chchen's picture
Training in progress, step 3325
3fbacf3 verified
{"current_steps": 10, "total_steps": 3325, "loss": 1.6617, "lr": 3.003003003003003e-06, "epoch": 0.015023474178403756, "percentage": 0.3, "elapsed_time": "0:01:26", "remaining_time": "7:56:05"}
{"current_steps": 20, "total_steps": 3325, "loss": 1.7717, "lr": 6.006006006006006e-06, "epoch": 0.03004694835680751, "percentage": 0.6, "elapsed_time": "0:02:52", "remaining_time": "7:54:06"}
{"current_steps": 30, "total_steps": 3325, "loss": 1.6755, "lr": 9.00900900900901e-06, "epoch": 0.04507042253521127, "percentage": 0.9, "elapsed_time": "0:04:19", "remaining_time": "7:55:09"}
{"current_steps": 40, "total_steps": 3325, "loss": 1.3133, "lr": 1.2012012012012012e-05, "epoch": 0.06009389671361502, "percentage": 1.2, "elapsed_time": "0:05:47", "remaining_time": "7:56:05"}
{"current_steps": 50, "total_steps": 3325, "loss": 0.8292, "lr": 1.5015015015015016e-05, "epoch": 0.07511737089201878, "percentage": 1.5, "elapsed_time": "0:07:16", "remaining_time": "7:56:38"}
{"current_steps": 50, "total_steps": 3325, "eval_loss": 0.6473308205604553, "epoch": 0.07511737089201878, "percentage": 1.5, "elapsed_time": "0:10:43", "remaining_time": "11:42:42"}
{"current_steps": 60, "total_steps": 3325, "loss": 0.6816, "lr": 1.801801801801802e-05, "epoch": 0.09014084507042254, "percentage": 1.8, "elapsed_time": "0:12:15", "remaining_time": "11:07:19"}
{"current_steps": 70, "total_steps": 3325, "loss": 0.4595, "lr": 2.102102102102102e-05, "epoch": 0.10516431924882629, "percentage": 2.11, "elapsed_time": "0:13:44", "remaining_time": "10:39:19"}
{"current_steps": 80, "total_steps": 3325, "loss": 0.2868, "lr": 2.4024024024024024e-05, "epoch": 0.12018779342723004, "percentage": 2.41, "elapsed_time": "0:15:14", "remaining_time": "10:18:12"}
{"current_steps": 90, "total_steps": 3325, "loss": 0.1691, "lr": 2.702702702702703e-05, "epoch": 0.1352112676056338, "percentage": 2.71, "elapsed_time": "0:16:43", "remaining_time": "10:01:25"}
{"current_steps": 100, "total_steps": 3325, "loss": 0.1595, "lr": 3.0030030030030033e-05, "epoch": 0.15023474178403756, "percentage": 3.01, "elapsed_time": "0:18:13", "remaining_time": "9:47:39"}
{"current_steps": 100, "total_steps": 3325, "eval_loss": 0.11690635979175568, "epoch": 0.15023474178403756, "percentage": 3.01, "elapsed_time": "0:21:41", "remaining_time": "11:39:29"}
{"current_steps": 110, "total_steps": 3325, "loss": 0.1358, "lr": 3.3033033033033035e-05, "epoch": 0.1652582159624413, "percentage": 3.31, "elapsed_time": "0:23:13", "remaining_time": "11:19:02"}
{"current_steps": 120, "total_steps": 3325, "loss": 0.1063, "lr": 3.603603603603604e-05, "epoch": 0.18028169014084508, "percentage": 3.61, "elapsed_time": "0:24:43", "remaining_time": "11:00:28"}
{"current_steps": 130, "total_steps": 3325, "loss": 0.0996, "lr": 3.903903903903904e-05, "epoch": 0.19530516431924883, "percentage": 3.91, "elapsed_time": "0:26:13", "remaining_time": "10:44:23"}
{"current_steps": 140, "total_steps": 3325, "loss": 0.1047, "lr": 4.204204204204204e-05, "epoch": 0.21032863849765257, "percentage": 4.21, "elapsed_time": "0:27:42", "remaining_time": "10:30:25"}
{"current_steps": 150, "total_steps": 3325, "loss": 0.0933, "lr": 4.5045045045045046e-05, "epoch": 0.22535211267605634, "percentage": 4.51, "elapsed_time": "0:29:12", "remaining_time": "10:18:10"}
{"current_steps": 150, "total_steps": 3325, "eval_loss": 0.07271095365285873, "epoch": 0.22535211267605634, "percentage": 4.51, "elapsed_time": "0:32:40", "remaining_time": "11:31:32"}
{"current_steps": 160, "total_steps": 3325, "loss": 0.0726, "lr": 4.804804804804805e-05, "epoch": 0.2403755868544601, "percentage": 4.81, "elapsed_time": "0:34:13", "remaining_time": "11:17:06"}
{"current_steps": 170, "total_steps": 3325, "loss": 0.0503, "lr": 5.105105105105106e-05, "epoch": 0.25539906103286386, "percentage": 5.11, "elapsed_time": "0:35:43", "remaining_time": "11:02:53"}
{"current_steps": 180, "total_steps": 3325, "loss": 0.0682, "lr": 5.405405405405406e-05, "epoch": 0.2704225352112676, "percentage": 5.41, "elapsed_time": "0:37:12", "remaining_time": "10:50:03"}
{"current_steps": 190, "total_steps": 3325, "loss": 0.0651, "lr": 5.705705705705706e-05, "epoch": 0.28544600938967135, "percentage": 5.71, "elapsed_time": "0:38:41", "remaining_time": "10:38:26"}
{"current_steps": 200, "total_steps": 3325, "loss": 0.0512, "lr": 6.0060060060060066e-05, "epoch": 0.3004694835680751, "percentage": 6.02, "elapsed_time": "0:40:10", "remaining_time": "10:27:49"}
{"current_steps": 200, "total_steps": 3325, "eval_loss": 0.05808287858963013, "epoch": 0.3004694835680751, "percentage": 6.02, "elapsed_time": "0:43:38", "remaining_time": "11:21:58"}
{"current_steps": 210, "total_steps": 3325, "loss": 0.0562, "lr": 6.306306306306306e-05, "epoch": 0.3154929577464789, "percentage": 6.32, "elapsed_time": "0:45:11", "remaining_time": "11:10:18"}
{"current_steps": 220, "total_steps": 3325, "loss": 0.0591, "lr": 6.606606606606607e-05, "epoch": 0.3305164319248826, "percentage": 6.62, "elapsed_time": "0:46:40", "remaining_time": "10:58:45"}
{"current_steps": 230, "total_steps": 3325, "loss": 0.0504, "lr": 6.906906906906907e-05, "epoch": 0.3455399061032864, "percentage": 6.92, "elapsed_time": "0:48:09", "remaining_time": "10:48:03"}
{"current_steps": 240, "total_steps": 3325, "loss": 0.0609, "lr": 7.207207207207208e-05, "epoch": 0.36056338028169016, "percentage": 7.22, "elapsed_time": "0:49:39", "remaining_time": "10:38:12"}
{"current_steps": 250, "total_steps": 3325, "loss": 0.0619, "lr": 7.507507507507507e-05, "epoch": 0.3755868544600939, "percentage": 7.52, "elapsed_time": "0:51:08", "remaining_time": "10:29:01"}
{"current_steps": 250, "total_steps": 3325, "eval_loss": 0.04737601429224014, "epoch": 0.3755868544600939, "percentage": 7.52, "elapsed_time": "0:54:36", "remaining_time": "11:11:38"}
{"current_steps": 260, "total_steps": 3325, "loss": 0.0473, "lr": 7.807807807807808e-05, "epoch": 0.39061032863849765, "percentage": 7.82, "elapsed_time": "0:56:09", "remaining_time": "11:01:55"}
{"current_steps": 270, "total_steps": 3325, "loss": 0.0552, "lr": 8.108108108108109e-05, "epoch": 0.4056338028169014, "percentage": 8.12, "elapsed_time": "0:57:38", "remaining_time": "10:52:07"}
{"current_steps": 280, "total_steps": 3325, "loss": 0.0567, "lr": 8.408408408408409e-05, "epoch": 0.42065727699530514, "percentage": 8.42, "elapsed_time": "0:59:07", "remaining_time": "10:42:58"}
{"current_steps": 290, "total_steps": 3325, "loss": 0.0572, "lr": 8.70870870870871e-05, "epoch": 0.4356807511737089, "percentage": 8.72, "elapsed_time": "1:00:36", "remaining_time": "10:34:17"}
{"current_steps": 300, "total_steps": 3325, "loss": 0.0395, "lr": 9.009009009009009e-05, "epoch": 0.4507042253521127, "percentage": 9.02, "elapsed_time": "1:02:06", "remaining_time": "10:26:11"}
{"current_steps": 300, "total_steps": 3325, "eval_loss": 0.04596845433115959, "epoch": 0.4507042253521127, "percentage": 9.02, "elapsed_time": "1:05:33", "remaining_time": "11:01:03"}
{"current_steps": 310, "total_steps": 3325, "loss": 0.0444, "lr": 9.30930930930931e-05, "epoch": 0.46572769953051646, "percentage": 9.32, "elapsed_time": "1:07:05", "remaining_time": "10:52:31"}
{"current_steps": 320, "total_steps": 3325, "loss": 0.063, "lr": 9.60960960960961e-05, "epoch": 0.4807511737089202, "percentage": 9.62, "elapsed_time": "1:08:34", "remaining_time": "10:44:01"}
{"current_steps": 330, "total_steps": 3325, "loss": 0.0497, "lr": 9.90990990990991e-05, "epoch": 0.49577464788732395, "percentage": 9.92, "elapsed_time": "1:10:04", "remaining_time": "10:35:55"}
{"current_steps": 340, "total_steps": 3325, "loss": 0.058, "lr": 9.999864944989638e-05, "epoch": 0.5107981220657277, "percentage": 10.23, "elapsed_time": "1:11:32", "remaining_time": "10:28:09"}
{"current_steps": 350, "total_steps": 3325, "loss": 0.0476, "lr": 9.999203468625017e-05, "epoch": 0.5258215962441315, "percentage": 10.53, "elapsed_time": "1:13:01", "remaining_time": "10:20:45"}
{"current_steps": 350, "total_steps": 3325, "eval_loss": 0.045401681214571, "epoch": 0.5258215962441315, "percentage": 10.53, "elapsed_time": "1:16:28", "remaining_time": "10:50:05"}
{"current_steps": 360, "total_steps": 3325, "loss": 0.055, "lr": 9.997990837719421e-05, "epoch": 0.5408450704225352, "percentage": 10.83, "elapsed_time": "1:18:02", "remaining_time": "10:42:44"}
{"current_steps": 370, "total_steps": 3325, "loss": 0.0412, "lr": 9.996227185963554e-05, "epoch": 0.5558685446009389, "percentage": 11.13, "elapsed_time": "1:19:32", "remaining_time": "10:35:12"}
{"current_steps": 380, "total_steps": 3325, "loss": 0.0448, "lr": 9.993912707797329e-05, "epoch": 0.5708920187793427, "percentage": 11.43, "elapsed_time": "1:21:01", "remaining_time": "10:27:55"}
{"current_steps": 390, "total_steps": 3325, "loss": 0.0423, "lr": 9.99104765838842e-05, "epoch": 0.5859154929577465, "percentage": 11.73, "elapsed_time": "1:22:30", "remaining_time": "10:20:56"}
{"current_steps": 400, "total_steps": 3325, "loss": 0.0444, "lr": 9.987632353604151e-05, "epoch": 0.6009389671361502, "percentage": 12.03, "elapsed_time": "1:23:59", "remaining_time": "10:14:10"}
{"current_steps": 400, "total_steps": 3325, "eval_loss": 0.0407099612057209, "epoch": 0.6009389671361502, "percentage": 12.03, "elapsed_time": "1:27:26", "remaining_time": "10:39:24"}
{"current_steps": 410, "total_steps": 3325, "loss": 0.0478, "lr": 9.98366716997665e-05, "epoch": 0.615962441314554, "percentage": 12.33, "elapsed_time": "1:28:58", "remaining_time": "10:32:38"}
{"current_steps": 420, "total_steps": 3325, "loss": 0.0381, "lr": 9.979152544661354e-05, "epoch": 0.6309859154929578, "percentage": 12.63, "elapsed_time": "1:30:27", "remaining_time": "10:25:43"}
