chchen's picture
Training in progress, step 3295
3338017 verified
{"current_steps": 10, "total_steps": 3295, "loss": 1.6899, "lr": 3.0303030303030305e-06, "epoch": 0.015164439389631315, "percentage": 0.3, "elapsed_time": "0:01:26", "remaining_time": "7:55:08"}
{"current_steps": 20, "total_steps": 3295, "loss": 1.7303, "lr": 6.060606060606061e-06, "epoch": 0.03032887877926263, "percentage": 0.61, "elapsed_time": "0:02:54", "remaining_time": "7:55:30"}
{"current_steps": 30, "total_steps": 3295, "loss": 1.5576, "lr": 9.090909090909091e-06, "epoch": 0.045493318168893945, "percentage": 0.91, "elapsed_time": "0:04:21", "remaining_time": "7:54:21"}
{"current_steps": 40, "total_steps": 3295, "loss": 1.3227, "lr": 1.2121212121212122e-05, "epoch": 0.06065775755852526, "percentage": 1.21, "elapsed_time": "0:05:48", "remaining_time": "7:53:18"}
{"current_steps": 50, "total_steps": 3295, "loss": 0.8816, "lr": 1.5151515151515153e-05, "epoch": 0.07582219694815658, "percentage": 1.52, "elapsed_time": "0:07:16", "remaining_time": "7:51:59"}
{"current_steps": 50, "total_steps": 3295, "eval_loss": 0.6520895957946777, "epoch": 0.07582219694815658, "percentage": 1.52, "elapsed_time": "0:10:37", "remaining_time": "11:29:26"}
{"current_steps": 60, "total_steps": 3295, "loss": 0.6344, "lr": 1.8181818181818182e-05, "epoch": 0.09098663633778789, "percentage": 1.82, "elapsed_time": "0:12:07", "remaining_time": "10:53:59"}
{"current_steps": 70, "total_steps": 3295, "loss": 0.4792, "lr": 2.1212121212121215e-05, "epoch": 0.1061510757274192, "percentage": 2.12, "elapsed_time": "0:13:34", "remaining_time": "10:25:43"}
{"current_steps": 80, "total_steps": 3295, "loss": 0.356, "lr": 2.4242424242424244e-05, "epoch": 0.12131551511705052, "percentage": 2.43, "elapsed_time": "0:15:01", "remaining_time": "10:04:07"}
{"current_steps": 90, "total_steps": 3295, "loss": 0.1739, "lr": 2.7272727272727273e-05, "epoch": 0.13647995450668182, "percentage": 2.73, "elapsed_time": "0:16:29", "remaining_time": "9:47:15"}
{"current_steps": 100, "total_steps": 3295, "loss": 0.1876, "lr": 3.0303030303030306e-05, "epoch": 0.15164439389631315, "percentage": 3.03, "elapsed_time": "0:17:56", "remaining_time": "9:33:13"}
{"current_steps": 100, "total_steps": 3295, "eval_loss": 0.12008175253868103, "epoch": 0.15164439389631315, "percentage": 3.03, "elapsed_time": "0:21:16", "remaining_time": "11:19:44"}
{"current_steps": 110, "total_steps": 3295, "loss": 0.1265, "lr": 3.3333333333333335e-05, "epoch": 0.16680883328594445, "percentage": 3.34, "elapsed_time": "0:22:46", "remaining_time": "10:59:28"}
{"current_steps": 120, "total_steps": 3295, "loss": 0.1058, "lr": 3.6363636363636364e-05, "epoch": 0.18197327267557578, "percentage": 3.64, "elapsed_time": "0:24:13", "remaining_time": "10:40:53"}
{"current_steps": 130, "total_steps": 3295, "loss": 0.0875, "lr": 3.939393939393939e-05, "epoch": 0.19713771206520708, "percentage": 3.95, "elapsed_time": "0:25:40", "remaining_time": "10:24:59"}
{"current_steps": 140, "total_steps": 3295, "loss": 0.0752, "lr": 4.242424242424243e-05, "epoch": 0.2123021514548384, "percentage": 4.25, "elapsed_time": "0:27:06", "remaining_time": "10:11:04"}
{"current_steps": 150, "total_steps": 3295, "loss": 0.0847, "lr": 4.545454545454546e-05, "epoch": 0.2274665908444697, "percentage": 4.55, "elapsed_time": "0:28:33", "remaining_time": "9:58:52"}
{"current_steps": 150, "total_steps": 3295, "eval_loss": 0.0709112212061882, "epoch": 0.2274665908444697, "percentage": 4.55, "elapsed_time": "0:31:53", "remaining_time": "11:08:43"}
{"current_steps": 160, "total_steps": 3295, "loss": 0.0758, "lr": 4.848484848484849e-05, "epoch": 0.24263103023410104, "percentage": 4.86, "elapsed_time": "0:33:23", "remaining_time": "10:54:19"}
{"current_steps": 170, "total_steps": 3295, "loss": 0.07, "lr": 5.151515151515152e-05, "epoch": 0.25779546962373234, "percentage": 5.16, "elapsed_time": "0:34:50", "remaining_time": "10:40:28"}
{"current_steps": 180, "total_steps": 3295, "loss": 0.0762, "lr": 5.4545454545454546e-05, "epoch": 0.27295990901336364, "percentage": 5.46, "elapsed_time": "0:36:17", "remaining_time": "10:27:55"}
{"current_steps": 190, "total_steps": 3295, "loss": 0.0737, "lr": 5.757575757575758e-05, "epoch": 0.288124348402995, "percentage": 5.77, "elapsed_time": "0:37:44", "remaining_time": "10:16:40"}
{"current_steps": 200, "total_steps": 3295, "loss": 0.0547, "lr": 6.060606060606061e-05, "epoch": 0.3032887877926263, "percentage": 6.07, "elapsed_time": "0:39:11", "remaining_time": "10:06:21"}
{"current_steps": 200, "total_steps": 3295, "eval_loss": 0.059455741196870804, "epoch": 0.3032887877926263, "percentage": 6.07, "elapsed_time": "0:42:31", "remaining_time": "10:57:58"}
{"current_steps": 210, "total_steps": 3295, "loss": 0.0504, "lr": 6.363636363636364e-05, "epoch": 0.3184532271822576, "percentage": 6.37, "elapsed_time": "0:44:00", "remaining_time": "10:46:34"}
{"current_steps": 220, "total_steps": 3295, "loss": 0.0545, "lr": 6.666666666666667e-05, "epoch": 0.3336176665718889, "percentage": 6.68, "elapsed_time": "0:45:27", "remaining_time": "10:35:27"}
{"current_steps": 230, "total_steps": 3295, "loss": 0.0623, "lr": 6.96969696969697e-05, "epoch": 0.34878210596152026, "percentage": 6.98, "elapsed_time": "0:46:54", "remaining_time": "10:25:07"}
{"current_steps": 240, "total_steps": 3295, "loss": 0.0462, "lr": 7.272727272727273e-05, "epoch": 0.36394654535115156, "percentage": 7.28, "elapsed_time": "0:48:21", "remaining_time": "10:15:30"}
{"current_steps": 250, "total_steps": 3295, "loss": 0.05, "lr": 7.575757575757576e-05, "epoch": 0.37911098474078286, "percentage": 7.59, "elapsed_time": "0:49:48", "remaining_time": "10:06:36"}
{"current_steps": 250, "total_steps": 3295, "eval_loss": 0.050985436886548996, "epoch": 0.37911098474078286, "percentage": 7.59, "elapsed_time": "0:53:08", "remaining_time": "10:47:11"}
{"current_steps": 260, "total_steps": 3295, "loss": 0.0441, "lr": 7.878787878787879e-05, "epoch": 0.39427542413041416, "percentage": 7.89, "elapsed_time": "0:54:38", "remaining_time": "10:37:45"}
{"current_steps": 270, "total_steps": 3295, "loss": 0.0763, "lr": 8.181818181818183e-05, "epoch": 0.40943986352004547, "percentage": 8.19, "elapsed_time": "0:56:04", "remaining_time": "10:28:19"}
{"current_steps": 280, "total_steps": 3295, "loss": 0.0493, "lr": 8.484848484848486e-05, "epoch": 0.4246043029096768, "percentage": 8.5, "elapsed_time": "0:57:31", "remaining_time": "10:19:25"}
{"current_steps": 290, "total_steps": 3295, "loss": 0.0421, "lr": 8.787878787878789e-05, "epoch": 0.4397687422993081, "percentage": 8.8, "elapsed_time": "0:58:58", "remaining_time": "10:11:03"}
{"current_steps": 300, "total_steps": 3295, "loss": 0.0566, "lr": 9.090909090909092e-05, "epoch": 0.4549331816889394, "percentage": 9.1, "elapsed_time": "1:00:25", "remaining_time": "10:03:12"}
{"current_steps": 300, "total_steps": 3295, "eval_loss": 0.04939843714237213, "epoch": 0.4549331816889394, "percentage": 9.1, "elapsed_time": "1:03:44", "remaining_time": "10:36:25"}
{"current_steps": 310, "total_steps": 3295, "loss": 0.0454, "lr": 9.393939393939395e-05, "epoch": 0.4700976210785707, "percentage": 9.41, "elapsed_time": "1:05:14", "remaining_time": "10:28:15"}
{"current_steps": 320, "total_steps": 3295, "loss": 0.0476, "lr": 9.696969696969698e-05, "epoch": 0.4852620604682021, "percentage": 9.71, "elapsed_time": "1:06:41", "remaining_time": "10:19:57"}
{"current_steps": 330, "total_steps": 3295, "loss": 0.0506, "lr": 0.0001, "epoch": 0.5004264998578334, "percentage": 10.02, "elapsed_time": "1:08:07", "remaining_time": "10:12:07"}
{"current_steps": 340, "total_steps": 3295, "loss": 0.0458, "lr": 9.999719336268101e-05, "epoch": 0.5155909392474647, "percentage": 10.32, "elapsed_time": "1:09:34", "remaining_time": "10:04:40"}
{"current_steps": 350, "total_steps": 3295, "loss": 0.057, "lr": 9.998877376581251e-05, "epoch": 0.530755378637096, "percentage": 10.62, "elapsed_time": "1:11:01", "remaining_time": "9:57:38"}
{"current_steps": 350, "total_steps": 3295, "eval_loss": 0.04613954573869705, "epoch": 0.530755378637096, "percentage": 10.62, "elapsed_time": "1:14:21", "remaining_time": "10:25:38"}
{"current_steps": 360, "total_steps": 3295, "loss": 0.0478, "lr": 9.997474215462472e-05, "epoch": 0.5459198180267273, "percentage": 10.93, "elapsed_time": "1:15:50", "remaining_time": "10:18:19"}
{"current_steps": 370, "total_steps": 3295, "loss": 0.0344, "lr": 9.995510010438337e-05, "epoch": 0.5610842574163586, "percentage": 11.23, "elapsed_time": "1:17:17", "remaining_time": "10:10:58"}
{"current_steps": 380, "total_steps": 3295, "loss": 0.0543, "lr": 9.992984982021295e-05, "epoch": 0.57624869680599, "percentage": 11.53, "elapsed_time": "1:18:43", "remaining_time": "10:03:55"}
{"current_steps": 390, "total_steps": 3295, "loss": 0.0408, "lr": 9.9898994136849e-05, "epoch": 0.5914131361956213, "percentage": 11.84, "elapsed_time": "1:20:10", "remaining_time": "9:57:10"}
{"current_steps": 400, "total_steps": 3295, "loss": 0.0323, "lr": 9.986253651832005e-05, "epoch": 0.6065775755852526, "percentage": 12.14, "elapsed_time": "1:21:36", "remaining_time": "9:50:41"}
{"current_steps": 400, "total_steps": 3295, "eval_loss": 0.04223111271858215, "epoch": 0.6065775755852526, "percentage": 12.14, "elapsed_time": "1:24:56", "remaining_time": "10:14:43"}
{"current_steps": 410, "total_steps": 3295, "loss": 0.0389, "lr": 9.982048105755859e-05, "epoch": 0.6217420149748839, "percentage": 12.44, "elapsed_time": "1:26:25", "remaining_time": "10:08:06"}
