| {"current_steps": 5, "total_steps": 1154, "loss": 1.0706, "learning_rate": 8.333333333333333e-07, "epoch": 0.004332755632582322, "percentage": 0.43, "elapsed_time": "0:01:32", "remaining_time": "5:52:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 10, "total_steps": 1154, "loss": 1.0694, "learning_rate": 1.6666666666666667e-06, "epoch": 0.008665511265164644, "percentage": 0.87, "elapsed_time": "0:03:05", "remaining_time": "5:53:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 15, "total_steps": 1154, "loss": 1.0706, "learning_rate": 2.5e-06, "epoch": 0.012998266897746967, "percentage": 1.3, "elapsed_time": "0:04:39", "remaining_time": "5:53:58", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 20, "total_steps": 1154, "loss": 1.0772, "learning_rate": 3.3333333333333333e-06, "epoch": 0.01733102253032929, "percentage": 1.73, "elapsed_time": "0:06:14", "remaining_time": "5:54:21", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 25, "total_steps": 1154, "loss": 1.0809, "learning_rate": 4.166666666666667e-06, "epoch": 0.021663778162911613, "percentage": 2.17, "elapsed_time": "0:07:49", "remaining_time": "5:53:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 30, "total_steps": 1154, "loss": 1.0737, "learning_rate": 5e-06, "epoch": 0.025996533795493933, "percentage": 2.6, "elapsed_time": "0:09:23", "remaining_time": "5:51:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 35, "total_steps": 1154, "loss": 1.0675, "learning_rate": 5.833333333333334e-06, "epoch": 0.030329289428076257, "percentage": 3.03, "elapsed_time": "0:10:57", "remaining_time": "5:50:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 40, "total_steps": 1154, "loss": 1.0638, "learning_rate": 6.666666666666667e-06, "epoch": 0.03466204506065858, "percentage": 3.47, "elapsed_time": "0:12:31", "remaining_time": "5:48:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 45, "total_steps": 1154, "loss": 1.0588, "learning_rate": 7.5e-06, "epoch": 0.0389948006932409, "percentage": 3.9, "elapsed_time": "0:14:05", "remaining_time": "5:47:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 50, "total_steps": 1154, "loss": 1.0512, "learning_rate": 8.333333333333334e-06, "epoch": 0.043327556325823226, "percentage": 4.33, "elapsed_time": "0:15:39", "remaining_time": "5:45:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 55, "total_steps": 1154, "loss": 1.0545, "learning_rate": 9.166666666666666e-06, "epoch": 0.047660311958405546, "percentage": 4.77, "elapsed_time": "0:17:13", "remaining_time": "5:44:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 60, "total_steps": 1154, "loss": 1.0357, "learning_rate": 1e-05, "epoch": 0.05199306759098787, "percentage": 5.2, "elapsed_time": "0:18:47", "remaining_time": "5:42:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 65, "total_steps": 1154, "loss": 1.0268, "learning_rate": 1.0833333333333334e-05, "epoch": 0.05632582322357019, "percentage": 5.63, "elapsed_time": "0:20:21", "remaining_time": "5:41:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 70, "total_steps": 1154, "loss": 1.0135, "learning_rate": 1.1666666666666668e-05, "epoch": 0.060658578856152515, "percentage": 6.07, "elapsed_time": "0:21:55", "remaining_time": "5:39:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 75, "total_steps": 1154, "loss": 1.0014, "learning_rate": 1.25e-05, "epoch": 0.06499133448873484, "percentage": 6.5, "elapsed_time": "0:23:29", "remaining_time": "5:37:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 80, "total_steps": 1154, "loss": 0.9824, "learning_rate": 1.3333333333333333e-05, "epoch": 0.06932409012131716, "percentage": 6.93, "elapsed_time": "0:25:03", "remaining_time": "5:36:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 85, "total_steps": 1154, "loss": 0.9676, "learning_rate": 1.4166666666666668e-05, "epoch": 0.07365684575389948, "percentage": 7.37, "elapsed_time": "0:26:37", "remaining_time": "5:34:45", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 90, "total_steps": 1154, "loss": 0.9463, "learning_rate": 1.5e-05, "epoch": 0.0779896013864818, "percentage": 7.8, "elapsed_time": "0:28:11", "remaining_time": "5:33:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 95, "total_steps": 1154, "loss": 0.9341, "learning_rate": 1.5833333333333333e-05, "epoch": 0.08232235701906412, "percentage": 8.23, "elapsed_time": "0:29:46", "remaining_time": "5:31:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 100, "total_steps": 1154, "loss": 0.9209, "learning_rate": 1.6666666666666667e-05, "epoch": 0.08665511265164645, "percentage": 8.67, "elapsed_time": "0:31:20", "remaining_time": "5:30:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 105, "total_steps": 1154, "loss": 0.8927, "learning_rate": 1.75e-05, "epoch": 0.09098786828422877, "percentage": 9.1, "elapsed_time": "0:32:54", "remaining_time": "5:28:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 110, "total_steps": 1154, "loss": 0.8755, "learning_rate": 1.8333333333333333e-05, "epoch": 0.09532062391681109, "percentage": 9.53, "elapsed_time": "0:34:28", "remaining_time": "5:27:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 115, "total_steps": 1154, "loss": 0.8453, "learning_rate": 1.9166666666666667e-05, "epoch": 0.09965337954939342, "percentage": 9.97, "elapsed_time": "0:36:02", "remaining_time": "5:25:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 120, "total_steps": 1154, "loss": 0.8081, "learning_rate": 2e-05, "epoch": 0.10398613518197573, "percentage": 10.4, "elapsed_time": "0:37:36", "remaining_time": "5:24:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 125, "total_steps": 1154, "loss": 0.7784, "learning_rate": 2.0833333333333336e-05, "epoch": 0.10831889081455806, "percentage": 10.83, "elapsed_time": "0:39:11", "remaining_time": "5:22:36", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 130, "total_steps": 1154, "loss": 0.7353, "learning_rate": 2.1666666666666667e-05, "epoch": 0.11265164644714037, "percentage": 11.27, "elapsed_time": "0:40:45", "remaining_time": "5:21:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 135, "total_steps": 1154, "loss": 0.7085, "learning_rate": 2.25e-05, "epoch": 0.1169844020797227, "percentage": 11.7, "elapsed_time": "0:42:20", "remaining_time": "5:19:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 140, "total_steps": 1154, "loss": 0.6753, "learning_rate": 2.3333333333333336e-05, "epoch": 0.12131715771230503, "percentage": 12.13, "elapsed_time": "0:43:54", "remaining_time": "5:17:59", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 145, "total_steps": 1154, "loss": 0.6448, "learning_rate": 2.4166666666666667e-05, "epoch": 0.12564991334488734, "percentage": 12.56, "elapsed_time": "0:45:28", "remaining_time": "5:16:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 150, "total_steps": 1154, "loss": 0.621, "learning_rate": 2.5e-05, "epoch": 0.12998266897746968, "percentage": 