stackexchange_math / trainer_log.jsonl
sedrickkeh's picture
Training in progress, epoch 2
ab43fcf verified
{"current_steps": 10, "total_steps": 126, "loss": 0.7715, "lr": 5e-06, "epoch": 0.23668639053254437, "percentage": 7.94, "elapsed_time": "0:05:25", "remaining_time": "1:02:51"}
{"current_steps": 20, "total_steps": 126, "loss": 0.6752, "lr": 5e-06, "epoch": 0.47337278106508873, "percentage": 15.87, "elapsed_time": "0:10:42", "remaining_time": "0:56:45"}
{"current_steps": 30, "total_steps": 126, "loss": 0.6462, "lr": 5e-06, "epoch": 0.7100591715976331, "percentage": 23.81, "elapsed_time": "0:16:01", "remaining_time": "0:51:16"}
{"current_steps": 40, "total_steps": 126, "loss": 0.629, "lr": 5e-06, "epoch": 0.9467455621301775, "percentage": 31.75, "elapsed_time": "0:21:20", "remaining_time": "0:45:53"}
{"current_steps": 42, "total_steps": 126, "eval_loss": 0.6316158175468445, "epoch": 0.9940828402366864, "percentage": 33.33, "elapsed_time": "0:23:01", "remaining_time": "0:46:02"}
{"current_steps": 50, "total_steps": 126, "loss": 0.6475, "lr": 5e-06, "epoch": 1.183431952662722, "percentage": 39.68, "elapsed_time": "0:28:20", "remaining_time": "0:43:05"}
{"current_steps": 60, "total_steps": 126, "loss": 0.5883, "lr": 5e-06, "epoch": 1.4201183431952662, "percentage": 47.62, "elapsed_time": "0:33:40", "remaining_time": "0:37:02"}
{"current_steps": 70, "total_steps": 126, "loss": 0.5813, "lr": 5e-06, "epoch": 1.6568047337278107, "percentage": 55.56, "elapsed_time": "0:39:00", "remaining_time": "0:31:12"}
{"current_steps": 80, "total_steps": 126, "loss": 0.5807, "lr": 5e-06, "epoch": 1.893491124260355, "percentage": 63.49, "elapsed_time": "0:44:19", "remaining_time": "0:25:29"}
{"current_steps": 84, "total_steps": 126, "eval_loss": 0.6118831038475037, "epoch": 1.9881656804733727, "percentage": 66.67, "elapsed_time": "0:47:10", "remaining_time": "0:23:35"}
{"current_steps": 90, "total_steps": 126, "loss": 0.5993, "lr": 5e-06, "epoch": 2.1301775147928996, "percentage": 71.43, "elapsed_time": "0:51:24", "remaining_time": "0:20:33"}
{"current_steps": 100, "total_steps": 126, "loss": 0.5376, "lr": 5e-06, "epoch": 2.366863905325444, "percentage": 79.37, "elapsed_time": "0:56:42", "remaining_time": "0:14:44"}
{"current_steps": 110, "total_steps": 126, "loss": 0.5388, "lr": 5e-06, "epoch": 2.603550295857988, "percentage": 87.3, "elapsed_time": "1:02:01", "remaining_time": "0:09:01"}
{"current_steps": 120, "total_steps": 126, "loss": 0.5369, "lr": 5e-06, "epoch": 2.8402366863905324, "percentage": 95.24, "elapsed_time": "1:07:19", "remaining_time": "0:03:21"}
{"current_steps": 126, "total_steps": 126, "eval_loss": 0.6119207143783569, "epoch": 2.9822485207100593, "percentage": 100.0, "elapsed_time": "1:12:18", "remaining_time": "0:00:00"}
{"current_steps": 126, "total_steps": 126, "epoch": 2.9822485207100593, "percentage": 100.0, "elapsed_time": "1:13:32", "remaining_time": "0:00:00"}