Files
Qwen2.5-0.5B-Math-SFT-1024/trainer_log.jsonl

23 lines
4.1 KiB
Plaintext
Raw Permalink Normal View History

{"current_steps": 10, "total_steps": 204, "loss": 0.8932, "lr": 1.6363636363636366e-05, "epoch": 0.14814814814814814, "percentage": 4.9, "elapsed_time": "0:00:14", "remaining_time": "0:04:50"}
{"current_steps": 20, "total_steps": 204, "loss": 0.7355, "lr": 1.991533160993366e-05, "epoch": 0.2962962962962963, "percentage": 9.8, "elapsed_time": "0:00:29", "remaining_time": "0:04:30"}
{"current_steps": 30, "total_steps": 204, "loss": 0.6846, "lr": 1.9573822052208013e-05, "epoch": 0.4444444444444444, "percentage": 14.71, "elapsed_time": "0:00:43", "remaining_time": "0:04:14"}
{"current_steps": 40, "total_steps": 204, "loss": 0.6672, "lr": 1.8979201311434434e-05, "epoch": 0.5925925925925926, "percentage": 19.61, "elapsed_time": "0:00:58", "remaining_time": "0:03:59"}
{"current_steps": 50, "total_steps": 204, "loss": 0.6607, "lr": 1.8147189876080463e-05, "epoch": 0.7407407407407407, "percentage": 24.51, "elapsed_time": "0:01:12", "remaining_time": "0:03:44"}
{"current_steps": 60, "total_steps": 204, "loss": 0.6538, "lr": 1.709978433199901e-05, "epoch": 0.8888888888888888, "percentage": 29.41, "elapsed_time": "0:01:27", "remaining_time": "0:03:29"}
{"current_steps": 70, "total_steps": 204, "loss": 0.6235, "lr": 1.586467582019392e-05, "epoch": 1.0296296296296297, "percentage": 34.31, "elapsed_time": "0:01:40", "remaining_time": "0:03:13"}
{"current_steps": 80, "total_steps": 204, "loss": 0.5681, "lr": 1.4474517942843173e-05, "epoch": 1.1777777777777778, "percentage": 39.22, "elapsed_time": "0:01:55", "remaining_time": "0:02:58"}
{"current_steps": 90, "total_steps": 204, "loss": 0.5632, "lr": 1.2966063472563686e-05, "epoch": 1.325925925925926, "percentage": 44.12, "elapsed_time": "0:02:09", "remaining_time": "0:02:44"}
{"current_steps": 100, "total_steps": 204, "loss": 0.5595, "lr": 1.137919268846074e-05, "epoch": 1.474074074074074, "percentage": 49.02, "elapsed_time": "0:02:24", "remaining_time": "0:02:30"}
{"current_steps": 110, "total_steps": 204, "loss": 0.5516, "lr": 9.755859027658848e-06, "epoch": 1.6222222222222222, "percentage": 53.92, "elapsed_time": "0:02:38", "remaining_time": "0:02:15"}
{"current_steps": 120, "total_steps": 204, "loss": 0.552, "lr": 8.138979927010964e-06, "epoch": 1.7703703703703704, "percentage": 58.82, "elapsed_time": "0:02:53", "remaining_time": "0:02:01"}
{"current_steps": 130, "total_steps": 204, "loss": 0.5492, "lr": 6.571302178736404e-06, "epoch": 1.9185185185185185, "percentage": 63.73, "elapsed_time": "0:03:07", "remaining_time": "0:01:46"}
{"current_steps": 140, "total_steps": 204, "loss": 0.5272, "lr": 5.094271797534404e-06, "epoch": 2.0592592592592593, "percentage": 68.63, "elapsed_time": "0:03:21", "remaining_time": "0:01:31"}
{"current_steps": 150, "total_steps": 204, "loss": 0.492, "lr": 3.74693827744665e-06, "epoch": 2.2074074074074073, "percentage": 73.53, "elapsed_time": "0:03:35", "remaining_time": "0:01:17"}
{"current_steps": 160, "total_steps": 204, "loss": 0.4981, "lr": 2.564922207551337e-06, "epoch": 2.3555555555555556, "percentage": 78.43, "elapsed_time": "0:03:49", "remaining_time": "0:01:03"}
{"current_steps": 170, "total_steps": 204, "loss": 0.4926, "lr": 1.579473540500751e-06, "epoch": 2.5037037037037035, "percentage": 83.33, "elapsed_time": "0:04:04", "remaining_time": "0:00:48"}
{"current_steps": 180, "total_steps": 204, "loss": 0.4956, "lr": 8.166454112487287e-07, "epoch": 2.651851851851852, "percentage": 88.24, "elapsed_time": "0:04:18", "remaining_time": "0:00:34"}
{"current_steps": 190, "total_steps": 204, "loss": 0.4903, "lr": 2.9660534841699175e-07, "epoch": 2.8, "percentage": 93.14, "elapsed_time": "0:04:32", "remaining_time": "0:00:20"}
{"current_steps": 200, "total_steps": 204, "loss": 0.4883, "lr": 3.310208838458562e-08, "epoch": 2.948148148148148, "percentage": 98.04, "elapsed_time": "0:04:47", "remaining_time": "0:00:05"}
{"current_steps": 200, "total_steps": 204, "eval_loss": 0.6684299111366272, "epoch": 2.948148148148148, "percentage": 98.04, "elapsed_time": "0:04:47", "remaining_time": "0:00:05"}
{"current_steps": 204, "total_steps": 204, "epoch": 3.0, "percentage": 100.0, "elapsed_time": "0:05:08", "remaining_time": "0:00:00"}