trainer_log.jsonl

{"current_steps": 10, "total_steps": 1938, "loss": 1.0792, "lr": 5e-06, "epoch": 0.015458937198067632, "percentage": 0.52, "elapsed_time": "0:09:43", "remaining_time": "1 day, 7:13:51"}
{"current_steps": 20, "total_steps": 1938, "loss": 0.9837, "lr": 5e-06, "epoch": 0.030917874396135265, "percentage": 1.03, "elapsed_time": "0:19:21", "remaining_time": "1 day, 6:57:00"}
{"current_steps": 30, "total_steps": 1938, "loss": 0.9474, "lr": 5e-06, "epoch": 0.0463768115942029, "percentage": 1.55, "elapsed_time": "0:29:00", "remaining_time": "1 day, 6:44:29"}
{"current_steps": 40, "total_steps": 1938, "loss": 0.9301, "lr": 5e-06, "epoch": 0.06183574879227053, "percentage": 2.06, "elapsed_time": "0:38:38", "remaining_time": "1 day, 6:33:22"}
{"current_steps": 50, "total_steps": 1938, "loss": 0.9155, "lr": 5e-06, "epoch": 0.07729468599033816, "percentage": 2.58, "elapsed_time": "0:48:16", "remaining_time": "1 day, 6:22:58"}
{"current_steps": 60, "total_steps": 1938, "loss": 0.9037, "lr": 5e-06, "epoch": 0.0927536231884058, "percentage": 3.1, "elapsed_time": "0:57:55", "remaining_time": "1 day, 6:13:07"}
{"current_steps": 70, "total_steps": 1938, "loss": 0.8967, "lr": 5e-06, "epoch": 0.10821256038647344, "percentage": 3.61, "elapsed_time": "1:07:34", "remaining_time": "1 day, 6:03:12"}
{"current_steps": 80, "total_steps": 1938, "loss": 0.8867, "lr": 5e-06, "epoch": 0.12367149758454106, "percentage": 4.13, "elapsed_time": "1:17:12", "remaining_time": "1 day, 5:53:11"}
{"current_steps": 90, "total_steps": 1938, "loss": 0.8792, "lr": 5e-06, "epoch": 0.1391304347826087, "percentage": 4.64, "elapsed_time": "1:26:51", "remaining_time": "1 day, 5:43:23"}
{"current_steps": 100, "total_steps": 1938, "loss": 0.8769, "lr": 5e-06, "epoch": 0.15458937198067632, "percentage": 5.16, "elapsed_time": "1:36:29", "remaining_time": "1 day, 5:33:27"}
{"current_steps": 110, "total_steps": 1938, "loss": 0.8717, "lr": 5e-06, "epoch": 0.17004830917874397, "percentage": 5.68, "elapsed_time": "1:46:07", "remaining_time": "1 day, 5:23:41"}
{"current_steps": 120, "total_steps": 1938, "loss": 0.8671, "lr": 5e-06, "epoch": 0.1855072463768116, "percentage": 6.19, "elapsed_time": "1:55:46", "remaining_time": "1 day, 5:13:53"}
{"current_steps": 130, "total_steps": 1938, "loss": 0.8642, "lr": 5e-06, "epoch": 0.20096618357487922, "percentage": 6.71, "elapsed_time": "2:05:24", "remaining_time": "1 day, 5:04:02"}
{"current_steps": 140, "total_steps": 1938, "loss": 0.863, "lr": 5e-06, "epoch": 0.21642512077294687, "percentage": 7.22, "elapsed_time": "2:15:01", "remaining_time": "1 day, 4:54:10"}
{"current_steps": 150, "total_steps": 1938, "loss": 0.861, "lr": 5e-06, "epoch": 0.2318840579710145, "percentage": 7.74, "elapsed_time": "2:24:39", "remaining_time": "1 day, 4:44:17"}
{"current_steps": 160, "total_steps": 1938, "loss": 0.864, "lr": 5e-06, "epoch": 0.24734299516908212, "percentage": 8.26, "elapsed_time": "2:34:16", "remaining_time": "1 day, 4:34:22"}
{"current_steps": 170, "total_steps": 1938, "loss": 0.8528, "lr": 5e-06, "epoch": 0.26280193236714977, "percentage": 8.77, "elapsed_time": "2:43:54", "remaining_time": "1 day, 4:24:33"}
{"current_steps": 180, "total_steps": 1938, "loss": 0.857, "lr": 5e-06, "epoch": 0.2782608695652174, "percentage": 9.29, "elapsed_time": "2:53:31", "remaining_time": "1 day, 4:14:46"}
{"current_steps": 190, "total_steps": 1938, "loss": 0.8511, "lr": 5e-06, "epoch": 0.293719806763285, "percentage": 9.8, "elapsed_time": "3:03:08", "remaining_time": "1 day, 4:04:52"}
{"current_steps": 200, "total_steps": 1938, "loss": 0.8485, "lr": 5e-06, "epoch": 0.30917874396135264, "percentage": 10.32, "elapsed_time": "3:12:46", "remaining_time": "1 day, 3:55:11"}
{"current_steps": 210, "total_steps": 1938, "loss": 0.8462, "lr": 5e-06, "epoch": 0.32463768115942027, "percentage": 10.84, "elapsed_time": "3:22:24", "remaining_time": "1 day, 3:45:32"}
{"current_steps": 220, "total_steps": 1938, "loss": 0.8446, "lr": 5e-06, "epoch": 0.34009661835748795, "percentage": 11.35, "elapsed_time": "3:32:02", "remaining_time": "1 day, 3:35:50"}
{"current_steps": 230, "total_steps": 1938, "loss": 0.8445, "lr": 5e-06, "epoch": 0.35555555555555557, "percentage": 11.87, "elapsed_time": "3:41:40", "remaining_time": "1 day, 3:26:13"}
{"current_steps": 240, "total_steps": 1938, "loss": 0.8497, "lr": 5e-06, "epoch": 0.3710144927536232, "percentage": 12.38, "elapsed_time": "3:51:19", "remaining_time": "1 day, 3:16:35"}
