410265744448e-06, + "loss": 2.8821, + "step": 28510 + }, + { + "epoch": 5.18, + "learning_rate": 1.4458864215507827e-06, + "loss": 0.2332, + "step": 28511 + }, + { + "epoch": 5.18, + "learning_rate": 1.4458318165271207e-06, + "loss": 0.0097, + "step": 28512 + }, + { + "epoch": 5.19, + "learning_rate": 1.4457772115034584e-06, + "loss": 0.0723, + "step": 28513 + }, + { + "epoch": 5.19, + "learning_rate": 1.4457226064797961e-06, + "loss": 0.154, + "step": 28514 + }, + { + "epoch": 5.19, + "learning_rate": 1.4456680014561339e-06, + "loss": 1.8152, + "step": 28515 + }, + { + "epoch": 5.19, + "learning_rate": 1.4456133964324718e-06, + "loss": 0.7129, + "step": 28516 + }, + { + "epoch": 5.19, + "learning_rate": 1.4455587914088097e-06, + "loss": 0.4323, + "step": 28517 + }, + { + "epoch": 5.19, + "learning_rate": 1.4455041863851475e-06, + "loss": 0.4969, + "step": 28518 + }, + { + "epoch": 5.19, + "learning_rate": 1.4454495813614854e-06, + "loss": 0.3905, + "step": 28519 + }, + { + "epoch": 5.19, + "learning_rate": 1.4453949763378232e-06, + "loss": 0.0066, + "step": 28520 + }, + { + "epoch": 5.19, + "learning_rate": 1.4453403713141609e-06, + "loss": 0.4805, + "step": 28521 + }, + { + "epoch": 5.19, + "learning_rate": 1.4452857662904986e-06, + "loss": 0.3821, + "step": 28522 + }, + { + "epoch": 5.19, + "learning_rate": 1.4452311612668366e-06, + "loss": 0.4654, + "step": 28523 + }, + { + "epoch": 5.19, + "learning_rate": 1.4451765562431745e-06, + "loss": 0.013, + "step": 28524 + }, + { + "epoch": 5.19, + "learning_rate": 1.4451219512195122e-06, + "loss": 1.2437, + "step": 28525 + }, + { + "epoch": 5.19, + "learning_rate": 1.4450673461958502e-06, + "loss": 0.0462, + "step": 28526 + }, + { + "epoch": 5.19, + "learning_rate": 1.445012741172188e-06, + "loss": 0.8501, + "step": 28527 + }, + { + "epoch": 5.19, + "learning_rate": 1.4449581361485256e-06, + "loss": 0.0048, + "step": 28528 + }, + { + "epoch": 5.19, + "learning_rate": 1.4449035311248634e-06, + "loss": 0.1166, + "step": 28529 + }, + { + "epoch": 5.19, + "learning_rate": 1.4448489261012013e-06, + "loss": 1.303, + "step": 28530 + }, + { + "epoch": 5.19, + "learning_rate": 1.4447943210775393e-06, + "loss": 0.3439, + "step": 28531 + }, + { + "epoch": 5.19, + "learning_rate": 1.444739716053877e-06, + "loss": 1.7065, + "step": 28532 + }, + { + "epoch": 5.19, + "learning_rate": 1.444685111030215e-06, + "loss": 0.2293, + "step": 28533 + }, + { + "epoch": 5.19, + "learning_rate": 1.4446305060065527e-06, + "loss": 0.0534, + "step": 28534 + }, + { + "epoch": 5.19, + "learning_rate": 1.4445759009828904e-06, + "loss": 0.0627, + "step": 28535 + }, + { + "epoch": 5.19, + "learning_rate": 1.4445212959592281e-06, + "loss": 0.0053, + "step": 28536 + }, + { + "epoch": 5.19, + "learning_rate": 1.444466690935566e-06, + "loss": 0.3219, + "step": 28537 + }, + { + "epoch": 5.19, + "learning_rate": 1.444412085911904e-06, + "loss": 0.1046, + "step": 28538 + }, + { + "epoch": 5.19, + "learning_rate": 1.4443574808882418e-06, + "loss": 0.1282, + "step": 28539 + }, + { + "epoch": 5.19, + "learning_rate": 1.4443028758645797e-06, + "loss": 0.379, + "step": 28540 + }, + { + "epoch": 5.19, + "learning_rate": 1.4442482708409174e-06, + "loss": 0.0114, + "step": 28541 + }, + { + "epoch": 5.19, + "learning_rate": 1.4441936658172552e-06, + "loss": 1.6936, + "step": 28542 + }, + { + "epoch": 5.19, + "learning_rate": 1.4441390607935929e-06, + "loss": 1.7181, + "step": 28543 + }, + { + "epoch": 5.19, + "learning_rate": 1.4440844557699308e-06, + "loss": 4.665, + "step": 28544 + }, + { + "epoch": 5.19, + "learning_rate": 1.4440298507462688e-06, + "loss": 0.1108, + "step": 28545 + }, + { + "epoch": 5.19, + "learning_rate": 1.4439752457226065e-06, + "loss": 0.9752, + "step": 28546 + }, + { + "epoch": 5.19, + "learning_rate": 1.4439206406989445e-06, + "loss": 0.013, + "step": 28547 + }, + { + "epoch": 5.19, + "learning_rate": 1.4438660356752822e-06, + "loss": 0.8656, + "step": 28548 + }, + { + "epoch": 5.19, + "learning_rate": 1.44381143065162e-06, + "loss": 0.1281, + "step": 28549 + }, + { + "epoch": 5.19, + "learning_rate": 1.4437568256279576e-06, + "loss": 0.3268, + "step": 28550 + }, + { + "epoch": 5.19, + "learning_rate": 1.4437022206042956e-06, + "loss": 1.8259, + "step": 28551 + }, + { + "epoch": 5.19, + "learning_rate": 1.4436476155806335e-06, + "loss": 1.9141, + "step": 28552 + }, + { + "epoch": 5.19, + "learning_rate": 1.4435930105569713e-06, + "loss": 0.0093, + "step": 28553 + }, + { + "epoch": 5.19, + "learning_rate": 1.4435384055333092e-06, + "loss": 0.1322, + "step": 28554 + }, + { + "epoch": 5.19, + "learning_rate": 1.443483800509647e-06, + "loss": 0.0904, + "step": 28555 + }, + { + "epoch": 5.19, + "learning_rate": 1.4434291954859847e-06, + "loss": 2.8503, + "step": 28556 + }, + { + "epoch": 5.19, + "learning_rate": 1.4433745904623224e-06, + "loss": 0.5452, + "step": 28557 + }, + { + "epoch": 5.19, + "learning_rate": 1.4433199854386603e-06, + "loss": 0.2823, + "step": 28558 + }, + { + "epoch": 5.19, + "learning_rate": 1.4432653804149983e-06, + "loss": 1.1088, + "step": 28559 + }, + { + "epoch": 5.19, + "learning_rate": 1.443210775391336e-06, + "loss": 0.302, + "step": 28560 + }, + { + "epoch": 5.19, + "learning_rate": 1.443156170367674e-06, + "loss": 1.8815, + "step": 28561 + }, + { + "epoch": 5.19, + "learning_rate": 1.4431015653440117e-06, + "loss": 0.329, + "step": 28562 + }, + { + "epoch": 5.19, + "learning_rate": 1.4430469603203494e-06, + "loss": 1.5724, + "step": 28563 + }, + { + "epoch": 5.19, + "learning_rate": 1.4429923552966874e-06, + "loss": 0.0731, + "step": 28564 + }, + { + "epoch": 5.19, + "learning_rate": 1.4429377502730253e-06, + "loss": 0.7966, + "step": 28565 + }, + { + "epoch": 5.19, + "learning_rate": 1.442883145249363e-06, + "loss": 0.0075, + "step": 28566 + }, + { + "epoch": 5.19, + "learning_rate": 1.4428285402257008e-06, + "loss": 1.3345, + "step": 28567 + }, + { + "epoch": 5.2, + "learning_rate": 1.4427739352020385e-06, + "loss": 0.5108, + "step": 28568 + }, + { + "epoch": 5.2, + "learning_rate": 1.4427193301783765e-06, + "loss": 0.4653, + "step": 28569 + }, + { + "epoch": 5.2, + "learning_rate": 1.4426647251547142e-06, + "loss": 0.5883, + "step": 28570 + }, + { + "epoch": 5.2, + "learning_rate": 1.4426101201310521e-06, + "loss": 0.8324, + "step": 28571 + }, + { + "epoch": 5.2, + "learning_rate": 1.44255551510739e-06, + "loss": 2.3044, + "step": 28572 + }, + { + "epoch": 5.2, + "learning_rate": 1.4425009100837278e-06, + "loss": 0.0144, + "step": 28573 + }, + { + "epoch": 5.2, + "learning_rate": 1.4424463050600655e-06, + "loss": 1.8828, + "step": 28574 + }, + { + "epoch": 5.2, + "learning_rate": 1.4423917000364033e-06, + "loss": 0.5391, + "step": 28575 + }, + { + "epoch": 5.2, + "learning_rate": 1.4423370950127412e-06, + "loss": 5.0092, + "step": 28576 + }, + { + "epoch": 5.2, + "learning_rate": 1.442282489989079e-06, + "loss": 0.009, + "step": 28577 + }, + { + "epoch": 5.2, + "learning_rate": 1.4422278849654169e-06, + "loss": 1.4164, + "step": 28578 + }, + { + "epoch": 5.2, + "learning_rate": 1.4421732799417548e-06, + "loss": 0.0549, + "step": 28579 + }, + { + "epoch": 5.2, + "learning_rate": 1.4421186749180926e-06, + "loss": 1.5773, + "step": 28580 + }, + { + "epoch": 5.2, + "learning_rate": 1.4420640698944303e-06, + "loss": 0.1381, + "step": 28581 + }, + { + "epoch": 5.2, + "learning_rate": 1.442009464870768e-06, + "loss": 0.0419, + "step": 28582 + }, + { + "epoch": 5.2, + "learning_rate": 1.441954859847106e-06, + "loss": 1.4111, + "step": 28583 + }, + { + "epoch": 5.2, + "learning_rate": 1.4419002548234437e-06, + "loss": 0.3527, + "step": 28584 + }, + { + "epoch": 5.2, + "learning_rate": 1.4418456497997816e-06, + "loss": 0.1118, + "step": 28585 + }, + { + "epoch": 5.2, + "learning_rate": 1.4417910447761196e-06, + "loss": 0.8555, + "step": 28586 + }, + { + "epoch": 5.2, + "learning_rate": 1.4417364397524573e-06, + "loss": 0.3001, + "step": 28587 + }, + { + "epoch": 5.2, + "learning_rate": 1.441681834728795e-06, + "loss": 0.126, + "step": 28588 + }, + { + "epoch": 5.2, + "learning_rate": 1.4416272297051328e-06, + "loss": 1.9832, + "step": 28589 + }, + { + "epoch": 5.2, + "learning_rate": 1.4415726246814707e-06, + "loss": 0.0642, + "step": 28590 + }, + { + "epoch": 5.2, + "learning_rate": 1.4415180196578085e-06, + "loss": 0.02, + "step": 28591 + }, + { + "epoch": 5.2, + "learning_rate": 1.4414634146341464e-06, + "loss": 0.6749, + "step": 28592 + }, + { + "epoch": 5.2, + "learning_rate": 1.4414088096104843e-06, + "loss": 0.6446, + "step": 28593 + }, + { + "epoch": 5.2, + "learning_rate": 1.441354204586822e-06, + "loss": 0.1169, + "step": 28594 + }, + { + "epoch": 5.2, + "learning_rate": 1.4412995995631598e-06, + "loss": 0.0302, + "step": 28595 + }, + { + "epoch": 5.2, + "learning_rate": 1.4412449945394975e-06, + "loss": 0.114, + "step": 28596 + }, + { + "epoch": 5.2, + "learning_rate": 1.4411903895158355e-06, + "loss": 0.1874, + "step": 28597 + }, + { + "epoch": 5.2, + "learning_rate": 1.4411357844921732e-06, + "loss": 0.2746, + "step": 28598 + }, + { + "epoch": 5.2, + "learning_rate": 1.4410811794685112e-06, + "loss": 1.7404, + "step": 28599 + }, + { + "epoch": 5.2, + "learning_rate": 1.441026574444849e-06, + "loss": 0.066, + "step": 28600 + }, + { + "epoch": 5.2, + "learning_rate": 1.4409719694211868e-06, + "loss": 0.5789, + "step": 28601 + }, + { + "epoch": 5.2, + "learning_rate": 1.4409173643975246e-06, + "loss": 0.7716, + "step": 28602 + }, + { + "epoch": 5.2, + "learning_rate": 1.4408627593738623e-06, + "loss": 0.3882, + "step": 28603 + }, + { + "epoch": 5.2, + "learning_rate": 1.4408081543502002e-06, + "loss": 0.1186, + "step": 28604 + }, + { + "epoch": 5.2, + "learning_rate": 1.4407535493265382e-06, + "loss": 0.9363, + "step": 28605 + }, + { + "epoch": 5.2, + "learning_rate": 1.440698944302876e-06, + "loss": 0.2993, + "step": 28606 + }, + { + "epoch": 5.2, + "learning_rate": 1.4406443392792139e-06, + "loss": 0.0686, + "step": 28607 + }, + { + "epoch": 5.2, + "learning_rate": 1.4405897342555516e-06, + "loss": 1.9773, + "step": 28608 + }, + { + "epoch": 5.2, + "learning_rate": 1.4405351292318893e-06, + "loss": 0.6453, + "step": 28609 + }, + { + "epoch": 5.2, + "learning_rate": 1.440480524208227e-06, + "loss": 2.6417, + "step": 28610 + }, + { + "epoch": 5.2, + "learning_rate": 1.440425919184565e-06, + "loss": 3.941, + "step": 28611 + }, + { + "epoch": 5.2, + "learning_rate": 1.440371314160903e-06, + "loss": 0.941, + "step": 28612 + }, + { + "epoch": 5.2, + "learning_rate": 1.4403167091372407e-06, + "loss": 0.4244, + "step": 28613 + }, + { + "epoch": 5.2, + "learning_rate": 1.4402621041135786e-06, + "loss": 1.5476, + "step": 28614 + }, + { + "epoch": 5.2, + "learning_rate": 1.4402074990899163e-06, + "loss": 1.4146, + "step": 28615 + }, + { + "epoch": 5.2, + "learning_rate": 1.440152894066254e-06, + "loss": 0.013, + "step": 28616 + }, + { + "epoch": 5.2, + "learning_rate": 1.4400982890425918e-06, + "loss": 1.4186, + "step": 28617 + }, + { + "epoch": 5.2, + "learning_rate": 1.4400436840189297e-06, + "loss": 0.1086, + "step": 28618 + }, + { + "epoch": 5.2, + "learning_rate": 1.4399890789952677e-06, + "loss": 0.8077, + "step": 28619 + }, + { + "epoch": 5.2, + "learning_rate": 1.4399344739716054e-06, + "loss": 2.9273, + "step": 28620 + }, + { + "epoch": 5.2, + "learning_rate": 1.4398798689479434e-06, + "loss": 0.0695, + "step": 28621 + }, + { + "epoch": 5.2, + "learning_rate": 1.439825263924281e-06, + "loss": 0.0165, + "step": 28622 + }, + { + "epoch": 5.21, + "learning_rate": 1.4397706589006188e-06, + "loss": 0.7798, + "step": 28623 + }, + { + "epoch": 5.21, + "learning_rate": 1.4397160538769566e-06, + "loss": 0.1294, + "step": 28624 + }, + { + "epoch": 5.21, + "learning_rate": 1.4396614488532945e-06, + "loss": 0.2153, + "step": 28625 + }, + { + "epoch": 5.21, + "learning_rate": 1.4396068438296324e-06, + "loss": 0.2431, + "step": 28626 + }, + { + "epoch": 5.21, + "learning_rate": 1.4395522388059702e-06, + "loss": 0.0075, + "step": 28627 + }, + { + "epoch": 5.21, + "learning_rate": 1.4394976337823081e-06, + "loss": 1.6719, + "step": 28628 + }, + { + "epoch": 5.21, + "learning_rate": 1.4394430287586459e-06, + "loss": 0.2968, + "step": 28629 + }, + { + "epoch": 5.21, + "learning_rate": 1.4393884237349836e-06, + "loss": 0.0178, + "step": 28630 + }, + { + "epoch": 5.21, + "learning_rate": 1.4393338187113213e-06, + "loss": 0.5381, + "step": 28631 + }, + { + "epoch": 5.21, + "learning_rate": 1.4392792136876593e-06, + "loss": 0.3003, + "step": 28632 + }, + { + "epoch": 5.21, + "learning_rate": 1.4392246086639972e-06, + "loss": 0.0281, + "step": 28633 + }, + { + "epoch": 5.21, + "learning_rate": 1.439170003640335e-06, + "loss": 3.0236, + "step": 28634 + }, + { + "epoch": 5.21, + "learning_rate": 1.4391153986166729e-06, + "loss": 1.464, + "step": 28635 + }, + { + "epoch": 5.21, + "learning_rate": 1.4390607935930106e-06, + "loss": 0.1534, + "step": 28636 + }, + { + "epoch": 5.21, + "learning_rate": 1.4390061885693483e-06, + "loss": 0.4978, + "step": 28637 + }, + { + "epoch": 5.21, + "learning_rate": 1.438951583545686e-06, + "loss": 0.8396, + "step": 28638 + }, + { + "epoch": 5.21, + "learning_rate": 1.438896978522024e-06, + "loss": 0.4041, + "step": 28639 + }, + { + "epoch": 5.21, + "learning_rate": 1.438842373498362e-06, + "loss": 2.3672, + "step": 28640 + }, + { + "epoch": 5.21, + "learning_rate": 1.4387877684746997e-06, + "loss": 0.9509, + "step": 28641 + }, + { + "epoch": 5.21, + "learning_rate": 1.4387331634510376e-06, + "loss": 3.1145, + "step": 28642 + }, + { + "epoch": 5.21, + "learning_rate": 1.4386785584273754e-06, + "loss": 0.4878, + "step": 28643 + }, + { + "epoch": 5.21, + "learning_rate": 1.438623953403713e-06, + "loss": 0.8127, + "step": 28644 + }, + { + "epoch": 5.21, + "learning_rate": 1.438569348380051e-06, + "loss": 1.4814, + "step": 28645 + }, + { + "epoch": 5.21, + "learning_rate": 1.438514743356389e-06, + "loss": 0.7258, + "step": 28646 + }, + { + "epoch": 5.21, + "learning_rate": 1.4384601383327267e-06, + "loss": 0.1784, + "step": 28647 + }, + { + "epoch": 5.21, + "learning_rate": 1.4384055333090645e-06, + "loss": 1.4107, + "step": 28648 + }, + { + "epoch": 5.21, + "learning_rate": 1.4383509282854022e-06, + "loss": 0.007, + "step": 28649 + }, + { + "epoch": 5.21, + "learning_rate": 1.4382963232617401e-06, + "loss": 0.0533, + "step": 28650 + }, + { + "epoch": 5.21, + "learning_rate": 1.4382417182380779e-06, + "loss": 0.1405, + "step": 28651 + }, + { + "epoch": 5.21, + "learning_rate": 1.4381871132144158e-06, + "loss": 0.1855, + "step": 28652 + }, + { + "epoch": 5.21, + "learning_rate": 1.4381325081907537e-06, + "loss": 0.211, + "step": 28653 + }, + { + "epoch": 5.21, + "learning_rate": 1.4380779031670915e-06, + "loss": 0.2207, + "step": 28654 + }, + { + "epoch": 5.21, + "learning_rate": 1.4380232981434292e-06, + "loss": 0.7746, + "step": 28655 + }, + { + "epoch": 5.21, + "learning_rate": 1.437968693119767e-06, + "loss": 0.6036, + "step": 28656 + }, + { + "epoch": 5.21, + "learning_rate": 1.4379140880961049e-06, + "loss": 0.1401, + "step": 28657 + }, + { + "epoch": 5.21, + "learning_rate": 1.4378594830724426e-06, + "loss": 0.0179, + "step": 28658 + }, + { + "epoch": 5.21, + "learning_rate": 1.4378048780487806e-06, + "loss": 0.6335, + "step": 28659 + }, + { + "epoch": 5.21, + "learning_rate": 1.4377502730251185e-06, + "loss": 0.1277, + "step": 28660 + }, + { + "epoch": 5.21, + "learning_rate": 1.4376956680014562e-06, + "loss": 1.2432, + "step": 28661 + }, + { + "epoch": 5.21, + "learning_rate": 1.437641062977794e-06, + "loss": 0.2235, + "step": 28662 + }, + { + "epoch": 5.21, + "learning_rate": 1.4375864579541317e-06, + "loss": 0.1905, + "step": 28663 + }, + { + "epoch": 5.21, + "learning_rate": 1.4375318529304696e-06, + "loss": 2.6174, + "step": 28664 + }, + { + "epoch": 5.21, + "learning_rate": 1.4374772479068074e-06, + "loss": 0.2189, + "step": 28665 + }, + { + "epoch": 5.21, + "learning_rate": 1.4374226428831453e-06, + "loss": 0.6098, + "step": 28666 + }, + { + "epoch": 5.21, + "learning_rate": 1.4373680378594833e-06, + "loss": 0.4097, + "step": 28667 + }, + { + "epoch": 5.21, + "learning_rate": 1.437313432835821e-06, + "loss": 0.2057, + "step": 28668 + }, + { + "epoch": 5.21, + "learning_rate": 1.4372588278121587e-06, + "loss": 0.2302, + "step": 28669 + }, + { + "epoch": 5.21, + "learning_rate": 1.4372042227884965e-06, + "loss": 0.0118, + "step": 28670 + }, + { + "epoch": 5.21, + "learning_rate": 1.4371496177648344e-06, + "loss": 0.0078, + "step": 28671 + }, + { + "epoch": 5.21, + "learning_rate": 1.4370950127411721e-06, + "loss": 1.6035, + "step": 28672 + }, + { + "epoch": 5.21, + "learning_rate": 1.43704040771751e-06, + "loss": 0.0877, + "step": 28673 + }, + { + "epoch": 5.21, + "learning_rate": 1.436985802693848e-06, + "loss": 1.0225, + "step": 28674 + }, + { + "epoch": 5.21, + "learning_rate": 1.4369311976701857e-06, + "loss": 1.9088, + "step": 28675 + }, + { + "epoch": 5.21, + "learning_rate": 1.4368765926465235e-06, + "loss": 0.0121, + "step": 28676 + }, + { + "epoch": 5.21, + "learning_rate": 1.4368219876228612e-06, + "loss": 0.0081, + "step": 28677 + }, + { + "epoch": 5.22, + "learning_rate": 1.4367673825991992e-06, + "loss": 0.9962, + "step": 28678 + }, + { + "epoch": 5.22, + "learning_rate": 1.4367127775755369e-06, + "loss": 3.7008, + "step": 28679 + }, + { + "epoch": 5.22, + "learning_rate": 1.4366581725518748e-06, + "loss": 0.4274, + "step": 28680 + }, + { + "epoch": 5.22, + "learning_rate": 1.4366035675282128e-06, + "loss": 2.2195, + "step": 28681 + }, + { + "epoch": 5.22, + "learning_rate": 1.4365489625045505e-06, + "loss": 0.068, + "step": 28682 + }, + { + "epoch": 5.22, + "learning_rate": 1.4364943574808882e-06, + "loss": 4.0087, + "step": 28683 + }, + { + "epoch": 5.22, + "learning_rate": 1.436439752457226e-06, + "loss": 0.5624, + "step": 28684 + }, + { + "epoch": 5.22, + "learning_rate": 1.436385147433564e-06, + "loss": 0.1487, + "step": 28685 + }, + { + "epoch": 5.22, + "learning_rate": 1.4363305424099019e-06, + "loss": 0.6189, + "step": 28686 + }, + { + "epoch": 5.22, + "learning_rate": 1.4362759373862396e-06, + "loss": 0.2614, + "step": 28687 + }, + { + "epoch": 5.22, + "learning_rate": 1.4362213323625775e-06, + "loss": 0.0188, + "step": 28688 + }, + { + "epoch": 5.22, + "learning_rate": 1.4361667273389153e-06, + "loss": 0.6371, + "step": 28689 + }, + { + "epoch": 5.22, + "learning_rate": 1.436112122315253e-06, + "loss": 0.2878, + "step": 28690 + }, + { + "epoch": 5.22, + "learning_rate": 1.4360575172915907e-06, + "loss": 0.5599, + "step": 28691 + }, + { + "epoch": 5.22, + "learning_rate": 1.4360029122679287e-06, + "loss": 0.1896, + "step": 28692 + }, + { + "epoch": 5.22, + "learning_rate": 1.4359483072442666e-06, + "loss": 1.2613, + "step": 28693 + }, + { + "epoch": 5.22, + "learning_rate": 1.4358937022206043e-06, + "loss": 0.0705, + "step": 28694 + }, + { + "epoch": 5.22, + "learning_rate": 1.4358390971969423e-06, + "loss": 0.5576, + "step": 28695 + }, + { + "epoch": 5.22, + "learning_rate": 1.43578449217328e-06, + "loss": 2.4314, + "step": 28696 + }, + { + "epoch": 5.22, + "learning_rate": 1.4357298871496177e-06, + "loss": 0.1866, + "step": 28697 + }, + { + "epoch": 5.22, + "learning_rate": 1.4356752821259555e-06, + "loss": 0.2293, + "step": 28698 + }, + { + "epoch": 5.22, + "learning_rate": 1.4356206771022934e-06, + "loss": 0.3041, + "step": 28699 + }, + { + "epoch": 5.22, + "learning_rate": 1.4355660720786314e-06, + "loss": 0.0624, + "step": 28700 + }, + { + "epoch": 5.22, + "learning_rate": 1.435511467054969e-06, + "loss": 0.0523, + "step": 28701 + }, + { + "epoch": 5.22, + "learning_rate": 1.435456862031307e-06, + "loss": 1.0494, + "step": 28702 + }, + { + "epoch": 5.22, + "learning_rate": 1.4354022570076448e-06, + "loss": 0.0883, + "step": 28703 + }, + { + "epoch": 5.22, + "learning_rate": 1.4353476519839825e-06, + "loss": 0.4102, + "step": 28704 + }, + { + "epoch": 5.22, + "learning_rate": 1.4352930469603202e-06, + "loss": 0.0103, + "step": 28705 + }, + { + "epoch": 5.22, + "learning_rate": 1.4352384419366582e-06, + "loss": 0.2825, + "step": 28706 + }, + { + "epoch": 5.22, + "learning_rate": 1.4351838369129961e-06, + "loss": 0.5822, + "step": 28707 + }, + { + "epoch": 5.22, + "learning_rate": 1.4351292318893339e-06, + "loss": 0.0204, + "step": 28708 + }, + { + "epoch": 5.22, + "learning_rate": 1.4350746268656718e-06, + "loss": 0.4256, + "step": 28709 + }, + { + "epoch": 5.22, + "learning_rate": 1.4350200218420095e-06, + "loss": 0.0626, + "step": 28710 + }, + { + "epoch": 5.22, + "learning_rate": 1.4349654168183473e-06, + "loss": 1.46, + "step": 28711 + }, + { + "epoch": 5.22, + "learning_rate": 1.434910811794685e-06, + "loss": 0.2321, + "step": 28712 + }, + { + "epoch": 5.22, + "learning_rate": 1.434856206771023e-06, + "loss": 1.6146, + "step": 28713 + }, + { + "epoch": 5.22, + "learning_rate": 1.4348016017473609e-06, + "loss": 0.1454, + "step": 28714 + }, + { + "epoch": 5.22, + "learning_rate": 1.4347469967236986e-06, + "loss": 0.9225, + "step": 28715 + }, + { + "epoch": 5.22, + "learning_rate": 1.4346923917000366e-06, + "loss": 1.6328, + "step": 28716 + }, + { + "epoch": 5.22, + "learning_rate": 1.4346377866763743e-06, + "loss": 0.8194, + "step": 28717 + }, + { + "epoch": 5.22, + "learning_rate": 1.434583181652712e-06, + "loss": 1.2208, + "step": 28718 + }, + { + "epoch": 5.22, + "learning_rate": 1.4345285766290498e-06, + "loss": 3.2369, + "step": 28719 + }, + { + "epoch": 5.22, + "learning_rate": 1.4344739716053877e-06, + "loss": 1.0582, + "step": 28720 + }, + { + "epoch": 5.22, + "learning_rate": 1.4344193665817256e-06, + "loss": 1.9448, + "step": 28721 + }, + { + "epoch": 5.22, + "learning_rate": 1.4343647615580634e-06, + "loss": 0.1519, + "step": 28722 + }, + { + "epoch": 5.22, + "learning_rate": 1.434310156534401e-06, + "loss": 2.8104, + "step": 28723 + }, + { + "epoch": 5.22, + "learning_rate": 1.434255551510739e-06, + "loss": 0.0107, + "step": 28724 + }, + { + "epoch": 5.22, + "learning_rate": 1.4342009464870768e-06, + "loss": 1.0656, + "step": 28725 + }, + { + "epoch": 5.22, + "learning_rate": 1.4341463414634145e-06, + "loss": 0.0115, + "step": 28726 + }, + { + "epoch": 5.22, + "learning_rate": 1.4340917364397525e-06, + "loss": 0.091, + "step": 28727 + }, + { + "epoch": 5.22, + "learning_rate": 1.4340371314160904e-06, + "loss": 1.811, + "step": 28728 + }, + { + "epoch": 5.22, + "learning_rate": 1.4339825263924281e-06, + "loss": 0.2884, + "step": 28729 + }, + { + "epoch": 5.22, + "learning_rate": 1.4339279213687659e-06, + "loss": 0.883, + "step": 28730 + }, + { + "epoch": 5.22, + "learning_rate": 1.4338733163451038e-06, + "loss": 0.0852, + "step": 28731 + }, + { + "epoch": 5.22, + "learning_rate": 1.4338187113214415e-06, + "loss": 1.2866, + "step": 28732 + }, + { + "epoch": 5.23, + "learning_rate": 1.4337641062977795e-06, + "loss": 1.6602, + "step": 28733 + }, + { + "epoch": 5.23, + "learning_rate": 1.4337095012741174e-06, + "loss": 0.3008, + "step": 28734 + }, + { + "epoch": 5.23, + "learning_rate": 1.4336548962504552e-06, + "loss": 0.7293, + "step": 28735 + }, + { + "epoch": 5.23, + "learning_rate": 1.4336002912267929e-06, + "loss": 0.3794, + "step": 28736 + }, + { + "epoch": 5.23, + "learning_rate": 1.4335456862031306e-06, + "loss": 0.6528, + "step": 28737 + }, + { + "epoch": 5.23, + "learning_rate": 1.4334910811794686e-06, + "loss": 2.2546, + "step": 28738 + }, + { + "epoch": 5.23, + "learning_rate": 1.4334364761558063e-06, + "loss": 0.1241, + "step": 28739 + }, + { + "epoch": 5.23, + "learning_rate": 1.4333818711321442e-06, + "loss": 0.2305, + "step": 28740 + }, + { + "epoch": 5.23, + "learning_rate": 1.4333272661084822e-06, + "loss": 1.0541, + "step": 28741 + }, + { + "epoch": 5.23, + "learning_rate": 1.43327266108482e-06, + "loss": 0.1866, + "step": 28742 + }, + { + "epoch": 5.23, + "learning_rate": 1.4332180560611576e-06, + "loss": 1.1945, + "step": 28743 + }, + { + "epoch": 5.23, + "learning_rate": 1.4331634510374954e-06, + "loss": 0.0157, + "step": 28744 + }, + { + "epoch": 5.23, + "learning_rate": 1.4331088460138333e-06, + "loss": 1.7728, + "step": 28745 + }, + { + "epoch": 5.23, + "learning_rate": 1.433054240990171e-06, + "loss": 0.2296, + "step": 28746 + }, + { + "epoch": 5.23, + "learning_rate": 1.432999635966509e-06, + "loss": 3.8634, + "step": 28747 + }, + { + "epoch": 5.23, + "learning_rate": 1.432945030942847e-06, + "loss": 1.5241, + "step": 28748 + }, + { + "epoch": 5.23, + "learning_rate": 1.4328904259191847e-06, + "loss": 0.2715, + "step": 28749 + }, + { + "epoch": 5.23, + "learning_rate": 1.4328358208955224e-06, + "loss": 0.0775, + "step": 28750 + }, + { + "epoch": 5.23, + "learning_rate": 1.4327812158718601e-06, + "loss": 0.148, + "step": 28751 + }, + { + "epoch": 5.23, + "learning_rate": 1.432726610848198e-06, + "loss": 0.1888, + "step": 28752 + }, + { + "epoch": 5.23, + "learning_rate": 1.4326720058245358e-06, + "loss": 0.2236, + "step": 28753 + }, + { + "epoch": 5.23, + "learning_rate": 1.4326174008008737e-06, + "loss": 0.7906, + "step": 28754 + }, + { + "epoch": 5.23, + "learning_rate": 1.4325627957772117e-06, + "loss": 0.4712, + "step": 28755 + }, + { + "epoch": 5.23, + "learning_rate": 1.4325081907535494e-06, + "loss": 1.0269, + "step": 28756 + }, + { + "epoch": 5.23, + "learning_rate": 1.4324535857298872e-06, + "loss": 1.4461, + "step": 28757 + }, + { + "epoch": 5.23, + "learning_rate": 1.4323989807062249e-06, + "loss": 0.4316, + "step": 28758 + }, + { + "epoch": 5.23, + "learning_rate": 1.4323443756825628e-06, + "loss": 1.7917, + "step": 28759 + }, + { + "epoch": 5.23, + "learning_rate": 1.4322897706589006e-06, + "loss": 0.555, + "step": 28760 + }, + { + "epoch": 5.23, + "learning_rate": 1.4322351656352385e-06, + "loss": 0.5173, + "step": 28761 + }, + { + "epoch": 5.23, + "learning_rate": 1.4321805606115764e-06, + "loss": 0.0151, + "step": 28762 + }, + { + "epoch": 5.23, + "learning_rate": 1.4321259555879142e-06, + "loss": 1.6025, + "step": 28763 + }, + { + "epoch": 5.23, + "learning_rate": 1.432071350564252e-06, + "loss": 0.011, + "step": 28764 + }, + { + "epoch": 5.23, + "learning_rate": 1.4320167455405896e-06, + "loss": 0.0663, + "step": 28765 + }, + { + "epoch": 5.23, + "learning_rate": 1.4319621405169276e-06, + "loss": 0.0714, + "step": 28766 + }, + { + "epoch": 5.23, + "learning_rate": 1.4319075354932653e-06, + "loss": 0.6022, + "step": 28767 + }, + { + "epoch": 5.23, + "learning_rate": 1.4318529304696033e-06, + "loss": 3.5152, + "step": 28768 + }, + { + "epoch": 5.23, + "learning_rate": 1.4317983254459412e-06, + "loss": 0.3569, + "step": 28769 + }, + { + "epoch": 5.23, + "learning_rate": 1.431743720422279e-06, + "loss": 0.0227, + "step": 28770 + }, + { + "epoch": 5.23, + "learning_rate": 1.4316891153986167e-06, + "loss": 3.2377, + "step": 28771 + }, + { + "epoch": 5.23, + "learning_rate": 1.4316345103749544e-06, + "loss": 0.0829, + "step": 28772 + }, + { + "epoch": 5.23, + "learning_rate": 1.4315799053512923e-06, + "loss": 0.3572, + "step": 28773 + }, + { + "epoch": 5.23, + "learning_rate": 1.4315253003276303e-06, + "loss": 0.3122, + "step": 28774 + }, + { + "epoch": 5.23, + "learning_rate": 1.431470695303968e-06, + "loss": 1.6983, + "step": 28775 + }, + { + "epoch": 5.23, + "learning_rate": 1.431416090280306e-06, + "loss": 0.8269, + "step": 28776 + }, + { + "epoch": 5.23, + "learning_rate": 1.4313614852566437e-06, + "loss": 0.3881, + "step": 28777 + }, + { + "epoch": 5.23, + "learning_rate": 1.4313068802329814e-06, + "loss": 0.0066, + "step": 28778 + }, + { + "epoch": 5.23, + "learning_rate": 1.4312522752093192e-06, + "loss": 4.1092, + "step": 28779 + }, + { + "epoch": 5.23, + "learning_rate": 1.431197670185657e-06, + "loss": 0.008, + "step": 28780 + }, + { + "epoch": 5.23, + "learning_rate": 1.431143065161995e-06, + "loss": 1.9521, + "step": 28781 + }, + { + "epoch": 5.23, + "learning_rate": 1.4310884601383328e-06, + "loss": 0.4836, + "step": 28782 + }, + { + "epoch": 5.23, + "learning_rate": 1.4310338551146707e-06, + "loss": 0.362, + "step": 28783 + }, + { + "epoch": 5.23, + "learning_rate": 1.4309792500910084e-06, + "loss": 0.0052, + "step": 28784 + }, + { + "epoch": 5.23, + "learning_rate": 1.4309246450673462e-06, + "loss": 1.5297, + "step": 28785 + }, + { + "epoch": 5.23, + "learning_rate": 1.430870040043684e-06, + "loss": 0.2246, + "step": 28786 + }, + { + "epoch": 5.23, + "learning_rate": 1.4308154350200219e-06, + "loss": 0.0694, + "step": 28787 + }, + { + "epoch": 5.24, + "learning_rate": 1.4307608299963598e-06, + "loss": 0.0088, + "step": 28788 + }, + { + "epoch": 5.24, + "learning_rate": 1.4307062249726975e-06, + "loss": 2.5952, + "step": 28789 + }, + { + "epoch": 5.24, + "learning_rate": 1.4306516199490355e-06, + "loss": 1.5265, + "step": 28790 + }, + { + "epoch": 5.24, + "learning_rate": 1.4305970149253732e-06, + "loss": 0.0099, + "step": 28791 + }, + { + "epoch": 5.24, + "learning_rate": 1.430542409901711e-06, + "loss": 0.0985, + "step": 28792 + }, + { + "epoch": 5.24, + "learning_rate": 1.4304878048780487e-06, + "loss": 0.0154, + "step": 28793 + }, + { + "epoch": 5.24, + "learning_rate": 1.4304331998543866e-06, + "loss": 0.0613, + "step": 28794 + }, + { + "epoch": 5.24, + "learning_rate": 1.4303785948307246e-06, + "loss": 3.5257, + "step": 28795 + }, + { + "epoch": 5.24, + "learning_rate": 1.4303239898070623e-06, + "loss": 0.1864, + "step": 28796 + }, + { + "epoch": 5.24, + "learning_rate": 1.4302693847834002e-06, + "loss": 0.3175, + "step": 28797 + }, + { + "epoch": 5.24, + "learning_rate": 1.430214779759738e-06, + "loss": 3.087, + "step": 28798 + }, + { + "epoch": 5.24, + "learning_rate": 1.4301601747360757e-06, + "loss": 2.0099, + "step": 28799 + }, + { + "epoch": 5.24, + "learning_rate": 1.4301055697124134e-06, + "loss": 1.0758, + "step": 28800 + }, + { + "epoch": 5.24, + "learning_rate": 1.4300509646887514e-06, + "loss": 0.0154, + "step": 28801 + }, + { + "epoch": 5.24, + "learning_rate": 1.4299963596650893e-06, + "loss": 0.0279, + "step": 28802 + }, + { + "epoch": 5.24, + "learning_rate": 1.429941754641427e-06, + "loss": 0.0143, + "step": 28803 + }, + { + "epoch": 5.24, + "learning_rate": 1.4298871496177648e-06, + "loss": 0.2179, + "step": 28804 + }, + { + "epoch": 5.24, + "learning_rate": 1.4298325445941027e-06, + "loss": 0.0594, + "step": 28805 + }, + { + "epoch": 5.24, + "learning_rate": 1.4297779395704404e-06, + "loss": 1.5802, + "step": 28806 + }, + { + "epoch": 5.24, + "learning_rate": 1.4297233345467782e-06, + "loss": 0.0664, + "step": 28807 + }, + { + "epoch": 5.24, + "learning_rate": 1.4296687295231161e-06, + "loss": 0.324, + "step": 28808 + }, + { + "epoch": 5.24, + "learning_rate": 1.429614124499454e-06, + "loss": 0.158, + "step": 28809 + }, + { + "epoch": 5.24, + "learning_rate": 1.4295595194757918e-06, + "loss": 0.7666, + "step": 28810 + }, + { + "epoch": 5.24, + "learning_rate": 1.4295049144521295e-06, + "loss": 0.0169, + "step": 28811 + }, + { + "epoch": 5.24, + "learning_rate": 1.4294503094284675e-06, + "loss": 0.0137, + "step": 28812 + }, + { + "epoch": 5.24, + "learning_rate": 1.4293957044048052e-06, + "loss": 0.8468, + "step": 28813 + }, + { + "epoch": 5.24, + "learning_rate": 1.4293410993811431e-06, + "loss": 1.3473, + "step": 28814 + }, + { + "epoch": 5.24, + "learning_rate": 1.429286494357481e-06, + "loss": 0.0126, + "step": 28815 + }, + { + "epoch": 5.24, + "learning_rate": 1.4292318893338188e-06, + "loss": 0.5601, + "step": 28816 + }, + { + "epoch": 5.24, + "learning_rate": 1.4291772843101566e-06, + "loss": 0.1579, + "step": 28817 + }, + { + "epoch": 5.24, + "learning_rate": 1.4291226792864943e-06, + "loss": 2.2398, + "step": 28818 + }, + { + "epoch": 5.24, + "learning_rate": 1.4290680742628322e-06, + "loss": 0.0495, + "step": 28819 + }, + { + "epoch": 5.24, + "learning_rate": 1.42901346923917e-06, + "loss": 0.6679, + "step": 28820 + }, + { + "epoch": 5.24, + "learning_rate": 1.428958864215508e-06, + "loss": 0.4097, + "step": 28821 + }, + { + "epoch": 5.24, + "learning_rate": 1.4289042591918459e-06, + "loss": 0.8194, + "step": 28822 + }, + { + "epoch": 5.24, + "learning_rate": 1.4288496541681836e-06, + "loss": 1.6894, + "step": 28823 + }, + { + "epoch": 5.24, + "learning_rate": 1.4287950491445213e-06, + "loss": 2.1888, + "step": 28824 + }, + { + "epoch": 5.24, + "learning_rate": 1.428740444120859e-06, + "loss": 0.5784, + "step": 28825 + }, + { + "epoch": 5.24, + "learning_rate": 1.428685839097197e-06, + "loss": 0.0564, + "step": 28826 + }, + { + "epoch": 5.24, + "learning_rate": 1.4286312340735347e-06, + "loss": 1.1815, + "step": 28827 + }, + { + "epoch": 5.24, + "learning_rate": 1.4285766290498727e-06, + "loss": 0.4338, + "step": 28828 + }, + { + "epoch": 5.24, + "learning_rate": 1.4285220240262106e-06, + "loss": 0.0049, + "step": 28829 + }, + { + "epoch": 5.24, + "learning_rate": 1.4284674190025483e-06, + "loss": 1.7311, + "step": 28830 + }, + { + "epoch": 5.24, + "learning_rate": 1.428412813978886e-06, + "loss": 0.5981, + "step": 28831 + }, + { + "epoch": 5.24, + "learning_rate": 1.4283582089552238e-06, + "loss": 0.7821, + "step": 28832 + }, + { + "epoch": 5.24, + "learning_rate": 1.4283036039315617e-06, + "loss": 0.2311, + "step": 28833 + }, + { + "epoch": 5.24, + "learning_rate": 1.4282489989078995e-06, + "loss": 0.4373, + "step": 28834 + }, + { + "epoch": 5.24, + "learning_rate": 1.4281943938842374e-06, + "loss": 1.7473, + "step": 28835 + }, + { + "epoch": 5.24, + "learning_rate": 1.4281397888605754e-06, + "loss": 1.8158, + "step": 28836 + }, + { + "epoch": 5.24, + "learning_rate": 1.428085183836913e-06, + "loss": 0.1641, + "step": 28837 + }, + { + "epoch": 5.24, + "learning_rate": 1.4280305788132508e-06, + "loss": 0.6367, + "step": 28838 + }, + { + "epoch": 5.24, + "learning_rate": 1.4279759737895886e-06, + "loss": 1.0167, + "step": 28839 + }, + { + "epoch": 5.24, + "learning_rate": 1.4279213687659265e-06, + "loss": 0.6827, + "step": 28840 + }, + { + "epoch": 5.24, + "learning_rate": 1.4278667637422642e-06, + "loss": 1.2967, + "step": 28841 + }, + { + "epoch": 5.24, + "learning_rate": 1.4278121587186022e-06, + "loss": 1.6972, + "step": 28842 + }, + { + "epoch": 5.25, + "learning_rate": 1.4277575536949401e-06, + "loss": 1.3459, + "step": 28843 + }, + { + "epoch": 5.25, + "learning_rate": 1.4277029486712779e-06, + "loss": 0.977, + "step": 28844 + }, + { + "epoch": 5.25, + "learning_rate": 1.4276483436476156e-06, + "loss": 1.2079, + "step": 28845 + }, + { + "epoch": 5.25, + "learning_rate": 1.4275937386239533e-06, + "loss": 0.5529, + "step": 28846 + }, + { + "epoch": 5.25, + "learning_rate": 1.4275391336002913e-06, + "loss": 2.1893, + "step": 28847 + }, + { + "epoch": 5.25, + "learning_rate": 1.427484528576629e-06, + "loss": 0.9756, + "step": 28848 + }, + { + "epoch": 5.25, + "learning_rate": 1.427429923552967e-06, + "loss": 0.0067, + "step": 28849 + }, + { + "epoch": 5.25, + "learning_rate": 1.4273753185293049e-06, + "loss": 2.6518, + "step": 28850 + }, + { + "epoch": 5.25, + "learning_rate": 1.4273207135056426e-06, + "loss": 0.2234, + "step": 28851 + }, + { + "epoch": 5.25, + "learning_rate": 1.4272661084819803e-06, + "loss": 0.8746, + "step": 28852 + }, + { + "epoch": 5.25, + "learning_rate": 1.427211503458318e-06, + "loss": 1.5249, + "step": 28853 + }, + { + "epoch": 5.25, + "learning_rate": 1.427156898434656e-06, + "loss": 0.0707, + "step": 28854 + }, + { + "epoch": 5.25, + "learning_rate": 1.4271022934109937e-06, + "loss": 0.2531, + "step": 28855 + }, + { + "epoch": 5.25, + "learning_rate": 1.4270476883873317e-06, + "loss": 0.1936, + "step": 28856 + }, + { + "epoch": 5.25, + "learning_rate": 1.4269930833636696e-06, + "loss": 2.344, + "step": 28857 + }, + { + "epoch": 5.25, + "learning_rate": 1.4269384783400074e-06, + "loss": 0.0592, + "step": 28858 + }, + { + "epoch": 5.25, + "learning_rate": 1.426883873316345e-06, + "loss": 0.6931, + "step": 28859 + }, + { + "epoch": 5.25, + "learning_rate": 1.4268292682926828e-06, + "loss": 0.4812, + "step": 28860 + }, + { + "epoch": 5.25, + "learning_rate": 1.4267746632690208e-06, + "loss": 0.0131, + "step": 28861 + }, + { + "epoch": 5.25, + "learning_rate": 1.4267200582453587e-06, + "loss": 0.1714, + "step": 28862 + }, + { + "epoch": 5.25, + "learning_rate": 1.4266654532216964e-06, + "loss": 0.0161, + "step": 28863 + }, + { + "epoch": 5.25, + "learning_rate": 1.4266108481980344e-06, + "loss": 1.696, + "step": 28864 + }, + { + "epoch": 5.25, + "learning_rate": 1.4265562431743721e-06, + "loss": 0.0374, + "step": 28865 + }, + { + "epoch": 5.25, + "learning_rate": 1.4265016381507099e-06, + "loss": 0.256, + "step": 28866 + }, + { + "epoch": 5.25, + "learning_rate": 1.4264470331270476e-06, + "loss": 0.8914, + "step": 28867 + }, + { + "epoch": 5.25, + "learning_rate": 1.4263924281033855e-06, + "loss": 1.0113, + "step": 28868 + }, + { + "epoch": 5.25, + "learning_rate": 1.4263378230797235e-06, + "loss": 0.9009, + "step": 28869 + }, + { + "epoch": 5.25, + "learning_rate": 1.4262832180560612e-06, + "loss": 0.2117, + "step": 28870 + }, + { + "epoch": 5.25, + "learning_rate": 1.4262286130323991e-06, + "loss": 0.0324, + "step": 28871 + }, + { + "epoch": 5.25, + "learning_rate": 1.4261740080087369e-06, + "loss": 0.0238, + "step": 28872 + }, + { + "epoch": 5.25, + "learning_rate": 1.4261194029850746e-06, + "loss": 0.4657, + "step": 28873 + }, + { + "epoch": 5.25, + "learning_rate": 1.4260647979614123e-06, + "loss": 0.0932, + "step": 28874 + }, + { + "epoch": 5.25, + "learning_rate": 1.4260101929377503e-06, + "loss": 0.4101, + "step": 28875 + }, + { + "epoch": 5.25, + "learning_rate": 1.4259555879140882e-06, + "loss": 0.6938, + "step": 28876 + }, + { + "epoch": 5.25, + "learning_rate": 1.425900982890426e-06, + "loss": 0.0133, + "step": 28877 + }, + { + "epoch": 5.25, + "learning_rate": 1.4258463778667637e-06, + "loss": 0.0373, + "step": 28878 + }, + { + "epoch": 5.25, + "learning_rate": 1.4257917728431016e-06, + "loss": 0.3577, + "step": 28879 + }, + { + "epoch": 5.25, + "learning_rate": 1.4257371678194394e-06, + "loss": 1.2708, + "step": 28880 + }, + { + "epoch": 5.25, + "learning_rate": 1.425682562795777e-06, + "loss": 0.0612, + "step": 28881 + }, + { + "epoch": 5.25, + "learning_rate": 1.425627957772115e-06, + "loss": 0.3023, + "step": 28882 + }, + { + "epoch": 5.25, + "learning_rate": 1.425573352748453e-06, + "loss": 0.0728, + "step": 28883 + }, + { + "epoch": 5.25, + "learning_rate": 1.4255187477247907e-06, + "loss": 1.9317, + "step": 28884 + }, + { + "epoch": 5.25, + "learning_rate": 1.4254641427011284e-06, + "loss": 2.1375, + "step": 28885 + }, + { + "epoch": 5.25, + "learning_rate": 1.4254095376774664e-06, + "loss": 0.7589, + "step": 28886 + }, + { + "epoch": 5.25, + "learning_rate": 1.4253549326538041e-06, + "loss": 1.1148, + "step": 28887 + }, + { + "epoch": 5.25, + "learning_rate": 1.4253003276301419e-06, + "loss": 0.1232, + "step": 28888 + }, + { + "epoch": 5.25, + "learning_rate": 1.4252457226064798e-06, + "loss": 0.2846, + "step": 28889 + }, + { + "epoch": 5.25, + "learning_rate": 1.4251911175828177e-06, + "loss": 0.2926, + "step": 28890 + }, + { + "epoch": 5.25, + "learning_rate": 1.4251365125591555e-06, + "loss": 2.2568, + "step": 28891 + }, + { + "epoch": 5.25, + "learning_rate": 1.4250819075354932e-06, + "loss": 0.5281, + "step": 28892 + }, + { + "epoch": 5.25, + "learning_rate": 1.4250273025118311e-06, + "loss": 1.657, + "step": 28893 + }, + { + "epoch": 5.25, + "learning_rate": 1.4249726974881689e-06, + "loss": 0.2062, + "step": 28894 + }, + { + "epoch": 5.25, + "learning_rate": 1.4249180924645066e-06, + "loss": 1.3346, + "step": 28895 + }, + { + "epoch": 5.25, + "learning_rate": 1.4248634874408446e-06, + "loss": 0.1021, + "step": 28896 + }, + { + "epoch": 5.25, + "learning_rate": 1.4248088824171825e-06, + "loss": 0.1852, + "step": 28897 + }, + { + "epoch": 5.26, + "learning_rate": 1.4247542773935202e-06, + "loss": 0.477, + "step": 28898 + }, + { + "epoch": 5.26, + "learning_rate": 1.424699672369858e-06, + "loss": 0.0305, + "step": 28899 + }, + { + "epoch": 5.26, + "learning_rate": 1.424645067346196e-06, + "loss": 0.2691, + "step": 28900 + }, + { + "epoch": 5.26, + "learning_rate": 1.4245904623225336e-06, + "loss": 0.5778, + "step": 28901 + }, + { + "epoch": 5.26, + "learning_rate": 1.4245358572988716e-06, + "loss": 0.3004, + "step": 28902 + }, + { + "epoch": 5.26, + "learning_rate": 1.4244812522752095e-06, + "loss": 2.5194, + "step": 28903 + }, + { + "epoch": 5.26, + "learning_rate": 1.4244266472515473e-06, + "loss": 0.7849, + "step": 28904 + }, + { + "epoch": 5.26, + "learning_rate": 1.424372042227885e-06, + "loss": 0.8566, + "step": 28905 + }, + { + "epoch": 5.26, + "learning_rate": 1.4243174372042227e-06, + "loss": 0.1144, + "step": 28906 + }, + { + "epoch": 5.26, + "learning_rate": 1.4242628321805607e-06, + "loss": 0.2013, + "step": 28907 + }, + { + "epoch": 5.26, + "learning_rate": 1.4242082271568984e-06, + "loss": 1.7478, + "step": 28908 + }, + { + "epoch": 5.26, + "learning_rate": 1.4241536221332363e-06, + "loss": 0.0487, + "step": 28909 + }, + { + "epoch": 5.26, + "learning_rate": 1.4240990171095743e-06, + "loss": 0.8438, + "step": 28910 + }, + { + "epoch": 5.26, + "learning_rate": 1.424044412085912e-06, + "loss": 0.6698, + "step": 28911 + }, + { + "epoch": 5.26, + "learning_rate": 1.4239898070622497e-06, + "loss": 0.0128, + "step": 28912 + }, + { + "epoch": 5.26, + "learning_rate": 1.4239352020385875e-06, + "loss": 0.0122, + "step": 28913 + }, + { + "epoch": 5.26, + "learning_rate": 1.4238805970149254e-06, + "loss": 0.2372, + "step": 28914 + }, + { + "epoch": 5.26, + "learning_rate": 1.4238259919912632e-06, + "loss": 2.1102, + "step": 28915 + }, + { + "epoch": 5.26, + "learning_rate": 1.423771386967601e-06, + "loss": 0.2552, + "step": 28916 + }, + { + "epoch": 5.26, + "learning_rate": 1.423716781943939e-06, + "loss": 0.185, + "step": 28917 + }, + { + "epoch": 5.26, + "learning_rate": 1.4236621769202768e-06, + "loss": 1.3571, + "step": 28918 + }, + { + "epoch": 5.26, + "learning_rate": 1.4236075718966145e-06, + "loss": 0.4863, + "step": 28919 + }, + { + "epoch": 5.26, + "learning_rate": 1.4235529668729522e-06, + "loss": 0.8653, + "step": 28920 + }, + { + "epoch": 5.26, + "learning_rate": 1.4234983618492902e-06, + "loss": 1.6463, + "step": 28921 + }, + { + "epoch": 5.26, + "learning_rate": 1.423443756825628e-06, + "loss": 0.6131, + "step": 28922 + }, + { + "epoch": 5.26, + "learning_rate": 1.4233891518019659e-06, + "loss": 0.1649, + "step": 28923 + }, + { + "epoch": 5.26, + "learning_rate": 1.4233345467783038e-06, + "loss": 0.0037, + "step": 28924 + }, + { + "epoch": 5.26, + "learning_rate": 1.4232799417546415e-06, + "loss": 1.6346, + "step": 28925 + }, + { + "epoch": 5.26, + "learning_rate": 1.4232253367309793e-06, + "loss": 0.0176, + "step": 28926 + }, + { + "epoch": 5.26, + "learning_rate": 1.423170731707317e-06, + "loss": 2.1612, + "step": 28927 + }, + { + "epoch": 5.26, + "learning_rate": 1.423116126683655e-06, + "loss": 0.026, + "step": 28928 + }, + { + "epoch": 5.26, + "learning_rate": 1.4230615216599927e-06, + "loss": 0.1182, + "step": 28929 + }, + { + "epoch": 5.26, + "learning_rate": 1.4230069166363306e-06, + "loss": 0.8793, + "step": 28930 + }, + { + "epoch": 5.26, + "learning_rate": 1.4229523116126686e-06, + "loss": 0.5889, + "step": 28931 + }, + { + "epoch": 5.26, + "learning_rate": 1.4228977065890063e-06, + "loss": 1.713, + "step": 28932 + }, + { + "epoch": 5.26, + "learning_rate": 1.422843101565344e-06, + "loss": 0.8469, + "step": 28933 + }, + { + "epoch": 5.26, + "learning_rate": 1.4227884965416817e-06, + "loss": 0.842, + "step": 28934 + }, + { + "epoch": 5.26, + "learning_rate": 1.4227338915180197e-06, + "loss": 3.2029, + "step": 28935 + }, + { + "epoch": 5.26, + "learning_rate": 1.4226792864943574e-06, + "loss": 0.5337, + "step": 28936 + }, + { + "epoch": 5.26, + "learning_rate": 1.4226246814706954e-06, + "loss": 0.4542, + "step": 28937 + }, + { + "epoch": 5.26, + "learning_rate": 1.4225700764470333e-06, + "loss": 0.3477, + "step": 28938 + }, + { + "epoch": 5.26, + "learning_rate": 1.422515471423371e-06, + "loss": 0.2815, + "step": 28939 + }, + { + "epoch": 5.26, + "learning_rate": 1.4224608663997088e-06, + "loss": 1.862, + "step": 28940 + }, + { + "epoch": 5.26, + "learning_rate": 1.4224062613760465e-06, + "loss": 1.9617, + "step": 28941 + }, + { + "epoch": 5.26, + "learning_rate": 1.4223516563523844e-06, + "loss": 0.2973, + "step": 28942 + }, + { + "epoch": 5.26, + "learning_rate": 1.4222970513287224e-06, + "loss": 0.078, + "step": 28943 + }, + { + "epoch": 5.26, + "learning_rate": 1.4222424463050601e-06, + "loss": 0.0562, + "step": 28944 + }, + { + "epoch": 5.26, + "learning_rate": 1.422187841281398e-06, + "loss": 1.3263, + "step": 28945 + }, + { + "epoch": 5.26, + "learning_rate": 1.4221332362577358e-06, + "loss": 0.0313, + "step": 28946 + }, + { + "epoch": 5.26, + "learning_rate": 1.4220786312340735e-06, + "loss": 0.0739, + "step": 28947 + }, + { + "epoch": 5.26, + "learning_rate": 1.4220240262104113e-06, + "loss": 0.7227, + "step": 28948 + }, + { + "epoch": 5.26, + "learning_rate": 1.4219694211867492e-06, + "loss": 0.0267, + "step": 28949 + }, + { + "epoch": 5.26, + "learning_rate": 1.4219148161630871e-06, + "loss": 0.8475, + "step": 28950 + }, + { + "epoch": 5.26, + "learning_rate": 1.4218602111394249e-06, + "loss": 1.6174, + "step": 28951 + }, + { + "epoch": 5.26, + "learning_rate": 1.4218056061157628e-06, + "loss": 1.3887, + "step": 28952 + }, + { + "epoch": 5.27, + "learning_rate": 1.4217510010921006e-06, + "loss": 0.5135, + "step": 28953 + }, + { + "epoch": 5.27, + "learning_rate": 1.4216963960684383e-06, + "loss": 0.3718, + "step": 28954 + }, + { + "epoch": 5.27, + "learning_rate": 1.421641791044776e-06, + "loss": 0.3227, + "step": 28955 + }, + { + "epoch": 5.27, + "learning_rate": 1.421587186021114e-06, + "loss": 1.3527, + "step": 28956 + }, + { + "epoch": 5.27, + "learning_rate": 1.421532580997452e-06, + "loss": 0.6782, + "step": 28957 + }, + { + "epoch": 5.27, + "learning_rate": 1.4214779759737896e-06, + "loss": 0.0463, + "step": 28958 + }, + { + "epoch": 5.27, + "learning_rate": 1.4214233709501274e-06, + "loss": 0.2253, + "step": 28959 + }, + { + "epoch": 5.27, + "learning_rate": 1.4213687659264653e-06, + "loss": 1.2236, + "step": 28960 + }, + { + "epoch": 5.27, + "learning_rate": 1.421314160902803e-06, + "loss": 1.6136, + "step": 28961 + }, + { + "epoch": 5.27, + "learning_rate": 1.4212595558791408e-06, + "loss": 1.4333, + "step": 28962 + }, + { + "epoch": 5.27, + "learning_rate": 1.4212049508554787e-06, + "loss": 0.4787, + "step": 28963 + }, + { + "epoch": 5.27, + "learning_rate": 1.4211503458318167e-06, + "loss": 1.2649, + "step": 28964 + }, + { + "epoch": 5.27, + "learning_rate": 1.4210957408081544e-06, + "loss": 0.005, + "step": 28965 + }, + { + "epoch": 5.27, + "learning_rate": 1.4210411357844921e-06, + "loss": 0.9954, + "step": 28966 + }, + { + "epoch": 5.27, + "learning_rate": 1.42098653076083e-06, + "loss": 0.3513, + "step": 28967 + }, + { + "epoch": 5.27, + "learning_rate": 1.4209319257371678e-06, + "loss": 0.0217, + "step": 28968 + }, + { + "epoch": 5.27, + "learning_rate": 1.4208773207135055e-06, + "loss": 2.0859, + "step": 28969 + }, + { + "epoch": 5.27, + "learning_rate": 1.4208227156898435e-06, + "loss": 0.9965, + "step": 28970 + }, + { + "epoch": 5.27, + "learning_rate": 1.4207681106661814e-06, + "loss": 0.1709, + "step": 28971 + }, + { + "epoch": 5.27, + "learning_rate": 1.4207135056425191e-06, + "loss": 0.53, + "step": 28972 + }, + { + "epoch": 5.27, + "learning_rate": 1.4206589006188569e-06, + "loss": 0.4126, + "step": 28973 + }, + { + "epoch": 5.27, + "learning_rate": 1.4206042955951948e-06, + "loss": 0.3649, + "step": 28974 + }, + { + "epoch": 5.27, + "learning_rate": 1.4205496905715326e-06, + "loss": 0.1431, + "step": 28975 + }, + { + "epoch": 5.27, + "learning_rate": 1.4204950855478703e-06, + "loss": 0.8497, + "step": 28976 + }, + { + "epoch": 5.27, + "learning_rate": 1.4204404805242082e-06, + "loss": 0.0105, + "step": 28977 + }, + { + "epoch": 5.27, + "learning_rate": 1.4203858755005462e-06, + "loss": 0.1454, + "step": 28978 + }, + { + "epoch": 5.27, + "learning_rate": 1.420331270476884e-06, + "loss": 1.5135, + "step": 28979 + }, + { + "epoch": 5.27, + "learning_rate": 1.4202766654532216e-06, + "loss": 0.2589, + "step": 28980 + }, + { + "epoch": 5.27, + "learning_rate": 1.4202220604295596e-06, + "loss": 2.7666, + "step": 28981 + }, + { + "epoch": 5.27, + "learning_rate": 1.4201674554058973e-06, + "loss": 0.0243, + "step": 28982 + }, + { + "epoch": 5.27, + "learning_rate": 1.4201128503822353e-06, + "loss": 0.5621, + "step": 28983 + }, + { + "epoch": 5.27, + "learning_rate": 1.4200582453585732e-06, + "loss": 0.834, + "step": 28984 + }, + { + "epoch": 5.27, + "learning_rate": 1.420003640334911e-06, + "loss": 1.6008, + "step": 28985 + }, + { + "epoch": 5.27, + "learning_rate": 1.4199490353112487e-06, + "loss": 2.0802, + "step": 28986 + }, + { + "epoch": 5.27, + "learning_rate": 1.4198944302875864e-06, + "loss": 0.0048, + "step": 28987 + }, + { + "epoch": 5.27, + "learning_rate": 1.4198398252639243e-06, + "loss": 0.0361, + "step": 28988 + }, + { + "epoch": 5.27, + "learning_rate": 1.419785220240262e-06, + "loss": 0.0498, + "step": 28989 + }, + { + "epoch": 5.27, + "learning_rate": 1.4197306152166e-06, + "loss": 0.2101, + "step": 28990 + }, + { + "epoch": 5.27, + "learning_rate": 1.419676010192938e-06, + "loss": 0.0034, + "step": 28991 + }, + { + "epoch": 5.27, + "learning_rate": 1.4196214051692757e-06, + "loss": 2.7257, + "step": 28992 + }, + { + "epoch": 5.27, + "learning_rate": 1.4195668001456134e-06, + "loss": 0.0134, + "step": 28993 + }, + { + "epoch": 5.27, + "learning_rate": 1.4195121951219512e-06, + "loss": 0.1262, + "step": 28994 + }, + { + "epoch": 5.27, + "learning_rate": 1.419457590098289e-06, + "loss": 0.0128, + "step": 28995 + }, + { + "epoch": 5.27, + "learning_rate": 1.4194029850746268e-06, + "loss": 0.2057, + "step": 28996 + }, + { + "epoch": 5.27, + "learning_rate": 1.4193483800509648e-06, + "loss": 3.1486, + "step": 28997 + }, + { + "epoch": 5.27, + "learning_rate": 1.4192937750273027e-06, + "loss": 0.013, + "step": 28998 + }, + { + "epoch": 5.27, + "learning_rate": 1.4192391700036404e-06, + "loss": 0.0223, + "step": 28999 + }, + { + "epoch": 5.27, + "learning_rate": 1.4191845649799782e-06, + "loss": 0.2188, + "step": 29000 + }, + { + "epoch": 5.27, + "learning_rate": 1.419129959956316e-06, + "loss": 0.0142, + "step": 29001 + }, + { + "epoch": 5.27, + "learning_rate": 1.4190753549326539e-06, + "loss": 2.2788, + "step": 29002 + }, + { + "epoch": 5.27, + "learning_rate": 1.4190207499089916e-06, + "loss": 0.213, + "step": 29003 + }, + { + "epoch": 5.27, + "learning_rate": 1.4189661448853295e-06, + "loss": 0.2228, + "step": 29004 + }, + { + "epoch": 5.27, + "learning_rate": 1.4189115398616675e-06, + "loss": 0.9399, + "step": 29005 + }, + { + "epoch": 5.27, + "learning_rate": 1.4188569348380052e-06, + "loss": 0.6322, + "step": 29006 + }, + { + "epoch": 5.27, + "learning_rate": 1.418802329814343e-06, + "loss": 2.3551, + "step": 29007 + }, + { + "epoch": 5.28, + "learning_rate": 1.4187477247906807e-06, + "loss": 2.6861, + "step": 29008 + }, + { + "epoch": 5.28, + "learning_rate": 1.4186931197670186e-06, + "loss": 1.5378, + "step": 29009 + }, + { + "epoch": 5.28, + "learning_rate": 1.4186385147433563e-06, + "loss": 0.9961, + "step": 29010 + }, + { + "epoch": 5.28, + "learning_rate": 1.4185839097196943e-06, + "loss": 0.5113, + "step": 29011 + }, + { + "epoch": 5.28, + "learning_rate": 1.4185293046960322e-06, + "loss": 0.0928, + "step": 29012 + }, + { + "epoch": 5.28, + "learning_rate": 1.41847469967237e-06, + "loss": 0.0289, + "step": 29013 + }, + { + "epoch": 5.28, + "learning_rate": 1.4184200946487077e-06, + "loss": 1.3518, + "step": 29014 + }, + { + "epoch": 5.28, + "learning_rate": 1.4183654896250454e-06, + "loss": 1.3353, + "step": 29015 + }, + { + "epoch": 5.28, + "learning_rate": 1.4183108846013834e-06, + "loss": 2.6684, + "step": 29016 + }, + { + "epoch": 5.28, + "learning_rate": 1.418256279577721e-06, + "loss": 0.2006, + "step": 29017 + }, + { + "epoch": 5.28, + "learning_rate": 1.418201674554059e-06, + "loss": 0.033, + "step": 29018 + }, + { + "epoch": 5.28, + "learning_rate": 1.418147069530397e-06, + "loss": 0.2416, + "step": 29019 + }, + { + "epoch": 5.28, + "learning_rate": 1.4180924645067347e-06, + "loss": 1.1483, + "step": 29020 + }, + { + "epoch": 5.28, + "learning_rate": 1.4180378594830724e-06, + "loss": 0.9496, + "step": 29021 + }, + { + "epoch": 5.28, + "learning_rate": 1.4179832544594102e-06, + "loss": 0.2277, + "step": 29022 + }, + { + "epoch": 5.28, + "learning_rate": 1.4179286494357481e-06, + "loss": 0.2212, + "step": 29023 + }, + { + "epoch": 5.28, + "learning_rate": 1.4178740444120859e-06, + "loss": 0.1609, + "step": 29024 + }, + { + "epoch": 5.28, + "learning_rate": 1.4178194393884238e-06, + "loss": 0.095, + "step": 29025 + }, + { + "epoch": 5.28, + "learning_rate": 1.4177648343647617e-06, + "loss": 0.0117, + "step": 29026 + }, + { + "epoch": 5.28, + "learning_rate": 1.4177102293410995e-06, + "loss": 1.063, + "step": 29027 + }, + { + "epoch": 5.28, + "learning_rate": 1.4176556243174372e-06, + "loss": 0.8555, + "step": 29028 + }, + { + "epoch": 5.28, + "learning_rate": 1.417601019293775e-06, + "loss": 0.0066, + "step": 29029 + }, + { + "epoch": 5.28, + "learning_rate": 1.4175464142701129e-06, + "loss": 0.1498, + "step": 29030 + }, + { + "epoch": 5.28, + "learning_rate": 1.4174918092464508e-06, + "loss": 0.5085, + "step": 29031 + }, + { + "epoch": 5.28, + "learning_rate": 1.4174372042227886e-06, + "loss": 0.1931, + "step": 29032 + }, + { + "epoch": 5.28, + "learning_rate": 1.4173825991991265e-06, + "loss": 2.2848, + "step": 29033 + }, + { + "epoch": 5.28, + "learning_rate": 1.4173279941754642e-06, + "loss": 0.6533, + "step": 29034 + }, + { + "epoch": 5.28, + "learning_rate": 1.417273389151802e-06, + "loss": 0.2999, + "step": 29035 + }, + { + "epoch": 5.28, + "learning_rate": 1.4172187841281397e-06, + "loss": 0.0085, + "step": 29036 + }, + { + "epoch": 5.28, + "learning_rate": 1.4171641791044776e-06, + "loss": 0.1146, + "step": 29037 + }, + { + "epoch": 5.28, + "learning_rate": 1.4171095740808156e-06, + "loss": 0.4755, + "step": 29038 + }, + { + "epoch": 5.28, + "learning_rate": 1.4170549690571533e-06, + "loss": 0.587, + "step": 29039 + }, + { + "epoch": 5.28, + "learning_rate": 1.417000364033491e-06, + "loss": 2.133, + "step": 29040 + }, + { + "epoch": 5.28, + "learning_rate": 1.416945759009829e-06, + "loss": 1.9173, + "step": 29041 + }, + { + "epoch": 5.28, + "learning_rate": 1.4168911539861667e-06, + "loss": 0.2329, + "step": 29042 + }, + { + "epoch": 5.28, + "learning_rate": 1.4168365489625044e-06, + "loss": 0.0078, + "step": 29043 + }, + { + "epoch": 5.28, + "learning_rate": 1.4167819439388424e-06, + "loss": 0.1761, + "step": 29044 + }, + { + "epoch": 5.28, + "learning_rate": 1.4167273389151803e-06, + "loss": 0.0276, + "step": 29045 + }, + { + "epoch": 5.28, + "learning_rate": 1.416672733891518e-06, + "loss": 0.3531, + "step": 29046 + }, + { + "epoch": 5.28, + "learning_rate": 1.4166181288678558e-06, + "loss": 0.3452, + "step": 29047 + }, + { + "epoch": 5.28, + "learning_rate": 1.4165635238441937e-06, + "loss": 1.8851, + "step": 29048 + }, + { + "epoch": 5.28, + "learning_rate": 1.4165089188205315e-06, + "loss": 0.8543, + "step": 29049 + }, + { + "epoch": 5.28, + "learning_rate": 1.4164543137968692e-06, + "loss": 0.0122, + "step": 29050 + }, + { + "epoch": 5.28, + "learning_rate": 1.4163997087732071e-06, + "loss": 0.0119, + "step": 29051 + }, + { + "epoch": 5.28, + "learning_rate": 1.416345103749545e-06, + "loss": 0.2973, + "step": 29052 + }, + { + "epoch": 5.28, + "learning_rate": 1.4162904987258828e-06, + "loss": 1.5278, + "step": 29053 + }, + { + "epoch": 5.28, + "learning_rate": 1.4162358937022206e-06, + "loss": 0.0655, + "step": 29054 + }, + { + "epoch": 5.28, + "learning_rate": 1.4161812886785585e-06, + "loss": 0.6531, + "step": 29055 + }, + { + "epoch": 5.28, + "learning_rate": 1.4161266836548962e-06, + "loss": 1.2076, + "step": 29056 + }, + { + "epoch": 5.28, + "learning_rate": 1.416072078631234e-06, + "loss": 0.0495, + "step": 29057 + }, + { + "epoch": 5.28, + "learning_rate": 1.416017473607572e-06, + "loss": 0.3495, + "step": 29058 + }, + { + "epoch": 5.28, + "learning_rate": 1.4159628685839098e-06, + "loss": 0.1067, + "step": 29059 + }, + { + "epoch": 5.28, + "learning_rate": 1.4159082635602476e-06, + "loss": 0.3705, + "step": 29060 + }, + { + "epoch": 5.28, + "learning_rate": 1.4158536585365853e-06, + "loss": 0.4038, + "step": 29061 + }, + { + "epoch": 5.28, + "learning_rate": 1.4157990535129233e-06, + "loss": 0.0072, + "step": 29062 + }, + { + "epoch": 5.29, + "learning_rate": 1.415744448489261e-06, + "loss": 1.9079, + "step": 29063 + }, + { + "epoch": 5.29, + "learning_rate": 1.4156898434655987e-06, + "loss": 0.0485, + "step": 29064 + }, + { + "epoch": 5.29, + "learning_rate": 1.4156352384419367e-06, + "loss": 1.296, + "step": 29065 + }, + { + "epoch": 5.29, + "learning_rate": 1.4155806334182746e-06, + "loss": 0.4733, + "step": 29066 + }, + { + "epoch": 5.29, + "learning_rate": 1.4155260283946123e-06, + "loss": 0.9591, + "step": 29067 + }, + { + "epoch": 5.29, + "learning_rate": 1.41547142337095e-06, + "loss": 0.2341, + "step": 29068 + }, + { + "epoch": 5.29, + "learning_rate": 1.415416818347288e-06, + "loss": 2.8478, + "step": 29069 + }, + { + "epoch": 5.29, + "learning_rate": 1.4153622133236257e-06, + "loss": 0.828, + "step": 29070 + }, + { + "epoch": 5.29, + "learning_rate": 1.4153076082999637e-06, + "loss": 0.4488, + "step": 29071 + }, + { + "epoch": 5.29, + "learning_rate": 1.4152530032763016e-06, + "loss": 0.494, + "step": 29072 + }, + { + "epoch": 5.29, + "learning_rate": 1.4151983982526394e-06, + "loss": 0.095, + "step": 29073 + }, + { + "epoch": 5.29, + "learning_rate": 1.415143793228977e-06, + "loss": 0.5754, + "step": 29074 + }, + { + "epoch": 5.29, + "learning_rate": 1.4150891882053148e-06, + "loss": 0.3159, + "step": 29075 + }, + { + "epoch": 5.29, + "learning_rate": 1.4150345831816528e-06, + "loss": 0.3978, + "step": 29076 + }, + { + "epoch": 5.29, + "learning_rate": 1.4149799781579905e-06, + "loss": 1.5389, + "step": 29077 + }, + { + "epoch": 5.29, + "learning_rate": 1.4149253731343284e-06, + "loss": 0.0288, + "step": 29078 + }, + { + "epoch": 5.29, + "learning_rate": 1.4148707681106664e-06, + "loss": 0.1969, + "step": 29079 + }, + { + "epoch": 5.29, + "learning_rate": 1.4148161630870041e-06, + "loss": 0.259, + "step": 29080 + }, + { + "epoch": 5.29, + "learning_rate": 1.4147615580633418e-06, + "loss": 0.4601, + "step": 29081 + }, + { + "epoch": 5.29, + "learning_rate": 1.4147069530396796e-06, + "loss": 0.2637, + "step": 29082 + }, + { + "epoch": 5.29, + "learning_rate": 1.4146523480160175e-06, + "loss": 0.2578, + "step": 29083 + }, + { + "epoch": 5.29, + "learning_rate": 1.4145977429923553e-06, + "loss": 0.2365, + "step": 29084 + }, + { + "epoch": 5.29, + "learning_rate": 1.4145431379686932e-06, + "loss": 0.1861, + "step": 29085 + }, + { + "epoch": 5.29, + "learning_rate": 1.4144885329450311e-06, + "loss": 0.3359, + "step": 29086 + }, + { + "epoch": 5.29, + "learning_rate": 1.4144339279213689e-06, + "loss": 2.0619, + "step": 29087 + }, + { + "epoch": 5.29, + "learning_rate": 1.4143793228977066e-06, + "loss": 0.4526, + "step": 29088 + }, + { + "epoch": 5.29, + "learning_rate": 1.4143247178740443e-06, + "loss": 0.23, + "step": 29089 + }, + { + "epoch": 5.29, + "learning_rate": 1.4142701128503823e-06, + "loss": 0.2918, + "step": 29090 + }, + { + "epoch": 5.29, + "learning_rate": 1.41421550782672e-06, + "loss": 0.2678, + "step": 29091 + }, + { + "epoch": 5.29, + "learning_rate": 1.414160902803058e-06, + "loss": 0.2545, + "step": 29092 + }, + { + "epoch": 5.29, + "learning_rate": 1.414106297779396e-06, + "loss": 1.1304, + "step": 29093 + }, + { + "epoch": 5.29, + "learning_rate": 1.4140516927557336e-06, + "loss": 0.5277, + "step": 29094 + }, + { + "epoch": 5.29, + "learning_rate": 1.4139970877320714e-06, + "loss": 1.3557, + "step": 29095 + }, + { + "epoch": 5.29, + "learning_rate": 1.413942482708409e-06, + "loss": 1.2962, + "step": 29096 + }, + { + "epoch": 5.29, + "learning_rate": 1.413887877684747e-06, + "loss": 0.2867, + "step": 29097 + }, + { + "epoch": 5.29, + "learning_rate": 1.4138332726610848e-06, + "loss": 0.6589, + "step": 29098 + }, + { + "epoch": 5.29, + "learning_rate": 1.4137786676374227e-06, + "loss": 1.6032, + "step": 29099 + }, + { + "epoch": 5.29, + "learning_rate": 1.4137240626137607e-06, + "loss": 0.4645, + "step": 29100 + }, + { + "epoch": 5.29, + "learning_rate": 1.4136694575900984e-06, + "loss": 0.135, + "step": 29101 + }, + { + "epoch": 5.29, + "learning_rate": 1.4136148525664361e-06, + "loss": 0.0573, + "step": 29102 + }, + { + "epoch": 5.29, + "learning_rate": 1.4135602475427739e-06, + "loss": 0.8963, + "step": 29103 + }, + { + "epoch": 5.29, + "learning_rate": 1.4135056425191118e-06, + "loss": 1.1882, + "step": 29104 + }, + { + "epoch": 5.29, + "learning_rate": 1.4134510374954495e-06, + "loss": 0.3893, + "step": 29105 + }, + { + "epoch": 5.29, + "learning_rate": 1.4133964324717875e-06, + "loss": 0.4287, + "step": 29106 + }, + { + "epoch": 5.29, + "learning_rate": 1.4133418274481254e-06, + "loss": 1.0788, + "step": 29107 + }, + { + "epoch": 5.29, + "learning_rate": 1.4132872224244631e-06, + "loss": 0.7037, + "step": 29108 + }, + { + "epoch": 5.29, + "learning_rate": 1.4132326174008009e-06, + "loss": 1.0794, + "step": 29109 + }, + { + "epoch": 5.29, + "learning_rate": 1.4131780123771386e-06, + "loss": 0.9583, + "step": 29110 + }, + { + "epoch": 5.29, + "learning_rate": 1.4131234073534766e-06, + "loss": 0.0281, + "step": 29111 + }, + { + "epoch": 5.29, + "learning_rate": 1.4130688023298145e-06, + "loss": 1.6409, + "step": 29112 + }, + { + "epoch": 5.29, + "learning_rate": 1.4130141973061522e-06, + "loss": 0.0131, + "step": 29113 + }, + { + "epoch": 5.29, + "learning_rate": 1.41295959228249e-06, + "loss": 0.0344, + "step": 29114 + }, + { + "epoch": 5.29, + "learning_rate": 1.412904987258828e-06, + "loss": 0.3827, + "step": 29115 + }, + { + "epoch": 5.29, + "learning_rate": 1.4128503822351656e-06, + "loss": 0.0056, + "step": 29116 + }, + { + "epoch": 5.29, + "learning_rate": 1.4127957772115034e-06, + "loss": 0.4483, + "step": 29117 + }, + { + "epoch": 5.3, + "learning_rate": 1.4127411721878413e-06, + "loss": 0.0394, + "step": 29118 + }, + { + "epoch": 5.3, + "learning_rate": 1.4126865671641793e-06, + "loss": 0.9182, + "step": 29119 + }, + { + "epoch": 5.3, + "learning_rate": 1.412631962140517e-06, + "loss": 0.5822, + "step": 29120 + }, + { + "epoch": 5.3, + "learning_rate": 1.4125773571168547e-06, + "loss": 2.0384, + "step": 29121 + }, + { + "epoch": 5.3, + "learning_rate": 1.4125227520931927e-06, + "loss": 0.3074, + "step": 29122 + }, + { + "epoch": 5.3, + "learning_rate": 1.4124681470695304e-06, + "loss": 1.2681, + "step": 29123 + }, + { + "epoch": 5.3, + "learning_rate": 1.4124135420458681e-06, + "loss": 0.3137, + "step": 29124 + }, + { + "epoch": 5.3, + "learning_rate": 1.412358937022206e-06, + "loss": 1.1312, + "step": 29125 + }, + { + "epoch": 5.3, + "learning_rate": 1.412304331998544e-06, + "loss": 1.8673, + "step": 29126 + }, + { + "epoch": 5.3, + "learning_rate": 1.4122497269748817e-06, + "loss": 0.0581, + "step": 29127 + }, + { + "epoch": 5.3, + "learning_rate": 1.4121951219512195e-06, + "loss": 0.0935, + "step": 29128 + }, + { + "epoch": 5.3, + "learning_rate": 1.4121405169275574e-06, + "loss": 0.0127, + "step": 29129 + }, + { + "epoch": 5.3, + "learning_rate": 1.4120859119038951e-06, + "loss": 0.0828, + "step": 29130 + }, + { + "epoch": 5.3, + "learning_rate": 1.4120313068802329e-06, + "loss": 0.2692, + "step": 29131 + }, + { + "epoch": 5.3, + "learning_rate": 1.4119767018565708e-06, + "loss": 0.9965, + "step": 29132 + }, + { + "epoch": 5.3, + "learning_rate": 1.4119220968329088e-06, + "loss": 0.4562, + "step": 29133 + }, + { + "epoch": 5.3, + "learning_rate": 1.4118674918092465e-06, + "loss": 0.7194, + "step": 29134 + }, + { + "epoch": 5.3, + "learning_rate": 1.4118128867855842e-06, + "loss": 0.1583, + "step": 29135 + }, + { + "epoch": 5.3, + "learning_rate": 1.4117582817619222e-06, + "loss": 0.5403, + "step": 29136 + }, + { + "epoch": 5.3, + "learning_rate": 1.41170367673826e-06, + "loss": 0.1213, + "step": 29137 + }, + { + "epoch": 5.3, + "learning_rate": 1.4116490717145976e-06, + "loss": 0.7984, + "step": 29138 + }, + { + "epoch": 5.3, + "learning_rate": 1.4115944666909356e-06, + "loss": 0.1527, + "step": 29139 + }, + { + "epoch": 5.3, + "learning_rate": 1.4115398616672735e-06, + "loss": 0.0127, + "step": 29140 + }, + { + "epoch": 5.3, + "learning_rate": 1.4114852566436113e-06, + "loss": 0.7547, + "step": 29141 + }, + { + "epoch": 5.3, + "learning_rate": 1.411430651619949e-06, + "loss": 0.4161, + "step": 29142 + }, + { + "epoch": 5.3, + "learning_rate": 1.411376046596287e-06, + "loss": 0.6239, + "step": 29143 + }, + { + "epoch": 5.3, + "learning_rate": 1.4113214415726247e-06, + "loss": 2.5904, + "step": 29144 + }, + { + "epoch": 5.3, + "learning_rate": 1.4112668365489624e-06, + "loss": 0.074, + "step": 29145 + }, + { + "epoch": 5.3, + "learning_rate": 1.4112122315253003e-06, + "loss": 1.5184, + "step": 29146 + }, + { + "epoch": 5.3, + "learning_rate": 1.4111576265016383e-06, + "loss": 0.2191, + "step": 29147 + }, + { + "epoch": 5.3, + "learning_rate": 1.411103021477976e-06, + "loss": 0.7232, + "step": 29148 + }, + { + "epoch": 5.3, + "learning_rate": 1.4110484164543137e-06, + "loss": 1.3674, + "step": 29149 + }, + { + "epoch": 5.3, + "learning_rate": 1.4109938114306517e-06, + "loss": 0.0054, + "step": 29150 + }, + { + "epoch": 5.3, + "learning_rate": 1.4109392064069894e-06, + "loss": 1.282, + "step": 29151 + }, + { + "epoch": 5.3, + "learning_rate": 1.4108846013833274e-06, + "loss": 0.2064, + "step": 29152 + }, + { + "epoch": 5.3, + "learning_rate": 1.410829996359665e-06, + "loss": 0.9471, + "step": 29153 + }, + { + "epoch": 5.3, + "learning_rate": 1.410775391336003e-06, + "loss": 0.0202, + "step": 29154 + }, + { + "epoch": 5.3, + "learning_rate": 1.4107207863123408e-06, + "loss": 0.6712, + "step": 29155 + }, + { + "epoch": 5.3, + "learning_rate": 1.4106661812886785e-06, + "loss": 2.7733, + "step": 29156 + }, + { + "epoch": 5.3, + "learning_rate": 1.4106115762650164e-06, + "loss": 0.0489, + "step": 29157 + }, + { + "epoch": 5.3, + "learning_rate": 1.4105569712413542e-06, + "loss": 0.0098, + "step": 29158 + }, + { + "epoch": 5.3, + "learning_rate": 1.4105023662176921e-06, + "loss": 0.0854, + "step": 29159 + }, + { + "epoch": 5.3, + "learning_rate": 1.41044776119403e-06, + "loss": 0.0134, + "step": 29160 + }, + { + "epoch": 5.3, + "learning_rate": 1.4103931561703678e-06, + "loss": 1.8472, + "step": 29161 + }, + { + "epoch": 5.3, + "learning_rate": 1.4103385511467055e-06, + "loss": 0.9335, + "step": 29162 + }, + { + "epoch": 5.3, + "learning_rate": 1.4102839461230433e-06, + "loss": 2.0479, + "step": 29163 + }, + { + "epoch": 5.3, + "learning_rate": 1.4102293410993812e-06, + "loss": 0.5566, + "step": 29164 + }, + { + "epoch": 5.3, + "learning_rate": 1.410174736075719e-06, + "loss": 0.2013, + "step": 29165 + }, + { + "epoch": 5.3, + "learning_rate": 1.4101201310520569e-06, + "loss": 0.2654, + "step": 29166 + }, + { + "epoch": 5.3, + "learning_rate": 1.4100655260283948e-06, + "loss": 0.1004, + "step": 29167 + }, + { + "epoch": 5.3, + "learning_rate": 1.4100109210047325e-06, + "loss": 0.9255, + "step": 29168 + }, + { + "epoch": 5.3, + "learning_rate": 1.4099563159810703e-06, + "loss": 0.0432, + "step": 29169 + }, + { + "epoch": 5.3, + "learning_rate": 1.409901710957408e-06, + "loss": 0.2915, + "step": 29170 + }, + { + "epoch": 5.3, + "learning_rate": 1.409847105933746e-06, + "loss": 0.2716, + "step": 29171 + }, + { + "epoch": 5.3, + "learning_rate": 1.4097925009100837e-06, + "loss": 1.4038, + "step": 29172 + }, + { + "epoch": 5.31, + "learning_rate": 1.4097378958864216e-06, + "loss": 0.1973, + "step": 29173 + }, + { + "epoch": 5.31, + "learning_rate": 1.4096832908627596e-06, + "loss": 0.3833, + "step": 29174 + }, + { + "epoch": 5.31, + "learning_rate": 1.4096286858390973e-06, + "loss": 1.2534, + "step": 29175 + }, + { + "epoch": 5.31, + "learning_rate": 1.409574080815435e-06, + "loss": 0.8123, + "step": 29176 + }, + { + "epoch": 5.31, + "learning_rate": 1.4095194757917728e-06, + "loss": 1.8848, + "step": 29177 + }, + { + "epoch": 5.31, + "learning_rate": 1.4094648707681107e-06, + "loss": 0.2333, + "step": 29178 + }, + { + "epoch": 5.31, + "learning_rate": 1.4094102657444484e-06, + "loss": 0.3053, + "step": 29179 + }, + { + "epoch": 5.31, + "learning_rate": 1.4093556607207864e-06, + "loss": 0.4107, + "step": 29180 + }, + { + "epoch": 5.31, + "learning_rate": 1.4093010556971243e-06, + "loss": 0.5657, + "step": 29181 + }, + { + "epoch": 5.31, + "learning_rate": 1.409246450673462e-06, + "loss": 0.0231, + "step": 29182 + }, + { + "epoch": 5.31, + "learning_rate": 1.4091918456497998e-06, + "loss": 0.0108, + "step": 29183 + }, + { + "epoch": 5.31, + "learning_rate": 1.4091372406261375e-06, + "loss": 0.387, + "step": 29184 + }, + { + "epoch": 5.31, + "learning_rate": 1.4090826356024755e-06, + "loss": 1.7322, + "step": 29185 + }, + { + "epoch": 5.31, + "learning_rate": 1.4090280305788132e-06, + "loss": 0.2481, + "step": 29186 + }, + { + "epoch": 5.31, + "learning_rate": 1.4089734255551511e-06, + "loss": 0.0154, + "step": 29187 + }, + { + "epoch": 5.31, + "learning_rate": 1.408918820531489e-06, + "loss": 2.0646, + "step": 29188 + }, + { + "epoch": 5.31, + "learning_rate": 1.4088642155078268e-06, + "loss": 0.6229, + "step": 29189 + }, + { + "epoch": 5.31, + "learning_rate": 1.4088096104841646e-06, + "loss": 0.9579, + "step": 29190 + }, + { + "epoch": 5.31, + "learning_rate": 1.4087550054605023e-06, + "loss": 0.2275, + "step": 29191 + }, + { + "epoch": 5.31, + "learning_rate": 1.4087004004368402e-06, + "loss": 0.2486, + "step": 29192 + }, + { + "epoch": 5.31, + "learning_rate": 1.408645795413178e-06, + "loss": 1.5653, + "step": 29193 + }, + { + "epoch": 5.31, + "learning_rate": 1.408591190389516e-06, + "loss": 1.9568, + "step": 29194 + }, + { + "epoch": 5.31, + "learning_rate": 1.4085365853658536e-06, + "loss": 0.0054, + "step": 29195 + }, + { + "epoch": 5.31, + "learning_rate": 1.4084819803421916e-06, + "loss": 0.224, + "step": 29196 + }, + { + "epoch": 5.31, + "learning_rate": 1.4084273753185293e-06, + "loss": 0.985, + "step": 29197 + }, + { + "epoch": 5.31, + "learning_rate": 1.408372770294867e-06, + "loss": 0.0103, + "step": 29198 + }, + { + "epoch": 5.31, + "learning_rate": 1.408318165271205e-06, + "loss": 1.3163, + "step": 29199 + }, + { + "epoch": 5.31, + "learning_rate": 1.408263560247543e-06, + "loss": 0.0118, + "step": 29200 + }, + { + "epoch": 5.31, + "learning_rate": 1.4082089552238807e-06, + "loss": 0.2012, + "step": 29201 + }, + { + "epoch": 5.31, + "learning_rate": 1.4081543502002184e-06, + "loss": 0.7077, + "step": 29202 + }, + { + "epoch": 5.31, + "learning_rate": 1.4080997451765563e-06, + "loss": 1.3382, + "step": 29203 + }, + { + "epoch": 5.31, + "learning_rate": 1.408045140152894e-06, + "loss": 2.6019, + "step": 29204 + }, + { + "epoch": 5.31, + "learning_rate": 1.4079905351292318e-06, + "loss": 0.0263, + "step": 29205 + }, + { + "epoch": 5.31, + "learning_rate": 1.4079359301055697e-06, + "loss": 0.4835, + "step": 29206 + }, + { + "epoch": 5.31, + "learning_rate": 1.4078813250819077e-06, + "loss": 2.8845, + "step": 29207 + }, + { + "epoch": 5.31, + "learning_rate": 1.4078267200582454e-06, + "loss": 2.6758, + "step": 29208 + }, + { + "epoch": 5.31, + "learning_rate": 1.4077721150345831e-06, + "loss": 2.9994, + "step": 29209 + }, + { + "epoch": 5.31, + "learning_rate": 1.407717510010921e-06, + "loss": 0.4406, + "step": 29210 + }, + { + "epoch": 5.31, + "learning_rate": 1.4076629049872588e-06, + "loss": 1.2458, + "step": 29211 + }, + { + "epoch": 5.31, + "learning_rate": 1.4076082999635966e-06, + "loss": 1.8291, + "step": 29212 + }, + { + "epoch": 5.31, + "learning_rate": 1.4075536949399345e-06, + "loss": 0.7664, + "step": 29213 + }, + { + "epoch": 5.31, + "learning_rate": 1.4074990899162724e-06, + "loss": 0.5944, + "step": 29214 + }, + { + "epoch": 5.31, + "learning_rate": 1.4074444848926102e-06, + "loss": 1.4903, + "step": 29215 + }, + { + "epoch": 5.31, + "learning_rate": 1.407389879868948e-06, + "loss": 0.5842, + "step": 29216 + }, + { + "epoch": 5.31, + "learning_rate": 1.4073352748452858e-06, + "loss": 0.6817, + "step": 29217 + }, + { + "epoch": 5.31, + "learning_rate": 1.4072806698216236e-06, + "loss": 0.3391, + "step": 29218 + }, + { + "epoch": 5.31, + "learning_rate": 1.4072260647979613e-06, + "loss": 0.089, + "step": 29219 + }, + { + "epoch": 5.31, + "learning_rate": 1.4071714597742993e-06, + "loss": 0.3778, + "step": 29220 + }, + { + "epoch": 5.31, + "learning_rate": 1.4071168547506372e-06, + "loss": 0.1999, + "step": 29221 + }, + { + "epoch": 5.31, + "learning_rate": 1.407062249726975e-06, + "loss": 0.5405, + "step": 29222 + }, + { + "epoch": 5.31, + "learning_rate": 1.4070076447033127e-06, + "loss": 0.1219, + "step": 29223 + }, + { + "epoch": 5.31, + "learning_rate": 1.4069530396796506e-06, + "loss": 0.0562, + "step": 29224 + }, + { + "epoch": 5.31, + "learning_rate": 1.4068984346559883e-06, + "loss": 1.5741, + "step": 29225 + }, + { + "epoch": 5.31, + "learning_rate": 1.406843829632326e-06, + "loss": 1.2455, + "step": 29226 + }, + { + "epoch": 5.31, + "learning_rate": 1.406789224608664e-06, + "loss": 1.475, + "step": 29227 + }, + { + "epoch": 5.32, + "learning_rate": 1.406734619585002e-06, + "loss": 0.1475, + "step": 29228 + }, + { + "epoch": 5.32, + "learning_rate": 1.4066800145613397e-06, + "loss": 0.7527, + "step": 29229 + }, + { + "epoch": 5.32, + "learning_rate": 1.4066254095376774e-06, + "loss": 0.1629, + "step": 29230 + }, + { + "epoch": 5.32, + "learning_rate": 1.4065708045140154e-06, + "loss": 0.0749, + "step": 29231 + }, + { + "epoch": 5.32, + "learning_rate": 1.406516199490353e-06, + "loss": 0.2414, + "step": 29232 + }, + { + "epoch": 5.32, + "learning_rate": 1.4064615944666908e-06, + "loss": 1.2028, + "step": 29233 + }, + { + "epoch": 5.32, + "learning_rate": 1.4064069894430288e-06, + "loss": 0.3237, + "step": 29234 + }, + { + "epoch": 5.32, + "learning_rate": 1.4063523844193667e-06, + "loss": 0.0053, + "step": 29235 + }, + { + "epoch": 5.32, + "learning_rate": 1.4062977793957044e-06, + "loss": 0.1662, + "step": 29236 + }, + { + "epoch": 5.32, + "learning_rate": 1.4062431743720422e-06, + "loss": 0.1326, + "step": 29237 + }, + { + "epoch": 5.32, + "learning_rate": 1.4061885693483801e-06, + "loss": 0.0988, + "step": 29238 + }, + { + "epoch": 5.32, + "learning_rate": 1.4061339643247178e-06, + "loss": 2.266, + "step": 29239 + }, + { + "epoch": 5.32, + "learning_rate": 1.4060793593010558e-06, + "loss": 0.7849, + "step": 29240 + }, + { + "epoch": 5.32, + "learning_rate": 1.4060247542773937e-06, + "loss": 0.3575, + "step": 29241 + }, + { + "epoch": 5.32, + "learning_rate": 1.4059701492537315e-06, + "loss": 0.0347, + "step": 29242 + }, + { + "epoch": 5.32, + "learning_rate": 1.4059155442300692e-06, + "loss": 1.7179, + "step": 29243 + }, + { + "epoch": 5.32, + "learning_rate": 1.405860939206407e-06, + "loss": 1.5344, + "step": 29244 + }, + { + "epoch": 5.32, + "learning_rate": 1.4058063341827449e-06, + "loss": 0.0685, + "step": 29245 + }, + { + "epoch": 5.32, + "learning_rate": 1.4057517291590826e-06, + "loss": 0.0067, + "step": 29246 + }, + { + "epoch": 5.32, + "learning_rate": 1.4056971241354205e-06, + "loss": 0.0897, + "step": 29247 + }, + { + "epoch": 5.32, + "learning_rate": 1.4056425191117585e-06, + "loss": 0.7694, + "step": 29248 + }, + { + "epoch": 5.32, + "learning_rate": 1.4055879140880962e-06, + "loss": 0.0208, + "step": 29249 + }, + { + "epoch": 5.32, + "learning_rate": 1.405533309064434e-06, + "loss": 0.2256, + "step": 29250 + }, + { + "epoch": 5.32, + "learning_rate": 1.4054787040407717e-06, + "loss": 0.0101, + "step": 29251 + }, + { + "epoch": 5.32, + "learning_rate": 1.4054240990171096e-06, + "loss": 0.8063, + "step": 29252 + }, + { + "epoch": 5.32, + "learning_rate": 1.4053694939934474e-06, + "loss": 0.1956, + "step": 29253 + }, + { + "epoch": 5.32, + "learning_rate": 1.4053148889697853e-06, + "loss": 0.2411, + "step": 29254 + }, + { + "epoch": 5.32, + "learning_rate": 1.4052602839461232e-06, + "loss": 0.3054, + "step": 29255 + }, + { + "epoch": 5.32, + "learning_rate": 1.405205678922461e-06, + "loss": 0.1554, + "step": 29256 + }, + { + "epoch": 5.32, + "learning_rate": 1.4051510738987987e-06, + "loss": 0.9042, + "step": 29257 + }, + { + "epoch": 5.32, + "learning_rate": 1.4050964688751364e-06, + "loss": 0.1289, + "step": 29258 + }, + { + "epoch": 5.32, + "learning_rate": 1.4050418638514744e-06, + "loss": 0.0115, + "step": 29259 + }, + { + "epoch": 5.32, + "learning_rate": 1.4049872588278121e-06, + "loss": 0.3471, + "step": 29260 + }, + { + "epoch": 5.32, + "learning_rate": 1.40493265380415e-06, + "loss": 0.249, + "step": 29261 + }, + { + "epoch": 5.32, + "learning_rate": 1.404878048780488e-06, + "loss": 0.9047, + "step": 29262 + }, + { + "epoch": 5.32, + "learning_rate": 1.4048234437568257e-06, + "loss": 0.0609, + "step": 29263 + }, + { + "epoch": 5.32, + "learning_rate": 1.4047688387331635e-06, + "loss": 1.1058, + "step": 29264 + }, + { + "epoch": 5.32, + "learning_rate": 1.4047142337095012e-06, + "loss": 0.9011, + "step": 29265 + }, + { + "epoch": 5.32, + "learning_rate": 1.4046596286858391e-06, + "loss": 0.0083, + "step": 29266 + }, + { + "epoch": 5.32, + "learning_rate": 1.4046050236621769e-06, + "loss": 1.237, + "step": 29267 + }, + { + "epoch": 5.32, + "learning_rate": 1.4045504186385148e-06, + "loss": 0.1929, + "step": 29268 + }, + { + "epoch": 5.32, + "learning_rate": 1.4044958136148525e-06, + "loss": 0.0118, + "step": 29269 + }, + { + "epoch": 5.32, + "learning_rate": 1.4044412085911905e-06, + "loss": 0.4767, + "step": 29270 + }, + { + "epoch": 5.32, + "learning_rate": 1.4043866035675282e-06, + "loss": 0.2678, + "step": 29271 + }, + { + "epoch": 5.32, + "learning_rate": 1.404331998543866e-06, + "loss": 0.4685, + "step": 29272 + }, + { + "epoch": 5.32, + "learning_rate": 1.404277393520204e-06, + "loss": 0.9957, + "step": 29273 + }, + { + "epoch": 5.32, + "learning_rate": 1.4042227884965416e-06, + "loss": 1.4843, + "step": 29274 + }, + { + "epoch": 5.32, + "learning_rate": 1.4041681834728796e-06, + "loss": 0.2552, + "step": 29275 + }, + { + "epoch": 5.32, + "learning_rate": 1.4041135784492173e-06, + "loss": 1.677, + "step": 29276 + }, + { + "epoch": 5.32, + "learning_rate": 1.4040589734255552e-06, + "loss": 0.2642, + "step": 29277 + }, + { + "epoch": 5.32, + "learning_rate": 1.404004368401893e-06, + "loss": 0.3064, + "step": 29278 + }, + { + "epoch": 5.32, + "learning_rate": 1.4039497633782307e-06, + "loss": 0.0203, + "step": 29279 + }, + { + "epoch": 5.32, + "learning_rate": 1.4038951583545687e-06, + "loss": 1.3276, + "step": 29280 + }, + { + "epoch": 5.32, + "learning_rate": 1.4038405533309066e-06, + "loss": 0.399, + "step": 29281 + }, + { + "epoch": 5.32, + "learning_rate": 1.4037859483072443e-06, + "loss": 0.4955, + "step": 29282 + }, + { + "epoch": 5.33, + "learning_rate": 1.403731343283582e-06, + "loss": 0.2813, + "step": 29283 + }, + { + "epoch": 5.33, + "learning_rate": 1.40367673825992e-06, + "loss": 0.303, + "step": 29284 + }, + { + "epoch": 5.33, + "learning_rate": 1.4036221332362577e-06, + "loss": 1.5878, + "step": 29285 + }, + { + "epoch": 5.33, + "learning_rate": 1.4035675282125955e-06, + "loss": 0.7341, + "step": 29286 + }, + { + "epoch": 5.33, + "learning_rate": 1.4035129231889334e-06, + "loss": 3.2852, + "step": 29287 + }, + { + "epoch": 5.33, + "learning_rate": 1.4034583181652714e-06, + "loss": 0.0451, + "step": 29288 + }, + { + "epoch": 5.33, + "learning_rate": 1.403403713141609e-06, + "loss": 0.7275, + "step": 29289 + }, + { + "epoch": 5.33, + "learning_rate": 1.4033491081179468e-06, + "loss": 1.3484, + "step": 29290 + }, + { + "epoch": 5.33, + "learning_rate": 1.4032945030942848e-06, + "loss": 0.6948, + "step": 29291 + }, + { + "epoch": 5.33, + "learning_rate": 1.4032398980706225e-06, + "loss": 0.6947, + "step": 29292 + }, + { + "epoch": 5.33, + "learning_rate": 1.4031852930469602e-06, + "loss": 1.3464, + "step": 29293 + }, + { + "epoch": 5.33, + "learning_rate": 1.4031306880232982e-06, + "loss": 0.5237, + "step": 29294 + }, + { + "epoch": 5.33, + "learning_rate": 1.4030760829996361e-06, + "loss": 0.5391, + "step": 29295 + }, + { + "epoch": 5.33, + "learning_rate": 1.4030214779759738e-06, + "loss": 0.0101, + "step": 29296 + }, + { + "epoch": 5.33, + "learning_rate": 1.4029668729523116e-06, + "loss": 1.1694, + "step": 29297 + }, + { + "epoch": 5.33, + "learning_rate": 1.4029122679286495e-06, + "loss": 0.3096, + "step": 29298 + }, + { + "epoch": 5.33, + "learning_rate": 1.4028576629049873e-06, + "loss": 1.4194, + "step": 29299 + }, + { + "epoch": 5.33, + "learning_rate": 1.402803057881325e-06, + "loss": 1.9842, + "step": 29300 + }, + { + "epoch": 5.33, + "learning_rate": 1.402748452857663e-06, + "loss": 0.6616, + "step": 29301 + }, + { + "epoch": 5.33, + "learning_rate": 1.4026938478340009e-06, + "loss": 0.3968, + "step": 29302 + }, + { + "epoch": 5.33, + "learning_rate": 1.4026392428103386e-06, + "loss": 0.0923, + "step": 29303 + }, + { + "epoch": 5.33, + "learning_rate": 1.4025846377866763e-06, + "loss": 0.1449, + "step": 29304 + }, + { + "epoch": 5.33, + "learning_rate": 1.4025300327630143e-06, + "loss": 0.0173, + "step": 29305 + }, + { + "epoch": 5.33, + "learning_rate": 1.402475427739352e-06, + "loss": 0.4765, + "step": 29306 + }, + { + "epoch": 5.33, + "learning_rate": 1.4024208227156897e-06, + "loss": 0.1521, + "step": 29307 + }, + { + "epoch": 5.33, + "learning_rate": 1.4023662176920277e-06, + "loss": 0.0278, + "step": 29308 + }, + { + "epoch": 5.33, + "learning_rate": 1.4023116126683656e-06, + "loss": 0.3676, + "step": 29309 + }, + { + "epoch": 5.33, + "learning_rate": 1.4022570076447034e-06, + "loss": 1.1051, + "step": 29310 + }, + { + "epoch": 5.33, + "learning_rate": 1.402202402621041e-06, + "loss": 0.1238, + "step": 29311 + }, + { + "epoch": 5.33, + "learning_rate": 1.402147797597379e-06, + "loss": 0.1349, + "step": 29312 + }, + { + "epoch": 5.33, + "learning_rate": 1.4020931925737168e-06, + "loss": 1.3723, + "step": 29313 + }, + { + "epoch": 5.33, + "learning_rate": 1.4020385875500545e-06, + "loss": 0.9304, + "step": 29314 + }, + { + "epoch": 5.33, + "learning_rate": 1.4019839825263924e-06, + "loss": 1.2749, + "step": 29315 + }, + { + "epoch": 5.33, + "learning_rate": 1.4019293775027304e-06, + "loss": 0.8871, + "step": 29316 + }, + { + "epoch": 5.33, + "learning_rate": 1.4018747724790681e-06, + "loss": 0.029, + "step": 29317 + }, + { + "epoch": 5.33, + "learning_rate": 1.4018201674554058e-06, + "loss": 1.3366, + "step": 29318 + }, + { + "epoch": 5.33, + "learning_rate": 1.4017655624317438e-06, + "loss": 0.0047, + "step": 29319 + }, + { + "epoch": 5.33, + "learning_rate": 1.4017109574080815e-06, + "loss": 0.0298, + "step": 29320 + }, + { + "epoch": 5.33, + "learning_rate": 1.4016563523844195e-06, + "loss": 0.1213, + "step": 29321 + }, + { + "epoch": 5.33, + "learning_rate": 1.4016017473607572e-06, + "loss": 0.0563, + "step": 29322 + }, + { + "epoch": 5.33, + "learning_rate": 1.4015471423370951e-06, + "loss": 1.201, + "step": 29323 + }, + { + "epoch": 5.33, + "learning_rate": 1.4014925373134329e-06, + "loss": 1.1562, + "step": 29324 + }, + { + "epoch": 5.33, + "learning_rate": 1.4014379322897706e-06, + "loss": 1.4006, + "step": 29325 + }, + { + "epoch": 5.33, + "learning_rate": 1.4013833272661085e-06, + "loss": 1.6949, + "step": 29326 + }, + { + "epoch": 5.33, + "learning_rate": 1.4013287222424463e-06, + "loss": 2.2615, + "step": 29327 + }, + { + "epoch": 5.33, + "learning_rate": 1.4012741172187842e-06, + "loss": 0.6421, + "step": 29328 + }, + { + "epoch": 5.33, + "learning_rate": 1.4012195121951222e-06, + "loss": 0.7747, + "step": 29329 + }, + { + "epoch": 5.33, + "learning_rate": 1.40116490717146e-06, + "loss": 0.2358, + "step": 29330 + }, + { + "epoch": 5.33, + "learning_rate": 1.4011103021477976e-06, + "loss": 2.2255, + "step": 29331 + }, + { + "epoch": 5.33, + "learning_rate": 1.4010556971241354e-06, + "loss": 0.0134, + "step": 29332 + }, + { + "epoch": 5.33, + "learning_rate": 1.4010010921004733e-06, + "loss": 0.5196, + "step": 29333 + }, + { + "epoch": 5.33, + "learning_rate": 1.400946487076811e-06, + "loss": 0.3395, + "step": 29334 + }, + { + "epoch": 5.33, + "learning_rate": 1.400891882053149e-06, + "loss": 0.0095, + "step": 29335 + }, + { + "epoch": 5.33, + "learning_rate": 1.400837277029487e-06, + "loss": 2.9344, + "step": 29336 + }, + { + "epoch": 5.33, + "learning_rate": 1.4007826720058247e-06, + "loss": 0.4355, + "step": 29337 + }, + { + "epoch": 5.34, + "learning_rate": 1.4007280669821624e-06, + "loss": 4.0447, + "step": 29338 + }, + { + "epoch": 5.34, + "learning_rate": 1.4006734619585001e-06, + "loss": 0.0802, + "step": 29339 + }, + { + "epoch": 5.34, + "learning_rate": 1.400618856934838e-06, + "loss": 1.3268, + "step": 29340 + }, + { + "epoch": 5.34, + "learning_rate": 1.4005642519111758e-06, + "loss": 0.0688, + "step": 29341 + }, + { + "epoch": 5.34, + "learning_rate": 1.4005096468875137e-06, + "loss": 1.025, + "step": 29342 + }, + { + "epoch": 5.34, + "learning_rate": 1.4004550418638517e-06, + "loss": 0.3403, + "step": 29343 + }, + { + "epoch": 5.34, + "learning_rate": 1.4004004368401894e-06, + "loss": 0.3834, + "step": 29344 + }, + { + "epoch": 5.34, + "learning_rate": 1.4003458318165271e-06, + "loss": 0.3441, + "step": 29345 + }, + { + "epoch": 5.34, + "learning_rate": 1.4002912267928649e-06, + "loss": 1.0938, + "step": 29346 + }, + { + "epoch": 5.34, + "learning_rate": 1.4002366217692028e-06, + "loss": 0.7091, + "step": 29347 + }, + { + "epoch": 5.34, + "learning_rate": 1.4001820167455405e-06, + "loss": 0.1993, + "step": 29348 + }, + { + "epoch": 5.34, + "learning_rate": 1.4001274117218785e-06, + "loss": 1.5476, + "step": 29349 + }, + { + "epoch": 5.34, + "learning_rate": 1.4000728066982162e-06, + "loss": 0.1146, + "step": 29350 + }, + { + "epoch": 5.34, + "learning_rate": 1.4000182016745542e-06, + "loss": 0.2131, + "step": 29351 + }, + { + "epoch": 5.34, + "learning_rate": 1.399963596650892e-06, + "loss": 0.2582, + "step": 29352 + }, + { + "epoch": 5.34, + "learning_rate": 1.3999089916272296e-06, + "loss": 2.6136, + "step": 29353 + }, + { + "epoch": 5.34, + "learning_rate": 1.3998543866035676e-06, + "loss": 0.3101, + "step": 29354 + }, + { + "epoch": 5.34, + "learning_rate": 1.3997997815799053e-06, + "loss": 0.9597, + "step": 29355 + }, + { + "epoch": 5.34, + "learning_rate": 1.3997451765562432e-06, + "loss": 0.7688, + "step": 29356 + }, + { + "epoch": 5.34, + "learning_rate": 1.399690571532581e-06, + "loss": 0.5609, + "step": 29357 + }, + { + "epoch": 5.34, + "learning_rate": 1.399635966508919e-06, + "loss": 1.2456, + "step": 29358 + }, + { + "epoch": 5.34, + "learning_rate": 1.3995813614852567e-06, + "loss": 1.8469, + "step": 29359 + }, + { + "epoch": 5.34, + "learning_rate": 1.3995267564615944e-06, + "loss": 2.3674, + "step": 29360 + }, + { + "epoch": 5.34, + "learning_rate": 1.3994721514379323e-06, + "loss": 0.8796, + "step": 29361 + }, + { + "epoch": 5.34, + "learning_rate": 1.39941754641427e-06, + "loss": 0.162, + "step": 29362 + }, + { + "epoch": 5.34, + "learning_rate": 1.399362941390608e-06, + "loss": 0.9064, + "step": 29363 + }, + { + "epoch": 5.34, + "learning_rate": 1.3993083363669457e-06, + "loss": 0.0095, + "step": 29364 + }, + { + "epoch": 5.34, + "learning_rate": 1.3992537313432837e-06, + "loss": 0.0319, + "step": 29365 + }, + { + "epoch": 5.34, + "learning_rate": 1.3991991263196214e-06, + "loss": 0.46, + "step": 29366 + }, + { + "epoch": 5.34, + "learning_rate": 1.3991445212959591e-06, + "loss": 1.6528, + "step": 29367 + }, + { + "epoch": 5.34, + "learning_rate": 1.399089916272297e-06, + "loss": 0.0083, + "step": 29368 + }, + { + "epoch": 5.34, + "learning_rate": 1.399035311248635e-06, + "loss": 0.0866, + "step": 29369 + }, + { + "epoch": 5.34, + "learning_rate": 1.3989807062249728e-06, + "loss": 0.1452, + "step": 29370 + }, + { + "epoch": 5.34, + "learning_rate": 1.3989261012013105e-06, + "loss": 0.2175, + "step": 29371 + }, + { + "epoch": 5.34, + "learning_rate": 1.3988714961776484e-06, + "loss": 0.2456, + "step": 29372 + }, + { + "epoch": 5.34, + "learning_rate": 1.3988168911539862e-06, + "loss": 1.9655, + "step": 29373 + }, + { + "epoch": 5.34, + "learning_rate": 1.398762286130324e-06, + "loss": 0.2088, + "step": 29374 + }, + { + "epoch": 5.34, + "learning_rate": 1.3987076811066618e-06, + "loss": 1.2475, + "step": 29375 + }, + { + "epoch": 5.34, + "learning_rate": 1.3986530760829998e-06, + "loss": 0.0139, + "step": 29376 + }, + { + "epoch": 5.34, + "learning_rate": 1.3985984710593375e-06, + "loss": 0.0657, + "step": 29377 + }, + { + "epoch": 5.34, + "learning_rate": 1.3985438660356753e-06, + "loss": 0.109, + "step": 29378 + }, + { + "epoch": 5.34, + "learning_rate": 1.3984892610120132e-06, + "loss": 0.2559, + "step": 29379 + }, + { + "epoch": 5.34, + "learning_rate": 1.398434655988351e-06, + "loss": 0.1194, + "step": 29380 + }, + { + "epoch": 5.34, + "learning_rate": 1.3983800509646887e-06, + "loss": 0.0534, + "step": 29381 + }, + { + "epoch": 5.34, + "learning_rate": 1.3983254459410266e-06, + "loss": 1.6905, + "step": 29382 + }, + { + "epoch": 5.34, + "learning_rate": 1.3982708409173645e-06, + "loss": 0.0085, + "step": 29383 + }, + { + "epoch": 5.34, + "learning_rate": 1.3982162358937023e-06, + "loss": 0.0575, + "step": 29384 + }, + { + "epoch": 5.34, + "learning_rate": 1.39816163087004e-06, + "loss": 4.4098, + "step": 29385 + }, + { + "epoch": 5.34, + "learning_rate": 1.398107025846378e-06, + "loss": 0.3557, + "step": 29386 + }, + { + "epoch": 5.34, + "learning_rate": 1.3980524208227157e-06, + "loss": 0.0163, + "step": 29387 + }, + { + "epoch": 5.34, + "learning_rate": 1.3979978157990534e-06, + "loss": 0.798, + "step": 29388 + }, + { + "epoch": 5.34, + "learning_rate": 1.3979432107753914e-06, + "loss": 0.1517, + "step": 29389 + }, + { + "epoch": 5.34, + "learning_rate": 1.3978886057517293e-06, + "loss": 0.0426, + "step": 29390 + }, + { + "epoch": 5.34, + "learning_rate": 1.397834000728067e-06, + "loss": 2.2597, + "step": 29391 + }, + { + "epoch": 5.34, + "learning_rate": 1.3977793957044048e-06, + "loss": 0.4117, + "step": 29392 + }, + { + "epoch": 5.35, + "learning_rate": 1.3977247906807427e-06, + "loss": 2.3254, + "step": 29393 + }, + { + "epoch": 5.35, + "learning_rate": 1.3976701856570804e-06, + "loss": 0.3676, + "step": 29394 + }, + { + "epoch": 5.35, + "learning_rate": 1.3976155806334182e-06, + "loss": 1.0012, + "step": 29395 + }, + { + "epoch": 5.35, + "learning_rate": 1.3975609756097561e-06, + "loss": 0.1675, + "step": 29396 + }, + { + "epoch": 5.35, + "learning_rate": 1.397506370586094e-06, + "loss": 3.0326, + "step": 29397 + }, + { + "epoch": 5.35, + "learning_rate": 1.3974517655624318e-06, + "loss": 0.313, + "step": 29398 + }, + { + "epoch": 5.35, + "learning_rate": 1.3973971605387695e-06, + "loss": 1.5284, + "step": 29399 + }, + { + "epoch": 5.35, + "learning_rate": 1.3973425555151075e-06, + "loss": 0.5862, + "step": 29400 + }, + { + "epoch": 5.35, + "learning_rate": 1.3972879504914452e-06, + "loss": 0.1495, + "step": 29401 + }, + { + "epoch": 5.35, + "learning_rate": 1.397233345467783e-06, + "loss": 0.9429, + "step": 29402 + }, + { + "epoch": 5.35, + "learning_rate": 1.3971787404441209e-06, + "loss": 1.1029, + "step": 29403 + }, + { + "epoch": 5.35, + "learning_rate": 1.3971241354204588e-06, + "loss": 0.698, + "step": 29404 + }, + { + "epoch": 5.35, + "learning_rate": 1.3970695303967965e-06, + "loss": 0.7566, + "step": 29405 + }, + { + "epoch": 5.35, + "learning_rate": 1.3970149253731343e-06, + "loss": 0.2935, + "step": 29406 + }, + { + "epoch": 5.35, + "learning_rate": 1.3969603203494722e-06, + "loss": 0.2494, + "step": 29407 + }, + { + "epoch": 5.35, + "learning_rate": 1.39690571532581e-06, + "loss": 0.2071, + "step": 29408 + }, + { + "epoch": 5.35, + "learning_rate": 1.396851110302148e-06, + "loss": 0.2105, + "step": 29409 + }, + { + "epoch": 5.35, + "learning_rate": 1.3967965052784858e-06, + "loss": 0.1485, + "step": 29410 + }, + { + "epoch": 5.35, + "learning_rate": 1.3967419002548236e-06, + "loss": 0.5241, + "step": 29411 + }, + { + "epoch": 5.35, + "learning_rate": 1.3966872952311613e-06, + "loss": 0.1292, + "step": 29412 + }, + { + "epoch": 5.35, + "learning_rate": 1.396632690207499e-06, + "loss": 0.5514, + "step": 29413 + }, + { + "epoch": 5.35, + "learning_rate": 1.396578085183837e-06, + "loss": 0.6382, + "step": 29414 + }, + { + "epoch": 5.35, + "learning_rate": 1.3965234801601747e-06, + "loss": 0.1163, + "step": 29415 + }, + { + "epoch": 5.35, + "learning_rate": 1.3964688751365127e-06, + "loss": 3.0249, + "step": 29416 + }, + { + "epoch": 5.35, + "learning_rate": 1.3964142701128506e-06, + "loss": 0.111, + "step": 29417 + }, + { + "epoch": 5.35, + "learning_rate": 1.3963596650891883e-06, + "loss": 0.3582, + "step": 29418 + }, + { + "epoch": 5.35, + "learning_rate": 1.396305060065526e-06, + "loss": 0.0451, + "step": 29419 + }, + { + "epoch": 5.35, + "learning_rate": 1.3962504550418638e-06, + "loss": 0.8285, + "step": 29420 + }, + { + "epoch": 5.35, + "learning_rate": 1.3961958500182017e-06, + "loss": 1.415, + "step": 29421 + }, + { + "epoch": 5.35, + "learning_rate": 1.3961412449945395e-06, + "loss": 0.1099, + "step": 29422 + }, + { + "epoch": 5.35, + "learning_rate": 1.3960866399708774e-06, + "loss": 0.2607, + "step": 29423 + }, + { + "epoch": 5.35, + "learning_rate": 1.3960320349472151e-06, + "loss": 0.6313, + "step": 29424 + }, + { + "epoch": 5.35, + "learning_rate": 1.395977429923553e-06, + "loss": 1.9383, + "step": 29425 + }, + { + "epoch": 5.35, + "learning_rate": 1.3959228248998908e-06, + "loss": 0.3647, + "step": 29426 + }, + { + "epoch": 5.35, + "learning_rate": 1.3958682198762285e-06, + "loss": 2.4336, + "step": 29427 + }, + { + "epoch": 5.35, + "learning_rate": 1.3958136148525665e-06, + "loss": 0.0124, + "step": 29428 + }, + { + "epoch": 5.35, + "learning_rate": 1.3957590098289042e-06, + "loss": 0.2763, + "step": 29429 + }, + { + "epoch": 5.35, + "learning_rate": 1.3957044048052422e-06, + "loss": 0.0565, + "step": 29430 + }, + { + "epoch": 5.35, + "learning_rate": 1.39564979978158e-06, + "loss": 2.2495, + "step": 29431 + }, + { + "epoch": 5.35, + "learning_rate": 1.3955951947579178e-06, + "loss": 0.1534, + "step": 29432 + }, + { + "epoch": 5.35, + "learning_rate": 1.3955405897342556e-06, + "loss": 0.0083, + "step": 29433 + }, + { + "epoch": 5.35, + "learning_rate": 1.3954859847105933e-06, + "loss": 0.1305, + "step": 29434 + }, + { + "epoch": 5.35, + "learning_rate": 1.3954313796869312e-06, + "loss": 0.9188, + "step": 29435 + }, + { + "epoch": 5.35, + "learning_rate": 1.395376774663269e-06, + "loss": 1.2737, + "step": 29436 + }, + { + "epoch": 5.35, + "learning_rate": 1.395322169639607e-06, + "loss": 0.806, + "step": 29437 + }, + { + "epoch": 5.35, + "learning_rate": 1.3952675646159447e-06, + "loss": 0.1613, + "step": 29438 + }, + { + "epoch": 5.35, + "learning_rate": 1.3952129595922826e-06, + "loss": 0.2196, + "step": 29439 + }, + { + "epoch": 5.35, + "learning_rate": 1.3951583545686203e-06, + "loss": 0.0219, + "step": 29440 + }, + { + "epoch": 5.35, + "learning_rate": 1.395103749544958e-06, + "loss": 1.0102, + "step": 29441 + }, + { + "epoch": 5.35, + "learning_rate": 1.395049144521296e-06, + "loss": 0.5245, + "step": 29442 + }, + { + "epoch": 5.35, + "learning_rate": 1.3949945394976337e-06, + "loss": 0.2335, + "step": 29443 + }, + { + "epoch": 5.35, + "learning_rate": 1.3949399344739717e-06, + "loss": 0.1933, + "step": 29444 + }, + { + "epoch": 5.35, + "learning_rate": 1.3948853294503094e-06, + "loss": 0.5467, + "step": 29445 + }, + { + "epoch": 5.35, + "learning_rate": 1.3948307244266474e-06, + "loss": 0.3194, + "step": 29446 + }, + { + "epoch": 5.35, + "learning_rate": 1.394776119402985e-06, + "loss": 0.1664, + "step": 29447 + }, + { + "epoch": 5.36, + "learning_rate": 1.3947215143793228e-06, + "loss": 1.1492, + "step": 29448 + }, + { + "epoch": 5.36, + "learning_rate": 1.3946669093556608e-06, + "loss": 1.6531, + "step": 29449 + }, + { + "epoch": 5.36, + "learning_rate": 1.3946123043319987e-06, + "loss": 0.1123, + "step": 29450 + }, + { + "epoch": 5.36, + "learning_rate": 1.3945576993083364e-06, + "loss": 0.0928, + "step": 29451 + }, + { + "epoch": 5.36, + "learning_rate": 1.3945030942846742e-06, + "loss": 0.9276, + "step": 29452 + }, + { + "epoch": 5.36, + "learning_rate": 1.3944484892610121e-06, + "loss": 0.837, + "step": 29453 + }, + { + "epoch": 5.36, + "learning_rate": 1.3943938842373498e-06, + "loss": 0.019, + "step": 29454 + }, + { + "epoch": 5.36, + "learning_rate": 1.3943392792136876e-06, + "loss": 0.0858, + "step": 29455 + }, + { + "epoch": 5.36, + "learning_rate": 1.3942846741900255e-06, + "loss": 0.0136, + "step": 29456 + }, + { + "epoch": 5.36, + "learning_rate": 1.3942300691663635e-06, + "loss": 1.0616, + "step": 29457 + }, + { + "epoch": 5.36, + "learning_rate": 1.3941754641427012e-06, + "loss": 0.2267, + "step": 29458 + }, + { + "epoch": 5.36, + "learning_rate": 1.394120859119039e-06, + "loss": 0.0596, + "step": 29459 + }, + { + "epoch": 5.36, + "learning_rate": 1.3940662540953769e-06, + "loss": 0.9804, + "step": 29460 + }, + { + "epoch": 5.36, + "learning_rate": 1.3940116490717146e-06, + "loss": 1.6152, + "step": 29461 + }, + { + "epoch": 5.36, + "learning_rate": 1.3939570440480523e-06, + "loss": 2.7368, + "step": 29462 + }, + { + "epoch": 5.36, + "learning_rate": 1.3939024390243903e-06, + "loss": 0.0094, + "step": 29463 + }, + { + "epoch": 5.36, + "learning_rate": 1.3938478340007282e-06, + "loss": 0.7, + "step": 29464 + }, + { + "epoch": 5.36, + "learning_rate": 1.393793228977066e-06, + "loss": 2.4761, + "step": 29465 + }, + { + "epoch": 5.36, + "learning_rate": 1.3937386239534037e-06, + "loss": 0.5816, + "step": 29466 + }, + { + "epoch": 5.36, + "learning_rate": 1.3936840189297416e-06, + "loss": 0.0075, + "step": 29467 + }, + { + "epoch": 5.36, + "learning_rate": 1.3936294139060794e-06, + "loss": 0.1662, + "step": 29468 + }, + { + "epoch": 5.36, + "learning_rate": 1.393574808882417e-06, + "loss": 0.1846, + "step": 29469 + }, + { + "epoch": 5.36, + "learning_rate": 1.393520203858755e-06, + "loss": 0.4466, + "step": 29470 + }, + { + "epoch": 5.36, + "learning_rate": 1.393465598835093e-06, + "loss": 0.1426, + "step": 29471 + }, + { + "epoch": 5.36, + "learning_rate": 1.3934109938114307e-06, + "loss": 1.0823, + "step": 29472 + }, + { + "epoch": 5.36, + "learning_rate": 1.3933563887877684e-06, + "loss": 0.1989, + "step": 29473 + }, + { + "epoch": 5.36, + "learning_rate": 1.3933017837641064e-06, + "loss": 0.02, + "step": 29474 + }, + { + "epoch": 5.36, + "learning_rate": 1.3932471787404441e-06, + "loss": 0.2059, + "step": 29475 + }, + { + "epoch": 5.36, + "learning_rate": 1.3931925737167818e-06, + "loss": 1.0609, + "step": 29476 + }, + { + "epoch": 5.36, + "learning_rate": 1.3931379686931198e-06, + "loss": 0.2693, + "step": 29477 + }, + { + "epoch": 5.36, + "learning_rate": 1.3930833636694577e-06, + "loss": 0.0229, + "step": 29478 + }, + { + "epoch": 5.36, + "learning_rate": 1.3930287586457955e-06, + "loss": 0.451, + "step": 29479 + }, + { + "epoch": 5.36, + "learning_rate": 1.3929741536221332e-06, + "loss": 0.0628, + "step": 29480 + }, + { + "epoch": 5.36, + "learning_rate": 1.3929195485984711e-06, + "loss": 2.8014, + "step": 29481 + }, + { + "epoch": 5.36, + "learning_rate": 1.3928649435748089e-06, + "loss": 0.109, + "step": 29482 + }, + { + "epoch": 5.36, + "learning_rate": 1.3928103385511466e-06, + "loss": 0.3609, + "step": 29483 + }, + { + "epoch": 5.36, + "learning_rate": 1.3927557335274845e-06, + "loss": 5.8115, + "step": 29484 + }, + { + "epoch": 5.36, + "learning_rate": 1.3927011285038225e-06, + "loss": 2.0657, + "step": 29485 + }, + { + "epoch": 5.36, + "learning_rate": 1.3926465234801602e-06, + "loss": 2.3876, + "step": 29486 + }, + { + "epoch": 5.36, + "learning_rate": 1.392591918456498e-06, + "loss": 2.8102, + "step": 29487 + }, + { + "epoch": 5.36, + "learning_rate": 1.392537313432836e-06, + "loss": 0.2179, + "step": 29488 + }, + { + "epoch": 5.36, + "learning_rate": 1.3924827084091736e-06, + "loss": 0.032, + "step": 29489 + }, + { + "epoch": 5.36, + "learning_rate": 1.3924281033855114e-06, + "loss": 0.9344, + "step": 29490 + }, + { + "epoch": 5.36, + "learning_rate": 1.3923734983618493e-06, + "loss": 0.1182, + "step": 29491 + }, + { + "epoch": 5.36, + "learning_rate": 1.3923188933381872e-06, + "loss": 0.5887, + "step": 29492 + }, + { + "epoch": 5.36, + "learning_rate": 1.392264288314525e-06, + "loss": 0.0109, + "step": 29493 + }, + { + "epoch": 5.36, + "learning_rate": 1.3922096832908627e-06, + "loss": 1.5868, + "step": 29494 + }, + { + "epoch": 5.36, + "learning_rate": 1.3921550782672007e-06, + "loss": 0.0106, + "step": 29495 + }, + { + "epoch": 5.36, + "learning_rate": 1.3921004732435384e-06, + "loss": 0.1824, + "step": 29496 + }, + { + "epoch": 5.36, + "learning_rate": 1.3920458682198763e-06, + "loss": 4.8788, + "step": 29497 + }, + { + "epoch": 5.36, + "learning_rate": 1.3919912631962143e-06, + "loss": 0.0822, + "step": 29498 + }, + { + "epoch": 5.36, + "learning_rate": 1.391936658172552e-06, + "loss": 0.5813, + "step": 29499 + }, + { + "epoch": 5.36, + "learning_rate": 1.3918820531488897e-06, + "loss": 1.6197, + "step": 29500 + }, + { + "epoch": 5.36, + "learning_rate": 1.3918274481252275e-06, + "loss": 0.597, + "step": 29501 + }, + { + "epoch": 5.36, + "learning_rate": 1.3917728431015654e-06, + "loss": 0.3634, + "step": 29502 + }, + { + "epoch": 5.37, + "learning_rate": 1.3917182380779031e-06, + "loss": 0.0404, + "step": 29503 + }, + { + "epoch": 5.37, + "learning_rate": 1.391663633054241e-06, + "loss": 1.1728, + "step": 29504 + }, + { + "epoch": 5.37, + "learning_rate": 1.3916090280305788e-06, + "loss": 0.3331, + "step": 29505 + }, + { + "epoch": 5.37, + "learning_rate": 1.3915544230069168e-06, + "loss": 0.0295, + "step": 29506 + }, + { + "epoch": 5.37, + "learning_rate": 1.3914998179832545e-06, + "loss": 0.3395, + "step": 29507 + }, + { + "epoch": 5.37, + "learning_rate": 1.3914452129595922e-06, + "loss": 0.4707, + "step": 29508 + }, + { + "epoch": 5.37, + "learning_rate": 1.3913906079359302e-06, + "loss": 0.0728, + "step": 29509 + }, + { + "epoch": 5.37, + "learning_rate": 1.391336002912268e-06, + "loss": 0.0645, + "step": 29510 + }, + { + "epoch": 5.37, + "learning_rate": 1.3912813978886058e-06, + "loss": 0.5218, + "step": 29511 + }, + { + "epoch": 5.37, + "learning_rate": 1.3912267928649436e-06, + "loss": 0.1735, + "step": 29512 + }, + { + "epoch": 5.37, + "learning_rate": 1.3911721878412815e-06, + "loss": 0.0294, + "step": 29513 + }, + { + "epoch": 5.37, + "learning_rate": 1.3911175828176192e-06, + "loss": 1.5499, + "step": 29514 + }, + { + "epoch": 5.37, + "learning_rate": 1.391062977793957e-06, + "loss": 0.0203, + "step": 29515 + }, + { + "epoch": 5.37, + "learning_rate": 1.391008372770295e-06, + "loss": 0.7286, + "step": 29516 + }, + { + "epoch": 5.37, + "learning_rate": 1.3909537677466327e-06, + "loss": 0.0046, + "step": 29517 + }, + { + "epoch": 5.37, + "learning_rate": 1.3908991627229706e-06, + "loss": 0.0268, + "step": 29518 + }, + { + "epoch": 5.37, + "learning_rate": 1.3908445576993083e-06, + "loss": 0.4683, + "step": 29519 + }, + { + "epoch": 5.37, + "learning_rate": 1.3907899526756463e-06, + "loss": 1.2364, + "step": 29520 + }, + { + "epoch": 5.37, + "learning_rate": 1.390735347651984e-06, + "loss": 0.7854, + "step": 29521 + }, + { + "epoch": 5.37, + "learning_rate": 1.3906807426283217e-06, + "loss": 0.4799, + "step": 29522 + }, + { + "epoch": 5.37, + "learning_rate": 1.3906261376046597e-06, + "loss": 0.0477, + "step": 29523 + }, + { + "epoch": 5.37, + "learning_rate": 1.3905715325809974e-06, + "loss": 1.7359, + "step": 29524 + }, + { + "epoch": 5.37, + "learning_rate": 1.3905169275573354e-06, + "loss": 0.272, + "step": 29525 + }, + { + "epoch": 5.37, + "learning_rate": 1.390462322533673e-06, + "loss": 0.1517, + "step": 29526 + }, + { + "epoch": 5.37, + "learning_rate": 1.390407717510011e-06, + "loss": 0.1653, + "step": 29527 + }, + { + "epoch": 5.37, + "learning_rate": 1.3903531124863488e-06, + "loss": 0.0081, + "step": 29528 + }, + { + "epoch": 5.37, + "learning_rate": 1.3902985074626865e-06, + "loss": 1.8966, + "step": 29529 + }, + { + "epoch": 5.37, + "learning_rate": 1.3902439024390244e-06, + "loss": 1.7487, + "step": 29530 + }, + { + "epoch": 5.37, + "learning_rate": 1.3901892974153622e-06, + "loss": 0.2015, + "step": 29531 + }, + { + "epoch": 5.37, + "learning_rate": 1.3901346923917001e-06, + "loss": 0.0084, + "step": 29532 + }, + { + "epoch": 5.37, + "learning_rate": 1.3900800873680378e-06, + "loss": 0.0413, + "step": 29533 + }, + { + "epoch": 5.37, + "learning_rate": 1.3900254823443758e-06, + "loss": 0.0258, + "step": 29534 + }, + { + "epoch": 5.37, + "learning_rate": 1.3899708773207135e-06, + "loss": 0.096, + "step": 29535 + }, + { + "epoch": 5.37, + "learning_rate": 1.3899162722970512e-06, + "loss": 0.2333, + "step": 29536 + }, + { + "epoch": 5.37, + "learning_rate": 1.3898616672733892e-06, + "loss": 0.0247, + "step": 29537 + }, + { + "epoch": 5.37, + "learning_rate": 1.3898070622497271e-06, + "loss": 0.0116, + "step": 29538 + }, + { + "epoch": 5.37, + "learning_rate": 1.3897524572260649e-06, + "loss": 0.633, + "step": 29539 + }, + { + "epoch": 5.37, + "learning_rate": 1.3896978522024026e-06, + "loss": 0.7372, + "step": 29540 + }, + { + "epoch": 5.37, + "learning_rate": 1.3896432471787405e-06, + "loss": 0.0121, + "step": 29541 + }, + { + "epoch": 5.37, + "learning_rate": 1.3895886421550783e-06, + "loss": 1.3029, + "step": 29542 + }, + { + "epoch": 5.37, + "learning_rate": 1.389534037131416e-06, + "loss": 0.8464, + "step": 29543 + }, + { + "epoch": 5.37, + "learning_rate": 1.389479432107754e-06, + "loss": 0.049, + "step": 29544 + }, + { + "epoch": 5.37, + "learning_rate": 1.3894248270840919e-06, + "loss": 0.2148, + "step": 29545 + }, + { + "epoch": 5.37, + "learning_rate": 1.3893702220604296e-06, + "loss": 1.1531, + "step": 29546 + }, + { + "epoch": 5.37, + "learning_rate": 1.3893156170367674e-06, + "loss": 0.5444, + "step": 29547 + }, + { + "epoch": 5.37, + "learning_rate": 1.3892610120131053e-06, + "loss": 1.8089, + "step": 29548 + }, + { + "epoch": 5.37, + "learning_rate": 1.389206406989443e-06, + "loss": 0.741, + "step": 29549 + }, + { + "epoch": 5.37, + "learning_rate": 1.3891518019657808e-06, + "loss": 1.1138, + "step": 29550 + }, + { + "epoch": 5.37, + "learning_rate": 1.3890971969421187e-06, + "loss": 0.008, + "step": 29551 + }, + { + "epoch": 5.37, + "learning_rate": 1.3890425919184566e-06, + "loss": 0.1338, + "step": 29552 + }, + { + "epoch": 5.37, + "learning_rate": 1.3889879868947944e-06, + "loss": 0.8652, + "step": 29553 + }, + { + "epoch": 5.37, + "learning_rate": 1.3889333818711321e-06, + "loss": 2.045, + "step": 29554 + }, + { + "epoch": 5.37, + "learning_rate": 1.38887877684747e-06, + "loss": 0.6068, + "step": 29555 + }, + { + "epoch": 5.37, + "learning_rate": 1.3888241718238078e-06, + "loss": 0.3222, + "step": 29556 + }, + { + "epoch": 5.37, + "learning_rate": 1.3887695668001455e-06, + "loss": 0.0137, + "step": 29557 + }, + { + "epoch": 5.38, + "learning_rate": 1.3887149617764835e-06, + "loss": 0.3281, + "step": 29558 + }, + { + "epoch": 5.38, + "learning_rate": 1.3886603567528214e-06, + "loss": 0.9131, + "step": 29559 + }, + { + "epoch": 5.38, + "learning_rate": 1.3886057517291591e-06, + "loss": 0.1159, + "step": 29560 + }, + { + "epoch": 5.38, + "learning_rate": 1.3885511467054969e-06, + "loss": 0.2837, + "step": 29561 + }, + { + "epoch": 5.38, + "learning_rate": 1.3884965416818348e-06, + "loss": 0.1758, + "step": 29562 + }, + { + "epoch": 5.38, + "learning_rate": 1.3884419366581725e-06, + "loss": 0.2444, + "step": 29563 + }, + { + "epoch": 5.38, + "learning_rate": 1.3883873316345103e-06, + "loss": 3.1317, + "step": 29564 + }, + { + "epoch": 5.38, + "learning_rate": 1.3883327266108482e-06, + "loss": 0.043, + "step": 29565 + }, + { + "epoch": 5.38, + "learning_rate": 1.3882781215871862e-06, + "loss": 0.8817, + "step": 29566 + }, + { + "epoch": 5.38, + "learning_rate": 1.388223516563524e-06, + "loss": 2.8019, + "step": 29567 + }, + { + "epoch": 5.38, + "learning_rate": 1.3881689115398616e-06, + "loss": 1.0423, + "step": 29568 + }, + { + "epoch": 5.38, + "learning_rate": 1.3881143065161996e-06, + "loss": 0.5353, + "step": 29569 + }, + { + "epoch": 5.38, + "learning_rate": 1.3880597014925373e-06, + "loss": 1.1103, + "step": 29570 + }, + { + "epoch": 5.38, + "learning_rate": 1.388005096468875e-06, + "loss": 1.269, + "step": 29571 + }, + { + "epoch": 5.38, + "learning_rate": 1.387950491445213e-06, + "loss": 2.3507, + "step": 29572 + }, + { + "epoch": 5.38, + "learning_rate": 1.387895886421551e-06, + "loss": 0.6484, + "step": 29573 + }, + { + "epoch": 5.38, + "learning_rate": 1.3878412813978887e-06, + "loss": 0.5862, + "step": 29574 + }, + { + "epoch": 5.38, + "learning_rate": 1.3877866763742264e-06, + "loss": 0.3308, + "step": 29575 + }, + { + "epoch": 5.38, + "learning_rate": 1.3877320713505643e-06, + "loss": 0.23, + "step": 29576 + }, + { + "epoch": 5.38, + "learning_rate": 1.387677466326902e-06, + "loss": 1.2288, + "step": 29577 + }, + { + "epoch": 5.38, + "learning_rate": 1.38762286130324e-06, + "loss": 1.1634, + "step": 29578 + }, + { + "epoch": 5.38, + "learning_rate": 1.3875682562795777e-06, + "loss": 0.07, + "step": 29579 + }, + { + "epoch": 5.38, + "learning_rate": 1.3875136512559157e-06, + "loss": 2.9333, + "step": 29580 + }, + { + "epoch": 5.38, + "learning_rate": 1.3874590462322534e-06, + "loss": 0.0061, + "step": 29581 + }, + { + "epoch": 5.38, + "learning_rate": 1.3874044412085911e-06, + "loss": 0.4301, + "step": 29582 + }, + { + "epoch": 5.38, + "learning_rate": 1.387349836184929e-06, + "loss": 1.2429, + "step": 29583 + }, + { + "epoch": 5.38, + "learning_rate": 1.3872952311612668e-06, + "loss": 0.2, + "step": 29584 + }, + { + "epoch": 5.38, + "learning_rate": 1.3872406261376048e-06, + "loss": 0.324, + "step": 29585 + }, + { + "epoch": 5.38, + "learning_rate": 1.3871860211139425e-06, + "loss": 0.9292, + "step": 29586 + }, + { + "epoch": 5.38, + "learning_rate": 1.3871314160902804e-06, + "loss": 0.084, + "step": 29587 + }, + { + "epoch": 5.38, + "learning_rate": 1.3870768110666182e-06, + "loss": 1.2409, + "step": 29588 + }, + { + "epoch": 5.38, + "learning_rate": 1.387022206042956e-06, + "loss": 0.2929, + "step": 29589 + }, + { + "epoch": 5.38, + "learning_rate": 1.3869676010192938e-06, + "loss": 0.0439, + "step": 29590 + }, + { + "epoch": 5.38, + "learning_rate": 1.3869129959956316e-06, + "loss": 1.0776, + "step": 29591 + }, + { + "epoch": 5.38, + "learning_rate": 1.3868583909719695e-06, + "loss": 0.977, + "step": 29592 + }, + { + "epoch": 5.38, + "learning_rate": 1.3868037859483072e-06, + "loss": 1.5885, + "step": 29593 + }, + { + "epoch": 5.38, + "learning_rate": 1.3867491809246452e-06, + "loss": 0.6369, + "step": 29594 + }, + { + "epoch": 5.38, + "learning_rate": 1.386694575900983e-06, + "loss": 0.0355, + "step": 29595 + }, + { + "epoch": 5.38, + "learning_rate": 1.3866399708773207e-06, + "loss": 0.1377, + "step": 29596 + }, + { + "epoch": 5.38, + "learning_rate": 1.3865853658536586e-06, + "loss": 0.0258, + "step": 29597 + }, + { + "epoch": 5.38, + "learning_rate": 1.3865307608299963e-06, + "loss": 2.4682, + "step": 29598 + }, + { + "epoch": 5.38, + "learning_rate": 1.3864761558063343e-06, + "loss": 0.0134, + "step": 29599 + }, + { + "epoch": 5.38, + "learning_rate": 1.386421550782672e-06, + "loss": 1.6251, + "step": 29600 + }, + { + "epoch": 5.38, + "learning_rate": 1.38636694575901e-06, + "loss": 0.8816, + "step": 29601 + }, + { + "epoch": 5.38, + "learning_rate": 1.3863123407353477e-06, + "loss": 0.3939, + "step": 29602 + }, + { + "epoch": 5.38, + "learning_rate": 1.3862577357116854e-06, + "loss": 1.5464, + "step": 29603 + }, + { + "epoch": 5.38, + "learning_rate": 1.3862031306880234e-06, + "loss": 0.0056, + "step": 29604 + }, + { + "epoch": 5.38, + "learning_rate": 1.386148525664361e-06, + "loss": 0.2089, + "step": 29605 + }, + { + "epoch": 5.38, + "learning_rate": 1.386093920640699e-06, + "loss": 0.0296, + "step": 29606 + }, + { + "epoch": 5.38, + "learning_rate": 1.3860393156170368e-06, + "loss": 0.0757, + "step": 29607 + }, + { + "epoch": 5.38, + "learning_rate": 1.3859847105933747e-06, + "loss": 0.5341, + "step": 29608 + }, + { + "epoch": 5.38, + "learning_rate": 1.3859301055697124e-06, + "loss": 0.0305, + "step": 29609 + }, + { + "epoch": 5.38, + "learning_rate": 1.3858755005460502e-06, + "loss": 2.1494, + "step": 29610 + }, + { + "epoch": 5.38, + "learning_rate": 1.3858208955223881e-06, + "loss": 1.007, + "step": 29611 + }, + { + "epoch": 5.38, + "learning_rate": 1.3857662904987258e-06, + "loss": 0.6274, + "step": 29612 + }, + { + "epoch": 5.39, + "learning_rate": 1.3857116854750638e-06, + "loss": 0.4404, + "step": 29613 + }, + { + "epoch": 5.39, + "learning_rate": 1.3856570804514015e-06, + "loss": 1.3394, + "step": 29614 + }, + { + "epoch": 5.39, + "learning_rate": 1.3856024754277395e-06, + "loss": 0.3155, + "step": 29615 + }, + { + "epoch": 5.39, + "learning_rate": 1.3855478704040772e-06, + "loss": 0.0149, + "step": 29616 + }, + { + "epoch": 5.39, + "learning_rate": 1.385493265380415e-06, + "loss": 0.0108, + "step": 29617 + }, + { + "epoch": 5.39, + "learning_rate": 1.3854386603567529e-06, + "loss": 0.3448, + "step": 29618 + }, + { + "epoch": 5.39, + "learning_rate": 1.3853840553330908e-06, + "loss": 0.201, + "step": 29619 + }, + { + "epoch": 5.39, + "learning_rate": 1.3853294503094285e-06, + "loss": 0.0635, + "step": 29620 + }, + { + "epoch": 5.39, + "learning_rate": 1.3852748452857663e-06, + "loss": 1.3739, + "step": 29621 + }, + { + "epoch": 5.39, + "learning_rate": 1.3852202402621042e-06, + "loss": 1.7058, + "step": 29622 + }, + { + "epoch": 5.39, + "learning_rate": 1.385165635238442e-06, + "loss": 1.6604, + "step": 29623 + }, + { + "epoch": 5.39, + "learning_rate": 1.3851110302147797e-06, + "loss": 0.2019, + "step": 29624 + }, + { + "epoch": 5.39, + "learning_rate": 1.3850564251911176e-06, + "loss": 0.2921, + "step": 29625 + }, + { + "epoch": 5.39, + "learning_rate": 1.3850018201674556e-06, + "loss": 0.3454, + "step": 29626 + }, + { + "epoch": 5.39, + "learning_rate": 1.3849472151437933e-06, + "loss": 2.6729, + "step": 29627 + }, + { + "epoch": 5.39, + "learning_rate": 1.384892610120131e-06, + "loss": 0.0108, + "step": 29628 + }, + { + "epoch": 5.39, + "learning_rate": 1.384838005096469e-06, + "loss": 1.3286, + "step": 29629 + }, + { + "epoch": 5.39, + "learning_rate": 1.3847834000728067e-06, + "loss": 1.3182, + "step": 29630 + }, + { + "epoch": 5.39, + "learning_rate": 1.3847287950491444e-06, + "loss": 3.7341, + "step": 29631 + }, + { + "epoch": 5.39, + "learning_rate": 1.3846741900254824e-06, + "loss": 0.1348, + "step": 29632 + }, + { + "epoch": 5.39, + "learning_rate": 1.3846195850018203e-06, + "loss": 0.4646, + "step": 29633 + }, + { + "epoch": 5.39, + "learning_rate": 1.384564979978158e-06, + "loss": 0.5574, + "step": 29634 + }, + { + "epoch": 5.39, + "learning_rate": 1.3845103749544958e-06, + "loss": 0.8351, + "step": 29635 + }, + { + "epoch": 5.39, + "learning_rate": 1.3844557699308337e-06, + "loss": 0.0111, + "step": 29636 + }, + { + "epoch": 5.39, + "learning_rate": 1.3844011649071715e-06, + "loss": 1.343, + "step": 29637 + }, + { + "epoch": 5.39, + "learning_rate": 1.3843465598835092e-06, + "loss": 0.102, + "step": 29638 + }, + { + "epoch": 5.39, + "learning_rate": 1.3842919548598471e-06, + "loss": 0.4922, + "step": 29639 + }, + { + "epoch": 5.39, + "learning_rate": 1.384237349836185e-06, + "loss": 0.1136, + "step": 29640 + }, + { + "epoch": 5.39, + "learning_rate": 1.3841827448125228e-06, + "loss": 0.0175, + "step": 29641 + }, + { + "epoch": 5.39, + "learning_rate": 1.3841281397888605e-06, + "loss": 1.0138, + "step": 29642 + }, + { + "epoch": 5.39, + "learning_rate": 1.3840735347651985e-06, + "loss": 0.13, + "step": 29643 + }, + { + "epoch": 5.39, + "learning_rate": 1.3840189297415362e-06, + "loss": 0.4855, + "step": 29644 + }, + { + "epoch": 5.39, + "learning_rate": 1.383964324717874e-06, + "loss": 0.3259, + "step": 29645 + }, + { + "epoch": 5.39, + "learning_rate": 1.3839097196942119e-06, + "loss": 0.7285, + "step": 29646 + }, + { + "epoch": 5.39, + "learning_rate": 1.3838551146705498e-06, + "loss": 0.8946, + "step": 29647 + }, + { + "epoch": 5.39, + "learning_rate": 1.3838005096468876e-06, + "loss": 0.0151, + "step": 29648 + }, + { + "epoch": 5.39, + "learning_rate": 1.3837459046232253e-06, + "loss": 0.6229, + "step": 29649 + }, + { + "epoch": 5.39, + "learning_rate": 1.3836912995995632e-06, + "loss": 0.2137, + "step": 29650 + }, + { + "epoch": 5.39, + "learning_rate": 1.383636694575901e-06, + "loss": 1.0357, + "step": 29651 + }, + { + "epoch": 5.39, + "learning_rate": 1.3835820895522387e-06, + "loss": 2.9082, + "step": 29652 + }, + { + "epoch": 5.39, + "learning_rate": 1.3835274845285767e-06, + "loss": 0.0084, + "step": 29653 + }, + { + "epoch": 5.39, + "learning_rate": 1.3834728795049146e-06, + "loss": 0.3852, + "step": 29654 + }, + { + "epoch": 5.39, + "learning_rate": 1.3834182744812523e-06, + "loss": 1.2949, + "step": 29655 + }, + { + "epoch": 5.39, + "learning_rate": 1.38336366945759e-06, + "loss": 0.1078, + "step": 29656 + }, + { + "epoch": 5.39, + "learning_rate": 1.383309064433928e-06, + "loss": 0.911, + "step": 29657 + }, + { + "epoch": 5.39, + "learning_rate": 1.3832544594102657e-06, + "loss": 2.9784, + "step": 29658 + }, + { + "epoch": 5.39, + "learning_rate": 1.3831998543866035e-06, + "loss": 0.2603, + "step": 29659 + }, + { + "epoch": 5.39, + "learning_rate": 1.3831452493629414e-06, + "loss": 0.1663, + "step": 29660 + }, + { + "epoch": 5.39, + "learning_rate": 1.3830906443392794e-06, + "loss": 0.0083, + "step": 29661 + }, + { + "epoch": 5.39, + "learning_rate": 1.383036039315617e-06, + "loss": 0.4825, + "step": 29662 + }, + { + "epoch": 5.39, + "learning_rate": 1.3829814342919548e-06, + "loss": 3.599, + "step": 29663 + }, + { + "epoch": 5.39, + "learning_rate": 1.3829268292682928e-06, + "loss": 0.8054, + "step": 29664 + }, + { + "epoch": 5.39, + "learning_rate": 1.3828722242446305e-06, + "loss": 0.228, + "step": 29665 + }, + { + "epoch": 5.39, + "learning_rate": 1.3828176192209684e-06, + "loss": 0.0322, + "step": 29666 + }, + { + "epoch": 5.39, + "learning_rate": 1.3827630141973062e-06, + "loss": 2.8963, + "step": 29667 + }, + { + "epoch": 5.4, + "learning_rate": 1.3827084091736441e-06, + "loss": 0.1302, + "step": 29668 + }, + { + "epoch": 5.4, + "learning_rate": 1.3826538041499818e-06, + "loss": 0.2293, + "step": 29669 + }, + { + "epoch": 5.4, + "learning_rate": 1.3825991991263196e-06, + "loss": 0.4228, + "step": 29670 + }, + { + "epoch": 5.4, + "learning_rate": 1.3825445941026575e-06, + "loss": 0.8073, + "step": 29671 + }, + { + "epoch": 5.4, + "learning_rate": 1.3824899890789952e-06, + "loss": 1.1375, + "step": 29672 + }, + { + "epoch": 5.4, + "learning_rate": 1.3824353840553332e-06, + "loss": 0.023, + "step": 29673 + }, + { + "epoch": 5.4, + "learning_rate": 1.382380779031671e-06, + "loss": 1.2452, + "step": 29674 + }, + { + "epoch": 5.4, + "learning_rate": 1.3823261740080089e-06, + "loss": 0.4403, + "step": 29675 + }, + { + "epoch": 5.4, + "learning_rate": 1.3822715689843466e-06, + "loss": 0.1358, + "step": 29676 + }, + { + "epoch": 5.4, + "learning_rate": 1.3822169639606843e-06, + "loss": 0.0439, + "step": 29677 + }, + { + "epoch": 5.4, + "learning_rate": 1.3821623589370223e-06, + "loss": 0.3141, + "step": 29678 + }, + { + "epoch": 5.4, + "learning_rate": 1.38210775391336e-06, + "loss": 0.3675, + "step": 29679 + }, + { + "epoch": 5.4, + "learning_rate": 1.382053148889698e-06, + "loss": 0.4882, + "step": 29680 + }, + { + "epoch": 5.4, + "learning_rate": 1.3819985438660357e-06, + "loss": 0.5363, + "step": 29681 + }, + { + "epoch": 5.4, + "learning_rate": 1.3819439388423736e-06, + "loss": 0.0656, + "step": 29682 + }, + { + "epoch": 5.4, + "learning_rate": 1.3818893338187114e-06, + "loss": 1.8531, + "step": 29683 + }, + { + "epoch": 5.4, + "learning_rate": 1.381834728795049e-06, + "loss": 0.5342, + "step": 29684 + }, + { + "epoch": 5.4, + "learning_rate": 1.381780123771387e-06, + "loss": 0.3597, + "step": 29685 + }, + { + "epoch": 5.4, + "learning_rate": 1.3817255187477248e-06, + "loss": 0.4651, + "step": 29686 + }, + { + "epoch": 5.4, + "learning_rate": 1.3816709137240627e-06, + "loss": 0.0121, + "step": 29687 + }, + { + "epoch": 5.4, + "learning_rate": 1.3816163087004004e-06, + "loss": 0.0719, + "step": 29688 + }, + { + "epoch": 5.4, + "learning_rate": 1.3815617036767384e-06, + "loss": 0.0333, + "step": 29689 + }, + { + "epoch": 5.4, + "learning_rate": 1.3815070986530761e-06, + "loss": 0.0129, + "step": 29690 + }, + { + "epoch": 5.4, + "learning_rate": 1.3814524936294138e-06, + "loss": 0.4418, + "step": 29691 + }, + { + "epoch": 5.4, + "learning_rate": 1.3813978886057518e-06, + "loss": 1.3879, + "step": 29692 + }, + { + "epoch": 5.4, + "learning_rate": 1.3813432835820895e-06, + "loss": 0.7824, + "step": 29693 + }, + { + "epoch": 5.4, + "learning_rate": 1.3812886785584275e-06, + "loss": 0.0378, + "step": 29694 + }, + { + "epoch": 5.4, + "learning_rate": 1.3812340735347652e-06, + "loss": 0.0635, + "step": 29695 + }, + { + "epoch": 5.4, + "learning_rate": 1.3811794685111031e-06, + "loss": 1.8721, + "step": 29696 + }, + { + "epoch": 5.4, + "learning_rate": 1.3811248634874409e-06, + "loss": 0.3689, + "step": 29697 + }, + { + "epoch": 5.4, + "learning_rate": 1.3810702584637786e-06, + "loss": 0.0135, + "step": 29698 + }, + { + "epoch": 5.4, + "learning_rate": 1.3810156534401165e-06, + "loss": 0.3782, + "step": 29699 + }, + { + "epoch": 5.4, + "learning_rate": 1.3809610484164543e-06, + "loss": 1.382, + "step": 29700 + }, + { + "epoch": 5.4, + "learning_rate": 1.3809064433927922e-06, + "loss": 2.7419, + "step": 29701 + }, + { + "epoch": 5.4, + "learning_rate": 1.38085183836913e-06, + "loss": 0.8145, + "step": 29702 + }, + { + "epoch": 5.4, + "learning_rate": 1.3807972333454679e-06, + "loss": 1.4104, + "step": 29703 + }, + { + "epoch": 5.4, + "learning_rate": 1.3807426283218056e-06, + "loss": 0.0899, + "step": 29704 + }, + { + "epoch": 5.4, + "learning_rate": 1.3806880232981434e-06, + "loss": 0.2753, + "step": 29705 + }, + { + "epoch": 5.4, + "learning_rate": 1.3806334182744813e-06, + "loss": 0.0078, + "step": 29706 + }, + { + "epoch": 5.4, + "learning_rate": 1.3805788132508192e-06, + "loss": 0.3146, + "step": 29707 + }, + { + "epoch": 5.4, + "learning_rate": 1.380524208227157e-06, + "loss": 0.1131, + "step": 29708 + }, + { + "epoch": 5.4, + "learning_rate": 1.3804696032034947e-06, + "loss": 1.1161, + "step": 29709 + }, + { + "epoch": 5.4, + "learning_rate": 1.3804149981798326e-06, + "loss": 2.104, + "step": 29710 + }, + { + "epoch": 5.4, + "learning_rate": 1.3803603931561704e-06, + "loss": 0.015, + "step": 29711 + }, + { + "epoch": 5.4, + "learning_rate": 1.3803057881325081e-06, + "loss": 0.3713, + "step": 29712 + }, + { + "epoch": 5.4, + "learning_rate": 1.380251183108846e-06, + "loss": 0.4825, + "step": 29713 + }, + { + "epoch": 5.4, + "learning_rate": 1.380196578085184e-06, + "loss": 2.1505, + "step": 29714 + }, + { + "epoch": 5.4, + "learning_rate": 1.3801419730615217e-06, + "loss": 0.3132, + "step": 29715 + }, + { + "epoch": 5.4, + "learning_rate": 1.3800873680378595e-06, + "loss": 2.6613, + "step": 29716 + }, + { + "epoch": 5.4, + "learning_rate": 1.3800327630141974e-06, + "loss": 0.8602, + "step": 29717 + }, + { + "epoch": 5.4, + "learning_rate": 1.3799781579905351e-06, + "loss": 0.5434, + "step": 29718 + }, + { + "epoch": 5.4, + "learning_rate": 1.3799235529668729e-06, + "loss": 0.1655, + "step": 29719 + }, + { + "epoch": 5.4, + "learning_rate": 1.3798689479432108e-06, + "loss": 0.4084, + "step": 29720 + }, + { + "epoch": 5.4, + "learning_rate": 1.3798143429195488e-06, + "loss": 1.655, + "step": 29721 + }, + { + "epoch": 5.4, + "learning_rate": 1.3797597378958865e-06, + "loss": 0.5507, + "step": 29722 + }, + { + "epoch": 5.41, + "learning_rate": 1.3797051328722242e-06, + "loss": 0.3386, + "step": 29723 + }, + { + "epoch": 5.41, + "learning_rate": 1.3796505278485622e-06, + "loss": 0.8119, + "step": 29724 + }, + { + "epoch": 5.41, + "learning_rate": 1.3795959228248999e-06, + "loss": 0.2336, + "step": 29725 + }, + { + "epoch": 5.41, + "learning_rate": 1.3795413178012376e-06, + "loss": 4.235, + "step": 29726 + }, + { + "epoch": 5.41, + "learning_rate": 1.3794867127775756e-06, + "loss": 1.2175, + "step": 29727 + }, + { + "epoch": 5.41, + "learning_rate": 1.3794321077539135e-06, + "loss": 1.2958, + "step": 29728 + }, + { + "epoch": 5.41, + "learning_rate": 1.3793775027302512e-06, + "loss": 0.0567, + "step": 29729 + }, + { + "epoch": 5.41, + "learning_rate": 1.379322897706589e-06, + "loss": 0.5563, + "step": 29730 + }, + { + "epoch": 5.41, + "learning_rate": 1.379268292682927e-06, + "loss": 1.468, + "step": 29731 + }, + { + "epoch": 5.41, + "learning_rate": 1.3792136876592646e-06, + "loss": 0.2076, + "step": 29732 + }, + { + "epoch": 5.41, + "learning_rate": 1.3791590826356024e-06, + "loss": 0.7605, + "step": 29733 + }, + { + "epoch": 5.41, + "learning_rate": 1.3791044776119403e-06, + "loss": 0.5117, + "step": 29734 + }, + { + "epoch": 5.41, + "learning_rate": 1.3790498725882783e-06, + "loss": 0.0298, + "step": 29735 + }, + { + "epoch": 5.41, + "learning_rate": 1.378995267564616e-06, + "loss": 0.1382, + "step": 29736 + }, + { + "epoch": 5.41, + "learning_rate": 1.3789406625409537e-06, + "loss": 0.1542, + "step": 29737 + }, + { + "epoch": 5.41, + "learning_rate": 1.3788860575172917e-06, + "loss": 2.3576, + "step": 29738 + }, + { + "epoch": 5.41, + "learning_rate": 1.3788314524936294e-06, + "loss": 0.2992, + "step": 29739 + }, + { + "epoch": 5.41, + "learning_rate": 1.3787768474699671e-06, + "loss": 0.7171, + "step": 29740 + }, + { + "epoch": 5.41, + "learning_rate": 1.378722242446305e-06, + "loss": 0.008, + "step": 29741 + }, + { + "epoch": 5.41, + "learning_rate": 1.378667637422643e-06, + "loss": 0.4395, + "step": 29742 + }, + { + "epoch": 5.41, + "learning_rate": 1.3786130323989808e-06, + "loss": 0.0305, + "step": 29743 + }, + { + "epoch": 5.41, + "learning_rate": 1.3785584273753185e-06, + "loss": 0.8315, + "step": 29744 + }, + { + "epoch": 5.41, + "learning_rate": 1.3785038223516564e-06, + "loss": 1.0994, + "step": 29745 + }, + { + "epoch": 5.41, + "learning_rate": 1.3784492173279942e-06, + "loss": 0.1125, + "step": 29746 + }, + { + "epoch": 5.41, + "learning_rate": 1.3783946123043321e-06, + "loss": 1.0545, + "step": 29747 + }, + { + "epoch": 5.41, + "learning_rate": 1.3783400072806698e-06, + "loss": 0.0276, + "step": 29748 + }, + { + "epoch": 5.41, + "learning_rate": 1.3782854022570078e-06, + "loss": 0.4918, + "step": 29749 + }, + { + "epoch": 5.41, + "learning_rate": 1.3782307972333455e-06, + "loss": 2.3737, + "step": 29750 + }, + { + "epoch": 5.41, + "learning_rate": 1.3781761922096832e-06, + "loss": 0.0182, + "step": 29751 + }, + { + "epoch": 5.41, + "learning_rate": 1.3781215871860212e-06, + "loss": 0.1003, + "step": 29752 + }, + { + "epoch": 5.41, + "learning_rate": 1.378066982162359e-06, + "loss": 0.923, + "step": 29753 + }, + { + "epoch": 5.41, + "learning_rate": 1.3780123771386969e-06, + "loss": 0.6606, + "step": 29754 + }, + { + "epoch": 5.41, + "learning_rate": 1.3779577721150346e-06, + "loss": 0.1445, + "step": 29755 + }, + { + "epoch": 5.41, + "learning_rate": 1.3779031670913725e-06, + "loss": 0.479, + "step": 29756 + }, + { + "epoch": 5.41, + "learning_rate": 1.3778485620677103e-06, + "loss": 0.0078, + "step": 29757 + }, + { + "epoch": 5.41, + "learning_rate": 1.377793957044048e-06, + "loss": 0.1174, + "step": 29758 + }, + { + "epoch": 5.41, + "learning_rate": 1.377739352020386e-06, + "loss": 0.433, + "step": 29759 + }, + { + "epoch": 5.41, + "learning_rate": 1.3776847469967237e-06, + "loss": 0.1564, + "step": 29760 + }, + { + "epoch": 5.41, + "learning_rate": 1.3776301419730616e-06, + "loss": 0.214, + "step": 29761 + }, + { + "epoch": 5.41, + "learning_rate": 1.3775755369493994e-06, + "loss": 0.1308, + "step": 29762 + }, + { + "epoch": 5.41, + "learning_rate": 1.3775209319257373e-06, + "loss": 1.1846, + "step": 29763 + }, + { + "epoch": 5.41, + "learning_rate": 1.377466326902075e-06, + "loss": 0.9338, + "step": 29764 + }, + { + "epoch": 5.41, + "learning_rate": 1.3774117218784128e-06, + "loss": 2.5608, + "step": 29765 + }, + { + "epoch": 5.41, + "learning_rate": 1.3773571168547507e-06, + "loss": 0.4045, + "step": 29766 + }, + { + "epoch": 5.41, + "learning_rate": 1.3773025118310884e-06, + "loss": 0.3264, + "step": 29767 + }, + { + "epoch": 5.41, + "learning_rate": 1.3772479068074264e-06, + "loss": 1.3496, + "step": 29768 + }, + { + "epoch": 5.41, + "learning_rate": 1.3771933017837641e-06, + "loss": 0.1208, + "step": 29769 + }, + { + "epoch": 5.41, + "learning_rate": 1.377138696760102e-06, + "loss": 0.1461, + "step": 29770 + }, + { + "epoch": 5.41, + "learning_rate": 1.3770840917364398e-06, + "loss": 1.2534, + "step": 29771 + }, + { + "epoch": 5.41, + "learning_rate": 1.3770294867127775e-06, + "loss": 0.7456, + "step": 29772 + }, + { + "epoch": 5.41, + "learning_rate": 1.3769748816891155e-06, + "loss": 0.9951, + "step": 29773 + }, + { + "epoch": 5.41, + "learning_rate": 1.3769202766654532e-06, + "loss": 1.0623, + "step": 29774 + }, + { + "epoch": 5.41, + "learning_rate": 1.3768656716417911e-06, + "loss": 0.3469, + "step": 29775 + }, + { + "epoch": 5.41, + "learning_rate": 1.3768110666181289e-06, + "loss": 0.1025, + "step": 29776 + }, + { + "epoch": 5.41, + "learning_rate": 1.3767564615944668e-06, + "loss": 0.0678, + "step": 29777 + }, + { + "epoch": 5.42, + "learning_rate": 1.3767018565708045e-06, + "loss": 1.1292, + "step": 29778 + }, + { + "epoch": 5.42, + "learning_rate": 1.3766472515471423e-06, + "loss": 0.0343, + "step": 29779 + }, + { + "epoch": 5.42, + "learning_rate": 1.3765926465234802e-06, + "loss": 1.3823, + "step": 29780 + }, + { + "epoch": 5.42, + "learning_rate": 1.376538041499818e-06, + "loss": 1.2881, + "step": 29781 + }, + { + "epoch": 5.42, + "learning_rate": 1.3764834364761559e-06, + "loss": 0.4278, + "step": 29782 + }, + { + "epoch": 5.42, + "learning_rate": 1.3764288314524936e-06, + "loss": 1.575, + "step": 29783 + }, + { + "epoch": 5.42, + "learning_rate": 1.3763742264288316e-06, + "loss": 0.1534, + "step": 29784 + }, + { + "epoch": 5.42, + "learning_rate": 1.3763196214051693e-06, + "loss": 0.6258, + "step": 29785 + }, + { + "epoch": 5.42, + "learning_rate": 1.376265016381507e-06, + "loss": 1.9602, + "step": 29786 + }, + { + "epoch": 5.42, + "learning_rate": 1.376210411357845e-06, + "loss": 0.288, + "step": 29787 + }, + { + "epoch": 5.42, + "learning_rate": 1.3761558063341827e-06, + "loss": 0.0355, + "step": 29788 + }, + { + "epoch": 5.42, + "learning_rate": 1.3761012013105206e-06, + "loss": 1.2717, + "step": 29789 + }, + { + "epoch": 5.42, + "learning_rate": 1.3760465962868584e-06, + "loss": 0.0243, + "step": 29790 + }, + { + "epoch": 5.42, + "learning_rate": 1.3759919912631963e-06, + "loss": 0.1251, + "step": 29791 + }, + { + "epoch": 5.42, + "learning_rate": 1.375937386239534e-06, + "loss": 0.0297, + "step": 29792 + }, + { + "epoch": 5.42, + "learning_rate": 1.3758827812158718e-06, + "loss": 2.5485, + "step": 29793 + }, + { + "epoch": 5.42, + "learning_rate": 1.3758281761922097e-06, + "loss": 0.15, + "step": 29794 + }, + { + "epoch": 5.42, + "learning_rate": 1.3757735711685477e-06, + "loss": 2.1903, + "step": 29795 + }, + { + "epoch": 5.42, + "learning_rate": 1.3757189661448854e-06, + "loss": 0.04, + "step": 29796 + }, + { + "epoch": 5.42, + "learning_rate": 1.3756643611212231e-06, + "loss": 0.2724, + "step": 29797 + }, + { + "epoch": 5.42, + "learning_rate": 1.375609756097561e-06, + "loss": 0.5787, + "step": 29798 + }, + { + "epoch": 5.42, + "learning_rate": 1.3755551510738988e-06, + "loss": 0.6362, + "step": 29799 + }, + { + "epoch": 5.42, + "learning_rate": 1.3755005460502365e-06, + "loss": 0.2063, + "step": 29800 + }, + { + "epoch": 5.42, + "learning_rate": 1.3754459410265745e-06, + "loss": 0.1881, + "step": 29801 + }, + { + "epoch": 5.42, + "learning_rate": 1.3753913360029124e-06, + "loss": 0.4991, + "step": 29802 + }, + { + "epoch": 5.42, + "learning_rate": 1.3753367309792502e-06, + "loss": 0.0125, + "step": 29803 + }, + { + "epoch": 5.42, + "learning_rate": 1.3752821259555879e-06, + "loss": 0.3238, + "step": 29804 + }, + { + "epoch": 5.42, + "learning_rate": 1.3752275209319258e-06, + "loss": 0.0149, + "step": 29805 + }, + { + "epoch": 5.42, + "learning_rate": 1.3751729159082636e-06, + "loss": 0.0842, + "step": 29806 + }, + { + "epoch": 5.42, + "learning_rate": 1.3751183108846013e-06, + "loss": 1.2552, + "step": 29807 + }, + { + "epoch": 5.42, + "learning_rate": 1.3750637058609392e-06, + "loss": 2.6972, + "step": 29808 + }, + { + "epoch": 5.42, + "learning_rate": 1.3750091008372772e-06, + "loss": 0.9455, + "step": 29809 + }, + { + "epoch": 5.42, + "learning_rate": 1.374954495813615e-06, + "loss": 0.019, + "step": 29810 + }, + { + "epoch": 5.42, + "learning_rate": 1.3748998907899526e-06, + "loss": 0.259, + "step": 29811 + }, + { + "epoch": 5.42, + "learning_rate": 1.3748452857662906e-06, + "loss": 1.3514, + "step": 29812 + }, + { + "epoch": 5.42, + "learning_rate": 1.3747906807426283e-06, + "loss": 0.0963, + "step": 29813 + }, + { + "epoch": 5.42, + "learning_rate": 1.374736075718966e-06, + "loss": 0.1953, + "step": 29814 + }, + { + "epoch": 5.42, + "learning_rate": 1.374681470695304e-06, + "loss": 0.6896, + "step": 29815 + }, + { + "epoch": 5.42, + "learning_rate": 1.374626865671642e-06, + "loss": 0.0208, + "step": 29816 + }, + { + "epoch": 5.42, + "learning_rate": 1.3745722606479797e-06, + "loss": 0.2549, + "step": 29817 + }, + { + "epoch": 5.42, + "learning_rate": 1.3745176556243174e-06, + "loss": 0.2103, + "step": 29818 + }, + { + "epoch": 5.42, + "learning_rate": 1.3744630506006553e-06, + "loss": 0.2233, + "step": 29819 + }, + { + "epoch": 5.42, + "learning_rate": 1.374408445576993e-06, + "loss": 2.1772, + "step": 29820 + }, + { + "epoch": 5.42, + "learning_rate": 1.3743538405533308e-06, + "loss": 0.016, + "step": 29821 + }, + { + "epoch": 5.42, + "learning_rate": 1.3742992355296688e-06, + "loss": 0.2691, + "step": 29822 + }, + { + "epoch": 5.42, + "learning_rate": 1.3742446305060067e-06, + "loss": 0.0344, + "step": 29823 + }, + { + "epoch": 5.42, + "learning_rate": 1.3741900254823444e-06, + "loss": 0.0943, + "step": 29824 + }, + { + "epoch": 5.42, + "learning_rate": 1.3741354204586822e-06, + "loss": 0.9243, + "step": 29825 + }, + { + "epoch": 5.42, + "learning_rate": 1.37408081543502e-06, + "loss": 0.1139, + "step": 29826 + }, + { + "epoch": 5.42, + "learning_rate": 1.3740262104113578e-06, + "loss": 0.0057, + "step": 29827 + }, + { + "epoch": 5.42, + "learning_rate": 1.3739716053876956e-06, + "loss": 1.3444, + "step": 29828 + }, + { + "epoch": 5.42, + "learning_rate": 1.3739170003640335e-06, + "loss": 0.6412, + "step": 29829 + }, + { + "epoch": 5.42, + "learning_rate": 1.3738623953403715e-06, + "loss": 2.8016, + "step": 29830 + }, + { + "epoch": 5.42, + "learning_rate": 1.3738077903167092e-06, + "loss": 0.3301, + "step": 29831 + }, + { + "epoch": 5.42, + "learning_rate": 1.373753185293047e-06, + "loss": 1.2777, + "step": 29832 + }, + { + "epoch": 5.43, + "learning_rate": 1.3736985802693849e-06, + "loss": 0.0208, + "step": 29833 + }, + { + "epoch": 5.43, + "learning_rate": 1.3736439752457226e-06, + "loss": 0.2076, + "step": 29834 + }, + { + "epoch": 5.43, + "learning_rate": 1.3735893702220605e-06, + "loss": 0.8919, + "step": 29835 + }, + { + "epoch": 5.43, + "learning_rate": 1.3735347651983983e-06, + "loss": 0.3233, + "step": 29836 + }, + { + "epoch": 5.43, + "learning_rate": 1.3734801601747362e-06, + "loss": 0.0273, + "step": 29837 + }, + { + "epoch": 5.43, + "learning_rate": 1.373425555151074e-06, + "loss": 0.6406, + "step": 29838 + }, + { + "epoch": 5.43, + "learning_rate": 1.3733709501274117e-06, + "loss": 0.7102, + "step": 29839 + }, + { + "epoch": 5.43, + "learning_rate": 1.3733163451037496e-06, + "loss": 0.2428, + "step": 29840 + }, + { + "epoch": 5.43, + "learning_rate": 1.3732617400800874e-06, + "loss": 1.3424, + "step": 29841 + }, + { + "epoch": 5.43, + "learning_rate": 1.3732071350564253e-06, + "loss": 0.5783, + "step": 29842 + }, + { + "epoch": 5.43, + "learning_rate": 1.373152530032763e-06, + "loss": 1.8168, + "step": 29843 + }, + { + "epoch": 5.43, + "learning_rate": 1.373097925009101e-06, + "loss": 0.3078, + "step": 29844 + }, + { + "epoch": 5.43, + "learning_rate": 1.3730433199854387e-06, + "loss": 0.5995, + "step": 29845 + }, + { + "epoch": 5.43, + "learning_rate": 1.3729887149617764e-06, + "loss": 1.7964, + "step": 29846 + }, + { + "epoch": 5.43, + "learning_rate": 1.3729341099381144e-06, + "loss": 0.0126, + "step": 29847 + }, + { + "epoch": 5.43, + "learning_rate": 1.3728795049144521e-06, + "loss": 0.0073, + "step": 29848 + }, + { + "epoch": 5.43, + "learning_rate": 1.37282489989079e-06, + "loss": 0.3261, + "step": 29849 + }, + { + "epoch": 5.43, + "learning_rate": 1.3727702948671278e-06, + "loss": 1.5385, + "step": 29850 + }, + { + "epoch": 5.43, + "learning_rate": 1.3727156898434657e-06, + "loss": 0.3926, + "step": 29851 + }, + { + "epoch": 5.43, + "learning_rate": 1.3726610848198035e-06, + "loss": 1.3495, + "step": 29852 + }, + { + "epoch": 5.43, + "learning_rate": 1.3726064797961412e-06, + "loss": 0.0099, + "step": 29853 + }, + { + "epoch": 5.43, + "learning_rate": 1.3725518747724791e-06, + "loss": 0.2964, + "step": 29854 + }, + { + "epoch": 5.43, + "learning_rate": 1.3724972697488169e-06, + "loss": 1.537, + "step": 29855 + }, + { + "epoch": 5.43, + "learning_rate": 1.3724426647251548e-06, + "loss": 0.1926, + "step": 29856 + }, + { + "epoch": 5.43, + "learning_rate": 1.3723880597014925e-06, + "loss": 0.4314, + "step": 29857 + }, + { + "epoch": 5.43, + "learning_rate": 1.3723334546778305e-06, + "loss": 0.3324, + "step": 29858 + }, + { + "epoch": 5.43, + "learning_rate": 1.3722788496541682e-06, + "loss": 0.2931, + "step": 29859 + }, + { + "epoch": 5.43, + "learning_rate": 1.372224244630506e-06, + "loss": 0.1863, + "step": 29860 + }, + { + "epoch": 5.43, + "learning_rate": 1.3721696396068439e-06, + "loss": 0.0315, + "step": 29861 + }, + { + "epoch": 5.43, + "learning_rate": 1.3721150345831816e-06, + "loss": 0.3908, + "step": 29862 + }, + { + "epoch": 5.43, + "learning_rate": 1.3720604295595196e-06, + "loss": 0.3073, + "step": 29863 + }, + { + "epoch": 5.43, + "learning_rate": 1.3720058245358573e-06, + "loss": 0.3579, + "step": 29864 + }, + { + "epoch": 5.43, + "learning_rate": 1.3719512195121952e-06, + "loss": 0.5081, + "step": 29865 + }, + { + "epoch": 5.43, + "learning_rate": 1.371896614488533e-06, + "loss": 1.2287, + "step": 29866 + }, + { + "epoch": 5.43, + "learning_rate": 1.3718420094648707e-06, + "loss": 0.7587, + "step": 29867 + }, + { + "epoch": 5.43, + "learning_rate": 1.3717874044412086e-06, + "loss": 0.117, + "step": 29868 + }, + { + "epoch": 5.43, + "learning_rate": 1.3717327994175464e-06, + "loss": 0.544, + "step": 29869 + }, + { + "epoch": 5.43, + "learning_rate": 1.3716781943938843e-06, + "loss": 1.2519, + "step": 29870 + }, + { + "epoch": 5.43, + "learning_rate": 1.371623589370222e-06, + "loss": 2.0891, + "step": 29871 + }, + { + "epoch": 5.43, + "learning_rate": 1.37156898434656e-06, + "loss": 0.0058, + "step": 29872 + }, + { + "epoch": 5.43, + "learning_rate": 1.3715143793228977e-06, + "loss": 0.1034, + "step": 29873 + }, + { + "epoch": 5.43, + "learning_rate": 1.3714597742992355e-06, + "loss": 0.2124, + "step": 29874 + }, + { + "epoch": 5.43, + "learning_rate": 1.3714051692755734e-06, + "loss": 1.1027, + "step": 29875 + }, + { + "epoch": 5.43, + "learning_rate": 1.3713505642519113e-06, + "loss": 0.1387, + "step": 29876 + }, + { + "epoch": 5.43, + "learning_rate": 1.371295959228249e-06, + "loss": 0.2494, + "step": 29877 + }, + { + "epoch": 5.43, + "learning_rate": 1.3712413542045868e-06, + "loss": 0.43, + "step": 29878 + }, + { + "epoch": 5.43, + "learning_rate": 1.3711867491809248e-06, + "loss": 0.8854, + "step": 29879 + }, + { + "epoch": 5.43, + "learning_rate": 1.3711321441572625e-06, + "loss": 0.2079, + "step": 29880 + }, + { + "epoch": 5.43, + "learning_rate": 1.3710775391336002e-06, + "loss": 0.1923, + "step": 29881 + }, + { + "epoch": 5.43, + "learning_rate": 1.3710229341099382e-06, + "loss": 1.5905, + "step": 29882 + }, + { + "epoch": 5.43, + "learning_rate": 1.370968329086276e-06, + "loss": 0.8292, + "step": 29883 + }, + { + "epoch": 5.43, + "learning_rate": 1.3709137240626138e-06, + "loss": 1.3216, + "step": 29884 + }, + { + "epoch": 5.43, + "learning_rate": 1.3708591190389516e-06, + "loss": 0.0079, + "step": 29885 + }, + { + "epoch": 5.43, + "learning_rate": 1.3708045140152895e-06, + "loss": 0.3592, + "step": 29886 + }, + { + "epoch": 5.43, + "learning_rate": 1.3707499089916272e-06, + "loss": 0.9159, + "step": 29887 + }, + { + "epoch": 5.44, + "learning_rate": 1.370695303967965e-06, + "loss": 1.5606, + "step": 29888 + }, + { + "epoch": 5.44, + "learning_rate": 1.370640698944303e-06, + "loss": 0.6397, + "step": 29889 + }, + { + "epoch": 5.44, + "learning_rate": 1.3705860939206409e-06, + "loss": 0.595, + "step": 29890 + }, + { + "epoch": 5.44, + "learning_rate": 1.3705314888969786e-06, + "loss": 0.0603, + "step": 29891 + }, + { + "epoch": 5.44, + "learning_rate": 1.3704768838733163e-06, + "loss": 1.9334, + "step": 29892 + }, + { + "epoch": 5.44, + "learning_rate": 1.3704222788496543e-06, + "loss": 0.1491, + "step": 29893 + }, + { + "epoch": 5.44, + "learning_rate": 1.370367673825992e-06, + "loss": 1.8367, + "step": 29894 + }, + { + "epoch": 5.44, + "learning_rate": 1.3703130688023297e-06, + "loss": 0.0122, + "step": 29895 + }, + { + "epoch": 5.44, + "learning_rate": 1.3702584637786677e-06, + "loss": 0.4653, + "step": 29896 + }, + { + "epoch": 5.44, + "learning_rate": 1.3702038587550056e-06, + "loss": 0.0183, + "step": 29897 + }, + { + "epoch": 5.44, + "learning_rate": 1.3701492537313433e-06, + "loss": 0.1422, + "step": 29898 + }, + { + "epoch": 5.44, + "learning_rate": 1.370094648707681e-06, + "loss": 0.622, + "step": 29899 + }, + { + "epoch": 5.44, + "learning_rate": 1.370040043684019e-06, + "loss": 0.8016, + "step": 29900 + }, + { + "epoch": 5.44, + "learning_rate": 1.3699854386603568e-06, + "loss": 0.8086, + "step": 29901 + }, + { + "epoch": 5.44, + "learning_rate": 1.3699308336366945e-06, + "loss": 0.586, + "step": 29902 + }, + { + "epoch": 5.44, + "learning_rate": 1.3698762286130324e-06, + "loss": 0.08, + "step": 29903 + }, + { + "epoch": 5.44, + "learning_rate": 1.3698216235893704e-06, + "loss": 1.504, + "step": 29904 + }, + { + "epoch": 5.44, + "learning_rate": 1.369767018565708e-06, + "loss": 1.3451, + "step": 29905 + }, + { + "epoch": 5.44, + "learning_rate": 1.3697124135420458e-06, + "loss": 0.8962, + "step": 29906 + }, + { + "epoch": 5.44, + "learning_rate": 1.3696578085183838e-06, + "loss": 0.2684, + "step": 29907 + }, + { + "epoch": 5.44, + "learning_rate": 1.3696032034947215e-06, + "loss": 0.6406, + "step": 29908 + }, + { + "epoch": 5.44, + "learning_rate": 1.3695485984710592e-06, + "loss": 0.2989, + "step": 29909 + }, + { + "epoch": 5.44, + "learning_rate": 1.3694939934473972e-06, + "loss": 0.2288, + "step": 29910 + }, + { + "epoch": 5.44, + "learning_rate": 1.3694393884237351e-06, + "loss": 0.0045, + "step": 29911 + }, + { + "epoch": 5.44, + "learning_rate": 1.3693847834000729e-06, + "loss": 0.2313, + "step": 29912 + }, + { + "epoch": 5.44, + "learning_rate": 1.3693301783764106e-06, + "loss": 0.3674, + "step": 29913 + }, + { + "epoch": 5.44, + "learning_rate": 1.3692755733527485e-06, + "loss": 2.5492, + "step": 29914 + }, + { + "epoch": 5.44, + "learning_rate": 1.3692209683290863e-06, + "loss": 0.0993, + "step": 29915 + }, + { + "epoch": 5.44, + "learning_rate": 1.3691663633054242e-06, + "loss": 0.03, + "step": 29916 + }, + { + "epoch": 5.44, + "learning_rate": 1.369111758281762e-06, + "loss": 0.1317, + "step": 29917 + }, + { + "epoch": 5.44, + "learning_rate": 1.3690571532580999e-06, + "loss": 0.0162, + "step": 29918 + }, + { + "epoch": 5.44, + "learning_rate": 1.3690025482344376e-06, + "loss": 0.24, + "step": 29919 + }, + { + "epoch": 5.44, + "learning_rate": 1.3689479432107753e-06, + "loss": 0.1673, + "step": 29920 + }, + { + "epoch": 5.44, + "learning_rate": 1.3688933381871133e-06, + "loss": 0.0083, + "step": 29921 + }, + { + "epoch": 5.44, + "learning_rate": 1.368838733163451e-06, + "loss": 0.3184, + "step": 29922 + }, + { + "epoch": 5.44, + "learning_rate": 1.368784128139789e-06, + "loss": 1.403, + "step": 29923 + }, + { + "epoch": 5.44, + "learning_rate": 1.3687295231161267e-06, + "loss": 1.6035, + "step": 29924 + }, + { + "epoch": 5.44, + "learning_rate": 1.3686749180924646e-06, + "loss": 0.1595, + "step": 29925 + }, + { + "epoch": 5.44, + "learning_rate": 1.3686203130688024e-06, + "loss": 0.0239, + "step": 29926 + }, + { + "epoch": 5.44, + "learning_rate": 1.3685657080451401e-06, + "loss": 0.0412, + "step": 29927 + }, + { + "epoch": 5.44, + "learning_rate": 1.368511103021478e-06, + "loss": 2.7684, + "step": 29928 + }, + { + "epoch": 5.44, + "learning_rate": 1.3684564979978158e-06, + "loss": 1.6816, + "step": 29929 + }, + { + "epoch": 5.44, + "learning_rate": 1.3684018929741537e-06, + "loss": 1.6813, + "step": 29930 + }, + { + "epoch": 5.44, + "learning_rate": 1.3683472879504915e-06, + "loss": 0.7064, + "step": 29931 + }, + { + "epoch": 5.44, + "learning_rate": 1.3682926829268294e-06, + "loss": 3.774, + "step": 29932 + }, + { + "epoch": 5.44, + "learning_rate": 1.3682380779031671e-06, + "loss": 1.9114, + "step": 29933 + }, + { + "epoch": 5.44, + "learning_rate": 1.3681834728795049e-06, + "loss": 0.0091, + "step": 29934 + }, + { + "epoch": 5.44, + "learning_rate": 1.3681288678558428e-06, + "loss": 0.0087, + "step": 29935 + }, + { + "epoch": 5.44, + "learning_rate": 1.3680742628321805e-06, + "loss": 1.2595, + "step": 29936 + }, + { + "epoch": 5.44, + "learning_rate": 1.3680196578085185e-06, + "loss": 0.0108, + "step": 29937 + }, + { + "epoch": 5.44, + "learning_rate": 1.3679650527848562e-06, + "loss": 0.1169, + "step": 29938 + }, + { + "epoch": 5.44, + "learning_rate": 1.3679104477611942e-06, + "loss": 0.0797, + "step": 29939 + }, + { + "epoch": 5.44, + "learning_rate": 1.3678558427375319e-06, + "loss": 1.0169, + "step": 29940 + }, + { + "epoch": 5.44, + "learning_rate": 1.3678012377138696e-06, + "loss": 0.818, + "step": 29941 + }, + { + "epoch": 5.44, + "learning_rate": 1.3677466326902076e-06, + "loss": 1.6751, + "step": 29942 + }, + { + "epoch": 5.45, + "learning_rate": 1.3676920276665453e-06, + "loss": 0.2755, + "step": 29943 + }, + { + "epoch": 5.45, + "learning_rate": 1.3676374226428832e-06, + "loss": 0.2917, + "step": 29944 + }, + { + "epoch": 5.45, + "learning_rate": 1.367582817619221e-06, + "loss": 0.0036, + "step": 29945 + }, + { + "epoch": 5.45, + "learning_rate": 1.367528212595559e-06, + "loss": 0.1672, + "step": 29946 + }, + { + "epoch": 5.45, + "learning_rate": 1.3674736075718966e-06, + "loss": 0.434, + "step": 29947 + }, + { + "epoch": 5.45, + "learning_rate": 1.3674190025482344e-06, + "loss": 0.0711, + "step": 29948 + }, + { + "epoch": 5.45, + "learning_rate": 1.3673643975245723e-06, + "loss": 1.973, + "step": 29949 + }, + { + "epoch": 5.45, + "learning_rate": 1.36730979250091e-06, + "loss": 0.2078, + "step": 29950 + }, + { + "epoch": 5.45, + "learning_rate": 1.367255187477248e-06, + "loss": 0.6158, + "step": 29951 + }, + { + "epoch": 5.45, + "learning_rate": 1.3672005824535857e-06, + "loss": 1.1694, + "step": 29952 + }, + { + "epoch": 5.45, + "learning_rate": 1.3671459774299237e-06, + "loss": 0.7815, + "step": 29953 + }, + { + "epoch": 5.45, + "learning_rate": 1.3670913724062614e-06, + "loss": 0.2355, + "step": 29954 + }, + { + "epoch": 5.45, + "learning_rate": 1.3670367673825991e-06, + "loss": 1.2168, + "step": 29955 + }, + { + "epoch": 5.45, + "learning_rate": 1.366982162358937e-06, + "loss": 0.0075, + "step": 29956 + }, + { + "epoch": 5.45, + "learning_rate": 1.3669275573352748e-06, + "loss": 1.2091, + "step": 29957 + }, + { + "epoch": 5.45, + "learning_rate": 1.3668729523116128e-06, + "loss": 0.0935, + "step": 29958 + }, + { + "epoch": 5.45, + "learning_rate": 1.3668183472879505e-06, + "loss": 0.9705, + "step": 29959 + }, + { + "epoch": 5.45, + "learning_rate": 1.3667637422642884e-06, + "loss": 0.3415, + "step": 29960 + }, + { + "epoch": 5.45, + "learning_rate": 1.3667091372406262e-06, + "loss": 1.6702, + "step": 29961 + }, + { + "epoch": 5.45, + "learning_rate": 1.3666545322169639e-06, + "loss": 0.0157, + "step": 29962 + }, + { + "epoch": 5.45, + "learning_rate": 1.3665999271933018e-06, + "loss": 1.3972, + "step": 29963 + }, + { + "epoch": 5.45, + "learning_rate": 1.3665453221696398e-06, + "loss": 0.9323, + "step": 29964 + }, + { + "epoch": 5.45, + "learning_rate": 1.3664907171459775e-06, + "loss": 0.1452, + "step": 29965 + }, + { + "epoch": 5.45, + "learning_rate": 1.3664361121223152e-06, + "loss": 0.0458, + "step": 29966 + }, + { + "epoch": 5.45, + "learning_rate": 1.3663815070986532e-06, + "loss": 0.1897, + "step": 29967 + }, + { + "epoch": 5.45, + "learning_rate": 1.366326902074991e-06, + "loss": 1.7522, + "step": 29968 + }, + { + "epoch": 5.45, + "learning_rate": 1.3662722970513286e-06, + "loss": 0.4504, + "step": 29969 + }, + { + "epoch": 5.45, + "learning_rate": 1.3662176920276666e-06, + "loss": 0.0187, + "step": 29970 + }, + { + "epoch": 5.45, + "learning_rate": 1.3661630870040045e-06, + "loss": 0.3343, + "step": 29971 + }, + { + "epoch": 5.45, + "learning_rate": 1.3661084819803423e-06, + "loss": 0.2672, + "step": 29972 + }, + { + "epoch": 5.45, + "learning_rate": 1.36605387695668e-06, + "loss": 1.3072, + "step": 29973 + }, + { + "epoch": 5.45, + "learning_rate": 1.365999271933018e-06, + "loss": 1.0652, + "step": 29974 + }, + { + "epoch": 5.45, + "learning_rate": 1.3659446669093557e-06, + "loss": 2.3506, + "step": 29975 + }, + { + "epoch": 5.45, + "learning_rate": 1.3658900618856934e-06, + "loss": 0.6485, + "step": 29976 + }, + { + "epoch": 5.45, + "learning_rate": 1.3658354568620313e-06, + "loss": 0.0096, + "step": 29977 + }, + { + "epoch": 5.45, + "learning_rate": 1.3657808518383693e-06, + "loss": 0.155, + "step": 29978 + }, + { + "epoch": 5.45, + "learning_rate": 1.365726246814707e-06, + "loss": 0.3096, + "step": 29979 + }, + { + "epoch": 5.45, + "learning_rate": 1.3656716417910448e-06, + "loss": 1.4365, + "step": 29980 + }, + { + "epoch": 5.45, + "learning_rate": 1.3656170367673827e-06, + "loss": 0.1383, + "step": 29981 + }, + { + "epoch": 5.45, + "learning_rate": 1.3655624317437204e-06, + "loss": 1.3551, + "step": 29982 + }, + { + "epoch": 5.45, + "learning_rate": 1.3655078267200582e-06, + "loss": 0.9596, + "step": 29983 + }, + { + "epoch": 5.45, + "learning_rate": 1.365453221696396e-06, + "loss": 0.1254, + "step": 29984 + }, + { + "epoch": 5.45, + "learning_rate": 1.365398616672734e-06, + "loss": 0.2128, + "step": 29985 + }, + { + "epoch": 5.45, + "learning_rate": 1.3653440116490718e-06, + "loss": 0.1383, + "step": 29986 + }, + { + "epoch": 5.45, + "learning_rate": 1.3652894066254095e-06, + "loss": 0.5619, + "step": 29987 + }, + { + "epoch": 5.45, + "learning_rate": 1.3652348016017475e-06, + "loss": 0.6785, + "step": 29988 + }, + { + "epoch": 5.45, + "learning_rate": 1.3651801965780852e-06, + "loss": 1.3965, + "step": 29989 + }, + { + "epoch": 5.45, + "learning_rate": 1.365125591554423e-06, + "loss": 0.2766, + "step": 29990 + }, + { + "epoch": 5.45, + "learning_rate": 1.3650709865307609e-06, + "loss": 0.1817, + "step": 29991 + }, + { + "epoch": 5.45, + "learning_rate": 1.3650163815070988e-06, + "loss": 0.0194, + "step": 29992 + }, + { + "epoch": 5.45, + "learning_rate": 1.3649617764834365e-06, + "loss": 0.4183, + "step": 29993 + }, + { + "epoch": 5.45, + "learning_rate": 1.3649071714597743e-06, + "loss": 0.0958, + "step": 29994 + }, + { + "epoch": 5.45, + "learning_rate": 1.3648525664361122e-06, + "loss": 2.0729, + "step": 29995 + }, + { + "epoch": 5.45, + "learning_rate": 1.36479796141245e-06, + "loss": 0.3713, + "step": 29996 + }, + { + "epoch": 5.45, + "learning_rate": 1.3647433563887877e-06, + "loss": 0.1549, + "step": 29997 + }, + { + "epoch": 5.46, + "learning_rate": 1.3646887513651256e-06, + "loss": 0.1411, + "step": 29998 + }, + { + "epoch": 5.46, + "learning_rate": 1.3646341463414636e-06, + "loss": 0.3822, + "step": 29999 + }, + { + "epoch": 5.46, + "learning_rate": 1.3645795413178013e-06, + "loss": 0.0256, + "step": 30000 + }, + { + "epoch": 5.46, + "learning_rate": 1.364524936294139e-06, + "loss": 0.026, + "step": 30001 + }, + { + "epoch": 5.46, + "learning_rate": 1.364470331270477e-06, + "loss": 0.8899, + "step": 30002 + }, + { + "epoch": 5.46, + "learning_rate": 1.3644157262468147e-06, + "loss": 0.9382, + "step": 30003 + }, + { + "epoch": 5.46, + "learning_rate": 1.3643611212231526e-06, + "loss": 1.6308, + "step": 30004 + }, + { + "epoch": 5.46, + "learning_rate": 1.3643065161994904e-06, + "loss": 0.4106, + "step": 30005 + }, + { + "epoch": 5.46, + "learning_rate": 1.3642519111758283e-06, + "loss": 0.0304, + "step": 30006 + }, + { + "epoch": 5.46, + "learning_rate": 1.364197306152166e-06, + "loss": 0.2686, + "step": 30007 + }, + { + "epoch": 5.46, + "learning_rate": 1.3641427011285038e-06, + "loss": 0.4764, + "step": 30008 + }, + { + "epoch": 5.46, + "learning_rate": 1.3640880961048417e-06, + "loss": 2.9322, + "step": 30009 + }, + { + "epoch": 5.46, + "learning_rate": 1.3640334910811795e-06, + "loss": 0.9536, + "step": 30010 + }, + { + "epoch": 5.46, + "learning_rate": 1.3639788860575174e-06, + "loss": 0.7224, + "step": 30011 + }, + { + "epoch": 5.46, + "learning_rate": 1.3639242810338551e-06, + "loss": 0.7367, + "step": 30012 + }, + { + "epoch": 5.46, + "learning_rate": 1.363869676010193e-06, + "loss": 0.1109, + "step": 30013 + }, + { + "epoch": 5.46, + "learning_rate": 1.3638150709865308e-06, + "loss": 0.048, + "step": 30014 + }, + { + "epoch": 5.46, + "learning_rate": 1.3637604659628685e-06, + "loss": 0.7623, + "step": 30015 + }, + { + "epoch": 5.46, + "learning_rate": 1.3637058609392065e-06, + "loss": 1.0545, + "step": 30016 + }, + { + "epoch": 5.46, + "learning_rate": 1.3636512559155442e-06, + "loss": 0.1579, + "step": 30017 + }, + { + "epoch": 5.46, + "learning_rate": 1.3635966508918822e-06, + "loss": 0.2796, + "step": 30018 + }, + { + "epoch": 5.46, + "learning_rate": 1.3635420458682199e-06, + "loss": 0.0853, + "step": 30019 + }, + { + "epoch": 5.46, + "learning_rate": 1.3634874408445578e-06, + "loss": 0.0165, + "step": 30020 + }, + { + "epoch": 5.46, + "learning_rate": 1.3634328358208956e-06, + "loss": 0.1457, + "step": 30021 + }, + { + "epoch": 5.46, + "learning_rate": 1.3633782307972333e-06, + "loss": 0.0525, + "step": 30022 + }, + { + "epoch": 5.46, + "learning_rate": 1.3633236257735712e-06, + "loss": 0.2079, + "step": 30023 + }, + { + "epoch": 5.46, + "learning_rate": 1.363269020749909e-06, + "loss": 1.1132, + "step": 30024 + }, + { + "epoch": 5.46, + "learning_rate": 1.363214415726247e-06, + "loss": 0.4776, + "step": 30025 + }, + { + "epoch": 5.46, + "learning_rate": 1.3631598107025846e-06, + "loss": 0.0157, + "step": 30026 + }, + { + "epoch": 5.46, + "learning_rate": 1.3631052056789226e-06, + "loss": 0.0211, + "step": 30027 + }, + { + "epoch": 5.46, + "learning_rate": 1.3630506006552603e-06, + "loss": 0.1812, + "step": 30028 + }, + { + "epoch": 5.46, + "learning_rate": 1.362995995631598e-06, + "loss": 0.1152, + "step": 30029 + }, + { + "epoch": 5.46, + "learning_rate": 1.362941390607936e-06, + "loss": 0.1787, + "step": 30030 + }, + { + "epoch": 5.46, + "learning_rate": 1.3628867855842737e-06, + "loss": 1.3108, + "step": 30031 + }, + { + "epoch": 5.46, + "learning_rate": 1.3628321805606117e-06, + "loss": 1.1959, + "step": 30032 + }, + { + "epoch": 5.46, + "learning_rate": 1.3627775755369494e-06, + "loss": 3.2506, + "step": 30033 + }, + { + "epoch": 5.46, + "learning_rate": 1.3627229705132873e-06, + "loss": 0.6512, + "step": 30034 + }, + { + "epoch": 5.46, + "learning_rate": 1.362668365489625e-06, + "loss": 0.0092, + "step": 30035 + }, + { + "epoch": 5.46, + "learning_rate": 1.3626137604659628e-06, + "loss": 1.2566, + "step": 30036 + }, + { + "epoch": 5.46, + "learning_rate": 1.3625591554423008e-06, + "loss": 0.4133, + "step": 30037 + }, + { + "epoch": 5.46, + "learning_rate": 1.3625045504186385e-06, + "loss": 0.2106, + "step": 30038 + }, + { + "epoch": 5.46, + "learning_rate": 1.3624499453949764e-06, + "loss": 0.1972, + "step": 30039 + }, + { + "epoch": 5.46, + "learning_rate": 1.3623953403713142e-06, + "loss": 1.0759, + "step": 30040 + }, + { + "epoch": 5.46, + "learning_rate": 1.362340735347652e-06, + "loss": 0.6766, + "step": 30041 + }, + { + "epoch": 5.46, + "learning_rate": 1.3622861303239898e-06, + "loss": 0.0617, + "step": 30042 + }, + { + "epoch": 5.46, + "learning_rate": 1.3622315253003276e-06, + "loss": 2.0872, + "step": 30043 + }, + { + "epoch": 5.46, + "learning_rate": 1.3621769202766655e-06, + "loss": 0.0085, + "step": 30044 + }, + { + "epoch": 5.46, + "learning_rate": 1.3621223152530035e-06, + "loss": 0.5773, + "step": 30045 + }, + { + "epoch": 5.46, + "learning_rate": 1.3620677102293412e-06, + "loss": 0.357, + "step": 30046 + }, + { + "epoch": 5.46, + "learning_rate": 1.362013105205679e-06, + "loss": 0.3301, + "step": 30047 + }, + { + "epoch": 5.46, + "learning_rate": 1.3619585001820169e-06, + "loss": 0.392, + "step": 30048 + }, + { + "epoch": 5.46, + "learning_rate": 1.3619038951583546e-06, + "loss": 0.9358, + "step": 30049 + }, + { + "epoch": 5.46, + "learning_rate": 1.3618492901346923e-06, + "loss": 0.3702, + "step": 30050 + }, + { + "epoch": 5.46, + "learning_rate": 1.3617946851110303e-06, + "loss": 0.0352, + "step": 30051 + }, + { + "epoch": 5.46, + "learning_rate": 1.3617400800873682e-06, + "loss": 0.7715, + "step": 30052 + }, + { + "epoch": 5.47, + "learning_rate": 1.361685475063706e-06, + "loss": 0.9907, + "step": 30053 + }, + { + "epoch": 5.47, + "learning_rate": 1.3616308700400437e-06, + "loss": 1.2698, + "step": 30054 + }, + { + "epoch": 5.47, + "learning_rate": 1.3615762650163816e-06, + "loss": 0.022, + "step": 30055 + }, + { + "epoch": 5.47, + "learning_rate": 1.3615216599927193e-06, + "loss": 1.1913, + "step": 30056 + }, + { + "epoch": 5.47, + "learning_rate": 1.361467054969057e-06, + "loss": 1.9564, + "step": 30057 + }, + { + "epoch": 5.47, + "learning_rate": 1.361412449945395e-06, + "loss": 2.7412, + "step": 30058 + }, + { + "epoch": 5.47, + "learning_rate": 1.361357844921733e-06, + "loss": 0.024, + "step": 30059 + }, + { + "epoch": 5.47, + "learning_rate": 1.3613032398980707e-06, + "loss": 1.4085, + "step": 30060 + }, + { + "epoch": 5.47, + "learning_rate": 1.3612486348744084e-06, + "loss": 0.1069, + "step": 30061 + }, + { + "epoch": 5.47, + "learning_rate": 1.3611940298507464e-06, + "loss": 0.4907, + "step": 30062 + }, + { + "epoch": 5.47, + "learning_rate": 1.361139424827084e-06, + "loss": 0.1351, + "step": 30063 + }, + { + "epoch": 5.47, + "learning_rate": 1.3610848198034218e-06, + "loss": 0.0451, + "step": 30064 + }, + { + "epoch": 5.47, + "learning_rate": 1.3610302147797598e-06, + "loss": 0.7372, + "step": 30065 + }, + { + "epoch": 5.47, + "learning_rate": 1.3609756097560977e-06, + "loss": 0.8939, + "step": 30066 + }, + { + "epoch": 5.47, + "learning_rate": 1.3609210047324355e-06, + "loss": 0.1863, + "step": 30067 + }, + { + "epoch": 5.47, + "learning_rate": 1.3608663997087732e-06, + "loss": 0.4459, + "step": 30068 + }, + { + "epoch": 5.47, + "learning_rate": 1.3608117946851111e-06, + "loss": 0.0432, + "step": 30069 + }, + { + "epoch": 5.47, + "learning_rate": 1.3607571896614489e-06, + "loss": 0.7041, + "step": 30070 + }, + { + "epoch": 5.47, + "learning_rate": 1.3607025846377866e-06, + "loss": 1.9129, + "step": 30071 + }, + { + "epoch": 5.47, + "learning_rate": 1.3606479796141245e-06, + "loss": 0.7701, + "step": 30072 + }, + { + "epoch": 5.47, + "learning_rate": 1.3605933745904625e-06, + "loss": 0.382, + "step": 30073 + }, + { + "epoch": 5.47, + "learning_rate": 1.3605387695668002e-06, + "loss": 0.9643, + "step": 30074 + }, + { + "epoch": 5.47, + "learning_rate": 1.360484164543138e-06, + "loss": 0.0987, + "step": 30075 + }, + { + "epoch": 5.47, + "learning_rate": 1.3604295595194759e-06, + "loss": 0.0099, + "step": 30076 + }, + { + "epoch": 5.47, + "learning_rate": 1.3603749544958136e-06, + "loss": 1.8204, + "step": 30077 + }, + { + "epoch": 5.47, + "learning_rate": 1.3603203494721513e-06, + "loss": 0.0507, + "step": 30078 + }, + { + "epoch": 5.47, + "learning_rate": 1.3602657444484893e-06, + "loss": 0.1277, + "step": 30079 + }, + { + "epoch": 5.47, + "learning_rate": 1.3602111394248272e-06, + "loss": 0.6271, + "step": 30080 + }, + { + "epoch": 5.47, + "learning_rate": 1.360156534401165e-06, + "loss": 0.6485, + "step": 30081 + }, + { + "epoch": 5.47, + "learning_rate": 1.3601019293775027e-06, + "loss": 1.0911, + "step": 30082 + }, + { + "epoch": 5.47, + "learning_rate": 1.3600473243538406e-06, + "loss": 0.5322, + "step": 30083 + }, + { + "epoch": 5.47, + "learning_rate": 1.3599927193301784e-06, + "loss": 0.1982, + "step": 30084 + }, + { + "epoch": 5.47, + "learning_rate": 1.3599381143065163e-06, + "loss": 1.2313, + "step": 30085 + }, + { + "epoch": 5.47, + "learning_rate": 1.359883509282854e-06, + "loss": 0.1233, + "step": 30086 + }, + { + "epoch": 5.47, + "learning_rate": 1.359828904259192e-06, + "loss": 0.4867, + "step": 30087 + }, + { + "epoch": 5.47, + "learning_rate": 1.3597742992355297e-06, + "loss": 0.1066, + "step": 30088 + }, + { + "epoch": 5.47, + "learning_rate": 1.3597196942118675e-06, + "loss": 0.2359, + "step": 30089 + }, + { + "epoch": 5.47, + "learning_rate": 1.3596650891882054e-06, + "loss": 0.0875, + "step": 30090 + }, + { + "epoch": 5.47, + "learning_rate": 1.3596104841645431e-06, + "loss": 1.539, + "step": 30091 + }, + { + "epoch": 5.47, + "learning_rate": 1.359555879140881e-06, + "loss": 0.4687, + "step": 30092 + }, + { + "epoch": 5.47, + "learning_rate": 1.3595012741172188e-06, + "loss": 1.3486, + "step": 30093 + }, + { + "epoch": 5.47, + "learning_rate": 1.3594466690935567e-06, + "loss": 0.1526, + "step": 30094 + }, + { + "epoch": 5.47, + "learning_rate": 1.3593920640698945e-06, + "loss": 0.1006, + "step": 30095 + }, + { + "epoch": 5.47, + "learning_rate": 1.3593374590462322e-06, + "loss": 0.0057, + "step": 30096 + }, + { + "epoch": 5.47, + "learning_rate": 1.3592828540225702e-06, + "loss": 0.1326, + "step": 30097 + }, + { + "epoch": 5.47, + "learning_rate": 1.3592282489989079e-06, + "loss": 0.0205, + "step": 30098 + }, + { + "epoch": 5.47, + "learning_rate": 1.3591736439752458e-06, + "loss": 2.6935, + "step": 30099 + }, + { + "epoch": 5.47, + "learning_rate": 1.3591190389515836e-06, + "loss": 0.3845, + "step": 30100 + }, + { + "epoch": 5.47, + "learning_rate": 1.3590644339279215e-06, + "loss": 0.0259, + "step": 30101 + }, + { + "epoch": 5.47, + "learning_rate": 1.3590098289042592e-06, + "loss": 0.5472, + "step": 30102 + }, + { + "epoch": 5.47, + "learning_rate": 1.358955223880597e-06, + "loss": 1.2497, + "step": 30103 + }, + { + "epoch": 5.47, + "learning_rate": 1.358900618856935e-06, + "loss": 0.0212, + "step": 30104 + }, + { + "epoch": 5.47, + "learning_rate": 1.3588460138332726e-06, + "loss": 0.5923, + "step": 30105 + }, + { + "epoch": 5.47, + "learning_rate": 1.3587914088096106e-06, + "loss": 0.3645, + "step": 30106 + }, + { + "epoch": 5.47, + "learning_rate": 1.3587368037859483e-06, + "loss": 0.1053, + "step": 30107 + }, + { + "epoch": 5.48, + "learning_rate": 1.3586821987622863e-06, + "loss": 0.1588, + "step": 30108 + }, + { + "epoch": 5.48, + "learning_rate": 1.358627593738624e-06, + "loss": 0.0129, + "step": 30109 + }, + { + "epoch": 5.48, + "learning_rate": 1.3585729887149617e-06, + "loss": 0.7306, + "step": 30110 + }, + { + "epoch": 5.48, + "learning_rate": 1.3585183836912997e-06, + "loss": 0.2267, + "step": 30111 + }, + { + "epoch": 5.48, + "learning_rate": 1.3584637786676374e-06, + "loss": 0.5167, + "step": 30112 + }, + { + "epoch": 5.48, + "learning_rate": 1.3584091736439753e-06, + "loss": 0.0087, + "step": 30113 + }, + { + "epoch": 5.48, + "learning_rate": 1.358354568620313e-06, + "loss": 1.7081, + "step": 30114 + }, + { + "epoch": 5.48, + "learning_rate": 1.358299963596651e-06, + "loss": 0.1088, + "step": 30115 + }, + { + "epoch": 5.48, + "learning_rate": 1.3582453585729888e-06, + "loss": 0.9238, + "step": 30116 + }, + { + "epoch": 5.48, + "learning_rate": 1.3581907535493265e-06, + "loss": 0.0185, + "step": 30117 + }, + { + "epoch": 5.48, + "learning_rate": 1.3581361485256644e-06, + "loss": 1.6442, + "step": 30118 + }, + { + "epoch": 5.48, + "learning_rate": 1.3580815435020022e-06, + "loss": 0.0087, + "step": 30119 + }, + { + "epoch": 5.48, + "learning_rate": 1.35802693847834e-06, + "loss": 1.9192, + "step": 30120 + }, + { + "epoch": 5.48, + "learning_rate": 1.3579723334546778e-06, + "loss": 0.3885, + "step": 30121 + }, + { + "epoch": 5.48, + "learning_rate": 1.3579177284310158e-06, + "loss": 1.6606, + "step": 30122 + }, + { + "epoch": 5.48, + "learning_rate": 1.3578631234073535e-06, + "loss": 0.9129, + "step": 30123 + }, + { + "epoch": 5.48, + "learning_rate": 1.3578085183836912e-06, + "loss": 1.5336, + "step": 30124 + }, + { + "epoch": 5.48, + "learning_rate": 1.357753913360029e-06, + "loss": 1.6136, + "step": 30125 + }, + { + "epoch": 5.48, + "learning_rate": 1.357699308336367e-06, + "loss": 0.1975, + "step": 30126 + }, + { + "epoch": 5.48, + "learning_rate": 1.3576447033127049e-06, + "loss": 0.1704, + "step": 30127 + }, + { + "epoch": 5.48, + "learning_rate": 1.3575900982890426e-06, + "loss": 0.3214, + "step": 30128 + }, + { + "epoch": 5.48, + "learning_rate": 1.3575354932653805e-06, + "loss": 2.0697, + "step": 30129 + }, + { + "epoch": 5.48, + "learning_rate": 1.3574808882417183e-06, + "loss": 0.5062, + "step": 30130 + }, + { + "epoch": 5.48, + "learning_rate": 1.357426283218056e-06, + "loss": 0.8964, + "step": 30131 + }, + { + "epoch": 5.48, + "learning_rate": 1.357371678194394e-06, + "loss": 0.0239, + "step": 30132 + }, + { + "epoch": 5.48, + "learning_rate": 1.3573170731707319e-06, + "loss": 0.4766, + "step": 30133 + }, + { + "epoch": 5.48, + "learning_rate": 1.3572624681470696e-06, + "loss": 3.4103, + "step": 30134 + }, + { + "epoch": 5.48, + "learning_rate": 1.3572078631234073e-06, + "loss": 3.137, + "step": 30135 + }, + { + "epoch": 5.48, + "learning_rate": 1.3571532580997453e-06, + "loss": 0.026, + "step": 30136 + }, + { + "epoch": 5.48, + "learning_rate": 1.357098653076083e-06, + "loss": 1.5497, + "step": 30137 + }, + { + "epoch": 5.48, + "learning_rate": 1.3570440480524208e-06, + "loss": 0.1736, + "step": 30138 + }, + { + "epoch": 5.48, + "learning_rate": 1.3569894430287587e-06, + "loss": 0.9137, + "step": 30139 + }, + { + "epoch": 5.48, + "learning_rate": 1.3569348380050966e-06, + "loss": 0.5949, + "step": 30140 + }, + { + "epoch": 5.48, + "learning_rate": 1.3568802329814344e-06, + "loss": 0.1987, + "step": 30141 + }, + { + "epoch": 5.48, + "learning_rate": 1.356825627957772e-06, + "loss": 0.1038, + "step": 30142 + }, + { + "epoch": 5.48, + "learning_rate": 1.35677102293411e-06, + "loss": 0.8167, + "step": 30143 + }, + { + "epoch": 5.48, + "learning_rate": 1.3567164179104478e-06, + "loss": 0.749, + "step": 30144 + }, + { + "epoch": 5.48, + "learning_rate": 1.3566618128867855e-06, + "loss": 0.295, + "step": 30145 + }, + { + "epoch": 5.48, + "learning_rate": 1.3566072078631235e-06, + "loss": 1.4919, + "step": 30146 + }, + { + "epoch": 5.48, + "learning_rate": 1.3565526028394614e-06, + "loss": 2.1215, + "step": 30147 + }, + { + "epoch": 5.48, + "learning_rate": 1.3564979978157991e-06, + "loss": 0.0526, + "step": 30148 + }, + { + "epoch": 5.48, + "learning_rate": 1.3564433927921369e-06, + "loss": 0.8794, + "step": 30149 + }, + { + "epoch": 5.48, + "learning_rate": 1.3563887877684748e-06, + "loss": 2.7182, + "step": 30150 + }, + { + "epoch": 5.48, + "learning_rate": 1.3563341827448125e-06, + "loss": 0.0804, + "step": 30151 + }, + { + "epoch": 5.48, + "learning_rate": 1.3562795777211503e-06, + "loss": 0.0294, + "step": 30152 + }, + { + "epoch": 5.48, + "learning_rate": 1.3562249726974882e-06, + "loss": 0.5631, + "step": 30153 + }, + { + "epoch": 5.48, + "learning_rate": 1.3561703676738262e-06, + "loss": 0.298, + "step": 30154 + }, + { + "epoch": 5.48, + "learning_rate": 1.3561157626501639e-06, + "loss": 2.0782, + "step": 30155 + }, + { + "epoch": 5.48, + "learning_rate": 1.3560611576265016e-06, + "loss": 0.9257, + "step": 30156 + }, + { + "epoch": 5.48, + "learning_rate": 1.3560065526028396e-06, + "loss": 1.1216, + "step": 30157 + }, + { + "epoch": 5.48, + "learning_rate": 1.3559519475791773e-06, + "loss": 0.2026, + "step": 30158 + }, + { + "epoch": 5.48, + "learning_rate": 1.355897342555515e-06, + "loss": 1.458, + "step": 30159 + }, + { + "epoch": 5.48, + "learning_rate": 1.355842737531853e-06, + "loss": 0.0041, + "step": 30160 + }, + { + "epoch": 5.48, + "learning_rate": 1.355788132508191e-06, + "loss": 0.7707, + "step": 30161 + }, + { + "epoch": 5.48, + "learning_rate": 1.3557335274845286e-06, + "loss": 0.2308, + "step": 30162 + }, + { + "epoch": 5.49, + "learning_rate": 1.3556789224608664e-06, + "loss": 1.9117, + "step": 30163 + }, + { + "epoch": 5.49, + "learning_rate": 1.3556243174372043e-06, + "loss": 1.3682, + "step": 30164 + }, + { + "epoch": 5.49, + "learning_rate": 1.355569712413542e-06, + "loss": 0.3457, + "step": 30165 + }, + { + "epoch": 5.49, + "learning_rate": 1.3555151073898798e-06, + "loss": 1.8024, + "step": 30166 + }, + { + "epoch": 5.49, + "learning_rate": 1.3554605023662177e-06, + "loss": 0.6317, + "step": 30167 + }, + { + "epoch": 5.49, + "learning_rate": 1.3554058973425557e-06, + "loss": 0.1495, + "step": 30168 + }, + { + "epoch": 5.49, + "learning_rate": 1.3553512923188934e-06, + "loss": 0.1504, + "step": 30169 + }, + { + "epoch": 5.49, + "learning_rate": 1.3552966872952311e-06, + "loss": 0.1347, + "step": 30170 + }, + { + "epoch": 5.49, + "learning_rate": 1.355242082271569e-06, + "loss": 0.2041, + "step": 30171 + }, + { + "epoch": 5.49, + "learning_rate": 1.3551874772479068e-06, + "loss": 1.6485, + "step": 30172 + }, + { + "epoch": 5.49, + "learning_rate": 1.3551328722242447e-06, + "loss": 0.321, + "step": 30173 + }, + { + "epoch": 5.49, + "learning_rate": 1.3550782672005825e-06, + "loss": 0.9237, + "step": 30174 + }, + { + "epoch": 5.49, + "learning_rate": 1.3550236621769204e-06, + "loss": 0.4191, + "step": 30175 + }, + { + "epoch": 5.49, + "learning_rate": 1.3549690571532582e-06, + "loss": 0.865, + "step": 30176 + }, + { + "epoch": 5.49, + "learning_rate": 1.3549144521295959e-06, + "loss": 0.4014, + "step": 30177 + }, + { + "epoch": 5.49, + "learning_rate": 1.3548598471059338e-06, + "loss": 0.0308, + "step": 30178 + }, + { + "epoch": 5.49, + "learning_rate": 1.3548052420822716e-06, + "loss": 0.0787, + "step": 30179 + }, + { + "epoch": 5.49, + "learning_rate": 1.3547506370586095e-06, + "loss": 1.4382, + "step": 30180 + }, + { + "epoch": 5.49, + "learning_rate": 1.3546960320349472e-06, + "loss": 0.814, + "step": 30181 + }, + { + "epoch": 5.49, + "learning_rate": 1.3546414270112852e-06, + "loss": 0.7916, + "step": 30182 + }, + { + "epoch": 5.49, + "learning_rate": 1.354586821987623e-06, + "loss": 2.3664, + "step": 30183 + }, + { + "epoch": 5.49, + "learning_rate": 1.3545322169639606e-06, + "loss": 0.3325, + "step": 30184 + }, + { + "epoch": 5.49, + "learning_rate": 1.3544776119402986e-06, + "loss": 0.43, + "step": 30185 + }, + { + "epoch": 5.49, + "learning_rate": 1.3544230069166363e-06, + "loss": 1.2042, + "step": 30186 + }, + { + "epoch": 5.49, + "learning_rate": 1.3543684018929743e-06, + "loss": 0.6542, + "step": 30187 + }, + { + "epoch": 5.49, + "learning_rate": 1.354313796869312e-06, + "loss": 0.334, + "step": 30188 + }, + { + "epoch": 5.49, + "learning_rate": 1.35425919184565e-06, + "loss": 0.803, + "step": 30189 + }, + { + "epoch": 5.49, + "learning_rate": 1.3542045868219877e-06, + "loss": 0.1548, + "step": 30190 + }, + { + "epoch": 5.49, + "learning_rate": 1.3541499817983254e-06, + "loss": 0.0225, + "step": 30191 + }, + { + "epoch": 5.49, + "learning_rate": 1.3540953767746633e-06, + "loss": 0.3723, + "step": 30192 + }, + { + "epoch": 5.49, + "learning_rate": 1.354040771751001e-06, + "loss": 0.9814, + "step": 30193 + }, + { + "epoch": 5.49, + "learning_rate": 1.353986166727339e-06, + "loss": 3.2021, + "step": 30194 + }, + { + "epoch": 5.49, + "learning_rate": 1.3539315617036767e-06, + "loss": 0.0105, + "step": 30195 + }, + { + "epoch": 5.49, + "learning_rate": 1.3538769566800147e-06, + "loss": 0.3226, + "step": 30196 + }, + { + "epoch": 5.49, + "learning_rate": 1.3538223516563524e-06, + "loss": 0.8927, + "step": 30197 + }, + { + "epoch": 5.49, + "learning_rate": 1.3537677466326902e-06, + "loss": 0.1717, + "step": 30198 + }, + { + "epoch": 5.49, + "learning_rate": 1.3537131416090279e-06, + "loss": 0.2002, + "step": 30199 + }, + { + "epoch": 5.49, + "learning_rate": 1.3536585365853658e-06, + "loss": 1.5152, + "step": 30200 + }, + { + "epoch": 5.49, + "learning_rate": 1.3536039315617038e-06, + "loss": 0.1535, + "step": 30201 + }, + { + "epoch": 5.49, + "learning_rate": 1.3535493265380415e-06, + "loss": 0.0359, + "step": 30202 + }, + { + "epoch": 5.49, + "learning_rate": 1.3534947215143794e-06, + "loss": 0.0043, + "step": 30203 + }, + { + "epoch": 5.49, + "learning_rate": 1.3534401164907172e-06, + "loss": 0.5869, + "step": 30204 + }, + { + "epoch": 5.49, + "learning_rate": 1.353385511467055e-06, + "loss": 0.1928, + "step": 30205 + }, + { + "epoch": 5.49, + "learning_rate": 1.3533309064433926e-06, + "loss": 0.624, + "step": 30206 + }, + { + "epoch": 5.49, + "learning_rate": 1.3532763014197306e-06, + "loss": 0.0173, + "step": 30207 + }, + { + "epoch": 5.49, + "learning_rate": 1.3532216963960685e-06, + "loss": 0.3799, + "step": 30208 + }, + { + "epoch": 5.49, + "learning_rate": 1.3531670913724063e-06, + "loss": 0.264, + "step": 30209 + }, + { + "epoch": 5.49, + "learning_rate": 1.3531124863487442e-06, + "loss": 0.4247, + "step": 30210 + }, + { + "epoch": 5.49, + "learning_rate": 1.353057881325082e-06, + "loss": 1.594, + "step": 30211 + }, + { + "epoch": 5.49, + "learning_rate": 1.3530032763014197e-06, + "loss": 1.7879, + "step": 30212 + }, + { + "epoch": 5.49, + "learning_rate": 1.3529486712777576e-06, + "loss": 0.1363, + "step": 30213 + }, + { + "epoch": 5.49, + "learning_rate": 1.3528940662540956e-06, + "loss": 0.1121, + "step": 30214 + }, + { + "epoch": 5.49, + "learning_rate": 1.3528394612304333e-06, + "loss": 0.0871, + "step": 30215 + }, + { + "epoch": 5.49, + "learning_rate": 1.352784856206771e-06, + "loss": 1.4902, + "step": 30216 + }, + { + "epoch": 5.49, + "learning_rate": 1.352730251183109e-06, + "loss": 0.2755, + "step": 30217 + }, + { + "epoch": 5.5, + "learning_rate": 1.3526756461594467e-06, + "loss": 0.5543, + "step": 30218 + }, + { + "epoch": 5.5, + "learning_rate": 1.3526210411357844e-06, + "loss": 0.2108, + "step": 30219 + }, + { + "epoch": 5.5, + "learning_rate": 1.3525664361121224e-06, + "loss": 0.0077, + "step": 30220 + }, + { + "epoch": 5.5, + "learning_rate": 1.3525118310884603e-06, + "loss": 0.2292, + "step": 30221 + }, + { + "epoch": 5.5, + "learning_rate": 1.352457226064798e-06, + "loss": 0.0049, + "step": 30222 + }, + { + "epoch": 5.5, + "learning_rate": 1.3524026210411358e-06, + "loss": 0.0101, + "step": 30223 + }, + { + "epoch": 5.5, + "learning_rate": 1.3523480160174737e-06, + "loss": 0.0112, + "step": 30224 + }, + { + "epoch": 5.5, + "learning_rate": 1.3522934109938115e-06, + "loss": 0.0067, + "step": 30225 + }, + { + "epoch": 5.5, + "learning_rate": 1.3522388059701492e-06, + "loss": 0.0067, + "step": 30226 + }, + { + "epoch": 5.5, + "learning_rate": 1.3521842009464871e-06, + "loss": 2.4317, + "step": 30227 + }, + { + "epoch": 5.5, + "learning_rate": 1.352129595922825e-06, + "loss": 1.403, + "step": 30228 + }, + { + "epoch": 5.5, + "learning_rate": 1.3520749908991628e-06, + "loss": 0.0084, + "step": 30229 + }, + { + "epoch": 5.5, + "learning_rate": 1.3520203858755005e-06, + "loss": 0.1137, + "step": 30230 + }, + { + "epoch": 5.5, + "learning_rate": 1.3519657808518385e-06, + "loss": 0.3605, + "step": 30231 + }, + { + "epoch": 5.5, + "learning_rate": 1.3519111758281762e-06, + "loss": 3.125, + "step": 30232 + }, + { + "epoch": 5.5, + "learning_rate": 1.351856570804514e-06, + "loss": 0.8891, + "step": 30233 + }, + { + "epoch": 5.5, + "learning_rate": 1.3518019657808519e-06, + "loss": 0.0058, + "step": 30234 + }, + { + "epoch": 5.5, + "learning_rate": 1.3517473607571898e-06, + "loss": 0.0195, + "step": 30235 + }, + { + "epoch": 5.5, + "learning_rate": 1.3516927557335276e-06, + "loss": 1.7299, + "step": 30236 + }, + { + "epoch": 5.5, + "learning_rate": 1.3516381507098653e-06, + "loss": 0.944, + "step": 30237 + }, + { + "epoch": 5.5, + "learning_rate": 1.3515835456862032e-06, + "loss": 1.2727, + "step": 30238 + }, + { + "epoch": 5.5, + "learning_rate": 1.351528940662541e-06, + "loss": 0.7282, + "step": 30239 + }, + { + "epoch": 5.5, + "learning_rate": 1.3514743356388787e-06, + "loss": 0.7428, + "step": 30240 + }, + { + "epoch": 5.5, + "learning_rate": 1.3514197306152166e-06, + "loss": 1.6821, + "step": 30241 + }, + { + "epoch": 5.5, + "learning_rate": 1.3513651255915546e-06, + "loss": 0.1851, + "step": 30242 + }, + { + "epoch": 5.5, + "learning_rate": 1.3513105205678923e-06, + "loss": 0.3582, + "step": 30243 + }, + { + "epoch": 5.5, + "learning_rate": 1.35125591554423e-06, + "loss": 0.0415, + "step": 30244 + }, + { + "epoch": 5.5, + "learning_rate": 1.351201310520568e-06, + "loss": 0.2905, + "step": 30245 + }, + { + "epoch": 5.5, + "learning_rate": 1.3511467054969057e-06, + "loss": 0.6246, + "step": 30246 + }, + { + "epoch": 5.5, + "learning_rate": 1.3510921004732435e-06, + "loss": 0.0933, + "step": 30247 + }, + { + "epoch": 5.5, + "learning_rate": 1.3510374954495814e-06, + "loss": 0.0225, + "step": 30248 + }, + { + "epoch": 5.5, + "learning_rate": 1.3509828904259193e-06, + "loss": 0.0688, + "step": 30249 + }, + { + "epoch": 5.5, + "learning_rate": 1.350928285402257e-06, + "loss": 0.6772, + "step": 30250 + }, + { + "epoch": 5.5, + "learning_rate": 1.3508736803785948e-06, + "loss": 0.0139, + "step": 30251 + }, + { + "epoch": 5.5, + "learning_rate": 1.3508190753549327e-06, + "loss": 1.5773, + "step": 30252 + }, + { + "epoch": 5.5, + "learning_rate": 1.3507644703312705e-06, + "loss": 0.6462, + "step": 30253 + }, + { + "epoch": 5.5, + "learning_rate": 1.3507098653076084e-06, + "loss": 0.8988, + "step": 30254 + }, + { + "epoch": 5.5, + "learning_rate": 1.3506552602839462e-06, + "loss": 0.2881, + "step": 30255 + }, + { + "epoch": 5.5, + "learning_rate": 1.350600655260284e-06, + "loss": 0.5045, + "step": 30256 + }, + { + "epoch": 5.5, + "learning_rate": 1.3505460502366218e-06, + "loss": 0.2475, + "step": 30257 + }, + { + "epoch": 5.5, + "learning_rate": 1.3504914452129596e-06, + "loss": 0.4175, + "step": 30258 + }, + { + "epoch": 5.5, + "learning_rate": 1.3504368401892975e-06, + "loss": 0.1399, + "step": 30259 + }, + { + "epoch": 5.5, + "learning_rate": 1.3503822351656352e-06, + "loss": 2.6463, + "step": 30260 + }, + { + "epoch": 5.5, + "learning_rate": 1.3503276301419732e-06, + "loss": 0.2805, + "step": 30261 + }, + { + "epoch": 5.5, + "learning_rate": 1.350273025118311e-06, + "loss": 0.0132, + "step": 30262 + }, + { + "epoch": 5.5, + "learning_rate": 1.3502184200946489e-06, + "loss": 0.0083, + "step": 30263 + }, + { + "epoch": 5.5, + "learning_rate": 1.3501638150709866e-06, + "loss": 2.1227, + "step": 30264 + }, + { + "epoch": 5.5, + "learning_rate": 1.3501092100473243e-06, + "loss": 0.4667, + "step": 30265 + }, + { + "epoch": 5.5, + "learning_rate": 1.3500546050236623e-06, + "loss": 0.0173, + "step": 30266 + }, + { + "epoch": 5.5, + "learning_rate": 1.35e-06, + "loss": 0.7635, + "step": 30267 + }, + { + "epoch": 5.5, + "learning_rate": 1.349945394976338e-06, + "loss": 0.9677, + "step": 30268 + }, + { + "epoch": 5.5, + "learning_rate": 1.3498907899526757e-06, + "loss": 1.0549, + "step": 30269 + }, + { + "epoch": 5.5, + "learning_rate": 1.3498361849290136e-06, + "loss": 0.2849, + "step": 30270 + }, + { + "epoch": 5.5, + "learning_rate": 1.3497815799053513e-06, + "loss": 1.0328, + "step": 30271 + }, + { + "epoch": 5.51, + "learning_rate": 1.349726974881689e-06, + "loss": 2.3096, + "step": 30272 + }, + { + "epoch": 5.51, + "learning_rate": 1.349672369858027e-06, + "loss": 0.6596, + "step": 30273 + }, + { + "epoch": 5.51, + "learning_rate": 1.3496177648343647e-06, + "loss": 0.1386, + "step": 30274 + }, + { + "epoch": 5.51, + "learning_rate": 1.3495631598107027e-06, + "loss": 0.1755, + "step": 30275 + }, + { + "epoch": 5.51, + "learning_rate": 1.3495085547870404e-06, + "loss": 0.8837, + "step": 30276 + }, + { + "epoch": 5.51, + "learning_rate": 1.3494539497633784e-06, + "loss": 0.2584, + "step": 30277 + }, + { + "epoch": 5.51, + "learning_rate": 1.349399344739716e-06, + "loss": 0.0056, + "step": 30278 + }, + { + "epoch": 5.51, + "learning_rate": 1.3493447397160538e-06, + "loss": 0.334, + "step": 30279 + }, + { + "epoch": 5.51, + "learning_rate": 1.3492901346923916e-06, + "loss": 0.5616, + "step": 30280 + }, + { + "epoch": 5.51, + "learning_rate": 1.3492355296687295e-06, + "loss": 0.2121, + "step": 30281 + }, + { + "epoch": 5.51, + "learning_rate": 1.3491809246450674e-06, + "loss": 0.0522, + "step": 30282 + }, + { + "epoch": 5.51, + "learning_rate": 1.3491263196214052e-06, + "loss": 0.3803, + "step": 30283 + }, + { + "epoch": 5.51, + "learning_rate": 1.3490717145977431e-06, + "loss": 0.0921, + "step": 30284 + }, + { + "epoch": 5.51, + "learning_rate": 1.3490171095740809e-06, + "loss": 1.5488, + "step": 30285 + }, + { + "epoch": 5.51, + "learning_rate": 1.3489625045504186e-06, + "loss": 0.3326, + "step": 30286 + }, + { + "epoch": 5.51, + "learning_rate": 1.3489078995267563e-06, + "loss": 0.1367, + "step": 30287 + }, + { + "epoch": 5.51, + "learning_rate": 1.3488532945030943e-06, + "loss": 1.7846, + "step": 30288 + }, + { + "epoch": 5.51, + "learning_rate": 1.3487986894794322e-06, + "loss": 0.3692, + "step": 30289 + }, + { + "epoch": 5.51, + "learning_rate": 1.34874408445577e-06, + "loss": 0.0111, + "step": 30290 + }, + { + "epoch": 5.51, + "learning_rate": 1.3486894794321079e-06, + "loss": 0.8266, + "step": 30291 + }, + { + "epoch": 5.51, + "learning_rate": 1.3486348744084456e-06, + "loss": 0.8468, + "step": 30292 + }, + { + "epoch": 5.51, + "learning_rate": 1.3485802693847833e-06, + "loss": 0.1063, + "step": 30293 + }, + { + "epoch": 5.51, + "learning_rate": 1.348525664361121e-06, + "loss": 0.1055, + "step": 30294 + }, + { + "epoch": 5.51, + "learning_rate": 1.348471059337459e-06, + "loss": 0.034, + "step": 30295 + }, + { + "epoch": 5.51, + "learning_rate": 1.348416454313797e-06, + "loss": 0.6334, + "step": 30296 + }, + { + "epoch": 5.51, + "learning_rate": 1.3483618492901347e-06, + "loss": 0.6792, + "step": 30297 + }, + { + "epoch": 5.51, + "learning_rate": 1.3483072442664726e-06, + "loss": 0.2396, + "step": 30298 + }, + { + "epoch": 5.51, + "learning_rate": 1.3482526392428104e-06, + "loss": 0.781, + "step": 30299 + }, + { + "epoch": 5.51, + "learning_rate": 1.348198034219148e-06, + "loss": 2.0121, + "step": 30300 + }, + { + "epoch": 5.51, + "learning_rate": 1.348143429195486e-06, + "loss": 0.3355, + "step": 30301 + }, + { + "epoch": 5.51, + "learning_rate": 1.348088824171824e-06, + "loss": 0.1236, + "step": 30302 + }, + { + "epoch": 5.51, + "learning_rate": 1.3480342191481617e-06, + "loss": 0.1823, + "step": 30303 + }, + { + "epoch": 5.51, + "learning_rate": 1.3479796141244995e-06, + "loss": 0.8737, + "step": 30304 + }, + { + "epoch": 5.51, + "learning_rate": 1.3479250091008374e-06, + "loss": 0.5383, + "step": 30305 + }, + { + "epoch": 5.51, + "learning_rate": 1.3478704040771751e-06, + "loss": 1.7817, + "step": 30306 + }, + { + "epoch": 5.51, + "learning_rate": 1.3478157990535129e-06, + "loss": 0.0149, + "step": 30307 + }, + { + "epoch": 5.51, + "learning_rate": 1.3477611940298508e-06, + "loss": 0.2795, + "step": 30308 + }, + { + "epoch": 5.51, + "learning_rate": 1.3477065890061887e-06, + "loss": 0.1681, + "step": 30309 + }, + { + "epoch": 5.51, + "learning_rate": 1.3476519839825265e-06, + "loss": 3.0036, + "step": 30310 + }, + { + "epoch": 5.51, + "learning_rate": 1.3475973789588642e-06, + "loss": 0.2688, + "step": 30311 + }, + { + "epoch": 5.51, + "learning_rate": 1.3475427739352022e-06, + "loss": 0.9634, + "step": 30312 + }, + { + "epoch": 5.51, + "learning_rate": 1.3474881689115399e-06, + "loss": 0.0326, + "step": 30313 + }, + { + "epoch": 5.51, + "learning_rate": 1.3474335638878776e-06, + "loss": 0.0129, + "step": 30314 + }, + { + "epoch": 5.51, + "learning_rate": 1.3473789588642156e-06, + "loss": 0.531, + "step": 30315 + }, + { + "epoch": 5.51, + "learning_rate": 1.3473243538405535e-06, + "loss": 0.1246, + "step": 30316 + }, + { + "epoch": 5.51, + "learning_rate": 1.3472697488168912e-06, + "loss": 0.7478, + "step": 30317 + }, + { + "epoch": 5.51, + "learning_rate": 1.347215143793229e-06, + "loss": 0.1201, + "step": 30318 + }, + { + "epoch": 5.51, + "learning_rate": 1.347160538769567e-06, + "loss": 0.208, + "step": 30319 + }, + { + "epoch": 5.51, + "learning_rate": 1.3471059337459046e-06, + "loss": 0.7874, + "step": 30320 + }, + { + "epoch": 5.51, + "learning_rate": 1.3470513287222424e-06, + "loss": 0.2879, + "step": 30321 + }, + { + "epoch": 5.51, + "learning_rate": 1.3469967236985803e-06, + "loss": 0.2898, + "step": 30322 + }, + { + "epoch": 5.51, + "learning_rate": 1.3469421186749183e-06, + "loss": 0.0109, + "step": 30323 + }, + { + "epoch": 5.51, + "learning_rate": 1.346887513651256e-06, + "loss": 0.4861, + "step": 30324 + }, + { + "epoch": 5.51, + "learning_rate": 1.3468329086275937e-06, + "loss": 0.2466, + "step": 30325 + }, + { + "epoch": 5.51, + "learning_rate": 1.3467783036039317e-06, + "loss": 0.0164, + "step": 30326 + }, + { + "epoch": 5.52, + "learning_rate": 1.3467236985802694e-06, + "loss": 0.2273, + "step": 30327 + }, + { + "epoch": 5.52, + "learning_rate": 1.3466690935566071e-06, + "loss": 0.5695, + "step": 30328 + }, + { + "epoch": 5.52, + "learning_rate": 1.346614488532945e-06, + "loss": 0.785, + "step": 30329 + }, + { + "epoch": 5.52, + "learning_rate": 1.346559883509283e-06, + "loss": 0.0219, + "step": 30330 + }, + { + "epoch": 5.52, + "learning_rate": 1.3465052784856207e-06, + "loss": 0.6445, + "step": 30331 + }, + { + "epoch": 5.52, + "learning_rate": 1.3464506734619585e-06, + "loss": 2.4117, + "step": 30332 + }, + { + "epoch": 5.52, + "learning_rate": 1.3463960684382964e-06, + "loss": 0.0858, + "step": 30333 + }, + { + "epoch": 5.52, + "learning_rate": 1.3463414634146342e-06, + "loss": 0.4271, + "step": 30334 + }, + { + "epoch": 5.52, + "learning_rate": 1.3462868583909719e-06, + "loss": 0.0061, + "step": 30335 + }, + { + "epoch": 5.52, + "learning_rate": 1.3462322533673098e-06, + "loss": 0.2758, + "step": 30336 + }, + { + "epoch": 5.52, + "learning_rate": 1.3461776483436478e-06, + "loss": 0.6722, + "step": 30337 + }, + { + "epoch": 5.52, + "learning_rate": 1.3461230433199855e-06, + "loss": 2.8447, + "step": 30338 + }, + { + "epoch": 5.52, + "learning_rate": 1.3460684382963232e-06, + "loss": 0.615, + "step": 30339 + }, + { + "epoch": 5.52, + "learning_rate": 1.3460138332726612e-06, + "loss": 0.0096, + "step": 30340 + }, + { + "epoch": 5.52, + "learning_rate": 1.345959228248999e-06, + "loss": 0.0183, + "step": 30341 + }, + { + "epoch": 5.52, + "learning_rate": 1.3459046232253369e-06, + "loss": 0.0061, + "step": 30342 + }, + { + "epoch": 5.52, + "learning_rate": 1.3458500182016746e-06, + "loss": 0.314, + "step": 30343 + }, + { + "epoch": 5.52, + "learning_rate": 1.3457954131780125e-06, + "loss": 0.3589, + "step": 30344 + }, + { + "epoch": 5.52, + "learning_rate": 1.3457408081543503e-06, + "loss": 0.0068, + "step": 30345 + }, + { + "epoch": 5.52, + "learning_rate": 1.345686203130688e-06, + "loss": 0.3431, + "step": 30346 + }, + { + "epoch": 5.52, + "learning_rate": 1.345631598107026e-06, + "loss": 0.3172, + "step": 30347 + }, + { + "epoch": 5.52, + "learning_rate": 1.3455769930833637e-06, + "loss": 0.1975, + "step": 30348 + }, + { + "epoch": 5.52, + "learning_rate": 1.3455223880597016e-06, + "loss": 0.0489, + "step": 30349 + }, + { + "epoch": 5.52, + "learning_rate": 1.3454677830360393e-06, + "loss": 0.9622, + "step": 30350 + }, + { + "epoch": 5.52, + "learning_rate": 1.3454131780123773e-06, + "loss": 0.2513, + "step": 30351 + }, + { + "epoch": 5.52, + "learning_rate": 1.345358572988715e-06, + "loss": 0.2819, + "step": 30352 + }, + { + "epoch": 5.52, + "learning_rate": 1.3453039679650527e-06, + "loss": 0.0629, + "step": 30353 + }, + { + "epoch": 5.52, + "learning_rate": 1.3452493629413905e-06, + "loss": 0.2741, + "step": 30354 + }, + { + "epoch": 5.52, + "learning_rate": 1.3451947579177284e-06, + "loss": 0.019, + "step": 30355 + }, + { + "epoch": 5.52, + "learning_rate": 1.3451401528940664e-06, + "loss": 0.3831, + "step": 30356 + }, + { + "epoch": 5.52, + "learning_rate": 1.345085547870404e-06, + "loss": 1.1593, + "step": 30357 + }, + { + "epoch": 5.52, + "learning_rate": 1.345030942846742e-06, + "loss": 0.1179, + "step": 30358 + }, + { + "epoch": 5.52, + "learning_rate": 1.3449763378230798e-06, + "loss": 1.3962, + "step": 30359 + }, + { + "epoch": 5.52, + "learning_rate": 1.3449217327994175e-06, + "loss": 0.5158, + "step": 30360 + }, + { + "epoch": 5.52, + "learning_rate": 1.3448671277757552e-06, + "loss": 0.6623, + "step": 30361 + }, + { + "epoch": 5.52, + "learning_rate": 1.3448125227520932e-06, + "loss": 0.4835, + "step": 30362 + }, + { + "epoch": 5.52, + "learning_rate": 1.3447579177284311e-06, + "loss": 0.2835, + "step": 30363 + }, + { + "epoch": 5.52, + "learning_rate": 1.3447033127047689e-06, + "loss": 1.0664, + "step": 30364 + }, + { + "epoch": 5.52, + "learning_rate": 1.3446487076811068e-06, + "loss": 4.235, + "step": 30365 + }, + { + "epoch": 5.52, + "learning_rate": 1.3445941026574445e-06, + "loss": 0.0098, + "step": 30366 + }, + { + "epoch": 5.52, + "learning_rate": 1.3445394976337823e-06, + "loss": 3.1064, + "step": 30367 + }, + { + "epoch": 5.52, + "learning_rate": 1.34448489261012e-06, + "loss": 1.3464, + "step": 30368 + }, + { + "epoch": 5.52, + "learning_rate": 1.344430287586458e-06, + "loss": 0.1647, + "step": 30369 + }, + { + "epoch": 5.52, + "learning_rate": 1.3443756825627959e-06, + "loss": 0.0064, + "step": 30370 + }, + { + "epoch": 5.52, + "learning_rate": 1.3443210775391336e-06, + "loss": 0.7728, + "step": 30371 + }, + { + "epoch": 5.52, + "learning_rate": 1.3442664725154716e-06, + "loss": 1.6306, + "step": 30372 + }, + { + "epoch": 5.52, + "learning_rate": 1.3442118674918093e-06, + "loss": 0.016, + "step": 30373 + }, + { + "epoch": 5.52, + "learning_rate": 1.344157262468147e-06, + "loss": 1.602, + "step": 30374 + }, + { + "epoch": 5.52, + "learning_rate": 1.3441026574444847e-06, + "loss": 0.7103, + "step": 30375 + }, + { + "epoch": 5.52, + "learning_rate": 1.3440480524208227e-06, + "loss": 0.2836, + "step": 30376 + }, + { + "epoch": 5.52, + "learning_rate": 1.3439934473971606e-06, + "loss": 2.3075, + "step": 30377 + }, + { + "epoch": 5.52, + "learning_rate": 1.3439388423734984e-06, + "loss": 0.2002, + "step": 30378 + }, + { + "epoch": 5.52, + "learning_rate": 1.3438842373498363e-06, + "loss": 0.4023, + "step": 30379 + }, + { + "epoch": 5.52, + "learning_rate": 1.343829632326174e-06, + "loss": 1.563, + "step": 30380 + }, + { + "epoch": 5.52, + "learning_rate": 1.3437750273025118e-06, + "loss": 1.3621, + "step": 30381 + }, + { + "epoch": 5.53, + "learning_rate": 1.3437204222788497e-06, + "loss": 0.2091, + "step": 30382 + }, + { + "epoch": 5.53, + "learning_rate": 1.3436658172551877e-06, + "loss": 0.3216, + "step": 30383 + }, + { + "epoch": 5.53, + "learning_rate": 1.3436112122315254e-06, + "loss": 0.3014, + "step": 30384 + }, + { + "epoch": 5.53, + "learning_rate": 1.3435566072078631e-06, + "loss": 0.2252, + "step": 30385 + }, + { + "epoch": 5.53, + "learning_rate": 1.343502002184201e-06, + "loss": 0.2264, + "step": 30386 + }, + { + "epoch": 5.53, + "learning_rate": 1.3434473971605388e-06, + "loss": 0.1669, + "step": 30387 + }, + { + "epoch": 5.53, + "learning_rate": 1.3433927921368765e-06, + "loss": 1.8817, + "step": 30388 + }, + { + "epoch": 5.53, + "learning_rate": 1.3433381871132145e-06, + "loss": 0.4753, + "step": 30389 + }, + { + "epoch": 5.53, + "learning_rate": 1.3432835820895524e-06, + "loss": 0.8464, + "step": 30390 + }, + { + "epoch": 5.53, + "learning_rate": 1.3432289770658901e-06, + "loss": 0.4992, + "step": 30391 + }, + { + "epoch": 5.53, + "learning_rate": 1.3431743720422279e-06, + "loss": 0.0115, + "step": 30392 + }, + { + "epoch": 5.53, + "learning_rate": 1.3431197670185658e-06, + "loss": 1.329, + "step": 30393 + }, + { + "epoch": 5.53, + "learning_rate": 1.3430651619949036e-06, + "loss": 0.3088, + "step": 30394 + }, + { + "epoch": 5.53, + "learning_rate": 1.3430105569712413e-06, + "loss": 0.526, + "step": 30395 + }, + { + "epoch": 5.53, + "learning_rate": 1.3429559519475792e-06, + "loss": 0.8018, + "step": 30396 + }, + { + "epoch": 5.53, + "learning_rate": 1.3429013469239172e-06, + "loss": 0.0056, + "step": 30397 + }, + { + "epoch": 5.53, + "learning_rate": 1.342846741900255e-06, + "loss": 0.028, + "step": 30398 + }, + { + "epoch": 5.53, + "learning_rate": 1.3427921368765926e-06, + "loss": 0.3148, + "step": 30399 + }, + { + "epoch": 5.53, + "learning_rate": 1.3427375318529306e-06, + "loss": 3.8443, + "step": 30400 + }, + { + "epoch": 5.53, + "learning_rate": 1.3426829268292683e-06, + "loss": 0.8026, + "step": 30401 + }, + { + "epoch": 5.53, + "learning_rate": 1.342628321805606e-06, + "loss": 0.1607, + "step": 30402 + }, + { + "epoch": 5.53, + "learning_rate": 1.342573716781944e-06, + "loss": 0.1158, + "step": 30403 + }, + { + "epoch": 5.53, + "learning_rate": 1.342519111758282e-06, + "loss": 0.2651, + "step": 30404 + }, + { + "epoch": 5.53, + "learning_rate": 1.3424645067346197e-06, + "loss": 3.1213, + "step": 30405 + }, + { + "epoch": 5.53, + "learning_rate": 1.3424099017109574e-06, + "loss": 0.2111, + "step": 30406 + }, + { + "epoch": 5.53, + "learning_rate": 1.3423552966872953e-06, + "loss": 1.3993, + "step": 30407 + }, + { + "epoch": 5.53, + "learning_rate": 1.342300691663633e-06, + "loss": 1.5247, + "step": 30408 + }, + { + "epoch": 5.53, + "learning_rate": 1.3422460866399708e-06, + "loss": 0.1362, + "step": 30409 + }, + { + "epoch": 5.53, + "learning_rate": 1.3421914816163087e-06, + "loss": 1.9543, + "step": 30410 + }, + { + "epoch": 5.53, + "learning_rate": 1.3421368765926467e-06, + "loss": 1.6917, + "step": 30411 + }, + { + "epoch": 5.53, + "learning_rate": 1.3420822715689844e-06, + "loss": 0.8837, + "step": 30412 + }, + { + "epoch": 5.53, + "learning_rate": 1.3420276665453222e-06, + "loss": 1.3489, + "step": 30413 + }, + { + "epoch": 5.53, + "learning_rate": 1.34197306152166e-06, + "loss": 0.7465, + "step": 30414 + }, + { + "epoch": 5.53, + "learning_rate": 1.3419184564979978e-06, + "loss": 0.5631, + "step": 30415 + }, + { + "epoch": 5.53, + "learning_rate": 1.3418638514743356e-06, + "loss": 0.4457, + "step": 30416 + }, + { + "epoch": 5.53, + "learning_rate": 1.3418092464506735e-06, + "loss": 0.3195, + "step": 30417 + }, + { + "epoch": 5.53, + "learning_rate": 1.3417546414270114e-06, + "loss": 0.6612, + "step": 30418 + }, + { + "epoch": 5.53, + "learning_rate": 1.3417000364033492e-06, + "loss": 0.0262, + "step": 30419 + }, + { + "epoch": 5.53, + "learning_rate": 1.341645431379687e-06, + "loss": 0.1439, + "step": 30420 + }, + { + "epoch": 5.53, + "learning_rate": 1.3415908263560249e-06, + "loss": 0.2195, + "step": 30421 + }, + { + "epoch": 5.53, + "learning_rate": 1.3415362213323626e-06, + "loss": 1.0132, + "step": 30422 + }, + { + "epoch": 5.53, + "learning_rate": 1.3414816163087003e-06, + "loss": 0.0918, + "step": 30423 + }, + { + "epoch": 5.53, + "learning_rate": 1.3414270112850383e-06, + "loss": 2.1759, + "step": 30424 + }, + { + "epoch": 5.53, + "learning_rate": 1.3413724062613762e-06, + "loss": 0.0073, + "step": 30425 + }, + { + "epoch": 5.53, + "learning_rate": 1.341317801237714e-06, + "loss": 0.0573, + "step": 30426 + }, + { + "epoch": 5.53, + "learning_rate": 1.3412631962140517e-06, + "loss": 0.0179, + "step": 30427 + }, + { + "epoch": 5.53, + "learning_rate": 1.3412085911903896e-06, + "loss": 0.7867, + "step": 30428 + }, + { + "epoch": 5.53, + "learning_rate": 1.3411539861667273e-06, + "loss": 0.0329, + "step": 30429 + }, + { + "epoch": 5.53, + "learning_rate": 1.3410993811430653e-06, + "loss": 1.6114, + "step": 30430 + }, + { + "epoch": 5.53, + "learning_rate": 1.341044776119403e-06, + "loss": 0.0165, + "step": 30431 + }, + { + "epoch": 5.53, + "learning_rate": 1.340990171095741e-06, + "loss": 0.0909, + "step": 30432 + }, + { + "epoch": 5.53, + "learning_rate": 1.3409355660720787e-06, + "loss": 0.5128, + "step": 30433 + }, + { + "epoch": 5.53, + "learning_rate": 1.3408809610484164e-06, + "loss": 0.1188, + "step": 30434 + }, + { + "epoch": 5.53, + "learning_rate": 1.3408263560247542e-06, + "loss": 0.3193, + "step": 30435 + }, + { + "epoch": 5.53, + "learning_rate": 1.340771751001092e-06, + "loss": 0.0516, + "step": 30436 + }, + { + "epoch": 5.54, + "learning_rate": 1.34071714597743e-06, + "loss": 0.3239, + "step": 30437 + }, + { + "epoch": 5.54, + "learning_rate": 1.3406625409537678e-06, + "loss": 0.9614, + "step": 30438 + }, + { + "epoch": 5.54, + "learning_rate": 1.3406079359301057e-06, + "loss": 0.1885, + "step": 30439 + }, + { + "epoch": 5.54, + "learning_rate": 1.3405533309064434e-06, + "loss": 0.7443, + "step": 30440 + }, + { + "epoch": 5.54, + "learning_rate": 1.3404987258827812e-06, + "loss": 0.0508, + "step": 30441 + }, + { + "epoch": 5.54, + "learning_rate": 1.340444120859119e-06, + "loss": 0.697, + "step": 30442 + }, + { + "epoch": 5.54, + "learning_rate": 1.3403895158354569e-06, + "loss": 0.0583, + "step": 30443 + }, + { + "epoch": 5.54, + "learning_rate": 1.3403349108117948e-06, + "loss": 0.837, + "step": 30444 + }, + { + "epoch": 5.54, + "learning_rate": 1.3402803057881325e-06, + "loss": 0.3437, + "step": 30445 + }, + { + "epoch": 5.54, + "learning_rate": 1.3402257007644705e-06, + "loss": 0.6464, + "step": 30446 + }, + { + "epoch": 5.54, + "learning_rate": 1.3401710957408082e-06, + "loss": 0.0051, + "step": 30447 + }, + { + "epoch": 5.54, + "learning_rate": 1.340116490717146e-06, + "loss": 0.4004, + "step": 30448 + }, + { + "epoch": 5.54, + "learning_rate": 1.3400618856934837e-06, + "loss": 0.3292, + "step": 30449 + }, + { + "epoch": 5.54, + "learning_rate": 1.3400072806698216e-06, + "loss": 0.0447, + "step": 30450 + }, + { + "epoch": 5.54, + "learning_rate": 1.3399526756461596e-06, + "loss": 0.4534, + "step": 30451 + }, + { + "epoch": 5.54, + "learning_rate": 1.3398980706224973e-06, + "loss": 0.3495, + "step": 30452 + }, + { + "epoch": 5.54, + "learning_rate": 1.3398434655988352e-06, + "loss": 0.2969, + "step": 30453 + }, + { + "epoch": 5.54, + "learning_rate": 1.339788860575173e-06, + "loss": 0.0227, + "step": 30454 + }, + { + "epoch": 5.54, + "learning_rate": 1.3397342555515107e-06, + "loss": 0.8745, + "step": 30455 + }, + { + "epoch": 5.54, + "learning_rate": 1.3396796505278484e-06, + "loss": 1.5217, + "step": 30456 + }, + { + "epoch": 5.54, + "learning_rate": 1.3396250455041864e-06, + "loss": 0.3548, + "step": 30457 + }, + { + "epoch": 5.54, + "learning_rate": 1.3395704404805243e-06, + "loss": 1.4165, + "step": 30458 + }, + { + "epoch": 5.54, + "learning_rate": 1.339515835456862e-06, + "loss": 0.0124, + "step": 30459 + }, + { + "epoch": 5.54, + "learning_rate": 1.3394612304332e-06, + "loss": 0.3833, + "step": 30460 + }, + { + "epoch": 5.54, + "learning_rate": 1.3394066254095377e-06, + "loss": 3.9851, + "step": 30461 + }, + { + "epoch": 5.54, + "learning_rate": 1.3393520203858754e-06, + "loss": 0.4458, + "step": 30462 + }, + { + "epoch": 5.54, + "learning_rate": 1.3392974153622132e-06, + "loss": 0.009, + "step": 30463 + }, + { + "epoch": 5.54, + "learning_rate": 1.3392428103385511e-06, + "loss": 0.501, + "step": 30464 + }, + { + "epoch": 5.54, + "learning_rate": 1.339188205314889e-06, + "loss": 0.4917, + "step": 30465 + }, + { + "epoch": 5.54, + "learning_rate": 1.3391336002912268e-06, + "loss": 1.008, + "step": 30466 + }, + { + "epoch": 5.54, + "learning_rate": 1.3390789952675647e-06, + "loss": 0.5871, + "step": 30467 + }, + { + "epoch": 5.54, + "learning_rate": 1.3390243902439025e-06, + "loss": 1.3483, + "step": 30468 + }, + { + "epoch": 5.54, + "learning_rate": 1.3389697852202402e-06, + "loss": 0.1425, + "step": 30469 + }, + { + "epoch": 5.54, + "learning_rate": 1.3389151801965781e-06, + "loss": 0.0294, + "step": 30470 + }, + { + "epoch": 5.54, + "learning_rate": 1.338860575172916e-06, + "loss": 1.0405, + "step": 30471 + }, + { + "epoch": 5.54, + "learning_rate": 1.3388059701492538e-06, + "loss": 1.6768, + "step": 30472 + }, + { + "epoch": 5.54, + "learning_rate": 1.3387513651255916e-06, + "loss": 1.4239, + "step": 30473 + }, + { + "epoch": 5.54, + "learning_rate": 1.3386967601019295e-06, + "loss": 1.5471, + "step": 30474 + }, + { + "epoch": 5.54, + "learning_rate": 1.3386421550782672e-06, + "loss": 0.1812, + "step": 30475 + }, + { + "epoch": 5.54, + "learning_rate": 1.338587550054605e-06, + "loss": 2.3515, + "step": 30476 + }, + { + "epoch": 5.54, + "learning_rate": 1.338532945030943e-06, + "loss": 1.014, + "step": 30477 + }, + { + "epoch": 5.54, + "learning_rate": 1.3384783400072808e-06, + "loss": 0.4361, + "step": 30478 + }, + { + "epoch": 5.54, + "learning_rate": 1.3384237349836186e-06, + "loss": 1.2977, + "step": 30479 + }, + { + "epoch": 5.54, + "learning_rate": 1.3383691299599563e-06, + "loss": 0.332, + "step": 30480 + }, + { + "epoch": 5.54, + "learning_rate": 1.3383145249362943e-06, + "loss": 0.7737, + "step": 30481 + }, + { + "epoch": 5.54, + "learning_rate": 1.338259919912632e-06, + "loss": 0.1051, + "step": 30482 + }, + { + "epoch": 5.54, + "learning_rate": 1.3382053148889697e-06, + "loss": 0.15, + "step": 30483 + }, + { + "epoch": 5.54, + "learning_rate": 1.3381507098653077e-06, + "loss": 0.1019, + "step": 30484 + }, + { + "epoch": 5.54, + "learning_rate": 1.3380961048416456e-06, + "loss": 0.4904, + "step": 30485 + }, + { + "epoch": 5.54, + "learning_rate": 1.3380414998179833e-06, + "loss": 1.0654, + "step": 30486 + }, + { + "epoch": 5.54, + "learning_rate": 1.337986894794321e-06, + "loss": 0.4528, + "step": 30487 + }, + { + "epoch": 5.54, + "learning_rate": 1.337932289770659e-06, + "loss": 0.2317, + "step": 30488 + }, + { + "epoch": 5.54, + "learning_rate": 1.3378776847469967e-06, + "loss": 0.6675, + "step": 30489 + }, + { + "epoch": 5.54, + "learning_rate": 1.3378230797233345e-06, + "loss": 0.5095, + "step": 30490 + }, + { + "epoch": 5.54, + "learning_rate": 1.3377684746996724e-06, + "loss": 1.7445, + "step": 30491 + }, + { + "epoch": 5.55, + "learning_rate": 1.3377138696760104e-06, + "loss": 0.2194, + "step": 30492 + }, + { + "epoch": 5.55, + "learning_rate": 1.337659264652348e-06, + "loss": 0.2138, + "step": 30493 + }, + { + "epoch": 5.55, + "learning_rate": 1.3376046596286858e-06, + "loss": 0.014, + "step": 30494 + }, + { + "epoch": 5.55, + "learning_rate": 1.3375500546050238e-06, + "loss": 0.198, + "step": 30495 + }, + { + "epoch": 5.55, + "learning_rate": 1.3374954495813615e-06, + "loss": 0.9696, + "step": 30496 + }, + { + "epoch": 5.55, + "learning_rate": 1.3374408445576992e-06, + "loss": 0.5276, + "step": 30497 + }, + { + "epoch": 5.55, + "learning_rate": 1.3373862395340372e-06, + "loss": 0.0231, + "step": 30498 + }, + { + "epoch": 5.55, + "learning_rate": 1.3373316345103751e-06, + "loss": 0.8713, + "step": 30499 + }, + { + "epoch": 5.55, + "learning_rate": 1.3372770294867129e-06, + "loss": 0.1596, + "step": 30500 + }, + { + "epoch": 5.55, + "learning_rate": 1.3372224244630506e-06, + "loss": 0.0126, + "step": 30501 + }, + { + "epoch": 5.55, + "learning_rate": 1.3371678194393885e-06, + "loss": 2.5388, + "step": 30502 + }, + { + "epoch": 5.55, + "learning_rate": 1.3371132144157263e-06, + "loss": 1.3607, + "step": 30503 + }, + { + "epoch": 5.55, + "learning_rate": 1.337058609392064e-06, + "loss": 0.2944, + "step": 30504 + }, + { + "epoch": 5.55, + "learning_rate": 1.337004004368402e-06, + "loss": 0.0128, + "step": 30505 + }, + { + "epoch": 5.55, + "learning_rate": 1.3369493993447399e-06, + "loss": 0.022, + "step": 30506 + }, + { + "epoch": 5.55, + "learning_rate": 1.3368947943210776e-06, + "loss": 1.8084, + "step": 30507 + }, + { + "epoch": 5.55, + "learning_rate": 1.3368401892974153e-06, + "loss": 1.3986, + "step": 30508 + }, + { + "epoch": 5.55, + "learning_rate": 1.336785584273753e-06, + "loss": 0.2131, + "step": 30509 + }, + { + "epoch": 5.55, + "learning_rate": 1.336730979250091e-06, + "loss": 0.1598, + "step": 30510 + }, + { + "epoch": 5.55, + "learning_rate": 1.336676374226429e-06, + "loss": 2.9859, + "step": 30511 + }, + { + "epoch": 5.55, + "learning_rate": 1.3366217692027667e-06, + "loss": 1.1671, + "step": 30512 + }, + { + "epoch": 5.55, + "learning_rate": 1.3365671641791046e-06, + "loss": 0.0307, + "step": 30513 + }, + { + "epoch": 5.55, + "learning_rate": 1.3365125591554424e-06, + "loss": 0.0155, + "step": 30514 + }, + { + "epoch": 5.55, + "learning_rate": 1.33645795413178e-06, + "loss": 1.0738, + "step": 30515 + }, + { + "epoch": 5.55, + "learning_rate": 1.3364033491081178e-06, + "loss": 0.5217, + "step": 30516 + }, + { + "epoch": 5.55, + "learning_rate": 1.3363487440844558e-06, + "loss": 0.5672, + "step": 30517 + }, + { + "epoch": 5.55, + "learning_rate": 1.3362941390607937e-06, + "loss": 0.1747, + "step": 30518 + }, + { + "epoch": 5.55, + "learning_rate": 1.3362395340371314e-06, + "loss": 0.0089, + "step": 30519 + }, + { + "epoch": 5.55, + "learning_rate": 1.3361849290134694e-06, + "loss": 0.1621, + "step": 30520 + }, + { + "epoch": 5.55, + "learning_rate": 1.3361303239898071e-06, + "loss": 1.3403, + "step": 30521 + }, + { + "epoch": 5.55, + "learning_rate": 1.3360757189661449e-06, + "loss": 0.7802, + "step": 30522 + }, + { + "epoch": 5.55, + "learning_rate": 1.3360211139424826e-06, + "loss": 1.5297, + "step": 30523 + }, + { + "epoch": 5.55, + "learning_rate": 1.3359665089188205e-06, + "loss": 1.0116, + "step": 30524 + }, + { + "epoch": 5.55, + "learning_rate": 1.3359119038951585e-06, + "loss": 1.2087, + "step": 30525 + }, + { + "epoch": 5.55, + "learning_rate": 1.3358572988714962e-06, + "loss": 0.1584, + "step": 30526 + }, + { + "epoch": 5.55, + "learning_rate": 1.3358026938478341e-06, + "loss": 1.291, + "step": 30527 + }, + { + "epoch": 5.55, + "learning_rate": 1.3357480888241719e-06, + "loss": 0.4859, + "step": 30528 + }, + { + "epoch": 5.55, + "learning_rate": 1.3356934838005096e-06, + "loss": 0.2612, + "step": 30529 + }, + { + "epoch": 5.55, + "learning_rate": 1.3356388787768473e-06, + "loss": 0.0428, + "step": 30530 + }, + { + "epoch": 5.55, + "learning_rate": 1.3355842737531853e-06, + "loss": 0.215, + "step": 30531 + }, + { + "epoch": 5.55, + "learning_rate": 1.3355296687295232e-06, + "loss": 0.9377, + "step": 30532 + }, + { + "epoch": 5.55, + "learning_rate": 1.335475063705861e-06, + "loss": 0.1343, + "step": 30533 + }, + { + "epoch": 5.55, + "learning_rate": 1.335420458682199e-06, + "loss": 0.1391, + "step": 30534 + }, + { + "epoch": 5.55, + "learning_rate": 1.3353658536585366e-06, + "loss": 1.4035, + "step": 30535 + }, + { + "epoch": 5.55, + "learning_rate": 1.3353112486348744e-06, + "loss": 1.3434, + "step": 30536 + }, + { + "epoch": 5.55, + "learning_rate": 1.335256643611212e-06, + "loss": 1.4244, + "step": 30537 + }, + { + "epoch": 5.55, + "learning_rate": 1.33520203858755e-06, + "loss": 1.0344, + "step": 30538 + }, + { + "epoch": 5.55, + "learning_rate": 1.335147433563888e-06, + "loss": 1.2241, + "step": 30539 + }, + { + "epoch": 5.55, + "learning_rate": 1.3350928285402257e-06, + "loss": 0.3245, + "step": 30540 + }, + { + "epoch": 5.55, + "learning_rate": 1.3350382235165637e-06, + "loss": 1.0676, + "step": 30541 + }, + { + "epoch": 5.55, + "learning_rate": 1.3349836184929014e-06, + "loss": 0.2561, + "step": 30542 + }, + { + "epoch": 5.55, + "learning_rate": 1.3349290134692391e-06, + "loss": 0.1022, + "step": 30543 + }, + { + "epoch": 5.55, + "learning_rate": 1.3348744084455769e-06, + "loss": 0.0717, + "step": 30544 + }, + { + "epoch": 5.55, + "learning_rate": 1.3348198034219148e-06, + "loss": 0.0971, + "step": 30545 + }, + { + "epoch": 5.55, + "learning_rate": 1.3347651983982527e-06, + "loss": 1.2824, + "step": 30546 + }, + { + "epoch": 5.56, + "learning_rate": 1.3347105933745905e-06, + "loss": 0.392, + "step": 30547 + }, + { + "epoch": 5.56, + "learning_rate": 1.3346559883509284e-06, + "loss": 0.0189, + "step": 30548 + }, + { + "epoch": 5.56, + "learning_rate": 1.3346013833272661e-06, + "loss": 0.7139, + "step": 30549 + }, + { + "epoch": 5.56, + "learning_rate": 1.3345467783036039e-06, + "loss": 0.1948, + "step": 30550 + }, + { + "epoch": 5.56, + "learning_rate": 1.3344921732799418e-06, + "loss": 0.1779, + "step": 30551 + }, + { + "epoch": 5.56, + "learning_rate": 1.3344375682562798e-06, + "loss": 0.0113, + "step": 30552 + }, + { + "epoch": 5.56, + "learning_rate": 1.3343829632326175e-06, + "loss": 0.2909, + "step": 30553 + }, + { + "epoch": 5.56, + "learning_rate": 1.3343283582089552e-06, + "loss": 0.4199, + "step": 30554 + }, + { + "epoch": 5.56, + "learning_rate": 1.3342737531852932e-06, + "loss": 0.0045, + "step": 30555 + }, + { + "epoch": 5.56, + "learning_rate": 1.334219148161631e-06, + "loss": 0.0738, + "step": 30556 + }, + { + "epoch": 5.56, + "learning_rate": 1.3341645431379686e-06, + "loss": 0.1793, + "step": 30557 + }, + { + "epoch": 5.56, + "learning_rate": 1.3341099381143066e-06, + "loss": 0.8248, + "step": 30558 + }, + { + "epoch": 5.56, + "learning_rate": 1.3340553330906445e-06, + "loss": 0.6359, + "step": 30559 + }, + { + "epoch": 5.56, + "learning_rate": 1.3340007280669823e-06, + "loss": 0.0407, + "step": 30560 + }, + { + "epoch": 5.56, + "learning_rate": 1.33394612304332e-06, + "loss": 0.0881, + "step": 30561 + }, + { + "epoch": 5.56, + "learning_rate": 1.333891518019658e-06, + "loss": 0.1175, + "step": 30562 + }, + { + "epoch": 5.56, + "learning_rate": 1.3338369129959957e-06, + "loss": 0.5322, + "step": 30563 + }, + { + "epoch": 5.56, + "learning_rate": 1.3337823079723334e-06, + "loss": 0.0737, + "step": 30564 + }, + { + "epoch": 5.56, + "learning_rate": 1.3337277029486713e-06, + "loss": 0.0048, + "step": 30565 + }, + { + "epoch": 5.56, + "learning_rate": 1.3336730979250093e-06, + "loss": 0.3724, + "step": 30566 + }, + { + "epoch": 5.56, + "learning_rate": 1.333618492901347e-06, + "loss": 0.0052, + "step": 30567 + }, + { + "epoch": 5.56, + "learning_rate": 1.3335638878776847e-06, + "loss": 0.1703, + "step": 30568 + }, + { + "epoch": 5.56, + "learning_rate": 1.3335092828540227e-06, + "loss": 1.1709, + "step": 30569 + }, + { + "epoch": 5.56, + "learning_rate": 1.3334546778303604e-06, + "loss": 0.145, + "step": 30570 + }, + { + "epoch": 5.56, + "learning_rate": 1.3334000728066982e-06, + "loss": 0.0518, + "step": 30571 + }, + { + "epoch": 5.56, + "learning_rate": 1.333345467783036e-06, + "loss": 2.5703, + "step": 30572 + }, + { + "epoch": 5.56, + "learning_rate": 1.333290862759374e-06, + "loss": 0.6434, + "step": 30573 + }, + { + "epoch": 5.56, + "learning_rate": 1.3332362577357118e-06, + "loss": 0.3362, + "step": 30574 + }, + { + "epoch": 5.56, + "learning_rate": 1.3331816527120495e-06, + "loss": 0.3184, + "step": 30575 + }, + { + "epoch": 5.56, + "learning_rate": 1.3331270476883874e-06, + "loss": 0.4439, + "step": 30576 + }, + { + "epoch": 5.56, + "learning_rate": 1.3330724426647252e-06, + "loss": 1.8886, + "step": 30577 + }, + { + "epoch": 5.56, + "learning_rate": 1.333017837641063e-06, + "loss": 0.2041, + "step": 30578 + }, + { + "epoch": 5.56, + "learning_rate": 1.3329632326174009e-06, + "loss": 0.501, + "step": 30579 + }, + { + "epoch": 5.56, + "learning_rate": 1.3329086275937388e-06, + "loss": 1.7576, + "step": 30580 + }, + { + "epoch": 5.56, + "learning_rate": 1.3328540225700765e-06, + "loss": 0.3522, + "step": 30581 + }, + { + "epoch": 5.56, + "learning_rate": 1.3327994175464143e-06, + "loss": 0.0082, + "step": 30582 + }, + { + "epoch": 5.56, + "learning_rate": 1.3327448125227522e-06, + "loss": 2.9678, + "step": 30583 + }, + { + "epoch": 5.56, + "learning_rate": 1.33269020749909e-06, + "loss": 0.2116, + "step": 30584 + }, + { + "epoch": 5.56, + "learning_rate": 1.3326356024754277e-06, + "loss": 0.22, + "step": 30585 + }, + { + "epoch": 5.56, + "learning_rate": 1.3325809974517656e-06, + "loss": 0.0164, + "step": 30586 + }, + { + "epoch": 5.56, + "learning_rate": 1.3325263924281036e-06, + "loss": 0.7049, + "step": 30587 + }, + { + "epoch": 5.56, + "learning_rate": 1.3324717874044413e-06, + "loss": 0.5741, + "step": 30588 + }, + { + "epoch": 5.56, + "learning_rate": 1.332417182380779e-06, + "loss": 0.0097, + "step": 30589 + }, + { + "epoch": 5.56, + "learning_rate": 1.3323625773571167e-06, + "loss": 1.2202, + "step": 30590 + }, + { + "epoch": 5.56, + "learning_rate": 1.3323079723334547e-06, + "loss": 0.025, + "step": 30591 + }, + { + "epoch": 5.56, + "learning_rate": 1.3322533673097924e-06, + "loss": 0.031, + "step": 30592 + }, + { + "epoch": 5.56, + "learning_rate": 1.3321987622861304e-06, + "loss": 0.005, + "step": 30593 + }, + { + "epoch": 5.56, + "learning_rate": 1.3321441572624683e-06, + "loss": 0.0218, + "step": 30594 + }, + { + "epoch": 5.56, + "learning_rate": 1.332089552238806e-06, + "loss": 1.1008, + "step": 30595 + }, + { + "epoch": 5.56, + "learning_rate": 1.3320349472151438e-06, + "loss": 0.6541, + "step": 30596 + }, + { + "epoch": 5.56, + "learning_rate": 1.3319803421914815e-06, + "loss": 0.306, + "step": 30597 + }, + { + "epoch": 5.56, + "learning_rate": 1.3319257371678194e-06, + "loss": 0.0993, + "step": 30598 + }, + { + "epoch": 5.56, + "learning_rate": 1.3318711321441574e-06, + "loss": 2.7972, + "step": 30599 + }, + { + "epoch": 5.56, + "learning_rate": 1.3318165271204951e-06, + "loss": 2.0435, + "step": 30600 + }, + { + "epoch": 5.56, + "learning_rate": 1.331761922096833e-06, + "loss": 0.0105, + "step": 30601 + }, + { + "epoch": 5.57, + "learning_rate": 1.3317073170731708e-06, + "loss": 0.3911, + "step": 30602 + }, + { + "epoch": 5.57, + "learning_rate": 1.3316527120495085e-06, + "loss": 0.1311, + "step": 30603 + }, + { + "epoch": 5.57, + "learning_rate": 1.3315981070258463e-06, + "loss": 0.0925, + "step": 30604 + }, + { + "epoch": 5.57, + "learning_rate": 1.3315435020021842e-06, + "loss": 0.0887, + "step": 30605 + }, + { + "epoch": 5.57, + "learning_rate": 1.3314888969785221e-06, + "loss": 1.2913, + "step": 30606 + }, + { + "epoch": 5.57, + "learning_rate": 1.3314342919548599e-06, + "loss": 0.0231, + "step": 30607 + }, + { + "epoch": 5.57, + "learning_rate": 1.3313796869311978e-06, + "loss": 0.0062, + "step": 30608 + }, + { + "epoch": 5.57, + "learning_rate": 1.3313250819075356e-06, + "loss": 0.145, + "step": 30609 + }, + { + "epoch": 5.57, + "learning_rate": 1.3312704768838733e-06, + "loss": 2.5608, + "step": 30610 + }, + { + "epoch": 5.57, + "learning_rate": 1.331215871860211e-06, + "loss": 0.9193, + "step": 30611 + }, + { + "epoch": 5.57, + "learning_rate": 1.331161266836549e-06, + "loss": 0.0115, + "step": 30612 + }, + { + "epoch": 5.57, + "learning_rate": 1.331106661812887e-06, + "loss": 0.7053, + "step": 30613 + }, + { + "epoch": 5.57, + "learning_rate": 1.3310520567892246e-06, + "loss": 0.0173, + "step": 30614 + }, + { + "epoch": 5.57, + "learning_rate": 1.3309974517655626e-06, + "loss": 0.0134, + "step": 30615 + }, + { + "epoch": 5.57, + "learning_rate": 1.3309428467419003e-06, + "loss": 0.0313, + "step": 30616 + }, + { + "epoch": 5.57, + "learning_rate": 1.330888241718238e-06, + "loss": 0.9279, + "step": 30617 + }, + { + "epoch": 5.57, + "learning_rate": 1.3308336366945758e-06, + "loss": 2.0863, + "step": 30618 + }, + { + "epoch": 5.57, + "learning_rate": 1.3307790316709137e-06, + "loss": 0.032, + "step": 30619 + }, + { + "epoch": 5.57, + "learning_rate": 1.3307244266472517e-06, + "loss": 0.7308, + "step": 30620 + }, + { + "epoch": 5.57, + "learning_rate": 1.3306698216235894e-06, + "loss": 0.2677, + "step": 30621 + }, + { + "epoch": 5.57, + "learning_rate": 1.3306152165999273e-06, + "loss": 0.3289, + "step": 30622 + }, + { + "epoch": 5.57, + "learning_rate": 1.330560611576265e-06, + "loss": 0.6523, + "step": 30623 + }, + { + "epoch": 5.57, + "learning_rate": 1.3305060065526028e-06, + "loss": 0.096, + "step": 30624 + }, + { + "epoch": 5.57, + "learning_rate": 1.3304514015289405e-06, + "loss": 0.2333, + "step": 30625 + }, + { + "epoch": 5.57, + "learning_rate": 1.3303967965052785e-06, + "loss": 2.5899, + "step": 30626 + }, + { + "epoch": 5.57, + "learning_rate": 1.3303421914816164e-06, + "loss": 0.0291, + "step": 30627 + }, + { + "epoch": 5.57, + "learning_rate": 1.3302875864579541e-06, + "loss": 0.0055, + "step": 30628 + }, + { + "epoch": 5.57, + "learning_rate": 1.330232981434292e-06, + "loss": 0.5888, + "step": 30629 + }, + { + "epoch": 5.57, + "learning_rate": 1.3301783764106298e-06, + "loss": 0.3152, + "step": 30630 + }, + { + "epoch": 5.57, + "learning_rate": 1.3301237713869676e-06, + "loss": 0.1689, + "step": 30631 + }, + { + "epoch": 5.57, + "learning_rate": 1.3300691663633053e-06, + "loss": 0.1377, + "step": 30632 + }, + { + "epoch": 5.57, + "learning_rate": 1.3300145613396432e-06, + "loss": 0.1102, + "step": 30633 + }, + { + "epoch": 5.57, + "learning_rate": 1.3299599563159812e-06, + "loss": 0.3881, + "step": 30634 + }, + { + "epoch": 5.57, + "learning_rate": 1.329905351292319e-06, + "loss": 0.355, + "step": 30635 + }, + { + "epoch": 5.57, + "learning_rate": 1.3298507462686568e-06, + "loss": 2.0545, + "step": 30636 + }, + { + "epoch": 5.57, + "learning_rate": 1.3297961412449946e-06, + "loss": 1.5156, + "step": 30637 + }, + { + "epoch": 5.57, + "learning_rate": 1.3297415362213323e-06, + "loss": 2.1351, + "step": 30638 + }, + { + "epoch": 5.57, + "learning_rate": 1.3296869311976703e-06, + "loss": 0.1406, + "step": 30639 + }, + { + "epoch": 5.57, + "learning_rate": 1.3296323261740082e-06, + "loss": 1.3205, + "step": 30640 + }, + { + "epoch": 5.57, + "learning_rate": 1.329577721150346e-06, + "loss": 0.0351, + "step": 30641 + }, + { + "epoch": 5.57, + "learning_rate": 1.3295231161266837e-06, + "loss": 0.7422, + "step": 30642 + }, + { + "epoch": 5.57, + "learning_rate": 1.3294685111030216e-06, + "loss": 0.0402, + "step": 30643 + }, + { + "epoch": 5.57, + "learning_rate": 1.3294139060793593e-06, + "loss": 0.0296, + "step": 30644 + }, + { + "epoch": 5.57, + "learning_rate": 1.329359301055697e-06, + "loss": 1.0208, + "step": 30645 + }, + { + "epoch": 5.57, + "learning_rate": 1.329304696032035e-06, + "loss": 0.0434, + "step": 30646 + }, + { + "epoch": 5.57, + "learning_rate": 1.329250091008373e-06, + "loss": 2.873, + "step": 30647 + }, + { + "epoch": 5.57, + "learning_rate": 1.3291954859847107e-06, + "loss": 2.3701, + "step": 30648 + }, + { + "epoch": 5.57, + "learning_rate": 1.3291408809610484e-06, + "loss": 0.0321, + "step": 30649 + }, + { + "epoch": 5.57, + "learning_rate": 1.3290862759373864e-06, + "loss": 0.2582, + "step": 30650 + }, + { + "epoch": 5.57, + "learning_rate": 1.329031670913724e-06, + "loss": 2.5219, + "step": 30651 + }, + { + "epoch": 5.57, + "learning_rate": 1.3289770658900618e-06, + "loss": 0.9308, + "step": 30652 + }, + { + "epoch": 5.57, + "learning_rate": 1.3289224608663998e-06, + "loss": 0.0097, + "step": 30653 + }, + { + "epoch": 5.57, + "learning_rate": 1.3288678558427377e-06, + "loss": 0.0353, + "step": 30654 + }, + { + "epoch": 5.57, + "learning_rate": 1.3288132508190754e-06, + "loss": 0.1434, + "step": 30655 + }, + { + "epoch": 5.57, + "learning_rate": 1.3287586457954132e-06, + "loss": 0.0294, + "step": 30656 + }, + { + "epoch": 5.58, + "learning_rate": 1.3287040407717511e-06, + "loss": 1.4108, + "step": 30657 + }, + { + "epoch": 5.58, + "learning_rate": 1.3286494357480888e-06, + "loss": 0.1139, + "step": 30658 + }, + { + "epoch": 5.58, + "learning_rate": 1.3285948307244266e-06, + "loss": 0.0848, + "step": 30659 + }, + { + "epoch": 5.58, + "learning_rate": 1.3285402257007645e-06, + "loss": 0.6632, + "step": 30660 + }, + { + "epoch": 5.58, + "learning_rate": 1.3284856206771025e-06, + "loss": 1.3311, + "step": 30661 + }, + { + "epoch": 5.58, + "learning_rate": 1.3284310156534402e-06, + "loss": 0.3401, + "step": 30662 + }, + { + "epoch": 5.58, + "learning_rate": 1.328376410629778e-06, + "loss": 0.3824, + "step": 30663 + }, + { + "epoch": 5.58, + "learning_rate": 1.3283218056061157e-06, + "loss": 0.4425, + "step": 30664 + }, + { + "epoch": 5.58, + "learning_rate": 1.3282672005824536e-06, + "loss": 0.3255, + "step": 30665 + }, + { + "epoch": 5.58, + "learning_rate": 1.3282125955587913e-06, + "loss": 0.5219, + "step": 30666 + }, + { + "epoch": 5.58, + "learning_rate": 1.3281579905351293e-06, + "loss": 0.2932, + "step": 30667 + }, + { + "epoch": 5.58, + "learning_rate": 1.3281033855114672e-06, + "loss": 1.2018, + "step": 30668 + }, + { + "epoch": 5.58, + "learning_rate": 1.328048780487805e-06, + "loss": 0.3315, + "step": 30669 + }, + { + "epoch": 5.58, + "learning_rate": 1.3279941754641427e-06, + "loss": 0.2016, + "step": 30670 + }, + { + "epoch": 5.58, + "learning_rate": 1.3279395704404804e-06, + "loss": 0.0247, + "step": 30671 + }, + { + "epoch": 5.58, + "learning_rate": 1.3278849654168184e-06, + "loss": 0.0175, + "step": 30672 + }, + { + "epoch": 5.58, + "learning_rate": 1.327830360393156e-06, + "loss": 0.1283, + "step": 30673 + }, + { + "epoch": 5.58, + "learning_rate": 1.327775755369494e-06, + "loss": 1.9506, + "step": 30674 + }, + { + "epoch": 5.58, + "learning_rate": 1.327721150345832e-06, + "loss": 0.3129, + "step": 30675 + }, + { + "epoch": 5.58, + "learning_rate": 1.3276665453221697e-06, + "loss": 2.338, + "step": 30676 + }, + { + "epoch": 5.58, + "learning_rate": 1.3276119402985074e-06, + "loss": 0.0615, + "step": 30677 + }, + { + "epoch": 5.58, + "learning_rate": 1.3275573352748452e-06, + "loss": 0.0437, + "step": 30678 + }, + { + "epoch": 5.58, + "learning_rate": 1.3275027302511831e-06, + "loss": 0.499, + "step": 30679 + }, + { + "epoch": 5.58, + "learning_rate": 1.327448125227521e-06, + "loss": 0.0073, + "step": 30680 + }, + { + "epoch": 5.58, + "learning_rate": 1.3273935202038588e-06, + "loss": 0.2282, + "step": 30681 + }, + { + "epoch": 5.58, + "learning_rate": 1.3273389151801967e-06, + "loss": 1.7458, + "step": 30682 + }, + { + "epoch": 5.58, + "learning_rate": 1.3272843101565345e-06, + "loss": 0.0218, + "step": 30683 + }, + { + "epoch": 5.58, + "learning_rate": 1.3272297051328722e-06, + "loss": 0.923, + "step": 30684 + }, + { + "epoch": 5.58, + "learning_rate": 1.32717510010921e-06, + "loss": 0.0701, + "step": 30685 + }, + { + "epoch": 5.58, + "learning_rate": 1.3271204950855479e-06, + "loss": 0.3276, + "step": 30686 + }, + { + "epoch": 5.58, + "learning_rate": 1.3270658900618858e-06, + "loss": 0.0072, + "step": 30687 + }, + { + "epoch": 5.58, + "learning_rate": 1.3270112850382236e-06, + "loss": 0.7031, + "step": 30688 + }, + { + "epoch": 5.58, + "learning_rate": 1.3269566800145615e-06, + "loss": 0.8393, + "step": 30689 + }, + { + "epoch": 5.58, + "learning_rate": 1.3269020749908992e-06, + "loss": 1.3318, + "step": 30690 + }, + { + "epoch": 5.58, + "learning_rate": 1.326847469967237e-06, + "loss": 0.102, + "step": 30691 + }, + { + "epoch": 5.58, + "learning_rate": 1.3267928649435747e-06, + "loss": 0.0149, + "step": 30692 + }, + { + "epoch": 5.58, + "learning_rate": 1.3267382599199126e-06, + "loss": 0.2553, + "step": 30693 + }, + { + "epoch": 5.58, + "learning_rate": 1.3266836548962506e-06, + "loss": 1.3473, + "step": 30694 + }, + { + "epoch": 5.58, + "learning_rate": 1.3266290498725883e-06, + "loss": 1.9744, + "step": 30695 + }, + { + "epoch": 5.58, + "learning_rate": 1.3265744448489263e-06, + "loss": 0.6802, + "step": 30696 + }, + { + "epoch": 5.58, + "learning_rate": 1.326519839825264e-06, + "loss": 0.0587, + "step": 30697 + }, + { + "epoch": 5.58, + "learning_rate": 1.3264652348016017e-06, + "loss": 0.1738, + "step": 30698 + }, + { + "epoch": 5.58, + "learning_rate": 1.3264106297779394e-06, + "loss": 0.7963, + "step": 30699 + }, + { + "epoch": 5.58, + "learning_rate": 1.3263560247542774e-06, + "loss": 0.443, + "step": 30700 + }, + { + "epoch": 5.58, + "learning_rate": 1.3263014197306153e-06, + "loss": 0.3999, + "step": 30701 + }, + { + "epoch": 5.58, + "learning_rate": 1.326246814706953e-06, + "loss": 1.9697, + "step": 30702 + }, + { + "epoch": 5.58, + "learning_rate": 1.326192209683291e-06, + "loss": 0.1019, + "step": 30703 + }, + { + "epoch": 5.58, + "learning_rate": 1.3261376046596287e-06, + "loss": 1.4371, + "step": 30704 + }, + { + "epoch": 5.58, + "learning_rate": 1.3260829996359665e-06, + "loss": 0.0994, + "step": 30705 + }, + { + "epoch": 5.58, + "learning_rate": 1.3260283946123042e-06, + "loss": 0.0826, + "step": 30706 + }, + { + "epoch": 5.58, + "learning_rate": 1.3259737895886421e-06, + "loss": 0.6197, + "step": 30707 + }, + { + "epoch": 5.58, + "learning_rate": 1.32591918456498e-06, + "loss": 0.8515, + "step": 30708 + }, + { + "epoch": 5.58, + "learning_rate": 1.3258645795413178e-06, + "loss": 0.8698, + "step": 30709 + }, + { + "epoch": 5.58, + "learning_rate": 1.3258099745176558e-06, + "loss": 0.998, + "step": 30710 + }, + { + "epoch": 5.58, + "learning_rate": 1.3257553694939935e-06, + "loss": 1.0616, + "step": 30711 + }, + { + "epoch": 5.59, + "learning_rate": 1.3257007644703312e-06, + "loss": 0.2797, + "step": 30712 + }, + { + "epoch": 5.59, + "learning_rate": 1.325646159446669e-06, + "loss": 0.7665, + "step": 30713 + }, + { + "epoch": 5.59, + "learning_rate": 1.325591554423007e-06, + "loss": 0.4242, + "step": 30714 + }, + { + "epoch": 5.59, + "learning_rate": 1.3255369493993448e-06, + "loss": 0.3904, + "step": 30715 + }, + { + "epoch": 5.59, + "learning_rate": 1.3254823443756826e-06, + "loss": 0.6508, + "step": 30716 + }, + { + "epoch": 5.59, + "learning_rate": 1.3254277393520205e-06, + "loss": 1.0353, + "step": 30717 + }, + { + "epoch": 5.59, + "learning_rate": 1.3253731343283583e-06, + "loss": 0.8781, + "step": 30718 + }, + { + "epoch": 5.59, + "learning_rate": 1.325318529304696e-06, + "loss": 0.0129, + "step": 30719 + }, + { + "epoch": 5.59, + "learning_rate": 1.325263924281034e-06, + "loss": 3.1026, + "step": 30720 + }, + { + "epoch": 5.59, + "learning_rate": 1.3252093192573717e-06, + "loss": 0.0073, + "step": 30721 + }, + { + "epoch": 5.59, + "learning_rate": 1.3251547142337096e-06, + "loss": 0.3371, + "step": 30722 + }, + { + "epoch": 5.59, + "learning_rate": 1.3251001092100473e-06, + "loss": 0.0963, + "step": 30723 + }, + { + "epoch": 5.59, + "learning_rate": 1.3250455041863853e-06, + "loss": 0.3075, + "step": 30724 + }, + { + "epoch": 5.59, + "learning_rate": 1.324990899162723e-06, + "loss": 0.0885, + "step": 30725 + }, + { + "epoch": 5.59, + "learning_rate": 1.3249362941390607e-06, + "loss": 1.8723, + "step": 30726 + }, + { + "epoch": 5.59, + "learning_rate": 1.3248816891153987e-06, + "loss": 0.2191, + "step": 30727 + }, + { + "epoch": 5.59, + "learning_rate": 1.3248270840917366e-06, + "loss": 0.0057, + "step": 30728 + }, + { + "epoch": 5.59, + "learning_rate": 1.3247724790680744e-06, + "loss": 0.4739, + "step": 30729 + }, + { + "epoch": 5.59, + "learning_rate": 1.324717874044412e-06, + "loss": 0.1378, + "step": 30730 + }, + { + "epoch": 5.59, + "learning_rate": 1.32466326902075e-06, + "loss": 0.0204, + "step": 30731 + }, + { + "epoch": 5.59, + "learning_rate": 1.3246086639970878e-06, + "loss": 2.9816, + "step": 30732 + }, + { + "epoch": 5.59, + "learning_rate": 1.3245540589734255e-06, + "loss": 2.2788, + "step": 30733 + }, + { + "epoch": 5.59, + "learning_rate": 1.3244994539497634e-06, + "loss": 0.0336, + "step": 30734 + }, + { + "epoch": 5.59, + "learning_rate": 1.3244448489261014e-06, + "loss": 0.0654, + "step": 30735 + }, + { + "epoch": 5.59, + "learning_rate": 1.3243902439024391e-06, + "loss": 1.5382, + "step": 30736 + }, + { + "epoch": 5.59, + "learning_rate": 1.3243356388787768e-06, + "loss": 0.4226, + "step": 30737 + }, + { + "epoch": 5.59, + "learning_rate": 1.3242810338551148e-06, + "loss": 2.6693, + "step": 30738 + }, + { + "epoch": 5.59, + "learning_rate": 1.3242264288314525e-06, + "loss": 0.4232, + "step": 30739 + }, + { + "epoch": 5.59, + "learning_rate": 1.3241718238077903e-06, + "loss": 0.1188, + "step": 30740 + }, + { + "epoch": 5.59, + "learning_rate": 1.3241172187841282e-06, + "loss": 0.5122, + "step": 30741 + }, + { + "epoch": 5.59, + "learning_rate": 1.3240626137604661e-06, + "loss": 0.8212, + "step": 30742 + }, + { + "epoch": 5.59, + "learning_rate": 1.3240080087368039e-06, + "loss": 0.7364, + "step": 30743 + }, + { + "epoch": 5.59, + "learning_rate": 1.3239534037131416e-06, + "loss": 1.0726, + "step": 30744 + }, + { + "epoch": 5.59, + "learning_rate": 1.3238987986894793e-06, + "loss": 0.1715, + "step": 30745 + }, + { + "epoch": 5.59, + "learning_rate": 1.3238441936658173e-06, + "loss": 0.1351, + "step": 30746 + }, + { + "epoch": 5.59, + "learning_rate": 1.323789588642155e-06, + "loss": 0.473, + "step": 30747 + }, + { + "epoch": 5.59, + "learning_rate": 1.323734983618493e-06, + "loss": 0.0123, + "step": 30748 + }, + { + "epoch": 5.59, + "learning_rate": 1.323680378594831e-06, + "loss": 0.0064, + "step": 30749 + }, + { + "epoch": 5.59, + "learning_rate": 1.3236257735711686e-06, + "loss": 2.424, + "step": 30750 + }, + { + "epoch": 5.59, + "learning_rate": 1.3235711685475064e-06, + "loss": 0.0082, + "step": 30751 + }, + { + "epoch": 5.59, + "learning_rate": 1.323516563523844e-06, + "loss": 0.8165, + "step": 30752 + }, + { + "epoch": 5.59, + "learning_rate": 1.323461958500182e-06, + "loss": 0.2709, + "step": 30753 + }, + { + "epoch": 5.59, + "learning_rate": 1.3234073534765198e-06, + "loss": 2.2639, + "step": 30754 + }, + { + "epoch": 5.59, + "learning_rate": 1.3233527484528577e-06, + "loss": 0.5412, + "step": 30755 + }, + { + "epoch": 5.59, + "learning_rate": 1.3232981434291957e-06, + "loss": 0.4787, + "step": 30756 + }, + { + "epoch": 5.59, + "learning_rate": 1.3232435384055334e-06, + "loss": 1.291, + "step": 30757 + }, + { + "epoch": 5.59, + "learning_rate": 1.3231889333818711e-06, + "loss": 0.4604, + "step": 30758 + }, + { + "epoch": 5.59, + "learning_rate": 1.3231343283582089e-06, + "loss": 0.0896, + "step": 30759 + }, + { + "epoch": 5.59, + "learning_rate": 1.3230797233345468e-06, + "loss": 0.711, + "step": 30760 + }, + { + "epoch": 5.59, + "learning_rate": 1.3230251183108845e-06, + "loss": 1.2939, + "step": 30761 + }, + { + "epoch": 5.59, + "learning_rate": 1.3229705132872225e-06, + "loss": 0.3056, + "step": 30762 + }, + { + "epoch": 5.59, + "learning_rate": 1.3229159082635604e-06, + "loss": 0.242, + "step": 30763 + }, + { + "epoch": 5.59, + "learning_rate": 1.3228613032398981e-06, + "loss": 0.0631, + "step": 30764 + }, + { + "epoch": 5.59, + "learning_rate": 1.3228066982162359e-06, + "loss": 0.2439, + "step": 30765 + }, + { + "epoch": 5.59, + "learning_rate": 1.3227520931925736e-06, + "loss": 0.1597, + "step": 30766 + }, + { + "epoch": 5.6, + "learning_rate": 1.3226974881689116e-06, + "loss": 0.1429, + "step": 30767 + }, + { + "epoch": 5.6, + "learning_rate": 1.3226428831452495e-06, + "loss": 0.4484, + "step": 30768 + }, + { + "epoch": 5.6, + "learning_rate": 1.3225882781215872e-06, + "loss": 1.1371, + "step": 30769 + }, + { + "epoch": 5.6, + "learning_rate": 1.3225336730979252e-06, + "loss": 1.3895, + "step": 30770 + }, + { + "epoch": 5.6, + "learning_rate": 1.322479068074263e-06, + "loss": 0.5202, + "step": 30771 + }, + { + "epoch": 5.6, + "learning_rate": 1.3224244630506006e-06, + "loss": 0.9267, + "step": 30772 + }, + { + "epoch": 5.6, + "learning_rate": 1.3223698580269384e-06, + "loss": 0.1109, + "step": 30773 + }, + { + "epoch": 5.6, + "learning_rate": 1.3223152530032763e-06, + "loss": 3.5707, + "step": 30774 + }, + { + "epoch": 5.6, + "learning_rate": 1.3222606479796143e-06, + "loss": 0.1537, + "step": 30775 + }, + { + "epoch": 5.6, + "learning_rate": 1.322206042955952e-06, + "loss": 1.4995, + "step": 30776 + }, + { + "epoch": 5.6, + "learning_rate": 1.32215143793229e-06, + "loss": 1.8856, + "step": 30777 + }, + { + "epoch": 5.6, + "learning_rate": 1.3220968329086277e-06, + "loss": 0.8646, + "step": 30778 + }, + { + "epoch": 5.6, + "learning_rate": 1.3220422278849654e-06, + "loss": 0.195, + "step": 30779 + }, + { + "epoch": 5.6, + "learning_rate": 1.3219876228613031e-06, + "loss": 0.3155, + "step": 30780 + }, + { + "epoch": 5.6, + "learning_rate": 1.321933017837641e-06, + "loss": 0.009, + "step": 30781 + }, + { + "epoch": 5.6, + "learning_rate": 1.321878412813979e-06, + "loss": 0.1947, + "step": 30782 + }, + { + "epoch": 5.6, + "learning_rate": 1.3218238077903167e-06, + "loss": 0.4225, + "step": 30783 + }, + { + "epoch": 5.6, + "learning_rate": 1.3217692027666547e-06, + "loss": 0.3166, + "step": 30784 + }, + { + "epoch": 5.6, + "learning_rate": 1.3217145977429924e-06, + "loss": 0.0927, + "step": 30785 + }, + { + "epoch": 5.6, + "learning_rate": 1.3216599927193301e-06, + "loss": 0.9635, + "step": 30786 + }, + { + "epoch": 5.6, + "learning_rate": 1.3216053876956679e-06, + "loss": 0.5602, + "step": 30787 + }, + { + "epoch": 5.6, + "learning_rate": 1.3215507826720058e-06, + "loss": 0.4122, + "step": 30788 + }, + { + "epoch": 5.6, + "learning_rate": 1.3214961776483438e-06, + "loss": 0.5521, + "step": 30789 + }, + { + "epoch": 5.6, + "learning_rate": 1.3214415726246815e-06, + "loss": 0.2513, + "step": 30790 + }, + { + "epoch": 5.6, + "learning_rate": 1.3213869676010194e-06, + "loss": 1.7118, + "step": 30791 + }, + { + "epoch": 5.6, + "learning_rate": 1.3213323625773572e-06, + "loss": 2.2231, + "step": 30792 + }, + { + "epoch": 5.6, + "learning_rate": 1.321277757553695e-06, + "loss": 0.0086, + "step": 30793 + }, + { + "epoch": 5.6, + "learning_rate": 1.3212231525300326e-06, + "loss": 0.2735, + "step": 30794 + }, + { + "epoch": 5.6, + "learning_rate": 1.3211685475063706e-06, + "loss": 0.4718, + "step": 30795 + }, + { + "epoch": 5.6, + "learning_rate": 1.3211139424827085e-06, + "loss": 0.9476, + "step": 30796 + }, + { + "epoch": 5.6, + "learning_rate": 1.3210593374590463e-06, + "loss": 0.8116, + "step": 30797 + }, + { + "epoch": 5.6, + "learning_rate": 1.3210047324353842e-06, + "loss": 0.0063, + "step": 30798 + }, + { + "epoch": 5.6, + "learning_rate": 1.320950127411722e-06, + "loss": 0.3552, + "step": 30799 + }, + { + "epoch": 5.6, + "learning_rate": 1.3208955223880597e-06, + "loss": 0.5475, + "step": 30800 + }, + { + "epoch": 5.6, + "learning_rate": 1.3208409173643974e-06, + "loss": 0.1516, + "step": 30801 + }, + { + "epoch": 5.6, + "learning_rate": 1.3207863123407353e-06, + "loss": 0.0063, + "step": 30802 + }, + { + "epoch": 5.6, + "learning_rate": 1.3207317073170733e-06, + "loss": 0.057, + "step": 30803 + }, + { + "epoch": 5.6, + "learning_rate": 1.320677102293411e-06, + "loss": 1.3907, + "step": 30804 + }, + { + "epoch": 5.6, + "learning_rate": 1.320622497269749e-06, + "loss": 0.1478, + "step": 30805 + }, + { + "epoch": 5.6, + "learning_rate": 1.3205678922460867e-06, + "loss": 0.5281, + "step": 30806 + }, + { + "epoch": 5.6, + "learning_rate": 1.3205132872224244e-06, + "loss": 0.0663, + "step": 30807 + }, + { + "epoch": 5.6, + "learning_rate": 1.3204586821987624e-06, + "loss": 0.0334, + "step": 30808 + }, + { + "epoch": 5.6, + "learning_rate": 1.3204040771751003e-06, + "loss": 0.2544, + "step": 30809 + }, + { + "epoch": 5.6, + "learning_rate": 1.320349472151438e-06, + "loss": 0.0948, + "step": 30810 + }, + { + "epoch": 5.6, + "learning_rate": 1.3202948671277758e-06, + "loss": 0.3281, + "step": 30811 + }, + { + "epoch": 5.6, + "learning_rate": 1.3202402621041137e-06, + "loss": 0.1485, + "step": 30812 + }, + { + "epoch": 5.6, + "learning_rate": 1.3201856570804514e-06, + "loss": 0.3537, + "step": 30813 + }, + { + "epoch": 5.6, + "learning_rate": 1.3201310520567892e-06, + "loss": 0.0071, + "step": 30814 + }, + { + "epoch": 5.6, + "learning_rate": 1.3200764470331271e-06, + "loss": 0.2041, + "step": 30815 + }, + { + "epoch": 5.6, + "learning_rate": 1.320021842009465e-06, + "loss": 1.6166, + "step": 30816 + }, + { + "epoch": 5.6, + "learning_rate": 1.3199672369858028e-06, + "loss": 0.8263, + "step": 30817 + }, + { + "epoch": 5.6, + "learning_rate": 1.3199126319621405e-06, + "loss": 1.1427, + "step": 30818 + }, + { + "epoch": 5.6, + "learning_rate": 1.3198580269384783e-06, + "loss": 0.2366, + "step": 30819 + }, + { + "epoch": 5.6, + "learning_rate": 1.3198034219148162e-06, + "loss": 0.9434, + "step": 30820 + }, + { + "epoch": 5.6, + "learning_rate": 1.319748816891154e-06, + "loss": 0.4932, + "step": 30821 + }, + { + "epoch": 5.61, + "learning_rate": 1.3196942118674919e-06, + "loss": 0.0105, + "step": 30822 + }, + { + "epoch": 5.61, + "learning_rate": 1.3196396068438298e-06, + "loss": 0.4849, + "step": 30823 + }, + { + "epoch": 5.61, + "learning_rate": 1.3195850018201675e-06, + "loss": 0.0259, + "step": 30824 + }, + { + "epoch": 5.61, + "learning_rate": 1.3195303967965053e-06, + "loss": 0.0043, + "step": 30825 + }, + { + "epoch": 5.61, + "learning_rate": 1.319475791772843e-06, + "loss": 0.046, + "step": 30826 + }, + { + "epoch": 5.61, + "learning_rate": 1.319421186749181e-06, + "loss": 1.8338, + "step": 30827 + }, + { + "epoch": 5.61, + "learning_rate": 1.3193665817255187e-06, + "loss": 0.0691, + "step": 30828 + }, + { + "epoch": 5.61, + "learning_rate": 1.3193119767018566e-06, + "loss": 0.6845, + "step": 30829 + }, + { + "epoch": 5.61, + "learning_rate": 1.3192573716781946e-06, + "loss": 0.0048, + "step": 30830 + }, + { + "epoch": 5.61, + "learning_rate": 1.3192027666545323e-06, + "loss": 0.0372, + "step": 30831 + }, + { + "epoch": 5.61, + "learning_rate": 1.31914816163087e-06, + "loss": 0.5999, + "step": 30832 + }, + { + "epoch": 5.61, + "learning_rate": 1.3190935566072078e-06, + "loss": 2.1804, + "step": 30833 + }, + { + "epoch": 5.61, + "learning_rate": 1.3190389515835457e-06, + "loss": 0.8846, + "step": 30834 + }, + { + "epoch": 5.61, + "learning_rate": 1.3189843465598834e-06, + "loss": 0.9774, + "step": 30835 + }, + { + "epoch": 5.61, + "learning_rate": 1.3189297415362214e-06, + "loss": 0.0083, + "step": 30836 + }, + { + "epoch": 5.61, + "learning_rate": 1.3188751365125593e-06, + "loss": 2.0579, + "step": 30837 + }, + { + "epoch": 5.61, + "learning_rate": 1.318820531488897e-06, + "loss": 0.0189, + "step": 30838 + }, + { + "epoch": 5.61, + "learning_rate": 1.3187659264652348e-06, + "loss": 0.4236, + "step": 30839 + }, + { + "epoch": 5.61, + "learning_rate": 1.3187113214415725e-06, + "loss": 0.1873, + "step": 30840 + }, + { + "epoch": 5.61, + "learning_rate": 1.3186567164179105e-06, + "loss": 0.0597, + "step": 30841 + }, + { + "epoch": 5.61, + "learning_rate": 1.3186021113942482e-06, + "loss": 0.7662, + "step": 30842 + }, + { + "epoch": 5.61, + "learning_rate": 1.3185475063705861e-06, + "loss": 0.4421, + "step": 30843 + }, + { + "epoch": 5.61, + "learning_rate": 1.318492901346924e-06, + "loss": 0.0761, + "step": 30844 + }, + { + "epoch": 5.61, + "learning_rate": 1.3184382963232618e-06, + "loss": 0.1499, + "step": 30845 + }, + { + "epoch": 5.61, + "learning_rate": 1.3183836912995995e-06, + "loss": 0.1906, + "step": 30846 + }, + { + "epoch": 5.61, + "learning_rate": 1.3183290862759373e-06, + "loss": 0.049, + "step": 30847 + }, + { + "epoch": 5.61, + "learning_rate": 1.3182744812522752e-06, + "loss": 0.7116, + "step": 30848 + }, + { + "epoch": 5.61, + "learning_rate": 1.3182198762286132e-06, + "loss": 0.2819, + "step": 30849 + }, + { + "epoch": 5.61, + "learning_rate": 1.318165271204951e-06, + "loss": 1.3071, + "step": 30850 + }, + { + "epoch": 5.61, + "learning_rate": 1.3181106661812888e-06, + "loss": 0.0088, + "step": 30851 + }, + { + "epoch": 5.61, + "learning_rate": 1.3180560611576266e-06, + "loss": 0.4027, + "step": 30852 + }, + { + "epoch": 5.61, + "learning_rate": 1.3180014561339643e-06, + "loss": 2.0215, + "step": 30853 + }, + { + "epoch": 5.61, + "learning_rate": 1.317946851110302e-06, + "loss": 0.0513, + "step": 30854 + }, + { + "epoch": 5.61, + "learning_rate": 1.31789224608664e-06, + "loss": 2.5008, + "step": 30855 + }, + { + "epoch": 5.61, + "learning_rate": 1.317837641062978e-06, + "loss": 0.4863, + "step": 30856 + }, + { + "epoch": 5.61, + "learning_rate": 1.3177830360393157e-06, + "loss": 0.067, + "step": 30857 + }, + { + "epoch": 5.61, + "learning_rate": 1.3177284310156536e-06, + "loss": 0.6161, + "step": 30858 + }, + { + "epoch": 5.61, + "learning_rate": 1.3176738259919913e-06, + "loss": 0.153, + "step": 30859 + }, + { + "epoch": 5.61, + "learning_rate": 1.317619220968329e-06, + "loss": 0.0208, + "step": 30860 + }, + { + "epoch": 5.61, + "learning_rate": 1.3175646159446668e-06, + "loss": 0.3897, + "step": 30861 + }, + { + "epoch": 5.61, + "learning_rate": 1.3175100109210047e-06, + "loss": 0.248, + "step": 30862 + }, + { + "epoch": 5.61, + "learning_rate": 1.3174554058973427e-06, + "loss": 0.7038, + "step": 30863 + }, + { + "epoch": 5.61, + "learning_rate": 1.3174008008736804e-06, + "loss": 0.7374, + "step": 30864 + }, + { + "epoch": 5.61, + "learning_rate": 1.3173461958500184e-06, + "loss": 0.719, + "step": 30865 + }, + { + "epoch": 5.61, + "learning_rate": 1.317291590826356e-06, + "loss": 0.0209, + "step": 30866 + }, + { + "epoch": 5.61, + "learning_rate": 1.3172369858026938e-06, + "loss": 0.1224, + "step": 30867 + }, + { + "epoch": 5.61, + "learning_rate": 1.3171823807790316e-06, + "loss": 0.4368, + "step": 30868 + }, + { + "epoch": 5.61, + "learning_rate": 1.3171277757553695e-06, + "loss": 1.4615, + "step": 30869 + }, + { + "epoch": 5.61, + "learning_rate": 1.3170731707317074e-06, + "loss": 0.7336, + "step": 30870 + }, + { + "epoch": 5.61, + "learning_rate": 1.3170185657080452e-06, + "loss": 0.3321, + "step": 30871 + }, + { + "epoch": 5.61, + "learning_rate": 1.3169639606843831e-06, + "loss": 0.0079, + "step": 30872 + }, + { + "epoch": 5.61, + "learning_rate": 1.3169093556607208e-06, + "loss": 0.1528, + "step": 30873 + }, + { + "epoch": 5.61, + "learning_rate": 1.3168547506370586e-06, + "loss": 0.1471, + "step": 30874 + }, + { + "epoch": 5.61, + "learning_rate": 1.3168001456133963e-06, + "loss": 1.6175, + "step": 30875 + }, + { + "epoch": 5.61, + "learning_rate": 1.3167455405897343e-06, + "loss": 2.9439, + "step": 30876 + }, + { + "epoch": 5.62, + "learning_rate": 1.3166909355660722e-06, + "loss": 1.4419, + "step": 30877 + }, + { + "epoch": 5.62, + "learning_rate": 1.31663633054241e-06, + "loss": 0.0123, + "step": 30878 + }, + { + "epoch": 5.62, + "learning_rate": 1.3165817255187479e-06, + "loss": 0.6808, + "step": 30879 + }, + { + "epoch": 5.62, + "learning_rate": 1.3165271204950856e-06, + "loss": 1.3641, + "step": 30880 + }, + { + "epoch": 5.62, + "learning_rate": 1.3164725154714233e-06, + "loss": 0.1938, + "step": 30881 + }, + { + "epoch": 5.62, + "learning_rate": 1.316417910447761e-06, + "loss": 0.3672, + "step": 30882 + }, + { + "epoch": 5.62, + "learning_rate": 1.316363305424099e-06, + "loss": 0.0319, + "step": 30883 + }, + { + "epoch": 5.62, + "learning_rate": 1.316308700400437e-06, + "loss": 0.0666, + "step": 30884 + }, + { + "epoch": 5.62, + "learning_rate": 1.3162540953767747e-06, + "loss": 0.055, + "step": 30885 + }, + { + "epoch": 5.62, + "learning_rate": 1.3161994903531126e-06, + "loss": 0.5234, + "step": 30886 + }, + { + "epoch": 5.62, + "learning_rate": 1.3161448853294504e-06, + "loss": 0.0123, + "step": 30887 + }, + { + "epoch": 5.62, + "learning_rate": 1.316090280305788e-06, + "loss": 0.6753, + "step": 30888 + }, + { + "epoch": 5.62, + "learning_rate": 1.3160356752821258e-06, + "loss": 0.8152, + "step": 30889 + }, + { + "epoch": 5.62, + "learning_rate": 1.3159810702584638e-06, + "loss": 0.3705, + "step": 30890 + }, + { + "epoch": 5.62, + "learning_rate": 1.3159264652348017e-06, + "loss": 0.4256, + "step": 30891 + }, + { + "epoch": 5.62, + "learning_rate": 1.3158718602111394e-06, + "loss": 0.1328, + "step": 30892 + }, + { + "epoch": 5.62, + "learning_rate": 1.3158172551874774e-06, + "loss": 1.1094, + "step": 30893 + }, + { + "epoch": 5.62, + "learning_rate": 1.3157626501638151e-06, + "loss": 0.1762, + "step": 30894 + }, + { + "epoch": 5.62, + "learning_rate": 1.3157080451401528e-06, + "loss": 4.5503, + "step": 30895 + }, + { + "epoch": 5.62, + "learning_rate": 1.3156534401164908e-06, + "loss": 0.022, + "step": 30896 + }, + { + "epoch": 5.62, + "learning_rate": 1.3155988350928287e-06, + "loss": 0.5317, + "step": 30897 + }, + { + "epoch": 5.62, + "learning_rate": 1.3155442300691665e-06, + "loss": 0.6254, + "step": 30898 + }, + { + "epoch": 5.62, + "learning_rate": 1.3154896250455042e-06, + "loss": 0.4968, + "step": 30899 + }, + { + "epoch": 5.62, + "learning_rate": 1.315435020021842e-06, + "loss": 0.4182, + "step": 30900 + }, + { + "epoch": 5.62, + "learning_rate": 1.3153804149981799e-06, + "loss": 0.4534, + "step": 30901 + }, + { + "epoch": 5.62, + "learning_rate": 1.3153258099745176e-06, + "loss": 0.0091, + "step": 30902 + }, + { + "epoch": 5.62, + "learning_rate": 1.3152712049508555e-06, + "loss": 0.2012, + "step": 30903 + }, + { + "epoch": 5.62, + "learning_rate": 1.3152165999271935e-06, + "loss": 2.2823, + "step": 30904 + }, + { + "epoch": 5.62, + "learning_rate": 1.3151619949035312e-06, + "loss": 0.0246, + "step": 30905 + }, + { + "epoch": 5.62, + "learning_rate": 1.315107389879869e-06, + "loss": 0.2602, + "step": 30906 + }, + { + "epoch": 5.62, + "learning_rate": 1.3150527848562067e-06, + "loss": 0.6021, + "step": 30907 + }, + { + "epoch": 5.62, + "learning_rate": 1.3149981798325446e-06, + "loss": 1.7068, + "step": 30908 + }, + { + "epoch": 5.62, + "learning_rate": 1.3149435748088824e-06, + "loss": 0.8972, + "step": 30909 + }, + { + "epoch": 5.62, + "learning_rate": 1.3148889697852203e-06, + "loss": 2.7695, + "step": 30910 + }, + { + "epoch": 5.62, + "learning_rate": 1.3148343647615582e-06, + "loss": 1.5274, + "step": 30911 + }, + { + "epoch": 5.62, + "learning_rate": 1.314779759737896e-06, + "loss": 1.3551, + "step": 30912 + }, + { + "epoch": 5.62, + "learning_rate": 1.3147251547142337e-06, + "loss": 0.0931, + "step": 30913 + }, + { + "epoch": 5.62, + "learning_rate": 1.3146705496905714e-06, + "loss": 1.0346, + "step": 30914 + }, + { + "epoch": 5.62, + "learning_rate": 1.3146159446669094e-06, + "loss": 2.1714, + "step": 30915 + }, + { + "epoch": 5.62, + "learning_rate": 1.3145613396432471e-06, + "loss": 0.0093, + "step": 30916 + }, + { + "epoch": 5.62, + "learning_rate": 1.314506734619585e-06, + "loss": 0.0081, + "step": 30917 + }, + { + "epoch": 5.62, + "learning_rate": 1.314452129595923e-06, + "loss": 1.1068, + "step": 30918 + }, + { + "epoch": 5.62, + "learning_rate": 1.3143975245722607e-06, + "loss": 0.4378, + "step": 30919 + }, + { + "epoch": 5.62, + "learning_rate": 1.3143429195485985e-06, + "loss": 0.155, + "step": 30920 + }, + { + "epoch": 5.62, + "learning_rate": 1.3142883145249362e-06, + "loss": 1.0656, + "step": 30921 + }, + { + "epoch": 5.62, + "learning_rate": 1.3142337095012741e-06, + "loss": 1.0384, + "step": 30922 + }, + { + "epoch": 5.62, + "learning_rate": 1.3141791044776119e-06, + "loss": 0.1869, + "step": 30923 + }, + { + "epoch": 5.62, + "learning_rate": 1.3141244994539498e-06, + "loss": 0.2416, + "step": 30924 + }, + { + "epoch": 5.62, + "learning_rate": 1.3140698944302878e-06, + "loss": 0.8263, + "step": 30925 + }, + { + "epoch": 5.62, + "learning_rate": 1.3140152894066255e-06, + "loss": 0.0987, + "step": 30926 + }, + { + "epoch": 5.62, + "learning_rate": 1.3139606843829632e-06, + "loss": 2.7346, + "step": 30927 + }, + { + "epoch": 5.62, + "learning_rate": 1.313906079359301e-06, + "loss": 0.0116, + "step": 30928 + }, + { + "epoch": 5.62, + "learning_rate": 1.313851474335639e-06, + "loss": 1.0655, + "step": 30929 + }, + { + "epoch": 5.62, + "learning_rate": 1.3137968693119766e-06, + "loss": 0.0079, + "step": 30930 + }, + { + "epoch": 5.62, + "learning_rate": 1.3137422642883146e-06, + "loss": 0.2051, + "step": 30931 + }, + { + "epoch": 5.63, + "learning_rate": 1.3136876592646525e-06, + "loss": 0.0891, + "step": 30932 + }, + { + "epoch": 5.63, + "learning_rate": 1.3136330542409902e-06, + "loss": 0.0089, + "step": 30933 + }, + { + "epoch": 5.63, + "learning_rate": 1.313578449217328e-06, + "loss": 0.2713, + "step": 30934 + }, + { + "epoch": 5.63, + "learning_rate": 1.3135238441936657e-06, + "loss": 0.0205, + "step": 30935 + }, + { + "epoch": 5.63, + "learning_rate": 1.3134692391700037e-06, + "loss": 0.8666, + "step": 30936 + }, + { + "epoch": 5.63, + "learning_rate": 1.3134146341463416e-06, + "loss": 1.0416, + "step": 30937 + }, + { + "epoch": 5.63, + "learning_rate": 1.3133600291226793e-06, + "loss": 0.4657, + "step": 30938 + }, + { + "epoch": 5.63, + "learning_rate": 1.3133054240990173e-06, + "loss": 0.0136, + "step": 30939 + }, + { + "epoch": 5.63, + "learning_rate": 1.313250819075355e-06, + "loss": 1.1064, + "step": 30940 + }, + { + "epoch": 5.63, + "learning_rate": 1.3131962140516927e-06, + "loss": 4.3533, + "step": 30941 + }, + { + "epoch": 5.63, + "learning_rate": 1.3131416090280305e-06, + "loss": 0.739, + "step": 30942 + }, + { + "epoch": 5.63, + "learning_rate": 1.3130870040043684e-06, + "loss": 0.0097, + "step": 30943 + }, + { + "epoch": 5.63, + "learning_rate": 1.3130323989807064e-06, + "loss": 0.1776, + "step": 30944 + }, + { + "epoch": 5.63, + "learning_rate": 1.312977793957044e-06, + "loss": 0.01, + "step": 30945 + }, + { + "epoch": 5.63, + "learning_rate": 1.312923188933382e-06, + "loss": 0.3188, + "step": 30946 + }, + { + "epoch": 5.63, + "learning_rate": 1.3128685839097198e-06, + "loss": 0.1342, + "step": 30947 + }, + { + "epoch": 5.63, + "learning_rate": 1.3128139788860575e-06, + "loss": 0.6679, + "step": 30948 + }, + { + "epoch": 5.63, + "learning_rate": 1.3127593738623952e-06, + "loss": 0.2571, + "step": 30949 + }, + { + "epoch": 5.63, + "learning_rate": 1.3127047688387332e-06, + "loss": 0.471, + "step": 30950 + }, + { + "epoch": 5.63, + "learning_rate": 1.3126501638150711e-06, + "loss": 0.3749, + "step": 30951 + }, + { + "epoch": 5.63, + "learning_rate": 1.3125955587914088e-06, + "loss": 0.0895, + "step": 30952 + }, + { + "epoch": 5.63, + "learning_rate": 1.3125409537677468e-06, + "loss": 0.2824, + "step": 30953 + }, + { + "epoch": 5.63, + "learning_rate": 1.3124863487440845e-06, + "loss": 0.0127, + "step": 30954 + }, + { + "epoch": 5.63, + "learning_rate": 1.3124317437204223e-06, + "loss": 0.2608, + "step": 30955 + }, + { + "epoch": 5.63, + "learning_rate": 1.31237713869676e-06, + "loss": 0.1685, + "step": 30956 + }, + { + "epoch": 5.63, + "learning_rate": 1.312322533673098e-06, + "loss": 0.2989, + "step": 30957 + }, + { + "epoch": 5.63, + "learning_rate": 1.3122679286494359e-06, + "loss": 0.3427, + "step": 30958 + }, + { + "epoch": 5.63, + "learning_rate": 1.3122133236257736e-06, + "loss": 0.4796, + "step": 30959 + }, + { + "epoch": 5.63, + "learning_rate": 1.3121587186021115e-06, + "loss": 0.0091, + "step": 30960 + }, + { + "epoch": 5.63, + "learning_rate": 1.3121041135784493e-06, + "loss": 0.007, + "step": 30961 + }, + { + "epoch": 5.63, + "learning_rate": 1.312049508554787e-06, + "loss": 3.963, + "step": 30962 + }, + { + "epoch": 5.63, + "learning_rate": 1.3119949035311247e-06, + "loss": 0.0071, + "step": 30963 + }, + { + "epoch": 5.63, + "learning_rate": 1.3119402985074627e-06, + "loss": 0.2576, + "step": 30964 + }, + { + "epoch": 5.63, + "learning_rate": 1.3118856934838006e-06, + "loss": 0.006, + "step": 30965 + }, + { + "epoch": 5.63, + "learning_rate": 1.3118310884601384e-06, + "loss": 2.002, + "step": 30966 + }, + { + "epoch": 5.63, + "learning_rate": 1.3117764834364763e-06, + "loss": 1.8043, + "step": 30967 + }, + { + "epoch": 5.63, + "learning_rate": 1.311721878412814e-06, + "loss": 0.6059, + "step": 30968 + }, + { + "epoch": 5.63, + "learning_rate": 1.3116672733891518e-06, + "loss": 0.0587, + "step": 30969 + }, + { + "epoch": 5.63, + "learning_rate": 1.3116126683654895e-06, + "loss": 0.0631, + "step": 30970 + }, + { + "epoch": 5.63, + "learning_rate": 1.3115580633418274e-06, + "loss": 0.4049, + "step": 30971 + }, + { + "epoch": 5.63, + "learning_rate": 1.3115034583181654e-06, + "loss": 0.2959, + "step": 30972 + }, + { + "epoch": 5.63, + "learning_rate": 1.3114488532945031e-06, + "loss": 1.4164, + "step": 30973 + }, + { + "epoch": 5.63, + "learning_rate": 1.3113942482708408e-06, + "loss": 0.0904, + "step": 30974 + }, + { + "epoch": 5.63, + "learning_rate": 1.3113396432471788e-06, + "loss": 0.9384, + "step": 30975 + }, + { + "epoch": 5.63, + "learning_rate": 1.3112850382235165e-06, + "loss": 0.9409, + "step": 30976 + }, + { + "epoch": 5.63, + "learning_rate": 1.3112304331998545e-06, + "loss": 0.3527, + "step": 30977 + }, + { + "epoch": 5.63, + "learning_rate": 1.3111758281761924e-06, + "loss": 0.1218, + "step": 30978 + }, + { + "epoch": 5.63, + "learning_rate": 1.3111212231525301e-06, + "loss": 0.1353, + "step": 30979 + }, + { + "epoch": 5.63, + "learning_rate": 1.3110666181288679e-06, + "loss": 1.063, + "step": 30980 + }, + { + "epoch": 5.63, + "learning_rate": 1.3110120131052056e-06, + "loss": 0.1151, + "step": 30981 + }, + { + "epoch": 5.63, + "learning_rate": 1.3109574080815435e-06, + "loss": 0.1824, + "step": 30982 + }, + { + "epoch": 5.63, + "learning_rate": 1.3109028030578813e-06, + "loss": 0.8141, + "step": 30983 + }, + { + "epoch": 5.63, + "learning_rate": 1.3108481980342192e-06, + "loss": 0.9411, + "step": 30984 + }, + { + "epoch": 5.63, + "learning_rate": 1.3107935930105572e-06, + "loss": 0.1863, + "step": 30985 + }, + { + "epoch": 5.63, + "learning_rate": 1.310738987986895e-06, + "loss": 0.1847, + "step": 30986 + }, + { + "epoch": 5.64, + "learning_rate": 1.3106843829632326e-06, + "loss": 1.4597, + "step": 30987 + }, + { + "epoch": 5.64, + "learning_rate": 1.3106297779395704e-06, + "loss": 0.0056, + "step": 30988 + }, + { + "epoch": 5.64, + "learning_rate": 1.3105751729159083e-06, + "loss": 0.0425, + "step": 30989 + }, + { + "epoch": 5.64, + "learning_rate": 1.310520567892246e-06, + "loss": 0.3032, + "step": 30990 + }, + { + "epoch": 5.64, + "learning_rate": 1.310465962868584e-06, + "loss": 0.8825, + "step": 30991 + }, + { + "epoch": 5.64, + "learning_rate": 1.310411357844922e-06, + "loss": 1.4994, + "step": 30992 + }, + { + "epoch": 5.64, + "learning_rate": 1.3103567528212597e-06, + "loss": 0.7746, + "step": 30993 + }, + { + "epoch": 5.64, + "learning_rate": 1.3103021477975974e-06, + "loss": 0.8151, + "step": 30994 + }, + { + "epoch": 5.64, + "learning_rate": 1.3102475427739351e-06, + "loss": 2.9595, + "step": 30995 + }, + { + "epoch": 5.64, + "learning_rate": 1.310192937750273e-06, + "loss": 2.8211, + "step": 30996 + }, + { + "epoch": 5.64, + "learning_rate": 1.3101383327266108e-06, + "loss": 0.3379, + "step": 30997 + }, + { + "epoch": 5.64, + "learning_rate": 1.3100837277029487e-06, + "loss": 0.5224, + "step": 30998 + }, + { + "epoch": 5.64, + "learning_rate": 1.3100291226792867e-06, + "loss": 0.3265, + "step": 30999 + }, + { + "epoch": 5.64, + "learning_rate": 1.3099745176556244e-06, + "loss": 0.0071, + "step": 31000 + }, + { + "epoch": 5.64, + "learning_rate": 1.3099199126319621e-06, + "loss": 0.5991, + "step": 31001 + }, + { + "epoch": 5.64, + "learning_rate": 1.3098653076082999e-06, + "loss": 0.4784, + "step": 31002 + }, + { + "epoch": 5.64, + "learning_rate": 1.3098107025846378e-06, + "loss": 0.4296, + "step": 31003 + }, + { + "epoch": 5.64, + "learning_rate": 1.3097560975609755e-06, + "loss": 0.741, + "step": 31004 + }, + { + "epoch": 5.64, + "learning_rate": 1.3097014925373135e-06, + "loss": 1.0706, + "step": 31005 + }, + { + "epoch": 5.64, + "learning_rate": 1.3096468875136514e-06, + "loss": 0.2018, + "step": 31006 + }, + { + "epoch": 5.64, + "learning_rate": 1.3095922824899892e-06, + "loss": 0.1903, + "step": 31007 + }, + { + "epoch": 5.64, + "learning_rate": 1.309537677466327e-06, + "loss": 1.1839, + "step": 31008 + }, + { + "epoch": 5.64, + "learning_rate": 1.3094830724426646e-06, + "loss": 0.5712, + "step": 31009 + }, + { + "epoch": 5.64, + "learning_rate": 1.3094284674190026e-06, + "loss": 0.1403, + "step": 31010 + }, + { + "epoch": 5.64, + "learning_rate": 1.3093738623953403e-06, + "loss": 0.6999, + "step": 31011 + }, + { + "epoch": 5.64, + "learning_rate": 1.3093192573716782e-06, + "loss": 1.1101, + "step": 31012 + }, + { + "epoch": 5.64, + "learning_rate": 1.3092646523480162e-06, + "loss": 0.6419, + "step": 31013 + }, + { + "epoch": 5.64, + "learning_rate": 1.309210047324354e-06, + "loss": 0.2671, + "step": 31014 + }, + { + "epoch": 5.64, + "learning_rate": 1.3091554423006917e-06, + "loss": 0.3923, + "step": 31015 + }, + { + "epoch": 5.64, + "learning_rate": 1.3091008372770294e-06, + "loss": 1.3906, + "step": 31016 + }, + { + "epoch": 5.64, + "learning_rate": 1.3090462322533673e-06, + "loss": 0.1158, + "step": 31017 + }, + { + "epoch": 5.64, + "learning_rate": 1.3089916272297053e-06, + "loss": 0.0675, + "step": 31018 + }, + { + "epoch": 5.64, + "learning_rate": 1.308937022206043e-06, + "loss": 0.6821, + "step": 31019 + }, + { + "epoch": 5.64, + "learning_rate": 1.308882417182381e-06, + "loss": 0.252, + "step": 31020 + }, + { + "epoch": 5.64, + "learning_rate": 1.3088278121587187e-06, + "loss": 0.5204, + "step": 31021 + }, + { + "epoch": 5.64, + "learning_rate": 1.3087732071350564e-06, + "loss": 0.0084, + "step": 31022 + }, + { + "epoch": 5.64, + "learning_rate": 1.3087186021113941e-06, + "loss": 0.3673, + "step": 31023 + }, + { + "epoch": 5.64, + "learning_rate": 1.308663997087732e-06, + "loss": 0.2857, + "step": 31024 + }, + { + "epoch": 5.64, + "learning_rate": 1.30860939206407e-06, + "loss": 1.2608, + "step": 31025 + }, + { + "epoch": 5.64, + "learning_rate": 1.3085547870404078e-06, + "loss": 1.3672, + "step": 31026 + }, + { + "epoch": 5.64, + "learning_rate": 1.3085001820167457e-06, + "loss": 1.1645, + "step": 31027 + }, + { + "epoch": 5.64, + "learning_rate": 1.3084455769930834e-06, + "loss": 0.9691, + "step": 31028 + }, + { + "epoch": 5.64, + "learning_rate": 1.3083909719694212e-06, + "loss": 0.2317, + "step": 31029 + }, + { + "epoch": 5.64, + "learning_rate": 1.308336366945759e-06, + "loss": 0.8252, + "step": 31030 + }, + { + "epoch": 5.64, + "learning_rate": 1.3082817619220968e-06, + "loss": 0.1153, + "step": 31031 + }, + { + "epoch": 5.64, + "learning_rate": 1.3082271568984348e-06, + "loss": 0.0835, + "step": 31032 + }, + { + "epoch": 5.64, + "learning_rate": 1.3081725518747725e-06, + "loss": 0.0068, + "step": 31033 + }, + { + "epoch": 5.64, + "learning_rate": 1.3081179468511105e-06, + "loss": 0.4643, + "step": 31034 + }, + { + "epoch": 5.64, + "learning_rate": 1.3080633418274482e-06, + "loss": 0.3203, + "step": 31035 + }, + { + "epoch": 5.64, + "learning_rate": 1.308008736803786e-06, + "loss": 0.7887, + "step": 31036 + }, + { + "epoch": 5.64, + "learning_rate": 1.3079541317801237e-06, + "loss": 2.2996, + "step": 31037 + }, + { + "epoch": 5.64, + "learning_rate": 1.3078995267564616e-06, + "loss": 1.4156, + "step": 31038 + }, + { + "epoch": 5.64, + "learning_rate": 1.3078449217327995e-06, + "loss": 1.2073, + "step": 31039 + }, + { + "epoch": 5.64, + "learning_rate": 1.3077903167091373e-06, + "loss": 0.2652, + "step": 31040 + }, + { + "epoch": 5.64, + "learning_rate": 1.3077357116854752e-06, + "loss": 0.2422, + "step": 31041 + }, + { + "epoch": 5.65, + "learning_rate": 1.307681106661813e-06, + "loss": 0.1803, + "step": 31042 + }, + { + "epoch": 5.65, + "learning_rate": 1.3076265016381507e-06, + "loss": 1.6136, + "step": 31043 + }, + { + "epoch": 5.65, + "learning_rate": 1.3075718966144884e-06, + "loss": 0.0045, + "step": 31044 + }, + { + "epoch": 5.65, + "learning_rate": 1.3075172915908264e-06, + "loss": 1.1763, + "step": 31045 + }, + { + "epoch": 5.65, + "learning_rate": 1.3074626865671643e-06, + "loss": 1.9249, + "step": 31046 + }, + { + "epoch": 5.65, + "learning_rate": 1.307408081543502e-06, + "loss": 0.2804, + "step": 31047 + }, + { + "epoch": 5.65, + "learning_rate": 1.30735347651984e-06, + "loss": 0.1402, + "step": 31048 + }, + { + "epoch": 5.65, + "learning_rate": 1.3072988714961777e-06, + "loss": 1.0242, + "step": 31049 + }, + { + "epoch": 5.65, + "learning_rate": 1.3072442664725154e-06, + "loss": 0.0662, + "step": 31050 + }, + { + "epoch": 5.65, + "learning_rate": 1.3071896614488532e-06, + "loss": 0.2946, + "step": 31051 + }, + { + "epoch": 5.65, + "learning_rate": 1.3071350564251911e-06, + "loss": 0.0052, + "step": 31052 + }, + { + "epoch": 5.65, + "learning_rate": 1.307080451401529e-06, + "loss": 0.0352, + "step": 31053 + }, + { + "epoch": 5.65, + "learning_rate": 1.3070258463778668e-06, + "loss": 2.975, + "step": 31054 + }, + { + "epoch": 5.65, + "learning_rate": 1.3069712413542045e-06, + "loss": 0.6256, + "step": 31055 + }, + { + "epoch": 5.65, + "learning_rate": 1.3069166363305425e-06, + "loss": 0.104, + "step": 31056 + }, + { + "epoch": 5.65, + "learning_rate": 1.3068620313068802e-06, + "loss": 0.4828, + "step": 31057 + }, + { + "epoch": 5.65, + "learning_rate": 1.306807426283218e-06, + "loss": 0.0352, + "step": 31058 + }, + { + "epoch": 5.65, + "learning_rate": 1.3067528212595559e-06, + "loss": 0.0077, + "step": 31059 + }, + { + "epoch": 5.65, + "learning_rate": 1.3066982162358938e-06, + "loss": 0.1595, + "step": 31060 + }, + { + "epoch": 5.65, + "learning_rate": 1.3066436112122315e-06, + "loss": 0.4301, + "step": 31061 + }, + { + "epoch": 5.65, + "learning_rate": 1.3065890061885693e-06, + "loss": 0.266, + "step": 31062 + }, + { + "epoch": 5.65, + "learning_rate": 1.3065344011649072e-06, + "loss": 0.0224, + "step": 31063 + }, + { + "epoch": 5.65, + "learning_rate": 1.306479796141245e-06, + "loss": 0.358, + "step": 31064 + }, + { + "epoch": 5.65, + "learning_rate": 1.306425191117583e-06, + "loss": 2.3222, + "step": 31065 + }, + { + "epoch": 5.65, + "learning_rate": 1.3063705860939208e-06, + "loss": 0.2003, + "step": 31066 + }, + { + "epoch": 5.65, + "learning_rate": 1.3063159810702586e-06, + "loss": 0.0093, + "step": 31067 + }, + { + "epoch": 5.65, + "learning_rate": 1.3062613760465963e-06, + "loss": 0.0856, + "step": 31068 + }, + { + "epoch": 5.65, + "learning_rate": 1.306206771022934e-06, + "loss": 0.0716, + "step": 31069 + }, + { + "epoch": 5.65, + "learning_rate": 1.306152165999272e-06, + "loss": 0.0067, + "step": 31070 + }, + { + "epoch": 5.65, + "learning_rate": 1.3060975609756097e-06, + "loss": 0.0576, + "step": 31071 + }, + { + "epoch": 5.65, + "learning_rate": 1.3060429559519477e-06, + "loss": 0.0178, + "step": 31072 + }, + { + "epoch": 5.65, + "learning_rate": 1.3059883509282856e-06, + "loss": 0.1804, + "step": 31073 + }, + { + "epoch": 5.65, + "learning_rate": 1.3059337459046233e-06, + "loss": 2.0721, + "step": 31074 + }, + { + "epoch": 5.65, + "learning_rate": 1.305879140880961e-06, + "loss": 1.0646, + "step": 31075 + }, + { + "epoch": 5.65, + "learning_rate": 1.3058245358572988e-06, + "loss": 0.7536, + "step": 31076 + }, + { + "epoch": 5.65, + "learning_rate": 1.3057699308336367e-06, + "loss": 0.761, + "step": 31077 + }, + { + "epoch": 5.65, + "learning_rate": 1.3057153258099745e-06, + "loss": 0.8427, + "step": 31078 + }, + { + "epoch": 5.65, + "learning_rate": 1.3056607207863124e-06, + "loss": 0.1667, + "step": 31079 + }, + { + "epoch": 5.65, + "learning_rate": 1.3056061157626504e-06, + "loss": 0.0662, + "step": 31080 + }, + { + "epoch": 5.65, + "learning_rate": 1.305551510738988e-06, + "loss": 0.2842, + "step": 31081 + }, + { + "epoch": 5.65, + "learning_rate": 1.3054969057153258e-06, + "loss": 2.0645, + "step": 31082 + }, + { + "epoch": 5.65, + "learning_rate": 1.3054423006916635e-06, + "loss": 1.0036, + "step": 31083 + }, + { + "epoch": 5.65, + "learning_rate": 1.3053876956680015e-06, + "loss": 0.398, + "step": 31084 + }, + { + "epoch": 5.65, + "learning_rate": 1.3053330906443392e-06, + "loss": 1.4402, + "step": 31085 + }, + { + "epoch": 5.65, + "learning_rate": 1.3052784856206772e-06, + "loss": 0.0525, + "step": 31086 + }, + { + "epoch": 5.65, + "learning_rate": 1.3052238805970151e-06, + "loss": 0.4177, + "step": 31087 + }, + { + "epoch": 5.65, + "learning_rate": 1.3051692755733528e-06, + "loss": 0.1544, + "step": 31088 + }, + { + "epoch": 5.65, + "learning_rate": 1.3051146705496906e-06, + "loss": 1.1651, + "step": 31089 + }, + { + "epoch": 5.65, + "learning_rate": 1.3050600655260283e-06, + "loss": 0.2909, + "step": 31090 + }, + { + "epoch": 5.65, + "learning_rate": 1.3050054605023662e-06, + "loss": 0.0098, + "step": 31091 + }, + { + "epoch": 5.65, + "learning_rate": 1.304950855478704e-06, + "loss": 0.1662, + "step": 31092 + }, + { + "epoch": 5.65, + "learning_rate": 1.304896250455042e-06, + "loss": 1.3283, + "step": 31093 + }, + { + "epoch": 5.65, + "learning_rate": 1.3048416454313799e-06, + "loss": 0.0999, + "step": 31094 + }, + { + "epoch": 5.65, + "learning_rate": 1.3047870404077176e-06, + "loss": 0.909, + "step": 31095 + }, + { + "epoch": 5.65, + "learning_rate": 1.3047324353840553e-06, + "loss": 0.0442, + "step": 31096 + }, + { + "epoch": 5.66, + "learning_rate": 1.304677830360393e-06, + "loss": 0.3516, + "step": 31097 + }, + { + "epoch": 5.66, + "learning_rate": 1.304623225336731e-06, + "loss": 0.2775, + "step": 31098 + }, + { + "epoch": 5.66, + "learning_rate": 1.3045686203130687e-06, + "loss": 0.6707, + "step": 31099 + }, + { + "epoch": 5.66, + "learning_rate": 1.3045140152894067e-06, + "loss": 0.7352, + "step": 31100 + }, + { + "epoch": 5.66, + "learning_rate": 1.3044594102657446e-06, + "loss": 0.2934, + "step": 31101 + }, + { + "epoch": 5.66, + "learning_rate": 1.3044048052420824e-06, + "loss": 0.3535, + "step": 31102 + }, + { + "epoch": 5.66, + "learning_rate": 1.30435020021842e-06, + "loss": 0.0793, + "step": 31103 + }, + { + "epoch": 5.66, + "learning_rate": 1.3042955951947578e-06, + "loss": 0.0052, + "step": 31104 + }, + { + "epoch": 5.66, + "learning_rate": 1.3042409901710958e-06, + "loss": 0.5382, + "step": 31105 + }, + { + "epoch": 5.66, + "learning_rate": 1.3041863851474337e-06, + "loss": 0.7757, + "step": 31106 + }, + { + "epoch": 5.66, + "learning_rate": 1.3041317801237714e-06, + "loss": 0.4135, + "step": 31107 + }, + { + "epoch": 5.66, + "learning_rate": 1.3040771751001094e-06, + "loss": 0.5113, + "step": 31108 + }, + { + "epoch": 5.66, + "learning_rate": 1.3040225700764471e-06, + "loss": 0.1458, + "step": 31109 + }, + { + "epoch": 5.66, + "learning_rate": 1.3039679650527848e-06, + "loss": 0.0939, + "step": 31110 + }, + { + "epoch": 5.66, + "learning_rate": 1.3039133600291226e-06, + "loss": 0.5447, + "step": 31111 + }, + { + "epoch": 5.66, + "learning_rate": 1.3038587550054605e-06, + "loss": 1.2228, + "step": 31112 + }, + { + "epoch": 5.66, + "learning_rate": 1.3038041499817985e-06, + "loss": 0.8355, + "step": 31113 + }, + { + "epoch": 5.66, + "learning_rate": 1.3037495449581362e-06, + "loss": 0.0081, + "step": 31114 + }, + { + "epoch": 5.66, + "learning_rate": 1.3036949399344741e-06, + "loss": 0.3336, + "step": 31115 + }, + { + "epoch": 5.66, + "learning_rate": 1.3036403349108119e-06, + "loss": 0.8653, + "step": 31116 + }, + { + "epoch": 5.66, + "learning_rate": 1.3035857298871496e-06, + "loss": 0.4099, + "step": 31117 + }, + { + "epoch": 5.66, + "learning_rate": 1.3035311248634873e-06, + "loss": 0.6178, + "step": 31118 + }, + { + "epoch": 5.66, + "learning_rate": 1.3034765198398253e-06, + "loss": 0.618, + "step": 31119 + }, + { + "epoch": 5.66, + "learning_rate": 1.3034219148161632e-06, + "loss": 0.7974, + "step": 31120 + }, + { + "epoch": 5.66, + "learning_rate": 1.303367309792501e-06, + "loss": 0.0964, + "step": 31121 + }, + { + "epoch": 5.66, + "learning_rate": 1.3033127047688389e-06, + "loss": 