| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.68504, | |
| "eval_steps": 500, | |
| "global_step": 8563, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 9.999999842086332e-06, | |
| "loss": 3.8678, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 9.99999605215876e-06, | |
| "loss": 3.1958, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 9.999984208641271e-06, | |
| "loss": 2.7416, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 9.999964469466236e-06, | |
| "loss": 2.7303, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 9.99993683466483e-06, | |
| "loss": 2.6118, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 9.999901304280686e-06, | |
| "loss": 2.5647, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 9.999857878369917e-06, | |
| "loss": 2.5395, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 9.999806557001092e-06, | |
| "loss": 2.4767, | |
| "step": 35 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 9.99974734025526e-06, | |
| "loss": 2.4768, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 9.99968022822593e-06, | |
| "loss": 2.4417, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 9.999605221019082e-06, | |
| "loss": 2.5207, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 9.99952231875316e-06, | |
| "loss": 2.4503, | |
| "step": 55 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 9.999431521559081e-06, | |
| "loss": 2.5273, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.999332829580227e-06, | |
| "loss": 2.5419, | |
| "step": 65 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.999226242972445e-06, | |
| "loss": 2.4085, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.999111761904046e-06, | |
| "loss": 2.4276, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.998989386555815e-06, | |
| "loss": 2.4457, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.998859117121e-06, | |
| "loss": 2.5333, | |
| "step": 85 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.998720953805312e-06, | |
| "loss": 2.5607, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.998574896826931e-06, | |
| "loss": 2.4188, | |
| "step": 95 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.9984209464165e-06, | |
| "loss": 2.3849, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.99825910281713e-06, | |
| "loss": 2.5013, | |
| "step": 105 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.998089366284392e-06, | |
| "loss": 2.4716, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.997911737086322e-06, | |
| "loss": 2.4551, | |
| "step": 115 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.997726215503422e-06, | |
| "loss": 2.4105, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.997532801828659e-06, | |
| "loss": 2.4917, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.997331496367455e-06, | |
| "loss": 2.5187, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.9971222994377e-06, | |
| "loss": 2.4179, | |
| "step": 135 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.996905211369748e-06, | |
| "loss": 2.4432, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.996680232506404e-06, | |
| "loss": 2.4748, | |
| "step": 145 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.996447363202947e-06, | |
| "loss": 2.4549, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.996206603827105e-06, | |
| "loss": 2.413, | |
| "step": 155 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.995957954759073e-06, | |
| "loss": 2.3685, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.9957014163915e-06, | |
| "loss": 2.4965, | |
| "step": 165 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.995436989129495e-06, | |
| "loss": 2.4297, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.995164673390624e-06, | |
| "loss": 2.4488, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.994884469604913e-06, | |
| "loss": 2.4427, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.99459637821484e-06, | |
| "loss": 2.4176, | |
| "step": 185 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.994300399675342e-06, | |
| "loss": 2.4354, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.993996534453812e-06, | |
| "loss": 2.4271, | |
| "step": 195 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.99368478303009e-06, | |
| "loss": 2.4104, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.993365145896473e-06, | |
| "loss": 2.472, | |
| "step": 205 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.993037623557716e-06, | |
| "loss": 2.3864, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.99270221653102e-06, | |
| "loss": 2.4398, | |
| "step": 215 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.99235892534604e-06, | |
| "loss": 2.4776, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.992007750544876e-06, | |
| "loss": 2.4641, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.991648692682083e-06, | |
| "loss": 2.418, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.991281752324664e-06, | |
| "loss": 2.4393, | |
| "step": 235 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.990906930052065e-06, | |
| "loss": 2.4516, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.990524226456182e-06, | |
| "loss": 2.5284, | |
| "step": 245 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.990133642141359e-06, | |
| "loss": 2.3922, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.989735177724378e-06, | |
| "loss": 2.4243, | |
| "step": 255 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.989328833834472e-06, | |
| "loss": 2.4688, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.988914611113311e-06, | |
| "loss": 2.4507, | |
| "step": 265 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.988492510215011e-06, | |
| "loss": 2.4323, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.988062531806127e-06, | |
| "loss": 2.4468, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.987624676565652e-06, | |
| "loss": 2.4449, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.987178945185019e-06, | |
| "loss": 2.4528, | |
| "step": 285 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.986725338368103e-06, | |
| "loss": 2.4404, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.986263856831204e-06, | |
| "loss": 2.3838, | |
| "step": 295 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.98579450130307e-06, | |
| "loss": 2.6499, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.985317272524876e-06, | |
| "loss": 2.7636, | |
| "step": 305 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.98483217125023e-06, | |
| "loss": 2.5432, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.984339198245175e-06, | |
| "loss": 2.5596, | |
| "step": 315 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.983838354288181e-06, | |
| "loss": 2.558, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.98332964017015e-06, | |
| "loss": 2.5666, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.982813056694411e-06, | |
| "loss": 2.5362, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.982288604676719e-06, | |
| "loss": 2.8506, | |
| "step": 335 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.981756284945256e-06, | |
| "loss": 2.7035, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.98121609834063e-06, | |
| "loss": 2.5648, | |
| "step": 345 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.980668045715864e-06, | |
| "loss": 2.9797, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.98011212793641e-06, | |
| "loss": 2.6807, | |
| "step": 355 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.979548345880142e-06, | |
| "loss": 2.694, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.978976700437341e-06, | |
| "loss": 2.5717, | |
| "step": 365 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.978397192510722e-06, | |
| "loss": 2.5549, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.9778098230154e-06, | |
| "loss": 2.5463, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.977214592878917e-06, | |
| "loss": 2.4522, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.976611503041218e-06, | |
| "loss": 2.4995, | |
| "step": 385 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.976000554454668e-06, | |
| "loss": 2.4276, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.975381748084035e-06, | |
| "loss": 2.4565, | |
| "step": 395 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.974755084906503e-06, | |
| "loss": 2.4484, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.974120565911653e-06, | |
| "loss": 2.4033, | |
| "step": 405 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.97347819210148e-06, | |
| "loss": 2.4396, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.972827964490382e-06, | |
| "loss": 2.4695, | |
| "step": 415 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.972169884105155e-06, | |
| "loss": 2.4338, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.971503951984996e-06, | |
| "loss": 2.4564, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.970830169181504e-06, | |
| "loss": 2.4482, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.970148536758678e-06, | |
| "loss": 2.5479, | |
| "step": 435 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.969459055792903e-06, | |
| "loss": 2.4472, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.968761727372965e-06, | |
| "loss": 2.4117, | |
| "step": 445 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.968056552600043e-06, | |
| "loss": 2.4558, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.967343532587701e-06, | |
| "loss": 2.4719, | |
| "step": 455 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.966622668461899e-06, | |
| "loss": 2.4085, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.965893961360977e-06, | |
| "loss": 2.4209, | |
| "step": 465 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.965157412435663e-06, | |
| "loss": 2.4407, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.964413022849069e-06, | |
| "loss": 2.3714, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.963660793776689e-06, | |
| "loss": 2.4234, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.96290072640639e-06, | |
| "loss": 2.423, | |
| "step": 485 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.96213282193843e-06, | |
| "loss": 2.4753, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.96135708158543e-06, | |
| "loss": 2.342, | |
| "step": 495 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.960573506572391e-06, | |
| "loss": 2.4249, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.959782098136683e-06, | |
| "loss": 2.3401, | |
| "step": 505 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.958982857528053e-06, | |
| "loss": 2.4531, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.958175786008605e-06, | |
| "loss": 2.4166, | |
| "step": 515 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.957360884852819e-06, | |
| "loss": 2.3662, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.956538155347534e-06, | |
| "loss": 2.3679, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.955707598791952e-06, | |
| "loss": 2.3647, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.954869216497636e-06, | |
| "loss": 2.41, | |
| "step": 535 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.954023009788505e-06, | |
| "loss": 2.3764, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.953168980000836e-06, | |
| "loss": 2.424, | |
| "step": 545 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.952307128483257e-06, | |
| "loss": 2.3777, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.951437456596751e-06, | |
| "loss": 2.4414, | |
| "step": 555 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.950559965714647e-06, | |
| "loss": 2.4281, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.949674657222624e-06, | |
| "loss": 2.3643, | |
| "step": 565 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.948781532518706e-06, | |
| "loss": 2.411, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.947880593013256e-06, | |
| "loss": 2.4174, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.946971840128982e-06, | |
| "loss": 2.3497, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.946055275300929e-06, | |
| "loss": 2.3523, | |
| "step": 585 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.945130899976477e-06, | |
| "loss": 2.372, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.944198715615343e-06, | |
| "loss": 2.3717, | |
| "step": 595 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.94325872368957e-06, | |
| "loss": 2.3737, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.942310925683538e-06, | |
| "loss": 2.4498, | |
| "step": 605 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.941355323093944e-06, | |
| "loss": 2.3515, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.94039191742982e-06, | |
| "loss": 2.3897, | |
| "step": 615 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.939420710212511e-06, | |
| "loss": 2.4276, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.938441702975689e-06, | |
| "loss": 2.4084, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.937454897265338e-06, | |
| "loss": 2.4149, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.93646029463976e-06, | |
| "loss": 2.3673, | |
| "step": 635 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.935457896669568e-06, | |
| "loss": 2.3262, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.934447704937684e-06, | |
| "loss": 2.3838, | |
| "step": 645 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.93342972103934e-06, | |
| "loss": 2.377, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.932403946582071e-06, | |
| "loss": 2.4281, | |
| "step": 655 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.931370383185717e-06, | |
| "loss": 2.3754, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.930329032482412e-06, | |
| "loss": 2.4245, | |
| "step": 665 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.929279896116595e-06, | |
| "loss": 2.3126, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.928222975744992e-06, | |
| "loss": 2.4046, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.927158273036624e-06, | |
| "loss": 2.3757, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.926085789672806e-06, | |
| "loss": 2.3906, | |
| "step": 685 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.925005527347132e-06, | |
| "loss": 2.3512, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.923917487765484e-06, | |
| "loss": 2.4564, | |
| "step": 695 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.922821672646028e-06, | |
| "loss": 2.3998, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.921718083719203e-06, | |
| "loss": 2.4962, | |
| "step": 705 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.920606722727726e-06, | |
| "loss": 2.3641, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.919487591426591e-06, | |
| "loss": 2.3407, | |
| "step": 715 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.918360691583056e-06, | |
| "loss": 2.3559, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.91722602497665e-06, | |
| "loss": 2.4193, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.916083593399167e-06, | |
| "loss": 2.3301, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.914933398654663e-06, | |
| "loss": 2.4055, | |
| "step": 735 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.913775442559451e-06, | |
| "loss": 2.4251, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.912609726942104e-06, | |
| "loss": 2.3905, | |
| "step": 745 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.911436253643445e-06, | |
| "loss": 2.3312, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.910255024516546e-06, | |
| "loss": 2.3348, | |
| "step": 755 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.909066041426733e-06, | |
| "loss": 2.3927, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.907869306251571e-06, | |
| "loss": 2.3886, | |
| "step": 765 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.906664820880869e-06, | |
| "loss": 2.382, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.90545258721667e-06, | |
| "loss": 2.3468, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.904232607173262e-06, | |
| "loss": 2.3527, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.903004882677157e-06, | |
| "loss": 2.3709, | |
| "step": 785 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.9017694156671e-06, | |
| "loss": 2.4618, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.900526208094061e-06, | |
| "loss": 2.341, | |
| "step": 795 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.899275261921236e-06, | |
| "loss": 2.3537, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.898016579124039e-06, | |
| "loss": 2.449, | |
| "step": 805 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.8967501616901e-06, | |
| "loss": 2.3372, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.895476011619269e-06, | |
| "loss": 2.4462, | |
| "step": 815 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.894194130923602e-06, | |
| "loss": 2.4556, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.89290452162736e-06, | |
| "loss": 2.462, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.891607185767018e-06, | |
| "loss": 2.3579, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.89030212539124e-06, | |
| "loss": 2.4308, | |
| "step": 835 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.8889893425609e-06, | |
| "loss": 2.4658, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.887668839349057e-06, | |
| "loss": 2.3401, | |
| "step": 845 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.886340617840968e-06, | |
| "loss": 2.4264, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.885004680134075e-06, | |
| "loss": 2.3856, | |
| "step": 855 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.883661028338009e-06, | |
| "loss": 2.3479, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.882309664574576e-06, | |
| "loss": 2.5498, | |
| "step": 865 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.880950590977764e-06, | |
| "loss": 2.5088, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.879583809693737e-06, | |
| "loss": 2.3513, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.87820932288083e-06, | |
| "loss": 2.3834, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.876827132709545e-06, | |
| "loss": 2.4223, | |
| "step": 885 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.875437241362546e-06, | |
| "loss": 2.3918, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.874039651034665e-06, | |
| "loss": 2.3935, | |
| "step": 895 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.872634363932887e-06, | |
| "loss": 2.3382, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.87122138227635e-06, | |
| "loss": 2.3507, | |
| "step": 905 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.869800708296347e-06, | |
| "loss": 2.3881, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.868372344236314e-06, | |
| "loss": 2.3536, | |
| "step": 915 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.866936292351837e-06, | |
| "loss": 2.3739, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.865492554910634e-06, | |
| "loss": 2.3488, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.864041134192563e-06, | |
| "loss": 2.3285, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.862582032489621e-06, | |
| "loss": 2.3309, | |
| "step": 935 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.861115252105922e-06, | |
| "loss": 2.3653, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.859640795357716e-06, | |
| "loss": 2.3722, | |
| "step": 945 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.85815866457337e-06, | |
| "loss": 2.3694, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.856668862093372e-06, | |
| "loss": 2.3212, | |
| "step": 955 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.855171390270325e-06, | |
| "loss": 2.3071, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.853666251468938e-06, | |
| "loss": 2.3044, | |
| "step": 965 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.852153448066031e-06, | |
| "loss": 2.3917, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.85063298245053e-06, | |
| "loss": 2.308, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.849104857023455e-06, | |
| "loss": 2.3165, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.847569074197927e-06, | |
| "loss": 2.2926, | |
| "step": 985 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.846025636399152e-06, | |
| "loss": 2.409, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.844474546064436e-06, | |
| "loss": 2.4206, | |
| "step": 995 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.842915805643156e-06, | |
| "loss": 2.4242, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.84134941759678e-06, | |
| "loss": 2.4045, | |
| "step": 1005 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.839775384398846e-06, | |
| "loss": 2.2667, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.838193708534969e-06, | |
| "loss": 2.348, | |
| "step": 1015 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.836604392502829e-06, | |
| "loss": 2.4119, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.835007438812177e-06, | |
| "loss": 2.349, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.833402849984815e-06, | |
| "loss": 2.37, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.831790628554613e-06, | |
| "loss": 2.3195, | |
| "step": 1035 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.830170777067486e-06, | |
| "loss": 2.4033, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.828543298081401e-06, | |
| "loss": 2.2899, | |
| "step": 1045 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.82690819416637e-06, | |
| "loss": 2.3335, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.825265467904446e-06, | |
| "loss": 2.3617, | |
| "step": 1055 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.823615121889716e-06, | |
| "loss": 2.4366, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.821957158728302e-06, | |
| "loss": 2.3295, | |
| "step": 1065 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.820291581038354e-06, | |
| "loss": 2.386, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.81861839145005e-06, | |
| "loss": 2.4164, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.81693759260558e-06, | |
| "loss": 2.4298, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.815249187159158e-06, | |
| "loss": 2.3613, | |
| "step": 1085 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.813553177777005e-06, | |
| "loss": 2.3239, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.811849567137351e-06, | |
| "loss": 2.3454, | |
| "step": 1095 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.81013835793043e-06, | |
| "loss": 2.3593, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.808419552858477e-06, | |
| "loss": 2.369, | |
| "step": 1105 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.806693154635719e-06, | |
| "loss": 2.3099, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.80495916598837e-06, | |
| "loss": 2.3498, | |
| "step": 1115 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.803217589654642e-06, | |
| "loss": 2.4043, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.801468428384716e-06, | |
| "loss": 2.3933, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.79971168494076e-06, | |
| "loss": 2.3453, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.797947362096909e-06, | |
| "loss": 2.3699, | |
| "step": 1135 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.796175462639273e-06, | |
| "loss": 2.3842, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.79439598936592e-06, | |
| "loss": 2.353, | |
| "step": 1145 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.79260894508688e-06, | |
| "loss": 2.3314, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.790814332624144e-06, | |
| "loss": 2.3427, | |
| "step": 1155 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.789012154811648e-06, | |
