| {"current_steps": 10, "total_steps": 453, "loss": 0.8555, "learning_rate": 1.0869565217391305e-05, "epoch": 0.022073532454990687, "percentage": 2.21, "elapsed_time": "1:05:25", "remaining_time": "2 days, 0:18:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 20, "total_steps": 453, "loss": 0.7997, "learning_rate": 2.173913043478261e-05, "epoch": 0.044147064909981375, "percentage": 4.42, "elapsed_time": "2:10:47", "remaining_time": "1 day, 23:11:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 30, "total_steps": 453, "loss": 0.7758, "learning_rate": 3.260869565217392e-05, "epoch": 0.06622059736497206, "percentage": 6.62, "elapsed_time": "3:15:43", "remaining_time": "1 day, 21:59:46", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 40, "total_steps": 453, "loss": 0.7616, "learning_rate": 4.347826086956522e-05, "epoch": 0.08829412981996275, "percentage": 8.83, "elapsed_time": "4:20:42", "remaining_time": "1 day, 20:51:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 50, "total_steps": 453, "loss": 0.7609, "learning_rate": 4.9988084660498037e-05, "epoch": 0.11036766227495344, "percentage": 11.04, "elapsed_time": "5:26:38", "remaining_time": "1 day, 19:52:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 60, "total_steps": 453, "loss": 0.7677, "learning_rate": 4.985416749673074e-05, "epoch": 0.13244119472994412, "percentage": 13.25, "elapsed_time": "6:30:27", "remaining_time": "1 day, 18:37:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 70, "total_steps": 453, "loss": 0.7558, "learning_rate": 4.957223915853709e-05, "epoch": 0.1545147271849348, "percentage": 15.45, "elapsed_time": "7:34:03", "remaining_time": "1 day, 17:24:21", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 80, "total_steps": 453, "loss": 0.7492, "learning_rate": 4.9143978581429445e-05, "epoch": 0.1765882596399255, "percentage": 17.66, "elapsed_time": "8:37:31", "remaining_time": "1 day, 16:12:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 90, "total_steps": 453, "loss": 0.7497, "learning_rate": 4.857193613652711e-05, "epoch": 0.1986617920949162, "percentage": 19.87, "elapsed_time": "9:40:39", "remaining_time": "1 day, 15:01:59", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 100, "total_steps": 453, "loss": 0.7433, "learning_rate": 4.78595184426236e-05, "epoch": 0.22073532454990688, "percentage": 22.08, "elapsed_time": "10:43:40", "remaining_time": "1 day, 13:52:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 110, "total_steps": 453, "loss": 0.744, "learning_rate": 4.7010968079140294e-05, "epoch": 0.24280885700489757, "percentage": 24.28, "elapsed_time": "11:46:44", "remaining_time": "1 day, 12:43:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 120, "total_steps": 453, "loss": 0.7378, "learning_rate": 4.6031338320779534e-05, "epoch": 0.26488238945988823, "percentage": 26.49, "elapsed_time": "12:49:47", "remaining_time": "1 day, 11:36:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 130, "total_steps": 453, "loss": 0.7388, "learning_rate": 4.492646304433711e-05, "epoch": 0.28695592191487895, "percentage": 28.7, "elapsed_time": "13:52:49", "remaining_time": "1 day, 10:29:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 140, "total_steps": 453, "loss": 0.7301, "learning_rate": 4.3702921986884574e-05, "epoch": 0.3090294543698696, "percentage": 30.91, "elapsed_time": "14:55:49", "remaining_time": "1 day, 9:22:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 150, "total_steps": 453, "loss": 0.7407, "learning_rate": 4.236800156221536e-05, "epoch": 0.33110298682486033, "percentage": 33.11, "elapsed_time": "15:58:47", "remaining_time": "1 day, 8:16:45", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 160, "total_steps": 453, "loss": 0.7319, "learning_rate": 4.092965146890002e-05, "epoch": 0.353176519279851, "percentage": 35.32, "elapsed_time": "17:01:45", "remaining_time": "1 day, 7:11:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 170, "total_steps": 453, "loss": 0.7292, "learning_rate": 3.9396437348357684e-05, "epoch": 0.3752500517348417, "percentage": 37.53, "elapsed_time": "18:04:38", "remaining_time": "1 day, 6:05:36", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 180, "total_steps": 453, "loss": 0.7488, "learning_rate": 3.777748977487366e-05, "epoch": 0.3973235841898324, "percentage": 39.74, "elapsed_time": "19:07:38", "remaining_time": "1 day, 5:00:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 190, "total_steps": 453, "loss": 0.7305, "learning_rate": 3.608244988133713e-05, "epoch": 0.41939711664482304, "percentage": 41.94, "elapsed_time": "20:10:54", "remaining_time": "1 day, 3:56:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 200, "total_steps": 453, "loss": 0.7329, "learning_rate": 3.432141194450772e-05, "epoch": 0.44147064909981376, "percentage": 44.15, "elapsed_time": "21:14:40", "remaining_time": "1 day, 2:52:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 210, "total_steps": 453, "loss": 0.729, "learning_rate": 3.2504863271726286e-05, "epoch": 0.4635441815548044, "percentage": 46.36, "elapsed_time": "22:17:56", "remaining_time": "1 day, 1:48:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 220, "total_steps": 453, "loss": 0.7392, "learning_rate": 3.064362174705578e-05, "epoch": 0.48561771400979514, "percentage": 48.57, "elapsed_time": "23:21:26", "remaining_time": "1 day, 0:44:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 230, "total_steps": 453, "loss": 0.7323, "learning_rate": 2.8748771408776466e-05, "epoch": 0.5076912464647858, "percentage": 50.77, "elapsed_time": "1 day, 0:27:42", "remaining_time": "23:43:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 