{"current_steps": 420, "total_steps": 3295, "loss": 0.0476, "lr": 9.977283247594166e-05, "epoch": 0.6369064543645152, "percentage": 12.75, "elapsed_time": "1:27:51", "remaining_time": "10:01:27"}
{"current_steps": 430, "total_steps": 3295, "loss": 0.045, "lr": 9.971959612276076e-05, "epoch": 0.6520708937541465, "percentage": 13.05, "elapsed_time": "1:29:18", "remaining_time": "9:55:01"}
{"current_steps": 440, "total_steps": 3295, "loss": 0.0382, "lr": 9.966077797462129e-05, "epoch": 0.6672353331437778, "percentage": 13.35, "elapsed_time": "1:30:45", "remaining_time": "9:48:52"}
{"current_steps": 450, "total_steps": 3295, "loss": 0.0331, "lr": 9.959638463477165e-05, "epoch": 0.6823997725334091, "percentage": 13.66, "elapsed_time": "1:32:11", "remaining_time": "9:42:53"}
{"current_steps": 450, "total_steps": 3295, "eval_loss": 0.03925846517086029, "epoch": 0.6823997725334091, "percentage": 13.66, "elapsed_time": "1:35:30", "remaining_time": "10:03:49"}
{"current_steps": 460, "total_steps": 3295, "loss": 0.0415, "lr": 9.952642333236186e-05, "epoch": 0.6975642119230405, "percentage": 13.96, "elapsed_time": "1:36:59", "remaining_time": "9:57:47"}
{"current_steps": 470, "total_steps": 3295, "loss": 0.0453, "lr": 9.945090192163202e-05, "epoch": 0.7127286513126718, "percentage": 14.26, "elapsed_time": "1:38:26", "remaining_time": "9:51:41"}
{"current_steps": 480, "total_steps": 3295, "loss": 0.034, "lr": 9.936982888103051e-05, "epoch": 0.7278930907023031, "percentage": 14.57, "elapsed_time": "1:39:52", "remaining_time": "9:45:46"}
{"current_steps": 490, "total_steps": 3295, "loss": 0.0417, "lr": 9.928321331226219e-05, "epoch": 0.7430575300919344, "percentage": 14.87, "elapsed_time": "1:41:20", "remaining_time": "9:40:06"}
{"current_steps": 500, "total_steps": 3295, "loss": 0.0339, "lr": 9.919106493926655e-05, "epoch": 0.7582219694815657, "percentage": 15.17, "elapsed_time": "1:42:47", "remaining_time": "9:34:33"}
{"current_steps": 500, "total_steps": 3295, "eval_loss": 0.04052235931158066, "epoch": 0.7582219694815657, "percentage": 15.17, "elapsed_time": "1:46:05", "remaining_time": "9:53:05"}
{"current_steps": 510, "total_steps": 3295, "loss": 0.0414, "lr": 9.909339410712612e-05, "epoch": 0.773386408871197, "percentage": 15.48, "elapsed_time": "1:47:35", "remaining_time": "9:47:34"}
{"current_steps": 520, "total_steps": 3295, "loss": 0.0364, "lr": 9.8990211780905e-05, "epoch": 0.7885508482608283, "percentage": 15.78, "elapsed_time": "1:49:02", "remaining_time": "9:41:54"}
{"current_steps": 530, "total_steps": 3295, "loss": 0.0384, "lr": 9.888152954441785e-05, "epoch": 0.8037152876504596, "percentage": 16.08, "elapsed_time": "1:50:29", "remaining_time": "9:36:24"}
{"current_steps": 540, "total_steps": 3295, "loss": 0.0354, "lr": 9.876735959892953e-05, "epoch": 0.8188797270400909, "percentage": 16.39, "elapsed_time": "1:51:55", "remaining_time": "9:31:03"}
{"current_steps": 550, "total_steps": 3295, "loss": 0.0432, "lr": 9.864771476178522e-05, "epoch": 0.8340441664297223, "percentage": 16.69, "elapsed_time": "1:53:22", "remaining_time": "9:25:52"}
{"current_steps": 550, "total_steps": 3295, "eval_loss": 0.03828958794474602, "epoch": 0.8340441664297223, "percentage": 16.69, "elapsed_time": "1:56:41", "remaining_time": "9:42:24"}
{"current_steps": 560, "total_steps": 3295, "loss": 0.0466, "lr": 9.852260846497153e-05, "epoch": 0.8492086058193536, "percentage": 17.0, "elapsed_time": "1:58:10", "remaining_time": "9:37:11"}
{"current_steps": 570, "total_steps": 3295, "loss": 0.0448, "lr": 9.839205475360851e-05, "epoch": 0.864373045208985, "percentage": 17.3, "elapsed_time": "1:59:37", "remaining_time": "9:31:51"}
{"current_steps": 580, "total_steps": 3295, "loss": 0.0482, "lr": 9.825606828437291e-05, "epoch": 0.8795374845986162, "percentage": 17.6, "elapsed_time": "2:01:03", "remaining_time": "9:26:40"}
{"current_steps": 590, "total_steps": 3295, "loss": 0.0341, "lr": 9.811466432385267e-05, "epoch": 0.8947019239882475, "percentage": 17.91, "elapsed_time": "2:02:29", "remaining_time": "9:21:37"}
{"current_steps": 600, "total_steps": 3295, "loss": 0.0332, "lr": 9.796785874683314e-05, "epoch": 0.9098663633778789, "percentage": 18.21, "elapsed_time": "2:03:56", "remaining_time": "9:16:41"}
{"current_steps": 600, "total_steps": 3295, "eval_loss": 0.03612194210290909, "epoch": 0.9098663633778789, "percentage": 18.21, "elapsed_time": "2:07:15", "remaining_time": "9:31:34"}
{"current_steps": 610, "total_steps": 3295, "loss": 0.0443, "lr": 9.781566803451475e-05, "epoch": 0.9250308027675102, "percentage": 18.51, "elapsed_time": "2:08:44", "remaining_time": "9:26:40"}
{"current_steps": 620, "total_steps": 3295, "loss": 0.0334, "lr": 9.765810927266281e-05, "epoch": 0.9401952421571415, "percentage": 18.82, "elapsed_time": "2:10:11", "remaining_time": "9:21:40"}
{"current_steps": 630, "total_steps": 3295, "loss": 0.0341, "lr": 9.749520014968934e-05, "epoch": 0.9553596815467729, "percentage": 19.12, "elapsed_time": "2:11:37", "remaining_time": "9:16:46"}
{"current_steps": 640, "total_steps": 3295, "loss": 0.0478, "lr": 9.732695895466735e-05, "epoch": 0.9705241209364042, "percentage": 19.42, "elapsed_time": "2:13:03", "remaining_time": "9:12:00"}
{"current_steps": 650, "total_steps": 3295, "loss": 0.0458, "lr": 9.715340457527746e-05, "epoch": 0.9856885603260355, "percentage": 19.73, "elapsed_time": "2:14:30", "remaining_time": "9:07:19"}
{"current_steps": 650, "total_steps": 3295, "eval_loss": 0.03808917477726936, "epoch": 0.9856885603260355, "percentage": 19.73, "elapsed_time": "2:17:49", "remaining_time": "9:20:50"}
{"current_steps": 660, "total_steps": 3295, "loss": 0.0345, "lr": 9.697455649568761e-05, "epoch": 1.0008529997156668, "percentage": 20.03, "elapsed_time": "2:19:19", "remaining_time": "9:16:16"}
{"current_steps": 670, "total_steps": 3295, "loss": 0.0253, "lr": 9.679043479436556e-05, "epoch": 1.016017439105298, "percentage": 20.33, "elapsed_time": "2:20:46", "remaining_time": "9:11:32"}
{"current_steps": 680, "total_steps": 3295, "loss": 0.0242, "lr": 9.660106014182489e-05, "epoch": 1.0311818784949294, "percentage": 20.64, "elapsed_time": "2:22:12", "remaining_time": "9:06:54"}
{"current_steps": 690, "total_steps": 3295, "loss": 0.0437, "lr": 9.640645379830424e-05, "epoch": 1.0463463178845607, "percentage": 20.94, "elapsed_time": "2:23:39", "remaining_time": "9:02:22"}
{"current_steps": 700, "total_steps": 3295, "loss": 0.0281, "lr": 9.620663761138067e-05, "epoch": 1.061510757274192, "percentage": 21.24, "elapsed_time": "2:25:06", "remaining_time": "8:57:56"}
{"current_steps": 700, "total_steps": 3295, "eval_loss": 0.03682916238903999, "epoch": 1.061510757274192, "percentage": 21.24, "elapsed_time": "2:28:27", "remaining_time": "9:10:19"}
{"current_steps": 710, "total_steps": 3295, "loss": 0.0201, "lr": 9.600163401351688e-05, "epoch": 1.0766751966638233, "percentage": 21.55, "elapsed_time": "2:29:57", "remaining_time": "9:05:57"}
{"current_steps": 720, "total_steps": 3295, "loss": 0.04, "lr": 9.579146601954276e-05, "epoch": 1.0918396360534546, "percentage": 21.85, "elapsed_time": "2:31:24", "remaining_time": "9:01:30"}
{"current_steps": 730, "total_steps": 3295, "loss": 0.0334, "lr": 9.557615722407177e-05, "epoch": 1.1070040754430859, "percentage": 22.15, "elapsed_time": "2:32:51", "remaining_time": "8:57:05"}
{"current_steps": 740, "total_steps": 3295, "loss": 0.0254, "lr": 9.535573179885191e-05, "epoch": 1.1221685148327172, "percentage": 22.46, "elapsed_time": "2:34:18", "remaining_time": "8:52:46"}
{"current_steps": 750, "total_steps": 3295, "loss": 0.0222, "lr": 9.513021449005214e-05, "epoch": 1.1373329542223485, "percentage": 22.76, "elapsed_time": "2:35:45", "remaining_time": "8:48:31"}
{"current_steps": 750, "total_steps": 3295, "eval_loss": 0.03816114366054535, "epoch": 1.1373329542223485, "percentage": 22.76, "elapsed_time": "2:39:06", "remaining_time": "8:59:54"}
{"current_steps": 760, "total_steps": 3295, "loss": 0.0355, "lr": 9.489963061548428e-05, "epoch": 1.15249739361198, "percentage": 23.07, "elapsed_time": "2:40:37", "remaining_time": "8:55:44"}
{"current_steps": 770, "total_steps": 3295, "loss": 0.0178, "lr": 9.466400606176062e-05, "epoch": 1.1676618330016113, "percentage": 23.37, "elapsed_time": "2:42:04", "remaining_time": "8:51:27"}
{"current_steps": 780, "total_steps": 3295, "loss": 0.0219, "lr": 9.442336728138779e-05, "epoch": 1.1828262723912426, "percentage": 23.67, "elapsed_time": "2:43:31", "remaining_time": "8:47:16"}
{"current_steps": 790, "total_steps": 3295, "loss": 0.0329, "lr": 9.417774128979706e-05, "epoch": 1.197990711780874, "percentage": 23.98, "elapsed_time": "2:44:59", "remaining_time": "8:43:10"}
{"current_steps": 800, "total_steps": 3295, "loss": 0.0272, "lr": 9.39271556623114e-05, "epoch": 1.2131551511705052, "percentage": 24.28, "elapsed_time": "2:46:27", "remaining_time": "8:39:08"}
{"current_steps": 800, "total_steps": 3295, "eval_loss": 0.03464984521269798, "epoch": 1.2131551511705052, "percentage": 24.28, "elapsed_time": "2:49:49", "remaining_time": "8:49:38"}
{"current_steps": 810, "total_steps": 3295, "loss": 0.032, "lr": 9.367163853104975e-05, "epoch": 1.2283195905601365, "percentage": 24.58, "elapsed_time": "2:51:20", "remaining_time": "8:45:39"}
{"current_steps": 820, "total_steps": 3295, "loss": 0.0353, "lr": 9.341121858176876e-05, "epoch": 1.2434840299497678, "percentage": 24.89, "elapsed_time": "2:52:48", "remaining_time": "8:41:34"}
{"current_steps": 830, "total_steps": 3295, "loss": 0.0286, "lr": 9.314592505064239e-05, "epoch": 1.2586484693393991, "percentage": 25.19, "elapsed_time": "2:54:16", "remaining_time": "8:37:34"}
