hp_ablations_qwen_lr1e-6 / trainer_log.jsonl
sedrickkeh's picture
Training in progress, epoch 2
5cdccfe verified
{"current_steps": 10, "total_steps": 1314, "loss": 0.8037, "learning_rate": 1e-06, "epoch": 0.022792022792022793, "percentage": 0.76, "elapsed_time": "0:08:48", "remaining_time": "19:09:40"}
{"current_steps": 20, "total_steps": 1314, "loss": 0.7481, "learning_rate": 1e-06, "epoch": 0.045584045584045586, "percentage": 1.52, "elapsed_time": "0:17:34", "remaining_time": "18:57:10"}
{"current_steps": 30, "total_steps": 1314, "loss": 0.7182, "learning_rate": 1e-06, "epoch": 0.06837606837606838, "percentage": 2.28, "elapsed_time": "0:26:19", "remaining_time": "18:46:23"}
{"current_steps": 40, "total_steps": 1314, "loss": 0.7142, "learning_rate": 1e-06, "epoch": 0.09116809116809117, "percentage": 3.04, "elapsed_time": "0:35:03", "remaining_time": "18:36:40"}
{"current_steps": 50, "total_steps": 1314, "loss": 0.7104, "learning_rate": 1e-06, "epoch": 0.11396011396011396, "percentage": 3.81, "elapsed_time": "0:43:48", "remaining_time": "18:27:32"}
{"current_steps": 60, "total_steps": 1314, "loss": 0.6912, "learning_rate": 1e-06, "epoch": 0.13675213675213677, "percentage": 4.57, "elapsed_time": "0:52:33", "remaining_time": "18:18:25"}
{"current_steps": 70, "total_steps": 1314, "loss": 0.69, "learning_rate": 1e-06, "epoch": 0.15954415954415954, "percentage": 5.33, "elapsed_time": "1:01:17", "remaining_time": "18:09:22"}
{"current_steps": 80, "total_steps": 1314, "loss": 0.6939, "learning_rate": 1e-06, "epoch": 0.18233618233618235, "percentage": 6.09, "elapsed_time": "1:10:03", "remaining_time": "18:00:36"}
{"current_steps": 90, "total_steps": 1314, "loss": 0.6785, "learning_rate": 1e-06, "epoch": 0.20512820512820512, "percentage": 6.85, "elapsed_time": "1:18:47", "remaining_time": "17:51:35"}
{"current_steps": 100, "total_steps": 1314, "loss": 0.6787, "learning_rate": 1e-06, "epoch": 0.22792022792022792, "percentage": 7.61, "elapsed_time": "1:27:32", "remaining_time": "17:42:47"}
{"current_steps": 110, "total_steps": 1314, "loss": 0.6685, "learning_rate": 1e-06, "epoch": 0.25071225071225073, "percentage": 8.37, "elapsed_time": "1:36:17", "remaining_time": "17:33:55"}
{"current_steps": 120, "total_steps": 1314, "loss": 0.6757, "learning_rate": 1e-06, "epoch": 0.27350427350427353, "percentage": 9.13, "elapsed_time": "1:44:59", "remaining_time": "17:24:43"}
{"current_steps": 130, "total_steps": 1314, "loss": 0.6753, "learning_rate": 1e-06, "epoch": 0.2962962962962963, "percentage": 9.89, "elapsed_time": "1:53:44", "remaining_time": "17:15:59"}
{"current_steps": 140, "total_steps": 1314, "loss": 0.6767, "learning_rate": 1e-06, "epoch": 0.3190883190883191, "percentage": 10.65, "elapsed_time": "2:02:30", "remaining_time": "17:07:19"}
{"current_steps": 150, "total_steps": 1314, "loss": 0.6681, "learning_rate": 1e-06, "epoch": 0.3418803418803419, "percentage": 11.42, "elapsed_time": "2:11:13", "remaining_time": "16:58:22"}
{"current_steps": 160, "total_steps": 1314, "loss": 0.6715, "learning_rate": 1e-06, "epoch": 0.3646723646723647, "percentage": 12.18, "elapsed_time": "2:19:58", "remaining_time": "16:49:33"}