| "loss": 2.2935, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.787202414495275e-06, | |
| "loss": 2.3644, | |
| "step": 1165 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.785385114532858e-06, | |
| "loss": 2.3309, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.783560257794153e-06, | |
| "loss": 2.3602, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.781727847160865e-06, | |
| "loss": 2.3894, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.779887885526616e-06, | |
| "loss": 2.382, | |
| "step": 1185 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.77804037579696e-06, | |
| "loss": 2.4115, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.776185320889364e-06, | |
| "loss": 2.3761, | |
| "step": 1195 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.774322723733216e-06, | |
| "loss": 2.3394, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.772452587269808e-06, | |
| "loss": 2.3536, | |
| "step": 1205 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.770574914452343e-06, | |
| "loss": 2.353, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.768689708245921e-06, | |
| "loss": 2.3405, | |
| "step": 1215 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.766796971627543e-06, | |
| "loss": 2.3894, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.764896707586095e-06, | |
| "loss": 2.3086, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.762988919122354e-06, | |
| "loss": 2.3898, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.761073609248981e-06, | |
| "loss": 2.3703, | |
| "step": 1235 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.759150780990508e-06, | |
| "loss": 2.3538, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.757220437383345e-06, | |
| "loss": 2.4508, | |
| "step": 1245 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.755282581475769e-06, | |
| "loss": 2.3082, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.753337216327917e-06, | |
| "loss": 2.4109, | |
| "step": 1255 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.751384345011787e-06, | |
| "loss": 2.366, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.749423970611232e-06, | |
| "loss": 2.3802, | |
| "step": 1265 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.747456096221946e-06, | |
| "loss": 2.3878, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.745480724951473e-06, | |
| "loss": 2.3319, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.743497859919196e-06, | |
| "loss": 2.3459, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.741507504256327e-06, | |
| "loss": 2.4305, | |
| "step": 1285 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.739509661105912e-06, | |
| "loss": 2.3212, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.737504333622814e-06, | |
| "loss": 2.3519, | |
| "step": 1295 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.735491524973723e-06, | |
| "loss": 2.4035, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.733471238337136e-06, | |
| "loss": 2.2749, | |
| "step": 1305 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.73144347690336e-06, | |
| "loss": 2.4223, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.729408243874511e-06, | |
| "loss": 2.3059, | |
| "step": 1315 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.727365542464498e-06, | |
| "loss": 2.3024, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.725315375899025e-06, | |
| "loss": 2.2939, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.723257747415584e-06, | |
| "loss": 2.3402, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.721192660263454e-06, | |
| "loss": 2.366, | |
| "step": 1335 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.719120117703688e-06, | |
| "loss": 2.4011, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.717040123009111e-06, | |
| "loss": 2.3492, | |
| "step": 1345 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.714952679464324e-06, | |
| "loss": 2.3405, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.71285779036568e-06, | |
| "loss": 2.3419, | |
| "step": 1355 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.710755459021297e-06, | |
| "loss": 2.3302, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.708645688751043e-06, | |
| "loss": 2.2838, | |
| "step": 1365 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.706528482886535e-06, | |
| "loss": 2.2852, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.704403844771128e-06, | |
| "loss": 2.3822, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.702271777759915e-06, | |
| "loss": 2.3764, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.700132285219724e-06, | |
| "loss": 2.3767, | |
| "step": 1385 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.697985370529101e-06, | |
| "loss": 2.383, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.695831037078323e-06, | |
| "loss": 2.2937, | |
| "step": 1395 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.693669288269371e-06, | |
| "loss": 2.3489, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.691500127515945e-06, | |
| "loss": 2.3873, | |
| "step": 1405 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.689323558243446e-06, | |
| "loss": 2.3831, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.687139583888971e-06, | |
| "loss": 2.3385, | |
| "step": 1415 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.684948207901315e-06, | |
| "loss": 2.3264, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.682749433740963e-06, | |
| "loss": 2.3455, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.680543264880075e-06, | |
| "loss": 2.3379, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.678329704802495e-06, | |
| "loss": 2.3679, | |
| "step": 1435 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.676108757003735e-06, | |
| "loss": 2.3293, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.673880424990978e-06, | |
| "loss": 2.3727, | |
| "step": 1445 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.671644712283061e-06, | |
| "loss": 2.3898, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.669401622410481e-06, | |
| "loss": 2.3139, | |
| "step": 1455 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.667151158915382e-06, | |
| "loss": 2.3811, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.664893325351556e-06, | |
| "loss": 2.2852, | |
| "step": 1465 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.662628125284426e-06, | |
| "loss": 2.3856, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.660355562291055e-06, | |
| "loss": 2.3889, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.65807563996013e-06, | |
| "loss": 2.32, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.65578836189196e-06, | |
| "loss": 2.4046, | |
| "step": 1485 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.653493731698467e-06, | |
| "loss": 2.317, | |
| "step": 1490 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.651191753003187e-06, | |
| "loss": 2.3666, | |
| "step": 1495 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.648882429441258e-06, | |
| "loss": 2.3768, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.646565764659418e-06, | |
| "loss": 2.3277, | |
| "step": 1505 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.644241762315995e-06, | |
| "loss": 2.3817, | |
| "step": 1510 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.641910426080909e-06, | |
| "loss": 2.3487, | |
| "step": 1515 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.639571759635655e-06, | |
| "loss": 2.3855, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.637225766673309e-06, | |
| "loss": 2.3123, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.634872450898511e-06, | |
| "loss": 2.3796, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.63251181602747e-06, | |
| "loss": 2.3167, | |
| "step": 1535 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.630143865787951e-06, | |
| "loss": 2.3651, | |
| "step": 1540 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.62776860391927e-06, | |
| "loss": 2.3516, | |
| "step": 1545 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.62538603417229e-06, | |
| "loss": 2.3768, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.622996160309415e-06, | |
| "loss": 2.4051, | |
| "step": 1555 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.620598986104578e-06, | |
| "loss": 2.392, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.61819451534325e-06, | |
| "loss": 2.2801, | |
| "step": 1565 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.615782751822413e-06, | |
| "loss": 2.3499, | |
| "step": 1570 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.613363699350575e-06, | |
| "loss": 2.4028, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.610937361747747e-06, | |
| "loss": 2.4018, | |
| "step": 1580 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.60850374284545e-06, | |
| "loss": 2.3549, | |
| "step": 1585 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.606062846486698e-06, | |
| "loss": 2.3292, | |
| "step": 1590 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.603614676526e-06, | |
| "loss": 2.3204, | |
| "step": 1595 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.601159236829353e-06, | |
| "loss": 2.3532, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.598696531274227e-06, | |
| "loss": 2.3334, | |
| "step": 1605 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.596226563749575e-06, | |
| "loss": 2.3013, | |
| "step": 1610 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.593749338155809e-06, | |
| "loss": 2.475, | |
| "step": 1615 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.591264858404809e-06, | |
| "loss": 2.3784, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.588773128419907e-06, | |
| "loss": 2.3192, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.586274152135883e-06, | |
| "loss": 2.31, | |
| "step": 1630 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.583767933498964e-06, | |
| "loss": 2.3635, | |
| "step": 1635 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.58125447646681e-06, | |
| "loss": 2.3542, | |
| "step": 1640 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.578733785008513e-06, | |
| "loss": 2.3946, | |
| "step": 1645 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.576205863104588e-06, | |
| "loss": 2.3609, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.573670714746973e-06, | |
| "loss": 2.3461, | |
| "step": 1655 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.571128343939006e-06, | |
| "loss": 2.3565, | |
| "step": 1660 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.568578754695443e-06, | |
| "loss": 2.3446, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.566021951042432e-06, | |
| "loss": 2.3353, | |
| "step": 1670 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.563457937017514e-06, | |
| "loss": 2.399, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.56088671666962e-06, | |
| "loss": 2.3797, | |
| "step": 1680 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.558308294059055e-06, | |
| "loss": 2.3472, | |
| "step": 1685 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.555722673257502e-06, | |
| "loss": 2.3961, | |
| "step": 1690 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.553129858348006e-06, | |
| "loss": 2.3554, | |
| "step": 1695 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.550529853424979e-06, | |
| "loss": 2.3868, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.547922662594183e-06, | |
| "loss": 2.2986, | |
| "step": 1705 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.545308289972727e-06, | |
| "loss": 2.3284, | |
| "step": 1710 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.542686739689064e-06, | |
| "loss": 2.3453, | |
| "step": 1715 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.54005801588298e-06, | |
| "loss": 2.3722, | |
| "step": 1720 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.537422122705585e-06, | |
| "loss": 2.377, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.534779064319318e-06, | |
| "loss": 2.3361, | |
| "step": 1730 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.532128844897928e-06, | |
| "loss": 2.2815, | |
| "step": 1735 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.529471468626472e-06, | |
| "loss": 2.3555, | |
| "step": 1740 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.52680693970131e-06, | |
| "loss": 2.3748, | |
| "step": 1745 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.524135262330098e-06, | |
| "loss": 2.3354, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.52145644073178e-06, | |
| "loss": 2.3775, | |
| "step": 1755 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.51877047913658e-06, | |
| "loss": 2.2927, | |
| "step": 1760 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.516077381785995e-06, | |
| "loss": 2.3331, | |
| "step": 1765 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.513377152932796e-06, | |
| "loss": 2.3614, | |
| "step": 1770 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.510669796841014e-06, | |
| "loss": 2.2829, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.507955317785935e-06, | |
| "loss": 2.3235, | |
| "step": 1780 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.505233720054086e-06, | |
| "loss": 2.4012, | |
| "step": 1785 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.502505007943248e-06, | |
| "loss": 2.32, | |
| "step": 1790 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.499769185762425e-06, | |
| "loss": 2.3578, | |
| "step": 1795 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.497026257831856e-06, | |
| "loss": 2.3312, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.494276228482998e-06, | |
| "loss": 2.3463, | |
| "step": 1805 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.491519102058523e-06, | |
| "loss": 2.4038, | |
| "step": 1810 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.48875488291231e-06, | |
| "loss": 2.4344, | |
| "step": 1815 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.48598357540944e-06, | |
| "loss": 2.3112, | |
| "step": 1820 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.48320518392618e-06, | |
| "loss": 2.3172, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.480419712849996e-06, | |
| "loss": 2.3116, | |
| "step": 1830 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.477627166579523e-06, | |
| "loss": 2.3388, | |
| "step": 1835 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.474827549524574e-06, | |
| "loss": 2.4493, | |
| "step": 1840 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.472020866106128e-06, | |
| "loss": 2.33, | |
| "step": 1845 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.46920712075632e-06, | |
| "loss": 2.343, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.466386317918436e-06, | |
| "loss": 2.3657, | |
| "step": 1855 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.463558462046912e-06, | |
| "loss": 2.3117, | |
| "step": 1860 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.460723557607317e-06, | |
| "loss": 2.3932, | |
| "step": 1865 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.457881609076352e-06, | |
| "loss": 2.3644, | |
| "step": 1870 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.45503262094184e-06, | |
| "loss": 2.3048, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.452176597702724e-06, | |
| "loss": 2.4008, | |
| "step": 1880 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.449313543869056e-06, | |
| "loss": 2.2881, | |
| "step": 1885 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.446443463961986e-06, | |
| "loss": 2.3549, | |
| "step": 1890 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.443566362513763e-06, | |
| "loss": 2.3225, | |
| "step": 1895 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.440682244067724e-06, | |
| "loss": 2.3483, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.437791113178283e-06, | |
| "loss": 2.2936, | |
| "step": 1905 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.434892974410932e-06, | |
| "loss": 2.4011, | |
| "step": 1910 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.431987832342228e-06, | |
| "loss": 2.3626, | |
| "step": 1915 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.429075691559788e-06, | |
| "loss": 2.3457, | |
| "step": 1920 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.426156556662276e-06, | |
| "loss": 2.2791, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.423230432259409e-06, | |
| "loss": 2.3234, | |
| "step": 1930 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.420297322971934e-06, | |
| "loss": 2.3283, | |
| "step": 1935 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.41735723343163e-06, | |
| "loss": 2.3014, | |
| "step": 1940 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.414410168281303e-06, | |
| "loss": 2.333, | |
| "step": 1945 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.411456132174768e-06, | |
| "loss": 2.3213, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.408495129776851e-06, | |
| "loss": 2.3893, | |
| "step": 1955 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.405527165763384e-06, | |
| "loss": 2.3552, | |
| "step": 1960 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.402552244821181e-06, | |
| "loss": 2.3828, | |
| "step": 1965 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.399570371648052e-06, | |
| "loss": 2.3239, | |
| "step": 1970 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.396581550952781e-06, | |
| "loss": 2.3524, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.393585787455125e-06, | |
| "loss": 2.3007, | |
| "step": 1980 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.3905830858858e-06, | |
| "loss": 2.3578, | |
| "step": 1985 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.387573450986485e-06, | |
| "loss": 2.3611, | |
| "step": 1990 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.384556887509802e-06, | |
| "loss": 2.3784, | |
| "step": 1995 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.381533400219319e-06, | |
| "loss": 2.3655, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.378502993889533e-06, | |
| "loss": 2.3281, | |
| "step": 2005 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.37546567330587e-06, | |
| "loss": 2.3777, | |
| "step": 2010 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.372421443264672e-06, | |
| "loss": 2.4091, | |
| "step": 2015 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.369370308573198e-06, | |
| "loss": 2.3085, | |
| "step": 2020 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.366312274049602e-06, | |
| "loss": 2.2421, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.363247344522939e-06, | |
| "loss": 2.3942, | |
| "step": 2030 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.360175524833153e-06, | |
| "loss": 2.317, | |
| "step": 2035 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.357096819831065e-06, | |
| "loss": 2.3567, | |
| "step": 2040 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.35401123437837e-06, | |
| "loss": 2.3427, | |
| "step": 2045 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.35091877334763e-06, | |
| "loss": 2.3237, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.347819441622261e-06, | |
| "loss": 2.3364, | |
| "step": 2055 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.344713244096533e-06, | |
| "loss": 2.3238, | |
| "step": 2060 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.341600185675555e-06, | |
| "loss": 2.2656, | |
| "step": 2065 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.33848027127527e-06, | |
| "loss": 2.2998, | |
| "step": 2070 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.33535350582245e-06, | |
| "loss": 2.3362, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.332219894254686e-06, | |
| "loss": 2.3953, | |
| "step": 2080 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.329079441520377e-06, | |
| "loss": 2.3724, | |
| "step": 2085 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.325932152578726e-06, | |
| "loss": 2.2792, | |
| "step": 2090 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.322778032399728e-06, | |
| "loss": 2.3206, | |
| "step": 2095 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.319617085964177e-06, | |
| "loss": 2.395, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.316449318263635e-06, | |
| "loss": 2.3439, | |
| "step": 2105 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.31327473430044e-06, | |
| "loss": 2.2661, | |
| "step": 2110 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.31009333908769e-06, | |
| "loss": 2.3384, | |
| "step": 2115 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.30690513764925e-06, | |
| "loss": 2.345, | |
| "step": 2120 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.30371013501972e-06, | |
| "loss": 2.348, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.300508336244443e-06, | |
| "loss": 2.3047, | |
| "step": 2130 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.297299746379503e-06, | |
| "loss": 2.3079, | |
| "step": 2135 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.294084370491695e-06, | |
| "loss": 2.3399, | |
| "step": 2140 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.29086221365854e-06, | |
| "loss": 2.3256, | |
| "step": 2145 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.287633280968263e-06, | |
| "loss": 2.3425, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.284397577519788e-06, | |
| "loss": 2.3558, | |
| "step": 2155 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.281155108422732e-06, | |
| "loss": 2.3339, | |
| "step": 2160 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.277905878797401e-06, | |
| "loss": 2.2839, | |
| "step": 2165 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.274649893774768e-06, | |
| "loss": 2.3632, | |
| "step": 2170 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.271387158496477e-06, | |
| "loss": 2.3651, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.268117678114833e-06, | |
| "loss": 2.3686, | |
| "step": 2180 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.264841457792795e-06, | |
| "loss": 2.3371, | |
| "step": 2185 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.26155850270396e-06, | |
| "loss": 2.2492, | |
| "step": 2190 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.25826881803256e-06, | |
| "loss": 2.3617, | |
| "step": 2195 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.25497240897346e-06, | |
| "loss": 2.3833, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.251669280732137e-06, | |
| "loss": 2.2811, | |
| "step": 2205 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.248359438524683e-06, | |
| "loss": 2.3552, | |
| "step": 2210 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.245042887577789e-06, | |
| "loss": 2.3454, | |
| "step": 2215 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.241719633128743e-06, | |
| "loss": 2.2956, | |
| "step": 2220 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.238389680425417e-06, | |
| "loss": 2.3762, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.235053034726261e-06, | |
| "loss": 2.3329, | |
| "step": 2230 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.231709701300293e-06, | |
| "loss": 2.2914, | |
| "step": 2235 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.228359685427095e-06, | |
| "loss": 2.295, | |
| "step": 2240 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.225002992396797e-06, | |
| "loss": 2.3728, | |
| "step": 2245 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.221639627510076e-06, | |
| "loss": 2.365, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.218269596078145e-06, | |
| "loss": 2.3951, | |
| "step": 2255 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.214892903422745e-06, | |
| "loss": 2.3543, | |
| "step": 2260 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.21150955487613e-06, | |
| "loss": 2.3172, | |
| "step": 2265 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.208119555781074e-06, | |