{"current_steps": 840, "total_steps": 3295, "loss": 0.0212, "lr": 9.28757877209796e-05, "epoch": 1.2738129087290304, "percentage": 25.49, "elapsed_time": "2:55:44", "remaining_time": "8:33:37"}
{"current_steps": 850, "total_steps": 3295, "loss": 0.0303, "lr": 9.260083691988084e-05, "epoch": 1.2889773481186617, "percentage": 25.8, "elapsed_time": "2:57:12", "remaining_time": "8:29:44"}
{"current_steps": 850, "total_steps": 3295, "eval_loss": 0.03515174984931946, "epoch": 1.2889773481186617, "percentage": 25.8, "elapsed_time": "3:00:35", "remaining_time": "8:39:29"}
{"current_steps": 860, "total_steps": 3295, "loss": 0.0323, "lr": 9.232110351483327e-05, "epoch": 1.304141787508293, "percentage": 26.1, "elapsed_time": "3:02:07", "remaining_time": "8:35:39"}
{"current_steps": 870, "total_steps": 3295, "loss": 0.0321, "lr": 9.203661891024547e-05, "epoch": 1.3193062268979243, "percentage": 26.4, "elapsed_time": "3:03:35", "remaining_time": "8:31:44"}
{"current_steps": 880, "total_steps": 3295, "loss": 0.0237, "lr": 9.174741504392173e-05, "epoch": 1.3344706662875556, "percentage": 26.71, "elapsed_time": "3:05:03", "remaining_time": "8:27:52"}
{"current_steps": 890, "total_steps": 3295, "loss": 0.0319, "lr": 9.145352438347662e-05, "epoch": 1.349635105677187, "percentage": 27.01, "elapsed_time": "3:06:32", "remaining_time": "8:24:05"}
{"current_steps": 900, "total_steps": 3295, "loss": 0.0318, "lr": 9.115497992268995e-05, "epoch": 1.3647995450668184, "percentage": 27.31, "elapsed_time": "3:08:01", "remaining_time": "8:20:21"}
{"current_steps": 900, "total_steps": 3295, "eval_loss": 0.035844411700963974, "epoch": 1.3647995450668184, "percentage": 27.31, "elapsed_time": "3:11:25", "remaining_time": "8:29:24"}
{"current_steps": 910, "total_steps": 3295, "loss": 0.0316, "lr": 9.085181517780273e-05, "epoch": 1.3799639844564497, "percentage": 27.62, "elapsed_time": "3:12:57", "remaining_time": "8:25:43"}
{"current_steps": 920, "total_steps": 3295, "loss": 0.0318, "lr": 9.054406418375443e-05, "epoch": 1.395128423846081, "percentage": 27.92, "elapsed_time": "3:14:26", "remaining_time": "8:21:56"}
{"current_steps": 930, "total_steps": 3295, "loss": 0.0266, "lr": 9.023176149036203e-05, "epoch": 1.4102928632357123, "percentage": 28.22, "elapsed_time": "3:15:55", "remaining_time": "8:18:13"}
{"current_steps": 940, "total_steps": 3295, "loss": 0.0358, "lr": 8.991494215844132e-05, "epoch": 1.4254573026253436, "percentage": 28.53, "elapsed_time": "3:17:24", "remaining_time": "8:14:33"}
{"current_steps": 950, "total_steps": 3295, "loss": 0.0233, "lr": 8.959364175587069e-05, "epoch": 1.440621742014975, "percentage": 28.83, "elapsed_time": "3:18:52", "remaining_time": "8:10:55"}
{"current_steps": 950, "total_steps": 3295, "eval_loss": 0.035272736102342606, "epoch": 1.440621742014975, "percentage": 28.83, "elapsed_time": "3:22:17", "remaining_time": "8:19:20"}
{"current_steps": 960, "total_steps": 3295, "loss": 0.0361, "lr": 8.926789635359817e-05, "epoch": 1.4557861814046062, "percentage": 29.14, "elapsed_time": "3:23:49", "remaining_time": "8:15:46"}
{"current_steps": 970, "total_steps": 3295, "loss": 0.025, "lr": 8.893774252159187e-05, "epoch": 1.4709506207942376, "percentage": 29.44, "elapsed_time": "3:25:18", "remaining_time": "8:12:07"}
{"current_steps": 980, "total_steps": 3295, "loss": 0.0328, "lr": 8.860321732473439e-05, "epoch": 1.4861150601838689, "percentage": 29.74, "elapsed_time": "3:26:48", "remaining_time": "8:08:30"}
{"current_steps": 990, "total_steps": 3295, "loss": 0.0338, "lr": 8.826435831866184e-05, "epoch": 1.5012794995735002, "percentage": 30.05, "elapsed_time": "3:28:16", "remaining_time": "8:04:56"}
{"current_steps": 1000, "total_steps": 3295, "loss": 0.0263, "lr": 8.79212035455475e-05, "epoch": 1.5164439389631315, "percentage": 30.35, "elapsed_time": "3:29:45", "remaining_time": "8:01:24"}
{"current_steps": 1000, "total_steps": 3295, "eval_loss": 0.03487205505371094, "epoch": 1.5164439389631315, "percentage": 30.35, "elapsed_time": "3:33:11", "remaining_time": "8:09:15"}
{"current_steps": 1010, "total_steps": 3295, "loss": 0.0245, "lr": 8.757379152983103e-05, "epoch": 1.5316083783527628, "percentage": 30.65, "elapsed_time": "3:34:43", "remaining_time": "8:05:47"}
{"current_steps": 1020, "total_steps": 3295, "loss": 0.028, "lr": 8.722216127389363e-05, "epoch": 1.546772817742394, "percentage": 30.96, "elapsed_time": "3:36:12", "remaining_time": "8:02:14"}
{"current_steps": 1030, "total_steps": 3295, "loss": 0.0317, "lr": 8.686635225367919e-05, "epoch": 1.5619372571320254, "percentage": 31.26, "elapsed_time": "3:37:42", "remaining_time": "7:58:44"}
{"current_steps": 1040, "total_steps": 3295, "loss": 0.0338, "lr": 8.650640441426274e-05, "epoch": 1.5771016965216567, "percentage": 31.56, "elapsed_time": "3:39:11", "remaining_time": "7:55:16"}
{"current_steps": 1050, "total_steps": 3295, "loss": 0.0381, "lr": 8.614235816536582e-05, "epoch": 1.592266135911288, "percentage": 31.87, "elapsed_time": "3:40:40", "remaining_time": "7:51:49"}
{"current_steps": 1050, "total_steps": 3295, "eval_loss": 0.035373929888010025, "epoch": 1.592266135911288, "percentage": 31.87, "elapsed_time": "3:44:06", "remaining_time": "7:59:09"}
{"current_steps": 1060, "total_steps": 3295, "loss": 0.039, "lr": 8.577425437681994e-05, "epoch": 1.6074305753009193, "percentage": 32.17, "elapsed_time": "3:45:39", "remaining_time": "7:55:48"}
{"current_steps": 1070, "total_steps": 3295, "loss": 0.03, "lr": 8.540213437397833e-05, "epoch": 1.6225950146905506, "percentage": 32.47, "elapsed_time": "3:47:08", "remaining_time": "7:52:19"}
{"current_steps": 1080, "total_steps": 3295, "loss": 0.027, "lr": 8.502603993307647e-05, "epoch": 1.6377594540801819, "percentage": 32.78, "elapsed_time": "3:48:37", "remaining_time": "7:48:54"}
{"current_steps": 1090, "total_steps": 3295, "loss": 0.023, "lr": 8.464601327654207e-05, "epoch": 1.6529238934698132, "percentage": 33.08, "elapsed_time": "3:50:07", "remaining_time": "7:45:31"}
{"current_steps": 1100, "total_steps": 3295, "loss": 0.0267, "lr": 8.4262097068255e-05, "epoch": 1.6680883328594445, "percentage": 33.38, "elapsed_time": "3:51:36", "remaining_time": "7:42:10"}
{"current_steps": 1100, "total_steps": 3295, "eval_loss": 0.0319027379155159, "epoch": 1.6680883328594445, "percentage": 33.38, "elapsed_time": "3:55:03", "remaining_time": "7:49:02"}
{"current_steps": 1110, "total_steps": 3295, "loss": 0.0296, "lr": 8.387433440875758e-05, "epoch": 1.6832527722490758, "percentage": 33.69, "elapsed_time": "3:56:35", "remaining_time": "7:45:43"}
{"current_steps": 1120, "total_steps": 3295, "loss": 0.0311, "lr": 8.348276883041583e-05, "epoch": 1.698417211638707, "percentage": 33.99, "elapsed_time": "3:58:04", "remaining_time": "7:42:20"}
{"current_steps": 1130, "total_steps": 3295, "loss": 0.0314, "lr": 8.308744429253238e-05, "epoch": 1.7135816510283386, "percentage": 34.29, "elapsed_time": "3:59:34", "remaining_time": "7:39:00"}
{"current_steps": 1140, "total_steps": 3295, "loss": 0.0292, "lr": 8.268840517641123e-05, "epoch": 1.72874609041797, "percentage": 34.6, "elapsed_time": "4:01:04", "remaining_time": "7:35:42"}
{"current_steps": 1150, "total_steps": 3295, "loss": 0.0335, "lr": 8.228569628037543e-05, "epoch": 1.7439105298076012, "percentage": 34.9, "elapsed_time": "4:02:34", "remaining_time": "7:32:26"}
{"current_steps": 1150, "total_steps": 3295, "eval_loss": 0.032032061368227005, "epoch": 1.7439105298076012, "percentage": 34.9, "elapsed_time": "4:06:01", "remaining_time": "7:38:52"}
{"current_steps": 1160, "total_steps": 3295, "loss": 0.0285, "lr": 8.187936281473758e-05, "epoch": 1.7590749691972325, "percentage": 35.2, "elapsed_time": "4:07:34", "remaining_time": "7:35:39"}
{"current_steps": 1170, "total_steps": 3295, "loss": 0.024, "lr": 8.146945039672446e-05, "epoch": 1.7742394085868638, "percentage": 35.51, "elapsed_time": "4:09:04", "remaining_time": "7:32:22"}
{"current_steps": 1180, "total_steps": 3295, "loss": 0.0292, "lr": 8.105600504535567e-05, "epoch": 1.789403847976495, "percentage": 35.81, "elapsed_time": "4:10:33", "remaining_time": "7:29:06"}
{"current_steps": 1190, "total_steps": 3295, "loss": 0.0379, "lr": 8.06390731762773e-05, "epoch": 1.8045682873661264, "percentage": 36.12, "elapsed_time": "4:12:03", "remaining_time": "7:25:52"}
{"current_steps": 1200, "total_steps": 3295, "loss": 0.0228, "lr": 8.021870159655109e-05, "epoch": 1.8197327267557577, "percentage": 36.42, "elapsed_time": "4:13:34", "remaining_time": "7:22:41"}
{"current_steps": 1200, "total_steps": 3295, "eval_loss": 0.03198724240064621, "epoch": 1.8197327267557577, "percentage": 36.42, "elapsed_time": "4:17:00", "remaining_time": "7:28:42"}
{"current_steps": 1210, "total_steps": 3295, "loss": 0.0262, "lr": 7.979493749939955e-05, "epoch": 1.8348971661453892, "percentage": 36.72, "elapsed_time": "4:18:33", "remaining_time": "7:25:32"}
{"current_steps": 1220, "total_steps": 3295, "loss": 0.0348, "lr": 7.93678284589079e-05, "epoch": 1.8500616055350205, "percentage": 37.03, "elapsed_time": "4:20:03", "remaining_time": "7:22:19"}
{"current_steps": 1230, "total_steps": 3295, "loss": 0.0278, "lr": 7.893742242468301e-05, "epoch": 1.8652260449246518, "percentage": 37.33, "elapsed_time": "4:21:33", "remaining_time": "7:19:07"}
{"current_steps": 1240, "total_steps": 3295, "loss": 0.0361, "lr": 7.850376771647038e-05, "epoch": 1.8803904843142831, "percentage": 37.63, "elapsed_time": "4:23:04", "remaining_time": "7:15:58"}