13.0, "elapsed_time": "0:47:02", "remaining_time": "5:14:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 155, "total_steps": 1154, "loss": 0.5997, "learning_rate": 2.5833333333333336e-05, "epoch": 0.134315424610052, "percentage": 13.43, "elapsed_time": "0:48:36", "remaining_time": "5:13:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 160, "total_steps": 1154, "loss": 0.5791, "learning_rate": 2.6666666666666667e-05, "epoch": 0.1386481802426343, "percentage": 13.86, "elapsed_time": "0:50:09", "remaining_time": "5:11:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 165, "total_steps": 1154, "loss": 0.57, "learning_rate": 2.7500000000000004e-05, "epoch": 0.14298093587521662, "percentage": 14.3, "elapsed_time": "0:51:43", "remaining_time": "5:10:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 170, "total_steps": 1154, "loss": 0.5645, "learning_rate": 2.8333333333333335e-05, "epoch": 0.14731369150779897, "percentage": 14.73, "elapsed_time": "0:53:17", "remaining_time": "5:08:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 175, "total_steps": 1154, "loss": 0.5583, "learning_rate": 2.916666666666667e-05, "epoch": 0.15164644714038128, "percentage": 15.16, "elapsed_time": "0:54:50", "remaining_time": "5:06:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 180, "total_steps": 1154, "loss": 0.555, "learning_rate": 3e-05, "epoch": 0.1559792027729636, "percentage": 15.6, "elapsed_time": "0:56:24", "remaining_time": "5:05:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 185, "total_steps": 1154, "loss": 0.5439, "learning_rate": 3.0833333333333335e-05, "epoch": 0.16031195840554593, "percentage": 16.03, "elapsed_time": "0:57:58", "remaining_time": "5:03:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 190, "total_steps": 1154, "loss": 0.537, "learning_rate": 3.1666666666666666e-05, "epoch": 0.16464471403812825, "percentage": 16.46, "elapsed_time": "0:59:31", "remaining_time": "5:02:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 195, "total_steps": 1154, "loss": 0.5381, "learning_rate": 3.2500000000000004e-05, "epoch": 0.16897746967071056, "percentage": 16.9, "elapsed_time": "1:01:05", "remaining_time": "5:00:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 200, "total_steps": 1154, "loss": 0.5344, "learning_rate": 3.3333333333333335e-05, "epoch": 0.1733102253032929, "percentage": 17.33, "elapsed_time": "1:02:39", "remaining_time": "4:58:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 205, "total_steps": 1154, "loss": 0.5291, "learning_rate": 3.4166666666666666e-05, "epoch": 0.17764298093587522, "percentage": 17.76, "elapsed_time": "1:04:12", "remaining_time": "4:57:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 210, "total_steps": 1154, "loss": 0.5342, "learning_rate": 3.5e-05, "epoch": 0.18197573656845753, "percentage": 18.2, "elapsed_time": "1:05:47", "remaining_time": "4:55:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 215, "total_steps": 1154, "loss": 0.5285, "learning_rate": 3.5833333333333335e-05, "epoch": 0.18630849220103987, "percentage": 18.63, "elapsed_time": "1:07:21", "remaining_time": "4:54:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 220, "total_steps": 1154, "loss": 0.5268, "learning_rate": 3.6666666666666666e-05, "epoch": 0.19064124783362218, "percentage": 19.06, "elapsed_time": "1:08:55", "remaining_time": "4:52:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 225, "total_steps": 1154, "loss": 0.5252, "learning_rate": 3.7500000000000003e-05, "epoch": 0.1949740034662045, "percentage": 19.5, "elapsed_time": "1:10:29", "remaining_time": "4:51:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 230, "total_steps": 1154, "loss": 0.5201, "learning_rate": 3.8333333333333334e-05, "epoch": 0.19930675909878684, "percentage": 19.93, "elapsed_time": "1:12:04", "remaining_time": "4:49:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 235, "total_steps": 1154, "loss": 0.5222, "learning_rate": 3.9166666666666665e-05, "epoch": 0.20363951473136915, "percentage": 20.36, "elapsed_time": "1:13:38", "remaining_time": "4:48:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 240, "total_steps": 1154, "loss": 0.5158, "learning_rate": 4e-05, "epoch": 0.20797227036395147, "percentage": 20.8, "elapsed_time": "1:15:13", "remaining_time": "4:46:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 245, "total_steps": 1154, "loss": 0.5127, "learning_rate": 4.0833333333333334e-05, "epoch": 0.2123050259965338, "percentage": 21.23, "elapsed_time": "1:16:47", "remaining_time": "4:44:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 250, "total_steps": 1154, "loss": 0.5181, "learning_rate": 4.166666666666667e-05, "epoch": 0.21663778162911612, "percentage": 21.66, "elapsed_time": "1:18:22", "remaining_time": "4:43:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 255, "total_steps": 1154, "loss": 0.5164, "learning_rate": 4.25e-05, "epoch": 0.22097053726169844, "percentage": 22.1, "elapsed_time": "1:19:56", "remaining_time": "4:41:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 260, "total_steps": 1154, "loss": 0.5149, "learning_rate": 4.3333333333333334e-05, "epoch": 0.22530329289428075, "percentage": 22.53, "elapsed_time": "1:21:29", "remaining_time": "4:40:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 265, "total_steps": 1154, "loss": 0.5137, "learning_rate": 4.4166666666666665e-05, "epoch": 0.2296360485268631, "percentage": 22.96, "elapsed_time": "1:23:04", "remaining_time": "4:38:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 270, "total_steps": 1154, "loss": 0.5066, "learning_rate": 4.5e-05, "epoch": 0.2339688041594454, "percentage": 23.4, "elapsed_time": "1:24:39", "remaining_time": "4:37:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 275, "total_steps": 1154, "loss": 0.5116, "learning_rate": 4.5833333333333334e-05, "epoch": 0.23830155979202772, "percentage": 23.83, "elapsed_time": "1:26:13", "remaining_time": "4:35:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 280, "total_steps": 1154, "loss": 0.5095, "learning_rate": 4.666666666666667e-05, "epoch": 0.24263431542461006, "percentage": 24.26, "elapsed_time": "1:27:47", "remaining_time": "4:34:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 285, "total_steps": 1154, "loss": 0.5077, "learning_rate": 4.75e-05, "epoch": 0.24696707105719237, "percentage": 24.7, "elapsed_time": "1:29:22", "remaining_time": "4:32:30", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 290, "total_steps": 1154, "loss": 0.5099, "learning_rate": 4.8333333333333334e-05, "epoch": 0.2512998266897747, "percentage": 25.13, "elapsed_time": "1:30:55", "remaining_time": "4:30:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 295, "total_steps": 1154, "loss": 0.5036, "learning_rate": 4.9166666666666665e-05, "epoch": 0.255632582322357, "percentage": 