{"current_steps": 250, "total_steps": 1938, "loss": 0.8438, "lr": 5e-06, "epoch": 0.3864734299516908, "percentage": 12.9, "elapsed_time": "4:00:57", "remaining_time": "1 day, 3:06:55"}
{"current_steps": 260, "total_steps": 1938, "loss": 0.8476, "lr": 5e-06, "epoch": 0.40193236714975844, "percentage": 13.42, "elapsed_time": "4:10:34", "remaining_time": "1 day, 2:57:13"}
{"current_steps": 270, "total_steps": 1938, "loss": 0.8442, "lr": 5e-06, "epoch": 0.41739130434782606, "percentage": 13.93, "elapsed_time": "4:20:12", "remaining_time": "1 day, 2:47:33"}
{"current_steps": 280, "total_steps": 1938, "loss": 0.8384, "lr": 5e-06, "epoch": 0.43285024154589374, "percentage": 14.45, "elapsed_time": "4:29:50", "remaining_time": "1 day, 2:37:53"}
{"current_steps": 290, "total_steps": 1938, "loss": 0.8416, "lr": 5e-06, "epoch": 0.44830917874396137, "percentage": 14.96, "elapsed_time": "4:39:29", "remaining_time": "1 day, 2:28:15"}
{"current_steps": 300, "total_steps": 1938, "loss": 0.8363, "lr": 5e-06, "epoch": 0.463768115942029, "percentage": 15.48, "elapsed_time": "4:49:07", "remaining_time": "1 day, 2:18:38"}
{"current_steps": 310, "total_steps": 1938, "loss": 0.8391, "lr": 5e-06, "epoch": 0.4792270531400966, "percentage": 16.0, "elapsed_time": "4:58:45", "remaining_time": "1 day, 2:08:55"}
{"current_steps": 320, "total_steps": 1938, "loss": 0.8384, "lr": 5e-06, "epoch": 0.49468599033816424, "percentage": 16.51, "elapsed_time": "5:08:23", "remaining_time": "1 day, 1:59:19"}
{"current_steps": 330, "total_steps": 1938, "loss": 0.8344, "lr": 5e-06, "epoch": 0.5101449275362319, "percentage": 17.03, "elapsed_time": "5:18:02", "remaining_time": "1 day, 1:49:41"}
{"current_steps": 340, "total_steps": 1938, "loss": 0.8352, "lr": 5e-06, "epoch": 0.5256038647342995, "percentage": 17.54, "elapsed_time": "5:27:40", "remaining_time": "1 day, 1:40:02"}
{"current_steps": 350, "total_steps": 1938, "loss": 0.8319, "lr": 5e-06, "epoch": 0.5410628019323671, "percentage": 18.06, "elapsed_time": "5:37:17", "remaining_time": "1 day, 1:30:20"}
{"current_steps": 360, "total_steps": 1938, "loss": 0.8305, "lr": 5e-06, "epoch": 0.5565217391304348, "percentage": 18.58, "elapsed_time": "5:46:54", "remaining_time": "1 day, 1:20:38"}
{"current_steps": 370, "total_steps": 1938, "loss": 0.8283, "lr": 5e-06, "epoch": 0.5719806763285025, "percentage": 19.09, "elapsed_time": "5:56:31", "remaining_time": "1 day, 1:10:54"}
{"current_steps": 380, "total_steps": 1938, "loss": 0.8341, "lr": 5e-06, "epoch": 0.58743961352657, "percentage": 19.61, "elapsed_time": "6:06:09", "remaining_time": "1 day, 1:01:13"}
{"current_steps": 390, "total_steps": 1938, "loss": 0.8291, "lr": 5e-06, "epoch": 0.6028985507246377, "percentage": 20.12, "elapsed_time": "6:15:47", "remaining_time": "1 day, 0:51:36"}
{"current_steps": 400, "total_steps": 1938, "loss": 0.8361, "lr": 5e-06, "epoch": 0.6183574879227053, "percentage": 20.64, "elapsed_time": "6:25:26", "remaining_time": "1 day, 0:42:01"}
{"current_steps": 410, "total_steps": 1938, "loss": 0.8291, "lr": 5e-06, "epoch": 0.633816425120773, "percentage": 21.16, "elapsed_time": "6:35:04", "remaining_time": "1 day, 0:32:22"}
{"current_steps": 420, "total_steps": 1938, "loss": 0.83, "lr": 5e-06, "epoch": 0.6492753623188405, "percentage": 21.67, "elapsed_time": "6:44:42", "remaining_time": "1 day, 0:22:44"}
{"current_steps": 430, "total_steps": 1938, "loss": 0.8237, "lr": 5e-06, "epoch": 0.6647342995169082, "percentage": 22.19, "elapsed_time": "6:54:20", "remaining_time": "1 day, 0:13:04"}
{"current_steps": 440, "total_steps": 1938, "loss": 0.8228, "lr": 5e-06, "epoch": 0.6801932367149759, "percentage": 22.7, "elapsed_time": "7:03:58", "remaining_time": "1 day, 0:03:27"}
{"current_steps": 450, "total_steps": 1938, "loss": 0.829, "lr": 5e-06, "epoch": 0.6956521739130435, "percentage": 23.22, "elapsed_time": "7:13:37", "remaining_time": "23:53:49"}
{"current_steps": 460, "total_steps": 1938, "loss": 0.8221, "lr": 5e-06, "epoch": 0.7111111111111111, "percentage": 23.74, "elapsed_time": "7:23:15", "remaining_time": "23:44:12"}
{"current_steps": 470, "total_steps": 1938, "loss": 0.8254, "lr": 5e-06, "epoch": 0.7265700483091787, "percentage": 24.25, "elapsed_time": "7:32:54", "remaining_time": "23:34:36"}
{"current_steps": 480, "total_steps": 1938, "loss": 0.8264, "lr": 5e-06, "epoch": 0.7420289855072464, "percentage": 24.77, "elapsed_time": "7:42:33", "remaining_time": "23:25:00"}
{"current_steps": 490, "total_steps": 1938, "loss": 0.8295, "lr": 5e-06, "epoch": 0.7574879227053141, "percentage": 25.28, "elapsed_time": "7:52:10", "remaining_time": "23:15:19"}
