trainer_log.jsonl

{"current_steps": 5, "total_steps": 62, "loss": 0.8189, "learning_rate": 4.920192797165511e-05, "epoch": 0.08, "percentage": 8.06, "elapsed_time": "0:04:59", "remaining_time": "0:56:59", "throughput": 180.63, "total_tokens": 54176}
{"current_steps": 10, "total_steps": 62, "loss": 0.6742, "learning_rate": 4.685866540361456e-05, "epoch": 0.16, "percentage": 16.13, "elapsed_time": "0:09:21", "remaining_time": "0:48:40", "throughput": 182.53, "total_tokens": 102528}
{"current_steps": 15, "total_steps": 62, "loss": 0.6816, "learning_rate": 4.3119819680728e-05, "epoch": 0.24, "percentage": 24.19, "elapsed_time": "0:14:05", "remaining_time": "0:44:09", "throughput": 180.81, "total_tokens": 152912}
{"current_steps": 20, "total_steps": 62, "loss": 0.6588, "learning_rate": 3.822410025817406e-05, "epoch": 0.32, "percentage": 32.26, "elapsed_time": "0:18:45", "remaining_time": "0:39:23", "throughput": 181.6, "total_tokens": 204432}
{"current_steps": 25, "total_steps": 62, "loss": 0.5882, "learning_rate": 3.2484078074333954e-05, "epoch": 0.4, "percentage": 40.32, "elapsed_time": "0:23:35", "remaining_time": "0:34:55", "throughput": 179.77, "total_tokens": 254512}
{"current_steps": 30, "total_steps": 62, "loss": 0.6473, "learning_rate": 2.6266229220967818e-05, "epoch": 0.48, "percentage": 48.39, "elapsed_time": "0:27:52", "remaining_time": "0:29:44", "throughput": 178.51, "total_tokens": 298608}
{"current_steps": 35, "total_steps": 62, "loss": 0.6343, "learning_rate": 1.9967536997783494e-05, "epoch": 0.56, "percentage": 56.45, "elapsed_time": "0:32:46", "remaining_time": "0:25:16", "throughput": 177.47, "total_tokens": 348928}
{"current_steps": 40, "total_steps": 62, "loss": 0.5067, "learning_rate": 1.399014621105914e-05, "epoch": 0.64, "percentage": 64.52, "elapsed_time": "0:37:00", "remaining_time": "0:20:21", "throughput": 178.92, "total_tokens": 397216}
{"current_steps": 45, "total_steps": 62, "loss": 0.5386, "learning_rate": 8.715687931944449e-06, "epoch": 0.72, "percentage": 72.58, "elapsed_time": "0:41:30", "remaining_time": "0:15:40", "throughput": 178.61, "total_tokens": 444832}
{"current_steps": 50, "total_steps": 62, "loss": 0.563, "learning_rate": 4.480913969818098e-06, "epoch": 0.8, "percentage": 80.65, "elapsed_time": "0:46:06", "remaining_time": "0:11:03", "throughput": 177.36, "total_tokens": 490640}
{"current_steps": 55, "total_steps": 62, "loss": 0.5484, "learning_rate": 1.5561966963229924e-06, "epoch": 0.88, "percentage": 88.71, "elapsed_time": "0:50:03", "remaining_time": "0:06:22", "throughput": 179.23, "total_tokens": 538400}
{"current_steps": 60, "total_steps": 62, "loss": 0.5979, "learning_rate": 1.2826691520262114e-07, "epoch": 0.96, "percentage": 96.77, "elapsed_time": "0:54:46", "remaining_time": "0:01:49", "throughput": 179.48, "total_tokens": 589856}
{"current_steps": 62, "total_steps": 62, "epoch": 0.992, "percentage": 100.0, "elapsed_time": "0:57:01", "remaining_time": "0:00:00", "throughput": 178.28, "total_tokens": 610016}