{"current_steps": 1250, "total_steps": 3295, "loss": 0.0311, "lr": 7.806691301872958e-05, "epoch": 1.8955549237039144, "percentage": 37.94, "elapsed_time": "4:24:34", "remaining_time": "7:12:50"}
{"current_steps": 1250, "total_steps": 3295, "eval_loss": 0.031337134540081024, "epoch": 1.8955549237039144, "percentage": 37.94, "elapsed_time": "4:28:01", "remaining_time": "7:18:28"}
{"current_steps": 1260, "total_steps": 3295, "loss": 0.0306, "lr": 7.762690737516846e-05, "epoch": 1.9107193630935457, "percentage": 38.24, "elapsed_time": "4:29:35", "remaining_time": "7:15:24"}
{"current_steps": 1270, "total_steps": 3295, "loss": 0.0233, "lr": 7.718380018323742e-05, "epoch": 1.925883802483177, "percentage": 38.54, "elapsed_time": "4:31:05", "remaining_time": "7:12:15"}
{"current_steps": 1280, "total_steps": 3295, "loss": 0.0242, "lr": 7.673764118858371e-05, "epoch": 1.9410482418728083, "percentage": 38.85, "elapsed_time": "4:32:35", "remaining_time": "7:09:07"}
{"current_steps": 1290, "total_steps": 3295, "loss": 0.0332, "lr": 7.628848047946675e-05, "epoch": 1.9562126812624396, "percentage": 39.15, "elapsed_time": "4:34:05", "remaining_time": "7:06:00"}
{"current_steps": 1300, "total_steps": 3295, "loss": 0.0228, "lr": 7.583636848113483e-05, "epoch": 1.971377120652071, "percentage": 39.45, "elapsed_time": "4:35:35", "remaining_time": "7:02:55"}
{"current_steps": 1300, "total_steps": 3295, "eval_loss": 0.03141847252845764, "epoch": 1.971377120652071, "percentage": 39.45, "elapsed_time": "4:39:02", "remaining_time": "7:08:13"}
{"current_steps": 1310, "total_steps": 3295, "loss": 0.0309, "lr": 7.538135595016423e-05, "epoch": 1.9865415600417022, "percentage": 39.76, "elapsed_time": "4:40:36", "remaining_time": "7:05:12"}
{"current_steps": 1320, "total_steps": 3295, "loss": 0.0279, "lr": 7.492349396876096e-05, "epoch": 2.0017059994313335, "percentage": 40.06, "elapsed_time": "4:42:06", "remaining_time": "7:02:06"}
{"current_steps": 1330, "total_steps": 3295, "loss": 0.0216, "lr": 7.4462833939026e-05, "epoch": 2.016870438820965, "percentage": 40.36, "elapsed_time": "4:43:37", "remaining_time": "6:59:02"}
{"current_steps": 1340, "total_steps": 3295, "loss": 0.0179, "lr": 7.399942757718455e-05, "epoch": 2.032034878210596, "percentage": 40.67, "elapsed_time": "4:45:07", "remaining_time": "6:55:58"}
{"current_steps": 1350, "total_steps": 3295, "loss": 0.0157, "lr": 7.35333269077802e-05, "epoch": 2.0471993176002274, "percentage": 40.97, "elapsed_time": "4:46:37", "remaining_time": "6:52:56"}
{"current_steps": 1350, "total_steps": 3295, "eval_loss": 0.03206400200724602, "epoch": 2.0471993176002274, "percentage": 40.97, "elapsed_time": "4:50:05", "remaining_time": "6:57:56"}
{"current_steps": 1360, "total_steps": 3295, "loss": 0.0242, "lr": 7.306458425783426e-05, "epoch": 2.0623637569898587, "percentage": 41.27, "elapsed_time": "4:51:38", "remaining_time": "6:54:57"}
{"current_steps": 1370, "total_steps": 3295, "loss": 0.0169, "lr": 7.25932522509713e-05, "epoch": 2.07752819637949, "percentage": 41.58, "elapsed_time": "4:53:09", "remaining_time": "6:51:54"}
{"current_steps": 1380, "total_steps": 3295, "loss": 0.0215, "lr": 7.211938380151133e-05, "epoch": 2.0926926357691213, "percentage": 41.88, "elapsed_time": "4:54:38", "remaining_time": "6:48:52"}
{"current_steps": 1390, "total_steps": 3295, "loss": 0.0241, "lr": 7.164303210852934e-05, "epoch": 2.1078570751587526, "percentage": 42.19, "elapsed_time": "4:56:08", "remaining_time": "6:45:51"}
{"current_steps": 1400, "total_steps": 3295, "loss": 0.0129, "lr": 7.116425064988286e-05, "epoch": 2.123021514548384, "percentage": 42.49, "elapsed_time": "4:57:38", "remaining_time": "6:42:53"}
{"current_steps": 1400, "total_steps": 3295, "eval_loss": 0.03161005303263664, "epoch": 2.123021514548384, "percentage": 42.49, "elapsed_time": "5:01:06", "remaining_time": "6:47:34"}
{"current_steps": 1410, "total_steps": 3295, "loss": 0.0204, "lr": 7.068309317620827e-05, "epoch": 2.1381859539380152, "percentage": 42.79, "elapsed_time": "5:02:40", "remaining_time": "6:44:38"}
{"current_steps": 1420, "total_steps": 3295, "loss": 0.0167, "lr": 7.019961370488645e-05, "epoch": 2.1533503933276466, "percentage": 43.1, "elapsed_time": "5:04:11", "remaining_time": "6:41:39"}
{"current_steps": 1430, "total_steps": 3295, "loss": 0.0165, "lr": 6.971386651397849e-05, "epoch": 2.168514832717278, "percentage": 43.4, "elapsed_time": "5:05:41", "remaining_time": "6:38:41"}
{"current_steps": 1440, "total_steps": 3295, "loss": 0.0148, "lr": 6.922590613613211e-05, "epoch": 2.183679272106909, "percentage": 43.7, "elapsed_time": "5:07:12", "remaining_time": "6:35:44"}
{"current_steps": 1450, "total_steps": 3295, "loss": 0.0208, "lr": 6.873578735245961e-05, "epoch": 2.1988437114965405, "percentage": 44.01, "elapsed_time": "5:08:43", "remaining_time": "6:32:49"}
{"current_steps": 1450, "total_steps": 3295, "eval_loss": 0.03302780166268349, "epoch": 2.1988437114965405, "percentage": 44.01, "elapsed_time": "5:12:12", "remaining_time": "6:37:15"}
{"current_steps": 1460, "total_steps": 3295, "loss": 0.0218, "lr": 6.824356518638775e-05, "epoch": 2.2140081508861718, "percentage": 44.31, "elapsed_time": "5:13:45", "remaining_time": "6:34:21"}
{"current_steps": 1470, "total_steps": 3295, "loss": 0.0175, "lr": 6.774929489748052e-05, "epoch": 2.229172590275803, "percentage": 44.61, "elapsed_time": "5:15:16", "remaining_time": "6:31:24"}
{"current_steps": 1480, "total_steps": 3295, "loss": 0.0269, "lr": 6.725303197523548e-05, "epoch": 2.2443370296654344, "percentage": 44.92, "elapsed_time": "5:16:46", "remaining_time": "6:28:28"}
{"current_steps": 1490, "total_steps": 3295, "loss": 0.017, "lr": 6.675483213285412e-05, "epoch": 2.259501469055066, "percentage": 45.22, "elapsed_time": "5:18:16", "remaining_time": "6:25:33"}
{"current_steps": 1500, "total_steps": 3295, "loss": 0.0191, "lr": 6.625475130098728e-05, "epoch": 2.274665908444697, "percentage": 45.52, "elapsed_time": "5:19:47", "remaining_time": "6:22:40"}
{"current_steps": 1500, "total_steps": 3295, "eval_loss": 0.03248446062207222, "epoch": 2.274665908444697, "percentage": 45.52, "elapsed_time": "5:23:15", "remaining_time": "6:26:50"}
{"current_steps": 1510, "total_steps": 3295, "loss": 0.0174, "lr": 6.575284562145593e-05, "epoch": 2.2898303478343287, "percentage": 45.83, "elapsed_time": "5:24:50", "remaining_time": "6:23:59"}
{"current_steps": 1520, "total_steps": 3295, "loss": 0.0242, "lr": 6.524917144094851e-05, "epoch": 2.30499478722396, "percentage": 46.13, "elapsed_time": "5:26:20", "remaining_time": "6:21:05"}
{"current_steps": 1530, "total_steps": 3295, "loss": 0.0237, "lr": 6.474378530469509e-05, "epoch": 2.3201592266135913, "percentage": 46.43, "elapsed_time": "5:27:50", "remaining_time": "6:18:12"}
{"current_steps": 1540, "total_steps": 3295, "loss": 0.0148, "lr": 6.42367439501193e-05, "epoch": 2.3353236660032226, "percentage": 46.74, "elapsed_time": "5:29:21", "remaining_time": "6:15:20"}
{"current_steps": 1550, "total_steps": 3295, "loss": 0.0189, "lr": 6.372810430046862e-05, "epoch": 2.350488105392854, "percentage": 47.04, "elapsed_time": "5:30:52", "remaining_time": "6:12:29"}
{"current_steps": 1550, "total_steps": 3295, "eval_loss": 0.03253987804055214, "epoch": 2.350488105392854, "percentage": 47.04, "elapsed_time": "5:34:20", "remaining_time": "6:16:24"}
{"current_steps": 1560, "total_steps": 3295, "loss": 0.0219, "lr": 6.321792345842402e-05, "epoch": 2.365652544782485, "percentage": 47.34, "elapsed_time": "5:35:54", "remaining_time": "6:13:35"}
{"current_steps": 1570, "total_steps": 3295, "loss": 0.0294, "lr": 6.270625869968906e-05, "epoch": 2.3808169841721165, "percentage": 47.65, "elapsed_time": "5:37:25", "remaining_time": "6:10:43"}
{"current_steps": 1580, "total_steps": 3295, "loss": 0.0216, "lr": 6.219316746656007e-05, "epoch": 2.395981423561748, "percentage": 47.95, "elapsed_time": "5:38:55", "remaining_time": "6:07:53"}
{"current_steps": 1590, "total_steps": 3295, "loss": 0.0217, "lr": 6.167870736147713e-05, "epoch": 2.411145862951379, "percentage": 48.25, "elapsed_time": "5:40:26", "remaining_time": "6:05:04"}
{"current_steps": 1600, "total_steps": 3295, "loss": 0.0161, "lr": 6.116293614055744e-05, "epoch": 2.4263103023410104, "percentage": 48.56, "elapsed_time": "5:41:57", "remaining_time": "6:02:15"}
{"current_steps": 1600, "total_steps": 3295, "eval_loss": 0.03252248466014862, "epoch": 2.4263103023410104, "percentage": 48.56, "elapsed_time": "5:45:26", "remaining_time": "6:05:56"}
{"current_steps": 1610, "total_steps": 3295, "loss": 0.0163, "lr": 6.06459117071113e-05, "epoch": 2.4414747417306417, "percentage": 48.86, "elapsed_time": "5:47:00", "remaining_time": "6:03:10"}
{"current_steps": 1620, "total_steps": 3295, "loss": 0.0174, "lr": 6.012769210514146e-05, "epoch": 2.456639181120273, "percentage": 49.17, "elapsed_time": "5:48:31", "remaining_time": "6:00:21"}
{"current_steps": 1630, "total_steps": 3295, "loss": 0.0178, "lr": 5.9608335512826915e-05, "epoch": 2.4718036205099043, "percentage": 49.47, "elapsed_time": "5:50:02", "remaining_time": "5:57:33"}
{"current_steps": 1640, "total_steps": 3295, "loss": 0.0117, "lr": 5.908790023599144e-05, "epoch": 2.4869680598995356, "percentage": 49.77, "elapsed_time": "5:51:33", "remaining_time": "5:54:45"}
{"current_steps": 1650, "total_steps": 3295, "loss": 0.0164, "lr": 5.856644470155781e-05, "epoch": 2.502132499289167, "percentage": 50.08, "elapsed_time": "5:53:03", "remaining_time": "5:51:59"}
{"current_steps": 1650, "total_steps": 3295, "eval_loss": 0.03559296950697899, "epoch": 2.502132499289167, "percentage": 50.08, "elapsed_time": "5:56:32", "remaining_time": "5:55:27"}