{"current_steps": 430, "total_steps": 3325, "loss": 0.0435, "lr": 9.974088975388802e-05, "epoch": 0.6460093896713615, "percentage": 12.93, "elapsed_time": "1:31:57", "remaining_time": "10:19:04"}
{"current_steps": 440, "total_steps": 3325, "loss": 0.0378, "lr": 9.968477020409766e-05, "epoch": 0.6610328638497652, "percentage": 13.23, "elapsed_time": "1:33:26", "remaining_time": "10:12:41"}
{"current_steps": 450, "total_steps": 3325, "loss": 0.0543, "lr": 9.962317298433705e-05, "epoch": 0.676056338028169, "percentage": 13.53, "elapsed_time": "1:34:55", "remaining_time": "10:06:27"}
{"current_steps": 450, "total_steps": 3325, "eval_loss": 0.042538370937108994, "epoch": 0.676056338028169, "percentage": 13.53, "elapsed_time": "1:38:22", "remaining_time": "10:28:31"}
{"current_steps": 460, "total_steps": 3325, "loss": 0.0529, "lr": 9.955610488560551e-05, "epoch": 0.6910798122065728, "percentage": 13.83, "elapsed_time": "1:39:55", "remaining_time": "10:22:20"}
{"current_steps": 470, "total_steps": 3325, "loss": 0.0488, "lr": 9.948357330205842e-05, "epoch": 0.7061032863849765, "percentage": 14.14, "elapsed_time": "1:41:24", "remaining_time": "10:15:57"}
{"current_steps": 480, "total_steps": 3325, "loss": 0.0282, "lr": 9.940558623019201e-05, "epoch": 0.7211267605633803, "percentage": 14.44, "elapsed_time": "1:42:52", "remaining_time": "10:09:47"}
{"current_steps": 490, "total_steps": 3325, "loss": 0.0377, "lr": 9.932215226796172e-05, "epoch": 0.7361502347417841, "percentage": 14.74, "elapsed_time": "1:44:21", "remaining_time": "10:03:49"}
{"current_steps": 500, "total_steps": 3325, "loss": 0.0454, "lr": 9.923328061383435e-05, "epoch": 0.7511737089201878, "percentage": 15.04, "elapsed_time": "1:45:50", "remaining_time": "9:58:01"}
{"current_steps": 500, "total_steps": 3325, "eval_loss": 0.0371524840593338, "epoch": 0.7511737089201878, "percentage": 15.04, "elapsed_time": "1:49:18", "remaining_time": "10:17:33"}
{"current_steps": 510, "total_steps": 3325, "loss": 0.0382, "lr": 9.913898106577393e-05, "epoch": 0.7661971830985915, "percentage": 15.34, "elapsed_time": "1:50:50", "remaining_time": "10:11:48"}
{"current_steps": 520, "total_steps": 3325, "loss": 0.0344, "lr": 9.903926402016153e-05, "epoch": 0.7812206572769953, "percentage": 15.64, "elapsed_time": "1:52:19", "remaining_time": "10:05:55"}
{"current_steps": 530, "total_steps": 3325, "loss": 0.0398, "lr": 9.893414047064897e-05, "epoch": 0.7962441314553991, "percentage": 15.94, "elapsed_time": "1:53:48", "remaining_time": "10:00:11"}
{"current_steps": 540, "total_steps": 3325, "loss": 0.0381, "lr": 9.88236220069469e-05, "epoch": 0.8112676056338028, "percentage": 16.24, "elapsed_time": "1:55:17", "remaining_time": "9:54:36"}
{"current_steps": 550, "total_steps": 3325, "loss": 0.0562, "lr": 9.870772081354705e-05, "epoch": 0.8262910798122066, "percentage": 16.54, "elapsed_time": "1:56:47", "remaining_time": "9:49:14"}
{"current_steps": 550, "total_steps": 3325, "eval_loss": 0.03766689449548721, "epoch": 0.8262910798122066, "percentage": 16.54, "elapsed_time": "2:00:14", "remaining_time": "10:06:42"}
{"current_steps": 560, "total_steps": 3325, "loss": 0.0406, "lr": 9.858644966837878e-05, "epoch": 0.8413145539906103, "percentage": 16.84, "elapsed_time": "2:01:47", "remaining_time": "10:01:21"}
{"current_steps": 570, "total_steps": 3325, "loss": 0.0381, "lr": 9.845982194140051e-05, "epoch": 0.856338028169014, "percentage": 17.14, "elapsed_time": "2:03:17", "remaining_time": "9:55:52"}
{"current_steps": 580, "total_steps": 3325, "loss": 0.0404, "lr": 9.832785159312559e-05, "epoch": 0.8713615023474178, "percentage": 17.44, "elapsed_time": "2:04:46", "remaining_time": "9:50:32"}
{"current_steps": 590, "total_steps": 3325, "loss": 0.0365, "lr": 9.819055317308317e-05, "epoch": 0.8863849765258216, "percentage": 17.74, "elapsed_time": "2:06:16", "remaining_time": "9:45:19"}
{"current_steps": 600, "total_steps": 3325, "loss": 0.0336, "lr": 9.804794181821422e-05, "epoch": 0.9014084507042254, "percentage": 18.05, "elapsed_time": "2:07:46", "remaining_time": "9:40:16"}
{"current_steps": 600, "total_steps": 3325, "eval_loss": 0.036089226603507996, "epoch": 0.9014084507042254, "percentage": 18.05, "elapsed_time": "2:11:14", "remaining_time": "9:56:04"}
{"current_steps": 610, "total_steps": 3325, "loss": 0.0375, "lr": 9.790003325120261e-05, "epoch": 0.9164319248826291, "percentage": 18.35, "elapsed_time": "2:12:47", "remaining_time": "9:51:02"}
{"current_steps": 620, "total_steps": 3325, "loss": 0.0466, "lr": 9.774684377874178e-05, "epoch": 0.9314553990610329, "percentage": 18.65, "elapsed_time": "2:14:17", "remaining_time": "9:45:53"}
{"current_steps": 630, "total_steps": 3325, "loss": 0.0308, "lr": 9.758839028973692e-05, "epoch": 0.9464788732394366, "percentage": 18.95, "elapsed_time": "2:15:47", "remaining_time": "9:40:53"}
{"current_steps": 640, "total_steps": 3325, "loss": 0.0416, "lr": 9.742469025344298e-05, "epoch": 0.9615023474178404, "percentage": 19.25, "elapsed_time": "2:17:17", "remaining_time": "9:35:58"}
{"current_steps": 650, "total_steps": 3325, "loss": 0.0494, "lr": 9.725576171753874e-05, "epoch": 0.9765258215962441, "percentage": 19.55, "elapsed_time": "2:18:47", "remaining_time": "9:31:12"}
{"current_steps": 650, "total_steps": 3325, "eval_loss": 0.03678149729967117, "epoch": 0.9765258215962441, "percentage": 19.55, "elapsed_time": "2:22:17", "remaining_time": "9:45:36"}
{"current_steps": 660, "total_steps": 3325, "loss": 0.0367, "lr": 9.708162330613708e-05, "epoch": 0.9915492957746479, "percentage": 19.85, "elapsed_time": "2:23:51", "remaining_time": "9:40:52"}
{"current_steps": 670, "total_steps": 3325, "loss": 0.0369, "lr": 9.690229421773167e-05, "epoch": 1.0065727699530516, "percentage": 20.15, "elapsed_time": "2:25:21", "remaining_time": "9:36:02"}
{"current_steps": 680, "total_steps": 3325, "loss": 0.0312, "lr": 9.67177942230804e-05, "epoch": 1.0215962441314554, "percentage": 20.45, "elapsed_time": "2:26:52", "remaining_time": "9:31:19"}
{"current_steps": 690, "total_steps": 3325, "loss": 0.0392, "lr": 9.652814366302568e-05, "epoch": 1.036619718309859, "percentage": 20.75, "elapsed_time": "2:28:23", "remaining_time": "9:26:41"}
{"current_steps": 700, "total_steps": 3325, "loss": 0.0354, "lr": 9.633336344625185e-05, "epoch": 1.051643192488263, "percentage": 21.05, "elapsed_time": "2:29:54", "remaining_time": "9:22:09"}
{"current_steps": 700, "total_steps": 3325, "eval_loss": 0.0386400930583477, "epoch": 1.051643192488263, "percentage": 21.05, "elapsed_time": "2:33:25", "remaining_time": "9:35:19"}
{"current_steps": 710, "total_steps": 3325, "loss": 0.0282, "lr": 9.61334750469801e-05, "epoch": 1.0666666666666667, "percentage": 21.35, "elapsed_time": "2:34:59", "remaining_time": "9:30:49"}
{"current_steps": 720, "total_steps": 3325, "loss": 0.0349, "lr": 9.592850050260089e-05, "epoch": 1.0816901408450703, "percentage": 21.65, "elapsed_time": "2:36:30", "remaining_time": "9:26:14"}
{"current_steps": 730, "total_steps": 3325, "loss": 0.0291, "lr": 9.571846241124446e-05, "epoch": 1.0967136150234742, "percentage": 21.95, "elapsed_time": "2:38:01", "remaining_time": "9:21:44"}
{"current_steps": 740, "total_steps": 3325, "loss": 0.0257, "lr": 9.55033839292893e-05, "epoch": 1.1117370892018779, "percentage": 22.26, "elapsed_time": "2:39:33", "remaining_time": "9:17:21"}
{"current_steps": 750, "total_steps": 3325, "loss": 0.029, "lr": 9.52832887688093e-05, "epoch": 1.1267605633802817, "percentage": 22.56, "elapsed_time": "2:41:04", "remaining_time": "9:13:00"}
{"current_steps": 750, "total_steps": 3325, "eval_loss": 0.037642158567905426, "epoch": 1.1267605633802817, "percentage": 22.56, "elapsed_time": "2:44:36", "remaining_time": "9:25:10"}
{"current_steps": 760, "total_steps": 3325, "loss": 0.0329, "lr": 9.50582011949595e-05, "epoch": 1.1417840375586854, "percentage": 22.86, "elapsed_time": "2:46:11", "remaining_time": "9:20:54"}
{"current_steps": 770, "total_steps": 3325, "loss": 0.0328, "lr": 9.482814602330084e-05, "epoch": 1.1568075117370893, "percentage": 23.16, "elapsed_time": "2:47:43", "remaining_time": "9:16:31"}
{"current_steps": 780, "total_steps": 3325, "loss": 0.0363, "lr": 9.459314861706435e-05, "epoch": 1.171830985915493, "percentage": 23.46, "elapsed_time": "2:49:14", "remaining_time": "9:12:12"}
{"current_steps": 790, "total_steps": 3325, "loss": 0.0279, "lr": 9.435323488435488e-05, "epoch": 1.1868544600938966, "percentage": 23.76, "elapsed_time": "2:50:45", "remaining_time": "9:07:57"}
{"current_steps": 800, "total_steps": 3325, "loss": 0.0301, "lr": 9.410843127529473e-05, "epoch": 1.2018779342723005, "percentage": 24.06, "elapsed_time": "2:52:17", "remaining_time": "9:03:48"}
{"current_steps": 800, "total_steps": 3325, "eval_loss": 0.03519855812191963, "epoch": 1.2018779342723005, "percentage": 24.06, "elapsed_time": "2:55:51", "remaining_time": "9:15:04"}
{"current_steps": 810, "total_steps": 3325, "loss": 0.0317, "lr": 9.385876477910765e-05, "epoch": 1.2169014084507042, "percentage": 24.36, "elapsed_time": "2:57:27", "remaining_time": "9:11:00"}
{"current_steps": 820, "total_steps": 3325, "loss": 0.0291, "lr": 9.360426292114314e-05, "epoch": 1.231924882629108, "percentage": 24.66, "elapsed_time": "2:58:59", "remaining_time": "9:06:48"}
{"current_steps": 830, "total_steps": 3325, "loss": 0.0279, "lr": 9.334495375984212e-05, "epoch": 1.2469483568075117, "percentage": 24.96, "elapsed_time": "3:00:31", "remaining_time": "9:02:39"}
{"current_steps": 840, "total_steps": 3325, "loss": 0.0293, "lr": 9.30808658836432e-05, "epoch": 1.2619718309859156, "percentage": 25.26, "elapsed_time": "3:02:03", "remaining_time": "8:58:35"}