{"current_steps": 170, "total_steps": 1314, "loss": 0.6634, "learning_rate": 1e-06, "epoch": 0.38746438746438744, "percentage": 12.94, "elapsed_time": "2:28:43", "remaining_time": "16:40:52"}
{"current_steps": 180, "total_steps": 1314, "loss": 0.6677, "learning_rate": 1e-06, "epoch": 0.41025641025641024, "percentage": 13.7, "elapsed_time": "2:37:29", "remaining_time": "16:32:14"}
{"current_steps": 190, "total_steps": 1314, "loss": 0.6591, "learning_rate": 1e-06, "epoch": 0.43304843304843305, "percentage": 14.46, "elapsed_time": "2:46:15", "remaining_time": "16:23:30"}
{"current_steps": 200, "total_steps": 1314, "loss": 0.6666, "learning_rate": 1e-06, "epoch": 0.45584045584045585, "percentage": 15.22, "elapsed_time": "2:55:00", "remaining_time": "16:14:46"}
{"current_steps": 210, "total_steps": 1314, "loss": 0.6618, "learning_rate": 1e-06, "epoch": 0.47863247863247865, "percentage": 15.98, "elapsed_time": "3:03:46", "remaining_time": "16:06:07"}
{"current_steps": 220, "total_steps": 1314, "loss": 0.6645, "learning_rate": 1e-06, "epoch": 0.5014245014245015, "percentage": 16.74, "elapsed_time": "3:12:31", "remaining_time": "15:57:24"}
{"current_steps": 230, "total_steps": 1314, "loss": 0.6587, "learning_rate": 1e-06, "epoch": 0.5242165242165242, "percentage": 17.5, "elapsed_time": "3:21:16", "remaining_time": "15:48:37"}
{"current_steps": 240, "total_steps": 1314, "loss": 0.6538, "learning_rate": 1e-06, "epoch": 0.5470085470085471, "percentage": 18.26, "elapsed_time": "3:30:02", "remaining_time": "15:39:57"}
{"current_steps": 250, "total_steps": 1314, "loss": 0.6582, "learning_rate": 1e-06, "epoch": 0.5698005698005698, "percentage": 19.03, "elapsed_time": "3:38:47", "remaining_time": "15:31:11"}
{"current_steps": 260, "total_steps": 1314, "loss": 0.6558, "learning_rate": 1e-06, "epoch": 0.5925925925925926, "percentage": 19.79, "elapsed_time": "3:47:32", "remaining_time": "15:22:26"}
{"current_steps": 270, "total_steps": 1314, "loss": 0.6522, "learning_rate": 1e-06, "epoch": 0.6153846153846154, "percentage": 20.55, "elapsed_time": "3:56:16", "remaining_time": "15:13:37"}
{"current_steps": 280, "total_steps": 1314, "loss": 0.6626, "learning_rate": 1e-06, "epoch": 0.6381766381766382, "percentage": 21.31, "elapsed_time": "4:05:00", "remaining_time": "15:04:46"}
{"current_steps": 290, "total_steps": 1314, "loss": 0.6506, "learning_rate": 1e-06, "epoch": 0.6609686609686609, "percentage": 22.07, "elapsed_time": "4:13:44", "remaining_time": "14:55:57"}
{"current_steps": 300, "total_steps": 1314, "loss": 0.6482, "learning_rate": 1e-06, "epoch": 0.6837606837606838, "percentage": 22.83, "elapsed_time": "4:22:27", "remaining_time": "14:47:07"}
{"current_steps": 310, "total_steps": 1314, "loss": 0.6476, "learning_rate": 1e-06, "epoch": 0.7065527065527065, "percentage": 23.59, "elapsed_time": "4:31:11", "remaining_time": "14:38:20"}
{"current_steps": 320, "total_steps": 1314, "loss": 0.6478, "learning_rate": 1e-06, "epoch": 0.7293447293447294, "percentage": 24.35, "elapsed_time": "4:39:54", "remaining_time": "14:29:29"}
{"current_steps": 330, "total_steps": 1314, "loss": 0.6443, "learning_rate": 1e-06, "epoch": 0.7521367521367521, "percentage": 25.11, "elapsed_time": "4:48:39", "remaining_time": "14:20:45"}
