llama-factory/saves/Mistral-7B-v0.3-Chinese-Chat/trainer_log.jsonl

{"current_steps": 5, "total_steps": 210, "loss": 2.7275, "learning_rate": 2.380952380952381e-05, "epoch": 0.1423487544483986, "percentage": 2.38, "elapsed_time": "0:01:25", "remaining_time": "0:58:22", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 10, "total_steps": 210, "loss": 2.3848, "learning_rate": 4.761904761904762e-05, "epoch": 0.2846975088967972, "percentage": 4.76, "elapsed_time": "0:02:55", "remaining_time": "0:58:24", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 15, "total_steps": 210, "loss": 1.9534, "learning_rate": 7.142857142857143e-05, "epoch": 0.42704626334519574, "percentage": 7.14, "elapsed_time": "0:04:16", "remaining_time": "0:55:39", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 20, "total_steps": 210, "loss": 1.7776, "learning_rate": 9.523809523809524e-05, "epoch": 0.5693950177935944, "percentage": 9.52, "elapsed_time": "0:05:39", "remaining_time": "0:53:41", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 25, "total_steps": 210, "loss": 1.6551, "learning_rate": 9.988952191691925e-05, "epoch": 0.7117437722419929, "percentage": 11.9, "elapsed_time": "0:07:12", "remaining_time": "0:53:22", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 30, "total_steps": 210, "loss": 1.6106, "learning_rate": 9.944154131125642e-05, "epoch": 0.8540925266903915, "percentage": 14.29, "elapsed_time": "0:08:37", "remaining_time": "0:51:46", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 35, "total_steps": 210, "loss": 1.6196, "learning_rate": 9.865224352899119e-05, "epoch": 0.99644128113879, "percentage": 16.67, "elapsed_time": "0:10:09", "remaining_time": "0:50:47", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 35, "total_steps": 210, "eval_loss": 1.531648874282837, "epoch": 0.99644128113879, "percentage": 16.67, "elapsed_time": "0:10:12", "remaining_time": "0:51:03", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 40, "total_steps": 210, "loss": 1.4904, "learning_rate": 9.752707744739145e-05, "epoch": 1.1387900355871885, "percentage": 19.05, "elapsed_time": "0:11:38", "remaining_time": "0:49:30", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 45, "total_steps": 210, "loss": 1.4244, "learning_rate": 9.607381059352038e-05, "epoch": 1.281138790035587, "percentage": 21.43, "elapsed_time": "0:13:11", "remaining_time": "0:48:21", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 50, "total_steps": 210, "loss": 1.4003, "learning_rate": 9.430247552150673e-05, "epoch": 1.4234875444839858, "percentage": 23.81, "elapsed_time": "0:14:41", "remaining_time": "0:46:59", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 55, "total_steps": 210, "loss": 1.3709, "learning_rate": 9.22253005533154e-05, "epoch": 1.5658362989323842, "percentage": 26.19, "elapsed_time": "0:16:07", "remaining_time": "0:45:26", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 60, "total_steps": 210, "loss": 1.3789, "learning_rate": 8.985662536114613e-05, "epoch": 1.708185053380783, "percentage": 28.57, "elapsed_time": "0:17:30", "remaining_time": "0:43:45", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 65, "total_steps": 210, "loss": 1.3579, "learning_rate": 8.721280197423258e-05, "epoch": 1.8505338078291815, "percentage": 30.95, "elapsed_time": "0:18:58", "remaining_time": "0:42:19", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 70, "total_steps": 210, "loss": 1.3654, "learning_rate": 8.43120818934367e-05, "epoch": 1.99288256227758, "percentage": 33.33, "elapsed_time": "0:20:24", "remaining_time": "0:40:48", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 70, "total_steps": 210, "eval_loss": 1.4582531452178955, "epoch": 1.99288256227758, "percentage": 33.33, "elapsed_time": "0:20:27", "remaining_time": "0:40:54", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 