trainer_log.jsonl

{"current_steps": 10, "total_steps": 1266, "loss": 0.8897, "lr": 5e-06, "epoch": 0.023661638568470866, "percentage": 0.79, "elapsed_time": "0:05:26", "remaining_time": "11:24:07"}
{"current_steps": 20, "total_steps": 1266, "loss": 0.7877, "lr": 5e-06, "epoch": 0.04732327713694173, "percentage": 1.58, "elapsed_time": "0:10:48", "remaining_time": "11:13:38"}
{"current_steps": 30, "total_steps": 1266, "loss": 0.7533, "lr": 5e-06, "epoch": 0.0709849157054126, "percentage": 2.37, "elapsed_time": "0:16:10", "remaining_time": "11:06:19"}
{"current_steps": 40, "total_steps": 1266, "loss": 0.7354, "lr": 5e-06, "epoch": 0.09464655427388347, "percentage": 3.16, "elapsed_time": "0:21:30", "remaining_time": "10:59:22"}
{"current_steps": 50, "total_steps": 1266, "loss": 0.726, "lr": 5e-06, "epoch": 0.11830819284235433, "percentage": 3.95, "elapsed_time": "0:26:52", "remaining_time": "10:53:27"}
{"current_steps": 60, "total_steps": 1266, "loss": 0.7152, "lr": 5e-06, "epoch": 0.1419698314108252, "percentage": 4.74, "elapsed_time": "0:32:13", "remaining_time": "10:47:45"}
{"current_steps": 70, "total_steps": 1266, "loss": 0.7101, "lr": 5e-06, "epoch": 0.16563146997929606, "percentage": 5.53, "elapsed_time": "0:37:35", "remaining_time": "10:42:15"}
{"current_steps": 80, "total_steps": 1266, "loss": 0.7007, "lr": 5e-06, "epoch": 0.18929310854776693, "percentage": 6.32, "elapsed_time": "0:42:55", "remaining_time": "10:36:19"}
{"current_steps": 90, "total_steps": 1266, "loss": 0.6856, "lr": 5e-06, "epoch": 0.2129547471162378, "percentage": 7.11, "elapsed_time": "0:48:17", "remaining_time": "10:30:55"}
{"current_steps": 100, "total_steps": 1266, "loss": 0.685, "lr": 5e-06, "epoch": 0.23661638568470866, "percentage": 7.9, "elapsed_time": "0:53:35", "remaining_time": "10:24:58"}
{"current_steps": 110, "total_steps": 1266, "loss": 0.6898, "lr": 5e-06, "epoch": 0.26027802425317953, "percentage": 8.69, "elapsed_time": "0:58:57", "remaining_time": "10:19:32"}
{"current_steps": 120, "total_steps": 1266, "loss": 0.6772, "lr": 5e-06, "epoch": 0.2839396628216504, "percentage": 9.48, "elapsed_time": "1:04:19", "remaining_time": "10:14:16"}
{"current_steps": 130, "total_steps": 1266, "loss": 0.6733, "lr": 5e-06, "epoch": 0.30760130139012126, "percentage": 10.27, "elapsed_time": "1:09:41", "remaining_time": "10:08:56"}
{"current_steps": 140, "total_steps": 1266, "loss": 0.6819, "lr": 5e-06, "epoch": 0.33126293995859213, "percentage": 11.06, "elapsed_time": "1:15:03", "remaining_time": "10:03:43"}
{"current_steps": 150, "total_steps": 1266, "loss": 0.6648, "lr": 5e-06, "epoch": 0.354924578527063, "percentage": 11.85, "elapsed_time": "1:20:26", "remaining_time": "9:58:26"}
{"current_steps": 160, "total_steps": 1266, "loss": 0.6691, "lr": 5e-06, "epoch": 0.37858621709553386, "percentage": 12.64, "elapsed_time": "1:25:46", "remaining_time": "9:52:58"}
{"current_steps": 170, "total_steps": 1266, "loss": 0.667, "lr": 5e-06, "epoch": 0.4022478556640047, "percentage": 13.43, "elapsed_time": "1:31:05", "remaining_time": "9:47:19"}
{"current_steps": 180, "total_steps": 1266, "loss": 0.6739, "lr": 5e-06, "epoch": 0.4259094942324756, "percentage": 14.22, "elapsed_time": "1:36:24", "remaining_time": "9:41:38"}
{"current_steps": 190, "total_steps": 1266, "loss": 0.6698, "lr": 5e-06, "epoch": 0.44957113280094646, "percentage": 15.01, "elapsed_time": "1:41:45", "remaining_time": "9:36:15"}
{"current_steps": 200, "total_steps": 1266, "loss": 0.6688, "lr": 5e-06, "epoch": 0.4732327713694173, "percentage": 15.8, "elapsed_time": "1:47:07", "remaining_time": "9:30:59"}
{"current_steps": 210, "total_steps": 1266, "loss": 0.6697, "lr": 5e-06, "epoch": 0.4968944099378882, "percentage": 16.59, "elapsed_time": "1:52:30", "remaining_time": "9:25:43"}