{"current_steps": 340, "total_steps": 1314, "loss": 0.6499, "learning_rate": 1e-06, "epoch": 0.7749287749287749, "percentage": 25.88, "elapsed_time": "4:57:25", "remaining_time": "14:12:01"}
{"current_steps": 350, "total_steps": 1314, "loss": 0.6572, "learning_rate": 1e-06, "epoch": 0.7977207977207977, "percentage": 26.64, "elapsed_time": "5:06:10", "remaining_time": "14:03:18"}
{"current_steps": 360, "total_steps": 1314, "loss": 0.6488, "learning_rate": 1e-06, "epoch": 0.8205128205128205, "percentage": 27.4, "elapsed_time": "5:14:56", "remaining_time": "13:54:34"}
{"current_steps": 370, "total_steps": 1314, "loss": 0.6423, "learning_rate": 1e-06, "epoch": 0.8433048433048433, "percentage": 28.16, "elapsed_time": "5:23:40", "remaining_time": "13:45:49"}
{"current_steps": 380, "total_steps": 1314, "loss": 0.6544, "learning_rate": 1e-06, "epoch": 0.8660968660968661, "percentage": 28.92, "elapsed_time": "5:32:24", "remaining_time": "13:37:00"}
{"current_steps": 390, "total_steps": 1314, "loss": 0.6529, "learning_rate": 1e-06, "epoch": 0.8888888888888888, "percentage": 29.68, "elapsed_time": "5:41:07", "remaining_time": "13:28:13"}
{"current_steps": 400, "total_steps": 1314, "loss": 0.6564, "learning_rate": 1e-06, "epoch": 0.9116809116809117, "percentage": 30.44, "elapsed_time": "5:49:52", "remaining_time": "13:19:27"}
{"current_steps": 410, "total_steps": 1314, "loss": 0.6524, "learning_rate": 1e-06, "epoch": 0.9344729344729344, "percentage": 31.2, "elapsed_time": "5:58:36", "remaining_time": "13:10:40"}
{"current_steps": 420, "total_steps": 1314, "loss": 0.639, "learning_rate": 1e-06, "epoch": 0.9572649572649573, "percentage": 31.96, "elapsed_time": "6:07:20", "remaining_time": "13:01:54"}
{"current_steps": 430, "total_steps": 1314, "loss": 0.6524, "learning_rate": 1e-06, "epoch": 0.98005698005698, "percentage": 32.72, "elapsed_time": "6:16:03", "remaining_time": "12:53:07"}
{"current_steps": 438, "total_steps": 1314, "eval_loss": 0.6427425146102905, "epoch": 0.9982905982905983, "percentage": 33.33, "elapsed_time": "6:30:50", "remaining_time": "13:01:41"}
{"current_steps": 440, "total_steps": 1314, "loss": 0.6687, "learning_rate": 1e-06, "epoch": 1.0034188034188034, "percentage": 33.49, "elapsed_time": "6:33:12", "remaining_time": "13:01:03"}
{"current_steps": 450, "total_steps": 1314, "loss": 0.6408, "learning_rate": 1e-06, "epoch": 1.0262108262108263, "percentage": 34.25, "elapsed_time": "6:41:57", "remaining_time": "12:51:46"}
{"current_steps": 460, "total_steps": 1314, "loss": 0.6342, "learning_rate": 1e-06, "epoch": 1.049002849002849, "percentage": 35.01, "elapsed_time": "6:50:41", "remaining_time": "12:42:27"}
{"current_steps": 470, "total_steps": 1314, "loss": 0.6354, "learning_rate": 1e-06, "epoch": 1.0717948717948718, "percentage": 35.77, "elapsed_time": "6:59:26", "remaining_time": "12:33:12"}
{"current_steps": 480, "total_steps": 1314, "loss": 0.6424, "learning_rate": 1e-06, "epoch": 1.0945868945868946, "percentage": 36.53, "elapsed_time": "7:08:11", "remaining_time": "12:23:59"}
{"current_steps": 490, "total_steps": 1314, "loss": 0.6475, "learning_rate": 1e-06, "epoch": 1.1173789173789175, "percentage": 37.29, "elapsed_time": "7:16:55", "remaining_time": "12:14:45"}
