OH_DCFT_V3_wo_evol_instruct_70k / trainer_log.jsonl
sedrickkeh's picture
Training in progress, epoch 0
d43f091 verified
{"current_steps": 10, "total_steps": 1173, "loss": 0.9117, "learning_rate": 5e-06, "epoch": 0.025559105431309903, "percentage": 0.85, "elapsed_time": "0:04:55", "remaining_time": "9:32:06"}
{"current_steps": 20, "total_steps": 1173, "loss": 0.803, "learning_rate": 5e-06, "epoch": 0.051118210862619806, "percentage": 1.71, "elapsed_time": "0:09:46", "remaining_time": "9:23:11"}
{"current_steps": 30, "total_steps": 1173, "loss": 0.761, "learning_rate": 5e-06, "epoch": 0.07667731629392971, "percentage": 2.56, "elapsed_time": "0:14:37", "remaining_time": "9:17:14"}
{"current_steps": 40, "total_steps": 1173, "loss": 0.7317, "learning_rate": 5e-06, "epoch": 0.10223642172523961, "percentage": 3.41, "elapsed_time": "0:19:28", "remaining_time": "9:11:37"}
{"current_steps": 50, "total_steps": 1173, "loss": 0.7193, "learning_rate": 5e-06, "epoch": 0.12779552715654952, "percentage": 4.26, "elapsed_time": "0:24:19", "remaining_time": "9:06:19"}
{"current_steps": 60, "total_steps": 1173, "loss": 0.7099, "learning_rate": 5e-06, "epoch": 0.15335463258785942, "percentage": 5.12, "elapsed_time": "0:29:10", "remaining_time": "9:01:19"}
{"current_steps": 70, "total_steps": 1173, "loss": 0.7024, "learning_rate": 5e-06, "epoch": 0.17891373801916932, "percentage": 5.97, "elapsed_time": "0:34:02", "remaining_time": "8:56:19"}
{"current_steps": 80, "total_steps": 1173, "loss": 0.6912, "learning_rate": 5e-06, "epoch": 0.20447284345047922, "percentage": 6.82, "elapsed_time": "0:38:53", "remaining_time": "8:51:20"}
{"current_steps": 90, "total_steps": 1173, "loss": 0.6887, "learning_rate": 5e-06, "epoch": 0.23003194888178913, "percentage": 7.67, "elapsed_time": "0:43:44", "remaining_time": "8:46:16"}
{"current_steps": 100, "total_steps": 1173, "loss": 0.685, "learning_rate": 5e-06, "epoch": 0.25559105431309903, "percentage": 8.53, "elapsed_time": "0:48:35", "remaining_time": "8:41:21"}
{"current_steps": 110, "total_steps": 1173, "loss": 0.6783, "learning_rate": 5e-06, "epoch": 0.28115015974440893, "percentage": 9.38, "elapsed_time": "0:53:25", "remaining_time": "8:36:18"}
{"current_steps": 120, "total_steps": 1173, "loss": 0.6778, "learning_rate": 5e-06, "epoch": 0.30670926517571884, "percentage": 10.23, "elapsed_time": "0:58:16", "remaining_time": "8:31:22"}
{"current_steps": 130, "total_steps": 1173, "loss": 0.6705, "learning_rate": 5e-06, "epoch": 0.33226837060702874, "percentage": 11.08, "elapsed_time": "1:03:07", "remaining_time": "8:26:28"}
{"current_steps": 140, "total_steps": 1173, "loss": 0.6636, "learning_rate": 5e-06, "epoch": 0.35782747603833864, "percentage": 11.94, "elapsed_time": "1:07:58", "remaining_time": "8:21:34"}
{"current_steps": 150, "total_steps": 1173, "loss": 0.6723, "learning_rate": 5e-06, "epoch": 0.38338658146964855, "percentage": 12.79, "elapsed_time": "1:12:49", "remaining_time": "8:16:40"}
{"current_steps": 160, "total_steps": 1173, "loss": 0.6709, "learning_rate": 5e-06, "epoch": 0.40894568690095845, "percentage": 13.64, "elapsed_time": "1:17:40", "remaining_time": "8:11:46"}
{"current_steps": 170, "total_steps": 1173, "loss": 0.668, "learning_rate": 5e-06, "epoch": 0.43450479233226835, "percentage": 14.49, "elapsed_time": "1:22:31", "remaining_time": "8:06:51"}
{"current_steps": 180, "total_steps": 1173, "loss": 0.6675, "learning_rate": 5e-06, "epoch": 0.46006389776357826, "percentage": 15.35, "elapsed_time": "1:27:21", "remaining_time": "8:01:57"}
{"current_steps": 190, "total_steps": 1173, "loss": 0.6565, "learning_rate": 5e-06, "epoch": 0.48562300319488816, "percentage": 16.2, "elapsed_time": "1:32:12", "remaining_time": "7:57:02"}
{"current_steps": 200, "total_steps": 1173, "loss": 0.6622, "learning_rate": 5e-06, "epoch": 0.5111821086261981, "percentage": 17.05, "elapsed_time": "1:37:02", "remaining_time": "7:52:08"}