{"current_steps": 850, "total_steps": 3325, "loss": 0.0321, "lr": 9.281202840783108e-05, "epoch": 1.2769953051643192, "percentage": 25.56, "elapsed_time": "3:03:35", "remaining_time": "8:54:34"}
{"current_steps": 850, "total_steps": 3325, "eval_loss": 0.034056928008794785, "epoch": 1.2769953051643192, "percentage": 25.56, "elapsed_time": "3:07:10", "remaining_time": "9:05:01"}
{"current_steps": 860, "total_steps": 3325, "loss": 0.0243, "lr": 9.253847097132655e-05, "epoch": 1.292018779342723, "percentage": 25.86, "elapsed_time": "3:08:45", "remaining_time": "9:01:02"}
{"current_steps": 870, "total_steps": 3325, "loss": 0.0247, "lr": 9.226022373341882e-05, "epoch": 1.3070422535211268, "percentage": 26.17, "elapsed_time": "3:10:17", "remaining_time": "8:56:59"}
{"current_steps": 880, "total_steps": 3325, "loss": 0.0245, "lr": 9.19773173704406e-05, "epoch": 1.3220657276995305, "percentage": 26.47, "elapsed_time": "3:11:50", "remaining_time": "8:53:00"}
{"current_steps": 890, "total_steps": 3325, "loss": 0.032, "lr": 9.168978307238594e-05, "epoch": 1.3370892018779343, "percentage": 26.77, "elapsed_time": "3:13:22", "remaining_time": "8:49:05"}
{"current_steps": 900, "total_steps": 3325, "loss": 0.0271, "lr": 9.13976525394717e-05, "epoch": 1.352112676056338, "percentage": 27.07, "elapsed_time": "3:14:55", "remaining_time": "8:45:13"}
{"current_steps": 900, "total_steps": 3325, "eval_loss": 0.03430046886205673, "epoch": 1.352112676056338, "percentage": 27.07, "elapsed_time": "3:18:31", "remaining_time": "8:54:54"}
{"current_steps": 910, "total_steps": 3325, "loss": 0.0224, "lr": 9.110095797864263e-05, "epoch": 1.3671361502347419, "percentage": 27.37, "elapsed_time": "3:20:07", "remaining_time": "8:51:06"}
{"current_steps": 920, "total_steps": 3325, "loss": 0.0318, "lr": 9.079973210002051e-05, "epoch": 1.3821596244131455, "percentage": 27.67, "elapsed_time": "3:21:39", "remaining_time": "8:47:10"}
{"current_steps": 930, "total_steps": 3325, "loss": 0.0322, "lr": 9.049400811329807e-05, "epoch": 1.3971830985915492, "percentage": 27.97, "elapsed_time": "3:23:13", "remaining_time": "8:43:20"}
{"current_steps": 940, "total_steps": 3325, "loss": 0.0305, "lr": 9.01838197240775e-05, "epoch": 1.412206572769953, "percentage": 28.27, "elapsed_time": "3:24:46", "remaining_time": "8:39:33"}
{"current_steps": 950, "total_steps": 3325, "loss": 0.0351, "lr": 8.986920113015461e-05, "epoch": 1.4272300469483568, "percentage": 28.57, "elapsed_time": "3:26:19", "remaining_time": "8:35:48"}
{"current_steps": 950, "total_steps": 3325, "eval_loss": 0.03298981115221977, "epoch": 1.4272300469483568, "percentage": 28.57, "elapsed_time": "3:29:55", "remaining_time": "8:44:48"}
{"current_steps": 960, "total_steps": 3325, "loss": 0.0262, "lr": 8.955018701774846e-05, "epoch": 1.4422535211267606, "percentage": 28.87, "elapsed_time": "3:31:32", "remaining_time": "8:41:08"}
{"current_steps": 970, "total_steps": 3325, "loss": 0.0347, "lr": 8.922681255767731e-05, "epoch": 1.4572769953051643, "percentage": 29.17, "elapsed_time": "3:33:06", "remaining_time": "8:37:22"}
{"current_steps": 980, "total_steps": 3325, "loss": 0.0308, "lr": 8.889911340148112e-05, "epoch": 1.4723004694835682, "percentage": 29.47, "elapsed_time": "3:34:39", "remaining_time": "8:33:38"}
{"current_steps": 990, "total_steps": 3325, "loss": 0.0278, "lr": 8.856712567749095e-05, "epoch": 1.4873239436619718, "percentage": 29.77, "elapsed_time": "3:36:12", "remaining_time": "8:29:56"}
{"current_steps": 1000, "total_steps": 3325, "loss": 0.0244, "lr": 8.82308859868459e-05, "epoch": 1.5023474178403755, "percentage": 30.08, "elapsed_time": "3:37:45", "remaining_time": "8:26:17"}
{"current_steps": 1000, "total_steps": 3325, "eval_loss": 0.033029552549123764, "epoch": 1.5023474178403755, "percentage": 30.08, "elapsed_time": "3:41:23", "remaining_time": "8:34:43"}
{"current_steps": 1010, "total_steps": 3325, "loss": 0.0298, "lr": 8.789043139945795e-05, "epoch": 1.5173708920187794, "percentage": 30.38, "elapsed_time": "3:43:00", "remaining_time": "8:31:08"}
{"current_steps": 1020, "total_steps": 3325, "loss": 0.036, "lr": 8.754579944992491e-05, "epoch": 1.532394366197183, "percentage": 30.68, "elapsed_time": "3:44:34", "remaining_time": "8:27:29"}
{"current_steps": 1030, "total_steps": 3325, "loss": 0.0288, "lr": 8.719702813339248e-05, "epoch": 1.5474178403755867, "percentage": 30.98, "elapsed_time": "3:46:08", "remaining_time": "8:23:52"}
{"current_steps": 1040, "total_steps": 3325, "loss": 0.0277, "lr": 8.684415590136518e-05, "epoch": 1.5624413145539906, "percentage": 31.28, "elapsed_time": "3:47:42", "remaining_time": "8:20:17"}
{"current_steps": 1050, "total_steps": 3325, "loss": 0.0277, "lr": 8.648722165746722e-05, "epoch": 1.5774647887323945, "percentage": 31.58, "elapsed_time": "3:49:15", "remaining_time": "8:16:44"}
{"current_steps": 1050, "total_steps": 3325, "eval_loss": 0.034053802490234375, "epoch": 1.5774647887323945, "percentage": 31.58, "elapsed_time": "3:52:55", "remaining_time": "8:24:39"}
{"current_steps": 1060, "total_steps": 3325, "loss": 0.0251, "lr": 8.61262647531534e-05, "epoch": 1.5924882629107981, "percentage": 31.88, "elapsed_time": "3:54:32", "remaining_time": "8:21:11"}
{"current_steps": 1070, "total_steps": 3325, "loss": 0.0362, "lr": 8.576132498337068e-05, "epoch": 1.6075117370892018, "percentage": 32.18, "elapsed_time": "3:56:06", "remaining_time": "8:17:36"}
{"current_steps": 1080, "total_steps": 3325, "loss": 0.0319, "lr": 8.539244258217088e-05, "epoch": 1.6225352112676057, "percentage": 32.48, "elapsed_time": "3:57:41", "remaining_time": "8:14:04"}
{"current_steps": 1090, "total_steps": 3325, "loss": 0.032, "lr": 8.501965821827485e-05, "epoch": 1.6375586854460094, "percentage": 32.78, "elapsed_time": "3:59:15", "remaining_time": "8:10:35"}
{"current_steps": 1100, "total_steps": 3325, "loss": 0.0231, "lr": 8.464301299058892e-05, "epoch": 1.652582159624413, "percentage": 33.08, "elapsed_time": "4:00:49", "remaining_time": "8:07:07"}
{"current_steps": 1100, "total_steps": 3325, "eval_loss": 0.034014929085969925, "epoch": 1.652582159624413, "percentage": 33.08, "elapsed_time": "4:04:29", "remaining_time": "8:14:32"}
{"current_steps": 1110, "total_steps": 3325, "loss": 0.0258, "lr": 8.426254842367374e-05, "epoch": 1.667605633802817, "percentage": 33.38, "elapsed_time": "4:06:06", "remaining_time": "8:11:06"}
{"current_steps": 1120, "total_steps": 3325, "loss": 0.0334, "lr": 8.387830646316623e-05, "epoch": 1.6826291079812208, "percentage": 33.68, "elapsed_time": "4:07:41", "remaining_time": "8:07:37"}
{"current_steps": 1130, "total_steps": 3325, "loss": 0.0249, "lr": 8.349032947115525e-05, "epoch": 1.6976525821596244, "percentage": 33.98, "elapsed_time": "4:09:16", "remaining_time": "8:04:11"}
{"current_steps": 1140, "total_steps": 3325, "loss": 0.0382, "lr": 8.309866022151107e-05, "epoch": 1.712676056338028, "percentage": 34.29, "elapsed_time": "4:10:51", "remaining_time": "8:00:48"}
{"current_steps": 1150, "total_steps": 3325, "loss": 0.0261, "lr": 8.270334189516983e-05, "epoch": 1.727699530516432, "percentage": 34.59, "elapsed_time": "4:12:25", "remaining_time": "7:57:25"}
{"current_steps": 1150, "total_steps": 3325, "eval_loss": 0.03269151970744133, "epoch": 1.727699530516432, "percentage": 34.59, "elapsed_time": "4:16:06", "remaining_time": "8:04:22"}
{"current_steps": 1160, "total_steps": 3325, "loss": 0.0393, "lr": 8.230441807537277e-05, "epoch": 1.7427230046948357, "percentage": 34.89, "elapsed_time": "4:17:45", "remaining_time": "8:01:03"}
{"current_steps": 1170, "total_steps": 3325, "loss": 0.026, "lr": 8.190193274286122e-05, "epoch": 1.7577464788732393, "percentage": 35.19, "elapsed_time": "4:19:19", "remaining_time": "7:57:38"}
{"current_steps": 1180, "total_steps": 3325, "loss": 0.0295, "lr": 8.149593027102789e-05, "epoch": 1.7727699530516432, "percentage": 35.49, "elapsed_time": "4:20:54", "remaining_time": "7:54:15"}
{"current_steps": 1190, "total_steps": 3325, "loss": 0.0255, "lr": 8.108645542102469e-05, "epoch": 1.787793427230047, "percentage": 35.79, "elapsed_time": "4:22:28", "remaining_time": "7:50:55"}
{"current_steps": 1200, "total_steps": 3325, "loss": 0.0297, "lr": 8.067355333682798e-05, "epoch": 1.8028169014084507, "percentage": 36.09, "elapsed_time": "4:24:03", "remaining_time": "7:47:36"}
{"current_steps": 1200, "total_steps": 3325, "eval_loss": 0.03482316806912422, "epoch": 1.8028169014084507, "percentage": 36.09, "elapsed_time": "4:27:44", "remaining_time": "7:54:06"}
{"current_steps": 1210, "total_steps": 3325, "loss": 0.0228, "lr": 8.025726954026138e-05, "epoch": 1.8178403755868544, "percentage": 36.39, "elapsed_time": "4:29:21", "remaining_time": "7:50:50"}
{"current_steps": 1220, "total_steps": 3325, "loss": 0.0272, "lr": 7.983764992597716e-05, "epoch": 1.8328638497652583, "percentage": 36.69, "elapsed_time": "4:30:56", "remaining_time": "7:47:29"}
{"current_steps": 1230, "total_steps": 3325, "loss": 0.0296, "lr": 7.94147407563964e-05, "epoch": 1.847887323943662, "percentage": 36.99, "elapsed_time": "4:32:31", "remaining_time": "7:44:10"}
{"current_steps": 1240, "total_steps": 3325, "loss": 0.0333, "lr": 7.89885886566086e-05, "epoch": 1.8629107981220656, "percentage": 37.29, "elapsed_time": "4:34:06", "remaining_time": "7:40:53"}
{"current_steps": 1250, "total_steps": 3325, "loss": 0.027, "lr": 7.855924060923141e-05, "epoch": 1.8779342723004695, "percentage": 37.59, "elapsed_time": "4:35:40", "remaining_time": "7:37:37"}
{"current_steps": 1250, "total_steps": 3325, "eval_loss": 0.03339933604001999, "epoch": 1.8779342723004695, "percentage": 37.59, "elapsed_time": "4:39:21", "remaining_time": "7:43:43"}