0.5063, + "step": 31122 + }, + { + "epoch": 5.66, + "learning_rate": 1.3032580997451766e-06, + "loss": 0.1654, + "step": 31123 + }, + { + "epoch": 5.66, + "learning_rate": 1.3032034947215144e-06, + "loss": 0.5039, + "step": 31124 + }, + { + "epoch": 5.66, + "learning_rate": 1.303148889697852e-06, + "loss": 0.0232, + "step": 31125 + }, + { + "epoch": 5.66, + "learning_rate": 1.30309428467419e-06, + "loss": 0.8295, + "step": 31126 + }, + { + "epoch": 5.66, + "learning_rate": 1.303039679650528e-06, + "loss": 1.5907, + "step": 31127 + }, + { + "epoch": 5.66, + "learning_rate": 1.3029850746268657e-06, + "loss": 1.142, + "step": 31128 + }, + { + "epoch": 5.66, + "learning_rate": 1.3029304696032034e-06, + "loss": 0.0215, + "step": 31129 + }, + { + "epoch": 5.66, + "learning_rate": 1.3028758645795414e-06, + "loss": 0.3199, + "step": 31130 + }, + { + "epoch": 5.66, + "learning_rate": 1.3028212595558791e-06, + "loss": 2.5628, + "step": 31131 + }, + { + "epoch": 5.66, + "learning_rate": 1.3027666545322168e-06, + "loss": 1.001, + "step": 31132 + }, + { + "epoch": 5.66, + "learning_rate": 1.3027120495085548e-06, + "loss": 1.6213, + "step": 31133 + }, + { + "epoch": 5.66, + "learning_rate": 1.3026574444848927e-06, + "loss": 0.1686, + "step": 31134 + }, + { + "epoch": 5.66, + "learning_rate": 1.3026028394612305e-06, + "loss": 0.3671, + "step": 31135 + }, + { + "epoch": 5.66, + "learning_rate": 1.3025482344375682e-06, + "loss": 1.6673, + "step": 31136 + }, + { + "epoch": 5.66, + "learning_rate": 1.3024936294139061e-06, + "loss": 0.2909, + "step": 31137 + }, + { + "epoch": 5.66, + "learning_rate": 1.3024390243902439e-06, + "loss": 3.7591, + "step": 31138 + }, + { + "epoch": 5.66, + "learning_rate": 1.3023844193665816e-06, + "loss": 0.6743, + "step": 31139 + }, + { + "epoch": 5.66, + "learning_rate": 1.3023298143429195e-06, + "loss": 0.0088, + "step": 31140 + }, + { + "epoch": 5.66, + "learning_rate": 1.3022752093192575e-06, + "loss": 0.182, + "step": 31141 + }, + { + "epoch": 5.66, + "learning_rate": 1.3022206042955952e-06, + "loss": 0.1187, + "step": 31142 + }, + { + "epoch": 5.66, + "learning_rate": 1.302165999271933e-06, + "loss": 0.7957, + "step": 31143 + }, + { + "epoch": 5.66, + "learning_rate": 1.302111394248271e-06, + "loss": 0.0873, + "step": 31144 + }, + { + "epoch": 5.66, + "learning_rate": 1.3020567892246086e-06, + "loss": 0.7465, + "step": 31145 + }, + { + "epoch": 5.66, + "learning_rate": 1.3020021842009466e-06, + "loss": 1.1554, + "step": 31146 + }, + { + "epoch": 5.66, + "learning_rate": 1.3019475791772845e-06, + "loss": 0.2136, + "step": 31147 + }, + { + "epoch": 5.66, + "learning_rate": 1.3018929741536222e-06, + "loss": 1.4828, + "step": 31148 + }, + { + "epoch": 5.66, + "learning_rate": 1.30183836912996e-06, + "loss": 0.5718, + "step": 31149 + }, + { + "epoch": 5.66, + "learning_rate": 1.3017837641062977e-06, + "loss": 1.4199, + "step": 31150 + }, + { + "epoch": 5.66, + "learning_rate": 1.3017291590826357e-06, + "loss": 1.9517, + "step": 31151 + }, + { + "epoch": 5.67, + "learning_rate": 1.3016745540589734e-06, + "loss": 0.1346, + "step": 31152 + }, + { + "epoch": 5.67, + "learning_rate": 1.3016199490353113e-06, + "loss": 0.6311, + "step": 31153 + }, + { + "epoch": 5.67, + "learning_rate": 1.3015653440116493e-06, + "loss": 1.7668, + "step": 31154 + }, + { + "epoch": 5.67, + "learning_rate": 1.301510738987987e-06, + "loss": 0.2656, + "step": 31155 + }, + { + "epoch": 5.67, + "learning_rate": 1.3014561339643247e-06, + "loss": 0.8218, + "step": 31156 + }, + { + "epoch": 5.67, + "learning_rate": 1.3014015289406625e-06, + "loss": 0.1267, + "step": 31157 + }, + { + "epoch": 5.67, + "learning_rate": 1.3013469239170004e-06, + "loss": 0.5469, + "step": 31158 + }, + { + "epoch": 5.67, + "learning_rate": 1.3012923188933381e-06, + "loss": 0.748, + "step": 31159 + }, + { + "epoch": 5.67, + "learning_rate": 1.301237713869676e-06, + "loss": 0.0156, + "step": 31160 + }, + { + "epoch": 5.67, + "learning_rate": 1.301183108846014e-06, + "loss": 0.0242, + "step": 31161 + }, + { + "epoch": 5.67, + "learning_rate": 1.3011285038223518e-06, + "loss": 0.5252, + "step": 31162 + }, + { + "epoch": 5.67, + "learning_rate": 1.3010738987986895e-06, + "loss": 0.0119, + "step": 31163 + }, + { + "epoch": 5.67, + "learning_rate": 1.3010192937750272e-06, + "loss": 1.5354, + "step": 31164 + }, + { + "epoch": 5.67, + "learning_rate": 1.3009646887513652e-06, + "loss": 1.3823, + "step": 31165 + }, + { + "epoch": 5.67, + "learning_rate": 1.300910083727703e-06, + "loss": 0.1461, + "step": 31166 + }, + { + "epoch": 5.67, + "learning_rate": 1.3008554787040408e-06, + "loss": 0.1751, + "step": 31167 + }, + { + "epoch": 5.67, + "learning_rate": 1.3008008736803788e-06, + "loss": 3.2919, + "step": 31168 + }, + { + "epoch": 5.67, + "learning_rate": 1.3007462686567165e-06, + "loss": 0.3123, + "step": 31169 + }, + { + "epoch": 5.67, + "learning_rate": 1.3006916636330542e-06, + "loss": 0.291, + "step": 31170 + }, + { + "epoch": 5.67, + "learning_rate": 1.300637058609392e-06, + "loss": 1.089, + "step": 31171 + }, + { + "epoch": 5.67, + "learning_rate": 1.30058245358573e-06, + "loss": 0.4254, + "step": 31172 + }, + { + "epoch": 5.67, + "learning_rate": 1.3005278485620677e-06, + "loss": 0.0295, + "step": 31173 + }, + { + "epoch": 5.67, + "learning_rate": 1.3004732435384056e-06, + "loss": 0.5494, + "step": 31174 + }, + { + "epoch": 5.67, + "learning_rate": 1.3004186385147435e-06, + "loss": 0.2592, + "step": 31175 + }, + { + "epoch": 5.67, + "learning_rate": 1.3003640334910813e-06, + "loss": 0.1934, + "step": 31176 + }, + { + "epoch": 5.67, + "learning_rate": 1.300309428467419e-06, + "loss": 2.6129, + "step": 31177 + }, + { + "epoch": 5.67, + "learning_rate": 1.3002548234437567e-06, + "loss": 0.1, + "step": 31178 + }, + { + "epoch": 5.67, + "learning_rate": 1.3002002184200947e-06, + "loss": 0.4158, + "step": 31179 + }, + { + "epoch": 5.67, + "learning_rate": 1.3001456133964324e-06, + "loss": 0.7703, + "step": 31180 + }, + { + "epoch": 5.67, + "learning_rate": 1.3000910083727704e-06, + "loss": 0.7372, + "step": 31181 + }, + { + "epoch": 5.67, + "learning_rate": 1.3000364033491083e-06, + "loss": 0.3276, + "step": 31182 + }, + { + "epoch": 5.67, + "learning_rate": 1.299981798325446e-06, + "loss": 0.0965, + "step": 31183 + }, + { + "epoch": 5.67, + "learning_rate": 1.2999271933017838e-06, + "loss": 0.1566, + "step": 31184 + }, + { + "epoch": 5.67, + "learning_rate": 1.2998725882781215e-06, + "loss": 2.1171, + "step": 31185 + }, + { + "epoch": 5.67, + "learning_rate": 1.2998179832544594e-06, + "loss": 0.0944, + "step": 31186 + }, + { + "epoch": 5.67, + "learning_rate": 1.2997633782307972e-06, + "loss": 0.6609, + "step": 31187 + }, + { + "epoch": 5.67, + "learning_rate": 1.2997087732071351e-06, + "loss": 1.5173, + "step": 31188 + }, + { + "epoch": 5.67, + "learning_rate": 1.299654168183473e-06, + "loss": 0.7883, + "step": 31189 + }, + { + "epoch": 5.67, + "learning_rate": 1.2995995631598108e-06, + "loss": 0.5445, + "step": 31190 + }, + { + "epoch": 5.67, + "learning_rate": 1.2995449581361485e-06, + "loss": 0.0514, + "step": 31191 + }, + { + "epoch": 5.67, + "learning_rate": 1.2994903531124862e-06, + "loss": 1.0473, + "step": 31192 + }, + { + "epoch": 5.67, + "learning_rate": 1.2994357480888242e-06, + "loss": 0.7451, + "step": 31193 + }, + { + "epoch": 5.67, + "learning_rate": 1.2993811430651621e-06, + "loss": 0.3663, + "step": 31194 + }, + { + "epoch": 5.67, + "learning_rate": 1.2993265380414999e-06, + "loss": 1.1754, + "step": 31195 + }, + { + "epoch": 5.67, + "learning_rate": 1.2992719330178378e-06, + "loss": 0.9561, + "step": 31196 + }, + { + "epoch": 5.67, + "learning_rate": 1.2992173279941755e-06, + "loss": 1.5588, + "step": 31197 + }, + { + "epoch": 5.67, + "learning_rate": 1.2991627229705133e-06, + "loss": 0.3753, + "step": 31198 + }, + { + "epoch": 5.67, + "learning_rate": 1.299108117946851e-06, + "loss": 0.0105, + "step": 31199 + }, + { + "epoch": 5.67, + "learning_rate": 1.299053512923189e-06, + "loss": 0.9109, + "step": 31200 + }, + { + "epoch": 5.67, + "learning_rate": 1.2989989078995269e-06, + "loss": 0.11, + "step": 31201 + }, + { + "epoch": 5.67, + "learning_rate": 1.2989443028758646e-06, + "loss": 1.5294, + "step": 31202 + }, + { + "epoch": 5.67, + "learning_rate": 1.2988896978522026e-06, + "loss": 0.032, + "step": 31203 + }, + { + "epoch": 5.67, + "learning_rate": 1.2988350928285403e-06, + "loss": 0.1733, + "step": 31204 + }, + { + "epoch": 5.67, + "learning_rate": 1.298780487804878e-06, + "loss": 0.9686, + "step": 31205 + }, + { + "epoch": 5.67, + "learning_rate": 1.2987258827812158e-06, + "loss": 0.0154, + "step": 31206 + }, + { + "epoch": 5.68, + "learning_rate": 1.2986712777575537e-06, + "loss": 0.2608, + "step": 31207 + }, + { + "epoch": 5.68, + "learning_rate": 1.2986166727338916e-06, + "loss": 0.9978, + "step": 31208 + }, + { + "epoch": 5.68, + "learning_rate": 1.2985620677102294e-06, + "loss": 0.3714, + "step": 31209 + }, + { + "epoch": 5.68, + "learning_rate": 1.2985074626865671e-06, + "loss": 0.5719, + "step": 31210 + }, + { + "epoch": 5.68, + "learning_rate": 1.298452857662905e-06, + "loss": 0.1825, + "step": 31211 + }, + { + "epoch": 5.68, + "learning_rate": 1.2983982526392428e-06, + "loss": 0.0541, + "step": 31212 + }, + { + "epoch": 5.68, + "learning_rate": 1.2983436476155805e-06, + "loss": 1.2407, + "step": 31213 + }, + { + "epoch": 5.68, + "learning_rate": 1.2982890425919185e-06, + "loss": 0.9106, + "step": 31214 + }, + { + "epoch": 5.68, + "learning_rate": 1.2982344375682564e-06, + "loss": 2.8991, + "step": 31215 + }, + { + "epoch": 5.68, + "learning_rate": 1.2981798325445941e-06, + "loss": 2.9218, + "step": 31216 + }, + { + "epoch": 5.68, + "learning_rate": 1.2981252275209319e-06, + "loss": 0.0096, + "step": 31217 + }, + { + "epoch": 5.68, + "learning_rate": 1.2980706224972698e-06, + "loss": 0.6128, + "step": 31218 + }, + { + "epoch": 5.68, + "learning_rate": 1.2980160174736075e-06, + "loss": 0.0392, + "step": 31219 + }, + { + "epoch": 5.68, + "learning_rate": 1.2979614124499453e-06, + "loss": 0.0519, + "step": 31220 + }, + { + "epoch": 5.68, + "learning_rate": 1.2979068074262832e-06, + "loss": 2.4443, + "step": 31221 + }, + { + "epoch": 5.68, + "learning_rate": 1.2978522024026212e-06, + "loss": 1.4381, + "step": 31222 + }, + { + "epoch": 5.68, + "learning_rate": 1.2977975973789589e-06, + "loss": 1.3082, + "step": 31223 + }, + { + "epoch": 5.68, + "learning_rate": 1.2977429923552966e-06, + "loss": 1.9518, + "step": 31224 + }, + { + "epoch": 5.68, + "learning_rate": 1.2976883873316346e-06, + "loss": 0.0845, + "step": 31225 + }, + { + "epoch": 5.68, + "learning_rate": 1.2976337823079723e-06, + "loss": 0.0216, + "step": 31226 + }, + { + "epoch": 5.68, + "learning_rate": 1.29757917728431e-06, + "loss": 0.0109, + "step": 31227 + }, + { + "epoch": 5.68, + "learning_rate": 1.297524572260648e-06, + "loss": 0.0424, + "step": 31228 + }, + { + "epoch": 5.68, + "learning_rate": 1.297469967236986e-06, + "loss": 0.4472, + "step": 31229 + }, + { + "epoch": 5.68, + "learning_rate": 1.2974153622133237e-06, + "loss": 0.37, + "step": 31230 + }, + { + "epoch": 5.68, + "learning_rate": 1.2973607571896614e-06, + "loss": 2.4434, + "step": 31231 + }, + { + "epoch": 5.68, + "learning_rate": 1.2973061521659993e-06, + "loss": 0.2854, + "step": 31232 + }, + { + "epoch": 5.68, + "learning_rate": 1.297251547142337e-06, + "loss": 0.3172, + "step": 31233 + }, + { + "epoch": 5.68, + "learning_rate": 1.297196942118675e-06, + "loss": 0.7364, + "step": 31234 + }, + { + "epoch": 5.68, + "learning_rate": 1.297142337095013e-06, + "loss": 0.8121, + "step": 31235 + }, + { + "epoch": 5.68, + "learning_rate": 1.2970877320713507e-06, + "loss": 1.8142, + "step": 31236 + }, + { + "epoch": 5.68, + "learning_rate": 1.2970331270476884e-06, + "loss": 0.0828, + "step": 31237 + }, + { + "epoch": 5.68, + "learning_rate": 1.2969785220240261e-06, + "loss": 1.721, + "step": 31238 + }, + { + "epoch": 5.68, + "learning_rate": 1.296923917000364e-06, + "loss": 0.4611, + "step": 31239 + }, + { + "epoch": 5.68, + "learning_rate": 1.2968693119767018e-06, + "loss": 0.7901, + "step": 31240 + }, + { + "epoch": 5.68, + "learning_rate": 1.2968147069530398e-06, + "loss": 0.6932, + "step": 31241 + }, + { + "epoch": 5.68, + "learning_rate": 1.2967601019293777e-06, + "loss": 1.2823, + "step": 31242 + }, + { + "epoch": 5.68, + "learning_rate": 1.2967054969057154e-06, + "loss": 0.9002, + "step": 31243 + }, + { + "epoch": 5.68, + "learning_rate": 1.2966508918820532e-06, + "loss": 0.3973, + "step": 31244 + }, + { + "epoch": 5.68, + "learning_rate": 1.296596286858391e-06, + "loss": 0.0042, + "step": 31245 + }, + { + "epoch": 5.68, + "learning_rate": 1.2965416818347288e-06, + "loss": 0.1638, + "step": 31246 + }, + { + "epoch": 5.68, + "learning_rate": 1.2964870768110666e-06, + "loss": 1.2217, + "step": 31247 + }, + { + "epoch": 5.68, + "learning_rate": 1.2964324717874045e-06, + "loss": 0.4723, + "step": 31248 + }, + { + "epoch": 5.68, + "learning_rate": 1.2963778667637425e-06, + "loss": 0.2561, + "step": 31249 + }, + { + "epoch": 5.68, + "learning_rate": 1.2963232617400802e-06, + "loss": 1.0353, + "step": 31250 + }, + { + "epoch": 5.68, + "learning_rate": 1.296268656716418e-06, + "loss": 0.5503, + "step": 31251 + }, + { + "epoch": 5.68, + "learning_rate": 1.2962140516927557e-06, + "loss": 0.0251, + "step": 31252 + }, + { + "epoch": 5.68, + "learning_rate": 1.2961594466690936e-06, + "loss": 0.0408, + "step": 31253 + }, + { + "epoch": 5.68, + "learning_rate": 1.2961048416454313e-06, + "loss": 0.8712, + "step": 31254 + }, + { + "epoch": 5.68, + "learning_rate": 1.2960502366217693e-06, + "loss": 2.3835, + "step": 31255 + }, + { + "epoch": 5.68, + "learning_rate": 1.2959956315981072e-06, + "loss": 0.915, + "step": 31256 + }, + { + "epoch": 5.68, + "learning_rate": 1.295941026574445e-06, + "loss": 0.1959, + "step": 31257 + }, + { + "epoch": 5.68, + "learning_rate": 1.2958864215507827e-06, + "loss": 0.1177, + "step": 31258 + }, + { + "epoch": 5.68, + "learning_rate": 1.2958318165271204e-06, + "loss": 0.5079, + "step": 31259 + }, + { + "epoch": 5.68, + "learning_rate": 1.2957772115034584e-06, + "loss": 2.8836, + "step": 31260 + }, + { + "epoch": 5.68, + "learning_rate": 1.295722606479796e-06, + "loss": 0.0124, + "step": 31261 + }, + { + "epoch": 5.69, + "learning_rate": 1.295668001456134e-06, + "loss": 0.2674, + "step": 31262 + }, + { + "epoch": 5.69, + "learning_rate": 1.295613396432472e-06, + "loss": 0.4844, + "step": 31263 + }, + { + "epoch": 5.69, + "learning_rate": 1.2955587914088097e-06, + "loss": 0.0058, + "step": 31264 + }, + { + "epoch": 5.69, + "learning_rate": 1.2955041863851474e-06, + "loss": 0.397, + "step": 31265 + }, + { + "epoch": 5.69, + "learning_rate": 1.2954495813614852e-06, + "loss": 0.0078, + "step": 31266 + }, + { + "epoch": 5.69, + "learning_rate": 1.2953949763378231e-06, + "loss": 0.7842, + "step": 31267 + }, + { + "epoch": 5.69, + "learning_rate": 1.2953403713141608e-06, + "loss": 0.6089, + "step": 31268 + }, + { + "epoch": 5.69, + "learning_rate": 1.2952857662904988e-06, + "loss": 1.5703, + "step": 31269 + }, + { + "epoch": 5.69, + "learning_rate": 1.2952311612668367e-06, + "loss": 0.0047, + "step": 31270 + }, + { + "epoch": 5.69, + "learning_rate": 1.2951765562431745e-06, + "loss": 0.0868, + "step": 31271 + }, + { + "epoch": 5.69, + "learning_rate": 1.2951219512195122e-06, + "loss": 0.087, + "step": 31272 + }, + { + "epoch": 5.69, + "learning_rate": 1.29506734619585e-06, + "loss": 0.0879, + "step": 31273 + }, + { + "epoch": 5.69, + "learning_rate": 1.2950127411721879e-06, + "loss": 0.0262, + "step": 31274 + }, + { + "epoch": 5.69, + "learning_rate": 1.2949581361485258e-06, + "loss": 0.2779, + "step": 31275 + }, + { + "epoch": 5.69, + "learning_rate": 1.2949035311248635e-06, + "loss": 1.6456, + "step": 31276 + }, + { + "epoch": 5.69, + "learning_rate": 1.2948489261012015e-06, + "loss": 0.0087, + "step": 31277 + }, + { + "epoch": 5.69, + "learning_rate": 1.2947943210775392e-06, + "loss": 0.4829, + "step": 31278 + }, + { + "epoch": 5.69, + "learning_rate": 1.294739716053877e-06, + "loss": 0.1185, + "step": 31279 + }, + { + "epoch": 5.69, + "learning_rate": 1.2946851110302147e-06, + "loss": 0.0133, + "step": 31280 + }, + { + "epoch": 5.69, + "learning_rate": 1.2946305060065526e-06, + "loss": 0.5227, + "step": 31281 + }, + { + "epoch": 5.69, + "learning_rate": 1.2945759009828906e-06, + "loss": 0.6972, + "step": 31282 + }, + { + "epoch": 5.69, + "learning_rate": 1.2945212959592283e-06, + "loss": 2.2659, + "step": 31283 + }, + { + "epoch": 5.69, + "learning_rate": 1.294466690935566e-06, + "loss": 1.4803, + "step": 31284 + }, + { + "epoch": 5.69, + "learning_rate": 1.294412085911904e-06, + "loss": 0.6101, + "step": 31285 + }, + { + "epoch": 5.69, + "learning_rate": 1.2943574808882417e-06, + "loss": 0.0066, + "step": 31286 + }, + { + "epoch": 5.69, + "learning_rate": 1.2943028758645794e-06, + "loss": 0.0713, + "step": 31287 + }, + { + "epoch": 5.69, + "learning_rate": 1.2942482708409174e-06, + "loss": 0.7198, + "step": 31288 + }, + { + "epoch": 5.69, + "learning_rate": 1.2941936658172553e-06, + "loss": 0.2018, + "step": 31289 + }, + { + "epoch": 5.69, + "learning_rate": 1.294139060793593e-06, + "loss": 0.2162, + "step": 31290 + }, + { + "epoch": 5.69, + "learning_rate": 1.2940844557699308e-06, + "loss": 1.2083, + "step": 31291 + }, + { + "epoch": 5.69, + "learning_rate": 1.2940298507462687e-06, + "loss": 0.714, + "step": 31292 + }, + { + "epoch": 5.69, + "learning_rate": 1.2939752457226065e-06, + "loss": 1.4799, + "step": 31293 + }, + { + "epoch": 5.69, + "learning_rate": 1.2939206406989442e-06, + "loss": 0.1289, + "step": 31294 + }, + { + "epoch": 5.69, + "learning_rate": 1.2938660356752821e-06, + "loss": 1.3633, + "step": 31295 + }, + { + "epoch": 5.69, + "learning_rate": 1.29381143065162e-06, + "loss": 1.2235, + "step": 31296 + }, + { + "epoch": 5.69, + "learning_rate": 1.2937568256279578e-06, + "loss": 1.157, + "step": 31297 + }, + { + "epoch": 5.69, + "learning_rate": 1.2937022206042955e-06, + "loss": 0.3837, + "step": 31298 + }, + { + "epoch": 5.69, + "learning_rate": 1.2936476155806335e-06, + "loss": 1.6117, + "step": 31299 + }, + { + "epoch": 5.69, + "learning_rate": 1.2935930105569712e-06, + "loss": 0.4343, + "step": 31300 + }, + { + "epoch": 5.69, + "learning_rate": 1.293538405533309e-06, + "loss": 0.0037, + "step": 31301 + }, + { + "epoch": 5.69, + "learning_rate": 1.2934838005096469e-06, + "loss": 1.2434, + "step": 31302 + }, + { + "epoch": 5.69, + "learning_rate": 1.2934291954859848e-06, + "loss": 1.3789, + "step": 31303 + }, + { + "epoch": 5.69, + "learning_rate": 1.2933745904623226e-06, + "loss": 0.4025, + "step": 31304 + }, + { + "epoch": 5.69, + "learning_rate": 1.2933199854386603e-06, + "loss": 0.1717, + "step": 31305 + }, + { + "epoch": 5.69, + "learning_rate": 1.2932653804149982e-06, + "loss": 0.2193, + "step": 31306 + }, + { + "epoch": 5.69, + "learning_rate": 1.293210775391336e-06, + "loss": 0.2408, + "step": 31307 + }, + { + "epoch": 5.69, + "learning_rate": 1.2931561703676737e-06, + "loss": 0.0443, + "step": 31308 + }, + { + "epoch": 5.69, + "learning_rate": 1.2931015653440116e-06, + "loss": 0.0957, + "step": 31309 + }, + { + "epoch": 5.69, + "learning_rate": 1.2930469603203496e-06, + "loss": 1.7338, + "step": 31310 + }, + { + "epoch": 5.69, + "learning_rate": 1.2929923552966873e-06, + "loss": 0.6078, + "step": 31311 + }, + { + "epoch": 5.69, + "learning_rate": 1.292937750273025e-06, + "loss": 0.152, + "step": 31312 + }, + { + "epoch": 5.69, + "learning_rate": 1.292883145249363e-06, + "loss": 0.2185, + "step": 31313 + }, + { + "epoch": 5.69, + "learning_rate": 1.2928285402257007e-06, + "loss": 0.0098, + "step": 31314 + }, + { + "epoch": 5.69, + "learning_rate": 1.2927739352020387e-06, + "loss": 0.245, + "step": 31315 + }, + { + "epoch": 5.69, + "learning_rate": 1.2927193301783766e-06, + "loss": 0.7567, + "step": 31316 + }, + { + "epoch": 5.7, + "learning_rate": 1.2926647251547143e-06, + "loss": 2.0639, + "step": 31317 + }, + { + "epoch": 5.7, + "learning_rate": 1.292610120131052e-06, + "loss": 0.0122, + "step": 31318 + }, + { + "epoch": 5.7, + "learning_rate": 1.2925555151073898e-06, + "loss": 0.4988, + "step": 31319 + }, + { + "epoch": 5.7, + "learning_rate": 1.2925009100837278e-06, + "loss": 0.2716, + "step": 31320 + }, + { + "epoch": 5.7, + "learning_rate": 1.2924463050600655e-06, + "loss": 0.0088, + "step": 31321 + }, + { + "epoch": 5.7, + "learning_rate": 1.2923917000364034e-06, + "loss": 0.069, + "step": 31322 + }, + { + "epoch": 5.7, + "learning_rate": 1.2923370950127414e-06, + "loss": 0.0191, + "step": 31323 + }, + { + "epoch": 5.7, + "learning_rate": 1.2922824899890791e-06, + "loss": 0.0166, + "step": 31324 + }, + { + "epoch": 5.7, + "learning_rate": 1.2922278849654168e-06, + "loss": 1.1624, + "step": 31325 + }, + { + "epoch": 5.7, + "learning_rate": 1.2921732799417546e-06, + "loss": 0.1544, + "step": 31326 + }, + { + "epoch": 5.7, + "learning_rate": 1.2921186749180925e-06, + "loss": 0.7682, + "step": 31327 + }, + { + "epoch": 5.7, + "learning_rate": 1.2920640698944302e-06, + "loss": 0.0284, + "step": 31328 + }, + { + "epoch": 5.7, + "learning_rate": 1.2920094648707682e-06, + "loss": 0.4168, + "step": 31329 + }, + { + "epoch": 5.7, + "learning_rate": 1.2919548598471061e-06, + "loss": 0.0113, + "step": 31330 + }, + { + "epoch": 5.7, + "learning_rate": 1.2919002548234439e-06, + "loss": 1.3583, + "step": 31331 + }, + { + "epoch": 5.7, + "learning_rate": 1.2918456497997816e-06, + "loss": 0.0188, + "step": 31332 + }, + { + "epoch": 5.7, + "learning_rate": 1.2917910447761193e-06, + "loss": 1.5157, + "step": 31333 + }, + { + "epoch": 5.7, + "learning_rate": 1.2917364397524573e-06, + "loss": 0.5328, + "step": 31334 + }, + { + "epoch": 5.7, + "learning_rate": 1.291681834728795e-06, + "loss": 1.2512, + "step": 31335 + }, + { + "epoch": 5.7, + "learning_rate": 1.291627229705133e-06, + "loss": 1.3172, + "step": 31336 + }, + { + "epoch": 5.7, + "learning_rate": 1.2915726246814709e-06, + "loss": 0.0186, + "step": 31337 + }, + { + "epoch": 5.7, + "learning_rate": 1.2915180196578086e-06, + "loss": 0.1427, + "step": 31338 + }, + { + "epoch": 5.7, + "learning_rate": 1.2914634146341464e-06, + "loss": 0.7641, + "step": 31339 + }, + { + "epoch": 5.7, + "learning_rate": 1.291408809610484e-06, + "loss": 0.2474, + "step": 31340 + }, + { + "epoch": 5.7, + "learning_rate": 1.291354204586822e-06, + "loss": 0.0142, + "step": 31341 + }, + { + "epoch": 5.7, + "learning_rate": 1.2912995995631598e-06, + "loss": 0.0062, + "step": 31342 + }, + { + "epoch": 5.7, + "learning_rate": 1.2912449945394977e-06, + "loss": 0.1783, + "step": 31343 + }, + { + "epoch": 5.7, + "learning_rate": 1.2911903895158356e-06, + "loss": 0.1042, + "step": 31344 + }, + { + "epoch": 5.7, + "learning_rate": 1.2911357844921734e-06, + "loss": 0.5584, + "step": 31345 + }, + { + "epoch": 5.7, + "learning_rate": 1.2910811794685111e-06, + "loss": 0.1369, + "step": 31346 + }, + { + "epoch": 5.7, + "learning_rate": 1.2910265744448488e-06, + "loss": 0.0171, + "step": 31347 + }, + { + "epoch": 5.7, + "learning_rate": 1.2909719694211868e-06, + "loss": 0.4828, + "step": 31348 + }, + { + "epoch": 5.7, + "learning_rate": 1.2909173643975245e-06, + "loss": 0.0151, + "step": 31349 + }, + { + "epoch": 5.7, + "learning_rate": 1.2908627593738625e-06, + "loss": 0.0812, + "step": 31350 + }, + { + "epoch": 5.7, + "learning_rate": 1.2908081543502004e-06, + "loss": 0.1938, + "step": 31351 + }, + { + "epoch": 5.7, + "learning_rate": 1.2907535493265381e-06, + "loss": 0.0933, + "step": 31352 + }, + { + "epoch": 5.7, + "learning_rate": 1.2906989443028759e-06, + "loss": 2.4409, + "step": 31353 + }, + { + "epoch": 5.7, + "learning_rate": 1.2906443392792136e-06, + "loss": 1.1766, + "step": 31354 + }, + { + "epoch": 5.7, + "learning_rate": 1.2905897342555515e-06, + "loss": 0.0529, + "step": 31355 + }, + { + "epoch": 5.7, + "learning_rate": 1.2905351292318893e-06, + "loss": 0.012, + "step": 31356 + }, + { + "epoch": 5.7, + "learning_rate": 1.2904805242082272e-06, + "loss": 0.0043, + "step": 31357 + }, + { + "epoch": 5.7, + "learning_rate": 1.2904259191845652e-06, + "loss": 0.8025, + "step": 31358 + }, + { + "epoch": 5.7, + "learning_rate": 1.2903713141609029e-06, + "loss": 1.2522, + "step": 31359 + }, + { + "epoch": 5.7, + "learning_rate": 1.2903167091372406e-06, + "loss": 0.5146, + "step": 31360 + }, + { + "epoch": 5.7, + "learning_rate": 1.2902621041135784e-06, + "loss": 0.7837, + "step": 31361 + }, + { + "epoch": 5.7, + "learning_rate": 1.2902074990899163e-06, + "loss": 0.0148, + "step": 31362 + }, + { + "epoch": 5.7, + "learning_rate": 1.2901528940662542e-06, + "loss": 0.8869, + "step": 31363 + }, + { + "epoch": 5.7, + "learning_rate": 1.290098289042592e-06, + "loss": 0.0099, + "step": 31364 + }, + { + "epoch": 5.7, + "learning_rate": 1.2900436840189297e-06, + "loss": 0.6533, + "step": 31365 + }, + { + "epoch": 5.7, + "learning_rate": 1.2899890789952676e-06, + "loss": 0.9248, + "step": 31366 + }, + { + "epoch": 5.7, + "learning_rate": 1.2899344739716054e-06, + "loss": 0.008, + "step": 31367 + }, + { + "epoch": 5.7, + "learning_rate": 1.2898798689479431e-06, + "loss": 0.0128, + "step": 31368 + }, + { + "epoch": 5.7, + "learning_rate": 1.289825263924281e-06, + "loss": 0.2945, + "step": 31369 + }, + { + "epoch": 5.7, + "learning_rate": 1.289770658900619e-06, + "loss": 0.011, + "step": 31370 + }, + { + "epoch": 5.7, + "learning_rate": 1.2897160538769567e-06, + "loss": 1.3506, + "step": 31371 + }, + { + "epoch": 5.71, + "learning_rate": 1.2896614488532945e-06, + "loss": 0.3997, + "step": 31372 + }, + { + "epoch": 5.71, + "learning_rate": 1.2896068438296324e-06, + "loss": 2.8071, + "step": 31373 + }, + { + "epoch": 5.71, + "learning_rate": 1.2895522388059701e-06, + "loss": 0.7837, + "step": 31374 + }, + { + "epoch": 5.71, + "learning_rate": 1.2894976337823079e-06, + "loss": 0.0078, + "step": 31375 + }, + { + "epoch": 5.71, + "learning_rate": 1.2894430287586458e-06, + "loss": 0.9488, + "step": 31376 + }, + { + "epoch": 5.71, + "learning_rate": 1.2893884237349838e-06, + "loss": 0.2867, + "step": 31377 + }, + { + "epoch": 5.71, + "learning_rate": 1.2893338187113215e-06, + "loss": 1.2929, + "step": 31378 + }, + { + "epoch": 5.71, + "learning_rate": 1.2892792136876592e-06, + "loss": 0.3454, + "step": 31379 + }, + { + "epoch": 5.71, + "learning_rate": 1.2892246086639972e-06, + "loss": 0.0154, + "step": 31380 + }, + { + "epoch": 5.71, + "learning_rate": 1.2891700036403349e-06, + "loss": 0.3329, + "step": 31381 + }, + { + "epoch": 5.71, + "learning_rate": 1.2891153986166726e-06, + "loss": 0.0072, + "step": 31382 + }, + { + "epoch": 5.71, + "learning_rate": 1.2890607935930106e-06, + "loss": 0.1314, + "step": 31383 + }, + { + "epoch": 5.71, + "learning_rate": 1.2890061885693485e-06, + "loss": 1.705, + "step": 31384 + }, + { + "epoch": 5.71, + "learning_rate": 1.2889515835456862e-06, + "loss": 0.8428, + "step": 31385 + }, + { + "epoch": 5.71, + "learning_rate": 1.288896978522024e-06, + "loss": 0.1434, + "step": 31386 + }, + { + "epoch": 5.71, + "learning_rate": 1.288842373498362e-06, + "loss": 1.6487, + "step": 31387 + }, + { + "epoch": 5.71, + "learning_rate": 1.2887877684746996e-06, + "loss": 0.208, + "step": 31388 + }, + { + "epoch": 5.71, + "learning_rate": 1.2887331634510374e-06, + "loss": 0.0266, + "step": 31389 + }, + { + "epoch": 5.71, + "learning_rate": 1.2886785584273753e-06, + "loss": 0.0829, + "step": 31390 + }, + { + "epoch": 5.71, + "learning_rate": 1.2886239534037133e-06, + "loss": 0.395, + "step": 31391 + }, + { + "epoch": 5.71, + "learning_rate": 1.288569348380051e-06, + "loss": 0.4512, + "step": 31392 + }, + { + "epoch": 5.71, + "learning_rate": 1.2885147433563887e-06, + "loss": 0.0075, + "step": 31393 + }, + { + "epoch": 5.71, + "learning_rate": 1.2884601383327267e-06, + "loss": 0.412, + "step": 31394 + }, + { + "epoch": 5.71, + "learning_rate": 1.2884055333090644e-06, + "loss": 1.0135, + "step": 31395 + }, + { + "epoch": 5.71, + "learning_rate": 1.2883509282854021e-06, + "loss": 0.4964, + "step": 31396 + }, + { + "epoch": 5.71, + "learning_rate": 1.28829632326174e-06, + "loss": 1.9081, + "step": 31397 + }, + { + "epoch": 5.71, + "learning_rate": 1.288241718238078e-06, + "loss": 0.9158, + "step": 31398 + }, + { + "epoch": 5.71, + "learning_rate": 1.2881871132144158e-06, + "loss": 0.0091, + "step": 31399 + }, + { + "epoch": 5.71, + "learning_rate": 1.2881325081907535e-06, + "loss": 0.2012, + "step": 31400 + }, + { + "epoch": 5.71, + "learning_rate": 1.2880779031670914e-06, + "loss": 0.4214, + "step": 31401 + }, + { + "epoch": 5.71, + "learning_rate": 1.2880232981434292e-06, + "loss": 0.4045, + "step": 31402 + }, + { + "epoch": 5.71, + "learning_rate": 1.287968693119767e-06, + "loss": 0.2271, + "step": 31403 + }, + { + "epoch": 5.71, + "learning_rate": 1.287914088096105e-06, + "loss": 0.0049, + "step": 31404 + }, + { + "epoch": 5.71, + "learning_rate": 1.2878594830724428e-06, + "loss": 0.164, + "step": 31405 + }, + { + "epoch": 5.71, + "learning_rate": 1.2878048780487805e-06, + "loss": 5.9516, + "step": 31406 + }, + { + "epoch": 5.71, + "learning_rate": 1.2877502730251182e-06, + "loss": 0.0964, + "step": 31407 + }, + { + "epoch": 5.71, + "learning_rate": 1.2876956680014562e-06, + "loss": 0.2374, + "step": 31408 + }, + { + "epoch": 5.71, + "learning_rate": 1.287641062977794e-06, + "loss": 0.1068, + "step": 31409 + }, + { + "epoch": 5.71, + "learning_rate": 1.2875864579541319e-06, + "loss": 0.6789, + "step": 31410 + }, + { + "epoch": 5.71, + "learning_rate": 1.2875318529304698e-06, + "loss": 1.1461, + "step": 31411 + }, + { + "epoch": 5.71, + "learning_rate": 1.2874772479068075e-06, + "loss": 0.0355, + "step": 31412 + }, + { + "epoch": 5.71, + "learning_rate": 1.2874226428831453e-06, + "loss": 1.3353, + "step": 31413 + }, + { + "epoch": 5.71, + "learning_rate": 1.287368037859483e-06, + "loss": 0.3041, + "step": 31414 + }, + { + "epoch": 5.71, + "learning_rate": 1.287313432835821e-06, + "loss": 0.3905, + "step": 31415 + }, + { + "epoch": 5.71, + "learning_rate": 1.2872588278121587e-06, + "loss": 0.1827, + "step": 31416 + }, + { + "epoch": 5.71, + "learning_rate": 1.2872042227884966e-06, + "loss": 0.016, + "step": 31417 + }, + { + "epoch": 5.71, + "learning_rate": 1.2871496177648346e-06, + "loss": 0.068, + "step": 31418 + }, + { + "epoch": 5.71, + "learning_rate": 1.2870950127411723e-06, + "loss": 1.0557, + "step": 31419 + }, + { + "epoch": 5.71, + "learning_rate": 1.28704040771751e-06, + "loss": 0.96, + "step": 31420 + }, + { + "epoch": 5.71, + "learning_rate": 1.2869858026938478e-06, + "loss": 2.0192, + "step": 31421 + }, + { + "epoch": 5.71, + "learning_rate": 1.2869311976701857e-06, + "loss": 0.6038, + "step": 31422 + }, + { + "epoch": 5.71, + "learning_rate": 1.2868765926465234e-06, + "loss": 0.1171, + "step": 31423 + }, + { + "epoch": 5.71, + "learning_rate": 1.2868219876228614e-06, + "loss": 2.3866, + "step": 31424 + }, + { + "epoch": 5.71, + "learning_rate": 1.2867673825991993e-06, + "loss": 0.0476, + "step": 31425 + }, + { + "epoch": 5.71, + "learning_rate": 1.286712777575537e-06, + "loss": 0.4626, + "step": 31426 + }, + { + "epoch": 5.72, + "learning_rate": 1.2866581725518748e-06, + "loss": 1.2331, + "step": 31427 + }, + { + "epoch": 5.72, + "learning_rate": 1.2866035675282125e-06, + "loss": 0.7805, + "step": 31428 + }, + { + "epoch": 5.72, + "learning_rate": 1.2865489625045505e-06, + "loss": 1.2115, + "step": 31429 + }, + { + "epoch": 5.72, + "learning_rate": 1.2864943574808882e-06, + "loss": 1.2128, + "step": 31430 + }, + { + "epoch": 5.72, + "learning_rate": 1.2864397524572261e-06, + "loss": 0.7397, + "step": 31431 + }, + { + "epoch": 5.72, + "learning_rate": 1.286385147433564e-06, + "loss": 0.5609, + "step": 31432 + }, + { + "epoch": 5.72, + "learning_rate": 1.2863305424099018e-06, + "loss": 0.6605, + "step": 31433 + }, + { + "epoch": 5.72, + "learning_rate": 1.2862759373862395e-06, + "loss": 0.1954, + "step": 31434 + }, + { + "epoch": 5.72, + "learning_rate": 1.2862213323625773e-06, + "loss": 2.5519, + "step": 31435 + }, + { + "epoch": 5.72, + "learning_rate": 1.2861667273389152e-06, + "loss": 1.0328, + "step": 31436 + }, + { + "epoch": 5.72, + "learning_rate": 1.286112122315253e-06, + "loss": 1.7514, + "step": 31437 + }, + { + "epoch": 5.72, + "learning_rate": 1.2860575172915909e-06, + "loss": 1.1361, + "step": 31438 + }, + { + "epoch": 5.72, + "learning_rate": 1.2860029122679286e-06, + "loss": 0.2429, + "step": 31439 + }, + { + "epoch": 5.72, + "learning_rate": 1.2859483072442666e-06, + "loss": 0.0151, + "step": 31440 + }, + { + "epoch": 5.72, + "learning_rate": 1.2858937022206043e-06, + "loss": 1.4334, + "step": 31441 + }, + { + "epoch": 5.72, + "learning_rate": 1.285839097196942e-06, + "loss": 0.4239, + "step": 31442 + }, + { + "epoch": 5.72, + "learning_rate": 1.28578449217328e-06, + "loss": 0.3554, + "step": 31443 + }, + { + "epoch": 5.72, + "learning_rate": 1.285729887149618e-06, + "loss": 0.2924, + "step": 31444 + }, + { + "epoch": 5.72, + "learning_rate": 1.2856752821259556e-06, + "loss": 2.2664, + "step": 31445 + }, + { + "epoch": 5.72, + "learning_rate": 1.2856206771022934e-06, + "loss": 0.3227, + "step": 31446 + }, + { + "epoch": 5.72, + "learning_rate": 1.2855660720786313e-06, + "loss": 0.305, + "step": 31447 + }, + { + "epoch": 5.72, + "learning_rate": 1.285511467054969e-06, + "loss": 0.5074, + "step": 31448 + }, + { + "epoch": 5.72, + "learning_rate": 1.2854568620313068e-06, + "loss": 0.5943, + "step": 31449 + }, + { + "epoch": 5.72, + "learning_rate": 1.2854022570076447e-06, + "loss": 0.7739, + "step": 31450 + }, + { + "epoch": 5.72, + "learning_rate": 1.2853476519839827e-06, + "loss": 1.2348, + "step": 31451 + }, + { + "epoch": 5.72, + "learning_rate": 1.2852930469603204e-06, + "loss": 0.1496, + "step": 31452 + }, + { + "epoch": 5.72, + "learning_rate": 1.2852384419366581e-06, + "loss": 0.4007, + "step": 31453 + }, + { + "epoch": 5.72, + "learning_rate": 1.285183836912996e-06, + "loss": 1.6852, + "step": 31454 + }, + { + "epoch": 5.72, + "learning_rate": 1.2851292318893338e-06, + "loss": 0.9985, + "step": 31455 + }, + { + "epoch": 5.72, + "learning_rate": 1.2850746268656715e-06, + "loss": 0.2093, + "step": 31456 + }, + { + "epoch": 5.72, + "learning_rate": 1.2850200218420095e-06, + "loss": 0.0262, + "step": 31457 + }, + { + "epoch": 5.72, + "learning_rate": 1.2849654168183474e-06, + "loss": 0.328, + "step": 31458 + }, + { + "epoch": 5.72, + "learning_rate": 1.2849108117946852e-06, + "loss": 0.0112, + "step": 31459 + }, + { + "epoch": 5.72, + "learning_rate": 1.2848562067710229e-06, + "loss": 0.0183, + "step": 31460 + }, + { + "epoch": 5.72, + "learning_rate": 1.2848016017473608e-06, + "loss": 0.4109, + "step": 31461 + }, + { + "epoch": 5.72, + "learning_rate": 1.2847469967236986e-06, + "loss": 0.3937, + "step": 31462 + }, + { + "epoch": 5.72, + "learning_rate": 1.2846923917000363e-06, + "loss": 0.4256, + "step": 31463 + }, + { + "epoch": 5.72, + "learning_rate": 1.2846377866763742e-06, + "loss": 1.5388, + "step": 31464 + }, + { + "epoch": 5.72, + "learning_rate": 1.2845831816527122e-06, + "loss": 0.2341, + "step": 31465 + }, + { + "epoch": 5.72, + "learning_rate": 1.28452857662905e-06, + "loss": 0.7577, + "step": 31466 + }, + { + "epoch": 5.72, + "learning_rate": 1.2844739716053876e-06, + "loss": 0.0264, + "step": 31467 + }, + { + "epoch": 5.72, + "learning_rate": 1.2844193665817256e-06, + "loss": 0.3495, + "step": 31468 + }, + { + "epoch": 5.72, + "learning_rate": 1.2843647615580633e-06, + "loss": 0.4072, + "step": 31469 + }, + { + "epoch": 5.72, + "learning_rate": 1.284310156534401e-06, + "loss": 0.3577, + "step": 31470 + }, + { + "epoch": 5.72, + "learning_rate": 1.284255551510739e-06, + "loss": 2.7638, + "step": 31471 + }, + { + "epoch": 5.72, + "learning_rate": 1.284200946487077e-06, + "loss": 0.321, + "step": 31472 + }, + { + "epoch": 5.72, + "learning_rate": 1.2841463414634147e-06, + "loss": 1.662, + "step": 31473 + }, + { + "epoch": 5.72, + "learning_rate": 1.2840917364397524e-06, + "loss": 1.1779, + "step": 31474 + }, + { + "epoch": 5.72, + "learning_rate": 1.2840371314160903e-06, + "loss": 0.2715, + "step": 31475 + }, + { + "epoch": 5.72, + "learning_rate": 1.283982526392428e-06, + "loss": 0.0078, + "step": 31476 + }, + { + "epoch": 5.72, + "learning_rate": 1.2839279213687658e-06, + "loss": 0.3844, + "step": 31477 + }, + { + "epoch": 5.72, + "learning_rate": 1.2838733163451038e-06, + "loss": 1.051, + "step": 31478 + }, + { + "epoch": 5.72, + "learning_rate": 1.2838187113214417e-06, + "loss": 0.1013, + "step": 31479 + }, + { + "epoch": 5.72, + "learning_rate": 1.2837641062977794e-06, + "loss": 0.4219, + "step": 31480 + }, + { + "epoch": 5.72, + "learning_rate": 1.2837095012741172e-06, + "loss": 0.5737, + "step": 31481 + }, + { + "epoch": 5.73, + "learning_rate": 1.283654896250455e-06, + "loss": 0.019, + "step": 31482 + }, + { + "epoch": 5.73, + "learning_rate": 1.2836002912267928e-06, + "loss": 3.1438, + "step": 31483 + }, + { + "epoch": 5.73, + "learning_rate": 1.2835456862031308e-06, + "loss": 0.4455, + "step": 31484 + }, + { + "epoch": 5.73, + "learning_rate": 1.2834910811794687e-06, + "loss": 0.0237, + "step": 31485 + }, + { + "epoch": 5.73, + "learning_rate": 1.2834364761558065e-06, + "loss": 0.0172, + "step": 31486 + }, + { + "epoch": 5.73, + "learning_rate": 1.2833818711321442e-06, + "loss": 0.0083, + "step": 31487 + }, + { + "epoch": 5.73, + "learning_rate": 1.283327266108482e-06, + "loss": 0.3912, + "step": 31488 + }, + { + "epoch": 5.73, + "learning_rate": 1.2832726610848199e-06, + "loss": 0.2527, + "step": 31489 + }, + { + "epoch": 5.73, + "learning_rate": 1.2832180560611576e-06, + "loss": 0.1134, + "step": 31490 + }, + { + "epoch": 5.73, + "learning_rate": 1.2831634510374955e-06, + "loss": 0.0116, + "step": 31491 + }, + { + "epoch": 5.73, + "learning_rate": 1.2831088460138335e-06, + "loss": 1.1264, + "step": 31492 + }, + { + "epoch": 5.73, + "learning_rate": 1.2830542409901712e-06, + "loss": 0.3564, + "step": 31493 + }, + { + "epoch": 5.73, + "learning_rate": 1.282999635966509e-06, + "loss": 0.013, + "step": 31494 + }, + { + "epoch": 5.73, + "learning_rate": 1.2829450309428467e-06, + "loss": 0.0865, + "step": 31495 + }, + { + "epoch": 5.73, + "learning_rate": 1.2828904259191846e-06, + "loss": 1.4853, + "step": 31496 + }, + { + "epoch": 5.73, + "learning_rate": 1.2828358208955223e-06, + "loss": 0.7055, + "step": 31497 + }, + { + "epoch": 5.73, + "learning_rate": 1.2827812158718603e-06, + "loss": 0.4081, + "step": 31498 + }, + { + "epoch": 5.73, + "learning_rate": 1.2827266108481982e-06, + "loss": 0.0064, + "step": 31499 + }, + { + "epoch": 5.73, + "learning_rate": 1.282672005824536e-06, + "loss": 0.2964, + "step": 31500 + }, + { + "epoch": 5.73, + "learning_rate": 1.2826174008008737e-06, + "loss": 0.4726, + "step": 31501 + }, + { + "epoch": 5.73, + "learning_rate": 1.2825627957772114e-06, + "loss": 0.2999, + "step": 31502 + }, + { + "epoch": 5.73, + "learning_rate": 1.2825081907535494e-06, + "loss": 0.3374, + "step": 31503 + }, + { + "epoch": 5.73, + "learning_rate": 1.2824535857298871e-06, + "loss": 0.0078, + "step": 31504 + }, + { + "epoch": 5.73, + "learning_rate": 1.282398980706225e-06, + "loss": 0.3112, + "step": 31505 + }, + { + "epoch": 5.73, + "learning_rate": 1.282344375682563e-06, + "loss": 0.6871, + "step": 31506 + }, + { + "epoch": 5.73, + "learning_rate": 1.2822897706589007e-06, + "loss": 0.0248, + "step": 31507 + }, + { + "epoch": 5.73, + "learning_rate": 1.2822351656352385e-06, + "loss": 0.168, + "step": 31508 + }, + { + "epoch": 5.73, + "learning_rate": 1.2821805606115762e-06, + "loss": 0.0713, + "step": 31509 + }, + { + "epoch": 5.73, + "learning_rate": 1.2821259555879141e-06, + "loss": 0.6867, + "step": 31510 + }, + { + "epoch": 5.73, + "learning_rate": 1.2820713505642519e-06, + "loss": 0.3338, + "step": 31511 + }, + { + "epoch": 5.73, + "learning_rate": 1.2820167455405898e-06, + "loss": 0.0251, + "step": 31512 + }, + { + "epoch": 5.73, + "learning_rate": 1.2819621405169278e-06, + "loss": 0.4439, + "step": 31513 + }, + { + "epoch": 5.73, + "learning_rate": 1.2819075354932655e-06, + "loss": 0.0222, + "step": 31514 + }, + { + "epoch": 5.73, + "learning_rate": 1.2818529304696032e-06, + "loss": 1.1549, + "step": 31515 + }, + { + "epoch": 5.73, + "learning_rate": 1.281798325445941e-06, + "loss": 0.2224, + "step": 31516 + }, + { + "epoch": 5.73, + "learning_rate": 1.2817437204222789e-06, + "loss": 0.318, + "step": 31517 + }, + { + "epoch": 5.73, + "learning_rate": 1.2816891153986166e-06, + "loss": 0.1514, + "step": 31518 + }, + { + "epoch": 5.73, + "learning_rate": 1.2816345103749546e-06, + "loss": 1.6739, + "step": 31519 + }, + { + "epoch": 5.73, + "learning_rate": 1.2815799053512923e-06, + "loss": 2.1666, + "step": 31520 + }, + { + "epoch": 5.73, + "learning_rate": 1.2815253003276302e-06, + "loss": 0.7987, + "step": 31521 + }, + { + "epoch": 5.73, + "learning_rate": 1.281470695303968e-06, + "loss": 0.2288, + "step": 31522 + }, + { + "epoch": 5.73, + "learning_rate": 1.2814160902803057e-06, + "loss": 0.0098, + "step": 31523 + }, + { + "epoch": 5.73, + "learning_rate": 1.2813614852566436e-06, + "loss": 0.0524, + "step": 31524 + }, + { + "epoch": 5.73, + "learning_rate": 1.2813068802329814e-06, + "loss": 1.0604, + "step": 31525 + }, + { + "epoch": 5.73, + "learning_rate": 1.2812522752093193e-06, + "loss": 0.4018, + "step": 31526 + }, + { + "epoch": 5.73, + "learning_rate": 1.281197670185657e-06, + "loss": 0.4191, + "step": 31527 + }, + { + "epoch": 5.73, + "learning_rate": 1.281143065161995e-06, + "loss": 1.0648, + "step": 31528 + }, + { + "epoch": 5.73, + "learning_rate": 1.2810884601383327e-06, + "loss": 0.2858, + "step": 31529 + }, + { + "epoch": 5.73, + "learning_rate": 1.2810338551146705e-06, + "loss": 0.0163, + "step": 31530 + }, + { + "epoch": 5.73, + "learning_rate": 1.2809792500910084e-06, + "loss": 1.7398, + "step": 31531 + }, + { + "epoch": 5.73, + "learning_rate": 1.2809246450673463e-06, + "loss": 0.8947, + "step": 31532 + }, + { + "epoch": 5.73, + "learning_rate": 1.280870040043684e-06, + "loss": 0.2143, + "step": 31533 + }, + { + "epoch": 5.73, + "learning_rate": 1.2808154350200218e-06, + "loss": 0.2656, + "step": 31534 + }, + { + "epoch": 5.73, + "learning_rate": 1.2807608299963598e-06, + "loss": 0.015, + "step": 31535 + }, + { + "epoch": 5.73, + "learning_rate": 1.2807062249726975e-06, + "loss": 0.0149, + "step": 31536 + }, + { + "epoch": 5.74, + "learning_rate": 1.2806516199490352e-06, + "loss": 0.73, + "step": 31537 + }, + { + "epoch": 5.74, + "learning_rate": 1.2805970149253732e-06, + "loss": 0.1396, + "step": 31538 + }, + { + "epoch": 5.74, + "learning_rate": 1.280542409901711e-06, + "loss": 0.7207, + "step": 31539 + }, + { + "epoch": 5.74, + "learning_rate": 1.2804878048780488e-06, + "loss": 0.0181, + "step": 31540 + }, + { + "epoch": 5.74, + "learning_rate": 1.2804331998543866e-06, + "loss": 0.1848, + "step": 31541 + }, + { + "epoch": 5.74, + "learning_rate": 1.2803785948307245e-06, + "loss": 2.0288, + "step": 31542 + }, + { + "epoch": 5.74, + "learning_rate": 1.2803239898070622e-06, + "loss": 0.4465, + "step": 31543 + }, + { + "epoch": 5.74, + "learning_rate": 1.2802693847834e-06, + "loss": 1.563, + "step": 31544 + }, + { + "epoch": 5.74, + "learning_rate": 1.280214779759738e-06, + "loss": 1.4991, + "step": 31545 + }, + { + "epoch": 5.74, + "learning_rate": 1.2801601747360759e-06, + "loss": 0.1391, + "step": 31546 + }, + { + "epoch": 5.74, + "learning_rate": 1.2801055697124136e-06, + "loss": 0.0078, + "step": 31547 + }, + { + "epoch": 5.74, + "learning_rate": 1.2800509646887513e-06, + "loss": 0.1749, + "step": 31548 + }, + { + "epoch": 5.74, + "learning_rate": 1.2799963596650893e-06, + "loss": 1.7821, + "step": 31549 + }, + { + "epoch": 5.74, + "learning_rate": 1.279941754641427e-06, + "loss": 0.2754, + "step": 31550 + }, + { + "epoch": 5.74, + "learning_rate": 1.2798871496177647e-06, + "loss": 0.0049, + "step": 31551 + }, + { + "epoch": 5.74, + "learning_rate": 1.2798325445941027e-06, + "loss": 0.9647, + "step": 31552 + }, + { + "epoch": 5.74, + "learning_rate": 1.2797779395704406e-06, + "loss": 0.0088, + "step": 31553 + }, + { + "epoch": 5.74, + "learning_rate": 1.2797233345467783e-06, + "loss": 0.4429, + "step": 31554 + }, + { + "epoch": 5.74, + "learning_rate": 1.279668729523116e-06, + "loss": 0.0489, + "step": 31555 + }, + { + "epoch": 5.74, + "learning_rate": 1.279614124499454e-06, + "loss": 1.5363, + "step": 31556 + }, + { + "epoch": 5.74, + "learning_rate": 1.2795595194757918e-06, + "loss": 1.0154, + "step": 31557 + }, + { + "epoch": 5.74, + "learning_rate": 1.2795049144521295e-06, + "loss": 0.0056, + "step": 31558 + }, + { + "epoch": 5.74, + "learning_rate": 1.2794503094284674e-06, + "loss": 0.2757, + "step": 31559 + }, + { + "epoch": 5.74, + "learning_rate": 1.2793957044048054e-06, + "loss": 1.7121, + "step": 31560 + }, + { + "epoch": 5.74, + "learning_rate": 1.279341099381143e-06, + "loss": 0.0995, + "step": 31561 + }, + { + "epoch": 5.74, + "learning_rate": 1.2792864943574808e-06, + "loss": 0.9549, + "step": 31562 + }, + { + "epoch": 5.74, + "learning_rate": 1.2792318893338188e-06, + "loss": 1.1818, + "step": 31563 + }, + { + "epoch": 5.74, + "learning_rate": 1.2791772843101565e-06, + "loss": 0.6611, + "step": 31564 + }, + { + "epoch": 5.74, + "learning_rate": 1.2791226792864942e-06, + "loss": 0.1689, + "step": 31565 + }, + { + "epoch": 5.74, + "learning_rate": 1.2790680742628322e-06, + "loss": 1.1689, + "step": 31566 + }, + { + "epoch": 5.74, + "learning_rate": 1.2790134692391701e-06, + "loss": 0.4422, + "step": 31567 + }, + { + "epoch": 5.74, + "learning_rate": 1.2789588642155079e-06, + "loss": 1.7071, + "step": 31568 + }, + { + "epoch": 5.74, + "learning_rate": 1.2789042591918456e-06, + "loss": 1.5746, + "step": 31569 + }, + { + "epoch": 5.74, + "learning_rate": 1.2788496541681835e-06, + "loss": 0.3538, + "step": 31570 + }, + { + "epoch": 5.74, + "learning_rate": 1.2787950491445213e-06, + "loss": 1.1817, + "step": 31571 + }, + { + "epoch": 5.74, + "learning_rate": 1.2787404441208592e-06, + "loss": 0.494, + "step": 31572 + }, + { + "epoch": 5.74, + "learning_rate": 1.2786858390971972e-06, + "loss": 0.1153, + "step": 31573 + }, + { + "epoch": 5.74, + "learning_rate": 1.2786312340735349e-06, + "loss": 0.3599, + "step": 31574 + }, + { + "epoch": 5.74, + "learning_rate": 1.2785766290498726e-06, + "loss": 2.7525, + "step": 31575 + }, + { + "epoch": 5.74, + "learning_rate": 1.2785220240262103e-06, + "loss": 0.1228, + "step": 31576 + }, + { + "epoch": 5.74, + "learning_rate": 1.2784674190025483e-06, + "loss": 0.7445, + "step": 31577 + }, + { + "epoch": 5.74, + "learning_rate": 1.278412813978886e-06, + "loss": 0.3919, + "step": 31578 + }, + { + "epoch": 5.74, + "learning_rate": 1.278358208955224e-06, + "loss": 0.3247, + "step": 31579 + }, + { + "epoch": 5.74, + "learning_rate": 1.278303603931562e-06, + "loss": 0.5351, + "step": 31580 + }, + { + "epoch": 5.74, + "learning_rate": 1.2782489989078996e-06, + "loss": 0.281, + "step": 31581 + }, + { + "epoch": 5.74, + "learning_rate": 1.2781943938842374e-06, + "loss": 0.371, + "step": 31582 + }, + { + "epoch": 5.74, + "learning_rate": 1.278139788860575e-06, + "loss": 0.596, + "step": 31583 + }, + { + "epoch": 5.74, + "learning_rate": 1.278085183836913e-06, + "loss": 0.7684, + "step": 31584 + }, + { + "epoch": 5.74, + "learning_rate": 1.2780305788132508e-06, + "loss": 0.6942, + "step": 31585 + }, + { + "epoch": 5.74, + "learning_rate": 1.2779759737895887e-06, + "loss": 0.8609, + "step": 31586 + }, + { + "epoch": 5.74, + "learning_rate": 1.2779213687659267e-06, + "loss": 0.0316, + "step": 31587 + }, + { + "epoch": 5.74, + "learning_rate": 1.2778667637422644e-06, + "loss": 0.1891, + "step": 31588 + }, + { + "epoch": 5.74, + "learning_rate": 1.2778121587186021e-06, + "loss": 0.1649, + "step": 31589 + }, + { + "epoch": 5.74, + "learning_rate": 1.2777575536949399e-06, + "loss": 1.5712, + "step": 31590 + }, + { + "epoch": 5.74, + "learning_rate": 1.2777029486712778e-06, + "loss": 1.4978, + "step": 31591 + }, + { + "epoch": 5.75, + "learning_rate": 1.2776483436476155e-06, + "loss": 0.7722, + "step": 31592 + }, + { + "epoch": 5.75, + "learning_rate": 1.2775937386239535e-06, + "loss": 1.0967, + "step": 31593 + }, + { + "epoch": 5.75, + "learning_rate": 1.2775391336002912e-06, + "loss": 0.3238, + "step": 31594 + }, + { + "epoch": 5.75, + "learning_rate": 1.2774845285766292e-06, + "loss": 0.6351, + "step": 31595 + }, + { + "epoch": 5.75, + "learning_rate": 1.2774299235529669e-06, + "loss": 0.8392, + "step": 31596 + }, + { + "epoch": 5.75, + "learning_rate": 1.2773753185293046e-06, + "loss": 1.1861, + "step": 31597 + }, + { + "epoch": 5.75, + "learning_rate": 1.2773207135056426e-06, + "loss": 1.9033, + "step": 31598 + }, + { + "epoch": 5.75, + "learning_rate": 1.2772661084819803e-06, + "loss": 0.0973, + "step": 31599 + }, + { + "epoch": 5.75, + "learning_rate": 1.2772115034583182e-06, + "loss": 0.0065, + "step": 31600 + }, + { + "epoch": 5.75, + "learning_rate": 1.277156898434656e-06, + "loss": 0.0491, + "step": 31601 + }, + { + "epoch": 5.75, + "learning_rate": 1.277102293410994e-06, + "loss": 1.275, + "step": 31602 + }, + { + "epoch": 5.75, + "learning_rate": 1.2770476883873316e-06, + "loss": 0.0033, + "step": 31603 + }, + { + "epoch": 5.75, + "learning_rate": 1.2769930833636694e-06, + "loss": 0.4528, + "step": 31604 + }, + { + "epoch": 5.75, + "learning_rate": 1.2769384783400073e-06, + "loss": 0.9913, + "step": 31605 + }, + { + "epoch": 5.75, + "learning_rate": 1.276883873316345e-06, + "loss": 0.13, + "step": 31606 + }, + { + "epoch": 5.75, + "learning_rate": 1.276829268292683e-06, + "loss": 0.4753, + "step": 31607 + }, + { + "epoch": 5.75, + "learning_rate": 1.2767746632690207e-06, + "loss": 1.6467, + "step": 31608 + }, + { + "epoch": 5.75, + "learning_rate": 1.2767200582453587e-06, + "loss": 0.0047, + "step": 31609 + }, + { + "epoch": 5.75, + "learning_rate": 1.2766654532216964e-06, + "loss": 0.4131, + "step": 31610 + }, + { + "epoch": 5.75, + "learning_rate": 1.2766108481980341e-06, + "loss": 0.1225, + "step": 31611 + }, + { + "epoch": 5.75, + "learning_rate": 1.276556243174372e-06, + "loss": 2.4351, + "step": 31612 + }, + { + "epoch": 5.75, + "learning_rate": 1.27650163815071e-06, + "loss": 1.4439, + "step": 31613 + }, + { + "epoch": 5.75, + "learning_rate": 1.2764470331270478e-06, + "loss": 1.2997, + "step": 31614 + }, + { + "epoch": 5.75, + "learning_rate": 1.2763924281033855e-06, + "loss": 0.5563, + "step": 31615 + }, + { + "epoch": 5.75, + "learning_rate": 1.2763378230797234e-06, + "loss": 2.5671, + "step": 31616 + }, + { + "epoch": 5.75, + "learning_rate": 1.2762832180560612e-06, + "loss": 1.6502, + "step": 31617 + }, + { + "epoch": 5.75, + "learning_rate": 1.2762286130323989e-06, + "loss": 0.1866, + "step": 31618 + }, + { + "epoch": 5.75, + "learning_rate": 1.2761740080087368e-06, + "loss": 0.0789, + "step": 31619 + }, + { + "epoch": 5.75, + "learning_rate": 1.2761194029850748e-06, + "loss": 2.2441, + "step": 31620 + }, + { + "epoch": 5.75, + "learning_rate": 1.2760647979614125e-06, + "loss": 2.4317, + "step": 31621 + }, + { + "epoch": 5.75, + "learning_rate": 1.2760101929377502e-06, + "loss": 0.5846, + "step": 31622 + }, + { + "epoch": 5.75, + "learning_rate": 1.2759555879140882e-06, + "loss": 0.0078, + "step": 31623 + }, + { + "epoch": 5.75, + "learning_rate": 1.275900982890426e-06, + "loss": 0.3096, + "step": 31624 + }, + { + "epoch": 5.75, + "learning_rate": 1.2758463778667636e-06, + "loss": 1.7425, + "step": 31625 + }, + { + "epoch": 5.75, + "learning_rate": 1.2757917728431016e-06, + "loss": 1.2552, + "step": 31626 + }, + { + "epoch": 5.75, + "learning_rate": 1.2757371678194395e-06, + "loss": 1.2286, + "step": 31627 + }, + { + "epoch": 5.75, + "learning_rate": 1.2756825627957773e-06, + "loss": 2.6081, + "step": 31628 + }, + { + "epoch": 5.75, + "learning_rate": 1.275627957772115e-06, + "loss": 0.3618, + "step": 31629 + }, + { + "epoch": 5.75, + "learning_rate": 1.275573352748453e-06, + "loss": 0.0707, + "step": 31630 + }, + { + "epoch": 5.75, + "learning_rate": 1.2755187477247907e-06, + "loss": 0.0095, + "step": 31631 + }, + { + "epoch": 5.75, + "learning_rate": 1.2754641427011284e-06, + "loss": 0.466, + "step": 31632 + }, + { + "epoch": 5.75, + "learning_rate": 1.2754095376774663e-06, + "loss": 0.444, + "step": 31633 + }, + { + "epoch": 5.75, + "learning_rate": 1.2753549326538043e-06, + "loss": 0.1193, + "step": 31634 + }, + { + "epoch": 5.75, + "learning_rate": 1.275300327630142e-06, + "loss": 0.1044, + "step": 31635 + }, + { + "epoch": 5.75, + "learning_rate": 1.2752457226064798e-06, + "loss": 0.852, + "step": 31636 + }, + { + "epoch": 5.75, + "learning_rate": 1.2751911175828177e-06, + "loss": 0.3682, + "step": 31637 + }, + { + "epoch": 5.75, + "learning_rate": 1.2751365125591554e-06, + "loss": 0.0401, + "step": 31638 + }, + { + "epoch": 5.75, + "learning_rate": 1.2750819075354932e-06, + "loss": 0.6849, + "step": 31639 + }, + { + "epoch": 5.75, + "learning_rate": 1.275027302511831e-06, + "loss": 0.0857, + "step": 31640 + }, + { + "epoch": 5.75, + "learning_rate": 1.274972697488169e-06, + "loss": 0.0427, + "step": 31641 + }, + { + "epoch": 5.75, + "learning_rate": 1.2749180924645068e-06, + "loss": 0.2598, + "step": 31642 + }, + { + "epoch": 5.75, + "learning_rate": 1.2748634874408445e-06, + "loss": 0.0086, + "step": 31643 + }, + { + "epoch": 5.75, + "learning_rate": 1.2748088824171825e-06, + "loss": 1.1806, + "step": 31644 + }, + { + "epoch": 5.75, + "learning_rate": 1.2747542773935202e-06, + "loss": 1.1078, + "step": 31645 + }, + { + "epoch": 5.75, + "learning_rate": 1.274699672369858e-06, + "loss": 0.7149, + "step": 31646 + }, + { + "epoch": 5.76, + "learning_rate": 1.2746450673461959e-06, + "loss": 0.8445, + "step": 31647 + }, + { + "epoch": 5.76, + "learning_rate": 1.2745904623225338e-06, + "loss": 0.6067, + "step": 31648 + }, + { + "epoch": 5.76, + "learning_rate": 1.2745358572988715e-06, + "loss": 0.1237, + "step": 31649 + }, + { + "epoch": 5.76, + "learning_rate": 1.2744812522752093e-06, + "loss": 1.757, + "step": 31650 + }, + { + "epoch": 5.76, + "learning_rate": 1.2744266472515472e-06, + "loss": 0.0821, + "step": 31651 + }, + { + "epoch": 5.76, + "learning_rate": 1.274372042227885e-06, + "loss": 0.1699, + "step": 31652 + }, + { + "epoch": 5.76, + "learning_rate": 1.2743174372042229e-06, + "loss": 0.9339, + "step": 31653 + }, + { + "epoch": 5.76, + "learning_rate": 1.2742628321805606e-06, + "loss": 1.0745, + "step": 31654 + }, + { + "epoch": 5.76, + "learning_rate": 1.2742082271568986e-06, + "loss": 0.0195, + "step": 31655 + }, + { + "epoch": 5.76, + "learning_rate": 1.2741536221332363e-06, + "loss": 0.0108, + "step": 31656 + }, + { + "epoch": 5.76, + "learning_rate": 1.274099017109574e-06, + "loss": 0.9167, + "step": 31657 + }, + { + "epoch": 5.76, + "learning_rate": 1.274044412085912e-06, + "loss": 0.5029, + "step": 31658 + }, + { + "epoch": 5.76, + "learning_rate": 1.2739898070622497e-06, + "loss": 2.1689, + "step": 31659 + }, + { + "epoch": 5.76, + "learning_rate": 1.2739352020385876e-06, + "loss": 0.0091, + "step": 31660 + }, + { + "epoch": 5.76, + "learning_rate": 1.2738805970149256e-06, + "loss": 0.4368, + "step": 31661 + }, + { + "epoch": 5.76, + "learning_rate": 1.2738259919912633e-06, + "loss": 1.6425, + "step": 31662 + }, + { + "epoch": 5.76, + "learning_rate": 1.273771386967601e-06, + "loss": 0.1052, + "step": 31663 + }, + { + "epoch": 5.76, + "learning_rate": 1.2737167819439388e-06, + "loss": 0.8483, + "step": 31664 + }, + { + "epoch": 5.76, + "learning_rate": 1.2736621769202767e-06, + "loss": 1.1881, + "step": 31665 + }, + { + "epoch": 5.76, + "learning_rate": 1.2736075718966145e-06, + "loss": 1.7002, + "step": 31666 + }, + { + "epoch": 5.76, + "learning_rate": 1.2735529668729524e-06, + "loss": 0.6254, + "step": 31667 + }, + { + "epoch": 5.76, + "learning_rate": 1.2734983618492903e-06, + "loss": 0.7331, + "step": 31668 + }, + { + "epoch": 5.76, + "learning_rate": 1.273443756825628e-06, + "loss": 0.8429, + "step": 31669 + }, + { + "epoch": 5.76, + "learning_rate": 1.2733891518019658e-06, + "loss": 0.0129, + "step": 31670 + }, + { + "epoch": 5.76, + "learning_rate": 1.2733345467783035e-06, + "loss": 2.7311, + "step": 31671 + }, + { + "epoch": 5.76, + "learning_rate": 1.2732799417546415e-06, + "loss": 1.2949, + "step": 31672 + }, + { + "epoch": 5.76, + "learning_rate": 1.2732253367309792e-06, + "loss": 0.4833, + "step": 31673 + }, + { + "epoch": 5.76, + "learning_rate": 1.2731707317073172e-06, + "loss": 0.0356, + "step": 31674 + }, + { + "epoch": 5.76, + "learning_rate": 1.2731161266836549e-06, + "loss": 2.8191, + "step": 31675 + }, + { + "epoch": 5.76, + "learning_rate": 1.2730615216599928e-06, + "loss": 0.6693, + "step": 31676 + }, + { + "epoch": 5.76, + "learning_rate": 1.2730069166363306e-06, + "loss": 1.9575, + "step": 31677 + }, + { + "epoch": 5.76, + "learning_rate": 1.2729523116126683e-06, + "loss": 0.1064, + "step": 31678 + }, + { + "epoch": 5.76, + "learning_rate": 1.2728977065890062e-06, + "loss": 0.0075, + "step": 31679 + }, + { + "epoch": 5.76, + "learning_rate": 1.272843101565344e-06, + "loss": 0.12, + "step": 31680 + }, + { + "epoch": 5.76, + "learning_rate": 1.272788496541682e-06, + "loss": 0.7982, + "step": 31681 + }, + { + "epoch": 5.76, + "learning_rate": 1.2727338915180196e-06, + "loss": 1.5879, + "step": 31682 + }, + { + "epoch": 5.76, + "learning_rate": 1.2726792864943576e-06, + "loss": 0.0078, + "step": 31683 + }, + { + "epoch": 5.76, + "learning_rate": 1.2726246814706953e-06, + "loss": 0.0351, + "step": 31684 + }, + { + "epoch": 5.76, + "learning_rate": 1.272570076447033e-06, + "loss": 0.3707, + "step": 31685 + }, + { + "epoch": 5.76, + "learning_rate": 1.272515471423371e-06, + "loss": 0.12, + "step": 31686 + }, + { + "epoch": 5.76, + "learning_rate": 1.2724608663997087e-06, + "loss": 0.2031, + "step": 31687 + }, + { + "epoch": 5.76, + "learning_rate": 1.2724062613760467e-06, + "loss": 0.0093, + "step": 31688 + }, + { + "epoch": 5.76, + "learning_rate": 1.2723516563523844e-06, + "loss": 0.3614, + "step": 31689 + }, + { + "epoch": 5.76, + "learning_rate": 1.2722970513287223e-06, + "loss": 0.144, + "step": 31690 + }, + { + "epoch": 5.76, + "learning_rate": 1.27224244630506e-06, + "loss": 0.7065, + "step": 31691 + }, + { + "epoch": 5.76, + "learning_rate": 1.2721878412813978e-06, + "loss": 2.395, + "step": 31692 + }, + { + "epoch": 5.76, + "learning_rate": 1.2721332362577358e-06, + "loss": 0.1371, + "step": 31693 + }, + { + "epoch": 5.76, + "learning_rate": 1.2720786312340735e-06, + "loss": 1.7887, + "step": 31694 + }, + { + "epoch": 5.76, + "learning_rate": 1.2720240262104114e-06, + "loss": 0.11, + "step": 31695 + }, + { + "epoch": 5.76, + "learning_rate": 1.2719694211867492e-06, + "loss": 0.3747, + "step": 31696 + }, + { + "epoch": 5.76, + "learning_rate": 1.271914816163087e-06, + "loss": 0.949, + "step": 31697 + }, + { + "epoch": 5.76, + "learning_rate": 1.2718602111394248e-06, + "loss": 1.2032, + "step": 31698 + }, + { + "epoch": 5.76, + "learning_rate": 1.2718056061157626e-06, + "loss": 0.0904, + "step": 31699 + }, + { + "epoch": 5.76, + "learning_rate": 1.2717510010921005e-06, + "loss": 1.0444, + "step": 31700 + }, + { + "epoch": 5.76, + "learning_rate": 1.2716963960684385e-06, + "loss": 1.7844, + "step": 31701 + }, + { + "epoch": 5.77, + "learning_rate": 1.2716417910447762e-06, + "loss": 2.4469, + "step": 31702 + }, + { + "epoch": 5.77, + "learning_rate": 1.271587186021114e-06, + "loss": 1.2368, + "step": 31703 + }, + { + "epoch": 5.77, + "learning_rate": 1.2715325809974519e-06, + "loss": 0.5643, + "step": 31704 + }, + { + "epoch": 5.77, + "learning_rate": 1.2714779759737896e-06, + "loss": 1.876, + "step": 31705 + }, + { + "epoch": 5.77, + "learning_rate": 1.2714233709501273e-06, + "loss": 0.0116, + "step": 31706 + }, + { + "epoch": 5.77, + "learning_rate": 1.2713687659264653e-06, + "loss": 0.5187, + "step": 31707 + }, + { + "epoch": 5.77, + "learning_rate": 1.2713141609028032e-06, + "loss": 0.5473, + "step": 31708 + }, + { + "epoch": 5.77, + "learning_rate": 1.271259555879141e-06, + "loss": 0.2452, + "step": 31709 + }, + { + "epoch": 5.77, + "learning_rate": 1.2712049508554787e-06, + "loss": 0.0351, + "step": 31710 + }, + { + "epoch": 5.77, + "learning_rate": 1.2711503458318166e-06, + "loss": 0.2979, + "step": 31711 + }, + { + "epoch": 5.77, + "learning_rate": 1.2710957408081543e-06, + "loss": 1.4725, + "step": 31712 + }, + { + "epoch": 5.77, + "learning_rate": 1.271041135784492e-06, + "loss": 0.4227, + "step": 31713 + }, + { + "epoch": 5.77, + "learning_rate": 1.27098653076083e-06, + "loss": 2.0846, + "step": 31714 + }, + { + "epoch": 5.77, + "learning_rate": 1.270931925737168e-06, + "loss": 0.0641, + "step": 31715 + }, + { + "epoch": 5.77, + "learning_rate": 1.2708773207135057e-06, + "loss": 1.3443, + "step": 31716 + }, + { + "epoch": 5.77, + "learning_rate": 1.2708227156898434e-06, + "loss": 0.2105, + "step": 31717 + }, + { + "epoch": 5.77, + "learning_rate": 1.2707681106661814e-06, + "loss": 0.5524, + "step": 31718 + }, + { + "epoch": 5.77, + "learning_rate": 1.270713505642519e-06, + "loss": 0.9613, + "step": 31719 + }, + { + "epoch": 5.77, + "learning_rate": 1.2706589006188568e-06, + "loss": 0.1842, + "step": 31720 + }, + { + "epoch": 5.77, + "learning_rate": 1.2706042955951948e-06, + "loss": 1.8847, + "step": 31721 + }, + { + "epoch": 5.77, + "learning_rate": 1.2705496905715327e-06, + "loss": 0.2349, + "step": 31722 + }, + { + "epoch": 5.77, + "learning_rate": 1.2704950855478705e-06, + "loss": 0.5701, + "step": 31723 + }, + { + "epoch": 5.77, + "learning_rate": 1.2704404805242082e-06, + "loss": 0.8887, + "step": 31724 + }, + { + "epoch": 5.77, + "learning_rate": 1.2703858755005461e-06, + "loss": 1.0571, + "step": 31725 + }, + { + "epoch": 5.77, + "learning_rate": 1.2703312704768839e-06, + "loss": 1.1284, + "step": 31726 + }, + { + "epoch": 5.77, + "learning_rate": 1.2702766654532216e-06, + "loss": 0.5898, + "step": 31727 + }, + { + "epoch": 5.77, + "learning_rate": 1.2702220604295595e-06, + "loss": 1.4455, + "step": 31728 + }, + { + "epoch": 5.77, + "learning_rate": 1.2701674554058975e-06, + "loss": 0.5696, + "step": 31729 + }, + { + "epoch": 5.77, + "learning_rate": 1.2701128503822352e-06, + "loss": 0.0435, + "step": 31730 + }, + { + "epoch": 5.77, + "learning_rate": 1.270058245358573e-06, + "loss": 1.8691, + "step": 31731 + }, + { + "epoch": 5.77, + "learning_rate": 1.2700036403349109e-06, + "loss": 0.1116, + "step": 31732 + }, + { + "epoch": 5.77, + "learning_rate": 1.2699490353112486e-06, + "loss": 0.9318, + "step": 31733 + }, + { + "epoch": 5.77, + "learning_rate": 1.2698944302875863e-06, + "loss": 1.2782, + "step": 31734 + }, + { + "epoch": 5.77, + "learning_rate": 1.2698398252639243e-06, + "loss": 0.023, + "step": 31735 + }, + { + "epoch": 5.77, + "learning_rate": 1.2697852202402622e-06, + "loss": 2.2813, + "step": 31736 + }, + { + "epoch": 5.77, + "learning_rate": 1.2697306152166e-06, + "loss": 2.5862, + "step": 31737 + }, + { + "epoch": 5.77, + "learning_rate": 1.2696760101929377e-06, + "loss": 0.6973, + "step": 31738 + }, + { + "epoch": 5.77, + "learning_rate": 1.2696214051692756e-06, + "loss": 0.2715, + "step": 31739 + }, + { + "epoch": 5.77, + "learning_rate": 1.2695668001456134e-06, + "loss": 0.1482, + "step": 31740 + }, + { + "epoch": 5.77, + "learning_rate": 1.2695121951219513e-06, + "loss": 1.485, + "step": 31741 + }, + { + "epoch": 5.77, + "learning_rate": 1.2694575900982893e-06, + "loss": 0.1919, + "step": 31742 + }, + { + "epoch": 5.77, + "learning_rate": 1.269402985074627e-06, + "loss": 0.0636, + "step": 31743 + }, + { + "epoch": 5.77, + "learning_rate": 1.2693483800509647e-06, + "loss": 0.1154, + "step": 31744 + }, + { + "epoch": 5.77, + "learning_rate": 1.2692937750273025e-06, + "loss": 0.3145, + "step": 31745 + }, + { + "epoch": 5.77, + "learning_rate": 1.2692391700036404e-06, + "loss": 0.0087, + "step": 31746 + }, + { + "epoch": 5.77, + "learning_rate": 1.2691845649799781e-06, + "loss": 0.5772, + "step": 31747 + }, + { + "epoch": 5.77, + "learning_rate": 1.269129959956316e-06, + "loss": 0.2247, + "step": 31748 + }, + { + "epoch": 5.77, + "learning_rate": 1.269075354932654e-06, + "loss": 0.7115, + "step": 31749 + }, + { + "epoch": 5.77, + "learning_rate": 1.2690207499089917e-06, + "loss": 0.1299, + "step": 31750 + }, + { + "epoch": 5.77, + "learning_rate": 1.2689661448853295e-06, + "loss": 0.5685, + "step": 31751 + }, + { + "epoch": 5.77, + "learning_rate": 1.2689115398616672e-06, + "loss": 2.2015, + "step": 31752 + }, + { + "epoch": 5.77, + "learning_rate": 1.2688569348380052e-06, + "loss": 0.2179, + "step": 31753 + }, + { + "epoch": 5.77, + "learning_rate": 1.2688023298143429e-06, + "loss": 0.456, + "step": 31754 + }, + { + "epoch": 5.77, + "learning_rate": 1.2687477247906808e-06, + "loss": 1.5445, + "step": 31755 + }, + { + "epoch": 5.77, + "learning_rate": 1.2686931197670186e-06, + "loss": 0.1816, + "step": 31756 + }, + { + "epoch": 5.78, + "learning_rate": 1.2686385147433565e-06, + "loss": 0.4803, + "step": 31757 + }, + { + "epoch": 5.78, + "learning_rate": 1.2685839097196942e-06, + "loss": 1.7067, + "step": 31758 + }, + { + "epoch": 5.78, + "learning_rate": 1.268529304696032e-06, + "loss": 0.5818, + "step": 31759 + }, + { + "epoch": 5.78, + "learning_rate": 1.26847469967237e-06, + "loss": 0.1507, + "step": 31760 + }, + { + "epoch": 5.78, + "learning_rate": 1.2684200946487076e-06, + "loss": 2.3448, + "step": 31761 + }, + { + "epoch": 5.78, + "learning_rate": 1.2683654896250456e-06, + "loss": 0.0321, + "step": 31762 + }, + { + "epoch": 5.78, + "learning_rate": 1.2683108846013833e-06, + "loss": 0.853, + "step": 31763 + }, + { + "epoch": 5.78, + "learning_rate": 1.2682562795777213e-06, + "loss": 1.872, + "step": 31764 + }, + { + "epoch": 5.78, + "learning_rate": 1.268201674554059e-06, + "loss": 1.2883, + "step": 31765 + }, + { + "epoch": 5.78, + "learning_rate": 1.2681470695303967e-06, + "loss": 1.4032, + "step": 31766 + }, + { + "epoch": 5.78, + "learning_rate": 1.2680924645067347e-06, + "loss": 1.105, + "step": 31767 + }, + { + "epoch": 5.78, + "learning_rate": 1.2680378594830724e-06, + "loss": 0.0057, + "step": 31768 + }, + { + "epoch": 5.78, + "learning_rate": 1.2679832544594103e-06, + "loss": 0.7422, + "step": 31769 + }, + { + "epoch": 5.78, + "learning_rate": 1.267928649435748e-06, + "loss": 0.311, + "step": 31770 + }, + { + "epoch": 5.78, + "learning_rate": 1.267874044412086e-06, + "loss": 0.4514, + "step": 31771 + }, + { + "epoch": 5.78, + "learning_rate": 1.2678194393884237e-06, + "loss": 0.9862, + "step": 31772 + }, + { + "epoch": 5.78, + "learning_rate": 1.2677648343647615e-06, + "loss": 0.0346, + "step": 31773 + }, + { + "epoch": 5.78, + "learning_rate": 1.2677102293410994e-06, + "loss": 0.0214, + "step": 31774 + }, + { + "epoch": 5.78, + "learning_rate": 1.2676556243174372e-06, + "loss": 2.4892, + "step": 31775 + }, + { + "epoch": 5.78, + "learning_rate": 1.267601019293775e-06, + "loss": 0.1544, + "step": 31776 + }, + { + "epoch": 5.78, + "learning_rate": 1.2675464142701128e-06, + "loss": 0.8472, + "step": 31777 + }, + { + "epoch": 5.78, + "learning_rate": 1.2674918092464508e-06, + "loss": 0.4421, + "step": 31778 + }, + { + "epoch": 5.78, + "learning_rate": 1.2674372042227885e-06, + "loss": 0.9741, + "step": 31779 + }, + { + "epoch": 5.78, + "learning_rate": 1.2673825991991262e-06, + "loss": 0.2131, + "step": 31780 + }, + { + "epoch": 5.78, + "learning_rate": 1.2673279941754642e-06, + "loss": 0.2894, + "step": 31781 + }, + { + "epoch": 5.78, + "learning_rate": 1.2672733891518021e-06, + "loss": 0.6671, + "step": 31782 + }, + { + "epoch": 5.78, + "learning_rate": 1.2672187841281399e-06, + "loss": 0.0502, + "step": 31783 + }, + { + "epoch": 5.78, + "learning_rate": 1.2671641791044776e-06, + "loss": 0.5101, + "step": 31784 + }, + { + "epoch": 5.78, + "learning_rate": 1.2671095740808155e-06, + "loss": 1.849, + "step": 31785 + }, + { + "epoch": 5.78, + "learning_rate": 1.2670549690571533e-06, + "loss": 0.008, + "step": 31786 + }, + { + "epoch": 5.78, + "learning_rate": 1.267000364033491e-06, + "loss": 0.1083, + "step": 31787 + }, + { + "epoch": 5.78, + "learning_rate": 1.266945759009829e-06, + "loss": 0.0306, + "step": 31788 + }, + { + "epoch": 5.78, + "learning_rate": 1.2668911539861669e-06, + "loss": 0.0795, + "step": 31789 + }, + { + "epoch": 5.78, + "learning_rate": 1.2668365489625046e-06, + "loss": 1.268, + "step": 31790 + }, + { + "epoch": 5.78, + "learning_rate": 1.2667819439388423e-06, + "loss": 0.68, + "step": 31791 + }, + { + "epoch": 5.78, + "learning_rate": 1.2667273389151803e-06, + "loss": 0.3919, + "step": 31792 + }, + { + "epoch": 5.78, + "learning_rate": 1.266672733891518e-06, + "loss": 0.5001, + "step": 31793 + }, + { + "epoch": 5.78, + "learning_rate": 1.2666181288678558e-06, + "loss": 0.0052, + "step": 31794 + }, + { + "epoch": 5.78, + "learning_rate": 1.2665635238441937e-06, + "loss": 0.2304, + "step": 31795 + }, + { + "epoch": 5.78, + "learning_rate": 1.2665089188205316e-06, + "loss": 0.2458, + "step": 31796 + }, + { + "epoch": 5.78, + "learning_rate": 1.2664543137968694e-06, + "loss": 0.1887, + "step": 31797 + }, + { + "epoch": 5.78, + "learning_rate": 1.266399708773207e-06, + "loss": 0.9987, + "step": 31798 + }, + { + "epoch": 5.78, + "learning_rate": 1.266345103749545e-06, + "loss": 0.0072, + "step": 31799 + }, + { + "epoch": 5.78, + "learning_rate": 1.2662904987258828e-06, + "loss": 0.0194, + "step": 31800 + }, + { + "epoch": 5.78, + "learning_rate": 1.2662358937022205e-06, + "loss": 1.6173, + "step": 31801 + }, + { + "epoch": 5.78, + "learning_rate": 1.2661812886785585e-06, + "loss": 0.139, + "step": 31802 + }, + { + "epoch": 5.78, + "learning_rate": 1.2661266836548964e-06, + "loss": 0.7821, + "step": 31803 + }, + { + "epoch": 5.78, + "learning_rate": 1.2660720786312341e-06, + "loss": 0.0342, + "step": 31804 + }, + { + "epoch": 5.78, + "learning_rate": 1.2660174736075719e-06, + "loss": 0.5615, + "step": 31805 + }, + { + "epoch": 5.78, + "learning_rate": 1.2659628685839098e-06, + "loss": 0.7763, + "step": 31806 + }, + { + "epoch": 5.78, + "learning_rate": 1.2659082635602475e-06, + "loss": 1.4059, + "step": 31807 + }, + { + "epoch": 5.78, + "learning_rate": 1.2658536585365853e-06, + "loss": 0.0224, + "step": 31808 + }, + { + "epoch": 5.78, + "learning_rate": 1.2657990535129232e-06, + "loss": 0.903, + "step": 31809 + }, + { + "epoch": 5.78, + "learning_rate": 1.2657444484892612e-06, + "loss": 0.0915, + "step": 31810 + }, + { + "epoch": 5.78, + "learning_rate": 1.2656898434655989e-06, + "loss": 1.2022, + "step": 31811 + }, + { + "epoch": 5.79, + "learning_rate": 1.2656352384419366e-06, + "loss": 0.0168, + "step": 31812 + }, + { + "epoch": 5.79, + "learning_rate": 1.2655806334182746e-06, + "loss": 0.2023, + "step": 31813 + }, + { + "epoch": 5.79, + "learning_rate": 1.2655260283946123e-06, + "loss": 0.2943, + "step": 31814 + }, + { + "epoch": 5.79, + "learning_rate": 1.26547142337095e-06, + "loss": 0.5827, + "step": 31815 + }, + { + "epoch": 5.79, + "learning_rate": 1.265416818347288e-06, + "loss": 1.2617, + "step": 31816 + }, + { + "epoch": 5.79, + "learning_rate": 1.265362213323626e-06, + "loss": 0.076, + "step": 31817 + }, + { + "epoch": 5.79, + "learning_rate": 1.2653076082999636e-06, + "loss": 0.2955, + "step": 31818 + }, + { + "epoch": 5.79, + "learning_rate": 1.2652530032763014e-06, + "loss": 1.4879, + "step": 31819 + }, + { + "epoch": 5.79, + "learning_rate": 1.2651983982526393e-06, + "loss": 0.4522, + "step": 31820 + }, + { + "epoch": 5.79, + "learning_rate": 1.265143793228977e-06, + "loss": 0.5147, + "step": 31821 + }, + { + "epoch": 5.79, + "learning_rate": 1.2650891882053148e-06, + "loss": 0.2152, + "step": 31822 + }, + { + "epoch": 5.79, + "learning_rate": 1.2650345831816527e-06, + "loss": 0.5092, + "step": 31823 + }, + { + "epoch": 5.79, + "learning_rate": 1.2649799781579907e-06, + "loss": 1.0742, + "step": 31824 + }, + { + "epoch": 5.79, + "learning_rate": 1.2649253731343284e-06, + "loss": 1.4298, + "step": 31825 + }, + { + "epoch": 5.79, + "learning_rate": 1.2648707681106661e-06, + "loss": 0.0073, + "step": 31826 + }, + { + "epoch": 5.79, + "learning_rate": 1.264816163087004e-06, + "loss": 1.7859, + "step": 31827 + }, + { + "epoch": 5.79, + "learning_rate": 1.2647615580633418e-06, + "loss": 0.0072, + "step": 31828 + }, + { + "epoch": 5.79, + "learning_rate": 1.2647069530396797e-06, + "loss": 1.3091, + "step": 31829 + }, + { + "epoch": 5.79, + "learning_rate": 1.2646523480160175e-06, + "loss": 2.1086, + "step": 31830 + }, + { + "epoch": 5.79, + "learning_rate": 1.2645977429923554e-06, + "loss": 0.7203, + "step": 31831 + }, + { + "epoch": 5.79, + "learning_rate": 1.2645431379686932e-06, + "loss": 0.022, + "step": 31832 + }, + { + "epoch": 5.79, + "learning_rate": 1.2644885329450309e-06, + "loss": 0.3686, + "step": 31833 + }, + { + "epoch": 5.79, + "learning_rate": 1.2644339279213688e-06, + "loss": 1.9998, + "step": 31834 + }, + { + "epoch": 5.79, + "learning_rate": 1.2643793228977066e-06, + "loss": 0.0794, + "step": 31835 + }, + { + "epoch": 5.79, + "learning_rate": 1.2643247178740445e-06, + "loss": 1.3117, + "step": 31836 + }, + { + "epoch": 5.79, + "learning_rate": 1.2642701128503822e-06, + "loss": 0.0346, + "step": 31837 + }, + { + "epoch": 5.79, + "learning_rate": 1.2642155078267202e-06, + "loss": 0.9229, + "step": 31838 + }, + { + "epoch": 5.79, + "learning_rate": 1.264160902803058e-06, + "loss": 0.0453, + "step": 31839 + }, + { + "epoch": 5.79, + "learning_rate": 1.2641062977793956e-06, + "loss": 2.825, + "step": 31840 + }, + { + "epoch": 5.79, + "learning_rate": 1.2640516927557336e-06, + "loss": 0.1396, + "step": 31841 + }, + { + "epoch": 5.79, + "learning_rate": 1.2639970877320713e-06, + "loss": 0.0837, + "step": 31842 + }, + { + "epoch": 5.79, + "learning_rate": 1.2639424827084093e-06, + "loss": 0.3878, + "step": 31843 + }, + { + "epoch": 5.79, + "learning_rate": 1.263887877684747e-06, + "loss": 0.0604, + "step": 31844 + }, + { + "epoch": 5.79, + "learning_rate": 1.263833272661085e-06, + "loss": 1.5969, + "step": 31845 + }, + { + "epoch": 5.79, + "learning_rate": 1.2637786676374227e-06, + "loss": 1.3356, + "step": 31846 + }, + { + "epoch": 5.79, + "learning_rate": 1.2637240626137604e-06, + "loss": 0.0909, + "step": 31847 + }, + { + "epoch": 5.79, + "learning_rate": 1.2636694575900983e-06, + "loss": 0.8571, + "step": 31848 + }, + { + "epoch": 5.79, + "learning_rate": 1.263614852566436e-06, + "loss": 1.1221, + "step": 31849 + }, + { + "epoch": 5.79, + "learning_rate": 1.263560247542774e-06, + "loss": 3.1257, + "step": 31850 + }, + { + "epoch": 5.79, + "learning_rate": 1.2635056425191117e-06, + "loss": 0.9699, + "step": 31851 + }, + { + "epoch": 5.79, + "learning_rate": 1.2634510374954497e-06, + "loss": 0.0072, + "step": 31852 + }, + { + "epoch": 5.79, + "learning_rate": 1.2633964324717874e-06, + "loss": 0.8409, + "step": 31853 + }, + { + "epoch": 5.79, + "learning_rate": 1.2633418274481252e-06, + "loss": 0.1457, + "step": 31854 + }, + { + "epoch": 5.79, + "learning_rate": 1.263287222424463e-06, + "loss": 1.7779, + "step": 31855 + }, + { + "epoch": 5.79, + "learning_rate": 1.2632326174008008e-06, + "loss": 0.0227, + "step": 31856 + }, + { + "epoch": 5.79, + "learning_rate": 1.2631780123771388e-06, + "loss": 0.2091, + "step": 31857 + }, + { + "epoch": 5.79, + "learning_rate": 1.2631234073534765e-06, + "loss": 2.5878, + "step": 31858 + }, + { + "epoch": 5.79, + "learning_rate": 1.2630688023298144e-06, + "loss": 0.0109, + "step": 31859 + }, + { + "epoch": 5.79, + "learning_rate": 1.2630141973061522e-06, + "loss": 0.9429, + "step": 31860 + }, + { + "epoch": 5.79, + "learning_rate": 1.26295959228249e-06, + "loss": 0.8531, + "step": 31861 + }, + { + "epoch": 5.79, + "learning_rate": 1.2629049872588279e-06, + "loss": 0.0052, + "step": 31862 + }, + { + "epoch": 5.79, + "learning_rate": 1.2628503822351656e-06, + "loss": 0.1485, + "step": 31863 + }, + { + "epoch": 5.79, + "learning_rate": 1.2627957772115035e-06, + "loss": 0.9694, + "step": 31864 + }, + { + "epoch": 5.79, + "learning_rate": 1.2627411721878413e-06, + "loss": 4.1445, + "step": 31865 + }, + { + "epoch": 5.79, + "learning_rate": 1.2626865671641792e-06, + "loss": 0.0197, + "step": 31866 + }, + { + "epoch": 5.8, + "learning_rate": 1.262631962140517e-06, + "loss": 0.3444, + "step": 31867 + }, + { + "epoch": 5.8, + "learning_rate": 1.2625773571168547e-06, + "loss": 0.1054, + "step": 31868 + }, + { + "epoch": 5.8, + "learning_rate": 1.2625227520931926e-06, + "loss": 0.5604, + "step": 31869 + }, + { + "epoch": 5.8, + "learning_rate": 1.2624681470695306e-06, + "loss": 0.0285, + "step": 31870 + }, + { + "epoch": 5.8, + "learning_rate": 1.2624135420458683e-06, + "loss": 1.8214, + "step": 31871 + }, + { + "epoch": 5.8, + "learning_rate": 1.262358937022206e-06, + "loss": 0.1795, + "step": 31872 + }, + { + "epoch": 5.8, + "learning_rate": 1.262304331998544e-06, + "loss": 0.6203, + "step": 31873 + }, + { + "epoch": 5.8, + "learning_rate": 1.2622497269748817e-06, + "loss": 1.8485, + "step": 31874 + }, + { + "epoch": 5.8, + "learning_rate": 1.2621951219512194e-06, + "loss": 0.8205, + "step": 31875 + }, + { + "epoch": 5.8, + "learning_rate": 1.2621405169275574e-06, + "loss": 0.1198, + "step": 31876 + }, + { + "epoch": 5.8, + "learning_rate": 1.2620859119038953e-06, + "loss": 0.0704, + "step": 31877 + }, + { + "epoch": 5.8, + "learning_rate": 1.262031306880233e-06, + "loss": 0.0132, + "step": 31878 + }, + { + "epoch": 5.8, + "learning_rate": 1.2619767018565708e-06, + "loss": 0.538, + "step": 31879 + }, + { + "epoch": 5.8, + "learning_rate": 1.2619220968329087e-06, + "loss": 0.0781, + "step": 31880 + }, + { + "epoch": 5.8, + "learning_rate": 1.2618674918092465e-06, + "loss": 0.599, + "step": 31881 + }, + { + "epoch": 5.8, + "learning_rate": 1.2618128867855842e-06, + "loss": 0.0239, + "step": 31882 + }, + { + "epoch": 5.8, + "learning_rate": 1.2617582817619221e-06, + "loss": 1.5197, + "step": 31883 + }, + { + "epoch": 5.8, + "learning_rate": 1.26170367673826e-06, + "loss": 0.0262, + "step": 31884 + }, + { + "epoch": 5.8, + "learning_rate": 1.2616490717145978e-06, + "loss": 2.3085, + "step": 31885 + }, + { + "epoch": 5.8, + "learning_rate": 1.2615944666909355e-06, + "loss": 0.9408, + "step": 31886 + }, + { + "epoch": 5.8, + "learning_rate": 1.2615398616672735e-06, + "loss": 0.1351, + "step": 31887 + }, + { + "epoch": 5.8, + "learning_rate": 1.2614852566436112e-06, + "loss": 0.0477, + "step": 31888 + }, + { + "epoch": 5.8, + "learning_rate": 1.261430651619949e-06, + "loss": 0.3524, + "step": 31889 + }, + { + "epoch": 5.8, + "learning_rate": 1.2613760465962869e-06, + "loss": 0.7937, + "step": 31890 + }, + { + "epoch": 5.8, + "learning_rate": 1.2613214415726248e-06, + "loss": 0.044, + "step": 31891 + }, + { + "epoch": 5.8, + "learning_rate": 1.2612668365489626e-06, + "loss": 0.3295, + "step": 31892 + }, + { + "epoch": 5.8, + "learning_rate": 1.2612122315253003e-06, + "loss": 0.0695, + "step": 31893 + }, + { + "epoch": 5.8, + "learning_rate": 1.2611576265016382e-06, + "loss": 0.1192, + "step": 31894 + }, + { + "epoch": 5.8, + "learning_rate": 1.261103021477976e-06, + "loss": 0.1806, + "step": 31895 + }, + { + "epoch": 5.8, + "learning_rate": 1.2610484164543137e-06, + "loss": 2.1391, + "step": 31896 + }, + { + "epoch": 5.8, + "learning_rate": 1.2609938114306516e-06, + "loss": 0.3895, + "step": 31897 + }, + { + "epoch": 5.8, + "learning_rate": 1.2609392064069896e-06, + "loss": 0.2045, + "step": 31898 + }, + { + "epoch": 5.8, + "learning_rate": 1.2608846013833273e-06, + "loss": 0.3032, + "step": 31899 + }, + { + "epoch": 5.8, + "learning_rate": 1.260829996359665e-06, + "loss": 1.7054, + "step": 31900 + }, + { + "epoch": 5.8, + "learning_rate": 1.260775391336003e-06, + "loss": 0.2302, + "step": 31901 + }, + { + "epoch": 5.8, + "learning_rate": 1.2607207863123407e-06, + "loss": 0.4667, + "step": 31902 + }, + { + "epoch": 5.8, + "learning_rate": 1.2606661812886785e-06, + "loss": 0.1284, + "step": 31903 + }, + { + "epoch": 5.8, + "learning_rate": 1.2606115762650164e-06, + "loss": 0.1066, + "step": 31904 + }, + { + "epoch": 5.8, + "learning_rate": 1.2605569712413543e-06, + "loss": 0.2144, + "step": 31905 + }, + { + "epoch": 5.8, + "learning_rate": 1.260502366217692e-06, + "loss": 0.1866, + "step": 31906 + }, + { + "epoch": 5.8, + "learning_rate": 1.2604477611940298e-06, + "loss": 0.1266, + "step": 31907 + }, + { + "epoch": 5.8, + "learning_rate": 1.2603931561703677e-06, + "loss": 1.4891, + "step": 31908 + }, + { + "epoch": 5.8, + "learning_rate": 1.2603385511467055e-06, + "loss": 1.5965, + "step": 31909 + }, + { + "epoch": 5.8, + "learning_rate": 1.2602839461230434e-06, + "loss": 0.8421, + "step": 31910 + }, + { + "epoch": 5.8, + "learning_rate": 1.2602293410993812e-06, + "loss": 0.3723, + "step": 31911 + }, + { + "epoch": 5.8, + "learning_rate": 1.260174736075719e-06, + "loss": 0.4017, + "step": 31912 + }, + { + "epoch": 5.8, + "learning_rate": 1.2601201310520568e-06, + "loss": 0.8474, + "step": 31913 + }, + { + "epoch": 5.8, + "learning_rate": 1.2600655260283946e-06, + "loss": 0.0082, + "step": 31914 + }, + { + "epoch": 5.8, + "learning_rate": 1.2600109210047325e-06, + "loss": 0.4832, + "step": 31915 + }, + { + "epoch": 5.8, + "learning_rate": 1.2599563159810702e-06, + "loss": 0.308, + "step": 31916 + }, + { + "epoch": 5.8, + "learning_rate": 1.2599017109574082e-06, + "loss": 0.0114, + "step": 31917 + }, + { + "epoch": 5.8, + "learning_rate": 1.259847105933746e-06, + "loss": 0.8603, + "step": 31918 + }, + { + "epoch": 5.8, + "learning_rate": 1.2597925009100839e-06, + "loss": 0.1112, + "step": 31919 + }, + { + "epoch": 5.8, + "learning_rate": 1.2597378958864216e-06, + "loss": 0.366, + "step": 31920 + }, + { + "epoch": 5.8, + "learning_rate": 1.2596832908627593e-06, + "loss": 0.5767, + "step": 31921 + }, + { + "epoch": 5.81, + "learning_rate": 1.2596286858390973e-06, + "loss": 0.0136, + "step": 31922 + }, + { + "epoch": 5.81, + "learning_rate": 1.259574080815435e-06, + "loss": 0.3117, + "step": 31923 + }, + { + "epoch": 5.81, + "learning_rate": 1.259519475791773e-06, + "loss": 1.1762, + "step": 31924 + }, + { + "epoch": 5.81, + "learning_rate": 1.2594648707681107e-06, + "loss": 0.3517, + "step": 31925 + }, + { + "epoch": 5.81, + "learning_rate": 1.2594102657444486e-06, + "loss": 1.8307, + "step": 31926 + }, + { + "epoch": 5.81, + "learning_rate": 1.2593556607207863e-06, + "loss": 3.4271, + "step": 31927 + }, + { + "epoch": 5.81, + "learning_rate": 1.259301055697124e-06, + "loss": 0.2303, + "step": 31928 + }, + { + "epoch": 5.81, + "learning_rate": 1.259246450673462e-06, + "loss": 0.2627, + "step": 31929 + }, + { + "epoch": 5.81, + "learning_rate": 1.2591918456497997e-06, + "loss": 0.7294, + "step": 31930 + }, + { + "epoch": 5.81, + "learning_rate": 1.2591372406261377e-06, + "loss": 0.0711, + "step": 31931 + }, + { + "epoch": 5.81, + "learning_rate": 1.2590826356024754e-06, + "loss": 0.2102, + "step": 31932 + }, + { + "epoch": 5.81, + "learning_rate": 1.2590280305788134e-06, + "loss": 0.3134, + "step": 31933 + }, + { + "epoch": 5.81, + "learning_rate": 1.258973425555151e-06, + "loss": 1.0568, + "step": 31934 + }, + { + "epoch": 5.81, + "learning_rate": 1.2589188205314888e-06, + "loss": 0.2312, + "step": 31935 + }, + { + "epoch": 5.81, + "learning_rate": 1.2588642155078268e-06, + "loss": 1.1787, + "step": 31936 + }, + { + "epoch": 5.81, + "learning_rate": 1.2588096104841645e-06, + "loss": 1.4123, + "step": 31937 + }, + { + "epoch": 5.81, + "learning_rate": 1.2587550054605024e-06, + "loss": 2.7428, + "step": 31938 + }, + { + "epoch": 5.81, + "learning_rate": 1.2587004004368402e-06, + "loss": 1.0087, + "step": 31939 + }, + { + "epoch": 5.81, + "learning_rate": 1.2586457954131781e-06, + "loss": 0.4647, + "step": 31940 + }, + { + "epoch": 5.81, + "learning_rate": 1.2585911903895159e-06, + "loss": 0.2052, + "step": 31941 + }, + { + "epoch": 5.81, + "learning_rate": 1.2585365853658536e-06, + "loss": 0.1591, + "step": 31942 + }, + { + "epoch": 5.81, + "learning_rate": 1.2584819803421915e-06, + "loss": 0.0878, + "step": 31943 + }, + { + "epoch": 5.81, + "learning_rate": 1.2584273753185293e-06, + "loss": 0.5642, + "step": 31944 + }, + { + "epoch": 5.81, + "learning_rate": 1.2583727702948672e-06, + "loss": 0.1195, + "step": 31945 + }, + { + "epoch": 5.81, + "learning_rate": 1.258318165271205e-06, + "loss": 1.2913, + "step": 31946 + }, + { + "epoch": 5.81, + "learning_rate": 1.2582635602475429e-06, + "loss": 0.0148, + "step": 31947 + }, + { + "epoch": 5.81, + "learning_rate": 1.2582089552238806e-06, + "loss": 1.0558, + "step": 31948 + }, + { + "epoch": 5.81, + "learning_rate": 1.2581543502002183e-06, + "loss": 0.0089, + "step": 31949 + }, + { + "epoch": 5.81, + "learning_rate": 1.2580997451765563e-06, + "loss": 1.1817, + "step": 31950 + }, + { + "epoch": 5.81, + "learning_rate": 1.2580451401528942e-06, + "loss": 0.0035, + "step": 31951 + }, + { + "epoch": 5.81, + "learning_rate": 1.257990535129232e-06, + "loss": 0.0126, + "step": 31952 + }, + { + "epoch": 5.81, + "learning_rate": 1.2579359301055697e-06, + "loss": 0.5208, + "step": 31953 + }, + { + "epoch": 5.81, + "learning_rate": 1.2578813250819076e-06, + "loss": 0.3112, + "step": 31954 + }, + { + "epoch": 5.81, + "learning_rate": 1.2578267200582454e-06, + "loss": 0.395, + "step": 31955 + }, + { + "epoch": 5.81, + "learning_rate": 1.257772115034583e-06, + "loss": 0.4063, + "step": 31956 + }, + { + "epoch": 5.81, + "learning_rate": 1.257717510010921e-06, + "loss": 2.4891, + "step": 31957 + }, + { + "epoch": 5.81, + "learning_rate": 1.257662904987259e-06, + "loss": 0.0131, + "step": 31958 + }, + { + "epoch": 5.81, + "learning_rate": 1.2576082999635967e-06, + "loss": 1.8611, + "step": 31959 + }, + { + "epoch": 5.81, + "learning_rate": 1.2575536949399344e-06, + "loss": 0.14, + "step": 31960 + }, + { + "epoch": 5.81, + "learning_rate": 1.2574990899162724e-06, + "loss": 0.0045, + "step": 31961 + }, + { + "epoch": 5.81, + "learning_rate": 1.2574444848926101e-06, + "loss": 0.1456, + "step": 31962 + }, + { + "epoch": 5.81, + "learning_rate": 1.2573898798689479e-06, + "loss": 1.0443, + "step": 31963 + }, + { + "epoch": 5.81, + "learning_rate": 1.2573352748452858e-06, + "loss": 0.2804, + "step": 31964 + }, + { + "epoch": 5.81, + "learning_rate": 1.2572806698216237e-06, + "loss": 0.9717, + "step": 31965 + }, + { + "epoch": 5.81, + "learning_rate": 1.2572260647979615e-06, + "loss": 0.4972, + "step": 31966 + }, + { + "epoch": 5.81, + "learning_rate": 1.2571714597742992e-06, + "loss": 1.0142, + "step": 31967 + }, + { + "epoch": 5.81, + "learning_rate": 1.2571168547506371e-06, + "loss": 0.6396, + "step": 31968 + }, + { + "epoch": 5.81, + "learning_rate": 1.2570622497269749e-06, + "loss": 0.3184, + "step": 31969 + }, + { + "epoch": 5.81, + "learning_rate": 1.2570076447033126e-06, + "loss": 0.9847, + "step": 31970 + }, + { + "epoch": 5.81, + "learning_rate": 1.2569530396796506e-06, + "loss": 0.3986, + "step": 31971 + }, + { + "epoch": 5.81, + "learning_rate": 1.2568984346559885e-06, + "loss": 0.1388, + "step": 31972 + }, + { + "epoch": 5.81, + "learning_rate": 1.2568438296323262e-06, + "loss": 0.1064, + "step": 31973 + }, + { + "epoch": 5.81, + "learning_rate": 1.256789224608664e-06, + "loss": 0.067, + "step": 31974 + }, + { + "epoch": 5.81, + "learning_rate": 1.256734619585002e-06, + "loss": 0.015, + "step": 31975 + }, + { + "epoch": 5.81, + "learning_rate": 1.2566800145613396e-06, + "loss": 1.7492, + "step": 31976 + }, + { + "epoch": 5.82, + "learning_rate": 1.2566254095376774e-06, + "loss": 0.0629, + "step": 31977 + }, + { + "epoch": 5.82, + "learning_rate": 1.2565708045140153e-06, + "loss": 0.391, + "step": 31978 + }, + { + "epoch": 5.82, + "learning_rate": 1.2565161994903533e-06, + "loss": 0.9322, + "step": 31979 + }, + { + "epoch": 5.82, + "learning_rate": 1.256461594466691e-06, + "loss": 0.008, + "step": 31980 + }, + { + "epoch": 5.82, + "learning_rate": 1.2564069894430287e-06, + "loss": 0.0106, + "step": 31981 + }, + { + "epoch": 5.82, + "learning_rate": 1.2563523844193667e-06, + "loss": 0.3897, + "step": 31982 + }, + { + "epoch": 5.82, + "learning_rate": 1.2562977793957044e-06, + "loss": 0.6867, + "step": 31983 + }, + { + "epoch": 5.82, + "learning_rate": 1.2562431743720421e-06, + "loss": 0.1718, + "step": 31984 + }, + { + "epoch": 5.82, + "learning_rate": 1.25618856934838e-06, + "loss": 1.6415, + "step": 31985 + }, + { + "epoch": 5.82, + "learning_rate": 1.256133964324718e-06, + "loss": 0.0093, + "step": 31986 + }, + { + "epoch": 5.82, + "learning_rate": 1.2560793593010557e-06, + "loss": 0.4741, + "step": 31987 + }, + { + "epoch": 5.82, + "learning_rate": 1.2560247542773935e-06, + "loss": 0.1992, + "step": 31988 + }, + { + "epoch": 5.82, + "learning_rate": 1.2559701492537314e-06, + "loss": 0.3796, + "step": 31989 + }, + { + "epoch": 5.82, + "learning_rate": 1.2559155442300692e-06, + "loss": 0.0125, + "step": 31990 + }, + { + "epoch": 5.82, + "learning_rate": 1.2558609392064069e-06, + "loss": 0.0079, + "step": 31991 + }, + { + "epoch": 5.82, + "learning_rate": 1.2558063341827448e-06, + "loss": 0.198, + "step": 31992 + }, + { + "epoch": 5.82, + "learning_rate": 1.2557517291590828e-06, + "loss": 0.9249, + "step": 31993 + }, + { + "epoch": 5.82, + "learning_rate": 1.2556971241354205e-06, + "loss": 0.007, + "step": 31994 + }, + { + "epoch": 5.82, + "learning_rate": 1.2556425191117582e-06, + "loss": 0.2033, + "step": 31995 + }, + { + "epoch": 5.82, + "learning_rate": 1.2555879140880962e-06, + "loss": 0.9027, + "step": 31996 + }, + { + "epoch": 5.82, + "learning_rate": 1.255533309064434e-06, + "loss": 0.0094, + "step": 31997 + }, + { + "epoch": 5.82, + "learning_rate": 1.2554787040407719e-06, + "loss": 0.0475, + "step": 31998 + }, + { + "epoch": 5.82, + "learning_rate": 1.2554240990171096e-06, + "loss": 1.1891, + "step": 31999 + }, + { + "epoch": 5.82, + "learning_rate": 1.2553694939934475e-06, + "loss": 0.2325, + "step": 32000 + }, + { + "epoch": 5.82, + "learning_rate": 1.2553148889697853e-06, + "loss": 0.1065, + "step": 32001 + }, + { + "epoch": 5.82, + "learning_rate": 1.255260283946123e-06, + "loss": 0.0051, + "step": 32002 + }, + { + "epoch": 5.82, + "learning_rate": 1.255205678922461e-06, + "loss": 0.15, + "step": 32003 + }, + { + "epoch": 5.82, + "learning_rate": 1.2551510738987987e-06, + "loss": 0.2984, + "step": 32004 + }, + { + "epoch": 5.82, + "learning_rate": 1.2550964688751366e-06, + "loss": 1.9786, + "step": 32005 + }, + { + "epoch": 5.82, + "learning_rate": 1.2550418638514743e-06, + "loss": 1.2331, + "step": 32006 + }, + { + "epoch": 5.82, + "learning_rate": 1.2549872588278123e-06, + "loss": 0.0088, + "step": 32007 + }, + { + "epoch": 5.82, + "learning_rate": 1.25493265380415e-06, + "loss": 0.5653, + "step": 32008 + }, + { + "epoch": 5.82, + "learning_rate": 1.2548780487804877e-06, + "loss": 1.7349, + "step": 32009 + }, + { + "epoch": 5.82, + "learning_rate": 1.2548234437568257e-06, + "loss": 0.4692, + "step": 32010 + }, + { + "epoch": 5.82, + "learning_rate": 1.2547688387331634e-06, + "loss": 0.3153, + "step": 32011 + }, + { + "epoch": 5.82, + "learning_rate": 1.2547142337095014e-06, + "loss": 0.6387, + "step": 32012 + }, + { + "epoch": 5.82, + "learning_rate": 1.254659628685839e-06, + "loss": 0.0957, + "step": 32013 + }, + { + "epoch": 5.82, + "learning_rate": 1.254605023662177e-06, + "loss": 0.2416, + "step": 32014 + }, + { + "epoch": 5.82, + "learning_rate": 1.2545504186385148e-06, + "loss": 3.2679, + "step": 32015 + }, + { + "epoch": 5.82, + "learning_rate": 1.2544958136148525e-06, + "loss": 0.0147, + "step": 32016 + }, + { + "epoch": 5.82, + "learning_rate": 1.2544412085911904e-06, + "loss": 0.5458, + "step": 32017 + }, + { + "epoch": 5.82, + "learning_rate": 1.2543866035675282e-06, + "loss": 0.1836, + "step": 32018 + }, + { + "epoch": 5.82, + "learning_rate": 1.2543319985438661e-06, + "loss": 5.9696, + "step": 32019 + }, + { + "epoch": 5.82, + "learning_rate": 1.2542773935202039e-06, + "loss": 1.6794, + "step": 32020 + }, + { + "epoch": 5.82, + "learning_rate": 1.2542227884965418e-06, + "loss": 0.3669, + "step": 32021 + }, + { + "epoch": 5.82, + "learning_rate": 1.2541681834728795e-06, + "loss": 0.2898, + "step": 32022 + }, + { + "epoch": 5.82, + "learning_rate": 1.2541135784492173e-06, + "loss": 0.3273, + "step": 32023 + }, + { + "epoch": 5.82, + "learning_rate": 1.2540589734255552e-06, + "loss": 0.4666, + "step": 32024 + }, + { + "epoch": 5.82, + "learning_rate": 1.254004368401893e-06, + "loss": 0.3412, + "step": 32025 + }, + { + "epoch": 5.82, + "learning_rate": 1.2539497633782309e-06, + "loss": 2.817, + "step": 32026 + }, + { + "epoch": 5.82, + "learning_rate": 1.2538951583545686e-06, + "loss": 7.5958, + "step": 32027 + }, + { + "epoch": 5.82, + "learning_rate": 1.2538405533309066e-06, + "loss": 0.0073, + "step": 32028 + }, + { + "epoch": 5.82, + "learning_rate": 1.2537859483072443e-06, + "loss": 0.0306, + "step": 32029 + }, + { + "epoch": 5.82, + "learning_rate": 1.253731343283582e-06, + "loss": 0.066, + "step": 32030 + }, + { + "epoch": 5.82, + "learning_rate": 1.25367673825992e-06, + "loss": 0.0473, + "step": 32031 + }, + { + "epoch": 5.83, + "learning_rate": 1.2536221332362577e-06, + "loss": 0.0114, + "step": 32032 + }, + { + "epoch": 5.83, + "learning_rate": 1.2535675282125956e-06, + "loss": 0.2075, + "step": 32033 + }, + { + "epoch": 5.83, + "learning_rate": 1.2535129231889334e-06, + "loss": 0.0423, + "step": 32034 + }, + { + "epoch": 5.83, + "learning_rate": 1.2534583181652713e-06, + "loss": 0.3101, + "step": 32035 + }, + { + "epoch": 5.83, + "learning_rate": 1.253403713141609e-06, + "loss": 0.9497, + "step": 32036 + }, + { + "epoch": 5.83, + "learning_rate": 1.2533491081179468e-06, + "loss": 1.344, + "step": 32037 + }, + { + "epoch": 5.83, + "learning_rate": 1.2532945030942847e-06, + "loss": 0.0136, + "step": 32038 + }, + { + "epoch": 5.83, + "learning_rate": 1.2532398980706227e-06, + "loss": 0.6393, + "step": 32039 + }, + { + "epoch": 5.83, + "learning_rate": 1.2531852930469604e-06, + "loss": 1.1947, + "step": 32040 + }, + { + "epoch": 5.83, + "learning_rate": 1.2531306880232981e-06, + "loss": 0.0793, + "step": 32041 + }, + { + "epoch": 5.83, + "learning_rate": 1.253076082999636e-06, + "loss": 0.3129, + "step": 32042 + }, + { + "epoch": 5.83, + "learning_rate": 1.2530214779759738e-06, + "loss": 1.2818, + "step": 32043 + }, + { + "epoch": 5.83, + "learning_rate": 1.2529668729523115e-06, + "loss": 0.0186, + "step": 32044 + }, + { + "epoch": 5.83, + "learning_rate": 1.2529122679286495e-06, + "loss": 0.1337, + "step": 32045 + }, + { + "epoch": 5.83, + "learning_rate": 1.2528576629049874e-06, + "loss": 0.4029, + "step": 32046 + }, + { + "epoch": 5.83, + "learning_rate": 1.2528030578813251e-06, + "loss": 0.0087, + "step": 32047 + }, + { + "epoch": 5.83, + "learning_rate": 1.2527484528576629e-06, + "loss": 0.057, + "step": 32048 + }, + { + "epoch": 5.83, + "learning_rate": 1.2526938478340008e-06, + "loss": 0.5953, + "step": 32049 + }, + { + "epoch": 5.83, + "learning_rate": 1.2526392428103386e-06, + "loss": 0.0537, + "step": 32050 + }, + { + "epoch": 5.83, + "learning_rate": 1.2525846377866763e-06, + "loss": 0.5946, + "step": 32051 + }, + { + "epoch": 5.83, + "learning_rate": 1.2525300327630142e-06, + "loss": 0.3247, + "step": 32052 + }, + { + "epoch": 5.83, + "learning_rate": 1.2524754277393522e-06, + "loss": 0.4313, + "step": 32053 + }, + { + "epoch": 5.83, + "learning_rate": 1.25242082271569e-06, + "loss": 0.0217, + "step": 32054 + }, + { + "epoch": 5.83, + "learning_rate": 1.2523662176920276e-06, + "loss": 1.674, + "step": 32055 + }, + { + "epoch": 5.83, + "learning_rate": 1.2523116126683656e-06, + "loss": 1.2172, + "step": 32056 + }, + { + "epoch": 5.83, + "learning_rate": 1.2522570076447033e-06, + "loss": 0.3167, + "step": 32057 + }, + { + "epoch": 5.83, + "learning_rate": 1.252202402621041e-06, + "loss": 0.4843, + "step": 32058 + }, + { + "epoch": 5.83, + "learning_rate": 1.252147797597379e-06, + "loss": 1.6714, + "step": 32059 + }, + { + "epoch": 5.83, + "learning_rate": 1.252093192573717e-06, + "loss": 0.4301, + "step": 32060 + }, + { + "epoch": 5.83, + "learning_rate": 1.2520385875500547e-06, + "loss": 0.0249, + "step": 32061 + }, + { + "epoch": 5.83, + "learning_rate": 1.2519839825263924e-06, + "loss": 0.1887, + "step": 32062 + }, + { + "epoch": 5.83, + "learning_rate": 1.2519293775027303e-06, + "loss": 0.8369, + "step": 32063 + }, + { + "epoch": 5.83, + "learning_rate": 1.251874772479068e-06, + "loss": 0.1447, + "step": 32064 + }, + { + "epoch": 5.83, + "learning_rate": 1.2518201674554058e-06, + "loss": 0.3261, + "step": 32065 + }, + { + "epoch": 5.83, + "learning_rate": 1.2517655624317437e-06, + "loss": 1.4967, + "step": 32066 + }, + { + "epoch": 5.83, + "learning_rate": 1.2517109574080817e-06, + "loss": 0.1176, + "step": 32067 + }, + { + "epoch": 5.83, + "learning_rate": 1.2516563523844194e-06, + "loss": 0.0337, + "step": 32068 + }, + { + "epoch": 5.83, + "learning_rate": 1.2516017473607572e-06, + "loss": 0.0187, + "step": 32069 + }, + { + "epoch": 5.83, + "learning_rate": 1.251547142337095e-06, + "loss": 0.6009, + "step": 32070 + }, + { + "epoch": 5.83, + "learning_rate": 1.2514925373134328e-06, + "loss": 0.0097, + "step": 32071 + }, + { + "epoch": 5.83, + "learning_rate": 1.2514379322897706e-06, + "loss": 0.5272, + "step": 32072 + }, + { + "epoch": 5.83, + "learning_rate": 1.2513833272661085e-06, + "loss": 0.3543, + "step": 32073 + }, + { + "epoch": 5.83, + "learning_rate": 1.2513287222424464e-06, + "loss": 0.0938, + "step": 32074 + }, + { + "epoch": 5.83, + "learning_rate": 1.2512741172187842e-06, + "loss": 0.0272, + "step": 32075 + }, + { + "epoch": 5.83, + "learning_rate": 1.251219512195122e-06, + "loss": 0.0135, + "step": 32076 + }, + { + "epoch": 5.83, + "learning_rate": 1.2511649071714599e-06, + "loss": 1.6086, + "step": 32077 + }, + { + "epoch": 5.83, + "learning_rate": 1.2511103021477976e-06, + "loss": 0.1401, + "step": 32078 + }, + { + "epoch": 5.83, + "learning_rate": 1.2510556971241355e-06, + "loss": 0.0122, + "step": 32079 + }, + { + "epoch": 5.83, + "learning_rate": 1.2510010921004733e-06, + "loss": 0.3951, + "step": 32080 + }, + { + "epoch": 5.83, + "learning_rate": 1.2509464870768112e-06, + "loss": 1.6494, + "step": 32081 + }, + { + "epoch": 5.83, + "learning_rate": 1.250891882053149e-06, + "loss": 0.0067, + "step": 32082 + }, + { + "epoch": 5.83, + "learning_rate": 1.2508372770294867e-06, + "loss": 0.1467, + "step": 32083 + }, + { + "epoch": 5.83, + "learning_rate": 1.2507826720058246e-06, + "loss": 0.1201, + "step": 32084 + }, + { + "epoch": 5.83, + "learning_rate": 1.2507280669821623e-06, + "loss": 0.0122, + "step": 32085 + }, + { + "epoch": 5.83, + "learning_rate": 1.2506734619585003e-06, + "loss": 0.2535, + "step": 32086 + }, + { + "epoch": 5.84, + "learning_rate": 1.250618856934838e-06, + "loss": 0.4151, + "step": 32087 + }, + { + "epoch": 5.84, + "learning_rate": 1.250564251911176e-06, + "loss": 0.008, + "step": 32088 + }, + { + "epoch": 5.84, + "learning_rate": 1.2505096468875137e-06, + "loss": 1.483, + "step": 32089 + }, + { + "epoch": 5.84, + "learning_rate": 1.2504550418638514e-06, + "loss": 0.4743, + "step": 32090 + }, + { + "epoch": 5.84, + "learning_rate": 1.2504004368401894e-06, + "loss": 0.0104, + "step": 32091 + }, + { + "epoch": 5.84, + "learning_rate": 1.250345831816527e-06, + "loss": 0.1657, + "step": 32092 + }, + { + "epoch": 5.84, + "learning_rate": 1.250291226792865e-06, + "loss": 0.03, + "step": 32093 + }, + { + "epoch": 5.84, + "learning_rate": 1.2502366217692028e-06, + "loss": 1.421, + "step": 32094 + }, + { + "epoch": 5.84, + "learning_rate": 1.2501820167455407e-06, + "loss": 1.3225, + "step": 32095 + }, + { + "epoch": 5.84, + "learning_rate": 1.2501274117218784e-06, + "loss": 0.0067, + "step": 32096 + }, + { + "epoch": 5.84, + "learning_rate": 1.2500728066982162e-06, + "loss": 2.0122, + "step": 32097 + }, + { + "epoch": 5.84, + "learning_rate": 1.2500182016745541e-06, + "loss": 0.6858, + "step": 32098 + }, + { + "epoch": 5.84, + "learning_rate": 1.2499635966508919e-06, + "loss": 0.1192, + "step": 32099 + }, + { + "epoch": 5.84, + "learning_rate": 1.2499089916272298e-06, + "loss": 2.6644, + "step": 32100 + }, + { + "epoch": 5.84, + "learning_rate": 1.2498543866035675e-06, + "loss": 0.4832, + "step": 32101 + }, + { + "epoch": 5.84, + "learning_rate": 1.2497997815799055e-06, + "loss": 0.0057, + "step": 32102 + }, + { + "epoch": 5.84, + "learning_rate": 1.2497451765562432e-06, + "loss": 0.8699, + "step": 32103 + }, + { + "epoch": 5.84, + "learning_rate": 1.249690571532581e-06, + "loss": 0.0117, + "step": 32104 + }, + { + "epoch": 5.84, + "learning_rate": 1.2496359665089189e-06, + "loss": 1.0986, + "step": 32105 + }, + { + "epoch": 5.84, + "learning_rate": 1.2495813614852566e-06, + "loss": 2.6673, + "step": 32106 + }, + { + "epoch": 5.84, + "learning_rate": 1.2495267564615946e-06, + "loss": 0.2882, + "step": 32107 + }, + { + "epoch": 5.84, + "learning_rate": 1.2494721514379323e-06, + "loss": 0.2515, + "step": 32108 + }, + { + "epoch": 5.84, + "learning_rate": 1.2494175464142702e-06, + "loss": 0.3861, + "step": 32109 + }, + { + "epoch": 5.84, + "learning_rate": 1.249362941390608e-06, + "loss": 1.4095, + "step": 32110 + }, + { + "epoch": 5.84, + "learning_rate": 1.2493083363669457e-06, + "loss": 0.0074, + "step": 32111 + }, + { + "epoch": 5.84, + "learning_rate": 1.2492537313432836e-06, + "loss": 0.0065, + "step": 32112 + }, + { + "epoch": 5.84, + "learning_rate": 1.2491991263196214e-06, + "loss": 0.3613, + "step": 32113 + }, + { + "epoch": 5.84, + "learning_rate": 1.2491445212959593e-06, + "loss": 0.0317, + "step": 32114 + }, + { + "epoch": 5.84, + "learning_rate": 1.249089916272297e-06, + "loss": 0.9469, + "step": 32115 + }, + { + "epoch": 5.84, + "learning_rate": 1.249035311248635e-06, + "loss": 0.3219, + "step": 32116 + }, + { + "epoch": 5.84, + "learning_rate": 1.2489807062249727e-06, + "loss": 0.1892, + "step": 32117 + }, + { + "epoch": 5.84, + "learning_rate": 1.2489261012013104e-06, + "loss": 0.3777, + "step": 32118 + }, + { + "epoch": 5.84, + "learning_rate": 1.2488714961776484e-06, + "loss": 0.465, + "step": 32119 + }, + { + "epoch": 5.84, + "learning_rate": 1.2488168911539861e-06, + "loss": 0.1793, + "step": 32120 + }, + { + "epoch": 5.84, + "learning_rate": 1.248762286130324e-06, + "loss": 0.11, + "step": 32121 + }, + { + "epoch": 5.84, + "learning_rate": 1.2487076811066618e-06, + "loss": 0.5394, + "step": 32122 + }, + { + "epoch": 5.84, + "learning_rate": 1.2486530760829997e-06, + "loss": 0.668, + "step": 32123 + }, + { + "epoch": 5.84, + "learning_rate": 1.2485984710593375e-06, + "loss": 1.1565, + "step": 32124 + }, + { + "epoch": 5.84, + "learning_rate": 1.2485438660356752e-06, + "loss": 0.0415, + "step": 32125 + }, + { + "epoch": 5.84, + "learning_rate": 1.2484892610120131e-06, + "loss": 0.0077, + "step": 32126 + }, + { + "epoch": 5.84, + "learning_rate": 1.248434655988351e-06, + "loss": 0.0463, + "step": 32127 + }, + { + "epoch": 5.84, + "learning_rate": 1.2483800509646888e-06, + "loss": 0.0295, + "step": 32128 + }, + { + "epoch": 5.84, + "learning_rate": 1.2483254459410266e-06, + "loss": 0.4074, + "step": 32129 + }, + { + "epoch": 5.84, + "learning_rate": 1.2482708409173645e-06, + "loss": 0.0239, + "step": 32130 + }, + { + "epoch": 5.84, + "learning_rate": 1.2482162358937022e-06, + "loss": 0.1632, + "step": 32131 + }, + { + "epoch": 5.84, + "learning_rate": 1.24816163087004e-06, + "loss": 1.4292, + "step": 32132 + }, + { + "epoch": 5.84, + "learning_rate": 1.248107025846378e-06, + "loss": 0.0994, + "step": 32133 + }, + { + "epoch": 5.84, + "learning_rate": 1.2480524208227158e-06, + "loss": 0.3615, + "step": 32134 + }, + { + "epoch": 5.84, + "learning_rate": 1.2479978157990536e-06, + "loss": 3.41, + "step": 32135 + }, + { + "epoch": 5.84, + "learning_rate": 1.2479432107753913e-06, + "loss": 1.1613, + "step": 32136 + }, + { + "epoch": 5.84, + "learning_rate": 1.2478886057517293e-06, + "loss": 1.9025, + "step": 32137 + }, + { + "epoch": 5.84, + "learning_rate": 1.247834000728067e-06, + "loss": 0.1629, + "step": 32138 + }, + { + "epoch": 5.84, + "learning_rate": 1.2477793957044047e-06, + "loss": 0.6417, + "step": 32139 + }, + { + "epoch": 5.84, + "learning_rate": 1.2477247906807427e-06, + "loss": 0.5479, + "step": 32140 + }, + { + "epoch": 5.84, + "learning_rate": 1.2476701856570806e-06, + "loss": 0.4535, + "step": 32141 + }, + { + "epoch": 5.85, + "learning_rate": 1.2476155806334183e-06, + "loss": 0.1435, + "step": 32142 + }, + { + "epoch": 5.85, + "learning_rate": 1.247560975609756e-06, + "loss": 0.0097, + "step": 32143 + }, + { + "epoch": 5.85, + "learning_rate": 1.247506370586094e-06, + "loss": 0.8272, + "step": 32144 + }, + { + "epoch": 5.85, + "learning_rate": 1.2474517655624317e-06, + "loss": 1.6616, + "step": 32145 + }, + { + "epoch": 5.85, + "learning_rate": 1.2473971605387695e-06, + "loss": 0.4096, + "step": 32146 + }, + { + "epoch": 5.85, + "learning_rate": 1.2473425555151074e-06, + "loss": 0.457, + "step": 32147 + }, + { + "epoch": 5.85, + "learning_rate": 1.2472879504914454e-06, + "loss": 0.5749, + "step": 32148 + }, + { + "epoch": 5.85, + "learning_rate": 1.247233345467783e-06, + "loss": 0.0946, + "step": 32149 + }, + { + "epoch": 5.85, + "learning_rate": 1.2471787404441208e-06, + "loss": 0.3082, + "step": 32150 + }, + { + "epoch": 5.85, + "learning_rate": 1.2471241354204588e-06, + "loss": 0.6105, + "step": 32151 + }, + { + "epoch": 5.85, + "learning_rate": 1.2470695303967965e-06, + "loss": 0.1863, + "step": 32152 + }, + { + "epoch": 5.85, + "learning_rate": 1.2470149253731342e-06, + "loss": 3.0753, + "step": 32153 + }, + { + "epoch": 5.85, + "learning_rate": 1.2469603203494722e-06, + "loss": 1.2875, + "step": 32154 + }, + { + "epoch": 5.85, + "learning_rate": 1.2469057153258101e-06, + "loss": 0.0373, + "step": 32155 + }, + { + "epoch": 5.85, + "learning_rate": 1.2468511103021479e-06, + "loss": 0.2726, + "step": 32156 + }, + { + "epoch": 5.85, + "learning_rate": 1.2467965052784856e-06, + "loss": 3.0325, + "step": 32157 + }, + { + "epoch": 5.85, + "learning_rate": 1.2467419002548235e-06, + "loss": 1.2831, + "step": 32158 + }, + { + "epoch": 5.85, + "learning_rate": 1.2466872952311613e-06, + "loss": 0.6413, + "step": 32159 + }, + { + "epoch": 5.85, + "learning_rate": 1.246632690207499e-06, + "loss": 0.1282, + "step": 32160 + }, + { + "epoch": 5.85, + "learning_rate": 1.246578085183837e-06, + "loss": 0.0575, + "step": 32161 + }, + { + "epoch": 5.85, + "learning_rate": 1.2465234801601749e-06, + "loss": 0.0572, + "step": 32162 + }, + { + "epoch": 5.85, + "learning_rate": 1.2464688751365126e-06, + "loss": 0.1467, + "step": 32163 + }, + { + "epoch": 5.85, + "learning_rate": 1.2464142701128503e-06, + "loss": 0.4631, + "step": 32164 + }, + { + "epoch": 5.85, + "learning_rate": 1.2463596650891883e-06, + "loss": 0.1394, + "step": 32165 + }, + { + "epoch": 5.85, + "learning_rate": 1.246305060065526e-06, + "loss": 0.604, + "step": 32166 + }, + { + "epoch": 5.85, + "learning_rate": 1.246250455041864e-06, + "loss": 1.1123, + "step": 32167 + }, + { + "epoch": 5.85, + "learning_rate": 1.2461958500182017e-06, + "loss": 0.1619, + "step": 32168 + }, + { + "epoch": 5.85, + "learning_rate": 1.2461412449945396e-06, + "loss": 0.9087, + "step": 32169 + }, + { + "epoch": 5.85, + "learning_rate": 1.2460866399708774e-06, + "loss": 1.4145, + "step": 32170 + }, + { + "epoch": 5.85, + "learning_rate": 1.246032034947215e-06, + "loss": 0.4871, + "step": 32171 + }, + { + "epoch": 5.85, + "learning_rate": 1.245977429923553e-06, + "loss": 0.0416, + "step": 32172 + }, + { + "epoch": 5.85, + "learning_rate": 1.2459228248998908e-06, + "loss": 1.0371, + "step": 32173 + }, + { + "epoch": 5.85, + "learning_rate": 1.2458682198762287e-06, + "loss": 0.4316, + "step": 32174 + }, + { + "epoch": 5.85, + "learning_rate": 1.2458136148525664e-06, + "loss": 0.1907, + "step": 32175 + }, + { + "epoch": 5.85, + "learning_rate": 1.2457590098289044e-06, + "loss": 0.0825, + "step": 32176 + }, + { + "epoch": 5.85, + "learning_rate": 1.2457044048052421e-06, + "loss": 0.2281, + "step": 32177 + }, + { + "epoch": 5.85, + "learning_rate": 1.2456497997815799e-06, + "loss": 0.0112, + "step": 32178 + }, + { + "epoch": 5.85, + "learning_rate": 1.2455951947579178e-06, + "loss": 0.9543, + "step": 32179 + }, + { + "epoch": 5.85, + "learning_rate": 1.2455405897342555e-06, + "loss": 0.4833, + "step": 32180 + }, + { + "epoch": 5.85, + "learning_rate": 1.2454859847105935e-06, + "loss": 0.5587, + "step": 32181 + }, + { + "epoch": 5.85, + "learning_rate": 1.2454313796869312e-06, + "loss": 1.6385, + "step": 32182 + }, + { + "epoch": 5.85, + "learning_rate": 1.2453767746632691e-06, + "loss": 0.0094, + "step": 32183 + }, + { + "epoch": 5.85, + "learning_rate": 1.2453221696396069e-06, + "loss": 0.0253, + "step": 32184 + }, + { + "epoch": 5.85, + "learning_rate": 1.2452675646159446e-06, + "loss": 0.7547, + "step": 32185 + }, + { + "epoch": 5.85, + "learning_rate": 1.2452129595922826e-06, + "loss": 0.0368, + "step": 32186 + }, + { + "epoch": 5.85, + "learning_rate": 1.2451583545686203e-06, + "loss": 0.5336, + "step": 32187 + }, + { + "epoch": 5.85, + "learning_rate": 1.2451037495449582e-06, + "loss": 0.0922, + "step": 32188 + }, + { + "epoch": 5.85, + "learning_rate": 1.245049144521296e-06, + "loss": 0.7291, + "step": 32189 + }, + { + "epoch": 5.85, + "learning_rate": 1.244994539497634e-06, + "loss": 0.0465, + "step": 32190 + }, + { + "epoch": 5.85, + "learning_rate": 1.2449399344739716e-06, + "loss": 0.4305, + "step": 32191 + }, + { + "epoch": 5.85, + "learning_rate": 1.2448853294503094e-06, + "loss": 0.1461, + "step": 32192 + }, + { + "epoch": 5.85, + "learning_rate": 1.2448307244266473e-06, + "loss": 0.3162, + "step": 32193 + }, + { + "epoch": 5.85, + "learning_rate": 1.244776119402985e-06, + "loss": 0.7321, + "step": 32194 + }, + { + "epoch": 5.85, + "learning_rate": 1.244721514379323e-06, + "loss": 0.0087, + "step": 32195 + }, + { + "epoch": 5.85, + "learning_rate": 1.2446669093556607e-06, + "loss": 2.0377, + "step": 32196 + }, + { + "epoch": 5.86, + "learning_rate": 1.2446123043319987e-06, + "loss": 1.9463, + "step": 32197 + }, + { + "epoch": 5.86, + "learning_rate": 1.2445576993083364e-06, + "loss": 0.4912, + "step": 32198 + }, + { + "epoch": 5.86, + "learning_rate": 1.2445030942846741e-06, + "loss": 1.4058, + "step": 32199 + }, + { + "epoch": 5.86, + "learning_rate": 1.244448489261012e-06, + "loss": 0.6171, + "step": 32200 + }, + { + "epoch": 5.86, + "learning_rate": 1.2443938842373498e-06, + "loss": 0.1351, + "step": 32201 + }, + { + "epoch": 5.86, + "learning_rate": 1.2443392792136877e-06, + "loss": 1.7538, + "step": 32202 + }, + { + "epoch": 5.86, + "learning_rate": 1.2442846741900255e-06, + "loss": 0.0233, + "step": 32203 + }, + { + "epoch": 5.86, + "learning_rate": 1.2442300691663634e-06, + "loss": 0.6415, + "step": 32204 + }, + { + "epoch": 5.86, + "learning_rate": 1.2441754641427011e-06, + "loss": 0.4307, + "step": 32205 + }, + { + "epoch": 5.86, + "learning_rate": 1.2441208591190389e-06, + "loss": 0.057, + "step": 32206 + }, + { + "epoch": 5.86, + "learning_rate": 1.2440662540953768e-06, + "loss": 0.0213, + "step": 32207 + }, + { + "epoch": 5.86, + "learning_rate": 1.2440116490717148e-06, + "loss": 0.7841, + "step": 32208 + }, + { + "epoch": 5.86, + "learning_rate": 1.2439570440480525e-06, + "loss": 0.0699, + "step": 32209 + }, + { + "epoch": 5.86, + "learning_rate": 1.2439024390243902e-06, + "loss": 0.1665, + "step": 32210 + }, + { + "epoch": 5.86, + "learning_rate": 1.2438478340007282e-06, + "loss": 0.0554, + "step": 32211 + }, + { + "epoch": 5.86, + "learning_rate": 1.243793228977066e-06, + "loss": 0.4746, + "step": 32212 + }, + { + "epoch": 5.86, + "learning_rate": 1.2437386239534036e-06, + "loss": 0.3597, + "step": 32213 + }, + { + "epoch": 5.86, + "learning_rate": 1.2436840189297416e-06, + "loss": 0.4998, + "step": 32214 + }, + { + "epoch": 5.86, + "learning_rate": 1.2436294139060795e-06, + "loss": 0.3047, + "step": 32215 + }, + { + "epoch": 5.86, + "learning_rate": 1.2435748088824173e-06, + "loss": 0.1821, + "step": 32216 + }, + { + "epoch": 5.86, + "learning_rate": 1.243520203858755e-06, + "loss": 4.6796, + "step": 32217 + }, + { + "epoch": 5.86, + "learning_rate": 1.243465598835093e-06, + "loss": 0.1338, + "step": 32218 + }, + { + "epoch": 5.86, + "learning_rate": 1.2434109938114307e-06, + "loss": 0.2886, + "step": 32219 + }, + { + "epoch": 5.86, + "learning_rate": 1.2433563887877684e-06, + "loss": 0.0305, + "step": 32220 + }, + { + "epoch": 5.86, + "learning_rate": 1.2433017837641063e-06, + "loss": 0.006, + "step": 32221 + }, + { + "epoch": 5.86, + "learning_rate": 1.2432471787404443e-06, + "loss": 0.1903, + "step": 32222 + }, + { + "epoch": 5.86, + "learning_rate": 1.243192573716782e-06, + "loss": 0.0548, + "step": 32223 + }, + { + "epoch": 5.86, + "learning_rate": 1.2431379686931197e-06, + "loss": 1.1297, + "step": 32224 + }, + { + "epoch": 5.86, + "learning_rate": 1.2430833636694577e-06, + "loss": 1.0072, + "step": 32225 + }, + { + "epoch": 5.86, + "learning_rate": 1.2430287586457954e-06, + "loss": 0.8165, + "step": 32226 + }, + { + "epoch": 5.86, + "learning_rate": 1.2429741536221331e-06, + "loss": 0.1005, + "step": 32227 + }, + { + "epoch": 5.86, + "learning_rate": 1.242919548598471e-06, + "loss": 0.0368, + "step": 32228 + }, + { + "epoch": 5.86, + "learning_rate": 1.242864943574809e-06, + "loss": 1.828, + "step": 32229 + }, + { + "epoch": 5.86, + "learning_rate": 1.2428103385511468e-06, + "loss": 1.0447, + "step": 32230 + }, + { + "epoch": 5.86, + "learning_rate": 1.2427557335274845e-06, + "loss": 1.0667, + "step": 32231 + }, + { + "epoch": 5.86, + "learning_rate": 1.2427011285038224e-06, + "loss": 0.757, + "step": 32232 + }, + { + "epoch": 5.86, + "learning_rate": 1.2426465234801602e-06, + "loss": 0.0045, + "step": 32233 + }, + { + "epoch": 5.86, + "learning_rate": 1.242591918456498e-06, + "loss": 0.6967, + "step": 32234 + }, + { + "epoch": 5.86, + "learning_rate": 1.2425373134328358e-06, + "loss": 1.3791, + "step": 32235 + }, + { + "epoch": 5.86, + "learning_rate": 1.2424827084091738e-06, + "loss": 0.087, + "step": 32236 + }, + { + "epoch": 5.86, + "learning_rate": 1.2424281033855115e-06, + "loss": 0.8142, + "step": 32237 + }, + { + "epoch": 5.86, + "learning_rate": 1.2423734983618493e-06, + "loss": 0.0333, + "step": 32238 + }, + { + "epoch": 5.86, + "learning_rate": 1.2423188933381872e-06, + "loss": 0.033, + "step": 32239 + }, + { + "epoch": 5.86, + "learning_rate": 1.242264288314525e-06, + "loss": 0.1253, + "step": 32240 + }, + { + "epoch": 5.86, + "learning_rate": 1.2422096832908627e-06, + "loss": 3.6436, + "step": 32241 + }, + { + "epoch": 5.86, + "learning_rate": 1.2421550782672006e-06, + "loss": 0.3322, + "step": 32242 + }, + { + "epoch": 5.86, + "learning_rate": 1.2421004732435385e-06, + "loss": 0.1526, + "step": 32243 + }, + { + "epoch": 5.86, + "learning_rate": 1.2420458682198763e-06, + "loss": 0.3383, + "step": 32244 + }, + { + "epoch": 5.86, + "learning_rate": 1.241991263196214e-06, + "loss": 0.2824, + "step": 32245 + }, + { + "epoch": 5.86, + "learning_rate": 1.241936658172552e-06, + "loss": 0.3227, + "step": 32246 + }, + { + "epoch": 5.86, + "learning_rate": 1.2418820531488897e-06, + "loss": 0.3409, + "step": 32247 + }, + { + "epoch": 5.86, + "learning_rate": 1.2418274481252276e-06, + "loss": 0.124, + "step": 32248 + }, + { + "epoch": 5.86, + "learning_rate": 1.2417728431015654e-06, + "loss": 0.925, + "step": 32249 + }, + { + "epoch": 5.86, + "learning_rate": 1.2417182380779033e-06, + "loss": 0.8357, + "step": 32250 + }, + { + "epoch": 5.86, + "learning_rate": 1.241663633054241e-06, + "loss": 0.1321, + "step": 32251 + }, + { + "epoch": 5.87, + "learning_rate": 1.2416090280305788e-06, + "loss": 0.5644, + "step": 32252 + }, + { + "epoch": 5.87, + "learning_rate": 1.2415544230069167e-06, + "loss": 0.3226, + "step": 32253 + }, + { + "epoch": 5.87, + "learning_rate": 1.2414998179832544e-06, + "loss": 0.4805, + "step": 32254 + }, + { + "epoch": 5.87, + "learning_rate": 1.2414452129595924e-06, + "loss": 0.0053, + "step": 32255 + }, + { + "epoch": 5.87, + "learning_rate": 1.2413906079359301e-06, + "loss": 0.1888, + "step": 32256 + }, + { + "epoch": 5.87, + "learning_rate": 1.241336002912268e-06, + "loss": 0.1159, + "step": 32257 + }, + { + "epoch": 5.87, + "learning_rate": 1.2412813978886058e-06, + "loss": 1.661, + "step": 32258 + }, + { + "epoch": 5.87, + "learning_rate": 1.2412267928649435e-06, + "loss": 0.0659, + "step": 32259 + }, + { + "epoch": 5.87, + "learning_rate": 1.2411721878412815e-06, + "loss": 1.0181, + "step": 32260 + }, + { + "epoch": 5.87, + "learning_rate": 1.2411175828176192e-06, + "loss": 0.1895, + "step": 32261 + }, + { + "epoch": 5.87, + "learning_rate": 1.2410629777939571e-06, + "loss": 0.0069, + "step": 32262 + }, + { + "epoch": 5.87, + "learning_rate": 1.2410083727702949e-06, + "loss": 0.5849, + "step": 32263 + }, + { + "epoch": 5.87, + "learning_rate": 1.2409537677466328e-06, + "loss": 0.0077, + "step": 32264 + }, + { + "epoch": 5.87, + "learning_rate": 1.2408991627229706e-06, + "loss": 0.1332, + "step": 32265 + }, + { + "epoch": 5.87, + "learning_rate": 1.2408445576993083e-06, + "loss": 0.0272, + "step": 32266 + }, + { + "epoch": 5.87, + "learning_rate": 1.2407899526756462e-06, + "loss": 1.8261, + "step": 32267 + }, + { + "epoch": 5.87, + "learning_rate": 1.240735347651984e-06, + "loss": 0.6842, + "step": 32268 + }, + { + "epoch": 5.87, + "learning_rate": 1.240680742628322e-06, + "loss": 0.1054, + "step": 32269 + }, + { + "epoch": 5.87, + "learning_rate": 1.2406261376046596e-06, + "loss": 0.2009, + "step": 32270 + }, + { + "epoch": 5.87, + "learning_rate": 1.2405715325809976e-06, + "loss": 0.0041, + "step": 32271 + }, + { + "epoch": 5.87, + "learning_rate": 1.2405169275573353e-06, + "loss": 0.6426, + "step": 32272 + }, + { + "epoch": 5.87, + "learning_rate": 1.240462322533673e-06, + "loss": 0.0878, + "step": 32273 + }, + { + "epoch": 5.87, + "learning_rate": 1.240407717510011e-06, + "loss": 0.1356, + "step": 32274 + }, + { + "epoch": 5.87, + "learning_rate": 1.2403531124863487e-06, + "loss": 0.9, + "step": 32275 + }, + { + "epoch": 5.87, + "learning_rate": 1.2402985074626867e-06, + "loss": 4.7851, + "step": 32276 + }, + { + "epoch": 5.87, + "learning_rate": 1.2402439024390244e-06, + "loss": 0.0152, + "step": 32277 + }, + { + "epoch": 5.87, + "learning_rate": 1.2401892974153623e-06, + "loss": 1.5833, + "step": 32278 + }, + { + "epoch": 5.87, + "learning_rate": 1.2401346923917e-06, + "loss": 0.1022, + "step": 32279 + }, + { + "epoch": 5.87, + "learning_rate": 1.2400800873680378e-06, + "loss": 0.0086, + "step": 32280 + }, + { + "epoch": 5.87, + "learning_rate": 1.2400254823443757e-06, + "loss": 0.2617, + "step": 32281 + }, + { + "epoch": 5.87, + "learning_rate": 1.2399708773207135e-06, + "loss": 0.0587, + "step": 32282 + }, + { + "epoch": 5.87, + "learning_rate": 1.2399162722970514e-06, + "loss": 0.011, + "step": 32283 + }, + { + "epoch": 5.87, + "learning_rate": 1.2398616672733891e-06, + "loss": 0.2518, + "step": 32284 + }, + { + "epoch": 5.87, + "learning_rate": 1.239807062249727e-06, + "loss": 0.8607, + "step": 32285 + }, + { + "epoch": 5.87, + "learning_rate": 1.2397524572260648e-06, + "loss": 0.0079, + "step": 32286 + }, + { + "epoch": 5.87, + "learning_rate": 1.2396978522024026e-06, + "loss": 1.2247, + "step": 32287 + }, + { + "epoch": 5.87, + "learning_rate": 1.2396432471787405e-06, + "loss": 1.1895, + "step": 32288 + }, + { + "epoch": 5.87, + "learning_rate": 1.2395886421550782e-06, + "loss": 0.171, + "step": 32289 + }, + { + "epoch": 5.87, + "learning_rate": 1.2395340371314162e-06, + "loss": 0.0692, + "step": 32290 + }, + { + "epoch": 5.87, + "learning_rate": 1.239479432107754e-06, + "loss": 0.2459, + "step": 32291 + }, + { + "epoch": 5.87, + "learning_rate": 1.2394248270840918e-06, + "loss": 0.9437, + "step": 32292 + }, + { + "epoch": 5.87, + "learning_rate": 1.2393702220604296e-06, + "loss": 0.0247, + "step": 32293 + }, + { + "epoch": 5.87, + "learning_rate": 1.2393156170367673e-06, + "loss": 0.0158, + "step": 32294 + }, + { + "epoch": 5.87, + "learning_rate": 1.2392610120131053e-06, + "loss": 3.1562, + "step": 32295 + }, + { + "epoch": 5.87, + "learning_rate": 1.2392064069894432e-06, + "loss": 1.0034, + "step": 32296 + }, + { + "epoch": 5.87, + "learning_rate": 1.239151801965781e-06, + "loss": 0.0088, + "step": 32297 + }, + { + "epoch": 5.87, + "learning_rate": 1.2390971969421187e-06, + "loss": 0.7669, + "step": 32298 + }, + { + "epoch": 5.87, + "learning_rate": 1.2390425919184566e-06, + "loss": 0.399, + "step": 32299 + }, + { + "epoch": 5.87, + "learning_rate": 1.2389879868947943e-06, + "loss": 0.0191, + "step": 32300 + }, + { + "epoch": 5.87, + "learning_rate": 1.238933381871132e-06, + "loss": 0.1876, + "step": 32301 + }, + { + "epoch": 5.87, + "learning_rate": 1.23887877684747e-06, + "loss": 0.0913, + "step": 32302 + }, + { + "epoch": 5.87, + "learning_rate": 1.238824171823808e-06, + "loss": 0.7868, + "step": 32303 + }, + { + "epoch": 5.87, + "learning_rate": 1.2387695668001457e-06, + "loss": 0.1086, + "step": 32304 + }, + { + "epoch": 5.87, + "learning_rate": 1.2387149617764834e-06, + "loss": 0.0785, + "step": 32305 + }, + { + "epoch": 5.87, + "learning_rate": 1.2386603567528214e-06, + "loss": 0.1478, + "step": 32306 + }, + { + "epoch": 5.88, + "learning_rate": 1.238605751729159e-06, + "loss": 0.1733, + "step": 32307 + }, + { + "epoch": 5.88, + "learning_rate": 1.2385511467054968e-06, + "loss": 0.0169, + "step": 32308 + }, + { + "epoch": 5.88, + "learning_rate": 1.2384965416818348e-06, + "loss": 0.7402, + "step": 32309 + }, + { + "epoch": 5.88, + "learning_rate": 1.2384419366581727e-06, + "loss": 2.5707, + "step": 32310 + }, + { + "epoch": 5.88, + "learning_rate": 1.2383873316345104e-06, + "loss": 1.0292, + "step": 32311 + }, + { + "epoch": 5.88, + "learning_rate": 1.2383327266108482e-06, + "loss": 0.0582, + "step": 32312 + }, + { + "epoch": 5.88, + "learning_rate": 1.2382781215871861e-06, + "loss": 0.0689, + "step": 32313 + }, + { + "epoch": 5.88, + "learning_rate": 1.2382235165635238e-06, + "loss": 0.2369, + "step": 32314 + }, + { + "epoch": 5.88, + "learning_rate": 1.2381689115398616e-06, + "loss": 0.1299, + "step": 32315 + }, + { + "epoch": 5.88, + "learning_rate": 1.2381143065161995e-06, + "loss": 0.4428, + "step": 32316 + }, + { + "epoch": 5.88, + "learning_rate": 1.2380597014925375e-06, + "loss": 1.1708, + "step": 32317 + }, + { + "epoch": 5.88, + "learning_rate": 1.2380050964688752e-06, + "loss": 0.247, + "step": 32318 + }, + { + "epoch": 5.88, + "learning_rate": 1.237950491445213e-06, + "loss": 0.114, + "step": 32319 + }, + { + "epoch": 5.88, + "learning_rate": 1.2378958864215509e-06, + "loss": 0.0864, + "step": 32320 + }, + { + "epoch": 5.88, + "learning_rate": 1.2378412813978886e-06, + "loss": 0.4717, + "step": 32321 + }, + { + "epoch": 5.88, + "learning_rate": 1.2377866763742263e-06, + "loss": 0.1613, + "step": 32322 + }, + { + "epoch": 5.88, + "learning_rate": 1.2377320713505643e-06, + "loss": 0.0233, + "step": 32323 + }, + { + "epoch": 5.88, + "learning_rate": 1.2376774663269022e-06, + "loss": 2.2206, + "step": 32324 + }, + { + "epoch": 5.88, + "learning_rate": 1.23762286130324e-06, + "loss": 0.0206, + "step": 32325 + }, + { + "epoch": 5.88, + "learning_rate": 1.2375682562795777e-06, + "loss": 0.5894, + "step": 32326 + }, + { + "epoch": 5.88, + "learning_rate": 1.2375136512559156e-06, + "loss": 0.4981, + "step": 32327 + }, + { + "epoch": 5.88, + "learning_rate": 1.2374590462322534e-06, + "loss": 0.3046, + "step": 32328 + }, + { + "epoch": 5.88, + "learning_rate": 1.237404441208591e-06, + "loss": 0.5841, + "step": 32329 + }, + { + "epoch": 5.88, + "learning_rate": 1.237349836184929e-06, + "loss": 2.5146, + "step": 32330 + }, + { + "epoch": 5.88, + "learning_rate": 1.237295231161267e-06, + "loss": 1.4928, + "step": 32331 + }, + { + "epoch": 5.88, + "learning_rate": 1.2372406261376047e-06, + "loss": 0.9524, + "step": 32332 + }, + { + "epoch": 5.88, + "learning_rate": 1.2371860211139424e-06, + "loss": 0.5291, + "step": 32333 + }, + { + "epoch": 5.88, + "learning_rate": 1.2371314160902804e-06, + "loss": 0.0751, + "step": 32334 + }, + { + "epoch": 5.88, + "learning_rate": 1.2370768110666181e-06, + "loss": 1.2146, + "step": 32335 + }, + { + "epoch": 5.88, + "learning_rate": 1.237022206042956e-06, + "loss": 0.7673, + "step": 32336 + }, + { + "epoch": 5.88, + "learning_rate": 1.2369676010192938e-06, + "loss": 0.1998, + "step": 32337 + }, + { + "epoch": 5.88, + "learning_rate": 1.2369129959956317e-06, + "loss": 0.2117, + "step": 32338 + }, + { + "epoch": 5.88, + "learning_rate": 1.2368583909719695e-06, + "loss": 1.0325, + "step": 32339 + }, + { + "epoch": 5.88, + "learning_rate": 1.2368037859483072e-06, + "loss": 0.2969, + "step": 32340 + }, + { + "epoch": 5.88, + "learning_rate": 1.2367491809246451e-06, + "loss": 1.3966, + "step": 32341 + }, + { + "epoch": 5.88, + "learning_rate": 1.2366945759009829e-06, + "loss": 0.0155, + "step": 32342 + }, + { + "epoch": 5.88, + "learning_rate": 1.2366399708773208e-06, + "loss": 1.8601, + "step": 32343 + }, + { + "epoch": 5.88, + "learning_rate": 1.2365853658536586e-06, + "loss": 0.7149, + "step": 32344 + }, + { + "epoch": 5.88, + "learning_rate": 1.2365307608299965e-06, + "loss": 0.0211, + "step": 32345 + }, + { + "epoch": 5.88, + "learning_rate": 1.2364761558063342e-06, + "loss": 1.4591, + "step": 32346 + }, + { + "epoch": 5.88, + "learning_rate": 1.236421550782672e-06, + "loss": 0.0762, + "step": 32347 + }, + { + "epoch": 5.88, + "learning_rate": 1.23636694575901e-06, + "loss": 0.3698, + "step": 32348 + }, + { + "epoch": 5.88, + "learning_rate": 1.2363123407353476e-06, + "loss": 0.3069, + "step": 32349 + }, + { + "epoch": 5.88, + "learning_rate": 1.2362577357116856e-06, + "loss": 1.5887, + "step": 32350 + }, + { + "epoch": 5.88, + "learning_rate": 1.2362031306880233e-06, + "loss": 0.1029, + "step": 32351 + }, + { + "epoch": 5.88, + "learning_rate": 1.2361485256643613e-06, + "loss": 0.0113, + "step": 32352 + }, + { + "epoch": 5.88, + "learning_rate": 1.236093920640699e-06, + "loss": 0.1962, + "step": 32353 + }, + { + "epoch": 5.88, + "learning_rate": 1.2360393156170367e-06, + "loss": 0.5202, + "step": 32354 + }, + { + "epoch": 5.88, + "learning_rate": 1.2359847105933747e-06, + "loss": 0.0172, + "step": 32355 + }, + { + "epoch": 5.88, + "learning_rate": 1.2359301055697124e-06, + "loss": 2.2862, + "step": 32356 + }, + { + "epoch": 5.88, + "learning_rate": 1.2358755005460503e-06, + "loss": 1.5765, + "step": 32357 + }, + { + "epoch": 5.88, + "learning_rate": 1.235820895522388e-06, + "loss": 0.4383, + "step": 32358 + }, + { + "epoch": 5.88, + "learning_rate": 1.235766290498726e-06, + "loss": 0.0989, + "step": 32359 + }, + { + "epoch": 5.88, + "learning_rate": 1.2357116854750637e-06, + "loss": 2.233, + "step": 32360 + }, + { + "epoch": 5.88, + "learning_rate": 1.2356570804514015e-06, + "loss": 0.1646, + "step": 32361 + }, + { + "epoch": 5.89, + "learning_rate": 1.2356024754277394e-06, + "loss": 0.9825, + "step": 32362 + }, + { + "epoch": 5.89, + "learning_rate": 1.2355478704040771e-06, + "loss": 0.0328, + "step": 32363 + }, + { + "epoch": 5.89, + "learning_rate": 1.235493265380415e-06, + "loss": 0.0364, + "step": 32364 + }, + { + "epoch": 5.89, + "learning_rate": 1.2354386603567528e-06, + "loss": 0.0055, + "step": 32365 + }, + { + "epoch": 5.89, + "learning_rate": 1.2353840553330908e-06, + "loss": 0.1718, + "step": 32366 + }, + { + "epoch": 5.89, + "learning_rate": 1.2353294503094285e-06, + "loss": 1.3735, + "step": 32367 + }, + { + "epoch": 5.89, + "learning_rate": 1.2352748452857662e-06, + "loss": 2.2179, + "step": 32368 + }, + { + "epoch": 5.89, + "learning_rate": 1.2352202402621042e-06, + "loss": 1.7477, + "step": 32369 + }, + { + "epoch": 5.89, + "learning_rate": 1.235165635238442e-06, + "loss": 0.0146, + "step": 32370 + }, + { + "epoch": 5.89, + "learning_rate": 1.2351110302147798e-06, + "loss": 0.19, + "step": 32371 + }, + { + "epoch": 5.89, + "learning_rate": 1.2350564251911176e-06, + "loss": 1.2356, + "step": 32372 + }, + { + "epoch": 5.89, + "learning_rate": 1.2350018201674555e-06, + "loss": 0.0171, + "step": 32373 + }, + { + "epoch": 5.89, + "learning_rate": 1.2349472151437933e-06, + "loss": 0.7813, + "step": 32374 + }, + { + "epoch": 5.89, + "learning_rate": 1.234892610120131e-06, + "loss": 0.5881, + "step": 32375 + }, + { + "epoch": 5.89, + "learning_rate": 1.234838005096469e-06, + "loss": 0.2315, + "step": 32376 + }, + { + "epoch": 5.89, + "learning_rate": 1.2347834000728069e-06, + "loss": 0.2686, + "step": 32377 + }, + { + "epoch": 5.89, + "learning_rate": 1.2347287950491446e-06, + "loss": 0.0205, + "step": 32378 + }, + { + "epoch": 5.89, + "learning_rate": 1.2346741900254823e-06, + "loss": 0.182, + "step": 32379 + }, + { + "epoch": 5.89, + "learning_rate": 1.2346195850018203e-06, + "loss": 0.3747, + "step": 32380 + }, + { + "epoch": 5.89, + "learning_rate": 1.234564979978158e-06, + "loss": 0.7538, + "step": 32381 + }, + { + "epoch": 5.89, + "learning_rate": 1.2345103749544957e-06, + "loss": 1.9014, + "step": 32382 + }, + { + "epoch": 5.89, + "learning_rate": 1.2344557699308337e-06, + "loss": 2.7826, + "step": 32383 + }, + { + "epoch": 5.89, + "learning_rate": 1.2344011649071716e-06, + "loss": 1.1318, + "step": 32384 + }, + { + "epoch": 5.89, + "learning_rate": 1.2343465598835094e-06, + "loss": 0.5827, + "step": 32385 + }, + { + "epoch": 5.89, + "learning_rate": 1.234291954859847e-06, + "loss": 0.0163, + "step": 32386 + }, + { + "epoch": 5.89, + "learning_rate": 1.234237349836185e-06, + "loss": 0.8319, + "step": 32387 + }, + { + "epoch": 5.89, + "learning_rate": 1.2341827448125228e-06, + "loss": 0.4003, + "step": 32388 + }, + { + "epoch": 5.89, + "learning_rate": 1.2341281397888605e-06, + "loss": 0.0523, + "step": 32389 + }, + { + "epoch": 5.89, + "learning_rate": 1.2340735347651984e-06, + "loss": 0.0231, + "step": 32390 + }, + { + "epoch": 5.89, + "learning_rate": 1.2340189297415364e-06, + "loss": 0.4367, + "step": 32391 + }, + { + "epoch": 5.89, + "learning_rate": 1.2339643247178741e-06, + "loss": 0.0293, + "step": 32392 + }, + { + "epoch": 5.89, + "learning_rate": 1.2339097196942118e-06, + "loss": 0.6348, + "step": 32393 + }, + { + "epoch": 5.89, + "learning_rate": 1.2338551146705498e-06, + "loss": 0.5574, + "step": 32394 + }, + { + "epoch": 5.89, + "learning_rate": 1.2338005096468875e-06, + "loss": 0.1292, + "step": 32395 + }, + { + "epoch": 5.89, + "learning_rate": 1.2337459046232253e-06, + "loss": 0.2506, + "step": 32396 + }, + { + "epoch": 5.89, + "learning_rate": 1.2336912995995632e-06, + "loss": 0.012, + "step": 32397 + }, + { + "epoch": 5.89, + "learning_rate": 1.2336366945759011e-06, + "loss": 0.1111, + "step": 32398 + }, + { + "epoch": 5.89, + "learning_rate": 1.2335820895522389e-06, + "loss": 0.0102, + "step": 32399 + }, + { + "epoch": 5.89, + "learning_rate": 1.2335274845285766e-06, + "loss": 0.3096, + "step": 32400 + }, + { + "epoch": 5.89, + "learning_rate": 1.2334728795049145e-06, + "loss": 0.216, + "step": 32401 + }, + { + "epoch": 5.89, + "learning_rate": 1.2334182744812523e-06, + "loss": 1.2558, + "step": 32402 + }, + { + "epoch": 5.89, + "learning_rate": 1.23336366945759e-06, + "loss": 0.0114, + "step": 32403 + }, + { + "epoch": 5.89, + "learning_rate": 1.233309064433928e-06, + "loss": 0.302, + "step": 32404 + }, + { + "epoch": 5.89, + "learning_rate": 1.233254459410266e-06, + "loss": 0.2274, + "step": 32405 + }, + { + "epoch": 5.89, + "learning_rate": 1.2331998543866036e-06, + "loss": 0.4662, + "step": 32406 + }, + { + "epoch": 5.89, + "learning_rate": 1.2331452493629414e-06, + "loss": 0.7688, + "step": 32407 + }, + { + "epoch": 5.89, + "learning_rate": 1.2330906443392793e-06, + "loss": 0.5262, + "step": 32408 + }, + { + "epoch": 5.89, + "learning_rate": 1.233036039315617e-06, + "loss": 0.075, + "step": 32409 + }, + { + "epoch": 5.89, + "learning_rate": 1.2329814342919548e-06, + "loss": 1.6332, + "step": 32410 + }, + { + "epoch": 5.89, + "learning_rate": 1.2329268292682927e-06, + "loss": 0.4735, + "step": 32411 + }, + { + "epoch": 5.89, + "learning_rate": 1.2328722242446307e-06, + "loss": 0.0063, + "step": 32412 + }, + { + "epoch": 5.89, + "learning_rate": 1.2328176192209684e-06, + "loss": 1.2907, + "step": 32413 + }, + { + "epoch": 5.89, + "learning_rate": 1.2327630141973061e-06, + "loss": 1.1546, + "step": 32414 + }, + { + "epoch": 5.89, + "learning_rate": 1.232708409173644e-06, + "loss": 2.0158, + "step": 32415 + }, + { + "epoch": 5.89, + "learning_rate": 1.2326538041499818e-06, + "loss": 0.1824, + "step": 32416 + }, + { + "epoch": 5.9, + "learning_rate": 1.2325991991263197e-06, + "loss": 0.4403, + "step": 32417 + }, + { + "epoch": 5.9, + "learning_rate": 1.2325445941026575e-06, + "loss": 0.1577, + "step": 32418 + }, + { + "epoch": 5.9, + "learning_rate": 1.2324899890789954e-06, + "loss": 0.0725, + "step": 32419 + }, + { + "epoch": 5.9, + "learning_rate": 1.2324353840553331e-06, + "loss": 1.476, + "step": 32420 + }, + { + "epoch": 5.9, + "learning_rate": 1.2323807790316709e-06, + "loss": 0.0214, + "step": 32421 + }, + { + "epoch": 5.9, + "learning_rate": 1.2323261740080088e-06, + "loss": 1.7308, + "step": 32422 + }, + { + "epoch": 5.9, + "learning_rate": 1.2322715689843465e-06, + "loss": 1.4931, + "step": 32423 + }, + { + "epoch": 5.9, + "learning_rate": 1.2322169639606845e-06, + "loss": 0.0096, + "step": 32424 + }, + { + "epoch": 5.9, + "learning_rate": 1.2321623589370222e-06, + "loss": 1.3508, + "step": 32425 + }, + { + "epoch": 5.9, + "learning_rate": 1.2321077539133602e-06, + "loss": 1.4328, + "step": 32426 + }, + { + "epoch": 5.9, + "learning_rate": 1.232053148889698e-06, + "loss": 1.0084, + "step": 32427 + }, + { + "epoch": 5.9, + "learning_rate": 1.2319985438660356e-06, + "loss": 0.1173, + "step": 32428 + }, + { + "epoch": 5.9, + "learning_rate": 1.2319439388423736e-06, + "loss": 0.2294, + "step": 32429 + }, + { + "epoch": 5.9, + "learning_rate": 1.2318893338187113e-06, + "loss": 0.5285, + "step": 32430 + }, + { + "epoch": 5.9, + "learning_rate": 1.2318347287950492e-06, + "loss": 0.0597, + "step": 32431 + }, + { + "epoch": 5.9, + "learning_rate": 1.231780123771387e-06, + "loss": 0.0471, + "step": 32432 + }, + { + "epoch": 5.9, + "learning_rate": 1.231725518747725e-06, + "loss": 4.2094, + "step": 32433 + }, + { + "epoch": 5.9, + "learning_rate": 1.2316709137240627e-06, + "loss": 1.6608, + "step": 32434 + }, + { + "epoch": 5.9, + "learning_rate": 1.2316163087004004e-06, + "loss": 0.0383, + "step": 32435 + }, + { + "epoch": 5.9, + "learning_rate": 1.2315617036767383e-06, + "loss": 0.1797, + "step": 32436 + }, + { + "epoch": 5.9, + "learning_rate": 1.231507098653076e-06, + "loss": 2.5014, + "step": 32437 + }, + { + "epoch": 5.9, + "learning_rate": 1.231452493629414e-06, + "loss": 0.9756, + "step": 32438 + }, + { + "epoch": 5.9, + "learning_rate": 1.2313978886057517e-06, + "loss": 0.0904, + "step": 32439 + }, + { + "epoch": 5.9, + "learning_rate": 1.2313432835820897e-06, + "loss": 0.0073, + "step": 32440 + }, + { + "epoch": 5.9, + "learning_rate": 1.2312886785584274e-06, + "loss": 1.0405, + "step": 32441 + }, + { + "epoch": 5.9, + "learning_rate": 1.2312340735347651e-06, + "loss": 0.0122, + "step": 32442 + }, + { + "epoch": 5.9, + "learning_rate": 1.231179468511103e-06, + "loss": 0.4331, + "step": 32443 + }, + { + "epoch": 5.9, + "learning_rate": 1.2311248634874408e-06, + "loss": 0.2309, + "step": 32444 + }, + { + "epoch": 5.9, + "learning_rate": 1.2310702584637788e-06, + "loss": 0.5955, + "step": 32445 + }, + { + "epoch": 5.9, + "learning_rate": 1.2310156534401165e-06, + "loss": 0.176, + "step": 32446 + }, + { + "epoch": 5.9, + "learning_rate": 1.2309610484164544e-06, + "loss": 1.207, + "step": 32447 + }, + { + "epoch": 5.9, + "learning_rate": 1.2309064433927922e-06, + "loss": 0.0223, + "step": 32448 + }, + { + "epoch": 5.9, + "learning_rate": 1.23085183836913e-06, + "loss": 0.5257, + "step": 32449 + }, + { + "epoch": 5.9, + "learning_rate": 1.2307972333454676e-06, + "loss": 0.0079, + "step": 32450 + }, + { + "epoch": 5.9, + "learning_rate": 1.2307426283218056e-06, + "loss": 0.0179, + "step": 32451 + }, + { + "epoch": 5.9, + "learning_rate": 1.2306880232981435e-06, + "loss": 0.2431, + "step": 32452 + }, + { + "epoch": 5.9, + "learning_rate": 1.2306334182744813e-06, + "loss": 0.9284, + "step": 32453 + }, + { + "epoch": 5.9, + "learning_rate": 1.2305788132508192e-06, + "loss": 2.8103, + "step": 32454 + }, + { + "epoch": 5.9, + "learning_rate": 1.230524208227157e-06, + "loss": 0.1873, + "step": 32455 + }, + { + "epoch": 5.9, + "learning_rate": 1.2304696032034947e-06, + "loss": 0.4131, + "step": 32456 + }, + { + "epoch": 5.9, + "learning_rate": 1.2304149981798324e-06, + "loss": 0.1779, + "step": 32457 + }, + { + "epoch": 5.9, + "learning_rate": 1.2303603931561703e-06, + "loss": 1.0037, + "step": 32458 + }, + { + "epoch": 5.9, + "learning_rate": 1.2303057881325083e-06, + "loss": 1.3061, + "step": 32459 + }, + { + "epoch": 5.9, + "learning_rate": 1.230251183108846e-06, + "loss": 0.1265, + "step": 32460 + }, + { + "epoch": 5.9, + "learning_rate": 1.230196578085184e-06, + "loss": 0.4855, + "step": 32461 + }, + { + "epoch": 5.9, + "learning_rate": 1.2301419730615217e-06, + "loss": 1.336, + "step": 32462 + }, + { + "epoch": 5.9, + "learning_rate": 1.2300873680378594e-06, + "loss": 0.2556, + "step": 32463 + }, + { + "epoch": 5.9, + "learning_rate": 1.2300327630141974e-06, + "loss": 0.4263, + "step": 32464 + }, + { + "epoch": 5.9, + "learning_rate": 1.2299781579905353e-06, + "loss": 0.048, + "step": 32465 + }, + { + "epoch": 5.9, + "learning_rate": 1.229923552966873e-06, + "loss": 0.1382, + "step": 32466 + }, + { + "epoch": 5.9, + "learning_rate": 1.2298689479432108e-06, + "loss": 0.0041, + "step": 32467 + }, + { + "epoch": 5.9, + "learning_rate": 1.2298143429195487e-06, + "loss": 2.1722, + "step": 32468 + }, + { + "epoch": 5.9, + "learning_rate": 1.2297597378958864e-06, + "loss": 0.1343, + "step": 32469 + }, + { + "epoch": 5.9, + "learning_rate": 1.2297051328722242e-06, + "loss": 2.5823, + "step": 32470 + }, + { + "epoch": 5.9, + "learning_rate": 1.2296505278485621e-06, + "loss": 0.7069, + "step": 32471 + }, + { + "epoch": 5.91, + "learning_rate": 1.2295959228249e-06, + "loss": 2.6183, + "step": 32472 + }, + { + "epoch": 5.91, + "learning_rate": 1.2295413178012378e-06, + "loss": 0.2687, + "step": 32473 + }, + { + "epoch": 5.91, + "learning_rate": 1.2294867127775755e-06, + "loss": 0.2836, + "step": 32474 + }, + { + "epoch": 5.91, + "learning_rate": 1.2294321077539135e-06, + "loss": 0.2741, + "step": 32475 + }, + { + "epoch": 5.91, + "learning_rate": 1.2293775027302512e-06, + "loss": 0.2336, + "step": 32476 + }, + { + "epoch": 5.91, + "learning_rate": 1.229322897706589e-06, + "loss": 1.4164, + "step": 32477 + }, + { + "epoch": 5.91, + "learning_rate": 1.2292682926829269e-06, + "loss": 0.3601, + "step": 32478 + }, + { + "epoch": 5.91, + "learning_rate": 1.2292136876592648e-06, + "loss": 0.11, + "step": 32479 + }, + { + "epoch": 5.91, + "learning_rate": 1.2291590826356025e-06, + "loss": 0.0086, + "step": 32480 + }, + { + "epoch": 5.91, + "learning_rate": 1.2291044776119403e-06, + "loss": 0.6031, + "step": 32481 + }, + { + "epoch": 5.91, + "learning_rate": 1.2290498725882782e-06, + "loss": 0.7443, + "step": 32482 + }, + { + "epoch": 5.91, + "learning_rate": 1.228995267564616e-06, + "loss": 1.5467, + "step": 32483 + }, + { + "epoch": 5.91, + "learning_rate": 1.2289406625409537e-06, + "loss": 1.4509, + "step": 32484 + }, + { + "epoch": 5.91, + "learning_rate": 1.2288860575172916e-06, + "loss": 1.872, + "step": 32485 + }, + { + "epoch": 5.91, + "learning_rate": 1.2288314524936296e-06, + "loss": 0.1695, + "step": 32486 + }, + { + "epoch": 5.91, + "learning_rate": 1.2287768474699673e-06, + "loss": 0.0086, + "step": 32487 + }, + { + "epoch": 5.91, + "learning_rate": 1.228722242446305e-06, + "loss": 1.6978, + "step": 32488 + }, + { + "epoch": 5.91, + "learning_rate": 1.228667637422643e-06, + "loss": 1.3867, + "step": 32489 + }, + { + "epoch": 5.91, + "learning_rate": 1.2286130323989807e-06, + "loss": 0.1767, + "step": 32490 + }, + { + "epoch": 5.91, + "learning_rate": 1.2285584273753184e-06, + "loss": 0.5425, + "step": 32491 + }, + { + "epoch": 5.91, + "learning_rate": 1.2285038223516564e-06, + "loss": 0.712, + "step": 32492 + }, + { + "epoch": 5.91, + "learning_rate": 1.2284492173279943e-06, + "loss": 0.1863, + "step": 32493 + }, + { + "epoch": 5.91, + "learning_rate": 1.228394612304332e-06, + "loss": 0.0201, + "step": 32494 + }, + { + "epoch": 5.91, + "learning_rate": 1.2283400072806698e-06, + "loss": 0.5521, + "step": 32495 + }, + { + "epoch": 5.91, + "learning_rate": 1.2282854022570077e-06, + "loss": 0.3017, + "step": 32496 + }, + { + "epoch": 5.91, + "learning_rate": 1.2282307972333455e-06, + "loss": 0.749, + "step": 32497 + }, + { + "epoch": 5.91, + "learning_rate": 1.2281761922096832e-06, + "loss": 1.0947, + "step": 32498 + }, + { + "epoch": 5.91, + "learning_rate": 1.2281215871860211e-06, + "loss": 0.5732, + "step": 32499 + }, + { + "epoch": 5.91, + "learning_rate": 1.228066982162359e-06, + "loss": 0.0086, + "step": 32500 + }, + { + "epoch": 5.91, + "learning_rate": 1.2280123771386968e-06, + "loss": 0.1878, + "step": 32501 + }, + { + "epoch": 5.91, + "learning_rate": 1.2279577721150345e-06, + "loss": 0.6565, + "step": 32502 + }, + { + "epoch": 5.91, + "learning_rate": 1.2279031670913725e-06, + "loss": 0.6186, + "step": 32503 + }, + { + "epoch": 5.91, + "learning_rate": 1.2278485620677102e-06, + "loss": 0.4759, + "step": 32504 + }, + { + "epoch": 5.91, + "learning_rate": 1.2277939570440482e-06, + "loss": 1.4088, + "step": 32505 + }, + { + "epoch": 5.91, + "learning_rate": 1.227739352020386e-06, + "loss": 0.0986, + "step": 32506 + }, + { + "epoch": 5.91, + "learning_rate": 1.2276847469967238e-06, + "loss": 0.922, + "step": 32507 + }, + { + "epoch": 5.91, + "learning_rate": 1.2276301419730616e-06, + "loss": 0.2437, + "step": 32508 + }, + { + "epoch": 5.91, + "learning_rate": 1.2275755369493993e-06, + "loss": 1.7504, + "step": 32509 + }, + { + "epoch": 5.91, + "learning_rate": 1.2275209319257372e-06, + "loss": 0.0114, + "step": 32510 + }, + { + "epoch": 5.91, + "learning_rate": 1.227466326902075e-06, + "loss": 0.0079, + "step": 32511 + }, + { + "epoch": 5.91, + "learning_rate": 1.227411721878413e-06, + "loss": 0.4484, + "step": 32512 + }, + { + "epoch": 5.91, + "learning_rate": 1.2273571168547507e-06, + "loss": 0.0811, + "step": 32513 + }, + { + "epoch": 5.91, + "learning_rate": 1.2273025118310886e-06, + "loss": 0.1306, + "step": 32514 + }, + { + "epoch": 5.91, + "learning_rate": 1.2272479068074263e-06, + "loss": 0.4067, + "step": 32515 + }, + { + "epoch": 5.91, + "learning_rate": 1.227193301783764e-06, + "loss": 0.8596, + "step": 32516 + }, + { + "epoch": 5.91, + "learning_rate": 1.227138696760102e-06, + "loss": 2.3105, + "step": 32517 + }, + { + "epoch": 5.91, + "learning_rate": 1.2270840917364397e-06, + "loss": 0.011, + "step": 32518 + }, + { + "epoch": 5.91, + "learning_rate": 1.2270294867127777e-06, + "loss": 0.0138, + "step": 32519 + }, + { + "epoch": 5.91, + "learning_rate": 1.2269748816891154e-06, + "loss": 0.2248, + "step": 32520 + }, + { + "epoch": 5.91, + "learning_rate": 1.2269202766654534e-06, + "loss": 4.6076, + "step": 32521 + }, + { + "epoch": 5.91, + "learning_rate": 1.226865671641791e-06, + "loss": 0.0092, + "step": 32522 + }, + { + "epoch": 5.91, + "learning_rate": 1.2268110666181288e-06, + "loss": 0.0723, + "step": 32523 + }, + { + "epoch": 5.91, + "learning_rate": 1.2267564615944668e-06, + "loss": 0.4275, + "step": 32524 + }, + { + "epoch": 5.91, + "learning_rate": 1.2267018565708045e-06, + "loss": 0.4806, + "step": 32525 + }, + { + "epoch": 5.91, + "learning_rate": 1.2266472515471424e-06, + "loss": 2.4954, + "step": 32526 + }, + { + "epoch": 5.92, + "learning_rate": 1.2265926465234802e-06, + "loss": 0.0739, + "step": 32527 + }, + { + "epoch": 5.92, + "learning_rate": 1.2265380414998181e-06, + "loss": 0.0921, + "step": 32528 + }, + { + "epoch": 5.92, + "learning_rate": 1.2264834364761558e-06, + "loss": 0.2292, + "step": 32529 + }, + { + "epoch": 5.92, + "learning_rate": 1.2264288314524936e-06, + "loss": 0.8593, + "step": 32530 + }, + { + "epoch": 5.92, + "learning_rate": 1.2263742264288313e-06, + "loss": 0.1935, + "step": 32531 + }, + { + "epoch": 5.92, + "learning_rate": 1.2263196214051693e-06, + "loss": 0.923, + "step": 32532 + }, + { + "epoch": 5.92, + "learning_rate": 1.2262650163815072e-06, + "loss": 1.4202, + "step": 32533 + }, + { + "epoch": 5.92, + "learning_rate": 1.226210411357845e-06, + "loss": 0.5943, + "step": 32534 + }, + { + "epoch": 5.92, + "learning_rate": 1.2261558063341829e-06, + "loss": 1.7905, + "step": 32535 + }, + { + "epoch": 5.92, + "learning_rate": 1.2261012013105206e-06, + "loss": 0.0165, + "step": 32536 + }, + { + "epoch": 5.92, + "learning_rate": 1.2260465962868583e-06, + "loss": 0.9278, + "step": 32537 + }, + { + "epoch": 5.92, + "learning_rate": 1.225991991263196e-06, + "loss": 0.5537, + "step": 32538 + }, + { + "epoch": 5.92, + "learning_rate": 1.225937386239534e-06, + "loss": 0.0614, + "step": 32539 + }, + { + "epoch": 5.92, + "learning_rate": 1.225882781215872e-06, + "loss": 1.0419, + "step": 32540 + }, + { + "epoch": 5.92, + "learning_rate": 1.2258281761922097e-06, + "loss": 0.0228, + "step": 32541 + }, + { + "epoch": 5.92, + "learning_rate": 1.2257735711685476e-06, + "loss": 0.7003, + "step": 32542 + }, + { + "epoch": 5.92, + "learning_rate": 1.2257189661448854e-06, + "loss": 1.1092, + "step": 32543 + }, + { + "epoch": 5.92, + "learning_rate": 1.225664361121223e-06, + "loss": 1.1874, + "step": 32544 + }, + { + "epoch": 5.92, + "learning_rate": 1.225609756097561e-06, + "loss": 3.0093, + "step": 32545 + }, + { + "epoch": 5.92, + "learning_rate": 1.225555151073899e-06, + "loss": 0.3258, + "step": 32546 + }, + { + "epoch": 5.92, + "learning_rate": 1.2255005460502367e-06, + "loss": 1.2675, + "step": 32547 + }, + { + "epoch": 5.92, + "learning_rate": 1.2254459410265744e-06, + "loss": 0.1004, + "step": 32548 + }, + { + "epoch": 5.92, + "learning_rate": 1.2253913360029124e-06, + "loss": 0.1528, + "step": 32549 + }, + { + "epoch": 5.92, + "learning_rate": 1.2253367309792501e-06, + "loss": 0.1366, + "step": 32550 + }, + { + "epoch": 5.92, + "learning_rate": 1.2252821259555878e-06, + "loss": 0.4703, + "step": 32551 + }, + { + "epoch": 5.92, + "learning_rate": 1.2252275209319258e-06, + "loss": 0.3042, + "step": 32552 + }, + { + "epoch": 5.92, + "learning_rate": 1.2251729159082637e-06, + "loss": 0.1811, + "step": 32553 + }, + { + "epoch": 5.92, + "learning_rate": 1.2251183108846015e-06, + "loss": 1.8556, + "step": 32554 + }, + { + "epoch": 5.92, + "learning_rate": 1.2250637058609392e-06, + "loss": 0.7624, + "step": 32555 + }, + { + "epoch": 5.92, + "learning_rate": 1.2250091008372771e-06, + "loss": 0.3093, + "step": 32556 + }, + { + "epoch": 5.92, + "learning_rate": 1.2249544958136149e-06, + "loss": 2.2951, + "step": 32557 + }, + { + "epoch": 5.92, + "learning_rate": 1.2248998907899526e-06, + "loss": 0.3203, + "step": 32558 + }, + { + "epoch": 5.92, + "learning_rate": 1.2248452857662905e-06, + "loss": 0.1979, + "step": 32559 + }, + { + "epoch": 5.92, + "learning_rate": 1.2247906807426285e-06, + "loss": 1.1908, + "step": 32560 + }, + { + "epoch": 5.92, + "learning_rate": 1.2247360757189662e-06, + "loss": 0.3586, + "step": 32561 + }, + { + "epoch": 5.92, + "learning_rate": 1.224681470695304e-06, + "loss": 0.7368, + "step": 32562 + }, + { + "epoch": 5.92, + "learning_rate": 1.224626865671642e-06, + "loss": 0.0294, + "step": 32563 + }, + { + "epoch": 5.92, + "learning_rate": 1.2245722606479796e-06, + "loss": 1.0891, + "step": 32564 + }, + { + "epoch": 5.92, + "learning_rate": 1.2245176556243174e-06, + "loss": 0.0089, + "step": 32565 + }, + { + "epoch": 5.92, + "learning_rate": 1.2244630506006553e-06, + "loss": 2.0455, + "step": 32566 + }, + { + "epoch": 5.92, + "learning_rate": 1.2244084455769932e-06, + "loss": 3.1383, + "step": 32567 + }, + { + "epoch": 5.92, + "learning_rate": 1.224353840553331e-06, + "loss": 0.948, + "step": 32568 + }, + { + "epoch": 5.92, + "learning_rate": 1.2242992355296687e-06, + "loss": 0.0286, + "step": 32569 + }, + { + "epoch": 5.92, + "learning_rate": 1.2242446305060067e-06, + "loss": 0.2125, + "step": 32570 + }, + { + "epoch": 5.92, + "learning_rate": 1.2241900254823444e-06, + "loss": 0.2651, + "step": 32571 + }, + { + "epoch": 5.92, + "learning_rate": 1.2241354204586821e-06, + "loss": 0.4231, + "step": 32572 + }, + { + "epoch": 5.92, + "learning_rate": 1.22408081543502e-06, + "loss": 0.9451, + "step": 32573 + }, + { + "epoch": 5.92, + "learning_rate": 1.224026210411358e-06, + "loss": 0.0275, + "step": 32574 + }, + { + "epoch": 5.92, + "learning_rate": 1.2239716053876957e-06, + "loss": 0.2053, + "step": 32575 + }, + { + "epoch": 5.92, + "learning_rate": 1.2239170003640335e-06, + "loss": 1.1431, + "step": 32576 + }, + { + "epoch": 5.92, + "learning_rate": 1.2238623953403714e-06, + "loss": 0.1027, + "step": 32577 + }, + { + "epoch": 5.92, + "learning_rate": 1.2238077903167091e-06, + "loss": 0.3559, + "step": 32578 + }, + { + "epoch": 5.92, + "learning_rate": 1.2237531852930469e-06, + "loss": 0.4982, + "step": 32579 + }, + { + "epoch": 5.92, + "learning_rate": 1.2236985802693848e-06, + "loss": 0.5038, + "step": 32580 + }, + { + "epoch": 5.92, + "learning_rate": 1.2236439752457228e-06, + "loss": 0.0595, + "step": 32581 + }, + { + "epoch": 5.93, + "learning_rate": 1.2235893702220605e-06, + "loss": 0.1061, + "step": 32582 + }, + { + "epoch": 5.93, + "learning_rate": 1.2235347651983982e-06, + "loss": 0.7873, + "step": 32583 + }, + { + "epoch": 5.93, + "learning_rate": 1.2234801601747362e-06, + "loss": 0.0064, + "step": 32584 + }, + { + "epoch": 5.93, + "learning_rate": 1.223425555151074e-06, + "loss": 0.4884, + "step": 32585 + }, + { + "epoch": 5.93, + "learning_rate": 1.2233709501274118e-06, + "loss": 0.2528, + "step": 32586 + }, + { + "epoch": 5.93, + "learning_rate": 1.2233163451037496e-06, + "loss": 0.9555, + "step": 32587 + }, + { + "epoch": 5.93, + "learning_rate": 1.2232617400800875e-06, + "loss": 0.6237, + "step": 32588 + }, + { + "epoch": 5.93, + "learning_rate": 1.2232071350564252e-06, + "loss": 0.9065, + "step": 32589 + }, + { + "epoch": 5.93, + "learning_rate": 1.223152530032763e-06, + "loss": 0.0317, + "step": 32590 + }, + { + "epoch": 5.93, + "learning_rate": 1.223097925009101e-06, + "loss": 0.0109, + "step": 32591 + }, + { + "epoch": 5.93, + "learning_rate": 1.2230433199854387e-06, + "loss": 0.0956, + "step": 32592 + }, + { + "epoch": 5.93, + "learning_rate": 1.2229887149617766e-06, + "loss": 0.3318, + "step": 32593 + }, + { + "epoch": 5.93, + "learning_rate": 1.2229341099381143e-06, + "loss": 0.4709, + "step": 32594 + }, + { + "epoch": 5.93, + "learning_rate": 1.2228795049144523e-06, + "loss": 0.8916, + "step": 32595 + }, + { + "epoch": 5.93, + "learning_rate": 1.22282489989079e-06, + "loss": 1.9097, + "step": 32596 + }, + { + "epoch": 5.93, + "learning_rate": 1.2227702948671277e-06, + "loss": 0.02, + "step": 32597 + }, + { + "epoch": 5.93, + "learning_rate": 1.2227156898434657e-06, + "loss": 0.6192, + "step": 32598 + }, + { + "epoch": 5.93, + "learning_rate": 1.2226610848198034e-06, + "loss": 3.3032, + "step": 32599 + }, + { + "epoch": 5.93, + "learning_rate": 1.2226064797961414e-06, + "loss": 1.8901, + "step": 32600 + }, + { + "epoch": 5.93, + "learning_rate": 1.222551874772479e-06, + "loss": 0.0577, + "step": 32601 + }, + { + "epoch": 5.93, + "learning_rate": 1.222497269748817e-06, + "loss": 1.1737, + "step": 32602 + }, + { + "epoch": 5.93, + "learning_rate": 1.2224426647251548e-06, + "loss": 1.5023, + "step": 32603 + }, + { + "epoch": 5.93, + "learning_rate": 1.2223880597014925e-06, + "loss": 0.0909, + "step": 32604 + }, + { + "epoch": 5.93, + "learning_rate": 1.2223334546778302e-06, + "loss": 0.0158, + "step": 32605 + }, + { + "epoch": 5.93, + "learning_rate": 1.2222788496541682e-06, + "loss": 0.3712, + "step": 32606 + }, + { + "epoch": 5.93, + "learning_rate": 1.2222242446305061e-06, + "loss": 0.08, + "step": 32607 + }, + { + "epoch": 5.93, + "learning_rate": 1.2221696396068438e-06, + "loss": 1.4474, + "step": 32608 + }, + { + "epoch": 5.93, + "learning_rate": 1.2221150345831818e-06, + "loss": 0.0069, + "step": 32609 + }, + { + "epoch": 5.93, + "learning_rate": 1.2220604295595195e-06, + "loss": 0.1993, + "step": 32610 + }, + { + "epoch": 5.93, + "learning_rate": 1.2220058245358572e-06, + "loss": 1.4736, + "step": 32611 + }, + { + "epoch": 5.93, + "learning_rate": 1.221951219512195e-06, + "loss": 0.9578, + "step": 32612 + }, + { + "epoch": 5.93, + "learning_rate": 1.221896614488533e-06, + "loss": 0.1764, + "step": 32613 + }, + { + "epoch": 5.93, + "learning_rate": 1.2218420094648709e-06, + "loss": 0.1671, + "step": 32614 + }, + { + "epoch": 5.93, + "learning_rate": 1.2217874044412086e-06, + "loss": 0.0072, + "step": 32615 + }, + { + "epoch": 5.93, + "learning_rate": 1.2217327994175465e-06, + "loss": 0.1271, + "step": 32616 + }, + { + "epoch": 5.93, + "learning_rate": 1.2216781943938843e-06, + "loss": 2.592, + "step": 32617 + }, + { + "epoch": 5.93, + "learning_rate": 1.221623589370222e-06, + "loss": 1.6795, + "step": 32618 + }, + { + "epoch": 5.93, + "learning_rate": 1.2215689843465597e-06, + "loss": 0.2192, + "step": 32619 + }, + { + "epoch": 5.93, + "learning_rate": 1.2215143793228977e-06, + "loss": 0.8002, + "step": 32620 + }, + { + "epoch": 5.93, + "learning_rate": 1.2214597742992356e-06, + "loss": 1.6513, + "step": 32621 + }, + { + "epoch": 5.93, + "learning_rate": 1.2214051692755734e-06, + "loss": 0.1791, + "step": 32622 + }, + { + "epoch": 5.93, + "learning_rate": 1.2213505642519113e-06, + "loss": 1.7382, + "step": 32623 + }, + { + "epoch": 5.93, + "learning_rate": 1.221295959228249e-06, + "loss": 0.4454, + "step": 32624 + }, + { + "epoch": 5.93, + "learning_rate": 1.2212413542045868e-06, + "loss": 0.9102, + "step": 32625 + }, + { + "epoch": 5.93, + "learning_rate": 1.2211867491809245e-06, + "loss": 0.0373, + "step": 32626 + }, + { + "epoch": 5.93, + "learning_rate": 1.2211321441572624e-06, + "loss": 0.0818, + "step": 32627 + }, + { + "epoch": 5.93, + "learning_rate": 1.2210775391336004e-06, + "loss": 0.6399, + "step": 32628 + }, + { + "epoch": 5.93, + "learning_rate": 1.2210229341099381e-06, + "loss": 0.466, + "step": 32629 + }, + { + "epoch": 5.93, + "learning_rate": 1.220968329086276e-06, + "loss": 0.0311, + "step": 32630 + }, + { + "epoch": 5.93, + "learning_rate": 1.2209137240626138e-06, + "loss": 0.074, + "step": 32631 + }, + { + "epoch": 5.93, + "learning_rate": 1.2208591190389515e-06, + "loss": 0.3743, + "step": 32632 + }, + { + "epoch": 5.93, + "learning_rate": 1.2208045140152895e-06, + "loss": 0.0201, + "step": 32633 + }, + { + "epoch": 5.93, + "learning_rate": 1.2207499089916274e-06, + "loss": 0.1302, + "step": 32634 + }, + { + "epoch": 5.93, + "learning_rate": 1.2206953039679651e-06, + "loss": 0.6663, + "step": 32635 + }, + { + "epoch": 5.93, + "learning_rate": 1.2206406989443029e-06, + "loss": 0.0713, + "step": 32636 + }, + { + "epoch": 5.94, + "learning_rate": 1.2205860939206408e-06, + "loss": 0.0485, + "step": 32637 + }, + { + "epoch": 5.94, + "learning_rate": 1.2205314888969785e-06, + "loss": 0.0215, + "step": 32638 + }, + { + "epoch": 5.94, + "learning_rate": 1.2204768838733163e-06, + "loss": 0.0844, + "step": 32639 + }, + { + "epoch": 5.94, + "learning_rate": 1.2204222788496542e-06, + "loss": 0.0258, + "step": 32640 + }, + { + "epoch": 5.94, + "learning_rate": 1.2203676738259922e-06, + "loss": 0.1885, + "step": 32641 + }, + { + "epoch": 5.94, + "learning_rate": 1.22031306880233e-06, + "loss": 1.2214, + "step": 32642 + }, + { + "epoch": 5.94, + "learning_rate": 1.2202584637786676e-06, + "loss": 0.2563, + "step": 32643 + }, + { + "epoch": 5.94, + "learning_rate": 1.2202038587550056e-06, + "loss": 3.3051, + "step": 32644 + }, + { + "epoch": 5.94, + "learning_rate": 1.2201492537313433e-06, + "loss": 0.2273, + "step": 32645 + }, + { + "epoch": 5.94, + "learning_rate": 1.220094648707681e-06, + "loss": 0.0506, + "step": 32646 + }, + { + "epoch": 5.94, + "learning_rate": 1.220040043684019e-06, + "loss": 0.0048, + "step": 32647 + }, + { + "epoch": 5.94, + "learning_rate": 1.219985438660357e-06, + "loss": 1.6626, + "step": 32648 + }, + { + "epoch": 5.94, + "learning_rate": 1.2199308336366947e-06, + "loss": 0.0578, + "step": 32649 + }, + { + "epoch": 5.94, + "learning_rate": 1.2198762286130324e-06, + "loss": 0.4348, + "step": 32650 + }, + { + "epoch": 5.94, + "learning_rate": 1.2198216235893703e-06, + "loss": 0.6709, + "step": 32651 + }, + { + "epoch": 5.94, + "learning_rate": 1.219767018565708e-06, + "loss": 0.3111, + "step": 32652 + }, + { + "epoch": 5.94, + "learning_rate": 1.2197124135420458e-06, + "loss": 0.1877, + "step": 32653 + }, + { + "epoch": 5.94, + "learning_rate": 1.2196578085183837e-06, + "loss": 0.133, + "step": 32654 + }, + { + "epoch": 5.94, + "learning_rate": 1.2196032034947217e-06, + "loss": 0.9889, + "step": 32655 + }, + { + "epoch": 5.94, + "learning_rate": 1.2195485984710594e-06, + "loss": 0.012, + "step": 32656 + }, + { + "epoch": 5.94, + "learning_rate": 1.2194939934473971e-06, + "loss": 0.0965, + "step": 32657 + }, + { + "epoch": 5.94, + "learning_rate": 1.219439388423735e-06, + "loss": 0.3619, + "step": 32658 + }, + { + "epoch": 5.94, + "learning_rate": 1.2193847834000728e-06, + "loss": 0.1533, + "step": 32659 + }, + { + "epoch": 5.94, + "learning_rate": 1.2193301783764105e-06, + "loss": 0.0062, + "step": 32660 + }, + { + "epoch": 5.94, + "learning_rate": 1.2192755733527485e-06, + "loss": 0.3939, + "step": 32661 + }, + { + "epoch": 5.94, + "learning_rate": 1.2192209683290864e-06, + "loss": 0.0117, + "step": 32662 + }, + { + "epoch": 5.94, + "learning_rate": 1.2191663633054242e-06, + "loss": 0.0096, + "step": 32663 + }, + { + "epoch": 5.94, + "learning_rate": 1.219111758281762e-06, + "loss": 0.6758, + "step": 32664 + }, + { + "epoch": 5.94, + "learning_rate": 1.2190571532580998e-06, + "loss": 0.0122, + "step": 32665 + }, + { + "epoch": 5.94, + "learning_rate": 1.2190025482344376e-06, + "loss": 0.1724, + "step": 32666 + }, + { + "epoch": 5.94, + "learning_rate": 1.2189479432107753e-06, + "loss": 0.0082, + "step": 32667 + }, + { + "epoch": 5.94, + "learning_rate": 1.2188933381871132e-06, + "loss": 0.3328, + "step": 32668 + }, + { + "epoch": 5.94, + "learning_rate": 1.2188387331634512e-06, + "loss": 1.3468, + "step": 32669 + }, + { + "epoch": 5.94, + "learning_rate": 1.218784128139789e-06, + "loss": 0.4132, + "step": 32670 + }, + { + "epoch": 5.94, + "learning_rate": 1.2187295231161267e-06, + "loss": 1.3696, + "step": 32671 + }, + { + "epoch": 5.94, + "learning_rate": 1.2186749180924646e-06, + "loss": 0.6299, + "step": 32672 + }, + { + "epoch": 5.94, + "learning_rate": 1.2186203130688023e-06, + "loss": 4.9934, + "step": 32673 + }, + { + "epoch": 5.94, + "learning_rate": 1.2185657080451403e-06, + "loss": 0.0967, + "step": 32674 + }, + { + "epoch": 5.94, + "learning_rate": 1.218511103021478e-06, + "loss": 2.9035, + "step": 32675 + }, + { + "epoch": 5.94, + "learning_rate": 1.218456497997816e-06, + "loss": 0.3457, + "step": 32676 + }, + { + "epoch": 5.94, + "learning_rate": 1.2184018929741537e-06, + "loss": 0.01, + "step": 32677 + }, + { + "epoch": 5.94, + "learning_rate": 1.2183472879504914e-06, + "loss": 0.0847, + "step": 32678 + }, + { + "epoch": 5.94, + "learning_rate": 1.2182926829268294e-06, + "loss": 1.3175, + "step": 32679 + }, + { + "epoch": 5.94, + "learning_rate": 1.218238077903167e-06, + "loss": 0.6695, + "step": 32680 + }, + { + "epoch": 5.94, + "learning_rate": 1.218183472879505e-06, + "loss": 0.9886, + "step": 32681 + }, + { + "epoch": 5.94, + "learning_rate": 1.2181288678558428e-06, + "loss": 0.4523, + "step": 32682 + }, + { + "epoch": 5.94, + "learning_rate": 1.2180742628321807e-06, + "loss": 2.2769, + "step": 32683 + }, + { + "epoch": 5.94, + "learning_rate": 1.2180196578085184e-06, + "loss": 0.0915, + "step": 32684 + }, + { + "epoch": 5.94, + "learning_rate": 1.2179650527848562e-06, + "loss": 0.143, + "step": 32685 + }, + { + "epoch": 5.94, + "learning_rate": 1.217910447761194e-06, + "loss": 0.0178, + "step": 32686 + }, + { + "epoch": 5.94, + "learning_rate": 1.2178558427375318e-06, + "loss": 0.0103, + "step": 32687 + }, + { + "epoch": 5.94, + "learning_rate": 1.2178012377138698e-06, + "loss": 0.1068, + "step": 32688 + }, + { + "epoch": 5.94, + "learning_rate": 1.2177466326902075e-06, + "loss": 1.0108, + "step": 32689 + }, + { + "epoch": 5.94, + "learning_rate": 1.2176920276665455e-06, + "loss": 0.3732, + "step": 32690 + }, + { + "epoch": 5.94, + "learning_rate": 1.2176374226428832e-06, + "loss": 0.0944, + "step": 32691 + }, + { + "epoch": 5.95, + "learning_rate": 1.217582817619221e-06, + "loss": 1.3642, + "step": 32692 + }, + { + "epoch": 5.95, + "learning_rate": 1.2175282125955587e-06, + "loss": 1.0249, + "step": 32693 + }, + { + "epoch": 5.95, + "learning_rate": 1.2174736075718966e-06, + "loss": 0.6694, + "step": 32694 + }, + { + "epoch": 5.95, + "learning_rate": 1.2174190025482345e-06, + "loss": 0.4518, + "step": 32695 + }, + { + "epoch": 5.95, + "learning_rate": 1.2173643975245723e-06, + "loss": 1.0707, + "step": 32696 + }, + { + "epoch": 5.95, + "learning_rate": 1.2173097925009102e-06, + "loss": 2.6135, + "step": 32697 + }, + { + "epoch": 5.95, + "learning_rate": 1.217255187477248e-06, + "loss": 0.0213, + "step": 32698 + }, + { + "epoch": 5.95, + "learning_rate": 1.2172005824535857e-06, + "loss": 1.8609, + "step": 32699 + }, + { + "epoch": 5.95, + "learning_rate": 1.2171459774299234e-06, + "loss": 0.0273, + "step": 32700 + }, + { + "epoch": 5.95, + "learning_rate": 1.2170913724062614e-06, + "loss": 0.1934, + "step": 32701 + }, + { + "epoch": 5.95, + "learning_rate": 1.2170367673825993e-06, + "loss": 0.0727, + "step": 32702 + }, + { + "epoch": 5.95, + "learning_rate": 1.216982162358937e-06, + "loss": 0.2763, + "step": 32703 + }, + { + "epoch": 5.95, + "learning_rate": 1.216927557335275e-06, + "loss": 0.4844, + "step": 32704 + }, + { + "epoch": 5.95, + "learning_rate": 1.2168729523116127e-06, + "loss": 1.3629, + "step": 32705 + }, + { + "epoch": 5.95, + "learning_rate": 1.2168183472879504e-06, + "loss": 0.1869, + "step": 32706 + }, + { + "epoch": 5.95, + "learning_rate": 1.2167637422642882e-06, + "loss": 0.0146, + "step": 32707 + }, + { + "epoch": 5.95, + "learning_rate": 1.2167091372406261e-06, + "loss": 0.0492, + "step": 32708 + }, + { + "epoch": 5.95, + "learning_rate": 1.216654532216964e-06, + "loss": 0.7196, + "step": 32709 + }, + { + "epoch": 5.95, + "learning_rate": 1.2165999271933018e-06, + "loss": 0.0816, + "step": 32710 + }, + { + "epoch": 5.95, + "learning_rate": 1.2165453221696397e-06, + "loss": 1.0181, + "step": 32711 + }, + { + "epoch": 5.95, + "learning_rate": 1.2164907171459775e-06, + "loss": 0.0077, + "step": 32712 + }, + { + "epoch": 5.95, + "learning_rate": 1.2164361121223152e-06, + "loss": 0.3856, + "step": 32713 + }, + { + "epoch": 5.95, + "learning_rate": 1.2163815070986531e-06, + "loss": 1.0102, + "step": 32714 + }, + { + "epoch": 5.95, + "learning_rate": 1.216326902074991e-06, + "loss": 3.7893, + "step": 32715 + }, + { + "epoch": 5.95, + "learning_rate": 1.2162722970513288e-06, + "loss": 0.2434, + "step": 32716 + }, + { + "epoch": 5.95, + "learning_rate": 1.2162176920276665e-06, + "loss": 0.1202, + "step": 32717 + }, + { + "epoch": 5.95, + "learning_rate": 1.2161630870040045e-06, + "loss": 1.4083, + "step": 32718 + }, + { + "epoch": 5.95, + "learning_rate": 1.2161084819803422e-06, + "loss": 1.4265, + "step": 32719 + }, + { + "epoch": 5.95, + "learning_rate": 1.21605387695668e-06, + "loss": 0.2236, + "step": 32720 + }, + { + "epoch": 5.95, + "learning_rate": 1.215999271933018e-06, + "loss": 0.1307, + "step": 32721 + }, + { + "epoch": 5.95, + "learning_rate": 1.2159446669093558e-06, + "loss": 1.3004, + "step": 32722 + }, + { + "epoch": 5.95, + "learning_rate": 1.2158900618856936e-06, + "loss": 1.1057, + "step": 32723 + }, + { + "epoch": 5.95, + "learning_rate": 1.2158354568620313e-06, + "loss": 0.4502, + "step": 32724 + }, + { + "epoch": 5.95, + "learning_rate": 1.2157808518383692e-06, + "loss": 1.2606, + "step": 32725 + }, + { + "epoch": 5.95, + "learning_rate": 1.215726246814707e-06, + "loss": 0.2095, + "step": 32726 + }, + { + "epoch": 5.95, + "learning_rate": 1.2156716417910447e-06, + "loss": 0.01, + "step": 32727 + }, + { + "epoch": 5.95, + "learning_rate": 1.2156170367673827e-06, + "loss": 0.2494, + "step": 32728 + }, + { + "epoch": 5.95, + "learning_rate": 1.2155624317437206e-06, + "loss": 0.1135, + "step": 32729 + }, + { + "epoch": 5.95, + "learning_rate": 1.2155078267200583e-06, + "loss": 0.2188, + "step": 32730 + }, + { + "epoch": 5.95, + "learning_rate": 1.215453221696396e-06, + "loss": 1.1763, + "step": 32731 + }, + { + "epoch": 5.95, + "learning_rate": 1.215398616672734e-06, + "loss": 1.3237, + "step": 32732 + }, + { + "epoch": 5.95, + "learning_rate": 1.2153440116490717e-06, + "loss": 0.4828, + "step": 32733 + }, + { + "epoch": 5.95, + "learning_rate": 1.2152894066254095e-06, + "loss": 0.1791, + "step": 32734 + }, + { + "epoch": 5.95, + "learning_rate": 1.2152348016017474e-06, + "loss": 0.7389, + "step": 32735 + }, + { + "epoch": 5.95, + "learning_rate": 1.2151801965780854e-06, + "loss": 0.1387, + "step": 32736 + }, + { + "epoch": 5.95, + "learning_rate": 1.215125591554423e-06, + "loss": 1.6523, + "step": 32737 + }, + { + "epoch": 5.95, + "learning_rate": 1.2150709865307608e-06, + "loss": 0.2163, + "step": 32738 + }, + { + "epoch": 5.95, + "learning_rate": 1.2150163815070988e-06, + "loss": 1.7433, + "step": 32739 + }, + { + "epoch": 5.95, + "learning_rate": 1.2149617764834365e-06, + "loss": 0.4811, + "step": 32740 + }, + { + "epoch": 5.95, + "learning_rate": 1.2149071714597742e-06, + "loss": 0.8866, + "step": 32741 + }, + { + "epoch": 5.95, + "learning_rate": 1.2148525664361122e-06, + "loss": 4.6893, + "step": 32742 + }, + { + "epoch": 5.95, + "learning_rate": 1.2147979614124501e-06, + "loss": 0.1843, + "step": 32743 + }, + { + "epoch": 5.95, + "learning_rate": 1.2147433563887878e-06, + "loss": 0.0717, + "step": 32744 + }, + { + "epoch": 5.95, + "learning_rate": 1.2146887513651256e-06, + "loss": 0.5702, + "step": 32745 + }, + { + "epoch": 5.95, + "learning_rate": 1.2146341463414635e-06, + "loss": 0.0254, + "step": 32746 + }, + { + "epoch": 5.96, + "learning_rate": 1.2145795413178012e-06, + "loss": 0.4132, + "step": 32747 + }, + { + "epoch": 5.96, + "learning_rate": 1.214524936294139e-06, + "loss": 0.3494, + "step": 32748 + }, + { + "epoch": 5.96, + "learning_rate": 1.214470331270477e-06, + "loss": 2.5766, + "step": 32749 + }, + { + "epoch": 5.96, + "learning_rate": 1.2144157262468149e-06, + "loss": 0.8697, + "step": 32750 + }, + { + "epoch": 5.96, + "learning_rate": 1.2143611212231526e-06, + "loss": 1.1227, + "step": 32751 + }, + { + "epoch": 5.96, + "learning_rate": 1.2143065161994903e-06, + "loss": 0.5597, + "step": 32752 + }, + { + "epoch": 5.96, + "learning_rate": 1.2142519111758283e-06, + "loss": 1.2236, + "step": 32753 + }, + { + "epoch": 5.96, + "learning_rate": 1.214197306152166e-06, + "loss": 1.6052, + "step": 32754 + }, + { + "epoch": 5.96, + "learning_rate": 1.2141427011285037e-06, + "loss": 2.0908, + "step": 32755 + }, + { + "epoch": 5.96, + "learning_rate": 1.2140880961048417e-06, + "loss": 0.0233, + "step": 32756 + }, + { + "epoch": 5.96, + "learning_rate": 1.2140334910811796e-06, + "loss": 0.5979, + "step": 32757 + }, + { + "epoch": 5.96, + "learning_rate": 1.2139788860575174e-06, + "loss": 0.0299, + "step": 32758 + }, + { + "epoch": 5.96, + "learning_rate": 1.213924281033855e-06, + "loss": 2.891, + "step": 32759 + }, + { + "epoch": 5.96, + "learning_rate": 1.2138696760101928e-06, + "loss": 0.7719, + "step": 32760 + }, + { + "epoch": 5.96, + "learning_rate": 1.2138150709865308e-06, + "loss": 1.0659, + "step": 32761 + }, + { + "epoch": 5.96, + "learning_rate": 1.2137604659628687e-06, + "loss": 0.4326, + "step": 32762 + }, + { + "epoch": 5.96, + "learning_rate": 1.2137058609392064e-06, + "loss": 0.1985, + "step": 32763 + }, + { + "epoch": 5.96, + "learning_rate": 1.2136512559155444e-06, + "loss": 0.0388, + "step": 32764 + }, + { + "epoch": 5.96, + "learning_rate": 1.2135966508918821e-06, + "loss": 0.0154, + "step": 32765 + }, + { + "epoch": 5.96, + "learning_rate": 1.2135420458682198e-06, + "loss": 0.0097, + "step": 32766 + }, + { + "epoch": 5.96, + "learning_rate": 1.2134874408445576e-06, + "loss": 0.3927, + "step": 32767 + }, + { + "epoch": 5.96, + "learning_rate": 1.2134328358208955e-06, + "loss": 0.401, + "step": 32768 + }, + { + "epoch": 5.96, + "learning_rate": 1.2133782307972335e-06, + "loss": 0.5221, + "step": 32769 + }, + { + "epoch": 5.96, + "learning_rate": 1.2133236257735712e-06, + "loss": 0.029, + "step": 32770 + }, + { + "epoch": 5.96, + "learning_rate": 1.2132690207499091e-06, + "loss": 0.7425, + "step": 32771 + }, + { + "epoch": 5.96, + "learning_rate": 1.2132144157262469e-06, + "loss": 0.4582, + "step": 32772 + }, + { + "epoch": 5.96, + "learning_rate": 1.2131598107025846e-06, + "loss": 0.2897, + "step": 32773 + }, + { + "epoch": 5.96, + "learning_rate": 1.2131052056789223e-06, + "loss": 3.2756, + "step": 32774 + }, + { + "epoch": 5.96, + "learning_rate": 1.2130506006552603e-06, + "loss": 0.5402, + "step": 32775 + }, + { + "epoch": 5.96, + "learning_rate": 1.2129959956315982e-06, + "loss": 1.9675, + "step": 32776 + }, + { + "epoch": 5.96, + "learning_rate": 1.212941390607936e-06, + "loss": 0.5659, + "step": 32777 + }, + { + "epoch": 5.96, + "learning_rate": 1.2128867855842739e-06, + "loss": 0.1557, + "step": 32778 + }, + { + "epoch": 5.96, + "learning_rate": 1.2128321805606116e-06, + "loss": 0.8986, + "step": 32779 + }, + { + "epoch": 5.96, + "learning_rate": 1.2127775755369494e-06, + "loss": 0.1778, + "step": 32780 + }, + { + "epoch": 5.96, + "learning_rate": 1.212722970513287e-06, + "loss": 1.2197, + "step": 32781 + }, + { + "epoch": 5.96, + "learning_rate": 1.212668365489625e-06, + "loss": 0.0172, + "step": 32782 + }, + { + "epoch": 5.96, + "learning_rate": 1.212613760465963e-06, + "loss": 3.6871, + "step": 32783 + }, + { + "epoch": 5.96, + "learning_rate": 1.2125591554423007e-06, + "loss": 0.0214, + "step": 32784 + }, + { + "epoch": 5.96, + "learning_rate": 1.2125045504186386e-06, + "loss": 0.0105, + "step": 32785 + }, + { + "epoch": 5.96, + "learning_rate": 1.2124499453949764e-06, + "loss": 0.0365, + "step": 32786 + }, + { + "epoch": 5.96, + "learning_rate": 1.2123953403713141e-06, + "loss": 1.0302, + "step": 32787 + }, + { + "epoch": 5.96, + "learning_rate": 1.2123407353476518e-06, + "loss": 0.0741, + "step": 32788 + }, + { + "epoch": 5.96, + "learning_rate": 1.2122861303239898e-06, + "loss": 0.0783, + "step": 32789 + }, + { + "epoch": 5.96, + "learning_rate": 1.2122315253003277e-06, + "loss": 0.1313, + "step": 32790 + }, + { + "epoch": 5.96, + "learning_rate": 1.2121769202766655e-06, + "loss": 1.7831, + "step": 32791 + }, + { + "epoch": 5.96, + "learning_rate": 1.2121223152530034e-06, + "loss": 0.9844, + "step": 32792 + }, + { + "epoch": 5.96, + "learning_rate": 1.2120677102293411e-06, + "loss": 0.0256, + "step": 32793 + }, + { + "epoch": 5.96, + "learning_rate": 1.2120131052056789e-06, + "loss": 0.6423, + "step": 32794 + }, + { + "epoch": 5.96, + "learning_rate": 1.2119585001820166e-06, + "loss": 1.1347, + "step": 32795 + }, + { + "epoch": 5.96, + "learning_rate": 1.2119038951583545e-06, + "loss": 1.1464, + "step": 32796 + }, + { + "epoch": 5.96, + "learning_rate": 1.2118492901346925e-06, + "loss": 0.3299, + "step": 32797 + }, + { + "epoch": 5.96, + "learning_rate": 1.2117946851110302e-06, + "loss": 0.701, + "step": 32798 + }, + { + "epoch": 5.96, + "learning_rate": 1.2117400800873682e-06, + "loss": 0.0113, + "step": 32799 + }, + { + "epoch": 5.96, + "learning_rate": 1.2116854750637059e-06, + "loss": 1.6612, + "step": 32800 + }, + { + "epoch": 5.96, + "learning_rate": 1.2116308700400436e-06, + "loss": 0.0895, + "step": 32801 + }, + { + "epoch": 5.97, + "learning_rate": 1.2115762650163816e-06, + "loss": 0.7216, + "step": 32802 + }, + { + "epoch": 5.97, + "learning_rate": 1.2115216599927195e-06, + "loss": 0.0989, + "step": 32803 + }, + { + "epoch": 5.97, + "learning_rate": 1.2114670549690572e-06, + "loss": 0.0404, + "step": 32804 + }, + { + "epoch": 5.97, + "learning_rate": 1.211412449945395e-06, + "loss": 0.6039, + "step": 32805 + }, + { + "epoch": 5.97, + "learning_rate": 1.211357844921733e-06, + "loss": 1.8117, + "step": 32806 + }, + { + "epoch": 5.97, + "learning_rate": 1.2113032398980707e-06, + "loss": 0.9251, + "step": 32807 + }, + { + "epoch": 5.97, + "learning_rate": 1.2112486348744084e-06, + "loss": 1.3519, + "step": 32808 + }, + { + "epoch": 5.97, + "learning_rate": 1.2111940298507463e-06, + "loss": 0.1488, + "step": 32809 + }, + { + "epoch": 5.97, + "learning_rate": 1.2111394248270843e-06, + "loss": 1.6798, + "step": 32810 + }, + { + "epoch": 5.97, + "learning_rate": 1.211084819803422e-06, + "loss": 1.2303, + "step": 32811 + }, + { + "epoch": 5.97, + "learning_rate": 1.2110302147797597e-06, + "loss": 0.4912, + "step": 32812 + }, + { + "epoch": 5.97, + "learning_rate": 1.2109756097560977e-06, + "loss": 0.2152, + "step": 32813 + }, + { + "epoch": 5.97, + "learning_rate": 1.2109210047324354e-06, + "loss": 0.1742, + "step": 32814 + }, + { + "epoch": 5.97, + "learning_rate": 1.2108663997087731e-06, + "loss": 1.5739, + "step": 32815 + }, + { + "epoch": 5.97, + "learning_rate": 1.210811794685111e-06, + "loss": 0.343, + "step": 32816 + }, + { + "epoch": 5.97, + "learning_rate": 1.210757189661449e-06, + "loss": 1.7142, + "step": 32817 + }, + { + "epoch": 5.97, + "learning_rate": 1.2107025846377868e-06, + "loss": 0.5449, + "step": 32818 + }, + { + "epoch": 5.97, + "learning_rate": 1.2106479796141245e-06, + "loss": 0.022, + "step": 32819 + }, + { + "epoch": 5.97, + "learning_rate": 1.2105933745904624e-06, + "loss": 0.4977, + "step": 32820 + }, + { + "epoch": 5.97, + "learning_rate": 1.2105387695668002e-06, + "loss": 0.3572, + "step": 32821 + }, + { + "epoch": 5.97, + "learning_rate": 1.210484164543138e-06, + "loss": 1.3968, + "step": 32822 + }, + { + "epoch": 5.97, + "learning_rate": 1.2104295595194758e-06, + "loss": 0.626, + "step": 32823 + }, + { + "epoch": 5.97, + "learning_rate": 1.2103749544958138e-06, + "loss": 0.1638, + "step": 32824 + }, + { + "epoch": 5.97, + "learning_rate": 1.2103203494721515e-06, + "loss": 1.3103, + "step": 32825 + }, + { + "epoch": 5.97, + "learning_rate": 1.2102657444484892e-06, + "loss": 0.0545, + "step": 32826 + }, + { + "epoch": 5.97, + "learning_rate": 1.2102111394248272e-06, + "loss": 0.1165, + "step": 32827 + }, + { + "epoch": 5.97, + "learning_rate": 1.210156534401165e-06, + "loss": 0.1677, + "step": 32828 + }, + { + "epoch": 5.97, + "learning_rate": 1.2101019293775027e-06, + "loss": 0.6199, + "step": 32829 + }, + { + "epoch": 5.97, + "learning_rate": 1.2100473243538406e-06, + "loss": 0.5878, + "step": 32830 + }, + { + "epoch": 5.97, + "learning_rate": 1.2099927193301785e-06, + "loss": 0.0319, + "step": 32831 + }, + { + "epoch": 5.97, + "learning_rate": 1.2099381143065163e-06, + "loss": 0.005, + "step": 32832 + }, + { + "epoch": 5.97, + "learning_rate": 1.209883509282854e-06, + "loss": 1.641, + "step": 32833 + }, + { + "epoch": 5.97, + "learning_rate": 1.209828904259192e-06, + "loss": 0.7953, + "step": 32834 + }, + { + "epoch": 5.97, + "learning_rate": 1.2097742992355297e-06, + "loss": 0.2665, + "step": 32835 + }, + { + "epoch": 5.97, + "learning_rate": 1.2097196942118674e-06, + "loss": 1.0545, + "step": 32836 + }, + { + "epoch": 5.97, + "learning_rate": 1.2096650891882054e-06, + "loss": 0.2016, + "step": 32837 + }, + { + "epoch": 5.97, + "learning_rate": 1.2096104841645433e-06, + "loss": 1.9091, + "step": 32838 + }, + { + "epoch": 5.97, + "learning_rate": 1.209555879140881e-06, + "loss": 0.0257, + "step": 32839 + }, + { + "epoch": 5.97, + "learning_rate": 1.2095012741172188e-06, + "loss": 0.0122, + "step": 32840 + }, + { + "epoch": 5.97, + "learning_rate": 1.2094466690935565e-06, + "loss": 0.1559, + "step": 32841 + }, + { + "epoch": 5.97, + "learning_rate": 1.2093920640698944e-06, + "loss": 0.4504, + "step": 32842 + }, + { + "epoch": 5.97, + "learning_rate": 1.2093374590462324e-06, + "loss": 0.0214, + "step": 32843 + }, + { + "epoch": 5.97, + "learning_rate": 1.2092828540225701e-06, + "loss": 0.0072, + "step": 32844 + }, + { + "epoch": 5.97, + "learning_rate": 1.209228248998908e-06, + "loss": 1.4981, + "step": 32845 + }, + { + "epoch": 5.97, + "learning_rate": 1.2091736439752458e-06, + "loss": 0.0233, + "step": 32846 + }, + { + "epoch": 5.97, + "learning_rate": 1.2091190389515835e-06, + "loss": 0.12, + "step": 32847 + }, + { + "epoch": 5.97, + "learning_rate": 1.2090644339279212e-06, + "loss": 0.2759, + "step": 32848 + }, + { + "epoch": 5.97, + "learning_rate": 1.2090098289042592e-06, + "loss": 0.2314, + "step": 32849 + }, + { + "epoch": 5.97, + "learning_rate": 1.2089552238805971e-06, + "loss": 0.4767, + "step": 32850 + }, + { + "epoch": 5.97, + "learning_rate": 1.2089006188569349e-06, + "loss": 2.1861, + "step": 32851 + }, + { + "epoch": 5.97, + "learning_rate": 1.2088460138332728e-06, + "loss": 0.01, + "step": 32852 + }, + { + "epoch": 5.97, + "learning_rate": 1.2087914088096105e-06, + "loss": 0.3772, + "step": 32853 + }, + { + "epoch": 5.97, + "learning_rate": 1.2087368037859483e-06, + "loss": 0.213, + "step": 32854 + }, + { + "epoch": 5.97, + "learning_rate": 1.208682198762286e-06, + "loss": 1.3463, + "step": 32855 + }, + { + "epoch": 5.97, + "learning_rate": 1.208627593738624e-06, + "loss": 0.1296, + "step": 32856 + }, + { + "epoch": 5.98, + "learning_rate": 1.2085729887149619e-06, + "loss": 0.9464, + "step": 32857 + }, + { + "epoch": 5.98, + "learning_rate": 1.2085183836912996e-06, + "loss": 0.1023, + "step": 32858 + }, + { + "epoch": 5.98, + "learning_rate": 1.2084637786676376e-06, + "loss": 1.7779, + "step": 32859 + }, + { + "epoch": 5.98, + "learning_rate": 1.2084091736439753e-06, + "loss": 0.2888, + "step": 32860 + }, + { + "epoch": 5.98, + "learning_rate": 1.208354568620313e-06, + "loss": 0.0453, + "step": 32861 + }, + { + "epoch": 5.98, + "learning_rate": 1.2082999635966508e-06, + "loss": 1.2442, + "step": 32862 + }, + { + "epoch": 5.98, + "learning_rate": 1.2082453585729887e-06, + "loss": 0.9643, + "step": 32863 + }, + { + "epoch": 5.98, + "learning_rate": 1.2081907535493266e-06, + "loss": 0.0187, + "step": 32864 + }, + { + "epoch": 5.98, + "learning_rate": 1.2081361485256644e-06, + "loss": 0.3405, + "step": 32865 + }, + { + "epoch": 5.98, + "learning_rate": 1.2080815435020023e-06, + "loss": 0.841, + "step": 32866 + }, + { + "epoch": 5.98, + "learning_rate": 1.20802693847834e-06, + "loss": 0.0199, + "step": 32867 + }, + { + "epoch": 5.98, + "learning_rate": 1.2079723334546778e-06, + "loss": 1.632, + "step": 32868 + }, + { + "epoch": 5.98, + "learning_rate": 1.2079177284310155e-06, + "loss": 0.0779, + "step": 32869 + }, + { + "epoch": 5.98, + "learning_rate": 1.2078631234073535e-06, + "loss": 0.6244, + "step": 32870 + }, + { + "epoch": 5.98, + "learning_rate": 1.2078085183836914e-06, + "loss": 0.1788, + "step": 32871 + }, + { + "epoch": 5.98, + "learning_rate": 1.2077539133600291e-06, + "loss": 0.0319, + "step": 32872 + }, + { + "epoch": 5.98, + "learning_rate": 1.207699308336367e-06, + "loss": 0.016, + "step": 32873 + }, + { + "epoch": 5.98, + "learning_rate": 1.2076447033127048e-06, + "loss": 2.3239, + "step": 32874 + }, + { + "epoch": 5.98, + "learning_rate": 1.2075900982890425e-06, + "loss": 1.1186, + "step": 32875 + }, + { + "epoch": 5.98, + "learning_rate": 1.2075354932653803e-06, + "loss": 0.8676, + "step": 32876 + }, + { + "epoch": 5.98, + "learning_rate": 1.2074808882417182e-06, + "loss": 1.1858, + "step": 32877 + }, + { + "epoch": 5.98, + "learning_rate": 1.2074262832180562e-06, + "loss": 0.1119, + "step": 32878 + }, + { + "epoch": 5.98, + "learning_rate": 1.2073716781943939e-06, + "loss": 2.0797, + "step": 32879 + }, + { + "epoch": 5.98, + "learning_rate": 1.2073170731707318e-06, + "loss": 3.0345, + "step": 32880 + }, + { + "epoch": 5.98, + "learning_rate": 1.2072624681470696e-06, + "loss": 0.0941, + "step": 32881 + }, + { + "epoch": 5.98, + "learning_rate": 1.2072078631234073e-06, + "loss": 1.0903, + "step": 32882 + }, + { + "epoch": 5.98, + "learning_rate": 1.2071532580997452e-06, + "loss": 1.6998, + "step": 32883 + }, + { + "epoch": 5.98, + "learning_rate": 1.2070986530760832e-06, + "loss": 1.2597, + "step": 32884 + }, + { + "epoch": 5.98, + "learning_rate": 1.207044048052421e-06, + "loss": 0.8421, + "step": 32885 + }, + { + "epoch": 5.98, + "learning_rate": 1.2069894430287586e-06, + "loss": 0.0037, + "step": 32886 + }, + { + "epoch": 5.98, + "learning_rate": 1.2069348380050966e-06, + "loss": 0.0067, + "step": 32887 + }, + { + "epoch": 5.98, + "learning_rate": 1.2068802329814343e-06, + "loss": 0.011, + "step": 32888 + }, + { + "epoch": 5.98, + "learning_rate": 1.206825627957772e-06, + "loss": 0.4745, + "step": 32889 + }, + { + "epoch": 5.98, + "learning_rate": 1.20677102293411e-06, + "loss": 0.7364, + "step": 32890 + }, + { + "epoch": 5.98, + "learning_rate": 1.206716417910448e-06, + "loss": 1.4817, + "step": 32891 + }, + { + "epoch": 5.98, + "learning_rate": 1.2066618128867857e-06, + "loss": 0.0096, + "step": 32892 + }, + { + "epoch": 5.98, + "learning_rate": 1.2066072078631234e-06, + "loss": 1.216, + "step": 32893 + }, + { + "epoch": 5.98, + "learning_rate": 1.2065526028394613e-06, + "loss": 1.82, + "step": 32894 + }, + { + "epoch": 5.98, + "learning_rate": 1.206497997815799e-06, + "loss": 3.2514, + "step": 32895 + }, + { + "epoch": 5.98, + "learning_rate": 1.2064433927921368e-06, + "loss": 0.0347, + "step": 32896 + }, + { + "epoch": 5.98, + "learning_rate": 1.2063887877684748e-06, + "loss": 0.8637, + "step": 32897 + }, + { + "epoch": 5.98, + "learning_rate": 1.2063341827448127e-06, + "loss": 3.1474, + "step": 32898 + }, + { + "epoch": 5.98, + "learning_rate": 1.2062795777211504e-06, + "loss": 0.009, + "step": 32899 + }, + { + "epoch": 5.98, + "learning_rate": 1.2062249726974882e-06, + "loss": 0.732, + "step": 32900 + }, + { + "epoch": 5.98, + "learning_rate": 1.2061703676738261e-06, + "loss": 0.0761, + "step": 32901 + }, + { + "epoch": 5.98, + "learning_rate": 1.2061157626501638e-06, + "loss": 0.2631, + "step": 32902 + }, + { + "epoch": 5.98, + "learning_rate": 1.2060611576265016e-06, + "loss": 0.0342, + "step": 32903 + }, + { + "epoch": 5.98, + "learning_rate": 1.2060065526028395e-06, + "loss": 0.0163, + "step": 32904 + }, + { + "epoch": 5.98, + "learning_rate": 1.2059519475791775e-06, + "loss": 0.3038, + "step": 32905 + }, + { + "epoch": 5.98, + "learning_rate": 1.2058973425555152e-06, + "loss": 0.4072, + "step": 32906 + }, + { + "epoch": 5.98, + "learning_rate": 1.205842737531853e-06, + "loss": 0.5813, + "step": 32907 + }, + { + "epoch": 5.98, + "learning_rate": 1.2057881325081909e-06, + "loss": 0.5983, + "step": 32908 + }, + { + "epoch": 5.98, + "learning_rate": 1.2057335274845286e-06, + "loss": 2.505, + "step": 32909 + }, + { + "epoch": 5.98, + "learning_rate": 1.2056789224608663e-06, + "loss": 1.248, + "step": 32910 + }, + { + "epoch": 5.98, + "learning_rate": 1.2056243174372043e-06, + "loss": 0.0914, + "step": 32911 + }, + { + "epoch": 5.99, + "learning_rate": 1.2055697124135422e-06, + "loss": 0.3191, + "step": 32912 + }, + { + "epoch": 5.99, + "learning_rate": 1.20551510738988e-06, + "loss": 2.0069, + "step": 32913 + }, + { + "epoch": 5.99, + "learning_rate": 1.2054605023662177e-06, + "loss": 0.8149, + "step": 32914 + }, + { + "epoch": 5.99, + "learning_rate": 1.2054058973425554e-06, + "loss": 0.665, + "step": 32915 + }, + { + "epoch": 5.99, + "learning_rate": 1.2053512923188934e-06, + "loss": 0.8036, + "step": 32916 + }, + { + "epoch": 5.99, + "learning_rate": 1.205296687295231e-06, + "loss": 0.1241, + "step": 32917 + }, + { + "epoch": 5.99, + "learning_rate": 1.205242082271569e-06, + "loss": 0.171, + "step": 32918 + }, + { + "epoch": 5.99, + "learning_rate": 1.205187477247907e-06, + "loss": 1.1219, + "step": 32919 + }, + { + "epoch": 5.99, + "learning_rate": 1.2051328722242447e-06, + "loss": 0.0227, + "step": 32920 + }, + { + "epoch": 5.99, + "learning_rate": 1.2050782672005824e-06, + "loss": 0.0098, + "step": 32921 + }, + { + "epoch": 5.99, + "learning_rate": 1.2050236621769202e-06, + "loss": 1.5674, + "step": 32922 + }, + { + "epoch": 5.99, + "learning_rate": 1.2049690571532581e-06, + "loss": 0.2767, + "step": 32923 + }, + { + "epoch": 5.99, + "learning_rate": 1.2049144521295958e-06, + "loss": 0.2427, + "step": 32924 + }, + { + "epoch": 5.99, + "learning_rate": 1.2048598471059338e-06, + "loss": 0.056, + "step": 32925 + }, + { + "epoch": 5.99, + "learning_rate": 1.2048052420822717e-06, + "loss": 2.9443, + "step": 32926 + }, + { + "epoch": 5.99, + "learning_rate": 1.2047506370586095e-06, + "loss": 0.0062, + "step": 32927 + }, + { + "epoch": 5.99, + "learning_rate": 1.2046960320349472e-06, + "loss": 0.8479, + "step": 32928 + }, + { + "epoch": 5.99, + "learning_rate": 1.204641427011285e-06, + "loss": 0.6026, + "step": 32929 + }, + { + "epoch": 5.99, + "learning_rate": 1.2045868219876229e-06, + "loss": 0.0121, + "step": 32930 + }, + { + "epoch": 5.99, + "learning_rate": 1.2045322169639608e-06, + "loss": 1.2418, + "step": 32931 + }, + { + "epoch": 5.99, + "learning_rate": 1.2044776119402985e-06, + "loss": 0.1669, + "step": 32932 + }, + { + "epoch": 5.99, + "learning_rate": 1.2044230069166365e-06, + "loss": 2.8086, + "step": 32933 + }, + { + "epoch": 5.99, + "learning_rate": 1.2043684018929742e-06, + "loss": 0.1646, + "step": 32934 + }, + { + "epoch": 5.99, + "learning_rate": 1.204313796869312e-06, + "loss": 1.1556, + "step": 32935 + }, + { + "epoch": 5.99, + "learning_rate": 1.2042591918456497e-06, + "loss": 0.3337, + "step": 32936 + }, + { + "epoch": 5.99, + "learning_rate": 1.2042045868219876e-06, + "loss": 0.0848, + "step": 32937 + }, + { + "epoch": 5.99, + "learning_rate": 1.2041499817983256e-06, + "loss": 0.5439, + "step": 32938 + }, + { + "epoch": 5.99, + "learning_rate": 1.2040953767746633e-06, + "loss": 0.0369, + "step": 32939 + }, + { + "epoch": 5.99, + "learning_rate": 1.2040407717510012e-06, + "loss": 0.3459, + "step": 32940 + }, + { + "epoch": 5.99, + "learning_rate": 1.203986166727339e-06, + "loss": 0.2431, + "step": 32941 + }, + { + "epoch": 5.99, + "learning_rate": 1.2039315617036767e-06, + "loss": 0.6769, + "step": 32942 + }, + { + "epoch": 5.99, + "learning_rate": 1.2038769566800144e-06, + "loss": 1.9889, + "step": 32943 + }, + { + "epoch": 5.99, + "learning_rate": 1.2038223516563524e-06, + "loss": 0.0093, + "step": 32944 + }, + { + "epoch": 5.99, + "learning_rate": 1.2037677466326903e-06, + "loss": 0.2207, + "step": 32945 + }, + { + "epoch": 5.99, + "learning_rate": 1.203713141609028e-06, + "loss": 0.0737, + "step": 32946 + }, + { + "epoch": 5.99, + "learning_rate": 1.203658536585366e-06, + "loss": 1.2263, + "step": 32947 + }, + { + "epoch": 5.99, + "learning_rate": 1.2036039315617037e-06, + "loss": 0.0078, + "step": 32948 + }, + { + "epoch": 5.99, + "learning_rate": 1.2035493265380415e-06, + "loss": 2.3232, + "step": 32949 + }, + { + "epoch": 5.99, + "learning_rate": 1.2034947215143792e-06, + "loss": 0.2948, + "step": 32950 + }, + { + "epoch": 5.99, + "learning_rate": 1.2034401164907171e-06, + "loss": 0.1072, + "step": 32951 + }, + { + "epoch": 5.99, + "learning_rate": 1.203385511467055e-06, + "loss": 0.149, + "step": 32952 + }, + { + "epoch": 5.99, + "learning_rate": 1.2033309064433928e-06, + "loss": 0.3355, + "step": 32953 + }, + { + "epoch": 5.99, + "learning_rate": 1.2032763014197308e-06, + "loss": 0.233, + "step": 32954 + }, + { + "epoch": 5.99, + "learning_rate": 1.2032216963960685e-06, + "loss": 0.2873, + "step": 32955 + }, + { + "epoch": 5.99, + "learning_rate": 1.2031670913724062e-06, + "loss": 0.3211, + "step": 32956 + }, + { + "epoch": 5.99, + "learning_rate": 1.203112486348744e-06, + "loss": 0.6971, + "step": 32957 + }, + { + "epoch": 5.99, + "learning_rate": 1.2030578813250819e-06, + "loss": 1.0513, + "step": 32958 + }, + { + "epoch": 5.99, + "learning_rate": 1.2030032763014198e-06, + "loss": 0.0832, + "step": 32959 + }, + { + "epoch": 5.99, + "learning_rate": 1.2029486712777576e-06, + "loss": 0.1466, + "step": 32960 + }, + { + "epoch": 5.99, + "learning_rate": 1.2028940662540955e-06, + "loss": 0.1496, + "step": 32961 + }, + { + "epoch": 5.99, + "learning_rate": 1.2028394612304332e-06, + "loss": 1.9544, + "step": 32962 + }, + { + "epoch": 5.99, + "learning_rate": 1.202784856206771e-06, + "loss": 2.2774, + "step": 32963 + }, + { + "epoch": 5.99, + "learning_rate": 1.2027302511831087e-06, + "loss": 1.1375, + "step": 32964 + }, + { + "epoch": 5.99, + "learning_rate": 1.2026756461594466e-06, + "loss": 0.069, + "step": 32965 + }, + { + "epoch": 5.99, + "learning_rate": 1.2026210411357846e-06, + "loss": 0.0953, + "step": 32966 + }, + { + "epoch": 6.0, + "learning_rate": 1.2025664361121223e-06, + "loss": 0.2262, + "step": 32967 + }, + { + "epoch": 6.0, + "learning_rate": 1.2025118310884603e-06, + "loss": 0.3242, + "step": 32968 + }, + { + "epoch": 6.0, + "learning_rate": 1.202457226064798e-06, + "loss": 0.7696, + "step": 32969 + }, + { + "epoch": 6.0, + "learning_rate": 1.2024026210411357e-06, + "loss": 0.3877, + "step": 32970 + }, + { + "epoch": 6.0, + "learning_rate": 1.2023480160174737e-06, + "loss": 1.5831, + "step": 32971 + }, + { + "epoch": 6.0, + "learning_rate": 1.2022934109938116e-06, + "loss": 0.1565, + "step": 32972 + }, + { + "epoch": 6.0, + "learning_rate": 1.2022388059701493e-06, + "loss": 0.4991, + "step": 32973 + }, + { + "epoch": 6.0, + "learning_rate": 1.202184200946487e-06, + "loss": 0.2411, + "step": 32974 + }, + { + "epoch": 6.0, + "learning_rate": 1.202129595922825e-06, + "loss": 0.0149, + "step": 32975 + }, + { + "epoch": 6.0, + "learning_rate": 1.2020749908991628e-06, + "loss": 1.6551, + "step": 32976 + }, + { + "epoch": 6.0, + "learning_rate": 1.2020203858755005e-06, + "loss": 0.3464, + "step": 32977 + }, + { + "epoch": 6.0, + "learning_rate": 1.2019657808518384e-06, + "loss": 1.2496, + "step": 32978 + }, + { + "epoch": 6.0, + "learning_rate": 1.2019111758281764e-06, + "loss": 0.1858, + "step": 32979 + }, + { + "epoch": 6.0, + "learning_rate": 1.201856570804514e-06, + "loss": 0.1534, + "step": 32980 + }, + { + "epoch": 6.0, + "learning_rate": 1.2018019657808518e-06, + "loss": 0.1379, + "step": 32981 + }, + { + "epoch": 6.0, + "learning_rate": 1.2017473607571898e-06, + "loss": 0.1319, + "step": 32982 + }, + { + "epoch": 6.0, + "learning_rate": 1.2016927557335275e-06, + "loss": 0.2225, + "step": 32983 + }, + { + "epoch": 6.0, + "learning_rate": 1.2016381507098652e-06, + "loss": 0.8536, + "step": 32984 + }, + { + "epoch": 6.0, + "learning_rate": 1.2015835456862032e-06, + "loss": 0.0043, + "step": 32985 + }, + { + "epoch": 6.0, + "learning_rate": 1.2015289406625411e-06, + "loss": 0.0972, + "step": 32986 + }, + { + "epoch": 6.0, + "learning_rate": 1.2014743356388789e-06, + "loss": 0.1778, + "step": 32987 + }, + { + "epoch": 6.0, + "learning_rate": 1.2014197306152166e-06, + "loss": 0.5174, + "step": 32988 + }, + { + "epoch": 6.0, + "learning_rate": 1.2013651255915545e-06, + "loss": 0.1617, + "step": 32989 + }, + { + "epoch": 6.0, + "learning_rate": 1.2013105205678923e-06, + "loss": 0.0826, + "step": 32990 + }, + { + "epoch": 6.0, + "learning_rate": 1.20125591554423e-06, + "loss": 2.1184, + "step": 32991 + }, + { + "epoch": 6.0, + "learning_rate": 1.201201310520568e-06, + "loss": 1.463, + "step": 32992 + }, + { + "epoch": 6.0, + "learning_rate": 1.2011467054969059e-06, + "loss": 0.0582, + "step": 32993 + }, + { + "epoch": 6.0, + "learning_rate": 1.2010921004732436e-06, + "loss": 0.0127, + "step": 32994 + }, + { + "epoch": 6.0, + "eval_accuracy": 0.8580196399345336, + "eval_loss": 0.6682834625244141, + "eval_runtime": 139.6383, + "eval_samples_per_second": 17.502, + "eval_steps_per_second": 4.376, + "step": 32994 + }, + { + "epoch": 6.0, + "learning_rate": 1.2010374954495814e-06, + "loss": 0.0132, + "step": 32995 + }, + { + "epoch": 6.0, + "learning_rate": 1.200982890425919e-06, + "loss": 0.0068, + "step": 32996 + }, + { + "epoch": 6.0, + "learning_rate": 1.200928285402257e-06, + "loss": 3.0781, + "step": 32997 + }, + { + "epoch": 6.0, + "learning_rate": 1.2008736803785948e-06, + "loss": 0.8521, + "step": 32998 + }, + { + "epoch": 6.0, + "learning_rate": 1.2008190753549327e-06, + "loss": 1.678, + "step": 32999 + }, + { + "epoch": 6.0, + "learning_rate": 1.2007644703312706e-06, + "loss": 0.1482, + "step": 33000 + }, + { + "epoch": 6.0, + "learning_rate": 1.2007098653076084e-06, + "loss": 0.0026, + "step": 33001 + }, + { + "epoch": 6.0, + "learning_rate": 1.2006552602839461e-06, + "loss": 0.2532, + "step": 33002 + }, + { + "epoch": 6.0, + "learning_rate": 1.2006006552602838e-06, + "loss": 0.4958, + "step": 33003 + }, + { + "epoch": 6.0, + "learning_rate": 1.2005460502366218e-06, + "loss": 0.0089, + "step": 33004 + }, + { + "epoch": 6.0, + "learning_rate": 1.2004914452129595e-06, + "loss": 0.4711, + "step": 33005 + }, + { + "epoch": 6.0, + "learning_rate": 1.2004368401892975e-06, + "loss": 0.0227, + "step": 33006 + }, + { + "epoch": 6.0, + "learning_rate": 1.2003822351656354e-06, + "loss": 1.1486, + "step": 33007 + }, + { + "epoch": 6.0, + "learning_rate": 1.2003276301419731e-06, + "loss": 0.008, + "step": 33008 + }, + { + "epoch": 6.0, + "learning_rate": 1.2002730251183109e-06, + "loss": 0.1321, + "step": 33009 + }, + { + "epoch": 6.0, + "learning_rate": 1.2002184200946486e-06, + "loss": 0.6351, + "step": 33010 + }, + { + "epoch": 6.0, + "learning_rate": 1.2001638150709865e-06, + "loss": 0.018, + "step": 33011 + }, + { + "epoch": 6.0, + "learning_rate": 1.2001092100473245e-06, + "loss": 0.5942, + "step": 33012 + }, + { + "epoch": 6.0, + "learning_rate": 1.2000546050236622e-06, + "loss": 0.2048, + "step": 33013 + }, + { + "epoch": 6.0, + "learning_rate": 1.2000000000000002e-06, + "loss": 2.6354, + "step": 33014 + }, + { + "epoch": 6.0, + "learning_rate": 1.1999453949763379e-06, + "loss": 0.0101, + "step": 33015 + }, + { + "epoch": 6.0, + "learning_rate": 1.1998907899526756e-06, + "loss": 0.6235, + "step": 33016 + }, + { + "epoch": 6.0, + "learning_rate": 1.1998361849290134e-06, + "loss": 0.0146, + "step": 33017 + }, + { + "epoch": 6.0, + "learning_rate": 1.1997815799053513e-06, + "loss": 1.1325, + "step": 33018 + }, + { + "epoch": 6.0, + "learning_rate": 1.1997269748816892e-06, + "loss": 0.624, + "step": 33019 + }, + { + "epoch": 6.0, + "learning_rate": 1.199672369858027e-06, + "loss": 0.0079, + "step": 33020 + }, + { + "epoch": 6.0, + "learning_rate": 1.199617764834365e-06, + "loss": 0.1517, + "step": 33021 + }, + { + "epoch": 6.01, + "learning_rate": 1.1995631598107026e-06, + "loss": 0.597, + "step": 33022 + }, + { + "epoch": 6.01, + "learning_rate": 1.1995085547870404e-06, + "loss": 2.7134, + "step": 33023 + }, + { + "epoch": 6.01, + "learning_rate": 1.1994539497633781e-06, + "loss": 0.5168, + "step": 33024 + }, + { + "epoch": 6.01, + "learning_rate": 1.199399344739716e-06, + "loss": 3.0973, + "step": 33025 + }, + { + "epoch": 6.01, + "learning_rate": 1.199344739716054e-06, + "loss": 1.2241, + "step": 33026 + }, + { + "epoch": 6.01, + "learning_rate": 1.1992901346923917e-06, + "loss": 0.331, + "step": 33027 + }, + { + "epoch": 6.01, + "learning_rate": 1.1992355296687297e-06, + "loss": 0.8531, + "step": 33028 + }, + { + "epoch": 6.01, + "learning_rate": 1.1991809246450674e-06, + "loss": 0.0177, + "step": 33029 + }, + { + "epoch": 6.01, + "learning_rate": 1.1991263196214051e-06, + "loss": 0.298, + "step": 33030 + }, + { + "epoch": 6.01, + "learning_rate": 1.1990717145977429e-06, + "loss": 0.0521, + "step": 33031 + }, + { + "epoch": 6.01, + "learning_rate": 1.1990171095740808e-06, + "loss": 0.1748, + "step": 33032 + }, + { + "epoch": 6.01, + "learning_rate": 1.1989625045504188e-06, + "loss": 0.0149, + "step": 33033 + }, + { + "epoch": 6.01, + "learning_rate": 1.1989078995267565e-06, + "loss": 0.5234, + "step": 33034 + }, + { + "epoch": 6.01, + "learning_rate": 1.1988532945030944e-06, + "loss": 0.2855, + "step": 33035 + }, + { + "epoch": 6.01, + "learning_rate": 1.1987986894794322e-06, + "loss": 0.5749, + "step": 33036 + }, + { + "epoch": 6.01, + "learning_rate": 1.1987440844557699e-06, + "loss": 0.4923, + "step": 33037 + }, + { + "epoch": 6.01, + "learning_rate": 1.1986894794321076e-06, + "loss": 0.4419, + "step": 33038 + }, + { + "epoch": 6.01, + "learning_rate": 1.1986348744084456e-06, + "loss": 0.0484, + "step": 33039 + }, + { + "epoch": 6.01, + "learning_rate": 1.1985802693847835e-06, + "loss": 0.2849, + "step": 33040 + }, + { + "epoch": 6.01, + "learning_rate": 1.1985256643611212e-06, + "loss": 0.0191, + "step": 33041 + }, + { + "epoch": 6.01, + "learning_rate": 1.1984710593374592e-06, + "loss": 1.1509, + "step": 33042 + }, + { + "epoch": 6.01, + "learning_rate": 1.198416454313797e-06, + "loss": 0.2205, + "step": 33043 + }, + { + "epoch": 6.01, + "learning_rate": 1.1983618492901346e-06, + "loss": 0.1864, + "step": 33044 + }, + { + "epoch": 6.01, + "learning_rate": 1.1983072442664724e-06, + "loss": 0.0106, + "step": 33045 + }, + { + "epoch": 6.01, + "learning_rate": 1.1982526392428103e-06, + "loss": 0.1898, + "step": 33046 + }, + { + "epoch": 6.01, + "learning_rate": 1.1981980342191483e-06, + "loss": 0.1929, + "step": 33047 + }, + { + "epoch": 6.01, + "learning_rate": 1.198143429195486e-06, + "loss": 1.8451, + "step": 33048 + }, + { + "epoch": 6.01, + "learning_rate": 1.198088824171824e-06, + "loss": 0.9796, + "step": 33049 + }, + { + "epoch": 6.01, + "learning_rate": 1.1980342191481617e-06, + "loss": 0.4629, + "step": 33050 + }, + { + "epoch": 6.01, + "learning_rate": 1.1979796141244994e-06, + "loss": 0.3458, + "step": 33051 + }, + { + "epoch": 6.01, + "learning_rate": 1.1979250091008373e-06, + "loss": 1.3634, + "step": 33052 + }, + { + "epoch": 6.01, + "learning_rate": 1.197870404077175e-06, + "loss": 0.5195, + "step": 33053 + }, + { + "epoch": 6.01, + "learning_rate": 1.197815799053513e-06, + "loss": 0.3475, + "step": 33054 + }, + { + "epoch": 6.01, + "learning_rate": 1.1977611940298508e-06, + "loss": 0.0147, + "step": 33055 + }, + { + "epoch": 6.01, + "learning_rate": 1.1977065890061887e-06, + "loss": 0.0692, + "step": 33056 + }, + { + "epoch": 6.01, + "learning_rate": 1.1976519839825264e-06, + "loss": 0.0177, + "step": 33057 + }, + { + "epoch": 6.01, + "learning_rate": 1.1975973789588642e-06, + "loss": 5.9423, + "step": 33058 + }, + { + "epoch": 6.01, + "learning_rate": 1.197542773935202e-06, + "loss": 0.0052, + "step": 33059 + }, + { + "epoch": 6.01, + "learning_rate": 1.19748816891154e-06, + "loss": 1.1093, + "step": 33060 + }, + { + "epoch": 6.01, + "learning_rate": 1.1974335638878778e-06, + "loss": 0.0139, + "step": 33061 + }, + { + "epoch": 6.01, + "learning_rate": 1.1973789588642155e-06, + "loss": 0.2322, + "step": 33062 + }, + { + "epoch": 6.01, + "learning_rate": 1.1973243538405535e-06, + "loss": 1.1844, + "step": 33063 + }, + { + "epoch": 6.01, + "learning_rate": 1.1972697488168912e-06, + "loss": 2.1413, + "step": 33064 + }, + { + "epoch": 6.01, + "learning_rate": 1.197215143793229e-06, + "loss": 1.142, + "step": 33065 + }, + { + "epoch": 6.01, + "learning_rate": 1.1971605387695669e-06, + "loss": 0.3932, + "step": 33066 + }, + { + "epoch": 6.01, + "learning_rate": 1.1971059337459048e-06, + "loss": 0.4721, + "step": 33067 + }, + { + "epoch": 6.01, + "learning_rate": 1.1970513287222425e-06, + "loss": 0.0611, + "step": 33068 + }, + { + "epoch": 6.01, + "learning_rate": 1.1969967236985803e-06, + "loss": 0.7933, + "step": 33069 + }, + { + "epoch": 6.01, + "learning_rate": 1.196942118674918e-06, + "loss": 0.736, + "step": 33070 + }, + { + "epoch": 6.01, + "learning_rate": 1.196887513651256e-06, + "loss": 0.1601, + "step": 33071 + }, + { + "epoch": 6.01, + "learning_rate": 1.1968329086275937e-06, + "loss": 1.2398, + "step": 33072 + }, + { + "epoch": 6.01, + "learning_rate": 1.1967783036039316e-06, + "loss": 0.0033, + "step": 33073 + }, + { + "epoch": 6.01, + "learning_rate": 1.1967236985802696e-06, + "loss": 1.6852, + "step": 33074 + }, + { + "epoch": 6.01, + "learning_rate": 1.1966690935566073e-06, + "loss": 1.8457, + "step": 33075 + }, + { + "epoch": 6.01, + "learning_rate": 1.196614488532945e-06, + "loss": 0.3212, + "step": 33076 + }, + { + "epoch": 6.02, + "learning_rate": 1.1965598835092828e-06, + "loss": 0.081, + "step": 33077 + }, + { + "epoch": 6.02, + "learning_rate": 1.1965052784856207e-06, + "loss": 0.0484, + "step": 33078 + }, + { + "epoch": 6.02, + "learning_rate": 1.1964506734619584e-06, + "loss": 0.2566, + "step": 33079 + }, + { + "epoch": 6.02, + "learning_rate": 1.1963960684382964e-06, + "loss": 1.8897, + "step": 33080 + }, + { + "epoch": 6.02, + "learning_rate": 1.1963414634146343e-06, + "loss": 2.1527, + "step": 33081 + }, + { + "epoch": 6.02, + "learning_rate": 1.196286858390972e-06, + "loss": 1.2352, + "step": 33082 + }, + { + "epoch": 6.02, + "learning_rate": 1.1962322533673098e-06, + "loss": 0.8389, + "step": 33083 + }, + { + "epoch": 6.02, + "learning_rate": 1.1961776483436475e-06, + "loss": 0.7893, + "step": 33084 + }, + { + "epoch": 6.02, + "learning_rate": 1.1961230433199855e-06, + "loss": 0.0252, + "step": 33085 + }, + { + "epoch": 6.02, + "learning_rate": 1.1960684382963232e-06, + "loss": 0.0132, + "step": 33086 + }, + { + "epoch": 6.02, + "learning_rate": 1.1960138332726611e-06, + "loss": 1.4122, + "step": 33087 + }, + { + "epoch": 6.02, + "learning_rate": 1.195959228248999e-06, + "loss": 1.1117, + "step": 33088 + }, + { + "epoch": 6.02, + "learning_rate": 1.1959046232253368e-06, + "loss": 0.0073, + "step": 33089 + }, + { + "epoch": 6.02, + "learning_rate": 1.1958500182016745e-06, + "loss": 0.0481, + "step": 33090 + }, + { + "epoch": 6.02, + "learning_rate": 1.1957954131780123e-06, + "loss": 0.0071, + "step": 33091 + }, + { + "epoch": 6.02, + "learning_rate": 1.1957408081543502e-06, + "loss": 1.6525, + "step": 33092 + }, + { + "epoch": 6.02, + "learning_rate": 1.195686203130688e-06, + "loss": 0.4039, + "step": 33093 + }, + { + "epoch": 6.02, + "learning_rate": 1.1956315981070259e-06, + "loss": 2.8699, + "step": 33094 + }, + { + "epoch": 6.02, + "learning_rate": 1.1955769930833638e-06, + "loss": 1.3576, + "step": 33095 + }, + { + "epoch": 6.02, + "learning_rate": 1.1955223880597016e-06, + "loss": 0.0201, + "step": 33096 + }, + { + "epoch": 6.02, + "learning_rate": 1.1954677830360393e-06, + "loss": 0.2041, + "step": 33097 + }, + { + "epoch": 6.02, + "learning_rate": 1.195413178012377e-06, + "loss": 1.0904, + "step": 33098 + }, + { + "epoch": 6.02, + "learning_rate": 1.195358572988715e-06, + "loss": 1.3954, + "step": 33099 + }, + { + "epoch": 6.02, + "learning_rate": 1.195303967965053e-06, + "loss": 0.0199, + "step": 33100 + }, + { + "epoch": 6.02, + "learning_rate": 1.1952493629413906e-06, + "loss": 0.1346, + "step": 33101 + }, + { + "epoch": 6.02, + "learning_rate": 1.1951947579177286e-06, + "loss": 1.5998, + "step": 33102 + }, + { + "epoch": 6.02, + "learning_rate": 1.1951401528940663e-06, + "loss": 0.1904, + "step": 33103 + }, + { + "epoch": 6.02, + "learning_rate": 1.195085547870404e-06, + "loss": 0.1873, + "step": 33104 + }, + { + "epoch": 6.02, + "learning_rate": 1.1950309428467418e-06, + "loss": 2.0113, + "step": 33105 + }, + { + "epoch": 6.02, + "learning_rate": 1.1949763378230797e-06, + "loss": 0.9281, + "step": 33106 + }, + { + "epoch": 6.02, + "learning_rate": 1.1949217327994177e-06, + "loss": 0.1116, + "step": 33107 + }, + { + "epoch": 6.02, + "learning_rate": 1.1948671277757554e-06, + "loss": 0.0541, + "step": 33108 + }, + { + "epoch": 6.02, + "learning_rate": 1.1948125227520933e-06, + "loss": 0.1935, + "step": 33109 + }, + { + "epoch": 6.02, + "learning_rate": 1.194757917728431e-06, + "loss": 0.634, + "step": 33110 + }, + { + "epoch": 6.02, + "learning_rate": 1.1947033127047688e-06, + "loss": 0.6104, + "step": 33111 + }, + { + "epoch": 6.02, + "learning_rate": 1.1946487076811065e-06, + "loss": 0.0087, + "step": 33112 + }, + { + "epoch": 6.02, + "learning_rate": 1.1945941026574445e-06, + "loss": 0.1918, + "step": 33113 + }, + { + "epoch": 6.02, + "learning_rate": 1.1945394976337824e-06, + "loss": 0.3722, + "step": 33114 + }, + { + "epoch": 6.02, + "learning_rate": 1.1944848926101202e-06, + "loss": 0.0187, + "step": 33115 + }, + { + "epoch": 6.02, + "learning_rate": 1.194430287586458e-06, + "loss": 0.0495, + "step": 33116 + }, + { + "epoch": 6.02, + "learning_rate": 1.1943756825627958e-06, + "loss": 0.0569, + "step": 33117 + }, + { + "epoch": 6.02, + "learning_rate": 1.1943210775391336e-06, + "loss": 0.9364, + "step": 33118 + }, + { + "epoch": 6.02, + "learning_rate": 1.1942664725154713e-06, + "loss": 2.4704, + "step": 33119 + }, + { + "epoch": 6.02, + "learning_rate": 1.1942118674918092e-06, + "loss": 0.1369, + "step": 33120 + }, + { + "epoch": 6.02, + "learning_rate": 1.1941572624681472e-06, + "loss": 2.4165, + "step": 33121 + }, + { + "epoch": 6.02, + "learning_rate": 1.194102657444485e-06, + "loss": 0.3512, + "step": 33122 + }, + { + "epoch": 6.02, + "learning_rate": 1.1940480524208229e-06, + "loss": 0.2248, + "step": 33123 + }, + { + "epoch": 6.02, + "learning_rate": 1.1939934473971606e-06, + "loss": 0.3847, + "step": 33124 + }, + { + "epoch": 6.02, + "learning_rate": 1.1939388423734983e-06, + "loss": 0.0198, + "step": 33125 + }, + { + "epoch": 6.02, + "learning_rate": 1.193884237349836e-06, + "loss": 0.1934, + "step": 33126 + }, + { + "epoch": 6.02, + "learning_rate": 1.193829632326174e-06, + "loss": 0.6371, + "step": 33127 + }, + { + "epoch": 6.02, + "learning_rate": 1.193775027302512e-06, + "loss": 0.4672, + "step": 33128 + }, + { + "epoch": 6.02, + "learning_rate": 1.1937204222788497e-06, + "loss": 2.5006, + "step": 33129 + }, + { + "epoch": 6.02, + "learning_rate": 1.1936658172551876e-06, + "loss": 0.3867, + "step": 33130 + }, + { + "epoch": 6.02, + "learning_rate": 1.1936112122315253e-06, + "loss": 0.0118, + "step": 33131 + }, + { + "epoch": 6.03, + "learning_rate": 1.193556607207863e-06, + "loss": 0.3521, + "step": 33132 + }, + { + "epoch": 6.03, + "learning_rate": 1.1935020021842008e-06, + "loss": 1.299, + "step": 33133 + }, + { + "epoch": 6.03, + "learning_rate": 1.1934473971605388e-06, + "loss": 0.334, + "step": 33134 + }, + { + "epoch": 6.03, + "learning_rate": 1.1933927921368767e-06, + "loss": 0.0253, + "step": 33135 + }, + { + "epoch": 6.03, + "learning_rate": 1.1933381871132144e-06, + "loss": 0.1129, + "step": 33136 + }, + { + "epoch": 6.03, + "learning_rate": 1.1932835820895524e-06, + "loss": 0.0399, + "step": 33137 + }, + { + "epoch": 6.03, + "learning_rate": 1.19322897706589e-06, + "loss": 0.8569, + "step": 33138 + }, + { + "epoch": 6.03, + "learning_rate": 1.1931743720422278e-06, + "loss": 0.0603, + "step": 33139 + }, + { + "epoch": 6.03, + "learning_rate": 1.1931197670185658e-06, + "loss": 0.0361, + "step": 33140 + }, + { + "epoch": 6.03, + "learning_rate": 1.1930651619949037e-06, + "loss": 1.0664, + "step": 33141 + }, + { + "epoch": 6.03, + "learning_rate": 1.1930105569712415e-06, + "loss": 0.5188, + "step": 33142 + }, + { + "epoch": 6.03, + "learning_rate": 1.1929559519475792e-06, + "loss": 1.6231, + "step": 33143 + }, + { + "epoch": 6.03, + "learning_rate": 1.1929013469239171e-06, + "loss": 0.1424, + "step": 33144 + }, + { + "epoch": 6.03, + "learning_rate": 1.1928467419002549e-06, + "loss": 1.7302, + "step": 33145 + }, + { + "epoch": 6.03, + "learning_rate": 1.1927921368765926e-06, + "loss": 0.7677, + "step": 33146 + }, + { + "epoch": 6.03, + "learning_rate": 1.1927375318529305e-06, + "loss": 0.7584, + "step": 33147 + }, + { + "epoch": 6.03, + "learning_rate": 1.1926829268292685e-06, + "loss": 0.509, + "step": 33148 + }, + { + "epoch": 6.03, + "learning_rate": 1.1926283218056062e-06, + "loss": 0.7941, + "step": 33149 + }, + { + "epoch": 6.03, + "learning_rate": 1.192573716781944e-06, + "loss": 0.9266, + "step": 33150 + }, + { + "epoch": 6.03, + "learning_rate": 1.1925191117582817e-06, + "loss": 0.2555, + "step": 33151 + }, + { + "epoch": 6.03, + "learning_rate": 1.1924645067346196e-06, + "loss": 0.7025, + "step": 33152 + }, + { + "epoch": 6.03, + "learning_rate": 1.1924099017109573e-06, + "loss": 0.1756, + "step": 33153 + }, + { + "epoch": 6.03, + "learning_rate": 1.1923552966872953e-06, + "loss": 0.3626, + "step": 33154 + }, + { + "epoch": 6.03, + "learning_rate": 1.1923006916636332e-06, + "loss": 0.3076, + "step": 33155 + }, + { + "epoch": 6.03, + "learning_rate": 1.192246086639971e-06, + "loss": 1.0673, + "step": 33156 + }, + { + "epoch": 6.03, + "learning_rate": 1.1921914816163087e-06, + "loss": 2.0086, + "step": 33157 + }, + { + "epoch": 6.03, + "learning_rate": 1.1921368765926464e-06, + "loss": 0.8337, + "step": 33158 + }, + { + "epoch": 6.03, + "learning_rate": 1.1920822715689844e-06, + "loss": 0.3581, + "step": 33159 + }, + { + "epoch": 6.03, + "learning_rate": 1.192027666545322e-06, + "loss": 0.3545, + "step": 33160 + }, + { + "epoch": 6.03, + "learning_rate": 1.19197306152166e-06, + "loss": 0.1052, + "step": 33161 + }, + { + "epoch": 6.03, + "learning_rate": 1.191918456497998e-06, + "loss": 0.8225, + "step": 33162 + }, + { + "epoch": 6.03, + "learning_rate": 1.1918638514743357e-06, + "loss": 0.3335, + "step": 33163 + }, + { + "epoch": 6.03, + "learning_rate": 1.1918092464506735e-06, + "loss": 0.1086, + "step": 33164 + }, + { + "epoch": 6.03, + "learning_rate": 1.1917546414270112e-06, + "loss": 2.5229, + "step": 33165 + }, + { + "epoch": 6.03, + "learning_rate": 1.1917000364033491e-06, + "loss": 0.0097, + "step": 33166 + }, + { + "epoch": 6.03, + "learning_rate": 1.1916454313796869e-06, + "loss": 0.0071, + "step": 33167 + }, + { + "epoch": 6.03, + "learning_rate": 1.1915908263560248e-06, + "loss": 0.0081, + "step": 33168 + }, + { + "epoch": 6.03, + "learning_rate": 1.1915362213323627e-06, + "loss": 0.0062, + "step": 33169 + }, + { + "epoch": 6.03, + "learning_rate": 1.1914816163087005e-06, + "loss": 0.5809, + "step": 33170 + }, + { + "epoch": 6.03, + "learning_rate": 1.1914270112850382e-06, + "loss": 0.1241, + "step": 33171 + }, + { + "epoch": 6.03, + "learning_rate": 1.191372406261376e-06, + "loss": 0.4554, + "step": 33172 + }, + { + "epoch": 6.03, + "learning_rate": 1.1913178012377139e-06, + "loss": 0.0525, + "step": 33173 + }, + { + "epoch": 6.03, + "learning_rate": 1.1912631962140516e-06, + "loss": 0.7128, + "step": 33174 + }, + { + "epoch": 6.03, + "learning_rate": 1.1912085911903896e-06, + "loss": 0.5155, + "step": 33175 + }, + { + "epoch": 6.03, + "learning_rate": 1.1911539861667275e-06, + "loss": 0.3327, + "step": 33176 + }, + { + "epoch": 6.03, + "learning_rate": 1.1910993811430652e-06, + "loss": 1.3684, + "step": 33177 + }, + { + "epoch": 6.03, + "learning_rate": 1.191044776119403e-06, + "loss": 0.0924, + "step": 33178 + }, + { + "epoch": 6.03, + "learning_rate": 1.1909901710957407e-06, + "loss": 1.0381, + "step": 33179 + }, + { + "epoch": 6.03, + "learning_rate": 1.1909355660720786e-06, + "loss": 0.0872, + "step": 33180 + }, + { + "epoch": 6.03, + "learning_rate": 1.1908809610484166e-06, + "loss": 1.0368, + "step": 33181 + }, + { + "epoch": 6.03, + "learning_rate": 1.1908263560247543e-06, + "loss": 1.8732, + "step": 33182 + }, + { + "epoch": 6.03, + "learning_rate": 1.1907717510010923e-06, + "loss": 0.0133, + "step": 33183 + }, + { + "epoch": 6.03, + "learning_rate": 1.19071714597743e-06, + "loss": 1.6135, + "step": 33184 + }, + { + "epoch": 6.03, + "learning_rate": 1.1906625409537677e-06, + "loss": 0.1884, + "step": 33185 + }, + { + "epoch": 6.03, + "learning_rate": 1.1906079359301055e-06, + "loss": 1.8148, + "step": 33186 + }, + { + "epoch": 6.04, + "learning_rate": 1.1905533309064434e-06, + "loss": 0.2175, + "step": 33187 + }, + { + "epoch": 6.04, + "learning_rate": 1.1904987258827813e-06, + "loss": 0.1475, + "step": 33188 + }, + { + "epoch": 6.04, + "learning_rate": 1.190444120859119e-06, + "loss": 0.0891, + "step": 33189 + }, + { + "epoch": 6.04, + "learning_rate": 1.190389515835457e-06, + "loss": 0.0471, + "step": 33190 + }, + { + "epoch": 6.04, + "learning_rate": 1.1903349108117948e-06, + "loss": 0.2725, + "step": 33191 + }, + { + "epoch": 6.04, + "learning_rate": 1.1902803057881325e-06, + "loss": 0.2557, + "step": 33192 + }, + { + "epoch": 6.04, + "learning_rate": 1.1902257007644702e-06, + "loss": 4.2051, + "step": 33193 + }, + { + "epoch": 6.04, + "learning_rate": 1.1901710957408082e-06, + "loss": 0.8976, + "step": 33194 + }, + { + "epoch": 6.04, + "learning_rate": 1.190116490717146e-06, + "loss": 0.0152, + "step": 33195 + }, + { + "epoch": 6.04, + "learning_rate": 1.1900618856934838e-06, + "loss": 1.3693, + "step": 33196 + }, + { + "epoch": 6.04, + "learning_rate": 1.1900072806698218e-06, + "loss": 0.0436, + "step": 33197 + }, + { + "epoch": 6.04, + "learning_rate": 1.1899526756461595e-06, + "loss": 1.501, + "step": 33198 + }, + { + "epoch": 6.04, + "learning_rate": 1.1898980706224972e-06, + "loss": 0.7651, + "step": 33199 + }, + { + "epoch": 6.04, + "learning_rate": 1.189843465598835e-06, + "loss": 0.4002, + "step": 33200 + }, + { + "epoch": 6.04, + "learning_rate": 1.189788860575173e-06, + "loss": 1.8502, + "step": 33201 + }, + { + "epoch": 6.04, + "learning_rate": 1.1897342555515109e-06, + "loss": 1.9293, + "step": 33202 + }, + { + "epoch": 6.04, + "learning_rate": 1.1896796505278486e-06, + "loss": 1.517, + "step": 33203 + }, + { + "epoch": 6.04, + "learning_rate": 1.1896250455041865e-06, + "loss": 2.9241, + "step": 33204 + }, + { + "epoch": 6.04, + "learning_rate": 1.1895704404805243e-06, + "loss": 0.4391, + "step": 33205 + }, + { + "epoch": 6.04, + "learning_rate": 1.189515835456862e-06, + "loss": 0.2354, + "step": 33206 + }, + { + "epoch": 6.04, + "learning_rate": 1.1894612304331997e-06, + "loss": 0.1909, + "step": 33207 + }, + { + "epoch": 6.04, + "learning_rate": 1.1894066254095377e-06, + "loss": 0.764, + "step": 33208 + }, + { + "epoch": 6.04, + "learning_rate": 1.1893520203858756e-06, + "loss": 0.7483, + "step": 33209 + }, + { + "epoch": 6.04, + "learning_rate": 1.1892974153622133e-06, + "loss": 0.2465, + "step": 33210 + }, + { + "epoch": 6.04, + "learning_rate": 1.1892428103385513e-06, + "loss": 0.1099, + "step": 33211 + }, + { + "epoch": 6.04, + "learning_rate": 1.189188205314889e-06, + "loss": 0.0066, + "step": 33212 + }, + { + "epoch": 6.04, + "learning_rate": 1.1891336002912268e-06, + "loss": 0.3791, + "step": 33213 + }, + { + "epoch": 6.04, + "learning_rate": 1.1890789952675645e-06, + "loss": 0.3104, + "step": 33214 + }, + { + "epoch": 6.04, + "learning_rate": 1.1890243902439024e-06, + "loss": 2.4643, + "step": 33215 + }, + { + "epoch": 6.04, + "learning_rate": 1.1889697852202404e-06, + "loss": 0.0441, + "step": 33216 + }, + { + "epoch": 6.04, + "learning_rate": 1.188915180196578e-06, + "loss": 0.6895, + "step": 33217 + }, + { + "epoch": 6.04, + "learning_rate": 1.188860575172916e-06, + "loss": 0.1314, + "step": 33218 + }, + { + "epoch": 6.04, + "learning_rate": 1.1888059701492538e-06, + "loss": 0.0171, + "step": 33219 + }, + { + "epoch": 6.04, + "learning_rate": 1.1887513651255915e-06, + "loss": 0.0643, + "step": 33220 + }, + { + "epoch": 6.04, + "learning_rate": 1.1886967601019295e-06, + "loss": 1.874, + "step": 33221 + }, + { + "epoch": 6.04, + "learning_rate": 1.1886421550782672e-06, + "loss": 0.5706, + "step": 33222 + }, + { + "epoch": 6.04, + "learning_rate": 1.1885875500546051e-06, + "loss": 0.7267, + "step": 33223 + }, + { + "epoch": 6.04, + "learning_rate": 1.1885329450309429e-06, + "loss": 0.0431, + "step": 33224 + }, + { + "epoch": 6.04, + "learning_rate": 1.1884783400072806e-06, + "loss": 0.6682, + "step": 33225 + }, + { + "epoch": 6.04, + "learning_rate": 1.1884237349836185e-06, + "loss": 0.1331, + "step": 33226 + }, + { + "epoch": 6.04, + "learning_rate": 1.1883691299599563e-06, + "loss": 0.0882, + "step": 33227 + }, + { + "epoch": 6.04, + "learning_rate": 1.1883145249362942e-06, + "loss": 0.0132, + "step": 33228 + }, + { + "epoch": 6.04, + "learning_rate": 1.1882599199126322e-06, + "loss": 0.281, + "step": 33229 + }, + { + "epoch": 6.04, + "learning_rate": 1.1882053148889699e-06, + "loss": 1.1759, + "step": 33230 + }, + { + "epoch": 6.04, + "learning_rate": 1.1881507098653076e-06, + "loss": 0.26, + "step": 33231 + }, + { + "epoch": 6.04, + "learning_rate": 1.1880961048416453e-06, + "loss": 0.1208, + "step": 33232 + }, + { + "epoch": 6.04, + "learning_rate": 1.1880414998179833e-06, + "loss": 1.5448, + "step": 33233 + }, + { + "epoch": 6.04, + "learning_rate": 1.187986894794321e-06, + "loss": 0.1334, + "step": 33234 + }, + { + "epoch": 6.04, + "learning_rate": 1.187932289770659e-06, + "loss": 2.0813, + "step": 33235 + }, + { + "epoch": 6.04, + "learning_rate": 1.187877684746997e-06, + "loss": 0.126, + "step": 33236 + }, + { + "epoch": 6.04, + "learning_rate": 1.1878230797233346e-06, + "loss": 0.973, + "step": 33237 + }, + { + "epoch": 6.04, + "learning_rate": 1.1877684746996724e-06, + "loss": 2.727, + "step": 33238 + }, + { + "epoch": 6.04, + "learning_rate": 1.18771386967601e-06, + "loss": 0.9099, + "step": 33239 + }, + { + "epoch": 6.04, + "learning_rate": 1.187659264652348e-06, + "loss": 0.0046, + "step": 33240 + }, + { + "epoch": 6.04, + "learning_rate": 1.1876046596286858e-06, + "loss": 0.2524, + "step": 33241 + }, + { + "epoch": 6.05, + "learning_rate": 1.1875500546050237e-06, + "loss": 0.0153, + "step": 33242 + }, + { + "epoch": 6.05, + "learning_rate": 1.1874954495813617e-06, + "loss": 1.3362, + "step": 33243 + }, + { + "epoch": 6.05, + "learning_rate": 1.1874408445576994e-06, + "loss": 0.56, + "step": 33244 + }, + { + "epoch": 6.05, + "learning_rate": 1.1873862395340371e-06, + "loss": 0.0376, + "step": 33245 + }, + { + "epoch": 6.05, + "learning_rate": 1.1873316345103749e-06, + "loss": 1.5053, + "step": 33246 + }, + { + "epoch": 6.05, + "learning_rate": 1.1872770294867128e-06, + "loss": 0.2053, + "step": 33247 + }, + { + "epoch": 6.05, + "learning_rate": 1.1872224244630505e-06, + "loss": 0.0804, + "step": 33248 + }, + { + "epoch": 6.05, + "learning_rate": 1.1871678194393885e-06, + "loss": 0.8371, + "step": 33249 + }, + { + "epoch": 6.05, + "learning_rate": 1.1871132144157264e-06, + "loss": 0.3554, + "step": 33250 + }, + { + "epoch": 6.05, + "learning_rate": 1.1870586093920642e-06, + "loss": 0.2992, + "step": 33251 + }, + { + "epoch": 6.05, + "learning_rate": 1.1870040043684019e-06, + "loss": 0.5428, + "step": 33252 + }, + { + "epoch": 6.05, + "learning_rate": 1.1869493993447396e-06, + "loss": 0.7323, + "step": 33253 + }, + { + "epoch": 6.05, + "learning_rate": 1.1868947943210776e-06, + "loss": 0.5087, + "step": 33254 + }, + { + "epoch": 6.05, + "learning_rate": 1.1868401892974153e-06, + "loss": 1.05, + "step": 33255 + }, + { + "epoch": 6.05, + "learning_rate": 1.1867855842737532e-06, + "loss": 0.3843, + "step": 33256 + }, + { + "epoch": 6.05, + "learning_rate": 1.1867309792500912e-06, + "loss": 0.6694, + "step": 33257 + }, + { + "epoch": 6.05, + "learning_rate": 1.186676374226429e-06, + "loss": 0.4157, + "step": 33258 + }, + { + "epoch": 6.05, + "learning_rate": 1.1866217692027666e-06, + "loss": 0.3499, + "step": 33259 + }, + { + "epoch": 6.05, + "learning_rate": 1.1865671641791044e-06, + "loss": 0.0278, + "step": 33260 + }, + { + "epoch": 6.05, + "learning_rate": 1.1865125591554423e-06, + "loss": 0.3078, + "step": 33261 + }, + { + "epoch": 6.05, + "learning_rate": 1.18645795413178e-06, + "loss": 0.6739, + "step": 33262 + }, + { + "epoch": 6.05, + "learning_rate": 1.186403349108118e-06, + "loss": 0.6455, + "step": 33263 + }, + { + "epoch": 6.05, + "learning_rate": 1.186348744084456e-06, + "loss": 1.3654, + "step": 33264 + }, + { + "epoch": 6.05, + "learning_rate": 1.1862941390607937e-06, + "loss": 0.0067, + "step": 33265 + }, + { + "epoch": 6.05, + "learning_rate": 1.1862395340371314e-06, + "loss": 0.6205, + "step": 33266 + }, + { + "epoch": 6.05, + "learning_rate": 1.1861849290134691e-06, + "loss": 0.5664, + "step": 33267 + }, + { + "epoch": 6.05, + "learning_rate": 1.186130323989807e-06, + "loss": 0.6918, + "step": 33268 + }, + { + "epoch": 6.05, + "learning_rate": 1.186075718966145e-06, + "loss": 0.5278, + "step": 33269 + }, + { + "epoch": 6.05, + "learning_rate": 1.1860211139424828e-06, + "loss": 1.0721, + "step": 33270 + }, + { + "epoch": 6.05, + "learning_rate": 1.1859665089188207e-06, + "loss": 0.0704, + "step": 33271 + }, + { + "epoch": 6.05, + "learning_rate": 1.1859119038951584e-06, + "loss": 1.7127, + "step": 33272 + }, + { + "epoch": 6.05, + "learning_rate": 1.1858572988714962e-06, + "loss": 1.0196, + "step": 33273 + }, + { + "epoch": 6.05, + "learning_rate": 1.1858026938478339e-06, + "loss": 0.1168, + "step": 33274 + }, + { + "epoch": 6.05, + "learning_rate": 1.1857480888241718e-06, + "loss": 0.2393, + "step": 33275 + }, + { + "epoch": 6.05, + "learning_rate": 1.1856934838005098e-06, + "loss": 0.186, + "step": 33276 + }, + { + "epoch": 6.05, + "learning_rate": 1.1856388787768475e-06, + "loss": 0.3704, + "step": 33277 + }, + { + "epoch": 6.05, + "learning_rate": 1.1855842737531855e-06, + "loss": 0.4952, + "step": 33278 + }, + { + "epoch": 6.05, + "learning_rate": 1.1855296687295232e-06, + "loss": 0.37, + "step": 33279 + }, + { + "epoch": 6.05, + "learning_rate": 1.185475063705861e-06, + "loss": 0.5235, + "step": 33280 + }, + { + "epoch": 6.05, + "learning_rate": 1.1854204586821986e-06, + "loss": 0.1957, + "step": 33281 + }, + { + "epoch": 6.05, + "learning_rate": 1.1853658536585366e-06, + "loss": 0.8172, + "step": 33282 + }, + { + "epoch": 6.05, + "learning_rate": 1.1853112486348745e-06, + "loss": 0.0297, + "step": 33283 + }, + { + "epoch": 6.05, + "learning_rate": 1.1852566436112123e-06, + "loss": 0.0026, + "step": 33284 + }, + { + "epoch": 6.05, + "learning_rate": 1.1852020385875502e-06, + "loss": 0.0074, + "step": 33285 + }, + { + "epoch": 6.05, + "learning_rate": 1.185147433563888e-06, + "loss": 1.674, + "step": 33286 + }, + { + "epoch": 6.05, + "learning_rate": 1.1850928285402257e-06, + "loss": 0.0854, + "step": 33287 + }, + { + "epoch": 6.05, + "learning_rate": 1.1850382235165634e-06, + "loss": 0.0577, + "step": 33288 + }, + { + "epoch": 6.05, + "learning_rate": 1.1849836184929013e-06, + "loss": 0.5068, + "step": 33289 + }, + { + "epoch": 6.05, + "learning_rate": 1.1849290134692393e-06, + "loss": 0.0093, + "step": 33290 + }, + { + "epoch": 6.05, + "learning_rate": 1.184874408445577e-06, + "loss": 0.4489, + "step": 33291 + }, + { + "epoch": 6.05, + "learning_rate": 1.184819803421915e-06, + "loss": 0.0831, + "step": 33292 + }, + { + "epoch": 6.05, + "learning_rate": 1.1847651983982527e-06, + "loss": 0.5057, + "step": 33293 + }, + { + "epoch": 6.05, + "learning_rate": 1.1847105933745904e-06, + "loss": 1.1534, + "step": 33294 + }, + { + "epoch": 6.05, + "learning_rate": 1.1846559883509282e-06, + "loss": 1.5433, + "step": 33295 + }, + { + "epoch": 6.05, + "learning_rate": 1.184601383327266e-06, + "loss": 0.3061, + "step": 33296 + }, + { + "epoch": 6.06, + "learning_rate": 1.184546778303604e-06, + "loss": 0.1205, + "step": 33297 + }, + { + "epoch": 6.06, + "learning_rate": 1.1844921732799418e-06, + "loss": 0.9342, + "step": 33298 + }, + { + "epoch": 6.06, + "learning_rate": 1.1844375682562797e-06, + "loss": 0.1579, + "step": 33299 + }, + { + "epoch": 6.06, + "learning_rate": 1.1843829632326175e-06, + "loss": 0.4021, + "step": 33300 + }, + { + "epoch": 6.06, + "learning_rate": 1.1843283582089552e-06, + "loss": 3.0782, + "step": 33301 + }, + { + "epoch": 6.06, + "learning_rate": 1.184273753185293e-06, + "loss": 0.1649, + "step": 33302 + }, + { + "epoch": 6.06, + "learning_rate": 1.1842191481616309e-06, + "loss": 0.0285, + "step": 33303 + }, + { + "epoch": 6.06, + "learning_rate": 1.1841645431379688e-06, + "loss": 1.2057, + "step": 33304 + }, + { + "epoch": 6.06, + "learning_rate": 1.1841099381143065e-06, + "loss": 0.6577, + "step": 33305 + }, + { + "epoch": 6.06, + "learning_rate": 1.1840553330906443e-06, + "loss": 0.6924, + "step": 33306 + }, + { + "epoch": 6.06, + "learning_rate": 1.1840007280669822e-06, + "loss": 0.4139, + "step": 33307 + }, + { + "epoch": 6.06, + "learning_rate": 1.18394612304332e-06, + "loss": 0.1334, + "step": 33308 + }, + { + "epoch": 6.06, + "learning_rate": 1.1838915180196579e-06, + "loss": 1.5419, + "step": 33309 + }, + { + "epoch": 6.06, + "learning_rate": 1.1838369129959958e-06, + "loss": 1.6162, + "step": 33310 + }, + { + "epoch": 6.06, + "learning_rate": 1.1837823079723336e-06, + "loss": 0.0307, + "step": 33311 + }, + { + "epoch": 6.06, + "learning_rate": 1.1837277029486713e-06, + "loss": 0.2142, + "step": 33312 + }, + { + "epoch": 6.06, + "learning_rate": 1.183673097925009e-06, + "loss": 0.0271, + "step": 33313 + }, + { + "epoch": 6.06, + "learning_rate": 1.183618492901347e-06, + "loss": 0.3228, + "step": 33314 + }, + { + "epoch": 6.06, + "learning_rate": 1.1835638878776847e-06, + "loss": 0.0109, + "step": 33315 + }, + { + "epoch": 6.06, + "learning_rate": 1.1835092828540226e-06, + "loss": 2.8467, + "step": 33316 + }, + { + "epoch": 6.06, + "learning_rate": 1.1834546778303606e-06, + "loss": 0.2374, + "step": 33317 + }, + { + "epoch": 6.06, + "learning_rate": 1.1834000728066983e-06, + "loss": 1.9696, + "step": 33318 + }, + { + "epoch": 6.06, + "learning_rate": 1.183345467783036e-06, + "loss": 0.5018, + "step": 33319 + }, + { + "epoch": 6.06, + "learning_rate": 1.1832908627593738e-06, + "loss": 0.1985, + "step": 33320 + }, + { + "epoch": 6.06, + "learning_rate": 1.1832362577357117e-06, + "loss": 0.0067, + "step": 33321 + }, + { + "epoch": 6.06, + "learning_rate": 1.1831816527120495e-06, + "loss": 0.1137, + "step": 33322 + }, + { + "epoch": 6.06, + "learning_rate": 1.1831270476883874e-06, + "loss": 0.0917, + "step": 33323 + }, + { + "epoch": 6.06, + "learning_rate": 1.1830724426647253e-06, + "loss": 0.2254, + "step": 33324 + }, + { + "epoch": 6.06, + "learning_rate": 1.183017837641063e-06, + "loss": 0.9053, + "step": 33325 + }, + { + "epoch": 6.06, + "learning_rate": 1.1829632326174008e-06, + "loss": 0.3601, + "step": 33326 + }, + { + "epoch": 6.06, + "learning_rate": 1.1829086275937385e-06, + "loss": 2.5128, + "step": 33327 + }, + { + "epoch": 6.06, + "learning_rate": 1.1828540225700765e-06, + "loss": 0.0072, + "step": 33328 + }, + { + "epoch": 6.06, + "learning_rate": 1.1827994175464142e-06, + "loss": 0.0555, + "step": 33329 + }, + { + "epoch": 6.06, + "learning_rate": 1.1827448125227522e-06, + "loss": 0.0506, + "step": 33330 + }, + { + "epoch": 6.06, + "learning_rate": 1.18269020749909e-06, + "loss": 1.3367, + "step": 33331 + }, + { + "epoch": 6.06, + "learning_rate": 1.1826356024754278e-06, + "loss": 1.8702, + "step": 33332 + }, + { + "epoch": 6.06, + "learning_rate": 1.1825809974517656e-06, + "loss": 0.9049, + "step": 33333 + }, + { + "epoch": 6.06, + "learning_rate": 1.1825263924281033e-06, + "loss": 0.0134, + "step": 33334 + }, + { + "epoch": 6.06, + "learning_rate": 1.1824717874044412e-06, + "loss": 0.1665, + "step": 33335 + }, + { + "epoch": 6.06, + "learning_rate": 1.182417182380779e-06, + "loss": 0.5587, + "step": 33336 + }, + { + "epoch": 6.06, + "learning_rate": 1.182362577357117e-06, + "loss": 0.7441, + "step": 33337 + }, + { + "epoch": 6.06, + "learning_rate": 1.1823079723334549e-06, + "loss": 0.0155, + "step": 33338 + }, + { + "epoch": 6.06, + "learning_rate": 1.1822533673097926e-06, + "loss": 0.1679, + "step": 33339 + }, + { + "epoch": 6.06, + "learning_rate": 1.1821987622861303e-06, + "loss": 0.0068, + "step": 33340 + }, + { + "epoch": 6.06, + "learning_rate": 1.182144157262468e-06, + "loss": 0.4384, + "step": 33341 + }, + { + "epoch": 6.06, + "learning_rate": 1.182089552238806e-06, + "loss": 1.0538, + "step": 33342 + }, + { + "epoch": 6.06, + "learning_rate": 1.1820349472151437e-06, + "loss": 1.9227, + "step": 33343 + }, + { + "epoch": 6.06, + "learning_rate": 1.1819803421914817e-06, + "loss": 0.2172, + "step": 33344 + }, + { + "epoch": 6.06, + "learning_rate": 1.1819257371678196e-06, + "loss": 0.6936, + "step": 33345 + }, + { + "epoch": 6.06, + "learning_rate": 1.1818711321441573e-06, + "loss": 0.1211, + "step": 33346 + }, + { + "epoch": 6.06, + "learning_rate": 1.181816527120495e-06, + "loss": 1.2299, + "step": 33347 + }, + { + "epoch": 6.06, + "learning_rate": 1.1817619220968328e-06, + "loss": 0.6595, + "step": 33348 + }, + { + "epoch": 6.06, + "learning_rate": 1.1817073170731707e-06, + "loss": 1.1251, + "step": 33349 + }, + { + "epoch": 6.06, + "learning_rate": 1.1816527120495087e-06, + "loss": 1.0152, + "step": 33350 + }, + { + "epoch": 6.06, + "learning_rate": 1.1815981070258464e-06, + "loss": 0.4487, + "step": 33351 + }, + { + "epoch": 6.07, + "learning_rate": 1.1815435020021844e-06, + "loss": 0.0164, + "step": 33352 + }, + { + "epoch": 6.07, + "learning_rate": 1.181488896978522e-06, + "loss": 1.3935, + "step": 33353 + }, + { + "epoch": 6.07, + "learning_rate": 1.1814342919548598e-06, + "loss": 0.0861, + "step": 33354 + }, + { + "epoch": 6.07, + "learning_rate": 1.1813796869311976e-06, + "loss": 0.0411, + "step": 33355 + }, + { + "epoch": 6.07, + "learning_rate": 1.1813250819075355e-06, + "loss": 1.1821, + "step": 33356 + }, + { + "epoch": 6.07, + "learning_rate": 1.1812704768838734e-06, + "loss": 1.0463, + "step": 33357 + }, + { + "epoch": 6.07, + "learning_rate": 1.1812158718602112e-06, + "loss": 0.0841, + "step": 33358 + }, + { + "epoch": 6.07, + "learning_rate": 1.1811612668365491e-06, + "loss": 0.4166, + "step": 33359 + }, + { + "epoch": 6.07, + "learning_rate": 1.1811066618128869e-06, + "loss": 0.2287, + "step": 33360 + }, + { + "epoch": 6.07, + "learning_rate": 1.1810520567892246e-06, + "loss": 0.1564, + "step": 33361 + }, + { + "epoch": 6.07, + "learning_rate": 1.1809974517655623e-06, + "loss": 2.049, + "step": 33362 + }, + { + "epoch": 6.07, + "learning_rate": 1.1809428467419003e-06, + "loss": 0.2443, + "step": 33363 + }, + { + "epoch": 6.07, + "learning_rate": 1.1808882417182382e-06, + "loss": 0.1249, + "step": 33364 + }, + { + "epoch": 6.07, + "learning_rate": 1.180833636694576e-06, + "loss": 0.1759, + "step": 33365 + }, + { + "epoch": 6.07, + "learning_rate": 1.1807790316709139e-06, + "loss": 1.3467, + "step": 33366 + }, + { + "epoch": 6.07, + "learning_rate": 1.1807244266472516e-06, + "loss": 0.0221, + "step": 33367 + }, + { + "epoch": 6.07, + "learning_rate": 1.1806698216235893e-06, + "loss": 1.6703, + "step": 33368 + }, + { + "epoch": 6.07, + "learning_rate": 1.180615216599927e-06, + "loss": 0.5866, + "step": 33369 + }, + { + "epoch": 6.07, + "learning_rate": 1.180560611576265e-06, + "loss": 0.3106, + "step": 33370 + }, + { + "epoch": 6.07, + "learning_rate": 1.180506006552603e-06, + "loss": 2.5425, + "step": 33371 + }, + { + "epoch": 6.07, + "learning_rate": 1.1804514015289407e-06, + "loss": 0.3464, + "step": 33372 + }, + { + "epoch": 6.07, + "learning_rate": 1.1803967965052786e-06, + "loss": 2.3407, + "step": 33373 + }, + { + "epoch": 6.07, + "learning_rate": 1.1803421914816164e-06, + "loss": 0.1736, + "step": 33374 + }, + { + "epoch": 6.07, + "learning_rate": 1.180287586457954e-06, + "loss": 0.7126, + "step": 33375 + }, + { + "epoch": 6.07, + "learning_rate": 1.1802329814342918e-06, + "loss": 0.0541, + "step": 33376 + }, + { + "epoch": 6.07, + "learning_rate": 1.1801783764106298e-06, + "loss": 0.2411, + "step": 33377 + }, + { + "epoch": 6.07, + "learning_rate": 1.1801237713869677e-06, + "loss": 1.7886, + "step": 33378 + }, + { + "epoch": 6.07, + "learning_rate": 1.1800691663633055e-06, + "loss": 0.0506, + "step": 33379 + }, + { + "epoch": 6.07, + "learning_rate": 1.1800145613396432e-06, + "loss": 1.214, + "step": 33380 + }, + { + "epoch": 6.07, + "learning_rate": 1.1799599563159811e-06, + "loss": 1.0248, + "step": 33381 + }, + { + "epoch": 6.07, + "learning_rate": 1.1799053512923189e-06, + "loss": 0.2587, + "step": 33382 + }, + { + "epoch": 6.07, + "learning_rate": 1.1798507462686566e-06, + "loss": 0.1028, + "step": 33383 + }, + { + "epoch": 6.07, + "learning_rate": 1.1797961412449945e-06, + "loss": 0.1142, + "step": 33384 + }, + { + "epoch": 6.07, + "learning_rate": 1.1797415362213325e-06, + "loss": 0.2296, + "step": 33385 + }, + { + "epoch": 6.07, + "learning_rate": 1.1796869311976702e-06, + "loss": 0.2495, + "step": 33386 + }, + { + "epoch": 6.07, + "learning_rate": 1.179632326174008e-06, + "loss": 0.0394, + "step": 33387 + }, + { + "epoch": 6.07, + "learning_rate": 1.1795777211503459e-06, + "loss": 1.5989, + "step": 33388 + }, + { + "epoch": 6.07, + "learning_rate": 1.1795231161266836e-06, + "loss": 0.0049, + "step": 33389 + }, + { + "epoch": 6.07, + "learning_rate": 1.1794685111030213e-06, + "loss": 0.0184, + "step": 33390 + }, + { + "epoch": 6.07, + "learning_rate": 1.1794139060793593e-06, + "loss": 0.7576, + "step": 33391 + }, + { + "epoch": 6.07, + "learning_rate": 1.1793593010556972e-06, + "loss": 0.3932, + "step": 33392 + }, + { + "epoch": 6.07, + "learning_rate": 1.179304696032035e-06, + "loss": 0.6505, + "step": 33393 + }, + { + "epoch": 6.07, + "learning_rate": 1.1792500910083727e-06, + "loss": 0.0647, + "step": 33394 + }, + { + "epoch": 6.07, + "learning_rate": 1.1791954859847106e-06, + "loss": 0.7451, + "step": 33395 + }, + { + "epoch": 6.07, + "learning_rate": 1.1791408809610484e-06, + "loss": 0.831, + "step": 33396 + }, + { + "epoch": 6.07, + "learning_rate": 1.1790862759373863e-06, + "loss": 0.4792, + "step": 33397 + }, + { + "epoch": 6.07, + "learning_rate": 1.1790316709137243e-06, + "loss": 2.2473, + "step": 33398 + }, + { + "epoch": 6.07, + "learning_rate": 1.178977065890062e-06, + "loss": 0.058, + "step": 33399 + }, + { + "epoch": 6.07, + "learning_rate": 1.1789224608663997e-06, + "loss": 0.0114, + "step": 33400 + }, + { + "epoch": 6.07, + "learning_rate": 1.1788678558427375e-06, + "loss": 0.5563, + "step": 33401 + }, + { + "epoch": 6.07, + "learning_rate": 1.1788132508190754e-06, + "loss": 0.0091, + "step": 33402 + }, + { + "epoch": 6.07, + "learning_rate": 1.1787586457954131e-06, + "loss": 0.0144, + "step": 33403 + }, + { + "epoch": 6.07, + "learning_rate": 1.178704040771751e-06, + "loss": 0.806, + "step": 33404 + }, + { + "epoch": 6.07, + "learning_rate": 1.178649435748089e-06, + "loss": 0.0213, + "step": 33405 + }, + { + "epoch": 6.07, + "learning_rate": 1.1785948307244267e-06, + "loss": 0.3657, + "step": 33406 + }, + { + "epoch": 6.08, + "learning_rate": 1.1785402257007645e-06, + "loss": 0.0583, + "step": 33407 + }, + { + "epoch": 6.08, + "learning_rate": 1.1784856206771022e-06, + "loss": 0.7705, + "step": 33408 + }, + { + "epoch": 6.08, + "learning_rate": 1.1784310156534402e-06, + "loss": 0.0257, + "step": 33409 + }, + { + "epoch": 6.08, + "learning_rate": 1.1783764106297779e-06, + "loss": 0.1833, + "step": 33410 + }, + { + "epoch": 6.08, + "learning_rate": 1.1783218056061158e-06, + "loss": 1.4155, + "step": 33411 + }, + { + "epoch": 6.08, + "learning_rate": 1.1782672005824538e-06, + "loss": 0.1812, + "step": 33412 + }, + { + "epoch": 6.08, + "learning_rate": 1.1782125955587915e-06, + "loss": 0.1336, + "step": 33413 + }, + { + "epoch": 6.08, + "learning_rate": 1.1781579905351292e-06, + "loss": 0.9905, + "step": 33414 + }, + { + "epoch": 6.08, + "learning_rate": 1.178103385511467e-06, + "loss": 0.1738, + "step": 33415 + }, + { + "epoch": 6.08, + "learning_rate": 1.178048780487805e-06, + "loss": 0.0393, + "step": 33416 + }, + { + "epoch": 6.08, + "learning_rate": 1.1779941754641426e-06, + "loss": 0.8399, + "step": 33417 + }, + { + "epoch": 6.08, + "learning_rate": 1.1779395704404806e-06, + "loss": 0.7918, + "step": 33418 + }, + { + "epoch": 6.08, + "learning_rate": 1.1778849654168185e-06, + "loss": 0.3656, + "step": 33419 + }, + { + "epoch": 6.08, + "learning_rate": 1.1778303603931563e-06, + "loss": 0.0271, + "step": 33420 + }, + { + "epoch": 6.08, + "learning_rate": 1.177775755369494e-06, + "loss": 0.826, + "step": 33421 + }, + { + "epoch": 6.08, + "learning_rate": 1.1777211503458317e-06, + "loss": 1.3931, + "step": 33422 + }, + { + "epoch": 6.08, + "learning_rate": 1.1776665453221697e-06, + "loss": 0.6443, + "step": 33423 + }, + { + "epoch": 6.08, + "learning_rate": 1.1776119402985074e-06, + "loss": 0.8695, + "step": 33424 + }, + { + "epoch": 6.08, + "learning_rate": 1.1775573352748453e-06, + "loss": 0.0068, + "step": 33425 + }, + { + "epoch": 6.08, + "learning_rate": 1.1775027302511833e-06, + "loss": 0.0234, + "step": 33426 + }, + { + "epoch": 6.08, + "learning_rate": 1.177448125227521e-06, + "loss": 0.0073, + "step": 33427 + }, + { + "epoch": 6.08, + "learning_rate": 1.1773935202038587e-06, + "loss": 0.5563, + "step": 33428 + }, + { + "epoch": 6.08, + "learning_rate": 1.1773389151801965e-06, + "loss": 0.4821, + "step": 33429 + }, + { + "epoch": 6.08, + "learning_rate": 1.1772843101565344e-06, + "loss": 0.4131, + "step": 33430 + }, + { + "epoch": 6.08, + "learning_rate": 1.1772297051328722e-06, + "loss": 0.1387, + "step": 33431 + }, + { + "epoch": 6.08, + "learning_rate": 1.17717510010921e-06, + "loss": 4.6246, + "step": 33432 + }, + { + "epoch": 6.08, + "learning_rate": 1.177120495085548e-06, + "loss": 0.1351, + "step": 33433 + }, + { + "epoch": 6.08, + "learning_rate": 1.1770658900618858e-06, + "loss": 0.5787, + "step": 33434 + }, + { + "epoch": 6.08, + "learning_rate": 1.1770112850382235e-06, + "loss": 4.3297, + "step": 33435 + }, + { + "epoch": 6.08, + "learning_rate": 1.1769566800145612e-06, + "loss": 2.0115, + "step": 33436 + }, + { + "epoch": 6.08, + "learning_rate": 1.1769020749908992e-06, + "loss": 0.273, + "step": 33437 + }, + { + "epoch": 6.08, + "learning_rate": 1.1768474699672371e-06, + "loss": 0.0659, + "step": 33438 + }, + { + "epoch": 6.08, + "learning_rate": 1.1767928649435749e-06, + "loss": 0.1864, + "step": 33439 + }, + { + "epoch": 6.08, + "learning_rate": 1.1767382599199128e-06, + "loss": 0.0316, + "step": 33440 + }, + { + "epoch": 6.08, + "learning_rate": 1.1766836548962505e-06, + "loss": 1.0372, + "step": 33441 + }, + { + "epoch": 6.08, + "learning_rate": 1.1766290498725883e-06, + "loss": 1.2854, + "step": 33442 + }, + { + "epoch": 6.08, + "learning_rate": 1.176574444848926e-06, + "loss": 1.0486, + "step": 33443 + }, + { + "epoch": 6.08, + "learning_rate": 1.176519839825264e-06, + "loss": 0.2775, + "step": 33444 + }, + { + "epoch": 6.08, + "learning_rate": 1.1764652348016019e-06, + "loss": 1.9062, + "step": 33445 + }, + { + "epoch": 6.08, + "learning_rate": 1.1764106297779396e-06, + "loss": 0.1078, + "step": 33446 + }, + { + "epoch": 6.08, + "learning_rate": 1.1763560247542776e-06, + "loss": 0.6902, + "step": 33447 + }, + { + "epoch": 6.08, + "learning_rate": 1.1763014197306153e-06, + "loss": 0.0061, + "step": 33448 + }, + { + "epoch": 6.08, + "learning_rate": 1.176246814706953e-06, + "loss": 0.013, + "step": 33449 + }, + { + "epoch": 6.08, + "learning_rate": 1.1761922096832908e-06, + "loss": 0.0065, + "step": 33450 + }, + { + "epoch": 6.08, + "learning_rate": 1.1761376046596287e-06, + "loss": 0.97, + "step": 33451 + }, + { + "epoch": 6.08, + "learning_rate": 1.1760829996359666e-06, + "loss": 0.0099, + "step": 33452 + }, + { + "epoch": 6.08, + "learning_rate": 1.1760283946123044e-06, + "loss": 0.0848, + "step": 33453 + }, + { + "epoch": 6.08, + "learning_rate": 1.1759737895886423e-06, + "loss": 0.7311, + "step": 33454 + }, + { + "epoch": 6.08, + "learning_rate": 1.17591918456498e-06, + "loss": 0.0616, + "step": 33455 + }, + { + "epoch": 6.08, + "learning_rate": 1.1758645795413178e-06, + "loss": 0.9013, + "step": 33456 + }, + { + "epoch": 6.08, + "learning_rate": 1.1758099745176555e-06, + "loss": 0.1422, + "step": 33457 + }, + { + "epoch": 6.08, + "learning_rate": 1.1757553694939935e-06, + "loss": 0.6611, + "step": 33458 + }, + { + "epoch": 6.08, + "learning_rate": 1.1757007644703314e-06, + "loss": 0.0168, + "step": 33459 + }, + { + "epoch": 6.08, + "learning_rate": 1.1756461594466691e-06, + "loss": 0.0072, + "step": 33460 + }, + { + "epoch": 6.08, + "learning_rate": 1.1755915544230069e-06, + "loss": 1.9069, + "step": 33461 + }, + { + "epoch": 6.09, + "learning_rate": 1.1755369493993448e-06, + "loss": 0.323, + "step": 33462 + }, + { + "epoch": 6.09, + "learning_rate": 1.1754823443756825e-06, + "loss": 0.0122, + "step": 33463 + }, + { + "epoch": 6.09, + "learning_rate": 1.1754277393520203e-06, + "loss": 0.7204, + "step": 33464 + }, + { + "epoch": 6.09, + "learning_rate": 1.1753731343283582e-06, + "loss": 1.2058, + "step": 33465 + }, + { + "epoch": 6.09, + "learning_rate": 1.1753185293046962e-06, + "loss": 1.5172, + "step": 33466 + }, + { + "epoch": 6.09, + "learning_rate": 1.1752639242810339e-06, + "loss": 0.4111, + "step": 33467 + }, + { + "epoch": 6.09, + "learning_rate": 1.1752093192573716e-06, + "loss": 0.0205, + "step": 33468 + }, + { + "epoch": 6.09, + "learning_rate": 1.1751547142337096e-06, + "loss": 0.1335, + "step": 33469 + }, + { + "epoch": 6.09, + "learning_rate": 1.1751001092100473e-06, + "loss": 1.1059, + "step": 33470 + }, + { + "epoch": 6.09, + "learning_rate": 1.175045504186385e-06, + "loss": 0.0773, + "step": 33471 + }, + { + "epoch": 6.09, + "learning_rate": 1.174990899162723e-06, + "loss": 0.079, + "step": 33472 + }, + { + "epoch": 6.09, + "learning_rate": 1.174936294139061e-06, + "loss": 0.1054, + "step": 33473 + }, + { + "epoch": 6.09, + "learning_rate": 1.1748816891153986e-06, + "loss": 0.0334, + "step": 33474 + }, + { + "epoch": 6.09, + "learning_rate": 1.1748270840917364e-06, + "loss": 1.0507, + "step": 33475 + }, + { + "epoch": 6.09, + "learning_rate": 1.1747724790680743e-06, + "loss": 0.1429, + "step": 33476 + }, + { + "epoch": 6.09, + "learning_rate": 1.174717874044412e-06, + "loss": 0.846, + "step": 33477 + }, + { + "epoch": 6.09, + "learning_rate": 1.17466326902075e-06, + "loss": 0.1246, + "step": 33478 + }, + { + "epoch": 6.09, + "learning_rate": 1.174608663997088e-06, + "loss": 0.0947, + "step": 33479 + }, + { + "epoch": 6.09, + "learning_rate": 1.1745540589734257e-06, + "loss": 0.2985, + "step": 33480 + }, + { + "epoch": 6.09, + "learning_rate": 1.1744994539497634e-06, + "loss": 0.0537, + "step": 33481 + }, + { + "epoch": 6.09, + "learning_rate": 1.1744448489261011e-06, + "loss": 0.4618, + "step": 33482 + }, + { + "epoch": 6.09, + "learning_rate": 1.174390243902439e-06, + "loss": 0.0313, + "step": 33483 + }, + { + "epoch": 6.09, + "learning_rate": 1.1743356388787768e-06, + "loss": 2.7765, + "step": 33484 + }, + { + "epoch": 6.09, + "learning_rate": 1.1742810338551147e-06, + "loss": 0.9827, + "step": 33485 + }, + { + "epoch": 6.09, + "learning_rate": 1.1742264288314527e-06, + "loss": 0.0576, + "step": 33486 + }, + { + "epoch": 6.09, + "learning_rate": 1.1741718238077904e-06, + "loss": 0.8161, + "step": 33487 + }, + { + "epoch": 6.09, + "learning_rate": 1.1741172187841282e-06, + "loss": 0.2366, + "step": 33488 + }, + { + "epoch": 6.09, + "learning_rate": 1.1740626137604659e-06, + "loss": 0.0069, + "step": 33489 + }, + { + "epoch": 6.09, + "learning_rate": 1.1740080087368038e-06, + "loss": 0.0073, + "step": 33490 + }, + { + "epoch": 6.09, + "learning_rate": 1.1739534037131416e-06, + "loss": 0.4409, + "step": 33491 + }, + { + "epoch": 6.09, + "learning_rate": 1.1738987986894795e-06, + "loss": 1.6906, + "step": 33492 + }, + { + "epoch": 6.09, + "learning_rate": 1.1738441936658174e-06, + "loss": 2.0634, + "step": 33493 + }, + { + "epoch": 6.09, + "learning_rate": 1.1737895886421552e-06, + "loss": 0.973, + "step": 33494 + }, + { + "epoch": 6.09, + "learning_rate": 1.173734983618493e-06, + "loss": 2.6069, + "step": 33495 + }, + { + "epoch": 6.09, + "learning_rate": 1.1736803785948306e-06, + "loss": 2.3842, + "step": 33496 + }, + { + "epoch": 6.09, + "learning_rate": 1.1736257735711686e-06, + "loss": 0.2633, + "step": 33497 + }, + { + "epoch": 6.09, + "learning_rate": 1.1735711685475063e-06, + "loss": 1.4949, + "step": 33498 + }, + { + "epoch": 6.09, + "learning_rate": 1.1735165635238443e-06, + "loss": 0.1211, + "step": 33499 + }, + { + "epoch": 6.09, + "learning_rate": 1.1734619585001822e-06, + "loss": 0.1393, + "step": 33500 + }, + { + "epoch": 6.09, + "learning_rate": 1.17340735347652e-06, + "loss": 0.8405, + "step": 33501 + }, + { + "epoch": 6.09, + "learning_rate": 1.1733527484528577e-06, + "loss": 2.7301, + "step": 33502 + }, + { + "epoch": 6.09, + "learning_rate": 1.1732981434291954e-06, + "loss": 0.4426, + "step": 33503 + }, + { + "epoch": 6.09, + "learning_rate": 1.1732435384055333e-06, + "loss": 0.1672, + "step": 33504 + }, + { + "epoch": 6.09, + "learning_rate": 1.173188933381871e-06, + "loss": 0.3901, + "step": 33505 + }, + { + "epoch": 6.09, + "learning_rate": 1.173134328358209e-06, + "loss": 3.3018, + "step": 33506 + }, + { + "epoch": 6.09, + "learning_rate": 1.173079723334547e-06, + "loss": 0.0458, + "step": 33507 + }, + { + "epoch": 6.09, + "learning_rate": 1.1730251183108847e-06, + "loss": 1.1613, + "step": 33508 + }, + { + "epoch": 6.09, + "learning_rate": 1.1729705132872224e-06, + "loss": 0.1093, + "step": 33509 + }, + { + "epoch": 6.09, + "learning_rate": 1.1729159082635602e-06, + "loss": 0.3692, + "step": 33510 + }, + { + "epoch": 6.09, + "learning_rate": 1.172861303239898e-06, + "loss": 0.0317, + "step": 33511 + }, + { + "epoch": 6.09, + "learning_rate": 1.1728066982162358e-06, + "loss": 0.7003, + "step": 33512 + }, + { + "epoch": 6.09, + "learning_rate": 1.1727520931925738e-06, + "loss": 0.0136, + "step": 33513 + }, + { + "epoch": 6.09, + "learning_rate": 1.1726974881689117e-06, + "loss": 0.0969, + "step": 33514 + }, + { + "epoch": 6.09, + "learning_rate": 1.1726428831452494e-06, + "loss": 0.3049, + "step": 33515 + }, + { + "epoch": 6.09, + "learning_rate": 1.1725882781215872e-06, + "loss": 0.8143, + "step": 33516 + }, + { + "epoch": 6.1, + "learning_rate": 1.172533673097925e-06, + "loss": 0.028, + "step": 33517 + }, + { + "epoch": 6.1, + "learning_rate": 1.1724790680742629e-06, + "loss": 0.3016, + "step": 33518 + }, + { + "epoch": 6.1, + "learning_rate": 1.1724244630506008e-06, + "loss": 0.0082, + "step": 33519 + }, + { + "epoch": 6.1, + "learning_rate": 1.1723698580269385e-06, + "loss": 0.075, + "step": 33520 + }, + { + "epoch": 6.1, + "learning_rate": 1.1723152530032765e-06, + "loss": 1.4854, + "step": 33521 + }, + { + "epoch": 6.1, + "learning_rate": 1.1722606479796142e-06, + "loss": 2.0148, + "step": 33522 + }, + { + "epoch": 6.1, + "learning_rate": 1.172206042955952e-06, + "loss": 0.6907, + "step": 33523 + }, + { + "epoch": 6.1, + "learning_rate": 1.1721514379322897e-06, + "loss": 0.0257, + "step": 33524 + }, + { + "epoch": 6.1, + "learning_rate": 1.1720968329086276e-06, + "loss": 0.0065, + "step": 33525 + }, + { + "epoch": 6.1, + "learning_rate": 1.1720422278849656e-06, + "loss": 1.8054, + "step": 33526 + }, + { + "epoch": 6.1, + "learning_rate": 1.1719876228613033e-06, + "loss": 0.3536, + "step": 33527 + }, + { + "epoch": 6.1, + "learning_rate": 1.1719330178376412e-06, + "loss": 0.0398, + "step": 33528 + }, + { + "epoch": 6.1, + "learning_rate": 1.171878412813979e-06, + "loss": 1.2062, + "step": 33529 + }, + { + "epoch": 6.1, + "learning_rate": 1.1718238077903167e-06, + "loss": 0.2771, + "step": 33530 + }, + { + "epoch": 6.1, + "learning_rate": 1.1717692027666544e-06, + "loss": 0.184, + "step": 33531 + }, + { + "epoch": 6.1, + "learning_rate": 1.1717145977429924e-06, + "loss": 0.2585, + "step": 33532 + }, + { + "epoch": 6.1, + "learning_rate": 1.1716599927193303e-06, + "loss": 0.2015, + "step": 33533 + }, + { + "epoch": 6.1, + "learning_rate": 1.171605387695668e-06, + "loss": 0.0942, + "step": 33534 + }, + { + "epoch": 6.1, + "learning_rate": 1.1715507826720058e-06, + "loss": 1.2406, + "step": 33535 + }, + { + "epoch": 6.1, + "learning_rate": 1.1714961776483437e-06, + "loss": 0.0837, + "step": 33536 + }, + { + "epoch": 6.1, + "learning_rate": 1.1714415726246814e-06, + "loss": 0.0558, + "step": 33537 + }, + { + "epoch": 6.1, + "learning_rate": 1.1713869676010192e-06, + "loss": 0.0569, + "step": 33538 + }, + { + "epoch": 6.1, + "learning_rate": 1.1713323625773571e-06, + "loss": 0.0956, + "step": 33539 + }, + { + "epoch": 6.1, + "learning_rate": 1.171277757553695e-06, + "loss": 0.6895, + "step": 33540 + }, + { + "epoch": 6.1, + "learning_rate": 1.1712231525300328e-06, + "loss": 0.3222, + "step": 33541 + }, + { + "epoch": 6.1, + "learning_rate": 1.1711685475063705e-06, + "loss": 0.0166, + "step": 33542 + }, + { + "epoch": 6.1, + "learning_rate": 1.1711139424827085e-06, + "loss": 0.0184, + "step": 33543 + }, + { + "epoch": 6.1, + "learning_rate": 1.1710593374590462e-06, + "loss": 0.0286, + "step": 33544 + }, + { + "epoch": 6.1, + "learning_rate": 1.171004732435384e-06, + "loss": 1.8336, + "step": 33545 + }, + { + "epoch": 6.1, + "learning_rate": 1.1709501274117219e-06, + "loss": 0.2416, + "step": 33546 + }, + { + "epoch": 6.1, + "learning_rate": 1.1708955223880598e-06, + "loss": 0.1385, + "step": 33547 + }, + { + "epoch": 6.1, + "learning_rate": 1.1708409173643976e-06, + "loss": 0.6282, + "step": 33548 + }, + { + "epoch": 6.1, + "learning_rate": 1.1707863123407353e-06, + "loss": 0.1761, + "step": 33549 + }, + { + "epoch": 6.1, + "learning_rate": 1.1707317073170732e-06, + "loss": 0.6003, + "step": 33550 + }, + { + "epoch": 6.1, + "learning_rate": 1.170677102293411e-06, + "loss": 1.9654, + "step": 33551 + }, + { + "epoch": 6.1, + "learning_rate": 1.1706224972697487e-06, + "loss": 0.6961, + "step": 33552 + }, + { + "epoch": 6.1, + "learning_rate": 1.1705678922460866e-06, + "loss": 0.0731, + "step": 33553 + }, + { + "epoch": 6.1, + "learning_rate": 1.1705132872224246e-06, + "loss": 0.0049, + "step": 33554 + }, + { + "epoch": 6.1, + "learning_rate": 1.1704586821987623e-06, + "loss": 0.0709, + "step": 33555 + }, + { + "epoch": 6.1, + "learning_rate": 1.1704040771751e-06, + "loss": 0.1799, + "step": 33556 + }, + { + "epoch": 6.1, + "learning_rate": 1.170349472151438e-06, + "loss": 0.9628, + "step": 33557 + }, + { + "epoch": 6.1, + "learning_rate": 1.1702948671277757e-06, + "loss": 1.4894, + "step": 33558 + }, + { + "epoch": 6.1, + "learning_rate": 1.1702402621041135e-06, + "loss": 2.147, + "step": 33559 + }, + { + "epoch": 6.1, + "learning_rate": 1.1701856570804514e-06, + "loss": 0.2699, + "step": 33560 + }, + { + "epoch": 6.1, + "learning_rate": 1.1701310520567893e-06, + "loss": 1.1378, + "step": 33561 + }, + { + "epoch": 6.1, + "learning_rate": 1.170076447033127e-06, + "loss": 0.0171, + "step": 33562 + }, + { + "epoch": 6.1, + "learning_rate": 1.1700218420094648e-06, + "loss": 0.0989, + "step": 33563 + }, + { + "epoch": 6.1, + "learning_rate": 1.1699672369858027e-06, + "loss": 0.0836, + "step": 33564 + }, + { + "epoch": 6.1, + "learning_rate": 1.1699126319621405e-06, + "loss": 2.2583, + "step": 33565 + }, + { + "epoch": 6.1, + "learning_rate": 1.1698580269384784e-06, + "loss": 1.7593, + "step": 33566 + }, + { + "epoch": 6.1, + "learning_rate": 1.1698034219148164e-06, + "loss": 1.499, + "step": 33567 + }, + { + "epoch": 6.1, + "learning_rate": 1.169748816891154e-06, + "loss": 1.3019, + "step": 33568 + }, + { + "epoch": 6.1, + "learning_rate": 1.1696942118674918e-06, + "loss": 0.4572, + "step": 33569 + }, + { + "epoch": 6.1, + "learning_rate": 1.1696396068438296e-06, + "loss": 0.0849, + "step": 33570 + }, + { + "epoch": 6.1, + "learning_rate": 1.1695850018201675e-06, + "loss": 0.9996, + "step": 33571 + }, + { + "epoch": 6.11, + "learning_rate": 1.1695303967965052e-06, + "loss": 0.0474, + "step": 33572 + }, + { + "epoch": 6.11, + "learning_rate": 1.1694757917728432e-06, + "loss": 0.081, + "step": 33573 + }, + { + "epoch": 6.11, + "learning_rate": 1.1694211867491811e-06, + "loss": 0.0109, + "step": 33574 + }, + { + "epoch": 6.11, + "learning_rate": 1.1693665817255189e-06, + "loss": 0.0823, + "step": 33575 + }, + { + "epoch": 6.11, + "learning_rate": 1.1693119767018566e-06, + "loss": 0.3895, + "step": 33576 + }, + { + "epoch": 6.11, + "learning_rate": 1.1692573716781943e-06, + "loss": 0.2138, + "step": 33577 + }, + { + "epoch": 6.11, + "learning_rate": 1.1692027666545323e-06, + "loss": 0.3763, + "step": 33578 + }, + { + "epoch": 6.11, + "learning_rate": 1.16914816163087e-06, + "loss": 0.0801, + "step": 33579 + }, + { + "epoch": 6.11, + "learning_rate": 1.169093556607208e-06, + "loss": 1.0749, + "step": 33580 + }, + { + "epoch": 6.11, + "learning_rate": 1.1690389515835459e-06, + "loss": 0.1553, + "step": 33581 + }, + { + "epoch": 6.11, + "learning_rate": 1.1689843465598836e-06, + "loss": 0.1597, + "step": 33582 + }, + { + "epoch": 6.11, + "learning_rate": 1.1689297415362213e-06, + "loss": 0.3009, + "step": 33583 + }, + { + "epoch": 6.11, + "learning_rate": 1.168875136512559e-06, + "loss": 4.4443, + "step": 33584 + }, + { + "epoch": 6.11, + "learning_rate": 1.168820531488897e-06, + "loss": 0.1158, + "step": 33585 + }, + { + "epoch": 6.11, + "learning_rate": 1.1687659264652347e-06, + "loss": 0.1394, + "step": 33586 + }, + { + "epoch": 6.11, + "learning_rate": 1.1687113214415727e-06, + "loss": 0.2504, + "step": 33587 + }, + { + "epoch": 6.11, + "learning_rate": 1.1686567164179106e-06, + "loss": 0.0062, + "step": 33588 + }, + { + "epoch": 6.11, + "learning_rate": 1.1686021113942484e-06, + "loss": 0.057, + "step": 33589 + }, + { + "epoch": 6.11, + "learning_rate": 1.168547506370586e-06, + "loss": 0.0179, + "step": 33590 + }, + { + "epoch": 6.11, + "learning_rate": 1.1684929013469238e-06, + "loss": 0.0308, + "step": 33591 + }, + { + "epoch": 6.11, + "learning_rate": 1.1684382963232618e-06, + "loss": 0.6432, + "step": 33592 + }, + { + "epoch": 6.11, + "learning_rate": 1.1683836912995995e-06, + "loss": 0.1895, + "step": 33593 + }, + { + "epoch": 6.11, + "learning_rate": 1.1683290862759374e-06, + "loss": 0.0175, + "step": 33594 + }, + { + "epoch": 6.11, + "learning_rate": 1.1682744812522754e-06, + "loss": 0.2383, + "step": 33595 + }, + { + "epoch": 6.11, + "learning_rate": 1.1682198762286131e-06, + "loss": 2.5955, + "step": 33596 + }, + { + "epoch": 6.11, + "learning_rate": 1.1681652712049509e-06, + "loss": 1.4046, + "step": 33597 + }, + { + "epoch": 6.11, + "learning_rate": 1.1681106661812886e-06, + "loss": 0.0635, + "step": 33598 + }, + { + "epoch": 6.11, + "learning_rate": 1.1680560611576265e-06, + "loss": 0.8406, + "step": 33599 + }, + { + "epoch": 6.11, + "learning_rate": 1.1680014561339643e-06, + "loss": 0.0064, + "step": 33600 + }, + { + "epoch": 6.11, + "learning_rate": 1.1679468511103022e-06, + "loss": 1.0735, + "step": 33601 + }, + { + "epoch": 6.11, + "learning_rate": 1.1678922460866401e-06, + "loss": 0.9035, + "step": 33602 + }, + { + "epoch": 6.11, + "learning_rate": 1.1678376410629779e-06, + "loss": 0.015, + "step": 33603 + }, + { + "epoch": 6.11, + "learning_rate": 1.1677830360393156e-06, + "loss": 0.2362, + "step": 33604 + }, + { + "epoch": 6.11, + "learning_rate": 1.1677284310156533e-06, + "loss": 0.0095, + "step": 33605 + }, + { + "epoch": 6.11, + "learning_rate": 1.1676738259919913e-06, + "loss": 0.776, + "step": 33606 + }, + { + "epoch": 6.11, + "learning_rate": 1.1676192209683292e-06, + "loss": 1.6087, + "step": 33607 + }, + { + "epoch": 6.11, + "learning_rate": 1.167564615944667e-06, + "loss": 0.659, + "step": 33608 + }, + { + "epoch": 6.11, + "learning_rate": 1.167510010921005e-06, + "loss": 0.1725, + "step": 33609 + }, + { + "epoch": 6.11, + "learning_rate": 1.1674554058973426e-06, + "loss": 1.6977, + "step": 33610 + }, + { + "epoch": 6.11, + "learning_rate": 1.1674008008736804e-06, + "loss": 2.5068, + "step": 33611 + }, + { + "epoch": 6.11, + "learning_rate": 1.167346195850018e-06, + "loss": 3.8523, + "step": 33612 + }, + { + "epoch": 6.11, + "learning_rate": 1.167291590826356e-06, + "loss": 0.0812, + "step": 33613 + }, + { + "epoch": 6.11, + "learning_rate": 1.167236985802694e-06, + "loss": 1.6645, + "step": 33614 + }, + { + "epoch": 6.11, + "learning_rate": 1.1671823807790317e-06, + "loss": 1.3704, + "step": 33615 + }, + { + "epoch": 6.11, + "learning_rate": 1.1671277757553694e-06, + "loss": 0.3087, + "step": 33616 + }, + { + "epoch": 6.11, + "learning_rate": 1.1670731707317074e-06, + "loss": 0.0051, + "step": 33617 + }, + { + "epoch": 6.11, + "learning_rate": 1.1670185657080451e-06, + "loss": 0.0086, + "step": 33618 + }, + { + "epoch": 6.11, + "learning_rate": 1.1669639606843829e-06, + "loss": 0.0143, + "step": 33619 + }, + { + "epoch": 6.11, + "learning_rate": 1.1669093556607208e-06, + "loss": 0.1998, + "step": 33620 + }, + { + "epoch": 6.11, + "learning_rate": 1.1668547506370587e-06, + "loss": 0.0448, + "step": 33621 + }, + { + "epoch": 6.11, + "learning_rate": 1.1668001456133965e-06, + "loss": 0.0679, + "step": 33622 + }, + { + "epoch": 6.11, + "learning_rate": 1.1667455405897342e-06, + "loss": 0.0051, + "step": 33623 + }, + { + "epoch": 6.11, + "learning_rate": 1.1666909355660721e-06, + "loss": 0.6089, + "step": 33624 + }, + { + "epoch": 6.11, + "learning_rate": 1.1666363305424099e-06, + "loss": 0.093, + "step": 33625 + }, + { + "epoch": 6.11, + "learning_rate": 1.1665817255187476e-06, + "loss": 0.0067, + "step": 33626 + }, + { + "epoch": 6.12, + "learning_rate": 1.1665271204950856e-06, + "loss": 2.0879, + "step": 33627 + }, + { + "epoch": 6.12, + "learning_rate": 1.1664725154714235e-06, + "loss": 0.0511, + "step": 33628 + }, + { + "epoch": 6.12, + "learning_rate": 1.1664179104477612e-06, + "loss": 0.3131, + "step": 33629 + }, + { + "epoch": 6.12, + "learning_rate": 1.166363305424099e-06, + "loss": 0.193, + "step": 33630 + }, + { + "epoch": 6.12, + "learning_rate": 1.166308700400437e-06, + "loss": 0.0049, + "step": 33631 + }, + { + "epoch": 6.12, + "learning_rate": 1.1662540953767746e-06, + "loss": 0.0291, + "step": 33632 + }, + { + "epoch": 6.12, + "learning_rate": 1.1661994903531124e-06, + "loss": 0.4786, + "step": 33633 + }, + { + "epoch": 6.12, + "learning_rate": 1.1661448853294503e-06, + "loss": 1.3952, + "step": 33634 + }, + { + "epoch": 6.12, + "learning_rate": 1.1660902803057883e-06, + "loss": 0.0748, + "step": 33635 + }, + { + "epoch": 6.12, + "learning_rate": 1.166035675282126e-06, + "loss": 0.0546, + "step": 33636 + }, + { + "epoch": 6.12, + "learning_rate": 1.1659810702584637e-06, + "loss": 1.0093, + "step": 33637 + }, + { + "epoch": 6.12, + "learning_rate": 1.1659264652348017e-06, + "loss": 0.6729, + "step": 33638 + }, + { + "epoch": 6.12, + "learning_rate": 1.1658718602111394e-06, + "loss": 0.3554, + "step": 33639 + }, + { + "epoch": 6.12, + "learning_rate": 1.1658172551874771e-06, + "loss": 2.1891, + "step": 33640 + }, + { + "epoch": 6.12, + "learning_rate": 1.165762650163815e-06, + "loss": 0.9101, + "step": 33641 + }, + { + "epoch": 6.12, + "learning_rate": 1.165708045140153e-06, + "loss": 0.606, + "step": 33642 + }, + { + "epoch": 6.12, + "learning_rate": 1.1656534401164907e-06, + "loss": 0.0831, + "step": 33643 + }, + { + "epoch": 6.12, + "learning_rate": 1.1655988350928285e-06, + "loss": 0.0066, + "step": 33644 + }, + { + "epoch": 6.12, + "learning_rate": 1.1655442300691664e-06, + "loss": 1.1921, + "step": 33645 + }, + { + "epoch": 6.12, + "learning_rate": 1.1654896250455042e-06, + "loss": 0.1603, + "step": 33646 + }, + { + "epoch": 6.12, + "learning_rate": 1.165435020021842e-06, + "loss": 0.2355, + "step": 33647 + }, + { + "epoch": 6.12, + "learning_rate": 1.16538041499818e-06, + "loss": 0.1984, + "step": 33648 + }, + { + "epoch": 6.12, + "learning_rate": 1.1653258099745178e-06, + "loss": 1.4225, + "step": 33649 + }, + { + "epoch": 6.12, + "learning_rate": 1.1652712049508555e-06, + "loss": 0.0112, + "step": 33650 + }, + { + "epoch": 6.12, + "learning_rate": 1.1652165999271932e-06, + "loss": 0.8832, + "step": 33651 + }, + { + "epoch": 6.12, + "learning_rate": 1.1651619949035312e-06, + "loss": 1.5934, + "step": 33652 + }, + { + "epoch": 6.12, + "learning_rate": 1.165107389879869e-06, + "loss": 0.0776, + "step": 33653 + }, + { + "epoch": 6.12, + "learning_rate": 1.1650527848562069e-06, + "loss": 0.8008, + "step": 33654 + }, + { + "epoch": 6.12, + "learning_rate": 1.1649981798325448e-06, + "loss": 0.138, + "step": 33655 + }, + { + "epoch": 6.12, + "learning_rate": 1.1649435748088825e-06, + "loss": 1.3182, + "step": 33656 + }, + { + "epoch": 6.12, + "learning_rate": 1.1648889697852203e-06, + "loss": 0.2787, + "step": 33657 + }, + { + "epoch": 6.12, + "learning_rate": 1.164834364761558e-06, + "loss": 0.1707, + "step": 33658 + }, + { + "epoch": 6.12, + "learning_rate": 1.164779759737896e-06, + "loss": 0.0891, + "step": 33659 + }, + { + "epoch": 6.12, + "learning_rate": 1.1647251547142337e-06, + "loss": 1.1395, + "step": 33660 + }, + { + "epoch": 6.12, + "learning_rate": 1.1646705496905716e-06, + "loss": 0.843, + "step": 33661 + }, + { + "epoch": 6.12, + "learning_rate": 1.1646159446669096e-06, + "loss": 0.4981, + "step": 33662 + }, + { + "epoch": 6.12, + "learning_rate": 1.1645613396432473e-06, + "loss": 0.1777, + "step": 33663 + }, + { + "epoch": 6.12, + "learning_rate": 1.164506734619585e-06, + "loss": 0.0082, + "step": 33664 + }, + { + "epoch": 6.12, + "learning_rate": 1.1644521295959227e-06, + "loss": 0.0038, + "step": 33665 + }, + { + "epoch": 6.12, + "learning_rate": 1.1643975245722607e-06, + "loss": 0.7629, + "step": 33666 + }, + { + "epoch": 6.12, + "learning_rate": 1.1643429195485984e-06, + "loss": 0.0801, + "step": 33667 + }, + { + "epoch": 6.12, + "learning_rate": 1.1642883145249364e-06, + "loss": 0.0275, + "step": 33668 + }, + { + "epoch": 6.12, + "learning_rate": 1.1642337095012743e-06, + "loss": 0.9776, + "step": 33669 + }, + { + "epoch": 6.12, + "learning_rate": 1.164179104477612e-06, + "loss": 0.1259, + "step": 33670 + }, + { + "epoch": 6.12, + "learning_rate": 1.1641244994539498e-06, + "loss": 0.2001, + "step": 33671 + }, + { + "epoch": 6.12, + "learning_rate": 1.1640698944302875e-06, + "loss": 0.1627, + "step": 33672 + }, + { + "epoch": 6.12, + "learning_rate": 1.1640152894066254e-06, + "loss": 0.6902, + "step": 33673 + }, + { + "epoch": 6.12, + "learning_rate": 1.1639606843829632e-06, + "loss": 0.0802, + "step": 33674 + }, + { + "epoch": 6.12, + "learning_rate": 1.1639060793593011e-06, + "loss": 0.4204, + "step": 33675 + }, + { + "epoch": 6.12, + "learning_rate": 1.163851474335639e-06, + "loss": 0.2847, + "step": 33676 + }, + { + "epoch": 6.12, + "learning_rate": 1.1637968693119768e-06, + "loss": 1.6766, + "step": 33677 + }, + { + "epoch": 6.12, + "learning_rate": 1.1637422642883145e-06, + "loss": 0.0913, + "step": 33678 + }, + { + "epoch": 6.12, + "learning_rate": 1.1636876592646523e-06, + "loss": 0.5555, + "step": 33679 + }, + { + "epoch": 6.12, + "learning_rate": 1.1636330542409902e-06, + "loss": 0.157, + "step": 33680 + }, + { + "epoch": 6.12, + "learning_rate": 1.163578449217328e-06, + "loss": 0.2291, + "step": 33681 + }, + { + "epoch": 6.13, + "learning_rate": 1.1635238441936659e-06, + "loss": 1.3603, + "step": 33682 + }, + { + "epoch": 6.13, + "learning_rate": 1.1634692391700038e-06, + "loss": 1.4931, + "step": 33683 + }, + { + "epoch": 6.13, + "learning_rate": 1.1634146341463416e-06, + "loss": 2.424, + "step": 33684 + }, + { + "epoch": 6.13, + "learning_rate": 1.1633600291226793e-06, + "loss": 0.0405, + "step": 33685 + }, + { + "epoch": 6.13, + "learning_rate": 1.163305424099017e-06, + "loss": 0.0056, + "step": 33686 + }, + { + "epoch": 6.13, + "learning_rate": 1.163250819075355e-06, + "loss": 0.0506, + "step": 33687 + }, + { + "epoch": 6.13, + "learning_rate": 1.1631962140516927e-06, + "loss": 2.4226, + "step": 33688 + }, + { + "epoch": 6.13, + "learning_rate": 1.1631416090280306e-06, + "loss": 0.6501, + "step": 33689 + }, + { + "epoch": 6.13, + "learning_rate": 1.1630870040043684e-06, + "loss": 1.5782, + "step": 33690 + }, + { + "epoch": 6.13, + "learning_rate": 1.1630323989807063e-06, + "loss": 0.87, + "step": 33691 + }, + { + "epoch": 6.13, + "learning_rate": 1.162977793957044e-06, + "loss": 0.9543, + "step": 33692 + }, + { + "epoch": 6.13, + "learning_rate": 1.1629231889333818e-06, + "loss": 3.0323, + "step": 33693 + }, + { + "epoch": 6.13, + "learning_rate": 1.1628685839097197e-06, + "loss": 0.4214, + "step": 33694 + }, + { + "epoch": 6.13, + "learning_rate": 1.1628139788860577e-06, + "loss": 0.0059, + "step": 33695 + }, + { + "epoch": 6.13, + "learning_rate": 1.1627593738623954e-06, + "loss": 0.0407, + "step": 33696 + }, + { + "epoch": 6.13, + "learning_rate": 1.1627047688387331e-06, + "loss": 0.0172, + "step": 33697 + }, + { + "epoch": 6.13, + "learning_rate": 1.162650163815071e-06, + "loss": 0.2419, + "step": 33698 + }, + { + "epoch": 6.13, + "learning_rate": 1.1625955587914088e-06, + "loss": 0.1465, + "step": 33699 + }, + { + "epoch": 6.13, + "learning_rate": 1.1625409537677465e-06, + "loss": 0.436, + "step": 33700 + }, + { + "epoch": 6.13, + "learning_rate": 1.1624863487440845e-06, + "loss": 1.7443, + "step": 33701 + }, + { + "epoch": 6.13, + "learning_rate": 1.1624317437204224e-06, + "loss": 0.3522, + "step": 33702 + }, + { + "epoch": 6.13, + "learning_rate": 1.1623771386967601e-06, + "loss": 3.1923, + "step": 33703 + }, + { + "epoch": 6.13, + "learning_rate": 1.1623225336730979e-06, + "loss": 0.0179, + "step": 33704 + }, + { + "epoch": 6.13, + "learning_rate": 1.1622679286494358e-06, + "loss": 0.4053, + "step": 33705 + }, + { + "epoch": 6.13, + "learning_rate": 1.1622133236257736e-06, + "loss": 0.0845, + "step": 33706 + }, + { + "epoch": 6.13, + "learning_rate": 1.1621587186021113e-06, + "loss": 0.7848, + "step": 33707 + }, + { + "epoch": 6.13, + "learning_rate": 1.1621041135784492e-06, + "loss": 1.7952, + "step": 33708 + }, + { + "epoch": 6.13, + "learning_rate": 1.1620495085547872e-06, + "loss": 0.1298, + "step": 33709 + }, + { + "epoch": 6.13, + "learning_rate": 1.161994903531125e-06, + "loss": 0.0487, + "step": 33710 + }, + { + "epoch": 6.13, + "learning_rate": 1.1619402985074626e-06, + "loss": 1.7024, + "step": 33711 + }, + { + "epoch": 6.13, + "learning_rate": 1.1618856934838006e-06, + "loss": 2.7442, + "step": 33712 + }, + { + "epoch": 6.13, + "learning_rate": 1.1618310884601383e-06, + "loss": 0.5651, + "step": 33713 + }, + { + "epoch": 6.13, + "learning_rate": 1.161776483436476e-06, + "loss": 0.1254, + "step": 33714 + }, + { + "epoch": 6.13, + "learning_rate": 1.161721878412814e-06, + "loss": 0.8124, + "step": 33715 + }, + { + "epoch": 6.13, + "learning_rate": 1.161667273389152e-06, + "loss": 0.2121, + "step": 33716 + }, + { + "epoch": 6.13, + "learning_rate": 1.1616126683654897e-06, + "loss": 1.2877, + "step": 33717 + }, + { + "epoch": 6.13, + "learning_rate": 1.1615580633418274e-06, + "loss": 1.5616, + "step": 33718 + }, + { + "epoch": 6.13, + "learning_rate": 1.1615034583181653e-06, + "loss": 0.0807, + "step": 33719 + }, + { + "epoch": 6.13, + "learning_rate": 1.161448853294503e-06, + "loss": 0.4155, + "step": 33720 + }, + { + "epoch": 6.13, + "learning_rate": 1.1613942482708408e-06, + "loss": 0.1558, + "step": 33721 + }, + { + "epoch": 6.13, + "learning_rate": 1.1613396432471787e-06, + "loss": 0.0397, + "step": 33722 + }, + { + "epoch": 6.13, + "learning_rate": 1.1612850382235167e-06, + "loss": 0.076, + "step": 33723 + }, + { + "epoch": 6.13, + "learning_rate": 1.1612304331998544e-06, + "loss": 0.2871, + "step": 33724 + }, + { + "epoch": 6.13, + "learning_rate": 1.1611758281761921e-06, + "loss": 0.8073, + "step": 33725 + }, + { + "epoch": 6.13, + "learning_rate": 1.16112122315253e-06, + "loss": 0.0096, + "step": 33726 + }, + { + "epoch": 6.13, + "learning_rate": 1.1610666181288678e-06, + "loss": 0.0378, + "step": 33727 + }, + { + "epoch": 6.13, + "learning_rate": 1.1610120131052056e-06, + "loss": 0.1962, + "step": 33728 + }, + { + "epoch": 6.13, + "learning_rate": 1.1609574080815435e-06, + "loss": 0.4005, + "step": 33729 + }, + { + "epoch": 6.13, + "learning_rate": 1.1609028030578814e-06, + "loss": 0.2447, + "step": 33730 + }, + { + "epoch": 6.13, + "learning_rate": 1.1608481980342192e-06, + "loss": 0.0417, + "step": 33731 + }, + { + "epoch": 6.13, + "learning_rate": 