25.56, "elapsed_time": "1:32:29", "remaining_time": "4:29:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 300, "total_steps": 1154, "loss": 0.5085, "learning_rate": 5e-05, "epoch": 0.25996533795493937, "percentage": 26.0, "elapsed_time": "1:34:04", "remaining_time": "4:27:46", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 305, "total_steps": 1154, "loss": 0.4978, "learning_rate": 4.999577115486055e-05, "epoch": 0.26429809358752165, "percentage": 26.43, "elapsed_time": "1:35:37", "remaining_time": "4:26:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 310, "total_steps": 1154, "loss": 0.5095, "learning_rate": 4.998308605009268e-05, "epoch": 0.268630849220104, "percentage": 26.86, "elapsed_time": "1:37:10", "remaining_time": "4:24:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 315, "total_steps": 1154, "loss": 0.5037, "learning_rate": 4.996194897716389e-05, "epoch": 0.2729636048526863, "percentage": 27.3, "elapsed_time": "1:38:44", "remaining_time": "4:22:59", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 320, "total_steps": 1154, "loss": 0.5058, "learning_rate": 4.993236708690683e-05, "epoch": 0.2772963604852686, "percentage": 27.73, "elapsed_time": "1:40:19", "remaining_time": "4:21:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 325, "total_steps": 1154, "loss": 0.4998, "learning_rate": 4.9894350387100126e-05, "epoch": 0.28162911611785096, "percentage": 28.16, "elapsed_time": "1:41:53", "remaining_time": "4:19:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 330, "total_steps": 1154, "loss": 0.5007, "learning_rate": 4.984791173908267e-05, "epoch": 0.28596187175043325, "percentage": 28.6, "elapsed_time": "1:43:27", "remaining_time": "4:18:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 335, "total_steps": 1154, "loss": 0.5038, "learning_rate": 4.9793066853402536e-05, "epoch": 0.2902946273830156, "percentage": 29.03, "elapsed_time": "1:45:01", "remaining_time": "4:16:46", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 340, "total_steps": 1154, "loss": 0.5062, "learning_rate": 4.9729834284501995e-05, "epoch": 0.29462738301559793, "percentage": 29.46, "elapsed_time": "1:46:36", "remaining_time": "4:15:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 345, "total_steps": 1154, "loss": 0.5025, "learning_rate": 4.965823542444037e-05, "epoch": 0.2989601386481802, "percentage": 29.9, "elapsed_time": "1:48:10", "remaining_time": "4:13:40", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 350, "total_steps": 1154, "loss": 0.4999, "learning_rate": 4.9578294495656965e-05, "epoch": 0.30329289428076256, "percentage": 30.33, "elapsed_time": "1:49:43", "remaining_time": "4:12:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 355, "total_steps": 1154, "loss": 0.4978, "learning_rate": 4.949003854277644e-05, "epoch": 0.3076256499133449, "percentage": 30.76, "elapsed_time": "1:51:18", "remaining_time": "4:10:30", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 360, "total_steps": 1154, "loss": 0.4997, "learning_rate": 4.9393497423459376e-05, "epoch": 0.3119584055459272, "percentage": 31.2, "elapsed_time": "1:52:52", "remaining_time": "4:08:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 365, "total_steps": 1154, "loss": 0.5037, "learning_rate": 4.928870379830124e-05, "epoch": 0.31629116117850953, "percentage": 31.63, "elapsed_time": "1:54:26", "remaining_time": "4:07:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 370, "total_steps": 1154, "loss": 0.4982, "learning_rate": 4.9175693119783013e-05, "epoch": 0.32062391681109187, "percentage": 32.06, "elapsed_time": "1:56:01", "remaining_time": "4:05:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 375, "total_steps": 1154, "loss": 0.5013, "learning_rate": 4.905450362027738e-05, "epoch": 0.32495667244367415, "percentage": 32.5, "elapsed_time": "1:57:34", "remaining_time": "4:04:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 380, "total_steps": 1154, "loss": 0.5008, "learning_rate": 4.8925176299114416e-05, "epoch": 0.3292894280762565, "percentage": 32.93, "elapsed_time": "1:59:10", "remaining_time": "4:02:45", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 385, "total_steps": 1154, "loss": 0.4975, "learning_rate": 4.878775490871121e-05, "epoch": 0.33362218370883884, "percentage": 33.36, "elapsed_time": "2:00:46", "remaining_time": "4:01:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 390, "total_steps": 1154, "loss": 0.5047, "learning_rate": 4.864228593977006e-05, "epoch": 0.3379549393414211, "percentage": 33.8, "elapsed_time": "2:02:22", "remaining_time": "3:59:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 395, "total_steps": 1154, "loss": 0.4986, "learning_rate": 4.848881860555035e-05, "epoch": 0.34228769497400346, "percentage": 34.23, "elapsed_time": "2:03:57", "remaining_time": "3:58:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 400, "total_steps": 1154, "loss": 0.4981, "learning_rate": 4.832740482521931e-05, "epoch": 0.3466204506065858, "percentage": 34.66, "elapsed_time": "2:05:33", "remaining_time": "3:56:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 405, "total_steps": 1154, "loss": 0.4984, "learning_rate": 4.815809920628738e-05, "epoch": 0.3509532062391681, "percentage": 35.1, "elapsed_time": "2:07:09", "remaining_time": "3:55:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 410, "total_steps": 1154, "loss": 0.4942, "learning_rate": 4.7980959026134044e-05, "epoch": 0.35528596187175043, "percentage": 35.53, "elapsed_time": "2:08:45", "remaining_time": "3:53:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 415, "total_steps": 1154, "loss": 0.4903, "learning_rate": 4.7796044212630486e-05, "epoch": 0.3596187175043328, "percentage": 35.96, "elapsed_time": "2:10:21", "remaining_time": "3:52:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 420, "total_steps": 1154, "loss": 0.4957, "learning_rate": 4.7603417323865547e-05, "epoch": 0.36395147313691506, "percentage": 36.4, "elapsed_time": "2:11:57", "remaining_time": "3:50:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 425, "total_steps": 1154, "loss": 0.499, "learning_rate": 4.74031435269818e-05, "epoch": 0.3682842287694974, "percentage": 36.83, "elapsed_time": "2:13:33", "remaining_time": "3:49:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 430, "total_steps": 1154, "loss": 0.4959, "learning_rate": 4.7195290576129034e-05, "epoch": 0.37261698440207974, "percentage": 37.26, "elapsed_time": "2:15:09", "remaining_time": "3:47:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 435, "total_steps": 1154, "loss": 0.4944, "learning_rate": 4.697992878954255e-05, "epoch": 0.37694974003466203, "percentage": 37.69, "elapsed_time": "2:16:45", "remaining_time": "3:46:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 440, "total_steps": 1154, "loss": 0.4909, "learning_rate": 4.6757131025753886e-05, "epoch": 0.38128249566724437, "percentage": 