hp_ablations_qwen_epoch5 / trainer_log.jsonl
sedrickkeh's picture
Training in progress, epoch 5
b4fe267 verified
raw
history blame
40.1 kB
{"current_steps": 10, "total_steps": 2195, "loss": 0.7557, "lr": 5e-06, "epoch": 0.022779043280182234, "percentage": 0.46, "elapsed_time": "0:02:19", "remaining_time": "8:27:03"}
{"current_steps": 20, "total_steps": 2195, "loss": 0.6991, "lr": 5e-06, "epoch": 0.04555808656036447, "percentage": 0.91, "elapsed_time": "0:04:33", "remaining_time": "8:16:36"}
{"current_steps": 30, "total_steps": 2195, "loss": 0.6807, "lr": 5e-06, "epoch": 0.0683371298405467, "percentage": 1.37, "elapsed_time": "0:06:48", "remaining_time": "8:11:55"}
{"current_steps": 40, "total_steps": 2195, "loss": 0.6795, "lr": 5e-06, "epoch": 0.09111617312072894, "percentage": 1.82, "elapsed_time": "0:09:04", "remaining_time": "8:08:36"}
{"current_steps": 50, "total_steps": 2195, "loss": 0.6751, "lr": 5e-06, "epoch": 0.11389521640091116, "percentage": 2.28, "elapsed_time": "0:11:19", "remaining_time": "8:05:41"}
{"current_steps": 60, "total_steps": 2195, "loss": 0.6551, "lr": 5e-06, "epoch": 0.1366742596810934, "percentage": 2.73, "elapsed_time": "0:13:34", "remaining_time": "8:03:01"}
{"current_steps": 70, "total_steps": 2195, "loss": 0.6519, "lr": 5e-06, "epoch": 0.15945330296127563, "percentage": 3.19, "elapsed_time": "0:15:49", "remaining_time": "8:00:25"}
{"current_steps": 80, "total_steps": 2195, "loss": 0.6552, "lr": 5e-06, "epoch": 0.18223234624145787, "percentage": 3.64, "elapsed_time": "0:18:04", "remaining_time": "7:57:55"}
{"current_steps": 90, "total_steps": 2195, "loss": 0.6394, "lr": 5e-06, "epoch": 0.20501138952164008, "percentage": 4.1, "elapsed_time": "0:20:19", "remaining_time": "7:55:31"}
{"current_steps": 100, "total_steps": 2195, "loss": 0.6413, "lr": 5e-06, "epoch": 0.22779043280182232, "percentage": 4.56, "elapsed_time": "0:22:35", "remaining_time": "7:53:07"}
{"current_steps": 110, "total_steps": 2195, "loss": 0.6325, "lr": 5e-06, "epoch": 0.2505694760820046, "percentage": 5.01, "elapsed_time": "0:24:50", "remaining_time": "7:50:46"}
{"current_steps": 120, "total_steps": 2195, "loss": 0.6411, "lr": 5e-06, "epoch": 0.2733485193621868, "percentage": 5.47, "elapsed_time": "0:27:05", "remaining_time": "7:48:26"}
{"current_steps": 130, "total_steps": 2195, "loss": 0.6422, "lr": 5e-06, "epoch": 0.296127562642369, "percentage": 5.92, "elapsed_time": "0:29:20", "remaining_time": "7:46:06"}
{"current_steps": 140, "total_steps": 2195, "loss": 0.6443, "lr": 5e-06, "epoch": 0.31890660592255127, "percentage": 6.38, "elapsed_time": "0:31:35", "remaining_time": "7:43:42"}
{"current_steps": 150, "total_steps": 2195, "loss": 0.6369, "lr": 5e-06, "epoch": 0.3416856492027335, "percentage": 6.83, "elapsed_time": "0:33:50", "remaining_time": "7:41:20"}
{"current_steps": 160, "total_steps": 2195, "loss": 0.6424, "lr": 5e-06, "epoch": 0.36446469248291574, "percentage": 7.29, "elapsed_time": "0:36:05", "remaining_time": "7:38:57"}
{"current_steps": 170, "total_steps": 2195, "loss": 0.6335, "lr": 5e-06, "epoch": 0.38724373576309795, "percentage": 7.74, "elapsed_time": "0:38:19", "remaining_time": "7:36:36"}
{"current_steps": 180, "total_steps": 2195, "loss": 0.6399, "lr": 5e-06, "epoch": 0.41002277904328016, "percentage": 8.2, "elapsed_time": "0:40:34", "remaining_time": "7:34:16"}
{"current_steps": 190, "total_steps": 2195, "loss": 0.6313, "lr": 5e-06, "epoch": 0.4328018223234624, "percentage": 8.66, "elapsed_time": "0:42:49", "remaining_time": "7:31:59"}
{"current_steps": 200, "total_steps": 2195, "loss": 0.6408, "lr": 5e-06, "epoch": 0.45558086560364464, "percentage": 9.11, "elapsed_time": "0:45:05", "remaining_time": "7:29:42"}
{"current_steps": 210, "total_steps": 2195, "loss": 0.6357, "lr": 5e-06, "epoch": 0.4783599088838269, "percentage": 9.57, "elapsed_time": "0:47:20", "remaining_time": "7:27:26"}
{"current_steps": 220, "total_steps": 2195, "loss": 0.6396, "lr": 5e-06, "epoch": 0.5011389521640092, "percentage": 10.02, "elapsed_time": "0:49:35", "remaining_time": "7:25:09"}
{"current_steps": 230, "total_steps": 2195, "loss": 0.6351, "lr": 5e-06, "epoch": 0.5239179954441914, "percentage": 10.48, "elapsed_time": "0:51:50", "remaining_time": "7:22:50"}
{"current_steps": 240, "total_steps": 2195, "loss": 0.6317, "lr": 5e-06, "epoch": 0.5466970387243736, "percentage": 10.93, "elapsed_time": "0:54:04", "remaining_time": "7:20:32"}
{"current_steps": 250, "total_steps": 2195, "loss": 0.6354, "lr": 5e-06, "epoch": 0.5694760820045558, "percentage": 11.39, "elapsed_time": "0:56:20", "remaining_time": "7:18:16"}
{"current_steps": 260, "total_steps": 2195, "loss": 0.6335, "lr": 5e-06, "epoch": 0.592255125284738, "percentage": 11.85, "elapsed_time": "0:58:35", "remaining_time": "7:16:00"}
{"current_steps": 270, "total_steps": 2195, "loss": 0.6313, "lr": 5e-06, "epoch": 0.6150341685649203, "percentage": 12.3, "elapsed_time": "1:00:50", "remaining_time": "7:13:44"}