{"current_steps": 500, "total_steps": 1938, "loss": 0.8223, "lr": 5e-06, "epoch": 0.7729468599033816, "percentage": 25.8, "elapsed_time": "8:01:49", "remaining_time": "23:05:43"}
{"current_steps": 510, "total_steps": 1938, "loss": 0.8257, "lr": 5e-06, "epoch": 0.7884057971014493, "percentage": 26.32, "elapsed_time": "8:11:27", "remaining_time": "22:56:05"}
{"current_steps": 520, "total_steps": 1938, "loss": 0.8265, "lr": 5e-06, "epoch": 0.8038647342995169, "percentage": 26.83, "elapsed_time": "8:21:05", "remaining_time": "22:46:27"}
{"current_steps": 530, "total_steps": 1938, "loss": 0.8224, "lr": 5e-06, "epoch": 0.8193236714975846, "percentage": 27.35, "elapsed_time": "8:30:44", "remaining_time": "22:36:49"}
{"current_steps": 540, "total_steps": 1938, "loss": 0.821, "lr": 5e-06, "epoch": 0.8347826086956521, "percentage": 27.86, "elapsed_time": "8:40:22", "remaining_time": "22:27:12"}
{"current_steps": 550, "total_steps": 1938, "loss": 0.8271, "lr": 5e-06, "epoch": 0.8502415458937198, "percentage": 28.38, "elapsed_time": "8:50:01", "remaining_time": "22:17:35"}
{"current_steps": 560, "total_steps": 1938, "loss": 0.8168, "lr": 5e-06, "epoch": 0.8657004830917875, "percentage": 28.9, "elapsed_time": "8:59:40", "remaining_time": "22:07:58"}
{"current_steps": 570, "total_steps": 1938, "loss": 0.8234, "lr": 5e-06, "epoch": 0.881159420289855, "percentage": 29.41, "elapsed_time": "9:09:18", "remaining_time": "21:58:20"}
{"current_steps": 580, "total_steps": 1938, "loss": 0.8198, "lr": 5e-06, "epoch": 0.8966183574879227, "percentage": 29.93, "elapsed_time": "9:18:57", "remaining_time": "21:48:44"}
{"current_steps": 590, "total_steps": 1938, "loss": 0.8169, "lr": 5e-06, "epoch": 0.9120772946859903, "percentage": 30.44, "elapsed_time": "9:28:35", "remaining_time": "21:39:06"}
{"current_steps": 600, "total_steps": 1938, "loss": 0.8225, "lr": 5e-06, "epoch": 0.927536231884058, "percentage": 30.96, "elapsed_time": "9:38:13", "remaining_time": "21:29:26"}
{"current_steps": 610, "total_steps": 1938, "loss": 0.8188, "lr": 5e-06, "epoch": 0.9429951690821256, "percentage": 31.48, "elapsed_time": "9:47:51", "remaining_time": "21:19:48"}
{"current_steps": 620, "total_steps": 1938, "loss": 0.8217, "lr": 5e-06, "epoch": 0.9584541062801932, "percentage": 31.99, "elapsed_time": "9:57:29", "remaining_time": "21:10:10"}
{"current_steps": 630, "total_steps": 1938, "loss": 0.8221, "lr": 5e-06, "epoch": 0.9739130434782609, "percentage": 32.51, "elapsed_time": "10:07:08", "remaining_time": "21:00:32"}
{"current_steps": 640, "total_steps": 1938, "loss": 0.8188, "lr": 5e-06, "epoch": 0.9893719806763285, "percentage": 33.02, "elapsed_time": "10:16:47", "remaining_time": "20:50:55"}
{"current_steps": 646, "total_steps": 1938, "eval_loss": 0.8183467984199524, "epoch": 0.9986473429951691, "percentage": 33.33, "elapsed_time": "10:34:44", "remaining_time": "21:09:28"}
{"current_steps": 650, "total_steps": 1938, "loss": 0.8235, "lr": 5e-06, "epoch": 1.0050241545893719, "percentage": 33.54, "elapsed_time": "10:39:07", "remaining_time": "21:06:26"}
{"current_steps": 660, "total_steps": 1938, "loss": 0.7771, "lr": 5e-06, "epoch": 1.0204830917874397, "percentage": 34.06, "elapsed_time": "10:48:45", "remaining_time": "20:56:13"}
{"current_steps": 670, "total_steps": 1938, "loss": 0.7802, "lr": 5e-06, "epoch": 1.0359420289855072, "percentage": 34.57, "elapsed_time": "10:58:24", "remaining_time": "20:46:02"}
{"current_steps": 680, "total_steps": 1938, "loss": 0.7777, "lr": 5e-06, "epoch": 1.0514009661835748, "percentage": 35.09, "elapsed_time": "11:08:02", "remaining_time": "20:35:52"}
{"current_steps": 690, "total_steps": 1938, "loss": 0.7751, "lr": 5e-06, "epoch": 1.0668599033816426, "percentage": 35.6, "elapsed_time": "11:17:40", "remaining_time": "20:25:43"}
{"current_steps": 700, "total_steps": 1938, "loss": 0.7808, "lr": 5e-06, "epoch": 1.0823188405797102, "percentage": 36.12, "elapsed_time": "11:27:20", "remaining_time": "20:15:36"}
{"current_steps": 710, "total_steps": 1938, "loss": 0.7769, "lr": 5e-06, "epoch": 1.0977777777777777, "percentage": 36.64, "elapsed_time": "11:36:59", "remaining_time": "20:05:30"}
{"current_steps": 720, "total_steps": 1938, "loss": 0.7775, "lr": 5e-06, "epoch": 1.1132367149758453, "percentage": 37.15, "elapsed_time": "11:46:38", "remaining_time": "19:55:24"}
{"current_steps": 730, "total_steps": 1938, "loss": 0.7787, "lr": 5e-06, "epoch": 1.128695652173913, "percentage": 37.67, "elapsed_time": "11:56:17", "remaining_time": "19:45:19"}