38.13, "elapsed_time": "2:18:21", "remaining_time": "3:44:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 445, "total_steps": 1154, "loss": 0.4966, "learning_rate": 4.652697265894228e-05, "epoch": 0.3856152512998267, "percentage": 38.56, "elapsed_time": "2:19:57", "remaining_time": "3:42:59", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 450, "total_steps": 1154, "loss": 0.4956, "learning_rate": 4.628953155343499e-05, "epoch": 0.389948006932409, "percentage": 38.99, "elapsed_time": "2:21:33", "remaining_time": "3:41:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 455, "total_steps": 1154, "loss": 0.4973, "learning_rate": 4.604488803736523e-05, "epoch": 0.39428076256499134, "percentage": 39.43, "elapsed_time": "2:23:09", "remaining_time": "3:39:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 460, "total_steps": 1154, "loss": 0.4903, "learning_rate": 4.579312487549649e-05, "epoch": 0.3986135181975737, "percentage": 39.86, "elapsed_time": "2:24:45", "remaining_time": "3:38:24", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 465, "total_steps": 1154, "loss": 0.4999, "learning_rate": 4.553432724122265e-05, "epoch": 0.40294627383015597, "percentage": 40.29, "elapsed_time": "2:26:21", "remaining_time": "3:36:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 470, "total_steps": 1154, "loss": 0.4967, "learning_rate": 4.526858268775313e-05, "epoch": 0.4072790294627383, "percentage": 40.73, "elapsed_time": "2:27:57", "remaining_time": "3:35:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 475, "total_steps": 1154, "loss": 0.4936, "learning_rate": 4.499598111849299e-05, "epoch": 0.41161178509532065, "percentage": 41.16, "elapsed_time": "2:29:33", "remaining_time": "3:33:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 480, "total_steps": 1154, "loss": 0.493, "learning_rate": 4.471661475662792e-05, "epoch": 0.41594454072790293, "percentage": 41.59, "elapsed_time": "2:31:09", "remaining_time": "3:32:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 485, "total_steps": 1154, "loss": 0.5002, "learning_rate": 4.443057811392445e-05, "epoch": 0.4202772963604853, "percentage": 42.03, "elapsed_time": "2:32:44", "remaining_time": "3:30:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 490, "total_steps": 1154, "loss": 0.4983, "learning_rate": 4.413796795875586e-05, "epoch": 0.4246100519930676, "percentage": 42.46, "elapsed_time": "2:34:20", "remaining_time": "3:29:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 495, "total_steps": 1154, "loss": 0.4949, "learning_rate": 4.383888328336476e-05, "epoch": 0.4289428076256499, "percentage": 42.89, "elapsed_time": "2:35:56", "remaining_time": "3:27:36", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 500, "total_steps": 1154, "loss": 0.4953, "learning_rate": 4.3533425270373216e-05, "epoch": 0.43327556325823224, "percentage": 43.33, "elapsed_time": "2:37:32", "remaining_time": "3:26:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 500, "total_steps": 1154, "eval_loss": 0.4938061535358429, "epoch": 0.43327556325823224, "percentage": 43.33, "elapsed_time": "2:39:53", "remaining_time": "3:29:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 505, "total_steps": 1154, "loss": 0.4934, "learning_rate": 4.3221697258551906e-05, "epoch": 0.4376083188908146, "percentage": 43.76, "elapsed_time": "2:41:29", "remaining_time": "3:27:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 510, "total_steps": 1154, "loss": 0.4934, "learning_rate": 4.2903804707859835e-05, "epoch": 0.44194107452339687, "percentage": 44.19, "elapsed_time": "2:43:05", "remaining_time": "3:25:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 515, "total_steps": 1154, "loss": 0.4927, "learning_rate": 4.257985516376644e-05, "epoch": 0.4462738301559792, "percentage": 44.63, "elapsed_time": "2:44:41", "remaining_time": "3:24:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 520, "total_steps": 1154, "loss": 0.4935, "learning_rate": 4.224995822086812e-05, "epoch": 0.4506065857885615, "percentage": 45.06, "elapsed_time": "2:46:16", "remaining_time": "3:22:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 525, "total_steps": 1154, "loss": 0.4947, "learning_rate": 4.191422548581154e-05, "epoch": 0.45493934142114384, "percentage": 45.49, "elapsed_time": "2:47:52", "remaining_time": "3:21:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 530, "total_steps": 1154, "loss": 0.4935, "learning_rate": 4.157277053953631e-05, "epoch": 0.4592720970537262, "percentage": 45.93, "elapsed_time": "2:49:28", "remaining_time": "3:19:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 535, "total_steps": 1154, "loss": 0.4975, "learning_rate": 4.1225708898849594e-05, "epoch": 0.46360485268630847, "percentage": 46.36, "elapsed_time": "2:51:04", "remaining_time": "3:17:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 540, "total_steps": 1154, "loss": 0.4923, "learning_rate": 4.0873157977346e-05, "epoch": 0.4679376083188908, "percentage": 46.79, "elapsed_time": "2:52:40", "remaining_time": "3:16:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 545, "total_steps": 1154, "loss": 0.4939, "learning_rate": 4.051523704568557e-05, "epoch": 0.47227036395147315, "percentage": 47.23, "elapsed_time": "2:54:16", "remaining_time": "3:14:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 550, "total_steps": 1154, "loss": 0.4927, "learning_rate": 4.0152067191243696e-05, "epoch": 0.47660311958405543, "percentage": 47.66, "elapsed_time": "2:55:52", "remaining_time": "3:13:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 555, "total_steps": 1154, "loss": 0.4908, "learning_rate": 3.978377127714628e-05, "epoch": 0.4809358752166378, "percentage": 48.09, "elapsed_time": "2:57:28", "remaining_time": "3:11:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 560, "total_steps": 1154, "loss": 0.4898, "learning_rate": 3.941047390070419e-05, "epoch": 0.4852686308492201, "percentage": 48.53, "elapsed_time": "2:59:04", "remaining_time": "3:09:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 565, "total_steps": 1154, "loss": 0.4899, "learning_rate": 3.903230135126113e-05, "epoch": 0.4896013864818024, "percentage": 48.96, "elapsed_time": "3:00:39", "remaining_time": "3:08:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 570, "total_steps": 1154, "loss": 0.4914, "learning_rate": 3.864938156746891e-05, "epoch": 0.49393414211438474, "percentage": 49.39, "elapsed_time": "3:02:15", "remaining_time": "3:06:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 575, "total_steps": 1154, "loss": 0.4922, "learning_rate": 3.8261844094004815e-05, "epoch": 0.4982668977469671, "percentage": 49.83, "elapsed_time": "3:03:51", "remaining_time": "3:05:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 580, "total_steps": 1154, "loss": 0.4964, "learning_rate": 3.7869820037745776e-05, "epoch": 0.5025996533795494, "percentage": 