Files
qwen25_plan_sft_10k/trainer_log.jsonl

35 lines
6.0 KiB
Plaintext
Raw Normal View History

{"current_steps": 610, "total_steps": 939, "loss": 0.3824, "lr": 3.314183409550293e-06, "epoch": 1.9504000000000001, "percentage": 64.96, "elapsed_time": "0:03:38", "remaining_time": "0:01:58"}
{"current_steps": 620, "total_steps": 939, "loss": 0.3858, "lr": 3.140380392754901e-06, "epoch": 1.9824000000000002, "percentage": 66.03, "elapsed_time": "0:07:00", "remaining_time": "0:03:36"}
{"current_steps": 630, "total_steps": 939, "loss": 0.4165, "lr": 2.969147537486175e-06, "epoch": 2.016, "percentage": 67.09, "elapsed_time": "0:10:42", "remaining_time": "0:05:15"}
{"current_steps": 640, "total_steps": 939, "loss": 0.3785, "lr": 2.800721502948506e-06, "epoch": 2.048, "percentage": 68.16, "elapsed_time": "0:14:10", "remaining_time": "0:06:37"}
{"current_steps": 650, "total_steps": 939, "loss": 0.3911, "lr": 2.635335069067617e-06, "epoch": 2.08, "percentage": 69.22, "elapsed_time": "0:17:42", "remaining_time": "0:07:52"}
{"current_steps": 660, "total_steps": 939, "loss": 0.3997, "lr": 2.4732168147677927e-06, "epoch": 2.112, "percentage": 70.29, "elapsed_time": "0:21:13", "remaining_time": "0:08:58"}
{"current_steps": 670, "total_steps": 939, "loss": 0.3896, "lr": 2.314590802055232e-06, "epoch": 2.144, "percentage": 71.35, "elapsed_time": "0:24:38", "remaining_time": "0:09:53"}
{"current_steps": 680, "total_steps": 939, "loss": 0.3966, "lr": 2.159676266344222e-06, "epoch": 2.176, "percentage": 72.42, "elapsed_time": "0:28:10", "remaining_time": "0:10:44"}
{"current_steps": 690, "total_steps": 939, "loss": 0.3791, "lr": 2.0086873134540626e-06, "epoch": 2.208, "percentage": 73.48, "elapsed_time": "0:31:42", "remaining_time": "0:11:26"}
{"current_steps": 700, "total_steps": 939, "loss": 0.3924, "lr": 1.8618326236955908e-06, "epoch": 2.24, "percentage": 74.55, "elapsed_time": "0:35:08", "remaining_time": "0:12:00"}
{"current_steps": 710, "total_steps": 939, "loss": 0.3884, "lr": 1.7193151634562071e-06, "epoch": 2.2720000000000002, "percentage": 75.61, "elapsed_time": "0:38:39", "remaining_time": "0:12:28"}
{"current_steps": 720, "total_steps": 939, "loss": 0.3825, "lr": 1.581331904682089e-06, "epoch": 2.304, "percentage": 76.68, "elapsed_time": "0:42:09", "remaining_time": "0:12:49"}
{"current_steps": 730, "total_steps": 939, "loss": 0.391, "lr": 1.4480735526452427e-06, "epoch": 2.336, "percentage": 77.74, "elapsed_time": "0:45:46", "remaining_time": "0:13:06"}
{"current_steps": 740, "total_steps": 939, "loss": 0.3858, "lr": 1.319724282371664e-06, "epoch": 2.368, "percentage": 78.81, "elapsed_time": "0:49:16", "remaining_time": "0:13:14"}
{"current_steps": 750, "total_steps": 939, "loss": 0.3903, "lr": 1.1964614840949002e-06, "epoch": 2.4, "percentage": 79.87, "elapsed_time": "0:52:41", "remaining_time": "0:13:16"}
{"current_steps": 760, "total_steps": 939, "loss": 0.3849, "lr": 1.078455518086784e-06, "epoch": 2.432, "percentage": 80.94, "elapsed_time": "0:56:18", "remaining_time": "0:13:15"}
{"current_steps": 770, "total_steps": 939, "loss": 0.3902, "lr": 9.658694792042284e-07, "epoch": 2.464, "percentage": 82.0, "elapsed_time": "0:59:48", "remaining_time": "0:13:07"}
