TianyiQ's picture
Upload ./trainer_log.jsonl with huggingface_hub
9c92438 verified
{"current_steps": 1, "total_steps": 40700, "loss": 2.7384, "eval_loss": null, "predict_loss": null, "reward": null, "learning_rate": 0.0, "epoch": 9.828009828009828e-05, "percentage": 0.0, "elapsed_time": "0:00:07", "remaining_time": "3 days, 10:46:01"}
{"current_steps": 2035, "total_steps": 40700, "loss": 2.4775, "eval_loss": null, "predict_loss": null, "reward": null, "learning_rate": 1.9927939731411727e-06, "epoch": 0.2, "percentage": 5.0, "elapsed_time": "0:48:37", "remaining_time": "15:23:44"}
{"current_steps": 4070, "total_steps": 40700, "loss": 2.404, "eval_loss": null, "predict_loss": null, "reward": null, "learning_rate": 2.238606233181306e-06, "epoch": 0.4, "percentage": 10.0, "elapsed_time": "1:34:37", "remaining_time": "14:11:37"}
{"current_steps": 4070, "total_steps": 40700, "loss": null, "eval_loss": 2.388622283935547, "predict_loss": null, "reward": null, "learning_rate": null, "epoch": 0.4, "percentage": 10.0, "elapsed_time": "1:34:37", "remaining_time": "14:11:37"}
{"current_steps": 6105, "total_steps": 40700, "loss": 2.3695, "eval_loss": null, "predict_loss": null, "reward": null, "learning_rate": 1.2183995585808034e-06, "epoch": 0.6, "percentage": 15.0, "elapsed_time": "2:27:16", "remaining_time": "13:54:33"}
{"current_steps": 8140, "total_steps": 40700, "loss": 2.3519, "eval_loss": null, "predict_loss": null, "reward": null, "learning_rate": 6.503021712782451e-07, "epoch": 0.8, "percentage": 20.0, "elapsed_time": "3:12:33", "remaining_time": "12:50:13"}
{"current_steps": 8140, "total_steps": 40700, "loss": null, "eval_loss": 2.347768545150757, "predict_loss": null, "reward": null, "learning_rate": null, "epoch": 0.8, "percentage": 20.0, "elapsed_time": "3:12:33", "remaining_time": "12:50:13"}
{"current_steps": 10175, "total_steps": 40700, "loss": 2.341, "eval_loss": null, "predict_loss": null, "reward": null, "learning_rate": 3.454633956957804e-07, "epoch": 1.0, "percentage": 25.0, "elapsed_time": "4:04:24", "remaining_time": "12:13:14"}
{"current_steps": 12210, "total_steps": 40700, "loss": 2.272, "eval_loss": null, "predict_loss": null, "reward": null, "learning_rate": 1.8843776358550853e-07, "epoch": 1.2, "percentage": 30.0, "elapsed_time": "4:49:33", "remaining_time": "11:15:37"}
{"current_steps": 12210, "total_steps": 40700, "loss": null, "eval_loss": 2.3399369716644287, "predict_loss": null, "reward": null, "learning_rate": null, "epoch": 1.2, "percentage": 30.0, "elapsed_time": "4:49:33", "remaining_time": "11:15:37"}
{"current_steps": 14245, "total_steps": 40700, "loss": 2.2758, "eval_loss": null, "predict_loss": null, "reward": null, "learning_rate": 1.1134838176940004e-07, "epoch": 1.4, "percentage": 35.0, "elapsed_time": "5:41:20", "remaining_time": "10:33:54"}
{"current_steps": 16280, "total_steps": 40700, "loss": 2.2737, "eval_loss": null, "predict_loss": null, "reward": null, "learning_rate": 7.549690625694449e-08, "epoch": 1.6, "percentage": 40.0, "elapsed_time": "6:26:27", "remaining_time": "9:39:41"}
{"current_steps": 16280, "total_steps": 40700, "loss": null, "eval_loss": 2.3373029232025146, "predict_loss": null, "reward": null, "learning_rate": null, "epoch": 1.6, "percentage": 40.0, "elapsed_time": "6:26:27", "remaining_time": "9:39:41"}
{"current_steps": 18315, "total_steps": 40700, "loss": 2.2725, "eval_loss": null, "predict_loss": null, "reward": null, "learning_rate": 5.980471884190691e-08, "epoch": 1.8, "percentage": 45.0, "elapsed_time": "7:18:12", "remaining_time": "8:55:34"}
{"current_steps": 20350, "total_steps": 40700, "loss": 2.2735, "eval_loss": null, "predict_loss": null, "reward": null, "learning_rate": 5.3444261786522375e-08, "epoch": 2.0, "percentage": 50.0, "elapsed_time": "8:03:21", "remaining_time": "8:03:21"}
{"current_steps": 20350, "total_steps": 40700, "loss": null, "eval_loss": 2.3360562324523926, "predict_loss": null, "reward": null, "learning_rate": null, "epoch": 2.0, "percentage": 50.0, "elapsed_time": "8:03:21", "remaining_time": "8:03:21"}