{"current_steps": 500, "total_steps": 1314, "loss": 0.6368, "learning_rate": 1e-06, "epoch": 1.1401709401709401, "percentage": 38.05, "elapsed_time": "7:25:40", "remaining_time": "12:05:34"}
{"current_steps": 510, "total_steps": 1314, "loss": 0.6342, "learning_rate": 1e-06, "epoch": 1.162962962962963, "percentage": 38.81, "elapsed_time": "7:34:25", "remaining_time": "11:56:23"}
{"current_steps": 520, "total_steps": 1314, "loss": 0.6397, "learning_rate": 1e-06, "epoch": 1.1857549857549858, "percentage": 39.57, "elapsed_time": "7:43:10", "remaining_time": "11:47:14"}
{"current_steps": 530, "total_steps": 1314, "loss": 0.6321, "learning_rate": 1e-06, "epoch": 1.2085470085470085, "percentage": 40.33, "elapsed_time": "7:51:56", "remaining_time": "11:38:06"}
{"current_steps": 540, "total_steps": 1314, "loss": 0.6426, "learning_rate": 1e-06, "epoch": 1.2313390313390313, "percentage": 41.1, "elapsed_time": "8:00:41", "remaining_time": "11:28:59"}
{"current_steps": 550, "total_steps": 1314, "loss": 0.6389, "learning_rate": 1e-06, "epoch": 1.2541310541310542, "percentage": 41.86, "elapsed_time": "8:09:26", "remaining_time": "11:19:52"}
{"current_steps": 560, "total_steps": 1314, "loss": 0.6402, "learning_rate": 1e-06, "epoch": 1.2769230769230768, "percentage": 42.62, "elapsed_time": "8:18:08", "remaining_time": "11:10:43"}
{"current_steps": 570, "total_steps": 1314, "loss": 0.6336, "learning_rate": 1e-06, "epoch": 1.2997150997150997, "percentage": 43.38, "elapsed_time": "8:26:53", "remaining_time": "11:01:37"}
{"current_steps": 580, "total_steps": 1314, "loss": 0.6324, "learning_rate": 1e-06, "epoch": 1.3225071225071225, "percentage": 44.14, "elapsed_time": "8:35:37", "remaining_time": "10:52:32"}
{"current_steps": 590, "total_steps": 1314, "loss": 0.632, "learning_rate": 1e-06, "epoch": 1.3452991452991454, "percentage": 44.9, "elapsed_time": "8:44:21", "remaining_time": "10:43:27"}
{"current_steps": 600, "total_steps": 1314, "loss": 0.6375, "learning_rate": 1e-06, "epoch": 1.368091168091168, "percentage": 45.66, "elapsed_time": "8:53:05", "remaining_time": "10:34:22"}
{"current_steps": 610, "total_steps": 1314, "loss": 0.6375, "learning_rate": 1e-06, "epoch": 1.390883190883191, "percentage": 46.42, "elapsed_time": "9:01:50", "remaining_time": "10:25:20"}
{"current_steps": 620, "total_steps": 1314, "loss": 0.6397, "learning_rate": 1e-06, "epoch": 1.4136752136752135, "percentage": 47.18, "elapsed_time": "9:10:35", "remaining_time": "10:16:18"}
{"current_steps": 630, "total_steps": 1314, "loss": 0.6376, "learning_rate": 1e-06, "epoch": 1.4364672364672364, "percentage": 47.95, "elapsed_time": "9:19:20", "remaining_time": "10:07:16"}
{"current_steps": 640, "total_steps": 1314, "loss": 0.6325, "learning_rate": 1e-06, "epoch": 1.4592592592592593, "percentage": 48.71, "elapsed_time": "9:28:02", "remaining_time": "9:58:12"}
{"current_steps": 650, "total_steps": 1314, "loss": 0.6301, "learning_rate": 1e-06, "epoch": 1.4820512820512821, "percentage": 49.47, "elapsed_time": "9:36:44", "remaining_time": "9:49:10"}
{"current_steps": 660, "total_steps": 1314, "loss": 0.6296, "learning_rate": 1e-06, "epoch": 1.504843304843305, "percentage": 50.23, "elapsed_time": "9:45:29", "remaining_time": "9:40:10"}