1.160793593010557e-06, + "loss": 0.4201, + "step": 33732 + }, + { + "epoch": 6.13, + "learning_rate": 1.1607389879868948e-06, + "loss": 2.5055, + "step": 33733 + }, + { + "epoch": 6.13, + "learning_rate": 1.1606843829632326e-06, + "loss": 0.1496, + "step": 33734 + }, + { + "epoch": 6.13, + "learning_rate": 1.1606297779395705e-06, + "loss": 0.6655, + "step": 33735 + }, + { + "epoch": 6.13, + "learning_rate": 1.1605751729159085e-06, + "loss": 0.0739, + "step": 33736 + }, + { + "epoch": 6.14, + "learning_rate": 1.1605205678922462e-06, + "loss": 0.6106, + "step": 33737 + }, + { + "epoch": 6.14, + "learning_rate": 1.160465962868584e-06, + "loss": 0.1019, + "step": 33738 + }, + { + "epoch": 6.14, + "learning_rate": 1.1604113578449217e-06, + "loss": 0.4618, + "step": 33739 + }, + { + "epoch": 6.14, + "learning_rate": 1.1603567528212596e-06, + "loss": 1.1248, + "step": 33740 + }, + { + "epoch": 6.14, + "learning_rate": 1.1603021477975973e-06, + "loss": 0.3632, + "step": 33741 + }, + { + "epoch": 6.14, + "learning_rate": 1.1602475427739353e-06, + "loss": 0.7588, + "step": 33742 + }, + { + "epoch": 6.14, + "learning_rate": 1.1601929377502732e-06, + "loss": 1.1451, + "step": 33743 + }, + { + "epoch": 6.14, + "learning_rate": 1.160138332726611e-06, + "loss": 0.0163, + "step": 33744 + }, + { + "epoch": 6.14, + "learning_rate": 1.1600837277029487e-06, + "loss": 0.0543, + "step": 33745 + }, + { + "epoch": 6.14, + "learning_rate": 1.1600291226792864e-06, + "loss": 0.6641, + "step": 33746 + }, + { + "epoch": 6.14, + "learning_rate": 1.1599745176556244e-06, + "loss": 1.9318, + "step": 33747 + }, + { + "epoch": 6.14, + "learning_rate": 1.159919912631962e-06, + "loss": 0.2257, + "step": 33748 + }, + { + "epoch": 6.14, + "learning_rate": 1.1598653076083e-06, + "loss": 1.7205, + "step": 33749 + }, + { + "epoch": 6.14, + "learning_rate": 1.159810702584638e-06, + "loss": 0.0105, + "step": 33750 + }, + { + "epoch": 6.14, + "learning_rate": 1.1597560975609757e-06, + "loss": 2.0398, + "step": 33751 + }, + { + "epoch": 6.14, + "learning_rate": 1.1597014925373134e-06, + "loss": 0.0671, + "step": 33752 + }, + { + "epoch": 6.14, + "learning_rate": 1.1596468875136512e-06, + "loss": 0.0185, + "step": 33753 + }, + { + "epoch": 6.14, + "learning_rate": 1.1595922824899891e-06, + "loss": 0.0559, + "step": 33754 + }, + { + "epoch": 6.14, + "learning_rate": 1.1595376774663269e-06, + "loss": 0.412, + "step": 33755 + }, + { + "epoch": 6.14, + "learning_rate": 1.1594830724426648e-06, + "loss": 0.0104, + "step": 33756 + }, + { + "epoch": 6.14, + "learning_rate": 1.1594284674190027e-06, + "loss": 0.5104, + "step": 33757 + }, + { + "epoch": 6.14, + "learning_rate": 1.1593738623953405e-06, + "loss": 1.1839, + "step": 33758 + }, + { + "epoch": 6.14, + "learning_rate": 1.1593192573716782e-06, + "loss": 0.1596, + "step": 33759 + }, + { + "epoch": 6.14, + "learning_rate": 1.159264652348016e-06, + "loss": 0.3283, + "step": 33760 + }, + { + "epoch": 6.14, + "learning_rate": 1.1592100473243539e-06, + "loss": 0.0327, + "step": 33761 + }, + { + "epoch": 6.14, + "learning_rate": 1.1591554423006916e-06, + "loss": 0.0118, + "step": 33762 + }, + { + "epoch": 6.14, + "learning_rate": 1.1591008372770296e-06, + "loss": 0.0062, + "step": 33763 + }, + { + "epoch": 6.14, + "learning_rate": 1.1590462322533675e-06, + "loss": 2.5432, + "step": 33764 + }, + { + "epoch": 6.14, + "learning_rate": 1.1589916272297052e-06, + "loss": 0.059, + "step": 33765 + }, + { + "epoch": 6.14, + "learning_rate": 1.158937022206043e-06, + "loss": 0.0941, + "step": 33766 + }, + { + "epoch": 6.14, + "learning_rate": 1.1588824171823807e-06, + "loss": 0.7584, + "step": 33767 + }, + { + "epoch": 6.14, + "learning_rate": 1.1588278121587186e-06, + "loss": 0.8079, + "step": 33768 + }, + { + "epoch": 6.14, + "learning_rate": 1.1587732071350564e-06, + "loss": 1.7265, + "step": 33769 + }, + { + "epoch": 6.14, + "learning_rate": 1.1587186021113943e-06, + "loss": 0.5604, + "step": 33770 + }, + { + "epoch": 6.14, + "learning_rate": 1.158663997087732e-06, + "loss": 0.1521, + "step": 33771 + }, + { + "epoch": 6.14, + "learning_rate": 1.15860939206407e-06, + "loss": 0.375, + "step": 33772 + }, + { + "epoch": 6.14, + "learning_rate": 1.1585547870404077e-06, + "loss": 1.5448, + "step": 33773 + }, + { + "epoch": 6.14, + "learning_rate": 1.1585001820167454e-06, + "loss": 0.0058, + "step": 33774 + }, + { + "epoch": 6.14, + "learning_rate": 1.1584455769930834e-06, + "loss": 0.4229, + "step": 33775 + }, + { + "epoch": 6.14, + "learning_rate": 1.1583909719694213e-06, + "loss": 0.0827, + "step": 33776 + }, + { + "epoch": 6.14, + "learning_rate": 1.158336366945759e-06, + "loss": 0.1218, + "step": 33777 + }, + { + "epoch": 6.14, + "learning_rate": 1.1582817619220968e-06, + "loss": 0.0082, + "step": 33778 + }, + { + "epoch": 6.14, + "learning_rate": 1.1582271568984347e-06, + "loss": 0.0266, + "step": 33779 + }, + { + "epoch": 6.14, + "learning_rate": 1.1581725518747725e-06, + "loss": 2.2807, + "step": 33780 + }, + { + "epoch": 6.14, + "learning_rate": 1.1581179468511102e-06, + "loss": 1.3261, + "step": 33781 + }, + { + "epoch": 6.14, + "learning_rate": 1.1580633418274481e-06, + "loss": 1.4396, + "step": 33782 + }, + { + "epoch": 6.14, + "learning_rate": 1.158008736803786e-06, + "loss": 0.0058, + "step": 33783 + }, + { + "epoch": 6.14, + "learning_rate": 1.1579541317801238e-06, + "loss": 0.0074, + "step": 33784 + }, + { + "epoch": 6.14, + "learning_rate": 1.1578995267564616e-06, + "loss": 0.9687, + "step": 33785 + }, + { + "epoch": 6.14, + "learning_rate": 1.1578449217327995e-06, + "loss": 0.0133, + "step": 33786 + }, + { + "epoch": 6.14, + "learning_rate": 1.1577903167091372e-06, + "loss": 0.3371, + "step": 33787 + }, + { + "epoch": 6.14, + "learning_rate": 1.157735711685475e-06, + "loss": 0.1487, + "step": 33788 + }, + { + "epoch": 6.14, + "learning_rate": 1.157681106661813e-06, + "loss": 0.9751, + "step": 33789 + }, + { + "epoch": 6.14, + "learning_rate": 1.1576265016381508e-06, + "loss": 1.1995, + "step": 33790 + }, + { + "epoch": 6.14, + "learning_rate": 1.1575718966144886e-06, + "loss": 0.1373, + "step": 33791 + }, + { + "epoch": 6.15, + "learning_rate": 1.1575172915908263e-06, + "loss": 0.3382, + "step": 33792 + }, + { + "epoch": 6.15, + "learning_rate": 1.1574626865671643e-06, + "loss": 1.4386, + "step": 33793 + }, + { + "epoch": 6.15, + "learning_rate": 1.157408081543502e-06, + "loss": 0.5593, + "step": 33794 + }, + { + "epoch": 6.15, + "learning_rate": 1.1573534765198397e-06, + "loss": 0.066, + "step": 33795 + }, + { + "epoch": 6.15, + "learning_rate": 1.1572988714961777e-06, + "loss": 0.3213, + "step": 33796 + }, + { + "epoch": 6.15, + "learning_rate": 1.1572442664725156e-06, + "loss": 1.8546, + "step": 33797 + }, + { + "epoch": 6.15, + "learning_rate": 1.1571896614488533e-06, + "loss": 0.8818, + "step": 33798 + }, + { + "epoch": 6.15, + "learning_rate": 1.157135056425191e-06, + "loss": 3.0553, + "step": 33799 + }, + { + "epoch": 6.15, + "learning_rate": 1.157080451401529e-06, + "loss": 0.0177, + "step": 33800 + }, + { + "epoch": 6.15, + "learning_rate": 1.1570258463778667e-06, + "loss": 0.8035, + "step": 33801 + }, + { + "epoch": 6.15, + "learning_rate": 1.1569712413542045e-06, + "loss": 0.2448, + "step": 33802 + }, + { + "epoch": 6.15, + "learning_rate": 1.1569166363305424e-06, + "loss": 0.1147, + "step": 33803 + }, + { + "epoch": 6.15, + "learning_rate": 1.1568620313068804e-06, + "loss": 0.2157, + "step": 33804 + }, + { + "epoch": 6.15, + "learning_rate": 1.156807426283218e-06, + "loss": 1.6355, + "step": 33805 + }, + { + "epoch": 6.15, + "learning_rate": 1.1567528212595558e-06, + "loss": 0.0229, + "step": 33806 + }, + { + "epoch": 6.15, + "learning_rate": 1.1566982162358938e-06, + "loss": 0.2217, + "step": 33807 + }, + { + "epoch": 6.15, + "learning_rate": 1.1566436112122315e-06, + "loss": 1.2077, + "step": 33808 + }, + { + "epoch": 6.15, + "learning_rate": 1.1565890061885692e-06, + "loss": 0.2673, + "step": 33809 + }, + { + "epoch": 6.15, + "learning_rate": 1.1565344011649072e-06, + "loss": 1.1834, + "step": 33810 + }, + { + "epoch": 6.15, + "learning_rate": 1.1564797961412451e-06, + "loss": 1.8171, + "step": 33811 + }, + { + "epoch": 6.15, + "learning_rate": 1.1564251911175828e-06, + "loss": 0.0091, + "step": 33812 + }, + { + "epoch": 6.15, + "learning_rate": 1.1563705860939206e-06, + "loss": 0.2336, + "step": 33813 + }, + { + "epoch": 6.15, + "learning_rate": 1.1563159810702585e-06, + "loss": 1.4119, + "step": 33814 + }, + { + "epoch": 6.15, + "learning_rate": 1.1562613760465963e-06, + "loss": 0.4012, + "step": 33815 + }, + { + "epoch": 6.15, + "learning_rate": 1.1562067710229342e-06, + "loss": 0.0066, + "step": 33816 + }, + { + "epoch": 6.15, + "learning_rate": 1.1561521659992721e-06, + "loss": 0.0418, + "step": 33817 + }, + { + "epoch": 6.15, + "learning_rate": 1.1560975609756099e-06, + "loss": 1.7825, + "step": 33818 + }, + { + "epoch": 6.15, + "learning_rate": 1.1560429559519476e-06, + "loss": 0.7698, + "step": 33819 + }, + { + "epoch": 6.15, + "learning_rate": 1.1559883509282853e-06, + "loss": 0.2891, + "step": 33820 + }, + { + "epoch": 6.15, + "learning_rate": 1.1559337459046233e-06, + "loss": 0.2363, + "step": 33821 + }, + { + "epoch": 6.15, + "learning_rate": 1.155879140880961e-06, + "loss": 0.1615, + "step": 33822 + }, + { + "epoch": 6.15, + "learning_rate": 1.155824535857299e-06, + "loss": 0.4857, + "step": 33823 + }, + { + "epoch": 6.15, + "learning_rate": 1.155769930833637e-06, + "loss": 1.3309, + "step": 33824 + }, + { + "epoch": 6.15, + "learning_rate": 1.1557153258099746e-06, + "loss": 0.0207, + "step": 33825 + }, + { + "epoch": 6.15, + "learning_rate": 1.1556607207863124e-06, + "loss": 0.2073, + "step": 33826 + }, + { + "epoch": 6.15, + "learning_rate": 1.15560611576265e-06, + "loss": 0.1314, + "step": 33827 + }, + { + "epoch": 6.15, + "learning_rate": 1.155551510738988e-06, + "loss": 0.008, + "step": 33828 + }, + { + "epoch": 6.15, + "learning_rate": 1.1554969057153258e-06, + "loss": 1.0979, + "step": 33829 + }, + { + "epoch": 6.15, + "learning_rate": 1.1554423006916637e-06, + "loss": 0.8727, + "step": 33830 + }, + { + "epoch": 6.15, + "learning_rate": 1.1553876956680017e-06, + "loss": 0.07, + "step": 33831 + }, + { + "epoch": 6.15, + "learning_rate": 1.1553330906443394e-06, + "loss": 0.2751, + "step": 33832 + }, + { + "epoch": 6.15, + "learning_rate": 1.1552784856206771e-06, + "loss": 0.9687, + "step": 33833 + }, + { + "epoch": 6.15, + "learning_rate": 1.1552238805970149e-06, + "loss": 0.4994, + "step": 33834 + }, + { + "epoch": 6.15, + "learning_rate": 1.1551692755733528e-06, + "loss": 0.8313, + "step": 33835 + }, + { + "epoch": 6.15, + "learning_rate": 1.1551146705496905e-06, + "loss": 0.1001, + "step": 33836 + }, + { + "epoch": 6.15, + "learning_rate": 1.1550600655260285e-06, + "loss": 0.4354, + "step": 33837 + }, + { + "epoch": 6.15, + "learning_rate": 1.1550054605023664e-06, + "loss": 0.4865, + "step": 33838 + }, + { + "epoch": 6.15, + "learning_rate": 1.1549508554787041e-06, + "loss": 0.3005, + "step": 33839 + }, + { + "epoch": 6.15, + "learning_rate": 1.1548962504550419e-06, + "loss": 0.0631, + "step": 33840 + }, + { + "epoch": 6.15, + "learning_rate": 1.1548416454313796e-06, + "loss": 0.0069, + "step": 33841 + }, + { + "epoch": 6.15, + "learning_rate": 1.1547870404077176e-06, + "loss": 0.0065, + "step": 33842 + }, + { + "epoch": 6.15, + "learning_rate": 1.1547324353840553e-06, + "loss": 0.4811, + "step": 33843 + }, + { + "epoch": 6.15, + "learning_rate": 1.1546778303603932e-06, + "loss": 0.0224, + "step": 33844 + }, + { + "epoch": 6.15, + "learning_rate": 1.154623225336731e-06, + "loss": 0.0952, + "step": 33845 + }, + { + "epoch": 6.15, + "learning_rate": 1.154568620313069e-06, + "loss": 0.1772, + "step": 33846 + }, + { + "epoch": 6.16, + "learning_rate": 1.1545140152894066e-06, + "loss": 0.7801, + "step": 33847 + }, + { + "epoch": 6.16, + "learning_rate": 1.1544594102657444e-06, + "loss": 1.8764, + "step": 33848 + }, + { + "epoch": 6.16, + "learning_rate": 1.1544048052420823e-06, + "loss": 0.1935, + "step": 33849 + }, + { + "epoch": 6.16, + "learning_rate": 1.15435020021842e-06, + "loss": 0.3875, + "step": 33850 + }, + { + "epoch": 6.16, + "learning_rate": 1.154295595194758e-06, + "loss": 1.0341, + "step": 33851 + }, + { + "epoch": 6.16, + "learning_rate": 1.1542409901710957e-06, + "loss": 1.0886, + "step": 33852 + }, + { + "epoch": 6.16, + "learning_rate": 1.1541863851474337e-06, + "loss": 0.5344, + "step": 33853 + }, + { + "epoch": 6.16, + "learning_rate": 1.1541317801237714e-06, + "loss": 0.1516, + "step": 33854 + }, + { + "epoch": 6.16, + "learning_rate": 1.1540771751001091e-06, + "loss": 0.0423, + "step": 33855 + }, + { + "epoch": 6.16, + "learning_rate": 1.154022570076447e-06, + "loss": 1.3228, + "step": 33856 + }, + { + "epoch": 6.16, + "learning_rate": 1.1539679650527848e-06, + "loss": 0.6701, + "step": 33857 + }, + { + "epoch": 6.16, + "learning_rate": 1.1539133600291227e-06, + "loss": 0.088, + "step": 33858 + }, + { + "epoch": 6.16, + "learning_rate": 1.1538587550054605e-06, + "loss": 2.9985, + "step": 33859 + }, + { + "epoch": 6.16, + "learning_rate": 1.1538041499817984e-06, + "loss": 0.3115, + "step": 33860 + }, + { + "epoch": 6.16, + "learning_rate": 1.1537495449581361e-06, + "loss": 0.0087, + "step": 33861 + }, + { + "epoch": 6.16, + "learning_rate": 1.1536949399344739e-06, + "loss": 0.1012, + "step": 33862 + }, + { + "epoch": 6.16, + "learning_rate": 1.1536403349108118e-06, + "loss": 0.7976, + "step": 33863 + }, + { + "epoch": 6.16, + "learning_rate": 1.1535857298871498e-06, + "loss": 0.4647, + "step": 33864 + }, + { + "epoch": 6.16, + "learning_rate": 1.1535311248634875e-06, + "loss": 0.0093, + "step": 33865 + }, + { + "epoch": 6.16, + "learning_rate": 1.1534765198398252e-06, + "loss": 0.0058, + "step": 33866 + }, + { + "epoch": 6.16, + "learning_rate": 1.1534219148161632e-06, + "loss": 1.1773, + "step": 33867 + }, + { + "epoch": 6.16, + "learning_rate": 1.153367309792501e-06, + "loss": 1.5241, + "step": 33868 + }, + { + "epoch": 6.16, + "learning_rate": 1.1533127047688386e-06, + "loss": 0.0153, + "step": 33869 + }, + { + "epoch": 6.16, + "learning_rate": 1.1532580997451766e-06, + "loss": 0.051, + "step": 33870 + }, + { + "epoch": 6.16, + "learning_rate": 1.1532034947215145e-06, + "loss": 1.0914, + "step": 33871 + }, + { + "epoch": 6.16, + "learning_rate": 1.1531488896978523e-06, + "loss": 0.0155, + "step": 33872 + }, + { + "epoch": 6.16, + "learning_rate": 1.15309428467419e-06, + "loss": 0.278, + "step": 33873 + }, + { + "epoch": 6.16, + "learning_rate": 1.153039679650528e-06, + "loss": 0.0132, + "step": 33874 + }, + { + "epoch": 6.16, + "learning_rate": 1.1529850746268657e-06, + "loss": 0.0097, + "step": 33875 + }, + { + "epoch": 6.16, + "learning_rate": 1.1529304696032034e-06, + "loss": 0.1299, + "step": 33876 + }, + { + "epoch": 6.16, + "learning_rate": 1.1528758645795413e-06, + "loss": 0.3298, + "step": 33877 + }, + { + "epoch": 6.16, + "learning_rate": 1.1528212595558793e-06, + "loss": 1.2766, + "step": 33878 + }, + { + "epoch": 6.16, + "learning_rate": 1.152766654532217e-06, + "loss": 0.0484, + "step": 33879 + }, + { + "epoch": 6.16, + "learning_rate": 1.1527120495085547e-06, + "loss": 0.2697, + "step": 33880 + }, + { + "epoch": 6.16, + "learning_rate": 1.1526574444848927e-06, + "loss": 0.0479, + "step": 33881 + }, + { + "epoch": 6.16, + "learning_rate": 1.1526028394612304e-06, + "loss": 1.1436, + "step": 33882 + }, + { + "epoch": 6.16, + "learning_rate": 1.1525482344375681e-06, + "loss": 2.3542, + "step": 33883 + }, + { + "epoch": 6.16, + "learning_rate": 1.152493629413906e-06, + "loss": 1.6915, + "step": 33884 + }, + { + "epoch": 6.16, + "learning_rate": 1.152439024390244e-06, + "loss": 0.3207, + "step": 33885 + }, + { + "epoch": 6.16, + "learning_rate": 1.1523844193665818e-06, + "loss": 0.4602, + "step": 33886 + }, + { + "epoch": 6.16, + "learning_rate": 1.1523298143429195e-06, + "loss": 0.5087, + "step": 33887 + }, + { + "epoch": 6.16, + "learning_rate": 1.1522752093192574e-06, + "loss": 1.2733, + "step": 33888 + }, + { + "epoch": 6.16, + "learning_rate": 1.1522206042955952e-06, + "loss": 0.3528, + "step": 33889 + }, + { + "epoch": 6.16, + "learning_rate": 1.152165999271933e-06, + "loss": 1.7325, + "step": 33890 + }, + { + "epoch": 6.16, + "learning_rate": 1.1521113942482708e-06, + "loss": 0.6857, + "step": 33891 + }, + { + "epoch": 6.16, + "learning_rate": 1.1520567892246088e-06, + "loss": 1.452, + "step": 33892 + }, + { + "epoch": 6.16, + "learning_rate": 1.1520021842009465e-06, + "loss": 0.1479, + "step": 33893 + }, + { + "epoch": 6.16, + "learning_rate": 1.1519475791772843e-06, + "loss": 2.7577, + "step": 33894 + }, + { + "epoch": 6.16, + "learning_rate": 1.1518929741536222e-06, + "loss": 0.5303, + "step": 33895 + }, + { + "epoch": 6.16, + "learning_rate": 1.15183836912996e-06, + "loss": 0.395, + "step": 33896 + }, + { + "epoch": 6.16, + "learning_rate": 1.1517837641062977e-06, + "loss": 0.5024, + "step": 33897 + }, + { + "epoch": 6.16, + "learning_rate": 1.1517291590826356e-06, + "loss": 0.3856, + "step": 33898 + }, + { + "epoch": 6.16, + "learning_rate": 1.1516745540589735e-06, + "loss": 0.2522, + "step": 33899 + }, + { + "epoch": 6.16, + "learning_rate": 1.1516199490353113e-06, + "loss": 0.1681, + "step": 33900 + }, + { + "epoch": 6.16, + "learning_rate": 1.151565344011649e-06, + "loss": 0.2574, + "step": 33901 + }, + { + "epoch": 6.17, + "learning_rate": 1.151510738987987e-06, + "loss": 1.0347, + "step": 33902 + }, + { + "epoch": 6.17, + "learning_rate": 1.1514561339643247e-06, + "loss": 0.4713, + "step": 33903 + }, + { + "epoch": 6.17, + "learning_rate": 1.1514015289406626e-06, + "loss": 0.2572, + "step": 33904 + }, + { + "epoch": 6.17, + "learning_rate": 1.1513469239170006e-06, + "loss": 0.0115, + "step": 33905 + }, + { + "epoch": 6.17, + "learning_rate": 1.1512923188933383e-06, + "loss": 0.2167, + "step": 33906 + }, + { + "epoch": 6.17, + "learning_rate": 1.151237713869676e-06, + "loss": 1.4984, + "step": 33907 + }, + { + "epoch": 6.17, + "learning_rate": 1.1511831088460138e-06, + "loss": 1.7404, + "step": 33908 + }, + { + "epoch": 6.17, + "learning_rate": 1.1511285038223517e-06, + "loss": 0.211, + "step": 33909 + }, + { + "epoch": 6.17, + "learning_rate": 1.1510738987986894e-06, + "loss": 0.3609, + "step": 33910 + }, + { + "epoch": 6.17, + "learning_rate": 1.1510192937750274e-06, + "loss": 0.262, + "step": 33911 + }, + { + "epoch": 6.17, + "learning_rate": 1.1509646887513653e-06, + "loss": 0.2471, + "step": 33912 + }, + { + "epoch": 6.17, + "learning_rate": 1.150910083727703e-06, + "loss": 1.2534, + "step": 33913 + }, + { + "epoch": 6.17, + "learning_rate": 1.1508554787040408e-06, + "loss": 1.1413, + "step": 33914 + }, + { + "epoch": 6.17, + "learning_rate": 1.1508008736803785e-06, + "loss": 0.1256, + "step": 33915 + }, + { + "epoch": 6.17, + "learning_rate": 1.1507462686567165e-06, + "loss": 0.0622, + "step": 33916 + }, + { + "epoch": 6.17, + "learning_rate": 1.1506916636330542e-06, + "loss": 0.5521, + "step": 33917 + }, + { + "epoch": 6.17, + "learning_rate": 1.1506370586093921e-06, + "loss": 0.4481, + "step": 33918 + }, + { + "epoch": 6.17, + "learning_rate": 1.15058245358573e-06, + "loss": 0.6896, + "step": 33919 + }, + { + "epoch": 6.17, + "learning_rate": 1.1505278485620678e-06, + "loss": 0.6455, + "step": 33920 + }, + { + "epoch": 6.17, + "learning_rate": 1.1504732435384056e-06, + "loss": 0.9911, + "step": 33921 + }, + { + "epoch": 6.17, + "learning_rate": 1.1504186385147433e-06, + "loss": 0.1119, + "step": 33922 + }, + { + "epoch": 6.17, + "learning_rate": 1.1503640334910812e-06, + "loss": 0.5207, + "step": 33923 + }, + { + "epoch": 6.17, + "learning_rate": 1.150309428467419e-06, + "loss": 0.0953, + "step": 33924 + }, + { + "epoch": 6.17, + "learning_rate": 1.150254823443757e-06, + "loss": 0.5708, + "step": 33925 + }, + { + "epoch": 6.17, + "learning_rate": 1.1502002184200946e-06, + "loss": 0.5325, + "step": 33926 + }, + { + "epoch": 6.17, + "learning_rate": 1.1501456133964326e-06, + "loss": 0.4784, + "step": 33927 + }, + { + "epoch": 6.17, + "learning_rate": 1.1500910083727703e-06, + "loss": 1.2349, + "step": 33928 + }, + { + "epoch": 6.17, + "learning_rate": 1.150036403349108e-06, + "loss": 0.2675, + "step": 33929 + }, + { + "epoch": 6.17, + "learning_rate": 1.149981798325446e-06, + "loss": 0.0216, + "step": 33930 + }, + { + "epoch": 6.17, + "learning_rate": 1.1499271933017837e-06, + "loss": 1.8274, + "step": 33931 + }, + { + "epoch": 6.17, + "learning_rate": 1.1498725882781217e-06, + "loss": 0.1757, + "step": 33932 + }, + { + "epoch": 6.17, + "learning_rate": 1.1498179832544594e-06, + "loss": 0.3635, + "step": 33933 + }, + { + "epoch": 6.17, + "learning_rate": 1.1497633782307973e-06, + "loss": 0.3276, + "step": 33934 + }, + { + "epoch": 6.17, + "learning_rate": 1.149708773207135e-06, + "loss": 0.9173, + "step": 33935 + }, + { + "epoch": 6.17, + "learning_rate": 1.1496541681834728e-06, + "loss": 0.1086, + "step": 33936 + }, + { + "epoch": 6.17, + "learning_rate": 1.1495995631598107e-06, + "loss": 0.0078, + "step": 33937 + }, + { + "epoch": 6.17, + "learning_rate": 1.1495449581361485e-06, + "loss": 1.8274, + "step": 33938 + }, + { + "epoch": 6.17, + "learning_rate": 1.1494903531124864e-06, + "loss": 0.8413, + "step": 33939 + }, + { + "epoch": 6.17, + "learning_rate": 1.1494357480888241e-06, + "loss": 1.4614, + "step": 33940 + }, + { + "epoch": 6.17, + "learning_rate": 1.149381143065162e-06, + "loss": 2.2404, + "step": 33941 + }, + { + "epoch": 6.17, + "learning_rate": 1.1493265380414998e-06, + "loss": 0.4025, + "step": 33942 + }, + { + "epoch": 6.17, + "learning_rate": 1.1492719330178376e-06, + "loss": 0.3793, + "step": 33943 + }, + { + "epoch": 6.17, + "learning_rate": 1.1492173279941755e-06, + "loss": 0.1208, + "step": 33944 + }, + { + "epoch": 6.17, + "learning_rate": 1.1491627229705134e-06, + "loss": 0.8671, + "step": 33945 + }, + { + "epoch": 6.17, + "learning_rate": 1.1491081179468512e-06, + "loss": 0.0128, + "step": 33946 + }, + { + "epoch": 6.17, + "learning_rate": 1.149053512923189e-06, + "loss": 1.6338, + "step": 33947 + }, + { + "epoch": 6.17, + "learning_rate": 1.1489989078995268e-06, + "loss": 0.0143, + "step": 33948 + }, + { + "epoch": 6.17, + "learning_rate": 1.1489443028758646e-06, + "loss": 0.2888, + "step": 33949 + }, + { + "epoch": 6.17, + "learning_rate": 1.1488896978522023e-06, + "loss": 0.9651, + "step": 33950 + }, + { + "epoch": 6.17, + "learning_rate": 1.1488350928285403e-06, + "loss": 0.0081, + "step": 33951 + }, + { + "epoch": 6.17, + "learning_rate": 1.1487804878048782e-06, + "loss": 1.0679, + "step": 33952 + }, + { + "epoch": 6.17, + "learning_rate": 1.148725882781216e-06, + "loss": 1.5889, + "step": 33953 + }, + { + "epoch": 6.17, + "learning_rate": 1.1486712777575537e-06, + "loss": 0.0076, + "step": 33954 + }, + { + "epoch": 6.17, + "learning_rate": 1.1486166727338916e-06, + "loss": 0.6921, + "step": 33955 + }, + { + "epoch": 6.17, + "learning_rate": 1.1485620677102293e-06, + "loss": 0.4623, + "step": 33956 + }, + { + "epoch": 6.18, + "learning_rate": 1.148507462686567e-06, + "loss": 0.0122, + "step": 33957 + }, + { + "epoch": 6.18, + "learning_rate": 1.148452857662905e-06, + "loss": 1.6626, + "step": 33958 + }, + { + "epoch": 6.18, + "learning_rate": 1.148398252639243e-06, + "loss": 0.3562, + "step": 33959 + }, + { + "epoch": 6.18, + "learning_rate": 1.1483436476155807e-06, + "loss": 0.3376, + "step": 33960 + }, + { + "epoch": 6.18, + "learning_rate": 1.1482890425919184e-06, + "loss": 0.2804, + "step": 33961 + }, + { + "epoch": 6.18, + "learning_rate": 1.1482344375682564e-06, + "loss": 2.4168, + "step": 33962 + }, + { + "epoch": 6.18, + "learning_rate": 1.148179832544594e-06, + "loss": 0.0197, + "step": 33963 + }, + { + "epoch": 6.18, + "learning_rate": 1.1481252275209318e-06, + "loss": 0.5223, + "step": 33964 + }, + { + "epoch": 6.18, + "learning_rate": 1.1480706224972698e-06, + "loss": 0.3173, + "step": 33965 + }, + { + "epoch": 6.18, + "learning_rate": 1.1480160174736077e-06, + "loss": 0.0049, + "step": 33966 + }, + { + "epoch": 6.18, + "learning_rate": 1.1479614124499454e-06, + "loss": 0.0074, + "step": 33967 + }, + { + "epoch": 6.18, + "learning_rate": 1.1479068074262832e-06, + "loss": 0.4974, + "step": 33968 + }, + { + "epoch": 6.18, + "learning_rate": 1.1478522024026211e-06, + "loss": 0.0031, + "step": 33969 + }, + { + "epoch": 6.18, + "learning_rate": 1.1477975973789588e-06, + "loss": 0.0463, + "step": 33970 + }, + { + "epoch": 6.18, + "learning_rate": 1.1477429923552966e-06, + "loss": 0.7248, + "step": 33971 + }, + { + "epoch": 6.18, + "learning_rate": 1.1476883873316345e-06, + "loss": 0.0198, + "step": 33972 + }, + { + "epoch": 6.18, + "learning_rate": 1.1476337823079725e-06, + "loss": 0.3318, + "step": 33973 + }, + { + "epoch": 6.18, + "learning_rate": 1.1475791772843102e-06, + "loss": 1.2248, + "step": 33974 + }, + { + "epoch": 6.18, + "learning_rate": 1.147524572260648e-06, + "loss": 0.4238, + "step": 33975 + }, + { + "epoch": 6.18, + "learning_rate": 1.1474699672369859e-06, + "loss": 0.0106, + "step": 33976 + }, + { + "epoch": 6.18, + "learning_rate": 1.1474153622133236e-06, + "loss": 0.3051, + "step": 33977 + }, + { + "epoch": 6.18, + "learning_rate": 1.1473607571896613e-06, + "loss": 0.3161, + "step": 33978 + }, + { + "epoch": 6.18, + "learning_rate": 1.1473061521659993e-06, + "loss": 0.2508, + "step": 33979 + }, + { + "epoch": 6.18, + "learning_rate": 1.1472515471423372e-06, + "loss": 0.3385, + "step": 33980 + }, + { + "epoch": 6.18, + "learning_rate": 1.147196942118675e-06, + "loss": 1.3232, + "step": 33981 + }, + { + "epoch": 6.18, + "learning_rate": 1.1471423370950127e-06, + "loss": 0.0091, + "step": 33982 + }, + { + "epoch": 6.18, + "learning_rate": 1.1470877320713506e-06, + "loss": 0.3325, + "step": 33983 + }, + { + "epoch": 6.18, + "learning_rate": 1.1470331270476884e-06, + "loss": 1.9506, + "step": 33984 + }, + { + "epoch": 6.18, + "learning_rate": 1.1469785220240263e-06, + "loss": 0.0764, + "step": 33985 + }, + { + "epoch": 6.18, + "learning_rate": 1.146923917000364e-06, + "loss": 1.6866, + "step": 33986 + }, + { + "epoch": 6.18, + "learning_rate": 1.146869311976702e-06, + "loss": 0.3434, + "step": 33987 + }, + { + "epoch": 6.18, + "learning_rate": 1.1468147069530397e-06, + "loss": 0.609, + "step": 33988 + }, + { + "epoch": 6.18, + "learning_rate": 1.1467601019293774e-06, + "loss": 0.026, + "step": 33989 + }, + { + "epoch": 6.18, + "learning_rate": 1.1467054969057154e-06, + "loss": 0.8373, + "step": 33990 + }, + { + "epoch": 6.18, + "learning_rate": 1.1466508918820531e-06, + "loss": 0.0386, + "step": 33991 + }, + { + "epoch": 6.18, + "learning_rate": 1.146596286858391e-06, + "loss": 0.0062, + "step": 33992 + }, + { + "epoch": 6.18, + "learning_rate": 1.146541681834729e-06, + "loss": 0.1857, + "step": 33993 + }, + { + "epoch": 6.18, + "learning_rate": 1.1464870768110667e-06, + "loss": 1.0248, + "step": 33994 + }, + { + "epoch": 6.18, + "learning_rate": 1.1464324717874045e-06, + "loss": 0.0059, + "step": 33995 + }, + { + "epoch": 6.18, + "learning_rate": 1.1463778667637422e-06, + "loss": 0.0119, + "step": 33996 + }, + { + "epoch": 6.18, + "learning_rate": 1.1463232617400801e-06, + "loss": 0.4862, + "step": 33997 + }, + { + "epoch": 6.18, + "learning_rate": 1.1462686567164179e-06, + "loss": 0.0127, + "step": 33998 + }, + { + "epoch": 6.18, + "learning_rate": 1.1462140516927558e-06, + "loss": 0.9054, + "step": 33999 + }, + { + "epoch": 6.18, + "learning_rate": 1.1461594466690935e-06, + "loss": 0.108, + "step": 34000 + }, + { + "epoch": 6.18, + "learning_rate": 1.1461048416454315e-06, + "loss": 0.0397, + "step": 34001 + }, + { + "epoch": 6.18, + "learning_rate": 1.1460502366217692e-06, + "loss": 0.357, + "step": 34002 + }, + { + "epoch": 6.18, + "learning_rate": 1.145995631598107e-06, + "loss": 0.4651, + "step": 34003 + }, + { + "epoch": 6.18, + "learning_rate": 1.145941026574445e-06, + "loss": 0.0046, + "step": 34004 + }, + { + "epoch": 6.18, + "learning_rate": 1.1458864215507826e-06, + "loss": 0.032, + "step": 34005 + }, + { + "epoch": 6.18, + "learning_rate": 1.1458318165271206e-06, + "loss": 0.555, + "step": 34006 + }, + { + "epoch": 6.18, + "learning_rate": 1.1457772115034583e-06, + "loss": 0.4906, + "step": 34007 + }, + { + "epoch": 6.18, + "learning_rate": 1.1457226064797962e-06, + "loss": 0.0226, + "step": 34008 + }, + { + "epoch": 6.18, + "learning_rate": 1.145668001456134e-06, + "loss": 1.4458, + "step": 34009 + }, + { + "epoch": 6.18, + "learning_rate": 1.1456133964324717e-06, + "loss": 0.0139, + "step": 34010 + }, + { + "epoch": 6.18, + "learning_rate": 1.1455587914088097e-06, + "loss": 0.3626, + "step": 34011 + }, + { + "epoch": 6.19, + "learning_rate": 1.1455041863851474e-06, + "loss": 0.0052, + "step": 34012 + }, + { + "epoch": 6.19, + "learning_rate": 1.1454495813614853e-06, + "loss": 0.2943, + "step": 34013 + }, + { + "epoch": 6.19, + "learning_rate": 1.145394976337823e-06, + "loss": 0.012, + "step": 34014 + }, + { + "epoch": 6.19, + "learning_rate": 1.145340371314161e-06, + "loss": 0.2973, + "step": 34015 + }, + { + "epoch": 6.19, + "learning_rate": 1.1452857662904987e-06, + "loss": 0.0168, + "step": 34016 + }, + { + "epoch": 6.19, + "learning_rate": 1.1452311612668365e-06, + "loss": 0.1921, + "step": 34017 + }, + { + "epoch": 6.19, + "learning_rate": 1.1451765562431744e-06, + "loss": 0.27, + "step": 34018 + }, + { + "epoch": 6.19, + "learning_rate": 1.1451219512195121e-06, + "loss": 0.1742, + "step": 34019 + }, + { + "epoch": 6.19, + "learning_rate": 1.14506734619585e-06, + "loss": 0.2876, + "step": 34020 + }, + { + "epoch": 6.19, + "learning_rate": 1.1450127411721878e-06, + "loss": 0.0216, + "step": 34021 + }, + { + "epoch": 6.19, + "learning_rate": 1.1449581361485258e-06, + "loss": 0.0062, + "step": 34022 + }, + { + "epoch": 6.19, + "learning_rate": 1.1449035311248635e-06, + "loss": 0.0076, + "step": 34023 + }, + { + "epoch": 6.19, + "learning_rate": 1.1448489261012012e-06, + "loss": 0.0353, + "step": 34024 + }, + { + "epoch": 6.19, + "learning_rate": 1.1447943210775392e-06, + "loss": 0.0611, + "step": 34025 + }, + { + "epoch": 6.19, + "learning_rate": 1.144739716053877e-06, + "loss": 2.059, + "step": 34026 + }, + { + "epoch": 6.19, + "learning_rate": 1.1446851110302148e-06, + "loss": 0.1701, + "step": 34027 + }, + { + "epoch": 6.19, + "learning_rate": 1.1446305060065526e-06, + "loss": 0.0145, + "step": 34028 + }, + { + "epoch": 6.19, + "learning_rate": 1.1445759009828905e-06, + "loss": 0.5809, + "step": 34029 + }, + { + "epoch": 6.19, + "learning_rate": 1.1445212959592283e-06, + "loss": 0.1425, + "step": 34030 + }, + { + "epoch": 6.19, + "learning_rate": 1.144466690935566e-06, + "loss": 1.4567, + "step": 34031 + }, + { + "epoch": 6.19, + "learning_rate": 1.144412085911904e-06, + "loss": 0.0193, + "step": 34032 + }, + { + "epoch": 6.19, + "learning_rate": 1.1443574808882419e-06, + "loss": 0.2646, + "step": 34033 + }, + { + "epoch": 6.19, + "learning_rate": 1.1443028758645796e-06, + "loss": 0.017, + "step": 34034 + }, + { + "epoch": 6.19, + "learning_rate": 1.1442482708409173e-06, + "loss": 0.3623, + "step": 34035 + }, + { + "epoch": 6.19, + "learning_rate": 1.1441936658172553e-06, + "loss": 0.0205, + "step": 34036 + }, + { + "epoch": 6.19, + "learning_rate": 1.144139060793593e-06, + "loss": 2.6793, + "step": 34037 + }, + { + "epoch": 6.19, + "learning_rate": 1.1440844557699307e-06, + "loss": 0.008, + "step": 34038 + }, + { + "epoch": 6.19, + "learning_rate": 1.1440298507462687e-06, + "loss": 0.3982, + "step": 34039 + }, + { + "epoch": 6.19, + "learning_rate": 1.1439752457226066e-06, + "loss": 0.3116, + "step": 34040 + }, + { + "epoch": 6.19, + "learning_rate": 1.1439206406989444e-06, + "loss": 0.2629, + "step": 34041 + }, + { + "epoch": 6.19, + "learning_rate": 1.143866035675282e-06, + "loss": 0.0091, + "step": 34042 + }, + { + "epoch": 6.19, + "learning_rate": 1.14381143065162e-06, + "loss": 0.0275, + "step": 34043 + }, + { + "epoch": 6.19, + "learning_rate": 1.1437568256279578e-06, + "loss": 2.9561, + "step": 34044 + }, + { + "epoch": 6.19, + "learning_rate": 1.1437022206042955e-06, + "loss": 0.0862, + "step": 34045 + }, + { + "epoch": 6.19, + "learning_rate": 1.1436476155806334e-06, + "loss": 0.0055, + "step": 34046 + }, + { + "epoch": 6.19, + "learning_rate": 1.1435930105569714e-06, + "loss": 0.1829, + "step": 34047 + }, + { + "epoch": 6.19, + "learning_rate": 1.1435384055333091e-06, + "loss": 0.042, + "step": 34048 + }, + { + "epoch": 6.19, + "learning_rate": 1.1434838005096468e-06, + "loss": 0.9623, + "step": 34049 + }, + { + "epoch": 6.19, + "learning_rate": 1.1434291954859848e-06, + "loss": 0.0919, + "step": 34050 + }, + { + "epoch": 6.19, + "learning_rate": 1.1433745904623225e-06, + "loss": 0.3451, + "step": 34051 + }, + { + "epoch": 6.19, + "learning_rate": 1.1433199854386603e-06, + "loss": 0.5035, + "step": 34052 + }, + { + "epoch": 6.19, + "learning_rate": 1.1432653804149982e-06, + "loss": 0.0072, + "step": 34053 + }, + { + "epoch": 6.19, + "learning_rate": 1.1432107753913361e-06, + "loss": 0.0892, + "step": 34054 + }, + { + "epoch": 6.19, + "learning_rate": 1.1431561703676739e-06, + "loss": 1.1645, + "step": 34055 + }, + { + "epoch": 6.19, + "learning_rate": 1.1431015653440116e-06, + "loss": 0.0406, + "step": 34056 + }, + { + "epoch": 6.19, + "learning_rate": 1.1430469603203495e-06, + "loss": 0.0972, + "step": 34057 + }, + { + "epoch": 6.19, + "learning_rate": 1.1429923552966873e-06, + "loss": 0.8669, + "step": 34058 + }, + { + "epoch": 6.19, + "learning_rate": 1.142937750273025e-06, + "loss": 0.0512, + "step": 34059 + }, + { + "epoch": 6.19, + "learning_rate": 1.142883145249363e-06, + "loss": 0.0071, + "step": 34060 + }, + { + "epoch": 6.19, + "learning_rate": 1.142828540225701e-06, + "loss": 0.5157, + "step": 34061 + }, + { + "epoch": 6.19, + "learning_rate": 1.1427739352020386e-06, + "loss": 0.7111, + "step": 34062 + }, + { + "epoch": 6.19, + "learning_rate": 1.1427193301783764e-06, + "loss": 0.2752, + "step": 34063 + }, + { + "epoch": 6.19, + "learning_rate": 1.1426647251547143e-06, + "loss": 0.6878, + "step": 34064 + }, + { + "epoch": 6.19, + "learning_rate": 1.142610120131052e-06, + "loss": 0.2582, + "step": 34065 + }, + { + "epoch": 6.19, + "learning_rate": 1.1425555151073898e-06, + "loss": 0.1821, + "step": 34066 + }, + { + "epoch": 6.2, + "learning_rate": 1.1425009100837277e-06, + "loss": 0.024, + "step": 34067 + }, + { + "epoch": 6.2, + "learning_rate": 1.1424463050600657e-06, + "loss": 0.0076, + "step": 34068 + }, + { + "epoch": 6.2, + "learning_rate": 1.1423917000364034e-06, + "loss": 1.4048, + "step": 34069 + }, + { + "epoch": 6.2, + "learning_rate": 1.1423370950127411e-06, + "loss": 0.1266, + "step": 34070 + }, + { + "epoch": 6.2, + "learning_rate": 1.142282489989079e-06, + "loss": 0.0699, + "step": 34071 + }, + { + "epoch": 6.2, + "learning_rate": 1.1422278849654168e-06, + "loss": 0.0041, + "step": 34072 + }, + { + "epoch": 6.2, + "learning_rate": 1.1421732799417547e-06, + "loss": 1.2177, + "step": 34073 + }, + { + "epoch": 6.2, + "learning_rate": 1.1421186749180927e-06, + "loss": 1.5738, + "step": 34074 + }, + { + "epoch": 6.2, + "learning_rate": 1.1420640698944304e-06, + "loss": 2.4071, + "step": 34075 + }, + { + "epoch": 6.2, + "learning_rate": 1.1420094648707681e-06, + "loss": 0.0138, + "step": 34076 + }, + { + "epoch": 6.2, + "learning_rate": 1.1419548598471059e-06, + "loss": 0.0341, + "step": 34077 + }, + { + "epoch": 6.2, + "learning_rate": 1.1419002548234438e-06, + "loss": 0.3721, + "step": 34078 + }, + { + "epoch": 6.2, + "learning_rate": 1.1418456497997815e-06, + "loss": 0.9372, + "step": 34079 + }, + { + "epoch": 6.2, + "learning_rate": 1.1417910447761195e-06, + "loss": 0.9875, + "step": 34080 + }, + { + "epoch": 6.2, + "learning_rate": 1.1417364397524572e-06, + "loss": 0.7562, + "step": 34081 + }, + { + "epoch": 6.2, + "learning_rate": 1.1416818347287952e-06, + "loss": 0.1368, + "step": 34082 + }, + { + "epoch": 6.2, + "learning_rate": 1.141627229705133e-06, + "loss": 0.35, + "step": 34083 + }, + { + "epoch": 6.2, + "learning_rate": 1.1415726246814706e-06, + "loss": 0.4234, + "step": 34084 + }, + { + "epoch": 6.2, + "learning_rate": 1.1415180196578086e-06, + "loss": 0.1025, + "step": 34085 + }, + { + "epoch": 6.2, + "learning_rate": 1.1414634146341463e-06, + "loss": 1.099, + "step": 34086 + }, + { + "epoch": 6.2, + "learning_rate": 1.1414088096104842e-06, + "loss": 2.5252, + "step": 34087 + }, + { + "epoch": 6.2, + "learning_rate": 1.141354204586822e-06, + "loss": 0.6406, + "step": 34088 + }, + { + "epoch": 6.2, + "learning_rate": 1.14129959956316e-06, + "loss": 0.7367, + "step": 34089 + }, + { + "epoch": 6.2, + "learning_rate": 1.1412449945394977e-06, + "loss": 0.1789, + "step": 34090 + }, + { + "epoch": 6.2, + "learning_rate": 1.1411903895158354e-06, + "loss": 0.653, + "step": 34091 + }, + { + "epoch": 6.2, + "learning_rate": 1.1411357844921733e-06, + "loss": 0.3125, + "step": 34092 + }, + { + "epoch": 6.2, + "learning_rate": 1.141081179468511e-06, + "loss": 0.8206, + "step": 34093 + }, + { + "epoch": 6.2, + "learning_rate": 1.141026574444849e-06, + "loss": 2.5238, + "step": 34094 + }, + { + "epoch": 6.2, + "learning_rate": 1.1409719694211867e-06, + "loss": 0.29, + "step": 34095 + }, + { + "epoch": 6.2, + "learning_rate": 1.1409173643975247e-06, + "loss": 3.4089, + "step": 34096 + }, + { + "epoch": 6.2, + "learning_rate": 1.1408627593738624e-06, + "loss": 0.117, + "step": 34097 + }, + { + "epoch": 6.2, + "learning_rate": 1.1408081543502001e-06, + "loss": 0.0042, + "step": 34098 + }, + { + "epoch": 6.2, + "learning_rate": 1.140753549326538e-06, + "loss": 0.2216, + "step": 34099 + }, + { + "epoch": 6.2, + "learning_rate": 1.1406989443028758e-06, + "loss": 2.2683, + "step": 34100 + }, + { + "epoch": 6.2, + "learning_rate": 1.1406443392792138e-06, + "loss": 0.4867, + "step": 34101 + }, + { + "epoch": 6.2, + "learning_rate": 1.1405897342555515e-06, + "loss": 0.075, + "step": 34102 + }, + { + "epoch": 6.2, + "learning_rate": 1.1405351292318894e-06, + "loss": 0.2868, + "step": 34103 + }, + { + "epoch": 6.2, + "learning_rate": 1.1404805242082272e-06, + "loss": 3.0368, + "step": 34104 + }, + { + "epoch": 6.2, + "learning_rate": 1.140425919184565e-06, + "loss": 0.0126, + "step": 34105 + }, + { + "epoch": 6.2, + "learning_rate": 1.1403713141609028e-06, + "loss": 3.561, + "step": 34106 + }, + { + "epoch": 6.2, + "learning_rate": 1.1403167091372406e-06, + "loss": 0.9422, + "step": 34107 + }, + { + "epoch": 6.2, + "learning_rate": 1.1402621041135785e-06, + "loss": 0.0305, + "step": 34108 + }, + { + "epoch": 6.2, + "learning_rate": 1.1402074990899163e-06, + "loss": 0.2047, + "step": 34109 + }, + { + "epoch": 6.2, + "learning_rate": 1.1401528940662542e-06, + "loss": 1.4359, + "step": 34110 + }, + { + "epoch": 6.2, + "learning_rate": 1.140098289042592e-06, + "loss": 0.0411, + "step": 34111 + }, + { + "epoch": 6.2, + "learning_rate": 1.1400436840189297e-06, + "loss": 1.7136, + "step": 34112 + }, + { + "epoch": 6.2, + "learning_rate": 1.1399890789952676e-06, + "loss": 0.4209, + "step": 34113 + }, + { + "epoch": 6.2, + "learning_rate": 1.1399344739716055e-06, + "loss": 0.1153, + "step": 34114 + }, + { + "epoch": 6.2, + "learning_rate": 1.1398798689479433e-06, + "loss": 0.1099, + "step": 34115 + }, + { + "epoch": 6.2, + "learning_rate": 1.139825263924281e-06, + "loss": 2.0263, + "step": 34116 + }, + { + "epoch": 6.2, + "learning_rate": 1.139770658900619e-06, + "loss": 0.2849, + "step": 34117 + }, + { + "epoch": 6.2, + "learning_rate": 1.1397160538769567e-06, + "loss": 0.8883, + "step": 34118 + }, + { + "epoch": 6.2, + "learning_rate": 1.1396614488532944e-06, + "loss": 0.2503, + "step": 34119 + }, + { + "epoch": 6.2, + "learning_rate": 1.1396068438296324e-06, + "loss": 0.9877, + "step": 34120 + }, + { + "epoch": 6.2, + "learning_rate": 1.1395522388059703e-06, + "loss": 0.0264, + "step": 34121 + }, + { + "epoch": 6.21, + "learning_rate": 1.139497633782308e-06, + "loss": 1.1826, + "step": 34122 + }, + { + "epoch": 6.21, + "learning_rate": 1.1394430287586458e-06, + "loss": 0.6776, + "step": 34123 + }, + { + "epoch": 6.21, + "learning_rate": 1.1393884237349837e-06, + "loss": 1.14, + "step": 34124 + }, + { + "epoch": 6.21, + "learning_rate": 1.1393338187113214e-06, + "loss": 0.3078, + "step": 34125 + }, + { + "epoch": 6.21, + "learning_rate": 1.1392792136876592e-06, + "loss": 0.5664, + "step": 34126 + }, + { + "epoch": 6.21, + "learning_rate": 1.1392246086639971e-06, + "loss": 0.9297, + "step": 34127 + }, + { + "epoch": 6.21, + "learning_rate": 1.139170003640335e-06, + "loss": 1.2156, + "step": 34128 + }, + { + "epoch": 6.21, + "learning_rate": 1.1391153986166728e-06, + "loss": 0.6063, + "step": 34129 + }, + { + "epoch": 6.21, + "learning_rate": 1.1390607935930105e-06, + "loss": 0.1122, + "step": 34130 + }, + { + "epoch": 6.21, + "learning_rate": 1.1390061885693485e-06, + "loss": 0.0304, + "step": 34131 + }, + { + "epoch": 6.21, + "learning_rate": 1.1389515835456862e-06, + "loss": 0.8423, + "step": 34132 + }, + { + "epoch": 6.21, + "learning_rate": 1.138896978522024e-06, + "loss": 0.0184, + "step": 34133 + }, + { + "epoch": 6.21, + "learning_rate": 1.1388423734983619e-06, + "loss": 2.4167, + "step": 34134 + }, + { + "epoch": 6.21, + "learning_rate": 1.1387877684746998e-06, + "loss": 2.5497, + "step": 34135 + }, + { + "epoch": 6.21, + "learning_rate": 1.1387331634510375e-06, + "loss": 4.1662, + "step": 34136 + }, + { + "epoch": 6.21, + "learning_rate": 1.1386785584273753e-06, + "loss": 0.4011, + "step": 34137 + }, + { + "epoch": 6.21, + "learning_rate": 1.1386239534037132e-06, + "loss": 0.0837, + "step": 34138 + }, + { + "epoch": 6.21, + "learning_rate": 1.138569348380051e-06, + "loss": 0.4778, + "step": 34139 + }, + { + "epoch": 6.21, + "learning_rate": 1.1385147433563887e-06, + "loss": 0.2063, + "step": 34140 + }, + { + "epoch": 6.21, + "learning_rate": 1.1384601383327266e-06, + "loss": 0.1995, + "step": 34141 + }, + { + "epoch": 6.21, + "learning_rate": 1.1384055333090646e-06, + "loss": 0.246, + "step": 34142 + }, + { + "epoch": 6.21, + "learning_rate": 1.1383509282854023e-06, + "loss": 1.2696, + "step": 34143 + }, + { + "epoch": 6.21, + "learning_rate": 1.13829632326174e-06, + "loss": 0.1551, + "step": 34144 + }, + { + "epoch": 6.21, + "learning_rate": 1.138241718238078e-06, + "loss": 0.4334, + "step": 34145 + }, + { + "epoch": 6.21, + "learning_rate": 1.1381871132144157e-06, + "loss": 0.9989, + "step": 34146 + }, + { + "epoch": 6.21, + "learning_rate": 1.1381325081907534e-06, + "loss": 1.2943, + "step": 34147 + }, + { + "epoch": 6.21, + "learning_rate": 1.1380779031670914e-06, + "loss": 1.7332, + "step": 34148 + }, + { + "epoch": 6.21, + "learning_rate": 1.1380232981434293e-06, + "loss": 0.2533, + "step": 34149 + }, + { + "epoch": 6.21, + "learning_rate": 1.137968693119767e-06, + "loss": 0.2287, + "step": 34150 + }, + { + "epoch": 6.21, + "learning_rate": 1.1379140880961048e-06, + "loss": 0.3973, + "step": 34151 + }, + { + "epoch": 6.21, + "learning_rate": 1.1378594830724427e-06, + "loss": 0.015, + "step": 34152 + }, + { + "epoch": 6.21, + "learning_rate": 1.1378048780487805e-06, + "loss": 0.4673, + "step": 34153 + }, + { + "epoch": 6.21, + "learning_rate": 1.1377502730251184e-06, + "loss": 0.0865, + "step": 34154 + }, + { + "epoch": 6.21, + "learning_rate": 1.1376956680014561e-06, + "loss": 2.0394, + "step": 34155 + }, + { + "epoch": 6.21, + "learning_rate": 1.137641062977794e-06, + "loss": 3.1846, + "step": 34156 + }, + { + "epoch": 6.21, + "learning_rate": 1.1375864579541318e-06, + "loss": 0.0794, + "step": 34157 + }, + { + "epoch": 6.21, + "learning_rate": 1.1375318529304695e-06, + "loss": 0.0676, + "step": 34158 + }, + { + "epoch": 6.21, + "learning_rate": 1.1374772479068075e-06, + "loss": 3.5027, + "step": 34159 + }, + { + "epoch": 6.21, + "learning_rate": 1.1374226428831452e-06, + "loss": 0.3031, + "step": 34160 + }, + { + "epoch": 6.21, + "learning_rate": 1.1373680378594832e-06, + "loss": 0.077, + "step": 34161 + }, + { + "epoch": 6.21, + "learning_rate": 1.137313432835821e-06, + "loss": 0.1672, + "step": 34162 + }, + { + "epoch": 6.21, + "learning_rate": 1.1372588278121588e-06, + "loss": 0.0088, + "step": 34163 + }, + { + "epoch": 6.21, + "learning_rate": 1.1372042227884966e-06, + "loss": 1.4912, + "step": 34164 + }, + { + "epoch": 6.21, + "learning_rate": 1.1371496177648343e-06, + "loss": 0.0648, + "step": 34165 + }, + { + "epoch": 6.21, + "learning_rate": 1.1370950127411722e-06, + "loss": 1.6797, + "step": 34166 + }, + { + "epoch": 6.21, + "learning_rate": 1.13704040771751e-06, + "loss": 0.167, + "step": 34167 + }, + { + "epoch": 6.21, + "learning_rate": 1.136985802693848e-06, + "loss": 1.2285, + "step": 34168 + }, + { + "epoch": 6.21, + "learning_rate": 1.1369311976701857e-06, + "loss": 0.8621, + "step": 34169 + }, + { + "epoch": 6.21, + "learning_rate": 1.1368765926465236e-06, + "loss": 0.3845, + "step": 34170 + }, + { + "epoch": 6.21, + "learning_rate": 1.1368219876228613e-06, + "loss": 0.8588, + "step": 34171 + }, + { + "epoch": 6.21, + "learning_rate": 1.136767382599199e-06, + "loss": 0.046, + "step": 34172 + }, + { + "epoch": 6.21, + "learning_rate": 1.136712777575537e-06, + "loss": 0.5585, + "step": 34173 + }, + { + "epoch": 6.21, + "learning_rate": 1.1366581725518747e-06, + "loss": 0.428, + "step": 34174 + }, + { + "epoch": 6.21, + "learning_rate": 1.1366035675282127e-06, + "loss": 0.5164, + "step": 34175 + }, + { + "epoch": 6.21, + "learning_rate": 1.1365489625045504e-06, + "loss": 2.0604, + "step": 34176 + }, + { + "epoch": 6.22, + "learning_rate": 1.1364943574808884e-06, + "loss": 0.1922, + "step": 34177 + }, + { + "epoch": 6.22, + "learning_rate": 1.136439752457226e-06, + "loss": 0.7391, + "step": 34178 + }, + { + "epoch": 6.22, + "learning_rate": 1.1363851474335638e-06, + "loss": 0.3922, + "step": 34179 + }, + { + "epoch": 6.22, + "learning_rate": 1.1363305424099018e-06, + "loss": 0.4582, + "step": 34180 + }, + { + "epoch": 6.22, + "learning_rate": 1.1362759373862395e-06, + "loss": 0.6575, + "step": 34181 + }, + { + "epoch": 6.22, + "learning_rate": 1.1362213323625774e-06, + "loss": 0.0087, + "step": 34182 + }, + { + "epoch": 6.22, + "learning_rate": 1.1361667273389152e-06, + "loss": 0.0041, + "step": 34183 + }, + { + "epoch": 6.22, + "learning_rate": 1.1361121223152531e-06, + "loss": 0.0042, + "step": 34184 + }, + { + "epoch": 6.22, + "learning_rate": 1.1360575172915908e-06, + "loss": 0.5177, + "step": 34185 + }, + { + "epoch": 6.22, + "learning_rate": 1.1360029122679286e-06, + "loss": 0.7088, + "step": 34186 + }, + { + "epoch": 6.22, + "learning_rate": 1.1359483072442665e-06, + "loss": 0.6725, + "step": 34187 + }, + { + "epoch": 6.22, + "learning_rate": 1.1358937022206042e-06, + "loss": 0.021, + "step": 34188 + }, + { + "epoch": 6.22, + "learning_rate": 1.1358390971969422e-06, + "loss": 0.0176, + "step": 34189 + }, + { + "epoch": 6.22, + "learning_rate": 1.13578449217328e-06, + "loss": 0.0168, + "step": 34190 + }, + { + "epoch": 6.22, + "learning_rate": 1.1357298871496179e-06, + "loss": 0.049, + "step": 34191 + }, + { + "epoch": 6.22, + "learning_rate": 1.1356752821259556e-06, + "loss": 0.1838, + "step": 34192 + }, + { + "epoch": 6.22, + "learning_rate": 1.1356206771022933e-06, + "loss": 0.2653, + "step": 34193 + }, + { + "epoch": 6.22, + "learning_rate": 1.1355660720786313e-06, + "loss": 0.1232, + "step": 34194 + }, + { + "epoch": 6.22, + "learning_rate": 1.135511467054969e-06, + "loss": 0.6382, + "step": 34195 + }, + { + "epoch": 6.22, + "learning_rate": 1.135456862031307e-06, + "loss": 0.6853, + "step": 34196 + }, + { + "epoch": 6.22, + "learning_rate": 1.1354022570076447e-06, + "loss": 0.2475, + "step": 34197 + }, + { + "epoch": 6.22, + "learning_rate": 1.1353476519839826e-06, + "loss": 1.148, + "step": 34198 + }, + { + "epoch": 6.22, + "learning_rate": 1.1352930469603204e-06, + "loss": 0.5468, + "step": 34199 + }, + { + "epoch": 6.22, + "learning_rate": 1.135238441936658e-06, + "loss": 0.4169, + "step": 34200 + }, + { + "epoch": 6.22, + "learning_rate": 1.135183836912996e-06, + "loss": 1.4147, + "step": 34201 + }, + { + "epoch": 6.22, + "learning_rate": 1.135129231889334e-06, + "loss": 0.1734, + "step": 34202 + }, + { + "epoch": 6.22, + "learning_rate": 1.1350746268656717e-06, + "loss": 0.7748, + "step": 34203 + }, + { + "epoch": 6.22, + "learning_rate": 1.1350200218420094e-06, + "loss": 0.0094, + "step": 34204 + }, + { + "epoch": 6.22, + "learning_rate": 1.1349654168183474e-06, + "loss": 1.3712, + "step": 34205 + }, + { + "epoch": 6.22, + "learning_rate": 1.1349108117946851e-06, + "loss": 0.1074, + "step": 34206 + }, + { + "epoch": 6.22, + "learning_rate": 1.1348562067710228e-06, + "loss": 0.158, + "step": 34207 + }, + { + "epoch": 6.22, + "learning_rate": 1.1348016017473608e-06, + "loss": 0.0447, + "step": 34208 + }, + { + "epoch": 6.22, + "learning_rate": 1.1347469967236987e-06, + "loss": 0.0633, + "step": 34209 + }, + { + "epoch": 6.22, + "learning_rate": 1.1346923917000365e-06, + "loss": 0.694, + "step": 34210 + }, + { + "epoch": 6.22, + "learning_rate": 1.1346377866763742e-06, + "loss": 0.217, + "step": 34211 + }, + { + "epoch": 6.22, + "learning_rate": 1.1345831816527121e-06, + "loss": 0.1805, + "step": 34212 + }, + { + "epoch": 6.22, + "learning_rate": 1.1345285766290499e-06, + "loss": 0.5657, + "step": 34213 + }, + { + "epoch": 6.22, + "learning_rate": 1.1344739716053876e-06, + "loss": 0.7144, + "step": 34214 + }, + { + "epoch": 6.22, + "learning_rate": 1.1344193665817255e-06, + "loss": 0.4975, + "step": 34215 + }, + { + "epoch": 6.22, + "learning_rate": 1.1343647615580635e-06, + "loss": 0.3077, + "step": 34216 + }, + { + "epoch": 6.22, + "learning_rate": 1.1343101565344012e-06, + "loss": 0.0739, + "step": 34217 + }, + { + "epoch": 6.22, + "learning_rate": 1.134255551510739e-06, + "loss": 0.1116, + "step": 34218 + }, + { + "epoch": 6.22, + "learning_rate": 1.134200946487077e-06, + "loss": 0.0051, + "step": 34219 + }, + { + "epoch": 6.22, + "learning_rate": 1.1341463414634146e-06, + "loss": 2.8743, + "step": 34220 + }, + { + "epoch": 6.22, + "learning_rate": 1.1340917364397524e-06, + "loss": 0.8247, + "step": 34221 + }, + { + "epoch": 6.22, + "learning_rate": 1.1340371314160903e-06, + "loss": 0.9715, + "step": 34222 + }, + { + "epoch": 6.22, + "learning_rate": 1.1339825263924282e-06, + "loss": 1.087, + "step": 34223 + }, + { + "epoch": 6.22, + "learning_rate": 1.133927921368766e-06, + "loss": 0.6566, + "step": 34224 + }, + { + "epoch": 6.22, + "learning_rate": 1.1338733163451037e-06, + "loss": 1.7245, + "step": 34225 + }, + { + "epoch": 6.22, + "learning_rate": 1.1338187113214417e-06, + "loss": 1.0729, + "step": 34226 + }, + { + "epoch": 6.22, + "learning_rate": 1.1337641062977794e-06, + "loss": 0.3525, + "step": 34227 + }, + { + "epoch": 6.22, + "learning_rate": 1.1337095012741171e-06, + "loss": 1.133, + "step": 34228 + }, + { + "epoch": 6.22, + "learning_rate": 1.133654896250455e-06, + "loss": 0.4866, + "step": 34229 + }, + { + "epoch": 6.22, + "learning_rate": 1.133600291226793e-06, + "loss": 0.1856, + "step": 34230 + }, + { + "epoch": 6.22, + "learning_rate": 1.1335456862031307e-06, + "loss": 0.1323, + "step": 34231 + }, + { + "epoch": 6.23, + "learning_rate": 1.1334910811794685e-06, + "loss": 0.1001, + "step": 34232 + }, + { + "epoch": 6.23, + "learning_rate": 1.1334364761558064e-06, + "loss": 0.2154, + "step": 34233 + }, + { + "epoch": 6.23, + "learning_rate": 1.1333818711321441e-06, + "loss": 0.1346, + "step": 34234 + }, + { + "epoch": 6.23, + "learning_rate": 1.1333272661084819e-06, + "loss": 0.026, + "step": 34235 + }, + { + "epoch": 6.23, + "learning_rate": 1.1332726610848198e-06, + "loss": 0.1661, + "step": 34236 + }, + { + "epoch": 6.23, + "learning_rate": 1.1332180560611578e-06, + "loss": 0.7857, + "step": 34237 + }, + { + "epoch": 6.23, + "learning_rate": 1.1331634510374955e-06, + "loss": 0.1585, + "step": 34238 + }, + { + "epoch": 6.23, + "learning_rate": 1.1331088460138332e-06, + "loss": 0.0086, + "step": 34239 + }, + { + "epoch": 6.23, + "learning_rate": 1.1330542409901712e-06, + "loss": 0.5378, + "step": 34240 + }, + { + "epoch": 6.23, + "learning_rate": 1.132999635966509e-06, + "loss": 0.23, + "step": 34241 + }, + { + "epoch": 6.23, + "learning_rate": 1.1329450309428468e-06, + "loss": 0.2365, + "step": 34242 + }, + { + "epoch": 6.23, + "learning_rate": 1.1328904259191846e-06, + "loss": 0.6681, + "step": 34243 + }, + { + "epoch": 6.23, + "learning_rate": 1.1328358208955225e-06, + "loss": 1.4805, + "step": 34244 + }, + { + "epoch": 6.23, + "learning_rate": 1.1327812158718602e-06, + "loss": 0.6765, + "step": 34245 + }, + { + "epoch": 6.23, + "learning_rate": 1.132726610848198e-06, + "loss": 0.3291, + "step": 34246 + }, + { + "epoch": 6.23, + "learning_rate": 1.132672005824536e-06, + "loss": 0.1952, + "step": 34247 + }, + { + "epoch": 6.23, + "learning_rate": 1.1326174008008737e-06, + "loss": 0.7477, + "step": 34248 + }, + { + "epoch": 6.23, + "learning_rate": 1.1325627957772116e-06, + "loss": 0.017, + "step": 34249 + }, + { + "epoch": 6.23, + "learning_rate": 1.1325081907535493e-06, + "loss": 0.255, + "step": 34250 + }, + { + "epoch": 6.23, + "learning_rate": 1.1324535857298873e-06, + "loss": 0.3455, + "step": 34251 + }, + { + "epoch": 6.23, + "learning_rate": 1.132398980706225e-06, + "loss": 0.0146, + "step": 34252 + }, + { + "epoch": 6.23, + "learning_rate": 1.1323443756825627e-06, + "loss": 0.8485, + "step": 34253 + }, + { + "epoch": 6.23, + "learning_rate": 1.1322897706589007e-06, + "loss": 0.371, + "step": 34254 + }, + { + "epoch": 6.23, + "learning_rate": 1.1322351656352384e-06, + "loss": 1.4644, + "step": 34255 + }, + { + "epoch": 6.23, + "learning_rate": 1.1321805606115764e-06, + "loss": 0.2775, + "step": 34256 + }, + { + "epoch": 6.23, + "learning_rate": 1.132125955587914e-06, + "loss": 1.9667, + "step": 34257 + }, + { + "epoch": 6.23, + "learning_rate": 1.132071350564252e-06, + "loss": 0.1681, + "step": 34258 + }, + { + "epoch": 6.23, + "learning_rate": 1.1320167455405898e-06, + "loss": 0.912, + "step": 34259 + }, + { + "epoch": 6.23, + "learning_rate": 1.1319621405169275e-06, + "loss": 0.5969, + "step": 34260 + }, + { + "epoch": 6.23, + "learning_rate": 1.1319075354932654e-06, + "loss": 0.106, + "step": 34261 + }, + { + "epoch": 6.23, + "learning_rate": 1.1318529304696032e-06, + "loss": 0.163, + "step": 34262 + }, + { + "epoch": 6.23, + "learning_rate": 1.1317983254459411e-06, + "loss": 0.3322, + "step": 34263 + }, + { + "epoch": 6.23, + "learning_rate": 1.1317437204222788e-06, + "loss": 0.1579, + "step": 34264 + }, + { + "epoch": 6.23, + "learning_rate": 1.1316891153986168e-06, + "loss": 0.7451, + "step": 34265 + }, + { + "epoch": 6.23, + "learning_rate": 1.1316345103749545e-06, + "loss": 0.3916, + "step": 34266 + }, + { + "epoch": 6.23, + "learning_rate": 1.1315799053512922e-06, + "loss": 1.126, + "step": 34267 + }, + { + "epoch": 6.23, + "learning_rate": 1.1315253003276302e-06, + "loss": 0.021, + "step": 34268 + }, + { + "epoch": 6.23, + "learning_rate": 1.131470695303968e-06, + "loss": 0.2191, + "step": 34269 + }, + { + "epoch": 6.23, + "learning_rate": 1.1314160902803059e-06, + "loss": 0.0043, + "step": 34270 + }, + { + "epoch": 6.23, + "learning_rate": 1.1313614852566436e-06, + "loss": 0.0322, + "step": 34271 + }, + { + "epoch": 6.23, + "learning_rate": 1.1313068802329815e-06, + "loss": 0.1324, + "step": 34272 + }, + { + "epoch": 6.23, + "learning_rate": 1.1312522752093193e-06, + "loss": 1.4468, + "step": 34273 + }, + { + "epoch": 6.23, + "learning_rate": 1.131197670185657e-06, + "loss": 4.2504, + "step": 34274 + }, + { + "epoch": 6.23, + "learning_rate": 1.131143065161995e-06, + "loss": 0.9319, + "step": 34275 + }, + { + "epoch": 6.23, + "learning_rate": 1.1310884601383327e-06, + "loss": 0.4918, + "step": 34276 + }, + { + "epoch": 6.23, + "learning_rate": 1.1310338551146706e-06, + "loss": 2.3277, + "step": 34277 + }, + { + "epoch": 6.23, + "learning_rate": 1.1309792500910084e-06, + "loss": 0.1894, + "step": 34278 + }, + { + "epoch": 6.23, + "learning_rate": 1.1309246450673463e-06, + "loss": 0.0114, + "step": 34279 + }, + { + "epoch": 6.23, + "learning_rate": 1.130870040043684e-06, + "loss": 1.0785, + "step": 34280 + }, + { + "epoch": 6.23, + "learning_rate": 1.1308154350200218e-06, + "loss": 0.2432, + "step": 34281 + }, + { + "epoch": 6.23, + "learning_rate": 1.1307608299963597e-06, + "loss": 0.0463, + "step": 34282 + }, + { + "epoch": 6.23, + "learning_rate": 1.1307062249726976e-06, + "loss": 2.1448, + "step": 34283 + }, + { + "epoch": 6.23, + "learning_rate": 1.1306516199490354e-06, + "loss": 0.7036, + "step": 34284 + }, + { + "epoch": 6.23, + "learning_rate": 1.1305970149253731e-06, + "loss": 0.0423, + "step": 34285 + }, + { + "epoch": 6.23, + "learning_rate": 1.130542409901711e-06, + "loss": 1.1609, + "step": 34286 + }, + { + "epoch": 6.24, + "learning_rate": 1.1304878048780488e-06, + "loss": 2.8153, + "step": 34287 + }, + { + "epoch": 6.24, + "learning_rate": 1.1304331998543865e-06, + "loss": 0.1091, + "step": 34288 + }, + { + "epoch": 6.24, + "learning_rate": 1.1303785948307245e-06, + "loss": 0.0708, + "step": 34289 + }, + { + "epoch": 6.24, + "learning_rate": 1.1303239898070624e-06, + "loss": 0.2761, + "step": 34290 + }, + { + "epoch": 6.24, + "learning_rate": 1.1302693847834001e-06, + "loss": 0.4014, + "step": 34291 + }, + { + "epoch": 6.24, + "learning_rate": 1.1302147797597379e-06, + "loss": 0.0967, + "step": 34292 + }, + { + "epoch": 6.24, + "learning_rate": 1.1301601747360758e-06, + "loss": 1.2106, + "step": 34293 + }, + { + "epoch": 6.24, + "learning_rate": 1.1301055697124135e-06, + "loss": 0.1431, + "step": 34294 + }, + { + "epoch": 6.24, + "learning_rate": 1.1300509646887513e-06, + "loss": 0.018, + "step": 34295 + }, + { + "epoch": 6.24, + "learning_rate": 1.1299963596650892e-06, + "loss": 1.2011, + "step": 34296 + }, + { + "epoch": 6.24, + "learning_rate": 1.1299417546414272e-06, + "loss": 0.4348, + "step": 34297 + }, + { + "epoch": 6.24, + "learning_rate": 1.129887149617765e-06, + "loss": 0.1674, + "step": 34298 + }, + { + "epoch": 6.24, + "learning_rate": 1.1298325445941026e-06, + "loss": 0.0101, + "step": 34299 + }, + { + "epoch": 6.24, + "learning_rate": 1.1297779395704406e-06, + "loss": 0.1547, + "step": 34300 + }, + { + "epoch": 6.24, + "learning_rate": 1.1297233345467783e-06, + "loss": 1.2625, + "step": 34301 + }, + { + "epoch": 6.24, + "learning_rate": 1.129668729523116e-06, + "loss": 0.1932, + "step": 34302 + }, + { + "epoch": 6.24, + "learning_rate": 1.129614124499454e-06, + "loss": 0.0089, + "step": 34303 + }, + { + "epoch": 6.24, + "learning_rate": 1.129559519475792e-06, + "loss": 0.4214, + "step": 34304 + }, + { + "epoch": 6.24, + "learning_rate": 1.1295049144521297e-06, + "loss": 0.005, + "step": 34305 + }, + { + "epoch": 6.24, + "learning_rate": 1.1294503094284674e-06, + "loss": 0.6333, + "step": 34306 + }, + { + "epoch": 6.24, + "learning_rate": 1.1293957044048053e-06, + "loss": 0.1046, + "step": 34307 + }, + { + "epoch": 6.24, + "learning_rate": 1.129341099381143e-06, + "loss": 2.5417, + "step": 34308 + }, + { + "epoch": 6.24, + "learning_rate": 1.1292864943574808e-06, + "loss": 0.415, + "step": 34309 + }, + { + "epoch": 6.24, + "learning_rate": 1.1292318893338187e-06, + "loss": 1.4165, + "step": 34310 + }, + { + "epoch": 6.24, + "learning_rate": 1.1291772843101567e-06, + "loss": 2.3774, + "step": 34311 + }, + { + "epoch": 6.24, + "learning_rate": 1.1291226792864944e-06, + "loss": 0.2168, + "step": 34312 + }, + { + "epoch": 6.24, + "learning_rate": 1.1290680742628321e-06, + "loss": 0.6727, + "step": 34313 + }, + { + "epoch": 6.24, + "learning_rate": 1.12901346923917e-06, + "loss": 0.3555, + "step": 34314 + }, + { + "epoch": 6.24, + "learning_rate": 1.1289588642155078e-06, + "loss": 0.284, + "step": 34315 + }, + { + "epoch": 6.24, + "learning_rate": 1.1289042591918455e-06, + "loss": 0.3565, + "step": 34316 + }, + { + "epoch": 6.24, + "learning_rate": 1.1288496541681835e-06, + "loss": 0.3031, + "step": 34317 + }, + { + "epoch": 6.24, + "learning_rate": 1.1287950491445214e-06, + "loss": 0.5378, + "step": 34318 + }, + { + "epoch": 6.24, + "learning_rate": 1.1287404441208592e-06, + "loss": 0.0067, + "step": 34319 + }, + { + "epoch": 6.24, + "learning_rate": 1.128685839097197e-06, + "loss": 0.1767, + "step": 34320 + }, + { + "epoch": 6.24, + "learning_rate": 1.1286312340735348e-06, + "loss": 0.1537, + "step": 34321 + }, + { + "epoch": 6.24, + "learning_rate": 1.1285766290498726e-06, + "loss": 0.0961, + "step": 34322 + }, + { + "epoch": 6.24, + "learning_rate": 1.1285220240262103e-06, + "loss": 0.2509, + "step": 34323 + }, + { + "epoch": 6.24, + "learning_rate": 1.1284674190025482e-06, + "loss": 0.3028, + "step": 34324 + }, + { + "epoch": 6.24, + "learning_rate": 1.1284128139788862e-06, + "loss": 0.8958, + "step": 34325 + }, + { + "epoch": 6.24, + "learning_rate": 1.128358208955224e-06, + "loss": 0.0156, + "step": 34326 + }, + { + "epoch": 6.24, + "learning_rate": 1.1283036039315617e-06, + "loss": 1.1144, + "step": 34327 + }, + { + "epoch": 6.24, + "learning_rate": 1.1282489989078996e-06, + "loss": 0.5264, + "step": 34328 + }, + { + "epoch": 6.24, + "learning_rate": 1.1281943938842373e-06, + "loss": 0.2796, + "step": 34329 + }, + { + "epoch": 6.24, + "learning_rate": 1.1281397888605753e-06, + "loss": 0.7052, + "step": 34330 + }, + { + "epoch": 6.24, + "learning_rate": 1.128085183836913e-06, + "loss": 0.0068, + "step": 34331 + }, + { + "epoch": 6.24, + "learning_rate": 1.128030578813251e-06, + "loss": 0.2759, + "step": 34332 + }, + { + "epoch": 6.24, + "learning_rate": 1.1279759737895887e-06, + "loss": 0.7058, + "step": 34333 + }, + { + "epoch": 6.24, + "learning_rate": 1.1279213687659264e-06, + "loss": 0.1261, + "step": 34334 + }, + { + "epoch": 6.24, + "learning_rate": 1.1278667637422644e-06, + "loss": 0.4713, + "step": 34335 + }, + { + "epoch": 6.24, + "learning_rate": 1.127812158718602e-06, + "loss": 1.9684, + "step": 34336 + }, + { + "epoch": 6.24, + "learning_rate": 1.12775755369494e-06, + "loss": 2.1016, + "step": 34337 + }, + { + "epoch": 6.24, + "learning_rate": 1.1277029486712778e-06, + "loss": 0.5246, + "step": 34338 + }, + { + "epoch": 6.24, + "learning_rate": 1.1276483436476157e-06, + "loss": 1.2555, + "step": 34339 + }, + { + "epoch": 6.24, + "learning_rate": 1.1275937386239534e-06, + "loss": 0.011, + "step": 34340 + }, + { + "epoch": 6.24, + "learning_rate": 1.1275391336002912e-06, + "loss": 0.0242, + "step": 34341 + }, + { + "epoch": 6.25, + "learning_rate": 1.1274845285766291e-06, + "loss": 2.0997, + "step": 34342 + }, + { + "epoch": 6.25, + "learning_rate": 1.1274299235529668e-06, + "loss": 0.6941, + "step": 34343 + }, + { + "epoch": 6.25, + "learning_rate": 1.1273753185293048e-06, + "loss": 0.7907, + "step": 34344 + }, + { + "epoch": 6.25, + "learning_rate": 1.1273207135056425e-06, + "loss": 0.1401, + "step": 34345 + }, + { + "epoch": 6.25, + "learning_rate": 1.1272661084819805e-06, + "loss": 0.1814, + "step": 34346 + }, + { + "epoch": 6.25, + "learning_rate": 1.1272115034583182e-06, + "loss": 0.1284, + "step": 34347 + }, + { + "epoch": 6.25, + "learning_rate": 1.127156898434656e-06, + "loss": 1.6358, + "step": 34348 + }, + { + "epoch": 6.25, + "learning_rate": 1.1271022934109939e-06, + "loss": 0.0709, + "step": 34349 + }, + { + "epoch": 6.25, + "learning_rate": 1.1270476883873316e-06, + "loss": 0.9853, + "step": 34350 + }, + { + "epoch": 6.25, + "learning_rate": 1.1269930833636695e-06, + "loss": 0.7911, + "step": 34351 + }, + { + "epoch": 6.25, + "learning_rate": 1.1269384783400073e-06, + "loss": 3.5171, + "step": 34352 + }, + { + "epoch": 6.25, + "learning_rate": 1.1268838733163452e-06, + "loss": 0.3529, + "step": 34353 + }, + { + "epoch": 6.25, + "learning_rate": 1.126829268292683e-06, + "loss": 0.0027, + "step": 34354 + }, + { + "epoch": 6.25, + "learning_rate": 1.1267746632690207e-06, + "loss": 1.0181, + "step": 34355 + }, + { + "epoch": 6.25, + "learning_rate": 1.1267200582453586e-06, + "loss": 0.2203, + "step": 34356 + }, + { + "epoch": 6.25, + "learning_rate": 1.1266654532216964e-06, + "loss": 0.0515, + "step": 34357 + }, + { + "epoch": 6.25, + "learning_rate": 1.1266108481980343e-06, + "loss": 2.9546, + "step": 34358 + }, + { + "epoch": 6.25, + "learning_rate": 1.126556243174372e-06, + "loss": 2.255, + "step": 34359 + }, + { + "epoch": 6.25, + "learning_rate": 1.12650163815071e-06, + "loss": 0.0168, + "step": 34360 + }, + { + "epoch": 6.25, + "learning_rate": 1.1264470331270477e-06, + "loss": 0.5906, + "step": 34361 + }, + { + "epoch": 6.25, + "learning_rate": 1.1263924281033854e-06, + "loss": 0.3294, + "step": 34362 + }, + { + "epoch": 6.25, + "learning_rate": 1.1263378230797234e-06, + "loss": 0.0335, + "step": 34363 + }, + { + "epoch": 6.25, + "learning_rate": 1.1262832180560611e-06, + "loss": 0.0087, + "step": 34364 + }, + { + "epoch": 6.25, + "learning_rate": 1.126228613032399e-06, + "loss": 0.0112, + "step": 34365 + }, + { + "epoch": 6.25, + "learning_rate": 1.1261740080087368e-06, + "loss": 1.7176, + "step": 34366 + }, + { + "epoch": 6.25, + "learning_rate": 1.1261194029850747e-06, + "loss": 0.0264, + "step": 34367 + }, + { + "epoch": 6.25, + "learning_rate": 1.1260647979614125e-06, + "loss": 0.0139, + "step": 34368 + }, + { + "epoch": 6.25, + "learning_rate": 1.1260101929377502e-06, + "loss": 0.0108, + "step": 34369 + }, + { + "epoch": 6.25, + "learning_rate": 1.1259555879140881e-06, + "loss": 0.0794, + "step": 34370 + }, + { + "epoch": 6.25, + "learning_rate": 1.125900982890426e-06, + "loss": 0.0275, + "step": 34371 + }, + { + "epoch": 6.25, + "learning_rate": 1.1258463778667638e-06, + "loss": 0.3144, + "step": 34372 + }, + { + "epoch": 6.25, + "learning_rate": 1.1257917728431015e-06, + "loss": 2.9219, + "step": 34373 + }, + { + "epoch": 6.25, + "learning_rate": 1.1257371678194395e-06, + "loss": 3.1315, + "step": 34374 + }, + { + "epoch": 6.25, + "learning_rate": 1.1256825627957772e-06, + "loss": 0.6026, + "step": 34375 + }, + { + "epoch": 6.25, + "learning_rate": 1.125627957772115e-06, + "loss": 0.0038, + "step": 34376 + }, + { + "epoch": 6.25, + "learning_rate": 1.1255733527484529e-06, + "loss": 0.0258, + "step": 34377 + }, + { + "epoch": 6.25, + "learning_rate": 1.1255187477247908e-06, + "loss": 0.0469, + "step": 34378 + }, + { + "epoch": 6.25, + "learning_rate": 1.1254641427011286e-06, + "loss": 0.3528, + "step": 34379 + }, + { + "epoch": 6.25, + "learning_rate": 1.1254095376774663e-06, + "loss": 1.3512, + "step": 34380 + }, + { + "epoch": 6.25, + "learning_rate": 1.1253549326538042e-06, + "loss": 0.0078, + "step": 34381 + }, + { + "epoch": 6.25, + "learning_rate": 1.125300327630142e-06, + "loss": 0.6409, + "step": 34382 + }, + { + "epoch": 6.25, + "learning_rate": 1.1252457226064797e-06, + "loss": 3.9813, + "step": 34383 + }, + { + "epoch": 6.25, + "learning_rate": 1.1251911175828177e-06, + "loss": 0.302, + "step": 34384 + }, + { + "epoch": 6.25, + "learning_rate": 1.1251365125591556e-06, + "loss": 1.3021, + "step": 34385 + }, + { + "epoch": 6.25, + "learning_rate": 1.1250819075354933e-06, + "loss": 1.9524, + "step": 34386 + }, + { + "epoch": 6.25, + "learning_rate": 1.125027302511831e-06, + "loss": 0.3489, + "step": 34387 + }, + { + "epoch": 6.25, + "learning_rate": 1.124972697488169e-06, + "loss": 0.2049, + "step": 34388 + }, + { + "epoch": 6.25, + "learning_rate": 1.1249180924645067e-06, + "loss": 0.9702, + "step": 34389 + }, + { + "epoch": 6.25, + "learning_rate": 1.1248634874408445e-06, + "loss": 0.2279, + "step": 34390 + }, + { + "epoch": 6.25, + "learning_rate": 1.1248088824171824e-06, + "loss": 0.3355, + "step": 34391 + }, + { + "epoch": 6.25, + "learning_rate": 1.1247542773935204e-06, + "loss": 0.0709, + "step": 34392 + }, + { + "epoch": 6.25, + "learning_rate": 1.124699672369858e-06, + "loss": 0.0111, + "step": 34393 + }, + { + "epoch": 6.25, + "learning_rate": 1.1246450673461958e-06, + "loss": 0.7084, + "step": 34394 + }, + { + "epoch": 6.25, + "learning_rate": 1.1245904623225338e-06, + "loss": 1.5899, + "step": 34395 + }, + { + "epoch": 6.25, + "learning_rate": 1.1245358572988715e-06, + "loss": 0.0774, + "step": 34396 + }, + { + "epoch": 6.26, + "learning_rate": 1.1244812522752092e-06, + "loss": 0.1827, + "step": 34397 + }, + { + "epoch": 6.26, + "learning_rate": 1.1244266472515472e-06, + "loss": 0.0146, + "step": 34398 + }, + { + "epoch": 6.26, + "learning_rate": 1.1243720422278851e-06, + "loss": 0.0066, + "step": 34399 + }, + { + "epoch": 6.26, + "learning_rate": 1.1243174372042228e-06, + "loss": 0.1133, + "step": 34400 + }, + { + "epoch": 6.26, + "learning_rate": 1.1242628321805606e-06, + "loss": 0.041, + "step": 34401 + }, + { + "epoch": 6.26, + "learning_rate": 1.1242082271568985e-06, + "loss": 0.9073, + "step": 34402 + }, + { + "epoch": 6.26, + "learning_rate": 1.1241536221332362e-06, + "loss": 0.1143, + "step": 34403 + }, + { + "epoch": 6.26, + "learning_rate": 1.124099017109574e-06, + "loss": 0.4114, + "step": 34404 + }, + { + "epoch": 6.26, + "learning_rate": 1.124044412085912e-06, + "loss": 1.4944, + "step": 34405 + }, + { + "epoch": 6.26, + "learning_rate": 1.1239898070622499e-06, + "loss": 0.0088, + "step": 34406 + }, + { + "epoch": 6.26, + "learning_rate": 1.1239352020385876e-06, + "loss": 0.625, + "step": 34407 + }, + { + "epoch": 6.26, + "learning_rate": 1.1238805970149253e-06, + "loss": 0.2279, + "step": 34408 + }, + { + "epoch": 6.26, + "learning_rate": 1.1238259919912633e-06, + "loss": 0.1091, + "step": 34409 + }, + { + "epoch": 6.26, + "learning_rate": 1.123771386967601e-06, + "loss": 0.4684, + "step": 34410 + }, + { + "epoch": 6.26, + "learning_rate": 1.123716781943939e-06, + "loss": 0.2087, + "step": 34411 + }, + { + "epoch": 6.26, + "learning_rate": 1.1236621769202767e-06, + "loss": 0.5225, + "step": 34412 + }, + { + "epoch": 6.26, + "learning_rate": 1.1236075718966146e-06, + "loss": 0.427, + "step": 34413 + }, + { + "epoch": 6.26, + "learning_rate": 1.1235529668729524e-06, + "loss": 1.5538, + "step": 34414 + }, + { + "epoch": 6.26, + "learning_rate": 1.12349836184929e-06, + "loss": 0.0966, + "step": 34415 + }, + { + "epoch": 6.26, + "learning_rate": 1.123443756825628e-06, + "loss": 0.4957, + "step": 34416 + }, + { + "epoch": 6.26, + "learning_rate": 1.1233891518019658e-06, + "loss": 1.3206, + "step": 34417 + }, + { + "epoch": 6.26, + "learning_rate": 1.1233345467783037e-06, + "loss": 0.471, + "step": 34418 + }, + { + "epoch": 6.26, + "learning_rate": 1.1232799417546414e-06, + "loss": 0.1582, + "step": 34419 + }, + { + "epoch": 6.26, + "learning_rate": 1.1232253367309794e-06, + "loss": 0.1936, + "step": 34420 + }, + { + "epoch": 6.26, + "learning_rate": 1.1231707317073171e-06, + "loss": 0.4326, + "step": 34421 + }, + { + "epoch": 6.26, + "learning_rate": 1.1231161266836548e-06, + "loss": 1.0492, + "step": 34422 + }, + { + "epoch": 6.26, + "learning_rate": 1.1230615216599928e-06, + "loss": 0.1882, + "step": 34423 + }, + { + "epoch": 6.26, + "learning_rate": 1.1230069166363305e-06, + "loss": 0.9502, + "step": 34424 + }, + { + "epoch": 6.26, + "learning_rate": 1.1229523116126685e-06, + "loss": 0.0138, + "step": 34425 + }, + { + "epoch": 6.26, + "learning_rate": 1.1228977065890062e-06, + "loss": 0.8412, + "step": 34426 + }, + { + "epoch": 6.26, + "learning_rate": 1.1228431015653441e-06, + "loss": 0.1832, + "step": 34427 + }, + { + "epoch": 6.26, + "learning_rate": 1.1227884965416819e-06, + "loss": 0.8404, + "step": 34428 + }, + { + "epoch": 6.26, + "learning_rate": 1.1227338915180196e-06, + "loss": 0.7127, + "step": 34429 + }, + { + "epoch": 6.26, + "learning_rate": 1.1226792864943575e-06, + "loss": 1.2351, + "step": 34430 + }, + { + "epoch": 6.26, + "learning_rate": 1.1226246814706953e-06, + "loss": 1.5195, + "step": 34431 + }, + { + "epoch": 6.26, + "learning_rate": 1.1225700764470332e-06, + "loss": 0.0213, + "step": 34432 + }, + { + "epoch": 6.26, + "learning_rate": 1.122515471423371e-06, + "loss": 0.0219, + "step": 34433 + }, + { + "epoch": 6.26, + "learning_rate": 1.1224608663997089e-06, + "loss": 0.3834, + "step": 34434 + }, + { + "epoch": 6.26, + "learning_rate": 1.1224062613760466e-06, + "loss": 5.09, + "step": 34435 + }, + { + "epoch": 6.26, + "learning_rate": 1.1223516563523844e-06, + "loss": 0.8523, + "step": 34436 + }, + { + "epoch": 6.26, + "learning_rate": 1.1222970513287223e-06, + "loss": 0.707, + "step": 34437 + }, + { + "epoch": 6.26, + "learning_rate": 1.12224244630506e-06, + "loss": 0.0027, + "step": 34438 + }, + { + "epoch": 6.26, + "learning_rate": 1.122187841281398e-06, + "loss": 0.0037, + "step": 34439 + }, + { + "epoch": 6.26, + "learning_rate": 1.1221332362577357e-06, + "loss": 0.1934, + "step": 34440 + }, + { + "epoch": 6.26, + "learning_rate": 1.1220786312340736e-06, + "loss": 1.0517, + "step": 34441 + }, + { + "epoch": 6.26, + "learning_rate": 1.1220240262104114e-06, + "loss": 0.2273, + "step": 34442 + }, + { + "epoch": 6.26, + "learning_rate": 1.1219694211867491e-06, + "loss": 0.1822, + "step": 34443 + }, + { + "epoch": 6.26, + "learning_rate": 1.121914816163087e-06, + "loss": 0.458, + "step": 34444 + }, + { + "epoch": 6.26, + "learning_rate": 1.1218602111394248e-06, + "loss": 0.0683, + "step": 34445 + }, + { + "epoch": 6.26, + "learning_rate": 1.1218056061157627e-06, + "loss": 0.7936, + "step": 34446 + }, + { + "epoch": 6.26, + "learning_rate": 1.1217510010921005e-06, + "loss": 0.4603, + "step": 34447 + }, + { + "epoch": 6.26, + "learning_rate": 1.1216963960684384e-06, + "loss": 0.139, + "step": 34448 + }, + { + "epoch": 6.26, + "learning_rate": 1.1216417910447761e-06, + "loss": 0.7136, + "step": 34449 + }, + { + "epoch": 6.26, + "learning_rate": 1.1215871860211139e-06, + "loss": 1.2533, + "step": 34450 + }, + { + "epoch": 6.26, + "learning_rate": 1.1215325809974518e-06, + "loss": 1.6615, + "step": 34451 + }, + { + "epoch": 6.27, + "learning_rate": 1.1214779759737898e-06, + "loss": 0.7219, + "step": 34452 + }, + { + "epoch": 6.27, + "learning_rate": 1.1214233709501275e-06, + "loss": 0.1633, + "step": 34453 + }, + { + "epoch": 6.27, + "learning_rate": 1.1213687659264652e-06, + "loss": 0.2485, + "step": 34454 + }, + { + "epoch": 6.27, + "learning_rate": 1.1213141609028032e-06, + "loss": 1.4086, + "step": 34455 + }, + { + "epoch": 6.27, + "learning_rate": 1.1212595558791409e-06, + "loss": 1.0447, + "step": 34456 + }, + { + "epoch": 6.27, + "learning_rate": 1.1212049508554786e-06, + "loss": 0.0058, + "step": 34457 + }, + { + "epoch": 6.27, + "learning_rate": 1.1211503458318166e-06, + "loss": 0.0728, + "step": 34458 + }, + { + "epoch": 6.27, + "learning_rate": 1.1210957408081545e-06, + "loss": 0.0163, + "step": 34459 + }, + { + "epoch": 6.27, + "learning_rate": 1.1210411357844922e-06, + "loss": 0.513, + "step": 34460 + }, + { + "epoch": 6.27, + "learning_rate": 1.12098653076083e-06, + "loss": 0.0143, + "step": 34461 + }, + { + "epoch": 6.27, + "learning_rate": 1.120931925737168e-06, + "loss": 0.0175, + "step": 34462 + }, + { + "epoch": 6.27, + "learning_rate": 1.1208773207135056e-06, + "loss": 0.0303, + "step": 34463 + }, + { + "epoch": 6.27, + "learning_rate": 1.1208227156898434e-06, + "loss": 0.0966, + "step": 34464 + }, + { + "epoch": 6.27, + "learning_rate": 1.1207681106661813e-06, + "loss": 0.4018, + "step": 34465 + }, + { + "epoch": 6.27, + "learning_rate": 1.1207135056425193e-06, + "loss": 0.1606, + "step": 34466 + }, + { + "epoch": 6.27, + "learning_rate": 1.120658900618857e-06, + "loss": 0.2508, + "step": 34467 + }, + { + "epoch": 6.27, + "learning_rate": 1.1206042955951947e-06, + "loss": 0.0051, + "step": 34468 + }, + { + "epoch": 6.27, + "learning_rate": 1.1205496905715327e-06, + "loss": 0.0075, + "step": 34469 + }, + { + "epoch": 6.27, + "learning_rate": 1.1204950855478704e-06, + "loss": 0.9165, + "step": 34470 + }, + { + "epoch": 6.27, + "learning_rate": 1.1204404805242081e-06, + "loss": 0.2361, + "step": 34471 + }, + { + "epoch": 6.27, + "learning_rate": 1.120385875500546e-06, + "loss": 0.3955, + "step": 34472 + }, + { + "epoch": 6.27, + "learning_rate": 1.120331270476884e-06, + "loss": 0.878, + "step": 34473 + }, + { + "epoch": 6.27, + "learning_rate": 1.1202766654532218e-06, + "loss": 0.1328, + "step": 34474 + }, + { + "epoch": 6.27, + "learning_rate": 1.1202220604295595e-06, + "loss": 0.5971, + "step": 34475 + }, + { + "epoch": 6.27, + "learning_rate": 1.1201674554058974e-06, + "loss": 0.0293, + "step": 34476 + }, + { + "epoch": 6.27, + "learning_rate": 1.1201128503822352e-06, + "loss": 0.45, + "step": 34477 + }, + { + "epoch": 6.27, + "learning_rate": 1.120058245358573e-06, + "loss": 0.5673, + "step": 34478 + }, + { + "epoch": 6.27, + "learning_rate": 1.1200036403349108e-06, + "loss": 0.008, + "step": 34479 + }, + { + "epoch": 6.27, + "learning_rate": 1.1199490353112488e-06, + "loss": 0.0033, + "step": 34480 + }, + { + "epoch": 6.27, + "learning_rate": 1.1198944302875865e-06, + "loss": 2.2714, + "step": 34481 + }, + { + "epoch": 6.27, + "learning_rate": 1.1198398252639242e-06, + "loss": 0.0071, + "step": 34482 + }, + { + "epoch": 6.27, + "learning_rate": 1.1197852202402622e-06, + "loss": 0.3653, + "step": 34483 + }, + { + "epoch": 6.27, + "learning_rate": 1.1197306152166e-06, + "loss": 3.919, + "step": 34484 + }, + { + "epoch": 6.27, + "learning_rate": 1.1196760101929377e-06, + "loss": 0.0161, + "step": 34485 + }, + { + "epoch": 6.27, + "learning_rate": 1.1196214051692756e-06, + "loss": 0.0483, + "step": 34486 + }, + { + "epoch": 6.27, + "learning_rate": 1.1195668001456135e-06, + "loss": 0.1346, + "step": 34487 + }, + { + "epoch": 6.27, + "learning_rate": 1.1195121951219513e-06, + "loss": 0.1588, + "step": 34488 + }, + { + "epoch": 6.27, + "learning_rate": 1.119457590098289e-06, + "loss": 0.2809, + "step": 34489 + }, + { + "epoch": 6.27, + "learning_rate": 1.119402985074627e-06, + "loss": 0.2717, + "step": 34490 + }, + { + "epoch": 6.27, + "learning_rate": 1.1193483800509647e-06, + "loss": 0.0913, + "step": 34491 + }, + { + "epoch": 6.27, + "learning_rate": 1.1192937750273024e-06, + "loss": 0.7101, + "step": 34492 + }, + { + "epoch": 6.27, + "learning_rate": 1.1192391700036404e-06, + "loss": 2.8962, + "step": 34493 + }, + { + "epoch": 6.27, + "learning_rate": 1.1191845649799783e-06, + "loss": 3.7041, + "step": 34494 + }, + { + "epoch": 6.27, + "learning_rate": 1.119129959956316e-06, + "loss": 0.1305, + "step": 34495 + }, + { + "epoch": 6.27, + "learning_rate": 1.1190753549326538e-06, + "loss": 3.9895, + "step": 34496 + }, + { + "epoch": 6.27, + "learning_rate": 1.1190207499089917e-06, + "loss": 0.368, + "step": 34497 + }, + { + "epoch": 6.27, + "learning_rate": 1.1189661448853294e-06, + "loss": 0.2378, + "step": 34498 + }, + { + "epoch": 6.27, + "learning_rate": 1.1189115398616674e-06, + "loss": 0.0055, + "step": 34499 + }, + { + "epoch": 6.27, + "learning_rate": 1.1188569348380051e-06, + "loss": 0.2488, + "step": 34500 + }, + { + "epoch": 6.27, + "learning_rate": 1.118802329814343e-06, + "loss": 0.655, + "step": 34501 + }, + { + "epoch": 6.27, + "learning_rate": 1.1187477247906808e-06, + "loss": 1.5059, + "step": 34502 + }, + { + "epoch": 6.27, + "learning_rate": 1.1186931197670185e-06, + "loss": 0.2951, + "step": 34503 + }, + { + "epoch": 6.27, + "learning_rate": 1.1186385147433565e-06, + "loss": 5.0128, + "step": 34504 + }, + { + "epoch": 6.27, + "learning_rate": 1.1185839097196942e-06, + "loss": 0.4959, + "step": 34505 + }, + { + "epoch": 6.27, + "learning_rate": 1.1185293046960321e-06, + "loss": 2.319, + "step": 34506 + }, + { + "epoch": 6.28, + "learning_rate": 1.1184746996723699e-06, + "loss": 0.7483, + "step": 34507 + }, + { + "epoch": 6.28, + "learning_rate": 1.1184200946487078e-06, + "loss": 0.4757, + "step": 34508 + }, + { + "epoch": 6.28, + "learning_rate": 1.1183654896250455e-06, + "loss": 0.0985, + "step": 34509 + }, + { + "epoch": 6.28, + "learning_rate": 1.1183108846013833e-06, + "loss": 0.2675, + "step": 34510 + }, + { + "epoch": 6.28, + "learning_rate": 1.1182562795777212e-06, + "loss": 0.0283, + "step": 34511 + }, + { + "epoch": 6.28, + "learning_rate": 1.118201674554059e-06, + "loss": 0.0728, + "step": 34512 + }, + { + "epoch": 6.28, + "learning_rate": 1.1181470695303969e-06, + "loss": 0.1605, + "step": 34513 + }, + { + "epoch": 6.28, + "learning_rate": 1.1180924645067346e-06, + "loss": 0.021, + "step": 34514 + }, + { + "epoch": 6.28, + "learning_rate": 1.1180378594830726e-06, + "loss": 0.0298, + "step": 34515 + }, + { + "epoch": 6.28, + "learning_rate": 1.1179832544594103e-06, + "loss": 0.0458, + "step": 34516 + }, + { + "epoch": 6.28, + "learning_rate": 1.117928649435748e-06, + "loss": 2.6714, + "step": 34517 + }, + { + "epoch": 6.28, + "learning_rate": 1.117874044412086e-06, + "loss": 1.0843, + "step": 34518 + }, + { + "epoch": 6.28, + "learning_rate": 1.1178194393884237e-06, + "loss": 0.6487, + "step": 34519 + }, + { + "epoch": 6.28, + "learning_rate": 1.1177648343647616e-06, + "loss": 0.0871, + "step": 34520 + }, + { + "epoch": 6.28, + "learning_rate": 1.1177102293410994e-06, + "loss": 0.5678, + "step": 34521 + }, + { + "epoch": 6.28, + "learning_rate": 1.1176556243174373e-06, + "loss": 0.3863, + "step": 34522 + }, + { + "epoch": 6.28, + "learning_rate": 1.117601019293775e-06, + "loss": 1.3136, + "step": 34523 + }, + { + "epoch": 6.28, + "learning_rate": 1.1175464142701128e-06, + "loss": 0.1986, + "step": 34524 + }, + { + "epoch": 6.28, + "learning_rate": 1.1174918092464507e-06, + "loss": 0.9105, + "step": 34525 + }, + { + "epoch": 6.28, + "learning_rate": 1.1174372042227885e-06, + "loss": 1.3556, + "step": 34526 + }, + { + "epoch": 6.28, + "learning_rate": 1.1173825991991264e-06, + "loss": 0.1023, + "step": 34527 + }, + { + "epoch": 6.28, + "learning_rate": 1.1173279941754641e-06, + "loss": 0.9648, + "step": 34528 + }, + { + "epoch": 6.28, + "learning_rate": 1.117273389151802e-06, + "loss": 0.2441, + "step": 34529 + }, + { + "epoch": 6.28, + "learning_rate": 1.1172187841281398e-06, + "loss": 2.1187, + "step": 34530 + }, + { + "epoch": 6.28, + "learning_rate": 1.1171641791044775e-06, + "loss": 0.2563, + "step": 34531 + }, + { + "epoch": 6.28, + "learning_rate": 1.1171095740808155e-06, + "loss": 0.0555, + "step": 34532 + }, + { + "epoch": 6.28, + "learning_rate": 1.1170549690571532e-06, + "loss": 0.0058, + "step": 34533 + }, + { + "epoch": 6.28, + "learning_rate": 1.1170003640334912e-06, + "loss": 3.2202, + "step": 34534 + }, + { + "epoch": 6.28, + "learning_rate": 1.1169457590098289e-06, + "loss": 0.1747, + "step": 34535 + }, + { + "epoch": 6.28, + "learning_rate": 1.1168911539861668e-06, + "loss": 0.3427, + "step": 34536 + }, + { + "epoch": 6.28, + "learning_rate": 1.1168365489625046e-06, + "loss": 0.441, + "step": 34537 + }, + { + "epoch": 6.28, + "learning_rate": 1.1167819439388423e-06, + "loss": 0.0503, + "step": 34538 + }, + { + "epoch": 6.28, + "learning_rate": 1.1167273389151802e-06, + "loss": 0.0964, + "step": 34539 + }, + { + "epoch": 6.28, + "learning_rate": 1.1166727338915182e-06, + "loss": 0.8748, + "step": 34540 + }, + { + "epoch": 6.28, + "learning_rate": 1.116618128867856e-06, + "loss": 0.1245, + "step": 34541 + }, + { + "epoch": 6.28, + "learning_rate": 1.1165635238441936e-06, + "loss": 0.0402, + "step": 34542 + }, + { + "epoch": 6.28, + "learning_rate": 1.1165089188205316e-06, + "loss": 0.5293, + "step": 34543 + }, + { + "epoch": 6.28, + "learning_rate": 1.1164543137968693e-06, + "loss": 0.8127, + "step": 34544 + }, + { + "epoch": 6.28, + "learning_rate": 1.116399708773207e-06, + "loss": 0.1559, + "step": 34545 + }, + { + "epoch": 6.28, + "learning_rate": 1.116345103749545e-06, + "loss": 3.0051, + "step": 34546 + }, + { + "epoch": 6.28, + "learning_rate": 1.116290498725883e-06, + "loss": 1.2461, + "step": 34547 + }, + { + "epoch": 6.28, + "learning_rate": 1.1162358937022207e-06, + "loss": 0.0072, + "step": 34548 + }, + { + "epoch": 6.28, + "learning_rate": 1.1161812886785584e-06, + "loss": 0.0148, + "step": 34549 + }, + { + "epoch": 6.28, + "learning_rate": 1.1161266836548963e-06, + "loss": 0.0336, + "step": 34550 + }, + { + "epoch": 6.28, + "learning_rate": 1.116072078631234e-06, + "loss": 0.2522, + "step": 34551 + }, + { + "epoch": 6.28, + "learning_rate": 1.1160174736075718e-06, + "loss": 1.3863, + "step": 34552 + }, + { + "epoch": 6.28, + "learning_rate": 1.1159628685839098e-06, + "loss": 1.8874, + "step": 34553 + }, + { + "epoch": 6.28, + "learning_rate": 1.1159082635602477e-06, + "loss": 0.5617, + "step": 34554 + }, + { + "epoch": 6.28, + "learning_rate": 1.1158536585365854e-06, + "loss": 0.7121, + "step": 34555 + }, + { + "epoch": 6.28, + "learning_rate": 1.1157990535129232e-06, + "loss": 1.2765, + "step": 34556 + }, + { + "epoch": 6.28, + "learning_rate": 1.115744448489261e-06, + "loss": 1.4838, + "step": 34557 + }, + { + "epoch": 6.28, + "learning_rate": 1.1156898434655988e-06, + "loss": 0.2204, + "step": 34558 + }, + { + "epoch": 6.28, + "learning_rate": 1.1156352384419366e-06, + "loss": 0.2836, + "step": 34559 + }, + { + "epoch": 6.28, + "learning_rate": 1.1155806334182745e-06, + "loss": 0.3921, + "step": 34560 + }, + { + "epoch": 6.28, + "learning_rate": 1.1155260283946125e-06, + "loss": 0.1687, + "step": 34561 + }, + { + "epoch": 6.29, + "learning_rate": 1.1154714233709502e-06, + "loss": 2.334, + "step": 34562 + }, + { + "epoch": 6.29, + "learning_rate": 1.115416818347288e-06, + "loss": 3.1563, + "step": 34563 + }, + { + "epoch": 6.29, + "learning_rate": 1.1153622133236259e-06, + "loss": 1.0092, + "step": 34564 + }, + { + "epoch": 6.29, + "learning_rate": 1.1153076082999636e-06, + "loss": 0.0158, + "step": 34565 + }, + { + "epoch": 6.29, + "learning_rate": 1.1152530032763013e-06, + "loss": 0.0287, + "step": 34566 + }, + { + "epoch": 6.29, + "learning_rate": 1.1151983982526393e-06, + "loss": 0.3358, + "step": 34567 + }, + { + "epoch": 6.29, + "learning_rate": 1.1151437932289772e-06, + "loss": 2.4515, + "step": 34568 + }, + { + "epoch": 6.29, + "learning_rate": 1.115089188205315e-06, + "loss": 0.2506, + "step": 34569 + }, + { + "epoch": 6.29, + "learning_rate": 1.1150345831816527e-06, + "loss": 2.1663, + "step": 34570 + }, + { + "epoch": 6.29, + "learning_rate": 1.1149799781579906e-06, + "loss": 0.3908, + "step": 34571 + }, + { + "epoch": 6.29, + "learning_rate": 1.1149253731343284e-06, + "loss": 0.9767, + "step": 34572 + }, + { + "epoch": 6.29, + "learning_rate": 1.114870768110666e-06, + "loss": 1.1132, + "step": 34573 + }, + { + "epoch": 6.29, + "learning_rate": 1.114816163087004e-06, + "loss": 0.5977, + "step": 34574 + }, + { + "epoch": 6.29, + "learning_rate": 1.114761558063342e-06, + "loss": 1.6298, + "step": 34575 + }, + { + "epoch": 6.29, + "learning_rate": 1.1147069530396797e-06, + "loss": 0.1862, + "step": 34576 + }, + { + "epoch": 6.29, + "learning_rate": 1.1146523480160174e-06, + "loss": 1.0721, + "step": 34577 + }, + { + "epoch": 6.29, + "learning_rate": 1.1145977429923554e-06, + "loss": 0.1536, + "step": 34578 + }, + { + "epoch": 6.29, + "learning_rate": 1.1145431379686931e-06, + "loss": 0.3449, + "step": 34579 + }, + { + "epoch": 6.29, + "learning_rate": 1.114488532945031e-06, + "loss": 1.6445, + "step": 34580 + }, + { + "epoch": 6.29, + "learning_rate": 1.1144339279213688e-06, + "loss": 1.9419, + "step": 34581 + }, + { + "epoch": 6.29, + "learning_rate": 1.1143793228977067e-06, + "loss": 0.0069, + "step": 34582 + }, + { + "epoch": 6.29, + "learning_rate": 1.1143247178740445e-06, + "loss": 1.5557, + "step": 34583 + }, + { + "epoch": 6.29, + "learning_rate": 1.1142701128503822e-06, + "loss": 0.66, + "step": 34584 + }, + { + "epoch": 6.29, + "learning_rate": 1.1142155078267201e-06, + "loss": 0.0383, + "step": 34585 + }, + { + "epoch": 6.29, + "learning_rate": 1.1141609028030579e-06, + "loss": 0.1659, + "step": 34586 + }, + { + "epoch": 6.29, + "learning_rate": 1.1141062977793958e-06, + "loss": 1.1963, + "step": 34587 + }, + { + "epoch": 6.29, + "learning_rate": 1.1140516927557335e-06, + "loss": 1.2274, + "step": 34588 + }, + { + "epoch": 6.29, + "learning_rate": 1.1139970877320715e-06, + "loss": 0.0075, + "step": 34589 + }, + { + "epoch": 6.29, + "learning_rate": 1.1139424827084092e-06, + "loss": 0.0169, + "step": 34590 + }, + { + "epoch": 6.29, + "learning_rate": 1.113887877684747e-06, + "loss": 0.1394, + "step": 34591 + }, + { + "epoch": 6.29, + "learning_rate": 1.1138332726610849e-06, + "loss": 0.3992, + "step": 34592 + }, + { + "epoch": 6.29, + "learning_rate": 1.1137786676374226e-06, + "loss": 0.3865, + "step": 34593 + }, + { + "epoch": 6.29, + "learning_rate": 1.1137240626137606e-06, + "loss": 2.1447, + "step": 34594 + }, + { + "epoch": 6.29, + "learning_rate": 1.1136694575900983e-06, + "loss": 1.1425, + "step": 34595 + }, + { + "epoch": 6.29, + "learning_rate": 1.1136148525664362e-06, + "loss": 0.0094, + "step": 34596 + }, + { + "epoch": 6.29, + "learning_rate": 1.113560247542774e-06, + "loss": 1.6663, + "step": 34597 + }, + { + "epoch": 6.29, + "learning_rate": 1.1135056425191117e-06, + "loss": 2.0496, + "step": 34598 + }, + { + "epoch": 6.29, + "learning_rate": 1.1134510374954496e-06, + "loss": 0.1263, + "step": 34599 + }, + { + "epoch": 6.29, + "learning_rate": 1.1133964324717874e-06, + "loss": 0.0359, + "step": 34600 + }, + { + "epoch": 6.29, + "learning_rate": 1.1133418274481253e-06, + "loss": 0.0262, + "step": 34601 + }, + { + "epoch": 6.29, + "learning_rate": 1.113287222424463e-06, + "loss": 0.5649, + "step": 34602 + }, + { + "epoch": 6.29, + "learning_rate": 1.113232617400801e-06, + "loss": 0.3425, + "step": 34603 + }, + { + "epoch": 6.29, + "learning_rate": 1.1131780123771387e-06, + "loss": 0.3517, + "step": 34604 + }, + { + "epoch": 6.29, + "learning_rate": 1.1131234073534765e-06, + "loss": 0.939, + "step": 34605 + }, + { + "epoch": 6.29, + "learning_rate": 1.1130688023298144e-06, + "loss": 0.0108, + "step": 34606 + }, + { + "epoch": 6.29, + "learning_rate": 1.1130141973061521e-06, + "loss": 0.6758, + "step": 34607 + }, + { + "epoch": 6.29, + "learning_rate": 1.11295959228249e-06, + "loss": 0.0115, + "step": 34608 + }, + { + "epoch": 6.29, + "learning_rate": 1.1129049872588278e-06, + "loss": 1.1667, + "step": 34609 + }, + { + "epoch": 6.29, + "learning_rate": 1.1128503822351658e-06, + "loss": 1.0387, + "step": 34610 + }, + { + "epoch": 6.29, + "learning_rate": 1.1127957772115035e-06, + "loss": 0.123, + "step": 34611 + }, + { + "epoch": 6.29, + "learning_rate": 1.1127411721878412e-06, + "loss": 0.0056, + "step": 34612 + }, + { + "epoch": 6.29, + "learning_rate": 1.1126865671641792e-06, + "loss": 0.0856, + "step": 34613 + }, + { + "epoch": 6.29, + "learning_rate": 1.1126319621405169e-06, + "loss": 0.0802, + "step": 34614 + }, + { + "epoch": 6.29, + "learning_rate": 1.1125773571168548e-06, + "loss": 0.3294, + "step": 34615 + }, + { + "epoch": 6.29, + "learning_rate": 1.1125227520931926e-06, + "loss": 0.0086, + "step": 34616 + }, + { + "epoch": 6.3, + "learning_rate": 1.1124681470695305e-06, + "loss": 0.06, + "step": 34617 + }, + { + "epoch": 6.3, + "learning_rate": 1.1124135420458682e-06, + "loss": 0.0072, + "step": 34618 + }, + { + "epoch": 6.3, + "learning_rate": 1.112358937022206e-06, + "loss": 0.6899, + "step": 34619 + }, + { + "epoch": 6.3, + "learning_rate": 1.112304331998544e-06, + "loss": 0.3386, + "step": 34620 + }, + { + "epoch": 6.3, + "learning_rate": 1.1122497269748816e-06, + "loss": 0.088, + "step": 34621 + }, + { + "epoch": 6.3, + "learning_rate": 1.1121951219512196e-06, + "loss": 0.1687, + "step": 34622 + }, + { + "epoch": 6.3, + "learning_rate": 1.1121405169275573e-06, + "loss": 0.0214, + "step": 34623 + }, + { + "epoch": 6.3, + "learning_rate": 1.1120859119038953e-06, + "loss": 0.0154, + "step": 34624 + }, + { + "epoch": 6.3, + "learning_rate": 1.112031306880233e-06, + "loss": 1.2288, + "step": 34625 + }, + { + "epoch": 6.3, + "learning_rate": 1.1119767018565707e-06, + "loss": 0.0041, + "step": 34626 + }, + { + "epoch": 6.3, + "learning_rate": 1.1119220968329087e-06, + "loss": 0.7448, + "step": 34627 + }, + { + "epoch": 6.3, + "learning_rate": 1.1118674918092466e-06, + "loss": 0.7191, + "step": 34628 + }, + { + "epoch": 6.3, + "learning_rate": 1.1118128867855843e-06, + "loss": 0.3107, + "step": 34629 + }, + { + "epoch": 6.3, + "learning_rate": 1.111758281761922e-06, + "loss": 1.3856, + "step": 34630 + }, + { + "epoch": 6.3, + "learning_rate": 1.11170367673826e-06, + "loss": 0.2341, + "step": 34631 + }, + { + "epoch": 6.3, + "learning_rate": 1.1116490717145978e-06, + "loss": 1.2684, + "step": 34632 + }, + { + "epoch": 6.3, + "learning_rate": 1.1115944666909355e-06, + "loss": 0.4771, + "step": 34633 + }, + { + "epoch": 6.3, + "learning_rate": 1.1115398616672734e-06, + "loss": 0.1519, + "step": 34634 + }, + { + "epoch": 6.3, + "learning_rate": 1.1114852566436114e-06, + "loss": 0.8093, + "step": 34635 + }, + { + "epoch": 6.3, + "learning_rate": 1.111430651619949e-06, + "loss": 0.7969, + "step": 34636 + }, + { + "epoch": 6.3, + "learning_rate": 1.1113760465962868e-06, + "loss": 0.2916, + "step": 34637 + }, + { + "epoch": 6.3, + "learning_rate": 1.1113214415726248e-06, + "loss": 0.3045, + "step": 34638 + }, + { + "epoch": 6.3, + "learning_rate": 1.1112668365489625e-06, + "loss": 1.5095, + "step": 34639 + }, + { + "epoch": 6.3, + "learning_rate": 1.1112122315253002e-06, + "loss": 0.0883, + "step": 34640 + }, + { + "epoch": 6.3, + "learning_rate": 1.1111576265016382e-06, + "loss": 0.3528, + "step": 34641 + }, + { + "epoch": 6.3, + "learning_rate": 1.1111030214779761e-06, + "loss": 0.2847, + "step": 34642 + }, + { + "epoch": 6.3, + "learning_rate": 1.1110484164543139e-06, + "loss": 0.2445, + "step": 34643 + }, + { + "epoch": 6.3, + "learning_rate": 1.1109938114306516e-06, + "loss": 0.0874, + "step": 34644 + }, + { + "epoch": 6.3, + "learning_rate": 1.1109392064069895e-06, + "loss": 0.1971, + "step": 34645 + }, + { + "epoch": 6.3, + "learning_rate": 1.1108846013833273e-06, + "loss": 0.1126, + "step": 34646 + }, + { + "epoch": 6.3, + "learning_rate": 1.110829996359665e-06, + "loss": 0.0383, + "step": 34647 + }, + { + "epoch": 6.3, + "learning_rate": 1.110775391336003e-06, + "loss": 0.4651, + "step": 34648 + }, + { + "epoch": 6.3, + "learning_rate": 1.1107207863123409e-06, + "loss": 0.3382, + "step": 34649 + }, + { + "epoch": 6.3, + "learning_rate": 1.1106661812886786e-06, + "loss": 0.32, + "step": 34650 + }, + { + "epoch": 6.3, + "learning_rate": 1.1106115762650163e-06, + "loss": 1.0551, + "step": 34651 + }, + { + "epoch": 6.3, + "learning_rate": 1.1105569712413543e-06, + "loss": 0.1118, + "step": 34652 + }, + { + "epoch": 6.3, + "learning_rate": 1.110502366217692e-06, + "loss": 3.6783, + "step": 34653 + }, + { + "epoch": 6.3, + "learning_rate": 1.1104477611940298e-06, + "loss": 0.5128, + "step": 34654 + }, + { + "epoch": 6.3, + "learning_rate": 1.1103931561703677e-06, + "loss": 1.4263, + "step": 34655 + }, + { + "epoch": 6.3, + "learning_rate": 1.1103385511467056e-06, + "loss": 1.0599, + "step": 34656 + }, + { + "epoch": 6.3, + "learning_rate": 1.1102839461230434e-06, + "loss": 0.0599, + "step": 34657 + }, + { + "epoch": 6.3, + "learning_rate": 1.1102293410993811e-06, + "loss": 1.1341, + "step": 34658 + }, + { + "epoch": 6.3, + "learning_rate": 1.110174736075719e-06, + "loss": 0.0122, + "step": 34659 + }, + { + "epoch": 6.3, + "learning_rate": 1.1101201310520568e-06, + "loss": 0.5614, + "step": 34660 + }, + { + "epoch": 6.3, + "learning_rate": 1.1100655260283945e-06, + "loss": 0.0568, + "step": 34661 + }, + { + "epoch": 6.3, + "learning_rate": 1.1100109210047325e-06, + "loss": 0.3015, + "step": 34662 + }, + { + "epoch": 6.3, + "learning_rate": 1.1099563159810704e-06, + "loss": 0.2796, + "step": 34663 + }, + { + "epoch": 6.3, + "learning_rate": 1.1099017109574081e-06, + "loss": 1.8622, + "step": 34664 + }, + { + "epoch": 6.3, + "learning_rate": 1.1098471059337459e-06, + "loss": 0.9086, + "step": 34665 + }, + { + "epoch": 6.3, + "learning_rate": 1.1097925009100838e-06, + "loss": 0.0112, + "step": 34666 + }, + { + "epoch": 6.3, + "learning_rate": 1.1097378958864215e-06, + "loss": 0.9272, + "step": 34667 + }, + { + "epoch": 6.3, + "learning_rate": 1.1096832908627595e-06, + "loss": 0.174, + "step": 34668 + }, + { + "epoch": 6.3, + "learning_rate": 1.1096286858390972e-06, + "loss": 0.7134, + "step": 34669 + }, + { + "epoch": 6.3, + "learning_rate": 1.1095740808154352e-06, + "loss": 0.2055, + "step": 34670 + }, + { + "epoch": 6.3, + "learning_rate": 1.1095194757917729e-06, + "loss": 0.0166, + "step": 34671 + }, + { + "epoch": 6.31, + "learning_rate": 1.1094648707681106e-06, + "loss": 0.3567, + "step": 34672 + }, + { + "epoch": 6.31, + "learning_rate": 1.1094102657444486e-06, + "loss": 0.9159, + "step": 34673 + }, + { + "epoch": 6.31, + "learning_rate": 1.1093556607207863e-06, + "loss": 0.1653, + "step": 34674 + }, + { + "epoch": 6.31, + "learning_rate": 1.1093010556971242e-06, + "loss": 3.4244, + "step": 34675 + }, + { + "epoch": 6.31, + "learning_rate": 1.109246450673462e-06, + "loss": 1.0314, + "step": 34676 + }, + { + "epoch": 6.31, + "learning_rate": 1.1091918456498e-06, + "loss": 0.1265, + "step": 34677 + }, + { + "epoch": 6.31, + "learning_rate": 1.1091372406261376e-06, + "loss": 0.6088, + "step": 34678 + }, + { + "epoch": 6.31, + "learning_rate": 1.1090826356024754e-06, + "loss": 0.004, + "step": 34679 + }, + { + "epoch": 6.31, + "learning_rate": 1.1090280305788133e-06, + "loss": 0.496, + "step": 34680 + }, + { + "epoch": 6.31, + "learning_rate": 1.108973425555151e-06, + "loss": 0.0781, + "step": 34681 + }, + { + "epoch": 6.31, + "learning_rate": 1.108918820531489e-06, + "loss": 0.0363, + "step": 34682 + }, + { + "epoch": 6.31, + "learning_rate": 1.1088642155078267e-06, + "loss": 0.2431, + "step": 34683 + }, + { + "epoch": 6.31, + "learning_rate": 1.1088096104841647e-06, + "loss": 2.1273, + "step": 34684 + }, + { + "epoch": 6.31, + "learning_rate": 1.1087550054605024e-06, + "loss": 0.0743, + "step": 34685 + }, + { + "epoch": 6.31, + "learning_rate": 1.1087004004368401e-06, + "loss": 0.4045, + "step": 34686 + }, + { + "epoch": 6.31, + "learning_rate": 1.108645795413178e-06, + "loss": 0.0859, + "step": 34687 + }, + { + "epoch": 6.31, + "learning_rate": 1.1085911903895158e-06, + "loss": 0.2732, + "step": 34688 + }, + { + "epoch": 6.31, + "learning_rate": 1.1085365853658538e-06, + "loss": 0.5722, + "step": 34689 + }, + { + "epoch": 6.31, + "learning_rate": 1.1084819803421915e-06, + "loss": 0.1949, + "step": 34690 + }, + { + "epoch": 6.31, + "learning_rate": 1.1084273753185294e-06, + "loss": 0.2098, + "step": 34691 + }, + { + "epoch": 6.31, + "learning_rate": 1.1083727702948672e-06, + "loss": 0.8669, + "step": 34692 + }, + { + "epoch": 6.31, + "learning_rate": 1.1083181652712049e-06, + "loss": 1.3199, + "step": 34693 + }, + { + "epoch": 6.31, + "learning_rate": 1.1082635602475428e-06, + "loss": 0.0337, + "step": 34694 + }, + { + "epoch": 6.31, + "learning_rate": 1.1082089552238806e-06, + "loss": 0.1321, + "step": 34695 + }, + { + "epoch": 6.31, + "learning_rate": 1.1081543502002185e-06, + "loss": 0.6264, + "step": 34696 + }, + { + "epoch": 6.31, + "learning_rate": 1.1080997451765562e-06, + "loss": 0.4483, + "step": 34697 + }, + { + "epoch": 6.31, + "learning_rate": 1.1080451401528942e-06, + "loss": 0.0991, + "step": 34698 + }, + { + "epoch": 6.31, + "learning_rate": 1.107990535129232e-06, + "loss": 1.3459, + "step": 34699 + }, + { + "epoch": 6.31, + "learning_rate": 1.1079359301055696e-06, + "loss": 0.1567, + "step": 34700 + }, + { + "epoch": 6.31, + "learning_rate": 1.1078813250819074e-06, + "loss": 2.0808, + "step": 34701 + }, + { + "epoch": 6.31, + "learning_rate": 1.1078267200582453e-06, + "loss": 0.1438, + "step": 34702 + }, + { + "epoch": 6.31, + "learning_rate": 1.1077721150345833e-06, + "loss": 0.4311, + "step": 34703 + }, + { + "epoch": 6.31, + "learning_rate": 1.107717510010921e-06, + "loss": 2.6603, + "step": 34704 + }, + { + "epoch": 6.31, + "learning_rate": 1.107662904987259e-06, + "loss": 0.1162, + "step": 34705 + }, + { + "epoch": 6.31, + "learning_rate": 1.1076082999635967e-06, + "loss": 1.1473, + "step": 34706 + }, + { + "epoch": 6.31, + "learning_rate": 1.1075536949399344e-06, + "loss": 0.1713, + "step": 34707 + }, + { + "epoch": 6.31, + "learning_rate": 1.1074990899162723e-06, + "loss": 0.825, + "step": 34708 + }, + { + "epoch": 6.31, + "learning_rate": 1.1074444848926103e-06, + "loss": 0.3043, + "step": 34709 + }, + { + "epoch": 6.31, + "learning_rate": 1.107389879868948e-06, + "loss": 1.0404, + "step": 34710 + }, + { + "epoch": 6.31, + "learning_rate": 1.1073352748452858e-06, + "loss": 0.6464, + "step": 34711 + }, + { + "epoch": 6.31, + "learning_rate": 1.1072806698216237e-06, + "loss": 0.3315, + "step": 34712 + }, + { + "epoch": 6.31, + "learning_rate": 1.1072260647979614e-06, + "loss": 0.0045, + "step": 34713 + }, + { + "epoch": 6.31, + "learning_rate": 1.1071714597742992e-06, + "loss": 0.0364, + "step": 34714 + }, + { + "epoch": 6.31, + "learning_rate": 1.107116854750637e-06, + "loss": 1.8721, + "step": 34715 + }, + { + "epoch": 6.31, + "learning_rate": 1.107062249726975e-06, + "loss": 1.9752, + "step": 34716 + }, + { + "epoch": 6.31, + "learning_rate": 1.1070076447033128e-06, + "loss": 0.7895, + "step": 34717 + }, + { + "epoch": 6.31, + "learning_rate": 1.1069530396796505e-06, + "loss": 1.0491, + "step": 34718 + }, + { + "epoch": 6.31, + "learning_rate": 1.1068984346559885e-06, + "loss": 0.0069, + "step": 34719 + }, + { + "epoch": 6.31, + "learning_rate": 1.1068438296323262e-06, + "loss": 0.0106, + "step": 34720 + }, + { + "epoch": 6.31, + "learning_rate": 1.106789224608664e-06, + "loss": 0.1323, + "step": 34721 + }, + { + "epoch": 6.31, + "learning_rate": 1.1067346195850019e-06, + "loss": 0.3352, + "step": 34722 + }, + { + "epoch": 6.31, + "learning_rate": 1.1066800145613398e-06, + "loss": 0.0678, + "step": 34723 + }, + { + "epoch": 6.31, + "learning_rate": 1.1066254095376775e-06, + "loss": 0.4226, + "step": 34724 + }, + { + "epoch": 6.31, + "learning_rate": 1.1065708045140153e-06, + "loss": 1.4221, + "step": 34725 + }, + { + "epoch": 6.31, + "learning_rate": 1.1065161994903532e-06, + "loss": 0.17, + "step": 34726 + }, + { + "epoch": 6.32, + "learning_rate": 1.106461594466691e-06, + "loss": 0.3494, + "step": 34727 + }, + { + "epoch": 6.32, + "learning_rate": 1.1064069894430287e-06, + "loss": 0.5818, + "step": 34728 + }, + { + "epoch": 6.32, + "learning_rate": 1.1063523844193666e-06, + "loss": 0.682, + "step": 34729 + }, + { + "epoch": 6.32, + "learning_rate": 1.1062977793957046e-06, + "loss": 0.0099, + "step": 34730 + }, + { + "epoch": 6.32, + "learning_rate": 1.1062431743720423e-06, + "loss": 0.716, + "step": 34731 + }, + { + "epoch": 6.32, + "learning_rate": 1.10618856934838e-06, + "loss": 0.7118, + "step": 34732 + }, + { + "epoch": 6.32, + "learning_rate": 1.106133964324718e-06, + "loss": 0.515, + "step": 34733 + }, + { + "epoch": 6.32, + "learning_rate": 1.1060793593010557e-06, + "loss": 0.4722, + "step": 34734 + }, + { + "epoch": 6.32, + "learning_rate": 1.1060247542773934e-06, + "loss": 1.4074, + "step": 34735 + }, + { + "epoch": 6.32, + "learning_rate": 1.1059701492537314e-06, + "loss": 0.0192, + "step": 34736 + }, + { + "epoch": 6.32, + "learning_rate": 1.1059155442300693e-06, + "loss": 0.0464, + "step": 34737 + }, + { + "epoch": 6.32, + "learning_rate": 1.105860939206407e-06, + "loss": 0.2815, + "step": 34738 + }, + { + "epoch": 6.32, + "learning_rate": 1.1058063341827448e-06, + "loss": 0.2273, + "step": 34739 + }, + { + "epoch": 6.32, + "learning_rate": 1.1057517291590827e-06, + "loss": 1.6314, + "step": 34740 + }, + { + "epoch": 6.32, + "learning_rate": 1.1056971241354205e-06, + "loss": 0.0192, + "step": 34741 + }, + { + "epoch": 6.32, + "learning_rate": 1.1056425191117582e-06, + "loss": 0.0099, + "step": 34742 + }, + { + "epoch": 6.32, + "learning_rate": 1.1055879140880961e-06, + "loss": 0.22, + "step": 34743 + }, + { + "epoch": 6.32, + "learning_rate": 1.105533309064434e-06, + "loss": 1.6361, + "step": 34744 + }, + { + "epoch": 6.32, + "learning_rate": 1.1054787040407718e-06, + "loss": 0.2618, + "step": 34745 + }, + { + "epoch": 6.32, + "learning_rate": 1.1054240990171095e-06, + "loss": 1.1622, + "step": 34746 + }, + { + "epoch": 6.32, + "learning_rate": 1.1053694939934475e-06, + "loss": 0.3314, + "step": 34747 + }, + { + "epoch": 6.32, + "learning_rate": 1.1053148889697852e-06, + "loss": 1.8289, + "step": 34748 + }, + { + "epoch": 6.32, + "learning_rate": 1.1052602839461232e-06, + "loss": 1.3215, + "step": 34749 + }, + { + "epoch": 6.32, + "learning_rate": 1.1052056789224609e-06, + "loss": 0.4242, + "step": 34750 + }, + { + "epoch": 6.32, + "learning_rate": 1.1051510738987988e-06, + "loss": 0.035, + "step": 34751 + }, + { + "epoch": 6.32, + "learning_rate": 1.1050964688751366e-06, + "loss": 1.2734, + "step": 34752 + }, + { + "epoch": 6.32, + "learning_rate": 1.1050418638514743e-06, + "loss": 0.0896, + "step": 34753 + }, + { + "epoch": 6.32, + "learning_rate": 1.1049872588278122e-06, + "loss": 0.4241, + "step": 34754 + }, + { + "epoch": 6.32, + "learning_rate": 1.10493265380415e-06, + "loss": 0.8703, + "step": 34755 + }, + { + "epoch": 6.32, + "learning_rate": 1.104878048780488e-06, + "loss": 2.5032, + "step": 34756 + }, + { + "epoch": 6.32, + "learning_rate": 1.1048234437568256e-06, + "loss": 0.8093, + "step": 34757 + }, + { + "epoch": 6.32, + "learning_rate": 1.1047688387331636e-06, + "loss": 0.6694, + "step": 34758 + }, + { + "epoch": 6.32, + "learning_rate": 1.1047142337095013e-06, + "loss": 0.0081, + "step": 34759 + }, + { + "epoch": 6.32, + "learning_rate": 1.104659628685839e-06, + "loss": 1.909, + "step": 34760 + }, + { + "epoch": 6.32, + "learning_rate": 1.104605023662177e-06, + "loss": 0.2525, + "step": 34761 + }, + { + "epoch": 6.32, + "learning_rate": 1.1045504186385147e-06, + "loss": 1.2725, + "step": 34762 + }, + { + "epoch": 6.32, + "learning_rate": 1.1044958136148527e-06, + "loss": 0.7621, + "step": 34763 + }, + { + "epoch": 6.32, + "learning_rate": 1.1044412085911904e-06, + "loss": 0.0603, + "step": 34764 + }, + { + "epoch": 6.32, + "learning_rate": 1.1043866035675283e-06, + "loss": 0.5137, + "step": 34765 + }, + { + "epoch": 6.32, + "learning_rate": 1.104331998543866e-06, + "loss": 0.076, + "step": 34766 + }, + { + "epoch": 6.32, + "learning_rate": 1.1042773935202038e-06, + "loss": 0.6032, + "step": 34767 + }, + { + "epoch": 6.32, + "learning_rate": 1.1042227884965418e-06, + "loss": 0.0057, + "step": 34768 + }, + { + "epoch": 6.32, + "learning_rate": 1.1041681834728795e-06, + "loss": 0.3412, + "step": 34769 + }, + { + "epoch": 6.32, + "learning_rate": 1.1041135784492174e-06, + "loss": 0.3066, + "step": 34770 + }, + { + "epoch": 6.32, + "learning_rate": 1.1040589734255552e-06, + "loss": 2.5783, + "step": 34771 + }, + { + "epoch": 6.32, + "learning_rate": 1.104004368401893e-06, + "loss": 1.1348, + "step": 34772 + }, + { + "epoch": 6.32, + "learning_rate": 1.1039497633782308e-06, + "loss": 0.1554, + "step": 34773 + }, + { + "epoch": 6.32, + "learning_rate": 1.1038951583545686e-06, + "loss": 0.0061, + "step": 34774 + }, + { + "epoch": 6.32, + "learning_rate": 1.1038405533309065e-06, + "loss": 1.4816, + "step": 34775 + }, + { + "epoch": 6.32, + "learning_rate": 1.1037859483072442e-06, + "loss": 0.4777, + "step": 34776 + }, + { + "epoch": 6.32, + "learning_rate": 1.1037313432835822e-06, + "loss": 0.0703, + "step": 34777 + }, + { + "epoch": 6.32, + "learning_rate": 1.10367673825992e-06, + "loss": 0.0818, + "step": 34778 + }, + { + "epoch": 6.32, + "learning_rate": 1.1036221332362579e-06, + "loss": 0.2525, + "step": 34779 + }, + { + "epoch": 6.32, + "learning_rate": 1.1035675282125956e-06, + "loss": 0.0057, + "step": 34780 + }, + { + "epoch": 6.32, + "learning_rate": 1.1035129231889333e-06, + "loss": 0.0462, + "step": 34781 + }, + { + "epoch": 6.33, + "learning_rate": 1.103458318165271e-06, + "loss": 0.1541, + "step": 34782 + }, + { + "epoch": 6.33, + "learning_rate": 1.103403713141609e-06, + "loss": 0.4108, + "step": 34783 + }, + { + "epoch": 6.33, + "learning_rate": 1.103349108117947e-06, + "loss": 0.2081, + "step": 34784 + }, + { + "epoch": 6.33, + "learning_rate": 1.1032945030942847e-06, + "loss": 1.7393, + "step": 34785 + }, + { + "epoch": 6.33, + "learning_rate": 1.1032398980706226e-06, + "loss": 0.172, + "step": 34786 + }, + { + "epoch": 6.33, + "learning_rate": 1.1031852930469603e-06, + "loss": 0.3509, + "step": 34787 + }, + { + "epoch": 6.33, + "learning_rate": 1.103130688023298e-06, + "loss": 0.0146, + "step": 34788 + }, + { + "epoch": 6.33, + "learning_rate": 1.103076082999636e-06, + "loss": 0.1008, + "step": 34789 + }, + { + "epoch": 6.33, + "learning_rate": 1.1030214779759738e-06, + "loss": 0.0759, + "step": 34790 + }, + { + "epoch": 6.33, + "learning_rate": 1.1029668729523117e-06, + "loss": 1.6391, + "step": 34791 + }, + { + "epoch": 6.33, + "learning_rate": 1.1029122679286494e-06, + "loss": 0.2095, + "step": 34792 + }, + { + "epoch": 6.33, + "learning_rate": 1.1028576629049874e-06, + "loss": 0.3026, + "step": 34793 + }, + { + "epoch": 6.33, + "learning_rate": 1.102803057881325e-06, + "loss": 1.0013, + "step": 34794 + }, + { + "epoch": 6.33, + "learning_rate": 1.1027484528576628e-06, + "loss": 1.3077, + "step": 34795 + }, + { + "epoch": 6.33, + "learning_rate": 1.1026938478340008e-06, + "loss": 0.2683, + "step": 34796 + }, + { + "epoch": 6.33, + "learning_rate": 1.1026392428103387e-06, + "loss": 2.1025, + "step": 34797 + }, + { + "epoch": 6.33, + "learning_rate": 1.1025846377866765e-06, + "loss": 1.911, + "step": 34798 + }, + { + "epoch": 6.33, + "learning_rate": 1.1025300327630142e-06, + "loss": 0.0506, + "step": 34799 + }, + { + "epoch": 6.33, + "learning_rate": 1.1024754277393521e-06, + "loss": 0.0108, + "step": 34800 + }, + { + "epoch": 6.33, + "learning_rate": 1.1024208227156899e-06, + "loss": 1.6181, + "step": 34801 + }, + { + "epoch": 6.33, + "learning_rate": 1.1023662176920276e-06, + "loss": 0.012, + "step": 34802 + }, + { + "epoch": 6.33, + "learning_rate": 1.1023116126683655e-06, + "loss": 0.1434, + "step": 34803 + }, + { + "epoch": 6.33, + "learning_rate": 1.1022570076447035e-06, + "loss": 2.0454, + "step": 34804 + }, + { + "epoch": 6.33, + "learning_rate": 1.1022024026210412e-06, + "loss": 0.1728, + "step": 34805 + }, + { + "epoch": 6.33, + "learning_rate": 1.102147797597379e-06, + "loss": 0.0106, + "step": 34806 + }, + { + "epoch": 6.33, + "learning_rate": 1.1020931925737169e-06, + "loss": 0.4294, + "step": 34807 + }, + { + "epoch": 6.33, + "learning_rate": 1.1020385875500546e-06, + "loss": 0.0592, + "step": 34808 + }, + { + "epoch": 6.33, + "learning_rate": 1.1019839825263923e-06, + "loss": 0.0042, + "step": 34809 + }, + { + "epoch": 6.33, + "learning_rate": 1.1019293775027303e-06, + "loss": 0.1256, + "step": 34810 + }, + { + "epoch": 6.33, + "learning_rate": 1.1018747724790682e-06, + "loss": 0.3248, + "step": 34811 + }, + { + "epoch": 6.33, + "learning_rate": 1.101820167455406e-06, + "loss": 0.1898, + "step": 34812 + }, + { + "epoch": 6.33, + "learning_rate": 1.1017655624317437e-06, + "loss": 2.4646, + "step": 34813 + }, + { + "epoch": 6.33, + "learning_rate": 1.1017109574080816e-06, + "loss": 0.87, + "step": 34814 + }, + { + "epoch": 6.33, + "learning_rate": 1.1016563523844194e-06, + "loss": 0.0726, + "step": 34815 + }, + { + "epoch": 6.33, + "learning_rate": 1.101601747360757e-06, + "loss": 0.0069, + "step": 34816 + }, + { + "epoch": 6.33, + "learning_rate": 1.101547142337095e-06, + "loss": 0.0803, + "step": 34817 + }, + { + "epoch": 6.33, + "learning_rate": 1.101492537313433e-06, + "loss": 1.448, + "step": 34818 + }, + { + "epoch": 6.33, + "learning_rate": 1.1014379322897707e-06, + "loss": 0.125, + "step": 34819 + }, + { + "epoch": 6.33, + "learning_rate": 1.1013833272661085e-06, + "loss": 0.0164, + "step": 34820 + }, + { + "epoch": 6.33, + "learning_rate": 1.1013287222424464e-06, + "loss": 0.3993, + "step": 34821 + }, + { + "epoch": 6.33, + "learning_rate": 1.1012741172187841e-06, + "loss": 1.6741, + "step": 34822 + }, + { + "epoch": 6.33, + "learning_rate": 1.1012195121951219e-06, + "loss": 0.2871, + "step": 34823 + }, + { + "epoch": 6.33, + "learning_rate": 1.1011649071714598e-06, + "loss": 0.0077, + "step": 34824 + }, + { + "epoch": 6.33, + "learning_rate": 1.1011103021477977e-06, + "loss": 0.1952, + "step": 34825 + }, + { + "epoch": 6.33, + "learning_rate": 1.1010556971241355e-06, + "loss": 3.7672, + "step": 34826 + }, + { + "epoch": 6.33, + "learning_rate": 1.1010010921004732e-06, + "loss": 0.6921, + "step": 34827 + }, + { + "epoch": 6.33, + "learning_rate": 1.1009464870768112e-06, + "loss": 1.5914, + "step": 34828 + }, + { + "epoch": 6.33, + "learning_rate": 1.1008918820531489e-06, + "loss": 0.1907, + "step": 34829 + }, + { + "epoch": 6.33, + "learning_rate": 1.1008372770294866e-06, + "loss": 0.019, + "step": 34830 + }, + { + "epoch": 6.33, + "learning_rate": 1.1007826720058246e-06, + "loss": 2.0379, + "step": 34831 + }, + { + "epoch": 6.33, + "learning_rate": 1.1007280669821625e-06, + "loss": 0.008, + "step": 34832 + }, + { + "epoch": 6.33, + "learning_rate": 1.1006734619585002e-06, + "loss": 0.3332, + "step": 34833 + }, + { + "epoch": 6.33, + "learning_rate": 1.100618856934838e-06, + "loss": 0.19, + "step": 34834 + }, + { + "epoch": 6.33, + "learning_rate": 1.100564251911176e-06, + "loss": 0.0104, + "step": 34835 + }, + { + "epoch": 6.33, + "learning_rate": 1.1005096468875136e-06, + "loss": 0.0176, + "step": 34836 + }, + { + "epoch": 6.34, + "learning_rate": 1.1004550418638516e-06, + "loss": 0.0181, + "step": 34837 + }, + { + "epoch": 6.34, + "learning_rate": 1.1004004368401893e-06, + "loss": 0.5122, + "step": 34838 + }, + { + "epoch": 6.34, + "learning_rate": 1.1003458318165273e-06, + "loss": 1.285, + "step": 34839 + }, + { + "epoch": 6.34, + "learning_rate": 1.100291226792865e-06, + "loss": 0.6596, + "step": 34840 + }, + { + "epoch": 6.34, + "learning_rate": 1.1002366217692027e-06, + "loss": 1.0188, + "step": 34841 + }, + { + "epoch": 6.34, + "learning_rate": 1.1001820167455407e-06, + "loss": 0.3776, + "step": 34842 + }, + { + "epoch": 6.34, + "learning_rate": 1.1001274117218784e-06, + "loss": 0.1061, + "step": 34843 + }, + { + "epoch": 6.34, + "learning_rate": 1.1000728066982163e-06, + "loss": 0.0724, + "step": 34844 + }, + { + "epoch": 6.34, + "learning_rate": 1.100018201674554e-06, + "loss": 0.6301, + "step": 34845 + }, + { + "epoch": 6.34, + "learning_rate": 1.099963596650892e-06, + "loss": 0.4255, + "step": 34846 + }, + { + "epoch": 6.34, + "learning_rate": 1.0999089916272298e-06, + "loss": 0.0195, + "step": 34847 + }, + { + "epoch": 6.34, + "learning_rate": 1.0998543866035675e-06, + "loss": 0.0252, + "step": 34848 + }, + { + "epoch": 6.34, + "learning_rate": 1.0997997815799054e-06, + "loss": 0.1747, + "step": 34849 + }, + { + "epoch": 6.34, + "learning_rate": 1.0997451765562432e-06, + "loss": 0.2018, + "step": 34850 + }, + { + "epoch": 6.34, + "learning_rate": 1.099690571532581e-06, + "loss": 0.0975, + "step": 34851 + }, + { + "epoch": 6.34, + "learning_rate": 1.0996359665089188e-06, + "loss": 0.0543, + "step": 34852 + }, + { + "epoch": 6.34, + "learning_rate": 1.0995813614852568e-06, + "loss": 0.425, + "step": 34853 + }, + { + "epoch": 6.34, + "learning_rate": 1.0995267564615945e-06, + "loss": 2.0227, + "step": 34854 + }, + { + "epoch": 6.34, + "learning_rate": 1.0994721514379322e-06, + "loss": 0.0106, + "step": 34855 + }, + { + "epoch": 6.34, + "learning_rate": 1.09941754641427e-06, + "loss": 0.4003, + "step": 34856 + }, + { + "epoch": 6.34, + "learning_rate": 1.099362941390608e-06, + "loss": 4.4347, + "step": 34857 + }, + { + "epoch": 6.34, + "learning_rate": 1.0993083363669459e-06, + "loss": 0.6826, + "step": 34858 + }, + { + "epoch": 6.34, + "learning_rate": 1.0992537313432836e-06, + "loss": 1.351, + "step": 34859 + }, + { + "epoch": 6.34, + "learning_rate": 1.0991991263196215e-06, + "loss": 0.0223, + "step": 34860 + }, + { + "epoch": 6.34, + "learning_rate": 1.0991445212959593e-06, + "loss": 0.6225, + "step": 34861 + }, + { + "epoch": 6.34, + "learning_rate": 1.099089916272297e-06, + "loss": 0.3565, + "step": 34862 + }, + { + "epoch": 6.34, + "learning_rate": 1.0990353112486347e-06, + "loss": 0.6081, + "step": 34863 + }, + { + "epoch": 6.34, + "learning_rate": 1.0989807062249727e-06, + "loss": 0.0734, + "step": 34864 + }, + { + "epoch": 6.34, + "learning_rate": 1.0989261012013106e-06, + "loss": 0.0134, + "step": 34865 + }, + { + "epoch": 6.34, + "learning_rate": 1.0988714961776483e-06, + "loss": 0.2805, + "step": 34866 + }, + { + "epoch": 6.34, + "learning_rate": 1.0988168911539863e-06, + "loss": 0.1127, + "step": 34867 + }, + { + "epoch": 6.34, + "learning_rate": 1.098762286130324e-06, + "loss": 1.2034, + "step": 34868 + }, + { + "epoch": 6.34, + "learning_rate": 1.0987076811066618e-06, + "loss": 0.5778, + "step": 34869 + }, + { + "epoch": 6.34, + "learning_rate": 1.0986530760829995e-06, + "loss": 0.1039, + "step": 34870 + }, + { + "epoch": 6.34, + "learning_rate": 1.0985984710593374e-06, + "loss": 0.18, + "step": 34871 + }, + { + "epoch": 6.34, + "learning_rate": 1.0985438660356754e-06, + "loss": 0.1307, + "step": 34872 + }, + { + "epoch": 6.34, + "learning_rate": 1.098489261012013e-06, + "loss": 1.0185, + "step": 34873 + }, + { + "epoch": 6.34, + "learning_rate": 1.098434655988351e-06, + "loss": 0.5735, + "step": 34874 + }, + { + "epoch": 6.34, + "learning_rate": 1.0983800509646888e-06, + "loss": 0.0078, + "step": 34875 + }, + { + "epoch": 6.34, + "learning_rate": 1.0983254459410265e-06, + "loss": 0.0803, + "step": 34876 + }, + { + "epoch": 6.34, + "learning_rate": 1.0982708409173645e-06, + "loss": 0.2591, + "step": 34877 + }, + { + "epoch": 6.34, + "learning_rate": 1.0982162358937024e-06, + "loss": 2.2912, + "step": 34878 + }, + { + "epoch": 6.34, + "learning_rate": 1.0981616308700401e-06, + "loss": 0.2217, + "step": 34879 + }, + { + "epoch": 6.34, + "learning_rate": 1.0981070258463779e-06, + "loss": 0.9771, + "step": 34880 + }, + { + "epoch": 6.34, + "learning_rate": 1.0980524208227158e-06, + "loss": 0.0607, + "step": 34881 + }, + { + "epoch": 6.34, + "learning_rate": 1.0979978157990535e-06, + "loss": 0.0437, + "step": 34882 + }, + { + "epoch": 6.34, + "learning_rate": 1.0979432107753913e-06, + "loss": 0.0043, + "step": 34883 + }, + { + "epoch": 6.34, + "learning_rate": 1.0978886057517292e-06, + "loss": 1.5534, + "step": 34884 + }, + { + "epoch": 6.34, + "learning_rate": 1.0978340007280672e-06, + "loss": 0.4345, + "step": 34885 + }, + { + "epoch": 6.34, + "learning_rate": 1.0977793957044049e-06, + "loss": 0.8857, + "step": 34886 + }, + { + "epoch": 6.34, + "learning_rate": 1.0977247906807426e-06, + "loss": 0.019, + "step": 34887 + }, + { + "epoch": 6.34, + "learning_rate": 1.0976701856570806e-06, + "loss": 0.0563, + "step": 34888 + }, + { + "epoch": 6.34, + "learning_rate": 1.0976155806334183e-06, + "loss": 0.2007, + "step": 34889 + }, + { + "epoch": 6.34, + "learning_rate": 1.097560975609756e-06, + "loss": 0.0042, + "step": 34890 + }, + { + "epoch": 6.34, + "learning_rate": 1.097506370586094e-06, + "loss": 0.4717, + "step": 34891 + }, + { + "epoch": 6.35, + "learning_rate": 1.097451765562432e-06, + "loss": 0.0201, + "step": 34892 + }, + { + "epoch": 6.35, + "learning_rate": 1.0973971605387696e-06, + "loss": 0.2071, + "step": 34893 + }, + { + "epoch": 6.35, + "learning_rate": 1.0973425555151074e-06, + "loss": 0.7099, + "step": 34894 + }, + { + "epoch": 6.35, + "learning_rate": 1.0972879504914453e-06, + "loss": 0.8717, + "step": 34895 + }, + { + "epoch": 6.35, + "learning_rate": 1.097233345467783e-06, + "loss": 1.8558, + "step": 34896 + }, + { + "epoch": 6.35, + "learning_rate": 1.0971787404441208e-06, + "loss": 1.056, + "step": 34897 + }, + { + "epoch": 6.35, + "learning_rate": 1.0971241354204587e-06, + "loss": 0.0163, + "step": 34898 + }, + { + "epoch": 6.35, + "learning_rate": 1.0970695303967967e-06, + "loss": 0.4706, + "step": 34899 + }, + { + "epoch": 6.35, + "learning_rate": 1.0970149253731344e-06, + "loss": 1.7789, + "step": 34900 + }, + { + "epoch": 6.35, + "learning_rate": 1.0969603203494721e-06, + "loss": 0.004, + "step": 34901 + }, + { + "epoch": 6.35, + "learning_rate": 1.09690571532581e-06, + "loss": 0.3154, + "step": 34902 + }, + { + "epoch": 6.35, + "learning_rate": 1.0968511103021478e-06, + "loss": 0.1011, + "step": 34903 + }, + { + "epoch": 6.35, + "learning_rate": 1.0967965052784855e-06, + "loss": 1.6226, + "step": 34904 + }, + { + "epoch": 6.35, + "learning_rate": 1.0967419002548235e-06, + "loss": 0.024, + "step": 34905 + }, + { + "epoch": 6.35, + "learning_rate": 1.0966872952311614e-06, + "loss": 0.3779, + "step": 34906 + }, + { + "epoch": 6.35, + "learning_rate": 1.0966326902074992e-06, + "loss": 0.0316, + "step": 34907 + }, + { + "epoch": 6.35, + "learning_rate": 1.0965780851838369e-06, + "loss": 0.4047, + "step": 34908 + }, + { + "epoch": 6.35, + "learning_rate": 1.0965234801601748e-06, + "loss": 1.086, + "step": 34909 + }, + { + "epoch": 6.35, + "learning_rate": 1.0964688751365126e-06, + "loss": 0.0246, + "step": 34910 + }, + { + "epoch": 6.35, + "learning_rate": 1.0964142701128503e-06, + "loss": 0.1219, + "step": 34911 + }, + { + "epoch": 6.35, + "learning_rate": 1.0963596650891882e-06, + "loss": 0.0908, + "step": 34912 + }, + { + "epoch": 6.35, + "learning_rate": 1.0963050600655262e-06, + "loss": 0.7739, + "step": 34913 + }, + { + "epoch": 6.35, + "learning_rate": 1.096250455041864e-06, + "loss": 0.4984, + "step": 34914 + }, + { + "epoch": 6.35, + "learning_rate": 1.0961958500182016e-06, + "loss": 0.0176, + "step": 34915 + }, + { + "epoch": 6.35, + "learning_rate": 1.0961412449945396e-06, + "loss": 0.5519, + "step": 34916 + }, + { + "epoch": 6.35, + "learning_rate": 1.0960866399708773e-06, + "loss": 0.2323, + "step": 34917 + }, + { + "epoch": 6.35, + "learning_rate": 1.0960320349472153e-06, + "loss": 0.012, + "step": 34918 + }, + { + "epoch": 6.35, + "learning_rate": 1.095977429923553e-06, + "loss": 1.6517, + "step": 34919 + }, + { + "epoch": 6.35, + "learning_rate": 1.095922824899891e-06, + "loss": 0.1854, + "step": 34920 + }, + { + "epoch": 6.35, + "learning_rate": 1.0958682198762287e-06, + "loss": 0.3589, + "step": 34921 + }, + { + "epoch": 6.35, + "learning_rate": 1.0958136148525664e-06, + "loss": 2.0549, + "step": 34922 + }, + { + "epoch": 6.35, + "learning_rate": 1.0957590098289043e-06, + "loss": 0.2707, + "step": 34923 + }, + { + "epoch": 6.35, + "learning_rate": 1.095704404805242e-06, + "loss": 0.0129, + "step": 34924 + }, + { + "epoch": 6.35, + "learning_rate": 1.09564979978158e-06, + "loss": 0.5171, + "step": 34925 + }, + { + "epoch": 6.35, + "learning_rate": 1.0955951947579177e-06, + "loss": 0.1044, + "step": 34926 + }, + { + "epoch": 6.35, + "learning_rate": 1.0955405897342557e-06, + "loss": 0.4562, + "step": 34927 + }, + { + "epoch": 6.35, + "learning_rate": 1.0954859847105934e-06, + "loss": 1.0049, + "step": 34928 + }, + { + "epoch": 6.35, + "learning_rate": 1.0954313796869312e-06, + "loss": 0.0098, + "step": 34929 + }, + { + "epoch": 6.35, + "learning_rate": 1.095376774663269e-06, + "loss": 0.1306, + "step": 34930 + }, + { + "epoch": 6.35, + "learning_rate": 1.0953221696396068e-06, + "loss": 1.9601, + "step": 34931 + }, + { + "epoch": 6.35, + "learning_rate": 1.0952675646159448e-06, + "loss": 0.0075, + "step": 34932 + }, + { + "epoch": 6.35, + "learning_rate": 1.0952129595922825e-06, + "loss": 0.0139, + "step": 34933 + }, + { + "epoch": 6.35, + "learning_rate": 1.0951583545686204e-06, + "loss": 0.8149, + "step": 34934 + }, + { + "epoch": 6.35, + "learning_rate": 1.0951037495449582e-06, + "loss": 0.5757, + "step": 34935 + }, + { + "epoch": 6.35, + "learning_rate": 1.095049144521296e-06, + "loss": 0.0682, + "step": 34936 + }, + { + "epoch": 6.35, + "learning_rate": 1.0949945394976336e-06, + "loss": 0.6079, + "step": 34937 + }, + { + "epoch": 6.35, + "learning_rate": 1.0949399344739716e-06, + "loss": 0.674, + "step": 34938 + }, + { + "epoch": 6.35, + "learning_rate": 1.0948853294503095e-06, + "loss": 0.0116, + "step": 34939 + }, + { + "epoch": 6.35, + "learning_rate": 1.0948307244266473e-06, + "loss": 0.6111, + "step": 34940 + }, + { + "epoch": 6.35, + "learning_rate": 1.0947761194029852e-06, + "loss": 0.9742, + "step": 34941 + }, + { + "epoch": 6.35, + "learning_rate": 1.094721514379323e-06, + "loss": 2.9981, + "step": 34942 + }, + { + "epoch": 6.35, + "learning_rate": 1.0946669093556607e-06, + "loss": 0.1183, + "step": 34943 + }, + { + "epoch": 6.35, + "learning_rate": 1.0946123043319984e-06, + "loss": 1.6032, + "step": 34944 + }, + { + "epoch": 6.35, + "learning_rate": 1.0945576993083363e-06, + "loss": 1.116, + "step": 34945 + }, + { + "epoch": 6.35, + "learning_rate": 1.0945030942846743e-06, + "loss": 0.0218, + "step": 34946 + }, + { + "epoch": 6.36, + "learning_rate": 1.094448489261012e-06, + "loss": 0.0543, + "step": 34947 + }, + { + "epoch": 6.36, + "learning_rate": 1.09439388423735e-06, + "loss": 0.2115, + "step": 34948 + }, + { + "epoch": 6.36, + "learning_rate": 1.0943392792136877e-06, + "loss": 2.2996, + "step": 34949 + }, + { + "epoch": 6.36, + "learning_rate": 1.0942846741900254e-06, + "loss": 0.4061, + "step": 34950 + }, + { + "epoch": 6.36, + "learning_rate": 1.0942300691663632e-06, + "loss": 0.1484, + "step": 34951 + }, + { + "epoch": 6.36, + "learning_rate": 1.094175464142701e-06, + "loss": 0.1653, + "step": 34952 + }, + { + "epoch": 6.36, + "learning_rate": 1.094120859119039e-06, + "loss": 0.0732, + "step": 34953 + }, + { + "epoch": 6.36, + "learning_rate": 1.0940662540953768e-06, + "loss": 0.5195, + "step": 34954 + }, + { + "epoch": 6.36, + "learning_rate": 1.0940116490717147e-06, + "loss": 0.0156, + "step": 34955 + }, + { + "epoch": 6.36, + "learning_rate": 1.0939570440480525e-06, + "loss": 0.1642, + "step": 34956 + }, + { + "epoch": 6.36, + "learning_rate": 1.0939024390243902e-06, + "loss": 0.4209, + "step": 34957 + }, + { + "epoch": 6.36, + "learning_rate": 1.093847834000728e-06, + "loss": 0.0867, + "step": 34958 + }, + { + "epoch": 6.36, + "learning_rate": 1.0937932289770659e-06, + "loss": 0.0094, + "step": 34959 + }, + { + "epoch": 6.36, + "learning_rate": 1.0937386239534038e-06, + "loss": 0.5961, + "step": 34960 + }, + { + "epoch": 6.36, + "learning_rate": 1.0936840189297415e-06, + "loss": 0.2612, + "step": 34961 + }, + { + "epoch": 6.36, + "learning_rate": 1.0936294139060795e-06, + "loss": 0.4024, + "step": 34962 + }, + { + "epoch": 6.36, + "learning_rate": 1.0935748088824172e-06, + "loss": 2.1685, + "step": 34963 + }, + { + "epoch": 6.36, + "learning_rate": 1.093520203858755e-06, + "loss": 1.7135, + "step": 34964 + }, + { + "epoch": 6.36, + "learning_rate": 1.0934655988350929e-06, + "loss": 1.8322, + "step": 34965 + }, + { + "epoch": 6.36, + "learning_rate": 1.0934109938114308e-06, + "loss": 0.6184, + "step": 34966 + }, + { + "epoch": 6.36, + "learning_rate": 1.0933563887877686e-06, + "loss": 1.3597, + "step": 34967 + }, + { + "epoch": 6.36, + "learning_rate": 1.0933017837641063e-06, + "loss": 1.1723, + "step": 34968 + }, + { + "epoch": 6.36, + "learning_rate": 1.0932471787404442e-06, + "loss": 0.2063, + "step": 34969 + }, + { + "epoch": 6.36, + "learning_rate": 1.093192573716782e-06, + "loss": 0.3882, + "step": 34970 + }, + { + "epoch": 6.36, + "learning_rate": 1.0931379686931197e-06, + "loss": 0.2744, + "step": 34971 + }, + { + "epoch": 6.36, + "learning_rate": 1.0930833636694576e-06, + "loss": 0.1752, + "step": 34972 + }, + { + "epoch": 6.36, + "learning_rate": 1.0930287586457956e-06, + "loss": 1.2013, + "step": 34973 + }, + { + "epoch": 6.36, + "learning_rate": 1.0929741536221333e-06, + "loss": 0.085, + "step": 34974 + }, + { + "epoch": 6.36, + "learning_rate": 1.092919548598471e-06, + "loss": 1.0883, + "step": 34975 + }, + { + "epoch": 6.36, + "learning_rate": 1.092864943574809e-06, + "loss": 0.5541, + "step": 34976 + }, + { + "epoch": 6.36, + "learning_rate": 1.0928103385511467e-06, + "loss": 0.0839, + "step": 34977 + }, + { + "epoch": 6.36, + "learning_rate": 1.0927557335274845e-06, + "loss": 0.9995, + "step": 34978 + }, + { + "epoch": 6.36, + "learning_rate": 1.0927011285038224e-06, + "loss": 0.1404, + "step": 34979 + }, + { + "epoch": 6.36, + "learning_rate": 1.0926465234801603e-06, + "loss": 0.3237, + "step": 34980 + }, + { + "epoch": 6.36, + "learning_rate": 1.092591918456498e-06, + "loss": 0.4153, + "step": 34981 + }, + { + "epoch": 6.36, + "learning_rate": 1.0925373134328358e-06, + "loss": 0.0507, + "step": 34982 + }, + { + "epoch": 6.36, + "learning_rate": 1.0924827084091737e-06, + "loss": 0.0871, + "step": 34983 + }, + { + "epoch": 6.36, + "learning_rate": 1.0924281033855115e-06, + "loss": 1.0499, + "step": 34984 + }, + { + "epoch": 6.36, + "learning_rate": 1.0923734983618492e-06, + "loss": 1.9038, + "step": 34985 + }, + { + "epoch": 6.36, + "learning_rate": 1.0923188933381872e-06, + "loss": 0.0067, + "step": 34986 + }, + { + "epoch": 6.36, + "learning_rate": 1.092264288314525e-06, + "loss": 0.3779, + "step": 34987 + }, + { + "epoch": 6.36, + "learning_rate": 1.0922096832908628e-06, + "loss": 0.1744, + "step": 34988 + }, + { + "epoch": 6.36, + "learning_rate": 1.0921550782672006e-06, + "loss": 0.7423, + "step": 34989 + }, + { + "epoch": 6.36, + "learning_rate": 1.0921004732435385e-06, + "loss": 0.4752, + "step": 34990 + }, + { + "epoch": 6.36, + "learning_rate": 1.0920458682198762e-06, + "loss": 0.6185, + "step": 34991 + }, + { + "epoch": 6.36, + "learning_rate": 1.091991263196214e-06, + "loss": 0.3289, + "step": 34992 + }, + { + "epoch": 6.36, + "learning_rate": 1.091936658172552e-06, + "loss": 3.4069, + "step": 34993 + }, + { + "epoch": 6.36, + "learning_rate": 1.0918820531488899e-06, + "loss": 0.3442, + "step": 34994 + }, + { + "epoch": 6.36, + "learning_rate": 1.0918274481252276e-06, + "loss": 0.0082, + "step": 34995 + }, + { + "epoch": 6.36, + "learning_rate": 1.0917728431015653e-06, + "loss": 1.2184, + "step": 34996 + }, + { + "epoch": 6.36, + "learning_rate": 1.0917182380779033e-06, + "loss": 0.7306, + "step": 34997 + }, + { + "epoch": 6.36, + "learning_rate": 1.091663633054241e-06, + "loss": 1.1604, + "step": 34998 + }, + { + "epoch": 6.36, + "learning_rate": 1.0916090280305787e-06, + "loss": 0.0465, + "step": 34999 + }, + { + "epoch": 6.36, + "learning_rate": 1.0915544230069167e-06, + "loss": 0.005, + "step": 35000 + }, + { + "epoch": 6.36, + "learning_rate": 1.0914998179832546e-06, + "loss": 0.9356, + "step": 35001 + }, + { + "epoch": 6.37, + "learning_rate": 1.0914452129595923e-06, + "loss": 0.3291, + "step": 35002 + }, + { + "epoch": 6.37, + "learning_rate": 1.09139060793593e-06, + "loss": 0.93, + "step": 35003 + }, + { + "epoch": 6.37, + "learning_rate": 1.091336002912268e-06, + "loss": 0.3702, + "step": 35004 + }, + { + "epoch": 6.37, + "learning_rate": 1.0912813978886057e-06, + "loss": 0.4048, + "step": 35005 + }, + { + "epoch": 6.37, + "learning_rate": 1.0912267928649437e-06, + "loss": 0.5628, + "step": 35006 + }, + { + "epoch": 6.37, + "learning_rate": 1.0911721878412814e-06, + "loss": 0.0048, + "step": 35007 + }, + { + "epoch": 6.37, + "learning_rate": 1.0911175828176194e-06, + "loss": 0.1103, + "step": 35008 + }, + { + "epoch": 6.37, + "learning_rate": 1.091062977793957e-06, + "loss": 1.151, + "step": 35009 + }, + { + "epoch": 6.37, + "learning_rate": 1.0910083727702948e-06, + "loss": 0.4954, + "step": 35010 + }, + { + "epoch": 6.37, + "learning_rate": 1.0909537677466326e-06, + "loss": 0.5609, + "step": 35011 + }, + { + "epoch": 6.37, + "learning_rate": 1.0908991627229705e-06, + "loss": 0.0083, + "step": 35012 + }, + { + "epoch": 6.37, + "learning_rate": 1.0908445576993084e-06, + "loss": 2.1501, + "step": 35013 + }, + { + "epoch": 6.37, + "learning_rate": 1.0907899526756462e-06, + "loss": 0.0618, + "step": 35014 + }, + { + "epoch": 6.37, + "learning_rate": 1.0907353476519841e-06, + "loss": 0.0647, + "step": 35015 + }, + { + "epoch": 6.37, + "learning_rate": 1.0906807426283219e-06, + "loss": 0.0105, + "step": 35016 + }, + { + "epoch": 6.37, + "learning_rate": 1.0906261376046596e-06, + "loss": 0.8692, + "step": 35017 + }, + { + "epoch": 6.37, + "learning_rate": 1.0905715325809973e-06, + "loss": 0.0119, + "step": 35018 + }, + { + "epoch": 6.37, + "learning_rate": 1.0905169275573353e-06, + "loss": 0.0086, + "step": 35019 + }, + { + "epoch": 6.37, + "learning_rate": 1.0904623225336732e-06, + "loss": 0.0219, + "step": 35020 + }, + { + "epoch": 6.37, + "learning_rate": 1.090407717510011e-06, + "loss": 0.3543, + "step": 35021 + }, + { + "epoch": 6.37, + "learning_rate": 1.0903531124863489e-06, + "loss": 0.9838, + "step": 35022 + }, + { + "epoch": 6.37, + "learning_rate": 1.0902985074626866e-06, + "loss": 1.1735, + "step": 35023 + }, + { + "epoch": 6.37, + "learning_rate": 1.0902439024390243e-06, + "loss": 0.1748, + "step": 35024 + }, + { + "epoch": 6.37, + "learning_rate": 1.090189297415362e-06, + "loss": 0.1471, + "step": 35025 + }, + { + "epoch": 6.37, + "learning_rate": 1.0901346923917e-06, + "loss": 1.7231, + "step": 35026 + }, + { + "epoch": 6.37, + "learning_rate": 1.090080087368038e-06, + "loss": 0.0085, + "step": 35027 + }, + { + "epoch": 6.37, + "learning_rate": 1.0900254823443757e-06, + "loss": 0.5726, + "step": 35028 + }, + { + "epoch": 6.37, + "learning_rate": 1.0899708773207136e-06, + "loss": 0.5736, + "step": 35029 + }, + { + "epoch": 6.37, + "learning_rate": 1.0899162722970514e-06, + "loss": 0.0086, + "step": 35030 + }, + { + "epoch": 6.37, + "learning_rate": 1.089861667273389e-06, + "loss": 0.507, + "step": 35031 + }, + { + "epoch": 6.37, + "learning_rate": 1.0898070622497268e-06, + "loss": 0.021, + "step": 35032 + }, + { + "epoch": 6.37, + "learning_rate": 1.0897524572260648e-06, + "loss": 0.2802, + "step": 35033 + }, + { + "epoch": 6.37, + "learning_rate": 1.0896978522024027e-06, + "loss": 0.5996, + "step": 35034 + }, + { + "epoch": 6.37, + "learning_rate": 1.0896432471787405e-06, + "loss": 0.4621, + "step": 35035 + }, + { + "epoch": 6.37, + "learning_rate": 1.0895886421550784e-06, + "loss": 0.0897, + "step": 35036 + }, + { + "epoch": 6.37, + "learning_rate": 1.0895340371314161e-06, + "loss": 0.0503, + "step": 35037 + }, + { + "epoch": 6.37, + "learning_rate": 1.0894794321077539e-06, + "loss": 1.5037, + "step": 35038 + }, + { + "epoch": 6.37, + "learning_rate": 1.0894248270840916e-06, + "loss": 0.0138, + "step": 35039 + }, + { + "epoch": 6.37, + "learning_rate": 1.0893702220604295e-06, + "loss": 0.008, + "step": 35040 + }, + { + "epoch": 6.37, + "learning_rate": 1.0893156170367675e-06, + "loss": 0.0046, + "step": 35041 + }, + { + "epoch": 6.37, + "learning_rate": 1.0892610120131052e-06, + "loss": 0.1449, + "step": 35042 + }, + { + "epoch": 6.37, + "learning_rate": 1.0892064069894432e-06, + "loss": 0.5284, + "step": 35043 + }, + { + "epoch": 6.37, + "learning_rate": 1.0891518019657809e-06, + "loss": 0.067, + "step": 35044 + }, + { + "epoch": 6.37, + "learning_rate": 1.0890971969421186e-06, + "loss": 1.8846, + "step": 35045 + }, + { + "epoch": 6.37, + "learning_rate": 1.0890425919184566e-06, + "loss": 0.0937, + "step": 35046 + }, + { + "epoch": 6.37, + "learning_rate": 1.0889879868947945e-06, + "loss": 0.3539, + "step": 35047 + }, + { + "epoch": 6.37, + "learning_rate": 1.0889333818711322e-06, + "loss": 0.5514, + "step": 35048 + }, + { + "epoch": 6.37, + "learning_rate": 1.08887877684747e-06, + "loss": 0.3771, + "step": 35049 + }, + { + "epoch": 6.37, + "learning_rate": 1.088824171823808e-06, + "loss": 0.2327, + "step": 35050 + }, + { + "epoch": 6.37, + "learning_rate": 1.0887695668001456e-06, + "loss": 0.4176, + "step": 35051 + }, + { + "epoch": 6.37, + "learning_rate": 1.0887149617764834e-06, + "loss": 0.1571, + "step": 35052 + }, + { + "epoch": 6.37, + "learning_rate": 1.0886603567528213e-06, + "loss": 0.484, + "step": 35053 + }, + { + "epoch": 6.37, + "learning_rate": 1.0886057517291593e-06, + "loss": 0.0069, + "step": 35054 + }, + { + "epoch": 6.37, + "learning_rate": 1.088551146705497e-06, + "loss": 0.1493, + "step": 35055 + }, + { + "epoch": 6.37, + "learning_rate": 1.0884965416818347e-06, + "loss": 0.0781, + "step": 35056 + }, + { + "epoch": 6.38, + "learning_rate": 1.0884419366581727e-06, + "loss": 0.9959, + "step": 35057 + }, + { + "epoch": 6.38, + "learning_rate": 1.0883873316345104e-06, + "loss": 0.149, + "step": 35058 + }, + { + "epoch": 6.38, + "learning_rate": 1.0883327266108481e-06, + "loss": 0.2589, + "step": 35059 + }, + { + "epoch": 6.38, + "learning_rate": 1.088278121587186e-06, + "loss": 0.3646, + "step": 35060 + }, + { + "epoch": 6.38, + "learning_rate": 1.088223516563524e-06, + "loss": 0.1226, + "step": 35061 + }, + { + "epoch": 6.38, + "learning_rate": 1.0881689115398617e-06, + "loss": 0.069, + "step": 35062 + }, + { + "epoch": 6.38, + "learning_rate": 1.0881143065161995e-06, + "loss": 1.2522, + "step": 35063 + }, + { + "epoch": 6.38, + "learning_rate": 1.0880597014925374e-06, + "loss": 1.5811, + "step": 35064 + }, + { + "epoch": 6.38, + "learning_rate": 1.0880050964688752e-06, + "loss": 1.2078, + "step": 35065 + }, + { + "epoch": 6.38, + "learning_rate": 1.0879504914452129e-06, + "loss": 0.7141, + "step": 35066 + }, + { + "epoch": 6.38, + "learning_rate": 1.0878958864215508e-06, + "loss": 1.0868, + "step": 35067 + }, + { + "epoch": 6.38, + "learning_rate": 1.0878412813978888e-06, + "loss": 0.0211, + "step": 35068 + }, + { + "epoch": 6.38, + "learning_rate": 1.0877866763742265e-06, + "loss": 0.1906, + "step": 35069 + }, + { + "epoch": 6.38, + "learning_rate": 1.0877320713505642e-06, + "loss": 0.0215, + "step": 35070 + }, + { + "epoch": 6.38, + "learning_rate": 1.0876774663269022e-06, + "loss": 0.0994, + "step": 35071 + }, + { + "epoch": 6.38, + "learning_rate": 1.08762286130324e-06, + "loss": 0.0076, + "step": 35072 + }, + { + "epoch": 6.38, + "learning_rate": 1.0875682562795776e-06, + "loss": 0.0521, + "step": 35073 + }, + { + "epoch": 6.38, + "learning_rate": 1.0875136512559156e-06, + "loss": 0.0111, + "step": 35074 + }, + { + "epoch": 6.38, + "learning_rate": 1.0874590462322535e-06, + "loss": 0.011, + "step": 35075 + }, + { + "epoch": 6.38, + "learning_rate": 1.0874044412085913e-06, + "loss": 0.4441, + "step": 35076 + }, + { + "epoch": 6.38, + "learning_rate": 1.087349836184929e-06, + "loss": 0.157, + "step": 35077 + }, + { + "epoch": 6.38, + "learning_rate": 1.087295231161267e-06, + "loss": 0.41, + "step": 35078 + }, + { + "epoch": 6.38, + "learning_rate": 1.0872406261376047e-06, + "loss": 0.05, + "step": 35079 + }, + { + "epoch": 6.38, + "learning_rate": 1.0871860211139424e-06, + "loss": 0.2962, + "step": 35080 + }, + { + "epoch": 6.38, + "learning_rate": 1.0871314160902803e-06, + "loss": 1.1779, + "step": 35081 + }, + { + "epoch": 6.38, + "learning_rate": 1.0870768110666183e-06, + "loss": 0.6173, + "step": 35082 + }, + { + "epoch": 6.38, + "learning_rate": 1.087022206042956e-06, + "loss": 0.5537, + "step": 35083 + }, + { + "epoch": 6.38, + "learning_rate": 1.0869676010192937e-06, + "loss": 0.0681, + "step": 35084 + }, + { + "epoch": 6.38, + "learning_rate": 1.0869129959956317e-06, + "loss": 0.8406, + "step": 35085 + }, + { + "epoch": 6.38, + "learning_rate": 1.0868583909719694e-06, + "loss": 0.1888, + "step": 35086 + }, + { + "epoch": 6.38, + "learning_rate": 1.0868037859483074e-06, + "loss": 0.0113, + "step": 35087 + }, + { + "epoch": 6.38, + "learning_rate": 1.086749180924645e-06, + "loss": 0.0674, + "step": 35088 + }, + { + "epoch": 6.38, + "learning_rate": 1.086694575900983e-06, + "loss": 0.1989, + "step": 35089 + }, + { + "epoch": 6.38, + "learning_rate": 1.0866399708773208e-06, + "loss": 0.0092, + "step": 35090 + }, + { + "epoch": 6.38, + "learning_rate": 1.0865853658536585e-06, + "loss": 0.9585, + "step": 35091 + }, + { + "epoch": 6.38, + "learning_rate": 1.0865307608299962e-06, + "loss": 0.4281, + "step": 35092 + }, + { + "epoch": 6.38, + "learning_rate": 1.0864761558063342e-06, + "loss": 0.43, + "step": 35093 + }, + { + "epoch": 6.38, + "learning_rate": 1.0864215507826721e-06, + "loss": 0.8831, + "step": 35094 + }, + { + "epoch": 6.38, + "learning_rate": 1.0863669457590099e-06, + "loss": 0.2067, + "step": 35095 + }, + { + "epoch": 6.38, + "learning_rate": 1.0863123407353478e-06, + "loss": 0.2302, + "step": 35096 + }, + { + "epoch": 6.38, + "learning_rate": 1.0862577357116855e-06, + "loss": 0.8216, + "step": 35097 + }, + { + "epoch": 6.38, + "learning_rate": 1.0862031306880233e-06, + "loss": 1.5679, + "step": 35098 + }, + { + "epoch": 6.38, + "learning_rate": 1.086148525664361e-06, + "loss": 2.6312, + "step": 35099 + }, + { + "epoch": 6.38, + "learning_rate": 1.086093920640699e-06, + "loss": 0.2105, + "step": 35100 + }, + { + "epoch": 6.38, + "learning_rate": 1.0860393156170369e-06, + "loss": 0.2192, + "step": 35101 + }, + { + "epoch": 6.38, + "learning_rate": 1.0859847105933746e-06, + "loss": 1.8354, + "step": 35102 + }, + { + "epoch": 6.38, + "learning_rate": 1.0859301055697126e-06, + "loss": 0.3471, + "step": 35103 + }, + { + "epoch": 6.38, + "learning_rate": 1.0858755005460503e-06, + "loss": 0.057, + "step": 35104 + }, + { + "epoch": 6.38, + "learning_rate": 1.085820895522388e-06, + "loss": 0.4912, + "step": 35105 + }, + { + "epoch": 6.38, + "learning_rate": 1.0857662904987257e-06, + "loss": 0.0978, + "step": 35106 + }, + { + "epoch": 6.38, + "learning_rate": 1.0857116854750637e-06, + "loss": 1.1302, + "step": 35107 + }, + { + "epoch": 6.38, + "learning_rate": 1.0856570804514016e-06, + "loss": 0.0737, + "step": 35108 + }, + { + "epoch": 6.38, + "learning_rate": 1.0856024754277394e-06, + "loss": 0.9797, + "step": 35109 + }, + { + "epoch": 6.38, + "learning_rate": 1.0855478704040773e-06, + "loss": 0.1513, + "step": 35110 + }, + { + "epoch": 6.38, + "learning_rate": 1.085493265380415e-06, + "loss": 0.3476, + "step": 35111 + }, + { + "epoch": 6.39, + "learning_rate": 1.0854386603567528e-06, + "loss": 4.3873, + "step": 35112 + }, + { + "epoch": 6.39, + "learning_rate": 1.0853840553330905e-06, + "loss": 0.4864, + "step": 35113 + }, + { + "epoch": 6.39, + "learning_rate": 1.0853294503094284e-06, + "loss": 0.9154, + "step": 35114 + }, + { + "epoch": 6.39, + "learning_rate": 1.0852748452857664e-06, + "loss": 0.9534, + "step": 35115 + }, + { + "epoch": 6.39, + "learning_rate": 1.0852202402621041e-06, + "loss": 2.1861, + "step": 35116 + }, + { + "epoch": 6.39, + "learning_rate": 1.085165635238442e-06, + "loss": 0.7133, + "step": 35117 + }, + { + "epoch": 6.39, + "learning_rate": 1.0851110302147798e-06, + "loss": 0.2727, + "step": 35118 + }, + { + "epoch": 6.39, + "learning_rate": 1.0850564251911175e-06, + "loss": 0.0473, + "step": 35119 + }, + { + "epoch": 6.39, + "learning_rate": 1.0850018201674553e-06, + "loss": 0.564, + "step": 35120 + }, + { + "epoch": 6.39, + "learning_rate": 1.0849472151437932e-06, + "loss": 1.663, + "step": 35121 + }, + { + "epoch": 6.39, + "learning_rate": 1.0848926101201311e-06, + "loss": 1.0908, + "step": 35122 + }, + { + "epoch": 6.39, + "learning_rate": 1.0848380050964689e-06, + "loss": 0.1852, + "step": 35123 + }, + { + "epoch": 6.39, + "learning_rate": 1.0847834000728068e-06, + "loss": 2.0302, + "step": 35124 + }, + { + "epoch": 6.39, + "learning_rate": 1.0847287950491446e-06, + "loss": 0.0372, + "step": 35125 + }, + { + "epoch": 6.39, + "learning_rate": 1.0846741900254823e-06, + "loss": 0.5721, + "step": 35126 + }, + { + "epoch": 6.39, + "learning_rate": 1.08461958500182e-06, + "loss": 0.0037, + "step": 35127 + }, + { + "epoch": 6.39, + "learning_rate": 1.084564979978158e-06, + "loss": 0.5851, + "step": 35128 + }, + { + "epoch": 6.39, + "learning_rate": 1.084510374954496e-06, + "loss": 0.0883, + "step": 35129 + }, + { + "epoch": 6.39, + "learning_rate": 1.0844557699308336e-06, + "loss": 0.1346, + "step": 35130 + }, + { + "epoch": 6.39, + "learning_rate": 1.0844011649071716e-06, + "loss": 1.0309, + "step": 35131 + }, + { + "epoch": 6.39, + "learning_rate": 1.0843465598835093e-06, + "loss": 1.7322, + "step": 35132 + }, + { + "epoch": 6.39, + "learning_rate": 1.084291954859847e-06, + "loss": 0.0586, + "step": 35133 + }, + { + "epoch": 6.39, + "learning_rate": 1.084237349836185e-06, + "loss": 2.4153, + "step": 35134 + }, + { + "epoch": 6.39, + "learning_rate": 1.084182744812523e-06, + "loss": 0.615, + "step": 35135 + }, + { + "epoch": 6.39, + "learning_rate": 1.0841281397888607e-06, + "loss": 1.1879, + "step": 35136 + }, + { + "epoch": 6.39, + "learning_rate": 1.0840735347651984e-06, + "loss": 0.1176, + "step": 35137 + }, + { + "epoch": 6.39, + "learning_rate": 1.0840189297415363e-06, + "loss": 0.0096, + "step": 35138 + }, + { + "epoch": 6.39, + "learning_rate": 1.083964324717874e-06, + "loss": 0.1631, + "step": 35139 + }, + { + "epoch": 6.39, + "learning_rate": 1.0839097196942118e-06, + "loss": 1.1755, + "step": 35140 + }, + { + "epoch": 6.39, + "learning_rate": 1.0838551146705497e-06, + "loss": 0.3358, + "step": 35141 + }, + { + "epoch": 6.39, + "learning_rate": 1.0838005096468877e-06, + "loss": 0.0051, + "step": 35142 + }, + { + "epoch": 6.39, + "learning_rate": 1.0837459046232254e-06, + "loss": 0.3609, + "step": 35143 + }, + { + "epoch": 6.39, + "learning_rate": 1.0836912995995632e-06, + "loss": 0.1175, + "step": 35144 + }, + { + "epoch": 6.39, + "learning_rate": 1.083636694575901e-06, + "loss": 0.0184, + "step": 35145 + }, + { + "epoch": 6.39, + "learning_rate": 1.0835820895522388e-06, + "loss": 0.0098, + "step": 35146 + }, + { + "epoch": 6.39, + "learning_rate": 1.0835274845285766e-06, + "loss": 0.6206, + "step": 35147 + }, + { + "epoch": 6.39, + "learning_rate": 1.0834728795049145e-06, + "loss": 0.1949, + "step": 35148 + }, + { + "epoch": 6.39, + "learning_rate": 1.0834182744812524e-06, + "loss": 1.1069, + "step": 35149 + }, + { + "epoch": 6.39, + "learning_rate": 1.0833636694575902e-06, + "loss": 0.3147, + "step": 35150 + }, + { + "epoch": 6.39, + "learning_rate": 1.083309064433928e-06, + "loss": 1.5549, + "step": 35151 + }, + { + "epoch": 6.39, + "learning_rate": 1.0832544594102659e-06, + "loss": 2.5865, + "step": 35152 + }, + { + "epoch": 6.39, + "learning_rate": 1.0831998543866036e-06, + "loss": 1.6487, + "step": 35153 + }, + { + "epoch": 6.39, + "learning_rate": 1.0831452493629413e-06, + "loss": 0.1015, + "step": 35154 + }, + { + "epoch": 6.39, + "learning_rate": 1.0830906443392793e-06, + "loss": 1.7714, + "step": 35155 + }, + { + "epoch": 6.39, + "learning_rate": 1.0830360393156172e-06, + "loss": 0.5959, + "step": 35156 + }, + { + "epoch": 6.39, + "learning_rate": 1.082981434291955e-06, + "loss": 0.0582, + "step": 35157 + }, + { + "epoch": 6.39, + "learning_rate": 1.0829268292682927e-06, + "loss": 0.4892, + "step": 35158 + }, + { + "epoch": 6.39, + "learning_rate": 1.0828722242446306e-06, + "loss": 0.148, + "step": 35159 + }, + { + "epoch": 6.39, + "learning_rate": 1.0828176192209683e-06, + "loss": 2.1924, + "step": 35160 + }, + { + "epoch": 6.39, + "learning_rate": 1.082763014197306e-06, + "loss": 0.8261, + "step": 35161 + }, + { + "epoch": 6.39, + "learning_rate": 1.082708409173644e-06, + "loss": 0.501, + "step": 35162 + }, + { + "epoch": 6.39, + "learning_rate": 1.082653804149982e-06, + "loss": 1.6334, + "step": 35163 + }, + { + "epoch": 6.39, + "learning_rate": 1.0825991991263197e-06, + "loss": 0.3864, + "step": 35164 + }, + { + "epoch": 6.39, + "learning_rate": 1.0825445941026574e-06, + "loss": 0.006, + "step": 35165 + }, + { + "epoch": 6.39, + "learning_rate": 1.0824899890789952e-06, + "loss": 0.0507, + "step": 35166 + }, + { + "epoch": 6.4, + "learning_rate": 1.082435384055333e-06, + "loss": 0.0464, + "step": 35167 + }, + { + "epoch": 6.4, + "learning_rate": 1.0823807790316708e-06, + "loss": 1.3288, + "step": 35168 + }, + { + "epoch": 6.4, + "learning_rate": 1.0823261740080088e-06, + "loss": 0.1358, + "step": 35169 + }, + { + "epoch": 6.4, + "learning_rate": 1.0822715689843467e-06, + "loss": 0.5566, + "step": 35170 + }, + { + "epoch": 6.4, + "learning_rate": 1.0822169639606844e-06, + "loss": 0.0608, + "step": 35171 + }, + { + "epoch": 6.4, + "learning_rate": 1.0821623589370222e-06, + "loss": 1.5484, + "step": 35172 + }, + { + "epoch": 6.4, + "learning_rate": 1.08210775391336e-06, + "loss": 0.2785, + "step": 35173 + }, + { + "epoch": 6.4, + "learning_rate": 1.0820531488896979e-06, + "loss": 0.5723, + "step": 35174 + }, + { + "epoch": 6.4, + "learning_rate": 1.0819985438660358e-06, + "loss": 0.0044, + "step": 35175 + }, + { + "epoch": 6.4, + "learning_rate": 1.0819439388423735e-06, + "loss": 0.5473, + "step": 35176 + }, + { + "epoch": 6.4, + "learning_rate": 1.0818893338187115e-06, + "loss": 0.417, + "step": 35177 + }, + { + "epoch": 6.4, + "learning_rate": 1.0818347287950492e-06, + "loss": 0.0687, + "step": 35178 + }, + { + "epoch": 6.4, + "learning_rate": 1.081780123771387e-06, + "loss": 0.0232, + "step": 35179 + }, + { + "epoch": 6.4, + "learning_rate": 1.0817255187477247e-06, + "loss": 0.1696, + "step": 35180 + }, + { + "epoch": 6.4, + "learning_rate": 1.0816709137240626e-06, + "loss": 0.0094, + "step": 35181 + }, + { + "epoch": 6.4, + "learning_rate": 1.0816163087004006e-06, + "loss": 0.0335, + "step": 35182 + }, + { + "epoch": 6.4, + "learning_rate": 1.0815617036767383e-06, + "loss": 0.0405, + "step": 35183 + }, + { + "epoch": 6.4, + "learning_rate": 1.0815070986530762e-06, + "loss": 1.2064, + "step": 35184 + }, + { + "epoch": 6.4, + "learning_rate": 1.081452493629414e-06, + "loss": 0.0081, + "step": 35185 + }, + { + "epoch": 6.4, + "learning_rate": 1.0813978886057517e-06, + "loss": 0.4499, + "step": 35186 + }, + { + "epoch": 6.4, + "learning_rate": 1.0813432835820894e-06, + "loss": 0.8223, + "step": 35187 + }, + { + "epoch": 6.4, + "learning_rate": 1.0812886785584274e-06, + "loss": 0.1304, + "step": 35188 + }, + { + "epoch": 6.4, + "learning_rate": 1.0812340735347653e-06, + "loss": 0.981, + "step": 35189 + }, + { + "epoch": 6.4, + "learning_rate": 1.081179468511103e-06, + "loss": 0.011, + "step": 35190 + }, + { + "epoch": 6.4, + "learning_rate": 1.081124863487441e-06, + "loss": 0.4479, + "step": 35191 + }, + { + "epoch": 6.4, + "learning_rate": 1.0810702584637787e-06, + "loss": 0.0452, + "step": 35192 + }, + { + "epoch": 6.4, + "learning_rate": 1.0810156534401164e-06, + "loss": 0.0086, + "step": 35193 + }, + { + "epoch": 6.4, + "learning_rate": 1.0809610484164542e-06, + "loss": 1.3048, + "step": 35194 + }, + { + "epoch": 6.4, + "learning_rate": 1.0809064433927921e-06, + "loss": 0.1951, + "step": 35195 + }, + { + "epoch": 6.4, + "learning_rate": 1.08085183836913e-06, + "loss": 2.5934, + "step": 35196 + }, + { + "epoch": 6.4, + "learning_rate": 1.0807972333454678e-06, + "loss": 0.4554, + "step": 35197 + }, + { + "epoch": 6.4, + "learning_rate": 1.0807426283218057e-06, + "loss": 1.7679, + "step": 35198 + }, + { + "epoch": 6.4, + "learning_rate": 1.0806880232981435e-06, + "loss": 0.482, + "step": 35199 + }, + { + "epoch": 6.4, + "learning_rate": 1.0806334182744812e-06, + "loss": 0.0738, + "step": 35200 + }, + { + "epoch": 6.4, + "learning_rate": 1.080578813250819e-06, + "loss": 0.1751, + "step": 35201 + }, + { + "epoch": 6.4, + "learning_rate": 1.0805242082271569e-06, + "loss": 1.7129, + "step": 35202 + }, + { + "epoch": 6.4, + "learning_rate": 1.0804696032034948e-06, + "loss": 0.1533, + "step": 35203 + }, + { + "epoch": 6.4, + "learning_rate": 1.0804149981798326e-06, + "loss": 0.7658, + "step": 35204 + }, + { + "epoch": 6.4, + "learning_rate": 1.0803603931561705e-06, + "loss": 0.2253, + "step": 35205 + }, + { + "epoch": 6.4, + "learning_rate": 1.0803057881325082e-06, + "loss": 1.8737, + "step": 35206 + }, + { + "epoch": 6.4, + "learning_rate": 1.080251183108846e-06, + "loss": 0.0449, + "step": 35207 + }, + { + "epoch": 6.4, + "learning_rate": 1.0801965780851837e-06, + "loss": 0.2363, + "step": 35208 + }, + { + "epoch": 6.4, + "learning_rate": 1.0801419730615216e-06, + "loss": 0.1376, + "step": 35209 + }, + { + "epoch": 6.4, + "learning_rate": 1.0800873680378596e-06, + "loss": 0.1732, + "step": 35210 + }, + { + "epoch": 6.4, + "learning_rate": 1.0800327630141973e-06, + "loss": 0.0183, + "step": 35211 + }, + { + "epoch": 6.4, + "learning_rate": 1.0799781579905353e-06, + "loss": 1.4634, + "step": 35212 + }, + { + "epoch": 6.4, + "learning_rate": 1.079923552966873e-06, + "loss": 1.902, + "step": 35213 + }, + { + "epoch": 6.4, + "learning_rate": 1.0798689479432107e-06, + "loss": 0.24, + "step": 35214 + }, + { + "epoch": 6.4, + "learning_rate": 1.0798143429195487e-06, + "loss": 1.8895, + "step": 35215 + }, + { + "epoch": 6.4, + "learning_rate": 1.0797597378958866e-06, + "loss": 0.0989, + "step": 35216 + }, + { + "epoch": 6.4, + "learning_rate": 1.0797051328722243e-06, + "loss": 0.0735, + "step": 35217 + }, + { + "epoch": 6.4, + "learning_rate": 1.079650527848562e-06, + "loss": 0.58, + "step": 35218 + }, + { + "epoch": 6.4, + "learning_rate": 1.0795959228249e-06, + "loss": 0.0038, + "step": 35219 + }, + { + "epoch": 6.4, + "learning_rate": 1.0795413178012377e-06, + "loss": 0.039, + "step": 35220 + }, + { + "epoch": 6.4, + "learning_rate": 1.0794867127775755e-06, + "loss": 0.2131, + "step": 35221 + }, + { + "epoch": 6.41, + "learning_rate": 1.0794321077539134e-06, + "loss": 0.0113, + "step": 35222 + }, + { + "epoch": 6.41, + "learning_rate": 1.0793775027302514e-06, + "loss": 0.2515, + "step": 35223 + }, + { + "epoch": 6.41, + "learning_rate": 1.079322897706589e-06, + "loss": 0.0081, + "step": 35224 + }, + { + "epoch": 6.41, + "learning_rate": 1.0792682926829268e-06, + "loss": 0.2547, + "step": 35225 + }, + { + "epoch": 6.41, + "learning_rate": 1.0792136876592648e-06, + "loss": 0.2761, + "step": 35226 + }, + { + "epoch": 6.41, + "learning_rate": 1.0791590826356025e-06, + "loss": 0.0041, + "step": 35227 + }, + { + "epoch": 6.41, + "learning_rate": 1.0791044776119402e-06, + "loss": 1.1945, + "step": 35228 + }, + { + "epoch": 6.41, + "learning_rate": 1.0790498725882782e-06, + "loss": 0.4986, + "step": 35229 + }, + { + "epoch": 6.41, + "learning_rate": 1.0789952675646161e-06, + "loss": 1.1957, + "step": 35230 + }, + { + "epoch": 6.41, + "learning_rate": 1.0789406625409539e-06, + "loss": 0.3245, + "step": 35231 + }, + { + "epoch": 6.41, + "learning_rate": 1.0788860575172916e-06, + "loss": 0.565, + "step": 35232 + }, + { + "epoch": 6.41, + "learning_rate": 1.0788314524936295e-06, + "loss": 0.1943, + "step": 35233 + }, + { + "epoch": 6.41, + "learning_rate": 1.0787768474699673e-06, + "loss": 0.0699, + "step": 35234 + }, + { + "epoch": 6.41, + "learning_rate": 1.078722242446305e-06, + "loss": 0.6466, + "step": 35235 + }, + { + "epoch": 6.41, + "learning_rate": 1.078667637422643e-06, + "loss": 1.3678, + "step": 35236 + }, + { + "epoch": 6.41, + "learning_rate": 1.0786130323989809e-06, + "loss": 1.1044, + "step": 35237 + }, + { + "epoch": 6.41, + "learning_rate": 1.0785584273753186e-06, + "loss": 0.3247, + "step": 35238 + }, + { + "epoch": 6.41, + "learning_rate": 1.0785038223516563e-06, + "loss": 0.1751, + "step": 35239 + }, + { + "epoch": 6.41, + "learning_rate": 1.0784492173279943e-06, + "loss": 0.0676, + "step": 35240 + }, + { + "epoch": 6.41, + "learning_rate": 1.078394612304332e-06, + "loss": 1.4234, + "step": 35241 + }, + { + "epoch": 6.41, + "learning_rate": 1.0783400072806697e-06, + "loss": 0.4359, + "step": 35242 + }, + { + "epoch": 6.41, + "learning_rate": 1.0782854022570077e-06, + "loss": 0.0042, + "step": 35243 + }, + { + "epoch": 6.41, + "learning_rate": 1.0782307972333456e-06, + "loss": 0.8922, + "step": 35244 + }, + { + "epoch": 6.41, + "learning_rate": 1.0781761922096834e-06, + "loss": 0.7085, + "step": 35245 + }, + { + "epoch": 6.41, + "learning_rate": 1.078121587186021e-06, + "loss": 0.1826, + "step": 35246 + }, + { + "epoch": 6.41, + "learning_rate": 1.0780669821623588e-06, + "loss": 0.247, + "step": 35247 + }, + { + "epoch": 6.41, + "learning_rate": 1.0780123771386968e-06, + "loss": 3.2948, + "step": 35248 + }, + { + "epoch": 6.41, + "learning_rate": 1.0779577721150345e-06, + "loss": 0.0111, + "step": 35249 + }, + { + "epoch": 6.41, + "learning_rate": 1.0779031670913724e-06, + "loss": 0.2249, + "step": 35250 + }, + { + "epoch": 6.41, + "learning_rate": 1.0778485620677104e-06, + "loss": 0.5835, + "step": 35251 + }, + { + "epoch": 6.41, + "learning_rate": 1.0777939570440481e-06, + "loss": 1.2511, + "step": 35252 + }, + { + "epoch": 6.41, + "learning_rate": 1.0777393520203859e-06, + "loss": 0.2719, + "step": 35253 + }, + { + "epoch": 6.41, + "learning_rate": 1.0776847469967236e-06, + "loss": 0.3342, + "step": 35254 + }, + { + "epoch": 6.41, + "learning_rate": 1.0776301419730615e-06, + "loss": 0.1499, + "step": 35255 + }, + { + "epoch": 6.41, + "learning_rate": 1.0775755369493993e-06, + "loss": 0.7724, + "step": 35256 + }, + { + "epoch": 6.41, + "learning_rate": 1.0775209319257372e-06, + "loss": 1.4654, + "step": 35257 + }, + { + "epoch": 6.41, + "learning_rate": 1.0774663269020751e-06, + "loss": 0.2595, + "step": 35258 + }, + { + "epoch": 6.41, + "learning_rate": 1.0774117218784129e-06, + "loss": 0.1481, + "step": 35259 + }, + { + "epoch": 6.41, + "learning_rate": 1.0773571168547506e-06, + "loss": 0.0103, + "step": 35260 + }, + { + "epoch": 6.41, + "learning_rate": 1.0773025118310883e-06, + "loss": 0.1789, + "step": 35261 + }, + { + "epoch": 6.41, + "learning_rate": 1.0772479068074263e-06, + "loss": 1.185, + "step": 35262 + }, + { + "epoch": 6.41, + "learning_rate": 1.0771933017837642e-06, + "loss": 0.1456, + "step": 35263 + }, + { + "epoch": 6.41, + "learning_rate": 1.077138696760102e-06, + "loss": 0.0859, + "step": 35264 + }, + { + "epoch": 6.41, + "learning_rate": 1.07708409173644e-06, + "loss": 0.0275, + "step": 35265 + }, + { + "epoch": 6.41, + "learning_rate": 1.0770294867127776e-06, + "loss": 1.043, + "step": 35266 + }, + { + "epoch": 6.41, + "learning_rate": 1.0769748816891154e-06, + "loss": 0.7577, + "step": 35267 + }, + { + "epoch": 6.41, + "learning_rate": 1.076920276665453e-06, + "loss": 0.357, + "step": 35268 + }, + { + "epoch": 6.41, + "learning_rate": 1.076865671641791e-06, + "loss": 0.7677, + "step": 35269 + }, + { + "epoch": 6.41, + "learning_rate": 1.076811066618129e-06, + "loss": 0.0115, + "step": 35270 + }, + { + "epoch": 6.41, + "learning_rate": 1.0767564615944667e-06, + "loss": 0.4961, + "step": 35271 + }, + { + "epoch": 6.41, + "learning_rate": 1.0767018565708047e-06, + "loss": 0.0811, + "step": 35272 + }, + { + "epoch": 6.41, + "learning_rate": 1.0766472515471424e-06, + "loss": 0.0687, + "step": 35273 + }, + { + "epoch": 6.41, + "learning_rate": 1.0765926465234801e-06, + "loss": 0.1906, + "step": 35274 + }, + { + "epoch": 6.41, + "learning_rate": 1.0765380414998179e-06, + "loss": 0.159, + "step": 35275 + }, + { + "epoch": 6.41, + "learning_rate": 1.0764834364761558e-06, + "loss": 0.0903, + "step": 35276 + }, + { + "epoch": 6.42, + "learning_rate": 1.0764288314524937e-06, + "loss": 0.3943, + "step": 35277 + }, + { + "epoch": 6.42, + "learning_rate": 1.0763742264288315e-06, + "loss": 0.2319, + "step": 35278 + }, + { + "epoch": 6.42, + "learning_rate": 1.0763196214051694e-06, + "loss": 0.0497, + "step": 35279 + }, + { + "epoch": 6.42, + "learning_rate": 1.0762650163815071e-06, + "loss": 0.3834, + "step": 35280 + }, + { + "epoch": 6.42, + "learning_rate": 1.0762104113578449e-06, + "loss": 0.0685, + "step": 35281 + }, + { + "epoch": 6.42, + "learning_rate": 1.0761558063341826e-06, + "loss": 1.1971, + "step": 35282 + }, + { + "epoch": 6.42, + "learning_rate": 1.0761012013105206e-06, + "loss": 0.8304, + "step": 35283 + }, + { + "epoch": 6.42, + "learning_rate": 1.0760465962868585e-06, + "loss": 0.1045, + "step": 35284 + }, + { + "epoch": 6.42, + "learning_rate": 1.0759919912631962e-06, + "loss": 0.8591, + "step": 35285 + }, + { + "epoch": 6.42, + "learning_rate": 1.0759373862395342e-06, + "loss": 0.6215, + "step": 35286 + }, + { + "epoch": 6.42, + "learning_rate": 1.075882781215872e-06, + "loss": 0.012, + "step": 35287 + }, + { + "epoch": 6.42, + "learning_rate": 1.0758281761922096e-06, + "loss": 0.0551, + "step": 35288 + }, + { + "epoch": 6.42, + "learning_rate": 1.0757735711685474e-06, + "loss": 0.636, + "step": 35289 + }, + { + "epoch": 6.42, + "learning_rate": 1.0757189661448853e-06, + "loss": 1.5619, + "step": 35290 + }, + { + "epoch": 6.42, + "learning_rate": 1.0756643611212233e-06, + "loss": 2.1979, + "step": 35291 + }, + { + "epoch": 6.42, + "learning_rate": 1.075609756097561e-06, + "loss": 0.1273, + "step": 35292 + }, + { + "epoch": 6.42, + "learning_rate": 1.075555151073899e-06, + "loss": 0.0825, + "step": 35293 + }, + { + "epoch": 6.42, + "learning_rate": 1.0755005460502367e-06, + "loss": 0.8466, + "step": 35294 + }, + { + "epoch": 6.42, + "learning_rate": 1.0754459410265744e-06, + "loss": 0.2836, + "step": 35295 + }, + { + "epoch": 6.42, + "learning_rate": 1.0753913360029121e-06, + "loss": 1.4748, + "step": 35296 + }, + { + "epoch": 6.42, + "learning_rate": 1.07533673097925e-06, + "loss": 1.528, + "step": 35297 + }, + { + "epoch": 6.42, + "learning_rate": 1.075282125955588e-06, + "loss": 0.3312, + "step": 35298 + }, + { + "epoch": 6.42, + "learning_rate": 1.0752275209319257e-06, + "loss": 1.1926, + "step": 35299 + }, + { + "epoch": 6.42, + "learning_rate": 1.0751729159082637e-06, + "loss": 0.0048, + "step": 35300 + }, + { + "epoch": 6.42, + "learning_rate": 1.0751183108846014e-06, + "loss": 1.1495, + "step": 35301 + }, + { + "epoch": 6.42, + "learning_rate": 1.0750637058609391e-06, + "loss": 0.018, + "step": 35302 + }, + { + "epoch": 6.42, + "learning_rate": 1.075009100837277e-06, + "loss": 0.4166, + "step": 35303 + }, + { + "epoch": 6.42, + "learning_rate": 1.074954495813615e-06, + "loss": 0.8803, + "step": 35304 + }, + { + "epoch": 6.42, + "learning_rate": 1.0748998907899528e-06, + "loss": 0.0947, + "step": 35305 + }, + { + "epoch": 6.42, + "learning_rate": 1.0748452857662905e-06, + "loss": 0.0033, + "step": 35306 + }, + { + "epoch": 6.42, + "learning_rate": 1.0747906807426284e-06, + "loss": 0.0094, + "step": 35307 + }, + { + "epoch": 6.42, + "learning_rate": 1.0747360757189662e-06, + "loss": 0.562, + "step": 35308 + }, + { + "epoch": 6.42, + "learning_rate": 1.074681470695304e-06, + "loss": 0.4509, + "step": 35309 + }, + { + "epoch": 6.42, + "learning_rate": 1.0746268656716418e-06, + "loss": 0.095, + "step": 35310 + }, + { + "epoch": 6.42, + "learning_rate": 1.0745722606479798e-06, + "loss": 0.5153, + "step": 35311 + }, + { + "epoch": 6.42, + "learning_rate": 1.0745176556243175e-06, + "loss": 1.3422, + "step": 35312 + }, + { + "epoch": 6.42, + "learning_rate": 1.0744630506006553e-06, + "loss": 1.0599, + "step": 35313 + }, + { + "epoch": 6.42, + "learning_rate": 1.0744084455769932e-06, + "loss": 1.395, + "step": 35314 + }, + { + "epoch": 6.42, + "learning_rate": 1.074353840553331e-06, + "loss": 0.0648, + "step": 35315 + }, + { + "epoch": 6.42, + "learning_rate": 1.0742992355296687e-06, + "loss": 0.266, + "step": 35316 + }, + { + "epoch": 6.42, + "learning_rate": 1.0742446305060066e-06, + "loss": 0.0127, + "step": 35317 + }, + { + "epoch": 6.42, + "learning_rate": 1.0741900254823446e-06, + "loss": 1.3002, + "step": 35318 + }, + { + "epoch": 6.42, + "learning_rate": 1.0741354204586823e-06, + "loss": 0.8765, + "step": 35319 + }, + { + "epoch": 6.42, + "learning_rate": 1.07408081543502e-06, + "loss": 0.9784, + "step": 35320 + }, + { + "epoch": 6.42, + "learning_rate": 1.0740262104113577e-06, + "loss": 0.1461, + "step": 35321 + }, + { + "epoch": 6.42, + "learning_rate": 1.0739716053876957e-06, + "loss": 0.0986, + "step": 35322 + }, + { + "epoch": 6.42, + "learning_rate": 1.0739170003640334e-06, + "loss": 0.2163, + "step": 35323 + }, + { + "epoch": 6.42, + "learning_rate": 1.0738623953403714e-06, + "loss": 0.0562, + "step": 35324 + }, + { + "epoch": 6.42, + "learning_rate": 1.0738077903167093e-06, + "loss": 0.037, + "step": 35325 + }, + { + "epoch": 6.42, + "learning_rate": 1.073753185293047e-06, + "loss": 1.7399, + "step": 35326 + }, + { + "epoch": 6.42, + "learning_rate": 1.0736985802693848e-06, + "loss": 0.1825, + "step": 35327 + }, + { + "epoch": 6.42, + "learning_rate": 1.0736439752457225e-06, + "loss": 0.7654, + "step": 35328 + }, + { + "epoch": 6.42, + "learning_rate": 1.0735893702220604e-06, + "loss": 0.2953, + "step": 35329 + }, + { + "epoch": 6.42, + "learning_rate": 1.0735347651983982e-06, + "loss": 1.6627, + "step": 35330 + }, + { + "epoch": 6.42, + "learning_rate": 1.0734801601747361e-06, + "loss": 0.4326, + "step": 35331 + }, + { + "epoch": 6.43, + "learning_rate": 1.073425555151074e-06, + "loss": 1.3209, + "step": 35332 + }, + { + "epoch": 6.43, + "learning_rate": 1.0733709501274118e-06, + "loss": 0.3564, + "step": 35333 + }, + { + "epoch": 6.43, + "learning_rate": 1.0733163451037495e-06, + "loss": 0.0064, + "step": 35334 + }, + { + "epoch": 6.43, + "learning_rate": 1.0732617400800873e-06, + "loss": 0.26, + "step": 35335 + }, + { + "epoch": 6.43, + "learning_rate": 1.0732071350564252e-06, + "loss": 0.0089, + "step": 35336 + }, + { + "epoch": 6.43, + "learning_rate": 1.073152530032763e-06, + "loss": 0.5249, + "step": 35337 + }, + { + "epoch": 6.43, + "learning_rate": 1.0730979250091009e-06, + "loss": 1.9814, + "step": 35338 + }, + { + "epoch": 6.43, + "learning_rate": 1.0730433199854388e-06, + "loss": 0.0071, + "step": 35339 + }, + { + "epoch": 6.43, + "learning_rate": 1.0729887149617766e-06, + "loss": 0.0164, + "step": 35340 + }, + { + "epoch": 6.43, + "learning_rate": 1.0729341099381143e-06, + "loss": 0.2766, + "step": 35341 + }, + { + "epoch": 6.43, + "learning_rate": 1.072879504914452e-06, + "loss": 0.0104, + "step": 35342 + }, + { + "epoch": 6.43, + "learning_rate": 1.07282489989079e-06, + "loss": 2.1298, + "step": 35343 + }, + { + "epoch": 6.43, + "learning_rate": 1.072770294867128e-06, + "loss": 0.865, + "step": 35344 + }, + { + "epoch": 6.43, + "learning_rate": 1.0727156898434656e-06, + "loss": 0.162, + "step": 35345 + }, + { + "epoch": 6.43, + "learning_rate": 1.0726610848198036e-06, + "loss": 0.0059, + "step": 35346 + }, + { + "epoch": 6.43, + "learning_rate": 1.0726064797961413e-06, + "loss": 0.0456, + "step": 35347 + }, + { + "epoch": 6.43, + "learning_rate": 1.072551874772479e-06, + "loss": 0.0334, + "step": 35348 + }, + { + "epoch": 6.43, + "learning_rate": 1.0724972697488168e-06, + "loss": 0.7951, + "step": 35349 + }, + { + "epoch": 6.43, + "learning_rate": 1.0724426647251547e-06, + "loss": 0.0076, + "step": 35350 + }, + { + "epoch": 6.43, + "learning_rate": 1.0723880597014927e-06, + "loss": 0.6438, + "step": 35351 + }, + { + "epoch": 6.43, + "learning_rate": 1.0723334546778304e-06, + "loss": 0.9821, + "step": 35352 + }, + { + "epoch": 6.43, + "learning_rate": 1.0722788496541683e-06, + "loss": 0.2992, + "step": 35353 + }, + { + "epoch": 6.43, + "learning_rate": 1.072224244630506e-06, + "loss": 0.0066, + "step": 35354 + }, + { + "epoch": 6.43, + "learning_rate": 1.0721696396068438e-06, + "loss": 0.0741, + "step": 35355 + }, + { + "epoch": 6.43, + "learning_rate": 1.0721150345831815e-06, + "loss": 1.4428, + "step": 35356 + }, + { + "epoch": 6.43, + "learning_rate": 1.0720604295595195e-06, + "loss": 0.4829, + "step": 35357 + }, + { + "epoch": 6.43, + "learning_rate": 1.0720058245358574e-06, + "loss": 0.2359, + "step": 35358 + }, + { + "epoch": 6.43, + "learning_rate": 1.0719512195121951e-06, + "loss": 2.8535, + "step": 35359 + }, + { + "epoch": 6.43, + "learning_rate": 1.071896614488533e-06, + "loss": 0.9654, + "step": 35360 + }, + { + "epoch": 6.43, + "learning_rate": 1.0718420094648708e-06, + "loss": 1.6069, + "step": 35361 + }, + { + "epoch": 6.43, + "learning_rate": 1.0717874044412086e-06, + "loss": 0.5407, + "step": 35362 + }, + { + "epoch": 6.43, + "learning_rate": 1.0717327994175463e-06, + "loss": 0.1212, + "step": 35363 + }, + { + "epoch": 6.43, + "learning_rate": 1.0716781943938842e-06, + "loss": 0.1574, + "step": 35364 + }, + { + "epoch": 6.43, + "learning_rate": 1.0716235893702222e-06, + "loss": 0.0093, + "step": 35365 + }, + { + "epoch": 6.43, + "learning_rate": 1.07156898434656e-06, + "loss": 0.1487, + "step": 35366 + }, + { + "epoch": 6.43, + "learning_rate": 1.0715143793228978e-06, + "loss": 0.8914, + "step": 35367 + }, + { + "epoch": 6.43, + "learning_rate": 1.0714597742992356e-06, + "loss": 0.0055, + "step": 35368 + }, + { + "epoch": 6.43, + "learning_rate": 1.0714051692755733e-06, + "loss": 0.0722, + "step": 35369 + }, + { + "epoch": 6.43, + "learning_rate": 1.071350564251911e-06, + "loss": 0.0261, + "step": 35370 + }, + { + "epoch": 6.43, + "learning_rate": 1.071295959228249e-06, + "loss": 0.4392, + "step": 35371 + }, + { + "epoch": 6.43, + "learning_rate": 1.071241354204587e-06, + "loss": 1.4582, + "step": 35372 + }, + { + "epoch": 6.43, + "learning_rate": 1.0711867491809247e-06, + "loss": 0.4323, + "step": 35373 + }, + { + "epoch": 6.43, + "learning_rate": 1.0711321441572626e-06, + "loss": 1.3028, + "step": 35374 + }, + { + "epoch": 6.43, + "learning_rate": 1.0710775391336003e-06, + "loss": 2.6038, + "step": 35375 + }, + { + "epoch": 6.43, + "learning_rate": 1.071022934109938e-06, + "loss": 0.3458, + "step": 35376 + }, + { + "epoch": 6.43, + "learning_rate": 1.0709683290862758e-06, + "loss": 0.2976, + "step": 35377 + }, + { + "epoch": 6.43, + "learning_rate": 1.0709137240626137e-06, + "loss": 0.4615, + "step": 35378 + }, + { + "epoch": 6.43, + "learning_rate": 1.0708591190389517e-06, + "loss": 0.0072, + "step": 35379 + }, + { + "epoch": 6.43, + "learning_rate": 1.0708045140152894e-06, + "loss": 0.5003, + "step": 35380 + }, + { + "epoch": 6.43, + "learning_rate": 1.0707499089916274e-06, + "loss": 1.5166, + "step": 35381 + }, + { + "epoch": 6.43, + "learning_rate": 1.070695303967965e-06, + "loss": 0.0728, + "step": 35382 + }, + { + "epoch": 6.43, + "learning_rate": 1.0706406989443028e-06, + "loss": 0.0107, + "step": 35383 + }, + { + "epoch": 6.43, + "learning_rate": 1.0705860939206408e-06, + "loss": 0.1711, + "step": 35384 + }, + { + "epoch": 6.43, + "learning_rate": 1.0705314888969787e-06, + "loss": 0.234, + "step": 35385 + }, + { + "epoch": 6.43, + "learning_rate": 1.0704768838733164e-06, + "loss": 0.2597, + "step": 35386 + }, + { + "epoch": 6.44, + "learning_rate": 1.0704222788496542e-06, + "loss": 0.3242, + "step": 35387 + }, + { + "epoch": 6.44, + "learning_rate": 1.0703676738259921e-06, + "loss": 0.3661, + "step": 35388 + }, + { + "epoch": 6.44, + "learning_rate": 1.0703130688023298e-06, + "loss": 0.1653, + "step": 35389 + }, + { + "epoch": 6.44, + "learning_rate": 1.0702584637786676e-06, + "loss": 1.5339, + "step": 35390 + }, + { + "epoch": 6.44, + "learning_rate": 1.0702038587550055e-06, + "loss": 0.1852, + "step": 35391 + }, + { + "epoch": 6.44, + "learning_rate": 1.0701492537313435e-06, + "loss": 0.6167, + "step": 35392 + }, + { + "epoch": 6.44, + "learning_rate": 1.0700946487076812e-06, + "loss": 1.828, + "step": 35393 + }, + { + "epoch": 6.44, + "learning_rate": 1.070040043684019e-06, + "loss": 0.3473, + "step": 35394 + }, + { + "epoch": 6.44, + "learning_rate": 1.0699854386603569e-06, + "loss": 0.0321, + "step": 35395 + }, + { + "epoch": 6.44, + "learning_rate": 1.0699308336366946e-06, + "loss": 0.1374, + "step": 35396 + }, + { + "epoch": 6.44, + "learning_rate": 1.0698762286130323e-06, + "loss": 2.7627, + "step": 35397 + }, + { + "epoch": 6.44, + "learning_rate": 1.0698216235893703e-06, + "loss": 0.1361, + "step": 35398 + }, + { + "epoch": 6.44, + "learning_rate": 1.0697670185657082e-06, + "loss": 0.5595, + "step": 35399 + }, + { + "epoch": 6.44, + "learning_rate": 1.069712413542046e-06, + "loss": 0.0511, + "step": 35400 + }, + { + "epoch": 6.44, + "learning_rate": 1.0696578085183837e-06, + "loss": 0.0685, + "step": 35401 + }, + { + "epoch": 6.44, + "learning_rate": 1.0696032034947214e-06, + "loss": 0.0033, + "step": 35402 + }, + { + "epoch": 6.44, + "learning_rate": 1.0695485984710594e-06, + "loss": 2.1925, + "step": 35403 + }, + { + "epoch": 6.44, + "learning_rate": 1.069493993447397e-06, + "loss": 1.756, + "step": 35404 + }, + { + "epoch": 6.44, + "learning_rate": 1.069439388423735e-06, + "loss": 0.0422, + "step": 35405 + }, + { + "epoch": 6.44, + "learning_rate": 1.069384783400073e-06, + "loss": 0.8903, + "step": 35406 + }, + { + "epoch": 6.44, + "learning_rate": 1.0693301783764107e-06, + "loss": 0.06, + "step": 35407 + }, + { + "epoch": 6.44, + "learning_rate": 1.0692755733527484e-06, + "loss": 0.2477, + "step": 35408 + }, + { + "epoch": 6.44, + "learning_rate": 1.0692209683290862e-06, + "loss": 0.2059, + "step": 35409 + }, + { + "epoch": 6.44, + "learning_rate": 1.0691663633054241e-06, + "loss": 0.0141, + "step": 35410 + }, + { + "epoch": 6.44, + "learning_rate": 1.0691117582817619e-06, + "loss": 1.0546, + "step": 35411 + }, + { + "epoch": 6.44, + "learning_rate": 1.0690571532580998e-06, + "loss": 0.2326, + "step": 35412 + }, + { + "epoch": 6.44, + "learning_rate": 1.0690025482344377e-06, + "loss": 0.5955, + "step": 35413 + }, + { + "epoch": 6.44, + "learning_rate": 1.0689479432107755e-06, + "loss": 0.3524, + "step": 35414 + }, + { + "epoch": 6.44, + "learning_rate": 1.0688933381871132e-06, + "loss": 1.3044, + "step": 35415 + }, + { + "epoch": 6.44, + "learning_rate": 1.068838733163451e-06, + "loss": 0.2778, + "step": 35416 + }, + { + "epoch": 6.44, + "learning_rate": 1.0687841281397889e-06, + "loss": 0.294, + "step": 35417 + }, + { + "epoch": 6.44, + "learning_rate": 1.0687295231161266e-06, + "loss": 1.3024, + "step": 35418 + }, + { + "epoch": 6.44, + "learning_rate": 1.0686749180924646e-06, + "loss": 0.2722, + "step": 35419 + }, + { + "epoch": 6.44, + "learning_rate": 1.0686203130688025e-06, + "loss": 1.2488, + "step": 35420 + }, + { + "epoch": 6.44, + "learning_rate": 1.0685657080451402e-06, + "loss": 0.3352, + "step": 35421 + }, + { + "epoch": 6.44, + "learning_rate": 1.068511103021478e-06, + "loss": 2.62, + "step": 35422 + }, + { + "epoch": 6.44, + "learning_rate": 1.0684564979978157e-06, + "loss": 0.0178, + "step": 35423 + }, + { + "epoch": 6.44, + "learning_rate": 1.0684018929741536e-06, + "loss": 1.3029, + "step": 35424 + }, + { + "epoch": 6.44, + "learning_rate": 1.0683472879504914e-06, + "loss": 0.9645, + "step": 35425 + }, + { + "epoch": 6.44, + "learning_rate": 1.0682926829268293e-06, + "loss": 1.5982, + "step": 35426 + }, + { + "epoch": 6.44, + "learning_rate": 1.0682380779031673e-06, + "loss": 0.0079, + "step": 35427 + }, + { + "epoch": 6.44, + "learning_rate": 1.068183472879505e-06, + "loss": 0.0528, + "step": 35428 + }, + { + "epoch": 6.44, + "learning_rate": 1.0681288678558427e-06, + "loss": 1.3054, + "step": 35429 + }, + { + "epoch": 6.44, + "learning_rate": 1.0680742628321804e-06, + "loss": 0.1098, + "step": 35430 + }, + { + "epoch": 6.44, + "learning_rate": 1.0680196578085184e-06, + "loss": 1.2522, + "step": 35431 + }, + { + "epoch": 6.44, + "learning_rate": 1.0679650527848563e-06, + "loss": 0.1095, + "step": 35432 + }, + { + "epoch": 6.44, + "learning_rate": 1.067910447761194e-06, + "loss": 0.4024, + "step": 35433 + }, + { + "epoch": 6.44, + "learning_rate": 1.067855842737532e-06, + "loss": 0.6107, + "step": 35434 + }, + { + "epoch": 6.44, + "learning_rate": 1.0678012377138697e-06, + "loss": 0.793, + "step": 35435 + }, + { + "epoch": 6.44, + "learning_rate": 1.0677466326902075e-06, + "loss": 0.0084, + "step": 35436 + }, + { + "epoch": 6.44, + "learning_rate": 1.0676920276665452e-06, + "loss": 0.1389, + "step": 35437 + }, + { + "epoch": 6.44, + "learning_rate": 1.0676374226428831e-06, + "loss": 0.4178, + "step": 35438 + }, + { + "epoch": 6.44, + "learning_rate": 1.067582817619221e-06, + "loss": 0.0048, + "step": 35439 + }, + { + "epoch": 6.44, + "learning_rate": 1.0675282125955588e-06, + "loss": 0.154, + "step": 35440 + }, + { + "epoch": 6.44, + "learning_rate": 1.0674736075718968e-06, + "loss": 0.0919, + "step": 35441 + }, + { + "epoch": 6.45, + "learning_rate": 1.0674190025482345e-06, + "loss": 0.9327, + "step": 35442 + }, + { + "epoch": 6.45, + "learning_rate": 1.0673643975245722e-06, + "loss": 0.2107, + "step": 35443 + }, + { + "epoch": 6.45, + "learning_rate": 1.06730979250091e-06, + "loss": 0.8663, + "step": 35444 + }, + { + "epoch": 6.45, + "learning_rate": 1.067255187477248e-06, + "loss": 0.1023, + "step": 35445 + }, + { + "epoch": 6.45, + "learning_rate": 1.0672005824535858e-06, + "loss": 0.5451, + "step": 35446 + }, + { + "epoch": 6.45, + "learning_rate": 1.0671459774299236e-06, + "loss": 0.0696, + "step": 35447 + }, + { + "epoch": 6.45, + "learning_rate": 1.0670913724062615e-06, + "loss": 0.5851, + "step": 35448 + }, + { + "epoch": 6.45, + "learning_rate": 1.0670367673825993e-06, + "loss": 1.2254, + "step": 35449 + }, + { + "epoch": 6.45, + "learning_rate": 1.066982162358937e-06, + "loss": 0.2216, + "step": 35450 + }, + { + "epoch": 6.45, + "learning_rate": 1.0669275573352747e-06, + "loss": 0.0345, + "step": 35451 + }, + { + "epoch": 6.45, + "learning_rate": 1.0668729523116127e-06, + "loss": 0.7127, + "step": 35452 + }, + { + "epoch": 6.45, + "learning_rate": 1.0668183472879506e-06, + "loss": 0.0289, + "step": 35453 + }, + { + "epoch": 6.45, + "learning_rate": 1.0667637422642883e-06, + "loss": 0.7361, + "step": 35454 + }, + { + "epoch": 6.45, + "learning_rate": 1.0667091372406263e-06, + "loss": 3.3095, + "step": 35455 + }, + { + "epoch": 6.45, + "learning_rate": 1.066654532216964e-06, + "loss": 0.2327, + "step": 35456 + }, + { + "epoch": 6.45, + "learning_rate": 1.0665999271933017e-06, + "loss": 0.6443, + "step": 35457 + }, + { + "epoch": 6.45, + "learning_rate": 1.0665453221696395e-06, + "loss": 0.1787, + "step": 35458 + }, + { + "epoch": 6.45, + "learning_rate": 1.0664907171459774e-06, + "loss": 1.8566, + "step": 35459 + }, + { + "epoch": 6.45, + "learning_rate": 1.0664361121223154e-06, + "loss": 0.0093, + "step": 35460 + }, + { + "epoch": 6.45, + "learning_rate": 1.066381507098653e-06, + "loss": 0.9829, + "step": 35461 + }, + { + "epoch": 6.45, + "learning_rate": 1.066326902074991e-06, + "loss": 3.2088, + "step": 35462 + }, + { + "epoch": 6.45, + "learning_rate": 1.0662722970513288e-06, + "loss": 0.0893, + "step": 35463 + }, + { + "epoch": 6.45, + "learning_rate": 1.0662176920276665e-06, + "loss": 1.2971, + "step": 35464 + }, + { + "epoch": 6.45, + "learning_rate": 1.0661630870040042e-06, + "loss": 0.0139, + "step": 35465 + }, + { + "epoch": 6.45, + "learning_rate": 1.0661084819803422e-06, + "loss": 0.117, + "step": 35466 + }, + { + "epoch": 6.45, + "learning_rate": 1.0660538769566801e-06, + "loss": 0.2407, + "step": 35467 + }, + { + "epoch": 6.45, + "learning_rate": 1.0659992719330178e-06, + "loss": 0.0286, + "step": 35468 + }, + { + "epoch": 6.45, + "learning_rate": 1.0659446669093558e-06, + "loss": 0.0282, + "step": 35469 + }, + { + "epoch": 6.45, + "learning_rate": 1.0658900618856935e-06, + "loss": 0.027, + "step": 35470 + }, + { + "epoch": 6.45, + "learning_rate": 1.0658354568620313e-06, + "loss": 0.1878, + "step": 35471 + }, + { + "epoch": 6.45, + "learning_rate": 1.0657808518383692e-06, + "loss": 0.0049, + "step": 35472 + }, + { + "epoch": 6.45, + "learning_rate": 1.0657262468147071e-06, + "loss": 0.2387, + "step": 35473 + }, + { + "epoch": 6.45, + "learning_rate": 1.0656716417910449e-06, + "loss": 0.0491, + "step": 35474 + }, + { + "epoch": 6.45, + "learning_rate": 1.0656170367673826e-06, + "loss": 1.9105, + "step": 35475 + }, + { + "epoch": 6.45, + "learning_rate": 1.0655624317437203e-06, + "loss": 0.1052, + "step": 35476 + }, + { + "epoch": 6.45, + "learning_rate": 1.0655078267200583e-06, + "loss": 0.0088, + "step": 35477 + }, + { + "epoch": 6.45, + "learning_rate": 1.065453221696396e-06, + "loss": 1.4868, + "step": 35478 + }, + { + "epoch": 6.45, + "learning_rate": 1.065398616672734e-06, + "loss": 2.6501, + "step": 35479 + }, + { + "epoch": 6.45, + "learning_rate": 1.065344011649072e-06, + "loss": 0.0683, + "step": 35480 + }, + { + "epoch": 6.45, + "learning_rate": 1.0652894066254096e-06, + "loss": 0.448, + "step": 35481 + }, + { + "epoch": 6.45, + "learning_rate": 1.0652348016017474e-06, + "loss": 0.2205, + "step": 35482 + }, + { + "epoch": 6.45, + "learning_rate": 1.065180196578085e-06, + "loss": 1.4596, + "step": 35483 + }, + { + "epoch": 6.45, + "learning_rate": 1.065125591554423e-06, + "loss": 0.0116, + "step": 35484 + }, + { + "epoch": 6.45, + "learning_rate": 1.0650709865307608e-06, + "loss": 2.2015, + "step": 35485 + }, + { + "epoch": 6.45, + "learning_rate": 1.0650163815070987e-06, + "loss": 1.2862, + "step": 35486 + }, + { + "epoch": 6.45, + "learning_rate": 1.0649617764834367e-06, + "loss": 0.2074, + "step": 35487 + }, + { + "epoch": 6.45, + "learning_rate": 1.0649071714597744e-06, + "loss": 0.5197, + "step": 35488 + }, + { + "epoch": 6.45, + "learning_rate": 1.0648525664361121e-06, + "loss": 0.3887, + "step": 35489 + }, + { + "epoch": 6.45, + "learning_rate": 1.0647979614124499e-06, + "loss": 1.1656, + "step": 35490 + }, + { + "epoch": 6.45, + "learning_rate": 1.0647433563887878e-06, + "loss": 0.0798, + "step": 35491 + }, + { + "epoch": 6.45, + "learning_rate": 1.0646887513651255e-06, + "loss": 1.377, + "step": 35492 + }, + { + "epoch": 6.45, + "learning_rate": 1.0646341463414635e-06, + "loss": 0.0163, + "step": 35493 + }, + { + "epoch": 6.45, + "learning_rate": 1.0645795413178014e-06, + "loss": 0.0076, + "step": 35494 + }, + { + "epoch": 6.45, + "learning_rate": 1.0645249362941391e-06, + "loss": 0.0251, + "step": 35495 + }, + { + "epoch": 6.45, + "learning_rate": 1.0644703312704769e-06, + "loss": 1.1031, + "step": 35496 + }, + { + "epoch": 6.46, + "learning_rate": 1.0644157262468146e-06, + "loss": 0.3357, + "step": 35497 + }, + { + "epoch": 6.46, + "learning_rate": 1.0643611212231526e-06, + "loss": 0.2769, + "step": 35498 + }, + { + "epoch": 6.46, + "learning_rate": 1.0643065161994903e-06, + "loss": 1.4092, + "step": 35499 + }, + { + "epoch": 6.46, + "learning_rate": 1.0642519111758282e-06, + "loss": 2.4674, + "step": 35500 + }, + { + "epoch": 6.46, + "learning_rate": 1.0641973061521662e-06, + "loss": 1.1922, + "step": 35501 + }, + { + "epoch": 6.46, + "learning_rate": 1.064142701128504e-06, + "loss": 0.0333, + "step": 35502 + }, + { + "epoch": 6.46, + "learning_rate": 1.0640880961048416e-06, + "loss": 2.9835, + "step": 35503 + }, + { + "epoch": 6.46, + "learning_rate": 1.0640334910811794e-06, + "loss": 0.3193, + "step": 35504 + }, + { + "epoch": 6.46, + "learning_rate": 1.0639788860575173e-06, + "loss": 0.3058, + "step": 35505 + }, + { + "epoch": 6.46, + "learning_rate": 1.063924281033855e-06, + "loss": 0.1546, + "step": 35506 + }, + { + "epoch": 6.46, + "learning_rate": 1.063869676010193e-06, + "loss": 0.0471, + "step": 35507 + }, + { + "epoch": 6.46, + "learning_rate": 1.063815070986531e-06, + "loss": 0.187, + "step": 35508 + }, + { + "epoch": 6.46, + "learning_rate": 1.0637604659628687e-06, + "loss": 0.0447, + "step": 35509 + }, + { + "epoch": 6.46, + "learning_rate": 1.0637058609392064e-06, + "loss": 1.4955, + "step": 35510 + }, + { + "epoch": 6.46, + "learning_rate": 1.0636512559155441e-06, + "loss": 0.406, + "step": 35511 + }, + { + "epoch": 6.46, + "learning_rate": 1.063596650891882e-06, + "loss": 0.0127, + "step": 35512 + }, + { + "epoch": 6.46, + "learning_rate": 1.06354204586822e-06, + "loss": 1.0183, + "step": 35513 + }, + { + "epoch": 6.46, + "learning_rate": 1.0634874408445577e-06, + "loss": 0.39, + "step": 35514 + }, + { + "epoch": 6.46, + "learning_rate": 1.0634328358208957e-06, + "loss": 0.1087, + "step": 35515 + }, + { + "epoch": 6.46, + "learning_rate": 1.0633782307972334e-06, + "loss": 0.3992, + "step": 35516 + }, + { + "epoch": 6.46, + "learning_rate": 1.0633236257735711e-06, + "loss": 0.3102, + "step": 35517 + }, + { + "epoch": 6.46, + "learning_rate": 1.0632690207499089e-06, + "loss": 0.0886, + "step": 35518 + }, + { + "epoch": 6.46, + "learning_rate": 1.0632144157262468e-06, + "loss": 0.2089, + "step": 35519 + }, + { + "epoch": 6.46, + "learning_rate": 1.0631598107025848e-06, + "loss": 1.2757, + "step": 35520 + }, + { + "epoch": 6.46, + "learning_rate": 1.0631052056789225e-06, + "loss": 2.0323, + "step": 35521 + }, + { + "epoch": 6.46, + "learning_rate": 1.0630506006552604e-06, + "loss": 0.0576, + "step": 35522 + }, + { + "epoch": 6.46, + "learning_rate": 1.0629959956315982e-06, + "loss": 0.0429, + "step": 35523 + }, + { + "epoch": 6.46, + "learning_rate": 1.062941390607936e-06, + "loss": 0.4451, + "step": 35524 + }, + { + "epoch": 6.46, + "learning_rate": 1.0628867855842736e-06, + "loss": 0.0482, + "step": 35525 + }, + { + "epoch": 6.46, + "learning_rate": 1.0628321805606116e-06, + "loss": 0.0922, + "step": 35526 + }, + { + "epoch": 6.46, + "learning_rate": 1.0627775755369495e-06, + "loss": 0.0638, + "step": 35527 + }, + { + "epoch": 6.46, + "learning_rate": 1.0627229705132873e-06, + "loss": 0.2814, + "step": 35528 + }, + { + "epoch": 6.46, + "learning_rate": 1.0626683654896252e-06, + "loss": 0.0409, + "step": 35529 + }, + { + "epoch": 6.46, + "learning_rate": 1.062613760465963e-06, + "loss": 1.1768, + "step": 35530 + }, + { + "epoch": 6.46, + "learning_rate": 1.0625591554423007e-06, + "loss": 1.3252, + "step": 35531 + }, + { + "epoch": 6.46, + "learning_rate": 1.0625045504186384e-06, + "loss": 2.3433, + "step": 35532 + }, + { + "epoch": 6.46, + "learning_rate": 1.0624499453949763e-06, + "loss": 2.3256, + "step": 35533 + }, + { + "epoch": 6.46, + "learning_rate": 1.0623953403713143e-06, + "loss": 1.4792, + "step": 35534 + }, + { + "epoch": 6.46, + "learning_rate": 1.062340735347652e-06, + "loss": 0.2826, + "step": 35535 + }, + { + "epoch": 6.46, + "learning_rate": 1.06228613032399e-06, + "loss": 1.0892, + "step": 35536 + }, + { + "epoch": 6.46, + "learning_rate": 1.0622315253003277e-06, + "loss": 0.6491, + "step": 35537 + }, + { + "epoch": 6.46, + "learning_rate": 1.0621769202766654e-06, + "loss": 1.1612, + "step": 35538 + }, + { + "epoch": 6.46, + "learning_rate": 1.0621223152530031e-06, + "loss": 0.1009, + "step": 35539 + }, + { + "epoch": 6.46, + "learning_rate": 1.062067710229341e-06, + "loss": 0.735, + "step": 35540 + }, + { + "epoch": 6.46, + "learning_rate": 1.062013105205679e-06, + "loss": 0.6041, + "step": 35541 + }, + { + "epoch": 6.46, + "learning_rate": 1.0619585001820168e-06, + "loss": 1.8156, + "step": 35542 + }, + { + "epoch": 6.46, + "learning_rate": 1.0619038951583547e-06, + "loss": 0.1357, + "step": 35543 + }, + { + "epoch": 6.46, + "learning_rate": 1.0618492901346924e-06, + "loss": 0.6039, + "step": 35544 + }, + { + "epoch": 6.46, + "learning_rate": 1.0617946851110302e-06, + "loss": 0.0169, + "step": 35545 + }, + { + "epoch": 6.46, + "learning_rate": 1.061740080087368e-06, + "loss": 0.8989, + "step": 35546 + }, + { + "epoch": 6.46, + "learning_rate": 1.0616854750637058e-06, + "loss": 0.7482, + "step": 35547 + }, + { + "epoch": 6.46, + "learning_rate": 1.0616308700400438e-06, + "loss": 0.5993, + "step": 35548 + }, + { + "epoch": 6.46, + "learning_rate": 1.0615762650163815e-06, + "loss": 0.1719, + "step": 35549 + }, + { + "epoch": 6.46, + "learning_rate": 1.0615216599927195e-06, + "loss": 0.0424, + "step": 35550 + }, + { + "epoch": 6.46, + "learning_rate": 1.0614670549690572e-06, + "loss": 0.3406, + "step": 35551 + }, + { + "epoch": 6.47, + "learning_rate": 1.061412449945395e-06, + "loss": 0.0394, + "step": 35552 + }, + { + "epoch": 6.47, + "learning_rate": 1.0613578449217329e-06, + "loss": 1.9277, + "step": 35553 + }, + { + "epoch": 6.47, + "learning_rate": 1.0613032398980706e-06, + "loss": 0.0402, + "step": 35554 + }, + { + "epoch": 6.47, + "learning_rate": 1.0612486348744085e-06, + "loss": 0.4102, + "step": 35555 + }, + { + "epoch": 6.47, + "learning_rate": 1.0611940298507463e-06, + "loss": 0.0331, + "step": 35556 + }, + { + "epoch": 6.47, + "learning_rate": 1.061139424827084e-06, + "loss": 1.3357, + "step": 35557 + }, + { + "epoch": 6.47, + "learning_rate": 1.061084819803422e-06, + "loss": 1.4863, + "step": 35558 + }, + { + "epoch": 6.47, + "learning_rate": 1.0610302147797597e-06, + "loss": 0.1584, + "step": 35559 + }, + { + "epoch": 6.47, + "learning_rate": 1.0609756097560976e-06, + "loss": 0.0077, + "step": 35560 + }, + { + "epoch": 6.47, + "learning_rate": 1.0609210047324356e-06, + "loss": 0.029, + "step": 35561 + }, + { + "epoch": 6.47, + "learning_rate": 1.0608663997087733e-06, + "loss": 1.0254, + "step": 35562 + }, + { + "epoch": 6.47, + "learning_rate": 1.060811794685111e-06, + "loss": 0.1547, + "step": 35563 + }, + { + "epoch": 6.47, + "learning_rate": 1.0607571896614488e-06, + "loss": 0.012, + "step": 35564 + }, + { + "epoch": 6.47, + "learning_rate": 1.0607025846377867e-06, + "loss": 0.0073, + "step": 35565 + }, + { + "epoch": 6.47, + "learning_rate": 1.0606479796141244e-06, + "loss": 0.2858, + "step": 35566 + }, + { + "epoch": 6.47, + "learning_rate": 1.0605933745904624e-06, + "loss": 0.5044, + "step": 35567 + }, + { + "epoch": 6.47, + "learning_rate": 1.0605387695668003e-06, + "loss": 0.1006, + "step": 35568 + }, + { + "epoch": 6.47, + "learning_rate": 1.060484164543138e-06, + "loss": 0.4122, + "step": 35569 + }, + { + "epoch": 6.47, + "learning_rate": 1.0604295595194758e-06, + "loss": 0.0229, + "step": 35570 + }, + { + "epoch": 6.47, + "learning_rate": 1.0603749544958135e-06, + "loss": 0.2744, + "step": 35571 + }, + { + "epoch": 6.47, + "learning_rate": 1.0603203494721515e-06, + "loss": 0.5214, + "step": 35572 + }, + { + "epoch": 6.47, + "learning_rate": 1.0602657444484892e-06, + "loss": 0.0614, + "step": 35573 + }, + { + "epoch": 6.47, + "learning_rate": 1.0602111394248271e-06, + "loss": 0.306, + "step": 35574 + }, + { + "epoch": 6.47, + "learning_rate": 1.060156534401165e-06, + "loss": 0.6652, + "step": 35575 + }, + { + "epoch": 6.47, + "learning_rate": 1.0601019293775028e-06, + "loss": 0.0172, + "step": 35576 + }, + { + "epoch": 6.47, + "learning_rate": 1.0600473243538405e-06, + "loss": 1.1072, + "step": 35577 + }, + { + "epoch": 6.47, + "learning_rate": 1.0599927193301783e-06, + "loss": 0.4149, + "step": 35578 + }, + { + "epoch": 6.47, + "learning_rate": 1.0599381143065162e-06, + "loss": 2.7793, + "step": 35579 + }, + { + "epoch": 6.47, + "learning_rate": 1.059883509282854e-06, + "loss": 0.5229, + "step": 35580 + }, + { + "epoch": 6.47, + "learning_rate": 1.059828904259192e-06, + "loss": 0.0798, + "step": 35581 + }, + { + "epoch": 6.47, + "learning_rate": 1.0597742992355298e-06, + "loss": 0.0433, + "step": 35582 + }, + { + "epoch": 6.47, + "learning_rate": 1.0597196942118676e-06, + "loss": 0.1807, + "step": 35583 + }, + { + "epoch": 6.47, + "learning_rate": 1.0596650891882053e-06, + "loss": 0.1554, + "step": 35584 + }, + { + "epoch": 6.47, + "learning_rate": 1.059610484164543e-06, + "loss": 0.2291, + "step": 35585 + }, + { + "epoch": 6.47, + "learning_rate": 1.059555879140881e-06, + "loss": 2.0013, + "step": 35586 + }, + { + "epoch": 6.47, + "learning_rate": 1.0595012741172187e-06, + "loss": 0.0126, + "step": 35587 + }, + { + "epoch": 6.47, + "learning_rate": 1.0594466690935567e-06, + "loss": 1.1014, + "step": 35588 + }, + { + "epoch": 6.47, + "learning_rate": 1.0593920640698946e-06, + "loss": 1.5384, + "step": 35589 + }, + { + "epoch": 6.47, + "learning_rate": 1.0593374590462323e-06, + "loss": 0.8068, + "step": 35590 + }, + { + "epoch": 6.47, + "learning_rate": 1.05928285402257e-06, + "loss": 0.3567, + "step": 35591 + }, + { + "epoch": 6.47, + "learning_rate": 1.0592282489989078e-06, + "loss": 0.631, + "step": 35592 + }, + { + "epoch": 6.47, + "learning_rate": 1.0591736439752457e-06, + "loss": 0.5363, + "step": 35593 + }, + { + "epoch": 6.47, + "learning_rate": 1.0591190389515835e-06, + "loss": 0.1114, + "step": 35594 + }, + { + "epoch": 6.47, + "learning_rate": 1.0590644339279214e-06, + "loss": 0.2259, + "step": 35595 + }, + { + "epoch": 6.47, + "learning_rate": 1.0590098289042594e-06, + "loss": 0.7927, + "step": 35596 + }, + { + "epoch": 6.47, + "learning_rate": 1.058955223880597e-06, + "loss": 0.6165, + "step": 35597 + }, + { + "epoch": 6.47, + "learning_rate": 1.0589006188569348e-06, + "loss": 0.1485, + "step": 35598 + }, + { + "epoch": 6.47, + "learning_rate": 1.0588460138332726e-06, + "loss": 1.27, + "step": 35599 + }, + { + "epoch": 6.47, + "learning_rate": 1.0587914088096105e-06, + "loss": 0.0068, + "step": 35600 + }, + { + "epoch": 6.47, + "learning_rate": 1.0587368037859484e-06, + "loss": 1.2039, + "step": 35601 + }, + { + "epoch": 6.47, + "learning_rate": 1.0586821987622862e-06, + "loss": 0.2166, + "step": 35602 + }, + { + "epoch": 6.47, + "learning_rate": 1.0586275937386241e-06, + "loss": 0.2628, + "step": 35603 + }, + { + "epoch": 6.47, + "learning_rate": 1.0585729887149618e-06, + "loss": 0.1365, + "step": 35604 + }, + { + "epoch": 6.47, + "learning_rate": 1.0585183836912996e-06, + "loss": 1.1598, + "step": 35605 + }, + { + "epoch": 6.47, + "learning_rate": 1.0584637786676373e-06, + "loss": 1.2558, + "step": 35606 + }, + { + "epoch": 6.48, + "learning_rate": 1.0584091736439753e-06, + "loss": 0.8899, + "step": 35607 + }, + { + "epoch": 6.48, + "learning_rate": 1.0583545686203132e-06, + "loss": 1.4135, + "step": 35608 + }, + { + "epoch": 6.48, + "learning_rate": 1.058299963596651e-06, + "loss": 2.4586, + "step": 35609 + }, + { + "epoch": 6.48, + "learning_rate": 1.0582453585729889e-06, + "loss": 0.9796, + "step": 35610 + }, + { + "epoch": 6.48, + "learning_rate": 1.0581907535493266e-06, + "loss": 2.0101, + "step": 35611 + }, + { + "epoch": 6.48, + "learning_rate": 1.0581361485256643e-06, + "loss": 0.9378, + "step": 35612 + }, + { + "epoch": 6.48, + "learning_rate": 1.058081543502002e-06, + "loss": 0.1957, + "step": 35613 + }, + { + "epoch": 6.48, + "learning_rate": 1.05802693847834e-06, + "loss": 0.6387, + "step": 35614 + }, + { + "epoch": 6.48, + "learning_rate": 1.057972333454678e-06, + "loss": 1.4317, + "step": 35615 + }, + { + "epoch": 6.48, + "learning_rate": 1.0579177284310157e-06, + "loss": 0.0853, + "step": 35616 + }, + { + "epoch": 6.48, + "learning_rate": 1.0578631234073536e-06, + "loss": 1.3965, + "step": 35617 + }, + { + "epoch": 6.48, + "learning_rate": 1.0578085183836914e-06, + "loss": 0.5162, + "step": 35618 + }, + { + "epoch": 6.48, + "learning_rate": 1.057753913360029e-06, + "loss": 0.4914, + "step": 35619 + }, + { + "epoch": 6.48, + "learning_rate": 1.0576993083363668e-06, + "loss": 0.0069, + "step": 35620 + }, + { + "epoch": 6.48, + "learning_rate": 1.0576447033127048e-06, + "loss": 0.187, + "step": 35621 + }, + { + "epoch": 6.48, + "learning_rate": 1.0575900982890427e-06, + "loss": 0.3102, + "step": 35622 + }, + { + "epoch": 6.48, + "learning_rate": 1.0575354932653804e-06, + "loss": 0.1438, + "step": 35623 + }, + { + "epoch": 6.48, + "learning_rate": 1.0574808882417184e-06, + "loss": 0.1001, + "step": 35624 + }, + { + "epoch": 6.48, + "learning_rate": 1.0574262832180561e-06, + "loss": 0.0543, + "step": 35625 + }, + { + "epoch": 6.48, + "learning_rate": 1.0573716781943938e-06, + "loss": 0.0804, + "step": 35626 + }, + { + "epoch": 6.48, + "learning_rate": 1.0573170731707316e-06, + "loss": 0.3955, + "step": 35627 + }, + { + "epoch": 6.48, + "learning_rate": 1.0572624681470695e-06, + "loss": 0.2239, + "step": 35628 + }, + { + "epoch": 6.48, + "learning_rate": 1.0572078631234075e-06, + "loss": 0.0742, + "step": 35629 + }, + { + "epoch": 6.48, + "learning_rate": 1.0571532580997452e-06, + "loss": 0.0332, + "step": 35630 + }, + { + "epoch": 6.48, + "learning_rate": 1.057098653076083e-06, + "loss": 3.1668, + "step": 35631 + }, + { + "epoch": 6.48, + "learning_rate": 1.0570440480524209e-06, + "loss": 0.0298, + "step": 35632 + }, + { + "epoch": 6.48, + "learning_rate": 1.0569894430287586e-06, + "loss": 0.0722, + "step": 35633 + }, + { + "epoch": 6.48, + "learning_rate": 1.0569348380050963e-06, + "loss": 1.3843, + "step": 35634 + }, + { + "epoch": 6.48, + "learning_rate": 1.0568802329814343e-06, + "loss": 0.7736, + "step": 35635 + }, + { + "epoch": 6.48, + "learning_rate": 1.0568256279577722e-06, + "loss": 2.1873, + "step": 35636 + }, + { + "epoch": 6.48, + "learning_rate": 1.05677102293411e-06, + "loss": 1.354, + "step": 35637 + }, + { + "epoch": 6.48, + "learning_rate": 1.0567164179104477e-06, + "loss": 1.1055, + "step": 35638 + }, + { + "epoch": 6.48, + "learning_rate": 1.0566618128867856e-06, + "loss": 0.0052, + "step": 35639 + }, + { + "epoch": 6.48, + "learning_rate": 1.0566072078631234e-06, + "loss": 0.7261, + "step": 35640 + }, + { + "epoch": 6.48, + "learning_rate": 1.0565526028394613e-06, + "loss": 0.0746, + "step": 35641 + }, + { + "epoch": 6.48, + "learning_rate": 1.0564979978157992e-06, + "loss": 0.0079, + "step": 35642 + }, + { + "epoch": 6.48, + "learning_rate": 1.056443392792137e-06, + "loss": 0.0469, + "step": 35643 + }, + { + "epoch": 6.48, + "learning_rate": 1.0563887877684747e-06, + "loss": 0.0122, + "step": 35644 + }, + { + "epoch": 6.48, + "learning_rate": 1.0563341827448124e-06, + "loss": 3.0174, + "step": 35645 + }, + { + "epoch": 6.48, + "learning_rate": 1.0562795777211504e-06, + "loss": 1.5751, + "step": 35646 + }, + { + "epoch": 6.48, + "learning_rate": 1.0562249726974881e-06, + "loss": 0.5681, + "step": 35647 + }, + { + "epoch": 6.48, + "learning_rate": 1.056170367673826e-06, + "loss": 1.5069, + "step": 35648 + }, + { + "epoch": 6.48, + "learning_rate": 1.056115762650164e-06, + "loss": 0.083, + "step": 35649 + }, + { + "epoch": 6.48, + "learning_rate": 1.0560611576265017e-06, + "loss": 0.4472, + "step": 35650 + }, + { + "epoch": 6.48, + "learning_rate": 1.0560065526028395e-06, + "loss": 1.2427, + "step": 35651 + }, + { + "epoch": 6.48, + "learning_rate": 1.0559519475791772e-06, + "loss": 1.6379, + "step": 35652 + }, + { + "epoch": 6.48, + "learning_rate": 1.0558973425555151e-06, + "loss": 1.0726, + "step": 35653 + }, + { + "epoch": 6.48, + "learning_rate": 1.0558427375318529e-06, + "loss": 3.562, + "step": 35654 + }, + { + "epoch": 6.48, + "learning_rate": 1.0557881325081908e-06, + "loss": 0.0101, + "step": 35655 + }, + { + "epoch": 6.48, + "learning_rate": 1.0557335274845288e-06, + "loss": 0.1302, + "step": 35656 + }, + { + "epoch": 6.48, + "learning_rate": 1.0556789224608665e-06, + "loss": 1.1949, + "step": 35657 + }, + { + "epoch": 6.48, + "learning_rate": 1.0556243174372042e-06, + "loss": 2.7077, + "step": 35658 + }, + { + "epoch": 6.48, + "learning_rate": 1.055569712413542e-06, + "loss": 0.4398, + "step": 35659 + }, + { + "epoch": 6.48, + "learning_rate": 1.05551510738988e-06, + "loss": 0.4803, + "step": 35660 + }, + { + "epoch": 6.48, + "learning_rate": 1.0554605023662176e-06, + "loss": 1.2378, + "step": 35661 + }, + { + "epoch": 6.49, + "learning_rate": 1.0554058973425556e-06, + "loss": 1.2973, + "step": 35662 + }, + { + "epoch": 6.49, + "learning_rate": 1.0553512923188935e-06, + "loss": 0.0174, + "step": 35663 + }, + { + "epoch": 6.49, + "learning_rate": 1.0552966872952312e-06, + "loss": 0.1959, + "step": 35664 + }, + { + "epoch": 6.49, + "learning_rate": 1.055242082271569e-06, + "loss": 0.2797, + "step": 35665 + }, + { + "epoch": 6.49, + "learning_rate": 1.0551874772479067e-06, + "loss": 2.577, + "step": 35666 + }, + { + "epoch": 6.49, + "learning_rate": 1.0551328722242447e-06, + "loss": 0.0102, + "step": 35667 + }, + { + "epoch": 6.49, + "learning_rate": 1.0550782672005824e-06, + "loss": 1.2895, + "step": 35668 + }, + { + "epoch": 6.49, + "learning_rate": 1.0550236621769203e-06, + "loss": 0.0083, + "step": 35669 + }, + { + "epoch": 6.49, + "learning_rate": 1.0549690571532583e-06, + "loss": 0.0055, + "step": 35670 + }, + { + "epoch": 6.49, + "learning_rate": 1.054914452129596e-06, + "loss": 0.4415, + "step": 35671 + }, + { + "epoch": 6.49, + "learning_rate": 1.0548598471059337e-06, + "loss": 0.3839, + "step": 35672 + }, + { + "epoch": 6.49, + "learning_rate": 1.0548052420822715e-06, + "loss": 1.8355, + "step": 35673 + }, + { + "epoch": 6.49, + "learning_rate": 1.0547506370586094e-06, + "loss": 0.0545, + "step": 35674 + }, + { + "epoch": 6.49, + "learning_rate": 1.0546960320349471e-06, + "loss": 2.991, + "step": 35675 + }, + { + "epoch": 6.49, + "learning_rate": 1.054641427011285e-06, + "loss": 0.1382, + "step": 35676 + }, + { + "epoch": 6.49, + "learning_rate": 1.054586821987623e-06, + "loss": 0.0188, + "step": 35677 + }, + { + "epoch": 6.49, + "learning_rate": 1.0545322169639608e-06, + "loss": 0.9245, + "step": 35678 + }, + { + "epoch": 6.49, + "learning_rate": 1.0544776119402985e-06, + "loss": 0.3832, + "step": 35679 + }, + { + "epoch": 6.49, + "learning_rate": 1.0544230069166362e-06, + "loss": 0.6679, + "step": 35680 + }, + { + "epoch": 6.49, + "learning_rate": 1.0543684018929742e-06, + "loss": 0.2221, + "step": 35681 + }, + { + "epoch": 6.49, + "learning_rate": 1.0543137968693121e-06, + "loss": 0.0465, + "step": 35682 + }, + { + "epoch": 6.49, + "learning_rate": 1.0542591918456498e-06, + "loss": 0.5273, + "step": 35683 + }, + { + "epoch": 6.49, + "learning_rate": 1.0542045868219878e-06, + "loss": 1.4663, + "step": 35684 + }, + { + "epoch": 6.49, + "learning_rate": 1.0541499817983255e-06, + "loss": 0.0064, + "step": 35685 + }, + { + "epoch": 6.49, + "learning_rate": 1.0540953767746633e-06, + "loss": 0.7123, + "step": 35686 + }, + { + "epoch": 6.49, + "learning_rate": 1.054040771751001e-06, + "loss": 1.1035, + "step": 35687 + }, + { + "epoch": 6.49, + "learning_rate": 1.053986166727339e-06, + "loss": 0.1023, + "step": 35688 + }, + { + "epoch": 6.49, + "learning_rate": 1.0539315617036769e-06, + "loss": 1.0535, + "step": 35689 + }, + { + "epoch": 6.49, + "learning_rate": 1.0538769566800146e-06, + "loss": 0.5465, + "step": 35690 + }, + { + "epoch": 6.49, + "learning_rate": 1.0538223516563525e-06, + "loss": 1.7536, + "step": 35691 + }, + { + "epoch": 6.49, + "learning_rate": 1.0537677466326903e-06, + "loss": 0.0088, + "step": 35692 + }, + { + "epoch": 6.49, + "learning_rate": 1.053713141609028e-06, + "loss": 0.2557, + "step": 35693 + }, + { + "epoch": 6.49, + "learning_rate": 1.0536585365853657e-06, + "loss": 0.0051, + "step": 35694 + }, + { + "epoch": 6.49, + "learning_rate": 1.0536039315617037e-06, + "loss": 0.1946, + "step": 35695 + }, + { + "epoch": 6.49, + "learning_rate": 1.0535493265380416e-06, + "loss": 0.6235, + "step": 35696 + }, + { + "epoch": 6.49, + "learning_rate": 1.0534947215143794e-06, + "loss": 1.349, + "step": 35697 + }, + { + "epoch": 6.49, + "learning_rate": 1.0534401164907173e-06, + "loss": 0.0052, + "step": 35698 + }, + { + "epoch": 6.49, + "learning_rate": 1.053385511467055e-06, + "loss": 0.552, + "step": 35699 + }, + { + "epoch": 6.49, + "learning_rate": 1.0533309064433928e-06, + "loss": 0.0162, + "step": 35700 + }, + { + "epoch": 6.49, + "learning_rate": 1.0532763014197305e-06, + "loss": 1.2605, + "step": 35701 + }, + { + "epoch": 6.49, + "learning_rate": 1.0532216963960684e-06, + "loss": 1.3697, + "step": 35702 + }, + { + "epoch": 6.49, + "learning_rate": 1.0531670913724064e-06, + "loss": 0.2777, + "step": 35703 + }, + { + "epoch": 6.49, + "learning_rate": 1.0531124863487441e-06, + "loss": 0.0083, + "step": 35704 + }, + { + "epoch": 6.49, + "learning_rate": 1.053057881325082e-06, + "loss": 0.0121, + "step": 35705 + }, + { + "epoch": 6.49, + "learning_rate": 1.0530032763014198e-06, + "loss": 0.1902, + "step": 35706 + }, + { + "epoch": 6.49, + "learning_rate": 1.0529486712777575e-06, + "loss": 0.3629, + "step": 35707 + }, + { + "epoch": 6.49, + "learning_rate": 1.0528940662540953e-06, + "loss": 5.77, + "step": 35708 + }, + { + "epoch": 6.49, + "learning_rate": 1.0528394612304332e-06, + "loss": 1.3822, + "step": 35709 + }, + { + "epoch": 6.49, + "learning_rate": 1.0527848562067711e-06, + "loss": 0.2303, + "step": 35710 + }, + { + "epoch": 6.49, + "learning_rate": 1.0527302511831089e-06, + "loss": 1.929, + "step": 35711 + }, + { + "epoch": 6.49, + "learning_rate": 1.0526756461594466e-06, + "loss": 0.005, + "step": 35712 + }, + { + "epoch": 6.49, + "learning_rate": 1.0526210411357845e-06, + "loss": 1.4912, + "step": 35713 + }, + { + "epoch": 6.49, + "learning_rate": 1.0525664361121223e-06, + "loss": 0.202, + "step": 35714 + }, + { + "epoch": 6.49, + "learning_rate": 1.05251183108846e-06, + "loss": 0.064, + "step": 35715 + }, + { + "epoch": 6.49, + "learning_rate": 1.052457226064798e-06, + "loss": 2.684, + "step": 35716 + }, + { + "epoch": 6.5, + "learning_rate": 1.052402621041136e-06, + "loss": 0.217, + "step": 35717 + }, + { + "epoch": 6.5, + "learning_rate": 1.0523480160174736e-06, + "loss": 0.1765, + "step": 35718 + }, + { + "epoch": 6.5, + "learning_rate": 1.0522934109938114e-06, + "loss": 1.4812, + "step": 35719 + }, + { + "epoch": 6.5, + "learning_rate": 1.0522388059701493e-06, + "loss": 0.2881, + "step": 35720 + }, + { + "epoch": 6.5, + "learning_rate": 1.052184200946487e-06, + "loss": 1.0635, + "step": 35721 + }, + { + "epoch": 6.5, + "learning_rate": 1.052129595922825e-06, + "loss": 0.0236, + "step": 35722 + }, + { + "epoch": 6.5, + "learning_rate": 1.0520749908991627e-06, + "loss": 0.0098, + "step": 35723 + }, + { + "epoch": 6.5, + "learning_rate": 1.0520203858755007e-06, + "loss": 0.8733, + "step": 35724 + }, + { + "epoch": 6.5, + "learning_rate": 1.0519657808518384e-06, + "loss": 1.3058, + "step": 35725 + }, + { + "epoch": 6.5, + "learning_rate": 1.0519111758281761e-06, + "loss": 0.0089, + "step": 35726 + }, + { + "epoch": 6.5, + "learning_rate": 1.051856570804514e-06, + "loss": 0.1884, + "step": 35727 + }, + { + "epoch": 6.5, + "learning_rate": 1.0518019657808518e-06, + "loss": 0.4177, + "step": 35728 + }, + { + "epoch": 6.5, + "learning_rate": 1.0517473607571897e-06, + "loss": 0.2613, + "step": 35729 + }, + { + "epoch": 6.5, + "learning_rate": 1.0516927557335277e-06, + "loss": 2.0022, + "step": 35730 + }, + { + "epoch": 6.5, + "learning_rate": 1.0516381507098654e-06, + "loss": 0.8996, + "step": 35731 + }, + { + "epoch": 6.5, + "learning_rate": 1.0515835456862031e-06, + "loss": 0.004, + "step": 35732 + }, + { + "epoch": 6.5, + "learning_rate": 1.0515289406625409e-06, + "loss": 1.2625, + "step": 35733 + }, + { + "epoch": 6.5, + "learning_rate": 1.0514743356388788e-06, + "loss": 0.0114, + "step": 35734 + }, + { + "epoch": 6.5, + "learning_rate": 1.0514197306152165e-06, + "loss": 0.6276, + "step": 35735 + }, + { + "epoch": 6.5, + "learning_rate": 1.0513651255915545e-06, + "loss": 0.1731, + "step": 35736 + }, + { + "epoch": 6.5, + "learning_rate": 1.0513105205678924e-06, + "loss": 1.3423, + "step": 35737 + }, + { + "epoch": 6.5, + "learning_rate": 1.0512559155442302e-06, + "loss": 0.0061, + "step": 35738 + }, + { + "epoch": 6.5, + "learning_rate": 1.051201310520568e-06, + "loss": 0.1991, + "step": 35739 + }, + { + "epoch": 6.5, + "learning_rate": 1.0511467054969056e-06, + "loss": 0.0162, + "step": 35740 + }, + { + "epoch": 6.5, + "learning_rate": 1.0510921004732436e-06, + "loss": 0.0827, + "step": 35741 + }, + { + "epoch": 6.5, + "learning_rate": 1.0510374954495813e-06, + "loss": 0.0424, + "step": 35742 + }, + { + "epoch": 6.5, + "learning_rate": 1.0509828904259192e-06, + "loss": 0.0147, + "step": 35743 + }, + { + "epoch": 6.5, + "learning_rate": 1.0509282854022572e-06, + "loss": 0.3722, + "step": 35744 + }, + { + "epoch": 6.5, + "learning_rate": 1.050873680378595e-06, + "loss": 0.006, + "step": 35745 + }, + { + "epoch": 6.5, + "learning_rate": 1.0508190753549327e-06, + "loss": 0.0174, + "step": 35746 + }, + { + "epoch": 6.5, + "learning_rate": 1.0507644703312704e-06, + "loss": 0.0922, + "step": 35747 + }, + { + "epoch": 6.5, + "learning_rate": 1.0507098653076083e-06, + "loss": 4.3776, + "step": 35748 + }, + { + "epoch": 6.5, + "learning_rate": 1.050655260283946e-06, + "loss": 0.6047, + "step": 35749 + }, + { + "epoch": 6.5, + "learning_rate": 1.050600655260284e-06, + "loss": 1.015, + "step": 35750 + }, + { + "epoch": 6.5, + "learning_rate": 1.050546050236622e-06, + "loss": 0.6385, + "step": 35751 + }, + { + "epoch": 6.5, + "learning_rate": 1.0504914452129597e-06, + "loss": 1.2901, + "step": 35752 + }, + { + "epoch": 6.5, + "learning_rate": 1.0504368401892974e-06, + "loss": 0.2285, + "step": 35753 + }, + { + "epoch": 6.5, + "learning_rate": 1.0503822351656351e-06, + "loss": 0.0172, + "step": 35754 + }, + { + "epoch": 6.5, + "learning_rate": 1.050327630141973e-06, + "loss": 0.2819, + "step": 35755 + }, + { + "epoch": 6.5, + "learning_rate": 1.0502730251183108e-06, + "loss": 0.8187, + "step": 35756 + }, + { + "epoch": 6.5, + "learning_rate": 1.0502184200946488e-06, + "loss": 0.5091, + "step": 35757 + }, + { + "epoch": 6.5, + "learning_rate": 1.0501638150709867e-06, + "loss": 0.8946, + "step": 35758 + }, + { + "epoch": 6.5, + "learning_rate": 1.0501092100473244e-06, + "loss": 0.7439, + "step": 35759 + }, + { + "epoch": 6.5, + "learning_rate": 1.0500546050236622e-06, + "loss": 0.331, + "step": 35760 + }, + { + "epoch": 6.5, + "learning_rate": 1.05e-06, + "loss": 0.1709, + "step": 35761 + }, + { + "epoch": 6.5, + "learning_rate": 1.0499453949763378e-06, + "loss": 0.7521, + "step": 35762 + }, + { + "epoch": 6.5, + "learning_rate": 1.0498907899526756e-06, + "loss": 0.3551, + "step": 35763 + }, + { + "epoch": 6.5, + "learning_rate": 1.0498361849290135e-06, + "loss": 0.008, + "step": 35764 + }, + { + "epoch": 6.5, + "learning_rate": 1.0497815799053515e-06, + "loss": 0.3657, + "step": 35765 + }, + { + "epoch": 6.5, + "learning_rate": 1.0497269748816892e-06, + "loss": 2.562, + "step": 35766 + }, + { + "epoch": 6.5, + "learning_rate": 1.049672369858027e-06, + "loss": 1.5877, + "step": 35767 + }, + { + "epoch": 6.5, + "learning_rate": 1.0496177648343647e-06, + "loss": 0.0044, + "step": 35768 + }, + { + "epoch": 6.5, + "learning_rate": 1.0495631598107026e-06, + "loss": 0.0486, + "step": 35769 + }, + { + "epoch": 6.5, + "learning_rate": 1.0495085547870405e-06, + "loss": 0.9379, + "step": 35770 + }, + { + "epoch": 6.51, + "learning_rate": 1.0494539497633783e-06, + "loss": 0.3411, + "step": 35771 + }, + { + "epoch": 6.51, + "learning_rate": 1.0493993447397162e-06, + "loss": 0.5606, + "step": 35772 + }, + { + "epoch": 6.51, + "learning_rate": 1.049344739716054e-06, + "loss": 2.5123, + "step": 35773 + }, + { + "epoch": 6.51, + "learning_rate": 1.0492901346923917e-06, + "loss": 0.0098, + "step": 35774 + }, + { + "epoch": 6.51, + "learning_rate": 1.0492355296687294e-06, + "loss": 0.5762, + "step": 35775 + }, + { + "epoch": 6.51, + "learning_rate": 1.0491809246450674e-06, + "loss": 1.1976, + "step": 35776 + }, + { + "epoch": 6.51, + "learning_rate": 1.0491263196214053e-06, + "loss": 0.0723, + "step": 35777 + }, + { + "epoch": 6.51, + "learning_rate": 1.049071714597743e-06, + "loss": 0.4296, + "step": 35778 + }, + { + "epoch": 6.51, + "learning_rate": 1.049017109574081e-06, + "loss": 0.1876, + "step": 35779 + }, + { + "epoch": 6.51, + "learning_rate": 1.0489625045504187e-06, + "loss": 0.5717, + "step": 35780 + }, + { + "epoch": 6.51, + "learning_rate": 1.0489078995267564e-06, + "loss": 0.3978, + "step": 35781 + }, + { + "epoch": 6.51, + "learning_rate": 1.0488532945030942e-06, + "loss": 0.6831, + "step": 35782 + }, + { + "epoch": 6.51, + "learning_rate": 1.0487986894794321e-06, + "loss": 0.5613, + "step": 35783 + }, + { + "epoch": 6.51, + "learning_rate": 1.04874408445577e-06, + "loss": 0.436, + "step": 35784 + }, + { + "epoch": 6.51, + "learning_rate": 1.0486894794321078e-06, + "loss": 1.1438, + "step": 35785 + }, + { + "epoch": 6.51, + "learning_rate": 1.0486348744084455e-06, + "loss": 0.4903, + "step": 35786 + }, + { + "epoch": 6.51, + "learning_rate": 1.0485802693847835e-06, + "loss": 1.1059, + "step": 35787 + }, + { + "epoch": 6.51, + "learning_rate": 1.0485256643611212e-06, + "loss": 1.4454, + "step": 35788 + }, + { + "epoch": 6.51, + "learning_rate": 1.048471059337459e-06, + "loss": 0.84, + "step": 35789 + }, + { + "epoch": 6.51, + "learning_rate": 1.0484164543137969e-06, + "loss": 0.1575, + "step": 35790 + }, + { + "epoch": 6.51, + "learning_rate": 1.0483618492901348e-06, + "loss": 0.1543, + "step": 35791 + }, + { + "epoch": 6.51, + "learning_rate": 1.0483072442664725e-06, + "loss": 0.1257, + "step": 35792 + }, + { + "epoch": 6.51, + "learning_rate": 1.0482526392428103e-06, + "loss": 0.3388, + "step": 35793 + }, + { + "epoch": 6.51, + "learning_rate": 1.0481980342191482e-06, + "loss": 1.6831, + "step": 35794 + }, + { + "epoch": 6.51, + "learning_rate": 1.048143429195486e-06, + "loss": 0.0664, + "step": 35795 + }, + { + "epoch": 6.51, + "learning_rate": 1.0480888241718237e-06, + "loss": 1.0428, + "step": 35796 + }, + { + "epoch": 6.51, + "learning_rate": 1.0480342191481616e-06, + "loss": 0.0497, + "step": 35797 + }, + { + "epoch": 6.51, + "learning_rate": 1.0479796141244996e-06, + "loss": 0.1652, + "step": 35798 + }, + { + "epoch": 6.51, + "learning_rate": 1.0479250091008373e-06, + "loss": 0.6591, + "step": 35799 + }, + { + "epoch": 6.51, + "learning_rate": 1.047870404077175e-06, + "loss": 0.0113, + "step": 35800 + }, + { + "epoch": 6.51, + "learning_rate": 1.047815799053513e-06, + "loss": 0.3293, + "step": 35801 + }, + { + "epoch": 6.51, + "learning_rate": 1.0477611940298507e-06, + "loss": 0.4647, + "step": 35802 + }, + { + "epoch": 6.51, + "learning_rate": 1.0477065890061884e-06, + "loss": 0.6065, + "step": 35803 + }, + { + "epoch": 6.51, + "learning_rate": 1.0476519839825264e-06, + "loss": 0.2562, + "step": 35804 + }, + { + "epoch": 6.51, + "learning_rate": 1.0475973789588643e-06, + "loss": 0.7222, + "step": 35805 + }, + { + "epoch": 6.51, + "learning_rate": 1.047542773935202e-06, + "loss": 0.01, + "step": 35806 + }, + { + "epoch": 6.51, + "learning_rate": 1.0474881689115398e-06, + "loss": 0.1152, + "step": 35807 + }, + { + "epoch": 6.51, + "learning_rate": 1.0474335638878777e-06, + "loss": 0.0669, + "step": 35808 + }, + { + "epoch": 6.51, + "learning_rate": 1.0473789588642155e-06, + "loss": 0.0192, + "step": 35809 + }, + { + "epoch": 6.51, + "learning_rate": 1.0473243538405534e-06, + "loss": 0.1619, + "step": 35810 + }, + { + "epoch": 6.51, + "learning_rate": 1.0472697488168914e-06, + "loss": 1.4104, + "step": 35811 + }, + { + "epoch": 6.51, + "learning_rate": 1.047215143793229e-06, + "loss": 0.8863, + "step": 35812 + }, + { + "epoch": 6.51, + "learning_rate": 1.0471605387695668e-06, + "loss": 0.0196, + "step": 35813 + }, + { + "epoch": 6.51, + "learning_rate": 1.0471059337459045e-06, + "loss": 0.7156, + "step": 35814 + }, + { + "epoch": 6.51, + "learning_rate": 1.0470513287222425e-06, + "loss": 0.741, + "step": 35815 + }, + { + "epoch": 6.51, + "learning_rate": 1.0469967236985802e-06, + "loss": 0.8195, + "step": 35816 + }, + { + "epoch": 6.51, + "learning_rate": 1.0469421186749182e-06, + "loss": 0.3522, + "step": 35817 + }, + { + "epoch": 6.51, + "learning_rate": 1.0468875136512561e-06, + "loss": 0.0225, + "step": 35818 + }, + { + "epoch": 6.51, + "learning_rate": 1.0468329086275938e-06, + "loss": 0.4313, + "step": 35819 + }, + { + "epoch": 6.51, + "learning_rate": 1.0467783036039316e-06, + "loss": 0.3827, + "step": 35820 + }, + { + "epoch": 6.51, + "learning_rate": 1.0467236985802693e-06, + "loss": 0.6059, + "step": 35821 + }, + { + "epoch": 6.51, + "learning_rate": 1.0466690935566072e-06, + "loss": 1.0904, + "step": 35822 + }, + { + "epoch": 6.51, + "learning_rate": 1.046614488532945e-06, + "loss": 0.581, + "step": 35823 + }, + { + "epoch": 6.51, + "learning_rate": 1.046559883509283e-06, + "loss": 0.4325, + "step": 35824 + }, + { + "epoch": 6.51, + "learning_rate": 1.0465052784856209e-06, + "loss": 1.5879, + "step": 35825 + }, + { + "epoch": 6.52, + "learning_rate": 1.0464506734619586e-06, + "loss": 1.2186, + "step": 35826 + }, + { + "epoch": 6.52, + "learning_rate": 1.0463960684382963e-06, + "loss": 1.0301, + "step": 35827 + }, + { + "epoch": 6.52, + "learning_rate": 1.046341463414634e-06, + "loss": 1.2631, + "step": 35828 + }, + { + "epoch": 6.52, + "learning_rate": 1.046286858390972e-06, + "loss": 1.629, + "step": 35829 + }, + { + "epoch": 6.52, + "learning_rate": 1.0462322533673097e-06, + "loss": 0.8031, + "step": 35830 + }, + { + "epoch": 6.52, + "learning_rate": 1.0461776483436477e-06, + "loss": 0.0846, + "step": 35831 + }, + { + "epoch": 6.52, + "learning_rate": 1.0461230433199856e-06, + "loss": 1.6876, + "step": 35832 + }, + { + "epoch": 6.52, + "learning_rate": 1.0460684382963234e-06, + "loss": 0.2273, + "step": 35833 + }, + { + "epoch": 6.52, + "learning_rate": 1.046013833272661e-06, + "loss": 0.0252, + "step": 35834 + }, + { + "epoch": 6.52, + "learning_rate": 1.0459592282489988e-06, + "loss": 1.3357, + "step": 35835 + }, + { + "epoch": 6.52, + "learning_rate": 1.0459046232253368e-06, + "loss": 0.0792, + "step": 35836 + }, + { + "epoch": 6.52, + "learning_rate": 1.0458500182016745e-06, + "loss": 0.0407, + "step": 35837 + }, + { + "epoch": 6.52, + "learning_rate": 1.0457954131780124e-06, + "loss": 0.4435, + "step": 35838 + }, + { + "epoch": 6.52, + "learning_rate": 1.0457408081543504e-06, + "loss": 0.0095, + "step": 35839 + }, + { + "epoch": 6.52, + "learning_rate": 1.0456862031306881e-06, + "loss": 0.0155, + "step": 35840 + }, + { + "epoch": 6.52, + "learning_rate": 1.0456315981070258e-06, + "loss": 0.1767, + "step": 35841 + }, + { + "epoch": 6.52, + "learning_rate": 1.0455769930833636e-06, + "loss": 1.1601, + "step": 35842 + }, + { + "epoch": 6.52, + "learning_rate": 1.0455223880597015e-06, + "loss": 0.247, + "step": 35843 + }, + { + "epoch": 6.52, + "learning_rate": 1.0454677830360392e-06, + "loss": 0.0165, + "step": 35844 + }, + { + "epoch": 6.52, + "learning_rate": 1.0454131780123772e-06, + "loss": 1.0365, + "step": 35845 + }, + { + "epoch": 6.52, + "learning_rate": 1.0453585729887151e-06, + "loss": 0.7878, + "step": 35846 + }, + { + "epoch": 6.52, + "learning_rate": 1.0453039679650529e-06, + "loss": 0.6586, + "step": 35847 + }, + { + "epoch": 6.52, + "learning_rate": 1.0452493629413906e-06, + "loss": 0.1565, + "step": 35848 + }, + { + "epoch": 6.52, + "learning_rate": 1.0451947579177283e-06, + "loss": 0.2339, + "step": 35849 + }, + { + "epoch": 6.52, + "learning_rate": 1.0451401528940663e-06, + "loss": 0.4992, + "step": 35850 + }, + { + "epoch": 6.52, + "learning_rate": 1.0450855478704042e-06, + "loss": 0.2926, + "step": 35851 + }, + { + "epoch": 6.52, + "learning_rate": 1.045030942846742e-06, + "loss": 0.1549, + "step": 35852 + }, + { + "epoch": 6.52, + "learning_rate": 1.0449763378230799e-06, + "loss": 0.0648, + "step": 35853 + }, + { + "epoch": 6.52, + "learning_rate": 1.0449217327994176e-06, + "loss": 1.7046, + "step": 35854 + }, + { + "epoch": 6.52, + "learning_rate": 1.0448671277757554e-06, + "loss": 0.0608, + "step": 35855 + }, + { + "epoch": 6.52, + "learning_rate": 1.044812522752093e-06, + "loss": 1.5774, + "step": 35856 + }, + { + "epoch": 6.52, + "learning_rate": 1.044757917728431e-06, + "loss": 0.9856, + "step": 35857 + }, + { + "epoch": 6.52, + "learning_rate": 1.044703312704769e-06, + "loss": 1.9341, + "step": 35858 + }, + { + "epoch": 6.52, + "learning_rate": 1.0446487076811067e-06, + "loss": 0.455, + "step": 35859 + }, + { + "epoch": 6.52, + "learning_rate": 1.0445941026574446e-06, + "loss": 2.1297, + "step": 35860 + }, + { + "epoch": 6.52, + "learning_rate": 1.0445394976337824e-06, + "loss": 0.0096, + "step": 35861 + }, + { + "epoch": 6.52, + "learning_rate": 1.0444848926101201e-06, + "loss": 2.2575, + "step": 35862 + }, + { + "epoch": 6.52, + "learning_rate": 1.0444302875864578e-06, + "loss": 0.0663, + "step": 35863 + }, + { + "epoch": 6.52, + "learning_rate": 1.0443756825627958e-06, + "loss": 0.6405, + "step": 35864 + }, + { + "epoch": 6.52, + "learning_rate": 1.0443210775391337e-06, + "loss": 0.1334, + "step": 35865 + }, + { + "epoch": 6.52, + "learning_rate": 1.0442664725154715e-06, + "loss": 0.0395, + "step": 35866 + }, + { + "epoch": 6.52, + "learning_rate": 1.0442118674918092e-06, + "loss": 0.0123, + "step": 35867 + }, + { + "epoch": 6.52, + "learning_rate": 1.0441572624681471e-06, + "loss": 0.4834, + "step": 35868 + }, + { + "epoch": 6.52, + "learning_rate": 1.0441026574444849e-06, + "loss": 0.0075, + "step": 35869 + }, + { + "epoch": 6.52, + "learning_rate": 1.0440480524208226e-06, + "loss": 1.5476, + "step": 35870 + }, + { + "epoch": 6.52, + "learning_rate": 1.0439934473971605e-06, + "loss": 0.0091, + "step": 35871 + }, + { + "epoch": 6.52, + "learning_rate": 1.0439388423734985e-06, + "loss": 0.011, + "step": 35872 + }, + { + "epoch": 6.52, + "learning_rate": 1.0438842373498362e-06, + "loss": 0.5895, + "step": 35873 + }, + { + "epoch": 6.52, + "learning_rate": 1.043829632326174e-06, + "loss": 0.126, + "step": 35874 + }, + { + "epoch": 6.52, + "learning_rate": 1.043775027302512e-06, + "loss": 0.57, + "step": 35875 + }, + { + "epoch": 6.52, + "learning_rate": 1.0437204222788496e-06, + "loss": 0.24, + "step": 35876 + }, + { + "epoch": 6.52, + "learning_rate": 1.0436658172551874e-06, + "loss": 0.383, + "step": 35877 + }, + { + "epoch": 6.52, + "learning_rate": 1.0436112122315253e-06, + "loss": 1.8617, + "step": 35878 + }, + { + "epoch": 6.52, + "learning_rate": 1.0435566072078632e-06, + "loss": 2.0228, + "step": 35879 + }, + { + "epoch": 6.52, + "learning_rate": 1.043502002184201e-06, + "loss": 0.5595, + "step": 35880 + }, + { + "epoch": 6.53, + "learning_rate": 1.0434473971605387e-06, + "loss": 0.6653, + "step": 35881 + }, + { + "epoch": 6.53, + "learning_rate": 1.0433927921368767e-06, + "loss": 0.8339, + "step": 35882 + }, + { + "epoch": 6.53, + "learning_rate": 1.0433381871132144e-06, + "loss": 2.4215, + "step": 35883 + }, + { + "epoch": 6.53, + "learning_rate": 1.0432835820895521e-06, + "loss": 0.4438, + "step": 35884 + }, + { + "epoch": 6.53, + "learning_rate": 1.04322897706589e-06, + "loss": 0.2543, + "step": 35885 + }, + { + "epoch": 6.53, + "learning_rate": 1.043174372042228e-06, + "loss": 0.0382, + "step": 35886 + }, + { + "epoch": 6.53, + "learning_rate": 1.0431197670185657e-06, + "loss": 0.0256, + "step": 35887 + }, + { + "epoch": 6.53, + "learning_rate": 1.0430651619949035e-06, + "loss": 0.1501, + "step": 35888 + }, + { + "epoch": 6.53, + "learning_rate": 1.0430105569712414e-06, + "loss": 0.0672, + "step": 35889 + }, + { + "epoch": 6.53, + "learning_rate": 1.0429559519475791e-06, + "loss": 1.1609, + "step": 35890 + }, + { + "epoch": 6.53, + "learning_rate": 1.0429013469239169e-06, + "loss": 0.3416, + "step": 35891 + }, + { + "epoch": 6.53, + "learning_rate": 1.0428467419002548e-06, + "loss": 0.1498, + "step": 35892 + }, + { + "epoch": 6.53, + "learning_rate": 1.0427921368765928e-06, + "loss": 0.0247, + "step": 35893 + }, + { + "epoch": 6.53, + "learning_rate": 1.0427375318529305e-06, + "loss": 0.4059, + "step": 35894 + }, + { + "epoch": 6.53, + "learning_rate": 1.0426829268292682e-06, + "loss": 0.1601, + "step": 35895 + }, + { + "epoch": 6.53, + "learning_rate": 1.0426283218056062e-06, + "loss": 0.1016, + "step": 35896 + }, + { + "epoch": 6.53, + "learning_rate": 1.042573716781944e-06, + "loss": 1.0954, + "step": 35897 + }, + { + "epoch": 6.53, + "learning_rate": 1.0425191117582818e-06, + "loss": 0.0083, + "step": 35898 + }, + { + "epoch": 6.53, + "learning_rate": 1.0424645067346198e-06, + "loss": 0.6412, + "step": 35899 + }, + { + "epoch": 6.53, + "learning_rate": 1.0424099017109575e-06, + "loss": 0.0439, + "step": 35900 + }, + { + "epoch": 6.53, + "learning_rate": 1.0423552966872952e-06, + "loss": 0.608, + "step": 35901 + }, + { + "epoch": 6.53, + "learning_rate": 1.042300691663633e-06, + "loss": 0.7306, + "step": 35902 + }, + { + "epoch": 6.53, + "learning_rate": 1.042246086639971e-06, + "loss": 0.5544, + "step": 35903 + }, + { + "epoch": 6.53, + "learning_rate": 1.0421914816163087e-06, + "loss": 0.1231, + "step": 35904 + }, + { + "epoch": 6.53, + "learning_rate": 1.0421368765926466e-06, + "loss": 0.1638, + "step": 35905 + }, + { + "epoch": 6.53, + "learning_rate": 1.0420822715689845e-06, + "loss": 0.326, + "step": 35906 + }, + { + "epoch": 6.53, + "learning_rate": 1.0420276665453223e-06, + "loss": 1.7121, + "step": 35907 + }, + { + "epoch": 6.53, + "learning_rate": 1.04197306152166e-06, + "loss": 4.7581, + "step": 35908 + }, + { + "epoch": 6.53, + "learning_rate": 1.0419184564979977e-06, + "loss": 0.9217, + "step": 35909 + }, + { + "epoch": 6.53, + "learning_rate": 1.0418638514743357e-06, + "loss": 0.0903, + "step": 35910 + }, + { + "epoch": 6.53, + "learning_rate": 1.0418092464506734e-06, + "loss": 0.0075, + "step": 35911 + }, + { + "epoch": 6.53, + "learning_rate": 1.0417546414270114e-06, + "loss": 0.3945, + "step": 35912 + }, + { + "epoch": 6.53, + "learning_rate": 1.0417000364033493e-06, + "loss": 0.0102, + "step": 35913 + }, + { + "epoch": 6.53, + "learning_rate": 1.041645431379687e-06, + "loss": 0.0035, + "step": 35914 + }, + { + "epoch": 6.53, + "learning_rate": 1.0415908263560248e-06, + "loss": 0.0508, + "step": 35915 + }, + { + "epoch": 6.53, + "learning_rate": 1.0415362213323625e-06, + "loss": 0.2736, + "step": 35916 + }, + { + "epoch": 6.53, + "learning_rate": 1.0414816163087004e-06, + "loss": 1.1625, + "step": 35917 + }, + { + "epoch": 6.53, + "learning_rate": 1.0414270112850382e-06, + "loss": 1.5909, + "step": 35918 + }, + { + "epoch": 6.53, + "learning_rate": 1.0413724062613761e-06, + "loss": 0.1592, + "step": 35919 + }, + { + "epoch": 6.53, + "learning_rate": 1.041317801237714e-06, + "loss": 1.7744, + "step": 35920 + }, + { + "epoch": 6.53, + "learning_rate": 1.0412631962140518e-06, + "loss": 1.4762, + "step": 35921 + }, + { + "epoch": 6.53, + "learning_rate": 1.0412085911903895e-06, + "loss": 0.0254, + "step": 35922 + }, + { + "epoch": 6.53, + "learning_rate": 1.0411539861667272e-06, + "loss": 0.9719, + "step": 35923 + }, + { + "epoch": 6.53, + "learning_rate": 1.0410993811430652e-06, + "loss": 1.1339, + "step": 35924 + }, + { + "epoch": 6.53, + "learning_rate": 1.041044776119403e-06, + "loss": 0.4271, + "step": 35925 + }, + { + "epoch": 6.53, + "learning_rate": 1.0409901710957409e-06, + "loss": 0.1959, + "step": 35926 + }, + { + "epoch": 6.53, + "learning_rate": 1.0409355660720788e-06, + "loss": 0.7588, + "step": 35927 + }, + { + "epoch": 6.53, + "learning_rate": 1.0408809610484165e-06, + "loss": 0.0119, + "step": 35928 + }, + { + "epoch": 6.53, + "learning_rate": 1.0408263560247543e-06, + "loss": 0.9126, + "step": 35929 + }, + { + "epoch": 6.53, + "learning_rate": 1.040771751001092e-06, + "loss": 0.0377, + "step": 35930 + }, + { + "epoch": 6.53, + "learning_rate": 1.04071714597743e-06, + "loss": 0.0117, + "step": 35931 + }, + { + "epoch": 6.53, + "learning_rate": 1.0406625409537677e-06, + "loss": 0.4572, + "step": 35932 + }, + { + "epoch": 6.53, + "learning_rate": 1.0406079359301056e-06, + "loss": 0.0144, + "step": 35933 + }, + { + "epoch": 6.53, + "learning_rate": 1.0405533309064436e-06, + "loss": 0.0211, + "step": 35934 + }, + { + "epoch": 6.53, + "learning_rate": 1.0404987258827813e-06, + "loss": 1.3505, + "step": 35935 + }, + { + "epoch": 6.54, + "learning_rate": 1.040444120859119e-06, + "loss": 1.1753, + "step": 35936 + }, + { + "epoch": 6.54, + "learning_rate": 1.0403895158354568e-06, + "loss": 0.005, + "step": 35937 + }, + { + "epoch": 6.54, + "learning_rate": 1.0403349108117947e-06, + "loss": 0.7374, + "step": 35938 + }, + { + "epoch": 6.54, + "learning_rate": 1.0402803057881326e-06, + "loss": 0.2463, + "step": 35939 + }, + { + "epoch": 6.54, + "learning_rate": 1.0402257007644704e-06, + "loss": 0.3504, + "step": 35940 + }, + { + "epoch": 6.54, + "learning_rate": 1.0401710957408081e-06, + "loss": 0.1274, + "step": 35941 + }, + { + "epoch": 6.54, + "learning_rate": 1.040116490717146e-06, + "loss": 0.2601, + "step": 35942 + }, + { + "epoch": 6.54, + "learning_rate": 1.0400618856934838e-06, + "loss": 0.577, + "step": 35943 + }, + { + "epoch": 6.54, + "learning_rate": 1.0400072806698215e-06, + "loss": 0.517, + "step": 35944 + }, + { + "epoch": 6.54, + "learning_rate": 1.0399526756461595e-06, + "loss": 0.3379, + "step": 35945 + }, + { + "epoch": 6.54, + "learning_rate": 1.0398980706224974e-06, + "loss": 1.1894, + "step": 35946 + }, + { + "epoch": 6.54, + "learning_rate": 1.0398434655988351e-06, + "loss": 1.2016, + "step": 35947 + }, + { + "epoch": 6.54, + "learning_rate": 1.0397888605751729e-06, + "loss": 0.1169, + "step": 35948 + }, + { + "epoch": 6.54, + "learning_rate": 1.0397342555515108e-06, + "loss": 4.1284, + "step": 35949 + }, + { + "epoch": 6.54, + "learning_rate": 1.0396796505278485e-06, + "loss": 0.3551, + "step": 35950 + }, + { + "epoch": 6.54, + "learning_rate": 1.0396250455041863e-06, + "loss": 0.3837, + "step": 35951 + }, + { + "epoch": 6.54, + "learning_rate": 1.0395704404805242e-06, + "loss": 0.2809, + "step": 35952 + }, + { + "epoch": 6.54, + "learning_rate": 1.0395158354568622e-06, + "loss": 1.0939, + "step": 35953 + }, + { + "epoch": 6.54, + "learning_rate": 1.0394612304331999e-06, + "loss": 0.3019, + "step": 35954 + }, + { + "epoch": 6.54, + "learning_rate": 1.0394066254095376e-06, + "loss": 0.005, + "step": 35955 + }, + { + "epoch": 6.54, + "learning_rate": 1.0393520203858756e-06, + "loss": 0.1985, + "step": 35956 + }, + { + "epoch": 6.54, + "learning_rate": 1.0392974153622133e-06, + "loss": 1.5521, + "step": 35957 + }, + { + "epoch": 6.54, + "learning_rate": 1.039242810338551e-06, + "loss": 0.5309, + "step": 35958 + }, + { + "epoch": 6.54, + "learning_rate": 1.039188205314889e-06, + "loss": 0.927, + "step": 35959 + }, + { + "epoch": 6.54, + "learning_rate": 1.039133600291227e-06, + "loss": 0.4873, + "step": 35960 + }, + { + "epoch": 6.54, + "learning_rate": 1.0390789952675647e-06, + "loss": 2.3415, + "step": 35961 + }, + { + "epoch": 6.54, + "learning_rate": 1.0390243902439024e-06, + "loss": 0.0959, + "step": 35962 + }, + { + "epoch": 6.54, + "learning_rate": 1.0389697852202403e-06, + "loss": 2.4356, + "step": 35963 + }, + { + "epoch": 6.54, + "learning_rate": 1.038915180196578e-06, + "loss": 0.0078, + "step": 35964 + }, + { + "epoch": 6.54, + "learning_rate": 1.0388605751729158e-06, + "loss": 0.7055, + "step": 35965 + }, + { + "epoch": 6.54, + "learning_rate": 1.0388059701492537e-06, + "loss": 0.0133, + "step": 35966 + }, + { + "epoch": 6.54, + "learning_rate": 1.0387513651255917e-06, + "loss": 0.0176, + "step": 35967 + }, + { + "epoch": 6.54, + "learning_rate": 1.0386967601019294e-06, + "loss": 0.4566, + "step": 35968 + }, + { + "epoch": 6.54, + "learning_rate": 1.0386421550782671e-06, + "loss": 0.3854, + "step": 35969 + }, + { + "epoch": 6.54, + "learning_rate": 1.038587550054605e-06, + "loss": 0.8639, + "step": 35970 + }, + { + "epoch": 6.54, + "learning_rate": 1.0385329450309428e-06, + "loss": 0.1045, + "step": 35971 + }, + { + "epoch": 6.54, + "learning_rate": 1.0384783400072805e-06, + "loss": 0.5247, + "step": 35972 + }, + { + "epoch": 6.54, + "learning_rate": 1.0384237349836185e-06, + "loss": 0.1, + "step": 35973 + }, + { + "epoch": 6.54, + "learning_rate": 1.0383691299599564e-06, + "loss": 0.1947, + "step": 35974 + }, + { + "epoch": 6.54, + "learning_rate": 1.0383145249362942e-06, + "loss": 0.0074, + "step": 35975 + }, + { + "epoch": 6.54, + "learning_rate": 1.038259919912632e-06, + "loss": 0.2336, + "step": 35976 + }, + { + "epoch": 6.54, + "learning_rate": 1.0382053148889698e-06, + "loss": 0.5224, + "step": 35977 + }, + { + "epoch": 6.54, + "learning_rate": 1.0381507098653076e-06, + "loss": 0.1026, + "step": 35978 + }, + { + "epoch": 6.54, + "learning_rate": 1.0380961048416455e-06, + "loss": 2.0177, + "step": 35979 + }, + { + "epoch": 6.54, + "learning_rate": 1.0380414998179835e-06, + "loss": 0.2817, + "step": 35980 + }, + { + "epoch": 6.54, + "learning_rate": 1.0379868947943212e-06, + "loss": 0.018, + "step": 35981 + }, + { + "epoch": 6.54, + "learning_rate": 1.037932289770659e-06, + "loss": 0.259, + "step": 35982 + }, + { + "epoch": 6.54, + "learning_rate": 1.0378776847469967e-06, + "loss": 0.0069, + "step": 35983 + }, + { + "epoch": 6.54, + "learning_rate": 1.0378230797233346e-06, + "loss": 0.2349, + "step": 35984 + }, + { + "epoch": 6.54, + "learning_rate": 1.0377684746996723e-06, + "loss": 3.4528, + "step": 35985 + }, + { + "epoch": 6.54, + "learning_rate": 1.0377138696760103e-06, + "loss": 3.0, + "step": 35986 + }, + { + "epoch": 6.54, + "learning_rate": 1.0376592646523482e-06, + "loss": 2.1496, + "step": 35987 + }, + { + "epoch": 6.54, + "learning_rate": 1.037604659628686e-06, + "loss": 2.2214, + "step": 35988 + }, + { + "epoch": 6.54, + "learning_rate": 1.0375500546050237e-06, + "loss": 0.0097, + "step": 35989 + }, + { + "epoch": 6.54, + "learning_rate": 1.0374954495813614e-06, + "loss": 0.2966, + "step": 35990 + }, + { + "epoch": 6.55, + "learning_rate": 1.0374408445576994e-06, + "loss": 0.0062, + "step": 35991 + }, + { + "epoch": 6.55, + "learning_rate": 1.037386239534037e-06, + "loss": 0.0911, + "step": 35992 + }, + { + "epoch": 6.55, + "learning_rate": 1.037331634510375e-06, + "loss": 0.0109, + "step": 35993 + }, + { + "epoch": 6.55, + "learning_rate": 1.037277029486713e-06, + "loss": 0.3305, + "step": 35994 + }, + { + "epoch": 6.55, + "learning_rate": 1.0372224244630507e-06, + "loss": 1.0379, + "step": 35995 + }, + { + "epoch": 6.55, + "learning_rate": 1.0371678194393884e-06, + "loss": 0.5676, + "step": 35996 + }, + { + "epoch": 6.55, + "learning_rate": 1.0371132144157262e-06, + "loss": 0.082, + "step": 35997 + }, + { + "epoch": 6.55, + "learning_rate": 1.0370586093920641e-06, + "loss": 0.06, + "step": 35998 + }, + { + "epoch": 6.55, + "learning_rate": 1.0370040043684018e-06, + "loss": 0.0914, + "step": 35999 + }, + { + "epoch": 6.55, + "learning_rate": 1.0369493993447398e-06, + "loss": 0.0187, + "step": 36000 + }, + { + "epoch": 6.55, + "learning_rate": 1.0368947943210777e-06, + "loss": 0.2212, + "step": 36001 + }, + { + "epoch": 6.55, + "learning_rate": 1.0368401892974155e-06, + "loss": 0.0106, + "step": 36002 + }, + { + "epoch": 6.55, + "learning_rate": 1.0367855842737532e-06, + "loss": 0.071, + "step": 36003 + }, + { + "epoch": 6.55, + "learning_rate": 1.036730979250091e-06, + "loss": 0.0773, + "step": 36004 + }, + { + "epoch": 6.55, + "learning_rate": 1.0366763742264289e-06, + "loss": 0.0112, + "step": 36005 + }, + { + "epoch": 6.55, + "learning_rate": 1.0366217692027666e-06, + "loss": 0.1426, + "step": 36006 + }, + { + "epoch": 6.55, + "learning_rate": 1.0365671641791045e-06, + "loss": 0.3743, + "step": 36007 + }, + { + "epoch": 6.55, + "learning_rate": 1.0365125591554425e-06, + "loss": 0.6107, + "step": 36008 + }, + { + "epoch": 6.55, + "learning_rate": 1.0364579541317802e-06, + "loss": 1.0575, + "step": 36009 + }, + { + "epoch": 6.55, + "learning_rate": 1.036403349108118e-06, + "loss": 0.0105, + "step": 36010 + }, + { + "epoch": 6.55, + "learning_rate": 1.0363487440844557e-06, + "loss": 1.2338, + "step": 36011 + }, + { + "epoch": 6.55, + "learning_rate": 1.0362941390607936e-06, + "loss": 1.2403, + "step": 36012 + }, + { + "epoch": 6.55, + "learning_rate": 1.0362395340371314e-06, + "loss": 0.5366, + "step": 36013 + }, + { + "epoch": 6.55, + "learning_rate": 1.0361849290134693e-06, + "loss": 0.0051, + "step": 36014 + }, + { + "epoch": 6.55, + "learning_rate": 1.0361303239898072e-06, + "loss": 0.0484, + "step": 36015 + }, + { + "epoch": 6.55, + "learning_rate": 1.036075718966145e-06, + "loss": 0.008, + "step": 36016 + }, + { + "epoch": 6.55, + "learning_rate": 1.0360211139424827e-06, + "loss": 0.4831, + "step": 36017 + }, + { + "epoch": 6.55, + "learning_rate": 1.0359665089188204e-06, + "loss": 0.6056, + "step": 36018 + }, + { + "epoch": 6.55, + "learning_rate": 1.0359119038951584e-06, + "loss": 0.4415, + "step": 36019 + }, + { + "epoch": 6.55, + "learning_rate": 1.0358572988714963e-06, + "loss": 0.1788, + "step": 36020 + }, + { + "epoch": 6.55, + "learning_rate": 1.035802693847834e-06, + "loss": 1.7984, + "step": 36021 + }, + { + "epoch": 6.55, + "learning_rate": 1.0357480888241718e-06, + "loss": 0.0506, + "step": 36022 + }, + { + "epoch": 6.55, + "learning_rate": 1.0356934838005097e-06, + "loss": 0.5915, + "step": 36023 + }, + { + "epoch": 6.55, + "learning_rate": 1.0356388787768475e-06, + "loss": 0.018, + "step": 36024 + }, + { + "epoch": 6.55, + "learning_rate": 1.0355842737531852e-06, + "loss": 0.0161, + "step": 36025 + }, + { + "epoch": 6.55, + "learning_rate": 1.0355296687295231e-06, + "loss": 0.0804, + "step": 36026 + }, + { + "epoch": 6.55, + "learning_rate": 1.035475063705861e-06, + "loss": 0.0839, + "step": 36027 + }, + { + "epoch": 6.55, + "learning_rate": 1.0354204586821988e-06, + "loss": 0.0146, + "step": 36028 + }, + { + "epoch": 6.55, + "learning_rate": 1.0353658536585365e-06, + "loss": 0.0058, + "step": 36029 + }, + { + "epoch": 6.55, + "learning_rate": 1.0353112486348745e-06, + "loss": 0.9905, + "step": 36030 + }, + { + "epoch": 6.55, + "learning_rate": 1.0352566436112122e-06, + "loss": 0.8951, + "step": 36031 + }, + { + "epoch": 6.55, + "learning_rate": 1.03520203858755e-06, + "loss": 0.7036, + "step": 36032 + }, + { + "epoch": 6.55, + "learning_rate": 1.0351474335638879e-06, + "loss": 2.6148, + "step": 36033 + }, + { + "epoch": 6.55, + "learning_rate": 1.0350928285402258e-06, + "loss": 1.5104, + "step": 36034 + }, + { + "epoch": 6.55, + "learning_rate": 1.0350382235165636e-06, + "loss": 0.0169, + "step": 36035 + }, + { + "epoch": 6.55, + "learning_rate": 1.0349836184929013e-06, + "loss": 0.4055, + "step": 36036 + }, + { + "epoch": 6.55, + "learning_rate": 1.0349290134692392e-06, + "loss": 0.0062, + "step": 36037 + }, + { + "epoch": 6.55, + "learning_rate": 1.034874408445577e-06, + "loss": 0.0657, + "step": 36038 + }, + { + "epoch": 6.55, + "learning_rate": 1.0348198034219147e-06, + "loss": 0.981, + "step": 36039 + }, + { + "epoch": 6.55, + "learning_rate": 1.0347651983982526e-06, + "loss": 2.2897, + "step": 36040 + }, + { + "epoch": 6.55, + "learning_rate": 1.0347105933745906e-06, + "loss": 2.6374, + "step": 36041 + }, + { + "epoch": 6.55, + "learning_rate": 1.0346559883509283e-06, + "loss": 2.1353, + "step": 36042 + }, + { + "epoch": 6.55, + "learning_rate": 1.034601383327266e-06, + "loss": 0.2283, + "step": 36043 + }, + { + "epoch": 6.55, + "learning_rate": 1.034546778303604e-06, + "loss": 1.3445, + "step": 36044 + }, + { + "epoch": 6.55, + "learning_rate": 1.0344921732799417e-06, + "loss": 0.0438, + "step": 36045 + }, + { + "epoch": 6.56, + "learning_rate": 1.0344375682562795e-06, + "loss": 1.4525, + "step": 36046 + }, + { + "epoch": 6.56, + "learning_rate": 1.0343829632326174e-06, + "loss": 0.3759, + "step": 36047 + }, + { + "epoch": 6.56, + "learning_rate": 1.0343283582089553e-06, + "loss": 0.2868, + "step": 36048 + }, + { + "epoch": 6.56, + "learning_rate": 1.034273753185293e-06, + "loss": 0.3389, + "step": 36049 + }, + { + "epoch": 6.56, + "learning_rate": 1.0342191481616308e-06, + "loss": 0.4831, + "step": 36050 + }, + { + "epoch": 6.56, + "learning_rate": 1.0341645431379688e-06, + "loss": 1.4904, + "step": 36051 + }, + { + "epoch": 6.56, + "learning_rate": 1.0341099381143065e-06, + "loss": 0.0137, + "step": 36052 + }, + { + "epoch": 6.56, + "learning_rate": 1.0340553330906442e-06, + "loss": 0.0067, + "step": 36053 + }, + { + "epoch": 6.56, + "learning_rate": 1.0340007280669822e-06, + "loss": 0.4159, + "step": 36054 + }, + { + "epoch": 6.56, + "learning_rate": 1.0339461230433201e-06, + "loss": 0.2825, + "step": 36055 + }, + { + "epoch": 6.56, + "learning_rate": 1.0338915180196578e-06, + "loss": 0.0027, + "step": 36056 + }, + { + "epoch": 6.56, + "learning_rate": 1.0338369129959956e-06, + "loss": 1.1757, + "step": 36057 + }, + { + "epoch": 6.56, + "learning_rate": 1.0337823079723335e-06, + "loss": 0.007, + "step": 36058 + }, + { + "epoch": 6.56, + "learning_rate": 1.0337277029486712e-06, + "loss": 0.1278, + "step": 36059 + }, + { + "epoch": 6.56, + "learning_rate": 1.033673097925009e-06, + "loss": 1.4172, + "step": 36060 + }, + { + "epoch": 6.56, + "learning_rate": 1.033618492901347e-06, + "loss": 0.13, + "step": 36061 + }, + { + "epoch": 6.56, + "learning_rate": 1.0335638878776849e-06, + "loss": 0.1287, + "step": 36062 + }, + { + "epoch": 6.56, + "learning_rate": 1.0335092828540226e-06, + "loss": 1.6063, + "step": 36063 + }, + { + "epoch": 6.56, + "learning_rate": 1.0334546778303603e-06, + "loss": 2.6811, + "step": 36064 + }, + { + "epoch": 6.56, + "learning_rate": 1.0334000728066983e-06, + "loss": 0.6384, + "step": 36065 + }, + { + "epoch": 6.56, + "learning_rate": 1.033345467783036e-06, + "loss": 0.3872, + "step": 36066 + }, + { + "epoch": 6.56, + "learning_rate": 1.033290862759374e-06, + "loss": 1.0094, + "step": 36067 + }, + { + "epoch": 6.56, + "learning_rate": 1.0332362577357119e-06, + "loss": 0.0243, + "step": 36068 + }, + { + "epoch": 6.56, + "learning_rate": 1.0331816527120496e-06, + "loss": 1.001, + "step": 36069 + }, + { + "epoch": 6.56, + "learning_rate": 1.0331270476883874e-06, + "loss": 0.9372, + "step": 36070 + }, + { + "epoch": 6.56, + "learning_rate": 1.033072442664725e-06, + "loss": 0.2543, + "step": 36071 + }, + { + "epoch": 6.56, + "learning_rate": 1.033017837641063e-06, + "loss": 1.3005, + "step": 36072 + }, + { + "epoch": 6.56, + "learning_rate": 1.0329632326174008e-06, + "loss": 0.466, + "step": 36073 + }, + { + "epoch": 6.56, + "learning_rate": 1.0329086275937387e-06, + "loss": 0.4048, + "step": 36074 + }, + { + "epoch": 6.56, + "learning_rate": 1.0328540225700766e-06, + "loss": 0.005, + "step": 36075 + }, + { + "epoch": 6.56, + "learning_rate": 1.0327994175464144e-06, + "loss": 3.4363, + "step": 36076 + }, + { + "epoch": 6.56, + "learning_rate": 1.0327448125227521e-06, + "loss": 0.4161, + "step": 36077 + }, + { + "epoch": 6.56, + "learning_rate": 1.0326902074990898e-06, + "loss": 0.5657, + "step": 36078 + }, + { + "epoch": 6.56, + "learning_rate": 1.0326356024754278e-06, + "loss": 0.0103, + "step": 36079 + }, + { + "epoch": 6.56, + "learning_rate": 1.0325809974517655e-06, + "loss": 0.0322, + "step": 36080 + }, + { + "epoch": 6.56, + "learning_rate": 1.0325263924281035e-06, + "loss": 0.8844, + "step": 36081 + }, + { + "epoch": 6.56, + "learning_rate": 1.0324717874044414e-06, + "loss": 1.1671, + "step": 36082 + }, + { + "epoch": 6.56, + "learning_rate": 1.0324171823807791e-06, + "loss": 0.4507, + "step": 36083 + }, + { + "epoch": 6.56, + "learning_rate": 1.0323625773571169e-06, + "loss": 0.5169, + "step": 36084 + }, + { + "epoch": 6.56, + "learning_rate": 1.0323079723334546e-06, + "loss": 0.4754, + "step": 36085 + }, + { + "epoch": 6.56, + "learning_rate": 1.0322533673097925e-06, + "loss": 0.044, + "step": 36086 + }, + { + "epoch": 6.56, + "learning_rate": 1.0321987622861303e-06, + "loss": 0.3871, + "step": 36087 + }, + { + "epoch": 6.56, + "learning_rate": 1.0321441572624682e-06, + "loss": 0.2658, + "step": 36088 + }, + { + "epoch": 6.56, + "learning_rate": 1.0320895522388062e-06, + "loss": 0.4266, + "step": 36089 + }, + { + "epoch": 6.56, + "learning_rate": 1.0320349472151439e-06, + "loss": 0.27, + "step": 36090 + }, + { + "epoch": 6.56, + "learning_rate": 1.0319803421914816e-06, + "loss": 0.0297, + "step": 36091 + }, + { + "epoch": 6.56, + "learning_rate": 1.0319257371678194e-06, + "loss": 0.9686, + "step": 36092 + }, + { + "epoch": 6.56, + "learning_rate": 1.0318711321441573e-06, + "loss": 0.0209, + "step": 36093 + }, + { + "epoch": 6.56, + "learning_rate": 1.031816527120495e-06, + "loss": 0.3052, + "step": 36094 + }, + { + "epoch": 6.56, + "learning_rate": 1.031761922096833e-06, + "loss": 0.2422, + "step": 36095 + }, + { + "epoch": 6.56, + "learning_rate": 1.0317073170731707e-06, + "loss": 0.0047, + "step": 36096 + }, + { + "epoch": 6.56, + "learning_rate": 1.0316527120495086e-06, + "loss": 2.3581, + "step": 36097 + }, + { + "epoch": 6.56, + "learning_rate": 1.0315981070258464e-06, + "loss": 0.0524, + "step": 36098 + }, + { + "epoch": 6.56, + "learning_rate": 1.0315435020021841e-06, + "loss": 0.0114, + "step": 36099 + }, + { + "epoch": 6.56, + "learning_rate": 1.031488896978522e-06, + "loss": 0.0298, + "step": 36100 + }, + { + "epoch": 6.57, + "learning_rate": 1.0314342919548598e-06, + "loss": 0.2059, + "step": 36101 + }, + { + "epoch": 6.57, + "learning_rate": 1.0313796869311977e-06, + "loss": 0.0865, + "step": 36102 + }, + { + "epoch": 6.57, + "learning_rate": 1.0313250819075355e-06, + "loss": 2.0575, + "step": 36103 + }, + { + "epoch": 6.57, + "learning_rate": 1.0312704768838734e-06, + "loss": 1.3782, + "step": 36104 + }, + { + "epoch": 6.57, + "learning_rate": 1.0312158718602111e-06, + "loss": 0.0152, + "step": 36105 + }, + { + "epoch": 6.57, + "learning_rate": 1.0311612668365489e-06, + "loss": 0.4636, + "step": 36106 + }, + { + "epoch": 6.57, + "learning_rate": 1.0311066618128868e-06, + "loss": 1.1777, + "step": 36107 + }, + { + "epoch": 6.57, + "learning_rate": 1.0310520567892248e-06, + "loss": 1.0824, + "step": 36108 + }, + { + "epoch": 6.57, + "learning_rate": 1.0309974517655625e-06, + "loss": 0.7596, + "step": 36109 + }, + { + "epoch": 6.57, + "learning_rate": 1.0309428467419002e-06, + "loss": 1.0991, + "step": 36110 + }, + { + "epoch": 6.57, + "learning_rate": 1.0308882417182382e-06, + "loss": 0.176, + "step": 36111 + }, + { + "epoch": 6.57, + "learning_rate": 1.0308336366945759e-06, + "loss": 0.0175, + "step": 36112 + }, + { + "epoch": 6.57, + "learning_rate": 1.0307790316709136e-06, + "loss": 0.0837, + "step": 36113 + }, + { + "epoch": 6.57, + "learning_rate": 1.0307244266472516e-06, + "loss": 0.0886, + "step": 36114 + }, + { + "epoch": 6.57, + "learning_rate": 1.0306698216235895e-06, + "loss": 0.3034, + "step": 36115 + }, + { + "epoch": 6.57, + "learning_rate": 1.0306152165999272e-06, + "loss": 0.0193, + "step": 36116 + }, + { + "epoch": 6.57, + "learning_rate": 1.030560611576265e-06, + "loss": 0.0578, + "step": 36117 + }, + { + "epoch": 6.57, + "learning_rate": 1.030506006552603e-06, + "loss": 0.1992, + "step": 36118 + }, + { + "epoch": 6.57, + "learning_rate": 1.0304514015289406e-06, + "loss": 0.4297, + "step": 36119 + }, + { + "epoch": 6.57, + "learning_rate": 1.0303967965052784e-06, + "loss": 0.1533, + "step": 36120 + }, + { + "epoch": 6.57, + "learning_rate": 1.0303421914816163e-06, + "loss": 1.2099, + "step": 36121 + }, + { + "epoch": 6.57, + "learning_rate": 1.0302875864579543e-06, + "loss": 0.2208, + "step": 36122 + }, + { + "epoch": 6.57, + "learning_rate": 1.030232981434292e-06, + "loss": 0.1099, + "step": 36123 + }, + { + "epoch": 6.57, + "learning_rate": 1.0301783764106297e-06, + "loss": 0.6093, + "step": 36124 + }, + { + "epoch": 6.57, + "learning_rate": 1.0301237713869677e-06, + "loss": 0.8572, + "step": 36125 + }, + { + "epoch": 6.57, + "learning_rate": 1.0300691663633054e-06, + "loss": 0.052, + "step": 36126 + }, + { + "epoch": 6.57, + "learning_rate": 1.0300145613396431e-06, + "loss": 1.1324, + "step": 36127 + }, + { + "epoch": 6.57, + "learning_rate": 1.029959956315981e-06, + "loss": 1.1461, + "step": 36128 + }, + { + "epoch": 6.57, + "learning_rate": 1.029905351292319e-06, + "loss": 0.0416, + "step": 36129 + }, + { + "epoch": 6.57, + "learning_rate": 1.0298507462686568e-06, + "loss": 0.0847, + "step": 36130 + }, + { + "epoch": 6.57, + "learning_rate": 1.0297961412449945e-06, + "loss": 0.5237, + "step": 36131 + }, + { + "epoch": 6.57, + "learning_rate": 1.0297415362213324e-06, + "loss": 1.6038, + "step": 36132 + }, + { + "epoch": 6.57, + "learning_rate": 1.0296869311976702e-06, + "loss": 1.1066, + "step": 36133 + }, + { + "epoch": 6.57, + "learning_rate": 1.0296323261740079e-06, + "loss": 0.0443, + "step": 36134 + }, + { + "epoch": 6.57, + "learning_rate": 1.0295777211503458e-06, + "loss": 0.8972, + "step": 36135 + }, + { + "epoch": 6.57, + "learning_rate": 1.0295231161266838e-06, + "loss": 0.0404, + "step": 36136 + }, + { + "epoch": 6.57, + "learning_rate": 1.0294685111030215e-06, + "loss": 0.735, + "step": 36137 + }, + { + "epoch": 6.57, + "learning_rate": 1.0294139060793592e-06, + "loss": 1.3012, + "step": 36138 + }, + { + "epoch": 6.57, + "learning_rate": 1.0293593010556972e-06, + "loss": 0.24, + "step": 36139 + }, + { + "epoch": 6.57, + "learning_rate": 1.029304696032035e-06, + "loss": 0.057, + "step": 36140 + }, + { + "epoch": 6.57, + "learning_rate": 1.0292500910083727e-06, + "loss": 0.5281, + "step": 36141 + }, + { + "epoch": 6.57, + "learning_rate": 1.0291954859847106e-06, + "loss": 1.0274, + "step": 36142 + }, + { + "epoch": 6.57, + "learning_rate": 1.0291408809610485e-06, + "loss": 0.8948, + "step": 36143 + }, + { + "epoch": 6.57, + "learning_rate": 1.0290862759373863e-06, + "loss": 0.3454, + "step": 36144 + }, + { + "epoch": 6.57, + "learning_rate": 1.029031670913724e-06, + "loss": 0.776, + "step": 36145 + }, + { + "epoch": 6.57, + "learning_rate": 1.028977065890062e-06, + "loss": 0.3654, + "step": 36146 + }, + { + "epoch": 6.57, + "learning_rate": 1.0289224608663997e-06, + "loss": 0.1353, + "step": 36147 + }, + { + "epoch": 6.57, + "learning_rate": 1.0288678558427376e-06, + "loss": 0.0418, + "step": 36148 + }, + { + "epoch": 6.57, + "learning_rate": 1.0288132508190756e-06, + "loss": 0.2257, + "step": 36149 + }, + { + "epoch": 6.57, + "learning_rate": 1.0287586457954133e-06, + "loss": 0.1781, + "step": 36150 + }, + { + "epoch": 6.57, + "learning_rate": 1.028704040771751e-06, + "loss": 0.1043, + "step": 36151 + }, + { + "epoch": 6.57, + "learning_rate": 1.0286494357480888e-06, + "loss": 0.0224, + "step": 36152 + }, + { + "epoch": 6.57, + "learning_rate": 1.0285948307244267e-06, + "loss": 0.0157, + "step": 36153 + }, + { + "epoch": 6.57, + "learning_rate": 1.0285402257007644e-06, + "loss": 0.0376, + "step": 36154 + }, + { + "epoch": 6.57, + "learning_rate": 1.0284856206771024e-06, + "loss": 0.426, + "step": 36155 + }, + { + "epoch": 6.58, + "learning_rate": 1.0284310156534403e-06, + "loss": 0.0055, + "step": 36156 + }, + { + "epoch": 6.58, + "learning_rate": 1.028376410629778e-06, + "loss": 2.4878, + "step": 36157 + }, + { + "epoch": 6.58, + "learning_rate": 1.0283218056061158e-06, + "loss": 1.0551, + "step": 36158 + }, + { + "epoch": 6.58, + "learning_rate": 1.0282672005824535e-06, + "loss": 2.288, + "step": 36159 + }, + { + "epoch": 6.58, + "learning_rate": 1.0282125955587915e-06, + "loss": 0.5424, + "step": 36160 + }, + { + "epoch": 6.58, + "learning_rate": 1.0281579905351292e-06, + "loss": 0.9032, + "step": 36161 + }, + { + "epoch": 6.58, + "learning_rate": 1.0281033855114671e-06, + "loss": 0.0346, + "step": 36162 + }, + { + "epoch": 6.58, + "learning_rate": 1.028048780487805e-06, + "loss": 0.087, + "step": 36163 + }, + { + "epoch": 6.58, + "learning_rate": 1.0279941754641428e-06, + "loss": 1.1774, + "step": 36164 + }, + { + "epoch": 6.58, + "learning_rate": 1.0279395704404805e-06, + "loss": 0.1051, + "step": 36165 + }, + { + "epoch": 6.58, + "learning_rate": 1.0278849654168183e-06, + "loss": 0.3138, + "step": 36166 + }, + { + "epoch": 6.58, + "learning_rate": 1.0278303603931562e-06, + "loss": 0.7229, + "step": 36167 + }, + { + "epoch": 6.58, + "learning_rate": 1.027775755369494e-06, + "loss": 0.8936, + "step": 36168 + }, + { + "epoch": 6.58, + "learning_rate": 1.0277211503458319e-06, + "loss": 0.4107, + "step": 36169 + }, + { + "epoch": 6.58, + "learning_rate": 1.0276665453221698e-06, + "loss": 0.1853, + "step": 36170 + }, + { + "epoch": 6.58, + "learning_rate": 1.0276119402985076e-06, + "loss": 0.6277, + "step": 36171 + }, + { + "epoch": 6.58, + "learning_rate": 1.0275573352748453e-06, + "loss": 0.081, + "step": 36172 + }, + { + "epoch": 6.58, + "learning_rate": 1.027502730251183e-06, + "loss": 0.298, + "step": 36173 + }, + { + "epoch": 6.58, + "learning_rate": 1.027448125227521e-06, + "loss": 0.5082, + "step": 36174 + }, + { + "epoch": 6.58, + "learning_rate": 1.0273935202038587e-06, + "loss": 0.9529, + "step": 36175 + }, + { + "epoch": 6.58, + "learning_rate": 1.0273389151801966e-06, + "loss": 0.0202, + "step": 36176 + }, + { + "epoch": 6.58, + "learning_rate": 1.0272843101565344e-06, + "loss": 0.0091, + "step": 36177 + }, + { + "epoch": 6.58, + "learning_rate": 1.0272297051328723e-06, + "loss": 1.6035, + "step": 36178 + }, + { + "epoch": 6.58, + "learning_rate": 1.02717510010921e-06, + "loss": 0.7845, + "step": 36179 + }, + { + "epoch": 6.58, + "learning_rate": 1.0271204950855478e-06, + "loss": 0.3632, + "step": 36180 + }, + { + "epoch": 6.58, + "learning_rate": 1.0270658900618857e-06, + "loss": 0.2958, + "step": 36181 + }, + { + "epoch": 6.58, + "learning_rate": 1.0270112850382235e-06, + "loss": 0.0053, + "step": 36182 + }, + { + "epoch": 6.58, + "learning_rate": 1.0269566800145614e-06, + "loss": 0.0082, + "step": 36183 + }, + { + "epoch": 6.58, + "learning_rate": 1.0269020749908991e-06, + "loss": 0.2927, + "step": 36184 + }, + { + "epoch": 6.58, + "learning_rate": 1.026847469967237e-06, + "loss": 1.1027, + "step": 36185 + }, + { + "epoch": 6.58, + "learning_rate": 1.0267928649435748e-06, + "loss": 0.3272, + "step": 36186 + }, + { + "epoch": 6.58, + "learning_rate": 1.0267382599199125e-06, + "loss": 0.0172, + "step": 36187 + }, + { + "epoch": 6.58, + "learning_rate": 1.0266836548962505e-06, + "loss": 1.809, + "step": 36188 + }, + { + "epoch": 6.58, + "learning_rate": 1.0266290498725882e-06, + "loss": 0.0168, + "step": 36189 + }, + { + "epoch": 6.58, + "learning_rate": 1.0265744448489262e-06, + "loss": 0.062, + "step": 36190 + }, + { + "epoch": 6.58, + "learning_rate": 1.0265198398252639e-06, + "loss": 0.0144, + "step": 36191 + }, + { + "epoch": 6.58, + "learning_rate": 1.0264652348016018e-06, + "loss": 0.1623, + "step": 36192 + }, + { + "epoch": 6.58, + "learning_rate": 1.0264106297779396e-06, + "loss": 1.2471, + "step": 36193 + }, + { + "epoch": 6.58, + "learning_rate": 1.0263560247542773e-06, + "loss": 1.1069, + "step": 36194 + }, + { + "epoch": 6.58, + "learning_rate": 1.0263014197306152e-06, + "loss": 0.9332, + "step": 36195 + }, + { + "epoch": 6.58, + "learning_rate": 1.0262468147069532e-06, + "loss": 1.191, + "step": 36196 + }, + { + "epoch": 6.58, + "learning_rate": 1.026192209683291e-06, + "loss": 1.3659, + "step": 36197 + }, + { + "epoch": 6.58, + "learning_rate": 1.0261376046596286e-06, + "loss": 0.0256, + "step": 36198 + }, + { + "epoch": 6.58, + "learning_rate": 1.0260829996359666e-06, + "loss": 0.187, + "step": 36199 + }, + { + "epoch": 6.58, + "learning_rate": 1.0260283946123043e-06, + "loss": 1.2496, + "step": 36200 + }, + { + "epoch": 6.58, + "learning_rate": 1.025973789588642e-06, + "loss": 0.0133, + "step": 36201 + }, + { + "epoch": 6.58, + "learning_rate": 1.02591918456498e-06, + "loss": 3.0457, + "step": 36202 + }, + { + "epoch": 6.58, + "learning_rate": 1.025864579541318e-06, + "loss": 0.134, + "step": 36203 + }, + { + "epoch": 6.58, + "learning_rate": 1.0258099745176557e-06, + "loss": 0.0169, + "step": 36204 + }, + { + "epoch": 6.58, + "learning_rate": 1.0257553694939934e-06, + "loss": 1.3391, + "step": 36205 + }, + { + "epoch": 6.58, + "learning_rate": 1.0257007644703313e-06, + "loss": 0.0165, + "step": 36206 + }, + { + "epoch": 6.58, + "learning_rate": 1.025646159446669e-06, + "loss": 0.3161, + "step": 36207 + }, + { + "epoch": 6.58, + "learning_rate": 1.0255915544230068e-06, + "loss": 0.0144, + "step": 36208 + }, + { + "epoch": 6.58, + "learning_rate": 1.0255369493993448e-06, + "loss": 0.9718, + "step": 36209 + }, + { + "epoch": 6.58, + "learning_rate": 1.0254823443756827e-06, + "loss": 0.0062, + "step": 36210 + }, + { + "epoch": 6.59, + "learning_rate": 1.0254277393520204e-06, + "loss": 0.2942, + "step": 36211 + }, + { + "epoch": 6.59, + "learning_rate": 1.0253731343283582e-06, + "loss": 0.0615, + "step": 36212 + }, + { + "epoch": 6.59, + "learning_rate": 1.025318529304696e-06, + "loss": 0.5374, + "step": 36213 + }, + { + "epoch": 6.59, + "learning_rate": 1.0252639242810338e-06, + "loss": 3.5005, + "step": 36214 + }, + { + "epoch": 6.59, + "learning_rate": 1.0252093192573716e-06, + "loss": 0.0109, + "step": 36215 + }, + { + "epoch": 6.59, + "learning_rate": 1.0251547142337095e-06, + "loss": 0.057, + "step": 36216 + }, + { + "epoch": 6.59, + "learning_rate": 1.0251001092100475e-06, + "loss": 0.4227, + "step": 36217 + }, + { + "epoch": 6.59, + "learning_rate": 1.0250455041863852e-06, + "loss": 0.3656, + "step": 36218 + }, + { + "epoch": 6.59, + "learning_rate": 1.024990899162723e-06, + "loss": 0.0039, + "step": 36219 + }, + { + "epoch": 6.59, + "learning_rate": 1.0249362941390609e-06, + "loss": 0.3429, + "step": 36220 + }, + { + "epoch": 6.59, + "learning_rate": 1.0248816891153986e-06, + "loss": 0.7244, + "step": 36221 + }, + { + "epoch": 6.59, + "learning_rate": 1.0248270840917363e-06, + "loss": 0.0083, + "step": 36222 + }, + { + "epoch": 6.59, + "learning_rate": 1.0247724790680743e-06, + "loss": 1.1864, + "step": 36223 + }, + { + "epoch": 6.59, + "learning_rate": 1.0247178740444122e-06, + "loss": 0.2044, + "step": 36224 + }, + { + "epoch": 6.59, + "learning_rate": 1.02466326902075e-06, + "loss": 0.0048, + "step": 36225 + }, + { + "epoch": 6.59, + "learning_rate": 1.0246086639970877e-06, + "loss": 0.2528, + "step": 36226 + }, + { + "epoch": 6.59, + "learning_rate": 1.0245540589734256e-06, + "loss": 0.255, + "step": 36227 + }, + { + "epoch": 6.59, + "learning_rate": 1.0244994539497633e-06, + "loss": 0.3391, + "step": 36228 + }, + { + "epoch": 6.59, + "learning_rate": 1.024444848926101e-06, + "loss": 0.3701, + "step": 36229 + }, + { + "epoch": 6.59, + "learning_rate": 1.024390243902439e-06, + "loss": 0.0103, + "step": 36230 + }, + { + "epoch": 6.59, + "learning_rate": 1.024335638878777e-06, + "loss": 1.3824, + "step": 36231 + }, + { + "epoch": 6.59, + "learning_rate": 1.0242810338551147e-06, + "loss": 0.2141, + "step": 36232 + }, + { + "epoch": 6.59, + "learning_rate": 1.0242264288314524e-06, + "loss": 0.0327, + "step": 36233 + }, + { + "epoch": 6.59, + "learning_rate": 1.0241718238077904e-06, + "loss": 0.0134, + "step": 36234 + }, + { + "epoch": 6.59, + "learning_rate": 1.0241172187841281e-06, + "loss": 2.4049, + "step": 36235 + }, + { + "epoch": 6.59, + "learning_rate": 1.024062613760466e-06, + "loss": 0.0852, + "step": 36236 + }, + { + "epoch": 6.59, + "learning_rate": 1.024008008736804e-06, + "loss": 0.229, + "step": 36237 + }, + { + "epoch": 6.59, + "learning_rate": 1.0239534037131417e-06, + "loss": 0.0243, + "step": 36238 + }, + { + "epoch": 6.59, + "learning_rate": 1.0238987986894795e-06, + "loss": 0.2772, + "step": 36239 + }, + { + "epoch": 6.59, + "learning_rate": 1.0238441936658172e-06, + "loss": 1.5368, + "step": 36240 + }, + { + "epoch": 6.59, + "learning_rate": 1.0237895886421551e-06, + "loss": 0.4025, + "step": 36241 + }, + { + "epoch": 6.59, + "learning_rate": 1.0237349836184929e-06, + "loss": 0.7792, + "step": 36242 + }, + { + "epoch": 6.59, + "learning_rate": 1.0236803785948308e-06, + "loss": 1.3278, + "step": 36243 + }, + { + "epoch": 6.59, + "learning_rate": 1.0236257735711687e-06, + "loss": 0.0163, + "step": 36244 + }, + { + "epoch": 6.59, + "learning_rate": 1.0235711685475065e-06, + "loss": 0.115, + "step": 36245 + }, + { + "epoch": 6.59, + "learning_rate": 1.0235165635238442e-06, + "loss": 0.0869, + "step": 36246 + }, + { + "epoch": 6.59, + "learning_rate": 1.023461958500182e-06, + "loss": 0.0805, + "step": 36247 + }, + { + "epoch": 6.59, + "learning_rate": 1.0234073534765199e-06, + "loss": 0.7163, + "step": 36248 + }, + { + "epoch": 6.59, + "learning_rate": 1.0233527484528576e-06, + "loss": 0.0832, + "step": 36249 + }, + { + "epoch": 6.59, + "learning_rate": 1.0232981434291956e-06, + "loss": 0.155, + "step": 36250 + }, + { + "epoch": 6.59, + "learning_rate": 1.0232435384055333e-06, + "loss": 0.5251, + "step": 36251 + }, + { + "epoch": 6.59, + "learning_rate": 1.0231889333818712e-06, + "loss": 0.2132, + "step": 36252 + }, + { + "epoch": 6.59, + "learning_rate": 1.023134328358209e-06, + "loss": 0.005, + "step": 36253 + }, + { + "epoch": 6.59, + "learning_rate": 1.0230797233345467e-06, + "loss": 0.1286, + "step": 36254 + }, + { + "epoch": 6.59, + "learning_rate": 1.0230251183108846e-06, + "loss": 0.0089, + "step": 36255 + }, + { + "epoch": 6.59, + "learning_rate": 1.0229705132872224e-06, + "loss": 1.319, + "step": 36256 + }, + { + "epoch": 6.59, + "learning_rate": 1.0229159082635603e-06, + "loss": 0.4533, + "step": 36257 + }, + { + "epoch": 6.59, + "learning_rate": 1.022861303239898e-06, + "loss": 2.0017, + "step": 36258 + }, + { + "epoch": 6.59, + "learning_rate": 1.022806698216236e-06, + "loss": 1.2466, + "step": 36259 + }, + { + "epoch": 6.59, + "learning_rate": 1.0227520931925737e-06, + "loss": 0.015, + "step": 36260 + }, + { + "epoch": 6.59, + "learning_rate": 1.0226974881689115e-06, + "loss": 0.1671, + "step": 36261 + }, + { + "epoch": 6.59, + "learning_rate": 1.0226428831452494e-06, + "loss": 0.1443, + "step": 36262 + }, + { + "epoch": 6.59, + "learning_rate": 1.0225882781215871e-06, + "loss": 0.0411, + "step": 36263 + }, + { + "epoch": 6.59, + "learning_rate": 1.022533673097925e-06, + "loss": 0.7151, + "step": 36264 + }, + { + "epoch": 6.59, + "learning_rate": 1.0224790680742628e-06, + "loss": 0.6076, + "step": 36265 + }, + { + "epoch": 6.6, + "learning_rate": 1.0224244630506008e-06, + "loss": 0.2396, + "step": 36266 + }, + { + "epoch": 6.6, + "learning_rate": 1.0223698580269385e-06, + "loss": 0.1941, + "step": 36267 + }, + { + "epoch": 6.6, + "learning_rate": 1.0223152530032762e-06, + "loss": 0.2521, + "step": 36268 + }, + { + "epoch": 6.6, + "learning_rate": 1.0222606479796142e-06, + "loss": 0.1275, + "step": 36269 + }, + { + "epoch": 6.6, + "learning_rate": 1.0222060429559519e-06, + "loss": 0.5253, + "step": 36270 + }, + { + "epoch": 6.6, + "learning_rate": 1.0221514379322898e-06, + "loss": 4.1694, + "step": 36271 + }, + { + "epoch": 6.6, + "learning_rate": 1.0220968329086276e-06, + "loss": 0.0948, + "step": 36272 + }, + { + "epoch": 6.6, + "learning_rate": 1.0220422278849655e-06, + "loss": 0.0437, + "step": 36273 + }, + { + "epoch": 6.6, + "learning_rate": 1.0219876228613032e-06, + "loss": 1.7693, + "step": 36274 + }, + { + "epoch": 6.6, + "learning_rate": 1.021933017837641e-06, + "loss": 0.9519, + "step": 36275 + }, + { + "epoch": 6.6, + "learning_rate": 1.021878412813979e-06, + "loss": 2.756, + "step": 36276 + }, + { + "epoch": 6.6, + "learning_rate": 1.0218238077903169e-06, + "loss": 0.0046, + "step": 36277 + }, + { + "epoch": 6.6, + "learning_rate": 1.0217692027666546e-06, + "loss": 0.0338, + "step": 36278 + }, + { + "epoch": 6.6, + "learning_rate": 1.0217145977429923e-06, + "loss": 1.5528, + "step": 36279 + }, + { + "epoch": 6.6, + "learning_rate": 1.0216599927193303e-06, + "loss": 0.5212, + "step": 36280 + }, + { + "epoch": 6.6, + "learning_rate": 1.021605387695668e-06, + "loss": 0.6254, + "step": 36281 + }, + { + "epoch": 6.6, + "learning_rate": 1.0215507826720057e-06, + "loss": 0.0836, + "step": 36282 + }, + { + "epoch": 6.6, + "learning_rate": 1.0214961776483437e-06, + "loss": 0.0359, + "step": 36283 + }, + { + "epoch": 6.6, + "learning_rate": 1.0214415726246816e-06, + "loss": 0.0054, + "step": 36284 + }, + { + "epoch": 6.6, + "learning_rate": 1.0213869676010193e-06, + "loss": 0.0756, + "step": 36285 + }, + { + "epoch": 6.6, + "learning_rate": 1.021332362577357e-06, + "loss": 0.3188, + "step": 36286 + }, + { + "epoch": 6.6, + "learning_rate": 1.021277757553695e-06, + "loss": 0.1589, + "step": 36287 + }, + { + "epoch": 6.6, + "learning_rate": 1.0212231525300328e-06, + "loss": 0.0484, + "step": 36288 + }, + { + "epoch": 6.6, + "learning_rate": 1.0211685475063705e-06, + "loss": 3.2212, + "step": 36289 + }, + { + "epoch": 6.6, + "learning_rate": 1.0211139424827084e-06, + "loss": 0.0072, + "step": 36290 + }, + { + "epoch": 6.6, + "learning_rate": 1.0210593374590464e-06, + "loss": 0.0602, + "step": 36291 + }, + { + "epoch": 6.6, + "learning_rate": 1.021004732435384e-06, + "loss": 0.2007, + "step": 36292 + }, + { + "epoch": 6.6, + "learning_rate": 1.0209501274117218e-06, + "loss": 0.9991, + "step": 36293 + }, + { + "epoch": 6.6, + "learning_rate": 1.0208955223880598e-06, + "loss": 1.0142, + "step": 36294 + }, + { + "epoch": 6.6, + "learning_rate": 1.0208409173643975e-06, + "loss": 0.0727, + "step": 36295 + }, + { + "epoch": 6.6, + "learning_rate": 1.0207863123407352e-06, + "loss": 0.3529, + "step": 36296 + }, + { + "epoch": 6.6, + "learning_rate": 1.0207317073170732e-06, + "loss": 1.1961, + "step": 36297 + }, + { + "epoch": 6.6, + "learning_rate": 1.0206771022934111e-06, + "loss": 0.021, + "step": 36298 + }, + { + "epoch": 6.6, + "learning_rate": 1.0206224972697489e-06, + "loss": 0.1798, + "step": 36299 + }, + { + "epoch": 6.6, + "learning_rate": 1.0205678922460866e-06, + "loss": 0.9722, + "step": 36300 + }, + { + "epoch": 6.6, + "learning_rate": 1.0205132872224245e-06, + "loss": 0.2509, + "step": 36301 + }, + { + "epoch": 6.6, + "learning_rate": 1.0204586821987623e-06, + "loss": 0.1489, + "step": 36302 + }, + { + "epoch": 6.6, + "learning_rate": 1.0204040771751e-06, + "loss": 0.0037, + "step": 36303 + }, + { + "epoch": 6.6, + "learning_rate": 1.020349472151438e-06, + "loss": 0.0969, + "step": 36304 + }, + { + "epoch": 6.6, + "learning_rate": 1.0202948671277759e-06, + "loss": 0.4806, + "step": 36305 + }, + { + "epoch": 6.6, + "learning_rate": 1.0202402621041136e-06, + "loss": 0.2238, + "step": 36306 + }, + { + "epoch": 6.6, + "learning_rate": 1.0201856570804513e-06, + "loss": 0.2061, + "step": 36307 + }, + { + "epoch": 6.6, + "learning_rate": 1.0201310520567893e-06, + "loss": 0.0661, + "step": 36308 + }, + { + "epoch": 6.6, + "learning_rate": 1.020076447033127e-06, + "loss": 0.153, + "step": 36309 + }, + { + "epoch": 6.6, + "learning_rate": 1.0200218420094648e-06, + "loss": 0.577, + "step": 36310 + }, + { + "epoch": 6.6, + "learning_rate": 1.0199672369858027e-06, + "loss": 0.2455, + "step": 36311 + }, + { + "epoch": 6.6, + "learning_rate": 1.0199126319621406e-06, + "loss": 0.7568, + "step": 36312 + }, + { + "epoch": 6.6, + "learning_rate": 1.0198580269384784e-06, + "loss": 0.4207, + "step": 36313 + }, + { + "epoch": 6.6, + "learning_rate": 1.019803421914816e-06, + "loss": 1.5189, + "step": 36314 + }, + { + "epoch": 6.6, + "learning_rate": 1.019748816891154e-06, + "loss": 1.1892, + "step": 36315 + }, + { + "epoch": 6.6, + "learning_rate": 1.0196942118674918e-06, + "loss": 0.1437, + "step": 36316 + }, + { + "epoch": 6.6, + "learning_rate": 1.0196396068438297e-06, + "loss": 0.0887, + "step": 36317 + }, + { + "epoch": 6.6, + "learning_rate": 1.0195850018201677e-06, + "loss": 0.0172, + "step": 36318 + }, + { + "epoch": 6.6, + "learning_rate": 1.0195303967965054e-06, + "loss": 0.0141, + "step": 36319 + }, + { + "epoch": 6.6, + "learning_rate": 1.0194757917728431e-06, + "loss": 0.4287, + "step": 36320 + }, + { + "epoch": 6.61, + "learning_rate": 1.0194211867491809e-06, + "loss": 0.1813, + "step": 36321 + }, + { + "epoch": 6.61, + "learning_rate": 1.0193665817255188e-06, + "loss": 0.6101, + "step": 36322 + }, + { + "epoch": 6.61, + "learning_rate": 1.0193119767018565e-06, + "loss": 0.0027, + "step": 36323 + }, + { + "epoch": 6.61, + "learning_rate": 1.0192573716781945e-06, + "loss": 0.3062, + "step": 36324 + }, + { + "epoch": 6.61, + "learning_rate": 1.0192027666545324e-06, + "loss": 0.7217, + "step": 36325 + }, + { + "epoch": 6.61, + "learning_rate": 1.0191481616308702e-06, + "loss": 0.0666, + "step": 36326 + }, + { + "epoch": 6.61, + "learning_rate": 1.0190935566072079e-06, + "loss": 0.0112, + "step": 36327 + }, + { + "epoch": 6.61, + "learning_rate": 1.0190389515835456e-06, + "loss": 1.0867, + "step": 36328 + }, + { + "epoch": 6.61, + "learning_rate": 1.0189843465598836e-06, + "loss": 0.0355, + "step": 36329 + }, + { + "epoch": 6.61, + "learning_rate": 1.0189297415362213e-06, + "loss": 0.0113, + "step": 36330 + }, + { + "epoch": 6.61, + "learning_rate": 1.0188751365125592e-06, + "loss": 2.0245, + "step": 36331 + }, + { + "epoch": 6.61, + "learning_rate": 1.018820531488897e-06, + "loss": 0.0967, + "step": 36332 + }, + { + "epoch": 6.61, + "learning_rate": 1.018765926465235e-06, + "loss": 0.576, + "step": 36333 + }, + { + "epoch": 6.61, + "learning_rate": 1.0187113214415726e-06, + "loss": 1.57, + "step": 36334 + }, + { + "epoch": 6.61, + "learning_rate": 1.0186567164179104e-06, + "loss": 1.3227, + "step": 36335 + }, + { + "epoch": 6.61, + "learning_rate": 1.0186021113942483e-06, + "loss": 0.7164, + "step": 36336 + }, + { + "epoch": 6.61, + "learning_rate": 1.018547506370586e-06, + "loss": 0.362, + "step": 36337 + }, + { + "epoch": 6.61, + "learning_rate": 1.018492901346924e-06, + "loss": 0.1253, + "step": 36338 + }, + { + "epoch": 6.61, + "learning_rate": 1.0184382963232617e-06, + "loss": 0.0077, + "step": 36339 + }, + { + "epoch": 6.61, + "learning_rate": 1.0183836912995997e-06, + "loss": 0.2208, + "step": 36340 + }, + { + "epoch": 6.61, + "learning_rate": 1.0183290862759374e-06, + "loss": 0.0066, + "step": 36341 + }, + { + "epoch": 6.61, + "learning_rate": 1.0182744812522751e-06, + "loss": 0.0057, + "step": 36342 + }, + { + "epoch": 6.61, + "learning_rate": 1.018219876228613e-06, + "loss": 0.0977, + "step": 36343 + }, + { + "epoch": 6.61, + "learning_rate": 1.0181652712049508e-06, + "loss": 0.4439, + "step": 36344 + }, + { + "epoch": 6.61, + "learning_rate": 1.0181106661812888e-06, + "loss": 0.0296, + "step": 36345 + }, + { + "epoch": 6.61, + "learning_rate": 1.0180560611576265e-06, + "loss": 0.1487, + "step": 36346 + }, + { + "epoch": 6.61, + "learning_rate": 1.0180014561339644e-06, + "loss": 0.3339, + "step": 36347 + }, + { + "epoch": 6.61, + "learning_rate": 1.0179468511103022e-06, + "loss": 0.0636, + "step": 36348 + }, + { + "epoch": 6.61, + "learning_rate": 1.0178922460866399e-06, + "loss": 0.2135, + "step": 36349 + }, + { + "epoch": 6.61, + "learning_rate": 1.0178376410629778e-06, + "loss": 0.6975, + "step": 36350 + }, + { + "epoch": 6.61, + "learning_rate": 1.0177830360393156e-06, + "loss": 0.5014, + "step": 36351 + }, + { + "epoch": 6.61, + "learning_rate": 1.0177284310156535e-06, + "loss": 0.0322, + "step": 36352 + }, + { + "epoch": 6.61, + "learning_rate": 1.0176738259919912e-06, + "loss": 1.492, + "step": 36353 + }, + { + "epoch": 6.61, + "learning_rate": 1.0176192209683292e-06, + "loss": 0.0901, + "step": 36354 + }, + { + "epoch": 6.61, + "learning_rate": 1.017564615944667e-06, + "loss": 0.4617, + "step": 36355 + }, + { + "epoch": 6.61, + "learning_rate": 1.0175100109210046e-06, + "loss": 0.931, + "step": 36356 + }, + { + "epoch": 6.61, + "learning_rate": 1.0174554058973426e-06, + "loss": 1.2751, + "step": 36357 + }, + { + "epoch": 6.61, + "learning_rate": 1.0174008008736803e-06, + "loss": 2.3066, + "step": 36358 + }, + { + "epoch": 6.61, + "learning_rate": 1.0173461958500183e-06, + "loss": 0.1052, + "step": 36359 + }, + { + "epoch": 6.61, + "learning_rate": 1.017291590826356e-06, + "loss": 0.29, + "step": 36360 + }, + { + "epoch": 6.61, + "learning_rate": 1.017236985802694e-06, + "loss": 0.005, + "step": 36361 + }, + { + "epoch": 6.61, + "learning_rate": 1.0171823807790317e-06, + "loss": 2.2887, + "step": 36362 + }, + { + "epoch": 6.61, + "learning_rate": 1.0171277757553694e-06, + "loss": 0.0737, + "step": 36363 + }, + { + "epoch": 6.61, + "learning_rate": 1.0170731707317073e-06, + "loss": 0.0686, + "step": 36364 + }, + { + "epoch": 6.61, + "learning_rate": 1.0170185657080453e-06, + "loss": 0.799, + "step": 36365 + }, + { + "epoch": 6.61, + "learning_rate": 1.016963960684383e-06, + "loss": 0.1358, + "step": 36366 + }, + { + "epoch": 6.61, + "learning_rate": 1.0169093556607208e-06, + "loss": 0.0065, + "step": 36367 + }, + { + "epoch": 6.61, + "learning_rate": 1.0168547506370587e-06, + "loss": 0.0254, + "step": 36368 + }, + { + "epoch": 6.61, + "learning_rate": 1.0168001456133964e-06, + "loss": 1.2189, + "step": 36369 + }, + { + "epoch": 6.61, + "learning_rate": 1.0167455405897342e-06, + "loss": 1.3055, + "step": 36370 + }, + { + "epoch": 6.61, + "learning_rate": 1.016690935566072e-06, + "loss": 0.2389, + "step": 36371 + }, + { + "epoch": 6.61, + "learning_rate": 1.01663633054241e-06, + "loss": 0.008, + "step": 36372 + }, + { + "epoch": 6.61, + "learning_rate": 1.0165817255187478e-06, + "loss": 0.4767, + "step": 36373 + }, + { + "epoch": 6.61, + "learning_rate": 1.0165271204950855e-06, + "loss": 0.1199, + "step": 36374 + }, + { + "epoch": 6.61, + "learning_rate": 1.0164725154714235e-06, + "loss": 0.0358, + "step": 36375 + }, + { + "epoch": 6.62, + "learning_rate": 1.0164179104477612e-06, + "loss": 1.239, + "step": 36376 + }, + { + "epoch": 6.62, + "learning_rate": 1.016363305424099e-06, + "loss": 0.0456, + "step": 36377 + }, + { + "epoch": 6.62, + "learning_rate": 1.0163087004004369e-06, + "loss": 0.0165, + "step": 36378 + }, + { + "epoch": 6.62, + "learning_rate": 1.0162540953767748e-06, + "loss": 2.2068, + "step": 36379 + }, + { + "epoch": 6.62, + "learning_rate": 1.0161994903531125e-06, + "loss": 0.5674, + "step": 36380 + }, + { + "epoch": 6.62, + "learning_rate": 1.0161448853294503e-06, + "loss": 0.2031, + "step": 36381 + }, + { + "epoch": 6.62, + "learning_rate": 1.0160902803057882e-06, + "loss": 0.0907, + "step": 36382 + }, + { + "epoch": 6.62, + "learning_rate": 1.016035675282126e-06, + "loss": 0.2478, + "step": 36383 + }, + { + "epoch": 6.62, + "learning_rate": 1.0159810702584637e-06, + "loss": 0.1406, + "step": 36384 + }, + { + "epoch": 6.62, + "learning_rate": 1.0159264652348016e-06, + "loss": 1.7319, + "step": 36385 + }, + { + "epoch": 6.62, + "learning_rate": 1.0158718602111396e-06, + "loss": 0.0219, + "step": 36386 + }, + { + "epoch": 6.62, + "learning_rate": 1.0158172551874773e-06, + "loss": 0.9761, + "step": 36387 + }, + { + "epoch": 6.62, + "learning_rate": 1.015762650163815e-06, + "loss": 0.8347, + "step": 36388 + }, + { + "epoch": 6.62, + "learning_rate": 1.015708045140153e-06, + "loss": 0.0161, + "step": 36389 + }, + { + "epoch": 6.62, + "learning_rate": 1.0156534401164907e-06, + "loss": 0.4411, + "step": 36390 + }, + { + "epoch": 6.62, + "learning_rate": 1.0155988350928284e-06, + "loss": 1.146, + "step": 36391 + }, + { + "epoch": 6.62, + "learning_rate": 1.0155442300691664e-06, + "loss": 0.0075, + "step": 36392 + }, + { + "epoch": 6.62, + "learning_rate": 1.0154896250455043e-06, + "loss": 0.1195, + "step": 36393 + }, + { + "epoch": 6.62, + "learning_rate": 1.015435020021842e-06, + "loss": 1.0881, + "step": 36394 + }, + { + "epoch": 6.62, + "learning_rate": 1.0153804149981798e-06, + "loss": 1.2423, + "step": 36395 + }, + { + "epoch": 6.62, + "learning_rate": 1.0153258099745177e-06, + "loss": 0.8648, + "step": 36396 + }, + { + "epoch": 6.62, + "learning_rate": 1.0152712049508555e-06, + "loss": 2.2992, + "step": 36397 + }, + { + "epoch": 6.62, + "learning_rate": 1.0152165999271932e-06, + "loss": 0.5904, + "step": 36398 + }, + { + "epoch": 6.62, + "learning_rate": 1.0151619949035311e-06, + "loss": 0.5655, + "step": 36399 + }, + { + "epoch": 6.62, + "learning_rate": 1.015107389879869e-06, + "loss": 0.5516, + "step": 36400 + }, + { + "epoch": 6.62, + "learning_rate": 1.0150527848562068e-06, + "loss": 0.0933, + "step": 36401 + }, + { + "epoch": 6.62, + "learning_rate": 1.0149981798325445e-06, + "loss": 0.0051, + "step": 36402 + }, + { + "epoch": 6.62, + "learning_rate": 1.0149435748088825e-06, + "loss": 0.2444, + "step": 36403 + }, + { + "epoch": 6.62, + "learning_rate": 1.0148889697852202e-06, + "loss": 0.3805, + "step": 36404 + }, + { + "epoch": 6.62, + "learning_rate": 1.0148343647615582e-06, + "loss": 2.3993, + "step": 36405 + }, + { + "epoch": 6.62, + "learning_rate": 1.0147797597378959e-06, + "loss": 0.1004, + "step": 36406 + }, + { + "epoch": 6.62, + "learning_rate": 1.0147251547142338e-06, + "loss": 0.3128, + "step": 36407 + }, + { + "epoch": 6.62, + "learning_rate": 1.0146705496905716e-06, + "loss": 0.2698, + "step": 36408 + }, + { + "epoch": 6.62, + "learning_rate": 1.0146159446669093e-06, + "loss": 0.0082, + "step": 36409 + }, + { + "epoch": 6.62, + "learning_rate": 1.0145613396432472e-06, + "loss": 1.3548, + "step": 36410 + }, + { + "epoch": 6.62, + "learning_rate": 1.014506734619585e-06, + "loss": 0.112, + "step": 36411 + }, + { + "epoch": 6.62, + "learning_rate": 1.014452129595923e-06, + "loss": 1.2406, + "step": 36412 + }, + { + "epoch": 6.62, + "learning_rate": 1.0143975245722606e-06, + "loss": 1.025, + "step": 36413 + }, + { + "epoch": 6.62, + "learning_rate": 1.0143429195485986e-06, + "loss": 0.6304, + "step": 36414 + }, + { + "epoch": 6.62, + "learning_rate": 1.0142883145249363e-06, + "loss": 0.3644, + "step": 36415 + }, + { + "epoch": 6.62, + "learning_rate": 1.014233709501274e-06, + "loss": 0.0339, + "step": 36416 + }, + { + "epoch": 6.62, + "learning_rate": 1.014179104477612e-06, + "loss": 1.1242, + "step": 36417 + }, + { + "epoch": 6.62, + "learning_rate": 1.0141244994539497e-06, + "loss": 1.5818, + "step": 36418 + }, + { + "epoch": 6.62, + "learning_rate": 1.0140698944302877e-06, + "loss": 0.1522, + "step": 36419 + }, + { + "epoch": 6.62, + "learning_rate": 1.0140152894066254e-06, + "loss": 0.0053, + "step": 36420 + }, + { + "epoch": 6.62, + "learning_rate": 1.0139606843829633e-06, + "loss": 0.8458, + "step": 36421 + }, + { + "epoch": 6.62, + "learning_rate": 1.013906079359301e-06, + "loss": 2.2499, + "step": 36422 + }, + { + "epoch": 6.62, + "learning_rate": 1.0138514743356388e-06, + "loss": 0.4454, + "step": 36423 + }, + { + "epoch": 6.62, + "learning_rate": 1.0137968693119767e-06, + "loss": 0.0893, + "step": 36424 + }, + { + "epoch": 6.62, + "learning_rate": 1.0137422642883145e-06, + "loss": 0.7448, + "step": 36425 + }, + { + "epoch": 6.62, + "learning_rate": 1.0136876592646524e-06, + "loss": 0.0078, + "step": 36426 + }, + { + "epoch": 6.62, + "learning_rate": 1.0136330542409902e-06, + "loss": 1.1015, + "step": 36427 + }, + { + "epoch": 6.62, + "learning_rate": 1.013578449217328e-06, + "loss": 0.0217, + "step": 36428 + }, + { + "epoch": 6.62, + "learning_rate": 1.0135238441936658e-06, + "loss": 0.1246, + "step": 36429 + }, + { + "epoch": 6.62, + "learning_rate": 1.0134692391700036e-06, + "loss": 0.3419, + "step": 36430 + }, + { + "epoch": 6.63, + "learning_rate": 1.0134146341463415e-06, + "loss": 1.8747, + "step": 36431 + }, + { + "epoch": 6.63, + "learning_rate": 1.0133600291226792e-06, + "loss": 0.0273, + "step": 36432 + }, + { + "epoch": 6.63, + "learning_rate": 1.0133054240990172e-06, + "loss": 0.256, + "step": 36433 + }, + { + "epoch": 6.63, + "learning_rate": 1.013250819075355e-06, + "loss": 0.4832, + "step": 36434 + }, + { + "epoch": 6.63, + "learning_rate": 1.0131962140516929e-06, + "loss": 1.3626, + "step": 36435 + }, + { + "epoch": 6.63, + "learning_rate": 1.0131416090280306e-06, + "loss": 0.0051, + "step": 36436 + }, + { + "epoch": 6.63, + "learning_rate": 1.0130870040043683e-06, + "loss": 0.4206, + "step": 36437 + }, + { + "epoch": 6.63, + "learning_rate": 1.0130323989807063e-06, + "loss": 0.405, + "step": 36438 + }, + { + "epoch": 6.63, + "learning_rate": 1.012977793957044e-06, + "loss": 0.2151, + "step": 36439 + }, + { + "epoch": 6.63, + "learning_rate": 1.012923188933382e-06, + "loss": 0.6675, + "step": 36440 + }, + { + "epoch": 6.63, + "learning_rate": 1.0128685839097197e-06, + "loss": 1.2193, + "step": 36441 + }, + { + "epoch": 6.63, + "learning_rate": 1.0128139788860576e-06, + "loss": 0.0607, + "step": 36442 + }, + { + "epoch": 6.63, + "learning_rate": 1.0127593738623953e-06, + "loss": 0.4, + "step": 36443 + }, + { + "epoch": 6.63, + "learning_rate": 1.012704768838733e-06, + "loss": 0.8742, + "step": 36444 + }, + { + "epoch": 6.63, + "learning_rate": 1.012650163815071e-06, + "loss": 0.7006, + "step": 36445 + }, + { + "epoch": 6.63, + "learning_rate": 1.012595558791409e-06, + "loss": 0.7048, + "step": 36446 + }, + { + "epoch": 6.63, + "learning_rate": 1.0125409537677467e-06, + "loss": 0.0159, + "step": 36447 + }, + { + "epoch": 6.63, + "learning_rate": 1.0124863487440844e-06, + "loss": 0.0239, + "step": 36448 + }, + { + "epoch": 6.63, + "learning_rate": 1.0124317437204224e-06, + "loss": 0.6402, + "step": 36449 + }, + { + "epoch": 6.63, + "learning_rate": 1.01237713869676e-06, + "loss": 0.7936, + "step": 36450 + }, + { + "epoch": 6.63, + "learning_rate": 1.0123225336730978e-06, + "loss": 0.1964, + "step": 36451 + }, + { + "epoch": 6.63, + "learning_rate": 1.0122679286494358e-06, + "loss": 0.0039, + "step": 36452 + }, + { + "epoch": 6.63, + "learning_rate": 1.0122133236257737e-06, + "loss": 0.2176, + "step": 36453 + }, + { + "epoch": 6.63, + "learning_rate": 1.0121587186021115e-06, + "loss": 0.7887, + "step": 36454 + }, + { + "epoch": 6.63, + "learning_rate": 1.0121041135784492e-06, + "loss": 0.917, + "step": 36455 + }, + { + "epoch": 6.63, + "learning_rate": 1.0120495085547871e-06, + "loss": 1.283, + "step": 36456 + }, + { + "epoch": 6.63, + "learning_rate": 1.0119949035311249e-06, + "loss": 0.4812, + "step": 36457 + }, + { + "epoch": 6.63, + "learning_rate": 1.0119402985074626e-06, + "loss": 0.0564, + "step": 36458 + }, + { + "epoch": 6.63, + "learning_rate": 1.0118856934838005e-06, + "loss": 0.0201, + "step": 36459 + }, + { + "epoch": 6.63, + "learning_rate": 1.0118310884601385e-06, + "loss": 1.9016, + "step": 36460 + }, + { + "epoch": 6.63, + "learning_rate": 1.0117764834364762e-06, + "loss": 0.0213, + "step": 36461 + }, + { + "epoch": 6.63, + "learning_rate": 1.011721878412814e-06, + "loss": 0.0446, + "step": 36462 + }, + { + "epoch": 6.63, + "learning_rate": 1.0116672733891519e-06, + "loss": 0.1021, + "step": 36463 + }, + { + "epoch": 6.63, + "learning_rate": 1.0116126683654896e-06, + "loss": 0.0093, + "step": 36464 + }, + { + "epoch": 6.63, + "learning_rate": 1.0115580633418273e-06, + "loss": 0.2759, + "step": 36465 + }, + { + "epoch": 6.63, + "learning_rate": 1.0115034583181653e-06, + "loss": 0.0163, + "step": 36466 + }, + { + "epoch": 6.63, + "learning_rate": 1.0114488532945032e-06, + "loss": 0.1744, + "step": 36467 + }, + { + "epoch": 6.63, + "learning_rate": 1.011394248270841e-06, + "loss": 0.2128, + "step": 36468 + }, + { + "epoch": 6.63, + "learning_rate": 1.0113396432471787e-06, + "loss": 1.7835, + "step": 36469 + }, + { + "epoch": 6.63, + "learning_rate": 1.0112850382235166e-06, + "loss": 2.0294, + "step": 36470 + }, + { + "epoch": 6.63, + "learning_rate": 1.0112304331998544e-06, + "loss": 2.98, + "step": 36471 + }, + { + "epoch": 6.63, + "learning_rate": 1.011175828176192e-06, + "loss": 0.222, + "step": 36472 + }, + { + "epoch": 6.63, + "learning_rate": 1.01112122315253e-06, + "loss": 0.859, + "step": 36473 + }, + { + "epoch": 6.63, + "learning_rate": 1.011066618128868e-06, + "loss": 0.0061, + "step": 36474 + }, + { + "epoch": 6.63, + "learning_rate": 1.0110120131052057e-06, + "loss": 0.2204, + "step": 36475 + }, + { + "epoch": 6.63, + "learning_rate": 1.0109574080815435e-06, + "loss": 0.0056, + "step": 36476 + }, + { + "epoch": 6.63, + "learning_rate": 1.0109028030578814e-06, + "loss": 0.3115, + "step": 36477 + }, + { + "epoch": 6.63, + "learning_rate": 1.0108481980342191e-06, + "loss": 0.4815, + "step": 36478 + }, + { + "epoch": 6.63, + "learning_rate": 1.0107935930105569e-06, + "loss": 0.2536, + "step": 36479 + }, + { + "epoch": 6.63, + "learning_rate": 1.0107389879868948e-06, + "loss": 0.2608, + "step": 36480 + }, + { + "epoch": 6.63, + "learning_rate": 1.0106843829632327e-06, + "loss": 0.6222, + "step": 36481 + }, + { + "epoch": 6.63, + "learning_rate": 1.0106297779395705e-06, + "loss": 2.8228, + "step": 36482 + }, + { + "epoch": 6.63, + "learning_rate": 1.0105751729159082e-06, + "loss": 0.0208, + "step": 36483 + }, + { + "epoch": 6.63, + "learning_rate": 1.0105205678922462e-06, + "loss": 0.0051, + "step": 36484 + }, + { + "epoch": 6.63, + "learning_rate": 1.0104659628685839e-06, + "loss": 0.0777, + "step": 36485 + }, + { + "epoch": 6.64, + "learning_rate": 1.0104113578449218e-06, + "loss": 0.1767, + "step": 36486 + }, + { + "epoch": 6.64, + "learning_rate": 1.0103567528212596e-06, + "loss": 0.3063, + "step": 36487 + }, + { + "epoch": 6.64, + "learning_rate": 1.0103021477975975e-06, + "loss": 0.1249, + "step": 36488 + }, + { + "epoch": 6.64, + "learning_rate": 1.0102475427739352e-06, + "loss": 3.0066, + "step": 36489 + }, + { + "epoch": 6.64, + "learning_rate": 1.010192937750273e-06, + "loss": 0.9764, + "step": 36490 + }, + { + "epoch": 6.64, + "learning_rate": 1.010138332726611e-06, + "loss": 1.2918, + "step": 36491 + }, + { + "epoch": 6.64, + "learning_rate": 1.0100837277029486e-06, + "loss": 2.1521, + "step": 36492 + }, + { + "epoch": 6.64, + "learning_rate": 1.0100291226792866e-06, + "loss": 0.329, + "step": 36493 + }, + { + "epoch": 6.64, + "learning_rate": 1.0099745176556243e-06, + "loss": 0.1973, + "step": 36494 + }, + { + "epoch": 6.64, + "learning_rate": 1.0099199126319623e-06, + "loss": 0.2336, + "step": 36495 + }, + { + "epoch": 6.64, + "learning_rate": 1.0098653076083e-06, + "loss": 0.2445, + "step": 36496 + }, + { + "epoch": 6.64, + "learning_rate": 1.0098107025846377e-06, + "loss": 0.0809, + "step": 36497 + }, + { + "epoch": 6.64, + "learning_rate": 1.0097560975609757e-06, + "loss": 1.9491, + "step": 36498 + }, + { + "epoch": 6.64, + "learning_rate": 1.0097014925373134e-06, + "loss": 0.0235, + "step": 36499 + }, + { + "epoch": 6.64, + "learning_rate": 1.0096468875136513e-06, + "loss": 0.1901, + "step": 36500 + }, + { + "epoch": 6.64, + "learning_rate": 1.009592282489989e-06, + "loss": 0.0066, + "step": 36501 + }, + { + "epoch": 6.64, + "learning_rate": 1.009537677466327e-06, + "loss": 0.4937, + "step": 36502 + }, + { + "epoch": 6.64, + "learning_rate": 1.0094830724426647e-06, + "loss": 0.4361, + "step": 36503 + }, + { + "epoch": 6.64, + "learning_rate": 1.0094284674190025e-06, + "loss": 0.2928, + "step": 36504 + }, + { + "epoch": 6.64, + "learning_rate": 1.0093738623953404e-06, + "loss": 0.1844, + "step": 36505 + }, + { + "epoch": 6.64, + "learning_rate": 1.0093192573716782e-06, + "loss": 0.0103, + "step": 36506 + }, + { + "epoch": 6.64, + "learning_rate": 1.009264652348016e-06, + "loss": 0.0035, + "step": 36507 + }, + { + "epoch": 6.64, + "learning_rate": 1.0092100473243538e-06, + "loss": 1.2497, + "step": 36508 + }, + { + "epoch": 6.64, + "learning_rate": 1.0091554423006918e-06, + "loss": 4.9608, + "step": 36509 + }, + { + "epoch": 6.64, + "learning_rate": 1.0091008372770295e-06, + "loss": 1.8959, + "step": 36510 + }, + { + "epoch": 6.64, + "learning_rate": 1.0090462322533672e-06, + "loss": 0.007, + "step": 36511 + }, + { + "epoch": 6.64, + "learning_rate": 1.0089916272297052e-06, + "loss": 0.0388, + "step": 36512 + }, + { + "epoch": 6.64, + "learning_rate": 1.008937022206043e-06, + "loss": 0.2679, + "step": 36513 + }, + { + "epoch": 6.64, + "learning_rate": 1.0088824171823809e-06, + "loss": 0.6997, + "step": 36514 + }, + { + "epoch": 6.64, + "learning_rate": 1.0088278121587186e-06, + "loss": 0.0223, + "step": 36515 + }, + { + "epoch": 6.64, + "learning_rate": 1.0087732071350565e-06, + "loss": 1.821, + "step": 36516 + }, + { + "epoch": 6.64, + "learning_rate": 1.0087186021113943e-06, + "loss": 0.7034, + "step": 36517 + }, + { + "epoch": 6.64, + "learning_rate": 1.008663997087732e-06, + "loss": 2.1535, + "step": 36518 + }, + { + "epoch": 6.64, + "learning_rate": 1.00860939206407e-06, + "loss": 0.0189, + "step": 36519 + }, + { + "epoch": 6.64, + "learning_rate": 1.0085547870404077e-06, + "loss": 0.2546, + "step": 36520 + }, + { + "epoch": 6.64, + "learning_rate": 1.0085001820167456e-06, + "loss": 0.0078, + "step": 36521 + }, + { + "epoch": 6.64, + "learning_rate": 1.0084455769930833e-06, + "loss": 0.6962, + "step": 36522 + }, + { + "epoch": 6.64, + "learning_rate": 1.0083909719694213e-06, + "loss": 0.0197, + "step": 36523 + }, + { + "epoch": 6.64, + "learning_rate": 1.008336366945759e-06, + "loss": 0.0956, + "step": 36524 + }, + { + "epoch": 6.64, + "learning_rate": 1.0082817619220968e-06, + "loss": 1.1817, + "step": 36525 + }, + { + "epoch": 6.64, + "learning_rate": 1.0082271568984347e-06, + "loss": 0.1502, + "step": 36526 + }, + { + "epoch": 6.64, + "learning_rate": 1.0081725518747724e-06, + "loss": 0.173, + "step": 36527 + }, + { + "epoch": 6.64, + "learning_rate": 1.0081179468511104e-06, + "loss": 2.5574, + "step": 36528 + }, + { + "epoch": 6.64, + "learning_rate": 1.008063341827448e-06, + "loss": 2.9397, + "step": 36529 + }, + { + "epoch": 6.64, + "learning_rate": 1.008008736803786e-06, + "loss": 0.9993, + "step": 36530 + }, + { + "epoch": 6.64, + "learning_rate": 1.0079541317801238e-06, + "loss": 1.201, + "step": 36531 + }, + { + "epoch": 6.64, + "learning_rate": 1.0078995267564615e-06, + "loss": 1.5914, + "step": 36532 + }, + { + "epoch": 6.64, + "learning_rate": 1.0078449217327995e-06, + "loss": 0.054, + "step": 36533 + }, + { + "epoch": 6.64, + "learning_rate": 1.0077903167091374e-06, + "loss": 0.0834, + "step": 36534 + }, + { + "epoch": 6.64, + "learning_rate": 1.0077357116854751e-06, + "loss": 0.2306, + "step": 36535 + }, + { + "epoch": 6.64, + "learning_rate": 1.0076811066618129e-06, + "loss": 1.6164, + "step": 36536 + }, + { + "epoch": 6.64, + "learning_rate": 1.0076265016381508e-06, + "loss": 0.0109, + "step": 36537 + }, + { + "epoch": 6.64, + "learning_rate": 1.0075718966144885e-06, + "loss": 1.4361, + "step": 36538 + }, + { + "epoch": 6.64, + "learning_rate": 1.0075172915908263e-06, + "loss": 1.5759, + "step": 36539 + }, + { + "epoch": 6.64, + "learning_rate": 1.0074626865671642e-06, + "loss": 0.015, + "step": 36540 + }, + { + "epoch": 6.65, + "learning_rate": 1.0074080815435022e-06, + "loss": 0.2993, + "step": 36541 + }, + { + "epoch": 6.65, + "learning_rate": 1.0073534765198399e-06, + "loss": 2.3412, + "step": 36542 + }, + { + "epoch": 6.65, + "learning_rate": 1.0072988714961776e-06, + "loss": 1.7361, + "step": 36543 + }, + { + "epoch": 6.65, + "learning_rate": 1.0072442664725156e-06, + "loss": 0.0048, + "step": 36544 + }, + { + "epoch": 6.65, + "learning_rate": 1.0071896614488533e-06, + "loss": 0.3677, + "step": 36545 + }, + { + "epoch": 6.65, + "learning_rate": 1.007135056425191e-06, + "loss": 0.0325, + "step": 36546 + }, + { + "epoch": 6.65, + "learning_rate": 1.007080451401529e-06, + "loss": 0.0655, + "step": 36547 + }, + { + "epoch": 6.65, + "learning_rate": 1.007025846377867e-06, + "loss": 0.8895, + "step": 36548 + }, + { + "epoch": 6.65, + "learning_rate": 1.0069712413542046e-06, + "loss": 0.2701, + "step": 36549 + }, + { + "epoch": 6.65, + "learning_rate": 1.0069166363305424e-06, + "loss": 0.0999, + "step": 36550 + }, + { + "epoch": 6.65, + "learning_rate": 1.0068620313068803e-06, + "loss": 0.2068, + "step": 36551 + }, + { + "epoch": 6.65, + "learning_rate": 1.006807426283218e-06, + "loss": 0.1714, + "step": 36552 + }, + { + "epoch": 6.65, + "learning_rate": 1.0067528212595558e-06, + "loss": 1.1692, + "step": 36553 + }, + { + "epoch": 6.65, + "learning_rate": 1.0066982162358937e-06, + "loss": 0.042, + "step": 36554 + }, + { + "epoch": 6.65, + "learning_rate": 1.0066436112122317e-06, + "loss": 0.1154, + "step": 36555 + }, + { + "epoch": 6.65, + "learning_rate": 1.0065890061885694e-06, + "loss": 0.0319, + "step": 36556 + }, + { + "epoch": 6.65, + "learning_rate": 1.0065344011649071e-06, + "loss": 0.3991, + "step": 36557 + }, + { + "epoch": 6.65, + "learning_rate": 1.006479796141245e-06, + "loss": 0.2663, + "step": 36558 + }, + { + "epoch": 6.65, + "learning_rate": 1.0064251911175828e-06, + "loss": 0.3656, + "step": 36559 + }, + { + "epoch": 6.65, + "learning_rate": 1.0063705860939205e-06, + "loss": 0.0049, + "step": 36560 + }, + { + "epoch": 6.65, + "learning_rate": 1.0063159810702585e-06, + "loss": 0.1872, + "step": 36561 + }, + { + "epoch": 6.65, + "learning_rate": 1.0062613760465964e-06, + "loss": 0.6821, + "step": 36562 + }, + { + "epoch": 6.65, + "learning_rate": 1.0062067710229342e-06, + "loss": 0.0629, + "step": 36563 + }, + { + "epoch": 6.65, + "learning_rate": 1.0061521659992719e-06, + "loss": 0.1157, + "step": 36564 + }, + { + "epoch": 6.65, + "learning_rate": 1.0060975609756098e-06, + "loss": 0.1994, + "step": 36565 + }, + { + "epoch": 6.65, + "learning_rate": 1.0060429559519476e-06, + "loss": 0.07, + "step": 36566 + }, + { + "epoch": 6.65, + "learning_rate": 1.0059883509282853e-06, + "loss": 0.438, + "step": 36567 + }, + { + "epoch": 6.65, + "learning_rate": 1.0059337459046232e-06, + "loss": 0.6972, + "step": 36568 + }, + { + "epoch": 6.65, + "learning_rate": 1.0058791408809612e-06, + "loss": 1.8879, + "step": 36569 + }, + { + "epoch": 6.65, + "learning_rate": 1.005824535857299e-06, + "loss": 0.9308, + "step": 36570 + }, + { + "epoch": 6.65, + "learning_rate": 1.0057699308336366e-06, + "loss": 1.553, + "step": 36571 + }, + { + "epoch": 6.65, + "learning_rate": 1.0057153258099746e-06, + "loss": 0.0883, + "step": 36572 + }, + { + "epoch": 6.65, + "learning_rate": 1.0056607207863123e-06, + "loss": 0.0102, + "step": 36573 + }, + { + "epoch": 6.65, + "learning_rate": 1.0056061157626503e-06, + "loss": 1.4216, + "step": 36574 + }, + { + "epoch": 6.65, + "learning_rate": 1.005551510738988e-06, + "loss": 0.0941, + "step": 36575 + }, + { + "epoch": 6.65, + "learning_rate": 1.005496905715326e-06, + "loss": 0.7589, + "step": 36576 + }, + { + "epoch": 6.65, + "learning_rate": 1.0054423006916637e-06, + "loss": 0.1671, + "step": 36577 + }, + { + "epoch": 6.65, + "learning_rate": 1.0053876956680014e-06, + "loss": 0.0902, + "step": 36578 + }, + { + "epoch": 6.65, + "learning_rate": 1.0053330906443393e-06, + "loss": 0.332, + "step": 36579 + }, + { + "epoch": 6.65, + "learning_rate": 1.005278485620677e-06, + "loss": 0.3599, + "step": 36580 + }, + { + "epoch": 6.65, + "learning_rate": 1.005223880597015e-06, + "loss": 2.461, + "step": 36581 + }, + { + "epoch": 6.65, + "learning_rate": 1.0051692755733527e-06, + "loss": 1.2748, + "step": 36582 + }, + { + "epoch": 6.65, + "learning_rate": 1.0051146705496907e-06, + "loss": 0.0096, + "step": 36583 + }, + { + "epoch": 6.65, + "learning_rate": 1.0050600655260284e-06, + "loss": 0.0109, + "step": 36584 + }, + { + "epoch": 6.65, + "learning_rate": 1.0050054605023662e-06, + "loss": 1.0728, + "step": 36585 + }, + { + "epoch": 6.65, + "learning_rate": 1.004950855478704e-06, + "loss": 1.0008, + "step": 36586 + }, + { + "epoch": 6.65, + "learning_rate": 1.0048962504550418e-06, + "loss": 1.7854, + "step": 36587 + }, + { + "epoch": 6.65, + "learning_rate": 1.0048416454313798e-06, + "loss": 0.0101, + "step": 36588 + }, + { + "epoch": 6.65, + "learning_rate": 1.0047870404077175e-06, + "loss": 0.6596, + "step": 36589 + }, + { + "epoch": 6.65, + "learning_rate": 1.0047324353840554e-06, + "loss": 1.1323, + "step": 36590 + }, + { + "epoch": 6.65, + "learning_rate": 1.0046778303603932e-06, + "loss": 0.4433, + "step": 36591 + }, + { + "epoch": 6.65, + "learning_rate": 1.004623225336731e-06, + "loss": 2.5151, + "step": 36592 + }, + { + "epoch": 6.65, + "learning_rate": 1.0045686203130689e-06, + "loss": 0.0059, + "step": 36593 + }, + { + "epoch": 6.65, + "learning_rate": 1.0045140152894066e-06, + "loss": 0.0445, + "step": 36594 + }, + { + "epoch": 6.65, + "learning_rate": 1.0044594102657445e-06, + "loss": 1.1312, + "step": 36595 + }, + { + "epoch": 6.66, + "learning_rate": 1.0044048052420823e-06, + "loss": 0.0267, + "step": 36596 + }, + { + "epoch": 6.66, + "learning_rate": 1.0043502002184202e-06, + "loss": 0.0107, + "step": 36597 + }, + { + "epoch": 6.66, + "learning_rate": 1.004295595194758e-06, + "loss": 1.1399, + "step": 36598 + }, + { + "epoch": 6.66, + "learning_rate": 1.0042409901710957e-06, + "loss": 0.4131, + "step": 36599 + }, + { + "epoch": 6.66, + "learning_rate": 1.0041863851474336e-06, + "loss": 1.6245, + "step": 36600 + }, + { + "epoch": 6.66, + "learning_rate": 1.0041317801237713e-06, + "loss": 0.0486, + "step": 36601 + }, + { + "epoch": 6.66, + "learning_rate": 1.0040771751001093e-06, + "loss": 2.4373, + "step": 36602 + }, + { + "epoch": 6.66, + "learning_rate": 1.004022570076447e-06, + "loss": 0.0147, + "step": 36603 + }, + { + "epoch": 6.66, + "learning_rate": 1.003967965052785e-06, + "loss": 0.0082, + "step": 36604 + }, + { + "epoch": 6.66, + "learning_rate": 1.0039133600291227e-06, + "loss": 0.0077, + "step": 36605 + }, + { + "epoch": 6.66, + "learning_rate": 1.0038587550054604e-06, + "loss": 0.3922, + "step": 36606 + }, + { + "epoch": 6.66, + "learning_rate": 1.0038041499817984e-06, + "loss": 0.004, + "step": 36607 + }, + { + "epoch": 6.66, + "learning_rate": 1.003749544958136e-06, + "loss": 0.493, + "step": 36608 + }, + { + "epoch": 6.66, + "learning_rate": 1.003694939934474e-06, + "loss": 0.0219, + "step": 36609 + }, + { + "epoch": 6.66, + "learning_rate": 1.0036403349108118e-06, + "loss": 0.0809, + "step": 36610 + }, + { + "epoch": 6.66, + "learning_rate": 1.0035857298871497e-06, + "loss": 2.3798, + "step": 36611 + }, + { + "epoch": 6.66, + "learning_rate": 1.0035311248634875e-06, + "loss": 1.8426, + "step": 36612 + }, + { + "epoch": 6.66, + "learning_rate": 1.0034765198398252e-06, + "loss": 0.2299, + "step": 36613 + }, + { + "epoch": 6.66, + "learning_rate": 1.0034219148161631e-06, + "loss": 0.0267, + "step": 36614 + }, + { + "epoch": 6.66, + "learning_rate": 1.003367309792501e-06, + "loss": 0.792, + "step": 36615 + }, + { + "epoch": 6.66, + "learning_rate": 1.0033127047688388e-06, + "loss": 0.0233, + "step": 36616 + }, + { + "epoch": 6.66, + "learning_rate": 1.0032580997451765e-06, + "loss": 0.0149, + "step": 36617 + }, + { + "epoch": 6.66, + "learning_rate": 1.0032034947215145e-06, + "loss": 0.1425, + "step": 36618 + }, + { + "epoch": 6.66, + "learning_rate": 1.0031488896978522e-06, + "loss": 1.3131, + "step": 36619 + }, + { + "epoch": 6.66, + "learning_rate": 1.00309428467419e-06, + "loss": 0.0649, + "step": 36620 + }, + { + "epoch": 6.66, + "learning_rate": 1.0030396796505279e-06, + "loss": 0.2812, + "step": 36621 + }, + { + "epoch": 6.66, + "learning_rate": 1.0029850746268658e-06, + "loss": 0.0782, + "step": 36622 + }, + { + "epoch": 6.66, + "learning_rate": 1.0029304696032036e-06, + "loss": 0.1004, + "step": 36623 + }, + { + "epoch": 6.66, + "learning_rate": 1.0028758645795413e-06, + "loss": 1.5489, + "step": 36624 + }, + { + "epoch": 6.66, + "learning_rate": 1.0028212595558792e-06, + "loss": 0.5325, + "step": 36625 + }, + { + "epoch": 6.66, + "learning_rate": 1.002766654532217e-06, + "loss": 0.0194, + "step": 36626 + }, + { + "epoch": 6.66, + "learning_rate": 1.0027120495085547e-06, + "loss": 0.1836, + "step": 36627 + }, + { + "epoch": 6.66, + "learning_rate": 1.0026574444848926e-06, + "loss": 0.1335, + "step": 36628 + }, + { + "epoch": 6.66, + "learning_rate": 1.0026028394612306e-06, + "loss": 0.2177, + "step": 36629 + }, + { + "epoch": 6.66, + "learning_rate": 1.0025482344375683e-06, + "loss": 0.0128, + "step": 36630 + }, + { + "epoch": 6.66, + "learning_rate": 1.002493629413906e-06, + "loss": 0.2536, + "step": 36631 + }, + { + "epoch": 6.66, + "learning_rate": 1.002439024390244e-06, + "loss": 0.1991, + "step": 36632 + }, + { + "epoch": 6.66, + "learning_rate": 1.0023844193665817e-06, + "loss": 0.6925, + "step": 36633 + }, + { + "epoch": 6.66, + "learning_rate": 1.0023298143429195e-06, + "loss": 0.1118, + "step": 36634 + }, + { + "epoch": 6.66, + "learning_rate": 1.0022752093192574e-06, + "loss": 1.2405, + "step": 36635 + }, + { + "epoch": 6.66, + "learning_rate": 1.0022206042955953e-06, + "loss": 0.7586, + "step": 36636 + }, + { + "epoch": 6.66, + "learning_rate": 1.002165999271933e-06, + "loss": 0.0208, + "step": 36637 + }, + { + "epoch": 6.66, + "learning_rate": 1.0021113942482708e-06, + "loss": 2.9149, + "step": 36638 + }, + { + "epoch": 6.66, + "learning_rate": 1.0020567892246087e-06, + "loss": 0.2923, + "step": 36639 + }, + { + "epoch": 6.66, + "learning_rate": 1.0020021842009465e-06, + "loss": 0.009, + "step": 36640 + }, + { + "epoch": 6.66, + "learning_rate": 1.0019475791772842e-06, + "loss": 0.7791, + "step": 36641 + }, + { + "epoch": 6.66, + "learning_rate": 1.0018929741536222e-06, + "loss": 0.1749, + "step": 36642 + }, + { + "epoch": 6.66, + "learning_rate": 1.00183836912996e-06, + "loss": 0.3307, + "step": 36643 + }, + { + "epoch": 6.66, + "learning_rate": 1.0017837641062978e-06, + "loss": 0.0802, + "step": 36644 + }, + { + "epoch": 6.66, + "learning_rate": 1.0017291590826356e-06, + "loss": 0.262, + "step": 36645 + }, + { + "epoch": 6.66, + "learning_rate": 1.0016745540589735e-06, + "loss": 0.0583, + "step": 36646 + }, + { + "epoch": 6.66, + "learning_rate": 1.0016199490353112e-06, + "loss": 0.2895, + "step": 36647 + }, + { + "epoch": 6.66, + "learning_rate": 1.001565344011649e-06, + "loss": 0.4756, + "step": 36648 + }, + { + "epoch": 6.66, + "learning_rate": 1.001510738987987e-06, + "loss": 1.2191, + "step": 36649 + }, + { + "epoch": 6.66, + "learning_rate": 1.0014561339643249e-06, + "loss": 0.2715, + "step": 36650 + }, + { + "epoch": 6.67, + "learning_rate": 1.0014015289406626e-06, + "loss": 0.0112, + "step": 36651 + }, + { + "epoch": 6.67, + "learning_rate": 1.0013469239170003e-06, + "loss": 0.1061, + "step": 36652 + }, + { + "epoch": 6.67, + "learning_rate": 1.0012923188933383e-06, + "loss": 0.0092, + "step": 36653 + }, + { + "epoch": 6.67, + "learning_rate": 1.001237713869676e-06, + "loss": 1.97, + "step": 36654 + }, + { + "epoch": 6.67, + "learning_rate": 1.001183108846014e-06, + "loss": 0.0434, + "step": 36655 + }, + { + "epoch": 6.67, + "learning_rate": 1.0011285038223517e-06, + "loss": 0.0809, + "step": 36656 + }, + { + "epoch": 6.67, + "learning_rate": 1.0010738987986896e-06, + "loss": 0.8646, + "step": 36657 + }, + { + "epoch": 6.67, + "learning_rate": 1.0010192937750273e-06, + "loss": 0.0154, + "step": 36658 + }, + { + "epoch": 6.67, + "learning_rate": 1.000964688751365e-06, + "loss": 0.0232, + "step": 36659 + }, + { + "epoch": 6.67, + "learning_rate": 1.000910083727703e-06, + "loss": 0.0142, + "step": 36660 + }, + { + "epoch": 6.67, + "learning_rate": 1.0008554787040407e-06, + "loss": 1.4022, + "step": 36661 + }, + { + "epoch": 6.67, + "learning_rate": 1.0008008736803787e-06, + "loss": 0.6384, + "step": 36662 + }, + { + "epoch": 6.67, + "learning_rate": 1.0007462686567164e-06, + "loss": 0.1794, + "step": 36663 + }, + { + "epoch": 6.67, + "learning_rate": 1.0006916636330544e-06, + "loss": 0.1301, + "step": 36664 + }, + { + "epoch": 6.67, + "learning_rate": 1.000637058609392e-06, + "loss": 1.8327, + "step": 36665 + }, + { + "epoch": 6.67, + "learning_rate": 1.0005824535857298e-06, + "loss": 0.5436, + "step": 36666 + }, + { + "epoch": 6.67, + "learning_rate": 1.0005278485620678e-06, + "loss": 0.5741, + "step": 36667 + }, + { + "epoch": 6.67, + "learning_rate": 1.0004732435384055e-06, + "loss": 0.0061, + "step": 36668 + }, + { + "epoch": 6.67, + "learning_rate": 1.0004186385147434e-06, + "loss": 0.2184, + "step": 36669 + }, + { + "epoch": 6.67, + "learning_rate": 1.0003640334910812e-06, + "loss": 0.454, + "step": 36670 + }, + { + "epoch": 6.67, + "learning_rate": 1.0003094284674191e-06, + "loss": 0.0615, + "step": 36671 + }, + { + "epoch": 6.67, + "learning_rate": 1.0002548234437569e-06, + "loss": 0.0396, + "step": 36672 + }, + { + "epoch": 6.67, + "learning_rate": 1.0002002184200946e-06, + "loss": 0.0385, + "step": 36673 + }, + { + "epoch": 6.67, + "learning_rate": 1.0001456133964325e-06, + "loss": 0.5615, + "step": 36674 + }, + { + "epoch": 6.67, + "learning_rate": 1.0000910083727703e-06, + "loss": 0.0059, + "step": 36675 + }, + { + "epoch": 6.67, + "learning_rate": 1.0000364033491082e-06, + "loss": 0.0588, + "step": 36676 + }, + { + "epoch": 6.67, + "learning_rate": 9.99981798325446e-07, + "loss": 1.1696, + "step": 36677 + }, + { + "epoch": 6.67, + "learning_rate": 9.999271933017839e-07, + "loss": 0.4195, + "step": 36678 + }, + { + "epoch": 6.67, + "learning_rate": 9.998725882781216e-07, + "loss": 1.4063, + "step": 36679 + }, + { + "epoch": 6.67, + "learning_rate": 9.998179832544593e-07, + "loss": 0.1232, + "step": 36680 + }, + { + "epoch": 6.67, + "learning_rate": 9.997633782307973e-07, + "loss": 0.1625, + "step": 36681 + }, + { + "epoch": 6.67, + "learning_rate": 9.99708773207135e-07, + "loss": 2.0103, + "step": 36682 + }, + { + "epoch": 6.67, + "learning_rate": 9.99654168183473e-07, + "loss": 0.0104, + "step": 36683 + }, + { + "epoch": 6.67, + "learning_rate": 9.995995631598107e-07, + "loss": 0.124, + "step": 36684 + }, + { + "epoch": 6.67, + "learning_rate": 9.995449581361486e-07, + "loss": 2.3016, + "step": 36685 + }, + { + "epoch": 6.67, + "learning_rate": 9.994903531124864e-07, + "loss": 0.1538, + "step": 36686 + }, + { + "epoch": 6.67, + "learning_rate": 9.99435748088824e-07, + "loss": 0.2774, + "step": 36687 + }, + { + "epoch": 6.67, + "learning_rate": 9.99381143065162e-07, + "loss": 1.0813, + "step": 36688 + }, + { + "epoch": 6.67, + "learning_rate": 9.993265380414998e-07, + "loss": 0.2034, + "step": 36689 + }, + { + "epoch": 6.67, + "learning_rate": 9.992719330178377e-07, + "loss": 0.0428, + "step": 36690 + }, + { + "epoch": 6.67, + "learning_rate": 9.992173279941754e-07, + "loss": 0.0302, + "step": 36691 + }, + { + "epoch": 6.67, + "learning_rate": 9.991627229705134e-07, + "loss": 0.0743, + "step": 36692 + }, + { + "epoch": 6.67, + "learning_rate": 9.991081179468511e-07, + "loss": 1.2099, + "step": 36693 + }, + { + "epoch": 6.67, + "learning_rate": 9.990535129231889e-07, + "loss": 0.2189, + "step": 36694 + }, + { + "epoch": 6.67, + "learning_rate": 9.989989078995268e-07, + "loss": 0.1194, + "step": 36695 + }, + { + "epoch": 6.67, + "learning_rate": 9.989443028758645e-07, + "loss": 0.1804, + "step": 36696 + }, + { + "epoch": 6.67, + "learning_rate": 9.988896978522025e-07, + "loss": 1.4292, + "step": 36697 + }, + { + "epoch": 6.67, + "learning_rate": 9.988350928285402e-07, + "loss": 0.1999, + "step": 36698 + }, + { + "epoch": 6.67, + "learning_rate": 9.987804878048781e-07, + "loss": 2.7542, + "step": 36699 + }, + { + "epoch": 6.67, + "learning_rate": 9.987258827812159e-07, + "loss": 0.0084, + "step": 36700 + }, + { + "epoch": 6.67, + "learning_rate": 9.986712777575536e-07, + "loss": 0.5988, + "step": 36701 + }, + { + "epoch": 6.67, + "learning_rate": 9.986166727338916e-07, + "loss": 1.0715, + "step": 36702 + }, + { + "epoch": 6.67, + "learning_rate": 9.985620677102295e-07, + "loss": 1.0884, + "step": 36703 + }, + { + "epoch": 6.67, + "learning_rate": 9.985074626865672e-07, + "loss": 0.2501, + "step": 36704 + }, + { + "epoch": 6.67, + "learning_rate": 9.98452857662905e-07, + "loss": 1.0961, + "step": 36705 + }, + { + "epoch": 6.68, + "learning_rate": 9.98398252639243e-07, + "loss": 1.0204, + "step": 36706 + }, + { + "epoch": 6.68, + "learning_rate": 9.983436476155806e-07, + "loss": 1.3256, + "step": 36707 + }, + { + "epoch": 6.68, + "learning_rate": 9.982890425919184e-07, + "loss": 0.1555, + "step": 36708 + }, + { + "epoch": 6.68, + "learning_rate": 9.982344375682563e-07, + "loss": 1.1579, + "step": 36709 + }, + { + "epoch": 6.68, + "learning_rate": 9.981798325445943e-07, + "loss": 1.2037, + "step": 36710 + }, + { + "epoch": 6.68, + "learning_rate": 9.98125227520932e-07, + "loss": 0.137, + "step": 36711 + }, + { + "epoch": 6.68, + "learning_rate": 9.980706224972697e-07, + "loss": 2.5186, + "step": 36712 + }, + { + "epoch": 6.68, + "learning_rate": 9.980160174736077e-07, + "loss": 0.0889, + "step": 36713 + }, + { + "epoch": 6.68, + "learning_rate": 9.979614124499454e-07, + "loss": 0.0378, + "step": 36714 + }, + { + "epoch": 6.68, + "learning_rate": 9.979068074262831e-07, + "loss": 0.0698, + "step": 36715 + }, + { + "epoch": 6.68, + "learning_rate": 9.97852202402621e-07, + "loss": 3.472, + "step": 36716 + }, + { + "epoch": 6.68, + "learning_rate": 9.97797597378959e-07, + "loss": 0.7164, + "step": 36717 + }, + { + "epoch": 6.68, + "learning_rate": 9.977429923552967e-07, + "loss": 0.008, + "step": 36718 + }, + { + "epoch": 6.68, + "learning_rate": 9.976883873316345e-07, + "loss": 2.5701, + "step": 36719 + }, + { + "epoch": 6.68, + "learning_rate": 9.976337823079724e-07, + "loss": 0.3759, + "step": 36720 + }, + { + "epoch": 6.68, + "learning_rate": 9.975791772843102e-07, + "loss": 0.454, + "step": 36721 + }, + { + "epoch": 6.68, + "learning_rate": 9.975245722606479e-07, + "loss": 0.8145, + "step": 36722 + }, + { + "epoch": 6.68, + "learning_rate": 9.974699672369858e-07, + "loss": 0.9401, + "step": 36723 + }, + { + "epoch": 6.68, + "learning_rate": 9.974153622133238e-07, + "loss": 0.1936, + "step": 36724 + }, + { + "epoch": 6.68, + "learning_rate": 9.973607571896615e-07, + "loss": 0.0809, + "step": 36725 + }, + { + "epoch": 6.68, + "learning_rate": 9.973061521659992e-07, + "loss": 0.7589, + "step": 36726 + }, + { + "epoch": 6.68, + "learning_rate": 9.972515471423372e-07, + "loss": 0.5043, + "step": 36727 + }, + { + "epoch": 6.68, + "learning_rate": 9.97196942118675e-07, + "loss": 0.111, + "step": 36728 + }, + { + "epoch": 6.68, + "learning_rate": 9.971423370950126e-07, + "loss": 2.0896, + "step": 36729 + }, + { + "epoch": 6.68, + "learning_rate": 9.970877320713506e-07, + "loss": 0.1506, + "step": 36730 + }, + { + "epoch": 6.68, + "learning_rate": 9.970331270476885e-07, + "loss": 0.6533, + "step": 36731 + }, + { + "epoch": 6.68, + "learning_rate": 9.969785220240263e-07, + "loss": 1.5059, + "step": 36732 + }, + { + "epoch": 6.68, + "learning_rate": 9.96923917000364e-07, + "loss": 0.0449, + "step": 36733 + }, + { + "epoch": 6.68, + "learning_rate": 9.96869311976702e-07, + "loss": 1.366, + "step": 36734 + }, + { + "epoch": 6.68, + "learning_rate": 9.968147069530397e-07, + "loss": 0.0046, + "step": 36735 + }, + { + "epoch": 6.68, + "learning_rate": 9.967601019293774e-07, + "loss": 0.0601, + "step": 36736 + }, + { + "epoch": 6.68, + "learning_rate": 9.967054969057153e-07, + "loss": 0.3764, + "step": 36737 + }, + { + "epoch": 6.68, + "learning_rate": 9.966508918820533e-07, + "loss": 0.018, + "step": 36738 + }, + { + "epoch": 6.68, + "learning_rate": 9.96596286858391e-07, + "loss": 0.0794, + "step": 36739 + }, + { + "epoch": 6.68, + "learning_rate": 9.965416818347287e-07, + "loss": 1.4748, + "step": 36740 + }, + { + "epoch": 6.68, + "learning_rate": 9.964870768110667e-07, + "loss": 1.106, + "step": 36741 + }, + { + "epoch": 6.68, + "learning_rate": 9.964324717874044e-07, + "loss": 0.0867, + "step": 36742 + }, + { + "epoch": 6.68, + "learning_rate": 9.963778667637424e-07, + "loss": 0.8868, + "step": 36743 + }, + { + "epoch": 6.68, + "learning_rate": 9.9632326174008e-07, + "loss": 0.0806, + "step": 36744 + }, + { + "epoch": 6.68, + "learning_rate": 9.96268656716418e-07, + "loss": 0.0079, + "step": 36745 + }, + { + "epoch": 6.68, + "learning_rate": 9.962140516927558e-07, + "loss": 2.6952, + "step": 36746 + }, + { + "epoch": 6.68, + "learning_rate": 9.961594466690935e-07, + "loss": 0.4958, + "step": 36747 + }, + { + "epoch": 6.68, + "learning_rate": 9.961048416454314e-07, + "loss": 0.6996, + "step": 36748 + }, + { + "epoch": 6.68, + "learning_rate": 9.960502366217692e-07, + "loss": 0.9234, + "step": 36749 + }, + { + "epoch": 6.68, + "learning_rate": 9.959956315981071e-07, + "loss": 0.9973, + "step": 36750 + }, + { + "epoch": 6.68, + "learning_rate": 9.959410265744449e-07, + "loss": 0.1579, + "step": 36751 + }, + { + "epoch": 6.68, + "learning_rate": 9.958864215507828e-07, + "loss": 0.6611, + "step": 36752 + }, + { + "epoch": 6.68, + "learning_rate": 9.958318165271205e-07, + "loss": 0.017, + "step": 36753 + }, + { + "epoch": 6.68, + "learning_rate": 9.957772115034583e-07, + "loss": 1.2356, + "step": 36754 + }, + { + "epoch": 6.68, + "learning_rate": 9.957226064797962e-07, + "loss": 0.7958, + "step": 36755 + }, + { + "epoch": 6.68, + "learning_rate": 9.95668001456134e-07, + "loss": 0.2925, + "step": 36756 + }, + { + "epoch": 6.68, + "learning_rate": 9.956133964324719e-07, + "loss": 0.0076, + "step": 36757 + }, + { + "epoch": 6.68, + "learning_rate": 9.955587914088096e-07, + "loss": 1.1194, + "step": 36758 + }, + { + "epoch": 6.68, + "learning_rate": 9.955041863851476e-07, + "loss": 1.1229, + "step": 36759 + }, + { + "epoch": 6.68, + "learning_rate": 9.954495813614853e-07, + "loss": 0.0132, + "step": 36760 + }, + { + "epoch": 6.69, + "learning_rate": 9.95394976337823e-07, + "loss": 2.0669, + "step": 36761 + }, + { + "epoch": 6.69, + "learning_rate": 9.95340371314161e-07, + "loss": 0.2109, + "step": 36762 + }, + { + "epoch": 6.69, + "learning_rate": 9.952857662904987e-07, + "loss": 0.1503, + "step": 36763 + }, + { + "epoch": 6.69, + "learning_rate": 9.952311612668366e-07, + "loss": 1.1821, + "step": 36764 + }, + { + "epoch": 6.69, + "learning_rate": 9.951765562431744e-07, + "loss": 0.0861, + "step": 36765 + }, + { + "epoch": 6.69, + "learning_rate": 9.951219512195123e-07, + "loss": 1.0923, + "step": 36766 + }, + { + "epoch": 6.69, + "learning_rate": 9.9506734619585e-07, + "loss": 1.5443, + "step": 36767 + }, + { + "epoch": 6.69, + "learning_rate": 9.950127411721878e-07, + "loss": 0.086, + "step": 36768 + }, + { + "epoch": 6.69, + "learning_rate": 9.949581361485257e-07, + "loss": 1.1724, + "step": 36769 + }, + { + "epoch": 6.69, + "learning_rate": 9.949035311248634e-07, + "loss": 0.0119, + "step": 36770 + }, + { + "epoch": 6.69, + "learning_rate": 9.948489261012014e-07, + "loss": 0.1571, + "step": 36771 + }, + { + "epoch": 6.69, + "learning_rate": 9.947943210775391e-07, + "loss": 0.1661, + "step": 36772 + }, + { + "epoch": 6.69, + "learning_rate": 9.94739716053877e-07, + "loss": 0.0085, + "step": 36773 + }, + { + "epoch": 6.69, + "learning_rate": 9.946851110302148e-07, + "loss": 0.485, + "step": 36774 + }, + { + "epoch": 6.69, + "learning_rate": 9.946305060065525e-07, + "loss": 0.0248, + "step": 36775 + }, + { + "epoch": 6.69, + "learning_rate": 9.945759009828905e-07, + "loss": 1.4743, + "step": 36776 + }, + { + "epoch": 6.69, + "learning_rate": 9.945212959592282e-07, + "loss": 1.0852, + "step": 36777 + }, + { + "epoch": 6.69, + "learning_rate": 9.944666909355661e-07, + "loss": 0.0766, + "step": 36778 + }, + { + "epoch": 6.69, + "learning_rate": 9.944120859119039e-07, + "loss": 0.4561, + "step": 36779 + }, + { + "epoch": 6.69, + "learning_rate": 9.943574808882418e-07, + "loss": 0.4327, + "step": 36780 + }, + { + "epoch": 6.69, + "learning_rate": 9.943028758645796e-07, + "loss": 0.0709, + "step": 36781 + }, + { + "epoch": 6.69, + "learning_rate": 9.942482708409173e-07, + "loss": 0.8513, + "step": 36782 + }, + { + "epoch": 6.69, + "learning_rate": 9.941936658172552e-07, + "loss": 0.2788, + "step": 36783 + }, + { + "epoch": 6.69, + "learning_rate": 9.941390607935932e-07, + "loss": 0.0172, + "step": 36784 + }, + { + "epoch": 6.69, + "learning_rate": 9.94084455769931e-07, + "loss": 0.2374, + "step": 36785 + }, + { + "epoch": 6.69, + "learning_rate": 9.940298507462686e-07, + "loss": 0.4942, + "step": 36786 + }, + { + "epoch": 6.69, + "learning_rate": 9.939752457226066e-07, + "loss": 0.8871, + "step": 36787 + }, + { + "epoch": 6.69, + "learning_rate": 9.939206406989443e-07, + "loss": 2.6318, + "step": 36788 + }, + { + "epoch": 6.69, + "learning_rate": 9.93866035675282e-07, + "loss": 0.1445, + "step": 36789 + }, + { + "epoch": 6.69, + "learning_rate": 9.9381143065162e-07, + "loss": 2.6683, + "step": 36790 + }, + { + "epoch": 6.69, + "learning_rate": 9.93756825627958e-07, + "loss": 0.0105, + "step": 36791 + }, + { + "epoch": 6.69, + "learning_rate": 9.937022206042957e-07, + "loss": 1.0624, + "step": 36792 + }, + { + "epoch": 6.69, + "learning_rate": 9.936476155806334e-07, + "loss": 1.5638, + "step": 36793 + }, + { + "epoch": 6.69, + "learning_rate": 9.935930105569713e-07, + "loss": 1.7173, + "step": 36794 + }, + { + "epoch": 6.69, + "learning_rate": 9.93538405533309e-07, + "loss": 0.9432, + "step": 36795 + }, + { + "epoch": 6.69, + "learning_rate": 9.934838005096468e-07, + "loss": 0.0841, + "step": 36796 + }, + { + "epoch": 6.69, + "learning_rate": 9.934291954859847e-07, + "loss": 0.0064, + "step": 36797 + }, + { + "epoch": 6.69, + "learning_rate": 9.933745904623227e-07, + "loss": 0.2589, + "step": 36798 + }, + { + "epoch": 6.69, + "learning_rate": 9.933199854386604e-07, + "loss": 0.4435, + "step": 36799 + }, + { + "epoch": 6.69, + "learning_rate": 9.932653804149982e-07, + "loss": 0.7357, + "step": 36800 + }, + { + "epoch": 6.69, + "learning_rate": 9.93210775391336e-07, + "loss": 0.0072, + "step": 36801 + }, + { + "epoch": 6.69, + "learning_rate": 9.931561703676738e-07, + "loss": 0.9278, + "step": 36802 + }, + { + "epoch": 6.69, + "learning_rate": 9.931015653440116e-07, + "loss": 0.0056, + "step": 36803 + }, + { + "epoch": 6.69, + "learning_rate": 9.930469603203495e-07, + "loss": 0.14, + "step": 36804 + }, + { + "epoch": 6.69, + "learning_rate": 9.929923552966874e-07, + "loss": 1.9285, + "step": 36805 + }, + { + "epoch": 6.69, + "learning_rate": 9.929377502730252e-07, + "loss": 0.7757, + "step": 36806 + }, + { + "epoch": 6.69, + "learning_rate": 9.92883145249363e-07, + "loss": 1.8822, + "step": 36807 + }, + { + "epoch": 6.69, + "learning_rate": 9.928285402257009e-07, + "loss": 0.2695, + "step": 36808 + }, + { + "epoch": 6.69, + "learning_rate": 9.927739352020386e-07, + "loss": 0.0117, + "step": 36809 + }, + { + "epoch": 6.69, + "learning_rate": 9.927193301783763e-07, + "loss": 0.0983, + "step": 36810 + }, + { + "epoch": 6.69, + "learning_rate": 9.926647251547143e-07, + "loss": 1.6037, + "step": 36811 + }, + { + "epoch": 6.69, + "learning_rate": 9.926101201310522e-07, + "loss": 0.2181, + "step": 36812 + }, + { + "epoch": 6.69, + "learning_rate": 9.9255551510739e-07, + "loss": 0.0067, + "step": 36813 + }, + { + "epoch": 6.69, + "learning_rate": 9.925009100837277e-07, + "loss": 0.6587, + "step": 36814 + }, + { + "epoch": 6.69, + "learning_rate": 9.924463050600656e-07, + "loss": 1.2277, + "step": 36815 + }, + { + "epoch": 6.7, + "learning_rate": 9.923917000364033e-07, + "loss": 0.0051, + "step": 36816 + }, + { + "epoch": 6.7, + "learning_rate": 9.92337095012741e-07, + "loss": 0.1852, + "step": 36817 + }, + { + "epoch": 6.7, + "learning_rate": 9.92282489989079e-07, + "loss": 0.3237, + "step": 36818 + }, + { + "epoch": 6.7, + "learning_rate": 9.92227884965417e-07, + "loss": 0.1223, + "step": 36819 + }, + { + "epoch": 6.7, + "learning_rate": 9.921732799417547e-07, + "loss": 0.1571, + "step": 36820 + }, + { + "epoch": 6.7, + "learning_rate": 9.921186749180924e-07, + "loss": 0.2653, + "step": 36821 + }, + { + "epoch": 6.7, + "learning_rate": 9.920640698944304e-07, + "loss": 0.6783, + "step": 36822 + }, + { + "epoch": 6.7, + "learning_rate": 9.92009464870768e-07, + "loss": 0.4476, + "step": 36823 + }, + { + "epoch": 6.7, + "learning_rate": 9.919548598471058e-07, + "loss": 0.1424, + "step": 36824 + }, + { + "epoch": 6.7, + "learning_rate": 9.919002548234438e-07, + "loss": 0.3762, + "step": 36825 + }, + { + "epoch": 6.7, + "learning_rate": 9.918456497997817e-07, + "loss": 1.0726, + "step": 36826 + }, + { + "epoch": 6.7, + "learning_rate": 9.917910447761194e-07, + "loss": 0.5701, + "step": 36827 + }, + { + "epoch": 6.7, + "learning_rate": 9.917364397524572e-07, + "loss": 0.0115, + "step": 36828 + }, + { + "epoch": 6.7, + "learning_rate": 9.916818347287951e-07, + "loss": 1.1864, + "step": 36829 + }, + { + "epoch": 6.7, + "learning_rate": 9.916272297051329e-07, + "loss": 0.0673, + "step": 36830 + }, + { + "epoch": 6.7, + "learning_rate": 9.915726246814708e-07, + "loss": 1.7464, + "step": 36831 + }, + { + "epoch": 6.7, + "learning_rate": 9.915180196578085e-07, + "loss": 0.1396, + "step": 36832 + }, + { + "epoch": 6.7, + "learning_rate": 9.914634146341465e-07, + "loss": 0.1856, + "step": 36833 + }, + { + "epoch": 6.7, + "learning_rate": 9.914088096104842e-07, + "loss": 0.1913, + "step": 36834 + }, + { + "epoch": 6.7, + "learning_rate": 9.91354204586822e-07, + "loss": 0.3587, + "step": 36835 + }, + { + "epoch": 6.7, + "learning_rate": 9.912995995631599e-07, + "loss": 5.4902, + "step": 36836 + }, + { + "epoch": 6.7, + "learning_rate": 9.912449945394976e-07, + "loss": 0.0081, + "step": 36837 + }, + { + "epoch": 6.7, + "learning_rate": 9.911903895158356e-07, + "loss": 0.3627, + "step": 36838 + }, + { + "epoch": 6.7, + "learning_rate": 9.911357844921733e-07, + "loss": 0.1262, + "step": 36839 + }, + { + "epoch": 6.7, + "learning_rate": 9.910811794685112e-07, + "loss": 0.3118, + "step": 36840 + }, + { + "epoch": 6.7, + "learning_rate": 9.91026574444849e-07, + "loss": 2.051, + "step": 36841 + }, + { + "epoch": 6.7, + "learning_rate": 9.909719694211867e-07, + "loss": 0.0099, + "step": 36842 + }, + { + "epoch": 6.7, + "learning_rate": 9.909173643975246e-07, + "loss": 0.0415, + "step": 36843 + }, + { + "epoch": 6.7, + "learning_rate": 9.908627593738624e-07, + "loss": 0.5125, + "step": 36844 + }, + { + "epoch": 6.7, + "learning_rate": 9.908081543502003e-07, + "loss": 0.2073, + "step": 36845 + }, + { + "epoch": 6.7, + "learning_rate": 9.90753549326538e-07, + "loss": 0.3126, + "step": 36846 + }, + { + "epoch": 6.7, + "learning_rate": 9.90698944302876e-07, + "loss": 0.6675, + "step": 36847 + }, + { + "epoch": 6.7, + "learning_rate": 9.906443392792137e-07, + "loss": 0.3511, + "step": 36848 + }, + { + "epoch": 6.7, + "learning_rate": 9.905897342555514e-07, + "loss": 0.1662, + "step": 36849 + }, + { + "epoch": 6.7, + "learning_rate": 9.905351292318894e-07, + "loss": 0.1219, + "step": 36850 + }, + { + "epoch": 6.7, + "learning_rate": 9.904805242082271e-07, + "loss": 0.897, + "step": 36851 + }, + { + "epoch": 6.7, + "learning_rate": 9.90425919184565e-07, + "loss": 0.2476, + "step": 36852 + }, + { + "epoch": 6.7, + "learning_rate": 9.903713141609028e-07, + "loss": 0.4065, + "step": 36853 + }, + { + "epoch": 6.7, + "learning_rate": 9.903167091372407e-07, + "loss": 0.2643, + "step": 36854 + }, + { + "epoch": 6.7, + "learning_rate": 9.902621041135785e-07, + "loss": 0.0654, + "step": 36855 + }, + { + "epoch": 6.7, + "learning_rate": 9.902074990899162e-07, + "loss": 0.0188, + "step": 36856 + }, + { + "epoch": 6.7, + "learning_rate": 9.901528940662541e-07, + "loss": 0.3991, + "step": 36857 + }, + { + "epoch": 6.7, + "learning_rate": 9.900982890425919e-07, + "loss": 1.0073, + "step": 36858 + }, + { + "epoch": 6.7, + "learning_rate": 9.900436840189298e-07, + "loss": 0.292, + "step": 36859 + }, + { + "epoch": 6.7, + "learning_rate": 9.899890789952676e-07, + "loss": 1.1654, + "step": 36860 + }, + { + "epoch": 6.7, + "learning_rate": 9.899344739716055e-07, + "loss": 0.3626, + "step": 36861 + }, + { + "epoch": 6.7, + "learning_rate": 9.898798689479432e-07, + "loss": 0.807, + "step": 36862 + }, + { + "epoch": 6.7, + "learning_rate": 9.89825263924281e-07, + "loss": 0.3016, + "step": 36863 + }, + { + "epoch": 6.7, + "learning_rate": 9.89770658900619e-07, + "loss": 0.0491, + "step": 36864 + }, + { + "epoch": 6.7, + "learning_rate": 9.897160538769566e-07, + "loss": 0.629, + "step": 36865 + }, + { + "epoch": 6.7, + "learning_rate": 9.896614488532946e-07, + "loss": 0.0103, + "step": 36866 + }, + { + "epoch": 6.7, + "learning_rate": 9.896068438296323e-07, + "loss": 0.1744, + "step": 36867 + }, + { + "epoch": 6.7, + "learning_rate": 9.895522388059703e-07, + "loss": 1.0533, + "step": 36868 + }, + { + "epoch": 6.7, + "learning_rate": 9.89497633782308e-07, + "loss": 0.2436, + "step": 36869 + }, + { + "epoch": 6.7, + "learning_rate": 9.894430287586457e-07, + "loss": 0.0426, + "step": 36870 + }, + { + "epoch": 6.71, + "learning_rate": 9.893884237349837e-07, + "loss": 1.6566, + "step": 36871 + }, + { + "epoch": 6.71, + "learning_rate": 9.893338187113216e-07, + "loss": 2.791, + "step": 36872 + }, + { + "epoch": 6.71, + "learning_rate": 9.892792136876593e-07, + "loss": 0.1682, + "step": 36873 + }, + { + "epoch": 6.71, + "learning_rate": 9.89224608663997e-07, + "loss": 0.5777, + "step": 36874 + }, + { + "epoch": 6.71, + "learning_rate": 9.89170003640335e-07, + "loss": 0.1576, + "step": 36875 + }, + { + "epoch": 6.71, + "learning_rate": 9.891153986166727e-07, + "loss": 0.0698, + "step": 36876 + }, + { + "epoch": 6.71, + "learning_rate": 9.890607935930105e-07, + "loss": 0.8002, + "step": 36877 + }, + { + "epoch": 6.71, + "learning_rate": 9.890061885693484e-07, + "loss": 0.0631, + "step": 36878 + }, + { + "epoch": 6.71, + "learning_rate": 9.889515835456864e-07, + "loss": 0.0119, + "step": 36879 + }, + { + "epoch": 6.71, + "learning_rate": 9.88896978522024e-07, + "loss": 1.3555, + "step": 36880 + }, + { + "epoch": 6.71, + "learning_rate": 9.888423734983618e-07, + "loss": 0.1369, + "step": 36881 + }, + { + "epoch": 6.71, + "learning_rate": 9.887877684746998e-07, + "loss": 1.351, + "step": 36882 + }, + { + "epoch": 6.71, + "learning_rate": 9.887331634510375e-07, + "loss": 2.3801, + "step": 36883 + }, + { + "epoch": 6.71, + "learning_rate": 9.886785584273752e-07, + "loss": 0.0869, + "step": 36884 + }, + { + "epoch": 6.71, + "learning_rate": 9.886239534037132e-07, + "loss": 0.0827, + "step": 36885 + }, + { + "epoch": 6.71, + "learning_rate": 9.885693483800511e-07, + "loss": 0.01, + "step": 36886 + }, + { + "epoch": 6.71, + "learning_rate": 9.885147433563888e-07, + "loss": 1.3384, + "step": 36887 + }, + { + "epoch": 6.71, + "learning_rate": 9.884601383327266e-07, + "loss": 0.2333, + "step": 36888 + }, + { + "epoch": 6.71, + "learning_rate": 9.884055333090645e-07, + "loss": 0.0464, + "step": 36889 + }, + { + "epoch": 6.71, + "learning_rate": 9.883509282854023e-07, + "loss": 0.4194, + "step": 36890 + }, + { + "epoch": 6.71, + "learning_rate": 9.8829632326174e-07, + "loss": 0.7505, + "step": 36891 + }, + { + "epoch": 6.71, + "learning_rate": 9.88241718238078e-07, + "loss": 0.7888, + "step": 36892 + }, + { + "epoch": 6.71, + "learning_rate": 9.881871132144159e-07, + "loss": 0.0393, + "step": 36893 + }, + { + "epoch": 6.71, + "learning_rate": 9.881325081907536e-07, + "loss": 0.0046, + "step": 36894 + }, + { + "epoch": 6.71, + "learning_rate": 9.880779031670913e-07, + "loss": 1.5898, + "step": 36895 + }, + { + "epoch": 6.71, + "learning_rate": 9.880232981434293e-07, + "loss": 0.1364, + "step": 36896 + }, + { + "epoch": 6.71, + "learning_rate": 9.87968693119767e-07, + "loss": 0.1655, + "step": 36897 + }, + { + "epoch": 6.71, + "learning_rate": 9.879140880961047e-07, + "loss": 0.0057, + "step": 36898 + }, + { + "epoch": 6.71, + "learning_rate": 9.878594830724427e-07, + "loss": 0.3134, + "step": 36899 + }, + { + "epoch": 6.71, + "learning_rate": 9.878048780487806e-07, + "loss": 0.0045, + "step": 36900 + }, + { + "epoch": 6.71, + "learning_rate": 9.877502730251184e-07, + "loss": 0.2719, + "step": 36901 + }, + { + "epoch": 6.71, + "learning_rate": 9.87695668001456e-07, + "loss": 1.5585, + "step": 36902 + }, + { + "epoch": 6.71, + "learning_rate": 9.87641062977794e-07, + "loss": 1.0878, + "step": 36903 + }, + { + "epoch": 6.71, + "learning_rate": 9.875864579541318e-07, + "loss": 2.1565, + "step": 36904 + }, + { + "epoch": 6.71, + "learning_rate": 9.875318529304695e-07, + "loss": 0.0076, + "step": 36905 + }, + { + "epoch": 6.71, + "learning_rate": 9.874772479068074e-07, + "loss": 0.2781, + "step": 36906 + }, + { + "epoch": 6.71, + "learning_rate": 9.874226428831454e-07, + "loss": 0.0774, + "step": 36907 + }, + { + "epoch": 6.71, + "learning_rate": 9.873680378594831e-07, + "loss": 0.0511, + "step": 36908 + }, + { + "epoch": 6.71, + "learning_rate": 9.873134328358209e-07, + "loss": 2.3632, + "step": 36909 + }, + { + "epoch": 6.71, + "learning_rate": 9.872588278121588e-07, + "loss": 2.0097, + "step": 36910 + }, + { + "epoch": 6.71, + "learning_rate": 9.872042227884965e-07, + "loss": 0.0281, + "step": 36911 + }, + { + "epoch": 6.71, + "learning_rate": 9.871496177648345e-07, + "loss": 0.5443, + "step": 36912 + }, + { + "epoch": 6.71, + "learning_rate": 9.870950127411722e-07, + "loss": 0.008, + "step": 36913 + }, + { + "epoch": 6.71, + "learning_rate": 9.870404077175101e-07, + "loss": 0.0926, + "step": 36914 + }, + { + "epoch": 6.71, + "learning_rate": 9.869858026938479e-07, + "loss": 0.0183, + "step": 36915 + }, + { + "epoch": 6.71, + "learning_rate": 9.869311976701856e-07, + "loss": 0.9701, + "step": 36916 + }, + { + "epoch": 6.71, + "learning_rate": 9.868765926465236e-07, + "loss": 0.1992, + "step": 36917 + }, + { + "epoch": 6.71, + "learning_rate": 9.868219876228613e-07, + "loss": 0.4026, + "step": 36918 + }, + { + "epoch": 6.71, + "learning_rate": 9.867673825991992e-07, + "loss": 0.2444, + "step": 36919 + }, + { + "epoch": 6.71, + "learning_rate": 9.86712777575537e-07, + "loss": 0.2387, + "step": 36920 + }, + { + "epoch": 6.71, + "learning_rate": 9.86658172551875e-07, + "loss": 0.5789, + "step": 36921 + }, + { + "epoch": 6.71, + "learning_rate": 9.866035675282126e-07, + "loss": 0.1092, + "step": 36922 + }, + { + "epoch": 6.71, + "learning_rate": 9.865489625045504e-07, + "loss": 1.4715, + "step": 36923 + }, + { + "epoch": 6.71, + "learning_rate": 9.864943574808883e-07, + "loss": 0.0258, + "step": 36924 + }, + { + "epoch": 6.71, + "learning_rate": 9.86439752457226e-07, + "loss": 1.5032, + "step": 36925 + }, + { + "epoch": 6.72, + "learning_rate": 9.86385147433564e-07, + "loss": 0.0572, + "step": 36926 + }, + { + "epoch": 6.72, + "learning_rate": 9.863305424099017e-07, + "loss": 0.567, + "step": 36927 + }, + { + "epoch": 6.72, + "learning_rate": 9.862759373862397e-07, + "loss": 1.0113, + "step": 36928 + }, + { + "epoch": 6.72, + "learning_rate": 9.862213323625774e-07, + "loss": 1.9765, + "step": 36929 + }, + { + "epoch": 6.72, + "learning_rate": 9.861667273389151e-07, + "loss": 0.5188, + "step": 36930 + }, + { + "epoch": 6.72, + "learning_rate": 9.86112122315253e-07, + "loss": 1.7502, + "step": 36931 + }, + { + "epoch": 6.72, + "learning_rate": 9.860575172915908e-07, + "loss": 0.1749, + "step": 36932 + }, + { + "epoch": 6.72, + "learning_rate": 9.860029122679287e-07, + "loss": 2.7491, + "step": 36933 + }, + { + "epoch": 6.72, + "learning_rate": 9.859483072442665e-07, + "loss": 1.2668, + "step": 36934 + }, + { + "epoch": 6.72, + "learning_rate": 9.858937022206044e-07, + "loss": 0.1384, + "step": 36935 + }, + { + "epoch": 6.72, + "learning_rate": 9.858390971969421e-07, + "loss": 1.3567, + "step": 36936 + }, + { + "epoch": 6.72, + "learning_rate": 9.857844921732799e-07, + "loss": 0.0972, + "step": 36937 + }, + { + "epoch": 6.72, + "learning_rate": 9.857298871496178e-07, + "loss": 0.342, + "step": 36938 + }, + { + "epoch": 6.72, + "learning_rate": 9.856752821259556e-07, + "loss": 0.0569, + "step": 36939 + }, + { + "epoch": 6.72, + "learning_rate": 9.856206771022935e-07, + "loss": 0.0071, + "step": 36940 + }, + { + "epoch": 6.72, + "learning_rate": 9.855660720786312e-07, + "loss": 0.7003, + "step": 36941 + }, + { + "epoch": 6.72, + "learning_rate": 9.855114670549692e-07, + "loss": 0.1191, + "step": 36942 + }, + { + "epoch": 6.72, + "learning_rate": 9.85456862031307e-07, + "loss": 1.5556, + "step": 36943 + }, + { + "epoch": 6.72, + "learning_rate": 9.854022570076446e-07, + "loss": 2.5432, + "step": 36944 + }, + { + "epoch": 6.72, + "learning_rate": 9.853476519839826e-07, + "loss": 0.7091, + "step": 36945 + }, + { + "epoch": 6.72, + "learning_rate": 9.852930469603203e-07, + "loss": 2.0627, + "step": 36946 + }, + { + "epoch": 6.72, + "learning_rate": 9.852384419366583e-07, + "loss": 0.2948, + "step": 36947 + }, + { + "epoch": 6.72, + "learning_rate": 9.85183836912996e-07, + "loss": 1.6938, + "step": 36948 + }, + { + "epoch": 6.72, + "learning_rate": 9.85129231889334e-07, + "loss": 0.1244, + "step": 36949 + }, + { + "epoch": 6.72, + "learning_rate": 9.850746268656717e-07, + "loss": 0.5962, + "step": 36950 + }, + { + "epoch": 6.72, + "learning_rate": 9.850200218420094e-07, + "loss": 0.1038, + "step": 36951 + }, + { + "epoch": 6.72, + "learning_rate": 9.849654168183473e-07, + "loss": 0.1002, + "step": 36952 + }, + { + "epoch": 6.72, + "learning_rate": 9.849108117946853e-07, + "loss": 1.0769, + "step": 36953 + }, + { + "epoch": 6.72, + "learning_rate": 9.84856206771023e-07, + "loss": 0.512, + "step": 36954 + }, + { + "epoch": 6.72, + "learning_rate": 9.848016017473607e-07, + "loss": 0.5224, + "step": 36955 + }, + { + "epoch": 6.72, + "learning_rate": 9.847469967236987e-07, + "loss": 0.1051, + "step": 36956 + }, + { + "epoch": 6.72, + "learning_rate": 9.846923917000364e-07, + "loss": 1.4792, + "step": 36957 + }, + { + "epoch": 6.72, + "learning_rate": 9.846377866763741e-07, + "loss": 0.0563, + "step": 36958 + }, + { + "epoch": 6.72, + "learning_rate": 9.84583181652712e-07, + "loss": 1.2529, + "step": 36959 + }, + { + "epoch": 6.72, + "learning_rate": 9.8452857662905e-07, + "loss": 0.2629, + "step": 36960 + }, + { + "epoch": 6.72, + "learning_rate": 9.844739716053878e-07, + "loss": 0.0997, + "step": 36961 + }, + { + "epoch": 6.72, + "learning_rate": 9.844193665817255e-07, + "loss": 0.4374, + "step": 36962 + }, + { + "epoch": 6.72, + "learning_rate": 9.843647615580634e-07, + "loss": 0.3971, + "step": 36963 + }, + { + "epoch": 6.72, + "learning_rate": 9.843101565344012e-07, + "loss": 0.7401, + "step": 36964 + }, + { + "epoch": 6.72, + "learning_rate": 9.84255551510739e-07, + "loss": 1.6251, + "step": 36965 + }, + { + "epoch": 6.72, + "learning_rate": 9.842009464870768e-07, + "loss": 0.7944, + "step": 36966 + }, + { + "epoch": 6.72, + "learning_rate": 9.841463414634148e-07, + "loss": 0.2687, + "step": 36967 + }, + { + "epoch": 6.72, + "learning_rate": 9.840917364397525e-07, + "loss": 0.0809, + "step": 36968 + }, + { + "epoch": 6.72, + "learning_rate": 9.840371314160903e-07, + "loss": 0.4092, + "step": 36969 + }, + { + "epoch": 6.72, + "learning_rate": 9.839825263924282e-07, + "loss": 0.8327, + "step": 36970 + }, + { + "epoch": 6.72, + "learning_rate": 9.83927921368766e-07, + "loss": 1.0312, + "step": 36971 + }, + { + "epoch": 6.72, + "learning_rate": 9.838733163451037e-07, + "loss": 0.0899, + "step": 36972 + }, + { + "epoch": 6.72, + "learning_rate": 9.838187113214416e-07, + "loss": 0.463, + "step": 36973 + }, + { + "epoch": 6.72, + "learning_rate": 9.837641062977795e-07, + "loss": 0.833, + "step": 36974 + }, + { + "epoch": 6.72, + "learning_rate": 9.837095012741173e-07, + "loss": 0.0464, + "step": 36975 + }, + { + "epoch": 6.72, + "learning_rate": 9.83654896250455e-07, + "loss": 0.0183, + "step": 36976 + }, + { + "epoch": 6.72, + "learning_rate": 9.83600291226793e-07, + "loss": 0.6943, + "step": 36977 + }, + { + "epoch": 6.72, + "learning_rate": 9.835456862031307e-07, + "loss": 0.1598, + "step": 36978 + }, + { + "epoch": 6.72, + "learning_rate": 9.834910811794684e-07, + "loss": 0.7077, + "step": 36979 + }, + { + "epoch": 6.72, + "learning_rate": 9.834364761558064e-07, + "loss": 0.0184, + "step": 36980 + }, + { + "epoch": 6.73, + "learning_rate": 9.833818711321443e-07, + "loss": 0.0215, + "step": 36981 + }, + { + "epoch": 6.73, + "learning_rate": 9.83327266108482e-07, + "loss": 0.0864, + "step": 36982 + }, + { + "epoch": 6.73, + "learning_rate": 9.832726610848198e-07, + "loss": 0.8258, + "step": 36983 + }, + { + "epoch": 6.73, + "learning_rate": 9.832180560611577e-07, + "loss": 2.4382, + "step": 36984 + }, + { + "epoch": 6.73, + "learning_rate": 9.831634510374954e-07, + "loss": 0.9063, + "step": 36985 + }, + { + "epoch": 6.73, + "learning_rate": 9.831088460138332e-07, + "loss": 1.5355, + "step": 36986 + }, + { + "epoch": 6.73, + "learning_rate": 9.830542409901711e-07, + "loss": 1.4162, + "step": 36987 + }, + { + "epoch": 6.73, + "learning_rate": 9.82999635966509e-07, + "loss": 0.0099, + "step": 36988 + }, + { + "epoch": 6.73, + "learning_rate": 9.829450309428468e-07, + "loss": 0.3621, + "step": 36989 + }, + { + "epoch": 6.73, + "learning_rate": 9.828904259191845e-07, + "loss": 0.2325, + "step": 36990 + }, + { + "epoch": 6.73, + "learning_rate": 9.828358208955225e-07, + "loss": 0.0397, + "step": 36991 + }, + { + "epoch": 6.73, + "learning_rate": 9.827812158718602e-07, + "loss": 0.4663, + "step": 36992 + }, + { + "epoch": 6.73, + "learning_rate": 9.82726610848198e-07, + "loss": 0.0773, + "step": 36993 + }, + { + "epoch": 6.73, + "learning_rate": 9.826720058245359e-07, + "loss": 0.0316, + "step": 36994 + }, + { + "epoch": 6.73, + "learning_rate": 9.826174008008738e-07, + "loss": 1.5444, + "step": 36995 + }, + { + "epoch": 6.73, + "learning_rate": 9.825627957772116e-07, + "loss": 0.4271, + "step": 36996 + }, + { + "epoch": 6.73, + "learning_rate": 9.825081907535493e-07, + "loss": 0.0924, + "step": 36997 + }, + { + "epoch": 6.73, + "learning_rate": 9.824535857298872e-07, + "loss": 0.179, + "step": 36998 + }, + { + "epoch": 6.73, + "learning_rate": 9.82398980706225e-07, + "loss": 0.0069, + "step": 36999 + }, + { + "epoch": 6.73, + "learning_rate": 9.82344375682563e-07, + "loss": 0.0074, + "step": 37000 + }, + { + "epoch": 6.73, + "learning_rate": 9.822897706589006e-07, + "loss": 0.0043, + "step": 37001 + }, + { + "epoch": 6.73, + "learning_rate": 9.822351656352386e-07, + "loss": 0.1766, + "step": 37002 + }, + { + "epoch": 6.73, + "learning_rate": 9.821805606115763e-07, + "loss": 0.0043, + "step": 37003 + }, + { + "epoch": 6.73, + "learning_rate": 9.82125955587914e-07, + "loss": 0.0981, + "step": 37004 + }, + { + "epoch": 6.73, + "learning_rate": 9.82071350564252e-07, + "loss": 0.0103, + "step": 37005 + }, + { + "epoch": 6.73, + "learning_rate": 9.820167455405897e-07, + "loss": 0.1723, + "step": 37006 + }, + { + "epoch": 6.73, + "learning_rate": 9.819621405169277e-07, + "loss": 0.0732, + "step": 37007 + }, + { + "epoch": 6.73, + "learning_rate": 9.819075354932654e-07, + "loss": 0.4342, + "step": 37008 + }, + { + "epoch": 6.73, + "learning_rate": 9.818529304696033e-07, + "loss": 2.5823, + "step": 37009 + }, + { + "epoch": 6.73, + "learning_rate": 9.81798325445941e-07, + "loss": 3.4993, + "step": 37010 + }, + { + "epoch": 6.73, + "learning_rate": 9.817437204222788e-07, + "loss": 1.4116, + "step": 37011 + }, + { + "epoch": 6.73, + "learning_rate": 9.816891153986167e-07, + "loss": 0.0111, + "step": 37012 + }, + { + "epoch": 6.73, + "learning_rate": 9.816345103749545e-07, + "loss": 0.0427, + "step": 37013 + }, + { + "epoch": 6.73, + "learning_rate": 9.815799053512924e-07, + "loss": 0.0855, + "step": 37014 + }, + { + "epoch": 6.73, + "learning_rate": 9.815253003276301e-07, + "loss": 0.7561, + "step": 37015 + }, + { + "epoch": 6.73, + "learning_rate": 9.81470695303968e-07, + "loss": 0.0769, + "step": 37016 + }, + { + "epoch": 6.73, + "learning_rate": 9.814160902803058e-07, + "loss": 0.0281, + "step": 37017 + }, + { + "epoch": 6.73, + "learning_rate": 9.813614852566436e-07, + "loss": 0.0063, + "step": 37018 + }, + { + "epoch": 6.73, + "learning_rate": 9.813068802329815e-07, + "loss": 1.9551, + "step": 37019 + }, + { + "epoch": 6.73, + "learning_rate": 9.812522752093192e-07, + "loss": 0.01, + "step": 37020 + }, + { + "epoch": 6.73, + "learning_rate": 9.811976701856572e-07, + "loss": 0.2424, + "step": 37021 + }, + { + "epoch": 6.73, + "learning_rate": 9.81143065161995e-07, + "loss": 0.5496, + "step": 37022 + }, + { + "epoch": 6.73, + "learning_rate": 9.810884601383328e-07, + "loss": 0.1673, + "step": 37023 + }, + { + "epoch": 6.73, + "learning_rate": 9.810338551146706e-07, + "loss": 0.7906, + "step": 37024 + }, + { + "epoch": 6.73, + "learning_rate": 9.809792500910083e-07, + "loss": 0.1789, + "step": 37025 + }, + { + "epoch": 6.73, + "learning_rate": 9.80924645067346e-07, + "loss": 0.6467, + "step": 37026 + }, + { + "epoch": 6.73, + "learning_rate": 9.80870040043684e-07, + "loss": 0.9618, + "step": 37027 + }, + { + "epoch": 6.73, + "learning_rate": 9.80815435020022e-07, + "loss": 2.3793, + "step": 37028 + }, + { + "epoch": 6.73, + "learning_rate": 9.807608299963597e-07, + "loss": 1.5357, + "step": 37029 + }, + { + "epoch": 6.73, + "learning_rate": 9.807062249726976e-07, + "loss": 0.4855, + "step": 37030 + }, + { + "epoch": 6.73, + "learning_rate": 9.806516199490353e-07, + "loss": 0.0036, + "step": 37031 + }, + { + "epoch": 6.73, + "learning_rate": 9.80597014925373e-07, + "loss": 0.0278, + "step": 37032 + }, + { + "epoch": 6.73, + "learning_rate": 9.805424099017108e-07, + "loss": 0.0035, + "step": 37033 + }, + { + "epoch": 6.73, + "learning_rate": 9.804878048780487e-07, + "loss": 1.3304, + "step": 37034 + }, + { + "epoch": 6.73, + "learning_rate": 9.804331998543867e-07, + "loss": 2.6547, + "step": 37035 + }, + { + "epoch": 6.74, + "learning_rate": 9.803785948307244e-07, + "loss": 0.7508, + "step": 37036 + }, + { + "epoch": 6.74, + "learning_rate": 9.803239898070624e-07, + "loss": 1.2345, + "step": 37037 + }, + { + "epoch": 6.74, + "learning_rate": 9.802693847834e-07, + "loss": 0.3262, + "step": 37038 + }, + { + "epoch": 6.74, + "learning_rate": 9.802147797597378e-07, + "loss": 0.3488, + "step": 37039 + }, + { + "epoch": 6.74, + "learning_rate": 9.801601747360758e-07, + "loss": 0.9028, + "step": 37040 + }, + { + "epoch": 6.74, + "learning_rate": 9.801055697124137e-07, + "loss": 0.7035, + "step": 37041 + }, + { + "epoch": 6.74, + "learning_rate": 9.800509646887514e-07, + "loss": 0.1009, + "step": 37042 + }, + { + "epoch": 6.74, + "learning_rate": 9.799963596650892e-07, + "loss": 0.2022, + "step": 37043 + }, + { + "epoch": 6.74, + "learning_rate": 9.799417546414271e-07, + "loss": 0.3892, + "step": 37044 + }, + { + "epoch": 6.74, + "learning_rate": 9.798871496177648e-07, + "loss": 0.1834, + "step": 37045 + }, + { + "epoch": 6.74, + "learning_rate": 9.798325445941026e-07, + "loss": 0.2505, + "step": 37046 + }, + { + "epoch": 6.74, + "learning_rate": 9.797779395704405e-07, + "loss": 0.8715, + "step": 37047 + }, + { + "epoch": 6.74, + "learning_rate": 9.797233345467785e-07, + "loss": 0.0104, + "step": 37048 + }, + { + "epoch": 6.74, + "learning_rate": 9.796687295231162e-07, + "loss": 1.0668, + "step": 37049 + }, + { + "epoch": 6.74, + "learning_rate": 9.79614124499454e-07, + "loss": 0.9937, + "step": 37050 + }, + { + "epoch": 6.74, + "learning_rate": 9.795595194757919e-07, + "loss": 0.2555, + "step": 37051 + }, + { + "epoch": 6.74, + "learning_rate": 9.795049144521296e-07, + "loss": 0.0206, + "step": 37052 + }, + { + "epoch": 6.74, + "learning_rate": 9.794503094284673e-07, + "loss": 0.5317, + "step": 37053 + }, + { + "epoch": 6.74, + "learning_rate": 9.793957044048053e-07, + "loss": 0.0396, + "step": 37054 + }, + { + "epoch": 6.74, + "learning_rate": 9.793410993811432e-07, + "loss": 3.3148, + "step": 37055 + }, + { + "epoch": 6.74, + "learning_rate": 9.79286494357481e-07, + "loss": 0.8315, + "step": 37056 + }, + { + "epoch": 6.74, + "learning_rate": 9.792318893338187e-07, + "loss": 0.1633, + "step": 37057 + }, + { + "epoch": 6.74, + "learning_rate": 9.791772843101566e-07, + "loss": 0.1154, + "step": 37058 + }, + { + "epoch": 6.74, + "learning_rate": 9.791226792864944e-07, + "loss": 0.0335, + "step": 37059 + }, + { + "epoch": 6.74, + "learning_rate": 9.79068074262832e-07, + "loss": 0.6016, + "step": 37060 + }, + { + "epoch": 6.74, + "learning_rate": 9.7901346923917e-07, + "loss": 0.0353, + "step": 37061 + }, + { + "epoch": 6.74, + "learning_rate": 9.78958864215508e-07, + "loss": 0.1123, + "step": 37062 + }, + { + "epoch": 6.74, + "learning_rate": 9.789042591918457e-07, + "loss": 0.2795, + "step": 37063 + }, + { + "epoch": 6.74, + "learning_rate": 9.788496541681834e-07, + "loss": 0.0907, + "step": 37064 + }, + { + "epoch": 6.74, + "learning_rate": 9.787950491445214e-07, + "loss": 0.1503, + "step": 37065 + }, + { + "epoch": 6.74, + "learning_rate": 9.787404441208591e-07, + "loss": 2.6271, + "step": 37066 + }, + { + "epoch": 6.74, + "learning_rate": 9.786858390971968e-07, + "loss": 1.1342, + "step": 37067 + }, + { + "epoch": 6.74, + "learning_rate": 9.786312340735348e-07, + "loss": 1.2914, + "step": 37068 + }, + { + "epoch": 6.74, + "learning_rate": 9.785766290498727e-07, + "loss": 0.096, + "step": 37069 + }, + { + "epoch": 6.74, + "learning_rate": 9.785220240262105e-07, + "loss": 0.01, + "step": 37070 + }, + { + "epoch": 6.74, + "learning_rate": 9.784674190025482e-07, + "loss": 0.0283, + "step": 37071 + }, + { + "epoch": 6.74, + "learning_rate": 9.784128139788861e-07, + "loss": 0.4014, + "step": 37072 + }, + { + "epoch": 6.74, + "learning_rate": 9.783582089552239e-07, + "loss": 0.1231, + "step": 37073 + }, + { + "epoch": 6.74, + "learning_rate": 9.783036039315616e-07, + "loss": 0.4834, + "step": 37074 + }, + { + "epoch": 6.74, + "learning_rate": 9.782489989078996e-07, + "loss": 0.1751, + "step": 37075 + }, + { + "epoch": 6.74, + "learning_rate": 9.781943938842375e-07, + "loss": 1.7626, + "step": 37076 + }, + { + "epoch": 6.74, + "learning_rate": 9.781397888605752e-07, + "loss": 0.5654, + "step": 37077 + }, + { + "epoch": 6.74, + "learning_rate": 9.78085183836913e-07, + "loss": 0.0124, + "step": 37078 + }, + { + "epoch": 6.74, + "learning_rate": 9.78030578813251e-07, + "loss": 0.0386, + "step": 37079 + }, + { + "epoch": 6.74, + "learning_rate": 9.779759737895886e-07, + "loss": 0.8089, + "step": 37080 + }, + { + "epoch": 6.74, + "learning_rate": 9.779213687659266e-07, + "loss": 0.59, + "step": 37081 + }, + { + "epoch": 6.74, + "learning_rate": 9.778667637422643e-07, + "loss": 0.6128, + "step": 37082 + }, + { + "epoch": 6.74, + "learning_rate": 9.778121587186023e-07, + "loss": 0.0216, + "step": 37083 + }, + { + "epoch": 6.74, + "learning_rate": 9.7775755369494e-07, + "loss": 0.0186, + "step": 37084 + }, + { + "epoch": 6.74, + "learning_rate": 9.777029486712777e-07, + "loss": 0.2564, + "step": 37085 + }, + { + "epoch": 6.74, + "learning_rate": 9.776483436476157e-07, + "loss": 1.5575, + "step": 37086 + }, + { + "epoch": 6.74, + "learning_rate": 9.775937386239534e-07, + "loss": 0.7782, + "step": 37087 + }, + { + "epoch": 6.74, + "learning_rate": 9.775391336002913e-07, + "loss": 0.0149, + "step": 37088 + }, + { + "epoch": 6.74, + "learning_rate": 9.77484528576629e-07, + "loss": 0.411, + "step": 37089 + }, + { + "epoch": 6.74, + "learning_rate": 9.77429923552967e-07, + "loss": 0.2513, + "step": 37090 + }, + { + "epoch": 6.75, + "learning_rate": 9.773753185293047e-07, + "loss": 1.0859, + "step": 37091 + }, + { + "epoch": 6.75, + "learning_rate": 9.773207135056425e-07, + "loss": 0.997, + "step": 37092 + }, + { + "epoch": 6.75, + "learning_rate": 9.772661084819804e-07, + "loss": 1.0375, + "step": 37093 + }, + { + "epoch": 6.75, + "learning_rate": 9.772115034583181e-07, + "loss": 0.117, + "step": 37094 + }, + { + "epoch": 6.75, + "learning_rate": 9.77156898434656e-07, + "loss": 0.0626, + "step": 37095 + }, + { + "epoch": 6.75, + "learning_rate": 9.771022934109938e-07, + "loss": 1.6154, + "step": 37096 + }, + { + "epoch": 6.75, + "learning_rate": 9.770476883873318e-07, + "loss": 0.7931, + "step": 37097 + }, + { + "epoch": 6.75, + "learning_rate": 9.769930833636695e-07, + "loss": 0.0136, + "step": 37098 + }, + { + "epoch": 6.75, + "learning_rate": 9.769384783400072e-07, + "loss": 0.2205, + "step": 37099 + }, + { + "epoch": 6.75, + "learning_rate": 9.768838733163452e-07, + "loss": 0.1913, + "step": 37100 + }, + { + "epoch": 6.75, + "learning_rate": 9.76829268292683e-07, + "loss": 1.6931, + "step": 37101 + }, + { + "epoch": 6.75, + "learning_rate": 9.767746632690208e-07, + "loss": 0.0149, + "step": 37102 + }, + { + "epoch": 6.75, + "learning_rate": 9.767200582453586e-07, + "loss": 0.2538, + "step": 37103 + }, + { + "epoch": 6.75, + "learning_rate": 9.766654532216965e-07, + "loss": 0.3588, + "step": 37104 + }, + { + "epoch": 6.75, + "learning_rate": 9.766108481980343e-07, + "loss": 0.5782, + "step": 37105 + }, + { + "epoch": 6.75, + "learning_rate": 9.76556243174372e-07, + "loss": 0.8287, + "step": 37106 + }, + { + "epoch": 6.75, + "learning_rate": 9.765016381507097e-07, + "loss": 0.0378, + "step": 37107 + }, + { + "epoch": 6.75, + "learning_rate": 9.764470331270477e-07, + "loss": 0.0596, + "step": 37108 + }, + { + "epoch": 6.75, + "learning_rate": 9.763924281033856e-07, + "loss": 1.0253, + "step": 37109 + }, + { + "epoch": 6.75, + "learning_rate": 9.763378230797233e-07, + "loss": 0.0873, + "step": 37110 + }, + { + "epoch": 6.75, + "learning_rate": 9.762832180560613e-07, + "loss": 1.3786, + "step": 37111 + }, + { + "epoch": 6.75, + "learning_rate": 9.76228613032399e-07, + "loss": 1.1749, + "step": 37112 + }, + { + "epoch": 6.75, + "learning_rate": 9.761740080087367e-07, + "loss": 0.2861, + "step": 37113 + }, + { + "epoch": 6.75, + "learning_rate": 9.761194029850745e-07, + "loss": 0.0133, + "step": 37114 + }, + { + "epoch": 6.75, + "learning_rate": 9.760647979614124e-07, + "loss": 0.0963, + "step": 37115 + }, + { + "epoch": 6.75, + "learning_rate": 9.760101929377504e-07, + "loss": 0.0619, + "step": 37116 + }, + { + "epoch": 6.75, + "learning_rate": 9.75955587914088e-07, + "loss": 0.0061, + "step": 37117 + }, + { + "epoch": 6.75, + "learning_rate": 9.75900982890426e-07, + "loss": 0.9126, + "step": 37118 + }, + { + "epoch": 6.75, + "learning_rate": 9.758463778667638e-07, + "loss": 0.2489, + "step": 37119 + }, + { + "epoch": 6.75, + "learning_rate": 9.757917728431015e-07, + "loss": 0.0082, + "step": 37120 + }, + { + "epoch": 6.75, + "learning_rate": 9.757371678194394e-07, + "loss": 2.5102, + "step": 37121 + }, + { + "epoch": 6.75, + "learning_rate": 9.756825627957772e-07, + "loss": 0.0785, + "step": 37122 + }, + { + "epoch": 6.75, + "learning_rate": 9.756279577721151e-07, + "loss": 0.9826, + "step": 37123 + }, + { + "epoch": 6.75, + "learning_rate": 9.755733527484528e-07, + "loss": 0.8828, + "step": 37124 + }, + { + "epoch": 6.75, + "learning_rate": 9.755187477247908e-07, + "loss": 0.8303, + "step": 37125 + }, + { + "epoch": 6.75, + "learning_rate": 9.754641427011285e-07, + "loss": 0.2062, + "step": 37126 + }, + { + "epoch": 6.75, + "learning_rate": 9.754095376774663e-07, + "loss": 0.0081, + "step": 37127 + }, + { + "epoch": 6.75, + "learning_rate": 9.753549326538042e-07, + "loss": 0.008, + "step": 37128 + }, + { + "epoch": 6.75, + "learning_rate": 9.753003276301421e-07, + "loss": 0.1537, + "step": 37129 + }, + { + "epoch": 6.75, + "learning_rate": 9.752457226064799e-07, + "loss": 0.2115, + "step": 37130 + }, + { + "epoch": 6.75, + "learning_rate": 9.751911175828176e-07, + "loss": 0.232, + "step": 37131 + }, + { + "epoch": 6.75, + "learning_rate": 9.751365125591555e-07, + "loss": 0.4153, + "step": 37132 + }, + { + "epoch": 6.75, + "learning_rate": 9.750819075354933e-07, + "loss": 1.6401, + "step": 37133 + }, + { + "epoch": 6.75, + "learning_rate": 9.75027302511831e-07, + "loss": 0.4962, + "step": 37134 + }, + { + "epoch": 6.75, + "learning_rate": 9.74972697488169e-07, + "loss": 0.0949, + "step": 37135 + }, + { + "epoch": 6.75, + "learning_rate": 9.74918092464507e-07, + "loss": 0.6437, + "step": 37136 + }, + { + "epoch": 6.75, + "learning_rate": 9.748634874408446e-07, + "loss": 0.1944, + "step": 37137 + }, + { + "epoch": 6.75, + "learning_rate": 9.748088824171824e-07, + "loss": 0.0331, + "step": 37138 + }, + { + "epoch": 6.75, + "learning_rate": 9.747542773935203e-07, + "loss": 0.2101, + "step": 37139 + }, + { + "epoch": 6.75, + "learning_rate": 9.74699672369858e-07, + "loss": 0.1918, + "step": 37140 + }, + { + "epoch": 6.75, + "learning_rate": 9.746450673461958e-07, + "loss": 0.1428, + "step": 37141 + }, + { + "epoch": 6.75, + "learning_rate": 9.745904623225337e-07, + "loss": 0.5487, + "step": 37142 + }, + { + "epoch": 6.75, + "learning_rate": 9.745358572988717e-07, + "loss": 0.0726, + "step": 37143 + }, + { + "epoch": 6.75, + "learning_rate": 9.744812522752094e-07, + "loss": 0.007, + "step": 37144 + }, + { + "epoch": 6.75, + "learning_rate": 9.744266472515471e-07, + "loss": 1.0352, + "step": 37145 + }, + { + "epoch": 6.76, + "learning_rate": 9.74372042227885e-07, + "loss": 0.1279, + "step": 37146 + }, + { + "epoch": 6.76, + "learning_rate": 9.743174372042228e-07, + "loss": 0.3024, + "step": 37147 + }, + { + "epoch": 6.76, + "learning_rate": 9.742628321805605e-07, + "loss": 0.1592, + "step": 37148 + }, + { + "epoch": 6.76, + "learning_rate": 9.742082271568985e-07, + "loss": 0.4085, + "step": 37149 + }, + { + "epoch": 6.76, + "learning_rate": 9.741536221332364e-07, + "loss": 1.955, + "step": 37150 + }, + { + "epoch": 6.76, + "learning_rate": 9.740990171095741e-07, + "loss": 0.5999, + "step": 37151 + }, + { + "epoch": 6.76, + "learning_rate": 9.740444120859119e-07, + "loss": 0.2792, + "step": 37152 + }, + { + "epoch": 6.76, + "learning_rate": 9.739898070622498e-07, + "loss": 0.7791, + "step": 37153 + }, + { + "epoch": 6.76, + "learning_rate": 9.739352020385875e-07, + "loss": 0.344, + "step": 37154 + }, + { + "epoch": 6.76, + "learning_rate": 9.738805970149253e-07, + "loss": 0.0304, + "step": 37155 + }, + { + "epoch": 6.76, + "learning_rate": 9.738259919912632e-07, + "loss": 0.0133, + "step": 37156 + }, + { + "epoch": 6.76, + "learning_rate": 9.737713869676012e-07, + "loss": 2.6192, + "step": 37157 + }, + { + "epoch": 6.76, + "learning_rate": 9.73716781943939e-07, + "loss": 0.1248, + "step": 37158 + }, + { + "epoch": 6.76, + "learning_rate": 9.736621769202766e-07, + "loss": 0.0766, + "step": 37159 + }, + { + "epoch": 6.76, + "learning_rate": 9.736075718966146e-07, + "loss": 1.662, + "step": 37160 + }, + { + "epoch": 6.76, + "learning_rate": 9.735529668729523e-07, + "loss": 0.0123, + "step": 37161 + }, + { + "epoch": 6.76, + "learning_rate": 9.7349836184929e-07, + "loss": 0.73, + "step": 37162 + }, + { + "epoch": 6.76, + "learning_rate": 9.73443756825628e-07, + "loss": 0.2062, + "step": 37163 + }, + { + "epoch": 6.76, + "learning_rate": 9.73389151801966e-07, + "loss": 0.0268, + "step": 37164 + }, + { + "epoch": 6.76, + "learning_rate": 9.733345467783037e-07, + "loss": 0.345, + "step": 37165 + }, + { + "epoch": 6.76, + "learning_rate": 9.732799417546414e-07, + "loss": 0.1979, + "step": 37166 + }, + { + "epoch": 6.76, + "learning_rate": 9.732253367309793e-07, + "loss": 0.486, + "step": 37167 + }, + { + "epoch": 6.76, + "learning_rate": 9.73170731707317e-07, + "loss": 0.0035, + "step": 37168 + }, + { + "epoch": 6.76, + "learning_rate": 9.73116126683655e-07, + "loss": 0.9578, + "step": 37169 + }, + { + "epoch": 6.76, + "learning_rate": 9.730615216599927e-07, + "loss": 0.1005, + "step": 37170 + }, + { + "epoch": 6.76, + "learning_rate": 9.730069166363307e-07, + "loss": 3.283, + "step": 37171 + }, + { + "epoch": 6.76, + "learning_rate": 9.729523116126684e-07, + "loss": 0.7633, + "step": 37172 + }, + { + "epoch": 6.76, + "learning_rate": 9.728977065890061e-07, + "loss": 0.01, + "step": 37173 + }, + { + "epoch": 6.76, + "learning_rate": 9.72843101565344e-07, + "loss": 1.1086, + "step": 37174 + }, + { + "epoch": 6.76, + "learning_rate": 9.727884965416818e-07, + "loss": 3.2082, + "step": 37175 + }, + { + "epoch": 6.76, + "learning_rate": 9.727338915180198e-07, + "loss": 0.2565, + "step": 37176 + }, + { + "epoch": 6.76, + "learning_rate": 9.726792864943575e-07, + "loss": 1.3938, + "step": 37177 + }, + { + "epoch": 6.76, + "learning_rate": 9.726246814706954e-07, + "loss": 0.1383, + "step": 37178 + }, + { + "epoch": 6.76, + "learning_rate": 9.725700764470332e-07, + "loss": 0.6479, + "step": 37179 + }, + { + "epoch": 6.76, + "learning_rate": 9.72515471423371e-07, + "loss": 1.4166, + "step": 37180 + }, + { + "epoch": 6.76, + "learning_rate": 9.724608663997088e-07, + "loss": 0.2777, + "step": 37181 + }, + { + "epoch": 6.76, + "learning_rate": 9.724062613760466e-07, + "loss": 0.133, + "step": 37182 + }, + { + "epoch": 6.76, + "learning_rate": 9.723516563523845e-07, + "loss": 0.0166, + "step": 37183 + }, + { + "epoch": 6.76, + "learning_rate": 9.722970513287223e-07, + "loss": 0.7825, + "step": 37184 + }, + { + "epoch": 6.76, + "learning_rate": 9.722424463050602e-07, + "loss": 0.1095, + "step": 37185 + }, + { + "epoch": 6.76, + "learning_rate": 9.72187841281398e-07, + "loss": 2.0692, + "step": 37186 + }, + { + "epoch": 6.76, + "learning_rate": 9.721332362577357e-07, + "loss": 0.0142, + "step": 37187 + }, + { + "epoch": 6.76, + "learning_rate": 9.720786312340734e-07, + "loss": 0.1745, + "step": 37188 + }, + { + "epoch": 6.76, + "learning_rate": 9.720240262104113e-07, + "loss": 0.1517, + "step": 37189 + }, + { + "epoch": 6.76, + "learning_rate": 9.719694211867493e-07, + "loss": 0.0637, + "step": 37190 + }, + { + "epoch": 6.76, + "learning_rate": 9.71914816163087e-07, + "loss": 0.1891, + "step": 37191 + }, + { + "epoch": 6.76, + "learning_rate": 9.71860211139425e-07, + "loss": 1.3499, + "step": 37192 + }, + { + "epoch": 6.76, + "learning_rate": 9.718056061157627e-07, + "loss": 0.1453, + "step": 37193 + }, + { + "epoch": 6.76, + "learning_rate": 9.717510010921004e-07, + "loss": 0.0152, + "step": 37194 + }, + { + "epoch": 6.76, + "learning_rate": 9.716963960684381e-07, + "loss": 0.0213, + "step": 37195 + }, + { + "epoch": 6.76, + "learning_rate": 9.71641791044776e-07, + "loss": 0.3596, + "step": 37196 + }, + { + "epoch": 6.76, + "learning_rate": 9.71587186021114e-07, + "loss": 0.2452, + "step": 37197 + }, + { + "epoch": 6.76, + "learning_rate": 9.715325809974518e-07, + "loss": 1.7588, + "step": 37198 + }, + { + "epoch": 6.76, + "learning_rate": 9.714779759737897e-07, + "loss": 1.5162, + "step": 37199 + }, + { + "epoch": 6.76, + "learning_rate": 9.714233709501274e-07, + "loss": 0.0073, + "step": 37200 + }, + { + "epoch": 6.77, + "learning_rate": 9.713687659264652e-07, + "loss": 0.5456, + "step": 37201 + }, + { + "epoch": 6.77, + "learning_rate": 9.71314160902803e-07, + "loss": 0.0499, + "step": 37202 + }, + { + "epoch": 6.77, + "learning_rate": 9.712595558791408e-07, + "loss": 1.409, + "step": 37203 + }, + { + "epoch": 6.77, + "learning_rate": 9.712049508554788e-07, + "loss": 0.8244, + "step": 37204 + }, + { + "epoch": 6.77, + "learning_rate": 9.711503458318165e-07, + "loss": 1.0664, + "step": 37205 + }, + { + "epoch": 6.77, + "learning_rate": 9.710957408081545e-07, + "loss": 0.9079, + "step": 37206 + }, + { + "epoch": 6.77, + "learning_rate": 9.710411357844922e-07, + "loss": 0.0481, + "step": 37207 + }, + { + "epoch": 6.77, + "learning_rate": 9.7098653076083e-07, + "loss": 0.2056, + "step": 37208 + }, + { + "epoch": 6.77, + "learning_rate": 9.709319257371679e-07, + "loss": 0.0693, + "step": 37209 + }, + { + "epoch": 6.77, + "learning_rate": 9.708773207135058e-07, + "loss": 0.3429, + "step": 37210 + }, + { + "epoch": 6.77, + "learning_rate": 9.708227156898435e-07, + "loss": 0.6025, + "step": 37211 + }, + { + "epoch": 6.77, + "learning_rate": 9.707681106661813e-07, + "loss": 0.0702, + "step": 37212 + }, + { + "epoch": 6.77, + "learning_rate": 9.707135056425192e-07, + "loss": 0.163, + "step": 37213 + }, + { + "epoch": 6.77, + "learning_rate": 9.70658900618857e-07, + "loss": 0.0116, + "step": 37214 + }, + { + "epoch": 6.77, + "learning_rate": 9.706042955951947e-07, + "loss": 1.3338, + "step": 37215 + }, + { + "epoch": 6.77, + "learning_rate": 9.705496905715326e-07, + "loss": 0.4222, + "step": 37216 + }, + { + "epoch": 6.77, + "learning_rate": 9.704950855478706e-07, + "loss": 0.4824, + "step": 37217 + }, + { + "epoch": 6.77, + "learning_rate": 9.704404805242083e-07, + "loss": 0.5751, + "step": 37218 + }, + { + "epoch": 6.77, + "learning_rate": 9.70385875500546e-07, + "loss": 0.7547, + "step": 37219 + }, + { + "epoch": 6.77, + "learning_rate": 9.70331270476884e-07, + "loss": 1.0341, + "step": 37220 + }, + { + "epoch": 6.77, + "learning_rate": 9.702766654532217e-07, + "loss": 0.2789, + "step": 37221 + }, + { + "epoch": 6.77, + "learning_rate": 9.702220604295594e-07, + "loss": 0.0185, + "step": 37222 + }, + { + "epoch": 6.77, + "learning_rate": 9.701674554058974e-07, + "loss": 0.1967, + "step": 37223 + }, + { + "epoch": 6.77, + "learning_rate": 9.701128503822353e-07, + "loss": 0.8596, + "step": 37224 + }, + { + "epoch": 6.77, + "learning_rate": 9.70058245358573e-07, + "loss": 0.0629, + "step": 37225 + }, + { + "epoch": 6.77, + "learning_rate": 9.700036403349108e-07, + "loss": 0.0671, + "step": 37226 + }, + { + "epoch": 6.77, + "learning_rate": 9.699490353112487e-07, + "loss": 3.1607, + "step": 37227 + }, + { + "epoch": 6.77, + "learning_rate": 9.698944302875865e-07, + "loss": 1.1152, + "step": 37228 + }, + { + "epoch": 6.77, + "learning_rate": 9.698398252639242e-07, + "loss": 0.5972, + "step": 37229 + }, + { + "epoch": 6.77, + "learning_rate": 9.697852202402621e-07, + "loss": 0.6011, + "step": 37230 + }, + { + "epoch": 6.77, + "learning_rate": 9.697306152166e-07, + "loss": 1.1977, + "step": 37231 + }, + { + "epoch": 6.77, + "learning_rate": 9.696760101929378e-07, + "loss": 0.0264, + "step": 37232 + }, + { + "epoch": 6.77, + "learning_rate": 9.696214051692755e-07, + "loss": 0.0414, + "step": 37233 + }, + { + "epoch": 6.77, + "learning_rate": 9.695668001456135e-07, + "loss": 0.0265, + "step": 37234 + }, + { + "epoch": 6.77, + "learning_rate": 9.695121951219512e-07, + "loss": 1.0696, + "step": 37235 + }, + { + "epoch": 6.77, + "learning_rate": 9.69457590098289e-07, + "loss": 1.2725, + "step": 37236 + }, + { + "epoch": 6.77, + "learning_rate": 9.69402985074627e-07, + "loss": 0.0207, + "step": 37237 + }, + { + "epoch": 6.77, + "learning_rate": 9.693483800509648e-07, + "loss": 0.3813, + "step": 37238 + }, + { + "epoch": 6.77, + "learning_rate": 9.692937750273026e-07, + "loss": 3.714, + "step": 37239 + }, + { + "epoch": 6.77, + "learning_rate": 9.692391700036403e-07, + "loss": 1.4537, + "step": 37240 + }, + { + "epoch": 6.77, + "learning_rate": 9.691845649799782e-07, + "loss": 0.2517, + "step": 37241 + }, + { + "epoch": 6.77, + "learning_rate": 9.69129959956316e-07, + "loss": 0.0065, + "step": 37242 + }, + { + "epoch": 6.77, + "learning_rate": 9.690753549326537e-07, + "loss": 3.2564, + "step": 37243 + }, + { + "epoch": 6.77, + "learning_rate": 9.690207499089917e-07, + "loss": 2.3176, + "step": 37244 + }, + { + "epoch": 6.77, + "learning_rate": 9.689661448853296e-07, + "loss": 0.3956, + "step": 37245 + }, + { + "epoch": 6.77, + "learning_rate": 9.689115398616673e-07, + "loss": 0.397, + "step": 37246 + }, + { + "epoch": 6.77, + "learning_rate": 9.68856934838005e-07, + "loss": 1.9815, + "step": 37247 + }, + { + "epoch": 6.77, + "learning_rate": 9.68802329814343e-07, + "loss": 0.4274, + "step": 37248 + }, + { + "epoch": 6.77, + "learning_rate": 9.687477247906807e-07, + "loss": 0.7784, + "step": 37249 + }, + { + "epoch": 6.77, + "learning_rate": 9.686931197670187e-07, + "loss": 1.5235, + "step": 37250 + }, + { + "epoch": 6.77, + "learning_rate": 9.686385147433564e-07, + "loss": 0.1398, + "step": 37251 + }, + { + "epoch": 6.77, + "learning_rate": 9.685839097196944e-07, + "loss": 0.9121, + "step": 37252 + }, + { + "epoch": 6.77, + "learning_rate": 9.68529304696032e-07, + "loss": 1.3684, + "step": 37253 + }, + { + "epoch": 6.77, + "learning_rate": 9.684746996723698e-07, + "loss": 0.6591, + "step": 37254 + }, + { + "epoch": 6.77, + "learning_rate": 9.684200946487078e-07, + "loss": 0.8727, + "step": 37255 + }, + { + "epoch": 6.78, + "learning_rate": 9.683654896250455e-07, + "loss": 0.0893, + "step": 37256 + }, + { + "epoch": 6.78, + "learning_rate": 9.683108846013834e-07, + "loss": 0.0066, + "step": 37257 + }, + { + "epoch": 6.78, + "learning_rate": 9.682562795777212e-07, + "loss": 0.6351, + "step": 37258 + }, + { + "epoch": 6.78, + "learning_rate": 9.682016745540591e-07, + "loss": 1.3167, + "step": 37259 + }, + { + "epoch": 6.78, + "learning_rate": 9.681470695303968e-07, + "loss": 0.6627, + "step": 37260 + }, + { + "epoch": 6.78, + "learning_rate": 9.680924645067346e-07, + "loss": 0.5897, + "step": 37261 + }, + { + "epoch": 6.78, + "learning_rate": 9.680378594830723e-07, + "loss": 0.4279, + "step": 37262 + }, + { + "epoch": 6.78, + "learning_rate": 9.679832544594103e-07, + "loss": 0.0222, + "step": 37263 + }, + { + "epoch": 6.78, + "learning_rate": 9.679286494357482e-07, + "loss": 2.521, + "step": 37264 + }, + { + "epoch": 6.78, + "learning_rate": 9.67874044412086e-07, + "loss": 0.0177, + "step": 37265 + }, + { + "epoch": 6.78, + "learning_rate": 9.678194393884239e-07, + "loss": 0.0658, + "step": 37266 + }, + { + "epoch": 6.78, + "learning_rate": 9.677648343647616e-07, + "loss": 0.0057, + "step": 37267 + }, + { + "epoch": 6.78, + "learning_rate": 9.677102293410993e-07, + "loss": 3.0517, + "step": 37268 + }, + { + "epoch": 6.78, + "learning_rate": 9.67655624317437e-07, + "loss": 2.3071, + "step": 37269 + }, + { + "epoch": 6.78, + "learning_rate": 9.67601019293775e-07, + "loss": 0.1228, + "step": 37270 + }, + { + "epoch": 6.78, + "learning_rate": 9.67546414270113e-07, + "loss": 0.5609, + "step": 37271 + }, + { + "epoch": 6.78, + "learning_rate": 9.674918092464507e-07, + "loss": 0.567, + "step": 37272 + }, + { + "epoch": 6.78, + "learning_rate": 9.674372042227886e-07, + "loss": 0.5331, + "step": 37273 + }, + { + "epoch": 6.78, + "learning_rate": 9.673825991991264e-07, + "loss": 0.16, + "step": 37274 + }, + { + "epoch": 6.78, + "learning_rate": 9.67327994175464e-07, + "loss": 1.0106, + "step": 37275 + }, + { + "epoch": 6.78, + "learning_rate": 9.672733891518018e-07, + "loss": 0.2706, + "step": 37276 + }, + { + "epoch": 6.78, + "learning_rate": 9.672187841281398e-07, + "loss": 0.2662, + "step": 37277 + }, + { + "epoch": 6.78, + "learning_rate": 9.671641791044777e-07, + "loss": 0.2026, + "step": 37278 + }, + { + "epoch": 6.78, + "learning_rate": 9.671095740808154e-07, + "loss": 0.5103, + "step": 37279 + }, + { + "epoch": 6.78, + "learning_rate": 9.670549690571534e-07, + "loss": 0.4034, + "step": 37280 + }, + { + "epoch": 6.78, + "learning_rate": 9.670003640334911e-07, + "loss": 0.9331, + "step": 37281 + }, + { + "epoch": 6.78, + "learning_rate": 9.669457590098288e-07, + "loss": 0.5208, + "step": 37282 + }, + { + "epoch": 6.78, + "learning_rate": 9.668911539861666e-07, + "loss": 1.4315, + "step": 37283 + }, + { + "epoch": 6.78, + "learning_rate": 9.668365489625045e-07, + "loss": 0.7833, + "step": 37284 + }, + { + "epoch": 6.78, + "learning_rate": 9.667819439388425e-07, + "loss": 0.0563, + "step": 37285 + }, + { + "epoch": 6.78, + "learning_rate": 9.667273389151802e-07, + "loss": 0.0261, + "step": 37286 + }, + { + "epoch": 6.78, + "learning_rate": 9.666727338915181e-07, + "loss": 0.0192, + "step": 37287 + }, + { + "epoch": 6.78, + "learning_rate": 9.666181288678559e-07, + "loss": 0.2259, + "step": 37288 + }, + { + "epoch": 6.78, + "learning_rate": 9.665635238441936e-07, + "loss": 0.9028, + "step": 37289 + }, + { + "epoch": 6.78, + "learning_rate": 9.665089188205313e-07, + "loss": 0.2451, + "step": 37290 + }, + { + "epoch": 6.78, + "learning_rate": 9.664543137968693e-07, + "loss": 0.1378, + "step": 37291 + }, + { + "epoch": 6.78, + "learning_rate": 9.663997087732072e-07, + "loss": 0.0628, + "step": 37292 + }, + { + "epoch": 6.78, + "learning_rate": 9.66345103749545e-07, + "loss": 0.0059, + "step": 37293 + }, + { + "epoch": 6.78, + "learning_rate": 9.66290498725883e-07, + "loss": 1.0253, + "step": 37294 + }, + { + "epoch": 6.78, + "learning_rate": 9.662358937022206e-07, + "loss": 0.9416, + "step": 37295 + }, + { + "epoch": 6.78, + "learning_rate": 9.661812886785584e-07, + "loss": 0.7332, + "step": 37296 + }, + { + "epoch": 6.78, + "learning_rate": 9.661266836548963e-07, + "loss": 1.4343, + "step": 37297 + }, + { + "epoch": 6.78, + "learning_rate": 9.660720786312342e-07, + "loss": 0.0077, + "step": 37298 + }, + { + "epoch": 6.78, + "learning_rate": 9.66017473607572e-07, + "loss": 0.0846, + "step": 37299 + }, + { + "epoch": 6.78, + "learning_rate": 9.659628685839097e-07, + "loss": 0.5614, + "step": 37300 + }, + { + "epoch": 6.78, + "learning_rate": 9.659082635602477e-07, + "loss": 0.2275, + "step": 37301 + }, + { + "epoch": 6.78, + "learning_rate": 9.658536585365854e-07, + "loss": 0.0646, + "step": 37302 + }, + { + "epoch": 6.78, + "learning_rate": 9.657990535129231e-07, + "loss": 2.5021, + "step": 37303 + }, + { + "epoch": 6.78, + "learning_rate": 9.65744448489261e-07, + "loss": 0.3371, + "step": 37304 + }, + { + "epoch": 6.78, + "learning_rate": 9.65689843465599e-07, + "loss": 1.3553, + "step": 37305 + }, + { + "epoch": 6.78, + "learning_rate": 9.656352384419367e-07, + "loss": 3.5406, + "step": 37306 + }, + { + "epoch": 6.78, + "learning_rate": 9.655806334182745e-07, + "loss": 0.0031, + "step": 37307 + }, + { + "epoch": 6.78, + "learning_rate": 9.655260283946124e-07, + "loss": 0.0904, + "step": 37308 + }, + { + "epoch": 6.78, + "learning_rate": 9.654714233709501e-07, + "loss": 0.1466, + "step": 37309 + }, + { + "epoch": 6.78, + "learning_rate": 9.654168183472879e-07, + "loss": 0.0107, + "step": 37310 + }, + { + "epoch": 6.79, + "learning_rate": 9.653622133236258e-07, + "loss": 0.7723, + "step": 37311 + }, + { + "epoch": 6.79, + "learning_rate": 9.653076082999638e-07, + "loss": 0.9032, + "step": 37312 + }, + { + "epoch": 6.79, + "learning_rate": 9.652530032763015e-07, + "loss": 3.9402, + "step": 37313 + }, + { + "epoch": 6.79, + "learning_rate": 9.651983982526392e-07, + "loss": 0.538, + "step": 37314 + }, + { + "epoch": 6.79, + "learning_rate": 9.651437932289772e-07, + "loss": 0.3791, + "step": 37315 + }, + { + "epoch": 6.79, + "learning_rate": 9.65089188205315e-07, + "loss": 0.5708, + "step": 37316 + }, + { + "epoch": 6.79, + "learning_rate": 9.650345831816526e-07, + "loss": 0.0719, + "step": 37317 + }, + { + "epoch": 6.79, + "learning_rate": 9.649799781579906e-07, + "loss": 5.3424, + "step": 37318 + }, + { + "epoch": 6.79, + "learning_rate": 9.649253731343285e-07, + "loss": 0.0066, + "step": 37319 + }, + { + "epoch": 6.79, + "learning_rate": 9.648707681106662e-07, + "loss": 0.005, + "step": 37320 + }, + { + "epoch": 6.79, + "learning_rate": 9.64816163087004e-07, + "loss": 0.262, + "step": 37321 + }, + { + "epoch": 6.79, + "learning_rate": 9.64761558063342e-07, + "loss": 2.9902, + "step": 37322 + }, + { + "epoch": 6.79, + "learning_rate": 9.647069530396797e-07, + "loss": 0.1438, + "step": 37323 + }, + { + "epoch": 6.79, + "learning_rate": 9.646523480160174e-07, + "loss": 0.0398, + "step": 37324 + }, + { + "epoch": 6.79, + "learning_rate": 9.645977429923553e-07, + "loss": 2.3746, + "step": 37325 + }, + { + "epoch": 6.79, + "learning_rate": 9.645431379686933e-07, + "loss": 0.0525, + "step": 37326 + }, + { + "epoch": 6.79, + "learning_rate": 9.64488532945031e-07, + "loss": 0.8568, + "step": 37327 + }, + { + "epoch": 6.79, + "learning_rate": 9.644339279213687e-07, + "loss": 0.0075, + "step": 37328 + }, + { + "epoch": 6.79, + "learning_rate": 9.643793228977067e-07, + "loss": 0.8146, + "step": 37329 + }, + { + "epoch": 6.79, + "learning_rate": 9.643247178740444e-07, + "loss": 0.2834, + "step": 37330 + }, + { + "epoch": 6.79, + "learning_rate": 9.642701128503821e-07, + "loss": 0.6872, + "step": 37331 + }, + { + "epoch": 6.79, + "learning_rate": 9.6421550782672e-07, + "loss": 0.8031, + "step": 37332 + }, + { + "epoch": 6.79, + "learning_rate": 9.64160902803058e-07, + "loss": 1.2517, + "step": 37333 + }, + { + "epoch": 6.79, + "learning_rate": 9.641062977793958e-07, + "loss": 0.4174, + "step": 37334 + }, + { + "epoch": 6.79, + "learning_rate": 9.640516927557335e-07, + "loss": 1.4909, + "step": 37335 + }, + { + "epoch": 6.79, + "learning_rate": 9.639970877320714e-07, + "loss": 0.0455, + "step": 37336 + }, + { + "epoch": 6.79, + "learning_rate": 9.639424827084092e-07, + "loss": 0.5401, + "step": 37337 + }, + { + "epoch": 6.79, + "learning_rate": 9.638878776847471e-07, + "loss": 2.1099, + "step": 37338 + }, + { + "epoch": 6.79, + "learning_rate": 9.638332726610848e-07, + "loss": 0.0513, + "step": 37339 + }, + { + "epoch": 6.79, + "learning_rate": 9.637786676374228e-07, + "loss": 1.3365, + "step": 37340 + }, + { + "epoch": 6.79, + "learning_rate": 9.637240626137605e-07, + "loss": 0.8041, + "step": 37341 + }, + { + "epoch": 6.79, + "learning_rate": 9.636694575900982e-07, + "loss": 0.2896, + "step": 37342 + }, + { + "epoch": 6.79, + "learning_rate": 9.63614852566436e-07, + "loss": 0.0675, + "step": 37343 + }, + { + "epoch": 6.79, + "learning_rate": 9.63560247542774e-07, + "loss": 1.0576, + "step": 37344 + }, + { + "epoch": 6.79, + "learning_rate": 9.635056425191119e-07, + "loss": 0.0098, + "step": 37345 + }, + { + "epoch": 6.79, + "learning_rate": 9.634510374954496e-07, + "loss": 0.5052, + "step": 37346 + }, + { + "epoch": 6.79, + "learning_rate": 9.633964324717875e-07, + "loss": 0.0392, + "step": 37347 + }, + { + "epoch": 6.79, + "learning_rate": 9.633418274481253e-07, + "loss": 2.2903, + "step": 37348 + }, + { + "epoch": 6.79, + "learning_rate": 9.63287222424463e-07, + "loss": 1.2353, + "step": 37349 + }, + { + "epoch": 6.79, + "learning_rate": 9.632326174008007e-07, + "loss": 2.2161, + "step": 37350 + }, + { + "epoch": 6.79, + "learning_rate": 9.631780123771387e-07, + "loss": 1.8046, + "step": 37351 + }, + { + "epoch": 6.79, + "learning_rate": 9.631234073534766e-07, + "loss": 0.7761, + "step": 37352 + }, + { + "epoch": 6.79, + "learning_rate": 9.630688023298144e-07, + "loss": 0.5332, + "step": 37353 + }, + { + "epoch": 6.79, + "learning_rate": 9.630141973061523e-07, + "loss": 0.3877, + "step": 37354 + }, + { + "epoch": 6.79, + "learning_rate": 9.6295959228249e-07, + "loss": 0.2012, + "step": 37355 + }, + { + "epoch": 6.79, + "learning_rate": 9.629049872588278e-07, + "loss": 0.1336, + "step": 37356 + }, + { + "epoch": 6.79, + "learning_rate": 9.628503822351655e-07, + "loss": 0.2914, + "step": 37357 + }, + { + "epoch": 6.79, + "learning_rate": 9.627957772115034e-07, + "loss": 0.0071, + "step": 37358 + }, + { + "epoch": 6.79, + "learning_rate": 9.627411721878414e-07, + "loss": 1.486, + "step": 37359 + }, + { + "epoch": 6.79, + "learning_rate": 9.626865671641791e-07, + "loss": 1.5414, + "step": 37360 + }, + { + "epoch": 6.79, + "learning_rate": 9.62631962140517e-07, + "loss": 2.5335, + "step": 37361 + }, + { + "epoch": 6.79, + "learning_rate": 9.625773571168548e-07, + "loss": 0.1705, + "step": 37362 + }, + { + "epoch": 6.79, + "learning_rate": 9.625227520931925e-07, + "loss": 0.4603, + "step": 37363 + }, + { + "epoch": 6.79, + "learning_rate": 9.624681470695303e-07, + "loss": 0.0555, + "step": 37364 + }, + { + "epoch": 6.79, + "learning_rate": 9.624135420458682e-07, + "loss": 1.5508, + "step": 37365 + }, + { + "epoch": 6.8, + "learning_rate": 9.623589370222061e-07, + "loss": 0.5818, + "step": 37366 + }, + { + "epoch": 6.8, + "learning_rate": 9.623043319985439e-07, + "loss": 0.0234, + "step": 37367 + }, + { + "epoch": 6.8, + "learning_rate": 9.622497269748818e-07, + "loss": 0.9201, + "step": 37368 + }, + { + "epoch": 6.8, + "learning_rate": 9.621951219512195e-07, + "loss": 2.1443, + "step": 37369 + }, + { + "epoch": 6.8, + "learning_rate": 9.621405169275573e-07, + "loss": 0.0153, + "step": 37370 + }, + { + "epoch": 6.8, + "learning_rate": 9.62085911903895e-07, + "loss": 1.2849, + "step": 37371 + }, + { + "epoch": 6.8, + "learning_rate": 9.62031306880233e-07, + "loss": 0.2765, + "step": 37372 + }, + { + "epoch": 6.8, + "learning_rate": 9.61976701856571e-07, + "loss": 0.0652, + "step": 37373 + }, + { + "epoch": 6.8, + "learning_rate": 9.619220968329086e-07, + "loss": 2.9457, + "step": 37374 + }, + { + "epoch": 6.8, + "learning_rate": 9.618674918092466e-07, + "loss": 1.2951, + "step": 37375 + }, + { + "epoch": 6.8, + "learning_rate": 9.618128867855843e-07, + "loss": 0.0974, + "step": 37376 + }, + { + "epoch": 6.8, + "learning_rate": 9.61758281761922e-07, + "loss": 0.0352, + "step": 37377 + }, + { + "epoch": 6.8, + "learning_rate": 9.6170367673826e-07, + "loss": 0.005, + "step": 37378 + }, + { + "epoch": 6.8, + "learning_rate": 9.61649071714598e-07, + "loss": 0.0188, + "step": 37379 + }, + { + "epoch": 6.8, + "learning_rate": 9.615944666909357e-07, + "loss": 0.0763, + "step": 37380 + }, + { + "epoch": 6.8, + "learning_rate": 9.615398616672734e-07, + "loss": 1.187, + "step": 37381 + }, + { + "epoch": 6.8, + "learning_rate": 9.614852566436113e-07, + "loss": 0.0597, + "step": 37382 + }, + { + "epoch": 6.8, + "learning_rate": 9.61430651619949e-07, + "loss": 1.3466, + "step": 37383 + }, + { + "epoch": 6.8, + "learning_rate": 9.613760465962868e-07, + "loss": 1.2597, + "step": 37384 + }, + { + "epoch": 6.8, + "learning_rate": 9.613214415726247e-07, + "loss": 0.5957, + "step": 37385 + }, + { + "epoch": 6.8, + "learning_rate": 9.612668365489627e-07, + "loss": 0.2361, + "step": 37386 + }, + { + "epoch": 6.8, + "learning_rate": 9.612122315253004e-07, + "loss": 1.8881, + "step": 37387 + }, + { + "epoch": 6.8, + "learning_rate": 9.611576265016381e-07, + "loss": 0.0324, + "step": 37388 + }, + { + "epoch": 6.8, + "learning_rate": 9.61103021477976e-07, + "loss": 0.7103, + "step": 37389 + }, + { + "epoch": 6.8, + "learning_rate": 9.610484164543138e-07, + "loss": 0.0145, + "step": 37390 + }, + { + "epoch": 6.8, + "learning_rate": 9.609938114306515e-07, + "loss": 2.5802, + "step": 37391 + }, + { + "epoch": 6.8, + "learning_rate": 9.609392064069895e-07, + "loss": 1.7497, + "step": 37392 + }, + { + "epoch": 6.8, + "learning_rate": 9.608846013833274e-07, + "loss": 1.78, + "step": 37393 + }, + { + "epoch": 6.8, + "learning_rate": 9.608299963596652e-07, + "loss": 0.0141, + "step": 37394 + }, + { + "epoch": 6.8, + "learning_rate": 9.60775391336003e-07, + "loss": 0.2073, + "step": 37395 + }, + { + "epoch": 6.8, + "learning_rate": 9.607207863123408e-07, + "loss": 0.8824, + "step": 37396 + }, + { + "epoch": 6.8, + "learning_rate": 9.606661812886786e-07, + "loss": 1.792, + "step": 37397 + }, + { + "epoch": 6.8, + "learning_rate": 9.606115762650163e-07, + "loss": 0.0075, + "step": 37398 + }, + { + "epoch": 6.8, + "learning_rate": 9.605569712413542e-07, + "loss": 0.2188, + "step": 37399 + }, + { + "epoch": 6.8, + "learning_rate": 9.605023662176922e-07, + "loss": 0.0445, + "step": 37400 + }, + { + "epoch": 6.8, + "learning_rate": 9.6044776119403e-07, + "loss": 0.1206, + "step": 37401 + }, + { + "epoch": 6.8, + "learning_rate": 9.603931561703677e-07, + "loss": 0.3801, + "step": 37402 + }, + { + "epoch": 6.8, + "learning_rate": 9.603385511467056e-07, + "loss": 0.3179, + "step": 37403 + }, + { + "epoch": 6.8, + "learning_rate": 9.602839461230433e-07, + "loss": 0.0443, + "step": 37404 + }, + { + "epoch": 6.8, + "learning_rate": 9.60229341099381e-07, + "loss": 1.2269, + "step": 37405 + }, + { + "epoch": 6.8, + "learning_rate": 9.60174736075719e-07, + "loss": 0.1038, + "step": 37406 + }, + { + "epoch": 6.8, + "learning_rate": 9.60120131052057e-07, + "loss": 0.2992, + "step": 37407 + }, + { + "epoch": 6.8, + "learning_rate": 9.600655260283947e-07, + "loss": 0.2669, + "step": 37408 + }, + { + "epoch": 6.8, + "learning_rate": 9.600109210047324e-07, + "loss": 0.2204, + "step": 37409 + }, + { + "epoch": 6.8, + "learning_rate": 9.599563159810704e-07, + "loss": 0.019, + "step": 37410 + }, + { + "epoch": 6.8, + "learning_rate": 9.59901710957408e-07, + "loss": 0.866, + "step": 37411 + }, + { + "epoch": 6.8, + "learning_rate": 9.598471059337458e-07, + "loss": 0.0616, + "step": 37412 + }, + { + "epoch": 6.8, + "learning_rate": 9.597925009100838e-07, + "loss": 0.5159, + "step": 37413 + }, + { + "epoch": 6.8, + "learning_rate": 9.597378958864217e-07, + "loss": 0.2847, + "step": 37414 + }, + { + "epoch": 6.8, + "learning_rate": 9.596832908627594e-07, + "loss": 0.0389, + "step": 37415 + }, + { + "epoch": 6.8, + "learning_rate": 9.596286858390972e-07, + "loss": 0.16, + "step": 37416 + }, + { + "epoch": 6.8, + "learning_rate": 9.59574080815435e-07, + "loss": 0.7864, + "step": 37417 + }, + { + "epoch": 6.8, + "learning_rate": 9.595194757917728e-07, + "loss": 0.2599, + "step": 37418 + }, + { + "epoch": 6.8, + "learning_rate": 9.594648707681108e-07, + "loss": 0.5848, + "step": 37419 + }, + { + "epoch": 6.8, + "learning_rate": 9.594102657444485e-07, + "loss": 0.3087, + "step": 37420 + }, + { + "epoch": 6.81, + "learning_rate": 9.593556607207865e-07, + "loss": 0.4212, + "step": 37421 + }, + { + "epoch": 6.81, + "learning_rate": 9.593010556971242e-07, + "loss": 0.0082, + "step": 37422 + }, + { + "epoch": 6.81, + "learning_rate": 9.59246450673462e-07, + "loss": 1.4841, + "step": 37423 + }, + { + "epoch": 6.81, + "learning_rate": 9.591918456497997e-07, + "loss": 0.0311, + "step": 37424 + }, + { + "epoch": 6.81, + "learning_rate": 9.591372406261376e-07, + "loss": 0.0142, + "step": 37425 + }, + { + "epoch": 6.81, + "learning_rate": 9.590826356024755e-07, + "loss": 1.4655, + "step": 37426 + }, + { + "epoch": 6.81, + "learning_rate": 9.590280305788133e-07, + "loss": 0.0064, + "step": 37427 + }, + { + "epoch": 6.81, + "learning_rate": 9.589734255551512e-07, + "loss": 0.0713, + "step": 37428 + }, + { + "epoch": 6.81, + "learning_rate": 9.58918820531489e-07, + "loss": 0.3553, + "step": 37429 + }, + { + "epoch": 6.81, + "learning_rate": 9.588642155078267e-07, + "loss": 1.2965, + "step": 37430 + }, + { + "epoch": 6.81, + "learning_rate": 9.588096104841644e-07, + "loss": 0.903, + "step": 37431 + }, + { + "epoch": 6.81, + "learning_rate": 9.587550054605024e-07, + "loss": 0.2163, + "step": 37432 + }, + { + "epoch": 6.81, + "learning_rate": 9.587004004368403e-07, + "loss": 0.612, + "step": 37433 + }, + { + "epoch": 6.81, + "learning_rate": 9.58645795413178e-07, + "loss": 1.9199, + "step": 37434 + }, + { + "epoch": 6.81, + "learning_rate": 9.58591190389516e-07, + "loss": 0.1658, + "step": 37435 + }, + { + "epoch": 6.81, + "learning_rate": 9.585365853658537e-07, + "loss": 0.0066, + "step": 37436 + }, + { + "epoch": 6.81, + "learning_rate": 9.584819803421914e-07, + "loss": 0.2326, + "step": 37437 + }, + { + "epoch": 6.81, + "learning_rate": 9.584273753185292e-07, + "loss": 0.6827, + "step": 37438 + }, + { + "epoch": 6.81, + "learning_rate": 9.583727702948671e-07, + "loss": 0.1091, + "step": 37439 + }, + { + "epoch": 6.81, + "learning_rate": 9.58318165271205e-07, + "loss": 0.3643, + "step": 37440 + }, + { + "epoch": 6.81, + "learning_rate": 9.582635602475428e-07, + "loss": 0.9458, + "step": 37441 + }, + { + "epoch": 6.81, + "learning_rate": 9.582089552238807e-07, + "loss": 0.0114, + "step": 37442 + }, + { + "epoch": 6.81, + "learning_rate": 9.581543502002185e-07, + "loss": 0.584, + "step": 37443 + }, + { + "epoch": 6.81, + "learning_rate": 9.580997451765562e-07, + "loss": 0.0587, + "step": 37444 + }, + { + "epoch": 6.81, + "learning_rate": 9.58045140152894e-07, + "loss": 2.335, + "step": 37445 + }, + { + "epoch": 6.81, + "learning_rate": 9.579905351292319e-07, + "loss": 0.6522, + "step": 37446 + }, + { + "epoch": 6.81, + "learning_rate": 9.579359301055698e-07, + "loss": 0.0508, + "step": 37447 + }, + { + "epoch": 6.81, + "learning_rate": 9.578813250819075e-07, + "loss": 0.3612, + "step": 37448 + }, + { + "epoch": 6.81, + "learning_rate": 9.578267200582455e-07, + "loss": 0.0328, + "step": 37449 + }, + { + "epoch": 6.81, + "learning_rate": 9.577721150345832e-07, + "loss": 0.0103, + "step": 37450 + }, + { + "epoch": 6.81, + "learning_rate": 9.57717510010921e-07, + "loss": 1.9175, + "step": 37451 + }, + { + "epoch": 6.81, + "learning_rate": 9.576629049872587e-07, + "loss": 0.8874, + "step": 37452 + }, + { + "epoch": 6.81, + "learning_rate": 9.576082999635966e-07, + "loss": 0.2402, + "step": 37453 + }, + { + "epoch": 6.81, + "learning_rate": 9.575536949399346e-07, + "loss": 0.0097, + "step": 37454 + }, + { + "epoch": 6.81, + "learning_rate": 9.574990899162723e-07, + "loss": 0.1019, + "step": 37455 + }, + { + "epoch": 6.81, + "learning_rate": 9.574444848926102e-07, + "loss": 1.313, + "step": 37456 + }, + { + "epoch": 6.81, + "learning_rate": 9.57389879868948e-07, + "loss": 0.846, + "step": 37457 + }, + { + "epoch": 6.81, + "learning_rate": 9.573352748452857e-07, + "loss": 0.4608, + "step": 37458 + }, + { + "epoch": 6.81, + "learning_rate": 9.572806698216234e-07, + "loss": 0.0094, + "step": 37459 + }, + { + "epoch": 6.81, + "learning_rate": 9.572260647979614e-07, + "loss": 1.1523, + "step": 37460 + }, + { + "epoch": 6.81, + "learning_rate": 9.571714597742993e-07, + "loss": 1.5891, + "step": 37461 + }, + { + "epoch": 6.81, + "learning_rate": 9.57116854750637e-07, + "loss": 1.5528, + "step": 37462 + }, + { + "epoch": 6.81, + "learning_rate": 9.57062249726975e-07, + "loss": 0.0052, + "step": 37463 + }, + { + "epoch": 6.81, + "learning_rate": 9.570076447033127e-07, + "loss": 0.0779, + "step": 37464 + }, + { + "epoch": 6.81, + "learning_rate": 9.569530396796505e-07, + "loss": 0.0054, + "step": 37465 + }, + { + "epoch": 6.81, + "learning_rate": 9.568984346559884e-07, + "loss": 0.0716, + "step": 37466 + }, + { + "epoch": 6.81, + "learning_rate": 9.568438296323264e-07, + "loss": 0.0637, + "step": 37467 + }, + { + "epoch": 6.81, + "learning_rate": 9.56789224608664e-07, + "loss": 0.027, + "step": 37468 + }, + { + "epoch": 6.81, + "learning_rate": 9.567346195850018e-07, + "loss": 0.8659, + "step": 37469 + }, + { + "epoch": 6.81, + "learning_rate": 9.566800145613398e-07, + "loss": 0.168, + "step": 37470 + }, + { + "epoch": 6.81, + "learning_rate": 9.566254095376775e-07, + "loss": 0.0126, + "step": 37471 + }, + { + "epoch": 6.81, + "learning_rate": 9.565708045140152e-07, + "loss": 1.9969, + "step": 37472 + }, + { + "epoch": 6.81, + "learning_rate": 9.565161994903532e-07, + "loss": 2.4408, + "step": 37473 + }, + { + "epoch": 6.81, + "learning_rate": 9.564615944666911e-07, + "loss": 1.6236, + "step": 37474 + }, + { + "epoch": 6.81, + "learning_rate": 9.564069894430288e-07, + "loss": 0.6482, + "step": 37475 + }, + { + "epoch": 6.82, + "learning_rate": 9.563523844193666e-07, + "loss": 0.1079, + "step": 37476 + }, + { + "epoch": 6.82, + "learning_rate": 9.562977793957045e-07, + "loss": 0.0178, + "step": 37477 + }, + { + "epoch": 6.82, + "learning_rate": 9.562431743720422e-07, + "loss": 0.0156, + "step": 37478 + }, + { + "epoch": 6.82, + "learning_rate": 9.5618856934838e-07, + "loss": 1.6689, + "step": 37479 + }, + { + "epoch": 6.82, + "learning_rate": 9.56133964324718e-07, + "loss": 1.5857, + "step": 37480 + }, + { + "epoch": 6.82, + "learning_rate": 9.560793593010559e-07, + "loss": 0.0057, + "step": 37481 + }, + { + "epoch": 6.82, + "learning_rate": 9.560247542773936e-07, + "loss": 0.0131, + "step": 37482 + }, + { + "epoch": 6.82, + "learning_rate": 9.559701492537313e-07, + "loss": 1.2653, + "step": 37483 + }, + { + "epoch": 6.82, + "learning_rate": 9.559155442300693e-07, + "loss": 0.8848, + "step": 37484 + }, + { + "epoch": 6.82, + "learning_rate": 9.55860939206407e-07, + "loss": 0.3825, + "step": 37485 + }, + { + "epoch": 6.82, + "learning_rate": 9.558063341827447e-07, + "loss": 0.1444, + "step": 37486 + }, + { + "epoch": 6.82, + "learning_rate": 9.557517291590827e-07, + "loss": 0.6209, + "step": 37487 + }, + { + "epoch": 6.82, + "learning_rate": 9.556971241354206e-07, + "loss": 0.3428, + "step": 37488 + }, + { + "epoch": 6.82, + "learning_rate": 9.556425191117584e-07, + "loss": 0.0091, + "step": 37489 + }, + { + "epoch": 6.82, + "learning_rate": 9.55587914088096e-07, + "loss": 0.495, + "step": 37490 + }, + { + "epoch": 6.82, + "learning_rate": 9.55533309064434e-07, + "loss": 0.3953, + "step": 37491 + }, + { + "epoch": 6.82, + "learning_rate": 9.554787040407718e-07, + "loss": 0.0817, + "step": 37492 + }, + { + "epoch": 6.82, + "learning_rate": 9.554240990171095e-07, + "loss": 1.1213, + "step": 37493 + }, + { + "epoch": 6.82, + "learning_rate": 9.553694939934474e-07, + "loss": 0.0115, + "step": 37494 + }, + { + "epoch": 6.82, + "learning_rate": 9.553148889697854e-07, + "loss": 1.1487, + "step": 37495 + }, + { + "epoch": 6.82, + "learning_rate": 9.552602839461231e-07, + "loss": 0.8753, + "step": 37496 + }, + { + "epoch": 6.82, + "learning_rate": 9.552056789224608e-07, + "loss": 1.5888, + "step": 37497 + }, + { + "epoch": 6.82, + "learning_rate": 9.551510738987986e-07, + "loss": 1.1458, + "step": 37498 + }, + { + "epoch": 6.82, + "learning_rate": 9.550964688751365e-07, + "loss": 2.9512, + "step": 37499 + }, + { + "epoch": 6.82, + "learning_rate": 9.550418638514742e-07, + "loss": 0.0595, + "step": 37500 + }, + { + "epoch": 6.82, + "learning_rate": 9.549872588278122e-07, + "loss": 0.0514, + "step": 37501 + }, + { + "epoch": 6.82, + "learning_rate": 9.549326538041501e-07, + "loss": 1.3536, + "step": 37502 + }, + { + "epoch": 6.82, + "learning_rate": 9.548780487804879e-07, + "loss": 0.0373, + "step": 37503 + }, + { + "epoch": 6.82, + "learning_rate": 9.548234437568256e-07, + "loss": 1.1979, + "step": 37504 + }, + { + "epoch": 6.82, + "learning_rate": 9.547688387331633e-07, + "loss": 0.1764, + "step": 37505 + }, + { + "epoch": 6.82, + "learning_rate": 9.547142337095013e-07, + "loss": 0.4864, + "step": 37506 + }, + { + "epoch": 6.82, + "learning_rate": 9.546596286858392e-07, + "loss": 2.6298, + "step": 37507 + }, + { + "epoch": 6.82, + "learning_rate": 9.54605023662177e-07, + "loss": 0.2849, + "step": 37508 + }, + { + "epoch": 6.82, + "learning_rate": 9.545504186385149e-07, + "loss": 0.0191, + "step": 37509 + }, + { + "epoch": 6.82, + "learning_rate": 9.544958136148526e-07, + "loss": 0.3953, + "step": 37510 + }, + { + "epoch": 6.82, + "learning_rate": 9.544412085911904e-07, + "loss": 0.0224, + "step": 37511 + }, + { + "epoch": 6.82, + "learning_rate": 9.54386603567528e-07, + "loss": 0.0652, + "step": 37512 + }, + { + "epoch": 6.82, + "learning_rate": 9.54331998543866e-07, + "loss": 0.2807, + "step": 37513 + }, + { + "epoch": 6.82, + "learning_rate": 9.54277393520204e-07, + "loss": 2.0405, + "step": 37514 + }, + { + "epoch": 6.82, + "learning_rate": 9.542227884965417e-07, + "loss": 0.6183, + "step": 37515 + }, + { + "epoch": 6.82, + "learning_rate": 9.541681834728796e-07, + "loss": 0.7423, + "step": 37516 + }, + { + "epoch": 6.82, + "learning_rate": 9.541135784492174e-07, + "loss": 0.6828, + "step": 37517 + }, + { + "epoch": 6.82, + "learning_rate": 9.540589734255551e-07, + "loss": 0.0172, + "step": 37518 + }, + { + "epoch": 6.82, + "learning_rate": 9.540043684018928e-07, + "loss": 0.2423, + "step": 37519 + }, + { + "epoch": 6.82, + "learning_rate": 9.539497633782308e-07, + "loss": 0.096, + "step": 37520 + }, + { + "epoch": 6.82, + "learning_rate": 9.538951583545687e-07, + "loss": 1.479, + "step": 37521 + }, + { + "epoch": 6.82, + "learning_rate": 9.538405533309065e-07, + "loss": 0.4031, + "step": 37522 + }, + { + "epoch": 6.82, + "learning_rate": 9.537859483072444e-07, + "loss": 0.3311, + "step": 37523 + }, + { + "epoch": 6.82, + "learning_rate": 9.537313432835821e-07, + "loss": 0.0434, + "step": 37524 + }, + { + "epoch": 6.82, + "learning_rate": 9.536767382599199e-07, + "loss": 0.1357, + "step": 37525 + }, + { + "epoch": 6.82, + "learning_rate": 9.536221332362577e-07, + "loss": 0.3879, + "step": 37526 + }, + { + "epoch": 6.82, + "learning_rate": 9.535675282125956e-07, + "loss": 0.1457, + "step": 37527 + }, + { + "epoch": 6.82, + "learning_rate": 9.535129231889334e-07, + "loss": 1.7373, + "step": 37528 + }, + { + "epoch": 6.82, + "learning_rate": 9.534583181652712e-07, + "loss": 0.0065, + "step": 37529 + }, + { + "epoch": 6.82, + "learning_rate": 9.534037131416092e-07, + "loss": 0.3629, + "step": 37530 + }, + { + "epoch": 6.83, + "learning_rate": 9.533491081179469e-07, + "loss": 0.4625, + "step": 37531 + }, + { + "epoch": 6.83, + "learning_rate": 9.532945030942846e-07, + "loss": 1.8504, + "step": 37532 + }, + { + "epoch": 6.83, + "learning_rate": 9.532398980706225e-07, + "loss": 0.3996, + "step": 37533 + }, + { + "epoch": 6.83, + "learning_rate": 9.531852930469604e-07, + "loss": 0.432, + "step": 37534 + }, + { + "epoch": 6.83, + "learning_rate": 9.531306880232981e-07, + "loss": 2.6837, + "step": 37535 + }, + { + "epoch": 6.83, + "learning_rate": 9.53076082999636e-07, + "loss": 0.8151, + "step": 37536 + }, + { + "epoch": 6.83, + "learning_rate": 9.530214779759739e-07, + "loss": 0.3698, + "step": 37537 + }, + { + "epoch": 6.83, + "learning_rate": 9.529668729523116e-07, + "loss": 0.1312, + "step": 37538 + }, + { + "epoch": 6.83, + "learning_rate": 9.529122679286495e-07, + "loss": 0.0408, + "step": 37539 + }, + { + "epoch": 6.83, + "learning_rate": 9.528576629049872e-07, + "loss": 0.1015, + "step": 37540 + }, + { + "epoch": 6.83, + "learning_rate": 9.528030578813252e-07, + "loss": 0.0063, + "step": 37541 + }, + { + "epoch": 6.83, + "learning_rate": 9.527484528576629e-07, + "loss": 0.1695, + "step": 37542 + }, + { + "epoch": 6.83, + "learning_rate": 9.526938478340007e-07, + "loss": 1.5945, + "step": 37543 + }, + { + "epoch": 6.83, + "learning_rate": 9.526392428103387e-07, + "loss": 0.1652, + "step": 37544 + }, + { + "epoch": 6.83, + "learning_rate": 9.525846377866764e-07, + "loss": 0.8831, + "step": 37545 + }, + { + "epoch": 6.83, + "learning_rate": 9.525300327630142e-07, + "loss": 0.1173, + "step": 37546 + }, + { + "epoch": 6.83, + "learning_rate": 9.52475427739352e-07, + "loss": 0.0182, + "step": 37547 + }, + { + "epoch": 6.83, + "learning_rate": 9.524208227156899e-07, + "loss": 0.0285, + "step": 37548 + }, + { + "epoch": 6.83, + "learning_rate": 9.523662176920277e-07, + "loss": 0.3401, + "step": 37549 + }, + { + "epoch": 6.83, + "learning_rate": 9.523116126683655e-07, + "loss": 0.0257, + "step": 37550 + }, + { + "epoch": 6.83, + "learning_rate": 9.522570076447034e-07, + "loss": 0.1481, + "step": 37551 + }, + { + "epoch": 6.83, + "learning_rate": 9.522024026210412e-07, + "loss": 0.3366, + "step": 37552 + }, + { + "epoch": 6.83, + "learning_rate": 9.52147797597379e-07, + "loss": 0.2402, + "step": 37553 + }, + { + "epoch": 6.83, + "learning_rate": 9.520931925737167e-07, + "loss": 0.858, + "step": 37554 + }, + { + "epoch": 6.83, + "learning_rate": 9.520385875500547e-07, + "loss": 0.6508, + "step": 37555 + }, + { + "epoch": 6.83, + "learning_rate": 9.519839825263924e-07, + "loss": 0.9669, + "step": 37556 + }, + { + "epoch": 6.83, + "learning_rate": 9.519293775027302e-07, + "loss": 0.0096, + "step": 37557 + }, + { + "epoch": 6.83, + "learning_rate": 9.518747724790682e-07, + "loss": 0.0911, + "step": 37558 + }, + { + "epoch": 6.83, + "learning_rate": 9.518201674554059e-07, + "loss": 1.4133, + "step": 37559 + }, + { + "epoch": 6.83, + "learning_rate": 9.517655624317438e-07, + "loss": 2.1128, + "step": 37560 + }, + { + "epoch": 6.83, + "learning_rate": 9.517109574080815e-07, + "loss": 0.4945, + "step": 37561 + }, + { + "epoch": 6.83, + "learning_rate": 9.516563523844194e-07, + "loss": 0.3594, + "step": 37562 + }, + { + "epoch": 6.83, + "learning_rate": 9.516017473607573e-07, + "loss": 0.8773, + "step": 37563 + }, + { + "epoch": 6.83, + "learning_rate": 9.51547142337095e-07, + "loss": 0.104, + "step": 37564 + }, + { + "epoch": 6.83, + "learning_rate": 9.514925373134329e-07, + "loss": 1.0947, + "step": 37565 + }, + { + "epoch": 6.83, + "learning_rate": 9.514379322897707e-07, + "loss": 0.0224, + "step": 37566 + }, + { + "epoch": 6.83, + "learning_rate": 9.513833272661085e-07, + "loss": 0.0963, + "step": 37567 + }, + { + "epoch": 6.83, + "learning_rate": 9.513287222424462e-07, + "loss": 0.6444, + "step": 37568 + }, + { + "epoch": 6.83, + "learning_rate": 9.512741172187842e-07, + "loss": 0.8606, + "step": 37569 + }, + { + "epoch": 6.83, + "learning_rate": 9.51219512195122e-07, + "loss": 0.2452, + "step": 37570 + }, + { + "epoch": 6.83, + "learning_rate": 9.511649071714598e-07, + "loss": 2.0625, + "step": 37571 + }, + { + "epoch": 6.83, + "learning_rate": 9.511103021477975e-07, + "loss": 1.3038, + "step": 37572 + }, + { + "epoch": 6.83, + "learning_rate": 9.510556971241354e-07, + "loss": 0.0422, + "step": 37573 + }, + { + "epoch": 6.83, + "learning_rate": 9.510010921004733e-07, + "loss": 2.0689, + "step": 37574 + }, + { + "epoch": 6.83, + "learning_rate": 9.50946487076811e-07, + "loss": 0.0115, + "step": 37575 + }, + { + "epoch": 6.83, + "learning_rate": 9.508918820531489e-07, + "loss": 0.0561, + "step": 37576 + }, + { + "epoch": 6.83, + "learning_rate": 9.508372770294868e-07, + "loss": 0.4066, + "step": 37577 + }, + { + "epoch": 6.83, + "learning_rate": 9.507826720058245e-07, + "loss": 0.0072, + "step": 37578 + }, + { + "epoch": 6.83, + "learning_rate": 9.507280669821624e-07, + "loss": 2.3351, + "step": 37579 + }, + { + "epoch": 6.83, + "learning_rate": 9.506734619585003e-07, + "loss": 1.402, + "step": 37580 + }, + { + "epoch": 6.83, + "learning_rate": 9.50618856934838e-07, + "loss": 0.0095, + "step": 37581 + }, + { + "epoch": 6.83, + "learning_rate": 9.505642519111758e-07, + "loss": 0.1861, + "step": 37582 + }, + { + "epoch": 6.83, + "learning_rate": 9.505096468875137e-07, + "loss": 0.6601, + "step": 37583 + }, + { + "epoch": 6.83, + "learning_rate": 9.504550418638515e-07, + "loss": 0.0392, + "step": 37584 + }, + { + "epoch": 6.83, + "learning_rate": 9.504004368401893e-07, + "loss": 1.0441, + "step": 37585 + }, + { + "epoch": 6.84, + "learning_rate": 9.503458318165271e-07, + "loss": 0.1776, + "step": 37586 + }, + { + "epoch": 6.84, + "learning_rate": 9.502912267928651e-07, + "loss": 2.3885, + "step": 37587 + }, + { + "epoch": 6.84, + "learning_rate": 9.502366217692028e-07, + "loss": 0.2135, + "step": 37588 + }, + { + "epoch": 6.84, + "learning_rate": 9.501820167455405e-07, + "loss": 1.2873, + "step": 37589 + }, + { + "epoch": 6.84, + "learning_rate": 9.501274117218785e-07, + "loss": 0.4779, + "step": 37590 + }, + { + "epoch": 6.84, + "learning_rate": 9.500728066982163e-07, + "loss": 0.1163, + "step": 37591 + }, + { + "epoch": 6.84, + "learning_rate": 9.50018201674554e-07, + "loss": 1.0495, + "step": 37592 + }, + { + "epoch": 6.84, + "learning_rate": 9.499635966508919e-07, + "loss": 0.7389, + "step": 37593 + }, + { + "epoch": 6.84, + "learning_rate": 9.499089916272298e-07, + "loss": 0.2286, + "step": 37594 + }, + { + "epoch": 6.84, + "learning_rate": 9.498543866035675e-07, + "loss": 1.373, + "step": 37595 + }, + { + "epoch": 6.84, + "learning_rate": 9.497997815799053e-07, + "loss": 0.1894, + "step": 37596 + }, + { + "epoch": 6.84, + "learning_rate": 9.497451765562432e-07, + "loss": 1.1796, + "step": 37597 + }, + { + "epoch": 6.84, + "learning_rate": 9.496905715325811e-07, + "loss": 2.1454, + "step": 37598 + }, + { + "epoch": 6.84, + "learning_rate": 9.496359665089188e-07, + "loss": 1.3622, + "step": 37599 + }, + { + "epoch": 6.84, + "learning_rate": 9.495813614852566e-07, + "loss": 1.3577, + "step": 37600 + }, + { + "epoch": 6.84, + "learning_rate": 9.495267564615946e-07, + "loss": 0.0089, + "step": 37601 + }, + { + "epoch": 6.84, + "learning_rate": 9.494721514379323e-07, + "loss": 0.0129, + "step": 37602 + }, + { + "epoch": 6.84, + "learning_rate": 9.494175464142701e-07, + "loss": 0.9167, + "step": 37603 + }, + { + "epoch": 6.84, + "learning_rate": 9.493629413906081e-07, + "loss": 1.231, + "step": 37604 + }, + { + "epoch": 6.84, + "learning_rate": 9.493083363669458e-07, + "loss": 2.2477, + "step": 37605 + }, + { + "epoch": 6.84, + "learning_rate": 9.492537313432835e-07, + "loss": 1.9517, + "step": 37606 + }, + { + "epoch": 6.84, + "learning_rate": 9.491991263196214e-07, + "loss": 0.0837, + "step": 37607 + }, + { + "epoch": 6.84, + "learning_rate": 9.491445212959593e-07, + "loss": 0.0048, + "step": 37608 + }, + { + "epoch": 6.84, + "learning_rate": 9.490899162722971e-07, + "loss": 0.9483, + "step": 37609 + }, + { + "epoch": 6.84, + "learning_rate": 9.490353112486349e-07, + "loss": 0.1573, + "step": 37610 + }, + { + "epoch": 6.84, + "learning_rate": 9.489807062249728e-07, + "loss": 0.2599, + "step": 37611 + }, + { + "epoch": 6.84, + "learning_rate": 9.489261012013106e-07, + "loss": 0.53, + "step": 37612 + }, + { + "epoch": 6.84, + "learning_rate": 9.488714961776483e-07, + "loss": 0.0089, + "step": 37613 + }, + { + "epoch": 6.84, + "learning_rate": 9.488168911539861e-07, + "loss": 0.0056, + "step": 37614 + }, + { + "epoch": 6.84, + "learning_rate": 9.487622861303241e-07, + "loss": 1.3597, + "step": 37615 + }, + { + "epoch": 6.84, + "learning_rate": 9.487076811066618e-07, + "loss": 0.1863, + "step": 37616 + }, + { + "epoch": 6.84, + "learning_rate": 9.486530760829996e-07, + "loss": 0.4002, + "step": 37617 + }, + { + "epoch": 6.84, + "learning_rate": 9.485984710593376e-07, + "loss": 0.9851, + "step": 37618 + }, + { + "epoch": 6.84, + "learning_rate": 9.485438660356753e-07, + "loss": 0.6173, + "step": 37619 + }, + { + "epoch": 6.84, + "learning_rate": 9.484892610120132e-07, + "loss": 0.2163, + "step": 37620 + }, + { + "epoch": 6.84, + "learning_rate": 9.484346559883509e-07, + "loss": 0.5099, + "step": 37621 + }, + { + "epoch": 6.84, + "learning_rate": 9.483800509646888e-07, + "loss": 0.3951, + "step": 37622 + }, + { + "epoch": 6.84, + "learning_rate": 9.483254459410266e-07, + "loss": 0.2528, + "step": 37623 + }, + { + "epoch": 6.84, + "learning_rate": 9.482708409173644e-07, + "loss": 0.7733, + "step": 37624 + }, + { + "epoch": 6.84, + "learning_rate": 9.482162358937023e-07, + "loss": 1.5681, + "step": 37625 + }, + { + "epoch": 6.84, + "learning_rate": 9.481616308700401e-07, + "loss": 1.3943, + "step": 37626 + }, + { + "epoch": 6.84, + "learning_rate": 9.481070258463779e-07, + "loss": 0.0079, + "step": 37627 + }, + { + "epoch": 6.84, + "learning_rate": 9.480524208227157e-07, + "loss": 1.5313, + "step": 37628 + }, + { + "epoch": 6.84, + "learning_rate": 9.479978157990536e-07, + "loss": 0.2668, + "step": 37629 + }, + { + "epoch": 6.84, + "learning_rate": 9.479432107753913e-07, + "loss": 0.8295, + "step": 37630 + }, + { + "epoch": 6.84, + "learning_rate": 9.478886057517292e-07, + "loss": 0.3117, + "step": 37631 + }, + { + "epoch": 6.84, + "learning_rate": 9.478340007280671e-07, + "loss": 0.7671, + "step": 37632 + }, + { + "epoch": 6.84, + "learning_rate": 9.477793957044048e-07, + "loss": 0.332, + "step": 37633 + }, + { + "epoch": 6.84, + "learning_rate": 9.477247906807427e-07, + "loss": 0.3433, + "step": 37634 + }, + { + "epoch": 6.84, + "learning_rate": 9.476701856570804e-07, + "loss": 1.42, + "step": 37635 + }, + { + "epoch": 6.84, + "learning_rate": 9.476155806334184e-07, + "loss": 0.4162, + "step": 37636 + }, + { + "epoch": 6.84, + "learning_rate": 9.475609756097561e-07, + "loss": 0.0039, + "step": 37637 + }, + { + "epoch": 6.84, + "learning_rate": 9.475063705860939e-07, + "loss": 0.0045, + "step": 37638 + }, + { + "epoch": 6.84, + "learning_rate": 9.474517655624319e-07, + "loss": 0.2585, + "step": 37639 + }, + { + "epoch": 6.84, + "learning_rate": 9.473971605387696e-07, + "loss": 0.1273, + "step": 37640 + }, + { + "epoch": 6.85, + "learning_rate": 9.473425555151074e-07, + "loss": 0.0302, + "step": 37641 + }, + { + "epoch": 6.85, + "learning_rate": 9.472879504914452e-07, + "loss": 1.727, + "step": 37642 + }, + { + "epoch": 6.85, + "learning_rate": 9.472333454677831e-07, + "loss": 0.6585, + "step": 37643 + }, + { + "epoch": 6.85, + "learning_rate": 9.471787404441209e-07, + "loss": 0.076, + "step": 37644 + }, + { + "epoch": 6.85, + "learning_rate": 9.471241354204587e-07, + "loss": 0.0135, + "step": 37645 + }, + { + "epoch": 6.85, + "learning_rate": 9.470695303967966e-07, + "loss": 1.1716, + "step": 37646 + }, + { + "epoch": 6.85, + "learning_rate": 9.470149253731344e-07, + "loss": 0.8237, + "step": 37647 + }, + { + "epoch": 6.85, + "learning_rate": 9.469603203494722e-07, + "loss": 2.755, + "step": 37648 + }, + { + "epoch": 6.85, + "learning_rate": 9.469057153258099e-07, + "loss": 0.1872, + "step": 37649 + }, + { + "epoch": 6.85, + "learning_rate": 9.468511103021479e-07, + "loss": 1.2811, + "step": 37650 + }, + { + "epoch": 6.85, + "learning_rate": 9.467965052784857e-07, + "loss": 1.1065, + "step": 37651 + }, + { + "epoch": 6.85, + "learning_rate": 9.467419002548234e-07, + "loss": 0.4591, + "step": 37652 + }, + { + "epoch": 6.85, + "learning_rate": 9.466872952311612e-07, + "loss": 0.0085, + "step": 37653 + }, + { + "epoch": 6.85, + "learning_rate": 9.466326902074991e-07, + "loss": 0.1459, + "step": 37654 + }, + { + "epoch": 6.85, + "learning_rate": 9.465780851838369e-07, + "loss": 0.2151, + "step": 37655 + }, + { + "epoch": 6.85, + "learning_rate": 9.465234801601747e-07, + "loss": 0.0417, + "step": 37656 + }, + { + "epoch": 6.85, + "learning_rate": 9.464688751365126e-07, + "loss": 0.412, + "step": 37657 + }, + { + "epoch": 6.85, + "learning_rate": 9.464142701128505e-07, + "loss": 1.0734, + "step": 37658 + }, + { + "epoch": 6.85, + "learning_rate": 9.463596650891882e-07, + "loss": 0.2887, + "step": 37659 + }, + { + "epoch": 6.85, + "learning_rate": 9.463050600655259e-07, + "loss": 0.102, + "step": 37660 + }, + { + "epoch": 6.85, + "learning_rate": 9.462504550418639e-07, + "loss": 0.3188, + "step": 37661 + }, + { + "epoch": 6.85, + "learning_rate": 9.461958500182017e-07, + "loss": 0.7815, + "step": 37662 + }, + { + "epoch": 6.85, + "learning_rate": 9.461412449945394e-07, + "loss": 0.2417, + "step": 37663 + }, + { + "epoch": 6.85, + "learning_rate": 9.460866399708774e-07, + "loss": 0.0859, + "step": 37664 + }, + { + "epoch": 6.85, + "learning_rate": 9.460320349472152e-07, + "loss": 2.709, + "step": 37665 + }, + { + "epoch": 6.85, + "learning_rate": 9.459774299235529e-07, + "loss": 0.0053, + "step": 37666 + }, + { + "epoch": 6.85, + "learning_rate": 9.459228248998908e-07, + "loss": 0.3253, + "step": 37667 + }, + { + "epoch": 6.85, + "learning_rate": 9.458682198762287e-07, + "loss": 0.7309, + "step": 37668 + }, + { + "epoch": 6.85, + "learning_rate": 9.458136148525665e-07, + "loss": 0.5466, + "step": 37669 + }, + { + "epoch": 6.85, + "learning_rate": 9.457590098289042e-07, + "loss": 0.2314, + "step": 37670 + }, + { + "epoch": 6.85, + "learning_rate": 9.457044048052421e-07, + "loss": 0.3025, + "step": 37671 + }, + { + "epoch": 6.85, + "learning_rate": 9.4564979978158e-07, + "loss": 0.0944, + "step": 37672 + }, + { + "epoch": 6.85, + "learning_rate": 9.455951947579177e-07, + "loss": 0.0177, + "step": 37673 + }, + { + "epoch": 6.85, + "learning_rate": 9.455405897342555e-07, + "loss": 0.7293, + "step": 37674 + }, + { + "epoch": 6.85, + "learning_rate": 9.454859847105935e-07, + "loss": 0.2977, + "step": 37675 + }, + { + "epoch": 6.85, + "learning_rate": 9.454313796869312e-07, + "loss": 0.1806, + "step": 37676 + }, + { + "epoch": 6.85, + "learning_rate": 9.45376774663269e-07, + "loss": 0.7686, + "step": 37677 + }, + { + "epoch": 6.85, + "learning_rate": 9.453221696396069e-07, + "loss": 0.0091, + "step": 37678 + }, + { + "epoch": 6.85, + "learning_rate": 9.452675646159447e-07, + "loss": 0.1152, + "step": 37679 + }, + { + "epoch": 6.85, + "learning_rate": 9.452129595922825e-07, + "loss": 0.2391, + "step": 37680 + }, + { + "epoch": 6.85, + "learning_rate": 9.451583545686203e-07, + "loss": 0.295, + "step": 37681 + }, + { + "epoch": 6.85, + "learning_rate": 9.451037495449582e-07, + "loss": 1.4597, + "step": 37682 + }, + { + "epoch": 6.85, + "learning_rate": 9.45049144521296e-07, + "loss": 0.8582, + "step": 37683 + }, + { + "epoch": 6.85, + "learning_rate": 9.449945394976338e-07, + "loss": 4.1176, + "step": 37684 + }, + { + "epoch": 6.85, + "learning_rate": 9.449399344739718e-07, + "loss": 0.1767, + "step": 37685 + }, + { + "epoch": 6.85, + "learning_rate": 9.448853294503095e-07, + "loss": 0.0049, + "step": 37686 + }, + { + "epoch": 6.85, + "learning_rate": 9.448307244266472e-07, + "loss": 0.1056, + "step": 37687 + }, + { + "epoch": 6.85, + "learning_rate": 9.447761194029851e-07, + "loss": 0.0536, + "step": 37688 + }, + { + "epoch": 6.85, + "learning_rate": 9.44721514379323e-07, + "loss": 0.0064, + "step": 37689 + }, + { + "epoch": 6.85, + "learning_rate": 9.446669093556607e-07, + "loss": 0.4804, + "step": 37690 + }, + { + "epoch": 6.85, + "learning_rate": 9.446123043319986e-07, + "loss": 0.2356, + "step": 37691 + }, + { + "epoch": 6.85, + "learning_rate": 9.445576993083365e-07, + "loss": 0.2485, + "step": 37692 + }, + { + "epoch": 6.85, + "learning_rate": 9.445030942846742e-07, + "loss": 1.4961, + "step": 37693 + }, + { + "epoch": 6.85, + "learning_rate": 9.44448489261012e-07, + "loss": 1.0442, + "step": 37694 + }, + { + "epoch": 6.85, + "learning_rate": 9.443938842373498e-07, + "loss": 0.5646, + "step": 37695 + }, + { + "epoch": 6.86, + "learning_rate": 9.443392792136878e-07, + "loss": 0.0071, + "step": 37696 + }, + { + "epoch": 6.86, + "learning_rate": 9.442846741900255e-07, + "loss": 0.7579, + "step": 37697 + }, + { + "epoch": 6.86, + "learning_rate": 9.442300691663633e-07, + "loss": 0.5583, + "step": 37698 + }, + { + "epoch": 6.86, + "learning_rate": 9.441754641427013e-07, + "loss": 1.8804, + "step": 37699 + }, + { + "epoch": 6.86, + "learning_rate": 9.44120859119039e-07, + "loss": 0.3938, + "step": 37700 + }, + { + "epoch": 6.86, + "learning_rate": 9.440662540953767e-07, + "loss": 0.6145, + "step": 37701 + }, + { + "epoch": 6.86, + "learning_rate": 9.440116490717146e-07, + "loss": 1.0956, + "step": 37702 + }, + { + "epoch": 6.86, + "learning_rate": 9.439570440480525e-07, + "loss": 0.3698, + "step": 37703 + }, + { + "epoch": 6.86, + "learning_rate": 9.439024390243902e-07, + "loss": 0.0163, + "step": 37704 + }, + { + "epoch": 6.86, + "learning_rate": 9.438478340007281e-07, + "loss": 0.0841, + "step": 37705 + }, + { + "epoch": 6.86, + "learning_rate": 9.43793228977066e-07, + "loss": 0.918, + "step": 37706 + }, + { + "epoch": 6.86, + "learning_rate": 9.437386239534038e-07, + "loss": 0.0099, + "step": 37707 + }, + { + "epoch": 6.86, + "learning_rate": 9.436840189297416e-07, + "loss": 0.2052, + "step": 37708 + }, + { + "epoch": 6.86, + "learning_rate": 9.436294139060793e-07, + "loss": 1.4886, + "step": 37709 + }, + { + "epoch": 6.86, + "learning_rate": 9.435748088824173e-07, + "loss": 0.3045, + "step": 37710 + }, + { + "epoch": 6.86, + "learning_rate": 9.43520203858755e-07, + "loss": 2.1261, + "step": 37711 + }, + { + "epoch": 6.86, + "learning_rate": 9.434655988350928e-07, + "loss": 0.0059, + "step": 37712 + }, + { + "epoch": 6.86, + "learning_rate": 9.434109938114308e-07, + "loss": 0.0368, + "step": 37713 + }, + { + "epoch": 6.86, + "learning_rate": 9.433563887877685e-07, + "loss": 0.3232, + "step": 37714 + }, + { + "epoch": 6.86, + "learning_rate": 9.433017837641063e-07, + "loss": 1.0787, + "step": 37715 + }, + { + "epoch": 6.86, + "learning_rate": 9.432471787404441e-07, + "loss": 0.1296, + "step": 37716 + }, + { + "epoch": 6.86, + "learning_rate": 9.43192573716782e-07, + "loss": 0.5227, + "step": 37717 + }, + { + "epoch": 6.86, + "learning_rate": 9.431379686931198e-07, + "loss": 4.9106, + "step": 37718 + }, + { + "epoch": 6.86, + "learning_rate": 9.430833636694576e-07, + "loss": 0.501, + "step": 37719 + }, + { + "epoch": 6.86, + "learning_rate": 9.430287586457955e-07, + "loss": 0.0034, + "step": 37720 + }, + { + "epoch": 6.86, + "learning_rate": 9.429741536221333e-07, + "loss": 0.0059, + "step": 37721 + }, + { + "epoch": 6.86, + "learning_rate": 9.429195485984711e-07, + "loss": 2.0985, + "step": 37722 + }, + { + "epoch": 6.86, + "learning_rate": 9.428649435748088e-07, + "loss": 0.01, + "step": 37723 + }, + { + "epoch": 6.86, + "learning_rate": 9.428103385511468e-07, + "loss": 2.4877, + "step": 37724 + }, + { + "epoch": 6.86, + "learning_rate": 9.427557335274845e-07, + "loss": 0.055, + "step": 37725 + }, + { + "epoch": 6.86, + "learning_rate": 9.427011285038224e-07, + "loss": 0.0323, + "step": 37726 + }, + { + "epoch": 6.86, + "learning_rate": 9.426465234801601e-07, + "loss": 0.3447, + "step": 37727 + }, + { + "epoch": 6.86, + "learning_rate": 9.42591918456498e-07, + "loss": 0.0426, + "step": 37728 + }, + { + "epoch": 6.86, + "learning_rate": 9.425373134328359e-07, + "loss": 0.2278, + "step": 37729 + }, + { + "epoch": 6.86, + "learning_rate": 9.424827084091736e-07, + "loss": 0.0302, + "step": 37730 + }, + { + "epoch": 6.86, + "learning_rate": 9.424281033855115e-07, + "loss": 2.7239, + "step": 37731 + }, + { + "epoch": 6.86, + "learning_rate": 9.423734983618494e-07, + "loss": 2.0583, + "step": 37732 + }, + { + "epoch": 6.86, + "learning_rate": 9.423188933381871e-07, + "loss": 0.3888, + "step": 37733 + }, + { + "epoch": 6.86, + "learning_rate": 9.422642883145248e-07, + "loss": 0.4047, + "step": 37734 + }, + { + "epoch": 6.86, + "learning_rate": 9.422096832908628e-07, + "loss": 0.2438, + "step": 37735 + }, + { + "epoch": 6.86, + "learning_rate": 9.421550782672006e-07, + "loss": 0.6573, + "step": 37736 + }, + { + "epoch": 6.86, + "learning_rate": 9.421004732435384e-07, + "loss": 0.1055, + "step": 37737 + }, + { + "epoch": 6.86, + "learning_rate": 9.420458682198763e-07, + "loss": 1.6185, + "step": 37738 + }, + { + "epoch": 6.86, + "learning_rate": 9.419912631962141e-07, + "loss": 1.5635, + "step": 37739 + }, + { + "epoch": 6.86, + "learning_rate": 9.419366581725519e-07, + "loss": 0.1183, + "step": 37740 + }, + { + "epoch": 6.86, + "learning_rate": 9.418820531488896e-07, + "loss": 0.0173, + "step": 37741 + }, + { + "epoch": 6.86, + "learning_rate": 9.418274481252275e-07, + "loss": 0.0056, + "step": 37742 + }, + { + "epoch": 6.86, + "learning_rate": 9.417728431015654e-07, + "loss": 0.7469, + "step": 37743 + }, + { + "epoch": 6.86, + "learning_rate": 9.417182380779031e-07, + "loss": 0.0165, + "step": 37744 + }, + { + "epoch": 6.86, + "learning_rate": 9.41663633054241e-07, + "loss": 1.2284, + "step": 37745 + }, + { + "epoch": 6.86, + "learning_rate": 9.416090280305789e-07, + "loss": 0.9098, + "step": 37746 + }, + { + "epoch": 6.86, + "learning_rate": 9.415544230069166e-07, + "loss": 0.1216, + "step": 37747 + }, + { + "epoch": 6.86, + "learning_rate": 9.414998179832545e-07, + "loss": 3.0608, + "step": 37748 + }, + { + "epoch": 6.86, + "learning_rate": 9.414452129595924e-07, + "loss": 0.4018, + "step": 37749 + }, + { + "epoch": 6.86, + "learning_rate": 9.413906079359301e-07, + "loss": 0.0135, + "step": 37750 + }, + { + "epoch": 6.87, + "learning_rate": 9.413360029122679e-07, + "loss": 0.2208, + "step": 37751 + }, + { + "epoch": 6.87, + "learning_rate": 9.412813978886058e-07, + "loss": 0.6496, + "step": 37752 + }, + { + "epoch": 6.87, + "learning_rate": 9.412267928649436e-07, + "loss": 0.0453, + "step": 37753 + }, + { + "epoch": 6.87, + "learning_rate": 9.411721878412814e-07, + "loss": 0.1051, + "step": 37754 + }, + { + "epoch": 6.87, + "learning_rate": 9.411175828176192e-07, + "loss": 4.7497, + "step": 37755 + }, + { + "epoch": 6.87, + "learning_rate": 9.410629777939572e-07, + "loss": 0.046, + "step": 37756 + }, + { + "epoch": 6.87, + "learning_rate": 9.410083727702949e-07, + "loss": 0.0068, + "step": 37757 + }, + { + "epoch": 6.87, + "learning_rate": 9.409537677466326e-07, + "loss": 0.6274, + "step": 37758 + }, + { + "epoch": 6.87, + "learning_rate": 9.408991627229706e-07, + "loss": 0.9977, + "step": 37759 + }, + { + "epoch": 6.87, + "learning_rate": 9.408445576993084e-07, + "loss": 0.0185, + "step": 37760 + }, + { + "epoch": 6.87, + "learning_rate": 9.407899526756461e-07, + "loss": 0.0064, + "step": 37761 + }, + { + "epoch": 6.87, + "learning_rate": 9.40735347651984e-07, + "loss": 0.0045, + "step": 37762 + }, + { + "epoch": 6.87, + "learning_rate": 9.406807426283219e-07, + "loss": 0.1562, + "step": 37763 + }, + { + "epoch": 6.87, + "learning_rate": 9.406261376046596e-07, + "loss": 0.6503, + "step": 37764 + }, + { + "epoch": 6.87, + "learning_rate": 9.405715325809974e-07, + "loss": 0.0293, + "step": 37765 + }, + { + "epoch": 6.87, + "learning_rate": 9.405169275573353e-07, + "loss": 0.5456, + "step": 37766 + }, + { + "epoch": 6.87, + "learning_rate": 9.404623225336732e-07, + "loss": 0.0407, + "step": 37767 + }, + { + "epoch": 6.87, + "learning_rate": 9.404077175100109e-07, + "loss": 2.7283, + "step": 37768 + }, + { + "epoch": 6.87, + "learning_rate": 9.403531124863487e-07, + "loss": 0.05, + "step": 37769 + }, + { + "epoch": 6.87, + "learning_rate": 9.402985074626867e-07, + "loss": 1.279, + "step": 37770 + }, + { + "epoch": 6.87, + "learning_rate": 9.402439024390244e-07, + "loss": 0.2481, + "step": 37771 + }, + { + "epoch": 6.87, + "learning_rate": 9.401892974153622e-07, + "loss": 1.6452, + "step": 37772 + }, + { + "epoch": 6.87, + "learning_rate": 9.401346923917002e-07, + "loss": 0.684, + "step": 37773 + }, + { + "epoch": 6.87, + "learning_rate": 9.400800873680379e-07, + "loss": 1.0789, + "step": 37774 + }, + { + "epoch": 6.87, + "learning_rate": 9.400254823443756e-07, + "loss": 0.1197, + "step": 37775 + }, + { + "epoch": 6.87, + "learning_rate": 9.399708773207135e-07, + "loss": 0.138, + "step": 37776 + }, + { + "epoch": 6.87, + "learning_rate": 9.399162722970514e-07, + "loss": 0.2782, + "step": 37777 + }, + { + "epoch": 6.87, + "learning_rate": 9.398616672733892e-07, + "loss": 0.607, + "step": 37778 + }, + { + "epoch": 6.87, + "learning_rate": 9.39807062249727e-07, + "loss": 0.954, + "step": 37779 + }, + { + "epoch": 6.87, + "learning_rate": 9.397524572260649e-07, + "loss": 0.1239, + "step": 37780 + }, + { + "epoch": 6.87, + "learning_rate": 9.396978522024027e-07, + "loss": 0.0646, + "step": 37781 + }, + { + "epoch": 6.87, + "learning_rate": 9.396432471787404e-07, + "loss": 1.6759, + "step": 37782 + }, + { + "epoch": 6.87, + "learning_rate": 9.395886421550782e-07, + "loss": 0.0576, + "step": 37783 + }, + { + "epoch": 6.87, + "learning_rate": 9.395340371314162e-07, + "loss": 1.8708, + "step": 37784 + }, + { + "epoch": 6.87, + "learning_rate": 9.394794321077539e-07, + "loss": 0.1823, + "step": 37785 + }, + { + "epoch": 6.87, + "learning_rate": 9.394248270840918e-07, + "loss": 0.0304, + "step": 37786 + }, + { + "epoch": 6.87, + "learning_rate": 9.393702220604297e-07, + "loss": 0.2169, + "step": 37787 + }, + { + "epoch": 6.87, + "learning_rate": 9.393156170367674e-07, + "loss": 0.5235, + "step": 37788 + }, + { + "epoch": 6.87, + "learning_rate": 9.392610120131052e-07, + "loss": 1.5066, + "step": 37789 + }, + { + "epoch": 6.87, + "learning_rate": 9.39206406989443e-07, + "loss": 0.0303, + "step": 37790 + }, + { + "epoch": 6.87, + "learning_rate": 9.391518019657809e-07, + "loss": 0.1631, + "step": 37791 + }, + { + "epoch": 6.87, + "learning_rate": 9.390971969421187e-07, + "loss": 0.0502, + "step": 37792 + }, + { + "epoch": 6.87, + "learning_rate": 9.390425919184565e-07, + "loss": 0.7963, + "step": 37793 + }, + { + "epoch": 6.87, + "learning_rate": 9.389879868947945e-07, + "loss": 0.7304, + "step": 37794 + }, + { + "epoch": 6.87, + "learning_rate": 9.389333818711322e-07, + "loss": 1.1951, + "step": 37795 + }, + { + "epoch": 6.87, + "learning_rate": 9.3887877684747e-07, + "loss": 0.3926, + "step": 37796 + }, + { + "epoch": 6.87, + "learning_rate": 9.388241718238078e-07, + "loss": 0.781, + "step": 37797 + }, + { + "epoch": 6.87, + "learning_rate": 9.387695668001457e-07, + "loss": 0.0739, + "step": 37798 + }, + { + "epoch": 6.87, + "learning_rate": 9.387149617764834e-07, + "loss": 1.9632, + "step": 37799 + }, + { + "epoch": 6.87, + "learning_rate": 9.386603567528213e-07, + "loss": 0.5224, + "step": 37800 + }, + { + "epoch": 6.87, + "learning_rate": 9.386057517291592e-07, + "loss": 0.4895, + "step": 37801 + }, + { + "epoch": 6.87, + "learning_rate": 9.385511467054969e-07, + "loss": 0.47, + "step": 37802 + }, + { + "epoch": 6.87, + "learning_rate": 9.384965416818348e-07, + "loss": 0.4155, + "step": 37803 + }, + { + "epoch": 6.87, + "learning_rate": 9.384419366581725e-07, + "loss": 0.6742, + "step": 37804 + }, + { + "epoch": 6.87, + "learning_rate": 9.383873316345105e-07, + "loss": 0.7044, + "step": 37805 + }, + { + "epoch": 6.88, + "learning_rate": 9.383327266108482e-07, + "loss": 0.1095, + "step": 37806 + }, + { + "epoch": 6.88, + "learning_rate": 9.38278121587186e-07, + "loss": 0.7723, + "step": 37807 + }, + { + "epoch": 6.88, + "learning_rate": 9.382235165635238e-07, + "loss": 0.0073, + "step": 37808 + }, + { + "epoch": 6.88, + "learning_rate": 9.381689115398617e-07, + "loss": 0.9821, + "step": 37809 + }, + { + "epoch": 6.88, + "learning_rate": 9.381143065161995e-07, + "loss": 1.6707, + "step": 37810 + }, + { + "epoch": 6.88, + "learning_rate": 9.380597014925373e-07, + "loss": 0.4489, + "step": 37811 + }, + { + "epoch": 6.88, + "learning_rate": 9.380050964688752e-07, + "loss": 0.715, + "step": 37812 + }, + { + "epoch": 6.88, + "learning_rate": 9.37950491445213e-07, + "loss": 0.5819, + "step": 37813 + }, + { + "epoch": 6.88, + "learning_rate": 9.378958864215508e-07, + "loss": 1.2652, + "step": 37814 + }, + { + "epoch": 6.88, + "learning_rate": 9.378412813978885e-07, + "loss": 0.0386, + "step": 37815 + }, + { + "epoch": 6.88, + "learning_rate": 9.377866763742265e-07, + "loss": 0.2214, + "step": 37816 + }, + { + "epoch": 6.88, + "learning_rate": 9.377320713505643e-07, + "loss": 0.2744, + "step": 37817 + }, + { + "epoch": 6.88, + "learning_rate": 9.37677466326902e-07, + "loss": 0.2751, + "step": 37818 + }, + { + "epoch": 6.88, + "learning_rate": 9.3762286130324e-07, + "loss": 0.5674, + "step": 37819 + }, + { + "epoch": 6.88, + "learning_rate": 9.375682562795778e-07, + "loss": 1.1282, + "step": 37820 + }, + { + "epoch": 6.88, + "learning_rate": 9.375136512559155e-07, + "loss": 0.1618, + "step": 37821 + }, + { + "epoch": 6.88, + "learning_rate": 9.374590462322533e-07, + "loss": 0.7065, + "step": 37822 + }, + { + "epoch": 6.88, + "learning_rate": 9.374044412085912e-07, + "loss": 0.0071, + "step": 37823 + }, + { + "epoch": 6.88, + "learning_rate": 9.37349836184929e-07, + "loss": 0.1322, + "step": 37824 + }, + { + "epoch": 6.88, + "learning_rate": 9.372952311612668e-07, + "loss": 0.0077, + "step": 37825 + }, + { + "epoch": 6.88, + "learning_rate": 9.372406261376047e-07, + "loss": 0.2023, + "step": 37826 + }, + { + "epoch": 6.88, + "learning_rate": 9.371860211139426e-07, + "loss": 0.2617, + "step": 37827 + }, + { + "epoch": 6.88, + "learning_rate": 9.371314160902803e-07, + "loss": 0.0536, + "step": 37828 + }, + { + "epoch": 6.88, + "learning_rate": 9.37076811066618e-07, + "loss": 1.7359, + "step": 37829 + }, + { + "epoch": 6.88, + "learning_rate": 9.37022206042956e-07, + "loss": 0.0936, + "step": 37830 + }, + { + "epoch": 6.88, + "learning_rate": 9.369676010192938e-07, + "loss": 2.775, + "step": 37831 + }, + { + "epoch": 6.88, + "learning_rate": 9.369129959956315e-07, + "loss": 0.8797, + "step": 37832 + }, + { + "epoch": 6.88, + "learning_rate": 9.368583909719695e-07, + "loss": 0.7204, + "step": 37833 + }, + { + "epoch": 6.88, + "learning_rate": 9.368037859483073e-07, + "loss": 0.182, + "step": 37834 + }, + { + "epoch": 6.88, + "learning_rate": 9.36749180924645e-07, + "loss": 0.0099, + "step": 37835 + }, + { + "epoch": 6.88, + "learning_rate": 9.366945759009829e-07, + "loss": 0.7973, + "step": 37836 + }, + { + "epoch": 6.88, + "learning_rate": 9.366399708773208e-07, + "loss": 0.5356, + "step": 37837 + }, + { + "epoch": 6.88, + "learning_rate": 9.365853658536586e-07, + "loss": 0.3246, + "step": 37838 + }, + { + "epoch": 6.88, + "learning_rate": 9.365307608299963e-07, + "loss": 0.1607, + "step": 37839 + }, + { + "epoch": 6.88, + "learning_rate": 9.364761558063342e-07, + "loss": 0.1475, + "step": 37840 + }, + { + "epoch": 6.88, + "learning_rate": 9.364215507826721e-07, + "loss": 0.9177, + "step": 37841 + }, + { + "epoch": 6.88, + "learning_rate": 9.363669457590098e-07, + "loss": 0.1331, + "step": 37842 + }, + { + "epoch": 6.88, + "learning_rate": 9.363123407353476e-07, + "loss": 0.0152, + "step": 37843 + }, + { + "epoch": 6.88, + "learning_rate": 9.362577357116856e-07, + "loss": 0.1294, + "step": 37844 + }, + { + "epoch": 6.88, + "learning_rate": 9.362031306880233e-07, + "loss": 1.5381, + "step": 37845 + }, + { + "epoch": 6.88, + "learning_rate": 9.36148525664361e-07, + "loss": 0.2375, + "step": 37846 + }, + { + "epoch": 6.88, + "learning_rate": 9.36093920640699e-07, + "loss": 0.3052, + "step": 37847 + }, + { + "epoch": 6.88, + "learning_rate": 9.360393156170368e-07, + "loss": 0.0182, + "step": 37848 + }, + { + "epoch": 6.88, + "learning_rate": 9.359847105933746e-07, + "loss": 0.0673, + "step": 37849 + }, + { + "epoch": 6.88, + "learning_rate": 9.359301055697124e-07, + "loss": 0.0104, + "step": 37850 + }, + { + "epoch": 6.88, + "learning_rate": 9.358755005460503e-07, + "loss": 1.4959, + "step": 37851 + }, + { + "epoch": 6.88, + "learning_rate": 9.358208955223881e-07, + "loss": 0.1899, + "step": 37852 + }, + { + "epoch": 6.88, + "learning_rate": 9.357662904987259e-07, + "loss": 0.0057, + "step": 37853 + }, + { + "epoch": 6.88, + "learning_rate": 9.357116854750638e-07, + "loss": 0.3397, + "step": 37854 + }, + { + "epoch": 6.88, + "learning_rate": 9.356570804514016e-07, + "loss": 0.1451, + "step": 37855 + }, + { + "epoch": 6.88, + "learning_rate": 9.356024754277393e-07, + "loss": 1.1896, + "step": 37856 + }, + { + "epoch": 6.88, + "learning_rate": 9.355478704040772e-07, + "loss": 0.0541, + "step": 37857 + }, + { + "epoch": 6.88, + "learning_rate": 9.354932653804151e-07, + "loss": 0.0049, + "step": 37858 + }, + { + "epoch": 6.88, + "learning_rate": 9.354386603567528e-07, + "loss": 1.13, + "step": 37859 + }, + { + "epoch": 6.88, + "learning_rate": 9.353840553330907e-07, + "loss": 0.1622, + "step": 37860 + }, + { + "epoch": 6.89, + "learning_rate": 9.353294503094286e-07, + "loss": 1.0001, + "step": 37861 + }, + { + "epoch": 6.89, + "learning_rate": 9.352748452857663e-07, + "loss": 2.4162, + "step": 37862 + }, + { + "epoch": 6.89, + "learning_rate": 9.352202402621041e-07, + "loss": 0.2482, + "step": 37863 + }, + { + "epoch": 6.89, + "learning_rate": 9.351656352384419e-07, + "loss": 0.6558, + "step": 37864 + }, + { + "epoch": 6.89, + "learning_rate": 9.351110302147799e-07, + "loss": 1.1539, + "step": 37865 + }, + { + "epoch": 6.89, + "learning_rate": 9.350564251911176e-07, + "loss": 0.2257, + "step": 37866 + }, + { + "epoch": 6.89, + "learning_rate": 9.350018201674554e-07, + "loss": 0.8072, + "step": 37867 + }, + { + "epoch": 6.89, + "learning_rate": 9.349472151437934e-07, + "loss": 1.3468, + "step": 37868 + }, + { + "epoch": 6.89, + "learning_rate": 9.348926101201311e-07, + "loss": 0.1215, + "step": 37869 + }, + { + "epoch": 6.89, + "learning_rate": 9.348380050964688e-07, + "loss": 0.0857, + "step": 37870 + }, + { + "epoch": 6.89, + "learning_rate": 9.347834000728067e-07, + "loss": 0.1827, + "step": 37871 + }, + { + "epoch": 6.89, + "learning_rate": 9.347287950491446e-07, + "loss": 0.5112, + "step": 37872 + }, + { + "epoch": 6.89, + "learning_rate": 9.346741900254823e-07, + "loss": 0.0308, + "step": 37873 + }, + { + "epoch": 6.89, + "learning_rate": 9.346195850018202e-07, + "loss": 2.6682, + "step": 37874 + }, + { + "epoch": 6.89, + "learning_rate": 9.345649799781581e-07, + "loss": 1.2053, + "step": 37875 + }, + { + "epoch": 6.89, + "learning_rate": 9.345103749544959e-07, + "loss": 0.5513, + "step": 37876 + }, + { + "epoch": 6.89, + "learning_rate": 9.344557699308337e-07, + "loss": 0.0153, + "step": 37877 + }, + { + "epoch": 6.89, + "learning_rate": 9.344011649071714e-07, + "loss": 1.2218, + "step": 37878 + }, + { + "epoch": 6.89, + "learning_rate": 9.343465598835094e-07, + "loss": 0.6142, + "step": 37879 + }, + { + "epoch": 6.89, + "learning_rate": 9.342919548598471e-07, + "loss": 1.0006, + "step": 37880 + }, + { + "epoch": 6.89, + "learning_rate": 9.342373498361849e-07, + "loss": 0.4155, + "step": 37881 + }, + { + "epoch": 6.89, + "learning_rate": 9.341827448125227e-07, + "loss": 0.0618, + "step": 37882 + }, + { + "epoch": 6.89, + "learning_rate": 9.341281397888606e-07, + "loss": 0.0512, + "step": 37883 + }, + { + "epoch": 6.89, + "learning_rate": 9.340735347651985e-07, + "loss": 0.0054, + "step": 37884 + }, + { + "epoch": 6.89, + "learning_rate": 9.340189297415362e-07, + "loss": 0.211, + "step": 37885 + }, + { + "epoch": 6.89, + "learning_rate": 9.339643247178741e-07, + "loss": 0.3915, + "step": 37886 + }, + { + "epoch": 6.89, + "learning_rate": 9.339097196942119e-07, + "loss": 0.0057, + "step": 37887 + }, + { + "epoch": 6.89, + "learning_rate": 9.338551146705497e-07, + "loss": 0.0297, + "step": 37888 + }, + { + "epoch": 6.89, + "learning_rate": 9.338005096468874e-07, + "loss": 0.2919, + "step": 37889 + }, + { + "epoch": 6.89, + "learning_rate": 9.337459046232254e-07, + "loss": 0.233, + "step": 37890 + }, + { + "epoch": 6.89, + "learning_rate": 9.336912995995632e-07, + "loss": 0.7077, + "step": 37891 + }, + { + "epoch": 6.89, + "learning_rate": 9.336366945759009e-07, + "loss": 0.1132, + "step": 37892 + }, + { + "epoch": 6.89, + "learning_rate": 9.335820895522389e-07, + "loss": 0.3487, + "step": 37893 + }, + { + "epoch": 6.89, + "learning_rate": 9.335274845285766e-07, + "loss": 2.1161, + "step": 37894 + }, + { + "epoch": 6.89, + "learning_rate": 9.334728795049145e-07, + "loss": 2.0105, + "step": 37895 + }, + { + "epoch": 6.89, + "learning_rate": 9.334182744812522e-07, + "loss": 0.0529, + "step": 37896 + }, + { + "epoch": 6.89, + "learning_rate": 9.333636694575901e-07, + "loss": 0.0327, + "step": 37897 + }, + { + "epoch": 6.89, + "learning_rate": 9.33309064433928e-07, + "loss": 0.8616, + "step": 37898 + }, + { + "epoch": 6.89, + "learning_rate": 9.332544594102657e-07, + "loss": 0.0787, + "step": 37899 + }, + { + "epoch": 6.89, + "learning_rate": 9.331998543866036e-07, + "loss": 0.0556, + "step": 37900 + }, + { + "epoch": 6.89, + "learning_rate": 9.331452493629415e-07, + "loss": 3.0957, + "step": 37901 + }, + { + "epoch": 6.89, + "learning_rate": 9.330906443392792e-07, + "loss": 0.1253, + "step": 37902 + }, + { + "epoch": 6.89, + "learning_rate": 9.330360393156169e-07, + "loss": 0.0034, + "step": 37903 + }, + { + "epoch": 6.89, + "learning_rate": 9.329814342919549e-07, + "loss": 0.0084, + "step": 37904 + }, + { + "epoch": 6.89, + "learning_rate": 9.329268292682927e-07, + "loss": 0.3008, + "step": 37905 + }, + { + "epoch": 6.89, + "learning_rate": 9.328722242446305e-07, + "loss": 1.0458, + "step": 37906 + }, + { + "epoch": 6.89, + "learning_rate": 9.328176192209684e-07, + "loss": 0.2861, + "step": 37907 + }, + { + "epoch": 6.89, + "learning_rate": 9.327630141973062e-07, + "loss": 0.7762, + "step": 37908 + }, + { + "epoch": 6.89, + "learning_rate": 9.32708409173644e-07, + "loss": 0.494, + "step": 37909 + }, + { + "epoch": 6.89, + "learning_rate": 9.326538041499817e-07, + "loss": 0.0928, + "step": 37910 + }, + { + "epoch": 6.89, + "learning_rate": 9.325991991263196e-07, + "loss": 0.6849, + "step": 37911 + }, + { + "epoch": 6.89, + "learning_rate": 9.325445941026575e-07, + "loss": 1.8577, + "step": 37912 + }, + { + "epoch": 6.89, + "learning_rate": 9.324899890789952e-07, + "loss": 0.0886, + "step": 37913 + }, + { + "epoch": 6.89, + "learning_rate": 9.324353840553332e-07, + "loss": 3.017, + "step": 37914 + }, + { + "epoch": 6.89, + "learning_rate": 9.32380779031671e-07, + "loss": 1.1806, + "step": 37915 + }, + { + "epoch": 6.9, + "learning_rate": 9.323261740080087e-07, + "loss": 0.1946, + "step": 37916 + }, + { + "epoch": 6.9, + "learning_rate": 9.322715689843466e-07, + "loss": 1.7299, + "step": 37917 + }, + { + "epoch": 6.9, + "learning_rate": 9.322169639606845e-07, + "loss": 2.0811, + "step": 37918 + }, + { + "epoch": 6.9, + "learning_rate": 9.321623589370222e-07, + "loss": 0.0391, + "step": 37919 + }, + { + "epoch": 6.9, + "learning_rate": 9.3210775391336e-07, + "loss": 0.0069, + "step": 37920 + }, + { + "epoch": 6.9, + "learning_rate": 9.320531488896979e-07, + "loss": 1.3909, + "step": 37921 + }, + { + "epoch": 6.9, + "learning_rate": 9.319985438660358e-07, + "loss": 0.4945, + "step": 37922 + }, + { + "epoch": 6.9, + "learning_rate": 9.319439388423735e-07, + "loss": 2.5993, + "step": 37923 + }, + { + "epoch": 6.9, + "learning_rate": 9.318893338187113e-07, + "loss": 2.6617, + "step": 37924 + }, + { + "epoch": 6.9, + "learning_rate": 9.318347287950493e-07, + "loss": 0.3529, + "step": 37925 + }, + { + "epoch": 6.9, + "learning_rate": 9.31780123771387e-07, + "loss": 0.2367, + "step": 37926 + }, + { + "epoch": 6.9, + "learning_rate": 9.317255187477247e-07, + "loss": 0.8238, + "step": 37927 + }, + { + "epoch": 6.9, + "learning_rate": 9.316709137240627e-07, + "loss": 0.3092, + "step": 37928 + }, + { + "epoch": 6.9, + "learning_rate": 9.316163087004005e-07, + "loss": 0.6736, + "step": 37929 + }, + { + "epoch": 6.9, + "learning_rate": 9.315617036767382e-07, + "loss": 2.7186, + "step": 37930 + }, + { + "epoch": 6.9, + "learning_rate": 9.315070986530761e-07, + "loss": 0.0496, + "step": 37931 + }, + { + "epoch": 6.9, + "learning_rate": 9.31452493629414e-07, + "loss": 0.2624, + "step": 37932 + }, + { + "epoch": 6.9, + "learning_rate": 9.313978886057518e-07, + "loss": 0.0166, + "step": 37933 + }, + { + "epoch": 6.9, + "learning_rate": 9.313432835820895e-07, + "loss": 1.0048, + "step": 37934 + }, + { + "epoch": 6.9, + "learning_rate": 9.312886785584274e-07, + "loss": 0.9213, + "step": 37935 + }, + { + "epoch": 6.9, + "learning_rate": 9.312340735347653e-07, + "loss": 2.2805, + "step": 37936 + }, + { + "epoch": 6.9, + "learning_rate": 9.31179468511103e-07, + "loss": 0.124, + "step": 37937 + }, + { + "epoch": 6.9, + "learning_rate": 9.311248634874408e-07, + "loss": 1.5036, + "step": 37938 + }, + { + "epoch": 6.9, + "learning_rate": 9.310702584637788e-07, + "loss": 0.0064, + "step": 37939 + }, + { + "epoch": 6.9, + "learning_rate": 9.310156534401165e-07, + "loss": 0.0768, + "step": 37940 + }, + { + "epoch": 6.9, + "learning_rate": 9.309610484164543e-07, + "loss": 1.121, + "step": 37941 + }, + { + "epoch": 6.9, + "learning_rate": 9.309064433927923e-07, + "loss": 0.0619, + "step": 37942 + }, + { + "epoch": 6.9, + "learning_rate": 9.3085183836913e-07, + "loss": 0.404, + "step": 37943 + }, + { + "epoch": 6.9, + "learning_rate": 9.307972333454678e-07, + "loss": 0.1172, + "step": 37944 + }, + { + "epoch": 6.9, + "learning_rate": 9.307426283218056e-07, + "loss": 0.0922, + "step": 37945 + }, + { + "epoch": 6.9, + "learning_rate": 9.306880232981435e-07, + "loss": 0.1657, + "step": 37946 + }, + { + "epoch": 6.9, + "learning_rate": 9.306334182744813e-07, + "loss": 0.2612, + "step": 37947 + }, + { + "epoch": 6.9, + "learning_rate": 9.305788132508191e-07, + "loss": 0.0084, + "step": 37948 + }, + { + "epoch": 6.9, + "learning_rate": 9.30524208227157e-07, + "loss": 0.238, + "step": 37949 + }, + { + "epoch": 6.9, + "learning_rate": 9.304696032034948e-07, + "loss": 0.7692, + "step": 37950 + }, + { + "epoch": 6.9, + "learning_rate": 9.304149981798325e-07, + "loss": 0.1496, + "step": 37951 + }, + { + "epoch": 6.9, + "learning_rate": 9.303603931561703e-07, + "loss": 0.2061, + "step": 37952 + }, + { + "epoch": 6.9, + "learning_rate": 9.303057881325083e-07, + "loss": 0.0067, + "step": 37953 + }, + { + "epoch": 6.9, + "learning_rate": 9.30251183108846e-07, + "loss": 0.8152, + "step": 37954 + }, + { + "epoch": 6.9, + "learning_rate": 9.301965780851839e-07, + "loss": 0.0039, + "step": 37955 + }, + { + "epoch": 6.9, + "learning_rate": 9.301419730615218e-07, + "loss": 1.4492, + "step": 37956 + }, + { + "epoch": 6.9, + "learning_rate": 9.300873680378595e-07, + "loss": 0.1179, + "step": 37957 + }, + { + "epoch": 6.9, + "learning_rate": 9.300327630141973e-07, + "loss": 0.5637, + "step": 37958 + }, + { + "epoch": 6.9, + "learning_rate": 9.299781579905351e-07, + "loss": 1.3281, + "step": 37959 + }, + { + "epoch": 6.9, + "learning_rate": 9.29923552966873e-07, + "loss": 1.5774, + "step": 37960 + }, + { + "epoch": 6.9, + "learning_rate": 9.298689479432108e-07, + "loss": 0.4171, + "step": 37961 + }, + { + "epoch": 6.9, + "learning_rate": 9.298143429195486e-07, + "loss": 1.4205, + "step": 37962 + }, + { + "epoch": 6.9, + "learning_rate": 9.297597378958863e-07, + "loss": 0.0276, + "step": 37963 + }, + { + "epoch": 6.9, + "learning_rate": 9.297051328722243e-07, + "loss": 0.6154, + "step": 37964 + }, + { + "epoch": 6.9, + "learning_rate": 9.296505278485621e-07, + "loss": 0.0779, + "step": 37965 + }, + { + "epoch": 6.9, + "learning_rate": 9.295959228248999e-07, + "loss": 1.0021, + "step": 37966 + }, + { + "epoch": 6.9, + "learning_rate": 9.295413178012378e-07, + "loss": 0.1543, + "step": 37967 + }, + { + "epoch": 6.9, + "learning_rate": 9.294867127775755e-07, + "loss": 0.0534, + "step": 37968 + }, + { + "epoch": 6.9, + "learning_rate": 9.294321077539134e-07, + "loss": 0.2994, + "step": 37969 + }, + { + "epoch": 6.9, + "learning_rate": 9.293775027302511e-07, + "loss": 0.0947, + "step": 37970 + }, + { + "epoch": 6.91, + "learning_rate": 9.29322897706589e-07, + "loss": 1.3275, + "step": 37971 + }, + { + "epoch": 6.91, + "learning_rate": 9.292682926829269e-07, + "loss": 0.3712, + "step": 37972 + }, + { + "epoch": 6.91, + "learning_rate": 9.292136876592646e-07, + "loss": 0.2709, + "step": 37973 + }, + { + "epoch": 6.91, + "learning_rate": 9.291590826356026e-07, + "loss": 0.1194, + "step": 37974 + }, + { + "epoch": 6.91, + "learning_rate": 9.291044776119403e-07, + "loss": 0.0768, + "step": 37975 + }, + { + "epoch": 6.91, + "learning_rate": 9.290498725882781e-07, + "loss": 0.3922, + "step": 37976 + }, + { + "epoch": 6.91, + "learning_rate": 9.289952675646159e-07, + "loss": 0.0101, + "step": 37977 + }, + { + "epoch": 6.91, + "learning_rate": 9.289406625409538e-07, + "loss": 1.3415, + "step": 37978 + }, + { + "epoch": 6.91, + "learning_rate": 9.288860575172916e-07, + "loss": 0.1734, + "step": 37979 + }, + { + "epoch": 6.91, + "learning_rate": 9.288314524936294e-07, + "loss": 0.3187, + "step": 37980 + }, + { + "epoch": 6.91, + "learning_rate": 9.287768474699673e-07, + "loss": 1.5982, + "step": 37981 + }, + { + "epoch": 6.91, + "learning_rate": 9.287222424463052e-07, + "loss": 0.025, + "step": 37982 + }, + { + "epoch": 6.91, + "learning_rate": 9.286676374226429e-07, + "loss": 0.0058, + "step": 37983 + }, + { + "epoch": 6.91, + "learning_rate": 9.286130323989806e-07, + "loss": 0.5229, + "step": 37984 + }, + { + "epoch": 6.91, + "learning_rate": 9.285584273753186e-07, + "loss": 2.3886, + "step": 37985 + }, + { + "epoch": 6.91, + "learning_rate": 9.285038223516564e-07, + "loss": 1.4745, + "step": 37986 + }, + { + "epoch": 6.91, + "learning_rate": 9.284492173279941e-07, + "loss": 0.4543, + "step": 37987 + }, + { + "epoch": 6.91, + "learning_rate": 9.283946123043321e-07, + "loss": 0.1312, + "step": 37988 + }, + { + "epoch": 6.91, + "learning_rate": 9.283400072806699e-07, + "loss": 1.5049, + "step": 37989 + }, + { + "epoch": 6.91, + "learning_rate": 9.282854022570076e-07, + "loss": 0.1555, + "step": 37990 + }, + { + "epoch": 6.91, + "learning_rate": 9.282307972333454e-07, + "loss": 1.493, + "step": 37991 + }, + { + "epoch": 6.91, + "learning_rate": 9.281761922096833e-07, + "loss": 0.7035, + "step": 37992 + }, + { + "epoch": 6.91, + "learning_rate": 9.281215871860212e-07, + "loss": 0.0054, + "step": 37993 + }, + { + "epoch": 6.91, + "learning_rate": 9.280669821623589e-07, + "loss": 1.3001, + "step": 37994 + }, + { + "epoch": 6.91, + "learning_rate": 9.280123771386968e-07, + "loss": 2.3119, + "step": 37995 + }, + { + "epoch": 6.91, + "learning_rate": 9.279577721150347e-07, + "loss": 1.8341, + "step": 37996 + }, + { + "epoch": 6.91, + "learning_rate": 9.279031670913724e-07, + "loss": 0.1841, + "step": 37997 + }, + { + "epoch": 6.91, + "learning_rate": 9.278485620677101e-07, + "loss": 0.2328, + "step": 37998 + }, + { + "epoch": 6.91, + "learning_rate": 9.277939570440481e-07, + "loss": 0.0638, + "step": 37999 + }, + { + "epoch": 6.91, + "learning_rate": 9.277393520203859e-07, + "loss": 0.0234, + "step": 38000 + }, + { + "epoch": 6.91, + "learning_rate": 9.276847469967236e-07, + "loss": 4.3929, + "step": 38001 + }, + { + "epoch": 6.91, + "learning_rate": 9.276301419730616e-07, + "loss": 0.0971, + "step": 38002 + }, + { + "epoch": 6.91, + "learning_rate": 9.275755369493994e-07, + "loss": 0.087, + "step": 38003 + }, + { + "epoch": 6.91, + "learning_rate": 9.275209319257372e-07, + "loss": 2.8353, + "step": 38004 + }, + { + "epoch": 6.91, + "learning_rate": 9.27466326902075e-07, + "loss": 1.0328, + "step": 38005 + }, + { + "epoch": 6.91, + "learning_rate": 9.274117218784129e-07, + "loss": 0.0075, + "step": 38006 + }, + { + "epoch": 6.91, + "learning_rate": 9.273571168547507e-07, + "loss": 0.2895, + "step": 38007 + }, + { + "epoch": 6.91, + "learning_rate": 9.273025118310884e-07, + "loss": 0.0039, + "step": 38008 + }, + { + "epoch": 6.91, + "learning_rate": 9.272479068074263e-07, + "loss": 0.5432, + "step": 38009 + }, + { + "epoch": 6.91, + "learning_rate": 9.271933017837642e-07, + "loss": 0.1166, + "step": 38010 + }, + { + "epoch": 6.91, + "learning_rate": 9.271386967601019e-07, + "loss": 1.5344, + "step": 38011 + }, + { + "epoch": 6.91, + "learning_rate": 9.270840917364398e-07, + "loss": 0.0597, + "step": 38012 + }, + { + "epoch": 6.91, + "learning_rate": 9.270294867127777e-07, + "loss": 0.4731, + "step": 38013 + }, + { + "epoch": 6.91, + "learning_rate": 9.269748816891154e-07, + "loss": 0.5687, + "step": 38014 + }, + { + "epoch": 6.91, + "learning_rate": 9.269202766654532e-07, + "loss": 0.3235, + "step": 38015 + }, + { + "epoch": 6.91, + "learning_rate": 9.268656716417911e-07, + "loss": 0.7548, + "step": 38016 + }, + { + "epoch": 6.91, + "learning_rate": 9.268110666181289e-07, + "loss": 0.0262, + "step": 38017 + }, + { + "epoch": 6.91, + "learning_rate": 9.267564615944667e-07, + "loss": 0.0537, + "step": 38018 + }, + { + "epoch": 6.91, + "learning_rate": 9.267018565708045e-07, + "loss": 0.0807, + "step": 38019 + }, + { + "epoch": 6.91, + "learning_rate": 9.266472515471425e-07, + "loss": 0.0086, + "step": 38020 + }, + { + "epoch": 6.91, + "learning_rate": 9.265926465234802e-07, + "loss": 0.9111, + "step": 38021 + }, + { + "epoch": 6.91, + "learning_rate": 9.26538041499818e-07, + "loss": 0.3332, + "step": 38022 + }, + { + "epoch": 6.91, + "learning_rate": 9.264834364761559e-07, + "loss": 0.1531, + "step": 38023 + }, + { + "epoch": 6.91, + "learning_rate": 9.264288314524937e-07, + "loss": 0.0054, + "step": 38024 + }, + { + "epoch": 6.91, + "learning_rate": 9.263742264288314e-07, + "loss": 0.0041, + "step": 38025 + }, + { + "epoch": 6.92, + "learning_rate": 9.263196214051693e-07, + "loss": 0.6231, + "step": 38026 + }, + { + "epoch": 6.92, + "learning_rate": 9.262650163815072e-07, + "loss": 0.4542, + "step": 38027 + }, + { + "epoch": 6.92, + "learning_rate": 9.262104113578449e-07, + "loss": 0.0641, + "step": 38028 + }, + { + "epoch": 6.92, + "learning_rate": 9.261558063341828e-07, + "loss": 0.651, + "step": 38029 + }, + { + "epoch": 6.92, + "learning_rate": 9.261012013105207e-07, + "loss": 0.1726, + "step": 38030 + }, + { + "epoch": 6.92, + "learning_rate": 9.260465962868585e-07, + "loss": 0.5583, + "step": 38031 + }, + { + "epoch": 6.92, + "learning_rate": 9.259919912631962e-07, + "loss": 0.0437, + "step": 38032 + }, + { + "epoch": 6.92, + "learning_rate": 9.25937386239534e-07, + "loss": 0.0362, + "step": 38033 + }, + { + "epoch": 6.92, + "learning_rate": 9.25882781215872e-07, + "loss": 0.3082, + "step": 38034 + }, + { + "epoch": 6.92, + "learning_rate": 9.258281761922097e-07, + "loss": 0.8318, + "step": 38035 + }, + { + "epoch": 6.92, + "learning_rate": 9.257735711685475e-07, + "loss": 0.2015, + "step": 38036 + }, + { + "epoch": 6.92, + "learning_rate": 9.257189661448853e-07, + "loss": 0.1823, + "step": 38037 + }, + { + "epoch": 6.92, + "learning_rate": 9.256643611212232e-07, + "loss": 0.4987, + "step": 38038 + }, + { + "epoch": 6.92, + "learning_rate": 9.256097560975609e-07, + "loss": 0.907, + "step": 38039 + }, + { + "epoch": 6.92, + "learning_rate": 9.255551510738988e-07, + "loss": 0.0463, + "step": 38040 + }, + { + "epoch": 6.92, + "learning_rate": 9.255005460502367e-07, + "loss": 0.0244, + "step": 38041 + }, + { + "epoch": 6.92, + "learning_rate": 9.254459410265745e-07, + "loss": 0.0462, + "step": 38042 + }, + { + "epoch": 6.92, + "learning_rate": 9.253913360029123e-07, + "loss": 0.2904, + "step": 38043 + }, + { + "epoch": 6.92, + "learning_rate": 9.2533673097925e-07, + "loss": 1.701, + "step": 38044 + }, + { + "epoch": 6.92, + "learning_rate": 9.25282125955588e-07, + "loss": 1.2019, + "step": 38045 + }, + { + "epoch": 6.92, + "learning_rate": 9.252275209319258e-07, + "loss": 1.4186, + "step": 38046 + }, + { + "epoch": 6.92, + "learning_rate": 9.251729159082635e-07, + "loss": 0.2047, + "step": 38047 + }, + { + "epoch": 6.92, + "learning_rate": 9.251183108846015e-07, + "loss": 0.0173, + "step": 38048 + }, + { + "epoch": 6.92, + "learning_rate": 9.250637058609392e-07, + "loss": 1.3435, + "step": 38049 + }, + { + "epoch": 6.92, + "learning_rate": 9.25009100837277e-07, + "loss": 0.0856, + "step": 38050 + }, + { + "epoch": 6.92, + "learning_rate": 9.249544958136148e-07, + "loss": 2.9326, + "step": 38051 + }, + { + "epoch": 6.92, + "learning_rate": 9.248998907899527e-07, + "loss": 1.3418, + "step": 38052 + }, + { + "epoch": 6.92, + "learning_rate": 9.248452857662906e-07, + "loss": 0.0092, + "step": 38053 + }, + { + "epoch": 6.92, + "learning_rate": 9.247906807426283e-07, + "loss": 0.0146, + "step": 38054 + }, + { + "epoch": 6.92, + "learning_rate": 9.247360757189662e-07, + "loss": 0.7746, + "step": 38055 + }, + { + "epoch": 6.92, + "learning_rate": 9.24681470695304e-07, + "loss": 0.5748, + "step": 38056 + }, + { + "epoch": 6.92, + "learning_rate": 9.246268656716418e-07, + "loss": 1.031, + "step": 38057 + }, + { + "epoch": 6.92, + "learning_rate": 9.245722606479795e-07, + "loss": 0.9038, + "step": 38058 + }, + { + "epoch": 6.92, + "learning_rate": 9.245176556243175e-07, + "loss": 0.2857, + "step": 38059 + }, + { + "epoch": 6.92, + "learning_rate": 9.244630506006553e-07, + "loss": 0.8171, + "step": 38060 + }, + { + "epoch": 6.92, + "learning_rate": 9.24408445576993e-07, + "loss": 1.3156, + "step": 38061 + }, + { + "epoch": 6.92, + "learning_rate": 9.24353840553331e-07, + "loss": 0.613, + "step": 38062 + }, + { + "epoch": 6.92, + "learning_rate": 9.242992355296687e-07, + "loss": 1.5877, + "step": 38063 + }, + { + "epoch": 6.92, + "learning_rate": 9.242446305060066e-07, + "loss": 1.3233, + "step": 38064 + }, + { + "epoch": 6.92, + "learning_rate": 9.241900254823443e-07, + "loss": 1.4695, + "step": 38065 + }, + { + "epoch": 6.92, + "learning_rate": 9.241354204586822e-07, + "loss": 0.009, + "step": 38066 + }, + { + "epoch": 6.92, + "learning_rate": 9.240808154350201e-07, + "loss": 0.135, + "step": 38067 + }, + { + "epoch": 6.92, + "learning_rate": 9.240262104113578e-07, + "loss": 0.2319, + "step": 38068 + }, + { + "epoch": 6.92, + "learning_rate": 9.239716053876957e-07, + "loss": 1.4211, + "step": 38069 + }, + { + "epoch": 6.92, + "learning_rate": 9.239170003640336e-07, + "loss": 0.1145, + "step": 38070 + }, + { + "epoch": 6.92, + "learning_rate": 9.238623953403713e-07, + "loss": 1.3716, + "step": 38071 + }, + { + "epoch": 6.92, + "learning_rate": 9.23807790316709e-07, + "loss": 0.9294, + "step": 38072 + }, + { + "epoch": 6.92, + "learning_rate": 9.23753185293047e-07, + "loss": 2.8881, + "step": 38073 + }, + { + "epoch": 6.92, + "learning_rate": 9.236985802693848e-07, + "loss": 0.1791, + "step": 38074 + }, + { + "epoch": 6.92, + "learning_rate": 9.236439752457226e-07, + "loss": 3.1477, + "step": 38075 + }, + { + "epoch": 6.92, + "learning_rate": 9.235893702220605e-07, + "loss": 0.5616, + "step": 38076 + }, + { + "epoch": 6.92, + "learning_rate": 9.235347651983983e-07, + "loss": 0.5987, + "step": 38077 + }, + { + "epoch": 6.92, + "learning_rate": 9.234801601747361e-07, + "loss": 2.5603, + "step": 38078 + }, + { + "epoch": 6.92, + "learning_rate": 9.234255551510738e-07, + "loss": 0.274, + "step": 38079 + }, + { + "epoch": 6.92, + "learning_rate": 9.233709501274117e-07, + "loss": 1.364, + "step": 38080 + }, + { + "epoch": 6.93, + "learning_rate": 9.233163451037496e-07, + "loss": 0.3599, + "step": 38081 + }, + { + "epoch": 6.93, + "learning_rate": 9.232617400800873e-07, + "loss": 0.7266, + "step": 38082 + }, + { + "epoch": 6.93, + "learning_rate": 9.232071350564253e-07, + "loss": 0.2824, + "step": 38083 + }, + { + "epoch": 6.93, + "learning_rate": 9.231525300327631e-07, + "loss": 0.0451, + "step": 38084 + }, + { + "epoch": 6.93, + "learning_rate": 9.230979250091008e-07, + "loss": 1.2411, + "step": 38085 + }, + { + "epoch": 6.93, + "learning_rate": 9.230433199854387e-07, + "loss": 5.1076, + "step": 38086 + }, + { + "epoch": 6.93, + "learning_rate": 9.229887149617765e-07, + "loss": 0.0157, + "step": 38087 + }, + { + "epoch": 6.93, + "learning_rate": 9.229341099381143e-07, + "loss": 1.4723, + "step": 38088 + }, + { + "epoch": 6.93, + "learning_rate": 9.228795049144521e-07, + "loss": 0.0308, + "step": 38089 + }, + { + "epoch": 6.93, + "learning_rate": 9.2282489989079e-07, + "loss": 0.694, + "step": 38090 + }, + { + "epoch": 6.93, + "learning_rate": 9.227702948671279e-07, + "loss": 0.7144, + "step": 38091 + }, + { + "epoch": 6.93, + "learning_rate": 9.227156898434656e-07, + "loss": 1.1811, + "step": 38092 + }, + { + "epoch": 6.93, + "learning_rate": 9.226610848198034e-07, + "loss": 0.3746, + "step": 38093 + }, + { + "epoch": 6.93, + "learning_rate": 9.226064797961414e-07, + "loss": 0.959, + "step": 38094 + }, + { + "epoch": 6.93, + "learning_rate": 9.225518747724791e-07, + "loss": 0.0647, + "step": 38095 + }, + { + "epoch": 6.93, + "learning_rate": 9.224972697488168e-07, + "loss": 0.1708, + "step": 38096 + }, + { + "epoch": 6.93, + "learning_rate": 9.224426647251548e-07, + "loss": 0.0357, + "step": 38097 + }, + { + "epoch": 6.93, + "learning_rate": 9.223880597014926e-07, + "loss": 1.3475, + "step": 38098 + }, + { + "epoch": 6.93, + "learning_rate": 9.223334546778303e-07, + "loss": 0.756, + "step": 38099 + }, + { + "epoch": 6.93, + "learning_rate": 9.222788496541682e-07, + "loss": 0.4301, + "step": 38100 + }, + { + "epoch": 6.93, + "learning_rate": 9.222242446305061e-07, + "loss": 0.0779, + "step": 38101 + }, + { + "epoch": 6.93, + "learning_rate": 9.221696396068439e-07, + "loss": 0.1888, + "step": 38102 + }, + { + "epoch": 6.93, + "learning_rate": 9.221150345831816e-07, + "loss": 0.0916, + "step": 38103 + }, + { + "epoch": 6.93, + "learning_rate": 9.220604295595195e-07, + "loss": 0.0151, + "step": 38104 + }, + { + "epoch": 6.93, + "learning_rate": 9.220058245358574e-07, + "loss": 0.0091, + "step": 38105 + }, + { + "epoch": 6.93, + "learning_rate": 9.219512195121951e-07, + "loss": 0.0127, + "step": 38106 + }, + { + "epoch": 6.93, + "learning_rate": 9.218966144885329e-07, + "loss": 0.0514, + "step": 38107 + }, + { + "epoch": 6.93, + "learning_rate": 9.218420094648709e-07, + "loss": 0.4752, + "step": 38108 + }, + { + "epoch": 6.93, + "learning_rate": 9.217874044412086e-07, + "loss": 0.0127, + "step": 38109 + }, + { + "epoch": 6.93, + "learning_rate": 9.217327994175465e-07, + "loss": 0.0463, + "step": 38110 + }, + { + "epoch": 6.93, + "learning_rate": 9.216781943938844e-07, + "loss": 0.4937, + "step": 38111 + }, + { + "epoch": 6.93, + "learning_rate": 9.216235893702221e-07, + "loss": 0.0031, + "step": 38112 + }, + { + "epoch": 6.93, + "learning_rate": 9.215689843465599e-07, + "loss": 0.2913, + "step": 38113 + }, + { + "epoch": 6.93, + "learning_rate": 9.215143793228977e-07, + "loss": 0.253, + "step": 38114 + }, + { + "epoch": 6.93, + "learning_rate": 9.214597742992356e-07, + "loss": 0.2431, + "step": 38115 + }, + { + "epoch": 6.93, + "learning_rate": 9.214051692755734e-07, + "loss": 1.5035, + "step": 38116 + }, + { + "epoch": 6.93, + "learning_rate": 9.213505642519112e-07, + "loss": 0.0039, + "step": 38117 + }, + { + "epoch": 6.93, + "learning_rate": 9.212959592282489e-07, + "loss": 0.0094, + "step": 38118 + }, + { + "epoch": 6.93, + "learning_rate": 9.212413542045869e-07, + "loss": 1.8973, + "step": 38119 + }, + { + "epoch": 6.93, + "learning_rate": 9.211867491809246e-07, + "loss": 0.0061, + "step": 38120 + }, + { + "epoch": 6.93, + "learning_rate": 9.211321441572625e-07, + "loss": 1.123, + "step": 38121 + }, + { + "epoch": 6.93, + "learning_rate": 9.210775391336004e-07, + "loss": 0.196, + "step": 38122 + }, + { + "epoch": 6.93, + "learning_rate": 9.210229341099381e-07, + "loss": 0.9253, + "step": 38123 + }, + { + "epoch": 6.93, + "learning_rate": 9.20968329086276e-07, + "loss": 0.056, + "step": 38124 + }, + { + "epoch": 6.93, + "learning_rate": 9.209137240626137e-07, + "loss": 0.943, + "step": 38125 + }, + { + "epoch": 6.93, + "learning_rate": 9.208591190389516e-07, + "loss": 0.2436, + "step": 38126 + }, + { + "epoch": 6.93, + "learning_rate": 9.208045140152894e-07, + "loss": 2.553, + "step": 38127 + }, + { + "epoch": 6.93, + "learning_rate": 9.207499089916272e-07, + "loss": 0.3961, + "step": 38128 + }, + { + "epoch": 6.93, + "learning_rate": 9.206953039679652e-07, + "loss": 0.0128, + "step": 38129 + }, + { + "epoch": 6.93, + "learning_rate": 9.206406989443029e-07, + "loss": 1.8952, + "step": 38130 + }, + { + "epoch": 6.93, + "learning_rate": 9.205860939206407e-07, + "loss": 1.5672, + "step": 38131 + }, + { + "epoch": 6.93, + "learning_rate": 9.205314888969785e-07, + "loss": 0.0116, + "step": 38132 + }, + { + "epoch": 6.93, + "learning_rate": 9.204768838733164e-07, + "loss": 0.3079, + "step": 38133 + }, + { + "epoch": 6.93, + "learning_rate": 9.204222788496542e-07, + "loss": 0.1378, + "step": 38134 + }, + { + "epoch": 6.93, + "learning_rate": 9.20367673825992e-07, + "loss": 0.006, + "step": 38135 + }, + { + "epoch": 6.94, + "learning_rate": 9.203130688023299e-07, + "loss": 0.4654, + "step": 38136 + }, + { + "epoch": 6.94, + "learning_rate": 9.202584637786676e-07, + "loss": 3.8145, + "step": 38137 + }, + { + "epoch": 6.94, + "learning_rate": 9.202038587550055e-07, + "loss": 0.1033, + "step": 38138 + }, + { + "epoch": 6.94, + "learning_rate": 9.201492537313432e-07, + "loss": 0.4216, + "step": 38139 + }, + { + "epoch": 6.94, + "learning_rate": 9.200946487076812e-07, + "loss": 0.162, + "step": 38140 + }, + { + "epoch": 6.94, + "learning_rate": 9.20040043684019e-07, + "loss": 0.2638, + "step": 38141 + }, + { + "epoch": 6.94, + "learning_rate": 9.199854386603567e-07, + "loss": 0.0418, + "step": 38142 + }, + { + "epoch": 6.94, + "learning_rate": 9.199308336366947e-07, + "loss": 3.5617, + "step": 38143 + }, + { + "epoch": 6.94, + "learning_rate": 9.198762286130324e-07, + "loss": 0.6305, + "step": 38144 + }, + { + "epoch": 6.94, + "learning_rate": 9.198216235893702e-07, + "loss": 0.224, + "step": 38145 + }, + { + "epoch": 6.94, + "learning_rate": 9.19767018565708e-07, + "loss": 0.4195, + "step": 38146 + }, + { + "epoch": 6.94, + "learning_rate": 9.197124135420459e-07, + "loss": 0.7669, + "step": 38147 + }, + { + "epoch": 6.94, + "learning_rate": 9.196578085183837e-07, + "loss": 0.012, + "step": 38148 + }, + { + "epoch": 6.94, + "learning_rate": 9.196032034947215e-07, + "loss": 0.0078, + "step": 38149 + }, + { + "epoch": 6.94, + "learning_rate": 9.195485984710594e-07, + "loss": 0.0534, + "step": 38150 + }, + { + "epoch": 6.94, + "learning_rate": 9.194939934473973e-07, + "loss": 0.0143, + "step": 38151 + }, + { + "epoch": 6.94, + "learning_rate": 9.19439388423735e-07, + "loss": 0.0136, + "step": 38152 + }, + { + "epoch": 6.94, + "learning_rate": 9.193847834000727e-07, + "loss": 0.0136, + "step": 38153 + }, + { + "epoch": 6.94, + "learning_rate": 9.193301783764107e-07, + "loss": 0.0153, + "step": 38154 + }, + { + "epoch": 6.94, + "learning_rate": 9.192755733527485e-07, + "loss": 0.0972, + "step": 38155 + }, + { + "epoch": 6.94, + "learning_rate": 9.192209683290862e-07, + "loss": 1.1736, + "step": 38156 + }, + { + "epoch": 6.94, + "learning_rate": 9.191663633054242e-07, + "loss": 0.1847, + "step": 38157 + }, + { + "epoch": 6.94, + "learning_rate": 9.19111758281762e-07, + "loss": 0.2669, + "step": 38158 + }, + { + "epoch": 6.94, + "learning_rate": 9.190571532580997e-07, + "loss": 0.0045, + "step": 38159 + }, + { + "epoch": 6.94, + "learning_rate": 9.190025482344375e-07, + "loss": 0.0067, + "step": 38160 + }, + { + "epoch": 6.94, + "learning_rate": 9.189479432107754e-07, + "loss": 0.3877, + "step": 38161 + }, + { + "epoch": 6.94, + "learning_rate": 9.188933381871133e-07, + "loss": 2.2603, + "step": 38162 + }, + { + "epoch": 6.94, + "learning_rate": 9.18838733163451e-07, + "loss": 0.4501, + "step": 38163 + }, + { + "epoch": 6.94, + "learning_rate": 9.187841281397889e-07, + "loss": 0.0109, + "step": 38164 + }, + { + "epoch": 6.94, + "learning_rate": 9.187295231161268e-07, + "loss": 0.61, + "step": 38165 + }, + { + "epoch": 6.94, + "learning_rate": 9.186749180924645e-07, + "loss": 0.5171, + "step": 38166 + }, + { + "epoch": 6.94, + "learning_rate": 9.186203130688022e-07, + "loss": 1.2015, + "step": 38167 + }, + { + "epoch": 6.94, + "learning_rate": 9.185657080451402e-07, + "loss": 0.3116, + "step": 38168 + }, + { + "epoch": 6.94, + "learning_rate": 9.18511103021478e-07, + "loss": 0.0603, + "step": 38169 + }, + { + "epoch": 6.94, + "learning_rate": 9.184564979978157e-07, + "loss": 0.0079, + "step": 38170 + }, + { + "epoch": 6.94, + "learning_rate": 9.184018929741537e-07, + "loss": 1.0136, + "step": 38171 + }, + { + "epoch": 6.94, + "learning_rate": 9.183472879504915e-07, + "loss": 0.4391, + "step": 38172 + }, + { + "epoch": 6.94, + "learning_rate": 9.182926829268293e-07, + "loss": 0.6532, + "step": 38173 + }, + { + "epoch": 6.94, + "learning_rate": 9.182380779031671e-07, + "loss": 0.0047, + "step": 38174 + }, + { + "epoch": 6.94, + "learning_rate": 9.18183472879505e-07, + "loss": 2.3499, + "step": 38175 + }, + { + "epoch": 6.94, + "learning_rate": 9.181288678558428e-07, + "loss": 1.3539, + "step": 38176 + }, + { + "epoch": 6.94, + "learning_rate": 9.180742628321805e-07, + "loss": 0.0102, + "step": 38177 + }, + { + "epoch": 6.94, + "learning_rate": 9.180196578085184e-07, + "loss": 0.265, + "step": 38178 + }, + { + "epoch": 6.94, + "learning_rate": 9.179650527848563e-07, + "loss": 0.2109, + "step": 38179 + }, + { + "epoch": 6.94, + "learning_rate": 9.17910447761194e-07, + "loss": 1.671, + "step": 38180 + }, + { + "epoch": 6.94, + "learning_rate": 9.178558427375319e-07, + "loss": 0.0429, + "step": 38181 + }, + { + "epoch": 6.94, + "learning_rate": 9.178012377138698e-07, + "loss": 1.4116, + "step": 38182 + }, + { + "epoch": 6.94, + "learning_rate": 9.177466326902075e-07, + "loss": 2.0614, + "step": 38183 + }, + { + "epoch": 6.94, + "learning_rate": 9.176920276665453e-07, + "loss": 0.0688, + "step": 38184 + }, + { + "epoch": 6.94, + "learning_rate": 9.176374226428832e-07, + "loss": 0.7559, + "step": 38185 + }, + { + "epoch": 6.94, + "learning_rate": 9.17582817619221e-07, + "loss": 0.4167, + "step": 38186 + }, + { + "epoch": 6.94, + "learning_rate": 9.175282125955588e-07, + "loss": 0.8639, + "step": 38187 + }, + { + "epoch": 6.94, + "learning_rate": 9.174736075718966e-07, + "loss": 1.1822, + "step": 38188 + }, + { + "epoch": 6.94, + "learning_rate": 9.174190025482346e-07, + "loss": 0.0391, + "step": 38189 + }, + { + "epoch": 6.94, + "learning_rate": 9.173643975245723e-07, + "loss": 0.5903, + "step": 38190 + }, + { + "epoch": 6.95, + "learning_rate": 9.1730979250091e-07, + "loss": 3.6752, + "step": 38191 + }, + { + "epoch": 6.95, + "learning_rate": 9.172551874772479e-07, + "loss": 0.0648, + "step": 38192 + }, + { + "epoch": 6.95, + "learning_rate": 9.172005824535858e-07, + "loss": 0.8412, + "step": 38193 + }, + { + "epoch": 6.95, + "learning_rate": 9.171459774299235e-07, + "loss": 0.0848, + "step": 38194 + }, + { + "epoch": 6.95, + "learning_rate": 9.170913724062614e-07, + "loss": 0.1381, + "step": 38195 + }, + { + "epoch": 6.95, + "learning_rate": 9.170367673825993e-07, + "loss": 1.207, + "step": 38196 + }, + { + "epoch": 6.95, + "learning_rate": 9.16982162358937e-07, + "loss": 0.1304, + "step": 38197 + }, + { + "epoch": 6.95, + "learning_rate": 9.169275573352749e-07, + "loss": 1.367, + "step": 38198 + }, + { + "epoch": 6.95, + "learning_rate": 9.168729523116126e-07, + "loss": 0.1456, + "step": 38199 + }, + { + "epoch": 6.95, + "learning_rate": 9.168183472879506e-07, + "loss": 0.7946, + "step": 38200 + }, + { + "epoch": 6.95, + "learning_rate": 9.167637422642883e-07, + "loss": 0.0746, + "step": 38201 + }, + { + "epoch": 6.95, + "learning_rate": 9.167091372406261e-07, + "loss": 0.1097, + "step": 38202 + }, + { + "epoch": 6.95, + "learning_rate": 9.166545322169641e-07, + "loss": 0.4552, + "step": 38203 + }, + { + "epoch": 6.95, + "learning_rate": 9.165999271933018e-07, + "loss": 0.6818, + "step": 38204 + }, + { + "epoch": 6.95, + "learning_rate": 9.165453221696396e-07, + "loss": 2.4549, + "step": 38205 + }, + { + "epoch": 6.95, + "learning_rate": 9.164907171459774e-07, + "loss": 0.3976, + "step": 38206 + }, + { + "epoch": 6.95, + "learning_rate": 9.164361121223153e-07, + "loss": 0.6524, + "step": 38207 + }, + { + "epoch": 6.95, + "learning_rate": 9.16381507098653e-07, + "loss": 0.506, + "step": 38208 + }, + { + "epoch": 6.95, + "learning_rate": 9.163269020749909e-07, + "loss": 0.5799, + "step": 38209 + }, + { + "epoch": 6.95, + "learning_rate": 9.162722970513288e-07, + "loss": 1.622, + "step": 38210 + }, + { + "epoch": 6.95, + "learning_rate": 9.162176920276666e-07, + "loss": 0.2752, + "step": 38211 + }, + { + "epoch": 6.95, + "learning_rate": 9.161630870040044e-07, + "loss": 0.2269, + "step": 38212 + }, + { + "epoch": 6.95, + "learning_rate": 9.161084819803421e-07, + "loss": 0.6031, + "step": 38213 + }, + { + "epoch": 6.95, + "learning_rate": 9.160538769566801e-07, + "loss": 0.2281, + "step": 38214 + }, + { + "epoch": 6.95, + "learning_rate": 9.159992719330179e-07, + "loss": 0.3056, + "step": 38215 + }, + { + "epoch": 6.95, + "learning_rate": 9.159446669093556e-07, + "loss": 0.1234, + "step": 38216 + }, + { + "epoch": 6.95, + "learning_rate": 9.158900618856936e-07, + "loss": 0.837, + "step": 38217 + }, + { + "epoch": 6.95, + "learning_rate": 9.158354568620313e-07, + "loss": 0.0421, + "step": 38218 + }, + { + "epoch": 6.95, + "learning_rate": 9.157808518383692e-07, + "loss": 3.5427, + "step": 38219 + }, + { + "epoch": 6.95, + "learning_rate": 9.157262468147069e-07, + "loss": 0.6551, + "step": 38220 + }, + { + "epoch": 6.95, + "learning_rate": 9.156716417910448e-07, + "loss": 4.1356, + "step": 38221 + }, + { + "epoch": 6.95, + "learning_rate": 9.156170367673827e-07, + "loss": 0.0538, + "step": 38222 + }, + { + "epoch": 6.95, + "learning_rate": 9.155624317437204e-07, + "loss": 0.3012, + "step": 38223 + }, + { + "epoch": 6.95, + "learning_rate": 9.155078267200583e-07, + "loss": 1.228, + "step": 38224 + }, + { + "epoch": 6.95, + "learning_rate": 9.154532216963961e-07, + "loss": 0.0188, + "step": 38225 + }, + { + "epoch": 6.95, + "learning_rate": 9.153986166727339e-07, + "loss": 0.5408, + "step": 38226 + }, + { + "epoch": 6.95, + "learning_rate": 9.153440116490716e-07, + "loss": 0.4331, + "step": 38227 + }, + { + "epoch": 6.95, + "learning_rate": 9.152894066254096e-07, + "loss": 0.2368, + "step": 38228 + }, + { + "epoch": 6.95, + "learning_rate": 9.152348016017474e-07, + "loss": 1.9544, + "step": 38229 + }, + { + "epoch": 6.95, + "learning_rate": 9.151801965780852e-07, + "loss": 0.3865, + "step": 38230 + }, + { + "epoch": 6.95, + "learning_rate": 9.151255915544231e-07, + "loss": 0.85, + "step": 38231 + }, + { + "epoch": 6.95, + "learning_rate": 9.150709865307608e-07, + "loss": 0.2123, + "step": 38232 + }, + { + "epoch": 6.95, + "learning_rate": 9.150163815070987e-07, + "loss": 1.2732, + "step": 38233 + }, + { + "epoch": 6.95, + "learning_rate": 9.149617764834364e-07, + "loss": 0.146, + "step": 38234 + }, + { + "epoch": 6.95, + "learning_rate": 9.149071714597743e-07, + "loss": 0.3407, + "step": 38235 + }, + { + "epoch": 6.95, + "learning_rate": 9.148525664361122e-07, + "loss": 0.4647, + "step": 38236 + }, + { + "epoch": 6.95, + "learning_rate": 9.147979614124499e-07, + "loss": 0.3103, + "step": 38237 + }, + { + "epoch": 6.95, + "learning_rate": 9.147433563887879e-07, + "loss": 0.6434, + "step": 38238 + }, + { + "epoch": 6.95, + "learning_rate": 9.146887513651257e-07, + "loss": 0.0946, + "step": 38239 + }, + { + "epoch": 6.95, + "learning_rate": 9.146341463414634e-07, + "loss": 0.2739, + "step": 38240 + }, + { + "epoch": 6.95, + "learning_rate": 9.145795413178012e-07, + "loss": 0.0417, + "step": 38241 + }, + { + "epoch": 6.95, + "learning_rate": 9.145249362941391e-07, + "loss": 1.169, + "step": 38242 + }, + { + "epoch": 6.95, + "learning_rate": 9.144703312704769e-07, + "loss": 0.0694, + "step": 38243 + }, + { + "epoch": 6.95, + "learning_rate": 9.144157262468147e-07, + "loss": 0.9176, + "step": 38244 + }, + { + "epoch": 6.95, + "learning_rate": 9.143611212231526e-07, + "loss": 0.1671, + "step": 38245 + }, + { + "epoch": 6.96, + "learning_rate": 9.143065161994904e-07, + "loss": 0.3219, + "step": 38246 + }, + { + "epoch": 6.96, + "learning_rate": 9.142519111758282e-07, + "loss": 0.0856, + "step": 38247 + }, + { + "epoch": 6.96, + "learning_rate": 9.141973061521659e-07, + "loss": 0.2052, + "step": 38248 + }, + { + "epoch": 6.96, + "learning_rate": 9.141427011285039e-07, + "loss": 0.3751, + "step": 38249 + }, + { + "epoch": 6.96, + "learning_rate": 9.140880961048417e-07, + "loss": 0.0049, + "step": 38250 + }, + { + "epoch": 6.96, + "learning_rate": 9.140334910811794e-07, + "loss": 0.0074, + "step": 38251 + }, + { + "epoch": 6.96, + "learning_rate": 9.139788860575174e-07, + "loss": 0.0718, + "step": 38252 + }, + { + "epoch": 6.96, + "learning_rate": 9.139242810338552e-07, + "loss": 0.4091, + "step": 38253 + }, + { + "epoch": 6.96, + "learning_rate": 9.138696760101929e-07, + "loss": 0.0656, + "step": 38254 + }, + { + "epoch": 6.96, + "learning_rate": 9.138150709865308e-07, + "loss": 0.0136, + "step": 38255 + }, + { + "epoch": 6.96, + "learning_rate": 9.137604659628686e-07, + "loss": 0.041, + "step": 38256 + }, + { + "epoch": 6.96, + "learning_rate": 9.137058609392064e-07, + "loss": 1.9059, + "step": 38257 + }, + { + "epoch": 6.96, + "learning_rate": 9.136512559155442e-07, + "loss": 1.4551, + "step": 38258 + }, + { + "epoch": 6.96, + "learning_rate": 9.135966508918821e-07, + "loss": 0.3652, + "step": 38259 + }, + { + "epoch": 6.96, + "learning_rate": 9.1354204586822e-07, + "loss": 0.4484, + "step": 38260 + }, + { + "epoch": 6.96, + "learning_rate": 9.134874408445577e-07, + "loss": 0.071, + "step": 38261 + }, + { + "epoch": 6.96, + "learning_rate": 9.134328358208955e-07, + "loss": 0.2613, + "step": 38262 + }, + { + "epoch": 6.96, + "learning_rate": 9.133782307972335e-07, + "loss": 0.0053, + "step": 38263 + }, + { + "epoch": 6.96, + "learning_rate": 9.133236257735712e-07, + "loss": 0.0122, + "step": 38264 + }, + { + "epoch": 6.96, + "learning_rate": 9.132690207499089e-07, + "loss": 0.1137, + "step": 38265 + }, + { + "epoch": 6.96, + "learning_rate": 9.132144157262469e-07, + "loss": 0.8727, + "step": 38266 + }, + { + "epoch": 6.96, + "learning_rate": 9.131598107025847e-07, + "loss": 0.7791, + "step": 38267 + }, + { + "epoch": 6.96, + "learning_rate": 9.131052056789224e-07, + "loss": 0.9051, + "step": 38268 + }, + { + "epoch": 6.96, + "learning_rate": 9.130506006552603e-07, + "loss": 0.7455, + "step": 38269 + }, + { + "epoch": 6.96, + "learning_rate": 9.129959956315982e-07, + "loss": 0.3395, + "step": 38270 + }, + { + "epoch": 6.96, + "learning_rate": 9.12941390607936e-07, + "loss": 0.0068, + "step": 38271 + }, + { + "epoch": 6.96, + "learning_rate": 9.128867855842737e-07, + "loss": 0.0273, + "step": 38272 + }, + { + "epoch": 6.96, + "learning_rate": 9.128321805606115e-07, + "loss": 0.1857, + "step": 38273 + }, + { + "epoch": 6.96, + "learning_rate": 9.127775755369495e-07, + "loss": 0.1236, + "step": 38274 + }, + { + "epoch": 6.96, + "learning_rate": 9.127229705132872e-07, + "loss": 0.2643, + "step": 38275 + }, + { + "epoch": 6.96, + "learning_rate": 9.12668365489625e-07, + "loss": 1.3097, + "step": 38276 + }, + { + "epoch": 6.96, + "learning_rate": 9.12613760465963e-07, + "loss": 1.1269, + "step": 38277 + }, + { + "epoch": 6.96, + "learning_rate": 9.125591554423007e-07, + "loss": 0.0608, + "step": 38278 + }, + { + "epoch": 6.96, + "learning_rate": 9.125045504186386e-07, + "loss": 0.0356, + "step": 38279 + }, + { + "epoch": 6.96, + "learning_rate": 9.124499453949763e-07, + "loss": 0.3879, + "step": 38280 + }, + { + "epoch": 6.96, + "learning_rate": 9.123953403713142e-07, + "loss": 2.1336, + "step": 38281 + }, + { + "epoch": 6.96, + "learning_rate": 9.12340735347652e-07, + "loss": 0.1572, + "step": 38282 + }, + { + "epoch": 6.96, + "learning_rate": 9.122861303239898e-07, + "loss": 0.0364, + "step": 38283 + }, + { + "epoch": 6.96, + "learning_rate": 9.122315253003277e-07, + "loss": 0.7862, + "step": 38284 + }, + { + "epoch": 6.96, + "learning_rate": 9.121769202766655e-07, + "loss": 0.2454, + "step": 38285 + }, + { + "epoch": 6.96, + "learning_rate": 9.121223152530033e-07, + "loss": 0.0094, + "step": 38286 + }, + { + "epoch": 6.96, + "learning_rate": 9.12067710229341e-07, + "loss": 3.4669, + "step": 38287 + }, + { + "epoch": 6.96, + "learning_rate": 9.12013105205679e-07, + "loss": 0.6871, + "step": 38288 + }, + { + "epoch": 6.96, + "learning_rate": 9.119585001820167e-07, + "loss": 0.9128, + "step": 38289 + }, + { + "epoch": 6.96, + "learning_rate": 9.119038951583546e-07, + "loss": 2.3337, + "step": 38290 + }, + { + "epoch": 6.96, + "learning_rate": 9.118492901346925e-07, + "loss": 0.0062, + "step": 38291 + }, + { + "epoch": 6.96, + "learning_rate": 9.117946851110302e-07, + "loss": 2.1758, + "step": 38292 + }, + { + "epoch": 6.96, + "learning_rate": 9.117400800873681e-07, + "loss": 0.0059, + "step": 38293 + }, + { + "epoch": 6.96, + "learning_rate": 9.116854750637058e-07, + "loss": 0.0372, + "step": 38294 + }, + { + "epoch": 6.96, + "learning_rate": 9.116308700400437e-07, + "loss": 0.2711, + "step": 38295 + }, + { + "epoch": 6.96, + "learning_rate": 9.115762650163815e-07, + "loss": 0.008, + "step": 38296 + }, + { + "epoch": 6.96, + "learning_rate": 9.115216599927193e-07, + "loss": 0.04, + "step": 38297 + }, + { + "epoch": 6.96, + "learning_rate": 9.114670549690573e-07, + "loss": 0.9722, + "step": 38298 + }, + { + "epoch": 6.96, + "learning_rate": 9.11412449945395e-07, + "loss": 0.0994, + "step": 38299 + }, + { + "epoch": 6.96, + "learning_rate": 9.113578449217328e-07, + "loss": 1.5587, + "step": 38300 + }, + { + "epoch": 6.97, + "learning_rate": 9.113032398980706e-07, + "loss": 1.3313, + "step": 38301 + }, + { + "epoch": 6.97, + "learning_rate": 9.112486348744085e-07, + "loss": 1.1784, + "step": 38302 + }, + { + "epoch": 6.97, + "learning_rate": 9.111940298507463e-07, + "loss": 0.0662, + "step": 38303 + }, + { + "epoch": 6.97, + "learning_rate": 9.111394248270841e-07, + "loss": 0.4374, + "step": 38304 + }, + { + "epoch": 6.97, + "learning_rate": 9.11084819803422e-07, + "loss": 0.4096, + "step": 38305 + }, + { + "epoch": 6.97, + "learning_rate": 9.110302147797597e-07, + "loss": 0.609, + "step": 38306 + }, + { + "epoch": 6.97, + "learning_rate": 9.109756097560976e-07, + "loss": 0.1327, + "step": 38307 + }, + { + "epoch": 6.97, + "learning_rate": 9.109210047324353e-07, + "loss": 0.0058, + "step": 38308 + }, + { + "epoch": 6.97, + "learning_rate": 9.108663997087733e-07, + "loss": 3.0851, + "step": 38309 + }, + { + "epoch": 6.97, + "learning_rate": 9.108117946851111e-07, + "loss": 0.0952, + "step": 38310 + }, + { + "epoch": 6.97, + "learning_rate": 9.107571896614488e-07, + "loss": 2.4071, + "step": 38311 + }, + { + "epoch": 6.97, + "learning_rate": 9.107025846377868e-07, + "loss": 2.5938, + "step": 38312 + }, + { + "epoch": 6.97, + "learning_rate": 9.106479796141245e-07, + "loss": 0.0367, + "step": 38313 + }, + { + "epoch": 6.97, + "learning_rate": 9.105933745904623e-07, + "loss": 0.0064, + "step": 38314 + }, + { + "epoch": 6.97, + "learning_rate": 9.105387695668001e-07, + "loss": 0.119, + "step": 38315 + }, + { + "epoch": 6.97, + "learning_rate": 9.10484164543138e-07, + "loss": 0.1706, + "step": 38316 + }, + { + "epoch": 6.97, + "learning_rate": 9.104295595194759e-07, + "loss": 0.0333, + "step": 38317 + }, + { + "epoch": 6.97, + "learning_rate": 9.103749544958136e-07, + "loss": 0.1368, + "step": 38318 + }, + { + "epoch": 6.97, + "learning_rate": 9.103203494721515e-07, + "loss": 0.0101, + "step": 38319 + }, + { + "epoch": 6.97, + "learning_rate": 9.102657444484894e-07, + "loss": 0.3172, + "step": 38320 + }, + { + "epoch": 6.97, + "learning_rate": 9.102111394248271e-07, + "loss": 0.3062, + "step": 38321 + }, + { + "epoch": 6.97, + "learning_rate": 9.101565344011648e-07, + "loss": 0.5183, + "step": 38322 + }, + { + "epoch": 6.97, + "learning_rate": 9.101019293775028e-07, + "loss": 1.7516, + "step": 38323 + }, + { + "epoch": 6.97, + "learning_rate": 9.100473243538406e-07, + "loss": 0.1414, + "step": 38324 + }, + { + "epoch": 6.97, + "learning_rate": 9.099927193301783e-07, + "loss": 1.4122, + "step": 38325 + }, + { + "epoch": 6.97, + "learning_rate": 9.099381143065163e-07, + "loss": 0.0072, + "step": 38326 + }, + { + "epoch": 6.97, + "learning_rate": 9.098835092828541e-07, + "loss": 0.499, + "step": 38327 + }, + { + "epoch": 6.97, + "learning_rate": 9.098289042591919e-07, + "loss": 1.8862, + "step": 38328 + }, + { + "epoch": 6.97, + "learning_rate": 9.097742992355296e-07, + "loss": 1.3864, + "step": 38329 + }, + { + "epoch": 6.97, + "learning_rate": 9.097196942118675e-07, + "loss": 1.0622, + "step": 38330 + }, + { + "epoch": 6.97, + "learning_rate": 9.096650891882054e-07, + "loss": 0.1683, + "step": 38331 + }, + { + "epoch": 6.97, + "learning_rate": 9.096104841645431e-07, + "loss": 1.1155, + "step": 38332 + }, + { + "epoch": 6.97, + "learning_rate": 9.09555879140881e-07, + "loss": 0.0615, + "step": 38333 + }, + { + "epoch": 6.97, + "learning_rate": 9.095012741172189e-07, + "loss": 3.1431, + "step": 38334 + }, + { + "epoch": 6.97, + "learning_rate": 9.094466690935566e-07, + "loss": 0.171, + "step": 38335 + }, + { + "epoch": 6.97, + "learning_rate": 9.093920640698943e-07, + "loss": 0.0231, + "step": 38336 + }, + { + "epoch": 6.97, + "learning_rate": 9.093374590462323e-07, + "loss": 0.1686, + "step": 38337 + }, + { + "epoch": 6.97, + "learning_rate": 9.092828540225701e-07, + "loss": 0.2175, + "step": 38338 + }, + { + "epoch": 6.97, + "learning_rate": 9.092282489989079e-07, + "loss": 0.0603, + "step": 38339 + }, + { + "epoch": 6.97, + "learning_rate": 9.091736439752458e-07, + "loss": 2.0535, + "step": 38340 + }, + { + "epoch": 6.97, + "learning_rate": 9.091190389515836e-07, + "loss": 0.2297, + "step": 38341 + }, + { + "epoch": 6.97, + "learning_rate": 9.090644339279214e-07, + "loss": 0.0649, + "step": 38342 + }, + { + "epoch": 6.97, + "learning_rate": 9.090098289042592e-07, + "loss": 0.3699, + "step": 38343 + }, + { + "epoch": 6.97, + "learning_rate": 9.089552238805971e-07, + "loss": 0.2481, + "step": 38344 + }, + { + "epoch": 6.97, + "learning_rate": 9.089006188569349e-07, + "loss": 1.1575, + "step": 38345 + }, + { + "epoch": 6.97, + "learning_rate": 9.088460138332726e-07, + "loss": 0.3374, + "step": 38346 + }, + { + "epoch": 6.97, + "learning_rate": 9.087914088096104e-07, + "loss": 0.2005, + "step": 38347 + }, + { + "epoch": 6.97, + "learning_rate": 9.087368037859484e-07, + "loss": 0.5288, + "step": 38348 + }, + { + "epoch": 6.97, + "learning_rate": 9.086821987622861e-07, + "loss": 0.827, + "step": 38349 + }, + { + "epoch": 6.97, + "learning_rate": 9.08627593738624e-07, + "loss": 0.9018, + "step": 38350 + }, + { + "epoch": 6.97, + "learning_rate": 9.085729887149619e-07, + "loss": 0.92, + "step": 38351 + }, + { + "epoch": 6.97, + "learning_rate": 9.085183836912996e-07, + "loss": 0.123, + "step": 38352 + }, + { + "epoch": 6.97, + "learning_rate": 9.084637786676374e-07, + "loss": 0.3831, + "step": 38353 + }, + { + "epoch": 6.97, + "learning_rate": 9.084091736439752e-07, + "loss": 0.1524, + "step": 38354 + }, + { + "epoch": 6.97, + "learning_rate": 9.083545686203131e-07, + "loss": 0.9804, + "step": 38355 + }, + { + "epoch": 6.98, + "learning_rate": 9.082999635966509e-07, + "loss": 0.0288, + "step": 38356 + }, + { + "epoch": 6.98, + "learning_rate": 9.082453585729887e-07, + "loss": 1.2223, + "step": 38357 + }, + { + "epoch": 6.98, + "learning_rate": 9.081907535493267e-07, + "loss": 0.015, + "step": 38358 + }, + { + "epoch": 6.98, + "learning_rate": 9.081361485256644e-07, + "loss": 0.0387, + "step": 38359 + }, + { + "epoch": 6.98, + "learning_rate": 9.080815435020021e-07, + "loss": 0.3566, + "step": 38360 + }, + { + "epoch": 6.98, + "learning_rate": 9.0802693847834e-07, + "loss": 0.9447, + "step": 38361 + }, + { + "epoch": 6.98, + "learning_rate": 9.079723334546779e-07, + "loss": 0.0069, + "step": 38362 + }, + { + "epoch": 6.98, + "learning_rate": 9.079177284310156e-07, + "loss": 0.1865, + "step": 38363 + }, + { + "epoch": 6.98, + "learning_rate": 9.078631234073535e-07, + "loss": 1.1329, + "step": 38364 + }, + { + "epoch": 6.98, + "learning_rate": 9.078085183836914e-07, + "loss": 0.453, + "step": 38365 + }, + { + "epoch": 6.98, + "learning_rate": 9.077539133600291e-07, + "loss": 0.0093, + "step": 38366 + }, + { + "epoch": 6.98, + "learning_rate": 9.07699308336367e-07, + "loss": 1.8566, + "step": 38367 + }, + { + "epoch": 6.98, + "learning_rate": 9.076447033127047e-07, + "loss": 3.5452, + "step": 38368 + }, + { + "epoch": 6.98, + "learning_rate": 9.075900982890427e-07, + "loss": 0.4924, + "step": 38369 + }, + { + "epoch": 6.98, + "learning_rate": 9.075354932653804e-07, + "loss": 0.4846, + "step": 38370 + }, + { + "epoch": 6.98, + "learning_rate": 9.074808882417182e-07, + "loss": 0.0563, + "step": 38371 + }, + { + "epoch": 6.98, + "learning_rate": 9.074262832180562e-07, + "loss": 0.1341, + "step": 38372 + }, + { + "epoch": 6.98, + "learning_rate": 9.073716781943939e-07, + "loss": 0.0116, + "step": 38373 + }, + { + "epoch": 6.98, + "learning_rate": 9.073170731707317e-07, + "loss": 0.6657, + "step": 38374 + }, + { + "epoch": 6.98, + "learning_rate": 9.072624681470695e-07, + "loss": 0.6249, + "step": 38375 + }, + { + "epoch": 6.98, + "learning_rate": 9.072078631234074e-07, + "loss": 3.0024, + "step": 38376 + }, + { + "epoch": 6.98, + "learning_rate": 9.071532580997452e-07, + "loss": 0.78, + "step": 38377 + }, + { + "epoch": 6.98, + "learning_rate": 9.07098653076083e-07, + "loss": 1.5031, + "step": 38378 + }, + { + "epoch": 6.98, + "learning_rate": 9.070440480524209e-07, + "loss": 0.0666, + "step": 38379 + }, + { + "epoch": 6.98, + "learning_rate": 9.069894430287587e-07, + "loss": 0.5491, + "step": 38380 + }, + { + "epoch": 6.98, + "learning_rate": 9.069348380050965e-07, + "loss": 0.0367, + "step": 38381 + }, + { + "epoch": 6.98, + "learning_rate": 9.068802329814342e-07, + "loss": 2.0253, + "step": 38382 + }, + { + "epoch": 6.98, + "learning_rate": 9.068256279577722e-07, + "loss": 0.0991, + "step": 38383 + }, + { + "epoch": 6.98, + "learning_rate": 9.0677102293411e-07, + "loss": 0.5559, + "step": 38384 + }, + { + "epoch": 6.98, + "learning_rate": 9.067164179104477e-07, + "loss": 0.1146, + "step": 38385 + }, + { + "epoch": 6.98, + "learning_rate": 9.066618128867857e-07, + "loss": 1.4913, + "step": 38386 + }, + { + "epoch": 6.98, + "learning_rate": 9.066072078631234e-07, + "loss": 0.0111, + "step": 38387 + }, + { + "epoch": 6.98, + "learning_rate": 9.065526028394613e-07, + "loss": 0.6406, + "step": 38388 + }, + { + "epoch": 6.98, + "learning_rate": 9.06497997815799e-07, + "loss": 0.2876, + "step": 38389 + }, + { + "epoch": 6.98, + "learning_rate": 9.064433927921369e-07, + "loss": 0.4248, + "step": 38390 + }, + { + "epoch": 6.98, + "learning_rate": 9.063887877684748e-07, + "loss": 0.2868, + "step": 38391 + }, + { + "epoch": 6.98, + "learning_rate": 9.063341827448125e-07, + "loss": 0.379, + "step": 38392 + }, + { + "epoch": 6.98, + "learning_rate": 9.062795777211504e-07, + "loss": 0.104, + "step": 38393 + }, + { + "epoch": 6.98, + "learning_rate": 9.062249726974882e-07, + "loss": 0.5382, + "step": 38394 + }, + { + "epoch": 6.98, + "learning_rate": 9.06170367673826e-07, + "loss": 1.2179, + "step": 38395 + }, + { + "epoch": 6.98, + "learning_rate": 9.061157626501637e-07, + "loss": 0.0055, + "step": 38396 + }, + { + "epoch": 6.98, + "learning_rate": 9.060611576265017e-07, + "loss": 1.5957, + "step": 38397 + }, + { + "epoch": 6.98, + "learning_rate": 9.060065526028395e-07, + "loss": 0.0367, + "step": 38398 + }, + { + "epoch": 6.98, + "learning_rate": 9.059519475791773e-07, + "loss": 0.6037, + "step": 38399 + }, + { + "epoch": 6.98, + "learning_rate": 9.058973425555152e-07, + "loss": 1.3791, + "step": 38400 + }, + { + "epoch": 6.98, + "learning_rate": 9.058427375318529e-07, + "loss": 1.932, + "step": 38401 + }, + { + "epoch": 6.98, + "learning_rate": 9.057881325081908e-07, + "loss": 2.8813, + "step": 38402 + }, + { + "epoch": 6.98, + "learning_rate": 9.057335274845285e-07, + "loss": 0.1472, + "step": 38403 + }, + { + "epoch": 6.98, + "learning_rate": 9.056789224608664e-07, + "loss": 0.6562, + "step": 38404 + }, + { + "epoch": 6.98, + "learning_rate": 9.056243174372043e-07, + "loss": 0.1513, + "step": 38405 + }, + { + "epoch": 6.98, + "learning_rate": 9.05569712413542e-07, + "loss": 2.143, + "step": 38406 + }, + { + "epoch": 6.98, + "learning_rate": 9.0551510738988e-07, + "loss": 0.3187, + "step": 38407 + }, + { + "epoch": 6.98, + "learning_rate": 9.054605023662178e-07, + "loss": 0.5434, + "step": 38408 + }, + { + "epoch": 6.98, + "learning_rate": 9.054058973425555e-07, + "loss": 2.1106, + "step": 38409 + }, + { + "epoch": 6.98, + "learning_rate": 9.053512923188933e-07, + "loss": 1.0182, + "step": 38410 + }, + { + "epoch": 6.99, + "learning_rate": 9.052966872952312e-07, + "loss": 2.6656, + "step": 38411 + }, + { + "epoch": 6.99, + "learning_rate": 9.05242082271569e-07, + "loss": 0.0169, + "step": 38412 + }, + { + "epoch": 6.99, + "learning_rate": 9.051874772479068e-07, + "loss": 3.5094, + "step": 38413 + }, + { + "epoch": 6.99, + "learning_rate": 9.051328722242447e-07, + "loss": 0.0664, + "step": 38414 + }, + { + "epoch": 6.99, + "learning_rate": 9.050782672005826e-07, + "loss": 0.3218, + "step": 38415 + }, + { + "epoch": 6.99, + "learning_rate": 9.050236621769203e-07, + "loss": 0.1448, + "step": 38416 + }, + { + "epoch": 6.99, + "learning_rate": 9.04969057153258e-07, + "loss": 1.3088, + "step": 38417 + }, + { + "epoch": 6.99, + "learning_rate": 9.04914452129596e-07, + "loss": 0.0193, + "step": 38418 + }, + { + "epoch": 6.99, + "learning_rate": 9.048598471059338e-07, + "loss": 0.0933, + "step": 38419 + }, + { + "epoch": 6.99, + "learning_rate": 9.048052420822715e-07, + "loss": 0.0058, + "step": 38420 + }, + { + "epoch": 6.99, + "learning_rate": 9.047506370586095e-07, + "loss": 0.196, + "step": 38421 + }, + { + "epoch": 6.99, + "learning_rate": 9.046960320349473e-07, + "loss": 0.1225, + "step": 38422 + }, + { + "epoch": 6.99, + "learning_rate": 9.04641427011285e-07, + "loss": 0.1312, + "step": 38423 + }, + { + "epoch": 6.99, + "learning_rate": 9.045868219876228e-07, + "loss": 0.1569, + "step": 38424 + }, + { + "epoch": 6.99, + "learning_rate": 9.045322169639607e-07, + "loss": 0.0388, + "step": 38425 + }, + { + "epoch": 6.99, + "learning_rate": 9.044776119402986e-07, + "loss": 0.5408, + "step": 38426 + }, + { + "epoch": 6.99, + "learning_rate": 9.044230069166363e-07, + "loss": 0.1204, + "step": 38427 + }, + { + "epoch": 6.99, + "learning_rate": 9.043684018929741e-07, + "loss": 0.268, + "step": 38428 + }, + { + "epoch": 6.99, + "learning_rate": 9.043137968693121e-07, + "loss": 0.4209, + "step": 38429 + }, + { + "epoch": 6.99, + "learning_rate": 9.042591918456498e-07, + "loss": 0.2756, + "step": 38430 + }, + { + "epoch": 6.99, + "learning_rate": 9.042045868219876e-07, + "loss": 0.6099, + "step": 38431 + }, + { + "epoch": 6.99, + "learning_rate": 9.041499817983256e-07, + "loss": 0.1543, + "step": 38432 + }, + { + "epoch": 6.99, + "learning_rate": 9.040953767746633e-07, + "loss": 1.2642, + "step": 38433 + }, + { + "epoch": 6.99, + "learning_rate": 9.04040771751001e-07, + "loss": 0.06, + "step": 38434 + }, + { + "epoch": 6.99, + "learning_rate": 9.039861667273389e-07, + "loss": 0.0086, + "step": 38435 + }, + { + "epoch": 6.99, + "learning_rate": 9.039315617036768e-07, + "loss": 0.2192, + "step": 38436 + }, + { + "epoch": 6.99, + "learning_rate": 9.038769566800146e-07, + "loss": 0.233, + "step": 38437 + }, + { + "epoch": 6.99, + "learning_rate": 9.038223516563524e-07, + "loss": 1.9725, + "step": 38438 + }, + { + "epoch": 6.99, + "learning_rate": 9.037677466326903e-07, + "loss": 2.7115, + "step": 38439 + }, + { + "epoch": 6.99, + "learning_rate": 9.037131416090281e-07, + "loss": 0.977, + "step": 38440 + }, + { + "epoch": 6.99, + "learning_rate": 9.036585365853658e-07, + "loss": 0.7456, + "step": 38441 + }, + { + "epoch": 6.99, + "learning_rate": 9.036039315617036e-07, + "loss": 1.7799, + "step": 38442 + }, + { + "epoch": 6.99, + "learning_rate": 9.035493265380416e-07, + "loss": 0.1234, + "step": 38443 + }, + { + "epoch": 6.99, + "learning_rate": 9.034947215143793e-07, + "loss": 1.5259, + "step": 38444 + }, + { + "epoch": 6.99, + "learning_rate": 9.034401164907171e-07, + "loss": 0.0442, + "step": 38445 + }, + { + "epoch": 6.99, + "learning_rate": 9.033855114670551e-07, + "loss": 0.0173, + "step": 38446 + }, + { + "epoch": 6.99, + "learning_rate": 9.033309064433928e-07, + "loss": 0.0213, + "step": 38447 + }, + { + "epoch": 6.99, + "learning_rate": 9.032763014197307e-07, + "loss": 0.0671, + "step": 38448 + }, + { + "epoch": 6.99, + "learning_rate": 9.032216963960684e-07, + "loss": 1.1854, + "step": 38449 + }, + { + "epoch": 6.99, + "learning_rate": 9.031670913724063e-07, + "loss": 0.9104, + "step": 38450 + }, + { + "epoch": 6.99, + "learning_rate": 9.031124863487441e-07, + "loss": 0.005, + "step": 38451 + }, + { + "epoch": 6.99, + "learning_rate": 9.030578813250819e-07, + "loss": 1.2382, + "step": 38452 + }, + { + "epoch": 6.99, + "learning_rate": 9.030032763014198e-07, + "loss": 0.0141, + "step": 38453 + }, + { + "epoch": 6.99, + "learning_rate": 9.029486712777576e-07, + "loss": 0.0802, + "step": 38454 + }, + { + "epoch": 6.99, + "learning_rate": 9.028940662540954e-07, + "loss": 0.0513, + "step": 38455 + }, + { + "epoch": 6.99, + "learning_rate": 9.028394612304331e-07, + "loss": 0.1646, + "step": 38456 + }, + { + "epoch": 6.99, + "learning_rate": 9.027848562067711e-07, + "loss": 0.1685, + "step": 38457 + }, + { + "epoch": 6.99, + "learning_rate": 9.027302511831088e-07, + "loss": 0.532, + "step": 38458 + }, + { + "epoch": 6.99, + "learning_rate": 9.026756461594467e-07, + "loss": 0.3574, + "step": 38459 + }, + { + "epoch": 6.99, + "learning_rate": 9.026210411357846e-07, + "loss": 0.1763, + "step": 38460 + }, + { + "epoch": 6.99, + "learning_rate": 9.025664361121223e-07, + "loss": 2.0312, + "step": 38461 + }, + { + "epoch": 6.99, + "learning_rate": 9.025118310884602e-07, + "loss": 1.5293, + "step": 38462 + }, + { + "epoch": 6.99, + "learning_rate": 9.024572260647979e-07, + "loss": 0.2997, + "step": 38463 + }, + { + "epoch": 6.99, + "learning_rate": 9.024026210411358e-07, + "loss": 0.2645, + "step": 38464 + }, + { + "epoch": 6.99, + "learning_rate": 9.023480160174736e-07, + "loss": 0.9492, + "step": 38465 + }, + { + "epoch": 7.0, + "learning_rate": 9.022934109938114e-07, + "loss": 0.5575, + "step": 38466 + }, + { + "epoch": 7.0, + "learning_rate": 9.022388059701494e-07, + "loss": 0.0579, + "step": 38467 + }, + { + "epoch": 7.0, + "learning_rate": 9.021842009464871e-07, + "loss": 3.1128, + "step": 38468 + }, + { + "epoch": 7.0, + "learning_rate": 9.021295959228249e-07, + "loss": 0.0059, + "step": 38469 + }, + { + "epoch": 7.0, + "learning_rate": 9.020749908991627e-07, + "loss": 1.1714, + "step": 38470 + }, + { + "epoch": 7.0, + "learning_rate": 9.020203858755006e-07, + "loss": 1.2849, + "step": 38471 + }, + { + "epoch": 7.0, + "learning_rate": 9.019657808518384e-07, + "loss": 0.1231, + "step": 38472 + }, + { + "epoch": 7.0, + "learning_rate": 9.019111758281762e-07, + "loss": 0.1987, + "step": 38473 + }, + { + "epoch": 7.0, + "learning_rate": 9.018565708045141e-07, + "loss": 2.5446, + "step": 38474 + }, + { + "epoch": 7.0, + "learning_rate": 9.018019657808519e-07, + "loss": 1.3022, + "step": 38475 + }, + { + "epoch": 7.0, + "learning_rate": 9.017473607571897e-07, + "loss": 0.063, + "step": 38476 + }, + { + "epoch": 7.0, + "learning_rate": 9.016927557335274e-07, + "loss": 0.4898, + "step": 38477 + }, + { + "epoch": 7.0, + "learning_rate": 9.016381507098654e-07, + "loss": 0.6171, + "step": 38478 + }, + { + "epoch": 7.0, + "learning_rate": 9.015835456862032e-07, + "loss": 1.2972, + "step": 38479 + }, + { + "epoch": 7.0, + "learning_rate": 9.015289406625409e-07, + "loss": 0.0949, + "step": 38480 + }, + { + "epoch": 7.0, + "learning_rate": 9.014743356388789e-07, + "loss": 0.2111, + "step": 38481 + }, + { + "epoch": 7.0, + "learning_rate": 9.014197306152166e-07, + "loss": 0.0141, + "step": 38482 + }, + { + "epoch": 7.0, + "learning_rate": 9.013651255915544e-07, + "loss": 0.7867, + "step": 38483 + }, + { + "epoch": 7.0, + "learning_rate": 9.013105205678922e-07, + "loss": 0.4737, + "step": 38484 + }, + { + "epoch": 7.0, + "learning_rate": 9.012559155442301e-07, + "loss": 0.2696, + "step": 38485 + }, + { + "epoch": 7.0, + "learning_rate": 9.01201310520568e-07, + "loss": 0.0038, + "step": 38486 + }, + { + "epoch": 7.0, + "learning_rate": 9.011467054969057e-07, + "loss": 0.4115, + "step": 38487 + }, + { + "epoch": 7.0, + "learning_rate": 9.010921004732436e-07, + "loss": 2.7476, + "step": 38488 + }, + { + "epoch": 7.0, + "learning_rate": 9.010374954495814e-07, + "loss": 0.0083, + "step": 38489 + }, + { + "epoch": 7.0, + "learning_rate": 9.009828904259192e-07, + "loss": 0.1506, + "step": 38490 + }, + { + "epoch": 7.0, + "learning_rate": 9.009282854022569e-07, + "loss": 0.1408, + "step": 38491 + }, + { + "epoch": 7.0, + "learning_rate": 9.008736803785949e-07, + "loss": 0.2826, + "step": 38492 + }, + { + "epoch": 7.0, + "learning_rate": 9.008190753549327e-07, + "loss": 0.7603, + "step": 38493 + }, + { + "epoch": 7.0, + "eval_accuracy": 0.8731587561374795, + "eval_loss": 0.615142822265625, + "eval_runtime": 138.81, + "eval_samples_per_second": 17.607, + "eval_steps_per_second": 4.402, + "step": 38493 + }, + { + "epoch": 7.0, + "learning_rate": 9.007644703312704e-07, + "loss": 0.4042, + "step": 38494 + }, + { + "epoch": 7.0, + "learning_rate": 9.007098653076084e-07, + "loss": 0.0376, + "step": 38495 + }, + { + "epoch": 7.0, + "learning_rate": 9.006552602839462e-07, + "loss": 0.2174, + "step": 38496 + }, + { + "epoch": 7.0, + "learning_rate": 9.00600655260284e-07, + "loss": 0.243, + "step": 38497 + }, + { + "epoch": 7.0, + "learning_rate": 9.005460502366217e-07, + "loss": 0.0512, + "step": 38498 + }, + { + "epoch": 7.0, + "learning_rate": 9.004914452129596e-07, + "loss": 0.0724, + "step": 38499 + }, + { + "epoch": 7.0, + "learning_rate": 9.004368401892975e-07, + "loss": 0.4848, + "step": 38500 + }, + { + "epoch": 7.0, + "learning_rate": 9.003822351656352e-07, + "loss": 0.0944, + "step": 38501 + }, + { + "epoch": 7.0, + "learning_rate": 9.00327630141973e-07, + "loss": 0.4445, + "step": 38502 + }, + { + "epoch": 7.0, + "learning_rate": 9.00273025118311e-07, + "loss": 0.0151, + "step": 38503 + }, + { + "epoch": 7.0, + "learning_rate": 9.002184200946487e-07, + "loss": 0.543, + "step": 38504 + }, + { + "epoch": 7.0, + "learning_rate": 9.001638150709864e-07, + "loss": 0.2112, + "step": 38505 + }, + { + "epoch": 7.0, + "learning_rate": 9.001092100473244e-07, + "loss": 1.2572, + "step": 38506 + }, + { + "epoch": 7.0, + "learning_rate": 9.000546050236622e-07, + "loss": 0.012, + "step": 38507 + }, + { + "epoch": 7.0, + "learning_rate": 9e-07, + "loss": 0.5363, + "step": 38508 + }, + { + "epoch": 7.0, + "learning_rate": 8.999453949763378e-07, + "loss": 0.2893, + "step": 38509 + }, + { + "epoch": 7.0, + "learning_rate": 8.998907899526757e-07, + "loss": 0.0544, + "step": 38510 + }, + { + "epoch": 7.0, + "learning_rate": 8.998361849290135e-07, + "loss": 0.2405, + "step": 38511 + }, + { + "epoch": 7.0, + "learning_rate": 8.997815799053513e-07, + "loss": 1.3016, + "step": 38512 + }, + { + "epoch": 7.0, + "learning_rate": 8.997269748816893e-07, + "loss": 0.6386, + "step": 38513 + }, + { + "epoch": 7.0, + "learning_rate": 8.99672369858027e-07, + "loss": 0.8204, + "step": 38514 + }, + { + "epoch": 7.0, + "learning_rate": 8.996177648343647e-07, + "loss": 0.5775, + "step": 38515 + }, + { + "epoch": 7.0, + "learning_rate": 8.995631598107026e-07, + "loss": 0.899, + "step": 38516 + }, + { + "epoch": 7.0, + "learning_rate": 8.995085547870405e-07, + "loss": 0.2594, + "step": 38517 + }, + { + "epoch": 7.0, + "learning_rate": 8.994539497633782e-07, + "loss": 0.0999, + "step": 38518 + }, + { + "epoch": 7.0, + "learning_rate": 8.993993447397161e-07, + "loss": 0.0119, + "step": 38519 + }, + { + "epoch": 7.0, + "learning_rate": 8.99344739716054e-07, + "loss": 1.2447, + "step": 38520 + }, + { + "epoch": 7.01, + "learning_rate": 8.992901346923917e-07, + "loss": 0.5799, + "step": 38521 + }, + { + "epoch": 7.01, + "learning_rate": 8.992355296687295e-07, + "loss": 0.3331, + "step": 38522 + }, + { + "epoch": 7.01, + "learning_rate": 8.991809246450673e-07, + "loss": 0.1765, + "step": 38523 + }, + { + "epoch": 7.01, + "learning_rate": 8.991263196214053e-07, + "loss": 0.0744, + "step": 38524 + }, + { + "epoch": 7.01, + "learning_rate": 8.99071714597743e-07, + "loss": 0.153, + "step": 38525 + }, + { + "epoch": 7.01, + "learning_rate": 8.990171095740808e-07, + "loss": 0.8074, + "step": 38526 + }, + { + "epoch": 7.01, + "learning_rate": 8.989625045504188e-07, + "loss": 0.0317, + "step": 38527 + }, + { + "epoch": 7.01, + "learning_rate": 8.989078995267565e-07, + "loss": 0.1318, + "step": 38528 + }, + { + "epoch": 7.01, + "learning_rate": 8.988532945030942e-07, + "loss": 2.0339, + "step": 38529 + }, + { + "epoch": 7.01, + "learning_rate": 8.987986894794321e-07, + "loss": 1.3043, + "step": 38530 + }, + { + "epoch": 7.01, + "learning_rate": 8.9874408445577e-07, + "loss": 0.0159, + "step": 38531 + }, + { + "epoch": 7.01, + "learning_rate": 8.986894794321077e-07, + "loss": 1.309, + "step": 38532 + }, + { + "epoch": 7.01, + "learning_rate": 8.986348744084456e-07, + "loss": 0.0168, + "step": 38533 + }, + { + "epoch": 7.01, + "learning_rate": 8.985802693847835e-07, + "loss": 0.3741, + "step": 38534 + }, + { + "epoch": 7.01, + "learning_rate": 8.985256643611213e-07, + "loss": 0.2125, + "step": 38535 + }, + { + "epoch": 7.01, + "learning_rate": 8.984710593374591e-07, + "loss": 1.0441, + "step": 38536 + }, + { + "epoch": 7.01, + "learning_rate": 8.984164543137968e-07, + "loss": 2.2264, + "step": 38537 + }, + { + "epoch": 7.01, + "learning_rate": 8.983618492901348e-07, + "loss": 3.4033, + "step": 38538 + }, + { + "epoch": 7.01, + "learning_rate": 8.983072442664725e-07, + "loss": 0.019, + "step": 38539 + }, + { + "epoch": 7.01, + "learning_rate": 8.982526392428103e-07, + "loss": 1.5409, + "step": 38540 + }, + { + "epoch": 7.01, + "learning_rate": 8.981980342191483e-07, + "loss": 0.5126, + "step": 38541 + }, + { + "epoch": 7.01, + "learning_rate": 8.98143429195486e-07, + "loss": 1.0673, + "step": 38542 + }, + { + "epoch": 7.01, + "learning_rate": 8.980888241718238e-07, + "loss": 0.6811, + "step": 38543 + }, + { + "epoch": 7.01, + "learning_rate": 8.980342191481616e-07, + "loss": 0.0035, + "step": 38544 + }, + { + "epoch": 7.01, + "learning_rate": 8.979796141244995e-07, + "loss": 1.5667, + "step": 38545 + }, + { + "epoch": 7.01, + "learning_rate": 8.979250091008373e-07, + "loss": 0.5295, + "step": 38546 + }, + { + "epoch": 7.01, + "learning_rate": 8.978704040771751e-07, + "loss": 0.0947, + "step": 38547 + }, + { + "epoch": 7.01, + "learning_rate": 8.97815799053513e-07, + "loss": 0.285, + "step": 38548 + }, + { + "epoch": 7.01, + "learning_rate": 8.977611940298508e-07, + "loss": 0.0525, + "step": 38549 + }, + { + "epoch": 7.01, + "learning_rate": 8.977065890061886e-07, + "loss": 0.4404, + "step": 38550 + }, + { + "epoch": 7.01, + "learning_rate": 8.976519839825263e-07, + "loss": 0.3262, + "step": 38551 + }, + { + "epoch": 7.01, + "learning_rate": 8.975973789588643e-07, + "loss": 0.0897, + "step": 38552 + }, + { + "epoch": 7.01, + "learning_rate": 8.975427739352021e-07, + "loss": 0.0085, + "step": 38553 + }, + { + "epoch": 7.01, + "learning_rate": 8.974881689115398e-07, + "loss": 0.1565, + "step": 38554 + }, + { + "epoch": 7.01, + "learning_rate": 8.974335638878778e-07, + "loss": 1.0817, + "step": 38555 + }, + { + "epoch": 7.01, + "learning_rate": 8.973789588642155e-07, + "loss": 0.2085, + "step": 38556 + }, + { + "epoch": 7.01, + "learning_rate": 8.973243538405534e-07, + "loss": 0.7108, + "step": 38557 + }, + { + "epoch": 7.01, + "learning_rate": 8.972697488168911e-07, + "loss": 0.0092, + "step": 38558 + }, + { + "epoch": 7.01, + "learning_rate": 8.97215143793229e-07, + "loss": 0.2464, + "step": 38559 + }, + { + "epoch": 7.01, + "learning_rate": 8.971605387695669e-07, + "loss": 0.0061, + "step": 38560 + }, + { + "epoch": 7.01, + "learning_rate": 8.971059337459046e-07, + "loss": 0.0883, + "step": 38561 + }, + { + "epoch": 7.01, + "learning_rate": 8.970513287222425e-07, + "loss": 0.6446, + "step": 38562 + }, + { + "epoch": 7.01, + "learning_rate": 8.969967236985803e-07, + "loss": 0.0061, + "step": 38563 + }, + { + "epoch": 7.01, + "learning_rate": 8.969421186749181e-07, + "loss": 0.3399, + "step": 38564 + }, + { + "epoch": 7.01, + "learning_rate": 8.968875136512559e-07, + "loss": 0.0232, + "step": 38565 + }, + { + "epoch": 7.01, + "learning_rate": 8.968329086275938e-07, + "loss": 0.0031, + "step": 38566 + }, + { + "epoch": 7.01, + "learning_rate": 8.967783036039316e-07, + "loss": 0.645, + "step": 38567 + }, + { + "epoch": 7.01, + "learning_rate": 8.967236985802694e-07, + "loss": 0.0436, + "step": 38568 + }, + { + "epoch": 7.01, + "learning_rate": 8.966690935566073e-07, + "loss": 1.1339, + "step": 38569 + }, + { + "epoch": 7.01, + "learning_rate": 8.96614488532945e-07, + "loss": 0.0621, + "step": 38570 + }, + { + "epoch": 7.01, + "learning_rate": 8.965598835092829e-07, + "loss": 0.0095, + "step": 38571 + }, + { + "epoch": 7.01, + "learning_rate": 8.965052784856206e-07, + "loss": 3.8909, + "step": 38572 + }, + { + "epoch": 7.01, + "learning_rate": 8.964506734619586e-07, + "loss": 0.2635, + "step": 38573 + }, + { + "epoch": 7.01, + "learning_rate": 8.963960684382964e-07, + "loss": 0.9476, + "step": 38574 + }, + { + "epoch": 7.01, + "learning_rate": 8.963414634146341e-07, + "loss": 0.3786, + "step": 38575 + }, + { + "epoch": 7.02, + "learning_rate": 8.962868583909721e-07, + "loss": 0.4761, + "step": 38576 + }, + { + "epoch": 7.02, + "learning_rate": 8.962322533673099e-07, + "loss": 0.1578, + "step": 38577 + }, + { + "epoch": 7.02, + "learning_rate": 8.961776483436476e-07, + "loss": 0.1319, + "step": 38578 + }, + { + "epoch": 7.02, + "learning_rate": 8.961230433199854e-07, + "loss": 0.0616, + "step": 38579 + }, + { + "epoch": 7.02, + "learning_rate": 8.960684382963233e-07, + "loss": 1.8584, + "step": 38580 + }, + { + "epoch": 7.02, + "learning_rate": 8.960138332726611e-07, + "loss": 1.8737, + "step": 38581 + }, + { + "epoch": 7.02, + "learning_rate": 8.959592282489989e-07, + "loss": 0.2206, + "step": 38582 + }, + { + "epoch": 7.02, + "learning_rate": 8.959046232253367e-07, + "loss": 0.0538, + "step": 38583 + }, + { + "epoch": 7.02, + "learning_rate": 8.958500182016747e-07, + "loss": 0.6797, + "step": 38584 + }, + { + "epoch": 7.02, + "learning_rate": 8.957954131780124e-07, + "loss": 2.4795, + "step": 38585 + }, + { + "epoch": 7.02, + "learning_rate": 8.957408081543501e-07, + "loss": 0.0385, + "step": 38586 + }, + { + "epoch": 7.02, + "learning_rate": 8.956862031306881e-07, + "loss": 0.0088, + "step": 38587 + }, + { + "epoch": 7.02, + "learning_rate": 8.956315981070259e-07, + "loss": 0.2964, + "step": 38588 + }, + { + "epoch": 7.02, + "learning_rate": 8.955769930833636e-07, + "loss": 0.0068, + "step": 38589 + }, + { + "epoch": 7.02, + "learning_rate": 8.955223880597015e-07, + "loss": 0.0173, + "step": 38590 + }, + { + "epoch": 7.02, + "learning_rate": 8.954677830360394e-07, + "loss": 0.0027, + "step": 38591 + }, + { + "epoch": 7.02, + "learning_rate": 8.954131780123771e-07, + "loss": 3.1772, + "step": 38592 + }, + { + "epoch": 7.02, + "learning_rate": 8.953585729887149e-07, + "loss": 1.0517, + "step": 38593 + }, + { + "epoch": 7.02, + "learning_rate": 8.953039679650528e-07, + "loss": 0.0095, + "step": 38594 + }, + { + "epoch": 7.02, + "learning_rate": 8.952493629413907e-07, + "loss": 0.0987, + "step": 38595 + }, + { + "epoch": 7.02, + "learning_rate": 8.951947579177284e-07, + "loss": 0.9798, + "step": 38596 + }, + { + "epoch": 7.02, + "learning_rate": 8.951401528940662e-07, + "loss": 0.1729, + "step": 38597 + }, + { + "epoch": 7.02, + "learning_rate": 8.950855478704042e-07, + "loss": 0.4785, + "step": 38598 + }, + { + "epoch": 7.02, + "learning_rate": 8.950309428467419e-07, + "loss": 0.2741, + "step": 38599 + }, + { + "epoch": 7.02, + "learning_rate": 8.949763378230797e-07, + "loss": 0.8744, + "step": 38600 + }, + { + "epoch": 7.02, + "learning_rate": 8.949217327994177e-07, + "loss": 0.1061, + "step": 38601 + }, + { + "epoch": 7.02, + "learning_rate": 8.948671277757554e-07, + "loss": 0.0052, + "step": 38602 + }, + { + "epoch": 7.02, + "learning_rate": 8.948125227520931e-07, + "loss": 0.138, + "step": 38603 + }, + { + "epoch": 7.02, + "learning_rate": 8.94757917728431e-07, + "loss": 0.0164, + "step": 38604 + }, + { + "epoch": 7.02, + "learning_rate": 8.947033127047689e-07, + "loss": 1.7941, + "step": 38605 + }, + { + "epoch": 7.02, + "learning_rate": 8.946487076811067e-07, + "loss": 0.4061, + "step": 38606 + }, + { + "epoch": 7.02, + "learning_rate": 8.945941026574445e-07, + "loss": 0.2646, + "step": 38607 + }, + { + "epoch": 7.02, + "learning_rate": 8.945394976337824e-07, + "loss": 0.6676, + "step": 38608 + }, + { + "epoch": 7.02, + "learning_rate": 8.944848926101202e-07, + "loss": 0.6536, + "step": 38609 + }, + { + "epoch": 7.02, + "learning_rate": 8.944302875864579e-07, + "loss": 0.0502, + "step": 38610 + }, + { + "epoch": 7.02, + "learning_rate": 8.943756825627957e-07, + "loss": 1.0992, + "step": 38611 + }, + { + "epoch": 7.02, + "learning_rate": 8.943210775391337e-07, + "loss": 0.402, + "step": 38612 + }, + { + "epoch": 7.02, + "learning_rate": 8.942664725154714e-07, + "loss": 0.6311, + "step": 38613 + }, + { + "epoch": 7.02, + "learning_rate": 8.942118674918093e-07, + "loss": 0.3407, + "step": 38614 + }, + { + "epoch": 7.02, + "learning_rate": 8.941572624681472e-07, + "loss": 0.0819, + "step": 38615 + }, + { + "epoch": 7.02, + "learning_rate": 8.941026574444849e-07, + "loss": 0.1588, + "step": 38616 + }, + { + "epoch": 7.02, + "learning_rate": 8.940480524208228e-07, + "loss": 0.2139, + "step": 38617 + }, + { + "epoch": 7.02, + "learning_rate": 8.939934473971605e-07, + "loss": 1.0431, + "step": 38618 + }, + { + "epoch": 7.02, + "learning_rate": 8.939388423734984e-07, + "loss": 0.1844, + "step": 38619 + }, + { + "epoch": 7.02, + "learning_rate": 8.938842373498362e-07, + "loss": 0.015, + "step": 38620 + }, + { + "epoch": 7.02, + "learning_rate": 8.93829632326174e-07, + "loss": 0.3783, + "step": 38621 + }, + { + "epoch": 7.02, + "learning_rate": 8.93775027302512e-07, + "loss": 0.8722, + "step": 38622 + }, + { + "epoch": 7.02, + "learning_rate": 8.937204222788497e-07, + "loss": 0.006, + "step": 38623 + }, + { + "epoch": 7.02, + "learning_rate": 8.936658172551875e-07, + "loss": 2.4719, + "step": 38624 + }, + { + "epoch": 7.02, + "learning_rate": 8.936112122315253e-07, + "loss": 0.3926, + "step": 38625 + }, + { + "epoch": 7.02, + "learning_rate": 8.935566072078632e-07, + "loss": 0.2131, + "step": 38626 + }, + { + "epoch": 7.02, + "learning_rate": 8.935020021842009e-07, + "loss": 0.0276, + "step": 38627 + }, + { + "epoch": 7.02, + "learning_rate": 8.934473971605388e-07, + "loss": 0.0726, + "step": 38628 + }, + { + "epoch": 7.02, + "learning_rate": 8.933927921368767e-07, + "loss": 0.0982, + "step": 38629 + }, + { + "epoch": 7.02, + "learning_rate": 8.933381871132144e-07, + "loss": 0.0493, + "step": 38630 + }, + { + "epoch": 7.03, + "learning_rate": 8.932835820895523e-07, + "loss": 0.4705, + "step": 38631 + }, + { + "epoch": 7.03, + "learning_rate": 8.9322897706589e-07, + "loss": 1.2417, + "step": 38632 + }, + { + "epoch": 7.03, + "learning_rate": 8.93174372042228e-07, + "loss": 0.5425, + "step": 38633 + }, + { + "epoch": 7.03, + "learning_rate": 8.931197670185657e-07, + "loss": 0.003, + "step": 38634 + }, + { + "epoch": 7.03, + "learning_rate": 8.930651619949035e-07, + "loss": 0.2862, + "step": 38635 + }, + { + "epoch": 7.03, + "learning_rate": 8.930105569712415e-07, + "loss": 1.1691, + "step": 38636 + }, + { + "epoch": 7.03, + "learning_rate": 8.929559519475792e-07, + "loss": 0.5146, + "step": 38637 + }, + { + "epoch": 7.03, + "learning_rate": 8.92901346923917e-07, + "loss": 0.1927, + "step": 38638 + }, + { + "epoch": 7.03, + "learning_rate": 8.928467419002548e-07, + "loss": 0.0658, + "step": 38639 + }, + { + "epoch": 7.03, + "learning_rate": 8.927921368765927e-07, + "loss": 0.2995, + "step": 38640 + }, + { + "epoch": 7.03, + "learning_rate": 8.927375318529305e-07, + "loss": 1.1898, + "step": 38641 + }, + { + "epoch": 7.03, + "learning_rate": 8.926829268292683e-07, + "loss": 0.0067, + "step": 38642 + }, + { + "epoch": 7.03, + "learning_rate": 8.926283218056062e-07, + "loss": 2.0143, + "step": 38643 + }, + { + "epoch": 7.03, + "learning_rate": 8.92573716781944e-07, + "loss": 0.0743, + "step": 38644 + }, + { + "epoch": 7.03, + "learning_rate": 8.925191117582818e-07, + "loss": 0.8334, + "step": 38645 + }, + { + "epoch": 7.03, + "learning_rate": 8.924645067346195e-07, + "loss": 0.1068, + "step": 38646 + }, + { + "epoch": 7.03, + "learning_rate": 8.924099017109575e-07, + "loss": 0.0512, + "step": 38647 + }, + { + "epoch": 7.03, + "learning_rate": 8.923552966872953e-07, + "loss": 2.1526, + "step": 38648 + }, + { + "epoch": 7.03, + "learning_rate": 8.92300691663633e-07, + "loss": 0.1134, + "step": 38649 + }, + { + "epoch": 7.03, + "learning_rate": 8.92246086639971e-07, + "loss": 0.0793, + "step": 38650 + }, + { + "epoch": 7.03, + "learning_rate": 8.921914816163087e-07, + "loss": 0.0579, + "step": 38651 + }, + { + "epoch": 7.03, + "learning_rate": 8.921368765926465e-07, + "loss": 0.0125, + "step": 38652 + }, + { + "epoch": 7.03, + "learning_rate": 8.920822715689843e-07, + "loss": 0.016, + "step": 38653 + }, + { + "epoch": 7.03, + "learning_rate": 8.920276665453222e-07, + "loss": 0.0394, + "step": 38654 + }, + { + "epoch": 7.03, + "learning_rate": 8.919730615216601e-07, + "loss": 0.2682, + "step": 38655 + }, + { + "epoch": 7.03, + "learning_rate": 8.919184564979978e-07, + "loss": 0.1392, + "step": 38656 + }, + { + "epoch": 7.03, + "learning_rate": 8.918638514743355e-07, + "loss": 0.1284, + "step": 38657 + }, + { + "epoch": 7.03, + "learning_rate": 8.918092464506735e-07, + "loss": 1.0751, + "step": 38658 + }, + { + "epoch": 7.03, + "learning_rate": 8.917546414270113e-07, + "loss": 1.8445, + "step": 38659 + }, + { + "epoch": 7.03, + "learning_rate": 8.91700036403349e-07, + "loss": 0.571, + "step": 38660 + }, + { + "epoch": 7.03, + "learning_rate": 8.91645431379687e-07, + "loss": 0.0364, + "step": 38661 + }, + { + "epoch": 7.03, + "learning_rate": 8.915908263560248e-07, + "loss": 0.1424, + "step": 38662 + }, + { + "epoch": 7.03, + "learning_rate": 8.915362213323626e-07, + "loss": 0.0524, + "step": 38663 + }, + { + "epoch": 7.03, + "learning_rate": 8.914816163087004e-07, + "loss": 0.0074, + "step": 38664 + }, + { + "epoch": 7.03, + "learning_rate": 8.914270112850383e-07, + "loss": 0.6685, + "step": 38665 + }, + { + "epoch": 7.03, + "learning_rate": 8.913724062613761e-07, + "loss": 0.4012, + "step": 38666 + }, + { + "epoch": 7.03, + "learning_rate": 8.913178012377138e-07, + "loss": 0.2235, + "step": 38667 + }, + { + "epoch": 7.03, + "learning_rate": 8.912631962140517e-07, + "loss": 0.5341, + "step": 38668 + }, + { + "epoch": 7.03, + "learning_rate": 8.912085911903896e-07, + "loss": 1.884, + "step": 38669 + }, + { + "epoch": 7.03, + "learning_rate": 8.911539861667273e-07, + "loss": 0.0075, + "step": 38670 + }, + { + "epoch": 7.03, + "learning_rate": 8.910993811430651e-07, + "loss": 0.4875, + "step": 38671 + }, + { + "epoch": 7.03, + "learning_rate": 8.910447761194031e-07, + "loss": 0.2248, + "step": 38672 + }, + { + "epoch": 7.03, + "learning_rate": 8.909901710957408e-07, + "loss": 0.5718, + "step": 38673 + }, + { + "epoch": 7.03, + "learning_rate": 8.909355660720786e-07, + "loss": 0.1703, + "step": 38674 + }, + { + "epoch": 7.03, + "learning_rate": 8.908809610484165e-07, + "loss": 0.0089, + "step": 38675 + }, + { + "epoch": 7.03, + "learning_rate": 8.908263560247543e-07, + "loss": 0.3451, + "step": 38676 + }, + { + "epoch": 7.03, + "learning_rate": 8.907717510010921e-07, + "loss": 0.0269, + "step": 38677 + }, + { + "epoch": 7.03, + "learning_rate": 8.907171459774299e-07, + "loss": 1.1875, + "step": 38678 + }, + { + "epoch": 7.03, + "learning_rate": 8.906625409537678e-07, + "loss": 0.2795, + "step": 38679 + }, + { + "epoch": 7.03, + "learning_rate": 8.906079359301056e-07, + "loss": 0.5475, + "step": 38680 + }, + { + "epoch": 7.03, + "learning_rate": 8.905533309064434e-07, + "loss": 0.7735, + "step": 38681 + }, + { + "epoch": 7.03, + "learning_rate": 8.904987258827814e-07, + "loss": 0.1976, + "step": 38682 + }, + { + "epoch": 7.03, + "learning_rate": 8.904441208591191e-07, + "loss": 0.6722, + "step": 38683 + }, + { + "epoch": 7.03, + "learning_rate": 8.903895158354568e-07, + "loss": 0.0174, + "step": 38684 + }, + { + "epoch": 7.03, + "learning_rate": 8.903349108117947e-07, + "loss": 0.0161, + "step": 38685 + }, + { + "epoch": 7.04, + "learning_rate": 8.902803057881326e-07, + "loss": 0.3716, + "step": 38686 + }, + { + "epoch": 7.04, + "learning_rate": 8.902257007644703e-07, + "loss": 0.5073, + "step": 38687 + }, + { + "epoch": 7.04, + "learning_rate": 8.901710957408082e-07, + "loss": 0.551, + "step": 38688 + }, + { + "epoch": 7.04, + "learning_rate": 8.901164907171461e-07, + "loss": 0.2008, + "step": 38689 + }, + { + "epoch": 7.04, + "learning_rate": 8.900618856934838e-07, + "loss": 0.7342, + "step": 38690 + }, + { + "epoch": 7.04, + "learning_rate": 8.900072806698216e-07, + "loss": 0.0047, + "step": 38691 + }, + { + "epoch": 7.04, + "learning_rate": 8.899526756461594e-07, + "loss": 0.0844, + "step": 38692 + }, + { + "epoch": 7.04, + "learning_rate": 8.898980706224974e-07, + "loss": 0.348, + "step": 38693 + }, + { + "epoch": 7.04, + "learning_rate": 8.898434655988351e-07, + "loss": 0.0203, + "step": 38694 + }, + { + "epoch": 7.04, + "learning_rate": 8.897888605751729e-07, + "loss": 0.0274, + "step": 38695 + }, + { + "epoch": 7.04, + "learning_rate": 8.897342555515109e-07, + "loss": 3.0433, + "step": 38696 + }, + { + "epoch": 7.04, + "learning_rate": 8.896796505278486e-07, + "loss": 1.2417, + "step": 38697 + }, + { + "epoch": 7.04, + "learning_rate": 8.896250455041863e-07, + "loss": 0.008, + "step": 38698 + }, + { + "epoch": 7.04, + "learning_rate": 8.895704404805242e-07, + "loss": 0.065, + "step": 38699 + }, + { + "epoch": 7.04, + "learning_rate": 8.895158354568621e-07, + "loss": 0.2776, + "step": 38700 + }, + { + "epoch": 7.04, + "learning_rate": 8.894612304331998e-07, + "loss": 0.3192, + "step": 38701 + }, + { + "epoch": 7.04, + "learning_rate": 8.894066254095377e-07, + "loss": 1.0091, + "step": 38702 + }, + { + "epoch": 7.04, + "learning_rate": 8.893520203858756e-07, + "loss": 0.2638, + "step": 38703 + }, + { + "epoch": 7.04, + "learning_rate": 8.892974153622134e-07, + "loss": 0.0074, + "step": 38704 + }, + { + "epoch": 7.04, + "learning_rate": 8.892428103385512e-07, + "loss": 0.1501, + "step": 38705 + }, + { + "epoch": 7.04, + "learning_rate": 8.891882053148889e-07, + "loss": 0.0078, + "step": 38706 + }, + { + "epoch": 7.04, + "learning_rate": 8.891336002912269e-07, + "loss": 0.056, + "step": 38707 + }, + { + "epoch": 7.04, + "learning_rate": 8.890789952675646e-07, + "loss": 1.7405, + "step": 38708 + }, + { + "epoch": 7.04, + "learning_rate": 8.890243902439024e-07, + "loss": 0.1431, + "step": 38709 + }, + { + "epoch": 7.04, + "learning_rate": 8.889697852202404e-07, + "loss": 0.8173, + "step": 38710 + }, + { + "epoch": 7.04, + "learning_rate": 8.889151801965781e-07, + "loss": 0.0053, + "step": 38711 + }, + { + "epoch": 7.04, + "learning_rate": 8.88860575172916e-07, + "loss": 0.1566, + "step": 38712 + }, + { + "epoch": 7.04, + "learning_rate": 8.888059701492537e-07, + "loss": 0.3195, + "step": 38713 + }, + { + "epoch": 7.04, + "learning_rate": 8.887513651255916e-07, + "loss": 0.0125, + "step": 38714 + }, + { + "epoch": 7.04, + "learning_rate": 8.886967601019294e-07, + "loss": 0.0639, + "step": 38715 + }, + { + "epoch": 7.04, + "learning_rate": 8.886421550782672e-07, + "loss": 0.0116, + "step": 38716 + }, + { + "epoch": 7.04, + "learning_rate": 8.885875500546051e-07, + "loss": 0.7397, + "step": 38717 + }, + { + "epoch": 7.04, + "learning_rate": 8.885329450309429e-07, + "loss": 0.3385, + "step": 38718 + }, + { + "epoch": 7.04, + "learning_rate": 8.884783400072807e-07, + "loss": 0.0165, + "step": 38719 + }, + { + "epoch": 7.04, + "learning_rate": 8.884237349836184e-07, + "loss": 0.8384, + "step": 38720 + }, + { + "epoch": 7.04, + "learning_rate": 8.883691299599564e-07, + "loss": 0.0403, + "step": 38721 + }, + { + "epoch": 7.04, + "learning_rate": 8.883145249362941e-07, + "loss": 0.0052, + "step": 38722 + }, + { + "epoch": 7.04, + "learning_rate": 8.88259919912632e-07, + "loss": 0.0525, + "step": 38723 + }, + { + "epoch": 7.04, + "learning_rate": 8.882053148889699e-07, + "loss": 0.0149, + "step": 38724 + }, + { + "epoch": 7.04, + "learning_rate": 8.881507098653076e-07, + "loss": 0.0114, + "step": 38725 + }, + { + "epoch": 7.04, + "learning_rate": 8.880961048416455e-07, + "loss": 1.0229, + "step": 38726 + }, + { + "epoch": 7.04, + "learning_rate": 8.880414998179832e-07, + "loss": 0.9804, + "step": 38727 + }, + { + "epoch": 7.04, + "learning_rate": 8.879868947943211e-07, + "loss": 0.0197, + "step": 38728 + }, + { + "epoch": 7.04, + "learning_rate": 8.87932289770659e-07, + "loss": 0.0055, + "step": 38729 + }, + { + "epoch": 7.04, + "learning_rate": 8.878776847469967e-07, + "loss": 0.0081, + "step": 38730 + }, + { + "epoch": 7.04, + "learning_rate": 8.878230797233347e-07, + "loss": 0.5823, + "step": 38731 + }, + { + "epoch": 7.04, + "learning_rate": 8.877684746996724e-07, + "loss": 0.1941, + "step": 38732 + }, + { + "epoch": 7.04, + "learning_rate": 8.877138696760102e-07, + "loss": 0.0593, + "step": 38733 + }, + { + "epoch": 7.04, + "learning_rate": 8.87659264652348e-07, + "loss": 0.5255, + "step": 38734 + }, + { + "epoch": 7.04, + "learning_rate": 8.876046596286859e-07, + "loss": 0.0136, + "step": 38735 + }, + { + "epoch": 7.04, + "learning_rate": 8.875500546050237e-07, + "loss": 0.1628, + "step": 38736 + }, + { + "epoch": 7.04, + "learning_rate": 8.874954495813615e-07, + "loss": 0.463, + "step": 38737 + }, + { + "epoch": 7.04, + "learning_rate": 8.874408445576992e-07, + "loss": 0.4563, + "step": 38738 + }, + { + "epoch": 7.04, + "learning_rate": 8.873862395340371e-07, + "loss": 0.0793, + "step": 38739 + }, + { + "epoch": 7.04, + "learning_rate": 8.87331634510375e-07, + "loss": 0.0278, + "step": 38740 + }, + { + "epoch": 7.05, + "learning_rate": 8.872770294867127e-07, + "loss": 0.9285, + "step": 38741 + }, + { + "epoch": 7.05, + "learning_rate": 8.872224244630507e-07, + "loss": 0.0086, + "step": 38742 + }, + { + "epoch": 7.05, + "learning_rate": 8.871678194393885e-07, + "loss": 0.6476, + "step": 38743 + }, + { + "epoch": 7.05, + "learning_rate": 8.871132144157262e-07, + "loss": 0.1615, + "step": 38744 + }, + { + "epoch": 7.05, + "learning_rate": 8.870586093920641e-07, + "loss": 0.2086, + "step": 38745 + }, + { + "epoch": 7.05, + "learning_rate": 8.87004004368402e-07, + "loss": 0.0823, + "step": 38746 + }, + { + "epoch": 7.05, + "learning_rate": 8.869493993447397e-07, + "loss": 0.6965, + "step": 38747 + }, + { + "epoch": 7.05, + "learning_rate": 8.868947943210775e-07, + "loss": 0.2993, + "step": 38748 + }, + { + "epoch": 7.05, + "learning_rate": 8.868401892974154e-07, + "loss": 0.0285, + "step": 38749 + }, + { + "epoch": 7.05, + "learning_rate": 8.867855842737533e-07, + "loss": 0.0817, + "step": 38750 + }, + { + "epoch": 7.05, + "learning_rate": 8.86730979250091e-07, + "loss": 0.0735, + "step": 38751 + }, + { + "epoch": 7.05, + "learning_rate": 8.866763742264288e-07, + "loss": 0.2674, + "step": 38752 + }, + { + "epoch": 7.05, + "learning_rate": 8.866217692027668e-07, + "loss": 0.0135, + "step": 38753 + }, + { + "epoch": 7.05, + "learning_rate": 8.865671641791045e-07, + "loss": 0.1812, + "step": 38754 + }, + { + "epoch": 7.05, + "learning_rate": 8.865125591554422e-07, + "loss": 0.0029, + "step": 38755 + }, + { + "epoch": 7.05, + "learning_rate": 8.864579541317802e-07, + "loss": 2.2383, + "step": 38756 + }, + { + "epoch": 7.05, + "learning_rate": 8.86403349108118e-07, + "loss": 0.0097, + "step": 38757 + }, + { + "epoch": 7.05, + "learning_rate": 8.863487440844557e-07, + "loss": 0.0055, + "step": 38758 + }, + { + "epoch": 7.05, + "learning_rate": 8.862941390607936e-07, + "loss": 0.0304, + "step": 38759 + }, + { + "epoch": 7.05, + "learning_rate": 8.862395340371315e-07, + "loss": 2.1543, + "step": 38760 + }, + { + "epoch": 7.05, + "learning_rate": 8.861849290134693e-07, + "loss": 0.1148, + "step": 38761 + }, + { + "epoch": 7.05, + "learning_rate": 8.86130323989807e-07, + "loss": 0.6494, + "step": 38762 + }, + { + "epoch": 7.05, + "learning_rate": 8.860757189661449e-07, + "loss": 0.8829, + "step": 38763 + }, + { + "epoch": 7.05, + "learning_rate": 8.860211139424828e-07, + "loss": 2.3027, + "step": 38764 + }, + { + "epoch": 7.05, + "learning_rate": 8.859665089188205e-07, + "loss": 0.0602, + "step": 38765 + }, + { + "epoch": 7.05, + "learning_rate": 8.859119038951583e-07, + "loss": 0.0086, + "step": 38766 + }, + { + "epoch": 7.05, + "learning_rate": 8.858572988714963e-07, + "loss": 0.2327, + "step": 38767 + }, + { + "epoch": 7.05, + "learning_rate": 8.85802693847834e-07, + "loss": 0.6194, + "step": 38768 + }, + { + "epoch": 7.05, + "learning_rate": 8.857480888241718e-07, + "loss": 0.1139, + "step": 38769 + }, + { + "epoch": 7.05, + "learning_rate": 8.856934838005098e-07, + "loss": 0.1164, + "step": 38770 + }, + { + "epoch": 7.05, + "learning_rate": 8.856388787768475e-07, + "loss": 0.0192, + "step": 38771 + }, + { + "epoch": 7.05, + "learning_rate": 8.855842737531853e-07, + "loss": 1.3155, + "step": 38772 + }, + { + "epoch": 7.05, + "learning_rate": 8.855296687295231e-07, + "loss": 0.0209, + "step": 38773 + }, + { + "epoch": 7.05, + "learning_rate": 8.85475063705861e-07, + "loss": 0.9846, + "step": 38774 + }, + { + "epoch": 7.05, + "learning_rate": 8.854204586821988e-07, + "loss": 0.1469, + "step": 38775 + }, + { + "epoch": 7.05, + "learning_rate": 8.853658536585366e-07, + "loss": 1.9255, + "step": 38776 + }, + { + "epoch": 7.05, + "learning_rate": 8.853112486348745e-07, + "loss": 0.117, + "step": 38777 + }, + { + "epoch": 7.05, + "learning_rate": 8.852566436112123e-07, + "loss": 0.0583, + "step": 38778 + }, + { + "epoch": 7.05, + "learning_rate": 8.8520203858755e-07, + "loss": 0.0526, + "step": 38779 + }, + { + "epoch": 7.05, + "learning_rate": 8.851474335638878e-07, + "loss": 2.788, + "step": 38780 + }, + { + "epoch": 7.05, + "learning_rate": 8.850928285402258e-07, + "loss": 0.1702, + "step": 38781 + }, + { + "epoch": 7.05, + "learning_rate": 8.850382235165635e-07, + "loss": 0.1332, + "step": 38782 + }, + { + "epoch": 7.05, + "learning_rate": 8.849836184929014e-07, + "loss": 0.3798, + "step": 38783 + }, + { + "epoch": 7.05, + "learning_rate": 8.849290134692393e-07, + "loss": 0.1271, + "step": 38784 + }, + { + "epoch": 7.05, + "learning_rate": 8.84874408445577e-07, + "loss": 0.0122, + "step": 38785 + }, + { + "epoch": 7.05, + "learning_rate": 8.848198034219149e-07, + "loss": 0.0822, + "step": 38786 + }, + { + "epoch": 7.05, + "learning_rate": 8.847651983982526e-07, + "loss": 0.2441, + "step": 38787 + }, + { + "epoch": 7.05, + "learning_rate": 8.847105933745905e-07, + "loss": 0.6449, + "step": 38788 + }, + { + "epoch": 7.05, + "learning_rate": 8.846559883509283e-07, + "loss": 0.022, + "step": 38789 + }, + { + "epoch": 7.05, + "learning_rate": 8.846013833272661e-07, + "loss": 0.1726, + "step": 38790 + }, + { + "epoch": 7.05, + "learning_rate": 8.845467783036041e-07, + "loss": 2.7497, + "step": 38791 + }, + { + "epoch": 7.05, + "learning_rate": 8.844921732799418e-07, + "loss": 0.0653, + "step": 38792 + }, + { + "epoch": 7.05, + "learning_rate": 8.844375682562796e-07, + "loss": 0.2784, + "step": 38793 + }, + { + "epoch": 7.05, + "learning_rate": 8.843829632326174e-07, + "loss": 0.1279, + "step": 38794 + }, + { + "epoch": 7.05, + "learning_rate": 8.843283582089553e-07, + "loss": 1.0707, + "step": 38795 + }, + { + "epoch": 7.06, + "learning_rate": 8.84273753185293e-07, + "loss": 0.2818, + "step": 38796 + }, + { + "epoch": 7.06, + "learning_rate": 8.842191481616309e-07, + "loss": 1.2485, + "step": 38797 + }, + { + "epoch": 7.06, + "learning_rate": 8.841645431379688e-07, + "loss": 0.46, + "step": 38798 + }, + { + "epoch": 7.06, + "learning_rate": 8.841099381143065e-07, + "loss": 0.0063, + "step": 38799 + }, + { + "epoch": 7.06, + "learning_rate": 8.840553330906444e-07, + "loss": 0.4742, + "step": 38800 + }, + { + "epoch": 7.06, + "learning_rate": 8.840007280669821e-07, + "loss": 0.2788, + "step": 38801 + }, + { + "epoch": 7.06, + "learning_rate": 8.839461230433201e-07, + "loss": 0.4015, + "step": 38802 + }, + { + "epoch": 7.06, + "learning_rate": 8.838915180196578e-07, + "loss": 0.8682, + "step": 38803 + }, + { + "epoch": 7.06, + "learning_rate": 8.838369129959956e-07, + "loss": 0.1689, + "step": 38804 + }, + { + "epoch": 7.06, + "learning_rate": 8.837823079723336e-07, + "loss": 0.2588, + "step": 38805 + }, + { + "epoch": 7.06, + "learning_rate": 8.837277029486713e-07, + "loss": 0.2428, + "step": 38806 + }, + { + "epoch": 7.06, + "learning_rate": 8.836730979250091e-07, + "loss": 0.8042, + "step": 38807 + }, + { + "epoch": 7.06, + "learning_rate": 8.836184929013469e-07, + "loss": 0.4308, + "step": 38808 + }, + { + "epoch": 7.06, + "learning_rate": 8.835638878776848e-07, + "loss": 0.1138, + "step": 38809 + }, + { + "epoch": 7.06, + "learning_rate": 8.835092828540227e-07, + "loss": 0.16, + "step": 38810 + }, + { + "epoch": 7.06, + "learning_rate": 8.834546778303604e-07, + "loss": 1.5394, + "step": 38811 + }, + { + "epoch": 7.06, + "learning_rate": 8.834000728066981e-07, + "loss": 0.6832, + "step": 38812 + }, + { + "epoch": 7.06, + "learning_rate": 8.833454677830361e-07, + "loss": 0.0197, + "step": 38813 + }, + { + "epoch": 7.06, + "learning_rate": 8.832908627593739e-07, + "loss": 0.8097, + "step": 38814 + }, + { + "epoch": 7.06, + "learning_rate": 8.832362577357116e-07, + "loss": 0.783, + "step": 38815 + }, + { + "epoch": 7.06, + "learning_rate": 8.831816527120496e-07, + "loss": 0.0713, + "step": 38816 + }, + { + "epoch": 7.06, + "learning_rate": 8.831270476883874e-07, + "loss": 1.4559, + "step": 38817 + }, + { + "epoch": 7.06, + "learning_rate": 8.830724426647251e-07, + "loss": 0.3424, + "step": 38818 + }, + { + "epoch": 7.06, + "learning_rate": 8.830178376410629e-07, + "loss": 0.0754, + "step": 38819 + }, + { + "epoch": 7.06, + "learning_rate": 8.829632326174008e-07, + "loss": 0.0062, + "step": 38820 + }, + { + "epoch": 7.06, + "learning_rate": 8.829086275937387e-07, + "loss": 0.153, + "step": 38821 + }, + { + "epoch": 7.06, + "learning_rate": 8.828540225700764e-07, + "loss": 1.4421, + "step": 38822 + }, + { + "epoch": 7.06, + "learning_rate": 8.827994175464143e-07, + "loss": 0.0817, + "step": 38823 + }, + { + "epoch": 7.06, + "learning_rate": 8.827448125227522e-07, + "loss": 3.8553, + "step": 38824 + }, + { + "epoch": 7.06, + "learning_rate": 8.826902074990899e-07, + "loss": 0.4729, + "step": 38825 + }, + { + "epoch": 7.06, + "learning_rate": 8.826356024754276e-07, + "loss": 0.2157, + "step": 38826 + }, + { + "epoch": 7.06, + "learning_rate": 8.825809974517656e-07, + "loss": 0.0077, + "step": 38827 + }, + { + "epoch": 7.06, + "learning_rate": 8.825263924281034e-07, + "loss": 1.2698, + "step": 38828 + }, + { + "epoch": 7.06, + "learning_rate": 8.824717874044411e-07, + "loss": 0.0263, + "step": 38829 + }, + { + "epoch": 7.06, + "learning_rate": 8.824171823807791e-07, + "loss": 0.0099, + "step": 38830 + }, + { + "epoch": 7.06, + "learning_rate": 8.823625773571169e-07, + "loss": 0.0832, + "step": 38831 + }, + { + "epoch": 7.06, + "learning_rate": 8.823079723334547e-07, + "loss": 0.437, + "step": 38832 + }, + { + "epoch": 7.06, + "learning_rate": 8.822533673097925e-07, + "loss": 0.5264, + "step": 38833 + }, + { + "epoch": 7.06, + "learning_rate": 8.821987622861304e-07, + "loss": 0.3943, + "step": 38834 + }, + { + "epoch": 7.06, + "learning_rate": 8.821441572624682e-07, + "loss": 0.0034, + "step": 38835 + }, + { + "epoch": 7.06, + "learning_rate": 8.820895522388059e-07, + "loss": 0.7024, + "step": 38836 + }, + { + "epoch": 7.06, + "learning_rate": 8.820349472151438e-07, + "loss": 0.4008, + "step": 38837 + }, + { + "epoch": 7.06, + "learning_rate": 8.819803421914817e-07, + "loss": 0.3117, + "step": 38838 + }, + { + "epoch": 7.06, + "learning_rate": 8.819257371678194e-07, + "loss": 0.9887, + "step": 38839 + }, + { + "epoch": 7.06, + "learning_rate": 8.818711321441573e-07, + "loss": 0.0022, + "step": 38840 + }, + { + "epoch": 7.06, + "learning_rate": 8.818165271204952e-07, + "loss": 0.3915, + "step": 38841 + }, + { + "epoch": 7.06, + "learning_rate": 8.817619220968329e-07, + "loss": 3.4269, + "step": 38842 + }, + { + "epoch": 7.06, + "learning_rate": 8.817073170731707e-07, + "loss": 0.0626, + "step": 38843 + }, + { + "epoch": 7.06, + "learning_rate": 8.816527120495086e-07, + "loss": 0.5141, + "step": 38844 + }, + { + "epoch": 7.06, + "learning_rate": 8.815981070258464e-07, + "loss": 1.8055, + "step": 38845 + }, + { + "epoch": 7.06, + "learning_rate": 8.815435020021842e-07, + "loss": 0.3277, + "step": 38846 + }, + { + "epoch": 7.06, + "learning_rate": 8.81488896978522e-07, + "loss": 0.9619, + "step": 38847 + }, + { + "epoch": 7.06, + "learning_rate": 8.8143429195486e-07, + "loss": 0.479, + "step": 38848 + }, + { + "epoch": 7.06, + "learning_rate": 8.813796869311977e-07, + "loss": 0.0443, + "step": 38849 + }, + { + "epoch": 7.06, + "learning_rate": 8.813250819075355e-07, + "loss": 0.458, + "step": 38850 + }, + { + "epoch": 7.07, + "learning_rate": 8.812704768838735e-07, + "loss": 0.0306, + "step": 38851 + }, + { + "epoch": 7.07, + "learning_rate": 8.812158718602112e-07, + "loss": 1.2638, + "step": 38852 + }, + { + "epoch": 7.07, + "learning_rate": 8.811612668365489e-07, + "loss": 0.2748, + "step": 38853 + }, + { + "epoch": 7.07, + "learning_rate": 8.811066618128868e-07, + "loss": 0.0474, + "step": 38854 + }, + { + "epoch": 7.07, + "learning_rate": 8.810520567892247e-07, + "loss": 2.5152, + "step": 38855 + }, + { + "epoch": 7.07, + "learning_rate": 8.809974517655624e-07, + "loss": 2.0278, + "step": 38856 + }, + { + "epoch": 7.07, + "learning_rate": 8.809428467419003e-07, + "loss": 0.3472, + "step": 38857 + }, + { + "epoch": 7.07, + "learning_rate": 8.808882417182382e-07, + "loss": 0.1745, + "step": 38858 + }, + { + "epoch": 7.07, + "learning_rate": 8.80833636694576e-07, + "loss": 0.0095, + "step": 38859 + }, + { + "epoch": 7.07, + "learning_rate": 8.807790316709137e-07, + "loss": 2.2998, + "step": 38860 + }, + { + "epoch": 7.07, + "learning_rate": 8.807244266472515e-07, + "loss": 0.2085, + "step": 38861 + }, + { + "epoch": 7.07, + "learning_rate": 8.806698216235895e-07, + "loss": 1.52, + "step": 38862 + }, + { + "epoch": 7.07, + "learning_rate": 8.806152165999272e-07, + "loss": 0.0141, + "step": 38863 + }, + { + "epoch": 7.07, + "learning_rate": 8.80560611576265e-07, + "loss": 0.664, + "step": 38864 + }, + { + "epoch": 7.07, + "learning_rate": 8.80506006552603e-07, + "loss": 0.4209, + "step": 38865 + }, + { + "epoch": 7.07, + "learning_rate": 8.804514015289407e-07, + "loss": 0.0431, + "step": 38866 + }, + { + "epoch": 7.07, + "learning_rate": 8.803967965052784e-07, + "loss": 0.0497, + "step": 38867 + }, + { + "epoch": 7.07, + "learning_rate": 8.803421914816163e-07, + "loss": 1.2925, + "step": 38868 + }, + { + "epoch": 7.07, + "learning_rate": 8.802875864579542e-07, + "loss": 0.016, + "step": 38869 + }, + { + "epoch": 7.07, + "learning_rate": 8.80232981434292e-07, + "loss": 0.0261, + "step": 38870 + }, + { + "epoch": 7.07, + "learning_rate": 8.801783764106298e-07, + "loss": 0.5219, + "step": 38871 + }, + { + "epoch": 7.07, + "learning_rate": 8.801237713869677e-07, + "loss": 0.2112, + "step": 38872 + }, + { + "epoch": 7.07, + "learning_rate": 8.800691663633055e-07, + "loss": 0.2216, + "step": 38873 + }, + { + "epoch": 7.07, + "learning_rate": 8.800145613396433e-07, + "loss": 0.881, + "step": 38874 + }, + { + "epoch": 7.07, + "learning_rate": 8.79959956315981e-07, + "loss": 0.433, + "step": 38875 + }, + { + "epoch": 7.07, + "learning_rate": 8.79905351292319e-07, + "loss": 0.2509, + "step": 38876 + }, + { + "epoch": 7.07, + "learning_rate": 8.798507462686567e-07, + "loss": 0.7129, + "step": 38877 + }, + { + "epoch": 7.07, + "learning_rate": 8.797961412449945e-07, + "loss": 0.6092, + "step": 38878 + }, + { + "epoch": 7.07, + "learning_rate": 8.797415362213325e-07, + "loss": 0.3571, + "step": 38879 + }, + { + "epoch": 7.07, + "learning_rate": 8.796869311976702e-07, + "loss": 0.2617, + "step": 38880 + }, + { + "epoch": 7.07, + "learning_rate": 8.796323261740081e-07, + "loss": 0.4234, + "step": 38881 + }, + { + "epoch": 7.07, + "learning_rate": 8.795777211503458e-07, + "loss": 1.3165, + "step": 38882 + }, + { + "epoch": 7.07, + "learning_rate": 8.795231161266837e-07, + "loss": 0.1972, + "step": 38883 + }, + { + "epoch": 7.07, + "learning_rate": 8.794685111030215e-07, + "loss": 0.4283, + "step": 38884 + }, + { + "epoch": 7.07, + "learning_rate": 8.794139060793593e-07, + "loss": 0.0932, + "step": 38885 + }, + { + "epoch": 7.07, + "learning_rate": 8.793593010556972e-07, + "loss": 1.0698, + "step": 38886 + }, + { + "epoch": 7.07, + "learning_rate": 8.79304696032035e-07, + "loss": 0.1287, + "step": 38887 + }, + { + "epoch": 7.07, + "learning_rate": 8.792500910083728e-07, + "loss": 0.0294, + "step": 38888 + }, + { + "epoch": 7.07, + "learning_rate": 8.791954859847105e-07, + "loss": 0.4356, + "step": 38889 + }, + { + "epoch": 7.07, + "learning_rate": 8.791408809610485e-07, + "loss": 0.0087, + "step": 38890 + }, + { + "epoch": 7.07, + "learning_rate": 8.790862759373862e-07, + "loss": 0.6854, + "step": 38891 + }, + { + "epoch": 7.07, + "learning_rate": 8.790316709137241e-07, + "loss": 0.0062, + "step": 38892 + }, + { + "epoch": 7.07, + "learning_rate": 8.789770658900618e-07, + "loss": 0.2256, + "step": 38893 + }, + { + "epoch": 7.07, + "learning_rate": 8.789224608663997e-07, + "loss": 0.0128, + "step": 38894 + }, + { + "epoch": 7.07, + "learning_rate": 8.788678558427376e-07, + "loss": 0.0078, + "step": 38895 + }, + { + "epoch": 7.07, + "learning_rate": 8.788132508190753e-07, + "loss": 0.0053, + "step": 38896 + }, + { + "epoch": 7.07, + "learning_rate": 8.787586457954132e-07, + "loss": 0.227, + "step": 38897 + }, + { + "epoch": 7.07, + "learning_rate": 8.787040407717511e-07, + "loss": 0.5479, + "step": 38898 + }, + { + "epoch": 7.07, + "learning_rate": 8.786494357480888e-07, + "loss": 2.7685, + "step": 38899 + }, + { + "epoch": 7.07, + "learning_rate": 8.785948307244265e-07, + "loss": 0.1845, + "step": 38900 + }, + { + "epoch": 7.07, + "learning_rate": 8.785402257007645e-07, + "loss": 1.5118, + "step": 38901 + }, + { + "epoch": 7.07, + "learning_rate": 8.784856206771023e-07, + "loss": 0.162, + "step": 38902 + }, + { + "epoch": 7.07, + "learning_rate": 8.784310156534401e-07, + "loss": 2.3732, + "step": 38903 + }, + { + "epoch": 7.07, + "learning_rate": 8.78376410629778e-07, + "loss": 0.0503, + "step": 38904 + }, + { + "epoch": 7.07, + "learning_rate": 8.783218056061158e-07, + "loss": 0.4149, + "step": 38905 + }, + { + "epoch": 7.08, + "learning_rate": 8.782672005824536e-07, + "loss": 0.1439, + "step": 38906 + }, + { + "epoch": 7.08, + "learning_rate": 8.782125955587913e-07, + "loss": 0.1028, + "step": 38907 + }, + { + "epoch": 7.08, + "learning_rate": 8.781579905351292e-07, + "loss": 0.1022, + "step": 38908 + }, + { + "epoch": 7.08, + "learning_rate": 8.781033855114671e-07, + "loss": 0.5506, + "step": 38909 + }, + { + "epoch": 7.08, + "learning_rate": 8.780487804878048e-07, + "loss": 0.0088, + "step": 38910 + }, + { + "epoch": 7.08, + "learning_rate": 8.779941754641428e-07, + "loss": 0.0955, + "step": 38911 + }, + { + "epoch": 7.08, + "learning_rate": 8.779395704404806e-07, + "loss": 1.4023, + "step": 38912 + }, + { + "epoch": 7.08, + "learning_rate": 8.778849654168183e-07, + "loss": 0.0067, + "step": 38913 + }, + { + "epoch": 7.08, + "learning_rate": 8.778303603931562e-07, + "loss": 0.0925, + "step": 38914 + }, + { + "epoch": 7.08, + "learning_rate": 8.777757553694941e-07, + "loss": 0.6431, + "step": 38915 + }, + { + "epoch": 7.08, + "learning_rate": 8.777211503458318e-07, + "loss": 0.2359, + "step": 38916 + }, + { + "epoch": 7.08, + "learning_rate": 8.776665453221696e-07, + "loss": 0.5981, + "step": 38917 + }, + { + "epoch": 7.08, + "learning_rate": 8.776119402985075e-07, + "loss": 0.6399, + "step": 38918 + }, + { + "epoch": 7.08, + "learning_rate": 8.775573352748454e-07, + "loss": 0.3003, + "step": 38919 + }, + { + "epoch": 7.08, + "learning_rate": 8.775027302511831e-07, + "loss": 0.2367, + "step": 38920 + }, + { + "epoch": 7.08, + "learning_rate": 8.774481252275209e-07, + "loss": 3.215, + "step": 38921 + }, + { + "epoch": 7.08, + "learning_rate": 8.773935202038589e-07, + "loss": 0.0146, + "step": 38922 + }, + { + "epoch": 7.08, + "learning_rate": 8.773389151801966e-07, + "loss": 0.1682, + "step": 38923 + }, + { + "epoch": 7.08, + "learning_rate": 8.772843101565343e-07, + "loss": 1.2339, + "step": 38924 + }, + { + "epoch": 7.08, + "learning_rate": 8.772297051328723e-07, + "loss": 0.17, + "step": 38925 + }, + { + "epoch": 7.08, + "learning_rate": 8.771751001092101e-07, + "loss": 0.1584, + "step": 38926 + }, + { + "epoch": 7.08, + "learning_rate": 8.771204950855478e-07, + "loss": 0.5769, + "step": 38927 + }, + { + "epoch": 7.08, + "learning_rate": 8.770658900618857e-07, + "loss": 0.0046, + "step": 38928 + }, + { + "epoch": 7.08, + "learning_rate": 8.770112850382236e-07, + "loss": 0.0864, + "step": 38929 + }, + { + "epoch": 7.08, + "learning_rate": 8.769566800145614e-07, + "loss": 0.0408, + "step": 38930 + }, + { + "epoch": 7.08, + "learning_rate": 8.769020749908991e-07, + "loss": 0.341, + "step": 38931 + }, + { + "epoch": 7.08, + "learning_rate": 8.76847469967237e-07, + "loss": 0.5269, + "step": 38932 + }, + { + "epoch": 7.08, + "learning_rate": 8.767928649435749e-07, + "loss": 0.0782, + "step": 38933 + }, + { + "epoch": 7.08, + "learning_rate": 8.767382599199126e-07, + "loss": 0.0976, + "step": 38934 + }, + { + "epoch": 7.08, + "learning_rate": 8.766836548962504e-07, + "loss": 0.2946, + "step": 38935 + }, + { + "epoch": 7.08, + "learning_rate": 8.766290498725884e-07, + "loss": 0.9516, + "step": 38936 + }, + { + "epoch": 7.08, + "learning_rate": 8.765744448489261e-07, + "loss": 1.3294, + "step": 38937 + }, + { + "epoch": 7.08, + "learning_rate": 8.76519839825264e-07, + "loss": 0.1903, + "step": 38938 + }, + { + "epoch": 7.08, + "learning_rate": 8.764652348016019e-07, + "loss": 0.0107, + "step": 38939 + }, + { + "epoch": 7.08, + "learning_rate": 8.764106297779396e-07, + "loss": 0.0788, + "step": 38940 + }, + { + "epoch": 7.08, + "learning_rate": 8.763560247542774e-07, + "loss": 0.1336, + "step": 38941 + }, + { + "epoch": 7.08, + "learning_rate": 8.763014197306152e-07, + "loss": 0.1347, + "step": 38942 + }, + { + "epoch": 7.08, + "learning_rate": 8.762468147069531e-07, + "loss": 0.0624, + "step": 38943 + }, + { + "epoch": 7.08, + "learning_rate": 8.761922096832909e-07, + "loss": 2.557, + "step": 38944 + }, + { + "epoch": 7.08, + "learning_rate": 8.761376046596287e-07, + "loss": 0.0133, + "step": 38945 + }, + { + "epoch": 7.08, + "learning_rate": 8.760829996359667e-07, + "loss": 3.2608, + "step": 38946 + }, + { + "epoch": 7.08, + "learning_rate": 8.760283946123044e-07, + "loss": 0.0713, + "step": 38947 + }, + { + "epoch": 7.08, + "learning_rate": 8.759737895886421e-07, + "loss": 0.9583, + "step": 38948 + }, + { + "epoch": 7.08, + "learning_rate": 8.7591918456498e-07, + "loss": 3.3772, + "step": 38949 + }, + { + "epoch": 7.08, + "learning_rate": 8.758645795413179e-07, + "loss": 0.2042, + "step": 38950 + }, + { + "epoch": 7.08, + "learning_rate": 8.758099745176556e-07, + "loss": 0.0461, + "step": 38951 + }, + { + "epoch": 7.08, + "learning_rate": 8.757553694939935e-07, + "loss": 2.8629, + "step": 38952 + }, + { + "epoch": 7.08, + "learning_rate": 8.757007644703314e-07, + "loss": 1.6485, + "step": 38953 + }, + { + "epoch": 7.08, + "learning_rate": 8.756461594466691e-07, + "loss": 0.0111, + "step": 38954 + }, + { + "epoch": 7.08, + "learning_rate": 8.75591554423007e-07, + "loss": 1.1568, + "step": 38955 + }, + { + "epoch": 7.08, + "learning_rate": 8.755369493993447e-07, + "loss": 1.7766, + "step": 38956 + }, + { + "epoch": 7.08, + "learning_rate": 8.754823443756827e-07, + "loss": 0.1301, + "step": 38957 + }, + { + "epoch": 7.08, + "learning_rate": 8.754277393520204e-07, + "loss": 0.8672, + "step": 38958 + }, + { + "epoch": 7.08, + "learning_rate": 8.753731343283582e-07, + "loss": 0.1266, + "step": 38959 + }, + { + "epoch": 7.08, + "learning_rate": 8.753185293046962e-07, + "loss": 0.7952, + "step": 38960 + }, + { + "epoch": 7.09, + "learning_rate": 8.752639242810339e-07, + "loss": 0.7041, + "step": 38961 + }, + { + "epoch": 7.09, + "learning_rate": 8.752093192573717e-07, + "loss": 1.7235, + "step": 38962 + }, + { + "epoch": 7.09, + "learning_rate": 8.751547142337095e-07, + "loss": 2.1843, + "step": 38963 + }, + { + "epoch": 7.09, + "learning_rate": 8.751001092100474e-07, + "loss": 0.278, + "step": 38964 + }, + { + "epoch": 7.09, + "learning_rate": 8.750455041863851e-07, + "loss": 1.0961, + "step": 38965 + }, + { + "epoch": 7.09, + "learning_rate": 8.74990899162723e-07, + "loss": 0.7261, + "step": 38966 + }, + { + "epoch": 7.09, + "learning_rate": 8.749362941390607e-07, + "loss": 0.0067, + "step": 38967 + }, + { + "epoch": 7.09, + "learning_rate": 8.748816891153987e-07, + "loss": 0.0872, + "step": 38968 + }, + { + "epoch": 7.09, + "learning_rate": 8.748270840917365e-07, + "loss": 0.7681, + "step": 38969 + }, + { + "epoch": 7.09, + "learning_rate": 8.747724790680742e-07, + "loss": 1.5859, + "step": 38970 + }, + { + "epoch": 7.09, + "learning_rate": 8.747178740444122e-07, + "loss": 0.2885, + "step": 38971 + }, + { + "epoch": 7.09, + "learning_rate": 8.746632690207499e-07, + "loss": 0.3126, + "step": 38972 + }, + { + "epoch": 7.09, + "learning_rate": 8.746086639970877e-07, + "loss": 0.5669, + "step": 38973 + }, + { + "epoch": 7.09, + "learning_rate": 8.745540589734255e-07, + "loss": 0.0141, + "step": 38974 + }, + { + "epoch": 7.09, + "learning_rate": 