| "loss": 2.338, | |
| "step": 2270 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.204722911490847e-06, | |
| "loss": 2.347, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.201319627369211e-06, | |
| "loss": 2.3313, | |
| "step": 2280 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.19790970879042e-06, | |
| "loss": 2.3183, | |
| "step": 2285 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.1944931611392e-06, | |
| "loss": 2.4201, | |
| "step": 2290 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.191069989810743e-06, | |
| "loss": 2.2605, | |
| "step": 2295 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.18764020021071e-06, | |
| "loss": 2.3619, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.1842037977552e-06, | |
| "loss": 2.3647, | |
| "step": 2305 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.180760787870766e-06, | |
| "loss": 2.3536, | |
| "step": 2310 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.17731117599439e-06, | |
| "loss": 2.3556, | |
| "step": 2315 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.173854967573479e-06, | |
| "loss": 2.3082, | |
| "step": 2320 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.170392168065858e-06, | |
| "loss": 2.3464, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.166922782939759e-06, | |
| "loss": 2.3798, | |
| "step": 2330 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.163446817673817e-06, | |
| "loss": 2.4031, | |
| "step": 2335 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.159964277757054e-06, | |
| "loss": 2.3262, | |
| "step": 2340 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.156475168688876e-06, | |
| "loss": 2.3394, | |
| "step": 2345 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.152979495979064e-06, | |
| "loss": 2.3255, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.149477265147762e-06, | |
| "loss": 2.2822, | |
| "step": 2355 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.145968481725466e-06, | |
| "loss": 2.3709, | |
| "step": 2360 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.142453151253033e-06, | |
| "loss": 2.3608, | |
| "step": 2365 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.13893127928164e-06, | |
| "loss": 2.3551, | |
| "step": 2370 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.13540287137281e-06, | |
| "loss": 2.2749, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.131867933098379e-06, | |
| "loss": 2.298, | |
| "step": 2380 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.128326470040495e-06, | |
| "loss": 2.2613, | |
| "step": 2385 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.124778487791615e-06, | |
| "loss": 2.3305, | |
| "step": 2390 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.121223991954484e-06, | |
| "loss": 2.2651, | |
| "step": 2395 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.117662988142138e-06, | |
| "loss": 2.2488, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.114095481977887e-06, | |
| "loss": 2.3113, | |
| "step": 2405 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.110521479095314e-06, | |
| "loss": 2.4102, | |
| "step": 2410 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.106940985138251e-06, | |
| "loss": 2.3155, | |
| "step": 2415 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.10335400576079e-06, | |
| "loss": 2.3533, | |
| "step": 2420 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.099760546627262e-06, | |
| "loss": 2.3666, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.096160613412228e-06, | |
| "loss": 2.3796, | |
| "step": 2430 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.092554211800476e-06, | |
| "loss": 2.3945, | |
| "step": 2435 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.088941347487004e-06, | |
| "loss": 2.2972, | |
| "step": 2440 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.085322026177017e-06, | |
| "loss": 2.278, | |
| "step": 2445 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.08169625358592e-06, | |
| "loss": 2.3882, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.078064035439303e-06, | |
| "loss": 2.2649, | |
| "step": 2455 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.074425377472932e-06, | |
| "loss": 2.4195, | |
| "step": 2460 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.070780285432746e-06, | |
| "loss": 2.3155, | |
| "step": 2465 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.067128765074842e-06, | |
| "loss": 2.3354, | |
| "step": 2470 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.06347082216547e-06, | |
| "loss": 2.3451, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.059806462481022e-06, | |
| "loss": 2.3193, | |
| "step": 2480 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.056135691808019e-06, | |
| "loss": 2.2897, | |
| "step": 2485 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.052458515943112e-06, | |
| "loss": 2.2659, | |
| "step": 2490 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.048774940693062e-06, | |
| "loss": 2.3363, | |
| "step": 2495 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.045084971874738e-06, | |
| "loss": 2.3887, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.041388615315102e-06, | |
| "loss": 2.3496, | |
| "step": 2505 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.037685876851211e-06, | |
| "loss": 2.3608, | |
| "step": 2510 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.033976762330189e-06, | |
| "loss": 2.2707, | |
| "step": 2515 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.030261277609235e-06, | |
| "loss": 2.2787, | |
| "step": 2520 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.026539428555609e-06, | |
| "loss": 2.2724, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.022811221046618e-06, | |
| "loss": 2.3896, | |
| "step": 2530 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.01907666096961e-06, | |
| "loss": 2.2743, | |
| "step": 2535 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.015335754221964e-06, | |
| "loss": 2.2859, | |
| "step": 2540 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.011588506711085e-06, | |
| "loss": 2.3019, | |
| "step": 2545 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.007834924354384e-06, | |
| "loss": 2.3834, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.004075013079284e-06, | |
| "loss": 2.3645, | |
| "step": 2555 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 9.000308778823196e-06, | |
| "loss": 2.3349, | |
| "step": 2560 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.996536227533519e-06, | |
| "loss": 2.2916, | |
| "step": 2565 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.992757365167625e-06, | |
| "loss": 2.323, | |
| "step": 2570 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.988972197692857e-06, | |
| "loss": 2.3197, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.985180731086505e-06, | |
| "loss": 2.4031, | |
| "step": 2580 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.98138297133582e-06, | |
| "loss": 2.3158, | |
| "step": 2585 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.977578924437976e-06, | |
| "loss": 2.3468, | |
| "step": 2590 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.973768596400085e-06, | |
| "loss": 2.2986, | |
| "step": 2595 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.969951993239177e-06, | |
| "loss": 2.3172, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.966129120982188e-06, | |
| "loss": 2.4022, | |
| "step": 2605 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.962299985665955e-06, | |
| "loss": 2.3028, | |
| "step": 2610 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.958464593337202e-06, | |
| "loss": 2.2815, | |
| "step": 2615 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.954622950052543e-06, | |
| "loss": 2.3018, | |
| "step": 2620 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.950775061878453e-06, | |
| "loss": 2.3578, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.946920934891274e-06, | |
| "loss": 2.3957, | |
| "step": 2630 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.943060575177197e-06, | |
| "loss": 2.3572, | |
| "step": 2635 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.939193988832261e-06, | |
| "loss": 2.2332, | |
| "step": 2640 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.93532118196233e-06, | |
| "loss": 2.2703, | |
| "step": 2645 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.931442160683094e-06, | |
| "loss": 2.342, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.92755693112006e-06, | |
| "loss": 2.3219, | |
| "step": 2655 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.923665499408535e-06, | |
| "loss": 2.4212, | |
| "step": 2660 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.91976787169362e-06, | |
| "loss": 2.317, | |
| "step": 2665 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.915864054130203e-06, | |
| "loss": 2.3502, | |
| "step": 2670 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.911954052882941e-06, | |
| "loss": 2.3477, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.908037874126263e-06, | |
| "loss": 2.3429, | |
| "step": 2680 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.904115524044349e-06, | |
| "loss": 2.3321, | |
| "step": 2685 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.900187008831124e-06, | |
| "loss": 2.2941, | |
| "step": 2690 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.896252334690251e-06, | |
| "loss": 2.3432, | |
| "step": 2695 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.892311507835118e-06, | |
| "loss": 2.3232, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.888364534488828e-06, | |
| "loss": 2.3243, | |
| "step": 2705 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.88441142088419e-06, | |
| "loss": 2.3276, | |
| "step": 2710 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.880452173263708e-06, | |
| "loss": 2.3487, | |
| "step": 2715 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.87648679787958e-06, | |
| "loss": 2.2962, | |
| "step": 2720 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.872515300993669e-06, | |
| "loss": 2.3847, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.868537688877516e-06, | |
| "loss": 2.366, | |
| "step": 2730 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.86455396781231e-06, | |
| "loss": 2.3688, | |
| "step": 2735 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.860564144088891e-06, | |
| "loss": 2.3808, | |
| "step": 2740 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.856568224007736e-06, | |
| "loss": 2.288, | |
| "step": 2745 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.852566213878947e-06, | |
| "loss": 2.3828, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.848558120022246e-06, | |
| "loss": 2.303, | |
| "step": 2755 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.844543948766958e-06, | |
| "loss": 2.3357, | |
| "step": 2760 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.84052370645201e-06, | |
| "loss": 2.3596, | |
| "step": 2765 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.83649739942591e-06, | |
| "loss": 2.4066, | |
| "step": 2770 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.83246503404675e-06, | |
| "loss": 2.308, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.828426616682184e-06, | |
| "loss": 2.3756, | |
| "step": 2780 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.824382153709423e-06, | |
| "loss": 2.3836, | |
| "step": 2785 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.820331651515226e-06, | |
| "loss": 2.2858, | |
| "step": 2790 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.816275116495891e-06, | |
| "loss": 2.2355, | |
| "step": 2795 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.81221255505724e-06, | |
| "loss": 2.3715, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.808143973614612e-06, | |
| "loss": 2.3173, | |
| "step": 2805 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.80406937859285e-06, | |
| "loss": 2.2888, | |
| "step": 2810 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.799988776426298e-06, | |
| "loss": 2.3617, | |
| "step": 2815 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.795902173558784e-06, | |
| "loss": 2.3934, | |
| "step": 2820 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.791809576443611e-06, | |
| "loss": 2.413, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.787710991543547e-06, | |
| "loss": 2.3531, | |
| "step": 2830 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.78360642533082e-06, | |
| "loss": 2.3705, | |
| "step": 2835 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.779495884287099e-06, | |
| "loss": 2.2574, | |
| "step": 2840 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.775379374903487e-06, | |
| "loss": 2.3226, | |
| "step": 2845 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.77125690368052e-06, | |
| "loss": 2.2892, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.767128477128138e-06, | |
| "loss": 2.3075, | |
| "step": 2855 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.76299410176569e-06, | |
| "loss": 2.3077, | |
| "step": 2860 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.758853784121921e-06, | |
| "loss": 2.3228, | |
| "step": 2865 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.754707530734958e-06, | |
| "loss": 2.3572, | |
| "step": 2870 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.750555348152299e-06, | |
| "loss": 2.3425, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.74639724293081e-06, | |
| "loss": 2.3524, | |
| "step": 2880 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.7422332216367e-06, | |
| "loss": 2.2912, | |
| "step": 2885 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.738063290845536e-06, | |
| "loss": 2.3643, | |
| "step": 2890 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.733887457142202e-06, | |
| "loss": 2.3362, | |
| "step": 2895 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.729705727120911e-06, | |
| "loss": 2.3051, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.725518107385188e-06, | |
| "loss": 2.3051, | |
| "step": 2905 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.721324604547851e-06, | |
| "loss": 2.2844, | |
| "step": 2910 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.717125225231018e-06, | |
| "loss": 2.3237, | |
| "step": 2915 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.712919976066078e-06, | |
| "loss": 2.3555, | |
| "step": 2920 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.708708863693696e-06, | |
| "loss": 2.3721, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.704491894763794e-06, | |
| "loss": 2.315, | |
| "step": 2930 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.700269075935542e-06, | |
| "loss": 2.2967, | |
| "step": 2935 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.696040413877344e-06, | |
| "loss": 2.3054, | |
| "step": 2940 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.691805915266836e-06, | |
| "loss": 2.307, | |
| "step": 2945 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.68756558679087e-06, | |
| "loss": 2.2983, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.683319435145503e-06, | |
| "loss": 2.3344, | |
| "step": 2955 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.679067467035989e-06, | |
| "loss": 2.3349, | |
| "step": 2960 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.674809689176765e-06, | |
| "loss": 2.3366, | |
| "step": 2965 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.670546108291443e-06, | |
| "loss": 2.2578, | |
| "step": 2970 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.666276731112802e-06, | |
| "loss": 2.2764, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.662001564382768e-06, | |
| "loss": 2.2654, | |
| "step": 2980 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.657720614852412e-06, | |
| "loss": 2.3795, | |
| "step": 2985 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.65343388928194e-06, | |
| "loss": 2.3317, | |
| "step": 2990 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.649141394440677e-06, | |
| "loss": 2.3353, | |
| "step": 2995 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.644843137107058e-06, | |
| "loss": 2.275, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.640539124068617e-06, | |
| "loss": 2.3396, | |
| "step": 3005 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.636229362121979e-06, | |
| "loss": 2.2529, | |
| "step": 3010 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.631913858072846e-06, | |
| "loss": 2.2705, | |
| "step": 3015 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.627592618735989e-06, | |
| "loss": 2.3156, | |
| "step": 3020 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.623265650935233e-06, | |
| "loss": 2.253, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.618932961503452e-06, | |
| "loss": 2.4288, | |
| "step": 3030 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.614594557282553e-06, | |
| "loss": 2.3461, | |
| "step": 3035 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.610250445123472e-06, | |
| "loss": 2.3063, | |
| "step": 3040 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.605900631886148e-06, | |
| "loss": 2.2712, | |
| "step": 3045 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.601545124439535e-06, | |
| "loss": 2.2447, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.597183929661573e-06, | |
| "loss": 2.344, | |
| "step": 3055 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.592817054439184e-06, | |
| "loss": 2.3218, | |
| "step": 3060 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.588444505668259e-06, | |
| "loss": 2.3292, | |
| "step": 3065 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.584066290253649e-06, | |
| "loss": 2.2895, | |
| "step": 3070 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.579682415109156e-06, | |
| "loss": 2.3987, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.575292887157515e-06, | |
| "loss": 2.3424, | |
| "step": 3080 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.570897713330392e-06, | |
| "loss": 2.3428, | |
| "step": 3085 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.566496900568364e-06, | |
| "loss": 2.335, | |
| "step": 3090 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.562090455820918e-06, | |
| "loss": 2.3462, | |
| "step": 3095 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.557678386046429e-06, | |
| "loss": 2.2932, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.553260698212156e-06, | |
| "loss": 2.2747, | |
| "step": 3105 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.548837399294235e-06, | |
| "loss": 2.3526, | |
| "step": 3110 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.544408496277657e-06, | |
| "loss": 2.3823, | |
| "step": 3115 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.539973996156265e-06, | |
| "loss": 2.3555, | |
| "step": 3120 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.535533905932739e-06, | |
| "loss": 2.3478, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.531088232618587e-06, | |
| "loss": 2.2629, | |
| "step": 3130 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.526636983234135e-06, | |
| "loss": 2.343, | |
| "step": 3135 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.522180164808515e-06, | |
| "loss": 2.2924, | |
| "step": 3140 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.51771778437965e-06, | |
| "loss": 2.2119, | |
| "step": 3145 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.513249848994248e-06, | |
| "loss": 2.2122, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.508776365707788e-06, | |
| "loss": 2.337, | |
| "step": 3155 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.504297341584509e-06, | |
| "loss": 2.3789, | |
| "step": 3160 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.499812783697406e-06, | |
| "loss": 2.3896, | |
| "step": 3165 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.495322699128206e-06, | |
| "loss": 2.3503, | |
| "step": 3170 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.490827094967364e-06, | |
| "loss": 2.3193, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.486325978314054e-06, | |
| "loss": 2.3593, | |
| "step": 3180 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.481819356276155e-06, | |
| "loss": 2.3143, | |
| "step": 3185 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.477307235970235e-06, | |
| "loss": 2.2559, | |
| "step": 3190 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.472789624521552e-06, | |
| "loss": 2.3604, | |
| "step": 3195 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.468266529064025e-06, | |
| "loss": 2.3362, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.463737956740246e-06, | |
| "loss": 2.3185, | |
| "step": 3205 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.459203914701444e-06, | |
| "loss": 2.3538, | |
| "step": 3210 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.454664410107492e-06, | |
| "loss": 2.3653, | |
| "step": 3215 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.450119450126889e-06, | |
| "loss": 2.2836, | |
| "step": 3220 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.445569041936743e-06, | |
| "loss": 2.2986, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.441013192722774e-06, | |
| "loss": 2.2899, | |
| "step": 3230 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.436451909679286e-06, | |
| "loss": 2.3433, | |
| "step": 3235 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.431885200009172e-06, | |
| "loss": 2.3198, | |
| "step": 3240 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.427313070923885e-06, | |
| "loss": 2.3295, | |
| "step": 3245 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.422735529643445e-06, | |
| "loss": 2.2613, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.418152583396411e-06, | |
| "loss": 2.3224, | |
| "step": 3255 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.413564239419883e-06, | |
| "loss": 2.3025, | |
| "step": 3260 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.40897050495948e-06, | |
| "loss": 2.3282, | |
| "step": 3265 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.404371387269341e-06, | |
| "loss": 2.3497, | |
| "step": 3270 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.399766893612096e-06, | |
| "loss": 2.2694, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.39515703125887e-06, | |
| "loss": 2.3458, | |
| "step": 3280 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.390541807489266e-06, | |
| "loss": 2.3306, | |
| "step": 3285 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.385921229591351e-06, | |
| "loss": 2.2168, | |
| "step": 3290 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.381295304861647e-06, | |
| "loss": 2.2668, | |
| "step": 3295 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.376664040605122e-06, | |
| "loss": 2.3674, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.372027444135176e-06, | |
| "loss": 2.3003, | |
| "step": 3305 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.367385522773625e-06, | |
| "loss": 2.3031, | |
| "step": 3310 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.3627382838507e-06, | |
| "loss": 2.3332, | |
| "step": 3315 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.358085734705021e-06, | |
| "loss": 2.3959, | |
| "step": 3320 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.353427882683601e-06, | |
| "loss": 2.3392, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.348764735141823e-06, | |
| "loss": 2.3216, | |
| "step": 3330 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.344096299443434e-06, | |
| "loss": 2.3151, | |
| "step": 3335 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.339422582960533e-06, | |
| "loss": 2.3178, | |
| "step": 3340 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.334743593073553e-06, | |
| "loss": 2.3906, | |