75, "total_steps": 210, "loss": 1.1069, "learning_rate": 8.117449009293668e-05, "epoch": 2.135231316725979, "percentage": 35.71, "elapsed_time": "0:21:53", "remaining_time": "0:39:24", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 80, "total_steps": 210, "loss": 1.0865, "learning_rate": 7.782168677883206e-05, "epoch": 2.277580071174377, "percentage": 38.1, "elapsed_time": "0:23:18", "remaining_time": "0:37:51", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 85, "total_steps": 210, "loss": 1.1173, "learning_rate": 7.427681785900761e-05, "epoch": 2.419928825622776, "percentage": 40.48, "elapsed_time": "0:24:43", "remaining_time": "0:36:21", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 90, "total_steps": 210, "loss": 1.0808, "learning_rate": 7.056435515653059e-05, "epoch": 2.562277580071174, "percentage": 42.86, "elapsed_time": "0:26:10", "remaining_time": "0:34:53", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 95, "total_steps": 210, "loss": 1.0482, "learning_rate": 6.670992746965938e-05, "epoch": 2.704626334519573, "percentage": 45.24, "elapsed_time": "0:27:40", "remaining_time": "0:33:30", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 100, "total_steps": 210, "loss": 1.0735, "learning_rate": 6.274014364473274e-05, "epoch": 2.8469750889679717, "percentage": 47.62, "elapsed_time": "0:29:05", "remaining_time": "0:32:00", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 105, "total_steps": 210, "loss": 1.0796, "learning_rate": 5.868240888334653e-05, "epoch": 2.9893238434163703, "percentage": 50.0, "elapsed_time": "0:30:31", "remaining_time": "0:30:31", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 105, "total_steps": 210, "eval_loss": 1.5236175060272217, "epoch": 2.9893238434163703, "percentage": 50.0, "elapsed_time": "0:30:34", "remaining_time": "0:30:34", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 110, "total_steps": 210, "loss": 0.855, "learning_rate": 5.456473555193242e-05, "epoch": 3.131672597864769, "percentage": 52.38, "elapsed_time": "0:31:59", "remaining_time": "0:29:05", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 115, "total_steps": 210, "loss": 0.8003, "learning_rate": 5.041554979980486e-05, "epoch": 3.2740213523131674, "percentage": 54.76, "elapsed_time": "0:33:25", "remaining_time": "0:27:36", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 120, "total_steps": 210, "loss": 0.8118, "learning_rate": 4.626349532067879e-05, "epoch": 3.416370106761566, "percentage": 57.14, "elapsed_time": "0:34:51", "remaining_time": "0:26:08", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 125, "total_steps": 210, "loss": 0.7847, "learning_rate": 4.213723561238074e-05, "epoch": 3.5587188612099645, "percentage": 59.52, "elapsed_time": "0:36:21", "remaining_time": "0:24:43", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 130, "total_steps": 210, "loss": 0.7769, "learning_rate": 3.806525609984312e-05, "epoch": 3.701067615658363, "percentage": 61.9, "elapsed_time": "0:37:56", "remaining_time": "0:23:20", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 135, "total_steps": 210, "loss": 0.7938, "learning_rate": 3.4075667487415785e-05, "epoch": 3.8434163701067616, "percentage": 64.29, "elapsed_time": "0:39:21", "remaining_time": "0:21:51", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 140, "total_steps": 210, "loss": 0.7613, "learning_rate": 3.019601169804216e-05, "epoch": 3.98576512455516, "percentage": 66.67, "elapsed_time": "0:40:43", "remaining_time": "0:20:21", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 140, "total_steps": 210, "eval_loss": 1.7009425163269043, "epoch": 3.98576512455516, "percentage": 66.67, "elapsed_time": "0:40:46", "remaining_time": "0:20:23", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 145, "total_steps": 210, "loss": 0.623, "learning_rate": 2.645307173898901e-05, "epoch": 