8.744994539497634e-07, + "loss": 1.0582, + "step": 38975 + }, + { + "epoch": 7.09, + "learning_rate": 8.744448489261012e-07, + "loss": 0.6103, + "step": 38976 + }, + { + "epoch": 7.09, + "learning_rate": 8.74390243902439e-07, + "loss": 0.0123, + "step": 38977 + }, + { + "epoch": 7.09, + "learning_rate": 8.743356388787769e-07, + "loss": 0.0564, + "step": 38978 + }, + { + "epoch": 7.09, + "learning_rate": 8.742810338551148e-07, + "loss": 0.2098, + "step": 38979 + }, + { + "epoch": 7.09, + "learning_rate": 8.742264288314525e-07, + "loss": 1.2529, + "step": 38980 + }, + { + "epoch": 7.09, + "learning_rate": 8.741718238077902e-07, + "loss": 3.5532, + "step": 38981 + }, + { + "epoch": 7.09, + "learning_rate": 8.741172187841282e-07, + "loss": 0.0074, + "step": 38982 + }, + { + "epoch": 7.09, + "learning_rate": 8.74062613760466e-07, + "loss": 0.0833, + "step": 38983 + }, + { + "epoch": 7.09, + "learning_rate": 8.740080087368037e-07, + "loss": 0.0702, + "step": 38984 + }, + { + "epoch": 7.09, + "learning_rate": 8.739534037131417e-07, + "loss": 0.4804, + "step": 38985 + }, + { + "epoch": 7.09, + "learning_rate": 8.738987986894795e-07, + "loss": 2.4676, + "step": 38986 + }, + { + "epoch": 7.09, + "learning_rate": 8.738441936658172e-07, + "loss": 0.315, + "step": 38987 + }, + { + "epoch": 7.09, + "learning_rate": 8.73789588642155e-07, + "loss": 0.0962, + "step": 38988 + }, + { + "epoch": 7.09, + "learning_rate": 8.737349836184929e-07, + "loss": 0.0922, + "step": 38989 + }, + { + "epoch": 7.09, + "learning_rate": 8.736803785948308e-07, + "loss": 0.0058, + "step": 38990 + }, + { + "epoch": 7.09, + "learning_rate": 8.736257735711685e-07, + "loss": 3.4557, + "step": 38991 + }, + { + "epoch": 7.09, + "learning_rate": 8.735711685475064e-07, + "loss": 1.6846, + "step": 38992 + }, + { + "epoch": 7.09, + "learning_rate": 8.735165635238443e-07, + "loss": 0.1025, + "step": 38993 + }, + { + "epoch": 7.09, + "learning_rate": 8.73461958500182e-07, + "loss": 0.1793, + "step": 38994 + }, + { + "epoch": 7.09, + "learning_rate": 8.734073534765197e-07, + "loss": 1.3274, + "step": 38995 + }, + { + "epoch": 7.09, + "learning_rate": 8.733527484528577e-07, + "loss": 0.0084, + "step": 38996 + }, + { + "epoch": 7.09, + "learning_rate": 8.732981434291955e-07, + "loss": 0.2231, + "step": 38997 + }, + { + "epoch": 7.09, + "learning_rate": 8.732435384055332e-07, + "loss": 1.3189, + "step": 38998 + }, + { + "epoch": 7.09, + "learning_rate": 8.731889333818712e-07, + "loss": 1.1952, + "step": 38999 + }, + { + "epoch": 7.09, + "learning_rate": 8.73134328358209e-07, + "loss": 0.0117, + "step": 39000 + }, + { + "epoch": 7.09, + "learning_rate": 8.730797233345468e-07, + "loss": 0.2884, + "step": 39001 + }, + { + "epoch": 7.09, + "learning_rate": 8.730251183108846e-07, + "loss": 0.1387, + "step": 39002 + }, + { + "epoch": 7.09, + "learning_rate": 8.729705132872225e-07, + "loss": 0.0145, + "step": 39003 + }, + { + "epoch": 7.09, + "learning_rate": 8.729159082635603e-07, + "loss": 0.1652, + "step": 39004 + }, + { + "epoch": 7.09, + "learning_rate": 8.72861303239898e-07, + "loss": 0.0766, + "step": 39005 + }, + { + "epoch": 7.09, + "learning_rate": 8.72806698216236e-07, + "loss": 0.3092, + "step": 39006 + }, + { + "epoch": 7.09, + "learning_rate": 8.727520931925738e-07, + "loss": 0.5016, + "step": 39007 + }, + { + "epoch": 7.09, + "learning_rate": 8.726974881689115e-07, + "loss": 0.7262, + "step": 39008 + }, + { + "epoch": 7.09, + "learning_rate": 8.726428831452494e-07, + "loss": 0.7225, + "step": 39009 + }, + { + "epoch": 7.09, + "learning_rate": 8.725882781215873e-07, + "loss": 0.3453, + "step": 39010 + }, + { + "epoch": 7.09, + "learning_rate": 8.72533673097925e-07, + "loss": 0.0476, + "step": 39011 + }, + { + "epoch": 7.09, + "learning_rate": 8.724790680742628e-07, + "loss": 2.3896, + "step": 39012 + }, + { + "epoch": 7.09, + "learning_rate": 8.724244630506007e-07, + "loss": 0.3351, + "step": 39013 + }, + { + "epoch": 7.09, + "learning_rate": 8.723698580269385e-07, + "loss": 0.0102, + "step": 39014 + }, + { + "epoch": 7.09, + "learning_rate": 8.723152530032763e-07, + "loss": 0.167, + "step": 39015 + }, + { + "epoch": 7.1, + "learning_rate": 8.722606479796141e-07, + "loss": 3.3217, + "step": 39016 + }, + { + "epoch": 7.1, + "learning_rate": 8.722060429559521e-07, + "loss": 0.0721, + "step": 39017 + }, + { + "epoch": 7.1, + "learning_rate": 8.721514379322898e-07, + "loss": 0.086, + "step": 39018 + }, + { + "epoch": 7.1, + "learning_rate": 8.720968329086276e-07, + "loss": 1.8743, + "step": 39019 + }, + { + "epoch": 7.1, + "learning_rate": 8.720422278849655e-07, + "loss": 0.1064, + "step": 39020 + }, + { + "epoch": 7.1, + "learning_rate": 8.719876228613033e-07, + "loss": 0.1298, + "step": 39021 + }, + { + "epoch": 7.1, + "learning_rate": 8.71933017837641e-07, + "loss": 0.9632, + "step": 39022 + }, + { + "epoch": 7.1, + "learning_rate": 8.718784128139789e-07, + "loss": 0.1312, + "step": 39023 + }, + { + "epoch": 7.1, + "learning_rate": 8.718238077903168e-07, + "loss": 0.2233, + "step": 39024 + }, + { + "epoch": 7.1, + "learning_rate": 8.717692027666545e-07, + "loss": 0.0133, + "step": 39025 + }, + { + "epoch": 7.1, + "learning_rate": 8.717145977429924e-07, + "loss": 0.2843, + "step": 39026 + }, + { + "epoch": 7.1, + "learning_rate": 8.716599927193303e-07, + "loss": 0.9798, + "step": 39027 + }, + { + "epoch": 7.1, + "learning_rate": 8.716053876956681e-07, + "loss": 0.1686, + "step": 39028 + }, + { + "epoch": 7.1, + "learning_rate": 8.715507826720058e-07, + "loss": 0.1378, + "step": 39029 + }, + { + "epoch": 7.1, + "learning_rate": 8.714961776483436e-07, + "loss": 0.4777, + "step": 39030 + }, + { + "epoch": 7.1, + "learning_rate": 8.714415726246816e-07, + "loss": 0.0455, + "step": 39031 + }, + { + "epoch": 7.1, + "learning_rate": 8.713869676010193e-07, + "loss": 0.7766, + "step": 39032 + }, + { + "epoch": 7.1, + "learning_rate": 8.713323625773571e-07, + "loss": 1.2846, + "step": 39033 + }, + { + "epoch": 7.1, + "learning_rate": 8.712777575536951e-07, + "loss": 0.0241, + "step": 39034 + }, + { + "epoch": 7.1, + "learning_rate": 8.712231525300328e-07, + "loss": 0.3436, + "step": 39035 + }, + { + "epoch": 7.1, + "learning_rate": 8.711685475063705e-07, + "loss": 1.1137, + "step": 39036 + }, + { + "epoch": 7.1, + "learning_rate": 8.711139424827084e-07, + "loss": 0.012, + "step": 39037 + }, + { + "epoch": 7.1, + "learning_rate": 8.710593374590463e-07, + "loss": 0.0794, + "step": 39038 + }, + { + "epoch": 7.1, + "learning_rate": 8.710047324353841e-07, + "loss": 0.2501, + "step": 39039 + }, + { + "epoch": 7.1, + "learning_rate": 8.709501274117219e-07, + "loss": 1.5443, + "step": 39040 + }, + { + "epoch": 7.1, + "learning_rate": 8.708955223880598e-07, + "loss": 0.0846, + "step": 39041 + }, + { + "epoch": 7.1, + "learning_rate": 8.708409173643976e-07, + "loss": 0.0072, + "step": 39042 + }, + { + "epoch": 7.1, + "learning_rate": 8.707863123407354e-07, + "loss": 0.0923, + "step": 39043 + }, + { + "epoch": 7.1, + "learning_rate": 8.707317073170731e-07, + "loss": 0.9419, + "step": 39044 + }, + { + "epoch": 7.1, + "learning_rate": 8.706771022934111e-07, + "loss": 1.9492, + "step": 39045 + }, + { + "epoch": 7.1, + "learning_rate": 8.706224972697488e-07, + "loss": 0.1481, + "step": 39046 + }, + { + "epoch": 7.1, + "learning_rate": 8.705678922460867e-07, + "loss": 0.0346, + "step": 39047 + }, + { + "epoch": 7.1, + "learning_rate": 8.705132872224244e-07, + "loss": 0.2623, + "step": 39048 + }, + { + "epoch": 7.1, + "learning_rate": 8.704586821987623e-07, + "loss": 0.0062, + "step": 39049 + }, + { + "epoch": 7.1, + "learning_rate": 8.704040771751002e-07, + "loss": 1.6304, + "step": 39050 + }, + { + "epoch": 7.1, + "learning_rate": 8.703494721514379e-07, + "loss": 0.0219, + "step": 39051 + }, + { + "epoch": 7.1, + "learning_rate": 8.702948671277758e-07, + "loss": 0.8073, + "step": 39052 + }, + { + "epoch": 7.1, + "learning_rate": 8.702402621041136e-07, + "loss": 0.735, + "step": 39053 + }, + { + "epoch": 7.1, + "learning_rate": 8.701856570804514e-07, + "loss": 0.0277, + "step": 39054 + }, + { + "epoch": 7.1, + "learning_rate": 8.701310520567891e-07, + "loss": 0.0148, + "step": 39055 + }, + { + "epoch": 7.1, + "learning_rate": 8.700764470331271e-07, + "loss": 0.1666, + "step": 39056 + }, + { + "epoch": 7.1, + "learning_rate": 8.700218420094649e-07, + "loss": 2.4273, + "step": 39057 + }, + { + "epoch": 7.1, + "learning_rate": 8.699672369858027e-07, + "loss": 0.0368, + "step": 39058 + }, + { + "epoch": 7.1, + "learning_rate": 8.699126319621406e-07, + "loss": 0.0141, + "step": 39059 + }, + { + "epoch": 7.1, + "learning_rate": 8.698580269384783e-07, + "loss": 1.369, + "step": 39060 + }, + { + "epoch": 7.1, + "learning_rate": 8.698034219148162e-07, + "loss": 0.6132, + "step": 39061 + }, + { + "epoch": 7.1, + "learning_rate": 8.697488168911539e-07, + "loss": 0.0206, + "step": 39062 + }, + { + "epoch": 7.1, + "learning_rate": 8.696942118674918e-07, + "loss": 0.2173, + "step": 39063 + }, + { + "epoch": 7.1, + "learning_rate": 8.696396068438297e-07, + "loss": 2.279, + "step": 39064 + }, + { + "epoch": 7.1, + "learning_rate": 8.695850018201674e-07, + "loss": 0.4811, + "step": 39065 + }, + { + "epoch": 7.1, + "learning_rate": 8.695303967965054e-07, + "loss": 0.6307, + "step": 39066 + }, + { + "epoch": 7.1, + "learning_rate": 8.694757917728432e-07, + "loss": 0.0957, + "step": 39067 + }, + { + "epoch": 7.1, + "learning_rate": 8.694211867491809e-07, + "loss": 0.2449, + "step": 39068 + }, + { + "epoch": 7.1, + "learning_rate": 8.693665817255187e-07, + "loss": 0.3356, + "step": 39069 + }, + { + "epoch": 7.1, + "learning_rate": 8.693119767018566e-07, + "loss": 0.0862, + "step": 39070 + }, + { + "epoch": 7.11, + "learning_rate": 8.692573716781944e-07, + "loss": 0.3421, + "step": 39071 + }, + { + "epoch": 7.11, + "learning_rate": 8.692027666545322e-07, + "loss": 0.4841, + "step": 39072 + }, + { + "epoch": 7.11, + "learning_rate": 8.691481616308701e-07, + "loss": 0.0469, + "step": 39073 + }, + { + "epoch": 7.11, + "learning_rate": 8.690935566072079e-07, + "loss": 0.3523, + "step": 39074 + }, + { + "epoch": 7.11, + "learning_rate": 8.690389515835457e-07, + "loss": 0.1013, + "step": 39075 + }, + { + "epoch": 7.11, + "learning_rate": 8.689843465598834e-07, + "loss": 0.7436, + "step": 39076 + }, + { + "epoch": 7.11, + "learning_rate": 8.689297415362214e-07, + "loss": 0.4973, + "step": 39077 + }, + { + "epoch": 7.11, + "learning_rate": 8.688751365125592e-07, + "loss": 2.8053, + "step": 39078 + }, + { + "epoch": 7.11, + "learning_rate": 8.688205314888969e-07, + "loss": 0.6917, + "step": 39079 + }, + { + "epoch": 7.11, + "learning_rate": 8.687659264652349e-07, + "loss": 1.3212, + "step": 39080 + }, + { + "epoch": 7.11, + "learning_rate": 8.687113214415727e-07, + "loss": 0.212, + "step": 39081 + }, + { + "epoch": 7.11, + "learning_rate": 8.686567164179104e-07, + "loss": 0.006, + "step": 39082 + }, + { + "epoch": 7.11, + "learning_rate": 8.686021113942483e-07, + "loss": 0.0084, + "step": 39083 + }, + { + "epoch": 7.11, + "learning_rate": 8.685475063705862e-07, + "loss": 1.8749, + "step": 39084 + }, + { + "epoch": 7.11, + "learning_rate": 8.68492901346924e-07, + "loss": 2.1166, + "step": 39085 + }, + { + "epoch": 7.11, + "learning_rate": 8.684382963232617e-07, + "loss": 0.2593, + "step": 39086 + }, + { + "epoch": 7.11, + "learning_rate": 8.683836912995996e-07, + "loss": 0.6349, + "step": 39087 + }, + { + "epoch": 7.11, + "learning_rate": 8.683290862759375e-07, + "loss": 0.7295, + "step": 39088 + }, + { + "epoch": 7.11, + "learning_rate": 8.682744812522752e-07, + "loss": 0.1856, + "step": 39089 + }, + { + "epoch": 7.11, + "learning_rate": 8.68219876228613e-07, + "loss": 1.616, + "step": 39090 + }, + { + "epoch": 7.11, + "learning_rate": 8.68165271204951e-07, + "loss": 0.6349, + "step": 39091 + }, + { + "epoch": 7.11, + "learning_rate": 8.681106661812887e-07, + "loss": 3.8428, + "step": 39092 + }, + { + "epoch": 7.11, + "learning_rate": 8.680560611576264e-07, + "loss": 0.034, + "step": 39093 + }, + { + "epoch": 7.11, + "learning_rate": 8.680014561339644e-07, + "loss": 1.6275, + "step": 39094 + }, + { + "epoch": 7.11, + "learning_rate": 8.679468511103022e-07, + "loss": 1.711, + "step": 39095 + }, + { + "epoch": 7.11, + "learning_rate": 8.6789224608664e-07, + "loss": 0.7021, + "step": 39096 + }, + { + "epoch": 7.11, + "learning_rate": 8.678376410629778e-07, + "loss": 0.0043, + "step": 39097 + }, + { + "epoch": 7.11, + "learning_rate": 8.677830360393157e-07, + "loss": 0.1198, + "step": 39098 + }, + { + "epoch": 7.11, + "learning_rate": 8.677284310156535e-07, + "loss": 0.9202, + "step": 39099 + }, + { + "epoch": 7.11, + "learning_rate": 8.676738259919912e-07, + "loss": 0.1065, + "step": 39100 + }, + { + "epoch": 7.11, + "learning_rate": 8.676192209683291e-07, + "loss": 0.0673, + "step": 39101 + }, + { + "epoch": 7.11, + "learning_rate": 8.67564615944667e-07, + "loss": 0.0069, + "step": 39102 + }, + { + "epoch": 7.11, + "learning_rate": 8.675100109210047e-07, + "loss": 0.0514, + "step": 39103 + }, + { + "epoch": 7.11, + "learning_rate": 8.674554058973425e-07, + "loss": 0.3026, + "step": 39104 + }, + { + "epoch": 7.11, + "learning_rate": 8.674008008736805e-07, + "loss": 0.191, + "step": 39105 + }, + { + "epoch": 7.11, + "learning_rate": 8.673461958500182e-07, + "loss": 0.4768, + "step": 39106 + }, + { + "epoch": 7.11, + "learning_rate": 8.672915908263561e-07, + "loss": 0.2769, + "step": 39107 + }, + { + "epoch": 7.11, + "learning_rate": 8.67236985802694e-07, + "loss": 0.1717, + "step": 39108 + }, + { + "epoch": 7.11, + "learning_rate": 8.671823807790317e-07, + "loss": 0.0061, + "step": 39109 + }, + { + "epoch": 7.11, + "learning_rate": 8.671277757553695e-07, + "loss": 0.014, + "step": 39110 + }, + { + "epoch": 7.11, + "learning_rate": 8.670731707317073e-07, + "loss": 0.297, + "step": 39111 + }, + { + "epoch": 7.11, + "learning_rate": 8.670185657080452e-07, + "loss": 1.1849, + "step": 39112 + }, + { + "epoch": 7.11, + "learning_rate": 8.66963960684383e-07, + "loss": 0.0887, + "step": 39113 + }, + { + "epoch": 7.11, + "learning_rate": 8.669093556607208e-07, + "loss": 0.6325, + "step": 39114 + }, + { + "epoch": 7.11, + "learning_rate": 8.668547506370588e-07, + "loss": 0.2282, + "step": 39115 + }, + { + "epoch": 7.11, + "learning_rate": 8.668001456133965e-07, + "loss": 0.2369, + "step": 39116 + }, + { + "epoch": 7.11, + "learning_rate": 8.667455405897342e-07, + "loss": 0.1417, + "step": 39117 + }, + { + "epoch": 7.11, + "learning_rate": 8.666909355660721e-07, + "loss": 0.3568, + "step": 39118 + }, + { + "epoch": 7.11, + "learning_rate": 8.6663633054241e-07, + "loss": 0.037, + "step": 39119 + }, + { + "epoch": 7.11, + "learning_rate": 8.665817255187477e-07, + "loss": 0.118, + "step": 39120 + }, + { + "epoch": 7.11, + "learning_rate": 8.665271204950856e-07, + "loss": 0.0049, + "step": 39121 + }, + { + "epoch": 7.11, + "learning_rate": 8.664725154714233e-07, + "loss": 0.085, + "step": 39122 + }, + { + "epoch": 7.11, + "learning_rate": 8.664179104477612e-07, + "loss": 0.2421, + "step": 39123 + }, + { + "epoch": 7.11, + "learning_rate": 8.66363305424099e-07, + "loss": 2.0878, + "step": 39124 + }, + { + "epoch": 7.11, + "learning_rate": 8.663087004004368e-07, + "loss": 1.8959, + "step": 39125 + }, + { + "epoch": 7.12, + "learning_rate": 8.662540953767748e-07, + "loss": 1.4015, + "step": 39126 + }, + { + "epoch": 7.12, + "learning_rate": 8.661994903531125e-07, + "loss": 0.0366, + "step": 39127 + }, + { + "epoch": 7.12, + "learning_rate": 8.661448853294503e-07, + "loss": 0.0186, + "step": 39128 + }, + { + "epoch": 7.12, + "learning_rate": 8.660902803057881e-07, + "loss": 0.2232, + "step": 39129 + }, + { + "epoch": 7.12, + "learning_rate": 8.66035675282126e-07, + "loss": 1.542, + "step": 39130 + }, + { + "epoch": 7.12, + "learning_rate": 8.659810702584638e-07, + "loss": 0.0089, + "step": 39131 + }, + { + "epoch": 7.12, + "learning_rate": 8.659264652348016e-07, + "loss": 0.0037, + "step": 39132 + }, + { + "epoch": 7.12, + "learning_rate": 8.658718602111395e-07, + "loss": 0.2777, + "step": 39133 + }, + { + "epoch": 7.12, + "learning_rate": 8.658172551874772e-07, + "loss": 0.6239, + "step": 39134 + }, + { + "epoch": 7.12, + "learning_rate": 8.657626501638151e-07, + "loss": 0.0847, + "step": 39135 + }, + { + "epoch": 7.12, + "learning_rate": 8.657080451401528e-07, + "loss": 0.0363, + "step": 39136 + }, + { + "epoch": 7.12, + "learning_rate": 8.656534401164908e-07, + "loss": 1.3739, + "step": 39137 + }, + { + "epoch": 7.12, + "learning_rate": 8.655988350928286e-07, + "loss": 0.6754, + "step": 39138 + }, + { + "epoch": 7.12, + "learning_rate": 8.655442300691663e-07, + "loss": 0.5746, + "step": 39139 + }, + { + "epoch": 7.12, + "learning_rate": 8.654896250455043e-07, + "loss": 1.3876, + "step": 39140 + }, + { + "epoch": 7.12, + "learning_rate": 8.65435020021842e-07, + "loss": 0.0051, + "step": 39141 + }, + { + "epoch": 7.12, + "learning_rate": 8.653804149981798e-07, + "loss": 0.6782, + "step": 39142 + }, + { + "epoch": 7.12, + "learning_rate": 8.653258099745176e-07, + "loss": 0.4381, + "step": 39143 + }, + { + "epoch": 7.12, + "learning_rate": 8.652712049508555e-07, + "loss": 0.0257, + "step": 39144 + }, + { + "epoch": 7.12, + "learning_rate": 8.652165999271934e-07, + "loss": 0.015, + "step": 39145 + }, + { + "epoch": 7.12, + "learning_rate": 8.651619949035311e-07, + "loss": 0.0065, + "step": 39146 + }, + { + "epoch": 7.12, + "learning_rate": 8.65107389879869e-07, + "loss": 0.7286, + "step": 39147 + }, + { + "epoch": 7.12, + "learning_rate": 8.650527848562069e-07, + "loss": 0.014, + "step": 39148 + }, + { + "epoch": 7.12, + "learning_rate": 8.649981798325446e-07, + "loss": 0.1135, + "step": 39149 + }, + { + "epoch": 7.12, + "learning_rate": 8.649435748088823e-07, + "loss": 0.0152, + "step": 39150 + }, + { + "epoch": 7.12, + "learning_rate": 8.648889697852203e-07, + "loss": 0.0123, + "step": 39151 + }, + { + "epoch": 7.12, + "learning_rate": 8.648343647615581e-07, + "loss": 0.1691, + "step": 39152 + }, + { + "epoch": 7.12, + "learning_rate": 8.647797597378958e-07, + "loss": 0.2232, + "step": 39153 + }, + { + "epoch": 7.12, + "learning_rate": 8.647251547142338e-07, + "loss": 0.4985, + "step": 39154 + }, + { + "epoch": 7.12, + "learning_rate": 8.646705496905716e-07, + "loss": 0.271, + "step": 39155 + }, + { + "epoch": 7.12, + "learning_rate": 8.646159446669094e-07, + "loss": 0.7458, + "step": 39156 + }, + { + "epoch": 7.12, + "learning_rate": 8.645613396432471e-07, + "loss": 0.25, + "step": 39157 + }, + { + "epoch": 7.12, + "learning_rate": 8.64506734619585e-07, + "loss": 0.1657, + "step": 39158 + }, + { + "epoch": 7.12, + "learning_rate": 8.644521295959229e-07, + "loss": 0.0367, + "step": 39159 + }, + { + "epoch": 7.12, + "learning_rate": 8.643975245722606e-07, + "loss": 0.0067, + "step": 39160 + }, + { + "epoch": 7.12, + "learning_rate": 8.643429195485985e-07, + "loss": 0.0181, + "step": 39161 + }, + { + "epoch": 7.12, + "learning_rate": 8.642883145249364e-07, + "loss": 0.1926, + "step": 39162 + }, + { + "epoch": 7.12, + "learning_rate": 8.642337095012741e-07, + "loss": 0.925, + "step": 39163 + }, + { + "epoch": 7.12, + "learning_rate": 8.641791044776118e-07, + "loss": 0.2379, + "step": 39164 + }, + { + "epoch": 7.12, + "learning_rate": 8.641244994539498e-07, + "loss": 0.2696, + "step": 39165 + }, + { + "epoch": 7.12, + "learning_rate": 8.640698944302876e-07, + "loss": 0.8726, + "step": 39166 + }, + { + "epoch": 7.12, + "learning_rate": 8.640152894066254e-07, + "loss": 0.2135, + "step": 39167 + }, + { + "epoch": 7.12, + "learning_rate": 8.639606843829633e-07, + "loss": 0.2189, + "step": 39168 + }, + { + "epoch": 7.12, + "learning_rate": 8.639060793593011e-07, + "loss": 0.7257, + "step": 39169 + }, + { + "epoch": 7.12, + "learning_rate": 8.638514743356389e-07, + "loss": 2.7348, + "step": 39170 + }, + { + "epoch": 7.12, + "learning_rate": 8.637968693119767e-07, + "loss": 0.0293, + "step": 39171 + }, + { + "epoch": 7.12, + "learning_rate": 8.637422642883146e-07, + "loss": 1.8987, + "step": 39172 + }, + { + "epoch": 7.12, + "learning_rate": 8.636876592646524e-07, + "loss": 0.0412, + "step": 39173 + }, + { + "epoch": 7.12, + "learning_rate": 8.636330542409901e-07, + "loss": 0.0068, + "step": 39174 + }, + { + "epoch": 7.12, + "learning_rate": 8.635784492173281e-07, + "loss": 2.3303, + "step": 39175 + }, + { + "epoch": 7.12, + "learning_rate": 8.635238441936659e-07, + "loss": 0.0045, + "step": 39176 + }, + { + "epoch": 7.12, + "learning_rate": 8.634692391700036e-07, + "loss": 0.3912, + "step": 39177 + }, + { + "epoch": 7.12, + "learning_rate": 8.634146341463415e-07, + "loss": 0.1858, + "step": 39178 + }, + { + "epoch": 7.12, + "learning_rate": 8.633600291226794e-07, + "loss": 0.005, + "step": 39179 + }, + { + "epoch": 7.12, + "learning_rate": 8.633054240990171e-07, + "loss": 0.4195, + "step": 39180 + }, + { + "epoch": 7.13, + "learning_rate": 8.632508190753549e-07, + "loss": 0.5205, + "step": 39181 + }, + { + "epoch": 7.13, + "learning_rate": 8.631962140516928e-07, + "loss": 0.3307, + "step": 39182 + }, + { + "epoch": 7.13, + "learning_rate": 8.631416090280306e-07, + "loss": 0.0764, + "step": 39183 + }, + { + "epoch": 7.13, + "learning_rate": 8.630870040043684e-07, + "loss": 1.6315, + "step": 39184 + }, + { + "epoch": 7.13, + "learning_rate": 8.630323989807062e-07, + "loss": 0.7289, + "step": 39185 + }, + { + "epoch": 7.13, + "learning_rate": 8.629777939570442e-07, + "loss": 2.099, + "step": 39186 + }, + { + "epoch": 7.13, + "learning_rate": 8.629231889333819e-07, + "loss": 0.7997, + "step": 39187 + }, + { + "epoch": 7.13, + "learning_rate": 8.628685839097197e-07, + "loss": 1.4533, + "step": 39188 + }, + { + "epoch": 7.13, + "learning_rate": 8.628139788860576e-07, + "loss": 0.0528, + "step": 39189 + }, + { + "epoch": 7.13, + "learning_rate": 8.627593738623954e-07, + "loss": 0.2671, + "step": 39190 + }, + { + "epoch": 7.13, + "learning_rate": 8.627047688387331e-07, + "loss": 1.4984, + "step": 39191 + }, + { + "epoch": 7.13, + "learning_rate": 8.62650163815071e-07, + "loss": 0.0098, + "step": 39192 + }, + { + "epoch": 7.13, + "learning_rate": 8.625955587914089e-07, + "loss": 0.6926, + "step": 39193 + }, + { + "epoch": 7.13, + "learning_rate": 8.625409537677466e-07, + "loss": 0.439, + "step": 39194 + }, + { + "epoch": 7.13, + "learning_rate": 8.624863487440845e-07, + "loss": 1.7082, + "step": 39195 + }, + { + "epoch": 7.13, + "learning_rate": 8.624317437204224e-07, + "loss": 0.0696, + "step": 39196 + }, + { + "epoch": 7.13, + "learning_rate": 8.623771386967602e-07, + "loss": 0.0071, + "step": 39197 + }, + { + "epoch": 7.13, + "learning_rate": 8.623225336730979e-07, + "loss": 0.0772, + "step": 39198 + }, + { + "epoch": 7.13, + "learning_rate": 8.622679286494357e-07, + "loss": 0.0741, + "step": 39199 + }, + { + "epoch": 7.13, + "learning_rate": 8.622133236257737e-07, + "loss": 0.0111, + "step": 39200 + }, + { + "epoch": 7.13, + "learning_rate": 8.621587186021114e-07, + "loss": 0.0063, + "step": 39201 + }, + { + "epoch": 7.13, + "learning_rate": 8.621041135784492e-07, + "loss": 0.0537, + "step": 39202 + }, + { + "epoch": 7.13, + "learning_rate": 8.62049508554787e-07, + "loss": 0.0115, + "step": 39203 + }, + { + "epoch": 7.13, + "learning_rate": 8.619949035311249e-07, + "loss": 0.098, + "step": 39204 + }, + { + "epoch": 7.13, + "learning_rate": 8.619402985074626e-07, + "loss": 0.0134, + "step": 39205 + }, + { + "epoch": 7.13, + "learning_rate": 8.618856934838005e-07, + "loss": 0.0131, + "step": 39206 + }, + { + "epoch": 7.13, + "learning_rate": 8.618310884601384e-07, + "loss": 0.0721, + "step": 39207 + }, + { + "epoch": 7.13, + "learning_rate": 8.617764834364762e-07, + "loss": 0.8307, + "step": 39208 + }, + { + "epoch": 7.13, + "learning_rate": 8.61721878412814e-07, + "loss": 1.5318, + "step": 39209 + }, + { + "epoch": 7.13, + "learning_rate": 8.616672733891517e-07, + "loss": 0.0078, + "step": 39210 + }, + { + "epoch": 7.13, + "learning_rate": 8.616126683654897e-07, + "loss": 0.0282, + "step": 39211 + }, + { + "epoch": 7.13, + "learning_rate": 8.615580633418275e-07, + "loss": 0.0529, + "step": 39212 + }, + { + "epoch": 7.13, + "learning_rate": 8.615034583181652e-07, + "loss": 0.0114, + "step": 39213 + }, + { + "epoch": 7.13, + "learning_rate": 8.614488532945032e-07, + "loss": 0.1956, + "step": 39214 + }, + { + "epoch": 7.13, + "learning_rate": 8.613942482708409e-07, + "loss": 0.0131, + "step": 39215 + }, + { + "epoch": 7.13, + "learning_rate": 8.613396432471788e-07, + "loss": 0.0087, + "step": 39216 + }, + { + "epoch": 7.13, + "learning_rate": 8.612850382235165e-07, + "loss": 1.9216, + "step": 39217 + }, + { + "epoch": 7.13, + "learning_rate": 8.612304331998544e-07, + "loss": 0.0449, + "step": 39218 + }, + { + "epoch": 7.13, + "learning_rate": 8.611758281761923e-07, + "loss": 0.0193, + "step": 39219 + }, + { + "epoch": 7.13, + "learning_rate": 8.6112122315253e-07, + "loss": 0.0094, + "step": 39220 + }, + { + "epoch": 7.13, + "learning_rate": 8.610666181288679e-07, + "loss": 0.6222, + "step": 39221 + }, + { + "epoch": 7.13, + "learning_rate": 8.610120131052057e-07, + "loss": 1.1871, + "step": 39222 + }, + { + "epoch": 7.13, + "learning_rate": 8.609574080815435e-07, + "loss": 0.4365, + "step": 39223 + }, + { + "epoch": 7.13, + "learning_rate": 8.609028030578812e-07, + "loss": 0.7711, + "step": 39224 + }, + { + "epoch": 7.13, + "learning_rate": 8.608481980342192e-07, + "loss": 0.9723, + "step": 39225 + }, + { + "epoch": 7.13, + "learning_rate": 8.60793593010557e-07, + "loss": 0.6448, + "step": 39226 + }, + { + "epoch": 7.13, + "learning_rate": 8.607389879868948e-07, + "loss": 3.4889, + "step": 39227 + }, + { + "epoch": 7.13, + "learning_rate": 8.606843829632327e-07, + "loss": 1.0114, + "step": 39228 + }, + { + "epoch": 7.13, + "learning_rate": 8.606297779395704e-07, + "loss": 0.3622, + "step": 39229 + }, + { + "epoch": 7.13, + "learning_rate": 8.605751729159083e-07, + "loss": 0.9756, + "step": 39230 + }, + { + "epoch": 7.13, + "learning_rate": 8.60520567892246e-07, + "loss": 0.0853, + "step": 39231 + }, + { + "epoch": 7.13, + "learning_rate": 8.604659628685839e-07, + "loss": 0.5753, + "step": 39232 + }, + { + "epoch": 7.13, + "learning_rate": 8.604113578449218e-07, + "loss": 0.3281, + "step": 39233 + }, + { + "epoch": 7.13, + "learning_rate": 8.603567528212595e-07, + "loss": 0.3385, + "step": 39234 + }, + { + "epoch": 7.13, + "learning_rate": 8.603021477975975e-07, + "loss": 0.1065, + "step": 39235 + }, + { + "epoch": 7.14, + "learning_rate": 8.602475427739353e-07, + "loss": 0.0864, + "step": 39236 + }, + { + "epoch": 7.14, + "learning_rate": 8.60192937750273e-07, + "loss": 0.7117, + "step": 39237 + }, + { + "epoch": 7.14, + "learning_rate": 8.601383327266108e-07, + "loss": 2.0257, + "step": 39238 + }, + { + "epoch": 7.14, + "learning_rate": 8.600837277029487e-07, + "loss": 0.5293, + "step": 39239 + }, + { + "epoch": 7.14, + "learning_rate": 8.600291226792865e-07, + "loss": 1.1727, + "step": 39240 + }, + { + "epoch": 7.14, + "learning_rate": 8.599745176556243e-07, + "loss": 0.0833, + "step": 39241 + }, + { + "epoch": 7.14, + "learning_rate": 8.599199126319622e-07, + "loss": 0.1535, + "step": 39242 + }, + { + "epoch": 7.14, + "learning_rate": 8.598653076083e-07, + "loss": 0.0674, + "step": 39243 + }, + { + "epoch": 7.14, + "learning_rate": 8.598107025846378e-07, + "loss": 0.0947, + "step": 39244 + }, + { + "epoch": 7.14, + "learning_rate": 8.597560975609755e-07, + "loss": 0.7823, + "step": 39245 + }, + { + "epoch": 7.14, + "learning_rate": 8.597014925373135e-07, + "loss": 0.1284, + "step": 39246 + }, + { + "epoch": 7.14, + "learning_rate": 8.596468875136513e-07, + "loss": 0.2948, + "step": 39247 + }, + { + "epoch": 7.14, + "learning_rate": 8.59592282489989e-07, + "loss": 0.1756, + "step": 39248 + }, + { + "epoch": 7.14, + "learning_rate": 8.59537677466327e-07, + "loss": 0.3532, + "step": 39249 + }, + { + "epoch": 7.14, + "learning_rate": 8.594830724426648e-07, + "loss": 0.4829, + "step": 39250 + }, + { + "epoch": 7.14, + "learning_rate": 8.594284674190025e-07, + "loss": 0.0357, + "step": 39251 + }, + { + "epoch": 7.14, + "learning_rate": 8.593738623953404e-07, + "loss": 0.4196, + "step": 39252 + }, + { + "epoch": 7.14, + "learning_rate": 8.593192573716783e-07, + "loss": 1.3921, + "step": 39253 + }, + { + "epoch": 7.14, + "learning_rate": 8.592646523480161e-07, + "loss": 0.1649, + "step": 39254 + }, + { + "epoch": 7.14, + "learning_rate": 8.592100473243538e-07, + "loss": 0.0552, + "step": 39255 + }, + { + "epoch": 7.14, + "learning_rate": 8.591554423006917e-07, + "loss": 0.0619, + "step": 39256 + }, + { + "epoch": 7.14, + "learning_rate": 8.591008372770296e-07, + "loss": 0.8353, + "step": 39257 + }, + { + "epoch": 7.14, + "learning_rate": 8.590462322533673e-07, + "loss": 0.0973, + "step": 39258 + }, + { + "epoch": 7.14, + "learning_rate": 8.589916272297051e-07, + "loss": 0.6064, + "step": 39259 + }, + { + "epoch": 7.14, + "learning_rate": 8.589370222060431e-07, + "loss": 0.0803, + "step": 39260 + }, + { + "epoch": 7.14, + "learning_rate": 8.588824171823808e-07, + "loss": 0.1604, + "step": 39261 + }, + { + "epoch": 7.14, + "learning_rate": 8.588278121587185e-07, + "loss": 0.6736, + "step": 39262 + }, + { + "epoch": 7.14, + "learning_rate": 8.587732071350565e-07, + "loss": 0.0644, + "step": 39263 + }, + { + "epoch": 7.14, + "learning_rate": 8.587186021113943e-07, + "loss": 0.4128, + "step": 39264 + }, + { + "epoch": 7.14, + "learning_rate": 8.586639970877321e-07, + "loss": 0.0693, + "step": 39265 + }, + { + "epoch": 7.14, + "learning_rate": 8.586093920640699e-07, + "loss": 0.159, + "step": 39266 + }, + { + "epoch": 7.14, + "learning_rate": 8.585547870404078e-07, + "loss": 0.3572, + "step": 39267 + }, + { + "epoch": 7.14, + "learning_rate": 8.585001820167456e-07, + "loss": 0.5125, + "step": 39268 + }, + { + "epoch": 7.14, + "learning_rate": 8.584455769930833e-07, + "loss": 0.9726, + "step": 39269 + }, + { + "epoch": 7.14, + "learning_rate": 8.583909719694212e-07, + "loss": 0.0067, + "step": 39270 + }, + { + "epoch": 7.14, + "learning_rate": 8.583363669457591e-07, + "loss": 0.0862, + "step": 39271 + }, + { + "epoch": 7.14, + "learning_rate": 8.582817619220968e-07, + "loss": 3.2225, + "step": 39272 + }, + { + "epoch": 7.14, + "learning_rate": 8.582271568984346e-07, + "loss": 0.6058, + "step": 39273 + }, + { + "epoch": 7.14, + "learning_rate": 8.581725518747726e-07, + "loss": 0.4685, + "step": 39274 + }, + { + "epoch": 7.14, + "learning_rate": 8.581179468511103e-07, + "loss": 0.1847, + "step": 39275 + }, + { + "epoch": 7.14, + "learning_rate": 8.580633418274482e-07, + "loss": 0.079, + "step": 39276 + }, + { + "epoch": 7.14, + "learning_rate": 8.580087368037859e-07, + "loss": 0.1629, + "step": 39277 + }, + { + "epoch": 7.14, + "learning_rate": 8.579541317801238e-07, + "loss": 0.3885, + "step": 39278 + }, + { + "epoch": 7.14, + "learning_rate": 8.578995267564616e-07, + "loss": 0.477, + "step": 39279 + }, + { + "epoch": 7.14, + "learning_rate": 8.578449217327994e-07, + "loss": 0.6611, + "step": 39280 + }, + { + "epoch": 7.14, + "learning_rate": 8.577903167091373e-07, + "loss": 0.2825, + "step": 39281 + }, + { + "epoch": 7.14, + "learning_rate": 8.577357116854751e-07, + "loss": 0.6653, + "step": 39282 + }, + { + "epoch": 7.14, + "learning_rate": 8.576811066618129e-07, + "loss": 1.4982, + "step": 39283 + }, + { + "epoch": 7.14, + "learning_rate": 8.576265016381506e-07, + "loss": 1.018, + "step": 39284 + }, + { + "epoch": 7.14, + "learning_rate": 8.575718966144886e-07, + "loss": 0.7208, + "step": 39285 + }, + { + "epoch": 7.14, + "learning_rate": 8.575172915908263e-07, + "loss": 0.1367, + "step": 39286 + }, + { + "epoch": 7.14, + "learning_rate": 8.574626865671642e-07, + "loss": 2.067, + "step": 39287 + }, + { + "epoch": 7.14, + "learning_rate": 8.574080815435021e-07, + "loss": 0.1657, + "step": 39288 + }, + { + "epoch": 7.14, + "learning_rate": 8.573534765198398e-07, + "loss": 0.0685, + "step": 39289 + }, + { + "epoch": 7.14, + "learning_rate": 8.572988714961777e-07, + "loss": 0.0092, + "step": 39290 + }, + { + "epoch": 7.15, + "learning_rate": 8.572442664725154e-07, + "loss": 0.2358, + "step": 39291 + }, + { + "epoch": 7.15, + "learning_rate": 8.571896614488533e-07, + "loss": 0.3522, + "step": 39292 + }, + { + "epoch": 7.15, + "learning_rate": 8.571350564251911e-07, + "loss": 0.7218, + "step": 39293 + }, + { + "epoch": 7.15, + "learning_rate": 8.570804514015289e-07, + "loss": 0.1825, + "step": 39294 + }, + { + "epoch": 7.15, + "learning_rate": 8.570258463778669e-07, + "loss": 0.0812, + "step": 39295 + }, + { + "epoch": 7.15, + "learning_rate": 8.569712413542046e-07, + "loss": 0.3532, + "step": 39296 + }, + { + "epoch": 7.15, + "learning_rate": 8.569166363305424e-07, + "loss": 0.0164, + "step": 39297 + }, + { + "epoch": 7.15, + "learning_rate": 8.568620313068802e-07, + "loss": 0.3247, + "step": 39298 + }, + { + "epoch": 7.15, + "learning_rate": 8.568074262832181e-07, + "loss": 0.4685, + "step": 39299 + }, + { + "epoch": 7.15, + "learning_rate": 8.567528212595559e-07, + "loss": 0.0659, + "step": 39300 + }, + { + "epoch": 7.15, + "learning_rate": 8.566982162358937e-07, + "loss": 1.6652, + "step": 39301 + }, + { + "epoch": 7.15, + "learning_rate": 8.566436112122316e-07, + "loss": 0.487, + "step": 39302 + }, + { + "epoch": 7.15, + "learning_rate": 8.565890061885694e-07, + "loss": 0.0182, + "step": 39303 + }, + { + "epoch": 7.15, + "learning_rate": 8.565344011649072e-07, + "loss": 1.2562, + "step": 39304 + }, + { + "epoch": 7.15, + "learning_rate": 8.564797961412449e-07, + "loss": 0.189, + "step": 39305 + }, + { + "epoch": 7.15, + "learning_rate": 8.564251911175829e-07, + "loss": 0.3644, + "step": 39306 + }, + { + "epoch": 7.15, + "learning_rate": 8.563705860939207e-07, + "loss": 0.4031, + "step": 39307 + }, + { + "epoch": 7.15, + "learning_rate": 8.563159810702584e-07, + "loss": 0.0046, + "step": 39308 + }, + { + "epoch": 7.15, + "learning_rate": 8.562613760465964e-07, + "loss": 0.4478, + "step": 39309 + }, + { + "epoch": 7.15, + "learning_rate": 8.562067710229341e-07, + "loss": 0.4554, + "step": 39310 + }, + { + "epoch": 7.15, + "learning_rate": 8.561521659992719e-07, + "loss": 0.1343, + "step": 39311 + }, + { + "epoch": 7.15, + "learning_rate": 8.560975609756097e-07, + "loss": 0.9578, + "step": 39312 + }, + { + "epoch": 7.15, + "learning_rate": 8.560429559519476e-07, + "loss": 0.9052, + "step": 39313 + }, + { + "epoch": 7.15, + "learning_rate": 8.559883509282855e-07, + "loss": 1.2072, + "step": 39314 + }, + { + "epoch": 7.15, + "learning_rate": 8.559337459046232e-07, + "loss": 2.7895, + "step": 39315 + }, + { + "epoch": 7.15, + "learning_rate": 8.558791408809611e-07, + "loss": 0.0222, + "step": 39316 + }, + { + "epoch": 7.15, + "learning_rate": 8.55824535857299e-07, + "loss": 1.1044, + "step": 39317 + }, + { + "epoch": 7.15, + "learning_rate": 8.557699308336367e-07, + "loss": 0.006, + "step": 39318 + }, + { + "epoch": 7.15, + "learning_rate": 8.557153258099744e-07, + "loss": 0.0329, + "step": 39319 + }, + { + "epoch": 7.15, + "learning_rate": 8.556607207863124e-07, + "loss": 0.0987, + "step": 39320 + }, + { + "epoch": 7.15, + "learning_rate": 8.556061157626502e-07, + "loss": 3.1981, + "step": 39321 + }, + { + "epoch": 7.15, + "learning_rate": 8.555515107389879e-07, + "loss": 1.3114, + "step": 39322 + }, + { + "epoch": 7.15, + "learning_rate": 8.554969057153259e-07, + "loss": 0.6928, + "step": 39323 + }, + { + "epoch": 7.15, + "learning_rate": 8.554423006916637e-07, + "loss": 0.6014, + "step": 39324 + }, + { + "epoch": 7.15, + "learning_rate": 8.553876956680015e-07, + "loss": 0.0543, + "step": 39325 + }, + { + "epoch": 7.15, + "learning_rate": 8.553330906443392e-07, + "loss": 0.0703, + "step": 39326 + }, + { + "epoch": 7.15, + "learning_rate": 8.552784856206771e-07, + "loss": 0.019, + "step": 39327 + }, + { + "epoch": 7.15, + "learning_rate": 8.55223880597015e-07, + "loss": 0.0562, + "step": 39328 + }, + { + "epoch": 7.15, + "learning_rate": 8.551692755733527e-07, + "loss": 0.0148, + "step": 39329 + }, + { + "epoch": 7.15, + "learning_rate": 8.551146705496906e-07, + "loss": 0.3678, + "step": 39330 + }, + { + "epoch": 7.15, + "learning_rate": 8.550600655260285e-07, + "loss": 0.428, + "step": 39331 + }, + { + "epoch": 7.15, + "learning_rate": 8.550054605023662e-07, + "loss": 1.8247, + "step": 39332 + }, + { + "epoch": 7.15, + "learning_rate": 8.549508554787039e-07, + "loss": 1.0068, + "step": 39333 + }, + { + "epoch": 7.15, + "learning_rate": 8.548962504550419e-07, + "loss": 0.1229, + "step": 39334 + }, + { + "epoch": 7.15, + "learning_rate": 8.548416454313797e-07, + "loss": 2.6903, + "step": 39335 + }, + { + "epoch": 7.15, + "learning_rate": 8.547870404077175e-07, + "loss": 0.0066, + "step": 39336 + }, + { + "epoch": 7.15, + "learning_rate": 8.547324353840554e-07, + "loss": 1.3727, + "step": 39337 + }, + { + "epoch": 7.15, + "learning_rate": 8.546778303603932e-07, + "loss": 0.1429, + "step": 39338 + }, + { + "epoch": 7.15, + "learning_rate": 8.54623225336731e-07, + "loss": 0.5503, + "step": 39339 + }, + { + "epoch": 7.15, + "learning_rate": 8.545686203130688e-07, + "loss": 0.0136, + "step": 39340 + }, + { + "epoch": 7.15, + "learning_rate": 8.545140152894068e-07, + "loss": 0.0899, + "step": 39341 + }, + { + "epoch": 7.15, + "learning_rate": 8.544594102657445e-07, + "loss": 0.0559, + "step": 39342 + }, + { + "epoch": 7.15, + "learning_rate": 8.544048052420822e-07, + "loss": 0.165, + "step": 39343 + }, + { + "epoch": 7.15, + "learning_rate": 8.543502002184202e-07, + "loss": 1.4046, + "step": 39344 + }, + { + "epoch": 7.15, + "learning_rate": 8.54295595194758e-07, + "loss": 0.1755, + "step": 39345 + }, + { + "epoch": 7.16, + "learning_rate": 8.542409901710957e-07, + "loss": 0.0604, + "step": 39346 + }, + { + "epoch": 7.16, + "learning_rate": 8.541863851474336e-07, + "loss": 0.8107, + "step": 39347 + }, + { + "epoch": 7.16, + "learning_rate": 8.541317801237715e-07, + "loss": 0.061, + "step": 39348 + }, + { + "epoch": 7.16, + "learning_rate": 8.540771751001092e-07, + "loss": 4.6072, + "step": 39349 + }, + { + "epoch": 7.16, + "learning_rate": 8.54022570076447e-07, + "loss": 0.1947, + "step": 39350 + }, + { + "epoch": 7.16, + "learning_rate": 8.539679650527849e-07, + "loss": 0.003, + "step": 39351 + }, + { + "epoch": 7.16, + "learning_rate": 8.539133600291228e-07, + "loss": 3.0806, + "step": 39352 + }, + { + "epoch": 7.16, + "learning_rate": 8.538587550054605e-07, + "loss": 0.9448, + "step": 39353 + }, + { + "epoch": 7.16, + "learning_rate": 8.538041499817983e-07, + "loss": 0.1178, + "step": 39354 + }, + { + "epoch": 7.16, + "learning_rate": 8.537495449581363e-07, + "loss": 0.3819, + "step": 39355 + }, + { + "epoch": 7.16, + "learning_rate": 8.53694939934474e-07, + "loss": 0.7397, + "step": 39356 + }, + { + "epoch": 7.16, + "learning_rate": 8.536403349108117e-07, + "loss": 0.2038, + "step": 39357 + }, + { + "epoch": 7.16, + "learning_rate": 8.535857298871496e-07, + "loss": 1.0448, + "step": 39358 + }, + { + "epoch": 7.16, + "learning_rate": 8.535311248634875e-07, + "loss": 0.4723, + "step": 39359 + }, + { + "epoch": 7.16, + "learning_rate": 8.534765198398252e-07, + "loss": 1.3125, + "step": 39360 + }, + { + "epoch": 7.16, + "learning_rate": 8.534219148161631e-07, + "loss": 0.4728, + "step": 39361 + }, + { + "epoch": 7.16, + "learning_rate": 8.53367309792501e-07, + "loss": 1.5283, + "step": 39362 + }, + { + "epoch": 7.16, + "learning_rate": 8.533127047688388e-07, + "loss": 0.2613, + "step": 39363 + }, + { + "epoch": 7.16, + "learning_rate": 8.532580997451766e-07, + "loss": 0.0969, + "step": 39364 + }, + { + "epoch": 7.16, + "learning_rate": 8.532034947215143e-07, + "loss": 0.0765, + "step": 39365 + }, + { + "epoch": 7.16, + "learning_rate": 8.531488896978523e-07, + "loss": 0.4969, + "step": 39366 + }, + { + "epoch": 7.16, + "learning_rate": 8.5309428467419e-07, + "loss": 0.0796, + "step": 39367 + }, + { + "epoch": 7.16, + "learning_rate": 8.530396796505278e-07, + "loss": 1.6931, + "step": 39368 + }, + { + "epoch": 7.16, + "learning_rate": 8.529850746268658e-07, + "loss": 3.1049, + "step": 39369 + }, + { + "epoch": 7.16, + "learning_rate": 8.529304696032035e-07, + "loss": 0.1449, + "step": 39370 + }, + { + "epoch": 7.16, + "learning_rate": 8.528758645795413e-07, + "loss": 0.0439, + "step": 39371 + }, + { + "epoch": 7.16, + "learning_rate": 8.528212595558791e-07, + "loss": 0.4193, + "step": 39372 + }, + { + "epoch": 7.16, + "learning_rate": 8.52766654532217e-07, + "loss": 0.0122, + "step": 39373 + }, + { + "epoch": 7.16, + "learning_rate": 8.527120495085548e-07, + "loss": 0.0088, + "step": 39374 + }, + { + "epoch": 7.16, + "learning_rate": 8.526574444848926e-07, + "loss": 1.5954, + "step": 39375 + }, + { + "epoch": 7.16, + "learning_rate": 8.526028394612305e-07, + "loss": 0.0869, + "step": 39376 + }, + { + "epoch": 7.16, + "learning_rate": 8.525482344375683e-07, + "loss": 4.4609, + "step": 39377 + }, + { + "epoch": 7.16, + "learning_rate": 8.524936294139061e-07, + "loss": 1.4716, + "step": 39378 + }, + { + "epoch": 7.16, + "learning_rate": 8.524390243902438e-07, + "loss": 1.4926, + "step": 39379 + }, + { + "epoch": 7.16, + "learning_rate": 8.523844193665818e-07, + "loss": 0.3435, + "step": 39380 + }, + { + "epoch": 7.16, + "learning_rate": 8.523298143429196e-07, + "loss": 0.0969, + "step": 39381 + }, + { + "epoch": 7.16, + "learning_rate": 8.522752093192573e-07, + "loss": 0.5716, + "step": 39382 + }, + { + "epoch": 7.16, + "learning_rate": 8.522206042955953e-07, + "loss": 0.0398, + "step": 39383 + }, + { + "epoch": 7.16, + "learning_rate": 8.52165999271933e-07, + "loss": 0.1439, + "step": 39384 + }, + { + "epoch": 7.16, + "learning_rate": 8.521113942482709e-07, + "loss": 2.2632, + "step": 39385 + }, + { + "epoch": 7.16, + "learning_rate": 8.520567892246086e-07, + "loss": 1.6811, + "step": 39386 + }, + { + "epoch": 7.16, + "learning_rate": 8.520021842009465e-07, + "loss": 0.1403, + "step": 39387 + }, + { + "epoch": 7.16, + "learning_rate": 8.519475791772844e-07, + "loss": 0.0815, + "step": 39388 + }, + { + "epoch": 7.16, + "learning_rate": 8.518929741536221e-07, + "loss": 0.1009, + "step": 39389 + }, + { + "epoch": 7.16, + "learning_rate": 8.5183836912996e-07, + "loss": 1.2421, + "step": 39390 + }, + { + "epoch": 7.16, + "learning_rate": 8.517837641062978e-07, + "loss": 0.4126, + "step": 39391 + }, + { + "epoch": 7.16, + "learning_rate": 8.517291590826356e-07, + "loss": 0.1921, + "step": 39392 + }, + { + "epoch": 7.16, + "learning_rate": 8.516745540589734e-07, + "loss": 0.9953, + "step": 39393 + }, + { + "epoch": 7.16, + "learning_rate": 8.516199490353113e-07, + "loss": 0.0052, + "step": 39394 + }, + { + "epoch": 7.16, + "learning_rate": 8.515653440116491e-07, + "loss": 1.3246, + "step": 39395 + }, + { + "epoch": 7.16, + "learning_rate": 8.515107389879869e-07, + "loss": 0.3855, + "step": 39396 + }, + { + "epoch": 7.16, + "learning_rate": 8.514561339643248e-07, + "loss": 1.3229, + "step": 39397 + }, + { + "epoch": 7.16, + "learning_rate": 8.514015289406625e-07, + "loss": 0.0376, + "step": 39398 + }, + { + "epoch": 7.16, + "learning_rate": 8.513469239170004e-07, + "loss": 0.3996, + "step": 39399 + }, + { + "epoch": 7.16, + "learning_rate": 8.512923188933381e-07, + "loss": 1.9284, + "step": 39400 + }, + { + "epoch": 7.17, + "learning_rate": 8.51237713869676e-07, + "loss": 0.0133, + "step": 39401 + }, + { + "epoch": 7.17, + "learning_rate": 8.511831088460139e-07, + "loss": 0.3313, + "step": 39402 + }, + { + "epoch": 7.17, + "learning_rate": 8.511285038223516e-07, + "loss": 0.0222, + "step": 39403 + }, + { + "epoch": 7.17, + "learning_rate": 8.510738987986896e-07, + "loss": 0.0079, + "step": 39404 + }, + { + "epoch": 7.17, + "learning_rate": 8.510192937750274e-07, + "loss": 1.4564, + "step": 39405 + }, + { + "epoch": 7.17, + "learning_rate": 8.509646887513651e-07, + "loss": 1.6131, + "step": 39406 + }, + { + "epoch": 7.17, + "learning_rate": 8.509100837277029e-07, + "loss": 0.9859, + "step": 39407 + }, + { + "epoch": 7.17, + "learning_rate": 8.508554787040408e-07, + "loss": 0.1625, + "step": 39408 + }, + { + "epoch": 7.17, + "learning_rate": 8.508008736803786e-07, + "loss": 0.0614, + "step": 39409 + }, + { + "epoch": 7.17, + "learning_rate": 8.507462686567164e-07, + "loss": 0.0186, + "step": 39410 + }, + { + "epoch": 7.17, + "learning_rate": 8.506916636330543e-07, + "loss": 0.1102, + "step": 39411 + }, + { + "epoch": 7.17, + "learning_rate": 8.506370586093922e-07, + "loss": 2.5679, + "step": 39412 + }, + { + "epoch": 7.17, + "learning_rate": 8.505824535857299e-07, + "loss": 1.0559, + "step": 39413 + }, + { + "epoch": 7.17, + "learning_rate": 8.505278485620676e-07, + "loss": 0.069, + "step": 39414 + }, + { + "epoch": 7.17, + "learning_rate": 8.504732435384056e-07, + "loss": 0.3874, + "step": 39415 + }, + { + "epoch": 7.17, + "learning_rate": 8.504186385147434e-07, + "loss": 0.2003, + "step": 39416 + }, + { + "epoch": 7.17, + "learning_rate": 8.503640334910811e-07, + "loss": 0.0659, + "step": 39417 + }, + { + "epoch": 7.17, + "learning_rate": 8.503094284674191e-07, + "loss": 0.0099, + "step": 39418 + }, + { + "epoch": 7.17, + "learning_rate": 8.502548234437569e-07, + "loss": 0.0047, + "step": 39419 + }, + { + "epoch": 7.17, + "learning_rate": 8.502002184200946e-07, + "loss": 1.2927, + "step": 39420 + }, + { + "epoch": 7.17, + "learning_rate": 8.501456133964325e-07, + "loss": 0.0093, + "step": 39421 + }, + { + "epoch": 7.17, + "learning_rate": 8.500910083727703e-07, + "loss": 0.0763, + "step": 39422 + }, + { + "epoch": 7.17, + "learning_rate": 8.500364033491082e-07, + "loss": 0.0063, + "step": 39423 + }, + { + "epoch": 7.17, + "learning_rate": 8.499817983254459e-07, + "loss": 0.3631, + "step": 39424 + }, + { + "epoch": 7.17, + "learning_rate": 8.499271933017838e-07, + "loss": 0.2302, + "step": 39425 + }, + { + "epoch": 7.17, + "learning_rate": 8.498725882781217e-07, + "loss": 0.4149, + "step": 39426 + }, + { + "epoch": 7.17, + "learning_rate": 8.498179832544594e-07, + "loss": 0.0129, + "step": 39427 + }, + { + "epoch": 7.17, + "learning_rate": 8.497633782307972e-07, + "loss": 0.7393, + "step": 39428 + }, + { + "epoch": 7.17, + "learning_rate": 8.497087732071352e-07, + "loss": 0.4161, + "step": 39429 + }, + { + "epoch": 7.17, + "learning_rate": 8.496541681834729e-07, + "loss": 0.0133, + "step": 39430 + }, + { + "epoch": 7.17, + "learning_rate": 8.495995631598106e-07, + "loss": 1.5626, + "step": 39431 + }, + { + "epoch": 7.17, + "learning_rate": 8.495449581361485e-07, + "loss": 2.0667, + "step": 39432 + }, + { + "epoch": 7.17, + "learning_rate": 8.494903531124864e-07, + "loss": 0.0842, + "step": 39433 + }, + { + "epoch": 7.17, + "learning_rate": 8.494357480888242e-07, + "loss": 0.4043, + "step": 39434 + }, + { + "epoch": 7.17, + "learning_rate": 8.49381143065162e-07, + "loss": 0.6622, + "step": 39435 + }, + { + "epoch": 7.17, + "learning_rate": 8.493265380414999e-07, + "loss": 1.3983, + "step": 39436 + }, + { + "epoch": 7.17, + "learning_rate": 8.492719330178377e-07, + "loss": 0.0378, + "step": 39437 + }, + { + "epoch": 7.17, + "learning_rate": 8.492173279941754e-07, + "loss": 0.0294, + "step": 39438 + }, + { + "epoch": 7.17, + "learning_rate": 8.491627229705132e-07, + "loss": 0.084, + "step": 39439 + }, + { + "epoch": 7.17, + "learning_rate": 8.491081179468512e-07, + "loss": 2.4628, + "step": 39440 + }, + { + "epoch": 7.17, + "learning_rate": 8.490535129231889e-07, + "loss": 0.8161, + "step": 39441 + }, + { + "epoch": 7.17, + "learning_rate": 8.489989078995268e-07, + "loss": 0.2303, + "step": 39442 + }, + { + "epoch": 7.17, + "learning_rate": 8.489443028758647e-07, + "loss": 0.391, + "step": 39443 + }, + { + "epoch": 7.17, + "learning_rate": 8.488896978522024e-07, + "loss": 0.0113, + "step": 39444 + }, + { + "epoch": 7.17, + "learning_rate": 8.488350928285403e-07, + "loss": 0.2749, + "step": 39445 + }, + { + "epoch": 7.17, + "learning_rate": 8.48780487804878e-07, + "loss": 1.3106, + "step": 39446 + }, + { + "epoch": 7.17, + "learning_rate": 8.487258827812159e-07, + "loss": 0.0145, + "step": 39447 + }, + { + "epoch": 7.17, + "learning_rate": 8.486712777575537e-07, + "loss": 1.3556, + "step": 39448 + }, + { + "epoch": 7.17, + "learning_rate": 8.486166727338915e-07, + "loss": 0.4794, + "step": 39449 + }, + { + "epoch": 7.17, + "learning_rate": 8.485620677102295e-07, + "loss": 1.17, + "step": 39450 + }, + { + "epoch": 7.17, + "learning_rate": 8.485074626865672e-07, + "loss": 2.9536, + "step": 39451 + }, + { + "epoch": 7.17, + "learning_rate": 8.48452857662905e-07, + "loss": 2.2727, + "step": 39452 + }, + { + "epoch": 7.17, + "learning_rate": 8.483982526392428e-07, + "loss": 1.0269, + "step": 39453 + }, + { + "epoch": 7.17, + "learning_rate": 8.483436476155807e-07, + "loss": 0.0066, + "step": 39454 + }, + { + "epoch": 7.17, + "learning_rate": 8.482890425919184e-07, + "loss": 2.2562, + "step": 39455 + }, + { + "epoch": 7.18, + "learning_rate": 8.482344375682563e-07, + "loss": 0.1258, + "step": 39456 + }, + { + "epoch": 7.18, + "learning_rate": 8.481798325445942e-07, + "loss": 0.3958, + "step": 39457 + }, + { + "epoch": 7.18, + "learning_rate": 8.481252275209319e-07, + "loss": 0.9042, + "step": 39458 + }, + { + "epoch": 7.18, + "learning_rate": 8.480706224972698e-07, + "loss": 2.9169, + "step": 39459 + }, + { + "epoch": 7.18, + "learning_rate": 8.480160174736075e-07, + "loss": 0.2471, + "step": 39460 + }, + { + "epoch": 7.18, + "learning_rate": 8.479614124499455e-07, + "loss": 0.9156, + "step": 39461 + }, + { + "epoch": 7.18, + "learning_rate": 8.479068074262832e-07, + "loss": 0.2996, + "step": 39462 + }, + { + "epoch": 7.18, + "learning_rate": 8.47852202402621e-07, + "loss": 1.2211, + "step": 39463 + }, + { + "epoch": 7.18, + "learning_rate": 8.47797597378959e-07, + "loss": 0.1943, + "step": 39464 + }, + { + "epoch": 7.18, + "learning_rate": 8.477429923552967e-07, + "loss": 0.9174, + "step": 39465 + }, + { + "epoch": 7.18, + "learning_rate": 8.476883873316345e-07, + "loss": 1.6279, + "step": 39466 + }, + { + "epoch": 7.18, + "learning_rate": 8.476337823079723e-07, + "loss": 0.1088, + "step": 39467 + }, + { + "epoch": 7.18, + "learning_rate": 8.475791772843102e-07, + "loss": 0.2127, + "step": 39468 + }, + { + "epoch": 7.18, + "learning_rate": 8.47524572260648e-07, + "loss": 0.0047, + "step": 39469 + }, + { + "epoch": 7.18, + "learning_rate": 8.474699672369858e-07, + "loss": 0.1917, + "step": 39470 + }, + { + "epoch": 7.18, + "learning_rate": 8.474153622133237e-07, + "loss": 0.0068, + "step": 39471 + }, + { + "epoch": 7.18, + "learning_rate": 8.473607571896615e-07, + "loss": 0.4358, + "step": 39472 + }, + { + "epoch": 7.18, + "learning_rate": 8.473061521659993e-07, + "loss": 0.0112, + "step": 39473 + }, + { + "epoch": 7.18, + "learning_rate": 8.47251547142337e-07, + "loss": 0.1493, + "step": 39474 + }, + { + "epoch": 7.18, + "learning_rate": 8.47196942118675e-07, + "loss": 0.0299, + "step": 39475 + }, + { + "epoch": 7.18, + "learning_rate": 8.471423370950128e-07, + "loss": 0.0825, + "step": 39476 + }, + { + "epoch": 7.18, + "learning_rate": 8.470877320713505e-07, + "loss": 0.0098, + "step": 39477 + }, + { + "epoch": 7.18, + "learning_rate": 8.470331270476885e-07, + "loss": 0.3771, + "step": 39478 + }, + { + "epoch": 7.18, + "learning_rate": 8.469785220240262e-07, + "loss": 0.5071, + "step": 39479 + }, + { + "epoch": 7.18, + "learning_rate": 8.46923917000364e-07, + "loss": 0.4691, + "step": 39480 + }, + { + "epoch": 7.18, + "learning_rate": 8.468693119767018e-07, + "loss": 0.168, + "step": 39481 + }, + { + "epoch": 7.18, + "learning_rate": 8.468147069530397e-07, + "loss": 0.0292, + "step": 39482 + }, + { + "epoch": 7.18, + "learning_rate": 8.467601019293776e-07, + "loss": 0.0114, + "step": 39483 + }, + { + "epoch": 7.18, + "learning_rate": 8.467054969057153e-07, + "loss": 0.3603, + "step": 39484 + }, + { + "epoch": 7.18, + "learning_rate": 8.466508918820532e-07, + "loss": 0.1661, + "step": 39485 + }, + { + "epoch": 7.18, + "learning_rate": 8.465962868583911e-07, + "loss": 1.2563, + "step": 39486 + }, + { + "epoch": 7.18, + "learning_rate": 8.465416818347288e-07, + "loss": 0.0473, + "step": 39487 + }, + { + "epoch": 7.18, + "learning_rate": 8.464870768110665e-07, + "loss": 1.0818, + "step": 39488 + }, + { + "epoch": 7.18, + "learning_rate": 8.464324717874045e-07, + "loss": 0.5419, + "step": 39489 + }, + { + "epoch": 7.18, + "learning_rate": 8.463778667637423e-07, + "loss": 2.3235, + "step": 39490 + }, + { + "epoch": 7.18, + "learning_rate": 8.4632326174008e-07, + "loss": 0.0995, + "step": 39491 + }, + { + "epoch": 7.18, + "learning_rate": 8.46268656716418e-07, + "loss": 0.277, + "step": 39492 + }, + { + "epoch": 7.18, + "learning_rate": 8.462140516927558e-07, + "loss": 0.471, + "step": 39493 + }, + { + "epoch": 7.18, + "learning_rate": 8.461594466690936e-07, + "loss": 0.5663, + "step": 39494 + }, + { + "epoch": 7.18, + "learning_rate": 8.461048416454313e-07, + "loss": 0.0069, + "step": 39495 + }, + { + "epoch": 7.18, + "learning_rate": 8.460502366217692e-07, + "loss": 0.8667, + "step": 39496 + }, + { + "epoch": 7.18, + "learning_rate": 8.459956315981071e-07, + "loss": 0.3182, + "step": 39497 + }, + { + "epoch": 7.18, + "learning_rate": 8.459410265744448e-07, + "loss": 0.9073, + "step": 39498 + }, + { + "epoch": 7.18, + "learning_rate": 8.458864215507828e-07, + "loss": 1.7266, + "step": 39499 + }, + { + "epoch": 7.18, + "learning_rate": 8.458318165271206e-07, + "loss": 0.4254, + "step": 39500 + }, + { + "epoch": 7.18, + "learning_rate": 8.457772115034583e-07, + "loss": 1.9526, + "step": 39501 + }, + { + "epoch": 7.18, + "learning_rate": 8.45722606479796e-07, + "loss": 1.8034, + "step": 39502 + }, + { + "epoch": 7.18, + "learning_rate": 8.45668001456134e-07, + "loss": 0.2247, + "step": 39503 + }, + { + "epoch": 7.18, + "learning_rate": 8.456133964324718e-07, + "loss": 1.0965, + "step": 39504 + }, + { + "epoch": 7.18, + "learning_rate": 8.455587914088096e-07, + "loss": 0.0464, + "step": 39505 + }, + { + "epoch": 7.18, + "learning_rate": 8.455041863851475e-07, + "loss": 0.0143, + "step": 39506 + }, + { + "epoch": 7.18, + "learning_rate": 8.454495813614853e-07, + "loss": 0.4292, + "step": 39507 + }, + { + "epoch": 7.18, + "learning_rate": 8.453949763378231e-07, + "loss": 0.1157, + "step": 39508 + }, + { + "epoch": 7.18, + "learning_rate": 8.453403713141609e-07, + "loss": 0.0248, + "step": 39509 + }, + { + "epoch": 7.18, + "learning_rate": 8.452857662904989e-07, + "loss": 0.1256, + "step": 39510 + }, + { + "epoch": 7.19, + "learning_rate": 8.452311612668366e-07, + "loss": 3.1834, + "step": 39511 + }, + { + "epoch": 7.19, + "learning_rate": 8.451765562431743e-07, + "loss": 2.5315, + "step": 39512 + }, + { + "epoch": 7.19, + "learning_rate": 8.451219512195122e-07, + "loss": 0.2017, + "step": 39513 + }, + { + "epoch": 7.19, + "learning_rate": 8.450673461958501e-07, + "loss": 0.8927, + "step": 39514 + }, + { + "epoch": 7.19, + "learning_rate": 8.450127411721878e-07, + "loss": 0.4382, + "step": 39515 + }, + { + "epoch": 7.19, + "learning_rate": 8.449581361485257e-07, + "loss": 0.9353, + "step": 39516 + }, + { + "epoch": 7.19, + "learning_rate": 8.449035311248636e-07, + "loss": 0.0052, + "step": 39517 + }, + { + "epoch": 7.19, + "learning_rate": 8.448489261012013e-07, + "loss": 0.0788, + "step": 39518 + }, + { + "epoch": 7.19, + "learning_rate": 8.447943210775391e-07, + "loss": 0.8356, + "step": 39519 + }, + { + "epoch": 7.19, + "learning_rate": 8.447397160538769e-07, + "loss": 0.0561, + "step": 39520 + }, + { + "epoch": 7.19, + "learning_rate": 8.446851110302149e-07, + "loss": 2.349, + "step": 39521 + }, + { + "epoch": 7.19, + "learning_rate": 8.446305060065526e-07, + "loss": 0.73, + "step": 39522 + }, + { + "epoch": 7.19, + "learning_rate": 8.445759009828904e-07, + "loss": 3.0091, + "step": 39523 + }, + { + "epoch": 7.19, + "learning_rate": 8.445212959592284e-07, + "loss": 0.6145, + "step": 39524 + }, + { + "epoch": 7.19, + "learning_rate": 8.444666909355661e-07, + "loss": 1.0267, + "step": 39525 + }, + { + "epoch": 7.19, + "learning_rate": 8.444120859119038e-07, + "loss": 1.3093, + "step": 39526 + }, + { + "epoch": 7.19, + "learning_rate": 8.443574808882417e-07, + "loss": 0.9218, + "step": 39527 + }, + { + "epoch": 7.19, + "learning_rate": 8.443028758645796e-07, + "loss": 0.0068, + "step": 39528 + }, + { + "epoch": 7.19, + "learning_rate": 8.442482708409173e-07, + "loss": 0.0425, + "step": 39529 + }, + { + "epoch": 7.19, + "learning_rate": 8.441936658172552e-07, + "loss": 0.0049, + "step": 39530 + }, + { + "epoch": 7.19, + "learning_rate": 8.441390607935931e-07, + "loss": 1.9382, + "step": 39531 + }, + { + "epoch": 7.19, + "learning_rate": 8.440844557699309e-07, + "loss": 2.4843, + "step": 39532 + }, + { + "epoch": 7.19, + "learning_rate": 8.440298507462687e-07, + "loss": 1.0557, + "step": 39533 + }, + { + "epoch": 7.19, + "learning_rate": 8.439752457226064e-07, + "loss": 0.2687, + "step": 39534 + }, + { + "epoch": 7.19, + "learning_rate": 8.439206406989444e-07, + "loss": 0.9306, + "step": 39535 + }, + { + "epoch": 7.19, + "learning_rate": 8.438660356752821e-07, + "loss": 1.6098, + "step": 39536 + }, + { + "epoch": 7.19, + "learning_rate": 8.438114306516199e-07, + "loss": 0.1585, + "step": 39537 + }, + { + "epoch": 7.19, + "learning_rate": 8.437568256279579e-07, + "loss": 0.342, + "step": 39538 + }, + { + "epoch": 7.19, + "learning_rate": 8.437022206042956e-07, + "loss": 0.2626, + "step": 39539 + }, + { + "epoch": 7.19, + "learning_rate": 8.436476155806335e-07, + "loss": 0.0252, + "step": 39540 + }, + { + "epoch": 7.19, + "learning_rate": 8.435930105569712e-07, + "loss": 0.5634, + "step": 39541 + }, + { + "epoch": 7.19, + "learning_rate": 8.435384055333091e-07, + "loss": 0.5191, + "step": 39542 + }, + { + "epoch": 7.19, + "learning_rate": 8.434838005096469e-07, + "loss": 0.1053, + "step": 39543 + }, + { + "epoch": 7.19, + "learning_rate": 8.434291954859847e-07, + "loss": 0.1074, + "step": 39544 + }, + { + "epoch": 7.19, + "learning_rate": 8.433745904623226e-07, + "loss": 1.898, + "step": 39545 + }, + { + "epoch": 7.19, + "learning_rate": 8.433199854386604e-07, + "loss": 0.3206, + "step": 39546 + }, + { + "epoch": 7.19, + "learning_rate": 8.432653804149982e-07, + "loss": 0.2749, + "step": 39547 + }, + { + "epoch": 7.19, + "learning_rate": 8.432107753913359e-07, + "loss": 1.0038, + "step": 39548 + }, + { + "epoch": 7.19, + "learning_rate": 8.431561703676739e-07, + "loss": 0.0617, + "step": 39549 + }, + { + "epoch": 7.19, + "learning_rate": 8.431015653440117e-07, + "loss": 0.4329, + "step": 39550 + }, + { + "epoch": 7.19, + "learning_rate": 8.430469603203495e-07, + "loss": 0.2643, + "step": 39551 + }, + { + "epoch": 7.19, + "learning_rate": 8.429923552966874e-07, + "loss": 0.9183, + "step": 39552 + }, + { + "epoch": 7.19, + "learning_rate": 8.429377502730251e-07, + "loss": 0.3699, + "step": 39553 + }, + { + "epoch": 7.19, + "learning_rate": 8.42883145249363e-07, + "loss": 0.0595, + "step": 39554 + }, + { + "epoch": 7.19, + "learning_rate": 8.428285402257007e-07, + "loss": 0.4755, + "step": 39555 + }, + { + "epoch": 7.19, + "learning_rate": 8.427739352020386e-07, + "loss": 0.6968, + "step": 39556 + }, + { + "epoch": 7.19, + "learning_rate": 8.427193301783765e-07, + "loss": 0.0062, + "step": 39557 + }, + { + "epoch": 7.19, + "learning_rate": 8.426647251547142e-07, + "loss": 1.5006, + "step": 39558 + }, + { + "epoch": 7.19, + "learning_rate": 8.426101201310522e-07, + "loss": 0.0045, + "step": 39559 + }, + { + "epoch": 7.19, + "learning_rate": 8.425555151073899e-07, + "loss": 1.253, + "step": 39560 + }, + { + "epoch": 7.19, + "learning_rate": 8.425009100837277e-07, + "loss": 0.1011, + "step": 39561 + }, + { + "epoch": 7.19, + "learning_rate": 8.424463050600655e-07, + "loss": 0.0571, + "step": 39562 + }, + { + "epoch": 7.19, + "learning_rate": 8.423917000364034e-07, + "loss": 0.9864, + "step": 39563 + }, + { + "epoch": 7.19, + "learning_rate": 8.423370950127412e-07, + "loss": 0.0687, + "step": 39564 + }, + { + "epoch": 7.19, + "learning_rate": 8.42282489989079e-07, + "loss": 0.1866, + "step": 39565 + }, + { + "epoch": 7.2, + "learning_rate": 8.422278849654169e-07, + "loss": 0.0092, + "step": 39566 + }, + { + "epoch": 7.2, + "learning_rate": 8.421732799417546e-07, + "loss": 0.0219, + "step": 39567 + }, + { + "epoch": 7.2, + "learning_rate": 8.421186749180925e-07, + "loss": 0.0154, + "step": 39568 + }, + { + "epoch": 7.2, + "learning_rate": 8.420640698944302e-07, + "loss": 1.9062, + "step": 39569 + }, + { + "epoch": 7.2, + "learning_rate": 8.420094648707682e-07, + "loss": 2.5161, + "step": 39570 + }, + { + "epoch": 7.2, + "learning_rate": 8.41954859847106e-07, + "loss": 1.0819, + "step": 39571 + }, + { + "epoch": 7.2, + "learning_rate": 8.419002548234437e-07, + "loss": 1.1475, + "step": 39572 + }, + { + "epoch": 7.2, + "learning_rate": 8.418456497997817e-07, + "loss": 0.0117, + "step": 39573 + }, + { + "epoch": 7.2, + "learning_rate": 8.417910447761195e-07, + "loss": 0.0044, + "step": 39574 + }, + { + "epoch": 7.2, + "learning_rate": 8.417364397524572e-07, + "loss": 0.0954, + "step": 39575 + }, + { + "epoch": 7.2, + "learning_rate": 8.41681834728795e-07, + "loss": 0.4927, + "step": 39576 + }, + { + "epoch": 7.2, + "learning_rate": 8.416272297051329e-07, + "loss": 0.2515, + "step": 39577 + }, + { + "epoch": 7.2, + "learning_rate": 8.415726246814707e-07, + "loss": 0.4671, + "step": 39578 + }, + { + "epoch": 7.2, + "learning_rate": 8.415180196578085e-07, + "loss": 0.6809, + "step": 39579 + }, + { + "epoch": 7.2, + "learning_rate": 8.414634146341464e-07, + "loss": 0.464, + "step": 39580 + }, + { + "epoch": 7.2, + "learning_rate": 8.414088096104843e-07, + "loss": 0.3965, + "step": 39581 + }, + { + "epoch": 7.2, + "learning_rate": 8.41354204586822e-07, + "loss": 0.0077, + "step": 39582 + }, + { + "epoch": 7.2, + "learning_rate": 8.412995995631597e-07, + "loss": 0.746, + "step": 39583 + }, + { + "epoch": 7.2, + "learning_rate": 8.412449945394977e-07, + "loss": 0.0048, + "step": 39584 + }, + { + "epoch": 7.2, + "learning_rate": 8.411903895158355e-07, + "loss": 1.1931, + "step": 39585 + }, + { + "epoch": 7.2, + "learning_rate": 8.411357844921732e-07, + "loss": 0.0928, + "step": 39586 + }, + { + "epoch": 7.2, + "learning_rate": 8.410811794685111e-07, + "loss": 0.1406, + "step": 39587 + }, + { + "epoch": 7.2, + "learning_rate": 8.41026574444849e-07, + "loss": 0.7539, + "step": 39588 + }, + { + "epoch": 7.2, + "learning_rate": 8.409719694211868e-07, + "loss": 1.8593, + "step": 39589 + }, + { + "epoch": 7.2, + "learning_rate": 8.409173643975245e-07, + "loss": 0.8577, + "step": 39590 + }, + { + "epoch": 7.2, + "learning_rate": 8.408627593738624e-07, + "loss": 2.0574, + "step": 39591 + }, + { + "epoch": 7.2, + "learning_rate": 8.408081543502003e-07, + "loss": 0.0047, + "step": 39592 + }, + { + "epoch": 7.2, + "learning_rate": 8.40753549326538e-07, + "loss": 0.1567, + "step": 39593 + }, + { + "epoch": 7.2, + "learning_rate": 8.406989443028758e-07, + "loss": 0.1994, + "step": 39594 + }, + { + "epoch": 7.2, + "learning_rate": 8.406443392792138e-07, + "loss": 1.4002, + "step": 39595 + }, + { + "epoch": 7.2, + "learning_rate": 8.405897342555515e-07, + "loss": 1.8832, + "step": 39596 + }, + { + "epoch": 7.2, + "learning_rate": 8.405351292318893e-07, + "loss": 0.9984, + "step": 39597 + }, + { + "epoch": 7.2, + "learning_rate": 8.404805242082273e-07, + "loss": 0.2594, + "step": 39598 + }, + { + "epoch": 7.2, + "learning_rate": 8.40425919184565e-07, + "loss": 0.0562, + "step": 39599 + }, + { + "epoch": 7.2, + "learning_rate": 8.403713141609028e-07, + "loss": 0.0927, + "step": 39600 + }, + { + "epoch": 7.2, + "learning_rate": 8.403167091372406e-07, + "loss": 0.055, + "step": 39601 + }, + { + "epoch": 7.2, + "learning_rate": 8.402621041135785e-07, + "loss": 0.0099, + "step": 39602 + }, + { + "epoch": 7.2, + "learning_rate": 8.402074990899163e-07, + "loss": 0.5736, + "step": 39603 + }, + { + "epoch": 7.2, + "learning_rate": 8.401528940662541e-07, + "loss": 0.4716, + "step": 39604 + }, + { + "epoch": 7.2, + "learning_rate": 8.40098289042592e-07, + "loss": 0.0098, + "step": 39605 + }, + { + "epoch": 7.2, + "learning_rate": 8.400436840189298e-07, + "loss": 0.4346, + "step": 39606 + }, + { + "epoch": 7.2, + "learning_rate": 8.399890789952675e-07, + "loss": 0.0342, + "step": 39607 + }, + { + "epoch": 7.2, + "learning_rate": 8.399344739716053e-07, + "loss": 0.554, + "step": 39608 + }, + { + "epoch": 7.2, + "learning_rate": 8.398798689479433e-07, + "loss": 0.4036, + "step": 39609 + }, + { + "epoch": 7.2, + "learning_rate": 8.39825263924281e-07, + "loss": 0.026, + "step": 39610 + }, + { + "epoch": 7.2, + "learning_rate": 8.397706589006189e-07, + "loss": 0.0593, + "step": 39611 + }, + { + "epoch": 7.2, + "learning_rate": 8.397160538769568e-07, + "loss": 0.0062, + "step": 39612 + }, + { + "epoch": 7.2, + "learning_rate": 8.396614488532945e-07, + "loss": 1.3302, + "step": 39613 + }, + { + "epoch": 7.2, + "learning_rate": 8.396068438296324e-07, + "loss": 0.0068, + "step": 39614 + }, + { + "epoch": 7.2, + "learning_rate": 8.395522388059701e-07, + "loss": 0.016, + "step": 39615 + }, + { + "epoch": 7.2, + "learning_rate": 8.39497633782308e-07, + "loss": 0.2672, + "step": 39616 + }, + { + "epoch": 7.2, + "learning_rate": 8.394430287586458e-07, + "loss": 0.5294, + "step": 39617 + }, + { + "epoch": 7.2, + "learning_rate": 8.393884237349836e-07, + "loss": 1.6418, + "step": 39618 + }, + { + "epoch": 7.2, + "learning_rate": 8.393338187113216e-07, + "loss": 0.0072, + "step": 39619 + }, + { + "epoch": 7.2, + "learning_rate": 8.392792136876593e-07, + "loss": 0.7622, + "step": 39620 + }, + { + "epoch": 7.21, + "learning_rate": 8.392246086639971e-07, + "loss": 2.4667, + "step": 39621 + }, + { + "epoch": 7.21, + "learning_rate": 8.391700036403349e-07, + "loss": 0.5319, + "step": 39622 + }, + { + "epoch": 7.21, + "learning_rate": 8.391153986166728e-07, + "loss": 0.1401, + "step": 39623 + }, + { + "epoch": 7.21, + "learning_rate": 8.390607935930105e-07, + "loss": 0.0193, + "step": 39624 + }, + { + "epoch": 7.21, + "learning_rate": 8.390061885693484e-07, + "loss": 1.165, + "step": 39625 + }, + { + "epoch": 7.21, + "learning_rate": 8.389515835456863e-07, + "loss": 1.1255, + "step": 39626 + }, + { + "epoch": 7.21, + "learning_rate": 8.38896978522024e-07, + "loss": 0.3322, + "step": 39627 + }, + { + "epoch": 7.21, + "learning_rate": 8.388423734983619e-07, + "loss": 0.3083, + "step": 39628 + }, + { + "epoch": 7.21, + "learning_rate": 8.387877684746996e-07, + "loss": 0.147, + "step": 39629 + }, + { + "epoch": 7.21, + "learning_rate": 8.387331634510376e-07, + "loss": 0.6605, + "step": 39630 + }, + { + "epoch": 7.21, + "learning_rate": 8.386785584273753e-07, + "loss": 0.0137, + "step": 39631 + }, + { + "epoch": 7.21, + "learning_rate": 8.386239534037131e-07, + "loss": 0.6374, + "step": 39632 + }, + { + "epoch": 7.21, + "learning_rate": 8.385693483800511e-07, + "loss": 0.0617, + "step": 39633 + }, + { + "epoch": 7.21, + "learning_rate": 8.385147433563888e-07, + "loss": 0.2413, + "step": 39634 + }, + { + "epoch": 7.21, + "learning_rate": 8.384601383327266e-07, + "loss": 1.5826, + "step": 39635 + }, + { + "epoch": 7.21, + "learning_rate": 8.384055333090644e-07, + "loss": 0.0381, + "step": 39636 + }, + { + "epoch": 7.21, + "learning_rate": 8.383509282854023e-07, + "loss": 0.0142, + "step": 39637 + }, + { + "epoch": 7.21, + "learning_rate": 8.382963232617402e-07, + "loss": 0.6988, + "step": 39638 + }, + { + "epoch": 7.21, + "learning_rate": 8.382417182380779e-07, + "loss": 0.579, + "step": 39639 + }, + { + "epoch": 7.21, + "learning_rate": 8.381871132144158e-07, + "loss": 0.0048, + "step": 39640 + }, + { + "epoch": 7.21, + "learning_rate": 8.381325081907536e-07, + "loss": 0.0303, + "step": 39641 + }, + { + "epoch": 7.21, + "learning_rate": 8.380779031670914e-07, + "loss": 0.0092, + "step": 39642 + }, + { + "epoch": 7.21, + "learning_rate": 8.380232981434291e-07, + "loss": 1.4651, + "step": 39643 + }, + { + "epoch": 7.21, + "learning_rate": 8.379686931197671e-07, + "loss": 0.0909, + "step": 39644 + }, + { + "epoch": 7.21, + "learning_rate": 8.379140880961049e-07, + "loss": 0.9796, + "step": 39645 + }, + { + "epoch": 7.21, + "learning_rate": 8.378594830724426e-07, + "loss": 0.0098, + "step": 39646 + }, + { + "epoch": 7.21, + "learning_rate": 8.378048780487806e-07, + "loss": 0.009, + "step": 39647 + }, + { + "epoch": 7.21, + "learning_rate": 8.377502730251183e-07, + "loss": 0.0527, + "step": 39648 + }, + { + "epoch": 7.21, + "learning_rate": 8.376956680014562e-07, + "loss": 0.0259, + "step": 39649 + }, + { + "epoch": 7.21, + "learning_rate": 8.376410629777939e-07, + "loss": 0.1176, + "step": 39650 + }, + { + "epoch": 7.21, + "learning_rate": 8.375864579541318e-07, + "loss": 0.0338, + "step": 39651 + }, + { + "epoch": 7.21, + "learning_rate": 8.375318529304697e-07, + "loss": 0.0059, + "step": 39652 + }, + { + "epoch": 7.21, + "learning_rate": 8.374772479068074e-07, + "loss": 0.2146, + "step": 39653 + }, + { + "epoch": 7.21, + "learning_rate": 8.374226428831453e-07, + "loss": 0.3806, + "step": 39654 + }, + { + "epoch": 7.21, + "learning_rate": 8.373680378594831e-07, + "loss": 0.0841, + "step": 39655 + }, + { + "epoch": 7.21, + "learning_rate": 8.373134328358209e-07, + "loss": 0.164, + "step": 39656 + }, + { + "epoch": 7.21, + "learning_rate": 8.372588278121586e-07, + "loss": 0.0288, + "step": 39657 + }, + { + "epoch": 7.21, + "learning_rate": 8.372042227884966e-07, + "loss": 0.3938, + "step": 39658 + }, + { + "epoch": 7.21, + "learning_rate": 8.371496177648344e-07, + "loss": 0.1655, + "step": 39659 + }, + { + "epoch": 7.21, + "learning_rate": 8.370950127411722e-07, + "loss": 0.17, + "step": 39660 + }, + { + "epoch": 7.21, + "learning_rate": 8.370404077175101e-07, + "loss": 0.0789, + "step": 39661 + }, + { + "epoch": 7.21, + "learning_rate": 8.369858026938479e-07, + "loss": 0.4524, + "step": 39662 + }, + { + "epoch": 7.21, + "learning_rate": 8.369311976701857e-07, + "loss": 2.1166, + "step": 39663 + }, + { + "epoch": 7.21, + "learning_rate": 8.368765926465234e-07, + "loss": 0.8909, + "step": 39664 + }, + { + "epoch": 7.21, + "learning_rate": 8.368219876228613e-07, + "loss": 0.0581, + "step": 39665 + }, + { + "epoch": 7.21, + "learning_rate": 8.367673825991992e-07, + "loss": 0.9763, + "step": 39666 + }, + { + "epoch": 7.21, + "learning_rate": 8.367127775755369e-07, + "loss": 0.3559, + "step": 39667 + }, + { + "epoch": 7.21, + "learning_rate": 8.366581725518747e-07, + "loss": 0.8805, + "step": 39668 + }, + { + "epoch": 7.21, + "learning_rate": 8.366035675282127e-07, + "loss": 0.0087, + "step": 39669 + }, + { + "epoch": 7.21, + "learning_rate": 8.365489625045504e-07, + "loss": 0.0154, + "step": 39670 + }, + { + "epoch": 7.21, + "learning_rate": 8.364943574808882e-07, + "loss": 0.4053, + "step": 39671 + }, + { + "epoch": 7.21, + "learning_rate": 8.364397524572261e-07, + "loss": 0.4498, + "step": 39672 + }, + { + "epoch": 7.21, + "learning_rate": 8.363851474335639e-07, + "loss": 0.9188, + "step": 39673 + }, + { + "epoch": 7.21, + "learning_rate": 8.363305424099017e-07, + "loss": 0.0904, + "step": 39674 + }, + { + "epoch": 7.21, + "learning_rate": 8.362759373862395e-07, + "loss": 0.0466, + "step": 39675 + }, + { + "epoch": 7.22, + "learning_rate": 8.362213323625774e-07, + "loss": 0.2884, + "step": 39676 + }, + { + "epoch": 7.22, + "learning_rate": 8.361667273389152e-07, + "loss": 0.0145, + "step": 39677 + }, + { + "epoch": 7.22, + "learning_rate": 8.36112122315253e-07, + "loss": 0.0536, + "step": 39678 + }, + { + "epoch": 7.22, + "learning_rate": 8.36057517291591e-07, + "loss": 0.9697, + "step": 39679 + }, + { + "epoch": 7.22, + "learning_rate": 8.360029122679287e-07, + "loss": 0.2149, + "step": 39680 + }, + { + "epoch": 7.22, + "learning_rate": 8.359483072442664e-07, + "loss": 0.1391, + "step": 39681 + }, + { + "epoch": 7.22, + "learning_rate": 8.358937022206043e-07, + "loss": 0.0311, + "step": 39682 + }, + { + "epoch": 7.22, + "learning_rate": 8.358390971969422e-07, + "loss": 0.6793, + "step": 39683 + }, + { + "epoch": 7.22, + "learning_rate": 8.357844921732799e-07, + "loss": 0.8066, + "step": 39684 + }, + { + "epoch": 7.22, + "learning_rate": 8.357298871496178e-07, + "loss": 0.8503, + "step": 39685 + }, + { + "epoch": 7.22, + "learning_rate": 8.356752821259557e-07, + "loss": 0.614, + "step": 39686 + }, + { + "epoch": 7.22, + "learning_rate": 8.356206771022935e-07, + "loss": 1.5217, + "step": 39687 + }, + { + "epoch": 7.22, + "learning_rate": 8.355660720786312e-07, + "loss": 0.3671, + "step": 39688 + }, + { + "epoch": 7.22, + "learning_rate": 8.35511467054969e-07, + "loss": 0.0655, + "step": 39689 + }, + { + "epoch": 7.22, + "learning_rate": 8.35456862031307e-07, + "loss": 0.0785, + "step": 39690 + }, + { + "epoch": 7.22, + "learning_rate": 8.354022570076447e-07, + "loss": 0.0235, + "step": 39691 + }, + { + "epoch": 7.22, + "learning_rate": 8.353476519839825e-07, + "loss": 0.0851, + "step": 39692 + }, + { + "epoch": 7.22, + "learning_rate": 8.352930469603205e-07, + "loss": 0.0061, + "step": 39693 + }, + { + "epoch": 7.22, + "learning_rate": 8.352384419366582e-07, + "loss": 0.5665, + "step": 39694 + }, + { + "epoch": 7.22, + "learning_rate": 8.351838369129959e-07, + "loss": 0.9195, + "step": 39695 + }, + { + "epoch": 7.22, + "learning_rate": 8.351292318893338e-07, + "loss": 0.0038, + "step": 39696 + }, + { + "epoch": 7.22, + "learning_rate": 8.350746268656717e-07, + "loss": 0.0071, + "step": 39697 + }, + { + "epoch": 7.22, + "learning_rate": 8.350200218420095e-07, + "loss": 0.1178, + "step": 39698 + }, + { + "epoch": 7.22, + "learning_rate": 8.349654168183473e-07, + "loss": 0.1865, + "step": 39699 + }, + { + "epoch": 7.22, + "learning_rate": 8.349108117946852e-07, + "loss": 0.0094, + "step": 39700 + }, + { + "epoch": 7.22, + "learning_rate": 8.34856206771023e-07, + "loss": 1.3971, + "step": 39701 + }, + { + "epoch": 7.22, + "learning_rate": 8.348016017473608e-07, + "loss": 0.0628, + "step": 39702 + }, + { + "epoch": 7.22, + "learning_rate": 8.347469967236985e-07, + "loss": 0.5687, + "step": 39703 + }, + { + "epoch": 7.22, + "learning_rate": 8.346923917000365e-07, + "loss": 0.4063, + "step": 39704 + }, + { + "epoch": 7.22, + "learning_rate": 8.346377866763742e-07, + "loss": 0.0332, + "step": 39705 + }, + { + "epoch": 7.22, + "learning_rate": 8.34583181652712e-07, + "loss": 0.0429, + "step": 39706 + }, + { + "epoch": 7.22, + "learning_rate": 8.3452857662905e-07, + "loss": 0.2615, + "step": 39707 + }, + { + "epoch": 7.22, + "learning_rate": 8.344739716053877e-07, + "loss": 0.1765, + "step": 39708 + }, + { + "epoch": 7.22, + "learning_rate": 8.344193665817256e-07, + "loss": 1.0308, + "step": 39709 + }, + { + "epoch": 7.22, + "learning_rate": 8.343647615580633e-07, + "loss": 0.011, + "step": 39710 + }, + { + "epoch": 7.22, + "learning_rate": 8.343101565344012e-07, + "loss": 0.2133, + "step": 39711 + }, + { + "epoch": 7.22, + "learning_rate": 8.34255551510739e-07, + "loss": 1.7642, + "step": 39712 + }, + { + "epoch": 7.22, + "learning_rate": 8.342009464870768e-07, + "loss": 0.4183, + "step": 39713 + }, + { + "epoch": 7.22, + "learning_rate": 8.341463414634147e-07, + "loss": 1.0706, + "step": 39714 + }, + { + "epoch": 7.22, + "learning_rate": 8.340917364397525e-07, + "loss": 0.0298, + "step": 39715 + }, + { + "epoch": 7.22, + "learning_rate": 8.340371314160903e-07, + "loss": 1.1618, + "step": 39716 + }, + { + "epoch": 7.22, + "learning_rate": 8.33982526392428e-07, + "loss": 0.1103, + "step": 39717 + }, + { + "epoch": 7.22, + "learning_rate": 8.33927921368766e-07, + "loss": 0.0077, + "step": 39718 + }, + { + "epoch": 7.22, + "learning_rate": 8.338733163451038e-07, + "loss": 0.0973, + "step": 39719 + }, + { + "epoch": 7.22, + "learning_rate": 8.338187113214416e-07, + "loss": 1.8666, + "step": 39720 + }, + { + "epoch": 7.22, + "learning_rate": 8.337641062977795e-07, + "loss": 0.858, + "step": 39721 + }, + { + "epoch": 7.22, + "learning_rate": 8.337095012741172e-07, + "loss": 2.5636, + "step": 39722 + }, + { + "epoch": 7.22, + "learning_rate": 8.336548962504551e-07, + "loss": 0.7468, + "step": 39723 + }, + { + "epoch": 7.22, + "learning_rate": 8.336002912267928e-07, + "loss": 1.8532, + "step": 39724 + }, + { + "epoch": 7.22, + "learning_rate": 8.335456862031307e-07, + "loss": 0.1849, + "step": 39725 + }, + { + "epoch": 7.22, + "learning_rate": 8.334910811794686e-07, + "loss": 1.4353, + "step": 39726 + }, + { + "epoch": 7.22, + "learning_rate": 8.334364761558063e-07, + "loss": 0.0281, + "step": 39727 + }, + { + "epoch": 7.22, + "learning_rate": 8.333818711321443e-07, + "loss": 0.385, + "step": 39728 + }, + { + "epoch": 7.22, + "learning_rate": 8.33327266108482e-07, + "loss": 0.4601, + "step": 39729 + }, + { + "epoch": 7.22, + "learning_rate": 8.332726610848198e-07, + "loss": 1.2201, + "step": 39730 + }, + { + "epoch": 7.23, + "learning_rate": 8.332180560611576e-07, + "loss": 0.1271, + "step": 39731 + }, + { + "epoch": 7.23, + "learning_rate": 8.331634510374955e-07, + "loss": 1.5682, + "step": 39732 + }, + { + "epoch": 7.23, + "learning_rate": 8.331088460138333e-07, + "loss": 0.6087, + "step": 39733 + }, + { + "epoch": 7.23, + "learning_rate": 8.330542409901711e-07, + "loss": 1.5373, + "step": 39734 + }, + { + "epoch": 7.23, + "learning_rate": 8.32999635966509e-07, + "loss": 0.0763, + "step": 39735 + }, + { + "epoch": 7.23, + "learning_rate": 8.329450309428467e-07, + "loss": 0.271, + "step": 39736 + }, + { + "epoch": 7.23, + "learning_rate": 8.328904259191846e-07, + "loss": 0.0652, + "step": 39737 + }, + { + "epoch": 7.23, + "learning_rate": 8.328358208955223e-07, + "loss": 0.6806, + "step": 39738 + }, + { + "epoch": 7.23, + "learning_rate": 8.327812158718603e-07, + "loss": 0.3857, + "step": 39739 + }, + { + "epoch": 7.23, + "learning_rate": 8.327266108481981e-07, + "loss": 0.43, + "step": 39740 + }, + { + "epoch": 7.23, + "learning_rate": 8.326720058245358e-07, + "loss": 2.1094, + "step": 39741 + }, + { + "epoch": 7.23, + "learning_rate": 8.326174008008737e-07, + "loss": 0.0276, + "step": 39742 + }, + { + "epoch": 7.23, + "learning_rate": 8.325627957772116e-07, + "loss": 0.1559, + "step": 39743 + }, + { + "epoch": 7.23, + "learning_rate": 8.325081907535493e-07, + "loss": 0.1403, + "step": 39744 + }, + { + "epoch": 7.23, + "learning_rate": 8.324535857298871e-07, + "loss": 0.3615, + "step": 39745 + }, + { + "epoch": 7.23, + "learning_rate": 8.32398980706225e-07, + "loss": 0.0206, + "step": 39746 + }, + { + "epoch": 7.23, + "learning_rate": 8.323443756825629e-07, + "loss": 1.1796, + "step": 39747 + }, + { + "epoch": 7.23, + "learning_rate": 8.322897706589006e-07, + "loss": 0.0053, + "step": 39748 + }, + { + "epoch": 7.23, + "learning_rate": 8.322351656352384e-07, + "loss": 0.0684, + "step": 39749 + }, + { + "epoch": 7.23, + "learning_rate": 8.321805606115764e-07, + "loss": 0.9541, + "step": 39750 + }, + { + "epoch": 7.23, + "learning_rate": 8.321259555879141e-07, + "loss": 0.0765, + "step": 39751 + }, + { + "epoch": 7.23, + "learning_rate": 8.320713505642518e-07, + "loss": 0.6658, + "step": 39752 + }, + { + "epoch": 7.23, + "learning_rate": 8.320167455405898e-07, + "loss": 2.5635, + "step": 39753 + }, + { + "epoch": 7.23, + "learning_rate": 8.319621405169276e-07, + "loss": 0.9414, + "step": 39754 + }, + { + "epoch": 7.23, + "learning_rate": 8.319075354932653e-07, + "loss": 0.1114, + "step": 39755 + }, + { + "epoch": 7.23, + "learning_rate": 8.318529304696032e-07, + "loss": 0.2563, + "step": 39756 + }, + { + "epoch": 7.23, + "learning_rate": 8.317983254459411e-07, + "loss": 0.3138, + "step": 39757 + }, + { + "epoch": 7.23, + "learning_rate": 8.317437204222789e-07, + "loss": 0.1036, + "step": 39758 + }, + { + "epoch": 7.23, + "learning_rate": 8.316891153986166e-07, + "loss": 0.0615, + "step": 39759 + }, + { + "epoch": 7.23, + "learning_rate": 8.316345103749545e-07, + "loss": 2.1647, + "step": 39760 + }, + { + "epoch": 7.23, + "learning_rate": 8.315799053512924e-07, + "loss": 0.0268, + "step": 39761 + }, + { + "epoch": 7.23, + "learning_rate": 8.315253003276301e-07, + "loss": 0.2106, + "step": 39762 + }, + { + "epoch": 7.23, + "learning_rate": 8.314706953039679e-07, + "loss": 0.1701, + "step": 39763 + }, + { + "epoch": 7.23, + "learning_rate": 8.314160902803059e-07, + "loss": 0.2269, + "step": 39764 + }, + { + "epoch": 7.23, + "learning_rate": 8.313614852566436e-07, + "loss": 0.8959, + "step": 39765 + }, + { + "epoch": 7.23, + "learning_rate": 8.313068802329815e-07, + "loss": 0.2518, + "step": 39766 + }, + { + "epoch": 7.23, + "learning_rate": 8.312522752093194e-07, + "loss": 0.0087, + "step": 39767 + }, + { + "epoch": 7.23, + "learning_rate": 8.311976701856571e-07, + "loss": 0.0038, + "step": 39768 + }, + { + "epoch": 7.23, + "learning_rate": 8.311430651619949e-07, + "loss": 0.0276, + "step": 39769 + }, + { + "epoch": 7.23, + "learning_rate": 8.310884601383327e-07, + "loss": 1.0854, + "step": 39770 + }, + { + "epoch": 7.23, + "learning_rate": 8.310338551146706e-07, + "loss": 0.1477, + "step": 39771 + }, + { + "epoch": 7.23, + "learning_rate": 8.309792500910084e-07, + "loss": 0.1396, + "step": 39772 + }, + { + "epoch": 7.23, + "learning_rate": 8.309246450673462e-07, + "loss": 0.4654, + "step": 39773 + }, + { + "epoch": 7.23, + "learning_rate": 8.308700400436842e-07, + "loss": 0.3145, + "step": 39774 + }, + { + "epoch": 7.23, + "learning_rate": 8.308154350200219e-07, + "loss": 0.5047, + "step": 39775 + }, + { + "epoch": 7.23, + "learning_rate": 8.307608299963596e-07, + "loss": 0.1315, + "step": 39776 + }, + { + "epoch": 7.23, + "learning_rate": 8.307062249726975e-07, + "loss": 1.1977, + "step": 39777 + }, + { + "epoch": 7.23, + "learning_rate": 8.306516199490354e-07, + "loss": 0.6705, + "step": 39778 + }, + { + "epoch": 7.23, + "learning_rate": 8.305970149253731e-07, + "loss": 0.0978, + "step": 39779 + }, + { + "epoch": 7.23, + "learning_rate": 8.30542409901711e-07, + "loss": 0.0699, + "step": 39780 + }, + { + "epoch": 7.23, + "learning_rate": 8.304878048780489e-07, + "loss": 0.2794, + "step": 39781 + }, + { + "epoch": 7.23, + "learning_rate": 8.304331998543866e-07, + "loss": 0.6263, + "step": 39782 + }, + { + "epoch": 7.23, + "learning_rate": 8.303785948307245e-07, + "loss": 0.6881, + "step": 39783 + }, + { + "epoch": 7.23, + "learning_rate": 8.303239898070622e-07, + "loss": 1.3401, + "step": 39784 + }, + { + "epoch": 7.23, + "learning_rate": 8.302693847834002e-07, + "loss": 0.319, + "step": 39785 + }, + { + "epoch": 7.24, + "learning_rate": 8.302147797597379e-07, + "loss": 0.8547, + "step": 39786 + }, + { + "epoch": 7.24, + "learning_rate": 8.301601747360757e-07, + "loss": 0.209, + "step": 39787 + }, + { + "epoch": 7.24, + "learning_rate": 8.301055697124137e-07, + "loss": 0.3277, + "step": 39788 + }, + { + "epoch": 7.24, + "learning_rate": 8.300509646887514e-07, + "loss": 0.4703, + "step": 39789 + }, + { + "epoch": 7.24, + "learning_rate": 8.299963596650892e-07, + "loss": 0.0638, + "step": 39790 + }, + { + "epoch": 7.24, + "learning_rate": 8.29941754641427e-07, + "loss": 1.5546, + "step": 39791 + }, + { + "epoch": 7.24, + "learning_rate": 8.298871496177649e-07, + "loss": 0.3019, + "step": 39792 + }, + { + "epoch": 7.24, + "learning_rate": 8.298325445941026e-07, + "loss": 0.0153, + "step": 39793 + }, + { + "epoch": 7.24, + "learning_rate": 8.297779395704405e-07, + "loss": 0.082, + "step": 39794 + }, + { + "epoch": 7.24, + "learning_rate": 8.297233345467784e-07, + "loss": 0.65, + "step": 39795 + }, + { + "epoch": 7.24, + "learning_rate": 8.296687295231162e-07, + "loss": 0.0076, + "step": 39796 + }, + { + "epoch": 7.24, + "learning_rate": 8.29614124499454e-07, + "loss": 1.2106, + "step": 39797 + }, + { + "epoch": 7.24, + "learning_rate": 8.295595194757917e-07, + "loss": 0.1235, + "step": 39798 + }, + { + "epoch": 7.24, + "learning_rate": 8.295049144521297e-07, + "loss": 0.4598, + "step": 39799 + }, + { + "epoch": 7.24, + "learning_rate": 8.294503094284674e-07, + "loss": 0.8252, + "step": 39800 + }, + { + "epoch": 7.24, + "learning_rate": 8.293957044048052e-07, + "loss": 0.846, + "step": 39801 + }, + { + "epoch": 7.24, + "learning_rate": 8.293410993811432e-07, + "loss": 0.0137, + "step": 39802 + }, + { + "epoch": 7.24, + "learning_rate": 8.292864943574809e-07, + "loss": 0.3838, + "step": 39803 + }, + { + "epoch": 7.24, + "learning_rate": 8.292318893338187e-07, + "loss": 0.715, + "step": 39804 + }, + { + "epoch": 7.24, + "learning_rate": 8.291772843101565e-07, + "loss": 0.0425, + "step": 39805 + }, + { + "epoch": 7.24, + "learning_rate": 8.291226792864944e-07, + "loss": 1.3792, + "step": 39806 + }, + { + "epoch": 7.24, + "learning_rate": 8.290680742628323e-07, + "loss": 0.0126, + "step": 39807 + }, + { + "epoch": 7.24, + "learning_rate": 8.2901346923917e-07, + "loss": 0.3251, + "step": 39808 + }, + { + "epoch": 7.24, + "learning_rate": 8.289588642155079e-07, + "loss": 0.084, + "step": 39809 + }, + { + "epoch": 7.24, + "learning_rate": 8.289042591918457e-07, + "loss": 1.2205, + "step": 39810 + }, + { + "epoch": 7.24, + "learning_rate": 8.288496541681835e-07, + "loss": 0.8075, + "step": 39811 + }, + { + "epoch": 7.24, + "learning_rate": 8.287950491445212e-07, + "loss": 0.3663, + "step": 39812 + }, + { + "epoch": 7.24, + "learning_rate": 8.287404441208592e-07, + "loss": 0.0661, + "step": 39813 + }, + { + "epoch": 7.24, + "learning_rate": 8.28685839097197e-07, + "loss": 0.1753, + "step": 39814 + }, + { + "epoch": 7.24, + "learning_rate": 8.286312340735347e-07, + "loss": 1.0084, + "step": 39815 + }, + { + "epoch": 7.24, + "learning_rate": 8.285766290498727e-07, + "loss": 0.0873, + "step": 39816 + }, + { + "epoch": 7.24, + "learning_rate": 8.285220240262104e-07, + "loss": 2.1332, + "step": 39817 + }, + { + "epoch": 7.24, + "learning_rate": 8.284674190025483e-07, + "loss": 0.0146, + "step": 39818 + }, + { + "epoch": 7.24, + "learning_rate": 8.28412813978886e-07, + "loss": 0.0107, + "step": 39819 + }, + { + "epoch": 7.24, + "learning_rate": 8.283582089552239e-07, + "loss": 2.1786, + "step": 39820 + }, + { + "epoch": 7.24, + "learning_rate": 8.283036039315618e-07, + "loss": 0.3906, + "step": 39821 + }, + { + "epoch": 7.24, + "learning_rate": 8.282489989078995e-07, + "loss": 1.6689, + "step": 39822 + }, + { + "epoch": 7.24, + "learning_rate": 8.281943938842373e-07, + "loss": 0.2056, + "step": 39823 + }, + { + "epoch": 7.24, + "learning_rate": 8.281397888605752e-07, + "loss": 0.0073, + "step": 39824 + }, + { + "epoch": 7.24, + "learning_rate": 8.28085183836913e-07, + "loss": 0.0802, + "step": 39825 + }, + { + "epoch": 7.24, + "learning_rate": 8.280305788132507e-07, + "loss": 0.3508, + "step": 39826 + }, + { + "epoch": 7.24, + "learning_rate": 8.279759737895887e-07, + "loss": 0.2298, + "step": 39827 + }, + { + "epoch": 7.24, + "learning_rate": 8.279213687659265e-07, + "loss": 1.6239, + "step": 39828 + }, + { + "epoch": 7.24, + "learning_rate": 8.278667637422643e-07, + "loss": 0.1855, + "step": 39829 + }, + { + "epoch": 7.24, + "learning_rate": 8.278121587186021e-07, + "loss": 2.3186, + "step": 39830 + }, + { + "epoch": 7.24, + "learning_rate": 8.2775755369494e-07, + "loss": 0.813, + "step": 39831 + }, + { + "epoch": 7.24, + "learning_rate": 8.277029486712778e-07, + "loss": 0.5591, + "step": 39832 + }, + { + "epoch": 7.24, + "learning_rate": 8.276483436476155e-07, + "loss": 0.6066, + "step": 39833 + }, + { + "epoch": 7.24, + "learning_rate": 8.275937386239534e-07, + "loss": 0.1369, + "step": 39834 + }, + { + "epoch": 7.24, + "learning_rate": 8.275391336002913e-07, + "loss": 0.4701, + "step": 39835 + }, + { + "epoch": 7.24, + "learning_rate": 8.27484528576629e-07, + "loss": 0.1984, + "step": 39836 + }, + { + "epoch": 7.24, + "learning_rate": 8.274299235529669e-07, + "loss": 0.6446, + "step": 39837 + }, + { + "epoch": 7.24, + "learning_rate": 8.273753185293048e-07, + "loss": 0.0067, + "step": 39838 + }, + { + "epoch": 7.24, + "learning_rate": 8.273207135056425e-07, + "loss": 0.5485, + "step": 39839 + }, + { + "epoch": 7.24, + "learning_rate": 8.272661084819803e-07, + "loss": 0.0127, + "step": 39840 + }, + { + "epoch": 7.25, + "learning_rate": 8.272115034583182e-07, + "loss": 0.585, + "step": 39841 + }, + { + "epoch": 7.25, + "learning_rate": 8.27156898434656e-07, + "loss": 0.834, + "step": 39842 + }, + { + "epoch": 7.25, + "learning_rate": 8.271022934109938e-07, + "loss": 0.9854, + "step": 39843 + }, + { + "epoch": 7.25, + "learning_rate": 8.270476883873316e-07, + "loss": 0.0763, + "step": 39844 + }, + { + "epoch": 7.25, + "learning_rate": 8.269930833636696e-07, + "loss": 0.0128, + "step": 39845 + }, + { + "epoch": 7.25, + "learning_rate": 8.269384783400073e-07, + "loss": 0.1835, + "step": 39846 + }, + { + "epoch": 7.25, + "learning_rate": 8.268838733163451e-07, + "loss": 0.0111, + "step": 39847 + }, + { + "epoch": 7.25, + "learning_rate": 8.268292682926831e-07, + "loss": 0.0048, + "step": 39848 + }, + { + "epoch": 7.25, + "learning_rate": 8.267746632690208e-07, + "loss": 0.3692, + "step": 39849 + }, + { + "epoch": 7.25, + "learning_rate": 8.267200582453585e-07, + "loss": 0.0074, + "step": 39850 + }, + { + "epoch": 7.25, + "learning_rate": 8.266654532216964e-07, + "loss": 0.0943, + "step": 39851 + }, + { + "epoch": 7.25, + "learning_rate": 8.266108481980343e-07, + "loss": 0.0805, + "step": 39852 + }, + { + "epoch": 7.25, + "learning_rate": 8.26556243174372e-07, + "loss": 0.9825, + "step": 39853 + }, + { + "epoch": 7.25, + "learning_rate": 8.265016381507099e-07, + "loss": 3.8144, + "step": 39854 + }, + { + "epoch": 7.25, + "learning_rate": 8.264470331270478e-07, + "loss": 1.5596, + "step": 39855 + }, + { + "epoch": 7.25, + "learning_rate": 8.263924281033856e-07, + "loss": 0.0066, + "step": 39856 + }, + { + "epoch": 7.25, + "learning_rate": 8.263378230797233e-07, + "loss": 0.7061, + "step": 39857 + }, + { + "epoch": 7.25, + "learning_rate": 8.262832180560611e-07, + "loss": 0.176, + "step": 39858 + }, + { + "epoch": 7.25, + "learning_rate": 8.262286130323991e-07, + "loss": 0.7394, + "step": 39859 + }, + { + "epoch": 7.25, + "learning_rate": 8.261740080087368e-07, + "loss": 0.7701, + "step": 39860 + }, + { + "epoch": 7.25, + "learning_rate": 8.261194029850746e-07, + "loss": 1.0972, + "step": 39861 + }, + { + "epoch": 7.25, + "learning_rate": 8.260647979614126e-07, + "loss": 0.0083, + "step": 39862 + }, + { + "epoch": 7.25, + "learning_rate": 8.260101929377503e-07, + "loss": 0.0601, + "step": 39863 + }, + { + "epoch": 7.25, + "learning_rate": 8.25955587914088e-07, + "loss": 1.6098, + "step": 39864 + }, + { + "epoch": 7.25, + "learning_rate": 8.259009828904259e-07, + "loss": 0.0579, + "step": 39865 + }, + { + "epoch": 7.25, + "learning_rate": 8.258463778667638e-07, + "loss": 0.4866, + "step": 39866 + }, + { + "epoch": 7.25, + "learning_rate": 8.257917728431016e-07, + "loss": 0.5544, + "step": 39867 + }, + { + "epoch": 7.25, + "learning_rate": 8.257371678194394e-07, + "loss": 2.2064, + "step": 39868 + }, + { + "epoch": 7.25, + "learning_rate": 8.256825627957773e-07, + "loss": 0.1093, + "step": 39869 + }, + { + "epoch": 7.25, + "learning_rate": 8.256279577721151e-07, + "loss": 1.6751, + "step": 39870 + }, + { + "epoch": 7.25, + "learning_rate": 8.255733527484529e-07, + "loss": 3.0173, + "step": 39871 + }, + { + "epoch": 7.25, + "learning_rate": 8.255187477247906e-07, + "loss": 0.0117, + "step": 39872 + }, + { + "epoch": 7.25, + "learning_rate": 8.254641427011286e-07, + "loss": 2.001, + "step": 39873 + }, + { + "epoch": 7.25, + "learning_rate": 8.254095376774663e-07, + "loss": 0.0933, + "step": 39874 + }, + { + "epoch": 7.25, + "learning_rate": 8.253549326538042e-07, + "loss": 0.3729, + "step": 39875 + }, + { + "epoch": 7.25, + "learning_rate": 8.253003276301421e-07, + "loss": 0.0086, + "step": 39876 + }, + { + "epoch": 7.25, + "learning_rate": 8.252457226064798e-07, + "loss": 0.1463, + "step": 39877 + }, + { + "epoch": 7.25, + "learning_rate": 8.251911175828177e-07, + "loss": 0.1237, + "step": 39878 + }, + { + "epoch": 7.25, + "learning_rate": 8.251365125591554e-07, + "loss": 0.0191, + "step": 39879 + }, + { + "epoch": 7.25, + "learning_rate": 8.250819075354933e-07, + "loss": 0.6714, + "step": 39880 + }, + { + "epoch": 7.25, + "learning_rate": 8.250273025118311e-07, + "loss": 0.0159, + "step": 39881 + }, + { + "epoch": 7.25, + "learning_rate": 8.249726974881689e-07, + "loss": 0.019, + "step": 39882 + }, + { + "epoch": 7.25, + "learning_rate": 8.249180924645069e-07, + "loss": 0.7322, + "step": 39883 + }, + { + "epoch": 7.25, + "learning_rate": 8.248634874408446e-07, + "loss": 0.7527, + "step": 39884 + }, + { + "epoch": 7.25, + "learning_rate": 8.248088824171824e-07, + "loss": 0.1354, + "step": 39885 + }, + { + "epoch": 7.25, + "learning_rate": 8.247542773935202e-07, + "loss": 0.3156, + "step": 39886 + }, + { + "epoch": 7.25, + "learning_rate": 8.246996723698581e-07, + "loss": 0.3968, + "step": 39887 + }, + { + "epoch": 7.25, + "learning_rate": 8.246450673461958e-07, + "loss": 0.0772, + "step": 39888 + }, + { + "epoch": 7.25, + "learning_rate": 8.245904623225337e-07, + "loss": 0.8502, + "step": 39889 + }, + { + "epoch": 7.25, + "learning_rate": 8.245358572988716e-07, + "loss": 0.1351, + "step": 39890 + }, + { + "epoch": 7.25, + "learning_rate": 8.244812522752093e-07, + "loss": 1.4084, + "step": 39891 + }, + { + "epoch": 7.25, + "learning_rate": 8.244266472515472e-07, + "loss": 0.1493, + "step": 39892 + }, + { + "epoch": 7.25, + "learning_rate": 8.243720422278849e-07, + "loss": 0.004, + "step": 39893 + }, + { + "epoch": 7.25, + "learning_rate": 8.243174372042229e-07, + "loss": 0.5676, + "step": 39894 + }, + { + "epoch": 7.25, + "learning_rate": 8.242628321805607e-07, + "loss": 0.3618, + "step": 39895 + }, + { + "epoch": 7.26, + "learning_rate": 8.242082271568984e-07, + "loss": 2.0829, + "step": 39896 + }, + { + "epoch": 7.26, + "learning_rate": 8.241536221332364e-07, + "loss": 2.1943, + "step": 39897 + }, + { + "epoch": 7.26, + "learning_rate": 8.240990171095741e-07, + "loss": 0.0067, + "step": 39898 + }, + { + "epoch": 7.26, + "learning_rate": 8.240444120859119e-07, + "loss": 0.4235, + "step": 39899 + }, + { + "epoch": 7.26, + "learning_rate": 8.239898070622497e-07, + "loss": 0.4553, + "step": 39900 + }, + { + "epoch": 7.26, + "learning_rate": 8.239352020385876e-07, + "loss": 0.7924, + "step": 39901 + }, + { + "epoch": 7.26, + "learning_rate": 8.238805970149254e-07, + "loss": 0.4851, + "step": 39902 + }, + { + "epoch": 7.26, + "learning_rate": 8.238259919912632e-07, + "loss": 0.19, + "step": 39903 + }, + { + "epoch": 7.26, + "learning_rate": 8.237713869676009e-07, + "loss": 0.0923, + "step": 39904 + }, + { + "epoch": 7.26, + "learning_rate": 8.237167819439389e-07, + "loss": 0.0134, + "step": 39905 + }, + { + "epoch": 7.26, + "learning_rate": 8.236621769202767e-07, + "loss": 2.0822, + "step": 39906 + }, + { + "epoch": 7.26, + "learning_rate": 8.236075718966144e-07, + "loss": 0.0167, + "step": 39907 + }, + { + "epoch": 7.26, + "learning_rate": 8.235529668729524e-07, + "loss": 0.4389, + "step": 39908 + }, + { + "epoch": 7.26, + "learning_rate": 8.234983618492902e-07, + "loss": 1.1907, + "step": 39909 + }, + { + "epoch": 7.26, + "learning_rate": 8.234437568256279e-07, + "loss": 0.2692, + "step": 39910 + }, + { + "epoch": 7.26, + "learning_rate": 8.233891518019658e-07, + "loss": 0.1319, + "step": 39911 + }, + { + "epoch": 7.26, + "learning_rate": 8.233345467783037e-07, + "loss": 0.0132, + "step": 39912 + }, + { + "epoch": 7.26, + "learning_rate": 8.232799417546414e-07, + "loss": 0.3407, + "step": 39913 + }, + { + "epoch": 7.26, + "learning_rate": 8.232253367309792e-07, + "loss": 1.7015, + "step": 39914 + }, + { + "epoch": 7.26, + "learning_rate": 8.231707317073171e-07, + "loss": 0.903, + "step": 39915 + }, + { + "epoch": 7.26, + "learning_rate": 8.23116126683655e-07, + "loss": 0.0579, + "step": 39916 + }, + { + "epoch": 7.26, + "learning_rate": 8.230615216599927e-07, + "loss": 0.0092, + "step": 39917 + }, + { + "epoch": 7.26, + "learning_rate": 8.230069166363305e-07, + "loss": 0.5047, + "step": 39918 + }, + { + "epoch": 7.26, + "learning_rate": 8.229523116126685e-07, + "loss": 0.0111, + "step": 39919 + }, + { + "epoch": 7.26, + "learning_rate": 8.228977065890062e-07, + "loss": 2.4168, + "step": 39920 + }, + { + "epoch": 7.26, + "learning_rate": 8.228431015653439e-07, + "loss": 0.1624, + "step": 39921 + }, + { + "epoch": 7.26, + "learning_rate": 8.227884965416819e-07, + "loss": 1.6914, + "step": 39922 + }, + { + "epoch": 7.26, + "learning_rate": 8.227338915180197e-07, + "loss": 0.0261, + "step": 39923 + }, + { + "epoch": 7.26, + "learning_rate": 8.226792864943574e-07, + "loss": 0.0191, + "step": 39924 + }, + { + "epoch": 7.26, + "learning_rate": 8.226246814706953e-07, + "loss": 2.0204, + "step": 39925 + }, + { + "epoch": 7.26, + "learning_rate": 8.225700764470332e-07, + "loss": 0.0099, + "step": 39926 + }, + { + "epoch": 7.26, + "learning_rate": 8.22515471423371e-07, + "loss": 0.1107, + "step": 39927 + }, + { + "epoch": 7.26, + "learning_rate": 8.224608663997087e-07, + "loss": 0.0047, + "step": 39928 + }, + { + "epoch": 7.26, + "learning_rate": 8.224062613760466e-07, + "loss": 0.9741, + "step": 39929 + }, + { + "epoch": 7.26, + "learning_rate": 8.223516563523845e-07, + "loss": 0.8465, + "step": 39930 + }, + { + "epoch": 7.26, + "learning_rate": 8.222970513287222e-07, + "loss": 0.0552, + "step": 39931 + }, + { + "epoch": 7.26, + "learning_rate": 8.2224244630506e-07, + "loss": 0.1091, + "step": 39932 + }, + { + "epoch": 7.26, + "learning_rate": 8.22187841281398e-07, + "loss": 0.0542, + "step": 39933 + }, + { + "epoch": 7.26, + "learning_rate": 8.221332362577357e-07, + "loss": 0.0069, + "step": 39934 + }, + { + "epoch": 7.26, + "learning_rate": 8.220786312340736e-07, + "loss": 0.3958, + "step": 39935 + }, + { + "epoch": 7.26, + "learning_rate": 8.220240262104115e-07, + "loss": 0.0055, + "step": 39936 + }, + { + "epoch": 7.26, + "learning_rate": 8.219694211867492e-07, + "loss": 0.6317, + "step": 39937 + }, + { + "epoch": 7.26, + "learning_rate": 8.21914816163087e-07, + "loss": 0.0285, + "step": 39938 + }, + { + "epoch": 7.26, + "learning_rate": 8.218602111394248e-07, + "loss": 1.108, + "step": 39939 + }, + { + "epoch": 7.26, + "learning_rate": 8.218056061157627e-07, + "loss": 0.3107, + "step": 39940 + }, + { + "epoch": 7.26, + "learning_rate": 8.217510010921005e-07, + "loss": 0.0951, + "step": 39941 + }, + { + "epoch": 7.26, + "learning_rate": 8.216963960684383e-07, + "loss": 1.6631, + "step": 39942 + }, + { + "epoch": 7.26, + "learning_rate": 8.216417910447763e-07, + "loss": 0.8542, + "step": 39943 + }, + { + "epoch": 7.26, + "learning_rate": 8.21587186021114e-07, + "loss": 1.707, + "step": 39944 + }, + { + "epoch": 7.26, + "learning_rate": 8.215325809974517e-07, + "loss": 0.3083, + "step": 39945 + }, + { + "epoch": 7.26, + "learning_rate": 8.214779759737896e-07, + "loss": 1.592, + "step": 39946 + }, + { + "epoch": 7.26, + "learning_rate": 8.214233709501275e-07, + "loss": 0.2578, + "step": 39947 + }, + { + "epoch": 7.26, + "learning_rate": 8.213687659264652e-07, + "loss": 2.1698, + "step": 39948 + }, + { + "epoch": 7.26, + "learning_rate": 8.213141609028031e-07, + "loss": 1.4226, + "step": 39949 + }, + { + "epoch": 7.26, + "learning_rate": 8.21259555879141e-07, + "loss": 0.0234, + "step": 39950 + }, + { + "epoch": 7.27, + "learning_rate": 8.212049508554787e-07, + "loss": 3.6051, + "step": 39951 + }, + { + "epoch": 7.27, + "learning_rate": 8.211503458318166e-07, + "loss": 0.0308, + "step": 39952 + }, + { + "epoch": 7.27, + "learning_rate": 8.210957408081543e-07, + "loss": 0.0171, + "step": 39953 + }, + { + "epoch": 7.27, + "learning_rate": 8.210411357844923e-07, + "loss": 0.4335, + "step": 39954 + }, + { + "epoch": 7.27, + "learning_rate": 8.2098653076083e-07, + "loss": 0.1583, + "step": 39955 + }, + { + "epoch": 7.27, + "learning_rate": 8.209319257371678e-07, + "loss": 1.3397, + "step": 39956 + }, + { + "epoch": 7.27, + "learning_rate": 8.208773207135058e-07, + "loss": 0.2395, + "step": 39957 + }, + { + "epoch": 7.27, + "learning_rate": 8.208227156898435e-07, + "loss": 1.6541, + "step": 39958 + }, + { + "epoch": 7.27, + "learning_rate": 8.207681106661813e-07, + "loss": 0.0082, + "step": 39959 + }, + { + "epoch": 7.27, + "learning_rate": 8.207135056425191e-07, + "loss": 0.0045, + "step": 39960 + }, + { + "epoch": 7.27, + "learning_rate": 8.20658900618857e-07, + "loss": 0.2189, + "step": 39961 + }, + { + "epoch": 7.27, + "learning_rate": 8.206042955951947e-07, + "loss": 0.0086, + "step": 39962 + }, + { + "epoch": 7.27, + "learning_rate": 8.205496905715326e-07, + "loss": 1.0207, + "step": 39963 + }, + { + "epoch": 7.27, + "learning_rate": 8.204950855478705e-07, + "loss": 0.1298, + "step": 39964 + }, + { + "epoch": 7.27, + "learning_rate": 8.204404805242083e-07, + "loss": 0.1665, + "step": 39965 + }, + { + "epoch": 7.27, + "learning_rate": 8.203858755005461e-07, + "loss": 0.3505, + "step": 39966 + }, + { + "epoch": 7.27, + "learning_rate": 8.203312704768838e-07, + "loss": 0.0693, + "step": 39967 + }, + { + "epoch": 7.27, + "learning_rate": 8.202766654532218e-07, + "loss": 1.3769, + "step": 39968 + }, + { + "epoch": 7.27, + "learning_rate": 8.202220604295595e-07, + "loss": 0.1572, + "step": 39969 + }, + { + "epoch": 7.27, + "learning_rate": 8.201674554058973e-07, + "loss": 0.4731, + "step": 39970 + }, + { + "epoch": 7.27, + "learning_rate": 8.201128503822353e-07, + "loss": 0.56, + "step": 39971 + }, + { + "epoch": 7.27, + "learning_rate": 8.20058245358573e-07, + "loss": 0.2974, + "step": 39972 + }, + { + "epoch": 7.27, + "learning_rate": 8.200036403349109e-07, + "loss": 1.0186, + "step": 39973 + }, + { + "epoch": 7.27, + "learning_rate": 8.199490353112486e-07, + "loss": 0.1055, + "step": 39974 + }, + { + "epoch": 7.27, + "learning_rate": 8.198944302875865e-07, + "loss": 0.1497, + "step": 39975 + }, + { + "epoch": 7.27, + "learning_rate": 8.198398252639244e-07, + "loss": 0.8227, + "step": 39976 + }, + { + "epoch": 7.27, + "learning_rate": 8.197852202402621e-07, + "loss": 0.6866, + "step": 39977 + }, + { + "epoch": 7.27, + "learning_rate": 8.197306152165998e-07, + "loss": 1.1748, + "step": 39978 + }, + { + "epoch": 7.27, + "learning_rate": 8.196760101929378e-07, + "loss": 0.1301, + "step": 39979 + }, + { + "epoch": 7.27, + "learning_rate": 8.196214051692756e-07, + "loss": 0.0198, + "step": 39980 + }, + { + "epoch": 7.27, + "learning_rate": 8.195668001456133e-07, + "loss": 2.2389, + "step": 39981 + }, + { + "epoch": 7.27, + "learning_rate": 8.195121951219513e-07, + "loss": 0.1902, + "step": 39982 + }, + { + "epoch": 7.27, + "learning_rate": 8.194575900982891e-07, + "loss": 0.4417, + "step": 39983 + }, + { + "epoch": 7.27, + "learning_rate": 8.194029850746269e-07, + "loss": 0.0544, + "step": 39984 + }, + { + "epoch": 7.27, + "learning_rate": 8.193483800509646e-07, + "loss": 1.7024, + "step": 39985 + }, + { + "epoch": 7.27, + "learning_rate": 8.192937750273025e-07, + "loss": 1.8735, + "step": 39986 + }, + { + "epoch": 7.27, + "learning_rate": 8.192391700036404e-07, + "loss": 1.4543, + "step": 39987 + }, + { + "epoch": 7.27, + "learning_rate": 8.191845649799781e-07, + "loss": 0.1762, + "step": 39988 + }, + { + "epoch": 7.27, + "learning_rate": 8.19129959956316e-07, + "loss": 0.2958, + "step": 39989 + }, + { + "epoch": 7.27, + "learning_rate": 8.190753549326539e-07, + "loss": 0.0163, + "step": 39990 + }, + { + "epoch": 7.27, + "learning_rate": 8.190207499089916e-07, + "loss": 0.009, + "step": 39991 + }, + { + "epoch": 7.27, + "learning_rate": 8.189661448853293e-07, + "loss": 0.6035, + "step": 39992 + }, + { + "epoch": 7.27, + "learning_rate": 8.189115398616673e-07, + "loss": 0.0489, + "step": 39993 + }, + { + "epoch": 7.27, + "learning_rate": 8.188569348380051e-07, + "loss": 1.1183, + "step": 39994 + }, + { + "epoch": 7.27, + "learning_rate": 8.188023298143429e-07, + "loss": 0.0151, + "step": 39995 + }, + { + "epoch": 7.27, + "learning_rate": 8.187477247906808e-07, + "loss": 0.0536, + "step": 39996 + }, + { + "epoch": 7.27, + "learning_rate": 8.186931197670186e-07, + "loss": 0.3675, + "step": 39997 + }, + { + "epoch": 7.27, + "learning_rate": 8.186385147433564e-07, + "loss": 0.0065, + "step": 39998 + }, + { + "epoch": 7.27, + "learning_rate": 8.185839097196942e-07, + "loss": 2.7602, + "step": 39999 + }, + { + "epoch": 7.27, + "learning_rate": 8.185293046960321e-07, + "loss": 0.8372, + "step": 40000 + }, + { + "epoch": 7.27, + "learning_rate": 8.184746996723699e-07, + "loss": 0.1365, + "step": 40001 + }, + { + "epoch": 7.27, + "learning_rate": 8.184200946487076e-07, + "loss": 2.1051, + "step": 40002 + }, + { + "epoch": 7.27, + "learning_rate": 8.183654896250456e-07, + "loss": 0.1241, + "step": 40003 + }, + { + "epoch": 7.27, + "learning_rate": 8.183108846013834e-07, + "loss": 0.0651, + "step": 40004 + }, + { + "epoch": 7.27, + "learning_rate": 8.182562795777211e-07, + "loss": 0.3216, + "step": 40005 + }, + { + "epoch": 7.28, + "learning_rate": 8.18201674554059e-07, + "loss": 0.0287, + "step": 40006 + }, + { + "epoch": 7.28, + "learning_rate": 8.181470695303969e-07, + "loss": 0.2223, + "step": 40007 + }, + { + "epoch": 7.28, + "learning_rate": 8.180924645067346e-07, + "loss": 0.1889, + "step": 40008 + }, + { + "epoch": 7.28, + "learning_rate": 8.180378594830724e-07, + "loss": 0.2095, + "step": 40009 + }, + { + "epoch": 7.28, + "learning_rate": 8.179832544594103e-07, + "loss": 0.9255, + "step": 40010 + }, + { + "epoch": 7.28, + "learning_rate": 8.179286494357481e-07, + "loss": 0.8838, + "step": 40011 + }, + { + "epoch": 7.28, + "learning_rate": 8.178740444120859e-07, + "loss": 0.8415, + "step": 40012 + }, + { + "epoch": 7.28, + "learning_rate": 8.178194393884237e-07, + "loss": 0.9865, + "step": 40013 + }, + { + "epoch": 7.28, + "learning_rate": 8.177648343647617e-07, + "loss": 0.0093, + "step": 40014 + }, + { + "epoch": 7.28, + "learning_rate": 8.177102293410994e-07, + "loss": 0.4377, + "step": 40015 + }, + { + "epoch": 7.28, + "learning_rate": 8.176556243174372e-07, + "loss": 0.0061, + "step": 40016 + }, + { + "epoch": 7.28, + "learning_rate": 8.176010192937752e-07, + "loss": 0.2701, + "step": 40017 + }, + { + "epoch": 7.28, + "learning_rate": 8.175464142701129e-07, + "loss": 1.4302, + "step": 40018 + }, + { + "epoch": 7.28, + "learning_rate": 8.174918092464506e-07, + "loss": 0.0655, + "step": 40019 + }, + { + "epoch": 7.28, + "learning_rate": 8.174372042227885e-07, + "loss": 0.9829, + "step": 40020 + }, + { + "epoch": 7.28, + "learning_rate": 8.173825991991264e-07, + "loss": 1.0164, + "step": 40021 + }, + { + "epoch": 7.28, + "learning_rate": 8.173279941754641e-07, + "loss": 0.0653, + "step": 40022 + }, + { + "epoch": 7.28, + "learning_rate": 8.17273389151802e-07, + "loss": 0.016, + "step": 40023 + }, + { + "epoch": 7.28, + "learning_rate": 8.172187841281399e-07, + "loss": 0.0103, + "step": 40024 + }, + { + "epoch": 7.28, + "learning_rate": 8.171641791044777e-07, + "loss": 0.6428, + "step": 40025 + }, + { + "epoch": 7.28, + "learning_rate": 8.171095740808154e-07, + "loss": 0.3922, + "step": 40026 + }, + { + "epoch": 7.28, + "learning_rate": 8.170549690571532e-07, + "loss": 0.1328, + "step": 40027 + }, + { + "epoch": 7.28, + "learning_rate": 8.170003640334912e-07, + "loss": 0.0225, + "step": 40028 + }, + { + "epoch": 7.28, + "learning_rate": 8.169457590098289e-07, + "loss": 0.4852, + "step": 40029 + }, + { + "epoch": 7.28, + "learning_rate": 8.168911539861667e-07, + "loss": 0.0184, + "step": 40030 + }, + { + "epoch": 7.28, + "learning_rate": 8.168365489625047e-07, + "loss": 0.5992, + "step": 40031 + }, + { + "epoch": 7.28, + "learning_rate": 8.167819439388424e-07, + "loss": 0.6471, + "step": 40032 + }, + { + "epoch": 7.28, + "learning_rate": 8.167273389151801e-07, + "loss": 0.13, + "step": 40033 + }, + { + "epoch": 7.28, + "learning_rate": 8.16672733891518e-07, + "loss": 0.1008, + "step": 40034 + }, + { + "epoch": 7.28, + "learning_rate": 8.166181288678559e-07, + "loss": 0.4699, + "step": 40035 + }, + { + "epoch": 7.28, + "learning_rate": 8.165635238441937e-07, + "loss": 0.7796, + "step": 40036 + }, + { + "epoch": 7.28, + "learning_rate": 8.165089188205315e-07, + "loss": 0.0104, + "step": 40037 + }, + { + "epoch": 7.28, + "learning_rate": 8.164543137968694e-07, + "loss": 0.0097, + "step": 40038 + }, + { + "epoch": 7.28, + "learning_rate": 8.163997087732072e-07, + "loss": 0.5851, + "step": 40039 + }, + { + "epoch": 7.28, + "learning_rate": 8.16345103749545e-07, + "loss": 0.2922, + "step": 40040 + }, + { + "epoch": 7.28, + "learning_rate": 8.162904987258827e-07, + "loss": 0.0214, + "step": 40041 + }, + { + "epoch": 7.28, + "learning_rate": 8.162358937022207e-07, + "loss": 0.0046, + "step": 40042 + }, + { + "epoch": 7.28, + "learning_rate": 8.161812886785584e-07, + "loss": 0.1169, + "step": 40043 + }, + { + "epoch": 7.28, + "learning_rate": 8.161266836548963e-07, + "loss": 0.3261, + "step": 40044 + }, + { + "epoch": 7.28, + "learning_rate": 8.160720786312342e-07, + "loss": 1.9542, + "step": 40045 + }, + { + "epoch": 7.28, + "learning_rate": 8.160174736075719e-07, + "loss": 0.3669, + "step": 40046 + }, + { + "epoch": 7.28, + "learning_rate": 8.159628685839098e-07, + "loss": 0.5335, + "step": 40047 + }, + { + "epoch": 7.28, + "learning_rate": 8.159082635602475e-07, + "loss": 1.7692, + "step": 40048 + }, + { + "epoch": 7.28, + "learning_rate": 8.158536585365854e-07, + "loss": 1.0837, + "step": 40049 + }, + { + "epoch": 7.28, + "learning_rate": 8.157990535129232e-07, + "loss": 1.151, + "step": 40050 + }, + { + "epoch": 7.28, + "learning_rate": 8.15744448489261e-07, + "loss": 0.053, + "step": 40051 + }, + { + "epoch": 7.28, + "learning_rate": 8.15689843465599e-07, + "loss": 0.0042, + "step": 40052 + }, + { + "epoch": 7.28, + "learning_rate": 8.156352384419367e-07, + "loss": 0.0046, + "step": 40053 + }, + { + "epoch": 7.28, + "learning_rate": 8.155806334182745e-07, + "loss": 1.0728, + "step": 40054 + }, + { + "epoch": 7.28, + "learning_rate": 8.155260283946123e-07, + "loss": 0.3311, + "step": 40055 + }, + { + "epoch": 7.28, + "learning_rate": 8.154714233709502e-07, + "loss": 0.5872, + "step": 40056 + }, + { + "epoch": 7.28, + "learning_rate": 8.154168183472879e-07, + "loss": 0.1039, + "step": 40057 + }, + { + "epoch": 7.28, + "learning_rate": 8.153622133236258e-07, + "loss": 0.1765, + "step": 40058 + }, + { + "epoch": 7.28, + "learning_rate": 8.153076082999635e-07, + "loss": 0.0753, + "step": 40059 + }, + { + "epoch": 7.28, + "learning_rate": 8.152530032763014e-07, + "loss": 0.2312, + "step": 40060 + }, + { + "epoch": 7.29, + "learning_rate": 8.151983982526393e-07, + "loss": 0.5896, + "step": 40061 + }, + { + "epoch": 7.29, + "learning_rate": 8.15143793228977e-07, + "loss": 0.1572, + "step": 40062 + }, + { + "epoch": 7.29, + "learning_rate": 8.15089188205315e-07, + "loss": 0.855, + "step": 40063 + }, + { + "epoch": 7.29, + "learning_rate": 8.150345831816528e-07, + "loss": 1.5398, + "step": 40064 + }, + { + "epoch": 7.29, + "learning_rate": 8.149799781579905e-07, + "loss": 0.0343, + "step": 40065 + }, + { + "epoch": 7.29, + "learning_rate": 8.149253731343283e-07, + "loss": 0.1283, + "step": 40066 + }, + { + "epoch": 7.29, + "learning_rate": 8.148707681106662e-07, + "loss": 0.0152, + "step": 40067 + }, + { + "epoch": 7.29, + "learning_rate": 8.14816163087004e-07, + "loss": 0.1145, + "step": 40068 + }, + { + "epoch": 7.29, + "learning_rate": 8.147615580633418e-07, + "loss": 0.1781, + "step": 40069 + }, + { + "epoch": 7.29, + "learning_rate": 8.147069530396797e-07, + "loss": 0.0955, + "step": 40070 + }, + { + "epoch": 7.29, + "learning_rate": 8.146523480160176e-07, + "loss": 0.0684, + "step": 40071 + }, + { + "epoch": 7.29, + "learning_rate": 8.145977429923553e-07, + "loss": 0.5926, + "step": 40072 + }, + { + "epoch": 7.29, + "learning_rate": 8.14543137968693e-07, + "loss": 2.2277, + "step": 40073 + }, + { + "epoch": 7.29, + "learning_rate": 8.14488532945031e-07, + "loss": 0.0324, + "step": 40074 + }, + { + "epoch": 7.29, + "learning_rate": 8.144339279213688e-07, + "loss": 0.0063, + "step": 40075 + }, + { + "epoch": 7.29, + "learning_rate": 8.143793228977065e-07, + "loss": 0.739, + "step": 40076 + }, + { + "epoch": 7.29, + "learning_rate": 8.143247178740445e-07, + "loss": 0.004, + "step": 40077 + }, + { + "epoch": 7.29, + "learning_rate": 8.142701128503823e-07, + "loss": 0.0051, + "step": 40078 + }, + { + "epoch": 7.29, + "learning_rate": 8.1421550782672e-07, + "loss": 1.2038, + "step": 40079 + }, + { + "epoch": 7.29, + "learning_rate": 8.141609028030579e-07, + "loss": 0.1563, + "step": 40080 + }, + { + "epoch": 7.29, + "learning_rate": 8.141062977793958e-07, + "loss": 0.0641, + "step": 40081 + }, + { + "epoch": 7.29, + "learning_rate": 8.140516927557336e-07, + "loss": 0.1386, + "step": 40082 + }, + { + "epoch": 7.29, + "learning_rate": 8.139970877320713e-07, + "loss": 0.0077, + "step": 40083 + }, + { + "epoch": 7.29, + "learning_rate": 8.139424827084092e-07, + "loss": 0.0912, + "step": 40084 + }, + { + "epoch": 7.29, + "learning_rate": 8.138878776847471e-07, + "loss": 0.5181, + "step": 40085 + }, + { + "epoch": 7.29, + "learning_rate": 8.138332726610848e-07, + "loss": 0.1632, + "step": 40086 + }, + { + "epoch": 7.29, + "learning_rate": 8.137786676374226e-07, + "loss": 0.0148, + "step": 40087 + }, + { + "epoch": 7.29, + "learning_rate": 8.137240626137606e-07, + "loss": 4.463, + "step": 40088 + }, + { + "epoch": 7.29, + "learning_rate": 8.136694575900983e-07, + "loss": 0.0259, + "step": 40089 + }, + { + "epoch": 7.29, + "learning_rate": 8.13614852566436e-07, + "loss": 0.6603, + "step": 40090 + }, + { + "epoch": 7.29, + "learning_rate": 8.13560247542774e-07, + "loss": 0.033, + "step": 40091 + }, + { + "epoch": 7.29, + "learning_rate": 8.135056425191118e-07, + "loss": 0.9636, + "step": 40092 + }, + { + "epoch": 7.29, + "learning_rate": 8.134510374954496e-07, + "loss": 0.0562, + "step": 40093 + }, + { + "epoch": 7.29, + "learning_rate": 8.133964324717874e-07, + "loss": 0.0083, + "step": 40094 + }, + { + "epoch": 7.29, + "learning_rate": 8.133418274481253e-07, + "loss": 0.0658, + "step": 40095 + }, + { + "epoch": 7.29, + "learning_rate": 8.132872224244631e-07, + "loss": 1.0623, + "step": 40096 + }, + { + "epoch": 7.29, + "learning_rate": 8.132326174008008e-07, + "loss": 0.0041, + "step": 40097 + }, + { + "epoch": 7.29, + "learning_rate": 8.131780123771387e-07, + "loss": 0.3451, + "step": 40098 + }, + { + "epoch": 7.29, + "learning_rate": 8.131234073534766e-07, + "loss": 0.0913, + "step": 40099 + }, + { + "epoch": 7.29, + "learning_rate": 8.130688023298143e-07, + "loss": 0.0611, + "step": 40100 + }, + { + "epoch": 7.29, + "learning_rate": 8.130141973061521e-07, + "loss": 0.1552, + "step": 40101 + }, + { + "epoch": 7.29, + "learning_rate": 8.129595922824901e-07, + "loss": 0.8939, + "step": 40102 + }, + { + "epoch": 7.29, + "learning_rate": 8.129049872588278e-07, + "loss": 0.4679, + "step": 40103 + }, + { + "epoch": 7.29, + "learning_rate": 8.128503822351657e-07, + "loss": 3.178, + "step": 40104 + }, + { + "epoch": 7.29, + "learning_rate": 8.127957772115036e-07, + "loss": 0.0729, + "step": 40105 + }, + { + "epoch": 7.29, + "learning_rate": 8.127411721878413e-07, + "loss": 0.3574, + "step": 40106 + }, + { + "epoch": 7.29, + "learning_rate": 8.126865671641791e-07, + "loss": 0.1834, + "step": 40107 + }, + { + "epoch": 7.29, + "learning_rate": 8.126319621405169e-07, + "loss": 0.3758, + "step": 40108 + }, + { + "epoch": 7.29, + "learning_rate": 8.125773571168548e-07, + "loss": 0.046, + "step": 40109 + }, + { + "epoch": 7.29, + "learning_rate": 8.125227520931926e-07, + "loss": 0.5796, + "step": 40110 + }, + { + "epoch": 7.29, + "learning_rate": 8.124681470695304e-07, + "loss": 0.0076, + "step": 40111 + }, + { + "epoch": 7.29, + "learning_rate": 8.124135420458684e-07, + "loss": 0.4723, + "step": 40112 + }, + { + "epoch": 7.29, + "learning_rate": 8.123589370222061e-07, + "loss": 1.3849, + "step": 40113 + }, + { + "epoch": 7.29, + "learning_rate": 8.123043319985438e-07, + "loss": 1.6724, + "step": 40114 + }, + { + "epoch": 7.29, + "learning_rate": 8.122497269748817e-07, + "loss": 0.1047, + "step": 40115 + }, + { + "epoch": 7.3, + "learning_rate": 8.121951219512196e-07, + "loss": 1.4398, + "step": 40116 + }, + { + "epoch": 7.3, + "learning_rate": 8.121405169275573e-07, + "loss": 2.5332, + "step": 40117 + }, + { + "epoch": 7.3, + "learning_rate": 8.120859119038952e-07, + "loss": 0.4383, + "step": 40118 + }, + { + "epoch": 7.3, + "learning_rate": 8.120313068802331e-07, + "loss": 0.5387, + "step": 40119 + }, + { + "epoch": 7.3, + "learning_rate": 8.119767018565708e-07, + "loss": 0.2429, + "step": 40120 + }, + { + "epoch": 7.3, + "learning_rate": 8.119220968329087e-07, + "loss": 1.0769, + "step": 40121 + }, + { + "epoch": 7.3, + "learning_rate": 8.118674918092464e-07, + "loss": 0.2882, + "step": 40122 + }, + { + "epoch": 7.3, + "learning_rate": 8.118128867855844e-07, + "loss": 1.6461, + "step": 40123 + }, + { + "epoch": 7.3, + "learning_rate": 8.117582817619221e-07, + "loss": 0.2313, + "step": 40124 + }, + { + "epoch": 7.3, + "learning_rate": 8.117036767382599e-07, + "loss": 1.5231, + "step": 40125 + }, + { + "epoch": 7.3, + "learning_rate": 8.116490717145979e-07, + "loss": 0.4697, + "step": 40126 + }, + { + "epoch": 7.3, + "learning_rate": 8.115944666909356e-07, + "loss": 0.1269, + "step": 40127 + }, + { + "epoch": 7.3, + "learning_rate": 8.115398616672734e-07, + "loss": 0.0064, + "step": 40128 + }, + { + "epoch": 7.3, + "learning_rate": 8.114852566436112e-07, + "loss": 1.5972, + "step": 40129 + }, + { + "epoch": 7.3, + "learning_rate": 8.114306516199491e-07, + "loss": 0.4606, + "step": 40130 + }, + { + "epoch": 7.3, + "learning_rate": 8.113760465962868e-07, + "loss": 0.0049, + "step": 40131 + }, + { + "epoch": 7.3, + "learning_rate": 8.113214415726247e-07, + "loss": 0.2289, + "step": 40132 + }, + { + "epoch": 7.3, + "learning_rate": 8.112668365489624e-07, + "loss": 0.4813, + "step": 40133 + }, + { + "epoch": 7.3, + "learning_rate": 8.112122315253004e-07, + "loss": 0.0616, + "step": 40134 + }, + { + "epoch": 7.3, + "learning_rate": 8.111576265016382e-07, + "loss": 0.553, + "step": 40135 + }, + { + "epoch": 7.3, + "learning_rate": 8.111030214779759e-07, + "loss": 0.3461, + "step": 40136 + }, + { + "epoch": 7.3, + "learning_rate": 8.110484164543139e-07, + "loss": 0.0803, + "step": 40137 + }, + { + "epoch": 7.3, + "learning_rate": 8.109938114306516e-07, + "loss": 0.0041, + "step": 40138 + }, + { + "epoch": 7.3, + "learning_rate": 8.109392064069894e-07, + "loss": 0.0093, + "step": 40139 + }, + { + "epoch": 7.3, + "learning_rate": 8.108846013833272e-07, + "loss": 0.0053, + "step": 40140 + }, + { + "epoch": 7.3, + "learning_rate": 8.108299963596651e-07, + "loss": 0.1324, + "step": 40141 + }, + { + "epoch": 7.3, + "learning_rate": 8.10775391336003e-07, + "loss": 1.2716, + "step": 40142 + }, + { + "epoch": 7.3, + "learning_rate": 8.107207863123407e-07, + "loss": 0.4231, + "step": 40143 + }, + { + "epoch": 7.3, + "learning_rate": 8.106661812886786e-07, + "loss": 0.0807, + "step": 40144 + }, + { + "epoch": 7.3, + "learning_rate": 8.106115762650165e-07, + "loss": 0.1516, + "step": 40145 + }, + { + "epoch": 7.3, + "learning_rate": 8.105569712413542e-07, + "loss": 0.3362, + "step": 40146 + }, + { + "epoch": 7.3, + "learning_rate": 8.105023662176919e-07, + "loss": 0.8032, + "step": 40147 + }, + { + "epoch": 7.3, + "learning_rate": 8.104477611940299e-07, + "loss": 0.203, + "step": 40148 + }, + { + "epoch": 7.3, + "learning_rate": 8.103931561703677e-07, + "loss": 0.2382, + "step": 40149 + }, + { + "epoch": 7.3, + "learning_rate": 8.103385511467054e-07, + "loss": 0.2026, + "step": 40150 + }, + { + "epoch": 7.3, + "learning_rate": 8.102839461230434e-07, + "loss": 0.39, + "step": 40151 + }, + { + "epoch": 7.3, + "learning_rate": 8.102293410993812e-07, + "loss": 1.5089, + "step": 40152 + }, + { + "epoch": 7.3, + "learning_rate": 8.10174736075719e-07, + "loss": 1.0503, + "step": 40153 + }, + { + "epoch": 7.3, + "learning_rate": 8.101201310520567e-07, + "loss": 0.1651, + "step": 40154 + }, + { + "epoch": 7.3, + "learning_rate": 8.100655260283946e-07, + "loss": 0.3097, + "step": 40155 + }, + { + "epoch": 7.3, + "learning_rate": 8.100109210047325e-07, + "loss": 0.0387, + "step": 40156 + }, + { + "epoch": 7.3, + "learning_rate": 8.099563159810702e-07, + "loss": 0.3103, + "step": 40157 + }, + { + "epoch": 7.3, + "learning_rate": 8.099017109574081e-07, + "loss": 0.0049, + "step": 40158 + }, + { + "epoch": 7.3, + "learning_rate": 8.09847105933746e-07, + "loss": 0.6423, + "step": 40159 + }, + { + "epoch": 7.3, + "learning_rate": 8.097925009100837e-07, + "loss": 0.0055, + "step": 40160 + }, + { + "epoch": 7.3, + "learning_rate": 8.097378958864214e-07, + "loss": 0.0129, + "step": 40161 + }, + { + "epoch": 7.3, + "learning_rate": 8.096832908627594e-07, + "loss": 0.1216, + "step": 40162 + }, + { + "epoch": 7.3, + "learning_rate": 8.096286858390972e-07, + "loss": 0.4157, + "step": 40163 + }, + { + "epoch": 7.3, + "learning_rate": 8.09574080815435e-07, + "loss": 0.0379, + "step": 40164 + }, + { + "epoch": 7.3, + "learning_rate": 8.095194757917729e-07, + "loss": 0.0177, + "step": 40165 + }, + { + "epoch": 7.3, + "learning_rate": 8.094648707681107e-07, + "loss": 0.0027, + "step": 40166 + }, + { + "epoch": 7.3, + "learning_rate": 8.094102657444485e-07, + "loss": 0.1444, + "step": 40167 + }, + { + "epoch": 7.3, + "learning_rate": 8.093556607207863e-07, + "loss": 0.9591, + "step": 40168 + }, + { + "epoch": 7.3, + "learning_rate": 8.093010556971243e-07, + "loss": 0.0275, + "step": 40169 + }, + { + "epoch": 7.3, + "learning_rate": 8.09246450673462e-07, + "loss": 0.3871, + "step": 40170 + }, + { + "epoch": 7.31, + "learning_rate": 8.091918456497997e-07, + "loss": 1.4657, + "step": 40171 + }, + { + "epoch": 7.31, + "learning_rate": 8.091372406261377e-07, + "loss": 2.9411, + "step": 40172 + }, + { + "epoch": 7.31, + "learning_rate": 8.090826356024755e-07, + "loss": 0.0485, + "step": 40173 + }, + { + "epoch": 7.31, + "learning_rate": 8.090280305788132e-07, + "loss": 0.0121, + "step": 40174 + }, + { + "epoch": 7.31, + "learning_rate": 8.089734255551511e-07, + "loss": 0.4041, + "step": 40175 + }, + { + "epoch": 7.31, + "learning_rate": 8.08918820531489e-07, + "loss": 0.3114, + "step": 40176 + }, + { + "epoch": 7.31, + "learning_rate": 8.088642155078267e-07, + "loss": 0.3972, + "step": 40177 + }, + { + "epoch": 7.31, + "learning_rate": 8.088096104841645e-07, + "loss": 0.0852, + "step": 40178 + }, + { + "epoch": 7.31, + "learning_rate": 8.087550054605024e-07, + "loss": 0.0773, + "step": 40179 + }, + { + "epoch": 7.31, + "learning_rate": 8.087004004368403e-07, + "loss": 0.1181, + "step": 40180 + }, + { + "epoch": 7.31, + "learning_rate": 8.08645795413178e-07, + "loss": 0.2559, + "step": 40181 + }, + { + "epoch": 7.31, + "learning_rate": 8.085911903895158e-07, + "loss": 0.0947, + "step": 40182 + }, + { + "epoch": 7.31, + "learning_rate": 8.085365853658538e-07, + "loss": 0.003, + "step": 40183 + }, + { + "epoch": 7.31, + "learning_rate": 8.084819803421915e-07, + "loss": 0.0106, + "step": 40184 + }, + { + "epoch": 7.31, + "learning_rate": 8.084273753185293e-07, + "loss": 0.837, + "step": 40185 + }, + { + "epoch": 7.31, + "learning_rate": 8.083727702948673e-07, + "loss": 0.2697, + "step": 40186 + }, + { + "epoch": 7.31, + "learning_rate": 8.08318165271205e-07, + "loss": 0.9901, + "step": 40187 + }, + { + "epoch": 7.31, + "learning_rate": 8.082635602475427e-07, + "loss": 0.0357, + "step": 40188 + }, + { + "epoch": 7.31, + "learning_rate": 8.082089552238806e-07, + "loss": 0.0088, + "step": 40189 + }, + { + "epoch": 7.31, + "learning_rate": 8.081543502002185e-07, + "loss": 0.1588, + "step": 40190 + }, + { + "epoch": 7.31, + "learning_rate": 8.080997451765563e-07, + "loss": 0.8195, + "step": 40191 + }, + { + "epoch": 7.31, + "learning_rate": 8.080451401528941e-07, + "loss": 0.0043, + "step": 40192 + }, + { + "epoch": 7.31, + "learning_rate": 8.07990535129232e-07, + "loss": 0.0512, + "step": 40193 + }, + { + "epoch": 7.31, + "learning_rate": 8.079359301055698e-07, + "loss": 0.5129, + "step": 40194 + }, + { + "epoch": 7.31, + "learning_rate": 8.078813250819075e-07, + "loss": 1.3406, + "step": 40195 + }, + { + "epoch": 7.31, + "learning_rate": 8.078267200582453e-07, + "loss": 0.0292, + "step": 40196 + }, + { + "epoch": 7.31, + "learning_rate": 8.077721150345833e-07, + "loss": 0.1435, + "step": 40197 + }, + { + "epoch": 7.31, + "learning_rate": 8.07717510010921e-07, + "loss": 0.0413, + "step": 40198 + }, + { + "epoch": 7.31, + "learning_rate": 8.076629049872588e-07, + "loss": 0.496, + "step": 40199 + }, + { + "epoch": 7.31, + "learning_rate": 8.076082999635968e-07, + "loss": 0.7679, + "step": 40200 + }, + { + "epoch": 7.31, + "learning_rate": 8.075536949399345e-07, + "loss": 0.3867, + "step": 40201 + }, + { + "epoch": 7.31, + "learning_rate": 8.074990899162723e-07, + "loss": 3.3788, + "step": 40202 + }, + { + "epoch": 7.31, + "learning_rate": 8.074444848926101e-07, + "loss": 0.0082, + "step": 40203 + }, + { + "epoch": 7.31, + "learning_rate": 8.07389879868948e-07, + "loss": 0.256, + "step": 40204 + }, + { + "epoch": 7.31, + "learning_rate": 8.073352748452858e-07, + "loss": 0.1143, + "step": 40205 + }, + { + "epoch": 7.31, + "learning_rate": 8.072806698216236e-07, + "loss": 0.0558, + "step": 40206 + }, + { + "epoch": 7.31, + "learning_rate": 8.072260647979615e-07, + "loss": 0.3015, + "step": 40207 + }, + { + "epoch": 7.31, + "learning_rate": 8.071714597742993e-07, + "loss": 1.0116, + "step": 40208 + }, + { + "epoch": 7.31, + "learning_rate": 8.071168547506371e-07, + "loss": 0.0174, + "step": 40209 + }, + { + "epoch": 7.31, + "learning_rate": 8.070622497269748e-07, + "loss": 1.1928, + "step": 40210 + }, + { + "epoch": 7.31, + "learning_rate": 8.070076447033128e-07, + "loss": 1.5753, + "step": 40211 + }, + { + "epoch": 7.31, + "learning_rate": 8.069530396796505e-07, + "loss": 0.0032, + "step": 40212 + }, + { + "epoch": 7.31, + "learning_rate": 8.068984346559884e-07, + "loss": 0.112, + "step": 40213 + }, + { + "epoch": 7.31, + "learning_rate": 8.068438296323261e-07, + "loss": 0.0172, + "step": 40214 + }, + { + "epoch": 7.31, + "learning_rate": 8.06789224608664e-07, + "loss": 0.0136, + "step": 40215 + }, + { + "epoch": 7.31, + "learning_rate": 8.067346195850019e-07, + "loss": 0.3101, + "step": 40216 + }, + { + "epoch": 7.31, + "learning_rate": 8.066800145613396e-07, + "loss": 0.0527, + "step": 40217 + }, + { + "epoch": 7.31, + "learning_rate": 8.066254095376775e-07, + "loss": 1.8256, + "step": 40218 + }, + { + "epoch": 7.31, + "learning_rate": 8.065708045140153e-07, + "loss": 1.9749, + "step": 40219 + }, + { + "epoch": 7.31, + "learning_rate": 8.065161994903531e-07, + "loss": 1.4807, + "step": 40220 + }, + { + "epoch": 7.31, + "learning_rate": 8.064615944666908e-07, + "loss": 2.0459, + "step": 40221 + }, + { + "epoch": 7.31, + "learning_rate": 8.064069894430288e-07, + "loss": 0.8549, + "step": 40222 + }, + { + "epoch": 7.31, + "learning_rate": 8.063523844193666e-07, + "loss": 2.3232, + "step": 40223 + }, + { + "epoch": 7.31, + "learning_rate": 8.062977793957044e-07, + "loss": 0.0068, + "step": 40224 + }, + { + "epoch": 7.31, + "learning_rate": 8.062431743720423e-07, + "loss": 0.1082, + "step": 40225 + }, + { + "epoch": 7.32, + "learning_rate": 8.0618856934838e-07, + "loss": 0.2225, + "step": 40226 + }, + { + "epoch": 7.32, + "learning_rate": 8.061339643247179e-07, + "loss": 0.0937, + "step": 40227 + }, + { + "epoch": 7.32, + "learning_rate": 8.060793593010556e-07, + "loss": 0.1282, + "step": 40228 + }, + { + "epoch": 7.32, + "learning_rate": 8.060247542773935e-07, + "loss": 2.4011, + "step": 40229 + }, + { + "epoch": 7.32, + "learning_rate": 8.059701492537314e-07, + "loss": 0.0472, + "step": 40230 + }, + { + "epoch": 7.32, + "learning_rate": 8.059155442300691e-07, + "loss": 0.0291, + "step": 40231 + }, + { + "epoch": 7.32, + "learning_rate": 8.058609392064071e-07, + "loss": 0.0114, + "step": 40232 + }, + { + "epoch": 7.32, + "learning_rate": 8.058063341827449e-07, + "loss": 0.092, + "step": 40233 + }, + { + "epoch": 7.32, + "learning_rate": 8.057517291590826e-07, + "loss": 0.4915, + "step": 40234 + }, + { + "epoch": 7.32, + "learning_rate": 8.056971241354204e-07, + "loss": 0.0197, + "step": 40235 + }, + { + "epoch": 7.32, + "learning_rate": 8.056425191117583e-07, + "loss": 0.0166, + "step": 40236 + }, + { + "epoch": 7.32, + "learning_rate": 8.055879140880961e-07, + "loss": 1.1414, + "step": 40237 + }, + { + "epoch": 7.32, + "learning_rate": 8.055333090644339e-07, + "loss": 0.0577, + "step": 40238 + }, + { + "epoch": 7.32, + "learning_rate": 8.054787040407718e-07, + "loss": 0.5732, + "step": 40239 + }, + { + "epoch": 7.32, + "learning_rate": 8.054240990171097e-07, + "loss": 0.8363, + "step": 40240 + }, + { + "epoch": 7.32, + "learning_rate": 8.053694939934474e-07, + "loss": 0.0413, + "step": 40241 + }, + { + "epoch": 7.32, + "learning_rate": 8.053148889697851e-07, + "loss": 0.2061, + "step": 40242 + }, + { + "epoch": 7.32, + "learning_rate": 8.052602839461231e-07, + "loss": 1.1731, + "step": 40243 + }, + { + "epoch": 7.32, + "learning_rate": 8.052056789224609e-07, + "loss": 0.0097, + "step": 40244 + }, + { + "epoch": 7.32, + "learning_rate": 8.051510738987986e-07, + "loss": 0.8055, + "step": 40245 + }, + { + "epoch": 7.32, + "learning_rate": 8.050964688751366e-07, + "loss": 0.0149, + "step": 40246 + }, + { + "epoch": 7.32, + "learning_rate": 8.050418638514744e-07, + "loss": 0.144, + "step": 40247 + }, + { + "epoch": 7.32, + "learning_rate": 8.049872588278121e-07, + "loss": 0.0387, + "step": 40248 + }, + { + "epoch": 7.32, + "learning_rate": 8.0493265380415e-07, + "loss": 1.0848, + "step": 40249 + }, + { + "epoch": 7.32, + "learning_rate": 8.048780487804879e-07, + "loss": 0.0037, + "step": 40250 + }, + { + "epoch": 7.32, + "learning_rate": 8.048234437568257e-07, + "loss": 0.075, + "step": 40251 + }, + { + "epoch": 7.32, + "learning_rate": 8.047688387331634e-07, + "loss": 0.0055, + "step": 40252 + }, + { + "epoch": 7.32, + "learning_rate": 8.047142337095013e-07, + "loss": 0.6232, + "step": 40253 + }, + { + "epoch": 7.32, + "learning_rate": 8.046596286858392e-07, + "loss": 0.0616, + "step": 40254 + }, + { + "epoch": 7.32, + "learning_rate": 8.046050236621769e-07, + "loss": 0.1459, + "step": 40255 + }, + { + "epoch": 7.32, + "learning_rate": 8.045504186385147e-07, + "loss": 0.2629, + "step": 40256 + }, + { + "epoch": 7.32, + "learning_rate": 8.044958136148527e-07, + "loss": 0.588, + "step": 40257 + }, + { + "epoch": 7.32, + "learning_rate": 8.044412085911904e-07, + "loss": 0.1866, + "step": 40258 + }, + { + "epoch": 7.32, + "learning_rate": 8.043866035675281e-07, + "loss": 2.9168, + "step": 40259 + }, + { + "epoch": 7.32, + "learning_rate": 8.043319985438661e-07, + "loss": 0.0723, + "step": 40260 + }, + { + "epoch": 7.32, + "learning_rate": 8.042773935202039e-07, + "loss": 0.0088, + "step": 40261 + }, + { + "epoch": 7.32, + "learning_rate": 8.042227884965417e-07, + "loss": 1.2048, + "step": 40262 + }, + { + "epoch": 7.32, + "learning_rate": 8.041681834728795e-07, + "loss": 0.1305, + "step": 40263 + }, + { + "epoch": 7.32, + "learning_rate": 8.041135784492174e-07, + "loss": 0.1661, + "step": 40264 + }, + { + "epoch": 7.32, + "learning_rate": 8.040589734255552e-07, + "loss": 0.0311, + "step": 40265 + }, + { + "epoch": 7.32, + "learning_rate": 8.040043684018929e-07, + "loss": 0.006, + "step": 40266 + }, + { + "epoch": 7.32, + "learning_rate": 8.039497633782308e-07, + "loss": 2.8656, + "step": 40267 + }, + { + "epoch": 7.32, + "learning_rate": 8.038951583545687e-07, + "loss": 0.7137, + "step": 40268 + }, + { + "epoch": 7.32, + "learning_rate": 8.038405533309064e-07, + "loss": 0.0743, + "step": 40269 + }, + { + "epoch": 7.32, + "learning_rate": 8.037859483072443e-07, + "loss": 0.12, + "step": 40270 + }, + { + "epoch": 7.32, + "learning_rate": 8.037313432835822e-07, + "loss": 0.0329, + "step": 40271 + }, + { + "epoch": 7.32, + "learning_rate": 8.036767382599199e-07, + "loss": 0.1196, + "step": 40272 + }, + { + "epoch": 7.32, + "learning_rate": 8.036221332362578e-07, + "loss": 0.0456, + "step": 40273 + }, + { + "epoch": 7.32, + "learning_rate": 8.035675282125957e-07, + "loss": 0.6985, + "step": 40274 + }, + { + "epoch": 7.32, + "learning_rate": 8.035129231889334e-07, + "loss": 0.0048, + "step": 40275 + }, + { + "epoch": 7.32, + "learning_rate": 8.034583181652712e-07, + "loss": 0.5814, + "step": 40276 + }, + { + "epoch": 7.32, + "learning_rate": 8.03403713141609e-07, + "loss": 1.0658, + "step": 40277 + }, + { + "epoch": 7.32, + "learning_rate": 8.03349108117947e-07, + "loss": 0.3614, + "step": 40278 + }, + { + "epoch": 7.32, + "learning_rate": 8.032945030942847e-07, + "loss": 0.374, + "step": 40279 + }, + { + "epoch": 7.32, + "learning_rate": 8.032398980706225e-07, + "loss": 0.089, + "step": 40280 + }, + { + "epoch": 7.33, + "learning_rate": 8.031852930469605e-07, + "loss": 0.7446, + "step": 40281 + }, + { + "epoch": 7.33, + "learning_rate": 8.031306880232982e-07, + "loss": 0.0076, + "step": 40282 + }, + { + "epoch": 7.33, + "learning_rate": 8.030760829996359e-07, + "loss": 0.3202, + "step": 40283 + }, + { + "epoch": 7.33, + "learning_rate": 8.030214779759738e-07, + "loss": 0.5341, + "step": 40284 + }, + { + "epoch": 7.33, + "learning_rate": 8.029668729523117e-07, + "loss": 2.5402, + "step": 40285 + }, + { + "epoch": 7.33, + "learning_rate": 8.029122679286494e-07, + "loss": 0.2009, + "step": 40286 + }, + { + "epoch": 7.33, + "learning_rate": 8.028576629049873e-07, + "loss": 0.2092, + "step": 40287 + }, + { + "epoch": 7.33, + "learning_rate": 8.02803057881325e-07, + "loss": 0.1607, + "step": 40288 + }, + { + "epoch": 7.33, + "learning_rate": 8.02748452857663e-07, + "loss": 0.017, + "step": 40289 + }, + { + "epoch": 7.33, + "learning_rate": 8.026938478340007e-07, + "loss": 1.0528, + "step": 40290 + }, + { + "epoch": 7.33, + "learning_rate": 8.026392428103385e-07, + "loss": 0.713, + "step": 40291 + }, + { + "epoch": 7.33, + "learning_rate": 8.025846377866765e-07, + "loss": 0.8077, + "step": 40292 + }, + { + "epoch": 7.33, + "learning_rate": 8.025300327630142e-07, + "loss": 0.1764, + "step": 40293 + }, + { + "epoch": 7.33, + "learning_rate": 8.02475427739352e-07, + "loss": 0.0801, + "step": 40294 + }, + { + "epoch": 7.33, + "learning_rate": 8.024208227156898e-07, + "loss": 0.5187, + "step": 40295 + }, + { + "epoch": 7.33, + "learning_rate": 8.023662176920277e-07, + "loss": 0.6757, + "step": 40296 + }, + { + "epoch": 7.33, + "learning_rate": 8.023116126683655e-07, + "loss": 0.011, + "step": 40297 + }, + { + "epoch": 7.33, + "learning_rate": 8.022570076447033e-07, + "loss": 0.8025, + "step": 40298 + }, + { + "epoch": 7.33, + "learning_rate": 8.022024026210412e-07, + "loss": 0.2854, + "step": 40299 + }, + { + "epoch": 7.33, + "learning_rate": 8.02147797597379e-07, + "loss": 0.0055, + "step": 40300 + }, + { + "epoch": 7.33, + "learning_rate": 8.020931925737168e-07, + "loss": 0.0082, + "step": 40301 + }, + { + "epoch": 7.33, + "learning_rate": 8.020385875500545e-07, + "loss": 0.1829, + "step": 40302 + }, + { + "epoch": 7.33, + "learning_rate": 8.019839825263925e-07, + "loss": 0.0474, + "step": 40303 + }, + { + "epoch": 7.33, + "learning_rate": 8.019293775027303e-07, + "loss": 1.0923, + "step": 40304 + }, + { + "epoch": 7.33, + "learning_rate": 8.01874772479068e-07, + "loss": 0.0735, + "step": 40305 + }, + { + "epoch": 7.33, + "learning_rate": 8.01820167455406e-07, + "loss": 0.6433, + "step": 40306 + }, + { + "epoch": 7.33, + "learning_rate": 8.017655624317437e-07, + "loss": 0.0059, + "step": 40307 + }, + { + "epoch": 7.33, + "learning_rate": 8.017109574080815e-07, + "loss": 0.2466, + "step": 40308 + }, + { + "epoch": 7.33, + "learning_rate": 8.016563523844193e-07, + "loss": 0.3334, + "step": 40309 + }, + { + "epoch": 7.33, + "learning_rate": 8.016017473607572e-07, + "loss": 0.0136, + "step": 40310 + }, + { + "epoch": 7.33, + "learning_rate": 8.015471423370951e-07, + "loss": 0.3236, + "step": 40311 + }, + { + "epoch": 7.33, + "learning_rate": 8.014925373134328e-07, + "loss": 0.0078, + "step": 40312 + }, + { + "epoch": 7.33, + "learning_rate": 8.014379322897707e-07, + "loss": 0.2244, + "step": 40313 + }, + { + "epoch": 7.33, + "learning_rate": 8.013833272661086e-07, + "loss": 1.2561, + "step": 40314 + }, + { + "epoch": 7.33, + "learning_rate": 8.013287222424463e-07, + "loss": 1.0955, + "step": 40315 + }, + { + "epoch": 7.33, + "learning_rate": 8.01274117218784e-07, + "loss": 0.0436, + "step": 40316 + }, + { + "epoch": 7.33, + "learning_rate": 8.01219512195122e-07, + "loss": 0.1713, + "step": 40317 + }, + { + "epoch": 7.33, + "learning_rate": 8.011649071714598e-07, + "loss": 0.0182, + "step": 40318 + }, + { + "epoch": 7.33, + "learning_rate": 8.011103021477975e-07, + "loss": 0.7051, + "step": 40319 + }, + { + "epoch": 7.33, + "learning_rate": 8.010556971241355e-07, + "loss": 0.4058, + "step": 40320 + }, + { + "epoch": 7.33, + "learning_rate": 8.010010921004733e-07, + "loss": 1.3861, + "step": 40321 + }, + { + "epoch": 7.33, + "learning_rate": 8.009464870768111e-07, + "loss": 0.5887, + "step": 40322 + }, + { + "epoch": 7.33, + "learning_rate": 8.008918820531488e-07, + "loss": 0.0061, + "step": 40323 + }, + { + "epoch": 7.33, + "learning_rate": 8.008372770294867e-07, + "loss": 0.2724, + "step": 40324 + }, + { + "epoch": 7.33, + "learning_rate": 8.007826720058246e-07, + "loss": 1.1321, + "step": 40325 + }, + { + "epoch": 7.33, + "learning_rate": 8.007280669821623e-07, + "loss": 0.0075, + "step": 40326 + }, + { + "epoch": 7.33, + "learning_rate": 8.006734619585003e-07, + "loss": 0.2339, + "step": 40327 + }, + { + "epoch": 7.33, + "learning_rate": 8.006188569348381e-07, + "loss": 0.4298, + "step": 40328 + }, + { + "epoch": 7.33, + "learning_rate": 8.005642519111758e-07, + "loss": 0.1035, + "step": 40329 + }, + { + "epoch": 7.33, + "learning_rate": 8.005096468875136e-07, + "loss": 0.0183, + "step": 40330 + }, + { + "epoch": 7.33, + "learning_rate": 8.004550418638515e-07, + "loss": 0.0054, + "step": 40331 + }, + { + "epoch": 7.33, + "learning_rate": 8.004004368401893e-07, + "loss": 0.3628, + "step": 40332 + }, + { + "epoch": 7.33, + "learning_rate": 8.003458318165271e-07, + "loss": 0.0151, + "step": 40333 + }, + { + "epoch": 7.33, + "learning_rate": 8.00291226792865e-07, + "loss": 1.5083, + "step": 40334 + }, + { + "epoch": 7.33, + "learning_rate": 8.002366217692028e-07, + "loss": 0.4266, + "step": 40335 + }, + { + "epoch": 7.34, + "learning_rate": 8.001820167455406e-07, + "loss": 0.7902, + "step": 40336 + }, + { + "epoch": 7.34, + "learning_rate": 8.001274117218784e-07, + "loss": 0.0085, + "step": 40337 + }, + { + "epoch": 7.34, + "learning_rate": 8.000728066982164e-07, + "loss": 0.1425, + "step": 40338 + }, + { + "epoch": 7.34, + "learning_rate": 8.000182016745541e-07, + "loss": 0.2585, + "step": 40339 + }, + { + "epoch": 7.34, + "learning_rate": 7.999635966508918e-07, + "loss": 1.8191, + "step": 40340 + }, + { + "epoch": 7.34, + "learning_rate": 7.999089916272298e-07, + "loss": 0.0226, + "step": 40341 + }, + { + "epoch": 7.34, + "learning_rate": 7.998543866035676e-07, + "loss": 0.0064, + "step": 40342 + }, + { + "epoch": 7.34, + "learning_rate": 7.997997815799053e-07, + "loss": 0.8798, + "step": 40343 + }, + { + "epoch": 7.34, + "learning_rate": 7.997451765562432e-07, + "loss": 0.6351, + "step": 40344 + }, + { + "epoch": 7.34, + "learning_rate": 7.996905715325811e-07, + "loss": 0.411, + "step": 40345 + }, + { + "epoch": 7.34, + "learning_rate": 7.996359665089188e-07, + "loss": 0.0123, + "step": 40346 + }, + { + "epoch": 7.34, + "learning_rate": 7.995813614852566e-07, + "loss": 0.0157, + "step": 40347 + }, + { + "epoch": 7.34, + "learning_rate": 7.995267564615945e-07, + "loss": 0.0092, + "step": 40348 + }, + { + "epoch": 7.34, + "learning_rate": 7.994721514379324e-07, + "loss": 0.492, + "step": 40349 + }, + { + "epoch": 7.34, + "learning_rate": 7.994175464142701e-07, + "loss": 0.091, + "step": 40350 + }, + { + "epoch": 7.34, + "learning_rate": 7.993629413906079e-07, + "loss": 0.3483, + "step": 40351 + }, + { + "epoch": 7.34, + "learning_rate": 7.993083363669459e-07, + "loss": 0.2049, + "step": 40352 + }, + { + "epoch": 7.34, + "learning_rate": 7.992537313432836e-07, + "loss": 0.2496, + "step": 40353 + }, + { + "epoch": 7.34, + "learning_rate": 7.991991263196214e-07, + "loss": 0.1742, + "step": 40354 + }, + { + "epoch": 7.34, + "learning_rate": 7.991445212959593e-07, + "loss": 1.6074, + "step": 40355 + }, + { + "epoch": 7.34, + "learning_rate": 7.990899162722971e-07, + "loss": 1.4088, + "step": 40356 + }, + { + "epoch": 7.34, + "learning_rate": 7.990353112486348e-07, + "loss": 3.1616, + "step": 40357 + }, + { + "epoch": 7.34, + "learning_rate": 7.989807062249727e-07, + "loss": 1.7228, + "step": 40358 + }, + { + "epoch": 7.34, + "learning_rate": 7.989261012013106e-07, + "loss": 0.2287, + "step": 40359 + }, + { + "epoch": 7.34, + "learning_rate": 7.988714961776484e-07, + "loss": 0.3005, + "step": 40360 + }, + { + "epoch": 7.34, + "learning_rate": 7.988168911539862e-07, + "loss": 0.2157, + "step": 40361 + }, + { + "epoch": 7.34, + "learning_rate": 7.987622861303241e-07, + "loss": 0.0211, + "step": 40362 + }, + { + "epoch": 7.34, + "learning_rate": 7.987076811066619e-07, + "loss": 0.0074, + "step": 40363 + }, + { + "epoch": 7.34, + "learning_rate": 7.986530760829996e-07, + "loss": 1.6763, + "step": 40364 + }, + { + "epoch": 7.34, + "learning_rate": 7.985984710593374e-07, + "loss": 1.7447, + "step": 40365 + }, + { + "epoch": 7.34, + "learning_rate": 7.985438660356754e-07, + "loss": 0.0732, + "step": 40366 + }, + { + "epoch": 7.34, + "learning_rate": 7.984892610120131e-07, + "loss": 0.3029, + "step": 40367 + }, + { + "epoch": 7.34, + "learning_rate": 7.98434655988351e-07, + "loss": 0.0332, + "step": 40368 + }, + { + "epoch": 7.34, + "learning_rate": 7.983800509646887e-07, + "loss": 1.0677, + "step": 40369 + }, + { + "epoch": 7.34, + "learning_rate": 7.983254459410266e-07, + "loss": 0.0617, + "step": 40370 + }, + { + "epoch": 7.34, + "learning_rate": 7.982708409173644e-07, + "loss": 1.5152, + "step": 40371 + }, + { + "epoch": 7.34, + "learning_rate": 7.982162358937022e-07, + "loss": 0.1066, + "step": 40372 + }, + { + "epoch": 7.34, + "learning_rate": 7.981616308700401e-07, + "loss": 1.8419, + "step": 40373 + }, + { + "epoch": 7.34, + "learning_rate": 7.981070258463779e-07, + "loss": 0.211, + "step": 40374 + }, + { + "epoch": 7.34, + "learning_rate": 7.980524208227157e-07, + "loss": 0.3039, + "step": 40375 + }, + { + "epoch": 7.34, + "learning_rate": 7.979978157990534e-07, + "loss": 1.4224, + "step": 40376 + }, + { + "epoch": 7.34, + "learning_rate": 7.979432107753914e-07, + "loss": 1.2116, + "step": 40377 + }, + { + "epoch": 7.34, + "learning_rate": 7.978886057517292e-07, + "loss": 1.858, + "step": 40378 + }, + { + "epoch": 7.34, + "learning_rate": 7.97834000728067e-07, + "loss": 0.0143, + "step": 40379 + }, + { + "epoch": 7.34, + "learning_rate": 7.977793957044049e-07, + "loss": 0.4285, + "step": 40380 + }, + { + "epoch": 7.34, + "learning_rate": 7.977247906807426e-07, + "loss": 0.0944, + "step": 40381 + }, + { + "epoch": 7.34, + "learning_rate": 7.976701856570805e-07, + "loss": 0.4548, + "step": 40382 + }, + { + "epoch": 7.34, + "learning_rate": 7.976155806334182e-07, + "loss": 0.539, + "step": 40383 + }, + { + "epoch": 7.34, + "learning_rate": 7.975609756097561e-07, + "loss": 0.5014, + "step": 40384 + }, + { + "epoch": 7.34, + "learning_rate": 7.97506370586094e-07, + "loss": 0.0549, + "step": 40385 + }, + { + "epoch": 7.34, + "learning_rate": 7.974517655624317e-07, + "loss": 1.0656, + "step": 40386 + }, + { + "epoch": 7.34, + "learning_rate": 7.973971605387697e-07, + "loss": 0.0741, + "step": 40387 + }, + { + "epoch": 7.34, + "learning_rate": 7.973425555151074e-07, + "loss": 0.8533, + "step": 40388 + }, + { + "epoch": 7.34, + "learning_rate": 7.972879504914452e-07, + "loss": 0.6443, + "step": 40389 + }, + { + "epoch": 7.34, + "learning_rate": 7.97233345467783e-07, + "loss": 1.1547, + "step": 40390 + }, + { + "epoch": 7.35, + "learning_rate": 7.971787404441209e-07, + "loss": 0.1312, + "step": 40391 + }, + { + "epoch": 7.35, + "learning_rate": 7.971241354204587e-07, + "loss": 0.8099, + "step": 40392 + }, + { + "epoch": 7.35, + "learning_rate": 7.970695303967965e-07, + "loss": 0.0222, + "step": 40393 + }, + { + "epoch": 7.35, + "learning_rate": 7.970149253731344e-07, + "loss": 0.0073, + "step": 40394 + }, + { + "epoch": 7.35, + "learning_rate": 7.969603203494721e-07, + "loss": 0.0761, + "step": 40395 + }, + { + "epoch": 7.35, + "learning_rate": 7.9690571532581e-07, + "loss": 0.2684, + "step": 40396 + }, + { + "epoch": 7.35, + "learning_rate": 7.968511103021477e-07, + "loss": 0.158, + "step": 40397 + }, + { + "epoch": 7.35, + "learning_rate": 7.967965052784857e-07, + "loss": 1.9297, + "step": 40398 + }, + { + "epoch": 7.35, + "learning_rate": 7.967419002548235e-07, + "loss": 0.0228, + "step": 40399 + }, + { + "epoch": 7.35, + "learning_rate": 7.966872952311612e-07, + "loss": 0.011, + "step": 40400 + }, + { + "epoch": 7.35, + "learning_rate": 7.966326902074992e-07, + "loss": 0.5642, + "step": 40401 + }, + { + "epoch": 7.35, + "learning_rate": 7.96578085183837e-07, + "loss": 0.4138, + "step": 40402 + }, + { + "epoch": 7.35, + "learning_rate": 7.965234801601747e-07, + "loss": 0.7287, + "step": 40403 + }, + { + "epoch": 7.35, + "learning_rate": 7.964688751365125e-07, + "loss": 0.3838, + "step": 40404 + }, + { + "epoch": 7.35, + "learning_rate": 7.964142701128504e-07, + "loss": 0.2996, + "step": 40405 + }, + { + "epoch": 7.35, + "learning_rate": 7.963596650891882e-07, + "loss": 0.5347, + "step": 40406 + }, + { + "epoch": 7.35, + "learning_rate": 7.96305060065526e-07, + "loss": 0.3876, + "step": 40407 + }, + { + "epoch": 7.35, + "learning_rate": 7.962504550418639e-07, + "loss": 0.0068, + "step": 40408 + }, + { + "epoch": 7.35, + "learning_rate": 7.961958500182018e-07, + "loss": 0.0096, + "step": 40409 + }, + { + "epoch": 7.35, + "learning_rate": 7.961412449945395e-07, + "loss": 0.34, + "step": 40410 + }, + { + "epoch": 7.35, + "learning_rate": 7.960866399708772e-07, + "loss": 0.0473, + "step": 40411 + }, + { + "epoch": 7.35, + "learning_rate": 7.960320349472152e-07, + "loss": 0.0111, + "step": 40412 + }, + { + "epoch": 7.35, + "learning_rate": 7.95977429923553e-07, + "loss": 1.2856, + "step": 40413 + }, + { + "epoch": 7.35, + "learning_rate": 7.959228248998907e-07, + "loss": 0.1112, + "step": 40414 + }, + { + "epoch": 7.35, + "learning_rate": 7.958682198762287e-07, + "loss": 0.6523, + "step": 40415 + }, + { + "epoch": 7.35, + "learning_rate": 7.958136148525665e-07, + "loss": 1.3848, + "step": 40416 + }, + { + "epoch": 7.35, + "learning_rate": 7.957590098289043e-07, + "loss": 0.0682, + "step": 40417 + }, + { + "epoch": 7.35, + "learning_rate": 7.957044048052421e-07, + "loss": 0.0713, + "step": 40418 + }, + { + "epoch": 7.35, + "learning_rate": 7.9564979978158e-07, + "loss": 0.3448, + "step": 40419 + }, + { + "epoch": 7.35, + "learning_rate": 7.955951947579178e-07, + "loss": 1.3637, + "step": 40420 + }, + { + "epoch": 7.35, + "learning_rate": 7.955405897342555e-07, + "loss": 0.0764, + "step": 40421 + }, + { + "epoch": 7.35, + "learning_rate": 7.954859847105934e-07, + "loss": 0.0193, + "step": 40422 + }, + { + "epoch": 7.35, + "learning_rate": 7.954313796869313e-07, + "loss": 0.0676, + "step": 40423 + }, + { + "epoch": 7.35, + "learning_rate": 7.95376774663269e-07, + "loss": 1.4101, + "step": 40424 + }, + { + "epoch": 7.35, + "learning_rate": 7.953221696396068e-07, + "loss": 1.2394, + "step": 40425 + }, + { + "epoch": 7.35, + "learning_rate": 7.952675646159448e-07, + "loss": 1.4973, + "step": 40426 + }, + { + "epoch": 7.35, + "learning_rate": 7.952129595922825e-07, + "loss": 0.4085, + "step": 40427 + }, + { + "epoch": 7.35, + "learning_rate": 7.951583545686203e-07, + "loss": 0.1662, + "step": 40428 + }, + { + "epoch": 7.35, + "learning_rate": 7.951037495449582e-07, + "loss": 0.7996, + "step": 40429 + }, + { + "epoch": 7.35, + "learning_rate": 7.95049144521296e-07, + "loss": 0.2846, + "step": 40430 + }, + { + "epoch": 7.35, + "learning_rate": 7.949945394976338e-07, + "loss": 0.0749, + "step": 40431 + }, + { + "epoch": 7.35, + "learning_rate": 7.949399344739716e-07, + "loss": 3.0187, + "step": 40432 + }, + { + "epoch": 7.35, + "learning_rate": 7.948853294503095e-07, + "loss": 2.9446, + "step": 40433 + }, + { + "epoch": 7.35, + "learning_rate": 7.948307244266473e-07, + "loss": 0.4131, + "step": 40434 + }, + { + "epoch": 7.35, + "learning_rate": 7.94776119402985e-07, + "loss": 0.0223, + "step": 40435 + }, + { + "epoch": 7.35, + "learning_rate": 7.94721514379323e-07, + "loss": 1.3571, + "step": 40436 + }, + { + "epoch": 7.35, + "learning_rate": 7.946669093556608e-07, + "loss": 0.2912, + "step": 40437 + }, + { + "epoch": 7.35, + "learning_rate": 7.946123043319985e-07, + "loss": 1.117, + "step": 40438 + }, + { + "epoch": 7.35, + "learning_rate": 7.945576993083364e-07, + "loss": 0.0097, + "step": 40439 + }, + { + "epoch": 7.35, + "learning_rate": 7.945030942846743e-07, + "loss": 0.3217, + "step": 40440 + }, + { + "epoch": 7.35, + "learning_rate": 7.94448489261012e-07, + "loss": 0.224, + "step": 40441 + }, + { + "epoch": 7.35, + "learning_rate": 7.943938842373499e-07, + "loss": 0.3209, + "step": 40442 + }, + { + "epoch": 7.35, + "learning_rate": 7.943392792136876e-07, + "loss": 0.7282, + "step": 40443 + }, + { + "epoch": 7.35, + "learning_rate": 7.942846741900255e-07, + "loss": 0.4381, + "step": 40444 + }, + { + "epoch": 7.35, + "learning_rate": 7.942300691663633e-07, + "loss": 0.2687, + "step": 40445 + }, + { + "epoch": 7.36, + "learning_rate": 7.941754641427011e-07, + "loss": 2.9216, + "step": 40446 + }, + { + "epoch": 7.36, + "learning_rate": 7.941208591190391e-07, + "loss": 0.2115, + "step": 40447 + }, + { + "epoch": 7.36, + "learning_rate": 7.940662540953768e-07, + "loss": 0.0064, + "step": 40448 + }, + { + "epoch": 7.36, + "learning_rate": 7.940116490717146e-07, + "loss": 2.1971, + "step": 40449 + }, + { + "epoch": 7.36, + "learning_rate": 7.939570440480524e-07, + "loss": 0.2688, + "step": 40450 + }, + { + "epoch": 7.36, + "learning_rate": 7.939024390243903e-07, + "loss": 0.0769, + "step": 40451 + }, + { + "epoch": 7.36, + "learning_rate": 7.93847834000728e-07, + "loss": 0.177, + "step": 40452 + }, + { + "epoch": 7.36, + "learning_rate": 7.937932289770659e-07, + "loss": 0.0879, + "step": 40453 + }, + { + "epoch": 7.36, + "learning_rate": 7.937386239534038e-07, + "loss": 0.4052, + "step": 40454 + }, + { + "epoch": 7.36, + "learning_rate": 7.936840189297415e-07, + "loss": 1.447, + "step": 40455 + }, + { + "epoch": 7.36, + "learning_rate": 7.936294139060794e-07, + "loss": 0.181, + "step": 40456 + }, + { + "epoch": 7.36, + "learning_rate": 7.935748088824171e-07, + "loss": 0.6248, + "step": 40457 + }, + { + "epoch": 7.36, + "learning_rate": 7.935202038587551e-07, + "loss": 0.0788, + "step": 40458 + }, + { + "epoch": 7.36, + "learning_rate": 7.934655988350928e-07, + "loss": 1.2552, + "step": 40459 + }, + { + "epoch": 7.36, + "learning_rate": 7.934109938114306e-07, + "loss": 0.0939, + "step": 40460 + }, + { + "epoch": 7.36, + "learning_rate": 7.933563887877686e-07, + "loss": 2.2476, + "step": 40461 + }, + { + "epoch": 7.36, + "learning_rate": 7.933017837641063e-07, + "loss": 1.206, + "step": 40462 + }, + { + "epoch": 7.36, + "learning_rate": 7.932471787404441e-07, + "loss": 0.0759, + "step": 40463 + }, + { + "epoch": 7.36, + "learning_rate": 7.931925737167819e-07, + "loss": 1.433, + "step": 40464 + }, + { + "epoch": 7.36, + "learning_rate": 7.931379686931198e-07, + "loss": 0.1809, + "step": 40465 + }, + { + "epoch": 7.36, + "learning_rate": 7.930833636694577e-07, + "loss": 0.0498, + "step": 40466 + }, + { + "epoch": 7.36, + "learning_rate": 7.930287586457954e-07, + "loss": 2.3374, + "step": 40467 + }, + { + "epoch": 7.36, + "learning_rate": 7.929741536221333e-07, + "loss": 0.1126, + "step": 40468 + }, + { + "epoch": 7.36, + "learning_rate": 7.929195485984711e-07, + "loss": 2.1689, + "step": 40469 + }, + { + "epoch": 7.36, + "learning_rate": 7.928649435748089e-07, + "loss": 0.3206, + "step": 40470 + }, + { + "epoch": 7.36, + "learning_rate": 7.928103385511466e-07, + "loss": 0.1805, + "step": 40471 + }, + { + "epoch": 7.36, + "learning_rate": 7.927557335274846e-07, + "loss": 0.0173, + "step": 40472 + }, + { + "epoch": 7.36, + "learning_rate": 7.927011285038224e-07, + "loss": 1.5313, + "step": 40473 + }, + { + "epoch": 7.36, + "learning_rate": 7.926465234801601e-07, + "loss": 0.175, + "step": 40474 + }, + { + "epoch": 7.36, + "learning_rate": 7.925919184564981e-07, + "loss": 0.0986, + "step": 40475 + }, + { + "epoch": 7.36, + "learning_rate": 7.925373134328358e-07, + "loss": 0.3627, + "step": 40476 + }, + { + "epoch": 7.36, + "learning_rate": 7.924827084091737e-07, + "loss": 0.0134, + "step": 40477 + }, + { + "epoch": 7.36, + "learning_rate": 7.924281033855114e-07, + "loss": 0.3734, + "step": 40478 + }, + { + "epoch": 7.36, + "learning_rate": 7.923734983618493e-07, + "loss": 0.2864, + "step": 40479 + }, + { + "epoch": 7.36, + "learning_rate": 7.923188933381872e-07, + "loss": 0.0035, + "step": 40480 + }, + { + "epoch": 7.36, + "learning_rate": 7.922642883145249e-07, + "loss": 0.1084, + "step": 40481 + }, + { + "epoch": 7.36, + "learning_rate": 7.922096832908628e-07, + "loss": 0.2077, + "step": 40482 + }, + { + "epoch": 7.36, + "learning_rate": 7.921550782672007e-07, + "loss": 2.3824, + "step": 40483 + }, + { + "epoch": 7.36, + "learning_rate": 7.921004732435384e-07, + "loss": 0.2954, + "step": 40484 + }, + { + "epoch": 7.36, + "learning_rate": 7.920458682198761e-07, + "loss": 0.2318, + "step": 40485 + }, + { + "epoch": 7.36, + "learning_rate": 7.919912631962141e-07, + "loss": 0.0049, + "step": 40486 + }, + { + "epoch": 7.36, + "learning_rate": 7.919366581725519e-07, + "loss": 0.0278, + "step": 40487 + }, + { + "epoch": 7.36, + "learning_rate": 7.918820531488897e-07, + "loss": 0.05, + "step": 40488 + }, + { + "epoch": 7.36, + "learning_rate": 7.918274481252276e-07, + "loss": 0.1567, + "step": 40489 + }, + { + "epoch": 7.36, + "learning_rate": 7.917728431015654e-07, + "loss": 0.7968, + "step": 40490 + }, + { + "epoch": 7.36, + "learning_rate": 7.917182380779032e-07, + "loss": 0.4325, + "step": 40491 + }, + { + "epoch": 7.36, + "learning_rate": 7.916636330542409e-07, + "loss": 0.048, + "step": 40492 + }, + { + "epoch": 7.36, + "learning_rate": 7.916090280305788e-07, + "loss": 0.3229, + "step": 40493 + }, + { + "epoch": 7.36, + "learning_rate": 7.915544230069167e-07, + "loss": 1.9097, + "step": 40494 + }, + { + "epoch": 7.36, + "learning_rate": 7.914998179832544e-07, + "loss": 0.0106, + "step": 40495 + }, + { + "epoch": 7.36, + "learning_rate": 7.914452129595924e-07, + "loss": 0.0513, + "step": 40496 + }, + { + "epoch": 7.36, + "learning_rate": 7.913906079359302e-07, + "loss": 0.0097, + "step": 40497 + }, + { + "epoch": 7.36, + "learning_rate": 7.913360029122679e-07, + "loss": 1.5321, + "step": 40498 + }, + { + "epoch": 7.36, + "learning_rate": 7.912813978886057e-07, + "loss": 1.1896, + "step": 40499 + }, + { + "epoch": 7.36, + "learning_rate": 7.912267928649436e-07, + "loss": 0.1164, + "step": 40500 + }, + { + "epoch": 7.37, + "learning_rate": 7.911721878412814e-07, + "loss": 1.3747, + "step": 40501 + }, + { + "epoch": 7.37, + "learning_rate": 7.911175828176192e-07, + "loss": 1.2339, + "step": 40502 + }, + { + "epoch": 7.37, + "learning_rate": 7.910629777939571e-07, + "loss": 0.3529, + "step": 40503 + }, + { + "epoch": 7.37, + "learning_rate": 7.91008372770295e-07, + "loss": 0.5189, + "step": 40504 + }, + { + "epoch": 7.37, + "learning_rate": 7.909537677466327e-07, + "loss": 0.2342, + "step": 40505 + }, + { + "epoch": 7.37, + "learning_rate": 7.908991627229705e-07, + "loss": 0.5165, + "step": 40506 + }, + { + "epoch": 7.37, + "learning_rate": 7.908445576993085e-07, + "loss": 0.4106, + "step": 40507 + }, + { + "epoch": 7.37, + "learning_rate": 7.907899526756462e-07, + "loss": 0.0274, + "step": 40508 + }, + { + "epoch": 7.37, + "learning_rate": 7.907353476519839e-07, + "loss": 0.0901, + "step": 40509 + }, + { + "epoch": 7.37, + "learning_rate": 7.906807426283219e-07, + "loss": 0.8266, + "step": 40510 + }, + { + "epoch": 7.37, + "learning_rate": 7.906261376046597e-07, + "loss": 1.3603, + "step": 40511 + }, + { + "epoch": 7.37, + "learning_rate": 7.905715325809974e-07, + "loss": 0.7743, + "step": 40512 + }, + { + "epoch": 7.37, + "learning_rate": 7.905169275573353e-07, + "loss": 0.7037, + "step": 40513 + }, + { + "epoch": 7.37, + "learning_rate": 7.904623225336732e-07, + "loss": 0.1266, + "step": 40514 + }, + { + "epoch": 7.37, + "learning_rate": 7.90407717510011e-07, + "loss": 0.0244, + "step": 40515 + }, + { + "epoch": 7.37, + "learning_rate": 7.903531124863487e-07, + "loss": 0.0036, + "step": 40516 + }, + { + "epoch": 7.37, + "learning_rate": 7.902985074626866e-07, + "loss": 1.2045, + "step": 40517 + }, + { + "epoch": 7.37, + "learning_rate": 7.902439024390245e-07, + "loss": 0.1141, + "step": 40518 + }, + { + "epoch": 7.37, + "learning_rate": 7.901892974153622e-07, + "loss": 0.0697, + "step": 40519 + }, + { + "epoch": 7.37, + "learning_rate": 7.901346923917e-07, + "loss": 0.4695, + "step": 40520 + }, + { + "epoch": 7.37, + "learning_rate": 7.90080087368038e-07, + "loss": 0.0198, + "step": 40521 + }, + { + "epoch": 7.37, + "learning_rate": 7.900254823443757e-07, + "loss": 0.1796, + "step": 40522 + }, + { + "epoch": 7.37, + "learning_rate": 7.899708773207134e-07, + "loss": 0.0039, + "step": 40523 + }, + { + "epoch": 7.37, + "learning_rate": 7.899162722970513e-07, + "loss": 0.7737, + "step": 40524 + }, + { + "epoch": 7.37, + "learning_rate": 7.898616672733892e-07, + "loss": 0.1441, + "step": 40525 + }, + { + "epoch": 7.37, + "learning_rate": 7.89807062249727e-07, + "loss": 0.1702, + "step": 40526 + }, + { + "epoch": 7.37, + "learning_rate": 7.897524572260648e-07, + "loss": 0.4183, + "step": 40527 + }, + { + "epoch": 7.37, + "learning_rate": 7.896978522024027e-07, + "loss": 0.0082, + "step": 40528 + }, + { + "epoch": 7.37, + "learning_rate": 7.896432471787405e-07, + "loss": 0.3616, + "step": 40529 + }, + { + "epoch": 7.37, + "learning_rate": 7.895886421550783e-07, + "loss": 0.0919, + "step": 40530 + }, + { + "epoch": 7.37, + "learning_rate": 7.89534037131416e-07, + "loss": 0.5543, + "step": 40531 + }, + { + "epoch": 7.37, + "learning_rate": 7.89479432107754e-07, + "loss": 0.01, + "step": 40532 + }, + { + "epoch": 7.37, + "learning_rate": 7.894248270840917e-07, + "loss": 0.014, + "step": 40533 + }, + { + "epoch": 7.37, + "learning_rate": 7.893702220604295e-07, + "loss": 0.9838, + "step": 40534 + }, + { + "epoch": 7.37, + "learning_rate": 7.893156170367675e-07, + "loss": 0.1284, + "step": 40535 + }, + { + "epoch": 7.37, + "learning_rate": 7.892610120131052e-07, + "loss": 0.0406, + "step": 40536 + }, + { + "epoch": 7.37, + "learning_rate": 7.892064069894431e-07, + "loss": 0.3188, + "step": 40537 + }, + { + "epoch": 7.37, + "learning_rate": 7.891518019657808e-07, + "loss": 0.1171, + "step": 40538 + }, + { + "epoch": 7.37, + "learning_rate": 7.890971969421187e-07, + "loss": 1.2064, + "step": 40539 + }, + { + "epoch": 7.37, + "learning_rate": 7.890425919184565e-07, + "loss": 0.298, + "step": 40540 + }, + { + "epoch": 7.37, + "learning_rate": 7.889879868947943e-07, + "loss": 0.0051, + "step": 40541 + }, + { + "epoch": 7.37, + "learning_rate": 7.889333818711322e-07, + "loss": 0.0136, + "step": 40542 + }, + { + "epoch": 7.37, + "learning_rate": 7.8887877684747e-07, + "loss": 0.86, + "step": 40543 + }, + { + "epoch": 7.37, + "learning_rate": 7.888241718238078e-07, + "loss": 0.2915, + "step": 40544 + }, + { + "epoch": 7.37, + "learning_rate": 7.887695668001455e-07, + "loss": 0.123, + "step": 40545 + }, + { + "epoch": 7.37, + "learning_rate": 7.887149617764835e-07, + "loss": 0.0796, + "step": 40546 + }, + { + "epoch": 7.37, + "learning_rate": 7.886603567528213e-07, + "loss": 0.1927, + "step": 40547 + }, + { + "epoch": 7.37, + "learning_rate": 7.886057517291591e-07, + "loss": 1.473, + "step": 40548 + }, + { + "epoch": 7.37, + "learning_rate": 7.88551146705497e-07, + "loss": 0.1319, + "step": 40549 + }, + { + "epoch": 7.37, + "learning_rate": 7.884965416818347e-07, + "loss": 0.2806, + "step": 40550 + }, + { + "epoch": 7.37, + "learning_rate": 7.884419366581726e-07, + "loss": 0.3496, + "step": 40551 + }, + { + "epoch": 7.37, + "learning_rate": 7.883873316345103e-07, + "loss": 0.8354, + "step": 40552 + }, + { + "epoch": 7.37, + "learning_rate": 7.883327266108482e-07, + "loss": 0.0739, + "step": 40553 + }, + { + "epoch": 7.37, + "learning_rate": 7.882781215871861e-07, + "loss": 0.5517, + "step": 40554 + }, + { + "epoch": 7.37, + "learning_rate": 7.882235165635238e-07, + "loss": 0.0207, + "step": 40555 + }, + { + "epoch": 7.38, + "learning_rate": 7.881689115398618e-07, + "loss": 0.5049, + "step": 40556 + }, + { + "epoch": 7.38, + "learning_rate": 7.881143065161995e-07, + "loss": 0.3774, + "step": 40557 + }, + { + "epoch": 7.38, + "learning_rate": 7.880597014925373e-07, + "loss": 0.0213, + "step": 40558 + }, + { + "epoch": 7.38, + "learning_rate": 7.880050964688751e-07, + "loss": 0.3779, + "step": 40559 + }, + { + "epoch": 7.38, + "learning_rate": 7.87950491445213e-07, + "loss": 0.0851, + "step": 40560 + }, + { + "epoch": 7.38, + "learning_rate": 7.878958864215508e-07, + "loss": 0.2402, + "step": 40561 + }, + { + "epoch": 7.38, + "learning_rate": 7.878412813978886e-07, + "loss": 0.5142, + "step": 40562 + }, + { + "epoch": 7.38, + "learning_rate": 7.877866763742265e-07, + "loss": 1.3811, + "step": 40563 + }, + { + "epoch": 7.38, + "learning_rate": 7.877320713505642e-07, + "loss": 0.5434, + "step": 40564 + }, + { + "epoch": 7.38, + "learning_rate": 7.876774663269021e-07, + "loss": 0.047, + "step": 40565 + }, + { + "epoch": 7.38, + "learning_rate": 7.876228613032398e-07, + "loss": 0.183, + "step": 40566 + }, + { + "epoch": 7.38, + "learning_rate": 7.875682562795778e-07, + "loss": 1.9747, + "step": 40567 + }, + { + "epoch": 7.38, + "learning_rate": 7.875136512559156e-07, + "loss": 1.5372, + "step": 40568 + }, + { + "epoch": 7.38, + "learning_rate": 7.874590462322533e-07, + "loss": 1.1596, + "step": 40569 + }, + { + "epoch": 7.38, + "learning_rate": 7.874044412085913e-07, + "loss": 0.2165, + "step": 40570 + }, + { + "epoch": 7.38, + "learning_rate": 7.873498361849291e-07, + "loss": 0.0463, + "step": 40571 + }, + { + "epoch": 7.38, + "learning_rate": 7.872952311612668e-07, + "loss": 0.1795, + "step": 40572 + }, + { + "epoch": 7.38, + "learning_rate": 7.872406261376046e-07, + "loss": 0.0732, + "step": 40573 + }, + { + "epoch": 7.38, + "learning_rate": 7.871860211139425e-07, + "loss": 0.0029, + "step": 40574 + }, + { + "epoch": 7.38, + "learning_rate": 7.871314160902804e-07, + "loss": 0.7758, + "step": 40575 + }, + { + "epoch": 7.38, + "learning_rate": 7.870768110666181e-07, + "loss": 0.2602, + "step": 40576 + }, + { + "epoch": 7.38, + "learning_rate": 7.87022206042956e-07, + "loss": 0.0049, + "step": 40577 + }, + { + "epoch": 7.38, + "learning_rate": 7.869676010192939e-07, + "loss": 0.6117, + "step": 40578 + }, + { + "epoch": 7.38, + "learning_rate": 7.869129959956316e-07, + "loss": 0.0181, + "step": 40579 + }, + { + "epoch": 7.38, + "learning_rate": 7.868583909719693e-07, + "loss": 1.2493, + "step": 40580 + }, + { + "epoch": 7.38, + "learning_rate": 7.868037859483073e-07, + "loss": 1.4052, + "step": 40581 + }, + { + "epoch": 7.38, + "learning_rate": 7.867491809246451e-07, + "loss": 0.0082, + "step": 40582 + }, + { + "epoch": 7.38, + "learning_rate": 7.866945759009828e-07, + "loss": 1.1302, + "step": 40583 + }, + { + "epoch": 7.38, + "learning_rate": 7.866399708773208e-07, + "loss": 1.2947, + "step": 40584 + }, + { + "epoch": 7.38, + "learning_rate": 7.865853658536586e-07, + "loss": 0.055, + "step": 40585 + }, + { + "epoch": 7.38, + "learning_rate": 7.865307608299964e-07, + "loss": 0.0854, + "step": 40586 + }, + { + "epoch": 7.38, + "learning_rate": 7.864761558063342e-07, + "loss": 0.0226, + "step": 40587 + }, + { + "epoch": 7.38, + "learning_rate": 7.86421550782672e-07, + "loss": 0.1913, + "step": 40588 + }, + { + "epoch": 7.38, + "learning_rate": 7.863669457590099e-07, + "loss": 0.0814, + "step": 40589 + }, + { + "epoch": 7.38, + "learning_rate": 7.863123407353476e-07, + "loss": 0.6099, + "step": 40590 + }, + { + "epoch": 7.38, + "learning_rate": 7.862577357116855e-07, + "loss": 0.0073, + "step": 40591 + }, + { + "epoch": 7.38, + "learning_rate": 7.862031306880234e-07, + "loss": 0.755, + "step": 40592 + }, + { + "epoch": 7.38, + "learning_rate": 7.861485256643611e-07, + "loss": 0.7844, + "step": 40593 + }, + { + "epoch": 7.38, + "learning_rate": 7.86093920640699e-07, + "loss": 0.0439, + "step": 40594 + }, + { + "epoch": 7.38, + "learning_rate": 7.860393156170369e-07, + "loss": 0.0054, + "step": 40595 + }, + { + "epoch": 7.38, + "learning_rate": 7.859847105933746e-07, + "loss": 0.1076, + "step": 40596 + }, + { + "epoch": 7.38, + "learning_rate": 7.859301055697124e-07, + "loss": 0.8826, + "step": 40597 + }, + { + "epoch": 7.38, + "learning_rate": 7.858755005460502e-07, + "loss": 0.0181, + "step": 40598 + }, + { + "epoch": 7.38, + "learning_rate": 7.858208955223881e-07, + "loss": 1.3103, + "step": 40599 + }, + { + "epoch": 7.38, + "learning_rate": 7.857662904987259e-07, + "loss": 0.8955, + "step": 40600 + }, + { + "epoch": 7.38, + "learning_rate": 7.857116854750637e-07, + "loss": 3.7994, + "step": 40601 + }, + { + "epoch": 7.38, + "learning_rate": 7.856570804514016e-07, + "loss": 0.1552, + "step": 40602 + }, + { + "epoch": 7.38, + "learning_rate": 7.856024754277394e-07, + "loss": 0.0431, + "step": 40603 + }, + { + "epoch": 7.38, + "learning_rate": 7.855478704040771e-07, + "loss": 2.313, + "step": 40604 + }, + { + "epoch": 7.38, + "learning_rate": 7.85493265380415e-07, + "loss": 0.2361, + "step": 40605 + }, + { + "epoch": 7.38, + "learning_rate": 7.854386603567529e-07, + "loss": 0.18, + "step": 40606 + }, + { + "epoch": 7.38, + "learning_rate": 7.853840553330906e-07, + "loss": 0.0451, + "step": 40607 + }, + { + "epoch": 7.38, + "learning_rate": 7.853294503094285e-07, + "loss": 0.4855, + "step": 40608 + }, + { + "epoch": 7.38, + "learning_rate": 7.852748452857664e-07, + "loss": 0.0747, + "step": 40609 + }, + { + "epoch": 7.38, + "learning_rate": 7.852202402621041e-07, + "loss": 0.1728, + "step": 40610 + }, + { + "epoch": 7.39, + "learning_rate": 7.85165635238442e-07, + "loss": 0.2993, + "step": 40611 + }, + { + "epoch": 7.39, + "learning_rate": 7.851110302147797e-07, + "loss": 1.1481, + "step": 40612 + }, + { + "epoch": 7.39, + "learning_rate": 7.850564251911177e-07, + "loss": 0.5129, + "step": 40613 + }, + { + "epoch": 7.39, + "learning_rate": 7.850018201674554e-07, + "loss": 0.0744, + "step": 40614 + }, + { + "epoch": 7.39, + "learning_rate": 7.849472151437932e-07, + "loss": 0.1638, + "step": 40615 + }, + { + "epoch": 7.39, + "learning_rate": 7.848926101201312e-07, + "loss": 2.3794, + "step": 40616 + }, + { + "epoch": 7.39, + "learning_rate": 7.848380050964689e-07, + "loss": 0.2691, + "step": 40617 + }, + { + "epoch": 7.39, + "learning_rate": 7.847834000728067e-07, + "loss": 0.2509, + "step": 40618 + }, + { + "epoch": 7.39, + "learning_rate": 7.847287950491445e-07, + "loss": 0.0837, + "step": 40619 + }, + { + "epoch": 7.39, + "learning_rate": 7.846741900254824e-07, + "loss": 0.4955, + "step": 40620 + }, + { + "epoch": 7.39, + "learning_rate": 7.846195850018201e-07, + "loss": 0.0502, + "step": 40621 + }, + { + "epoch": 7.39, + "learning_rate": 7.84564979978158e-07, + "loss": 0.0177, + "step": 40622 + }, + { + "epoch": 7.39, + "learning_rate": 7.845103749544959e-07, + "loss": 0.0057, + "step": 40623 + }, + { + "epoch": 7.39, + "learning_rate": 7.844557699308337e-07, + "loss": 0.0128, + "step": 40624 + }, + { + "epoch": 7.39, + "learning_rate": 7.844011649071715e-07, + "loss": 0.2024, + "step": 40625 + }, + { + "epoch": 7.39, + "learning_rate": 7.843465598835092e-07, + "loss": 0.0167, + "step": 40626 + }, + { + "epoch": 7.39, + "learning_rate": 7.842919548598472e-07, + "loss": 1.4331, + "step": 40627 + }, + { + "epoch": 7.39, + "learning_rate": 7.842373498361849e-07, + "loss": 0.0293, + "step": 40628 + }, + { + "epoch": 7.39, + "learning_rate": 7.841827448125227e-07, + "loss": 0.6689, + "step": 40629 + }, + { + "epoch": 7.39, + "learning_rate": 7.841281397888607e-07, + "loss": 0.1873, + "step": 40630 + }, + { + "epoch": 7.39, + "learning_rate": 7.840735347651984e-07, + "loss": 0.2772, + "step": 40631 + }, + { + "epoch": 7.39, + "learning_rate": 7.840189297415362e-07, + "loss": 0.2227, + "step": 40632 + }, + { + "epoch": 7.39, + "learning_rate": 7.83964324717874e-07, + "loss": 0.3537, + "step": 40633 + }, + { + "epoch": 7.39, + "learning_rate": 7.839097196942119e-07, + "loss": 0.148, + "step": 40634 + }, + { + "epoch": 7.39, + "learning_rate": 7.838551146705498e-07, + "loss": 0.0107, + "step": 40635 + }, + { + "epoch": 7.39, + "learning_rate": 7.838005096468875e-07, + "loss": 0.2728, + "step": 40636 + }, + { + "epoch": 7.39, + "learning_rate": 7.837459046232254e-07, + "loss": 1.719, + "step": 40637 + }, + { + "epoch": 7.39, + "learning_rate": 7.836912995995632e-07, + "loss": 1.0307, + "step": 40638 + }, + { + "epoch": 7.39, + "learning_rate": 7.83636694575901e-07, + "loss": 1.7952, + "step": 40639 + }, + { + "epoch": 7.39, + "learning_rate": 7.835820895522387e-07, + "loss": 0.2725, + "step": 40640 + }, + { + "epoch": 7.39, + "learning_rate": 7.835274845285767e-07, + "loss": 0.4249, + "step": 40641 + }, + { + "epoch": 7.39, + "learning_rate": 7.834728795049145e-07, + "loss": 0.226, + "step": 40642 + }, + { + "epoch": 7.39, + "learning_rate": 7.834182744812522e-07, + "loss": 0.2041, + "step": 40643 + }, + { + "epoch": 7.39, + "learning_rate": 7.833636694575902e-07, + "loss": 0.5245, + "step": 40644 + }, + { + "epoch": 7.39, + "learning_rate": 7.833090644339279e-07, + "loss": 2.0791, + "step": 40645 + }, + { + "epoch": 7.39, + "learning_rate": 7.832544594102658e-07, + "loss": 0.3749, + "step": 40646 + }, + { + "epoch": 7.39, + "learning_rate": 7.831998543866035e-07, + "loss": 1.5514, + "step": 40647 + }, + { + "epoch": 7.39, + "learning_rate": 7.831452493629414e-07, + "loss": 0.9458, + "step": 40648 + }, + { + "epoch": 7.39, + "learning_rate": 7.830906443392793e-07, + "loss": 0.0099, + "step": 40649 + }, + { + "epoch": 7.39, + "learning_rate": 7.83036039315617e-07, + "loss": 0.0555, + "step": 40650 + }, + { + "epoch": 7.39, + "learning_rate": 7.829814342919549e-07, + "loss": 1.5803, + "step": 40651 + }, + { + "epoch": 7.39, + "learning_rate": 7.829268292682928e-07, + "loss": 0.5916, + "step": 40652 + }, + { + "epoch": 7.39, + "learning_rate": 7.828722242446305e-07, + "loss": 1.2889, + "step": 40653 + }, + { + "epoch": 7.39, + "learning_rate": 7.828176192209682e-07, + "loss": 0.0281, + "step": 40654 + }, + { + "epoch": 7.39, + "learning_rate": 7.827630141973062e-07, + "loss": 0.0104, + "step": 40655 + }, + { + "epoch": 7.39, + "learning_rate": 7.82708409173644e-07, + "loss": 0.2495, + "step": 40656 + }, + { + "epoch": 7.39, + "learning_rate": 7.826538041499818e-07, + "loss": 0.1308, + "step": 40657 + }, + { + "epoch": 7.39, + "learning_rate": 7.825991991263197e-07, + "loss": 0.3943, + "step": 40658 + }, + { + "epoch": 7.39, + "learning_rate": 7.825445941026575e-07, + "loss": 0.1774, + "step": 40659 + }, + { + "epoch": 7.39, + "learning_rate": 7.824899890789953e-07, + "loss": 0.1027, + "step": 40660 + }, + { + "epoch": 7.39, + "learning_rate": 7.82435384055333e-07, + "loss": 0.2065, + "step": 40661 + }, + { + "epoch": 7.39, + "learning_rate": 7.82380779031671e-07, + "loss": 0.1396, + "step": 40662 + }, + { + "epoch": 7.39, + "learning_rate": 7.823261740080088e-07, + "loss": 0.2887, + "step": 40663 + }, + { + "epoch": 7.39, + "learning_rate": 7.822715689843465e-07, + "loss": 0.006, + "step": 40664 + }, + { + "epoch": 7.39, + "learning_rate": 7.822169639606845e-07, + "loss": 0.0232, + "step": 40665 + }, + { + "epoch": 7.4, + "learning_rate": 7.821623589370223e-07, + "loss": 0.1283, + "step": 40666 + }, + { + "epoch": 7.4, + "learning_rate": 7.8210775391336e-07, + "loss": 0.0765, + "step": 40667 + }, + { + "epoch": 7.4, + "learning_rate": 7.820531488896978e-07, + "loss": 0.1805, + "step": 40668 + }, + { + "epoch": 7.4, + "learning_rate": 7.819985438660357e-07, + "loss": 0.0072, + "step": 40669 + }, + { + "epoch": 7.4, + "learning_rate": 7.819439388423735e-07, + "loss": 0.0156, + "step": 40670 + }, + { + "epoch": 7.4, + "learning_rate": 7.818893338187113e-07, + "loss": 0.0037, + "step": 40671 + }, + { + "epoch": 7.4, + "learning_rate": 7.818347287950492e-07, + "loss": 0.3739, + "step": 40672 + }, + { + "epoch": 7.4, + "learning_rate": 7.817801237713871e-07, + "loss": 0.0486, + "step": 40673 + }, + { + "epoch": 7.4, + "learning_rate": 7.817255187477248e-07, + "loss": 0.9715, + "step": 40674 + }, + { + "epoch": 7.4, + "learning_rate": 7.816709137240626e-07, + "loss": 0.0188, + "step": 40675 + }, + { + "epoch": 7.4, + "learning_rate": 7.816163087004006e-07, + "loss": 1.9578, + "step": 40676 + }, + { + "epoch": 7.4, + "learning_rate": 7.815617036767383e-07, + "loss": 0.2552, + "step": 40677 + }, + { + "epoch": 7.4, + "learning_rate": 7.81507098653076e-07, + "loss": 0.1854, + "step": 40678 + }, + { + "epoch": 7.4, + "learning_rate": 7.814524936294139e-07, + "loss": 0.6461, + "step": 40679 + }, + { + "epoch": 7.4, + "learning_rate": 7.813978886057518e-07, + "loss": 1.3688, + "step": 40680 + }, + { + "epoch": 7.4, + "learning_rate": 7.813432835820895e-07, + "loss": 0.0859, + "step": 40681 + }, + { + "epoch": 7.4, + "learning_rate": 7.812886785584274e-07, + "loss": 0.8193, + "step": 40682 + }, + { + "epoch": 7.4, + "learning_rate": 7.812340735347653e-07, + "loss": 0.7366, + "step": 40683 + }, + { + "epoch": 7.4, + "learning_rate": 7.811794685111031e-07, + "loss": 0.0865, + "step": 40684 + }, + { + "epoch": 7.4, + "learning_rate": 7.811248634874408e-07, + "loss": 0.0151, + "step": 40685 + }, + { + "epoch": 7.4, + "learning_rate": 7.810702584637786e-07, + "loss": 0.6959, + "step": 40686 + }, + { + "epoch": 7.4, + "learning_rate": 7.810156534401166e-07, + "loss": 1.1965, + "step": 40687 + }, + { + "epoch": 7.4, + "learning_rate": 7.809610484164543e-07, + "loss": 0.5116, + "step": 40688 + }, + { + "epoch": 7.4, + "learning_rate": 7.809064433927921e-07, + "loss": 0.0214, + "step": 40689 + }, + { + "epoch": 7.4, + "learning_rate": 7.808518383691301e-07, + "loss": 0.1982, + "step": 40690 + }, + { + "epoch": 7.4, + "learning_rate": 7.807972333454678e-07, + "loss": 0.6872, + "step": 40691 + }, + { + "epoch": 7.4, + "learning_rate": 7.807426283218055e-07, + "loss": 0.7124, + "step": 40692 + }, + { + "epoch": 7.4, + "learning_rate": 7.806880232981434e-07, + "loss": 0.0782, + "step": 40693 + }, + { + "epoch": 7.4, + "learning_rate": 7.806334182744813e-07, + "loss": 2.5072, + "step": 40694 + }, + { + "epoch": 7.4, + "learning_rate": 7.805788132508191e-07, + "loss": 0.2181, + "step": 40695 + }, + { + "epoch": 7.4, + "learning_rate": 7.805242082271569e-07, + "loss": 1.2453, + "step": 40696 + }, + { + "epoch": 7.4, + "learning_rate": 7.804696032034948e-07, + "loss": 0.1518, + "step": 40697 + }, + { + "epoch": 7.4, + "learning_rate": 7.804149981798326e-07, + "loss": 0.9021, + "step": 40698 + }, + { + "epoch": 7.4, + "learning_rate": 7.803603931561704e-07, + "loss": 0.1583, + "step": 40699 + }, + { + "epoch": 7.4, + "learning_rate": 7.803057881325081e-07, + "loss": 2.0919, + "step": 40700 + }, + { + "epoch": 7.4, + "learning_rate": 7.802511831088461e-07, + "loss": 1.1979, + "step": 40701 + }, + { + "epoch": 7.4, + "learning_rate": 7.801965780851838e-07, + "loss": 0.2532, + "step": 40702 + }, + { + "epoch": 7.4, + "learning_rate": 7.801419730615217e-07, + "loss": 0.7451, + "step": 40703 + }, + { + "epoch": 7.4, + "learning_rate": 7.800873680378596e-07, + "loss": 0.134, + "step": 40704 + }, + { + "epoch": 7.4, + "learning_rate": 7.800327630141973e-07, + "loss": 1.4157, + "step": 40705 + }, + { + "epoch": 7.4, + "learning_rate": 7.799781579905352e-07, + "loss": 0.0078, + "step": 40706 + }, + { + "epoch": 7.4, + "learning_rate": 7.799235529668729e-07, + "loss": 0.6936, + "step": 40707 + }, + { + "epoch": 7.4, + "learning_rate": 7.798689479432108e-07, + "loss": 0.0133, + "step": 40708 + }, + { + "epoch": 7.4, + "learning_rate": 7.798143429195486e-07, + "loss": 0.5127, + "step": 40709 + }, + { + "epoch": 7.4, + "learning_rate": 7.797597378958864e-07, + "loss": 0.4081, + "step": 40710 + }, + { + "epoch": 7.4, + "learning_rate": 7.797051328722244e-07, + "loss": 1.2613, + "step": 40711 + }, + { + "epoch": 7.4, + "learning_rate": 7.796505278485621e-07, + "loss": 1.8447, + "step": 40712 + }, + { + "epoch": 7.4, + "learning_rate": 7.795959228248999e-07, + "loss": 0.799, + "step": 40713 + }, + { + "epoch": 7.4, + "learning_rate": 7.795413178012377e-07, + "loss": 0.0141, + "step": 40714 + }, + { + "epoch": 7.4, + "learning_rate": 7.794867127775756e-07, + "loss": 0.1277, + "step": 40715 + }, + { + "epoch": 7.4, + "learning_rate": 7.794321077539134e-07, + "loss": 0.0714, + "step": 40716 + }, + { + "epoch": 7.4, + "learning_rate": 7.793775027302512e-07, + "loss": 0.3525, + "step": 40717 + }, + { + "epoch": 7.4, + "learning_rate": 7.793228977065891e-07, + "loss": 0.5905, + "step": 40718 + }, + { + "epoch": 7.4, + "learning_rate": 7.792682926829268e-07, + "loss": 0.2068, + "step": 40719 + }, + { + "epoch": 7.4, + "learning_rate": 7.792136876592647e-07, + "loss": 0.1828, + "step": 40720 + }, + { + "epoch": 7.41, + "learning_rate": 7.791590826356024e-07, + "loss": 1.5911, + "step": 40721 + }, + { + "epoch": 7.41, + "learning_rate": 7.791044776119404e-07, + "loss": 0.6087, + "step": 40722 + }, + { + "epoch": 7.41, + "learning_rate": 7.790498725882782e-07, + "loss": 0.0101, + "step": 40723 + }, + { + "epoch": 7.41, + "learning_rate": 7.789952675646159e-07, + "loss": 0.0295, + "step": 40724 + }, + { + "epoch": 7.41, + "learning_rate": 7.789406625409539e-07, + "loss": 0.035, + "step": 40725 + }, + { + "epoch": 7.41, + "learning_rate": 7.788860575172916e-07, + "loss": 0.4553, + "step": 40726 + }, + { + "epoch": 7.41, + "learning_rate": 7.788314524936294e-07, + "loss": 3.5165, + "step": 40727 + }, + { + "epoch": 7.41, + "learning_rate": 7.787768474699672e-07, + "loss": 2.0587, + "step": 40728 + }, + { + "epoch": 7.41, + "learning_rate": 7.787222424463051e-07, + "loss": 0.0883, + "step": 40729 + }, + { + "epoch": 7.41, + "learning_rate": 7.786676374226429e-07, + "loss": 0.0075, + "step": 40730 + }, + { + "epoch": 7.41, + "learning_rate": 7.786130323989807e-07, + "loss": 0.8772, + "step": 40731 + }, + { + "epoch": 7.41, + "learning_rate": 7.785584273753186e-07, + "loss": 3.0669, + "step": 40732 + }, + { + "epoch": 7.41, + "learning_rate": 7.785038223516564e-07, + "loss": 0.1832, + "step": 40733 + }, + { + "epoch": 7.41, + "learning_rate": 7.784492173279942e-07, + "loss": 0.809, + "step": 40734 + }, + { + "epoch": 7.41, + "learning_rate": 7.783946123043319e-07, + "loss": 0.0632, + "step": 40735 + }, + { + "epoch": 7.41, + "learning_rate": 7.783400072806699e-07, + "loss": 0.0164, + "step": 40736 + }, + { + "epoch": 7.41, + "learning_rate": 7.782854022570077e-07, + "loss": 0.2366, + "step": 40737 + }, + { + "epoch": 7.41, + "learning_rate": 7.782307972333454e-07, + "loss": 0.008, + "step": 40738 + }, + { + "epoch": 7.41, + "learning_rate": 7.781761922096834e-07, + "loss": 2.4437, + "step": 40739 + }, + { + "epoch": 7.41, + "learning_rate": 7.781215871860212e-07, + "loss": 0.3967, + "step": 40740 + }, + { + "epoch": 7.41, + "learning_rate": 7.780669821623589e-07, + "loss": 1.1621, + "step": 40741 + }, + { + "epoch": 7.41, + "learning_rate": 7.780123771386967e-07, + "loss": 0.1018, + "step": 40742 + }, + { + "epoch": 7.41, + "learning_rate": 7.779577721150346e-07, + "loss": 1.7063, + "step": 40743 + }, + { + "epoch": 7.41, + "learning_rate": 7.779031670913725e-07, + "loss": 0.59, + "step": 40744 + }, + { + "epoch": 7.41, + "learning_rate": 7.778485620677102e-07, + "loss": 0.009, + "step": 40745 + }, + { + "epoch": 7.41, + "learning_rate": 7.777939570440481e-07, + "loss": 0.0119, + "step": 40746 + }, + { + "epoch": 7.41, + "learning_rate": 7.77739352020386e-07, + "loss": 1.2136, + "step": 40747 + }, + { + "epoch": 7.41, + "learning_rate": 7.776847469967237e-07, + "loss": 3.4513, + "step": 40748 + }, + { + "epoch": 7.41, + "learning_rate": 7.776301419730614e-07, + "loss": 0.4602, + "step": 40749 + }, + { + "epoch": 7.41, + "learning_rate": 7.775755369493994e-07, + "loss": 0.4177, + "step": 40750 + }, + { + "epoch": 7.41, + "learning_rate": 7.775209319257372e-07, + "loss": 0.0848, + "step": 40751 + }, + { + "epoch": 7.41, + "learning_rate": 7.77466326902075e-07, + "loss": 2.7861, + "step": 40752 + }, + { + "epoch": 7.41, + "learning_rate": 7.774117218784128e-07, + "loss": 0.0318, + "step": 40753 + }, + { + "epoch": 7.41, + "learning_rate": 7.773571168547507e-07, + "loss": 0.8324, + "step": 40754 + }, + { + "epoch": 7.41, + "learning_rate": 7.773025118310885e-07, + "loss": 0.0133, + "step": 40755 + }, + { + "epoch": 7.41, + "learning_rate": 7.772479068074263e-07, + "loss": 0.1582, + "step": 40756 + }, + { + "epoch": 7.41, + "learning_rate": 7.771933017837641e-07, + "loss": 0.0056, + "step": 40757 + }, + { + "epoch": 7.41, + "learning_rate": 7.77138696760102e-07, + "loss": 1.8613, + "step": 40758 + }, + { + "epoch": 7.41, + "learning_rate": 7.770840917364397e-07, + "loss": 0.3755, + "step": 40759 + }, + { + "epoch": 7.41, + "learning_rate": 7.770294867127775e-07, + "loss": 2.9039, + "step": 40760 + }, + { + "epoch": 7.41, + "learning_rate": 7.769748816891155e-07, + "loss": 0.1129, + "step": 40761 + }, + { + "epoch": 7.41, + "learning_rate": 7.769202766654532e-07, + "loss": 0.0559, + "step": 40762 + }, + { + "epoch": 7.41, + "learning_rate": 7.768656716417911e-07, + "loss": 0.0531, + "step": 40763 + }, + { + "epoch": 7.41, + "learning_rate": 7.76811066618129e-07, + "loss": 0.136, + "step": 40764 + }, + { + "epoch": 7.41, + "learning_rate": 7.767564615944667e-07, + "loss": 0.4894, + "step": 40765 + }, + { + "epoch": 7.41, + "learning_rate": 7.767018565708045e-07, + "loss": 0.0147, + "step": 40766 + }, + { + "epoch": 7.41, + "learning_rate": 7.766472515471423e-07, + "loss": 0.3576, + "step": 40767 + }, + { + "epoch": 7.41, + "learning_rate": 7.765926465234802e-07, + "loss": 0.014, + "step": 40768 + }, + { + "epoch": 7.41, + "learning_rate": 7.76538041499818e-07, + "loss": 0.7071, + "step": 40769 + }, + { + "epoch": 7.41, + "learning_rate": 7.764834364761558e-07, + "loss": 0.1597, + "step": 40770 + }, + { + "epoch": 7.41, + "learning_rate": 7.764288314524938e-07, + "loss": 0.0942, + "step": 40771 + }, + { + "epoch": 7.41, + "learning_rate": 7.763742264288315e-07, + "loss": 0.1375, + "step": 40772 + }, + { + "epoch": 7.41, + "learning_rate": 7.763196214051692e-07, + "loss": 0.8482, + "step": 40773 + }, + { + "epoch": 7.41, + "learning_rate": 7.762650163815071e-07, + "loss": 0.7928, + "step": 40774 + }, + { + "epoch": 7.41, + "learning_rate": 7.76210411357845e-07, + "loss": 0.8152, + "step": 40775 + }, + { + "epoch": 7.42, + "learning_rate": 7.761558063341827e-07, + "loss": 1.2369, + "step": 40776 + }, + { + "epoch": 7.42, + "learning_rate": 7.761012013105206e-07, + "loss": 0.0627, + "step": 40777 + }, + { + "epoch": 7.42, + "learning_rate": 7.760465962868585e-07, + "loss": 0.3055, + "step": 40778 + }, + { + "epoch": 7.42, + "learning_rate": 7.759919912631962e-07, + "loss": 0.0343, + "step": 40779 + }, + { + "epoch": 7.42, + "learning_rate": 7.759373862395341e-07, + "loss": 0.0342, + "step": 40780 + }, + { + "epoch": 7.42, + "learning_rate": 7.758827812158718e-07, + "loss": 0.0068, + "step": 40781 + }, + { + "epoch": 7.42, + "learning_rate": 7.758281761922098e-07, + "loss": 0.0345, + "step": 40782 + }, + { + "epoch": 7.42, + "learning_rate": 7.757735711685475e-07, + "loss": 0.3175, + "step": 40783 + }, + { + "epoch": 7.42, + "learning_rate": 7.757189661448853e-07, + "loss": 0.4357, + "step": 40784 + }, + { + "epoch": 7.42, + "learning_rate": 7.756643611212233e-07, + "loss": 1.4623, + "step": 40785 + }, + { + "epoch": 7.42, + "learning_rate": 7.75609756097561e-07, + "loss": 0.451, + "step": 40786 + }, + { + "epoch": 7.42, + "learning_rate": 7.755551510738988e-07, + "loss": 0.5877, + "step": 40787 + }, + { + "epoch": 7.42, + "learning_rate": 7.755005460502366e-07, + "loss": 0.795, + "step": 40788 + }, + { + "epoch": 7.42, + "learning_rate": 7.754459410265745e-07, + "loss": 0.0083, + "step": 40789 + }, + { + "epoch": 7.42, + "learning_rate": 7.753913360029122e-07, + "loss": 0.3177, + "step": 40790 + }, + { + "epoch": 7.42, + "learning_rate": 7.753367309792501e-07, + "loss": 1.0493, + "step": 40791 + }, + { + "epoch": 7.42, + "learning_rate": 7.75282125955588e-07, + "loss": 1.1411, + "step": 40792 + }, + { + "epoch": 7.42, + "learning_rate": 7.752275209319258e-07, + "loss": 0.9115, + "step": 40793 + }, + { + "epoch": 7.42, + "learning_rate": 7.751729159082636e-07, + "loss": 1.784, + "step": 40794 + }, + { + "epoch": 7.42, + "learning_rate": 7.751183108846013e-07, + "loss": 0.3902, + "step": 40795 + }, + { + "epoch": 7.42, + "learning_rate": 7.750637058609393e-07, + "loss": 0.4952, + "step": 40796 + }, + { + "epoch": 7.42, + "learning_rate": 7.75009100837277e-07, + "loss": 1.516, + "step": 40797 + }, + { + "epoch": 7.42, + "learning_rate": 7.749544958136148e-07, + "loss": 0.0502, + "step": 40798 + }, + { + "epoch": 7.42, + "learning_rate": 7.748998907899528e-07, + "loss": 0.0215, + "step": 40799 + }, + { + "epoch": 7.42, + "learning_rate": 7.748452857662905e-07, + "loss": 0.1479, + "step": 40800 + }, + { + "epoch": 7.42, + "learning_rate": 7.747906807426284e-07, + "loss": 0.1331, + "step": 40801 + }, + { + "epoch": 7.42, + "learning_rate": 7.747360757189661e-07, + "loss": 0.4506, + "step": 40802 + }, + { + "epoch": 7.42, + "learning_rate": 7.74681470695304e-07, + "loss": 0.4688, + "step": 40803 + }, + { + "epoch": 7.42, + "learning_rate": 7.746268656716419e-07, + "loss": 0.3669, + "step": 40804 + }, + { + "epoch": 7.42, + "learning_rate": 7.745722606479796e-07, + "loss": 0.1889, + "step": 40805 + }, + { + "epoch": 7.42, + "learning_rate": 7.745176556243175e-07, + "loss": 0.0109, + "step": 40806 + }, + { + "epoch": 7.42, + "learning_rate": 7.744630506006553e-07, + "loss": 1.6597, + "step": 40807 + }, + { + "epoch": 7.42, + "learning_rate": 7.744084455769931e-07, + "loss": 0.5608, + "step": 40808 + }, + { + "epoch": 7.42, + "learning_rate": 7.743538405533308e-07, + "loss": 0.0035, + "step": 40809 + }, + { + "epoch": 7.42, + "learning_rate": 7.742992355296688e-07, + "loss": 0.815, + "step": 40810 + }, + { + "epoch": 7.42, + "learning_rate": 7.742446305060066e-07, + "loss": 0.8003, + "step": 40811 + }, + { + "epoch": 7.42, + "learning_rate": 7.741900254823444e-07, + "loss": 0.0458, + "step": 40812 + }, + { + "epoch": 7.42, + "learning_rate": 7.741354204586823e-07, + "loss": 1.071, + "step": 40813 + }, + { + "epoch": 7.42, + "learning_rate": 7.7408081543502e-07, + "loss": 0.0413, + "step": 40814 + }, + { + "epoch": 7.42, + "learning_rate": 7.740262104113579e-07, + "loss": 0.2707, + "step": 40815 + }, + { + "epoch": 7.42, + "learning_rate": 7.739716053876956e-07, + "loss": 0.0214, + "step": 40816 + }, + { + "epoch": 7.42, + "learning_rate": 7.739170003640335e-07, + "loss": 0.5471, + "step": 40817 + }, + { + "epoch": 7.42, + "learning_rate": 7.738623953403714e-07, + "loss": 0.1135, + "step": 40818 + }, + { + "epoch": 7.42, + "learning_rate": 7.738077903167091e-07, + "loss": 1.8557, + "step": 40819 + }, + { + "epoch": 7.42, + "learning_rate": 7.73753185293047e-07, + "loss": 1.1889, + "step": 40820 + }, + { + "epoch": 7.42, + "learning_rate": 7.736985802693848e-07, + "loss": 0.0055, + "step": 40821 + }, + { + "epoch": 7.42, + "learning_rate": 7.736439752457226e-07, + "loss": 0.0072, + "step": 40822 + }, + { + "epoch": 7.42, + "learning_rate": 7.735893702220604e-07, + "loss": 0.7345, + "step": 40823 + }, + { + "epoch": 7.42, + "learning_rate": 7.735347651983983e-07, + "loss": 0.1207, + "step": 40824 + }, + { + "epoch": 7.42, + "learning_rate": 7.734801601747361e-07, + "loss": 0.1255, + "step": 40825 + }, + { + "epoch": 7.42, + "learning_rate": 7.734255551510739e-07, + "loss": 0.1953, + "step": 40826 + }, + { + "epoch": 7.42, + "learning_rate": 7.733709501274118e-07, + "loss": 0.028, + "step": 40827 + }, + { + "epoch": 7.42, + "learning_rate": 7.733163451037496e-07, + "loss": 3.1768, + "step": 40828 + }, + { + "epoch": 7.42, + "learning_rate": 7.732617400800874e-07, + "loss": 0.6704, + "step": 40829 + }, + { + "epoch": 7.42, + "learning_rate": 7.732071350564251e-07, + "loss": 3.8029, + "step": 40830 + }, + { + "epoch": 7.43, + "learning_rate": 7.73152530032763e-07, + "loss": 0.8994, + "step": 40831 + }, + { + "epoch": 7.43, + "learning_rate": 7.730979250091009e-07, + "loss": 0.2291, + "step": 40832 + }, + { + "epoch": 7.43, + "learning_rate": 7.730433199854386e-07, + "loss": 0.3299, + "step": 40833 + }, + { + "epoch": 7.43, + "learning_rate": 7.729887149617765e-07, + "loss": 0.8149, + "step": 40834 + }, + { + "epoch": 7.43, + "learning_rate": 7.729341099381144e-07, + "loss": 0.1146, + "step": 40835 + }, + { + "epoch": 7.43, + "learning_rate": 7.728795049144521e-07, + "loss": 1.1024, + "step": 40836 + }, + { + "epoch": 7.43, + "learning_rate": 7.728248998907899e-07, + "loss": 0.7643, + "step": 40837 + }, + { + "epoch": 7.43, + "learning_rate": 7.727702948671278e-07, + "loss": 0.7042, + "step": 40838 + }, + { + "epoch": 7.43, + "learning_rate": 7.727156898434656e-07, + "loss": 0.5765, + "step": 40839 + }, + { + "epoch": 7.43, + "learning_rate": 7.726610848198034e-07, + "loss": 0.9672, + "step": 40840 + }, + { + "epoch": 7.43, + "learning_rate": 7.726064797961412e-07, + "loss": 0.7429, + "step": 40841 + }, + { + "epoch": 7.43, + "learning_rate": 7.725518747724792e-07, + "loss": 2.4977, + "step": 40842 + }, + { + "epoch": 7.43, + "learning_rate": 7.724972697488169e-07, + "loss": 0.3054, + "step": 40843 + }, + { + "epoch": 7.43, + "learning_rate": 7.724426647251547e-07, + "loss": 0.1743, + "step": 40844 + }, + { + "epoch": 7.43, + "learning_rate": 7.723880597014927e-07, + "loss": 0.005, + "step": 40845 + }, + { + "epoch": 7.43, + "learning_rate": 7.723334546778304e-07, + "loss": 0.1236, + "step": 40846 + }, + { + "epoch": 7.43, + "learning_rate": 7.722788496541681e-07, + "loss": 0.0764, + "step": 40847 + }, + { + "epoch": 7.43, + "learning_rate": 7.72224244630506e-07, + "loss": 0.058, + "step": 40848 + }, + { + "epoch": 7.43, + "learning_rate": 7.721696396068439e-07, + "loss": 0.0094, + "step": 40849 + }, + { + "epoch": 7.43, + "learning_rate": 7.721150345831816e-07, + "loss": 0.0064, + "step": 40850 + }, + { + "epoch": 7.43, + "learning_rate": 7.720604295595195e-07, + "loss": 0.8825, + "step": 40851 + }, + { + "epoch": 7.43, + "learning_rate": 7.720058245358574e-07, + "loss": 0.7112, + "step": 40852 + }, + { + "epoch": 7.43, + "learning_rate": 7.719512195121952e-07, + "loss": 0.0591, + "step": 40853 + }, + { + "epoch": 7.43, + "learning_rate": 7.718966144885329e-07, + "loss": 0.3299, + "step": 40854 + }, + { + "epoch": 7.43, + "learning_rate": 7.718420094648707e-07, + "loss": 0.0269, + "step": 40855 + }, + { + "epoch": 7.43, + "learning_rate": 7.717874044412087e-07, + "loss": 0.34, + "step": 40856 + }, + { + "epoch": 7.43, + "learning_rate": 7.717327994175464e-07, + "loss": 1.1649, + "step": 40857 + }, + { + "epoch": 7.43, + "learning_rate": 7.716781943938842e-07, + "loss": 0.33, + "step": 40858 + }, + { + "epoch": 7.43, + "learning_rate": 7.716235893702222e-07, + "loss": 0.0938, + "step": 40859 + }, + { + "epoch": 7.43, + "learning_rate": 7.715689843465599e-07, + "loss": 0.0258, + "step": 40860 + }, + { + "epoch": 7.43, + "learning_rate": 7.715143793228976e-07, + "loss": 0.0106, + "step": 40861 + }, + { + "epoch": 7.43, + "learning_rate": 7.714597742992355e-07, + "loss": 1.2189, + "step": 40862 + }, + { + "epoch": 7.43, + "learning_rate": 7.714051692755734e-07, + "loss": 0.5395, + "step": 40863 + }, + { + "epoch": 7.43, + "learning_rate": 7.713505642519112e-07, + "loss": 1.5742, + "step": 40864 + }, + { + "epoch": 7.43, + "learning_rate": 7.71295959228249e-07, + "loss": 0.506, + "step": 40865 + }, + { + "epoch": 7.43, + "learning_rate": 7.712413542045869e-07, + "loss": 0.7329, + "step": 40866 + }, + { + "epoch": 7.43, + "learning_rate": 7.711867491809247e-07, + "loss": 0.6495, + "step": 40867 + }, + { + "epoch": 7.43, + "learning_rate": 7.711321441572625e-07, + "loss": 0.2722, + "step": 40868 + }, + { + "epoch": 7.43, + "learning_rate": 7.710775391336002e-07, + "loss": 1.0519, + "step": 40869 + }, + { + "epoch": 7.43, + "learning_rate": 7.710229341099382e-07, + "loss": 0.84, + "step": 40870 + }, + { + "epoch": 7.43, + "learning_rate": 7.709683290862759e-07, + "loss": 1.2827, + "step": 40871 + }, + { + "epoch": 7.43, + "learning_rate": 7.709137240626138e-07, + "loss": 0.7798, + "step": 40872 + }, + { + "epoch": 7.43, + "learning_rate": 7.708591190389517e-07, + "loss": 1.716, + "step": 40873 + }, + { + "epoch": 7.43, + "learning_rate": 7.708045140152894e-07, + "loss": 0.0419, + "step": 40874 + }, + { + "epoch": 7.43, + "learning_rate": 7.707499089916273e-07, + "loss": 1.6227, + "step": 40875 + }, + { + "epoch": 7.43, + "learning_rate": 7.70695303967965e-07, + "loss": 1.3101, + "step": 40876 + }, + { + "epoch": 7.43, + "learning_rate": 7.706406989443029e-07, + "loss": 0.3598, + "step": 40877 + }, + { + "epoch": 7.43, + "learning_rate": 7.705860939206407e-07, + "loss": 0.56, + "step": 40878 + }, + { + "epoch": 7.43, + "learning_rate": 7.705314888969785e-07, + "loss": 0.1631, + "step": 40879 + }, + { + "epoch": 7.43, + "learning_rate": 7.704768838733165e-07, + "loss": 0.0076, + "step": 40880 + }, + { + "epoch": 7.43, + "learning_rate": 7.704222788496542e-07, + "loss": 0.8643, + "step": 40881 + }, + { + "epoch": 7.43, + "learning_rate": 7.70367673825992e-07, + "loss": 0.0385, + "step": 40882 + }, + { + "epoch": 7.43, + "learning_rate": 7.703130688023298e-07, + "loss": 1.8206, + "step": 40883 + }, + { + "epoch": 7.43, + "learning_rate": 7.702584637786677e-07, + "loss": 0.13, + "step": 40884 + }, + { + "epoch": 7.43, + "learning_rate": 7.702038587550055e-07, + "loss": 0.1692, + "step": 40885 + }, + { + "epoch": 7.44, + "learning_rate": 7.701492537313433e-07, + "loss": 0.1735, + "step": 40886 + }, + { + "epoch": 7.44, + "learning_rate": 7.700946487076812e-07, + "loss": 0.024, + "step": 40887 + }, + { + "epoch": 7.44, + "learning_rate": 7.700400436840189e-07, + "loss": 0.046, + "step": 40888 + }, + { + "epoch": 7.44, + "learning_rate": 7.699854386603568e-07, + "loss": 0.299, + "step": 40889 + }, + { + "epoch": 7.44, + "learning_rate": 7.699308336366945e-07, + "loss": 1.3399, + "step": 40890 + }, + { + "epoch": 7.44, + "learning_rate": 7.698762286130325e-07, + "loss": 0.4469, + "step": 40891 + }, + { + "epoch": 7.44, + "learning_rate": 7.698216235893703e-07, + "loss": 0.5759, + "step": 40892 + }, + { + "epoch": 7.44, + "learning_rate": 7.69767018565708e-07, + "loss": 0.0169, + "step": 40893 + }, + { + "epoch": 7.44, + "learning_rate": 7.69712413542046e-07, + "loss": 1.793, + "step": 40894 + }, + { + "epoch": 7.44, + "learning_rate": 7.696578085183837e-07, + "loss": 0.315, + "step": 40895 + }, + { + "epoch": 7.44, + "learning_rate": 7.696032034947215e-07, + "loss": 0.0031, + "step": 40896 + }, + { + "epoch": 7.44, + "learning_rate": 7.695485984710593e-07, + "loss": 0.118, + "step": 40897 + }, + { + "epoch": 7.44, + "learning_rate": 7.694939934473972e-07, + "loss": 0.2111, + "step": 40898 + }, + { + "epoch": 7.44, + "learning_rate": 7.69439388423735e-07, + "loss": 0.01, + "step": 40899 + }, + { + "epoch": 7.44, + "learning_rate": 7.693847834000728e-07, + "loss": 0.0863, + "step": 40900 + }, + { + "epoch": 7.44, + "learning_rate": 7.693301783764107e-07, + "loss": 0.5138, + "step": 40901 + }, + { + "epoch": 7.44, + "learning_rate": 7.692755733527485e-07, + "loss": 0.6753, + "step": 40902 + }, + { + "epoch": 7.44, + "learning_rate": 7.692209683290863e-07, + "loss": 0.3976, + "step": 40903 + }, + { + "epoch": 7.44, + "learning_rate": 7.69166363305424e-07, + "loss": 0.0083, + "step": 40904 + }, + { + "epoch": 7.44, + "learning_rate": 7.69111758281762e-07, + "loss": 0.0062, + "step": 40905 + }, + { + "epoch": 7.44, + "learning_rate": 7.690571532580998e-07, + "loss": 2.3358, + "step": 40906 + }, + { + "epoch": 7.44, + "learning_rate": 7.690025482344375e-07, + "loss": 0.0377, + "step": 40907 + }, + { + "epoch": 7.44, + "learning_rate": 7.689479432107754e-07, + "loss": 0.0862, + "step": 40908 + }, + { + "epoch": 7.44, + "learning_rate": 7.688933381871133e-07, + "loss": 0.1803, + "step": 40909 + }, + { + "epoch": 7.44, + "learning_rate": 7.68838733163451e-07, + "loss": 0.0548, + "step": 40910 + }, + { + "epoch": 7.44, + "learning_rate": 7.687841281397888e-07, + "loss": 0.0515, + "step": 40911 + }, + { + "epoch": 7.44, + "learning_rate": 7.687295231161267e-07, + "loss": 0.4585, + "step": 40912 + }, + { + "epoch": 7.44, + "learning_rate": 7.686749180924646e-07, + "loss": 0.013, + "step": 40913 + }, + { + "epoch": 7.44, + "learning_rate": 7.686203130688023e-07, + "loss": 1.1343, + "step": 40914 + }, + { + "epoch": 7.44, + "learning_rate": 7.685657080451401e-07, + "loss": 0.777, + "step": 40915 + }, + { + "epoch": 7.44, + "learning_rate": 7.685111030214781e-07, + "loss": 0.0492, + "step": 40916 + }, + { + "epoch": 7.44, + "learning_rate": 7.684564979978158e-07, + "loss": 0.4041, + "step": 40917 + }, + { + "epoch": 7.44, + "learning_rate": 7.684018929741535e-07, + "loss": 0.1077, + "step": 40918 + }, + { + "epoch": 7.44, + "learning_rate": 7.683472879504915e-07, + "loss": 0.313, + "step": 40919 + }, + { + "epoch": 7.44, + "learning_rate": 7.682926829268293e-07, + "loss": 0.1082, + "step": 40920 + }, + { + "epoch": 7.44, + "learning_rate": 7.682380779031671e-07, + "loss": 0.0575, + "step": 40921 + }, + { + "epoch": 7.44, + "learning_rate": 7.681834728795049e-07, + "loss": 0.5188, + "step": 40922 + }, + { + "epoch": 7.44, + "learning_rate": 7.681288678558428e-07, + "loss": 0.0044, + "step": 40923 + }, + { + "epoch": 7.44, + "learning_rate": 7.680742628321806e-07, + "loss": 0.3338, + "step": 40924 + }, + { + "epoch": 7.44, + "learning_rate": 7.680196578085183e-07, + "loss": 0.1663, + "step": 40925 + }, + { + "epoch": 7.44, + "learning_rate": 7.679650527848562e-07, + "loss": 0.0034, + "step": 40926 + }, + { + "epoch": 7.44, + "learning_rate": 7.679104477611941e-07, + "loss": 0.0146, + "step": 40927 + }, + { + "epoch": 7.44, + "learning_rate": 7.678558427375318e-07, + "loss": 0.1958, + "step": 40928 + }, + { + "epoch": 7.44, + "learning_rate": 7.678012377138696e-07, + "loss": 0.2303, + "step": 40929 + }, + { + "epoch": 7.44, + "learning_rate": 7.677466326902076e-07, + "loss": 0.0661, + "step": 40930 + }, + { + "epoch": 7.44, + "learning_rate": 7.676920276665453e-07, + "loss": 0.0558, + "step": 40931 + }, + { + "epoch": 7.44, + "learning_rate": 7.676374226428832e-07, + "loss": 0.9078, + "step": 40932 + }, + { + "epoch": 7.44, + "learning_rate": 7.675828176192211e-07, + "loss": 0.9487, + "step": 40933 + }, + { + "epoch": 7.44, + "learning_rate": 7.675282125955588e-07, + "loss": 0.1503, + "step": 40934 + }, + { + "epoch": 7.44, + "learning_rate": 7.674736075718966e-07, + "loss": 0.1315, + "step": 40935 + }, + { + "epoch": 7.44, + "learning_rate": 7.674190025482344e-07, + "loss": 1.5717, + "step": 40936 + }, + { + "epoch": 7.44, + "learning_rate": 7.673643975245723e-07, + "loss": 0.0079, + "step": 40937 + }, + { + "epoch": 7.44, + "learning_rate": 7.673097925009101e-07, + "loss": 0.0383, + "step": 40938 + }, + { + "epoch": 7.44, + "learning_rate": 7.672551874772479e-07, + "loss": 2.7948, + "step": 40939 + }, + { + "epoch": 7.44, + "learning_rate": 7.672005824535859e-07, + "loss": 0.3844, + "step": 40940 + }, + { + "epoch": 7.45, + "learning_rate": 7.671459774299236e-07, + "loss": 0.0887, + "step": 40941 + }, + { + "epoch": 7.45, + "learning_rate": 7.670913724062613e-07, + "loss": 0.0099, + "step": 40942 + }, + { + "epoch": 7.45, + "learning_rate": 7.670367673825992e-07, + "loss": 0.8587, + "step": 40943 + }, + { + "epoch": 7.45, + "learning_rate": 7.669821623589371e-07, + "loss": 0.0509, + "step": 40944 + }, + { + "epoch": 7.45, + "learning_rate": 7.669275573352748e-07, + "loss": 1.7041, + "step": 40945 + }, + { + "epoch": 7.45, + "learning_rate": 7.668729523116127e-07, + "loss": 0.0972, + "step": 40946 + }, + { + "epoch": 7.45, + "learning_rate": 7.668183472879506e-07, + "loss": 0.0052, + "step": 40947 + }, + { + "epoch": 7.45, + "learning_rate": 7.667637422642883e-07, + "loss": 0.2405, + "step": 40948 + }, + { + "epoch": 7.45, + "learning_rate": 7.667091372406262e-07, + "loss": 1.31, + "step": 40949 + }, + { + "epoch": 7.45, + "learning_rate": 7.666545322169639e-07, + "loss": 0.0159, + "step": 40950 + }, + { + "epoch": 7.45, + "learning_rate": 7.665999271933019e-07, + "loss": 0.2073, + "step": 40951 + }, + { + "epoch": 7.45, + "learning_rate": 7.665453221696396e-07, + "loss": 1.4135, + "step": 40952 + }, + { + "epoch": 7.45, + "learning_rate": 7.664907171459774e-07, + "loss": 0.0113, + "step": 40953 + }, + { + "epoch": 7.45, + "learning_rate": 7.664361121223154e-07, + "loss": 0.4173, + "step": 40954 + }, + { + "epoch": 7.45, + "learning_rate": 7.663815070986531e-07, + "loss": 0.0512, + "step": 40955 + }, + { + "epoch": 7.45, + "learning_rate": 7.663269020749909e-07, + "loss": 0.0919, + "step": 40956 + }, + { + "epoch": 7.45, + "learning_rate": 7.662722970513287e-07, + "loss": 0.8263, + "step": 40957 + }, + { + "epoch": 7.45, + "learning_rate": 7.662176920276666e-07, + "loss": 0.004, + "step": 40958 + }, + { + "epoch": 7.45, + "learning_rate": 7.661630870040043e-07, + "loss": 2.6487, + "step": 40959 + }, + { + "epoch": 7.45, + "learning_rate": 7.661084819803422e-07, + "loss": 1.5458, + "step": 40960 + }, + { + "epoch": 7.45, + "learning_rate": 7.660538769566801e-07, + "loss": 0.4671, + "step": 40961 + }, + { + "epoch": 7.45, + "learning_rate": 7.659992719330179e-07, + "loss": 0.1471, + "step": 40962 + }, + { + "epoch": 7.45, + "learning_rate": 7.659446669093557e-07, + "loss": 0.2633, + "step": 40963 + }, + { + "epoch": 7.45, + "learning_rate": 7.658900618856934e-07, + "loss": 0.4631, + "step": 40964 + }, + { + "epoch": 7.45, + "learning_rate": 7.658354568620314e-07, + "loss": 2.8098, + "step": 40965 + }, + { + "epoch": 7.45, + "learning_rate": 7.657808518383691e-07, + "loss": 1.798, + "step": 40966 + }, + { + "epoch": 7.45, + "learning_rate": 7.657262468147069e-07, + "loss": 0.2287, + "step": 40967 + }, + { + "epoch": 7.45, + "learning_rate": 7.656716417910449e-07, + "loss": 0.0118, + "step": 40968 + }, + { + "epoch": 7.45, + "learning_rate": 7.656170367673826e-07, + "loss": 0.0397, + "step": 40969 + }, + { + "epoch": 7.45, + "learning_rate": 7.655624317437205e-07, + "loss": 0.7792, + "step": 40970 + }, + { + "epoch": 7.45, + "learning_rate": 7.655078267200582e-07, + "loss": 1.171, + "step": 40971 + }, + { + "epoch": 7.45, + "learning_rate": 7.654532216963961e-07, + "loss": 0.5995, + "step": 40972 + }, + { + "epoch": 7.45, + "learning_rate": 7.65398616672734e-07, + "loss": 0.0082, + "step": 40973 + }, + { + "epoch": 7.45, + "learning_rate": 7.653440116490717e-07, + "loss": 0.8067, + "step": 40974 + }, + { + "epoch": 7.45, + "learning_rate": 7.652894066254096e-07, + "loss": 0.1212, + "step": 40975 + }, + { + "epoch": 7.45, + "learning_rate": 7.652348016017474e-07, + "loss": 0.2467, + "step": 40976 + }, + { + "epoch": 7.45, + "learning_rate": 7.651801965780852e-07, + "loss": 0.3702, + "step": 40977 + }, + { + "epoch": 7.45, + "learning_rate": 7.651255915544229e-07, + "loss": 0.0171, + "step": 40978 + }, + { + "epoch": 7.45, + "learning_rate": 7.650709865307609e-07, + "loss": 0.0641, + "step": 40979 + }, + { + "epoch": 7.45, + "learning_rate": 7.650163815070987e-07, + "loss": 0.2668, + "step": 40980 + }, + { + "epoch": 7.45, + "learning_rate": 7.649617764834365e-07, + "loss": 0.144, + "step": 40981 + }, + { + "epoch": 7.45, + "learning_rate": 7.649071714597744e-07, + "loss": 0.0081, + "step": 40982 + }, + { + "epoch": 7.45, + "learning_rate": 7.648525664361121e-07, + "loss": 0.0556, + "step": 40983 + }, + { + "epoch": 7.45, + "learning_rate": 7.6479796141245e-07, + "loss": 0.0119, + "step": 40984 + }, + { + "epoch": 7.45, + "learning_rate": 7.647433563887877e-07, + "loss": 0.0924, + "step": 40985 + }, + { + "epoch": 7.45, + "learning_rate": 7.646887513651256e-07, + "loss": 0.3593, + "step": 40986 + }, + { + "epoch": 7.45, + "learning_rate": 7.646341463414635e-07, + "loss": 1.203, + "step": 40987 + }, + { + "epoch": 7.45, + "learning_rate": 7.645795413178012e-07, + "loss": 0.1709, + "step": 40988 + }, + { + "epoch": 7.45, + "learning_rate": 7.64524936294139e-07, + "loss": 0.2731, + "step": 40989 + }, + { + "epoch": 7.45, + "learning_rate": 7.644703312704769e-07, + "loss": 0.1128, + "step": 40990 + }, + { + "epoch": 7.45, + "learning_rate": 7.644157262468147e-07, + "loss": 2.3894, + "step": 40991 + }, + { + "epoch": 7.45, + "learning_rate": 7.643611212231525e-07, + "loss": 0.1056, + "step": 40992 + }, + { + "epoch": 7.45, + "learning_rate": 7.643065161994904e-07, + "loss": 0.6587, + "step": 40993 + }, + { + "epoch": 7.45, + "learning_rate": 7.642519111758282e-07, + "loss": 0.2757, + "step": 40994 + }, + { + "epoch": 7.45, + "learning_rate": 7.64197306152166e-07, + "loss": 0.7786, + "step": 40995 + }, + { + "epoch": 7.46, + "learning_rate": 7.641427011285038e-07, + "loss": 0.0377, + "step": 40996 + }, + { + "epoch": 7.46, + "learning_rate": 7.640880961048418e-07, + "loss": 0.0164, + "step": 40997 + }, + { + "epoch": 7.46, + "learning_rate": 7.640334910811795e-07, + "loss": 0.0244, + "step": 40998 + }, + { + "epoch": 7.46, + "learning_rate": 7.639788860575172e-07, + "loss": 0.0118, + "step": 40999 + }, + { + "epoch": 7.46, + "learning_rate": 7.639242810338552e-07, + "loss": 1.7734, + "step": 41000 + }, + { + "epoch": 7.46, + "learning_rate": 7.63869676010193e-07, + "loss": 0.008, + "step": 41001 + }, + { + "epoch": 7.46, + "learning_rate": 7.638150709865307e-07, + "loss": 0.8713, + "step": 41002 + }, + { + "epoch": 7.46, + "learning_rate": 7.637604659628686e-07, + "loss": 0.0251, + "step": 41003 + }, + { + "epoch": 7.46, + "learning_rate": 7.637058609392065e-07, + "loss": 0.2269, + "step": 41004 + }, + { + "epoch": 7.46, + "learning_rate": 7.636512559155442e-07, + "loss": 0.4775, + "step": 41005 + }, + { + "epoch": 7.46, + "learning_rate": 7.63596650891882e-07, + "loss": 0.0731, + "step": 41006 + }, + { + "epoch": 7.46, + "learning_rate": 7.635420458682199e-07, + "loss": 0.2919, + "step": 41007 + }, + { + "epoch": 7.46, + "learning_rate": 7.634874408445578e-07, + "loss": 0.0262, + "step": 41008 + }, + { + "epoch": 7.46, + "learning_rate": 7.634328358208955e-07, + "loss": 4.4757, + "step": 41009 + }, + { + "epoch": 7.46, + "learning_rate": 7.633782307972333e-07, + "loss": 0.0121, + "step": 41010 + }, + { + "epoch": 7.46, + "learning_rate": 7.633236257735713e-07, + "loss": 0.3277, + "step": 41011 + }, + { + "epoch": 7.46, + "learning_rate": 7.63269020749909e-07, + "loss": 0.5813, + "step": 41012 + }, + { + "epoch": 7.46, + "learning_rate": 7.632144157262468e-07, + "loss": 0.0091, + "step": 41013 + }, + { + "epoch": 7.46, + "learning_rate": 7.631598107025848e-07, + "loss": 0.8672, + "step": 41014 + }, + { + "epoch": 7.46, + "learning_rate": 7.631052056789225e-07, + "loss": 0.4525, + "step": 41015 + }, + { + "epoch": 7.46, + "learning_rate": 7.630506006552602e-07, + "loss": 0.0044, + "step": 41016 + }, + { + "epoch": 7.46, + "learning_rate": 7.629959956315981e-07, + "loss": 0.4637, + "step": 41017 + }, + { + "epoch": 7.46, + "learning_rate": 7.62941390607936e-07, + "loss": 0.4265, + "step": 41018 + }, + { + "epoch": 7.46, + "learning_rate": 7.628867855842738e-07, + "loss": 2.0745, + "step": 41019 + }, + { + "epoch": 7.46, + "learning_rate": 7.628321805606116e-07, + "loss": 0.3604, + "step": 41020 + }, + { + "epoch": 7.46, + "learning_rate": 7.627775755369495e-07, + "loss": 0.2225, + "step": 41021 + }, + { + "epoch": 7.46, + "learning_rate": 7.627229705132873e-07, + "loss": 1.7985, + "step": 41022 + }, + { + "epoch": 7.46, + "learning_rate": 7.62668365489625e-07, + "loss": 0.4079, + "step": 41023 + }, + { + "epoch": 7.46, + "learning_rate": 7.626137604659628e-07, + "loss": 0.2741, + "step": 41024 + }, + { + "epoch": 7.46, + "learning_rate": 7.625591554423008e-07, + "loss": 0.7939, + "step": 41025 + }, + { + "epoch": 7.46, + "learning_rate": 7.625045504186385e-07, + "loss": 0.7965, + "step": 41026 + }, + { + "epoch": 7.46, + "learning_rate": 7.624499453949763e-07, + "loss": 0.1737, + "step": 41027 + }, + { + "epoch": 7.46, + "learning_rate": 7.623953403713143e-07, + "loss": 0.4559, + "step": 41028 + }, + { + "epoch": 7.46, + "learning_rate": 7.62340735347652e-07, + "loss": 0.003, + "step": 41029 + }, + { + "epoch": 7.46, + "learning_rate": 7.622861303239898e-07, + "loss": 0.0051, + "step": 41030 + }, + { + "epoch": 7.46, + "learning_rate": 7.622315253003276e-07, + "loss": 1.5559, + "step": 41031 + }, + { + "epoch": 7.46, + "learning_rate": 7.621769202766655e-07, + "loss": 0.9442, + "step": 41032 + }, + { + "epoch": 7.46, + "learning_rate": 7.621223152530033e-07, + "loss": 0.608, + "step": 41033 + }, + { + "epoch": 7.46, + "learning_rate": 7.620677102293411e-07, + "loss": 0.0038, + "step": 41034 + }, + { + "epoch": 7.46, + "learning_rate": 7.62013105205679e-07, + "loss": 0.3512, + "step": 41035 + }, + { + "epoch": 7.46, + "learning_rate": 7.619585001820168e-07, + "loss": 0.9222, + "step": 41036 + }, + { + "epoch": 7.46, + "learning_rate": 7.619038951583546e-07, + "loss": 2.0255, + "step": 41037 + }, + { + "epoch": 7.46, + "learning_rate": 7.618492901346923e-07, + "loss": 0.5274, + "step": 41038 + }, + { + "epoch": 7.46, + "learning_rate": 7.617946851110303e-07, + "loss": 0.1181, + "step": 41039 + }, + { + "epoch": 7.46, + "learning_rate": 7.61740080087368e-07, + "loss": 0.2582, + "step": 41040 + }, + { + "epoch": 7.46, + "learning_rate": 7.616854750637059e-07, + "loss": 0.3253, + "step": 41041 + }, + { + "epoch": 7.46, + "learning_rate": 7.616308700400438e-07, + "loss": 0.1785, + "step": 41042 + }, + { + "epoch": 7.46, + "learning_rate": 7.615762650163815e-07, + "loss": 0.0295, + "step": 41043 + }, + { + "epoch": 7.46, + "learning_rate": 7.615216599927194e-07, + "loss": 0.045, + "step": 41044 + }, + { + "epoch": 7.46, + "learning_rate": 7.614670549690571e-07, + "loss": 0.9077, + "step": 41045 + }, + { + "epoch": 7.46, + "learning_rate": 7.61412449945395e-07, + "loss": 0.9655, + "step": 41046 + }, + { + "epoch": 7.46, + "learning_rate": 7.613578449217328e-07, + "loss": 1.4217, + "step": 41047 + }, + { + "epoch": 7.46, + "learning_rate": 7.613032398980706e-07, + "loss": 0.0647, + "step": 41048 + }, + { + "epoch": 7.46, + "learning_rate": 7.612486348744086e-07, + "loss": 0.7436, + "step": 41049 + }, + { + "epoch": 7.46, + "learning_rate": 7.611940298507463e-07, + "loss": 0.0536, + "step": 41050 + }, + { + "epoch": 7.47, + "learning_rate": 7.611394248270841e-07, + "loss": 0.315, + "step": 41051 + }, + { + "epoch": 7.47, + "learning_rate": 7.610848198034219e-07, + "loss": 0.0144, + "step": 41052 + }, + { + "epoch": 7.47, + "learning_rate": 7.610302147797598e-07, + "loss": 0.609, + "step": 41053 + }, + { + "epoch": 7.47, + "learning_rate": 7.609756097560976e-07, + "loss": 1.0373, + "step": 41054 + }, + { + "epoch": 7.47, + "learning_rate": 7.609210047324354e-07, + "loss": 0.1631, + "step": 41055 + }, + { + "epoch": 7.47, + "learning_rate": 7.608663997087733e-07, + "loss": 0.0241, + "step": 41056 + }, + { + "epoch": 7.47, + "learning_rate": 7.60811794685111e-07, + "loss": 1.5423, + "step": 41057 + }, + { + "epoch": 7.47, + "learning_rate": 7.607571896614489e-07, + "loss": 1.3535, + "step": 41058 + }, + { + "epoch": 7.47, + "learning_rate": 7.607025846377866e-07, + "loss": 0.5018, + "step": 41059 + }, + { + "epoch": 7.47, + "learning_rate": 7.606479796141246e-07, + "loss": 2.1137, + "step": 41060 + }, + { + "epoch": 7.47, + "learning_rate": 7.605933745904624e-07, + "loss": 0.1999, + "step": 41061 + }, + { + "epoch": 7.47, + "learning_rate": 7.605387695668001e-07, + "loss": 0.529, + "step": 41062 + }, + { + "epoch": 7.47, + "learning_rate": 7.604841645431379e-07, + "loss": 0.1738, + "step": 41063 + }, + { + "epoch": 7.47, + "learning_rate": 7.604295595194758e-07, + "loss": 1.157, + "step": 41064 + }, + { + "epoch": 7.47, + "learning_rate": 7.603749544958136e-07, + "loss": 0.4291, + "step": 41065 + }, + { + "epoch": 7.47, + "learning_rate": 7.603203494721514e-07, + "loss": 1.4625, + "step": 41066 + }, + { + "epoch": 7.47, + "learning_rate": 7.602657444484893e-07, + "loss": 0.1182, + "step": 41067 + }, + { + "epoch": 7.47, + "learning_rate": 7.602111394248272e-07, + "loss": 1.4292, + "step": 41068 + }, + { + "epoch": 7.47, + "learning_rate": 7.601565344011649e-07, + "loss": 0.0038, + "step": 41069 + }, + { + "epoch": 7.47, + "learning_rate": 7.601019293775026e-07, + "loss": 0.8475, + "step": 41070 + }, + { + "epoch": 7.47, + "learning_rate": 7.600473243538406e-07, + "loss": 0.5406, + "step": 41071 + }, + { + "epoch": 7.47, + "learning_rate": 7.599927193301784e-07, + "loss": 0.4469, + "step": 41072 + }, + { + "epoch": 7.47, + "learning_rate": 7.599381143065161e-07, + "loss": 0.8728, + "step": 41073 + }, + { + "epoch": 7.47, + "learning_rate": 7.598835092828541e-07, + "loss": 0.0037, + "step": 41074 + }, + { + "epoch": 7.47, + "learning_rate": 7.598289042591919e-07, + "loss": 0.2957, + "step": 41075 + }, + { + "epoch": 7.47, + "learning_rate": 7.597742992355296e-07, + "loss": 0.2396, + "step": 41076 + }, + { + "epoch": 7.47, + "learning_rate": 7.597196942118675e-07, + "loss": 1.371, + "step": 41077 + }, + { + "epoch": 7.47, + "learning_rate": 7.596650891882054e-07, + "loss": 0.0079, + "step": 41078 + }, + { + "epoch": 7.47, + "learning_rate": 7.596104841645432e-07, + "loss": 0.0074, + "step": 41079 + }, + { + "epoch": 7.47, + "learning_rate": 7.595558791408809e-07, + "loss": 0.0056, + "step": 41080 + }, + { + "epoch": 7.47, + "learning_rate": 7.595012741172188e-07, + "loss": 0.9053, + "step": 41081 + }, + { + "epoch": 7.47, + "learning_rate": 7.594466690935567e-07, + "loss": 0.6488, + "step": 41082 + }, + { + "epoch": 7.47, + "learning_rate": 7.593920640698944e-07, + "loss": 0.166, + "step": 41083 + }, + { + "epoch": 7.47, + "learning_rate": 7.593374590462322e-07, + "loss": 0.0439, + "step": 41084 + }, + { + "epoch": 7.47, + "learning_rate": 7.592828540225702e-07, + "loss": 0.2496, + "step": 41085 + }, + { + "epoch": 7.47, + "learning_rate": 7.592282489989079e-07, + "loss": 2.3618, + "step": 41086 + }, + { + "epoch": 7.47, + "learning_rate": 7.591736439752456e-07, + "loss": 0.6588, + "step": 41087 + }, + { + "epoch": 7.47, + "learning_rate": 7.591190389515836e-07, + "loss": 0.2476, + "step": 41088 + }, + { + "epoch": 7.47, + "learning_rate": 7.590644339279214e-07, + "loss": 1.0113, + "step": 41089 + }, + { + "epoch": 7.47, + "learning_rate": 7.590098289042592e-07, + "loss": 1.1492, + "step": 41090 + }, + { + "epoch": 7.47, + "learning_rate": 7.58955223880597e-07, + "loss": 2.1168, + "step": 41091 + }, + { + "epoch": 7.47, + "learning_rate": 7.589006188569349e-07, + "loss": 0.9588, + "step": 41092 + }, + { + "epoch": 7.47, + "learning_rate": 7.588460138332727e-07, + "loss": 0.4347, + "step": 41093 + }, + { + "epoch": 7.47, + "learning_rate": 7.587914088096104e-07, + "loss": 2.6514, + "step": 41094 + }, + { + "epoch": 7.47, + "learning_rate": 7.587368037859483e-07, + "loss": 0.0135, + "step": 41095 + }, + { + "epoch": 7.47, + "learning_rate": 7.586821987622862e-07, + "loss": 0.0263, + "step": 41096 + }, + { + "epoch": 7.47, + "learning_rate": 7.586275937386239e-07, + "loss": 2.0448, + "step": 41097 + }, + { + "epoch": 7.47, + "learning_rate": 7.585729887149618e-07, + "loss": 0.0119, + "step": 41098 + }, + { + "epoch": 7.47, + "learning_rate": 7.585183836912997e-07, + "loss": 1.9564, + "step": 41099 + }, + { + "epoch": 7.47, + "learning_rate": 7.584637786676374e-07, + "loss": 0.1611, + "step": 41100 + }, + { + "epoch": 7.47, + "learning_rate": 7.584091736439753e-07, + "loss": 0.308, + "step": 41101 + }, + { + "epoch": 7.47, + "learning_rate": 7.583545686203132e-07, + "loss": 0.009, + "step": 41102 + }, + { + "epoch": 7.47, + "learning_rate": 7.582999635966509e-07, + "loss": 0.11, + "step": 41103 + }, + { + "epoch": 7.47, + "learning_rate": 7.582453585729887e-07, + "loss": 0.8757, + "step": 41104 + }, + { + "epoch": 7.47, + "learning_rate": 7.581907535493265e-07, + "loss": 0.4482, + "step": 41105 + }, + { + "epoch": 7.48, + "learning_rate": 7.581361485256645e-07, + "loss": 0.6813, + "step": 41106 + }, + { + "epoch": 7.48, + "learning_rate": 7.580815435020022e-07, + "loss": 0.5285, + "step": 41107 + }, + { + "epoch": 7.48, + "learning_rate": 7.5802693847834e-07, + "loss": 0.3884, + "step": 41108 + }, + { + "epoch": 7.48, + "learning_rate": 7.57972333454678e-07, + "loss": 0.2956, + "step": 41109 + }, + { + "epoch": 7.48, + "learning_rate": 7.579177284310157e-07, + "loss": 0.5129, + "step": 41110 + }, + { + "epoch": 7.48, + "learning_rate": 7.578631234073534e-07, + "loss": 0.1837, + "step": 41111 + }, + { + "epoch": 7.48, + "learning_rate": 7.578085183836913e-07, + "loss": 0.0251, + "step": 41112 + }, + { + "epoch": 7.48, + "learning_rate": 7.577539133600292e-07, + "loss": 1.9352, + "step": 41113 + }, + { + "epoch": 7.48, + "learning_rate": 7.576993083363669e-07, + "loss": 0.1265, + "step": 41114 + }, + { + "epoch": 7.48, + "learning_rate": 7.576447033127048e-07, + "loss": 1.052, + "step": 41115 + }, + { + "epoch": 7.48, + "learning_rate": 7.575900982890427e-07, + "loss": 0.119, + "step": 41116 + }, + { + "epoch": 7.48, + "learning_rate": 7.575354932653805e-07, + "loss": 0.1658, + "step": 41117 + }, + { + "epoch": 7.48, + "learning_rate": 7.574808882417183e-07, + "loss": 0.0224, + "step": 41118 + }, + { + "epoch": 7.48, + "learning_rate": 7.57426283218056e-07, + "loss": 0.0138, + "step": 41119 + }, + { + "epoch": 7.48, + "learning_rate": 7.57371678194394e-07, + "loss": 1.0467, + "step": 41120 + }, + { + "epoch": 7.48, + "learning_rate": 7.573170731707317e-07, + "loss": 0.1483, + "step": 41121 + }, + { + "epoch": 7.48, + "learning_rate": 7.572624681470695e-07, + "loss": 0.677, + "step": 41122 + }, + { + "epoch": 7.48, + "learning_rate": 7.572078631234075e-07, + "loss": 0.081, + "step": 41123 + }, + { + "epoch": 7.48, + "learning_rate": 7.571532580997452e-07, + "loss": 0.3192, + "step": 41124 + }, + { + "epoch": 7.48, + "learning_rate": 7.57098653076083e-07, + "loss": 0.0843, + "step": 41125 + }, + { + "epoch": 7.48, + "learning_rate": 7.570440480524208e-07, + "loss": 0.1786, + "step": 41126 + }, + { + "epoch": 7.48, + "learning_rate": 7.569894430287587e-07, + "loss": 0.0104, + "step": 41127 + }, + { + "epoch": 7.48, + "learning_rate": 7.569348380050965e-07, + "loss": 0.3648, + "step": 41128 + }, + { + "epoch": 7.48, + "learning_rate": 7.568802329814343e-07, + "loss": 0.0612, + "step": 41129 + }, + { + "epoch": 7.48, + "learning_rate": 7.568256279577722e-07, + "loss": 0.0763, + "step": 41130 + }, + { + "epoch": 7.48, + "learning_rate": 7.5677102293411e-07, + "loss": 1.041, + "step": 41131 + }, + { + "epoch": 7.48, + "learning_rate": 7.567164179104478e-07, + "loss": 0.9532, + "step": 41132 + }, + { + "epoch": 7.48, + "learning_rate": 7.566618128867855e-07, + "loss": 0.5841, + "step": 41133 + }, + { + "epoch": 7.48, + "learning_rate": 7.566072078631235e-07, + "loss": 0.3638, + "step": 41134 + }, + { + "epoch": 7.48, + "learning_rate": 7.565526028394612e-07, + "loss": 0.7692, + "step": 41135 + }, + { + "epoch": 7.48, + "learning_rate": 7.56497997815799e-07, + "loss": 0.1047, + "step": 41136 + }, + { + "epoch": 7.48, + "learning_rate": 7.56443392792137e-07, + "loss": 0.1693, + "step": 41137 + }, + { + "epoch": 7.48, + "learning_rate": 7.563887877684747e-07, + "loss": 1.2335, + "step": 41138 + }, + { + "epoch": 7.48, + "learning_rate": 7.563341827448126e-07, + "loss": 1.9699, + "step": 41139 + }, + { + "epoch": 7.48, + "learning_rate": 7.562795777211503e-07, + "loss": 0.8018, + "step": 41140 + }, + { + "epoch": 7.48, + "learning_rate": 7.562249726974882e-07, + "loss": 0.0678, + "step": 41141 + }, + { + "epoch": 7.48, + "learning_rate": 7.561703676738261e-07, + "loss": 0.0458, + "step": 41142 + }, + { + "epoch": 7.48, + "learning_rate": 7.561157626501638e-07, + "loss": 1.5657, + "step": 41143 + }, + { + "epoch": 7.48, + "learning_rate": 7.560611576265015e-07, + "loss": 0.0484, + "step": 41144 + }, + { + "epoch": 7.48, + "learning_rate": 7.560065526028395e-07, + "loss": 0.1212, + "step": 41145 + }, + { + "epoch": 7.48, + "learning_rate": 7.559519475791773e-07, + "loss": 0.0045, + "step": 41146 + }, + { + "epoch": 7.48, + "learning_rate": 7.55897342555515e-07, + "loss": 0.1075, + "step": 41147 + }, + { + "epoch": 7.48, + "learning_rate": 7.55842737531853e-07, + "loss": 0.3081, + "step": 41148 + }, + { + "epoch": 7.48, + "learning_rate": 7.557881325081908e-07, + "loss": 0.3884, + "step": 41149 + }, + { + "epoch": 7.48, + "learning_rate": 7.557335274845286e-07, + "loss": 1.1888, + "step": 41150 + }, + { + "epoch": 7.48, + "learning_rate": 7.556789224608663e-07, + "loss": 0.0038, + "step": 41151 + }, + { + "epoch": 7.48, + "learning_rate": 7.556243174372042e-07, + "loss": 1.8195, + "step": 41152 + }, + { + "epoch": 7.48, + "learning_rate": 7.555697124135421e-07, + "loss": 0.0036, + "step": 41153 + }, + { + "epoch": 7.48, + "learning_rate": 7.555151073898798e-07, + "loss": 0.3931, + "step": 41154 + }, + { + "epoch": 7.48, + "learning_rate": 7.554605023662177e-07, + "loss": 1.4242, + "step": 41155 + }, + { + "epoch": 7.48, + "learning_rate": 7.554058973425556e-07, + "loss": 0.0981, + "step": 41156 + }, + { + "epoch": 7.48, + "learning_rate": 7.553512923188933e-07, + "loss": 0.03, + "step": 41157 + }, + { + "epoch": 7.48, + "learning_rate": 7.55296687295231e-07, + "loss": 0.0209, + "step": 41158 + }, + { + "epoch": 7.48, + "learning_rate": 7.55242082271569e-07, + "loss": 0.0135, + "step": 41159 + }, + { + "epoch": 7.48, + "learning_rate": 7.551874772479068e-07, + "loss": 0.9923, + "step": 41160 + }, + { + "epoch": 7.49, + "learning_rate": 7.551328722242446e-07, + "loss": 0.1867, + "step": 41161 + }, + { + "epoch": 7.49, + "learning_rate": 7.550782672005825e-07, + "loss": 0.0082, + "step": 41162 + }, + { + "epoch": 7.49, + "learning_rate": 7.550236621769203e-07, + "loss": 0.6412, + "step": 41163 + }, + { + "epoch": 7.49, + "learning_rate": 7.549690571532581e-07, + "loss": 0.2711, + "step": 41164 + }, + { + "epoch": 7.49, + "learning_rate": 7.549144521295959e-07, + "loss": 0.8602, + "step": 41165 + }, + { + "epoch": 7.49, + "learning_rate": 7.548598471059339e-07, + "loss": 0.05, + "step": 41166 + }, + { + "epoch": 7.49, + "learning_rate": 7.548052420822716e-07, + "loss": 0.127, + "step": 41167 + }, + { + "epoch": 7.49, + "learning_rate": 7.547506370586093e-07, + "loss": 0.7404, + "step": 41168 + }, + { + "epoch": 7.49, + "learning_rate": 7.546960320349473e-07, + "loss": 0.4455, + "step": 41169 + }, + { + "epoch": 7.49, + "learning_rate": 7.546414270112851e-07, + "loss": 0.151, + "step": 41170 + }, + { + "epoch": 7.49, + "learning_rate": 7.545868219876228e-07, + "loss": 0.211, + "step": 41171 + }, + { + "epoch": 7.49, + "learning_rate": 7.545322169639607e-07, + "loss": 1.3365, + "step": 41172 + }, + { + "epoch": 7.49, + "learning_rate": 7.544776119402986e-07, + "loss": 0.0543, + "step": 41173 + }, + { + "epoch": 7.49, + "learning_rate": 7.544230069166363e-07, + "loss": 0.0526, + "step": 41174 + }, + { + "epoch": 7.49, + "learning_rate": 7.543684018929741e-07, + "loss": 0.1643, + "step": 41175 + }, + { + "epoch": 7.49, + "learning_rate": 7.54313796869312e-07, + "loss": 0.9623, + "step": 41176 + }, + { + "epoch": 7.49, + "learning_rate": 7.542591918456499e-07, + "loss": 0.0065, + "step": 41177 + }, + { + "epoch": 7.49, + "learning_rate": 7.542045868219876e-07, + "loss": 0.7931, + "step": 41178 + }, + { + "epoch": 7.49, + "learning_rate": 7.541499817983254e-07, + "loss": 0.0097, + "step": 41179 + }, + { + "epoch": 7.49, + "learning_rate": 7.540953767746634e-07, + "loss": 0.1696, + "step": 41180 + }, + { + "epoch": 7.49, + "learning_rate": 7.540407717510011e-07, + "loss": 0.3596, + "step": 41181 + }, + { + "epoch": 7.49, + "learning_rate": 7.539861667273389e-07, + "loss": 0.422, + "step": 41182 + }, + { + "epoch": 7.49, + "learning_rate": 7.539315617036769e-07, + "loss": 1.1341, + "step": 41183 + }, + { + "epoch": 7.49, + "learning_rate": 7.538769566800146e-07, + "loss": 1.4267, + "step": 41184 + }, + { + "epoch": 7.49, + "learning_rate": 7.538223516563523e-07, + "loss": 0.4157, + "step": 41185 + }, + { + "epoch": 7.49, + "learning_rate": 7.537677466326902e-07, + "loss": 0.9316, + "step": 41186 + }, + { + "epoch": 7.49, + "learning_rate": 7.537131416090281e-07, + "loss": 2.15, + "step": 41187 + }, + { + "epoch": 7.49, + "learning_rate": 7.536585365853659e-07, + "loss": 1.2777, + "step": 41188 + }, + { + "epoch": 7.49, + "learning_rate": 7.536039315617037e-07, + "loss": 0.4973, + "step": 41189 + }, + { + "epoch": 7.49, + "learning_rate": 7.535493265380416e-07, + "loss": 2.1995, + "step": 41190 + }, + { + "epoch": 7.49, + "learning_rate": 7.534947215143794e-07, + "loss": 0.6903, + "step": 41191 + }, + { + "epoch": 7.49, + "learning_rate": 7.534401164907171e-07, + "loss": 0.1527, + "step": 41192 + }, + { + "epoch": 7.49, + "learning_rate": 7.533855114670549e-07, + "loss": 0.0982, + "step": 41193 + }, + { + "epoch": 7.49, + "learning_rate": 7.533309064433929e-07, + "loss": 0.0738, + "step": 41194 + }, + { + "epoch": 7.49, + "learning_rate": 7.532763014197306e-07, + "loss": 0.3128, + "step": 41195 + }, + { + "epoch": 7.49, + "learning_rate": 7.532216963960685e-07, + "loss": 0.0869, + "step": 41196 + }, + { + "epoch": 7.49, + "learning_rate": 7.531670913724064e-07, + "loss": 0.6464, + "step": 41197 + }, + { + "epoch": 7.49, + "learning_rate": 7.531124863487441e-07, + "loss": 0.4394, + "step": 41198 + }, + { + "epoch": 7.49, + "learning_rate": 7.530578813250819e-07, + "loss": 0.0725, + "step": 41199 + }, + { + "epoch": 7.49, + "learning_rate": 7.530032763014197e-07, + "loss": 1.3019, + "step": 41200 + }, + { + "epoch": 7.49, + "learning_rate": 7.529486712777576e-07, + "loss": 0.3972, + "step": 41201 + }, + { + "epoch": 7.49, + "learning_rate": 7.528940662540954e-07, + "loss": 0.0146, + "step": 41202 + }, + { + "epoch": 7.49, + "learning_rate": 7.528394612304332e-07, + "loss": 2.5251, + "step": 41203 + }, + { + "epoch": 7.49, + "learning_rate": 7.527848562067712e-07, + "loss": 0.145, + "step": 41204 + }, + { + "epoch": 7.49, + "learning_rate": 7.527302511831089e-07, + "loss": 0.0495, + "step": 41205 + }, + { + "epoch": 7.49, + "learning_rate": 7.526756461594467e-07, + "loss": 0.4511, + "step": 41206 + }, + { + "epoch": 7.49, + "learning_rate": 7.526210411357845e-07, + "loss": 1.6525, + "step": 41207 + }, + { + "epoch": 7.49, + "learning_rate": 7.525664361121224e-07, + "loss": 0.032, + "step": 41208 + }, + { + "epoch": 7.49, + "learning_rate": 7.525118310884601e-07, + "loss": 0.0536, + "step": 41209 + }, + { + "epoch": 7.49, + "learning_rate": 7.52457226064798e-07, + "loss": 0.0095, + "step": 41210 + }, + { + "epoch": 7.49, + "learning_rate": 7.524026210411359e-07, + "loss": 0.1111, + "step": 41211 + }, + { + "epoch": 7.49, + "learning_rate": 7.523480160174736e-07, + "loss": 0.0901, + "step": 41212 + }, + { + "epoch": 7.49, + "learning_rate": 7.522934109938115e-07, + "loss": 0.1988, + "step": 41213 + }, + { + "epoch": 7.49, + "learning_rate": 7.522388059701492e-07, + "loss": 2.0594, + "step": 41214 + }, + { + "epoch": 7.49, + "learning_rate": 7.521842009464872e-07, + "loss": 0.0631, + "step": 41215 + }, + { + "epoch": 7.5, + "learning_rate": 7.521295959228249e-07, + "loss": 0.0045, + "step": 41216 + }, + { + "epoch": 7.5, + "learning_rate": 7.520749908991627e-07, + "loss": 0.1783, + "step": 41217 + }, + { + "epoch": 7.5, + "learning_rate": 7.520203858755005e-07, + "loss": 0.0308, + "step": 41218 + }, + { + "epoch": 7.5, + "learning_rate": 7.519657808518384e-07, + "loss": 1.1219, + "step": 41219 + }, + { + "epoch": 7.5, + "learning_rate": 7.519111758281762e-07, + "loss": 0.5116, + "step": 41220 + }, + { + "epoch": 7.5, + "learning_rate": 7.51856570804514e-07, + "loss": 0.931, + "step": 41221 + }, + { + "epoch": 7.5, + "learning_rate": 7.518019657808519e-07, + "loss": 0.0162, + "step": 41222 + }, + { + "epoch": 7.5, + "learning_rate": 7.517473607571896e-07, + "loss": 0.0912, + "step": 41223 + }, + { + "epoch": 7.5, + "learning_rate": 7.516927557335275e-07, + "loss": 0.014, + "step": 41224 + }, + { + "epoch": 7.5, + "learning_rate": 7.516381507098652e-07, + "loss": 1.2139, + "step": 41225 + }, + { + "epoch": 7.5, + "learning_rate": 7.515835456862032e-07, + "loss": 1.7965, + "step": 41226 + }, + { + "epoch": 7.5, + "learning_rate": 7.51528940662541e-07, + "loss": 1.4457, + "step": 41227 + }, + { + "epoch": 7.5, + "learning_rate": 7.514743356388787e-07, + "loss": 0.0175, + "step": 41228 + }, + { + "epoch": 7.5, + "learning_rate": 7.514197306152167e-07, + "loss": 0.0115, + "step": 41229 + }, + { + "epoch": 7.5, + "learning_rate": 7.513651255915545e-07, + "loss": 0.1579, + "step": 41230 + }, + { + "epoch": 7.5, + "learning_rate": 7.513105205678922e-07, + "loss": 0.0044, + "step": 41231 + }, + { + "epoch": 7.5, + "learning_rate": 7.5125591554423e-07, + "loss": 0.8203, + "step": 41232 + }, + { + "epoch": 7.5, + "learning_rate": 7.512013105205679e-07, + "loss": 0.1148, + "step": 41233 + }, + { + "epoch": 7.5, + "learning_rate": 7.511467054969057e-07, + "loss": 1.1287, + "step": 41234 + }, + { + "epoch": 7.5, + "learning_rate": 7.510921004732435e-07, + "loss": 0.0115, + "step": 41235 + }, + { + "epoch": 7.5, + "learning_rate": 7.510374954495814e-07, + "loss": 0.0361, + "step": 41236 + }, + { + "epoch": 7.5, + "learning_rate": 7.509828904259193e-07, + "loss": 0.1965, + "step": 41237 + }, + { + "epoch": 7.5, + "learning_rate": 7.50928285402257e-07, + "loss": 0.2086, + "step": 41238 + }, + { + "epoch": 7.5, + "learning_rate": 7.508736803785947e-07, + "loss": 0.2019, + "step": 41239 + }, + { + "epoch": 7.5, + "learning_rate": 7.508190753549327e-07, + "loss": 1.2851, + "step": 41240 + }, + { + "epoch": 7.5, + "learning_rate": 7.507644703312705e-07, + "loss": 0.2573, + "step": 41241 + }, + { + "epoch": 7.5, + "learning_rate": 7.507098653076082e-07, + "loss": 1.7558, + "step": 41242 + }, + { + "epoch": 7.5, + "learning_rate": 7.506552602839462e-07, + "loss": 0.014, + "step": 41243 + }, + { + "epoch": 7.5, + "learning_rate": 7.50600655260284e-07, + "loss": 1.3221, + "step": 41244 + }, + { + "epoch": 7.5, + "learning_rate": 7.505460502366217e-07, + "loss": 0.2109, + "step": 41245 + }, + { + "epoch": 7.5, + "learning_rate": 7.504914452129596e-07, + "loss": 3.0597, + "step": 41246 + }, + { + "epoch": 7.5, + "learning_rate": 7.504368401892975e-07, + "loss": 0.0457, + "step": 41247 + }, + { + "epoch": 7.5, + "learning_rate": 7.503822351656353e-07, + "loss": 1.3109, + "step": 41248 + }, + { + "epoch": 7.5, + "learning_rate": 7.50327630141973e-07, + "loss": 1.2511, + "step": 41249 + }, + { + "epoch": 7.5, + "learning_rate": 7.502730251183109e-07, + "loss": 0.1151, + "step": 41250 + }, + { + "epoch": 7.5, + "learning_rate": 7.502184200946488e-07, + "loss": 0.0696, + "step": 41251 + }, + { + "epoch": 7.5, + "learning_rate": 7.501638150709865e-07, + "loss": 0.0089, + "step": 41252 + }, + { + "epoch": 7.5, + "learning_rate": 7.501092100473243e-07, + "loss": 0.3391, + "step": 41253 + }, + { + "epoch": 7.5, + "learning_rate": 7.500546050236623e-07, + "loss": 0.0546, + "step": 41254 + }, + { + "epoch": 7.5, + "learning_rate": 7.5e-07, + "loss": 1.094, + "step": 41255 + }, + { + "epoch": 7.5, + "learning_rate": 7.499453949763379e-07, + "loss": 1.3119, + "step": 41256 + }, + { + "epoch": 7.5, + "learning_rate": 7.498907899526757e-07, + "loss": 0.5606, + "step": 41257 + }, + { + "epoch": 7.5, + "learning_rate": 7.498361849290135e-07, + "loss": 1.062, + "step": 41258 + }, + { + "epoch": 7.5, + "learning_rate": 7.497815799053513e-07, + "loss": 0.4587, + "step": 41259 + }, + { + "epoch": 7.5, + "learning_rate": 7.497269748816891e-07, + "loss": 2.6438, + "step": 41260 + }, + { + "epoch": 7.5, + "learning_rate": 7.496723698580269e-07, + "loss": 0.1967, + "step": 41261 + }, + { + "epoch": 7.5, + "learning_rate": 7.496177648343648e-07, + "loss": 3.7637, + "step": 41262 + }, + { + "epoch": 7.5, + "learning_rate": 7.495631598107026e-07, + "loss": 0.0088, + "step": 41263 + }, + { + "epoch": 7.5, + "learning_rate": 7.495085547870405e-07, + "loss": 1.9771, + "step": 41264 + }, + { + "epoch": 7.5, + "learning_rate": 7.494539497633783e-07, + "loss": 0.0096, + "step": 41265 + }, + { + "epoch": 7.5, + "learning_rate": 7.49399344739716e-07, + "loss": 1.2617, + "step": 41266 + }, + { + "epoch": 7.5, + "learning_rate": 7.49344739716054e-07, + "loss": 0.4255, + "step": 41267 + }, + { + "epoch": 7.5, + "learning_rate": 7.492901346923917e-07, + "loss": 0.0536, + "step": 41268 + }, + { + "epoch": 7.5, + "learning_rate": 7.492355296687295e-07, + "loss": 0.0053, + "step": 41269 + }, + { + "epoch": 7.51, + "learning_rate": 7.491809246450674e-07, + "loss": 0.9045, + "step": 41270 + }, + { + "epoch": 7.51, + "learning_rate": 7.491263196214052e-07, + "loss": 0.082, + "step": 41271 + }, + { + "epoch": 7.51, + "learning_rate": 7.49071714597743e-07, + "loss": 0.1199, + "step": 41272 + }, + { + "epoch": 7.51, + "learning_rate": 7.490171095740808e-07, + "loss": 1.3726, + "step": 41273 + }, + { + "epoch": 7.51, + "learning_rate": 7.489625045504187e-07, + "loss": 3.0881, + "step": 41274 + }, + { + "epoch": 7.51, + "learning_rate": 7.489078995267565e-07, + "loss": 0.9903, + "step": 41275 + }, + { + "epoch": 7.51, + "learning_rate": 7.488532945030943e-07, + "loss": 0.0192, + "step": 41276 + }, + { + "epoch": 7.51, + "learning_rate": 7.487986894794321e-07, + "loss": 1.1811, + "step": 41277 + }, + { + "epoch": 7.51, + "learning_rate": 7.4874408445577e-07, + "loss": 0.3689, + "step": 41278 + }, + { + "epoch": 7.51, + "learning_rate": 7.486894794321078e-07, + "loss": 2.8067, + "step": 41279 + }, + { + "epoch": 7.51, + "learning_rate": 7.486348744084455e-07, + "loss": 0.1152, + "step": 41280 + }, + { + "epoch": 7.51, + "learning_rate": 7.485802693847835e-07, + "loss": 0.1163, + "step": 41281 + }, + { + "epoch": 7.51, + "learning_rate": 7.485256643611212e-07, + "loss": 0.0332, + "step": 41282 + }, + { + "epoch": 7.51, + "learning_rate": 7.48471059337459e-07, + "loss": 0.8213, + "step": 41283 + }, + { + "epoch": 7.51, + "learning_rate": 7.484164543137969e-07, + "loss": 0.0286, + "step": 41284 + }, + { + "epoch": 7.51, + "learning_rate": 7.483618492901347e-07, + "loss": 0.4909, + "step": 41285 + }, + { + "epoch": 7.51, + "learning_rate": 7.483072442664726e-07, + "loss": 1.3468, + "step": 41286 + }, + { + "epoch": 7.51, + "learning_rate": 7.482526392428104e-07, + "loss": 0.1693, + "step": 41287 + }, + { + "epoch": 7.51, + "learning_rate": 7.481980342191482e-07, + "loss": 0.0203, + "step": 41288 + }, + { + "epoch": 7.51, + "learning_rate": 7.48143429195486e-07, + "loss": 0.5003, + "step": 41289 + }, + { + "epoch": 7.51, + "learning_rate": 7.480888241718238e-07, + "loss": 1.6913, + "step": 41290 + }, + { + "epoch": 7.51, + "learning_rate": 7.480342191481616e-07, + "loss": 0.1372, + "step": 41291 + }, + { + "epoch": 7.51, + "learning_rate": 7.479796141244995e-07, + "loss": 0.0207, + "step": 41292 + }, + { + "epoch": 7.51, + "learning_rate": 7.479250091008373e-07, + "loss": 2.3257, + "step": 41293 + }, + { + "epoch": 7.51, + "learning_rate": 7.478704040771752e-07, + "loss": 0.9121, + "step": 41294 + }, + { + "epoch": 7.51, + "learning_rate": 7.47815799053513e-07, + "loss": 0.0702, + "step": 41295 + }, + { + "epoch": 7.51, + "learning_rate": 7.477611940298507e-07, + "loss": 0.0738, + "step": 41296 + }, + { + "epoch": 7.51, + "learning_rate": 7.477065890061886e-07, + "loss": 0.0697, + "step": 41297 + }, + { + "epoch": 7.51, + "learning_rate": 7.476519839825264e-07, + "loss": 0.382, + "step": 41298 + }, + { + "epoch": 7.51, + "learning_rate": 7.475973789588642e-07, + "loss": 0.2195, + "step": 41299 + }, + { + "epoch": 7.51, + "learning_rate": 7.475427739352021e-07, + "loss": 0.4849, + "step": 41300 + }, + { + "epoch": 7.51, + "learning_rate": 7.474881689115399e-07, + "loss": 0.2327, + "step": 41301 + }, + { + "epoch": 7.51, + "learning_rate": 7.474335638878777e-07, + "loss": 1.8739, + "step": 41302 + }, + { + "epoch": 7.51, + "learning_rate": 7.473789588642155e-07, + "loss": 3.4233, + "step": 41303 + }, + { + "epoch": 7.51, + "learning_rate": 7.473243538405533e-07, + "loss": 1.751, + "step": 41304 + }, + { + "epoch": 7.51, + "learning_rate": 7.472697488168912e-07, + "loss": 1.5067, + "step": 41305 + }, + { + "epoch": 7.51, + "learning_rate": 7.47215143793229e-07, + "loss": 0.0141, + "step": 41306 + }, + { + "epoch": 7.51, + "learning_rate": 7.471605387695668e-07, + "loss": 0.8936, + "step": 41307 + }, + { + "epoch": 7.51, + "learning_rate": 7.471059337459047e-07, + "loss": 0.1104, + "step": 41308 + }, + { + "epoch": 7.51, + "learning_rate": 7.470513287222425e-07, + "loss": 0.4731, + "step": 41309 + }, + { + "epoch": 7.51, + "learning_rate": 7.469967236985802e-07, + "loss": 0.0059, + "step": 41310 + }, + { + "epoch": 7.51, + "learning_rate": 7.469421186749182e-07, + "loss": 0.0439, + "step": 41311 + }, + { + "epoch": 7.51, + "learning_rate": 7.468875136512559e-07, + "loss": 0.2741, + "step": 41312 + }, + { + "epoch": 7.51, + "learning_rate": 7.468329086275937e-07, + "loss": 0.1154, + "step": 41313 + }, + { + "epoch": 7.51, + "learning_rate": 7.467783036039316e-07, + "loss": 0.8686, + "step": 41314 + }, + { + "epoch": 7.51, + "learning_rate": 7.467236985802694e-07, + "loss": 0.0998, + "step": 41315 + }, + { + "epoch": 7.51, + "learning_rate": 7.466690935566073e-07, + "loss": 0.4866, + "step": 41316 + }, + { + "epoch": 7.51, + "learning_rate": 7.46614488532945e-07, + "loss": 0.4573, + "step": 41317 + }, + { + "epoch": 7.51, + "learning_rate": 7.465598835092829e-07, + "loss": 0.0536, + "step": 41318 + }, + { + "epoch": 7.51, + "learning_rate": 7.465052784856207e-07, + "loss": 0.0614, + "step": 41319 + }, + { + "epoch": 7.51, + "learning_rate": 7.464506734619585e-07, + "loss": 0.2147, + "step": 41320 + }, + { + "epoch": 7.51, + "learning_rate": 7.463960684382963e-07, + "loss": 0.5394, + "step": 41321 + }, + { + "epoch": 7.51, + "learning_rate": 7.463414634146342e-07, + "loss": 0.0617, + "step": 41322 + }, + { + "epoch": 7.51, + "learning_rate": 7.46286858390972e-07, + "loss": 0.1562, + "step": 41323 + }, + { + "epoch": 7.51, + "learning_rate": 7.462322533673097e-07, + "loss": 1.4799, + "step": 41324 + }, + { + "epoch": 7.52, + "learning_rate": 7.461776483436477e-07, + "loss": 0.0187, + "step": 41325 + }, + { + "epoch": 7.52, + "learning_rate": 7.461230433199854e-07, + "loss": 0.0031, + "step": 41326 + }, + { + "epoch": 7.52, + "learning_rate": 7.460684382963233e-07, + "loss": 0.1498, + "step": 41327 + }, + { + "epoch": 7.52, + "learning_rate": 7.460138332726611e-07, + "loss": 0.1799, + "step": 41328 + }, + { + "epoch": 7.52, + "learning_rate": 7.459592282489989e-07, + "loss": 0.7757, + "step": 41329 + }, + { + "epoch": 7.52, + "learning_rate": 7.459046232253368e-07, + "loss": 0.1293, + "step": 41330 + }, + { + "epoch": 7.52, + "learning_rate": 7.458500182016746e-07, + "loss": 0.0078, + "step": 41331 + }, + { + "epoch": 7.52, + "learning_rate": 7.457954131780124e-07, + "loss": 0.3916, + "step": 41332 + }, + { + "epoch": 7.52, + "learning_rate": 7.457408081543502e-07, + "loss": 0.5876, + "step": 41333 + }, + { + "epoch": 7.52, + "learning_rate": 7.45686203130688e-07, + "loss": 0.0551, + "step": 41334 + }, + { + "epoch": 7.52, + "learning_rate": 7.456315981070259e-07, + "loss": 0.0612, + "step": 41335 + }, + { + "epoch": 7.52, + "learning_rate": 7.455769930833637e-07, + "loss": 0.204, + "step": 41336 + }, + { + "epoch": 7.52, + "learning_rate": 7.455223880597015e-07, + "loss": 1.516, + "step": 41337 + }, + { + "epoch": 7.52, + "learning_rate": 7.454677830360394e-07, + "loss": 1.5505, + "step": 41338 + }, + { + "epoch": 7.52, + "learning_rate": 7.454131780123772e-07, + "loss": 1.2332, + "step": 41339 + }, + { + "epoch": 7.52, + "learning_rate": 7.453585729887149e-07, + "loss": 0.5875, + "step": 41340 + }, + { + "epoch": 7.52, + "learning_rate": 7.453039679650528e-07, + "loss": 0.5207, + "step": 41341 + }, + { + "epoch": 7.52, + "learning_rate": 7.452493629413906e-07, + "loss": 0.4094, + "step": 41342 + }, + { + "epoch": 7.52, + "learning_rate": 7.451947579177284e-07, + "loss": 1.2104, + "step": 41343 + }, + { + "epoch": 7.52, + "learning_rate": 7.451401528940663e-07, + "loss": 0.1023, + "step": 41344 + }, + { + "epoch": 7.52, + "learning_rate": 7.450855478704041e-07, + "loss": 0.0087, + "step": 41345 + }, + { + "epoch": 7.52, + "learning_rate": 7.45030942846742e-07, + "loss": 0.9121, + "step": 41346 + }, + { + "epoch": 7.52, + "learning_rate": 7.449763378230797e-07, + "loss": 2.0681, + "step": 41347 + }, + { + "epoch": 7.52, + "learning_rate": 7.449217327994175e-07, + "loss": 0.0084, + "step": 41348 + }, + { + "epoch": 7.52, + "learning_rate": 7.448671277757554e-07, + "loss": 0.1283, + "step": 41349 + }, + { + "epoch": 7.52, + "learning_rate": 7.448125227520932e-07, + "loss": 0.1854, + "step": 41350 + }, + { + "epoch": 7.52, + "learning_rate": 7.44757917728431e-07, + "loss": 0.0054, + "step": 41351 + }, + { + "epoch": 7.52, + "learning_rate": 7.447033127047689e-07, + "loss": 0.1248, + "step": 41352 + }, + { + "epoch": 7.52, + "learning_rate": 7.446487076811067e-07, + "loss": 0.0479, + "step": 41353 + }, + { + "epoch": 7.52, + "learning_rate": 7.445941026574445e-07, + "loss": 0.0215, + "step": 41354 + }, + { + "epoch": 7.52, + "learning_rate": 7.445394976337824e-07, + "loss": 0.0107, + "step": 41355 + }, + { + "epoch": 7.52, + "learning_rate": 7.444848926101201e-07, + "loss": 0.0797, + "step": 41356 + }, + { + "epoch": 7.52, + "learning_rate": 7.44430287586458e-07, + "loss": 0.0905, + "step": 41357 + }, + { + "epoch": 7.52, + "learning_rate": 7.443756825627958e-07, + "loss": 0.6353, + "step": 41358 + }, + { + "epoch": 7.52, + "learning_rate": 7.443210775391336e-07, + "loss": 0.0218, + "step": 41359 + }, + { + "epoch": 7.52, + "learning_rate": 7.442664725154715e-07, + "loss": 0.2284, + "step": 41360 + }, + { + "epoch": 7.52, + "learning_rate": 7.442118674918092e-07, + "loss": 0.018, + "step": 41361 + }, + { + "epoch": 7.52, + "learning_rate": 7.441572624681472e-07, + "loss": 0.0073, + "step": 41362 + }, + { + "epoch": 7.52, + "learning_rate": 7.441026574444849e-07, + "loss": 0.1929, + "step": 41363 + }, + { + "epoch": 7.52, + "learning_rate": 7.440480524208227e-07, + "loss": 0.056, + "step": 41364 + }, + { + "epoch": 7.52, + "learning_rate": 7.439934473971606e-07, + "loss": 0.0147, + "step": 41365 + }, + { + "epoch": 7.52, + "learning_rate": 7.439388423734984e-07, + "loss": 1.1405, + "step": 41366 + }, + { + "epoch": 7.52, + "learning_rate": 7.438842373498362e-07, + "loss": 0.6679, + "step": 41367 + }, + { + "epoch": 7.52, + "learning_rate": 7.43829632326174e-07, + "loss": 0.0414, + "step": 41368 + }, + { + "epoch": 7.52, + "learning_rate": 7.437750273025119e-07, + "loss": 0.0882, + "step": 41369 + }, + { + "epoch": 7.52, + "learning_rate": 7.437204222788496e-07, + "loss": 0.4179, + "step": 41370 + }, + { + "epoch": 7.52, + "learning_rate": 7.436658172551875e-07, + "loss": 2.8685, + "step": 41371 + }, + { + "epoch": 7.52, + "learning_rate": 7.436112122315253e-07, + "loss": 1.1228, + "step": 41372 + }, + { + "epoch": 7.52, + "learning_rate": 7.435566072078632e-07, + "loss": 0.6582, + "step": 41373 + }, + { + "epoch": 7.52, + "learning_rate": 7.43502002184201e-07, + "loss": 0.006, + "step": 41374 + }, + { + "epoch": 7.52, + "learning_rate": 7.434473971605388e-07, + "loss": 0.5897, + "step": 41375 + }, + { + "epoch": 7.52, + "learning_rate": 7.433927921368767e-07, + "loss": 2.9357, + "step": 41376 + }, + { + "epoch": 7.52, + "learning_rate": 7.433381871132144e-07, + "loss": 0.1659, + "step": 41377 + }, + { + "epoch": 7.52, + "learning_rate": 7.432835820895522e-07, + "loss": 0.379, + "step": 41378 + }, + { + "epoch": 7.52, + "learning_rate": 7.432289770658901e-07, + "loss": 0.2127, + "step": 41379 + }, + { + "epoch": 7.53, + "learning_rate": 7.431743720422279e-07, + "loss": 1.1533, + "step": 41380 + }, + { + "epoch": 7.53, + "learning_rate": 7.431197670185657e-07, + "loss": 0.0059, + "step": 41381 + }, + { + "epoch": 7.53, + "learning_rate": 7.430651619949036e-07, + "loss": 0.4096, + "step": 41382 + }, + { + "epoch": 7.53, + "learning_rate": 7.430105569712414e-07, + "loss": 0.4658, + "step": 41383 + }, + { + "epoch": 7.53, + "learning_rate": 7.429559519475792e-07, + "loss": 2.6246, + "step": 41384 + }, + { + "epoch": 7.53, + "learning_rate": 7.42901346923917e-07, + "loss": 0.6021, + "step": 41385 + }, + { + "epoch": 7.53, + "learning_rate": 7.428467419002548e-07, + "loss": 0.5396, + "step": 41386 + }, + { + "epoch": 7.53, + "learning_rate": 7.427921368765927e-07, + "loss": 0.1651, + "step": 41387 + }, + { + "epoch": 7.53, + "learning_rate": 7.427375318529305e-07, + "loss": 1.0738, + "step": 41388 + }, + { + "epoch": 7.53, + "learning_rate": 7.426829268292683e-07, + "loss": 0.0111, + "step": 41389 + }, + { + "epoch": 7.53, + "learning_rate": 7.426283218056062e-07, + "loss": 0.0246, + "step": 41390 + }, + { + "epoch": 7.53, + "learning_rate": 7.425737167819439e-07, + "loss": 0.0897, + "step": 41391 + }, + { + "epoch": 7.53, + "learning_rate": 7.425191117582817e-07, + "loss": 0.41, + "step": 41392 + }, + { + "epoch": 7.53, + "learning_rate": 7.424645067346196e-07, + "loss": 0.8931, + "step": 41393 + }, + { + "epoch": 7.53, + "learning_rate": 7.424099017109574e-07, + "loss": 0.2464, + "step": 41394 + }, + { + "epoch": 7.53, + "learning_rate": 7.423552966872953e-07, + "loss": 0.008, + "step": 41395 + }, + { + "epoch": 7.53, + "learning_rate": 7.423006916636331e-07, + "loss": 2.2928, + "step": 41396 + }, + { + "epoch": 7.53, + "learning_rate": 7.422460866399709e-07, + "loss": 0.0484, + "step": 41397 + }, + { + "epoch": 7.53, + "learning_rate": 7.421914816163087e-07, + "loss": 0.172, + "step": 41398 + }, + { + "epoch": 7.53, + "learning_rate": 7.421368765926466e-07, + "loss": 0.0492, + "step": 41399 + }, + { + "epoch": 7.53, + "learning_rate": 7.420822715689843e-07, + "loss": 0.3738, + "step": 41400 + }, + { + "epoch": 7.53, + "learning_rate": 7.420276665453222e-07, + "loss": 0.503, + "step": 41401 + }, + { + "epoch": 7.53, + "learning_rate": 7.4197306152166e-07, + "loss": 0.0224, + "step": 41402 + }, + { + "epoch": 7.53, + "learning_rate": 7.419184564979979e-07, + "loss": 1.319, + "step": 41403 + }, + { + "epoch": 7.53, + "learning_rate": 7.418638514743357e-07, + "loss": 0.0135, + "step": 41404 + }, + { + "epoch": 7.53, + "learning_rate": 7.418092464506734e-07, + "loss": 0.389, + "step": 41405 + }, + { + "epoch": 7.53, + "learning_rate": 7.417546414270114e-07, + "loss": 0.0948, + "step": 41406 + }, + { + "epoch": 7.53, + "learning_rate": 7.417000364033491e-07, + "loss": 0.0239, + "step": 41407 + }, + { + "epoch": 7.53, + "learning_rate": 7.416454313796869e-07, + "loss": 0.2981, + "step": 41408 + }, + { + "epoch": 7.53, + "learning_rate": 7.415908263560248e-07, + "loss": 2.5834, + "step": 41409 + }, + { + "epoch": 7.53, + "learning_rate": 7.415362213323626e-07, + "loss": 0.6991, + "step": 41410 + }, + { + "epoch": 7.53, + "learning_rate": 7.414816163087004e-07, + "loss": 0.1403, + "step": 41411 + }, + { + "epoch": 7.53, + "learning_rate": 7.414270112850382e-07, + "loss": 0.5711, + "step": 41412 + }, + { + "epoch": 7.53, + "learning_rate": 7.413724062613761e-07, + "loss": 1.043, + "step": 41413 + }, + { + "epoch": 7.53, + "learning_rate": 7.413178012377139e-07, + "loss": 0.7692, + "step": 41414 + }, + { + "epoch": 7.53, + "learning_rate": 7.412631962140517e-07, + "loss": 2.1928, + "step": 41415 + }, + { + "epoch": 7.53, + "learning_rate": 7.412085911903895e-07, + "loss": 1.7413, + "step": 41416 + }, + { + "epoch": 7.53, + "learning_rate": 7.411539861667274e-07, + "loss": 0.0188, + "step": 41417 + }, + { + "epoch": 7.53, + "learning_rate": 7.410993811430652e-07, + "loss": 0.0043, + "step": 41418 + }, + { + "epoch": 7.53, + "learning_rate": 7.41044776119403e-07, + "loss": 1.2607, + "step": 41419 + }, + { + "epoch": 7.53, + "learning_rate": 7.409901710957409e-07, + "loss": 0.2881, + "step": 41420 + }, + { + "epoch": 7.53, + "learning_rate": 7.409355660720786e-07, + "loss": 0.08, + "step": 41421 + }, + { + "epoch": 7.53, + "learning_rate": 7.408809610484164e-07, + "loss": 0.0539, + "step": 41422 + }, + { + "epoch": 7.53, + "learning_rate": 7.408263560247543e-07, + "loss": 0.0102, + "step": 41423 + }, + { + "epoch": 7.53, + "learning_rate": 7.407717510010921e-07, + "loss": 0.5839, + "step": 41424 + }, + { + "epoch": 7.53, + "learning_rate": 7.4071714597743e-07, + "loss": 0.0092, + "step": 41425 + }, + { + "epoch": 7.53, + "learning_rate": 7.406625409537678e-07, + "loss": 0.2217, + "step": 41426 + }, + { + "epoch": 7.53, + "learning_rate": 7.406079359301056e-07, + "loss": 1.0483, + "step": 41427 + }, + { + "epoch": 7.53, + "learning_rate": 7.405533309064434e-07, + "loss": 0.0081, + "step": 41428 + }, + { + "epoch": 7.53, + "learning_rate": 7.404987258827812e-07, + "loss": 0.0475, + "step": 41429 + }, + { + "epoch": 7.53, + "learning_rate": 7.40444120859119e-07, + "loss": 0.2442, + "step": 41430 + }, + { + "epoch": 7.53, + "learning_rate": 7.403895158354569e-07, + "loss": 0.717, + "step": 41431 + }, + { + "epoch": 7.53, + "learning_rate": 7.403349108117946e-07, + "loss": 0.2965, + "step": 41432 + }, + { + "epoch": 7.53, + "learning_rate": 7.402803057881326e-07, + "loss": 0.5943, + "step": 41433 + }, + { + "epoch": 7.53, + "learning_rate": 7.402257007644704e-07, + "loss": 0.3261, + "step": 41434 + }, + { + "epoch": 7.54, + "learning_rate": 7.401710957408081e-07, + "loss": 1.4061, + "step": 41435 + }, + { + "epoch": 7.54, + "learning_rate": 7.401164907171461e-07, + "loss": 3.027, + "step": 41436 + }, + { + "epoch": 7.54, + "learning_rate": 7.400618856934838e-07, + "loss": 0.1657, + "step": 41437 + }, + { + "epoch": 7.54, + "learning_rate": 7.400072806698216e-07, + "loss": 0.3137, + "step": 41438 + }, + { + "epoch": 7.54, + "learning_rate": 7.399526756461595e-07, + "loss": 0.3252, + "step": 41439 + }, + { + "epoch": 7.54, + "learning_rate": 7.398980706224973e-07, + "loss": 1.703, + "step": 41440 + }, + { + "epoch": 7.54, + "learning_rate": 7.398434655988352e-07, + "loss": 0.8045, + "step": 41441 + }, + { + "epoch": 7.54, + "learning_rate": 7.397888605751729e-07, + "loss": 1.1573, + "step": 41442 + }, + { + "epoch": 7.54, + "learning_rate": 7.397342555515108e-07, + "loss": 0.7692, + "step": 41443 + }, + { + "epoch": 7.54, + "learning_rate": 7.396796505278486e-07, + "loss": 0.2178, + "step": 41444 + }, + { + "epoch": 7.54, + "learning_rate": 7.396250455041864e-07, + "loss": 1.7526, + "step": 41445 + }, + { + "epoch": 7.54, + "learning_rate": 7.395704404805242e-07, + "loss": 0.0763, + "step": 41446 + }, + { + "epoch": 7.54, + "learning_rate": 7.395158354568621e-07, + "loss": 0.8176, + "step": 41447 + }, + { + "epoch": 7.54, + "learning_rate": 7.394612304331999e-07, + "loss": 0.0051, + "step": 41448 + }, + { + "epoch": 7.54, + "learning_rate": 7.394066254095376e-07, + "loss": 0.743, + "step": 41449 + }, + { + "epoch": 7.54, + "learning_rate": 7.393520203858756e-07, + "loss": 0.0121, + "step": 41450 + }, + { + "epoch": 7.54, + "learning_rate": 7.392974153622133e-07, + "loss": 0.4973, + "step": 41451 + }, + { + "epoch": 7.54, + "learning_rate": 7.392428103385512e-07, + "loss": 0.778, + "step": 41452 + }, + { + "epoch": 7.54, + "learning_rate": 7.39188205314889e-07, + "loss": 0.7037, + "step": 41453 + }, + { + "epoch": 7.54, + "learning_rate": 7.391336002912268e-07, + "loss": 0.7026, + "step": 41454 + }, + { + "epoch": 7.54, + "learning_rate": 7.390789952675647e-07, + "loss": 0.2082, + "step": 41455 + }, + { + "epoch": 7.54, + "learning_rate": 7.390243902439024e-07, + "loss": 0.0554, + "step": 41456 + }, + { + "epoch": 7.54, + "learning_rate": 7.389697852202403e-07, + "loss": 0.026, + "step": 41457 + }, + { + "epoch": 7.54, + "learning_rate": 7.389151801965781e-07, + "loss": 0.9274, + "step": 41458 + }, + { + "epoch": 7.54, + "learning_rate": 7.388605751729159e-07, + "loss": 0.2882, + "step": 41459 + }, + { + "epoch": 7.54, + "learning_rate": 7.388059701492537e-07, + "loss": 0.0063, + "step": 41460 + }, + { + "epoch": 7.54, + "learning_rate": 7.387513651255916e-07, + "loss": 1.0517, + "step": 41461 + }, + { + "epoch": 7.54, + "learning_rate": 7.386967601019294e-07, + "loss": 1.0174, + "step": 41462 + }, + { + "epoch": 7.54, + "learning_rate": 7.386421550782673e-07, + "loss": 0.0219, + "step": 41463 + }, + { + "epoch": 7.54, + "learning_rate": 7.385875500546051e-07, + "loss": 0.0105, + "step": 41464 + }, + { + "epoch": 7.54, + "learning_rate": 7.385329450309428e-07, + "loss": 0.8762, + "step": 41465 + }, + { + "epoch": 7.54, + "learning_rate": 7.384783400072807e-07, + "loss": 0.0238, + "step": 41466 + }, + { + "epoch": 7.54, + "learning_rate": 7.384237349836185e-07, + "loss": 0.0142, + "step": 41467 + }, + { + "epoch": 7.54, + "learning_rate": 7.383691299599563e-07, + "loss": 0.1409, + "step": 41468 + }, + { + "epoch": 7.54, + "learning_rate": 7.383145249362941e-07, + "loss": 0.1305, + "step": 41469 + }, + { + "epoch": 7.54, + "learning_rate": 7.38259919912632e-07, + "loss": 0.155, + "step": 41470 + }, + { + "epoch": 7.54, + "learning_rate": 7.382053148889699e-07, + "loss": 0.4718, + "step": 41471 + }, + { + "epoch": 7.54, + "learning_rate": 7.381507098653076e-07, + "loss": 2.1009, + "step": 41472 + }, + { + "epoch": 7.54, + "learning_rate": 7.380961048416454e-07, + "loss": 1.1046, + "step": 41473 + }, + { + "epoch": 7.54, + "learning_rate": 7.380414998179833e-07, + "loss": 0.0065, + "step": 41474 + }, + { + "epoch": 7.54, + "learning_rate": 7.379868947943211e-07, + "loss": 0.0209, + "step": 41475 + }, + { + "epoch": 7.54, + "learning_rate": 7.379322897706588e-07, + "loss": 0.1727, + "step": 41476 + }, + { + "epoch": 7.54, + "learning_rate": 7.378776847469968e-07, + "loss": 0.0656, + "step": 41477 + }, + { + "epoch": 7.54, + "learning_rate": 7.378230797233346e-07, + "loss": 0.0927, + "step": 41478 + }, + { + "epoch": 7.54, + "learning_rate": 7.377684746996723e-07, + "loss": 0.0045, + "step": 41479 + }, + { + "epoch": 7.54, + "learning_rate": 7.377138696760103e-07, + "loss": 0.9957, + "step": 41480 + }, + { + "epoch": 7.54, + "learning_rate": 7.37659264652348e-07, + "loss": 0.346, + "step": 41481 + }, + { + "epoch": 7.54, + "learning_rate": 7.376046596286859e-07, + "loss": 0.1927, + "step": 41482 + }, + { + "epoch": 7.54, + "learning_rate": 7.375500546050237e-07, + "loss": 0.2948, + "step": 41483 + }, + { + "epoch": 7.54, + "learning_rate": 7.374954495813615e-07, + "loss": 0.0086, + "step": 41484 + }, + { + "epoch": 7.54, + "learning_rate": 7.374408445576994e-07, + "loss": 0.5101, + "step": 41485 + }, + { + "epoch": 7.54, + "learning_rate": 7.373862395340371e-07, + "loss": 1.039, + "step": 41486 + }, + { + "epoch": 7.54, + "learning_rate": 7.37331634510375e-07, + "loss": 2.7648, + "step": 41487 + }, + { + "epoch": 7.54, + "learning_rate": 7.372770294867128e-07, + "loss": 0.0094, + "step": 41488 + }, + { + "epoch": 7.54, + "learning_rate": 7.372224244630506e-07, + "loss": 0.357, + "step": 41489 + }, + { + "epoch": 7.55, + "learning_rate": 7.371678194393884e-07, + "loss": 0.7813, + "step": 41490 + }, + { + "epoch": 7.55, + "learning_rate": 7.371132144157263e-07, + "loss": 0.0124, + "step": 41491 + }, + { + "epoch": 7.55, + "learning_rate": 7.370586093920641e-07, + "loss": 0.0601, + "step": 41492 + }, + { + "epoch": 7.55, + "learning_rate": 7.370040043684019e-07, + "loss": 0.1541, + "step": 41493 + }, + { + "epoch": 7.55, + "learning_rate": 7.369493993447398e-07, + "loss": 0.1538, + "step": 41494 + }, + { + "epoch": 7.55, + "learning_rate": 7.368947943210775e-07, + "loss": 0.0757, + "step": 41495 + }, + { + "epoch": 7.55, + "learning_rate": 7.368401892974154e-07, + "loss": 0.2808, + "step": 41496 + }, + { + "epoch": 7.55, + "learning_rate": 7.367855842737532e-07, + "loss": 0.005, + "step": 41497 + }, + { + "epoch": 7.55, + "learning_rate": 7.36730979250091e-07, + "loss": 1.3034, + "step": 41498 + }, + { + "epoch": 7.55, + "learning_rate": 7.366763742264289e-07, + "loss": 1.9021, + "step": 41499 + }, + { + "epoch": 7.55, + "learning_rate": 7.366217692027667e-07, + "loss": 0.0471, + "step": 41500 + }, + { + "epoch": 7.55, + "learning_rate": 7.365671641791046e-07, + "loss": 0.0137, + "step": 41501 + }, + { + "epoch": 7.55, + "learning_rate": 7.365125591554423e-07, + "loss": 2.3293, + "step": 41502 + }, + { + "epoch": 7.55, + "learning_rate": 7.364579541317801e-07, + "loss": 0.5399, + "step": 41503 + }, + { + "epoch": 7.55, + "learning_rate": 7.36403349108118e-07, + "loss": 0.5599, + "step": 41504 + }, + { + "epoch": 7.55, + "learning_rate": 7.363487440844558e-07, + "loss": 0.5472, + "step": 41505 + }, + { + "epoch": 7.55, + "learning_rate": 7.362941390607936e-07, + "loss": 1.511, + "step": 41506 + }, + { + "epoch": 7.55, + "learning_rate": 7.362395340371315e-07, + "loss": 0.007, + "step": 41507 + }, + { + "epoch": 7.55, + "learning_rate": 7.361849290134693e-07, + "loss": 0.2887, + "step": 41508 + }, + { + "epoch": 7.55, + "learning_rate": 7.36130323989807e-07, + "loss": 1.054, + "step": 41509 + }, + { + "epoch": 7.55, + "learning_rate": 7.360757189661449e-07, + "loss": 2.2896, + "step": 41510 + }, + { + "epoch": 7.55, + "learning_rate": 7.360211139424827e-07, + "loss": 0.9889, + "step": 41511 + }, + { + "epoch": 7.55, + "learning_rate": 7.359665089188206e-07, + "loss": 0.0043, + "step": 41512 + }, + { + "epoch": 7.55, + "learning_rate": 7.359119038951583e-07, + "loss": 0.6534, + "step": 41513 + }, + { + "epoch": 7.55, + "learning_rate": 7.358572988714962e-07, + "loss": 0.0472, + "step": 41514 + }, + { + "epoch": 7.55, + "learning_rate": 7.358026938478341e-07, + "loss": 0.0396, + "step": 41515 + }, + { + "epoch": 7.55, + "learning_rate": 7.357480888241718e-07, + "loss": 0.1639, + "step": 41516 + }, + { + "epoch": 7.55, + "learning_rate": 7.356934838005096e-07, + "loss": 2.1155, + "step": 41517 + }, + { + "epoch": 7.55, + "learning_rate": 7.356388787768475e-07, + "loss": 2.2229, + "step": 41518 + }, + { + "epoch": 7.55, + "learning_rate": 7.355842737531853e-07, + "loss": 0.4288, + "step": 41519 + }, + { + "epoch": 7.55, + "learning_rate": 7.355296687295231e-07, + "loss": 0.4091, + "step": 41520 + }, + { + "epoch": 7.55, + "learning_rate": 7.35475063705861e-07, + "loss": 0.0086, + "step": 41521 + }, + { + "epoch": 7.55, + "learning_rate": 7.354204586821988e-07, + "loss": 0.0272, + "step": 41522 + }, + { + "epoch": 7.55, + "learning_rate": 7.353658536585366e-07, + "loss": 0.227, + "step": 41523 + }, + { + "epoch": 7.55, + "learning_rate": 7.353112486348745e-07, + "loss": 1.4322, + "step": 41524 + }, + { + "epoch": 7.55, + "learning_rate": 7.352566436112122e-07, + "loss": 0.1209, + "step": 41525 + }, + { + "epoch": 7.55, + "learning_rate": 7.352020385875501e-07, + "loss": 0.2517, + "step": 41526 + }, + { + "epoch": 7.55, + "learning_rate": 7.351474335638879e-07, + "loss": 0.2542, + "step": 41527 + }, + { + "epoch": 7.55, + "learning_rate": 7.350928285402257e-07, + "loss": 0.2653, + "step": 41528 + }, + { + "epoch": 7.55, + "learning_rate": 7.350382235165636e-07, + "loss": 0.4274, + "step": 41529 + }, + { + "epoch": 7.55, + "learning_rate": 7.349836184929013e-07, + "loss": 2.1097, + "step": 41530 + }, + { + "epoch": 7.55, + "learning_rate": 7.349290134692393e-07, + "loss": 0.7724, + "step": 41531 + }, + { + "epoch": 7.55, + "learning_rate": 7.34874408445577e-07, + "loss": 0.4146, + "step": 41532 + }, + { + "epoch": 7.55, + "learning_rate": 7.348198034219148e-07, + "loss": 0.1034, + "step": 41533 + }, + { + "epoch": 7.55, + "learning_rate": 7.347651983982527e-07, + "loss": 3.3931, + "step": 41534 + }, + { + "epoch": 7.55, + "learning_rate": 7.347105933745905e-07, + "loss": 0.2558, + "step": 41535 + }, + { + "epoch": 7.55, + "learning_rate": 7.346559883509283e-07, + "loss": 0.0162, + "step": 41536 + }, + { + "epoch": 7.55, + "learning_rate": 7.346013833272661e-07, + "loss": 1.4358, + "step": 41537 + }, + { + "epoch": 7.55, + "learning_rate": 7.34546778303604e-07, + "loss": 0.0381, + "step": 41538 + }, + { + "epoch": 7.55, + "learning_rate": 7.344921732799417e-07, + "loss": 0.132, + "step": 41539 + }, + { + "epoch": 7.55, + "learning_rate": 7.344375682562796e-07, + "loss": 0.0252, + "step": 41540 + }, + { + "epoch": 7.55, + "learning_rate": 7.343829632326174e-07, + "loss": 0.387, + "step": 41541 + }, + { + "epoch": 7.55, + "learning_rate": 7.343283582089553e-07, + "loss": 0.0597, + "step": 41542 + }, + { + "epoch": 7.55, + "learning_rate": 7.342737531852931e-07, + "loss": 0.6896, + "step": 41543 + }, + { + "epoch": 7.55, + "learning_rate": 7.342191481616309e-07, + "loss": 2.0766, + "step": 41544 + }, + { + "epoch": 7.56, + "learning_rate": 7.341645431379688e-07, + "loss": 0.1775, + "step": 41545 + }, + { + "epoch": 7.56, + "learning_rate": 7.341099381143065e-07, + "loss": 0.0349, + "step": 41546 + }, + { + "epoch": 7.56, + "learning_rate": 7.340553330906443e-07, + "loss": 0.6082, + "step": 41547 + }, + { + "epoch": 7.56, + "learning_rate": 7.340007280669822e-07, + "loss": 1.2902, + "step": 41548 + }, + { + "epoch": 7.56, + "learning_rate": 7.3394612304332e-07, + "loss": 0.0041, + "step": 41549 + }, + { + "epoch": 7.56, + "learning_rate": 7.338915180196577e-07, + "loss": 0.0308, + "step": 41550 + }, + { + "epoch": 7.56, + "learning_rate": 7.338369129959957e-07, + "loss": 0.1748, + "step": 41551 + }, + { + "epoch": 7.56, + "learning_rate": 7.337823079723335e-07, + "loss": 0.6945, + "step": 41552 + }, + { + "epoch": 7.56, + "learning_rate": 7.337277029486713e-07, + "loss": 1.2212, + "step": 41553 + }, + { + "epoch": 7.56, + "learning_rate": 7.336730979250091e-07, + "loss": 1.2487, + "step": 41554 + }, + { + "epoch": 7.56, + "learning_rate": 7.336184929013469e-07, + "loss": 1.5641, + "step": 41555 + }, + { + "epoch": 7.56, + "learning_rate": 7.335638878776848e-07, + "loss": 0.1255, + "step": 41556 + }, + { + "epoch": 7.56, + "learning_rate": 7.335092828540225e-07, + "loss": 0.3817, + "step": 41557 + }, + { + "epoch": 7.56, + "learning_rate": 7.334546778303604e-07, + "loss": 0.7459, + "step": 41558 + }, + { + "epoch": 7.56, + "learning_rate": 7.334000728066983e-07, + "loss": 0.2074, + "step": 41559 + }, + { + "epoch": 7.56, + "learning_rate": 7.33345467783036e-07, + "loss": 0.0635, + "step": 41560 + }, + { + "epoch": 7.56, + "learning_rate": 7.332908627593739e-07, + "loss": 0.8042, + "step": 41561 + }, + { + "epoch": 7.56, + "learning_rate": 7.332362577357117e-07, + "loss": 2.0364, + "step": 41562 + }, + { + "epoch": 7.56, + "learning_rate": 7.331816527120495e-07, + "loss": 1.3695, + "step": 41563 + }, + { + "epoch": 7.56, + "learning_rate": 7.331270476883874e-07, + "loss": 0.4792, + "step": 41564 + }, + { + "epoch": 7.56, + "learning_rate": 7.330724426647252e-07, + "loss": 2.2068, + "step": 41565 + }, + { + "epoch": 7.56, + "learning_rate": 7.33017837641063e-07, + "loss": 0.9533, + "step": 41566 + }, + { + "epoch": 7.56, + "learning_rate": 7.329632326174008e-07, + "loss": 1.4008, + "step": 41567 + }, + { + "epoch": 7.56, + "learning_rate": 7.329086275937387e-07, + "loss": 0.0197, + "step": 41568 + }, + { + "epoch": 7.56, + "learning_rate": 7.328540225700764e-07, + "loss": 0.3719, + "step": 41569 + }, + { + "epoch": 7.56, + "learning_rate": 7.327994175464143e-07, + "loss": 1.0697, + "step": 41570 + }, + { + "epoch": 7.56, + "learning_rate": 7.327448125227521e-07, + "loss": 0.0289, + "step": 41571 + }, + { + "epoch": 7.56, + "learning_rate": 7.3269020749909e-07, + "loss": 1.1955, + "step": 41572 + }, + { + "epoch": 7.56, + "learning_rate": 7.326356024754278e-07, + "loss": 1.1675, + "step": 41573 + }, + { + "epoch": 7.56, + "learning_rate": 7.325809974517655e-07, + "loss": 1.9457, + "step": 41574 + }, + { + "epoch": 7.56, + "learning_rate": 7.325263924281035e-07, + "loss": 0.3449, + "step": 41575 + }, + { + "epoch": 7.56, + "learning_rate": 7.324717874044412e-07, + "loss": 0.1648, + "step": 41576 + }, + { + "epoch": 7.56, + "learning_rate": 7.32417182380779e-07, + "loss": 0.7844, + "step": 41577 + }, + { + "epoch": 7.56, + "learning_rate": 7.323625773571169e-07, + "loss": 0.9511, + "step": 41578 + }, + { + "epoch": 7.56, + "learning_rate": 7.323079723334547e-07, + "loss": 0.3061, + "step": 41579 + }, + { + "epoch": 7.56, + "learning_rate": 7.322533673097926e-07, + "loss": 0.0871, + "step": 41580 + }, + { + "epoch": 7.56, + "learning_rate": 7.321987622861303e-07, + "loss": 2.1435, + "step": 41581 + }, + { + "epoch": 7.56, + "learning_rate": 7.321441572624682e-07, + "loss": 2.0621, + "step": 41582 + }, + { + "epoch": 7.56, + "learning_rate": 7.32089552238806e-07, + "loss": 0.1488, + "step": 41583 + }, + { + "epoch": 7.56, + "learning_rate": 7.320349472151438e-07, + "loss": 0.0072, + "step": 41584 + }, + { + "epoch": 7.56, + "learning_rate": 7.319803421914816e-07, + "loss": 0.6724, + "step": 41585 + }, + { + "epoch": 7.56, + "learning_rate": 7.319257371678195e-07, + "loss": 0.238, + "step": 41586 + }, + { + "epoch": 7.56, + "learning_rate": 7.318711321441572e-07, + "loss": 0.1264, + "step": 41587 + }, + { + "epoch": 7.56, + "learning_rate": 7.318165271204951e-07, + "loss": 1.3357, + "step": 41588 + }, + { + "epoch": 7.56, + "learning_rate": 7.31761922096833e-07, + "loss": 0.0318, + "step": 41589 + }, + { + "epoch": 7.56, + "learning_rate": 7.317073170731707e-07, + "loss": 1.4613, + "step": 41590 + }, + { + "epoch": 7.56, + "learning_rate": 7.316527120495086e-07, + "loss": 0.0164, + "step": 41591 + }, + { + "epoch": 7.56, + "learning_rate": 7.315981070258464e-07, + "loss": 1.628, + "step": 41592 + }, + { + "epoch": 7.56, + "learning_rate": 7.315435020021842e-07, + "loss": 0.2694, + "step": 41593 + }, + { + "epoch": 7.56, + "learning_rate": 7.31488896978522e-07, + "loss": 0.5806, + "step": 41594 + }, + { + "epoch": 7.56, + "learning_rate": 7.314342919548599e-07, + "loss": 0.4213, + "step": 41595 + }, + { + "epoch": 7.56, + "learning_rate": 7.313796869311977e-07, + "loss": 0.857, + "step": 41596 + }, + { + "epoch": 7.56, + "learning_rate": 7.313250819075355e-07, + "loss": 0.0675, + "step": 41597 + }, + { + "epoch": 7.56, + "learning_rate": 7.312704768838733e-07, + "loss": 0.5023, + "step": 41598 + }, + { + "epoch": 7.56, + "learning_rate": 7.312158718602111e-07, + "loss": 0.2462, + "step": 41599 + }, + { + "epoch": 7.57, + "learning_rate": 7.31161266836549e-07, + "loss": 0.0101, + "step": 41600 + }, + { + "epoch": 7.57, + "learning_rate": 7.311066618128867e-07, + "loss": 0.0802, + "step": 41601 + }, + { + "epoch": 7.57, + "learning_rate": 7.310520567892247e-07, + "loss": 0.7952, + "step": 41602 + }, + { + "epoch": 7.57, + "learning_rate": 7.309974517655625e-07, + "loss": 0.3425, + "step": 41603 + }, + { + "epoch": 7.57, + "learning_rate": 7.309428467419002e-07, + "loss": 0.0494, + "step": 41604 + }, + { + "epoch": 7.57, + "learning_rate": 7.308882417182381e-07, + "loss": 0.0082, + "step": 41605 + }, + { + "epoch": 7.57, + "learning_rate": 7.308336366945759e-07, + "loss": 2.9927, + "step": 41606 + }, + { + "epoch": 7.57, + "learning_rate": 7.307790316709137e-07, + "loss": 0.1425, + "step": 41607 + }, + { + "epoch": 7.57, + "learning_rate": 7.307244266472516e-07, + "loss": 0.2802, + "step": 41608 + }, + { + "epoch": 7.57, + "learning_rate": 7.306698216235894e-07, + "loss": 0.0088, + "step": 41609 + }, + { + "epoch": 7.57, + "learning_rate": 