{"current_steps": 1660, "total_steps": 3295, "loss": 0.021, "lr": 5.8044027450988546e-05, "epoch": 2.5172969386787982, "percentage": 50.38, "elapsed_time": "5:58:06", "remaining_time": "5:52:43"}
{"current_steps": 1670, "total_steps": 3295, "loss": 0.0147, "lr": 5.752070713371371e-05, "epoch": 2.5324613780684295, "percentage": 50.68, "elapsed_time": "5:59:37", "remaining_time": "5:49:56"}
{"current_steps": 1680, "total_steps": 3295, "loss": 0.0143, "lr": 5.699654250054662e-05, "epoch": 2.547625817458061, "percentage": 50.99, "elapsed_time": "6:01:08", "remaining_time": "5:47:09"}
{"current_steps": 1690, "total_steps": 3295, "loss": 0.0153, "lr": 5.647159239708809e-05, "epoch": 2.562790256847692, "percentage": 51.29, "elapsed_time": "6:02:38", "remaining_time": "5:44:24"}
{"current_steps": 1700, "total_steps": 3295, "loss": 0.0144, "lr": 5.5945915757120146e-05, "epoch": 2.5779546962373234, "percentage": 51.59, "elapsed_time": "6:04:09", "remaining_time": "5:41:40"}
{"current_steps": 1700, "total_steps": 3295, "eval_loss": 0.0337643064558506, "epoch": 2.5779546962373234, "percentage": 51.59, "elapsed_time": "6:07:39", "remaining_time": "5:44:56"}
{"current_steps": 1710, "total_steps": 3295, "loss": 0.0176, "lr": 5.5419571595989825e-05, "epoch": 2.5931191356269547, "percentage": 51.9, "elapsed_time": "6:09:13", "remaining_time": "5:42:13"}
{"current_steps": 1720, "total_steps": 3295, "loss": 0.02, "lr": 5.4892619003983734e-05, "epoch": 2.608283575016586, "percentage": 52.2, "elapsed_time": "6:10:44", "remaining_time": "5:39:29"}
{"current_steps": 1730, "total_steps": 3295, "loss": 0.0134, "lr": 5.436511713969428e-05, "epoch": 2.6234480144062173, "percentage": 52.5, "elapsed_time": "6:12:15", "remaining_time": "5:36:45"}
{"current_steps": 1740, "total_steps": 3295, "loss": 0.022, "lr": 5.383712522337817e-05, "epoch": 2.6386124537958486, "percentage": 52.81, "elapsed_time": "6:13:46", "remaining_time": "5:34:01"}
{"current_steps": 1750, "total_steps": 3295, "loss": 0.0211, "lr": 5.3308702530308076e-05, "epoch": 2.65377689318548, "percentage": 53.11, "elapsed_time": "6:15:17", "remaining_time": "5:31:19"}
{"current_steps": 1750, "total_steps": 3295, "eval_loss": 0.032832950353622437, "epoch": 2.65377689318548, "percentage": 53.11, "elapsed_time": "6:18:47", "remaining_time": "5:34:24"}
{"current_steps": 1760, "total_steps": 3295, "loss": 0.0233, "lr": 5.2779908384118025e-05, "epoch": 2.6689413325751112, "percentage": 53.41, "elapsed_time": "6:20:21", "remaining_time": "5:31:44"}
{"current_steps": 1770, "total_steps": 3295, "loss": 0.0233, "lr": 5.22508021501434e-05, "epoch": 2.6841057719647425, "percentage": 53.72, "elapsed_time": "6:21:52", "remaining_time": "5:29:01"}
{"current_steps": 1780, "total_steps": 3295, "loss": 0.015, "lr": 5.1721443228756284e-05, "epoch": 2.699270211354374, "percentage": 54.02, "elapsed_time": "6:23:23", "remaining_time": "5:26:18"}
{"current_steps": 1790, "total_steps": 3295, "loss": 0.0175, "lr": 5.119189104869683e-05, "epoch": 2.714434650744005, "percentage": 54.32, "elapsed_time": "6:24:53", "remaining_time": "5:23:36"}
{"current_steps": 1800, "total_steps": 3295, "loss": 0.0195, "lr": 5.066220506040148e-05, "epoch": 2.729599090133637, "percentage": 54.63, "elapsed_time": "6:26:23", "remaining_time": "5:20:55"}
{"current_steps": 1800, "total_steps": 3295, "eval_loss": 0.03220284357666969, "epoch": 2.729599090133637, "percentage": 54.63, "elapsed_time": "6:29:49", "remaining_time": "5:23:46"}
{"current_steps": 1810, "total_steps": 3295, "loss": 0.0202, "lr": 5.013244472932872e-05, "epoch": 2.7447635295232677, "percentage": 54.93, "elapsed_time": "6:31:23", "remaining_time": "5:21:06"}
{"current_steps": 1820, "total_steps": 3295, "loss": 0.0179, "lr": 4.960266952928316e-05, "epoch": 2.7599279689128995, "percentage": 55.24, "elapsed_time": "6:32:54", "remaining_time": "5:18:25"}
{"current_steps": 1830, "total_steps": 3295, "loss": 0.0163, "lr": 4.907293893573867e-05, "epoch": 2.7750924083025303, "percentage": 55.54, "elapsed_time": "6:34:25", "remaining_time": "5:15:45"}
{"current_steps": 1840, "total_steps": 3295, "loss": 0.0182, "lr": 4.8543312419161396e-05, "epoch": 2.790256847692162, "percentage": 55.84, "elapsed_time": "6:35:56", "remaining_time": "5:13:05"}
{"current_steps": 1850, "total_steps": 3295, "loss": 0.0148, "lr": 4.8013849438333165e-05, "epoch": 2.805421287081793, "percentage": 56.15, "elapsed_time": "6:37:27", "remaining_time": "5:10:26"}
{"current_steps": 1850, "total_steps": 3295, "eval_loss": 0.033808089792728424, "epoch": 2.805421287081793, "percentage": 56.15, "elapsed_time": "6:40:56", "remaining_time": "5:13:10"}
{"current_steps": 1860, "total_steps": 3295, "loss": 0.0189, "lr": 4.748460943367643e-05, "epoch": 2.8205857264714247, "percentage": 56.45, "elapsed_time": "6:42:31", "remaining_time": "5:10:33"}
{"current_steps": 1870, "total_steps": 3295, "loss": 0.0256, "lr": 4.695565182058113e-05, "epoch": 2.8357501658610555, "percentage": 56.75, "elapsed_time": "6:44:02", "remaining_time": "5:07:53"}
{"current_steps": 1880, "total_steps": 3295, "loss": 0.017, "lr": 4.642703598273431e-05, "epoch": 2.8509146052506873, "percentage": 57.06, "elapsed_time": "6:45:33", "remaining_time": "5:05:14"}
{"current_steps": 1890, "total_steps": 3295, "loss": 0.0189, "lr": 4.589882126545352e-05, "epoch": 2.8660790446403186, "percentage": 57.36, "elapsed_time": "6:47:03", "remaining_time": "5:02:36"}
{"current_steps": 1900, "total_steps": 3295, "loss": 0.0249, "lr": 4.537106696902425e-05, "epoch": 2.88124348402995, "percentage": 57.66, "elapsed_time": "6:48:34", "remaining_time": "4:59:58"}
{"current_steps": 1900, "total_steps": 3295, "eval_loss": 0.03270639479160309, "epoch": 2.88124348402995, "percentage": 57.66, "elapsed_time": "6:52:04", "remaining_time": "5:02:32"}
{"current_steps": 1910, "total_steps": 3295, "loss": 0.022, "lr": 4.484383234204266e-05, "epoch": 2.896407923419581, "percentage": 57.97, "elapsed_time": "6:53:38", "remaining_time": "4:59:56"}
{"current_steps": 1920, "total_steps": 3295, "loss": 0.0184, "lr": 4.4317176574763935e-05, "epoch": 2.9115723628092125, "percentage": 58.27, "elapsed_time": "6:55:09", "remaining_time": "4:57:19"}
{"current_steps": 1930, "total_steps": 3295, "loss": 0.0211, "lr": 4.3791158792457334e-05, "epoch": 2.926736802198844, "percentage": 58.57, "elapsed_time": "6:56:40", "remaining_time": "4:54:41"}
{"current_steps": 1940, "total_steps": 3295, "loss": 0.0168, "lr": 4.3265838048768334e-05, "epoch": 2.941901241588475, "percentage": 58.88, "elapsed_time": "6:58:11", "remaining_time": "4:52:05"}
{"current_steps": 1950, "total_steps": 3295, "loss": 0.0152, "lr": 4.274127331908915e-05, "epoch": 2.9570656809781064, "percentage": 59.18, "elapsed_time": "6:59:43", "remaining_time": "4:49:30"}
{"current_steps": 1950, "total_steps": 3295, "eval_loss": 0.03197691962122917, "epoch": 2.9570656809781064, "percentage": 59.18, "elapsed_time": "7:03:13", "remaining_time": "4:51:55"}
{"current_steps": 1960, "total_steps": 3295, "loss": 0.0152, "lr": 4.221752349393758e-05, "epoch": 2.9722301203677377, "percentage": 59.48, "elapsed_time": "7:04:48", "remaining_time": "4:49:20"}
{"current_steps": 1970, "total_steps": 3295, "loss": 0.0172, "lr": 4.1694647372345886e-05, "epoch": 2.987394559757369, "percentage": 59.79, "elapsed_time": "7:06:18", "remaining_time": "4:46:44"}
{"current_steps": 1980, "total_steps": 3295, "loss": 0.0181, "lr": 4.117270365525946e-05, "epoch": 3.0025589991470003, "percentage": 60.09, "elapsed_time": "7:07:50", "remaining_time": "4:44:08"}
{"current_steps": 1990, "total_steps": 3295, "loss": 0.0104, "lr": 4.065175093894694e-05, "epoch": 3.0177234385366316, "percentage": 60.39, "elapsed_time": "7:09:21", "remaining_time": "4:41:33"}
{"current_steps": 2000, "total_steps": 3295, "loss": 0.0136, "lr": 4.013184770842167e-05, "epoch": 3.032887877926263, "percentage": 60.7, "elapsed_time": "7:10:53", "remaining_time": "4:38:59"}
{"current_steps": 2000, "total_steps": 3295, "eval_loss": 0.03291192650794983, "epoch": 3.032887877926263, "percentage": 60.7, "elapsed_time": "7:14:23", "remaining_time": "4:41:16"}
{"current_steps": 2010, "total_steps": 3295, "loss": 0.0118, "lr": 3.9613052330876064e-05, "epoch": 3.048052317315894, "percentage": 61.0, "elapsed_time": "7:15:57", "remaining_time": "4:38:42"}
{"current_steps": 2020, "total_steps": 3295, "loss": 0.0137, "lr": 3.909542304912881e-05, "epoch": 3.0632167567055255, "percentage": 61.31, "elapsed_time": "7:17:28", "remaining_time": "4:36:07"}
{"current_steps": 2030, "total_steps": 3295, "loss": 0.0156, "lr": 3.857901797508628e-05, "epoch": 3.078381196095157, "percentage": 61.61, "elapsed_time": "7:18:59", "remaining_time": "4:33:33"}
{"current_steps": 2040, "total_steps": 3295, "loss": 0.0182, "lr": 3.80638950832186e-05, "epoch": 3.093545635484788, "percentage": 61.91, "elapsed_time": "7:20:31", "remaining_time": "4:31:00"}
{"current_steps": 2050, "total_steps": 3295, "loss": 0.009, "lr": 3.7550112204051014e-05, "epoch": 3.1087100748744194, "percentage": 62.22, "elapsed_time": "7:22:02", "remaining_time": "4:28:27"}
{"current_steps": 2050, "total_steps": 3295, "eval_loss": 0.03408639878034592, "epoch": 3.1087100748744194, "percentage": 62.22, "elapsed_time": "7:25:32", "remaining_time": "4:30:35"}
{"current_steps": 2060, "total_steps": 3295, "loss": 0.0073, "lr": 3.703772701767167e-05, "epoch": 3.1238745142640507, "percentage": 62.52, "elapsed_time": "7:27:07", "remaining_time": "4:28:03"}