{"current_steps": 280, "total_steps": 2195, "loss": 0.6429, "lr": 5e-06, "epoch": 0.6378132118451025, "percentage": 12.76, "elapsed_time": "1:03:05", "remaining_time": "7:11:28"}
{"current_steps": 290, "total_steps": 2195, "loss": 0.6306, "lr": 5e-06, "epoch": 0.6605922551252847, "percentage": 13.21, "elapsed_time": "1:05:20", "remaining_time": "7:09:12"}
{"current_steps": 300, "total_steps": 2195, "loss": 0.6288, "lr": 5e-06, "epoch": 0.683371298405467, "percentage": 13.67, "elapsed_time": "1:07:35", "remaining_time": "7:06:57"}
{"current_steps": 310, "total_steps": 2195, "loss": 0.6286, "lr": 5e-06, "epoch": 0.7061503416856492, "percentage": 14.12, "elapsed_time": "1:09:50", "remaining_time": "7:04:42"}
{"current_steps": 320, "total_steps": 2195, "loss": 0.6304, "lr": 5e-06, "epoch": 0.7289293849658315, "percentage": 14.58, "elapsed_time": "1:12:06", "remaining_time": "7:02:27"}
{"current_steps": 330, "total_steps": 2195, "loss": 0.6259, "lr": 5e-06, "epoch": 0.7517084282460137, "percentage": 15.03, "elapsed_time": "1:14:21", "remaining_time": "7:00:12"}
{"current_steps": 340, "total_steps": 2195, "loss": 0.6316, "lr": 5e-06, "epoch": 0.7744874715261959, "percentage": 15.49, "elapsed_time": "1:16:36", "remaining_time": "6:57:56"}
{"current_steps": 350, "total_steps": 2195, "loss": 0.6396, "lr": 5e-06, "epoch": 0.7972665148063781, "percentage": 15.95, "elapsed_time": "1:18:51", "remaining_time": "6:55:41"}
{"current_steps": 360, "total_steps": 2195, "loss": 0.6313, "lr": 5e-06, "epoch": 0.8200455580865603, "percentage": 16.4, "elapsed_time": "1:21:06", "remaining_time": "6:53:25"}
{"current_steps": 370, "total_steps": 2195, "loss": 0.6246, "lr": 5e-06, "epoch": 0.8428246013667426, "percentage": 16.86, "elapsed_time": "1:23:21", "remaining_time": "6:51:09"}
{"current_steps": 380, "total_steps": 2195, "loss": 0.6364, "lr": 5e-06, "epoch": 0.8656036446469249, "percentage": 17.31, "elapsed_time": "1:25:36", "remaining_time": "6:48:53"}
{"current_steps": 390, "total_steps": 2195, "loss": 0.6356, "lr": 5e-06, "epoch": 0.8883826879271071, "percentage": 17.77, "elapsed_time": "1:27:51", "remaining_time": "6:46:38"}
{"current_steps": 400, "total_steps": 2195, "loss": 0.6389, "lr": 5e-06, "epoch": 0.9111617312072893, "percentage": 18.22, "elapsed_time": "1:30:06", "remaining_time": "6:44:23"}
{"current_steps": 410, "total_steps": 2195, "loss": 0.6344, "lr": 5e-06, "epoch": 0.9339407744874715, "percentage": 18.68, "elapsed_time": "1:32:21", "remaining_time": "6:42:07"}
{"current_steps": 420, "total_steps": 2195, "loss": 0.6218, "lr": 5e-06, "epoch": 0.9567198177676538, "percentage": 19.13, "elapsed_time": "1:34:36", "remaining_time": "6:39:50"}
{"current_steps": 430, "total_steps": 2195, "loss": 0.6342, "lr": 5e-06, "epoch": 0.979498861047836, "percentage": 19.59, "elapsed_time": "1:36:51", "remaining_time": "6:37:35"}
{"current_steps": 439, "total_steps": 2195, "eval_loss": 0.6256291270256042, "epoch": 1.0, "percentage": 20.0, "elapsed_time": "1:40:45", "remaining_time": "6:43:03"}
{"current_steps": 440, "total_steps": 2195, "loss": 0.6154, "lr": 5e-06, "epoch": 1.0022779043280183, "percentage": 20.05, "elapsed_time": "1:41:45", "remaining_time": "6:45:52"}
{"current_steps": 450, "total_steps": 2195, "loss": 0.6008, "lr": 5e-06, "epoch": 1.0250569476082005, "percentage": 20.5, "elapsed_time": "1:44:00", "remaining_time": "6:43:20"}
{"current_steps": 460, "total_steps": 2195, "loss": 0.5988, "lr": 5e-06, "epoch": 1.0478359908883828, "percentage": 20.96, "elapsed_time": "1:46:16", "remaining_time": "6:40:48"}
{"current_steps": 470, "total_steps": 2195, "loss": 0.5963, "lr": 5e-06, "epoch": 1.070615034168565, "percentage": 21.41, "elapsed_time": "1:48:31", "remaining_time": "6:38:18"}
{"current_steps": 480, "total_steps": 2195, "loss": 0.6032, "lr": 5e-06, "epoch": 1.0933940774487472, "percentage": 21.87, "elapsed_time": "1:50:46", "remaining_time": "6:35:48"}
{"current_steps": 490, "total_steps": 2195, "loss": 0.6081, "lr": 5e-06, "epoch": 1.1161731207289294, "percentage": 22.32, "elapsed_time": "1:53:01", "remaining_time": "6:33:18"}
{"current_steps": 500, "total_steps": 2195, "loss": 0.5983, "lr": 5e-06, "epoch": 1.1389521640091116, "percentage": 22.78, "elapsed_time": "1:55:17", "remaining_time": "6:30:49"}
{"current_steps": 510, "total_steps": 2195, "loss": 0.5955, "lr": 5e-06, "epoch": 1.1617312072892938, "percentage": 23.23, "elapsed_time": "1:57:32", "remaining_time": "6:28:20"}
{"current_steps": 520, "total_steps": 2195, "loss": 0.6023, "lr": 5e-06, "epoch": 1.184510250569476, "percentage": 23.69, "elapsed_time": "1:59:47", "remaining_time": "6:25:52"}
{"current_steps": 530, "total_steps": 2195, "loss": 0.5979, "lr": 5e-06, "epoch": 1.2072892938496582, "percentage": 24.15, "elapsed_time": "2:02:02", "remaining_time": "6:23:24"}
{"current_steps": 540, "total_steps": 2195, "loss": 0.6031, "lr": 5e-06, "epoch": 1.2300683371298406, "percentage": 24.6, "elapsed_time": "2:04:17", "remaining_time": "6:20:56"}
{"current_steps": 550, "total_steps": 2195, "loss": 0.6057, "lr": 5e-06, "epoch": 1.2528473804100229, "percentage": 25.06, "elapsed_time": "2:06:32", "remaining_time": "6:18:29"}
