chchen's picture
Training in progress, step 3225
deb479f verified
{"current_steps": 10, "total_steps": 3225, "loss": 1.3797, "lr": 3.0959752321981426e-06, "epoch": 0.015503875968992248, "percentage": 0.31, "elapsed_time": "0:01:17", "remaining_time": "6:55:05"}
{"current_steps": 20, "total_steps": 3225, "loss": 1.277, "lr": 6.191950464396285e-06, "epoch": 0.031007751937984496, "percentage": 0.62, "elapsed_time": "0:02:33", "remaining_time": "6:49:18"}
{"current_steps": 30, "total_steps": 3225, "loss": 1.1112, "lr": 9.287925696594429e-06, "epoch": 0.046511627906976744, "percentage": 0.93, "elapsed_time": "0:03:49", "remaining_time": "6:47:01"}
{"current_steps": 40, "total_steps": 3225, "loss": 0.7071, "lr": 1.238390092879257e-05, "epoch": 0.06201550387596899, "percentage": 1.24, "elapsed_time": "0:05:05", "remaining_time": "6:45:12"}
{"current_steps": 50, "total_steps": 3225, "loss": 0.4179, "lr": 1.5479876160990712e-05, "epoch": 0.07751937984496124, "percentage": 1.55, "elapsed_time": "0:06:21", "remaining_time": "6:43:47"}
{"current_steps": 50, "total_steps": 3225, "eval_loss": 0.34340164065361023, "epoch": 0.07751937984496124, "percentage": 1.55, "elapsed_time": "0:08:55", "remaining_time": "9:26:25"}
{"current_steps": 60, "total_steps": 3225, "loss": 0.3034, "lr": 1.8575851393188857e-05, "epoch": 0.09302325581395349, "percentage": 1.86, "elapsed_time": "0:10:15", "remaining_time": "9:01:01"}
{"current_steps": 70, "total_steps": 3225, "loss": 0.2055, "lr": 2.1671826625387e-05, "epoch": 0.10852713178294573, "percentage": 2.17, "elapsed_time": "0:11:31", "remaining_time": "8:39:23"}
{"current_steps": 80, "total_steps": 3225, "loss": 0.1391, "lr": 2.476780185758514e-05, "epoch": 0.12403100775193798, "percentage": 2.48, "elapsed_time": "0:12:47", "remaining_time": "8:22:57"}
{"current_steps": 90, "total_steps": 3225, "loss": 0.1286, "lr": 2.7863777089783283e-05, "epoch": 0.13953488372093023, "percentage": 2.79, "elapsed_time": "0:14:03", "remaining_time": "8:09:55"}
{"current_steps": 100, "total_steps": 3225, "loss": 0.0933, "lr": 3.0959752321981425e-05, "epoch": 0.15503875968992248, "percentage": 3.1, "elapsed_time": "0:15:20", "remaining_time": "7:59:15"}
{"current_steps": 100, "total_steps": 3225, "eval_loss": 0.08204219490289688, "epoch": 0.15503875968992248, "percentage": 3.1, "elapsed_time": "0:17:53", "remaining_time": "9:19:16"}
{"current_steps": 110, "total_steps": 3225, "loss": 0.1012, "lr": 3.4055727554179566e-05, "epoch": 0.17054263565891473, "percentage": 3.41, "elapsed_time": "0:19:15", "remaining_time": "9:05:34"}
{"current_steps": 120, "total_steps": 3225, "loss": 0.0923, "lr": 3.7151702786377715e-05, "epoch": 0.18604651162790697, "percentage": 3.72, "elapsed_time": "0:20:32", "remaining_time": "8:51:18"}
{"current_steps": 130, "total_steps": 3225, "loss": 0.1105, "lr": 4.024767801857585e-05, "epoch": 0.20155038759689922, "percentage": 4.03, "elapsed_time": "0:21:47", "remaining_time": "8:38:56"}
{"current_steps": 140, "total_steps": 3225, "loss": 0.0983, "lr": 4.3343653250774e-05, "epoch": 0.21705426356589147, "percentage": 4.34, "elapsed_time": "0:23:03", "remaining_time": "8:28:11"}
{"current_steps": 150, "total_steps": 3225, "loss": 0.0662, "lr": 4.6439628482972134e-05, "epoch": 0.23255813953488372, "percentage": 4.65, "elapsed_time": "0:24:19", "remaining_time": "8:18:41"}
{"current_steps": 150, "total_steps": 3225, "eval_loss": 0.0667666420340538, "epoch": 0.23255813953488372, "percentage": 4.65, "elapsed_time": "0:26:53", "remaining_time": "9:11:09"}
{"current_steps": 160, "total_steps": 3225, "loss": 0.079, "lr": 4.953560371517028e-05, "epoch": 0.24806201550387597, "percentage": 4.96, "elapsed_time": "0:28:13", "remaining_time": "9:00:34"}