{"current_steps": 2070, "total_steps": 3295, "loss": 0.007, "lr": 3.652679704725596e-05, "epoch": 3.139038953653682, "percentage": 62.82, "elapsed_time": "7:28:38", "remaining_time": "4:25:29"}
{"current_steps": 2080, "total_steps": 3295, "loss": 0.0079, "lr": 3.601737965260882e-05, "epoch": 3.1542033930433133, "percentage": 63.13, "elapsed_time": "7:30:09", "remaining_time": "4:22:57"}
{"current_steps": 2090, "total_steps": 3295, "loss": 0.0061, "lr": 3.550953202372503e-05, "epoch": 3.1693678324329446, "percentage": 63.43, "elapsed_time": "7:31:40", "remaining_time": "4:20:25"}
{"current_steps": 2100, "total_steps": 3295, "loss": 0.0089, "lr": 3.500331117436895e-05, "epoch": 3.184532271822576, "percentage": 63.73, "elapsed_time": "7:33:11", "remaining_time": "4:17:53"}
{"current_steps": 2100, "total_steps": 3295, "eval_loss": 0.03669964522123337, "epoch": 3.184532271822576, "percentage": 63.73, "elapsed_time": "7:36:42", "remaining_time": "4:19:53"}
{"current_steps": 2110, "total_steps": 3295, "loss": 0.0101, "lr": 3.4498773935673626e-05, "epoch": 3.1996967112122072, "percentage": 64.04, "elapsed_time": "7:38:17", "remaining_time": "4:17:22"}
{"current_steps": 2120, "total_steps": 3295, "loss": 0.0154, "lr": 3.399597694976081e-05, "epoch": 3.2148611506018385, "percentage": 64.34, "elapsed_time": "7:39:48", "remaining_time": "4:14:50"}
{"current_steps": 2130, "total_steps": 3295, "loss": 0.0109, "lr": 3.349497666338187e-05, "epoch": 3.23002558999147, "percentage": 64.64, "elapsed_time": "7:41:19", "remaining_time": "4:12:19"}
{"current_steps": 2140, "total_steps": 3295, "loss": 0.0107, "lr": 3.299582932158085e-05, "epoch": 3.245190029381101, "percentage": 64.95, "elapsed_time": "7:42:51", "remaining_time": "4:09:48"}
{"current_steps": 2150, "total_steps": 3295, "loss": 0.0127, "lr": 3.2498590961379996e-05, "epoch": 3.260354468770733, "percentage": 65.25, "elapsed_time": "7:44:22", "remaining_time": "4:07:18"}
{"current_steps": 2150, "total_steps": 3295, "eval_loss": 0.0364246740937233, "epoch": 3.260354468770733, "percentage": 65.25, "elapsed_time": "7:47:52", "remaining_time": "4:09:10"}
{"current_steps": 2160, "total_steps": 3295, "loss": 0.0121, "lr": 3.200331740548887e-05, "epoch": 3.2755189081603637, "percentage": 65.55, "elapsed_time": "7:49:26", "remaining_time": "4:06:40"}
{"current_steps": 2170, "total_steps": 3295, "loss": 0.0125, "lr": 3.1510064256037274e-05, "epoch": 3.2906833475499955, "percentage": 65.86, "elapsed_time": "7:50:57", "remaining_time": "4:04:09"}
{"current_steps": 2180, "total_steps": 3295, "loss": 0.0088, "lr": 3.1018886888333065e-05, "epoch": 3.3058477869396263, "percentage": 66.16, "elapsed_time": "7:52:28", "remaining_time": "4:01:39"}
{"current_steps": 2190, "total_steps": 3295, "loss": 0.0109, "lr": 3.052984044464548e-05, "epoch": 3.321012226329258, "percentage": 66.46, "elapsed_time": "7:53:59", "remaining_time": "3:59:09"}
{"current_steps": 2200, "total_steps": 3295, "loss": 0.0119, "lr": 3.0042979828014496e-05, "epoch": 3.3361766657188894, "percentage": 66.77, "elapsed_time": "7:55:31", "remaining_time": "3:56:40"}
{"current_steps": 2200, "total_steps": 3295, "eval_loss": 0.03581894189119339, "epoch": 3.3361766657188894, "percentage": 66.77, "elapsed_time": "7:59:01", "remaining_time": "3:58:25"}
{"current_steps": 2210, "total_steps": 3295, "loss": 0.0134, "lr": 2.9558359696087152e-05, "epoch": 3.3513411051085207, "percentage": 67.07, "elapsed_time": "8:00:36", "remaining_time": "3:55:57"}
{"current_steps": 2220, "total_steps": 3295, "loss": 0.0116, "lr": 2.9076034454981332e-05, "epoch": 3.366505544498152, "percentage": 67.37, "elapsed_time": "8:02:07", "remaining_time": "3:53:27"}
{"current_steps": 2230, "total_steps": 3295, "loss": 0.0105, "lr": 2.8596058253177932e-05, "epoch": 3.3816699838877833, "percentage": 67.68, "elapsed_time": "8:03:37", "remaining_time": "3:50:58"}
{"current_steps": 2240, "total_steps": 3295, "loss": 0.0119, "lr": 2.811848497544175e-05, "epoch": 3.3968344232774146, "percentage": 67.98, "elapsed_time": "8:05:08", "remaining_time": "3:48:29"}
{"current_steps": 2250, "total_steps": 3295, "loss": 0.0118, "lr": 2.764336823677216e-05, "epoch": 3.411998862667046, "percentage": 68.29, "elapsed_time": "8:06:39", "remaining_time": "3:46:01"}
{"current_steps": 2250, "total_steps": 3295, "eval_loss": 0.0358290821313858, "epoch": 3.411998862667046, "percentage": 68.29, "elapsed_time": "8:10:10", "remaining_time": "3:47:39"}
{"current_steps": 2260, "total_steps": 3295, "loss": 0.0116, "lr": 2.717076137638388e-05, "epoch": 3.427163302056677, "percentage": 68.59, "elapsed_time": "8:11:45", "remaining_time": "3:45:12"}
{"current_steps": 2270, "total_steps": 3295, "loss": 0.011, "lr": 2.6700717451719066e-05, "epoch": 3.4423277414463085, "percentage": 68.89, "elapsed_time": "8:13:17", "remaining_time": "3:42:44"}
{"current_steps": 2280, "total_steps": 3295, "loss": 0.0121, "lr": 2.6233289232490447e-05, "epoch": 3.45749218083594, "percentage": 69.2, "elapsed_time": "8:14:48", "remaining_time": "3:40:16"}
{"current_steps": 2290, "total_steps": 3295, "loss": 0.0077, "lr": 2.5768529194757474e-05, "epoch": 3.472656620225571, "percentage": 69.5, "elapsed_time": "8:16:20", "remaining_time": "3:37:49"}
{"current_steps": 2300, "total_steps": 3295, "loss": 0.0084, "lr": 2.5306489515034713e-05, "epoch": 3.4878210596152024, "percentage": 69.8, "elapsed_time": "8:17:52", "remaining_time": "3:35:23"}
{"current_steps": 2300, "total_steps": 3295, "eval_loss": 0.0370686799287796, "epoch": 3.4878210596152024, "percentage": 69.8, "elapsed_time": "8:21:23", "remaining_time": "3:36:54"}
{"current_steps": 2310, "total_steps": 3295, "loss": 0.016, "lr": 2.484722206443455e-05, "epoch": 3.5029854990048337, "percentage": 70.11, "elapsed_time": "8:22:57", "remaining_time": "3:34:27"}
{"current_steps": 2320, "total_steps": 3295, "loss": 0.0107, "lr": 2.4390778402843605e-05, "epoch": 3.518149938394465, "percentage": 70.41, "elapsed_time": "8:24:28", "remaining_time": "3:32:00"}
{"current_steps": 2330, "total_steps": 3295, "loss": 0.0092, "lr": 2.3937209773134472e-05, "epoch": 3.5333143777840963, "percentage": 70.71, "elapsed_time": "8:25:59", "remaining_time": "3:29:33"}
{"current_steps": 2340, "total_steps": 3295, "loss": 0.0096, "lr": 2.3486567095412864e-05, "epoch": 3.5484788171737276, "percentage": 71.02, "elapsed_time": "8:27:31", "remaining_time": "3:27:07"}
{"current_steps": 2350, "total_steps": 3295, "loss": 0.0137, "lr": 2.3038900961301053e-05, "epoch": 3.563643256563359, "percentage": 71.32, "elapsed_time": "8:29:03", "remaining_time": "3:24:42"}
{"current_steps": 2350, "total_steps": 3295, "eval_loss": 0.03648369759321213, "epoch": 3.563643256563359, "percentage": 71.32, "elapsed_time": "8:32:33", "remaining_time": "3:26:06"}
{"current_steps": 2360, "total_steps": 3295, "loss": 0.0082, "lr": 2.259426162825819e-05, "epoch": 3.57880769595299, "percentage": 71.62, "elapsed_time": "8:34:09", "remaining_time": "3:23:42"}
{"current_steps": 2370, "total_steps": 3295, "loss": 0.0099, "lr": 2.215269901393805e-05, "epoch": 3.5939721353426215, "percentage": 71.93, "elapsed_time": "8:35:40", "remaining_time": "3:21:15"}
{"current_steps": 2380, "total_steps": 3295, "loss": 0.0079, "lr": 2.1714262690585192e-05, "epoch": 3.609136574732253, "percentage": 72.23, "elapsed_time": "8:37:12", "remaining_time": "3:18:50"}
{"current_steps": 2390, "total_steps": 3295, "loss": 0.0082, "lr": 2.1279001879469424e-05, "epoch": 3.624301014121884, "percentage": 72.53, "elapsed_time": "8:38:44", "remaining_time": "3:16:25"}
{"current_steps": 2400, "total_steps": 3295, "loss": 0.0093, "lr": 2.084696544536025e-05, "epoch": 3.6394654535115154, "percentage": 72.84, "elapsed_time": "8:40:15", "remaining_time": "3:14:00"}
{"current_steps": 2400, "total_steps": 3295, "eval_loss": 0.036392148584127426, "epoch": 3.6394654535115154, "percentage": 72.84, "elapsed_time": "8:43:46", "remaining_time": "3:15:19"}
{"current_steps": 2410, "total_steps": 3295, "loss": 0.0105, "lr": 2.0418201891040778e-05, "epoch": 3.6546298929011467, "percentage": 73.14, "elapsed_time": "8:45:20", "remaining_time": "3:12:55"}
{"current_steps": 2420, "total_steps": 3295, "loss": 0.0164, "lr": 1.9992759351862772e-05, "epoch": 3.669794332290778, "percentage": 73.44, "elapsed_time": "8:46:52", "remaining_time": "3:10:30"}
{"current_steps": 2430, "total_steps": 3295, "loss": 0.0073, "lr": 1.9570685590342463e-05, "epoch": 3.6849587716804093, "percentage": 73.75, "elapsed_time": "8:48:24", "remaining_time": "3:08:05"}
{"current_steps": 2440, "total_steps": 3295, "loss": 0.0109, "lr": 1.9152027990798748e-05, "epoch": 3.7001232110700406, "percentage": 74.05, "elapsed_time": "8:49:55", "remaining_time": "3:05:41"}
{"current_steps": 2450, "total_steps": 3295, "loss": 0.0095, "lr": 1.8736833554033263e-05, "epoch": 3.715287650459672, "percentage": 74.36, "elapsed_time": "8:51:27", "remaining_time": "3:03:17"}
{"current_steps": 2450, "total_steps": 3295, "eval_loss": 0.0359375923871994, "epoch": 3.715287650459672, "percentage": 74.36, "elapsed_time": "8:54:57", "remaining_time": "3:04:30"}
{"current_steps": 2460, "total_steps": 3295, "loss": 0.0097, "lr": 1.8325148892054093e-05, "epoch": 3.7304520898493037, "percentage": 74.66, "elapsed_time": "8:56:33", "remaining_time": "3:02:07"}
{"current_steps": 2470, "total_steps": 3295, "loss": 0.0109, "lr": 1.7917020222842662e-05, "epoch": 3.7456165292389345, "percentage": 74.96, "elapsed_time": "8:58:04", "remaining_time": "2:59:43"}