{"current_steps": 780, "total_steps": 939, "loss": 0.3774, "lr": 8.58858971477457e-07, "epoch": 2.496, "percentage": 83.07, "elapsed_time": "1:03:07", "remaining_time": "0:12:52"}
{"current_steps": 790, "total_steps": 939, "loss": 0.3899, "lr": 7.575718930512516e-07, "epoch": 2.528, "percentage": 84.13, "elapsed_time": "1:06:32", "remaining_time": "0:12:33"}
{"current_steps": 800, "total_steps": 939, "loss": 0.3954, "lr": 6.621482317764105e-07, "epoch": 2.56, "percentage": 85.2, "elapsed_time": "1:09:56", "remaining_time": "0:12:09"}
{"current_steps": 810, "total_steps": 939, "loss": 0.3809, "lr": 5.727198717339511e-07, "epoch": 2.592, "percentage": 86.26, "elapsed_time": "1:16:15", "remaining_time": "0:12:08"}
{"current_steps": 820, "total_steps": 939, "loss": 0.3881, "lr": 4.894104109594466e-07, "epoch": 2.624, "percentage": 87.33, "elapsed_time": "1:19:44", "remaining_time": "0:11:34"}
{"current_steps": 830, "total_steps": 939, "loss": 0.3869, "lr": 4.123349906194357e-07, "epoch": 2.656, "percentage": 88.39, "elapsed_time": "1:23:04", "remaining_time": "0:10:54"}
{"current_steps": 840, "total_steps": 939, "loss": 0.3859, "lr": 3.416001358759635e-07, "epoch": 2.6879999999999997, "percentage": 89.46, "elapsed_time": "1:26:36", "remaining_time": "0:10:12"}
{"current_steps": 850, "total_steps": 939, "loss": 0.3855, "lr": 2.7730360865923954e-07, "epoch": 2.7199999999999998, "percentage": 90.52, "elapsed_time": "1:29:59", "remaining_time": "0:09:25"}
{"current_steps": 860, "total_steps": 939, "loss": 0.384, "lr": 2.1953427255185122e-07, "epoch": 2.752, "percentage": 91.59, "elapsed_time": "1:33:29", "remaining_time": "0:08:35"}
{"current_steps": 870, "total_steps": 939, "loss": 0.3866, "lr": 1.6837196997130434e-07, "epoch": 2.784, "percentage": 92.65, "elapsed_time": "1:36:59", "remaining_time": "0:07:41"}
{"current_steps": 880, "total_steps": 939, "loss": 0.3924, "lr": 1.2388741182062348e-07, "epoch": 2.816, "percentage": 93.72, "elapsed_time": "1:40:23", "remaining_time": "0:06:43"}
{"current_steps": 890, "total_steps": 939, "loss": 0.391, "lr": 8.614207975952083e-08, "epoch": 2.848, "percentage": 94.78, "elapsed_time": "1:43:55", "remaining_time": "0:05:43"}
{"current_steps": 900, "total_steps": 939, "loss": 0.3651, "lr": 5.518814123121885e-08, "epoch": 2.88, "percentage": 95.85, "elapsed_time": "1:47:24", "remaining_time": "0:04:39"}
{"current_steps": 910, "total_steps": 939, "loss": 0.3891, "lr": 3.10683773623488e-08, "epoch": 2.912, "percentage": 96.91, "elapsed_time": "1:51:03", "remaining_time": "0:03:32"}
{"current_steps": 920, "total_steps": 939, "loss": 0.3982, "lr": 1.3816123835588835e-08, "epoch": 2.944, "percentage": 97.98, "elapsed_time": "1:54:39", "remaining_time": "0:02:22"}
{"current_steps": 930, "total_steps": 939, "loss": 0.3864, "lr": 3.4552248167507576e-09, "epoch": 2.976, "percentage": 99.04, "elapsed_time": "1:58:12", "remaining_time": "0:01:08"}
{"current_steps": 938, "total_steps": 939, "epoch": 3.0, "percentage": 99.89, "elapsed_time": "2:00:47", "remaining_time": "0:00:07"}