50.26, "elapsed_time": "3:05:27", "remaining_time": "3:03:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 585, "total_steps": 1154, "loss": 0.4945, "learning_rate": 3.747344202341373e-05, "epoch": 0.5069324090121318, "percentage": 50.69, "elapsed_time": "3:07:03", "remaining_time": "3:01:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 590, "total_steps": 1154, "loss": 0.499, "learning_rate": 3.707284414870786e-05, "epoch": 0.511265164644714, "percentage": 51.13, "elapsed_time": "3:08:38", "remaining_time": "3:00:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 595, "total_steps": 1154, "loss": 0.4947, "learning_rate": 3.666816193893817e-05, "epoch": 0.5155979202772963, "percentage": 51.56, "elapsed_time": "3:10:14", "remaining_time": "2:58:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 600, "total_steps": 1154, "loss": 0.4905, "learning_rate": 3.6259532301176335e-05, "epoch": 0.5199306759098787, "percentage": 51.99, "elapsed_time": "3:11:48", "remaining_time": "2:57:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 605, "total_steps": 1154, "loss": 0.4974, "learning_rate": 3.5847093477938956e-05, "epoch": 0.524263431542461, "percentage": 52.43, "elapsed_time": "3:13:23", "remaining_time": "2:55:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 610, "total_steps": 1154, "loss": 0.4927, "learning_rate": 3.543098500041906e-05, "epoch": 0.5285961871750433, "percentage": 52.86, "elapsed_time": "3:14:57", "remaining_time": "2:53:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 615, "total_steps": 1154, "loss": 0.4907, "learning_rate": 3.501134764128167e-05, "epoch": 0.5329289428076257, "percentage": 53.29, "elapsed_time": "3:16:32", "remaining_time": "2:52:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 620, "total_steps": 1154, "loss": 0.491, "learning_rate": 3.458832336703929e-05, "epoch": 0.537261698440208, "percentage": 53.73, "elapsed_time": "3:18:07", "remaining_time": "2:50:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 625, "total_steps": 1154, "loss": 0.4942, "learning_rate": 3.416205529002363e-05, "epoch": 0.5415944540727903, "percentage": 54.16, "elapsed_time": "3:19:43", "remaining_time": "2:49:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 630, "total_steps": 1154, "loss": 0.493, "learning_rate": 3.37326876199695e-05, "epoch": 0.5459272097053726, "percentage": 54.59, "elapsed_time": "3:21:19", "remaining_time": "2:47:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 635, "total_steps": 1154, "loss": 0.4942, "learning_rate": 3.3300365615227685e-05, "epoch": 0.550259965337955, "percentage": 55.03, "elapsed_time": "3:22:55", "remaining_time": "2:45:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 640, "total_steps": 1154, "loss": 0.4946, "learning_rate": 3.286523553362287e-05, "epoch": 0.5545927209705372, "percentage": 55.46, "elapsed_time": "3:24:31", "remaining_time": "2:44:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 645, "total_steps": 1154, "loss": 0.4907, "learning_rate": 3.242744458297348e-05, "epoch": 0.5589254766031195, "percentage": 55.89, "elapsed_time": "3:26:07", "remaining_time": "2:42:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 650, "total_steps": 1154, "loss": 0.4889, "learning_rate": 3.1987140871290236e-05, "epoch": 0.5632582322357019, "percentage": 56.33, "elapsed_time": "3:27:41", "remaining_time": "2:41:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 655, "total_steps": 1154, "loss": 0.4935, "learning_rate": 3.154447335667001e-05, "epoch": 0.5675909878682842, "percentage": 56.76, "elapsed_time": "3:29:17", "remaining_time": "2:39:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 660, "total_steps": 1154, "loss": 0.4925, "learning_rate": 3.1099591796902215e-05, "epoch": 0.5719237435008665, "percentage": 57.19, "elapsed_time": "3:30:53", "remaining_time": "2:37:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 665, "total_steps": 1154, "loss": 0.4892, "learning_rate": 3.0652646698804585e-05, "epoch": 0.5762564991334489, "percentage": 57.63, "elapsed_time": "3:32:29", "remaining_time": "2:36:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 670, "total_steps": 1154, "loss": 0.4918, "learning_rate": 3.0203789267305567e-05, "epoch": 0.5805892547660312, "percentage": 58.06, "elapsed_time": "3:34:04", "remaining_time": "2:34:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 675, "total_steps": 1154, "loss": 0.4885, "learning_rate": 2.975317135429056e-05, "epoch": 0.5849220103986135, "percentage": 58.49, "elapsed_time": "3:35:40", "remaining_time": "2:33:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 680, "total_steps": 1154, "loss": 0.4889, "learning_rate": 2.930094540722927e-05, "epoch": 0.5892547660311959, "percentage": 58.93, "elapsed_time": "3:37:16", "remaining_time": "2:31:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 685, "total_steps": 1154, "loss": 0.4936, "learning_rate": 2.884726441760155e-05, "epoch": 0.5935875216637782, "percentage": 59.36, "elapsed_time": "3:38:52", "remaining_time": "2:29:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 690, "total_steps": 1154, "loss": 0.4925, "learning_rate": 2.8392281869139213e-05, "epoch": 0.5979202772963604, "percentage": 59.79, "elapsed_time": "3:40:28", "remaining_time": "2:28:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 695, "total_steps": 1154, "loss": 0.49, "learning_rate": 2.7936151685901373e-05, "epoch": 0.6022530329289428, "percentage": 60.23, "elapsed_time": "3:42:04", "remaining_time": "2:26:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 700, "total_steps": 1154, "loss": 0.4919, "learning_rate": 2.747902818020067e-05, "epoch": 0.6065857885615251, "percentage": 60.66, "elapsed_time": "3:43:39", "remaining_time": "2:25:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 705, "total_steps": 1154, "loss": 0.4928, "learning_rate": 2.7021066000398344e-05, "epoch": 0.6109185441941074, "percentage": 61.09, "elapsed_time": "3:45:15", "remaining_time": "2:23:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 710, "total_steps": 1154, "loss": 0.489, "learning_rate": 2.6562420078585433e-05, "epoch": 0.6152512998266898, "percentage": 61.53, "elapsed_time": "3:46:51", "remaining_time": "2:21:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 715, "total_steps": 1154, "loss": 0.4919, "learning_rate": 2.6103245578168106e-05, "epoch": 0.6195840554592721, "percentage": 61.96, "elapsed_time": "3:48:27", "remaining_time": "2:20:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 720, "total_steps": 1154, "loss": 0.4883, "learning_rate": 2.564369784137472e-05, "epoch": 0.6239168110918544, "percentage": 62.39, "elapsed_time": "3:50:03", "remaining_time": "2:18:40", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 725, "total_steps": 1154, "loss": 0.4909, "learning_rate": 2.5183932336702297e-05, "epoch": 0.6282495667244368, "percentage": 