7.306152165999273e-07, + "loss": 0.0705, + "step": 41610 + }, + { + "epoch": 7.57, + "learning_rate": 7.30560611576265e-07, + "loss": 0.9581, + "step": 41611 + }, + { + "epoch": 7.57, + "learning_rate": 7.305060065526029e-07, + "loss": 2.8633, + "step": 41612 + }, + { + "epoch": 7.57, + "learning_rate": 7.304514015289407e-07, + "loss": 0.4368, + "step": 41613 + }, + { + "epoch": 7.57, + "learning_rate": 7.303967965052785e-07, + "loss": 2.7614, + "step": 41614 + }, + { + "epoch": 7.57, + "learning_rate": 7.303421914816163e-07, + "loss": 0.9087, + "step": 41615 + }, + { + "epoch": 7.57, + "learning_rate": 7.302875864579542e-07, + "loss": 1.2773, + "step": 41616 + }, + { + "epoch": 7.57, + "learning_rate": 7.30232981434292e-07, + "loss": 0.1751, + "step": 41617 + }, + { + "epoch": 7.57, + "learning_rate": 7.301783764106297e-07, + "loss": 0.5694, + "step": 41618 + }, + { + "epoch": 7.57, + "learning_rate": 7.301237713869677e-07, + "loss": 0.3566, + "step": 41619 + }, + { + "epoch": 7.57, + "learning_rate": 7.300691663633054e-07, + "loss": 0.1202, + "step": 41620 + }, + { + "epoch": 7.57, + "learning_rate": 7.300145613396433e-07, + "loss": 1.3948, + "step": 41621 + }, + { + "epoch": 7.57, + "learning_rate": 7.299599563159811e-07, + "loss": 0.0242, + "step": 41622 + }, + { + "epoch": 7.57, + "learning_rate": 7.299053512923189e-07, + "loss": 0.0169, + "step": 41623 + }, + { + "epoch": 7.57, + "learning_rate": 7.298507462686567e-07, + "loss": 0.0074, + "step": 41624 + }, + { + "epoch": 7.57, + "learning_rate": 7.297961412449945e-07, + "loss": 0.0539, + "step": 41625 + }, + { + "epoch": 7.57, + "learning_rate": 7.297415362213324e-07, + "loss": 0.036, + "step": 41626 + }, + { + "epoch": 7.57, + "learning_rate": 7.296869311976702e-07, + "loss": 0.0453, + "step": 41627 + }, + { + "epoch": 7.57, + "learning_rate": 7.29632326174008e-07, + "loss": 1.375, + "step": 41628 + }, + { + "epoch": 7.57, + "learning_rate": 7.295777211503459e-07, + "loss": 0.0202, + "step": 41629 + }, + { + "epoch": 7.57, + "learning_rate": 7.295231161266837e-07, + "loss": 0.4247, + "step": 41630 + }, + { + "epoch": 7.57, + "learning_rate": 7.294685111030214e-07, + "loss": 0.0755, + "step": 41631 + }, + { + "epoch": 7.57, + "learning_rate": 7.294139060793594e-07, + "loss": 0.1676, + "step": 41632 + }, + { + "epoch": 7.57, + "learning_rate": 7.293593010556972e-07, + "loss": 0.1454, + "step": 41633 + }, + { + "epoch": 7.57, + "learning_rate": 7.293046960320349e-07, + "loss": 0.0812, + "step": 41634 + }, + { + "epoch": 7.57, + "learning_rate": 7.292500910083728e-07, + "loss": 0.3886, + "step": 41635 + }, + { + "epoch": 7.57, + "learning_rate": 7.291954859847106e-07, + "loss": 0.5789, + "step": 41636 + }, + { + "epoch": 7.57, + "learning_rate": 7.291408809610484e-07, + "loss": 0.1427, + "step": 41637 + }, + { + "epoch": 7.57, + "learning_rate": 7.290862759373862e-07, + "loss": 0.3407, + "step": 41638 + }, + { + "epoch": 7.57, + "learning_rate": 7.290316709137241e-07, + "loss": 0.1165, + "step": 41639 + }, + { + "epoch": 7.57, + "learning_rate": 7.28977065890062e-07, + "loss": 0.912, + "step": 41640 + }, + { + "epoch": 7.57, + "learning_rate": 7.289224608663997e-07, + "loss": 0.711, + "step": 41641 + }, + { + "epoch": 7.57, + "learning_rate": 7.288678558427375e-07, + "loss": 0.1567, + "step": 41642 + }, + { + "epoch": 7.57, + "learning_rate": 7.288132508190754e-07, + "loss": 0.4549, + "step": 41643 + }, + { + "epoch": 7.57, + "learning_rate": 7.287586457954132e-07, + "loss": 0.1572, + "step": 41644 + }, + { + "epoch": 7.57, + "learning_rate": 7.287040407717509e-07, + "loss": 1.3479, + "step": 41645 + }, + { + "epoch": 7.57, + "learning_rate": 7.286494357480889e-07, + "loss": 0.0379, + "step": 41646 + }, + { + "epoch": 7.57, + "learning_rate": 7.285948307244267e-07, + "loss": 0.0696, + "step": 41647 + }, + { + "epoch": 7.57, + "learning_rate": 7.285402257007644e-07, + "loss": 0.2293, + "step": 41648 + }, + { + "epoch": 7.57, + "learning_rate": 7.284856206771024e-07, + "loss": 0.4887, + "step": 41649 + }, + { + "epoch": 7.57, + "learning_rate": 7.284310156534401e-07, + "loss": 0.3051, + "step": 41650 + }, + { + "epoch": 7.57, + "learning_rate": 7.28376410629778e-07, + "loss": 0.2716, + "step": 41651 + }, + { + "epoch": 7.57, + "learning_rate": 7.283218056061158e-07, + "loss": 0.0118, + "step": 41652 + }, + { + "epoch": 7.57, + "learning_rate": 7.282672005824536e-07, + "loss": 0.0818, + "step": 41653 + }, + { + "epoch": 7.57, + "learning_rate": 7.282125955587915e-07, + "loss": 1.2958, + "step": 41654 + }, + { + "epoch": 7.58, + "learning_rate": 7.281579905351292e-07, + "loss": 0.9646, + "step": 41655 + }, + { + "epoch": 7.58, + "learning_rate": 7.281033855114671e-07, + "loss": 0.0939, + "step": 41656 + }, + { + "epoch": 7.58, + "learning_rate": 7.280487804878049e-07, + "loss": 0.0102, + "step": 41657 + }, + { + "epoch": 7.58, + "learning_rate": 7.279941754641427e-07, + "loss": 0.2361, + "step": 41658 + }, + { + "epoch": 7.58, + "learning_rate": 7.279395704404806e-07, + "loss": 0.8006, + "step": 41659 + }, + { + "epoch": 7.58, + "learning_rate": 7.278849654168184e-07, + "loss": 0.3266, + "step": 41660 + }, + { + "epoch": 7.58, + "learning_rate": 7.278303603931562e-07, + "loss": 0.5897, + "step": 41661 + }, + { + "epoch": 7.58, + "learning_rate": 7.27775755369494e-07, + "loss": 0.0944, + "step": 41662 + }, + { + "epoch": 7.58, + "learning_rate": 7.277211503458319e-07, + "loss": 0.2073, + "step": 41663 + }, + { + "epoch": 7.58, + "learning_rate": 7.276665453221696e-07, + "loss": 0.0108, + "step": 41664 + }, + { + "epoch": 7.58, + "learning_rate": 7.276119402985075e-07, + "loss": 0.0962, + "step": 41665 + }, + { + "epoch": 7.58, + "learning_rate": 7.275573352748453e-07, + "loss": 0.1889, + "step": 41666 + }, + { + "epoch": 7.58, + "learning_rate": 7.275027302511831e-07, + "loss": 0.6292, + "step": 41667 + }, + { + "epoch": 7.58, + "learning_rate": 7.274481252275209e-07, + "loss": 0.681, + "step": 41668 + }, + { + "epoch": 7.58, + "learning_rate": 7.273935202038588e-07, + "loss": 0.2118, + "step": 41669 + }, + { + "epoch": 7.58, + "learning_rate": 7.273389151801967e-07, + "loss": 1.4108, + "step": 41670 + }, + { + "epoch": 7.58, + "learning_rate": 7.272843101565344e-07, + "loss": 0.0643, + "step": 41671 + }, + { + "epoch": 7.58, + "learning_rate": 7.272297051328722e-07, + "loss": 0.0036, + "step": 41672 + }, + { + "epoch": 7.58, + "learning_rate": 7.271751001092101e-07, + "loss": 0.0998, + "step": 41673 + }, + { + "epoch": 7.58, + "learning_rate": 7.271204950855479e-07, + "loss": 0.1661, + "step": 41674 + }, + { + "epoch": 7.58, + "learning_rate": 7.270658900618856e-07, + "loss": 0.345, + "step": 41675 + }, + { + "epoch": 7.58, + "learning_rate": 7.270112850382236e-07, + "loss": 0.9851, + "step": 41676 + }, + { + "epoch": 7.58, + "learning_rate": 7.269566800145614e-07, + "loss": 0.7594, + "step": 41677 + }, + { + "epoch": 7.58, + "learning_rate": 7.269020749908991e-07, + "loss": 0.1825, + "step": 41678 + }, + { + "epoch": 7.58, + "learning_rate": 7.26847469967237e-07, + "loss": 2.8813, + "step": 41679 + }, + { + "epoch": 7.58, + "learning_rate": 7.267928649435748e-07, + "loss": 0.096, + "step": 41680 + }, + { + "epoch": 7.58, + "learning_rate": 7.267382599199127e-07, + "loss": 0.0079, + "step": 41681 + }, + { + "epoch": 7.58, + "learning_rate": 7.266836548962504e-07, + "loss": 0.1714, + "step": 41682 + }, + { + "epoch": 7.58, + "learning_rate": 7.266290498725883e-07, + "loss": 0.2445, + "step": 41683 + }, + { + "epoch": 7.58, + "learning_rate": 7.265744448489262e-07, + "loss": 0.0126, + "step": 41684 + }, + { + "epoch": 7.58, + "learning_rate": 7.265198398252639e-07, + "loss": 0.5168, + "step": 41685 + }, + { + "epoch": 7.58, + "learning_rate": 7.264652348016017e-07, + "loss": 0.2532, + "step": 41686 + }, + { + "epoch": 7.58, + "learning_rate": 7.264106297779396e-07, + "loss": 1.3425, + "step": 41687 + }, + { + "epoch": 7.58, + "learning_rate": 7.263560247542774e-07, + "loss": 0.7639, + "step": 41688 + }, + { + "epoch": 7.58, + "learning_rate": 7.263014197306153e-07, + "loss": 0.2397, + "step": 41689 + }, + { + "epoch": 7.58, + "learning_rate": 7.262468147069531e-07, + "loss": 0.0051, + "step": 41690 + }, + { + "epoch": 7.58, + "learning_rate": 7.261922096832909e-07, + "loss": 1.3257, + "step": 41691 + }, + { + "epoch": 7.58, + "learning_rate": 7.261376046596287e-07, + "loss": 0.0688, + "step": 41692 + }, + { + "epoch": 7.58, + "learning_rate": 7.260829996359666e-07, + "loss": 2.1215, + "step": 41693 + }, + { + "epoch": 7.58, + "learning_rate": 7.260283946123043e-07, + "loss": 0.4011, + "step": 41694 + }, + { + "epoch": 7.58, + "learning_rate": 7.259737895886422e-07, + "loss": 2.6926, + "step": 41695 + }, + { + "epoch": 7.58, + "learning_rate": 7.2591918456498e-07, + "loss": 0.6302, + "step": 41696 + }, + { + "epoch": 7.58, + "learning_rate": 7.258645795413178e-07, + "loss": 0.0845, + "step": 41697 + }, + { + "epoch": 7.58, + "learning_rate": 7.258099745176557e-07, + "loss": 1.5529, + "step": 41698 + }, + { + "epoch": 7.58, + "learning_rate": 7.257553694939934e-07, + "loss": 1.9756, + "step": 41699 + }, + { + "epoch": 7.58, + "learning_rate": 7.257007644703314e-07, + "loss": 0.0146, + "step": 41700 + }, + { + "epoch": 7.58, + "learning_rate": 7.256461594466691e-07, + "loss": 0.2007, + "step": 41701 + }, + { + "epoch": 7.58, + "learning_rate": 7.255915544230069e-07, + "loss": 0.2178, + "step": 41702 + }, + { + "epoch": 7.58, + "learning_rate": 7.255369493993448e-07, + "loss": 0.1455, + "step": 41703 + }, + { + "epoch": 7.58, + "learning_rate": 7.254823443756826e-07, + "loss": 0.3058, + "step": 41704 + }, + { + "epoch": 7.58, + "learning_rate": 7.254277393520203e-07, + "loss": 0.2344, + "step": 41705 + }, + { + "epoch": 7.58, + "learning_rate": 7.253731343283582e-07, + "loss": 0.0114, + "step": 41706 + }, + { + "epoch": 7.58, + "learning_rate": 7.253185293046961e-07, + "loss": 0.5747, + "step": 41707 + }, + { + "epoch": 7.58, + "learning_rate": 7.252639242810338e-07, + "loss": 0.363, + "step": 41708 + }, + { + "epoch": 7.58, + "learning_rate": 7.252093192573717e-07, + "loss": 0.3668, + "step": 41709 + }, + { + "epoch": 7.59, + "learning_rate": 7.251547142337095e-07, + "loss": 0.5282, + "step": 41710 + }, + { + "epoch": 7.59, + "learning_rate": 7.251001092100474e-07, + "loss": 0.0493, + "step": 41711 + }, + { + "epoch": 7.59, + "learning_rate": 7.250455041863851e-07, + "loss": 0.0116, + "step": 41712 + }, + { + "epoch": 7.59, + "learning_rate": 7.24990899162723e-07, + "loss": 0.0051, + "step": 41713 + }, + { + "epoch": 7.59, + "learning_rate": 7.249362941390609e-07, + "loss": 0.2035, + "step": 41714 + }, + { + "epoch": 7.59, + "learning_rate": 7.248816891153986e-07, + "loss": 0.4573, + "step": 41715 + }, + { + "epoch": 7.59, + "learning_rate": 7.248270840917364e-07, + "loss": 0.1009, + "step": 41716 + }, + { + "epoch": 7.59, + "learning_rate": 7.247724790680743e-07, + "loss": 0.1954, + "step": 41717 + }, + { + "epoch": 7.59, + "learning_rate": 7.247178740444121e-07, + "loss": 0.8802, + "step": 41718 + }, + { + "epoch": 7.59, + "learning_rate": 7.246632690207499e-07, + "loss": 0.088, + "step": 41719 + }, + { + "epoch": 7.59, + "learning_rate": 7.246086639970878e-07, + "loss": 1.7446, + "step": 41720 + }, + { + "epoch": 7.59, + "learning_rate": 7.245540589734256e-07, + "loss": 0.0083, + "step": 41721 + }, + { + "epoch": 7.59, + "learning_rate": 7.244994539497634e-07, + "loss": 2.5658, + "step": 41722 + }, + { + "epoch": 7.59, + "learning_rate": 7.244448489261012e-07, + "loss": 0.26, + "step": 41723 + }, + { + "epoch": 7.59, + "learning_rate": 7.24390243902439e-07, + "loss": 0.9983, + "step": 41724 + }, + { + "epoch": 7.59, + "learning_rate": 7.243356388787769e-07, + "loss": 0.0969, + "step": 41725 + }, + { + "epoch": 7.59, + "learning_rate": 7.242810338551146e-07, + "loss": 1.3128, + "step": 41726 + }, + { + "epoch": 7.59, + "learning_rate": 7.242264288314526e-07, + "loss": 0.1655, + "step": 41727 + }, + { + "epoch": 7.59, + "learning_rate": 7.241718238077904e-07, + "loss": 0.4778, + "step": 41728 + }, + { + "epoch": 7.59, + "learning_rate": 7.241172187841281e-07, + "loss": 0.0436, + "step": 41729 + }, + { + "epoch": 7.59, + "learning_rate": 7.24062613760466e-07, + "loss": 0.4352, + "step": 41730 + }, + { + "epoch": 7.59, + "learning_rate": 7.240080087368038e-07, + "loss": 0.1238, + "step": 41731 + }, + { + "epoch": 7.59, + "learning_rate": 7.239534037131416e-07, + "loss": 0.0044, + "step": 41732 + }, + { + "epoch": 7.59, + "learning_rate": 7.238987986894795e-07, + "loss": 0.2795, + "step": 41733 + }, + { + "epoch": 7.59, + "learning_rate": 7.238441936658173e-07, + "loss": 0.0078, + "step": 41734 + }, + { + "epoch": 7.59, + "learning_rate": 7.237895886421551e-07, + "loss": 0.2073, + "step": 41735 + }, + { + "epoch": 7.59, + "learning_rate": 7.237349836184929e-07, + "loss": 0.1466, + "step": 41736 + }, + { + "epoch": 7.59, + "learning_rate": 7.236803785948308e-07, + "loss": 0.0079, + "step": 41737 + }, + { + "epoch": 7.59, + "learning_rate": 7.236257735711686e-07, + "loss": 0.0057, + "step": 41738 + }, + { + "epoch": 7.59, + "learning_rate": 7.235711685475064e-07, + "loss": 0.1328, + "step": 41739 + }, + { + "epoch": 7.59, + "learning_rate": 7.235165635238442e-07, + "loss": 0.082, + "step": 41740 + }, + { + "epoch": 7.59, + "learning_rate": 7.234619585001821e-07, + "loss": 0.4169, + "step": 41741 + }, + { + "epoch": 7.59, + "learning_rate": 7.234073534765198e-07, + "loss": 0.0456, + "step": 41742 + }, + { + "epoch": 7.59, + "learning_rate": 7.233527484528576e-07, + "loss": 0.0148, + "step": 41743 + }, + { + "epoch": 7.59, + "learning_rate": 7.232981434291956e-07, + "loss": 2.3717, + "step": 41744 + }, + { + "epoch": 7.59, + "learning_rate": 7.232435384055333e-07, + "loss": 2.6181, + "step": 41745 + }, + { + "epoch": 7.59, + "learning_rate": 7.231889333818711e-07, + "loss": 0.9152, + "step": 41746 + }, + { + "epoch": 7.59, + "learning_rate": 7.23134328358209e-07, + "loss": 0.3293, + "step": 41747 + }, + { + "epoch": 7.59, + "learning_rate": 7.230797233345468e-07, + "loss": 3.1219, + "step": 41748 + }, + { + "epoch": 7.59, + "learning_rate": 7.230251183108846e-07, + "loss": 0.0099, + "step": 41749 + }, + { + "epoch": 7.59, + "learning_rate": 7.229705132872224e-07, + "loss": 0.7461, + "step": 41750 + }, + { + "epoch": 7.59, + "learning_rate": 7.229159082635603e-07, + "loss": 0.0442, + "step": 41751 + }, + { + "epoch": 7.59, + "learning_rate": 7.228613032398981e-07, + "loss": 1.7445, + "step": 41752 + }, + { + "epoch": 7.59, + "learning_rate": 7.228066982162359e-07, + "loss": 0.3334, + "step": 41753 + }, + { + "epoch": 7.59, + "learning_rate": 7.227520931925737e-07, + "loss": 0.394, + "step": 41754 + }, + { + "epoch": 7.59, + "learning_rate": 7.226974881689116e-07, + "loss": 0.0075, + "step": 41755 + }, + { + "epoch": 7.59, + "learning_rate": 7.226428831452493e-07, + "loss": 0.7094, + "step": 41756 + }, + { + "epoch": 7.59, + "learning_rate": 7.225882781215873e-07, + "loss": 0.7681, + "step": 41757 + }, + { + "epoch": 7.59, + "learning_rate": 7.225336730979251e-07, + "loss": 0.0085, + "step": 41758 + }, + { + "epoch": 7.59, + "learning_rate": 7.224790680742628e-07, + "loss": 0.0233, + "step": 41759 + }, + { + "epoch": 7.59, + "learning_rate": 7.224244630506007e-07, + "loss": 2.205, + "step": 41760 + }, + { + "epoch": 7.59, + "learning_rate": 7.223698580269385e-07, + "loss": 0.1197, + "step": 41761 + }, + { + "epoch": 7.59, + "learning_rate": 7.223152530032763e-07, + "loss": 0.3585, + "step": 41762 + }, + { + "epoch": 7.59, + "learning_rate": 7.222606479796141e-07, + "loss": 3.8333, + "step": 41763 + }, + { + "epoch": 7.59, + "learning_rate": 7.22206042955952e-07, + "loss": 0.0262, + "step": 41764 + }, + { + "epoch": 7.6, + "learning_rate": 7.221514379322898e-07, + "loss": 0.1446, + "step": 41765 + }, + { + "epoch": 7.6, + "learning_rate": 7.220968329086276e-07, + "loss": 1.4099, + "step": 41766 + }, + { + "epoch": 7.6, + "learning_rate": 7.220422278849654e-07, + "loss": 0.0259, + "step": 41767 + }, + { + "epoch": 7.6, + "learning_rate": 7.219876228613033e-07, + "loss": 0.4359, + "step": 41768 + }, + { + "epoch": 7.6, + "learning_rate": 7.219330178376411e-07, + "loss": 0.0164, + "step": 41769 + }, + { + "epoch": 7.6, + "learning_rate": 7.218784128139788e-07, + "loss": 1.0108, + "step": 41770 + }, + { + "epoch": 7.6, + "learning_rate": 7.218238077903168e-07, + "loss": 1.3668, + "step": 41771 + }, + { + "epoch": 7.6, + "learning_rate": 7.217692027666546e-07, + "loss": 1.1214, + "step": 41772 + }, + { + "epoch": 7.6, + "learning_rate": 7.217145977429923e-07, + "loss": 0.6349, + "step": 41773 + }, + { + "epoch": 7.6, + "learning_rate": 7.216599927193302e-07, + "loss": 0.0112, + "step": 41774 + }, + { + "epoch": 7.6, + "learning_rate": 7.21605387695668e-07, + "loss": 0.9025, + "step": 41775 + }, + { + "epoch": 7.6, + "learning_rate": 7.215507826720058e-07, + "loss": 0.6401, + "step": 41776 + }, + { + "epoch": 7.6, + "learning_rate": 7.214961776483437e-07, + "loss": 0.0752, + "step": 41777 + }, + { + "epoch": 7.6, + "learning_rate": 7.214415726246815e-07, + "loss": 0.0362, + "step": 41778 + }, + { + "epoch": 7.6, + "learning_rate": 7.213869676010193e-07, + "loss": 1.2365, + "step": 41779 + }, + { + "epoch": 7.6, + "learning_rate": 7.213323625773571e-07, + "loss": 0.0047, + "step": 41780 + }, + { + "epoch": 7.6, + "learning_rate": 7.21277757553695e-07, + "loss": 0.0447, + "step": 41781 + }, + { + "epoch": 7.6, + "learning_rate": 7.212231525300328e-07, + "loss": 0.0991, + "step": 41782 + }, + { + "epoch": 7.6, + "learning_rate": 7.211685475063706e-07, + "loss": 0.4119, + "step": 41783 + }, + { + "epoch": 7.6, + "learning_rate": 7.211139424827084e-07, + "loss": 0.9855, + "step": 41784 + }, + { + "epoch": 7.6, + "learning_rate": 7.210593374590463e-07, + "loss": 0.0075, + "step": 41785 + }, + { + "epoch": 7.6, + "learning_rate": 7.21004732435384e-07, + "loss": 0.0458, + "step": 41786 + }, + { + "epoch": 7.6, + "learning_rate": 7.209501274117218e-07, + "loss": 1.3633, + "step": 41787 + }, + { + "epoch": 7.6, + "learning_rate": 7.208955223880598e-07, + "loss": 0.2599, + "step": 41788 + }, + { + "epoch": 7.6, + "learning_rate": 7.208409173643975e-07, + "loss": 1.0509, + "step": 41789 + }, + { + "epoch": 7.6, + "learning_rate": 7.207863123407354e-07, + "loss": 0.0297, + "step": 41790 + }, + { + "epoch": 7.6, + "learning_rate": 7.207317073170732e-07, + "loss": 1.5823, + "step": 41791 + }, + { + "epoch": 7.6, + "learning_rate": 7.20677102293411e-07, + "loss": 0.1201, + "step": 41792 + }, + { + "epoch": 7.6, + "learning_rate": 7.206224972697488e-07, + "loss": 0.0514, + "step": 41793 + }, + { + "epoch": 7.6, + "learning_rate": 7.205678922460866e-07, + "loss": 2.323, + "step": 41794 + }, + { + "epoch": 7.6, + "learning_rate": 7.205132872224245e-07, + "loss": 1.8434, + "step": 41795 + }, + { + "epoch": 7.6, + "learning_rate": 7.204586821987623e-07, + "loss": 0.0716, + "step": 41796 + }, + { + "epoch": 7.6, + "learning_rate": 7.204040771751001e-07, + "loss": 1.513, + "step": 41797 + }, + { + "epoch": 7.6, + "learning_rate": 7.20349472151438e-07, + "loss": 1.4576, + "step": 41798 + }, + { + "epoch": 7.6, + "learning_rate": 7.202948671277758e-07, + "loss": 0.0086, + "step": 41799 + }, + { + "epoch": 7.6, + "learning_rate": 7.202402621041135e-07, + "loss": 1.1459, + "step": 41800 + }, + { + "epoch": 7.6, + "learning_rate": 7.201856570804515e-07, + "loss": 0.3542, + "step": 41801 + }, + { + "epoch": 7.6, + "learning_rate": 7.201310520567893e-07, + "loss": 1.5715, + "step": 41802 + }, + { + "epoch": 7.6, + "learning_rate": 7.20076447033127e-07, + "loss": 1.8496, + "step": 41803 + }, + { + "epoch": 7.6, + "learning_rate": 7.200218420094649e-07, + "loss": 0.0105, + "step": 41804 + }, + { + "epoch": 7.6, + "learning_rate": 7.199672369858027e-07, + "loss": 0.0927, + "step": 41805 + }, + { + "epoch": 7.6, + "learning_rate": 7.199126319621405e-07, + "loss": 1.207, + "step": 41806 + }, + { + "epoch": 7.6, + "learning_rate": 7.198580269384783e-07, + "loss": 0.4445, + "step": 41807 + }, + { + "epoch": 7.6, + "learning_rate": 7.198034219148162e-07, + "loss": 0.066, + "step": 41808 + }, + { + "epoch": 7.6, + "learning_rate": 7.197488168911541e-07, + "loss": 1.1695, + "step": 41809 + }, + { + "epoch": 7.6, + "learning_rate": 7.196942118674918e-07, + "loss": 0.0074, + "step": 41810 + }, + { + "epoch": 7.6, + "learning_rate": 7.196396068438296e-07, + "loss": 0.0727, + "step": 41811 + }, + { + "epoch": 7.6, + "learning_rate": 7.195850018201675e-07, + "loss": 0.0044, + "step": 41812 + }, + { + "epoch": 7.6, + "learning_rate": 7.195303967965053e-07, + "loss": 3.0916, + "step": 41813 + }, + { + "epoch": 7.6, + "learning_rate": 7.19475791772843e-07, + "loss": 0.0196, + "step": 41814 + }, + { + "epoch": 7.6, + "learning_rate": 7.19421186749181e-07, + "loss": 0.0504, + "step": 41815 + }, + { + "epoch": 7.6, + "learning_rate": 7.193665817255188e-07, + "loss": 1.3659, + "step": 41816 + }, + { + "epoch": 7.6, + "learning_rate": 7.193119767018566e-07, + "loss": 0.771, + "step": 41817 + }, + { + "epoch": 7.6, + "learning_rate": 7.192573716781945e-07, + "loss": 0.1083, + "step": 41818 + }, + { + "epoch": 7.6, + "learning_rate": 7.192027666545322e-07, + "loss": 0.4261, + "step": 41819 + }, + { + "epoch": 7.61, + "learning_rate": 7.191481616308701e-07, + "loss": 0.1401, + "step": 41820 + }, + { + "epoch": 7.61, + "learning_rate": 7.190935566072079e-07, + "loss": 0.2614, + "step": 41821 + }, + { + "epoch": 7.61, + "learning_rate": 7.190389515835457e-07, + "loss": 0.3267, + "step": 41822 + }, + { + "epoch": 7.61, + "learning_rate": 7.189843465598835e-07, + "loss": 0.13, + "step": 41823 + }, + { + "epoch": 7.61, + "learning_rate": 7.189297415362213e-07, + "loss": 0.262, + "step": 41824 + }, + { + "epoch": 7.61, + "learning_rate": 7.188751365125593e-07, + "loss": 0.1741, + "step": 41825 + }, + { + "epoch": 7.61, + "learning_rate": 7.18820531488897e-07, + "loss": 0.0045, + "step": 41826 + }, + { + "epoch": 7.61, + "learning_rate": 7.187659264652348e-07, + "loss": 1.3352, + "step": 41827 + }, + { + "epoch": 7.61, + "learning_rate": 7.187113214415727e-07, + "loss": 0.0154, + "step": 41828 + }, + { + "epoch": 7.61, + "learning_rate": 7.186567164179105e-07, + "loss": 1.0101, + "step": 41829 + }, + { + "epoch": 7.61, + "learning_rate": 7.186021113942482e-07, + "loss": 0.873, + "step": 41830 + }, + { + "epoch": 7.61, + "learning_rate": 7.185475063705861e-07, + "loss": 1.2095, + "step": 41831 + }, + { + "epoch": 7.61, + "learning_rate": 7.18492901346924e-07, + "loss": 0.6771, + "step": 41832 + }, + { + "epoch": 7.61, + "learning_rate": 7.184382963232617e-07, + "loss": 0.0141, + "step": 41833 + }, + { + "epoch": 7.61, + "learning_rate": 7.183836912995996e-07, + "loss": 2.1223, + "step": 41834 + }, + { + "epoch": 7.61, + "learning_rate": 7.183290862759374e-07, + "loss": 0.1958, + "step": 41835 + }, + { + "epoch": 7.61, + "learning_rate": 7.182744812522753e-07, + "loss": 0.0079, + "step": 41836 + }, + { + "epoch": 7.61, + "learning_rate": 7.18219876228613e-07, + "loss": 0.5593, + "step": 41837 + }, + { + "epoch": 7.61, + "learning_rate": 7.181652712049509e-07, + "loss": 0.0993, + "step": 41838 + }, + { + "epoch": 7.61, + "learning_rate": 7.181106661812888e-07, + "loss": 0.4187, + "step": 41839 + }, + { + "epoch": 7.61, + "learning_rate": 7.180560611576265e-07, + "loss": 0.0145, + "step": 41840 + }, + { + "epoch": 7.61, + "learning_rate": 7.180014561339643e-07, + "loss": 0.2756, + "step": 41841 + }, + { + "epoch": 7.61, + "learning_rate": 7.179468511103022e-07, + "loss": 0.4723, + "step": 41842 + }, + { + "epoch": 7.61, + "learning_rate": 7.1789224608664e-07, + "loss": 1.0302, + "step": 41843 + }, + { + "epoch": 7.61, + "learning_rate": 7.178376410629777e-07, + "loss": 0.1527, + "step": 41844 + }, + { + "epoch": 7.61, + "learning_rate": 7.177830360393157e-07, + "loss": 0.0122, + "step": 41845 + }, + { + "epoch": 7.61, + "learning_rate": 7.177284310156535e-07, + "loss": 0.7555, + "step": 41846 + }, + { + "epoch": 7.61, + "learning_rate": 7.176738259919913e-07, + "loss": 1.3007, + "step": 41847 + }, + { + "epoch": 7.61, + "learning_rate": 7.176192209683291e-07, + "loss": 0.004, + "step": 41848 + }, + { + "epoch": 7.61, + "learning_rate": 7.175646159446669e-07, + "loss": 0.0372, + "step": 41849 + }, + { + "epoch": 7.61, + "learning_rate": 7.175100109210048e-07, + "loss": 1.1322, + "step": 41850 + }, + { + "epoch": 7.61, + "learning_rate": 7.174554058973425e-07, + "loss": 0.0833, + "step": 41851 + }, + { + "epoch": 7.61, + "learning_rate": 7.174008008736804e-07, + "loss": 0.1247, + "step": 41852 + }, + { + "epoch": 7.61, + "learning_rate": 7.173461958500183e-07, + "loss": 2.4699, + "step": 41853 + }, + { + "epoch": 7.61, + "learning_rate": 7.17291590826356e-07, + "loss": 0.3368, + "step": 41854 + }, + { + "epoch": 7.61, + "learning_rate": 7.172369858026938e-07, + "loss": 0.0203, + "step": 41855 + }, + { + "epoch": 7.61, + "learning_rate": 7.171823807790317e-07, + "loss": 0.3539, + "step": 41856 + }, + { + "epoch": 7.61, + "learning_rate": 7.171277757553695e-07, + "loss": 0.0196, + "step": 41857 + }, + { + "epoch": 7.61, + "learning_rate": 7.170731707317073e-07, + "loss": 0.3004, + "step": 41858 + }, + { + "epoch": 7.61, + "learning_rate": 7.170185657080452e-07, + "loss": 0.2405, + "step": 41859 + }, + { + "epoch": 7.61, + "learning_rate": 7.169639606843829e-07, + "loss": 0.0149, + "step": 41860 + }, + { + "epoch": 7.61, + "learning_rate": 7.169093556607208e-07, + "loss": 0.0245, + "step": 41861 + }, + { + "epoch": 7.61, + "learning_rate": 7.168547506370587e-07, + "loss": 0.2127, + "step": 41862 + }, + { + "epoch": 7.61, + "learning_rate": 7.168001456133964e-07, + "loss": 1.2726, + "step": 41863 + }, + { + "epoch": 7.61, + "learning_rate": 7.167455405897343e-07, + "loss": 0.0071, + "step": 41864 + }, + { + "epoch": 7.61, + "learning_rate": 7.166909355660721e-07, + "loss": 2.655, + "step": 41865 + }, + { + "epoch": 7.61, + "learning_rate": 7.1663633054241e-07, + "loss": 0.1404, + "step": 41866 + }, + { + "epoch": 7.61, + "learning_rate": 7.165817255187477e-07, + "loss": 0.0046, + "step": 41867 + }, + { + "epoch": 7.61, + "learning_rate": 7.165271204950855e-07, + "loss": 1.0284, + "step": 41868 + }, + { + "epoch": 7.61, + "learning_rate": 7.164725154714235e-07, + "loss": 2.5303, + "step": 41869 + }, + { + "epoch": 7.61, + "learning_rate": 7.164179104477612e-07, + "loss": 0.3003, + "step": 41870 + }, + { + "epoch": 7.61, + "learning_rate": 7.16363305424099e-07, + "loss": 0.2111, + "step": 41871 + }, + { + "epoch": 7.61, + "learning_rate": 7.163087004004369e-07, + "loss": 0.0124, + "step": 41872 + }, + { + "epoch": 7.61, + "learning_rate": 7.162540953767747e-07, + "loss": 0.1388, + "step": 41873 + }, + { + "epoch": 7.61, + "learning_rate": 7.161994903531124e-07, + "loss": 1.7266, + "step": 41874 + }, + { + "epoch": 7.62, + "learning_rate": 7.161448853294503e-07, + "loss": 1.9535, + "step": 41875 + }, + { + "epoch": 7.62, + "learning_rate": 7.160902803057882e-07, + "loss": 3.0669, + "step": 41876 + }, + { + "epoch": 7.62, + "learning_rate": 7.16035675282126e-07, + "loss": 0.1349, + "step": 41877 + }, + { + "epoch": 7.62, + "learning_rate": 7.159810702584638e-07, + "loss": 0.8506, + "step": 41878 + }, + { + "epoch": 7.62, + "learning_rate": 7.159264652348016e-07, + "loss": 2.2262, + "step": 41879 + }, + { + "epoch": 7.62, + "learning_rate": 7.158718602111395e-07, + "loss": 0.0105, + "step": 41880 + }, + { + "epoch": 7.62, + "learning_rate": 7.158172551874772e-07, + "loss": 0.3234, + "step": 41881 + }, + { + "epoch": 7.62, + "learning_rate": 7.157626501638151e-07, + "loss": 0.1084, + "step": 41882 + }, + { + "epoch": 7.62, + "learning_rate": 7.15708045140153e-07, + "loss": 0.0269, + "step": 41883 + }, + { + "epoch": 7.62, + "learning_rate": 7.156534401164907e-07, + "loss": 1.114, + "step": 41884 + }, + { + "epoch": 7.62, + "learning_rate": 7.155988350928285e-07, + "loss": 0.9888, + "step": 41885 + }, + { + "epoch": 7.62, + "learning_rate": 7.155442300691664e-07, + "loss": 0.0042, + "step": 41886 + }, + { + "epoch": 7.62, + "learning_rate": 7.154896250455042e-07, + "loss": 0.0095, + "step": 41887 + }, + { + "epoch": 7.62, + "learning_rate": 7.15435020021842e-07, + "loss": 0.1985, + "step": 41888 + }, + { + "epoch": 7.62, + "learning_rate": 7.153804149981799e-07, + "loss": 1.0159, + "step": 41889 + }, + { + "epoch": 7.62, + "learning_rate": 7.153258099745177e-07, + "loss": 0.0067, + "step": 41890 + }, + { + "epoch": 7.62, + "learning_rate": 7.152712049508555e-07, + "loss": 2.0268, + "step": 41891 + }, + { + "epoch": 7.62, + "learning_rate": 7.152165999271933e-07, + "loss": 0.089, + "step": 41892 + }, + { + "epoch": 7.62, + "learning_rate": 7.151619949035311e-07, + "loss": 0.0419, + "step": 41893 + }, + { + "epoch": 7.62, + "learning_rate": 7.15107389879869e-07, + "loss": 0.0813, + "step": 41894 + }, + { + "epoch": 7.62, + "learning_rate": 7.150527848562067e-07, + "loss": 0.0306, + "step": 41895 + }, + { + "epoch": 7.62, + "learning_rate": 7.149981798325447e-07, + "loss": 0.1017, + "step": 41896 + }, + { + "epoch": 7.62, + "learning_rate": 7.149435748088824e-07, + "loss": 0.1741, + "step": 41897 + }, + { + "epoch": 7.62, + "learning_rate": 7.148889697852202e-07, + "loss": 0.189, + "step": 41898 + }, + { + "epoch": 7.62, + "learning_rate": 7.148343647615581e-07, + "loss": 0.1994, + "step": 41899 + }, + { + "epoch": 7.62, + "learning_rate": 7.147797597378959e-07, + "loss": 0.0779, + "step": 41900 + }, + { + "epoch": 7.62, + "learning_rate": 7.147251547142337e-07, + "loss": 0.079, + "step": 41901 + }, + { + "epoch": 7.62, + "learning_rate": 7.146705496905716e-07, + "loss": 0.0065, + "step": 41902 + }, + { + "epoch": 7.62, + "learning_rate": 7.146159446669094e-07, + "loss": 0.088, + "step": 41903 + }, + { + "epoch": 7.62, + "learning_rate": 7.145613396432471e-07, + "loss": 0.076, + "step": 41904 + }, + { + "epoch": 7.62, + "learning_rate": 7.14506734619585e-07, + "loss": 0.0052, + "step": 41905 + }, + { + "epoch": 7.62, + "learning_rate": 7.144521295959229e-07, + "loss": 0.3558, + "step": 41906 + }, + { + "epoch": 7.62, + "learning_rate": 7.143975245722607e-07, + "loss": 0.4947, + "step": 41907 + }, + { + "epoch": 7.62, + "learning_rate": 7.143429195485985e-07, + "loss": 0.3698, + "step": 41908 + }, + { + "epoch": 7.62, + "learning_rate": 7.142883145249363e-07, + "loss": 0.014, + "step": 41909 + }, + { + "epoch": 7.62, + "learning_rate": 7.142337095012742e-07, + "loss": 0.0096, + "step": 41910 + }, + { + "epoch": 7.62, + "learning_rate": 7.141791044776119e-07, + "loss": 0.1459, + "step": 41911 + }, + { + "epoch": 7.62, + "learning_rate": 7.141244994539497e-07, + "loss": 0.2043, + "step": 41912 + }, + { + "epoch": 7.62, + "learning_rate": 7.140698944302877e-07, + "loss": 0.7909, + "step": 41913 + }, + { + "epoch": 7.62, + "learning_rate": 7.140152894066254e-07, + "loss": 0.6188, + "step": 41914 + }, + { + "epoch": 7.62, + "learning_rate": 7.139606843829633e-07, + "loss": 0.0196, + "step": 41915 + }, + { + "epoch": 7.62, + "learning_rate": 7.139060793593011e-07, + "loss": 0.1334, + "step": 41916 + }, + { + "epoch": 7.62, + "learning_rate": 7.138514743356389e-07, + "loss": 0.579, + "step": 41917 + }, + { + "epoch": 7.62, + "learning_rate": 7.137968693119767e-07, + "loss": 0.5211, + "step": 41918 + }, + { + "epoch": 7.62, + "learning_rate": 7.137422642883145e-07, + "loss": 0.0358, + "step": 41919 + }, + { + "epoch": 7.62, + "learning_rate": 7.136876592646524e-07, + "loss": 0.2286, + "step": 41920 + }, + { + "epoch": 7.62, + "learning_rate": 7.136330542409902e-07, + "loss": 0.6239, + "step": 41921 + }, + { + "epoch": 7.62, + "learning_rate": 7.13578449217328e-07, + "loss": 0.6766, + "step": 41922 + }, + { + "epoch": 7.62, + "learning_rate": 7.135238441936658e-07, + "loss": 0.0474, + "step": 41923 + }, + { + "epoch": 7.62, + "learning_rate": 7.134692391700037e-07, + "loss": 0.9244, + "step": 41924 + }, + { + "epoch": 7.62, + "learning_rate": 7.134146341463414e-07, + "loss": 0.5207, + "step": 41925 + }, + { + "epoch": 7.62, + "learning_rate": 7.133600291226794e-07, + "loss": 0.4551, + "step": 41926 + }, + { + "epoch": 7.62, + "learning_rate": 7.133054240990172e-07, + "loss": 0.2831, + "step": 41927 + }, + { + "epoch": 7.62, + "learning_rate": 7.132508190753549e-07, + "loss": 0.7138, + "step": 41928 + }, + { + "epoch": 7.62, + "learning_rate": 7.131962140516928e-07, + "loss": 0.4049, + "step": 41929 + }, + { + "epoch": 7.63, + "learning_rate": 7.131416090280306e-07, + "loss": 0.2127, + "step": 41930 + }, + { + "epoch": 7.63, + "learning_rate": 7.130870040043684e-07, + "loss": 0.4189, + "step": 41931 + }, + { + "epoch": 7.63, + "learning_rate": 7.130323989807062e-07, + "loss": 1.0664, + "step": 41932 + }, + { + "epoch": 7.63, + "learning_rate": 7.129777939570441e-07, + "loss": 3.3709, + "step": 41933 + }, + { + "epoch": 7.63, + "learning_rate": 7.129231889333818e-07, + "loss": 0.3783, + "step": 41934 + }, + { + "epoch": 7.63, + "learning_rate": 7.128685839097197e-07, + "loss": 1.6059, + "step": 41935 + }, + { + "epoch": 7.63, + "learning_rate": 7.128139788860575e-07, + "loss": 0.9267, + "step": 41936 + }, + { + "epoch": 7.63, + "learning_rate": 7.127593738623954e-07, + "loss": 0.9448, + "step": 41937 + }, + { + "epoch": 7.63, + "learning_rate": 7.127047688387332e-07, + "loss": 1.0542, + "step": 41938 + }, + { + "epoch": 7.63, + "learning_rate": 7.126501638150709e-07, + "loss": 0.4298, + "step": 41939 + }, + { + "epoch": 7.63, + "learning_rate": 7.125955587914089e-07, + "loss": 2.5059, + "step": 41940 + }, + { + "epoch": 7.63, + "learning_rate": 7.125409537677466e-07, + "loss": 0.0163, + "step": 41941 + }, + { + "epoch": 7.63, + "learning_rate": 7.124863487440844e-07, + "loss": 0.7187, + "step": 41942 + }, + { + "epoch": 7.63, + "learning_rate": 7.124317437204223e-07, + "loss": 1.1231, + "step": 41943 + }, + { + "epoch": 7.63, + "learning_rate": 7.123771386967601e-07, + "loss": 0.7327, + "step": 41944 + }, + { + "epoch": 7.63, + "learning_rate": 7.12322533673098e-07, + "loss": 0.3261, + "step": 41945 + }, + { + "epoch": 7.63, + "learning_rate": 7.122679286494358e-07, + "loss": 0.091, + "step": 41946 + }, + { + "epoch": 7.63, + "learning_rate": 7.122133236257736e-07, + "loss": 0.2554, + "step": 41947 + }, + { + "epoch": 7.63, + "learning_rate": 7.121587186021114e-07, + "loss": 1.4103, + "step": 41948 + }, + { + "epoch": 7.63, + "learning_rate": 7.121041135784492e-07, + "loss": 0.7878, + "step": 41949 + }, + { + "epoch": 7.63, + "learning_rate": 7.120495085547871e-07, + "loss": 0.0133, + "step": 41950 + }, + { + "epoch": 7.63, + "learning_rate": 7.119949035311249e-07, + "loss": 0.0084, + "step": 41951 + }, + { + "epoch": 7.63, + "learning_rate": 7.119402985074627e-07, + "loss": 0.6589, + "step": 41952 + }, + { + "epoch": 7.63, + "learning_rate": 7.118856934838005e-07, + "loss": 0.1253, + "step": 41953 + }, + { + "epoch": 7.63, + "learning_rate": 7.118310884601384e-07, + "loss": 0.3884, + "step": 41954 + }, + { + "epoch": 7.63, + "learning_rate": 7.117764834364761e-07, + "loss": 0.0296, + "step": 41955 + }, + { + "epoch": 7.63, + "learning_rate": 7.11721878412814e-07, + "loss": 0.069, + "step": 41956 + }, + { + "epoch": 7.63, + "learning_rate": 7.116672733891519e-07, + "loss": 0.9314, + "step": 41957 + }, + { + "epoch": 7.63, + "learning_rate": 7.116126683654896e-07, + "loss": 0.0104, + "step": 41958 + }, + { + "epoch": 7.63, + "learning_rate": 7.115580633418275e-07, + "loss": 0.1844, + "step": 41959 + }, + { + "epoch": 7.63, + "learning_rate": 7.115034583181653e-07, + "loss": 0.8194, + "step": 41960 + }, + { + "epoch": 7.63, + "learning_rate": 7.114488532945031e-07, + "loss": 0.1165, + "step": 41961 + }, + { + "epoch": 7.63, + "learning_rate": 7.113942482708409e-07, + "loss": 0.5243, + "step": 41962 + }, + { + "epoch": 7.63, + "learning_rate": 7.113396432471787e-07, + "loss": 0.0075, + "step": 41963 + }, + { + "epoch": 7.63, + "learning_rate": 7.112850382235167e-07, + "loss": 0.4332, + "step": 41964 + }, + { + "epoch": 7.63, + "learning_rate": 7.112304331998544e-07, + "loss": 0.1102, + "step": 41965 + }, + { + "epoch": 7.63, + "learning_rate": 7.111758281761922e-07, + "loss": 0.0085, + "step": 41966 + }, + { + "epoch": 7.63, + "learning_rate": 7.111212231525301e-07, + "loss": 0.5388, + "step": 41967 + }, + { + "epoch": 7.63, + "learning_rate": 7.110666181288679e-07, + "loss": 0.2335, + "step": 41968 + }, + { + "epoch": 7.63, + "learning_rate": 7.110120131052056e-07, + "loss": 0.0105, + "step": 41969 + }, + { + "epoch": 7.63, + "learning_rate": 7.109574080815436e-07, + "loss": 1.1455, + "step": 41970 + }, + { + "epoch": 7.63, + "learning_rate": 7.109028030578814e-07, + "loss": 1.3221, + "step": 41971 + }, + { + "epoch": 7.63, + "learning_rate": 7.108481980342191e-07, + "loss": 0.0232, + "step": 41972 + }, + { + "epoch": 7.63, + "learning_rate": 7.10793593010557e-07, + "loss": 0.0085, + "step": 41973 + }, + { + "epoch": 7.63, + "learning_rate": 7.107389879868948e-07, + "loss": 0.4853, + "step": 41974 + }, + { + "epoch": 7.63, + "learning_rate": 7.106843829632327e-07, + "loss": 3.5166, + "step": 41975 + }, + { + "epoch": 7.63, + "learning_rate": 7.106297779395704e-07, + "loss": 1.6688, + "step": 41976 + }, + { + "epoch": 7.63, + "learning_rate": 7.105751729159083e-07, + "loss": 0.0129, + "step": 41977 + }, + { + "epoch": 7.63, + "learning_rate": 7.105205678922461e-07, + "loss": 2.0395, + "step": 41978 + }, + { + "epoch": 7.63, + "learning_rate": 7.104659628685839e-07, + "loss": 0.1174, + "step": 41979 + }, + { + "epoch": 7.63, + "learning_rate": 7.104113578449217e-07, + "loss": 0.0759, + "step": 41980 + }, + { + "epoch": 7.63, + "learning_rate": 7.103567528212596e-07, + "loss": 0.0064, + "step": 41981 + }, + { + "epoch": 7.63, + "learning_rate": 7.103021477975974e-07, + "loss": 0.015, + "step": 41982 + }, + { + "epoch": 7.63, + "learning_rate": 7.102475427739351e-07, + "loss": 0.0287, + "step": 41983 + }, + { + "epoch": 7.63, + "learning_rate": 7.101929377502731e-07, + "loss": 1.0454, + "step": 41984 + }, + { + "epoch": 7.64, + "learning_rate": 7.101383327266108e-07, + "loss": 0.0702, + "step": 41985 + }, + { + "epoch": 7.64, + "learning_rate": 7.100837277029487e-07, + "loss": 0.7, + "step": 41986 + }, + { + "epoch": 7.64, + "learning_rate": 7.100291226792866e-07, + "loss": 0.709, + "step": 41987 + }, + { + "epoch": 7.64, + "learning_rate": 7.099745176556243e-07, + "loss": 2.1916, + "step": 41988 + }, + { + "epoch": 7.64, + "learning_rate": 7.099199126319622e-07, + "loss": 2.1389, + "step": 41989 + }, + { + "epoch": 7.64, + "learning_rate": 7.098653076083e-07, + "loss": 0.8036, + "step": 41990 + }, + { + "epoch": 7.64, + "learning_rate": 7.098107025846378e-07, + "loss": 0.961, + "step": 41991 + }, + { + "epoch": 7.64, + "learning_rate": 7.097560975609756e-07, + "loss": 0.0171, + "step": 41992 + }, + { + "epoch": 7.64, + "learning_rate": 7.097014925373134e-07, + "loss": 0.9465, + "step": 41993 + }, + { + "epoch": 7.64, + "learning_rate": 7.096468875136514e-07, + "loss": 0.1781, + "step": 41994 + }, + { + "epoch": 7.64, + "learning_rate": 7.095922824899891e-07, + "loss": 0.0072, + "step": 41995 + }, + { + "epoch": 7.64, + "learning_rate": 7.095376774663269e-07, + "loss": 0.3175, + "step": 41996 + }, + { + "epoch": 7.64, + "learning_rate": 7.094830724426648e-07, + "loss": 0.4446, + "step": 41997 + }, + { + "epoch": 7.64, + "learning_rate": 7.094284674190026e-07, + "loss": 0.1143, + "step": 41998 + }, + { + "epoch": 7.64, + "learning_rate": 7.093738623953403e-07, + "loss": 0.1882, + "step": 41999 + }, + { + "epoch": 7.64, + "learning_rate": 7.093192573716782e-07, + "loss": 2.2427, + "step": 42000 + }, + { + "epoch": 7.64, + "learning_rate": 7.092646523480161e-07, + "loss": 0.0161, + "step": 42001 + }, + { + "epoch": 7.64, + "learning_rate": 7.092100473243538e-07, + "loss": 0.0129, + "step": 42002 + }, + { + "epoch": 7.64, + "learning_rate": 7.091554423006917e-07, + "loss": 0.1742, + "step": 42003 + }, + { + "epoch": 7.64, + "learning_rate": 7.091008372770295e-07, + "loss": 0.2688, + "step": 42004 + }, + { + "epoch": 7.64, + "learning_rate": 7.090462322533674e-07, + "loss": 0.0119, + "step": 42005 + }, + { + "epoch": 7.64, + "learning_rate": 7.089916272297051e-07, + "loss": 0.4595, + "step": 42006 + }, + { + "epoch": 7.64, + "learning_rate": 7.089370222060429e-07, + "loss": 0.0715, + "step": 42007 + }, + { + "epoch": 7.64, + "learning_rate": 7.088824171823809e-07, + "loss": 0.0038, + "step": 42008 + }, + { + "epoch": 7.64, + "learning_rate": 7.088278121587186e-07, + "loss": 0.8304, + "step": 42009 + }, + { + "epoch": 7.64, + "learning_rate": 7.087732071350564e-07, + "loss": 0.8743, + "step": 42010 + }, + { + "epoch": 7.64, + "learning_rate": 7.087186021113943e-07, + "loss": 0.1408, + "step": 42011 + }, + { + "epoch": 7.64, + "learning_rate": 7.086639970877321e-07, + "loss": 1.1119, + "step": 42012 + }, + { + "epoch": 7.64, + "learning_rate": 7.086093920640698e-07, + "loss": 0.9371, + "step": 42013 + }, + { + "epoch": 7.64, + "learning_rate": 7.085547870404078e-07, + "loss": 0.0125, + "step": 42014 + }, + { + "epoch": 7.64, + "learning_rate": 7.085001820167455e-07, + "loss": 0.3175, + "step": 42015 + }, + { + "epoch": 7.64, + "learning_rate": 7.084455769930834e-07, + "loss": 0.0093, + "step": 42016 + }, + { + "epoch": 7.64, + "learning_rate": 7.083909719694212e-07, + "loss": 0.0802, + "step": 42017 + }, + { + "epoch": 7.64, + "learning_rate": 7.08336366945759e-07, + "loss": 0.0176, + "step": 42018 + }, + { + "epoch": 7.64, + "learning_rate": 7.082817619220969e-07, + "loss": 0.0125, + "step": 42019 + }, + { + "epoch": 7.64, + "learning_rate": 7.082271568984346e-07, + "loss": 0.0041, + "step": 42020 + }, + { + "epoch": 7.64, + "learning_rate": 7.081725518747725e-07, + "loss": 1.0024, + "step": 42021 + }, + { + "epoch": 7.64, + "learning_rate": 7.081179468511103e-07, + "loss": 1.3144, + "step": 42022 + }, + { + "epoch": 7.64, + "learning_rate": 7.080633418274481e-07, + "loss": 0.0335, + "step": 42023 + }, + { + "epoch": 7.64, + "learning_rate": 7.08008736803786e-07, + "loss": 2.2601, + "step": 42024 + }, + { + "epoch": 7.64, + "learning_rate": 7.079541317801238e-07, + "loss": 1.6421, + "step": 42025 + }, + { + "epoch": 7.64, + "learning_rate": 7.078995267564616e-07, + "loss": 1.2539, + "step": 42026 + }, + { + "epoch": 7.64, + "learning_rate": 7.078449217327994e-07, + "loss": 0.8614, + "step": 42027 + }, + { + "epoch": 7.64, + "learning_rate": 7.077903167091373e-07, + "loss": 1.0507, + "step": 42028 + }, + { + "epoch": 7.64, + "learning_rate": 7.07735711685475e-07, + "loss": 0.0786, + "step": 42029 + }, + { + "epoch": 7.64, + "learning_rate": 7.076811066618129e-07, + "loss": 0.0366, + "step": 42030 + }, + { + "epoch": 7.64, + "learning_rate": 7.076265016381508e-07, + "loss": 1.5476, + "step": 42031 + }, + { + "epoch": 7.64, + "learning_rate": 7.075718966144885e-07, + "loss": 0.7606, + "step": 42032 + }, + { + "epoch": 7.64, + "learning_rate": 7.075172915908264e-07, + "loss": 0.1224, + "step": 42033 + }, + { + "epoch": 7.64, + "learning_rate": 7.074626865671642e-07, + "loss": 0.4651, + "step": 42034 + }, + { + "epoch": 7.64, + "learning_rate": 7.074080815435021e-07, + "loss": 0.0036, + "step": 42035 + }, + { + "epoch": 7.64, + "learning_rate": 7.073534765198398e-07, + "loss": 0.343, + "step": 42036 + }, + { + "epoch": 7.64, + "learning_rate": 7.072988714961776e-07, + "loss": 1.0675, + "step": 42037 + }, + { + "epoch": 7.64, + "learning_rate": 7.072442664725156e-07, + "loss": 0.2598, + "step": 42038 + }, + { + "epoch": 7.64, + "learning_rate": 7.071896614488533e-07, + "loss": 0.7658, + "step": 42039 + }, + { + "epoch": 7.65, + "learning_rate": 7.071350564251911e-07, + "loss": 0.1499, + "step": 42040 + }, + { + "epoch": 7.65, + "learning_rate": 7.07080451401529e-07, + "loss": 1.9737, + "step": 42041 + }, + { + "epoch": 7.65, + "learning_rate": 7.070258463778668e-07, + "loss": 1.2963, + "step": 42042 + }, + { + "epoch": 7.65, + "learning_rate": 7.069712413542045e-07, + "loss": 0.2548, + "step": 42043 + }, + { + "epoch": 7.65, + "learning_rate": 7.069166363305424e-07, + "loss": 0.1231, + "step": 42044 + }, + { + "epoch": 7.65, + "learning_rate": 7.068620313068803e-07, + "loss": 0.2448, + "step": 42045 + }, + { + "epoch": 7.65, + "learning_rate": 7.068074262832181e-07, + "loss": 0.8734, + "step": 42046 + }, + { + "epoch": 7.65, + "learning_rate": 7.067528212595559e-07, + "loss": 0.2014, + "step": 42047 + }, + { + "epoch": 7.65, + "learning_rate": 7.066982162358937e-07, + "loss": 0.2108, + "step": 42048 + }, + { + "epoch": 7.65, + "learning_rate": 7.066436112122316e-07, + "loss": 1.276, + "step": 42049 + }, + { + "epoch": 7.65, + "learning_rate": 7.065890061885693e-07, + "loss": 0.5148, + "step": 42050 + }, + { + "epoch": 7.65, + "learning_rate": 7.065344011649072e-07, + "loss": 0.0357, + "step": 42051 + }, + { + "epoch": 7.65, + "learning_rate": 7.06479796141245e-07, + "loss": 0.0548, + "step": 42052 + }, + { + "epoch": 7.65, + "learning_rate": 7.064251911175828e-07, + "loss": 0.1472, + "step": 42053 + }, + { + "epoch": 7.65, + "learning_rate": 7.063705860939207e-07, + "loss": 2.6697, + "step": 42054 + }, + { + "epoch": 7.65, + "learning_rate": 7.063159810702585e-07, + "loss": 0.4711, + "step": 42055 + }, + { + "epoch": 7.65, + "learning_rate": 7.062613760465963e-07, + "loss": 0.1052, + "step": 42056 + }, + { + "epoch": 7.65, + "learning_rate": 7.062067710229341e-07, + "loss": 0.1128, + "step": 42057 + }, + { + "epoch": 7.65, + "learning_rate": 7.06152165999272e-07, + "loss": 0.0081, + "step": 42058 + }, + { + "epoch": 7.65, + "learning_rate": 7.060975609756097e-07, + "loss": 0.2313, + "step": 42059 + }, + { + "epoch": 7.65, + "learning_rate": 7.060429559519476e-07, + "loss": 0.4712, + "step": 42060 + }, + { + "epoch": 7.65, + "learning_rate": 7.059883509282854e-07, + "loss": 0.0057, + "step": 42061 + }, + { + "epoch": 7.65, + "learning_rate": 7.059337459046232e-07, + "loss": 0.1255, + "step": 42062 + }, + { + "epoch": 7.65, + "learning_rate": 7.058791408809611e-07, + "loss": 0.3847, + "step": 42063 + }, + { + "epoch": 7.65, + "learning_rate": 7.058245358572988e-07, + "loss": 1.0044, + "step": 42064 + }, + { + "epoch": 7.65, + "learning_rate": 7.057699308336368e-07, + "loss": 0.0738, + "step": 42065 + }, + { + "epoch": 7.65, + "learning_rate": 7.057153258099745e-07, + "loss": 1.4018, + "step": 42066 + }, + { + "epoch": 7.65, + "learning_rate": 7.056607207863123e-07, + "loss": 0.0192, + "step": 42067 + }, + { + "epoch": 7.65, + "learning_rate": 7.056061157626502e-07, + "loss": 1.7078, + "step": 42068 + }, + { + "epoch": 7.65, + "learning_rate": 7.05551510738988e-07, + "loss": 0.1164, + "step": 42069 + }, + { + "epoch": 7.65, + "learning_rate": 7.054969057153258e-07, + "loss": 2.3546, + "step": 42070 + }, + { + "epoch": 7.65, + "learning_rate": 7.054423006916637e-07, + "loss": 1.6606, + "step": 42071 + }, + { + "epoch": 7.65, + "learning_rate": 7.053876956680015e-07, + "loss": 0.0477, + "step": 42072 + }, + { + "epoch": 7.65, + "learning_rate": 7.053330906443392e-07, + "loss": 0.022, + "step": 42073 + }, + { + "epoch": 7.65, + "learning_rate": 7.052784856206771e-07, + "loss": 0.2072, + "step": 42074 + }, + { + "epoch": 7.65, + "learning_rate": 7.05223880597015e-07, + "loss": 0.0169, + "step": 42075 + }, + { + "epoch": 7.65, + "learning_rate": 7.051692755733528e-07, + "loss": 1.4256, + "step": 42076 + }, + { + "epoch": 7.65, + "learning_rate": 7.051146705496906e-07, + "loss": 0.2545, + "step": 42077 + }, + { + "epoch": 7.65, + "learning_rate": 7.050600655260284e-07, + "loss": 2.1055, + "step": 42078 + }, + { + "epoch": 7.65, + "learning_rate": 7.050054605023663e-07, + "loss": 0.0063, + "step": 42079 + }, + { + "epoch": 7.65, + "learning_rate": 7.04950855478704e-07, + "loss": 0.7793, + "step": 42080 + }, + { + "epoch": 7.65, + "learning_rate": 7.048962504550418e-07, + "loss": 0.1929, + "step": 42081 + }, + { + "epoch": 7.65, + "learning_rate": 7.048416454313798e-07, + "loss": 0.0837, + "step": 42082 + }, + { + "epoch": 7.65, + "learning_rate": 7.047870404077175e-07, + "loss": 0.1512, + "step": 42083 + }, + { + "epoch": 7.65, + "learning_rate": 7.047324353840554e-07, + "loss": 0.0067, + "step": 42084 + }, + { + "epoch": 7.65, + "learning_rate": 7.046778303603932e-07, + "loss": 0.0143, + "step": 42085 + }, + { + "epoch": 7.65, + "learning_rate": 7.04623225336731e-07, + "loss": 0.3886, + "step": 42086 + }, + { + "epoch": 7.65, + "learning_rate": 7.045686203130688e-07, + "loss": 0.2274, + "step": 42087 + }, + { + "epoch": 7.65, + "learning_rate": 7.045140152894066e-07, + "loss": 1.2302, + "step": 42088 + }, + { + "epoch": 7.65, + "learning_rate": 7.044594102657445e-07, + "loss": 0.8635, + "step": 42089 + }, + { + "epoch": 7.65, + "learning_rate": 7.044048052420823e-07, + "loss": 0.0058, + "step": 42090 + }, + { + "epoch": 7.65, + "learning_rate": 7.043502002184201e-07, + "loss": 1.1669, + "step": 42091 + }, + { + "epoch": 7.65, + "learning_rate": 7.04295595194758e-07, + "loss": 0.3908, + "step": 42092 + }, + { + "epoch": 7.65, + "learning_rate": 7.042409901710958e-07, + "loss": 0.0441, + "step": 42093 + }, + { + "epoch": 7.65, + "learning_rate": 7.041863851474335e-07, + "loss": 0.2868, + "step": 42094 + }, + { + "epoch": 7.66, + "learning_rate": 7.041317801237715e-07, + "loss": 0.3126, + "step": 42095 + }, + { + "epoch": 7.66, + "learning_rate": 7.040771751001092e-07, + "loss": 0.0831, + "step": 42096 + }, + { + "epoch": 7.66, + "learning_rate": 7.04022570076447e-07, + "loss": 0.9431, + "step": 42097 + }, + { + "epoch": 7.66, + "learning_rate": 7.039679650527849e-07, + "loss": 0.6167, + "step": 42098 + }, + { + "epoch": 7.66, + "learning_rate": 7.039133600291227e-07, + "loss": 1.1389, + "step": 42099 + }, + { + "epoch": 7.66, + "learning_rate": 7.038587550054605e-07, + "loss": 1.4514, + "step": 42100 + }, + { + "epoch": 7.66, + "learning_rate": 7.038041499817983e-07, + "loss": 0.5095, + "step": 42101 + }, + { + "epoch": 7.66, + "learning_rate": 7.037495449581362e-07, + "loss": 0.2585, + "step": 42102 + }, + { + "epoch": 7.66, + "learning_rate": 7.03694939934474e-07, + "loss": 0.4627, + "step": 42103 + }, + { + "epoch": 7.66, + "learning_rate": 7.036403349108118e-07, + "loss": 0.3776, + "step": 42104 + }, + { + "epoch": 7.66, + "learning_rate": 7.035857298871496e-07, + "loss": 0.0121, + "step": 42105 + }, + { + "epoch": 7.66, + "learning_rate": 7.035311248634875e-07, + "loss": 0.1452, + "step": 42106 + }, + { + "epoch": 7.66, + "learning_rate": 7.034765198398253e-07, + "loss": 1.6548, + "step": 42107 + }, + { + "epoch": 7.66, + "learning_rate": 7.03421914816163e-07, + "loss": 0.0547, + "step": 42108 + }, + { + "epoch": 7.66, + "learning_rate": 7.03367309792501e-07, + "loss": 0.0368, + "step": 42109 + }, + { + "epoch": 7.66, + "learning_rate": 7.033127047688387e-07, + "loss": 1.8263, + "step": 42110 + }, + { + "epoch": 7.66, + "learning_rate": 7.032580997451765e-07, + "loss": 1.8317, + "step": 42111 + }, + { + "epoch": 7.66, + "learning_rate": 7.032034947215144e-07, + "loss": 0.1612, + "step": 42112 + }, + { + "epoch": 7.66, + "learning_rate": 7.031488896978522e-07, + "loss": 1.2733, + "step": 42113 + }, + { + "epoch": 7.66, + "learning_rate": 7.030942846741901e-07, + "loss": 0.2832, + "step": 42114 + }, + { + "epoch": 7.66, + "learning_rate": 7.030396796505279e-07, + "loss": 0.3669, + "step": 42115 + }, + { + "epoch": 7.66, + "learning_rate": 7.029850746268657e-07, + "loss": 0.7692, + "step": 42116 + }, + { + "epoch": 7.66, + "learning_rate": 7.029304696032035e-07, + "loss": 0.1279, + "step": 42117 + }, + { + "epoch": 7.66, + "learning_rate": 7.028758645795413e-07, + "loss": 0.832, + "step": 42118 + }, + { + "epoch": 7.66, + "learning_rate": 7.028212595558792e-07, + "loss": 0.1056, + "step": 42119 + }, + { + "epoch": 7.66, + "learning_rate": 7.02766654532217e-07, + "loss": 0.409, + "step": 42120 + }, + { + "epoch": 7.66, + "learning_rate": 7.027120495085548e-07, + "loss": 0.1869, + "step": 42121 + }, + { + "epoch": 7.66, + "learning_rate": 7.026574444848927e-07, + "loss": 1.6277, + "step": 42122 + }, + { + "epoch": 7.66, + "learning_rate": 7.026028394612305e-07, + "loss": 2.7633, + "step": 42123 + }, + { + "epoch": 7.66, + "learning_rate": 7.025482344375682e-07, + "loss": 0.0127, + "step": 42124 + }, + { + "epoch": 7.66, + "learning_rate": 7.024936294139061e-07, + "loss": 0.0302, + "step": 42125 + }, + { + "epoch": 7.66, + "learning_rate": 7.02439024390244e-07, + "loss": 0.7942, + "step": 42126 + }, + { + "epoch": 7.66, + "learning_rate": 7.023844193665817e-07, + "loss": 0.346, + "step": 42127 + }, + { + "epoch": 7.66, + "learning_rate": 7.023298143429196e-07, + "loss": 0.777, + "step": 42128 + }, + { + "epoch": 7.66, + "learning_rate": 7.022752093192574e-07, + "loss": 0.0443, + "step": 42129 + }, + { + "epoch": 7.66, + "learning_rate": 7.022206042955952e-07, + "loss": 0.0047, + "step": 42130 + }, + { + "epoch": 7.66, + "learning_rate": 7.02165999271933e-07, + "loss": 0.1888, + "step": 42131 + }, + { + "epoch": 7.66, + "learning_rate": 7.021113942482708e-07, + "loss": 0.0539, + "step": 42132 + }, + { + "epoch": 7.66, + "learning_rate": 7.020567892246087e-07, + "loss": 0.0109, + "step": 42133 + }, + { + "epoch": 7.66, + "learning_rate": 7.020021842009465e-07, + "loss": 0.0075, + "step": 42134 + }, + { + "epoch": 7.66, + "learning_rate": 7.019475791772843e-07, + "loss": 1.1052, + "step": 42135 + }, + { + "epoch": 7.66, + "learning_rate": 7.018929741536222e-07, + "loss": 0.0841, + "step": 42136 + }, + { + "epoch": 7.66, + "learning_rate": 7.0183836912996e-07, + "loss": 0.946, + "step": 42137 + }, + { + "epoch": 7.66, + "learning_rate": 7.017837641062977e-07, + "loss": 0.5183, + "step": 42138 + }, + { + "epoch": 7.66, + "learning_rate": 7.017291590826357e-07, + "loss": 0.0576, + "step": 42139 + }, + { + "epoch": 7.66, + "learning_rate": 7.016745540589734e-07, + "loss": 1.366, + "step": 42140 + }, + { + "epoch": 7.66, + "learning_rate": 7.016199490353112e-07, + "loss": 1.3872, + "step": 42141 + }, + { + "epoch": 7.66, + "learning_rate": 7.015653440116491e-07, + "loss": 0.1754, + "step": 42142 + }, + { + "epoch": 7.66, + "learning_rate": 7.015107389879869e-07, + "loss": 0.3149, + "step": 42143 + }, + { + "epoch": 7.66, + "learning_rate": 7.014561339643248e-07, + "loss": 0.3318, + "step": 42144 + }, + { + "epoch": 7.66, + "learning_rate": 7.014015289406625e-07, + "loss": 0.2376, + "step": 42145 + }, + { + "epoch": 7.66, + "learning_rate": 7.013469239170004e-07, + "loss": 0.0335, + "step": 42146 + }, + { + "epoch": 7.66, + "learning_rate": 7.012923188933382e-07, + "loss": 0.3253, + "step": 42147 + }, + { + "epoch": 7.66, + "learning_rate": 7.01237713869676e-07, + "loss": 1.6653, + "step": 42148 + }, + { + "epoch": 7.66, + "learning_rate": 7.011831088460138e-07, + "loss": 0.2913, + "step": 42149 + }, + { + "epoch": 7.67, + "learning_rate": 7.011285038223517e-07, + "loss": 0.581, + "step": 42150 + }, + { + "epoch": 7.67, + "learning_rate": 7.010738987986895e-07, + "loss": 0.0126, + "step": 42151 + }, + { + "epoch": 7.67, + "learning_rate": 7.010192937750272e-07, + "loss": 0.029, + "step": 42152 + }, + { + "epoch": 7.67, + "learning_rate": 7.009646887513652e-07, + "loss": 0.7662, + "step": 42153 + }, + { + "epoch": 7.67, + "learning_rate": 7.009100837277029e-07, + "loss": 0.0823, + "step": 42154 + }, + { + "epoch": 7.67, + "learning_rate": 7.008554787040408e-07, + "loss": 0.075, + "step": 42155 + }, + { + "epoch": 7.67, + "learning_rate": 7.008008736803786e-07, + "loss": 2.9087, + "step": 42156 + }, + { + "epoch": 7.67, + "learning_rate": 7.007462686567164e-07, + "loss": 2.1406, + "step": 42157 + }, + { + "epoch": 7.67, + "learning_rate": 7.006916636330543e-07, + "loss": 0.5152, + "step": 42158 + }, + { + "epoch": 7.67, + "learning_rate": 7.006370586093921e-07, + "loss": 0.1678, + "step": 42159 + }, + { + "epoch": 7.67, + "learning_rate": 7.0058245358573e-07, + "loss": 0.0417, + "step": 42160 + }, + { + "epoch": 7.67, + "learning_rate": 7.005278485620677e-07, + "loss": 0.2555, + "step": 42161 + }, + { + "epoch": 7.67, + "learning_rate": 7.004732435384055e-07, + "loss": 0.5028, + "step": 42162 + }, + { + "epoch": 7.67, + "learning_rate": 7.004186385147435e-07, + "loss": 0.2765, + "step": 42163 + }, + { + "epoch": 7.67, + "learning_rate": 7.003640334910812e-07, + "loss": 0.0691, + "step": 42164 + }, + { + "epoch": 7.67, + "learning_rate": 7.00309428467419e-07, + "loss": 0.0618, + "step": 42165 + }, + { + "epoch": 7.67, + "learning_rate": 7.002548234437569e-07, + "loss": 0.014, + "step": 42166 + }, + { + "epoch": 7.67, + "learning_rate": 7.002002184200947e-07, + "loss": 0.0043, + "step": 42167 + }, + { + "epoch": 7.67, + "learning_rate": 7.001456133964324e-07, + "loss": 0.2298, + "step": 42168 + }, + { + "epoch": 7.67, + "learning_rate": 7.000910083727703e-07, + "loss": 0.0033, + "step": 42169 + }, + { + "epoch": 7.67, + "learning_rate": 7.000364033491081e-07, + "loss": 0.4997, + "step": 42170 + }, + { + "epoch": 7.67, + "learning_rate": 6.99981798325446e-07, + "loss": 0.0348, + "step": 42171 + }, + { + "epoch": 7.67, + "learning_rate": 6.999271933017838e-07, + "loss": 1.2285, + "step": 42172 + }, + { + "epoch": 7.67, + "learning_rate": 6.998725882781216e-07, + "loss": 1.8123, + "step": 42173 + }, + { + "epoch": 7.67, + "learning_rate": 6.998179832544595e-07, + "loss": 0.9775, + "step": 42174 + }, + { + "epoch": 7.67, + "learning_rate": 6.997633782307972e-07, + "loss": 0.0355, + "step": 42175 + }, + { + "epoch": 7.67, + "learning_rate": 6.99708773207135e-07, + "loss": 0.8208, + "step": 42176 + }, + { + "epoch": 7.67, + "learning_rate": 6.996541681834729e-07, + "loss": 0.0644, + "step": 42177 + }, + { + "epoch": 7.67, + "learning_rate": 6.995995631598107e-07, + "loss": 0.0794, + "step": 42178 + }, + { + "epoch": 7.67, + "learning_rate": 6.995449581361485e-07, + "loss": 0.0053, + "step": 42179 + }, + { + "epoch": 7.67, + "learning_rate": 6.994903531124864e-07, + "loss": 0.0623, + "step": 42180 + }, + { + "epoch": 7.67, + "learning_rate": 6.994357480888242e-07, + "loss": 1.0856, + "step": 42181 + }, + { + "epoch": 7.67, + "learning_rate": 6.99381143065162e-07, + "loss": 1.6376, + "step": 42182 + }, + { + "epoch": 7.67, + "learning_rate": 6.993265380414999e-07, + "loss": 0.2875, + "step": 42183 + }, + { + "epoch": 7.67, + "learning_rate": 6.992719330178376e-07, + "loss": 0.439, + "step": 42184 + }, + { + "epoch": 7.67, + "learning_rate": 6.992173279941755e-07, + "loss": 0.1948, + "step": 42185 + }, + { + "epoch": 7.67, + "learning_rate": 6.991627229705133e-07, + "loss": 0.467, + "step": 42186 + }, + { + "epoch": 7.67, + "learning_rate": 6.991081179468511e-07, + "loss": 0.6148, + "step": 42187 + }, + { + "epoch": 7.67, + "learning_rate": 6.99053512923189e-07, + "loss": 0.0689, + "step": 42188 + }, + { + "epoch": 7.67, + "learning_rate": 6.989989078995267e-07, + "loss": 0.0433, + "step": 42189 + }, + { + "epoch": 7.67, + "learning_rate": 6.989443028758647e-07, + "loss": 0.7783, + "step": 42190 + }, + { + "epoch": 7.67, + "learning_rate": 6.988896978522024e-07, + "loss": 0.1672, + "step": 42191 + }, + { + "epoch": 7.67, + "learning_rate": 6.988350928285402e-07, + "loss": 0.1311, + "step": 42192 + }, + { + "epoch": 7.67, + "learning_rate": 6.987804878048781e-07, + "loss": 0.1369, + "step": 42193 + }, + { + "epoch": 7.67, + "learning_rate": 6.987258827812159e-07, + "loss": 0.6153, + "step": 42194 + }, + { + "epoch": 7.67, + "learning_rate": 6.986712777575537e-07, + "loss": 0.4262, + "step": 42195 + }, + { + "epoch": 7.67, + "learning_rate": 6.986166727338915e-07, + "loss": 0.241, + "step": 42196 + }, + { + "epoch": 7.67, + "learning_rate": 6.985620677102294e-07, + "loss": 0.4531, + "step": 42197 + }, + { + "epoch": 7.67, + "learning_rate": 6.985074626865671e-07, + "loss": 0.0832, + "step": 42198 + }, + { + "epoch": 7.67, + "learning_rate": 6.98452857662905e-07, + "loss": 0.0283, + "step": 42199 + }, + { + "epoch": 7.67, + "learning_rate": 6.983982526392429e-07, + "loss": 0.084, + "step": 42200 + }, + { + "epoch": 7.67, + "learning_rate": 6.983436476155807e-07, + "loss": 0.4436, + "step": 42201 + }, + { + "epoch": 7.67, + "learning_rate": 6.982890425919185e-07, + "loss": 0.1328, + "step": 42202 + }, + { + "epoch": 7.67, + "learning_rate": 6.982344375682563e-07, + "loss": 0.8071, + "step": 42203 + }, + { + "epoch": 7.67, + "learning_rate": 6.981798325445942e-07, + "loss": 0.1223, + "step": 42204 + }, + { + "epoch": 7.68, + "learning_rate": 6.981252275209319e-07, + "loss": 0.2847, + "step": 42205 + }, + { + "epoch": 7.68, + "learning_rate": 6.980706224972697e-07, + "loss": 0.525, + "step": 42206 + }, + { + "epoch": 7.68, + "learning_rate": 6.980160174736076e-07, + "loss": 0.6804, + "step": 42207 + }, + { + "epoch": 7.68, + "learning_rate": 6.979614124499454e-07, + "loss": 2.8209, + "step": 42208 + }, + { + "epoch": 7.68, + "learning_rate": 6.979068074262832e-07, + "loss": 0.1113, + "step": 42209 + }, + { + "epoch": 7.68, + "learning_rate": 6.978522024026211e-07, + "loss": 0.1239, + "step": 42210 + }, + { + "epoch": 7.68, + "learning_rate": 6.977975973789589e-07, + "loss": 0.54, + "step": 42211 + }, + { + "epoch": 7.68, + "learning_rate": 6.977429923552967e-07, + "loss": 0.0103, + "step": 42212 + }, + { + "epoch": 7.68, + "learning_rate": 6.976883873316345e-07, + "loss": 0.0171, + "step": 42213 + }, + { + "epoch": 7.68, + "learning_rate": 6.976337823079723e-07, + "loss": 0.0368, + "step": 42214 + }, + { + "epoch": 7.68, + "learning_rate": 6.975791772843102e-07, + "loss": 0.0068, + "step": 42215 + }, + { + "epoch": 7.68, + "learning_rate": 6.97524572260648e-07, + "loss": 1.2276, + "step": 42216 + }, + { + "epoch": 7.68, + "learning_rate": 6.974699672369858e-07, + "loss": 0.0106, + "step": 42217 + }, + { + "epoch": 7.68, + "learning_rate": 6.974153622133237e-07, + "loss": 4.1393, + "step": 42218 + }, + { + "epoch": 7.68, + "learning_rate": 6.973607571896614e-07, + "loss": 1.2738, + "step": 42219 + }, + { + "epoch": 7.68, + "learning_rate": 6.973061521659994e-07, + "loss": 0.6376, + "step": 42220 + }, + { + "epoch": 7.68, + "learning_rate": 6.972515471423371e-07, + "loss": 0.0071, + "step": 42221 + }, + { + "epoch": 7.68, + "learning_rate": 6.971969421186749e-07, + "loss": 1.0958, + "step": 42222 + }, + { + "epoch": 7.68, + "learning_rate": 6.971423370950128e-07, + "loss": 0.0196, + "step": 42223 + }, + { + "epoch": 7.68, + "learning_rate": 6.970877320713506e-07, + "loss": 0.1114, + "step": 42224 + }, + { + "epoch": 7.68, + "learning_rate": 6.970331270476884e-07, + "loss": 0.0603, + "step": 42225 + }, + { + "epoch": 7.68, + "learning_rate": 6.969785220240262e-07, + "loss": 1.727, + "step": 42226 + }, + { + "epoch": 7.68, + "learning_rate": 6.969239170003641e-07, + "loss": 0.1197, + "step": 42227 + }, + { + "epoch": 7.68, + "learning_rate": 6.968693119767018e-07, + "loss": 0.1687, + "step": 42228 + }, + { + "epoch": 7.68, + "learning_rate": 6.968147069530397e-07, + "loss": 0.0952, + "step": 42229 + }, + { + "epoch": 7.68, + "learning_rate": 6.967601019293775e-07, + "loss": 0.0137, + "step": 42230 + }, + { + "epoch": 7.68, + "learning_rate": 6.967054969057154e-07, + "loss": 0.0259, + "step": 42231 + }, + { + "epoch": 7.68, + "learning_rate": 6.966508918820532e-07, + "loss": 0.5644, + "step": 42232 + }, + { + "epoch": 7.68, + "learning_rate": 6.965962868583909e-07, + "loss": 0.4996, + "step": 42233 + }, + { + "epoch": 7.68, + "learning_rate": 6.965416818347289e-07, + "loss": 0.0739, + "step": 42234 + }, + { + "epoch": 7.68, + "learning_rate": 6.964870768110666e-07, + "loss": 0.0091, + "step": 42235 + }, + { + "epoch": 7.68, + "learning_rate": 6.964324717874044e-07, + "loss": 0.3619, + "step": 42236 + }, + { + "epoch": 7.68, + "learning_rate": 6.963778667637423e-07, + "loss": 0.1415, + "step": 42237 + }, + { + "epoch": 7.68, + "learning_rate": 6.963232617400801e-07, + "loss": 0.0516, + "step": 42238 + }, + { + "epoch": 7.68, + "learning_rate": 6.96268656716418e-07, + "loss": 3.1525, + "step": 42239 + }, + { + "epoch": 7.68, + "learning_rate": 6.962140516927557e-07, + "loss": 0.1504, + "step": 42240 + }, + { + "epoch": 7.68, + "learning_rate": 6.961594466690936e-07, + "loss": 1.0213, + "step": 42241 + }, + { + "epoch": 7.68, + "learning_rate": 6.961048416454314e-07, + "loss": 0.5041, + "step": 42242 + }, + { + "epoch": 7.68, + "learning_rate": 6.960502366217692e-07, + "loss": 0.0712, + "step": 42243 + }, + { + "epoch": 7.68, + "learning_rate": 6.959956315981071e-07, + "loss": 0.2412, + "step": 42244 + }, + { + "epoch": 7.68, + "learning_rate": 6.959410265744449e-07, + "loss": 0.0323, + "step": 42245 + }, + { + "epoch": 7.68, + "learning_rate": 6.958864215507827e-07, + "loss": 0.1582, + "step": 42246 + }, + { + "epoch": 7.68, + "learning_rate": 6.958318165271205e-07, + "loss": 0.1516, + "step": 42247 + }, + { + "epoch": 7.68, + "learning_rate": 6.957772115034584e-07, + "loss": 0.2142, + "step": 42248 + }, + { + "epoch": 7.68, + "learning_rate": 6.957226064797961e-07, + "loss": 0.4587, + "step": 42249 + }, + { + "epoch": 7.68, + "learning_rate": 6.95668001456134e-07, + "loss": 0.2753, + "step": 42250 + }, + { + "epoch": 7.68, + "learning_rate": 6.956133964324718e-07, + "loss": 0.3449, + "step": 42251 + }, + { + "epoch": 7.68, + "learning_rate": 6.955587914088096e-07, + "loss": 0.0949, + "step": 42252 + }, + { + "epoch": 7.68, + "learning_rate": 6.955041863851475e-07, + "loss": 2.9253, + "step": 42253 + }, + { + "epoch": 7.68, + "learning_rate": 6.954495813614853e-07, + "loss": 0.0934, + "step": 42254 + }, + { + "epoch": 7.68, + "learning_rate": 6.953949763378231e-07, + "loss": 0.0056, + "step": 42255 + }, + { + "epoch": 7.68, + "learning_rate": 6.953403713141609e-07, + "loss": 1.5823, + "step": 42256 + }, + { + "epoch": 7.68, + "learning_rate": 6.952857662904987e-07, + "loss": 2.0531, + "step": 42257 + }, + { + "epoch": 7.68, + "learning_rate": 6.952311612668365e-07, + "loss": 0.2063, + "step": 42258 + }, + { + "epoch": 7.68, + "learning_rate": 6.951765562431744e-07, + "loss": 0.9707, + "step": 42259 + }, + { + "epoch": 7.69, + "learning_rate": 6.951219512195122e-07, + "loss": 0.5358, + "step": 42260 + }, + { + "epoch": 7.69, + "learning_rate": 6.950673461958501e-07, + "loss": 0.0539, + "step": 42261 + }, + { + "epoch": 7.69, + "learning_rate": 6.950127411721879e-07, + "loss": 0.2841, + "step": 42262 + }, + { + "epoch": 7.69, + "learning_rate": 6.949581361485256e-07, + "loss": 0.0098, + "step": 42263 + }, + { + "epoch": 7.69, + "learning_rate": 6.949035311248636e-07, + "loss": 0.1519, + "step": 42264 + }, + { + "epoch": 7.69, + "learning_rate": 6.948489261012013e-07, + "loss": 0.0114, + "step": 42265 + }, + { + "epoch": 7.69, + "learning_rate": 6.947943210775391e-07, + "loss": 0.0446, + "step": 42266 + }, + { + "epoch": 7.69, + "learning_rate": 6.94739716053877e-07, + "loss": 0.0123, + "step": 42267 + }, + { + "epoch": 7.69, + "learning_rate": 6.946851110302148e-07, + "loss": 0.8941, + "step": 42268 + }, + { + "epoch": 7.69, + "learning_rate": 6.946305060065526e-07, + "loss": 0.734, + "step": 42269 + }, + { + "epoch": 7.69, + "learning_rate": 6.945759009828904e-07, + "loss": 0.5899, + "step": 42270 + }, + { + "epoch": 7.69, + "learning_rate": 6.945212959592283e-07, + "loss": 0.5701, + "step": 42271 + }, + { + "epoch": 7.69, + "learning_rate": 6.944666909355661e-07, + "loss": 0.505, + "step": 42272 + }, + { + "epoch": 7.69, + "learning_rate": 6.944120859119039e-07, + "loss": 0.3048, + "step": 42273 + }, + { + "epoch": 7.69, + "learning_rate": 6.943574808882417e-07, + "loss": 1.809, + "step": 42274 + }, + { + "epoch": 7.69, + "learning_rate": 6.943028758645796e-07, + "loss": 1.0679, + "step": 42275 + }, + { + "epoch": 7.69, + "learning_rate": 6.942482708409174e-07, + "loss": 0.0235, + "step": 42276 + }, + { + "epoch": 7.69, + "learning_rate": 6.941936658172551e-07, + "loss": 0.2229, + "step": 42277 + }, + { + "epoch": 7.69, + "learning_rate": 6.941390607935931e-07, + "loss": 0.3469, + "step": 42278 + }, + { + "epoch": 7.69, + "learning_rate": 6.940844557699308e-07, + "loss": 0.5481, + "step": 42279 + }, + { + "epoch": 7.69, + "learning_rate": 6.940298507462687e-07, + "loss": 0.3138, + "step": 42280 + }, + { + "epoch": 7.69, + "learning_rate": 6.939752457226065e-07, + "loss": 0.0474, + "step": 42281 + }, + { + "epoch": 7.69, + "learning_rate": 6.939206406989443e-07, + "loss": 0.1898, + "step": 42282 + }, + { + "epoch": 7.69, + "learning_rate": 6.938660356752822e-07, + "loss": 0.0643, + "step": 42283 + }, + { + "epoch": 7.69, + "learning_rate": 6.9381143065162e-07, + "loss": 0.4107, + "step": 42284 + }, + { + "epoch": 7.69, + "learning_rate": 6.937568256279578e-07, + "loss": 0.014, + "step": 42285 + }, + { + "epoch": 7.69, + "learning_rate": 6.937022206042956e-07, + "loss": 0.1598, + "step": 42286 + }, + { + "epoch": 7.69, + "learning_rate": 6.936476155806334e-07, + "loss": 1.326, + "step": 42287 + }, + { + "epoch": 7.69, + "learning_rate": 6.935930105569712e-07, + "loss": 0.9474, + "step": 42288 + }, + { + "epoch": 7.69, + "learning_rate": 6.935384055333091e-07, + "loss": 0.661, + "step": 42289 + }, + { + "epoch": 7.69, + "learning_rate": 6.934838005096469e-07, + "loss": 0.4693, + "step": 42290 + }, + { + "epoch": 7.69, + "learning_rate": 6.934291954859848e-07, + "loss": 0.6302, + "step": 42291 + }, + { + "epoch": 7.69, + "learning_rate": 6.933745904623226e-07, + "loss": 0.4534, + "step": 42292 + }, + { + "epoch": 7.69, + "learning_rate": 6.933199854386603e-07, + "loss": 0.3599, + "step": 42293 + }, + { + "epoch": 7.69, + "learning_rate": 6.932653804149982e-07, + "loss": 0.8018, + "step": 42294 + }, + { + "epoch": 7.69, + "learning_rate": 6.93210775391336e-07, + "loss": 1.1008, + "step": 42295 + }, + { + "epoch": 7.69, + "learning_rate": 6.931561703676738e-07, + "loss": 1.1652, + "step": 42296 + }, + { + "epoch": 7.69, + "learning_rate": 6.931015653440117e-07, + "loss": 0.1121, + "step": 42297 + }, + { + "epoch": 7.69, + "learning_rate": 6.930469603203495e-07, + "loss": 0.5179, + "step": 42298 + }, + { + "epoch": 7.69, + "learning_rate": 6.929923552966874e-07, + "loss": 0.2831, + "step": 42299 + }, + { + "epoch": 7.69, + "learning_rate": 6.929377502730251e-07, + "loss": 0.571, + "step": 42300 + }, + { + "epoch": 7.69, + "learning_rate": 6.928831452493629e-07, + "loss": 0.5468, + "step": 42301 + }, + { + "epoch": 7.69, + "learning_rate": 6.928285402257008e-07, + "loss": 1.6176, + "step": 42302 + }, + { + "epoch": 7.69, + "learning_rate": 6.927739352020386e-07, + "loss": 0.0458, + "step": 42303 + }, + { + "epoch": 7.69, + "learning_rate": 6.927193301783764e-07, + "loss": 0.0637, + "step": 42304 + }, + { + "epoch": 7.69, + "learning_rate": 6.926647251547143e-07, + "loss": 0.2788, + "step": 42305 + }, + { + "epoch": 7.69, + "learning_rate": 6.926101201310521e-07, + "loss": 1.5617, + "step": 42306 + }, + { + "epoch": 7.69, + "learning_rate": 6.925555151073898e-07, + "loss": 0.0797, + "step": 42307 + }, + { + "epoch": 7.69, + "learning_rate": 6.925009100837278e-07, + "loss": 0.2203, + "step": 42308 + }, + { + "epoch": 7.69, + "learning_rate": 6.924463050600655e-07, + "loss": 0.0356, + "step": 42309 + }, + { + "epoch": 7.69, + "learning_rate": 6.923917000364034e-07, + "loss": 0.06, + "step": 42310 + }, + { + "epoch": 7.69, + "learning_rate": 6.923370950127412e-07, + "loss": 1.2196, + "step": 42311 + }, + { + "epoch": 7.69, + "learning_rate": 6.92282489989079e-07, + "loss": 0.0072, + "step": 42312 + }, + { + "epoch": 7.69, + "learning_rate": 6.922278849654169e-07, + "loss": 0.0751, + "step": 42313 + }, + { + "epoch": 7.69, + "learning_rate": 6.921732799417546e-07, + "loss": 0.976, + "step": 42314 + }, + { + "epoch": 7.7, + "learning_rate": 6.921186749180925e-07, + "loss": 0.1356, + "step": 42315 + }, + { + "epoch": 7.7, + "learning_rate": 6.920640698944303e-07, + "loss": 0.0859, + "step": 42316 + }, + { + "epoch": 7.7, + "learning_rate": 6.920094648707681e-07, + "loss": 0.3189, + "step": 42317 + }, + { + "epoch": 7.7, + "learning_rate": 6.919548598471059e-07, + "loss": 2.7706, + "step": 42318 + }, + { + "epoch": 7.7, + "learning_rate": 6.919002548234438e-07, + "loss": 0.3704, + "step": 42319 + }, + { + "epoch": 7.7, + "learning_rate": 6.918456497997816e-07, + "loss": 0.1502, + "step": 42320 + }, + { + "epoch": 7.7, + "learning_rate": 6.917910447761194e-07, + "loss": 0.007, + "step": 42321 + }, + { + "epoch": 7.7, + "learning_rate": 6.917364397524573e-07, + "loss": 0.561, + "step": 42322 + }, + { + "epoch": 7.7, + "learning_rate": 6.91681834728795e-07, + "loss": 0.6006, + "step": 42323 + }, + { + "epoch": 7.7, + "learning_rate": 6.916272297051329e-07, + "loss": 0.4969, + "step": 42324 + }, + { + "epoch": 7.7, + "learning_rate": 6.915726246814707e-07, + "loss": 0.2008, + "step": 42325 + }, + { + "epoch": 7.7, + "learning_rate": 6.915180196578085e-07, + "loss": 0.1524, + "step": 42326 + }, + { + "epoch": 7.7, + "learning_rate": 6.914634146341464e-07, + "loss": 0.7373, + "step": 42327 + }, + { + "epoch": 7.7, + "learning_rate": 6.914088096104842e-07, + "loss": 3.1026, + "step": 42328 + }, + { + "epoch": 7.7, + "learning_rate": 6.913542045868221e-07, + "loss": 1.3539, + "step": 42329 + }, + { + "epoch": 7.7, + "learning_rate": 6.912995995631598e-07, + "loss": 0.0722, + "step": 42330 + }, + { + "epoch": 7.7, + "learning_rate": 6.912449945394976e-07, + "loss": 1.3754, + "step": 42331 + }, + { + "epoch": 7.7, + "learning_rate": 6.911903895158355e-07, + "loss": 1.4618, + "step": 42332 + }, + { + "epoch": 7.7, + "learning_rate": 6.911357844921733e-07, + "loss": 0.4527, + "step": 42333 + }, + { + "epoch": 7.7, + "learning_rate": 6.910811794685111e-07, + "loss": 0.1589, + "step": 42334 + }, + { + "epoch": 7.7, + "learning_rate": 6.91026574444849e-07, + "loss": 0.2347, + "step": 42335 + }, + { + "epoch": 7.7, + "learning_rate": 6.909719694211868e-07, + "loss": 2.8142, + "step": 42336 + }, + { + "epoch": 7.7, + "learning_rate": 6.909173643975245e-07, + "loss": 0.0782, + "step": 42337 + }, + { + "epoch": 7.7, + "learning_rate": 6.908627593738624e-07, + "loss": 0.0049, + "step": 42338 + }, + { + "epoch": 7.7, + "learning_rate": 6.908081543502002e-07, + "loss": 1.0834, + "step": 42339 + }, + { + "epoch": 7.7, + "learning_rate": 6.907535493265381e-07, + "loss": 0.0562, + "step": 42340 + }, + { + "epoch": 7.7, + "learning_rate": 6.906989443028759e-07, + "loss": 0.1016, + "step": 42341 + }, + { + "epoch": 7.7, + "learning_rate": 6.906443392792137e-07, + "loss": 1.5847, + "step": 42342 + }, + { + "epoch": 7.7, + "learning_rate": 6.905897342555516e-07, + "loss": 0.2023, + "step": 42343 + }, + { + "epoch": 7.7, + "learning_rate": 6.905351292318893e-07, + "loss": 0.0048, + "step": 42344 + }, + { + "epoch": 7.7, + "learning_rate": 6.904805242082271e-07, + "loss": 0.9581, + "step": 42345 + }, + { + "epoch": 7.7, + "learning_rate": 6.90425919184565e-07, + "loss": 0.1251, + "step": 42346 + }, + { + "epoch": 7.7, + "learning_rate": 6.903713141609028e-07, + "loss": 0.5009, + "step": 42347 + }, + { + "epoch": 7.7, + "learning_rate": 6.903167091372406e-07, + "loss": 1.1232, + "step": 42348 + }, + { + "epoch": 7.7, + "learning_rate": 6.902621041135785e-07, + "loss": 0.0415, + "step": 42349 + }, + { + "epoch": 7.7, + "learning_rate": 6.902074990899163e-07, + "loss": 0.1052, + "step": 42350 + }, + { + "epoch": 7.7, + "learning_rate": 6.901528940662541e-07, + "loss": 0.0109, + "step": 42351 + }, + { + "epoch": 7.7, + "learning_rate": 6.90098289042592e-07, + "loss": 0.7317, + "step": 42352 + }, + { + "epoch": 7.7, + "learning_rate": 6.900436840189297e-07, + "loss": 0.0381, + "step": 42353 + }, + { + "epoch": 7.7, + "learning_rate": 6.899890789952676e-07, + "loss": 0.0168, + "step": 42354 + }, + { + "epoch": 7.7, + "learning_rate": 6.899344739716054e-07, + "loss": 0.0259, + "step": 42355 + }, + { + "epoch": 7.7, + "learning_rate": 6.898798689479432e-07, + "loss": 1.0074, + "step": 42356 + }, + { + "epoch": 7.7, + "learning_rate": 6.898252639242811e-07, + "loss": 0.1923, + "step": 42357 + }, + { + "epoch": 7.7, + "learning_rate": 6.897706589006188e-07, + "loss": 1.1981, + "step": 42358 + }, + { + "epoch": 7.7, + "learning_rate": 6.897160538769568e-07, + "loss": 0.8821, + "step": 42359 + }, + { + "epoch": 7.7, + "learning_rate": 6.896614488532945e-07, + "loss": 0.0937, + "step": 42360 + }, + { + "epoch": 7.7, + "learning_rate": 6.896068438296323e-07, + "loss": 2.8467, + "step": 42361 + }, + { + "epoch": 7.7, + "learning_rate": 6.895522388059702e-07, + "loss": 0.1015, + "step": 42362 + }, + { + "epoch": 7.7, + "learning_rate": 6.89497633782308e-07, + "loss": 0.6659, + "step": 42363 + }, + { + "epoch": 7.7, + "learning_rate": 6.894430287586458e-07, + "loss": 0.0107, + "step": 42364 + }, + { + "epoch": 7.7, + "learning_rate": 6.893884237349836e-07, + "loss": 0.0101, + "step": 42365 + }, + { + "epoch": 7.7, + "learning_rate": 6.893338187113215e-07, + "loss": 0.9662, + "step": 42366 + }, + { + "epoch": 7.7, + "learning_rate": 6.892792136876592e-07, + "loss": 0.0326, + "step": 42367 + }, + { + "epoch": 7.7, + "learning_rate": 6.892246086639971e-07, + "loss": 0.0099, + "step": 42368 + }, + { + "epoch": 7.7, + "learning_rate": 6.891700036403349e-07, + "loss": 1.2611, + "step": 42369 + }, + { + "epoch": 7.71, + "learning_rate": 6.891153986166728e-07, + "loss": 1.3089, + "step": 42370 + }, + { + "epoch": 7.71, + "learning_rate": 6.890607935930106e-07, + "loss": 0.2172, + "step": 42371 + }, + { + "epoch": 7.71, + "learning_rate": 6.890061885693484e-07, + "loss": 0.1303, + "step": 42372 + }, + { + "epoch": 7.71, + "learning_rate": 6.889515835456863e-07, + "loss": 0.0711, + "step": 42373 + }, + { + "epoch": 7.71, + "learning_rate": 6.88896978522024e-07, + "loss": 1.3203, + "step": 42374 + }, + { + "epoch": 7.71, + "learning_rate": 6.888423734983618e-07, + "loss": 0.0979, + "step": 42375 + }, + { + "epoch": 7.71, + "learning_rate": 6.887877684746997e-07, + "loss": 0.0093, + "step": 42376 + }, + { + "epoch": 7.71, + "learning_rate": 6.887331634510375e-07, + "loss": 1.8243, + "step": 42377 + }, + { + "epoch": 7.71, + "learning_rate": 6.886785584273754e-07, + "loss": 2.783, + "step": 42378 + }, + { + "epoch": 7.71, + "learning_rate": 6.886239534037132e-07, + "loss": 0.0105, + "step": 42379 + }, + { + "epoch": 7.71, + "learning_rate": 6.88569348380051e-07, + "loss": 0.2442, + "step": 42380 + }, + { + "epoch": 7.71, + "learning_rate": 6.885147433563888e-07, + "loss": 2.2079, + "step": 42381 + }, + { + "epoch": 7.71, + "learning_rate": 6.884601383327266e-07, + "loss": 0.0379, + "step": 42382 + }, + { + "epoch": 7.71, + "learning_rate": 6.884055333090644e-07, + "loss": 0.1255, + "step": 42383 + }, + { + "epoch": 7.71, + "learning_rate": 6.883509282854023e-07, + "loss": 0.1251, + "step": 42384 + }, + { + "epoch": 7.71, + "learning_rate": 6.882963232617401e-07, + "loss": 0.0523, + "step": 42385 + }, + { + "epoch": 7.71, + "learning_rate": 6.882417182380779e-07, + "loss": 0.6526, + "step": 42386 + }, + { + "epoch": 7.71, + "learning_rate": 6.881871132144158e-07, + "loss": 0.7481, + "step": 42387 + }, + { + "epoch": 7.71, + "learning_rate": 6.881325081907535e-07, + "loss": 0.1383, + "step": 42388 + }, + { + "epoch": 7.71, + "learning_rate": 6.880779031670914e-07, + "loss": 0.1508, + "step": 42389 + }, + { + "epoch": 7.71, + "learning_rate": 6.880232981434292e-07, + "loss": 0.0573, + "step": 42390 + }, + { + "epoch": 7.71, + "learning_rate": 6.87968693119767e-07, + "loss": 0.008, + "step": 42391 + }, + { + "epoch": 7.71, + "learning_rate": 6.879140880961049e-07, + "loss": 3.7191, + "step": 42392 + }, + { + "epoch": 7.71, + "learning_rate": 6.878594830724427e-07, + "loss": 1.075, + "step": 42393 + }, + { + "epoch": 7.71, + "learning_rate": 6.878048780487805e-07, + "loss": 0.22, + "step": 42394 + }, + { + "epoch": 7.71, + "learning_rate": 6.877502730251183e-07, + "loss": 0.1385, + "step": 42395 + }, + { + "epoch": 7.71, + "learning_rate": 6.876956680014562e-07, + "loss": 1.8649, + "step": 42396 + }, + { + "epoch": 7.71, + "learning_rate": 6.876410629777939e-07, + "loss": 2.0811, + "step": 42397 + }, + { + "epoch": 7.71, + "learning_rate": 6.875864579541318e-07, + "loss": 2.1117, + "step": 42398 + }, + { + "epoch": 7.71, + "learning_rate": 6.875318529304696e-07, + "loss": 0.3117, + "step": 42399 + }, + { + "epoch": 7.71, + "learning_rate": 6.874772479068075e-07, + "loss": 0.1363, + "step": 42400 + }, + { + "epoch": 7.71, + "learning_rate": 6.874226428831453e-07, + "loss": 0.0138, + "step": 42401 + }, + { + "epoch": 7.71, + "learning_rate": 6.87368037859483e-07, + "loss": 0.9457, + "step": 42402 + }, + { + "epoch": 7.71, + "learning_rate": 6.87313432835821e-07, + "loss": 0.056, + "step": 42403 + }, + { + "epoch": 7.71, + "learning_rate": 6.872588278121587e-07, + "loss": 0.0705, + "step": 42404 + }, + { + "epoch": 7.71, + "learning_rate": 6.872042227884965e-07, + "loss": 1.8242, + "step": 42405 + }, + { + "epoch": 7.71, + "learning_rate": 6.871496177648344e-07, + "loss": 0.0095, + "step": 42406 + }, + { + "epoch": 7.71, + "learning_rate": 6.870950127411722e-07, + "loss": 0.0082, + "step": 42407 + }, + { + "epoch": 7.71, + "learning_rate": 6.8704040771751e-07, + "loss": 0.0681, + "step": 42408 + }, + { + "epoch": 7.71, + "learning_rate": 6.869858026938478e-07, + "loss": 0.1072, + "step": 42409 + }, + { + "epoch": 7.71, + "learning_rate": 6.869311976701857e-07, + "loss": 0.012, + "step": 42410 + }, + { + "epoch": 7.71, + "learning_rate": 6.868765926465235e-07, + "loss": 1.5813, + "step": 42411 + }, + { + "epoch": 7.71, + "learning_rate": 6.868219876228613e-07, + "loss": 0.1793, + "step": 42412 + }, + { + "epoch": 7.71, + "learning_rate": 6.867673825991991e-07, + "loss": 0.7361, + "step": 42413 + }, + { + "epoch": 7.71, + "learning_rate": 6.86712777575537e-07, + "loss": 0.2444, + "step": 42414 + }, + { + "epoch": 7.71, + "learning_rate": 6.866581725518748e-07, + "loss": 0.0082, + "step": 42415 + }, + { + "epoch": 7.71, + "learning_rate": 6.866035675282126e-07, + "loss": 0.4903, + "step": 42416 + }, + { + "epoch": 7.71, + "learning_rate": 6.865489625045505e-07, + "loss": 0.6071, + "step": 42417 + }, + { + "epoch": 7.71, + "learning_rate": 6.864943574808882e-07, + "loss": 1.7908, + "step": 42418 + }, + { + "epoch": 7.71, + "learning_rate": 6.864397524572261e-07, + "loss": 3.1454, + "step": 42419 + }, + { + "epoch": 7.71, + "learning_rate": 6.863851474335639e-07, + "loss": 0.0123, + "step": 42420 + }, + { + "epoch": 7.71, + "learning_rate": 6.863305424099017e-07, + "loss": 0.76, + "step": 42421 + }, + { + "epoch": 7.71, + "learning_rate": 6.862759373862396e-07, + "loss": 1.3474, + "step": 42422 + }, + { + "epoch": 7.71, + "learning_rate": 6.862213323625774e-07, + "loss": 2.75, + "step": 42423 + }, + { + "epoch": 7.71, + "learning_rate": 6.861667273389152e-07, + "loss": 1.7615, + "step": 42424 + }, + { + "epoch": 7.72, + "learning_rate": 6.86112122315253e-07, + "loss": 0.3898, + "step": 42425 + }, + { + "epoch": 7.72, + "learning_rate": 6.860575172915908e-07, + "loss": 1.9274, + "step": 42426 + }, + { + "epoch": 7.72, + "learning_rate": 6.860029122679286e-07, + "loss": 0.2656, + "step": 42427 + }, + { + "epoch": 7.72, + "learning_rate": 6.859483072442665e-07, + "loss": 1.4091, + "step": 42428 + }, + { + "epoch": 7.72, + "learning_rate": 6.858937022206043e-07, + "loss": 1.0433, + "step": 42429 + }, + { + "epoch": 7.72, + "learning_rate": 6.858390971969422e-07, + "loss": 0.153, + "step": 42430 + }, + { + "epoch": 7.72, + "learning_rate": 6.8578449217328e-07, + "loss": 0.1759, + "step": 42431 + }, + { + "epoch": 7.72, + "learning_rate": 6.857298871496177e-07, + "loss": 3.0792, + "step": 42432 + }, + { + "epoch": 7.72, + "learning_rate": 6.856752821259557e-07, + "loss": 0.1892, + "step": 42433 + }, + { + "epoch": 7.72, + "learning_rate": 6.856206771022934e-07, + "loss": 0.0324, + "step": 42434 + }, + { + "epoch": 7.72, + "learning_rate": 6.855660720786312e-07, + "loss": 0.634, + "step": 42435 + }, + { + "epoch": 7.72, + "learning_rate": 6.855114670549691e-07, + "loss": 0.2729, + "step": 42436 + }, + { + "epoch": 7.72, + "learning_rate": 6.854568620313069e-07, + "loss": 0.2692, + "step": 42437 + }, + { + "epoch": 7.72, + "learning_rate": 6.854022570076448e-07, + "loss": 0.8327, + "step": 42438 + }, + { + "epoch": 7.72, + "learning_rate": 6.853476519839825e-07, + "loss": 0.0078, + "step": 42439 + }, + { + "epoch": 7.72, + "learning_rate": 6.852930469603204e-07, + "loss": 0.0054, + "step": 42440 + }, + { + "epoch": 7.72, + "learning_rate": 6.852384419366582e-07, + "loss": 0.6118, + "step": 42441 + }, + { + "epoch": 7.72, + "learning_rate": 6.85183836912996e-07, + "loss": 0.8604, + "step": 42442 + }, + { + "epoch": 7.72, + "learning_rate": 6.851292318893338e-07, + "loss": 0.819, + "step": 42443 + }, + { + "epoch": 7.72, + "learning_rate": 6.850746268656717e-07, + "loss": 0.0104, + "step": 42444 + }, + { + "epoch": 7.72, + "learning_rate": 6.850200218420095e-07, + "loss": 0.1279, + "step": 42445 + }, + { + "epoch": 7.72, + "learning_rate": 6.849654168183472e-07, + "loss": 3.4682, + "step": 42446 + }, + { + "epoch": 7.72, + "learning_rate": 6.849108117946852e-07, + "loss": 0.6031, + "step": 42447 + }, + { + "epoch": 7.72, + "learning_rate": 6.848562067710229e-07, + "loss": 0.0042, + "step": 42448 + }, + { + "epoch": 7.72, + "learning_rate": 6.848016017473608e-07, + "loss": 0.3998, + "step": 42449 + }, + { + "epoch": 7.72, + "learning_rate": 6.847469967236986e-07, + "loss": 0.052, + "step": 42450 + }, + { + "epoch": 7.72, + "learning_rate": 6.846923917000364e-07, + "loss": 0.1674, + "step": 42451 + }, + { + "epoch": 7.72, + "learning_rate": 6.846377866763743e-07, + "loss": 0.4938, + "step": 42452 + }, + { + "epoch": 7.72, + "learning_rate": 6.845831816527121e-07, + "loss": 0.075, + "step": 42453 + }, + { + "epoch": 7.72, + "learning_rate": 6.845285766290499e-07, + "loss": 0.0275, + "step": 42454 + }, + { + "epoch": 7.72, + "learning_rate": 6.844739716053877e-07, + "loss": 1.1848, + "step": 42455 + }, + { + "epoch": 7.72, + "learning_rate": 6.844193665817255e-07, + "loss": 0.0053, + "step": 42456 + }, + { + "epoch": 7.72, + "learning_rate": 6.843647615580633e-07, + "loss": 0.9719, + "step": 42457 + }, + { + "epoch": 7.72, + "learning_rate": 6.843101565344012e-07, + "loss": 2.3624, + "step": 42458 + }, + { + "epoch": 7.72, + "learning_rate": 6.84255551510739e-07, + "loss": 1.0829, + "step": 42459 + }, + { + "epoch": 7.72, + "learning_rate": 6.842009464870769e-07, + "loss": 0.4208, + "step": 42460 + }, + { + "epoch": 7.72, + "learning_rate": 6.841463414634147e-07, + "loss": 0.4192, + "step": 42461 + }, + { + "epoch": 7.72, + "learning_rate": 6.840917364397524e-07, + "loss": 0.1169, + "step": 42462 + }, + { + "epoch": 7.72, + "learning_rate": 6.840371314160903e-07, + "loss": 0.3133, + "step": 42463 + }, + { + "epoch": 7.72, + "learning_rate": 6.839825263924281e-07, + "loss": 1.213, + "step": 42464 + }, + { + "epoch": 7.72, + "learning_rate": 6.839279213687659e-07, + "loss": 2.0208, + "step": 42465 + }, + { + "epoch": 7.72, + "learning_rate": 6.838733163451038e-07, + "loss": 1.6335, + "step": 42466 + }, + { + "epoch": 7.72, + "learning_rate": 6.838187113214416e-07, + "loss": 0.0129, + "step": 42467 + }, + { + "epoch": 7.72, + "learning_rate": 6.837641062977795e-07, + "loss": 0.5937, + "step": 42468 + }, + { + "epoch": 7.72, + "learning_rate": 6.837095012741172e-07, + "loss": 0.0039, + "step": 42469 + }, + { + "epoch": 7.72, + "learning_rate": 6.83654896250455e-07, + "loss": 0.2655, + "step": 42470 + }, + { + "epoch": 7.72, + "learning_rate": 6.836002912267929e-07, + "loss": 1.0248, + "step": 42471 + }, + { + "epoch": 7.72, + "learning_rate": 6.835456862031307e-07, + "loss": 0.0041, + "step": 42472 + }, + { + "epoch": 7.72, + "learning_rate": 6.834910811794685e-07, + "loss": 2.3032, + "step": 42473 + }, + { + "epoch": 7.72, + "learning_rate": 6.834364761558064e-07, + "loss": 0.0093, + "step": 42474 + }, + { + "epoch": 7.72, + "learning_rate": 6.833818711321442e-07, + "loss": 0.3841, + "step": 42475 + }, + { + "epoch": 7.72, + "learning_rate": 6.833272661084819e-07, + "loss": 0.0501, + "step": 42476 + }, + { + "epoch": 7.72, + "learning_rate": 6.832726610848199e-07, + "loss": 1.4485, + "step": 42477 + }, + { + "epoch": 7.72, + "learning_rate": 6.832180560611576e-07, + "loss": 0.2545, + "step": 42478 + }, + { + "epoch": 7.72, + "learning_rate": 6.831634510374955e-07, + "loss": 0.1914, + "step": 42479 + }, + { + "epoch": 7.73, + "learning_rate": 6.831088460138333e-07, + "loss": 2.0576, + "step": 42480 + }, + { + "epoch": 7.73, + "learning_rate": 6.830542409901711e-07, + "loss": 0.023, + "step": 42481 + }, + { + "epoch": 7.73, + "learning_rate": 6.82999635966509e-07, + "loss": 0.0042, + "step": 42482 + }, + { + "epoch": 7.73, + "learning_rate": 6.829450309428467e-07, + "loss": 0.1918, + "step": 42483 + }, + { + "epoch": 7.73, + "learning_rate": 6.828904259191846e-07, + "loss": 0.4547, + "step": 42484 + }, + { + "epoch": 7.73, + "learning_rate": 6.828358208955224e-07, + "loss": 4.0988, + "step": 42485 + }, + { + "epoch": 7.73, + "learning_rate": 6.827812158718602e-07, + "loss": 0.1334, + "step": 42486 + }, + { + "epoch": 7.73, + "learning_rate": 6.82726610848198e-07, + "loss": 0.0209, + "step": 42487 + }, + { + "epoch": 7.73, + "learning_rate": 6.826720058245359e-07, + "loss": 0.1272, + "step": 42488 + }, + { + "epoch": 7.73, + "learning_rate": 6.826174008008737e-07, + "loss": 1.0324, + "step": 42489 + }, + { + "epoch": 7.73, + "learning_rate": 6.825627957772115e-07, + "loss": 0.3079, + "step": 42490 + }, + { + "epoch": 7.73, + "learning_rate": 6.825081907535494e-07, + "loss": 1.0212, + "step": 42491 + }, + { + "epoch": 7.73, + "learning_rate": 6.824535857298871e-07, + "loss": 1.1231, + "step": 42492 + }, + { + "epoch": 7.73, + "learning_rate": 6.82398980706225e-07, + "loss": 4.5534, + "step": 42493 + }, + { + "epoch": 7.73, + "learning_rate": 6.823443756825628e-07, + "loss": 0.0098, + "step": 42494 + }, + { + "epoch": 7.73, + "learning_rate": 6.822897706589006e-07, + "loss": 1.8698, + "step": 42495 + }, + { + "epoch": 7.73, + "learning_rate": 6.822351656352385e-07, + "loss": 0.5612, + "step": 42496 + }, + { + "epoch": 7.73, + "learning_rate": 6.821805606115763e-07, + "loss": 2.6417, + "step": 42497 + }, + { + "epoch": 7.73, + "learning_rate": 6.821259555879142e-07, + "loss": 2.5296, + "step": 42498 + }, + { + "epoch": 7.73, + "learning_rate": 6.820713505642519e-07, + "loss": 0.1487, + "step": 42499 + }, + { + "epoch": 7.73, + "learning_rate": 6.820167455405897e-07, + "loss": 0.7157, + "step": 42500 + }, + { + "epoch": 7.73, + "learning_rate": 6.819621405169276e-07, + "loss": 0.5313, + "step": 42501 + }, + { + "epoch": 7.73, + "learning_rate": 6.819075354932654e-07, + "loss": 0.4566, + "step": 42502 + }, + { + "epoch": 7.73, + "learning_rate": 6.818529304696032e-07, + "loss": 0.3276, + "step": 42503 + }, + { + "epoch": 7.73, + "learning_rate": 6.817983254459411e-07, + "loss": 0.0463, + "step": 42504 + }, + { + "epoch": 7.73, + "learning_rate": 6.817437204222789e-07, + "loss": 0.2886, + "step": 42505 + }, + { + "epoch": 7.73, + "learning_rate": 6.816891153986166e-07, + "loss": 2.8824, + "step": 42506 + }, + { + "epoch": 7.73, + "learning_rate": 6.816345103749545e-07, + "loss": 0.011, + "step": 42507 + }, + { + "epoch": 7.73, + "learning_rate": 6.815799053512923e-07, + "loss": 0.5699, + "step": 42508 + }, + { + "epoch": 7.73, + "learning_rate": 6.815253003276302e-07, + "loss": 0.5594, + "step": 42509 + }, + { + "epoch": 7.73, + "learning_rate": 6.81470695303968e-07, + "loss": 0.4628, + "step": 42510 + }, + { + "epoch": 7.73, + "learning_rate": 6.814160902803058e-07, + "loss": 1.7006, + "step": 42511 + }, + { + "epoch": 7.73, + "learning_rate": 6.813614852566437e-07, + "loss": 1.1695, + "step": 42512 + }, + { + "epoch": 7.73, + "learning_rate": 6.813068802329814e-07, + "loss": 0.0346, + "step": 42513 + }, + { + "epoch": 7.73, + "learning_rate": 6.812522752093192e-07, + "loss": 0.1583, + "step": 42514 + }, + { + "epoch": 7.73, + "learning_rate": 6.811976701856571e-07, + "loss": 0.192, + "step": 42515 + }, + { + "epoch": 7.73, + "learning_rate": 6.811430651619949e-07, + "loss": 0.0328, + "step": 42516 + }, + { + "epoch": 7.73, + "learning_rate": 6.810884601383328e-07, + "loss": 0.0083, + "step": 42517 + }, + { + "epoch": 7.73, + "learning_rate": 6.810338551146706e-07, + "loss": 1.3064, + "step": 42518 + }, + { + "epoch": 7.73, + "learning_rate": 6.809792500910084e-07, + "loss": 0.0435, + "step": 42519 + }, + { + "epoch": 7.73, + "learning_rate": 6.809246450673462e-07, + "loss": 0.3997, + "step": 42520 + }, + { + "epoch": 7.73, + "learning_rate": 6.808700400436841e-07, + "loss": 0.0443, + "step": 42521 + }, + { + "epoch": 7.73, + "learning_rate": 6.808154350200218e-07, + "loss": 0.6831, + "step": 42522 + }, + { + "epoch": 7.73, + "learning_rate": 6.807608299963597e-07, + "loss": 0.2669, + "step": 42523 + }, + { + "epoch": 7.73, + "learning_rate": 6.807062249726975e-07, + "loss": 2.2527, + "step": 42524 + }, + { + "epoch": 7.73, + "learning_rate": 6.806516199490353e-07, + "loss": 0.093, + "step": 42525 + }, + { + "epoch": 7.73, + "learning_rate": 6.805970149253732e-07, + "loss": 0.1616, + "step": 42526 + }, + { + "epoch": 7.73, + "learning_rate": 6.805424099017109e-07, + "loss": 0.0043, + "step": 42527 + }, + { + "epoch": 7.73, + "learning_rate": 6.804878048780489e-07, + "loss": 0.4079, + "step": 42528 + }, + { + "epoch": 7.73, + "learning_rate": 6.804331998543866e-07, + "loss": 1.0429, + "step": 42529 + }, + { + "epoch": 7.73, + "learning_rate": 6.803785948307244e-07, + "loss": 0.8058, + "step": 42530 + }, + { + "epoch": 7.73, + "learning_rate": 6.803239898070623e-07, + "loss": 0.2926, + "step": 42531 + }, + { + "epoch": 7.73, + "learning_rate": 6.802693847834001e-07, + "loss": 0.7832, + "step": 42532 + }, + { + "epoch": 7.73, + "learning_rate": 6.802147797597379e-07, + "loss": 0.7118, + "step": 42533 + }, + { + "epoch": 7.73, + "learning_rate": 6.801601747360757e-07, + "loss": 0.4579, + "step": 42534 + }, + { + "epoch": 7.74, + "learning_rate": 6.801055697124136e-07, + "loss": 0.5847, + "step": 42535 + }, + { + "epoch": 7.74, + "learning_rate": 6.800509646887513e-07, + "loss": 3.9926, + "step": 42536 + }, + { + "epoch": 7.74, + "learning_rate": 6.799963596650892e-07, + "loss": 0.2828, + "step": 42537 + }, + { + "epoch": 7.74, + "learning_rate": 6.79941754641427e-07, + "loss": 0.8732, + "step": 42538 + }, + { + "epoch": 7.74, + "learning_rate": 6.798871496177649e-07, + "loss": 0.1199, + "step": 42539 + }, + { + "epoch": 7.74, + "learning_rate": 6.798325445941027e-07, + "loss": 0.0135, + "step": 42540 + }, + { + "epoch": 7.74, + "learning_rate": 6.797779395704405e-07, + "loss": 0.2055, + "step": 42541 + }, + { + "epoch": 7.74, + "learning_rate": 6.797233345467784e-07, + "loss": 0.0064, + "step": 42542 + }, + { + "epoch": 7.74, + "learning_rate": 6.796687295231161e-07, + "loss": 0.5718, + "step": 42543 + }, + { + "epoch": 7.74, + "learning_rate": 6.796141244994539e-07, + "loss": 0.3151, + "step": 42544 + }, + { + "epoch": 7.74, + "learning_rate": 6.795595194757918e-07, + "loss": 0.704, + "step": 42545 + }, + { + "epoch": 7.74, + "learning_rate": 6.795049144521296e-07, + "loss": 0.0096, + "step": 42546 + }, + { + "epoch": 7.74, + "learning_rate": 6.794503094284675e-07, + "loss": 1.5934, + "step": 42547 + }, + { + "epoch": 7.74, + "learning_rate": 6.793957044048053e-07, + "loss": 0.0345, + "step": 42548 + }, + { + "epoch": 7.74, + "learning_rate": 6.793410993811431e-07, + "loss": 0.862, + "step": 42549 + }, + { + "epoch": 7.74, + "learning_rate": 6.792864943574809e-07, + "loss": 1.1586, + "step": 42550 + }, + { + "epoch": 7.74, + "learning_rate": 6.792318893338187e-07, + "loss": 2.777, + "step": 42551 + }, + { + "epoch": 7.74, + "learning_rate": 6.791772843101565e-07, + "loss": 0.0586, + "step": 42552 + }, + { + "epoch": 7.74, + "learning_rate": 6.791226792864944e-07, + "loss": 0.0659, + "step": 42553 + }, + { + "epoch": 7.74, + "learning_rate": 6.790680742628322e-07, + "loss": 0.6694, + "step": 42554 + }, + { + "epoch": 7.74, + "learning_rate": 6.7901346923917e-07, + "loss": 0.914, + "step": 42555 + }, + { + "epoch": 7.74, + "learning_rate": 6.789588642155079e-07, + "loss": 1.4846, + "step": 42556 + }, + { + "epoch": 7.74, + "learning_rate": 6.789042591918456e-07, + "loss": 0.6539, + "step": 42557 + }, + { + "epoch": 7.74, + "learning_rate": 6.788496541681835e-07, + "loss": 0.7684, + "step": 42558 + }, + { + "epoch": 7.74, + "learning_rate": 6.787950491445213e-07, + "loss": 0.0807, + "step": 42559 + }, + { + "epoch": 7.74, + "learning_rate": 6.787404441208591e-07, + "loss": 0.9684, + "step": 42560 + }, + { + "epoch": 7.74, + "learning_rate": 6.78685839097197e-07, + "loss": 0.4476, + "step": 42561 + }, + { + "epoch": 7.74, + "learning_rate": 6.786312340735348e-07, + "loss": 1.2817, + "step": 42562 + }, + { + "epoch": 7.74, + "learning_rate": 6.785766290498726e-07, + "loss": 0.1478, + "step": 42563 + }, + { + "epoch": 7.74, + "learning_rate": 6.785220240262104e-07, + "loss": 0.6292, + "step": 42564 + }, + { + "epoch": 7.74, + "learning_rate": 6.784674190025483e-07, + "loss": 0.6604, + "step": 42565 + }, + { + "epoch": 7.74, + "learning_rate": 6.78412813978886e-07, + "loss": 0.602, + "step": 42566 + }, + { + "epoch": 7.74, + "learning_rate": 6.783582089552239e-07, + "loss": 0.7716, + "step": 42567 + }, + { + "epoch": 7.74, + "learning_rate": 6.783036039315617e-07, + "loss": 0.1292, + "step": 42568 + }, + { + "epoch": 7.74, + "learning_rate": 6.782489989078996e-07, + "loss": 0.8373, + "step": 42569 + }, + { + "epoch": 7.74, + "learning_rate": 6.781943938842374e-07, + "loss": 0.0062, + "step": 42570 + }, + { + "epoch": 7.74, + "learning_rate": 6.781397888605751e-07, + "loss": 1.7979, + "step": 42571 + }, + { + "epoch": 7.74, + "learning_rate": 6.780851838369131e-07, + "loss": 1.0136, + "step": 42572 + }, + { + "epoch": 7.74, + "learning_rate": 6.780305788132508e-07, + "loss": 0.3271, + "step": 42573 + }, + { + "epoch": 7.74, + "learning_rate": 6.779759737895886e-07, + "loss": 0.24, + "step": 42574 + }, + { + "epoch": 7.74, + "learning_rate": 6.779213687659265e-07, + "loss": 0.3563, + "step": 42575 + }, + { + "epoch": 7.74, + "learning_rate": 6.778667637422643e-07, + "loss": 0.0648, + "step": 42576 + }, + { + "epoch": 7.74, + "learning_rate": 6.778121587186022e-07, + "loss": 0.256, + "step": 42577 + }, + { + "epoch": 7.74, + "learning_rate": 6.777575536949399e-07, + "loss": 0.7658, + "step": 42578 + }, + { + "epoch": 7.74, + "learning_rate": 6.777029486712778e-07, + "loss": 0.2446, + "step": 42579 + }, + { + "epoch": 7.74, + "learning_rate": 6.776483436476156e-07, + "loss": 0.2043, + "step": 42580 + }, + { + "epoch": 7.74, + "learning_rate": 6.775937386239534e-07, + "loss": 0.274, + "step": 42581 + }, + { + "epoch": 7.74, + "learning_rate": 6.775391336002912e-07, + "loss": 0.1055, + "step": 42582 + }, + { + "epoch": 7.74, + "learning_rate": 6.774845285766291e-07, + "loss": 0.2057, + "step": 42583 + }, + { + "epoch": 7.74, + "learning_rate": 6.774299235529669e-07, + "loss": 1.2611, + "step": 42584 + }, + { + "epoch": 7.74, + "learning_rate": 6.773753185293048e-07, + "loss": 1.6039, + "step": 42585 + }, + { + "epoch": 7.74, + "learning_rate": 6.773207135056426e-07, + "loss": 0.007, + "step": 42586 + }, + { + "epoch": 7.74, + "learning_rate": 6.772661084819803e-07, + "loss": 0.0287, + "step": 42587 + }, + { + "epoch": 7.74, + "learning_rate": 6.772115034583182e-07, + "loss": 2.8495, + "step": 42588 + }, + { + "epoch": 7.74, + "learning_rate": 6.77156898434656e-07, + "loss": 0.0091, + "step": 42589 + }, + { + "epoch": 7.75, + "learning_rate": 6.771022934109938e-07, + "loss": 0.0181, + "step": 42590 + }, + { + "epoch": 7.75, + "learning_rate": 6.770476883873317e-07, + "loss": 0.0108, + "step": 42591 + }, + { + "epoch": 7.75, + "learning_rate": 6.769930833636695e-07, + "loss": 1.3973, + "step": 42592 + }, + { + "epoch": 7.75, + "learning_rate": 6.769384783400073e-07, + "loss": 0.9216, + "step": 42593 + }, + { + "epoch": 7.75, + "learning_rate": 6.768838733163451e-07, + "loss": 0.0044, + "step": 42594 + }, + { + "epoch": 7.75, + "learning_rate": 6.768292682926829e-07, + "loss": 0.5185, + "step": 42595 + }, + { + "epoch": 7.75, + "learning_rate": 6.767746632690208e-07, + "loss": 0.0514, + "step": 42596 + }, + { + "epoch": 7.75, + "learning_rate": 6.767200582453586e-07, + "loss": 0.0185, + "step": 42597 + }, + { + "epoch": 7.75, + "learning_rate": 6.766654532216963e-07, + "loss": 0.3053, + "step": 42598 + }, + { + "epoch": 7.75, + "learning_rate": 6.766108481980343e-07, + "loss": 1.2013, + "step": 42599 + }, + { + "epoch": 7.75, + "learning_rate": 6.765562431743721e-07, + "loss": 0.4392, + "step": 42600 + }, + { + "epoch": 7.75, + "learning_rate": 6.765016381507098e-07, + "loss": 0.0142, + "step": 42601 + }, + { + "epoch": 7.75, + "learning_rate": 6.764470331270478e-07, + "loss": 0.0568, + "step": 42602 + }, + { + "epoch": 7.75, + "learning_rate": 6.763924281033855e-07, + "loss": 0.2808, + "step": 42603 + }, + { + "epoch": 7.75, + "learning_rate": 6.763378230797233e-07, + "loss": 0.0747, + "step": 42604 + }, + { + "epoch": 7.75, + "learning_rate": 6.762832180560612e-07, + "loss": 0.1571, + "step": 42605 + }, + { + "epoch": 7.75, + "learning_rate": 6.76228613032399e-07, + "loss": 0.571, + "step": 42606 + }, + { + "epoch": 7.75, + "learning_rate": 6.761740080087369e-07, + "loss": 0.049, + "step": 42607 + }, + { + "epoch": 7.75, + "learning_rate": 6.761194029850746e-07, + "loss": 0.8337, + "step": 42608 + }, + { + "epoch": 7.75, + "learning_rate": 6.760647979614125e-07, + "loss": 0.0148, + "step": 42609 + }, + { + "epoch": 7.75, + "learning_rate": 6.760101929377503e-07, + "loss": 1.3816, + "step": 42610 + }, + { + "epoch": 7.75, + "learning_rate": 6.759555879140881e-07, + "loss": 0.1196, + "step": 42611 + }, + { + "epoch": 7.75, + "learning_rate": 6.759009828904259e-07, + "loss": 0.1825, + "step": 42612 + }, + { + "epoch": 7.75, + "learning_rate": 6.758463778667638e-07, + "loss": 0.835, + "step": 42613 + }, + { + "epoch": 7.75, + "learning_rate": 6.757917728431016e-07, + "loss": 0.3093, + "step": 42614 + }, + { + "epoch": 7.75, + "learning_rate": 6.757371678194393e-07, + "loss": 0.8024, + "step": 42615 + }, + { + "epoch": 7.75, + "learning_rate": 6.756825627957773e-07, + "loss": 0.022, + "step": 42616 + }, + { + "epoch": 7.75, + "learning_rate": 6.75627957772115e-07, + "loss": 0.2788, + "step": 42617 + }, + { + "epoch": 7.75, + "learning_rate": 6.755733527484529e-07, + "loss": 0.0724, + "step": 42618 + }, + { + "epoch": 7.75, + "learning_rate": 6.755187477247907e-07, + "loss": 0.0053, + "step": 42619 + }, + { + "epoch": 7.75, + "learning_rate": 6.754641427011285e-07, + "loss": 0.016, + "step": 42620 + }, + { + "epoch": 7.75, + "learning_rate": 6.754095376774664e-07, + "loss": 0.0551, + "step": 42621 + }, + { + "epoch": 7.75, + "learning_rate": 6.753549326538042e-07, + "loss": 2.2476, + "step": 42622 + }, + { + "epoch": 7.75, + "learning_rate": 6.75300327630142e-07, + "loss": 0.5995, + "step": 42623 + }, + { + "epoch": 7.75, + "learning_rate": 6.752457226064798e-07, + "loss": 1.1011, + "step": 42624 + }, + { + "epoch": 7.75, + "learning_rate": 6.751911175828176e-07, + "loss": 0.2055, + "step": 42625 + }, + { + "epoch": 7.75, + "learning_rate": 6.751365125591555e-07, + "loss": 1.7253, + "step": 42626 + }, + { + "epoch": 7.75, + "learning_rate": 6.750819075354933e-07, + "loss": 0.0212, + "step": 42627 + }, + { + "epoch": 7.75, + "learning_rate": 6.750273025118311e-07, + "loss": 2.464, + "step": 42628 + }, + { + "epoch": 7.75, + "learning_rate": 6.74972697488169e-07, + "loss": 0.214, + "step": 42629 + }, + { + "epoch": 7.75, + "learning_rate": 6.749180924645068e-07, + "loss": 0.1931, + "step": 42630 + }, + { + "epoch": 7.75, + "learning_rate": 6.748634874408445e-07, + "loss": 1.7124, + "step": 42631 + }, + { + "epoch": 7.75, + "learning_rate": 6.748088824171824e-07, + "loss": 0.9442, + "step": 42632 + }, + { + "epoch": 7.75, + "learning_rate": 6.747542773935202e-07, + "loss": 0.2699, + "step": 42633 + }, + { + "epoch": 7.75, + "learning_rate": 6.74699672369858e-07, + "loss": 0.5067, + "step": 42634 + }, + { + "epoch": 7.75, + "learning_rate": 6.746450673461958e-07, + "loss": 0.4359, + "step": 42635 + }, + { + "epoch": 7.75, + "learning_rate": 6.745904623225337e-07, + "loss": 0.9915, + "step": 42636 + }, + { + "epoch": 7.75, + "learning_rate": 6.745358572988716e-07, + "loss": 0.1046, + "step": 42637 + }, + { + "epoch": 7.75, + "learning_rate": 6.744812522752093e-07, + "loss": 0.1762, + "step": 42638 + }, + { + "epoch": 7.75, + "learning_rate": 6.744266472515471e-07, + "loss": 0.0857, + "step": 42639 + }, + { + "epoch": 7.75, + "learning_rate": 6.74372042227885e-07, + "loss": 0.3089, + "step": 42640 + }, + { + "epoch": 7.75, + "learning_rate": 6.743174372042228e-07, + "loss": 0.4466, + "step": 42641 + }, + { + "epoch": 7.75, + "learning_rate": 6.742628321805605e-07, + "loss": 0.5411, + "step": 42642 + }, + { + "epoch": 7.75, + "learning_rate": 6.742082271568985e-07, + "loss": 0.0101, + "step": 42643 + }, + { + "epoch": 7.75, + "learning_rate": 6.741536221332363e-07, + "loss": 0.9216, + "step": 42644 + }, + { + "epoch": 7.76, + "learning_rate": 6.74099017109574e-07, + "loss": 0.0067, + "step": 42645 + }, + { + "epoch": 7.76, + "learning_rate": 6.74044412085912e-07, + "loss": 0.4405, + "step": 42646 + }, + { + "epoch": 7.76, + "learning_rate": 6.739898070622497e-07, + "loss": 0.0056, + "step": 42647 + }, + { + "epoch": 7.76, + "learning_rate": 6.739352020385876e-07, + "loss": 1.1502, + "step": 42648 + }, + { + "epoch": 7.76, + "learning_rate": 6.738805970149254e-07, + "loss": 0.7703, + "step": 42649 + }, + { + "epoch": 7.76, + "learning_rate": 6.738259919912632e-07, + "loss": 1.7827, + "step": 42650 + }, + { + "epoch": 7.76, + "learning_rate": 6.737713869676011e-07, + "loss": 0.0054, + "step": 42651 + }, + { + "epoch": 7.76, + "learning_rate": 6.737167819439388e-07, + "loss": 0.046, + "step": 42652 + }, + { + "epoch": 7.76, + "learning_rate": 6.736621769202768e-07, + "loss": 0.9862, + "step": 42653 + }, + { + "epoch": 7.76, + "learning_rate": 6.736075718966145e-07, + "loss": 0.449, + "step": 42654 + }, + { + "epoch": 7.76, + "learning_rate": 6.735529668729523e-07, + "loss": 0.065, + "step": 42655 + }, + { + "epoch": 7.76, + "learning_rate": 6.734983618492902e-07, + "loss": 0.0452, + "step": 42656 + }, + { + "epoch": 7.76, + "learning_rate": 6.73443756825628e-07, + "loss": 1.2205, + "step": 42657 + }, + { + "epoch": 7.76, + "learning_rate": 6.733891518019658e-07, + "loss": 0.0562, + "step": 42658 + }, + { + "epoch": 7.76, + "learning_rate": 6.733345467783036e-07, + "loss": 0.0658, + "step": 42659 + }, + { + "epoch": 7.76, + "learning_rate": 6.732799417546415e-07, + "loss": 1.0217, + "step": 42660 + }, + { + "epoch": 7.76, + "learning_rate": 6.732253367309792e-07, + "loss": 0.0077, + "step": 42661 + }, + { + "epoch": 7.76, + "learning_rate": 6.731707317073171e-07, + "loss": 0.8336, + "step": 42662 + }, + { + "epoch": 7.76, + "learning_rate": 6.731161266836549e-07, + "loss": 0.2689, + "step": 42663 + }, + { + "epoch": 7.76, + "learning_rate": 6.730615216599928e-07, + "loss": 0.0413, + "step": 42664 + }, + { + "epoch": 7.76, + "learning_rate": 6.730069166363306e-07, + "loss": 0.006, + "step": 42665 + }, + { + "epoch": 7.76, + "learning_rate": 6.729523116126684e-07, + "loss": 1.2196, + "step": 42666 + }, + { + "epoch": 7.76, + "learning_rate": 6.728977065890063e-07, + "loss": 0.2224, + "step": 42667 + }, + { + "epoch": 7.76, + "learning_rate": 6.72843101565344e-07, + "loss": 0.4893, + "step": 42668 + }, + { + "epoch": 7.76, + "learning_rate": 6.727884965416818e-07, + "loss": 0.2613, + "step": 42669 + }, + { + "epoch": 7.76, + "learning_rate": 6.727338915180197e-07, + "loss": 0.0951, + "step": 42670 + }, + { + "epoch": 7.76, + "learning_rate": 6.726792864943575e-07, + "loss": 0.1928, + "step": 42671 + }, + { + "epoch": 7.76, + "learning_rate": 6.726246814706952e-07, + "loss": 0.0415, + "step": 42672 + }, + { + "epoch": 7.76, + "learning_rate": 6.725700764470332e-07, + "loss": 0.0624, + "step": 42673 + }, + { + "epoch": 7.76, + "learning_rate": 6.72515471423371e-07, + "loss": 0.0128, + "step": 42674 + }, + { + "epoch": 7.76, + "learning_rate": 6.724608663997088e-07, + "loss": 1.0102, + "step": 42675 + }, + { + "epoch": 7.76, + "learning_rate": 6.724062613760466e-07, + "loss": 0.2593, + "step": 42676 + }, + { + "epoch": 7.76, + "learning_rate": 6.723516563523844e-07, + "loss": 0.4521, + "step": 42677 + }, + { + "epoch": 7.76, + "learning_rate": 6.722970513287223e-07, + "loss": 0.0142, + "step": 42678 + }, + { + "epoch": 7.76, + "learning_rate": 6.7224244630506e-07, + "loss": 0.0056, + "step": 42679 + }, + { + "epoch": 7.76, + "learning_rate": 6.721878412813979e-07, + "loss": 0.375, + "step": 42680 + }, + { + "epoch": 7.76, + "learning_rate": 6.721332362577358e-07, + "loss": 0.5301, + "step": 42681 + }, + { + "epoch": 7.76, + "learning_rate": 6.720786312340735e-07, + "loss": 3.152, + "step": 42682 + }, + { + "epoch": 7.76, + "learning_rate": 6.720240262104113e-07, + "loss": 0.008, + "step": 42683 + }, + { + "epoch": 7.76, + "learning_rate": 6.719694211867492e-07, + "loss": 0.4811, + "step": 42684 + }, + { + "epoch": 7.76, + "learning_rate": 6.71914816163087e-07, + "loss": 1.4809, + "step": 42685 + }, + { + "epoch": 7.76, + "learning_rate": 6.718602111394249e-07, + "loss": 0.0196, + "step": 42686 + }, + { + "epoch": 7.76, + "learning_rate": 6.718056061157627e-07, + "loss": 0.6495, + "step": 42687 + }, + { + "epoch": 7.76, + "learning_rate": 6.717510010921005e-07, + "loss": 1.885, + "step": 42688 + }, + { + "epoch": 7.76, + "learning_rate": 6.716963960684383e-07, + "loss": 0.7925, + "step": 42689 + }, + { + "epoch": 7.76, + "learning_rate": 6.716417910447762e-07, + "loss": 0.196, + "step": 42690 + }, + { + "epoch": 7.76, + "learning_rate": 6.715871860211139e-07, + "loss": 0.0603, + "step": 42691 + }, + { + "epoch": 7.76, + "learning_rate": 6.715325809974518e-07, + "loss": 0.244, + "step": 42692 + }, + { + "epoch": 7.76, + "learning_rate": 6.714779759737896e-07, + "loss": 1.5733, + "step": 42693 + }, + { + "epoch": 7.76, + "learning_rate": 6.714233709501275e-07, + "loss": 0.1335, + "step": 42694 + }, + { + "epoch": 7.76, + "learning_rate": 6.713687659264653e-07, + "loss": 1.2731, + "step": 42695 + }, + { + "epoch": 7.76, + "learning_rate": 6.71314160902803e-07, + "loss": 0.0767, + "step": 42696 + }, + { + "epoch": 7.76, + "learning_rate": 6.71259555879141e-07, + "loss": 0.0697, + "step": 42697 + }, + { + "epoch": 7.76, + "learning_rate": 6.712049508554787e-07, + "loss": 0.4815, + "step": 42698 + }, + { + "epoch": 7.76, + "learning_rate": 6.711503458318165e-07, + "loss": 3.8704, + "step": 42699 + }, + { + "epoch": 7.77, + "learning_rate": 6.710957408081544e-07, + "loss": 1.172, + "step": 42700 + }, + { + "epoch": 7.77, + "learning_rate": 6.710411357844922e-07, + "loss": 0.7268, + "step": 42701 + }, + { + "epoch": 7.77, + "learning_rate": 6.7098653076083e-07, + "loss": 1.9789, + "step": 42702 + }, + { + "epoch": 7.77, + "learning_rate": 6.709319257371678e-07, + "loss": 0.4406, + "step": 42703 + }, + { + "epoch": 7.77, + "learning_rate": 6.708773207135057e-07, + "loss": 0.1516, + "step": 42704 + }, + { + "epoch": 7.77, + "learning_rate": 6.708227156898435e-07, + "loss": 2.531, + "step": 42705 + }, + { + "epoch": 7.77, + "learning_rate": 6.707681106661813e-07, + "loss": 0.3044, + "step": 42706 + }, + { + "epoch": 7.77, + "learning_rate": 6.707135056425191e-07, + "loss": 2.5934, + "step": 42707 + }, + { + "epoch": 7.77, + "learning_rate": 6.70658900618857e-07, + "loss": 0.61, + "step": 42708 + }, + { + "epoch": 7.77, + "learning_rate": 6.706042955951948e-07, + "loss": 0.0061, + "step": 42709 + }, + { + "epoch": 7.77, + "learning_rate": 6.705496905715326e-07, + "loss": 0.0066, + "step": 42710 + }, + { + "epoch": 7.77, + "learning_rate": 6.704950855478705e-07, + "loss": 1.1873, + "step": 42711 + }, + { + "epoch": 7.77, + "learning_rate": 6.704404805242082e-07, + "loss": 1.9621, + "step": 42712 + }, + { + "epoch": 7.77, + "learning_rate": 6.70385875500546e-07, + "loss": 0.0751, + "step": 42713 + }, + { + "epoch": 7.77, + "learning_rate": 6.703312704768839e-07, + "loss": 0.0692, + "step": 42714 + }, + { + "epoch": 7.77, + "learning_rate": 6.702766654532217e-07, + "loss": 1.1976, + "step": 42715 + }, + { + "epoch": 7.77, + "learning_rate": 6.702220604295595e-07, + "loss": 0.1326, + "step": 42716 + }, + { + "epoch": 7.77, + "learning_rate": 6.701674554058974e-07, + "loss": 1.7497, + "step": 42717 + }, + { + "epoch": 7.77, + "learning_rate": 6.701128503822352e-07, + "loss": 2.0424, + "step": 42718 + }, + { + "epoch": 7.77, + "learning_rate": 6.70058245358573e-07, + "loss": 0.0127, + "step": 42719 + }, + { + "epoch": 7.77, + "learning_rate": 6.700036403349108e-07, + "loss": 0.0804, + "step": 42720 + }, + { + "epoch": 7.77, + "learning_rate": 6.699490353112486e-07, + "loss": 1.07, + "step": 42721 + }, + { + "epoch": 7.77, + "learning_rate": 6.698944302875865e-07, + "loss": 0.0085, + "step": 42722 + }, + { + "epoch": 7.77, + "learning_rate": 6.698398252639242e-07, + "loss": 0.1465, + "step": 42723 + }, + { + "epoch": 7.77, + "learning_rate": 6.697852202402622e-07, + "loss": 0.0034, + "step": 42724 + }, + { + "epoch": 7.77, + "learning_rate": 6.697306152166e-07, + "loss": 1.4856, + "step": 42725 + }, + { + "epoch": 7.77, + "learning_rate": 6.696760101929377e-07, + "loss": 0.9703, + "step": 42726 + }, + { + "epoch": 7.77, + "learning_rate": 6.696214051692756e-07, + "loss": 0.3563, + "step": 42727 + }, + { + "epoch": 7.77, + "learning_rate": 6.695668001456134e-07, + "loss": 0.4675, + "step": 42728 + }, + { + "epoch": 7.77, + "learning_rate": 6.695121951219512e-07, + "loss": 0.7542, + "step": 42729 + }, + { + "epoch": 7.77, + "learning_rate": 6.694575900982891e-07, + "loss": 0.1747, + "step": 42730 + }, + { + "epoch": 7.77, + "learning_rate": 6.694029850746269e-07, + "loss": 1.2371, + "step": 42731 + }, + { + "epoch": 7.77, + "learning_rate": 6.693483800509647e-07, + "loss": 0.5676, + "step": 42732 + }, + { + "epoch": 7.77, + "learning_rate": 6.692937750273025e-07, + "loss": 0.5491, + "step": 42733 + }, + { + "epoch": 7.77, + "learning_rate": 6.692391700036404e-07, + "loss": 0.006, + "step": 42734 + }, + { + "epoch": 7.77, + "learning_rate": 6.691845649799782e-07, + "loss": 0.0448, + "step": 42735 + }, + { + "epoch": 7.77, + "learning_rate": 6.69129959956316e-07, + "loss": 0.0791, + "step": 42736 + }, + { + "epoch": 7.77, + "learning_rate": 6.690753549326538e-07, + "loss": 1.1407, + "step": 42737 + }, + { + "epoch": 7.77, + "learning_rate": 6.690207499089917e-07, + "loss": 0.4359, + "step": 42738 + }, + { + "epoch": 7.77, + "learning_rate": 6.689661448853295e-07, + "loss": 0.0511, + "step": 42739 + }, + { + "epoch": 7.77, + "learning_rate": 6.689115398616672e-07, + "loss": 1.0875, + "step": 42740 + }, + { + "epoch": 7.77, + "learning_rate": 6.688569348380052e-07, + "loss": 1.3426, + "step": 42741 + }, + { + "epoch": 7.77, + "learning_rate": 6.688023298143429e-07, + "loss": 0.8772, + "step": 42742 + }, + { + "epoch": 7.77, + "learning_rate": 6.687477247906808e-07, + "loss": 0.0051, + "step": 42743 + }, + { + "epoch": 7.77, + "learning_rate": 6.686931197670186e-07, + "loss": 0.0264, + "step": 42744 + }, + { + "epoch": 7.77, + "learning_rate": 6.686385147433564e-07, + "loss": 0.0189, + "step": 42745 + }, + { + "epoch": 7.77, + "learning_rate": 6.685839097196943e-07, + "loss": 0.0058, + "step": 42746 + }, + { + "epoch": 7.77, + "learning_rate": 6.68529304696032e-07, + "loss": 0.5404, + "step": 42747 + }, + { + "epoch": 7.77, + "learning_rate": 6.684746996723699e-07, + "loss": 1.1929, + "step": 42748 + }, + { + "epoch": 7.77, + "learning_rate": 6.684200946487077e-07, + "loss": 0.1379, + "step": 42749 + }, + { + "epoch": 7.77, + "learning_rate": 6.683654896250455e-07, + "loss": 0.9648, + "step": 42750 + }, + { + "epoch": 7.77, + "learning_rate": 6.683108846013833e-07, + "loss": 0.5838, + "step": 42751 + }, + { + "epoch": 7.77, + "learning_rate": 6.682562795777212e-07, + "loss": 4.4545, + "step": 42752 + }, + { + "epoch": 7.77, + "learning_rate": 6.682016745540589e-07, + "loss": 0.9708, + "step": 42753 + }, + { + "epoch": 7.77, + "learning_rate": 6.681470695303969e-07, + "loss": 1.3838, + "step": 42754 + }, + { + "epoch": 7.78, + "learning_rate": 6.680924645067347e-07, + "loss": 0.2223, + "step": 42755 + }, + { + "epoch": 7.78, + "learning_rate": 6.680378594830724e-07, + "loss": 0.1473, + "step": 42756 + }, + { + "epoch": 7.78, + "learning_rate": 6.679832544594103e-07, + "loss": 2.7749, + "step": 42757 + }, + { + "epoch": 7.78, + "learning_rate": 6.679286494357481e-07, + "loss": 0.0742, + "step": 42758 + }, + { + "epoch": 7.78, + "learning_rate": 6.678740444120859e-07, + "loss": 0.5872, + "step": 42759 + }, + { + "epoch": 7.78, + "learning_rate": 6.678194393884237e-07, + "loss": 0.0587, + "step": 42760 + }, + { + "epoch": 7.78, + "learning_rate": 6.677648343647616e-07, + "loss": 0.1547, + "step": 42761 + }, + { + "epoch": 7.78, + "learning_rate": 6.677102293410995e-07, + "loss": 0.163, + "step": 42762 + }, + { + "epoch": 7.78, + "learning_rate": 6.676556243174372e-07, + "loss": 0.3398, + "step": 42763 + }, + { + "epoch": 7.78, + "learning_rate": 6.67601019293775e-07, + "loss": 1.3752, + "step": 42764 + }, + { + "epoch": 7.78, + "learning_rate": 6.675464142701129e-07, + "loss": 1.5199, + "step": 42765 + }, + { + "epoch": 7.78, + "learning_rate": 6.674918092464507e-07, + "loss": 0.715, + "step": 42766 + }, + { + "epoch": 7.78, + "learning_rate": 6.674372042227884e-07, + "loss": 0.0082, + "step": 42767 + }, + { + "epoch": 7.78, + "learning_rate": 6.673825991991264e-07, + "loss": 0.7035, + "step": 42768 + }, + { + "epoch": 7.78, + "learning_rate": 6.673279941754642e-07, + "loss": 0.4853, + "step": 42769 + }, + { + "epoch": 7.78, + "learning_rate": 6.672733891518019e-07, + "loss": 0.7635, + "step": 42770 + }, + { + "epoch": 7.78, + "learning_rate": 6.672187841281399e-07, + "loss": 1.3403, + "step": 42771 + }, + { + "epoch": 7.78, + "learning_rate": 6.671641791044776e-07, + "loss": 0.072, + "step": 42772 + }, + { + "epoch": 7.78, + "learning_rate": 6.671095740808155e-07, + "loss": 0.0664, + "step": 42773 + }, + { + "epoch": 7.78, + "learning_rate": 6.670549690571533e-07, + "loss": 1.142, + "step": 42774 + }, + { + "epoch": 7.78, + "learning_rate": 6.670003640334911e-07, + "loss": 0.5843, + "step": 42775 + }, + { + "epoch": 7.78, + "learning_rate": 6.66945759009829e-07, + "loss": 0.0042, + "step": 42776 + }, + { + "epoch": 7.78, + "learning_rate": 6.668911539861667e-07, + "loss": 1.3253, + "step": 42777 + }, + { + "epoch": 7.78, + "learning_rate": 6.668365489625046e-07, + "loss": 0.5956, + "step": 42778 + }, + { + "epoch": 7.78, + "learning_rate": 6.667819439388424e-07, + "loss": 0.1814, + "step": 42779 + }, + { + "epoch": 7.78, + "learning_rate": 6.667273389151802e-07, + "loss": 0.4846, + "step": 42780 + }, + { + "epoch": 7.78, + "learning_rate": 6.66672733891518e-07, + "loss": 0.4426, + "step": 42781 + }, + { + "epoch": 7.78, + "learning_rate": 6.666181288678559e-07, + "loss": 0.0616, + "step": 42782 + }, + { + "epoch": 7.78, + "learning_rate": 6.665635238441937e-07, + "loss": 0.1692, + "step": 42783 + }, + { + "epoch": 7.78, + "learning_rate": 6.665089188205315e-07, + "loss": 0.1648, + "step": 42784 + }, + { + "epoch": 7.78, + "learning_rate": 6.664543137968694e-07, + "loss": 0.0224, + "step": 42785 + }, + { + "epoch": 7.78, + "learning_rate": 6.663997087732071e-07, + "loss": 0.2463, + "step": 42786 + }, + { + "epoch": 7.78, + "learning_rate": 6.66345103749545e-07, + "loss": 0.2332, + "step": 42787 + }, + { + "epoch": 7.78, + "learning_rate": 6.662904987258828e-07, + "loss": 1.7787, + "step": 42788 + }, + { + "epoch": 7.78, + "learning_rate": 6.662358937022206e-07, + "loss": 0.1461, + "step": 42789 + }, + { + "epoch": 7.78, + "learning_rate": 6.661812886785584e-07, + "loss": 0.4824, + "step": 42790 + }, + { + "epoch": 7.78, + "learning_rate": 6.661266836548962e-07, + "loss": 1.0525, + "step": 42791 + }, + { + "epoch": 7.78, + "learning_rate": 6.660720786312342e-07, + "loss": 3.6529, + "step": 42792 + }, + { + "epoch": 7.78, + "learning_rate": 6.660174736075719e-07, + "loss": 0.0991, + "step": 42793 + }, + { + "epoch": 7.78, + "learning_rate": 6.659628685839097e-07, + "loss": 0.9675, + "step": 42794 + }, + { + "epoch": 7.78, + "learning_rate": 6.659082635602476e-07, + "loss": 0.0047, + "step": 42795 + }, + { + "epoch": 7.78, + "learning_rate": 6.658536585365854e-07, + "loss": 1.5538, + "step": 42796 + }, + { + "epoch": 7.78, + "learning_rate": 6.657990535129231e-07, + "loss": 1.2691, + "step": 42797 + }, + { + "epoch": 7.78, + "learning_rate": 6.657444484892611e-07, + "loss": 0.124, + "step": 42798 + }, + { + "epoch": 7.78, + "learning_rate": 6.656898434655989e-07, + "loss": 1.5611, + "step": 42799 + }, + { + "epoch": 7.78, + "learning_rate": 6.656352384419366e-07, + "loss": 0.0086, + "step": 42800 + }, + { + "epoch": 7.78, + "learning_rate": 6.655806334182745e-07, + "loss": 0.0603, + "step": 42801 + }, + { + "epoch": 7.78, + "learning_rate": 6.655260283946123e-07, + "loss": 0.1948, + "step": 42802 + }, + { + "epoch": 7.78, + "learning_rate": 6.654714233709502e-07, + "loss": 0.0097, + "step": 42803 + }, + { + "epoch": 7.78, + "learning_rate": 6.654168183472879e-07, + "loss": 0.8826, + "step": 42804 + }, + { + "epoch": 7.78, + "learning_rate": 6.653622133236258e-07, + "loss": 0.6019, + "step": 42805 + }, + { + "epoch": 7.78, + "learning_rate": 6.653076082999637e-07, + "loss": 0.0662, + "step": 42806 + }, + { + "epoch": 7.78, + "learning_rate": 6.652530032763014e-07, + "loss": 0.1332, + "step": 42807 + }, + { + "epoch": 7.78, + "learning_rate": 6.651983982526392e-07, + "loss": 0.848, + "step": 42808 + }, + { + "epoch": 7.78, + "learning_rate": 6.651437932289771e-07, + "loss": 0.18, + "step": 42809 + }, + { + "epoch": 7.79, + "learning_rate": 6.650891882053149e-07, + "loss": 0.0062, + "step": 42810 + }, + { + "epoch": 7.79, + "learning_rate": 6.650345831816526e-07, + "loss": 1.1731, + "step": 42811 + }, + { + "epoch": 7.79, + "learning_rate": 6.649799781579906e-07, + "loss": 2.0353, + "step": 42812 + }, + { + "epoch": 7.79, + "learning_rate": 6.649253731343284e-07, + "loss": 3.1615, + "step": 42813 + }, + { + "epoch": 7.79, + "learning_rate": 6.648707681106662e-07, + "loss": 0.2202, + "step": 42814 + }, + { + "epoch": 7.79, + "learning_rate": 6.648161630870041e-07, + "loss": 0.4298, + "step": 42815 + }, + { + "epoch": 7.79, + "learning_rate": 6.647615580633418e-07, + "loss": 0.2801, + "step": 42816 + }, + { + "epoch": 7.79, + "learning_rate": 6.647069530396797e-07, + "loss": 0.4419, + "step": 42817 + }, + { + "epoch": 7.79, + "learning_rate": 6.646523480160175e-07, + "loss": 0.0338, + "step": 42818 + }, + { + "epoch": 7.79, + "learning_rate": 6.645977429923553e-07, + "loss": 0.0172, + "step": 42819 + }, + { + "epoch": 7.79, + "learning_rate": 6.645431379686932e-07, + "loss": 0.151, + "step": 42820 + }, + { + "epoch": 7.79, + "learning_rate": 6.644885329450309e-07, + "loss": 0.0222, + "step": 42821 + }, + { + "epoch": 7.79, + "learning_rate": 6.644339279213689e-07, + "loss": 1.576, + "step": 42822 + }, + { + "epoch": 7.79, + "learning_rate": 6.643793228977066e-07, + "loss": 0.2848, + "step": 42823 + }, + { + "epoch": 7.79, + "learning_rate": 6.643247178740444e-07, + "loss": 0.0915, + "step": 42824 + }, + { + "epoch": 7.79, + "learning_rate": 6.642701128503823e-07, + "loss": 1.0672, + "step": 42825 + }, + { + "epoch": 7.79, + "learning_rate": 6.642155078267201e-07, + "loss": 1.4498, + "step": 42826 + }, + { + "epoch": 7.79, + "learning_rate": 6.641609028030578e-07, + "loss": 0.0705, + "step": 42827 + }, + { + "epoch": 7.79, + "learning_rate": 6.641062977793957e-07, + "loss": 0.2458, + "step": 42828 + }, + { + "epoch": 7.79, + "learning_rate": 6.640516927557336e-07, + "loss": 0.0131, + "step": 42829 + }, + { + "epoch": 7.79, + "learning_rate": 6.639970877320713e-07, + "loss": 0.2787, + "step": 42830 + }, + { + "epoch": 7.79, + "learning_rate": 6.639424827084092e-07, + "loss": 0.0036, + "step": 42831 + }, + { + "epoch": 7.79, + "learning_rate": 6.63887877684747e-07, + "loss": 0.0547, + "step": 42832 + }, + { + "epoch": 7.79, + "learning_rate": 6.638332726610849e-07, + "loss": 0.0047, + "step": 42833 + }, + { + "epoch": 7.79, + "learning_rate": 6.637786676374226e-07, + "loss": 0.7974, + "step": 42834 + }, + { + "epoch": 7.79, + "learning_rate": 6.637240626137605e-07, + "loss": 0.1497, + "step": 42835 + }, + { + "epoch": 7.79, + "learning_rate": 6.636694575900984e-07, + "loss": 3.2694, + "step": 42836 + }, + { + "epoch": 7.79, + "learning_rate": 6.636148525664361e-07, + "loss": 0.8279, + "step": 42837 + }, + { + "epoch": 7.79, + "learning_rate": 6.635602475427739e-07, + "loss": 0.6307, + "step": 42838 + }, + { + "epoch": 7.79, + "learning_rate": 6.635056425191118e-07, + "loss": 0.1567, + "step": 42839 + }, + { + "epoch": 7.79, + "learning_rate": 6.634510374954496e-07, + "loss": 0.4977, + "step": 42840 + }, + { + "epoch": 7.79, + "learning_rate": 6.633964324717873e-07, + "loss": 0.1483, + "step": 42841 + }, + { + "epoch": 7.79, + "learning_rate": 6.633418274481253e-07, + "loss": 0.0135, + "step": 42842 + }, + { + "epoch": 7.79, + "learning_rate": 6.632872224244631e-07, + "loss": 1.4787, + "step": 42843 + }, + { + "epoch": 7.79, + "learning_rate": 6.632326174008009e-07, + "loss": 0.0158, + "step": 42844 + }, + { + "epoch": 7.79, + "learning_rate": 6.631780123771387e-07, + "loss": 1.7731, + "step": 42845 + }, + { + "epoch": 7.79, + "learning_rate": 6.631234073534765e-07, + "loss": 0.3417, + "step": 42846 + }, + { + "epoch": 7.79, + "learning_rate": 6.630688023298144e-07, + "loss": 0.1314, + "step": 42847 + }, + { + "epoch": 7.79, + "learning_rate": 6.630141973061521e-07, + "loss": 0.014, + "step": 42848 + }, + { + "epoch": 7.79, + "learning_rate": 6.6295959228249e-07, + "loss": 0.1072, + "step": 42849 + }, + { + "epoch": 7.79, + "learning_rate": 6.629049872588279e-07, + "loss": 0.6903, + "step": 42850 + }, + { + "epoch": 7.79, + "learning_rate": 6.628503822351656e-07, + "loss": 2.5724, + "step": 42851 + }, + { + "epoch": 7.79, + "learning_rate": 6.627957772115035e-07, + "loss": 0.3406, + "step": 42852 + }, + { + "epoch": 7.79, + "learning_rate": 6.627411721878413e-07, + "loss": 1.2657, + "step": 42853 + }, + { + "epoch": 7.79, + "learning_rate": 6.626865671641791e-07, + "loss": 0.0675, + "step": 42854 + }, + { + "epoch": 7.79, + "learning_rate": 6.62631962140517e-07, + "loss": 0.6013, + "step": 42855 + }, + { + "epoch": 7.79, + "learning_rate": 6.625773571168548e-07, + "loss": 0.1529, + "step": 42856 + }, + { + "epoch": 7.79, + "learning_rate": 6.625227520931926e-07, + "loss": 0.8522, + "step": 42857 + }, + { + "epoch": 7.79, + "learning_rate": 6.624681470695304e-07, + "loss": 0.3413, + "step": 42858 + }, + { + "epoch": 7.79, + "learning_rate": 6.624135420458683e-07, + "loss": 0.0623, + "step": 42859 + }, + { + "epoch": 7.79, + "learning_rate": 6.62358937022206e-07, + "loss": 0.1368, + "step": 42860 + }, + { + "epoch": 7.79, + "learning_rate": 6.623043319985439e-07, + "loss": 3.02, + "step": 42861 + }, + { + "epoch": 7.79, + "learning_rate": 6.622497269748817e-07, + "loss": 1.0531, + "step": 42862 + }, + { + "epoch": 7.79, + "learning_rate": 6.621951219512196e-07, + "loss": 0.1791, + "step": 42863 + }, + { + "epoch": 7.79, + "learning_rate": 6.621405169275574e-07, + "loss": 0.1436, + "step": 42864 + }, + { + "epoch": 7.8, + "learning_rate": 6.620859119038951e-07, + "loss": 0.2855, + "step": 42865 + }, + { + "epoch": 7.8, + "learning_rate": 6.620313068802331e-07, + "loss": 0.0113, + "step": 42866 + }, + { + "epoch": 7.8, + "learning_rate": 6.619767018565708e-07, + "loss": 0.0478, + "step": 42867 + }, + { + "epoch": 7.8, + "learning_rate": 6.619220968329086e-07, + "loss": 0.1691, + "step": 42868 + }, + { + "epoch": 7.8, + "learning_rate": 6.618674918092465e-07, + "loss": 0.4658, + "step": 42869 + }, + { + "epoch": 7.8, + "learning_rate": 6.618128867855843e-07, + "loss": 2.4228, + "step": 42870 + }, + { + "epoch": 7.8, + "learning_rate": 6.61758281761922e-07, + "loss": 0.3437, + "step": 42871 + }, + { + "epoch": 7.8, + "learning_rate": 6.617036767382599e-07, + "loss": 0.3309, + "step": 42872 + }, + { + "epoch": 7.8, + "learning_rate": 6.616490717145978e-07, + "loss": 1.3296, + "step": 42873 + }, + { + "epoch": 7.8, + "learning_rate": 6.615944666909356e-07, + "loss": 0.423, + "step": 42874 + }, + { + "epoch": 7.8, + "learning_rate": 6.615398616672734e-07, + "loss": 1.5908, + "step": 42875 + }, + { + "epoch": 7.8, + "learning_rate": 6.614852566436112e-07, + "loss": 0.0188, + "step": 42876 + }, + { + "epoch": 7.8, + "learning_rate": 6.614306516199491e-07, + "loss": 0.0222, + "step": 42877 + }, + { + "epoch": 7.8, + "learning_rate": 6.613760465962868e-07, + "loss": 1.7926, + "step": 42878 + }, + { + "epoch": 7.8, + "learning_rate": 6.613214415726247e-07, + "loss": 0.0763, + "step": 42879 + }, + { + "epoch": 7.8, + "learning_rate": 6.612668365489626e-07, + "loss": 0.302, + "step": 42880 + }, + { + "epoch": 7.8, + "learning_rate": 6.612122315253003e-07, + "loss": 1.3176, + "step": 42881 + }, + { + "epoch": 7.8, + "learning_rate": 6.611576265016382e-07, + "loss": 0.9966, + "step": 42882 + }, + { + "epoch": 7.8, + "learning_rate": 6.61103021477976e-07, + "loss": 0.1976, + "step": 42883 + }, + { + "epoch": 7.8, + "learning_rate": 6.610484164543138e-07, + "loss": 0.9842, + "step": 42884 + }, + { + "epoch": 7.8, + "learning_rate": 6.609938114306516e-07, + "loss": 0.0117, + "step": 42885 + }, + { + "epoch": 7.8, + "learning_rate": 6.609392064069895e-07, + "loss": 0.0097, + "step": 42886 + }, + { + "epoch": 7.8, + "learning_rate": 6.608846013833273e-07, + "loss": 0.5818, + "step": 42887 + }, + { + "epoch": 7.8, + "learning_rate": 6.608299963596651e-07, + "loss": 0.0338, + "step": 42888 + }, + { + "epoch": 7.8, + "learning_rate": 6.607753913360029e-07, + "loss": 0.9382, + "step": 42889 + }, + { + "epoch": 7.8, + "learning_rate": 6.607207863123407e-07, + "loss": 0.0053, + "step": 42890 + }, + { + "epoch": 7.8, + "learning_rate": 6.606661812886786e-07, + "loss": 0.0498, + "step": 42891 + }, + { + "epoch": 7.8, + "learning_rate": 6.606115762650163e-07, + "loss": 1.5477, + "step": 42892 + }, + { + "epoch": 7.8, + "learning_rate": 6.605569712413543e-07, + "loss": 2.6922, + "step": 42893 + }, + { + "epoch": 7.8, + "learning_rate": 6.605023662176921e-07, + "loss": 0.0208, + "step": 42894 + }, + { + "epoch": 7.8, + "learning_rate": 6.604477611940298e-07, + "loss": 0.3247, + "step": 42895 + }, + { + "epoch": 7.8, + "learning_rate": 6.603931561703677e-07, + "loss": 0.0415, + "step": 42896 + }, + { + "epoch": 7.8, + "learning_rate": 6.603385511467055e-07, + "loss": 0.0095, + "step": 42897 + }, + { + "epoch": 7.8, + "learning_rate": 6.602839461230433e-07, + "loss": 1.1052, + "step": 42898 + }, + { + "epoch": 7.8, + "learning_rate": 6.602293410993812e-07, + "loss": 0.7232, + "step": 42899 + }, + { + "epoch": 7.8, + "learning_rate": 6.60174736075719e-07, + "loss": 0.0089, + "step": 42900 + }, + { + "epoch": 7.8, + "learning_rate": 6.601201310520569e-07, + "loss": 0.0107, + "step": 42901 + }, + { + "epoch": 7.8, + "learning_rate": 6.600655260283946e-07, + "loss": 0.3948, + "step": 42902 + }, + { + "epoch": 7.8, + "learning_rate": 6.600109210047325e-07, + "loss": 0.2916, + "step": 42903 + }, + { + "epoch": 7.8, + "learning_rate": 6.599563159810703e-07, + "loss": 1.7705, + "step": 42904 + }, + { + "epoch": 7.8, + "learning_rate": 6.599017109574081e-07, + "loss": 0.2354, + "step": 42905 + }, + { + "epoch": 7.8, + "learning_rate": 6.598471059337459e-07, + "loss": 1.3187, + "step": 42906 + }, + { + "epoch": 7.8, + "learning_rate": 6.597925009100838e-07, + "loss": 1.2322, + "step": 42907 + }, + { + "epoch": 7.8, + "learning_rate": 6.597378958864215e-07, + "loss": 0.0512, + "step": 42908 + }, + { + "epoch": 7.8, + "learning_rate": 6.596832908627593e-07, + "loss": 0.0894, + "step": 42909 + }, + { + "epoch": 7.8, + "learning_rate": 6.596286858390973e-07, + "loss": 1.1321, + "step": 42910 + }, + { + "epoch": 7.8, + "learning_rate": 6.59574080815435e-07, + "loss": 0.0127, + "step": 42911 + }, + { + "epoch": 7.8, + "learning_rate": 6.595194757917729e-07, + "loss": 0.1461, + "step": 42912 + }, + { + "epoch": 7.8, + "learning_rate": 6.594648707681107e-07, + "loss": 0.1029, + "step": 42913 + }, + { + "epoch": 7.8, + "learning_rate": 6.594102657444485e-07, + "loss": 0.4564, + "step": 42914 + }, + { + "epoch": 7.8, + "learning_rate": 6.593556607207863e-07, + "loss": 0.2546, + "step": 42915 + }, + { + "epoch": 7.8, + "learning_rate": 6.593010556971241e-07, + "loss": 0.0126, + "step": 42916 + }, + { + "epoch": 7.8, + "learning_rate": 6.59246450673462e-07, + "loss": 1.0048, + "step": 42917 + }, + { + "epoch": 7.8, + "learning_rate": 6.591918456497998e-07, + "loss": 0.0162, + "step": 42918 + }, + { + "epoch": 7.8, + "learning_rate": 6.591372406261376e-07, + "loss": 0.8364, + "step": 42919 + }, + { + "epoch": 7.81, + "learning_rate": 6.590826356024754e-07, + "loss": 0.0569, + "step": 42920 + }, + { + "epoch": 7.81, + "learning_rate": 6.590280305788133e-07, + "loss": 0.0913, + "step": 42921 + }, + { + "epoch": 7.81, + "learning_rate": 6.58973425555151e-07, + "loss": 3.9312, + "step": 42922 + }, + { + "epoch": 7.81, + "learning_rate": 6.58918820531489e-07, + "loss": 0.0112, + "step": 42923 + }, + { + "epoch": 7.81, + "learning_rate": 6.588642155078268e-07, + "loss": 1.3075, + "step": 42924 + }, + { + "epoch": 7.81, + "learning_rate": 6.588096104841645e-07, + "loss": 0.3288, + "step": 42925 + }, + { + "epoch": 7.81, + "learning_rate": 6.587550054605024e-07, + "loss": 0.1126, + "step": 42926 + }, + { + "epoch": 7.81, + "learning_rate": 6.587004004368402e-07, + "loss": 0.3041, + "step": 42927 + }, + { + "epoch": 7.81, + "learning_rate": 6.58645795413178e-07, + "loss": 3.9509, + "step": 42928 + }, + { + "epoch": 7.81, + "learning_rate": 6.585911903895158e-07, + "loss": 1.0535, + "step": 42929 + }, + { + "epoch": 7.81, + "learning_rate": 6.585365853658537e-07, + "loss": 1.0249, + "step": 42930 + }, + { + "epoch": 7.81, + "learning_rate": 6.584819803421916e-07, + "loss": 0.5134, + "step": 42931 + }, + { + "epoch": 7.81, + "learning_rate": 6.584273753185293e-07, + "loss": 0.0821, + "step": 42932 + }, + { + "epoch": 7.81, + "learning_rate": 6.583727702948671e-07, + "loss": 0.2078, + "step": 42933 + }, + { + "epoch": 7.81, + "learning_rate": 6.58318165271205e-07, + "loss": 1.3789, + "step": 42934 + }, + { + "epoch": 7.81, + "learning_rate": 6.582635602475428e-07, + "loss": 1.3802, + "step": 42935 + }, + { + "epoch": 7.81, + "learning_rate": 6.582089552238805e-07, + "loss": 0.4022, + "step": 42936 + }, + { + "epoch": 7.81, + "learning_rate": 6.581543502002185e-07, + "loss": 0.6378, + "step": 42937 + }, + { + "epoch": 7.81, + "learning_rate": 6.580997451765563e-07, + "loss": 1.7643, + "step": 42938 + }, + { + "epoch": 7.81, + "learning_rate": 6.58045140152894e-07, + "loss": 0.5975, + "step": 42939 + }, + { + "epoch": 7.81, + "learning_rate": 6.579905351292319e-07, + "loss": 0.0468, + "step": 42940 + }, + { + "epoch": 7.81, + "learning_rate": 6.579359301055697e-07, + "loss": 0.0068, + "step": 42941 + }, + { + "epoch": 7.81, + "learning_rate": 6.578813250819076e-07, + "loss": 0.0062, + "step": 42942 + }, + { + "epoch": 7.81, + "learning_rate": 6.578267200582454e-07, + "loss": 0.0861, + "step": 42943 + }, + { + "epoch": 7.81, + "learning_rate": 6.577721150345832e-07, + "loss": 1.9243, + "step": 42944 + }, + { + "epoch": 7.81, + "learning_rate": 6.57717510010921e-07, + "loss": 0.0083, + "step": 42945 + }, + { + "epoch": 7.81, + "learning_rate": 6.576629049872588e-07, + "loss": 0.0796, + "step": 42946 + }, + { + "epoch": 7.81, + "learning_rate": 6.576082999635967e-07, + "loss": 0.0208, + "step": 42947 + }, + { + "epoch": 7.81, + "learning_rate": 6.575536949399345e-07, + "loss": 0.2169, + "step": 42948 + }, + { + "epoch": 7.81, + "learning_rate": 6.574990899162723e-07, + "loss": 0.0212, + "step": 42949 + }, + { + "epoch": 7.81, + "learning_rate": 6.574444848926102e-07, + "loss": 0.0988, + "step": 42950 + }, + { + "epoch": 7.81, + "learning_rate": 6.57389879868948e-07, + "loss": 0.0228, + "step": 42951 + }, + { + "epoch": 7.81, + "learning_rate": 6.573352748452857e-07, + "loss": 1.7486, + "step": 42952 + }, + { + "epoch": 7.81, + "learning_rate": 6.572806698216236e-07, + "loss": 0.1712, + "step": 42953 + }, + { + "epoch": 7.81, + "learning_rate": 6.572260647979615e-07, + "loss": 0.8159, + "step": 42954 + }, + { + "epoch": 7.81, + "learning_rate": 6.571714597742992e-07, + "loss": 0.6906, + "step": 42955 + }, + { + "epoch": 7.81, + "learning_rate": 6.571168547506371e-07, + "loss": 0.2988, + "step": 42956 + }, + { + "epoch": 7.81, + "learning_rate": 6.570622497269749e-07, + "loss": 0.2408, + "step": 42957 + }, + { + "epoch": 7.81, + "learning_rate": 6.570076447033127e-07, + "loss": 1.6395, + "step": 42958 + }, + { + "epoch": 7.81, + "learning_rate": 6.569530396796505e-07, + "loss": 0.0046, + "step": 42959 + }, + { + "epoch": 7.81, + "learning_rate": 6.568984346559883e-07, + "loss": 0.0379, + "step": 42960 + }, + { + "epoch": 7.81, + "learning_rate": 6.568438296323263e-07, + "loss": 0.027, + "step": 42961 + }, + { + "epoch": 7.81, + "learning_rate": 6.56789224608664e-07, + "loss": 0.0263, + "step": 42962 + }, + { + "epoch": 7.81, + "learning_rate": 6.567346195850018e-07, + "loss": 0.5021, + "step": 42963 + }, + { + "epoch": 7.81, + "learning_rate": 6.566800145613397e-07, + "loss": 0.4289, + "step": 42964 + }, + { + "epoch": 7.81, + "learning_rate": 6.566254095376775e-07, + "loss": 0.9506, + "step": 42965 + }, + { + "epoch": 7.81, + "learning_rate": 6.565708045140152e-07, + "loss": 0.0155, + "step": 42966 + }, + { + "epoch": 7.81, + "learning_rate": 6.565161994903532e-07, + "loss": 0.954, + "step": 42967 + }, + { + "epoch": 7.81, + "learning_rate": 6.56461594466691e-07, + "loss": 0.8878, + "step": 42968 + }, + { + "epoch": 7.81, + "learning_rate": 6.564069894430287e-07, + "loss": 0.3346, + "step": 42969 + }, + { + "epoch": 7.81, + "learning_rate": 6.563523844193666e-07, + "loss": 0.0409, + "step": 42970 + }, + { + "epoch": 7.81, + "learning_rate": 6.562977793957044e-07, + "loss": 0.9082, + "step": 42971 + }, + { + "epoch": 7.81, + "learning_rate": 6.562431743720423e-07, + "loss": 0.4787, + "step": 42972 + }, + { + "epoch": 7.81, + "learning_rate": 6.5618856934838e-07, + "loss": 0.0025, + "step": 42973 + }, + { + "epoch": 7.81, + "learning_rate": 6.561339643247179e-07, + "loss": 0.3215, + "step": 42974 + }, + { + "epoch": 7.82, + "learning_rate": 6.560793593010558e-07, + "loss": 0.1558, + "step": 42975 + }, + { + "epoch": 7.82, + "learning_rate": 6.560247542773935e-07, + "loss": 0.8778, + "step": 42976 + }, + { + "epoch": 7.82, + "learning_rate": 6.559701492537313e-07, + "loss": 1.5709, + "step": 42977 + }, + { + "epoch": 7.82, + "learning_rate": 6.559155442300692e-07, + "loss": 0.2418, + "step": 42978 + }, + { + "epoch": 7.82, + "learning_rate": 6.55860939206407e-07, + "loss": 0.0352, + "step": 42979 + }, + { + "epoch": 7.82, + "learning_rate": 6.558063341827447e-07, + "loss": 0.3604, + "step": 42980 + }, + { + "epoch": 7.82, + "learning_rate": 6.557517291590827e-07, + "loss": 0.0415, + "step": 42981 + }, + { + "epoch": 7.82, + "learning_rate": 6.556971241354204e-07, + "loss": 0.0928, + "step": 42982 + }, + { + "epoch": 7.82, + "learning_rate": 6.556425191117583e-07, + "loss": 0.8561, + "step": 42983 + }, + { + "epoch": 7.82, + "learning_rate": 6.555879140880962e-07, + "loss": 0.0287, + "step": 42984 + }, + { + "epoch": 7.82, + "learning_rate": 6.555333090644339e-07, + "loss": 1.18, + "step": 42985 + }, + { + "epoch": 7.82, + "learning_rate": 6.554787040407718e-07, + "loss": 1.1694, + "step": 42986 + }, + { + "epoch": 7.82, + "learning_rate": 6.554240990171096e-07, + "loss": 0.3373, + "step": 42987 + }, + { + "epoch": 7.82, + "learning_rate": 6.553694939934474e-07, + "loss": 0.1369, + "step": 42988 + }, + { + "epoch": 7.82, + "learning_rate": 6.553148889697852e-07, + "loss": 0.4327, + "step": 42989 + }, + { + "epoch": 7.82, + "learning_rate": 6.55260283946123e-07, + "loss": 0.0834, + "step": 42990 + }, + { + "epoch": 7.82, + "learning_rate": 6.55205678922461e-07, + "loss": 0.426, + "step": 42991 + }, + { + "epoch": 7.82, + "learning_rate": 6.551510738987987e-07, + "loss": 0.0047, + "step": 42992 + }, + { + "epoch": 7.82, + "learning_rate": 6.550964688751365e-07, + "loss": 0.6529, + "step": 42993 + }, + { + "epoch": 7.82, + "learning_rate": 6.550418638514744e-07, + "loss": 0.8583, + "step": 42994 + }, + { + "epoch": 7.82, + "learning_rate": 6.549872588278122e-07, + "loss": 0.3125, + "step": 42995 + }, + { + "epoch": 7.82, + "learning_rate": 6.549326538041499e-07, + "loss": 1.8187, + "step": 42996 + }, + { + "epoch": 7.82, + "learning_rate": 6.548780487804878e-07, + "loss": 0.1849, + "step": 42997 + }, + { + "epoch": 7.82, + "learning_rate": 6.548234437568257e-07, + "loss": 2.6282, + "step": 42998 + }, + { + "epoch": 7.82, + "learning_rate": 6.547688387331634e-07, + "loss": 1.4449, + "step": 42999 + }, + { + "epoch": 7.82, + "learning_rate": 6.547142337095013e-07, + "loss": 0.8818, + "step": 43000 + }, + { + "epoch": 7.82, + "learning_rate": 6.546596286858391e-07, + "loss": 0.3867, + "step": 43001 + }, + { + "epoch": 7.82, + "learning_rate": 6.54605023662177e-07, + "loss": 0.0068, + "step": 43002 + }, + { + "epoch": 7.82, + "learning_rate": 6.545504186385147e-07, + "loss": 0.1333, + "step": 43003 + }, + { + "epoch": 7.82, + "learning_rate": 6.544958136148526e-07, + "loss": 0.4652, + "step": 43004 + }, + { + "epoch": 7.82, + "learning_rate": 6.544412085911905e-07, + "loss": 0.5199, + "step": 43005 + }, + { + "epoch": 7.82, + "learning_rate": 6.543866035675282e-07, + "loss": 0.5213, + "step": 43006 + }, + { + "epoch": 7.82, + "learning_rate": 6.54331998543866e-07, + "loss": 0.396, + "step": 43007 + }, + { + "epoch": 7.82, + "learning_rate": 6.542773935202039e-07, + "loss": 0.1846, + "step": 43008 + }, + { + "epoch": 7.82, + "learning_rate": 6.542227884965417e-07, + "loss": 0.1306, + "step": 43009 + }, + { + "epoch": 7.82, + "learning_rate": 6.541681834728794e-07, + "loss": 0.2151, + "step": 43010 + }, + { + "epoch": 7.82, + "learning_rate": 6.541135784492174e-07, + "loss": 1.1361, + "step": 43011 + }, + { + "epoch": 7.82, + "learning_rate": 6.540589734255552e-07, + "loss": 0.0595, + "step": 43012 + }, + { + "epoch": 7.82, + "learning_rate": 6.54004368401893e-07, + "loss": 0.8404, + "step": 43013 + }, + { + "epoch": 7.82, + "learning_rate": 6.539497633782308e-07, + "loss": 0.0116, + "step": 43014 + }, + { + "epoch": 7.82, + "learning_rate": 6.538951583545686e-07, + "loss": 0.121, + "step": 43015 + }, + { + "epoch": 7.82, + "learning_rate": 6.538405533309065e-07, + "loss": 0.2611, + "step": 43016 + }, + { + "epoch": 7.82, + "learning_rate": 6.537859483072442e-07, + "loss": 0.0039, + "step": 43017 + }, + { + "epoch": 7.82, + "learning_rate": 6.537313432835822e-07, + "loss": 0.0907, + "step": 43018 + }, + { + "epoch": 7.82, + "learning_rate": 6.5367673825992e-07, + "loss": 0.0072, + "step": 43019 + }, + { + "epoch": 7.82, + "learning_rate": 6.536221332362577e-07, + "loss": 0.2802, + "step": 43020 + }, + { + "epoch": 7.82, + "learning_rate": 6.535675282125956e-07, + "loss": 0.0373, + "step": 43021 + }, + { + "epoch": 7.82, + "learning_rate": 6.535129231889334e-07, + "loss": 0.0413, + "step": 43022 + }, + { + "epoch": 7.82, + "learning_rate": 6.534583181652712e-07, + "loss": 0.0042, + "step": 43023 + }, + { + "epoch": 7.82, + "learning_rate": 6.53403713141609e-07, + "loss": 0.0644, + "step": 43024 + }, + { + "epoch": 7.82, + "learning_rate": 6.533491081179469e-07, + "loss": 0.4642, + "step": 43025 + }, + { + "epoch": 7.82, + "learning_rate": 6.532945030942846e-07, + "loss": 0.0068, + "step": 43026 + }, + { + "epoch": 7.82, + "learning_rate": 6.532398980706225e-07, + "loss": 1.2081, + "step": 43027 + }, + { + "epoch": 7.82, + "learning_rate": 6.531852930469604e-07, + "loss": 0.401, + "step": 43028 + }, + { + "epoch": 7.82, + "learning_rate": 6.531306880232982e-07, + "loss": 0.0083, + "step": 43029 + }, + { + "epoch": 7.83, + "learning_rate": 6.53076082999636e-07, + "loss": 0.2197, + "step": 43030 + }, + { + "epoch": 7.83, + "learning_rate": 6.530214779759738e-07, + "loss": 0.0985, + "step": 43031 + }, + { + "epoch": 7.83, + "learning_rate": 6.529668729523117e-07, + "loss": 0.5042, + "step": 43032 + }, + { + "epoch": 7.83, + "learning_rate": 6.529122679286494e-07, + "loss": 0.1872, + "step": 43033 + }, + { + "epoch": 7.83, + "learning_rate": 6.528576629049872e-07, + "loss": 0.0683, + "step": 43034 + }, + { + "epoch": 7.83, + "learning_rate": 6.528030578813252e-07, + "loss": 0.3258, + "step": 43035 + }, + { + "epoch": 7.83, + "learning_rate": 6.527484528576629e-07, + "loss": 0.6944, + "step": 43036 + }, + { + "epoch": 7.83, + "learning_rate": 6.526938478340007e-07, + "loss": 1.2624, + "step": 43037 + }, + { + "epoch": 7.83, + "learning_rate": 6.526392428103386e-07, + "loss": 0.6518, + "step": 43038 + }, + { + "epoch": 7.83, + "learning_rate": 6.525846377866764e-07, + "loss": 1.3831, + "step": 43039 + }, + { + "epoch": 7.83, + "learning_rate": 6.525300327630142e-07, + "loss": 0.4833, + "step": 43040 + }, + { + "epoch": 7.83, + "learning_rate": 6.52475427739352e-07, + "loss": 0.0034, + "step": 43041 + }, + { + "epoch": 7.83, + "learning_rate": 6.524208227156899e-07, + "loss": 0.0562, + "step": 43042 + }, + { + "epoch": 7.83, + "learning_rate": 6.523662176920277e-07, + "loss": 1.1136, + "step": 43043 + }, + { + "epoch": 7.83, + "learning_rate": 6.523116126683655e-07, + "loss": 0.038, + "step": 43044 + }, + { + "epoch": 7.83, + "learning_rate": 6.522570076447033e-07, + "loss": 1.5528, + "step": 43045 + }, + { + "epoch": 7.83, + "learning_rate": 6.522024026210412e-07, + "loss": 0.4495, + "step": 43046 + }, + { + "epoch": 7.83, + "learning_rate": 6.521477975973789e-07, + "loss": 1.6439, + "step": 43047 + }, + { + "epoch": 7.83, + "learning_rate": 6.520931925737169e-07, + "loss": 0.0081, + "step": 43048 + }, + { + "epoch": 7.83, + "learning_rate": 6.520385875500547e-07, + "loss": 0.6636, + "step": 43049 + }, + { + "epoch": 7.83, + "learning_rate": 6.519839825263924e-07, + "loss": 1.2976, + "step": 43050 + }, + { + "epoch": 7.83, + "learning_rate": 6.519293775027303e-07, + "loss": 0.3453, + "step": 43051 + }, + { + "epoch": 7.83, + "learning_rate": 6.518747724790681e-07, + "loss": 1.8331, + "step": 43052 + }, + { + "epoch": 7.83, + "learning_rate": 6.518201674554059e-07, + "loss": 0.2868, + "step": 43053 + }, + { + "epoch": 7.83, + "learning_rate": 6.517655624317437e-07, + "loss": 0.592, + "step": 43054 + }, + { + "epoch": 7.83, + "learning_rate": 6.517109574080816e-07, + "loss": 0.0429, + "step": 43055 + }, + { + "epoch": 7.83, + "learning_rate": 6.516563523844194e-07, + "loss": 1.5884, + "step": 43056 + }, + { + "epoch": 7.83, + "learning_rate": 6.516017473607572e-07, + "loss": 0.1007, + "step": 43057 + }, + { + "epoch": 7.83, + "learning_rate": 6.51547142337095e-07, + "loss": 0.9365, + "step": 43058 + }, + { + "epoch": 7.83, + "learning_rate": 6.514925373134329e-07, + "loss": 0.0131, + "step": 43059 + }, + { + "epoch": 7.83, + "learning_rate": 6.514379322897707e-07, + "loss": 0.0203, + "step": 43060 + }, + { + "epoch": 7.83, + "learning_rate": 6.513833272661084e-07, + "loss": 0.0078, + "step": 43061 + }, + { + "epoch": 7.83, + "learning_rate": 6.513287222424464e-07, + "loss": 0.0064, + "step": 43062 + }, + { + "epoch": 7.83, + "learning_rate": 6.512741172187841e-07, + "loss": 0.0049, + "step": 43063 + }, + { + "epoch": 7.83, + "learning_rate": 6.512195121951219e-07, + "loss": 0.8959, + "step": 43064 + }, + { + "epoch": 7.83, + "learning_rate": 6.511649071714598e-07, + "loss": 0.052, + "step": 43065 + }, + { + "epoch": 7.83, + "learning_rate": 6.511103021477976e-07, + "loss": 0.1789, + "step": 43066 + }, + { + "epoch": 7.83, + "learning_rate": 6.510556971241354e-07, + "loss": 0.0557, + "step": 43067 + }, + { + "epoch": 7.83, + "learning_rate": 6.510010921004733e-07, + "loss": 0.4728, + "step": 43068 + }, + { + "epoch": 7.83, + "learning_rate": 6.509464870768111e-07, + "loss": 0.264, + "step": 43069 + }, + { + "epoch": 7.83, + "learning_rate": 6.508918820531489e-07, + "loss": 1.3414, + "step": 43070 + }, + { + "epoch": 7.83, + "learning_rate": 6.508372770294867e-07, + "loss": 0.5565, + "step": 43071 + }, + { + "epoch": 7.83, + "learning_rate": 6.507826720058246e-07, + "loss": 0.0238, + "step": 43072 + }, + { + "epoch": 7.83, + "learning_rate": 6.507280669821624e-07, + "loss": 1.9859, + "step": 43073 + }, + { + "epoch": 7.83, + "learning_rate": 6.506734619585002e-07, + "loss": 0.1911, + "step": 43074 + }, + { + "epoch": 7.83, + "learning_rate": 6.50618856934838e-07, + "loss": 0.0049, + "step": 43075 + }, + { + "epoch": 7.83, + "learning_rate": 6.505642519111759e-07, + "loss": 0.5549, + "step": 43076 + }, + { + "epoch": 7.83, + "learning_rate": 6.505096468875136e-07, + "loss": 1.3217, + "step": 43077 + }, + { + "epoch": 7.83, + "learning_rate": 6.504550418638514e-07, + "loss": 0.1077, + "step": 43078 + }, + { + "epoch": 7.83, + "learning_rate": 6.504004368401894e-07, + "loss": 0.2983, + "step": 43079 + }, + { + "epoch": 7.83, + "learning_rate": 6.503458318165271e-07, + "loss": 0.0075, + "step": 43080 + }, + { + "epoch": 7.83, + "learning_rate": 6.50291226792865e-07, + "loss": 3.6838, + "step": 43081 + }, + { + "epoch": 7.83, + "learning_rate": 6.502366217692028e-07, + "loss": 1.3418, + "step": 43082 + }, + { + "epoch": 7.83, + "learning_rate": 6.501820167455406e-07, + "loss": 0.194, + "step": 43083 + }, + { + "epoch": 7.83, + "learning_rate": 6.501274117218784e-07, + "loss": 1.0913, + "step": 43084 + }, + { + "epoch": 7.84, + "learning_rate": 6.500728066982162e-07, + "loss": 0.6737, + "step": 43085 + }, + { + "epoch": 7.84, + "learning_rate": 6.500182016745541e-07, + "loss": 0.1153, + "step": 43086 + }, + { + "epoch": 7.84, + "learning_rate": 6.499635966508919e-07, + "loss": 0.5474, + "step": 43087 + }, + { + "epoch": 7.84, + "learning_rate": 6.499089916272297e-07, + "loss": 0.0238, + "step": 43088 + }, + { + "epoch": 7.84, + "learning_rate": 6.498543866035676e-07, + "loss": 1.254, + "step": 43089 + }, + { + "epoch": 7.84, + "learning_rate": 6.497997815799054e-07, + "loss": 0.4271, + "step": 43090 + }, + { + "epoch": 7.84, + "learning_rate": 6.497451765562431e-07, + "loss": 0.4985, + "step": 43091 + }, + { + "epoch": 7.84, + "learning_rate": 6.496905715325811e-07, + "loss": 0.0408, + "step": 43092 + }, + { + "epoch": 7.84, + "learning_rate": 6.496359665089189e-07, + "loss": 0.0186, + "step": 43093 + }, + { + "epoch": 7.84, + "learning_rate": 6.495813614852566e-07, + "loss": 0.0994, + "step": 43094 + }, + { + "epoch": 7.84, + "learning_rate": 6.495267564615945e-07, + "loss": 0.0341, + "step": 43095 + }, + { + "epoch": 7.84, + "learning_rate": 6.494721514379323e-07, + "loss": 0.1732, + "step": 43096 + }, + { + "epoch": 7.84, + "learning_rate": 6.494175464142701e-07, + "loss": 0.0197, + "step": 43097 + }, + { + "epoch": 7.84, + "learning_rate": 6.493629413906079e-07, + "loss": 0.1291, + "step": 43098 + }, + { + "epoch": 7.84, + "learning_rate": 6.493083363669458e-07, + "loss": 0.7578, + "step": 43099 + }, + { + "epoch": 7.84, + "learning_rate": 6.492537313432836e-07, + "loss": 0.6727, + "step": 43100 + }, + { + "epoch": 7.84, + "learning_rate": 6.491991263196214e-07, + "loss": 0.3411, + "step": 43101 + }, + { + "epoch": 7.84, + "learning_rate": 6.491445212959592e-07, + "loss": 0.1256, + "step": 43102 + }, + { + "epoch": 7.84, + "learning_rate": 6.490899162722971e-07, + "loss": 0.0043, + "step": 43103 + }, + { + "epoch": 7.84, + "learning_rate": 6.490353112486349e-07, + "loss": 0.1143, + "step": 43104 + }, + { + "epoch": 7.84, + "learning_rate": 6.489807062249726e-07, + "loss": 1.2843, + "step": 43105 + }, + { + "epoch": 7.84, + "learning_rate": 6.489261012013106e-07, + "loss": 1.2472, + "step": 43106 + }, + { + "epoch": 7.84, + "learning_rate": 6.488714961776483e-07, + "loss": 0.0109, + "step": 43107 + }, + { + "epoch": 7.84, + "learning_rate": 6.488168911539862e-07, + "loss": 0.4639, + "step": 43108 + }, + { + "epoch": 7.84, + "learning_rate": 6.48762286130324e-07, + "loss": 0.9035, + "step": 43109 + }, + { + "epoch": 7.84, + "learning_rate": 6.487076811066618e-07, + "loss": 0.9322, + "step": 43110 + }, + { + "epoch": 7.84, + "learning_rate": 6.486530760829997e-07, + "loss": 1.1169, + "step": 43111 + }, + { + "epoch": 7.84, + "learning_rate": 6.485984710593375e-07, + "loss": 2.3924, + "step": 43112 + }, + { + "epoch": 7.84, + "learning_rate": 6.485438660356753e-07, + "loss": 0.2318, + "step": 43113 + }, + { + "epoch": 7.84, + "learning_rate": 6.484892610120131e-07, + "loss": 1.0366, + "step": 43114 + }, + { + "epoch": 7.84, + "learning_rate": 6.484346559883509e-07, + "loss": 0.9481, + "step": 43115 + }, + { + "epoch": 7.84, + "learning_rate": 6.483800509646889e-07, + "loss": 0.0089, + "step": 43116 + }, + { + "epoch": 7.84, + "learning_rate": 6.483254459410266e-07, + "loss": 0.5311, + "step": 43117 + }, + { + "epoch": 7.84, + "learning_rate": 6.482708409173644e-07, + "loss": 1.6933, + "step": 43118 + }, + { + "epoch": 7.84, + "learning_rate": 6.482162358937023e-07, + "loss": 2.0079, + "step": 43119 + }, + { + "epoch": 7.84, + "learning_rate": 6.481616308700401e-07, + "loss": 1.3953, + "step": 43120 + }, + { + "epoch": 7.84, + "learning_rate": 6.481070258463778e-07, + "loss": 3.5902, + "step": 43121 + }, + { + "epoch": 7.84, + "learning_rate": 6.480524208227157e-07, + "loss": 0.0188, + "step": 43122 + }, + { + "epoch": 7.84, + "learning_rate": 6.479978157990536e-07, + "loss": 0.1964, + "step": 43123 + }, + { + "epoch": 7.84, + "learning_rate": 6.479432107753913e-07, + "loss": 0.6972, + "step": 43124 + }, + { + "epoch": 7.84, + "learning_rate": 6.478886057517292e-07, + "loss": 0.0823, + "step": 43125 + }, + { + "epoch": 7.84, + "learning_rate": 6.47834000728067e-07, + "loss": 0.8177, + "step": 43126 + }, + { + "epoch": 7.84, + "learning_rate": 6.477793957044049e-07, + "loss": 0.1361, + "step": 43127 + }, + { + "epoch": 7.84, + "learning_rate": 6.477247906807426e-07, + "loss": 0.006, + "step": 43128 + }, + { + "epoch": 7.84, + "learning_rate": 6.476701856570804e-07, + "loss": 0.6228, + "step": 43129 + }, + { + "epoch": 7.84, + "learning_rate": 6.476155806334184e-07, + "loss": 0.3452, + "step": 43130 + }, + { + "epoch": 7.84, + "learning_rate": 6.475609756097561e-07, + "loss": 0.1367, + "step": 43131 + }, + { + "epoch": 7.84, + "learning_rate": 6.475063705860939e-07, + "loss": 0.0803, + "step": 43132 + }, + { + "epoch": 7.84, + "learning_rate": 6.474517655624318e-07, + "loss": 1.5582, + "step": 43133 + }, + { + "epoch": 7.84, + "learning_rate": 6.473971605387696e-07, + "loss": 2.3915, + "step": 43134 + }, + { + "epoch": 7.84, + "learning_rate": 6.473425555151073e-07, + "loss": 0.2924, + "step": 43135 + }, + { + "epoch": 7.84, + "learning_rate": 6.472879504914453e-07, + "loss": 3.0696, + "step": 43136 + }, + { + "epoch": 7.84, + "learning_rate": 6.47233345467783e-07, + "loss": 0.2323, + "step": 43137 + }, + { + "epoch": 7.84, + "learning_rate": 6.471787404441209e-07, + "loss": 0.2603, + "step": 43138 + }, + { + "epoch": 7.84, + "learning_rate": 6.471241354204587e-07, + "loss": 2.0414, + "step": 43139 + }, + { + "epoch": 7.85, + "learning_rate": 6.470695303967965e-07, + "loss": 0.0083, + "step": 43140 + }, + { + "epoch": 7.85, + "learning_rate": 6.470149253731344e-07, + "loss": 0.0059, + "step": 43141 + }, + { + "epoch": 7.85, + "learning_rate": 6.469603203494721e-07, + "loss": 0.01, + "step": 43142 + }, + { + "epoch": 7.85, + "learning_rate": 6.4690571532581e-07, + "loss": 2.7295, + "step": 43143 + }, + { + "epoch": 7.85, + "learning_rate": 6.468511103021478e-07, + "loss": 0.4125, + "step": 43144 + }, + { + "epoch": 7.85, + "learning_rate": 6.467965052784856e-07, + "loss": 1.5503, + "step": 43145 + }, + { + "epoch": 7.85, + "learning_rate": 6.467419002548234e-07, + "loss": 0.0048, + "step": 43146 + }, + { + "epoch": 7.85, + "learning_rate": 6.466872952311613e-07, + "loss": 1.9107, + "step": 43147 + }, + { + "epoch": 7.85, + "learning_rate": 6.466326902074991e-07, + "loss": 0.0106, + "step": 43148 + }, + { + "epoch": 7.85, + "learning_rate": 6.465780851838369e-07, + "loss": 0.0123, + "step": 43149 + }, + { + "epoch": 7.85, + "learning_rate": 6.465234801601748e-07, + "loss": 0.2843, + "step": 43150 + }, + { + "epoch": 7.85, + "learning_rate": 6.464688751365125e-07, + "loss": 1.7299, + "step": 43151 + }, + { + "epoch": 7.85, + "learning_rate": 6.464142701128504e-07, + "loss": 0.5238, + "step": 43152 + }, + { + "epoch": 7.85, + "learning_rate": 6.463596650891883e-07, + "loss": 1.48, + "step": 43153 + }, + { + "epoch": 7.85, + "learning_rate": 6.46305060065526e-07, + "loss": 2.7384, + "step": 43154 + }, + { + "epoch": 7.85, + "learning_rate": 6.462504550418639e-07, + "loss": 0.0237, + "step": 43155 + }, + { + "epoch": 7.85, + "learning_rate": 6.461958500182017e-07, + "loss": 0.3771, + "step": 43156 + }, + { + "epoch": 7.85, + "learning_rate": 6.461412449945396e-07, + "loss": 0.004, + "step": 43157 + }, + { + "epoch": 7.85, + "learning_rate": 6.460866399708773e-07, + "loss": 0.0969, + "step": 43158 + }, + { + "epoch": 7.85, + "learning_rate": 6.460320349472151e-07, + "loss": 0.6316, + "step": 43159 + }, + { + "epoch": 7.85, + "learning_rate": 6.459774299235531e-07, + "loss": 0.2258, + "step": 43160 + }, + { + "epoch": 7.85, + "learning_rate": 6.459228248998908e-07, + "loss": 0.6364, + "step": 43161 + }, + { + "epoch": 7.85, + "learning_rate": 6.458682198762286e-07, + "loss": 0.0077, + "step": 43162 + }, + { + "epoch": 7.85, + "learning_rate": 6.458136148525665e-07, + "loss": 0.2379, + "step": 43163 + }, + { + "epoch": 7.85, + "learning_rate": 6.457590098289043e-07, + "loss": 1.6762, + "step": 43164 + }, + { + "epoch": 7.85, + "learning_rate": 6.45704404805242e-07, + "loss": 0.0264, + "step": 43165 + }, + { + "epoch": 7.85, + "learning_rate": 6.456497997815799e-07, + "loss": 0.0562, + "step": 43166 + }, + { + "epoch": 7.85, + "learning_rate": 6.455951947579178e-07, + "loss": 0.9742, + "step": 43167 + }, + { + "epoch": 7.85, + "learning_rate": 6.455405897342556e-07, + "loss": 0.2208, + "step": 43168 + }, + { + "epoch": 7.85, + "learning_rate": 6.454859847105934e-07, + "loss": 1.5555, + "step": 43169 + }, + { + "epoch": 7.85, + "learning_rate": 6.454313796869312e-07, + "loss": 1.7933, + "step": 43170 + }, + { + "epoch": 7.85, + "learning_rate": 6.453767746632691e-07, + "loss": 0.0055, + "step": 43171 + }, + { + "epoch": 7.85, + "learning_rate": 6.453221696396068e-07, + "loss": 0.6428, + "step": 43172 + }, + { + "epoch": 7.85, + "learning_rate": 6.452675646159446e-07, + "loss": 0.9603, + "step": 43173 + }, + { + "epoch": 7.85, + "learning_rate": 6.452129595922826e-07, + "loss": 0.4109, + "step": 43174 + }, + { + "epoch": 7.85, + "learning_rate": 6.451583545686203e-07, + "loss": 0.0862, + "step": 43175 + }, + { + "epoch": 7.85, + "learning_rate": 6.451037495449581e-07, + "loss": 0.993, + "step": 43176 + }, + { + "epoch": 7.85, + "learning_rate": 6.45049144521296e-07, + "loss": 0.0094, + "step": 43177 + }, + { + "epoch": 7.85, + "learning_rate": 6.449945394976338e-07, + "loss": 1.1572, + "step": 43178 + }, + { + "epoch": 7.85, + "learning_rate": 6.449399344739716e-07, + "loss": 0.3178, + "step": 43179 + }, + { + "epoch": 7.85, + "learning_rate": 6.448853294503095e-07, + "loss": 0.0039, + "step": 43180 + }, + { + "epoch": 7.85, + "learning_rate": 6.448307244266472e-07, + "loss": 0.316, + "step": 43181 + }, + { + "epoch": 7.85, + "learning_rate": 6.447761194029851e-07, + "loss": 1.0245, + "step": 43182 + }, + { + "epoch": 7.85, + "learning_rate": 6.447215143793229e-07, + "loss": 0.2146, + "step": 43183 + }, + { + "epoch": 7.85, + "learning_rate": 6.446669093556607e-07, + "loss": 0.1893, + "step": 43184 + }, + { + "epoch": 7.85, + "learning_rate": 6.446123043319986e-07, + "loss": 0.5161, + "step": 43185 + }, + { + "epoch": 7.85, + "learning_rate": 6.445576993083363e-07, + "loss": 0.0293, + "step": 43186 + }, + { + "epoch": 7.85, + "learning_rate": 6.445030942846743e-07, + "loss": 0.0211, + "step": 43187 + }, + { + "epoch": 7.85, + "learning_rate": 6.44448489261012e-07, + "loss": 1.6972, + "step": 43188 + }, + { + "epoch": 7.85, + "learning_rate": 6.443938842373498e-07, + "loss": 1.5673, + "step": 43189 + }, + { + "epoch": 7.85, + "learning_rate": 6.443392792136877e-07, + "loss": 0.4837, + "step": 43190 + }, + { + "epoch": 7.85, + "learning_rate": 6.442846741900255e-07, + "loss": 1.4388, + "step": 43191 + }, + { + "epoch": 7.85, + "learning_rate": 6.442300691663633e-07, + "loss": 0.5762, + "step": 43192 + }, + { + "epoch": 7.85, + "learning_rate": 6.441754641427011e-07, + "loss": 0.4995, + "step": 43193 + }, + { + "epoch": 7.85, + "learning_rate": 6.44120859119039e-07, + "loss": 0.3085, + "step": 43194 + }, + { + "epoch": 7.86, + "learning_rate": 6.440662540953767e-07, + "loss": 0.04, + "step": 43195 + }, + { + "epoch": 7.86, + "learning_rate": 6.440116490717146e-07, + "loss": 0.0048, + "step": 43196 + }, + { + "epoch": 7.86, + "learning_rate": 6.439570440480525e-07, + "loss": 2.2319, + "step": 43197 + }, + { + "epoch": 7.86, + "learning_rate": 6.439024390243903e-07, + "loss": 0.1092, + "step": 43198 + }, + { + "epoch": 7.86, + "learning_rate": 6.438478340007281e-07, + "loss": 0.2894, + "step": 43199 + }, + { + "epoch": 7.86, + "learning_rate": 6.437932289770659e-07, + "loss": 0.175, + "step": 43200 + }, + { + "epoch": 7.86, + "learning_rate": 6.437386239534038e-07, + "loss": 0.1274, + "step": 43201 + }, + { + "epoch": 7.86, + "learning_rate": 6.436840189297415e-07, + "loss": 0.0824, + "step": 43202 + }, + { + "epoch": 7.86, + "learning_rate": 6.436294139060793e-07, + "loss": 0.0046, + "step": 43203 + }, + { + "epoch": 7.86, + "learning_rate": 6.435748088824173e-07, + "loss": 0.9326, + "step": 43204 + }, + { + "epoch": 7.86, + "learning_rate": 6.43520203858755e-07, + "loss": 0.0145, + "step": 43205 + }, + { + "epoch": 7.86, + "learning_rate": 6.434655988350929e-07, + "loss": 1.3837, + "step": 43206 + }, + { + "epoch": 7.86, + "learning_rate": 6.434109938114307e-07, + "loss": 0.4404, + "step": 43207 + }, + { + "epoch": 7.86, + "learning_rate": 6.433563887877685e-07, + "loss": 0.7785, + "step": 43208 + }, + { + "epoch": 7.86, + "learning_rate": 6.433017837641063e-07, + "loss": 0.0083, + "step": 43209 + }, + { + "epoch": 7.86, + "learning_rate": 6.432471787404441e-07, + "loss": 1.8585, + "step": 43210 + }, + { + "epoch": 7.86, + "learning_rate": 6.43192573716782e-07, + "loss": 0.1748, + "step": 43211 + }, + { + "epoch": 7.86, + "learning_rate": 6.431379686931198e-07, + "loss": 0.5032, + "step": 43212 + }, + { + "epoch": 7.86, + "learning_rate": 6.430833636694576e-07, + "loss": 0.0136, + "step": 43213 + }, + { + "epoch": 7.86, + "learning_rate": 6.430287586457954e-07, + "loss": 0.0585, + "step": 43214 + }, + { + "epoch": 7.86, + "learning_rate": 6.429741536221333e-07, + "loss": 0.429, + "step": 43215 + }, + { + "epoch": 7.86, + "learning_rate": 6.42919548598471e-07, + "loss": 0.1107, + "step": 43216 + }, + { + "epoch": 7.86, + "learning_rate": 6.42864943574809e-07, + "loss": 2.7347, + "step": 43217 + }, + { + "epoch": 7.86, + "learning_rate": 6.428103385511467e-07, + "loss": 1.0279, + "step": 43218 + }, + { + "epoch": 7.86, + "learning_rate": 6.427557335274845e-07, + "loss": 0.3339, + "step": 43219 + }, + { + "epoch": 7.86, + "learning_rate": 6.427011285038224e-07, + "loss": 0.0073, + "step": 43220 + }, + { + "epoch": 7.86, + "learning_rate": 6.426465234801602e-07, + "loss": 0.0305, + "step": 43221 + }, + { + "epoch": 7.86, + "learning_rate": 6.42591918456498e-07, + "loss": 0.0076, + "step": 43222 + }, + { + "epoch": 7.86, + "learning_rate": 6.425373134328358e-07, + "loss": 0.0138, + "step": 43223 + }, + { + "epoch": 7.86, + "learning_rate": 6.424827084091737e-07, + "loss": 0.5734, + "step": 43224 + }, + { + "epoch": 7.86, + "learning_rate": 6.424281033855114e-07, + "loss": 0.0464, + "step": 43225 + }, + { + "epoch": 7.86, + "learning_rate": 6.423734983618493e-07, + "loss": 0.2819, + "step": 43226 + }, + { + "epoch": 7.86, + "learning_rate": 6.423188933381871e-07, + "loss": 0.6833, + "step": 43227 + }, + { + "epoch": 7.86, + "learning_rate": 6.42264288314525e-07, + "loss": 0.2056, + "step": 43228 + }, + { + "epoch": 7.86, + "learning_rate": 6.422096832908628e-07, + "loss": 0.4776, + "step": 43229 + }, + { + "epoch": 7.86, + "learning_rate": 6.421550782672005e-07, + "loss": 0.0877, + "step": 43230 + }, + { + "epoch": 7.86, + "learning_rate": 6.421004732435385e-07, + "loss": 0.0156, + "step": 43231 + }, + { + "epoch": 7.86, + "learning_rate": 6.420458682198762e-07, + "loss": 0.1796, + "step": 43232 + }, + { + "epoch": 7.86, + "learning_rate": 6.41991263196214e-07, + "loss": 0.0352, + "step": 43233 + }, + { + "epoch": 7.86, + "learning_rate": 6.419366581725519e-07, + "loss": 0.9174, + "step": 43234 + }, + { + "epoch": 7.86, + "learning_rate": 6.418820531488897e-07, + "loss": 0.0076, + "step": 43235 + }, + { + "epoch": 7.86, + "learning_rate": 6.418274481252276e-07, + "loss": 0.0399, + "step": 43236 + }, + { + "epoch": 7.86, + "learning_rate": 6.417728431015654e-07, + "loss": 0.9954, + "step": 43237 + }, + { + "epoch": 7.86, + "learning_rate": 6.417182380779032e-07, + "loss": 0.1227, + "step": 43238 + }, + { + "epoch": 7.86, + "learning_rate": 6.41663633054241e-07, + "loss": 0.8147, + "step": 43239 + }, + { + "epoch": 7.86, + "learning_rate": 6.416090280305788e-07, + "loss": 0.791, + "step": 43240 + }, + { + "epoch": 7.86, + "learning_rate": 6.415544230069167e-07, + "loss": 0.0088, + "step": 43241 + }, + { + "epoch": 7.86, + "learning_rate": 6.414998179832545e-07, + "loss": 1.314, + "step": 43242 + }, + { + "epoch": 7.86, + "learning_rate": 6.414452129595923e-07, + "loss": 0.3675, + "step": 43243 + }, + { + "epoch": 7.86, + "learning_rate": 6.413906079359301e-07, + "loss": 0.5669, + "step": 43244 + }, + { + "epoch": 7.86, + "learning_rate": 6.41336002912268e-07, + "loss": 0.0276, + "step": 43245 + }, + { + "epoch": 7.86, + "learning_rate": 6.412813978886057e-07, + "loss": 0.1202, + "step": 43246 + }, + { + "epoch": 7.86, + "learning_rate": 6.412267928649436e-07, + "loss": 0.1534, + "step": 43247 + }, + { + "epoch": 7.86, + "learning_rate": 6.411721878412815e-07, + "loss": 1.4758, + "step": 43248 + }, + { + "epoch": 7.86, + "learning_rate": 6.411175828176192e-07, + "loss": 0.9294, + "step": 43249 + }, + { + "epoch": 7.87, + "learning_rate": 6.410629777939571e-07, + "loss": 0.0192, + "step": 43250 + }, + { + "epoch": 7.87, + "learning_rate": 6.410083727702949e-07, + "loss": 1.4057, + "step": 43251 + }, + { + "epoch": 7.87, + "learning_rate": 6.409537677466327e-07, + "loss": 0.6249, + "step": 43252 + }, + { + "epoch": 7.87, + "learning_rate": 6.408991627229705e-07, + "loss": 0.7684, + "step": 43253 + }, + { + "epoch": 7.87, + "learning_rate": 6.408445576993083e-07, + "loss": 0.1268, + "step": 43254 + }, + { + "epoch": 7.87, + "learning_rate": 6.407899526756461e-07, + "loss": 0.3572, + "step": 43255 + }, + { + "epoch": 7.87, + "learning_rate": 6.40735347651984e-07, + "loss": 0.1791, + "step": 43256 + }, + { + "epoch": 7.87, + "learning_rate": 6.406807426283218e-07, + "loss": 0.2455, + "step": 43257 + }, + { + "epoch": 7.87, + "learning_rate": 6.406261376046597e-07, + "loss": 2.8393, + "step": 43258 + }, + { + "epoch": 7.87, + "learning_rate": 6.405715325809975e-07, + "loss": 2.6573, + "step": 43259 + }, + { + "epoch": 7.87, + "learning_rate": 6.405169275573352e-07, + "loss": 0.3517, + "step": 43260 + }, + { + "epoch": 7.87, + "learning_rate": 6.404623225336732e-07, + "loss": 0.2067, + "step": 43261 + }, + { + "epoch": 7.87, + "learning_rate": 6.404077175100109e-07, + "loss": 0.0325, + "step": 43262 + }, + { + "epoch": 7.87, + "learning_rate": 6.403531124863487e-07, + "loss": 1.388, + "step": 43263 + }, + { + "epoch": 7.87, + "learning_rate": 6.402985074626866e-07, + "loss": 0.4736, + "step": 43264 + }, + { + "epoch": 7.87, + "learning_rate": 6.402439024390244e-07, + "loss": 0.1356, + "step": 43265 + }, + { + "epoch": 7.87, + "learning_rate": 6.401892974153623e-07, + "loss": 0.0087, + "step": 43266 + }, + { + "epoch": 7.87, + "learning_rate": 6.401346923917e-07, + "loss": 0.683, + "step": 43267 + }, + { + "epoch": 7.87, + "learning_rate": 6.400800873680379e-07, + "loss": 0.8663, + "step": 43268 + }, + { + "epoch": 7.87, + "learning_rate": 6.400254823443757e-07, + "loss": 0.2042, + "step": 43269 + }, + { + "epoch": 7.87, + "learning_rate": 6.399708773207135e-07, + "loss": 0.8049, + "step": 43270 + }, + { + "epoch": 7.87, + "learning_rate": 6.399162722970513e-07, + "loss": 0.0053, + "step": 43271 + }, + { + "epoch": 7.87, + "learning_rate": 6.398616672733892e-07, + "loss": 0.7921, + "step": 43272 + }, + { + "epoch": 7.87, + "learning_rate": 6.39807062249727e-07, + "loss": 0.0053, + "step": 43273 + }, + { + "epoch": 7.87, + "learning_rate": 6.397524572260647e-07, + "loss": 1.7636, + "step": 43274 + }, + { + "epoch": 7.87, + "learning_rate": 6.396978522024027e-07, + "loss": 1.8379, + "step": 43275 + }, + { + "epoch": 7.87, + "learning_rate": 6.396432471787404e-07, + "loss": 0.3979, + "step": 43276 + }, + { + "epoch": 7.87, + "learning_rate": 6.395886421550783e-07, + "loss": 0.0261, + "step": 43277 + }, + { + "epoch": 7.87, + "learning_rate": 6.395340371314161e-07, + "loss": 0.1208, + "step": 43278 + }, + { + "epoch": 7.87, + "learning_rate": 6.394794321077539e-07, + "loss": 0.6493, + "step": 43279 + }, + { + "epoch": 7.87, + "learning_rate": 6.394248270840918e-07, + "loss": 2.3208, + "step": 43280 + }, + { + "epoch": 7.87, + "learning_rate": 6.393702220604296e-07, + "loss": 0.014, + "step": 43281 + }, + { + "epoch": 7.87, + "learning_rate": 6.393156170367674e-07, + "loss": 0.0429, + "step": 43282 + }, + { + "epoch": 7.87, + "learning_rate": 6.392610120131052e-07, + "loss": 1.5329, + "step": 43283 + }, + { + "epoch": 7.87, + "learning_rate": 6.39206406989443e-07, + "loss": 0.4395, + "step": 43284 + }, + { + "epoch": 7.87, + "learning_rate": 6.39151801965781e-07, + "loss": 0.1401, + "step": 43285 + }, + { + "epoch": 7.87, + "learning_rate": 6.390971969421187e-07, + "loss": 1.7444, + "step": 43286 + }, + { + "epoch": 7.87, + "learning_rate": 6.390425919184565e-07, + "loss": 0.5858, + "step": 43287 + }, + { + "epoch": 7.87, + "learning_rate": 6.389879868947944e-07, + "loss": 4.5116, + "step": 43288 + }, + { + "epoch": 7.87, + "learning_rate": 6.389333818711322e-07, + "loss": 0.7075, + "step": 43289 + }, + { + "epoch": 7.87, + "learning_rate": 6.388787768474699e-07, + "loss": 0.0487, + "step": 43290 + }, + { + "epoch": 7.87, + "learning_rate": 6.388241718238078e-07, + "loss": 0.0084, + "step": 43291 + }, + { + "epoch": 7.87, + "learning_rate": 6.387695668001456e-07, + "loss": 0.0126, + "step": 43292 + }, + { + "epoch": 7.87, + "learning_rate": 6.387149617764834e-07, + "loss": 1.1452, + "step": 43293 + }, + { + "epoch": 7.87, + "learning_rate": 6.386603567528213e-07, + "loss": 0.0838, + "step": 43294 + }, + { + "epoch": 7.87, + "learning_rate": 6.386057517291591e-07, + "loss": 0.0088, + "step": 43295 + }, + { + "epoch": 7.87, + "learning_rate": 6.38551146705497e-07, + "loss": 0.3837, + "step": 43296 + }, + { + "epoch": 7.87, + "learning_rate": 6.384965416818347e-07, + "loss": 0.3024, + "step": 43297 + }, + { + "epoch": 7.87, + "learning_rate": 6.384419366581725e-07, + "loss": 1.0782, + "step": 43298 + }, + { + "epoch": 7.87, + "learning_rate": 6.383873316345104e-07, + "loss": 0.3282, + "step": 43299 + }, + { + "epoch": 7.87, + "learning_rate": 6.383327266108482e-07, + "loss": 0.7877, + "step": 43300 + }, + { + "epoch": 7.87, + "learning_rate": 6.38278121587186e-07, + "loss": 0.2471, + "step": 43301 + }, + { + "epoch": 7.87, + "learning_rate": 6.382235165635239e-07, + "loss": 1.1998, + "step": 43302 + }, + { + "epoch": 7.87, + "learning_rate": 6.381689115398617e-07, + "loss": 0.0058, + "step": 43303 + }, + { + "epoch": 7.87, + "learning_rate": 6.381143065161994e-07, + "loss": 0.2248, + "step": 43304 + }, + { + "epoch": 7.88, + "learning_rate": 6.380597014925374e-07, + "loss": 0.8125, + "step": 43305 + }, + { + "epoch": 7.88, + "learning_rate": 6.380050964688751e-07, + "loss": 0.1054, + "step": 43306 + }, + { + "epoch": 7.88, + "learning_rate": 6.37950491445213e-07, + "loss": 0.0113, + "step": 43307 + }, + { + "epoch": 7.88, + "learning_rate": 6.378958864215508e-07, + "loss": 0.025, + "step": 43308 + }, + { + "epoch": 7.88, + "learning_rate": 6.378412813978886e-07, + "loss": 0.5053, + "step": 43309 + }, + { + "epoch": 7.88, + "learning_rate": 6.377866763742265e-07, + "loss": 0.4962, + "step": 43310 + }, + { + "epoch": 7.88, + "learning_rate": 6.377320713505642e-07, + "loss": 1.3215, + "step": 43311 + }, + { + "epoch": 7.88, + "learning_rate": 6.376774663269021e-07, + "loss": 2.913, + "step": 43312 + }, + { + "epoch": 7.88, + "learning_rate": 6.376228613032399e-07, + "loss": 0.0914, + "step": 43313 + }, + { + "epoch": 7.88, + "learning_rate": 6.375682562795777e-07, + "loss": 1.1463, + "step": 43314 + }, + { + "epoch": 7.88, + "learning_rate": 6.375136512559156e-07, + "loss": 0.0095, + "step": 43315 + }, + { + "epoch": 7.88, + "learning_rate": 6.374590462322534e-07, + "loss": 0.2568, + "step": 43316 + }, + { + "epoch": 7.88, + "learning_rate": 6.374044412085912e-07, + "loss": 0.0589, + "step": 43317 + }, + { + "epoch": 7.88, + "learning_rate": 6.37349836184929e-07, + "loss": 0.0138, + "step": 43318 + }, + { + "epoch": 7.88, + "learning_rate": 6.372952311612669e-07, + "loss": 0.0841, + "step": 43319 + }, + { + "epoch": 7.88, + "learning_rate": 6.372406261376046e-07, + "loss": 0.3519, + "step": 43320 + }, + { + "epoch": 7.88, + "learning_rate": 6.371860211139425e-07, + "loss": 0.2877, + "step": 43321 + }, + { + "epoch": 7.88, + "learning_rate": 6.371314160902803e-07, + "loss": 0.2994, + "step": 43322 + }, + { + "epoch": 7.88, + "learning_rate": 6.370768110666181e-07, + "loss": 1.298, + "step": 43323 + }, + { + "epoch": 7.88, + "learning_rate": 6.37022206042956e-07, + "loss": 0.054, + "step": 43324 + }, + { + "epoch": 7.88, + "learning_rate": 6.369676010192938e-07, + "loss": 0.131, + "step": 43325 + }, + { + "epoch": 7.88, + "learning_rate": 6.369129959956317e-07, + "loss": 4.6146, + "step": 43326 + }, + { + "epoch": 7.88, + "learning_rate": 6.368583909719694e-07, + "loss": 0.0195, + "step": 43327 + }, + { + "epoch": 7.88, + "learning_rate": 6.368037859483072e-07, + "loss": 0.0054, + "step": 43328 + }, + { + "epoch": 7.88, + "learning_rate": 6.367491809246452e-07, + "loss": 0.1554, + "step": 43329 + }, + { + "epoch": 7.88, + "learning_rate": 6.366945759009829e-07, + "loss": 0.0857, + "step": 43330 + }, + { + "epoch": 7.88, + "learning_rate": 6.366399708773207e-07, + "loss": 0.8155, + "step": 43331 + }, + { + "epoch": 7.88, + "learning_rate": 6.365853658536586e-07, + "loss": 0.7265, + "step": 43332 + }, + { + "epoch": 7.88, + "learning_rate": 6.365307608299964e-07, + "loss": 3.5228, + "step": 43333 + }, + { + "epoch": 7.88, + "learning_rate": 6.364761558063341e-07, + "loss": 0.0135, + "step": 43334 + }, + { + "epoch": 7.88, + "learning_rate": 6.36421550782672e-07, + "loss": 0.9404, + "step": 43335 + }, + { + "epoch": 7.88, + "learning_rate": 6.363669457590098e-07, + "loss": 0.1893, + "step": 43336 + }, + { + "epoch": 7.88, + "learning_rate": 6.363123407353477e-07, + "loss": 0.0307, + "step": 43337 + }, + { + "epoch": 7.88, + "learning_rate": 6.362577357116855e-07, + "loss": 0.9142, + "step": 43338 + }, + { + "epoch": 7.88, + "learning_rate": 6.362031306880233e-07, + "loss": 0.0266, + "step": 43339 + }, + { + "epoch": 7.88, + "learning_rate": 6.361485256643612e-07, + "loss": 0.1982, + "step": 43340 + }, + { + "epoch": 7.88, + "learning_rate": 6.360939206406989e-07, + "loss": 0.5973, + "step": 43341 + }, + { + "epoch": 7.88, + "learning_rate": 6.360393156170367e-07, + "loss": 2.2733, + "step": 43342 + }, + { + "epoch": 7.88, + "learning_rate": 6.359847105933746e-07, + "loss": 0.0089, + "step": 43343 + }, + { + "epoch": 7.88, + "learning_rate": 6.359301055697124e-07, + "loss": 0.2808, + "step": 43344 + }, + { + "epoch": 7.88, + "learning_rate": 6.358755005460503e-07, + "loss": 0.1603, + "step": 43345 + }, + { + "epoch": 7.88, + "learning_rate": 6.358208955223881e-07, + "loss": 1.0145, + "step": 43346 + }, + { + "epoch": 7.88, + "learning_rate": 6.357662904987259e-07, + "loss": 0.9746, + "step": 43347 + }, + { + "epoch": 7.88, + "learning_rate": 6.357116854750637e-07, + "loss": 0.4025, + "step": 43348 + }, + { + "epoch": 7.88, + "learning_rate": 6.356570804514016e-07, + "loss": 0.1386, + "step": 43349 + }, + { + "epoch": 7.88, + "learning_rate": 6.356024754277393e-07, + "loss": 0.1278, + "step": 43350 + }, + { + "epoch": 7.88, + "learning_rate": 6.355478704040772e-07, + "loss": 0.0699, + "step": 43351 + }, + { + "epoch": 7.88, + "learning_rate": 6.35493265380415e-07, + "loss": 0.6083, + "step": 43352 + }, + { + "epoch": 7.88, + "learning_rate": 6.354386603567528e-07, + "loss": 0.272, + "step": 43353 + }, + { + "epoch": 7.88, + "learning_rate": 6.353840553330907e-07, + "loss": 0.1311, + "step": 43354 + }, + { + "epoch": 7.88, + "learning_rate": 6.353294503094284e-07, + "loss": 0.14, + "step": 43355 + }, + { + "epoch": 7.88, + "learning_rate": 6.352748452857664e-07, + "loss": 0.3429, + "step": 43356 + }, + { + "epoch": 7.88, + "learning_rate": 6.352202402621041e-07, + "loss": 0.0779, + "step": 43357 + }, + { + "epoch": 7.88, + "learning_rate": 6.351656352384419e-07, + "loss": 0.6626, + "step": 43358 + }, + { + "epoch": 7.88, + "learning_rate": 6.351110302147798e-07, + "loss": 0.1311, + "step": 43359 + }, + { + "epoch": 7.89, + "learning_rate": 6.350564251911176e-07, + "loss": 0.6628, + "step": 43360 + }, + { + "epoch": 7.89, + "learning_rate": 6.350018201674554e-07, + "loss": 0.3864, + "step": 43361 + }, + { + "epoch": 7.89, + "learning_rate": 6.349472151437932e-07, + "loss": 0.2988, + "step": 43362 + }, + { + "epoch": 7.89, + "learning_rate": 6.348926101201311e-07, + "loss": 0.0164, + "step": 43363 + }, + { + "epoch": 7.89, + "learning_rate": 6.348380050964688e-07, + "loss": 0.8259, + "step": 43364 + }, + { + "epoch": 7.89, + "learning_rate": 6.347834000728067e-07, + "loss": 0.0349, + "step": 43365 + }, + { + "epoch": 7.89, + "learning_rate": 6.347287950491446e-07, + "loss": 1.1119, + "step": 43366 + }, + { + "epoch": 7.89, + "learning_rate": 6.346741900254824e-07, + "loss": 0.0051, + "step": 43367 + }, + { + "epoch": 7.89, + "learning_rate": 6.346195850018202e-07, + "loss": 0.6427, + "step": 43368 + }, + { + "epoch": 7.89, + "learning_rate": 6.34564979978158e-07, + "loss": 0.1986, + "step": 43369 + }, + { + "epoch": 7.89, + "learning_rate": 6.345103749544959e-07, + "loss": 0.222, + "step": 43370 + }, + { + "epoch": 7.89, + "learning_rate": 6.344557699308336e-07, + "loss": 0.2422, + "step": 43371 + }, + { + "epoch": 7.89, + "learning_rate": 6.344011649071714e-07, + "loss": 2.9963, + "step": 43372 + }, + { + "epoch": 7.89, + "learning_rate": 6.343465598835093e-07, + "loss": 0.0083, + "step": 43373 + }, + { + "epoch": 7.89, + "learning_rate": 6.342919548598471e-07, + "loss": 0.0684, + "step": 43374 + }, + { + "epoch": 7.89, + "learning_rate": 6.34237349836185e-07, + "loss": 0.0884, + "step": 43375 + }, + { + "epoch": 7.89, + "learning_rate": 6.341827448125228e-07, + "loss": 0.0046, + "step": 43376 + }, + { + "epoch": 7.89, + "learning_rate": 6.341281397888606e-07, + "loss": 0.015, + "step": 43377 + }, + { + "epoch": 7.89, + "learning_rate": 6.340735347651984e-07, + "loss": 2.2255, + "step": 43378 + }, + { + "epoch": 7.89, + "learning_rate": 6.340189297415362e-07, + "loss": 0.6525, + "step": 43379 + }, + { + "epoch": 7.89, + "learning_rate": 6.33964324717874e-07, + "loss": 0.4608, + "step": 43380 + }, + { + "epoch": 7.89, + "learning_rate": 6.339097196942119e-07, + "loss": 0.0078, + "step": 43381 + }, + { + "epoch": 7.89, + "learning_rate": 6.338551146705497e-07, + "loss": 0.0055, + "step": 43382 + }, + { + "epoch": 7.89, + "learning_rate": 6.338005096468875e-07, + "loss": 0.1256, + "step": 43383 + }, + { + "epoch": 7.89, + "learning_rate": 6.337459046232254e-07, + "loss": 0.9632, + "step": 43384 + }, + { + "epoch": 7.89, + "learning_rate": 6.336912995995631e-07, + "loss": 0.1171, + "step": 43385 + }, + { + "epoch": 7.89, + "learning_rate": 6.336366945759011e-07, + "loss": 1.4338, + "step": 43386 + }, + { + "epoch": 7.89, + "learning_rate": 6.335820895522388e-07, + "loss": 0.2255, + "step": 43387 + }, + { + "epoch": 7.89, + "learning_rate": 6.335274845285766e-07, + "loss": 1.1712, + "step": 43388 + }, + { + "epoch": 7.89, + "learning_rate": 6.334728795049145e-07, + "loss": 0.7973, + "step": 43389 + }, + { + "epoch": 7.89, + "learning_rate": 6.334182744812523e-07, + "loss": 0.0376, + "step": 43390 + }, + { + "epoch": 7.89, + "learning_rate": 6.333636694575901e-07, + "loss": 0.1225, + "step": 43391 + }, + { + "epoch": 7.89, + "learning_rate": 6.333090644339279e-07, + "loss": 1.3925, + "step": 43392 + }, + { + "epoch": 7.89, + "learning_rate": 6.332544594102658e-07, + "loss": 1.074, + "step": 43393 + }, + { + "epoch": 7.89, + "learning_rate": 6.331998543866036e-07, + "loss": 0.0736, + "step": 43394 + }, + { + "epoch": 7.89, + "learning_rate": 6.331452493629414e-07, + "loss": 0.0089, + "step": 43395 + }, + { + "epoch": 7.89, + "learning_rate": 6.330906443392792e-07, + "loss": 0.737, + "step": 43396 + }, + { + "epoch": 7.89, + "learning_rate": 6.330360393156171e-07, + "loss": 0.5398, + "step": 43397 + }, + { + "epoch": 7.89, + "learning_rate": 6.329814342919549e-07, + "loss": 0.0307, + "step": 43398 + }, + { + "epoch": 7.89, + "learning_rate": 6.329268292682926e-07, + "loss": 0.8374, + "step": 43399 + }, + { + "epoch": 7.89, + "learning_rate": 6.328722242446306e-07, + "loss": 0.4228, + "step": 43400 + }, + { + "epoch": 7.89, + "learning_rate": 6.328176192209683e-07, + "loss": 2.7801, + "step": 43401 + }, + { + "epoch": 7.89, + "learning_rate": 6.327630141973061e-07, + "loss": 0.3098, + "step": 43402 + }, + { + "epoch": 7.89, + "learning_rate": 6.32708409173644e-07, + "loss": 2.5676, + "step": 43403 + }, + { + "epoch": 7.89, + "learning_rate": 6.326538041499818e-07, + "loss": 0.0461, + "step": 43404 + }, + { + "epoch": 7.89, + "learning_rate": 6.325991991263197e-07, + "loss": 0.0548, + "step": 43405 + }, + { + "epoch": 7.89, + "learning_rate": 6.325445941026574e-07, + "loss": 0.4974, + "step": 43406 + }, + { + "epoch": 7.89, + "learning_rate": 6.324899890789953e-07, + "loss": 0.0051, + "step": 43407 + }, + { + "epoch": 7.89, + "learning_rate": 6.324353840553331e-07, + "loss": 0.2198, + "step": 43408 + }, + { + "epoch": 7.89, + "learning_rate": 6.323807790316709e-07, + "loss": 0.1196, + "step": 43409 + }, + { + "epoch": 7.89, + "learning_rate": 6.323261740080087e-07, + "loss": 0.5927, + "step": 43410 + }, + { + "epoch": 7.89, + "learning_rate": 6.322715689843466e-07, + "loss": 0.0528, + "step": 43411 + }, + { + "epoch": 7.89, + "learning_rate": 6.322169639606844e-07, + "loss": 0.8713, + "step": 43412 + }, + { + "epoch": 7.89, + "learning_rate": 6.321623589370223e-07, + "loss": 0.0564, + "step": 43413 + }, + { + "epoch": 7.89, + "learning_rate": 6.321077539133601e-07, + "loss": 0.0203, + "step": 43414 + }, + { + "epoch": 7.9, + "learning_rate": 6.320531488896978e-07, + "loss": 0.1842, + "step": 43415 + }, + { + "epoch": 7.9, + "learning_rate": 6.319985438660357e-07, + "loss": 0.048, + "step": 43416 + }, + { + "epoch": 7.9, + "learning_rate": 6.319439388423735e-07, + "loss": 0.4816, + "step": 43417 + }, + { + "epoch": 7.9, + "learning_rate": 6.318893338187113e-07, + "loss": 0.5704, + "step": 43418 + }, + { + "epoch": 7.9, + "learning_rate": 6.318347287950492e-07, + "loss": 0.015, + "step": 43419 + }, + { + "epoch": 7.9, + "learning_rate": 6.31780123771387e-07, + "loss": 0.3924, + "step": 43420 + }, + { + "epoch": 7.9, + "learning_rate": 6.317255187477248e-07, + "loss": 2.6561, + "step": 43421 + }, + { + "epoch": 7.9, + "learning_rate": 6.316709137240626e-07, + "loss": 0.0138, + "step": 43422 + }, + { + "epoch": 7.9, + "learning_rate": 6.316163087004004e-07, + "loss": 0.0098, + "step": 43423 + }, + { + "epoch": 7.9, + "learning_rate": 6.315617036767383e-07, + "loss": 0.6923, + "step": 43424 + }, + { + "epoch": 7.9, + "learning_rate": 6.315070986530761e-07, + "loss": 1.546, + "step": 43425 + }, + { + "epoch": 7.9, + "learning_rate": 6.314524936294139e-07, + "loss": 0.1443, + "step": 43426 + }, + { + "epoch": 7.9, + "learning_rate": 6.313978886057518e-07, + "loss": 0.0203, + "step": 43427 + }, + { + "epoch": 7.9, + "learning_rate": 6.313432835820896e-07, + "loss": 0.2231, + "step": 43428 + }, + { + "epoch": 7.9, + "learning_rate": 6.312886785584273e-07, + "loss": 0.1431, + "step": 43429 + }, + { + "epoch": 7.9, + "learning_rate": 6.312340735347653e-07, + "loss": 0.8617, + "step": 43430 + }, + { + "epoch": 7.9, + "learning_rate": 6.31179468511103e-07, + "loss": 1.4492, + "step": 43431 + }, + { + "epoch": 7.9, + "learning_rate": 6.311248634874408e-07, + "loss": 0.1367, + "step": 43432 + }, + { + "epoch": 7.9, + "learning_rate": 6.310702584637787e-07, + "loss": 0.0145, + "step": 43433 + }, + { + "epoch": 7.9, + "learning_rate": 6.310156534401165e-07, + "loss": 0.6604, + "step": 43434 + }, + { + "epoch": 7.9, + "learning_rate": 6.309610484164544e-07, + "loss": 1.9651, + "step": 43435 + }, + { + "epoch": 7.9, + "learning_rate": 6.309064433927921e-07, + "loss": 0.0906, + "step": 43436 + }, + { + "epoch": 7.9, + "learning_rate": 6.3085183836913e-07, + "loss": 0.0315, + "step": 43437 + }, + { + "epoch": 7.9, + "learning_rate": 6.307972333454678e-07, + "loss": 0.889, + "step": 43438 + }, + { + "epoch": 7.9, + "learning_rate": 6.307426283218056e-07, + "loss": 1.4093, + "step": 43439 + }, + { + "epoch": 7.9, + "learning_rate": 6.306880232981434e-07, + "loss": 0.0072, + "step": 43440 + }, + { + "epoch": 7.9, + "learning_rate": 6.306334182744813e-07, + "loss": 0.0907, + "step": 43441 + }, + { + "epoch": 7.9, + "learning_rate": 6.305788132508191e-07, + "loss": 0.1798, + "step": 43442 + }, + { + "epoch": 7.9, + "learning_rate": 6.305242082271568e-07, + "loss": 0.0052, + "step": 43443 + }, + { + "epoch": 7.9, + "learning_rate": 6.304696032034948e-07, + "loss": 0.0922, + "step": 43444 + }, + { + "epoch": 7.9, + "learning_rate": 6.304149981798325e-07, + "loss": 0.8761, + "step": 43445 + }, + { + "epoch": 7.9, + "learning_rate": 6.303603931561704e-07, + "loss": 0.0065, + "step": 43446 + }, + { + "epoch": 7.9, + "learning_rate": 6.303057881325082e-07, + "loss": 0.2197, + "step": 43447 + }, + { + "epoch": 7.9, + "learning_rate": 6.30251183108846e-07, + "loss": 1.0334, + "step": 43448 + }, + { + "epoch": 7.9, + "learning_rate": 6.301965780851839e-07, + "loss": 0.7101, + "step": 43449 + }, + { + "epoch": 7.9, + "learning_rate": 6.301419730615217e-07, + "loss": 0.6484, + "step": 43450 + }, + { + "epoch": 7.9, + "learning_rate": 6.300873680378595e-07, + "loss": 0.3542, + "step": 43451 + }, + { + "epoch": 7.9, + "learning_rate": 6.300327630141973e-07, + "loss": 0.9037, + "step": 43452 + }, + { + "epoch": 7.9, + "learning_rate": 6.299781579905351e-07, + "loss": 0.0728, + "step": 43453 + }, + { + "epoch": 7.9, + "learning_rate": 6.29923552966873e-07, + "loss": 0.0131, + "step": 43454 + }, + { + "epoch": 7.9, + "learning_rate": 6.298689479432108e-07, + "loss": 0.1556, + "step": 43455 + }, + { + "epoch": 7.9, + "learning_rate": 6.298143429195486e-07, + "loss": 0.0936, + "step": 43456 + }, + { + "epoch": 7.9, + "learning_rate": 6.297597378958865e-07, + "loss": 0.0071, + "step": 43457 + }, + { + "epoch": 7.9, + "learning_rate": 6.297051328722243e-07, + "loss": 0.0064, + "step": 43458 + }, + { + "epoch": 7.9, + "learning_rate": 6.29650527848562e-07, + "loss": 0.0651, + "step": 43459 + }, + { + "epoch": 7.9, + "learning_rate": 6.295959228248999e-07, + "loss": 0.4291, + "step": 43460 + }, + { + "epoch": 7.9, + "learning_rate": 6.295413178012377e-07, + "loss": 1.3722, + "step": 43461 + }, + { + "epoch": 7.9, + "learning_rate": 6.294867127775755e-07, + "loss": 1.8815, + "step": 43462 + }, + { + "epoch": 7.9, + "learning_rate": 6.294321077539134e-07, + "loss": 1.1862, + "step": 43463 + }, + { + "epoch": 7.9, + "learning_rate": 6.293775027302512e-07, + "loss": 0.5531, + "step": 43464 + }, + { + "epoch": 7.9, + "learning_rate": 6.293228977065891e-07, + "loss": 0.0683, + "step": 43465 + }, + { + "epoch": 7.9, + "learning_rate": 6.292682926829268e-07, + "loss": 1.9658, + "step": 43466 + }, + { + "epoch": 7.9, + "learning_rate": 6.292136876592646e-07, + "loss": 0.4046, + "step": 43467 + }, + { + "epoch": 7.9, + "learning_rate": 6.291590826356025e-07, + "loss": 1.2603, + "step": 43468 + }, + { + "epoch": 7.9, + "learning_rate": 6.291044776119403e-07, + "loss": 0.5897, + "step": 43469 + }, + { + "epoch": 7.91, + "learning_rate": 6.290498725882781e-07, + "loss": 0.3626, + "step": 43470 + }, + { + "epoch": 7.91, + "learning_rate": 6.28995267564616e-07, + "loss": 0.0113, + "step": 43471 + }, + { + "epoch": 7.91, + "learning_rate": 6.289406625409538e-07, + "loss": 0.0697, + "step": 43472 + }, + { + "epoch": 7.91, + "learning_rate": 6.288860575172915e-07, + "loss": 0.2117, + "step": 43473 + }, + { + "epoch": 7.91, + "learning_rate": 6.288314524936295e-07, + "loss": 0.2417, + "step": 43474 + }, + { + "epoch": 7.91, + "learning_rate": 6.287768474699672e-07, + "loss": 0.3687, + "step": 43475 + }, + { + "epoch": 7.91, + "learning_rate": 6.287222424463051e-07, + "loss": 0.1004, + "step": 43476 + }, + { + "epoch": 7.91, + "learning_rate": 6.286676374226429e-07, + "loss": 0.0035, + "step": 43477 + }, + { + "epoch": 7.91, + "learning_rate": 6.286130323989807e-07, + "loss": 0.2454, + "step": 43478 + }, + { + "epoch": 7.91, + "learning_rate": 6.285584273753186e-07, + "loss": 0.5455, + "step": 43479 + }, + { + "epoch": 7.91, + "learning_rate": 6.285038223516563e-07, + "loss": 0.0108, + "step": 43480 + }, + { + "epoch": 7.91, + "learning_rate": 6.284492173279942e-07, + "loss": 0.9196, + "step": 43481 + }, + { + "epoch": 7.91, + "learning_rate": 6.28394612304332e-07, + "loss": 0.0098, + "step": 43482 + }, + { + "epoch": 7.91, + "learning_rate": 6.283400072806698e-07, + "loss": 0.0373, + "step": 43483 + }, + { + "epoch": 7.91, + "learning_rate": 6.282854022570077e-07, + "loss": 0.1281, + "step": 43484 + }, + { + "epoch": 7.91, + "learning_rate": 6.282307972333455e-07, + "loss": 0.0056, + "step": 43485 + }, + { + "epoch": 7.91, + "learning_rate": 6.281761922096833e-07, + "loss": 0.1812, + "step": 43486 + }, + { + "epoch": 7.91, + "learning_rate": 6.281215871860211e-07, + "loss": 0.8977, + "step": 43487 + }, + { + "epoch": 7.91, + "learning_rate": 6.28066982162359e-07, + "loss": 1.18, + "step": 43488 + }, + { + "epoch": 7.91, + "learning_rate": 6.280123771386967e-07, + "loss": 0.289, + "step": 43489 + }, + { + "epoch": 7.91, + "learning_rate": 6.279577721150346e-07, + "loss": 0.283, + "step": 43490 + }, + { + "epoch": 7.91, + "learning_rate": 6.279031670913724e-07, + "loss": 0.2087, + "step": 43491 + }, + { + "epoch": 7.91, + "learning_rate": 6.278485620677103e-07, + "loss": 0.302, + "step": 43492 + }, + { + "epoch": 7.91, + "learning_rate": 6.277939570440481e-07, + "loss": 0.1609, + "step": 43493 + }, + { + "epoch": 7.91, + "learning_rate": 6.277393520203859e-07, + "loss": 2.4674, + "step": 43494 + }, + { + "epoch": 7.91, + "learning_rate": 6.276847469967238e-07, + "loss": 0.0178, + "step": 43495 + }, + { + "epoch": 7.91, + "learning_rate": 6.276301419730615e-07, + "loss": 0.43, + "step": 43496 + }, + { + "epoch": 7.91, + "learning_rate": 6.275755369493993e-07, + "loss": 0.0096, + "step": 43497 + }, + { + "epoch": 7.91, + "learning_rate": 6.275209319257372e-07, + "loss": 0.0529, + "step": 43498 + }, + { + "epoch": 7.91, + "learning_rate": 6.27466326902075e-07, + "loss": 0.075, + "step": 43499 + }, + { + "epoch": 7.91, + "learning_rate": 6.274117218784128e-07, + "loss": 0.3884, + "step": 43500 + }, + { + "epoch": 7.91, + "learning_rate": 6.273571168547507e-07, + "loss": 0.8287, + "step": 43501 + }, + { + "epoch": 7.91, + "learning_rate": 6.273025118310885e-07, + "loss": 0.2583, + "step": 43502 + }, + { + "epoch": 7.91, + "learning_rate": 6.272479068074263e-07, + "loss": 1.8689, + "step": 43503 + }, + { + "epoch": 7.91, + "learning_rate": 6.271933017837641e-07, + "loss": 0.2129, + "step": 43504 + }, + { + "epoch": 7.91, + "learning_rate": 6.271386967601019e-07, + "loss": 1.0946, + "step": 43505 + }, + { + "epoch": 7.91, + "learning_rate": 6.270840917364398e-07, + "loss": 0.3157, + "step": 43506 + }, + { + "epoch": 7.91, + "learning_rate": 6.270294867127776e-07, + "loss": 0.6471, + "step": 43507 + }, + { + "epoch": 7.91, + "learning_rate": 6.269748816891154e-07, + "loss": 0.221, + "step": 43508 + }, + { + "epoch": 7.91, + "learning_rate": 6.269202766654533e-07, + "loss": 0.0061, + "step": 43509 + }, + { + "epoch": 7.91, + "learning_rate": 6.26865671641791e-07, + "loss": 2.4223, + "step": 43510 + }, + { + "epoch": 7.91, + "learning_rate": 6.268110666181288e-07, + "loss": 0.5526, + "step": 43511 + }, + { + "epoch": 7.91, + "learning_rate": 6.267564615944667e-07, + "loss": 0.9763, + "step": 43512 + }, + { + "epoch": 7.91, + "learning_rate": 6.267018565708045e-07, + "loss": 0.3082, + "step": 43513 + }, + { + "epoch": 7.91, + "learning_rate": 6.266472515471424e-07, + "loss": 0.1613, + "step": 43514 + }, + { + "epoch": 7.91, + "learning_rate": 6.265926465234802e-07, + "loss": 0.3912, + "step": 43515 + }, + { + "epoch": 7.91, + "learning_rate": 6.26538041499818e-07, + "loss": 0.1926, + "step": 43516 + }, + { + "epoch": 7.91, + "learning_rate": 6.264834364761558e-07, + "loss": 1.6494, + "step": 43517 + }, + { + "epoch": 7.91, + "learning_rate": 6.264288314524937e-07, + "loss": 0.0121, + "step": 43518 + }, + { + "epoch": 7.91, + "learning_rate": 6.263742264288314e-07, + "loss": 0.0101, + "step": 43519 + }, + { + "epoch": 7.91, + "learning_rate": 6.263196214051693e-07, + "loss": 0.0095, + "step": 43520 + }, + { + "epoch": 7.91, + "learning_rate": 6.262650163815071e-07, + "loss": 0.1137, + "step": 43521 + }, + { + "epoch": 7.91, + "learning_rate": 6.26210411357845e-07, + "loss": 1.032, + "step": 43522 + }, + { + "epoch": 7.91, + "learning_rate": 6.261558063341828e-07, + "loss": 2.4702, + "step": 43523 + }, + { + "epoch": 7.91, + "learning_rate": 6.261012013105205e-07, + "loss": 0.4223, + "step": 43524 + }, + { + "epoch": 7.92, + "learning_rate": 6.260465962868585e-07, + "loss": 0.2413, + "step": 43525 + }, + { + "epoch": 7.92, + "learning_rate": 6.259919912631962e-07, + "loss": 0.191, + "step": 43526 + }, + { + "epoch": 7.92, + "learning_rate": 6.25937386239534e-07, + "loss": 0.0445, + "step": 43527 + }, + { + "epoch": 7.92, + "learning_rate": 6.258827812158719e-07, + "loss": 0.4699, + "step": 43528 + }, + { + "epoch": 7.92, + "learning_rate": 6.258281761922097e-07, + "loss": 0.1309, + "step": 43529 + }, + { + "epoch": 7.92, + "learning_rate": 6.257735711685475e-07, + "loss": 1.0481, + "step": 43530 + }, + { + "epoch": 7.92, + "learning_rate": 6.257189661448853e-07, + "loss": 0.5521, + "step": 43531 + }, + { + "epoch": 7.92, + "learning_rate": 6.256643611212232e-07, + "loss": 0.7154, + "step": 43532 + }, + { + "epoch": 7.92, + "learning_rate": 6.25609756097561e-07, + "loss": 1.6671, + "step": 43533 + }, + { + "epoch": 7.92, + "learning_rate": 6.255551510738988e-07, + "loss": 0.0461, + "step": 43534 + }, + { + "epoch": 7.92, + "learning_rate": 6.255005460502366e-07, + "loss": 1.0699, + "step": 43535 + }, + { + "epoch": 7.92, + "learning_rate": 6.254459410265745e-07, + "loss": 1.0994, + "step": 43536 + }, + { + "epoch": 7.92, + "learning_rate": 6.253913360029123e-07, + "loss": 0.0507, + "step": 43537 + }, + { + "epoch": 7.92, + "learning_rate": 6.253367309792501e-07, + "loss": 0.8195, + "step": 43538 + }, + { + "epoch": 7.92, + "learning_rate": 6.25282125955588e-07, + "loss": 0.0137, + "step": 43539 + }, + { + "epoch": 7.92, + "learning_rate": 6.252275209319257e-07, + "loss": 0.2684, + "step": 43540 + }, + { + "epoch": 7.92, + "learning_rate": 6.251729159082635e-07, + "loss": 0.8912, + "step": 43541 + }, + { + "epoch": 7.92, + "learning_rate": 6.251183108846014e-07, + "loss": 0.2702, + "step": 43542 + }, + { + "epoch": 7.92, + "learning_rate": 6.250637058609392e-07, + "loss": 0.0183, + "step": 43543 + }, + { + "epoch": 7.92, + "learning_rate": 6.250091008372771e-07, + "loss": 0.4024, + "step": 43544 + }, + { + "epoch": 7.92, + "learning_rate": 6.249544958136149e-07, + "loss": 0.0325, + "step": 43545 + }, + { + "epoch": 7.92, + "learning_rate": 6.248998907899527e-07, + "loss": 2.8737, + "step": 43546 + }, + { + "epoch": 7.92, + "learning_rate": 6.248452857662905e-07, + "loss": 0.0075, + "step": 43547 + }, + { + "epoch": 7.92, + "learning_rate": 6.247906807426283e-07, + "loss": 0.6666, + "step": 43548 + }, + { + "epoch": 7.92, + "learning_rate": 6.247360757189661e-07, + "loss": 1.4584, + "step": 43549 + }, + { + "epoch": 7.92, + "learning_rate": 6.24681470695304e-07, + "loss": 0.1494, + "step": 43550 + }, + { + "epoch": 7.92, + "learning_rate": 6.246268656716418e-07, + "loss": 0.0596, + "step": 43551 + }, + { + "epoch": 7.92, + "learning_rate": 6.245722606479797e-07, + "loss": 0.0239, + "step": 43552 + }, + { + "epoch": 7.92, + "learning_rate": 6.245176556243175e-07, + "loss": 0.1104, + "step": 43553 + }, + { + "epoch": 7.92, + "learning_rate": 6.244630506006552e-07, + "loss": 0.0057, + "step": 43554 + }, + { + "epoch": 7.92, + "learning_rate": 6.244084455769931e-07, + "loss": 0.0155, + "step": 43555 + }, + { + "epoch": 7.92, + "learning_rate": 6.243538405533309e-07, + "loss": 0.0683, + "step": 43556 + }, + { + "epoch": 7.92, + "learning_rate": 6.242992355296687e-07, + "loss": 1.8703, + "step": 43557 + }, + { + "epoch": 7.92, + "learning_rate": 6.242446305060066e-07, + "loss": 0.5308, + "step": 43558 + }, + { + "epoch": 7.92, + "learning_rate": 6.241900254823444e-07, + "loss": 1.1969, + "step": 43559 + }, + { + "epoch": 7.92, + "learning_rate": 6.241354204586822e-07, + "loss": 0.003, + "step": 43560 + }, + { + "epoch": 7.92, + "learning_rate": 6.2408081543502e-07, + "loss": 0.2444, + "step": 43561 + }, + { + "epoch": 7.92, + "learning_rate": 6.240262104113579e-07, + "loss": 0.7361, + "step": 43562 + }, + { + "epoch": 7.92, + "learning_rate": 6.239716053876957e-07, + "loss": 0.9909, + "step": 43563 + }, + { + "epoch": 7.92, + "learning_rate": 6.239170003640335e-07, + "loss": 3.7377, + "step": 43564 + }, + { + "epoch": 7.92, + "learning_rate": 6.238623953403713e-07, + "loss": 0.1058, + "step": 43565 + }, + { + "epoch": 7.92, + "learning_rate": 6.238077903167092e-07, + "loss": 0.0073, + "step": 43566 + }, + { + "epoch": 7.92, + "learning_rate": 6.23753185293047e-07, + "loss": 0.4552, + "step": 43567 + }, + { + "epoch": 7.92, + "learning_rate": 6.236985802693847e-07, + "loss": 0.3369, + "step": 43568 + }, + { + "epoch": 7.92, + "learning_rate": 6.236439752457227e-07, + "loss": 0.6442, + "step": 43569 + }, + { + "epoch": 7.92, + "learning_rate": 6.235893702220604e-07, + "loss": 0.0052, + "step": 43570 + }, + { + "epoch": 7.92, + "learning_rate": 6.235347651983982e-07, + "loss": 1.128, + "step": 43571 + }, + { + "epoch": 7.92, + "learning_rate": 6.234801601747361e-07, + "loss": 0.2657, + "step": 43572 + }, + { + "epoch": 7.92, + "learning_rate": 6.234255551510739e-07, + "loss": 0.0063, + "step": 43573 + }, + { + "epoch": 7.92, + "learning_rate": 6.233709501274118e-07, + "loss": 0.119, + "step": 43574 + }, + { + "epoch": 7.92, + "learning_rate": 6.233163451037495e-07, + "loss": 0.0299, + "step": 43575 + }, + { + "epoch": 7.92, + "learning_rate": 6.232617400800874e-07, + "loss": 2.1758, + "step": 43576 + }, + { + "epoch": 7.92, + "learning_rate": 6.232071350564252e-07, + "loss": 1.5089, + "step": 43577 + }, + { + "epoch": 7.92, + "learning_rate": 6.23152530032763e-07, + "loss": 0.0938, + "step": 43578 + }, + { + "epoch": 7.92, + "learning_rate": 6.230979250091008e-07, + "loss": 0.068, + "step": 43579 + }, + { + "epoch": 7.93, + "learning_rate": 6.230433199854387e-07, + "loss": 0.0388, + "step": 43580 + }, + { + "epoch": 7.93, + "learning_rate": 6.229887149617765e-07, + "loss": 0.4014, + "step": 43581 + }, + { + "epoch": 7.93, + "learning_rate": 6.229341099381144e-07, + "loss": 1.8055, + "step": 43582 + }, + { + "epoch": 7.93, + "learning_rate": 6.228795049144522e-07, + "loss": 0.0218, + "step": 43583 + }, + { + "epoch": 7.93, + "learning_rate": 6.228248998907899e-07, + "loss": 0.0343, + "step": 43584 + }, + { + "epoch": 7.93, + "learning_rate": 6.227702948671278e-07, + "loss": 0.9608, + "step": 43585 + }, + { + "epoch": 7.93, + "learning_rate": 6.227156898434656e-07, + "loss": 0.2269, + "step": 43586 + }, + { + "epoch": 7.93, + "learning_rate": 6.226610848198034e-07, + "loss": 1.5326, + "step": 43587 + }, + { + "epoch": 7.93, + "learning_rate": 6.226064797961413e-07, + "loss": 0.0032, + "step": 43588 + }, + { + "epoch": 7.93, + "learning_rate": 6.225518747724791e-07, + "loss": 0.761, + "step": 43589 + }, + { + "epoch": 7.93, + "learning_rate": 6.22497269748817e-07, + "loss": 0.4782, + "step": 43590 + }, + { + "epoch": 7.93, + "learning_rate": 6.224426647251547e-07, + "loss": 0.1891, + "step": 43591 + }, + { + "epoch": 7.93, + "learning_rate": 6.223880597014925e-07, + "loss": 0.4495, + "step": 43592 + }, + { + "epoch": 7.93, + "learning_rate": 6.223334546778304e-07, + "loss": 0.073, + "step": 43593 + }, + { + "epoch": 7.93, + "learning_rate": 6.222788496541682e-07, + "loss": 0.1645, + "step": 43594 + }, + { + "epoch": 7.93, + "learning_rate": 6.22224244630506e-07, + "loss": 0.2494, + "step": 43595 + }, + { + "epoch": 7.93, + "learning_rate": 6.221696396068439e-07, + "loss": 0.1013, + "step": 43596 + }, + { + "epoch": 7.93, + "learning_rate": 6.221150345831817e-07, + "loss": 0.4129, + "step": 43597 + }, + { + "epoch": 7.93, + "learning_rate": 6.220604295595194e-07, + "loss": 0.0043, + "step": 43598 + }, + { + "epoch": 7.93, + "learning_rate": 6.220058245358574e-07, + "loss": 0.4971, + "step": 43599 + }, + { + "epoch": 7.93, + "learning_rate": 6.219512195121951e-07, + "loss": 3.4575, + "step": 43600 + }, + { + "epoch": 7.93, + "learning_rate": 6.21896614488533e-07, + "loss": 0.0295, + "step": 43601 + }, + { + "epoch": 7.93, + "learning_rate": 6.218420094648708e-07, + "loss": 0.1731, + "step": 43602 + }, + { + "epoch": 7.93, + "learning_rate": 6.217874044412086e-07, + "loss": 0.8828, + "step": 43603 + }, + { + "epoch": 7.93, + "learning_rate": 6.217327994175465e-07, + "loss": 0.0981, + "step": 43604 + }, + { + "epoch": 7.93, + "learning_rate": 6.216781943938842e-07, + "loss": 0.138, + "step": 43605 + }, + { + "epoch": 7.93, + "learning_rate": 6.216235893702221e-07, + "loss": 0.1923, + "step": 43606 + }, + { + "epoch": 7.93, + "learning_rate": 6.215689843465599e-07, + "loss": 0.0056, + "step": 43607 + }, + { + "epoch": 7.93, + "learning_rate": 6.215143793228977e-07, + "loss": 0.6314, + "step": 43608 + }, + { + "epoch": 7.93, + "learning_rate": 6.214597742992355e-07, + "loss": 2.4236, + "step": 43609 + }, + { + "epoch": 7.93, + "learning_rate": 6.214051692755734e-07, + "loss": 1.0826, + "step": 43610 + }, + { + "epoch": 7.93, + "learning_rate": 6.213505642519112e-07, + "loss": 0.505, + "step": 43611 + }, + { + "epoch": 7.93, + "learning_rate": 6.21295959228249e-07, + "loss": 0.0661, + "step": 43612 + }, + { + "epoch": 7.93, + "learning_rate": 6.212413542045869e-07, + "loss": 0.3181, + "step": 43613 + }, + { + "epoch": 7.93, + "learning_rate": 6.211867491809246e-07, + "loss": 0.1485, + "step": 43614 + }, + { + "epoch": 7.93, + "learning_rate": 6.211321441572625e-07, + "loss": 0.0104, + "step": 43615 + }, + { + "epoch": 7.93, + "learning_rate": 6.210775391336003e-07, + "loss": 2.0801, + "step": 43616 + }, + { + "epoch": 7.93, + "learning_rate": 6.210229341099381e-07, + "loss": 0.5377, + "step": 43617 + }, + { + "epoch": 7.93, + "learning_rate": 6.20968329086276e-07, + "loss": 1.1166, + "step": 43618 + }, + { + "epoch": 7.93, + "learning_rate": 6.209137240626138e-07, + "loss": 0.306, + "step": 43619 + }, + { + "epoch": 7.93, + "learning_rate": 6.208591190389517e-07, + "loss": 1.2886, + "step": 43620 + }, + { + "epoch": 7.93, + "learning_rate": 6.208045140152894e-07, + "loss": 0.1553, + "step": 43621 + }, + { + "epoch": 7.93, + "learning_rate": 6.207499089916272e-07, + "loss": 0.0537, + "step": 43622 + }, + { + "epoch": 7.93, + "learning_rate": 6.206953039679651e-07, + "loss": 1.1109, + "step": 43623 + }, + { + "epoch": 7.93, + "learning_rate": 6.206406989443029e-07, + "loss": 0.0063, + "step": 43624 + }, + { + "epoch": 7.93, + "learning_rate": 6.205860939206407e-07, + "loss": 0.2944, + "step": 43625 + }, + { + "epoch": 7.93, + "learning_rate": 6.205314888969786e-07, + "loss": 0.0669, + "step": 43626 + }, + { + "epoch": 7.93, + "learning_rate": 6.204768838733164e-07, + "loss": 0.0736, + "step": 43627 + }, + { + "epoch": 7.93, + "learning_rate": 6.204222788496541e-07, + "loss": 0.0788, + "step": 43628 + }, + { + "epoch": 7.93, + "learning_rate": 6.20367673825992e-07, + "loss": 0.0171, + "step": 43629 + }, + { + "epoch": 7.93, + "learning_rate": 6.203130688023298e-07, + "loss": 0.0099, + "step": 43630 + }, + { + "epoch": 7.93, + "learning_rate": 6.202584637786677e-07, + "loss": 0.6608, + "step": 43631 + }, + { + "epoch": 7.93, + "learning_rate": 6.202038587550055e-07, + "loss": 0.0411, + "step": 43632 + }, + { + "epoch": 7.93, + "learning_rate": 6.201492537313433e-07, + "loss": 0.0562, + "step": 43633 + }, + { + "epoch": 7.93, + "learning_rate": 6.200946487076812e-07, + "loss": 0.151, + "step": 43634 + }, + { + "epoch": 7.94, + "learning_rate": 6.200400436840189e-07, + "loss": 0.0082, + "step": 43635 + }, + { + "epoch": 7.94, + "learning_rate": 6.199854386603567e-07, + "loss": 0.0048, + "step": 43636 + }, + { + "epoch": 7.94, + "learning_rate": 6.199308336366946e-07, + "loss": 0.3744, + "step": 43637 + }, + { + "epoch": 7.94, + "learning_rate": 6.198762286130324e-07, + "loss": 0.0291, + "step": 43638 + }, + { + "epoch": 7.94, + "learning_rate": 6.198216235893702e-07, + "loss": 0.9278, + "step": 43639 + }, + { + "epoch": 7.94, + "learning_rate": 6.197670185657081e-07, + "loss": 0.6868, + "step": 43640 + }, + { + "epoch": 7.94, + "learning_rate": 6.197124135420459e-07, + "loss": 1.002, + "step": 43641 + }, + { + "epoch": 7.94, + "learning_rate": 6.196578085183837e-07, + "loss": 0.0119, + "step": 43642 + }, + { + "epoch": 7.94, + "learning_rate": 6.196032034947216e-07, + "loss": 0.859, + "step": 43643 + }, + { + "epoch": 7.94, + "learning_rate": 6.195485984710593e-07, + "loss": 0.1424, + "step": 43644 + }, + { + "epoch": 7.94, + "learning_rate": 6.194939934473972e-07, + "loss": 0.0063, + "step": 43645 + }, + { + "epoch": 7.94, + "learning_rate": 6.19439388423735e-07, + "loss": 0.33, + "step": 43646 + }, + { + "epoch": 7.94, + "learning_rate": 6.193847834000728e-07, + "loss": 0.0051, + "step": 43647 + }, + { + "epoch": 7.94, + "learning_rate": 6.193301783764107e-07, + "loss": 0.0246, + "step": 43648 + }, + { + "epoch": 7.94, + "learning_rate": 6.192755733527484e-07, + "loss": 0.8562, + "step": 43649 + }, + { + "epoch": 7.94, + "learning_rate": 6.192209683290864e-07, + "loss": 0.0025, + "step": 43650 + }, + { + "epoch": 7.94, + "learning_rate": 6.191663633054241e-07, + "loss": 1.2003, + "step": 43651 + }, + { + "epoch": 7.94, + "learning_rate": 6.191117582817619e-07, + "loss": 2.0221, + "step": 43652 + }, + { + "epoch": 7.94, + "learning_rate": 6.190571532580998e-07, + "loss": 0.1608, + "step": 43653 + }, + { + "epoch": 7.94, + "learning_rate": 6.190025482344376e-07, + "loss": 1.0885, + "step": 43654 + }, + { + "epoch": 7.94, + "learning_rate": 6.189479432107754e-07, + "loss": 0.0906, + "step": 43655 + }, + { + "epoch": 7.94, + "learning_rate": 6.188933381871132e-07, + "loss": 0.279, + "step": 43656 + }, + { + "epoch": 7.94, + "learning_rate": 6.188387331634511e-07, + "loss": 0.2011, + "step": 43657 + }, + { + "epoch": 7.94, + "learning_rate": 6.187841281397888e-07, + "loss": 0.2572, + "step": 43658 + }, + { + "epoch": 7.94, + "learning_rate": 6.187295231161267e-07, + "loss": 0.1022, + "step": 43659 + }, + { + "epoch": 7.94, + "learning_rate": 6.186749180924645e-07, + "loss": 0.4288, + "step": 43660 + }, + { + "epoch": 7.94, + "learning_rate": 6.186203130688024e-07, + "loss": 1.2689, + "step": 43661 + }, + { + "epoch": 7.94, + "learning_rate": 6.185657080451402e-07, + "loss": 0.1641, + "step": 43662 + }, + { + "epoch": 7.94, + "learning_rate": 6.18511103021478e-07, + "loss": 3.2069, + "step": 43663 + }, + { + "epoch": 7.94, + "learning_rate": 6.184564979978159e-07, + "loss": 0.0126, + "step": 43664 + }, + { + "epoch": 7.94, + "learning_rate": 6.184018929741536e-07, + "loss": 2.3189, + "step": 43665 + }, + { + "epoch": 7.94, + "learning_rate": 6.183472879504914e-07, + "loss": 4.7903, + "step": 43666 + }, + { + "epoch": 7.94, + "learning_rate": 6.182926829268293e-07, + "loss": 1.0757, + "step": 43667 + }, + { + "epoch": 7.94, + "learning_rate": 6.182380779031671e-07, + "loss": 1.5174, + "step": 43668 + }, + { + "epoch": 7.94, + "learning_rate": 6.18183472879505e-07, + "loss": 1.4758, + "step": 43669 + }, + { + "epoch": 7.94, + "learning_rate": 6.181288678558428e-07, + "loss": 0.0483, + "step": 43670 + }, + { + "epoch": 7.94, + "learning_rate": 6.180742628321806e-07, + "loss": 1.3888, + "step": 43671 + }, + { + "epoch": 7.94, + "learning_rate": 6.180196578085184e-07, + "loss": 3.1388, + "step": 43672 + }, + { + "epoch": 7.94, + "learning_rate": 6.179650527848562e-07, + "loss": 0.3911, + "step": 43673 + }, + { + "epoch": 7.94, + "learning_rate": 6.17910447761194e-07, + "loss": 0.7385, + "step": 43674 + }, + { + "epoch": 7.94, + "learning_rate": 6.178558427375319e-07, + "loss": 0.0181, + "step": 43675 + }, + { + "epoch": 7.94, + "learning_rate": 6.178012377138697e-07, + "loss": 0.2018, + "step": 43676 + }, + { + "epoch": 7.94, + "learning_rate": 6.177466326902075e-07, + "loss": 0.0031, + "step": 43677 + }, + { + "epoch": 7.94, + "learning_rate": 6.176920276665454e-07, + "loss": 0.0882, + "step": 43678 + }, + { + "epoch": 7.94, + "learning_rate": 6.176374226428831e-07, + "loss": 0.0751, + "step": 43679 + }, + { + "epoch": 7.94, + "learning_rate": 6.17582817619221e-07, + "loss": 1.3973, + "step": 43680 + }, + { + "epoch": 7.94, + "learning_rate": 6.175282125955588e-07, + "loss": 1.7186, + "step": 43681 + }, + { + "epoch": 7.94, + "learning_rate": 6.174736075718966e-07, + "loss": 0.0634, + "step": 43682 + }, + { + "epoch": 7.94, + "learning_rate": 6.174190025482345e-07, + "loss": 0.0516, + "step": 43683 + }, + { + "epoch": 7.94, + "learning_rate": 6.173643975245723e-07, + "loss": 0.1693, + "step": 43684 + }, + { + "epoch": 7.94, + "learning_rate": 6.173097925009101e-07, + "loss": 0.0026, + "step": 43685 + }, + { + "epoch": 7.94, + "learning_rate": 6.172551874772479e-07, + "loss": 0.1982, + "step": 43686 + }, + { + "epoch": 7.94, + "learning_rate": 6.172005824535858e-07, + "loss": 0.2358, + "step": 43687 + }, + { + "epoch": 7.94, + "learning_rate": 6.171459774299235e-07, + "loss": 0.2204, + "step": 43688 + }, + { + "epoch": 7.94, + "learning_rate": 6.170913724062614e-07, + "loss": 0.1196, + "step": 43689 + }, + { + "epoch": 7.95, + "learning_rate": 6.170367673825992e-07, + "loss": 0.0092, + "step": 43690 + }, + { + "epoch": 7.95, + "learning_rate": 6.169821623589371e-07, + "loss": 0.8263, + "step": 43691 + }, + { + "epoch": 7.95, + "learning_rate": 6.169275573352749e-07, + "loss": 0.0157, + "step": 43692 + }, + { + "epoch": 7.95, + "learning_rate": 6.168729523116126e-07, + "loss": 0.2041, + "step": 43693 + }, + { + "epoch": 7.95, + "learning_rate": 6.168183472879506e-07, + "loss": 0.0066, + "step": 43694 + }, + { + "epoch": 7.95, + "learning_rate": 6.167637422642883e-07, + "loss": 0.1227, + "step": 43695 + }, + { + "epoch": 7.95, + "learning_rate": 6.167091372406261e-07, + "loss": 0.0983, + "step": 43696 + }, + { + "epoch": 7.95, + "learning_rate": 6.16654532216964e-07, + "loss": 1.2906, + "step": 43697 + }, + { + "epoch": 7.95, + "learning_rate": 6.165999271933018e-07, + "loss": 0.0027, + "step": 43698 + }, + { + "epoch": 7.95, + "learning_rate": 6.165453221696397e-07, + "loss": 0.5031, + "step": 43699 + }, + { + "epoch": 7.95, + "learning_rate": 6.164907171459774e-07, + "loss": 0.5888, + "step": 43700 + }, + { + "epoch": 7.95, + "learning_rate": 6.164361121223153e-07, + "loss": 0.1446, + "step": 43701 + }, + { + "epoch": 7.95, + "learning_rate": 6.163815070986531e-07, + "loss": 1.5777, + "step": 43702 + }, + { + "epoch": 7.95, + "learning_rate": 6.163269020749909e-07, + "loss": 0.5888, + "step": 43703 + }, + { + "epoch": 7.95, + "learning_rate": 6.162722970513287e-07, + "loss": 0.7201, + "step": 43704 + }, + { + "epoch": 7.95, + "learning_rate": 6.162176920276666e-07, + "loss": 1.4281, + "step": 43705 + }, + { + "epoch": 7.95, + "learning_rate": 6.161630870040044e-07, + "loss": 0.0657, + "step": 43706 + }, + { + "epoch": 7.95, + "learning_rate": 6.161084819803422e-07, + "loss": 2.1943, + "step": 43707 + }, + { + "epoch": 7.95, + "learning_rate": 6.160538769566801e-07, + "loss": 0.1025, + "step": 43708 + }, + { + "epoch": 7.95, + "learning_rate": 6.159992719330178e-07, + "loss": 0.0524, + "step": 43709 + }, + { + "epoch": 7.95, + "learning_rate": 6.159446669093557e-07, + "loss": 0.8252, + "step": 43710 + }, + { + "epoch": 7.95, + "learning_rate": 6.158900618856935e-07, + "loss": 1.6374, + "step": 43711 + }, + { + "epoch": 7.95, + "learning_rate": 6.158354568620313e-07, + "loss": 0.8879, + "step": 43712 + }, + { + "epoch": 7.95, + "learning_rate": 6.157808518383692e-07, + "loss": 0.2072, + "step": 43713 + }, + { + "epoch": 7.95, + "learning_rate": 6.15726246814707e-07, + "loss": 0.2013, + "step": 43714 + }, + { + "epoch": 7.95, + "learning_rate": 6.156716417910448e-07, + "loss": 0.0261, + "step": 43715 + }, + { + "epoch": 7.95, + "learning_rate": 6.156170367673826e-07, + "loss": 0.047, + "step": 43716 + }, + { + "epoch": 7.95, + "learning_rate": 6.155624317437204e-07, + "loss": 0.007, + "step": 43717 + }, + { + "epoch": 7.95, + "learning_rate": 6.155078267200582e-07, + "loss": 0.0404, + "step": 43718 + }, + { + "epoch": 7.95, + "learning_rate": 6.154532216963961e-07, + "loss": 0.0452, + "step": 43719 + }, + { + "epoch": 7.95, + "learning_rate": 6.153986166727338e-07, + "loss": 0.0168, + "step": 43720 + }, + { + "epoch": 7.95, + "learning_rate": 6.153440116490718e-07, + "loss": 0.0577, + "step": 43721 + }, + { + "epoch": 7.95, + "learning_rate": 6.152894066254096e-07, + "loss": 0.074, + "step": 43722 + }, + { + "epoch": 7.95, + "learning_rate": 6.152348016017473e-07, + "loss": 0.0233, + "step": 43723 + }, + { + "epoch": 7.95, + "learning_rate": 6.151801965780852e-07, + "loss": 0.4039, + "step": 43724 + }, + { + "epoch": 7.95, + "learning_rate": 6.15125591554423e-07, + "loss": 0.3235, + "step": 43725 + }, + { + "epoch": 7.95, + "learning_rate": 6.150709865307608e-07, + "loss": 0.3951, + "step": 43726 + }, + { + "epoch": 7.95, + "learning_rate": 6.150163815070987e-07, + "loss": 0.017, + "step": 43727 + }, + { + "epoch": 7.95, + "learning_rate": 6.149617764834365e-07, + "loss": 1.7852, + "step": 43728 + }, + { + "epoch": 7.95, + "learning_rate": 6.149071714597744e-07, + "loss": 0.8142, + "step": 43729 + }, + { + "epoch": 7.95, + "learning_rate": 6.148525664361121e-07, + "loss": 2.313, + "step": 43730 + }, + { + "epoch": 7.95, + "learning_rate": 6.1479796141245e-07, + "loss": 0.0069, + "step": 43731 + }, + { + "epoch": 7.95, + "learning_rate": 6.147433563887878e-07, + "loss": 0.1536, + "step": 43732 + }, + { + "epoch": 7.95, + "learning_rate": 6.146887513651256e-07, + "loss": 0.1231, + "step": 43733 + }, + { + "epoch": 7.95, + "learning_rate": 6.146341463414634e-07, + "loss": 0.0698, + "step": 43734 + }, + { + "epoch": 7.95, + "learning_rate": 6.145795413178013e-07, + "loss": 0.1366, + "step": 43735 + }, + { + "epoch": 7.95, + "learning_rate": 6.145249362941391e-07, + "loss": 0.0356, + "step": 43736 + }, + { + "epoch": 7.95, + "learning_rate": 6.144703312704768e-07, + "loss": 0.0907, + "step": 43737 + }, + { + "epoch": 7.95, + "learning_rate": 6.144157262468148e-07, + "loss": 0.0976, + "step": 43738 + }, + { + "epoch": 7.95, + "learning_rate": 6.143611212231525e-07, + "loss": 0.4534, + "step": 43739 + }, + { + "epoch": 7.95, + "learning_rate": 6.143065161994904e-07, + "loss": 0.2782, + "step": 43740 + }, + { + "epoch": 7.95, + "learning_rate": 6.142519111758282e-07, + "loss": 0.1918, + "step": 43741 + }, + { + "epoch": 7.95, + "learning_rate": 6.14197306152166e-07, + "loss": 0.1902, + "step": 43742 + }, + { + "epoch": 7.95, + "learning_rate": 6.141427011285039e-07, + "loss": 0.1363, + "step": 43743 + }, + { + "epoch": 7.95, + "learning_rate": 6.140880961048416e-07, + "loss": 0.0061, + "step": 43744 + }, + { + "epoch": 7.96, + "learning_rate": 6.140334910811795e-07, + "loss": 0.7709, + "step": 43745 + }, + { + "epoch": 7.96, + "learning_rate": 6.139788860575173e-07, + "loss": 3.2152, + "step": 43746 + }, + { + "epoch": 7.96, + "learning_rate": 6.139242810338551e-07, + "loss": 0.7154, + "step": 43747 + }, + { + "epoch": 7.96, + "learning_rate": 6.13869676010193e-07, + "loss": 0.4246, + "step": 43748 + }, + { + "epoch": 7.96, + "learning_rate": 6.138150709865308e-07, + "loss": 3.0383, + "step": 43749 + }, + { + "epoch": 7.96, + "learning_rate": 6.137604659628686e-07, + "loss": 0.2531, + "step": 43750 + }, + { + "epoch": 7.96, + "learning_rate": 6.137058609392065e-07, + "loss": 0.0481, + "step": 43751 + }, + { + "epoch": 7.96, + "learning_rate": 6.136512559155443e-07, + "loss": 1.0087, + "step": 43752 + }, + { + "epoch": 7.96, + "learning_rate": 6.13596650891882e-07, + "loss": 0.0291, + "step": 43753 + }, + { + "epoch": 7.96, + "learning_rate": 6.135420458682199e-07, + "loss": 1.6657, + "step": 43754 + }, + { + "epoch": 7.96, + "learning_rate": 6.134874408445577e-07, + "loss": 0.0473, + "step": 43755 + }, + { + "epoch": 7.96, + "learning_rate": 6.134328358208955e-07, + "loss": 0.068, + "step": 43756 + }, + { + "epoch": 7.96, + "learning_rate": 6.133782307972334e-07, + "loss": 0.1346, + "step": 43757 + }, + { + "epoch": 7.96, + "learning_rate": 6.133236257735712e-07, + "loss": 0.1131, + "step": 43758 + }, + { + "epoch": 7.96, + "learning_rate": 6.132690207499091e-07, + "loss": 0.0073, + "step": 43759 + }, + { + "epoch": 7.96, + "learning_rate": 6.132144157262468e-07, + "loss": 0.007, + "step": 43760 + }, + { + "epoch": 7.96, + "learning_rate": 6.131598107025846e-07, + "loss": 0.2339, + "step": 43761 + }, + { + "epoch": 7.96, + "learning_rate": 6.131052056789225e-07, + "loss": 0.1434, + "step": 43762 + }, + { + "epoch": 7.96, + "learning_rate": 6.130506006552603e-07, + "loss": 0.7157, + "step": 43763 + }, + { + "epoch": 7.96, + "learning_rate": 6.12995995631598e-07, + "loss": 0.7148, + "step": 43764 + }, + { + "epoch": 7.96, + "learning_rate": 6.12941390607936e-07, + "loss": 0.8942, + "step": 43765 + }, + { + "epoch": 7.96, + "learning_rate": 6.128867855842738e-07, + "loss": 2.1475, + "step": 43766 + }, + { + "epoch": 7.96, + "learning_rate": 6.128321805606115e-07, + "loss": 0.6286, + "step": 43767 + }, + { + "epoch": 7.96, + "learning_rate": 6.127775755369495e-07, + "loss": 2.2189, + "step": 43768 + }, + { + "epoch": 7.96, + "learning_rate": 6.127229705132872e-07, + "loss": 1.6109, + "step": 43769 + }, + { + "epoch": 7.96, + "learning_rate": 6.126683654896251e-07, + "loss": 2.5234, + "step": 43770 + }, + { + "epoch": 7.96, + "learning_rate": 6.126137604659629e-07, + "loss": 0.3107, + "step": 43771 + }, + { + "epoch": 7.96, + "learning_rate": 6.125591554423007e-07, + "loss": 0.026, + "step": 43772 + }, + { + "epoch": 7.96, + "learning_rate": 6.125045504186386e-07, + "loss": 0.1997, + "step": 43773 + }, + { + "epoch": 7.96, + "learning_rate": 6.124499453949763e-07, + "loss": 0.0103, + "step": 43774 + }, + { + "epoch": 7.96, + "learning_rate": 6.123953403713142e-07, + "loss": 0.1691, + "step": 43775 + }, + { + "epoch": 7.96, + "learning_rate": 6.12340735347652e-07, + "loss": 0.136, + "step": 43776 + }, + { + "epoch": 7.96, + "learning_rate": 6.122861303239898e-07, + "loss": 0.3411, + "step": 43777 + }, + { + "epoch": 7.96, + "learning_rate": 6.122315253003277e-07, + "loss": 0.1839, + "step": 43778 + }, + { + "epoch": 7.96, + "learning_rate": 6.121769202766655e-07, + "loss": 0.4713, + "step": 43779 + }, + { + "epoch": 7.96, + "learning_rate": 6.121223152530033e-07, + "loss": 2.6324, + "step": 43780 + }, + { + "epoch": 7.96, + "learning_rate": 6.120677102293411e-07, + "loss": 0.5953, + "step": 43781 + }, + { + "epoch": 7.96, + "learning_rate": 6.12013105205679e-07, + "loss": 0.6776, + "step": 43782 + }, + { + "epoch": 7.96, + "learning_rate": 6.119585001820167e-07, + "loss": 0.1125, + "step": 43783 + }, + { + "epoch": 7.96, + "learning_rate": 6.119038951583546e-07, + "loss": 0.8387, + "step": 43784 + }, + { + "epoch": 7.96, + "learning_rate": 6.118492901346924e-07, + "loss": 1.3898, + "step": 43785 + }, + { + "epoch": 7.96, + "learning_rate": 6.117946851110302e-07, + "loss": 0.3425, + "step": 43786 + }, + { + "epoch": 7.96, + "learning_rate": 6.117400800873681e-07, + "loss": 2.2493, + "step": 43787 + }, + { + "epoch": 7.96, + "learning_rate": 6.116854750637059e-07, + "loss": 0.1529, + "step": 43788 + }, + { + "epoch": 7.96, + "learning_rate": 6.116308700400438e-07, + "loss": 0.1491, + "step": 43789 + }, + { + "epoch": 7.96, + "learning_rate": 6.115762650163815e-07, + "loss": 0.0404, + "step": 43790 + }, + { + "epoch": 7.96, + "learning_rate": 6.115216599927193e-07, + "loss": 2.825, + "step": 43791 + }, + { + "epoch": 7.96, + "learning_rate": 6.114670549690572e-07, + "loss": 1.6448, + "step": 43792 + }, + { + "epoch": 7.96, + "learning_rate": 6.11412449945395e-07, + "loss": 0.0844, + "step": 43793 + }, + { + "epoch": 7.96, + "learning_rate": 6.113578449217328e-07, + "loss": 0.0779, + "step": 43794 + }, + { + "epoch": 7.96, + "learning_rate": 6.113032398980707e-07, + "loss": 0.0755, + "step": 43795 + }, + { + "epoch": 7.96, + "learning_rate": 6.112486348744085e-07, + "loss": 0.0251, + "step": 43796 + }, + { + "epoch": 7.96, + "learning_rate": 6.111940298507462e-07, + "loss": 0.006, + "step": 43797 + }, + { + "epoch": 7.96, + "learning_rate": 6.111394248270841e-07, + "loss": 0.0442, + "step": 43798 + }, + { + "epoch": 7.96, + "learning_rate": 6.110848198034219e-07, + "loss": 0.571, + "step": 43799 + }, + { + "epoch": 7.97, + "learning_rate": 6.110302147797598e-07, + "loss": 1.1254, + "step": 43800 + }, + { + "epoch": 7.97, + "learning_rate": 6.109756097560975e-07, + "loss": 0.013, + "step": 43801 + }, + { + "epoch": 7.97, + "learning_rate": 6.109210047324354e-07, + "loss": 0.8601, + "step": 43802 + }, + { + "epoch": 7.97, + "learning_rate": 6.108663997087733e-07, + "loss": 1.9448, + "step": 43803 + }, + { + "epoch": 7.97, + "learning_rate": 6.10811794685111e-07, + "loss": 1.7496, + "step": 43804 + }, + { + "epoch": 7.97, + "learning_rate": 6.107571896614488e-07, + "loss": 0.1791, + "step": 43805 + }, + { + "epoch": 7.97, + "learning_rate": 6.107025846377867e-07, + "loss": 0.5372, + "step": 43806 + }, + { + "epoch": 7.97, + "learning_rate": 6.106479796141245e-07, + "loss": 0.007, + "step": 43807 + }, + { + "epoch": 7.97, + "learning_rate": 6.105933745904622e-07, + "loss": 0.23, + "step": 43808 + }, + { + "epoch": 7.97, + "learning_rate": 6.105387695668002e-07, + "loss": 0.0485, + "step": 43809 + }, + { + "epoch": 7.97, + "learning_rate": 6.10484164543138e-07, + "loss": 0.8448, + "step": 43810 + }, + { + "epoch": 7.97, + "learning_rate": 6.104295595194758e-07, + "loss": 0.0438, + "step": 43811 + }, + { + "epoch": 7.97, + "learning_rate": 6.103749544958137e-07, + "loss": 0.1664, + "step": 43812 + }, + { + "epoch": 7.97, + "learning_rate": 6.103203494721514e-07, + "loss": 0.0043, + "step": 43813 + }, + { + "epoch": 7.97, + "learning_rate": 6.102657444484893e-07, + "loss": 0.4431, + "step": 43814 + }, + { + "epoch": 7.97, + "learning_rate": 6.102111394248271e-07, + "loss": 0.0627, + "step": 43815 + }, + { + "epoch": 7.97, + "learning_rate": 6.10156534401165e-07, + "loss": 0.0131, + "step": 43816 + }, + { + "epoch": 7.97, + "learning_rate": 6.101019293775028e-07, + "loss": 0.6693, + "step": 43817 + }, + { + "epoch": 7.97, + "learning_rate": 6.100473243538405e-07, + "loss": 0.1519, + "step": 43818 + }, + { + "epoch": 7.97, + "learning_rate": 6.099927193301785e-07, + "loss": 1.6342, + "step": 43819 + }, + { + "epoch": 7.97, + "learning_rate": 6.099381143065162e-07, + "loss": 0.1697, + "step": 43820 + }, + { + "epoch": 7.97, + "learning_rate": 6.09883509282854e-07, + "loss": 0.6854, + "step": 43821 + }, + { + "epoch": 7.97, + "learning_rate": 6.098289042591919e-07, + "loss": 1.0419, + "step": 43822 + }, + { + "epoch": 7.97, + "learning_rate": 6.097742992355297e-07, + "loss": 1.4067, + "step": 43823 + }, + { + "epoch": 7.97, + "learning_rate": 6.097196942118675e-07, + "loss": 0.205, + "step": 43824 + }, + { + "epoch": 7.97, + "learning_rate": 6.096650891882053e-07, + "loss": 0.0072, + "step": 43825 + }, + { + "epoch": 7.97, + "learning_rate": 6.096104841645432e-07, + "loss": 0.1218, + "step": 43826 + }, + { + "epoch": 7.97, + "learning_rate": 6.09555879140881e-07, + "loss": 0.6398, + "step": 43827 + }, + { + "epoch": 7.97, + "learning_rate": 6.095012741172188e-07, + "loss": 0.4406, + "step": 43828 + }, + { + "epoch": 7.97, + "learning_rate": 6.094466690935566e-07, + "loss": 1.1977, + "step": 43829 + }, + { + "epoch": 7.97, + "learning_rate": 6.093920640698945e-07, + "loss": 0.0566, + "step": 43830 + }, + { + "epoch": 7.97, + "learning_rate": 6.093374590462323e-07, + "loss": 3.2589, + "step": 43831 + }, + { + "epoch": 7.97, + "learning_rate": 6.092828540225701e-07, + "loss": 0.2528, + "step": 43832 + }, + { + "epoch": 7.97, + "learning_rate": 6.09228248998908e-07, + "loss": 0.4999, + "step": 43833 + }, + { + "epoch": 7.97, + "learning_rate": 6.091736439752457e-07, + "loss": 0.3383, + "step": 43834 + }, + { + "epoch": 7.97, + "learning_rate": 6.091190389515835e-07, + "loss": 0.983, + "step": 43835 + }, + { + "epoch": 7.97, + "learning_rate": 6.090644339279214e-07, + "loss": 0.0643, + "step": 43836 + }, + { + "epoch": 7.97, + "learning_rate": 6.090098289042592e-07, + "loss": 2.2053, + "step": 43837 + }, + { + "epoch": 7.97, + "learning_rate": 6.08955223880597e-07, + "loss": 0.9249, + "step": 43838 + }, + { + "epoch": 7.97, + "learning_rate": 6.089006188569349e-07, + "loss": 0.2219, + "step": 43839 + }, + { + "epoch": 7.97, + "learning_rate": 6.088460138332727e-07, + "loss": 1.0028, + "step": 43840 + }, + { + "epoch": 7.97, + "learning_rate": 6.087914088096105e-07, + "loss": 0.0323, + "step": 43841 + }, + { + "epoch": 7.97, + "learning_rate": 6.087368037859483e-07, + "loss": 0.5264, + "step": 43842 + }, + { + "epoch": 7.97, + "learning_rate": 6.086821987622861e-07, + "loss": 0.5047, + "step": 43843 + }, + { + "epoch": 7.97, + "learning_rate": 6.08627593738624e-07, + "loss": 0.11, + "step": 43844 + }, + { + "epoch": 7.97, + "learning_rate": 6.085729887149617e-07, + "loss": 1.1316, + "step": 43845 + }, + { + "epoch": 7.97, + "learning_rate": 6.085183836912996e-07, + "loss": 0.2137, + "step": 43846 + }, + { + "epoch": 7.97, + "learning_rate": 6.084637786676375e-07, + "loss": 0.0082, + "step": 43847 + }, + { + "epoch": 7.97, + "learning_rate": 6.084091736439752e-07, + "loss": 1.9978, + "step": 43848 + }, + { + "epoch": 7.97, + "learning_rate": 6.083545686203131e-07, + "loss": 0.1306, + "step": 43849 + }, + { + "epoch": 7.97, + "learning_rate": 6.082999635966509e-07, + "loss": 2.748, + "step": 43850 + }, + { + "epoch": 7.97, + "learning_rate": 6.082453585729887e-07, + "loss": 1.1887, + "step": 43851 + }, + { + "epoch": 7.97, + "learning_rate": 6.081907535493266e-07, + "loss": 0.0486, + "step": 43852 + }, + { + "epoch": 7.97, + "learning_rate": 6.081361485256644e-07, + "loss": 0.5948, + "step": 43853 + }, + { + "epoch": 7.97, + "learning_rate": 6.080815435020022e-07, + "loss": 0.2398, + "step": 43854 + }, + { + "epoch": 7.98, + "learning_rate": 6.0802693847834e-07, + "loss": 0.7076, + "step": 43855 + }, + { + "epoch": 7.98, + "learning_rate": 6.079723334546779e-07, + "loss": 0.0048, + "step": 43856 + }, + { + "epoch": 7.98, + "learning_rate": 6.079177284310157e-07, + "loss": 0.1342, + "step": 43857 + }, + { + "epoch": 7.98, + "learning_rate": 6.078631234073535e-07, + "loss": 0.0588, + "step": 43858 + }, + { + "epoch": 7.98, + "learning_rate": 6.078085183836913e-07, + "loss": 0.1324, + "step": 43859 + }, + { + "epoch": 7.98, + "learning_rate": 6.077539133600292e-07, + "loss": 0.0823, + "step": 43860 + }, + { + "epoch": 7.98, + "learning_rate": 6.07699308336367e-07, + "loss": 0.0061, + "step": 43861 + }, + { + "epoch": 7.98, + "learning_rate": 6.076447033127047e-07, + "loss": 0.1659, + "step": 43862 + }, + { + "epoch": 7.98, + "learning_rate": 6.075900982890427e-07, + "loss": 0.3835, + "step": 43863 + }, + { + "epoch": 7.98, + "learning_rate": 6.075354932653804e-07, + "loss": 0.0367, + "step": 43864 + }, + { + "epoch": 7.98, + "learning_rate": 6.074808882417182e-07, + "loss": 1.6451, + "step": 43865 + }, + { + "epoch": 7.98, + "learning_rate": 6.074262832180561e-07, + "loss": 0.0104, + "step": 43866 + }, + { + "epoch": 7.98, + "learning_rate": 6.073716781943939e-07, + "loss": 1.3603, + "step": 43867 + }, + { + "epoch": 7.98, + "learning_rate": 6.073170731707318e-07, + "loss": 0.9983, + "step": 43868 + }, + { + "epoch": 7.98, + "learning_rate": 6.072624681470695e-07, + "loss": 1.8689, + "step": 43869 + }, + { + "epoch": 7.98, + "learning_rate": 6.072078631234074e-07, + "loss": 1.5738, + "step": 43870 + }, + { + "epoch": 7.98, + "learning_rate": 6.071532580997452e-07, + "loss": 0.4974, + "step": 43871 + }, + { + "epoch": 7.98, + "learning_rate": 6.07098653076083e-07, + "loss": 0.0042, + "step": 43872 + }, + { + "epoch": 7.98, + "learning_rate": 6.070440480524208e-07, + "loss": 0.3217, + "step": 43873 + }, + { + "epoch": 7.98, + "learning_rate": 6.069894430287587e-07, + "loss": 0.5809, + "step": 43874 + }, + { + "epoch": 7.98, + "learning_rate": 6.069348380050964e-07, + "loss": 0.0102, + "step": 43875 + }, + { + "epoch": 7.98, + "learning_rate": 6.068802329814344e-07, + "loss": 3.1138, + "step": 43876 + }, + { + "epoch": 7.98, + "learning_rate": 6.068256279577722e-07, + "loss": 0.8787, + "step": 43877 + }, + { + "epoch": 7.98, + "learning_rate": 6.067710229341099e-07, + "loss": 0.3932, + "step": 43878 + }, + { + "epoch": 7.98, + "learning_rate": 6.067164179104478e-07, + "loss": 2.8206, + "step": 43879 + }, + { + "epoch": 7.98, + "learning_rate": 6.066618128867856e-07, + "loss": 1.4766, + "step": 43880 + }, + { + "epoch": 7.98, + "learning_rate": 6.066072078631234e-07, + "loss": 0.0739, + "step": 43881 + }, + { + "epoch": 7.98, + "learning_rate": 6.065526028394612e-07, + "loss": 0.1261, + "step": 43882 + }, + { + "epoch": 7.98, + "learning_rate": 6.064979978157991e-07, + "loss": 0.0969, + "step": 43883 + }, + { + "epoch": 7.98, + "learning_rate": 6.064433927921369e-07, + "loss": 0.9505, + "step": 43884 + }, + { + "epoch": 7.98, + "learning_rate": 6.063887877684747e-07, + "loss": 0.4496, + "step": 43885 + }, + { + "epoch": 7.98, + "learning_rate": 6.063341827448125e-07, + "loss": 0.618, + "step": 43886 + }, + { + "epoch": 7.98, + "learning_rate": 6.062795777211504e-07, + "loss": 0.0248, + "step": 43887 + }, + { + "epoch": 7.98, + "learning_rate": 6.062249726974882e-07, + "loss": 1.2727, + "step": 43888 + }, + { + "epoch": 7.98, + "learning_rate": 6.061703676738259e-07, + "loss": 0.0444, + "step": 43889 + }, + { + "epoch": 7.98, + "learning_rate": 6.061157626501639e-07, + "loss": 0.0232, + "step": 43890 + }, + { + "epoch": 7.98, + "learning_rate": 6.060611576265017e-07, + "loss": 0.8281, + "step": 43891 + }, + { + "epoch": 7.98, + "learning_rate": 6.060065526028394e-07, + "loss": 0.017, + "step": 43892 + }, + { + "epoch": 7.98, + "learning_rate": 6.059519475791773e-07, + "loss": 0.0818, + "step": 43893 + }, + { + "epoch": 7.98, + "learning_rate": 6.058973425555151e-07, + "loss": 0.0064, + "step": 43894 + }, + { + "epoch": 7.98, + "learning_rate": 6.058427375318529e-07, + "loss": 0.9065, + "step": 43895 + }, + { + "epoch": 7.98, + "learning_rate": 6.057881325081908e-07, + "loss": 1.5147, + "step": 43896 + }, + { + "epoch": 7.98, + "learning_rate": 6.057335274845286e-07, + "loss": 0.2003, + "step": 43897 + }, + { + "epoch": 7.98, + "learning_rate": 6.056789224608665e-07, + "loss": 0.1459, + "step": 43898 + }, + { + "epoch": 7.98, + "learning_rate": 6.056243174372042e-07, + "loss": 0.9613, + "step": 43899 + }, + { + "epoch": 7.98, + "learning_rate": 6.055697124135421e-07, + "loss": 3.838, + "step": 43900 + }, + { + "epoch": 7.98, + "learning_rate": 6.055151073898799e-07, + "loss": 0.9083, + "step": 43901 + }, + { + "epoch": 7.98, + "learning_rate": 6.054605023662177e-07, + "loss": 0.0067, + "step": 43902 + }, + { + "epoch": 7.98, + "learning_rate": 6.054058973425555e-07, + "loss": 1.6434, + "step": 43903 + }, + { + "epoch": 7.98, + "learning_rate": 6.053512923188934e-07, + "loss": 0.2232, + "step": 43904 + }, + { + "epoch": 7.98, + "learning_rate": 6.052966872952312e-07, + "loss": 0.1075, + "step": 43905 + }, + { + "epoch": 7.98, + "learning_rate": 6.05242082271569e-07, + "loss": 0.9052, + "step": 43906 + }, + { + "epoch": 7.98, + "learning_rate": 6.051874772479069e-07, + "loss": 3.0832, + "step": 43907 + }, + { + "epoch": 7.98, + "learning_rate": 6.051328722242446e-07, + "loss": 1.5859, + "step": 43908 + }, + { + "epoch": 7.98, + "learning_rate": 6.050782672005825e-07, + "loss": 0.4425, + "step": 43909 + }, + { + "epoch": 7.99, + "learning_rate": 6.050236621769203e-07, + "loss": 0.0139, + "step": 43910 + }, + { + "epoch": 7.99, + "learning_rate": 6.049690571532581e-07, + "loss": 0.0073, + "step": 43911 + }, + { + "epoch": 7.99, + "learning_rate": 6.04914452129596e-07, + "loss": 0.0072, + "step": 43912 + }, + { + "epoch": 7.99, + "learning_rate": 6.048598471059337e-07, + "loss": 0.0462, + "step": 43913 + }, + { + "epoch": 7.99, + "learning_rate": 6.048052420822716e-07, + "loss": 1.6171, + "step": 43914 + }, + { + "epoch": 7.99, + "learning_rate": 6.047506370586094e-07, + "loss": 0.1601, + "step": 43915 + }, + { + "epoch": 7.99, + "learning_rate": 6.046960320349472e-07, + "loss": 0.9744, + "step": 43916 + }, + { + "epoch": 7.99, + "learning_rate": 6.046414270112851e-07, + "loss": 0.1456, + "step": 43917 + }, + { + "epoch": 7.99, + "learning_rate": 6.045868219876229e-07, + "loss": 0.2269, + "step": 43918 + }, + { + "epoch": 7.99, + "learning_rate": 6.045322169639606e-07, + "loss": 0.2865, + "step": 43919 + }, + { + "epoch": 7.99, + "learning_rate": 6.044776119402986e-07, + "loss": 0.0805, + "step": 43920 + }, + { + "epoch": 7.99, + "learning_rate": 6.044230069166364e-07, + "loss": 2.1893, + "step": 43921 + }, + { + "epoch": 7.99, + "learning_rate": 6.043684018929741e-07, + "loss": 1.2292, + "step": 43922 + }, + { + "epoch": 7.99, + "learning_rate": 6.04313796869312e-07, + "loss": 0.9201, + "step": 43923 + }, + { + "epoch": 7.99, + "learning_rate": 6.042591918456498e-07, + "loss": 0.0978, + "step": 43924 + }, + { + "epoch": 7.99, + "learning_rate": 6.042045868219876e-07, + "loss": 0.0194, + "step": 43925 + }, + { + "epoch": 7.99, + "learning_rate": 6.041499817983254e-07, + "loss": 0.6007, + "step": 43926 + }, + { + "epoch": 7.99, + "learning_rate": 6.040953767746633e-07, + "loss": 1.5294, + "step": 43927 + }, + { + "epoch": 7.99, + "learning_rate": 6.040407717510012e-07, + "loss": 0.6551, + "step": 43928 + }, + { + "epoch": 7.99, + "learning_rate": 6.039861667273389e-07, + "loss": 0.4729, + "step": 43929 + }, + { + "epoch": 7.99, + "learning_rate": 6.039315617036767e-07, + "loss": 0.3702, + "step": 43930 + }, + { + "epoch": 7.99, + "learning_rate": 6.038769566800146e-07, + "loss": 1.0506, + "step": 43931 + }, + { + "epoch": 7.99, + "learning_rate": 6.038223516563524e-07, + "loss": 0.0518, + "step": 43932 + }, + { + "epoch": 7.99, + "learning_rate": 6.037677466326901e-07, + "loss": 1.1031, + "step": 43933 + }, + { + "epoch": 7.99, + "learning_rate": 6.037131416090281e-07, + "loss": 0.0069, + "step": 43934 + }, + { + "epoch": 7.99, + "learning_rate": 6.036585365853659e-07, + "loss": 0.8682, + "step": 43935 + }, + { + "epoch": 7.99, + "learning_rate": 6.036039315617036e-07, + "loss": 0.0104, + "step": 43936 + }, + { + "epoch": 7.99, + "learning_rate": 6.035493265380416e-07, + "loss": 1.8928, + "step": 43937 + }, + { + "epoch": 7.99, + "learning_rate": 6.034947215143793e-07, + "loss": 0.126, + "step": 43938 + }, + { + "epoch": 7.99, + "learning_rate": 6.034401164907172e-07, + "loss": 0.2901, + "step": 43939 + }, + { + "epoch": 7.99, + "learning_rate": 6.03385511467055e-07, + "loss": 0.0075, + "step": 43940 + }, + { + "epoch": 7.99, + "learning_rate": 6.033309064433928e-07, + "loss": 0.0045, + "step": 43941 + }, + { + "epoch": 7.99, + "learning_rate": 6.032763014197307e-07, + "loss": 0.9257, + "step": 43942 + }, + { + "epoch": 7.99, + "learning_rate": 6.032216963960684e-07, + "loss": 0.0094, + "step": 43943 + }, + { + "epoch": 7.99, + "learning_rate": 6.031670913724063e-07, + "loss": 0.5177, + "step": 43944 + }, + { + "epoch": 7.99, + "learning_rate": 6.031124863487441e-07, + "loss": 0.4006, + "step": 43945 + }, + { + "epoch": 7.99, + "learning_rate": 6.030578813250819e-07, + "loss": 0.4312, + "step": 43946 + }, + { + "epoch": 7.99, + "learning_rate": 6.030032763014198e-07, + "loss": 0.3673, + "step": 43947 + }, + { + "epoch": 7.99, + "learning_rate": 6.029486712777576e-07, + "loss": 0.2922, + "step": 43948 + }, + { + "epoch": 7.99, + "learning_rate": 6.028940662540954e-07, + "loss": 0.7652, + "step": 43949 + }, + { + "epoch": 7.99, + "learning_rate": 6.028394612304332e-07, + "loss": 0.2438, + "step": 43950 + }, + { + "epoch": 7.99, + "learning_rate": 6.027848562067711e-07, + "loss": 0.1454, + "step": 43951 + }, + { + "epoch": 7.99, + "learning_rate": 6.027302511831088e-07, + "loss": 0.3462, + "step": 43952 + }, + { + "epoch": 7.99, + "learning_rate": 6.026756461594467e-07, + "loss": 0.0135, + "step": 43953 + }, + { + "epoch": 7.99, + "learning_rate": 6.026210411357845e-07, + "loss": 0.3974, + "step": 43954 + }, + { + "epoch": 7.99, + "learning_rate": 6.025664361121224e-07, + "loss": 3.2372, + "step": 43955 + }, + { + "epoch": 7.99, + "learning_rate": 6.025118310884601e-07, + "loss": 0.2933, + "step": 43956 + }, + { + "epoch": 7.99, + "learning_rate": 6.024572260647979e-07, + "loss": 0.2671, + "step": 43957 + }, + { + "epoch": 7.99, + "learning_rate": 6.024026210411359e-07, + "loss": 0.6212, + "step": 43958 + }, + { + "epoch": 7.99, + "learning_rate": 6.023480160174736e-07, + "loss": 0.6554, + "step": 43959 + }, + { + "epoch": 7.99, + "learning_rate": 6.022934109938114e-07, + "loss": 0.0117, + "step": 43960 + }, + { + "epoch": 7.99, + "learning_rate": 6.022388059701493e-07, + "loss": 0.4113, + "step": 43961 + }, + { + "epoch": 7.99, + "learning_rate": 6.021842009464871e-07, + "loss": 0.004, + "step": 43962 + }, + { + "epoch": 7.99, + "learning_rate": 6.021295959228248e-07, + "loss": 1.2827, + "step": 43963 + }, + { + "epoch": 7.99, + "learning_rate": 6.020749908991628e-07, + "loss": 0.1747, + "step": 43964 + }, + { + "epoch": 8.0, + "learning_rate": 6.020203858755006e-07, + "loss": 1.9948, + "step": 43965 + }, + { + "epoch": 8.0, + "learning_rate": 6.019657808518384e-07, + "loss": 0.0054, + "step": 43966 + }, + { + "epoch": 8.0, + "learning_rate": 6.019111758281762e-07, + "loss": 0.0181, + "step": 43967 + }, + { + "epoch": 8.0, + "learning_rate": 6.01856570804514e-07, + "loss": 1.3424, + "step": 43968 + }, + { + "epoch": 8.0, + "learning_rate": 6.018019657808519e-07, + "loss": 0.0209, + "step": 43969 + }, + { + "epoch": 8.0, + "learning_rate": 6.017473607571896e-07, + "loss": 0.3291, + "step": 43970 + }, + { + "epoch": 8.0, + "learning_rate": 6.016927557335275e-07, + "loss": 0.0421, + "step": 43971 + }, + { + "epoch": 8.0, + "learning_rate": 6.016381507098654e-07, + "loss": 0.2909, + "step": 43972 + }, + { + "epoch": 8.0, + "learning_rate": 6.015835456862031e-07, + "loss": 0.2513, + "step": 43973 + }, + { + "epoch": 8.0, + "learning_rate": 6.015289406625409e-07, + "loss": 0.0745, + "step": 43974 + }, + { + "epoch": 8.0, + "learning_rate": 6.014743356388788e-07, + "loss": 0.3737, + "step": 43975 + }, + { + "epoch": 8.0, + "learning_rate": 6.014197306152166e-07, + "loss": 0.009, + "step": 43976 + }, + { + "epoch": 8.0, + "learning_rate": 6.013651255915544e-07, + "loss": 3.7021, + "step": 43977 + }, + { + "epoch": 8.0, + "learning_rate": 6.013105205678923e-07, + "loss": 0.8659, + "step": 43978 + }, + { + "epoch": 8.0, + "learning_rate": 6.012559155442301e-07, + "loss": 1.0643, + "step": 43979 + }, + { + "epoch": 8.0, + "learning_rate": 6.012013105205679e-07, + "loss": 0.0491, + "step": 43980 + }, + { + "epoch": 8.0, + "learning_rate": 6.011467054969058e-07, + "loss": 0.6438, + "step": 43981 + }, + { + "epoch": 8.0, + "learning_rate": 6.010921004732435e-07, + "loss": 1.24, + "step": 43982 + }, + { + "epoch": 8.0, + "learning_rate": 6.010374954495814e-07, + "loss": 0.0041, + "step": 43983 + }, + { + "epoch": 8.0, + "learning_rate": 6.009828904259192e-07, + "loss": 0.8082, + "step": 43984 + }, + { + "epoch": 8.0, + "learning_rate": 6.00928285402257e-07, + "loss": 0.8242, + "step": 43985 + }, + { + "epoch": 8.0, + "learning_rate": 6.008736803785949e-07, + "loss": 0.3129, + "step": 43986 + }, + { + "epoch": 8.0, + "learning_rate": 6.008190753549326e-07, + "loss": 0.0044, + "step": 43987 + }, + { + "epoch": 8.0, + "learning_rate": 6.007644703312706e-07, + "loss": 1.9294, + "step": 43988 + }, + { + "epoch": 8.0, + "learning_rate": 6.007098653076083e-07, + "loss": 1.2815, + "step": 43989 + }, + { + "epoch": 8.0, + "learning_rate": 6.006552602839461e-07, + "loss": 1.6297, + "step": 43990 + }, + { + "epoch": 8.0, + "learning_rate": 6.00600655260284e-07, + "loss": 0.2424, + "step": 43991 + }, + { + "epoch": 8.0, + "learning_rate": 6.005460502366218e-07, + "loss": 0.0203, + "step": 43992 + }, + { + "epoch": 8.0, + "eval_accuracy": 0.8715220949263502, + "eval_loss": 0.6241586804389954, + "eval_runtime": 139.0346, + "eval_samples_per_second": 17.578, + "eval_steps_per_second": 4.395, + "step": 43992 + }, + { + "epoch": 8.0, + "learning_rate": 6.004914452129595e-07, + "loss": 0.864, + "step": 43993 + }, + { + "epoch": 8.0, + "learning_rate": 6.004368401892974e-07, + "loss": 2.0349, + "step": 43994 + }, + { + "epoch": 8.0, + "learning_rate": 6.003822351656353e-07, + "loss": 1.6851, + "step": 43995 + }, + { + "epoch": 8.0, + "learning_rate": 6.003276301419731e-07, + "loss": 0.3919, + "step": 43996 + }, + { + "epoch": 8.0, + "learning_rate": 6.002730251183109e-07, + "loss": 0.4955, + "step": 43997 + }, + { + "epoch": 8.0, + "learning_rate": 6.002184200946487e-07, + "loss": 0.0063, + "step": 43998 + }, + { + "epoch": 8.0, + "learning_rate": 6.001638150709866e-07, + "loss": 0.3534, + "step": 43999 + }, + { + "epoch": 8.0, + "learning_rate": 6.001092100473243e-07, + "loss": 0.0218, + "step": 44000 + }, + { + "epoch": 8.0, + "learning_rate": 6.000546050236622e-07, + "loss": 0.4863, + "step": 44001 + }, + { + "epoch": 8.0, + "learning_rate": 6.000000000000001e-07, + "loss": 0.107, + "step": 44002 + }, + { + "epoch": 8.0, + "learning_rate": 5.999453949763378e-07, + "loss": 1.8743, + "step": 44003 + }, + { + "epoch": 8.0, + "learning_rate": 5.998907899526756e-07, + "loss": 1.3521, + "step": 44004 + }, + { + "epoch": 8.0, + "learning_rate": 5.998361849290135e-07, + "loss": 0.3202, + "step": 44005 + }, + { + "epoch": 8.0, + "learning_rate": 5.997815799053513e-07, + "loss": 0.0655, + "step": 44006 + }, + { + "epoch": 8.0, + "learning_rate": 5.997269748816891e-07, + "loss": 0.6852, + "step": 44007 + }, + { + "epoch": 8.0, + "learning_rate": 5.99672369858027e-07, + "loss": 0.2226, + "step": 44008 + }, + { + "epoch": 8.0, + "learning_rate": 5.996177648343648e-07, + "loss": 0.141, + "step": 44009 + }, + { + "epoch": 8.0, + "learning_rate": 5.995631598107026e-07, + "loss": 4.6825, + "step": 44010 + }, + { + "epoch": 8.0, + "learning_rate": 5.995085547870404e-07, + "loss": 0.8682, + "step": 44011 + }, + { + "epoch": 8.0, + "learning_rate": 5.994539497633782e-07, + "loss": 0.0024, + "step": 44012 + }, + { + "epoch": 8.0, + "learning_rate": 5.993993447397161e-07, + "loss": 0.8304, + "step": 44013 + }, + { + "epoch": 8.0, + "learning_rate": 5.993447397160538e-07, + "loss": 0.6204, + "step": 44014 + }, + { + "epoch": 8.0, + "learning_rate": 5.992901346923918e-07, + "loss": 0.063, + "step": 44015 + }, + { + "epoch": 8.0, + "learning_rate": 5.992355296687296e-07, + "loss": 0.0322, + "step": 44016 + }, + { + "epoch": 8.0, + "learning_rate": 5.991809246450673e-07, + "loss": 0.7229, + "step": 44017 + }, + { + "epoch": 8.0, + "learning_rate": 5.991263196214052e-07, + "loss": 0.1218, + "step": 44018 + }, + { + "epoch": 8.0, + "learning_rate": 5.99071714597743e-07, + "loss": 0.0736, + "step": 44019 + }, + { + "epoch": 8.01, + "learning_rate": 5.990171095740808e-07, + "loss": 0.4521, + "step": 44020 + }, + { + "epoch": 8.01, + "learning_rate": 5.989625045504187e-07, + "loss": 0.0797, + "step": 44021 + }, + { + "epoch": 8.01, + "learning_rate": 5.989078995267565e-07, + "loss": 0.5665, + "step": 44022 + }, + { + "epoch": 8.01, + "learning_rate": 5.988532945030943e-07, + "loss": 0.9927, + "step": 44023 + }, + { + "epoch": 8.01, + "learning_rate": 5.987986894794321e-07, + "loss": 0.0056, + "step": 44024 + }, + { + "epoch": 8.01, + "learning_rate": 5.9874408445577e-07, + "loss": 0.0866, + "step": 44025 + }, + { + "epoch": 8.01, + "learning_rate": 5.986894794321078e-07, + "loss": 0.1488, + "step": 44026 + }, + { + "epoch": 8.01, + "learning_rate": 5.986348744084456e-07, + "loss": 0.1443, + "step": 44027 + }, + { + "epoch": 8.01, + "learning_rate": 5.985802693847834e-07, + "loss": 0.4058, + "step": 44028 + }, + { + "epoch": 8.01, + "learning_rate": 5.985256643611213e-07, + "loss": 0.4558, + "step": 44029 + }, + { + "epoch": 8.01, + "learning_rate": 5.98471059337459e-07, + "loss": 0.0479, + "step": 44030 + }, + { + "epoch": 8.01, + "learning_rate": 5.984164543137968e-07, + "loss": 2.386, + "step": 44031 + }, + { + "epoch": 8.01, + "learning_rate": 5.983618492901348e-07, + "loss": 0.069, + "step": 44032 + }, + { + "epoch": 8.01, + "learning_rate": 5.983072442664725e-07, + "loss": 0.5186, + "step": 44033 + }, + { + "epoch": 8.01, + "learning_rate": 5.982526392428103e-07, + "loss": 0.2828, + "step": 44034 + }, + { + "epoch": 8.01, + "learning_rate": 5.981980342191482e-07, + "loss": 0.0104, + "step": 44035 + }, + { + "epoch": 8.01, + "learning_rate": 5.98143429195486e-07, + "loss": 0.0943, + "step": 44036 + }, + { + "epoch": 8.01, + "learning_rate": 5.980888241718238e-07, + "loss": 0.1406, + "step": 44037 + }, + { + "epoch": 8.01, + "learning_rate": 5.980342191481616e-07, + "loss": 0.0668, + "step": 44038 + }, + { + "epoch": 8.01, + "learning_rate": 5.979796141244995e-07, + "loss": 0.7444, + "step": 44039 + }, + { + "epoch": 8.01, + "learning_rate": 5.979250091008373e-07, + "loss": 0.7657, + "step": 44040 + }, + { + "epoch": 8.01, + "learning_rate": 5.978704040771751e-07, + "loss": 0.1552, + "step": 44041 + }, + { + "epoch": 8.01, + "learning_rate": 5.978157990535129e-07, + "loss": 0.0966, + "step": 44042 + }, + { + "epoch": 8.01, + "learning_rate": 5.977611940298508e-07, + "loss": 0.0867, + "step": 44043 + }, + { + "epoch": 8.01, + "learning_rate": 5.977065890061885e-07, + "loss": 0.1887, + "step": 44044 + }, + { + "epoch": 8.01, + "learning_rate": 5.976519839825265e-07, + "loss": 0.3375, + "step": 44045 + }, + { + "epoch": 8.01, + "learning_rate": 5.975973789588643e-07, + "loss": 1.3073, + "step": 44046 + }, + { + "epoch": 8.01, + "learning_rate": 5.97542773935202e-07, + "loss": 2.3522, + "step": 44047 + }, + { + "epoch": 8.01, + "learning_rate": 5.974881689115399e-07, + "loss": 0.2932, + "step": 44048 + }, + { + "epoch": 8.01, + "learning_rate": 5.974335638878777e-07, + "loss": 1.7828, + "step": 44049 + }, + { + "epoch": 8.01, + "learning_rate": 5.973789588642155e-07, + "loss": 0.0913, + "step": 44050 + }, + { + "epoch": 8.01, + "learning_rate": 5.973243538405533e-07, + "loss": 0.061, + "step": 44051 + }, + { + "epoch": 8.01, + "learning_rate": 5.972697488168912e-07, + "loss": 0.382, + "step": 44052 + }, + { + "epoch": 8.01, + "learning_rate": 5.97215143793229e-07, + "loss": 1.8326, + "step": 44053 + }, + { + "epoch": 8.01, + "learning_rate": 5.971605387695668e-07, + "loss": 0.0475, + "step": 44054 + }, + { + "epoch": 8.01, + "learning_rate": 5.971059337459046e-07, + "loss": 0.0831, + "step": 44055 + }, + { + "epoch": 8.01, + "learning_rate": 5.970513287222425e-07, + "loss": 0.016, + "step": 44056 + }, + { + "epoch": 8.01, + "learning_rate": 5.969967236985803e-07, + "loss": 0.0115, + "step": 44057 + }, + { + "epoch": 8.01, + "learning_rate": 5.96942118674918e-07, + "loss": 0.0121, + "step": 44058 + }, + { + "epoch": 8.01, + "learning_rate": 5.96887513651256e-07, + "loss": 1.6032, + "step": 44059 + }, + { + "epoch": 8.01, + "learning_rate": 5.968329086275938e-07, + "loss": 0.009, + "step": 44060 + }, + { + "epoch": 8.01, + "learning_rate": 5.967783036039315e-07, + "loss": 1.6584, + "step": 44061 + }, + { + "epoch": 8.01, + "learning_rate": 5.967236985802694e-07, + "loss": 1.1986, + "step": 44062 + }, + { + "epoch": 8.01, + "learning_rate": 5.966690935566072e-07, + "loss": 0.0042, + "step": 44063 + }, + { + "epoch": 8.01, + "learning_rate": 5.96614488532945e-07, + "loss": 1.1521, + "step": 44064 + }, + { + "epoch": 8.01, + "learning_rate": 5.965598835092829e-07, + "loss": 0.6706, + "step": 44065 + }, + { + "epoch": 8.01, + "learning_rate": 5.965052784856207e-07, + "loss": 0.0105, + "step": 44066 + }, + { + "epoch": 8.01, + "learning_rate": 5.964506734619586e-07, + "loss": 0.0349, + "step": 44067 + }, + { + "epoch": 8.01, + "learning_rate": 5.963960684382963e-07, + "loss": 1.5985, + "step": 44068 + }, + { + "epoch": 8.01, + "learning_rate": 5.963414634146342e-07, + "loss": 0.7921, + "step": 44069 + }, + { + "epoch": 8.01, + "learning_rate": 5.96286858390972e-07, + "loss": 0.0737, + "step": 44070 + }, + { + "epoch": 8.01, + "learning_rate": 5.962322533673098e-07, + "loss": 0.5624, + "step": 44071 + }, + { + "epoch": 8.01, + "learning_rate": 5.961776483436476e-07, + "loss": 0.0828, + "step": 44072 + }, + { + "epoch": 8.01, + "learning_rate": 5.961230433199855e-07, + "loss": 0.3589, + "step": 44073 + }, + { + "epoch": 8.01, + "learning_rate": 5.960684382963232e-07, + "loss": 0.2817, + "step": 44074 + }, + { + "epoch": 8.02, + "learning_rate": 5.96013833272661e-07, + "loss": 0.68, + "step": 44075 + }, + { + "epoch": 8.02, + "learning_rate": 5.95959228248999e-07, + "loss": 0.4214, + "step": 44076 + }, + { + "epoch": 8.02, + "learning_rate": 5.959046232253367e-07, + "loss": 0.6309, + "step": 44077 + }, + { + "epoch": 8.02, + "learning_rate": 5.958500182016746e-07, + "loss": 0.0377, + "step": 44078 + }, + { + "epoch": 8.02, + "learning_rate": 5.957954131780124e-07, + "loss": 0.1469, + "step": 44079 + }, + { + "epoch": 8.02, + "learning_rate": 5.957408081543502e-07, + "loss": 0.4106, + "step": 44080 + }, + { + "epoch": 8.02, + "learning_rate": 5.95686203130688e-07, + "loss": 0.0117, + "step": 44081 + }, + { + "epoch": 8.02, + "learning_rate": 5.956315981070258e-07, + "loss": 1.2904, + "step": 44082 + }, + { + "epoch": 8.02, + "learning_rate": 5.955769930833638e-07, + "loss": 0.5553, + "step": 44083 + }, + { + "epoch": 8.02, + "learning_rate": 5.955223880597015e-07, + "loss": 2.0649, + "step": 44084 + }, + { + "epoch": 8.02, + "learning_rate": 5.954677830360393e-07, + "loss": 0.0133, + "step": 44085 + }, + { + "epoch": 8.02, + "learning_rate": 5.954131780123772e-07, + "loss": 0.4891, + "step": 44086 + }, + { + "epoch": 8.02, + "learning_rate": 5.95358572988715e-07, + "loss": 0.3895, + "step": 44087 + }, + { + "epoch": 8.02, + "learning_rate": 5.953039679650527e-07, + "loss": 0.1566, + "step": 44088 + }, + { + "epoch": 8.02, + "learning_rate": 5.952493629413907e-07, + "loss": 0.0162, + "step": 44089 + }, + { + "epoch": 8.02, + "learning_rate": 5.951947579177285e-07, + "loss": 1.709, + "step": 44090 + }, + { + "epoch": 8.02, + "learning_rate": 5.951401528940662e-07, + "loss": 0.502, + "step": 44091 + }, + { + "epoch": 8.02, + "learning_rate": 5.950855478704041e-07, + "loss": 0.0261, + "step": 44092 + }, + { + "epoch": 8.02, + "learning_rate": 5.950309428467419e-07, + "loss": 1.0571, + "step": 44093 + }, + { + "epoch": 8.02, + "learning_rate": 5.949763378230798e-07, + "loss": 1.6237, + "step": 44094 + }, + { + "epoch": 8.02, + "learning_rate": 5.949217327994175e-07, + "loss": 0.3319, + "step": 44095 + }, + { + "epoch": 8.02, + "learning_rate": 5.948671277757554e-07, + "loss": 1.0529, + "step": 44096 + }, + { + "epoch": 8.02, + "learning_rate": 5.948125227520933e-07, + "loss": 0.0583, + "step": 44097 + }, + { + "epoch": 8.02, + "learning_rate": 5.94757917728431e-07, + "loss": 0.0042, + "step": 44098 + }, + { + "epoch": 8.02, + "learning_rate": 5.947033127047688e-07, + "loss": 0.0619, + "step": 44099 + }, + { + "epoch": 8.02, + "learning_rate": 5.946487076811067e-07, + "loss": 0.7141, + "step": 44100 + }, + { + "epoch": 8.02, + "learning_rate": 5.945941026574445e-07, + "loss": 0.0518, + "step": 44101 + }, + { + "epoch": 8.02, + "learning_rate": 5.945394976337822e-07, + "loss": 3.0118, + "step": 44102 + }, + { + "epoch": 8.02, + "learning_rate": 5.944848926101202e-07, + "loss": 0.1782, + "step": 44103 + }, + { + "epoch": 8.02, + "learning_rate": 5.94430287586458e-07, + "loss": 0.0305, + "step": 44104 + }, + { + "epoch": 8.02, + "learning_rate": 5.943756825627958e-07, + "loss": 0.0928, + "step": 44105 + }, + { + "epoch": 8.02, + "learning_rate": 5.943210775391336e-07, + "loss": 0.0079, + "step": 44106 + }, + { + "epoch": 8.02, + "learning_rate": 5.942664725154714e-07, + "loss": 0.5162, + "step": 44107 + }, + { + "epoch": 8.02, + "learning_rate": 5.942118674918093e-07, + "loss": 1.3019, + "step": 44108 + }, + { + "epoch": 8.02, + "learning_rate": 5.941572624681471e-07, + "loss": 0.0647, + "step": 44109 + }, + { + "epoch": 8.02, + "learning_rate": 5.941026574444849e-07, + "loss": 0.0615, + "step": 44110 + }, + { + "epoch": 8.02, + "learning_rate": 5.940480524208227e-07, + "loss": 0.0062, + "step": 44111 + }, + { + "epoch": 8.02, + "learning_rate": 5.939934473971605e-07, + "loss": 0.0035, + "step": 44112 + }, + { + "epoch": 8.02, + "learning_rate": 5.939388423734985e-07, + "loss": 0.0896, + "step": 44113 + }, + { + "epoch": 8.02, + "learning_rate": 5.938842373498362e-07, + "loss": 0.0937, + "step": 44114 + }, + { + "epoch": 8.02, + "learning_rate": 5.93829632326174e-07, + "loss": 0.0244, + "step": 44115 + }, + { + "epoch": 8.02, + "learning_rate": 5.937750273025119e-07, + "loss": 0.08, + "step": 44116 + }, + { + "epoch": 8.02, + "learning_rate": 5.937204222788497e-07, + "loss": 1.3284, + "step": 44117 + }, + { + "epoch": 8.02, + "learning_rate": 5.936658172551874e-07, + "loss": 0.1507, + "step": 44118 + }, + { + "epoch": 8.02, + "learning_rate": 5.936112122315253e-07, + "loss": 0.1823, + "step": 44119 + }, + { + "epoch": 8.02, + "learning_rate": 5.935566072078632e-07, + "loss": 0.0579, + "step": 44120 + }, + { + "epoch": 8.02, + "learning_rate": 5.935020021842009e-07, + "loss": 0.0132, + "step": 44121 + }, + { + "epoch": 8.02, + "learning_rate": 5.934473971605388e-07, + "loss": 1.0037, + "step": 44122 + }, + { + "epoch": 8.02, + "learning_rate": 5.933927921368766e-07, + "loss": 0.2168, + "step": 44123 + }, + { + "epoch": 8.02, + "learning_rate": 5.933381871132145e-07, + "loss": 1.9415, + "step": 44124 + }, + { + "epoch": 8.02, + "learning_rate": 5.932835820895522e-07, + "loss": 0.0107, + "step": 44125 + }, + { + "epoch": 8.02, + "learning_rate": 5.9322897706589e-07, + "loss": 0.977, + "step": 44126 + }, + { + "epoch": 8.02, + "learning_rate": 5.93174372042228e-07, + "loss": 0.2808, + "step": 44127 + }, + { + "epoch": 8.02, + "learning_rate": 5.931197670185657e-07, + "loss": 0.0044, + "step": 44128 + }, + { + "epoch": 8.02, + "learning_rate": 5.930651619949035e-07, + "loss": 0.0068, + "step": 44129 + }, + { + "epoch": 8.03, + "learning_rate": 5.930105569712414e-07, + "loss": 0.0138, + "step": 44130 + }, + { + "epoch": 8.03, + "learning_rate": 5.929559519475792e-07, + "loss": 0.1931, + "step": 44131 + }, + { + "epoch": 8.03, + "learning_rate": 5.929013469239169e-07, + "loss": 0.0447, + "step": 44132 + }, + { + "epoch": 8.03, + "learning_rate": 5.928467419002549e-07, + "loss": 0.0501, + "step": 44133 + }, + { + "epoch": 8.03, + "learning_rate": 5.927921368765927e-07, + "loss": 0.1828, + "step": 44134 + }, + { + "epoch": 8.03, + "learning_rate": 5.927375318529305e-07, + "loss": 0.1301, + "step": 44135 + }, + { + "epoch": 8.03, + "learning_rate": 5.926829268292683e-07, + "loss": 0.0753, + "step": 44136 + }, + { + "epoch": 8.03, + "learning_rate": 5.926283218056061e-07, + "loss": 0.0103, + "step": 44137 + }, + { + "epoch": 8.03, + "learning_rate": 5.92573716781944e-07, + "loss": 0.0046, + "step": 44138 + }, + { + "epoch": 8.03, + "learning_rate": 5.925191117582817e-07, + "loss": 2.9729, + "step": 44139 + }, + { + "epoch": 8.03, + "learning_rate": 5.924645067346196e-07, + "loss": 0.3133, + "step": 44140 + }, + { + "epoch": 8.03, + "learning_rate": 5.924099017109575e-07, + "loss": 3.258, + "step": 44141 + }, + { + "epoch": 8.03, + "learning_rate": 5.923552966872952e-07, + "loss": 0.0091, + "step": 44142 + }, + { + "epoch": 8.03, + "learning_rate": 5.92300691663633e-07, + "loss": 0.0093, + "step": 44143 + }, + { + "epoch": 8.03, + "learning_rate": 5.922460866399709e-07, + "loss": 0.0046, + "step": 44144 + }, + { + "epoch": 8.03, + "learning_rate": 5.921914816163087e-07, + "loss": 0.0212, + "step": 44145 + }, + { + "epoch": 8.03, + "learning_rate": 5.921368765926465e-07, + "loss": 0.3175, + "step": 44146 + }, + { + "epoch": 8.03, + "learning_rate": 5.920822715689844e-07, + "loss": 1.6884, + "step": 44147 + }, + { + "epoch": 8.03, + "learning_rate": 5.920276665453221e-07, + "loss": 0.2168, + "step": 44148 + }, + { + "epoch": 8.03, + "learning_rate": 5.9197306152166e-07, + "loss": 0.3053, + "step": 44149 + }, + { + "epoch": 8.03, + "learning_rate": 5.919184564979979e-07, + "loss": 1.9497, + "step": 44150 + }, + { + "epoch": 8.03, + "learning_rate": 5.918638514743356e-07, + "loss": 0.5307, + "step": 44151 + }, + { + "epoch": 8.03, + "learning_rate": 5.918092464506735e-07, + "loss": 0.1044, + "step": 44152 + }, + { + "epoch": 8.03, + "learning_rate": 5.917546414270113e-07, + "loss": 0.6891, + "step": 44153 + }, + { + "epoch": 8.03, + "learning_rate": 5.917000364033492e-07, + "loss": 0.0326, + "step": 44154 + }, + { + "epoch": 8.03, + "learning_rate": 5.916454313796869e-07, + "loss": 2.6128, + "step": 44155 + }, + { + "epoch": 8.03, + "learning_rate": 5.915908263560247e-07, + "loss": 1.0148, + "step": 44156 + }, + { + "epoch": 8.03, + "learning_rate": 5.915362213323627e-07, + "loss": 0.0497, + "step": 44157 + }, + { + "epoch": 8.03, + "learning_rate": 5.914816163087004e-07, + "loss": 1.5695, + "step": 44158 + }, + { + "epoch": 8.03, + "learning_rate": 5.914270112850382e-07, + "loss": 0.0571, + "step": 44159 + }, + { + "epoch": 8.03, + "learning_rate": 5.913724062613761e-07, + "loss": 0.1051, + "step": 44160 + }, + { + "epoch": 8.03, + "learning_rate": 5.913178012377139e-07, + "loss": 0.1712, + "step": 44161 + }, + { + "epoch": 8.03, + "learning_rate": 5.912631962140516e-07, + "loss": 0.454, + "step": 44162 + }, + { + "epoch": 8.03, + "learning_rate": 5.912085911903895e-07, + "loss": 0.1436, + "step": 44163 + }, + { + "epoch": 8.03, + "learning_rate": 5.911539861667274e-07, + "loss": 0.1505, + "step": 44164 + }, + { + "epoch": 8.03, + "learning_rate": 5.910993811430652e-07, + "loss": 0.8113, + "step": 44165 + }, + { + "epoch": 8.03, + "learning_rate": 5.91044776119403e-07, + "loss": 0.0083, + "step": 44166 + }, + { + "epoch": 8.03, + "learning_rate": 5.909901710957408e-07, + "loss": 0.0057, + "step": 44167 + }, + { + "epoch": 8.03, + "learning_rate": 5.909355660720787e-07, + "loss": 0.6445, + "step": 44168 + }, + { + "epoch": 8.03, + "learning_rate": 5.908809610484164e-07, + "loss": 0.0642, + "step": 44169 + }, + { + "epoch": 8.03, + "learning_rate": 5.908263560247543e-07, + "loss": 0.5134, + "step": 44170 + }, + { + "epoch": 8.03, + "learning_rate": 5.907717510010922e-07, + "loss": 0.995, + "step": 44171 + }, + { + "epoch": 8.03, + "learning_rate": 5.907171459774299e-07, + "loss": 1.3876, + "step": 44172 + }, + { + "epoch": 8.03, + "learning_rate": 5.906625409537678e-07, + "loss": 1.6378, + "step": 44173 + }, + { + "epoch": 8.03, + "learning_rate": 5.906079359301056e-07, + "loss": 0.9336, + "step": 44174 + }, + { + "epoch": 8.03, + "learning_rate": 5.905533309064434e-07, + "loss": 1.3749, + "step": 44175 + }, + { + "epoch": 8.03, + "learning_rate": 5.904987258827812e-07, + "loss": 0.0103, + "step": 44176 + }, + { + "epoch": 8.03, + "learning_rate": 5.904441208591191e-07, + "loss": 0.357, + "step": 44177 + }, + { + "epoch": 8.03, + "learning_rate": 5.903895158354569e-07, + "loss": 1.0849, + "step": 44178 + }, + { + "epoch": 8.03, + "learning_rate": 5.903349108117947e-07, + "loss": 0.1682, + "step": 44179 + }, + { + "epoch": 8.03, + "learning_rate": 5.902803057881325e-07, + "loss": 0.3087, + "step": 44180 + }, + { + "epoch": 8.03, + "learning_rate": 5.902257007644703e-07, + "loss": 0.0133, + "step": 44181 + }, + { + "epoch": 8.03, + "learning_rate": 5.901710957408082e-07, + "loss": 0.1986, + "step": 44182 + }, + { + "epoch": 8.03, + "learning_rate": 5.901164907171459e-07, + "loss": 0.0086, + "step": 44183 + }, + { + "epoch": 8.03, + "learning_rate": 5.900618856934839e-07, + "loss": 0.4331, + "step": 44184 + }, + { + "epoch": 8.04, + "learning_rate": 5.900072806698216e-07, + "loss": 0.0316, + "step": 44185 + }, + { + "epoch": 8.04, + "learning_rate": 5.899526756461594e-07, + "loss": 0.0295, + "step": 44186 + }, + { + "epoch": 8.04, + "learning_rate": 5.898980706224973e-07, + "loss": 0.2976, + "step": 44187 + }, + { + "epoch": 8.04, + "learning_rate": 5.898434655988351e-07, + "loss": 2.7078, + "step": 44188 + }, + { + "epoch": 8.04, + "learning_rate": 5.897888605751729e-07, + "loss": 0.0082, + "step": 44189 + }, + { + "epoch": 8.04, + "learning_rate": 5.897342555515107e-07, + "loss": 0.1572, + "step": 44190 + }, + { + "epoch": 8.04, + "learning_rate": 5.896796505278486e-07, + "loss": 0.0463, + "step": 44191 + }, + { + "epoch": 8.04, + "learning_rate": 5.896250455041863e-07, + "loss": 0.1401, + "step": 44192 + }, + { + "epoch": 8.04, + "learning_rate": 5.895704404805242e-07, + "loss": 0.8684, + "step": 44193 + }, + { + "epoch": 8.04, + "learning_rate": 5.895158354568621e-07, + "loss": 1.526, + "step": 44194 + }, + { + "epoch": 8.04, + "learning_rate": 5.894612304331999e-07, + "loss": 0.0676, + "step": 44195 + }, + { + "epoch": 8.04, + "learning_rate": 5.894066254095377e-07, + "loss": 0.3556, + "step": 44196 + }, + { + "epoch": 8.04, + "learning_rate": 5.893520203858755e-07, + "loss": 1.7252, + "step": 44197 + }, + { + "epoch": 8.04, + "learning_rate": 5.892974153622134e-07, + "loss": 0.4215, + "step": 44198 + }, + { + "epoch": 8.04, + "learning_rate": 5.892428103385511e-07, + "loss": 1.4587, + "step": 44199 + }, + { + "epoch": 8.04, + "learning_rate": 5.891882053148889e-07, + "loss": 2.231, + "step": 44200 + }, + { + "epoch": 8.04, + "learning_rate": 5.891336002912269e-07, + "loss": 0.4743, + "step": 44201 + }, + { + "epoch": 8.04, + "learning_rate": 5.890789952675646e-07, + "loss": 1.0882, + "step": 44202 + }, + { + "epoch": 8.04, + "learning_rate": 5.890243902439025e-07, + "loss": 0.2505, + "step": 44203 + }, + { + "epoch": 8.04, + "learning_rate": 5.889697852202403e-07, + "loss": 0.0021, + "step": 44204 + }, + { + "epoch": 8.04, + "learning_rate": 5.889151801965781e-07, + "loss": 0.1801, + "step": 44205 + }, + { + "epoch": 8.04, + "learning_rate": 5.888605751729159e-07, + "loss": 0.0088, + "step": 44206 + }, + { + "epoch": 8.04, + "learning_rate": 5.888059701492537e-07, + "loss": 0.5088, + "step": 44207 + }, + { + "epoch": 8.04, + "learning_rate": 5.887513651255916e-07, + "loss": 0.4437, + "step": 44208 + }, + { + "epoch": 8.04, + "learning_rate": 5.886967601019294e-07, + "loss": 2.323, + "step": 44209 + }, + { + "epoch": 8.04, + "learning_rate": 5.886421550782672e-07, + "loss": 0.3926, + "step": 44210 + }, + { + "epoch": 8.04, + "learning_rate": 5.88587550054605e-07, + "loss": 0.0699, + "step": 44211 + }, + { + "epoch": 8.04, + "learning_rate": 5.885329450309429e-07, + "loss": 0.4443, + "step": 44212 + }, + { + "epoch": 8.04, + "learning_rate": 5.884783400072806e-07, + "loss": 0.029, + "step": 44213 + }, + { + "epoch": 8.04, + "learning_rate": 5.884237349836186e-07, + "loss": 0.0587, + "step": 44214 + }, + { + "epoch": 8.04, + "learning_rate": 5.883691299599564e-07, + "loss": 0.007, + "step": 44215 + }, + { + "epoch": 8.04, + "learning_rate": 5.883145249362941e-07, + "loss": 0.7314, + "step": 44216 + }, + { + "epoch": 8.04, + "learning_rate": 5.88259919912632e-07, + "loss": 0.219, + "step": 44217 + }, + { + "epoch": 8.04, + "learning_rate": 5.882053148889698e-07, + "loss": 1.4164, + "step": 44218 + }, + { + "epoch": 8.04, + "learning_rate": 5.881507098653076e-07, + "loss": 1.669, + "step": 44219 + }, + { + "epoch": 8.04, + "learning_rate": 5.880961048416454e-07, + "loss": 0.0224, + "step": 44220 + }, + { + "epoch": 8.04, + "learning_rate": 5.880414998179833e-07, + "loss": 0.1043, + "step": 44221 + }, + { + "epoch": 8.04, + "learning_rate": 5.879868947943212e-07, + "loss": 0.8062, + "step": 44222 + }, + { + "epoch": 8.04, + "learning_rate": 5.879322897706589e-07, + "loss": 0.1502, + "step": 44223 + }, + { + "epoch": 8.04, + "learning_rate": 5.878776847469967e-07, + "loss": 0.0152, + "step": 44224 + }, + { + "epoch": 8.04, + "learning_rate": 5.878230797233346e-07, + "loss": 0.3104, + "step": 44225 + }, + { + "epoch": 8.04, + "learning_rate": 5.877684746996724e-07, + "loss": 0.0306, + "step": 44226 + }, + { + "epoch": 8.04, + "learning_rate": 5.877138696760101e-07, + "loss": 0.0445, + "step": 44227 + }, + { + "epoch": 8.04, + "learning_rate": 5.876592646523481e-07, + "loss": 0.2033, + "step": 44228 + }, + { + "epoch": 8.04, + "learning_rate": 5.876046596286858e-07, + "loss": 0.1416, + "step": 44229 + }, + { + "epoch": 8.04, + "learning_rate": 5.875500546050236e-07, + "loss": 0.0426, + "step": 44230 + }, + { + "epoch": 8.04, + "learning_rate": 5.874954495813615e-07, + "loss": 1.3884, + "step": 44231 + }, + { + "epoch": 8.04, + "learning_rate": 5.874408445576993e-07, + "loss": 0.3805, + "step": 44232 + }, + { + "epoch": 8.04, + "learning_rate": 5.873862395340372e-07, + "loss": 0.253, + "step": 44233 + }, + { + "epoch": 8.04, + "learning_rate": 5.87331634510375e-07, + "loss": 0.4242, + "step": 44234 + }, + { + "epoch": 8.04, + "learning_rate": 5.872770294867128e-07, + "loss": 0.7586, + "step": 44235 + }, + { + "epoch": 8.04, + "learning_rate": 5.872224244630506e-07, + "loss": 0.0054, + "step": 44236 + }, + { + "epoch": 8.04, + "learning_rate": 5.871678194393884e-07, + "loss": 0.0036, + "step": 44237 + }, + { + "epoch": 8.04, + "learning_rate": 5.871132144157263e-07, + "loss": 0.9406, + "step": 44238 + }, + { + "epoch": 8.04, + "learning_rate": 5.870586093920641e-07, + "loss": 0.4321, + "step": 44239 + }, + { + "epoch": 8.05, + "learning_rate": 5.870040043684019e-07, + "loss": 0.0496, + "step": 44240 + }, + { + "epoch": 8.05, + "learning_rate": 5.869493993447398e-07, + "loss": 0.0502, + "step": 44241 + }, + { + "epoch": 8.05, + "learning_rate": 5.868947943210776e-07, + "loss": 0.3875, + "step": 