{"current_steps": 560, "total_steps": 2195, "loss": 0.6016, "lr": 5e-06, "epoch": 1.275626423690205, "percentage": 25.51, "elapsed_time": "2:08:47", "remaining_time": "6:16:02"}
{"current_steps": 570, "total_steps": 2195, "loss": 0.5977, "lr": 5e-06, "epoch": 1.2984054669703873, "percentage": 25.97, "elapsed_time": "2:11:03", "remaining_time": "6:13:36"}
{"current_steps": 580, "total_steps": 2195, "loss": 0.599, "lr": 5e-06, "epoch": 1.3211845102505695, "percentage": 26.42, "elapsed_time": "2:13:18", "remaining_time": "6:11:10"}
{"current_steps": 590, "total_steps": 2195, "loss": 0.5936, "lr": 5e-06, "epoch": 1.3439635535307517, "percentage": 26.88, "elapsed_time": "2:15:33", "remaining_time": "6:08:44"}
{"current_steps": 600, "total_steps": 2195, "loss": 0.6035, "lr": 5e-06, "epoch": 1.366742596810934, "percentage": 27.33, "elapsed_time": "2:17:48", "remaining_time": "6:06:19"}
{"current_steps": 610, "total_steps": 2195, "loss": 0.5998, "lr": 5e-06, "epoch": 1.3895216400911161, "percentage": 27.79, "elapsed_time": "2:20:03", "remaining_time": "6:03:54"}
{"current_steps": 620, "total_steps": 2195, "loss": 0.6048, "lr": 5e-06, "epoch": 1.4123006833712983, "percentage": 28.25, "elapsed_time": "2:22:18", "remaining_time": "6:01:30"}
{"current_steps": 630, "total_steps": 2195, "loss": 0.604, "lr": 5e-06, "epoch": 1.4350797266514808, "percentage": 28.7, "elapsed_time": "2:24:33", "remaining_time": "5:59:06"}
{"current_steps": 640, "total_steps": 2195, "loss": 0.5943, "lr": 5e-06, "epoch": 1.4578587699316627, "percentage": 29.16, "elapsed_time": "2:26:48", "remaining_time": "5:56:43"}
{"current_steps": 650, "total_steps": 2195, "loss": 0.5969, "lr": 5e-06, "epoch": 1.4806378132118452, "percentage": 29.61, "elapsed_time": "2:29:04", "remaining_time": "5:54:19"}
{"current_steps": 660, "total_steps": 2195, "loss": 0.5934, "lr": 5e-06, "epoch": 1.5034168564920274, "percentage": 30.07, "elapsed_time": "2:31:19", "remaining_time": "5:51:56"}
{"current_steps": 670, "total_steps": 2195, "loss": 0.6031, "lr": 5e-06, "epoch": 1.5261958997722096, "percentage": 30.52, "elapsed_time": "2:33:34", "remaining_time": "5:49:34"}
{"current_steps": 680, "total_steps": 2195, "loss": 0.5933, "lr": 5e-06, "epoch": 1.5489749430523918, "percentage": 30.98, "elapsed_time": "2:35:50", "remaining_time": "5:47:11"}
{"current_steps": 690, "total_steps": 2195, "loss": 0.5927, "lr": 5e-06, "epoch": 1.571753986332574, "percentage": 31.44, "elapsed_time": "2:38:05", "remaining_time": "5:44:49"}
{"current_steps": 700, "total_steps": 2195, "loss": 0.6108, "lr": 5e-06, "epoch": 1.5945330296127562, "percentage": 31.89, "elapsed_time": "2:40:20", "remaining_time": "5:42:27"}
{"current_steps": 710, "total_steps": 2195, "loss": 0.5971, "lr": 5e-06, "epoch": 1.6173120728929384, "percentage": 32.35, "elapsed_time": "2:42:35", "remaining_time": "5:40:05"}
{"current_steps": 720, "total_steps": 2195, "loss": 0.599, "lr": 5e-06, "epoch": 1.6400911161731209, "percentage": 32.8, "elapsed_time": "2:44:51", "remaining_time": "5:37:43"}
{"current_steps": 730, "total_steps": 2195, "loss": 0.5988, "lr": 5e-06, "epoch": 1.6628701594533029, "percentage": 33.26, "elapsed_time": "2:47:06", "remaining_time": "5:35:21"}
{"current_steps": 740, "total_steps": 2195, "loss": 0.5942, "lr": 5e-06, "epoch": 1.6856492027334853, "percentage": 33.71, "elapsed_time": "2:49:21", "remaining_time": "5:33:00"}
{"current_steps": 750, "total_steps": 2195, "loss": 0.5976, "lr": 5e-06, "epoch": 1.7084282460136673, "percentage": 34.17, "elapsed_time": "2:51:36", "remaining_time": "5:30:38"}
{"current_steps": 760, "total_steps": 2195, "loss": 0.5948, "lr": 5e-06, "epoch": 1.7312072892938497, "percentage": 34.62, "elapsed_time": "2:53:51", "remaining_time": "5:28:17"}
{"current_steps": 770, "total_steps": 2195, "loss": 0.5949, "lr": 5e-06, "epoch": 1.753986332574032, "percentage": 35.08, "elapsed_time": "2:56:07", "remaining_time": "5:25:55"}
{"current_steps": 780, "total_steps": 2195, "loss": 0.6001, "lr": 5e-06, "epoch": 1.7767653758542141, "percentage": 35.54, "elapsed_time": "2:58:22", "remaining_time": "5:23:34"}
{"current_steps": 790, "total_steps": 2195, "loss": 0.5981, "lr": 5e-06, "epoch": 1.7995444191343963, "percentage": 35.99, "elapsed_time": "3:00:37", "remaining_time": "5:21:13"}
{"current_steps": 800, "total_steps": 2195, "loss": 0.5894, "lr": 5e-06, "epoch": 1.8223234624145785, "percentage": 36.45, "elapsed_time": "3:02:52", "remaining_time": "5:18:52"}
{"current_steps": 810, "total_steps": 2195, "loss": 0.59, "lr": 5e-06, "epoch": 1.845102505694761, "percentage": 36.9, "elapsed_time": "3:05:07", "remaining_time": "5:16:31"}
{"current_steps": 820, "total_steps": 2195, "loss": 0.5906, "lr": 5e-06, "epoch": 1.867881548974943, "percentage": 37.36, "elapsed_time": "3:07:22", "remaining_time": "5:14:11"}
{"current_steps": 830, "total_steps": 2195, "loss": 0.597, "lr": 5e-06, "epoch": 1.8906605922551254, "percentage": 37.81, "elapsed_time": "3:09:37", "remaining_time": "5:11:50"}