{"current_steps": 2480, "total_steps": 3295, "loss": 0.01, "lr": 1.751249336516513e-05, "epoch": 3.7607809686285663, "percentage": 75.27, "elapsed_time": "8:59:36", "remaining_time": "2:57:19"}
{"current_steps": 2490, "total_steps": 3295, "loss": 0.009, "lr": 1.7111613733428522e-05, "epoch": 3.775945408018197, "percentage": 75.57, "elapsed_time": "9:01:07", "remaining_time": "2:54:56"}
{"current_steps": 2500, "total_steps": 3295, "loss": 0.0102, "lr": 1.671442633258218e-05, "epoch": 3.791109847407829, "percentage": 75.87, "elapsed_time": "9:02:38", "remaining_time": "2:52:33"}
{"current_steps": 2500, "total_steps": 3295, "eval_loss": 0.03812110051512718, "epoch": 3.791109847407829, "percentage": 75.87, "elapsed_time": "9:06:09", "remaining_time": "2:53:40"}
{"current_steps": 2510, "total_steps": 3295, "loss": 0.0085, "lr": 1.6320975753065453e-05, "epoch": 3.8062742867974597, "percentage": 76.18, "elapsed_time": "9:07:44", "remaining_time": "2:51:18"}
{"current_steps": 2520, "total_steps": 3295, "loss": 0.0063, "lr": 1.5931306165801452e-05, "epoch": 3.8214387261870915, "percentage": 76.48, "elapsed_time": "9:09:16", "remaining_time": "2:48:55"}
{"current_steps": 2530, "total_steps": 3295, "loss": 0.0103, "lr": 1.554546131723848e-05, "epoch": 3.8366031655767223, "percentage": 76.78, "elapsed_time": "9:10:47", "remaining_time": "2:46:32"}
{"current_steps": 2540, "total_steps": 3295, "loss": 0.009, "lr": 1.5163484524438516e-05, "epoch": 3.851767604966354, "percentage": 77.09, "elapsed_time": "9:12:18", "remaining_time": "2:44:10"}
{"current_steps": 2550, "total_steps": 3295, "loss": 0.0137, "lr": 1.4785418670214496e-05, "epoch": 3.8669320443559854, "percentage": 77.39, "elapsed_time": "9:13:50", "remaining_time": "2:41:48"}
{"current_steps": 2550, "total_steps": 3295, "eval_loss": 0.03836597874760628, "epoch": 3.8669320443559854, "percentage": 77.39, "elapsed_time": "9:17:21", "remaining_time": "2:42:50"}
{"current_steps": 2560, "total_steps": 3295, "loss": 0.0075, "lr": 1.4411306198315777e-05, "epoch": 3.8820964837456167, "percentage": 77.69, "elapsed_time": "9:18:55", "remaining_time": "2:40:28"}
{"current_steps": 2570, "total_steps": 3295, "loss": 0.0106, "lr": 1.4041189108663421e-05, "epoch": 3.897260923135248, "percentage": 78.0, "elapsed_time": "9:20:27", "remaining_time": "2:38:06"}
{"current_steps": 2580, "total_steps": 3295, "loss": 0.0082, "lr": 1.3675108952634824e-05, "epoch": 3.9124253625248793, "percentage": 78.3, "elapsed_time": "9:21:58", "remaining_time": "2:35:44"}
{"current_steps": 2590, "total_steps": 3295, "loss": 0.0091, "lr": 1.3313106828399147e-05, "epoch": 3.9275898019145106, "percentage": 78.6, "elapsed_time": "9:23:29", "remaining_time": "2:33:23"}
{"current_steps": 2600, "total_steps": 3295, "loss": 0.01, "lr": 1.2955223376303205e-05, "epoch": 3.942754241304142, "percentage": 78.91, "elapsed_time": "9:25:01", "remaining_time": "2:31:02"}
{"current_steps": 2600, "total_steps": 3295, "eval_loss": 0.03762030228972435, "epoch": 3.942754241304142, "percentage": 78.91, "elapsed_time": "9:28:32", "remaining_time": "2:31:58"}
{"current_steps": 2610, "total_steps": 3295, "loss": 0.0098, "lr": 1.2601498774309112e-05, "epoch": 3.957918680693773, "percentage": 79.21, "elapsed_time": "9:30:06", "remaining_time": "2:29:37"}
{"current_steps": 2620, "total_steps": 3295, "loss": 0.011, "lr": 1.2251972733483612e-05, "epoch": 3.9730831200834045, "percentage": 79.51, "elapsed_time": "9:31:37", "remaining_time": "2:27:16"}
{"current_steps": 2630, "total_steps": 3295, "loss": 0.0112, "lr": 1.19066844935399e-05, "epoch": 3.9882475594730358, "percentage": 79.82, "elapsed_time": "9:33:09", "remaining_time": "2:24:55"}
{"current_steps": 2640, "total_steps": 3295, "loss": 0.01, "lr": 1.156567281843241e-05, "epoch": 4.003411998862667, "percentage": 80.12, "elapsed_time": "9:34:41", "remaining_time": "2:22:34"}
{"current_steps": 2650, "total_steps": 3295, "loss": 0.0058, "lr": 1.1228975992004842e-05, "epoch": 4.018576438252298, "percentage": 80.42, "elapsed_time": "9:36:12", "remaining_time": "2:20:14"}
{"current_steps": 2650, "total_steps": 3295, "eval_loss": 0.0389280766248703, "epoch": 4.018576438252298, "percentage": 80.42, "elapsed_time": "9:39:43", "remaining_time": "2:21:06"}
{"current_steps": 2660, "total_steps": 3295, "loss": 0.0074, "lr": 1.0896631813692377e-05, "epoch": 4.03374087764193, "percentage": 80.73, "elapsed_time": "9:41:18", "remaining_time": "2:18:46"}
{"current_steps": 2670, "total_steps": 3295, "loss": 0.0068, "lr": 1.0568677594277899e-05, "epoch": 4.048905317031561, "percentage": 81.03, "elapsed_time": "9:42:49", "remaining_time": "2:16:25"}
{"current_steps": 2680, "total_steps": 3295, "loss": 0.004, "lr": 1.0245150151703509e-05, "epoch": 4.064069756421192, "percentage": 81.34, "elapsed_time": "9:44:21", "remaining_time": "2:14:05"}
{"current_steps": 2690, "total_steps": 3295, "loss": 0.0054, "lr": 9.926085806936918e-06, "epoch": 4.079234195810824, "percentage": 81.64, "elapsed_time": "9:45:53", "remaining_time": "2:11:46"}
{"current_steps": 2700, "total_steps": 3295, "loss": 0.0071, "lr": 9.611520379894068e-06, "epoch": 4.094398635200455, "percentage": 81.94, "elapsed_time": "9:47:25", "remaining_time": "2:09:26"}
{"current_steps": 2700, "total_steps": 3295, "eval_loss": 0.040721941739320755, "epoch": 4.094398635200455, "percentage": 81.94, "elapsed_time": "9:50:55", "remaining_time": "2:10:13"}
{"current_steps": 2710, "total_steps": 3295, "loss": 0.0051, "lr": 9.301489185417578e-06, "epoch": 4.109563074590087, "percentage": 82.25, "elapsed_time": "9:52:30", "remaining_time": "2:07:54"}
{"current_steps": 2720, "total_steps": 3295, "loss": 0.0037, "lr": 8.996027029312314e-06, "epoch": 4.1247275139797175, "percentage": 82.55, "elapsed_time": "9:54:01", "remaining_time": "2:05:34"}
{"current_steps": 2730, "total_steps": 3295, "loss": 0.0096, "lr": 8.695168204437721e-06, "epoch": 4.139891953369349, "percentage": 82.85, "elapsed_time": "9:55:33", "remaining_time": "2:03:15"}
{"current_steps": 2740, "total_steps": 3295, "loss": 0.005, "lr": 8.398946486858029e-06, "epoch": 4.15505639275898, "percentage": 83.16, "elapsed_time": "9:57:05", "remaining_time": "2:00:56"}
{"current_steps": 2750, "total_steps": 3295, "loss": 0.0051, "lr": 8.107395132050333e-06, "epoch": 4.170220832148612, "percentage": 83.46, "elapsed_time": "9:58:36", "remaining_time": "1:58:38"}
{"current_steps": 2750, "total_steps": 3295, "eval_loss": 0.04139946773648262, "epoch": 4.170220832148612, "percentage": 83.46, "elapsed_time": "10:02:07", "remaining_time": "1:59:19"}
{"current_steps": 2760, "total_steps": 3295, "loss": 0.0047, "lr": 7.820546871171142e-06, "epoch": 4.185385271538243, "percentage": 83.76, "elapsed_time": "10:03:42", "remaining_time": "1:57:01"}
{"current_steps": 2770, "total_steps": 3295, "loss": 0.0039, "lr": 7.5384339073818145e-06, "epoch": 4.200549710927874, "percentage": 84.07, "elapsed_time": "10:05:13", "remaining_time": "1:54:42"}
{"current_steps": 2780, "total_steps": 3295, "loss": 0.0084, "lr": 7.261087912233228e-06, "epoch": 4.215714150317505, "percentage": 84.37, "elapsed_time": "10:06:45", "remaining_time": "1:52:24"}
{"current_steps": 2790, "total_steps": 3295, "loss": 0.0048, "lr": 6.988540022110235e-06, "epoch": 4.230878589707137, "percentage": 84.67, "elapsed_time": "10:08:16", "remaining_time": "1:50:05"}
{"current_steps": 2800, "total_steps": 3295, "loss": 0.0077, "lr": 6.720820834735969e-06, "epoch": 4.246043029096768, "percentage": 84.98, "elapsed_time": "10:09:47", "remaining_time": "1:47:48"}
{"current_steps": 2800, "total_steps": 3295, "eval_loss": 0.041931215673685074, "epoch": 4.246043029096768, "percentage": 84.98, "elapsed_time": "10:13:18", "remaining_time": "1:48:25"}
{"current_steps": 2810, "total_steps": 3295, "loss": 0.0073, "lr": 6.457960405736968e-06, "epoch": 4.2612074684864, "percentage": 85.28, "elapsed_time": "10:14:53", "remaining_time": "1:46:07"}
{"current_steps": 2820, "total_steps": 3295, "loss": 0.0031, "lr": 6.199988245268778e-06, "epoch": 4.2763719078760305, "percentage": 85.58, "elapsed_time": "10:16:24", "remaining_time": "1:43:49"}
{"current_steps": 2830, "total_steps": 3295, "loss": 0.0046, "lr": 5.946933314703157e-06, "epoch": 4.291536347265662, "percentage": 85.89, "elapsed_time": "10:17:56", "remaining_time": "1:41:32"}
{"current_steps": 2840, "total_steps": 3295, "loss": 0.0047, "lr": 5.698824023376531e-06, "epoch": 4.306700786655293, "percentage": 86.19, "elapsed_time": "10:19:27", "remaining_time": "1:39:14"}
{"current_steps": 2850, "total_steps": 3295, "loss": 0.0039, "lr": 5.455688225400802e-06, "epoch": 4.321865226044925, "percentage": 86.49, "elapsed_time": "10:20:58", "remaining_time": "1:36:57"}
{"current_steps": 2850, "total_steps": 3295, "eval_loss": 0.042382657527923584, "epoch": 4.321865226044925, "percentage": 86.49, "elapsed_time": "10:24:30", "remaining_time": "1:37:30"}
{"current_steps": 2860, "total_steps": 3295, "loss": 0.008, "lr": 5.217553216536098e-06, "epoch": 4.337029665434556, "percentage": 86.8, "elapsed_time": "10:26:05", "remaining_time": "1:35:13"}
{"current_steps": 2870, "total_steps": 3295, "loss": 0.0085, "lr": 4.984445731126542e-06, "epoch": 4.3521941048241874, "percentage": 87.1, "elapsed_time": "10:27:36", "remaining_time": "1:32:56"}
{"current_steps": 2880, "total_steps": 3295, "loss": 0.0038, "lr": 4.756391939098853e-06, "epoch": 4.367358544213818, "percentage": 87.41, "elapsed_time": "10:29:08", "remaining_time": "1:30:39"}
{"current_steps": 2890, "total_steps": 3295, "loss": 0.0069, "lr": 4.533417443024374e-06, "epoch": 4.38252298360345, "percentage": 87.71, "elapsed_time": "10:30:39", "remaining_time": "1:28:22"}