{"current_steps": 740, "total_steps": 1938, "loss": 0.7816, "lr": 5e-06, "epoch": 1.1441545893719807, "percentage": 38.18, "elapsed_time": "12:05:56", "remaining_time": "19:35:14"}
{"current_steps": 750, "total_steps": 1938, "loss": 0.7811, "lr": 5e-06, "epoch": 1.1596135265700482, "percentage": 38.7, "elapsed_time": "12:15:35", "remaining_time": "19:25:10"}
{"current_steps": 760, "total_steps": 1938, "loss": 0.7763, "lr": 5e-06, "epoch": 1.175072463768116, "percentage": 39.22, "elapsed_time": "12:25:13", "remaining_time": "19:15:06"}
{"current_steps": 770, "total_steps": 1938, "loss": 0.7821, "lr": 5e-06, "epoch": 1.1905314009661836, "percentage": 39.73, "elapsed_time": "12:34:52", "remaining_time": "19:05:03"}
{"current_steps": 780, "total_steps": 1938, "loss": 0.7771, "lr": 5e-06, "epoch": 1.2059903381642512, "percentage": 40.25, "elapsed_time": "12:44:31", "remaining_time": "18:55:01"}
{"current_steps": 790, "total_steps": 1938, "loss": 0.7832, "lr": 5e-06, "epoch": 1.221449275362319, "percentage": 40.76, "elapsed_time": "12:54:10", "remaining_time": "18:44:59"}
{"current_steps": 800, "total_steps": 1938, "loss": 0.7795, "lr": 5e-06, "epoch": 1.2369082125603865, "percentage": 41.28, "elapsed_time": "13:03:49", "remaining_time": "18:34:59"}
{"current_steps": 810, "total_steps": 1938, "loss": 0.7767, "lr": 5e-06, "epoch": 1.252367149758454, "percentage": 41.8, "elapsed_time": "13:13:29", "remaining_time": "18:25:00"}
{"current_steps": 820, "total_steps": 1938, "loss": 0.7784, "lr": 5e-06, "epoch": 1.2678260869565217, "percentage": 42.31, "elapsed_time": "13:23:09", "remaining_time": "18:15:01"}
{"current_steps": 830, "total_steps": 1938, "loss": 0.7765, "lr": 5e-06, "epoch": 1.2832850241545894, "percentage": 42.83, "elapsed_time": "13:32:48", "remaining_time": "18:05:02"}
{"current_steps": 840, "total_steps": 1938, "loss": 0.7791, "lr": 5e-06, "epoch": 1.298743961352657, "percentage": 43.34, "elapsed_time": "13:42:26", "remaining_time": "17:55:02"}
{"current_steps": 850, "total_steps": 1938, "loss": 0.7775, "lr": 5e-06, "epoch": 1.3142028985507246, "percentage": 43.86, "elapsed_time": "13:52:05", "remaining_time": "17:45:04"}
{"current_steps": 860, "total_steps": 1938, "loss": 0.7743, "lr": 5e-06, "epoch": 1.3296618357487922, "percentage": 44.38, "elapsed_time": "14:01:44", "remaining_time": "17:35:06"}
{"current_steps": 870, "total_steps": 1938, "loss": 0.7768, "lr": 5e-06, "epoch": 1.34512077294686, "percentage": 44.89, "elapsed_time": "14:11:23", "remaining_time": "17:25:09"}
{"current_steps": 880, "total_steps": 1938, "loss": 0.773, "lr": 5e-06, "epoch": 1.3605797101449275, "percentage": 45.41, "elapsed_time": "14:21:03", "remaining_time": "17:15:13"}
{"current_steps": 890, "total_steps": 1938, "loss": 0.774, "lr": 5e-06, "epoch": 1.376038647342995, "percentage": 45.92, "elapsed_time": "14:30:40", "remaining_time": "17:05:15"}
{"current_steps": 900, "total_steps": 1938, "loss": 0.7802, "lr": 5e-06, "epoch": 1.3914975845410629, "percentage": 46.44, "elapsed_time": "14:40:19", "remaining_time": "16:55:18"}
{"current_steps": 910, "total_steps": 1938, "loss": 0.7787, "lr": 5e-06, "epoch": 1.4069565217391304, "percentage": 46.96, "elapsed_time": "14:49:58", "remaining_time": "16:45:22"}
{"current_steps": 920, "total_steps": 1938, "loss": 0.7738, "lr": 5e-06, "epoch": 1.422415458937198, "percentage": 47.47, "elapsed_time": "14:59:38", "remaining_time": "16:35:28"}
{"current_steps": 930, "total_steps": 1938, "loss": 0.7773, "lr": 5e-06, "epoch": 1.4378743961352658, "percentage": 47.99, "elapsed_time": "15:09:17", "remaining_time": "16:25:32"}
{"current_steps": 940, "total_steps": 1938, "loss": 0.7767, "lr": 5e-06, "epoch": 1.4533333333333334, "percentage": 48.5, "elapsed_time": "15:18:55", "remaining_time": "16:15:37"}
{"current_steps": 950, "total_steps": 1938, "loss": 0.7765, "lr": 5e-06, "epoch": 1.468792270531401, "percentage": 49.02, "elapsed_time": "15:28:34", "remaining_time": "16:05:43"}
{"current_steps": 960, "total_steps": 1938, "loss": 0.7728, "lr": 5e-06, "epoch": 1.4842512077294687, "percentage": 49.54, "elapsed_time": "15:38:13", "remaining_time": "15:55:48"}
{"current_steps": 970, "total_steps": 1938, "loss": 0.7777, "lr": 5e-06, "epoch": 1.4997101449275363, "percentage": 50.05, "elapsed_time": "15:47:52", "remaining_time": "15:45:55"}
{"current_steps": 980, "total_steps": 1938, "loss": 0.7747, "lr": 5e-06, "epoch": 1.5151690821256039, "percentage": 50.57, "elapsed_time": "15:57:31", "remaining_time": "15:36:02"}