240, "total_steps": 453, "loss": 0.7391, "learning_rate": 2.683159644188339e-05, "epoch": 0.5297647789197765, "percentage": 52.98, "elapsed_time": "1 day, 1:35:35", "remaining_time": "22:42:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 250, "total_steps": 453, "loss": 0.7304, "learning_rate": 2.4903513978673077e-05, "epoch": 0.5518383113747672, "percentage": 55.19, "elapsed_time": "1 day, 2:43:22", "remaining_time": "21:41:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 260, "total_steps": 453, "loss": 0.7248, "learning_rate": 2.2976006107604482e-05, "epoch": 0.5739118438297579, "percentage": 57.4, "elapsed_time": "1 day, 3:47:40", "remaining_time": "20:37:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 270, "total_steps": 453, "loss": 0.7361, "learning_rate": 2.1060551495333818e-05, "epoch": 0.5959853762847486, "percentage": 59.6, "elapsed_time": "1 day, 4:51:30", "remaining_time": "19:33:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 280, "total_steps": 453, "loss": 0.7205, "learning_rate": 1.9168557029126963e-05, "epoch": 0.6180589087397392, "percentage": 61.81, "elapsed_time": "1 day, 5:55:20", "remaining_time": "18:29:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 290, "total_steps": 453, "loss": 0.7215, "learning_rate": 1.7311289886731408e-05, "epoch": 0.6401324411947299, "percentage": 64.02, "elapsed_time": "1 day, 6:59:03", "remaining_time": "17:24:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 300, "total_steps": 453, "loss": 0.7204, "learning_rate": 1.549981043824425e-05, "epoch": 0.6622059736497207, "percentage": 66.23, "elapsed_time": "1 day, 8:02:33", "remaining_time": "16:20:30", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 310, "total_steps": 453, "loss": 0.7217, "learning_rate": 1.3744906379558165e-05, "epoch": 0.6842795061047113, "percentage": 68.43, "elapsed_time": "1 day, 9:05:54", "remaining_time": "15:16:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 320, "total_steps": 453, "loss": 0.7288, "learning_rate": 1.2057028489632682e-05, "epoch": 0.706353038559702, "percentage": 70.64, "elapsed_time": "1 day, 10:09:19", "remaining_time": "14:11:45", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 330, "total_steps": 453, "loss": 0.7128, "learning_rate": 1.0446228394168356e-05, "epoch": 0.7284265710146927, "percentage": 72.85, "elapsed_time": "1 day, 11:12:40", "remaining_time": "13:07:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 340, "total_steps": 453, "loss": 0.7212, "learning_rate": 8.922098706312548e-06, "epoch": 0.7505001034696834, "percentage": 75.06, "elapsed_time": "1 day, 12:16:03", "remaining_time": "12:03:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 350, "total_steps": 453, "loss": 0.7109, "learning_rate": 7.493715900870027e-06, "epoch": 0.7725736359246741, "percentage": 77.26, "elapsed_time": "1 day, 13:19:26", "remaining_time": "10:59:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 360, "total_steps": 453, "loss": 0.7206, "learning_rate": 6.169586262213081e-06, "epoch": 0.7946471683796648, "percentage": 79.47, "elapsed_time": "1 day, 14:22:50", "remaining_time": "9:54:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 370, "total_steps": 453, "loss": 0.7166, "learning_rate": 4.957595227781395e-06, "epoch": 0.8167207008346554, "percentage": 81.68, "elapsed_time": "1 day, 15:26:17", "remaining_time": "8:50:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 380, "total_steps": 453, "loss": 0.7095, "learning_rate": 3.864960428840375e-06, "epoch": 0.8387942332896461, "percentage": 83.89, "elapsed_time": "1 day, 16:29:44", "remaining_time": "7:46:46", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 390, "total_steps": 453, "loss": 0.7088, "learning_rate": 2.8981887081491576e-06, "epoch": 0.8608677657446369, "percentage": 86.09, "elapsed_time": "1 day, 17:33:28", "remaining_time": "6:42:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 400, "total_steps": 453, "loss": 0.7189, "learning_rate": 2.0630373705058407e-06, "epoch": 0.8829412981996275, "percentage": 88.3, "elapsed_time": "1 day, 18:36:48", "remaining_time": "5:38:46", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 410, "total_steps": 453, "loss": 0.7215, "learning_rate": 1.3644798969302403e-06, "epoch": 0.9050148306546182, "percentage": 90.51, "elapsed_time": "1 day, 19:40:02", "remaining_time": "4:34:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 420, "total_steps": 453, "loss": 0.7155, "learning_rate": 8.066763266625282e-07, "epoch": 0.9270883631096088, "percentage": 92.72, "elapsed_time": "1 day, 20:43:43", "remaining_time": "3:30:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 430, "total_steps": 453, "loss": 0.7076, "learning_rate": 3.929484833584546e-07, "epoch": 0.9491618955645996, "percentage": 94.92, "elapsed_time": "1 day, 21:47:27", "remaining_time": "2:26:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 440, "total_steps": 453, "loss": 0.7176, "learning_rate": 1.2576019301373532e-07, "epoch": 0.9712354280195903, "percentage": 97.13, "elapsed_time": "1 day, 22:51:11", "remaining_time": "1:23:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 450, "total_steps": 453, "loss": 0.7082, "learning_rate": 6.702611423550775e-09, "epoch": 0.993308960474581, "percentage": 99.34, "elapsed_time": "1 day, 23:54:56", "remaining_time": "0:19:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 453, "total_steps": 453, "epoch": 0.9999310202110782, "percentage": 100.0, "elapsed_time": "2 days, 0:18:26", "remaining_time": "0:00:00", "throughput": "0.00", "total_tokens": 0} | |