{"current_steps": 840, "total_steps": 2195, "loss": 0.5986, "lr": 5e-06, "epoch": 1.9134396355353074, "percentage": 38.27, "elapsed_time": "3:11:52", "remaining_time": "5:09:30"}
{"current_steps": 850, "total_steps": 2195, "loss": 0.5973, "lr": 5e-06, "epoch": 1.9362186788154898, "percentage": 38.72, "elapsed_time": "3:14:07", "remaining_time": "5:07:10"}
{"current_steps": 860, "total_steps": 2195, "loss": 0.5976, "lr": 5e-06, "epoch": 1.958997722095672, "percentage": 39.18, "elapsed_time": "3:16:22", "remaining_time": "5:04:50"}
{"current_steps": 870, "total_steps": 2195, "loss": 0.5978, "lr": 5e-06, "epoch": 1.9817767653758542, "percentage": 39.64, "elapsed_time": "3:18:37", "remaining_time": "5:02:30"}
{"current_steps": 878, "total_steps": 2195, "eval_loss": 0.6193838715553284, "epoch": 2.0, "percentage": 40.0, "elapsed_time": "3:22:17", "remaining_time": "5:03:26"}
{"current_steps": 880, "total_steps": 2195, "loss": 0.5872, "lr": 5e-06, "epoch": 2.0045558086560367, "percentage": 40.09, "elapsed_time": "3:23:34", "remaining_time": "5:04:12"}
{"current_steps": 890, "total_steps": 2195, "loss": 0.556, "lr": 5e-06, "epoch": 2.0273348519362187, "percentage": 40.55, "elapsed_time": "3:25:49", "remaining_time": "5:01:48"}
{"current_steps": 900, "total_steps": 2195, "loss": 0.5683, "lr": 5e-06, "epoch": 2.050113895216401, "percentage": 41.0, "elapsed_time": "3:28:04", "remaining_time": "4:59:24"}
{"current_steps": 910, "total_steps": 2195, "loss": 0.5671, "lr": 5e-06, "epoch": 2.072892938496583, "percentage": 41.46, "elapsed_time": "3:30:20", "remaining_time": "4:57:00"}
{"current_steps": 920, "total_steps": 2195, "loss": 0.5729, "lr": 5e-06, "epoch": 2.0956719817767655, "percentage": 41.91, "elapsed_time": "3:32:35", "remaining_time": "4:54:37"}
{"current_steps": 930, "total_steps": 2195, "loss": 0.5655, "lr": 5e-06, "epoch": 2.1184510250569475, "percentage": 42.37, "elapsed_time": "3:34:50", "remaining_time": "4:52:13"}
{"current_steps": 940, "total_steps": 2195, "loss": 0.5684, "lr": 5e-06, "epoch": 2.14123006833713, "percentage": 42.82, "elapsed_time": "3:37:05", "remaining_time": "4:49:50"}
{"current_steps": 950, "total_steps": 2195, "loss": 0.5635, "lr": 5e-06, "epoch": 2.164009111617312, "percentage": 43.28, "elapsed_time": "3:39:20", "remaining_time": "4:47:26"}
{"current_steps": 960, "total_steps": 2195, "loss": 0.5684, "lr": 5e-06, "epoch": 2.1867881548974943, "percentage": 43.74, "elapsed_time": "3:41:35", "remaining_time": "4:45:04"}
{"current_steps": 970, "total_steps": 2195, "loss": 0.5694, "lr": 5e-06, "epoch": 2.2095671981776768, "percentage": 44.19, "elapsed_time": "3:43:50", "remaining_time": "4:42:41"}
{"current_steps": 980, "total_steps": 2195, "loss": 0.5618, "lr": 5e-06, "epoch": 2.2323462414578588, "percentage": 44.65, "elapsed_time": "3:46:05", "remaining_time": "4:40:18"}
{"current_steps": 990, "total_steps": 2195, "loss": 0.5659, "lr": 5e-06, "epoch": 2.255125284738041, "percentage": 45.1, "elapsed_time": "3:48:20", "remaining_time": "4:37:56"}
{"current_steps": 1000, "total_steps": 2195, "loss": 0.5677, "lr": 5e-06, "epoch": 2.277904328018223, "percentage": 45.56, "elapsed_time": "3:50:36", "remaining_time": "4:35:34"}
{"current_steps": 1010, "total_steps": 2195, "loss": 0.5657, "lr": 5e-06, "epoch": 2.3006833712984056, "percentage": 46.01, "elapsed_time": "3:52:51", "remaining_time": "4:33:12"}
{"current_steps": 1020, "total_steps": 2195, "loss": 0.5695, "lr": 5e-06, "epoch": 2.3234624145785876, "percentage": 46.47, "elapsed_time": "3:55:06", "remaining_time": "4:30:50"}
{"current_steps": 1030, "total_steps": 2195, "loss": 0.57, "lr": 5e-06, "epoch": 2.34624145785877, "percentage": 46.92, "elapsed_time": "3:57:21", "remaining_time": "4:28:28"}
{"current_steps": 1040, "total_steps": 2195, "loss": 0.5678, "lr": 5e-06, "epoch": 2.369020501138952, "percentage": 47.38, "elapsed_time": "3:59:36", "remaining_time": "4:26:06"}
{"current_steps": 1050, "total_steps": 2195, "loss": 0.5655, "lr": 5e-06, "epoch": 2.3917995444191344, "percentage": 47.84, "elapsed_time": "4:01:51", "remaining_time": "4:23:44"}
{"current_steps": 1060, "total_steps": 2195, "loss": 0.5764, "lr": 5e-06, "epoch": 2.4145785876993164, "percentage": 48.29, "elapsed_time": "4:04:07", "remaining_time": "4:21:23"}
{"current_steps": 1070, "total_steps": 2195, "loss": 0.5791, "lr": 5e-06, "epoch": 2.437357630979499, "percentage": 48.75, "elapsed_time": "4:06:22", "remaining_time": "4:19:02"}
{"current_steps": 1080, "total_steps": 2195, "loss": 0.5743, "lr": 5e-06, "epoch": 2.4601366742596813, "percentage": 49.2, "elapsed_time": "4:08:37", "remaining_time": "4:16:40"}
{"current_steps": 1090, "total_steps": 2195, "loss": 0.5726, "lr": 5e-06, "epoch": 2.4829157175398633, "percentage": 49.66, "elapsed_time": "4:10:52", "remaining_time": "4:14:19"}
{"current_steps": 1100, "total_steps": 2195, "loss": 0.5779, "lr": 5e-06, "epoch": 2.5056947608200457, "percentage": 50.11, "elapsed_time": "4:13:07", "remaining_time": "4:11:58"}