{"current_steps": 670, "total_steps": 1314, "loss": 0.6377, "learning_rate": 1e-06, "epoch": 1.5276353276353276, "percentage": 50.99, "elapsed_time": "9:54:15", "remaining_time": "9:31:11"}
{"current_steps": 680, "total_steps": 1314, "loss": 0.6256, "learning_rate": 1e-06, "epoch": 1.5504273504273505, "percentage": 51.75, "elapsed_time": "10:02:58", "remaining_time": "9:22:11"}
{"current_steps": 690, "total_steps": 1314, "loss": 0.6298, "learning_rate": 1e-06, "epoch": 1.573219373219373, "percentage": 52.51, "elapsed_time": "10:11:43", "remaining_time": "9:13:13"}
{"current_steps": 700, "total_steps": 1314, "loss": 0.6453, "learning_rate": 1e-06, "epoch": 1.596011396011396, "percentage": 53.27, "elapsed_time": "10:20:28", "remaining_time": "9:04:14"}
{"current_steps": 710, "total_steps": 1314, "loss": 0.632, "learning_rate": 1e-06, "epoch": 1.6188034188034188, "percentage": 54.03, "elapsed_time": "10:29:13", "remaining_time": "8:55:16"}
{"current_steps": 720, "total_steps": 1314, "loss": 0.6368, "learning_rate": 1e-06, "epoch": 1.6415954415954417, "percentage": 54.79, "elapsed_time": "10:37:58", "remaining_time": "8:46:19"}
{"current_steps": 730, "total_steps": 1314, "loss": 0.6294, "learning_rate": 1e-06, "epoch": 1.6643874643874645, "percentage": 55.56, "elapsed_time": "10:46:43", "remaining_time": "8:37:22"}
{"current_steps": 740, "total_steps": 1314, "loss": 0.6305, "learning_rate": 1e-06, "epoch": 1.6871794871794872, "percentage": 56.32, "elapsed_time": "10:55:26", "remaining_time": "8:28:24"}
{"current_steps": 750, "total_steps": 1314, "loss": 0.6275, "learning_rate": 1e-06, "epoch": 1.7099715099715098, "percentage": 57.08, "elapsed_time": "11:04:11", "remaining_time": "8:19:28"}
{"current_steps": 760, "total_steps": 1314, "loss": 0.632, "learning_rate": 1e-06, "epoch": 1.7327635327635327, "percentage": 57.84, "elapsed_time": "11:12:57", "remaining_time": "8:10:32"}
{"current_steps": 770, "total_steps": 1314, "loss": 0.628, "learning_rate": 1e-06, "epoch": 1.7555555555555555, "percentage": 58.6, "elapsed_time": "11:21:42", "remaining_time": "8:01:37"}
{"current_steps": 780, "total_steps": 1314, "loss": 0.6352, "learning_rate": 1e-06, "epoch": 1.7783475783475784, "percentage": 59.36, "elapsed_time": "11:30:26", "remaining_time": "7:52:41"}
{"current_steps": 790, "total_steps": 1314, "loss": 0.6323, "learning_rate": 1e-06, "epoch": 1.8011396011396013, "percentage": 60.12, "elapsed_time": "11:39:09", "remaining_time": "7:43:45"}
{"current_steps": 800, "total_steps": 1314, "loss": 0.6187, "learning_rate": 1e-06, "epoch": 1.823931623931624, "percentage": 60.88, "elapsed_time": "11:47:54", "remaining_time": "7:34:49"}
{"current_steps": 810, "total_steps": 1314, "loss": 0.6268, "learning_rate": 1e-06, "epoch": 1.8467236467236468, "percentage": 61.64, "elapsed_time": "11:56:39", "remaining_time": "7:25:55"}
{"current_steps": 820, "total_steps": 1314, "loss": 0.6245, "learning_rate": 1e-06, "epoch": 1.8695156695156694, "percentage": 62.4, "elapsed_time": "12:05:24", "remaining_time": "7:17:00"}
{"current_steps": 830, "total_steps": 1314, "loss": 0.6286, "learning_rate": 1e-06, "epoch": 1.8923076923076922, "percentage": 63.17, "elapsed_time": "12:14:10", "remaining_time": "7:08:07"}