44242 + }, + { + "epoch": 8.05, + "learning_rate": 5.868401892974153e-07, + "loss": 1.0327, + "step": 44243 + }, + { + "epoch": 8.05, + "learning_rate": 5.867855842737532e-07, + "loss": 0.0207, + "step": 44244 + }, + { + "epoch": 8.05, + "learning_rate": 5.867309792500911e-07, + "loss": 0.8826, + "step": 44245 + }, + { + "epoch": 8.05, + "learning_rate": 5.866763742264288e-07, + "loss": 0.1039, + "step": 44246 + }, + { + "epoch": 8.05, + "learning_rate": 5.866217692027667e-07, + "loss": 0.0421, + "step": 44247 + }, + { + "epoch": 8.05, + "learning_rate": 5.865671641791045e-07, + "loss": 2.23, + "step": 44248 + }, + { + "epoch": 8.05, + "learning_rate": 5.865125591554423e-07, + "loss": 0.3365, + "step": 44249 + }, + { + "epoch": 8.05, + "learning_rate": 5.864579541317801e-07, + "loss": 0.443, + "step": 44250 + }, + { + "epoch": 8.05, + "learning_rate": 5.864033491081179e-07, + "loss": 0.8968, + "step": 44251 + }, + { + "epoch": 8.05, + "learning_rate": 5.863487440844559e-07, + "loss": 0.125, + "step": 44252 + }, + { + "epoch": 8.05, + "learning_rate": 5.862941390607936e-07, + "loss": 0.0042, + "step": 44253 + }, + { + "epoch": 8.05, + "learning_rate": 5.862395340371314e-07, + "loss": 0.0045, + "step": 44254 + }, + { + "epoch": 8.05, + "learning_rate": 5.861849290134693e-07, + "loss": 0.2123, + "step": 44255 + }, + { + "epoch": 8.05, + "learning_rate": 5.861303239898071e-07, + "loss": 0.2054, + "step": 44256 + }, + { + "epoch": 8.05, + "learning_rate": 5.860757189661448e-07, + "loss": 0.1259, + "step": 44257 + }, + { + "epoch": 8.05, + "learning_rate": 5.860211139424828e-07, + "loss": 0.0123, + "step": 44258 + }, + { + "epoch": 8.05, + "learning_rate": 5.859665089188206e-07, + "loss": 0.1745, + "step": 44259 + }, + { + "epoch": 8.05, + "learning_rate": 5.859119038951583e-07, + "loss": 0.5398, + "step": 44260 + }, + { + "epoch": 8.05, + "learning_rate": 5.858572988714962e-07, + "loss": 0.01, + "step": 44261 + }, + { + "epoch": 8.05, + "learning_rate": 5.85802693847834e-07, + "loss": 1.3064, + "step": 44262 + }, + { + "epoch": 8.05, + "learning_rate": 5.857480888241719e-07, + "loss": 0.442, + "step": 44263 + }, + { + "epoch": 8.05, + "learning_rate": 5.856934838005096e-07, + "loss": 0.7098, + "step": 44264 + }, + { + "epoch": 8.05, + "learning_rate": 5.856388787768475e-07, + "loss": 0.0235, + "step": 44265 + }, + { + "epoch": 8.05, + "learning_rate": 5.855842737531853e-07, + "loss": 2.2727, + "step": 44266 + }, + { + "epoch": 8.05, + "learning_rate": 5.855296687295231e-07, + "loss": 0.3781, + "step": 44267 + }, + { + "epoch": 8.05, + "learning_rate": 5.854750637058609e-07, + "loss": 0.0758, + "step": 44268 + }, + { + "epoch": 8.05, + "learning_rate": 5.854204586821988e-07, + "loss": 0.2139, + "step": 44269 + }, + { + "epoch": 8.05, + "learning_rate": 5.853658536585366e-07, + "loss": 0.1476, + "step": 44270 + }, + { + "epoch": 8.05, + "learning_rate": 5.853112486348743e-07, + "loss": 0.0384, + "step": 44271 + }, + { + "epoch": 8.05, + "learning_rate": 5.852566436112123e-07, + "loss": 0.0057, + "step": 44272 + }, + { + "epoch": 8.05, + "learning_rate": 5.8520203858755e-07, + "loss": 0.0077, + "step": 44273 + }, + { + "epoch": 8.05, + "learning_rate": 5.851474335638879e-07, + "loss": 2.6029, + "step": 44274 + }, + { + "epoch": 8.05, + "learning_rate": 5.850928285402257e-07, + "loss": 0.1424, + "step": 44275 + }, + { + "epoch": 8.05, + "learning_rate": 5.850382235165635e-07, + "loss": 0.1467, + "step": 44276 + }, + { + "epoch": 8.05, + "learning_rate": 5.849836184929014e-07, + "loss": 0.6075, + "step": 44277 + }, + { + "epoch": 8.05, + "learning_rate": 5.849290134692392e-07, + "loss": 0.0441, + "step": 44278 + }, + { + "epoch": 8.05, + "learning_rate": 5.84874408445577e-07, + "loss": 0.2105, + "step": 44279 + }, + { + "epoch": 8.05, + "learning_rate": 5.848198034219148e-07, + "loss": 0.4072, + "step": 44280 + }, + { + "epoch": 8.05, + "learning_rate": 5.847651983982526e-07, + "loss": 0.6204, + "step": 44281 + }, + { + "epoch": 8.05, + "learning_rate": 5.847105933745906e-07, + "loss": 0.0153, + "step": 44282 + }, + { + "epoch": 8.05, + "learning_rate": 5.846559883509283e-07, + "loss": 0.0179, + "step": 44283 + }, + { + "epoch": 8.05, + "learning_rate": 5.846013833272661e-07, + "loss": 0.1852, + "step": 44284 + }, + { + "epoch": 8.05, + "learning_rate": 5.84546778303604e-07, + "loss": 0.4918, + "step": 44285 + }, + { + "epoch": 8.05, + "learning_rate": 5.844921732799418e-07, + "loss": 1.6824, + "step": 44286 + }, + { + "epoch": 8.05, + "learning_rate": 5.844375682562795e-07, + "loss": 0.1366, + "step": 44287 + }, + { + "epoch": 8.05, + "learning_rate": 5.843829632326174e-07, + "loss": 2.0135, + "step": 44288 + }, + { + "epoch": 8.05, + "learning_rate": 5.843283582089553e-07, + "loss": 0.0046, + "step": 44289 + }, + { + "epoch": 8.05, + "learning_rate": 5.84273753185293e-07, + "loss": 0.0313, + "step": 44290 + }, + { + "epoch": 8.05, + "learning_rate": 5.842191481616309e-07, + "loss": 0.0623, + "step": 44291 + }, + { + "epoch": 8.05, + "learning_rate": 5.841645431379687e-07, + "loss": 0.1276, + "step": 44292 + }, + { + "epoch": 8.05, + "learning_rate": 5.841099381143066e-07, + "loss": 0.409, + "step": 44293 + }, + { + "epoch": 8.05, + "learning_rate": 5.840553330906443e-07, + "loss": 0.0524, + "step": 44294 + }, + { + "epoch": 8.06, + "learning_rate": 5.840007280669821e-07, + "loss": 0.0054, + "step": 44295 + }, + { + "epoch": 8.06, + "learning_rate": 5.839461230433201e-07, + "loss": 1.7567, + "step": 44296 + }, + { + "epoch": 8.06, + "learning_rate": 5.838915180196578e-07, + "loss": 0.3236, + "step": 44297 + }, + { + "epoch": 8.06, + "learning_rate": 5.838369129959956e-07, + "loss": 0.1795, + "step": 44298 + }, + { + "epoch": 8.06, + "learning_rate": 5.837823079723335e-07, + "loss": 0.3692, + "step": 44299 + }, + { + "epoch": 8.06, + "learning_rate": 5.837277029486713e-07, + "loss": 0.0119, + "step": 44300 + }, + { + "epoch": 8.06, + "learning_rate": 5.83673097925009e-07, + "loss": 1.2291, + "step": 44301 + }, + { + "epoch": 8.06, + "learning_rate": 5.83618492901347e-07, + "loss": 2.2131, + "step": 44302 + }, + { + "epoch": 8.06, + "learning_rate": 5.835638878776847e-07, + "loss": 0.9326, + "step": 44303 + }, + { + "epoch": 8.06, + "learning_rate": 5.835092828540226e-07, + "loss": 0.1086, + "step": 44304 + }, + { + "epoch": 8.06, + "learning_rate": 5.834546778303604e-07, + "loss": 0.0114, + "step": 44305 + }, + { + "epoch": 8.06, + "learning_rate": 5.834000728066982e-07, + "loss": 0.7691, + "step": 44306 + }, + { + "epoch": 8.06, + "learning_rate": 5.833454677830361e-07, + "loss": 0.4294, + "step": 44307 + }, + { + "epoch": 8.06, + "learning_rate": 5.832908627593738e-07, + "loss": 0.0289, + "step": 44308 + }, + { + "epoch": 8.06, + "learning_rate": 5.832362577357117e-07, + "loss": 0.3229, + "step": 44309 + }, + { + "epoch": 8.06, + "learning_rate": 5.831816527120495e-07, + "loss": 5.1371, + "step": 44310 + }, + { + "epoch": 8.06, + "learning_rate": 5.831270476883873e-07, + "loss": 0.0452, + "step": 44311 + }, + { + "epoch": 8.06, + "learning_rate": 5.830724426647252e-07, + "loss": 0.0045, + "step": 44312 + }, + { + "epoch": 8.06, + "learning_rate": 5.83017837641063e-07, + "loss": 0.9392, + "step": 44313 + }, + { + "epoch": 8.06, + "learning_rate": 5.829632326174008e-07, + "loss": 5.3647, + "step": 44314 + }, + { + "epoch": 8.06, + "learning_rate": 5.829086275937386e-07, + "loss": 0.0554, + "step": 44315 + }, + { + "epoch": 8.06, + "learning_rate": 5.828540225700765e-07, + "loss": 1.1281, + "step": 44316 + }, + { + "epoch": 8.06, + "learning_rate": 5.827994175464142e-07, + "loss": 0.0401, + "step": 44317 + }, + { + "epoch": 8.06, + "learning_rate": 5.827448125227521e-07, + "loss": 0.6283, + "step": 44318 + }, + { + "epoch": 8.06, + "learning_rate": 5.8269020749909e-07, + "loss": 0.0623, + "step": 44319 + }, + { + "epoch": 8.06, + "learning_rate": 5.826356024754278e-07, + "loss": 0.3062, + "step": 44320 + }, + { + "epoch": 8.06, + "learning_rate": 5.825809974517656e-07, + "loss": 0.0608, + "step": 44321 + }, + { + "epoch": 8.06, + "learning_rate": 5.825263924281034e-07, + "loss": 1.2329, + "step": 44322 + }, + { + "epoch": 8.06, + "learning_rate": 5.824717874044413e-07, + "loss": 0.1801, + "step": 44323 + }, + { + "epoch": 8.06, + "learning_rate": 5.82417182380779e-07, + "loss": 0.01, + "step": 44324 + }, + { + "epoch": 8.06, + "learning_rate": 5.823625773571168e-07, + "loss": 0.1352, + "step": 44325 + }, + { + "epoch": 8.06, + "learning_rate": 5.823079723334548e-07, + "loss": 0.7783, + "step": 44326 + }, + { + "epoch": 8.06, + "learning_rate": 5.822533673097925e-07, + "loss": 0.0319, + "step": 44327 + }, + { + "epoch": 8.06, + "learning_rate": 5.821987622861303e-07, + "loss": 0.0383, + "step": 44328 + }, + { + "epoch": 8.06, + "learning_rate": 5.821441572624682e-07, + "loss": 0.0051, + "step": 44329 + }, + { + "epoch": 8.06, + "learning_rate": 5.82089552238806e-07, + "loss": 1.0707, + "step": 44330 + }, + { + "epoch": 8.06, + "learning_rate": 5.820349472151438e-07, + "loss": 1.5689, + "step": 44331 + }, + { + "epoch": 8.06, + "learning_rate": 5.819803421914816e-07, + "loss": 0.0043, + "step": 44332 + }, + { + "epoch": 8.06, + "learning_rate": 5.819257371678195e-07, + "loss": 0.0655, + "step": 44333 + }, + { + "epoch": 8.06, + "learning_rate": 5.818711321441573e-07, + "loss": 2.1957, + "step": 44334 + }, + { + "epoch": 8.06, + "learning_rate": 5.818165271204951e-07, + "loss": 2.2764, + "step": 44335 + }, + { + "epoch": 8.06, + "learning_rate": 5.817619220968329e-07, + "loss": 0.7977, + "step": 44336 + }, + { + "epoch": 8.06, + "learning_rate": 5.817073170731708e-07, + "loss": 0.3822, + "step": 44337 + }, + { + "epoch": 8.06, + "learning_rate": 5.816527120495085e-07, + "loss": 0.189, + "step": 44338 + }, + { + "epoch": 8.06, + "learning_rate": 5.815981070258463e-07, + "loss": 0.6441, + "step": 44339 + }, + { + "epoch": 8.06, + "learning_rate": 5.815435020021842e-07, + "loss": 0.0134, + "step": 44340 + }, + { + "epoch": 8.06, + "learning_rate": 5.81488896978522e-07, + "loss": 0.0431, + "step": 44341 + }, + { + "epoch": 8.06, + "learning_rate": 5.814342919548599e-07, + "loss": 0.9388, + "step": 44342 + }, + { + "epoch": 8.06, + "learning_rate": 5.813796869311977e-07, + "loss": 0.359, + "step": 44343 + }, + { + "epoch": 8.06, + "learning_rate": 5.813250819075355e-07, + "loss": 1.6963, + "step": 44344 + }, + { + "epoch": 8.06, + "learning_rate": 5.812704768838733e-07, + "loss": 0.9602, + "step": 44345 + }, + { + "epoch": 8.06, + "learning_rate": 5.812158718602112e-07, + "loss": 0.6122, + "step": 44346 + }, + { + "epoch": 8.06, + "learning_rate": 5.811612668365489e-07, + "loss": 0.1045, + "step": 44347 + }, + { + "epoch": 8.06, + "learning_rate": 5.811066618128868e-07, + "loss": 0.0371, + "step": 44348 + }, + { + "epoch": 8.06, + "learning_rate": 5.810520567892246e-07, + "loss": 3.8452, + "step": 44349 + }, + { + "epoch": 8.07, + "learning_rate": 5.809974517655625e-07, + "loss": 0.0667, + "step": 44350 + }, + { + "epoch": 8.07, + "learning_rate": 5.809428467419003e-07, + "loss": 1.4338, + "step": 44351 + }, + { + "epoch": 8.07, + "learning_rate": 5.80888241718238e-07, + "loss": 2.9533, + "step": 44352 + }, + { + "epoch": 8.07, + "learning_rate": 5.80833636694576e-07, + "loss": 0.304, + "step": 44353 + }, + { + "epoch": 8.07, + "learning_rate": 5.807790316709137e-07, + "loss": 0.3614, + "step": 44354 + }, + { + "epoch": 8.07, + "learning_rate": 5.807244266472515e-07, + "loss": 0.2511, + "step": 44355 + }, + { + "epoch": 8.07, + "learning_rate": 5.806698216235894e-07, + "loss": 0.033, + "step": 44356 + }, + { + "epoch": 8.07, + "learning_rate": 5.806152165999272e-07, + "loss": 0.7936, + "step": 44357 + }, + { + "epoch": 8.07, + "learning_rate": 5.80560611576265e-07, + "loss": 1.1206, + "step": 44358 + }, + { + "epoch": 8.07, + "learning_rate": 5.805060065526028e-07, + "loss": 0.0712, + "step": 44359 + }, + { + "epoch": 8.07, + "learning_rate": 5.804514015289407e-07, + "loss": 1.0938, + "step": 44360 + }, + { + "epoch": 8.07, + "learning_rate": 5.803967965052785e-07, + "loss": 0.2464, + "step": 44361 + }, + { + "epoch": 8.07, + "learning_rate": 5.803421914816163e-07, + "loss": 1.2696, + "step": 44362 + }, + { + "epoch": 8.07, + "learning_rate": 5.802875864579542e-07, + "loss": 1.6936, + "step": 44363 + }, + { + "epoch": 8.07, + "learning_rate": 5.80232981434292e-07, + "loss": 0.0902, + "step": 44364 + }, + { + "epoch": 8.07, + "learning_rate": 5.801783764106298e-07, + "loss": 0.0138, + "step": 44365 + }, + { + "epoch": 8.07, + "learning_rate": 5.801237713869676e-07, + "loss": 0.0922, + "step": 44366 + }, + { + "epoch": 8.07, + "learning_rate": 5.800691663633055e-07, + "loss": 0.7117, + "step": 44367 + }, + { + "epoch": 8.07, + "learning_rate": 5.800145613396432e-07, + "loss": 0.7453, + "step": 44368 + }, + { + "epoch": 8.07, + "learning_rate": 5.79959956315981e-07, + "loss": 0.6942, + "step": 44369 + }, + { + "epoch": 8.07, + "learning_rate": 5.79905351292319e-07, + "loss": 0.3026, + "step": 44370 + }, + { + "epoch": 8.07, + "learning_rate": 5.798507462686567e-07, + "loss": 0.319, + "step": 44371 + }, + { + "epoch": 8.07, + "learning_rate": 5.797961412449946e-07, + "loss": 0.184, + "step": 44372 + }, + { + "epoch": 8.07, + "learning_rate": 5.797415362213324e-07, + "loss": 0.9103, + "step": 44373 + }, + { + "epoch": 8.07, + "learning_rate": 5.796869311976702e-07, + "loss": 1.871, + "step": 44374 + }, + { + "epoch": 8.07, + "learning_rate": 5.79632326174008e-07, + "loss": 0.1457, + "step": 44375 + }, + { + "epoch": 8.07, + "learning_rate": 5.795777211503458e-07, + "loss": 0.0066, + "step": 44376 + }, + { + "epoch": 8.07, + "learning_rate": 5.795231161266837e-07, + "loss": 0.0706, + "step": 44377 + }, + { + "epoch": 8.07, + "learning_rate": 5.794685111030215e-07, + "loss": 0.3976, + "step": 44378 + }, + { + "epoch": 8.07, + "learning_rate": 5.794139060793593e-07, + "loss": 0.747, + "step": 44379 + }, + { + "epoch": 8.07, + "learning_rate": 5.793593010556972e-07, + "loss": 3.107, + "step": 44380 + }, + { + "epoch": 8.07, + "learning_rate": 5.79304696032035e-07, + "loss": 0.0953, + "step": 44381 + }, + { + "epoch": 8.07, + "learning_rate": 5.792500910083727e-07, + "loss": 3.4558, + "step": 44382 + }, + { + "epoch": 8.07, + "learning_rate": 5.791954859847107e-07, + "loss": 0.3205, + "step": 44383 + }, + { + "epoch": 8.07, + "learning_rate": 5.791408809610484e-07, + "loss": 0.9773, + "step": 44384 + }, + { + "epoch": 8.07, + "learning_rate": 5.790862759373862e-07, + "loss": 1.4326, + "step": 44385 + }, + { + "epoch": 8.07, + "learning_rate": 5.790316709137241e-07, + "loss": 2.4115, + "step": 44386 + }, + { + "epoch": 8.07, + "learning_rate": 5.789770658900619e-07, + "loss": 0.0103, + "step": 44387 + }, + { + "epoch": 8.07, + "learning_rate": 5.789224608663997e-07, + "loss": 0.2134, + "step": 44388 + }, + { + "epoch": 8.07, + "learning_rate": 5.788678558427375e-07, + "loss": 0.1184, + "step": 44389 + }, + { + "epoch": 8.07, + "learning_rate": 5.788132508190754e-07, + "loss": 0.0199, + "step": 44390 + }, + { + "epoch": 8.07, + "learning_rate": 5.787586457954132e-07, + "loss": 0.1805, + "step": 44391 + }, + { + "epoch": 8.07, + "learning_rate": 5.78704040771751e-07, + "loss": 0.1797, + "step": 44392 + }, + { + "epoch": 8.07, + "learning_rate": 5.786494357480888e-07, + "loss": 0.6204, + "step": 44393 + }, + { + "epoch": 8.07, + "learning_rate": 5.785948307244267e-07, + "loss": 0.2404, + "step": 44394 + }, + { + "epoch": 8.07, + "learning_rate": 5.785402257007645e-07, + "loss": 0.0807, + "step": 44395 + }, + { + "epoch": 8.07, + "learning_rate": 5.784856206771022e-07, + "loss": 0.3401, + "step": 44396 + }, + { + "epoch": 8.07, + "learning_rate": 5.784310156534402e-07, + "loss": 0.3506, + "step": 44397 + }, + { + "epoch": 8.07, + "learning_rate": 5.783764106297779e-07, + "loss": 0.0858, + "step": 44398 + }, + { + "epoch": 8.07, + "learning_rate": 5.783218056061157e-07, + "loss": 0.1967, + "step": 44399 + }, + { + "epoch": 8.07, + "learning_rate": 5.782672005824536e-07, + "loss": 0.2976, + "step": 44400 + }, + { + "epoch": 8.07, + "learning_rate": 5.782125955587914e-07, + "loss": 0.6084, + "step": 44401 + }, + { + "epoch": 8.07, + "learning_rate": 5.781579905351293e-07, + "loss": 1.9873, + "step": 44402 + }, + { + "epoch": 8.07, + "learning_rate": 5.781033855114671e-07, + "loss": 0.6335, + "step": 44403 + }, + { + "epoch": 8.07, + "learning_rate": 5.780487804878049e-07, + "loss": 1.6527, + "step": 44404 + }, + { + "epoch": 8.08, + "learning_rate": 5.779941754641427e-07, + "loss": 0.2115, + "step": 44405 + }, + { + "epoch": 8.08, + "learning_rate": 5.779395704404805e-07, + "loss": 0.7118, + "step": 44406 + }, + { + "epoch": 8.08, + "learning_rate": 5.778849654168184e-07, + "loss": 0.164, + "step": 44407 + }, + { + "epoch": 8.08, + "learning_rate": 5.778303603931562e-07, + "loss": 0.849, + "step": 44408 + }, + { + "epoch": 8.08, + "learning_rate": 5.77775755369494e-07, + "loss": 0.1909, + "step": 44409 + }, + { + "epoch": 8.08, + "learning_rate": 5.777211503458319e-07, + "loss": 0.051, + "step": 44410 + }, + { + "epoch": 8.08, + "learning_rate": 5.776665453221697e-07, + "loss": 0.579, + "step": 44411 + }, + { + "epoch": 8.08, + "learning_rate": 5.776119402985074e-07, + "loss": 2.8515, + "step": 44412 + }, + { + "epoch": 8.08, + "learning_rate": 5.775573352748453e-07, + "loss": 0.1918, + "step": 44413 + }, + { + "epoch": 8.08, + "learning_rate": 5.775027302511832e-07, + "loss": 0.0307, + "step": 44414 + }, + { + "epoch": 8.08, + "learning_rate": 5.774481252275209e-07, + "loss": 0.0675, + "step": 44415 + }, + { + "epoch": 8.08, + "learning_rate": 5.773935202038588e-07, + "loss": 0.0546, + "step": 44416 + }, + { + "epoch": 8.08, + "learning_rate": 5.773389151801966e-07, + "loss": 0.8805, + "step": 44417 + }, + { + "epoch": 8.08, + "learning_rate": 5.772843101565345e-07, + "loss": 0.1599, + "step": 44418 + }, + { + "epoch": 8.08, + "learning_rate": 5.772297051328722e-07, + "loss": 0.9844, + "step": 44419 + }, + { + "epoch": 8.08, + "learning_rate": 5.7717510010921e-07, + "loss": 0.4635, + "step": 44420 + }, + { + "epoch": 8.08, + "learning_rate": 5.771204950855479e-07, + "loss": 1.4548, + "step": 44421 + }, + { + "epoch": 8.08, + "learning_rate": 5.770658900618857e-07, + "loss": 0.4903, + "step": 44422 + }, + { + "epoch": 8.08, + "learning_rate": 5.770112850382235e-07, + "loss": 0.0182, + "step": 44423 + }, + { + "epoch": 8.08, + "learning_rate": 5.769566800145614e-07, + "loss": 0.7174, + "step": 44424 + }, + { + "epoch": 8.08, + "learning_rate": 5.769020749908992e-07, + "loss": 1.172, + "step": 44425 + }, + { + "epoch": 8.08, + "learning_rate": 5.768474699672369e-07, + "loss": 1.2501, + "step": 44426 + }, + { + "epoch": 8.08, + "learning_rate": 5.767928649435749e-07, + "loss": 0.0867, + "step": 44427 + }, + { + "epoch": 8.08, + "learning_rate": 5.767382599199126e-07, + "loss": 0.4755, + "step": 44428 + }, + { + "epoch": 8.08, + "learning_rate": 5.766836548962505e-07, + "loss": 0.3328, + "step": 44429 + }, + { + "epoch": 8.08, + "learning_rate": 5.766290498725883e-07, + "loss": 0.9088, + "step": 44430 + }, + { + "epoch": 8.08, + "learning_rate": 5.765744448489261e-07, + "loss": 0.1822, + "step": 44431 + }, + { + "epoch": 8.08, + "learning_rate": 5.76519839825264e-07, + "loss": 0.363, + "step": 44432 + }, + { + "epoch": 8.08, + "learning_rate": 5.764652348016017e-07, + "loss": 1.269, + "step": 44433 + }, + { + "epoch": 8.08, + "learning_rate": 5.764106297779396e-07, + "loss": 0.0048, + "step": 44434 + }, + { + "epoch": 8.08, + "learning_rate": 5.763560247542774e-07, + "loss": 1.1243, + "step": 44435 + }, + { + "epoch": 8.08, + "learning_rate": 5.763014197306152e-07, + "loss": 1.4442, + "step": 44436 + }, + { + "epoch": 8.08, + "learning_rate": 5.76246814706953e-07, + "loss": 0.1829, + "step": 44437 + }, + { + "epoch": 8.08, + "learning_rate": 5.761922096832909e-07, + "loss": 0.7195, + "step": 44438 + }, + { + "epoch": 8.08, + "learning_rate": 5.761376046596287e-07, + "loss": 0.0853, + "step": 44439 + }, + { + "epoch": 8.08, + "learning_rate": 5.760829996359665e-07, + "loss": 0.1195, + "step": 44440 + }, + { + "epoch": 8.08, + "learning_rate": 5.760283946123044e-07, + "loss": 1.1192, + "step": 44441 + }, + { + "epoch": 8.08, + "learning_rate": 5.759737895886421e-07, + "loss": 1.327, + "step": 44442 + }, + { + "epoch": 8.08, + "learning_rate": 5.7591918456498e-07, + "loss": 1.7118, + "step": 44443 + }, + { + "epoch": 8.08, + "learning_rate": 5.758645795413178e-07, + "loss": 1.742, + "step": 44444 + }, + { + "epoch": 8.08, + "learning_rate": 5.758099745176556e-07, + "loss": 0.7142, + "step": 44445 + }, + { + "epoch": 8.08, + "learning_rate": 5.757553694939935e-07, + "loss": 0.3383, + "step": 44446 + }, + { + "epoch": 8.08, + "learning_rate": 5.757007644703313e-07, + "loss": 0.9588, + "step": 44447 + }, + { + "epoch": 8.08, + "learning_rate": 5.756461594466692e-07, + "loss": 0.5128, + "step": 44448 + }, + { + "epoch": 8.08, + "learning_rate": 5.755915544230069e-07, + "loss": 0.0106, + "step": 44449 + }, + { + "epoch": 8.08, + "learning_rate": 5.755369493993447e-07, + "loss": 0.0088, + "step": 44450 + }, + { + "epoch": 8.08, + "learning_rate": 5.754823443756827e-07, + "loss": 0.7287, + "step": 44451 + }, + { + "epoch": 8.08, + "learning_rate": 5.754277393520204e-07, + "loss": 0.0098, + "step": 44452 + }, + { + "epoch": 8.08, + "learning_rate": 5.753731343283582e-07, + "loss": 0.1786, + "step": 44453 + }, + { + "epoch": 8.08, + "learning_rate": 5.753185293046961e-07, + "loss": 0.0143, + "step": 44454 + }, + { + "epoch": 8.08, + "learning_rate": 5.752639242810339e-07, + "loss": 0.0051, + "step": 44455 + }, + { + "epoch": 8.08, + "learning_rate": 5.752093192573716e-07, + "loss": 1.3004, + "step": 44456 + }, + { + "epoch": 8.08, + "learning_rate": 5.751547142337095e-07, + "loss": 0.0678, + "step": 44457 + }, + { + "epoch": 8.08, + "learning_rate": 5.751001092100473e-07, + "loss": 1.2163, + "step": 44458 + }, + { + "epoch": 8.08, + "learning_rate": 5.750455041863852e-07, + "loss": 0.0106, + "step": 44459 + }, + { + "epoch": 8.09, + "learning_rate": 5.74990899162723e-07, + "loss": 1.1529, + "step": 44460 + }, + { + "epoch": 8.09, + "learning_rate": 5.749362941390608e-07, + "loss": 0.0048, + "step": 44461 + }, + { + "epoch": 8.09, + "learning_rate": 5.748816891153987e-07, + "loss": 1.6658, + "step": 44462 + }, + { + "epoch": 8.09, + "learning_rate": 5.748270840917364e-07, + "loss": 4.5666, + "step": 44463 + }, + { + "epoch": 8.09, + "learning_rate": 5.747724790680742e-07, + "loss": 1.951, + "step": 44464 + }, + { + "epoch": 8.09, + "learning_rate": 5.747178740444121e-07, + "loss": 0.155, + "step": 44465 + }, + { + "epoch": 8.09, + "learning_rate": 5.746632690207499e-07, + "loss": 0.0173, + "step": 44466 + }, + { + "epoch": 8.09, + "learning_rate": 5.746086639970877e-07, + "loss": 0.0283, + "step": 44467 + }, + { + "epoch": 8.09, + "learning_rate": 5.745540589734256e-07, + "loss": 0.6897, + "step": 44468 + }, + { + "epoch": 8.09, + "learning_rate": 5.744994539497634e-07, + "loss": 0.1489, + "step": 44469 + }, + { + "epoch": 8.09, + "learning_rate": 5.744448489261012e-07, + "loss": 0.0286, + "step": 44470 + }, + { + "epoch": 8.09, + "learning_rate": 5.743902439024391e-07, + "loss": 0.0398, + "step": 44471 + }, + { + "epoch": 8.09, + "learning_rate": 5.743356388787768e-07, + "loss": 0.0293, + "step": 44472 + }, + { + "epoch": 8.09, + "learning_rate": 5.742810338551147e-07, + "loss": 0.2489, + "step": 44473 + }, + { + "epoch": 8.09, + "learning_rate": 5.742264288314525e-07, + "loss": 0.0047, + "step": 44474 + }, + { + "epoch": 8.09, + "learning_rate": 5.741718238077903e-07, + "loss": 0.0545, + "step": 44475 + }, + { + "epoch": 8.09, + "learning_rate": 5.741172187841282e-07, + "loss": 0.0068, + "step": 44476 + }, + { + "epoch": 8.09, + "learning_rate": 5.740626137604659e-07, + "loss": 0.7897, + "step": 44477 + }, + { + "epoch": 8.09, + "learning_rate": 5.740080087368039e-07, + "loss": 0.2934, + "step": 44478 + }, + { + "epoch": 8.09, + "learning_rate": 5.739534037131416e-07, + "loss": 0.1509, + "step": 44479 + }, + { + "epoch": 8.09, + "learning_rate": 5.738987986894794e-07, + "loss": 0.1808, + "step": 44480 + }, + { + "epoch": 8.09, + "learning_rate": 5.738441936658173e-07, + "loss": 0.123, + "step": 44481 + }, + { + "epoch": 8.09, + "learning_rate": 5.737895886421551e-07, + "loss": 2.2691, + "step": 44482 + }, + { + "epoch": 8.09, + "learning_rate": 5.737349836184929e-07, + "loss": 0.0121, + "step": 44483 + }, + { + "epoch": 8.09, + "learning_rate": 5.736803785948307e-07, + "loss": 3.2936, + "step": 44484 + }, + { + "epoch": 8.09, + "learning_rate": 5.736257735711686e-07, + "loss": 1.4382, + "step": 44485 + }, + { + "epoch": 8.09, + "learning_rate": 5.735711685475063e-07, + "loss": 0.9734, + "step": 44486 + }, + { + "epoch": 8.09, + "learning_rate": 5.735165635238442e-07, + "loss": 1.5005, + "step": 44487 + }, + { + "epoch": 8.09, + "learning_rate": 5.73461958500182e-07, + "loss": 3.3718, + "step": 44488 + }, + { + "epoch": 8.09, + "learning_rate": 5.734073534765199e-07, + "loss": 3.0903, + "step": 44489 + }, + { + "epoch": 8.09, + "learning_rate": 5.733527484528577e-07, + "loss": 1.91, + "step": 44490 + }, + { + "epoch": 8.09, + "learning_rate": 5.732981434291955e-07, + "loss": 0.3577, + "step": 44491 + }, + { + "epoch": 8.09, + "learning_rate": 5.732435384055334e-07, + "loss": 1.3668, + "step": 44492 + }, + { + "epoch": 8.09, + "learning_rate": 5.731889333818711e-07, + "loss": 0.0055, + "step": 44493 + }, + { + "epoch": 8.09, + "learning_rate": 5.731343283582089e-07, + "loss": 0.0034, + "step": 44494 + }, + { + "epoch": 8.09, + "learning_rate": 5.730797233345468e-07, + "loss": 0.1844, + "step": 44495 + }, + { + "epoch": 8.09, + "learning_rate": 5.730251183108846e-07, + "loss": 0.1421, + "step": 44496 + }, + { + "epoch": 8.09, + "learning_rate": 5.729705132872224e-07, + "loss": 0.1725, + "step": 44497 + }, + { + "epoch": 8.09, + "learning_rate": 5.729159082635603e-07, + "loss": 0.088, + "step": 44498 + }, + { + "epoch": 8.09, + "learning_rate": 5.728613032398981e-07, + "loss": 0.3534, + "step": 44499 + }, + { + "epoch": 8.09, + "learning_rate": 5.728066982162359e-07, + "loss": 0.0137, + "step": 44500 + }, + { + "epoch": 8.09, + "learning_rate": 5.727520931925737e-07, + "loss": 0.3716, + "step": 44501 + }, + { + "epoch": 8.09, + "learning_rate": 5.726974881689115e-07, + "loss": 1.1691, + "step": 44502 + }, + { + "epoch": 8.09, + "learning_rate": 5.726428831452494e-07, + "loss": 0.0041, + "step": 44503 + }, + { + "epoch": 8.09, + "learning_rate": 5.725882781215872e-07, + "loss": 0.0037, + "step": 44504 + }, + { + "epoch": 8.09, + "learning_rate": 5.72533673097925e-07, + "loss": 1.0907, + "step": 44505 + }, + { + "epoch": 8.09, + "learning_rate": 5.724790680742629e-07, + "loss": 0.1028, + "step": 44506 + }, + { + "epoch": 8.09, + "learning_rate": 5.724244630506006e-07, + "loss": 1.6685, + "step": 44507 + }, + { + "epoch": 8.09, + "learning_rate": 5.723698580269385e-07, + "loss": 2.818, + "step": 44508 + }, + { + "epoch": 8.09, + "learning_rate": 5.723152530032763e-07, + "loss": 1.5399, + "step": 44509 + }, + { + "epoch": 8.09, + "learning_rate": 5.722606479796141e-07, + "loss": 0.8425, + "step": 44510 + }, + { + "epoch": 8.09, + "learning_rate": 5.72206042955952e-07, + "loss": 0.0093, + "step": 44511 + }, + { + "epoch": 8.09, + "learning_rate": 5.721514379322898e-07, + "loss": 0.2922, + "step": 44512 + }, + { + "epoch": 8.09, + "learning_rate": 5.720968329086276e-07, + "loss": 1.2037, + "step": 44513 + }, + { + "epoch": 8.09, + "learning_rate": 5.720422278849654e-07, + "loss": 1.2431, + "step": 44514 + }, + { + "epoch": 8.1, + "learning_rate": 5.719876228613033e-07, + "loss": 1.923, + "step": 44515 + }, + { + "epoch": 8.1, + "learning_rate": 5.71933017837641e-07, + "loss": 0.7568, + "step": 44516 + }, + { + "epoch": 8.1, + "learning_rate": 5.718784128139789e-07, + "loss": 0.7096, + "step": 44517 + }, + { + "epoch": 8.1, + "learning_rate": 5.718238077903167e-07, + "loss": 0.9529, + "step": 44518 + }, + { + "epoch": 8.1, + "learning_rate": 5.717692027666546e-07, + "loss": 1.4961, + "step": 44519 + }, + { + "epoch": 8.1, + "learning_rate": 5.717145977429924e-07, + "loss": 0.059, + "step": 44520 + }, + { + "epoch": 8.1, + "learning_rate": 5.716599927193301e-07, + "loss": 0.4009, + "step": 44521 + }, + { + "epoch": 8.1, + "learning_rate": 5.716053876956681e-07, + "loss": 0.2101, + "step": 44522 + }, + { + "epoch": 8.1, + "learning_rate": 5.715507826720058e-07, + "loss": 0.2279, + "step": 44523 + }, + { + "epoch": 8.1, + "learning_rate": 5.714961776483436e-07, + "loss": 0.2832, + "step": 44524 + }, + { + "epoch": 8.1, + "learning_rate": 5.714415726246815e-07, + "loss": 0.9751, + "step": 44525 + }, + { + "epoch": 8.1, + "learning_rate": 5.713869676010193e-07, + "loss": 0.547, + "step": 44526 + }, + { + "epoch": 8.1, + "learning_rate": 5.713323625773572e-07, + "loss": 0.0049, + "step": 44527 + }, + { + "epoch": 8.1, + "learning_rate": 5.712777575536949e-07, + "loss": 0.8949, + "step": 44528 + }, + { + "epoch": 8.1, + "learning_rate": 5.712231525300328e-07, + "loss": 0.0391, + "step": 44529 + }, + { + "epoch": 8.1, + "learning_rate": 5.711685475063706e-07, + "loss": 1.9735, + "step": 44530 + }, + { + "epoch": 8.1, + "learning_rate": 5.711139424827084e-07, + "loss": 2.818, + "step": 44531 + }, + { + "epoch": 8.1, + "learning_rate": 5.710593374590463e-07, + "loss": 0.2142, + "step": 44532 + }, + { + "epoch": 8.1, + "learning_rate": 5.710047324353841e-07, + "loss": 1.0958, + "step": 44533 + }, + { + "epoch": 8.1, + "learning_rate": 5.709501274117219e-07, + "loss": 0.0476, + "step": 44534 + }, + { + "epoch": 8.1, + "learning_rate": 5.708955223880597e-07, + "loss": 0.0828, + "step": 44535 + }, + { + "epoch": 8.1, + "learning_rate": 5.708409173643976e-07, + "loss": 0.0202, + "step": 44536 + }, + { + "epoch": 8.1, + "learning_rate": 5.707863123407353e-07, + "loss": 0.0377, + "step": 44537 + }, + { + "epoch": 8.1, + "learning_rate": 5.707317073170732e-07, + "loss": 0.0606, + "step": 44538 + }, + { + "epoch": 8.1, + "learning_rate": 5.70677102293411e-07, + "loss": 0.6316, + "step": 44539 + }, + { + "epoch": 8.1, + "learning_rate": 5.706224972697488e-07, + "loss": 0.361, + "step": 44540 + }, + { + "epoch": 8.1, + "learning_rate": 5.705678922460867e-07, + "loss": 0.5999, + "step": 44541 + }, + { + "epoch": 8.1, + "learning_rate": 5.705132872224245e-07, + "loss": 0.5966, + "step": 44542 + }, + { + "epoch": 8.1, + "learning_rate": 5.704586821987623e-07, + "loss": 0.1884, + "step": 44543 + }, + { + "epoch": 8.1, + "learning_rate": 5.704040771751001e-07, + "loss": 0.0088, + "step": 44544 + }, + { + "epoch": 8.1, + "learning_rate": 5.703494721514379e-07, + "loss": 0.0062, + "step": 44545 + }, + { + "epoch": 8.1, + "learning_rate": 5.702948671277757e-07, + "loss": 0.0105, + "step": 44546 + }, + { + "epoch": 8.1, + "learning_rate": 5.702402621041136e-07, + "loss": 0.3405, + "step": 44547 + }, + { + "epoch": 8.1, + "learning_rate": 5.701856570804514e-07, + "loss": 0.1615, + "step": 44548 + }, + { + "epoch": 8.1, + "learning_rate": 5.701310520567893e-07, + "loss": 1.5022, + "step": 44549 + }, + { + "epoch": 8.1, + "learning_rate": 5.700764470331271e-07, + "loss": 0.0186, + "step": 44550 + }, + { + "epoch": 8.1, + "learning_rate": 5.700218420094648e-07, + "loss": 0.1655, + "step": 44551 + }, + { + "epoch": 8.1, + "learning_rate": 5.699672369858028e-07, + "loss": 0.0035, + "step": 44552 + }, + { + "epoch": 8.1, + "learning_rate": 5.699126319621405e-07, + "loss": 0.0706, + "step": 44553 + }, + { + "epoch": 8.1, + "learning_rate": 5.698580269384783e-07, + "loss": 0.8454, + "step": 44554 + }, + { + "epoch": 8.1, + "learning_rate": 5.698034219148162e-07, + "loss": 0.073, + "step": 44555 + }, + { + "epoch": 8.1, + "learning_rate": 5.69748816891154e-07, + "loss": 2.6793, + "step": 44556 + }, + { + "epoch": 8.1, + "learning_rate": 5.696942118674919e-07, + "loss": 1.6493, + "step": 44557 + }, + { + "epoch": 8.1, + "learning_rate": 5.696396068438296e-07, + "loss": 0.0864, + "step": 44558 + }, + { + "epoch": 8.1, + "learning_rate": 5.695850018201675e-07, + "loss": 0.6822, + "step": 44559 + }, + { + "epoch": 8.1, + "learning_rate": 5.695303967965053e-07, + "loss": 0.0621, + "step": 44560 + }, + { + "epoch": 8.1, + "learning_rate": 5.694757917728431e-07, + "loss": 0.0135, + "step": 44561 + }, + { + "epoch": 8.1, + "learning_rate": 5.694211867491809e-07, + "loss": 0.1972, + "step": 44562 + }, + { + "epoch": 8.1, + "learning_rate": 5.693665817255188e-07, + "loss": 0.026, + "step": 44563 + }, + { + "epoch": 8.1, + "learning_rate": 5.693119767018566e-07, + "loss": 0.4664, + "step": 44564 + }, + { + "epoch": 8.1, + "learning_rate": 5.692573716781943e-07, + "loss": 2.9419, + "step": 44565 + }, + { + "epoch": 8.1, + "learning_rate": 5.692027666545323e-07, + "loss": 0.072, + "step": 44566 + }, + { + "epoch": 8.1, + "learning_rate": 5.6914816163087e-07, + "loss": 1.2743, + "step": 44567 + }, + { + "epoch": 8.1, + "learning_rate": 5.690935566072079e-07, + "loss": 0.8002, + "step": 44568 + }, + { + "epoch": 8.1, + "learning_rate": 5.690389515835457e-07, + "loss": 0.8871, + "step": 44569 + }, + { + "epoch": 8.11, + "learning_rate": 5.689843465598835e-07, + "loss": 0.0026, + "step": 44570 + }, + { + "epoch": 8.11, + "learning_rate": 5.689297415362214e-07, + "loss": 0.6442, + "step": 44571 + }, + { + "epoch": 8.11, + "learning_rate": 5.688751365125592e-07, + "loss": 0.1234, + "step": 44572 + }, + { + "epoch": 8.11, + "learning_rate": 5.68820531488897e-07, + "loss": 0.8263, + "step": 44573 + }, + { + "epoch": 8.11, + "learning_rate": 5.687659264652348e-07, + "loss": 0.1961, + "step": 44574 + }, + { + "epoch": 8.11, + "learning_rate": 5.687113214415726e-07, + "loss": 0.2582, + "step": 44575 + }, + { + "epoch": 8.11, + "learning_rate": 5.686567164179104e-07, + "loss": 0.4092, + "step": 44576 + }, + { + "epoch": 8.11, + "learning_rate": 5.686021113942483e-07, + "loss": 2.2663, + "step": 44577 + }, + { + "epoch": 8.11, + "learning_rate": 5.685475063705861e-07, + "loss": 0.0069, + "step": 44578 + }, + { + "epoch": 8.11, + "learning_rate": 5.68492901346924e-07, + "loss": 0.1267, + "step": 44579 + }, + { + "epoch": 8.11, + "learning_rate": 5.684382963232618e-07, + "loss": 1.187, + "step": 44580 + }, + { + "epoch": 8.11, + "learning_rate": 5.683836912995995e-07, + "loss": 0.4944, + "step": 44581 + }, + { + "epoch": 8.11, + "learning_rate": 5.683290862759374e-07, + "loss": 0.0048, + "step": 44582 + }, + { + "epoch": 8.11, + "learning_rate": 5.682744812522752e-07, + "loss": 0.0432, + "step": 44583 + }, + { + "epoch": 8.11, + "learning_rate": 5.68219876228613e-07, + "loss": 1.5617, + "step": 44584 + }, + { + "epoch": 8.11, + "learning_rate": 5.681652712049509e-07, + "loss": 0.0079, + "step": 44585 + }, + { + "epoch": 8.11, + "learning_rate": 5.681106661812887e-07, + "loss": 0.1835, + "step": 44586 + }, + { + "epoch": 8.11, + "learning_rate": 5.680560611576266e-07, + "loss": 0.1751, + "step": 44587 + }, + { + "epoch": 8.11, + "learning_rate": 5.680014561339643e-07, + "loss": 1.8481, + "step": 44588 + }, + { + "epoch": 8.11, + "learning_rate": 5.679468511103021e-07, + "loss": 0.0901, + "step": 44589 + }, + { + "epoch": 8.11, + "learning_rate": 5.6789224608664e-07, + "loss": 0.1556, + "step": 44590 + }, + { + "epoch": 8.11, + "learning_rate": 5.678376410629778e-07, + "loss": 0.3383, + "step": 44591 + }, + { + "epoch": 8.11, + "learning_rate": 5.677830360393156e-07, + "loss": 0.1129, + "step": 44592 + }, + { + "epoch": 8.11, + "learning_rate": 5.677284310156535e-07, + "loss": 0.2407, + "step": 44593 + }, + { + "epoch": 8.11, + "learning_rate": 5.676738259919913e-07, + "loss": 0.0942, + "step": 44594 + }, + { + "epoch": 8.11, + "learning_rate": 5.67619220968329e-07, + "loss": 0.009, + "step": 44595 + }, + { + "epoch": 8.11, + "learning_rate": 5.67564615944667e-07, + "loss": 0.8537, + "step": 44596 + }, + { + "epoch": 8.11, + "learning_rate": 5.675100109210047e-07, + "loss": 0.9169, + "step": 44597 + }, + { + "epoch": 8.11, + "learning_rate": 5.674554058973426e-07, + "loss": 0.0277, + "step": 44598 + }, + { + "epoch": 8.11, + "learning_rate": 5.674008008736804e-07, + "loss": 2.2465, + "step": 44599 + }, + { + "epoch": 8.11, + "learning_rate": 5.673461958500182e-07, + "loss": 0.0139, + "step": 44600 + }, + { + "epoch": 8.11, + "learning_rate": 5.672915908263561e-07, + "loss": 1.4869, + "step": 44601 + }, + { + "epoch": 8.11, + "learning_rate": 5.672369858026938e-07, + "loss": 0.007, + "step": 44602 + }, + { + "epoch": 8.11, + "learning_rate": 5.671823807790317e-07, + "loss": 0.0047, + "step": 44603 + }, + { + "epoch": 8.11, + "learning_rate": 5.671277757553695e-07, + "loss": 0.4, + "step": 44604 + }, + { + "epoch": 8.11, + "learning_rate": 5.670731707317073e-07, + "loss": 0.0071, + "step": 44605 + }, + { + "epoch": 8.11, + "learning_rate": 5.670185657080452e-07, + "loss": 0.0738, + "step": 44606 + }, + { + "epoch": 8.11, + "learning_rate": 5.66963960684383e-07, + "loss": 0.0448, + "step": 44607 + }, + { + "epoch": 8.11, + "learning_rate": 5.669093556607208e-07, + "loss": 0.2906, + "step": 44608 + }, + { + "epoch": 8.11, + "learning_rate": 5.668547506370586e-07, + "loss": 0.0121, + "step": 44609 + }, + { + "epoch": 8.11, + "learning_rate": 5.668001456133965e-07, + "loss": 0.6885, + "step": 44610 + }, + { + "epoch": 8.11, + "learning_rate": 5.667455405897342e-07, + "loss": 0.6048, + "step": 44611 + }, + { + "epoch": 8.11, + "learning_rate": 5.666909355660721e-07, + "loss": 0.1128, + "step": 44612 + }, + { + "epoch": 8.11, + "learning_rate": 5.666363305424099e-07, + "loss": 0.364, + "step": 44613 + }, + { + "epoch": 8.11, + "learning_rate": 5.665817255187477e-07, + "loss": 0.469, + "step": 44614 + }, + { + "epoch": 8.11, + "learning_rate": 5.665271204950856e-07, + "loss": 0.0257, + "step": 44615 + }, + { + "epoch": 8.11, + "learning_rate": 5.664725154714234e-07, + "loss": 0.3224, + "step": 44616 + }, + { + "epoch": 8.11, + "learning_rate": 5.664179104477613e-07, + "loss": 0.3619, + "step": 44617 + }, + { + "epoch": 8.11, + "learning_rate": 5.66363305424099e-07, + "loss": 1.3728, + "step": 44618 + }, + { + "epoch": 8.11, + "learning_rate": 5.663087004004368e-07, + "loss": 0.0134, + "step": 44619 + }, + { + "epoch": 8.11, + "learning_rate": 5.662540953767747e-07, + "loss": 0.2159, + "step": 44620 + }, + { + "epoch": 8.11, + "learning_rate": 5.661994903531125e-07, + "loss": 0.0132, + "step": 44621 + }, + { + "epoch": 8.11, + "learning_rate": 5.661448853294503e-07, + "loss": 1.7629, + "step": 44622 + }, + { + "epoch": 8.11, + "learning_rate": 5.660902803057882e-07, + "loss": 0.7221, + "step": 44623 + }, + { + "epoch": 8.11, + "learning_rate": 5.66035675282126e-07, + "loss": 0.8526, + "step": 44624 + }, + { + "epoch": 8.12, + "learning_rate": 5.659810702584637e-07, + "loss": 0.3401, + "step": 44625 + }, + { + "epoch": 8.12, + "learning_rate": 5.659264652348016e-07, + "loss": 0.0062, + "step": 44626 + }, + { + "epoch": 8.12, + "learning_rate": 5.658718602111394e-07, + "loss": 0.3805, + "step": 44627 + }, + { + "epoch": 8.12, + "learning_rate": 5.658172551874773e-07, + "loss": 0.2163, + "step": 44628 + }, + { + "epoch": 8.12, + "learning_rate": 5.657626501638151e-07, + "loss": 0.1448, + "step": 44629 + }, + { + "epoch": 8.12, + "learning_rate": 5.657080451401529e-07, + "loss": 0.153, + "step": 44630 + }, + { + "epoch": 8.12, + "learning_rate": 5.656534401164908e-07, + "loss": 2.5267, + "step": 44631 + }, + { + "epoch": 8.12, + "learning_rate": 5.655988350928285e-07, + "loss": 1.8323, + "step": 44632 + }, + { + "epoch": 8.12, + "learning_rate": 5.655442300691663e-07, + "loss": 0.0087, + "step": 44633 + }, + { + "epoch": 8.12, + "learning_rate": 5.654896250455042e-07, + "loss": 0.5372, + "step": 44634 + }, + { + "epoch": 8.12, + "learning_rate": 5.65435020021842e-07, + "loss": 1.7311, + "step": 44635 + }, + { + "epoch": 8.12, + "learning_rate": 5.653804149981799e-07, + "loss": 0.0261, + "step": 44636 + }, + { + "epoch": 8.12, + "learning_rate": 5.653258099745177e-07, + "loss": 0.401, + "step": 44637 + }, + { + "epoch": 8.12, + "learning_rate": 5.652712049508555e-07, + "loss": 0.3672, + "step": 44638 + }, + { + "epoch": 8.12, + "learning_rate": 5.652165999271933e-07, + "loss": 0.0154, + "step": 44639 + }, + { + "epoch": 8.12, + "learning_rate": 5.651619949035312e-07, + "loss": 0.1497, + "step": 44640 + }, + { + "epoch": 8.12, + "learning_rate": 5.651073898798689e-07, + "loss": 1.6062, + "step": 44641 + }, + { + "epoch": 8.12, + "learning_rate": 5.650527848562068e-07, + "loss": 0.1481, + "step": 44642 + }, + { + "epoch": 8.12, + "learning_rate": 5.649981798325446e-07, + "loss": 0.0059, + "step": 44643 + }, + { + "epoch": 8.12, + "learning_rate": 5.649435748088824e-07, + "loss": 0.3601, + "step": 44644 + }, + { + "epoch": 8.12, + "learning_rate": 5.648889697852203e-07, + "loss": 0.3952, + "step": 44645 + }, + { + "epoch": 8.12, + "learning_rate": 5.64834364761558e-07, + "loss": 0.0101, + "step": 44646 + }, + { + "epoch": 8.12, + "learning_rate": 5.64779759737896e-07, + "loss": 0.3597, + "step": 44647 + }, + { + "epoch": 8.12, + "learning_rate": 5.647251547142337e-07, + "loss": 1.4967, + "step": 44648 + }, + { + "epoch": 8.12, + "learning_rate": 5.646705496905715e-07, + "loss": 0.1699, + "step": 44649 + }, + { + "epoch": 8.12, + "learning_rate": 5.646159446669094e-07, + "loss": 0.2855, + "step": 44650 + }, + { + "epoch": 8.12, + "learning_rate": 5.645613396432472e-07, + "loss": 0.4321, + "step": 44651 + }, + { + "epoch": 8.12, + "learning_rate": 5.64506734619585e-07, + "loss": 0.1078, + "step": 44652 + }, + { + "epoch": 8.12, + "learning_rate": 5.644521295959228e-07, + "loss": 2.7082, + "step": 44653 + }, + { + "epoch": 8.12, + "learning_rate": 5.643975245722607e-07, + "loss": 1.5689, + "step": 44654 + }, + { + "epoch": 8.12, + "learning_rate": 5.643429195485984e-07, + "loss": 0.3655, + "step": 44655 + }, + { + "epoch": 8.12, + "learning_rate": 5.642883145249363e-07, + "loss": 0.0054, + "step": 44656 + }, + { + "epoch": 8.12, + "learning_rate": 5.642337095012741e-07, + "loss": 0.0427, + "step": 44657 + }, + { + "epoch": 8.12, + "learning_rate": 5.64179104477612e-07, + "loss": 0.0114, + "step": 44658 + }, + { + "epoch": 8.12, + "learning_rate": 5.641244994539498e-07, + "loss": 0.2981, + "step": 44659 + }, + { + "epoch": 8.12, + "learning_rate": 5.640698944302876e-07, + "loss": 0.93, + "step": 44660 + }, + { + "epoch": 8.12, + "learning_rate": 5.640152894066255e-07, + "loss": 0.0538, + "step": 44661 + }, + { + "epoch": 8.12, + "learning_rate": 5.639606843829632e-07, + "loss": 0.37, + "step": 44662 + }, + { + "epoch": 8.12, + "learning_rate": 5.63906079359301e-07, + "loss": 2.0295, + "step": 44663 + }, + { + "epoch": 8.12, + "learning_rate": 5.638514743356389e-07, + "loss": 0.2476, + "step": 44664 + }, + { + "epoch": 8.12, + "learning_rate": 5.637968693119767e-07, + "loss": 0.1162, + "step": 44665 + }, + { + "epoch": 8.12, + "learning_rate": 5.637422642883146e-07, + "loss": 0.5666, + "step": 44666 + }, + { + "epoch": 8.12, + "learning_rate": 5.636876592646524e-07, + "loss": 0.3177, + "step": 44667 + }, + { + "epoch": 8.12, + "learning_rate": 5.636330542409902e-07, + "loss": 0.1906, + "step": 44668 + }, + { + "epoch": 8.12, + "learning_rate": 5.63578449217328e-07, + "loss": 0.0947, + "step": 44669 + }, + { + "epoch": 8.12, + "learning_rate": 5.635238441936658e-07, + "loss": 1.7452, + "step": 44670 + }, + { + "epoch": 8.12, + "learning_rate": 5.634692391700036e-07, + "loss": 0.2476, + "step": 44671 + }, + { + "epoch": 8.12, + "learning_rate": 5.634146341463415e-07, + "loss": 0.1476, + "step": 44672 + }, + { + "epoch": 8.12, + "learning_rate": 5.633600291226793e-07, + "loss": 0.4244, + "step": 44673 + }, + { + "epoch": 8.12, + "learning_rate": 5.633054240990171e-07, + "loss": 0.0393, + "step": 44674 + }, + { + "epoch": 8.12, + "learning_rate": 5.63250819075355e-07, + "loss": 0.7367, + "step": 44675 + }, + { + "epoch": 8.12, + "learning_rate": 5.631962140516927e-07, + "loss": 0.4511, + "step": 44676 + }, + { + "epoch": 8.12, + "learning_rate": 5.631416090280306e-07, + "loss": 0.2368, + "step": 44677 + }, + { + "epoch": 8.12, + "learning_rate": 5.630870040043684e-07, + "loss": 0.0637, + "step": 44678 + }, + { + "epoch": 8.12, + "learning_rate": 5.630323989807062e-07, + "loss": 0.111, + "step": 44679 + }, + { + "epoch": 8.13, + "learning_rate": 5.629777939570441e-07, + "loss": 1.1164, + "step": 44680 + }, + { + "epoch": 8.13, + "learning_rate": 5.629231889333819e-07, + "loss": 0.1854, + "step": 44681 + }, + { + "epoch": 8.13, + "learning_rate": 5.628685839097197e-07, + "loss": 0.2243, + "step": 44682 + }, + { + "epoch": 8.13, + "learning_rate": 5.628139788860575e-07, + "loss": 0.0143, + "step": 44683 + }, + { + "epoch": 8.13, + "learning_rate": 5.627593738623954e-07, + "loss": 0.5644, + "step": 44684 + }, + { + "epoch": 8.13, + "learning_rate": 5.627047688387332e-07, + "loss": 0.8419, + "step": 44685 + }, + { + "epoch": 8.13, + "learning_rate": 5.62650163815071e-07, + "loss": 0.0541, + "step": 44686 + }, + { + "epoch": 8.13, + "learning_rate": 5.625955587914088e-07, + "loss": 1.4614, + "step": 44687 + }, + { + "epoch": 8.13, + "learning_rate": 5.625409537677467e-07, + "loss": 0.8123, + "step": 44688 + }, + { + "epoch": 8.13, + "learning_rate": 5.624863487440845e-07, + "loss": 0.0253, + "step": 44689 + }, + { + "epoch": 8.13, + "learning_rate": 5.624317437204222e-07, + "loss": 1.1462, + "step": 44690 + }, + { + "epoch": 8.13, + "learning_rate": 5.623771386967602e-07, + "loss": 0.0245, + "step": 44691 + }, + { + "epoch": 8.13, + "learning_rate": 5.623225336730979e-07, + "loss": 0.4378, + "step": 44692 + }, + { + "epoch": 8.13, + "learning_rate": 5.622679286494357e-07, + "loss": 0.029, + "step": 44693 + }, + { + "epoch": 8.13, + "learning_rate": 5.622133236257736e-07, + "loss": 0.3605, + "step": 44694 + }, + { + "epoch": 8.13, + "learning_rate": 5.621587186021114e-07, + "loss": 0.6754, + "step": 44695 + }, + { + "epoch": 8.13, + "learning_rate": 5.621041135784493e-07, + "loss": 0.3104, + "step": 44696 + }, + { + "epoch": 8.13, + "learning_rate": 5.62049508554787e-07, + "loss": 0.0276, + "step": 44697 + }, + { + "epoch": 8.13, + "learning_rate": 5.619949035311249e-07, + "loss": 0.4485, + "step": 44698 + }, + { + "epoch": 8.13, + "learning_rate": 5.619402985074627e-07, + "loss": 2.9692, + "step": 44699 + }, + { + "epoch": 8.13, + "learning_rate": 5.618856934838005e-07, + "loss": 1.1481, + "step": 44700 + }, + { + "epoch": 8.13, + "learning_rate": 5.618310884601383e-07, + "loss": 1.1335, + "step": 44701 + }, + { + "epoch": 8.13, + "learning_rate": 5.617764834364762e-07, + "loss": 0.4796, + "step": 44702 + }, + { + "epoch": 8.13, + "learning_rate": 5.61721878412814e-07, + "loss": 0.8315, + "step": 44703 + }, + { + "epoch": 8.13, + "learning_rate": 5.616672733891519e-07, + "loss": 0.2519, + "step": 44704 + }, + { + "epoch": 8.13, + "learning_rate": 5.616126683654897e-07, + "loss": 1.463, + "step": 44705 + }, + { + "epoch": 8.13, + "learning_rate": 5.615580633418274e-07, + "loss": 0.1734, + "step": 44706 + }, + { + "epoch": 8.13, + "learning_rate": 5.615034583181653e-07, + "loss": 0.0056, + "step": 44707 + }, + { + "epoch": 8.13, + "learning_rate": 5.614488532945031e-07, + "loss": 0.399, + "step": 44708 + }, + { + "epoch": 8.13, + "learning_rate": 5.613942482708409e-07, + "loss": 0.151, + "step": 44709 + }, + { + "epoch": 8.13, + "learning_rate": 5.613396432471788e-07, + "loss": 0.0394, + "step": 44710 + }, + { + "epoch": 8.13, + "learning_rate": 5.612850382235166e-07, + "loss": 1.3909, + "step": 44711 + }, + { + "epoch": 8.13, + "learning_rate": 5.612304331998544e-07, + "loss": 0.1133, + "step": 44712 + }, + { + "epoch": 8.13, + "learning_rate": 5.611758281761922e-07, + "loss": 0.6777, + "step": 44713 + }, + { + "epoch": 8.13, + "learning_rate": 5.6112122315253e-07, + "loss": 0.1605, + "step": 44714 + }, + { + "epoch": 8.13, + "learning_rate": 5.610666181288679e-07, + "loss": 1.463, + "step": 44715 + }, + { + "epoch": 8.13, + "learning_rate": 5.610120131052057e-07, + "loss": 0.6912, + "step": 44716 + }, + { + "epoch": 8.13, + "learning_rate": 5.609574080815435e-07, + "loss": 0.0372, + "step": 44717 + }, + { + "epoch": 8.13, + "learning_rate": 5.609028030578814e-07, + "loss": 0.2282, + "step": 44718 + }, + { + "epoch": 8.13, + "learning_rate": 5.608481980342192e-07, + "loss": 0.0165, + "step": 44719 + }, + { + "epoch": 8.13, + "learning_rate": 5.607935930105569e-07, + "loss": 1.6853, + "step": 44720 + }, + { + "epoch": 8.13, + "learning_rate": 5.607389879868949e-07, + "loss": 0.0137, + "step": 44721 + }, + { + "epoch": 8.13, + "learning_rate": 5.606843829632326e-07, + "loss": 0.4443, + "step": 44722 + }, + { + "epoch": 8.13, + "learning_rate": 5.606297779395704e-07, + "loss": 0.006, + "step": 44723 + }, + { + "epoch": 8.13, + "learning_rate": 5.605751729159083e-07, + "loss": 0.1271, + "step": 44724 + }, + { + "epoch": 8.13, + "learning_rate": 5.605205678922461e-07, + "loss": 0.0147, + "step": 44725 + }, + { + "epoch": 8.13, + "learning_rate": 5.60465962868584e-07, + "loss": 0.347, + "step": 44726 + }, + { + "epoch": 8.13, + "learning_rate": 5.604113578449217e-07, + "loss": 0.5048, + "step": 44727 + }, + { + "epoch": 8.13, + "learning_rate": 5.603567528212596e-07, + "loss": 0.0663, + "step": 44728 + }, + { + "epoch": 8.13, + "learning_rate": 5.603021477975974e-07, + "loss": 0.1021, + "step": 44729 + }, + { + "epoch": 8.13, + "learning_rate": 5.602475427739352e-07, + "loss": 0.0093, + "step": 44730 + }, + { + "epoch": 8.13, + "learning_rate": 5.60192937750273e-07, + "loss": 0.9517, + "step": 44731 + }, + { + "epoch": 8.13, + "learning_rate": 5.601383327266109e-07, + "loss": 0.2972, + "step": 44732 + }, + { + "epoch": 8.13, + "learning_rate": 5.600837277029487e-07, + "loss": 0.068, + "step": 44733 + }, + { + "epoch": 8.13, + "learning_rate": 5.600291226792864e-07, + "loss": 0.1505, + "step": 44734 + }, + { + "epoch": 8.14, + "learning_rate": 5.599745176556244e-07, + "loss": 0.3177, + "step": 44735 + }, + { + "epoch": 8.14, + "learning_rate": 5.599199126319621e-07, + "loss": 1.4734, + "step": 44736 + }, + { + "epoch": 8.14, + "learning_rate": 5.598653076083e-07, + "loss": 0.2164, + "step": 44737 + }, + { + "epoch": 8.14, + "learning_rate": 5.598107025846378e-07, + "loss": 0.0559, + "step": 44738 + }, + { + "epoch": 8.14, + "learning_rate": 5.597560975609756e-07, + "loss": 1.2848, + "step": 44739 + }, + { + "epoch": 8.14, + "learning_rate": 5.597014925373135e-07, + "loss": 0.1924, + "step": 44740 + }, + { + "epoch": 8.14, + "learning_rate": 5.596468875136512e-07, + "loss": 0.0873, + "step": 44741 + }, + { + "epoch": 8.14, + "learning_rate": 5.595922824899891e-07, + "loss": 0.0106, + "step": 44742 + }, + { + "epoch": 8.14, + "learning_rate": 5.595376774663269e-07, + "loss": 0.0125, + "step": 44743 + }, + { + "epoch": 8.14, + "learning_rate": 5.594830724426647e-07, + "loss": 0.0373, + "step": 44744 + }, + { + "epoch": 8.14, + "learning_rate": 5.594284674190026e-07, + "loss": 0.3894, + "step": 44745 + }, + { + "epoch": 8.14, + "learning_rate": 5.593738623953404e-07, + "loss": 0.9424, + "step": 44746 + }, + { + "epoch": 8.14, + "learning_rate": 5.593192573716782e-07, + "loss": 0.1166, + "step": 44747 + }, + { + "epoch": 8.14, + "learning_rate": 5.592646523480161e-07, + "loss": 0.8288, + "step": 44748 + }, + { + "epoch": 8.14, + "learning_rate": 5.592100473243539e-07, + "loss": 0.3184, + "step": 44749 + }, + { + "epoch": 8.14, + "learning_rate": 5.591554423006916e-07, + "loss": 0.0107, + "step": 44750 + }, + { + "epoch": 8.14, + "learning_rate": 5.591008372770295e-07, + "loss": 1.2236, + "step": 44751 + }, + { + "epoch": 8.14, + "learning_rate": 5.590462322533673e-07, + "loss": 0.0059, + "step": 44752 + }, + { + "epoch": 8.14, + "learning_rate": 5.589916272297051e-07, + "loss": 1.2761, + "step": 44753 + }, + { + "epoch": 8.14, + "learning_rate": 5.58937022206043e-07, + "loss": 0.0377, + "step": 44754 + }, + { + "epoch": 8.14, + "learning_rate": 5.588824171823808e-07, + "loss": 2.8609, + "step": 44755 + }, + { + "epoch": 8.14, + "learning_rate": 5.588278121587187e-07, + "loss": 1.8449, + "step": 44756 + }, + { + "epoch": 8.14, + "learning_rate": 5.587732071350564e-07, + "loss": 0.0106, + "step": 44757 + }, + { + "epoch": 8.14, + "learning_rate": 5.587186021113942e-07, + "loss": 0.8113, + "step": 44758 + }, + { + "epoch": 8.14, + "learning_rate": 5.586639970877321e-07, + "loss": 0.0541, + "step": 44759 + }, + { + "epoch": 8.14, + "learning_rate": 5.586093920640699e-07, + "loss": 0.1354, + "step": 44760 + }, + { + "epoch": 8.14, + "learning_rate": 5.585547870404077e-07, + "loss": 1.0957, + "step": 44761 + }, + { + "epoch": 8.14, + "learning_rate": 5.585001820167456e-07, + "loss": 0.0495, + "step": 44762 + }, + { + "epoch": 8.14, + "learning_rate": 5.584455769930834e-07, + "loss": 0.28, + "step": 44763 + }, + { + "epoch": 8.14, + "learning_rate": 5.583909719694211e-07, + "loss": 0.7168, + "step": 44764 + }, + { + "epoch": 8.14, + "learning_rate": 5.583363669457591e-07, + "loss": 0.1882, + "step": 44765 + }, + { + "epoch": 8.14, + "learning_rate": 5.582817619220968e-07, + "loss": 3.6312, + "step": 44766 + }, + { + "epoch": 8.14, + "learning_rate": 5.582271568984347e-07, + "loss": 3.1775, + "step": 44767 + }, + { + "epoch": 8.14, + "learning_rate": 5.581725518747725e-07, + "loss": 0.0059, + "step": 44768 + }, + { + "epoch": 8.14, + "learning_rate": 5.581179468511103e-07, + "loss": 0.1549, + "step": 44769 + }, + { + "epoch": 8.14, + "learning_rate": 5.580633418274482e-07, + "loss": 0.0773, + "step": 44770 + }, + { + "epoch": 8.14, + "learning_rate": 5.580087368037859e-07, + "loss": 0.0057, + "step": 44771 + }, + { + "epoch": 8.14, + "learning_rate": 5.579541317801238e-07, + "loss": 0.1016, + "step": 44772 + }, + { + "epoch": 8.14, + "learning_rate": 5.578995267564616e-07, + "loss": 0.0046, + "step": 44773 + }, + { + "epoch": 8.14, + "learning_rate": 5.578449217327994e-07, + "loss": 0.0947, + "step": 44774 + }, + { + "epoch": 8.14, + "learning_rate": 5.577903167091373e-07, + "loss": 0.3664, + "step": 44775 + }, + { + "epoch": 8.14, + "learning_rate": 5.577357116854751e-07, + "loss": 0.1891, + "step": 44776 + }, + { + "epoch": 8.14, + "learning_rate": 5.576811066618129e-07, + "loss": 0.3301, + "step": 44777 + }, + { + "epoch": 8.14, + "learning_rate": 5.576265016381507e-07, + "loss": 0.1255, + "step": 44778 + }, + { + "epoch": 8.14, + "learning_rate": 5.575718966144886e-07, + "loss": 0.0162, + "step": 44779 + }, + { + "epoch": 8.14, + "learning_rate": 5.575172915908263e-07, + "loss": 0.3053, + "step": 44780 + }, + { + "epoch": 8.14, + "learning_rate": 5.574626865671642e-07, + "loss": 0.0892, + "step": 44781 + }, + { + "epoch": 8.14, + "learning_rate": 5.57408081543502e-07, + "loss": 0.1326, + "step": 44782 + }, + { + "epoch": 8.14, + "learning_rate": 5.573534765198399e-07, + "loss": 0.1227, + "step": 44783 + }, + { + "epoch": 8.14, + "learning_rate": 5.572988714961777e-07, + "loss": 0.5078, + "step": 44784 + }, + { + "epoch": 8.14, + "learning_rate": 5.572442664725155e-07, + "loss": 0.0068, + "step": 44785 + }, + { + "epoch": 8.14, + "learning_rate": 5.571896614488534e-07, + "loss": 0.2331, + "step": 44786 + }, + { + "epoch": 8.14, + "learning_rate": 5.571350564251911e-07, + "loss": 2.2191, + "step": 44787 + }, + { + "epoch": 8.14, + "learning_rate": 5.570804514015289e-07, + "loss": 0.3344, + "step": 44788 + }, + { + "epoch": 8.14, + "learning_rate": 5.570258463778668e-07, + "loss": 0.9759, + "step": 44789 + }, + { + "epoch": 8.15, + "learning_rate": 5.569712413542046e-07, + "loss": 0.2855, + "step": 44790 + }, + { + "epoch": 8.15, + "learning_rate": 5.569166363305424e-07, + "loss": 0.1903, + "step": 44791 + }, + { + "epoch": 8.15, + "learning_rate": 5.568620313068803e-07, + "loss": 1.3157, + "step": 44792 + }, + { + "epoch": 8.15, + "learning_rate": 5.568074262832181e-07, + "loss": 0.2051, + "step": 44793 + }, + { + "epoch": 8.15, + "learning_rate": 5.567528212595559e-07, + "loss": 0.4875, + "step": 44794 + }, + { + "epoch": 8.15, + "learning_rate": 5.566982162358937e-07, + "loss": 0.272, + "step": 44795 + }, + { + "epoch": 8.15, + "learning_rate": 5.566436112122315e-07, + "loss": 0.0506, + "step": 44796 + }, + { + "epoch": 8.15, + "learning_rate": 5.565890061885694e-07, + "loss": 1.2171, + "step": 44797 + }, + { + "epoch": 8.15, + "learning_rate": 5.565344011649072e-07, + "loss": 0.4317, + "step": 44798 + }, + { + "epoch": 8.15, + "learning_rate": 5.56479796141245e-07, + "loss": 0.2548, + "step": 44799 + }, + { + "epoch": 8.15, + "learning_rate": 5.564251911175829e-07, + "loss": 0.0209, + "step": 44800 + }, + { + "epoch": 8.15, + "learning_rate": 5.563705860939206e-07, + "loss": 0.0703, + "step": 44801 + }, + { + "epoch": 8.15, + "learning_rate": 5.563159810702584e-07, + "loss": 0.1688, + "step": 44802 + }, + { + "epoch": 8.15, + "learning_rate": 5.562613760465963e-07, + "loss": 1.409, + "step": 44803 + }, + { + "epoch": 8.15, + "learning_rate": 5.562067710229341e-07, + "loss": 0.4717, + "step": 44804 + }, + { + "epoch": 8.15, + "learning_rate": 5.56152165999272e-07, + "loss": 1.8049, + "step": 44805 + }, + { + "epoch": 8.15, + "learning_rate": 5.560975609756098e-07, + "loss": 0.2622, + "step": 44806 + }, + { + "epoch": 8.15, + "learning_rate": 5.560429559519476e-07, + "loss": 0.0498, + "step": 44807 + }, + { + "epoch": 8.15, + "learning_rate": 5.559883509282854e-07, + "loss": 0.1205, + "step": 44808 + }, + { + "epoch": 8.15, + "learning_rate": 5.559337459046233e-07, + "loss": 0.7567, + "step": 44809 + }, + { + "epoch": 8.15, + "learning_rate": 5.55879140880961e-07, + "loss": 1.4536, + "step": 44810 + }, + { + "epoch": 8.15, + "learning_rate": 5.558245358572989e-07, + "loss": 0.504, + "step": 44811 + }, + { + "epoch": 8.15, + "learning_rate": 5.557699308336367e-07, + "loss": 0.0405, + "step": 44812 + }, + { + "epoch": 8.15, + "learning_rate": 5.557153258099746e-07, + "loss": 0.1895, + "step": 44813 + }, + { + "epoch": 8.15, + "learning_rate": 5.556607207863124e-07, + "loss": 0.0068, + "step": 44814 + }, + { + "epoch": 8.15, + "learning_rate": 5.556061157626501e-07, + "loss": 0.0531, + "step": 44815 + }, + { + "epoch": 8.15, + "learning_rate": 5.555515107389881e-07, + "loss": 0.8255, + "step": 44816 + }, + { + "epoch": 8.15, + "learning_rate": 5.554969057153258e-07, + "loss": 0.1323, + "step": 44817 + }, + { + "epoch": 8.15, + "learning_rate": 5.554423006916636e-07, + "loss": 0.0715, + "step": 44818 + }, + { + "epoch": 8.15, + "learning_rate": 5.553876956680015e-07, + "loss": 0.0293, + "step": 44819 + }, + { + "epoch": 8.15, + "learning_rate": 5.553330906443393e-07, + "loss": 5.3451, + "step": 44820 + }, + { + "epoch": 8.15, + "learning_rate": 5.552784856206771e-07, + "loss": 0.4219, + "step": 44821 + }, + { + "epoch": 8.15, + "learning_rate": 5.552238805970149e-07, + "loss": 0.9813, + "step": 44822 + }, + { + "epoch": 8.15, + "learning_rate": 5.551692755733528e-07, + "loss": 0.0322, + "step": 44823 + }, + { + "epoch": 8.15, + "learning_rate": 5.551146705496906e-07, + "loss": 0.0429, + "step": 44824 + }, + { + "epoch": 8.15, + "learning_rate": 5.550600655260284e-07, + "loss": 0.0367, + "step": 44825 + }, + { + "epoch": 8.15, + "learning_rate": 5.550054605023662e-07, + "loss": 0.0429, + "step": 44826 + }, + { + "epoch": 8.15, + "learning_rate": 5.549508554787041e-07, + "loss": 0.6575, + "step": 44827 + }, + { + "epoch": 8.15, + "learning_rate": 5.548962504550419e-07, + "loss": 0.0155, + "step": 44828 + }, + { + "epoch": 8.15, + "learning_rate": 5.548416454313797e-07, + "loss": 1.4659, + "step": 44829 + }, + { + "epoch": 8.15, + "learning_rate": 5.547870404077176e-07, + "loss": 0.0514, + "step": 44830 + }, + { + "epoch": 8.15, + "learning_rate": 5.547324353840553e-07, + "loss": 0.7491, + "step": 44831 + }, + { + "epoch": 8.15, + "learning_rate": 5.546778303603931e-07, + "loss": 0.6802, + "step": 44832 + }, + { + "epoch": 8.15, + "learning_rate": 5.54623225336731e-07, + "loss": 0.1992, + "step": 44833 + }, + { + "epoch": 8.15, + "learning_rate": 5.545686203130688e-07, + "loss": 0.1716, + "step": 44834 + }, + { + "epoch": 8.15, + "learning_rate": 5.545140152894067e-07, + "loss": 0.0421, + "step": 44835 + }, + { + "epoch": 8.15, + "learning_rate": 5.544594102657445e-07, + "loss": 0.056, + "step": 44836 + }, + { + "epoch": 8.15, + "learning_rate": 5.544048052420823e-07, + "loss": 0.4423, + "step": 44837 + }, + { + "epoch": 8.15, + "learning_rate": 5.543502002184201e-07, + "loss": 0.1531, + "step": 44838 + }, + { + "epoch": 8.15, + "learning_rate": 5.542955951947579e-07, + "loss": 0.0099, + "step": 44839 + }, + { + "epoch": 8.15, + "learning_rate": 5.542409901710957e-07, + "loss": 0.2911, + "step": 44840 + }, + { + "epoch": 8.15, + "learning_rate": 5.541863851474336e-07, + "loss": 1.4951, + "step": 44841 + }, + { + "epoch": 8.15, + "learning_rate": 5.541317801237714e-07, + "loss": 1.76, + "step": 44842 + }, + { + "epoch": 8.15, + "learning_rate": 5.540771751001093e-07, + "loss": 0.0408, + "step": 44843 + }, + { + "epoch": 8.15, + "learning_rate": 5.540225700764471e-07, + "loss": 0.0053, + "step": 44844 + }, + { + "epoch": 8.16, + "learning_rate": 5.539679650527848e-07, + "loss": 0.6858, + "step": 44845 + }, + { + "epoch": 8.16, + "learning_rate": 5.539133600291227e-07, + "loss": 2.8875, + "step": 44846 + }, + { + "epoch": 8.16, + "learning_rate": 5.538587550054605e-07, + "loss": 0.0713, + "step": 44847 + }, + { + "epoch": 8.16, + "learning_rate": 5.538041499817983e-07, + "loss": 0.6369, + "step": 44848 + }, + { + "epoch": 8.16, + "learning_rate": 5.537495449581362e-07, + "loss": 1.8892, + "step": 44849 + }, + { + "epoch": 8.16, + "learning_rate": 5.53694939934474e-07, + "loss": 0.2683, + "step": 44850 + }, + { + "epoch": 8.16, + "learning_rate": 5.536403349108118e-07, + "loss": 0.0462, + "step": 44851 + }, + { + "epoch": 8.16, + "learning_rate": 5.535857298871496e-07, + "loss": 0.0249, + "step": 44852 + }, + { + "epoch": 8.16, + "learning_rate": 5.535311248634875e-07, + "loss": 0.4513, + "step": 44853 + }, + { + "epoch": 8.16, + "learning_rate": 5.534765198398253e-07, + "loss": 0.3777, + "step": 44854 + }, + { + "epoch": 8.16, + "learning_rate": 5.534219148161631e-07, + "loss": 1.4687, + "step": 44855 + }, + { + "epoch": 8.16, + "learning_rate": 5.533673097925009e-07, + "loss": 0.2461, + "step": 44856 + }, + { + "epoch": 8.16, + "learning_rate": 5.533127047688388e-07, + "loss": 0.4464, + "step": 44857 + }, + { + "epoch": 8.16, + "learning_rate": 5.532580997451766e-07, + "loss": 0.5661, + "step": 44858 + }, + { + "epoch": 8.16, + "learning_rate": 5.532034947215143e-07, + "loss": 1.9842, + "step": 44859 + }, + { + "epoch": 8.16, + "learning_rate": 5.531488896978523e-07, + "loss": 0.8948, + "step": 44860 + }, + { + "epoch": 8.16, + "learning_rate": 5.5309428467419e-07, + "loss": 0.1092, + "step": 44861 + }, + { + "epoch": 8.16, + "learning_rate": 5.530396796505278e-07, + "loss": 0.1578, + "step": 44862 + }, + { + "epoch": 8.16, + "learning_rate": 5.529850746268657e-07, + "loss": 0.6628, + "step": 44863 + }, + { + "epoch": 8.16, + "learning_rate": 5.529304696032035e-07, + "loss": 3.6929, + "step": 44864 + }, + { + "epoch": 8.16, + "learning_rate": 5.528758645795414e-07, + "loss": 1.1461, + "step": 44865 + }, + { + "epoch": 8.16, + "learning_rate": 5.528212595558791e-07, + "loss": 0.0077, + "step": 44866 + }, + { + "epoch": 8.16, + "learning_rate": 5.52766654532217e-07, + "loss": 0.0073, + "step": 44867 + }, + { + "epoch": 8.16, + "learning_rate": 5.527120495085548e-07, + "loss": 0.0989, + "step": 44868 + }, + { + "epoch": 8.16, + "learning_rate": 5.526574444848926e-07, + "loss": 0.0085, + "step": 44869 + }, + { + "epoch": 8.16, + "learning_rate": 5.526028394612304e-07, + "loss": 0.0023, + "step": 44870 + }, + { + "epoch": 8.16, + "learning_rate": 5.525482344375683e-07, + "loss": 0.0043, + "step": 44871 + }, + { + "epoch": 8.16, + "learning_rate": 5.524936294139061e-07, + "loss": 0.1417, + "step": 44872 + }, + { + "epoch": 8.16, + "learning_rate": 5.52439024390244e-07, + "loss": 0.5446, + "step": 44873 + }, + { + "epoch": 8.16, + "learning_rate": 5.523844193665818e-07, + "loss": 0.8736, + "step": 44874 + }, + { + "epoch": 8.16, + "learning_rate": 5.523298143429195e-07, + "loss": 0.8273, + "step": 44875 + }, + { + "epoch": 8.16, + "learning_rate": 5.522752093192574e-07, + "loss": 0.131, + "step": 44876 + }, + { + "epoch": 8.16, + "learning_rate": 5.522206042955952e-07, + "loss": 1.2266, + "step": 44877 + }, + { + "epoch": 8.16, + "learning_rate": 5.52165999271933e-07, + "loss": 0.292, + "step": 44878 + }, + { + "epoch": 8.16, + "learning_rate": 5.521113942482709e-07, + "loss": 0.202, + "step": 44879 + }, + { + "epoch": 8.16, + "learning_rate": 5.520567892246087e-07, + "loss": 0.1563, + "step": 44880 + }, + { + "epoch": 8.16, + "learning_rate": 5.520021842009466e-07, + "loss": 1.1334, + "step": 44881 + }, + { + "epoch": 8.16, + "learning_rate": 5.519475791772843e-07, + "loss": 0.0259, + "step": 44882 + }, + { + "epoch": 8.16, + "learning_rate": 5.518929741536221e-07, + "loss": 3.6233, + "step": 44883 + }, + { + "epoch": 8.16, + "learning_rate": 5.5183836912996e-07, + "loss": 0.3675, + "step": 44884 + }, + { + "epoch": 8.16, + "learning_rate": 5.517837641062978e-07, + "loss": 1.0662, + "step": 44885 + }, + { + "epoch": 8.16, + "learning_rate": 5.517291590826355e-07, + "loss": 0.4149, + "step": 44886 + }, + { + "epoch": 8.16, + "learning_rate": 5.516745540589735e-07, + "loss": 0.034, + "step": 44887 + }, + { + "epoch": 8.16, + "learning_rate": 5.516199490353113e-07, + "loss": 0.7831, + "step": 44888 + }, + { + "epoch": 8.16, + "learning_rate": 5.51565344011649e-07, + "loss": 0.9321, + "step": 44889 + }, + { + "epoch": 8.16, + "learning_rate": 5.515107389879869e-07, + "loss": 0.6851, + "step": 44890 + }, + { + "epoch": 8.16, + "learning_rate": 5.514561339643247e-07, + "loss": 0.0359, + "step": 44891 + }, + { + "epoch": 8.16, + "learning_rate": 5.514015289406626e-07, + "loss": 0.1319, + "step": 44892 + }, + { + "epoch": 8.16, + "learning_rate": 5.513469239170004e-07, + "loss": 0.7123, + "step": 44893 + }, + { + "epoch": 8.16, + "learning_rate": 5.512923188933382e-07, + "loss": 0.4549, + "step": 44894 + }, + { + "epoch": 8.16, + "learning_rate": 5.512377138696761e-07, + "loss": 0.0049, + "step": 44895 + }, + { + "epoch": 8.16, + "learning_rate": 5.511831088460138e-07, + "loss": 0.0577, + "step": 44896 + }, + { + "epoch": 8.16, + "learning_rate": 5.511285038223517e-07, + "loss": 1.343, + "step": 44897 + }, + { + "epoch": 8.16, + "learning_rate": 5.510738987986895e-07, + "loss": 0.7905, + "step": 44898 + }, + { + "epoch": 8.16, + "learning_rate": 5.510192937750273e-07, + "loss": 1.2929, + "step": 44899 + }, + { + "epoch": 8.17, + "learning_rate": 5.509646887513651e-07, + "loss": 0.1755, + "step": 44900 + }, + { + "epoch": 8.17, + "learning_rate": 5.50910083727703e-07, + "loss": 0.2195, + "step": 44901 + }, + { + "epoch": 8.17, + "learning_rate": 5.508554787040408e-07, + "loss": 0.0055, + "step": 44902 + }, + { + "epoch": 8.17, + "learning_rate": 5.508008736803786e-07, + "loss": 2.4449, + "step": 44903 + }, + { + "epoch": 8.17, + "learning_rate": 5.507462686567165e-07, + "loss": 0.3513, + "step": 44904 + }, + { + "epoch": 8.17, + "learning_rate": 5.506916636330542e-07, + "loss": 0.0071, + "step": 44905 + }, + { + "epoch": 8.17, + "learning_rate": 5.506370586093921e-07, + "loss": 1.5521, + "step": 44906 + }, + { + "epoch": 8.17, + "learning_rate": 5.505824535857299e-07, + "loss": 0.8251, + "step": 44907 + }, + { + "epoch": 8.17, + "learning_rate": 5.505278485620677e-07, + "loss": 0.0269, + "step": 44908 + }, + { + "epoch": 8.17, + "learning_rate": 5.504732435384056e-07, + "loss": 0.4846, + "step": 44909 + }, + { + "epoch": 8.17, + "learning_rate": 5.504186385147433e-07, + "loss": 0.6479, + "step": 44910 + }, + { + "epoch": 8.17, + "learning_rate": 5.503640334910813e-07, + "loss": 0.3934, + "step": 44911 + }, + { + "epoch": 8.17, + "learning_rate": 5.50309428467419e-07, + "loss": 0.0851, + "step": 44912 + }, + { + "epoch": 8.17, + "learning_rate": 5.502548234437568e-07, + "loss": 0.8472, + "step": 44913 + }, + { + "epoch": 8.17, + "learning_rate": 5.502002184200947e-07, + "loss": 0.3037, + "step": 44914 + }, + { + "epoch": 8.17, + "learning_rate": 5.501456133964325e-07, + "loss": 0.1794, + "step": 44915 + }, + { + "epoch": 8.17, + "learning_rate": 5.500910083727703e-07, + "loss": 0.1979, + "step": 44916 + }, + { + "epoch": 8.17, + "learning_rate": 5.500364033491082e-07, + "loss": 0.0063, + "step": 44917 + }, + { + "epoch": 8.17, + "learning_rate": 5.49981798325446e-07, + "loss": 1.5855, + "step": 44918 + }, + { + "epoch": 8.17, + "learning_rate": 5.499271933017837e-07, + "loss": 1.3337, + "step": 44919 + }, + { + "epoch": 8.17, + "learning_rate": 5.498725882781216e-07, + "loss": 1.4838, + "step": 44920 + }, + { + "epoch": 8.17, + "learning_rate": 5.498179832544594e-07, + "loss": 1.398, + "step": 44921 + }, + { + "epoch": 8.17, + "learning_rate": 5.497633782307973e-07, + "loss": 0.0047, + "step": 44922 + }, + { + "epoch": 8.17, + "learning_rate": 5.49708773207135e-07, + "loss": 0.0024, + "step": 44923 + }, + { + "epoch": 8.17, + "learning_rate": 5.496541681834729e-07, + "loss": 0.6814, + "step": 44924 + }, + { + "epoch": 8.17, + "learning_rate": 5.495995631598108e-07, + "loss": 1.1496, + "step": 44925 + }, + { + "epoch": 8.17, + "learning_rate": 5.495449581361485e-07, + "loss": 1.4704, + "step": 44926 + }, + { + "epoch": 8.17, + "learning_rate": 5.494903531124863e-07, + "loss": 3.6879, + "step": 44927 + }, + { + "epoch": 8.17, + "learning_rate": 5.494357480888242e-07, + "loss": 0.0283, + "step": 44928 + }, + { + "epoch": 8.17, + "learning_rate": 5.49381143065162e-07, + "loss": 0.3229, + "step": 44929 + }, + { + "epoch": 8.17, + "learning_rate": 5.493265380414997e-07, + "loss": 0.0056, + "step": 44930 + }, + { + "epoch": 8.17, + "learning_rate": 5.492719330178377e-07, + "loss": 0.941, + "step": 44931 + }, + { + "epoch": 8.17, + "learning_rate": 5.492173279941755e-07, + "loss": 0.1505, + "step": 44932 + }, + { + "epoch": 8.17, + "learning_rate": 5.491627229705133e-07, + "loss": 0.1317, + "step": 44933 + }, + { + "epoch": 8.17, + "learning_rate": 5.491081179468512e-07, + "loss": 1.3258, + "step": 44934 + }, + { + "epoch": 8.17, + "learning_rate": 5.490535129231889e-07, + "loss": 0.3753, + "step": 44935 + }, + { + "epoch": 8.17, + "learning_rate": 5.489989078995268e-07, + "loss": 0.03, + "step": 44936 + }, + { + "epoch": 8.17, + "learning_rate": 5.489443028758646e-07, + "loss": 0.2259, + "step": 44937 + }, + { + "epoch": 8.17, + "learning_rate": 5.488896978522024e-07, + "loss": 0.0152, + "step": 44938 + }, + { + "epoch": 8.17, + "learning_rate": 5.488350928285403e-07, + "loss": 0.045, + "step": 44939 + }, + { + "epoch": 8.17, + "learning_rate": 5.48780487804878e-07, + "loss": 0.1625, + "step": 44940 + }, + { + "epoch": 8.17, + "learning_rate": 5.48725882781216e-07, + "loss": 0.0032, + "step": 44941 + }, + { + "epoch": 8.17, + "learning_rate": 5.486712777575537e-07, + "loss": 0.0067, + "step": 44942 + }, + { + "epoch": 8.17, + "learning_rate": 5.486166727338915e-07, + "loss": 0.941, + "step": 44943 + }, + { + "epoch": 8.17, + "learning_rate": 5.485620677102294e-07, + "loss": 0.8192, + "step": 44944 + }, + { + "epoch": 8.17, + "learning_rate": 5.485074626865672e-07, + "loss": 0.3533, + "step": 44945 + }, + { + "epoch": 8.17, + "learning_rate": 5.48452857662905e-07, + "loss": 1.127, + "step": 44946 + }, + { + "epoch": 8.17, + "learning_rate": 5.483982526392428e-07, + "loss": 0.1249, + "step": 44947 + }, + { + "epoch": 8.17, + "learning_rate": 5.483436476155807e-07, + "loss": 0.172, + "step": 44948 + }, + { + "epoch": 8.17, + "learning_rate": 5.482890425919184e-07, + "loss": 1.3684, + "step": 44949 + }, + { + "epoch": 8.17, + "learning_rate": 5.482344375682563e-07, + "loss": 3.7186, + "step": 44950 + }, + { + "epoch": 8.17, + "learning_rate": 5.481798325445941e-07, + "loss": 1.2804, + "step": 44951 + }, + { + "epoch": 8.17, + "learning_rate": 5.48125227520932e-07, + "loss": 0.3363, + "step": 44952 + }, + { + "epoch": 8.17, + "learning_rate": 5.480706224972698e-07, + "loss": 0.3676, + "step": 44953 + }, + { + "epoch": 8.17, + "learning_rate": 5.480160174736076e-07, + "loss": 0.0714, + "step": 44954 + }, + { + "epoch": 8.18, + "learning_rate": 5.479614124499455e-07, + "loss": 0.3125, + "step": 44955 + }, + { + "epoch": 8.18, + "learning_rate": 5.479068074262832e-07, + "loss": 1.4267, + "step": 44956 + }, + { + "epoch": 8.18, + "learning_rate": 5.47852202402621e-07, + "loss": 0.7297, + "step": 44957 + }, + { + "epoch": 8.18, + "learning_rate": 5.477975973789589e-07, + "loss": 2.3978, + "step": 44958 + }, + { + "epoch": 8.18, + "learning_rate": 5.477429923552967e-07, + "loss": 0.0033, + "step": 44959 + }, + { + "epoch": 8.18, + "learning_rate": 5.476883873316345e-07, + "loss": 0.0719, + "step": 44960 + }, + { + "epoch": 8.18, + "learning_rate": 5.476337823079724e-07, + "loss": 1.0793, + "step": 44961 + }, + { + "epoch": 8.18, + "learning_rate": 5.475791772843102e-07, + "loss": 0.0524, + "step": 44962 + }, + { + "epoch": 8.18, + "learning_rate": 5.47524572260648e-07, + "loss": 0.6223, + "step": 44963 + }, + { + "epoch": 8.18, + "learning_rate": 5.474699672369858e-07, + "loss": 0.2799, + "step": 44964 + }, + { + "epoch": 8.18, + "learning_rate": 5.474153622133236e-07, + "loss": 0.0849, + "step": 44965 + }, + { + "epoch": 8.18, + "learning_rate": 5.473607571896615e-07, + "loss": 0.0056, + "step": 44966 + }, + { + "epoch": 8.18, + "learning_rate": 5.473061521659992e-07, + "loss": 0.0053, + "step": 44967 + }, + { + "epoch": 8.18, + "learning_rate": 5.472515471423371e-07, + "loss": 0.0046, + "step": 44968 + }, + { + "epoch": 8.18, + "learning_rate": 5.47196942118675e-07, + "loss": 2.4713, + "step": 44969 + }, + { + "epoch": 8.18, + "learning_rate": 5.471423370950127e-07, + "loss": 2.2206, + "step": 44970 + }, + { + "epoch": 8.18, + "learning_rate": 5.470877320713506e-07, + "loss": 0.0547, + "step": 44971 + }, + { + "epoch": 8.18, + "learning_rate": 5.470331270476884e-07, + "loss": 0.7163, + "step": 44972 + }, + { + "epoch": 8.18, + "learning_rate": 5.469785220240262e-07, + "loss": 2.9134, + "step": 44973 + }, + { + "epoch": 8.18, + "learning_rate": 5.46923917000364e-07, + "loss": 1.6357, + "step": 44974 + }, + { + "epoch": 8.18, + "learning_rate": 5.468693119767019e-07, + "loss": 0.01, + "step": 44975 + }, + { + "epoch": 8.18, + "learning_rate": 5.468147069530397e-07, + "loss": 0.8728, + "step": 44976 + }, + { + "epoch": 8.18, + "learning_rate": 5.467601019293775e-07, + "loss": 2.1131, + "step": 44977 + }, + { + "epoch": 8.18, + "learning_rate": 5.467054969057154e-07, + "loss": 0.0137, + "step": 44978 + }, + { + "epoch": 8.18, + "learning_rate": 5.466508918820531e-07, + "loss": 0.0087, + "step": 44979 + }, + { + "epoch": 8.18, + "learning_rate": 5.46596286858391e-07, + "loss": 0.1006, + "step": 44980 + }, + { + "epoch": 8.18, + "learning_rate": 5.465416818347288e-07, + "loss": 0.1685, + "step": 44981 + }, + { + "epoch": 8.18, + "learning_rate": 5.464870768110667e-07, + "loss": 1.0522, + "step": 44982 + }, + { + "epoch": 8.18, + "learning_rate": 5.464324717874045e-07, + "loss": 0.0159, + "step": 44983 + }, + { + "epoch": 8.18, + "learning_rate": 5.463778667637422e-07, + "loss": 1.0622, + "step": 44984 + }, + { + "epoch": 8.18, + "learning_rate": 5.463232617400802e-07, + "loss": 0.0141, + "step": 44985 + }, + { + "epoch": 8.18, + "learning_rate": 5.462686567164179e-07, + "loss": 0.0589, + "step": 44986 + }, + { + "epoch": 8.18, + "learning_rate": 5.462140516927557e-07, + "loss": 0.0156, + "step": 44987 + }, + { + "epoch": 8.18, + "learning_rate": 5.461594466690936e-07, + "loss": 0.0037, + "step": 44988 + }, + { + "epoch": 8.18, + "learning_rate": 5.461048416454314e-07, + "loss": 0.9122, + "step": 44989 + }, + { + "epoch": 8.18, + "learning_rate": 5.460502366217693e-07, + "loss": 0.7237, + "step": 44990 + }, + { + "epoch": 8.18, + "learning_rate": 5.45995631598107e-07, + "loss": 1.4848, + "step": 44991 + }, + { + "epoch": 8.18, + "learning_rate": 5.459410265744449e-07, + "loss": 0.0457, + "step": 44992 + }, + { + "epoch": 8.18, + "learning_rate": 5.458864215507827e-07, + "loss": 2.0205, + "step": 44993 + }, + { + "epoch": 8.18, + "learning_rate": 5.458318165271205e-07, + "loss": 1.3746, + "step": 44994 + }, + { + "epoch": 8.18, + "learning_rate": 5.457772115034583e-07, + "loss": 0.2593, + "step": 44995 + }, + { + "epoch": 8.18, + "learning_rate": 5.457226064797962e-07, + "loss": 0.6417, + "step": 44996 + }, + { + "epoch": 8.18, + "learning_rate": 5.45668001456134e-07, + "loss": 0.0045, + "step": 44997 + }, + { + "epoch": 8.18, + "learning_rate": 5.456133964324718e-07, + "loss": 0.1114, + "step": 44998 + }, + { + "epoch": 8.18, + "learning_rate": 5.455587914088097e-07, + "loss": 1.9351, + "step": 44999 + }, + { + "epoch": 8.18, + "learning_rate": 5.455041863851474e-07, + "loss": 0.0619, + "step": 45000 + }, + { + "epoch": 8.18, + "learning_rate": 5.454495813614853e-07, + "loss": 0.8093, + "step": 45001 + }, + { + "epoch": 8.18, + "learning_rate": 5.453949763378231e-07, + "loss": 0.0299, + "step": 45002 + }, + { + "epoch": 8.18, + "learning_rate": 5.453403713141609e-07, + "loss": 0.01, + "step": 45003 + }, + { + "epoch": 8.18, + "learning_rate": 5.452857662904987e-07, + "loss": 0.2505, + "step": 45004 + }, + { + "epoch": 8.18, + "learning_rate": 5.452311612668366e-07, + "loss": 2.8481, + "step": 45005 + }, + { + "epoch": 8.18, + "learning_rate": 5.451765562431744e-07, + "loss": 0.0201, + "step": 45006 + }, + { + "epoch": 8.18, + "learning_rate": 5.451219512195122e-07, + "loss": 1.7112, + "step": 45007 + }, + { + "epoch": 8.18, + "learning_rate": 5.4506734619585e-07, + "loss": 1.8687, + "step": 45008 + }, + { + "epoch": 8.18, + "learning_rate": 5.450127411721878e-07, + "loss": 0.1273, + "step": 45009 + }, + { + "epoch": 8.19, + "learning_rate": 5.449581361485257e-07, + "loss": 0.3564, + "step": 45010 + }, + { + "epoch": 8.19, + "learning_rate": 5.449035311248634e-07, + "loss": 0.1479, + "step": 45011 + }, + { + "epoch": 8.19, + "learning_rate": 5.448489261012014e-07, + "loss": 0.3609, + "step": 45012 + }, + { + "epoch": 8.19, + "learning_rate": 5.447943210775392e-07, + "loss": 0.1319, + "step": 45013 + }, + { + "epoch": 8.19, + "learning_rate": 5.447397160538769e-07, + "loss": 0.5716, + "step": 45014 + }, + { + "epoch": 8.19, + "learning_rate": 5.446851110302148e-07, + "loss": 0.9146, + "step": 45015 + }, + { + "epoch": 8.19, + "learning_rate": 5.446305060065526e-07, + "loss": 0.0095, + "step": 45016 + }, + { + "epoch": 8.19, + "learning_rate": 5.445759009828904e-07, + "loss": 0.0037, + "step": 45017 + }, + { + "epoch": 8.19, + "learning_rate": 5.445212959592283e-07, + "loss": 0.7881, + "step": 45018 + }, + { + "epoch": 8.19, + "learning_rate": 5.444666909355661e-07, + "loss": 0.0035, + "step": 45019 + }, + { + "epoch": 8.19, + "learning_rate": 5.44412085911904e-07, + "loss": 0.1344, + "step": 45020 + }, + { + "epoch": 8.19, + "learning_rate": 5.443574808882417e-07, + "loss": 2.1991, + "step": 45021 + }, + { + "epoch": 8.19, + "learning_rate": 5.443028758645796e-07, + "loss": 0.0328, + "step": 45022 + }, + { + "epoch": 8.19, + "learning_rate": 5.442482708409174e-07, + "loss": 0.0058, + "step": 45023 + }, + { + "epoch": 8.19, + "learning_rate": 5.441936658172552e-07, + "loss": 0.0089, + "step": 45024 + }, + { + "epoch": 8.19, + "learning_rate": 5.44139060793593e-07, + "loss": 0.0095, + "step": 45025 + }, + { + "epoch": 8.19, + "learning_rate": 5.440844557699309e-07, + "loss": 0.3829, + "step": 45026 + }, + { + "epoch": 8.19, + "learning_rate": 5.440298507462687e-07, + "loss": 1.1834, + "step": 45027 + }, + { + "epoch": 8.19, + "learning_rate": 5.439752457226064e-07, + "loss": 0.7794, + "step": 45028 + }, + { + "epoch": 8.19, + "learning_rate": 5.439206406989444e-07, + "loss": 0.9287, + "step": 45029 + }, + { + "epoch": 8.19, + "learning_rate": 5.438660356752821e-07, + "loss": 1.5563, + "step": 45030 + }, + { + "epoch": 8.19, + "learning_rate": 5.4381143065162e-07, + "loss": 0.4062, + "step": 45031 + }, + { + "epoch": 8.19, + "learning_rate": 5.437568256279578e-07, + "loss": 0.0567, + "step": 45032 + }, + { + "epoch": 8.19, + "learning_rate": 5.437022206042956e-07, + "loss": 3.0685, + "step": 45033 + }, + { + "epoch": 8.19, + "learning_rate": 5.436476155806335e-07, + "loss": 2.3306, + "step": 45034 + }, + { + "epoch": 8.19, + "learning_rate": 5.435930105569712e-07, + "loss": 0.007, + "step": 45035 + }, + { + "epoch": 8.19, + "learning_rate": 5.435384055333091e-07, + "loss": 0.1263, + "step": 45036 + }, + { + "epoch": 8.19, + "learning_rate": 5.434838005096469e-07, + "loss": 0.4891, + "step": 45037 + }, + { + "epoch": 8.19, + "learning_rate": 5.434291954859847e-07, + "loss": 0.8656, + "step": 45038 + }, + { + "epoch": 8.19, + "learning_rate": 5.433745904623225e-07, + "loss": 0.2066, + "step": 45039 + }, + { + "epoch": 8.19, + "learning_rate": 5.433199854386604e-07, + "loss": 0.0046, + "step": 45040 + }, + { + "epoch": 8.19, + "learning_rate": 5.432653804149981e-07, + "loss": 0.041, + "step": 45041 + }, + { + "epoch": 8.19, + "learning_rate": 5.432107753913361e-07, + "loss": 0.6743, + "step": 45042 + }, + { + "epoch": 8.19, + "learning_rate": 5.431561703676739e-07, + "loss": 1.6781, + "step": 45043 + }, + { + "epoch": 8.19, + "learning_rate": 5.431015653440116e-07, + "loss": 0.0056, + "step": 45044 + }, + { + "epoch": 8.19, + "learning_rate": 5.430469603203495e-07, + "loss": 0.0194, + "step": 45045 + }, + { + "epoch": 8.19, + "learning_rate": 5.429923552966873e-07, + "loss": 0.3979, + "step": 45046 + }, + { + "epoch": 8.19, + "learning_rate": 5.429377502730251e-07, + "loss": 0.0042, + "step": 45047 + }, + { + "epoch": 8.19, + "learning_rate": 5.428831452493629e-07, + "loss": 0.915, + "step": 45048 + }, + { + "epoch": 8.19, + "learning_rate": 5.428285402257008e-07, + "loss": 1.4564, + "step": 45049 + }, + { + "epoch": 8.19, + "learning_rate": 5.427739352020387e-07, + "loss": 0.5046, + "step": 45050 + }, + { + "epoch": 8.19, + "learning_rate": 5.427193301783764e-07, + "loss": 0.0046, + "step": 45051 + }, + { + "epoch": 8.19, + "learning_rate": 5.426647251547142e-07, + "loss": 0.015, + "step": 45052 + }, + { + "epoch": 8.19, + "learning_rate": 5.426101201310521e-07, + "loss": 1.5716, + "step": 45053 + }, + { + "epoch": 8.19, + "learning_rate": 5.425555151073899e-07, + "loss": 0.8188, + "step": 45054 + }, + { + "epoch": 8.19, + "learning_rate": 5.425009100837276e-07, + "loss": 1.2808, + "step": 45055 + }, + { + "epoch": 8.19, + "learning_rate": 5.424463050600656e-07, + "loss": 0.0068, + "step": 45056 + }, + { + "epoch": 8.19, + "learning_rate": 5.423917000364034e-07, + "loss": 0.0054, + "step": 45057 + }, + { + "epoch": 8.19, + "learning_rate": 5.423370950127411e-07, + "loss": 0.0628, + "step": 45058 + }, + { + "epoch": 8.19, + "learning_rate": 5.42282489989079e-07, + "loss": 0.0027, + "step": 45059 + }, + { + "epoch": 8.19, + "learning_rate": 5.422278849654168e-07, + "loss": 0.3254, + "step": 45060 + }, + { + "epoch": 8.19, + "learning_rate": 5.421732799417547e-07, + "loss": 0.8784, + "step": 45061 + }, + { + "epoch": 8.19, + "learning_rate": 5.421186749180925e-07, + "loss": 0.1506, + "step": 45062 + }, + { + "epoch": 8.19, + "learning_rate": 5.420640698944303e-07, + "loss": 0.0074, + "step": 45063 + }, + { + "epoch": 8.19, + "learning_rate": 5.420094648707682e-07, + "loss": 0.1599, + "step": 45064 + }, + { + "epoch": 8.2, + "learning_rate": 5.419548598471059e-07, + "loss": 0.0599, + "step": 45065 + }, + { + "epoch": 8.2, + "learning_rate": 5.419002548234438e-07, + "loss": 0.0058, + "step": 45066 + }, + { + "epoch": 8.2, + "learning_rate": 5.418456497997816e-07, + "loss": 0.0397, + "step": 45067 + }, + { + "epoch": 8.2, + "learning_rate": 5.417910447761194e-07, + "loss": 0.6539, + "step": 45068 + }, + { + "epoch": 8.2, + "learning_rate": 5.417364397524573e-07, + "loss": 0.1393, + "step": 45069 + }, + { + "epoch": 8.2, + "learning_rate": 5.416818347287951e-07, + "loss": 1.0862, + "step": 45070 + }, + { + "epoch": 8.2, + "learning_rate": 5.416272297051329e-07, + "loss": 1.6876, + "step": 45071 + }, + { + "epoch": 8.2, + "learning_rate": 5.415726246814707e-07, + "loss": 1.0993, + "step": 45072 + }, + { + "epoch": 8.2, + "learning_rate": 5.415180196578086e-07, + "loss": 0.9226, + "step": 45073 + }, + { + "epoch": 8.2, + "learning_rate": 5.414634146341463e-07, + "loss": 0.0428, + "step": 45074 + }, + { + "epoch": 8.2, + "learning_rate": 5.414088096104842e-07, + "loss": 0.239, + "step": 45075 + }, + { + "epoch": 8.2, + "learning_rate": 5.41354204586822e-07, + "loss": 0.4976, + "step": 45076 + }, + { + "epoch": 8.2, + "learning_rate": 5.412995995631598e-07, + "loss": 0.2762, + "step": 45077 + }, + { + "epoch": 8.2, + "learning_rate": 5.412449945394976e-07, + "loss": 0.6285, + "step": 45078 + }, + { + "epoch": 8.2, + "learning_rate": 5.411903895158354e-07, + "loss": 0.1816, + "step": 45079 + }, + { + "epoch": 8.2, + "learning_rate": 5.411357844921734e-07, + "loss": 0.2475, + "step": 45080 + }, + { + "epoch": 8.2, + "learning_rate": 5.410811794685111e-07, + "loss": 3.1107, + "step": 45081 + }, + { + "epoch": 8.2, + "learning_rate": 5.410265744448489e-07, + "loss": 0.0282, + "step": 45082 + }, + { + "epoch": 8.2, + "learning_rate": 5.409719694211868e-07, + "loss": 0.2048, + "step": 45083 + }, + { + "epoch": 8.2, + "learning_rate": 5.409173643975246e-07, + "loss": 1.1209, + "step": 45084 + }, + { + "epoch": 8.2, + "learning_rate": 5.408627593738623e-07, + "loss": 0.0059, + "step": 45085 + }, + { + "epoch": 8.2, + "learning_rate": 5.408081543502003e-07, + "loss": 0.7318, + "step": 45086 + }, + { + "epoch": 8.2, + "learning_rate": 5.407535493265381e-07, + "loss": 1.3315, + "step": 45087 + }, + { + "epoch": 8.2, + "learning_rate": 5.406989443028758e-07, + "loss": 0.1059, + "step": 45088 + }, + { + "epoch": 8.2, + "learning_rate": 5.406443392792137e-07, + "loss": 0.2052, + "step": 45089 + }, + { + "epoch": 8.2, + "learning_rate": 5.405897342555515e-07, + "loss": 0.2137, + "step": 45090 + }, + { + "epoch": 8.2, + "learning_rate": 5.405351292318894e-07, + "loss": 0.0114, + "step": 45091 + }, + { + "epoch": 8.2, + "learning_rate": 5.404805242082271e-07, + "loss": 0.0138, + "step": 45092 + }, + { + "epoch": 8.2, + "learning_rate": 5.40425919184565e-07, + "loss": 0.1821, + "step": 45093 + }, + { + "epoch": 8.2, + "learning_rate": 5.403713141609029e-07, + "loss": 0.459, + "step": 45094 + }, + { + "epoch": 8.2, + "learning_rate": 5.403167091372406e-07, + "loss": 0.0051, + "step": 45095 + }, + { + "epoch": 8.2, + "learning_rate": 5.402621041135784e-07, + "loss": 1.177, + "step": 45096 + }, + { + "epoch": 8.2, + "learning_rate": 5.402074990899163e-07, + "loss": 0.1422, + "step": 45097 + }, + { + "epoch": 8.2, + "learning_rate": 5.401528940662541e-07, + "loss": 0.3108, + "step": 45098 + }, + { + "epoch": 8.2, + "learning_rate": 5.400982890425918e-07, + "loss": 0.0373, + "step": 45099 + }, + { + "epoch": 8.2, + "learning_rate": 5.400436840189298e-07, + "loss": 1.833, + "step": 45100 + }, + { + "epoch": 8.2, + "learning_rate": 5.399890789952676e-07, + "loss": 0.0139, + "step": 45101 + }, + { + "epoch": 8.2, + "learning_rate": 5.399344739716054e-07, + "loss": 0.1095, + "step": 45102 + }, + { + "epoch": 8.2, + "learning_rate": 5.398798689479433e-07, + "loss": 0.7737, + "step": 45103 + }, + { + "epoch": 8.2, + "learning_rate": 5.39825263924281e-07, + "loss": 0.0551, + "step": 45104 + }, + { + "epoch": 8.2, + "learning_rate": 5.397706589006189e-07, + "loss": 0.4244, + "step": 45105 + }, + { + "epoch": 8.2, + "learning_rate": 5.397160538769567e-07, + "loss": 0.1707, + "step": 45106 + }, + { + "epoch": 8.2, + "learning_rate": 5.396614488532945e-07, + "loss": 0.0084, + "step": 45107 + }, + { + "epoch": 8.2, + "learning_rate": 5.396068438296324e-07, + "loss": 0.4276, + "step": 45108 + }, + { + "epoch": 8.2, + "learning_rate": 5.395522388059701e-07, + "loss": 0.1638, + "step": 45109 + }, + { + "epoch": 8.2, + "learning_rate": 5.394976337823081e-07, + "loss": 0.3905, + "step": 45110 + }, + { + "epoch": 8.2, + "learning_rate": 5.394430287586458e-07, + "loss": 0.0995, + "step": 45111 + }, + { + "epoch": 8.2, + "learning_rate": 5.393884237349836e-07, + "loss": 0.1329, + "step": 45112 + }, + { + "epoch": 8.2, + "learning_rate": 5.393338187113215e-07, + "loss": 0.7008, + "step": 45113 + }, + { + "epoch": 8.2, + "learning_rate": 5.392792136876593e-07, + "loss": 0.0964, + "step": 45114 + }, + { + "epoch": 8.2, + "learning_rate": 5.392246086639971e-07, + "loss": 0.6371, + "step": 45115 + }, + { + "epoch": 8.2, + "learning_rate": 5.391700036403349e-07, + "loss": 0.0054, + "step": 45116 + }, + { + "epoch": 8.2, + "learning_rate": 5.391153986166728e-07, + "loss": 0.2747, + "step": 45117 + }, + { + "epoch": 8.2, + "learning_rate": 5.390607935930105e-07, + "loss": 0.061, + "step": 45118 + }, + { + "epoch": 8.2, + "learning_rate": 5.390061885693484e-07, + "loss": 0.5629, + "step": 45119 + }, + { + "epoch": 8.21, + "learning_rate": 5.389515835456862e-07, + "loss": 0.0176, + "step": 45120 + }, + { + "epoch": 8.21, + "learning_rate": 5.388969785220241e-07, + "loss": 0.1066, + "step": 45121 + }, + { + "epoch": 8.21, + "learning_rate": 5.388423734983618e-07, + "loss": 0.1987, + "step": 45122 + }, + { + "epoch": 8.21, + "learning_rate": 5.387877684746996e-07, + "loss": 0.092, + "step": 45123 + }, + { + "epoch": 8.21, + "learning_rate": 5.387331634510376e-07, + "loss": 0.0058, + "step": 45124 + }, + { + "epoch": 8.21, + "learning_rate": 5.386785584273753e-07, + "loss": 0.4999, + "step": 45125 + }, + { + "epoch": 8.21, + "learning_rate": 5.386239534037131e-07, + "loss": 0.0974, + "step": 45126 + }, + { + "epoch": 8.21, + "learning_rate": 5.38569348380051e-07, + "loss": 0.0168, + "step": 45127 + }, + { + "epoch": 8.21, + "learning_rate": 5.385147433563888e-07, + "loss": 0.065, + "step": 45128 + }, + { + "epoch": 8.21, + "learning_rate": 5.384601383327265e-07, + "loss": 0.0039, + "step": 45129 + }, + { + "epoch": 8.21, + "learning_rate": 5.384055333090645e-07, + "loss": 0.0081, + "step": 45130 + }, + { + "epoch": 8.21, + "learning_rate": 5.383509282854023e-07, + "loss": 0.282, + "step": 45131 + }, + { + "epoch": 8.21, + "learning_rate": 5.382963232617401e-07, + "loss": 0.0369, + "step": 45132 + }, + { + "epoch": 8.21, + "learning_rate": 5.382417182380779e-07, + "loss": 0.0781, + "step": 45133 + }, + { + "epoch": 8.21, + "learning_rate": 5.381871132144157e-07, + "loss": 0.2478, + "step": 45134 + }, + { + "epoch": 8.21, + "learning_rate": 5.381325081907536e-07, + "loss": 0.1923, + "step": 45135 + }, + { + "epoch": 8.21, + "learning_rate": 5.380779031670913e-07, + "loss": 0.8153, + "step": 45136 + }, + { + "epoch": 8.21, + "learning_rate": 5.380232981434292e-07, + "loss": 2.9085, + "step": 45137 + }, + { + "epoch": 8.21, + "learning_rate": 5.379686931197671e-07, + "loss": 0.2922, + "step": 45138 + }, + { + "epoch": 8.21, + "learning_rate": 5.379140880961048e-07, + "loss": 1.9945, + "step": 45139 + }, + { + "epoch": 8.21, + "learning_rate": 5.378594830724427e-07, + "loss": 0.1953, + "step": 45140 + }, + { + "epoch": 8.21, + "learning_rate": 5.378048780487805e-07, + "loss": 0.9424, + "step": 45141 + }, + { + "epoch": 8.21, + "learning_rate": 5.377502730251183e-07, + "loss": 0.1189, + "step": 45142 + }, + { + "epoch": 8.21, + "learning_rate": 5.376956680014561e-07, + "loss": 0.5753, + "step": 45143 + }, + { + "epoch": 8.21, + "learning_rate": 5.37641062977794e-07, + "loss": 0.7517, + "step": 45144 + }, + { + "epoch": 8.21, + "learning_rate": 5.375864579541318e-07, + "loss": 0.005, + "step": 45145 + }, + { + "epoch": 8.21, + "learning_rate": 5.375318529304696e-07, + "loss": 2.2325, + "step": 45146 + }, + { + "epoch": 8.21, + "learning_rate": 5.374772479068075e-07, + "loss": 0.1671, + "step": 45147 + }, + { + "epoch": 8.21, + "learning_rate": 5.374226428831452e-07, + "loss": 1.3727, + "step": 45148 + }, + { + "epoch": 8.21, + "learning_rate": 5.373680378594831e-07, + "loss": 2.2733, + "step": 45149 + }, + { + "epoch": 8.21, + "learning_rate": 5.373134328358209e-07, + "loss": 1.3368, + "step": 45150 + }, + { + "epoch": 8.21, + "learning_rate": 5.372588278121588e-07, + "loss": 0.0097, + "step": 45151 + }, + { + "epoch": 8.21, + "learning_rate": 5.372042227884966e-07, + "loss": 0.254, + "step": 45152 + }, + { + "epoch": 8.21, + "learning_rate": 5.371496177648343e-07, + "loss": 0.0076, + "step": 45153 + }, + { + "epoch": 8.21, + "learning_rate": 5.370950127411723e-07, + "loss": 0.6812, + "step": 45154 + }, + { + "epoch": 8.21, + "learning_rate": 5.3704040771751e-07, + "loss": 0.1581, + "step": 45155 + }, + { + "epoch": 8.21, + "learning_rate": 5.369858026938478e-07, + "loss": 2.3891, + "step": 45156 + }, + { + "epoch": 8.21, + "learning_rate": 5.369311976701857e-07, + "loss": 0.9972, + "step": 45157 + }, + { + "epoch": 8.21, + "learning_rate": 5.368765926465235e-07, + "loss": 0.2237, + "step": 45158 + }, + { + "epoch": 8.21, + "learning_rate": 5.368219876228613e-07, + "loss": 0.0059, + "step": 45159 + }, + { + "epoch": 8.21, + "learning_rate": 5.367673825991991e-07, + "loss": 0.8397, + "step": 45160 + }, + { + "epoch": 8.21, + "learning_rate": 5.36712777575537e-07, + "loss": 0.2603, + "step": 45161 + }, + { + "epoch": 8.21, + "learning_rate": 5.366581725518748e-07, + "loss": 0.1179, + "step": 45162 + }, + { + "epoch": 8.21, + "learning_rate": 5.366035675282126e-07, + "loss": 0.2313, + "step": 45163 + }, + { + "epoch": 8.21, + "learning_rate": 5.365489625045504e-07, + "loss": 3.1538, + "step": 45164 + }, + { + "epoch": 8.21, + "learning_rate": 5.364943574808883e-07, + "loss": 0.4073, + "step": 45165 + }, + { + "epoch": 8.21, + "learning_rate": 5.36439752457226e-07, + "loss": 1.6757, + "step": 45166 + }, + { + "epoch": 8.21, + "learning_rate": 5.36385147433564e-07, + "loss": 0.4063, + "step": 45167 + }, + { + "epoch": 8.21, + "learning_rate": 5.363305424099018e-07, + "loss": 0.0068, + "step": 45168 + }, + { + "epoch": 8.21, + "learning_rate": 5.362759373862395e-07, + "loss": 0.014, + "step": 45169 + }, + { + "epoch": 8.21, + "learning_rate": 5.362213323625774e-07, + "loss": 0.2378, + "step": 45170 + }, + { + "epoch": 8.21, + "learning_rate": 5.361667273389152e-07, + "loss": 0.0123, + "step": 45171 + }, + { + "epoch": 8.21, + "learning_rate": 5.36112122315253e-07, + "loss": 0.1224, + "step": 45172 + }, + { + "epoch": 8.21, + "learning_rate": 5.360575172915908e-07, + "loss": 0.131, + "step": 45173 + }, + { + "epoch": 8.21, + "learning_rate": 5.360029122679287e-07, + "loss": 1.0907, + "step": 45174 + }, + { + "epoch": 8.22, + "learning_rate": 5.359483072442665e-07, + "loss": 0.7497, + "step": 45175 + }, + { + "epoch": 8.22, + "learning_rate": 5.358937022206043e-07, + "loss": 0.0282, + "step": 45176 + }, + { + "epoch": 8.22, + "learning_rate": 5.358390971969421e-07, + "loss": 0.3054, + "step": 45177 + }, + { + "epoch": 8.22, + "learning_rate": 5.3578449217328e-07, + "loss": 0.0107, + "step": 45178 + }, + { + "epoch": 8.22, + "learning_rate": 5.357298871496178e-07, + "loss": 0.0186, + "step": 45179 + }, + { + "epoch": 8.22, + "learning_rate": 5.356752821259555e-07, + "loss": 0.4007, + "step": 45180 + }, + { + "epoch": 8.22, + "learning_rate": 5.356206771022935e-07, + "loss": 0.0083, + "step": 45181 + }, + { + "epoch": 8.22, + "learning_rate": 5.355660720786313e-07, + "loss": 0.5945, + "step": 45182 + }, + { + "epoch": 8.22, + "learning_rate": 5.35511467054969e-07, + "loss": 0.3714, + "step": 45183 + }, + { + "epoch": 8.22, + "learning_rate": 5.354568620313069e-07, + "loss": 1.3829, + "step": 45184 + }, + { + "epoch": 8.22, + "learning_rate": 5.354022570076447e-07, + "loss": 0.0428, + "step": 45185 + }, + { + "epoch": 8.22, + "learning_rate": 5.353476519839825e-07, + "loss": 0.3969, + "step": 45186 + }, + { + "epoch": 8.22, + "learning_rate": 5.352930469603204e-07, + "loss": 0.0828, + "step": 45187 + }, + { + "epoch": 8.22, + "learning_rate": 5.352384419366582e-07, + "loss": 1.546, + "step": 45188 + }, + { + "epoch": 8.22, + "learning_rate": 5.351838369129961e-07, + "loss": 0.082, + "step": 45189 + }, + { + "epoch": 8.22, + "learning_rate": 5.351292318893338e-07, + "loss": 0.0051, + "step": 45190 + }, + { + "epoch": 8.22, + "learning_rate": 5.350746268656717e-07, + "loss": 1.7854, + "step": 45191 + }, + { + "epoch": 8.22, + "learning_rate": 5.350200218420095e-07, + "loss": 0.4637, + "step": 45192 + }, + { + "epoch": 8.22, + "learning_rate": 5.349654168183473e-07, + "loss": 1.1571, + "step": 45193 + }, + { + "epoch": 8.22, + "learning_rate": 5.349108117946851e-07, + "loss": 0.0971, + "step": 45194 + }, + { + "epoch": 8.22, + "learning_rate": 5.34856206771023e-07, + "loss": 0.0085, + "step": 45195 + }, + { + "epoch": 8.22, + "learning_rate": 5.348016017473607e-07, + "loss": 1.4594, + "step": 45196 + }, + { + "epoch": 8.22, + "learning_rate": 5.347469967236985e-07, + "loss": 0.0512, + "step": 45197 + }, + { + "epoch": 8.22, + "learning_rate": 5.346923917000365e-07, + "loss": 0.3585, + "step": 45198 + }, + { + "epoch": 8.22, + "learning_rate": 5.346377866763742e-07, + "loss": 0.4172, + "step": 45199 + }, + { + "epoch": 8.22, + "learning_rate": 5.345831816527121e-07, + "loss": 0.1158, + "step": 45200 + }, + { + "epoch": 8.22, + "learning_rate": 5.345285766290499e-07, + "loss": 1.598, + "step": 45201 + }, + { + "epoch": 8.22, + "learning_rate": 5.344739716053877e-07, + "loss": 0.825, + "step": 45202 + }, + { + "epoch": 8.22, + "learning_rate": 5.344193665817255e-07, + "loss": 0.316, + "step": 45203 + }, + { + "epoch": 8.22, + "learning_rate": 5.343647615580633e-07, + "loss": 0.0425, + "step": 45204 + }, + { + "epoch": 8.22, + "learning_rate": 5.343101565344012e-07, + "loss": 0.019, + "step": 45205 + }, + { + "epoch": 8.22, + "learning_rate": 5.34255551510739e-07, + "loss": 0.562, + "step": 45206 + }, + { + "epoch": 8.22, + "learning_rate": 5.342009464870768e-07, + "loss": 0.7862, + "step": 45207 + }, + { + "epoch": 8.22, + "learning_rate": 5.341463414634147e-07, + "loss": 0.8938, + "step": 45208 + }, + { + "epoch": 8.22, + "learning_rate": 5.340917364397525e-07, + "loss": 0.154, + "step": 45209 + }, + { + "epoch": 8.22, + "learning_rate": 5.340371314160902e-07, + "loss": 0.538, + "step": 45210 + }, + { + "epoch": 8.22, + "learning_rate": 5.339825263924282e-07, + "loss": 0.0723, + "step": 45211 + }, + { + "epoch": 8.22, + "learning_rate": 5.33927921368766e-07, + "loss": 2.2426, + "step": 45212 + }, + { + "epoch": 8.22, + "learning_rate": 5.338733163451037e-07, + "loss": 0.0078, + "step": 45213 + }, + { + "epoch": 8.22, + "learning_rate": 5.338187113214416e-07, + "loss": 0.0037, + "step": 45214 + }, + { + "epoch": 8.22, + "learning_rate": 5.337641062977794e-07, + "loss": 0.2735, + "step": 45215 + }, + { + "epoch": 8.22, + "learning_rate": 5.337095012741172e-07, + "loss": 1.1218, + "step": 45216 + }, + { + "epoch": 8.22, + "learning_rate": 5.33654896250455e-07, + "loss": 0.0659, + "step": 45217 + }, + { + "epoch": 8.22, + "learning_rate": 5.336002912267929e-07, + "loss": 0.0097, + "step": 45218 + }, + { + "epoch": 8.22, + "learning_rate": 5.335456862031308e-07, + "loss": 0.2111, + "step": 45219 + }, + { + "epoch": 8.22, + "learning_rate": 5.334910811794685e-07, + "loss": 0.0808, + "step": 45220 + }, + { + "epoch": 8.22, + "learning_rate": 5.334364761558063e-07, + "loss": 0.2979, + "step": 45221 + }, + { + "epoch": 8.22, + "learning_rate": 5.333818711321442e-07, + "loss": 1.1078, + "step": 45222 + }, + { + "epoch": 8.22, + "learning_rate": 5.33327266108482e-07, + "loss": 1.8682, + "step": 45223 + }, + { + "epoch": 8.22, + "learning_rate": 5.332726610848197e-07, + "loss": 0.4193, + "step": 45224 + }, + { + "epoch": 8.22, + "learning_rate": 5.332180560611577e-07, + "loss": 0.0053, + "step": 45225 + }, + { + "epoch": 8.22, + "learning_rate": 5.331634510374955e-07, + "loss": 0.1725, + "step": 45226 + }, + { + "epoch": 8.22, + "learning_rate": 5.331088460138332e-07, + "loss": 0.6359, + "step": 45227 + }, + { + "epoch": 8.22, + "learning_rate": 5.330542409901711e-07, + "loss": 0.0098, + "step": 45228 + }, + { + "epoch": 8.22, + "learning_rate": 5.329996359665089e-07, + "loss": 0.4602, + "step": 45229 + }, + { + "epoch": 8.23, + "learning_rate": 5.329450309428468e-07, + "loss": 1.0657, + "step": 45230 + }, + { + "epoch": 8.23, + "learning_rate": 5.328904259191846e-07, + "loss": 0.0388, + "step": 45231 + }, + { + "epoch": 8.23, + "learning_rate": 5.328358208955224e-07, + "loss": 0.2741, + "step": 45232 + }, + { + "epoch": 8.23, + "learning_rate": 5.327812158718602e-07, + "loss": 0.1736, + "step": 45233 + }, + { + "epoch": 8.23, + "learning_rate": 5.32726610848198e-07, + "loss": 0.0663, + "step": 45234 + }, + { + "epoch": 8.23, + "learning_rate": 5.32672005824536e-07, + "loss": 0.0438, + "step": 45235 + }, + { + "epoch": 8.23, + "learning_rate": 5.326174008008737e-07, + "loss": 0.1891, + "step": 45236 + }, + { + "epoch": 8.23, + "learning_rate": 5.325627957772115e-07, + "loss": 0.0746, + "step": 45237 + }, + { + "epoch": 8.23, + "learning_rate": 5.325081907535494e-07, + "loss": 0.3009, + "step": 45238 + }, + { + "epoch": 8.23, + "learning_rate": 5.324535857298872e-07, + "loss": 0.0062, + "step": 45239 + }, + { + "epoch": 8.23, + "learning_rate": 5.323989807062249e-07, + "loss": 2.3946, + "step": 45240 + }, + { + "epoch": 8.23, + "learning_rate": 5.323443756825628e-07, + "loss": 0.0163, + "step": 45241 + }, + { + "epoch": 8.23, + "learning_rate": 5.322897706589007e-07, + "loss": 0.056, + "step": 45242 + }, + { + "epoch": 8.23, + "learning_rate": 5.322351656352384e-07, + "loss": 0.1042, + "step": 45243 + }, + { + "epoch": 8.23, + "learning_rate": 5.321805606115763e-07, + "loss": 0.1636, + "step": 45244 + }, + { + "epoch": 8.23, + "learning_rate": 5.321259555879141e-07, + "loss": 0.0253, + "step": 45245 + }, + { + "epoch": 8.23, + "learning_rate": 5.32071350564252e-07, + "loss": 0.0065, + "step": 45246 + }, + { + "epoch": 8.23, + "learning_rate": 5.320167455405897e-07, + "loss": 0.0658, + "step": 45247 + }, + { + "epoch": 8.23, + "learning_rate": 5.319621405169275e-07, + "loss": 0.0098, + "step": 45248 + }, + { + "epoch": 8.23, + "learning_rate": 5.319075354932655e-07, + "loss": 0.7543, + "step": 45249 + }, + { + "epoch": 8.23, + "learning_rate": 5.318529304696032e-07, + "loss": 0.0104, + "step": 45250 + }, + { + "epoch": 8.23, + "learning_rate": 5.31798325445941e-07, + "loss": 0.9474, + "step": 45251 + }, + { + "epoch": 8.23, + "learning_rate": 5.317437204222789e-07, + "loss": 0.1388, + "step": 45252 + }, + { + "epoch": 8.23, + "learning_rate": 5.316891153986167e-07, + "loss": 0.4781, + "step": 45253 + }, + { + "epoch": 8.23, + "learning_rate": 5.316345103749544e-07, + "loss": 1.1799, + "step": 45254 + }, + { + "epoch": 8.23, + "learning_rate": 5.315799053512924e-07, + "loss": 0.0349, + "step": 45255 + }, + { + "epoch": 8.23, + "learning_rate": 5.315253003276302e-07, + "loss": 0.0137, + "step": 45256 + }, + { + "epoch": 8.23, + "learning_rate": 5.31470695303968e-07, + "loss": 0.7558, + "step": 45257 + }, + { + "epoch": 8.23, + "learning_rate": 5.314160902803058e-07, + "loss": 0.0548, + "step": 45258 + }, + { + "epoch": 8.23, + "learning_rate": 5.313614852566436e-07, + "loss": 0.005, + "step": 45259 + }, + { + "epoch": 8.23, + "learning_rate": 5.313068802329815e-07, + "loss": 0.009, + "step": 45260 + }, + { + "epoch": 8.23, + "learning_rate": 5.312522752093192e-07, + "loss": 1.4939, + "step": 45261 + }, + { + "epoch": 8.23, + "learning_rate": 5.311976701856571e-07, + "loss": 0.5061, + "step": 45262 + }, + { + "epoch": 8.23, + "learning_rate": 5.31143065161995e-07, + "loss": 0.0327, + "step": 45263 + }, + { + "epoch": 8.23, + "learning_rate": 5.310884601383327e-07, + "loss": 0.0045, + "step": 45264 + }, + { + "epoch": 8.23, + "learning_rate": 5.310338551146705e-07, + "loss": 0.0531, + "step": 45265 + }, + { + "epoch": 8.23, + "learning_rate": 5.309792500910084e-07, + "loss": 0.5226, + "step": 45266 + }, + { + "epoch": 8.23, + "learning_rate": 5.309246450673462e-07, + "loss": 1.3309, + "step": 45267 + }, + { + "epoch": 8.23, + "learning_rate": 5.30870040043684e-07, + "loss": 0.7271, + "step": 45268 + }, + { + "epoch": 8.23, + "learning_rate": 5.308154350200219e-07, + "loss": 1.4649, + "step": 45269 + }, + { + "epoch": 8.23, + "learning_rate": 5.307608299963597e-07, + "loss": 0.2452, + "step": 45270 + }, + { + "epoch": 8.23, + "learning_rate": 5.307062249726975e-07, + "loss": 1.1628, + "step": 45271 + }, + { + "epoch": 8.23, + "learning_rate": 5.306516199490353e-07, + "loss": 0.3048, + "step": 45272 + }, + { + "epoch": 8.23, + "learning_rate": 5.305970149253731e-07, + "loss": 0.0395, + "step": 45273 + }, + { + "epoch": 8.23, + "learning_rate": 5.30542409901711e-07, + "loss": 0.1044, + "step": 45274 + }, + { + "epoch": 8.23, + "learning_rate": 5.304878048780488e-07, + "loss": 0.0039, + "step": 45275 + }, + { + "epoch": 8.23, + "learning_rate": 5.304331998543867e-07, + "loss": 0.0068, + "step": 45276 + }, + { + "epoch": 8.23, + "learning_rate": 5.303785948307244e-07, + "loss": 0.4434, + "step": 45277 + }, + { + "epoch": 8.23, + "learning_rate": 5.303239898070622e-07, + "loss": 0.3427, + "step": 45278 + }, + { + "epoch": 8.23, + "learning_rate": 5.302693847834002e-07, + "loss": 0.8065, + "step": 45279 + }, + { + "epoch": 8.23, + "learning_rate": 5.302147797597379e-07, + "loss": 0.0048, + "step": 45280 + }, + { + "epoch": 8.23, + "learning_rate": 5.301601747360757e-07, + "loss": 0.3591, + "step": 45281 + }, + { + "epoch": 8.23, + "learning_rate": 5.301055697124136e-07, + "loss": 0.3014, + "step": 45282 + }, + { + "epoch": 8.23, + "learning_rate": 5.300509646887514e-07, + "loss": 0.1022, + "step": 45283 + }, + { + "epoch": 8.23, + "learning_rate": 5.299963596650891e-07, + "loss": 1.3558, + "step": 45284 + }, + { + "epoch": 8.24, + "learning_rate": 5.29941754641427e-07, + "loss": 0.5842, + "step": 45285 + }, + { + "epoch": 8.24, + "learning_rate": 5.298871496177649e-07, + "loss": 0.0191, + "step": 45286 + }, + { + "epoch": 8.24, + "learning_rate": 5.298325445941027e-07, + "loss": 1.4235, + "step": 45287 + }, + { + "epoch": 8.24, + "learning_rate": 5.297779395704405e-07, + "loss": 0.0559, + "step": 45288 + }, + { + "epoch": 8.24, + "learning_rate": 5.297233345467783e-07, + "loss": 0.2001, + "step": 45289 + }, + { + "epoch": 8.24, + "learning_rate": 5.296687295231162e-07, + "loss": 0.5286, + "step": 45290 + }, + { + "epoch": 8.24, + "learning_rate": 5.296141244994539e-07, + "loss": 0.3613, + "step": 45291 + }, + { + "epoch": 8.24, + "learning_rate": 5.295595194757917e-07, + "loss": 0.5308, + "step": 45292 + }, + { + "epoch": 8.24, + "learning_rate": 5.295049144521297e-07, + "loss": 0.303, + "step": 45293 + }, + { + "epoch": 8.24, + "learning_rate": 5.294503094284674e-07, + "loss": 1.4013, + "step": 45294 + }, + { + "epoch": 8.24, + "learning_rate": 5.293957044048052e-07, + "loss": 1.1499, + "step": 45295 + }, + { + "epoch": 8.24, + "learning_rate": 5.293410993811431e-07, + "loss": 1.438, + "step": 45296 + }, + { + "epoch": 8.24, + "learning_rate": 5.292864943574809e-07, + "loss": 0.2964, + "step": 45297 + }, + { + "epoch": 8.24, + "learning_rate": 5.292318893338187e-07, + "loss": 1.0118, + "step": 45298 + }, + { + "epoch": 8.24, + "learning_rate": 5.291772843101566e-07, + "loss": 0.0502, + "step": 45299 + }, + { + "epoch": 8.24, + "learning_rate": 5.291226792864944e-07, + "loss": 0.0273, + "step": 45300 + }, + { + "epoch": 8.24, + "learning_rate": 5.290680742628322e-07, + "loss": 0.2912, + "step": 45301 + }, + { + "epoch": 8.24, + "learning_rate": 5.2901346923917e-07, + "loss": 1.885, + "step": 45302 + }, + { + "epoch": 8.24, + "learning_rate": 5.289588642155078e-07, + "loss": 0.0649, + "step": 45303 + }, + { + "epoch": 8.24, + "learning_rate": 5.289042591918457e-07, + "loss": 1.9994, + "step": 45304 + }, + { + "epoch": 8.24, + "learning_rate": 5.288496541681834e-07, + "loss": 0.0227, + "step": 45305 + }, + { + "epoch": 8.24, + "learning_rate": 5.287950491445214e-07, + "loss": 0.7099, + "step": 45306 + }, + { + "epoch": 8.24, + "learning_rate": 5.287404441208592e-07, + "loss": 1.5098, + "step": 45307 + }, + { + "epoch": 8.24, + "learning_rate": 5.286858390971969e-07, + "loss": 0.029, + "step": 45308 + }, + { + "epoch": 8.24, + "learning_rate": 5.286312340735348e-07, + "loss": 0.1551, + "step": 45309 + }, + { + "epoch": 8.24, + "learning_rate": 5.285766290498726e-07, + "loss": 0.4499, + "step": 45310 + }, + { + "epoch": 8.24, + "learning_rate": 5.285220240262104e-07, + "loss": 0.1305, + "step": 45311 + }, + { + "epoch": 8.24, + "learning_rate": 5.284674190025482e-07, + "loss": 0.8601, + "step": 45312 + }, + { + "epoch": 8.24, + "learning_rate": 5.284128139788861e-07, + "loss": 0.0506, + "step": 45313 + }, + { + "epoch": 8.24, + "learning_rate": 5.283582089552238e-07, + "loss": 0.4065, + "step": 45314 + }, + { + "epoch": 8.24, + "learning_rate": 5.283036039315617e-07, + "loss": 0.0913, + "step": 45315 + }, + { + "epoch": 8.24, + "learning_rate": 5.282489989078996e-07, + "loss": 1.1487, + "step": 45316 + }, + { + "epoch": 8.24, + "learning_rate": 5.281943938842374e-07, + "loss": 0.0442, + "step": 45317 + }, + { + "epoch": 8.24, + "learning_rate": 5.281397888605752e-07, + "loss": 0.0515, + "step": 45318 + }, + { + "epoch": 8.24, + "learning_rate": 5.28085183836913e-07, + "loss": 0.0148, + "step": 45319 + }, + { + "epoch": 8.24, + "learning_rate": 5.280305788132509e-07, + "loss": 1.2358, + "step": 45320 + }, + { + "epoch": 8.24, + "learning_rate": 5.279759737895886e-07, + "loss": 0.1414, + "step": 45321 + }, + { + "epoch": 8.24, + "learning_rate": 5.279213687659264e-07, + "loss": 0.7226, + "step": 45322 + }, + { + "epoch": 8.24, + "learning_rate": 5.278667637422644e-07, + "loss": 0.1394, + "step": 45323 + }, + { + "epoch": 8.24, + "learning_rate": 5.278121587186021e-07, + "loss": 1.7639, + "step": 45324 + }, + { + "epoch": 8.24, + "learning_rate": 5.2775755369494e-07, + "loss": 0.2477, + "step": 45325 + }, + { + "epoch": 8.24, + "learning_rate": 5.277029486712778e-07, + "loss": 0.1829, + "step": 45326 + }, + { + "epoch": 8.24, + "learning_rate": 5.276483436476156e-07, + "loss": 0.8897, + "step": 45327 + }, + { + "epoch": 8.24, + "learning_rate": 5.275937386239534e-07, + "loss": 0.0716, + "step": 45328 + }, + { + "epoch": 8.24, + "learning_rate": 5.275391336002912e-07, + "loss": 0.4423, + "step": 45329 + }, + { + "epoch": 8.24, + "learning_rate": 5.274845285766291e-07, + "loss": 0.6417, + "step": 45330 + }, + { + "epoch": 8.24, + "learning_rate": 5.274299235529669e-07, + "loss": 1.5973, + "step": 45331 + }, + { + "epoch": 8.24, + "learning_rate": 5.273753185293047e-07, + "loss": 0.6581, + "step": 45332 + }, + { + "epoch": 8.24, + "learning_rate": 5.273207135056425e-07, + "loss": 0.1523, + "step": 45333 + }, + { + "epoch": 8.24, + "learning_rate": 5.272661084819804e-07, + "loss": 0.4876, + "step": 45334 + }, + { + "epoch": 8.24, + "learning_rate": 5.272115034583181e-07, + "loss": 0.3776, + "step": 45335 + }, + { + "epoch": 8.24, + "learning_rate": 5.271568984346561e-07, + "loss": 1.931, + "step": 45336 + }, + { + "epoch": 8.24, + "learning_rate": 5.271022934109939e-07, + "loss": 0.7635, + "step": 45337 + }, + { + "epoch": 8.24, + "learning_rate": 5.270476883873316e-07, + "loss": 2.3959, + "step": 45338 + }, + { + "epoch": 8.24, + "learning_rate": 5.269930833636695e-07, + "loss": 0.0662, + "step": 45339 + }, + { + "epoch": 8.25, + "learning_rate": 5.269384783400073e-07, + "loss": 0.0336, + "step": 45340 + }, + { + "epoch": 8.25, + "learning_rate": 5.268838733163451e-07, + "loss": 1.2872, + "step": 45341 + }, + { + "epoch": 8.25, + "learning_rate": 5.268292682926829e-07, + "loss": 0.0083, + "step": 45342 + }, + { + "epoch": 8.25, + "learning_rate": 5.267746632690208e-07, + "loss": 3.443, + "step": 45343 + }, + { + "epoch": 8.25, + "learning_rate": 5.267200582453587e-07, + "loss": 0.4361, + "step": 45344 + }, + { + "epoch": 8.25, + "learning_rate": 5.266654532216964e-07, + "loss": 1.1141, + "step": 45345 + }, + { + "epoch": 8.25, + "learning_rate": 5.266108481980342e-07, + "loss": 0.0152, + "step": 45346 + }, + { + "epoch": 8.25, + "learning_rate": 5.265562431743721e-07, + "loss": 1.153, + "step": 45347 + }, + { + "epoch": 8.25, + "learning_rate": 5.265016381507099e-07, + "loss": 0.0182, + "step": 45348 + }, + { + "epoch": 8.25, + "learning_rate": 5.264470331270476e-07, + "loss": 0.3572, + "step": 45349 + }, + { + "epoch": 8.25, + "learning_rate": 5.263924281033856e-07, + "loss": 0.0729, + "step": 45350 + }, + { + "epoch": 8.25, + "learning_rate": 5.263378230797233e-07, + "loss": 0.1142, + "step": 45351 + }, + { + "epoch": 8.25, + "learning_rate": 5.262832180560611e-07, + "loss": 1.3151, + "step": 45352 + }, + { + "epoch": 8.25, + "learning_rate": 5.26228613032399e-07, + "loss": 2.1489, + "step": 45353 + }, + { + "epoch": 8.25, + "learning_rate": 5.261740080087368e-07, + "loss": 1.1838, + "step": 45354 + }, + { + "epoch": 8.25, + "learning_rate": 5.261194029850747e-07, + "loss": 1.3695, + "step": 45355 + }, + { + "epoch": 8.25, + "learning_rate": 5.260647979614125e-07, + "loss": 0.0293, + "step": 45356 + }, + { + "epoch": 8.25, + "learning_rate": 5.260101929377503e-07, + "loss": 0.0038, + "step": 45357 + }, + { + "epoch": 8.25, + "learning_rate": 5.259555879140881e-07, + "loss": 1.6129, + "step": 45358 + }, + { + "epoch": 8.25, + "learning_rate": 5.259009828904259e-07, + "loss": 2.4872, + "step": 45359 + }, + { + "epoch": 8.25, + "learning_rate": 5.258463778667638e-07, + "loss": 0.4405, + "step": 45360 + }, + { + "epoch": 8.25, + "learning_rate": 5.257917728431016e-07, + "loss": 0.0058, + "step": 45361 + }, + { + "epoch": 8.25, + "learning_rate": 5.257371678194394e-07, + "loss": 1.4653, + "step": 45362 + }, + { + "epoch": 8.25, + "learning_rate": 5.256825627957772e-07, + "loss": 0.0396, + "step": 45363 + }, + { + "epoch": 8.25, + "learning_rate": 5.256279577721151e-07, + "loss": 0.043, + "step": 45364 + }, + { + "epoch": 8.25, + "learning_rate": 5.255733527484528e-07, + "loss": 1.2035, + "step": 45365 + }, + { + "epoch": 8.25, + "learning_rate": 5.255187477247907e-07, + "loss": 1.7837, + "step": 45366 + }, + { + "epoch": 8.25, + "learning_rate": 5.254641427011286e-07, + "loss": 0.0087, + "step": 45367 + }, + { + "epoch": 8.25, + "learning_rate": 5.254095376774663e-07, + "loss": 0.742, + "step": 45368 + }, + { + "epoch": 8.25, + "learning_rate": 5.253549326538042e-07, + "loss": 0.017, + "step": 45369 + }, + { + "epoch": 8.25, + "learning_rate": 5.25300327630142e-07, + "loss": 0.4512, + "step": 45370 + }, + { + "epoch": 8.25, + "learning_rate": 5.252457226064798e-07, + "loss": 0.3766, + "step": 45371 + }, + { + "epoch": 8.25, + "learning_rate": 5.251911175828176e-07, + "loss": 0.0086, + "step": 45372 + }, + { + "epoch": 8.25, + "learning_rate": 5.251365125591554e-07, + "loss": 0.5232, + "step": 45373 + }, + { + "epoch": 8.25, + "learning_rate": 5.250819075354934e-07, + "loss": 0.6316, + "step": 45374 + }, + { + "epoch": 8.25, + "learning_rate": 5.250273025118311e-07, + "loss": 0.0194, + "step": 45375 + }, + { + "epoch": 8.25, + "learning_rate": 5.249726974881689e-07, + "loss": 0.0256, + "step": 45376 + }, + { + "epoch": 8.25, + "learning_rate": 5.249180924645068e-07, + "loss": 0.3044, + "step": 45377 + }, + { + "epoch": 8.25, + "learning_rate": 5.248634874408446e-07, + "loss": 0.002, + "step": 45378 + }, + { + "epoch": 8.25, + "learning_rate": 5.248088824171823e-07, + "loss": 0.0635, + "step": 45379 + }, + { + "epoch": 8.25, + "learning_rate": 5.247542773935203e-07, + "loss": 0.0043, + "step": 45380 + }, + { + "epoch": 8.25, + "learning_rate": 5.246996723698581e-07, + "loss": 0.2655, + "step": 45381 + }, + { + "epoch": 8.25, + "learning_rate": 5.246450673461958e-07, + "loss": 0.0292, + "step": 45382 + }, + { + "epoch": 8.25, + "learning_rate": 5.245904623225337e-07, + "loss": 0.0104, + "step": 45383 + }, + { + "epoch": 8.25, + "learning_rate": 5.245358572988715e-07, + "loss": 0.7006, + "step": 45384 + }, + { + "epoch": 8.25, + "learning_rate": 5.244812522752094e-07, + "loss": 0.0852, + "step": 45385 + }, + { + "epoch": 8.25, + "learning_rate": 5.244266472515471e-07, + "loss": 2.2973, + "step": 45386 + }, + { + "epoch": 8.25, + "learning_rate": 5.24372042227885e-07, + "loss": 0.0871, + "step": 45387 + }, + { + "epoch": 8.25, + "learning_rate": 5.243174372042228e-07, + "loss": 0.2406, + "step": 45388 + }, + { + "epoch": 8.25, + "learning_rate": 5.242628321805606e-07, + "loss": 1.7603, + "step": 45389 + }, + { + "epoch": 8.25, + "learning_rate": 5.242082271568984e-07, + "loss": 0.004, + "step": 45390 + }, + { + "epoch": 8.25, + "learning_rate": 5.241536221332363e-07, + "loss": 0.2134, + "step": 45391 + }, + { + "epoch": 8.25, + "learning_rate": 5.240990171095741e-07, + "loss": 0.9507, + "step": 45392 + }, + { + "epoch": 8.25, + "learning_rate": 5.240444120859118e-07, + "loss": 0.4054, + "step": 45393 + }, + { + "epoch": 8.25, + "learning_rate": 5.239898070622498e-07, + "loss": 1.5841, + "step": 45394 + }, + { + "epoch": 8.26, + "learning_rate": 5.239352020385875e-07, + "loss": 0.4708, + "step": 45395 + }, + { + "epoch": 8.26, + "learning_rate": 5.238805970149254e-07, + "loss": 0.7456, + "step": 45396 + }, + { + "epoch": 8.26, + "learning_rate": 5.238259919912632e-07, + "loss": 2.0787, + "step": 45397 + }, + { + "epoch": 8.26, + "learning_rate": 5.23771386967601e-07, + "loss": 0.0752, + "step": 45398 + }, + { + "epoch": 8.26, + "learning_rate": 5.237167819439389e-07, + "loss": 0.7462, + "step": 45399 + }, + { + "epoch": 8.26, + "learning_rate": 5.236621769202767e-07, + "loss": 0.0764, + "step": 45400 + }, + { + "epoch": 8.26, + "learning_rate": 5.236075718966145e-07, + "loss": 0.0502, + "step": 45401 + }, + { + "epoch": 8.26, + "learning_rate": 5.235529668729523e-07, + "loss": 0.6878, + "step": 45402 + }, + { + "epoch": 8.26, + "learning_rate": 5.234983618492901e-07, + "loss": 0.3096, + "step": 45403 + }, + { + "epoch": 8.26, + "learning_rate": 5.234437568256281e-07, + "loss": 0.0323, + "step": 45404 + }, + { + "epoch": 8.26, + "learning_rate": 5.233891518019658e-07, + "loss": 0.5572, + "step": 45405 + }, + { + "epoch": 8.26, + "learning_rate": 5.233345467783036e-07, + "loss": 0.0295, + "step": 45406 + }, + { + "epoch": 8.26, + "learning_rate": 5.232799417546415e-07, + "loss": 1.7754, + "step": 45407 + }, + { + "epoch": 8.26, + "learning_rate": 5.232253367309793e-07, + "loss": 0.0086, + "step": 45408 + }, + { + "epoch": 8.26, + "learning_rate": 5.23170731707317e-07, + "loss": 0.2388, + "step": 45409 + }, + { + "epoch": 8.26, + "learning_rate": 5.231161266836549e-07, + "loss": 1.2277, + "step": 45410 + }, + { + "epoch": 8.26, + "learning_rate": 5.230615216599928e-07, + "loss": 0.0484, + "step": 45411 + }, + { + "epoch": 8.26, + "learning_rate": 5.230069166363305e-07, + "loss": 2.5622, + "step": 45412 + }, + { + "epoch": 8.26, + "learning_rate": 5.229523116126684e-07, + "loss": 0.1813, + "step": 45413 + }, + { + "epoch": 8.26, + "learning_rate": 5.228977065890062e-07, + "loss": 0.8063, + "step": 45414 + }, + { + "epoch": 8.26, + "learning_rate": 5.228431015653441e-07, + "loss": 2.1337, + "step": 45415 + }, + { + "epoch": 8.26, + "learning_rate": 5.227884965416818e-07, + "loss": 0.1167, + "step": 45416 + }, + { + "epoch": 8.26, + "learning_rate": 5.227338915180196e-07, + "loss": 1.2773, + "step": 45417 + }, + { + "epoch": 8.26, + "learning_rate": 5.226792864943576e-07, + "loss": 0.0129, + "step": 45418 + }, + { + "epoch": 8.26, + "learning_rate": 5.226246814706953e-07, + "loss": 0.0689, + "step": 45419 + }, + { + "epoch": 8.26, + "learning_rate": 5.225700764470331e-07, + "loss": 0.0048, + "step": 45420 + }, + { + "epoch": 8.26, + "learning_rate": 5.22515471423371e-07, + "loss": 0.1748, + "step": 45421 + }, + { + "epoch": 8.26, + "learning_rate": 5.224608663997088e-07, + "loss": 0.3898, + "step": 45422 + }, + { + "epoch": 8.26, + "learning_rate": 5.224062613760465e-07, + "loss": 0.0401, + "step": 45423 + }, + { + "epoch": 8.26, + "learning_rate": 5.223516563523845e-07, + "loss": 0.1036, + "step": 45424 + }, + { + "epoch": 8.26, + "learning_rate": 5.222970513287223e-07, + "loss": 0.1618, + "step": 45425 + }, + { + "epoch": 8.26, + "learning_rate": 5.222424463050601e-07, + "loss": 1.0761, + "step": 45426 + }, + { + "epoch": 8.26, + "learning_rate": 5.221878412813979e-07, + "loss": 0.0038, + "step": 45427 + }, + { + "epoch": 8.26, + "learning_rate": 5.221332362577357e-07, + "loss": 1.1955, + "step": 45428 + }, + { + "epoch": 8.26, + "learning_rate": 5.220786312340736e-07, + "loss": 0.0109, + "step": 45429 + }, + { + "epoch": 8.26, + "learning_rate": 5.220240262104113e-07, + "loss": 0.0916, + "step": 45430 + }, + { + "epoch": 8.26, + "learning_rate": 5.219694211867492e-07, + "loss": 0.548, + "step": 45431 + }, + { + "epoch": 8.26, + "learning_rate": 5.21914816163087e-07, + "loss": 0.3289, + "step": 45432 + }, + { + "epoch": 8.26, + "learning_rate": 5.218602111394248e-07, + "loss": 0.0111, + "step": 45433 + }, + { + "epoch": 8.26, + "learning_rate": 5.218056061157627e-07, + "loss": 2.3222, + "step": 45434 + }, + { + "epoch": 8.26, + "learning_rate": 5.217510010921005e-07, + "loss": 0.9086, + "step": 45435 + }, + { + "epoch": 8.26, + "learning_rate": 5.216963960684383e-07, + "loss": 0.8565, + "step": 45436 + }, + { + "epoch": 8.26, + "learning_rate": 5.216417910447761e-07, + "loss": 0.009, + "step": 45437 + }, + { + "epoch": 8.26, + "learning_rate": 5.21587186021114e-07, + "loss": 0.0133, + "step": 45438 + }, + { + "epoch": 8.26, + "learning_rate": 5.215325809974517e-07, + "loss": 0.0103, + "step": 45439 + }, + { + "epoch": 8.26, + "learning_rate": 5.214779759737896e-07, + "loss": 0.1358, + "step": 45440 + }, + { + "epoch": 8.26, + "learning_rate": 5.214233709501274e-07, + "loss": 0.0312, + "step": 45441 + }, + { + "epoch": 8.26, + "learning_rate": 5.213687659264652e-07, + "loss": 1.4676, + "step": 45442 + }, + { + "epoch": 8.26, + "learning_rate": 5.213141609028031e-07, + "loss": 0.1179, + "step": 45443 + }, + { + "epoch": 8.26, + "learning_rate": 5.212595558791409e-07, + "loss": 1.5308, + "step": 45444 + }, + { + "epoch": 8.26, + "learning_rate": 5.212049508554788e-07, + "loss": 0.3048, + "step": 45445 + }, + { + "epoch": 8.26, + "learning_rate": 5.211503458318165e-07, + "loss": 0.777, + "step": 45446 + }, + { + "epoch": 8.26, + "learning_rate": 5.210957408081543e-07, + "loss": 0.0029, + "step": 45447 + }, + { + "epoch": 8.26, + "learning_rate": 5.210411357844923e-07, + "loss": 0.0091, + "step": 45448 + }, + { + "epoch": 8.26, + "learning_rate": 5.2098653076083e-07, + "loss": 0.0947, + "step": 45449 + }, + { + "epoch": 8.27, + "learning_rate": 5.209319257371678e-07, + "loss": 0.0451, + "step": 45450 + }, + { + "epoch": 8.27, + "learning_rate": 5.208773207135057e-07, + "loss": 1.5546, + "step": 45451 + }, + { + "epoch": 8.27, + "learning_rate": 5.208227156898435e-07, + "loss": 1.6057, + "step": 45452 + }, + { + "epoch": 8.27, + "learning_rate": 5.207681106661812e-07, + "loss": 0.0634, + "step": 45453 + }, + { + "epoch": 8.27, + "learning_rate": 5.207135056425191e-07, + "loss": 1.1595, + "step": 45454 + }, + { + "epoch": 8.27, + "learning_rate": 5.20658900618857e-07, + "loss": 1.9632, + "step": 45455 + }, + { + "epoch": 8.27, + "learning_rate": 5.206042955951948e-07, + "loss": 0.1528, + "step": 45456 + }, + { + "epoch": 8.27, + "learning_rate": 5.205496905715326e-07, + "loss": 0.0108, + "step": 45457 + }, + { + "epoch": 8.27, + "learning_rate": 5.204950855478704e-07, + "loss": 0.1499, + "step": 45458 + }, + { + "epoch": 8.27, + "learning_rate": 5.204404805242083e-07, + "loss": 0.1651, + "step": 45459 + }, + { + "epoch": 8.27, + "learning_rate": 5.20385875500546e-07, + "loss": 0.4524, + "step": 45460 + }, + { + "epoch": 8.27, + "learning_rate": 5.203312704768838e-07, + "loss": 0.7315, + "step": 45461 + }, + { + "epoch": 8.27, + "learning_rate": 5.202766654532218e-07, + "loss": 0.2957, + "step": 45462 + }, + { + "epoch": 8.27, + "learning_rate": 5.202220604295595e-07, + "loss": 0.0783, + "step": 45463 + }, + { + "epoch": 8.27, + "learning_rate": 5.201674554058974e-07, + "loss": 0.8555, + "step": 45464 + }, + { + "epoch": 8.27, + "learning_rate": 5.201128503822352e-07, + "loss": 0.1039, + "step": 45465 + }, + { + "epoch": 8.27, + "learning_rate": 5.20058245358573e-07, + "loss": 0.199, + "step": 45466 + }, + { + "epoch": 8.27, + "learning_rate": 5.200036403349108e-07, + "loss": 0.0182, + "step": 45467 + }, + { + "epoch": 8.27, + "learning_rate": 5.199490353112487e-07, + "loss": 0.8376, + "step": 45468 + }, + { + "epoch": 8.27, + "learning_rate": 5.198944302875864e-07, + "loss": 0.1194, + "step": 45469 + }, + { + "epoch": 8.27, + "learning_rate": 5.198398252639243e-07, + "loss": 3.5751, + "step": 45470 + }, + { + "epoch": 8.27, + "learning_rate": 5.197852202402621e-07, + "loss": 0.482, + "step": 45471 + }, + { + "epoch": 8.27, + "learning_rate": 5.197306152165999e-07, + "loss": 1.6679, + "step": 45472 + }, + { + "epoch": 8.27, + "learning_rate": 5.196760101929378e-07, + "loss": 0.8454, + "step": 45473 + }, + { + "epoch": 8.27, + "learning_rate": 5.196214051692755e-07, + "loss": 0.0366, + "step": 45474 + }, + { + "epoch": 8.27, + "learning_rate": 5.195668001456135e-07, + "loss": 0.0051, + "step": 45475 + }, + { + "epoch": 8.27, + "learning_rate": 5.195121951219512e-07, + "loss": 0.6822, + "step": 45476 + }, + { + "epoch": 8.27, + "learning_rate": 5.19457590098289e-07, + "loss": 0.8502, + "step": 45477 + }, + { + "epoch": 8.27, + "learning_rate": 5.194029850746269e-07, + "loss": 0.1122, + "step": 45478 + }, + { + "epoch": 8.27, + "learning_rate": 5.193483800509647e-07, + "loss": 0.0695, + "step": 45479 + }, + { + "epoch": 8.27, + "learning_rate": 5.192937750273025e-07, + "loss": 0.2752, + "step": 45480 + }, + { + "epoch": 8.27, + "learning_rate": 5.192391700036403e-07, + "loss": 0.0041, + "step": 45481 + }, + { + "epoch": 8.27, + "learning_rate": 5.191845649799782e-07, + "loss": 1.2459, + "step": 45482 + }, + { + "epoch": 8.27, + "learning_rate": 5.19129959956316e-07, + "loss": 0.0726, + "step": 45483 + }, + { + "epoch": 8.27, + "learning_rate": 5.190753549326538e-07, + "loss": 0.4382, + "step": 45484 + }, + { + "epoch": 8.27, + "learning_rate": 5.190207499089917e-07, + "loss": 0.3725, + "step": 45485 + }, + { + "epoch": 8.27, + "learning_rate": 5.189661448853295e-07, + "loss": 0.0098, + "step": 45486 + }, + { + "epoch": 8.27, + "learning_rate": 5.189115398616673e-07, + "loss": 0.015, + "step": 45487 + }, + { + "epoch": 8.27, + "learning_rate": 5.188569348380051e-07, + "loss": 0.4029, + "step": 45488 + }, + { + "epoch": 8.27, + "learning_rate": 5.18802329814343e-07, + "loss": 0.9327, + "step": 45489 + }, + { + "epoch": 8.27, + "learning_rate": 5.187477247906807e-07, + "loss": 0.2613, + "step": 45490 + }, + { + "epoch": 8.27, + "learning_rate": 5.186931197670185e-07, + "loss": 1.3651, + "step": 45491 + }, + { + "epoch": 8.27, + "learning_rate": 5.186385147433565e-07, + "loss": 1.1849, + "step": 45492 + }, + { + "epoch": 8.27, + "learning_rate": 5.185839097196942e-07, + "loss": 1.8539, + "step": 45493 + }, + { + "epoch": 8.27, + "learning_rate": 5.185293046960321e-07, + "loss": 1.837, + "step": 45494 + }, + { + "epoch": 8.27, + "learning_rate": 5.184746996723699e-07, + "loss": 0.1332, + "step": 45495 + }, + { + "epoch": 8.27, + "learning_rate": 5.184200946487077e-07, + "loss": 0.0289, + "step": 45496 + }, + { + "epoch": 8.27, + "learning_rate": 5.183654896250455e-07, + "loss": 0.0061, + "step": 45497 + }, + { + "epoch": 8.27, + "learning_rate": 5.183108846013833e-07, + "loss": 0.0873, + "step": 45498 + }, + { + "epoch": 8.27, + "learning_rate": 5.182562795777212e-07, + "loss": 0.107, + "step": 45499 + }, + { + "epoch": 8.27, + "learning_rate": 5.18201674554059e-07, + "loss": 0.7842, + "step": 45500 + }, + { + "epoch": 8.27, + "learning_rate": 5.181470695303968e-07, + "loss": 0.5704, + "step": 45501 + }, + { + "epoch": 8.27, + "learning_rate": 5.180924645067346e-07, + "loss": 0.05, + "step": 45502 + }, + { + "epoch": 8.27, + "learning_rate": 5.180378594830725e-07, + "loss": 0.2365, + "step": 45503 + }, + { + "epoch": 8.27, + "learning_rate": 5.179832544594102e-07, + "loss": 0.1413, + "step": 45504 + }, + { + "epoch": 8.28, + "learning_rate": 5.179286494357482e-07, + "loss": 2.4723, + "step": 45505 + }, + { + "epoch": 8.28, + "learning_rate": 5.178740444120859e-07, + "loss": 0.1492, + "step": 45506 + }, + { + "epoch": 8.28, + "learning_rate": 5.178194393884237e-07, + "loss": 0.3131, + "step": 45507 + }, + { + "epoch": 8.28, + "learning_rate": 5.177648343647616e-07, + "loss": 0.3866, + "step": 45508 + }, + { + "epoch": 8.28, + "learning_rate": 5.177102293410994e-07, + "loss": 0.6257, + "step": 45509 + }, + { + "epoch": 8.28, + "learning_rate": 5.176556243174372e-07, + "loss": 0.9919, + "step": 45510 + }, + { + "epoch": 8.28, + "learning_rate": 5.17601019293775e-07, + "loss": 0.2075, + "step": 45511 + }, + { + "epoch": 8.28, + "learning_rate": 5.175464142701129e-07, + "loss": 0.5749, + "step": 45512 + }, + { + "epoch": 8.28, + "learning_rate": 5.174918092464506e-07, + "loss": 0.3225, + "step": 45513 + }, + { + "epoch": 8.28, + "learning_rate": 5.174372042227885e-07, + "loss": 0.3684, + "step": 45514 + }, + { + "epoch": 8.28, + "learning_rate": 5.173825991991263e-07, + "loss": 0.9181, + "step": 45515 + }, + { + "epoch": 8.28, + "learning_rate": 5.173279941754642e-07, + "loss": 0.0044, + "step": 45516 + }, + { + "epoch": 8.28, + "learning_rate": 5.17273389151802e-07, + "loss": 0.0046, + "step": 45517 + }, + { + "epoch": 8.28, + "learning_rate": 5.172187841281397e-07, + "loss": 1.411, + "step": 45518 + }, + { + "epoch": 8.28, + "learning_rate": 5.171641791044777e-07, + "loss": 0.4683, + "step": 45519 + }, + { + "epoch": 8.28, + "learning_rate": 5.171095740808154e-07, + "loss": 0.3437, + "step": 45520 + }, + { + "epoch": 8.28, + "learning_rate": 5.170549690571532e-07, + "loss": 0.0057, + "step": 45521 + }, + { + "epoch": 8.28, + "learning_rate": 5.170003640334911e-07, + "loss": 1.5132, + "step": 45522 + }, + { + "epoch": 8.28, + "learning_rate": 5.169457590098289e-07, + "loss": 0.364, + "step": 45523 + }, + { + "epoch": 8.28, + "learning_rate": 5.168911539861668e-07, + "loss": 0.0132, + "step": 45524 + }, + { + "epoch": 8.28, + "learning_rate": 5.168365489625045e-07, + "loss": 0.0838, + "step": 45525 + }, + { + "epoch": 8.28, + "learning_rate": 5.167819439388424e-07, + "loss": 0.0399, + "step": 45526 + }, + { + "epoch": 8.28, + "learning_rate": 5.167273389151802e-07, + "loss": 0.0483, + "step": 45527 + }, + { + "epoch": 8.28, + "learning_rate": 5.16672733891518e-07, + "loss": 0.687, + "step": 45528 + }, + { + "epoch": 8.28, + "learning_rate": 5.166181288678559e-07, + "loss": 2.174, + "step": 45529 + }, + { + "epoch": 8.28, + "learning_rate": 5.165635238441937e-07, + "loss": 0.0088, + "step": 45530 + }, + { + "epoch": 8.28, + "learning_rate": 5.165089188205315e-07, + "loss": 0.0064, + "step": 45531 + }, + { + "epoch": 8.28, + "learning_rate": 5.164543137968694e-07, + "loss": 0.0127, + "step": 45532 + }, + { + "epoch": 8.28, + "learning_rate": 5.163997087732072e-07, + "loss": 0.1966, + "step": 45533 + }, + { + "epoch": 8.28, + "learning_rate": 5.163451037495449e-07, + "loss": 1.2216, + "step": 45534 + }, + { + "epoch": 8.28, + "learning_rate": 5.162904987258828e-07, + "loss": 0.7917, + "step": 45535 + }, + { + "epoch": 8.28, + "learning_rate": 5.162358937022207e-07, + "loss": 0.0684, + "step": 45536 + }, + { + "epoch": 8.28, + "learning_rate": 5.161812886785584e-07, + "loss": 0.0059, + "step": 45537 + }, + { + "epoch": 8.28, + "learning_rate": 5.161266836548963e-07, + "loss": 0.9946, + "step": 45538 + }, + { + "epoch": 8.28, + "learning_rate": 5.160720786312341e-07, + "loss": 1.8001, + "step": 45539 + }, + { + "epoch": 8.28, + "learning_rate": 5.160174736075719e-07, + "loss": 0.2391, + "step": 45540 + }, + { + "epoch": 8.28, + "learning_rate": 5.159628685839097e-07, + "loss": 0.1348, + "step": 45541 + }, + { + "epoch": 8.28, + "learning_rate": 5.159082635602475e-07, + "loss": 0.0197, + "step": 45542 + }, + { + "epoch": 8.28, + "learning_rate": 5.158536585365854e-07, + "loss": 1.0633, + "step": 45543 + }, + { + "epoch": 8.28, + "learning_rate": 5.157990535129232e-07, + "loss": 0.1169, + "step": 45544 + }, + { + "epoch": 8.28, + "learning_rate": 5.15744448489261e-07, + "loss": 0.0316, + "step": 45545 + }, + { + "epoch": 8.28, + "learning_rate": 5.156898434655989e-07, + "loss": 0.8908, + "step": 45546 + }, + { + "epoch": 8.28, + "learning_rate": 5.156352384419367e-07, + "loss": 2.6178, + "step": 45547 + }, + { + "epoch": 8.28, + "learning_rate": 5.155806334182744e-07, + "loss": 0.1474, + "step": 45548 + }, + { + "epoch": 8.28, + "learning_rate": 5.155260283946124e-07, + "loss": 0.3902, + "step": 45549 + }, + { + "epoch": 8.28, + "learning_rate": 5.154714233709501e-07, + "loss": 0.116, + "step": 45550 + }, + { + "epoch": 8.28, + "learning_rate": 5.154168183472879e-07, + "loss": 1.3838, + "step": 45551 + }, + { + "epoch": 8.28, + "learning_rate": 5.153622133236258e-07, + "loss": 0.0821, + "step": 45552 + }, + { + "epoch": 8.28, + "learning_rate": 5.153076082999636e-07, + "loss": 0.1371, + "step": 45553 + }, + { + "epoch": 8.28, + "learning_rate": 5.152530032763015e-07, + "loss": 0.0775, + "step": 45554 + }, + { + "epoch": 8.28, + "learning_rate": 5.151983982526392e-07, + "loss": 0.1808, + "step": 45555 + }, + { + "epoch": 8.28, + "learning_rate": 5.151437932289771e-07, + "loss": 0.8668, + "step": 45556 + }, + { + "epoch": 8.28, + "learning_rate": 5.150891882053149e-07, + "loss": 0.9423, + "step": 45557 + }, + { + "epoch": 8.28, + "learning_rate": 5.150345831816527e-07, + "loss": 0.2304, + "step": 45558 + }, + { + "epoch": 8.28, + "learning_rate": 5.149799781579905e-07, + "loss": 0.5081, + "step": 45559 + }, + { + "epoch": 8.29, + "learning_rate": 5.149253731343284e-07, + "loss": 0.3024, + "step": 45560 + }, + { + "epoch": 8.29, + "learning_rate": 5.148707681106662e-07, + "loss": 0.0421, + "step": 45561 + }, + { + "epoch": 8.29, + "learning_rate": 5.148161630870039e-07, + "loss": 0.0572, + "step": 45562 + }, + { + "epoch": 8.29, + "learning_rate": 5.147615580633419e-07, + "loss": 0.1913, + "step": 45563 + }, + { + "epoch": 8.29, + "learning_rate": 5.147069530396796e-07, + "loss": 0.0364, + "step": 45564 + }, + { + "epoch": 8.29, + "learning_rate": 5.146523480160175e-07, + "loss": 0.0335, + "step": 45565 + }, + { + "epoch": 8.29, + "learning_rate": 5.145977429923553e-07, + "loss": 0.2382, + "step": 45566 + }, + { + "epoch": 8.29, + "learning_rate": 5.145431379686931e-07, + "loss": 0.0534, + "step": 45567 + }, + { + "epoch": 8.29, + "learning_rate": 5.14488532945031e-07, + "loss": 0.1533, + "step": 45568 + }, + { + "epoch": 8.29, + "learning_rate": 5.144339279213688e-07, + "loss": 0.0065, + "step": 45569 + }, + { + "epoch": 8.29, + "learning_rate": 5.143793228977066e-07, + "loss": 0.6078, + "step": 45570 + }, + { + "epoch": 8.29, + "learning_rate": 5.143247178740444e-07, + "loss": 0.5496, + "step": 45571 + }, + { + "epoch": 8.29, + "learning_rate": 5.142701128503822e-07, + "loss": 0.0293, + "step": 45572 + }, + { + "epoch": 8.29, + "learning_rate": 5.142155078267202e-07, + "loss": 1.0399, + "step": 45573 + }, + { + "epoch": 8.29, + "learning_rate": 5.141609028030579e-07, + "loss": 1.9122, + "step": 45574 + }, + { + "epoch": 8.29, + "learning_rate": 5.141062977793957e-07, + "loss": 0.0221, + "step": 45575 + }, + { + "epoch": 8.29, + "learning_rate": 5.140516927557336e-07, + "loss": 1.2429, + "step": 45576 + }, + { + "epoch": 8.29, + "learning_rate": 5.139970877320714e-07, + "loss": 0.4485, + "step": 45577 + }, + { + "epoch": 8.29, + "learning_rate": 5.139424827084091e-07, + "loss": 0.0159, + "step": 45578 + }, + { + "epoch": 8.29, + "learning_rate": 5.13887877684747e-07, + "loss": 0.01, + "step": 45579 + }, + { + "epoch": 8.29, + "learning_rate": 5.138332726610849e-07, + "loss": 0.0921, + "step": 45580 + }, + { + "epoch": 8.29, + "learning_rate": 5.137786676374226e-07, + "loss": 0.6044, + "step": 45581 + }, + { + "epoch": 8.29, + "learning_rate": 5.137240626137605e-07, + "loss": 1.1994, + "step": 45582 + }, + { + "epoch": 8.29, + "learning_rate": 5.136694575900983e-07, + "loss": 1.0792, + "step": 45583 + }, + { + "epoch": 8.29, + "learning_rate": 5.136148525664362e-07, + "loss": 0.0123, + "step": 45584 + }, + { + "epoch": 8.29, + "learning_rate": 5.135602475427739e-07, + "loss": 0.0049, + "step": 45585 + }, + { + "epoch": 8.29, + "learning_rate": 5.135056425191117e-07, + "loss": 0.1788, + "step": 45586 + }, + { + "epoch": 8.29, + "learning_rate": 5.134510374954496e-07, + "loss": 0.1257, + "step": 45587 + }, + { + "epoch": 8.29, + "learning_rate": 5.133964324717874e-07, + "loss": 0.0082, + "step": 45588 + }, + { + "epoch": 8.29, + "learning_rate": 5.133418274481252e-07, + "loss": 0.0208, + "step": 45589 + }, + { + "epoch": 8.29, + "learning_rate": 5.132872224244631e-07, + "loss": 0.0352, + "step": 45590 + }, + { + "epoch": 8.29, + "learning_rate": 5.132326174008009e-07, + "loss": 0.4025, + "step": 45591 + }, + { + "epoch": 8.29, + "learning_rate": 5.131780123771386e-07, + "loss": 0.777, + "step": 45592 + }, + { + "epoch": 8.29, + "learning_rate": 5.131234073534766e-07, + "loss": 0.1371, + "step": 45593 + }, + { + "epoch": 8.29, + "learning_rate": 5.130688023298143e-07, + "loss": 0.2225, + "step": 45594 + }, + { + "epoch": 8.29, + "learning_rate": 5.130141973061522e-07, + "loss": 0.0095, + "step": 45595 + }, + { + "epoch": 8.29, + "learning_rate": 5.1295959228249e-07, + "loss": 0.0055, + "step": 45596 + }, + { + "epoch": 8.29, + "learning_rate": 5.129049872588278e-07, + "loss": 0.0132, + "step": 45597 + }, + { + "epoch": 8.29, + "learning_rate": 5.128503822351657e-07, + "loss": 0.319, + "step": 45598 + }, + { + "epoch": 8.29, + "learning_rate": 5.127957772115034e-07, + "loss": 2.0161, + "step": 45599 + }, + { + "epoch": 8.29, + "learning_rate": 5.127411721878413e-07, + "loss": 0.9931, + "step": 45600 + }, + { + "epoch": 8.29, + "learning_rate": 5.126865671641791e-07, + "loss": 0.0149, + "step": 45601 + }, + { + "epoch": 8.29, + "learning_rate": 5.126319621405169e-07, + "loss": 2.3264, + "step": 45602 + }, + { + "epoch": 8.29, + "learning_rate": 5.125773571168548e-07, + "loss": 0.0442, + "step": 45603 + }, + { + "epoch": 8.29, + "learning_rate": 5.125227520931926e-07, + "loss": 0.2889, + "step": 45604 + }, + { + "epoch": 8.29, + "learning_rate": 5.124681470695304e-07, + "loss": 0.7144, + "step": 45605 + }, + { + "epoch": 8.29, + "learning_rate": 5.124135420458682e-07, + "loss": 0.1267, + "step": 45606 + }, + { + "epoch": 8.29, + "learning_rate": 5.123589370222061e-07, + "loss": 0.4741, + "step": 45607 + }, + { + "epoch": 8.29, + "learning_rate": 5.123043319985438e-07, + "loss": 0.2451, + "step": 45608 + }, + { + "epoch": 8.29, + "learning_rate": 5.122497269748817e-07, + "loss": 0.0193, + "step": 45609 + }, + { + "epoch": 8.29, + "learning_rate": 5.121951219512195e-07, + "loss": 0.0401, + "step": 45610 + }, + { + "epoch": 8.29, + "learning_rate": 5.121405169275573e-07, + "loss": 0.0123, + "step": 45611 + }, + { + "epoch": 8.29, + "learning_rate": 5.120859119038952e-07, + "loss": 1.1503, + "step": 45612 + }, + { + "epoch": 8.29, + "learning_rate": 5.12031306880233e-07, + "loss": 0.0203, + "step": 45613 + }, + { + "epoch": 8.29, + "learning_rate": 5.119767018565709e-07, + "loss": 0.3461, + "step": 45614 + }, + { + "epoch": 8.3, + "learning_rate": 5.119220968329086e-07, + "loss": 0.5279, + "step": 45615 + }, + { + "epoch": 8.3, + "learning_rate": 5.118674918092464e-07, + "loss": 0.8495, + "step": 45616 + }, + { + "epoch": 8.3, + "learning_rate": 5.118128867855844e-07, + "loss": 0.0875, + "step": 45617 + }, + { + "epoch": 8.3, + "learning_rate": 5.117582817619221e-07, + "loss": 0.3752, + "step": 45618 + }, + { + "epoch": 8.3, + "learning_rate": 5.117036767382599e-07, + "loss": 0.1207, + "step": 45619 + }, + { + "epoch": 8.3, + "learning_rate": 5.116490717145978e-07, + "loss": 0.4679, + "step": 45620 + }, + { + "epoch": 8.3, + "learning_rate": 5.115944666909356e-07, + "loss": 0.9269, + "step": 45621 + }, + { + "epoch": 8.3, + "learning_rate": 5.115398616672734e-07, + "loss": 0.3798, + "step": 45622 + }, + { + "epoch": 8.3, + "learning_rate": 5.114852566436112e-07, + "loss": 1.7469, + "step": 45623 + }, + { + "epoch": 8.3, + "learning_rate": 5.11430651619949e-07, + "loss": 0.0062, + "step": 45624 + }, + { + "epoch": 8.3, + "learning_rate": 5.113760465962869e-07, + "loss": 1.6574, + "step": 45625 + }, + { + "epoch": 8.3, + "learning_rate": 5.113214415726247e-07, + "loss": 0.0547, + "step": 45626 + }, + { + "epoch": 8.3, + "learning_rate": 5.112668365489625e-07, + "loss": 1.4199, + "step": 45627 + }, + { + "epoch": 8.3, + "learning_rate": 5.112122315253004e-07, + "loss": 0.0764, + "step": 45628 + }, + { + "epoch": 8.3, + "learning_rate": 5.111576265016381e-07, + "loss": 0.0693, + "step": 45629 + }, + { + "epoch": 8.3, + "learning_rate": 5.111030214779759e-07, + "loss": 1.0312, + "step": 45630 + }, + { + "epoch": 8.3, + "learning_rate": 5.110484164543138e-07, + "loss": 0.0084, + "step": 45631 + }, + { + "epoch": 8.3, + "learning_rate": 5.109938114306516e-07, + "loss": 0.7798, + "step": 45632 + }, + { + "epoch": 8.3, + "learning_rate": 5.109392064069895e-07, + "loss": 1.4373, + "step": 45633 + }, + { + "epoch": 8.3, + "learning_rate": 5.108846013833273e-07, + "loss": 0.4519, + "step": 45634 + }, + { + "epoch": 8.3, + "learning_rate": 5.108299963596651e-07, + "loss": 0.3491, + "step": 45635 + }, + { + "epoch": 8.3, + "learning_rate": 5.107753913360029e-07, + "loss": 0.1648, + "step": 45636 + }, + { + "epoch": 8.3, + "learning_rate": 5.107207863123408e-07, + "loss": 0.7597, + "step": 45637 + }, + { + "epoch": 8.3, + "learning_rate": 5.106661812886785e-07, + "loss": 0.2539, + "step": 45638 + }, + { + "epoch": 8.3, + "learning_rate": 5.106115762650164e-07, + "loss": 0.8379, + "step": 45639 + }, + { + "epoch": 8.3, + "learning_rate": 5.105569712413542e-07, + "loss": 0.9226, + "step": 45640 + }, + { + "epoch": 8.3, + "learning_rate": 5.10502366217692e-07, + "loss": 0.6094, + "step": 45641 + }, + { + "epoch": 8.3, + "learning_rate": 5.104477611940299e-07, + "loss": 0.7527, + "step": 45642 + }, + { + "epoch": 8.3, + "learning_rate": 5.103931561703676e-07, + "loss": 1.1404, + "step": 45643 + }, + { + "epoch": 8.3, + "learning_rate": 5.103385511467056e-07, + "loss": 0.0048, + "step": 45644 + }, + { + "epoch": 8.3, + "learning_rate": 5.102839461230433e-07, + "loss": 0.2541, + "step": 45645 + }, + { + "epoch": 8.3, + "learning_rate": 5.102293410993811e-07, + "loss": 0.3375, + "step": 45646 + }, + { + "epoch": 8.3, + "learning_rate": 5.10174736075719e-07, + "loss": 1.2644, + "step": 45647 + }, + { + "epoch": 8.3, + "learning_rate": 5.101201310520568e-07, + "loss": 0.4431, + "step": 45648 + }, + { + "epoch": 8.3, + "learning_rate": 5.100655260283946e-07, + "loss": 0.0739, + "step": 45649 + }, + { + "epoch": 8.3, + "learning_rate": 5.100109210047324e-07, + "loss": 0.0924, + "step": 45650 + }, + { + "epoch": 8.3, + "learning_rate": 5.099563159810703e-07, + "loss": 1.1441, + "step": 45651 + }, + { + "epoch": 8.3, + "learning_rate": 5.09901710957408e-07, + "loss": 0.4759, + "step": 45652 + }, + { + "epoch": 8.3, + "learning_rate": 5.098471059337459e-07, + "loss": 0.287, + "step": 45653 + }, + { + "epoch": 8.3, + "learning_rate": 5.097925009100838e-07, + "loss": 0.0036, + "step": 45654 + }, + { + "epoch": 8.3, + "learning_rate": 5.097378958864216e-07, + "loss": 0.0935, + "step": 45655 + }, + { + "epoch": 8.3, + "learning_rate": 5.096832908627594e-07, + "loss": 1.5039, + "step": 45656 + }, + { + "epoch": 8.3, + "learning_rate": 5.096286858390972e-07, + "loss": 0.3816, + "step": 45657 + }, + { + "epoch": 8.3, + "learning_rate": 5.095740808154351e-07, + "loss": 0.0228, + "step": 45658 + }, + { + "epoch": 8.3, + "learning_rate": 5.095194757917728e-07, + "loss": 1.1698, + "step": 45659 + }, + { + "epoch": 8.3, + "learning_rate": 5.094648707681106e-07, + "loss": 0.8069, + "step": 45660 + }, + { + "epoch": 8.3, + "learning_rate": 5.094102657444485e-07, + "loss": 0.1326, + "step": 45661 + }, + { + "epoch": 8.3, + "learning_rate": 5.093556607207863e-07, + "loss": 0.0137, + "step": 45662 + }, + { + "epoch": 8.3, + "learning_rate": 5.093010556971242e-07, + "loss": 1.3932, + "step": 45663 + }, + { + "epoch": 8.3, + "learning_rate": 5.09246450673462e-07, + "loss": 0.0395, + "step": 45664 + }, + { + "epoch": 8.3, + "learning_rate": 5.091918456497998e-07, + "loss": 0.2489, + "step": 45665 + }, + { + "epoch": 8.3, + "learning_rate": 5.091372406261376e-07, + "loss": 0.486, + "step": 45666 + }, + { + "epoch": 8.3, + "learning_rate": 5.090826356024754e-07, + "loss": 1.7501, + "step": 45667 + }, + { + "epoch": 8.3, + "learning_rate": 5.090280305788132e-07, + "loss": 0.5899, + "step": 45668 + }, + { + "epoch": 8.3, + "learning_rate": 5.089734255551511e-07, + "loss": 0.4717, + "step": 45669 + }, + { + "epoch": 8.31, + "learning_rate": 5.089188205314889e-07, + "loss": 0.7357, + "step": 45670 + }, + { + "epoch": 8.31, + "learning_rate": 5.088642155078268e-07, + "loss": 0.0415, + "step": 45671 + }, + { + "epoch": 8.31, + "learning_rate": 5.088096104841646e-07, + "loss": 0.8243, + "step": 45672 + }, + { + "epoch": 8.31, + "learning_rate": 5.087550054605023e-07, + "loss": 0.2814, + "step": 45673 + }, + { + "epoch": 8.31, + "learning_rate": 5.087004004368402e-07, + "loss": 0.008, + "step": 45674 + }, + { + "epoch": 8.31, + "learning_rate": 5.08645795413178e-07, + "loss": 1.127, + "step": 45675 + }, + { + "epoch": 8.31, + "learning_rate": 5.085911903895158e-07, + "loss": 0.2068, + "step": 45676 + }, + { + "epoch": 8.31, + "learning_rate": 5.085365853658537e-07, + "loss": 0.7042, + "step": 45677 + }, + { + "epoch": 8.31, + "learning_rate": 5.084819803421915e-07, + "loss": 0.0533, + "step": 45678 + }, + { + "epoch": 8.31, + "learning_rate": 5.084273753185293e-07, + "loss": 0.1132, + "step": 45679 + }, + { + "epoch": 8.31, + "learning_rate": 5.083727702948671e-07, + "loss": 1.0573, + "step": 45680 + }, + { + "epoch": 8.31, + "learning_rate": 5.08318165271205e-07, + "loss": 1.3072, + "step": 45681 + }, + { + "epoch": 8.31, + "learning_rate": 5.082635602475428e-07, + "loss": 0.123, + "step": 45682 + }, + { + "epoch": 8.31, + "learning_rate": 5.082089552238806e-07, + "loss": 0.0512, + "step": 45683 + }, + { + "epoch": 8.31, + "learning_rate": 5.081543502002184e-07, + "loss": 1.1109, + "step": 45684 + }, + { + "epoch": 8.31, + "learning_rate": 5.080997451765563e-07, + "loss": 0.2886, + "step": 45685 + }, + { + "epoch": 8.31, + "learning_rate": 5.080451401528941e-07, + "loss": 0.2947, + "step": 45686 + }, + { + "epoch": 8.31, + "learning_rate": 5.079905351292318e-07, + "loss": 0.311, + "step": 45687 + }, + { + "epoch": 8.31, + "learning_rate": 5.079359301055698e-07, + "loss": 0.0737, + "step": 45688 + }, + { + "epoch": 8.31, + "learning_rate": 5.078813250819075e-07, + "loss": 0.083, + "step": 45689 + }, + { + "epoch": 8.31, + "learning_rate": 5.078267200582453e-07, + "loss": 0.4667, + "step": 45690 + }, + { + "epoch": 8.31, + "learning_rate": 5.077721150345832e-07, + "loss": 0.3392, + "step": 45691 + }, + { + "epoch": 8.31, + "learning_rate": 5.07717510010921e-07, + "loss": 1.4988, + "step": 45692 + }, + { + "epoch": 8.31, + "learning_rate": 5.076629049872589e-07, + "loss": 0.0194, + "step": 45693 + }, + { + "epoch": 8.31, + "learning_rate": 5.076082999635966e-07, + "loss": 1.5355, + "step": 45694 + }, + { + "epoch": 8.31, + "learning_rate": 5.075536949399345e-07, + "loss": 0.0039, + "step": 45695 + }, + { + "epoch": 8.31, + "learning_rate": 5.074990899162723e-07, + "loss": 1.1022, + "step": 45696 + }, + { + "epoch": 8.31, + "learning_rate": 5.074444848926101e-07, + "loss": 0.0942, + "step": 45697 + }, + { + "epoch": 8.31, + "learning_rate": 5.073898798689479e-07, + "loss": 0.9952, + "step": 45698 + }, + { + "epoch": 8.31, + "learning_rate": 5.073352748452858e-07, + "loss": 0.5291, + "step": 45699 + }, + { + "epoch": 8.31, + "learning_rate": 5.072806698216236e-07, + "loss": 0.1063, + "step": 45700 + }, + { + "epoch": 8.31, + "learning_rate": 5.072260647979615e-07, + "loss": 2.7958, + "step": 45701 + }, + { + "epoch": 8.31, + "learning_rate": 5.071714597742993e-07, + "loss": 0.3333, + "step": 45702 + }, + { + "epoch": 8.31, + "learning_rate": 5.07116854750637e-07, + "loss": 1.7646, + "step": 45703 + }, + { + "epoch": 8.31, + "learning_rate": 5.070622497269749e-07, + "loss": 1.2603, + "step": 45704 + }, + { + "epoch": 8.31, + "learning_rate": 5.070076447033127e-07, + "loss": 1.9644, + "step": 45705 + }, + { + "epoch": 8.31, + "learning_rate": 5.069530396796505e-07, + "loss": 0.6848, + "step": 45706 + }, + { + "epoch": 8.31, + "learning_rate": 5.068984346559884e-07, + "loss": 1.1064, + "step": 45707 + }, + { + "epoch": 8.31, + "learning_rate": 5.068438296323262e-07, + "loss": 1.0761, + "step": 45708 + }, + { + "epoch": 8.31, + "learning_rate": 5.06789224608664e-07, + "loss": 0.1161, + "step": 45709 + }, + { + "epoch": 8.31, + "learning_rate": 5.067346195850018e-07, + "loss": 0.0154, + "step": 45710 + }, + { + "epoch": 8.31, + "learning_rate": 5.066800145613396e-07, + "loss": 0.87, + "step": 45711 + }, + { + "epoch": 8.31, + "learning_rate": 5.066254095376775e-07, + "loss": 0.0099, + "step": 45712 + }, + { + "epoch": 8.31, + "learning_rate": 5.065708045140153e-07, + "loss": 0.0128, + "step": 45713 + }, + { + "epoch": 8.31, + "learning_rate": 5.065161994903531e-07, + "loss": 0.1542, + "step": 45714 + }, + { + "epoch": 8.31, + "learning_rate": 5.06461594466691e-07, + "loss": 1.7718, + "step": 45715 + }, + { + "epoch": 8.31, + "learning_rate": 5.064069894430288e-07, + "loss": 2.7826, + "step": 45716 + }, + { + "epoch": 8.31, + "learning_rate": 5.063523844193665e-07, + "loss": 0.0688, + "step": 45717 + }, + { + "epoch": 8.31, + "learning_rate": 5.062977793957045e-07, + "loss": 1.0223, + "step": 45718 + }, + { + "epoch": 8.31, + "learning_rate": 5.062431743720422e-07, + "loss": 0.0054, + "step": 45719 + }, + { + "epoch": 8.31, + "learning_rate": 5.0618856934838e-07, + "loss": 1.5679, + "step": 45720 + }, + { + "epoch": 8.31, + "learning_rate": 5.061339643247179e-07, + "loss": 0.7807, + "step": 45721 + }, + { + "epoch": 8.31, + "learning_rate": 5.060793593010557e-07, + "loss": 1.1375, + "step": 45722 + }, + { + "epoch": 8.31, + "learning_rate": 5.060247542773936e-07, + "loss": 1.3708, + "step": 45723 + }, + { + "epoch": 8.31, + "learning_rate": 5.059701492537313e-07, + "loss": 3.4076, + "step": 45724 + }, + { + "epoch": 8.32, + "learning_rate": 5.059155442300692e-07, + "loss": 2.0033, + "step": 45725 + }, + { + "epoch": 8.32, + "learning_rate": 5.05860939206407e-07, + "loss": 0.7452, + "step": 45726 + }, + { + "epoch": 8.32, + "learning_rate": 5.058063341827448e-07, + "loss": 1.3501, + "step": 45727 + }, + { + "epoch": 8.32, + "learning_rate": 5.057517291590826e-07, + "loss": 0.0901, + "step": 45728 + }, + { + "epoch": 8.32, + "learning_rate": 5.056971241354205e-07, + "loss": 0.48, + "step": 45729 + }, + { + "epoch": 8.32, + "learning_rate": 5.056425191117583e-07, + "loss": 0.0183, + "step": 45730 + }, + { + "epoch": 8.32, + "learning_rate": 5.05587914088096e-07, + "loss": 0.0096, + "step": 45731 + }, + { + "epoch": 8.32, + "learning_rate": 5.05533309064434e-07, + "loss": 0.8495, + "step": 45732 + }, + { + "epoch": 8.32, + "learning_rate": 5.054787040407717e-07, + "loss": 0.2901, + "step": 45733 + }, + { + "epoch": 8.32, + "learning_rate": 5.054240990171096e-07, + "loss": 0.2408, + "step": 45734 + }, + { + "epoch": 8.32, + "learning_rate": 5.053694939934474e-07, + "loss": 0.8592, + "step": 45735 + }, + { + "epoch": 8.32, + "learning_rate": 5.053148889697852e-07, + "loss": 0.356, + "step": 45736 + }, + { + "epoch": 8.32, + "learning_rate": 5.052602839461231e-07, + "loss": 0.1866, + "step": 45737 + }, + { + "epoch": 8.32, + "learning_rate": 5.052056789224609e-07, + "loss": 0.0064, + "step": 45738 + }, + { + "epoch": 8.32, + "learning_rate": 5.051510738987988e-07, + "loss": 0.0928, + "step": 45739 + }, + { + "epoch": 8.32, + "learning_rate": 5.050964688751365e-07, + "loss": 0.5295, + "step": 45740 + }, + { + "epoch": 8.32, + "learning_rate": 5.050418638514743e-07, + "loss": 0.1423, + "step": 45741 + }, + { + "epoch": 8.32, + "learning_rate": 5.049872588278122e-07, + "loss": 0.356, + "step": 45742 + }, + { + "epoch": 8.32, + "learning_rate": 5.0493265380415e-07, + "loss": 0.0711, + "step": 45743 + }, + { + "epoch": 8.32, + "learning_rate": 5.048780487804878e-07, + "loss": 0.4444, + "step": 45744 + }, + { + "epoch": 8.32, + "learning_rate": 5.048234437568257e-07, + "loss": 0.0379, + "step": 45745 + }, + { + "epoch": 8.32, + "learning_rate": 5.047688387331635e-07, + "loss": 0.0258, + "step": 45746 + }, + { + "epoch": 8.32, + "learning_rate": 5.047142337095012e-07, + "loss": 0.1047, + "step": 45747 + }, + { + "epoch": 8.32, + "learning_rate": 5.046596286858391e-07, + "loss": 0.3195, + "step": 45748 + }, + { + "epoch": 8.32, + "learning_rate": 5.046050236621769e-07, + "loss": 3.0465, + "step": 45749 + }, + { + "epoch": 8.32, + "learning_rate": 5.045504186385148e-07, + "loss": 0.007, + "step": 45750 + }, + { + "epoch": 8.32, + "learning_rate": 5.044958136148526e-07, + "loss": 0.9652, + "step": 45751 + }, + { + "epoch": 8.32, + "learning_rate": 5.044412085911904e-07, + "loss": 0.0128, + "step": 45752 + }, + { + "epoch": 8.32, + "learning_rate": 5.043866035675283e-07, + "loss": 0.6698, + "step": 45753 + }, + { + "epoch": 8.32, + "learning_rate": 5.04331998543866e-07, + "loss": 0.2633, + "step": 45754 + }, + { + "epoch": 8.32, + "learning_rate": 5.042773935202038e-07, + "loss": 1.5227, + "step": 45755 + }, + { + "epoch": 8.32, + "learning_rate": 5.042227884965417e-07, + "loss": 0.0056, + "step": 45756 + }, + { + "epoch": 8.32, + "learning_rate": 5.041681834728795e-07, + "loss": 0.7766, + "step": 45757 + }, + { + "epoch": 8.32, + "learning_rate": 5.041135784492173e-07, + "loss": 0.0268, + "step": 45758 + }, + { + "epoch": 8.32, + "learning_rate": 5.040589734255552e-07, + "loss": 1.3766, + "step": 45759 + }, + { + "epoch": 8.32, + "learning_rate": 5.04004368401893e-07, + "loss": 4.3584, + "step": 45760 + }, + { + "epoch": 8.32, + "learning_rate": 5.039497633782308e-07, + "loss": 0.1364, + "step": 45761 + }, + { + "epoch": 8.32, + "learning_rate": 5.038951583545687e-07, + "loss": 0.0637, + "step": 45762 + }, + { + "epoch": 8.32, + "learning_rate": 5.038405533309064e-07, + "loss": 0.0235, + "step": 45763 + }, + { + "epoch": 8.32, + "learning_rate": 5.037859483072443e-07, + "loss": 0.8577, + "step": 45764 + }, + { + "epoch": 8.32, + "learning_rate": 5.037313432835821e-07, + "loss": 0.1907, + "step": 45765 + }, + { + "epoch": 8.32, + "learning_rate": 5.036767382599199e-07, + "loss": 0.5919, + "step": 45766 + }, + { + "epoch": 8.32, + "learning_rate": 5.036221332362578e-07, + "loss": 0.0108, + "step": 45767 + }, + { + "epoch": 8.32, + "learning_rate": 5.035675282125955e-07, + "loss": 0.2623, + "step": 45768 + }, + { + "epoch": 8.32, + "learning_rate": 5.035129231889335e-07, + "loss": 0.013, + "step": 45769 + }, + { + "epoch": 8.32, + "learning_rate": 5.034583181652712e-07, + "loss": 0.5686, + "step": 45770 + }, + { + "epoch": 8.32, + "learning_rate": 5.03403713141609e-07, + "loss": 0.0059, + "step": 45771 + }, + { + "epoch": 8.32, + "learning_rate": 5.033491081179469e-07, + "loss": 1.4674, + "step": 45772 + }, + { + "epoch": 8.32, + "learning_rate": 5.032945030942847e-07, + "loss": 0.355, + "step": 45773 + }, + { + "epoch": 8.32, + "learning_rate": 5.032398980706225e-07, + "loss": 0.005, + "step": 45774 + }, + { + "epoch": 8.32, + "learning_rate": 5.031852930469603e-07, + "loss": 0.2617, + "step": 45775 + }, + { + "epoch": 8.32, + "learning_rate": 5.031306880232982e-07, + "loss": 0.1516, + "step": 45776 + }, + { + "epoch": 8.32, + "learning_rate": 5.030760829996359e-07, + "loss": 0.7643, + "step": 45777 + }, + { + "epoch": 8.32, + "learning_rate": 5.030214779759738e-07, + "loss": 0.1835, + "step": 45778 + }, + { + "epoch": 8.32, + "learning_rate": 5.029668729523116e-07, + "loss": 1.8904, + "step": 45779 + }, + { + "epoch": 8.33, + "learning_rate": 5.029122679286495e-07, + "loss": 0.2258, + "step": 45780 + }, + { + "epoch": 8.33, + "learning_rate": 5.028576629049873e-07, + "loss": 0.0475, + "step": 45781 + }, + { + "epoch": 8.33, + "learning_rate": 5.028030578813251e-07, + "loss": 0.4183, + "step": 45782 + }, + { + "epoch": 8.33, + "learning_rate": 5.02748452857663e-07, + "loss": 1.3676, + "step": 45783 + }, + { + "epoch": 8.33, + "learning_rate": 5.026938478340007e-07, + "loss": 1.8933, + "step": 45784 + }, + { + "epoch": 8.33, + "learning_rate": 5.026392428103385e-07, + "loss": 0.1264, + "step": 45785 + }, + { + "epoch": 8.33, + "learning_rate": 5.025846377866764e-07, + "loss": 0.1691, + "step": 45786 + }, + { + "epoch": 8.33, + "learning_rate": 5.025300327630142e-07, + "loss": 0.0098, + "step": 45787 + }, + { + "epoch": 8.33, + "learning_rate": 5.02475427739352e-07, + "loss": 0.2451, + "step": 45788 + }, + { + "epoch": 8.33, + "learning_rate": 5.024208227156899e-07, + "loss": 0.0248, + "step": 45789 + }, + { + "epoch": 8.33, + "learning_rate": 5.023662176920277e-07, + "loss": 1.8714, + "step": 45790 + }, + { + "epoch": 8.33, + "learning_rate": 5.023116126683655e-07, + "loss": 0.6396, + "step": 45791 + }, + { + "epoch": 8.33, + "learning_rate": 5.022570076447033e-07, + "loss": 0.009, + "step": 45792 + }, + { + "epoch": 8.33, + "learning_rate": 5.022024026210411e-07, + "loss": 0.8254, + "step": 45793 + }, + { + "epoch": 8.33, + "learning_rate": 5.02147797597379e-07, + "loss": 0.0462, + "step": 45794 + }, + { + "epoch": 8.33, + "learning_rate": 5.020931925737168e-07, + "loss": 2.0623, + "step": 45795 + }, + { + "epoch": 8.33, + "learning_rate": 5.020385875500546e-07, + "loss": 0.1561, + "step": 45796 + }, + { + "epoch": 8.33, + "learning_rate": 5.019839825263925e-07, + "loss": 0.2438, + "step": 45797 + }, + { + "epoch": 8.33, + "learning_rate": 5.019293775027302e-07, + "loss": 0.0215, + "step": 45798 + }, + { + "epoch": 8.33, + "learning_rate": 5.01874772479068e-07, + "loss": 0.0041, + "step": 45799 + }, + { + "epoch": 8.33, + "learning_rate": 5.018201674554059e-07, + "loss": 0.3078, + "step": 45800 + }, + { + "epoch": 8.33, + "learning_rate": 5.017655624317437e-07, + "loss": 0.192, + "step": 45801 + }, + { + "epoch": 8.33, + "learning_rate": 5.017109574080816e-07, + "loss": 3.4319, + "step": 45802 + }, + { + "epoch": 8.33, + "learning_rate": 5.016563523844194e-07, + "loss": 1.3302, + "step": 45803 + }, + { + "epoch": 8.33, + "learning_rate": 5.016017473607572e-07, + "loss": 0.0151, + "step": 45804 + }, + { + "epoch": 8.33, + "learning_rate": 5.01547142337095e-07, + "loss": 0.1993, + "step": 45805 + }, + { + "epoch": 8.33, + "learning_rate": 5.014925373134329e-07, + "loss": 0.0616, + "step": 45806 + }, + { + "epoch": 8.33, + "learning_rate": 5.014379322897706e-07, + "loss": 0.6262, + "step": 45807 + }, + { + "epoch": 8.33, + "learning_rate": 5.013833272661085e-07, + "loss": 0.371, + "step": 45808 + }, + { + "epoch": 8.33, + "learning_rate": 5.013287222424463e-07, + "loss": 0.1233, + "step": 45809 + }, + { + "epoch": 8.33, + "learning_rate": 5.012741172187842e-07, + "loss": 0.6822, + "step": 45810 + }, + { + "epoch": 8.33, + "learning_rate": 5.01219512195122e-07, + "loss": 0.0036, + "step": 45811 + }, + { + "epoch": 8.33, + "learning_rate": 5.011649071714597e-07, + "loss": 0.1196, + "step": 45812 + }, + { + "epoch": 8.33, + "learning_rate": 5.011103021477977e-07, + "loss": 0.0745, + "step": 45813 + }, + { + "epoch": 8.33, + "learning_rate": 5.010556971241354e-07, + "loss": 1.2464, + "step": 45814 + }, + { + "epoch": 8.33, + "learning_rate": 5.010010921004732e-07, + "loss": 0.8298, + "step": 45815 + }, + { + "epoch": 8.33, + "learning_rate": 5.009464870768111e-07, + "loss": 0.1291, + "step": 45816 + }, + { + "epoch": 8.33, + "learning_rate": 5.008918820531489e-07, + "loss": 0.371, + "step": 45817 + }, + { + "epoch": 8.33, + "learning_rate": 5.008372770294868e-07, + "loss": 0.1246, + "step": 45818 + }, + { + "epoch": 8.33, + "learning_rate": 5.007826720058245e-07, + "loss": 0.0114, + "step": 45819 + }, + { + "epoch": 8.33, + "learning_rate": 5.007280669821624e-07, + "loss": 2.0097, + "step": 45820 + }, + { + "epoch": 8.33, + "learning_rate": 5.006734619585002e-07, + "loss": 0.9158, + "step": 45821 + }, + { + "epoch": 8.33, + "learning_rate": 5.00618856934838e-07, + "loss": 0.054, + "step": 45822 + }, + { + "epoch": 8.33, + "learning_rate": 5.005642519111758e-07, + "loss": 0.053, + "step": 45823 + }, + { + "epoch": 8.33, + "learning_rate": 5.005096468875137e-07, + "loss": 0.2073, + "step": 45824 + }, + { + "epoch": 8.33, + "learning_rate": 5.004550418638515e-07, + "loss": 0.0204, + "step": 45825 + }, + { + "epoch": 8.33, + "learning_rate": 5.004004368401893e-07, + "loss": 0.226, + "step": 45826 + }, + { + "epoch": 8.33, + "learning_rate": 5.003458318165272e-07, + "loss": 1.4615, + "step": 45827 + }, + { + "epoch": 8.33, + "learning_rate": 5.002912267928649e-07, + "loss": 0.6401, + "step": 45828 + }, + { + "epoch": 8.33, + "learning_rate": 5.002366217692028e-07, + "loss": 0.3524, + "step": 45829 + }, + { + "epoch": 8.33, + "learning_rate": 5.001820167455406e-07, + "loss": 0.2256, + "step": 45830 + }, + { + "epoch": 8.33, + "learning_rate": 5.001274117218784e-07, + "loss": 0.2438, + "step": 45831 + }, + { + "epoch": 8.33, + "learning_rate": 5.000728066982163e-07, + "loss": 0.018, + "step": 45832 + }, + { + "epoch": 8.33, + "learning_rate": 5.000182016745541e-07, + "loss": 0.0112, + "step": 45833 + }, + { + "epoch": 8.33, + "learning_rate": 4.999635966508919e-07, + "loss": 0.0344, + "step": 45834 + }, + { + "epoch": 8.34, + "learning_rate": 4.999089916272297e-07, + "loss": 0.006, + "step": 45835 + }, + { + "epoch": 8.34, + "learning_rate": 4.998543866035675e-07, + "loss": 0.1454, + "step": 45836 + }, + { + "epoch": 8.34, + "learning_rate": 4.997997815799053e-07, + "loss": 2.6459, + "step": 45837 + }, + { + "epoch": 8.34, + "learning_rate": 4.997451765562432e-07, + "loss": 0.013, + "step": 45838 + }, + { + "epoch": 8.34, + "learning_rate": 4.99690571532581e-07, + "loss": 0.1045, + "step": 45839 + }, + { + "epoch": 8.34, + "learning_rate": 4.996359665089189e-07, + "loss": 0.014, + "step": 45840 + }, + { + "epoch": 8.34, + "learning_rate": 4.995813614852567e-07, + "loss": 0.1382, + "step": 45841 + }, + { + "epoch": 8.34, + "learning_rate": 4.995267564615944e-07, + "loss": 3.8709, + "step": 45842 + }, + { + "epoch": 8.34, + "learning_rate": 4.994721514379323e-07, + "loss": 1.3083, + "step": 45843 + }, + { + "epoch": 8.34, + "learning_rate": 4.994175464142701e-07, + "loss": 0.471, + "step": 45844 + }, + { + "epoch": 8.34, + "learning_rate": 4.993629413906079e-07, + "loss": 0.9607, + "step": 45845 + }, + { + "epoch": 8.34, + "learning_rate": 4.993083363669458e-07, + "loss": 0.5809, + "step": 45846 + }, + { + "epoch": 8.34, + "learning_rate": 4.992537313432836e-07, + "loss": 0.011, + "step": 45847 + }, + { + "epoch": 8.34, + "learning_rate": 4.991991263196215e-07, + "loss": 0.4367, + "step": 45848 + }, + { + "epoch": 8.34, + "learning_rate": 4.991445212959592e-07, + "loss": 0.6672, + "step": 45849 + }, + { + "epoch": 8.34, + "learning_rate": 4.990899162722971e-07, + "loss": 0.0211, + "step": 45850 + }, + { + "epoch": 8.34, + "learning_rate": 4.990353112486349e-07, + "loss": 0.0088, + "step": 45851 + }, + { + "epoch": 8.34, + "learning_rate": 4.989807062249727e-07, + "loss": 0.007, + "step": 45852 + }, + { + "epoch": 8.34, + "learning_rate": 4.989261012013105e-07, + "loss": 0.0724, + "step": 45853 + }, + { + "epoch": 8.34, + "learning_rate": 4.988714961776484e-07, + "loss": 0.0372, + "step": 45854 + }, + { + "epoch": 8.34, + "learning_rate": 4.988168911539862e-07, + "loss": 0.1541, + "step": 45855 + }, + { + "epoch": 8.34, + "learning_rate": 4.987622861303239e-07, + "loss": 0.3765, + "step": 45856 + }, + { + "epoch": 8.34, + "learning_rate": 4.987076811066619e-07, + "loss": 0.1122, + "step": 45857 + }, + { + "epoch": 8.34, + "learning_rate": 4.986530760829996e-07, + "loss": 1.3025, + "step": 45858 + }, + { + "epoch": 8.34, + "learning_rate": 4.985984710593375e-07, + "loss": 1.5478, + "step": 45859 + }, + { + "epoch": 8.34, + "learning_rate": 4.985438660356753e-07, + "loss": 0.1236, + "step": 45860 + }, + { + "epoch": 8.34, + "learning_rate": 4.984892610120131e-07, + "loss": 0.7142, + "step": 45861 + }, + { + "epoch": 8.34, + "learning_rate": 4.98434655988351e-07, + "loss": 0.0042, + "step": 45862 + }, + { + "epoch": 8.34, + "learning_rate": 4.983800509646887e-07, + "loss": 0.6575, + "step": 45863 + }, + { + "epoch": 8.34, + "learning_rate": 4.983254459410266e-07, + "loss": 0.1016, + "step": 45864 + }, + { + "epoch": 8.34, + "learning_rate": 4.982708409173644e-07, + "loss": 0.1682, + "step": 45865 + }, + { + "epoch": 8.34, + "learning_rate": 4.982162358937022e-07, + "loss": 0.0244, + "step": 45866 + }, + { + "epoch": 8.34, + "learning_rate": 4.9816163087004e-07, + "loss": 0.0045, + "step": 45867 + }, + { + "epoch": 8.34, + "learning_rate": 4.981070258463779e-07, + "loss": 0.8371, + "step": 45868 + }, + { + "epoch": 8.34, + "learning_rate": 4.980524208227157e-07, + "loss": 0.034, + "step": 45869 + }, + { + "epoch": 8.34, + "learning_rate": 4.979978157990536e-07, + "loss": 0.2227, + "step": 45870 + }, + { + "epoch": 8.34, + "learning_rate": 4.979432107753914e-07, + "loss": 0.2378, + "step": 45871 + }, + { + "epoch": 8.34, + "learning_rate": 4.978886057517291e-07, + "loss": 0.0598, + "step": 45872 + }, + { + "epoch": 8.34, + "learning_rate": 4.97834000728067e-07, + "loss": 0.0106, + "step": 45873 + }, + { + "epoch": 8.34, + "learning_rate": 4.977793957044048e-07, + "loss": 2.5048, + "step": 45874 + }, + { + "epoch": 8.34, + "learning_rate": 4.977247906807426e-07, + "loss": 0.601, + "step": 45875 + }, + { + "epoch": 8.34, + "learning_rate": 4.976701856570805e-07, + "loss": 0.4249, + "step": 45876 + }, + { + "epoch": 8.34, + "learning_rate": 4.976155806334183e-07, + "loss": 0.0328, + "step": 45877 + }, + { + "epoch": 8.34, + "learning_rate": 4.975609756097562e-07, + "loss": 2.3854, + "step": 45878 + }, + { + "epoch": 8.34, + "learning_rate": 4.975063705860939e-07, + "loss": 0.0471, + "step": 45879 + }, + { + "epoch": 8.34, + "learning_rate": 4.974517655624317e-07, + "loss": 1.7061, + "step": 45880 + }, + { + "epoch": 8.34, + "learning_rate": 4.973971605387696e-07, + "loss": 0.0813, + "step": 45881 + }, + { + "epoch": 8.34, + "learning_rate": 4.973425555151074e-07, + "loss": 0.0291, + "step": 45882 + }, + { + "epoch": 8.34, + "learning_rate": 4.972879504914452e-07, + "loss": 2.0639, + "step": 45883 + }, + { + "epoch": 8.34, + "learning_rate": 4.972333454677831e-07, + "loss": 1.9881, + "step": 45884 + }, + { + "epoch": 8.34, + "learning_rate": 4.971787404441209e-07, + "loss": 1.5298, + "step": 45885 + }, + { + "epoch": 8.34, + "learning_rate": 4.971241354204586e-07, + "loss": 0.0261, + "step": 45886 + }, + { + "epoch": 8.34, + "learning_rate": 4.970695303967966e-07, + "loss": 1.1289, + "step": 45887 + }, + { + "epoch": 8.34, + "learning_rate": 4.970149253731343e-07, + "loss": 0.0317, + "step": 45888 + }, + { + "epoch": 8.34, + "learning_rate": 4.969603203494722e-07, + "loss": 0.2248, + "step": 45889 + }, + { + "epoch": 8.35, + "learning_rate": 4.9690571532581e-07, + "loss": 4.2283, + "step": 45890 + }, + { + "epoch": 8.35, + "learning_rate": 4.968511103021478e-07, + "loss": 0.0038, + "step": 45891 + }, + { + "epoch": 8.35, + "learning_rate": 4.967965052784857e-07, + "loss": 0.0081, + "step": 45892 + }, + { + "epoch": 8.35, + "learning_rate": 4.967419002548234e-07, + "loss": 0.0589, + "step": 45893 + }, + { + "epoch": 8.35, + "learning_rate": 4.966872952311613e-07, + "loss": 0.0455, + "step": 45894 + }, + { + "epoch": 8.35, + "learning_rate": 4.966326902074991e-07, + "loss": 0.0062, + "step": 45895 + }, + { + "epoch": 8.35, + "learning_rate": 4.965780851838369e-07, + "loss": 0.0373, + "step": 45896 + }, + { + "epoch": 8.35, + "learning_rate": 4.965234801601748e-07, + "loss": 0.0779, + "step": 45897 + }, + { + "epoch": 8.35, + "learning_rate": 4.964688751365126e-07, + "loss": 0.0498, + "step": 45898 + }, + { + "epoch": 8.35, + "learning_rate": 4.964142701128504e-07, + "loss": 2.5989, + "step": 45899 + }, + { + "epoch": 8.35, + "learning_rate": 4.963596650891882e-07, + "loss": 0.008, + "step": 45900 + }, + { + "epoch": 8.35, + "learning_rate": 4.963050600655261e-07, + "loss": 0.1273, + "step": 45901 + }, + { + "epoch": 8.35, + "learning_rate": 4.962504550418638e-07, + "loss": 0.0069, + "step": 45902 + }, + { + "epoch": 8.35, + "learning_rate": 4.961958500182017e-07, + "loss": 1.6284, + "step": 45903 + }, + { + "epoch": 8.35, + "learning_rate": 4.961412449945395e-07, + "loss": 0.0081, + "step": 45904 + }, + { + "epoch": 8.35, + "learning_rate": 4.960866399708773e-07, + "loss": 0.0043, + "step": 45905 + }, + { + "epoch": 8.35, + "learning_rate": 4.960320349472152e-07, + "loss": 0.6286, + "step": 45906 + }, + { + "epoch": 8.35, + "learning_rate": 4.959774299235529e-07, + "loss": 0.9571, + "step": 45907 + }, + { + "epoch": 8.35, + "learning_rate": 4.959228248998909e-07, + "loss": 0.0782, + "step": 45908 + }, + { + "epoch": 8.35, + "learning_rate": 4.958682198762286e-07, + "loss": 1.0447, + "step": 45909 + }, + { + "epoch": 8.35, + "learning_rate": 4.958136148525664e-07, + "loss": 2.5656, + "step": 45910 + }, + { + "epoch": 8.35, + "learning_rate": 4.957590098289043e-07, + "loss": 0.6426, + "step": 45911 + }, + { + "epoch": 8.35, + "learning_rate": 4.957044048052421e-07, + "loss": 0.4877, + "step": 45912 + }, + { + "epoch": 8.35, + "learning_rate": 4.956497997815799e-07, + "loss": 0.007, + "step": 45913 + }, + { + "epoch": 8.35, + "learning_rate": 4.955951947579178e-07, + "loss": 0.6166, + "step": 45914 + }, + { + "epoch": 8.35, + "learning_rate": 4.955405897342556e-07, + "loss": 0.4254, + "step": 45915 + }, + { + "epoch": 8.35, + "learning_rate": 4.954859847105933e-07, + "loss": 1.2713, + "step": 45916 + }, + { + "epoch": 8.35, + "learning_rate": 4.954313796869312e-07, + "loss": 0.0553, + "step": 45917 + }, + { + "epoch": 8.35, + "learning_rate": 4.95376774663269e-07, + "loss": 0.6575, + "step": 45918 + }, + { + "epoch": 8.35, + "learning_rate": 4.953221696396069e-07, + "loss": 0.0068, + "step": 45919 + }, + { + "epoch": 8.35, + "learning_rate": 4.952675646159447e-07, + "loss": 0.5895, + "step": 45920 + }, + { + "epoch": 8.35, + "learning_rate": 4.952129595922825e-07, + "loss": 0.0698, + "step": 45921 + }, + { + "epoch": 8.35, + "learning_rate": 4.951583545686204e-07, + "loss": 1.3492, + "step": 45922 + }, + { + "epoch": 8.35, + "learning_rate": 4.951037495449581e-07, + "loss": 3.5674, + "step": 45923 + }, + { + "epoch": 8.35, + "learning_rate": 4.950491445212959e-07, + "loss": 0.0132, + "step": 45924 + }, + { + "epoch": 8.35, + "learning_rate": 4.949945394976338e-07, + "loss": 0.4076, + "step": 45925 + }, + { + "epoch": 8.35, + "learning_rate": 4.949399344739716e-07, + "loss": 1.3726, + "step": 45926 + }, + { + "epoch": 8.35, + "learning_rate": 4.948853294503095e-07, + "loss": 0.0084, + "step": 45927 + }, + { + "epoch": 8.35, + "learning_rate": 4.948307244266473e-07, + "loss": 0.0432, + "step": 45928 + }, + { + "epoch": 8.35, + "learning_rate": 4.947761194029851e-07, + "loss": 1.0574, + "step": 45929 + }, + { + "epoch": 8.35, + "learning_rate": 4.947215143793229e-07, + "loss": 1.2389, + "step": 45930 + }, + { + "epoch": 8.35, + "learning_rate": 4.946669093556608e-07, + "loss": 0.0672, + "step": 45931 + }, + { + "epoch": 8.35, + "learning_rate": 4.946123043319985e-07, + "loss": 0.4626, + "step": 45932 + }, + { + "epoch": 8.35, + "learning_rate": 4.945576993083364e-07, + "loss": 0.103, + "step": 45933 + }, + { + "epoch": 8.35, + "learning_rate": 4.945030942846742e-07, + "loss": 0.0184, + "step": 45934 + }, + { + "epoch": 8.35, + "learning_rate": 4.94448489261012e-07, + "loss": 2.7942, + "step": 45935 + }, + { + "epoch": 8.35, + "learning_rate": 4.943938842373499e-07, + "loss": 0.6999, + "step": 45936 + }, + { + "epoch": 8.35, + "learning_rate": 4.943392792136876e-07, + "loss": 0.3561, + "step": 45937 + }, + { + "epoch": 8.35, + "learning_rate": 4.942846741900256e-07, + "loss": 0.014, + "step": 45938 + }, + { + "epoch": 8.35, + "learning_rate": 4.942300691663633e-07, + "loss": 0.0623, + "step": 45939 + }, + { + "epoch": 8.35, + "learning_rate": 4.941754641427011e-07, + "loss": 0.7136, + "step": 45940 + }, + { + "epoch": 8.35, + "learning_rate": 4.94120859119039e-07, + "loss": 1.0912, + "step": 45941 + }, + { + "epoch": 8.35, + "learning_rate": 4.940662540953768e-07, + "loss": 0.1728, + "step": 45942 + }, + { + "epoch": 8.35, + "learning_rate": 4.940116490717146e-07, + "loss": 0.0386, + "step": 45943 + }, + { + "epoch": 8.35, + "learning_rate": 4.939570440480524e-07, + "loss": 0.7693, + "step": 45944 + }, + { + "epoch": 8.36, + "learning_rate": 4.939024390243903e-07, + "loss": 0.1051, + "step": 45945 + }, + { + "epoch": 8.36, + "learning_rate": 4.93847834000728e-07, + "loss": 0.6342, + "step": 45946 + }, + { + "epoch": 8.36, + "learning_rate": 4.937932289770659e-07, + "loss": 0.9481, + "step": 45947 + }, + { + "epoch": 8.36, + "learning_rate": 4.937386239534037e-07, + "loss": 1.7606, + "step": 45948 + }, + { + "epoch": 8.36, + "learning_rate": 4.936840189297416e-07, + "loss": 0.5308, + "step": 45949 + }, + { + "epoch": 8.36, + "learning_rate": 4.936294139060794e-07, + "loss": 1.5084, + "step": 45950 + }, + { + "epoch": 8.36, + "learning_rate": 4.935748088824172e-07, + "loss": 0.3834, + "step": 45951 + }, + { + "epoch": 8.36, + "learning_rate": 4.935202038587551e-07, + "loss": 0.0037, + "step": 45952 + }, + { + "epoch": 8.36, + "learning_rate": 4.934655988350928e-07, + "loss": 0.1762, + "step": 45953 + }, + { + "epoch": 8.36, + "learning_rate": 4.934109938114306e-07, + "loss": 0.053, + "step": 45954 + }, + { + "epoch": 8.36, + "learning_rate": 4.933563887877685e-07, + "loss": 0.3151, + "step": 45955 + }, + { + "epoch": 8.36, + "learning_rate": 4.933017837641063e-07, + "loss": 0.9365, + "step": 45956 + }, + { + "epoch": 8.36, + "learning_rate": 4.932471787404442e-07, + "loss": 0.1193, + "step": 45957 + }, + { + "epoch": 8.36, + "learning_rate": 4.93192573716782e-07, + "loss": 1.585, + "step": 45958 + }, + { + "epoch": 8.36, + "learning_rate": 4.931379686931198e-07, + "loss": 0.4428, + "step": 45959 + }, + { + "epoch": 8.36, + "learning_rate": 4.930833636694576e-07, + "loss": 0.1277, + "step": 45960 + }, + { + "epoch": 8.36, + "learning_rate": 4.930287586457954e-07, + "loss": 1.4544, + "step": 45961 + }, + { + "epoch": 8.36, + "learning_rate": 4.929741536221332e-07, + "loss": 0.0927, + "step": 45962 + }, + { + "epoch": 8.36, + "learning_rate": 4.929195485984711e-07, + "loss": 0.5304, + "step": 45963 + }, + { + "epoch": 8.36, + "learning_rate": 4.928649435748089e-07, + "loss": 0.4106, + "step": 45964 + }, + { + "epoch": 8.36, + "learning_rate": 4.928103385511467e-07, + "loss": 0.5473, + "step": 45965 + }, + { + "epoch": 8.36, + "learning_rate": 4.927557335274846e-07, + "loss": 0.078, + "step": 45966 + }, + { + "epoch": 8.36, + "learning_rate": 4.927011285038223e-07, + "loss": 0.0972, + "step": 45967 + }, + { + "epoch": 8.36, + "learning_rate": 4.926465234801602e-07, + "loss": 0.4101, + "step": 45968 + }, + { + "epoch": 8.36, + "learning_rate": 4.92591918456498e-07, + "loss": 0.0625, + "step": 45969 + }, + { + "epoch": 8.36, + "learning_rate": 4.925373134328358e-07, + "loss": 0.4443, + "step": 45970 + }, + { + "epoch": 8.36, + "learning_rate": 4.924827084091737e-07, + "loss": 0.0726, + "step": 45971 + }, + { + "epoch": 8.36, + "learning_rate": 4.924281033855115e-07, + "loss": 0.3363, + "step": 45972 + }, + { + "epoch": 8.36, + "learning_rate": 4.923734983618493e-07, + "loss": 1.003, + "step": 45973 + }, + { + "epoch": 8.36, + "learning_rate": 4.923188933381871e-07, + "loss": 1.0639, + "step": 45974 + }, + { + "epoch": 8.36, + "learning_rate": 4.92264288314525e-07, + "loss": 0.2963, + "step": 45975 + }, + { + "epoch": 8.36, + "learning_rate": 4.922096832908627e-07, + "loss": 0.2424, + "step": 45976 + }, + { + "epoch": 8.36, + "learning_rate": 4.921550782672006e-07, + "loss": 0.0325, + "step": 45977 + }, + { + "epoch": 8.36, + "learning_rate": 4.921004732435384e-07, + "loss": 0.0253, + "step": 45978 + }, + { + "epoch": 8.36, + "learning_rate": 4.920458682198763e-07, + "loss": 2.6426, + "step": 45979 + }, + { + "epoch": 8.36, + "learning_rate": 4.919912631962141e-07, + "loss": 0.0061, + "step": 45980 + }, + { + "epoch": 8.36, + "learning_rate": 4.919366581725518e-07, + "loss": 0.9638, + "step": 45981 + }, + { + "epoch": 8.36, + "learning_rate": 4.918820531488898e-07, + "loss": 0.1015, + "step": 45982 + }, + { + "epoch": 8.36, + "learning_rate": 4.918274481252275e-07, + "loss": 0.0672, + "step": 45983 + }, + { + "epoch": 8.36, + "learning_rate": 4.917728431015653e-07, + "loss": 1.4493, + "step": 45984 + }, + { + "epoch": 8.36, + "learning_rate": 4.917182380779032e-07, + "loss": 0.0986, + "step": 45985 + }, + { + "epoch": 8.36, + "learning_rate": 4.91663633054241e-07, + "loss": 0.0355, + "step": 45986 + }, + { + "epoch": 8.36, + "learning_rate": 4.916090280305789e-07, + "loss": 1.6954, + "step": 45987 + }, + { + "epoch": 8.36, + "learning_rate": 4.915544230069166e-07, + "loss": 0.0129, + "step": 45988 + }, + { + "epoch": 8.36, + "learning_rate": 4.914998179832545e-07, + "loss": 0.6845, + "step": 45989 + }, + { + "epoch": 8.36, + "learning_rate": 4.914452129595923e-07, + "loss": 0.0261, + "step": 45990 + }, + { + "epoch": 8.36, + "learning_rate": 4.913906079359301e-07, + "loss": 0.7312, + "step": 45991 + }, + { + "epoch": 8.36, + "learning_rate": 4.913360029122679e-07, + "loss": 0.0647, + "step": 45992 + }, + { + "epoch": 8.36, + "learning_rate": 4.912813978886058e-07, + "loss": 1.0529, + "step": 45993 + }, + { + "epoch": 8.36, + "learning_rate": 4.912267928649436e-07, + "loss": 1.8801, + "step": 45994 + }, + { + "epoch": 8.36, + "learning_rate": 4.911721878412815e-07, + "loss": 1.8659, + "step": 45995 + }, + { + "epoch": 8.36, + "learning_rate": 4.911175828176193e-07, + "loss": 0.1116, + "step": 45996 + }, + { + "epoch": 8.36, + "learning_rate": 4.91062977793957e-07, + "loss": 1.5402, + "step": 45997 + }, + { + "epoch": 8.36, + "learning_rate": 4.910083727702949e-07, + "loss": 0.1281, + "step": 45998 + }, + { + "epoch": 8.36, + "learning_rate": 4.909537677466327e-07, + "loss": 0.806, + "step": 45999 + }, + { + "epoch": 8.37, + "learning_rate": 4.908991627229705e-07, + "loss": 0.0603, + "step": 46000 + }, + { + "epoch": 8.37, + "learning_rate": 4.908445576993084e-07, + "loss": 0.0711, + "step": 46001 + }, + { + "epoch": 8.37, + "learning_rate": 4.907899526756462e-07, + "loss": 0.2801, + "step": 46002 + }, + { + "epoch": 8.37, + "learning_rate": 4.90735347651984e-07, + "loss": 1.5274, + "step": 46003 + }, + { + "epoch": 8.37, + "learning_rate": 4.906807426283218e-07, + "loss": 0.5557, + "step": 46004 + }, + { + "epoch": 8.37, + "learning_rate": 4.906261376046596e-07, + "loss": 0.2443, + "step": 46005 + }, + { + "epoch": 8.37, + "learning_rate": 4.905715325809975e-07, + "loss": 1.8637, + "step": 46006 + }, + { + "epoch": 8.37, + "learning_rate": 4.905169275573353e-07, + "loss": 1.6962, + "step": 46007 + }, + { + "epoch": 8.37, + "learning_rate": 4.90462322533673e-07, + "loss": 0.0538, + "step": 46008 + }, + { + "epoch": 8.37, + "learning_rate": 4.90407717510011e-07, + "loss": 4.9432, + "step": 46009 + }, + { + "epoch": 8.37, + "learning_rate": 4.903531124863488e-07, + "loss": 0.2248, + "step": 46010 + }, + { + "epoch": 8.37, + "learning_rate": 4.902985074626865e-07, + "loss": 1.0614, + "step": 46011 + }, + { + "epoch": 8.37, + "learning_rate": 4.902439024390244e-07, + "loss": 0.1013, + "step": 46012 + }, + { + "epoch": 8.37, + "learning_rate": 4.901892974153622e-07, + "loss": 1.6467, + "step": 46013 + }, + { + "epoch": 8.37, + "learning_rate": 4.901346923917e-07, + "loss": 1.111, + "step": 46014 + }, + { + "epoch": 8.37, + "learning_rate": 4.900800873680379e-07, + "loss": 1.2979, + "step": 46015 + }, + { + "epoch": 8.37, + "learning_rate": 4.900254823443757e-07, + "loss": 0.0122, + "step": 46016 + }, + { + "epoch": 8.37, + "learning_rate": 4.899708773207136e-07, + "loss": 0.0715, + "step": 46017 + }, + { + "epoch": 8.37, + "learning_rate": 4.899162722970513e-07, + "loss": 1.5803, + "step": 46018 + }, + { + "epoch": 8.37, + "learning_rate": 4.898616672733892e-07, + "loss": 0.6513, + "step": 46019 + }, + { + "epoch": 8.37, + "learning_rate": 4.89807062249727e-07, + "loss": 0.1011, + "step": 46020 + }, + { + "epoch": 8.37, + "learning_rate": 4.897524572260648e-07, + "loss": 0.0742, + "step": 46021 + }, + { + "epoch": 8.37, + "learning_rate": 4.896978522024026e-07, + "loss": 0.3518, + "step": 46022 + }, + { + "epoch": 8.37, + "learning_rate": 4.896432471787405e-07, + "loss": 0.0068, + "step": 46023 + }, + { + "epoch": 8.37, + "learning_rate": 4.895886421550783e-07, + "loss": 0.013, + "step": 46024 + }, + { + "epoch": 8.37, + "learning_rate": 4.89534037131416e-07, + "loss": 0.6823, + "step": 46025 + }, + { + "epoch": 8.37, + "learning_rate": 4.89479432107754e-07, + "loss": 1.0948, + "step": 46026 + }, + { + "epoch": 8.37, + "learning_rate": 4.894248270840917e-07, + "loss": 0.1111, + "step": 46027 + }, + { + "epoch": 8.37, + "learning_rate": 4.893702220604296e-07, + "loss": 0.1605, + "step": 46028 + }, + { + "epoch": 8.37, + "learning_rate": 4.893156170367674e-07, + "loss": 1.4178, + "step": 46029 + }, + { + "epoch": 8.37, + "learning_rate": 4.892610120131052e-07, + "loss": 0.0086, + "step": 46030 + }, + { + "epoch": 8.37, + "learning_rate": 4.892064069894431e-07, + "loss": 0.6166, + "step": 46031 + }, + { + "epoch": 8.37, + "learning_rate": 4.891518019657808e-07, + "loss": 0.0125, + "step": 46032 + }, + { + "epoch": 8.37, + "learning_rate": 4.890971969421187e-07, + "loss": 0.0333, + "step": 46033 + }, + { + "epoch": 8.37, + "learning_rate": 4.890425919184565e-07, + "loss": 0.6581, + "step": 46034 + }, + { + "epoch": 8.37, + "learning_rate": 4.889879868947943e-07, + "loss": 0.9301, + "step": 46035 + }, + { + "epoch": 8.37, + "learning_rate": 4.889333818711322e-07, + "loss": 0.0829, + "step": 46036 + }, + { + "epoch": 8.37, + "learning_rate": 4.8887877684747e-07, + "loss": 0.2393, + "step": 46037 + }, + { + "epoch": 8.37, + "learning_rate": 4.888241718238078e-07, + "loss": 0.1417, + "step": 46038 + }, + { + "epoch": 8.37, + "learning_rate": 4.887695668001457e-07, + "loss": 0.4684, + "step": 46039 + }, + { + "epoch": 8.37, + "learning_rate": 4.887149617764835e-07, + "loss": 0.1641, + "step": 46040 + }, + { + "epoch": 8.37, + "learning_rate": 4.886603567528212e-07, + "loss": 0.2432, + "step": 46041 + }, + { + "epoch": 8.37, + "learning_rate": 4.886057517291591e-07, + "loss": 0.0212, + "step": 46042 + }, + { + "epoch": 8.37, + "learning_rate": 4.885511467054969e-07, + "loss": 1.103, + "step": 46043 + }, + { + "epoch": 8.37, + "learning_rate": 4.884965416818347e-07, + "loss": 0.4184, + "step": 46044 + }, + { + "epoch": 8.37, + "learning_rate": 4.884419366581726e-07, + "loss": 0.1165, + "step": 46045 + }, + { + "epoch": 8.37, + "learning_rate": 4.883873316345104e-07, + "loss": 0.8973, + "step": 46046 + }, + { + "epoch": 8.37, + "learning_rate": 4.883327266108483e-07, + "loss": 0.0755, + "step": 46047 + }, + { + "epoch": 8.37, + "learning_rate": 4.88278121587186e-07, + "loss": 1.4803, + "step": 46048 + }, + { + "epoch": 8.37, + "learning_rate": 4.882235165635238e-07, + "loss": 0.5524, + "step": 46049 + }, + { + "epoch": 8.37, + "learning_rate": 4.881689115398617e-07, + "loss": 0.0632, + "step": 46050 + }, + { + "epoch": 8.37, + "learning_rate": 4.881143065161995e-07, + "loss": 0.0188, + "step": 46051 + }, + { + "epoch": 8.37, + "learning_rate": 4.880597014925372e-07, + "loss": 0.1383, + "step": 46052 + }, + { + "epoch": 8.37, + "learning_rate": 4.880050964688752e-07, + "loss": 0.1327, + "step": 46053 + }, + { + "epoch": 8.37, + "learning_rate": 4.87950491445213e-07, + "loss": 2.1131, + "step": 46054 + }, + { + "epoch": 8.38, + "learning_rate": 4.878958864215507e-07, + "loss": 0.1183, + "step": 46055 + }, + { + "epoch": 8.38, + "learning_rate": 4.878412813978886e-07, + "loss": 0.3639, + "step": 46056 + }, + { + "epoch": 8.38, + "learning_rate": 4.877866763742264e-07, + "loss": 0.0721, + "step": 46057 + }, + { + "epoch": 8.38, + "learning_rate": 4.877320713505643e-07, + "loss": 0.385, + "step": 46058 + }, + { + "epoch": 8.38, + "learning_rate": 4.876774663269021e-07, + "loss": 0.0046, + "step": 46059 + }, + { + "epoch": 8.38, + "learning_rate": 4.876228613032399e-07, + "loss": 0.0033, + "step": 46060 + }, + { + "epoch": 8.38, + "learning_rate": 4.875682562795778e-07, + "loss": 0.9464, + "step": 46061 + }, + { + "epoch": 8.38, + "learning_rate": 4.875136512559155e-07, + "loss": 0.2462, + "step": 46062 + }, + { + "epoch": 8.38, + "learning_rate": 4.874590462322534e-07, + "loss": 0.6771, + "step": 46063 + }, + { + "epoch": 8.38, + "learning_rate": 4.874044412085912e-07, + "loss": 0.0027, + "step": 46064 + }, + { + "epoch": 8.38, + "learning_rate": 4.87349836184929e-07, + "loss": 0.068, + "step": 46065 + }, + { + "epoch": 8.38, + "learning_rate": 4.872952311612669e-07, + "loss": 0.9378, + "step": 46066 + }, + { + "epoch": 8.38, + "learning_rate": 4.872406261376047e-07, + "loss": 3.1372, + "step": 46067 + }, + { + "epoch": 8.38, + "learning_rate": 4.871860211139425e-07, + "loss": 0.5403, + "step": 46068 + }, + { + "epoch": 8.38, + "learning_rate": 4.871314160902803e-07, + "loss": 1.7509, + "step": 46069 + }, + { + "epoch": 8.38, + "learning_rate": 4.870768110666182e-07, + "loss": 0.218, + "step": 46070 + }, + { + "epoch": 8.38, + "learning_rate": 4.870222060429559e-07, + "loss": 0.2649, + "step": 46071 + }, + { + "epoch": 8.38, + "learning_rate": 4.869676010192938e-07, + "loss": 0.0299, + "step": 46072 + }, + { + "epoch": 8.38, + "learning_rate": 4.869129959956316e-07, + "loss": 0.185, + "step": 46073 + }, + { + "epoch": 8.38, + "learning_rate": 4.868583909719694e-07, + "loss": 0.1113, + "step": 46074 + }, + { + "epoch": 8.38, + "learning_rate": 4.868037859483073e-07, + "loss": 0.8591, + "step": 46075 + }, + { + "epoch": 8.38, + "learning_rate": 4.86749180924645e-07, + "loss": 1.356, + "step": 46076 + }, + { + "epoch": 8.38, + "learning_rate": 4.86694575900983e-07, + "loss": 1.333, + "step": 46077 + }, + { + "epoch": 8.38, + "learning_rate": 4.866399708773207e-07, + "loss": 0.0163, + "step": 46078 + }, + { + "epoch": 8.38, + "learning_rate": 4.865853658536585e-07, + "loss": 0.1181, + "step": 46079 + }, + { + "epoch": 8.38, + "learning_rate": 4.865307608299964e-07, + "loss": 0.9105, + "step": 46080 + }, + { + "epoch": 8.38, + "learning_rate": 4.864761558063342e-07, + "loss": 0.211, + "step": 46081 + }, + { + "epoch": 8.38, + "learning_rate": 4.86421550782672e-07, + "loss": 0.258, + "step": 46082 + }, + { + "epoch": 8.38, + "learning_rate": 4.863669457590099e-07, + "loss": 0.0693, + "step": 46083 + }, + { + "epoch": 8.38, + "learning_rate": 4.863123407353477e-07, + "loss": 0.1035, + "step": 46084 + }, + { + "epoch": 8.38, + "learning_rate": 4.862577357116855e-07, + "loss": 0.7539, + "step": 46085 + }, + { + "epoch": 8.38, + "learning_rate": 4.862031306880233e-07, + "loss": 0.461, + "step": 46086 + }, + { + "epoch": 8.38, + "learning_rate": 4.861485256643611e-07, + "loss": 0.7384, + "step": 46087 + }, + { + "epoch": 8.38, + "learning_rate": 4.86093920640699e-07, + "loss": 0.0448, + "step": 46088 + }, + { + "epoch": 8.38, + "learning_rate": 4.860393156170367e-07, + "loss": 0.0992, + "step": 46089 + }, + { + "epoch": 8.38, + "learning_rate": 4.859847105933746e-07, + "loss": 0.9061, + "step": 46090 + }, + { + "epoch": 8.38, + "learning_rate": 4.859301055697125e-07, + "loss": 0.2649, + "step": 46091 + }, + { + "epoch": 8.38, + "learning_rate": 4.858755005460502e-07, + "loss": 0.0588, + "step": 46092 + }, + { + "epoch": 8.38, + "learning_rate": 4.85820895522388e-07, + "loss": 0.2198, + "step": 46093 + }, + { + "epoch": 8.38, + "learning_rate": 4.857662904987259e-07, + "loss": 0.2587, + "step": 46094 + }, + { + "epoch": 8.38, + "learning_rate": 4.857116854750637e-07, + "loss": 0.435, + "step": 46095 + }, + { + "epoch": 8.38, + "learning_rate": 4.856570804514015e-07, + "loss": 0.0029, + "step": 46096 + }, + { + "epoch": 8.38, + "learning_rate": 4.856024754277394e-07, + "loss": 0.0063, + "step": 46097 + }, + { + "epoch": 8.38, + "learning_rate": 4.855478704040772e-07, + "loss": 2.8903, + "step": 46098 + }, + { + "epoch": 8.38, + "learning_rate": 4.85493265380415e-07, + "loss": 0.1423, + "step": 46099 + }, + { + "epoch": 8.38, + "learning_rate": 4.854386603567529e-07, + "loss": 0.6815, + "step": 46100 + }, + { + "epoch": 8.38, + "learning_rate": 4.853840553330906e-07, + "loss": 0.0278, + "step": 46101 + }, + { + "epoch": 8.38, + "learning_rate": 4.853294503094285e-07, + "loss": 0.8038, + "step": 46102 + }, + { + "epoch": 8.38, + "learning_rate": 4.852748452857663e-07, + "loss": 0.2765, + "step": 46103 + }, + { + "epoch": 8.38, + "learning_rate": 4.852202402621042e-07, + "loss": 0.7059, + "step": 46104 + }, + { + "epoch": 8.38, + "learning_rate": 4.85165635238442e-07, + "loss": 0.0101, + "step": 46105 + }, + { + "epoch": 8.38, + "learning_rate": 4.851110302147797e-07, + "loss": 1.4674, + "step": 46106 + }, + { + "epoch": 8.38, + "learning_rate": 4.850564251911177e-07, + "loss": 0.0637, + "step": 46107 + }, + { + "epoch": 8.38, + "learning_rate": 4.850018201674554e-07, + "loss": 0.1454, + "step": 46108 + }, + { + "epoch": 8.38, + "learning_rate": 4.849472151437932e-07, + "loss": 0.4256, + "step": 46109 + }, + { + "epoch": 8.39, + "learning_rate": 4.848926101201311e-07, + "loss": 2.6948, + "step": 46110 + }, + { + "epoch": 8.39, + "learning_rate": 4.848380050964689e-07, + "loss": 0.6564, + "step": 46111 + }, + { + "epoch": 8.39, + "learning_rate": 4.847834000728067e-07, + "loss": 0.51, + "step": 46112 + }, + { + "epoch": 8.39, + "learning_rate": 4.847287950491445e-07, + "loss": 0.2871, + "step": 46113 + }, + { + "epoch": 8.39, + "learning_rate": 4.846741900254824e-07, + "loss": 0.4279, + "step": 46114 + }, + { + "epoch": 8.39, + "learning_rate": 4.846195850018202e-07, + "loss": 0.0572, + "step": 46115 + }, + { + "epoch": 8.39, + "learning_rate": 4.84564979978158e-07, + "loss": 0.0837, + "step": 46116 + }, + { + "epoch": 8.39, + "learning_rate": 4.845103749544958e-07, + "loss": 0.0076, + "step": 46117 + }, + { + "epoch": 8.39, + "learning_rate": 4.844557699308337e-07, + "loss": 0.5445, + "step": 46118 + }, + { + "epoch": 8.39, + "learning_rate": 4.844011649071715e-07, + "loss": 0.1195, + "step": 46119 + }, + { + "epoch": 8.39, + "learning_rate": 4.843465598835093e-07, + "loss": 0.5851, + "step": 46120 + }, + { + "epoch": 8.39, + "learning_rate": 4.842919548598472e-07, + "loss": 0.0084, + "step": 46121 + }, + { + "epoch": 8.39, + "learning_rate": 4.842373498361849e-07, + "loss": 0.8706, + "step": 46122 + }, + { + "epoch": 8.39, + "learning_rate": 4.841827448125227e-07, + "loss": 0.0842, + "step": 46123 + }, + { + "epoch": 8.39, + "learning_rate": 4.841281397888606e-07, + "loss": 0.4038, + "step": 46124 + }, + { + "epoch": 8.39, + "learning_rate": 4.840735347651984e-07, + "loss": 0.8929, + "step": 46125 + }, + { + "epoch": 8.39, + "learning_rate": 4.840189297415362e-07, + "loss": 0.1422, + "step": 46126 + }, + { + "epoch": 8.39, + "learning_rate": 4.839643247178741e-07, + "loss": 0.2714, + "step": 46127 + }, + { + "epoch": 8.39, + "learning_rate": 4.839097196942119e-07, + "loss": 1.3855, + "step": 46128 + }, + { + "epoch": 8.39, + "learning_rate": 4.838551146705497e-07, + "loss": 0.0159, + "step": 46129 + }, + { + "epoch": 8.39, + "learning_rate": 4.838005096468875e-07, + "loss": 0.3414, + "step": 46130 + }, + { + "epoch": 8.39, + "learning_rate": 4.837459046232253e-07, + "loss": 0.2583, + "step": 46131 + }, + { + "epoch": 8.39, + "learning_rate": 4.836912995995632e-07, + "loss": 0.9633, + "step": 46132 + }, + { + "epoch": 8.39, + "learning_rate": 4.836366945759009e-07, + "loss": 0.3076, + "step": 46133 + }, + { + "epoch": 8.39, + "learning_rate": 4.835820895522389e-07, + "loss": 0.6601, + "step": 46134 + }, + { + "epoch": 8.39, + "learning_rate": 4.835274845285767e-07, + "loss": 2.4022, + "step": 46135 + }, + { + "epoch": 8.39, + "learning_rate": 4.834728795049144e-07, + "loss": 1.6191, + "step": 46136 + }, + { + "epoch": 8.39, + "learning_rate": 4.834182744812523e-07, + "loss": 0.0503, + "step": 46137 + }, + { + "epoch": 8.39, + "learning_rate": 4.833636694575901e-07, + "loss": 0.1195, + "step": 46138 + }, + { + "epoch": 8.39, + "learning_rate": 4.833090644339279e-07, + "loss": 0.2505, + "step": 46139 + }, + { + "epoch": 8.39, + "learning_rate": 4.832544594102657e-07, + "loss": 2.1354, + "step": 46140 + }, + { + "epoch": 8.39, + "learning_rate": 4.831998543866036e-07, + "loss": 0.1135, + "step": 46141 + }, + { + "epoch": 8.39, + "learning_rate": 4.831452493629414e-07, + "loss": 0.2094, + "step": 46142 + }, + { + "epoch": 8.39, + "learning_rate": 4.830906443392792e-07, + "loss": 1.4112, + "step": 46143 + }, + { + "epoch": 8.39, + "learning_rate": 4.830360393156171e-07, + "loss": 0.0189, + "step": 46144 + }, + { + "epoch": 8.39, + "learning_rate": 4.829814342919549e-07, + "loss": 0.0131, + "step": 46145 + }, + { + "epoch": 8.39, + "learning_rate": 4.829268292682927e-07, + "loss": 1.0604, + "step": 46146 + }, + { + "epoch": 8.39, + "learning_rate": 4.828722242446305e-07, + "loss": 4.6025, + "step": 46147 + }, + { + "epoch": 8.39, + "learning_rate": 4.828176192209684e-07, + "loss": 0.9884, + "step": 46148 + }, + { + "epoch": 8.39, + "learning_rate": 4.827630141973062e-07, + "loss": 1.2282, + "step": 46149 + }, + { + "epoch": 8.39, + "learning_rate": 4.827084091736439e-07, + "loss": 0.0052, + "step": 46150 + }, + { + "epoch": 8.39, + "learning_rate": 4.826538041499819e-07, + "loss": 0.0322, + "step": 46151 + }, + { + "epoch": 8.39, + "learning_rate": 4.825991991263196e-07, + "loss": 0.012, + "step": 46152 + }, + { + "epoch": 8.39, + "learning_rate": 4.825445941026574e-07, + "loss": 0.0791, + "step": 46153 + }, + { + "epoch": 8.39, + "learning_rate": 4.824899890789953e-07, + "loss": 0.2837, + "step": 46154 + }, + { + "epoch": 8.39, + "learning_rate": 4.824353840553331e-07, + "loss": 1.8545, + "step": 46155 + }, + { + "epoch": 8.39, + "learning_rate": 4.82380779031671e-07, + "loss": 0.3665, + "step": 46156 + }, + { + "epoch": 8.39, + "learning_rate": 4.823261740080087e-07, + "loss": 0.3532, + "step": 46157 + }, + { + "epoch": 8.39, + "learning_rate": 4.822715689843466e-07, + "loss": 0.0475, + "step": 46158 + }, + { + "epoch": 8.39, + "learning_rate": 4.822169639606844e-07, + "loss": 1.0972, + "step": 46159 + }, + { + "epoch": 8.39, + "learning_rate": 4.821623589370222e-07, + "loss": 0.0089, + "step": 46160 + }, + { + "epoch": 8.39, + "learning_rate": 4.8210775391336e-07, + "loss": 0.1384, + "step": 46161 + }, + { + "epoch": 8.39, + "learning_rate": 4.820531488896979e-07, + "loss": 0.1039, + "step": 46162 + }, + { + "epoch": 8.39, + "learning_rate": 4.819985438660357e-07, + "loss": 0.8255, + "step": 46163 + }, + { + "epoch": 8.39, + "learning_rate": 4.819439388423736e-07, + "loss": 0.0512, + "step": 46164 + }, + { + "epoch": 8.4, + "learning_rate": 4.818893338187114e-07, + "loss": 0.5401, + "step": 46165 + }, + { + "epoch": 8.4, + "learning_rate": 4.818347287950491e-07, + "loss": 0.1198, + "step": 46166 + }, + { + "epoch": 8.4, + "learning_rate": 4.81780123771387e-07, + "loss": 0.3167, + "step": 46167 + }, + { + "epoch": 8.4, + "learning_rate": 4.817255187477248e-07, + "loss": 0.0059, + "step": 46168 + }, + { + "epoch": 8.4, + "learning_rate": 4.816709137240626e-07, + "loss": 0.093, + "step": 46169 + }, + { + "epoch": 8.4, + "learning_rate": 4.816163087004004e-07, + "loss": 2.2362, + "step": 46170 + }, + { + "epoch": 8.4, + "learning_rate": 4.815617036767383e-07, + "loss": 0.0039, + "step": 46171 + }, + { + "epoch": 8.4, + "learning_rate": 4.815070986530761e-07, + "loss": 0.6874, + "step": 46172 + }, + { + "epoch": 8.4, + "learning_rate": 4.814524936294139e-07, + "loss": 0.3833, + "step": 46173 + }, + { + "epoch": 8.4, + "learning_rate": 4.813978886057517e-07, + "loss": 0.0122, + "step": 46174 + }, + { + "epoch": 8.4, + "learning_rate": 4.813432835820896e-07, + "loss": 0.0099, + "step": 46175 + }, + { + "epoch": 8.4, + "learning_rate": 4.812886785584274e-07, + "loss": 1.5358, + "step": 46176 + }, + { + "epoch": 8.4, + "learning_rate": 4.812340735347651e-07, + "loss": 0.5957, + "step": 46177 + }, + { + "epoch": 8.4, + "learning_rate": 4.811794685111031e-07, + "loss": 0.2245, + "step": 46178 + }, + { + "epoch": 8.4, + "learning_rate": 4.811248634874409e-07, + "loss": 0.0494, + "step": 46179 + }, + { + "epoch": 8.4, + "learning_rate": 4.810702584637786e-07, + "loss": 0.0062, + "step": 46180 + }, + { + "epoch": 8.4, + "learning_rate": 4.810156534401165e-07, + "loss": 0.6266, + "step": 46181 + }, + { + "epoch": 8.4, + "learning_rate": 4.809610484164543e-07, + "loss": 1.4146, + "step": 46182 + }, + { + "epoch": 8.4, + "learning_rate": 4.809064433927922e-07, + "loss": 0.0189, + "step": 46183 + }, + { + "epoch": 8.4, + "learning_rate": 4.8085183836913e-07, + "loss": 0.1294, + "step": 46184 + }, + { + "epoch": 8.4, + "learning_rate": 4.807972333454678e-07, + "loss": 1.4309, + "step": 46185 + }, + { + "epoch": 8.4, + "learning_rate": 4.807426283218057e-07, + "loss": 0.794, + "step": 46186 + }, + { + "epoch": 8.4, + "learning_rate": 4.806880232981434e-07, + "loss": 2.2045, + "step": 46187 + }, + { + "epoch": 8.4, + "learning_rate": 4.806334182744813e-07, + "loss": 0.0616, + "step": 46188 + }, + { + "epoch": 8.4, + "learning_rate": 4.805788132508191e-07, + "loss": 0.0362, + "step": 46189 + }, + { + "epoch": 8.4, + "learning_rate": 4.805242082271569e-07, + "loss": 0.0318, + "step": 46190 + }, + { + "epoch": 8.4, + "learning_rate": 4.804696032034947e-07, + "loss": 0.0092, + "step": 46191 + }, + { + "epoch": 8.4, + "learning_rate": 4.804149981798326e-07, + "loss": 0.0075, + "step": 46192 + }, + { + "epoch": 8.4, + "learning_rate": 4.803603931561704e-07, + "loss": 0.0754, + "step": 46193 + }, + { + "epoch": 8.4, + "learning_rate": 4.803057881325082e-07, + "loss": 1.262, + "step": 46194 + }, + { + "epoch": 8.4, + "learning_rate": 4.802511831088461e-07, + "loss": 0.4178, + "step": 46195 + }, + { + "epoch": 8.4, + "learning_rate": 4.801965780851838e-07, + "loss": 0.4676, + "step": 46196 + }, + { + "epoch": 8.4, + "learning_rate": 4.801419730615217e-07, + "loss": 0.0829, + "step": 46197 + }, + { + "epoch": 8.4, + "learning_rate": 4.800873680378595e-07, + "loss": 0.0409, + "step": 46198 + }, + { + "epoch": 8.4, + "learning_rate": 4.800327630141973e-07, + "loss": 0.2027, + "step": 46199 + }, + { + "epoch": 8.4, + "learning_rate": 4.799781579905352e-07, + "loss": 0.0466, + "step": 46200 + }, + { + "epoch": 8.4, + "learning_rate": 4.799235529668729e-07, + "loss": 0.0623, + "step": 46201 + }, + { + "epoch": 8.4, + "learning_rate": 4.798689479432109e-07, + "loss": 0.0064, + "step": 46202 + }, + { + "epoch": 8.4, + "learning_rate": 4.798143429195486e-07, + "loss": 0.1917, + "step": 46203 + }, + { + "epoch": 8.4, + "learning_rate": 4.797597378958864e-07, + "loss": 0.0962, + "step": 46204 + }, + { + "epoch": 8.4, + "learning_rate": 4.797051328722243e-07, + "loss": 1.0398, + "step": 46205 + }, + { + "epoch": 8.4, + "learning_rate": 4.796505278485621e-07, + "loss": 0.2166, + "step": 46206 + }, + { + "epoch": 8.4, + "learning_rate": 4.795959228248998e-07, + "loss": 1.4162, + "step": 46207 + }, + { + "epoch": 8.4, + "learning_rate": 4.795413178012378e-07, + "loss": 0.0496, + "step": 46208 + }, + { + "epoch": 8.4, + "learning_rate": 4.794867127775756e-07, + "loss": 0.3411, + "step": 46209 + }, + { + "epoch": 8.4, + "learning_rate": 4.794321077539133e-07, + "loss": 1.1417, + "step": 46210 + }, + { + "epoch": 8.4, + "learning_rate": 4.793775027302512e-07, + "loss": 1.0829, + "step": 46211 + }, + { + "epoch": 8.4, + "learning_rate": 4.79322897706589e-07, + "loss": 0.0275, + "step": 46212 + }, + { + "epoch": 8.4, + "learning_rate": 4.792682926829269e-07, + "loss": 0.783, + "step": 46213 + }, + { + "epoch": 8.4, + "learning_rate": 4.792136876592646e-07, + "loss": 0.6131, + "step": 46214 + }, + { + "epoch": 8.4, + "learning_rate": 4.791590826356025e-07, + "loss": 0.1691, + "step": 46215 + }, + { + "epoch": 8.4, + "learning_rate": 4.791044776119404e-07, + "loss": 0.8681, + "step": 46216 + }, + { + "epoch": 8.4, + "learning_rate": 4.790498725882781e-07, + "loss": 0.0624, + "step": 46217 + }, + { + "epoch": 8.4, + "learning_rate": 4.789952675646159e-07, + "loss": 0.1407, + "step": 46218 + }, + { + "epoch": 8.4, + "learning_rate": 4.789406625409538e-07, + "loss": 0.0727, + "step": 46219 + }, + { + "epoch": 8.41, + "learning_rate": 4.788860575172916e-07, + "loss": 0.5097, + "step": 46220 + }, + { + "epoch": 8.41, + "learning_rate": 4.788314524936293e-07, + "loss": 0.1268, + "step": 46221 + }, + { + "epoch": 8.41, + "learning_rate": 4.787768474699673e-07, + "loss": 0.7758, + "step": 46222 + }, + { + "epoch": 8.41, + "learning_rate": 4.787222424463051e-07, + "loss": 0.3784, + "step": 46223 + }, + { + "epoch": 8.41, + "learning_rate": 4.786676374226429e-07, + "loss": 0.1824, + "step": 46224 + }, + { + "epoch": 8.41, + "learning_rate": 4.786130323989807e-07, + "loss": 0.037, + "step": 46225 + }, + { + "epoch": 8.41, + "learning_rate": 4.785584273753185e-07, + "loss": 0.002, + "step": 46226 + }, + { + "epoch": 8.41, + "learning_rate": 4.785038223516564e-07, + "loss": 0.0295, + "step": 46227 + }, + { + "epoch": 8.41, + "learning_rate": 4.784492173279942e-07, + "loss": 0.1899, + "step": 46228 + }, + { + "epoch": 8.41, + "learning_rate": 4.78394612304332e-07, + "loss": 0.7642, + "step": 46229 + }, + { + "epoch": 8.41, + "learning_rate": 4.783400072806699e-07, + "loss": 0.0172, + "step": 46230 + }, + { + "epoch": 8.41, + "learning_rate": 4.782854022570076e-07, + "loss": 0.0422, + "step": 46231 + }, + { + "epoch": 8.41, + "learning_rate": 4.782307972333456e-07, + "loss": 0.6507, + "step": 46232 + }, + { + "epoch": 8.41, + "learning_rate": 4.781761922096833e-07, + "loss": 3.4975, + "step": 46233 + }, + { + "epoch": 8.41, + "learning_rate": 4.781215871860211e-07, + "loss": 3.7232, + "step": 46234 + }, + { + "epoch": 8.41, + "learning_rate": 4.78066982162359e-07, + "loss": 0.0482, + "step": 46235 + }, + { + "epoch": 8.41, + "learning_rate": 4.780123771386968e-07, + "loss": 0.0152, + "step": 46236 + }, + { + "epoch": 8.41, + "learning_rate": 4.779577721150346e-07, + "loss": 0.1711, + "step": 46237 + }, + { + "epoch": 8.41, + "learning_rate": 4.779031670913724e-07, + "loss": 0.7502, + "step": 46238 + }, + { + "epoch": 8.41, + "learning_rate": 4.778485620677103e-07, + "loss": 0.0025, + "step": 46239 + }, + { + "epoch": 8.41, + "learning_rate": 4.77793957044048e-07, + "loss": 0.4552, + "step": 46240 + }, + { + "epoch": 8.41, + "learning_rate": 4.777393520203859e-07, + "loss": 1.4374, + "step": 46241 + }, + { + "epoch": 8.41, + "learning_rate": 4.776847469967237e-07, + "loss": 2.8637, + "step": 46242 + }, + { + "epoch": 8.41, + "learning_rate": 4.776301419730616e-07, + "loss": 0.1007, + "step": 46243 + }, + { + "epoch": 8.41, + "learning_rate": 4.775755369493993e-07, + "loss": 0.0533, + "step": 46244 + }, + { + "epoch": 8.41, + "learning_rate": 4.775209319257371e-07, + "loss": 0.0038, + "step": 46245 + }, + { + "epoch": 8.41, + "learning_rate": 4.774663269020751e-07, + "loss": 0.1712, + "step": 46246 + }, + { + "epoch": 8.41, + "learning_rate": 4.774117218784128e-07, + "loss": 0.2741, + "step": 46247 + }, + { + "epoch": 8.41, + "learning_rate": 4.773571168547506e-07, + "loss": 0.1473, + "step": 46248 + }, + { + "epoch": 8.41, + "learning_rate": 4.773025118310885e-07, + "loss": 2.3484, + "step": 46249 + }, + { + "epoch": 8.41, + "learning_rate": 4.772479068074263e-07, + "loss": 0.0848, + "step": 46250 + }, + { + "epoch": 8.41, + "learning_rate": 4.77193301783764e-07, + "loss": 0.2639, + "step": 46251 + }, + { + "epoch": 8.41, + "learning_rate": 4.77138696760102e-07, + "loss": 0.1159, + "step": 46252 + }, + { + "epoch": 8.41, + "learning_rate": 4.770840917364398e-07, + "loss": 0.2036, + "step": 46253 + }, + { + "epoch": 8.41, + "learning_rate": 4.770294867127776e-07, + "loss": 0.8657, + "step": 46254 + }, + { + "epoch": 8.41, + "learning_rate": 4.769748816891154e-07, + "loss": 0.2386, + "step": 46255 + }, + { + "epoch": 8.41, + "learning_rate": 4.769202766654532e-07, + "loss": 0.0063, + "step": 46256 + }, + { + "epoch": 8.41, + "learning_rate": 4.768656716417911e-07, + "loss": 0.1498, + "step": 46257 + }, + { + "epoch": 8.41, + "learning_rate": 4.7681106661812885e-07, + "loss": 0.1235, + "step": 46258 + }, + { + "epoch": 8.41, + "learning_rate": 4.767564615944667e-07, + "loss": 1.2407, + "step": 46259 + }, + { + "epoch": 8.41, + "learning_rate": 4.767018565708046e-07, + "loss": 0.0619, + "step": 46260 + }, + { + "epoch": 8.41, + "learning_rate": 4.766472515471423e-07, + "loss": 0.0876, + "step": 46261 + }, + { + "epoch": 8.41, + "learning_rate": 4.765926465234802e-07, + "loss": 0.3056, + "step": 46262 + }, + { + "epoch": 8.41, + "learning_rate": 4.76538041499818e-07, + "loss": 0.0305, + "step": 46263 + }, + { + "epoch": 8.41, + "learning_rate": 4.764834364761558e-07, + "loss": 1.572, + "step": 46264 + }, + { + "epoch": 8.41, + "learning_rate": 4.764288314524936e-07, + "loss": 0.2495, + "step": 46265 + }, + { + "epoch": 8.41, + "learning_rate": 4.7637422642883145e-07, + "loss": 0.0396, + "step": 46266 + }, + { + "epoch": 8.41, + "learning_rate": 4.7631962140516934e-07, + "loss": 0.0065, + "step": 46267 + }, + { + "epoch": 8.41, + "learning_rate": 4.762650163815071e-07, + "loss": 0.0948, + "step": 46268 + }, + { + "epoch": 8.41, + "learning_rate": 4.7621041135784496e-07, + "loss": 0.0044, + "step": 46269 + }, + { + "epoch": 8.41, + "learning_rate": 4.7615580633418274e-07, + "loss": 0.6451, + "step": 46270 + }, + { + "epoch": 8.41, + "learning_rate": 4.761012013105206e-07, + "loss": 0.1787, + "step": 46271 + }, + { + "epoch": 8.41, + "learning_rate": 4.7604659628685837e-07, + "loss": 0.1424, + "step": 46272 + }, + { + "epoch": 8.41, + "learning_rate": 4.759919912631962e-07, + "loss": 0.0062, + "step": 46273 + }, + { + "epoch": 8.41, + "learning_rate": 4.759373862395341e-07, + "loss": 0.046, + "step": 46274 + }, + { + "epoch": 8.42, + "learning_rate": 4.758827812158719e-07, + "loss": 0.0088, + "step": 46275 + }, + { + "epoch": 8.42, + "learning_rate": 4.758281761922097e-07, + "loss": 2.133, + "step": 46276 + }, + { + "epoch": 8.42, + "learning_rate": 4.757735711685475e-07, + "loss": 1.3458, + "step": 46277 + }, + { + "epoch": 8.42, + "learning_rate": 4.7571896614488534e-07, + "loss": 1.7416, + "step": 46278 + }, + { + "epoch": 8.42, + "learning_rate": 4.756643611212231e-07, + "loss": 0.0126, + "step": 46279 + }, + { + "epoch": 8.42, + "learning_rate": 4.75609756097561e-07, + "loss": 0.0142, + "step": 46280 + }, + { + "epoch": 8.42, + "learning_rate": 4.7555515107389875e-07, + "loss": 0.1541, + "step": 46281 + }, + { + "epoch": 8.42, + "learning_rate": 4.7550054605023664e-07, + "loss": 1.0672, + "step": 46282 + }, + { + "epoch": 8.42, + "learning_rate": 4.7544594102657447e-07, + "loss": 0.3425, + "step": 46283 + }, + { + "epoch": 8.42, + "learning_rate": 4.7539133600291226e-07, + "loss": 0.0061, + "step": 46284 + }, + { + "epoch": 8.42, + "learning_rate": 4.7533673097925015e-07, + "loss": 0.0077, + "step": 46285 + }, + { + "epoch": 8.42, + "learning_rate": 4.752821259555879e-07, + "loss": 2.0144, + "step": 46286 + }, + { + "epoch": 8.42, + "learning_rate": 4.7522752093192577e-07, + "loss": 0.424, + "step": 46287 + }, + { + "epoch": 8.42, + "learning_rate": 4.7517291590826355e-07, + "loss": 1.0342, + "step": 46288 + }, + { + "epoch": 8.42, + "learning_rate": 4.751183108846014e-07, + "loss": 2.5021, + "step": 46289 + }, + { + "epoch": 8.42, + "learning_rate": 4.7506370586093923e-07, + "loss": 0.4068, + "step": 46290 + }, + { + "epoch": 8.42, + "learning_rate": 4.75009100837277e-07, + "loss": 0.0076, + "step": 46291 + }, + { + "epoch": 8.42, + "learning_rate": 4.749544958136149e-07, + "loss": 0.0281, + "step": 46292 + }, + { + "epoch": 8.42, + "learning_rate": 4.7489989078995264e-07, + "loss": 1.1396, + "step": 46293 + }, + { + "epoch": 8.42, + "learning_rate": 4.7484528576629053e-07, + "loss": 0.4224, + "step": 46294 + }, + { + "epoch": 8.42, + "learning_rate": 4.747906807426283e-07, + "loss": 0.0293, + "step": 46295 + }, + { + "epoch": 8.42, + "learning_rate": 4.7473607571896615e-07, + "loss": 0.7807, + "step": 46296 + }, + { + "epoch": 8.42, + "learning_rate": 4.7468147069530404e-07, + "loss": 0.1808, + "step": 46297 + }, + { + "epoch": 8.42, + "learning_rate": 4.7462686567164177e-07, + "loss": 0.0135, + "step": 46298 + }, + { + "epoch": 8.42, + "learning_rate": 4.7457226064797966e-07, + "loss": 0.8491, + "step": 46299 + }, + { + "epoch": 8.42, + "learning_rate": 4.7451765562431745e-07, + "loss": 0.0086, + "step": 46300 + }, + { + "epoch": 8.42, + "learning_rate": 4.744630506006553e-07, + "loss": 0.2938, + "step": 46301 + }, + { + "epoch": 8.42, + "learning_rate": 4.7440844557699307e-07, + "loss": 0.0071, + "step": 46302 + }, + { + "epoch": 8.42, + "learning_rate": 4.743538405533309e-07, + "loss": 0.0895, + "step": 46303 + }, + { + "epoch": 8.42, + "learning_rate": 4.742992355296688e-07, + "loss": 0.7401, + "step": 46304 + }, + { + "epoch": 8.42, + "learning_rate": 4.742446305060066e-07, + "loss": 0.7461, + "step": 46305 + }, + { + "epoch": 8.42, + "learning_rate": 4.741900254823444e-07, + "loss": 0.0671, + "step": 46306 + }, + { + "epoch": 8.42, + "learning_rate": 4.741354204586822e-07, + "loss": 3.2229, + "step": 46307 + }, + { + "epoch": 8.42, + "learning_rate": 4.7408081543502004e-07, + "loss": 0.0697, + "step": 46308 + }, + { + "epoch": 8.42, + "learning_rate": 4.740262104113578e-07, + "loss": 0.893, + "step": 46309 + }, + { + "epoch": 8.42, + "learning_rate": 4.7397160538769566e-07, + "loss": 0.5346, + "step": 46310 + }, + { + "epoch": 8.42, + "learning_rate": 4.7391700036403355e-07, + "loss": 0.6208, + "step": 46311 + }, + { + "epoch": 8.42, + "learning_rate": 4.7386239534037134e-07, + "loss": 0.0869, + "step": 46312 + }, + { + "epoch": 8.42, + "learning_rate": 4.738077903167092e-07, + "loss": 0.6383, + "step": 46313 + }, + { + "epoch": 8.42, + "learning_rate": 4.7375318529304696e-07, + "loss": 0.7459, + "step": 46314 + }, + { + "epoch": 8.42, + "learning_rate": 4.736985802693848e-07, + "loss": 0.0606, + "step": 46315 + }, + { + "epoch": 8.42, + "learning_rate": 4.736439752457226e-07, + "loss": 1.4192, + "step": 46316 + }, + { + "epoch": 8.42, + "learning_rate": 4.7358937022206047e-07, + "loss": 0.6879, + "step": 46317 + }, + { + "epoch": 8.42, + "learning_rate": 4.735347651983983e-07, + "loss": 0.2081, + "step": 46318 + }, + { + "epoch": 8.42, + "learning_rate": 4.734801601747361e-07, + "loss": 1.4811, + "step": 46319 + }, + { + "epoch": 8.42, + "learning_rate": 4.7342555515107393e-07, + "loss": 0.5156, + "step": 46320 + }, + { + "epoch": 8.42, + "learning_rate": 4.733709501274117e-07, + "loss": 0.7327, + "step": 46321 + }, + { + "epoch": 8.42, + "learning_rate": 4.7331634510374955e-07, + "loss": 0.0049, + "step": 46322 + }, + { + "epoch": 8.42, + "learning_rate": 4.7326174008008734e-07, + "loss": 1.3124, + "step": 46323 + }, + { + "epoch": 8.42, + "learning_rate": 4.7320713505642523e-07, + "loss": 1.2974, + "step": 46324 + }, + { + "epoch": 8.42, + "learning_rate": 4.7315253003276296e-07, + "loss": 0.7661, + "step": 46325 + }, + { + "epoch": 8.42, + "learning_rate": 4.7309792500910085e-07, + "loss": 0.1715, + "step": 46326 + }, + { + "epoch": 8.42, + "learning_rate": 4.730433199854387e-07, + "loss": 1.1625, + "step": 46327 + }, + { + "epoch": 8.42, + "learning_rate": 4.7298871496177647e-07, + "loss": 0.0944, + "step": 46328 + }, + { + "epoch": 8.42, + "learning_rate": 4.7293410993811436e-07, + "loss": 0.0164, + "step": 46329 + }, + { + "epoch": 8.43, + "learning_rate": 4.728795049144521e-07, + "loss": 0.1077, + "step": 46330 + }, + { + "epoch": 8.43, + "learning_rate": 4.7282489989079e-07, + "loss": 0.0051, + "step": 46331 + }, + { + "epoch": 8.43, + "learning_rate": 4.7277029486712777e-07, + "loss": 0.0558, + "step": 46332 + }, + { + "epoch": 8.43, + "learning_rate": 4.727156898434656e-07, + "loss": 0.2279, + "step": 46333 + }, + { + "epoch": 8.43, + "learning_rate": 4.7266108481980345e-07, + "loss": 0.0048, + "step": 46334 + }, + { + "epoch": 8.43, + "learning_rate": 4.7260647979614123e-07, + "loss": 0.0197, + "step": 46335 + }, + { + "epoch": 8.43, + "learning_rate": 4.725518747724791e-07, + "loss": 0.0973, + "step": 46336 + }, + { + "epoch": 8.43, + "learning_rate": 4.724972697488169e-07, + "loss": 0.0236, + "step": 46337 + }, + { + "epoch": 8.43, + "learning_rate": 4.7244266472515474e-07, + "loss": 0.0261, + "step": 46338 + }, + { + "epoch": 8.43, + "learning_rate": 4.7238805970149253e-07, + "loss": 0.1478, + "step": 46339 + }, + { + "epoch": 8.43, + "learning_rate": 4.7233345467783036e-07, + "loss": 0.2879, + "step": 46340 + }, + { + "epoch": 8.43, + "learning_rate": 4.7227884965416825e-07, + "loss": 2.0285, + "step": 46341 + }, + { + "epoch": 8.43, + "learning_rate": 4.72224244630506e-07, + "loss": 0.0039, + "step": 46342 + }, + { + "epoch": 8.43, + "learning_rate": 4.721696396068439e-07, + "loss": 1.0227, + "step": 46343 + }, + { + "epoch": 8.43, + "learning_rate": 4.7211503458318166e-07, + "loss": 0.5507, + "step": 46344 + }, + { + "epoch": 8.43, + "learning_rate": 4.720604295595195e-07, + "loss": 1.5203, + "step": 46345 + }, + { + "epoch": 8.43, + "learning_rate": 4.720058245358573e-07, + "loss": 2.0302, + "step": 46346 + }, + { + "epoch": 8.43, + "learning_rate": 4.719512195121951e-07, + "loss": 0.1724, + "step": 46347 + }, + { + "epoch": 8.43, + "learning_rate": 4.71896614488533e-07, + "loss": 1.1701, + "step": 46348 + }, + { + "epoch": 8.43, + "learning_rate": 4.718420094648708e-07, + "loss": 1.3331, + "step": 46349 + }, + { + "epoch": 8.43, + "learning_rate": 4.7178740444120863e-07, + "loss": 0.1354, + "step": 46350 + }, + { + "epoch": 8.43, + "learning_rate": 4.717327994175464e-07, + "loss": 0.081, + "step": 46351 + }, + { + "epoch": 8.43, + "learning_rate": 4.7167819439388426e-07, + "loss": 0.2859, + "step": 46352 + }, + { + "epoch": 8.43, + "learning_rate": 4.7162358937022204e-07, + "loss": 0.1085, + "step": 46353 + }, + { + "epoch": 8.43, + "learning_rate": 4.715689843465599e-07, + "loss": 0.2005, + "step": 46354 + }, + { + "epoch": 8.43, + "learning_rate": 4.7151437932289777e-07, + "loss": 0.0866, + "step": 46355 + }, + { + "epoch": 8.43, + "learning_rate": 4.7145977429923555e-07, + "loss": 0.0147, + "step": 46356 + }, + { + "epoch": 8.43, + "learning_rate": 4.714051692755734e-07, + "loss": 0.8857, + "step": 46357 + }, + { + "epoch": 8.43, + "learning_rate": 4.713505642519112e-07, + "loss": 0.2645, + "step": 46358 + }, + { + "epoch": 8.43, + "learning_rate": 4.71295959228249e-07, + "loss": 0.6137, + "step": 46359 + }, + { + "epoch": 8.43, + "learning_rate": 4.712413542045868e-07, + "loss": 0.0133, + "step": 46360 + }, + { + "epoch": 8.43, + "learning_rate": 4.711867491809247e-07, + "loss": 0.6917, + "step": 46361 + }, + { + "epoch": 8.43, + "learning_rate": 4.711321441572624e-07, + "loss": 0.8545, + "step": 46362 + }, + { + "epoch": 8.43, + "learning_rate": 4.710775391336003e-07, + "loss": 0.5977, + "step": 46363 + }, + { + "epoch": 8.43, + "learning_rate": 4.7102293410993815e-07, + "loss": 1.0213, + "step": 46364 + }, + { + "epoch": 8.43, + "learning_rate": 4.7096832908627593e-07, + "loss": 0.375, + "step": 46365 + }, + { + "epoch": 8.43, + "learning_rate": 4.7091372406261377e-07, + "loss": 0.2039, + "step": 46366 + }, + { + "epoch": 8.43, + "learning_rate": 4.7085911903895155e-07, + "loss": 0.1411, + "step": 46367 + }, + { + "epoch": 8.43, + "learning_rate": 4.7080451401528944e-07, + "loss": 1.5708, + "step": 46368 + }, + { + "epoch": 8.43, + "learning_rate": 4.7074990899162723e-07, + "loss": 0.0058, + "step": 46369 + }, + { + "epoch": 8.43, + "learning_rate": 4.7069530396796507e-07, + "loss": 0.1564, + "step": 46370 + }, + { + "epoch": 8.43, + "learning_rate": 4.706406989443029e-07, + "loss": 1.0126, + "step": 46371 + }, + { + "epoch": 8.43, + "learning_rate": 4.705860939206407e-07, + "loss": 0.1281, + "step": 46372 + }, + { + "epoch": 8.43, + "learning_rate": 4.705314888969786e-07, + "loss": 0.8823, + "step": 46373 + }, + { + "epoch": 8.43, + "learning_rate": 4.704768838733163e-07, + "loss": 0.1027, + "step": 46374 + }, + { + "epoch": 8.43, + "learning_rate": 4.704222788496542e-07, + "loss": 0.0784, + "step": 46375 + }, + { + "epoch": 8.43, + "learning_rate": 4.70367673825992e-07, + "loss": 1.8316, + "step": 46376 + }, + { + "epoch": 8.43, + "learning_rate": 4.703130688023298e-07, + "loss": 0.1198, + "step": 46377 + }, + { + "epoch": 8.43, + "learning_rate": 4.7025846377866766e-07, + "loss": 0.0429, + "step": 46378 + }, + { + "epoch": 8.43, + "learning_rate": 4.7020385875500545e-07, + "loss": 0.0453, + "step": 46379 + }, + { + "epoch": 8.43, + "learning_rate": 4.7014925373134334e-07, + "loss": 0.0416, + "step": 46380 + }, + { + "epoch": 8.43, + "learning_rate": 4.700946487076811e-07, + "loss": 0.0052, + "step": 46381 + }, + { + "epoch": 8.43, + "learning_rate": 4.7004004368401896e-07, + "loss": 0.0274, + "step": 46382 + }, + { + "epoch": 8.43, + "learning_rate": 4.6998543866035674e-07, + "loss": 3.814, + "step": 46383 + }, + { + "epoch": 8.43, + "learning_rate": 4.699308336366946e-07, + "loss": 1.5619, + "step": 46384 + }, + { + "epoch": 8.44, + "learning_rate": 4.6987622861303247e-07, + "loss": 0.5251, + "step": 46385 + }, + { + "epoch": 8.44, + "learning_rate": 4.698216235893702e-07, + "loss": 0.0216, + "step": 46386 + }, + { + "epoch": 8.44, + "learning_rate": 4.697670185657081e-07, + "loss": 0.0084, + "step": 46387 + }, + { + "epoch": 8.44, + "learning_rate": 4.697124135420459e-07, + "loss": 4.2035, + "step": 46388 + }, + { + "epoch": 8.44, + "learning_rate": 4.696578085183837e-07, + "loss": 0.7712, + "step": 46389 + }, + { + "epoch": 8.44, + "learning_rate": 4.696032034947215e-07, + "loss": 0.0482, + "step": 46390 + }, + { + "epoch": 8.44, + "learning_rate": 4.6954859847105934e-07, + "loss": 0.1445, + "step": 46391 + }, + { + "epoch": 8.44, + "learning_rate": 4.6949399344739723e-07, + "loss": 0.2867, + "step": 46392 + }, + { + "epoch": 8.44, + "learning_rate": 4.69439388423735e-07, + "loss": 0.3544, + "step": 46393 + }, + { + "epoch": 8.44, + "learning_rate": 4.6938478340007285e-07, + "loss": 0.787, + "step": 46394 + }, + { + "epoch": 8.44, + "learning_rate": 4.6933017837641063e-07, + "loss": 0.3998, + "step": 46395 + }, + { + "epoch": 8.44, + "learning_rate": 4.6927557335274847e-07, + "loss": 0.3452, + "step": 46396 + }, + { + "epoch": 8.44, + "learning_rate": 4.6922096832908626e-07, + "loss": 0.1518, + "step": 46397 + }, + { + "epoch": 8.44, + "learning_rate": 4.691663633054241e-07, + "loss": 0.7592, + "step": 46398 + }, + { + "epoch": 8.44, + "learning_rate": 4.691117582817619e-07, + "loss": 1.6229, + "step": 46399 + }, + { + "epoch": 8.44, + "learning_rate": 4.6905715325809977e-07, + "loss": 0.0053, + "step": 46400 + }, + { + "epoch": 8.44, + "learning_rate": 4.690025482344376e-07, + "loss": 0.4916, + "step": 46401 + }, + { + "epoch": 8.44, + "learning_rate": 4.689479432107754e-07, + "loss": 0.0039, + "step": 46402 + }, + { + "epoch": 8.44, + "learning_rate": 4.6889333818711323e-07, + "loss": 1.1688, + "step": 46403 + }, + { + "epoch": 8.44, + "learning_rate": 4.68838733163451e-07, + "loss": 1.3409, + "step": 46404 + }, + { + "epoch": 8.44, + "learning_rate": 4.687841281397889e-07, + "loss": 0.0682, + "step": 46405 + }, + { + "epoch": 8.44, + "learning_rate": 4.6872952311612663e-07, + "loss": 0.4628, + "step": 46406 + }, + { + "epoch": 8.44, + "learning_rate": 4.686749180924645e-07, + "loss": 0.5304, + "step": 46407 + }, + { + "epoch": 8.44, + "learning_rate": 4.6862031306880236e-07, + "loss": 0.0113, + "step": 46408 + }, + { + "epoch": 8.44, + "learning_rate": 4.6856570804514015e-07, + "loss": 0.0046, + "step": 46409 + }, + { + "epoch": 8.44, + "learning_rate": 4.68511103021478e-07, + "loss": 0.0441, + "step": 46410 + }, + { + "epoch": 8.44, + "learning_rate": 4.6845649799781577e-07, + "loss": 0.0586, + "step": 46411 + }, + { + "epoch": 8.44, + "learning_rate": 4.6840189297415366e-07, + "loss": 0.3266, + "step": 46412 + }, + { + "epoch": 8.44, + "learning_rate": 4.6834728795049144e-07, + "loss": 0.8223, + "step": 46413 + }, + { + "epoch": 8.44, + "learning_rate": 4.682926829268293e-07, + "loss": 0.0703, + "step": 46414 + }, + { + "epoch": 8.44, + "learning_rate": 4.682380779031671e-07, + "loss": 0.0173, + "step": 46415 + }, + { + "epoch": 8.44, + "learning_rate": 4.681834728795049e-07, + "loss": 0.2662, + "step": 46416 + }, + { + "epoch": 8.44, + "learning_rate": 4.681288678558428e-07, + "loss": 3.0187, + "step": 46417 + }, + { + "epoch": 8.44, + "learning_rate": 4.680742628321805e-07, + "loss": 0.8537, + "step": 46418 + }, + { + "epoch": 8.44, + "learning_rate": 4.680196578085184e-07, + "loss": 2.3439, + "step": 46419 + }, + { + "epoch": 8.44, + "learning_rate": 4.679650527848562e-07, + "loss": 0.9883, + "step": 46420 + }, + { + "epoch": 8.44, + "learning_rate": 4.6791044776119404e-07, + "loss": 0.0123, + "step": 46421 + }, + { + "epoch": 8.44, + "learning_rate": 4.678558427375319e-07, + "loss": 0.0415, + "step": 46422 + }, + { + "epoch": 8.44, + "learning_rate": 4.6780123771386966e-07, + "loss": 0.1202, + "step": 46423 + }, + { + "epoch": 8.44, + "learning_rate": 4.6774663269020755e-07, + "loss": 0.0051, + "step": 46424 + }, + { + "epoch": 8.44, + "learning_rate": 4.6769202766654534e-07, + "loss": 0.0267, + "step": 46425 + }, + { + "epoch": 8.44, + "learning_rate": 4.6763742264288317e-07, + "loss": 0.6071, + "step": 46426 + }, + { + "epoch": 8.44, + "learning_rate": 4.6758281761922096e-07, + "loss": 1.1699, + "step": 46427 + }, + { + "epoch": 8.44, + "learning_rate": 4.675282125955588e-07, + "loss": 0.05, + "step": 46428 + }, + { + "epoch": 8.44, + "learning_rate": 4.674736075718967e-07, + "loss": 1.5099, + "step": 46429 + }, + { + "epoch": 8.44, + "learning_rate": 4.674190025482344e-07, + "loss": 0.4605, + "step": 46430 + }, + { + "epoch": 8.44, + "learning_rate": 4.673643975245723e-07, + "loss": 1.623, + "step": 46431 + }, + { + "epoch": 8.44, + "learning_rate": 4.673097925009101e-07, + "loss": 0.1686, + "step": 46432 + }, + { + "epoch": 8.44, + "learning_rate": 4.6725518747724793e-07, + "loss": 1.5708, + "step": 46433 + }, + { + "epoch": 8.44, + "learning_rate": 4.672005824535857e-07, + "loss": 1.1218, + "step": 46434 + }, + { + "epoch": 8.44, + "learning_rate": 4.6714597742992355e-07, + "loss": 0.2515, + "step": 46435 + }, + { + "epoch": 8.44, + "learning_rate": 4.6709137240626134e-07, + "loss": 0.0359, + "step": 46436 + }, + { + "epoch": 8.44, + "learning_rate": 4.6703676738259923e-07, + "loss": 2.7005, + "step": 46437 + }, + { + "epoch": 8.44, + "learning_rate": 4.6698216235893706e-07, + "loss": 0.694, + "step": 46438 + }, + { + "epoch": 8.44, + "learning_rate": 4.6692755733527485e-07, + "loss": 2.0023, + "step": 46439 + }, + { + "epoch": 8.45, + "learning_rate": 4.668729523116127e-07, + "loss": 0.0073, + "step": 46440 + }, + { + "epoch": 8.45, + "learning_rate": 4.6681834728795047e-07, + "loss": 1.6903, + "step": 46441 + }, + { + "epoch": 8.45, + "learning_rate": 4.667637422642883e-07, + "loss": 1.3309, + "step": 46442 + }, + { + "epoch": 8.45, + "learning_rate": 4.667091372406261e-07, + "loss": 0.0967, + "step": 46443 + }, + { + "epoch": 8.45, + "learning_rate": 4.66654532216964e-07, + "loss": 0.4045, + "step": 46444 + }, + { + "epoch": 8.45, + "learning_rate": 4.665999271933018e-07, + "loss": 0.1228, + "step": 46445 + }, + { + "epoch": 8.45, + "learning_rate": 4.665453221696396e-07, + "loss": 0.7018, + "step": 46446 + }, + { + "epoch": 8.45, + "learning_rate": 4.6649071714597744e-07, + "loss": 0.0124, + "step": 46447 + }, + { + "epoch": 8.45, + "learning_rate": 4.6643611212231523e-07, + "loss": 1.3869, + "step": 46448 + }, + { + "epoch": 8.45, + "learning_rate": 4.663815070986531e-07, + "loss": 0.1033, + "step": 46449 + }, + { + "epoch": 8.45, + "learning_rate": 4.6632690207499085e-07, + "loss": 0.9003, + "step": 46450 + }, + { + "epoch": 8.45, + "learning_rate": 4.6627229705132874e-07, + "loss": 1.8786, + "step": 46451 + }, + { + "epoch": 8.45, + "learning_rate": 4.662176920276666e-07, + "loss": 0.067, + "step": 46452 + }, + { + "epoch": 8.45, + "learning_rate": 4.6616308700400436e-07, + "loss": 0.3496, + "step": 46453 + }, + { + "epoch": 8.45, + "learning_rate": 4.6610848198034225e-07, + "loss": 0.0153, + "step": 46454 + }, + { + "epoch": 8.45, + "learning_rate": 4.6605387695668e-07, + "loss": 1.8221, + "step": 46455 + }, + { + "epoch": 8.45, + "learning_rate": 4.659992719330179e-07, + "loss": 2.0501, + "step": 46456 + }, + { + "epoch": 8.45, + "learning_rate": 4.6594466690935566e-07, + "loss": 0.3687, + "step": 46457 + }, + { + "epoch": 8.45, + "learning_rate": 4.658900618856935e-07, + "loss": 0.671, + "step": 46458 + }, + { + "epoch": 8.45, + "learning_rate": 4.6583545686203133e-07, + "loss": 0.0098, + "step": 46459 + }, + { + "epoch": 8.45, + "learning_rate": 4.657808518383691e-07, + "loss": 1.1752, + "step": 46460 + }, + { + "epoch": 8.45, + "learning_rate": 4.65726246814707e-07, + "loss": 0.4925, + "step": 46461 + }, + { + "epoch": 8.45, + "learning_rate": 4.6567164179104474e-07, + "loss": 0.0081, + "step": 46462 + }, + { + "epoch": 8.45, + "learning_rate": 4.6561703676738263e-07, + "loss": 0.8994, + "step": 46463 + }, + { + "epoch": 8.45, + "learning_rate": 4.655624317437204e-07, + "loss": 0.1308, + "step": 46464 + }, + { + "epoch": 8.45, + "learning_rate": 4.6550782672005825e-07, + "loss": 0.5672, + "step": 46465 + }, + { + "epoch": 8.45, + "learning_rate": 4.6545322169639614e-07, + "loss": 0.4185, + "step": 46466 + }, + { + "epoch": 8.45, + "learning_rate": 4.653986166727339e-07, + "loss": 0.0035, + "step": 46467 + }, + { + "epoch": 8.45, + "learning_rate": 4.6534401164907177e-07, + "loss": 0.0601, + "step": 46468 + }, + { + "epoch": 8.45, + "learning_rate": 4.6528940662540955e-07, + "loss": 0.7593, + "step": 46469 + }, + { + "epoch": 8.45, + "learning_rate": 4.652348016017474e-07, + "loss": 0.8515, + "step": 46470 + }, + { + "epoch": 8.45, + "learning_rate": 4.6518019657808517e-07, + "loss": 1.7959, + "step": 46471 + }, + { + "epoch": 8.45, + "learning_rate": 4.65125591554423e-07, + "loss": 0.0221, + "step": 46472 + }, + { + "epoch": 8.45, + "learning_rate": 4.650709865307609e-07, + "loss": 0.9772, + "step": 46473 + }, + { + "epoch": 8.45, + "learning_rate": 4.6501638150709863e-07, + "loss": 0.7112, + "step": 46474 + }, + { + "epoch": 8.45, + "learning_rate": 4.649617764834365e-07, + "loss": 0.0109, + "step": 46475 + }, + { + "epoch": 8.45, + "learning_rate": 4.649071714597743e-07, + "loss": 1.7003, + "step": 46476 + }, + { + "epoch": 8.45, + "learning_rate": 4.6485256643611215e-07, + "loss": 0.0217, + "step": 46477 + }, + { + "epoch": 8.45, + "learning_rate": 4.6479796141244993e-07, + "loss": 0.1848, + "step": 46478 + }, + { + "epoch": 8.45, + "learning_rate": 4.6474335638878777e-07, + "loss": 0.4361, + "step": 46479 + }, + { + "epoch": 8.45, + "learning_rate": 4.6468875136512555e-07, + "loss": 1.1652, + "step": 46480 + }, + { + "epoch": 8.45, + "learning_rate": 4.6463414634146344e-07, + "loss": 0.4641, + "step": 46481 + }, + { + "epoch": 8.45, + "learning_rate": 4.645795413178013e-07, + "loss": 0.4052, + "step": 46482 + }, + { + "epoch": 8.45, + "learning_rate": 4.6452493629413906e-07, + "loss": 0.0129, + "step": 46483 + }, + { + "epoch": 8.45, + "learning_rate": 4.644703312704769e-07, + "loss": 0.1166, + "step": 46484 + }, + { + "epoch": 8.45, + "learning_rate": 4.644157262468147e-07, + "loss": 0.0431, + "step": 46485 + }, + { + "epoch": 8.45, + "learning_rate": 4.643611212231526e-07, + "loss": 2.6381, + "step": 46486 + }, + { + "epoch": 8.45, + "learning_rate": 4.643065161994903e-07, + "loss": 0.0143, + "step": 46487 + }, + { + "epoch": 8.45, + "learning_rate": 4.642519111758282e-07, + "loss": 0.0043, + "step": 46488 + }, + { + "epoch": 8.45, + "learning_rate": 4.6419730615216604e-07, + "loss": 0.3313, + "step": 46489 + }, + { + "epoch": 8.45, + "learning_rate": 4.641427011285038e-07, + "loss": 0.3829, + "step": 46490 + }, + { + "epoch": 8.45, + "learning_rate": 4.6408809610484166e-07, + "loss": 0.121, + "step": 46491 + }, + { + "epoch": 8.45, + "learning_rate": 4.6403349108117944e-07, + "loss": 0.7518, + "step": 46492 + }, + { + "epoch": 8.45, + "learning_rate": 4.6397888605751733e-07, + "loss": 0.2249, + "step": 46493 + }, + { + "epoch": 8.45, + "learning_rate": 4.6392428103385507e-07, + "loss": 1.1854, + "step": 46494 + }, + { + "epoch": 8.46, + "learning_rate": 4.6386967601019296e-07, + "loss": 0.0726, + "step": 46495 + }, + { + "epoch": 8.46, + "learning_rate": 4.638150709865308e-07, + "loss": 0.8371, + "step": 46496 + }, + { + "epoch": 8.46, + "learning_rate": 4.637604659628686e-07, + "loss": 0.7278, + "step": 46497 + }, + { + "epoch": 8.46, + "learning_rate": 4.6370586093920647e-07, + "loss": 0.1243, + "step": 46498 + }, + { + "epoch": 8.46, + "learning_rate": 4.636512559155442e-07, + "loss": 0.2706, + "step": 46499 + }, + { + "epoch": 8.46, + "learning_rate": 4.635966508918821e-07, + "loss": 0.0122, + "step": 46500 + }, + { + "epoch": 8.46, + "learning_rate": 4.635420458682199e-07, + "loss": 2.7974, + "step": 46501 + }, + { + "epoch": 8.46, + "learning_rate": 4.634874408445577e-07, + "loss": 1.4143, + "step": 46502 + }, + { + "epoch": 8.46, + "learning_rate": 4.6343283582089555e-07, + "loss": 0.6025, + "step": 46503 + }, + { + "epoch": 8.46, + "learning_rate": 4.6337823079723333e-07, + "loss": 0.0349, + "step": 46504 + }, + { + "epoch": 8.46, + "learning_rate": 4.633236257735712e-07, + "loss": 0.008, + "step": 46505 + }, + { + "epoch": 8.46, + "learning_rate": 4.63269020749909e-07, + "loss": 2.218, + "step": 46506 + }, + { + "epoch": 8.46, + "learning_rate": 4.6321441572624685e-07, + "loss": 2.2572, + "step": 46507 + }, + { + "epoch": 8.46, + "learning_rate": 4.6315981070258463e-07, + "loss": 0.3609, + "step": 46508 + }, + { + "epoch": 8.46, + "learning_rate": 4.6310520567892247e-07, + "loss": 0.4931, + "step": 46509 + }, + { + "epoch": 8.46, + "learning_rate": 4.6305060065526036e-07, + "loss": 0.0197, + "step": 46510 + }, + { + "epoch": 8.46, + "learning_rate": 4.629959956315981e-07, + "loss": 0.2173, + "step": 46511 + }, + { + "epoch": 8.46, + "learning_rate": 4.62941390607936e-07, + "loss": 0.6272, + "step": 46512 + }, + { + "epoch": 8.46, + "learning_rate": 4.6288678558427377e-07, + "loss": 0.2558, + "step": 46513 + }, + { + "epoch": 8.46, + "learning_rate": 4.628321805606116e-07, + "loss": 0.7349, + "step": 46514 + }, + { + "epoch": 8.46, + "learning_rate": 4.627775755369494e-07, + "loss": 0.0057, + "step": 46515 + }, + { + "epoch": 8.46, + "learning_rate": 4.627229705132872e-07, + "loss": 0.0058, + "step": 46516 + }, + { + "epoch": 8.46, + "learning_rate": 4.62668365489625e-07, + "loss": 1.7595, + "step": 46517 + }, + { + "epoch": 8.46, + "learning_rate": 4.626137604659629e-07, + "loss": 0.1224, + "step": 46518 + }, + { + "epoch": 8.46, + "learning_rate": 4.6255915544230074e-07, + "loss": 0.029, + "step": 46519 + }, + { + "epoch": 8.46, + "learning_rate": 4.625045504186385e-07, + "loss": 0.386, + "step": 46520 + }, + { + "epoch": 8.46, + "learning_rate": 4.6244994539497636e-07, + "loss": 0.034, + "step": 46521 + }, + { + "epoch": 8.46, + "learning_rate": 4.6239534037131415e-07, + "loss": 0.0903, + "step": 46522 + }, + { + "epoch": 8.46, + "learning_rate": 4.62340735347652e-07, + "loss": 0.7971, + "step": 46523 + }, + { + "epoch": 8.46, + "learning_rate": 4.6228613032398977e-07, + "loss": 2.5244, + "step": 46524 + }, + { + "epoch": 8.46, + "learning_rate": 4.6223152530032766e-07, + "loss": 0.0258, + "step": 46525 + }, + { + "epoch": 8.46, + "learning_rate": 4.621769202766655e-07, + "loss": 1.5526, + "step": 46526 + }, + { + "epoch": 8.46, + "learning_rate": 4.621223152530033e-07, + "loss": 0.0112, + "step": 46527 + }, + { + "epoch": 8.46, + "learning_rate": 4.620677102293411e-07, + "loss": 1.442, + "step": 46528 + }, + { + "epoch": 8.46, + "learning_rate": 4.620131052056789e-07, + "loss": 0.1182, + "step": 46529 + }, + { + "epoch": 8.46, + "learning_rate": 4.619585001820168e-07, + "loss": 0.0607, + "step": 46530 + }, + { + "epoch": 8.46, + "learning_rate": 4.619038951583545e-07, + "loss": 2.6246, + "step": 46531 + }, + { + "epoch": 8.46, + "learning_rate": 4.618492901346924e-07, + "loss": 2.0403, + "step": 46532 + }, + { + "epoch": 8.46, + "learning_rate": 4.6179468511103025e-07, + "loss": 0.0302, + "step": 46533 + }, + { + "epoch": 8.46, + "learning_rate": 4.6174008008736804e-07, + "loss": 0.0888, + "step": 46534 + }, + { + "epoch": 8.46, + "learning_rate": 4.616854750637059e-07, + "loss": 0.0047, + "step": 46535 + }, + { + "epoch": 8.46, + "learning_rate": 4.6163087004004366e-07, + "loss": 1.7773, + "step": 46536 + }, + { + "epoch": 8.46, + "learning_rate": 4.6157626501638155e-07, + "loss": 0.0503, + "step": 46537 + }, + { + "epoch": 8.46, + "learning_rate": 4.6152165999271933e-07, + "loss": 0.0187, + "step": 46538 + }, + { + "epoch": 8.46, + "learning_rate": 4.6146705496905717e-07, + "loss": 1.0717, + "step": 46539 + }, + { + "epoch": 8.46, + "learning_rate": 4.61412449945395e-07, + "loss": 0.0198, + "step": 46540 + }, + { + "epoch": 8.46, + "learning_rate": 4.613578449217328e-07, + "loss": 0.7495, + "step": 46541 + }, + { + "epoch": 8.46, + "learning_rate": 4.613032398980707e-07, + "loss": 0.026, + "step": 46542 + }, + { + "epoch": 8.46, + "learning_rate": 4.612486348744084e-07, + "loss": 3.0652, + "step": 46543 + }, + { + "epoch": 8.46, + "learning_rate": 4.611940298507463e-07, + "loss": 0.1047, + "step": 46544 + }, + { + "epoch": 8.46, + "learning_rate": 4.611394248270841e-07, + "loss": 0.0078, + "step": 46545 + }, + { + "epoch": 8.46, + "learning_rate": 4.6108481980342193e-07, + "loss": 0.0037, + "step": 46546 + }, + { + "epoch": 8.46, + "learning_rate": 4.6103021477975977e-07, + "loss": 0.1087, + "step": 46547 + }, + { + "epoch": 8.46, + "learning_rate": 4.6097560975609755e-07, + "loss": 0.2851, + "step": 46548 + }, + { + "epoch": 8.46, + "learning_rate": 4.6092100473243544e-07, + "loss": 1.4273, + "step": 46549 + }, + { + "epoch": 8.47, + "learning_rate": 4.608663997087732e-07, + "loss": 0.3428, + "step": 46550 + }, + { + "epoch": 8.47, + "learning_rate": 4.6081179468511106e-07, + "loss": 0.3955, + "step": 46551 + }, + { + "epoch": 8.47, + "learning_rate": 4.6075718966144885e-07, + "loss": 0.8777, + "step": 46552 + }, + { + "epoch": 8.47, + "learning_rate": 4.607025846377867e-07, + "loss": 0.06, + "step": 46553 + }, + { + "epoch": 8.47, + "learning_rate": 4.6064797961412447e-07, + "loss": 0.2799, + "step": 46554 + }, + { + "epoch": 8.47, + "learning_rate": 4.605933745904623e-07, + "loss": 0.0848, + "step": 46555 + }, + { + "epoch": 8.47, + "learning_rate": 4.605387695668002e-07, + "loss": 1.1557, + "step": 46556 + }, + { + "epoch": 8.47, + "learning_rate": 4.60484164543138e-07, + "loss": 0.158, + "step": 46557 + }, + { + "epoch": 8.47, + "learning_rate": 4.604295595194758e-07, + "loss": 3.2023, + "step": 46558 + }, + { + "epoch": 8.47, + "learning_rate": 4.603749544958136e-07, + "loss": 0.4642, + "step": 46559 + }, + { + "epoch": 8.47, + "learning_rate": 4.6032034947215144e-07, + "loss": 0.5853, + "step": 46560 + }, + { + "epoch": 8.47, + "learning_rate": 4.602657444484892e-07, + "loss": 0.9745, + "step": 46561 + }, + { + "epoch": 8.47, + "learning_rate": 4.602111394248271e-07, + "loss": 0.968, + "step": 46562 + }, + { + "epoch": 8.47, + "learning_rate": 4.6015653440116495e-07, + "loss": 0.0439, + "step": 46563 + }, + { + "epoch": 8.47, + "learning_rate": 4.6010192937750274e-07, + "loss": 0.803, + "step": 46564 + }, + { + "epoch": 8.47, + "learning_rate": 4.600473243538406e-07, + "loss": 0.4097, + "step": 46565 + }, + { + "epoch": 8.47, + "learning_rate": 4.5999271933017836e-07, + "loss": 0.0198, + "step": 46566 + }, + { + "epoch": 8.47, + "learning_rate": 4.599381143065162e-07, + "loss": 0.746, + "step": 46567 + }, + { + "epoch": 8.47, + "learning_rate": 4.59883509282854e-07, + "loss": 1.4118, + "step": 46568 + }, + { + "epoch": 8.47, + "learning_rate": 4.598289042591919e-07, + "loss": 0.7183, + "step": 46569 + }, + { + "epoch": 8.47, + "learning_rate": 4.597742992355297e-07, + "loss": 1.1069, + "step": 46570 + }, + { + "epoch": 8.47, + "learning_rate": 4.597196942118675e-07, + "loss": 1.4487, + "step": 46571 + }, + { + "epoch": 8.47, + "learning_rate": 4.5966508918820533e-07, + "loss": 0.0068, + "step": 46572 + }, + { + "epoch": 8.47, + "learning_rate": 4.596104841645431e-07, + "loss": 0.0055, + "step": 46573 + }, + { + "epoch": 8.47, + "learning_rate": 4.59555879140881e-07, + "loss": 0.1831, + "step": 46574 + }, + { + "epoch": 8.47, + "learning_rate": 4.5950127411721874e-07, + "loss": 0.0419, + "step": 46575 + }, + { + "epoch": 8.47, + "learning_rate": 4.5944666909355663e-07, + "loss": 0.1334, + "step": 46576 + }, + { + "epoch": 8.47, + "learning_rate": 4.5939206406989447e-07, + "loss": 0.2923, + "step": 46577 + }, + { + "epoch": 8.47, + "learning_rate": 4.5933745904623225e-07, + "loss": 2.3909, + "step": 46578 + }, + { + "epoch": 8.47, + "learning_rate": 4.592828540225701e-07, + "loss": 0.004, + "step": 46579 + }, + { + "epoch": 8.47, + "learning_rate": 4.592282489989079e-07, + "loss": 1.2745, + "step": 46580 + }, + { + "epoch": 8.47, + "learning_rate": 4.5917364397524576e-07, + "loss": 0.0984, + "step": 46581 + }, + { + "epoch": 8.47, + "learning_rate": 4.5911903895158355e-07, + "loss": 0.0369, + "step": 46582 + }, + { + "epoch": 8.47, + "learning_rate": 4.590644339279214e-07, + "loss": 3.0402, + "step": 46583 + }, + { + "epoch": 8.47, + "learning_rate": 4.590098289042592e-07, + "loss": 0.3116, + "step": 46584 + }, + { + "epoch": 8.47, + "learning_rate": 4.58955223880597e-07, + "loss": 1.1122, + "step": 46585 + }, + { + "epoch": 8.47, + "learning_rate": 4.589006188569349e-07, + "loss": 1.8055, + "step": 46586 + }, + { + "epoch": 8.47, + "learning_rate": 4.5884601383327263e-07, + "loss": 1.326, + "step": 46587 + }, + { + "epoch": 8.47, + "learning_rate": 4.587914088096105e-07, + "loss": 0.0251, + "step": 46588 + }, + { + "epoch": 8.47, + "learning_rate": 4.587368037859483e-07, + "loss": 0.0039, + "step": 46589 + }, + { + "epoch": 8.47, + "learning_rate": 4.5868219876228614e-07, + "loss": 1.4466, + "step": 46590 + }, + { + "epoch": 8.47, + "learning_rate": 4.5862759373862393e-07, + "loss": 0.2475, + "step": 46591 + }, + { + "epoch": 8.47, + "learning_rate": 4.5857298871496177e-07, + "loss": 0.0485, + "step": 46592 + }, + { + "epoch": 8.47, + "learning_rate": 4.5851838369129966e-07, + "loss": 1.3207, + "step": 46593 + }, + { + "epoch": 8.47, + "learning_rate": 4.5846377866763744e-07, + "loss": 0.032, + "step": 46594 + }, + { + "epoch": 8.47, + "learning_rate": 4.584091736439753e-07, + "loss": 0.0049, + "step": 46595 + }, + { + "epoch": 8.47, + "learning_rate": 4.5835456862031306e-07, + "loss": 2.712, + "step": 46596 + }, + { + "epoch": 8.47, + "learning_rate": 4.582999635966509e-07, + "loss": 3.5712, + "step": 46597 + }, + { + "epoch": 8.47, + "learning_rate": 4.582453585729887e-07, + "loss": 0.2113, + "step": 46598 + }, + { + "epoch": 8.47, + "learning_rate": 4.581907535493265e-07, + "loss": 1.4927, + "step": 46599 + }, + { + "epoch": 8.47, + "learning_rate": 4.581361485256644e-07, + "loss": 0.0625, + "step": 46600 + }, + { + "epoch": 8.47, + "learning_rate": 4.580815435020022e-07, + "loss": 0.7317, + "step": 46601 + }, + { + "epoch": 8.47, + "learning_rate": 4.5802693847834004e-07, + "loss": 0.1883, + "step": 46602 + }, + { + "epoch": 8.47, + "learning_rate": 4.579723334546778e-07, + "loss": 0.0233, + "step": 46603 + }, + { + "epoch": 8.47, + "learning_rate": 4.5791772843101566e-07, + "loss": 1.816, + "step": 46604 + }, + { + "epoch": 8.48, + "learning_rate": 4.5786312340735344e-07, + "loss": 0.5152, + "step": 46605 + }, + { + "epoch": 8.48, + "learning_rate": 4.5780851838369133e-07, + "loss": 0.4911, + "step": 46606 + }, + { + "epoch": 8.48, + "learning_rate": 4.5775391336002917e-07, + "loss": 0.1855, + "step": 46607 + }, + { + "epoch": 8.48, + "learning_rate": 4.5769930833636695e-07, + "loss": 1.5032, + "step": 46608 + }, + { + "epoch": 8.48, + "learning_rate": 4.576447033127048e-07, + "loss": 0.0119, + "step": 46609 + }, + { + "epoch": 8.48, + "learning_rate": 4.575900982890426e-07, + "loss": 0.0086, + "step": 46610 + }, + { + "epoch": 8.48, + "learning_rate": 4.575354932653804e-07, + "loss": 0.0845, + "step": 46611 + }, + { + "epoch": 8.48, + "learning_rate": 4.574808882417182e-07, + "loss": 0.3371, + "step": 46612 + }, + { + "epoch": 8.48, + "learning_rate": 4.574262832180561e-07, + "loss": 0.0069, + "step": 46613 + }, + { + "epoch": 8.48, + "learning_rate": 4.5737167819439393e-07, + "loss": 0.788, + "step": 46614 + }, + { + "epoch": 8.48, + "learning_rate": 4.573170731707317e-07, + "loss": 1.2942, + "step": 46615 + }, + { + "epoch": 8.48, + "learning_rate": 4.5726246814706955e-07, + "loss": 0.5558, + "step": 46616 + }, + { + "epoch": 8.48, + "learning_rate": 4.5720786312340733e-07, + "loss": 0.4979, + "step": 46617 + }, + { + "epoch": 8.48, + "learning_rate": 4.571532580997452e-07, + "loss": 0.4128, + "step": 46618 + }, + { + "epoch": 8.48, + "learning_rate": 4.5709865307608296e-07, + "loss": 1.5242, + "step": 46619 + }, + { + "epoch": 8.48, + "learning_rate": 4.5704404805242085e-07, + "loss": 1.764, + "step": 46620 + }, + { + "epoch": 8.48, + "learning_rate": 4.569894430287587e-07, + "loss": 0.6685, + "step": 46621 + }, + { + "epoch": 8.48, + "learning_rate": 4.5693483800509647e-07, + "loss": 0.0042, + "step": 46622 + }, + { + "epoch": 8.48, + "learning_rate": 4.568802329814343e-07, + "loss": 0.6467, + "step": 46623 + }, + { + "epoch": 8.48, + "learning_rate": 4.568256279577721e-07, + "loss": 1.0502, + "step": 46624 + }, + { + "epoch": 8.48, + "learning_rate": 4.5677102293411e-07, + "loss": 1.8992, + "step": 46625 + }, + { + "epoch": 8.48, + "learning_rate": 4.5671641791044777e-07, + "loss": 0.0157, + "step": 46626 + }, + { + "epoch": 8.48, + "learning_rate": 4.566618128867856e-07, + "loss": 1.5031, + "step": 46627 + }, + { + "epoch": 8.48, + "learning_rate": 4.5660720786312344e-07, + "loss": 0.0079, + "step": 46628 + }, + { + "epoch": 8.48, + "learning_rate": 4.565526028394612e-07, + "loss": 0.0074, + "step": 46629 + }, + { + "epoch": 8.48, + "learning_rate": 4.564979978157991e-07, + "loss": 0.0173, + "step": 46630 + }, + { + "epoch": 8.48, + "learning_rate": 4.5644339279213685e-07, + "loss": 0.4643, + "step": 46631 + }, + { + "epoch": 8.48, + "learning_rate": 4.5638878776847474e-07, + "loss": 0.1998, + "step": 46632 + }, + { + "epoch": 8.48, + "learning_rate": 4.563341827448125e-07, + "loss": 0.6428, + "step": 46633 + }, + { + "epoch": 8.48, + "learning_rate": 4.5627957772115036e-07, + "loss": 0.0867, + "step": 46634 + }, + { + "epoch": 8.48, + "learning_rate": 4.5622497269748814e-07, + "loss": 0.2416, + "step": 46635 + }, + { + "epoch": 8.48, + "learning_rate": 4.56170367673826e-07, + "loss": 0.8119, + "step": 46636 + }, + { + "epoch": 8.48, + "learning_rate": 4.5611576265016387e-07, + "loss": 1.7289, + "step": 46637 + }, + { + "epoch": 8.48, + "learning_rate": 4.5606115762650166e-07, + "loss": 0.2353, + "step": 46638 + }, + { + "epoch": 8.48, + "learning_rate": 4.560065526028395e-07, + "loss": 1.1561, + "step": 46639 + }, + { + "epoch": 8.48, + "learning_rate": 4.559519475791773e-07, + "loss": 1.4538, + "step": 46640 + }, + { + "epoch": 8.48, + "learning_rate": 4.558973425555151e-07, + "loss": 0.1644, + "step": 46641 + }, + { + "epoch": 8.48, + "learning_rate": 4.558427375318529e-07, + "loss": 0.4023, + "step": 46642 + }, + { + "epoch": 8.48, + "learning_rate": 4.5578813250819074e-07, + "loss": 0.7349, + "step": 46643 + }, + { + "epoch": 8.48, + "learning_rate": 4.5573352748452863e-07, + "loss": 0.2357, + "step": 46644 + }, + { + "epoch": 8.48, + "learning_rate": 4.556789224608664e-07, + "loss": 0.0175, + "step": 46645 + }, + { + "epoch": 8.48, + "learning_rate": 4.5562431743720425e-07, + "loss": 0.041, + "step": 46646 + }, + { + "epoch": 8.48, + "learning_rate": 4.5556971241354204e-07, + "loss": 0.3186, + "step": 46647 + }, + { + "epoch": 8.48, + "learning_rate": 4.5551510738987987e-07, + "loss": 1.4389, + "step": 46648 + }, + { + "epoch": 8.48, + "learning_rate": 4.5546050236621766e-07, + "loss": 2.4223, + "step": 46649 + }, + { + "epoch": 8.48, + "learning_rate": 4.5540589734255555e-07, + "loss": 0.0217, + "step": 46650 + }, + { + "epoch": 8.48, + "learning_rate": 4.553512923188934e-07, + "loss": 0.3853, + "step": 46651 + }, + { + "epoch": 8.48, + "learning_rate": 4.5529668729523117e-07, + "loss": 1.9667, + "step": 46652 + }, + { + "epoch": 8.48, + "learning_rate": 4.55242082271569e-07, + "loss": 0.0126, + "step": 46653 + }, + { + "epoch": 8.48, + "learning_rate": 4.551874772479068e-07, + "loss": 0.1528, + "step": 46654 + }, + { + "epoch": 8.48, + "learning_rate": 4.551328722242447e-07, + "loss": 0.4723, + "step": 46655 + }, + { + "epoch": 8.48, + "learning_rate": 4.550782672005824e-07, + "loss": 0.1667, + "step": 46656 + }, + { + "epoch": 8.48, + "learning_rate": 4.550236621769203e-07, + "loss": 0.3662, + "step": 46657 + }, + { + "epoch": 8.48, + "learning_rate": 4.5496905715325814e-07, + "loss": 0.7663, + "step": 46658 + }, + { + "epoch": 8.48, + "learning_rate": 4.5491445212959593e-07, + "loss": 0.7581, + "step": 46659 + }, + { + "epoch": 8.49, + "learning_rate": 4.5485984710593376e-07, + "loss": 0.0182, + "step": 46660 + }, + { + "epoch": 8.49, + "learning_rate": 4.5480524208227155e-07, + "loss": 0.7159, + "step": 46661 + }, + { + "epoch": 8.49, + "learning_rate": 4.5475063705860944e-07, + "loss": 0.6247, + "step": 46662 + }, + { + "epoch": 8.49, + "learning_rate": 4.5469603203494717e-07, + "loss": 0.258, + "step": 46663 + }, + { + "epoch": 8.49, + "learning_rate": 4.5464142701128506e-07, + "loss": 0.6009, + "step": 46664 + }, + { + "epoch": 8.49, + "learning_rate": 4.545868219876229e-07, + "loss": 1.5253, + "step": 46665 + }, + { + "epoch": 8.49, + "learning_rate": 4.545322169639607e-07, + "loss": 0.055, + "step": 46666 + }, + { + "epoch": 8.49, + "learning_rate": 4.544776119402986e-07, + "loss": 0.4079, + "step": 46667 + }, + { + "epoch": 8.49, + "learning_rate": 4.544230069166363e-07, + "loss": 1.0863, + "step": 46668 + }, + { + "epoch": 8.49, + "learning_rate": 4.543684018929742e-07, + "loss": 0.1919, + "step": 46669 + }, + { + "epoch": 8.49, + "learning_rate": 4.54313796869312e-07, + "loss": 0.306, + "step": 46670 + }, + { + "epoch": 8.49, + "learning_rate": 4.542591918456498e-07, + "loss": 0.4374, + "step": 46671 + }, + { + "epoch": 8.49, + "learning_rate": 4.542045868219876e-07, + "loss": 1.0346, + "step": 46672 + }, + { + "epoch": 8.49, + "learning_rate": 4.5414998179832544e-07, + "loss": 0.234, + "step": 46673 + }, + { + "epoch": 8.49, + "learning_rate": 4.5409537677466333e-07, + "loss": 0.4174, + "step": 46674 + }, + { + "epoch": 8.49, + "learning_rate": 4.5404077175100106e-07, + "loss": 6.1367, + "step": 46675 + }, + { + "epoch": 8.49, + "learning_rate": 4.5398616672733895e-07, + "loss": 0.2331, + "step": 46676 + }, + { + "epoch": 8.49, + "learning_rate": 4.5393156170367674e-07, + "loss": 0.9667, + "step": 46677 + }, + { + "epoch": 8.49, + "learning_rate": 4.538769566800146e-07, + "loss": 0.0301, + "step": 46678 + }, + { + "epoch": 8.49, + "learning_rate": 4.5382235165635236e-07, + "loss": 2.217, + "step": 46679 + }, + { + "epoch": 8.49, + "learning_rate": 4.537677466326902e-07, + "loss": 0.0305, + "step": 46680 + }, + { + "epoch": 8.49, + "learning_rate": 4.537131416090281e-07, + "loss": 0.0068, + "step": 46681 + }, + { + "epoch": 8.49, + "learning_rate": 4.5365853658536587e-07, + "loss": 0.6006, + "step": 46682 + }, + { + "epoch": 8.49, + "learning_rate": 4.536039315617037e-07, + "loss": 0.0057, + "step": 46683 + }, + { + "epoch": 8.49, + "learning_rate": 4.535493265380415e-07, + "loss": 0.0082, + "step": 46684 + }, + { + "epoch": 8.49, + "learning_rate": 4.5349472151437933e-07, + "loss": 0.58, + "step": 46685 + }, + { + "epoch": 8.49, + "learning_rate": 4.534401164907171e-07, + "loss": 0.0392, + "step": 46686 + }, + { + "epoch": 8.49, + "learning_rate": 4.53385511467055e-07, + "loss": 0.0472, + "step": 46687 + }, + { + "epoch": 8.49, + "learning_rate": 4.5333090644339284e-07, + "loss": 0.7369, + "step": 46688 + }, + { + "epoch": 8.49, + "learning_rate": 4.5327630141973063e-07, + "loss": 0.0091, + "step": 46689 + }, + { + "epoch": 8.49, + "learning_rate": 4.5322169639606847e-07, + "loss": 2.2459, + "step": 46690 + }, + { + "epoch": 8.49, + "learning_rate": 4.5316709137240625e-07, + "loss": 1.2612, + "step": 46691 + }, + { + "epoch": 8.49, + "learning_rate": 4.531124863487441e-07, + "loss": 1.2077, + "step": 46692 + }, + { + "epoch": 8.49, + "learning_rate": 4.5305788132508187e-07, + "loss": 0.1149, + "step": 46693 + }, + { + "epoch": 8.49, + "learning_rate": 4.5300327630141976e-07, + "loss": 0.1796, + "step": 46694 + }, + { + "epoch": 8.49, + "learning_rate": 4.529486712777576e-07, + "loss": 0.2866, + "step": 46695 + }, + { + "epoch": 8.49, + "learning_rate": 4.528940662540954e-07, + "loss": 0.0928, + "step": 46696 + }, + { + "epoch": 8.49, + "learning_rate": 4.528394612304332e-07, + "loss": 2.8503, + "step": 46697 + }, + { + "epoch": 8.49, + "learning_rate": 4.52784856206771e-07, + "loss": 0.6993, + "step": 46698 + }, + { + "epoch": 8.49, + "learning_rate": 4.527302511831089e-07, + "loss": 0.2122, + "step": 46699 + }, + { + "epoch": 8.49, + "learning_rate": 4.5267564615944663e-07, + "loss": 0.8993, + "step": 46700 + }, + { + "epoch": 8.49, + "learning_rate": 4.526210411357845e-07, + "loss": 0.4245, + "step": 46701 + }, + { + "epoch": 8.49, + "learning_rate": 4.5256643611212236e-07, + "loss": 0.0048, + "step": 46702 + }, + { + "epoch": 8.49, + "learning_rate": 4.5251183108846014e-07, + "loss": 0.0034, + "step": 46703 + }, + { + "epoch": 8.49, + "learning_rate": 4.52457226064798e-07, + "loss": 0.0052, + "step": 46704 + }, + { + "epoch": 8.49, + "learning_rate": 4.5240262104113576e-07, + "loss": 0.995, + "step": 46705 + }, + { + "epoch": 8.49, + "learning_rate": 4.5234801601747365e-07, + "loss": 0.0082, + "step": 46706 + }, + { + "epoch": 8.49, + "learning_rate": 4.522934109938114e-07, + "loss": 0.0223, + "step": 46707 + }, + { + "epoch": 8.49, + "learning_rate": 4.522388059701493e-07, + "loss": 0.0115, + "step": 46708 + }, + { + "epoch": 8.49, + "learning_rate": 4.5218420094648706e-07, + "loss": 0.0054, + "step": 46709 + }, + { + "epoch": 8.49, + "learning_rate": 4.521295959228249e-07, + "loss": 0.022, + "step": 46710 + }, + { + "epoch": 8.49, + "learning_rate": 4.520749908991628e-07, + "loss": 0.0504, + "step": 46711 + }, + { + "epoch": 8.49, + "learning_rate": 4.520203858755005e-07, + "loss": 0.2146, + "step": 46712 + }, + { + "epoch": 8.49, + "learning_rate": 4.519657808518384e-07, + "loss": 0.0131, + "step": 46713 + }, + { + "epoch": 8.49, + "learning_rate": 4.519111758281762e-07, + "loss": 0.1574, + "step": 46714 + }, + { + "epoch": 8.5, + "learning_rate": 4.5185657080451403e-07, + "loss": 1.2251, + "step": 46715 + }, + { + "epoch": 8.5, + "learning_rate": 4.518019657808518e-07, + "loss": 2.0639, + "step": 46716 + }, + { + "epoch": 8.5, + "learning_rate": 4.5174736075718966e-07, + "loss": 0.0072, + "step": 46717 + }, + { + "epoch": 8.5, + "learning_rate": 4.5169275573352755e-07, + "loss": 0.2752, + "step": 46718 + }, + { + "epoch": 8.5, + "learning_rate": 4.5163815070986533e-07, + "loss": 0.3185, + "step": 46719 + }, + { + "epoch": 8.5, + "learning_rate": 4.5158354568620317e-07, + "loss": 0.0194, + "step": 46720 + }, + { + "epoch": 8.5, + "learning_rate": 4.5152894066254095e-07, + "loss": 0.0102, + "step": 46721 + }, + { + "epoch": 8.5, + "learning_rate": 4.514743356388788e-07, + "loss": 0.0948, + "step": 46722 + }, + { + "epoch": 8.5, + "learning_rate": 4.514197306152166e-07, + "loss": 2.1312, + "step": 46723 + }, + { + "epoch": 8.5, + "learning_rate": 4.513651255915544e-07, + "loss": 1.4006, + "step": 46724 + }, + { + "epoch": 8.5, + "learning_rate": 4.513105205678923e-07, + "loss": 0.0058, + "step": 46725 + }, + { + "epoch": 8.5, + "learning_rate": 4.512559155442301e-07, + "loss": 1.5278, + "step": 46726 + }, + { + "epoch": 8.5, + "learning_rate": 4.512013105205679e-07, + "loss": 1.9128, + "step": 46727 + }, + { + "epoch": 8.5, + "learning_rate": 4.511467054969057e-07, + "loss": 0.007, + "step": 46728 + }, + { + "epoch": 8.5, + "learning_rate": 4.5109210047324355e-07, + "loss": 0.1102, + "step": 46729 + }, + { + "epoch": 8.5, + "learning_rate": 4.5103749544958133e-07, + "loss": 4.4243, + "step": 46730 + }, + { + "epoch": 8.5, + "learning_rate": 4.509828904259192e-07, + "loss": 0.5237, + "step": 46731 + }, + { + "epoch": 8.5, + "learning_rate": 4.5092828540225706e-07, + "loss": 0.7463, + "step": 46732 + }, + { + "epoch": 8.5, + "learning_rate": 4.5087368037859484e-07, + "loss": 0.0673, + "step": 46733 + }, + { + "epoch": 8.5, + "learning_rate": 4.508190753549327e-07, + "loss": 0.0186, + "step": 46734 + }, + { + "epoch": 8.5, + "learning_rate": 4.5076447033127047e-07, + "loss": 2.0138, + "step": 46735 + }, + { + "epoch": 8.5, + "learning_rate": 4.507098653076083e-07, + "loss": 0.756, + "step": 46736 + }, + { + "epoch": 8.5, + "learning_rate": 4.506552602839461e-07, + "loss": 1.1027, + "step": 46737 + }, + { + "epoch": 8.5, + "learning_rate": 4.50600655260284e-07, + "loss": 0.0036, + "step": 46738 + }, + { + "epoch": 8.5, + "learning_rate": 4.505460502366218e-07, + "loss": 1.1864, + "step": 46739 + }, + { + "epoch": 8.5, + "learning_rate": 4.504914452129596e-07, + "loss": 0.3239, + "step": 46740 + }, + { + "epoch": 8.5, + "learning_rate": 4.5043684018929744e-07, + "loss": 0.5757, + "step": 46741 + }, + { + "epoch": 8.5, + "learning_rate": 4.503822351656352e-07, + "loss": 0.0455, + "step": 46742 + }, + { + "epoch": 8.5, + "learning_rate": 4.503276301419731e-07, + "loss": 0.8341, + "step": 46743 + }, + { + "epoch": 8.5, + "learning_rate": 4.5027302511831084e-07, + "loss": 0.0053, + "step": 46744 + }, + { + "epoch": 8.5, + "learning_rate": 4.5021842009464874e-07, + "loss": 0.0068, + "step": 46745 + }, + { + "epoch": 8.5, + "learning_rate": 4.501638150709865e-07, + "loss": 1.7236, + "step": 46746 + }, + { + "epoch": 8.5, + "learning_rate": 4.5010921004732436e-07, + "loss": 0.1198, + "step": 46747 + }, + { + "epoch": 8.5, + "learning_rate": 4.500546050236622e-07, + "loss": 0.5317, + "step": 46748 + }, + { + "epoch": 8.5, + "learning_rate": 4.5e-07, + "loss": 0.0099, + "step": 46749 + }, + { + "epoch": 8.5, + "learning_rate": 4.4994539497633787e-07, + "loss": 0.1188, + "step": 46750 + }, + { + "epoch": 8.5, + "learning_rate": 4.4989078995267565e-07, + "loss": 0.0191, + "step": 46751 + }, + { + "epoch": 8.5, + "learning_rate": 4.498361849290135e-07, + "loss": 0.0177, + "step": 46752 + }, + { + "epoch": 8.5, + "learning_rate": 4.497815799053513e-07, + "loss": 0.0182, + "step": 46753 + }, + { + "epoch": 8.5, + "learning_rate": 4.497269748816891e-07, + "loss": 0.6677, + "step": 46754 + }, + { + "epoch": 8.5, + "learning_rate": 4.49672369858027e-07, + "loss": 1.1742, + "step": 46755 + }, + { + "epoch": 8.5, + "learning_rate": 4.4961776483436474e-07, + "loss": 0.0189, + "step": 46756 + }, + { + "epoch": 8.5, + "learning_rate": 4.4956315981070263e-07, + "loss": 0.0122, + "step": 46757 + }, + { + "epoch": 8.5, + "learning_rate": 4.495085547870404e-07, + "loss": 0.0939, + "step": 46758 + }, + { + "epoch": 8.5, + "learning_rate": 4.4945394976337825e-07, + "loss": 0.8459, + "step": 46759 + }, + { + "epoch": 8.5, + "learning_rate": 4.4939934473971603e-07, + "loss": 0.0933, + "step": 46760 + }, + { + "epoch": 8.5, + "learning_rate": 4.4934473971605387e-07, + "loss": 0.0142, + "step": 46761 + }, + { + "epoch": 8.5, + "learning_rate": 4.4929013469239176e-07, + "loss": 0.2986, + "step": 46762 + }, + { + "epoch": 8.5, + "learning_rate": 4.4923552966872955e-07, + "loss": 0.0155, + "step": 46763 + }, + { + "epoch": 8.5, + "learning_rate": 4.491809246450674e-07, + "loss": 0.9128, + "step": 46764 + }, + { + "epoch": 8.5, + "learning_rate": 4.4912631962140517e-07, + "loss": 0.029, + "step": 46765 + }, + { + "epoch": 8.5, + "learning_rate": 4.49071714597743e-07, + "loss": 0.4205, + "step": 46766 + }, + { + "epoch": 8.5, + "learning_rate": 4.490171095740808e-07, + "loss": 0.0043, + "step": 46767 + }, + { + "epoch": 8.5, + "learning_rate": 4.4896250455041863e-07, + "loss": 0.3092, + "step": 46768 + }, + { + "epoch": 8.51, + "learning_rate": 4.489078995267565e-07, + "loss": 1.4613, + "step": 46769 + }, + { + "epoch": 8.51, + "learning_rate": 4.488532945030943e-07, + "loss": 0.0738, + "step": 46770 + }, + { + "epoch": 8.51, + "learning_rate": 4.4879868947943214e-07, + "loss": 0.0414, + "step": 46771 + }, + { + "epoch": 8.51, + "learning_rate": 4.487440844557699e-07, + "loss": 0.1461, + "step": 46772 + }, + { + "epoch": 8.51, + "learning_rate": 4.4868947943210776e-07, + "loss": 1.168, + "step": 46773 + }, + { + "epoch": 8.51, + "learning_rate": 4.4863487440844555e-07, + "loss": 0.0045, + "step": 46774 + }, + { + "epoch": 8.51, + "learning_rate": 4.4858026938478344e-07, + "loss": 0.0088, + "step": 46775 + }, + { + "epoch": 8.51, + "learning_rate": 4.485256643611213e-07, + "loss": 0.65, + "step": 46776 + }, + { + "epoch": 8.51, + "learning_rate": 4.4847105933745906e-07, + "loss": 0.1005, + "step": 46777 + }, + { + "epoch": 8.51, + "learning_rate": 4.484164543137969e-07, + "loss": 1.1368, + "step": 46778 + }, + { + "epoch": 8.51, + "learning_rate": 4.483618492901347e-07, + "loss": 0.0555, + "step": 46779 + }, + { + "epoch": 8.51, + "learning_rate": 4.483072442664725e-07, + "loss": 1.4329, + "step": 46780 + }, + { + "epoch": 8.51, + "learning_rate": 4.482526392428103e-07, + "loss": 2.9325, + "step": 46781 + }, + { + "epoch": 8.51, + "learning_rate": 4.481980342191482e-07, + "loss": 1.2786, + "step": 46782 + }, + { + "epoch": 8.51, + "learning_rate": 4.4814342919548603e-07, + "loss": 0.0063, + "step": 46783 + }, + { + "epoch": 8.51, + "learning_rate": 4.480888241718238e-07, + "loss": 0.2896, + "step": 46784 + }, + { + "epoch": 8.51, + "learning_rate": 4.4803421914816165e-07, + "loss": 1.037, + "step": 46785 + }, + { + "epoch": 8.51, + "learning_rate": 4.4797961412449944e-07, + "loss": 0.1372, + "step": 46786 + }, + { + "epoch": 8.51, + "learning_rate": 4.4792500910083733e-07, + "loss": 0.3882, + "step": 46787 + }, + { + "epoch": 8.51, + "learning_rate": 4.4787040407717506e-07, + "loss": 0.0063, + "step": 46788 + }, + { + "epoch": 8.51, + "learning_rate": 4.4781579905351295e-07, + "loss": 0.0052, + "step": 46789 + }, + { + "epoch": 8.51, + "learning_rate": 4.4776119402985074e-07, + "loss": 0.0057, + "step": 46790 + }, + { + "epoch": 8.51, + "learning_rate": 4.4770658900618857e-07, + "loss": 0.2635, + "step": 46791 + }, + { + "epoch": 8.51, + "learning_rate": 4.476519839825264e-07, + "loss": 0.0826, + "step": 46792 + }, + { + "epoch": 8.51, + "learning_rate": 4.475973789588642e-07, + "loss": 1.3869, + "step": 46793 + }, + { + "epoch": 8.51, + "learning_rate": 4.475427739352021e-07, + "loss": 1.714, + "step": 46794 + }, + { + "epoch": 8.51, + "learning_rate": 4.4748816891153987e-07, + "loss": 0.0731, + "step": 46795 + }, + { + "epoch": 8.51, + "learning_rate": 4.474335638878777e-07, + "loss": 1.2195, + "step": 46796 + }, + { + "epoch": 8.51, + "learning_rate": 4.473789588642155e-07, + "loss": 0.411, + "step": 46797 + }, + { + "epoch": 8.51, + "learning_rate": 4.4732435384055333e-07, + "loss": 0.0434, + "step": 46798 + }, + { + "epoch": 8.51, + "learning_rate": 4.472697488168912e-07, + "loss": 0.0041, + "step": 46799 + }, + { + "epoch": 8.51, + "learning_rate": 4.4721514379322895e-07, + "loss": 0.0041, + "step": 46800 + }, + { + "epoch": 8.51, + "learning_rate": 4.4716053876956684e-07, + "loss": 0.0183, + "step": 46801 + }, + { + "epoch": 8.51, + "learning_rate": 4.4710593374590463e-07, + "loss": 0.4084, + "step": 46802 + }, + { + "epoch": 8.51, + "learning_rate": 4.4705132872224246e-07, + "loss": 0.2574, + "step": 46803 + }, + { + "epoch": 8.51, + "learning_rate": 4.4699672369858025e-07, + "loss": 0.0881, + "step": 46804 + }, + { + "epoch": 8.51, + "learning_rate": 4.469421186749181e-07, + "loss": 0.3629, + "step": 46805 + }, + { + "epoch": 8.51, + "learning_rate": 4.46887513651256e-07, + "loss": 0.2532, + "step": 46806 + }, + { + "epoch": 8.51, + "learning_rate": 4.4683290862759376e-07, + "loss": 0.0145, + "step": 46807 + }, + { + "epoch": 8.51, + "learning_rate": 4.467783036039316e-07, + "loss": 0.773, + "step": 46808 + }, + { + "epoch": 8.51, + "learning_rate": 4.467236985802694e-07, + "loss": 2.0208, + "step": 46809 + }, + { + "epoch": 8.51, + "learning_rate": 4.466690935566072e-07, + "loss": 0.1814, + "step": 46810 + }, + { + "epoch": 8.51, + "learning_rate": 4.46614488532945e-07, + "loss": 0.8258, + "step": 46811 + }, + { + "epoch": 8.51, + "learning_rate": 4.4655988350928284e-07, + "loss": 0.3103, + "step": 46812 + }, + { + "epoch": 8.51, + "learning_rate": 4.4650527848562073e-07, + "loss": 0.2282, + "step": 46813 + }, + { + "epoch": 8.51, + "learning_rate": 4.464506734619585e-07, + "loss": 0.7652, + "step": 46814 + }, + { + "epoch": 8.51, + "learning_rate": 4.4639606843829636e-07, + "loss": 0.0134, + "step": 46815 + }, + { + "epoch": 8.51, + "learning_rate": 4.4634146341463414e-07, + "loss": 0.0631, + "step": 46816 + }, + { + "epoch": 8.51, + "learning_rate": 4.46286858390972e-07, + "loss": 0.0198, + "step": 46817 + }, + { + "epoch": 8.51, + "learning_rate": 4.4623225336730976e-07, + "loss": 1.2346, + "step": 46818 + }, + { + "epoch": 8.51, + "learning_rate": 4.4617764834364765e-07, + "loss": 2.2033, + "step": 46819 + }, + { + "epoch": 8.51, + "learning_rate": 4.461230433199855e-07, + "loss": 0.1143, + "step": 46820 + }, + { + "epoch": 8.51, + "learning_rate": 4.460684382963233e-07, + "loss": 2.4869, + "step": 46821 + }, + { + "epoch": 8.51, + "learning_rate": 4.460138332726611e-07, + "loss": 0.161, + "step": 46822 + }, + { + "epoch": 8.51, + "learning_rate": 4.459592282489989e-07, + "loss": 0.236, + "step": 46823 + }, + { + "epoch": 8.52, + "learning_rate": 4.4590462322533673e-07, + "loss": 0.1275, + "step": 46824 + }, + { + "epoch": 8.52, + "learning_rate": 4.458500182016745e-07, + "loss": 0.9877, + "step": 46825 + }, + { + "epoch": 8.52, + "learning_rate": 4.457954131780124e-07, + "loss": 0.1148, + "step": 46826 + }, + { + "epoch": 8.52, + "learning_rate": 4.457408081543502e-07, + "loss": 0.0635, + "step": 46827 + }, + { + "epoch": 8.52, + "learning_rate": 4.4568620313068803e-07, + "loss": 0.6243, + "step": 46828 + }, + { + "epoch": 8.52, + "learning_rate": 4.4563159810702587e-07, + "loss": 0.3692, + "step": 46829 + }, + { + "epoch": 8.52, + "learning_rate": 4.4557699308336365e-07, + "loss": 4.4078, + "step": 46830 + }, + { + "epoch": 8.52, + "learning_rate": 4.4552238805970154e-07, + "loss": 0.006, + "step": 46831 + }, + { + "epoch": 8.52, + "learning_rate": 4.454677830360393e-07, + "loss": 0.036, + "step": 46832 + }, + { + "epoch": 8.52, + "learning_rate": 4.4541317801237717e-07, + "loss": 1.0713, + "step": 46833 + }, + { + "epoch": 8.52, + "learning_rate": 4.4535857298871495e-07, + "loss": 1.4519, + "step": 46834 + }, + { + "epoch": 8.52, + "learning_rate": 4.453039679650528e-07, + "loss": 1.2199, + "step": 46835 + }, + { + "epoch": 8.52, + "learning_rate": 4.452493629413907e-07, + "loss": 4.2903, + "step": 46836 + }, + { + "epoch": 8.52, + "learning_rate": 4.451947579177284e-07, + "loss": 0.0028, + "step": 46837 + }, + { + "epoch": 8.52, + "learning_rate": 4.451401528940663e-07, + "loss": 0.0329, + "step": 46838 + }, + { + "epoch": 8.52, + "learning_rate": 4.450855478704041e-07, + "loss": 0.1148, + "step": 46839 + }, + { + "epoch": 8.52, + "learning_rate": 4.450309428467419e-07, + "loss": 0.0095, + "step": 46840 + }, + { + "epoch": 8.52, + "learning_rate": 4.449763378230797e-07, + "loss": 1.2033, + "step": 46841 + }, + { + "epoch": 8.52, + "learning_rate": 4.4492173279941755e-07, + "loss": 0.0816, + "step": 46842 + }, + { + "epoch": 8.52, + "learning_rate": 4.4486712777575544e-07, + "loss": 0.0674, + "step": 46843 + }, + { + "epoch": 8.52, + "learning_rate": 4.4481252275209317e-07, + "loss": 0.1808, + "step": 46844 + }, + { + "epoch": 8.52, + "learning_rate": 4.4475791772843106e-07, + "loss": 0.5749, + "step": 46845 + }, + { + "epoch": 8.52, + "learning_rate": 4.4470331270476884e-07, + "loss": 0.7074, + "step": 46846 + }, + { + "epoch": 8.52, + "learning_rate": 4.446487076811067e-07, + "loss": 1.6253, + "step": 46847 + }, + { + "epoch": 8.52, + "learning_rate": 4.4459410265744446e-07, + "loss": 0.0634, + "step": 46848 + }, + { + "epoch": 8.52, + "learning_rate": 4.445394976337823e-07, + "loss": 0.0118, + "step": 46849 + }, + { + "epoch": 8.52, + "learning_rate": 4.444848926101202e-07, + "loss": 0.1226, + "step": 46850 + }, + { + "epoch": 8.52, + "learning_rate": 4.44430287586458e-07, + "loss": 0.0505, + "step": 46851 + }, + { + "epoch": 8.52, + "learning_rate": 4.443756825627958e-07, + "loss": 1.7159, + "step": 46852 + }, + { + "epoch": 8.52, + "learning_rate": 4.443210775391336e-07, + "loss": 0.3013, + "step": 46853 + }, + { + "epoch": 8.52, + "learning_rate": 4.4426647251547144e-07, + "loss": 0.2145, + "step": 46854 + }, + { + "epoch": 8.52, + "learning_rate": 4.442118674918092e-07, + "loss": 0.3842, + "step": 46855 + }, + { + "epoch": 8.52, + "learning_rate": 4.4415726246814706e-07, + "loss": 0.0139, + "step": 46856 + }, + { + "epoch": 8.52, + "learning_rate": 4.4410265744448495e-07, + "loss": 0.035, + "step": 46857 + }, + { + "epoch": 8.52, + "learning_rate": 4.4404805242082273e-07, + "loss": 0.0083, + "step": 46858 + }, + { + "epoch": 8.52, + "learning_rate": 4.4399344739716057e-07, + "loss": 0.0537, + "step": 46859 + }, + { + "epoch": 8.52, + "learning_rate": 4.4393884237349836e-07, + "loss": 0.0122, + "step": 46860 + }, + { + "epoch": 8.52, + "learning_rate": 4.438842373498362e-07, + "loss": 0.3961, + "step": 46861 + }, + { + "epoch": 8.52, + "learning_rate": 4.43829632326174e-07, + "loss": 0.1424, + "step": 46862 + }, + { + "epoch": 8.52, + "learning_rate": 4.4377502730251187e-07, + "loss": 1.0692, + "step": 46863 + }, + { + "epoch": 8.52, + "learning_rate": 4.437204222788496e-07, + "loss": 0.4128, + "step": 46864 + }, + { + "epoch": 8.52, + "learning_rate": 4.436658172551875e-07, + "loss": 1.1798, + "step": 46865 + }, + { + "epoch": 8.52, + "learning_rate": 4.4361121223152533e-07, + "loss": 0.0648, + "step": 46866 + }, + { + "epoch": 8.52, + "learning_rate": 4.435566072078631e-07, + "loss": 0.0232, + "step": 46867 + }, + { + "epoch": 8.52, + "learning_rate": 4.43502002184201e-07, + "loss": 0.1419, + "step": 46868 + }, + { + "epoch": 8.52, + "learning_rate": 4.4344739716053873e-07, + "loss": 0.0199, + "step": 46869 + }, + { + "epoch": 8.52, + "learning_rate": 4.433927921368766e-07, + "loss": 2.1529, + "step": 46870 + }, + { + "epoch": 8.52, + "learning_rate": 4.433381871132144e-07, + "loss": 0.032, + "step": 46871 + }, + { + "epoch": 8.52, + "learning_rate": 4.4328358208955225e-07, + "loss": 0.5996, + "step": 46872 + }, + { + "epoch": 8.52, + "learning_rate": 4.432289770658901e-07, + "loss": 0.7407, + "step": 46873 + }, + { + "epoch": 8.52, + "learning_rate": 