{"current_steps": 990, "total_steps": 1938, "loss": 0.7786, "lr": 5e-06, "epoch": 1.5306280193236717, "percentage": 51.08, "elapsed_time": "16:07:10", "remaining_time": "15:26:08"}
{"current_steps": 1000, "total_steps": 1938, "loss": 0.7823, "lr": 5e-06, "epoch": 1.546086956521739, "percentage": 51.6, "elapsed_time": "16:16:49", "remaining_time": "15:16:15"}
{"current_steps": 1010, "total_steps": 1938, "loss": 0.7704, "lr": 5e-06, "epoch": 1.5615458937198068, "percentage": 52.12, "elapsed_time": "16:26:28", "remaining_time": "15:06:23"}
{"current_steps": 1020, "total_steps": 1938, "loss": 0.7758, "lr": 5e-06, "epoch": 1.5770048309178744, "percentage": 52.63, "elapsed_time": "16:36:07", "remaining_time": "14:56:30"}
{"current_steps": 1030, "total_steps": 1938, "loss": 0.7751, "lr": 5e-06, "epoch": 1.592463768115942, "percentage": 53.15, "elapsed_time": "16:45:46", "remaining_time": "14:46:38"}
{"current_steps": 1040, "total_steps": 1938, "loss": 0.7742, "lr": 5e-06, "epoch": 1.6079227053140097, "percentage": 53.66, "elapsed_time": "16:55:25", "remaining_time": "14:36:46"}
{"current_steps": 1050, "total_steps": 1938, "loss": 0.777, "lr": 5e-06, "epoch": 1.6233816425120773, "percentage": 54.18, "elapsed_time": "17:05:04", "remaining_time": "14:26:54"}
{"current_steps": 1060, "total_steps": 1938, "loss": 0.7757, "lr": 5e-06, "epoch": 1.6388405797101449, "percentage": 54.7, "elapsed_time": "17:14:43", "remaining_time": "14:17:03"}
{"current_steps": 1070, "total_steps": 1938, "loss": 0.7744, "lr": 5e-06, "epoch": 1.6542995169082126, "percentage": 55.21, "elapsed_time": "17:24:22", "remaining_time": "14:07:12"}
{"current_steps": 1080, "total_steps": 1938, "loss": 0.7753, "lr": 5e-06, "epoch": 1.6697584541062802, "percentage": 55.73, "elapsed_time": "17:34:02", "remaining_time": "13:57:22"}
{"current_steps": 1090, "total_steps": 1938, "loss": 0.7753, "lr": 5e-06, "epoch": 1.6852173913043478, "percentage": 56.24, "elapsed_time": "17:43:41", "remaining_time": "13:47:32"}
{"current_steps": 1100, "total_steps": 1938, "loss": 0.774, "lr": 5e-06, "epoch": 1.7006763285024156, "percentage": 56.76, "elapsed_time": "17:53:20", "remaining_time": "13:37:41"}
{"current_steps": 1110, "total_steps": 1938, "loss": 0.7772, "lr": 5e-06, "epoch": 1.7161352657004831, "percentage": 57.28, "elapsed_time": "18:02:59", "remaining_time": "13:27:51"}
{"current_steps": 1120, "total_steps": 1938, "loss": 0.781, "lr": 5e-06, "epoch": 1.7315942028985507, "percentage": 57.79, "elapsed_time": "18:12:38", "remaining_time": "13:18:01"}
{"current_steps": 1130, "total_steps": 1938, "loss": 0.7749, "lr": 5e-06, "epoch": 1.7470531400966185, "percentage": 58.31, "elapsed_time": "18:22:17", "remaining_time": "13:08:10"}
{"current_steps": 1140, "total_steps": 1938, "loss": 0.7719, "lr": 5e-06, "epoch": 1.7625120772946858, "percentage": 58.82, "elapsed_time": "18:31:55", "remaining_time": "12:58:20"}
{"current_steps": 1150, "total_steps": 1938, "loss": 0.7753, "lr": 5e-06, "epoch": 1.7779710144927536, "percentage": 59.34, "elapsed_time": "18:41:34", "remaining_time": "12:48:31"}
{"current_steps": 1160, "total_steps": 1938, "loss": 0.7777, "lr": 5e-06, "epoch": 1.7934299516908214, "percentage": 59.86, "elapsed_time": "18:51:14", "remaining_time": "12:38:42"}
{"current_steps": 1170, "total_steps": 1938, "loss": 0.7774, "lr": 5e-06, "epoch": 1.8088888888888888, "percentage": 60.37, "elapsed_time": "19:00:52", "remaining_time": "12:28:52"}
{"current_steps": 1180, "total_steps": 1938, "loss": 0.7734, "lr": 5e-06, "epoch": 1.8243478260869566, "percentage": 60.89, "elapsed_time": "19:10:31", "remaining_time": "12:19:04"}
{"current_steps": 1190, "total_steps": 1938, "loss": 0.7743, "lr": 5e-06, "epoch": 1.8398067632850241, "percentage": 61.4, "elapsed_time": "19:20:11", "remaining_time": "12:09:15"}
{"current_steps": 1200, "total_steps": 1938, "loss": 0.7728, "lr": 5e-06, "epoch": 1.8552657004830917, "percentage": 61.92, "elapsed_time": "19:29:49", "remaining_time": "11:59:26"}
{"current_steps": 1210, "total_steps": 1938, "loss": 0.774, "lr": 5e-06, "epoch": 1.8707246376811595, "percentage": 62.44, "elapsed_time": "19:39:28", "remaining_time": "11:49:38"}
{"current_steps": 1220, "total_steps": 1938, "loss": 0.7746, "lr": 5e-06, "epoch": 1.886183574879227, "percentage": 62.95, "elapsed_time": "19:49:07", "remaining_time": "11:39:49"}
{"current_steps": 1230, "total_steps": 1938, "loss": 0.7693, "lr": 5e-06, "epoch": 1.9016425120772946, "percentage": 63.47, "elapsed_time": "19:58:47", "remaining_time": "11:30:02"}