{"current_steps": 220, "total_steps": 1266, "loss": 0.6681, "lr": 5e-06, "epoch": 0.5205560485063591, "percentage": 17.38, "elapsed_time": "1:57:53", "remaining_time": "9:20:29"}
{"current_steps": 230, "total_steps": 1266, "loss": 0.6603, "lr": 5e-06, "epoch": 0.54421768707483, "percentage": 18.17, "elapsed_time": "2:03:14", "remaining_time": "9:15:05"}
{"current_steps": 240, "total_steps": 1266, "loss": 0.6645, "lr": 5e-06, "epoch": 0.5678793256433008, "percentage": 18.96, "elapsed_time": "2:08:35", "remaining_time": "9:09:45"}
{"current_steps": 250, "total_steps": 1266, "loss": 0.6616, "lr": 5e-06, "epoch": 0.5915409642117717, "percentage": 19.75, "elapsed_time": "2:13:56", "remaining_time": "9:04:20"}
{"current_steps": 260, "total_steps": 1266, "loss": 0.652, "lr": 5e-06, "epoch": 0.6152026027802425, "percentage": 20.54, "elapsed_time": "2:19:19", "remaining_time": "8:59:03"}
{"current_steps": 270, "total_steps": 1266, "loss": 0.6641, "lr": 5e-06, "epoch": 0.6388642413487134, "percentage": 21.33, "elapsed_time": "2:24:40", "remaining_time": "8:53:41"}
{"current_steps": 280, "total_steps": 1266, "loss": 0.6652, "lr": 5e-06, "epoch": 0.6625258799171843, "percentage": 22.12, "elapsed_time": "2:30:01", "remaining_time": "8:48:18"}
{"current_steps": 290, "total_steps": 1266, "loss": 0.6597, "lr": 5e-06, "epoch": 0.6861875184856552, "percentage": 22.91, "elapsed_time": "2:35:23", "remaining_time": "8:42:57"}
{"current_steps": 300, "total_steps": 1266, "loss": 0.6535, "lr": 5e-06, "epoch": 0.709849157054126, "percentage": 23.7, "elapsed_time": "2:40:43", "remaining_time": "8:37:31"}
{"current_steps": 310, "total_steps": 1266, "loss": 0.6619, "lr": 5e-06, "epoch": 0.7335107956225969, "percentage": 24.49, "elapsed_time": "2:46:04", "remaining_time": "8:32:09"}
{"current_steps": 320, "total_steps": 1266, "loss": 0.6556, "lr": 5e-06, "epoch": 0.7571724341910677, "percentage": 25.28, "elapsed_time": "2:51:25", "remaining_time": "8:26:47"}
{"current_steps": 330, "total_steps": 1266, "loss": 0.6541, "lr": 5e-06, "epoch": 0.7808340727595386, "percentage": 26.07, "elapsed_time": "2:56:48", "remaining_time": "8:21:28"}
{"current_steps": 340, "total_steps": 1266, "loss": 0.6538, "lr": 5e-06, "epoch": 0.8044957113280095, "percentage": 26.86, "elapsed_time": "3:02:10", "remaining_time": "8:16:09"}
{"current_steps": 350, "total_steps": 1266, "loss": 0.6602, "lr": 5e-06, "epoch": 0.8281573498964804, "percentage": 27.65, "elapsed_time": "3:07:31", "remaining_time": "8:10:47"}
{"current_steps": 360, "total_steps": 1266, "loss": 0.6538, "lr": 5e-06, "epoch": 0.8518189884649512, "percentage": 28.44, "elapsed_time": "3:12:54", "remaining_time": "8:05:28"}
{"current_steps": 370, "total_steps": 1266, "loss": 0.6567, "lr": 5e-06, "epoch": 0.8754806270334221, "percentage": 29.23, "elapsed_time": "3:18:15", "remaining_time": "8:00:07"}
{"current_steps": 380, "total_steps": 1266, "loss": 0.6478, "lr": 5e-06, "epoch": 0.8991422656018929, "percentage": 30.02, "elapsed_time": "3:23:35", "remaining_time": "7:54:42"}
{"current_steps": 390, "total_steps": 1266, "loss": 0.6553, "lr": 5e-06, "epoch": 0.9228039041703638, "percentage": 30.81, "elapsed_time": "3:28:58", "remaining_time": "7:49:22"}
{"current_steps": 400, "total_steps": 1266, "loss": 0.6466, "lr": 5e-06, "epoch": 0.9464655427388347, "percentage": 31.6, "elapsed_time": "3:34:17", "remaining_time": "7:43:57"}
{"current_steps": 410, "total_steps": 1266, "loss": 0.6491, "lr": 5e-06, "epoch": 0.9701271813073056, "percentage": 32.39, "elapsed_time": "3:39:39", "remaining_time": "7:38:35"}
{"current_steps": 420, "total_steps": 1266, "loss": 0.648, "lr": 5e-06, "epoch": 0.9937888198757764, "percentage": 33.18, "elapsed_time": "3:44:59", "remaining_time": "7:33:11"}
{"current_steps": 422, "total_steps": 1266, "eval_loss": 0.6503860950469971, "epoch": 0.9985211475894705, "percentage": 33.33, "elapsed_time": "3:51:15", "remaining_time": "7:42:30"}