| "step": 3345 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.33005933717126e-06, | |
| "loss": 2.2768, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.325369822650731e-06, | |
| "loss": 2.3562, | |
| "step": 3355 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.320675056917353e-06, | |
| "loss": 2.2956, | |
| "step": 3360 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.315975047384798e-06, | |
| "loss": 2.3581, | |
| "step": 3365 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.311269801475026e-06, | |
| "loss": 2.339, | |
| "step": 3370 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.30655932661826e-06, | |
| "loss": 2.3081, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.301843630252986e-06, | |
| "loss": 2.347, | |
| "step": 3380 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.297122719825928e-06, | |
| "loss": 2.3676, | |
| "step": 3385 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.29239660279205e-06, | |
| "loss": 2.3438, | |
| "step": 3390 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.287665286614538e-06, | |
| "loss": 2.3044, | |
| "step": 3395 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.282928778764783e-06, | |
| "loss": 2.2677, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.278187086722378e-06, | |
| "loss": 2.287, | |
| "step": 3405 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.273440217975103e-06, | |
| "loss": 2.3155, | |
| "step": 3410 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.268688180018911e-06, | |
| "loss": 2.3335, | |
| "step": 3415 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.26393098035792e-06, | |
| "loss": 2.3436, | |
| "step": 3420 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.259168626504395e-06, | |
| "loss": 2.2565, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.254401125978744e-06, | |
| "loss": 2.3239, | |
| "step": 3430 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.2496284863095e-06, | |
| "loss": 2.2819, | |
| "step": 3435 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.244850715033316e-06, | |
| "loss": 2.336, | |
| "step": 3440 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.240067819694941e-06, | |
| "loss": 2.2427, | |
| "step": 3445 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.235279807847223e-06, | |
| "loss": 2.3044, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.230486687051082e-06, | |
| "loss": 2.3128, | |
| "step": 3455 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.225688464875514e-06, | |
| "loss": 2.2754, | |
| "step": 3460 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.220885148897566e-06, | |
| "loss": 2.2479, | |
| "step": 3465 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.216076746702327e-06, | |
| "loss": 2.3225, | |
| "step": 3470 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.211263265882923e-06, | |
| "loss": 2.3474, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.206444714040496e-06, | |
| "loss": 2.2638, | |
| "step": 3480 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.201621098784198e-06, | |
| "loss": 2.282, | |
| "step": 3485 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.196792427731175e-06, | |
| "loss": 2.3049, | |
| "step": 3490 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.191958708506557e-06, | |
| "loss": 2.2981, | |
| "step": 3495 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.18711994874345e-06, | |
| "loss": 2.3496, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.182276156082911e-06, | |
| "loss": 2.3574, | |
| "step": 3505 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.177427338173955e-06, | |
| "loss": 2.2538, | |
| "step": 3510 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.172573502673522e-06, | |
| "loss": 2.3052, | |
| "step": 3515 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.167714657246486e-06, | |
| "loss": 2.3976, | |
| "step": 3520 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.162850809565623e-06, | |
| "loss": 2.2584, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.157981967311614e-06, | |
| "loss": 2.3171, | |
| "step": 3530 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.153108138173027e-06, | |
| "loss": 2.2602, | |
| "step": 3535 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.1482293298463e-06, | |
| "loss": 2.2461, | |
| "step": 3540 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.143345550035742e-06, | |
| "loss": 2.2912, | |
| "step": 3545 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.138456806453503e-06, | |
| "loss": 2.33, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.133563106819579e-06, | |
| "loss": 2.3271, | |
| "step": 3555 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.12866445886179e-06, | |
| "loss": 2.3225, | |
| "step": 3560 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.12376087031577e-06, | |
| "loss": 2.4378, | |
| "step": 3565 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.118852348924951e-06, | |
| "loss": 2.2822, | |
| "step": 3570 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.113938902440563e-06, | |
| "loss": 2.2497, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.109020538621607e-06, | |
| "loss": 2.3472, | |
| "step": 3580 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.10409726523485e-06, | |
| "loss": 2.3614, | |
| "step": 3585 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.099169090054812e-06, | |
| "loss": 2.3647, | |
| "step": 3590 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.094236020863758e-06, | |
| "loss": 2.324, | |
| "step": 3595 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.089298065451673e-06, | |
| "loss": 2.3319, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.084355231616266e-06, | |
| "loss": 2.2663, | |
| "step": 3605 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.079407527162944e-06, | |
| "loss": 2.319, | |
| "step": 3610 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.074454959904807e-06, | |
| "loss": 2.3662, | |
| "step": 3615 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.069497537662638e-06, | |
| "loss": 2.2951, | |
| "step": 3620 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.064535268264883e-06, | |
| "loss": 2.2674, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.05956815954764e-06, | |
| "loss": 2.2552, | |
| "step": 3630 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.054596219354655e-06, | |
| "loss": 2.3214, | |
| "step": 3635 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.049619455537296e-06, | |
| "loss": 2.3588, | |
| "step": 3640 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.044637875954556e-06, | |
| "loss": 2.2673, | |
| "step": 3645 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.039651488473028e-06, | |
| "loss": 2.32, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.034660300966898e-06, | |
| "loss": 2.2775, | |
| "step": 3655 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.029664321317932e-06, | |
| "loss": 2.2553, | |
| "step": 3660 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.024663557415466e-06, | |
| "loss": 2.3138, | |
| "step": 3665 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.019658017156384e-06, | |
| "loss": 2.272, | |
| "step": 3670 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.014647708445124e-06, | |
| "loss": 2.2805, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.009632639193643e-06, | |
| "loss": 2.3127, | |
| "step": 3680 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 8.004612817321419e-06, | |
| "loss": 2.3578, | |
| "step": 3685 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.999588250755442e-06, | |
| "loss": 2.3113, | |
| "step": 3690 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.99455894743018e-06, | |
| "loss": 2.3345, | |
| "step": 3695 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.989524915287595e-06, | |
| "loss": 2.3073, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.98448616227711e-06, | |
| "loss": 2.3059, | |
| "step": 3705 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.979442696355601e-06, | |
| "loss": 2.3148, | |
| "step": 3710 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.974394525487395e-06, | |
| "loss": 2.3477, | |
| "step": 3715 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.969341657644236e-06, | |
| "loss": 2.2966, | |
| "step": 3720 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.964284100805297e-06, | |
| "loss": 2.302, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.959221862957149e-06, | |
| "loss": 2.2996, | |
| "step": 3730 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.954154952093754e-06, | |
| "loss": 2.2492, | |
| "step": 3735 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.94908337621646e-06, | |
| "loss": 2.3366, | |
| "step": 3740 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.944007143333976e-06, | |
| "loss": 2.3723, | |
| "step": 3745 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.938926261462366e-06, | |
| "loss": 2.3513, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.933840738625035e-06, | |
| "loss": 2.2887, | |
| "step": 3755 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.928750582852722e-06, | |
| "loss": 2.3003, | |
| "step": 3760 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.923655802183475e-06, | |
| "loss": 2.3103, | |
| "step": 3765 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.918556404662645e-06, | |
| "loss": 2.2769, | |
| "step": 3770 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.913452398342882e-06, | |
| "loss": 2.3517, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.908343791284104e-06, | |
| "loss": 2.3953, | |
| "step": 3780 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.903230591553504e-06, | |
| "loss": 2.2957, | |
| "step": 3785 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.898112807225517e-06, | |
| "loss": 2.3143, | |
| "step": 3790 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.892990446381828e-06, | |
| "loss": 2.3442, | |
| "step": 3795 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.887863517111337e-06, | |
| "loss": 2.3237, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.882732027510174e-06, | |
| "loss": 2.3686, | |
| "step": 3805 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.877595985681656e-06, | |
| "loss": 2.3272, | |
| "step": 3810 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.872455399736295e-06, | |
| "loss": 2.2526, | |
| "step": 3815 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.867310277791778e-06, | |
| "loss": 2.2496, | |
| "step": 3820 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.862160627972956e-06, | |
| "loss": 2.3442, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.857006458411826e-06, | |
| "loss": 2.3039, | |
| "step": 3830 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.851847777247528e-06, | |
| "loss": 2.3293, | |
| "step": 3835 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.846684592626324e-06, | |
| "loss": 2.3605, | |
| "step": 3840 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.841516912701585e-06, | |
| "loss": 2.286, | |
| "step": 3845 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.836344745633785e-06, | |
| "loss": 2.3416, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.831168099590478e-06, | |
| "loss": 2.3096, | |
| "step": 3855 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.8259869827463e-06, | |
| "loss": 2.3253, | |
| "step": 3860 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.82080140328294e-06, | |
| "loss": 2.2856, | |
| "step": 3865 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.815611369389134e-06, | |
| "loss": 2.2835, | |
| "step": 3870 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.810416889260653e-06, | |
| "loss": 2.2696, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.805217971100295e-06, | |
| "loss": 2.3515, | |
| "step": 3880 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.800014623117858e-06, | |
| "loss": 2.3351, | |
| "step": 3885 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.794806853530139e-06, | |
| "loss": 2.3044, | |
| "step": 3890 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.789594670560917e-06, | |
| "loss": 2.329, | |
| "step": 3895 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.78437808244094e-06, | |
| "loss": 2.3173, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.779157097407914e-06, | |
| "loss": 2.2906, | |
| "step": 3905 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.773931723706487e-06, | |
| "loss": 2.3314, | |
| "step": 3910 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.768701969588237e-06, | |
| "loss": 2.2308, | |
| "step": 3915 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.763467843311658e-06, | |
| "loss": 2.3625, | |
| "step": 3920 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.758229353142153e-06, | |
| "loss": 2.3067, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.752986507352009e-06, | |
| "loss": 2.3529, | |
| "step": 3930 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.747739314220398e-06, | |
| "loss": 2.3094, | |
| "step": 3935 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.742487782033352e-06, | |
| "loss": 2.337, | |
| "step": 3940 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.737231919083761e-06, | |
| "loss": 2.3259, | |
| "step": 3945 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.731971733671347e-06, | |
| "loss": 2.3052, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.72670723410266e-06, | |
| "loss": 2.3158, | |
| "step": 3955 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.721438428691065e-06, | |
| "loss": 2.3282, | |
| "step": 3960 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.716165325756727e-06, | |
| "loss": 2.3027, | |
| "step": 3965 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.71088793362659e-06, | |
| "loss": 2.3143, | |
| "step": 3970 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.70560626063438e-06, | |
| "loss": 2.3918, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.70032031512058e-06, | |
| "loss": 2.2726, | |
| "step": 3980 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.695030105432417e-06, | |
| "loss": 2.2606, | |
| "step": 3985 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.689735639923857e-06, | |
| "loss": 2.3089, | |
| "step": 3990 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.684436926955584e-06, | |
| "loss": 2.2965, | |
| "step": 3995 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.679133974894984e-06, | |
| "loss": 2.2936, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.673826792116146e-06, | |
| "loss": 2.2601, | |
| "step": 4005 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.668515386999837e-06, | |
| "loss": 2.3488, | |
| "step": 4010 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.663199767933489e-06, | |
| "loss": 2.3245, | |
| "step": 4015 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.65787994331119e-06, | |
| "loss": 2.2646, | |
| "step": 4020 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.652555921533671e-06, | |
| "loss": 2.4197, | |
| "step": 4025 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.647227711008288e-06, | |
| "loss": 2.3521, | |
| "step": 4030 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.641895320149008e-06, | |
| "loss": 2.3344, | |
| "step": 4035 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.636558757376413e-06, | |
| "loss": 2.2585, | |
| "step": 4040 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.631218031117658e-06, | |
| "loss": 2.2581, | |
| "step": 4045 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.6258731498064796e-06, | |
| "loss": 2.2796, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.620524121883175e-06, | |
| "loss": 2.2735, | |
| "step": 4055 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.615170955794592e-06, | |
| "loss": 2.2992, | |
| "step": 4060 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.609813659994107e-06, | |
| "loss": 2.2521, | |
| "step": 4065 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.604452242941622e-06, | |
| "loss": 2.3223, | |
| "step": 4070 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.5990867131035474e-06, | |
| "loss": 2.3207, | |
| "step": 4075 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.593717078952788e-06, | |
| "loss": 2.3283, | |
| "step": 4080 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.588343348968728e-06, | |
| "loss": 2.2824, | |
| "step": 4085 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.582965531637221e-06, | |
| "loss": 2.2998, | |
| "step": 4090 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.577583635450572e-06, | |
| "loss": 2.2794, | |
| "step": 4095 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.572197668907533e-06, | |
| "loss": 2.2927, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.566807640513278e-06, | |
| "loss": 2.3367, | |
| "step": 4105 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.561413558779401e-06, | |
| "loss": 2.3393, | |
| "step": 4110 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.55601543222389e-06, | |
| "loss": 2.2557, | |
| "step": 4115 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.550613269371124e-06, | |
| "loss": 2.2634, | |
| "step": 4120 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.545207078751858e-06, | |
| "loss": 2.2909, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.5397968689032e-06, | |
| "loss": 2.3116, | |
| "step": 4130 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.534382648368617e-06, | |
| "loss": 2.2445, | |
| "step": 4135 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.528964425697895e-06, | |
| "loss": 2.3044, | |
| "step": 4140 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.523542209447152e-06, | |
| "loss": 2.2561, | |
| "step": 4145 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.518116008178805e-06, | |
| "loss": 2.3586, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.512685830461568e-06, | |
| "loss": 2.3003, | |
| "step": 4155 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.507251684870433e-06, | |
| "loss": 2.3108, | |
| "step": 4160 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.501813579986655e-06, | |
| "loss": 2.3008, | |
| "step": 4165 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.496371524397747e-06, | |
| "loss": 2.2967, | |
| "step": 4170 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.490925526697455e-06, | |
| "loss": 2.2927, | |
| "step": 4175 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.485475595485756e-06, | |
| "loss": 2.3512, | |
| "step": 4180 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.480021739368831e-06, | |
| "loss": 2.3197, | |
| "step": 4185 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.474563966959068e-06, | |
| "loss": 2.2785, | |
| "step": 4190 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.469102286875029e-06, | |
| "loss": 2.3029, | |
| "step": 4195 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.463636707741458e-06, | |
| "loss": 2.3051, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.458167238189249e-06, | |
| "loss": 2.3209, | |
| "step": 4205 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.452693886855438e-06, | |
| "loss": 2.2582, | |
| "step": 4210 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.447216662383196e-06, | |
| "loss": 2.3513, | |
| "step": 4215 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.4417355734218085e-06, | |
| "loss": 2.3232, | |
| "step": 4220 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.436250628626662e-06, | |
| "loss": 2.3386, | |
| "step": 4225 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.430761836659235e-06, | |
| "loss": 2.3386, | |
| "step": 4230 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.425269206187076e-06, | |
| "loss": 2.2413, | |
| "step": 4235 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.4197727458837995e-06, | |
| "loss": 2.2431, | |
| "step": 4240 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.414272464429068e-06, | |
| "loss": 2.2852, | |
| "step": 4245 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.408768370508577e-06, | |
| "loss": 2.2154, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.40326047281404e-06, | |
| "loss": 2.2376, | |
| "step": 4255 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.397748780043179e-06, | |
| "loss": 2.3474, | |
| "step": 4260 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.392233300899712e-06, | |
| "loss": 2.3257, | |
| "step": 4265 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.386714044093331e-06, | |
| "loss": 2.3226, | |
| "step": 4270 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.381191018339697e-06, | |
| "loss": 2.3102, | |
| "step": 4275 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.375664232360421e-06, | |
| "loss": 2.2428, | |
| "step": 4280 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.370133694883052e-06, | |
| "loss": 2.3311, | |
| "step": 4285 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.364599414641064e-06, | |
| "loss": 2.2936, | |
| "step": 4290 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.359061400373841e-06, | |
| "loss": 2.2807, | |
| "step": 4295 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.353519660826665e-06, | |
| "loss": 2.3075, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.3479742047506955e-06, | |
| "loss": 2.2143, | |
| "step": 4305 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.342425040902967e-06, | |
| "loss": 2.3901, | |
| "step": 4310 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.336872178046368e-06, | |
| "loss": 2.2464, | |
| "step": 4315 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.331315624949624e-06, | |
| "loss": 2.2425, | |
| "step": 4320 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.325755390387293e-06, | |
| "loss": 2.321, | |
| "step": 4325 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.320191483139742e-06, | |
| "loss": 2.2928, | |
| "step": 4330 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.314623911993143e-06, | |
| "loss": 2.2613, | |
| "step": 4335 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.309052685739448e-06, | |
| "loss": 2.2948, | |
| "step": 4340 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.303477813176385e-06, | |
| "loss": 2.2389, | |
| "step": 4345 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.297899303107441e-06, | |
| "loss": 2.2945, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.29231716434184e-06, | |
| "loss": 2.2881, | |
| "step": 4355 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.286731405694544e-06, | |
| "loss": 2.2987, | |
| "step": 4360 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.281142035986227e-06, | |
| "loss": 2.2413, | |
| "step": 4365 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.275549064043269e-06, | |
| "loss": 2.1828, | |
| "step": 4370 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.269952498697734e-06, | |
| "loss": 2.2939, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.264352348787364e-06, | |
| "loss": 2.2431, | |
| "step": 4380 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.258748623155558e-06, | |
| "loss": 2.2975, | |
| "step": 4385 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.253141330651367e-06, | |
| "loss": 2.2863, | |
| "step": 4390 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.247530480129469e-06, | |
| "loss": 2.3238, | |
| "step": 4395 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.241916080450163e-06, | |
| "loss": 2.2852, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.236298140479352e-06, | |
| "loss": 2.2699, | |
| "step": 4405 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.23067666908853e-06, | |
| "loss": 2.3141, | |
| "step": 4410 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.225051675154768e-06, | |
| "loss": 2.2845, | |
| "step": 4415 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.219423167560701e-06, | |
| "loss": 2.2736, | |