{"current_steps": 1260, "total_steps": 3325, "loss": 0.03, "lr": 7.812674394923077e-05, "epoch": 1.8929577464788734, "percentage": 37.89, "elapsed_time": "4:40:59", "remaining_time": "7:40:31"}
{"current_steps": 1270, "total_steps": 3325, "loss": 0.0275, "lr": 7.769114635870231e-05, "epoch": 1.907981220657277, "percentage": 38.2, "elapsed_time": "4:42:34", "remaining_time": "7:37:14"}
{"current_steps": 1280, "total_steps": 3325, "loss": 0.0294, "lr": 7.725249586161463e-05, "epoch": 1.9230046948356807, "percentage": 38.5, "elapsed_time": "4:44:09", "remaining_time": "7:33:59"}
{"current_steps": 1290, "total_steps": 3325, "loss": 0.031, "lr": 7.68108408185145e-05, "epoch": 1.9380281690140846, "percentage": 38.8, "elapsed_time": "4:45:44", "remaining_time": "7:30:46"}
{"current_steps": 1300, "total_steps": 3325, "loss": 0.0417, "lr": 7.636622992119536e-05, "epoch": 1.9530516431924883, "percentage": 39.1, "elapsed_time": "4:47:20", "remaining_time": "7:27:34"}
{"current_steps": 1300, "total_steps": 3325, "eval_loss": 0.03484388440847397, "epoch": 1.9530516431924883, "percentage": 39.1, "elapsed_time": "4:51:01", "remaining_time": "7:33:19"}
{"current_steps": 1310, "total_steps": 3325, "loss": 0.0329, "lr": 7.591871218732902e-05, "epoch": 1.968075117370892, "percentage": 39.4, "elapsed_time": "4:52:39", "remaining_time": "7:30:10"}
{"current_steps": 1320, "total_steps": 3325, "loss": 0.0364, "lr": 7.54683369550616e-05, "epoch": 1.9830985915492958, "percentage": 39.7, "elapsed_time": "4:54:15", "remaining_time": "7:26:57"}
{"current_steps": 1330, "total_steps": 3325, "loss": 0.0269, "lr": 7.501515387757404e-05, "epoch": 1.9981220657276997, "percentage": 40.0, "elapsed_time": "4:55:50", "remaining_time": "7:23:45"}
{"current_steps": 1340, "total_steps": 3325, "loss": 0.0313, "lr": 7.455921291760796e-05, "epoch": 2.013145539906103, "percentage": 40.3, "elapsed_time": "4:57:25", "remaining_time": "7:20:34"}
{"current_steps": 1350, "total_steps": 3325, "loss": 0.0173, "lr": 7.410056434195725e-05, "epoch": 2.028169014084507, "percentage": 40.6, "elapsed_time": "4:59:00", "remaining_time": "7:17:25"}
{"current_steps": 1350, "total_steps": 3325, "eval_loss": 0.03278001397848129, "epoch": 2.028169014084507, "percentage": 40.6, "elapsed_time": "5:02:41", "remaining_time": "7:22:49"}
{"current_steps": 1360, "total_steps": 3325, "loss": 0.0205, "lr": 7.363925871592629e-05, "epoch": 2.043192488262911, "percentage": 40.9, "elapsed_time": "5:04:19", "remaining_time": "7:19:41"}
{"current_steps": 1370, "total_steps": 3325, "loss": 0.0201, "lr": 7.317534689775528e-05, "epoch": 2.0582159624413148, "percentage": 41.2, "elapsed_time": "5:05:54", "remaining_time": "7:16:32"}
{"current_steps": 1380, "total_steps": 3325, "loss": 0.0159, "lr": 7.270888003301304e-05, "epoch": 2.073239436619718, "percentage": 41.5, "elapsed_time": "5:07:29", "remaining_time": "7:13:23"}
{"current_steps": 1390, "total_steps": 3325, "loss": 0.0222, "lr": 7.22399095489584e-05, "epoch": 2.088262910798122, "percentage": 41.8, "elapsed_time": "5:09:05", "remaining_time": "7:10:16"}
{"current_steps": 1400, "total_steps": 3325, "loss": 0.0207, "lr": 7.176848714887042e-05, "epoch": 2.103286384976526, "percentage": 42.11, "elapsed_time": "5:10:41", "remaining_time": "7:07:11"}
{"current_steps": 1400, "total_steps": 3325, "eval_loss": 0.03226166218519211, "epoch": 2.103286384976526, "percentage": 42.11, "elapsed_time": "5:14:23", "remaining_time": "7:12:17"}
{"current_steps": 1410, "total_steps": 3325, "loss": 0.0225, "lr": 7.129466480634806e-05, "epoch": 2.1183098591549294, "percentage": 42.41, "elapsed_time": "5:16:02", "remaining_time": "7:09:13"}
{"current_steps": 1420, "total_steps": 3325, "loss": 0.0187, "lr": 7.081849475958042e-05, "epoch": 2.1333333333333333, "percentage": 42.71, "elapsed_time": "5:17:38", "remaining_time": "7:06:07"}
{"current_steps": 1430, "total_steps": 3325, "loss": 0.0185, "lr": 7.034002950558723e-05, "epoch": 2.148356807511737, "percentage": 43.01, "elapsed_time": "5:19:13", "remaining_time": "7:03:01"}
{"current_steps": 1440, "total_steps": 3325, "loss": 0.0192, "lr": 6.985932179443144e-05, "epoch": 2.1633802816901406, "percentage": 43.31, "elapsed_time": "5:20:49", "remaining_time": "6:59:57"}
{"current_steps": 1450, "total_steps": 3325, "loss": 0.0223, "lr": 6.937642462340342e-05, "epoch": 2.1784037558685445, "percentage": 43.61, "elapsed_time": "5:22:24", "remaining_time": "6:56:54"}
{"current_steps": 1450, "total_steps": 3325, "eval_loss": 0.03254189342260361, "epoch": 2.1784037558685445, "percentage": 43.61, "elapsed_time": "5:26:06", "remaining_time": "7:01:41"}
{"current_steps": 1460, "total_steps": 3325, "loss": 0.0191, "lr": 6.889139123117817e-05, "epoch": 2.1934272300469484, "percentage": 43.91, "elapsed_time": "5:27:45", "remaining_time": "6:58:40"}
{"current_steps": 1470, "total_steps": 3325, "loss": 0.0209, "lr": 6.840427509194575e-05, "epoch": 2.2084507042253523, "percentage": 44.21, "elapsed_time": "5:29:21", "remaining_time": "6:55:37"}
{"current_steps": 1480, "total_steps": 3325, "loss": 0.0149, "lr": 6.791512990951597e-05, "epoch": 2.2234741784037557, "percentage": 44.51, "elapsed_time": "5:30:56", "remaining_time": "6:52:33"}
{"current_steps": 1490, "total_steps": 3325, "loss": 0.0211, "lr": 6.74240096113975e-05, "epoch": 2.2384976525821596, "percentage": 44.81, "elapsed_time": "5:32:32", "remaining_time": "6:49:32"}
{"current_steps": 1500, "total_steps": 3325, "loss": 0.0107, "lr": 6.693096834285256e-05, "epoch": 2.2535211267605635, "percentage": 45.11, "elapsed_time": "5:34:07", "remaining_time": "6:46:31"}
{"current_steps": 1500, "total_steps": 3325, "eval_loss": 0.035872478038072586, "epoch": 2.2535211267605635, "percentage": 45.11, "elapsed_time": "5:37:51", "remaining_time": "6:51:03"}
{"current_steps": 1510, "total_steps": 3325, "loss": 0.0226, "lr": 6.643606046092732e-05, "epoch": 2.2685446009389674, "percentage": 45.41, "elapsed_time": "5:39:29", "remaining_time": "6:48:04"}
{"current_steps": 1520, "total_steps": 3325, "loss": 0.0253, "lr": 6.593934052845929e-05, "epoch": 2.283568075117371, "percentage": 45.71, "elapsed_time": "5:41:05", "remaining_time": "6:45:02"}
{"current_steps": 1530, "total_steps": 3325, "loss": 0.0205, "lr": 6.544086330806181e-05, "epoch": 2.2985915492957747, "percentage": 46.02, "elapsed_time": "5:42:40", "remaining_time": "6:42:01"}
{"current_steps": 1540, "total_steps": 3325, "loss": 0.0197, "lr": 6.494068375608646e-05, "epoch": 2.3136150234741786, "percentage": 46.32, "elapsed_time": "5:44:16", "remaining_time": "6:39:02"}
{"current_steps": 1550, "total_steps": 3325, "loss": 0.0182, "lr": 6.443885701656432e-05, "epoch": 2.328638497652582, "percentage": 46.62, "elapsed_time": "5:45:51", "remaining_time": "6:36:04"}
{"current_steps": 1550, "total_steps": 3325, "eval_loss": 0.03318268060684204, "epoch": 2.328638497652582, "percentage": 46.62, "elapsed_time": "5:49:35", "remaining_time": "6:40:19"}
{"current_steps": 1560, "total_steps": 3325, "loss": 0.02, "lr": 6.393543841512632e-05, "epoch": 2.343661971830986, "percentage": 46.92, "elapsed_time": "5:51:13", "remaining_time": "6:37:23"}
{"current_steps": 1570, "total_steps": 3325, "loss": 0.0144, "lr": 6.343048345290386e-05, "epoch": 2.35868544600939, "percentage": 47.22, "elapsed_time": "5:52:49", "remaining_time": "6:34:23"}
{"current_steps": 1580, "total_steps": 3325, "loss": 0.023, "lr": 6.292404780040961e-05, "epoch": 2.3737089201877932, "percentage": 47.52, "elapsed_time": "5:54:24", "remaining_time": "6:31:25"}
{"current_steps": 1590, "total_steps": 3325, "loss": 0.0157, "lr": 6.241618729140018e-05, "epoch": 2.388732394366197, "percentage": 47.82, "elapsed_time": "5:56:00", "remaining_time": "6:28:28"}
{"current_steps": 1600, "total_steps": 3325, "loss": 0.0187, "lr": 6.190695791672042e-05, "epoch": 2.403755868544601, "percentage": 48.12, "elapsed_time": "5:57:36", "remaining_time": "6:25:33"}
{"current_steps": 1600, "total_steps": 3325, "eval_loss": 0.032255951315164566, "epoch": 2.403755868544601, "percentage": 48.12, "elapsed_time": "6:01:19", "remaining_time": "6:29:33"}
{"current_steps": 1610, "total_steps": 3325, "loss": 0.0183, "lr": 6.139641581813052e-05, "epoch": 2.418779342723005, "percentage": 48.42, "elapsed_time": "6:02:58", "remaining_time": "6:26:38"}
{"current_steps": 1620, "total_steps": 3325, "loss": 0.0177, "lr": 6.088461728211642e-05, "epoch": 2.4338028169014083, "percentage": 48.72, "elapsed_time": "6:04:33", "remaining_time": "6:23:41"}
{"current_steps": 1630, "total_steps": 3325, "loss": 0.0117, "lr": 6.0371618733684474e-05, "epoch": 2.448826291079812, "percentage": 49.02, "elapsed_time": "6:06:09", "remaining_time": "6:20:45"}
{"current_steps": 1640, "total_steps": 3325, "loss": 0.0175, "lr": 5.9857476730140485e-05, "epoch": 2.463849765258216, "percentage": 49.32, "elapsed_time": "6:07:44", "remaining_time": "6:17:49"}
{"current_steps": 1650, "total_steps": 3325, "loss": 0.018, "lr": 5.9342247954854466e-05, "epoch": 2.4788732394366195, "percentage": 49.62, "elapsed_time": "6:09:19", "remaining_time": "6:14:55"}
{"current_steps": 1650, "total_steps": 3325, "eval_loss": 0.03270378336310387, "epoch": 2.4788732394366195, "percentage": 49.62, "elapsed_time": "6:13:01", "remaining_time": "6:18:40"}
{"current_steps": 1660, "total_steps": 3325, "loss": 0.022, "lr": 5.8825989211011335e-05, "epoch": 2.4938967136150234, "percentage": 49.92, "elapsed_time": "6:14:39", "remaining_time": "6:15:47"}
{"current_steps": 1670, "total_steps": 3325, "loss": 0.0159, "lr": 5.830875741534852e-05, "epoch": 2.5089201877934273, "percentage": 50.23, "elapsed_time": "6:16:15", "remaining_time": "6:12:52"}