{"current_steps": 2900, "total_steps": 3295, "loss": 0.002, "lr": 4.315547275244769e-06, "epoch": 4.397687422993081, "percentage": 88.01, "elapsed_time": "10:32:11", "remaining_time": "1:26:06"}
{"current_steps": 2900, "total_steps": 3295, "eval_loss": 0.04292920231819153, "epoch": 4.397687422993081, "percentage": 88.01, "elapsed_time": "10:35:42", "remaining_time": "1:26:35"}
{"current_steps": 2910, "total_steps": 3295, "loss": 0.006, "lr": 4.10280589506179e-06, "epoch": 4.412851862382713, "percentage": 88.32, "elapsed_time": "10:37:17", "remaining_time": "1:24:18"}
{"current_steps": 2920, "total_steps": 3295, "loss": 0.0046, "lr": 3.895217185991312e-06, "epoch": 4.4280163017723435, "percentage": 88.62, "elapsed_time": "10:38:48", "remaining_time": "1:22:02"}
{"current_steps": 2930, "total_steps": 3295, "loss": 0.0048, "lr": 3.692804453082038e-06, "epoch": 4.443180741161975, "percentage": 88.92, "elapsed_time": "10:40:20", "remaining_time": "1:19:46"}
{"current_steps": 2940, "total_steps": 3295, "loss": 0.0046, "lr": 3.495590420299194e-06, "epoch": 4.458345180551606, "percentage": 89.23, "elapsed_time": "10:41:52", "remaining_time": "1:17:30"}
{"current_steps": 2950, "total_steps": 3295, "loss": 0.0036, "lr": 3.3035972279733207e-06, "epoch": 4.473509619941238, "percentage": 89.53, "elapsed_time": "10:43:24", "remaining_time": "1:15:14"}
{"current_steps": 2950, "total_steps": 3295, "eval_loss": 0.0432889424264431, "epoch": 4.473509619941238, "percentage": 89.53, "elapsed_time": "10:46:55", "remaining_time": "1:15:39"}
{"current_steps": 2960, "total_steps": 3295, "loss": 0.0062, "lr": 3.116846430314796e-06, "epoch": 4.488674059330869, "percentage": 89.83, "elapsed_time": "10:48:29", "remaining_time": "1:13:23"}
{"current_steps": 2970, "total_steps": 3295, "loss": 0.0037, "lr": 2.9353589929939087e-06, "epoch": 4.5038384987205005, "percentage": 90.14, "elapsed_time": "10:50:01", "remaining_time": "1:11:07"}
{"current_steps": 2980, "total_steps": 3295, "loss": 0.0036, "lr": 2.7591552907872574e-06, "epoch": 4.519002938110132, "percentage": 90.44, "elapsed_time": "10:51:33", "remaining_time": "1:08:52"}
{"current_steps": 2990, "total_steps": 3295, "loss": 0.0081, "lr": 2.5882551052902883e-06, "epoch": 4.534167377499763, "percentage": 90.74, "elapsed_time": "10:53:05", "remaining_time": "1:06:37"}
{"current_steps": 3000, "total_steps": 3295, "loss": 0.0083, "lr": 2.4226776226965453e-06, "epoch": 4.549331816889394, "percentage": 91.05, "elapsed_time": "10:54:37", "remaining_time": "1:04:22"}
{"current_steps": 3000, "total_steps": 3295, "eval_loss": 0.043931957334280014, "epoch": 4.549331816889394, "percentage": 91.05, "elapsed_time": "10:58:07", "remaining_time": "1:04:42"}
{"current_steps": 3010, "total_steps": 3295, "loss": 0.0064, "lr": 2.262441431643697e-06, "epoch": 4.564496256279026, "percentage": 91.35, "elapsed_time": "10:59:43", "remaining_time": "1:02:27"}
{"current_steps": 3020, "total_steps": 3295, "loss": 0.0025, "lr": 2.1075645211266927e-06, "epoch": 4.579660695668657, "percentage": 91.65, "elapsed_time": "11:01:14", "remaining_time": "1:00:12"}
{"current_steps": 3030, "total_steps": 3295, "loss": 0.0056, "lr": 1.9580642784782053e-06, "epoch": 4.594825135058288, "percentage": 91.96, "elapsed_time": "11:02:46", "remaining_time": "0:57:57"}
{"current_steps": 3040, "total_steps": 3295, "loss": 0.003, "lr": 1.813957487416651e-06, "epoch": 4.60998957444792, "percentage": 92.26, "elapsed_time": "11:04:17", "remaining_time": "0:55:43"}
{"current_steps": 3050, "total_steps": 3295, "loss": 0.0049, "lr": 1.6752603261619315e-06, "epoch": 4.625154013837551, "percentage": 92.56, "elapsed_time": "11:05:49", "remaining_time": "0:53:29"}
{"current_steps": 3050, "total_steps": 3295, "eval_loss": 0.043927934020757675, "epoch": 4.625154013837551, "percentage": 92.56, "elapsed_time": "11:09:20", "remaining_time": "0:53:46"}
{"current_steps": 3060, "total_steps": 3295, "loss": 0.0065, "lr": 1.541988365619207e-06, "epoch": 4.640318453227183, "percentage": 92.87, "elapsed_time": "11:10:56", "remaining_time": "0:51:31"}
{"current_steps": 3070, "total_steps": 3295, "loss": 0.0047, "lr": 1.4141565676307865e-06, "epoch": 4.6554828926168135, "percentage": 93.17, "elapsed_time": "11:12:28", "remaining_time": "0:49:17"}
{"current_steps": 3080, "total_steps": 3295, "loss": 0.0026, "lr": 1.291779283296468e-06, "epoch": 4.670647332006445, "percentage": 93.47, "elapsed_time": "11:13:59", "remaining_time": "0:47:02"}
{"current_steps": 3090, "total_steps": 3295, "loss": 0.0042, "lr": 1.1748702513623922e-06, "epoch": 4.685811771396076, "percentage": 93.78, "elapsed_time": "11:15:31", "remaining_time": "0:44:48"}
{"current_steps": 3100, "total_steps": 3295, "loss": 0.0033, "lr": 1.0634425966786155e-06, "epoch": 4.700976210785708, "percentage": 94.08, "elapsed_time": "11:17:02", "remaining_time": "0:42:35"}
{"current_steps": 3100, "total_steps": 3295, "eval_loss": 0.04425197094678879, "epoch": 4.700976210785708, "percentage": 94.08, "elapsed_time": "11:20:33", "remaining_time": "0:42:48"}
{"current_steps": 3110, "total_steps": 3295, "loss": 0.0045, "lr": 9.575088287257118e-07, "epoch": 4.716140650175339, "percentage": 94.39, "elapsed_time": "11:22:08", "remaining_time": "0:40:34"}
{"current_steps": 3120, "total_steps": 3295, "loss": 0.0057, "lr": 8.570808402103392e-07, "epoch": 4.73130508956497, "percentage": 94.69, "elapsed_time": "11:23:40", "remaining_time": "0:38:20"}
{"current_steps": 3130, "total_steps": 3295, "loss": 0.0032, "lr": 7.62169905730109e-07, "epoch": 4.746469528954601, "percentage": 94.99, "elapsed_time": "11:25:11", "remaining_time": "0:36:07"}
{"current_steps": 3140, "total_steps": 3295, "loss": 0.0035, "lr": 6.727866805078531e-07, "epoch": 4.761633968344233, "percentage": 95.3, "elapsed_time": "11:26:43", "remaining_time": "0:33:53"}
{"current_steps": 3150, "total_steps": 3295, "loss": 0.0034, "lr": 5.889411991953975e-07, "epoch": 4.776798407733864, "percentage": 95.6, "elapsed_time": "11:28:15", "remaining_time": "0:31:40"}
{"current_steps": 3150, "total_steps": 3295, "eval_loss": 0.044278133660554886, "epoch": 4.776798407733864, "percentage": 95.6, "elapsed_time": "11:31:46", "remaining_time": "0:31:50"}
{"current_steps": 3160, "total_steps": 3295, "loss": 0.0043, "lr": 5.106428747470137e-07, "epoch": 4.791962847123496, "percentage": 95.9, "elapsed_time": "11:33:21", "remaining_time": "0:29:37"}
{"current_steps": 3170, "total_steps": 3295, "loss": 0.0049, "lr": 4.37900497362681e-07, "epoch": 4.8071272865131265, "percentage": 96.21, "elapsed_time": "11:34:53", "remaining_time": "0:27:24"}
{"current_steps": 3180, "total_steps": 3295, "loss": 0.0036, "lr": 3.7072223350124167e-07, "epoch": 4.822291725902758, "percentage": 96.51, "elapsed_time": "11:36:25", "remaining_time": "0:25:11"}
{"current_steps": 3190, "total_steps": 3295, "loss": 0.0055, "lr": 3.0911562496358517e-07, "epoch": 4.837456165292389, "percentage": 96.81, "elapsed_time": "11:37:57", "remaining_time": "0:22:58"}
{"current_steps": 3200, "total_steps": 3295, "loss": 0.0038, "lr": 2.53087588045986e-07, "epoch": 4.852620604682021, "percentage": 97.12, "elapsed_time": "11:39:29", "remaining_time": "0:20:45"}
{"current_steps": 3200, "total_steps": 3295, "eval_loss": 0.04416884854435921, "epoch": 4.852620604682021, "percentage": 97.12, "elapsed_time": "11:43:01", "remaining_time": "0:20:52"}
{"current_steps": 3210, "total_steps": 3295, "loss": 0.0034, "lr": 2.026444127636029e-07, "epoch": 4.867785044071652, "percentage": 97.42, "elapsed_time": "11:44:35", "remaining_time": "0:18:39"}
{"current_steps": 3220, "total_steps": 3295, "loss": 0.004, "lr": 1.577917621443825e-07, "epoch": 4.882949483461283, "percentage": 97.72, "elapsed_time": "11:46:07", "remaining_time": "0:16:26"}
{"current_steps": 3230, "total_steps": 3295, "loss": 0.0064, "lr": 1.185346715932345e-07, "epoch": 4.898113922850914, "percentage": 98.03, "elapsed_time": "11:47:39", "remaining_time": "0:14:14"}
{"current_steps": 3240, "total_steps": 3295, "loss": 0.0052, "lr": 8.48775483267783e-08, "epoch": 4.913278362240546, "percentage": 98.33, "elapsed_time": "11:49:11", "remaining_time": "0:12:02"}
{"current_steps": 3250, "total_steps": 3295, "loss": 0.0029, "lr": 5.6824170878544414e-08, "epoch": 4.928442801630177, "percentage": 98.63, "elapsed_time": "11:50:43", "remaining_time": "0:09:50"}
{"current_steps": 3250, "total_steps": 3295, "eval_loss": 0.04429050534963608, "epoch": 4.928442801630177, "percentage": 98.63, "elapsed_time": "11:54:14", "remaining_time": "0:09:53"}
{"current_steps": 3260, "total_steps": 3295, "loss": 0.0034, "lr": 3.4377688674774866e-08, "epoch": 4.943607241019809, "percentage": 98.94, "elapsed_time": "11:55:48", "remaining_time": "0:07:41"}
{"current_steps": 3270, "total_steps": 3295, "loss": 0.0059, "lr": 1.7540621680850466e-08, "epoch": 4.9587716804094395, "percentage": 99.24, "elapsed_time": "11:57:20", "remaining_time": "0:05:29"}
{"current_steps": 3280, "total_steps": 3295, "loss": 0.0044, "lr": 6.314860118400434e-09, "epoch": 4.973936119799071, "percentage": 99.54, "elapsed_time": "11:58:52", "remaining_time": "0:03:17"}
{"current_steps": 3290, "total_steps": 3295, "loss": 0.0055, "lr": 7.016642530777162e-10, "epoch": 4.989100559188703, "percentage": 99.85, "elapsed_time": "12:00:24", "remaining_time": "0:01:05"}
{"current_steps": 3295, "total_steps": 3295, "epoch": 4.996682778883518, "percentage": 100.0, "elapsed_time": "12:01:14", "remaining_time": "0:00:00"}