OH_DCFT_V3_wo_airoboros / trainer_log.jsonl
sedrickkeh's picture
Training in progress, epoch 1
0991ea3 verified
{"current_steps": 10, "total_steps": 1221, "loss": 0.9151, "learning_rate": 5e-06, "epoch": 0.02457002457002457, "percentage": 0.82, "elapsed_time": "0:04:58", "remaining_time": "10:03:12"}
{"current_steps": 20, "total_steps": 1221, "loss": 0.8186, "learning_rate": 5e-06, "epoch": 0.04914004914004914, "percentage": 1.64, "elapsed_time": "0:09:53", "remaining_time": "9:53:40"}
{"current_steps": 30, "total_steps": 1221, "loss": 0.7773, "learning_rate": 5e-06, "epoch": 0.07371007371007371, "percentage": 2.46, "elapsed_time": "0:14:47", "remaining_time": "9:47:08"}
{"current_steps": 40, "total_steps": 1221, "loss": 0.752, "learning_rate": 5e-06, "epoch": 0.09828009828009827, "percentage": 3.28, "elapsed_time": "0:19:41", "remaining_time": "9:41:37"}
{"current_steps": 50, "total_steps": 1221, "loss": 0.7311, "learning_rate": 5e-06, "epoch": 0.12285012285012285, "percentage": 4.1, "elapsed_time": "0:24:36", "remaining_time": "9:36:20"}
{"current_steps": 60, "total_steps": 1221, "loss": 0.7249, "learning_rate": 5e-06, "epoch": 0.14742014742014742, "percentage": 4.91, "elapsed_time": "0:29:30", "remaining_time": "9:31:04"}
{"current_steps": 70, "total_steps": 1221, "loss": 0.7188, "learning_rate": 5e-06, "epoch": 0.171990171990172, "percentage": 5.73, "elapsed_time": "0:34:25", "remaining_time": "9:25:57"}
{"current_steps": 80, "total_steps": 1221, "loss": 0.7116, "learning_rate": 5e-06, "epoch": 0.19656019656019655, "percentage": 6.55, "elapsed_time": "0:39:19", "remaining_time": "9:20:57"}
{"current_steps": 90, "total_steps": 1221, "loss": 0.7015, "learning_rate": 5e-06, "epoch": 0.22113022113022113, "percentage": 7.37, "elapsed_time": "0:44:14", "remaining_time": "9:15:58"}
{"current_steps": 100, "total_steps": 1221, "loss": 0.6967, "learning_rate": 5e-06, "epoch": 0.2457002457002457, "percentage": 8.19, "elapsed_time": "0:49:09", "remaining_time": "9:11:00"}
{"current_steps": 110, "total_steps": 1221, "loss": 0.6902, "learning_rate": 5e-06, "epoch": 0.2702702702702703, "percentage": 9.01, "elapsed_time": "0:54:03", "remaining_time": "9:05:55"}
{"current_steps": 120, "total_steps": 1221, "loss": 0.6934, "learning_rate": 5e-06, "epoch": 0.29484029484029484, "percentage": 9.83, "elapsed_time": "0:58:57", "remaining_time": "9:00:58"}
{"current_steps": 130, "total_steps": 1221, "loss": 0.6765, "learning_rate": 5e-06, "epoch": 0.3194103194103194, "percentage": 10.65, "elapsed_time": "1:03:51", "remaining_time": "8:55:55"}
{"current_steps": 140, "total_steps": 1221, "loss": 0.6819, "learning_rate": 5e-06, "epoch": 0.343980343980344, "percentage": 11.47, "elapsed_time": "1:08:45", "remaining_time": "8:50:55"}
{"current_steps": 150, "total_steps": 1221, "loss": 0.6715, "learning_rate": 5e-06, "epoch": 0.36855036855036855, "percentage": 12.29, "elapsed_time": "1:13:40", "remaining_time": "8:46:00"}
{"current_steps": 160, "total_steps": 1221, "loss": 0.6793, "learning_rate": 5e-06, "epoch": 0.3931203931203931, "percentage": 13.1, "elapsed_time": "1:18:34", "remaining_time": "8:41:00"}
{"current_steps": 170, "total_steps": 1221, "loss": 0.6747, "learning_rate": 5e-06, "epoch": 0.4176904176904177, "percentage": 13.92, "elapsed_time": "1:23:28", "remaining_time": "8:36:05"}
{"current_steps": 180, "total_steps": 1221, "loss": 0.6814, "learning_rate": 5e-06, "epoch": 0.44226044226044225, "percentage": 14.74, "elapsed_time": "1:28:23", "remaining_time": "8:31:12"}
{"current_steps": 190, "total_steps": 1221, "loss": 0.6687, "learning_rate": 5e-06, "epoch": 0.4668304668304668, "percentage": 15.56, "elapsed_time": "1:33:18", "remaining_time": "8:26:16"}
{"current_steps": 200, "total_steps": 1221, "loss": 0.6757, "learning_rate": 5e-06, "epoch": 0.4914004914004914, "percentage": 16.38, "elapsed_time": "1:38:11", "remaining_time": "8:21:17"}
{"current_steps": 210, "total_steps": 1221, "loss": 0.6652, "learning_rate": 5e-06, "epoch": 0.515970515970516, "percentage": 17.2, "elapsed_time": "1:43:05", "remaining_time": "8:16:19"}