{"current_steps": 1680, "total_steps": 3325, "loss": 0.0239, "lr": 5.7790609591880826e-05, "epoch": 2.523943661971831, "percentage": 50.53, "elapsed_time": "6:17:50", "remaining_time": "6:09:58"}
{"current_steps": 1690, "total_steps": 3325, "loss": 0.0224, "lr": 5.727160286561386e-05, "epoch": 2.5389671361502346, "percentage": 50.83, "elapsed_time": "6:19:25", "remaining_time": "6:07:04"}
{"current_steps": 1700, "total_steps": 3325, "loss": 0.0205, "lr": 5.675179445624581e-05, "epoch": 2.5539906103286385, "percentage": 51.13, "elapsed_time": "6:21:01", "remaining_time": "6:04:12"}
{"current_steps": 1700, "total_steps": 3325, "eval_loss": 0.03496889770030975, "epoch": 2.5539906103286385, "percentage": 51.13, "elapsed_time": "6:24:42", "remaining_time": "6:07:44"}
{"current_steps": 1710, "total_steps": 3325, "loss": 0.0164, "lr": 5.62312416718593e-05, "epoch": 2.5690140845070424, "percentage": 51.43, "elapsed_time": "6:26:21", "remaining_time": "6:04:53"}
{"current_steps": 1720, "total_steps": 3325, "loss": 0.0257, "lr": 5.5710001902603116e-05, "epoch": 2.584037558685446, "percentage": 51.73, "elapsed_time": "6:27:56", "remaining_time": "6:02:00"}
{"current_steps": 1730, "total_steps": 3325, "loss": 0.0198, "lr": 5.5188132614365094e-05, "epoch": 2.5990610328638497, "percentage": 52.03, "elapsed_time": "6:29:31", "remaining_time": "5:59:08"}
{"current_steps": 1740, "total_steps": 3325, "loss": 0.0152, "lr": 5.4665691342436565e-05, "epoch": 2.6140845070422536, "percentage": 52.33, "elapsed_time": "6:31:07", "remaining_time": "5:56:16"}
{"current_steps": 1750, "total_steps": 3325, "loss": 0.0182, "lr": 5.414273568516919e-05, "epoch": 2.629107981220657, "percentage": 52.63, "elapsed_time": "6:32:43", "remaining_time": "5:53:26"}
{"current_steps": 1750, "total_steps": 3325, "eval_loss": 0.032272376120090485, "epoch": 2.629107981220657, "percentage": 52.63, "elapsed_time": "6:36:24", "remaining_time": "5:56:46"}
{"current_steps": 1760, "total_steps": 3325, "loss": 0.0197, "lr": 5.361932329762481e-05, "epoch": 2.644131455399061, "percentage": 52.93, "elapsed_time": "6:38:03", "remaining_time": "5:53:56"}
{"current_steps": 1770, "total_steps": 3325, "loss": 0.0197, "lr": 5.309551188521914e-05, "epoch": 2.659154929577465, "percentage": 53.23, "elapsed_time": "6:39:37", "remaining_time": "5:51:05"}
{"current_steps": 1780, "total_steps": 3325, "loss": 0.0161, "lr": 5.2571359197359704e-05, "epoch": 2.6741784037558687, "percentage": 53.53, "elapsed_time": "6:41:13", "remaining_time": "5:48:15"}
{"current_steps": 1790, "total_steps": 3325, "loss": 0.0151, "lr": 5.2046923021079175e-05, "epoch": 2.6892018779342726, "percentage": 53.83, "elapsed_time": "6:42:48", "remaining_time": "5:45:25"}
{"current_steps": 1800, "total_steps": 3325, "loss": 0.0202, "lr": 5.1522261174664346e-05, "epoch": 2.704225352112676, "percentage": 54.14, "elapsed_time": "6:44:23", "remaining_time": "5:42:36"}
{"current_steps": 1800, "total_steps": 3325, "eval_loss": 0.032467469573020935, "epoch": 2.704225352112676, "percentage": 54.14, "elapsed_time": "6:48:05", "remaining_time": "5:45:44"}
{"current_steps": 1810, "total_steps": 3325, "loss": 0.0262, "lr": 5.0997431501281835e-05, "epoch": 2.71924882629108, "percentage": 54.44, "elapsed_time": "6:49:43", "remaining_time": "5:42:56"}
{"current_steps": 1820, "total_steps": 3325, "loss": 0.0202, "lr": 5.0472491862600915e-05, "epoch": 2.7342723004694838, "percentage": 54.74, "elapsed_time": "6:51:17", "remaining_time": "5:40:06"}
{"current_steps": 1830, "total_steps": 3325, "loss": 0.022, "lr": 4.994750013241435e-05, "epoch": 2.749295774647887, "percentage": 55.04, "elapsed_time": "6:52:53", "remaining_time": "5:37:18"}
{"current_steps": 1840, "total_steps": 3325, "loss": 0.0201, "lr": 4.9422514190257974e-05, "epoch": 2.764319248826291, "percentage": 55.34, "elapsed_time": "6:54:28", "remaining_time": "5:34:30"}
{"current_steps": 1850, "total_steps": 3325, "loss": 0.0218, "lr": 4.88975919150294e-05, "epoch": 2.779342723004695, "percentage": 55.64, "elapsed_time": "6:56:04", "remaining_time": "5:31:43"}
{"current_steps": 1850, "total_steps": 3325, "eval_loss": 0.03232384845614433, "epoch": 2.779342723004695, "percentage": 55.64, "elapsed_time": "6:59:45", "remaining_time": "5:34:40"}
{"current_steps": 1860, "total_steps": 3325, "loss": 0.022, "lr": 4.83727911786071e-05, "epoch": 2.7943661971830984, "percentage": 55.94, "elapsed_time": "7:01:23", "remaining_time": "5:31:54"}
{"current_steps": 1870, "total_steps": 3325, "loss": 0.0224, "lr": 4.7848169839470145e-05, "epoch": 2.8093896713615023, "percentage": 56.24, "elapsed_time": "7:02:59", "remaining_time": "5:29:06"}
{"current_steps": 1880, "total_steps": 3325, "loss": 0.021, "lr": 4.7323785736319244e-05, "epoch": 2.824413145539906, "percentage": 56.54, "elapsed_time": "7:04:34", "remaining_time": "5:26:20"}
{"current_steps": 1890, "total_steps": 3325, "loss": 0.0232, "lr": 4.679969668170024e-05, "epoch": 2.8394366197183096, "percentage": 56.84, "elapsed_time": "7:06:09", "remaining_time": "5:23:33"}
{"current_steps": 1900, "total_steps": 3325, "loss": 0.0179, "lr": 4.627596045563031e-05, "epoch": 2.8544600938967135, "percentage": 57.14, "elapsed_time": "7:07:44", "remaining_time": "5:20:48"}
{"current_steps": 1900, "total_steps": 3325, "eval_loss": 0.03188331797719002, "epoch": 2.8544600938967135, "percentage": 57.14, "elapsed_time": "7:11:25", "remaining_time": "5:23:34"}
{"current_steps": 1910, "total_steps": 3325, "loss": 0.0147, "lr": 4.575263479922783e-05, "epoch": 2.8694835680751174, "percentage": 57.44, "elapsed_time": "7:13:04", "remaining_time": "5:20:49"}
{"current_steps": 1920, "total_steps": 3325, "loss": 0.0118, "lr": 4.522977740834651e-05, "epoch": 2.8845070422535213, "percentage": 57.74, "elapsed_time": "7:14:39", "remaining_time": "5:18:04"}
{"current_steps": 1930, "total_steps": 3325, "loss": 0.0126, "lr": 4.4707445927214456e-05, "epoch": 2.8995305164319247, "percentage": 58.05, "elapsed_time": "7:16:14", "remaining_time": "5:15:18"}
{"current_steps": 1940, "total_steps": 3325, "loss": 0.0199, "lr": 4.4185697942079115e-05, "epoch": 2.9145539906103286, "percentage": 58.35, "elapsed_time": "7:17:49", "remaining_time": "5:12:34"}
{"current_steps": 1950, "total_steps": 3325, "loss": 0.0213, "lr": 4.366459097485832e-05, "epoch": 2.9295774647887325, "percentage": 58.65, "elapsed_time": "7:19:24", "remaining_time": "5:09:50"}
{"current_steps": 1950, "total_steps": 3325, "eval_loss": 0.03299366310238838, "epoch": 2.9295774647887325, "percentage": 58.65, "elapsed_time": "7:23:06", "remaining_time": "5:12:26"}
{"current_steps": 1960, "total_steps": 3325, "loss": 0.0231, "lr": 4.314418247679866e-05, "epoch": 2.9446009389671364, "percentage": 58.95, "elapsed_time": "7:24:44", "remaining_time": "5:09:44"}
{"current_steps": 1970, "total_steps": 3325, "loss": 0.0141, "lr": 4.26245298221416e-05, "epoch": 2.95962441314554, "percentage": 59.25, "elapsed_time": "7:26:19", "remaining_time": "5:06:59"}
{"current_steps": 1980, "total_steps": 3325, "loss": 0.0143, "lr": 4.2105690301798014e-05, "epoch": 2.9746478873239437, "percentage": 59.55, "elapsed_time": "7:27:54", "remaining_time": "5:04:15"}
{"current_steps": 1990, "total_steps": 3325, "loss": 0.0167, "lr": 4.158772111703194e-05, "epoch": 2.9896713615023476, "percentage": 59.85, "elapsed_time": "7:29:29", "remaining_time": "5:01:32"}
{"current_steps": 2000, "total_steps": 3325, "loss": 0.0104, "lr": 4.107067937315429e-05, "epoch": 3.004694835680751, "percentage": 60.15, "elapsed_time": "7:31:05", "remaining_time": "4:58:50"}
{"current_steps": 2000, "total_steps": 3325, "eval_loss": 0.03279737010598183, "epoch": 3.004694835680751, "percentage": 60.15, "elapsed_time": "7:34:47", "remaining_time": "5:01:17"}
{"current_steps": 2010, "total_steps": 3325, "loss": 0.01, "lr": 4.055462207322698e-05, "epoch": 3.019718309859155, "percentage": 60.45, "elapsed_time": "7:36:25", "remaining_time": "4:58:36"}
{"current_steps": 2020, "total_steps": 3325, "loss": 0.0101, "lr": 4.003960611177855e-05, "epoch": 3.034741784037559, "percentage": 60.75, "elapsed_time": "7:38:00", "remaining_time": "4:55:53"}
{"current_steps": 2030, "total_steps": 3325, "loss": 0.0094, "lr": 3.952568826853152e-05, "epoch": 3.0497652582159622, "percentage": 61.05, "elapsed_time": "7:39:35", "remaining_time": "4:53:11"}
{"current_steps": 2040, "total_steps": 3325, "loss": 0.0068, "lr": 3.901292520214256e-05, "epoch": 3.064788732394366, "percentage": 61.35, "elapsed_time": "7:41:10", "remaining_time": "4:50:29"}
{"current_steps": 2050, "total_steps": 3325, "loss": 0.0097, "lr": 3.850137344395598e-05, "epoch": 3.07981220657277, "percentage": 61.65, "elapsed_time": "7:42:45", "remaining_time": "4:47:48"}
{"current_steps": 2050, "total_steps": 3325, "eval_loss": 0.035948336124420166, "epoch": 3.07981220657277, "percentage": 61.65, "elapsed_time": "7:46:27", "remaining_time": "4:50:06"}
{"current_steps": 2060, "total_steps": 3325, "loss": 0.0076, "lr": 3.799108939177118e-05, "epoch": 3.094835680751174, "percentage": 61.95, "elapsed_time": "7:48:05", "remaining_time": "4:47:26"}
{"current_steps": 2070, "total_steps": 3325, "loss": 0.0148, "lr": 3.7482129303624934e-05, "epoch": 3.1098591549295773, "percentage": 62.26, "elapsed_time": "7:49:41", "remaining_time": "4:44:45"}
{"current_steps": 2080, "total_steps": 3325, "loss": 0.0115, "lr": 3.697454929158901e-05, "epoch": 3.124882629107981, "percentage": 62.56, "elapsed_time": "7:51:16", "remaining_time": "4:42:05"}