{"current_steps": 1110, "total_steps": 2195, "loss": 0.5696, "lr": 5e-06, "epoch": 2.5284738041002277, "percentage": 50.57, "elapsed_time": "4:15:23", "remaining_time": "4:09:37"}
{"current_steps": 1120, "total_steps": 2195, "loss": 0.5708, "lr": 5e-06, "epoch": 2.55125284738041, "percentage": 51.03, "elapsed_time": "4:17:38", "remaining_time": "4:07:17"}
{"current_steps": 1130, "total_steps": 2195, "loss": 0.5687, "lr": 5e-06, "epoch": 2.574031890660592, "percentage": 51.48, "elapsed_time": "4:19:53", "remaining_time": "4:04:56"}
{"current_steps": 1140, "total_steps": 2195, "loss": 0.57, "lr": 5e-06, "epoch": 2.5968109339407746, "percentage": 51.94, "elapsed_time": "4:22:08", "remaining_time": "4:02:35"}
{"current_steps": 1150, "total_steps": 2195, "loss": 0.5687, "lr": 5e-06, "epoch": 2.619589977220957, "percentage": 52.39, "elapsed_time": "4:24:23", "remaining_time": "4:00:15"}
{"current_steps": 1160, "total_steps": 2195, "loss": 0.5683, "lr": 5e-06, "epoch": 2.642369020501139, "percentage": 52.85, "elapsed_time": "4:26:38", "remaining_time": "3:57:54"}
{"current_steps": 1170, "total_steps": 2195, "loss": 0.5756, "lr": 5e-06, "epoch": 2.665148063781321, "percentage": 53.3, "elapsed_time": "4:28:53", "remaining_time": "3:55:34"}
{"current_steps": 1180, "total_steps": 2195, "loss": 0.5688, "lr": 5e-06, "epoch": 2.6879271070615034, "percentage": 53.76, "elapsed_time": "4:31:09", "remaining_time": "3:53:14"}
{"current_steps": 1190, "total_steps": 2195, "loss": 0.5668, "lr": 5e-06, "epoch": 2.710706150341686, "percentage": 54.21, "elapsed_time": "4:33:24", "remaining_time": "3:50:54"}
{"current_steps": 1200, "total_steps": 2195, "loss": 0.5735, "lr": 5e-06, "epoch": 2.733485193621868, "percentage": 54.67, "elapsed_time": "4:35:39", "remaining_time": "3:48:34"}
{"current_steps": 1210, "total_steps": 2195, "loss": 0.5746, "lr": 5e-06, "epoch": 2.7562642369020502, "percentage": 55.13, "elapsed_time": "4:37:54", "remaining_time": "3:46:14"}
{"current_steps": 1220, "total_steps": 2195, "loss": 0.5645, "lr": 5e-06, "epoch": 2.7790432801822322, "percentage": 55.58, "elapsed_time": "4:40:10", "remaining_time": "3:43:54"}
{"current_steps": 1230, "total_steps": 2195, "loss": 0.5662, "lr": 5e-06, "epoch": 2.8018223234624147, "percentage": 56.04, "elapsed_time": "4:42:25", "remaining_time": "3:41:34"}
{"current_steps": 1240, "total_steps": 2195, "loss": 0.5776, "lr": 5e-06, "epoch": 2.8246013667425967, "percentage": 56.49, "elapsed_time": "4:44:40", "remaining_time": "3:39:14"}
{"current_steps": 1250, "total_steps": 2195, "loss": 0.5766, "lr": 5e-06, "epoch": 2.847380410022779, "percentage": 56.95, "elapsed_time": "4:46:55", "remaining_time": "3:36:55"}
{"current_steps": 1260, "total_steps": 2195, "loss": 0.5712, "lr": 5e-06, "epoch": 2.8701594533029615, "percentage": 57.4, "elapsed_time": "4:49:11", "remaining_time": "3:34:35"}
{"current_steps": 1270, "total_steps": 2195, "loss": 0.5739, "lr": 5e-06, "epoch": 2.8929384965831435, "percentage": 57.86, "elapsed_time": "4:51:26", "remaining_time": "3:32:16"}
{"current_steps": 1280, "total_steps": 2195, "loss": 0.5664, "lr": 5e-06, "epoch": 2.9157175398633255, "percentage": 58.31, "elapsed_time": "4:53:42", "remaining_time": "3:29:57"}
{"current_steps": 1290, "total_steps": 2195, "loss": 0.5603, "lr": 5e-06, "epoch": 2.938496583143508, "percentage": 58.77, "elapsed_time": "4:55:57", "remaining_time": "3:27:37"}
{"current_steps": 1300, "total_steps": 2195, "loss": 0.5668, "lr": 5e-06, "epoch": 2.9612756264236904, "percentage": 59.23, "elapsed_time": "4:58:12", "remaining_time": "3:25:18"}
{"current_steps": 1310, "total_steps": 2195, "loss": 0.5732, "lr": 5e-06, "epoch": 2.9840546697038723, "percentage": 59.68, "elapsed_time": "5:00:28", "remaining_time": "3:22:59"}
{"current_steps": 1317, "total_steps": 2195, "eval_loss": 0.6193576455116272, "epoch": 3.0, "percentage": 60.0, "elapsed_time": "5:03:55", "remaining_time": "3:22:36"}
{"current_steps": 1320, "total_steps": 2195, "loss": 0.5578, "lr": 5e-06, "epoch": 3.0068337129840548, "percentage": 60.14, "elapsed_time": "5:05:26", "remaining_time": "3:22:27"}
{"current_steps": 1330, "total_steps": 2195, "loss": 0.5379, "lr": 5e-06, "epoch": 3.0296127562642368, "percentage": 60.59, "elapsed_time": "5:07:41", "remaining_time": "3:20:06"}
{"current_steps": 1340, "total_steps": 2195, "loss": 0.5411, "lr": 5e-06, "epoch": 3.052391799544419, "percentage": 61.05, "elapsed_time": "5:09:56", "remaining_time": "3:17:45"}
{"current_steps": 1350, "total_steps": 2195, "loss": 0.5424, "lr": 5e-06, "epoch": 3.075170842824601, "percentage": 61.5, "elapsed_time": "5:12:11", "remaining_time": "3:15:24"}
{"current_steps": 1360, "total_steps": 2195, "loss": 0.5382, "lr": 5e-06, "epoch": 3.0979498861047836, "percentage": 61.96, "elapsed_time": "5:14:26", "remaining_time": "3:13:03"}
{"current_steps": 1370, "total_steps": 2195, "loss": 0.5396, "lr": 5e-06, "epoch": 3.120728929384966, "percentage": 62.41, "elapsed_time": "5:16:41", "remaining_time": "3:10:42"}