{"current_steps": 220, "total_steps": 1221, "loss": 0.661, "learning_rate": 5e-06, "epoch": 0.5405405405405406, "percentage": 18.02, "elapsed_time": "1:48:00", "remaining_time": "8:11:25"}
{"current_steps": 230, "total_steps": 1221, "loss": 0.6693, "learning_rate": 5e-06, "epoch": 0.5651105651105651, "percentage": 18.84, "elapsed_time": "1:52:54", "remaining_time": "8:06:31"}
{"current_steps": 240, "total_steps": 1221, "loss": 0.6627, "learning_rate": 5e-06, "epoch": 0.5896805896805897, "percentage": 19.66, "elapsed_time": "1:57:49", "remaining_time": "8:01:36"}
{"current_steps": 250, "total_steps": 1221, "loss": 0.6516, "learning_rate": 5e-06, "epoch": 0.6142506142506142, "percentage": 20.48, "elapsed_time": "2:02:43", "remaining_time": "7:56:40"}
{"current_steps": 260, "total_steps": 1221, "loss": 0.6652, "learning_rate": 5e-06, "epoch": 0.6388206388206388, "percentage": 21.29, "elapsed_time": "2:07:38", "remaining_time": "7:51:46"}
{"current_steps": 270, "total_steps": 1221, "loss": 0.6701, "learning_rate": 5e-06, "epoch": 0.6633906633906634, "percentage": 22.11, "elapsed_time": "2:12:32", "remaining_time": "7:46:49"}
{"current_steps": 280, "total_steps": 1221, "loss": 0.6617, "learning_rate": 5e-06, "epoch": 0.687960687960688, "percentage": 22.93, "elapsed_time": "2:17:26", "remaining_time": "7:41:55"}
{"current_steps": 290, "total_steps": 1221, "loss": 0.6577, "learning_rate": 5e-06, "epoch": 0.7125307125307125, "percentage": 23.75, "elapsed_time": "2:22:21", "remaining_time": "7:37:01"}
{"current_steps": 300, "total_steps": 1221, "loss": 0.6577, "learning_rate": 5e-06, "epoch": 0.7371007371007371, "percentage": 24.57, "elapsed_time": "2:27:15", "remaining_time": "7:32:05"}
{"current_steps": 310, "total_steps": 1221, "loss": 0.6624, "learning_rate": 5e-06, "epoch": 0.7616707616707616, "percentage": 25.39, "elapsed_time": "2:32:09", "remaining_time": "7:27:08"}
{"current_steps": 320, "total_steps": 1221, "loss": 0.6511, "learning_rate": 5e-06, "epoch": 0.7862407862407862, "percentage": 26.21, "elapsed_time": "2:37:03", "remaining_time": "7:22:12"}
{"current_steps": 330, "total_steps": 1221, "loss": 0.6553, "learning_rate": 5e-06, "epoch": 0.8108108108108109, "percentage": 27.03, "elapsed_time": "2:41:57", "remaining_time": "7:17:17"}
{"current_steps": 340, "total_steps": 1221, "loss": 0.6536, "learning_rate": 5e-06, "epoch": 0.8353808353808354, "percentage": 27.85, "elapsed_time": "2:46:52", "remaining_time": "7:12:23"}
{"current_steps": 350, "total_steps": 1221, "loss": 0.6546, "learning_rate": 5e-06, "epoch": 0.85995085995086, "percentage": 28.67, "elapsed_time": "2:51:47", "remaining_time": "7:07:29"}
{"current_steps": 360, "total_steps": 1221, "loss": 0.6549, "learning_rate": 5e-06, "epoch": 0.8845208845208845, "percentage": 29.48, "elapsed_time": "2:56:41", "remaining_time": "7:02:36"}
{"current_steps": 370, "total_steps": 1221, "loss": 0.6558, "learning_rate": 5e-06, "epoch": 0.9090909090909091, "percentage": 30.3, "elapsed_time": "3:01:36", "remaining_time": "6:57:42"}
{"current_steps": 380, "total_steps": 1221, "loss": 0.6524, "learning_rate": 5e-06, "epoch": 0.9336609336609336, "percentage": 31.12, "elapsed_time": "3:06:31", "remaining_time": "6:52:48"}
{"current_steps": 390, "total_steps": 1221, "loss": 0.6573, "learning_rate": 5e-06, "epoch": 0.9582309582309583, "percentage": 31.94, "elapsed_time": "3:11:25", "remaining_time": "6:47:53"}
{"current_steps": 400, "total_steps": 1221, "loss": 0.6513, "learning_rate": 5e-06, "epoch": 0.9828009828009828, "percentage": 32.76, "elapsed_time": "3:16:19", "remaining_time": "6:42:57"}
{"current_steps": 407, "total_steps": 1221, "eval_loss": 0.6469405889511108, "epoch": 1.0, "percentage": 33.33, "elapsed_time": "3:23:25", "remaining_time": "6:46:50"}
{"current_steps": 410, "total_steps": 1221, "loss": 0.6392, "learning_rate": 5e-06, "epoch": 1.0073710073710074, "percentage": 33.58, "elapsed_time": "3:25:50", "remaining_time": "6:47:10"}