| "step": 4420 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.21379115519451e-06, | |
| "loss": 2.2909, | |
| "step": 4425 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.208155646949908e-06, | |
| "loss": 2.3407, | |
| "step": 4430 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.202516651726135e-06, | |
| "loss": 2.2754, | |
| "step": 4435 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.196874178427933e-06, | |
| "loss": 2.3012, | |
| "step": 4440 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.191228235965539e-06, | |
| "loss": 2.2588, | |
| "step": 4445 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.185578833254665e-06, | |
| "loss": 2.2662, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.1799259792164914e-06, | |
| "loss": 2.2842, | |
| "step": 4455 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.1742696827776415e-06, | |
| "loss": 2.3201, | |
| "step": 4460 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.168609952870185e-06, | |
| "loss": 2.2799, | |
| "step": 4465 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.162946798431605e-06, | |
| "loss": 2.3022, | |
| "step": 4470 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.157280228404796e-06, | |
| "loss": 2.2339, | |
| "step": 4475 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.151610251738045e-06, | |
| "loss": 2.3019, | |
| "step": 4480 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.145936877385019e-06, | |
| "loss": 2.33, | |
| "step": 4485 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.1402601143047514e-06, | |
| "loss": 2.3497, | |
| "step": 4490 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.134579971461627e-06, | |
| "loss": 2.2592, | |
| "step": 4495 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.128896457825364e-06, | |
| "loss": 2.278, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.1232095823710064e-06, | |
| "loss": 2.1947, | |
| "step": 4505 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.11751935407891e-06, | |
| "loss": 2.2497, | |
| "step": 4510 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.111825781934719e-06, | |
| "loss": 2.2872, | |
| "step": 4515 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.106128874929364e-06, | |
| "loss": 2.2975, | |
| "step": 4520 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.100428642059033e-06, | |
| "loss": 2.3236, | |
| "step": 4525 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.094725092325177e-06, | |
| "loss": 2.3056, | |
| "step": 4530 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.089018234734476e-06, | |
| "loss": 2.2794, | |
| "step": 4535 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.08330807829884e-06, | |
| "loss": 2.2758, | |
| "step": 4540 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.077594632035385e-06, | |
| "loss": 2.283, | |
| "step": 4545 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.071877904966422e-06, | |
| "loss": 2.3237, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.066157906119442e-06, | |
| "loss": 2.3569, | |
| "step": 4555 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.060434644527105e-06, | |
| "loss": 2.3726, | |
| "step": 4560 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 7.054708129227225e-06, | |
| "loss": 2.3648, | |
| "step": 4565 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 7.048978369262747e-06, | |
| "loss": 2.2151, | |
| "step": 4570 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 7.043245373681746e-06, | |
| "loss": 2.2749, | |
| "step": 4575 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 7.037509151537404e-06, | |
| "loss": 2.3678, | |
| "step": 4580 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 7.031769711887999e-06, | |
| "loss": 2.3461, | |
| "step": 4585 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 7.026027063796891e-06, | |
| "loss": 2.2884, | |
| "step": 4590 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 7.020281216332503e-06, | |
| "loss": 2.2799, | |
| "step": 4595 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 7.014532178568314e-06, | |
| "loss": 2.3168, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 7.008779959582838e-06, | |
| "loss": 2.2814, | |
| "step": 4605 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 7.003024568459614e-06, | |
| "loss": 2.3394, | |
| "step": 4610 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 6.997266014287193e-06, | |
| "loss": 2.2883, | |
| "step": 4615 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 6.991504306159115e-06, | |
| "loss": 2.2861, | |
| "step": 4620 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 6.985739453173903e-06, | |
| "loss": 2.3097, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 6.9799714644350504e-06, | |
| "loss": 2.2177, | |
| "step": 4630 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 6.974200349050996e-06, | |
| "loss": 2.3081, | |
| "step": 4635 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 6.968426116135118e-06, | |
| "loss": 2.2604, | |
| "step": 4640 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 6.96264877480572e-06, | |
| "loss": 2.2846, | |
| "step": 4645 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 6.9568683341860135e-06, | |
| "loss": 2.2789, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 6.9510848034041e-06, | |
| "loss": 2.2993, | |
| "step": 4655 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 6.945298191592967e-06, | |
| "loss": 2.2946, | |
| "step": 4660 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 6.939508507890464e-06, | |
| "loss": 2.2999, | |
| "step": 4665 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 6.93371576143929e-06, | |
| "loss": 2.3512, | |
| "step": 4670 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 6.927919961386984e-06, | |
| "loss": 2.274, | |
| "step": 4675 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 6.922121116885905e-06, | |
| "loss": 2.29, | |
| "step": 4680 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 6.916319237093219e-06, | |
| "loss": 2.3253, | |
| "step": 4685 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.910514331170888e-06, | |
| "loss": 2.3136, | |
| "step": 4690 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.904706408285649e-06, | |
| "loss": 2.1658, | |
| "step": 4695 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.898895477609007e-06, | |
| "loss": 2.327, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.893081548317212e-06, | |
| "loss": 2.2537, | |
| "step": 4705 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.887264629591254e-06, | |
| "loss": 2.2299, | |
| "step": 4710 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.881444730616842e-06, | |
| "loss": 2.2907, | |
| "step": 4715 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.875621860584389e-06, | |
| "loss": 2.1995, | |
| "step": 4720 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.869796028689002e-06, | |
| "loss": 2.3364, | |
| "step": 4725 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.863967244130467e-06, | |
| "loss": 2.2934, | |
| "step": 4730 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.858135516113226e-06, | |
| "loss": 2.2679, | |
| "step": 4735 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.852300853846381e-06, | |
| "loss": 2.2466, | |
| "step": 4740 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.846463266543653e-06, | |
| "loss": 2.2422, | |
| "step": 4745 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.840622763423391e-06, | |
| "loss": 2.2817, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.8347793537085474e-06, | |
| "loss": 2.3266, | |
| "step": 4755 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.8289330466266635e-06, | |
| "loss": 2.3688, | |
| "step": 4760 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.823083851409857e-06, | |
| "loss": 2.3133, | |
| "step": 4765 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.817231777294804e-06, | |
| "loss": 2.2128, | |
| "step": 4770 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.811376833522729e-06, | |
| "loss": 2.3112, | |
| "step": 4775 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.805519029339388e-06, | |
| "loss": 2.2092, | |
| "step": 4780 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.799658373995054e-06, | |
| "loss": 2.3035, | |
| "step": 4785 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.793794876744499e-06, | |
| "loss": 2.3186, | |
| "step": 4790 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.787928546846987e-06, | |
| "loss": 2.3321, | |
| "step": 4795 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.782059393566254e-06, | |
| "loss": 2.2435, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.776187426170494e-06, | |
| "loss": 2.3579, | |
| "step": 4805 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.770312653932346e-06, | |
| "loss": 2.3404, | |
| "step": 4810 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.764435086128876e-06, | |
| "loss": 2.2645, | |
| "step": 4815 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.758554732041564e-06, | |
| "loss": 2.2836, | |
| "step": 4820 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.752671600956295e-06, | |
| "loss": 2.2955, | |
| "step": 4825 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.7467857021633354e-06, | |
| "loss": 2.2942, | |
| "step": 4830 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.740897044957322e-06, | |
| "loss": 2.2712, | |
| "step": 4835 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.7350056386372485e-06, | |
| "loss": 2.2603, | |
| "step": 4840 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.72911149250645e-06, | |
| "loss": 2.305, | |
| "step": 4845 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.723214615872585e-06, | |
| "loss": 2.3159, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.717315018047631e-06, | |
| "loss": 2.3539, | |
| "step": 4855 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.711412708347857e-06, | |
| "loss": 2.2216, | |
| "step": 4860 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.7055076960938135e-06, | |
| "loss": 2.3397, | |
| "step": 4865 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.699599990610324e-06, | |
| "loss": 2.2257, | |
| "step": 4870 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.693689601226458e-06, | |
| "loss": 2.3418, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.68777653727553e-06, | |
| "loss": 2.2535, | |
| "step": 4880 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.681860808095074e-06, | |
| "loss": 2.2368, | |
| "step": 4885 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.675942423026834e-06, | |
| "loss": 2.2917, | |
| "step": 4890 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.6700213914167485e-06, | |
| "loss": 2.2268, | |
| "step": 4895 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.664097722614934e-06, | |
| "loss": 2.2911, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.658171425975673e-06, | |
| "loss": 2.3062, | |
| "step": 4905 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.652242510857395e-06, | |
| "loss": 2.2728, | |
| "step": 4910 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.6463109866226675e-06, | |
| "loss": 2.3351, | |
| "step": 4915 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.640376862638176e-06, | |
| "loss": 2.3026, | |
| "step": 4920 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.634440148274712e-06, | |
| "loss": 2.2661, | |
| "step": 4925 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.6285008529071615e-06, | |
| "loss": 2.3494, | |
| "step": 4930 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.622558985914478e-06, | |
| "loss": 2.2568, | |
| "step": 4935 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.616614556679684e-06, | |
| "loss": 2.3232, | |
| "step": 4940 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.610667574589841e-06, | |
| "loss": 2.2727, | |
| "step": 4945 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.604718049036047e-06, | |
| "loss": 2.3194, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.598765989413419e-06, | |
| "loss": 2.2589, | |
| "step": 4955 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.592811405121064e-06, | |
| "loss": 2.2509, | |
| "step": 4960 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.5868543055620895e-06, | |
| "loss": 2.2405, | |
| "step": 4965 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.580894700143565e-06, | |
| "loss": 2.2483, | |
| "step": 4970 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.574932598276524e-06, | |
| "loss": 2.3299, | |
| "step": 4975 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.568968009375938e-06, | |
| "loss": 2.2897, | |
| "step": 4980 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.5630009428607065e-06, | |
| "loss": 2.315, | |
| "step": 4985 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.557031408153642e-06, | |
| "loss": 2.2338, | |
| "step": 4990 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.551059414681455e-06, | |
| "loss": 2.3174, | |
| "step": 4995 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.545084971874738e-06, | |
| "loss": 2.289, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.539108089167953e-06, | |
| "loss": 2.2682, | |
| "step": 5005 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.533128775999411e-06, | |
| "loss": 2.3725, | |
| "step": 5010 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.527147041811266e-06, | |
| "loss": 2.2756, | |
| "step": 5015 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.521162896049491e-06, | |
| "loss": 2.3028, | |
| "step": 5020 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.5151763481638705e-06, | |
| "loss": 2.3266, | |
| "step": 5025 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.509187407607981e-06, | |
| "loss": 2.2829, | |
| "step": 5030 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.503196083839175e-06, | |
| "loss": 2.3261, | |
| "step": 5035 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.497202386318573e-06, | |
| "loss": 2.2781, | |
| "step": 5040 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.491206324511039e-06, | |
| "loss": 2.242, | |
| "step": 5045 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.485207907885175e-06, | |
| "loss": 2.3122, | |
| "step": 5050 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.4792071459133e-06, | |
| "loss": 2.2544, | |
| "step": 5055 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.473204048071433e-06, | |
| "loss": 2.2366, | |
| "step": 5060 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.467198623839288e-06, | |
| "loss": 2.2598, | |
| "step": 5065 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.4611908827002504e-06, | |
| "loss": 2.2665, | |
| "step": 5070 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.455180834141359e-06, | |
| "loss": 2.3059, | |
| "step": 5075 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.449168487653305e-06, | |
| "loss": 2.3231, | |
| "step": 5080 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.443153852730404e-06, | |
| "loss": 2.3079, | |
| "step": 5085 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.437136938870583e-06, | |
| "loss": 2.2658, | |
| "step": 5090 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.431117755575371e-06, | |
| "loss": 2.2425, | |
| "step": 5095 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.425096312349881e-06, | |
| "loss": 2.2776, | |
| "step": 5100 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.419072618702794e-06, | |
| "loss": 2.2867, | |
| "step": 5105 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.413046684146343e-06, | |
| "loss": 2.316, | |
| "step": 5110 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.407018518196303e-06, | |
| "loss": 2.2536, | |
| "step": 5115 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.400988130371969e-06, | |
| "loss": 2.2747, | |
| "step": 5120 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.3949555301961474e-06, | |
| "loss": 2.2994, | |
| "step": 5125 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.388920727195138e-06, | |
| "loss": 2.2803, | |
| "step": 5130 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.382883730898717e-06, | |
| "loss": 2.2863, | |
| "step": 5135 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.376844550840126e-06, | |
| "loss": 2.2517, | |
| "step": 5140 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.3708031965560545e-06, | |
| "loss": 2.2053, | |
| "step": 5145 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.364759677586627e-06, | |
| "loss": 2.3057, | |
| "step": 5150 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.3587140034753836e-06, | |
| "loss": 2.2284, | |
| "step": 5155 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.352666183769269e-06, | |
| "loss": 2.2562, | |
| "step": 5160 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.3466162280186164e-06, | |
| "loss": 2.267, | |
| "step": 5165 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.340564145777131e-06, | |
| "loss": 2.2915, | |
| "step": 5170 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.334509946601879e-06, | |
| "loss": 2.2691, | |
| "step": 5175 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.328453640053264e-06, | |
| "loss": 2.3175, | |
| "step": 5180 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.322395235695022e-06, | |
| "loss": 2.2416, | |
| "step": 5185 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.316334743094201e-06, | |
| "loss": 2.259, | |
| "step": 5190 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.310272171821145e-06, | |
| "loss": 2.2588, | |
| "step": 5195 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.304207531449486e-06, | |
| "loss": 2.211, | |
| "step": 5200 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.298140831556112e-06, | |
| "loss": 2.3189, | |
| "step": 5205 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.292072081721173e-06, | |
| "loss": 2.2851, | |
| "step": 5210 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.286001291528056e-06, | |
| "loss": 2.3112, | |
| "step": 5215 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.279928470563365e-06, | |
| "loss": 2.2273, | |
| "step": 5220 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.273853628416911e-06, | |
| "loss": 2.2543, | |
| "step": 5225 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.267776774681703e-06, | |
| "loss": 2.2217, | |
| "step": 5230 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.261697918953922e-06, | |
| "loss": 2.2516, | |
| "step": 5235 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.255617070832908e-06, | |
| "loss": 2.2491, | |
| "step": 5240 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.249534239921154e-06, | |
| "loss": 2.285, | |
| "step": 5245 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.243449435824276e-06, | |
| "loss": 2.2831, | |
| "step": 5250 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.237362668151013e-06, | |
| "loss": 2.2987, | |
| "step": 5255 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.231273946513201e-06, | |
| "loss": 2.3151, | |
| "step": 5260 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.225183280525763e-06, | |
| "loss": 2.2287, | |
| "step": 5265 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.219090679806694e-06, | |
| "loss": 2.2534, | |
| "step": 5270 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.212996153977038e-06, | |
| "loss": 2.2452, | |
| "step": 5275 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.206899712660887e-06, | |
| "loss": 2.2368, | |
| "step": 5280 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.2008013654853505e-06, | |
| "loss": 2.2477, | |
| "step": 5285 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.1947011220805535e-06, | |
| "loss": 2.3123, | |
| "step": 5290 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.188598992079613e-06, | |
| "loss": 2.1981, | |
| "step": 5295 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.182494985118625e-06, | |
| "loss": 2.2761, | |
| "step": 5300 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.176389110836647e-06, | |
| "loss": 2.2693, | |
| "step": 5305 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.170281378875692e-06, | |
| "loss": 2.2794, | |
| "step": 5310 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.1641717988807006e-06, | |
| "loss": 2.32, | |
| "step": 5315 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.158060380499533e-06, | |
| "loss": 2.373, | |
| "step": 5320 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.151947133382954e-06, | |
| "loss": 2.2373, | |
| "step": 5325 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.145832067184614e-06, | |
| "loss": 2.2928, | |
| "step": 5330 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.139715191561038e-06, | |
| "loss": 2.3444, | |
| "step": 5335 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.133596516171609e-06, | |
| "loss": 2.2909, | |
| "step": 5340 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.127476050678548e-06, | |
| "loss": 2.2982, | |
| "step": 5345 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.121353804746907e-06, | |
| "loss": 2.2556, | |
| "step": 5350 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.1152297880445476e-06, | |
| "loss": 2.3198, | |
| "step": 5355 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.109104010242127e-06, | |
| "loss": 2.2375, | |
| "step": 5360 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.102976481013086e-06, | |
| "loss": 2.2558, | |
| "step": 5365 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.09684721003363e-06, | |
| "loss": 2.2616, | |
| "step": 5370 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.090716206982714e-06, | |
| "loss": 2.2913, | |
| "step": 5375 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.084583481542028e-06, | |
| "loss": 2.2505, | |
| "step": 5380 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.078449043395982e-06, | |
| "loss": 2.3346, | |
| "step": 5385 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.072312902231692e-06, | |
| "loss": 2.2542, | |
| "step": 5390 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.066175067738964e-06, | |
| "loss": 2.2899, | |
| "step": 5395 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.060035549610275e-06, | |
| "loss": 2.3113, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.053894357540761e-06, | |
| "loss": 2.2564, | |
| "step": 5405 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.047751501228203e-06, | |
| "loss": 2.2299, | |
| "step": 5410 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.041606990373012e-06, | |
| "loss": 2.3287, | |
| "step": 5415 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.0354608346782075e-06, | |
| "loss": 2.2343, | |
| "step": 5420 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.029313043849407e-06, | |
| "loss": 2.2489, | |
| "step": 5425 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.023163627594813e-06, | |
| "loss": 2.2848, | |
| "step": 5430 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.0170125956251935e-06, | |
| "loss": 2.2716, | |
| "step": 5435 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 6.010859957653869e-06, | |
| "loss": 2.2821, | |
| "step": 5440 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 6.00470572339669e-06, | |
| "loss": 2.2096, | |
| "step": 5445 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.9985499025720354e-06, | |
| "loss": 2.2687, | |
| "step": 5450 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.992392504900786e-06, | |
| "loss": 2.2843, | |
| "step": 5455 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.986233540106315e-06, | |
| "loss": 2.224, | |
| "step": 5460 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.9800730179144665e-06, | |
| "loss": 2.3012, | |
| "step": 5465 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.973910948053545e-06, | |
| "loss": 2.2787, | |