{"current_steps": 840, "total_steps": 1314, "loss": 0.633, "learning_rate": 1e-06, "epoch": 1.915099715099715, "percentage": 63.93, "elapsed_time": "12:22:56", "remaining_time": "6:59:13"}
{"current_steps": 850, "total_steps": 1314, "loss": 0.6308, "learning_rate": 1e-06, "epoch": 1.937891737891738, "percentage": 64.69, "elapsed_time": "12:31:40", "remaining_time": "6:50:19"}
{"current_steps": 860, "total_steps": 1314, "loss": 0.6322, "learning_rate": 1e-06, "epoch": 1.9606837606837608, "percentage": 65.45, "elapsed_time": "12:40:25", "remaining_time": "6:41:25"}
{"current_steps": 870, "total_steps": 1314, "loss": 0.6297, "learning_rate": 1e-06, "epoch": 1.9834757834757835, "percentage": 66.21, "elapsed_time": "12:49:09", "remaining_time": "6:32:32"}
{"current_steps": 877, "total_steps": 1314, "eval_loss": 0.6338208317756653, "epoch": 1.9994301994301993, "percentage": 66.74, "elapsed_time": "13:02:36", "remaining_time": "6:29:57"}
{"current_steps": 880, "total_steps": 1314, "loss": 0.6567, "learning_rate": 1e-06, "epoch": 2.006837606837607, "percentage": 66.97, "elapsed_time": "13:06:22", "remaining_time": "6:27:49"}
{"current_steps": 890, "total_steps": 1314, "loss": 0.6093, "learning_rate": 1e-06, "epoch": 2.0296296296296297, "percentage": 67.73, "elapsed_time": "13:15:05", "remaining_time": "6:18:47"}
{"current_steps": 900, "total_steps": 1314, "loss": 0.6259, "learning_rate": 1e-06, "epoch": 2.0524216524216525, "percentage": 68.49, "elapsed_time": "13:23:48", "remaining_time": "6:09:44"}
{"current_steps": 910, "total_steps": 1314, "loss": 0.6235, "learning_rate": 1e-06, "epoch": 2.0752136752136754, "percentage": 69.25, "elapsed_time": "13:32:33", "remaining_time": "6:00:44"}
{"current_steps": 920, "total_steps": 1314, "loss": 0.6269, "learning_rate": 1e-06, "epoch": 2.098005698005698, "percentage": 70.02, "elapsed_time": "13:41:17", "remaining_time": "5:51:43"}
{"current_steps": 930, "total_steps": 1314, "loss": 0.6239, "learning_rate": 1e-06, "epoch": 2.1207977207977207, "percentage": 70.78, "elapsed_time": "13:50:02", "remaining_time": "5:42:43"}
{"current_steps": 940, "total_steps": 1314, "loss": 0.6184, "learning_rate": 1e-06, "epoch": 2.1435897435897435, "percentage": 71.54, "elapsed_time": "13:58:47", "remaining_time": "5:33:44"}
{"current_steps": 950, "total_steps": 1314, "loss": 0.6196, "learning_rate": 1e-06, "epoch": 2.1663817663817664, "percentage": 72.3, "elapsed_time": "14:07:31", "remaining_time": "5:24:44"}
{"current_steps": 960, "total_steps": 1314, "loss": 0.6245, "learning_rate": 1e-06, "epoch": 2.1891737891737892, "percentage": 73.06, "elapsed_time": "14:16:14", "remaining_time": "5:15:44"}
{"current_steps": 970, "total_steps": 1314, "loss": 0.6233, "learning_rate": 1e-06, "epoch": 2.211965811965812, "percentage": 73.82, "elapsed_time": "14:24:58", "remaining_time": "5:06:45"}
{"current_steps": 980, "total_steps": 1314, "loss": 0.6119, "learning_rate": 1e-06, "epoch": 2.234757834757835, "percentage": 74.58, "elapsed_time": "14:33:43", "remaining_time": "4:57:46"}
{"current_steps": 990, "total_steps": 1314, "loss": 0.6209, "learning_rate": 1e-06, "epoch": 2.2575498575498574, "percentage": 75.34, "elapsed_time": "14:42:26", "remaining_time": "4:48:47"}