4.128113879003559, "percentage": 69.05, "elapsed_time": "0:42:09", "remaining_time": "0:18:53", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 150, "total_steps": 210, "loss": 0.5953, "learning_rate": 2.2872686806712035e-05, "epoch": 4.270462633451958, "percentage": 71.43, "elapsed_time": "0:43:37", "remaining_time": "0:17:26", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 155, "total_steps": 210, "loss": 0.5766, "learning_rate": 1.947957390727185e-05, "epoch": 4.412811387900356, "percentage": 73.81, "elapsed_time": "0:45:04", "remaining_time": "0:15:59", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 160, "total_steps": 210, "loss": 0.5776, "learning_rate": 1.629715722373423e-05, "epoch": 4.555160142348754, "percentage": 76.19, "elapsed_time": "0:46:32", "remaining_time": "0:14:32", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 165, "total_steps": 210, "loss": 0.5623, "learning_rate": 1.3347406408508695e-05, "epoch": 4.697508896797153, "percentage": 78.57, "elapsed_time": "0:47:58", "remaining_time": "0:13:04", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 170, "total_steps": 210, "loss": 0.5588, "learning_rate": 1.0650684916965559e-05, "epoch": 4.839857651245552, "percentage": 80.95, "elapsed_time": "0:49:19", "remaining_time": "0:11:36", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 175, "total_steps": 210, "loss": 0.5639, "learning_rate": 8.225609429353187e-06, "epoch": 4.98220640569395, "percentage": 83.33, "elapsed_time": "0:50:51", "remaining_time": "0:10:10", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 175, "total_steps": 210, "eval_loss": 1.8933742046356201, "epoch": 4.98220640569395, "percentage": 83.33, "elapsed_time": "0:50:55", "remaining_time": "0:10:11", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 180, "total_steps": 210, "loss": 0.4838, "learning_rate": 6.088921331488568e-06, "epoch": 5.124555160142349, "percentage": 85.71, "elapsed_time": "0:52:22", "remaining_time": "0:08:43", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 185, "total_steps": 210, "loss": 0.4729, "learning_rate": 4.255371141448272e-06, "epoch": 5.266903914590747, "percentage": 88.1, "elapsed_time": "0:53:48", "remaining_time": "0:07:16", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 190, "total_steps": 210, "loss": 0.4591, "learning_rate": 2.737616680113758e-06, "epoch": 5.409252669039146, "percentage": 90.48, "elapsed_time": "0:55:10", "remaining_time": "0:05:48", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 195, "total_steps": 210, "loss": 0.473, "learning_rate": 1.5461356885461075e-06, "epoch": 5.551601423487544, "percentage": 92.86, "elapsed_time": "0:56:43", "remaining_time": "0:04:21", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 200, "total_steps": 210, "loss": 0.4638, "learning_rate": 6.891534954310885e-07, "epoch": 5.693950177935943, "percentage": 95.24, "elapsed_time": "0:58:12", "remaining_time": "0:02:54", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 205, "total_steps": 210, "loss": 0.4604, "learning_rate": 1.725862339392259e-07, "epoch": 5.8362989323843415, "percentage": 97.62, "elapsed_time": "0:59:35", "remaining_time": "0:01:27", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 210, "total_steps": 210, "loss": 0.4648, "learning_rate": 0.0, "epoch": 5.9786476868327405, "percentage": 100.0, "elapsed_time": "1:01:03", "remaining_time": "0:00:00", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 210, "total_steps": 210, "eval_loss": 2.0316998958587646, "epoch": 5.9786476868327405, "percentage": 100.0, "elapsed_time": "1:01:06", "remaining_time": "0:00:00", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 210, "total_steps": 210, "epoch": 5.9786476868327405, "percentage": 100.0, "elapsed_time": "1:01:08", "remaining_time": "0:00:00", "throughput": "0.00", "total_tokens": 0}