| "step": 5470 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.967747340254303e-06, | |
| "loss": 2.2809, | |
| "step": 5475 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.961582204249915e-06, | |
| "loss": 2.2944, | |
| "step": 5480 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.955415549775975e-06, | |
| "loss": 2.2772, | |
| "step": 5485 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.949247386570471e-06, | |
| "loss": 2.2152, | |
| "step": 5490 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.9430777243737744e-06, | |
| "loss": 2.3105, | |
| "step": 5495 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.936906572928625e-06, | |
| "loss": 2.2854, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.930733941980111e-06, | |
| "loss": 2.2787, | |
| "step": 5505 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.924559841275661e-06, | |
| "loss": 2.2922, | |
| "step": 5510 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.918384280565025e-06, | |
| "loss": 2.3135, | |
| "step": 5515 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.912207269600252e-06, | |
| "loss": 2.2973, | |
| "step": 5520 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.906028818135687e-06, | |
| "loss": 2.3105, | |
| "step": 5525 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.89984893592795e-06, | |
| "loss": 2.2825, | |
| "step": 5530 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.893667632735915e-06, | |
| "loss": 2.3134, | |
| "step": 5535 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.887484918320708e-06, | |
| "loss": 2.2026, | |
| "step": 5540 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.881300802445675e-06, | |
| "loss": 2.3041, | |
| "step": 5545 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.8751152948763815e-06, | |
| "loss": 2.3399, | |
| "step": 5550 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.868928405380585e-06, | |
| "loss": 2.316, | |
| "step": 5555 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.8627401437282334e-06, | |
| "loss": 2.2602, | |
| "step": 5560 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.856550519691433e-06, | |
| "loss": 2.2303, | |
| "step": 5565 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.850359543044446e-06, | |
| "loss": 2.1945, | |
| "step": 5570 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.844167223563669e-06, | |
| "loss": 2.2801, | |
| "step": 5575 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.837973571027621e-06, | |
| "loss": 2.2166, | |
| "step": 5580 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.8317785952169245e-06, | |
| "loss": 2.3138, | |
| "step": 5585 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.82558230591429e-06, | |
| "loss": 2.2406, | |
| "step": 5590 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.819384712904508e-06, | |
| "loss": 2.2364, | |
| "step": 5595 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.813185825974419e-06, | |
| "loss": 2.2341, | |
| "step": 5600 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.806985654912915e-06, | |
| "loss": 2.3449, | |
| "step": 5605 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.80078420951091e-06, | |
| "loss": 2.2424, | |
| "step": 5610 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.794581499561335e-06, | |
| "loss": 2.2441, | |
| "step": 5615 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.7883775348591146e-06, | |
| "loss": 2.2408, | |
| "step": 5620 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.782172325201155e-06, | |
| "loss": 2.2169, | |
| "step": 5625 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.77596588038633e-06, | |
| "loss": 2.2021, | |
| "step": 5630 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.769758210215466e-06, | |
| "loss": 2.3004, | |
| "step": 5635 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.763549324491317e-06, | |
| "loss": 2.1895, | |
| "step": 5640 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.757339233018563e-06, | |
| "loss": 2.1797, | |
| "step": 5645 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.751127945603786e-06, | |
| "loss": 2.2288, | |
| "step": 5650 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.744915472055457e-06, | |
| "loss": 2.3038, | |
| "step": 5655 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.7387018221839195e-06, | |
| "loss": 2.2304, | |
| "step": 5660 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.7324870058013736e-06, | |
| "loss": 2.2384, | |
| "step": 5665 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.726271032721864e-06, | |
| "loss": 2.3377, | |
| "step": 5670 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.720053912761261e-06, | |
| "loss": 2.3201, | |
| "step": 5675 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.7138356557372444e-06, | |
| "loss": 2.3159, | |
| "step": 5680 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.707616271469293e-06, | |
| "loss": 2.3207, | |
| "step": 5685 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.70139576977866e-06, | |
| "loss": 2.2416, | |
| "step": 5690 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.695174160488369e-06, | |
| "loss": 2.2577, | |
| "step": 5695 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.68895145342319e-06, | |
| "loss": 2.3121, | |
| "step": 5700 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.682727658409628e-06, | |
| "loss": 2.1992, | |
| "step": 5705 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.6765027852759015e-06, | |
| "loss": 2.2995, | |
| "step": 5710 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.670276843851939e-06, | |
| "loss": 2.2898, | |
| "step": 5715 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.664049843969348e-06, | |
| "loss": 2.3188, | |
| "step": 5720 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.657821795461413e-06, | |
| "loss": 2.2783, | |
| "step": 5725 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.651592708163074e-06, | |
| "loss": 2.2751, | |
| "step": 5730 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.645362591910908e-06, | |
| "loss": 2.3223, | |
| "step": 5735 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.639131456543119e-06, | |
| "loss": 2.2557, | |
| "step": 5740 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.6328993118995215e-06, | |
| "loss": 2.2473, | |
| "step": 5745 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.626666167821522e-06, | |
| "loss": 2.232, | |
| "step": 5750 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.620432034152107e-06, | |
| "loss": 2.2924, | |
| "step": 5755 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.614196920735822e-06, | |
| "loss": 2.2849, | |
| "step": 5760 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.607960837418763e-06, | |
| "loss": 2.3238, | |
| "step": 5765 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.601723794048558e-06, | |
| "loss": 2.3151, | |
| "step": 5770 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.59548580047435e-06, | |
| "loss": 2.2851, | |
| "step": 5775 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.58924686654678e-06, | |
| "loss": 2.2321, | |
| "step": 5780 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.5830070021179785e-06, | |
| "loss": 2.2784, | |
| "step": 5785 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.576766217041541e-06, | |
| "loss": 2.2757, | |
| "step": 5790 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.570524521172523e-06, | |
| "loss": 2.2105, | |
| "step": 5795 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.5642819243674085e-06, | |
| "loss": 2.3106, | |
| "step": 5800 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.558038436484116e-06, | |
| "loss": 2.1958, | |
| "step": 5805 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 5.551794067381959e-06, | |
| "loss": 2.2666, | |
| "step": 5810 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.545548826921653e-06, | |
| "loss": 2.1864, | |
| "step": 5815 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.5393027249652844e-06, | |
| "loss": 2.2005, | |
| "step": 5820 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.5330557713763e-06, | |
| "loss": 2.2718, | |
| "step": 5825 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.526807976019492e-06, | |
| "loss": 2.2996, | |
| "step": 5830 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.520559348760984e-06, | |
| "loss": 2.3001, | |
| "step": 5835 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.514309899468209e-06, | |
| "loss": 2.2747, | |
| "step": 5840 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.5080596380099e-06, | |
| "loss": 2.1979, | |
| "step": 5845 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.5018085742560745e-06, | |
| "loss": 2.3155, | |
| "step": 5850 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.495556718078017e-06, | |
| "loss": 2.2549, | |
| "step": 5855 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.489304079348259e-06, | |
| "loss": 2.2184, | |
| "step": 5860 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.483050667940571e-06, | |
| "loss": 2.2718, | |
| "step": 5865 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.476796493729943e-06, | |
| "loss": 2.2518, | |
| "step": 5870 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.470541566592573e-06, | |
| "loss": 2.2502, | |
| "step": 5875 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.46428589640584e-06, | |
| "loss": 2.2822, | |
| "step": 5880 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.458029493048303e-06, | |
| "loss": 2.255, | |
| "step": 5885 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.451772366399678e-06, | |
| "loss": 2.2627, | |
| "step": 5890 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.445514526340822e-06, | |
| "loss": 2.3074, | |
| "step": 5895 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.439255982753717e-06, | |
| "loss": 2.192, | |
| "step": 5900 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.432996745521458e-06, | |
| "loss": 2.2741, | |
| "step": 5905 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.426736824528236e-06, | |
| "loss": 2.2495, | |
| "step": 5910 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.420476229659319e-06, | |
| "loss": 2.2634, | |
| "step": 5915 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.414214970801041e-06, | |
| "loss": 2.252, | |
| "step": 5920 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.4079530578407895e-06, | |
| "loss": 2.2395, | |
| "step": 5925 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.401690500666972e-06, | |
| "loss": 2.2953, | |
| "step": 5930 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5.3954273091690245e-06, | |
| "loss": 2.2015, | |
| "step": 5935 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.389163493237382e-06, | |
| "loss": 2.3292, | |
| "step": 5940 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.3828990627634655e-06, | |
| "loss": 2.2712, | |
| "step": 5945 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.376634027639664e-06, | |
| "loss": 2.2003, | |
| "step": 5950 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.370368397759324e-06, | |
| "loss": 2.1814, | |
| "step": 5955 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.36410218301673e-06, | |
| "loss": 2.2471, | |
| "step": 5960 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.357835393307089e-06, | |
| "loss": 2.17, | |
| "step": 5965 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.35156803852652e-06, | |
| "loss": 2.2586, | |
| "step": 5970 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.345300128572031e-06, | |
| "loss": 2.2792, | |
| "step": 5975 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.339031673341505e-06, | |
| "loss": 2.3279, | |
| "step": 5980 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.3327626827336906e-06, | |
| "loss": 2.2893, | |
| "step": 5985 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.326493166648179e-06, | |
| "loss": 2.286, | |
| "step": 5990 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.320223134985393e-06, | |
| "loss": 2.2495, | |
| "step": 5995 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.3139525976465675e-06, | |
| "loss": 2.2329, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.307681564533736e-06, | |
| "loss": 2.2677, | |
| "step": 6005 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.301410045549719e-06, | |
| "loss": 2.247, | |
| "step": 6010 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.295138050598097e-06, | |
| "loss": 2.3168, | |
| "step": 6015 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.2888655895832075e-06, | |
| "loss": 2.2643, | |
| "step": 6020 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.282592672410124e-06, | |
| "loss": 2.2419, | |
| "step": 6025 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.276319308984637e-06, | |
| "loss": 2.2945, | |
| "step": 6030 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.270045509213244e-06, | |
| "loss": 2.2598, | |
| "step": 6035 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.263771283003133e-06, | |
| "loss": 2.1945, | |
| "step": 6040 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.2574966402621615e-06, | |
| "loss": 2.2878, | |
| "step": 6045 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.251221590898848e-06, | |
| "loss": 2.2699, | |
| "step": 6050 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.244946144822351e-06, | |
| "loss": 2.3284, | |
| "step": 6055 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.238670311942459e-06, | |
| "loss": 2.2116, | |
| "step": 6060 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.232394102169566e-06, | |
| "loss": 2.2753, | |
| "step": 6065 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.226117525414663e-06, | |
| "loss": 2.3424, | |
| "step": 6070 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.219840591589325e-06, | |
| "loss": 2.2138, | |
| "step": 6075 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.213563310605686e-06, | |
| "loss": 2.2668, | |
| "step": 6080 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.207285692376427e-06, | |
| "loss": 2.2332, | |
| "step": 6085 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.201007746814767e-06, | |
| "loss": 2.3016, | |
| "step": 6090 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.194729483834438e-06, | |
| "loss": 2.2858, | |
| "step": 6095 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.188450913349674e-06, | |
| "loss": 2.229, | |
| "step": 6100 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.1821720452751945e-06, | |
| "loss": 2.2278, | |
| "step": 6105 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.175892889526189e-06, | |
| "loss": 2.1993, | |
| "step": 6110 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.1696134560183045e-06, | |
| "loss": 2.2403, | |
| "step": 6115 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.16333375466762e-06, | |
| "loss": 2.3565, | |
| "step": 6120 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.157053795390642e-06, | |
| "loss": 2.2705, | |
| "step": 6125 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.150773588104284e-06, | |
| "loss": 2.265, | |
| "step": 6130 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.144493142725851e-06, | |
| "loss": 2.3347, | |
| "step": 6135 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.138212469173022e-06, | |
| "loss": 2.2875, | |
| "step": 6140 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.13193157736384e-06, | |
| "loss": 2.2841, | |
| "step": 6145 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.1256504772166885e-06, | |
| "loss": 2.2417, | |
| "step": 6150 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.119369178650282e-06, | |
| "loss": 2.2779, | |
| "step": 6155 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.1130876915836495e-06, | |
| "loss": 2.2376, | |
| "step": 6160 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.1068060259361155e-06, | |
| "loss": 2.2326, | |
| "step": 6165 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.100524191627289e-06, | |
| "loss": 2.1924, | |
| "step": 6170 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.0942421985770415e-06, | |
| "loss": 2.2584, | |
| "step": 6175 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.087960056705499e-06, | |
| "loss": 2.2568, | |
| "step": 6180 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.0816777759330215e-06, | |
| "loss": 2.3209, | |
| "step": 6185 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 5.075395366180186e-06, | |
| "loss": 2.2195, | |
| "step": 6190 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 5.069112837367777e-06, | |
| "loss": 2.2729, | |
| "step": 6195 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 5.062830199416764e-06, | |
| "loss": 2.1644, | |
| "step": 6200 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 5.05654746224829e-06, | |
| "loss": 2.275, | |
| "step": 6205 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 5.050264635783654e-06, | |
| "loss": 2.2803, | |
| "step": 6210 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 5.043981729944298e-06, | |
| "loss": 2.2699, | |
| "step": 6215 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 5.037698754651786e-06, | |
| "loss": 2.2054, | |
| "step": 6220 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 5.031415719827796e-06, | |
| "loss": 2.2891, | |
| "step": 6225 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 5.025132635394095e-06, | |
| "loss": 2.264, | |
| "step": 6230 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 5.018849511272532e-06, | |
| "loss": 2.206, | |
| "step": 6235 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 5.0125663573850204e-06, | |
| "loss": 2.2726, | |
| "step": 6240 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 5.006283183653513e-06, | |
| "loss": 2.249, | |
| "step": 6245 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 5e-06, | |
| "loss": 2.2258, | |
| "step": 6250 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.99371681634649e-06, | |
| "loss": 2.261, | |
| "step": 6255 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.987433642614981e-06, | |
| "loss": 2.3, | |
| "step": 6260 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.981150488727469e-06, | |
| "loss": 2.2315, | |
| "step": 6265 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.974867364605906e-06, | |
| "loss": 2.1771, | |
| "step": 6270 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.968584280172206e-06, | |
| "loss": 2.212, | |
| "step": 6275 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.962301245348215e-06, | |
| "loss": 2.2349, | |
| "step": 6280 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.956018270055703e-06, | |
| "loss": 2.2151, | |
| "step": 6285 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.949735364216348e-06, | |
| "loss": 2.2273, | |
| "step": 6290 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.9434525377517115e-06, | |
| "loss": 2.2515, | |
| "step": 6295 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.937169800583237e-06, | |
| "loss": 2.2963, | |
| "step": 6300 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.930887162632225e-06, | |
| "loss": 2.188, | |
| "step": 6305 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.924604633819815e-06, | |
| "loss": 2.1956, | |
| "step": 6310 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.91832222406698e-06, | |
| "loss": 2.1879, | |
| "step": 6315 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.912039943294502e-06, | |
| "loss": 2.2168, | |
| "step": 6320 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.90575780142296e-06, | |
| "loss": 2.222, | |
| "step": 6325 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.899475808372714e-06, | |
| "loss": 2.2744, | |
| "step": 6330 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.893193974063885e-06, | |
| "loss": 2.2352, | |
| "step": 6335 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.886912308416353e-06, | |
| "loss": 2.2829, | |
| "step": 6340 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.880630821349718e-06, | |
| "loss": 2.2319, | |
| "step": 6345 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.874349522783313e-06, | |
| "loss": 2.3262, | |
| "step": 6350 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.8680684226361624e-06, | |
| "loss": 2.1806, | |
| "step": 6355 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.861787530826979e-06, | |
| "loss": 2.3417, | |
| "step": 6360 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.85550685727415e-06, | |
| "loss": 2.1901, | |
| "step": 6365 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.8492264118957165e-06, | |
| "loss": 2.2282, | |
| "step": 6370 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.842946204609359e-06, | |
| "loss": 2.2469, | |
| "step": 6375 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.8366662453323826e-06, | |
| "loss": 2.2872, | |
| "step": 6380 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.830386543981696e-06, | |
| "loss": 2.273, | |
| "step": 6385 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.8241071104738115e-06, | |
| "loss": 2.2575, | |
| "step": 6390 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.8178279547248055e-06, | |
| "loss": 2.2889, | |
| "step": 6395 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.811549086650327e-06, | |
| "loss": 2.265, | |
| "step": 6400 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.805270516165564e-06, | |
| "loss": 2.1987, | |
| "step": 6405 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.798992253185233e-06, | |
| "loss": 2.2337, | |
| "step": 6410 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.792714307623574e-06, | |
| "loss": 2.3008, | |
| "step": 6415 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.786436689394317e-06, | |
| "loss": 2.2412, | |
| "step": 6420 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.780159408410677e-06, | |
| "loss": 2.2265, | |
| "step": 6425 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.773882474585338e-06, | |
| "loss": 2.2141, | |
| "step": 6430 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.767605897830436e-06, | |
| "loss": 2.1922, | |
| "step": 6435 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.761329688057543e-06, | |
| "loss": 2.2825, | |
| "step": 6440 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.7550538551776495e-06, | |
| "loss": 2.221, | |
| "step": 6445 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.748778409101153e-06, | |
| "loss": 2.295, | |
| "step": 6450 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.742503359737841e-06, | |
| "loss": 2.2723, | |
| "step": 6455 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.736228716996868e-06, | |
| "loss": 2.2652, | |
| "step": 6460 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.7299544907867576e-06, | |
| "loss": 2.2184, | |
| "step": 6465 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.723680691015366e-06, | |
| "loss": 2.174, | |
| "step": 6470 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.717407327589878e-06, | |
| "loss": 2.2772, | |
| "step": 6475 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.711134410416794e-06, | |
| "loss": 2.2606, | |
| "step": 6480 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.704861949401904e-06, | |
| "loss": 2.2625, | |
| "step": 6485 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.6985899544502835e-06, | |
| "loss": 2.2356, | |
| "step": 6490 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.692318435466265e-06, | |
| "loss": 2.205, | |
| "step": 6495 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.686047402353433e-06, | |
| "loss": 2.2931, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.679776865014609e-06, | |
| "loss": 2.2384, | |
| "step": 6505 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.673506833351821e-06, | |
| "loss": 2.2201, | |
| "step": 6510 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.667237317266311e-06, | |