{"current_steps": 1380, "total_steps": 2195, "loss": 0.5379, "lr": 5e-06, "epoch": 3.143507972665148, "percentage": 62.87, "elapsed_time": "5:18:57", "remaining_time": "3:08:22"}
{"current_steps": 1390, "total_steps": 2195, "loss": 0.5378, "lr": 5e-06, "epoch": 3.1662870159453305, "percentage": 63.33, "elapsed_time": "5:21:12", "remaining_time": "3:06:01"}
{"current_steps": 1400, "total_steps": 2195, "loss": 0.5355, "lr": 5e-06, "epoch": 3.1890660592255125, "percentage": 63.78, "elapsed_time": "5:23:27", "remaining_time": "3:03:40"}
{"current_steps": 1410, "total_steps": 2195, "loss": 0.5392, "lr": 5e-06, "epoch": 3.211845102505695, "percentage": 64.24, "elapsed_time": "5:25:43", "remaining_time": "3:01:20"}
{"current_steps": 1420, "total_steps": 2195, "loss": 0.5453, "lr": 5e-06, "epoch": 3.234624145785877, "percentage": 64.69, "elapsed_time": "5:27:58", "remaining_time": "2:58:59"}
{"current_steps": 1430, "total_steps": 2195, "loss": 0.5462, "lr": 5e-06, "epoch": 3.2574031890660593, "percentage": 65.15, "elapsed_time": "5:30:13", "remaining_time": "2:56:39"}
{"current_steps": 1440, "total_steps": 2195, "loss": 0.5389, "lr": 5e-06, "epoch": 3.2801822323462413, "percentage": 65.6, "elapsed_time": "5:32:28", "remaining_time": "2:54:19"}
{"current_steps": 1450, "total_steps": 2195, "loss": 0.536, "lr": 5e-06, "epoch": 3.3029612756264237, "percentage": 66.06, "elapsed_time": "5:34:43", "remaining_time": "2:51:58"}
{"current_steps": 1460, "total_steps": 2195, "loss": 0.5386, "lr": 5e-06, "epoch": 3.3257403189066057, "percentage": 66.51, "elapsed_time": "5:36:58", "remaining_time": "2:49:38"}
{"current_steps": 1470, "total_steps": 2195, "loss": 0.5399, "lr": 5e-06, "epoch": 3.348519362186788, "percentage": 66.97, "elapsed_time": "5:39:14", "remaining_time": "2:47:18"}
{"current_steps": 1480, "total_steps": 2195, "loss": 0.5424, "lr": 5e-06, "epoch": 3.3712984054669706, "percentage": 67.43, "elapsed_time": "5:41:29", "remaining_time": "2:44:58"}
{"current_steps": 1490, "total_steps": 2195, "loss": 0.541, "lr": 5e-06, "epoch": 3.3940774487471526, "percentage": 67.88, "elapsed_time": "5:43:45", "remaining_time": "2:42:38"}
{"current_steps": 1500, "total_steps": 2195, "loss": 0.5383, "lr": 5e-06, "epoch": 3.416856492027335, "percentage": 68.34, "elapsed_time": "5:46:00", "remaining_time": "2:40:18"}
{"current_steps": 1510, "total_steps": 2195, "loss": 0.543, "lr": 5e-06, "epoch": 3.439635535307517, "percentage": 68.79, "elapsed_time": "5:48:15", "remaining_time": "2:37:59"}
{"current_steps": 1520, "total_steps": 2195, "loss": 0.5402, "lr": 5e-06, "epoch": 3.4624145785876994, "percentage": 69.25, "elapsed_time": "5:50:31", "remaining_time": "2:35:39"}
{"current_steps": 1530, "total_steps": 2195, "loss": 0.548, "lr": 5e-06, "epoch": 3.4851936218678814, "percentage": 69.7, "elapsed_time": "5:52:46", "remaining_time": "2:33:19"}
{"current_steps": 1540, "total_steps": 2195, "loss": 0.5481, "lr": 5e-06, "epoch": 3.507972665148064, "percentage": 70.16, "elapsed_time": "5:55:01", "remaining_time": "2:31:00"}
{"current_steps": 1550, "total_steps": 2195, "loss": 0.5436, "lr": 5e-06, "epoch": 3.5307517084282463, "percentage": 70.62, "elapsed_time": "5:57:16", "remaining_time": "2:28:40"}
{"current_steps": 1560, "total_steps": 2195, "loss": 0.5429, "lr": 5e-06, "epoch": 3.5535307517084282, "percentage": 71.07, "elapsed_time": "5:59:32", "remaining_time": "2:26:20"}
{"current_steps": 1570, "total_steps": 2195, "loss": 0.5458, "lr": 5e-06, "epoch": 3.5763097949886102, "percentage": 71.53, "elapsed_time": "6:01:47", "remaining_time": "2:24:01"}
{"current_steps": 1580, "total_steps": 2195, "loss": 0.546, "lr": 5e-06, "epoch": 3.5990888382687927, "percentage": 71.98, "elapsed_time": "6:04:02", "remaining_time": "2:21:42"}
{"current_steps": 1590, "total_steps": 2195, "loss": 0.5422, "lr": 5e-06, "epoch": 3.621867881548975, "percentage": 72.44, "elapsed_time": "6:06:17", "remaining_time": "2:19:22"}
{"current_steps": 1600, "total_steps": 2195, "loss": 0.5359, "lr": 5e-06, "epoch": 3.644646924829157, "percentage": 72.89, "elapsed_time": "6:08:33", "remaining_time": "2:17:03"}
{"current_steps": 1610, "total_steps": 2195, "loss": 0.5422, "lr": 5e-06, "epoch": 3.6674259681093395, "percentage": 73.35, "elapsed_time": "6:10:48", "remaining_time": "2:14:44"}
{"current_steps": 1620, "total_steps": 2195, "loss": 0.5427, "lr": 5e-06, "epoch": 3.6902050113895215, "percentage": 73.8, "elapsed_time": "6:13:03", "remaining_time": "2:12:24"}
{"current_steps": 1630, "total_steps": 2195, "loss": 0.5419, "lr": 5e-06, "epoch": 3.712984054669704, "percentage": 74.26, "elapsed_time": "6:15:19", "remaining_time": "2:10:05"}
{"current_steps": 1640, "total_steps": 2195, "loss": 0.5384, "lr": 5e-06, "epoch": 3.735763097949886, "percentage": 74.72, "elapsed_time": "6:17:34", "remaining_time": "2:07:46"}
{"current_steps": 1650, "total_steps": 2195, "loss": 0.552, "lr": 5e-06, "epoch": 3.7585421412300684, "percentage": 75.17, "elapsed_time": "6:19:49", "remaining_time": "2:05:27"}