{"current_steps": 1000, "total_steps": 1314, "loss": 0.6211, "learning_rate": 1e-06, "epoch": 2.2803418803418802, "percentage": 76.1, "elapsed_time": "14:51:11", "remaining_time": "4:39:49"}
{"current_steps": 1010, "total_steps": 1314, "loss": 0.6183, "learning_rate": 1e-06, "epoch": 2.303133903133903, "percentage": 76.86, "elapsed_time": "14:59:55", "remaining_time": "4:30:52"}
{"current_steps": 1020, "total_steps": 1314, "loss": 0.6216, "learning_rate": 1e-06, "epoch": 2.325925925925926, "percentage": 77.63, "elapsed_time": "15:08:40", "remaining_time": "4:21:54"}
{"current_steps": 1030, "total_steps": 1314, "loss": 0.6233, "learning_rate": 1e-06, "epoch": 2.348717948717949, "percentage": 78.39, "elapsed_time": "15:17:23", "remaining_time": "4:12:56"}
{"current_steps": 1040, "total_steps": 1314, "loss": 0.6201, "learning_rate": 1e-06, "epoch": 2.3715099715099717, "percentage": 79.15, "elapsed_time": "15:26:07", "remaining_time": "4:03:59"}
{"current_steps": 1050, "total_steps": 1314, "loss": 0.6195, "learning_rate": 1e-06, "epoch": 2.394301994301994, "percentage": 79.91, "elapsed_time": "15:34:52", "remaining_time": "3:55:03"}
{"current_steps": 1060, "total_steps": 1314, "loss": 0.6284, "learning_rate": 1e-06, "epoch": 2.417094017094017, "percentage": 80.67, "elapsed_time": "15:43:36", "remaining_time": "3:46:06"}
{"current_steps": 1070, "total_steps": 1314, "loss": 0.6316, "learning_rate": 1e-06, "epoch": 2.43988603988604, "percentage": 81.43, "elapsed_time": "15:52:21", "remaining_time": "3:37:10"}
{"current_steps": 1080, "total_steps": 1314, "loss": 0.6264, "learning_rate": 1e-06, "epoch": 2.4626780626780627, "percentage": 82.19, "elapsed_time": "16:01:04", "remaining_time": "3:28:13"}
{"current_steps": 1090, "total_steps": 1314, "loss": 0.6233, "learning_rate": 1e-06, "epoch": 2.4854700854700855, "percentage": 82.95, "elapsed_time": "16:09:48", "remaining_time": "3:19:18"}
{"current_steps": 1100, "total_steps": 1314, "loss": 0.6307, "learning_rate": 1e-06, "epoch": 2.5082621082621084, "percentage": 83.71, "elapsed_time": "16:18:33", "remaining_time": "3:10:22"}
{"current_steps": 1110, "total_steps": 1314, "loss": 0.6229, "learning_rate": 1e-06, "epoch": 2.5310541310541312, "percentage": 84.47, "elapsed_time": "16:27:18", "remaining_time": "3:01:27"}
{"current_steps": 1120, "total_steps": 1314, "loss": 0.621, "learning_rate": 1e-06, "epoch": 2.5538461538461537, "percentage": 85.24, "elapsed_time": "16:36:02", "remaining_time": "2:52:31"}
{"current_steps": 1130, "total_steps": 1314, "loss": 0.62, "learning_rate": 1e-06, "epoch": 2.5766381766381765, "percentage": 86.0, "elapsed_time": "16:44:47", "remaining_time": "2:43:36"}
{"current_steps": 1140, "total_steps": 1314, "loss": 0.6197, "learning_rate": 1e-06, "epoch": 2.5994301994301994, "percentage": 86.76, "elapsed_time": "16:53:31", "remaining_time": "2:34:41"}
{"current_steps": 1150, "total_steps": 1314, "loss": 0.6198, "learning_rate": 1e-06, "epoch": 2.6222222222222222, "percentage": 87.52, "elapsed_time": "17:02:13", "remaining_time": "2:25:46"}
{"current_steps": 1160, "total_steps": 1314, "loss": 0.6202, "learning_rate": 1e-06, "epoch": 2.645014245014245, "percentage": 88.28, "elapsed_time": "17:10:57", "remaining_time": "2:16:52"}
