trainer_log.jsonl

{"current_steps": 10, "total_steps": 195, "loss": 1.2733, "learning_rate": 2.5e-05, "epoch": 0.15384615384615385, "percentage": 5.13, "elapsed_time": "0:01:24", "remaining_time": "0:26:00", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 20, "total_steps": 195, "loss": 1.1302, "learning_rate": 5e-05, "epoch": 0.3076923076923077, "percentage": 10.26, "elapsed_time": "0:02:52", "remaining_time": "0:25:10", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 30, "total_steps": 195, "loss": 1.0472, "learning_rate": 4.959823971496574e-05, "epoch": 0.46153846153846156, "percentage": 15.38, "elapsed_time": "0:04:27", "remaining_time": "0:24:31", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 40, "total_steps": 195, "loss": 1.0394, "learning_rate": 4.8405871765993433e-05, "epoch": 0.6153846153846154, "percentage": 20.51, "elapsed_time": "0:06:03", "remaining_time": "0:23:27", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 50, "total_steps": 195, "loss": 0.9386, "learning_rate": 4.6461219840046654e-05, "epoch": 0.7692307692307693, "percentage": 25.64, "elapsed_time": "0:07:44", "remaining_time": "0:22:25", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 60, "total_steps": 195, "loss": 1.175, "learning_rate": 4.382678665009028e-05, "epoch": 0.9230769230769231, "percentage": 30.77, "elapsed_time": "0:09:20", "remaining_time": "0:21:01", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 70, "total_steps": 195, "loss": 0.8893, "learning_rate": 4.058724504646834e-05, "epoch": 1.0769230769230769, "percentage": 35.9, "elapsed_time": "0:10:59", "remaining_time": "0:19:37", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 80, "total_steps": 195, "loss": 0.8154, "learning_rate": 3.6846716561824965e-05, "epoch": 1.2307692307692308, "percentage": 41.03, "elapsed_time": "0:12:37", "remaining_time": "0:18:08", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 90, "total_steps": 195, "loss": 0.7051, "learning_rate": 3.272542485937369e-05, "epoch": 1.3846153846153846, "percentage": 46.15, "elapsed_time": "0:14:15", "remaining_time": "0:16:38", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 100, "total_steps": 195, "loss": 0.6895, "learning_rate": 2.8355831645441388e-05, "epoch": 1.5384615384615383, "percentage": 51.28, "elapsed_time": "0:15:45", "remaining_time": "0:14:58", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 110, "total_steps": 195, "loss": 0.6998, "learning_rate": 2.3878379241237136e-05, "epoch": 1.6923076923076923, "percentage": 56.41, "elapsed_time": "0:17:18", "remaining_time": "0:13:22", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 120, "total_steps": 195, "loss": 0.6849, "learning_rate": 1.9875396155194242e-05, "epoch": 1.8461538461538463, "percentage": 61.54, "elapsed_time": "0:18:50", "remaining_time": "0:11:46", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 130, "total_steps": 195, "loss": 0.7386, "learning_rate": 1.55886183656402e-05, "epoch": 2.0, "percentage": 66.67, "elapsed_time": "0:20:28", "remaining_time": "0:10:14", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 140, "total_steps": 195, "loss": 0.5885, "learning_rate": 1.1604330125525079e-05, "epoch": 2.1538461538461537, "percentage": 71.79, "elapsed_time": "0:22:02", "remaining_time": "0:08:39", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 150, "total_steps": 195, "loss": 0.4572, "learning_rate": 8.050589737169485e-06, "epoch": 2.3076923076923075, "percentage": 76.92, "elapsed_time": "0:23:40", "remaining_time": "0:07:06", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 160, "total_steps": 195, "loss": 0.4833, "learning_rate": 5.041617340682467e-06, "epoch": 2.4615384615384617, "percentage": 82.05, "elapsed_time": "0:25:17", "remaining_time": "0:05:31", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 170, "total_steps": 195, "loss": 0.5649, "learning_rate": 2.6741237846778676e-06, "epoch": 2.6153846153846154, "percentage": 87.18, "elapsed_time": "0:27:00", "remaining_time": "0:03:58", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 180, "total_steps": 195, "loss": 0.4536, "learning_rate": 1.024202260025861e-06, "epoch": 2.769230769230769, "percentage": 92.31, "elapsed_time": "0:28:30", "remaining_time": "0:02:22", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 190, "total_steps": 195, "loss": 0.4979, "learning_rate": 1.4488260208871397e-07, "epoch": 2.9230769230769234, "percentage": 97.44, "elapsed_time": "0:30:05", "remaining_time": "0:00:47", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 195, "total_steps": 195, "epoch": 3.0, "percentage": 100.0, "elapsed_time": "0:30:51", "remaining_time": "0:00:00", "throughput": "0.00", "total_tokens": 0}