{"current_steps": 1240, "total_steps": 1938, "loss": 0.7717, "lr": 5e-06, "epoch": 1.9171014492753624, "percentage": 63.98, "elapsed_time": "20:08:25", "remaining_time": "11:20:13"}
{"current_steps": 1250, "total_steps": 1938, "loss": 0.7754, "lr": 5e-06, "epoch": 1.93256038647343, "percentage": 64.5, "elapsed_time": "20:18:04", "remaining_time": "11:10:25"}
{"current_steps": 1260, "total_steps": 1938, "loss": 0.7755, "lr": 5e-06, "epoch": 1.9480193236714975, "percentage": 65.02, "elapsed_time": "20:27:43", "remaining_time": "11:00:38"}
{"current_steps": 1270, "total_steps": 1938, "loss": 0.7722, "lr": 5e-06, "epoch": 1.9634782608695653, "percentage": 65.53, "elapsed_time": "20:37:23", "remaining_time": "10:50:50"}
{"current_steps": 1280, "total_steps": 1938, "loss": 0.7755, "lr": 5e-06, "epoch": 1.9789371980676327, "percentage": 66.05, "elapsed_time": "20:47:01", "remaining_time": "10:41:02"}
{"current_steps": 1290, "total_steps": 1938, "loss": 0.7709, "lr": 5e-06, "epoch": 1.9943961352657005, "percentage": 66.56, "elapsed_time": "20:56:39", "remaining_time": "10:31:15"}
{"current_steps": 1293, "total_steps": 1938, "eval_loss": 0.8051349520683289, "epoch": 1.9990338164251207, "percentage": 66.72, "elapsed_time": "21:11:26", "remaining_time": "10:34:14"}
{"current_steps": 1300, "total_steps": 1938, "loss": 0.7604, "lr": 5e-06, "epoch": 2.0100483091787438, "percentage": 67.08, "elapsed_time": "21:19:00", "remaining_time": "10:27:41"}
{"current_steps": 1310, "total_steps": 1938, "loss": 0.7299, "lr": 5e-06, "epoch": 2.0255072463768116, "percentage": 67.6, "elapsed_time": "21:28:40", "remaining_time": "10:17:46"}
{"current_steps": 1320, "total_steps": 1938, "loss": 0.7254, "lr": 5e-06, "epoch": 2.0409661835748794, "percentage": 68.11, "elapsed_time": "21:38:19", "remaining_time": "10:07:51"}
{"current_steps": 1330, "total_steps": 1938, "loss": 0.7312, "lr": 5e-06, "epoch": 2.0564251207729467, "percentage": 68.63, "elapsed_time": "21:47:58", "remaining_time": "9:57:55"}
{"current_steps": 1340, "total_steps": 1938, "loss": 0.7283, "lr": 5e-06, "epoch": 2.0718840579710145, "percentage": 69.14, "elapsed_time": "21:57:37", "remaining_time": "9:48:00"}
{"current_steps": 1350, "total_steps": 1938, "loss": 0.7322, "lr": 5e-06, "epoch": 2.0873429951690823, "percentage": 69.66, "elapsed_time": "22:07:17", "remaining_time": "9:38:06"}
{"current_steps": 1360, "total_steps": 1938, "loss": 0.7317, "lr": 5e-06, "epoch": 2.1028019323671496, "percentage": 70.18, "elapsed_time": "22:16:56", "remaining_time": "9:28:12"}
{"current_steps": 1370, "total_steps": 1938, "loss": 0.7331, "lr": 5e-06, "epoch": 2.1182608695652174, "percentage": 70.69, "elapsed_time": "22:26:35", "remaining_time": "9:18:17"}
{"current_steps": 1380, "total_steps": 1938, "loss": 0.7346, "lr": 5e-06, "epoch": 2.133719806763285, "percentage": 71.21, "elapsed_time": "22:36:15", "remaining_time": "9:08:23"}
{"current_steps": 1390, "total_steps": 1938, "loss": 0.7336, "lr": 5e-06, "epoch": 2.1491787439613526, "percentage": 71.72, "elapsed_time": "22:45:54", "remaining_time": "8:58:29"}
{"current_steps": 1400, "total_steps": 1938, "loss": 0.7299, "lr": 5e-06, "epoch": 2.1646376811594203, "percentage": 72.24, "elapsed_time": "22:55:33", "remaining_time": "8:48:36"}
{"current_steps": 1410, "total_steps": 1938, "loss": 0.7333, "lr": 5e-06, "epoch": 2.1800966183574877, "percentage": 72.76, "elapsed_time": "23:05:13", "remaining_time": "8:38:43"}
{"current_steps": 1420, "total_steps": 1938, "loss": 0.7371, "lr": 5e-06, "epoch": 2.1955555555555555, "percentage": 73.27, "elapsed_time": "23:14:53", "remaining_time": "8:28:50"}
{"current_steps": 1430, "total_steps": 1938, "loss": 0.7314, "lr": 5e-06, "epoch": 2.2110144927536233, "percentage": 73.79, "elapsed_time": "23:24:34", "remaining_time": "8:18:57"}
{"current_steps": 1440, "total_steps": 1938, "loss": 0.7322, "lr": 5e-06, "epoch": 2.2264734299516906, "percentage": 74.3, "elapsed_time": "23:34:14", "remaining_time": "8:09:05"}
{"current_steps": 1450, "total_steps": 1938, "loss": 0.7356, "lr": 5e-06, "epoch": 2.2419323671497584, "percentage": 74.82, "elapsed_time": "23:43:54", "remaining_time": "7:59:13"}
{"current_steps": 1460, "total_steps": 1938, "loss": 0.7418, "lr": 5e-06, "epoch": 2.257391304347826, "percentage": 75.34, "elapsed_time": "23:53:34", "remaining_time": "7:49:20"}