{"current_steps": 210, "total_steps": 1173, "loss": 0.6567, "learning_rate": 5e-06, "epoch": 0.536741214057508, "percentage": 17.9, "elapsed_time": "1:41:54", "remaining_time": "7:47:17"}
{"current_steps": 220, "total_steps": 1173, "loss": 0.6564, "learning_rate": 5e-06, "epoch": 0.5623003194888179, "percentage": 18.76, "elapsed_time": "1:46:45", "remaining_time": "7:42:26"}
{"current_steps": 230, "total_steps": 1173, "loss": 0.6596, "learning_rate": 5e-06, "epoch": 0.5878594249201278, "percentage": 19.61, "elapsed_time": "1:51:35", "remaining_time": "7:37:32"}
{"current_steps": 240, "total_steps": 1173, "loss": 0.6568, "learning_rate": 5e-06, "epoch": 0.6134185303514377, "percentage": 20.46, "elapsed_time": "1:56:26", "remaining_time": "7:32:40"}
{"current_steps": 250, "total_steps": 1173, "loss": 0.6554, "learning_rate": 5e-06, "epoch": 0.6389776357827476, "percentage": 21.31, "elapsed_time": "2:01:17", "remaining_time": "7:27:49"}
{"current_steps": 260, "total_steps": 1173, "loss": 0.6579, "learning_rate": 5e-06, "epoch": 0.6645367412140575, "percentage": 22.17, "elapsed_time": "2:06:09", "remaining_time": "7:22:59"}
{"current_steps": 270, "total_steps": 1173, "loss": 0.6519, "learning_rate": 5e-06, "epoch": 0.6900958466453674, "percentage": 23.02, "elapsed_time": "2:10:59", "remaining_time": "7:18:05"}
{"current_steps": 280, "total_steps": 1173, "loss": 0.6514, "learning_rate": 5e-06, "epoch": 0.7156549520766773, "percentage": 23.87, "elapsed_time": "2:15:50", "remaining_time": "7:13:13"}
{"current_steps": 290, "total_steps": 1173, "loss": 0.6505, "learning_rate": 5e-06, "epoch": 0.7412140575079872, "percentage": 24.72, "elapsed_time": "2:20:41", "remaining_time": "7:08:22"}
{"current_steps": 300, "total_steps": 1173, "loss": 0.6517, "learning_rate": 5e-06, "epoch": 0.7667731629392971, "percentage": 25.58, "elapsed_time": "2:25:32", "remaining_time": "7:03:31"}
{"current_steps": 310, "total_steps": 1173, "loss": 0.6473, "learning_rate": 5e-06, "epoch": 0.792332268370607, "percentage": 26.43, "elapsed_time": "2:30:23", "remaining_time": "6:58:40"}
{"current_steps": 320, "total_steps": 1173, "loss": 0.6431, "learning_rate": 5e-06, "epoch": 0.8178913738019169, "percentage": 27.28, "elapsed_time": "2:35:13", "remaining_time": "6:53:47"}
{"current_steps": 330, "total_steps": 1173, "loss": 0.6428, "learning_rate": 5e-06, "epoch": 0.8434504792332268, "percentage": 28.13, "elapsed_time": "2:40:05", "remaining_time": "6:48:56"}
{"current_steps": 340, "total_steps": 1173, "loss": 0.6452, "learning_rate": 5e-06, "epoch": 0.8690095846645367, "percentage": 28.99, "elapsed_time": "2:44:56", "remaining_time": "6:44:05"}
{"current_steps": 350, "total_steps": 1173, "loss": 0.6409, "learning_rate": 5e-06, "epoch": 0.8945686900958466, "percentage": 29.84, "elapsed_time": "2:49:47", "remaining_time": "6:39:14"}
{"current_steps": 360, "total_steps": 1173, "loss": 0.6473, "learning_rate": 5e-06, "epoch": 0.9201277955271565, "percentage": 30.69, "elapsed_time": "2:54:38", "remaining_time": "6:34:24"}
{"current_steps": 370, "total_steps": 1173, "loss": 0.6472, "learning_rate": 5e-06, "epoch": 0.9456869009584664, "percentage": 31.54, "elapsed_time": "2:59:29", "remaining_time": "6:29:33"}
{"current_steps": 380, "total_steps": 1173, "loss": 0.6533, "learning_rate": 5e-06, "epoch": 0.9712460063897763, "percentage": 32.4, "elapsed_time": "3:04:20", "remaining_time": "6:24:41"}
{"current_steps": 390, "total_steps": 1173, "loss": 0.6456, "learning_rate": 5e-06, "epoch": 0.9968051118210862, "percentage": 33.25, "elapsed_time": "3:09:11", "remaining_time": "6:19:50"}
{"current_steps": 391, "total_steps": 1173, "eval_loss": 0.6414024829864502, "epoch": 0.9993610223642172, "percentage": 33.33, "elapsed_time": "3:13:16", "remaining_time": "6:26:33"}