4.4317437204222787e-07, + "loss": 1.5104, + "step": 46874 + }, + { + "epoch": 8.52, + "learning_rate": 4.4311976701856576e-07, + "loss": 0.0978, + "step": 46875 + }, + { + "epoch": 8.52, + "learning_rate": 4.430651619949035e-07, + "loss": 0.0977, + "step": 46876 + }, + { + "epoch": 8.52, + "learning_rate": 4.430105569712414e-07, + "loss": 0.8371, + "step": 46877 + }, + { + "epoch": 8.52, + "learning_rate": 4.4295595194757917e-07, + "loss": 0.6836, + "step": 46878 + }, + { + "epoch": 8.53, + "learning_rate": 4.42901346923917e-07, + "loss": 0.1626, + "step": 46879 + }, + { + "epoch": 8.53, + "learning_rate": 4.428467419002549e-07, + "loss": 3.0215, + "step": 46880 + }, + { + "epoch": 8.53, + "learning_rate": 4.427921368765926e-07, + "loss": 0.5976, + "step": 46881 + }, + { + "epoch": 8.53, + "learning_rate": 4.427375318529305e-07, + "loss": 0.0184, + "step": 46882 + }, + { + "epoch": 8.53, + "learning_rate": 4.426829268292683e-07, + "loss": 0.2662, + "step": 46883 + }, + { + "epoch": 8.53, + "learning_rate": 4.4262832180560614e-07, + "loss": 0.9959, + "step": 46884 + }, + { + "epoch": 8.53, + "learning_rate": 4.425737167819439e-07, + "loss": 2.1074, + "step": 46885 + }, + { + "epoch": 8.53, + "learning_rate": 4.4251911175828176e-07, + "loss": 1.2355, + "step": 46886 + }, + { + "epoch": 8.53, + "learning_rate": 4.4246450673461965e-07, + "loss": 0.0661, + "step": 46887 + }, + { + "epoch": 8.53, + "learning_rate": 4.4240990171095744e-07, + "loss": 0.5117, + "step": 46888 + }, + { + "epoch": 8.53, + "learning_rate": 4.4235529668729527e-07, + "loss": 0.6479, + "step": 46889 + }, + { + "epoch": 8.53, + "learning_rate": 4.4230069166363306e-07, + "loss": 0.0645, + "step": 46890 + }, + { + "epoch": 8.53, + "learning_rate": 4.422460866399709e-07, + "loss": 0.7199, + "step": 46891 + }, + { + "epoch": 8.53, + "learning_rate": 4.421914816163087e-07, + "loss": 2.4099, + "step": 46892 + }, + { + "epoch": 8.53, + "learning_rate": 4.421368765926465e-07, + "loss": 2.5535, + "step": 46893 + }, + { + "epoch": 8.53, + "learning_rate": 4.420822715689844e-07, + "loss": 2.0782, + "step": 46894 + }, + { + "epoch": 8.53, + "learning_rate": 4.420276665453222e-07, + "loss": 0.0038, + "step": 46895 + }, + { + "epoch": 8.53, + "learning_rate": 4.4197306152166003e-07, + "loss": 0.8698, + "step": 46896 + }, + { + "epoch": 8.53, + "learning_rate": 4.419184564979978e-07, + "loss": 0.6901, + "step": 46897 + }, + { + "epoch": 8.53, + "learning_rate": 4.4186385147433565e-07, + "loss": 1.4883, + "step": 46898 + }, + { + "epoch": 8.53, + "learning_rate": 4.4180924645067344e-07, + "loss": 2.4377, + "step": 46899 + }, + { + "epoch": 8.53, + "learning_rate": 4.4175464142701133e-07, + "loss": 0.0197, + "step": 46900 + }, + { + "epoch": 8.53, + "learning_rate": 4.4170003640334906e-07, + "loss": 0.771, + "step": 46901 + }, + { + "epoch": 8.53, + "learning_rate": 4.4164543137968695e-07, + "loss": 0.1876, + "step": 46902 + }, + { + "epoch": 8.53, + "learning_rate": 4.415908263560248e-07, + "loss": 1.1802, + "step": 46903 + }, + { + "epoch": 8.53, + "learning_rate": 4.4153622133236257e-07, + "loss": 0.1475, + "step": 46904 + }, + { + "epoch": 8.53, + "learning_rate": 4.414816163087004e-07, + "loss": 0.016, + "step": 46905 + }, + { + "epoch": 8.53, + "learning_rate": 4.414270112850382e-07, + "loss": 0.4973, + "step": 46906 + }, + { + "epoch": 8.53, + "learning_rate": 4.413724062613761e-07, + "loss": 0.3023, + "step": 46907 + }, + { + "epoch": 8.53, + "learning_rate": 4.413178012377138e-07, + "loss": 1.974, + "step": 46908 + }, + { + "epoch": 8.53, + "learning_rate": 4.412631962140517e-07, + "loss": 0.3838, + "step": 46909 + }, + { + "epoch": 8.53, + "learning_rate": 4.4120859119038954e-07, + "loss": 0.2078, + "step": 46910 + }, + { + "epoch": 8.53, + "learning_rate": 4.4115398616672733e-07, + "loss": 1.0865, + "step": 46911 + }, + { + "epoch": 8.53, + "learning_rate": 4.410993811430652e-07, + "loss": 0.1955, + "step": 46912 + }, + { + "epoch": 8.53, + "learning_rate": 4.4104477611940295e-07, + "loss": 1.0304, + "step": 46913 + }, + { + "epoch": 8.53, + "learning_rate": 4.4099017109574084e-07, + "loss": 0.0052, + "step": 46914 + }, + { + "epoch": 8.53, + "learning_rate": 4.409355660720786e-07, + "loss": 0.2141, + "step": 46915 + }, + { + "epoch": 8.53, + "learning_rate": 4.4088096104841646e-07, + "loss": 0.0045, + "step": 46916 + }, + { + "epoch": 8.53, + "learning_rate": 4.408263560247543e-07, + "loss": 0.0496, + "step": 46917 + }, + { + "epoch": 8.53, + "learning_rate": 4.407717510010921e-07, + "loss": 0.0484, + "step": 46918 + }, + { + "epoch": 8.53, + "learning_rate": 4.4071714597743e-07, + "loss": 0.125, + "step": 46919 + }, + { + "epoch": 8.53, + "learning_rate": 4.4066254095376776e-07, + "loss": 0.1515, + "step": 46920 + }, + { + "epoch": 8.53, + "learning_rate": 4.406079359301056e-07, + "loss": 1.0992, + "step": 46921 + }, + { + "epoch": 8.53, + "learning_rate": 4.405533309064434e-07, + "loss": 0.2013, + "step": 46922 + }, + { + "epoch": 8.53, + "learning_rate": 4.404987258827812e-07, + "loss": 0.0574, + "step": 46923 + }, + { + "epoch": 8.53, + "learning_rate": 4.404441208591191e-07, + "loss": 0.434, + "step": 46924 + }, + { + "epoch": 8.53, + "learning_rate": 4.4038951583545684e-07, + "loss": 2.4729, + "step": 46925 + }, + { + "epoch": 8.53, + "learning_rate": 4.4033491081179473e-07, + "loss": 0.0874, + "step": 46926 + }, + { + "epoch": 8.53, + "learning_rate": 4.402803057881325e-07, + "loss": 0.0989, + "step": 46927 + }, + { + "epoch": 8.53, + "learning_rate": 4.4022570076447035e-07, + "loss": 0.0907, + "step": 46928 + }, + { + "epoch": 8.53, + "learning_rate": 4.4017109574080814e-07, + "loss": 0.7006, + "step": 46929 + }, + { + "epoch": 8.53, + "learning_rate": 4.40116490717146e-07, + "loss": 1.9539, + "step": 46930 + }, + { + "epoch": 8.53, + "learning_rate": 4.4006188569348387e-07, + "loss": 0.0527, + "step": 46931 + }, + { + "epoch": 8.53, + "learning_rate": 4.4000728066982165e-07, + "loss": 0.8444, + "step": 46932 + }, + { + "epoch": 8.53, + "learning_rate": 4.399526756461595e-07, + "loss": 0.0485, + "step": 46933 + }, + { + "epoch": 8.54, + "learning_rate": 4.3989807062249727e-07, + "loss": 0.7437, + "step": 46934 + }, + { + "epoch": 8.54, + "learning_rate": 4.398434655988351e-07, + "loss": 1.3561, + "step": 46935 + }, + { + "epoch": 8.54, + "learning_rate": 4.397888605751729e-07, + "loss": 1.5296, + "step": 46936 + }, + { + "epoch": 8.54, + "learning_rate": 4.3973425555151073e-07, + "loss": 1.3357, + "step": 46937 + }, + { + "epoch": 8.54, + "learning_rate": 4.396796505278486e-07, + "loss": 0.1068, + "step": 46938 + }, + { + "epoch": 8.54, + "learning_rate": 4.396250455041864e-07, + "loss": 1.3324, + "step": 46939 + }, + { + "epoch": 8.54, + "learning_rate": 4.3957044048052425e-07, + "loss": 0.8763, + "step": 46940 + }, + { + "epoch": 8.54, + "learning_rate": 4.3951583545686203e-07, + "loss": 0.0499, + "step": 46941 + }, + { + "epoch": 8.54, + "learning_rate": 4.3946123043319987e-07, + "loss": 0.0141, + "step": 46942 + }, + { + "epoch": 8.54, + "learning_rate": 4.3940662540953765e-07, + "loss": 0.0627, + "step": 46943 + }, + { + "epoch": 8.54, + "learning_rate": 4.3935202038587554e-07, + "loss": 2.2895, + "step": 46944 + }, + { + "epoch": 8.54, + "learning_rate": 4.392974153622133e-07, + "loss": 0.1427, + "step": 46945 + }, + { + "epoch": 8.54, + "learning_rate": 4.3924281033855116e-07, + "loss": 0.2898, + "step": 46946 + }, + { + "epoch": 8.54, + "learning_rate": 4.39188205314889e-07, + "loss": 0.7855, + "step": 46947 + }, + { + "epoch": 8.54, + "learning_rate": 4.391336002912268e-07, + "loss": 0.1618, + "step": 46948 + }, + { + "epoch": 8.54, + "learning_rate": 4.390789952675646e-07, + "loss": 0.3069, + "step": 46949 + }, + { + "epoch": 8.54, + "learning_rate": 4.390243902439024e-07, + "loss": 0.0309, + "step": 46950 + }, + { + "epoch": 8.54, + "learning_rate": 4.389697852202403e-07, + "loss": 0.4145, + "step": 46951 + }, + { + "epoch": 8.54, + "learning_rate": 4.389151801965781e-07, + "loss": 0.0094, + "step": 46952 + }, + { + "epoch": 8.54, + "learning_rate": 4.388605751729159e-07, + "loss": 0.0544, + "step": 46953 + }, + { + "epoch": 8.54, + "learning_rate": 4.3880597014925376e-07, + "loss": 0.4029, + "step": 46954 + }, + { + "epoch": 8.54, + "learning_rate": 4.3875136512559154e-07, + "loss": 0.2685, + "step": 46955 + }, + { + "epoch": 8.54, + "learning_rate": 4.3869676010192943e-07, + "loss": 1.7916, + "step": 46956 + }, + { + "epoch": 8.54, + "learning_rate": 4.3864215507826717e-07, + "loss": 0.0529, + "step": 46957 + }, + { + "epoch": 8.54, + "learning_rate": 4.3858755005460506e-07, + "loss": 1.3764, + "step": 46958 + }, + { + "epoch": 8.54, + "learning_rate": 4.3853294503094284e-07, + "loss": 0.1622, + "step": 46959 + }, + { + "epoch": 8.54, + "learning_rate": 4.384783400072807e-07, + "loss": 0.0037, + "step": 46960 + }, + { + "epoch": 8.54, + "learning_rate": 4.384237349836185e-07, + "loss": 0.5848, + "step": 46961 + }, + { + "epoch": 8.54, + "learning_rate": 4.383691299599563e-07, + "loss": 0.3484, + "step": 46962 + }, + { + "epoch": 8.54, + "learning_rate": 4.383145249362942e-07, + "loss": 0.0109, + "step": 46963 + }, + { + "epoch": 8.54, + "learning_rate": 4.38259919912632e-07, + "loss": 0.0517, + "step": 46964 + }, + { + "epoch": 8.54, + "learning_rate": 4.382053148889698e-07, + "loss": 0.0527, + "step": 46965 + }, + { + "epoch": 8.54, + "learning_rate": 4.381507098653076e-07, + "loss": 0.0101, + "step": 46966 + }, + { + "epoch": 8.54, + "learning_rate": 4.3809610484164543e-07, + "loss": 0.0054, + "step": 46967 + }, + { + "epoch": 8.54, + "learning_rate": 4.380414998179833e-07, + "loss": 0.0062, + "step": 46968 + }, + { + "epoch": 8.54, + "learning_rate": 4.3798689479432106e-07, + "loss": 0.4117, + "step": 46969 + }, + { + "epoch": 8.54, + "learning_rate": 4.3793228977065895e-07, + "loss": 0.0104, + "step": 46970 + }, + { + "epoch": 8.54, + "learning_rate": 4.3787768474699673e-07, + "loss": 0.0069, + "step": 46971 + }, + { + "epoch": 8.54, + "learning_rate": 4.3782307972333457e-07, + "loss": 0.4216, + "step": 46972 + }, + { + "epoch": 8.54, + "learning_rate": 4.3776847469967235e-07, + "loss": 0.0959, + "step": 46973 + }, + { + "epoch": 8.54, + "learning_rate": 4.377138696760102e-07, + "loss": 0.0066, + "step": 46974 + }, + { + "epoch": 8.54, + "learning_rate": 4.376592646523481e-07, + "loss": 2.438, + "step": 46975 + }, + { + "epoch": 8.54, + "learning_rate": 4.3760465962868587e-07, + "loss": 1.2271, + "step": 46976 + }, + { + "epoch": 8.54, + "learning_rate": 4.375500546050237e-07, + "loss": 0.0113, + "step": 46977 + }, + { + "epoch": 8.54, + "learning_rate": 4.374954495813615e-07, + "loss": 0.0078, + "step": 46978 + }, + { + "epoch": 8.54, + "learning_rate": 4.374408445576993e-07, + "loss": 0.4936, + "step": 46979 + }, + { + "epoch": 8.54, + "learning_rate": 4.373862395340371e-07, + "loss": 0.0038, + "step": 46980 + }, + { + "epoch": 8.54, + "learning_rate": 4.3733163451037495e-07, + "loss": 2.2057, + "step": 46981 + }, + { + "epoch": 8.54, + "learning_rate": 4.3727702948671273e-07, + "loss": 0.0062, + "step": 46982 + }, + { + "epoch": 8.54, + "learning_rate": 4.372224244630506e-07, + "loss": 1.6015, + "step": 46983 + }, + { + "epoch": 8.54, + "learning_rate": 4.3716781943938846e-07, + "loss": 1.2333, + "step": 46984 + }, + { + "epoch": 8.54, + "learning_rate": 4.3711321441572625e-07, + "loss": 0.1385, + "step": 46985 + }, + { + "epoch": 8.54, + "learning_rate": 4.370586093920641e-07, + "loss": 0.3634, + "step": 46986 + }, + { + "epoch": 8.54, + "learning_rate": 4.3700400436840187e-07, + "loss": 0.0059, + "step": 46987 + }, + { + "epoch": 8.54, + "learning_rate": 4.3694939934473976e-07, + "loss": 2.8078, + "step": 46988 + }, + { + "epoch": 8.55, + "learning_rate": 4.368947943210775e-07, + "loss": 0.3108, + "step": 46989 + }, + { + "epoch": 8.55, + "learning_rate": 4.368401892974154e-07, + "loss": 0.9657, + "step": 46990 + }, + { + "epoch": 8.55, + "learning_rate": 4.367855842737532e-07, + "loss": 2.9927, + "step": 46991 + }, + { + "epoch": 8.55, + "learning_rate": 4.36730979250091e-07, + "loss": 0.0584, + "step": 46992 + }, + { + "epoch": 8.55, + "learning_rate": 4.3667637422642884e-07, + "loss": 0.4122, + "step": 46993 + }, + { + "epoch": 8.55, + "learning_rate": 4.366217692027666e-07, + "loss": 2.2326, + "step": 46994 + }, + { + "epoch": 8.55, + "learning_rate": 4.365671641791045e-07, + "loss": 0.0354, + "step": 46995 + }, + { + "epoch": 8.55, + "learning_rate": 4.365125591554423e-07, + "loss": 2.1002, + "step": 46996 + }, + { + "epoch": 8.55, + "learning_rate": 4.3645795413178014e-07, + "loss": 0.9013, + "step": 46997 + }, + { + "epoch": 8.55, + "learning_rate": 4.36403349108118e-07, + "loss": 0.7647, + "step": 46998 + }, + { + "epoch": 8.55, + "learning_rate": 4.3634874408445576e-07, + "loss": 0.246, + "step": 46999 + }, + { + "epoch": 8.55, + "learning_rate": 4.3629413906079365e-07, + "loss": 0.1143, + "step": 47000 + }, + { + "epoch": 8.55, + "learning_rate": 4.362395340371314e-07, + "loss": 2.2438, + "step": 47001 + }, + { + "epoch": 8.55, + "learning_rate": 4.3618492901346927e-07, + "loss": 0.1024, + "step": 47002 + }, + { + "epoch": 8.55, + "learning_rate": 4.3613032398980706e-07, + "loss": 1.2095, + "step": 47003 + }, + { + "epoch": 8.55, + "learning_rate": 4.360757189661449e-07, + "loss": 0.1322, + "step": 47004 + }, + { + "epoch": 8.55, + "learning_rate": 4.3602111394248273e-07, + "loss": 0.0023, + "step": 47005 + }, + { + "epoch": 8.55, + "learning_rate": 4.359665089188205e-07, + "loss": 1.0713, + "step": 47006 + }, + { + "epoch": 8.55, + "learning_rate": 4.359119038951584e-07, + "loss": 0.1358, + "step": 47007 + }, + { + "epoch": 8.55, + "learning_rate": 4.358572988714962e-07, + "loss": 0.0086, + "step": 47008 + }, + { + "epoch": 8.55, + "learning_rate": 4.3580269384783403e-07, + "loss": 0.0829, + "step": 47009 + }, + { + "epoch": 8.55, + "learning_rate": 4.357480888241718e-07, + "loss": 0.1442, + "step": 47010 + }, + { + "epoch": 8.55, + "learning_rate": 4.3569348380050965e-07, + "loss": 0.0439, + "step": 47011 + }, + { + "epoch": 8.55, + "learning_rate": 4.3563887877684754e-07, + "loss": 1.852, + "step": 47012 + }, + { + "epoch": 8.55, + "learning_rate": 4.3558427375318527e-07, + "loss": 0.1018, + "step": 47013 + }, + { + "epoch": 8.55, + "learning_rate": 4.3552966872952316e-07, + "loss": 1.5122, + "step": 47014 + }, + { + "epoch": 8.55, + "learning_rate": 4.3547506370586095e-07, + "loss": 0.0142, + "step": 47015 + }, + { + "epoch": 8.55, + "learning_rate": 4.354204586821988e-07, + "loss": 3.1809, + "step": 47016 + }, + { + "epoch": 8.55, + "learning_rate": 4.3536585365853657e-07, + "loss": 0.3006, + "step": 47017 + }, + { + "epoch": 8.55, + "learning_rate": 4.353112486348744e-07, + "loss": 0.0056, + "step": 47018 + }, + { + "epoch": 8.55, + "learning_rate": 4.352566436112122e-07, + "loss": 0.0049, + "step": 47019 + }, + { + "epoch": 8.55, + "learning_rate": 4.352020385875501e-07, + "loss": 0.0052, + "step": 47020 + }, + { + "epoch": 8.55, + "learning_rate": 4.351474335638879e-07, + "loss": 0.0069, + "step": 47021 + }, + { + "epoch": 8.55, + "learning_rate": 4.350928285402257e-07, + "loss": 2.4884, + "step": 47022 + }, + { + "epoch": 8.55, + "learning_rate": 4.3503822351656354e-07, + "loss": 0.0423, + "step": 47023 + }, + { + "epoch": 8.55, + "learning_rate": 4.349836184929013e-07, + "loss": 0.0043, + "step": 47024 + }, + { + "epoch": 8.55, + "learning_rate": 4.3492901346923916e-07, + "loss": 0.0044, + "step": 47025 + }, + { + "epoch": 8.55, + "learning_rate": 4.3487440844557695e-07, + "loss": 0.7191, + "step": 47026 + }, + { + "epoch": 8.55, + "learning_rate": 4.3481980342191484e-07, + "loss": 0.4222, + "step": 47027 + }, + { + "epoch": 8.55, + "learning_rate": 4.347651983982527e-07, + "loss": 0.2499, + "step": 47028 + }, + { + "epoch": 8.55, + "learning_rate": 4.3471059337459046e-07, + "loss": 1.489, + "step": 47029 + }, + { + "epoch": 8.55, + "learning_rate": 4.346559883509283e-07, + "loss": 0.4476, + "step": 47030 + }, + { + "epoch": 8.55, + "learning_rate": 4.346013833272661e-07, + "loss": 0.0234, + "step": 47031 + }, + { + "epoch": 8.55, + "learning_rate": 4.3454677830360397e-07, + "loss": 0.3518, + "step": 47032 + }, + { + "epoch": 8.55, + "learning_rate": 4.344921732799417e-07, + "loss": 0.6462, + "step": 47033 + }, + { + "epoch": 8.55, + "learning_rate": 4.344375682562796e-07, + "loss": 0.8012, + "step": 47034 + }, + { + "epoch": 8.55, + "learning_rate": 4.3438296323261743e-07, + "loss": 1.4665, + "step": 47035 + }, + { + "epoch": 8.55, + "learning_rate": 4.343283582089552e-07, + "loss": 0.0073, + "step": 47036 + }, + { + "epoch": 8.55, + "learning_rate": 4.342737531852931e-07, + "loss": 1.2761, + "step": 47037 + }, + { + "epoch": 8.55, + "learning_rate": 4.3421914816163084e-07, + "loss": 0.0588, + "step": 47038 + }, + { + "epoch": 8.55, + "learning_rate": 4.3416454313796873e-07, + "loss": 1.9954, + "step": 47039 + }, + { + "epoch": 8.55, + "learning_rate": 4.341099381143065e-07, + "loss": 0.3463, + "step": 47040 + }, + { + "epoch": 8.55, + "learning_rate": 4.3405533309064435e-07, + "loss": 0.66, + "step": 47041 + }, + { + "epoch": 8.55, + "learning_rate": 4.340007280669822e-07, + "loss": 1.5947, + "step": 47042 + }, + { + "epoch": 8.55, + "learning_rate": 4.3394612304332e-07, + "loss": 0.3883, + "step": 47043 + }, + { + "epoch": 8.56, + "learning_rate": 4.3389151801965786e-07, + "loss": 0.0916, + "step": 47044 + }, + { + "epoch": 8.56, + "learning_rate": 4.338369129959956e-07, + "loss": 0.2178, + "step": 47045 + }, + { + "epoch": 8.56, + "learning_rate": 4.337823079723335e-07, + "loss": 1.0026, + "step": 47046 + }, + { + "epoch": 8.56, + "learning_rate": 4.3372770294867127e-07, + "loss": 0.0063, + "step": 47047 + }, + { + "epoch": 8.56, + "learning_rate": 4.336730979250091e-07, + "loss": 0.4016, + "step": 47048 + }, + { + "epoch": 8.56, + "learning_rate": 4.33618492901347e-07, + "loss": 0.0205, + "step": 47049 + }, + { + "epoch": 8.56, + "learning_rate": 4.3356388787768473e-07, + "loss": 2.445, + "step": 47050 + }, + { + "epoch": 8.56, + "learning_rate": 4.335092828540226e-07, + "loss": 0.0604, + "step": 47051 + }, + { + "epoch": 8.56, + "learning_rate": 4.334546778303604e-07, + "loss": 0.0188, + "step": 47052 + }, + { + "epoch": 8.56, + "learning_rate": 4.3340007280669824e-07, + "loss": 2.6674, + "step": 47053 + }, + { + "epoch": 8.56, + "learning_rate": 4.3334546778303603e-07, + "loss": 0.01, + "step": 47054 + }, + { + "epoch": 8.56, + "learning_rate": 4.3329086275937387e-07, + "loss": 2.5606, + "step": 47055 + }, + { + "epoch": 8.56, + "learning_rate": 4.3323625773571165e-07, + "loss": 1.5548, + "step": 47056 + }, + { + "epoch": 8.56, + "learning_rate": 4.331816527120495e-07, + "loss": 2.4015, + "step": 47057 + }, + { + "epoch": 8.56, + "learning_rate": 4.331270476883874e-07, + "loss": 0.0103, + "step": 47058 + }, + { + "epoch": 8.56, + "learning_rate": 4.3307244266472516e-07, + "loss": 0.019, + "step": 47059 + }, + { + "epoch": 8.56, + "learning_rate": 4.33017837641063e-07, + "loss": 2.217, + "step": 47060 + }, + { + "epoch": 8.56, + "learning_rate": 4.329632326174008e-07, + "loss": 0.3977, + "step": 47061 + }, + { + "epoch": 8.56, + "learning_rate": 4.329086275937386e-07, + "loss": 0.2437, + "step": 47062 + }, + { + "epoch": 8.56, + "learning_rate": 4.328540225700764e-07, + "loss": 1.1678, + "step": 47063 + }, + { + "epoch": 8.56, + "learning_rate": 4.327994175464143e-07, + "loss": 0.4573, + "step": 47064 + }, + { + "epoch": 8.56, + "learning_rate": 4.3274481252275214e-07, + "loss": 1.478, + "step": 47065 + }, + { + "epoch": 8.56, + "learning_rate": 4.326902074990899e-07, + "loss": 1.0718, + "step": 47066 + }, + { + "epoch": 8.56, + "learning_rate": 4.3263560247542776e-07, + "loss": 0.0388, + "step": 47067 + }, + { + "epoch": 8.56, + "learning_rate": 4.3258099745176554e-07, + "loss": 0.0687, + "step": 47068 + }, + { + "epoch": 8.56, + "learning_rate": 4.3252639242810343e-07, + "loss": 0.7825, + "step": 47069 + }, + { + "epoch": 8.56, + "learning_rate": 4.3247178740444116e-07, + "loss": 0.0058, + "step": 47070 + }, + { + "epoch": 8.56, + "learning_rate": 4.3241718238077905e-07, + "loss": 0.0495, + "step": 47071 + }, + { + "epoch": 8.56, + "learning_rate": 4.323625773571169e-07, + "loss": 0.2589, + "step": 47072 + }, + { + "epoch": 8.56, + "learning_rate": 4.323079723334547e-07, + "loss": 0.2987, + "step": 47073 + }, + { + "epoch": 8.56, + "learning_rate": 4.322533673097925e-07, + "loss": 0.8349, + "step": 47074 + }, + { + "epoch": 8.56, + "learning_rate": 4.321987622861303e-07, + "loss": 0.0117, + "step": 47075 + }, + { + "epoch": 8.56, + "learning_rate": 4.321441572624682e-07, + "loss": 1.8336, + "step": 47076 + }, + { + "epoch": 8.56, + "learning_rate": 4.320895522388059e-07, + "loss": 0.0069, + "step": 47077 + }, + { + "epoch": 8.56, + "learning_rate": 4.320349472151438e-07, + "loss": 0.0055, + "step": 47078 + }, + { + "epoch": 8.56, + "learning_rate": 4.3198034219148165e-07, + "loss": 0.081, + "step": 47079 + }, + { + "epoch": 8.56, + "learning_rate": 4.3192573716781943e-07, + "loss": 0.0097, + "step": 47080 + }, + { + "epoch": 8.56, + "learning_rate": 4.318711321441573e-07, + "loss": 0.0627, + "step": 47081 + }, + { + "epoch": 8.56, + "learning_rate": 4.3181652712049506e-07, + "loss": 0.0121, + "step": 47082 + }, + { + "epoch": 8.56, + "learning_rate": 4.3176192209683295e-07, + "loss": 2.3079, + "step": 47083 + }, + { + "epoch": 8.56, + "learning_rate": 4.3170731707317073e-07, + "loss": 0.0048, + "step": 47084 + }, + { + "epoch": 8.56, + "learning_rate": 4.3165271204950857e-07, + "loss": 1.5008, + "step": 47085 + }, + { + "epoch": 8.56, + "learning_rate": 4.315981070258464e-07, + "loss": 0.3362, + "step": 47086 + }, + { + "epoch": 8.56, + "learning_rate": 4.315435020021842e-07, + "loss": 0.323, + "step": 47087 + }, + { + "epoch": 8.56, + "learning_rate": 4.314888969785221e-07, + "loss": 0.3484, + "step": 47088 + }, + { + "epoch": 8.56, + "learning_rate": 4.3143429195485986e-07, + "loss": 0.4919, + "step": 47089 + }, + { + "epoch": 8.56, + "learning_rate": 4.313796869311977e-07, + "loss": 0.1578, + "step": 47090 + }, + { + "epoch": 8.56, + "learning_rate": 4.313250819075355e-07, + "loss": 0.2373, + "step": 47091 + }, + { + "epoch": 8.56, + "learning_rate": 4.312704768838733e-07, + "loss": 0.0385, + "step": 47092 + }, + { + "epoch": 8.56, + "learning_rate": 4.312158718602112e-07, + "loss": 0.159, + "step": 47093 + }, + { + "epoch": 8.56, + "learning_rate": 4.3116126683654895e-07, + "loss": 0.0168, + "step": 47094 + }, + { + "epoch": 8.56, + "learning_rate": 4.3110666181288684e-07, + "loss": 2.1877, + "step": 47095 + }, + { + "epoch": 8.56, + "learning_rate": 4.310520567892246e-07, + "loss": 0.0243, + "step": 47096 + }, + { + "epoch": 8.56, + "learning_rate": 4.3099745176556246e-07, + "loss": 0.3973, + "step": 47097 + }, + { + "epoch": 8.56, + "learning_rate": 4.3094284674190024e-07, + "loss": 1.2002, + "step": 47098 + }, + { + "epoch": 8.57, + "learning_rate": 4.308882417182381e-07, + "loss": 0.0181, + "step": 47099 + }, + { + "epoch": 8.57, + "learning_rate": 4.3083363669457587e-07, + "loss": 0.014, + "step": 47100 + }, + { + "epoch": 8.57, + "learning_rate": 4.3077903167091376e-07, + "loss": 0.3145, + "step": 47101 + }, + { + "epoch": 8.57, + "learning_rate": 4.307244266472516e-07, + "loss": 1.4656, + "step": 47102 + }, + { + "epoch": 8.57, + "learning_rate": 4.306698216235894e-07, + "loss": 0.3118, + "step": 47103 + }, + { + "epoch": 8.57, + "learning_rate": 4.306152165999272e-07, + "loss": 0.0392, + "step": 47104 + }, + { + "epoch": 8.57, + "learning_rate": 4.30560611576265e-07, + "loss": 0.0033, + "step": 47105 + }, + { + "epoch": 8.57, + "learning_rate": 4.3050600655260284e-07, + "loss": 0.0592, + "step": 47106 + }, + { + "epoch": 8.57, + "learning_rate": 4.304514015289406e-07, + "loss": 0.7633, + "step": 47107 + }, + { + "epoch": 8.57, + "learning_rate": 4.303967965052785e-07, + "loss": 0.0053, + "step": 47108 + }, + { + "epoch": 8.57, + "learning_rate": 4.3034219148161635e-07, + "loss": 1.0933, + "step": 47109 + }, + { + "epoch": 8.57, + "learning_rate": 4.3028758645795414e-07, + "loss": 0.078, + "step": 47110 + }, + { + "epoch": 8.57, + "learning_rate": 4.3023298143429197e-07, + "loss": 0.4099, + "step": 47111 + }, + { + "epoch": 8.57, + "learning_rate": 4.3017837641062976e-07, + "loss": 0.0238, + "step": 47112 + }, + { + "epoch": 8.57, + "learning_rate": 4.3012377138696765e-07, + "loss": 0.9275, + "step": 47113 + }, + { + "epoch": 8.57, + "learning_rate": 4.300691663633054e-07, + "loss": 0.1547, + "step": 47114 + }, + { + "epoch": 8.57, + "learning_rate": 4.3001456133964327e-07, + "loss": 0.069, + "step": 47115 + }, + { + "epoch": 8.57, + "learning_rate": 4.299599563159811e-07, + "loss": 0.0063, + "step": 47116 + }, + { + "epoch": 8.57, + "learning_rate": 4.299053512923189e-07, + "loss": 0.2541, + "step": 47117 + }, + { + "epoch": 8.57, + "learning_rate": 4.2985074626865673e-07, + "loss": 0.931, + "step": 47118 + }, + { + "epoch": 8.57, + "learning_rate": 4.297961412449945e-07, + "loss": 0.0354, + "step": 47119 + }, + { + "epoch": 8.57, + "learning_rate": 4.297415362213324e-07, + "loss": 0.0675, + "step": 47120 + }, + { + "epoch": 8.57, + "learning_rate": 4.296869311976702e-07, + "loss": 0.4905, + "step": 47121 + }, + { + "epoch": 8.57, + "learning_rate": 4.2963232617400803e-07, + "loss": 0.159, + "step": 47122 + }, + { + "epoch": 8.57, + "learning_rate": 4.2957772115034586e-07, + "loss": 1.9809, + "step": 47123 + }, + { + "epoch": 8.57, + "learning_rate": 4.2952311612668365e-07, + "loss": 1.2569, + "step": 47124 + }, + { + "epoch": 8.57, + "learning_rate": 4.2946851110302154e-07, + "loss": 0.1723, + "step": 47125 + }, + { + "epoch": 8.57, + "learning_rate": 4.2941390607935927e-07, + "loss": 0.1363, + "step": 47126 + }, + { + "epoch": 8.57, + "learning_rate": 4.2935930105569716e-07, + "loss": 0.3055, + "step": 47127 + }, + { + "epoch": 8.57, + "learning_rate": 4.2930469603203495e-07, + "loss": 0.0313, + "step": 47128 + }, + { + "epoch": 8.57, + "learning_rate": 4.292500910083728e-07, + "loss": 0.085, + "step": 47129 + }, + { + "epoch": 8.57, + "learning_rate": 4.291954859847106e-07, + "loss": 0.1025, + "step": 47130 + }, + { + "epoch": 8.57, + "learning_rate": 4.291408809610484e-07, + "loss": 2.1426, + "step": 47131 + }, + { + "epoch": 8.57, + "learning_rate": 4.290862759373863e-07, + "loss": 0.2068, + "step": 47132 + }, + { + "epoch": 8.57, + "learning_rate": 4.290316709137241e-07, + "loss": 0.3482, + "step": 47133 + }, + { + "epoch": 8.57, + "learning_rate": 4.289770658900619e-07, + "loss": 0.0061, + "step": 47134 + }, + { + "epoch": 8.57, + "learning_rate": 4.289224608663997e-07, + "loss": 0.0127, + "step": 47135 + }, + { + "epoch": 8.57, + "learning_rate": 4.2886785584273754e-07, + "loss": 0.0485, + "step": 47136 + }, + { + "epoch": 8.57, + "learning_rate": 4.288132508190753e-07, + "loss": 0.2381, + "step": 47137 + }, + { + "epoch": 8.57, + "learning_rate": 4.2875864579541316e-07, + "loss": 0.0141, + "step": 47138 + }, + { + "epoch": 8.57, + "learning_rate": 4.2870404077175105e-07, + "loss": 0.6282, + "step": 47139 + }, + { + "epoch": 8.57, + "learning_rate": 4.2864943574808884e-07, + "loss": 0.2511, + "step": 47140 + }, + { + "epoch": 8.57, + "learning_rate": 4.285948307244267e-07, + "loss": 0.0074, + "step": 47141 + }, + { + "epoch": 8.57, + "learning_rate": 4.2854022570076446e-07, + "loss": 0.0094, + "step": 47142 + }, + { + "epoch": 8.57, + "learning_rate": 4.284856206771023e-07, + "loss": 0.6488, + "step": 47143 + }, + { + "epoch": 8.57, + "learning_rate": 4.284310156534401e-07, + "loss": 0.014, + "step": 47144 + }, + { + "epoch": 8.57, + "learning_rate": 4.2837641062977797e-07, + "loss": 0.8459, + "step": 47145 + }, + { + "epoch": 8.57, + "learning_rate": 4.283218056061158e-07, + "loss": 0.1112, + "step": 47146 + }, + { + "epoch": 8.57, + "learning_rate": 4.282672005824536e-07, + "loss": 0.6503, + "step": 47147 + }, + { + "epoch": 8.57, + "learning_rate": 4.2821259555879143e-07, + "loss": 0.97, + "step": 47148 + }, + { + "epoch": 8.57, + "learning_rate": 4.281579905351292e-07, + "loss": 0.3653, + "step": 47149 + }, + { + "epoch": 8.57, + "learning_rate": 4.2810338551146705e-07, + "loss": 0.0031, + "step": 47150 + }, + { + "epoch": 8.57, + "learning_rate": 4.2804878048780484e-07, + "loss": 0.6991, + "step": 47151 + }, + { + "epoch": 8.57, + "learning_rate": 4.2799417546414273e-07, + "loss": 1.0625, + "step": 47152 + }, + { + "epoch": 8.57, + "learning_rate": 4.2793957044048057e-07, + "loss": 1.3744, + "step": 47153 + }, + { + "epoch": 8.58, + "learning_rate": 4.2788496541681835e-07, + "loss": 0.0103, + "step": 47154 + }, + { + "epoch": 8.58, + "learning_rate": 4.278303603931562e-07, + "loss": 0.0079, + "step": 47155 + }, + { + "epoch": 8.58, + "learning_rate": 4.2777575536949397e-07, + "loss": 0.2397, + "step": 47156 + }, + { + "epoch": 8.58, + "learning_rate": 4.2772115034583186e-07, + "loss": 0.1174, + "step": 47157 + }, + { + "epoch": 8.58, + "learning_rate": 4.276665453221696e-07, + "loss": 0.0997, + "step": 47158 + }, + { + "epoch": 8.58, + "learning_rate": 4.276119402985075e-07, + "loss": 0.1713, + "step": 47159 + }, + { + "epoch": 8.58, + "learning_rate": 4.275573352748453e-07, + "loss": 1.3364, + "step": 47160 + }, + { + "epoch": 8.58, + "learning_rate": 4.275027302511831e-07, + "loss": 0.1607, + "step": 47161 + }, + { + "epoch": 8.58, + "learning_rate": 4.2744812522752094e-07, + "loss": 0.4684, + "step": 47162 + }, + { + "epoch": 8.58, + "learning_rate": 4.2739352020385873e-07, + "loss": 1.7969, + "step": 47163 + }, + { + "epoch": 8.58, + "learning_rate": 4.273389151801966e-07, + "loss": 0.4026, + "step": 47164 + }, + { + "epoch": 8.58, + "learning_rate": 4.272843101565344e-07, + "loss": 0.9142, + "step": 47165 + }, + { + "epoch": 8.58, + "learning_rate": 4.2722970513287224e-07, + "loss": 0.1331, + "step": 47166 + }, + { + "epoch": 8.58, + "learning_rate": 4.271751001092101e-07, + "loss": 1.4178, + "step": 47167 + }, + { + "epoch": 8.58, + "learning_rate": 4.2712049508554786e-07, + "loss": 0.6352, + "step": 47168 + }, + { + "epoch": 8.58, + "learning_rate": 4.2706589006188575e-07, + "loss": 0.0459, + "step": 47169 + }, + { + "epoch": 8.58, + "learning_rate": 4.270112850382235e-07, + "loss": 1.0756, + "step": 47170 + }, + { + "epoch": 8.58, + "learning_rate": 4.269566800145614e-07, + "loss": 0.2068, + "step": 47171 + }, + { + "epoch": 8.58, + "learning_rate": 4.2690207499089916e-07, + "loss": 0.0088, + "step": 47172 + }, + { + "epoch": 8.58, + "learning_rate": 4.26847469967237e-07, + "loss": 1.5969, + "step": 47173 + }, + { + "epoch": 8.58, + "learning_rate": 4.267928649435748e-07, + "loss": 0.1435, + "step": 47174 + }, + { + "epoch": 8.58, + "learning_rate": 4.267382599199126e-07, + "loss": 0.9448, + "step": 47175 + }, + { + "epoch": 8.58, + "learning_rate": 4.266836548962505e-07, + "loss": 0.0034, + "step": 47176 + }, + { + "epoch": 8.58, + "learning_rate": 4.266290498725883e-07, + "loss": 0.0073, + "step": 47177 + }, + { + "epoch": 8.58, + "learning_rate": 4.2657444484892613e-07, + "loss": 0.0043, + "step": 47178 + }, + { + "epoch": 8.58, + "learning_rate": 4.265198398252639e-07, + "loss": 0.0515, + "step": 47179 + }, + { + "epoch": 8.58, + "learning_rate": 4.2646523480160176e-07, + "loss": 1.625, + "step": 47180 + }, + { + "epoch": 8.58, + "learning_rate": 4.2641062977793954e-07, + "loss": 0.0267, + "step": 47181 + }, + { + "epoch": 8.58, + "learning_rate": 4.263560247542774e-07, + "loss": 0.1637, + "step": 47182 + }, + { + "epoch": 8.58, + "learning_rate": 4.2630141973061527e-07, + "loss": 0.0444, + "step": 47183 + }, + { + "epoch": 8.58, + "learning_rate": 4.2624681470695305e-07, + "loss": 0.0235, + "step": 47184 + }, + { + "epoch": 8.58, + "learning_rate": 4.261922096832909e-07, + "loss": 0.5235, + "step": 47185 + }, + { + "epoch": 8.58, + "learning_rate": 4.261376046596287e-07, + "loss": 0.2234, + "step": 47186 + }, + { + "epoch": 8.58, + "learning_rate": 4.260829996359665e-07, + "loss": 0.9378, + "step": 47187 + }, + { + "epoch": 8.58, + "learning_rate": 4.260283946123043e-07, + "loss": 1.1257, + "step": 47188 + }, + { + "epoch": 8.58, + "learning_rate": 4.259737895886422e-07, + "loss": 0.6214, + "step": 47189 + }, + { + "epoch": 8.58, + "learning_rate": 4.2591918456498e-07, + "loss": 0.108, + "step": 47190 + }, + { + "epoch": 8.58, + "learning_rate": 4.258645795413178e-07, + "loss": 0.0191, + "step": 47191 + }, + { + "epoch": 8.58, + "learning_rate": 4.2580997451765565e-07, + "loss": 0.392, + "step": 47192 + }, + { + "epoch": 8.58, + "learning_rate": 4.2575536949399343e-07, + "loss": 2.761, + "step": 47193 + }, + { + "epoch": 8.58, + "learning_rate": 4.2570076447033127e-07, + "loss": 0.1064, + "step": 47194 + }, + { + "epoch": 8.58, + "learning_rate": 4.2564615944666905e-07, + "loss": 0.0447, + "step": 47195 + }, + { + "epoch": 8.58, + "learning_rate": 4.2559155442300694e-07, + "loss": 0.7046, + "step": 47196 + }, + { + "epoch": 8.58, + "learning_rate": 4.255369493993448e-07, + "loss": 0.0047, + "step": 47197 + }, + { + "epoch": 8.58, + "learning_rate": 4.2548234437568257e-07, + "loss": 0.099, + "step": 47198 + }, + { + "epoch": 8.58, + "learning_rate": 4.254277393520204e-07, + "loss": 0.017, + "step": 47199 + }, + { + "epoch": 8.58, + "learning_rate": 4.253731343283582e-07, + "loss": 1.7659, + "step": 47200 + }, + { + "epoch": 8.58, + "learning_rate": 4.253185293046961e-07, + "loss": 0.5064, + "step": 47201 + }, + { + "epoch": 8.58, + "learning_rate": 4.252639242810338e-07, + "loss": 0.1804, + "step": 47202 + }, + { + "epoch": 8.58, + "learning_rate": 4.252093192573717e-07, + "loss": 0.01, + "step": 47203 + }, + { + "epoch": 8.58, + "learning_rate": 4.2515471423370954e-07, + "loss": 0.4762, + "step": 47204 + }, + { + "epoch": 8.58, + "learning_rate": 4.251001092100473e-07, + "loss": 0.0426, + "step": 47205 + }, + { + "epoch": 8.58, + "learning_rate": 4.2504550418638516e-07, + "loss": 0.0764, + "step": 47206 + }, + { + "epoch": 8.58, + "learning_rate": 4.2499089916272294e-07, + "loss": 1.3401, + "step": 47207 + }, + { + "epoch": 8.58, + "learning_rate": 4.2493629413906084e-07, + "loss": 0.8837, + "step": 47208 + }, + { + "epoch": 8.59, + "learning_rate": 4.248816891153986e-07, + "loss": 0.3944, + "step": 47209 + }, + { + "epoch": 8.59, + "learning_rate": 4.2482708409173646e-07, + "loss": 0.0834, + "step": 47210 + }, + { + "epoch": 8.59, + "learning_rate": 4.2477247906807424e-07, + "loss": 0.0506, + "step": 47211 + }, + { + "epoch": 8.59, + "learning_rate": 4.247178740444121e-07, + "loss": 0.1858, + "step": 47212 + }, + { + "epoch": 8.59, + "learning_rate": 4.2466326902074997e-07, + "loss": 0.8222, + "step": 47213 + }, + { + "epoch": 8.59, + "learning_rate": 4.246086639970877e-07, + "loss": 0.4496, + "step": 47214 + }, + { + "epoch": 8.59, + "learning_rate": 4.245540589734256e-07, + "loss": 0.4975, + "step": 47215 + }, + { + "epoch": 8.59, + "learning_rate": 4.244994539497634e-07, + "loss": 1.9371, + "step": 47216 + }, + { + "epoch": 8.59, + "learning_rate": 4.244448489261012e-07, + "loss": 0.417, + "step": 47217 + }, + { + "epoch": 8.59, + "learning_rate": 4.24390243902439e-07, + "loss": 0.027, + "step": 47218 + }, + { + "epoch": 8.59, + "learning_rate": 4.2433563887877684e-07, + "loss": 0.0053, + "step": 47219 + }, + { + "epoch": 8.59, + "learning_rate": 4.2428103385511473e-07, + "loss": 0.0699, + "step": 47220 + }, + { + "epoch": 8.59, + "learning_rate": 4.242264288314525e-07, + "loss": 1.907, + "step": 47221 + }, + { + "epoch": 8.59, + "learning_rate": 4.2417182380779035e-07, + "loss": 1.2535, + "step": 47222 + }, + { + "epoch": 8.59, + "learning_rate": 4.2411721878412813e-07, + "loss": 3.9136, + "step": 47223 + }, + { + "epoch": 8.59, + "learning_rate": 4.2406261376046597e-07, + "loss": 1.23, + "step": 47224 + }, + { + "epoch": 8.59, + "learning_rate": 4.2400800873680376e-07, + "loss": 0.0248, + "step": 47225 + }, + { + "epoch": 8.59, + "learning_rate": 4.239534037131416e-07, + "loss": 0.0382, + "step": 47226 + }, + { + "epoch": 8.59, + "learning_rate": 4.238987986894795e-07, + "loss": 0.081, + "step": 47227 + }, + { + "epoch": 8.59, + "learning_rate": 4.2384419366581727e-07, + "loss": 0.0396, + "step": 47228 + }, + { + "epoch": 8.59, + "learning_rate": 4.237895886421551e-07, + "loss": 0.5242, + "step": 47229 + }, + { + "epoch": 8.59, + "learning_rate": 4.237349836184929e-07, + "loss": 2.263, + "step": 47230 + }, + { + "epoch": 8.59, + "learning_rate": 4.2368037859483073e-07, + "loss": 0.013, + "step": 47231 + }, + { + "epoch": 8.59, + "learning_rate": 4.236257735711685e-07, + "loss": 0.1054, + "step": 47232 + }, + { + "epoch": 8.59, + "learning_rate": 4.235711685475064e-07, + "loss": 0.204, + "step": 47233 + }, + { + "epoch": 8.59, + "learning_rate": 4.2351656352384424e-07, + "loss": 0.1612, + "step": 47234 + }, + { + "epoch": 8.59, + "learning_rate": 4.23461958500182e-07, + "loss": 0.7042, + "step": 47235 + }, + { + "epoch": 8.59, + "learning_rate": 4.2340735347651986e-07, + "loss": 0.1415, + "step": 47236 + }, + { + "epoch": 8.59, + "learning_rate": 4.2335274845285765e-07, + "loss": 0.0731, + "step": 47237 + }, + { + "epoch": 8.59, + "learning_rate": 4.2329814342919554e-07, + "loss": 1.1444, + "step": 47238 + }, + { + "epoch": 8.59, + "learning_rate": 4.2324353840553327e-07, + "loss": 0.0453, + "step": 47239 + }, + { + "epoch": 8.59, + "learning_rate": 4.2318893338187116e-07, + "loss": 0.1448, + "step": 47240 + }, + { + "epoch": 8.59, + "learning_rate": 4.23134328358209e-07, + "loss": 1.0471, + "step": 47241 + }, + { + "epoch": 8.59, + "learning_rate": 4.230797233345468e-07, + "loss": 0.3734, + "step": 47242 + }, + { + "epoch": 8.59, + "learning_rate": 4.230251183108846e-07, + "loss": 0.7342, + "step": 47243 + }, + { + "epoch": 8.59, + "learning_rate": 4.229705132872224e-07, + "loss": 0.1631, + "step": 47244 + }, + { + "epoch": 8.59, + "learning_rate": 4.229159082635603e-07, + "loss": 0.0869, + "step": 47245 + }, + { + "epoch": 8.59, + "learning_rate": 4.22861303239898e-07, + "loss": 0.2978, + "step": 47246 + }, + { + "epoch": 8.59, + "learning_rate": 4.228066982162359e-07, + "loss": 0.0241, + "step": 47247 + }, + { + "epoch": 8.59, + "learning_rate": 4.2275209319257375e-07, + "loss": 1.1368, + "step": 47248 + }, + { + "epoch": 8.59, + "learning_rate": 4.2269748816891154e-07, + "loss": 1.453, + "step": 47249 + }, + { + "epoch": 8.59, + "learning_rate": 4.2264288314524943e-07, + "loss": 0.7226, + "step": 47250 + }, + { + "epoch": 8.59, + "learning_rate": 4.2258827812158716e-07, + "loss": 0.0212, + "step": 47251 + }, + { + "epoch": 8.59, + "learning_rate": 4.2253367309792505e-07, + "loss": 0.8644, + "step": 47252 + }, + { + "epoch": 8.59, + "learning_rate": 4.2247906807426284e-07, + "loss": 1.5499, + "step": 47253 + }, + { + "epoch": 8.59, + "learning_rate": 4.2242446305060067e-07, + "loss": 0.2893, + "step": 47254 + }, + { + "epoch": 8.59, + "learning_rate": 4.2236985802693846e-07, + "loss": 0.1168, + "step": 47255 + }, + { + "epoch": 8.59, + "learning_rate": 4.223152530032763e-07, + "loss": 0.1338, + "step": 47256 + }, + { + "epoch": 8.59, + "learning_rate": 4.222606479796142e-07, + "loss": 0.1912, + "step": 47257 + }, + { + "epoch": 8.59, + "learning_rate": 4.222060429559519e-07, + "loss": 0.2798, + "step": 47258 + }, + { + "epoch": 8.59, + "learning_rate": 4.221514379322898e-07, + "loss": 0.0527, + "step": 47259 + }, + { + "epoch": 8.59, + "learning_rate": 4.220968329086276e-07, + "loss": 0.005, + "step": 47260 + }, + { + "epoch": 8.59, + "learning_rate": 4.2204222788496543e-07, + "loss": 0.5523, + "step": 47261 + }, + { + "epoch": 8.59, + "learning_rate": 4.219876228613032e-07, + "loss": 0.027, + "step": 47262 + }, + { + "epoch": 8.59, + "learning_rate": 4.2193301783764105e-07, + "loss": 0.0063, + "step": 47263 + }, + { + "epoch": 8.6, + "learning_rate": 4.2187841281397894e-07, + "loss": 1.4177, + "step": 47264 + }, + { + "epoch": 8.6, + "learning_rate": 4.2182380779031673e-07, + "loss": 0.2037, + "step": 47265 + }, + { + "epoch": 8.6, + "learning_rate": 4.2176920276665456e-07, + "loss": 1.7836, + "step": 47266 + }, + { + "epoch": 8.6, + "learning_rate": 4.2171459774299235e-07, + "loss": 0.0503, + "step": 47267 + }, + { + "epoch": 8.6, + "learning_rate": 4.216599927193302e-07, + "loss": 0.1523, + "step": 47268 + }, + { + "epoch": 8.6, + "learning_rate": 4.2160538769566797e-07, + "loss": 2.7518, + "step": 47269 + }, + { + "epoch": 8.6, + "learning_rate": 4.2155078267200586e-07, + "loss": 0.435, + "step": 47270 + }, + { + "epoch": 8.6, + "learning_rate": 4.214961776483437e-07, + "loss": 0.2835, + "step": 47271 + }, + { + "epoch": 8.6, + "learning_rate": 4.214415726246815e-07, + "loss": 0.0066, + "step": 47272 + }, + { + "epoch": 8.6, + "learning_rate": 4.213869676010193e-07, + "loss": 0.2681, + "step": 47273 + }, + { + "epoch": 8.6, + "learning_rate": 4.213323625773571e-07, + "loss": 0.7377, + "step": 47274 + }, + { + "epoch": 8.6, + "learning_rate": 4.2127775755369494e-07, + "loss": 0.0023, + "step": 47275 + }, + { + "epoch": 8.6, + "learning_rate": 4.2122315253003273e-07, + "loss": 0.015, + "step": 47276 + }, + { + "epoch": 8.6, + "learning_rate": 4.211685475063706e-07, + "loss": 1.5885, + "step": 47277 + }, + { + "epoch": 8.6, + "learning_rate": 4.2111394248270846e-07, + "loss": 0.0759, + "step": 47278 + }, + { + "epoch": 8.6, + "learning_rate": 4.2105933745904624e-07, + "loss": 0.7985, + "step": 47279 + }, + { + "epoch": 8.6, + "learning_rate": 4.210047324353841e-07, + "loss": 3.5911, + "step": 47280 + }, + { + "epoch": 8.6, + "learning_rate": 4.2095012741172186e-07, + "loss": 0.0084, + "step": 47281 + }, + { + "epoch": 8.6, + "learning_rate": 4.2089552238805975e-07, + "loss": 1.4193, + "step": 47282 + }, + { + "epoch": 8.6, + "learning_rate": 4.208409173643975e-07, + "loss": 0.0094, + "step": 47283 + }, + { + "epoch": 8.6, + "learning_rate": 4.207863123407354e-07, + "loss": 1.3016, + "step": 47284 + }, + { + "epoch": 8.6, + "learning_rate": 4.207317073170732e-07, + "loss": 0.3147, + "step": 47285 + }, + { + "epoch": 8.6, + "learning_rate": 4.20677102293411e-07, + "loss": 0.0201, + "step": 47286 + }, + { + "epoch": 8.6, + "learning_rate": 4.2062249726974883e-07, + "loss": 0.0667, + "step": 47287 + }, + { + "epoch": 8.6, + "learning_rate": 4.205678922460866e-07, + "loss": 0.1292, + "step": 47288 + }, + { + "epoch": 8.6, + "learning_rate": 4.205132872224245e-07, + "loss": 0.0584, + "step": 47289 + }, + { + "epoch": 8.6, + "learning_rate": 4.2045868219876224e-07, + "loss": 0.0274, + "step": 47290 + }, + { + "epoch": 8.6, + "learning_rate": 4.2040407717510013e-07, + "loss": 0.1392, + "step": 47291 + }, + { + "epoch": 8.6, + "learning_rate": 4.203494721514379e-07, + "loss": 0.0536, + "step": 47292 + }, + { + "epoch": 8.6, + "learning_rate": 4.2029486712777575e-07, + "loss": 0.0084, + "step": 47293 + }, + { + "epoch": 8.6, + "learning_rate": 4.2024026210411364e-07, + "loss": 0.0602, + "step": 47294 + }, + { + "epoch": 8.6, + "learning_rate": 4.201856570804514e-07, + "loss": 0.168, + "step": 47295 + }, + { + "epoch": 8.6, + "learning_rate": 4.2013105205678927e-07, + "loss": 0.392, + "step": 47296 + }, + { + "epoch": 8.6, + "learning_rate": 4.2007644703312705e-07, + "loss": 0.0836, + "step": 47297 + }, + { + "epoch": 8.6, + "learning_rate": 4.200218420094649e-07, + "loss": 0.0536, + "step": 47298 + }, + { + "epoch": 8.6, + "learning_rate": 4.1996723698580267e-07, + "loss": 0.4169, + "step": 47299 + }, + { + "epoch": 8.6, + "learning_rate": 4.199126319621405e-07, + "loss": 0.3582, + "step": 47300 + }, + { + "epoch": 8.6, + "learning_rate": 4.198580269384784e-07, + "loss": 1.3206, + "step": 47301 + }, + { + "epoch": 8.6, + "learning_rate": 4.198034219148162e-07, + "loss": 0.112, + "step": 47302 + }, + { + "epoch": 8.6, + "learning_rate": 4.19748816891154e-07, + "loss": 2.0006, + "step": 47303 + }, + { + "epoch": 8.6, + "learning_rate": 4.196942118674918e-07, + "loss": 0.4225, + "step": 47304 + }, + { + "epoch": 8.6, + "learning_rate": 4.1963960684382965e-07, + "loss": 2.5394, + "step": 47305 + }, + { + "epoch": 8.6, + "learning_rate": 4.1958500182016743e-07, + "loss": 0.6401, + "step": 47306 + }, + { + "epoch": 8.6, + "learning_rate": 4.1953039679650527e-07, + "loss": 0.006, + "step": 47307 + }, + { + "epoch": 8.6, + "learning_rate": 4.1947579177284316e-07, + "loss": 0.007, + "step": 47308 + }, + { + "epoch": 8.6, + "learning_rate": 4.1942118674918094e-07, + "loss": 0.0474, + "step": 47309 + }, + { + "epoch": 8.6, + "learning_rate": 4.193665817255188e-07, + "loss": 2.9975, + "step": 47310 + }, + { + "epoch": 8.6, + "learning_rate": 4.1931197670185656e-07, + "loss": 1.6057, + "step": 47311 + }, + { + "epoch": 8.6, + "learning_rate": 4.192573716781944e-07, + "loss": 0.0064, + "step": 47312 + }, + { + "epoch": 8.6, + "learning_rate": 4.192027666545322e-07, + "loss": 0.0738, + "step": 47313 + }, + { + "epoch": 8.6, + "learning_rate": 4.191481616308701e-07, + "loss": 0.1762, + "step": 47314 + }, + { + "epoch": 8.6, + "learning_rate": 4.190935566072079e-07, + "loss": 0.172, + "step": 47315 + }, + { + "epoch": 8.6, + "learning_rate": 4.190389515835457e-07, + "loss": 1.4006, + "step": 47316 + }, + { + "epoch": 8.6, + "learning_rate": 4.1898434655988354e-07, + "loss": 0.0044, + "step": 47317 + }, + { + "epoch": 8.6, + "learning_rate": 4.189297415362213e-07, + "loss": 0.2443, + "step": 47318 + }, + { + "epoch": 8.61, + "learning_rate": 4.1887513651255916e-07, + "loss": 0.1418, + "step": 47319 + }, + { + "epoch": 8.61, + "learning_rate": 4.1882053148889694e-07, + "loss": 2.9492, + "step": 47320 + }, + { + "epoch": 8.61, + "learning_rate": 4.1876592646523483e-07, + "loss": 0.028, + "step": 47321 + }, + { + "epoch": 8.61, + "learning_rate": 4.1871132144157267e-07, + "loss": 0.1798, + "step": 47322 + }, + { + "epoch": 8.61, + "learning_rate": 4.1865671641791046e-07, + "loss": 0.0224, + "step": 47323 + }, + { + "epoch": 8.61, + "learning_rate": 4.186021113942483e-07, + "loss": 0.3818, + "step": 47324 + }, + { + "epoch": 8.61, + "learning_rate": 4.185475063705861e-07, + "loss": 1.455, + "step": 47325 + }, + { + "epoch": 8.61, + "learning_rate": 4.1849290134692397e-07, + "loss": 0.0384, + "step": 47326 + }, + { + "epoch": 8.61, + "learning_rate": 4.184382963232617e-07, + "loss": 0.0698, + "step": 47327 + }, + { + "epoch": 8.61, + "learning_rate": 4.183836912995996e-07, + "loss": 0.0106, + "step": 47328 + }, + { + "epoch": 8.61, + "learning_rate": 4.183290862759374e-07, + "loss": 1.4198, + "step": 47329 + }, + { + "epoch": 8.61, + "learning_rate": 4.182744812522752e-07, + "loss": 0.0431, + "step": 47330 + }, + { + "epoch": 8.61, + "learning_rate": 4.1821987622861305e-07, + "loss": 0.0047, + "step": 47331 + }, + { + "epoch": 8.61, + "learning_rate": 4.1816527120495083e-07, + "loss": 0.0143, + "step": 47332 + }, + { + "epoch": 8.61, + "learning_rate": 4.181106661812887e-07, + "loss": 1.5396, + "step": 47333 + }, + { + "epoch": 8.61, + "learning_rate": 4.180560611576265e-07, + "loss": 0.535, + "step": 47334 + }, + { + "epoch": 8.61, + "learning_rate": 4.1800145613396435e-07, + "loss": 0.1589, + "step": 47335 + }, + { + "epoch": 8.61, + "learning_rate": 4.1794685111030213e-07, + "loss": 0.3341, + "step": 47336 + }, + { + "epoch": 8.61, + "learning_rate": 4.1789224608663997e-07, + "loss": 0.1496, + "step": 47337 + }, + { + "epoch": 8.61, + "learning_rate": 4.1783764106297786e-07, + "loss": 1.6265, + "step": 47338 + }, + { + "epoch": 8.61, + "learning_rate": 4.177830360393156e-07, + "loss": 0.1042, + "step": 47339 + }, + { + "epoch": 8.61, + "learning_rate": 4.177284310156535e-07, + "loss": 0.164, + "step": 47340 + }, + { + "epoch": 8.61, + "learning_rate": 4.1767382599199127e-07, + "loss": 0.1954, + "step": 47341 + }, + { + "epoch": 8.61, + "learning_rate": 4.176192209683291e-07, + "loss": 0.9591, + "step": 47342 + }, + { + "epoch": 8.61, + "learning_rate": 4.175646159446669e-07, + "loss": 0.2461, + "step": 47343 + }, + { + "epoch": 8.61, + "learning_rate": 4.175100109210047e-07, + "loss": 0.5264, + "step": 47344 + }, + { + "epoch": 8.61, + "learning_rate": 4.174554058973426e-07, + "loss": 0.3451, + "step": 47345 + }, + { + "epoch": 8.61, + "learning_rate": 4.174008008736804e-07, + "loss": 0.2015, + "step": 47346 + }, + { + "epoch": 8.61, + "learning_rate": 4.1734619585001824e-07, + "loss": 1.3471, + "step": 47347 + }, + { + "epoch": 8.61, + "learning_rate": 4.17291590826356e-07, + "loss": 0.6738, + "step": 47348 + }, + { + "epoch": 8.61, + "learning_rate": 4.1723698580269386e-07, + "loss": 1.2548, + "step": 47349 + }, + { + "epoch": 8.61, + "learning_rate": 4.1718238077903165e-07, + "loss": 0.1832, + "step": 47350 + }, + { + "epoch": 8.61, + "learning_rate": 4.171277757553695e-07, + "loss": 0.3646, + "step": 47351 + }, + { + "epoch": 8.61, + "learning_rate": 4.1707317073170737e-07, + "loss": 0.0851, + "step": 47352 + }, + { + "epoch": 8.61, + "learning_rate": 4.1701856570804516e-07, + "loss": 0.005, + "step": 47353 + }, + { + "epoch": 8.61, + "learning_rate": 4.16963960684383e-07, + "loss": 0.1227, + "step": 47354 + }, + { + "epoch": 8.61, + "learning_rate": 4.169093556607208e-07, + "loss": 0.0885, + "step": 47355 + }, + { + "epoch": 8.61, + "learning_rate": 4.168547506370586e-07, + "loss": 0.0547, + "step": 47356 + }, + { + "epoch": 8.61, + "learning_rate": 4.168001456133964e-07, + "loss": 0.8998, + "step": 47357 + }, + { + "epoch": 8.61, + "learning_rate": 4.167455405897343e-07, + "loss": 0.0489, + "step": 47358 + }, + { + "epoch": 8.61, + "learning_rate": 4.1669093556607213e-07, + "loss": 1.0084, + "step": 47359 + }, + { + "epoch": 8.61, + "learning_rate": 4.166363305424099e-07, + "loss": 0.1466, + "step": 47360 + }, + { + "epoch": 8.61, + "learning_rate": 4.1658172551874775e-07, + "loss": 1.9016, + "step": 47361 + }, + { + "epoch": 8.61, + "learning_rate": 4.1652712049508554e-07, + "loss": 0.1115, + "step": 47362 + }, + { + "epoch": 8.61, + "learning_rate": 4.164725154714234e-07, + "loss": 0.1879, + "step": 47363 + }, + { + "epoch": 8.61, + "learning_rate": 4.1641791044776116e-07, + "loss": 0.0088, + "step": 47364 + }, + { + "epoch": 8.61, + "learning_rate": 4.1636330542409905e-07, + "loss": 0.0153, + "step": 47365 + }, + { + "epoch": 8.61, + "learning_rate": 4.1630870040043683e-07, + "loss": 0.4764, + "step": 47366 + }, + { + "epoch": 8.61, + "learning_rate": 4.1625409537677467e-07, + "loss": 2.9577, + "step": 47367 + }, + { + "epoch": 8.61, + "learning_rate": 4.161994903531125e-07, + "loss": 0.4405, + "step": 47368 + }, + { + "epoch": 8.61, + "learning_rate": 4.161448853294503e-07, + "loss": 1.5702, + "step": 47369 + }, + { + "epoch": 8.61, + "learning_rate": 4.160902803057882e-07, + "loss": 0.2951, + "step": 47370 + }, + { + "epoch": 8.61, + "learning_rate": 4.160356752821259e-07, + "loss": 0.0662, + "step": 47371 + }, + { + "epoch": 8.61, + "learning_rate": 4.159810702584638e-07, + "loss": 0.874, + "step": 47372 + }, + { + "epoch": 8.61, + "learning_rate": 4.159264652348016e-07, + "loss": 0.0357, + "step": 47373 + }, + { + "epoch": 8.62, + "learning_rate": 4.1587186021113943e-07, + "loss": 0.7588, + "step": 47374 + }, + { + "epoch": 8.62, + "learning_rate": 4.1581725518747727e-07, + "loss": 0.3727, + "step": 47375 + }, + { + "epoch": 8.62, + "learning_rate": 4.1576265016381505e-07, + "loss": 0.0139, + "step": 47376 + }, + { + "epoch": 8.62, + "learning_rate": 4.1570804514015294e-07, + "loss": 0.0443, + "step": 47377 + }, + { + "epoch": 8.62, + "learning_rate": 4.156534401164907e-07, + "loss": 0.1328, + "step": 47378 + }, + { + "epoch": 8.62, + "learning_rate": 4.1559883509282856e-07, + "loss": 0.2948, + "step": 47379 + }, + { + "epoch": 8.62, + "learning_rate": 4.1554423006916635e-07, + "loss": 1.2356, + "step": 47380 + }, + { + "epoch": 8.62, + "learning_rate": 4.154896250455042e-07, + "loss": 0.5932, + "step": 47381 + }, + { + "epoch": 8.62, + "learning_rate": 4.154350200218421e-07, + "loss": 0.006, + "step": 47382 + }, + { + "epoch": 8.62, + "learning_rate": 4.153804149981798e-07, + "loss": 0.4318, + "step": 47383 + }, + { + "epoch": 8.62, + "learning_rate": 4.153258099745177e-07, + "loss": 0.0837, + "step": 47384 + }, + { + "epoch": 8.62, + "learning_rate": 4.152712049508555e-07, + "loss": 0.085, + "step": 47385 + }, + { + "epoch": 8.62, + "learning_rate": 4.152165999271933e-07, + "loss": 1.1855, + "step": 47386 + }, + { + "epoch": 8.62, + "learning_rate": 4.151619949035311e-07, + "loss": 0.0279, + "step": 47387 + }, + { + "epoch": 8.62, + "learning_rate": 4.1510738987986894e-07, + "loss": 0.0634, + "step": 47388 + }, + { + "epoch": 8.62, + "learning_rate": 4.1505278485620683e-07, + "loss": 2.7642, + "step": 47389 + }, + { + "epoch": 8.62, + "learning_rate": 4.149981798325446e-07, + "loss": 2.259, + "step": 47390 + }, + { + "epoch": 8.62, + "learning_rate": 4.1494357480888245e-07, + "loss": 0.1583, + "step": 47391 + }, + { + "epoch": 8.62, + "learning_rate": 4.1488896978522024e-07, + "loss": 0.0061, + "step": 47392 + }, + { + "epoch": 8.62, + "learning_rate": 4.148343647615581e-07, + "loss": 0.0538, + "step": 47393 + }, + { + "epoch": 8.62, + "learning_rate": 4.1477975973789586e-07, + "loss": 0.8724, + "step": 47394 + }, + { + "epoch": 8.62, + "learning_rate": 4.147251547142337e-07, + "loss": 0.7733, + "step": 47395 + }, + { + "epoch": 8.62, + "learning_rate": 4.146705496905716e-07, + "loss": 0.0454, + "step": 47396 + }, + { + "epoch": 8.62, + "learning_rate": 4.1461594466690937e-07, + "loss": 0.0082, + "step": 47397 + }, + { + "epoch": 8.62, + "learning_rate": 4.145613396432472e-07, + "loss": 1.2165, + "step": 47398 + }, + { + "epoch": 8.62, + "learning_rate": 4.14506734619585e-07, + "loss": 0.9922, + "step": 47399 + }, + { + "epoch": 8.62, + "learning_rate": 4.1445212959592283e-07, + "loss": 0.0913, + "step": 47400 + }, + { + "epoch": 8.62, + "learning_rate": 4.143975245722606e-07, + "loss": 0.0394, + "step": 47401 + }, + { + "epoch": 8.62, + "learning_rate": 4.143429195485985e-07, + "loss": 0.42, + "step": 47402 + }, + { + "epoch": 8.62, + "learning_rate": 4.1428831452493635e-07, + "loss": 0.0979, + "step": 47403 + }, + { + "epoch": 8.62, + "learning_rate": 4.1423370950127413e-07, + "loss": 0.0069, + "step": 47404 + }, + { + "epoch": 8.62, + "learning_rate": 4.1417910447761197e-07, + "loss": 0.7972, + "step": 47405 + }, + { + "epoch": 8.62, + "learning_rate": 4.1412449945394975e-07, + "loss": 0.0137, + "step": 47406 + }, + { + "epoch": 8.62, + "learning_rate": 4.140698944302876e-07, + "loss": 1.3172, + "step": 47407 + }, + { + "epoch": 8.62, + "learning_rate": 4.140152894066254e-07, + "loss": 0.0556, + "step": 47408 + }, + { + "epoch": 8.62, + "learning_rate": 4.1396068438296326e-07, + "loss": 0.4351, + "step": 47409 + }, + { + "epoch": 8.62, + "learning_rate": 4.1390607935930105e-07, + "loss": 0.0239, + "step": 47410 + }, + { + "epoch": 8.62, + "learning_rate": 4.138514743356389e-07, + "loss": 0.0023, + "step": 47411 + }, + { + "epoch": 8.62, + "learning_rate": 4.137968693119767e-07, + "loss": 1.3784, + "step": 47412 + }, + { + "epoch": 8.62, + "learning_rate": 4.137422642883145e-07, + "loss": 0.324, + "step": 47413 + }, + { + "epoch": 8.62, + "learning_rate": 4.136876592646524e-07, + "loss": 1.8398, + "step": 47414 + }, + { + "epoch": 8.62, + "learning_rate": 4.1363305424099013e-07, + "loss": 0.1325, + "step": 47415 + }, + { + "epoch": 8.62, + "learning_rate": 4.13578449217328e-07, + "loss": 0.0552, + "step": 47416 + }, + { + "epoch": 8.62, + "learning_rate": 4.135238441936658e-07, + "loss": 0.1635, + "step": 47417 + }, + { + "epoch": 8.62, + "learning_rate": 4.1346923917000364e-07, + "loss": 0.3343, + "step": 47418 + }, + { + "epoch": 8.62, + "learning_rate": 4.1341463414634153e-07, + "loss": 1.1896, + "step": 47419 + }, + { + "epoch": 8.62, + "learning_rate": 4.1336002912267927e-07, + "loss": 0.0623, + "step": 47420 + }, + { + "epoch": 8.62, + "learning_rate": 4.1330542409901716e-07, + "loss": 0.0065, + "step": 47421 + }, + { + "epoch": 8.62, + "learning_rate": 4.1325081907535494e-07, + "loss": 0.0049, + "step": 47422 + }, + { + "epoch": 8.62, + "learning_rate": 4.131962140516928e-07, + "loss": 0.2263, + "step": 47423 + }, + { + "epoch": 8.62, + "learning_rate": 4.1314160902803056e-07, + "loss": 0.0294, + "step": 47424 + }, + { + "epoch": 8.62, + "learning_rate": 4.130870040043684e-07, + "loss": 0.0151, + "step": 47425 + }, + { + "epoch": 8.62, + "learning_rate": 4.130323989807063e-07, + "loss": 2.3614, + "step": 47426 + }, + { + "epoch": 8.62, + "learning_rate": 4.12977793957044e-07, + "loss": 0.0788, + "step": 47427 + }, + { + "epoch": 8.62, + "learning_rate": 4.129231889333819e-07, + "loss": 0.6455, + "step": 47428 + }, + { + "epoch": 8.63, + "learning_rate": 4.128685839097197e-07, + "loss": 0.624, + "step": 47429 + }, + { + "epoch": 8.63, + "learning_rate": 4.1281397888605753e-07, + "loss": 0.2938, + "step": 47430 + }, + { + "epoch": 8.63, + "learning_rate": 4.127593738623953e-07, + "loss": 2.5736, + "step": 47431 + }, + { + "epoch": 8.63, + "learning_rate": 4.1270476883873316e-07, + "loss": 0.0783, + "step": 47432 + }, + { + "epoch": 8.63, + "learning_rate": 4.1265016381507105e-07, + "loss": 0.004, + "step": 47433 + }, + { + "epoch": 8.63, + "learning_rate": 4.1259555879140883e-07, + "loss": 0.0099, + "step": 47434 + }, + { + "epoch": 8.63, + "learning_rate": 4.1254095376774667e-07, + "loss": 0.5708, + "step": 47435 + }, + { + "epoch": 8.63, + "learning_rate": 4.1248634874408445e-07, + "loss": 0.1124, + "step": 47436 + }, + { + "epoch": 8.63, + "learning_rate": 4.124317437204223e-07, + "loss": 0.0802, + "step": 47437 + }, + { + "epoch": 8.63, + "learning_rate": 4.123771386967601e-07, + "loss": 2.441, + "step": 47438 + }, + { + "epoch": 8.63, + "learning_rate": 4.123225336730979e-07, + "loss": 0.0722, + "step": 47439 + }, + { + "epoch": 8.63, + "learning_rate": 4.122679286494358e-07, + "loss": 0.4886, + "step": 47440 + }, + { + "epoch": 8.63, + "learning_rate": 4.122133236257736e-07, + "loss": 0.0059, + "step": 47441 + }, + { + "epoch": 8.63, + "learning_rate": 4.121587186021114e-07, + "loss": 0.1144, + "step": 47442 + }, + { + "epoch": 8.63, + "learning_rate": 4.121041135784492e-07, + "loss": 0.9465, + "step": 47443 + }, + { + "epoch": 8.63, + "learning_rate": 4.1204950855478705e-07, + "loss": 0.0114, + "step": 47444 + }, + { + "epoch": 8.63, + "learning_rate": 4.1199490353112483e-07, + "loss": 0.0117, + "step": 47445 + }, + { + "epoch": 8.63, + "learning_rate": 4.119402985074627e-07, + "loss": 0.1023, + "step": 47446 + }, + { + "epoch": 8.63, + "learning_rate": 4.1188569348380045e-07, + "loss": 0.103, + "step": 47447 + }, + { + "epoch": 8.63, + "learning_rate": 4.1183108846013835e-07, + "loss": 0.0719, + "step": 47448 + }, + { + "epoch": 8.63, + "learning_rate": 4.117764834364762e-07, + "loss": 0.159, + "step": 47449 + }, + { + "epoch": 8.63, + "learning_rate": 4.1172187841281397e-07, + "loss": 0.0628, + "step": 47450 + }, + { + "epoch": 8.63, + "learning_rate": 4.1166727338915186e-07, + "loss": 0.0432, + "step": 47451 + }, + { + "epoch": 8.63, + "learning_rate": 4.116126683654896e-07, + "loss": 0.0796, + "step": 47452 + }, + { + "epoch": 8.63, + "learning_rate": 4.115580633418275e-07, + "loss": 0.0065, + "step": 47453 + }, + { + "epoch": 8.63, + "learning_rate": 4.1150345831816526e-07, + "loss": 1.6475, + "step": 47454 + }, + { + "epoch": 8.63, + "learning_rate": 4.114488532945031e-07, + "loss": 0.0076, + "step": 47455 + }, + { + "epoch": 8.63, + "learning_rate": 4.1139424827084094e-07, + "loss": 1.29, + "step": 47456 + }, + { + "epoch": 8.63, + "learning_rate": 4.113396432471787e-07, + "loss": 0.1054, + "step": 47457 + }, + { + "epoch": 8.63, + "learning_rate": 4.112850382235166e-07, + "loss": 0.0205, + "step": 47458 + }, + { + "epoch": 8.63, + "learning_rate": 4.1123043319985435e-07, + "loss": 0.0223, + "step": 47459 + }, + { + "epoch": 8.63, + "learning_rate": 4.1117582817619224e-07, + "loss": 0.8223, + "step": 47460 + }, + { + "epoch": 8.63, + "learning_rate": 4.1112122315253e-07, + "loss": 0.0254, + "step": 47461 + }, + { + "epoch": 8.63, + "learning_rate": 4.1106661812886786e-07, + "loss": 0.0739, + "step": 47462 + }, + { + "epoch": 8.63, + "learning_rate": 4.1101201310520575e-07, + "loss": 0.5655, + "step": 47463 + }, + { + "epoch": 8.63, + "learning_rate": 4.109574080815435e-07, + "loss": 0.0658, + "step": 47464 + }, + { + "epoch": 8.63, + "learning_rate": 4.1090280305788137e-07, + "loss": 0.1175, + "step": 47465 + }, + { + "epoch": 8.63, + "learning_rate": 4.1084819803421916e-07, + "loss": 0.005, + "step": 47466 + }, + { + "epoch": 8.63, + "learning_rate": 4.10793593010557e-07, + "loss": 0.0237, + "step": 47467 + }, + { + "epoch": 8.63, + "learning_rate": 4.107389879868948e-07, + "loss": 0.9763, + "step": 47468 + }, + { + "epoch": 8.63, + "learning_rate": 4.106843829632326e-07, + "loss": 0.0708, + "step": 47469 + }, + { + "epoch": 8.63, + "learning_rate": 4.106297779395705e-07, + "loss": 4.4837, + "step": 47470 + }, + { + "epoch": 8.63, + "learning_rate": 4.105751729159083e-07, + "loss": 0.036, + "step": 47471 + }, + { + "epoch": 8.63, + "learning_rate": 4.1052056789224613e-07, + "loss": 1.3521, + "step": 47472 + }, + { + "epoch": 8.63, + "learning_rate": 4.104659628685839e-07, + "loss": 0.1221, + "step": 47473 + }, + { + "epoch": 8.63, + "learning_rate": 4.1041135784492175e-07, + "loss": 0.0472, + "step": 47474 + }, + { + "epoch": 8.63, + "learning_rate": 4.1035675282125953e-07, + "loss": 1.127, + "step": 47475 + }, + { + "epoch": 8.63, + "learning_rate": 4.1030214779759737e-07, + "loss": 1.0495, + "step": 47476 + }, + { + "epoch": 8.63, + "learning_rate": 4.1024754277393526e-07, + "loss": 0.6347, + "step": 47477 + }, + { + "epoch": 8.63, + "learning_rate": 4.1019293775027305e-07, + "loss": 2.3284, + "step": 47478 + }, + { + "epoch": 8.63, + "learning_rate": 4.101383327266109e-07, + "loss": 0.0576, + "step": 47479 + }, + { + "epoch": 8.63, + "learning_rate": 4.1008372770294867e-07, + "loss": 0.0461, + "step": 47480 + }, + { + "epoch": 8.63, + "learning_rate": 4.100291226792865e-07, + "loss": 0.4396, + "step": 47481 + }, + { + "epoch": 8.63, + "learning_rate": 4.099745176556243e-07, + "loss": 2.1928, + "step": 47482 + }, + { + "epoch": 8.63, + "learning_rate": 4.099199126319622e-07, + "loss": 1.3717, + "step": 47483 + }, + { + "epoch": 8.64, + "learning_rate": 4.098653076082999e-07, + "loss": 1.8189, + "step": 47484 + }, + { + "epoch": 8.64, + "learning_rate": 4.098107025846378e-07, + "loss": 0.01, + "step": 47485 + }, + { + "epoch": 8.64, + "learning_rate": 4.0975609756097564e-07, + "loss": 0.005, + "step": 47486 + }, + { + "epoch": 8.64, + "learning_rate": 4.097014925373134e-07, + "loss": 0.0257, + "step": 47487 + }, + { + "epoch": 8.64, + "learning_rate": 4.0964688751365126e-07, + "loss": 0.1807, + "step": 47488 + }, + { + "epoch": 8.64, + "learning_rate": 4.0959228248998905e-07, + "loss": 0.0554, + "step": 47489 + }, + { + "epoch": 8.64, + "learning_rate": 4.0953767746632694e-07, + "loss": 0.0107, + "step": 47490 + }, + { + "epoch": 8.64, + "learning_rate": 4.0948307244266467e-07, + "loss": 0.0105, + "step": 47491 + }, + { + "epoch": 8.64, + "learning_rate": 4.0942846741900256e-07, + "loss": 0.1332, + "step": 47492 + }, + { + "epoch": 8.64, + "learning_rate": 4.093738623953404e-07, + "loss": 0.5478, + "step": 47493 + }, + { + "epoch": 8.64, + "learning_rate": 4.093192573716782e-07, + "loss": 0.3139, + "step": 47494 + }, + { + "epoch": 8.64, + "learning_rate": 4.0926465234801607e-07, + "loss": 0.0026, + "step": 47495 + }, + { + "epoch": 8.64, + "learning_rate": 4.092100473243538e-07, + "loss": 0.0029, + "step": 47496 + }, + { + "epoch": 8.64, + "learning_rate": 4.091554423006917e-07, + "loss": 0.7553, + "step": 47497 + }, + { + "epoch": 8.64, + "learning_rate": 4.091008372770295e-07, + "loss": 0.0116, + "step": 47498 + }, + { + "epoch": 8.64, + "learning_rate": 4.090462322533673e-07, + "loss": 0.034, + "step": 47499 + }, + { + "epoch": 8.64, + "learning_rate": 4.0899162722970516e-07, + "loss": 0.3007, + "step": 47500 + }, + { + "epoch": 8.64, + "learning_rate": 4.0893702220604294e-07, + "loss": 0.128, + "step": 47501 + }, + { + "epoch": 8.64, + "learning_rate": 4.0888241718238083e-07, + "loss": 0.06, + "step": 47502 + }, + { + "epoch": 8.64, + "learning_rate": 4.088278121587186e-07, + "loss": 0.4708, + "step": 47503 + }, + { + "epoch": 8.64, + "learning_rate": 4.0877320713505645e-07, + "loss": 0.2856, + "step": 47504 + }, + { + "epoch": 8.64, + "learning_rate": 4.0871860211139424e-07, + "loss": 0.6187, + "step": 47505 + }, + { + "epoch": 8.64, + "learning_rate": 4.086639970877321e-07, + "loss": 0.1563, + "step": 47506 + }, + { + "epoch": 8.64, + "learning_rate": 4.0860939206406996e-07, + "loss": 0.7417, + "step": 47507 + }, + { + "epoch": 8.64, + "learning_rate": 4.085547870404077e-07, + "loss": 0.038, + "step": 47508 + }, + { + "epoch": 8.64, + "learning_rate": 4.085001820167456e-07, + "loss": 0.0506, + "step": 47509 + }, + { + "epoch": 8.64, + "learning_rate": 4.0844557699308337e-07, + "loss": 0.0381, + "step": 47510 + }, + { + "epoch": 8.64, + "learning_rate": 4.083909719694212e-07, + "loss": 1.6998, + "step": 47511 + }, + { + "epoch": 8.64, + "learning_rate": 4.08336366945759e-07, + "loss": 0.1232, + "step": 47512 + }, + { + "epoch": 8.64, + "learning_rate": 4.0828176192209683e-07, + "loss": 0.1704, + "step": 47513 + }, + { + "epoch": 8.64, + "learning_rate": 4.082271568984347e-07, + "loss": 0.8076, + "step": 47514 + }, + { + "epoch": 8.64, + "learning_rate": 4.081725518747725e-07, + "loss": 1.2573, + "step": 47515 + }, + { + "epoch": 8.64, + "learning_rate": 4.0811794685111034e-07, + "loss": 0.6613, + "step": 47516 + }, + { + "epoch": 8.64, + "learning_rate": 4.0806334182744813e-07, + "loss": 2.5712, + "step": 47517 + }, + { + "epoch": 8.64, + "learning_rate": 4.0800873680378597e-07, + "loss": 0.2968, + "step": 47518 + }, + { + "epoch": 8.64, + "learning_rate": 4.0795413178012375e-07, + "loss": 0.0484, + "step": 47519 + }, + { + "epoch": 8.64, + "learning_rate": 4.078995267564616e-07, + "loss": 0.8039, + "step": 47520 + }, + { + "epoch": 8.64, + "learning_rate": 4.078449217327995e-07, + "loss": 0.5406, + "step": 47521 + }, + { + "epoch": 8.64, + "learning_rate": 4.0779031670913726e-07, + "loss": 0.0796, + "step": 47522 + }, + { + "epoch": 8.64, + "learning_rate": 4.077357116854751e-07, + "loss": 0.0055, + "step": 47523 + }, + { + "epoch": 8.64, + "learning_rate": 4.076811066618129e-07, + "loss": 0.4161, + "step": 47524 + }, + { + "epoch": 8.64, + "learning_rate": 4.076265016381507e-07, + "loss": 0.0046, + "step": 47525 + }, + { + "epoch": 8.64, + "learning_rate": 4.075718966144885e-07, + "loss": 1.4965, + "step": 47526 + }, + { + "epoch": 8.64, + "learning_rate": 4.075172915908264e-07, + "loss": 0.009, + "step": 47527 + }, + { + "epoch": 8.64, + "learning_rate": 4.0746268656716413e-07, + "loss": 0.9037, + "step": 47528 + }, + { + "epoch": 8.64, + "learning_rate": 4.07408081543502e-07, + "loss": 0.474, + "step": 47529 + }, + { + "epoch": 8.64, + "learning_rate": 4.0735347651983986e-07, + "loss": 2.4388, + "step": 47530 + }, + { + "epoch": 8.64, + "learning_rate": 4.0729887149617764e-07, + "loss": 1.114, + "step": 47531 + }, + { + "epoch": 8.64, + "learning_rate": 4.072442664725155e-07, + "loss": 1.307, + "step": 47532 + }, + { + "epoch": 8.64, + "learning_rate": 4.0718966144885326e-07, + "loss": 0.6403, + "step": 47533 + }, + { + "epoch": 8.64, + "learning_rate": 4.0713505642519115e-07, + "loss": 0.4298, + "step": 47534 + }, + { + "epoch": 8.64, + "learning_rate": 4.0708045140152894e-07, + "loss": 0.7512, + "step": 47535 + }, + { + "epoch": 8.64, + "learning_rate": 4.070258463778668e-07, + "loss": 0.1284, + "step": 47536 + }, + { + "epoch": 8.64, + "learning_rate": 4.069712413542046e-07, + "loss": 0.4445, + "step": 47537 + }, + { + "epoch": 8.64, + "learning_rate": 4.069166363305424e-07, + "loss": 0.6908, + "step": 47538 + }, + { + "epoch": 8.65, + "learning_rate": 4.068620313068803e-07, + "loss": 0.4582, + "step": 47539 + }, + { + "epoch": 8.65, + "learning_rate": 4.06807426283218e-07, + "loss": 1.2756, + "step": 47540 + }, + { + "epoch": 8.65, + "learning_rate": 4.067528212595559e-07, + "loss": 0.4091, + "step": 47541 + }, + { + "epoch": 8.65, + "learning_rate": 4.066982162358937e-07, + "loss": 0.4428, + "step": 47542 + }, + { + "epoch": 8.65, + "learning_rate": 4.0664361121223153e-07, + "loss": 0.5822, + "step": 47543 + }, + { + "epoch": 8.65, + "learning_rate": 4.0658900618856937e-07, + "loss": 0.09, + "step": 47544 + }, + { + "epoch": 8.65, + "learning_rate": 4.0653440116490716e-07, + "loss": 0.0566, + "step": 47545 + }, + { + "epoch": 8.65, + "learning_rate": 4.0647979614124505e-07, + "loss": 0.9063, + "step": 47546 + }, + { + "epoch": 8.65, + "learning_rate": 4.0642519111758283e-07, + "loss": 0.0103, + "step": 47547 + }, + { + "epoch": 8.65, + "learning_rate": 4.0637058609392067e-07, + "loss": 0.1183, + "step": 47548 + }, + { + "epoch": 8.65, + "learning_rate": 4.0631598107025845e-07, + "loss": 0.0076, + "step": 47549 + }, + { + "epoch": 8.65, + "learning_rate": 4.062613760465963e-07, + "loss": 0.0048, + "step": 47550 + }, + { + "epoch": 8.65, + "learning_rate": 4.062067710229342e-07, + "loss": 1.4533, + "step": 47551 + }, + { + "epoch": 8.65, + "learning_rate": 4.061521659992719e-07, + "loss": 0.2396, + "step": 47552 + }, + { + "epoch": 8.65, + "learning_rate": 4.060975609756098e-07, + "loss": 0.0138, + "step": 47553 + }, + { + "epoch": 8.65, + "learning_rate": 4.060429559519476e-07, + "loss": 0.3569, + "step": 47554 + }, + { + "epoch": 8.65, + "learning_rate": 4.059883509282854e-07, + "loss": 0.2495, + "step": 47555 + }, + { + "epoch": 8.65, + "learning_rate": 4.059337459046232e-07, + "loss": 0.0077, + "step": 47556 + }, + { + "epoch": 8.65, + "learning_rate": 4.0587914088096105e-07, + "loss": 0.1113, + "step": 47557 + }, + { + "epoch": 8.65, + "learning_rate": 4.0582453585729894e-07, + "loss": 0.0657, + "step": 47558 + }, + { + "epoch": 8.65, + "learning_rate": 4.057699308336367e-07, + "loss": 0.0023, + "step": 47559 + }, + { + "epoch": 8.65, + "learning_rate": 4.0571532580997456e-07, + "loss": 1.4275, + "step": 47560 + }, + { + "epoch": 8.65, + "learning_rate": 4.0566072078631234e-07, + "loss": 0.062, + "step": 47561 + }, + { + "epoch": 8.65, + "learning_rate": 4.056061157626502e-07, + "loss": 0.0103, + "step": 47562 + }, + { + "epoch": 8.65, + "learning_rate": 4.0555151073898797e-07, + "loss": 0.0833, + "step": 47563 + }, + { + "epoch": 8.65, + "learning_rate": 4.054969057153258e-07, + "loss": 0.0111, + "step": 47564 + }, + { + "epoch": 8.65, + "learning_rate": 4.054423006916636e-07, + "loss": 0.2142, + "step": 47565 + }, + { + "epoch": 8.65, + "learning_rate": 4.053876956680015e-07, + "loss": 0.0603, + "step": 47566 + }, + { + "epoch": 8.65, + "learning_rate": 4.053330906443393e-07, + "loss": 0.0677, + "step": 47567 + }, + { + "epoch": 8.65, + "learning_rate": 4.052784856206771e-07, + "loss": 1.4475, + "step": 47568 + }, + { + "epoch": 8.65, + "learning_rate": 4.0522388059701494e-07, + "loss": 0.7531, + "step": 47569 + }, + { + "epoch": 8.65, + "learning_rate": 4.051692755733527e-07, + "loss": 1.3361, + "step": 47570 + }, + { + "epoch": 8.65, + "learning_rate": 4.051146705496906e-07, + "loss": 0.0071, + "step": 47571 + }, + { + "epoch": 8.65, + "learning_rate": 4.0506006552602834e-07, + "loss": 1.1275, + "step": 47572 + }, + { + "epoch": 8.65, + "learning_rate": 4.0500546050236623e-07, + "loss": 0.4082, + "step": 47573 + }, + { + "epoch": 8.65, + "learning_rate": 4.0495085547870407e-07, + "loss": 0.1085, + "step": 47574 + }, + { + "epoch": 8.65, + "learning_rate": 4.0489625045504186e-07, + "loss": 0.02, + "step": 47575 + }, + { + "epoch": 8.65, + "learning_rate": 4.048416454313797e-07, + "loss": 0.2793, + "step": 47576 + }, + { + "epoch": 8.65, + "learning_rate": 4.047870404077175e-07, + "loss": 0.0959, + "step": 47577 + }, + { + "epoch": 8.65, + "learning_rate": 4.0473243538405537e-07, + "loss": 0.5534, + "step": 47578 + }, + { + "epoch": 8.65, + "learning_rate": 4.0467783036039315e-07, + "loss": 0.5629, + "step": 47579 + }, + { + "epoch": 8.65, + "learning_rate": 4.04623225336731e-07, + "loss": 2.9354, + "step": 47580 + }, + { + "epoch": 8.65, + "learning_rate": 4.0456862031306883e-07, + "loss": 0.0553, + "step": 47581 + }, + { + "epoch": 8.65, + "learning_rate": 4.045140152894066e-07, + "loss": 0.0284, + "step": 47582 + }, + { + "epoch": 8.65, + "learning_rate": 4.044594102657445e-07, + "loss": 0.0058, + "step": 47583 + }, + { + "epoch": 8.65, + "learning_rate": 4.0440480524208224e-07, + "loss": 0.0613, + "step": 47584 + }, + { + "epoch": 8.65, + "learning_rate": 4.043502002184201e-07, + "loss": 0.0514, + "step": 47585 + }, + { + "epoch": 8.65, + "learning_rate": 4.042955951947579e-07, + "loss": 0.7868, + "step": 47586 + }, + { + "epoch": 8.65, + "learning_rate": 4.0424099017109575e-07, + "loss": 1.138, + "step": 47587 + }, + { + "epoch": 8.65, + "learning_rate": 4.0418638514743364e-07, + "loss": 0.1415, + "step": 47588 + }, + { + "epoch": 8.65, + "learning_rate": 4.0413178012377137e-07, + "loss": 0.1878, + "step": 47589 + }, + { + "epoch": 8.65, + "learning_rate": 4.0407717510010926e-07, + "loss": 0.7364, + "step": 47590 + }, + { + "epoch": 8.65, + "learning_rate": 4.0402257007644705e-07, + "loss": 0.0082, + "step": 47591 + }, + { + "epoch": 8.65, + "learning_rate": 4.039679650527849e-07, + "loss": 0.0187, + "step": 47592 + }, + { + "epoch": 8.65, + "learning_rate": 4.0391336002912267e-07, + "loss": 0.3638, + "step": 47593 + }, + { + "epoch": 8.66, + "learning_rate": 4.038587550054605e-07, + "loss": 0.2664, + "step": 47594 + }, + { + "epoch": 8.66, + "learning_rate": 4.038041499817984e-07, + "loss": 0.0067, + "step": 47595 + }, + { + "epoch": 8.66, + "learning_rate": 4.0374954495813613e-07, + "loss": 0.0449, + "step": 47596 + }, + { + "epoch": 8.66, + "learning_rate": 4.03694939934474e-07, + "loss": 1.5113, + "step": 47597 + }, + { + "epoch": 8.66, + "learning_rate": 4.036403349108118e-07, + "loss": 0.3301, + "step": 47598 + }, + { + "epoch": 8.66, + "learning_rate": 4.0358572988714964e-07, + "loss": 0.0045, + "step": 47599 + }, + { + "epoch": 8.66, + "learning_rate": 4.035311248634874e-07, + "loss": 0.1231, + "step": 47600 + }, + { + "epoch": 8.66, + "learning_rate": 4.0347651983982526e-07, + "loss": 0.7918, + "step": 47601 + }, + { + "epoch": 8.66, + "learning_rate": 4.0342191481616305e-07, + "loss": 0.8989, + "step": 47602 + }, + { + "epoch": 8.66, + "learning_rate": 4.0336730979250094e-07, + "loss": 0.0085, + "step": 47603 + }, + { + "epoch": 8.66, + "learning_rate": 4.033127047688388e-07, + "loss": 0.0441, + "step": 47604 + }, + { + "epoch": 8.66, + "learning_rate": 4.0325809974517656e-07, + "loss": 0.3427, + "step": 47605 + }, + { + "epoch": 8.66, + "learning_rate": 4.032034947215144e-07, + "loss": 0.1967, + "step": 47606 + }, + { + "epoch": 8.66, + "learning_rate": 4.031488896978522e-07, + "loss": 0.1802, + "step": 47607 + }, + { + "epoch": 8.66, + "learning_rate": 4.0309428467419e-07, + "loss": 0.7762, + "step": 47608 + }, + { + "epoch": 8.66, + "learning_rate": 4.030396796505278e-07, + "loss": 0.9693, + "step": 47609 + }, + { + "epoch": 8.66, + "learning_rate": 4.029850746268657e-07, + "loss": 0.0172, + "step": 47610 + }, + { + "epoch": 8.66, + "learning_rate": 4.0293046960320353e-07, + "loss": 1.4978, + "step": 47611 + }, + { + "epoch": 8.66, + "learning_rate": 4.028758645795413e-07, + "loss": 0.3438, + "step": 47612 + }, + { + "epoch": 8.66, + "learning_rate": 4.0282125955587915e-07, + "loss": 3.796, + "step": 47613 + }, + { + "epoch": 8.66, + "learning_rate": 4.0276665453221694e-07, + "loss": 0.6154, + "step": 47614 + }, + { + "epoch": 8.66, + "learning_rate": 4.0271204950855483e-07, + "loss": 0.1454, + "step": 47615 + }, + { + "epoch": 8.66, + "learning_rate": 4.0265744448489256e-07, + "loss": 0.9474, + "step": 47616 + }, + { + "epoch": 8.66, + "learning_rate": 4.0260283946123045e-07, + "loss": 0.2189, + "step": 47617 + }, + { + "epoch": 8.66, + "learning_rate": 4.025482344375683e-07, + "loss": 0.0747, + "step": 47618 + }, + { + "epoch": 8.66, + "learning_rate": 4.0249362941390607e-07, + "loss": 0.0137, + "step": 47619 + }, + { + "epoch": 8.66, + "learning_rate": 4.0243902439024396e-07, + "loss": 0.0139, + "step": 47620 + }, + { + "epoch": 8.66, + "learning_rate": 4.023844193665817e-07, + "loss": 0.4479, + "step": 47621 + }, + { + "epoch": 8.66, + "learning_rate": 4.023298143429196e-07, + "loss": 1.6222, + "step": 47622 + }, + { + "epoch": 8.66, + "learning_rate": 4.0227520931925737e-07, + "loss": 0.9397, + "step": 47623 + }, + { + "epoch": 8.66, + "learning_rate": 4.022206042955952e-07, + "loss": 0.0304, + "step": 47624 + }, + { + "epoch": 8.66, + "learning_rate": 4.0216599927193304e-07, + "loss": 0.9257, + "step": 47625 + }, + { + "epoch": 8.66, + "learning_rate": 4.0211139424827083e-07, + "loss": 0.1829, + "step": 47626 + }, + { + "epoch": 8.66, + "learning_rate": 4.020567892246087e-07, + "loss": 1.6464, + "step": 47627 + }, + { + "epoch": 8.66, + "learning_rate": 4.0200218420094645e-07, + "loss": 0.1366, + "step": 47628 + }, + { + "epoch": 8.66, + "learning_rate": 4.0194757917728434e-07, + "loss": 1.5056, + "step": 47629 + }, + { + "epoch": 8.66, + "learning_rate": 4.018929741536221e-07, + "loss": 1.3036, + "step": 47630 + }, + { + "epoch": 8.66, + "learning_rate": 4.0183836912995996e-07, + "loss": 0.1367, + "step": 47631 + }, + { + "epoch": 8.66, + "learning_rate": 4.0178376410629785e-07, + "loss": 0.3103, + "step": 47632 + }, + { + "epoch": 8.66, + "learning_rate": 4.017291590826356e-07, + "loss": 0.5094, + "step": 47633 + }, + { + "epoch": 8.66, + "learning_rate": 4.016745540589735e-07, + "loss": 0.5974, + "step": 47634 + }, + { + "epoch": 8.66, + "learning_rate": 4.0161994903531126e-07, + "loss": 0.3258, + "step": 47635 + }, + { + "epoch": 8.66, + "learning_rate": 4.015653440116491e-07, + "loss": 0.5629, + "step": 47636 + }, + { + "epoch": 8.66, + "learning_rate": 4.015107389879869e-07, + "loss": 0.0088, + "step": 47637 + }, + { + "epoch": 8.66, + "learning_rate": 4.014561339643247e-07, + "loss": 0.0606, + "step": 47638 + }, + { + "epoch": 8.66, + "learning_rate": 4.014015289406625e-07, + "loss": 0.27, + "step": 47639 + }, + { + "epoch": 8.66, + "learning_rate": 4.0134692391700034e-07, + "loss": 0.7011, + "step": 47640 + }, + { + "epoch": 8.66, + "learning_rate": 4.0129231889333823e-07, + "loss": 1.4107, + "step": 47641 + }, + { + "epoch": 8.66, + "learning_rate": 4.01237713869676e-07, + "loss": 0.1706, + "step": 47642 + }, + { + "epoch": 8.66, + "learning_rate": 4.0118310884601386e-07, + "loss": 0.0206, + "step": 47643 + }, + { + "epoch": 8.66, + "learning_rate": 4.0112850382235164e-07, + "loss": 1.1455, + "step": 47644 + }, + { + "epoch": 8.66, + "learning_rate": 4.010738987986895e-07, + "loss": 1.874, + "step": 47645 + }, + { + "epoch": 8.66, + "learning_rate": 4.0101929377502726e-07, + "loss": 0.167, + "step": 47646 + }, + { + "epoch": 8.66, + "learning_rate": 4.0096468875136515e-07, + "loss": 0.8709, + "step": 47647 + }, + { + "epoch": 8.66, + "learning_rate": 4.00910083727703e-07, + "loss": 0.0058, + "step": 47648 + }, + { + "epoch": 8.67, + "learning_rate": 4.008554787040408e-07, + "loss": 0.3044, + "step": 47649 + }, + { + "epoch": 8.67, + "learning_rate": 4.008008736803786e-07, + "loss": 0.8185, + "step": 47650 + }, + { + "epoch": 8.67, + "learning_rate": 4.007462686567164e-07, + "loss": 0.9552, + "step": 47651 + }, + { + "epoch": 8.67, + "learning_rate": 4.006916636330543e-07, + "loss": 0.0934, + "step": 47652 + }, + { + "epoch": 8.67, + "learning_rate": 4.00637058609392e-07, + "loss": 0.2714, + "step": 47653 + }, + { + "epoch": 8.67, + "learning_rate": 4.005824535857299e-07, + "loss": 0.903, + "step": 47654 + }, + { + "epoch": 8.67, + "learning_rate": 4.0052784856206775e-07, + "loss": 0.0035, + "step": 47655 + }, + { + "epoch": 8.67, + "learning_rate": 4.0047324353840553e-07, + "loss": 0.4191, + "step": 47656 + }, + { + "epoch": 8.67, + "learning_rate": 4.0041863851474337e-07, + "loss": 0.0534, + "step": 47657 + }, + { + "epoch": 8.67, + "learning_rate": 4.0036403349108115e-07, + "loss": 0.9586, + "step": 47658 + }, + { + "epoch": 8.67, + "learning_rate": 4.0030942846741904e-07, + "loss": 0.2791, + "step": 47659 + }, + { + "epoch": 8.67, + "learning_rate": 4.002548234437568e-07, + "loss": 0.3911, + "step": 47660 + }, + { + "epoch": 8.67, + "learning_rate": 4.0020021842009467e-07, + "loss": 0.189, + "step": 47661 + }, + { + "epoch": 8.67, + "learning_rate": 4.001456133964325e-07, + "loss": 0.2854, + "step": 47662 + }, + { + "epoch": 8.67, + "learning_rate": 4.000910083727703e-07, + "loss": 1.1976, + "step": 47663 + }, + { + "epoch": 8.67, + "learning_rate": 4.000364033491082e-07, + "loss": 0.8051, + "step": 47664 + }, + { + "epoch": 8.67, + "learning_rate": 3.999817983254459e-07, + "loss": 1.1622, + "step": 47665 + }, + { + "epoch": 8.67, + "learning_rate": 3.999271933017838e-07, + "loss": 0.6398, + "step": 47666 + }, + { + "epoch": 8.67, + "learning_rate": 3.998725882781216e-07, + "loss": 0.2671, + "step": 47667 + }, + { + "epoch": 8.67, + "learning_rate": 3.998179832544594e-07, + "loss": 0.059, + "step": 47668 + }, + { + "epoch": 8.67, + "learning_rate": 3.9976337823079726e-07, + "loss": 0.7495, + "step": 47669 + }, + { + "epoch": 8.67, + "learning_rate": 3.9970877320713504e-07, + "loss": 0.0552, + "step": 47670 + }, + { + "epoch": 8.67, + "learning_rate": 3.9965416818347294e-07, + "loss": 0.0055, + "step": 47671 + }, + { + "epoch": 8.67, + "learning_rate": 3.995995631598107e-07, + "loss": 0.006, + "step": 47672 + }, + { + "epoch": 8.67, + "learning_rate": 3.9954495813614856e-07, + "loss": 0.0066, + "step": 47673 + }, + { + "epoch": 8.67, + "learning_rate": 3.9949035311248634e-07, + "loss": 0.3493, + "step": 47674 + }, + { + "epoch": 8.67, + "learning_rate": 3.994357480888242e-07, + "loss": 1.1752, + "step": 47675 + }, + { + "epoch": 8.67, + "learning_rate": 3.9938114306516207e-07, + "loss": 1.0967, + "step": 47676 + }, + { + "epoch": 8.67, + "learning_rate": 3.993265380414998e-07, + "loss": 1.2427, + "step": 47677 + }, + { + "epoch": 8.67, + "learning_rate": 3.992719330178377e-07, + "loss": 0.5784, + "step": 47678 + }, + { + "epoch": 8.67, + "learning_rate": 3.992173279941755e-07, + "loss": 0.2161, + "step": 47679 + }, + { + "epoch": 8.67, + "learning_rate": 3.991627229705133e-07, + "loss": 0.0046, + "step": 47680 + }, + { + "epoch": 8.67, + "learning_rate": 3.991081179468511e-07, + "loss": 0.011, + "step": 47681 + }, + { + "epoch": 8.67, + "learning_rate": 3.9905351292318894e-07, + "loss": 0.9186, + "step": 47682 + }, + { + "epoch": 8.67, + "learning_rate": 3.989989078995267e-07, + "loss": 0.1153, + "step": 47683 + }, + { + "epoch": 8.67, + "learning_rate": 3.989443028758646e-07, + "loss": 0.0455, + "step": 47684 + }, + { + "epoch": 8.67, + "learning_rate": 3.9888969785220245e-07, + "loss": 0.0086, + "step": 47685 + }, + { + "epoch": 8.67, + "learning_rate": 3.9883509282854023e-07, + "loss": 0.2321, + "step": 47686 + }, + { + "epoch": 8.67, + "learning_rate": 3.9878048780487807e-07, + "loss": 0.0021, + "step": 47687 + }, + { + "epoch": 8.67, + "learning_rate": 3.9872588278121586e-07, + "loss": 0.1014, + "step": 47688 + }, + { + "epoch": 8.67, + "learning_rate": 3.986712777575537e-07, + "loss": 1.6182, + "step": 47689 + }, + { + "epoch": 8.67, + "learning_rate": 3.986166727338915e-07, + "loss": 0.0063, + "step": 47690 + }, + { + "epoch": 8.67, + "learning_rate": 3.9856206771022937e-07, + "loss": 1.6113, + "step": 47691 + }, + { + "epoch": 8.67, + "learning_rate": 3.985074626865672e-07, + "loss": 1.131, + "step": 47692 + }, + { + "epoch": 8.67, + "learning_rate": 3.98452857662905e-07, + "loss": 0.3807, + "step": 47693 + }, + { + "epoch": 8.67, + "learning_rate": 3.9839825263924283e-07, + "loss": 0.0537, + "step": 47694 + }, + { + "epoch": 8.67, + "learning_rate": 3.983436476155806e-07, + "loss": 0.4748, + "step": 47695 + }, + { + "epoch": 8.67, + "learning_rate": 3.982890425919185e-07, + "loss": 0.0639, + "step": 47696 + }, + { + "epoch": 8.67, + "learning_rate": 3.9823443756825623e-07, + "loss": 0.1415, + "step": 47697 + }, + { + "epoch": 8.67, + "learning_rate": 3.981798325445941e-07, + "loss": 0.1708, + "step": 47698 + }, + { + "epoch": 8.67, + "learning_rate": 3.9812522752093196e-07, + "loss": 1.1418, + "step": 47699 + }, + { + "epoch": 8.67, + "learning_rate": 3.9807062249726975e-07, + "loss": 0.76, + "step": 47700 + }, + { + "epoch": 8.67, + "learning_rate": 3.980160174736076e-07, + "loss": 0.1963, + "step": 47701 + }, + { + "epoch": 8.67, + "learning_rate": 3.9796141244994537e-07, + "loss": 1.0164, + "step": 47702 + }, + { + "epoch": 8.67, + "learning_rate": 3.9790680742628326e-07, + "loss": 0.0122, + "step": 47703 + }, + { + "epoch": 8.68, + "learning_rate": 3.9785220240262104e-07, + "loss": 0.0176, + "step": 47704 + }, + { + "epoch": 8.68, + "learning_rate": 3.977975973789589e-07, + "loss": 1.1142, + "step": 47705 + }, + { + "epoch": 8.68, + "learning_rate": 3.977429923552967e-07, + "loss": 0.3056, + "step": 47706 + }, + { + "epoch": 8.68, + "learning_rate": 3.976883873316345e-07, + "loss": 0.3765, + "step": 47707 + }, + { + "epoch": 8.68, + "learning_rate": 3.976337823079724e-07, + "loss": 0.7895, + "step": 47708 + }, + { + "epoch": 8.68, + "learning_rate": 3.975791772843101e-07, + "loss": 0.3389, + "step": 47709 + }, + { + "epoch": 8.68, + "learning_rate": 3.97524572260648e-07, + "loss": 0.1994, + "step": 47710 + }, + { + "epoch": 8.68, + "learning_rate": 3.974699672369858e-07, + "loss": 0.0074, + "step": 47711 + }, + { + "epoch": 8.68, + "learning_rate": 3.9741536221332364e-07, + "loss": 0.0856, + "step": 47712 + }, + { + "epoch": 8.68, + "learning_rate": 3.973607571896615e-07, + "loss": 1.369, + "step": 47713 + }, + { + "epoch": 8.68, + "learning_rate": 3.9730615216599926e-07, + "loss": 0.0063, + "step": 47714 + }, + { + "epoch": 8.68, + "learning_rate": 3.9725154714233715e-07, + "loss": 0.2882, + "step": 47715 + }, + { + "epoch": 8.68, + "learning_rate": 3.9719694211867494e-07, + "loss": 0.7447, + "step": 47716 + }, + { + "epoch": 8.68, + "learning_rate": 3.9714233709501277e-07, + "loss": 0.1361, + "step": 47717 + }, + { + "epoch": 8.68, + "learning_rate": 3.9708773207135056e-07, + "loss": 0.6892, + "step": 47718 + }, + { + "epoch": 8.68, + "learning_rate": 3.970331270476884e-07, + "loss": 0.0127, + "step": 47719 + }, + { + "epoch": 8.68, + "learning_rate": 3.969785220240262e-07, + "loss": 0.1679, + "step": 47720 + }, + { + "epoch": 8.68, + "learning_rate": 3.96923917000364e-07, + "loss": 0.2937, + "step": 47721 + }, + { + "epoch": 8.68, + "learning_rate": 3.968693119767019e-07, + "loss": 0.0431, + "step": 47722 + }, + { + "epoch": 8.68, + "learning_rate": 3.968147069530397e-07, + "loss": 0.0048, + "step": 47723 + }, + { + "epoch": 8.68, + "learning_rate": 3.9676010192937753e-07, + "loss": 0.1464, + "step": 47724 + }, + { + "epoch": 8.68, + "learning_rate": 3.967054969057153e-07, + "loss": 0.3464, + "step": 47725 + }, + { + "epoch": 8.68, + "learning_rate": 3.9665089188205315e-07, + "loss": 0.1514, + "step": 47726 + }, + { + "epoch": 8.68, + "learning_rate": 3.9659628685839094e-07, + "loss": 0.7236, + "step": 47727 + }, + { + "epoch": 8.68, + "learning_rate": 3.9654168183472883e-07, + "loss": 0.4507, + "step": 47728 + }, + { + "epoch": 8.68, + "learning_rate": 3.9648707681106666e-07, + "loss": 0.1192, + "step": 47729 + }, + { + "epoch": 8.68, + "learning_rate": 3.9643247178740445e-07, + "loss": 0.2318, + "step": 47730 + }, + { + "epoch": 8.68, + "learning_rate": 3.963778667637423e-07, + "loss": 0.0153, + "step": 47731 + }, + { + "epoch": 8.68, + "learning_rate": 3.9632326174008007e-07, + "loss": 0.0175, + "step": 47732 + }, + { + "epoch": 8.68, + "learning_rate": 3.962686567164179e-07, + "loss": 1.1479, + "step": 47733 + }, + { + "epoch": 8.68, + "learning_rate": 3.962140516927557e-07, + "loss": 0.4753, + "step": 47734 + }, + { + "epoch": 8.68, + "learning_rate": 3.961594466690936e-07, + "loss": 0.2241, + "step": 47735 + }, + { + "epoch": 8.68, + "learning_rate": 3.961048416454314e-07, + "loss": 0.3929, + "step": 47736 + }, + { + "epoch": 8.68, + "learning_rate": 3.960502366217692e-07, + "loss": 1.5578, + "step": 47737 + }, + { + "epoch": 8.68, + "learning_rate": 3.9599563159810704e-07, + "loss": 1.2576, + "step": 47738 + }, + { + "epoch": 8.68, + "learning_rate": 3.9594102657444483e-07, + "loss": 0.0129, + "step": 47739 + }, + { + "epoch": 8.68, + "learning_rate": 3.958864215507827e-07, + "loss": 0.2876, + "step": 47740 + }, + { + "epoch": 8.68, + "learning_rate": 3.9583181652712045e-07, + "loss": 0.4582, + "step": 47741 + }, + { + "epoch": 8.68, + "learning_rate": 3.9577721150345834e-07, + "loss": 0.2549, + "step": 47742 + }, + { + "epoch": 8.68, + "learning_rate": 3.957226064797962e-07, + "loss": 1.24, + "step": 47743 + }, + { + "epoch": 8.68, + "learning_rate": 3.9566800145613396e-07, + "loss": 0.2463, + "step": 47744 + }, + { + "epoch": 8.68, + "learning_rate": 3.956133964324718e-07, + "loss": 1.0888, + "step": 47745 + }, + { + "epoch": 8.68, + "learning_rate": 3.955587914088096e-07, + "loss": 0.0428, + "step": 47746 + }, + { + "epoch": 8.68, + "learning_rate": 3.955041863851475e-07, + "loss": 0.9747, + "step": 47747 + }, + { + "epoch": 8.68, + "learning_rate": 3.9544958136148526e-07, + "loss": 0.0552, + "step": 47748 + }, + { + "epoch": 8.68, + "learning_rate": 3.953949763378231e-07, + "loss": 1.6308, + "step": 47749 + }, + { + "epoch": 8.68, + "learning_rate": 3.9534037131416093e-07, + "loss": 2.4283, + "step": 47750 + }, + { + "epoch": 8.68, + "learning_rate": 3.952857662904987e-07, + "loss": 0.0595, + "step": 47751 + }, + { + "epoch": 8.68, + "learning_rate": 3.952311612668366e-07, + "loss": 3.1374, + "step": 47752 + }, + { + "epoch": 8.68, + "learning_rate": 3.9517655624317434e-07, + "loss": 0.6426, + "step": 47753 + }, + { + "epoch": 8.68, + "learning_rate": 3.9512195121951223e-07, + "loss": 0.0126, + "step": 47754 + }, + { + "epoch": 8.68, + "learning_rate": 3.9506734619585e-07, + "loss": 0.9056, + "step": 47755 + }, + { + "epoch": 8.68, + "learning_rate": 3.9501274117218785e-07, + "loss": 0.0359, + "step": 47756 + }, + { + "epoch": 8.68, + "learning_rate": 3.9495813614852564e-07, + "loss": 1.4727, + "step": 47757 + }, + { + "epoch": 8.68, + "learning_rate": 3.949035311248635e-07, + "loss": 2.802, + "step": 47758 + }, + { + "epoch": 8.69, + "learning_rate": 3.9484892610120137e-07, + "loss": 1.5636, + "step": 47759 + }, + { + "epoch": 8.69, + "learning_rate": 3.9479432107753915e-07, + "loss": 0.0456, + "step": 47760 + }, + { + "epoch": 8.69, + "learning_rate": 3.94739716053877e-07, + "loss": 0.199, + "step": 47761 + }, + { + "epoch": 8.69, + "learning_rate": 3.9468511103021477e-07, + "loss": 0.0032, + "step": 47762 + }, + { + "epoch": 8.69, + "learning_rate": 3.946305060065526e-07, + "loss": 0.0107, + "step": 47763 + }, + { + "epoch": 8.69, + "learning_rate": 3.945759009828904e-07, + "loss": 0.0125, + "step": 47764 + }, + { + "epoch": 8.69, + "learning_rate": 3.9452129595922823e-07, + "loss": 0.1181, + "step": 47765 + }, + { + "epoch": 8.69, + "learning_rate": 3.944666909355661e-07, + "loss": 0.0592, + "step": 47766 + }, + { + "epoch": 8.69, + "learning_rate": 3.944120859119039e-07, + "loss": 0.0138, + "step": 47767 + }, + { + "epoch": 8.69, + "learning_rate": 3.9435748088824174e-07, + "loss": 0.0126, + "step": 47768 + }, + { + "epoch": 8.69, + "learning_rate": 3.9430287586457953e-07, + "loss": 0.7332, + "step": 47769 + }, + { + "epoch": 8.69, + "learning_rate": 3.9424827084091737e-07, + "loss": 0.0098, + "step": 47770 + }, + { + "epoch": 8.69, + "learning_rate": 3.9419366581725515e-07, + "loss": 0.4897, + "step": 47771 + }, + { + "epoch": 8.69, + "learning_rate": 3.9413906079359304e-07, + "loss": 0.3652, + "step": 47772 + }, + { + "epoch": 8.69, + "learning_rate": 3.940844557699309e-07, + "loss": 1.4675, + "step": 47773 + }, + { + "epoch": 8.69, + "learning_rate": 3.9402985074626866e-07, + "loss": 0.1507, + "step": 47774 + }, + { + "epoch": 8.69, + "learning_rate": 3.939752457226065e-07, + "loss": 3.7199, + "step": 47775 + }, + { + "epoch": 8.69, + "learning_rate": 3.939206406989443e-07, + "loss": 0.5857, + "step": 47776 + }, + { + "epoch": 8.69, + "learning_rate": 3.938660356752821e-07, + "loss": 0.0156, + "step": 47777 + }, + { + "epoch": 8.69, + "learning_rate": 3.938114306516199e-07, + "loss": 0.5632, + "step": 47778 + }, + { + "epoch": 8.69, + "learning_rate": 3.937568256279578e-07, + "loss": 0.0091, + "step": 47779 + }, + { + "epoch": 8.69, + "learning_rate": 3.9370222060429564e-07, + "loss": 0.3595, + "step": 47780 + }, + { + "epoch": 8.69, + "learning_rate": 3.936476155806334e-07, + "loss": 5.6478, + "step": 47781 + }, + { + "epoch": 8.69, + "learning_rate": 3.9359301055697126e-07, + "loss": 0.0499, + "step": 47782 + }, + { + "epoch": 8.69, + "learning_rate": 3.9353840553330904e-07, + "loss": 2.0909, + "step": 47783 + }, + { + "epoch": 8.69, + "learning_rate": 3.9348380050964693e-07, + "loss": 0.1482, + "step": 47784 + }, + { + "epoch": 8.69, + "learning_rate": 3.9342919548598467e-07, + "loss": 0.014, + "step": 47785 + }, + { + "epoch": 8.69, + "learning_rate": 3.9337459046232256e-07, + "loss": 1.2961, + "step": 47786 + }, + { + "epoch": 8.69, + "learning_rate": 3.933199854386604e-07, + "loss": 0.0079, + "step": 47787 + }, + { + "epoch": 8.69, + "learning_rate": 3.932653804149982e-07, + "loss": 0.1452, + "step": 47788 + }, + { + "epoch": 8.69, + "learning_rate": 3.93210775391336e-07, + "loss": 0.2042, + "step": 47789 + }, + { + "epoch": 8.69, + "learning_rate": 3.931561703676738e-07, + "loss": 0.1331, + "step": 47790 + }, + { + "epoch": 8.69, + "learning_rate": 3.931015653440117e-07, + "loss": 0.1082, + "step": 47791 + }, + { + "epoch": 8.69, + "learning_rate": 3.930469603203495e-07, + "loss": 0.1631, + "step": 47792 + }, + { + "epoch": 8.69, + "learning_rate": 3.929923552966873e-07, + "loss": 0.0757, + "step": 47793 + }, + { + "epoch": 8.69, + "learning_rate": 3.929377502730251e-07, + "loss": 1.4382, + "step": 47794 + }, + { + "epoch": 8.69, + "learning_rate": 3.9288314524936293e-07, + "loss": 0.0079, + "step": 47795 + }, + { + "epoch": 8.69, + "learning_rate": 3.928285402257008e-07, + "loss": 0.0406, + "step": 47796 + }, + { + "epoch": 8.69, + "learning_rate": 3.9277393520203856e-07, + "loss": 0.1622, + "step": 47797 + }, + { + "epoch": 8.69, + "learning_rate": 3.9271933017837645e-07, + "loss": 0.0113, + "step": 47798 + }, + { + "epoch": 8.69, + "learning_rate": 3.9266472515471423e-07, + "loss": 0.0046, + "step": 47799 + }, + { + "epoch": 8.69, + "learning_rate": 3.9261012013105207e-07, + "loss": 0.0664, + "step": 47800 + }, + { + "epoch": 8.69, + "learning_rate": 3.9255551510738985e-07, + "loss": 0.0204, + "step": 47801 + }, + { + "epoch": 8.69, + "learning_rate": 3.925009100837277e-07, + "loss": 2.4509, + "step": 47802 + }, + { + "epoch": 8.69, + "learning_rate": 3.924463050600656e-07, + "loss": 0.1632, + "step": 47803 + }, + { + "epoch": 8.69, + "learning_rate": 3.9239170003640337e-07, + "loss": 0.0046, + "step": 47804 + }, + { + "epoch": 8.69, + "learning_rate": 3.923370950127412e-07, + "loss": 0.0397, + "step": 47805 + }, + { + "epoch": 8.69, + "learning_rate": 3.92282489989079e-07, + "loss": 2.216, + "step": 47806 + }, + { + "epoch": 8.69, + "learning_rate": 3.922278849654168e-07, + "loss": 1.526, + "step": 47807 + }, + { + "epoch": 8.69, + "learning_rate": 3.921732799417546e-07, + "loss": 1.0448, + "step": 47808 + }, + { + "epoch": 8.69, + "learning_rate": 3.9211867491809245e-07, + "loss": 0.7026, + "step": 47809 + }, + { + "epoch": 8.69, + "learning_rate": 3.9206406989443034e-07, + "loss": 1.2463, + "step": 47810 + }, + { + "epoch": 8.69, + "learning_rate": 3.920094648707681e-07, + "loss": 1.2713, + "step": 47811 + }, + { + "epoch": 8.69, + "learning_rate": 3.9195485984710596e-07, + "loss": 0.657, + "step": 47812 + }, + { + "epoch": 8.69, + "learning_rate": 3.9190025482344374e-07, + "loss": 0.0062, + "step": 47813 + }, + { + "epoch": 8.7, + "learning_rate": 3.918456497997816e-07, + "loss": 0.622, + "step": 47814 + }, + { + "epoch": 8.7, + "learning_rate": 3.9179104477611937e-07, + "loss": 2.4245, + "step": 47815 + }, + { + "epoch": 8.7, + "learning_rate": 3.9173643975245726e-07, + "loss": 0.9049, + "step": 47816 + }, + { + "epoch": 8.7, + "learning_rate": 3.916818347287951e-07, + "loss": 1.1041, + "step": 47817 + }, + { + "epoch": 8.7, + "learning_rate": 3.916272297051329e-07, + "loss": 0.0118, + "step": 47818 + }, + { + "epoch": 8.7, + "learning_rate": 3.915726246814707e-07, + "loss": 0.0136, + "step": 47819 + }, + { + "epoch": 8.7, + "learning_rate": 3.915180196578085e-07, + "loss": 0.0701, + "step": 47820 + }, + { + "epoch": 8.7, + "learning_rate": 3.914634146341464e-07, + "loss": 0.0648, + "step": 47821 + }, + { + "epoch": 8.7, + "learning_rate": 3.914088096104841e-07, + "loss": 0.0622, + "step": 47822 + }, + { + "epoch": 8.7, + "learning_rate": 3.91354204586822e-07, + "loss": 4.2114, + "step": 47823 + }, + { + "epoch": 8.7, + "learning_rate": 3.9129959956315985e-07, + "loss": 0.7034, + "step": 47824 + }, + { + "epoch": 8.7, + "learning_rate": 3.9124499453949764e-07, + "loss": 1.3346, + "step": 47825 + }, + { + "epoch": 8.7, + "learning_rate": 3.911903895158355e-07, + "loss": 0.1047, + "step": 47826 + }, + { + "epoch": 8.7, + "learning_rate": 3.9113578449217326e-07, + "loss": 0.5802, + "step": 47827 + }, + { + "epoch": 8.7, + "learning_rate": 3.9108117946851115e-07, + "loss": 1.429, + "step": 47828 + }, + { + "epoch": 8.7, + "learning_rate": 3.910265744448489e-07, + "loss": 0.6792, + "step": 47829 + }, + { + "epoch": 8.7, + "learning_rate": 3.9097196942118677e-07, + "loss": 0.0072, + "step": 47830 + }, + { + "epoch": 8.7, + "learning_rate": 3.909173643975246e-07, + "loss": 0.0295, + "step": 47831 + }, + { + "epoch": 8.7, + "learning_rate": 3.908627593738624e-07, + "loss": 0.2529, + "step": 47832 + }, + { + "epoch": 8.7, + "learning_rate": 3.908081543502003e-07, + "loss": 0.0312, + "step": 47833 + }, + { + "epoch": 8.7, + "learning_rate": 3.90753549326538e-07, + "loss": 0.4188, + "step": 47834 + }, + { + "epoch": 8.7, + "learning_rate": 3.906989443028759e-07, + "loss": 0.2844, + "step": 47835 + }, + { + "epoch": 8.7, + "learning_rate": 3.906443392792137e-07, + "loss": 0.709, + "step": 47836 + }, + { + "epoch": 8.7, + "learning_rate": 3.9058973425555153e-07, + "loss": 0.0557, + "step": 47837 + }, + { + "epoch": 8.7, + "learning_rate": 3.905351292318893e-07, + "loss": 0.4149, + "step": 47838 + }, + { + "epoch": 8.7, + "learning_rate": 3.9048052420822715e-07, + "loss": 0.7533, + "step": 47839 + }, + { + "epoch": 8.7, + "learning_rate": 3.9042591918456504e-07, + "loss": 1.0, + "step": 47840 + }, + { + "epoch": 8.7, + "learning_rate": 3.9037131416090277e-07, + "loss": 0.0144, + "step": 47841 + }, + { + "epoch": 8.7, + "learning_rate": 3.9031670913724066e-07, + "loss": 0.0259, + "step": 47842 + }, + { + "epoch": 8.7, + "learning_rate": 3.9026210411357845e-07, + "loss": 0.0299, + "step": 47843 + }, + { + "epoch": 8.7, + "learning_rate": 3.902074990899163e-07, + "loss": 0.8753, + "step": 47844 + }, + { + "epoch": 8.7, + "learning_rate": 3.9015289406625407e-07, + "loss": 0.0398, + "step": 47845 + }, + { + "epoch": 8.7, + "learning_rate": 3.900982890425919e-07, + "loss": 0.0365, + "step": 47846 + }, + { + "epoch": 8.7, + "learning_rate": 3.900436840189298e-07, + "loss": 0.2774, + "step": 47847 + }, + { + "epoch": 8.7, + "learning_rate": 3.899890789952676e-07, + "loss": 0.1525, + "step": 47848 + }, + { + "epoch": 8.7, + "learning_rate": 3.899344739716054e-07, + "loss": 0.0597, + "step": 47849 + }, + { + "epoch": 8.7, + "learning_rate": 3.898798689479432e-07, + "loss": 0.6796, + "step": 47850 + }, + { + "epoch": 8.7, + "learning_rate": 3.8982526392428104e-07, + "loss": 0.0088, + "step": 47851 + }, + { + "epoch": 8.7, + "learning_rate": 3.897706589006188e-07, + "loss": 0.1318, + "step": 47852 + }, + { + "epoch": 8.7, + "learning_rate": 3.897160538769567e-07, + "loss": 0.0079, + "step": 47853 + }, + { + "epoch": 8.7, + "learning_rate": 3.8966144885329455e-07, + "loss": 0.1043, + "step": 47854 + }, + { + "epoch": 8.7, + "learning_rate": 3.8960684382963234e-07, + "loss": 0.0316, + "step": 47855 + }, + { + "epoch": 8.7, + "learning_rate": 3.895522388059702e-07, + "loss": 0.5463, + "step": 47856 + }, + { + "epoch": 8.7, + "learning_rate": 3.8949763378230796e-07, + "loss": 0.2792, + "step": 47857 + }, + { + "epoch": 8.7, + "learning_rate": 3.894430287586458e-07, + "loss": 0.0644, + "step": 47858 + }, + { + "epoch": 8.7, + "learning_rate": 3.893884237349836e-07, + "loss": 1.5681, + "step": 47859 + }, + { + "epoch": 8.7, + "learning_rate": 3.8933381871132147e-07, + "loss": 0.1613, + "step": 47860 + }, + { + "epoch": 8.7, + "learning_rate": 3.892792136876593e-07, + "loss": 1.3506, + "step": 47861 + }, + { + "epoch": 8.7, + "learning_rate": 3.892246086639971e-07, + "loss": 0.1827, + "step": 47862 + }, + { + "epoch": 8.7, + "learning_rate": 3.8917000364033493e-07, + "loss": 0.0031, + "step": 47863 + }, + { + "epoch": 8.7, + "learning_rate": 3.891153986166727e-07, + "loss": 0.2183, + "step": 47864 + }, + { + "epoch": 8.7, + "learning_rate": 3.890607935930106e-07, + "loss": 0.4463, + "step": 47865 + }, + { + "epoch": 8.7, + "learning_rate": 3.8900618856934834e-07, + "loss": 0.1785, + "step": 47866 + }, + { + "epoch": 8.7, + "learning_rate": 3.8895158354568623e-07, + "loss": 0.0612, + "step": 47867 + }, + { + "epoch": 8.7, + "learning_rate": 3.8889697852202407e-07, + "loss": 1.313, + "step": 47868 + }, + { + "epoch": 8.71, + "learning_rate": 3.8884237349836185e-07, + "loss": 1.4509, + "step": 47869 + }, + { + "epoch": 8.71, + "learning_rate": 3.887877684746997e-07, + "loss": 0.1604, + "step": 47870 + }, + { + "epoch": 8.71, + "learning_rate": 3.887331634510375e-07, + "loss": 0.6139, + "step": 47871 + }, + { + "epoch": 8.71, + "learning_rate": 3.8867855842737536e-07, + "loss": 0.0823, + "step": 47872 + }, + { + "epoch": 8.71, + "learning_rate": 3.8862395340371315e-07, + "loss": 0.2221, + "step": 47873 + }, + { + "epoch": 8.71, + "learning_rate": 3.88569348380051e-07, + "loss": 0.987, + "step": 47874 + }, + { + "epoch": 8.71, + "learning_rate": 3.8851474335638877e-07, + "loss": 0.3595, + "step": 47875 + }, + { + "epoch": 8.71, + "learning_rate": 3.884601383327266e-07, + "loss": 1.5114, + "step": 47876 + }, + { + "epoch": 8.71, + "learning_rate": 3.884055333090645e-07, + "loss": 1.2001, + "step": 47877 + }, + { + "epoch": 8.71, + "learning_rate": 3.8835092828540223e-07, + "loss": 0.9922, + "step": 47878 + }, + { + "epoch": 8.71, + "learning_rate": 3.882963232617401e-07, + "loss": 0.0066, + "step": 47879 + }, + { + "epoch": 8.71, + "learning_rate": 3.882417182380779e-07, + "loss": 0.0391, + "step": 47880 + }, + { + "epoch": 8.71, + "learning_rate": 3.8818711321441574e-07, + "loss": 0.8154, + "step": 47881 + }, + { + "epoch": 8.71, + "learning_rate": 3.8813250819075353e-07, + "loss": 0.0465, + "step": 47882 + }, + { + "epoch": 8.71, + "learning_rate": 3.8807790316709137e-07, + "loss": 0.1984, + "step": 47883 + }, + { + "epoch": 8.71, + "learning_rate": 3.8802329814342926e-07, + "loss": 0.6266, + "step": 47884 + }, + { + "epoch": 8.71, + "learning_rate": 3.8796869311976704e-07, + "loss": 0.0242, + "step": 47885 + }, + { + "epoch": 8.71, + "learning_rate": 3.879140880961049e-07, + "loss": 1.0271, + "step": 47886 + }, + { + "epoch": 8.71, + "learning_rate": 3.8785948307244266e-07, + "loss": 0.4306, + "step": 47887 + }, + { + "epoch": 8.71, + "learning_rate": 3.878048780487805e-07, + "loss": 0.5503, + "step": 47888 + }, + { + "epoch": 8.71, + "learning_rate": 3.877502730251183e-07, + "loss": 1.3393, + "step": 47889 + }, + { + "epoch": 8.71, + "learning_rate": 3.876956680014561e-07, + "loss": 0.0852, + "step": 47890 + }, + { + "epoch": 8.71, + "learning_rate": 3.87641062977794e-07, + "loss": 0.0074, + "step": 47891 + }, + { + "epoch": 8.71, + "learning_rate": 3.875864579541318e-07, + "loss": 0.1118, + "step": 47892 + }, + { + "epoch": 8.71, + "learning_rate": 3.8753185293046963e-07, + "loss": 0.798, + "step": 47893 + }, + { + "epoch": 8.71, + "learning_rate": 3.874772479068074e-07, + "loss": 0.8818, + "step": 47894 + }, + { + "epoch": 8.71, + "learning_rate": 3.8742264288314526e-07, + "loss": 1.2784, + "step": 47895 + }, + { + "epoch": 8.71, + "learning_rate": 3.8736803785948304e-07, + "loss": 0.145, + "step": 47896 + }, + { + "epoch": 8.71, + "learning_rate": 3.8731343283582093e-07, + "loss": 2.6294, + "step": 47897 + }, + { + "epoch": 8.71, + "learning_rate": 3.8725882781215877e-07, + "loss": 0.1405, + "step": 47898 + }, + { + "epoch": 8.71, + "learning_rate": 3.8720422278849655e-07, + "loss": 1.3215, + "step": 47899 + }, + { + "epoch": 8.71, + "learning_rate": 3.871496177648344e-07, + "loss": 0.7693, + "step": 47900 + }, + { + "epoch": 8.71, + "learning_rate": 3.870950127411722e-07, + "loss": 0.0197, + "step": 47901 + }, + { + "epoch": 8.71, + "learning_rate": 3.8704040771751e-07, + "loss": 3.0832, + "step": 47902 + }, + { + "epoch": 8.71, + "learning_rate": 3.869858026938478e-07, + "loss": 0.0089, + "step": 47903 + }, + { + "epoch": 8.71, + "learning_rate": 3.869311976701857e-07, + "loss": 1.4574, + "step": 47904 + }, + { + "epoch": 8.71, + "learning_rate": 3.868765926465235e-07, + "loss": 0.1016, + "step": 47905 + }, + { + "epoch": 8.71, + "learning_rate": 3.868219876228613e-07, + "loss": 0.0212, + "step": 47906 + }, + { + "epoch": 8.71, + "learning_rate": 3.8676738259919915e-07, + "loss": 0.4372, + "step": 47907 + }, + { + "epoch": 8.71, + "learning_rate": 3.8671277757553693e-07, + "loss": 0.6656, + "step": 47908 + }, + { + "epoch": 8.71, + "learning_rate": 3.866581725518748e-07, + "loss": 0.5385, + "step": 47909 + }, + { + "epoch": 8.71, + "learning_rate": 3.8660356752821255e-07, + "loss": 0.428, + "step": 47910 + }, + { + "epoch": 8.71, + "learning_rate": 3.8654896250455045e-07, + "loss": 1.3567, + "step": 47911 + }, + { + "epoch": 8.71, + "learning_rate": 3.8649435748088823e-07, + "loss": 0.272, + "step": 47912 + }, + { + "epoch": 8.71, + "learning_rate": 3.8643975245722607e-07, + "loss": 0.0096, + "step": 47913 + }, + { + "epoch": 8.71, + "learning_rate": 3.863851474335639e-07, + "loss": 1.0101, + "step": 47914 + }, + { + "epoch": 8.71, + "learning_rate": 3.863305424099017e-07, + "loss": 0.0213, + "step": 47915 + }, + { + "epoch": 8.71, + "learning_rate": 3.862759373862396e-07, + "loss": 0.6534, + "step": 47916 + }, + { + "epoch": 8.71, + "learning_rate": 3.8622133236257736e-07, + "loss": 1.8962, + "step": 47917 + }, + { + "epoch": 8.71, + "learning_rate": 3.861667273389152e-07, + "loss": 0.1043, + "step": 47918 + }, + { + "epoch": 8.71, + "learning_rate": 3.86112122315253e-07, + "loss": 1.4133, + "step": 47919 + }, + { + "epoch": 8.71, + "learning_rate": 3.860575172915908e-07, + "loss": 0.3752, + "step": 47920 + }, + { + "epoch": 8.71, + "learning_rate": 3.860029122679287e-07, + "loss": 0.2558, + "step": 47921 + }, + { + "epoch": 8.71, + "learning_rate": 3.8594830724426645e-07, + "loss": 2.8966, + "step": 47922 + }, + { + "epoch": 8.71, + "learning_rate": 3.8589370222060434e-07, + "loss": 0.1267, + "step": 47923 + }, + { + "epoch": 8.72, + "learning_rate": 3.858390971969421e-07, + "loss": 0.5133, + "step": 47924 + }, + { + "epoch": 8.72, + "learning_rate": 3.8578449217327996e-07, + "loss": 0.0113, + "step": 47925 + }, + { + "epoch": 8.72, + "learning_rate": 3.8572988714961774e-07, + "loss": 0.1126, + "step": 47926 + }, + { + "epoch": 8.72, + "learning_rate": 3.856752821259556e-07, + "loss": 0.4189, + "step": 47927 + }, + { + "epoch": 8.72, + "learning_rate": 3.8562067710229347e-07, + "loss": 2.1575, + "step": 47928 + }, + { + "epoch": 8.72, + "learning_rate": 3.8556607207863126e-07, + "loss": 0.0101, + "step": 47929 + }, + { + "epoch": 8.72, + "learning_rate": 3.855114670549691e-07, + "loss": 1.4311, + "step": 47930 + }, + { + "epoch": 8.72, + "learning_rate": 3.854568620313069e-07, + "loss": 0.3547, + "step": 47931 + }, + { + "epoch": 8.72, + "learning_rate": 3.854022570076447e-07, + "loss": 0.0226, + "step": 47932 + }, + { + "epoch": 8.72, + "learning_rate": 3.853476519839825e-07, + "loss": 0.2238, + "step": 47933 + }, + { + "epoch": 8.72, + "learning_rate": 3.8529304696032034e-07, + "loss": 0.5031, + "step": 47934 + }, + { + "epoch": 8.72, + "learning_rate": 3.8523844193665823e-07, + "loss": 0.1134, + "step": 47935 + }, + { + "epoch": 8.72, + "learning_rate": 3.85183836912996e-07, + "loss": 0.2963, + "step": 47936 + }, + { + "epoch": 8.72, + "learning_rate": 3.8512923188933385e-07, + "loss": 2.0868, + "step": 47937 + }, + { + "epoch": 8.72, + "learning_rate": 3.8507462686567163e-07, + "loss": 0.0158, + "step": 47938 + }, + { + "epoch": 8.72, + "learning_rate": 3.8502002184200947e-07, + "loss": 1.4022, + "step": 47939 + }, + { + "epoch": 8.72, + "learning_rate": 3.8496541681834726e-07, + "loss": 0.8371, + "step": 47940 + }, + { + "epoch": 8.72, + "learning_rate": 3.8491081179468515e-07, + "loss": 0.7009, + "step": 47941 + }, + { + "epoch": 8.72, + "learning_rate": 3.84856206771023e-07, + "loss": 2.5949, + "step": 47942 + }, + { + "epoch": 8.72, + "learning_rate": 3.8480160174736077e-07, + "loss": 0.188, + "step": 47943 + }, + { + "epoch": 8.72, + "learning_rate": 3.847469967236986e-07, + "loss": 0.0749, + "step": 47944 + }, + { + "epoch": 8.72, + "learning_rate": 3.846923917000364e-07, + "loss": 0.4601, + "step": 47945 + }, + { + "epoch": 8.72, + "learning_rate": 3.8463778667637423e-07, + "loss": 0.3194, + "step": 47946 + }, + { + "epoch": 8.72, + "learning_rate": 3.84583181652712e-07, + "loss": 0.0576, + "step": 47947 + }, + { + "epoch": 8.72, + "learning_rate": 3.845285766290499e-07, + "loss": 0.2085, + "step": 47948 + }, + { + "epoch": 8.72, + "learning_rate": 3.844739716053877e-07, + "loss": 1.9681, + "step": 47949 + }, + { + "epoch": 8.72, + "learning_rate": 3.844193665817255e-07, + "loss": 0.221, + "step": 47950 + }, + { + "epoch": 8.72, + "learning_rate": 3.8436476155806336e-07, + "loss": 0.0706, + "step": 47951 + }, + { + "epoch": 8.72, + "learning_rate": 3.8431015653440115e-07, + "loss": 0.0077, + "step": 47952 + }, + { + "epoch": 8.72, + "learning_rate": 3.8425555151073904e-07, + "loss": 0.3624, + "step": 47953 + }, + { + "epoch": 8.72, + "learning_rate": 3.8420094648707677e-07, + "loss": 0.5584, + "step": 47954 + }, + { + "epoch": 8.72, + "learning_rate": 3.8414634146341466e-07, + "loss": 0.0112, + "step": 47955 + }, + { + "epoch": 8.72, + "learning_rate": 3.8409173643975245e-07, + "loss": 2.7912, + "step": 47956 + }, + { + "epoch": 8.72, + "learning_rate": 3.840371314160903e-07, + "loss": 0.1406, + "step": 47957 + }, + { + "epoch": 8.72, + "learning_rate": 3.839825263924281e-07, + "loss": 0.7276, + "step": 47958 + }, + { + "epoch": 8.72, + "learning_rate": 3.839279213687659e-07, + "loss": 0.4278, + "step": 47959 + }, + { + "epoch": 8.72, + "learning_rate": 3.838733163451038e-07, + "loss": 1.0357, + "step": 47960 + }, + { + "epoch": 8.72, + "learning_rate": 3.838187113214416e-07, + "loss": 0.3762, + "step": 47961 + }, + { + "epoch": 8.72, + "learning_rate": 3.837641062977794e-07, + "loss": 0.0633, + "step": 47962 + }, + { + "epoch": 8.72, + "learning_rate": 3.837095012741172e-07, + "loss": 0.1786, + "step": 47963 + }, + { + "epoch": 8.72, + "learning_rate": 3.8365489625045504e-07, + "loss": 0.8318, + "step": 47964 + }, + { + "epoch": 8.72, + "learning_rate": 3.8360029122679293e-07, + "loss": 0.004, + "step": 47965 + }, + { + "epoch": 8.72, + "learning_rate": 3.8354568620313066e-07, + "loss": 4.5647, + "step": 47966 + }, + { + "epoch": 8.72, + "learning_rate": 3.8349108117946855e-07, + "loss": 0.3992, + "step": 47967 + }, + { + "epoch": 8.72, + "learning_rate": 3.8343647615580634e-07, + "loss": 0.5376, + "step": 47968 + }, + { + "epoch": 8.72, + "learning_rate": 3.833818711321442e-07, + "loss": 0.5026, + "step": 47969 + }, + { + "epoch": 8.72, + "learning_rate": 3.8332726610848196e-07, + "loss": 0.0626, + "step": 47970 + }, + { + "epoch": 8.72, + "learning_rate": 3.832726610848198e-07, + "loss": 0.0423, + "step": 47971 + }, + { + "epoch": 8.72, + "learning_rate": 3.832180560611577e-07, + "loss": 0.173, + "step": 47972 + }, + { + "epoch": 8.72, + "learning_rate": 3.8316345103749547e-07, + "loss": 0.3133, + "step": 47973 + }, + { + "epoch": 8.72, + "learning_rate": 3.831088460138333e-07, + "loss": 0.0255, + "step": 47974 + }, + { + "epoch": 8.72, + "learning_rate": 3.830542409901711e-07, + "loss": 0.1464, + "step": 47975 + }, + { + "epoch": 8.72, + "learning_rate": 3.8299963596650893e-07, + "loss": 0.6241, + "step": 47976 + }, + { + "epoch": 8.72, + "learning_rate": 3.829450309428467e-07, + "loss": 1.0339, + "step": 47977 + }, + { + "epoch": 8.72, + "learning_rate": 3.8289042591918455e-07, + "loss": 0.4653, + "step": 47978 + }, + { + "epoch": 8.73, + "learning_rate": 3.8283582089552244e-07, + "loss": 0.7811, + "step": 47979 + }, + { + "epoch": 8.73, + "learning_rate": 3.8278121587186023e-07, + "loss": 0.0673, + "step": 47980 + }, + { + "epoch": 8.73, + "learning_rate": 3.8272661084819807e-07, + "loss": 0.0514, + "step": 47981 + }, + { + "epoch": 8.73, + "learning_rate": 3.8267200582453585e-07, + "loss": 2.3393, + "step": 47982 + }, + { + "epoch": 8.73, + "learning_rate": 3.826174008008737e-07, + "loss": 0.0056, + "step": 47983 + }, + { + "epoch": 8.73, + "learning_rate": 3.8256279577721147e-07, + "loss": 0.5865, + "step": 47984 + }, + { + "epoch": 8.73, + "learning_rate": 3.8250819075354936e-07, + "loss": 0.027, + "step": 47985 + }, + { + "epoch": 8.73, + "learning_rate": 3.824535857298872e-07, + "loss": 0.0116, + "step": 47986 + }, + { + "epoch": 8.73, + "learning_rate": 3.82398980706225e-07, + "loss": 1.2308, + "step": 47987 + }, + { + "epoch": 8.73, + "learning_rate": 3.823443756825628e-07, + "loss": 0.0219, + "step": 47988 + }, + { + "epoch": 8.73, + "learning_rate": 3.822897706589006e-07, + "loss": 4.4785, + "step": 47989 + }, + { + "epoch": 8.73, + "learning_rate": 3.8223516563523844e-07, + "loss": 0.1076, + "step": 47990 + }, + { + "epoch": 8.73, + "learning_rate": 3.8218056061157623e-07, + "loss": 0.4108, + "step": 47991 + }, + { + "epoch": 8.73, + "learning_rate": 3.821259555879141e-07, + "loss": 2.4998, + "step": 47992 + }, + { + "epoch": 8.73, + "learning_rate": 3.820713505642519e-07, + "loss": 0.0043, + "step": 47993 + }, + { + "epoch": 8.73, + "learning_rate": 3.8201674554058974e-07, + "loss": 0.14, + "step": 47994 + }, + { + "epoch": 8.73, + "learning_rate": 3.819621405169276e-07, + "loss": 0.6625, + "step": 47995 + }, + { + "epoch": 8.73, + "learning_rate": 3.8190753549326536e-07, + "loss": 0.114, + "step": 47996 + }, + { + "epoch": 8.73, + "learning_rate": 3.8185293046960325e-07, + "loss": 1.7599, + "step": 47997 + }, + { + "epoch": 8.73, + "learning_rate": 3.81798325445941e-07, + "loss": 0.3043, + "step": 47998 + }, + { + "epoch": 8.73, + "learning_rate": 3.817437204222789e-07, + "loss": 0.102, + "step": 47999 + }, + { + "epoch": 8.73, + "learning_rate": 3.8168911539861666e-07, + "loss": 0.3626, + "step": 48000 + }, + { + "epoch": 8.73, + "learning_rate": 3.816345103749545e-07, + "loss": 0.1342, + "step": 48001 + }, + { + "epoch": 8.73, + "learning_rate": 3.815799053512924e-07, + "loss": 0.2732, + "step": 48002 + }, + { + "epoch": 8.73, + "learning_rate": 3.815253003276301e-07, + "loss": 0.0594, + "step": 48003 + }, + { + "epoch": 8.73, + "learning_rate": 3.81470695303968e-07, + "loss": 0.0729, + "step": 48004 + }, + { + "epoch": 8.73, + "learning_rate": 3.814160902803058e-07, + "loss": 0.6768, + "step": 48005 + }, + { + "epoch": 8.73, + "learning_rate": 3.8136148525664363e-07, + "loss": 0.0064, + "step": 48006 + }, + { + "epoch": 8.73, + "learning_rate": 3.813068802329814e-07, + "loss": 0.0112, + "step": 48007 + }, + { + "epoch": 8.73, + "learning_rate": 3.8125227520931925e-07, + "loss": 0.0063, + "step": 48008 + }, + { + "epoch": 8.73, + "learning_rate": 3.8119767018565715e-07, + "loss": 1.8817, + "step": 48009 + }, + { + "epoch": 8.73, + "learning_rate": 3.811430651619949e-07, + "loss": 0.1376, + "step": 48010 + }, + { + "epoch": 8.73, + "learning_rate": 3.8108846013833277e-07, + "loss": 0.0455, + "step": 48011 + }, + { + "epoch": 8.73, + "learning_rate": 3.8103385511467055e-07, + "loss": 1.5406, + "step": 48012 + }, + { + "epoch": 8.73, + "learning_rate": 3.809792500910084e-07, + "loss": 1.4734, + "step": 48013 + }, + { + "epoch": 8.73, + "learning_rate": 3.809246450673462e-07, + "loss": 0.1327, + "step": 48014 + }, + { + "epoch": 8.73, + "learning_rate": 3.80870040043684e-07, + "loss": 0.0151, + "step": 48015 + }, + { + "epoch": 8.73, + "learning_rate": 3.808154350200219e-07, + "loss": 0.6656, + "step": 48016 + }, + { + "epoch": 8.73, + "learning_rate": 3.807608299963597e-07, + "loss": 0.4332, + "step": 48017 + }, + { + "epoch": 8.73, + "learning_rate": 3.807062249726975e-07, + "loss": 0.0096, + "step": 48018 + }, + { + "epoch": 8.73, + "learning_rate": 3.806516199490353e-07, + "loss": 0.2032, + "step": 48019 + }, + { + "epoch": 8.73, + "learning_rate": 3.8059701492537315e-07, + "loss": 0.059, + "step": 48020 + }, + { + "epoch": 8.73, + "learning_rate": 3.8054240990171093e-07, + "loss": 0.1962, + "step": 48021 + }, + { + "epoch": 8.73, + "learning_rate": 3.804878048780488e-07, + "loss": 0.0071, + "step": 48022 + }, + { + "epoch": 8.73, + "learning_rate": 3.8043319985438666e-07, + "loss": 0.5785, + "step": 48023 + }, + { + "epoch": 8.73, + "learning_rate": 3.8037859483072444e-07, + "loss": 0.0037, + "step": 48024 + }, + { + "epoch": 8.73, + "learning_rate": 3.803239898070623e-07, + "loss": 0.5191, + "step": 48025 + }, + { + "epoch": 8.73, + "learning_rate": 3.8026938478340007e-07, + "loss": 0.0443, + "step": 48026 + }, + { + "epoch": 8.73, + "learning_rate": 3.802147797597379e-07, + "loss": 0.253, + "step": 48027 + }, + { + "epoch": 8.73, + "learning_rate": 3.801601747360757e-07, + "loss": 0.2258, + "step": 48028 + }, + { + "epoch": 8.73, + "learning_rate": 3.801055697124136e-07, + "loss": 1.0273, + "step": 48029 + }, + { + "epoch": 8.73, + "learning_rate": 3.800509646887513e-07, + "loss": 0.0352, + "step": 48030 + }, + { + "epoch": 8.73, + "learning_rate": 3.799963596650892e-07, + "loss": 0.0417, + "step": 48031 + }, + { + "epoch": 8.73, + "learning_rate": 3.7994175464142704e-07, + "loss": 0.6871, + "step": 48032 + }, + { + "epoch": 8.73, + "learning_rate": 3.798871496177648e-07, + "loss": 0.004, + "step": 48033 + }, + { + "epoch": 8.74, + "learning_rate": 3.798325445941027e-07, + "loss": 0.0659, + "step": 48034 + }, + { + "epoch": 8.74, + "learning_rate": 3.7977793957044044e-07, + "loss": 0.1198, + "step": 48035 + }, + { + "epoch": 8.74, + "learning_rate": 3.7972333454677833e-07, + "loss": 0.011, + "step": 48036 + }, + { + "epoch": 8.74, + "learning_rate": 3.796687295231161e-07, + "loss": 0.0075, + "step": 48037 + }, + { + "epoch": 8.74, + "learning_rate": 3.7961412449945396e-07, + "loss": 1.8229, + "step": 48038 + }, + { + "epoch": 8.74, + "learning_rate": 3.795595194757918e-07, + "loss": 0.5626, + "step": 48039 + }, + { + "epoch": 8.74, + "learning_rate": 3.795049144521296e-07, + "loss": 2.329, + "step": 48040 + }, + { + "epoch": 8.74, + "learning_rate": 3.7945030942846747e-07, + "loss": 0.2496, + "step": 48041 + }, + { + "epoch": 8.74, + "learning_rate": 3.793957044048052e-07, + "loss": 0.0049, + "step": 48042 + }, + { + "epoch": 8.74, + "learning_rate": 3.793410993811431e-07, + "loss": 0.3313, + "step": 48043 + }, + { + "epoch": 8.74, + "learning_rate": 3.792864943574809e-07, + "loss": 1.5819, + "step": 48044 + }, + { + "epoch": 8.74, + "learning_rate": 3.792318893338187e-07, + "loss": 0.033, + "step": 48045 + }, + { + "epoch": 8.74, + "learning_rate": 3.791772843101566e-07, + "loss": 0.8842, + "step": 48046 + }, + { + "epoch": 8.74, + "learning_rate": 3.7912267928649434e-07, + "loss": 0.118, + "step": 48047 + }, + { + "epoch": 8.74, + "learning_rate": 3.790680742628322e-07, + "loss": 1.4073, + "step": 48048 + }, + { + "epoch": 8.74, + "learning_rate": 3.7901346923917e-07, + "loss": 0.6029, + "step": 48049 + }, + { + "epoch": 8.74, + "learning_rate": 3.7895886421550785e-07, + "loss": 0.0043, + "step": 48050 + }, + { + "epoch": 8.74, + "learning_rate": 3.7890425919184563e-07, + "loss": 0.5695, + "step": 48051 + }, + { + "epoch": 8.74, + "learning_rate": 3.7884965416818347e-07, + "loss": 0.0303, + "step": 48052 + }, + { + "epoch": 8.74, + "learning_rate": 3.7879504914452136e-07, + "loss": 0.0561, + "step": 48053 + }, + { + "epoch": 8.74, + "learning_rate": 3.7874044412085915e-07, + "loss": 2.0357, + "step": 48054 + }, + { + "epoch": 8.74, + "learning_rate": 3.78685839097197e-07, + "loss": 0.1491, + "step": 48055 + }, + { + "epoch": 8.74, + "learning_rate": 3.7863123407353477e-07, + "loss": 0.4838, + "step": 48056 + }, + { + "epoch": 8.74, + "learning_rate": 3.785766290498726e-07, + "loss": 1.4829, + "step": 48057 + }, + { + "epoch": 8.74, + "learning_rate": 3.785220240262104e-07, + "loss": 0.2745, + "step": 48058 + }, + { + "epoch": 8.74, + "learning_rate": 3.7846741900254823e-07, + "loss": 0.1129, + "step": 48059 + }, + { + "epoch": 8.74, + "learning_rate": 3.784128139788861e-07, + "loss": 0.017, + "step": 48060 + }, + { + "epoch": 8.74, + "learning_rate": 3.783582089552239e-07, + "loss": 0.6763, + "step": 48061 + }, + { + "epoch": 8.74, + "learning_rate": 3.7830360393156174e-07, + "loss": 2.6862, + "step": 48062 + }, + { + "epoch": 8.74, + "learning_rate": 3.782489989078995e-07, + "loss": 0.5405, + "step": 48063 + }, + { + "epoch": 8.74, + "learning_rate": 3.7819439388423736e-07, + "loss": 0.1531, + "step": 48064 + }, + { + "epoch": 8.74, + "learning_rate": 3.7813978886057515e-07, + "loss": 1.1117, + "step": 48065 + }, + { + "epoch": 8.74, + "learning_rate": 3.7808518383691304e-07, + "loss": 1.1354, + "step": 48066 + }, + { + "epoch": 8.74, + "learning_rate": 3.7803057881325077e-07, + "loss": 0.1777, + "step": 48067 + }, + { + "epoch": 8.74, + "learning_rate": 3.7797597378958866e-07, + "loss": 0.3499, + "step": 48068 + }, + { + "epoch": 8.74, + "learning_rate": 3.779213687659265e-07, + "loss": 0.1177, + "step": 48069 + }, + { + "epoch": 8.74, + "learning_rate": 3.778667637422643e-07, + "loss": 0.1166, + "step": 48070 + }, + { + "epoch": 8.74, + "learning_rate": 3.778121587186021e-07, + "loss": 0.0364, + "step": 48071 + }, + { + "epoch": 8.74, + "learning_rate": 3.777575536949399e-07, + "loss": 0.21, + "step": 48072 + }, + { + "epoch": 8.74, + "learning_rate": 3.777029486712778e-07, + "loss": 1.1683, + "step": 48073 + }, + { + "epoch": 8.74, + "learning_rate": 3.776483436476155e-07, + "loss": 0.1674, + "step": 48074 + }, + { + "epoch": 8.74, + "learning_rate": 3.775937386239534e-07, + "loss": 1.8756, + "step": 48075 + }, + { + "epoch": 8.74, + "learning_rate": 3.7753913360029125e-07, + "loss": 0.7999, + "step": 48076 + }, + { + "epoch": 8.74, + "learning_rate": 3.7748452857662904e-07, + "loss": 0.1449, + "step": 48077 + }, + { + "epoch": 8.74, + "learning_rate": 3.7742992355296693e-07, + "loss": 0.2028, + "step": 48078 + }, + { + "epoch": 8.74, + "learning_rate": 3.7737531852930466e-07, + "loss": 0.0998, + "step": 48079 + }, + { + "epoch": 8.74, + "learning_rate": 3.7732071350564255e-07, + "loss": 0.0058, + "step": 48080 + }, + { + "epoch": 8.74, + "learning_rate": 3.7726610848198033e-07, + "loss": 0.6062, + "step": 48081 + }, + { + "epoch": 8.74, + "learning_rate": 3.7721150345831817e-07, + "loss": 0.0032, + "step": 48082 + }, + { + "epoch": 8.74, + "learning_rate": 3.77156898434656e-07, + "loss": 1.0002, + "step": 48083 + }, + { + "epoch": 8.74, + "learning_rate": 3.771022934109938e-07, + "loss": 1.6799, + "step": 48084 + }, + { + "epoch": 8.74, + "learning_rate": 3.770476883873317e-07, + "loss": 1.5069, + "step": 48085 + }, + { + "epoch": 8.74, + "learning_rate": 3.7699308336366947e-07, + "loss": 2.5369, + "step": 48086 + }, + { + "epoch": 8.74, + "learning_rate": 3.769384783400073e-07, + "loss": 0.9388, + "step": 48087 + }, + { + "epoch": 8.74, + "learning_rate": 3.768838733163451e-07, + "loss": 0.6207, + "step": 48088 + }, + { + "epoch": 8.75, + "learning_rate": 3.7682926829268293e-07, + "loss": 0.6078, + "step": 48089 + }, + { + "epoch": 8.75, + "learning_rate": 3.767746632690208e-07, + "loss": 1.341, + "step": 48090 + }, + { + "epoch": 8.75, + "learning_rate": 3.7672005824535855e-07, + "loss": 0.0287, + "step": 48091 + }, + { + "epoch": 8.75, + "learning_rate": 3.7666545322169644e-07, + "loss": 0.1188, + "step": 48092 + }, + { + "epoch": 8.75, + "learning_rate": 3.766108481980342e-07, + "loss": 0.0085, + "step": 48093 + }, + { + "epoch": 8.75, + "learning_rate": 3.7655624317437206e-07, + "loss": 0.6961, + "step": 48094 + }, + { + "epoch": 8.75, + "learning_rate": 3.7650163815070985e-07, + "loss": 0.2034, + "step": 48095 + }, + { + "epoch": 8.75, + "learning_rate": 3.764470331270477e-07, + "loss": 0.0831, + "step": 48096 + }, + { + "epoch": 8.75, + "learning_rate": 3.763924281033856e-07, + "loss": 0.0239, + "step": 48097 + }, + { + "epoch": 8.75, + "learning_rate": 3.7633782307972336e-07, + "loss": 0.6637, + "step": 48098 + }, + { + "epoch": 8.75, + "learning_rate": 3.762832180560612e-07, + "loss": 0.2828, + "step": 48099 + }, + { + "epoch": 8.75, + "learning_rate": 3.76228613032399e-07, + "loss": 0.3793, + "step": 48100 + }, + { + "epoch": 8.75, + "learning_rate": 3.761740080087368e-07, + "loss": 0.1585, + "step": 48101 + }, + { + "epoch": 8.75, + "learning_rate": 3.761194029850746e-07, + "loss": 1.0718, + "step": 48102 + }, + { + "epoch": 8.75, + "learning_rate": 3.7606479796141244e-07, + "loss": 1.0085, + "step": 48103 + }, + { + "epoch": 8.75, + "learning_rate": 3.7601019293775023e-07, + "loss": 0.0241, + "step": 48104 + }, + { + "epoch": 8.75, + "learning_rate": 3.759555879140881e-07, + "loss": 2.7581, + "step": 48105 + }, + { + "epoch": 8.75, + "learning_rate": 3.7590098289042596e-07, + "loss": 1.0785, + "step": 48106 + }, + { + "epoch": 8.75, + "learning_rate": 3.7584637786676374e-07, + "loss": 0.1795, + "step": 48107 + }, + { + "epoch": 8.75, + "learning_rate": 3.757917728431016e-07, + "loss": 0.0045, + "step": 48108 + }, + { + "epoch": 8.75, + "learning_rate": 3.7573716781943936e-07, + "loss": 0.0404, + "step": 48109 + }, + { + "epoch": 8.75, + "learning_rate": 3.7568256279577725e-07, + "loss": 1.9662, + "step": 48110 + }, + { + "epoch": 8.75, + "learning_rate": 3.75627957772115e-07, + "loss": 1.0971, + "step": 48111 + }, + { + "epoch": 8.75, + "learning_rate": 3.755733527484529e-07, + "loss": 0.0359, + "step": 48112 + }, + { + "epoch": 8.75, + "learning_rate": 3.755187477247907e-07, + "loss": 2.5626, + "step": 48113 + }, + { + "epoch": 8.75, + "learning_rate": 3.754641427011285e-07, + "loss": 0.0835, + "step": 48114 + }, + { + "epoch": 8.75, + "learning_rate": 3.7540953767746633e-07, + "loss": 1.2772, + "step": 48115 + }, + { + "epoch": 8.75, + "learning_rate": 3.753549326538041e-07, + "loss": 0.0122, + "step": 48116 + }, + { + "epoch": 8.75, + "learning_rate": 3.75300327630142e-07, + "loss": 2.9819, + "step": 48117 + }, + { + "epoch": 8.75, + "learning_rate": 3.752457226064798e-07, + "loss": 0.7059, + "step": 48118 + }, + { + "epoch": 8.75, + "learning_rate": 3.7519111758281763e-07, + "loss": 0.253, + "step": 48119 + }, + { + "epoch": 8.75, + "learning_rate": 3.7513651255915547e-07, + "loss": 0.0522, + "step": 48120 + }, + { + "epoch": 8.75, + "learning_rate": 3.7508190753549325e-07, + "loss": 1.4389, + "step": 48121 + }, + { + "epoch": 8.75, + "learning_rate": 3.7502730251183114e-07, + "loss": 0.0715, + "step": 48122 + }, + { + "epoch": 8.75, + "learning_rate": 3.7497269748816893e-07, + "loss": 1.8735, + "step": 48123 + }, + { + "epoch": 8.75, + "learning_rate": 3.7491809246450677e-07, + "loss": 0.3771, + "step": 48124 + }, + { + "epoch": 8.75, + "learning_rate": 3.7486348744084455e-07, + "loss": 0.0513, + "step": 48125 + }, + { + "epoch": 8.75, + "learning_rate": 3.748088824171824e-07, + "loss": 0.0294, + "step": 48126 + }, + { + "epoch": 8.75, + "learning_rate": 3.747542773935202e-07, + "loss": 0.8199, + "step": 48127 + }, + { + "epoch": 8.75, + "learning_rate": 3.74699672369858e-07, + "loss": 0.0617, + "step": 48128 + }, + { + "epoch": 8.75, + "learning_rate": 3.7464506734619585e-07, + "loss": 0.3508, + "step": 48129 + }, + { + "epoch": 8.75, + "learning_rate": 3.745904623225337e-07, + "loss": 0.6481, + "step": 48130 + }, + { + "epoch": 8.75, + "learning_rate": 3.745358572988715e-07, + "loss": 0.0053, + "step": 48131 + }, + { + "epoch": 8.75, + "learning_rate": 3.7448125227520936e-07, + "loss": 0.1607, + "step": 48132 + }, + { + "epoch": 8.75, + "learning_rate": 3.7442664725154714e-07, + "loss": 1.5378, + "step": 48133 + }, + { + "epoch": 8.75, + "learning_rate": 3.74372042227885e-07, + "loss": 0.4612, + "step": 48134 + }, + { + "epoch": 8.75, + "learning_rate": 3.7431743720422277e-07, + "loss": 0.5631, + "step": 48135 + }, + { + "epoch": 8.75, + "learning_rate": 3.742628321805606e-07, + "loss": 1.4364, + "step": 48136 + }, + { + "epoch": 8.75, + "learning_rate": 3.7420822715689844e-07, + "loss": 0.1195, + "step": 48137 + }, + { + "epoch": 8.75, + "learning_rate": 3.741536221332363e-07, + "loss": 0.7784, + "step": 48138 + }, + { + "epoch": 8.75, + "learning_rate": 3.740990171095741e-07, + "loss": 0.0546, + "step": 48139 + }, + { + "epoch": 8.75, + "learning_rate": 3.740444120859119e-07, + "loss": 1.4265, + "step": 48140 + }, + { + "epoch": 8.75, + "learning_rate": 3.7398980706224974e-07, + "loss": 1.6021, + "step": 48141 + }, + { + "epoch": 8.75, + "learning_rate": 3.739352020385876e-07, + "loss": 0.1197, + "step": 48142 + }, + { + "epoch": 8.75, + "learning_rate": 3.7388059701492536e-07, + "loss": 1.8626, + "step": 48143 + }, + { + "epoch": 8.76, + "learning_rate": 3.738259919912632e-07, + "loss": 0.0742, + "step": 48144 + }, + { + "epoch": 8.76, + "learning_rate": 3.7377138696760104e-07, + "loss": 0.0129, + "step": 48145 + }, + { + "epoch": 8.76, + "learning_rate": 3.7371678194393887e-07, + "loss": 0.0408, + "step": 48146 + }, + { + "epoch": 8.76, + "learning_rate": 3.7366217692027666e-07, + "loss": 0.151, + "step": 48147 + }, + { + "epoch": 8.76, + "learning_rate": 3.736075718966145e-07, + "loss": 0.3344, + "step": 48148 + }, + { + "epoch": 8.76, + "learning_rate": 3.7355296687295233e-07, + "loss": 0.0049, + "step": 48149 + }, + { + "epoch": 8.76, + "learning_rate": 3.734983618492901e-07, + "loss": 0.0325, + "step": 48150 + }, + { + "epoch": 8.76, + "learning_rate": 3.7344375682562796e-07, + "loss": 0.1527, + "step": 48151 + }, + { + "epoch": 8.76, + "learning_rate": 3.733891518019658e-07, + "loss": 0.0505, + "step": 48152 + }, + { + "epoch": 8.76, + "learning_rate": 3.7333454677830363e-07, + "loss": 0.5133, + "step": 48153 + }, + { + "epoch": 8.76, + "learning_rate": 3.7327994175464147e-07, + "loss": 0.4775, + "step": 48154 + }, + { + "epoch": 8.76, + "learning_rate": 3.7322533673097925e-07, + "loss": 0.0033, + "step": 48155 + }, + { + "epoch": 8.76, + "learning_rate": 3.731707317073171e-07, + "loss": 0.2779, + "step": 48156 + }, + { + "epoch": 8.76, + "learning_rate": 3.731161266836549e-07, + "loss": 0.2165, + "step": 48157 + }, + { + "epoch": 8.76, + "learning_rate": 3.730615216599927e-07, + "loss": 0.0092, + "step": 48158 + }, + { + "epoch": 8.76, + "learning_rate": 3.7300691663633055e-07, + "loss": 0.188, + "step": 48159 + }, + { + "epoch": 8.76, + "learning_rate": 3.729523116126684e-07, + "loss": 0.0098, + "step": 48160 + }, + { + "epoch": 8.76, + "learning_rate": 3.728977065890062e-07, + "loss": 0.5538, + "step": 48161 + }, + { + "epoch": 8.76, + "learning_rate": 3.72843101565344e-07, + "loss": 0.0125, + "step": 48162 + }, + { + "epoch": 8.76, + "learning_rate": 3.7278849654168185e-07, + "loss": 0.3996, + "step": 48163 + }, + { + "epoch": 8.76, + "learning_rate": 3.727338915180197e-07, + "loss": 0.3013, + "step": 48164 + }, + { + "epoch": 8.76, + "learning_rate": 3.7267928649435747e-07, + "loss": 0.0493, + "step": 48165 + }, + { + "epoch": 8.76, + "learning_rate": 3.726246814706953e-07, + "loss": 0.0074, + "step": 48166 + }, + { + "epoch": 8.76, + "learning_rate": 3.7257007644703314e-07, + "loss": 1.3146, + "step": 48167 + }, + { + "epoch": 8.76, + "learning_rate": 3.72515471423371e-07, + "loss": 0.0241, + "step": 48168 + }, + { + "epoch": 8.76, + "learning_rate": 3.7246086639970877e-07, + "loss": 0.0504, + "step": 48169 + }, + { + "epoch": 8.76, + "learning_rate": 3.724062613760466e-07, + "loss": 0.7744, + "step": 48170 + }, + { + "epoch": 8.76, + "learning_rate": 3.7235165635238444e-07, + "loss": 0.5547, + "step": 48171 + }, + { + "epoch": 8.76, + "learning_rate": 3.722970513287222e-07, + "loss": 1.6631, + "step": 48172 + }, + { + "epoch": 8.76, + "learning_rate": 3.7224244630506006e-07, + "loss": 0.9943, + "step": 48173 + }, + { + "epoch": 8.76, + "learning_rate": 3.721878412813979e-07, + "loss": 3.5055, + "step": 48174 + }, + { + "epoch": 8.76, + "learning_rate": 3.7213323625773574e-07, + "loss": 0.1746, + "step": 48175 + }, + { + "epoch": 8.76, + "learning_rate": 3.720786312340736e-07, + "loss": 0.0715, + "step": 48176 + }, + { + "epoch": 8.76, + "learning_rate": 3.7202402621041136e-07, + "loss": 0.248, + "step": 48177 + }, + { + "epoch": 8.76, + "learning_rate": 3.719694211867492e-07, + "loss": 0.092, + "step": 48178 + }, + { + "epoch": 8.76, + "learning_rate": 3.71914816163087e-07, + "loss": 0.7249, + "step": 48179 + }, + { + "epoch": 8.76, + "learning_rate": 3.718602111394248e-07, + "loss": 0.0474, + "step": 48180 + }, + { + "epoch": 8.76, + "learning_rate": 3.7180560611576266e-07, + "loss": 0.6373, + "step": 48181 + }, + { + "epoch": 8.76, + "learning_rate": 3.717510010921005e-07, + "loss": 1.1297, + "step": 48182 + }, + { + "epoch": 8.76, + "learning_rate": 3.7169639606843833e-07, + "loss": 0.998, + "step": 48183 + }, + { + "epoch": 8.76, + "learning_rate": 3.716417910447761e-07, + "loss": 1.4636, + "step": 48184 + }, + { + "epoch": 8.76, + "learning_rate": 3.7158718602111395e-07, + "loss": 0.0703, + "step": 48185 + }, + { + "epoch": 8.76, + "learning_rate": 3.715325809974518e-07, + "loss": 0.0478, + "step": 48186 + }, + { + "epoch": 8.76, + "learning_rate": 3.714779759737896e-07, + "loss": 1.8642, + "step": 48187 + }, + { + "epoch": 8.76, + "learning_rate": 3.714233709501274e-07, + "loss": 0.0114, + "step": 48188 + }, + { + "epoch": 8.76, + "learning_rate": 3.7136876592646525e-07, + "loss": 0.0079, + "step": 48189 + }, + { + "epoch": 8.76, + "learning_rate": 3.713141609028031e-07, + "loss": 1.0073, + "step": 48190 + }, + { + "epoch": 8.76, + "learning_rate": 3.712595558791409e-07, + "loss": 0.0412, + "step": 48191 + }, + { + "epoch": 8.76, + "learning_rate": 3.712049508554787e-07, + "loss": 0.0055, + "step": 48192 + }, + { + "epoch": 8.76, + "learning_rate": 3.7115034583181655e-07, + "loss": 0.0082, + "step": 48193 + }, + { + "epoch": 8.76, + "learning_rate": 3.7109574080815433e-07, + "loss": 0.0526, + "step": 48194 + }, + { + "epoch": 8.76, + "learning_rate": 3.7104113578449217e-07, + "loss": 1.1401, + "step": 48195 + }, + { + "epoch": 8.76, + "learning_rate": 3.7098653076083e-07, + "loss": 0.0075, + "step": 48196 + }, + { + "epoch": 8.76, + "learning_rate": 3.7093192573716785e-07, + "loss": 0.1114, + "step": 48197 + }, + { + "epoch": 8.76, + "learning_rate": 3.708773207135057e-07, + "loss": 0.055, + "step": 48198 + }, + { + "epoch": 8.77, + "learning_rate": 3.7082271568984347e-07, + "loss": 0.3431, + "step": 48199 + }, + { + "epoch": 8.77, + "learning_rate": 3.707681106661813e-07, + "loss": 0.569, + "step": 48200 + }, + { + "epoch": 8.77, + "learning_rate": 3.707135056425191e-07, + "loss": 0.2383, + "step": 48201 + }, + { + "epoch": 8.77, + "learning_rate": 3.7065890061885693e-07, + "loss": 0.706, + "step": 48202 + }, + { + "epoch": 8.77, + "learning_rate": 3.7060429559519476e-07, + "loss": 0.2499, + "step": 48203 + }, + { + "epoch": 8.77, + "learning_rate": 3.705496905715326e-07, + "loss": 0.2876, + "step": 48204 + }, + { + "epoch": 8.77, + "learning_rate": 3.7049508554787044e-07, + "loss": 0.1357, + "step": 48205 + }, + { + "epoch": 8.77, + "learning_rate": 3.704404805242082e-07, + "loss": 0.0204, + "step": 48206 + }, + { + "epoch": 8.77, + "learning_rate": 3.7038587550054606e-07, + "loss": 0.0059, + "step": 48207 + }, + { + "epoch": 8.77, + "learning_rate": 3.703312704768839e-07, + "loss": 2.5642, + "step": 48208 + }, + { + "epoch": 8.77, + "learning_rate": 3.702766654532217e-07, + "loss": 0.0697, + "step": 48209 + }, + { + "epoch": 8.77, + "learning_rate": 3.702220604295595e-07, + "loss": 1.283, + "step": 48210 + }, + { + "epoch": 8.77, + "learning_rate": 3.701674554058973e-07, + "loss": 0.2872, + "step": 48211 + }, + { + "epoch": 8.77, + "learning_rate": 3.701128503822352e-07, + "loss": 0.0658, + "step": 48212 + }, + { + "epoch": 8.77, + "learning_rate": 3.7005824535857303e-07, + "loss": 0.1386, + "step": 48213 + }, + { + "epoch": 8.77, + "learning_rate": 3.700036403349108e-07, + "loss": 0.1825, + "step": 48214 + }, + { + "epoch": 8.77, + "learning_rate": 3.6994903531124866e-07, + "loss": 0.0984, + "step": 48215 + }, + { + "epoch": 8.77, + "learning_rate": 3.6989443028758644e-07, + "loss": 0.0671, + "step": 48216 + }, + { + "epoch": 8.77, + "learning_rate": 3.698398252639243e-07, + "loss": 0.3822, + "step": 48217 + }, + { + "epoch": 8.77, + "learning_rate": 3.697852202402621e-07, + "loss": 1.7167, + "step": 48218 + }, + { + "epoch": 8.77, + "learning_rate": 3.6973061521659995e-07, + "loss": 0.005, + "step": 48219 + }, + { + "epoch": 8.77, + "learning_rate": 3.696760101929378e-07, + "loss": 0.4523, + "step": 48220 + }, + { + "epoch": 8.77, + "learning_rate": 3.696214051692756e-07, + "loss": 0.1988, + "step": 48221 + }, + { + "epoch": 8.77, + "learning_rate": 3.695668001456134e-07, + "loss": 0.163, + "step": 48222 + }, + { + "epoch": 8.77, + "learning_rate": 3.695121951219512e-07, + "loss": 0.0203, + "step": 48223 + }, + { + "epoch": 8.77, + "learning_rate": 3.6945759009828904e-07, + "loss": 0.1993, + "step": 48224 + }, + { + "epoch": 8.77, + "learning_rate": 3.6940298507462687e-07, + "loss": 0.0047, + "step": 48225 + }, + { + "epoch": 8.77, + "learning_rate": 3.693483800509647e-07, + "loss": 0.5573, + "step": 48226 + }, + { + "epoch": 8.77, + "learning_rate": 3.6929377502730255e-07, + "loss": 0.0049, + "step": 48227 + }, + { + "epoch": 8.77, + "learning_rate": 3.6923917000364033e-07, + "loss": 0.0158, + "step": 48228 + }, + { + "epoch": 8.77, + "learning_rate": 3.6918456497997817e-07, + "loss": 2.753, + "step": 48229 + }, + { + "epoch": 8.77, + "learning_rate": 3.69129959956316e-07, + "loss": 0.787, + "step": 48230 + }, + { + "epoch": 8.77, + "learning_rate": 3.690753549326538e-07, + "loss": 0.0242, + "step": 48231 + }, + { + "epoch": 8.77, + "learning_rate": 3.6902074990899163e-07, + "loss": 0.0071, + "step": 48232 + }, + { + "epoch": 8.77, + "learning_rate": 3.689661448853294e-07, + "loss": 0.5297, + "step": 48233 + }, + { + "epoch": 8.77, + "learning_rate": 3.689115398616673e-07, + "loss": 0.2795, + "step": 48234 + }, + { + "epoch": 8.77, + "learning_rate": 3.6885693483800514e-07, + "loss": 0.084, + "step": 48235 + }, + { + "epoch": 8.77, + "learning_rate": 3.6880232981434293e-07, + "loss": 1.3918, + "step": 48236 + }, + { + "epoch": 8.77, + "learning_rate": 3.6874772479068076e-07, + "loss": 0.1632, + "step": 48237 + }, + { + "epoch": 8.77, + "learning_rate": 3.6869311976701855e-07, + "loss": 0.3657, + "step": 48238 + }, + { + "epoch": 8.77, + "learning_rate": 3.686385147433564e-07, + "loss": 0.0849, + "step": 48239 + }, + { + "epoch": 8.77, + "learning_rate": 3.685839097196942e-07, + "loss": 0.1271, + "step": 48240 + }, + { + "epoch": 8.77, + "learning_rate": 3.6852930469603206e-07, + "loss": 0.7014, + "step": 48241 + }, + { + "epoch": 8.77, + "learning_rate": 3.684746996723699e-07, + "loss": 0.367, + "step": 48242 + }, + { + "epoch": 8.77, + "learning_rate": 3.684200946487077e-07, + "loss": 0.1587, + "step": 48243 + }, + { + "epoch": 8.77, + "learning_rate": 3.683654896250455e-07, + "loss": 0.2412, + "step": 48244 + }, + { + "epoch": 8.77, + "learning_rate": 3.6831088460138336e-07, + "loss": 0.4116, + "step": 48245 + }, + { + "epoch": 8.77, + "learning_rate": 3.6825627957772114e-07, + "loss": 0.0117, + "step": 48246 + }, + { + "epoch": 8.77, + "learning_rate": 3.68201674554059e-07, + "loss": 0.0555, + "step": 48247 + }, + { + "epoch": 8.77, + "learning_rate": 3.681470695303968e-07, + "loss": 0.2582, + "step": 48248 + }, + { + "epoch": 8.77, + "learning_rate": 3.6809246450673466e-07, + "loss": 0.9775, + "step": 48249 + }, + { + "epoch": 8.77, + "learning_rate": 3.6803785948307244e-07, + "loss": 0.0222, + "step": 48250 + }, + { + "epoch": 8.77, + "learning_rate": 3.679832544594103e-07, + "loss": 0.3533, + "step": 48251 + }, + { + "epoch": 8.77, + "learning_rate": 3.679286494357481e-07, + "loss": 0.0994, + "step": 48252 + }, + { + "epoch": 8.77, + "learning_rate": 3.678740444120859e-07, + "loss": 1.1328, + "step": 48253 + }, + { + "epoch": 8.78, + "learning_rate": 3.6781943938842374e-07, + "loss": 0.1969, + "step": 48254 + }, + { + "epoch": 8.78, + "learning_rate": 3.677648343647616e-07, + "loss": 0.7043, + "step": 48255 + }, + { + "epoch": 8.78, + "learning_rate": 3.677102293410994e-07, + "loss": 0.3931, + "step": 48256 + }, + { + "epoch": 8.78, + "learning_rate": 3.6765562431743725e-07, + "loss": 0.7722, + "step": 48257 + }, + { + "epoch": 8.78, + "learning_rate": 3.6760101929377503e-07, + "loss": 0.0636, + "step": 48258 + }, + { + "epoch": 8.78, + "learning_rate": 3.6754641427011287e-07, + "loss": 0.0679, + "step": 48259 + }, + { + "epoch": 8.78, + "learning_rate": 3.6749180924645066e-07, + "loss": 1.6543, + "step": 48260 + }, + { + "epoch": 8.78, + "learning_rate": 3.674372042227885e-07, + "loss": 0.2163, + "step": 48261 + }, + { + "epoch": 8.78, + "learning_rate": 3.6738259919912633e-07, + "loss": 0.1542, + "step": 48262 + }, + { + "epoch": 8.78, + "learning_rate": 3.6732799417546417e-07, + "loss": 1.7276, + "step": 48263 + }, + { + "epoch": 8.78, + "learning_rate": 3.67273389151802e-07, + "loss": 3.0816, + "step": 48264 + }, + { + "epoch": 8.78, + "learning_rate": 3.672187841281398e-07, + "loss": 0.0465, + "step": 48265 + }, + { + "epoch": 8.78, + "learning_rate": 3.6716417910447763e-07, + "loss": 0.0093, + "step": 48266 + }, + { + "epoch": 8.78, + "learning_rate": 3.6710957408081547e-07, + "loss": 3.1018, + "step": 48267 + }, + { + "epoch": 8.78, + "learning_rate": 3.6705496905715325e-07, + "loss": 0.2113, + "step": 48268 + }, + { + "epoch": 8.78, + "learning_rate": 3.670003640334911e-07, + "loss": 0.7079, + "step": 48269 + }, + { + "epoch": 8.78, + "learning_rate": 3.6694575900982887e-07, + "loss": 0.6375, + "step": 48270 + }, + { + "epoch": 8.78, + "learning_rate": 3.6689115398616676e-07, + "loss": 0.3589, + "step": 48271 + }, + { + "epoch": 8.78, + "learning_rate": 3.6683654896250455e-07, + "loss": 0.0237, + "step": 48272 + }, + { + "epoch": 8.78, + "learning_rate": 3.667819439388424e-07, + "loss": 2.2383, + "step": 48273 + }, + { + "epoch": 8.78, + "learning_rate": 3.667273389151802e-07, + "loss": 0.0065, + "step": 48274 + }, + { + "epoch": 8.78, + "learning_rate": 3.66672733891518e-07, + "loss": 1.9822, + "step": 48275 + }, + { + "epoch": 8.78, + "learning_rate": 3.6661812886785584e-07, + "loss": 1.0287, + "step": 48276 + }, + { + "epoch": 8.78, + "learning_rate": 3.665635238441937e-07, + "loss": 0.0823, + "step": 48277 + }, + { + "epoch": 8.78, + "learning_rate": 3.665089188205315e-07, + "loss": 0.175, + "step": 48278 + }, + { + "epoch": 8.78, + "learning_rate": 3.6645431379686936e-07, + "loss": 0.0578, + "step": 48279 + }, + { + "epoch": 8.78, + "learning_rate": 3.6639970877320714e-07, + "loss": 2.3097, + "step": 48280 + }, + { + "epoch": 8.78, + "learning_rate": 3.66345103749545e-07, + "loss": 1.0771, + "step": 48281 + }, + { + "epoch": 8.78, + "learning_rate": 3.6629049872588276e-07, + "loss": 0.3756, + "step": 48282 + }, + { + "epoch": 8.78, + "learning_rate": 3.662358937022206e-07, + "loss": 0.6892, + "step": 48283 + }, + { + "epoch": 8.78, + "learning_rate": 3.6618128867855844e-07, + "loss": 0.041, + "step": 48284 + }, + { + "epoch": 8.78, + "learning_rate": 3.661266836548963e-07, + "loss": 0.129, + "step": 48285 + }, + { + "epoch": 8.78, + "learning_rate": 3.660720786312341e-07, + "loss": 0.0077, + "step": 48286 + }, + { + "epoch": 8.78, + "learning_rate": 3.660174736075719e-07, + "loss": 2.1734, + "step": 48287 + }, + { + "epoch": 8.78, + "learning_rate": 3.6596286858390974e-07, + "loss": 0.055, + "step": 48288 + }, + { + "epoch": 8.78, + "learning_rate": 3.659082635602476e-07, + "loss": 1.004, + "step": 48289 + }, + { + "epoch": 8.78, + "learning_rate": 3.6585365853658536e-07, + "loss": 0.009, + "step": 48290 + }, + { + "epoch": 8.78, + "learning_rate": 3.657990535129232e-07, + "loss": 0.0215, + "step": 48291 + }, + { + "epoch": 8.78, + "learning_rate": 3.65744448489261e-07, + "loss": 0.0773, + "step": 48292 + }, + { + "epoch": 8.78, + "learning_rate": 3.6568984346559887e-07, + "loss": 0.0101, + "step": 48293 + }, + { + "epoch": 8.78, + "learning_rate": 3.6563523844193666e-07, + "loss": 1.0321, + "step": 48294 + }, + { + "epoch": 8.78, + "learning_rate": 3.655806334182745e-07, + "loss": 0.5798, + "step": 48295 + }, + { + "epoch": 8.78, + "learning_rate": 3.6552602839461233e-07, + "loss": 0.0614, + "step": 48296 + }, + { + "epoch": 8.78, + "learning_rate": 3.654714233709501e-07, + "loss": 0.8732, + "step": 48297 + }, + { + "epoch": 8.78, + "learning_rate": 3.6541681834728795e-07, + "loss": 0.0599, + "step": 48298 + }, + { + "epoch": 8.78, + "learning_rate": 3.653622133236258e-07, + "loss": 0.0535, + "step": 48299 + }, + { + "epoch": 8.78, + "learning_rate": 3.6530760829996363e-07, + "loss": 0.2231, + "step": 48300 + }, + { + "epoch": 8.78, + "learning_rate": 3.6525300327630147e-07, + "loss": 0.0141, + "step": 48301 + }, + { + "epoch": 8.78, + "learning_rate": 3.6519839825263925e-07, + "loss": 1.5551, + "step": 48302 + }, + { + "epoch": 8.78, + "learning_rate": 3.651437932289771e-07, + "loss": 0.3914, + "step": 48303 + }, + { + "epoch": 8.78, + "learning_rate": 3.6508918820531487e-07, + "loss": 0.0052, + "step": 48304 + }, + { + "epoch": 8.78, + "learning_rate": 3.650345831816527e-07, + "loss": 0.0166, + "step": 48305 + }, + { + "epoch": 8.78, + "learning_rate": 3.6497997815799055e-07, + "loss": 0.0988, + "step": 48306 + }, + { + "epoch": 8.78, + "learning_rate": 3.6492537313432833e-07, + "loss": 0.2667, + "step": 48307 + }, + { + "epoch": 8.78, + "learning_rate": 3.648707681106662e-07, + "loss": 2.2032, + "step": 48308 + }, + { + "epoch": 8.79, + "learning_rate": 3.64816163087004e-07, + "loss": 0.6036, + "step": 48309 + }, + { + "epoch": 8.79, + "learning_rate": 3.6476155806334184e-07, + "loss": 0.7152, + "step": 48310 + }, + { + "epoch": 8.79, + "learning_rate": 3.647069530396797e-07, + "loss": 0.2796, + "step": 48311 + }, + { + "epoch": 8.79, + "learning_rate": 3.6465234801601747e-07, + "loss": 0.1656, + "step": 48312 + }, + { + "epoch": 8.79, + "learning_rate": 3.645977429923553e-07, + "loss": 0.0537, + "step": 48313 + }, + { + "epoch": 8.79, + "learning_rate": 3.645431379686931e-07, + "loss": 3.1722, + "step": 48314 + }, + { + "epoch": 8.79, + "learning_rate": 3.64488532945031e-07, + "loss": 0.0247, + "step": 48315 + }, + { + "epoch": 8.79, + "learning_rate": 3.6443392792136876e-07, + "loss": 0.0167, + "step": 48316 + }, + { + "epoch": 8.79, + "learning_rate": 3.643793228977066e-07, + "loss": 2.9742, + "step": 48317 + }, + { + "epoch": 8.79, + "learning_rate": 3.6432471787404444e-07, + "loss": 0.1772, + "step": 48318 + }, + { + "epoch": 8.79, + "learning_rate": 3.642701128503822e-07, + "loss": 0.0815, + "step": 48319 + }, + { + "epoch": 8.79, + "learning_rate": 3.6421550782672006e-07, + "loss": 0.0077, + "step": 48320 + }, + { + "epoch": 8.79, + "learning_rate": 3.641609028030579e-07, + "loss": 0.9855, + "step": 48321 + }, + { + "epoch": 8.79, + "learning_rate": 3.6410629777939574e-07, + "loss": 0.0076, + "step": 48322 + }, + { + "epoch": 8.79, + "learning_rate": 3.6405169275573357e-07, + "loss": 2.8321, + "step": 48323 + }, + { + "epoch": 8.79, + "learning_rate": 3.6399708773207136e-07, + "loss": 0.0089, + "step": 48324 + }, + { + "epoch": 8.79, + "learning_rate": 3.639424827084092e-07, + "loss": 0.5949, + "step": 48325 + }, + { + "epoch": 8.79, + "learning_rate": 3.63887877684747e-07, + "loss": 0.1182, + "step": 48326 + }, + { + "epoch": 8.79, + "learning_rate": 3.638332726610848e-07, + "loss": 0.0184, + "step": 48327 + }, + { + "epoch": 8.79, + "learning_rate": 3.6377866763742265e-07, + "loss": 0.0096, + "step": 48328 + }, + { + "epoch": 8.79, + "learning_rate": 3.6372406261376044e-07, + "loss": 0.3922, + "step": 48329 + }, + { + "epoch": 8.79, + "learning_rate": 3.6366945759009833e-07, + "loss": 1.5455, + "step": 48330 + }, + { + "epoch": 8.79, + "learning_rate": 3.636148525664361e-07, + "loss": 0.1816, + "step": 48331 + }, + { + "epoch": 8.79, + "learning_rate": 3.6356024754277395e-07, + "loss": 0.0077, + "step": 48332 + }, + { + "epoch": 8.79, + "learning_rate": 3.635056425191118e-07, + "loss": 0.0551, + "step": 48333 + }, + { + "epoch": 8.79, + "learning_rate": 3.634510374954496e-07, + "loss": 0.0052, + "step": 48334 + }, + { + "epoch": 8.79, + "learning_rate": 3.633964324717874e-07, + "loss": 2.1045, + "step": 48335 + }, + { + "epoch": 8.79, + "learning_rate": 3.633418274481252e-07, + "loss": 0.1102, + "step": 48336 + }, + { + "epoch": 8.79, + "learning_rate": 3.632872224244631e-07, + "loss": 0.9832, + "step": 48337 + }, + { + "epoch": 8.79, + "learning_rate": 3.6323261740080087e-07, + "loss": 0.1378, + "step": 48338 + }, + { + "epoch": 8.79, + "learning_rate": 3.631780123771387e-07, + "loss": 0.2991, + "step": 48339 + }, + { + "epoch": 8.79, + "learning_rate": 3.6312340735347655e-07, + "loss": 0.2867, + "step": 48340 + }, + { + "epoch": 8.79, + "learning_rate": 3.6306880232981433e-07, + "loss": 1.5897, + "step": 48341 + }, + { + "epoch": 8.79, + "learning_rate": 3.6301419730615217e-07, + "loss": 1.0508, + "step": 48342 + }, + { + "epoch": 8.79, + "learning_rate": 3.6295959228249e-07, + "loss": 2.2054, + "step": 48343 + }, + { + "epoch": 8.79, + "learning_rate": 3.6290498725882784e-07, + "loss": 1.7081, + "step": 48344 + }, + { + "epoch": 8.79, + "learning_rate": 3.628503822351657e-07, + "loss": 0.2896, + "step": 48345 + }, + { + "epoch": 8.79, + "learning_rate": 3.6279577721150347e-07, + "loss": 0.2332, + "step": 48346 + }, + { + "epoch": 8.79, + "learning_rate": 3.627411721878413e-07, + "loss": 0.3116, + "step": 48347 + }, + { + "epoch": 8.79, + "learning_rate": 3.626865671641791e-07, + "loss": 0.0051, + "step": 48348 + }, + { + "epoch": 8.79, + "learning_rate": 3.626319621405169e-07, + "loss": 0.0098, + "step": 48349 + }, + { + "epoch": 8.79, + "learning_rate": 3.6257735711685476e-07, + "loss": 0.0024, + "step": 48350 + }, + { + "epoch": 8.79, + "learning_rate": 3.6252275209319255e-07, + "loss": 1.9532, + "step": 48351 + }, + { + "epoch": 8.79, + "learning_rate": 3.6246814706953044e-07, + "loss": 0.0062, + "step": 48352 + }, + { + "epoch": 8.79, + "learning_rate": 3.624135420458682e-07, + "loss": 0.0037, + "step": 48353 + }, + { + "epoch": 8.79, + "learning_rate": 3.6235893702220606e-07, + "loss": 0.1844, + "step": 48354 + }, + { + "epoch": 8.79, + "learning_rate": 3.623043319985439e-07, + "loss": 0.45, + "step": 48355 + }, + { + "epoch": 8.79, + "learning_rate": 3.622497269748817e-07, + "loss": 0.017, + "step": 48356 + }, + { + "epoch": 8.79, + "learning_rate": 3.621951219512195e-07, + "loss": 1.0267, + "step": 48357 + }, + { + "epoch": 8.79, + "learning_rate": 3.621405169275573e-07, + "loss": 1.5682, + "step": 48358 + }, + { + "epoch": 8.79, + "learning_rate": 3.620859119038952e-07, + "loss": 1.0631, + "step": 48359 + }, + { + "epoch": 8.79, + "learning_rate": 3.62031306880233e-07, + "loss": 1.7887, + "step": 48360 + }, + { + "epoch": 8.79, + "learning_rate": 3.619767018565708e-07, + "loss": 0.0905, + "step": 48361 + }, + { + "epoch": 8.79, + "learning_rate": 3.6192209683290865e-07, + "loss": 0.5914, + "step": 48362 + }, + { + "epoch": 8.79, + "learning_rate": 3.6186749180924644e-07, + "loss": 0.0942, + "step": 48363 + }, + { + "epoch": 8.8, + "learning_rate": 3.618128867855843e-07, + "loss": 1.5574, + "step": 48364 + }, + { + "epoch": 8.8, + "learning_rate": 3.617582817619221e-07, + "loss": 0.2924, + "step": 48365 + }, + { + "epoch": 8.8, + "learning_rate": 3.617036767382599e-07, + "loss": 0.0046, + "step": 48366 + }, + { + "epoch": 8.8, + "learning_rate": 3.616490717145978e-07, + "loss": 0.2488, + "step": 48367 + }, + { + "epoch": 8.8, + "learning_rate": 3.6159446669093557e-07, + "loss": 0.0422, + "step": 48368 + }, + { + "epoch": 8.8, + "learning_rate": 3.615398616672734e-07, + "loss": 2.3437, + "step": 48369 + }, + { + "epoch": 8.8, + "learning_rate": 3.614852566436112e-07, + "loss": 0.0047, + "step": 48370 + }, + { + "epoch": 8.8, + "learning_rate": 3.6143065161994903e-07, + "loss": 2.5958, + "step": 48371 + }, + { + "epoch": 8.8, + "learning_rate": 3.6137604659628687e-07, + "loss": 0.6784, + "step": 48372 + }, + { + "epoch": 8.8, + "learning_rate": 3.6132144157262465e-07, + "loss": 0.252, + "step": 48373 + }, + { + "epoch": 8.8, + "learning_rate": 3.6126683654896255e-07, + "loss": 0.0109, + "step": 48374 + }, + { + "epoch": 8.8, + "learning_rate": 3.6121223152530033e-07, + "loss": 0.0249, + "step": 48375 + }, + { + "epoch": 8.8, + "learning_rate": 3.6115762650163817e-07, + "loss": 0.035, + "step": 48376 + }, + { + "epoch": 8.8, + "learning_rate": 3.61103021477976e-07, + "loss": 2.5044, + "step": 48377 + }, + { + "epoch": 8.8, + "learning_rate": 3.610484164543138e-07, + "loss": 0.7183, + "step": 48378 + }, + { + "epoch": 8.8, + "learning_rate": 3.6099381143065163e-07, + "loss": 2.2924, + "step": 48379 + }, + { + "epoch": 8.8, + "learning_rate": 3.609392064069894e-07, + "loss": 0.4726, + "step": 48380 + }, + { + "epoch": 8.8, + "learning_rate": 3.608846013833273e-07, + "loss": 0.5997, + "step": 48381 + }, + { + "epoch": 8.8, + "learning_rate": 3.608299963596651e-07, + "loss": 0.0392, + "step": 48382 + }, + { + "epoch": 8.8, + "learning_rate": 3.607753913360029e-07, + "loss": 0.0193, + "step": 48383 + }, + { + "epoch": 8.8, + "learning_rate": 3.6072078631234076e-07, + "loss": 1.9552, + "step": 48384 + }, + { + "epoch": 8.8, + "learning_rate": 3.6066618128867855e-07, + "loss": 0.003, + "step": 48385 + }, + { + "epoch": 8.8, + "learning_rate": 3.606115762650164e-07, + "loss": 0.5493, + "step": 48386 + }, + { + "epoch": 8.8, + "learning_rate": 3.605569712413542e-07, + "loss": 1.0208, + "step": 48387 + }, + { + "epoch": 8.8, + "learning_rate": 3.60502366217692e-07, + "loss": 0.2037, + "step": 48388 + }, + { + "epoch": 8.8, + "learning_rate": 3.604477611940299e-07, + "loss": 0.0095, + "step": 48389 + }, + { + "epoch": 8.8, + "learning_rate": 3.603931561703677e-07, + "loss": 0.04, + "step": 48390 + }, + { + "epoch": 8.8, + "learning_rate": 3.603385511467055e-07, + "loss": 0.4448, + "step": 48391 + }, + { + "epoch": 8.8, + "learning_rate": 3.602839461230433e-07, + "loss": 0.4155, + "step": 48392 + }, + { + "epoch": 8.8, + "learning_rate": 3.6022934109938114e-07, + "loss": 1.2295, + "step": 48393 + }, + { + "epoch": 8.8, + "learning_rate": 3.60174736075719e-07, + "loss": 0.0503, + "step": 48394 + }, + { + "epoch": 8.8, + "learning_rate": 3.6012013105205676e-07, + "loss": 0.1831, + "step": 48395 + }, + { + "epoch": 8.8, + "learning_rate": 3.6006552602839465e-07, + "loss": 0.0697, + "step": 48396 + }, + { + "epoch": 8.8, + "learning_rate": 3.6001092100473244e-07, + "loss": 0.007, + "step": 48397 + }, + { + "epoch": 8.8, + "learning_rate": 3.599563159810703e-07, + "loss": 0.1762, + "step": 48398 + }, + { + "epoch": 8.8, + "learning_rate": 3.599017109574081e-07, + "loss": 0.8164, + "step": 48399 + }, + { + "epoch": 8.8, + "learning_rate": 3.598471059337459e-07, + "loss": 0.7852, + "step": 48400 + }, + { + "epoch": 8.8, + "learning_rate": 3.5979250091008373e-07, + "loss": 2.4202, + "step": 48401 + }, + { + "epoch": 8.8, + "learning_rate": 3.597378958864215e-07, + "loss": 0.2357, + "step": 48402 + }, + { + "epoch": 8.8, + "learning_rate": 3.596832908627594e-07, + "loss": 0.9968, + "step": 48403 + }, + { + "epoch": 8.8, + "learning_rate": 3.5962868583909725e-07, + "loss": 0.182, + "step": 48404 + }, + { + "epoch": 8.8, + "learning_rate": 3.5957408081543503e-07, + "loss": 1.5788, + "step": 48405 + }, + { + "epoch": 8.8, + "learning_rate": 3.5951947579177287e-07, + "loss": 0.2708, + "step": 48406 + }, + { + "epoch": 8.8, + "learning_rate": 3.5946487076811065e-07, + "loss": 0.5359, + "step": 48407 + }, + { + "epoch": 8.8, + "learning_rate": 3.594102657444485e-07, + "loss": 0.1433, + "step": 48408 + }, + { + "epoch": 8.8, + "learning_rate": 3.5935566072078633e-07, + "loss": 1.1946, + "step": 48409 + }, + { + "epoch": 8.8, + "learning_rate": 3.593010556971241e-07, + "loss": 0.1337, + "step": 48410 + }, + { + "epoch": 8.8, + "learning_rate": 3.59246450673462e-07, + "loss": 0.0318, + "step": 48411 + }, + { + "epoch": 8.8, + "learning_rate": 3.591918456497998e-07, + "loss": 0.0853, + "step": 48412 + }, + { + "epoch": 8.8, + "learning_rate": 3.591372406261376e-07, + "loss": 0.1416, + "step": 48413 + }, + { + "epoch": 8.8, + "learning_rate": 3.5908263560247546e-07, + "loss": 0.0078, + "step": 48414 + }, + { + "epoch": 8.8, + "learning_rate": 3.5902803057881325e-07, + "loss": 0.9711, + "step": 48415 + }, + { + "epoch": 8.8, + "learning_rate": 3.589734255551511e-07, + "loss": 0.0448, + "step": 48416 + }, + { + "epoch": 8.8, + "learning_rate": 3.5891882053148887e-07, + "loss": 0.2446, + "step": 48417 + }, + { + "epoch": 8.8, + "learning_rate": 3.5886421550782676e-07, + "loss": 0.2262, + "step": 48418 + }, + { + "epoch": 8.81, + "learning_rate": 3.5880961048416455e-07, + "loss": 0.9485, + "step": 48419 + }, + { + "epoch": 8.81, + "learning_rate": 3.587550054605024e-07, + "loss": 0.6525, + "step": 48420 + }, + { + "epoch": 8.81, + "learning_rate": 3.587004004368402e-07, + "loss": 0.2332, + "step": 48421 + }, + { + "epoch": 8.81, + "learning_rate": 3.58645795413178e-07, + "loss": 0.6494, + "step": 48422 + }, + { + "epoch": 8.81, + "learning_rate": 3.5859119038951584e-07, + "loss": 0.0992, + "step": 48423 + }, + { + "epoch": 8.81, + "learning_rate": 3.5853658536585363e-07, + "loss": 0.0383, + "step": 48424 + }, + { + "epoch": 8.81, + "learning_rate": 3.5848198034219146e-07, + "loss": 1.1674, + "step": 48425 + }, + { + "epoch": 8.81, + "learning_rate": 3.5842737531852935e-07, + "loss": 0.2283, + "step": 48426 + }, + { + "epoch": 8.81, + "learning_rate": 3.5837277029486714e-07, + "loss": 0.0094, + "step": 48427 + }, + { + "epoch": 8.81, + "learning_rate": 3.58318165271205e-07, + "loss": 0.0924, + "step": 48428 + }, + { + "epoch": 8.81, + "learning_rate": 3.5826356024754276e-07, + "loss": 0.019, + "step": 48429 + }, + { + "epoch": 8.81, + "learning_rate": 3.582089552238806e-07, + "loss": 0.0699, + "step": 48430 + }, + { + "epoch": 8.81, + "learning_rate": 3.5815435020021844e-07, + "loss": 0.0051, + "step": 48431 + }, + { + "epoch": 8.81, + "learning_rate": 3.580997451765562e-07, + "loss": 0.6393, + "step": 48432 + }, + { + "epoch": 8.81, + "learning_rate": 3.580451401528941e-07, + "loss": 0.2722, + "step": 48433 + }, + { + "epoch": 8.81, + "learning_rate": 3.579905351292319e-07, + "loss": 0.0971, + "step": 48434 + }, + { + "epoch": 8.81, + "learning_rate": 3.5793593010556973e-07, + "loss": 1.4462, + "step": 48435 + }, + { + "epoch": 8.81, + "learning_rate": 3.5788132508190757e-07, + "loss": 1.1722, + "step": 48436 + }, + { + "epoch": 8.81, + "learning_rate": 3.5782672005824536e-07, + "loss": 0.0571, + "step": 48437 + }, + { + "epoch": 8.81, + "learning_rate": 3.577721150345832e-07, + "loss": 0.2252, + "step": 48438 + }, + { + "epoch": 8.81, + "learning_rate": 3.57717510010921e-07, + "loss": 0.0079, + "step": 48439 + }, + { + "epoch": 8.81, + "learning_rate": 3.5766290498725887e-07, + "loss": 0.0078, + "step": 48440 + }, + { + "epoch": 8.81, + "learning_rate": 3.5760829996359665e-07, + "loss": 0.005, + "step": 48441 + }, + { + "epoch": 8.81, + "learning_rate": 3.575536949399345e-07, + "loss": 0.0273, + "step": 48442 + }, + { + "epoch": 8.81, + "learning_rate": 3.5749908991627233e-07, + "loss": 0.0079, + "step": 48443 + }, + { + "epoch": 8.81, + "learning_rate": 3.574444848926101e-07, + "loss": 0.0755, + "step": 48444 + }, + { + "epoch": 8.81, + "learning_rate": 3.5738987986894795e-07, + "loss": 2.1231, + "step": 48445 + }, + { + "epoch": 8.81, + "learning_rate": 3.573352748452858e-07, + "loss": 0.5255, + "step": 48446 + }, + { + "epoch": 8.81, + "learning_rate": 3.5728066982162357e-07, + "loss": 0.1755, + "step": 48447 + }, + { + "epoch": 8.81, + "learning_rate": 3.5722606479796146e-07, + "loss": 1.8545, + "step": 48448 + }, + { + "epoch": 8.81, + "learning_rate": 3.5717145977429925e-07, + "loss": 1.5659, + "step": 48449 + }, + { + "epoch": 8.81, + "learning_rate": 3.571168547506371e-07, + "loss": 0.5468, + "step": 48450 + }, + { + "epoch": 8.81, + "learning_rate": 3.5706224972697487e-07, + "loss": 1.1623, + "step": 48451 + }, + { + "epoch": 8.81, + "learning_rate": 3.570076447033127e-07, + "loss": 0.073, + "step": 48452 + }, + { + "epoch": 8.81, + "learning_rate": 3.5695303967965054e-07, + "loss": 0.8297, + "step": 48453 + }, + { + "epoch": 8.81, + "learning_rate": 3.5689843465598833e-07, + "loss": 0.005, + "step": 48454 + }, + { + "epoch": 8.81, + "learning_rate": 3.568438296323262e-07, + "loss": 0.7594, + "step": 48455 + }, + { + "epoch": 8.81, + "learning_rate": 3.56789224608664e-07, + "loss": 0.0109, + "step": 48456 + }, + { + "epoch": 8.81, + "learning_rate": 3.5673461958500184e-07, + "loss": 0.3232, + "step": 48457 + }, + { + "epoch": 8.81, + "learning_rate": 3.566800145613397e-07, + "loss": 0.1564, + "step": 48458 + }, + { + "epoch": 8.81, + "learning_rate": 3.5662540953767746e-07, + "loss": 0.1956, + "step": 48459 + }, + { + "epoch": 8.81, + "learning_rate": 3.565708045140153e-07, + "loss": 1.0485, + "step": 48460 + }, + { + "epoch": 8.81, + "learning_rate": 3.565161994903531e-07, + "loss": 0.6298, + "step": 48461 + }, + { + "epoch": 8.81, + "learning_rate": 3.564615944666909e-07, + "loss": 0.2531, + "step": 48462 + }, + { + "epoch": 8.81, + "learning_rate": 3.5640698944302876e-07, + "loss": 0.2334, + "step": 48463 + }, + { + "epoch": 8.81, + "learning_rate": 3.563523844193666e-07, + "loss": 0.1454, + "step": 48464 + }, + { + "epoch": 8.81, + "learning_rate": 3.5629777939570444e-07, + "loss": 1.4366, + "step": 48465 + }, + { + "epoch": 8.81, + "learning_rate": 3.562431743720422e-07, + "loss": 2.1839, + "step": 48466 + }, + { + "epoch": 8.81, + "learning_rate": 3.5618856934838006e-07, + "loss": 1.3686, + "step": 48467 + }, + { + "epoch": 8.81, + "learning_rate": 3.561339643247179e-07, + "loss": 0.0129, + "step": 48468 + }, + { + "epoch": 8.81, + "learning_rate": 3.560793593010557e-07, + "loss": 1.9778, + "step": 48469 + }, + { + "epoch": 8.81, + "learning_rate": 3.5602475427739357e-07, + "loss": 0.0125, + "step": 48470 + }, + { + "epoch": 8.81, + "learning_rate": 3.5597014925373135e-07, + "loss": 0.8181, + "step": 48471 + }, + { + "epoch": 8.81, + "learning_rate": 3.559155442300692e-07, + "loss": 0.0045, + "step": 48472 + }, + { + "epoch": 8.81, + "learning_rate": 3.55860939206407e-07, + "loss": 1.2372, + "step": 48473 + }, + { + "epoch": 8.82, + "learning_rate": 3.558063341827448e-07, + "loss": 0.0045, + "step": 48474 + }, + { + "epoch": 8.82, + "learning_rate": 3.5575172915908265e-07, + "loss": 0.2834, + "step": 48475 + }, + { + "epoch": 8.82, + "learning_rate": 3.5569712413542044e-07, + "loss": 3.5645, + "step": 48476 + }, + { + "epoch": 8.82, + "learning_rate": 3.5564251911175833e-07, + "loss": 1.7024, + "step": 48477 + }, + { + "epoch": 8.82, + "learning_rate": 3.555879140880961e-07, + "loss": 0.3823, + "step": 48478 + }, + { + "epoch": 8.82, + "learning_rate": 3.5553330906443395e-07, + "loss": 0.0437, + "step": 48479 + }, + { + "epoch": 8.82, + "learning_rate": 3.554787040407718e-07, + "loss": 0.005, + "step": 48480 + }, + { + "epoch": 8.82, + "learning_rate": 3.5542409901710957e-07, + "loss": 0.1679, + "step": 48481 + }, + { + "epoch": 8.82, + "learning_rate": 3.553694939934474e-07, + "loss": 0.0077, + "step": 48482 + }, + { + "epoch": 8.82, + "learning_rate": 3.553148889697852e-07, + "loss": 1.1831, + "step": 48483 + }, + { + "epoch": 8.82, + "learning_rate": 3.5526028394612303e-07, + "loss": 1.0967, + "step": 48484 + }, + { + "epoch": 8.82, + "learning_rate": 3.5520567892246087e-07, + "loss": 0.4951, + "step": 48485 + }, + { + "epoch": 8.82, + "learning_rate": 3.551510738987987e-07, + "loss": 0.036, + "step": 48486 + }, + { + "epoch": 8.82, + "learning_rate": 3.5509646887513654e-07, + "loss": 1.0511, + "step": 48487 + }, + { + "epoch": 8.82, + "learning_rate": 3.5504186385147433e-07, + "loss": 1.0223, + "step": 48488 + }, + { + "epoch": 8.82, + "learning_rate": 3.5498725882781217e-07, + "loss": 0.6462, + "step": 48489 + }, + { + "epoch": 8.82, + "learning_rate": 3.5493265380415e-07, + "loss": 1.8, + "step": 48490 + }, + { + "epoch": 8.82, + "learning_rate": 3.548780487804878e-07, + "loss": 0.0895, + "step": 48491 + }, + { + "epoch": 8.82, + "learning_rate": 3.548234437568257e-07, + "loss": 0.2022, + "step": 48492 + }, + { + "epoch": 8.82, + "learning_rate": 3.5476883873316346e-07, + "loss": 0.035, + "step": 48493 + }, + { + "epoch": 8.82, + "learning_rate": 3.547142337095013e-07, + "loss": 0.0142, + "step": 48494 + }, + { + "epoch": 8.82, + "learning_rate": 3.546596286858391e-07, + "loss": 2.9591, + "step": 48495 + }, + { + "epoch": 8.82, + "learning_rate": 3.546050236621769e-07, + "loss": 0.0988, + "step": 48496 + }, + { + "epoch": 8.82, + "learning_rate": 3.5455041863851476e-07, + "loss": 2.0995, + "step": 48497 + }, + { + "epoch": 8.82, + "learning_rate": 3.5449581361485254e-07, + "loss": 0.2357, + "step": 48498 + }, + { + "epoch": 8.82, + "learning_rate": 3.5444120859119043e-07, + "loss": 0.0388, + "step": 48499 + }, + { + "epoch": 8.82, + "learning_rate": 3.543866035675282e-07, + "loss": 0.0095, + "step": 48500 + }, + { + "epoch": 8.82, + "learning_rate": 3.5433199854386606e-07, + "loss": 0.2413, + "step": 48501 + }, + { + "epoch": 8.82, + "learning_rate": 3.542773935202039e-07, + "loss": 0.0147, + "step": 48502 + }, + { + "epoch": 8.82, + "learning_rate": 3.542227884965417e-07, + "loss": 0.0175, + "step": 48503 + }, + { + "epoch": 8.82, + "learning_rate": 3.541681834728795e-07, + "loss": 0.1581, + "step": 48504 + }, + { + "epoch": 8.82, + "learning_rate": 3.541135784492173e-07, + "loss": 0.9023, + "step": 48505 + }, + { + "epoch": 8.82, + "learning_rate": 3.5405897342555514e-07, + "loss": 0.0103, + "step": 48506 + }, + { + "epoch": 8.82, + "learning_rate": 3.54004368401893e-07, + "loss": 0.2763, + "step": 48507 + }, + { + "epoch": 8.82, + "learning_rate": 3.539497633782308e-07, + "loss": 0.1423, + "step": 48508 + }, + { + "epoch": 8.82, + "learning_rate": 3.5389515835456865e-07, + "loss": 0.8076, + "step": 48509 + }, + { + "epoch": 8.82, + "learning_rate": 3.5384055333090644e-07, + "loss": 0.277, + "step": 48510 + }, + { + "epoch": 8.82, + "learning_rate": 3.5378594830724427e-07, + "loss": 1.4033, + "step": 48511 + }, + { + "epoch": 8.82, + "learning_rate": 3.537313432835821e-07, + "loss": 0.0277, + "step": 48512 + }, + { + "epoch": 8.82, + "learning_rate": 3.536767382599199e-07, + "loss": 0.3613, + "step": 48513 + }, + { + "epoch": 8.82, + "learning_rate": 3.536221332362578e-07, + "loss": 0.1139, + "step": 48514 + }, + { + "epoch": 8.82, + "learning_rate": 3.5356752821259557e-07, + "loss": 0.0041, + "step": 48515 + }, + { + "epoch": 8.82, + "learning_rate": 3.535129231889334e-07, + "loss": 1.1191, + "step": 48516 + }, + { + "epoch": 8.82, + "learning_rate": 3.534583181652712e-07, + "loss": 0.4007, + "step": 48517 + }, + { + "epoch": 8.82, + "learning_rate": 3.5340371314160903e-07, + "loss": 2.0047, + "step": 48518 + }, + { + "epoch": 8.82, + "learning_rate": 3.5334910811794687e-07, + "loss": 0.2712, + "step": 48519 + }, + { + "epoch": 8.82, + "learning_rate": 3.5329450309428465e-07, + "loss": 0.016, + "step": 48520 + }, + { + "epoch": 8.82, + "learning_rate": 3.532398980706225e-07, + "loss": 0.0121, + "step": 48521 + }, + { + "epoch": 8.82, + "learning_rate": 3.5318529304696033e-07, + "loss": 0.4532, + "step": 48522 + }, + { + "epoch": 8.82, + "learning_rate": 3.5313068802329816e-07, + "loss": 0.0302, + "step": 48523 + }, + { + "epoch": 8.82, + "learning_rate": 3.53076082999636e-07, + "loss": 0.1752, + "step": 48524 + }, + { + "epoch": 8.82, + "learning_rate": 3.530214779759738e-07, + "loss": 1.2076, + "step": 48525 + }, + { + "epoch": 8.82, + "learning_rate": 3.529668729523116e-07, + "loss": 0.0695, + "step": 48526 + }, + { + "epoch": 8.82, + "learning_rate": 3.529122679286494e-07, + "loss": 1.8545, + "step": 48527 + }, + { + "epoch": 8.82, + "learning_rate": 3.5285766290498725e-07, + "loss": 0.0204, + "step": 48528 + }, + { + "epoch": 8.83, + "learning_rate": 3.528030578813251e-07, + "loss": 0.8015, + "step": 48529 + }, + { + "epoch": 8.83, + "learning_rate": 3.527484528576629e-07, + "loss": 0.1368, + "step": 48530 + }, + { + "epoch": 8.83, + "learning_rate": 3.5269384783400076e-07, + "loss": 1.1742, + "step": 48531 + }, + { + "epoch": 8.83, + "learning_rate": 3.5263924281033854e-07, + "loss": 0.0042, + "step": 48532 + }, + { + "epoch": 8.83, + "learning_rate": 3.525846377866764e-07, + "loss": 0.2002, + "step": 48533 + }, + { + "epoch": 8.83, + "learning_rate": 3.525300327630142e-07, + "loss": 0.9458, + "step": 48534 + }, + { + "epoch": 8.83, + "learning_rate": 3.52475427739352e-07, + "loss": 1.4395, + "step": 48535 + }, + { + "epoch": 8.83, + "learning_rate": 3.524208227156899e-07, + "loss": 0.8186, + "step": 48536 + }, + { + "epoch": 8.83, + "learning_rate": 3.523662176920277e-07, + "loss": 0.3259, + "step": 48537 + }, + { + "epoch": 8.83, + "learning_rate": 3.523116126683655e-07, + "loss": 1.0674, + "step": 48538 + }, + { + "epoch": 8.83, + "learning_rate": 3.522570076447033e-07, + "loss": 0.725, + "step": 48539 + }, + { + "epoch": 8.83, + "learning_rate": 3.5220240262104114e-07, + "loss": 0.0039, + "step": 48540 + }, + { + "epoch": 8.83, + "learning_rate": 3.52147797597379e-07, + "loss": 0.632, + "step": 48541 + }, + { + "epoch": 8.83, + "learning_rate": 3.5209319257371676e-07, + "loss": 2.068, + "step": 48542 + }, + { + "epoch": 8.83, + "learning_rate": 3.520385875500546e-07, + "loss": 0.2359, + "step": 48543 + }, + { + "epoch": 8.83, + "learning_rate": 3.5198398252639243e-07, + "loss": 1.4474, + "step": 48544 + }, + { + "epoch": 8.83, + "learning_rate": 3.5192937750273027e-07, + "loss": 0.1699, + "step": 48545 + }, + { + "epoch": 8.83, + "learning_rate": 3.518747724790681e-07, + "loss": 0.0796, + "step": 48546 + }, + { + "epoch": 8.83, + "learning_rate": 3.518201674554059e-07, + "loss": 0.1544, + "step": 48547 + }, + { + "epoch": 8.83, + "learning_rate": 3.5176556243174373e-07, + "loss": 0.1951, + "step": 48548 + }, + { + "epoch": 8.83, + "learning_rate": 3.517109574080815e-07, + "loss": 0.1506, + "step": 48549 + }, + { + "epoch": 8.83, + "learning_rate": 3.5165635238441935e-07, + "loss": 0.0074, + "step": 48550 + }, + { + "epoch": 8.83, + "learning_rate": 3.516017473607572e-07, + "loss": 0.8205, + "step": 48551 + }, + { + "epoch": 8.83, + "learning_rate": 3.5154714233709503e-07, + "loss": 0.004, + "step": 48552 + }, + { + "epoch": 8.83, + "learning_rate": 3.5149253731343287e-07, + "loss": 0.1944, + "step": 48553 + }, + { + "epoch": 8.83, + "learning_rate": 3.5143793228977065e-07, + "loss": 0.0225, + "step": 48554 + }, + { + "epoch": 8.83, + "learning_rate": 3.513833272661085e-07, + "loss": 0.0741, + "step": 48555 + }, + { + "epoch": 8.83, + "learning_rate": 3.513287222424463e-07, + "loss": 3.3603, + "step": 48556 + }, + { + "epoch": 8.83, + "learning_rate": 3.512741172187841e-07, + "loss": 0.113, + "step": 48557 + }, + { + "epoch": 8.83, + "learning_rate": 3.51219512195122e-07, + "loss": 0.0109, + "step": 48558 + }, + { + "epoch": 8.83, + "learning_rate": 3.511649071714598e-07, + "loss": 0.1092, + "step": 48559 + }, + { + "epoch": 8.83, + "learning_rate": 3.511103021477976e-07, + "loss": 2.4087, + "step": 48560 + }, + { + "epoch": 8.83, + "learning_rate": 3.510556971241354e-07, + "loss": 2.4546, + "step": 48561 + }, + { + "epoch": 8.83, + "learning_rate": 3.5100109210047325e-07, + "loss": 0.0995, + "step": 48562 + }, + { + "epoch": 8.83, + "learning_rate": 3.509464870768111e-07, + "loss": 0.8678, + "step": 48563 + }, + { + "epoch": 8.83, + "learning_rate": 3.5089188205314887e-07, + "loss": 0.2843, + "step": 48564 + }, + { + "epoch": 8.83, + "learning_rate": 3.508372770294867e-07, + "loss": 0.2204, + "step": 48565 + }, + { + "epoch": 8.83, + "learning_rate": 3.5078267200582454e-07, + "loss": 1.5468, + "step": 48566 + }, + { + "epoch": 8.83, + "learning_rate": 3.507280669821624e-07, + "loss": 0.0058, + "step": 48567 + }, + { + "epoch": 8.83, + "learning_rate": 3.506734619585002e-07, + "loss": 0.1299, + "step": 48568 + }, + { + "epoch": 8.83, + "learning_rate": 3.50618856934838e-07, + "loss": 1.5702, + "step": 48569 + }, + { + "epoch": 8.83, + "learning_rate": 3.5056425191117584e-07, + "loss": 0.1523, + "step": 48570 + }, + { + "epoch": 8.83, + "learning_rate": 3.505096468875136e-07, + "loss": 0.136, + "step": 48571 + }, + { + "epoch": 8.83, + "learning_rate": 3.5045504186385146e-07, + "loss": 0.9503, + "step": 48572 + }, + { + "epoch": 8.83, + "learning_rate": 3.504004368401893e-07, + "loss": 1.2187, + "step": 48573 + }, + { + "epoch": 8.83, + "learning_rate": 3.5034583181652714e-07, + "loss": 0.018, + "step": 48574 + }, + { + "epoch": 8.83, + "learning_rate": 3.50291226792865e-07, + "loss": 2.2107, + "step": 48575 + }, + { + "epoch": 8.83, + "learning_rate": 3.5023662176920276e-07, + "loss": 1.0706, + "step": 48576 + }, + { + "epoch": 8.83, + "learning_rate": 3.501820167455406e-07, + "loss": 0.1941, + "step": 48577 + }, + { + "epoch": 8.83, + "learning_rate": 3.5012741172187843e-07, + "loss": 0.6367, + "step": 48578 + }, + { + "epoch": 8.83, + "learning_rate": 3.500728066982162e-07, + "loss": 0.0137, + "step": 48579 + }, + { + "epoch": 8.83, + "learning_rate": 3.5001820167455406e-07, + "loss": 0.099, + "step": 48580 + }, + { + "epoch": 8.83, + "learning_rate": 3.499635966508919e-07, + "loss": 0.1518, + "step": 48581 + }, + { + "epoch": 8.83, + "learning_rate": 3.4990899162722973e-07, + "loss": 0.2965, + "step": 48582 + }, + { + "epoch": 8.83, + "learning_rate": 3.498543866035675e-07, + "loss": 1.4513, + "step": 48583 + }, + { + "epoch": 8.84, + "learning_rate": 3.4979978157990535e-07, + "loss": 1.1876, + "step": 48584 + }, + { + "epoch": 8.84, + "learning_rate": 3.497451765562432e-07, + "loss": 1.1997, + "step": 48585 + }, + { + "epoch": 8.84, + "learning_rate": 3.49690571532581e-07, + "loss": 0.1434, + "step": 48586 + }, + { + "epoch": 8.84, + "learning_rate": 3.496359665089188e-07, + "loss": 0.2009, + "step": 48587 + }, + { + "epoch": 8.84, + "learning_rate": 3.4958136148525665e-07, + "loss": 1.7059, + "step": 48588 + }, + { + "epoch": 8.84, + "learning_rate": 3.495267564615945e-07, + "loss": 0.1797, + "step": 48589 + }, + { + "epoch": 8.84, + "learning_rate": 3.494721514379323e-07, + "loss": 0.0104, + "step": 48590 + }, + { + "epoch": 8.84, + "learning_rate": 3.494175464142701e-07, + "loss": 1.1106, + "step": 48591 + }, + { + "epoch": 8.84, + "learning_rate": 3.4936294139060795e-07, + "loss": 0.4261, + "step": 48592 + }, + { + "epoch": 8.84, + "learning_rate": 3.4930833636694573e-07, + "loss": 0.0082, + "step": 48593 + }, + { + "epoch": 8.84, + "learning_rate": 3.4925373134328357e-07, + "loss": 0.036, + "step": 48594 + }, + { + "epoch": 8.84, + "learning_rate": 3.4919912631962146e-07, + "loss": 0.0081, + "step": 48595 + }, + { + "epoch": 8.84, + "learning_rate": 3.4914452129595924e-07, + "loss": 0.1024, + "step": 48596 + }, + { + "epoch": 8.84, + "learning_rate": 3.490899162722971e-07, + "loss": 1.3106, + "step": 48597 + }, + { + "epoch": 8.84, + "learning_rate": 3.4903531124863487e-07, + "loss": 0.3369, + "step": 48598 + }, + { + "epoch": 8.84, + "learning_rate": 3.489807062249727e-07, + "loss": 0.3017, + "step": 48599 + }, + { + "epoch": 8.84, + "learning_rate": 3.4892610120131054e-07, + "loss": 1.2602, + "step": 48600 + }, + { + "epoch": 8.84, + "learning_rate": 3.488714961776483e-07, + "loss": 0.0077, + "step": 48601 + }, + { + "epoch": 8.84, + "learning_rate": 3.4881689115398616e-07, + "loss": 0.0381, + "step": 48602 + }, + { + "epoch": 8.84, + "learning_rate": 3.48762286130324e-07, + "loss": 1.1839, + "step": 48603 + }, + { + "epoch": 8.84, + "learning_rate": 3.4870768110666184e-07, + "loss": 1.9329, + "step": 48604 + }, + { + "epoch": 8.84, + "learning_rate": 3.486530760829997e-07, + "loss": 0.117, + "step": 48605 + }, + { + "epoch": 8.84, + "learning_rate": 3.4859847105933746e-07, + "loss": 1.6408, + "step": 48606 + }, + { + "epoch": 8.84, + "learning_rate": 3.485438660356753e-07, + "loss": 1.4281, + "step": 48607 + }, + { + "epoch": 8.84, + "learning_rate": 3.484892610120131e-07, + "loss": 0.2407, + "step": 48608 + }, + { + "epoch": 8.84, + "learning_rate": 3.484346559883509e-07, + "loss": 0.1501, + "step": 48609 + }, + { + "epoch": 8.84, + "learning_rate": 3.4838005096468876e-07, + "loss": 0.0058, + "step": 48610 + }, + { + "epoch": 8.84, + "learning_rate": 3.483254459410266e-07, + "loss": 0.8011, + "step": 48611 + }, + { + "epoch": 8.84, + "learning_rate": 3.4827084091736443e-07, + "loss": 0.0731, + "step": 48612 + }, + { + "epoch": 8.84, + "learning_rate": 3.482162358937022e-07, + "loss": 0.0671, + "step": 48613 + }, + { + "epoch": 8.84, + "learning_rate": 3.4816163087004006e-07, + "loss": 0.2604, + "step": 48614 + }, + { + "epoch": 8.84, + "learning_rate": 3.4810702584637784e-07, + "loss": 1.2807, + "step": 48615 + }, + { + "epoch": 8.84, + "learning_rate": 3.480524208227157e-07, + "loss": 0.0352, + "step": 48616 + }, + { + "epoch": 8.84, + "learning_rate": 3.4799781579905357e-07, + "loss": 3.2266, + "step": 48617 + }, + { + "epoch": 8.84, + "learning_rate": 3.4794321077539135e-07, + "loss": 1.372, + "step": 48618 + }, + { + "epoch": 8.84, + "learning_rate": 3.478886057517292e-07, + "loss": 0.0071, + "step": 48619 + }, + { + "epoch": 8.84, + "learning_rate": 3.47834000728067e-07, + "loss": 0.8682, + "step": 48620 + }, + { + "epoch": 8.84, + "learning_rate": 3.477793957044048e-07, + "loss": 0.1768, + "step": 48621 + }, + { + "epoch": 8.84, + "learning_rate": 3.4772479068074265e-07, + "loss": 0.4438, + "step": 48622 + }, + { + "epoch": 8.84, + "learning_rate": 3.4767018565708043e-07, + "loss": 0.1893, + "step": 48623 + }, + { + "epoch": 8.84, + "learning_rate": 3.4761558063341827e-07, + "loss": 0.0439, + "step": 48624 + }, + { + "epoch": 8.84, + "learning_rate": 3.475609756097561e-07, + "loss": 0.0416, + "step": 48625 + }, + { + "epoch": 8.84, + "learning_rate": 3.4750637058609395e-07, + "loss": 0.7194, + "step": 48626 + }, + { + "epoch": 8.84, + "learning_rate": 3.474517655624318e-07, + "loss": 0.0059, + "step": 48627 + }, + { + "epoch": 8.84, + "learning_rate": 3.4739716053876957e-07, + "loss": 0.1187, + "step": 48628 + }, + { + "epoch": 8.84, + "learning_rate": 3.473425555151074e-07, + "loss": 0.0399, + "step": 48629 + }, + { + "epoch": 8.84, + "learning_rate": 3.472879504914452e-07, + "loss": 0.7146, + "step": 48630 + }, + { + "epoch": 8.84, + "learning_rate": 3.4723334546778303e-07, + "loss": 3.6299, + "step": 48631 + }, + { + "epoch": 8.84, + "learning_rate": 3.4717874044412087e-07, + "loss": 0.3742, + "step": 48632 + }, + { + "epoch": 8.84, + "learning_rate": 3.471241354204587e-07, + "loss": 4.5473, + "step": 48633 + }, + { + "epoch": 8.84, + "learning_rate": 3.4706953039679654e-07, + "loss": 2.778, + "step": 48634 + }, + { + "epoch": 8.84, + "learning_rate": 3.470149253731343e-07, + "loss": 0.006, + "step": 48635 + }, + { + "epoch": 8.84, + "learning_rate": 3.4696032034947216e-07, + "loss": 0.0405, + "step": 48636 + }, + { + "epoch": 8.84, + "learning_rate": 3.4690571532581e-07, + "loss": 0.1041, + "step": 48637 + }, + { + "epoch": 8.84, + "learning_rate": 3.468511103021478e-07, + "loss": 0.3282, + "step": 48638 + }, + { + "epoch": 8.85, + "learning_rate": 3.467965052784856e-07, + "loss": 0.3382, + "step": 48639 + }, + { + "epoch": 8.85, + "learning_rate": 3.4674190025482346e-07, + "loss": 0.255, + "step": 48640 + }, + { + "epoch": 8.85, + "learning_rate": 3.466872952311613e-07, + "loss": 0.4128, + "step": 48641 + }, + { + "epoch": 8.85, + "learning_rate": 3.466326902074991e-07, + "loss": 2.4393, + "step": 48642 + }, + { + "epoch": 8.85, + "learning_rate": 3.465780851838369e-07, + "loss": 0.4305, + "step": 48643 + }, + { + "epoch": 8.85, + "learning_rate": 3.4652348016017476e-07, + "loss": 0.1104, + "step": 48644 + }, + { + "epoch": 8.85, + "learning_rate": 3.4646887513651254e-07, + "loss": 0.1388, + "step": 48645 + }, + { + "epoch": 8.85, + "learning_rate": 3.464142701128504e-07, + "loss": 0.0543, + "step": 48646 + }, + { + "epoch": 8.85, + "learning_rate": 3.463596650891882e-07, + "loss": 1.1615, + "step": 48647 + }, + { + "epoch": 8.85, + "learning_rate": 3.4630506006552605e-07, + "loss": 0.3046, + "step": 48648 + }, + { + "epoch": 8.85, + "learning_rate": 3.462504550418639e-07, + "loss": 0.0923, + "step": 48649 + }, + { + "epoch": 8.85, + "learning_rate": 3.461958500182017e-07, + "loss": 0.062, + "step": 48650 + }, + { + "epoch": 8.85, + "learning_rate": 3.461412449945395e-07, + "loss": 0.053, + "step": 48651 + }, + { + "epoch": 8.85, + "learning_rate": 3.460866399708773e-07, + "loss": 0.7536, + "step": 48652 + }, + { + "epoch": 8.85, + "learning_rate": 3.4603203494721514e-07, + "loss": 0.0559, + "step": 48653 + }, + { + "epoch": 8.85, + "learning_rate": 3.4597742992355297e-07, + "loss": 0.3197, + "step": 48654 + }, + { + "epoch": 8.85, + "learning_rate": 3.459228248998908e-07, + "loss": 0.5173, + "step": 48655 + }, + { + "epoch": 8.85, + "learning_rate": 3.4586821987622865e-07, + "loss": 1.47, + "step": 48656 + }, + { + "epoch": 8.85, + "learning_rate": 3.4581361485256643e-07, + "loss": 0.619, + "step": 48657 + }, + { + "epoch": 8.85, + "learning_rate": 3.4575900982890427e-07, + "loss": 0.0182, + "step": 48658 + }, + { + "epoch": 8.85, + "learning_rate": 3.457044048052421e-07, + "loss": 2.8074, + "step": 48659 + }, + { + "epoch": 8.85, + "learning_rate": 3.456497997815799e-07, + "loss": 0.2909, + "step": 48660 + }, + { + "epoch": 8.85, + "learning_rate": 3.4559519475791773e-07, + "loss": 0.0031, + "step": 48661 + }, + { + "epoch": 8.85, + "learning_rate": 3.4554058973425557e-07, + "loss": 0.194, + "step": 48662 + }, + { + "epoch": 8.85, + "learning_rate": 3.454859847105934e-07, + "loss": 0.4215, + "step": 48663 + }, + { + "epoch": 8.85, + "learning_rate": 3.454313796869312e-07, + "loss": 0.0357, + "step": 48664 + }, + { + "epoch": 8.85, + "learning_rate": 3.4537677466326903e-07, + "loss": 0.0053, + "step": 48665 + }, + { + "epoch": 8.85, + "learning_rate": 3.4532216963960686e-07, + "loss": 0.8103, + "step": 48666 + }, + { + "epoch": 8.85, + "learning_rate": 3.4526756461594465e-07, + "loss": 0.1553, + "step": 48667 + }, + { + "epoch": 8.85, + "learning_rate": 3.452129595922825e-07, + "loss": 0.0126, + "step": 48668 + }, + { + "epoch": 8.85, + "learning_rate": 3.451583545686203e-07, + "loss": 0.0146, + "step": 48669 + }, + { + "epoch": 8.85, + "learning_rate": 3.4510374954495816e-07, + "loss": 0.4843, + "step": 48670 + }, + { + "epoch": 8.85, + "learning_rate": 3.45049144521296e-07, + "loss": 0.0828, + "step": 48671 + }, + { + "epoch": 8.85, + "learning_rate": 3.449945394976338e-07, + "loss": 0.0088, + "step": 48672 + }, + { + "epoch": 8.85, + "learning_rate": 3.449399344739716e-07, + "loss": 0.1215, + "step": 48673 + }, + { + "epoch": 8.85, + "learning_rate": 3.448853294503094e-07, + "loss": 0.0036, + "step": 48674 + }, + { + "epoch": 8.85, + "learning_rate": 3.4483072442664724e-07, + "loss": 0.651, + "step": 48675 + }, + { + "epoch": 8.85, + "learning_rate": 3.447761194029851e-07, + "loss": 0.0197, + "step": 48676 + }, + { + "epoch": 8.85, + "learning_rate": 3.447215143793229e-07, + "loss": 0.2941, + "step": 48677 + }, + { + "epoch": 8.85, + "learning_rate": 3.4466690935566076e-07, + "loss": 0.1062, + "step": 48678 + }, + { + "epoch": 8.85, + "learning_rate": 3.4461230433199854e-07, + "loss": 0.1493, + "step": 48679 + }, + { + "epoch": 8.85, + "learning_rate": 3.445576993083364e-07, + "loss": 0.1639, + "step": 48680 + }, + { + "epoch": 8.85, + "learning_rate": 3.445030942846742e-07, + "loss": 0.5598, + "step": 48681 + }, + { + "epoch": 8.85, + "learning_rate": 3.44448489261012e-07, + "loss": 1.2884, + "step": 48682 + }, + { + "epoch": 8.85, + "learning_rate": 3.4439388423734984e-07, + "loss": 0.9663, + "step": 48683 + }, + { + "epoch": 8.85, + "learning_rate": 3.443392792136877e-07, + "loss": 0.3399, + "step": 48684 + }, + { + "epoch": 8.85, + "learning_rate": 3.442846741900255e-07, + "loss": 0.0133, + "step": 48685 + }, + { + "epoch": 8.85, + "learning_rate": 3.442300691663633e-07, + "loss": 0.0066, + "step": 48686 + }, + { + "epoch": 8.85, + "learning_rate": 3.4417546414270114e-07, + "loss": 2.0645, + "step": 48687 + }, + { + "epoch": 8.85, + "learning_rate": 3.4412085911903897e-07, + "loss": 0.0866, + "step": 48688 + }, + { + "epoch": 8.85, + "learning_rate": 3.4406625409537676e-07, + "loss": 0.1208, + "step": 48689 + }, + { + "epoch": 8.85, + "learning_rate": 3.440116490717146e-07, + "loss": 0.6113, + "step": 48690 + }, + { + "epoch": 8.85, + "learning_rate": 3.4395704404805243e-07, + "loss": 0.0117, + "step": 48691 + }, + { + "epoch": 8.85, + "learning_rate": 3.4390243902439027e-07, + "loss": 0.1992, + "step": 48692 + }, + { + "epoch": 8.85, + "learning_rate": 3.438478340007281e-07, + "loss": 0.2178, + "step": 48693 + }, + { + "epoch": 8.86, + "learning_rate": 3.437932289770659e-07, + "loss": 0.0103, + "step": 48694 + }, + { + "epoch": 8.86, + "learning_rate": 3.4373862395340373e-07, + "loss": 0.0499, + "step": 48695 + }, + { + "epoch": 8.86, + "learning_rate": 3.436840189297415e-07, + "loss": 1.0844, + "step": 48696 + }, + { + "epoch": 8.86, + "learning_rate": 3.4362941390607935e-07, + "loss": 0.161, + "step": 48697 + }, + { + "epoch": 8.86, + "learning_rate": 3.435748088824172e-07, + "loss": 0.1142, + "step": 48698 + }, + { + "epoch": 8.86, + "learning_rate": 3.43520203858755e-07, + "loss": 0.7731, + "step": 48699 + }, + { + "epoch": 8.86, + "learning_rate": 3.4346559883509286e-07, + "loss": 0.7951, + "step": 48700 + }, + { + "epoch": 8.86, + "learning_rate": 3.4341099381143065e-07, + "loss": 0.0035, + "step": 48701 + }, + { + "epoch": 8.86, + "learning_rate": 3.433563887877685e-07, + "loss": 0.3457, + "step": 48702 + }, + { + "epoch": 8.86, + "learning_rate": 3.433017837641063e-07, + "loss": 2.5084, + "step": 48703 + }, + { + "epoch": 8.86, + "learning_rate": 3.432471787404441e-07, + "loss": 0.2545, + "step": 48704 + }, + { + "epoch": 8.86, + "learning_rate": 3.4319257371678195e-07, + "loss": 0.2213, + "step": 48705 + }, + { + "epoch": 8.86, + "learning_rate": 3.431379686931198e-07, + "loss": 0.2552, + "step": 48706 + }, + { + "epoch": 8.86, + "learning_rate": 3.430833636694576e-07, + "loss": 0.0031, + "step": 48707 + }, + { + "epoch": 8.86, + "learning_rate": 3.430287586457954e-07, + "loss": 0.1367, + "step": 48708 + }, + { + "epoch": 8.86, + "learning_rate": 3.4297415362213324e-07, + "loss": 0.4526, + "step": 48709 + }, + { + "epoch": 8.86, + "learning_rate": 3.429195485984711e-07, + "loss": 0.0939, + "step": 48710 + }, + { + "epoch": 8.86, + "learning_rate": 3.4286494357480886e-07, + "loss": 0.3513, + "step": 48711 + }, + { + "epoch": 8.86, + "learning_rate": 3.428103385511467e-07, + "loss": 0.0037, + "step": 48712 + }, + { + "epoch": 8.86, + "learning_rate": 3.4275573352748454e-07, + "loss": 0.1022, + "step": 48713 + }, + { + "epoch": 8.86, + "learning_rate": 3.427011285038224e-07, + "loss": 0.4472, + "step": 48714 + }, + { + "epoch": 8.86, + "learning_rate": 3.426465234801602e-07, + "loss": 0.2496, + "step": 48715 + }, + { + "epoch": 8.86, + "learning_rate": 3.42591918456498e-07, + "loss": 0.3071, + "step": 48716 + }, + { + "epoch": 8.86, + "learning_rate": 3.4253731343283584e-07, + "loss": 0.1901, + "step": 48717 + }, + { + "epoch": 8.86, + "learning_rate": 3.424827084091736e-07, + "loss": 0.054, + "step": 48718 + }, + { + "epoch": 8.86, + "learning_rate": 3.4242810338551146e-07, + "loss": 0.1049, + "step": 48719 + }, + { + "epoch": 8.86, + "learning_rate": 3.423734983618493e-07, + "loss": 1.55, + "step": 48720 + }, + { + "epoch": 8.86, + "learning_rate": 3.4231889333818713e-07, + "loss": 0.5287, + "step": 48721 + }, + { + "epoch": 8.86, + "learning_rate": 3.4226428831452497e-07, + "loss": 0.0454, + "step": 48722 + }, + { + "epoch": 8.86, + "learning_rate": 3.4220968329086276e-07, + "loss": 0.3442, + "step": 48723 + }, + { + "epoch": 8.86, + "learning_rate": 3.421550782672006e-07, + "loss": 0.4008, + "step": 48724 + }, + { + "epoch": 8.86, + "learning_rate": 3.4210047324353843e-07, + "loss": 0.3239, + "step": 48725 + }, + { + "epoch": 8.86, + "learning_rate": 3.420458682198762e-07, + "loss": 0.7986, + "step": 48726 + }, + { + "epoch": 8.86, + "learning_rate": 3.4199126319621405e-07, + "loss": 0.301, + "step": 48727 + }, + { + "epoch": 8.86, + "learning_rate": 3.419366581725519e-07, + "loss": 0.8434, + "step": 48728 + }, + { + "epoch": 8.86, + "learning_rate": 3.4188205314888973e-07, + "loss": 0.0446, + "step": 48729 + }, + { + "epoch": 8.86, + "learning_rate": 3.418274481252275e-07, + "loss": 0.3945, + "step": 48730 + }, + { + "epoch": 8.86, + "learning_rate": 3.4177284310156535e-07, + "loss": 0.0073, + "step": 48731 + }, + { + "epoch": 8.86, + "learning_rate": 3.417182380779032e-07, + "loss": 0.3822, + "step": 48732 + }, + { + "epoch": 8.86, + "learning_rate": 3.4166363305424097e-07, + "loss": 0.9325, + "step": 48733 + }, + { + "epoch": 8.86, + "learning_rate": 3.416090280305788e-07, + "loss": 0.3084, + "step": 48734 + }, + { + "epoch": 8.86, + "learning_rate": 3.4155442300691665e-07, + "loss": 0.8044, + "step": 48735 + }, + { + "epoch": 8.86, + "learning_rate": 3.414998179832545e-07, + "loss": 1.2499, + "step": 48736 + }, + { + "epoch": 8.86, + "learning_rate": 3.414452129595923e-07, + "loss": 0.7758, + "step": 48737 + }, + { + "epoch": 8.86, + "learning_rate": 3.413906079359301e-07, + "loss": 0.0063, + "step": 48738 + }, + { + "epoch": 8.86, + "learning_rate": 3.4133600291226794e-07, + "loss": 3.8947, + "step": 48739 + }, + { + "epoch": 8.86, + "learning_rate": 3.4128139788860573e-07, + "loss": 0.1303, + "step": 48740 + }, + { + "epoch": 8.86, + "learning_rate": 3.4122679286494357e-07, + "loss": 0.487, + "step": 48741 + }, + { + "epoch": 8.86, + "learning_rate": 3.411721878412814e-07, + "loss": 0.0127, + "step": 48742 + }, + { + "epoch": 8.86, + "learning_rate": 3.4111758281761924e-07, + "loss": 0.9664, + "step": 48743 + }, + { + "epoch": 8.86, + "learning_rate": 3.410629777939571e-07, + "loss": 0.1743, + "step": 48744 + }, + { + "epoch": 8.86, + "learning_rate": 3.4100837277029486e-07, + "loss": 0.5702, + "step": 48745 + }, + { + "epoch": 8.86, + "learning_rate": 3.409537677466327e-07, + "loss": 0.0097, + "step": 48746 + }, + { + "epoch": 8.86, + "learning_rate": 3.4089916272297054e-07, + "loss": 0.288, + "step": 48747 + }, + { + "epoch": 8.86, + "learning_rate": 3.408445576993083e-07, + "loss": 0.0953, + "step": 48748 + }, + { + "epoch": 8.87, + "learning_rate": 3.4078995267564616e-07, + "loss": 0.9092, + "step": 48749 + }, + { + "epoch": 8.87, + "learning_rate": 3.40735347651984e-07, + "loss": 0.6077, + "step": 48750 + }, + { + "epoch": 8.87, + "learning_rate": 3.4068074262832184e-07, + "loss": 0.0688, + "step": 48751 + }, + { + "epoch": 8.87, + "learning_rate": 3.406261376046596e-07, + "loss": 0.2599, + "step": 48752 + }, + { + "epoch": 8.87, + "learning_rate": 3.4057153258099746e-07, + "loss": 0.3749, + "step": 48753 + }, + { + "epoch": 8.87, + "learning_rate": 3.405169275573353e-07, + "loss": 0.0057, + "step": 48754 + }, + { + "epoch": 8.87, + "learning_rate": 3.404623225336731e-07, + "loss": 0.1361, + "step": 48755 + }, + { + "epoch": 8.87, + "learning_rate": 3.404077175100109e-07, + "loss": 0.3425, + "step": 48756 + }, + { + "epoch": 8.87, + "learning_rate": 3.4035311248634876e-07, + "loss": 0.0717, + "step": 48757 + }, + { + "epoch": 8.87, + "learning_rate": 3.402985074626866e-07, + "loss": 0.8826, + "step": 48758 + }, + { + "epoch": 8.87, + "learning_rate": 3.4024390243902443e-07, + "loss": 0.1726, + "step": 48759 + }, + { + "epoch": 8.87, + "learning_rate": 3.401892974153622e-07, + "loss": 0.0794, + "step": 48760 + }, + { + "epoch": 8.87, + "learning_rate": 3.4013469239170005e-07, + "loss": 0.0464, + "step": 48761 + }, + { + "epoch": 8.87, + "learning_rate": 3.4008008736803784e-07, + "loss": 1.1091, + "step": 48762 + }, + { + "epoch": 8.87, + "learning_rate": 3.400254823443757e-07, + "loss": 0.9317, + "step": 48763 + }, + { + "epoch": 8.87, + "learning_rate": 3.399708773207135e-07, + "loss": 0.0621, + "step": 48764 + }, + { + "epoch": 8.87, + "learning_rate": 3.3991627229705135e-07, + "loss": 0.1353, + "step": 48765 + }, + { + "epoch": 8.87, + "learning_rate": 3.398616672733892e-07, + "loss": 1.1226, + "step": 48766 + }, + { + "epoch": 8.87, + "learning_rate": 3.3980706224972697e-07, + "loss": 1.5539, + "step": 48767 + }, + { + "epoch": 8.87, + "learning_rate": 3.397524572260648e-07, + "loss": 0.6655, + "step": 48768 + }, + { + "epoch": 8.87, + "learning_rate": 3.3969785220240265e-07, + "loss": 0.5998, + "step": 48769 + }, + { + "epoch": 8.87, + "learning_rate": 3.3964324717874043e-07, + "loss": 0.0168, + "step": 48770 + }, + { + "epoch": 8.87, + "learning_rate": 3.3958864215507827e-07, + "loss": 1.1746, + "step": 48771 + }, + { + "epoch": 8.87, + "learning_rate": 3.395340371314161e-07, + "loss": 0.0063, + "step": 48772 + }, + { + "epoch": 8.87, + "learning_rate": 3.3947943210775394e-07, + "loss": 0.9233, + "step": 48773 + }, + { + "epoch": 8.87, + "learning_rate": 3.3942482708409173e-07, + "loss": 0.0123, + "step": 48774 + }, + { + "epoch": 8.87, + "learning_rate": 3.3937022206042957e-07, + "loss": 0.3925, + "step": 48775 + }, + { + "epoch": 8.87, + "learning_rate": 3.393156170367674e-07, + "loss": 1.5412, + "step": 48776 + }, + { + "epoch": 8.87, + "learning_rate": 3.392610120131052e-07, + "loss": 1.403, + "step": 48777 + }, + { + "epoch": 8.87, + "learning_rate": 3.39206406989443e-07, + "loss": 1.1092, + "step": 48778 + }, + { + "epoch": 8.87, + "learning_rate": 3.3915180196578086e-07, + "loss": 0.1233, + "step": 48779 + }, + { + "epoch": 8.87, + "learning_rate": 3.390971969421187e-07, + "loss": 0.0066, + "step": 48780 + }, + { + "epoch": 8.87, + "learning_rate": 3.3904259191845654e-07, + "loss": 0.0385, + "step": 48781 + }, + { + "epoch": 8.87, + "learning_rate": 3.389879868947943e-07, + "loss": 1.0209, + "step": 48782 + }, + { + "epoch": 8.87, + "learning_rate": 3.3893338187113216e-07, + "loss": 0.8305, + "step": 48783 + }, + { + "epoch": 8.87, + "learning_rate": 3.3887877684746994e-07, + "loss": 0.0472, + "step": 48784 + }, + { + "epoch": 8.87, + "learning_rate": 3.388241718238078e-07, + "loss": 0.048, + "step": 48785 + }, + { + "epoch": 8.87, + "learning_rate": 3.387695668001456e-07, + "loss": 0.0093, + "step": 48786 + }, + { + "epoch": 8.87, + "learning_rate": 3.3871496177648346e-07, + "loss": 0.1854, + "step": 48787 + }, + { + "epoch": 8.87, + "learning_rate": 3.386603567528213e-07, + "loss": 0.0078, + "step": 48788 + }, + { + "epoch": 8.87, + "learning_rate": 3.386057517291591e-07, + "loss": 1.4037, + "step": 48789 + }, + { + "epoch": 8.87, + "learning_rate": 3.385511467054969e-07, + "loss": 0.6382, + "step": 48790 + }, + { + "epoch": 8.87, + "learning_rate": 3.3849654168183475e-07, + "loss": 1.5818, + "step": 48791 + }, + { + "epoch": 8.87, + "learning_rate": 3.3844193665817254e-07, + "loss": 0.5387, + "step": 48792 + }, + { + "epoch": 8.87, + "learning_rate": 3.383873316345104e-07, + "loss": 0.6472, + "step": 48793 + }, + { + "epoch": 8.87, + "learning_rate": 3.3833272661084816e-07, + "loss": 0.4329, + "step": 48794 + }, + { + "epoch": 8.87, + "learning_rate": 3.3827812158718605e-07, + "loss": 1.2588, + "step": 48795 + }, + { + "epoch": 8.87, + "learning_rate": 3.382235165635239e-07, + "loss": 1.2916, + "step": 48796 + }, + { + "epoch": 8.87, + "learning_rate": 3.381689115398617e-07, + "loss": 0.0542, + "step": 48797 + }, + { + "epoch": 8.87, + "learning_rate": 3.381143065161995e-07, + "loss": 0.1374, + "step": 48798 + }, + { + "epoch": 8.87, + "learning_rate": 3.380597014925373e-07, + "loss": 0.0045, + "step": 48799 + }, + { + "epoch": 8.87, + "learning_rate": 3.3800509646887513e-07, + "loss": 0.0503, + "step": 48800 + }, + { + "epoch": 8.87, + "learning_rate": 3.3795049144521297e-07, + "loss": 0.012, + "step": 48801 + }, + { + "epoch": 8.87, + "learning_rate": 3.378958864215508e-07, + "loss": 1.4991, + "step": 48802 + }, + { + "epoch": 8.87, + "learning_rate": 3.3784128139788865e-07, + "loss": 1.0792, + "step": 48803 + }, + { + "epoch": 8.88, + "learning_rate": 3.3778667637422643e-07, + "loss": 0.534, + "step": 48804 + }, + { + "epoch": 8.88, + "learning_rate": 3.3773207135056427e-07, + "loss": 0.2138, + "step": 48805 + }, + { + "epoch": 8.88, + "learning_rate": 3.376774663269021e-07, + "loss": 0.0513, + "step": 48806 + }, + { + "epoch": 8.88, + "learning_rate": 3.376228613032399e-07, + "loss": 0.3358, + "step": 48807 + }, + { + "epoch": 8.88, + "learning_rate": 3.3756825627957773e-07, + "loss": 0.0566, + "step": 48808 + }, + { + "epoch": 8.88, + "learning_rate": 3.3751365125591557e-07, + "loss": 0.2308, + "step": 48809 + }, + { + "epoch": 8.88, + "learning_rate": 3.374590462322534e-07, + "loss": 0.125, + "step": 48810 + }, + { + "epoch": 8.88, + "learning_rate": 3.374044412085912e-07, + "loss": 0.6297, + "step": 48811 + }, + { + "epoch": 8.88, + "learning_rate": 3.37349836184929e-07, + "loss": 4.0615, + "step": 48812 + }, + { + "epoch": 8.88, + "learning_rate": 3.3729523116126686e-07, + "loss": 0.2392, + "step": 48813 + }, + { + "epoch": 8.88, + "learning_rate": 3.3724062613760465e-07, + "loss": 0.0548, + "step": 48814 + }, + { + "epoch": 8.88, + "learning_rate": 3.371860211139425e-07, + "loss": 0.2738, + "step": 48815 + }, + { + "epoch": 8.88, + "learning_rate": 3.3713141609028027e-07, + "loss": 0.0652, + "step": 48816 + }, + { + "epoch": 8.88, + "learning_rate": 3.3707681106661816e-07, + "loss": 0.0667, + "step": 48817 + }, + { + "epoch": 8.88, + "learning_rate": 3.37022206042956e-07, + "loss": 0.1555, + "step": 48818 + }, + { + "epoch": 8.88, + "learning_rate": 3.369676010192938e-07, + "loss": 1.8158, + "step": 48819 + }, + { + "epoch": 8.88, + "learning_rate": 3.369129959956316e-07, + "loss": 0.3064, + "step": 48820 + }, + { + "epoch": 8.88, + "learning_rate": 3.368583909719694e-07, + "loss": 0.0123, + "step": 48821 + }, + { + "epoch": 8.88, + "learning_rate": 3.3680378594830724e-07, + "loss": 0.0619, + "step": 48822 + }, + { + "epoch": 8.88, + "learning_rate": 3.367491809246451e-07, + "loss": 0.3946, + "step": 48823 + }, + { + "epoch": 8.88, + "learning_rate": 3.366945759009829e-07, + "loss": 0.5852, + "step": 48824 + }, + { + "epoch": 8.88, + "learning_rate": 3.3663997087732075e-07, + "loss": 1.5225, + "step": 48825 + }, + { + "epoch": 8.88, + "learning_rate": 3.3658536585365854e-07, + "loss": 0.062, + "step": 48826 + }, + { + "epoch": 8.88, + "learning_rate": 3.365307608299964e-07, + "loss": 0.9619, + "step": 48827 + }, + { + "epoch": 8.88, + "learning_rate": 3.364761558063342e-07, + "loss": 1.7142, + "step": 48828 + }, + { + "epoch": 8.88, + "learning_rate": 3.36421550782672e-07, + "loss": 0.0339, + "step": 48829 + }, + { + "epoch": 8.88, + "learning_rate": 3.3636694575900984e-07, + "loss": 1.0282, + "step": 48830 + }, + { + "epoch": 8.88, + "learning_rate": 3.363123407353476e-07, + "loss": 0.006, + "step": 48831 + }, + { + "epoch": 8.88, + "learning_rate": 3.362577357116855e-07, + "loss": 0.0867, + "step": 48832 + }, + { + "epoch": 8.88, + "learning_rate": 3.362031306880233e-07, + "loss": 1.6915, + "step": 48833 + }, + { + "epoch": 8.88, + "learning_rate": 3.3614852566436113e-07, + "loss": 0.0086, + "step": 48834 + }, + { + "epoch": 8.88, + "learning_rate": 3.3609392064069897e-07, + "loss": 1.8673, + "step": 48835 + }, + { + "epoch": 8.88, + "learning_rate": 3.3603931561703675e-07, + "loss": 0.6365, + "step": 48836 + }, + { + "epoch": 8.88, + "learning_rate": 3.359847105933746e-07, + "loss": 0.1683, + "step": 48837 + }, + { + "epoch": 8.88, + "learning_rate": 3.3593010556971243e-07, + "loss": 0.2034, + "step": 48838 + }, + { + "epoch": 8.88, + "learning_rate": 3.3587550054605027e-07, + "loss": 1.0129, + "step": 48839 + }, + { + "epoch": 8.88, + "learning_rate": 3.358208955223881e-07, + "loss": 0.0589, + "step": 48840 + }, + { + "epoch": 8.88, + "learning_rate": 3.357662904987259e-07, + "loss": 0.101, + "step": 48841 + }, + { + "epoch": 8.88, + "learning_rate": 3.3571168547506373e-07, + "loss": 2.3247, + "step": 48842 + }, + { + "epoch": 8.88, + "learning_rate": 3.356570804514015e-07, + "loss": 0.3413, + "step": 48843 + }, + { + "epoch": 8.88, + "learning_rate": 3.3560247542773935e-07, + "loss": 0.0245, + "step": 48844 + }, + { + "epoch": 8.88, + "learning_rate": 3.355478704040772e-07, + "loss": 0.7381, + "step": 48845 + }, + { + "epoch": 8.88, + "learning_rate": 3.35493265380415e-07, + "loss": 1.7545, + "step": 48846 + }, + { + "epoch": 8.88, + "learning_rate": 3.3543866035675286e-07, + "loss": 0.1312, + "step": 48847 + }, + { + "epoch": 8.88, + "learning_rate": 3.3538405533309065e-07, + "loss": 0.0132, + "step": 48848 + }, + { + "epoch": 8.88, + "learning_rate": 3.353294503094285e-07, + "loss": 0.7654, + "step": 48849 + }, + { + "epoch": 8.88, + "learning_rate": 3.352748452857663e-07, + "loss": 0.1062, + "step": 48850 + }, + { + "epoch": 8.88, + "learning_rate": 3.352202402621041e-07, + "loss": 0.0205, + "step": 48851 + }, + { + "epoch": 8.88, + "learning_rate": 3.3516563523844194e-07, + "loss": 0.0751, + "step": 48852 + }, + { + "epoch": 8.88, + "learning_rate": 3.3511103021477973e-07, + "loss": 0.0748, + "step": 48853 + }, + { + "epoch": 8.88, + "learning_rate": 3.350564251911176e-07, + "loss": 0.0047, + "step": 48854 + }, + { + "epoch": 8.88, + "learning_rate": 3.350018201674554e-07, + "loss": 0.0207, + "step": 48855 + }, + { + "epoch": 8.88, + "learning_rate": 3.3494721514379324e-07, + "loss": 0.2833, + "step": 48856 + }, + { + "epoch": 8.88, + "learning_rate": 3.348926101201311e-07, + "loss": 0.9083, + "step": 48857 + }, + { + "epoch": 8.88, + "learning_rate": 3.3483800509646886e-07, + "loss": 0.0505, + "step": 48858 + }, + { + "epoch": 8.89, + "learning_rate": 3.347834000728067e-07, + "loss": 0.0026, + "step": 48859 + }, + { + "epoch": 8.89, + "learning_rate": 3.3472879504914454e-07, + "loss": 1.7124, + "step": 48860 + }, + { + "epoch": 8.89, + "learning_rate": 3.346741900254824e-07, + "loss": 2.2208, + "step": 48861 + }, + { + "epoch": 8.89, + "learning_rate": 3.346195850018202e-07, + "loss": 4.1378, + "step": 48862 + }, + { + "epoch": 8.89, + "learning_rate": 3.34564979978158e-07, + "loss": 0.0079, + "step": 48863 + }, + { + "epoch": 8.89, + "learning_rate": 3.3451037495449583e-07, + "loss": 1.501, + "step": 48864 + }, + { + "epoch": 8.89, + "learning_rate": 3.344557699308336e-07, + "loss": 0.0175, + "step": 48865 + }, + { + "epoch": 8.89, + "learning_rate": 3.3440116490717146e-07, + "loss": 2.4017, + "step": 48866 + }, + { + "epoch": 8.89, + "learning_rate": 3.343465598835093e-07, + "loss": 1.0674, + "step": 48867 + }, + { + "epoch": 8.89, + "learning_rate": 3.3429195485984713e-07, + "loss": 1.2985, + "step": 48868 + }, + { + "epoch": 8.89, + "learning_rate": 3.3423734983618497e-07, + "loss": 0.973, + "step": 48869 + }, + { + "epoch": 8.89, + "learning_rate": 3.3418274481252275e-07, + "loss": 1.383, + "step": 48870 + }, + { + "epoch": 8.89, + "learning_rate": 3.341281397888606e-07, + "loss": 0.0881, + "step": 48871 + }, + { + "epoch": 8.89, + "learning_rate": 3.3407353476519843e-07, + "loss": 1.4852, + "step": 48872 + }, + { + "epoch": 8.89, + "learning_rate": 3.340189297415362e-07, + "loss": 0.0737, + "step": 48873 + }, + { + "epoch": 8.89, + "learning_rate": 3.3396432471787405e-07, + "loss": 0.2796, + "step": 48874 + }, + { + "epoch": 8.89, + "learning_rate": 3.3390971969421184e-07, + "loss": 0.0177, + "step": 48875 + }, + { + "epoch": 8.89, + "learning_rate": 3.338551146705497e-07, + "loss": 0.7016, + "step": 48876 + }, + { + "epoch": 8.89, + "learning_rate": 3.338005096468875e-07, + "loss": 2.1812, + "step": 48877 + }, + { + "epoch": 8.89, + "learning_rate": 3.3374590462322535e-07, + "loss": 0.8671, + "step": 48878 + }, + { + "epoch": 8.89, + "learning_rate": 3.336912995995632e-07, + "loss": 0.0947, + "step": 48879 + }, + { + "epoch": 8.89, + "learning_rate": 3.3363669457590097e-07, + "loss": 1.6761, + "step": 48880 + }, + { + "epoch": 8.89, + "learning_rate": 3.335820895522388e-07, + "loss": 0.1924, + "step": 48881 + }, + { + "epoch": 8.89, + "learning_rate": 3.3352748452857664e-07, + "loss": 3.9578, + "step": 48882 + }, + { + "epoch": 8.89, + "learning_rate": 3.334728795049145e-07, + "loss": 0.7755, + "step": 48883 + }, + { + "epoch": 8.89, + "learning_rate": 3.334182744812523e-07, + "loss": 0.3994, + "step": 48884 + }, + { + "epoch": 8.89, + "learning_rate": 3.333636694575901e-07, + "loss": 0.135, + "step": 48885 + }, + { + "epoch": 8.89, + "learning_rate": 3.3330906443392794e-07, + "loss": 0.1719, + "step": 48886 + }, + { + "epoch": 8.89, + "learning_rate": 3.3325445941026573e-07, + "loss": 0.1575, + "step": 48887 + }, + { + "epoch": 8.89, + "learning_rate": 3.3319985438660356e-07, + "loss": 0.5758, + "step": 48888 + }, + { + "epoch": 8.89, + "learning_rate": 3.331452493629414e-07, + "loss": 0.3965, + "step": 48889 + }, + { + "epoch": 8.89, + "learning_rate": 3.330906443392792e-07, + "loss": 2.0665, + "step": 48890 + }, + { + "epoch": 8.89, + "learning_rate": 3.330360393156171e-07, + "loss": 0.8281, + "step": 48891 + }, + { + "epoch": 8.89, + "learning_rate": 3.3298143429195486e-07, + "loss": 0.0354, + "step": 48892 + }, + { + "epoch": 8.89, + "learning_rate": 3.329268292682927e-07, + "loss": 0.2745, + "step": 48893 + }, + { + "epoch": 8.89, + "learning_rate": 3.3287222424463054e-07, + "loss": 0.2993, + "step": 48894 + }, + { + "epoch": 8.89, + "learning_rate": 3.328176192209683e-07, + "loss": 0.4654, + "step": 48895 + }, + { + "epoch": 8.89, + "learning_rate": 3.3276301419730616e-07, + "loss": 0.392, + "step": 48896 + }, + { + "epoch": 8.89, + "learning_rate": 3.3270840917364394e-07, + "loss": 1.2725, + "step": 48897 + }, + { + "epoch": 8.89, + "learning_rate": 3.3265380414998183e-07, + "loss": 0.3837, + "step": 48898 + }, + { + "epoch": 8.89, + "learning_rate": 3.325991991263196e-07, + "loss": 1.1886, + "step": 48899 + }, + { + "epoch": 8.89, + "learning_rate": 3.3254459410265746e-07, + "loss": 0.1386, + "step": 48900 + }, + { + "epoch": 8.89, + "learning_rate": 3.324899890789953e-07, + "loss": 0.0116, + "step": 48901 + }, + { + "epoch": 8.89, + "learning_rate": 3.324353840553331e-07, + "loss": 0.2209, + "step": 48902 + }, + { + "epoch": 8.89, + "learning_rate": 3.323807790316709e-07, + "loss": 0.081, + "step": 48903 + }, + { + "epoch": 8.89, + "learning_rate": 3.3232617400800875e-07, + "loss": 0.011, + "step": 48904 + }, + { + "epoch": 8.89, + "learning_rate": 3.322715689843466e-07, + "loss": 0.9867, + "step": 48905 + }, + { + "epoch": 8.89, + "learning_rate": 3.3221696396068443e-07, + "loss": 0.1348, + "step": 48906 + }, + { + "epoch": 8.89, + "learning_rate": 3.321623589370222e-07, + "loss": 0.0933, + "step": 48907 + }, + { + "epoch": 8.89, + "learning_rate": 3.3210775391336005e-07, + "loss": 0.2404, + "step": 48908 + }, + { + "epoch": 8.89, + "learning_rate": 3.3205314888969783e-07, + "loss": 0.5646, + "step": 48909 + }, + { + "epoch": 8.89, + "learning_rate": 3.3199854386603567e-07, + "loss": 0.0055, + "step": 48910 + }, + { + "epoch": 8.89, + "learning_rate": 3.319439388423735e-07, + "loss": 1.4714, + "step": 48911 + }, + { + "epoch": 8.89, + "learning_rate": 3.318893338187113e-07, + "loss": 0.0771, + "step": 48912 + }, + { + "epoch": 8.89, + "learning_rate": 3.318347287950492e-07, + "loss": 0.012, + "step": 48913 + }, + { + "epoch": 8.9, + "learning_rate": 3.3178012377138697e-07, + "loss": 0.5781, + "step": 48914 + }, + { + "epoch": 8.9, + "learning_rate": 3.317255187477248e-07, + "loss": 0.7658, + "step": 48915 + }, + { + "epoch": 8.9, + "learning_rate": 3.3167091372406264e-07, + "loss": 0.0439, + "step": 48916 + }, + { + "epoch": 8.9, + "learning_rate": 3.3161630870040043e-07, + "loss": 0.2034, + "step": 48917 + }, + { + "epoch": 8.9, + "learning_rate": 3.3156170367673827e-07, + "loss": 0.071, + "step": 48918 + }, + { + "epoch": 8.9, + "learning_rate": 3.3150709865307605e-07, + "loss": 1.9876, + "step": 48919 + }, + { + "epoch": 8.9, + "learning_rate": 3.3145249362941394e-07, + "loss": 0.6244, + "step": 48920 + }, + { + "epoch": 8.9, + "learning_rate": 3.313978886057517e-07, + "loss": 0.0075, + "step": 48921 + }, + { + "epoch": 8.9, + "learning_rate": 3.3134328358208956e-07, + "loss": 0.7038, + "step": 48922 + }, + { + "epoch": 8.9, + "learning_rate": 3.312886785584274e-07, + "loss": 2.7581, + "step": 48923 + }, + { + "epoch": 8.9, + "learning_rate": 3.312340735347652e-07, + "loss": 0.0879, + "step": 48924 + }, + { + "epoch": 8.9, + "learning_rate": 3.31179468511103e-07, + "loss": 1.4718, + "step": 48925 + }, + { + "epoch": 8.9, + "learning_rate": 3.3112486348744086e-07, + "loss": 0.1089, + "step": 48926 + }, + { + "epoch": 8.9, + "learning_rate": 3.310702584637787e-07, + "loss": 0.4984, + "step": 48927 + }, + { + "epoch": 8.9, + "learning_rate": 3.3101565344011654e-07, + "loss": 0.2206, + "step": 48928 + }, + { + "epoch": 8.9, + "learning_rate": 3.309610484164543e-07, + "loss": 0.067, + "step": 48929 + }, + { + "epoch": 8.9, + "learning_rate": 3.3090644339279216e-07, + "loss": 0.0876, + "step": 48930 + }, + { + "epoch": 8.9, + "learning_rate": 3.3085183836912994e-07, + "loss": 0.139, + "step": 48931 + }, + { + "epoch": 8.9, + "learning_rate": 3.307972333454678e-07, + "loss": 0.0372, + "step": 48932 + }, + { + "epoch": 8.9, + "learning_rate": 3.307426283218056e-07, + "loss": 0.0917, + "step": 48933 + }, + { + "epoch": 8.9, + "learning_rate": 3.306880232981434e-07, + "loss": 0.1956, + "step": 48934 + }, + { + "epoch": 8.9, + "learning_rate": 3.306334182744813e-07, + "loss": 2.1405, + "step": 48935 + }, + { + "epoch": 8.9, + "learning_rate": 3.305788132508191e-07, + "loss": 0.5591, + "step": 48936 + }, + { + "epoch": 8.9, + "learning_rate": 3.305242082271569e-07, + "loss": 0.7113, + "step": 48937 + }, + { + "epoch": 8.9, + "learning_rate": 3.3046960320349475e-07, + "loss": 0.0109, + "step": 48938 + }, + { + "epoch": 8.9, + "learning_rate": 3.3041499817983254e-07, + "loss": 2.592, + "step": 48939 + }, + { + "epoch": 8.9, + "learning_rate": 3.303603931561704e-07, + "loss": 0.7038, + "step": 48940 + }, + { + "epoch": 8.9, + "learning_rate": 3.3030578813250816e-07, + "loss": 3.2818, + "step": 48941 + }, + { + "epoch": 8.9, + "learning_rate": 3.3025118310884605e-07, + "loss": 3.4831, + "step": 48942 + }, + { + "epoch": 8.9, + "learning_rate": 3.3019657808518383e-07, + "loss": 0.2385, + "step": 48943 + }, + { + "epoch": 8.9, + "learning_rate": 3.3014197306152167e-07, + "loss": 1.7197, + "step": 48944 + }, + { + "epoch": 8.9, + "learning_rate": 3.300873680378595e-07, + "loss": 1.9565, + "step": 48945 + }, + { + "epoch": 8.9, + "learning_rate": 3.300327630141973e-07, + "loss": 1.3446, + "step": 48946 + }, + { + "epoch": 8.9, + "learning_rate": 3.2997815799053513e-07, + "loss": 0.2401, + "step": 48947 + }, + { + "epoch": 8.9, + "learning_rate": 3.2992355296687297e-07, + "loss": 0.0042, + "step": 48948 + }, + { + "epoch": 8.9, + "learning_rate": 3.2986894794321075e-07, + "loss": 1.5048, + "step": 48949 + }, + { + "epoch": 8.9, + "learning_rate": 3.2981434291954864e-07, + "loss": 0.0457, + "step": 48950 + }, + { + "epoch": 8.9, + "learning_rate": 3.2975973789588643e-07, + "loss": 0.8308, + "step": 48951 + }, + { + "epoch": 8.9, + "learning_rate": 3.2970513287222427e-07, + "loss": 0.0083, + "step": 48952 + }, + { + "epoch": 8.9, + "learning_rate": 3.2965052784856205e-07, + "loss": 0.0975, + "step": 48953 + }, + { + "epoch": 8.9, + "learning_rate": 3.295959228248999e-07, + "loss": 0.1697, + "step": 48954 + }, + { + "epoch": 8.9, + "learning_rate": 3.295413178012377e-07, + "loss": 0.0076, + "step": 48955 + }, + { + "epoch": 8.9, + "learning_rate": 3.294867127775755e-07, + "loss": 1.3513, + "step": 48956 + }, + { + "epoch": 8.9, + "learning_rate": 3.294321077539134e-07, + "loss": 2.4411, + "step": 48957 + }, + { + "epoch": 8.9, + "learning_rate": 3.293775027302512e-07, + "loss": 0.1061, + "step": 48958 + }, + { + "epoch": 8.9, + "learning_rate": 3.29322897706589e-07, + "loss": 0.3187, + "step": 48959 + }, + { + "epoch": 8.9, + "learning_rate": 3.2926829268292686e-07, + "loss": 1.7225, + "step": 48960 + }, + { + "epoch": 8.9, + "learning_rate": 3.2921368765926464e-07, + "loss": 0.2366, + "step": 48961 + }, + { + "epoch": 8.9, + "learning_rate": 3.291590826356025e-07, + "loss": 0.0064, + "step": 48962 + }, + { + "epoch": 8.9, + "learning_rate": 3.2910447761194027e-07, + "loss": 0.3202, + "step": 48963 + }, + { + "epoch": 8.9, + "learning_rate": 3.2904987258827816e-07, + "loss": 0.7912, + "step": 48964 + }, + { + "epoch": 8.9, + "learning_rate": 3.2899526756461594e-07, + "loss": 0.0989, + "step": 48965 + }, + { + "epoch": 8.9, + "learning_rate": 3.289406625409538e-07, + "loss": 4.1224, + "step": 48966 + }, + { + "epoch": 8.9, + "learning_rate": 3.288860575172916e-07, + "loss": 0.9687, + "step": 48967 + }, + { + "epoch": 8.9, + "learning_rate": 3.288314524936294e-07, + "loss": 0.0852, + "step": 48968 + }, + { + "epoch": 8.91, + "learning_rate": 3.2877684746996724e-07, + "loss": 1.0303, + "step": 48969 + }, + { + "epoch": 8.91, + "learning_rate": 3.287222424463051e-07, + "loss": 0.0584, + "step": 48970 + }, + { + "epoch": 8.91, + "learning_rate": 3.2866763742264286e-07, + "loss": 0.7613, + "step": 48971 + }, + { + "epoch": 8.91, + "learning_rate": 3.2861303239898075e-07, + "loss": 3.8207, + "step": 48972 + }, + { + "epoch": 8.91, + "learning_rate": 3.2855842737531854e-07, + "loss": 0.1392, + "step": 48973 + }, + { + "epoch": 8.91, + "learning_rate": 3.2850382235165637e-07, + "loss": 0.2043, + "step": 48974 + }, + { + "epoch": 8.91, + "learning_rate": 3.2844921732799416e-07, + "loss": 0.4453, + "step": 48975 + }, + { + "epoch": 8.91, + "learning_rate": 3.28394612304332e-07, + "loss": 0.0078, + "step": 48976 + }, + { + "epoch": 8.91, + "learning_rate": 3.2834000728066983e-07, + "loss": 0.1269, + "step": 48977 + }, + { + "epoch": 8.91, + "learning_rate": 3.282854022570076e-07, + "loss": 0.5006, + "step": 48978 + }, + { + "epoch": 8.91, + "learning_rate": 3.282307972333455e-07, + "loss": 2.7883, + "step": 48979 + }, + { + "epoch": 8.91, + "learning_rate": 3.281761922096833e-07, + "loss": 0.0183, + "step": 48980 + }, + { + "epoch": 8.91, + "learning_rate": 3.2812158718602113e-07, + "loss": 0.949, + "step": 48981 + }, + { + "epoch": 8.91, + "learning_rate": 3.2806698216235897e-07, + "loss": 1.5093, + "step": 48982 + }, + { + "epoch": 8.91, + "learning_rate": 3.2801237713869675e-07, + "loss": 0.1352, + "step": 48983 + }, + { + "epoch": 8.91, + "learning_rate": 3.279577721150346e-07, + "loss": 1.3192, + "step": 48984 + }, + { + "epoch": 8.91, + "learning_rate": 3.279031670913724e-07, + "loss": 0.2074, + "step": 48985 + }, + { + "epoch": 8.91, + "learning_rate": 3.278485620677102e-07, + "loss": 0.1664, + "step": 48986 + }, + { + "epoch": 8.91, + "learning_rate": 3.277939570440481e-07, + "loss": 0.0081, + "step": 48987 + }, + { + "epoch": 8.91, + "learning_rate": 3.277393520203859e-07, + "loss": 0.204, + "step": 48988 + }, + { + "epoch": 8.91, + "learning_rate": 3.276847469967237e-07, + "loss": 0.0021, + "step": 48989 + }, + { + "epoch": 8.91, + "learning_rate": 3.276301419730615e-07, + "loss": 0.0432, + "step": 48990 + }, + { + "epoch": 8.91, + "learning_rate": 3.2757553694939935e-07, + "loss": 1.0013, + "step": 48991 + }, + { + "epoch": 8.91, + "learning_rate": 3.275209319257372e-07, + "loss": 0.0098, + "step": 48992 + }, + { + "epoch": 8.91, + "learning_rate": 3.2746632690207497e-07, + "loss": 0.3883, + "step": 48993 + }, + { + "epoch": 8.91, + "learning_rate": 3.2741172187841286e-07, + "loss": 0.4502, + "step": 48994 + }, + { + "epoch": 8.91, + "learning_rate": 3.2735711685475064e-07, + "loss": 0.0527, + "step": 48995 + }, + { + "epoch": 8.91, + "learning_rate": 3.273025118310885e-07, + "loss": 0.8893, + "step": 48996 + }, + { + "epoch": 8.91, + "learning_rate": 3.272479068074263e-07, + "loss": 0.0739, + "step": 48997 + }, + { + "epoch": 8.91, + "learning_rate": 3.271933017837641e-07, + "loss": 0.3933, + "step": 48998 + }, + { + "epoch": 8.91, + "learning_rate": 3.2713869676010194e-07, + "loss": 0.0105, + "step": 48999 + }, + { + "epoch": 8.91, + "learning_rate": 3.270840917364397e-07, + "loss": 0.0085, + "step": 49000 + }, + { + "epoch": 8.91, + "learning_rate": 3.270294867127776e-07, + "loss": 0.3165, + "step": 49001 + }, + { + "epoch": 8.91, + "learning_rate": 3.269748816891154e-07, + "loss": 1.8015, + "step": 49002 + }, + { + "epoch": 8.91, + "learning_rate": 3.2692027666545324e-07, + "loss": 0.6788, + "step": 49003 + }, + { + "epoch": 8.91, + "learning_rate": 3.268656716417911e-07, + "loss": 0.0098, + "step": 49004 + }, + { + "epoch": 8.91, + "learning_rate": 3.2681106661812886e-07, + "loss": 0.3617, + "step": 49005 + }, + { + "epoch": 8.91, + "learning_rate": 3.267564615944667e-07, + "loss": 0.6946, + "step": 49006 + }, + { + "epoch": 8.91, + "learning_rate": 3.267018565708045e-07, + "loss": 0.2685, + "step": 49007 + }, + { + "epoch": 8.91, + "learning_rate": 3.266472515471423e-07, + "loss": 3.9564, + "step": 49008 + }, + { + "epoch": 8.91, + "learning_rate": 3.265926465234802e-07, + "loss": 0.5636, + "step": 49009 + }, + { + "epoch": 8.91, + "learning_rate": 3.26538041499818e-07, + "loss": 0.0043, + "step": 49010 + }, + { + "epoch": 8.91, + "learning_rate": 3.2648343647615583e-07, + "loss": 0.0113, + "step": 49011 + }, + { + "epoch": 8.91, + "learning_rate": 3.264288314524936e-07, + "loss": 0.2071, + "step": 49012 + }, + { + "epoch": 8.91, + "learning_rate": 3.2637422642883145e-07, + "loss": 0.0602, + "step": 49013 + }, + { + "epoch": 8.91, + "learning_rate": 3.263196214051693e-07, + "loss": 0.0454, + "step": 49014 + }, + { + "epoch": 8.91, + "learning_rate": 3.262650163815071e-07, + "loss": 0.0083, + "step": 49015 + }, + { + "epoch": 8.91, + "learning_rate": 3.2621041135784497e-07, + "loss": 0.0031, + "step": 49016 + }, + { + "epoch": 8.91, + "learning_rate": 3.2615580633418275e-07, + "loss": 0.3635, + "step": 49017 + }, + { + "epoch": 8.91, + "learning_rate": 3.261012013105206e-07, + "loss": 0.0875, + "step": 49018 + }, + { + "epoch": 8.91, + "learning_rate": 3.260465962868584e-07, + "loss": 0.0454, + "step": 49019 + }, + { + "epoch": 8.91, + "learning_rate": 3.259919912631962e-07, + "loss": 0.51, + "step": 49020 + }, + { + "epoch": 8.91, + "learning_rate": 3.2593738623953405e-07, + "loss": 0.086, + "step": 49021 + }, + { + "epoch": 8.91, + "learning_rate": 3.2588278121587183e-07, + "loss": 0.9645, + "step": 49022 + }, + { + "epoch": 8.91, + "learning_rate": 3.258281761922097e-07, + "loss": 0.0039, + "step": 49023 + }, + { + "epoch": 8.92, + "learning_rate": 3.257735711685475e-07, + "loss": 0.3096, + "step": 49024 + }, + { + "epoch": 8.92, + "learning_rate": 3.2571896614488535e-07, + "loss": 0.0381, + "step": 49025 + }, + { + "epoch": 8.92, + "learning_rate": 3.256643611212232e-07, + "loss": 1.1379, + "step": 49026 + }, + { + "epoch": 8.92, + "learning_rate": 3.2560975609756097e-07, + "loss": 0.394, + "step": 49027 + }, + { + "epoch": 8.92, + "learning_rate": 3.255551510738988e-07, + "loss": 0.0156, + "step": 49028 + }, + { + "epoch": 8.92, + "learning_rate": 3.2550054605023664e-07, + "loss": 0.3981, + "step": 49029 + }, + { + "epoch": 8.92, + "learning_rate": 3.2544594102657443e-07, + "loss": 0.8232, + "step": 49030 + }, + { + "epoch": 8.92, + "learning_rate": 3.253913360029123e-07, + "loss": 1.5291, + "step": 49031 + }, + { + "epoch": 8.92, + "learning_rate": 3.253367309792501e-07, + "loss": 0.0652, + "step": 49032 + }, + { + "epoch": 8.92, + "learning_rate": 3.2528212595558794e-07, + "loss": 0.2366, + "step": 49033 + }, + { + "epoch": 8.92, + "learning_rate": 3.252275209319257e-07, + "loss": 1.8039, + "step": 49034 + }, + { + "epoch": 8.92, + "learning_rate": 3.2517291590826356e-07, + "loss": 0.0098, + "step": 49035 + }, + { + "epoch": 8.92, + "learning_rate": 3.251183108846014e-07, + "loss": 0.0561, + "step": 49036 + }, + { + "epoch": 8.92, + "learning_rate": 3.250637058609392e-07, + "loss": 0.0839, + "step": 49037 + }, + { + "epoch": 8.92, + "learning_rate": 3.250091008372771e-07, + "loss": 0.0042, + "step": 49038 + }, + { + "epoch": 8.92, + "learning_rate": 3.2495449581361486e-07, + "loss": 0.0045, + "step": 49039 + }, + { + "epoch": 8.92, + "learning_rate": 3.248998907899527e-07, + "loss": 0.0341, + "step": 49040 + }, + { + "epoch": 8.92, + "learning_rate": 3.2484528576629053e-07, + "loss": 0.2077, + "step": 49041 + }, + { + "epoch": 8.92, + "learning_rate": 3.247906807426283e-07, + "loss": 0.7851, + "step": 49042 + }, + { + "epoch": 8.92, + "learning_rate": 3.2473607571896616e-07, + "loss": 0.0646, + "step": 49043 + }, + { + "epoch": 8.92, + "learning_rate": 3.2468147069530394e-07, + "loss": 0.7666, + "step": 49044 + }, + { + "epoch": 8.92, + "learning_rate": 3.246268656716418e-07, + "loss": 0.1013, + "step": 49045 + }, + { + "epoch": 8.92, + "learning_rate": 3.245722606479796e-07, + "loss": 1.9855, + "step": 49046 + }, + { + "epoch": 8.92, + "learning_rate": 3.2451765562431745e-07, + "loss": 0.0274, + "step": 49047 + }, + { + "epoch": 8.92, + "learning_rate": 3.244630506006553e-07, + "loss": 0.421, + "step": 49048 + }, + { + "epoch": 8.92, + "learning_rate": 3.244084455769931e-07, + "loss": 0.3336, + "step": 49049 + }, + { + "epoch": 8.92, + "learning_rate": 3.243538405533309e-07, + "loss": 0.1523, + "step": 49050 + }, + { + "epoch": 8.92, + "learning_rate": 3.2429923552966875e-07, + "loss": 0.0088, + "step": 49051 + }, + { + "epoch": 8.92, + "learning_rate": 3.2424463050600653e-07, + "loss": 0.5935, + "step": 49052 + }, + { + "epoch": 8.92, + "learning_rate": 3.241900254823444e-07, + "loss": 0.0056, + "step": 49053 + }, + { + "epoch": 8.92, + "learning_rate": 3.241354204586822e-07, + "loss": 0.0404, + "step": 49054 + }, + { + "epoch": 8.92, + "learning_rate": 3.2408081543502005e-07, + "loss": 0.0052, + "step": 49055 + }, + { + "epoch": 8.92, + "learning_rate": 3.2402621041135783e-07, + "loss": 0.0169, + "step": 49056 + }, + { + "epoch": 8.92, + "learning_rate": 3.2397160538769567e-07, + "loss": 0.216, + "step": 49057 + }, + { + "epoch": 8.92, + "learning_rate": 3.239170003640335e-07, + "loss": 1.1689, + "step": 49058 + }, + { + "epoch": 8.92, + "learning_rate": 3.238623953403713e-07, + "loss": 0.0992, + "step": 49059 + }, + { + "epoch": 8.92, + "learning_rate": 3.238077903167092e-07, + "loss": 0.01, + "step": 49060 + }, + { + "epoch": 8.92, + "learning_rate": 3.2375318529304697e-07, + "loss": 0.0386, + "step": 49061 + }, + { + "epoch": 8.92, + "learning_rate": 3.236985802693848e-07, + "loss": 0.3816, + "step": 49062 + }, + { + "epoch": 8.92, + "learning_rate": 3.2364397524572264e-07, + "loss": 1.405, + "step": 49063 + }, + { + "epoch": 8.92, + "learning_rate": 3.235893702220604e-07, + "loss": 0.0984, + "step": 49064 + }, + { + "epoch": 8.92, + "learning_rate": 3.2353476519839826e-07, + "loss": 0.265, + "step": 49065 + }, + { + "epoch": 8.92, + "learning_rate": 3.2348016017473605e-07, + "loss": 0.0076, + "step": 49066 + }, + { + "epoch": 8.92, + "learning_rate": 3.234255551510739e-07, + "loss": 1.7916, + "step": 49067 + }, + { + "epoch": 8.92, + "learning_rate": 3.233709501274117e-07, + "loss": 0.4172, + "step": 49068 + }, + { + "epoch": 8.92, + "learning_rate": 3.2331634510374956e-07, + "loss": 0.0183, + "step": 49069 + }, + { + "epoch": 8.92, + "learning_rate": 3.232617400800874e-07, + "loss": 0.7462, + "step": 49070 + }, + { + "epoch": 8.92, + "learning_rate": 3.232071350564252e-07, + "loss": 0.087, + "step": 49071 + }, + { + "epoch": 8.92, + "learning_rate": 3.23152530032763e-07, + "loss": 0.3645, + "step": 49072 + }, + { + "epoch": 8.92, + "learning_rate": 3.2309792500910086e-07, + "loss": 2.7122, + "step": 49073 + }, + { + "epoch": 8.92, + "learning_rate": 3.2304331998543864e-07, + "loss": 0.0348, + "step": 49074 + }, + { + "epoch": 8.92, + "learning_rate": 3.2298871496177653e-07, + "loss": 0.1178, + "step": 49075 + }, + { + "epoch": 8.92, + "learning_rate": 3.229341099381143e-07, + "loss": 0.006, + "step": 49076 + }, + { + "epoch": 8.92, + "learning_rate": 3.2287950491445215e-07, + "loss": 0.0402, + "step": 49077 + }, + { + "epoch": 8.92, + "learning_rate": 3.2282489989078994e-07, + "loss": 0.013, + "step": 49078 + }, + { + "epoch": 8.93, + "learning_rate": 3.227702948671278e-07, + "loss": 2.6751, + "step": 49079 + }, + { + "epoch": 8.93, + "learning_rate": 3.227156898434656e-07, + "loss": 0.3815, + "step": 49080 + }, + { + "epoch": 8.93, + "learning_rate": 3.226610848198034e-07, + "loss": 0.7825, + "step": 49081 + }, + { + "epoch": 8.93, + "learning_rate": 3.226064797961413e-07, + "loss": 0.7514, + "step": 49082 + }, + { + "epoch": 8.93, + "learning_rate": 3.225518747724791e-07, + "loss": 0.1889, + "step": 49083 + }, + { + "epoch": 8.93, + "learning_rate": 3.224972697488169e-07, + "loss": 0.2579, + "step": 49084 + }, + { + "epoch": 8.93, + "learning_rate": 3.2244266472515475e-07, + "loss": 1.3551, + "step": 49085 + }, + { + "epoch": 8.93, + "learning_rate": 3.2238805970149253e-07, + "loss": 0.2315, + "step": 49086 + }, + { + "epoch": 8.93, + "learning_rate": 3.2233345467783037e-07, + "loss": 0.0975, + "step": 49087 + }, + { + "epoch": 8.93, + "learning_rate": 3.2227884965416816e-07, + "loss": 0.0161, + "step": 49088 + }, + { + "epoch": 8.93, + "learning_rate": 3.22224244630506e-07, + "loss": 0.0036, + "step": 49089 + }, + { + "epoch": 8.93, + "learning_rate": 3.2216963960684383e-07, + "loss": 0.0385, + "step": 49090 + }, + { + "epoch": 8.93, + "learning_rate": 3.2211503458318167e-07, + "loss": 0.1271, + "step": 49091 + }, + { + "epoch": 8.93, + "learning_rate": 3.220604295595195e-07, + "loss": 0.5078, + "step": 49092 + }, + { + "epoch": 8.93, + "learning_rate": 3.220058245358573e-07, + "loss": 1.3683, + "step": 49093 + }, + { + "epoch": 8.93, + "learning_rate": 3.2195121951219513e-07, + "loss": 0.0762, + "step": 49094 + }, + { + "epoch": 8.93, + "learning_rate": 3.2189661448853297e-07, + "loss": 0.5469, + "step": 49095 + }, + { + "epoch": 8.93, + "learning_rate": 3.2184200946487075e-07, + "loss": 0.1438, + "step": 49096 + }, + { + "epoch": 8.93, + "learning_rate": 3.2178740444120864e-07, + "loss": 1.9709, + "step": 49097 + }, + { + "epoch": 8.93, + "learning_rate": 3.217327994175464e-07, + "loss": 0.4146, + "step": 49098 + }, + { + "epoch": 8.93, + "learning_rate": 3.2167819439388426e-07, + "loss": 0.0102, + "step": 49099 + }, + { + "epoch": 8.93, + "learning_rate": 3.2162358937022205e-07, + "loss": 0.4947, + "step": 49100 + }, + { + "epoch": 8.93, + "learning_rate": 3.215689843465599e-07, + "loss": 0.3552, + "step": 49101 + }, + { + "epoch": 8.93, + "learning_rate": 3.215143793228977e-07, + "loss": 0.2102, + "step": 49102 + }, + { + "epoch": 8.93, + "learning_rate": 3.214597742992355e-07, + "loss": 0.7207, + "step": 49103 + }, + { + "epoch": 8.93, + "learning_rate": 3.2140516927557334e-07, + "loss": 0.2383, + "step": 49104 + }, + { + "epoch": 8.93, + "learning_rate": 3.213505642519112e-07, + "loss": 0.038, + "step": 49105 + }, + { + "epoch": 8.93, + "learning_rate": 3.21295959228249e-07, + "loss": 1.1268, + "step": 49106 + }, + { + "epoch": 8.93, + "learning_rate": 3.2124135420458686e-07, + "loss": 0.007, + "step": 49107 + }, + { + "epoch": 8.93, + "learning_rate": 3.2118674918092464e-07, + "loss": 3.1533, + "step": 49108 + }, + { + "epoch": 8.93, + "learning_rate": 3.211321441572625e-07, + "loss": 0.6646, + "step": 49109 + }, + { + "epoch": 8.93, + "learning_rate": 3.2107753913360026e-07, + "loss": 0.18, + "step": 49110 + }, + { + "epoch": 8.93, + "learning_rate": 3.210229341099381e-07, + "loss": 0.0126, + "step": 49111 + }, + { + "epoch": 8.93, + "learning_rate": 3.2096832908627594e-07, + "loss": 1.8353, + "step": 49112 + }, + { + "epoch": 8.93, + "learning_rate": 3.209137240626138e-07, + "loss": 0.0681, + "step": 49113 + }, + { + "epoch": 8.93, + "learning_rate": 3.208591190389516e-07, + "loss": 0.1939, + "step": 49114 + }, + { + "epoch": 8.93, + "learning_rate": 3.208045140152894e-07, + "loss": 0.7957, + "step": 49115 + }, + { + "epoch": 8.93, + "learning_rate": 3.2074990899162724e-07, + "loss": 0.0103, + "step": 49116 + }, + { + "epoch": 8.93, + "learning_rate": 3.2069530396796507e-07, + "loss": 0.0069, + "step": 49117 + }, + { + "epoch": 8.93, + "learning_rate": 3.2064069894430286e-07, + "loss": 0.0022, + "step": 49118 + }, + { + "epoch": 8.93, + "learning_rate": 3.2058609392064075e-07, + "loss": 0.429, + "step": 49119 + }, + { + "epoch": 8.93, + "learning_rate": 3.2053148889697853e-07, + "loss": 1.081, + "step": 49120 + }, + { + "epoch": 8.93, + "learning_rate": 3.2047688387331637e-07, + "loss": 0.0788, + "step": 49121 + }, + { + "epoch": 8.93, + "learning_rate": 3.2042227884965415e-07, + "loss": 0.7828, + "step": 49122 + }, + { + "epoch": 8.93, + "learning_rate": 3.20367673825992e-07, + "loss": 2.3564, + "step": 49123 + }, + { + "epoch": 8.93, + "learning_rate": 3.2031306880232983e-07, + "loss": 0.097, + "step": 49124 + }, + { + "epoch": 8.93, + "learning_rate": 3.202584637786676e-07, + "loss": 0.0186, + "step": 49125 + }, + { + "epoch": 8.93, + "learning_rate": 3.2020385875500545e-07, + "loss": 0.242, + "step": 49126 + }, + { + "epoch": 8.93, + "learning_rate": 3.201492537313433e-07, + "loss": 1.2193, + "step": 49127 + }, + { + "epoch": 8.93, + "learning_rate": 3.2009464870768113e-07, + "loss": 0.1356, + "step": 49128 + }, + { + "epoch": 8.93, + "learning_rate": 3.2004004368401896e-07, + "loss": 0.3119, + "step": 49129 + }, + { + "epoch": 8.93, + "learning_rate": 3.1998543866035675e-07, + "loss": 0.0244, + "step": 49130 + }, + { + "epoch": 8.93, + "learning_rate": 3.199308336366946e-07, + "loss": 0.6072, + "step": 49131 + }, + { + "epoch": 8.93, + "learning_rate": 3.1987622861303237e-07, + "loss": 0.0088, + "step": 49132 + }, + { + "epoch": 8.93, + "learning_rate": 3.198216235893702e-07, + "loss": 0.6903, + "step": 49133 + }, + { + "epoch": 8.94, + "learning_rate": 3.1976701856570805e-07, + "loss": 2.0477, + "step": 49134 + }, + { + "epoch": 8.94, + "learning_rate": 3.197124135420459e-07, + "loss": 0.0191, + "step": 49135 + }, + { + "epoch": 8.94, + "learning_rate": 3.196578085183837e-07, + "loss": 0.1942, + "step": 49136 + }, + { + "epoch": 8.94, + "learning_rate": 3.196032034947215e-07, + "loss": 0.005, + "step": 49137 + }, + { + "epoch": 8.94, + "learning_rate": 3.1954859847105934e-07, + "loss": 0.0485, + "step": 49138 + }, + { + "epoch": 8.94, + "learning_rate": 3.194939934473972e-07, + "loss": 0.5116, + "step": 49139 + }, + { + "epoch": 8.94, + "learning_rate": 3.1943938842373497e-07, + "loss": 0.0043, + "step": 49140 + }, + { + "epoch": 8.94, + "learning_rate": 3.193847834000728e-07, + "loss": 0.1618, + "step": 49141 + }, + { + "epoch": 8.94, + "learning_rate": 3.1933017837641064e-07, + "loss": 0.1031, + "step": 49142 + }, + { + "epoch": 8.94, + "learning_rate": 3.192755733527485e-07, + "loss": 0.4001, + "step": 49143 + }, + { + "epoch": 8.94, + "learning_rate": 3.1922096832908626e-07, + "loss": 0.3809, + "step": 49144 + }, + { + "epoch": 8.94, + "learning_rate": 3.191663633054241e-07, + "loss": 0.0841, + "step": 49145 + }, + { + "epoch": 8.94, + "learning_rate": 3.1911175828176194e-07, + "loss": 0.0441, + "step": 49146 + }, + { + "epoch": 8.94, + "learning_rate": 3.190571532580997e-07, + "loss": 0.4428, + "step": 49147 + }, + { + "epoch": 8.94, + "learning_rate": 3.1900254823443756e-07, + "loss": 0.0278, + "step": 49148 + }, + { + "epoch": 8.94, + "learning_rate": 3.189479432107754e-07, + "loss": 0.3598, + "step": 49149 + }, + { + "epoch": 8.94, + "learning_rate": 3.1889333818711323e-07, + "loss": 0.1185, + "step": 49150 + }, + { + "epoch": 8.94, + "learning_rate": 3.1883873316345107e-07, + "loss": 0.0168, + "step": 49151 + }, + { + "epoch": 8.94, + "learning_rate": 3.1878412813978886e-07, + "loss": 0.0106, + "step": 49152 + }, + { + "epoch": 8.94, + "learning_rate": 3.187295231161267e-07, + "loss": 0.2582, + "step": 49153 + }, + { + "epoch": 8.94, + "learning_rate": 3.186749180924645e-07, + "loss": 0.2526, + "step": 49154 + }, + { + "epoch": 8.94, + "learning_rate": 3.186203130688023e-07, + "loss": 0.4491, + "step": 49155 + }, + { + "epoch": 8.94, + "learning_rate": 3.1856570804514015e-07, + "loss": 0.1144, + "step": 49156 + }, + { + "epoch": 8.94, + "learning_rate": 3.18511103021478e-07, + "loss": 0.3731, + "step": 49157 + }, + { + "epoch": 8.94, + "learning_rate": 3.1845649799781583e-07, + "loss": 0.0083, + "step": 49158 + }, + { + "epoch": 8.94, + "learning_rate": 3.184018929741536e-07, + "loss": 2.7622, + "step": 49159 + }, + { + "epoch": 8.94, + "learning_rate": 3.1834728795049145e-07, + "loss": 0.0744, + "step": 49160 + }, + { + "epoch": 8.94, + "learning_rate": 3.182926829268293e-07, + "loss": 0.4631, + "step": 49161 + }, + { + "epoch": 8.94, + "learning_rate": 3.1823807790316707e-07, + "loss": 0.2817, + "step": 49162 + }, + { + "epoch": 8.94, + "learning_rate": 3.181834728795049e-07, + "loss": 0.523, + "step": 49163 + }, + { + "epoch": 8.94, + "learning_rate": 3.1812886785584275e-07, + "loss": 2.9029, + "step": 49164 + }, + { + "epoch": 8.94, + "learning_rate": 3.180742628321806e-07, + "loss": 0.0599, + "step": 49165 + }, + { + "epoch": 8.94, + "learning_rate": 3.1801965780851837e-07, + "loss": 0.2677, + "step": 49166 + }, + { + "epoch": 8.94, + "learning_rate": 3.179650527848562e-07, + "loss": 2.4631, + "step": 49167 + }, + { + "epoch": 8.94, + "learning_rate": 3.1791044776119405e-07, + "loss": 0.6603, + "step": 49168 + }, + { + "epoch": 8.94, + "learning_rate": 3.1785584273753183e-07, + "loss": 2.753, + "step": 49169 + }, + { + "epoch": 8.94, + "learning_rate": 3.1780123771386967e-07, + "loss": 0.1275, + "step": 49170 + }, + { + "epoch": 8.94, + "learning_rate": 3.177466326902075e-07, + "loss": 1.2562, + "step": 49171 + }, + { + "epoch": 8.94, + "learning_rate": 3.1769202766654534e-07, + "loss": 0.7874, + "step": 49172 + }, + { + "epoch": 8.94, + "learning_rate": 3.176374226428832e-07, + "loss": 0.0558, + "step": 49173 + }, + { + "epoch": 8.94, + "learning_rate": 3.1758281761922096e-07, + "loss": 0.1186, + "step": 49174 + }, + { + "epoch": 8.94, + "learning_rate": 3.175282125955588e-07, + "loss": 0.2056, + "step": 49175 + }, + { + "epoch": 8.94, + "learning_rate": 3.174736075718966e-07, + "loss": 0.1836, + "step": 49176 + }, + { + "epoch": 8.94, + "learning_rate": 3.174190025482344e-07, + "loss": 0.4307, + "step": 49177 + }, + { + "epoch": 8.94, + "learning_rate": 3.173643975245723e-07, + "loss": 0.2312, + "step": 49178 + }, + { + "epoch": 8.94, + "learning_rate": 3.173097925009101e-07, + "loss": 0.0979, + "step": 49179 + }, + { + "epoch": 8.94, + "learning_rate": 3.1725518747724794e-07, + "loss": 0.4417, + "step": 49180 + }, + { + "epoch": 8.94, + "learning_rate": 3.172005824535857e-07, + "loss": 1.1694, + "step": 49181 + }, + { + "epoch": 8.94, + "learning_rate": 3.1714597742992356e-07, + "loss": 0.977, + "step": 49182 + }, + { + "epoch": 8.94, + "learning_rate": 3.170913724062614e-07, + "loss": 0.1057, + "step": 49183 + }, + { + "epoch": 8.94, + "learning_rate": 3.170367673825992e-07, + "loss": 0.3215, + "step": 49184 + }, + { + "epoch": 8.94, + "learning_rate": 3.16982162358937e-07, + "loss": 0.009, + "step": 49185 + }, + { + "epoch": 8.94, + "learning_rate": 3.1692755733527486e-07, + "loss": 0.0437, + "step": 49186 + }, + { + "epoch": 8.94, + "learning_rate": 3.168729523116127e-07, + "loss": 0.0142, + "step": 49187 + }, + { + "epoch": 8.94, + "learning_rate": 3.1681834728795053e-07, + "loss": 0.0045, + "step": 49188 + }, + { + "epoch": 8.95, + "learning_rate": 3.167637422642883e-07, + "loss": 0.1546, + "step": 49189 + }, + { + "epoch": 8.95, + "learning_rate": 3.1670913724062615e-07, + "loss": 0.3642, + "step": 49190 + }, + { + "epoch": 8.95, + "learning_rate": 3.1665453221696394e-07, + "loss": 0.0039, + "step": 49191 + }, + { + "epoch": 8.95, + "learning_rate": 3.165999271933018e-07, + "loss": 2.503, + "step": 49192 + }, + { + "epoch": 8.95, + "learning_rate": 3.165453221696396e-07, + "loss": 0.1947, + "step": 49193 + }, + { + "epoch": 8.95, + "learning_rate": 3.1649071714597745e-07, + "loss": 0.0044, + "step": 49194 + }, + { + "epoch": 8.95, + "learning_rate": 3.164361121223153e-07, + "loss": 0.0062, + "step": 49195 + }, + { + "epoch": 8.95, + "learning_rate": 3.1638150709865307e-07, + "loss": 0.2409, + "step": 49196 + }, + { + "epoch": 8.95, + "learning_rate": 3.163269020749909e-07, + "loss": 0.3177, + "step": 49197 + }, + { + "epoch": 8.95, + "learning_rate": 3.162722970513287e-07, + "loss": 0.082, + "step": 49198 + }, + { + "epoch": 8.95, + "learning_rate": 3.1621769202766653e-07, + "loss": 0.7291, + "step": 49199 + }, + { + "epoch": 8.95, + "learning_rate": 3.1616308700400437e-07, + "loss": 0.2188, + "step": 49200 + }, + { + "epoch": 8.95, + "learning_rate": 3.161084819803422e-07, + "loss": 0.0047, + "step": 49201 + }, + { + "epoch": 8.95, + "learning_rate": 3.1605387695668004e-07, + "loss": 1.369, + "step": 49202 + }, + { + "epoch": 8.95, + "learning_rate": 3.1599927193301783e-07, + "loss": 0.6238, + "step": 49203 + }, + { + "epoch": 8.95, + "learning_rate": 3.1594466690935567e-07, + "loss": 0.2577, + "step": 49204 + }, + { + "epoch": 8.95, + "learning_rate": 3.158900618856935e-07, + "loss": 0.0103, + "step": 49205 + }, + { + "epoch": 8.95, + "learning_rate": 3.158354568620313e-07, + "loss": 0.0742, + "step": 49206 + }, + { + "epoch": 8.95, + "learning_rate": 3.157808518383691e-07, + "loss": 0.0102, + "step": 49207 + }, + { + "epoch": 8.95, + "learning_rate": 3.1572624681470696e-07, + "loss": 0.2728, + "step": 49208 + }, + { + "epoch": 8.95, + "learning_rate": 3.156716417910448e-07, + "loss": 0.6503, + "step": 49209 + }, + { + "epoch": 8.95, + "learning_rate": 3.1561703676738264e-07, + "loss": 0.1143, + "step": 49210 + }, + { + "epoch": 8.95, + "learning_rate": 3.155624317437204e-07, + "loss": 1.4429, + "step": 49211 + }, + { + "epoch": 8.95, + "learning_rate": 3.1550782672005826e-07, + "loss": 0.087, + "step": 49212 + }, + { + "epoch": 8.95, + "learning_rate": 3.1545322169639605e-07, + "loss": 0.0576, + "step": 49213 + }, + { + "epoch": 8.95, + "learning_rate": 3.153986166727339e-07, + "loss": 0.013, + "step": 49214 + }, + { + "epoch": 8.95, + "learning_rate": 3.153440116490717e-07, + "loss": 0.0074, + "step": 49215 + }, + { + "epoch": 8.95, + "learning_rate": 3.1528940662540956e-07, + "loss": 0.0091, + "step": 49216 + }, + { + "epoch": 8.95, + "learning_rate": 3.152348016017474e-07, + "loss": 1.7743, + "step": 49217 + }, + { + "epoch": 8.95, + "learning_rate": 3.151801965780852e-07, + "loss": 0.0046, + "step": 49218 + }, + { + "epoch": 8.95, + "learning_rate": 3.15125591554423e-07, + "loss": 0.0263, + "step": 49219 + }, + { + "epoch": 8.95, + "learning_rate": 3.1507098653076086e-07, + "loss": 0.6429, + "step": 49220 + }, + { + "epoch": 8.95, + "learning_rate": 3.1501638150709864e-07, + "loss": 1.5844, + "step": 49221 + }, + { + "epoch": 8.95, + "learning_rate": 3.149617764834365e-07, + "loss": 0.0071, + "step": 49222 + }, + { + "epoch": 8.95, + "learning_rate": 3.149071714597743e-07, + "loss": 0.0991, + "step": 49223 + }, + { + "epoch": 8.95, + "learning_rate": 3.1485256643611215e-07, + "loss": 1.1999, + "step": 49224 + }, + { + "epoch": 8.95, + "learning_rate": 3.1479796141244994e-07, + "loss": 0.0584, + "step": 49225 + }, + { + "epoch": 8.95, + "learning_rate": 3.147433563887878e-07, + "loss": 1.7386, + "step": 49226 + }, + { + "epoch": 8.95, + "learning_rate": 3.146887513651256e-07, + "loss": 0.0043, + "step": 49227 + }, + { + "epoch": 8.95, + "learning_rate": 3.146341463414634e-07, + "loss": 0.257, + "step": 49228 + }, + { + "epoch": 8.95, + "learning_rate": 3.1457954131780123e-07, + "loss": 0.226, + "step": 49229 + }, + { + "epoch": 8.95, + "learning_rate": 3.1452493629413907e-07, + "loss": 0.3327, + "step": 49230 + }, + { + "epoch": 8.95, + "learning_rate": 3.144703312704769e-07, + "loss": 0.0061, + "step": 49231 + }, + { + "epoch": 8.95, + "learning_rate": 3.1441572624681475e-07, + "loss": 0.2425, + "step": 49232 + }, + { + "epoch": 8.95, + "learning_rate": 3.1436112122315253e-07, + "loss": 0.0114, + "step": 49233 + }, + { + "epoch": 8.95, + "learning_rate": 3.1430651619949037e-07, + "loss": 0.0427, + "step": 49234 + }, + { + "epoch": 8.95, + "learning_rate": 3.1425191117582815e-07, + "loss": 0.6931, + "step": 49235 + }, + { + "epoch": 8.95, + "learning_rate": 3.14197306152166e-07, + "loss": 0.8844, + "step": 49236 + }, + { + "epoch": 8.95, + "learning_rate": 3.1414270112850383e-07, + "loss": 0.0028, + "step": 49237 + }, + { + "epoch": 8.95, + "learning_rate": 3.1408809610484167e-07, + "loss": 0.2253, + "step": 49238 + }, + { + "epoch": 8.95, + "learning_rate": 3.140334910811795e-07, + "loss": 1.0955, + "step": 49239 + }, + { + "epoch": 8.95, + "learning_rate": 3.139788860575173e-07, + "loss": 0.1887, + "step": 49240 + }, + { + "epoch": 8.95, + "learning_rate": 3.139242810338551e-07, + "loss": 0.1197, + "step": 49241 + }, + { + "epoch": 8.95, + "learning_rate": 3.1386967601019296e-07, + "loss": 0.9777, + "step": 49242 + }, + { + "epoch": 8.95, + "learning_rate": 3.1381507098653075e-07, + "loss": 0.434, + "step": 49243 + }, + { + "epoch": 8.96, + "learning_rate": 3.137604659628686e-07, + "loss": 0.1263, + "step": 49244 + }, + { + "epoch": 8.96, + "learning_rate": 3.137058609392064e-07, + "loss": 1.6627, + "step": 49245 + }, + { + "epoch": 8.96, + "learning_rate": 3.1365125591554426e-07, + "loss": 0.1628, + "step": 49246 + }, + { + "epoch": 8.96, + "learning_rate": 3.1359665089188204e-07, + "loss": 0.0053, + "step": 49247 + }, + { + "epoch": 8.96, + "learning_rate": 3.135420458682199e-07, + "loss": 0.6671, + "step": 49248 + }, + { + "epoch": 8.96, + "learning_rate": 3.134874408445577e-07, + "loss": 0.1175, + "step": 49249 + }, + { + "epoch": 8.96, + "learning_rate": 3.134328358208955e-07, + "loss": 0.7649, + "step": 49250 + }, + { + "epoch": 8.96, + "learning_rate": 3.1337823079723334e-07, + "loss": 0.0104, + "step": 49251 + }, + { + "epoch": 8.96, + "learning_rate": 3.133236257735712e-07, + "loss": 0.0064, + "step": 49252 + }, + { + "epoch": 8.96, + "learning_rate": 3.13269020749909e-07, + "loss": 1.7304, + "step": 49253 + }, + { + "epoch": 8.96, + "learning_rate": 3.1321441572624685e-07, + "loss": 2.3394, + "step": 49254 + }, + { + "epoch": 8.96, + "learning_rate": 3.1315981070258464e-07, + "loss": 0.0073, + "step": 49255 + }, + { + "epoch": 8.96, + "learning_rate": 3.131052056789225e-07, + "loss": 0.0367, + "step": 49256 + }, + { + "epoch": 8.96, + "learning_rate": 3.1305060065526026e-07, + "loss": 1.3871, + "step": 49257 + }, + { + "epoch": 8.96, + "learning_rate": 3.129959956315981e-07, + "loss": 0.0941, + "step": 49258 + }, + { + "epoch": 8.96, + "learning_rate": 3.1294139060793594e-07, + "loss": 1.1268, + "step": 49259 + }, + { + "epoch": 8.96, + "learning_rate": 3.128867855842738e-07, + "loss": 3.7852, + "step": 49260 + }, + { + "epoch": 8.96, + "learning_rate": 3.128321805606116e-07, + "loss": 0.174, + "step": 49261 + }, + { + "epoch": 8.96, + "learning_rate": 3.127775755369494e-07, + "loss": 0.1329, + "step": 49262 + }, + { + "epoch": 8.96, + "learning_rate": 3.1272297051328723e-07, + "loss": 1.2649, + "step": 49263 + }, + { + "epoch": 8.96, + "learning_rate": 3.1266836548962507e-07, + "loss": 0.2173, + "step": 49264 + }, + { + "epoch": 8.96, + "learning_rate": 3.1261376046596286e-07, + "loss": 1.4338, + "step": 49265 + }, + { + "epoch": 8.96, + "learning_rate": 3.125591554423007e-07, + "loss": 1.7948, + "step": 49266 + }, + { + "epoch": 8.96, + "learning_rate": 3.1250455041863853e-07, + "loss": 0.0237, + "step": 49267 + }, + { + "epoch": 8.96, + "learning_rate": 3.1244994539497637e-07, + "loss": 0.0129, + "step": 49268 + }, + { + "epoch": 8.96, + "learning_rate": 3.1239534037131415e-07, + "loss": 0.0276, + "step": 49269 + }, + { + "epoch": 8.96, + "learning_rate": 3.12340735347652e-07, + "loss": 3.972, + "step": 49270 + }, + { + "epoch": 8.96, + "learning_rate": 3.1228613032398983e-07, + "loss": 0.0688, + "step": 49271 + }, + { + "epoch": 8.96, + "learning_rate": 3.122315253003276e-07, + "loss": 0.5494, + "step": 49272 + }, + { + "epoch": 8.96, + "learning_rate": 3.1217692027666545e-07, + "loss": 0.04, + "step": 49273 + }, + { + "epoch": 8.96, + "learning_rate": 3.121223152530033e-07, + "loss": 0.0045, + "step": 49274 + }, + { + "epoch": 8.96, + "learning_rate": 3.120677102293411e-07, + "loss": 0.4987, + "step": 49275 + }, + { + "epoch": 8.96, + "learning_rate": 3.1201310520567896e-07, + "loss": 1.6345, + "step": 49276 + }, + { + "epoch": 8.96, + "learning_rate": 3.1195850018201675e-07, + "loss": 0.7065, + "step": 49277 + }, + { + "epoch": 8.96, + "learning_rate": 3.119038951583546e-07, + "loss": 0.0391, + "step": 49278 + }, + { + "epoch": 8.96, + "learning_rate": 3.1184929013469237e-07, + "loss": 0.0702, + "step": 49279 + }, + { + "epoch": 8.96, + "learning_rate": 3.117946851110302e-07, + "loss": 1.3835, + "step": 49280 + }, + { + "epoch": 8.96, + "learning_rate": 3.1174008008736804e-07, + "loss": 0.0191, + "step": 49281 + }, + { + "epoch": 8.96, + "learning_rate": 3.116854750637059e-07, + "loss": 0.1842, + "step": 49282 + }, + { + "epoch": 8.96, + "learning_rate": 3.116308700400437e-07, + "loss": 2.9941, + "step": 49283 + }, + { + "epoch": 8.96, + "learning_rate": 3.115762650163815e-07, + "loss": 0.0094, + "step": 49284 + }, + { + "epoch": 8.96, + "learning_rate": 3.1152165999271934e-07, + "loss": 0.1659, + "step": 49285 + }, + { + "epoch": 8.96, + "learning_rate": 3.114670549690572e-07, + "loss": 0.143, + "step": 49286 + }, + { + "epoch": 8.96, + "learning_rate": 3.1141244994539496e-07, + "loss": 0.971, + "step": 49287 + }, + { + "epoch": 8.96, + "learning_rate": 3.113578449217328e-07, + "loss": 0.0855, + "step": 49288 + }, + { + "epoch": 8.96, + "learning_rate": 3.1130323989807064e-07, + "loss": 0.0534, + "step": 49289 + }, + { + "epoch": 8.96, + "learning_rate": 3.112486348744085e-07, + "loss": 0.0109, + "step": 49290 + }, + { + "epoch": 8.96, + "learning_rate": 3.1119402985074626e-07, + "loss": 0.9836, + "step": 49291 + }, + { + "epoch": 8.96, + "learning_rate": 3.111394248270841e-07, + "loss": 1.3579, + "step": 49292 + }, + { + "epoch": 8.96, + "learning_rate": 3.1108481980342194e-07, + "loss": 2.0505, + "step": 49293 + }, + { + "epoch": 8.96, + "learning_rate": 3.110302147797597e-07, + "loss": 0.2214, + "step": 49294 + }, + { + "epoch": 8.96, + "learning_rate": 3.1097560975609756e-07, + "loss": 0.0059, + "step": 49295 + }, + { + "epoch": 8.96, + "learning_rate": 3.109210047324354e-07, + "loss": 1.787, + "step": 49296 + }, + { + "epoch": 8.96, + "learning_rate": 3.1086639970877323e-07, + "loss": 1.7419, + "step": 49297 + }, + { + "epoch": 8.96, + "learning_rate": 3.1081179468511107e-07, + "loss": 0.1146, + "step": 49298 + }, + { + "epoch": 8.97, + "learning_rate": 3.1075718966144885e-07, + "loss": 1.8537, + "step": 49299 + }, + { + "epoch": 8.97, + "learning_rate": 3.107025846377867e-07, + "loss": 0.168, + "step": 49300 + }, + { + "epoch": 8.97, + "learning_rate": 3.106479796141245e-07, + "loss": 0.0124, + "step": 49301 + }, + { + "epoch": 8.97, + "learning_rate": 3.105933745904623e-07, + "loss": 1.3347, + "step": 49302 + }, + { + "epoch": 8.97, + "learning_rate": 3.1053876956680015e-07, + "loss": 1.0202, + "step": 49303 + }, + { + "epoch": 8.97, + "learning_rate": 3.10484164543138e-07, + "loss": 0.0172, + "step": 49304 + }, + { + "epoch": 8.97, + "learning_rate": 3.1042955951947583e-07, + "loss": 0.6233, + "step": 49305 + }, + { + "epoch": 8.97, + "learning_rate": 3.103749544958136e-07, + "loss": 1.0718, + "step": 49306 + }, + { + "epoch": 8.97, + "learning_rate": 3.1032034947215145e-07, + "loss": 0.4126, + "step": 49307 + }, + { + "epoch": 8.97, + "learning_rate": 3.102657444484893e-07, + "loss": 1.3362, + "step": 49308 + }, + { + "epoch": 8.97, + "learning_rate": 3.1021113942482707e-07, + "loss": 2.6836, + "step": 49309 + }, + { + "epoch": 8.97, + "learning_rate": 3.101565344011649e-07, + "loss": 0.3908, + "step": 49310 + }, + { + "epoch": 8.97, + "learning_rate": 3.1010192937750275e-07, + "loss": 0.2231, + "step": 49311 + }, + { + "epoch": 8.97, + "learning_rate": 3.100473243538406e-07, + "loss": 0.494, + "step": 49312 + }, + { + "epoch": 8.97, + "learning_rate": 3.0999271933017837e-07, + "loss": 0.0741, + "step": 49313 + }, + { + "epoch": 8.97, + "learning_rate": 3.099381143065162e-07, + "loss": 1.43, + "step": 49314 + }, + { + "epoch": 8.97, + "learning_rate": 3.0988350928285404e-07, + "loss": 1.6854, + "step": 49315 + }, + { + "epoch": 8.97, + "learning_rate": 3.0982890425919183e-07, + "loss": 0.0538, + "step": 49316 + }, + { + "epoch": 8.97, + "learning_rate": 3.0977429923552966e-07, + "loss": 0.0379, + "step": 49317 + }, + { + "epoch": 8.97, + "learning_rate": 3.097196942118675e-07, + "loss": 0.5654, + "step": 49318 + }, + { + "epoch": 8.97, + "learning_rate": 3.0966508918820534e-07, + "loss": 0.0919, + "step": 49319 + }, + { + "epoch": 8.97, + "learning_rate": 3.096104841645432e-07, + "loss": 2.6879, + "step": 49320 + }, + { + "epoch": 8.97, + "learning_rate": 3.0955587914088096e-07, + "loss": 0.1135, + "step": 49321 + }, + { + "epoch": 8.97, + "learning_rate": 3.095012741172188e-07, + "loss": 0.0446, + "step": 49322 + }, + { + "epoch": 8.97, + "learning_rate": 3.094466690935566e-07, + "loss": 0.0268, + "step": 49323 + }, + { + "epoch": 8.97, + "learning_rate": 3.093920640698944e-07, + "loss": 0.3216, + "step": 49324 + }, + { + "epoch": 8.97, + "learning_rate": 3.0933745904623226e-07, + "loss": 0.0456, + "step": 49325 + }, + { + "epoch": 8.97, + "learning_rate": 3.092828540225701e-07, + "loss": 0.6396, + "step": 49326 + }, + { + "epoch": 8.97, + "learning_rate": 3.0922824899890793e-07, + "loss": 2.5704, + "step": 49327 + }, + { + "epoch": 8.97, + "learning_rate": 3.091736439752457e-07, + "loss": 1.1423, + "step": 49328 + }, + { + "epoch": 8.97, + "learning_rate": 3.0911903895158356e-07, + "loss": 0.0298, + "step": 49329 + }, + { + "epoch": 8.97, + "learning_rate": 3.090644339279214e-07, + "loss": 0.0846, + "step": 49330 + }, + { + "epoch": 8.97, + "learning_rate": 3.090098289042592e-07, + "loss": 0.0455, + "step": 49331 + }, + { + "epoch": 8.97, + "learning_rate": 3.08955223880597e-07, + "loss": 0.0043, + "step": 49332 + }, + { + "epoch": 8.97, + "learning_rate": 3.0890061885693485e-07, + "loss": 1.1375, + "step": 49333 + }, + { + "epoch": 8.97, + "learning_rate": 3.088460138332727e-07, + "loss": 0.1854, + "step": 49334 + }, + { + "epoch": 8.97, + "learning_rate": 3.087914088096105e-07, + "loss": 0.1339, + "step": 49335 + }, + { + "epoch": 8.97, + "learning_rate": 3.087368037859483e-07, + "loss": 0.0036, + "step": 49336 + }, + { + "epoch": 8.97, + "learning_rate": 3.0868219876228615e-07, + "loss": 0.3785, + "step": 49337 + }, + { + "epoch": 8.97, + "learning_rate": 3.0862759373862394e-07, + "loss": 0.0072, + "step": 49338 + }, + { + "epoch": 8.97, + "learning_rate": 3.0857298871496177e-07, + "loss": 0.0457, + "step": 49339 + }, + { + "epoch": 8.97, + "learning_rate": 3.085183836912996e-07, + "loss": 0.0538, + "step": 49340 + }, + { + "epoch": 8.97, + "learning_rate": 3.0846377866763745e-07, + "loss": 0.0563, + "step": 49341 + }, + { + "epoch": 8.97, + "learning_rate": 3.084091736439753e-07, + "loss": 2.0876, + "step": 49342 + }, + { + "epoch": 8.97, + "learning_rate": 3.0835456862031307e-07, + "loss": 0.0405, + "step": 49343 + }, + { + "epoch": 8.97, + "learning_rate": 3.082999635966509e-07, + "loss": 0.4718, + "step": 49344 + }, + { + "epoch": 8.97, + "learning_rate": 3.082453585729887e-07, + "loss": 0.0685, + "step": 49345 + }, + { + "epoch": 8.97, + "learning_rate": 3.0819075354932653e-07, + "loss": 0.2524, + "step": 49346 + }, + { + "epoch": 8.97, + "learning_rate": 3.0813614852566437e-07, + "loss": 1.2821, + "step": 49347 + }, + { + "epoch": 8.97, + "learning_rate": 3.080815435020022e-07, + "loss": 0.4078, + "step": 49348 + }, + { + "epoch": 8.97, + "learning_rate": 3.0802693847834004e-07, + "loss": 0.0038, + "step": 49349 + }, + { + "epoch": 8.97, + "learning_rate": 3.0797233345467783e-07, + "loss": 0.0466, + "step": 49350 + }, + { + "epoch": 8.97, + "learning_rate": 3.0791772843101566e-07, + "loss": 1.9845, + "step": 49351 + }, + { + "epoch": 8.97, + "learning_rate": 3.078631234073535e-07, + "loss": 1.783, + "step": 49352 + }, + { + "epoch": 8.97, + "learning_rate": 3.078085183836913e-07, + "loss": 0.9434, + "step": 49353 + }, + { + "epoch": 8.98, + "learning_rate": 3.077539133600291e-07, + "loss": 0.16, + "step": 49354 + }, + { + "epoch": 8.98, + "learning_rate": 3.076993083363669e-07, + "loss": 2.1342, + "step": 49355 + }, + { + "epoch": 8.98, + "learning_rate": 3.076447033127048e-07, + "loss": 1.0277, + "step": 49356 + }, + { + "epoch": 8.98, + "learning_rate": 3.075900982890426e-07, + "loss": 0.0625, + "step": 49357 + }, + { + "epoch": 8.98, + "learning_rate": 3.075354932653804e-07, + "loss": 0.4095, + "step": 49358 + }, + { + "epoch": 8.98, + "learning_rate": 3.0748088824171826e-07, + "loss": 0.0731, + "step": 49359 + }, + { + "epoch": 8.98, + "learning_rate": 3.0742628321805604e-07, + "loss": 1.008, + "step": 49360 + }, + { + "epoch": 8.98, + "learning_rate": 3.073716781943939e-07, + "loss": 0.9634, + "step": 49361 + }, + { + "epoch": 8.98, + "learning_rate": 3.073170731707317e-07, + "loss": 6.8061, + "step": 49362 + }, + { + "epoch": 8.98, + "learning_rate": 3.0726246814706956e-07, + "loss": 0.113, + "step": 49363 + }, + { + "epoch": 8.98, + "learning_rate": 3.072078631234074e-07, + "loss": 0.01, + "step": 49364 + }, + { + "epoch": 8.98, + "learning_rate": 3.071532580997452e-07, + "loss": 0.015, + "step": 49365 + }, + { + "epoch": 8.98, + "learning_rate": 3.07098653076083e-07, + "loss": 0.0046, + "step": 49366 + }, + { + "epoch": 8.98, + "learning_rate": 3.070440480524208e-07, + "loss": 0.2069, + "step": 49367 + }, + { + "epoch": 8.98, + "learning_rate": 3.0698944302875864e-07, + "loss": 1.1856, + "step": 49368 + }, + { + "epoch": 8.98, + "learning_rate": 3.069348380050965e-07, + "loss": 0.0073, + "step": 49369 + }, + { + "epoch": 8.98, + "learning_rate": 3.068802329814343e-07, + "loss": 0.0114, + "step": 49370 + }, + { + "epoch": 8.98, + "learning_rate": 3.0682562795777215e-07, + "loss": 0.8594, + "step": 49371 + }, + { + "epoch": 8.98, + "learning_rate": 3.0677102293410993e-07, + "loss": 0.4394, + "step": 49372 + }, + { + "epoch": 8.98, + "learning_rate": 3.0671641791044777e-07, + "loss": 0.0055, + "step": 49373 + }, + { + "epoch": 8.98, + "learning_rate": 3.066618128867856e-07, + "loss": 0.3444, + "step": 49374 + }, + { + "epoch": 8.98, + "learning_rate": 3.066072078631234e-07, + "loss": 0.5302, + "step": 49375 + }, + { + "epoch": 8.98, + "learning_rate": 3.0655260283946123e-07, + "loss": 0.9222, + "step": 49376 + }, + { + "epoch": 8.98, + "learning_rate": 3.06497997815799e-07, + "loss": 0.6393, + "step": 49377 + }, + { + "epoch": 8.98, + "learning_rate": 3.064433927921369e-07, + "loss": 0.3746, + "step": 49378 + }, + { + "epoch": 8.98, + "learning_rate": 3.0638878776847474e-07, + "loss": 0.9856, + "step": 49379 + }, + { + "epoch": 8.98, + "learning_rate": 3.0633418274481253e-07, + "loss": 0.0623, + "step": 49380 + }, + { + "epoch": 8.98, + "learning_rate": 3.0627957772115037e-07, + "loss": 2.529, + "step": 49381 + }, + { + "epoch": 8.98, + "learning_rate": 3.0622497269748815e-07, + "loss": 2.2664, + "step": 49382 + }, + { + "epoch": 8.98, + "learning_rate": 3.06170367673826e-07, + "loss": 0.1927, + "step": 49383 + }, + { + "epoch": 8.98, + "learning_rate": 3.061157626501638e-07, + "loss": 0.0582, + "step": 49384 + }, + { + "epoch": 8.98, + "learning_rate": 3.0606115762650166e-07, + "loss": 3.1781, + "step": 49385 + }, + { + "epoch": 8.98, + "learning_rate": 3.060065526028395e-07, + "loss": 0.6818, + "step": 49386 + }, + { + "epoch": 8.98, + "learning_rate": 3.059519475791773e-07, + "loss": 0.3303, + "step": 49387 + }, + { + "epoch": 8.98, + "learning_rate": 3.058973425555151e-07, + "loss": 0.0151, + "step": 49388 + }, + { + "epoch": 8.98, + "learning_rate": 3.0584273753185296e-07, + "loss": 0.1482, + "step": 49389 + }, + { + "epoch": 8.98, + "learning_rate": 3.0578813250819074e-07, + "loss": 0.0062, + "step": 49390 + }, + { + "epoch": 8.98, + "learning_rate": 3.057335274845286e-07, + "loss": 1.5402, + "step": 49391 + }, + { + "epoch": 8.98, + "learning_rate": 3.056789224608664e-07, + "loss": 0.0273, + "step": 49392 + }, + { + "epoch": 8.98, + "learning_rate": 3.0562431743720426e-07, + "loss": 0.7657, + "step": 49393 + }, + { + "epoch": 8.98, + "learning_rate": 3.0556971241354204e-07, + "loss": 0.0435, + "step": 49394 + }, + { + "epoch": 8.98, + "learning_rate": 3.055151073898799e-07, + "loss": 0.9514, + "step": 49395 + }, + { + "epoch": 8.98, + "learning_rate": 3.054605023662177e-07, + "loss": 0.1759, + "step": 49396 + }, + { + "epoch": 8.98, + "learning_rate": 3.054058973425555e-07, + "loss": 0.138, + "step": 49397 + }, + { + "epoch": 8.98, + "learning_rate": 3.0535129231889334e-07, + "loss": 0.093, + "step": 49398 + }, + { + "epoch": 8.98, + "learning_rate": 3.052966872952311e-07, + "loss": 0.1194, + "step": 49399 + }, + { + "epoch": 8.98, + "learning_rate": 3.05242082271569e-07, + "loss": 0.0772, + "step": 49400 + }, + { + "epoch": 8.98, + "learning_rate": 3.0518747724790685e-07, + "loss": 0.0248, + "step": 49401 + }, + { + "epoch": 8.98, + "learning_rate": 3.0513287222424464e-07, + "loss": 0.1689, + "step": 49402 + }, + { + "epoch": 8.98, + "learning_rate": 3.050782672005825e-07, + "loss": 0.2687, + "step": 49403 + }, + { + "epoch": 8.98, + "learning_rate": 3.0502366217692026e-07, + "loss": 0.0043, + "step": 49404 + }, + { + "epoch": 8.98, + "learning_rate": 3.049690571532581e-07, + "loss": 0.0436, + "step": 49405 + }, + { + "epoch": 8.98, + "learning_rate": 3.0491445212959593e-07, + "loss": 0.9799, + "step": 49406 + }, + { + "epoch": 8.98, + "learning_rate": 3.0485984710593377e-07, + "loss": 2.1162, + "step": 49407 + }, + { + "epoch": 8.98, + "learning_rate": 3.048052420822716e-07, + "loss": 0.3649, + "step": 49408 + }, + { + "epoch": 8.99, + "learning_rate": 3.047506370586094e-07, + "loss": 0.6537, + "step": 49409 + }, + { + "epoch": 8.99, + "learning_rate": 3.0469603203494723e-07, + "loss": 0.1714, + "step": 49410 + }, + { + "epoch": 8.99, + "learning_rate": 3.0464142701128507e-07, + "loss": 1.0327, + "step": 49411 + }, + { + "epoch": 8.99, + "learning_rate": 3.0458682198762285e-07, + "loss": 1.3536, + "step": 49412 + }, + { + "epoch": 8.99, + "learning_rate": 3.045322169639607e-07, + "loss": 0.1249, + "step": 49413 + }, + { + "epoch": 8.99, + "learning_rate": 3.044776119402985e-07, + "loss": 0.3002, + "step": 49414 + }, + { + "epoch": 8.99, + "learning_rate": 3.0442300691663637e-07, + "loss": 0.3976, + "step": 49415 + }, + { + "epoch": 8.99, + "learning_rate": 3.0436840189297415e-07, + "loss": 1.0586, + "step": 49416 + }, + { + "epoch": 8.99, + "learning_rate": 3.04313796869312e-07, + "loss": 0.5218, + "step": 49417 + }, + { + "epoch": 8.99, + "learning_rate": 3.042591918456498e-07, + "loss": 0.0037, + "step": 49418 + }, + { + "epoch": 8.99, + "learning_rate": 3.042045868219876e-07, + "loss": 0.6433, + "step": 49419 + }, + { + "epoch": 8.99, + "learning_rate": 3.0414998179832545e-07, + "loss": 0.0034, + "step": 49420 + }, + { + "epoch": 8.99, + "learning_rate": 3.040953767746633e-07, + "loss": 1.8166, + "step": 49421 + }, + { + "epoch": 8.99, + "learning_rate": 3.040407717510011e-07, + "loss": 0.0036, + "step": 49422 + }, + { + "epoch": 8.99, + "learning_rate": 3.0398616672733896e-07, + "loss": 0.6738, + "step": 49423 + }, + { + "epoch": 8.99, + "learning_rate": 3.0393156170367674e-07, + "loss": 0.1787, + "step": 49424 + }, + { + "epoch": 8.99, + "learning_rate": 3.038769566800146e-07, + "loss": 0.0096, + "step": 49425 + }, + { + "epoch": 8.99, + "learning_rate": 3.0382235165635237e-07, + "loss": 0.0151, + "step": 49426 + }, + { + "epoch": 8.99, + "learning_rate": 3.037677466326902e-07, + "loss": 0.0694, + "step": 49427 + }, + { + "epoch": 8.99, + "learning_rate": 3.0371314160902804e-07, + "loss": 0.0073, + "step": 49428 + }, + { + "epoch": 8.99, + "learning_rate": 3.036585365853659e-07, + "loss": 0.2296, + "step": 49429 + }, + { + "epoch": 8.99, + "learning_rate": 3.036039315617037e-07, + "loss": 3.655, + "step": 49430 + }, + { + "epoch": 8.99, + "learning_rate": 3.035493265380415e-07, + "loss": 0.4573, + "step": 49431 + }, + { + "epoch": 8.99, + "learning_rate": 3.0349472151437934e-07, + "loss": 0.0044, + "step": 49432 + }, + { + "epoch": 8.99, + "learning_rate": 3.034401164907172e-07, + "loss": 0.8503, + "step": 49433 + }, + { + "epoch": 8.99, + "learning_rate": 3.0338551146705496e-07, + "loss": 0.0036, + "step": 49434 + }, + { + "epoch": 8.99, + "learning_rate": 3.033309064433928e-07, + "loss": 0.1658, + "step": 49435 + }, + { + "epoch": 8.99, + "learning_rate": 3.032763014197306e-07, + "loss": 0.1548, + "step": 49436 + }, + { + "epoch": 8.99, + "learning_rate": 3.0322169639606847e-07, + "loss": 0.3439, + "step": 49437 + }, + { + "epoch": 8.99, + "learning_rate": 3.0316709137240626e-07, + "loss": 0.8749, + "step": 49438 + }, + { + "epoch": 8.99, + "learning_rate": 3.031124863487441e-07, + "loss": 0.0437, + "step": 49439 + }, + { + "epoch": 8.99, + "learning_rate": 3.0305788132508193e-07, + "loss": 0.7777, + "step": 49440 + }, + { + "epoch": 8.99, + "learning_rate": 3.030032763014197e-07, + "loss": 0.8215, + "step": 49441 + }, + { + "epoch": 8.99, + "learning_rate": 3.0294867127775755e-07, + "loss": 0.3941, + "step": 49442 + }, + { + "epoch": 8.99, + "learning_rate": 3.028940662540954e-07, + "loss": 0.0042, + "step": 49443 + }, + { + "epoch": 8.99, + "learning_rate": 3.0283946123043323e-07, + "loss": 0.1404, + "step": 49444 + }, + { + "epoch": 8.99, + "learning_rate": 3.0278485620677107e-07, + "loss": 0.0112, + "step": 49445 + }, + { + "epoch": 8.99, + "learning_rate": 3.0273025118310885e-07, + "loss": 0.1905, + "step": 49446 + }, + { + "epoch": 8.99, + "learning_rate": 3.026756461594467e-07, + "loss": 1.0135, + "step": 49447 + }, + { + "epoch": 8.99, + "learning_rate": 3.026210411357845e-07, + "loss": 0.1259, + "step": 49448 + }, + { + "epoch": 8.99, + "learning_rate": 3.025664361121223e-07, + "loss": 0.5177, + "step": 49449 + }, + { + "epoch": 8.99, + "learning_rate": 3.0251183108846015e-07, + "loss": 0.0136, + "step": 49450 + }, + { + "epoch": 8.99, + "learning_rate": 3.02457226064798e-07, + "loss": 0.0621, + "step": 49451 + }, + { + "epoch": 8.99, + "learning_rate": 3.024026210411358e-07, + "loss": 0.056, + "step": 49452 + }, + { + "epoch": 8.99, + "learning_rate": 3.023480160174736e-07, + "loss": 0.1391, + "step": 49453 + }, + { + "epoch": 8.99, + "learning_rate": 3.0229341099381145e-07, + "loss": 1.3727, + "step": 49454 + }, + { + "epoch": 8.99, + "learning_rate": 3.022388059701493e-07, + "loss": 0.0815, + "step": 49455 + }, + { + "epoch": 8.99, + "learning_rate": 3.0218420094648707e-07, + "loss": 1.5839, + "step": 49456 + }, + { + "epoch": 8.99, + "learning_rate": 3.021295959228249e-07, + "loss": 0.0904, + "step": 49457 + }, + { + "epoch": 8.99, + "learning_rate": 3.020749908991627e-07, + "loss": 0.0074, + "step": 49458 + }, + { + "epoch": 8.99, + "learning_rate": 3.020203858755006e-07, + "loss": 0.0112, + "step": 49459 + }, + { + "epoch": 8.99, + "learning_rate": 3.0196578085183837e-07, + "loss": 0.0073, + "step": 49460 + }, + { + "epoch": 8.99, + "learning_rate": 3.019111758281762e-07, + "loss": 1.2661, + "step": 49461 + }, + { + "epoch": 8.99, + "learning_rate": 3.0185657080451404e-07, + "loss": 1.2855, + "step": 49462 + }, + { + "epoch": 8.99, + "learning_rate": 3.018019657808518e-07, + "loss": 0.0633, + "step": 49463 + }, + { + "epoch": 9.0, + "learning_rate": 3.0174736075718966e-07, + "loss": 0.091, + "step": 49464 + }, + { + "epoch": 9.0, + "learning_rate": 3.016927557335275e-07, + "loss": 0.3119, + "step": 49465 + }, + { + "epoch": 9.0, + "learning_rate": 3.0163815070986534e-07, + "loss": 0.0836, + "step": 49466 + }, + { + "epoch": 9.0, + "learning_rate": 3.015835456862032e-07, + "loss": 1.2472, + "step": 49467 + }, + { + "epoch": 9.0, + "learning_rate": 3.0152894066254096e-07, + "loss": 0.7834, + "step": 49468 + }, + { + "epoch": 9.0, + "learning_rate": 3.014743356388788e-07, + "loss": 0.0145, + "step": 49469 + }, + { + "epoch": 9.0, + "learning_rate": 3.014197306152166e-07, + "loss": 0.0159, + "step": 49470 + }, + { + "epoch": 9.0, + "learning_rate": 3.013651255915544e-07, + "loss": 1.3437, + "step": 49471 + }, + { + "epoch": 9.0, + "learning_rate": 3.0131052056789226e-07, + "loss": 0.5207, + "step": 49472 + }, + { + "epoch": 9.0, + "learning_rate": 3.0125591554423004e-07, + "loss": 0.6736, + "step": 49473 + }, + { + "epoch": 9.0, + "learning_rate": 3.0120131052056793e-07, + "loss": 0.0184, + "step": 49474 + }, + { + "epoch": 9.0, + "learning_rate": 3.011467054969057e-07, + "loss": 2.1588, + "step": 49475 + }, + { + "epoch": 9.0, + "learning_rate": 3.0109210047324355e-07, + "loss": 0.327, + "step": 49476 + }, + { + "epoch": 9.0, + "learning_rate": 3.010374954495814e-07, + "loss": 0.0477, + "step": 49477 + }, + { + "epoch": 9.0, + "learning_rate": 3.009828904259192e-07, + "loss": 2.1312, + "step": 49478 + }, + { + "epoch": 9.0, + "learning_rate": 3.00928285402257e-07, + "loss": 0.189, + "step": 49479 + }, + { + "epoch": 9.0, + "learning_rate": 3.008736803785948e-07, + "loss": 0.2052, + "step": 49480 + }, + { + "epoch": 9.0, + "learning_rate": 3.008190753549327e-07, + "loss": 0.0387, + "step": 49481 + }, + { + "epoch": 9.0, + "learning_rate": 3.0076447033127047e-07, + "loss": 0.0061, + "step": 49482 + }, + { + "epoch": 9.0, + "learning_rate": 3.007098653076083e-07, + "loss": 0.0403, + "step": 49483 + }, + { + "epoch": 9.0, + "learning_rate": 3.0065526028394615e-07, + "loss": 0.7661, + "step": 49484 + }, + { + "epoch": 9.0, + "learning_rate": 3.0060065526028393e-07, + "loss": 0.2637, + "step": 49485 + }, + { + "epoch": 9.0, + "learning_rate": 3.0054605023662177e-07, + "loss": 0.0431, + "step": 49486 + }, + { + "epoch": 9.0, + "learning_rate": 3.004914452129596e-07, + "loss": 0.0231, + "step": 49487 + }, + { + "epoch": 9.0, + "learning_rate": 3.0043684018929745e-07, + "loss": 0.0703, + "step": 49488 + }, + { + "epoch": 9.0, + "learning_rate": 3.003822351656353e-07, + "loss": 0.5345, + "step": 49489 + }, + { + "epoch": 9.0, + "learning_rate": 3.0032763014197307e-07, + "loss": 1.0435, + "step": 49490 + }, + { + "epoch": 9.0, + "learning_rate": 3.002730251183109e-07, + "loss": 2.271, + "step": 49491 + }, + { + "epoch": 9.0, + "eval_accuracy": 0.8821603927986906, + "eval_loss": 0.5778843760490417, + "eval_runtime": 139.3416, + "eval_samples_per_second": 17.54, + "eval_steps_per_second": 4.385, + "step": 49491 + } + ], + "max_steps": 54990, + "num_train_epochs": 10, + "total_flos": 6.834735824793512e+18, + "trial_name": null, + "trial_params": null +}