{"current_steps": 1170, "total_steps": 1314, "loss": 0.6238, "learning_rate": 1e-06, "epoch": 2.667806267806268, "percentage": 89.04, "elapsed_time": "17:19:42", "remaining_time": "2:07:57"}
{"current_steps": 1180, "total_steps": 1314, "loss": 0.6185, "learning_rate": 1e-06, "epoch": 2.690598290598291, "percentage": 89.8, "elapsed_time": "17:28:24", "remaining_time": "1:59:03"}
{"current_steps": 1190, "total_steps": 1314, "loss": 0.6191, "learning_rate": 1e-06, "epoch": 2.7133903133903132, "percentage": 90.56, "elapsed_time": "17:37:08", "remaining_time": "1:50:09"}
{"current_steps": 1200, "total_steps": 1314, "loss": 0.6227, "learning_rate": 1e-06, "epoch": 2.736182336182336, "percentage": 91.32, "elapsed_time": "17:45:53", "remaining_time": "1:41:15"}
{"current_steps": 1210, "total_steps": 1314, "loss": 0.6259, "learning_rate": 1e-06, "epoch": 2.758974358974359, "percentage": 92.09, "elapsed_time": "17:54:38", "remaining_time": "1:32:21"}
{"current_steps": 1220, "total_steps": 1314, "loss": 0.6152, "learning_rate": 1e-06, "epoch": 2.781766381766382, "percentage": 92.85, "elapsed_time": "18:03:22", "remaining_time": "1:23:28"}
{"current_steps": 1230, "total_steps": 1314, "loss": 0.6136, "learning_rate": 1e-06, "epoch": 2.8045584045584047, "percentage": 93.61, "elapsed_time": "18:12:07", "remaining_time": "1:14:35"}
{"current_steps": 1240, "total_steps": 1314, "loss": 0.6293, "learning_rate": 1e-06, "epoch": 2.827350427350427, "percentage": 94.37, "elapsed_time": "18:20:49", "remaining_time": "1:05:41"}
{"current_steps": 1250, "total_steps": 1314, "loss": 0.6258, "learning_rate": 1e-06, "epoch": 2.8501424501424504, "percentage": 95.13, "elapsed_time": "18:29:32", "remaining_time": "0:56:48"}
{"current_steps": 1260, "total_steps": 1314, "loss": 0.6225, "learning_rate": 1e-06, "epoch": 2.872934472934473, "percentage": 95.89, "elapsed_time": "18:38:16", "remaining_time": "0:47:55"}
{"current_steps": 1270, "total_steps": 1314, "loss": 0.6233, "learning_rate": 1e-06, "epoch": 2.8957264957264957, "percentage": 96.65, "elapsed_time": "18:46:59", "remaining_time": "0:39:02"}
{"current_steps": 1280, "total_steps": 1314, "loss": 0.6132, "learning_rate": 1e-06, "epoch": 2.9185185185185185, "percentage": 97.41, "elapsed_time": "18:55:45", "remaining_time": "0:30:10"}
{"current_steps": 1290, "total_steps": 1314, "loss": 0.6129, "learning_rate": 1e-06, "epoch": 2.9413105413105414, "percentage": 98.17, "elapsed_time": "19:04:29", "remaining_time": "0:21:17"}
{"current_steps": 1300, "total_steps": 1314, "loss": 0.6144, "learning_rate": 1e-06, "epoch": 2.9641025641025642, "percentage": 98.93, "elapsed_time": "19:13:12", "remaining_time": "0:12:25"}
{"current_steps": 1310, "total_steps": 1314, "loss": 0.625, "learning_rate": 1e-06, "epoch": 2.9868945868945866, "percentage": 99.7, "elapsed_time": "19:21:56", "remaining_time": "0:03:32"}
{"current_steps": 1314, "total_steps": 1314, "eval_loss": 0.6290851831436157, "epoch": 2.996011396011396, "percentage": 100.0, "elapsed_time": "19:33:57", "remaining_time": "0:00:00"}
{"current_steps": 1314, "total_steps": 1314, "epoch": 2.996011396011396, "percentage": 100.0, "elapsed_time": "19:35:19", "remaining_time": "0:00:00"}