{"current_steps": 1660, "total_steps": 2195, "loss": 0.5416, "lr": 5e-06, "epoch": 3.781321184510251, "percentage": 75.63, "elapsed_time": "6:22:05", "remaining_time": "2:03:08"}
{"current_steps": 1670, "total_steps": 2195, "loss": 0.5497, "lr": 5e-06, "epoch": 3.8041002277904328, "percentage": 76.08, "elapsed_time": "6:24:20", "remaining_time": "2:00:49"}
{"current_steps": 1680, "total_steps": 2195, "loss": 0.5429, "lr": 5e-06, "epoch": 3.8268792710706148, "percentage": 76.54, "elapsed_time": "6:26:35", "remaining_time": "1:58:30"}
{"current_steps": 1690, "total_steps": 2195, "loss": 0.5451, "lr": 5e-06, "epoch": 3.849658314350797, "percentage": 76.99, "elapsed_time": "6:28:51", "remaining_time": "1:56:11"}
{"current_steps": 1700, "total_steps": 2195, "loss": 0.5422, "lr": 5e-06, "epoch": 3.8724373576309796, "percentage": 77.45, "elapsed_time": "6:31:06", "remaining_time": "1:53:52"}
{"current_steps": 1710, "total_steps": 2195, "loss": 0.5463, "lr": 5e-06, "epoch": 3.8952164009111616, "percentage": 77.9, "elapsed_time": "6:33:21", "remaining_time": "1:51:34"}
{"current_steps": 1720, "total_steps": 2195, "loss": 0.5505, "lr": 5e-06, "epoch": 3.917995444191344, "percentage": 78.36, "elapsed_time": "6:35:36", "remaining_time": "1:49:15"}
{"current_steps": 1730, "total_steps": 2195, "loss": 0.5496, "lr": 5e-06, "epoch": 3.940774487471526, "percentage": 78.82, "elapsed_time": "6:37:51", "remaining_time": "1:46:56"}
{"current_steps": 1740, "total_steps": 2195, "loss": 0.5511, "lr": 5e-06, "epoch": 3.9635535307517085, "percentage": 79.27, "elapsed_time": "6:40:06", "remaining_time": "1:44:37"}
{"current_steps": 1750, "total_steps": 2195, "loss": 0.5461, "lr": 5e-06, "epoch": 3.9863325740318905, "percentage": 79.73, "elapsed_time": "6:42:22", "remaining_time": "1:42:18"}
{"current_steps": 1756, "total_steps": 2195, "eval_loss": 0.6239981055259705, "epoch": 4.0, "percentage": 80.0, "elapsed_time": "6:45:35", "remaining_time": "1:41:23"}
{"current_steps": 1760, "total_steps": 2195, "loss": 0.5355, "lr": 5e-06, "epoch": 4.009111617312073, "percentage": 80.18, "elapsed_time": "6:47:20", "remaining_time": "1:40:40"}
{"current_steps": 1770, "total_steps": 2195, "loss": 0.5112, "lr": 5e-06, "epoch": 4.031890660592255, "percentage": 80.64, "elapsed_time": "6:49:35", "remaining_time": "1:38:20"}
{"current_steps": 1780, "total_steps": 2195, "loss": 0.5155, "lr": 5e-06, "epoch": 4.054669703872437, "percentage": 81.09, "elapsed_time": "6:51:50", "remaining_time": "1:36:01"}
{"current_steps": 1790, "total_steps": 2195, "loss": 0.5107, "lr": 5e-06, "epoch": 4.077448747152619, "percentage": 81.55, "elapsed_time": "6:54:05", "remaining_time": "1:33:41"}
{"current_steps": 1800, "total_steps": 2195, "loss": 0.5168, "lr": 5e-06, "epoch": 4.100227790432802, "percentage": 82.0, "elapsed_time": "6:56:21", "remaining_time": "1:31:21"}
{"current_steps": 1810, "total_steps": 2195, "loss": 0.5126, "lr": 5e-06, "epoch": 4.123006833712984, "percentage": 82.46, "elapsed_time": "6:58:36", "remaining_time": "1:29:02"}
{"current_steps": 1820, "total_steps": 2195, "loss": 0.5134, "lr": 5e-06, "epoch": 4.145785876993166, "percentage": 82.92, "elapsed_time": "7:00:51", "remaining_time": "1:26:42"}
{"current_steps": 1830, "total_steps": 2195, "loss": 0.5093, "lr": 5e-06, "epoch": 4.168564920273348, "percentage": 83.37, "elapsed_time": "7:03:06", "remaining_time": "1:24:23"}
{"current_steps": 1840, "total_steps": 2195, "loss": 0.5174, "lr": 5e-06, "epoch": 4.191343963553531, "percentage": 83.83, "elapsed_time": "7:05:21", "remaining_time": "1:22:03"}
{"current_steps": 1850, "total_steps": 2195, "loss": 0.5172, "lr": 5e-06, "epoch": 4.214123006833713, "percentage": 84.28, "elapsed_time": "7:07:36", "remaining_time": "1:19:44"}
{"current_steps": 1860, "total_steps": 2195, "loss": 0.5121, "lr": 5e-06, "epoch": 4.236902050113895, "percentage": 84.74, "elapsed_time": "7:09:52", "remaining_time": "1:17:25"}
{"current_steps": 1870, "total_steps": 2195, "loss": 0.5101, "lr": 5e-06, "epoch": 4.259681093394078, "percentage": 85.19, "elapsed_time": "7:12:07", "remaining_time": "1:15:06"}
{"current_steps": 1880, "total_steps": 2195, "loss": 0.5144, "lr": 5e-06, "epoch": 4.28246013667426, "percentage": 85.65, "elapsed_time": "7:14:22", "remaining_time": "1:12:46"}
{"current_steps": 1890, "total_steps": 2195, "loss": 0.517, "lr": 5e-06, "epoch": 4.305239179954442, "percentage": 86.1, "elapsed_time": "7:16:38", "remaining_time": "1:10:27"}
{"current_steps": 1900, "total_steps": 2195, "loss": 0.5163, "lr": 5e-06, "epoch": 4.328018223234624, "percentage": 86.56, "elapsed_time": "7:18:53", "remaining_time": "1:08:08"}
{"current_steps": 1910, "total_steps": 2195, "loss": 0.517, "lr": 5e-06, "epoch": 4.350797266514807, "percentage": 87.02, "elapsed_time": "7:21:09", "remaining_time": "1:05:49"}
{"current_steps": 1920, "total_steps": 2195, "loss": 0.5148, "lr": 5e-06, "epoch": 4.373576309794989, "percentage": 87.47, "elapsed_time": "7:23:24", "remaining_time": "1:03:30"}
{"current_steps": 1930, "total_steps": 2195, "loss": 0.5167, "lr": 5e-06, "epoch": 4.396355353075171, "percentage": 87.93, "elapsed_time": "7:25:39", "remaining_time": "1:01:11"}