{"current_steps": 1470, "total_steps": 1938, "loss": 0.7319, "lr": 5e-06, "epoch": 2.2728502415458935, "percentage": 75.85, "elapsed_time": "1 day, 0:03:14", "remaining_time": "7:39:28"}
{"current_steps": 1480, "total_steps": 1938, "loss": 0.7332, "lr": 5e-06, "epoch": 2.2883091787439613, "percentage": 76.37, "elapsed_time": "1 day, 0:12:55", "remaining_time": "7:29:37"}
{"current_steps": 1490, "total_steps": 1938, "loss": 0.7342, "lr": 5e-06, "epoch": 2.303768115942029, "percentage": 76.88, "elapsed_time": "1 day, 0:22:35", "remaining_time": "7:19:45"}
{"current_steps": 1500, "total_steps": 1938, "loss": 0.7345, "lr": 5e-06, "epoch": 2.3192270531400965, "percentage": 77.4, "elapsed_time": "1 day, 0:32:14", "remaining_time": "7:09:53"}
{"current_steps": 1510, "total_steps": 1938, "loss": 0.7357, "lr": 5e-06, "epoch": 2.3346859903381643, "percentage": 77.92, "elapsed_time": "1 day, 0:41:55", "remaining_time": "7:00:02"}
{"current_steps": 1520, "total_steps": 1938, "loss": 0.7339, "lr": 5e-06, "epoch": 2.350144927536232, "percentage": 78.43, "elapsed_time": "1 day, 0:51:35", "remaining_time": "6:50:11"}
{"current_steps": 1530, "total_steps": 1938, "loss": 0.7297, "lr": 5e-06, "epoch": 2.3656038647342994, "percentage": 78.95, "elapsed_time": "1 day, 1:01:14", "remaining_time": "6:40:19"}
{"current_steps": 1540, "total_steps": 1938, "loss": 0.7298, "lr": 5e-06, "epoch": 2.381062801932367, "percentage": 79.46, "elapsed_time": "1 day, 1:10:55", "remaining_time": "6:30:29"}
{"current_steps": 1550, "total_steps": 1938, "loss": 0.7368, "lr": 5e-06, "epoch": 2.396521739130435, "percentage": 79.98, "elapsed_time": "1 day, 1:20:35", "remaining_time": "6:20:38"}
{"current_steps": 1560, "total_steps": 1938, "loss": 0.7353, "lr": 5e-06, "epoch": 2.4119806763285023, "percentage": 80.5, "elapsed_time": "1 day, 1:30:15", "remaining_time": "6:10:47"}
{"current_steps": 1570, "total_steps": 1938, "loss": 0.7338, "lr": 5e-06, "epoch": 2.42743961352657, "percentage": 81.01, "elapsed_time": "1 day, 1:39:55", "remaining_time": "6:00:57"}
{"current_steps": 1580, "total_steps": 1938, "loss": 0.7329, "lr": 5e-06, "epoch": 2.442898550724638, "percentage": 81.53, "elapsed_time": "1 day, 1:49:35", "remaining_time": "5:51:06"}
{"current_steps": 1590, "total_steps": 1938, "loss": 0.7367, "lr": 5e-06, "epoch": 2.4583574879227053, "percentage": 82.04, "elapsed_time": "1 day, 1:59:15", "remaining_time": "5:41:16"}
{"current_steps": 1600, "total_steps": 1938, "loss": 0.733, "lr": 5e-06, "epoch": 2.473816425120773, "percentage": 82.56, "elapsed_time": "1 day, 2:08:55", "remaining_time": "5:31:26"}
{"current_steps": 1610, "total_steps": 1938, "loss": 0.7375, "lr": 5e-06, "epoch": 2.4892753623188404, "percentage": 83.08, "elapsed_time": "1 day, 2:18:35", "remaining_time": "5:21:36"}
{"current_steps": 1620, "total_steps": 1938, "loss": 0.7374, "lr": 5e-06, "epoch": 2.504734299516908, "percentage": 83.59, "elapsed_time": "1 day, 2:28:15", "remaining_time": "5:11:46"}
{"current_steps": 1630, "total_steps": 1938, "loss": 0.7311, "lr": 5e-06, "epoch": 2.520193236714976, "percentage": 84.11, "elapsed_time": "1 day, 2:37:54", "remaining_time": "5:01:56"}
{"current_steps": 1640, "total_steps": 1938, "loss": 0.7308, "lr": 5e-06, "epoch": 2.5356521739130433, "percentage": 84.62, "elapsed_time": "1 day, 2:47:34", "remaining_time": "4:52:06"}
{"current_steps": 1650, "total_steps": 1938, "loss": 0.7337, "lr": 5e-06, "epoch": 2.551111111111111, "percentage": 85.14, "elapsed_time": "1 day, 2:57:14", "remaining_time": "4:42:16"}
{"current_steps": 1660, "total_steps": 1938, "loss": 0.7318, "lr": 5e-06, "epoch": 2.566570048309179, "percentage": 85.66, "elapsed_time": "1 day, 3:06:53", "remaining_time": "4:32:27"}
{"current_steps": 1670, "total_steps": 1938, "loss": 0.736, "lr": 5e-06, "epoch": 2.5820289855072462, "percentage": 86.17, "elapsed_time": "1 day, 3:16:34", "remaining_time": "4:22:38"}
{"current_steps": 1680, "total_steps": 1938, "loss": 0.7325, "lr": 5e-06, "epoch": 2.597487922705314, "percentage": 86.69, "elapsed_time": "1 day, 3:26:15", "remaining_time": "4:12:49"}
{"current_steps": 1690, "total_steps": 1938, "loss": 0.7375, "lr": 5e-06, "epoch": 2.6129468599033814, "percentage": 87.2, "elapsed_time": "1 day, 3:35:55", "remaining_time": "4:03:00"}
{"current_steps": 1700, "total_steps": 1938, "loss": 0.7366, "lr": 5e-06, "epoch": 2.628405797101449, "percentage": 87.72, "elapsed_time": "1 day, 3:45:34", "remaining_time": "3:53:10"}
{"current_steps": 1710, "total_steps": 1938, "loss": 0.7372, "lr": 5e-06, "epoch": 2.643864734299517, "percentage": 88.24, "elapsed_time": "1 day, 3:55:14", "remaining_time": "3:43:21"}