{"current_steps": 22385, "total_steps": 40700, "loss": 2.2617, "eval_loss": null, "predict_loss": null, "reward": null, "learning_rate": 5.108383189787308e-08, "epoch": 2.2, "percentage": 55.0, "elapsed_time": "8:55:06", "remaining_time": "7:17:48"}
{"current_steps": 24420, "total_steps": 40700, "loss": 2.2629, "eval_loss": null, "predict_loss": null, "reward": null, "learning_rate": 5.029810361474273e-08, "epoch": 2.4, "percentage": 60.0, "elapsed_time": "9:40:13", "remaining_time": "6:26:48"}
{"current_steps": 24420, "total_steps": 40700, "loss": null, "eval_loss": 2.3360707759857178, "predict_loss": null, "reward": null, "learning_rate": null, "epoch": 2.4, "percentage": 60.0, "elapsed_time": "9:40:13", "remaining_time": "6:26:48"}
{"current_steps": 26455, "total_steps": 40700, "loss": 2.2617, "eval_loss": null, "predict_loss": null, "reward": null, "learning_rate": 5.006893878600473e-08, "epoch": 2.6, "percentage": 65.0, "elapsed_time": "10:31:55", "remaining_time": "5:40:16"}
{"current_steps": 28490, "total_steps": 40700, "loss": 2.2655, "eval_loss": null, "predict_loss": null, "reward": null, "learning_rate": 5.001272696084112e-08, "epoch": 2.8, "percentage": 70.0, "elapsed_time": "11:16:59", "remaining_time": "4:50:08"}
{"current_steps": 28490, "total_steps": 40700, "loss": null, "eval_loss": 2.3352677822113037, "predict_loss": null, "reward": null, "learning_rate": null, "epoch": 2.8, "percentage": 70.0, "elapsed_time": "11:16:59", "remaining_time": "4:50:08"}
{"current_steps": 30525, "total_steps": 40700, "loss": 2.2615, "eval_loss": null, "predict_loss": null, "reward": null, "learning_rate": 5.0001729586514593e-08, "epoch": 3.0, "percentage": 75.0, "elapsed_time": "12:08:41", "remaining_time": "4:02:53"}
{"current_steps": 32560, "total_steps": 40700, "loss": 2.2567, "eval_loss": null, "predict_loss": null, "reward": null, "learning_rate": 5.0000150744726626e-08, "epoch": 3.2, "percentage": 80.0, "elapsed_time": "12:53:45", "remaining_time": "3:13:26"}
{"current_steps": 32560, "total_steps": 40700, "loss": null, "eval_loss": 2.334810733795166, "predict_loss": null, "reward": null, "learning_rate": null, "epoch": 3.2, "percentage": 80.0, "elapsed_time": "12:53:45", "remaining_time": "3:13:26"}
{"current_steps": 34595, "total_steps": 40700, "loss": 2.2555, "eval_loss": null, "predict_loss": null, "reward": null, "learning_rate": 5.000000648758777e-08, "epoch": 3.4, "percentage": 85.0, "elapsed_time": "13:45:25", "remaining_time": "2:25:39"}
{"current_steps": 36630, "total_steps": 40700, "loss": 2.2581, "eval_loss": null, "predict_loss": null, "reward": null, "learning_rate": 5.000000007849755e-08, "epoch": 3.6, "percentage": 90.0, "elapsed_time": "14:30:28", "remaining_time": "1:36:43"}
{"current_steps": 36630, "total_steps": 40700, "loss": null, "eval_loss": 2.3342058658599854, "predict_loss": null, "reward": null, "learning_rate": null, "epoch": 3.6, "percentage": 90.0, "elapsed_time": "14:30:28", "remaining_time": "1:36:43"}
{"current_steps": 38665, "total_steps": 40700, "loss": 2.2592, "eval_loss": null, "predict_loss": null, "reward": null, "learning_rate": 5.000000000004389e-08, "epoch": 3.8, "percentage": 95.0, "elapsed_time": "15:22:10", "remaining_time": "0:48:32"}
{"current_steps": 40700, "total_steps": 40700, "loss": 2.2607, "eval_loss": null, "predict_loss": null, "reward": null, "learning_rate": 5e-08, "epoch": 4.0, "percentage": 100.0, "elapsed_time": "16:07:14", "remaining_time": "0:00:00"}
{"current_steps": 40700, "total_steps": 40700, "loss": null, "eval_loss": 2.3333284854888916, "predict_loss": null, "reward": null, "learning_rate": null, "epoch": 4.0, "percentage": 100.0, "elapsed_time": "16:07:14", "remaining_time": "0:00:00"}
{"current_steps": 40700, "total_steps": 40700, "loss": null, "eval_loss": null, "predict_loss": null, "reward": null, "learning_rate": null, "epoch": 4.0, "percentage": 100.0, "elapsed_time": "16:07:14", "remaining_time": "0:00:00"}
{"current_steps": 566, "total_steps": 566, "loss": null, "eval_loss": 2.3333284854888916, "predict_loss": null, "reward": null, "learning_rate": null, "epoch": 4.0, "percentage": 100.0, "elapsed_time": "16:20:14", "remaining_time": "0:00:00"}