{"current_steps": 2090, "total_steps": 3325, "loss": 0.0147, "lr": 3.6468405315583854e-05, "epoch": 3.139906103286385, "percentage": 62.86, "elapsed_time": "7:52:52", "remaining_time": "4:39:25"}
{"current_steps": 2100, "total_steps": 3325, "loss": 0.0103, "lr": 3.59637531772092e-05, "epoch": 3.1549295774647885, "percentage": 63.16, "elapsed_time": "7:54:27", "remaining_time": "4:36:46"}
{"current_steps": 2100, "total_steps": 3325, "eval_loss": 0.0363353006541729, "epoch": 3.1549295774647885, "percentage": 63.16, "elapsed_time": "7:58:08", "remaining_time": "4:38:54"}
{"current_steps": 2110, "total_steps": 3325, "loss": 0.0082, "lr": 3.546064851359192e-05, "epoch": 3.1699530516431924, "percentage": 63.46, "elapsed_time": "7:59:47", "remaining_time": "4:36:16"}
{"current_steps": 2120, "total_steps": 3325, "loss": 0.0097, "lr": 3.495914679125212e-05, "epoch": 3.1849765258215963, "percentage": 63.76, "elapsed_time": "8:01:22", "remaining_time": "4:33:36"}
{"current_steps": 2130, "total_steps": 3325, "loss": 0.0103, "lr": 3.445930329998819e-05, "epoch": 3.2, "percentage": 64.06, "elapsed_time": "8:02:57", "remaining_time": "4:30:57"}
{"current_steps": 2140, "total_steps": 3325, "loss": 0.0142, "lr": 3.396117314678097e-05, "epoch": 3.2150234741784036, "percentage": 64.36, "elapsed_time": "8:04:33", "remaining_time": "4:28:19"}
{"current_steps": 2150, "total_steps": 3325, "loss": 0.0131, "lr": 3.3464811249718474e-05, "epoch": 3.2300469483568075, "percentage": 64.66, "elapsed_time": "8:06:08", "remaining_time": "4:25:40"}
{"current_steps": 2150, "total_steps": 3325, "eval_loss": 0.03587425872683525, "epoch": 3.2300469483568075, "percentage": 64.66, "elapsed_time": "8:09:49", "remaining_time": "4:27:41"}
{"current_steps": 2160, "total_steps": 3325, "loss": 0.0126, "lr": 3.297027233194114e-05, "epoch": 3.2450704225352114, "percentage": 64.96, "elapsed_time": "8:11:28", "remaining_time": "4:25:04"}
{"current_steps": 2170, "total_steps": 3325, "loss": 0.0081, "lr": 3.2477610915608704e-05, "epoch": 3.260093896713615, "percentage": 65.26, "elapsed_time": "8:13:04", "remaining_time": "4:22:26"}
{"current_steps": 2180, "total_steps": 3325, "loss": 0.0091, "lr": 3.1986881315889315e-05, "epoch": 3.2751173708920187, "percentage": 65.56, "elapsed_time": "8:14:39", "remaining_time": "4:19:48"}
{"current_steps": 2190, "total_steps": 3325, "loss": 0.0155, "lr": 3.149813763497124e-05, "epoch": 3.2901408450704226, "percentage": 65.86, "elapsed_time": "8:16:14", "remaining_time": "4:17:11"}
{"current_steps": 2200, "total_steps": 3325, "loss": 0.0149, "lr": 3.101143375609818e-05, "epoch": 3.3051643192488265, "percentage": 66.17, "elapsed_time": "8:17:50", "remaining_time": "4:14:34"}
{"current_steps": 2200, "total_steps": 3325, "eval_loss": 0.036220405250787735, "epoch": 3.3051643192488265, "percentage": 66.17, "elapsed_time": "8:21:32", "remaining_time": "4:16:28"}
{"current_steps": 2210, "total_steps": 3325, "loss": 0.0129, "lr": 3.0526823337628915e-05, "epoch": 3.32018779342723, "percentage": 66.47, "elapsed_time": "8:23:09", "remaining_time": "4:13:51"}
{"current_steps": 2220, "total_steps": 3325, "loss": 0.0098, "lr": 3.004435980712129e-05, "epoch": 3.335211267605634, "percentage": 66.77, "elapsed_time": "8:24:44", "remaining_time": "4:11:14"}
{"current_steps": 2230, "total_steps": 3325, "loss": 0.0128, "lr": 2.9564096355442116e-05, "epoch": 3.3502347417840377, "percentage": 67.07, "elapsed_time": "8:26:20", "remaining_time": "4:08:37"}
{"current_steps": 2240, "total_steps": 3325, "loss": 0.0137, "lr": 2.9086085930902824e-05, "epoch": 3.365258215962441, "percentage": 67.37, "elapsed_time": "8:27:55", "remaining_time": "4:06:01"}
{"current_steps": 2250, "total_steps": 3325, "loss": 0.0083, "lr": 2.8610381233422058e-05, "epoch": 3.380281690140845, "percentage": 67.67, "elapsed_time": "8:29:30", "remaining_time": "4:03:26"}
{"current_steps": 2250, "total_steps": 3325, "eval_loss": 0.03651123493909836, "epoch": 3.380281690140845, "percentage": 67.67, "elapsed_time": "8:33:12", "remaining_time": "4:05:12"}
{"current_steps": 2260, "total_steps": 3325, "loss": 0.0098, "lr": 2.8137034708715592e-05, "epoch": 3.395305164319249, "percentage": 67.97, "elapsed_time": "8:34:50", "remaining_time": "4:02:36"}
{"current_steps": 2270, "total_steps": 3325, "loss": 0.0096, "lr": 2.7666098542514273e-05, "epoch": 3.4103286384976528, "percentage": 68.27, "elapsed_time": "8:36:25", "remaining_time": "4:00:00"}
{"current_steps": 2280, "total_steps": 3325, "loss": 0.0123, "lr": 2.719762465481055e-05, "epoch": 3.425352112676056, "percentage": 68.57, "elapsed_time": "8:38:00", "remaining_time": "3:57:25"}
{"current_steps": 2290, "total_steps": 3325, "loss": 0.0118, "lr": 2.6731664694134473e-05, "epoch": 3.44037558685446, "percentage": 68.87, "elapsed_time": "8:39:35", "remaining_time": "3:54:50"}
{"current_steps": 2300, "total_steps": 3325, "loss": 0.0115, "lr": 2.6268270031859476e-05, "epoch": 3.455399061032864, "percentage": 69.17, "elapsed_time": "8:41:10", "remaining_time": "3:52:15"}
{"current_steps": 2300, "total_steps": 3325, "eval_loss": 0.035897351801395416, "epoch": 3.455399061032864, "percentage": 69.17, "elapsed_time": "8:44:52", "remaining_time": "3:53:54"}
{"current_steps": 2310, "total_steps": 3325, "loss": 0.0097, "lr": 2.580749175653877e-05, "epoch": 3.4704225352112674, "percentage": 69.47, "elapsed_time": "8:46:31", "remaining_time": "3:51:20"}
{"current_steps": 2320, "total_steps": 3325, "loss": 0.0122, "lr": 2.5349380668272905e-05, "epoch": 3.4854460093896713, "percentage": 69.77, "elapsed_time": "8:48:06", "remaining_time": "3:48:46"}
{"current_steps": 2330, "total_steps": 3325, "loss": 0.0096, "lr": 2.489398727310908e-05, "epoch": 3.500469483568075, "percentage": 70.08, "elapsed_time": "8:49:41", "remaining_time": "3:46:12"}
{"current_steps": 2340, "total_steps": 3325, "loss": 0.0088, "lr": 2.4441361777473066e-05, "epoch": 3.5154929577464786, "percentage": 70.38, "elapsed_time": "8:51:17", "remaining_time": "3:43:38"}
{"current_steps": 2350, "total_steps": 3325, "loss": 0.0111, "lr": 2.3991554082633912e-05, "epoch": 3.5305164319248825, "percentage": 70.68, "elapsed_time": "8:52:52", "remaining_time": "3:41:05"}
{"current_steps": 2350, "total_steps": 3325, "eval_loss": 0.038713712245225906, "epoch": 3.5305164319248825, "percentage": 70.68, "elapsed_time": "8:56:33", "remaining_time": "3:42:37"}
{"current_steps": 2360, "total_steps": 3325, "loss": 0.009, "lr": 2.354461377920239e-05, "epoch": 3.5455399061032864, "percentage": 70.98, "elapsed_time": "8:58:11", "remaining_time": "3:40:04"}
{"current_steps": 2370, "total_steps": 3325, "loss": 0.0123, "lr": 2.3100590141663807e-05, "epoch": 3.5605633802816903, "percentage": 71.28, "elapsed_time": "8:59:47", "remaining_time": "3:37:30"}
{"current_steps": 2380, "total_steps": 3325, "loss": 0.0143, "lr": 2.265953212294551e-05, "epoch": 3.575586854460094, "percentage": 71.58, "elapsed_time": "9:01:22", "remaining_time": "3:34:57"}
{"current_steps": 2390, "total_steps": 3325, "loss": 0.015, "lr": 2.2221488349019903e-05, "epoch": 3.5906103286384976, "percentage": 71.88, "elapsed_time": "9:02:57", "remaining_time": "3:32:24"}
{"current_steps": 2400, "total_steps": 3325, "loss": 0.0094, "lr": 2.1786507113543457e-05, "epoch": 3.6056338028169015, "percentage": 72.18, "elapsed_time": "9:04:33", "remaining_time": "3:29:53"}
{"current_steps": 2400, "total_steps": 3325, "eval_loss": 0.03762707859277725, "epoch": 3.6056338028169015, "percentage": 72.18, "elapsed_time": "9:08:16", "remaining_time": "3:31:18"}
{"current_steps": 2410, "total_steps": 3325, "loss": 0.0149, "lr": 2.1354636372532523e-05, "epoch": 3.6206572769953054, "percentage": 72.48, "elapsed_time": "9:09:55", "remaining_time": "3:28:47"}
{"current_steps": 2420, "total_steps": 3325, "loss": 0.0106, "lr": 2.092592373907617e-05, "epoch": 3.635680751173709, "percentage": 72.78, "elapsed_time": "9:11:31", "remaining_time": "3:26:15"}
{"current_steps": 2430, "total_steps": 3325, "loss": 0.0079, "lr": 2.0500416478086932e-05, "epoch": 3.6507042253521127, "percentage": 73.08, "elapsed_time": "9:13:06", "remaining_time": "3:23:43"}
{"current_steps": 2440, "total_steps": 3325, "loss": 0.0106, "lr": 2.0078161501089954e-05, "epoch": 3.6657276995305166, "percentage": 73.38, "elapsed_time": "9:14:42", "remaining_time": "3:21:11"}
{"current_steps": 2450, "total_steps": 3325, "loss": 0.0051, "lr": 1.9659205361050982e-05, "epoch": 3.68075117370892, "percentage": 73.68, "elapsed_time": "9:16:17", "remaining_time": "3:18:40"}
{"current_steps": 2450, "total_steps": 3325, "eval_loss": 0.03760899230837822, "epoch": 3.68075117370892, "percentage": 73.68, "elapsed_time": "9:19:59", "remaining_time": "3:19:59"}
{"current_steps": 2460, "total_steps": 3325, "loss": 0.0143, "lr": 1.924359424724408e-05, "epoch": 3.695774647887324, "percentage": 73.98, "elapsed_time": "9:21:39", "remaining_time": "3:17:29"}
{"current_steps": 2470, "total_steps": 3325, "loss": 0.0129, "lr": 1.8831373980159296e-05, "epoch": 3.710798122065728, "percentage": 74.29, "elapsed_time": "9:23:14", "remaining_time": "3:14:58"}
{"current_steps": 2480, "total_steps": 3325, "loss": 0.0062, "lr": 1.8422590006450947e-05, "epoch": 3.7258215962441312, "percentage": 74.59, "elapsed_time": "9:24:49", "remaining_time": "3:12:27"}
{"current_steps": 2490, "total_steps": 3325, "loss": 0.0087, "lr": 1.801728739392731e-05, "epoch": 3.740845070422535, "percentage": 74.89, "elapsed_time": "9:26:24", "remaining_time": "3:09:56"}
{"current_steps": 2500, "total_steps": 3325, "loss": 0.0053, "lr": 1.7615510826581904e-05, "epoch": 3.755868544600939, "percentage": 75.19, "elapsed_time": "9:28:00", "remaining_time": "3:07:26"}
{"current_steps": 2500, "total_steps": 3325, "eval_loss": 0.03752273693680763, "epoch": 3.755868544600939, "percentage": 75.19, "elapsed_time": "9:31:42", "remaining_time": "3:08:39"}
{"current_steps": 2510, "total_steps": 3325, "loss": 0.0111, "lr": 1.7217304599667146e-05, "epoch": 3.770892018779343, "percentage": 75.49, "elapsed_time": "9:33:21", "remaining_time": "3:06:10"}
{"current_steps": 2520, "total_steps": 3325, "loss": 0.0101, "lr": 1.6822712614810893e-05, "epoch": 3.7859154929577463, "percentage": 75.79, "elapsed_time": "9:34:56", "remaining_time": "3:03:39"}
{"current_steps": 2530, "total_steps": 3325, "loss": 0.0093, "lr": 1.643177837517631e-05, "epoch": 3.80093896713615, "percentage": 76.09, "elapsed_time": "9:36:32", "remaining_time": "3:01:09"}
{"current_steps": 2540, "total_steps": 3325, "loss": 0.0125, "lr": 1.6044544980665767e-05, "epoch": 3.815962441314554, "percentage": 76.39, "elapsed_time": "9:38:08", "remaining_time": "2:58:40"}
{"current_steps": 2550, "total_steps": 3325, "loss": 0.0078, "lr": 1.5661055123169126e-05, "epoch": 3.830985915492958, "percentage": 76.69, "elapsed_time": "9:39:44", "remaining_time": "2:56:11"}
{"current_steps": 2550, "total_steps": 3325, "eval_loss": 0.037725865840911865, "epoch": 3.830985915492958, "percentage": 76.69, "elapsed_time": "9:43:26", "remaining_time": "2:57:19"}
{"current_steps": 2560, "total_steps": 3325, "loss": 0.0118, "lr": 1.5281351081856974e-05, "epoch": 3.8460093896713614, "percentage": 76.99, "elapsed_time": "9:45:04", "remaining_time": "2:54:50"}
{"current_steps": 2570, "total_steps": 3325, "loss": 0.007, "lr": 1.4905474718519491e-05, "epoch": 3.8610328638497653, "percentage": 77.29, "elapsed_time": "9:46:40", "remaining_time": "2:52:20"}
{"current_steps": 2580, "total_steps": 3325, "loss": 0.0141, "lr": 1.453346747295119e-05, "epoch": 3.876056338028169, "percentage": 77.59, "elapsed_time": "9:48:15", "remaining_time": "2:49:52"}
{"current_steps": 2590, "total_steps": 3325, "loss": 0.0093, "lr": 1.4165370358382274e-05, "epoch": 3.8910798122065726, "percentage": 77.89, "elapsed_time": "9:49:51", "remaining_time": "2:47:23"}
{"current_steps": 2600, "total_steps": 3325, "loss": 0.0105, "lr": 1.3801223956956994e-05, "epoch": 3.9061032863849765, "percentage": 78.2, "elapsed_time": "9:51:26", "remaining_time": "2:44:55"}
{"current_steps": 2600, "total_steps": 3325, "eval_loss": 0.0372486487030983, "epoch": 3.9061032863849765, "percentage": 78.2, "elapsed_time": "9:55:08", "remaining_time": "2:45:57"}
{"current_steps": 2610, "total_steps": 3325, "loss": 0.0141, "lr": 1.344106841525946e-05, "epoch": 3.9211267605633804, "percentage": 78.5, "elapsed_time": "9:56:47", "remaining_time": "2:43:29"}
{"current_steps": 2620, "total_steps": 3325, "loss": 0.0085, "lr": 1.3084943439887659e-05, "epoch": 3.936150234741784, "percentage": 78.8, "elapsed_time": "9:58:22", "remaining_time": "2:41:00"}
{"current_steps": 2630, "total_steps": 3325, "loss": 0.0054, "lr": 1.273288829307579e-05, "epoch": 3.9511737089201877, "percentage": 79.1, "elapsed_time": "9:59:58", "remaining_time": "2:38:32"}
{"current_steps": 2640, "total_steps": 3325, "loss": 0.0078, "lr": 1.2384941788365622e-05, "epoch": 3.9661971830985916, "percentage": 79.4, "elapsed_time": "10:01:33", "remaining_time": "2:36:05"}
{"current_steps": 2650, "total_steps": 3325, "loss": 0.0105, "lr": 1.2041142286327477e-05, "epoch": 3.981220657276995, "percentage": 79.7, "elapsed_time": "10:03:08", "remaining_time": "2:33:37"}
{"current_steps": 2650, "total_steps": 3325, "eval_loss": 0.03705061599612236, "epoch": 3.981220657276995, "percentage": 79.7, "elapsed_time": "10:06:51", "remaining_time": "2:34:34"}
{"current_steps": 2660, "total_steps": 3325, "loss": 0.0085, "lr": 1.170152769033095e-05, "epoch": 3.996244131455399, "percentage": 80.0, "elapsed_time": "10:08:30", "remaining_time": "2:32:07"}
{"current_steps": 2670, "total_steps": 3325, "loss": 0.0057, "lr": 1.1366135442366127e-05, "epoch": 4.011267605633803, "percentage": 80.3, "elapsed_time": "10:10:05", "remaining_time": "2:29:40"}
{"current_steps": 2680, "total_steps": 3325, "loss": 0.0074, "lr": 1.103500251891571e-05, "epoch": 4.026291079812206, "percentage": 80.6, "elapsed_time": "10:11:41", "remaining_time": "2:27:12"}
{"current_steps": 2690, "total_steps": 3325, "loss": 0.0049, "lr": 1.0708165426878325e-05, "epoch": 4.041314553990611, "percentage": 80.9, "elapsed_time": "10:13:17", "remaining_time": "2:24:46"}
{"current_steps": 2700, "total_steps": 3325, "loss": 0.0064, "lr": 1.0385660199543812e-05, "epoch": 4.056338028169014, "percentage": 81.2, "elapsed_time": "10:14:52", "remaining_time": "2:22:20"}
{"current_steps": 2700, "total_steps": 3325, "eval_loss": 0.038154859095811844, "epoch": 4.056338028169014, "percentage": 81.2, "elapsed_time": "10:18:33", "remaining_time": "2:23:11"}
{"current_steps": 2710, "total_steps": 3325, "loss": 0.0052, "lr": 1.0067522392620537e-05, "epoch": 4.0713615023474174, "percentage": 81.5, "elapsed_time": "10:20:11", "remaining_time": "2:20:44"}
{"current_steps": 2720, "total_steps": 3325, "loss": 0.0075, "lr": 9.753787080315385e-06, "epoch": 4.086384976525822, "percentage": 81.8, "elapsed_time": "10:21:43", "remaining_time": "2:18:17"}
{"current_steps": 2730, "total_steps": 3325, "loss": 0.0059, "lr": 9.444488851467042e-06, "epoch": 4.101408450704225, "percentage": 82.11, "elapsed_time": "10:23:17", "remaining_time": "2:15:50"}
{"current_steps": 2740, "total_steps": 3325, "loss": 0.0042, "lr": 9.139661805732435e-06, "epoch": 4.1164319248826295, "percentage": 82.41, "elapsed_time": "10:24:52", "remaining_time": "2:13:24"}
{"current_steps": 2750, "total_steps": 3325, "loss": 0.0048, "lr": 8.839339549827397e-06, "epoch": 4.131455399061033, "percentage": 82.71, "elapsed_time": "10:26:28", "remaining_time": "2:10:59"}
{"current_steps": 2750, "total_steps": 3325, "eval_loss": 0.03982308879494667, "epoch": 4.131455399061033, "percentage": 82.71, "elapsed_time": "10:30:10", "remaining_time": "2:11:45"}
{"current_steps": 2760, "total_steps": 3325, "loss": 0.0075, "lr": 8.543555193821634e-06, "epoch": 4.146478873239436, "percentage": 83.01, "elapsed_time": "10:31:48", "remaining_time": "2:09:20"}
{"current_steps": 2770, "total_steps": 3325, "loss": 0.0043, "lr": 8.252341347488251e-06, "epoch": 4.161502347417841, "percentage": 83.31, "elapsed_time": "10:33:23", "remaining_time": "2:06:54"}
{"current_steps": 2780, "total_steps": 3325, "loss": 0.008, "lr": 7.965730116708681e-06, "epoch": 4.176525821596244, "percentage": 83.61, "elapsed_time": "10:34:59", "remaining_time": "2:04:29"}
{"current_steps": 2790, "total_steps": 3325, "loss": 0.0031, "lr": 7.68375309993304e-06, "epoch": 4.191549295774648, "percentage": 83.91, "elapsed_time": "10:36:34", "remaining_time": "2:02:04"}
{"current_steps": 2800, "total_steps": 3325, "loss": 0.0065, "lr": 7.406441384696372e-06, "epoch": 4.206572769953052, "percentage": 84.21, "elapsed_time": "10:38:10", "remaining_time": "1:59:39"}
{"current_steps": 2800, "total_steps": 3325, "eval_loss": 0.04066622629761696, "epoch": 4.206572769953052, "percentage": 84.21, "elapsed_time": "10:41:52", "remaining_time": "2:00:21"}
{"current_steps": 2810, "total_steps": 3325, "loss": 0.0095, "lr": 7.133825544191464e-06, "epoch": 4.221596244131455, "percentage": 84.51, "elapsed_time": "10:43:31", "remaining_time": "1:57:56"}
{"current_steps": 2820, "total_steps": 3325, "loss": 0.004, "lr": 6.865935633897996e-06, "epoch": 4.236619718309859, "percentage": 84.81, "elapsed_time": "10:45:06", "remaining_time": "1:55:31"}
{"current_steps": 2830, "total_steps": 3325, "loss": 0.0064, "lr": 6.602801188269081e-06, "epoch": 4.251643192488263, "percentage": 85.11, "elapsed_time": "10:46:41", "remaining_time": "1:53:06"}
{"current_steps": 2840, "total_steps": 3325, "loss": 0.0047, "lr": 6.344451217475183e-06, "epoch": 4.266666666666667, "percentage": 85.41, "elapsed_time": "10:48:17", "remaining_time": "1:50:42"}
{"current_steps": 2850, "total_steps": 3325, "loss": 0.0031, "lr": 6.090914204205655e-06, "epoch": 4.28169014084507, "percentage": 85.71, "elapsed_time": "10:49:52", "remaining_time": "1:48:18"}
{"current_steps": 2850, "total_steps": 3325, "eval_loss": 0.04169252887368202, "epoch": 4.28169014084507, "percentage": 85.71, "elapsed_time": "10:53:34", "remaining_time": "1:48:55"}
{"current_steps": 2860, "total_steps": 3325, "loss": 0.0052, "lr": 5.842218100528679e-06, "epoch": 4.296713615023474, "percentage": 86.02, "elapsed_time": "10:55:12", "remaining_time": "1:46:31"}
{"current_steps": 2870, "total_steps": 3325, "loss": 0.0046, "lr": 5.598390324809555e-06, "epoch": 4.311737089201878, "percentage": 86.32, "elapsed_time": "10:56:48", "remaining_time": "1:44:07"}
{"current_steps": 2880, "total_steps": 3325, "loss": 0.0089, "lr": 5.359457758687841e-06, "epoch": 4.326760563380281, "percentage": 86.62, "elapsed_time": "10:58:23", "remaining_time": "1:41:43"}
{"current_steps": 2890, "total_steps": 3325, "loss": 0.0057, "lr": 5.125446744113743e-06, "epoch": 4.341784037558686, "percentage": 86.92, "elapsed_time": "10:59:58", "remaining_time": "1:39:20"}
{"current_steps": 2900, "total_steps": 3325, "loss": 0.0028, "lr": 4.896383080443934e-06, "epoch": 4.356807511737089, "percentage": 87.22, "elapsed_time": "11:01:34", "remaining_time": "1:36:57"}
{"current_steps": 2900, "total_steps": 3325, "eval_loss": 0.041960135102272034, "epoch": 4.356807511737089, "percentage": 87.22, "elapsed_time": "11:05:16", "remaining_time": "1:37:29"}
{"current_steps": 2910, "total_steps": 3325, "loss": 0.0056, "lr": 4.672292021597174e-06, "epoch": 4.371830985915493, "percentage": 87.52, "elapsed_time": "11:06:55", "remaining_time": "1:35:06"}
{"current_steps": 2920, "total_steps": 3325, "loss": 0.0064, "lr": 4.4531982732702145e-06, "epoch": 4.386854460093897, "percentage": 87.82, "elapsed_time": "11:08:30", "remaining_time": "1:32:43"}
{"current_steps": 2930, "total_steps": 3325, "loss": 0.0039, "lr": 4.239125990213883e-06, "epoch": 4.4018779342723, "percentage": 88.12, "elapsed_time": "11:10:05", "remaining_time": "1:30:20"}
{"current_steps": 2940, "total_steps": 3325, "loss": 0.0016, "lr": 4.030098773570174e-06, "epoch": 4.416901408450705, "percentage": 88.42, "elapsed_time": "11:11:41", "remaining_time": "1:27:57"}
{"current_steps": 2950, "total_steps": 3325, "loss": 0.0043, "lr": 3.826139668270234e-06, "epoch": 4.431924882629108, "percentage": 88.72, "elapsed_time": "11:13:16", "remaining_time": "1:25:35"}
{"current_steps": 2950, "total_steps": 3325, "eval_loss": 0.042107198387384415, "epoch": 4.431924882629108, "percentage": 88.72, "elapsed_time": "11:16:59", "remaining_time": "1:26:03"}
{"current_steps": 2960, "total_steps": 3325, "loss": 0.0049, "lr": 3.6272711604936504e-06, "epoch": 4.446948356807511, "percentage": 89.02, "elapsed_time": "11:18:37", "remaining_time": "1:23:40"}
{"current_steps": 2970, "total_steps": 3325, "loss": 0.0053, "lr": 3.433515175189428e-06, "epoch": 4.461971830985916, "percentage": 89.32, "elapsed_time": "11:20:13", "remaining_time": "1:21:18"}
{"current_steps": 2980, "total_steps": 3325, "loss": 0.0077, "lr": 3.2448930736588e-06, "epoch": 4.476995305164319, "percentage": 89.62, "elapsed_time": "11:21:48", "remaining_time": "1:18:56"}
{"current_steps": 2990, "total_steps": 3325, "loss": 0.0038, "lr": 3.061425651200117e-06, "epoch": 4.492018779342723, "percentage": 89.92, "elapsed_time": "11:23:24", "remaining_time": "1:16:34"}
{"current_steps": 3000, "total_steps": 3325, "loss": 0.0048, "lr": 2.883133134816296e-06, "epoch": 4.507042253521127, "percentage": 90.23, "elapsed_time": "11:25:00", "remaining_time": "1:14:12"}
{"current_steps": 3000, "total_steps": 3325, "eval_loss": 0.04240112751722336, "epoch": 4.507042253521127, "percentage": 90.23, "elapsed_time": "11:28:42", "remaining_time": "1:14:36"}
{"current_steps": 3010, "total_steps": 3325, "loss": 0.0046, "lr": 2.7100351809847326e-06, "epoch": 4.52206572769953, "percentage": 90.53, "elapsed_time": "11:30:20", "remaining_time": "1:12:14"}
{"current_steps": 3020, "total_steps": 3325, "loss": 0.0028, "lr": 2.542150873490251e-06, "epoch": 4.537089201877935, "percentage": 90.83, "elapsed_time": "11:31:56", "remaining_time": "1:09:52"}
{"current_steps": 3030, "total_steps": 3325, "loss": 0.0055, "lr": 2.3794987213211383e-06, "epoch": 4.552112676056338, "percentage": 91.13, "elapsed_time": "11:33:31", "remaining_time": "1:07:31"}
{"current_steps": 3040, "total_steps": 3325, "loss": 0.0035, "lr": 2.222096656628547e-06, "epoch": 4.567136150234742, "percentage": 91.43, "elapsed_time": "11:35:07", "remaining_time": "1:05:10"}
{"current_steps": 3050, "total_steps": 3325, "loss": 0.0038, "lr": 2.0699620327495174e-06, "epoch": 4.582159624413146, "percentage": 91.73, "elapsed_time": "11:36:43", "remaining_time": "1:02:49"}
{"current_steps": 3050, "total_steps": 3325, "eval_loss": 0.04280637204647064, "epoch": 4.582159624413146, "percentage": 91.73, "elapsed_time": "11:40:25", "remaining_time": "1:03:09"}
{"current_steps": 3060, "total_steps": 3325, "loss": 0.0025, "lr": 1.9231116222937996e-06, "epoch": 4.597183098591549, "percentage": 92.03, "elapsed_time": "11:42:04", "remaining_time": "1:00:48"}
{"current_steps": 3070, "total_steps": 3325, "loss": 0.0045, "lr": 1.7815616152946523e-06, "epoch": 4.612206572769953, "percentage": 92.33, "elapsed_time": "11:43:40", "remaining_time": "0:58:26"}
{"current_steps": 3080, "total_steps": 3325, "loss": 0.004, "lr": 1.6453276174240195e-06, "epoch": 4.627230046948357, "percentage": 92.63, "elapsed_time": "11:45:16", "remaining_time": "0:56:06"}
{"current_steps": 3090, "total_steps": 3325, "loss": 0.0043, "lr": 1.5144246482719114e-06, "epoch": 4.642253521126761, "percentage": 92.93, "elapsed_time": "11:46:51", "remaining_time": "0:53:45"}
{"current_steps": 3100, "total_steps": 3325, "loss": 0.0041, "lr": 1.3888671396905805e-06, "epoch": 4.657276995305164, "percentage": 93.23, "elapsed_time": "11:48:26", "remaining_time": "0:51:25"}
{"current_steps": 3100, "total_steps": 3325, "eval_loss": 0.0430205836892128, "epoch": 4.657276995305164, "percentage": 93.23, "elapsed_time": "11:52:08", "remaining_time": "0:51:41"}
{"current_steps": 3110, "total_steps": 3325, "loss": 0.0082, "lr": 1.2686689342034431e-06, "epoch": 4.672300469483568, "percentage": 93.53, "elapsed_time": "11:53:47", "remaining_time": "0:49:20"}
{"current_steps": 3120, "total_steps": 3325, "loss": 0.0054, "lr": 1.1538432834789227e-06, "epoch": 4.687323943661972, "percentage": 93.83, "elapsed_time": "11:55:23", "remaining_time": "0:47:00"}
{"current_steps": 3130, "total_steps": 3325, "loss": 0.0049, "lr": 1.044402846869491e-06, "epoch": 4.702347417840375, "percentage": 94.14, "elapsed_time": "11:56:59", "remaining_time": "0:44:40"}
{"current_steps": 3140, "total_steps": 3325, "loss": 0.0034, "lr": 9.403596900160073e-07, "epoch": 4.71737089201878, "percentage": 94.44, "elapsed_time": "11:58:34", "remaining_time": "0:42:20"}
{"current_steps": 3150, "total_steps": 3325, "loss": 0.0066, "lr": 8.417252835174749e-07, "epoch": 4.732394366197183, "percentage": 94.74, "elapsed_time": "12:00:09", "remaining_time": "0:40:00"}
{"current_steps": 3150, "total_steps": 3325, "eval_loss": 0.043101657181978226, "epoch": 4.732394366197183, "percentage": 94.74, "elapsed_time": "12:03:52", "remaining_time": "0:40:12"}
{"current_steps": 3160, "total_steps": 3325, "loss": 0.0073, "lr": 7.48510501666455e-07, "epoch": 4.7474178403755865, "percentage": 95.04, "elapsed_time": "12:05:31", "remaining_time": "0:37:53"}
{"current_steps": 3170, "total_steps": 3325, "loss": 0.0067, "lr": 6.607256212501578e-07, "epoch": 4.762441314553991, "percentage": 95.34, "elapsed_time": "12:07:06", "remaining_time": "0:35:33"}
{"current_steps": 3180, "total_steps": 3325, "loss": 0.0054, "lr": 5.783803204174654e-07, "epoch": 4.777464788732394, "percentage": 95.64, "elapsed_time": "12:08:42", "remaining_time": "0:33:13"}
{"current_steps": 3190, "total_steps": 3325, "loss": 0.0051, "lr": 5.014836776119358e-07, "epoch": 4.792488262910798, "percentage": 95.94, "elapsed_time": "12:10:17", "remaining_time": "0:30:54"}
{"current_steps": 3200, "total_steps": 3325, "loss": 0.003, "lr": 4.300441705708924e-07, "epoch": 4.807511737089202, "percentage": 96.24, "elapsed_time": "12:11:53", "remaining_time": "0:28:35"}
{"current_steps": 3200, "total_steps": 3325, "eval_loss": 0.043016187846660614, "epoch": 4.807511737089202, "percentage": 96.24, "elapsed_time": "12:15:36", "remaining_time": "0:28:44"}
{"current_steps": 3210, "total_steps": 3325, "loss": 0.0042, "lr": 3.6406967539078796e-07, "epoch": 4.822535211267605, "percentage": 96.54, "elapsed_time": "12:17:15", "remaining_time": "0:26:24"}
{"current_steps": 3220, "total_steps": 3325, "loss": 0.0056, "lr": 3.0356746565887715e-07, "epoch": 4.83755868544601, "percentage": 96.84, "elapsed_time": "12:18:51", "remaining_time": "0:24:05"}
{"current_steps": 3230, "total_steps": 3325, "loss": 0.0048, "lr": 2.485442116513026e-07, "epoch": 4.852582159624413, "percentage": 97.14, "elapsed_time": "12:20:27", "remaining_time": "0:21:46"}
{"current_steps": 3240, "total_steps": 3325, "loss": 0.0079, "lr": 1.9900597959770507e-07, "epoch": 4.867605633802817, "percentage": 97.44, "elapsed_time": "12:22:03", "remaining_time": "0:19:28"}
{"current_steps": 3250, "total_steps": 3325, "loss": 0.0031, "lr": 1.5495823101245866e-07, "epoch": 4.882629107981221, "percentage": 97.74, "elapsed_time": "12:23:38", "remaining_time": "0:17:09"}
{"current_steps": 3250, "total_steps": 3325, "eval_loss": 0.042939841747283936, "epoch": 4.882629107981221, "percentage": 97.74, "elapsed_time": "12:27:21", "remaining_time": "0:17:14"}
{"current_steps": 3260, "total_steps": 3325, "loss": 0.0043, "lr": 1.164058220925135e-07, "epoch": 4.897652582159624, "percentage": 98.05, "elapsed_time": "12:29:00", "remaining_time": "0:14:56"}
{"current_steps": 3270, "total_steps": 3325, "loss": 0.0074, "lr": 8.335300318201844e-08, "epoch": 4.912676056338028, "percentage": 98.35, "elapsed_time": "12:30:35", "remaining_time": "0:12:37"}
{"current_steps": 3280, "total_steps": 3325, "loss": 0.005, "lr": 5.5803418303745917e-08, "epoch": 4.927699530516432, "percentage": 98.65, "elapsed_time": "12:32:10", "remaining_time": "0:10:19"}
{"current_steps": 3290, "total_steps": 3325, "loss": 0.003, "lr": 3.3760104757313284e-08, "epoch": 4.942723004694836, "percentage": 98.95, "elapsed_time": "12:33:46", "remaining_time": "0:08:01"}
{"current_steps": 3300, "total_steps": 3325, "loss": 0.0046, "lr": 1.7225492784345156e-08, "epoch": 4.957746478873239, "percentage": 99.25, "elapsed_time": "12:35:22", "remaining_time": "0:05:43"}
{"current_steps": 3300, "total_steps": 3325, "eval_loss": 0.042976681143045425, "epoch": 4.957746478873239, "percentage": 99.25, "elapsed_time": "12:39:04", "remaining_time": "0:05:45"}
{"current_steps": 3310, "total_steps": 3325, "loss": 0.0058, "lr": 6.201405300532148e-09, "epoch": 4.972769953051643, "percentage": 99.55, "elapsed_time": "12:40:44", "remaining_time": "0:03:26"}
{"current_steps": 3320, "total_steps": 3325, "loss": 0.0043, "lr": 6.890576946805282e-10, "epoch": 4.987793427230047, "percentage": 99.85, "elapsed_time": "12:42:20", "remaining_time": "0:01:08"}
{"current_steps": 3325, "total_steps": 3325, "epoch": 4.995305164319249, "percentage": 100.0, "elapsed_time": "12:43:12", "remaining_time": "0:00:00"}