| "loss": 2.1928, | |
| "step": 6515 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.660968326658497e-06, | |
| "loss": 2.3077, | |
| "step": 6520 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.654699871427972e-06, | |
| "loss": 2.274, | |
| "step": 6525 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.648431961473482e-06, | |
| "loss": 2.261, | |
| "step": 6530 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.642164606692912e-06, | |
| "loss": 2.1873, | |
| "step": 6535 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.635897816983272e-06, | |
| "loss": 2.2949, | |
| "step": 6540 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.629631602240678e-06, | |
| "loss": 2.2936, | |
| "step": 6545 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.6233659723603374e-06, | |
| "loss": 2.2299, | |
| "step": 6550 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.617100937236535e-06, | |
| "loss": 2.2462, | |
| "step": 6555 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.610836506762618e-06, | |
| "loss": 2.242, | |
| "step": 6560 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.604572690830976e-06, | |
| "loss": 2.3226, | |
| "step": 6565 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.59830949933303e-06, | |
| "loss": 2.2171, | |
| "step": 6570 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.592046942159213e-06, | |
| "loss": 2.2503, | |
| "step": 6575 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.5857850291989596e-06, | |
| "loss": 2.2439, | |
| "step": 6580 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.579523770340681e-06, | |
| "loss": 2.3189, | |
| "step": 6585 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.573263175471766e-06, | |
| "loss": 2.1639, | |
| "step": 6590 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.567003254478545e-06, | |
| "loss": 2.261, | |
| "step": 6595 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.560744017246284e-06, | |
| "loss": 2.2563, | |
| "step": 6600 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.55448547365918e-06, | |
| "loss": 2.2097, | |
| "step": 6605 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.548227633600322e-06, | |
| "loss": 2.1462, | |
| "step": 6610 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.541970506951698e-06, | |
| "loss": 2.2125, | |
| "step": 6615 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.535714103594162e-06, | |
| "loss": 2.246, | |
| "step": 6620 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.529458433407429e-06, | |
| "loss": 2.2367, | |
| "step": 6625 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.523203506270058e-06, | |
| "loss": 2.2105, | |
| "step": 6630 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.516949332059429e-06, | |
| "loss": 2.2816, | |
| "step": 6635 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.510695920651742e-06, | |
| "loss": 2.2368, | |
| "step": 6640 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.504443281921985e-06, | |
| "loss": 2.1866, | |
| "step": 6645 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.4981914257439254e-06, | |
| "loss": 2.2128, | |
| "step": 6650 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.491940361990101e-06, | |
| "loss": 2.2162, | |
| "step": 6655 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.485690100531793e-06, | |
| "loss": 2.1871, | |
| "step": 6660 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.4794406512390175e-06, | |
| "loss": 2.2568, | |
| "step": 6665 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.473192023980509e-06, | |
| "loss": 2.2777, | |
| "step": 6670 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.466944228623701e-06, | |
| "loss": 2.2439, | |
| "step": 6675 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.460697275034717e-06, | |
| "loss": 2.1973, | |
| "step": 6680 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.454451173078347e-06, | |
| "loss": 2.2348, | |
| "step": 6685 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.448205932618042e-06, | |
| "loss": 2.176, | |
| "step": 6690 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.4419615635158875e-06, | |
| "loss": 2.2864, | |
| "step": 6695 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.4357180756325915e-06, | |
| "loss": 2.2244, | |
| "step": 6700 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.42947547882748e-06, | |
| "loss": 2.2324, | |
| "step": 6705 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.423233782958459e-06, | |
| "loss": 2.2881, | |
| "step": 6710 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.416992997882023e-06, | |
| "loss": 2.2689, | |
| "step": 6715 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.410753133453222e-06, | |
| "loss": 2.2598, | |
| "step": 6720 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.404514199525651e-06, | |
| "loss": 2.2269, | |
| "step": 6725 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.398276205951443e-06, | |
| "loss": 2.2365, | |
| "step": 6730 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.392039162581239e-06, | |
| "loss": 2.1676, | |
| "step": 6735 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.38580307926418e-06, | |
| "loss": 2.2274, | |
| "step": 6740 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.379567965847896e-06, | |
| "loss": 2.2549, | |
| "step": 6745 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.373333832178478e-06, | |
| "loss": 2.2574, | |
| "step": 6750 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.36710068810048e-06, | |
| "loss": 2.2525, | |
| "step": 6755 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.360868543456883e-06, | |
| "loss": 2.2346, | |
| "step": 6760 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.354637408089093e-06, | |
| "loss": 2.2808, | |
| "step": 6765 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.348407291836928e-06, | |
| "loss": 2.2716, | |
| "step": 6770 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.342178204538588e-06, | |
| "loss": 2.2281, | |
| "step": 6775 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.335950156030653e-06, | |
| "loss": 2.2172, | |
| "step": 6780 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.329723156148064e-06, | |
| "loss": 2.2134, | |
| "step": 6785 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.323497214724099e-06, | |
| "loss": 2.2721, | |
| "step": 6790 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.317272341590373e-06, | |
| "loss": 2.2303, | |
| "step": 6795 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.31104854657681e-06, | |
| "loss": 2.2113, | |
| "step": 6800 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.3048258395116326e-06, | |
| "loss": 2.2363, | |
| "step": 6805 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.298604230221341e-06, | |
| "loss": 2.2702, | |
| "step": 6810 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.2923837285307085e-06, | |
| "loss": 2.2306, | |
| "step": 6815 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.286164344262756e-06, | |
| "loss": 2.2728, | |
| "step": 6820 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.279946087238739e-06, | |
| "loss": 2.2265, | |
| "step": 6825 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.273728967278137e-06, | |
| "loss": 2.1786, | |
| "step": 6830 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.267512994198629e-06, | |
| "loss": 2.3435, | |
| "step": 6835 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.261298177816082e-06, | |
| "loss": 2.2276, | |
| "step": 6840 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.2550845279445455e-06, | |
| "loss": 2.2196, | |
| "step": 6845 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.248872054396215e-06, | |
| "loss": 2.2195, | |
| "step": 6850 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.242660766981439e-06, | |
| "loss": 2.2251, | |
| "step": 6855 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.2364506755086856e-06, | |
| "loss": 2.2161, | |
| "step": 6860 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.230241789784535e-06, | |
| "loss": 2.1537, | |
| "step": 6865 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.224034119613671e-06, | |
| "loss": 2.213, | |
| "step": 6870 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.217827674798845e-06, | |
| "loss": 2.2898, | |
| "step": 6875 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.211622465140887e-06, | |
| "loss": 2.2036, | |
| "step": 6880 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.2054185004386675e-06, | |
| "loss": 2.1947, | |
| "step": 6885 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.199215790489091e-06, | |
| "loss": 2.3359, | |
| "step": 6890 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.193014345087088e-06, | |
| "loss": 2.2807, | |
| "step": 6895 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.186814174025582e-06, | |
| "loss": 2.2464, | |
| "step": 6900 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.180615287095494e-06, | |
| "loss": 2.2259, | |
| "step": 6905 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.174417694085711e-06, | |
| "loss": 2.1961, | |
| "step": 6910 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.168221404783076e-06, | |
| "loss": 2.1827, | |
| "step": 6915 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.16202642897238e-06, | |
| "loss": 2.2115, | |
| "step": 6920 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.155832776436331e-06, | |
| "loss": 2.2251, | |
| "step": 6925 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.149640456955555e-06, | |
| "loss": 2.3135, | |
| "step": 6930 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.143449480308569e-06, | |
| "loss": 2.2528, | |
| "step": 6935 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.137259856271767e-06, | |
| "loss": 2.34, | |
| "step": 6940 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.131071594619416e-06, | |
| "loss": 2.2743, | |
| "step": 6945 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.124884705123619e-06, | |
| "loss": 2.2834, | |
| "step": 6950 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.118699197554327e-06, | |
| "loss": 2.2718, | |
| "step": 6955 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.112515081679295e-06, | |
| "loss": 2.1518, | |
| "step": 6960 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.106332367264085e-06, | |
| "loss": 2.2409, | |
| "step": 6965 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.1001510640720525e-06, | |
| "loss": 2.2183, | |
| "step": 6970 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.093971181864313e-06, | |
| "loss": 2.2012, | |
| "step": 6975 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.087792730399749e-06, | |
| "loss": 2.2158, | |
| "step": 6980 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.081615719434978e-06, | |
| "loss": 2.1918, | |
| "step": 6985 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.075440158724339e-06, | |
| "loss": 2.2587, | |
| "step": 6990 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.0692660580198905e-06, | |
| "loss": 2.1974, | |
| "step": 6995 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.063093427071376e-06, | |
| "loss": 2.2154, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.056922275626227e-06, | |
| "loss": 2.1753, | |
| "step": 7005 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.0507526134295314e-06, | |
| "loss": 2.2454, | |
| "step": 7010 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.044584450224026e-06, | |
| "loss": 2.1729, | |
| "step": 7015 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.038417795750086e-06, | |
| "loss": 2.2229, | |
| "step": 7020 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.032252659745699e-06, | |
| "loss": 2.2797, | |
| "step": 7025 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.0260890519464565e-06, | |
| "loss": 2.2983, | |
| "step": 7030 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.019926982085536e-06, | |
| "loss": 2.155, | |
| "step": 7035 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.013766459893686e-06, | |
| "loss": 2.2354, | |
| "step": 7040 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.007607495099215e-06, | |
| "loss": 2.2573, | |
| "step": 7045 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.001450097427965e-06, | |
| "loss": 2.2761, | |
| "step": 7050 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 3.995294276603312e-06, | |
| "loss": 2.3164, | |
| "step": 7055 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 3.989140042346134e-06, | |
| "loss": 2.2678, | |
| "step": 7060 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.9829874043748064e-06, | |
| "loss": 2.1791, | |
| "step": 7065 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.9768363724051875e-06, | |
| "loss": 2.2229, | |
| "step": 7070 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.970686956150595e-06, | |
| "loss": 2.1566, | |
| "step": 7075 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.964539165321795e-06, | |
| "loss": 2.2904, | |
| "step": 7080 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.95839300962699e-06, | |
| "loss": 2.2081, | |
| "step": 7085 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.952248498771797e-06, | |
| "loss": 2.2489, | |
| "step": 7090 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.946105642459241e-06, | |
| "loss": 2.2803, | |
| "step": 7095 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.939964450389728e-06, | |
| "loss": 2.1665, | |
| "step": 7100 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.9338249322610375e-06, | |
| "loss": 2.1943, | |
| "step": 7105 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.927687097768309e-06, | |
| "loss": 2.2534, | |
| "step": 7110 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.921550956604019e-06, | |
| "loss": 2.1374, | |
| "step": 7115 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.915416518457974e-06, | |
| "loss": 2.2036, | |
| "step": 7120 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.909283793017289e-06, | |
| "loss": 2.2206, | |
| "step": 7125 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.9031527899663705e-06, | |
| "loss": 2.2535, | |
| "step": 7130 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.897023518986915e-06, | |
| "loss": 2.2163, | |
| "step": 7135 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.890895989757874e-06, | |
| "loss": 2.1598, | |
| "step": 7140 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.884770211955454e-06, | |
| "loss": 2.202, | |
| "step": 7145 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.8786461952530955e-06, | |
| "loss": 2.1922, | |
| "step": 7150 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.872523949321454e-06, | |
| "loss": 2.1665, | |
| "step": 7155 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.866403483828392e-06, | |
| "loss": 2.2083, | |
| "step": 7160 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.860284808438962e-06, | |
| "loss": 2.2245, | |
| "step": 7165 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.854167932815387e-06, | |
| "loss": 2.2825, | |
| "step": 7170 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.8480528666170495e-06, | |
| "loss": 2.2044, | |
| "step": 7175 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.841939619500468e-06, | |
| "loss": 2.2104, | |
| "step": 7180 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.835828201119302e-06, | |
| "loss": 2.1428, | |
| "step": 7185 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.8297186211243085e-06, | |
| "loss": 2.3107, | |
| "step": 7190 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.823610889163354e-06, | |
| "loss": 2.165, | |
| "step": 7195 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.817505014881378e-06, | |
| "loss": 2.238, | |
| "step": 7200 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.8114010079203877e-06, | |
| "loss": 2.2269, | |
| "step": 7205 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.8052988779194478e-06, | |
| "loss": 2.2185, | |
| "step": 7210 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.7991986345146503e-06, | |
| "loss": 2.2225, | |
| "step": 7215 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.7931002873391156e-06, | |
| "loss": 2.2309, | |
| "step": 7220 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.787003846022964e-06, | |
| "loss": 2.268, | |
| "step": 7225 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.7809093201933078e-06, | |
| "loss": 2.2521, | |
| "step": 7230 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.774816719474238e-06, | |
| "loss": 2.1898, | |
| "step": 7235 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.7687260534868e-06, | |
| "loss": 2.2248, | |
| "step": 7240 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.762637331848989e-06, | |
| "loss": 2.1709, | |
| "step": 7245 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.756550564175727e-06, | |
| "loss": 2.216, | |
| "step": 7250 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.7504657600788484e-06, | |
| "loss": 2.1816, | |
| "step": 7255 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.744382929167094e-06, | |
| "loss": 2.2197, | |
| "step": 7260 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.73830208104608e-06, | |
| "loss": 2.2808, | |
| "step": 7265 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.7322232253182984e-06, | |
| "loss": 2.1773, | |
| "step": 7270 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.7261463715830902e-06, | |
| "loss": 2.1414, | |
| "step": 7275 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.7200715294366376e-06, | |
| "loss": 2.3004, | |
| "step": 7280 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.7139987084719463e-06, | |
| "loss": 2.2929, | |
| "step": 7285 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.7079279182788263e-06, | |
| "loss": 2.2374, | |
| "step": 7290 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.70185916844389e-06, | |
| "loss": 2.1595, | |
| "step": 7295 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.695792468550517e-06, | |
| "loss": 2.3007, | |
| "step": 7300 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.689727828178854e-06, | |
| "loss": 2.3113, | |
| "step": 7305 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.6836652569057994e-06, | |
| "loss": 2.2021, | |
| "step": 7310 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.6776047643049777e-06, | |
| "loss": 2.1749, | |
| "step": 7315 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.6715463599467372e-06, | |
| "loss": 2.186, | |
| "step": 7320 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.6654900533981234e-06, | |
| "loss": 2.2751, | |
| "step": 7325 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.659435854222869e-06, | |
| "loss": 2.1464, | |
| "step": 7330 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.653383771981385e-06, | |
| "loss": 2.1889, | |
| "step": 7335 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.6473338162307314e-06, | |
| "loss": 2.1604, | |
| "step": 7340 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.6412859965246173e-06, | |
| "loss": 2.2495, | |
| "step": 7345 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.635240322413375e-06, | |
| "loss": 2.164, | |
| "step": 7350 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.6291968034439463e-06, | |
| "loss": 2.2169, | |
| "step": 7355 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.6231554491598766e-06, | |
| "loss": 2.2339, | |
| "step": 7360 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.617116269101286e-06, | |
| "loss": 2.0963, | |
| "step": 7365 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.6110792728048636e-06, | |
| "loss": 2.2447, | |
| "step": 7370 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.6050444698038547e-06, | |
| "loss": 2.1749, | |
| "step": 7375 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.599011869628033e-06, | |
| "loss": 2.2506, | |
| "step": 7380 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.592981481803699e-06, | |
| "loss": 2.2727, | |
| "step": 7385 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.5869533158536583e-06, | |
| "loss": 2.1917, | |
| "step": 7390 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.5809273812972078e-06, | |
| "loss": 2.205, | |
| "step": 7395 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.5749036876501196e-06, | |
| "loss": 2.2134, | |
| "step": 7400 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.5688822444246297e-06, | |
| "loss": 2.268, | |
| "step": 7405 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.562863061129419e-06, | |
| "loss": 2.2294, | |
| "step": 7410 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.556846147269598e-06, | |
| "loss": 2.2682, | |
| "step": 7415 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.550831512346695e-06, | |
| "loss": 2.1912, | |
| "step": 7420 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.5448191658586423e-06, | |
| "loss": 2.177, | |
| "step": 7425 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.538809117299751e-06, | |
| "loss": 2.2358, | |
| "step": 7430 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.532801376160713e-06, | |
| "loss": 2.2285, | |
| "step": 7435 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.526795951928569e-06, | |
| "loss": 2.1081, | |
| "step": 7440 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.520792854086702e-06, | |
| "loss": 2.2432, | |
| "step": 7445 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.5147920921148267e-06, | |
| "loss": 2.183, | |
| "step": 7450 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.5087936754889614e-06, | |
| "loss": 2.2032, | |
| "step": 7455 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.502797613681429e-06, | |
| "loss": 2.2021, | |
| "step": 7460 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.496803916160827e-06, | |
| "loss": 2.1604, | |
| "step": 7465 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.4908125923920204e-06, | |
| "loss": 2.2151, | |
| "step": 7470 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.484823651836131e-06, | |
| "loss": 2.1931, | |
| "step": 7475 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.478837103950509e-06, | |
| "loss": 2.2241, | |
| "step": 7480 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.472852958188736e-06, | |
| "loss": 2.2083, | |
| "step": 7485 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.4668712240005912e-06, | |
| "loss": 2.2133, | |
| "step": 7490 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.4608919108320488e-06, | |
| "loss": 2.2385, | |
| "step": 7495 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.4549150281252635e-06, | |
| "loss": 2.2044, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.4489405853185465e-06, | |
| "loss": 2.1845, | |
| "step": 7505 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.442968591846359e-06, | |
| "loss": 2.261, | |
| "step": 7510 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.436999057139295e-06, | |
| "loss": 2.2565, | |
| "step": 7515 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.431031990624063e-06, | |
| "loss": 2.2248, | |
| "step": 7520 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.4250674017234774e-06, | |
| "loss": 2.223, | |
| "step": 7525 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.4191052998564344e-06, | |
| "loss": 2.2012, | |
| "step": 7530 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.4131456944379126e-06, | |
| "loss": 2.2049, | |
| "step": 7535 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.407188594878938e-06, | |
| "loss": 2.2636, | |
| "step": 7540 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.401234010586583e-06, | |
| "loss": 2.2088, | |
| "step": 7545 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.3952819509639534e-06, | |
| "loss": 2.2409, | |
| "step": 7550 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.38933242541016e-06, | |