{"current_steps": 1940, "total_steps": 2195, "loss": 0.51, "lr": 5e-06, "epoch": 4.4191343963553535, "percentage": 88.38, "elapsed_time": "7:27:54", "remaining_time": "0:58:52"}
{"current_steps": 1950, "total_steps": 2195, "loss": 0.5172, "lr": 5e-06, "epoch": 4.4419134396355355, "percentage": 88.84, "elapsed_time": "7:30:10", "remaining_time": "0:56:33"}
{"current_steps": 1960, "total_steps": 2195, "loss": 0.5134, "lr": 5e-06, "epoch": 4.4646924829157175, "percentage": 89.29, "elapsed_time": "7:32:25", "remaining_time": "0:54:14"}
{"current_steps": 1970, "total_steps": 2195, "loss": 0.5123, "lr": 5e-06, "epoch": 4.4874715261958995, "percentage": 89.75, "elapsed_time": "7:34:40", "remaining_time": "0:51:55"}
{"current_steps": 1980, "total_steps": 2195, "loss": 0.5177, "lr": 5e-06, "epoch": 4.510250569476082, "percentage": 90.21, "elapsed_time": "7:36:56", "remaining_time": "0:49:37"}
{"current_steps": 1990, "total_steps": 2195, "loss": 0.5095, "lr": 5e-06, "epoch": 4.533029612756264, "percentage": 90.66, "elapsed_time": "7:39:11", "remaining_time": "0:47:18"}
{"current_steps": 2000, "total_steps": 2195, "loss": 0.5126, "lr": 5e-06, "epoch": 4.555808656036446, "percentage": 91.12, "elapsed_time": "7:41:26", "remaining_time": "0:44:59"}
{"current_steps": 2010, "total_steps": 2195, "loss": 0.516, "lr": 5e-06, "epoch": 4.578587699316628, "percentage": 91.57, "elapsed_time": "7:43:41", "remaining_time": "0:42:40"}
{"current_steps": 2020, "total_steps": 2195, "loss": 0.516, "lr": 5e-06, "epoch": 4.601366742596811, "percentage": 92.03, "elapsed_time": "7:45:57", "remaining_time": "0:40:22"}
{"current_steps": 2030, "total_steps": 2195, "loss": 0.5159, "lr": 5e-06, "epoch": 4.624145785876993, "percentage": 92.48, "elapsed_time": "7:48:12", "remaining_time": "0:38:03"}
{"current_steps": 2040, "total_steps": 2195, "loss": 0.5103, "lr": 5e-06, "epoch": 4.646924829157175, "percentage": 92.94, "elapsed_time": "7:50:27", "remaining_time": "0:35:44"}
{"current_steps": 2050, "total_steps": 2195, "loss": 0.526, "lr": 5e-06, "epoch": 4.669703872437358, "percentage": 93.39, "elapsed_time": "7:52:42", "remaining_time": "0:33:26"}
{"current_steps": 2060, "total_steps": 2195, "loss": 0.5205, "lr": 5e-06, "epoch": 4.69248291571754, "percentage": 93.85, "elapsed_time": "7:54:58", "remaining_time": "0:31:07"}
{"current_steps": 2070, "total_steps": 2195, "loss": 0.5128, "lr": 5e-06, "epoch": 4.715261958997722, "percentage": 94.31, "elapsed_time": "7:57:13", "remaining_time": "0:28:49"}
{"current_steps": 2080, "total_steps": 2195, "loss": 0.5195, "lr": 5e-06, "epoch": 4.738041002277904, "percentage": 94.76, "elapsed_time": "7:59:28", "remaining_time": "0:26:30"}
{"current_steps": 2090, "total_steps": 2195, "loss": 0.5192, "lr": 5e-06, "epoch": 4.760820045558087, "percentage": 95.22, "elapsed_time": "8:01:43", "remaining_time": "0:24:12"}
{"current_steps": 2100, "total_steps": 2195, "loss": 0.5155, "lr": 5e-06, "epoch": 4.783599088838269, "percentage": 95.67, "elapsed_time": "8:03:59", "remaining_time": "0:21:53"}
{"current_steps": 2110, "total_steps": 2195, "loss": 0.5157, "lr": 5e-06, "epoch": 4.806378132118451, "percentage": 96.13, "elapsed_time": "8:06:14", "remaining_time": "0:19:35"}
{"current_steps": 2120, "total_steps": 2195, "loss": 0.515, "lr": 5e-06, "epoch": 4.829157175398633, "percentage": 96.58, "elapsed_time": "8:08:29", "remaining_time": "0:17:16"}
{"current_steps": 2130, "total_steps": 2195, "loss": 0.52, "lr": 5e-06, "epoch": 4.851936218678816, "percentage": 97.04, "elapsed_time": "8:10:44", "remaining_time": "0:14:58"}
{"current_steps": 2140, "total_steps": 2195, "loss": 0.5214, "lr": 5e-06, "epoch": 4.874715261958998, "percentage": 97.49, "elapsed_time": "8:13:00", "remaining_time": "0:12:40"}
{"current_steps": 2150, "total_steps": 2195, "loss": 0.5141, "lr": 5e-06, "epoch": 4.89749430523918, "percentage": 97.95, "elapsed_time": "8:15:15", "remaining_time": "0:10:21"}
{"current_steps": 2160, "total_steps": 2195, "loss": 0.5191, "lr": 5e-06, "epoch": 4.920273348519363, "percentage": 98.41, "elapsed_time": "8:17:30", "remaining_time": "0:08:03"}
{"current_steps": 2170, "total_steps": 2195, "loss": 0.5204, "lr": 5e-06, "epoch": 4.943052391799545, "percentage": 98.86, "elapsed_time": "8:19:45", "remaining_time": "0:05:45"}
{"current_steps": 2180, "total_steps": 2195, "loss": 0.5214, "lr": 5e-06, "epoch": 4.965831435079727, "percentage": 99.32, "elapsed_time": "8:22:01", "remaining_time": "0:03:27"}
{"current_steps": 2190, "total_steps": 2195, "loss": 0.5179, "lr": 5e-06, "epoch": 4.988610478359909, "percentage": 99.77, "elapsed_time": "8:24:16", "remaining_time": "0:01:09"}
{"current_steps": 2195, "total_steps": 2195, "eval_loss": 0.6346195340156555, "epoch": 5.0, "percentage": 100.0, "elapsed_time": "8:28:05", "remaining_time": "0:00:00"}
{"current_steps": 2195, "total_steps": 2195, "epoch": 5.0, "percentage": 100.0, "elapsed_time": "8:28:48", "remaining_time": "0:00:00"}