{"current_steps": 1720, "total_steps": 1938, "loss": 0.7351, "lr": 5e-06, "epoch": 2.6593236714975843, "percentage": 88.75, "elapsed_time": "1 day, 4:04:54", "remaining_time": "3:33:33"}
{"current_steps": 1730, "total_steps": 1938, "loss": 0.7384, "lr": 5e-06, "epoch": 2.674782608695652, "percentage": 89.27, "elapsed_time": "1 day, 4:14:34", "remaining_time": "3:23:44"}
{"current_steps": 1740, "total_steps": 1938, "loss": 0.7361, "lr": 5e-06, "epoch": 2.69024154589372, "percentage": 89.78, "elapsed_time": "1 day, 4:24:13", "remaining_time": "3:13:55"}
{"current_steps": 1750, "total_steps": 1938, "loss": 0.7309, "lr": 5e-06, "epoch": 2.7057004830917872, "percentage": 90.3, "elapsed_time": "1 day, 4:33:53", "remaining_time": "3:04:07"}
{"current_steps": 1760, "total_steps": 1938, "loss": 0.7296, "lr": 5e-06, "epoch": 2.721159420289855, "percentage": 90.82, "elapsed_time": "1 day, 4:43:32", "remaining_time": "2:54:18"}
{"current_steps": 1770, "total_steps": 1938, "loss": 0.7406, "lr": 5e-06, "epoch": 2.736618357487923, "percentage": 91.33, "elapsed_time": "1 day, 4:53:11", "remaining_time": "2:44:30"}
{"current_steps": 1780, "total_steps": 1938, "loss": 0.7324, "lr": 5e-06, "epoch": 2.75207729468599, "percentage": 91.85, "elapsed_time": "1 day, 5:02:50", "remaining_time": "2:34:42"}
{"current_steps": 1790, "total_steps": 1938, "loss": 0.7331, "lr": 5e-06, "epoch": 2.767536231884058, "percentage": 92.36, "elapsed_time": "1 day, 5:12:29", "remaining_time": "2:24:53"}
{"current_steps": 1800, "total_steps": 1938, "loss": 0.7372, "lr": 5e-06, "epoch": 2.7829951690821257, "percentage": 92.88, "elapsed_time": "1 day, 5:22:08", "remaining_time": "2:15:05"}
{"current_steps": 1810, "total_steps": 1938, "loss": 0.7359, "lr": 5e-06, "epoch": 2.798454106280193, "percentage": 93.4, "elapsed_time": "1 day, 5:31:47", "remaining_time": "2:05:17"}
{"current_steps": 1820, "total_steps": 1938, "loss": 0.7384, "lr": 5e-06, "epoch": 2.813913043478261, "percentage": 93.91, "elapsed_time": "1 day, 5:41:27", "remaining_time": "1:55:30"}
{"current_steps": 1830, "total_steps": 1938, "loss": 0.7324, "lr": 5e-06, "epoch": 2.8293719806763287, "percentage": 94.43, "elapsed_time": "1 day, 5:51:07", "remaining_time": "1:45:42"}
{"current_steps": 1840, "total_steps": 1938, "loss": 0.736, "lr": 5e-06, "epoch": 2.844830917874396, "percentage": 94.94, "elapsed_time": "1 day, 6:00:47", "remaining_time": "1:35:54"}
{"current_steps": 1850, "total_steps": 1938, "loss": 0.7327, "lr": 5e-06, "epoch": 2.860289855072464, "percentage": 95.46, "elapsed_time": "1 day, 6:10:26", "remaining_time": "1:26:07"}
{"current_steps": 1860, "total_steps": 1938, "loss": 0.7374, "lr": 5e-06, "epoch": 2.8757487922705316, "percentage": 95.98, "elapsed_time": "1 day, 6:20:06", "remaining_time": "1:16:19"}
{"current_steps": 1870, "total_steps": 1938, "loss": 0.7351, "lr": 5e-06, "epoch": 2.891207729468599, "percentage": 96.49, "elapsed_time": "1 day, 6:29:47", "remaining_time": "1:06:32"}
{"current_steps": 1880, "total_steps": 1938, "loss": 0.7326, "lr": 5e-06, "epoch": 2.9066666666666667, "percentage": 97.01, "elapsed_time": "1 day, 6:39:27", "remaining_time": "0:56:44"}
{"current_steps": 1890, "total_steps": 1938, "loss": 0.7321, "lr": 5e-06, "epoch": 2.9221256038647345, "percentage": 97.52, "elapsed_time": "1 day, 6:49:07", "remaining_time": "0:46:57"}
{"current_steps": 1900, "total_steps": 1938, "loss": 0.7395, "lr": 5e-06, "epoch": 2.937584541062802, "percentage": 98.04, "elapsed_time": "1 day, 6:58:47", "remaining_time": "0:37:10"}
{"current_steps": 1910, "total_steps": 1938, "loss": 0.7368, "lr": 5e-06, "epoch": 2.9530434782608697, "percentage": 98.56, "elapsed_time": "1 day, 7:08:27", "remaining_time": "0:27:23"}
{"current_steps": 1920, "total_steps": 1938, "loss": 0.7339, "lr": 5e-06, "epoch": 2.9685024154589374, "percentage": 99.07, "elapsed_time": "1 day, 7:18:07", "remaining_time": "0:17:36"}
{"current_steps": 1930, "total_steps": 1938, "loss": 0.7386, "lr": 5e-06, "epoch": 2.983961352657005, "percentage": 99.59, "elapsed_time": "1 day, 7:27:47", "remaining_time": "0:07:49"}
{"current_steps": 1938, "total_steps": 1938, "eval_loss": 0.8025317192077637, "epoch": 2.9963285024154587, "percentage": 100.0, "elapsed_time": "1 day, 7:48:20", "remaining_time": "0:00:00"}
{"current_steps": 1938, "total_steps": 1938, "epoch": 2.9963285024154587, "percentage": 100.0, "elapsed_time": "1 day, 7:49:49", "remaining_time": "0:00:00"}