stackexchange_mathematica / trainer_log.jsonl
sedrickkeh's picture
Training in progress, epoch 3
fad8d5c verified
{"current_steps": 10, "total_steps": 120, "loss": 0.9372, "lr": 5e-06, "epoch": 0.25, "percentage": 8.33, "elapsed_time": "0:05:25", "remaining_time": "0:59:38"}
{"current_steps": 20, "total_steps": 120, "loss": 0.8283, "lr": 5e-06, "epoch": 0.5, "percentage": 16.67, "elapsed_time": "0:10:41", "remaining_time": "0:53:26"}
{"current_steps": 30, "total_steps": 120, "loss": 0.7811, "lr": 5e-06, "epoch": 0.75, "percentage": 25.0, "elapsed_time": "0:15:59", "remaining_time": "0:47:59"}
{"current_steps": 40, "total_steps": 120, "loss": 0.7595, "lr": 5e-06, "epoch": 1.0, "percentage": 33.33, "elapsed_time": "0:21:16", "remaining_time": "0:42:33"}
{"current_steps": 40, "total_steps": 120, "eval_loss": 0.7516233921051025, "epoch": 1.0, "percentage": 33.33, "elapsed_time": "0:21:44", "remaining_time": "0:43:28"}
{"current_steps": 50, "total_steps": 120, "loss": 0.7169, "lr": 5e-06, "epoch": 1.25, "percentage": 41.67, "elapsed_time": "0:28:10", "remaining_time": "0:39:27"}
{"current_steps": 60, "total_steps": 120, "loss": 0.7105, "lr": 5e-06, "epoch": 1.5, "percentage": 50.0, "elapsed_time": "0:33:27", "remaining_time": "0:33:27"}
{"current_steps": 70, "total_steps": 120, "loss": 0.6992, "lr": 5e-06, "epoch": 1.75, "percentage": 58.33, "elapsed_time": "0:38:43", "remaining_time": "0:27:39"}
{"current_steps": 80, "total_steps": 120, "loss": 0.6946, "lr": 5e-06, "epoch": 2.0, "percentage": 66.67, "elapsed_time": "0:44:01", "remaining_time": "0:22:00"}
{"current_steps": 80, "total_steps": 120, "eval_loss": 0.7181093096733093, "epoch": 2.0, "percentage": 66.67, "elapsed_time": "0:44:28", "remaining_time": "0:22:14"}
{"current_steps": 90, "total_steps": 120, "loss": 0.651, "lr": 5e-06, "epoch": 2.25, "percentage": 75.0, "elapsed_time": "0:51:04", "remaining_time": "0:17:01"}
{"current_steps": 100, "total_steps": 120, "loss": 0.6462, "lr": 5e-06, "epoch": 2.5, "percentage": 83.33, "elapsed_time": "0:56:23", "remaining_time": "0:11:16"}
{"current_steps": 110, "total_steps": 120, "loss": 0.6444, "lr": 5e-06, "epoch": 2.75, "percentage": 91.67, "elapsed_time": "1:01:42", "remaining_time": "0:05:36"}
{"current_steps": 120, "total_steps": 120, "loss": 0.6445, "lr": 5e-06, "epoch": 3.0, "percentage": 100.0, "elapsed_time": "1:06:58", "remaining_time": "0:00:00"}
{"current_steps": 120, "total_steps": 120, "eval_loss": 0.7114487886428833, "epoch": 3.0, "percentage": 100.0, "elapsed_time": "1:08:43", "remaining_time": "0:00:00"}
{"current_steps": 120, "total_steps": 120, "epoch": 3.0, "percentage": 100.0, "elapsed_time": "1:09:58", "remaining_time": "0:00:00"}