62.82, "elapsed_time": "3:51:37", "remaining_time": "2:17:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 730, "total_steps": 1154, "loss": 0.4878, "learning_rate": 2.4724104606320445e-05, "epoch": 0.6325823223570191, "percentage": 63.26, "elapsed_time": "3:53:11", "remaining_time": "2:15:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 735, "total_steps": 1154, "loss": 0.4901, "learning_rate": 2.426437021345015e-05, "epoch": 0.6369150779896013, "percentage": 63.69, "elapsed_time": "3:54:46", "remaining_time": "2:13:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 740, "total_steps": 1154, "loss": 0.4906, "learning_rate": 2.3804884689735642e-05, "epoch": 0.6412478336221837, "percentage": 64.12, "elapsed_time": "3:56:20", "remaining_time": "2:12:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 745, "total_steps": 1154, "loss": 0.494, "learning_rate": 2.3345803482626797e-05, "epoch": 0.645580589254766, "percentage": 64.56, "elapsed_time": "3:57:55", "remaining_time": "2:10:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 750, "total_steps": 1154, "loss": 0.4886, "learning_rate": 2.288728190279014e-05, "epoch": 0.6499133448873483, "percentage": 64.99, "elapsed_time": "3:59:30", "remaining_time": "2:09:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 755, "total_steps": 1154, "loss": 0.4908, "learning_rate": 2.2429475071565987e-05, "epoch": 0.6542461005199307, "percentage": 65.42, "elapsed_time": "4:01:02", "remaining_time": "2:07:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 760, "total_steps": 1154, "loss": 0.4896, "learning_rate": 2.1972537868489797e-05, "epoch": 0.658578856152513, "percentage": 65.86, "elapsed_time": "4:02:38", "remaining_time": "2:05:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 765, "total_steps": 1154, "loss": 0.4885, "learning_rate": 2.151662487889518e-05, "epoch": 0.6629116117850953, "percentage": 66.29, "elapsed_time": "4:04:13", "remaining_time": "2:04:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 770, "total_steps": 1154, "loss": 0.4873, "learning_rate": 2.1061890341616558e-05, "epoch": 0.6672443674176777, "percentage": 66.72, "elapsed_time": "4:05:49", "remaining_time": "2:02:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 775, "total_steps": 1154, "loss": 0.4853, "learning_rate": 2.060848809680893e-05, "epoch": 0.67157712305026, "percentage": 67.16, "elapsed_time": "4:07:25", "remaining_time": "2:01:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 780, "total_steps": 1154, "loss": 0.4849, "learning_rate": 2.0156571533902627e-05, "epoch": 0.6759098786828422, "percentage": 67.59, "elapsed_time": "4:09:01", "remaining_time": "1:59:24", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 785, "total_steps": 1154, "loss": 0.488, "learning_rate": 1.97062935397105e-05, "epoch": 0.6802426343154246, "percentage": 68.02, "elapsed_time": "4:10:35", "remaining_time": "1:57:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 790, "total_steps": 1154, "loss": 0.4867, "learning_rate": 1.9257806446705116e-05, "epoch": 0.6845753899480069, "percentage": 68.46, "elapsed_time": "4:12:11", "remaining_time": "1:56:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 795, "total_steps": 1154, "loss": 0.4911, "learning_rate": 1.8811261981483548e-05, "epoch": 0.6889081455805892, "percentage": 68.89, "elapsed_time": "4:13:46", "remaining_time": "1:54:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 800, "total_steps": 1154, "loss": 0.4888, "learning_rate": 1.8366811213437092e-05, "epoch": 0.6932409012131716, "percentage": 69.32, "elapsed_time": "4:15:22", "remaining_time": "1:53:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 805, "total_steps": 1154, "loss": 0.4903, "learning_rate": 1.7924604503643367e-05, "epoch": 0.6975736568457539, "percentage": 69.76, "elapsed_time": "4:16:58", "remaining_time": "1:51:24", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 810, "total_steps": 1154, "loss": 0.4874, "learning_rate": 1.7484791453998006e-05, "epoch": 0.7019064124783362, "percentage": 70.19, "elapsed_time": "4:18:34", "remaining_time": "1:49:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 815, "total_steps": 1154, "loss": 0.4895, "learning_rate": 1.7047520856603183e-05, "epoch": 0.7062391681109186, "percentage": 70.62, "elapsed_time": "4:20:10", "remaining_time": "1:48:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 820, "total_steps": 1154, "loss": 0.4909, "learning_rate": 1.6612940643430138e-05, "epoch": 0.7105719237435009, "percentage": 71.06, "elapsed_time": "4:21:45", "remaining_time": "1:46:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 825, "total_steps": 1154, "loss": 0.4898, "learning_rate": 1.618119783627263e-05, "epoch": 0.7149046793760832, "percentage": 71.49, "elapsed_time": "4:23:21", "remaining_time": "1:45:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 830, "total_steps": 1154, "loss": 0.4886, "learning_rate": 1.5752438497008405e-05, "epoch": 0.7192374350086655, "percentage": 71.92, "elapsed_time": "4:24:57", "remaining_time": "1:43:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 835, "total_steps": 1154, "loss": 0.4893, "learning_rate": 1.5326807678185267e-05, "epoch": 0.7235701906412478, "percentage": 72.36, "elapsed_time": "4:26:33", "remaining_time": "1:41:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 840, "total_steps": 1154, "loss": 0.4891, "learning_rate": 1.490444937394879e-05, "epoch": 0.7279029462738301, "percentage": 72.79, "elapsed_time": "4:28:09", "remaining_time": "1:40:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 845, "total_steps": 1154, "loss": 0.4968, "learning_rate": 1.4485506471327914e-05, "epoch": 0.7322357019064125, "percentage": 73.22, "elapsed_time": "4:29:45", "remaining_time": "1:38:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 850, "total_steps": 1154, "loss": 0.4861, "learning_rate": 1.407012070189524e-05, "epoch": 0.7365684575389948, "percentage": 73.66, "elapsed_time": "4:31:20", "remaining_time": "1:37:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 855, "total_steps": 1154, "loss": 0.4888, "learning_rate": 1.3658432593818149e-05, "epoch": 0.7409012131715771, "percentage": 74.09, "elapsed_time": "4:32:53", "remaining_time": "1:35:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 860, "total_steps": 1154, "loss": 0.4879, "learning_rate": 1.325058142431701e-05, "epoch": 0.7452339688041595, "percentage": 74.52, "elapsed_time": "4:34:28", "remaining_time": "1:33:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 865, "total_steps": 1154, "loss": 0.4909, "learning_rate": 1.2846705172546675e-05, "epoch": 0.7495667244367418, "percentage": 74.96, "elapsed_time": "4:36:02", "remaining_time": "1:32:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 870, "total_steps": 1154, "loss": 0.485, "learning_rate": 1.2446940472917099e-05, "epoch": 0.7538994800693241, "percentage": 75.39, "elapsed_time": "4:37:37", "remaining_time": "1:30:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 875, "total_steps": 1154, "loss": 0.4891, "learning_rate": 1.2051422568868833e-05, "epoch": 0.7582322357019065, "percentage": 75.82, "elapsed_time": "4:39:11", "remaining_time": "1:29:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 880, "total_steps": 1154, "loss": 0.4903, "learning_rate": 1.1660285267119167e-05, "epoch": 0.7625649913344887, "percentage": 76.26, "elapsed_time": "4:40:45", "remaining_time": "1:27:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 885, "total_steps": 1154, "loss": 0.4895, "learning_rate": 1.1273660892394297e-05, "epoch": 0.766897746967071, "percentage": 76.69, "elapsed_time": "4:42:20", "remaining_time": "1:25:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 890, "total_steps": 1154, "loss": 0.4843, "learning_rate": 1.0891680242662835e-05, "epoch": 0.7712305025996534, "percentage": 77.12, "elapsed_time": "4:43:54", "remaining_time": "1:24:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 895, "total_steps": 1154, "loss": 0.4894, "learning_rate": 1.051447254488591e-05, "epoch": 0.7755632582322357, "percentage": 77.56, "elapsed_time": "4:45:28", "remaining_time": "1:22:36", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 900, "total_steps": 1154, "loss": 0.4884, "learning_rate": 1.0142165411298662e-05, "epoch": 0.779896013864818, "percentage": 77.99, "elapsed_time": "4:47:03", "remaining_time": "1:21:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 905, "total_steps": 1154, "loss": 0.49, "learning_rate": 9.774884796238085e-06, "epoch": 0.7842287694974004, "percentage": 78.42, "elapsed_time": "4:48:37", "remaining_time": "1:19:24", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 910, "total_steps": 1154, "loss": 0.4884, "learning_rate": 9.412754953531663e-06, "epoch": 0.7885615251299827, "percentage": 78.86, "elapsed_time": "4:50:10", "remaining_time": "1:17:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 915, "total_steps": 1154, "loss": 0.4883, "learning_rate": 9.055898394461423e-06, "epoch": 0.792894280762565, "percentage": 79.29, "elapsed_time": "4:51:45", "remaining_time": "1:16:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 920, "total_steps": 1154, "loss": 0.4869, "learning_rate": 8.704435846317386e-06, "epoch": 0.7972270363951474, "percentage": 79.72, "elapsed_time": "4:53:21", "remaining_time": "1:14:36", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 925, "total_steps": 1154, "loss": 0.4887, "learning_rate": 8.358486211554637e-06, "epoch": 0.8015597920277296, "percentage": 80.16, "elapsed_time": "4:54:57", "remaining_time": "1:13:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 930, "total_steps": 1154, "loss": 0.4852, "learning_rate": 8.018166527567672e-06, "epoch": 0.8058925476603119, "percentage": 80.59, "elapsed_time": "4:56:32", "remaining_time": "1:11:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 935, "total_steps": 1154, "loss": 0.4865, "learning_rate": 7.683591927095824e-06, "epoch": 0.8102253032928943, "percentage": 81.02, "elapsed_time": "4:58:08", "remaining_time": "1:09:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 940, "total_steps": 1154, "loss": 0.4902, "learning_rate": 7.354875599272928e-06, "epoch": 0.8145580589254766, "percentage": 81.46, "elapsed_time": "4:59:43", "remaining_time": "1:08:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 945, "total_steps": 1154, "loss": 0.4879, "learning_rate": 7.0321287513346074e-06, "epoch": 0.8188908145580589, "percentage": 81.89, "elapsed_time": "5:01:19", "remaining_time": "1:06:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 950, "total_steps": 1154, "loss": 0.4905, "learning_rate": 6.715460570995988e-06, "epoch": 0.8232235701906413, "percentage": 82.32, "elapsed_time": "5:02:55", "remaining_time": "1:05:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 955, "total_steps": 1154, "loss": 0.492, "learning_rate": 6.404978189512681e-06, "epoch": 0.8275563258232236, "percentage": 82.76, "elapsed_time": "5:04:31", "remaining_time": "1:03:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 960, "total_steps": 1154, "loss": 0.4855, "learning_rate": 6.100786645437481e-06, "epoch": 0.8318890814558059, "percentage": 83.19, "elapsed_time": "5:06:06", "remaining_time": "1:01:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 965, "total_steps": 1154, "loss": 0.4942, "learning_rate": 5.8029888490850005e-06, "epoch": 0.8362218370883883, "percentage": 83.62, "elapsed_time": "5:07:41", "remaining_time": "1:00:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 970, "total_steps": 1154, "loss": 0.4829, "learning_rate": 5.511685547716328e-06, "epoch": 0.8405545927209706, "percentage": 84.06, "elapsed_time": "5:09:17", "remaining_time": "0:58:40", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 975, "total_steps": 1154, "loss": 0.4856, "learning_rate": 5.226975291455477e-06, "epoch": 0.8448873483535528, "percentage": 84.49, "elapsed_time": "5:10:52", "remaining_time": "0:57:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 980, "total_steps": 1154, "loss": 0.4852, "learning_rate": 4.9489543999491045e-06, "epoch": 0.8492201039861352, "percentage": 84.92, "elapsed_time": "5:12:27", "remaining_time": "0:55:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 985, "total_steps": 1154, "loss": 0.4885, "learning_rate": 4.67771692978087e-06, "epoch": 0.8535528596187175, "percentage": 85.36, "elapsed_time": "5:14:02", "remaining_time": "0:53:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 990, "total_steps": 1154, "loss": 0.4867, "learning_rate": 4.413354642651369e-06, "epoch": 0.8578856152512998, "percentage": 85.79, "elapsed_time": "5:15:38", "remaining_time": "0:52:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 995, "total_steps": 1154, "loss": 0.4839, "learning_rate": 4.1559569743344405e-06, "epoch": 0.8622183708838822, "percentage": 86.22, "elapsed_time": "5:17:13", "remaining_time": "0:50:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1000, "total_steps": 1154, "loss": 0.4909, "learning_rate": 3.90561100442036e-06, "epoch": 0.8665511265164645, "percentage": 86.66, "elapsed_time": "5:18:48", "remaining_time": "0:49:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1000, "total_steps": 1154, "eval_loss": 0.48741263151168823, "epoch": 0.8665511265164645, "percentage": 86.66, "elapsed_time": "5:21:07", "remaining_time": "0:49:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1005, "total_steps": 1154, "loss": 0.4912, "learning_rate": 3.662401426856177e-06, "epoch": 0.8708838821490468, "percentage": 87.09, "elapsed_time": "5:22:43", "remaining_time": "0:47:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1010, "total_steps": 1154, "loss": 0.4882, "learning_rate": 3.4264105212930915e-06, "epoch": 0.8752166377816292, "percentage": 87.52, "elapsed_time": "5:24:18", "remaining_time": "0:46:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1015, "total_steps": 1154, "loss": 0.4834, "learning_rate": 3.197718125250618e-06, "epoch": 0.8795493934142115, "percentage": 87.95, "elapsed_time": "5:25:54", "remaining_time": "0:44:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1020, "total_steps": 1154, "loss": 0.4881, "learning_rate": 2.9764016071069434e-06, "epoch": 0.8838821490467937, "percentage": 88.39, "elapsed_time": "5:27:30", "remaining_time": "0:43:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1025, "total_steps": 1154, "loss": 0.4852, "learning_rate": 2.7625358399246376e-06, "epoch": 0.8882149046793761, "percentage": 88.82, "elapsed_time": "5:29:06", "remaining_time": "0:41:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1030, "total_steps": 1154, "loss": 0.4879, "learning_rate": 2.5561931761205082e-06, "epoch": 0.8925476603119584, "percentage": 89.25, "elapsed_time": "5:30:41", "remaining_time": "0:39:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1035, "total_steps": 1154, "loss": 0.4865, "learning_rate": 2.3574434229882145e-06, "epoch": 0.8968804159445407, "percentage": 89.69, "elapsed_time": "5:32:17", "remaining_time": "0:38:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1040, "total_steps": 1154, "loss": 0.4855, "learning_rate": 2.166353819081968e-06, "epoch": 0.901213171577123, "percentage": 90.12, "elapsed_time": "5:33:53", "remaining_time": "0:36:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1045, "total_steps": 1154, "loss": 0.4868, "learning_rate": 1.982989011469172e-06, "epoch": 0.9055459272097054, "percentage": 90.55, "elapsed_time": "5:35:28", "remaining_time": "0:34:59", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1050, "total_steps": 1154, "loss": 0.487, "learning_rate": 1.8074110338598682e-06, "epoch": 0.9098786828422877, "percentage": 90.99, "elapsed_time": "5:36:59", "remaining_time": "0:33:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1055, "total_steps": 1154, "loss": 0.4881, "learning_rate": 1.6396792856202298e-06, "epoch": 0.91421143847487, "percentage": 91.42, "elapsed_time": "5:38:32", "remaining_time": "0:31:46", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1060, "total_steps": 1154, "loss": 0.4877, "learning_rate": 1.479850511677322e-06, "epoch": 0.9185441941074524, "percentage": 91.85, "elapsed_time": "5:40:09", "remaining_time": "0:30:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1065, "total_steps": 1154, "loss": 0.4886, "learning_rate": 1.3279787833218537e-06, "epoch": 0.9228769497400346, "percentage": 92.29, "elapsed_time": "5:41:43", "remaining_time": "0:28:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1070, "total_steps": 1154, "loss": 0.4856, "learning_rate": 1.1841154799154374e-06, "epoch": 0.9272097053726169, "percentage": 92.72, "elapsed_time": "5:43:18", "remaining_time": "0:26:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1075, "total_steps": 1154, "loss": 0.4872, "learning_rate": 1.0483092715085879e-06, "epoch": 0.9315424610051993, "percentage": 93.15, "elapsed_time": "5:44:54", "remaining_time": "0:25:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1080, "total_steps": 1154, "loss": 0.4867, "learning_rate": 9.206061023752516e-07, "epoch": 0.9358752166377816, "percentage": 93.59, "elapsed_time": "5:46:30", "remaining_time": "0:23:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1085, "total_steps": 1154, "loss": 0.4868, "learning_rate": 8.010491754695177e-07, "epoch": 0.9402079722703639, "percentage": 94.02, "elapsed_time": "5:48:05", "remaining_time": "0:22:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1090, "total_steps": 1154, "loss": 0.4911, "learning_rate": 6.896789378097179e-07, "epoch": 0.9445407279029463, "percentage": 94.45, "elapsed_time": "5:49:41", "remaining_time": "0:20:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1095, "total_steps": 1154, "loss": 0.485, "learning_rate": 5.865330667949115e-07, "epoch": 0.9488734835355286, "percentage": 94.89, "elapsed_time": "5:51:17", "remaining_time": "0:18:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1100, "total_steps": 1154, "loss": 0.4864, "learning_rate": 4.916464574583251e-07, "epoch": 0.9532062391681109, "percentage": 95.32, "elapsed_time": "5:52:52", "remaining_time": "0:17:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1105, "total_steps": 1154, "loss": 0.4873, "learning_rate": 4.050512106620913e-07, "epoch": 0.9575389948006933, "percentage": 95.75, "elapsed_time": "5:54:28", "remaining_time": "0:15:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1110, "total_steps": 1154, "loss": 0.4885, "learning_rate": 3.267766222372931e-07, "epoch": 0.9618717504332756, "percentage": 96.19, "elapsed_time": "5:56:04", "remaining_time": "0:14:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1115, "total_steps": 1154, "loss": 0.4869, "learning_rate": 2.568491730729539e-07, "epoch": 0.9662045060658578, "percentage": 96.62, "elapsed_time": "5:57:40", "remaining_time": "0:12:30", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1120, "total_steps": 1154, "loss": 0.489, "learning_rate": 1.95292520157353e-07, "epoch": 0.9705372616984402, "percentage": 97.05, "elapsed_time": "5:59:16", "remaining_time": "0:10:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1125, "total_steps": 1154, "loss": 0.4871, "learning_rate": 1.4212748857468926e-07, "epoch": 0.9748700173310225, "percentage": 97.49, "elapsed_time": "6:00:51", "remaining_time": "0:09:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1130, "total_steps": 1154, "loss": 0.4875, "learning_rate": 9.737206445979696e-08, "epoch": 0.9792027729636048, "percentage": 97.92, "elapsed_time": "6:02:27", "remaining_time": "0:07:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1135, "total_steps": 1154, "loss": 0.4895, "learning_rate": 6.104138891329659e-08, "epoch": 0.9835355285961872, "percentage": 98.35, "elapsed_time": "6:04:03", "remaining_time": "0:06:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1140, "total_steps": 1154, "loss": 0.4857, "learning_rate": 3.314775287923677e-08, "epoch": 0.9878682842287695, "percentage": 98.79, "elapsed_time": "6:05:39", "remaining_time": "0:04:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1145, "total_steps": 1154, "loss": 0.4866, "learning_rate": 1.3700592986998217e-08, "epoch": 0.9922010398613518, "percentage": 99.22, "elapsed_time": "6:07:15", "remaining_time": "0:02:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1150, "total_steps": 1154, "loss": 0.4893, "learning_rate": 2.7064883587807345e-09, "epoch": 0.9965337954939342, "percentage": 99.65, "elapsed_time": "6:08:51", "remaining_time": "0:01:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1154, "total_steps": 1154, "epoch": 1.0, "percentage": 100.0, "elapsed_time": "6:10:08", "remaining_time": "0:00:00", "throughput": "0.00", "total_tokens": 0} | |