| "loss": 2.2077, | |
| "step": 7555 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.3833854433203185e-06, | |
| "loss": 2.2122, | |
| "step": 7560 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.377441014085524e-06, | |
| "loss": 2.2117, | |
| "step": 7565 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.3714991470928393e-06, | |
| "loss": 2.2256, | |
| "step": 7570 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.3655598517252886e-06, | |
| "loss": 2.2005, | |
| "step": 7575 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.359623137361825e-06, | |
| "loss": 2.2018, | |
| "step": 7580 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.3536890133773346e-06, | |
| "loss": 2.1866, | |
| "step": 7585 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.347757489142608e-06, | |
| "loss": 2.175, | |
| "step": 7590 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.3418285740243285e-06, | |
| "loss": 2.147, | |
| "step": 7595 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.3359022773850673e-06, | |
| "loss": 2.1949, | |
| "step": 7600 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.329978608583252e-06, | |
| "loss": 2.1693, | |
| "step": 7605 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.3240575769731662e-06, | |
| "loss": 2.1607, | |
| "step": 7610 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.3181391919049277e-06, | |
| "loss": 2.2739, | |
| "step": 7615 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.312223462724472e-06, | |
| "loss": 2.2886, | |
| "step": 7620 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.3063103987735433e-06, | |
| "loss": 2.2069, | |
| "step": 7625 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.300400009389678e-06, | |
| "loss": 2.2316, | |
| "step": 7630 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.294492303906188e-06, | |
| "loss": 2.1939, | |
| "step": 7635 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.2885872916521445e-06, | |
| "loss": 2.2077, | |
| "step": 7640 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.282684981952369e-06, | |
| "loss": 2.2212, | |
| "step": 7645 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.2767853841274154e-06, | |
| "loss": 2.2299, | |
| "step": 7650 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.2708885074935515e-06, | |
| "loss": 2.1994, | |
| "step": 7655 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.264994361362753e-06, | |
| "loss": 2.1668, | |
| "step": 7660 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.25910295504268e-06, | |
| "loss": 2.2223, | |
| "step": 7665 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.2532142978366654e-06, | |
| "loss": 2.1745, | |
| "step": 7670 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.247328399043706e-06, | |
| "loss": 2.2487, | |
| "step": 7675 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.241445267958438e-06, | |
| "loss": 2.2104, | |
| "step": 7680 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.235564913871126e-06, | |
| "loss": 2.2327, | |
| "step": 7685 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.2296873460676557e-06, | |
| "loss": 2.2531, | |
| "step": 7690 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.2238125738295063e-06, | |
| "loss": 2.2159, | |
| "step": 7695 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.217940606433747e-06, | |
| "loss": 2.1774, | |
| "step": 7700 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.212071453153015e-06, | |
| "loss": 2.2625, | |
| "step": 7705 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.2062051232555024e-06, | |
| "loss": 2.2579, | |
| "step": 7710 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.2003416260049493e-06, | |
| "loss": 2.1515, | |
| "step": 7715 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.1944809706606123e-06, | |
| "loss": 2.1691, | |
| "step": 7720 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.188623166477272e-06, | |
| "loss": 2.2366, | |
| "step": 7725 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.182768222705198e-06, | |
| "loss": 2.2738, | |
| "step": 7730 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.1769161485901445e-06, | |
| "loss": 2.1898, | |
| "step": 7735 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.171066953373338e-06, | |
| "loss": 2.2526, | |
| "step": 7740 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.1652206462914542e-06, | |
| "loss": 2.2128, | |
| "step": 7745 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.1593772365766107e-06, | |
| "loss": 2.2041, | |
| "step": 7750 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.1535367334563493e-06, | |
| "loss": 2.1675, | |
| "step": 7755 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.147699146153621e-06, | |
| "loss": 2.2181, | |
| "step": 7760 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.141864483886774e-06, | |
| "loss": 2.2141, | |
| "step": 7765 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.1360327558695336e-06, | |
| "loss": 2.1343, | |
| "step": 7770 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.130203971310999e-06, | |
| "loss": 2.1392, | |
| "step": 7775 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.1243781394156138e-06, | |
| "loss": 2.2489, | |
| "step": 7780 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.1185552693831595e-06, | |
| "loss": 2.2357, | |
| "step": 7785 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.1127353704087477e-06, | |
| "loss": 2.1943, | |
| "step": 7790 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.106918451682789e-06, | |
| "loss": 2.2423, | |
| "step": 7795 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.1011045223909954e-06, | |
| "loss": 2.1739, | |
| "step": 7800 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.0952935917143533e-06, | |
| "loss": 2.1657, | |
| "step": 7805 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.089485668829113e-06, | |
| "loss": 2.2285, | |
| "step": 7810 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 3.0836807629067828e-06, | |
| "loss": 2.1396, | |
| "step": 7815 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 3.077878883114096e-06, | |
| "loss": 2.2545, | |
| "step": 7820 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 3.0720800386130176e-06, | |
| "loss": 2.1563, | |
| "step": 7825 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 3.066284238560713e-06, | |
| "loss": 2.2612, | |
| "step": 7830 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 3.0604914921095373e-06, | |
| "loss": 2.2927, | |
| "step": 7835 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 3.0547018084070344e-06, | |
| "loss": 2.1302, | |
| "step": 7840 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 3.0489151965958998e-06, | |
| "loss": 2.2431, | |
| "step": 7845 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 3.043131665813988e-06, | |
| "loss": 2.2243, | |
| "step": 7850 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 3.0373512251942817e-06, | |
| "loss": 2.2663, | |
| "step": 7855 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 3.031573883864882e-06, | |
| "loss": 2.1598, | |
| "step": 7860 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 3.025799650949006e-06, | |
| "loss": 2.2245, | |
| "step": 7865 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 3.0200285355649504e-06, | |
| "loss": 2.162, | |
| "step": 7870 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 3.0142605468260976e-06, | |
| "loss": 2.1349, | |
| "step": 7875 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 3.0084956938408873e-06, | |
| "loss": 2.1707, | |
| "step": 7880 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 3.002733985712808e-06, | |
| "loss": 2.1705, | |
| "step": 7885 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 2.9969754315403865e-06, | |
| "loss": 2.1642, | |
| "step": 7890 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 2.991220040417162e-06, | |
| "loss": 2.1568, | |
| "step": 7895 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 2.9854678214316875e-06, | |
| "loss": 2.1443, | |
| "step": 7900 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 2.979718783667499e-06, | |
| "loss": 2.1753, | |
| "step": 7905 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 2.97397293620311e-06, | |
| "loss": 2.1515, | |
| "step": 7910 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 2.968230288112002e-06, | |
| "loss": 2.1942, | |
| "step": 7915 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 2.962490848462596e-06, | |
| "loss": 2.1404, | |
| "step": 7920 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 2.9567546263182554e-06, | |
| "loss": 2.2624, | |
| "step": 7925 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 2.951021630737255e-06, | |
| "loss": 2.2003, | |
| "step": 7930 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 2.945291870772776e-06, | |
| "loss": 2.2547, | |
| "step": 7935 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.9395653554728955e-06, | |
| "loss": 2.2449, | |
| "step": 7940 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.933842093880558e-06, | |
| "loss": 2.1933, | |
| "step": 7945 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.92812209503358e-06, | |
| "loss": 2.1803, | |
| "step": 7950 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.922405367964617e-06, | |
| "loss": 2.2251, | |
| "step": 7955 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.91669192170116e-06, | |
| "loss": 2.1468, | |
| "step": 7960 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.9109817652655253e-06, | |
| "loss": 2.1795, | |
| "step": 7965 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.9052749076748266e-06, | |
| "loss": 2.1931, | |
| "step": 7970 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.899571357940969e-06, | |
| "loss": 2.1697, | |
| "step": 7975 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.8938711250706397e-06, | |
| "loss": 2.2354, | |
| "step": 7980 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.8881742180652813e-06, | |
| "loss": 2.1965, | |
| "step": 7985 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.8824806459210907e-06, | |
| "loss": 2.2709, | |
| "step": 7990 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.876790417628994e-06, | |
| "loss": 2.221, | |
| "step": 7995 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.871103542174637e-06, | |
| "loss": 2.255, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.865420028538375e-06, | |
| "loss": 2.1331, | |
| "step": 8005 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.8597398856952473e-06, | |
| "loss": 2.1633, | |
| "step": 8010 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.8540631226149813e-06, | |
| "loss": 2.255, | |
| "step": 8015 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.8483897482619566e-06, | |
| "loss": 2.2264, | |
| "step": 8020 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.8427197715952047e-06, | |
| "loss": 2.1511, | |
| "step": 8025 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.837053201568396e-06, | |
| "loss": 2.14, | |
| "step": 8030 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.831390047129815e-06, | |
| "loss": 2.1392, | |
| "step": 8035 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.825730317222358e-06, | |
| "loss": 2.1658, | |
| "step": 8040 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.820074020783511e-06, | |
| "loss": 2.1934, | |
| "step": 8045 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.814421166745337e-06, | |
| "loss": 2.2368, | |
| "step": 8050 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.808771764034462e-06, | |
| "loss": 2.2083, | |
| "step": 8055 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.803125821572068e-06, | |
| "loss": 2.2294, | |
| "step": 8060 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.7974833482738674e-06, | |
| "loss": 2.1725, | |
| "step": 8065 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.791844353050094e-06, | |
| "loss": 2.1908, | |
| "step": 8070 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.7862088448054936e-06, | |
| "loss": 2.1769, | |
| "step": 8075 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.7805768324393017e-06, | |
| "loss": 2.1606, | |
| "step": 8080 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.7749483248452324e-06, | |
| "loss": 2.1116, | |
| "step": 8085 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.769323330911472e-06, | |
| "loss": 2.1355, | |
| "step": 8090 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.763701859520652e-06, | |
| "loss": 2.1257, | |
| "step": 8095 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.7580839195498397e-06, | |
| "loss": 2.2196, | |
| "step": 8100 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.752469519870534e-06, | |
| "loss": 2.2401, | |
| "step": 8105 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.746858669348634e-06, | |
| "loss": 2.1424, | |
| "step": 8110 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.741251376844443e-06, | |
| "loss": 2.2377, | |
| "step": 8115 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.7356476512126386e-06, | |
| "loss": 2.2156, | |
| "step": 8120 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.7300475013022666e-06, | |
| "loss": 2.2278, | |
| "step": 8125 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.724450935956733e-06, | |
| "loss": 2.2109, | |
| "step": 8130 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.718857964013773e-06, | |
| "loss": 2.1578, | |
| "step": 8135 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.713268594305458e-06, | |
| "loss": 2.1792, | |
| "step": 8140 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.707682835658163e-06, | |
| "loss": 2.1204, | |
| "step": 8145 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.7021006968925613e-06, | |
| "loss": 2.2116, | |
| "step": 8150 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.6965221868236156e-06, | |
| "loss": 2.2175, | |
| "step": 8155 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.6909473142605522e-06, | |
| "loss": 2.1413, | |
| "step": 8160 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.6853760880068587e-06, | |
| "loss": 2.1436, | |
| "step": 8165 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.6798085168602595e-06, | |
| "loss": 2.1733, | |
| "step": 8170 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.6742446096127086e-06, | |
| "loss": 2.2105, | |
| "step": 8175 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.668684375050378e-06, | |
| "loss": 2.2215, | |
| "step": 8180 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.663127821953633e-06, | |
| "loss": 2.2253, | |
| "step": 8185 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.6575749590970336e-06, | |
| "loss": 2.2247, | |
| "step": 8190 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.6520257952493066e-06, | |
| "loss": 2.2185, | |
| "step": 8195 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.646480339173337e-06, | |
| "loss": 2.2234, | |
| "step": 8200 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.6409385996261606e-06, | |
| "loss": 2.1679, | |
| "step": 8205 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.635400585358937e-06, | |
| "loss": 2.1984, | |
| "step": 8210 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.62986630511695e-06, | |
| "loss": 2.1892, | |
| "step": 8215 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.624335767639582e-06, | |
| "loss": 2.1375, | |
| "step": 8220 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.618808981660304e-06, | |
| "loss": 2.2053, | |
| "step": 8225 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.6132859559066704e-06, | |
| "loss": 2.1921, | |
| "step": 8230 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.607766699100288e-06, | |
| "loss": 2.2098, | |
| "step": 8235 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.6022512199568205e-06, | |
| "loss": 2.2212, | |
| "step": 8240 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.5967395271859614e-06, | |
| "loss": 2.1665, | |
| "step": 8245 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.5912316294914232e-06, | |
| "loss": 2.1644, | |
| "step": 8250 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.5857275355709317e-06, | |
| "loss": 2.1334, | |
| "step": 8255 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.580227254116199e-06, | |
| "loss": 2.1471, | |
| "step": 8260 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.5747307938129245e-06, | |
| "loss": 2.1673, | |
| "step": 8265 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.5692381633407672e-06, | |
| "loss": 2.21, | |
| "step": 8270 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.5637493713733376e-06, | |
| "loss": 2.1586, | |
| "step": 8275 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.558264426578192e-06, | |
| "loss": 2.1746, | |
| "step": 8280 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.5527833376168055e-06, | |
| "loss": 2.1255, | |
| "step": 8285 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.547306113144564e-06, | |
| "loss": 2.1898, | |
| "step": 8290 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.541832761810753e-06, | |
| "loss": 2.2576, | |
| "step": 8295 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.536363292258543e-06, | |
| "loss": 2.1903, | |
| "step": 8300 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.5308977131249724e-06, | |
| "loss": 2.2267, | |
| "step": 8305 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.5254360330409343e-06, | |
| "loss": 2.2118, | |
| "step": 8310 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.5199782606311708e-06, | |
| "loss": 2.1397, | |
| "step": 8315 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.514524404514248e-06, | |
| "loss": 2.2176, | |
| "step": 8320 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.509074473302546e-06, | |
| "loss": 2.1814, | |
| "step": 8325 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.503628475602256e-06, | |
| "loss": 2.1132, | |
| "step": 8330 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.4981864200133483e-06, | |
| "loss": 2.1979, | |
| "step": 8335 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.49274831512957e-06, | |
| "loss": 2.2007, | |
| "step": 8340 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.4873141695384346e-06, | |
| "loss": 2.1319, | |
| "step": 8345 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.4818839918211963e-06, | |
| "loss": 2.2091, | |
| "step": 8350 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.4764577905528503e-06, | |
| "loss": 2.1447, | |
| "step": 8355 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.4710355743021077e-06, | |
| "loss": 2.2068, | |
| "step": 8360 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.4656173516313852e-06, | |
| "loss": 2.1397, | |
| "step": 8365 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.4602031310968013e-06, | |
| "loss": 2.136, | |
| "step": 8370 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.4547929212481436e-06, | |
| "loss": 2.1956, | |
| "step": 8375 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.4493867306288772e-06, | |
| "loss": 2.2745, | |
| "step": 8380 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.4439845677761124e-06, | |
| "loss": 2.2172, | |
| "step": 8385 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.4385864412206e-06, | |
| "loss": 2.2382, | |
| "step": 8390 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.433192359486723e-06, | |
| "loss": 2.1553, | |
| "step": 8395 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.4278023310924676e-06, | |
| "loss": 2.2337, | |
| "step": 8400 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.422416364549429e-06, | |
| "loss": 2.1927, | |
| "step": 8405 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.417034468362782e-06, | |
| "loss": 2.1936, | |
| "step": 8410 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.4116566510312734e-06, | |
| "loss": 2.2425, | |
| "step": 8415 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.406282921047213e-06, | |
| "loss": 2.1399, | |
| "step": 8420 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.4009132868964525e-06, | |
| "loss": 2.203, | |
| "step": 8425 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.395547757058379e-06, | |
| "loss": 2.1916, | |
| "step": 8430 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.3901863400058954e-06, | |
| "loss": 2.1335, | |
| "step": 8435 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.3848290442054096e-06, | |
| "loss": 2.1704, | |
| "step": 8440 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.379475878116826e-06, | |
| "loss": 2.1943, | |
| "step": 8445 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.3741268501935212e-06, | |
| "loss": 2.2924, | |
| "step": 8450 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.368781968882343e-06, | |
| "loss": 2.2466, | |
| "step": 8455 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.3634412426235886e-06, | |
| "loss": 2.191, | |
| "step": 8460 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.358104679850991e-06, | |
| "loss": 2.1468, | |
| "step": 8465 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.3527722889917147e-06, | |
| "loss": 2.178, | |
| "step": 8470 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.3474440784663287e-06, | |
| "loss": 2.2116, | |
| "step": 8475 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.3421200566888096e-06, | |
| "loss": 2.1915, | |
| "step": 8480 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.3368002320665118e-06, | |
| "loss": 2.165, | |
| "step": 8485 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.3314846130001622e-06, | |
| "loss": 2.2258, | |
| "step": 8490 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.326173207883854e-06, | |
| "loss": 2.1779, | |
| "step": 8495 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.320866025105016e-06, | |
| "loss": 2.1818, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.3155630730444185e-06, | |
| "loss": 2.2147, | |
| "step": 8505 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.3102643600761445e-06, | |
| "loss": 2.1821, | |
| "step": 8510 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.3049698945675826e-06, | |
| "loss": 2.1946, | |
| "step": 8515 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.299679684879421e-06, | |
| "loss": 2.1554, | |
| "step": 8520 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.294393739365621e-06, | |
| "loss": 2.2281, | |
| "step": 8525 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.289112066373411e-06, | |
| "loss": 2.1658, | |
| "step": 8530 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.2838346742432753e-06, | |
| "loss": 2.1312, | |
| "step": 8535 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.2785615713089363e-06, | |
| "loss": 2.1439, | |
| "step": 8540 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.2732927658973427e-06, | |
| "loss": 2.1456, | |
| "step": 8545 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.268028266328655e-06, | |
| "loss": 2.1033, | |
| "step": 8550 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.262768080916241e-06, | |
| "loss": 2.1671, | |
| "step": 8555 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.25751221796665e-06, | |
| "loss": 2.1956, | |
| "step": 8560 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "eval_loss": 2.187166213989258, | |
| "eval_runtime": 653.3252, | |
| "eval_samples_per_second": 7.653, | |
| "eval_steps_per_second": 0.479, | |
| "step": 8563 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "step": 8563, | |
| "total_flos": 224101662720000.0, | |
| "train_loss": 2.299378549998349, | |
| "train_runtime": 31433.0779, | |
| "train_samples_per_second": 1.591, | |
| "train_steps_per_second": 0.398 | |
| } | |
| ], | |
| "logging_steps": 5, | |
| "max_steps": 12500, | |
| "num_train_epochs": 1, | |
| "save_steps": 500, | |
| "total_flos": 224101662720000.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |