laaj_5k_3epochs_all_Scode / trainer_log.jsonl
Rakancorle1's picture
Training in progress, step 1875
664f2da verified
{"current_steps": 10, "total_steps": 1875, "loss": 1.6713, "lr": 9.999298177883903e-05, "epoch": 0.016, "percentage": 0.53, "elapsed_time": "0:01:31", "remaining_time": "4:44:47"}
{"current_steps": 20, "total_steps": 1875, "loss": 0.9562, "lr": 9.997192908557323e-05, "epoch": 0.032, "percentage": 1.07, "elapsed_time": "0:02:57", "remaining_time": "4:33:54"}
{"current_steps": 30, "total_steps": 1875, "loss": 0.7998, "lr": 9.993684783030088e-05, "epoch": 0.048, "percentage": 1.6, "elapsed_time": "0:04:22", "remaining_time": "4:29:11"}
{"current_steps": 40, "total_steps": 1875, "loss": 0.7019, "lr": 9.988774786134234e-05, "epoch": 0.064, "percentage": 2.13, "elapsed_time": "0:05:47", "remaining_time": "4:26:03"}
{"current_steps": 50, "total_steps": 1875, "loss": 0.6965, "lr": 9.982464296247522e-05, "epoch": 0.08, "percentage": 2.67, "elapsed_time": "0:07:13", "remaining_time": "4:23:41"}
{"current_steps": 60, "total_steps": 1875, "loss": 0.6417, "lr": 9.974755084906502e-05, "epoch": 0.096, "percentage": 3.2, "elapsed_time": "0:08:38", "remaining_time": "4:21:33"}
{"current_steps": 70, "total_steps": 1875, "loss": 0.6388, "lr": 9.965649316309178e-05, "epoch": 0.112, "percentage": 3.73, "elapsed_time": "0:10:04", "remaining_time": "4:19:40"}
{"current_steps": 80, "total_steps": 1875, "loss": 0.6665, "lr": 9.955149546707465e-05, "epoch": 0.128, "percentage": 4.27, "elapsed_time": "0:11:29", "remaining_time": "4:17:44"}
{"current_steps": 90, "total_steps": 1875, "loss": 0.6014, "lr": 9.94325872368957e-05, "epoch": 0.144, "percentage": 4.8, "elapsed_time": "0:12:54", "remaining_time": "4:15:55"}
{"current_steps": 100, "total_steps": 1875, "loss": 0.5971, "lr": 9.929980185352526e-05, "epoch": 0.16, "percentage": 5.33, "elapsed_time": "0:14:19", "remaining_time": "4:14:11"}
{"current_steps": 110, "total_steps": 1875, "loss": 0.6206, "lr": 9.915317659365077e-05, "epoch": 0.176, "percentage": 5.87, "elapsed_time": "0:15:44", "remaining_time": "4:12:29"}
{"current_steps": 120, "total_steps": 1875, "loss": 0.5597, "lr": 9.899275261921234e-05, "epoch": 0.192, "percentage": 6.4, "elapsed_time": "0:17:09", "remaining_time": "4:10:56"}
{"current_steps": 130, "total_steps": 1875, "loss": 0.5774, "lr": 9.881857496584726e-05, "epoch": 0.208, "percentage": 6.93, "elapsed_time": "0:18:35", "remaining_time": "4:09:28"}
{"current_steps": 140, "total_steps": 1875, "loss": 0.5656, "lr": 9.863069253024719e-05, "epoch": 0.224, "percentage": 7.47, "elapsed_time": "0:20:00", "remaining_time": "4:07:52"}
{"current_steps": 150, "total_steps": 1875, "loss": 0.5724, "lr": 9.842915805643155e-05, "epoch": 0.24, "percentage": 8.0, "elapsed_time": "0:21:25", "remaining_time": "4:06:21"}
{"current_steps": 160, "total_steps": 1875, "loss": 0.5796, "lr": 9.821402812094073e-05, "epoch": 0.256, "percentage": 8.53, "elapsed_time": "0:22:50", "remaining_time": "4:04:46"}
{"current_steps": 170, "total_steps": 1875, "loss": 0.5556, "lr": 9.798536311695334e-05, "epoch": 0.272, "percentage": 9.07, "elapsed_time": "0:24:15", "remaining_time": "4:03:14"}
{"current_steps": 180, "total_steps": 1875, "loss": 0.5451, "lr": 9.774322723733216e-05, "epoch": 0.288, "percentage": 9.6, "elapsed_time": "0:25:40", "remaining_time": "4:01:42"}
{"current_steps": 190, "total_steps": 1875, "loss": 0.5605, "lr": 9.748768845660334e-05, "epoch": 0.304, "percentage": 10.13, "elapsed_time": "0:27:05", "remaining_time": "4:00:14"}
{"current_steps": 200, "total_steps": 1875, "loss": 0.5644, "lr": 9.721881851187406e-05, "epoch": 0.32, "percentage": 10.67, "elapsed_time": "0:28:30", "remaining_time": "3:58:46"}
{"current_steps": 210, "total_steps": 1875, "loss": 0.5658, "lr": 9.693669288269372e-05, "epoch": 0.336, "percentage": 11.2, "elapsed_time": "0:29:55", "remaining_time": "3:57:17"}
{"current_steps": 220, "total_steps": 1875, "loss": 0.5483, "lr": 9.664139076986473e-05, "epoch": 0.352, "percentage": 11.73, "elapsed_time": "0:31:20", "remaining_time": "3:55:50"}
{"current_steps": 230, "total_steps": 1875, "loss": 0.5376, "lr": 9.63329950732086e-05, "epoch": 0.368, "percentage": 12.27, "elapsed_time": "0:32:46", "remaining_time": "3:54:21"}
{"current_steps": 240, "total_steps": 1875, "loss": 0.5402, "lr": 9.601159236829352e-05, "epoch": 0.384, "percentage": 12.8, "elapsed_time": "0:34:11", "remaining_time": "3:52:53"}
{"current_steps": 250, "total_steps": 1875, "loss": 0.5552, "lr": 9.567727288213005e-05, "epoch": 0.4, "percentage": 13.33, "elapsed_time": "0:35:36", "remaining_time": "3:51:24"}
{"current_steps": 260, "total_steps": 1875, "loss": 0.5118, "lr": 9.533013046784189e-05, "epoch": 0.416, "percentage": 13.87, "elapsed_time": "0:37:01", "remaining_time": "3:49:57"}
{"current_steps": 270, "total_steps": 1875, "loss": 0.5479, "lr": 9.497026257831855e-05, "epoch": 0.432, "percentage": 14.4, "elapsed_time": "0:38:26", "remaining_time": "3:48:30"}
{"current_steps": 280, "total_steps": 1875, "loss": 0.4773, "lr": 9.459777023885755e-05, "epoch": 0.448, "percentage": 14.93, "elapsed_time": "0:39:51", "remaining_time": "3:47:02"}
{"current_steps": 290, "total_steps": 1875, "loss": 0.5149, "lr": 9.421275801880362e-05, "epoch": 0.464, "percentage": 15.47, "elapsed_time": "0:41:16", "remaining_time": "3:45:36"}
{"current_steps": 300, "total_steps": 1875, "loss": 0.5209, "lr": 9.381533400219318e-05, "epoch": 0.48, "percentage": 16.0, "elapsed_time": "0:42:42", "remaining_time": "3:44:13"}
{"current_steps": 310, "total_steps": 1875, "loss": 0.5065, "lr": 9.340560975741197e-05, "epoch": 0.496, "percentage": 16.53, "elapsed_time": "0:44:08", "remaining_time": "3:42:49"}
{"current_steps": 320, "total_steps": 1875, "loss": 0.515, "lr": 9.298370030587456e-05, "epoch": 0.512, "percentage": 17.07, "elapsed_time": "0:45:33", "remaining_time": "3:41:25"}
{"current_steps": 330, "total_steps": 1875, "loss": 0.5375, "lr": 9.254972408973461e-05, "epoch": 0.528, "percentage": 17.6, "elapsed_time": "0:46:59", "remaining_time": "3:40:00"}
{"current_steps": 340, "total_steps": 1875, "loss": 0.5085, "lr": 9.210380293863462e-05, "epoch": 0.544, "percentage": 18.13, "elapsed_time": "0:48:25", "remaining_time": "3:38:35"}
{"current_steps": 350, "total_steps": 1875, "loss": 0.5137, "lr": 9.164606203550497e-05, "epoch": 0.56, "percentage": 18.67, "elapsed_time": "0:49:50", "remaining_time": "3:37:10"}
{"current_steps": 360, "total_steps": 1875, "loss": 0.4906, "lr": 9.117662988142138e-05, "epoch": 0.576, "percentage": 19.2, "elapsed_time": "0:51:15", "remaining_time": "3:35:44"}
{"current_steps": 370, "total_steps": 1875, "loss": 0.4966, "lr": 9.069563825953092e-05, "epoch": 0.592, "percentage": 19.73, "elapsed_time": "0:52:41", "remaining_time": "3:34:18"}
{"current_steps": 380, "total_steps": 1875, "loss": 0.5063, "lr": 9.020322219805674e-05, "epoch": 0.608, "percentage": 20.27, "elapsed_time": "0:54:07", "remaining_time": "3:32:54"}
{"current_steps": 390, "total_steps": 1875, "loss": 0.4864, "lr": 8.969951993239177e-05, "epoch": 0.624, "percentage": 20.8, "elapsed_time": "0:55:32", "remaining_time": "3:31:29"}
{"current_steps": 400, "total_steps": 1875, "loss": 0.4632, "lr": 8.9184672866292e-05, "epoch": 0.64, "percentage": 21.33, "elapsed_time": "0:56:58", "remaining_time": "3:30:04"}
{"current_steps": 410, "total_steps": 1875, "loss": 0.5097, "lr": 8.865882553218037e-05, "epoch": 0.656, "percentage": 21.87, "elapsed_time": "0:58:23", "remaining_time": "3:28:38"}
{"current_steps": 420, "total_steps": 1875, "loss": 0.4795, "lr": 8.81221255505724e-05, "epoch": 0.672, "percentage": 22.4, "elapsed_time": "0:59:49", "remaining_time": "3:27:13"}
{"current_steps": 430, "total_steps": 1875, "loss": 0.5006, "lr": 8.757472358863481e-05, "epoch": 0.688, "percentage": 22.93, "elapsed_time": "1:01:14", "remaining_time": "3:25:48"}
{"current_steps": 440, "total_steps": 1875, "loss": 0.4822, "lr": 8.701677331788891e-05, "epoch": 0.704, "percentage": 23.47, "elapsed_time": "1:02:39", "remaining_time": "3:24:22"}
{"current_steps": 450, "total_steps": 1875, "loss": 0.4856, "lr": 8.644843137107059e-05, "epoch": 0.72, "percentage": 24.0, "elapsed_time": "1:04:04", "remaining_time": "3:22:55"}
{"current_steps": 460, "total_steps": 1875, "loss": 0.4648, "lr": 8.586985729815894e-05, "epoch": 0.736, "percentage": 24.53, "elapsed_time": "1:05:30", "remaining_time": "3:21:29"}
{"current_steps": 470, "total_steps": 1875, "loss": 0.488, "lr": 8.528121352158604e-05, "epoch": 0.752, "percentage": 25.07, "elapsed_time": "1:06:55", "remaining_time": "3:20:04"}
{"current_steps": 480, "total_steps": 1875, "loss": 0.5039, "lr": 8.468266529064025e-05, "epoch": 0.768, "percentage": 25.6, "elapsed_time": "1:08:21", "remaining_time": "3:18:38"}
{"current_steps": 490, "total_steps": 1875, "loss": 0.4737, "lr": 8.4074380635076e-05, "epoch": 0.784, "percentage": 26.13, "elapsed_time": "1:09:46", "remaining_time": "3:17:13"}
{"current_steps": 500, "total_steps": 1875, "loss": 0.5296, "lr": 8.345653031794292e-05, "epoch": 0.8, "percentage": 26.67, "elapsed_time": "1:11:12", "remaining_time": "3:15:48"}
{"current_steps": 510, "total_steps": 1875, "loss": 0.4688, "lr": 8.282928778764783e-05, "epoch": 0.816, "percentage": 27.2, "elapsed_time": "1:12:47", "remaining_time": "3:14:49"}
{"current_steps": 520, "total_steps": 1875, "loss": 0.4675, "lr": 8.21928291292627e-05, "epoch": 0.832, "percentage": 27.73, "elapsed_time": "1:14:12", "remaining_time": "3:13:21"}
{"current_steps": 530, "total_steps": 1875, "loss": 0.4408, "lr": 8.154733301509248e-05, "epoch": 0.848, "percentage": 28.27, "elapsed_time": "1:15:37", "remaining_time": "3:11:54"}
{"current_steps": 540, "total_steps": 1875, "loss": 0.4608, "lr": 8.089298065451672e-05, "epoch": 0.864, "percentage": 28.8, "elapsed_time": "1:17:02", "remaining_time": "3:10:27"}
{"current_steps": 550, "total_steps": 1875, "loss": 0.4476, "lr": 8.022995574311876e-05, "epoch": 0.88, "percentage": 29.33, "elapsed_time": "1:18:26", "remaining_time": "3:08:58"}
{"current_steps": 560, "total_steps": 1875, "loss": 0.5027, "lr": 7.95584444111171e-05, "epoch": 0.896, "percentage": 29.87, "elapsed_time": "1:19:50", "remaining_time": "3:07:29"}
{"current_steps": 570, "total_steps": 1875, "loss": 0.4929, "lr": 7.887863517111338e-05, "epoch": 0.912, "percentage": 30.4, "elapsed_time": "1:21:15", "remaining_time": "3:06:01"}
{"current_steps": 580, "total_steps": 1875, "loss": 0.4739, "lr": 7.819071886517134e-05, "epoch": 0.928, "percentage": 30.93, "elapsed_time": "1:22:39", "remaining_time": "3:04:33"}
{"current_steps": 590, "total_steps": 1875, "loss": 0.4529, "lr": 7.7494888611242e-05, "epoch": 0.944, "percentage": 31.47, "elapsed_time": "1:24:03", "remaining_time": "3:03:05"}
{"current_steps": 600, "total_steps": 1875, "loss": 0.4619, "lr": 7.679133974894983e-05, "epoch": 0.96, "percentage": 32.0, "elapsed_time": "1:25:28", "remaining_time": "3:01:37"}
{"current_steps": 610, "total_steps": 1875, "loss": 0.462, "lr": 7.60802697847554e-05, "epoch": 0.976, "percentage": 32.53, "elapsed_time": "1:26:52", "remaining_time": "3:00:10"}
{"current_steps": 620, "total_steps": 1875, "loss": 0.4478, "lr": 7.536187833650947e-05, "epoch": 0.992, "percentage": 33.07, "elapsed_time": "1:28:17", "remaining_time": "2:58:43"}
{"current_steps": 630, "total_steps": 1875, "loss": 0.4304, "lr": 7.463636707741458e-05, "epoch": 1.008, "percentage": 33.6, "elapsed_time": "1:29:42", "remaining_time": "2:57:17"}
{"current_steps": 640, "total_steps": 1875, "loss": 0.4138, "lr": 7.390393967940962e-05, "epoch": 1.024, "percentage": 34.13, "elapsed_time": "1:31:07", "remaining_time": "2:55:50"}
{"current_steps": 650, "total_steps": 1875, "loss": 0.3909, "lr": 7.316480175599309e-05, "epoch": 1.04, "percentage": 34.67, "elapsed_time": "1:32:32", "remaining_time": "2:54:23"}
{"current_steps": 660, "total_steps": 1875, "loss": 0.3873, "lr": 7.241916080450163e-05, "epoch": 1.056, "percentage": 35.2, "elapsed_time": "1:33:56", "remaining_time": "2:52:56"}
{"current_steps": 670, "total_steps": 1875, "loss": 0.3744, "lr": 7.166722614785937e-05, "epoch": 1.072, "percentage": 35.73, "elapsed_time": "1:35:21", "remaining_time": "2:51:29"}
{"current_steps": 680, "total_steps": 1875, "loss": 0.3641, "lr": 7.090920887581506e-05, "epoch": 1.088, "percentage": 36.27, "elapsed_time": "1:36:45", "remaining_time": "2:50:02"}
{"current_steps": 690, "total_steps": 1875, "loss": 0.3708, "lr": 7.014532178568314e-05, "epoch": 1.104, "percentage": 36.8, "elapsed_time": "1:38:10", "remaining_time": "2:48:35"}
{"current_steps": 700, "total_steps": 1875, "loss": 0.4147, "lr": 6.937577932260515e-05, "epoch": 1.12, "percentage": 37.33, "elapsed_time": "1:39:34", "remaining_time": "2:47:08"}
{"current_steps": 710, "total_steps": 1875, "loss": 0.3698, "lr": 6.860079751934908e-05, "epoch": 1.1360000000000001, "percentage": 37.87, "elapsed_time": "1:40:59", "remaining_time": "2:45:42"}
{"current_steps": 720, "total_steps": 1875, "loss": 0.3768, "lr": 6.782059393566253e-05, "epoch": 1.152, "percentage": 38.4, "elapsed_time": "1:42:24", "remaining_time": "2:44:16"}
{"current_steps": 730, "total_steps": 1875, "loss": 0.3713, "lr": 6.70353875971976e-05, "epoch": 1.168, "percentage": 38.93, "elapsed_time": "1:43:48", "remaining_time": "2:42:49"}
{"current_steps": 740, "total_steps": 1875, "loss": 0.376, "lr": 6.624539893402382e-05, "epoch": 1.184, "percentage": 39.47, "elapsed_time": "1:45:12", "remaining_time": "2:41:22"}
{"current_steps": 750, "total_steps": 1875, "loss": 0.3762, "lr": 6.545084971874738e-05, "epoch": 1.2, "percentage": 40.0, "elapsed_time": "1:46:37", "remaining_time": "2:39:56"}
{"current_steps": 760, "total_steps": 1875, "loss": 0.3891, "lr": 6.465196300425287e-05, "epoch": 1.216, "percentage": 40.53, "elapsed_time": "1:48:01", "remaining_time": "2:38:29"}
{"current_steps": 770, "total_steps": 1875, "loss": 0.3772, "lr": 6.384896306108612e-05, "epoch": 1.232, "percentage": 41.07, "elapsed_time": "1:49:26", "remaining_time": "2:37:02"}
{"current_steps": 780, "total_steps": 1875, "loss": 0.3678, "lr": 6.304207531449486e-05, "epoch": 1.248, "percentage": 41.6, "elapsed_time": "1:50:50", "remaining_time": "2:35:36"}
{"current_steps": 790, "total_steps": 1875, "loss": 0.3648, "lr": 6.223152628114537e-05, "epoch": 1.264, "percentage": 42.13, "elapsed_time": "1:52:14", "remaining_time": "2:34:09"}
{"current_steps": 800, "total_steps": 1875, "loss": 0.3909, "lr": 6.141754350553279e-05, "epoch": 1.28, "percentage": 42.67, "elapsed_time": "1:53:39", "remaining_time": "2:32:44"}
{"current_steps": 810, "total_steps": 1875, "loss": 0.383, "lr": 6.0600355496102745e-05, "epoch": 1.296, "percentage": 43.2, "elapsed_time": "1:55:05", "remaining_time": "2:31:19"}
{"current_steps": 820, "total_steps": 1875, "loss": 0.3802, "lr": 5.9780191661102415e-05, "epoch": 1.312, "percentage": 43.73, "elapsed_time": "1:56:30", "remaining_time": "2:29:54"}
{"current_steps": 830, "total_steps": 1875, "loss": 0.423, "lr": 5.8957282244179124e-05, "epoch": 1.328, "percentage": 44.27, "elapsed_time": "1:57:56", "remaining_time": "2:28:29"}
{"current_steps": 840, "total_steps": 1875, "loss": 0.3641, "lr": 5.813185825974419e-05, "epoch": 1.3439999999999999, "percentage": 44.8, "elapsed_time": "1:59:21", "remaining_time": "2:27:03"}
{"current_steps": 850, "total_steps": 1875, "loss": 0.3519, "lr": 5.730415142812059e-05, "epoch": 1.3599999999999999, "percentage": 45.33, "elapsed_time": "2:00:46", "remaining_time": "2:25:38"}
{"current_steps": 860, "total_steps": 1875, "loss": 0.3732, "lr": 5.6474394110492344e-05, "epoch": 1.376, "percentage": 45.87, "elapsed_time": "2:02:12", "remaining_time": "2:24:13"}
{"current_steps": 870, "total_steps": 1875, "loss": 0.3954, "lr": 5.564281924367408e-05, "epoch": 1.392, "percentage": 46.4, "elapsed_time": "2:03:37", "remaining_time": "2:22:48"}
{"current_steps": 880, "total_steps": 1875, "loss": 0.3669, "lr": 5.480966027471889e-05, "epoch": 1.408, "percentage": 46.93, "elapsed_time": "2:05:02", "remaining_time": "2:21:23"}
{"current_steps": 890, "total_steps": 1875, "loss": 0.3994, "lr": 5.3975151095382995e-05, "epoch": 1.424, "percentage": 47.47, "elapsed_time": "2:06:28", "remaining_time": "2:19:57"}
{"current_steps": 900, "total_steps": 1875, "loss": 0.3904, "lr": 5.313952597646568e-05, "epoch": 1.44, "percentage": 48.0, "elapsed_time": "2:07:53", "remaining_time": "2:18:32"}
{"current_steps": 910, "total_steps": 1875, "loss": 0.3821, "lr": 5.230301950204262e-05, "epoch": 1.456, "percentage": 48.53, "elapsed_time": "2:09:18", "remaining_time": "2:17:07"}
{"current_steps": 920, "total_steps": 1875, "loss": 0.3713, "lr": 5.1465866503611426e-05, "epoch": 1.472, "percentage": 49.07, "elapsed_time": "2:10:42", "remaining_time": "2:15:41"}
{"current_steps": 930, "total_steps": 1875, "loss": 0.3785, "lr": 5.062830199416764e-05, "epoch": 1.488, "percentage": 49.6, "elapsed_time": "2:12:07", "remaining_time": "2:14:14"}
{"current_steps": 940, "total_steps": 1875, "loss": 0.3851, "lr": 4.979056110222981e-05, "epoch": 1.504, "percentage": 50.13, "elapsed_time": "2:13:31", "remaining_time": "2:12:48"}
{"current_steps": 950, "total_steps": 1875, "loss": 0.3991, "lr": 4.895287900583216e-05, "epoch": 1.52, "percentage": 50.67, "elapsed_time": "2:14:55", "remaining_time": "2:11:22"}
{"current_steps": 960, "total_steps": 1875, "loss": 0.3726, "lr": 4.811549086650327e-05, "epoch": 1.536, "percentage": 51.2, "elapsed_time": "2:16:20", "remaining_time": "2:09:57"}
{"current_steps": 970, "total_steps": 1875, "loss": 0.3703, "lr": 4.7278631763249554e-05, "epoch": 1.552, "percentage": 51.73, "elapsed_time": "2:17:45", "remaining_time": "2:08:31"}
{"current_steps": 980, "total_steps": 1875, "loss": 0.3462, "lr": 4.6442536626561675e-05, "epoch": 1.568, "percentage": 52.27, "elapsed_time": "2:19:09", "remaining_time": "2:07:05"}
{"current_steps": 990, "total_steps": 1875, "loss": 0.3486, "lr": 4.560744017246284e-05, "epoch": 1.584, "percentage": 52.8, "elapsed_time": "2:20:34", "remaining_time": "2:05:39"}
{"current_steps": 1000, "total_steps": 1875, "loss": 0.3384, "lr": 4.477357683661734e-05, "epoch": 1.6, "percentage": 53.33, "elapsed_time": "2:21:58", "remaining_time": "2:04:13"}
{"current_steps": 1010, "total_steps": 1875, "loss": 0.3976, "lr": 4.394118070851749e-05, "epoch": 1.616, "percentage": 53.87, "elapsed_time": "2:23:32", "remaining_time": "2:02:55"}
{"current_steps": 1020, "total_steps": 1875, "loss": 0.3607, "lr": 4.31104854657681e-05, "epoch": 1.6320000000000001, "percentage": 54.4, "elapsed_time": "2:24:55", "remaining_time": "2:01:29"}
{"current_steps": 1030, "total_steps": 1875, "loss": 0.371, "lr": 4.228172430848644e-05, "epoch": 1.6480000000000001, "percentage": 54.93, "elapsed_time": "2:26:19", "remaining_time": "2:00:02"}
{"current_steps": 1040, "total_steps": 1875, "loss": 0.3627, "lr": 4.1455129893836174e-05, "epoch": 1.6640000000000001, "percentage": 55.47, "elapsed_time": "2:27:42", "remaining_time": "1:58:35"}
{"current_steps": 1050, "total_steps": 1875, "loss": 0.3725, "lr": 4.063093427071376e-05, "epoch": 1.6800000000000002, "percentage": 56.0, "elapsed_time": "2:29:06", "remaining_time": "1:57:09"}
{"current_steps": 1060, "total_steps": 1875, "loss": 0.3466, "lr": 3.9809368814605766e-05, "epoch": 1.696, "percentage": 56.53, "elapsed_time": "2:30:30", "remaining_time": "1:55:42"}
{"current_steps": 1070, "total_steps": 1875, "loss": 0.3629, "lr": 3.899066416263493e-05, "epoch": 1.712, "percentage": 57.07, "elapsed_time": "2:31:53", "remaining_time": "1:54:16"}
{"current_steps": 1080, "total_steps": 1875, "loss": 0.3941, "lr": 3.817505014881378e-05, "epoch": 1.728, "percentage": 57.6, "elapsed_time": "2:33:17", "remaining_time": "1:52:50"}
{"current_steps": 1090, "total_steps": 1875, "loss": 0.3362, "lr": 3.736275573952354e-05, "epoch": 1.744, "percentage": 58.13, "elapsed_time": "2:34:41", "remaining_time": "1:51:24"}
{"current_steps": 1100, "total_steps": 1875, "loss": 0.3197, "lr": 3.655400896923672e-05, "epoch": 1.76, "percentage": 58.67, "elapsed_time": "2:36:05", "remaining_time": "1:49:58"}
{"current_steps": 1110, "total_steps": 1875, "loss": 0.3447, "lr": 3.5749036876501194e-05, "epoch": 1.776, "percentage": 59.2, "elapsed_time": "2:37:29", "remaining_time": "1:48:32"}
{"current_steps": 1120, "total_steps": 1875, "loss": 0.3535, "lr": 3.494806544020398e-05, "epoch": 1.792, "percentage": 59.73, "elapsed_time": "2:38:53", "remaining_time": "1:47:06"}
{"current_steps": 1130, "total_steps": 1875, "loss": 0.3447, "lr": 3.4151319516132416e-05, "epoch": 1.808, "percentage": 60.27, "elapsed_time": "2:40:16", "remaining_time": "1:45:40"}
{"current_steps": 1140, "total_steps": 1875, "loss": 0.338, "lr": 3.335902277385067e-05, "epoch": 1.8239999999999998, "percentage": 60.8, "elapsed_time": "2:41:40", "remaining_time": "1:44:14"}
{"current_steps": 1150, "total_steps": 1875, "loss": 0.3915, "lr": 3.257139763390925e-05, "epoch": 1.8399999999999999, "percentage": 61.33, "elapsed_time": "2:43:04", "remaining_time": "1:42:48"}
{"current_steps": 1160, "total_steps": 1875, "loss": 0.332, "lr": 3.178866520540509e-05, "epoch": 1.8559999999999999, "percentage": 61.87, "elapsed_time": "2:44:28", "remaining_time": "1:41:22"}
{"current_steps": 1170, "total_steps": 1875, "loss": 0.3734, "lr": 3.101104522390995e-05, "epoch": 1.8719999999999999, "percentage": 62.4, "elapsed_time": "2:45:52", "remaining_time": "1:39:56"}
{"current_steps": 1180, "total_steps": 1875, "loss": 0.3409, "lr": 3.023875598978419e-05, "epoch": 1.888, "percentage": 62.93, "elapsed_time": "2:47:16", "remaining_time": "1:38:31"}
{"current_steps": 1190, "total_steps": 1875, "loss": 0.3333, "lr": 2.9472014306893603e-05, "epoch": 1.904, "percentage": 63.47, "elapsed_time": "2:48:40", "remaining_time": "1:37:05"}
{"current_steps": 1200, "total_steps": 1875, "loss": 0.369, "lr": 2.8711035421746367e-05, "epoch": 1.92, "percentage": 64.0, "elapsed_time": "2:50:04", "remaining_time": "1:35:39"}
{"current_steps": 1210, "total_steps": 1875, "loss": 0.3449, "lr": 2.795603296306708e-05, "epoch": 1.936, "percentage": 64.53, "elapsed_time": "2:51:28", "remaining_time": "1:34:14"}
{"current_steps": 1220, "total_steps": 1875, "loss": 0.3662, "lr": 2.7207218881825014e-05, "epoch": 1.952, "percentage": 65.07, "elapsed_time": "2:52:52", "remaining_time": "1:32:48"}
{"current_steps": 1230, "total_steps": 1875, "loss": 0.3773, "lr": 2.6464803391733374e-05, "epoch": 1.968, "percentage": 65.6, "elapsed_time": "2:54:16", "remaining_time": "1:31:23"}
{"current_steps": 1240, "total_steps": 1875, "loss": 0.3245, "lr": 2.5728994910236304e-05, "epoch": 1.984, "percentage": 66.13, "elapsed_time": "2:55:40", "remaining_time": "1:29:57"}
{"current_steps": 1250, "total_steps": 1875, "loss": 0.3412, "lr": 2.500000000000001e-05, "epoch": 2.0, "percentage": 66.67, "elapsed_time": "2:57:04", "remaining_time": "1:28:32"}
{"current_steps": 1260, "total_steps": 1875, "loss": 0.3027, "lr": 2.4278023310924673e-05, "epoch": 2.016, "percentage": 67.2, "elapsed_time": "2:58:28", "remaining_time": "1:27:06"}
{"current_steps": 1270, "total_steps": 1875, "loss": 0.2745, "lr": 2.3563267522693415e-05, "epoch": 2.032, "percentage": 67.73, "elapsed_time": "2:59:52", "remaining_time": "1:25:41"}
{"current_steps": 1280, "total_steps": 1875, "loss": 0.2686, "lr": 2.2855933287874138e-05, "epoch": 2.048, "percentage": 68.27, "elapsed_time": "3:01:16", "remaining_time": "1:24:15"}
{"current_steps": 1290, "total_steps": 1875, "loss": 0.2628, "lr": 2.215621917559062e-05, "epoch": 2.064, "percentage": 68.8, "elapsed_time": "3:02:40", "remaining_time": "1:22:50"}
{"current_steps": 1300, "total_steps": 1875, "loss": 0.274, "lr": 2.1464321615778422e-05, "epoch": 2.08, "percentage": 69.33, "elapsed_time": "3:04:04", "remaining_time": "1:21:25"}
{"current_steps": 1310, "total_steps": 1875, "loss": 0.2711, "lr": 2.07804348440414e-05, "epoch": 2.096, "percentage": 69.87, "elapsed_time": "3:05:28", "remaining_time": "1:19:59"}
{"current_steps": 1320, "total_steps": 1875, "loss": 0.2662, "lr": 2.0104750847124075e-05, "epoch": 2.112, "percentage": 70.4, "elapsed_time": "3:06:52", "remaining_time": "1:18:34"}
{"current_steps": 1330, "total_steps": 1875, "loss": 0.2722, "lr": 1.9437459309015427e-05, "epoch": 2.128, "percentage": 70.93, "elapsed_time": "3:08:16", "remaining_time": "1:17:08"}
{"current_steps": 1340, "total_steps": 1875, "loss": 0.2965, "lr": 1.8778747557699224e-05, "epoch": 2.144, "percentage": 71.47, "elapsed_time": "3:09:40", "remaining_time": "1:15:43"}
{"current_steps": 1350, "total_steps": 1875, "loss": 0.2789, "lr": 1.8128800512565513e-05, "epoch": 2.16, "percentage": 72.0, "elapsed_time": "3:11:04", "remaining_time": "1:14:18"}
{"current_steps": 1360, "total_steps": 1875, "loss": 0.2727, "lr": 1.7487800632498545e-05, "epoch": 2.176, "percentage": 72.53, "elapsed_time": "3:12:27", "remaining_time": "1:12:52"}
{"current_steps": 1370, "total_steps": 1875, "loss": 0.2579, "lr": 1.685592786465524e-05, "epoch": 2.192, "percentage": 73.07, "elapsed_time": "3:13:51", "remaining_time": "1:11:27"}
{"current_steps": 1380, "total_steps": 1875, "loss": 0.2639, "lr": 1.6233359593948777e-05, "epoch": 2.208, "percentage": 73.6, "elapsed_time": "3:15:15", "remaining_time": "1:10:02"}
{"current_steps": 1390, "total_steps": 1875, "loss": 0.2909, "lr": 1.5620270593251635e-05, "epoch": 2.224, "percentage": 74.13, "elapsed_time": "3:16:39", "remaining_time": "1:08:37"}
{"current_steps": 1400, "total_steps": 1875, "loss": 0.2675, "lr": 1.5016832974331724e-05, "epoch": 2.24, "percentage": 74.67, "elapsed_time": "3:18:03", "remaining_time": "1:07:11"}
{"current_steps": 1410, "total_steps": 1875, "loss": 0.2501, "lr": 1.4423216139535734e-05, "epoch": 2.2560000000000002, "percentage": 75.2, "elapsed_time": "3:19:26", "remaining_time": "1:05:46"}
{"current_steps": 1420, "total_steps": 1875, "loss": 0.28, "lr": 1.3839586734232906e-05, "epoch": 2.2720000000000002, "percentage": 75.73, "elapsed_time": "3:20:50", "remaining_time": "1:04:21"}
{"current_steps": 1430, "total_steps": 1875, "loss": 0.2946, "lr": 1.3266108600032929e-05, "epoch": 2.288, "percentage": 76.27, "elapsed_time": "3:22:14", "remaining_time": "1:02:56"}
{"current_steps": 1440, "total_steps": 1875, "loss": 0.2801, "lr": 1.2702942728790895e-05, "epoch": 2.304, "percentage": 76.8, "elapsed_time": "3:23:38", "remaining_time": "1:01:30"}
{"current_steps": 1450, "total_steps": 1875, "loss": 0.2686, "lr": 1.2150247217412186e-05, "epoch": 2.32, "percentage": 77.33, "elapsed_time": "3:25:02", "remaining_time": "1:00:05"}
{"current_steps": 1460, "total_steps": 1875, "loss": 0.2623, "lr": 1.160817722347014e-05, "epoch": 2.336, "percentage": 77.87, "elapsed_time": "3:26:26", "remaining_time": "0:58:40"}
{"current_steps": 1470, "total_steps": 1875, "loss": 0.2865, "lr": 1.1076884921648834e-05, "epoch": 2.352, "percentage": 78.4, "elapsed_time": "3:27:50", "remaining_time": "0:57:15"}
{"current_steps": 1480, "total_steps": 1875, "loss": 0.2677, "lr": 1.0556519461023301e-05, "epoch": 2.368, "percentage": 78.93, "elapsed_time": "3:29:14", "remaining_time": "0:55:50"}
{"current_steps": 1490, "total_steps": 1875, "loss": 0.2821, "lr": 1.0047226923189024e-05, "epoch": 2.384, "percentage": 79.47, "elapsed_time": "3:30:38", "remaining_time": "0:54:25"}
{"current_steps": 1500, "total_steps": 1875, "loss": 0.2628, "lr": 9.549150281252633e-06, "epoch": 2.4, "percentage": 80.0, "elapsed_time": "3:32:02", "remaining_time": "0:53:00"}
{"current_steps": 1510, "total_steps": 1875, "loss": 0.2648, "lr": 9.06242935969528e-06, "epoch": 2.416, "percentage": 80.53, "elapsed_time": "3:33:36", "remaining_time": "0:51:38"}
{"current_steps": 1520, "total_steps": 1875, "loss": 0.2718, "lr": 8.587200795119793e-06, "epoch": 2.432, "percentage": 81.07, "elapsed_time": "3:35:00", "remaining_time": "0:50:13"}
{"current_steps": 1530, "total_steps": 1875, "loss": 0.2779, "lr": 8.123597997892918e-06, "epoch": 2.448, "percentage": 81.6, "elapsed_time": "3:36:25", "remaining_time": "0:48:48"}
{"current_steps": 1540, "total_steps": 1875, "loss": 0.2522, "lr": 7.671751114693104e-06, "epoch": 2.464, "percentage": 82.13, "elapsed_time": "3:37:50", "remaining_time": "0:47:23"}
{"current_steps": 1550, "total_steps": 1875, "loss": 0.2668, "lr": 7.2317869919746705e-06, "epoch": 2.48, "percentage": 82.67, "elapsed_time": "3:39:14", "remaining_time": "0:45:58"}
{"current_steps": 1560, "total_steps": 1875, "loss": 0.2582, "lr": 6.803829140358237e-06, "epoch": 2.496, "percentage": 83.2, "elapsed_time": "3:40:38", "remaining_time": "0:44:33"}
{"current_steps": 1570, "total_steps": 1875, "loss": 0.2509, "lr": 6.3879976999578154e-06, "epoch": 2.512, "percentage": 83.73, "elapsed_time": "3:42:02", "remaining_time": "0:43:08"}
{"current_steps": 1580, "total_steps": 1875, "loss": 0.2743, "lr": 5.98440940665399e-06, "epoch": 2.528, "percentage": 84.27, "elapsed_time": "3:43:26", "remaining_time": "0:41:43"}
{"current_steps": 1590, "total_steps": 1875, "loss": 0.2833, "lr": 5.593177559322777e-06, "epoch": 2.544, "percentage": 84.8, "elapsed_time": "3:44:49", "remaining_time": "0:40:17"}
{"current_steps": 1600, "total_steps": 1875, "loss": 0.2693, "lr": 5.214411988029355e-06, "epoch": 2.56, "percentage": 85.33, "elapsed_time": "3:46:12", "remaining_time": "0:38:52"}
{"current_steps": 1610, "total_steps": 1875, "loss": 0.2607, "lr": 4.848219023195644e-06, "epoch": 2.576, "percentage": 85.87, "elapsed_time": "3:47:35", "remaining_time": "0:37:27"}
{"current_steps": 1620, "total_steps": 1875, "loss": 0.2632, "lr": 4.494701465750217e-06, "epoch": 2.592, "percentage": 86.4, "elapsed_time": "3:48:59", "remaining_time": "0:36:02"}
{"current_steps": 1630, "total_steps": 1875, "loss": 0.2573, "lr": 4.153958558269189e-06, "epoch": 2.608, "percentage": 86.93, "elapsed_time": "3:50:22", "remaining_time": "0:34:37"}
{"current_steps": 1640, "total_steps": 1875, "loss": 0.2411, "lr": 3.826085957115888e-06, "epoch": 2.624, "percentage": 87.47, "elapsed_time": "3:51:45", "remaining_time": "0:33:12"}
{"current_steps": 1650, "total_steps": 1875, "loss": 0.2601, "lr": 3.511175705587433e-06, "epoch": 2.64, "percentage": 88.0, "elapsed_time": "3:53:08", "remaining_time": "0:31:47"}
{"current_steps": 1660, "total_steps": 1875, "loss": 0.2832, "lr": 3.2093162080754637e-06, "epoch": 2.656, "percentage": 88.53, "elapsed_time": "3:54:32", "remaining_time": "0:30:22"}
{"current_steps": 1670, "total_steps": 1875, "loss": 0.2725, "lr": 2.9205922052484958e-06, "epoch": 2.672, "percentage": 89.07, "elapsed_time": "3:55:55", "remaining_time": "0:28:57"}
{"current_steps": 1680, "total_steps": 1875, "loss": 0.2654, "lr": 2.6450847502627884e-06, "epoch": 2.6879999999999997, "percentage": 89.6, "elapsed_time": "3:57:18", "remaining_time": "0:27:32"}
{"current_steps": 1690, "total_steps": 1875, "loss": 0.2784, "lr": 2.3828711860083674e-06, "epoch": 2.7039999999999997, "percentage": 90.13, "elapsed_time": "3:58:41", "remaining_time": "0:26:07"}
{"current_steps": 1700, "total_steps": 1875, "loss": 0.2758, "lr": 2.134025123396638e-06, "epoch": 2.7199999999999998, "percentage": 90.67, "elapsed_time": "4:00:05", "remaining_time": "0:24:42"}
{"current_steps": 1710, "total_steps": 1875, "loss": 0.2533, "lr": 1.8986164206957035e-06, "epoch": 2.7359999999999998, "percentage": 91.2, "elapsed_time": "4:01:28", "remaining_time": "0:23:17"}
{"current_steps": 1720, "total_steps": 1875, "loss": 0.268, "lr": 1.6767111639191202e-06, "epoch": 2.752, "percentage": 91.73, "elapsed_time": "4:02:51", "remaining_time": "0:21:53"}
{"current_steps": 1730, "total_steps": 1875, "loss": 0.28, "lr": 1.4683716482736366e-06, "epoch": 2.768, "percentage": 92.27, "elapsed_time": "4:04:14", "remaining_time": "0:20:28"}
{"current_steps": 1740, "total_steps": 1875, "loss": 0.2797, "lr": 1.2736563606711382e-06, "epoch": 2.784, "percentage": 92.8, "elapsed_time": "4:05:37", "remaining_time": "0:19:03"}
{"current_steps": 1750, "total_steps": 1875, "loss": 0.2831, "lr": 1.0926199633097157e-06, "epoch": 2.8, "percentage": 93.33, "elapsed_time": "4:07:00", "remaining_time": "0:17:38"}
{"current_steps": 1760, "total_steps": 1875, "loss": 0.2558, "lr": 9.253132783283547e-07, "epoch": 2.816, "percentage": 93.87, "elapsed_time": "4:08:23", "remaining_time": "0:16:13"}
{"current_steps": 1770, "total_steps": 1875, "loss": 0.2635, "lr": 7.717832735397335e-07, "epoch": 2.832, "percentage": 94.4, "elapsed_time": "4:09:48", "remaining_time": "0:14:49"}
{"current_steps": 1780, "total_steps": 1875, "loss": 0.2639, "lr": 6.3207304924498e-07, "epoch": 2.848, "percentage": 94.93, "elapsed_time": "4:11:12", "remaining_time": "0:13:24"}
{"current_steps": 1790, "total_steps": 1875, "loss": 0.2532, "lr": 5.062218261342122e-07, "epoch": 2.864, "percentage": 95.47, "elapsed_time": "4:12:36", "remaining_time": "0:11:59"}
{"current_steps": 1800, "total_steps": 1875, "loss": 0.2327, "lr": 3.9426493427611177e-07, "epoch": 2.88, "percentage": 96.0, "elapsed_time": "4:14:00", "remaining_time": "0:10:35"}
{"current_steps": 1810, "total_steps": 1875, "loss": 0.2621, "lr": 2.962338031997691e-07, "epoch": 2.896, "percentage": 96.53, "elapsed_time": "4:15:23", "remaining_time": "0:09:10"}
{"current_steps": 1820, "total_steps": 1875, "loss": 0.256, "lr": 2.1215595307154667e-07, "epoch": 2.912, "percentage": 97.07, "elapsed_time": "4:16:47", "remaining_time": "0:07:45"}
{"current_steps": 1830, "total_steps": 1875, "loss": 0.2545, "lr": 1.420549869693033e-07, "epoch": 2.928, "percentage": 97.6, "elapsed_time": "4:18:11", "remaining_time": "0:06:20"}
{"current_steps": 1840, "total_steps": 1875, "loss": 0.2596, "lr": 8.595058425640013e-08, "epoch": 2.944, "percentage": 98.13, "elapsed_time": "4:19:34", "remaining_time": "0:04:56"}
{"current_steps": 1850, "total_steps": 1875, "loss": 0.2584, "lr": 4.385849505708084e-08, "epoch": 2.96, "percentage": 98.67, "elapsed_time": "4:20:58", "remaining_time": "0:03:31"}
{"current_steps": 1860, "total_steps": 1875, "loss": 0.2752, "lr": 1.5790535835003008e-08, "epoch": 2.976, "percentage": 99.2, "elapsed_time": "4:22:21", "remaining_time": "0:02:06"}
{"current_steps": 1870, "total_steps": 1875, "loss": 0.2778, "lr": 1.7545860759693445e-09, "epoch": 2.992, "percentage": 99.73, "elapsed_time": "4:23:45", "remaining_time": "0:00:42"}
{"current_steps": 1875, "total_steps": 1875, "epoch": 3.0, "percentage": 100.0, "elapsed_time": "4:24:37", "remaining_time": "0:00:00"}