{"current_steps": 170, "total_steps": 3225, "loss": 0.0753, "lr": 5.2631578947368424e-05, "epoch": 0.26356589147286824, "percentage": 5.27, "elapsed_time": "0:29:29", "remaining_time": "8:49:52"}
{"current_steps": 180, "total_steps": 3225, "loss": 0.0691, "lr": 5.5727554179566566e-05, "epoch": 0.27906976744186046, "percentage": 5.58, "elapsed_time": "0:30:45", "remaining_time": "8:40:13"}
{"current_steps": 190, "total_steps": 3225, "loss": 0.0706, "lr": 5.882352941176471e-05, "epoch": 0.29457364341085274, "percentage": 5.89, "elapsed_time": "0:32:01", "remaining_time": "8:31:27"}
{"current_steps": 200, "total_steps": 3225, "loss": 0.0584, "lr": 6.191950464396285e-05, "epoch": 0.31007751937984496, "percentage": 6.2, "elapsed_time": "0:33:16", "remaining_time": "8:23:23"}
{"current_steps": 200, "total_steps": 3225, "eval_loss": 0.058870695531368256, "epoch": 0.31007751937984496, "percentage": 6.2, "elapsed_time": "0:35:50", "remaining_time": "9:01:58"}
{"current_steps": 210, "total_steps": 3225, "loss": 0.0659, "lr": 6.501547987616098e-05, "epoch": 0.32558139534883723, "percentage": 6.51, "elapsed_time": "0:37:10", "remaining_time": "8:53:37"}
{"current_steps": 220, "total_steps": 3225, "loss": 0.0563, "lr": 6.811145510835913e-05, "epoch": 0.34108527131782945, "percentage": 6.82, "elapsed_time": "0:38:25", "remaining_time": "8:44:54"}
{"current_steps": 230, "total_steps": 3225, "loss": 0.06, "lr": 7.120743034055728e-05, "epoch": 0.35658914728682173, "percentage": 7.13, "elapsed_time": "0:39:41", "remaining_time": "8:36:47"}
{"current_steps": 240, "total_steps": 3225, "loss": 0.0844, "lr": 7.430340557275543e-05, "epoch": 0.37209302325581395, "percentage": 7.44, "elapsed_time": "0:40:57", "remaining_time": "8:29:20"}
{"current_steps": 250, "total_steps": 3225, "loss": 0.0666, "lr": 7.739938080495357e-05, "epoch": 0.3875968992248062, "percentage": 7.75, "elapsed_time": "0:42:13", "remaining_time": "8:22:26"}
{"current_steps": 250, "total_steps": 3225, "eval_loss": 0.05274031311273575, "epoch": 0.3875968992248062, "percentage": 7.75, "elapsed_time": "0:44:46", "remaining_time": "8:52:49"}
{"current_steps": 260, "total_steps": 3225, "loss": 0.0531, "lr": 8.04953560371517e-05, "epoch": 0.40310077519379844, "percentage": 8.06, "elapsed_time": "0:46:07", "remaining_time": "8:45:55"}
{"current_steps": 270, "total_steps": 3225, "loss": 0.0542, "lr": 8.359133126934985e-05, "epoch": 0.4186046511627907, "percentage": 8.37, "elapsed_time": "0:47:23", "remaining_time": "8:38:36"}
{"current_steps": 280, "total_steps": 3225, "loss": 0.0835, "lr": 8.6687306501548e-05, "epoch": 0.43410852713178294, "percentage": 8.68, "elapsed_time": "0:48:38", "remaining_time": "8:31:40"}
{"current_steps": 290, "total_steps": 3225, "loss": 0.0561, "lr": 8.978328173374613e-05, "epoch": 0.4496124031007752, "percentage": 8.99, "elapsed_time": "0:49:54", "remaining_time": "8:25:09"}
{"current_steps": 300, "total_steps": 3225, "loss": 0.0448, "lr": 9.287925696594427e-05, "epoch": 0.46511627906976744, "percentage": 9.3, "elapsed_time": "0:51:10", "remaining_time": "8:18:59"}
{"current_steps": 300, "total_steps": 3225, "eval_loss": 0.052115585654973984, "epoch": 0.46511627906976744, "percentage": 9.3, "elapsed_time": "0:53:43", "remaining_time": "8:43:52"}
{"current_steps": 310, "total_steps": 3225, "loss": 0.0597, "lr": 9.597523219814242e-05, "epoch": 0.4806201550387597, "percentage": 9.61, "elapsed_time": "0:55:04", "remaining_time": "8:37:48"}
{"current_steps": 320, "total_steps": 3225, "loss": 0.0406, "lr": 9.907120743034056e-05, "epoch": 0.49612403100775193, "percentage": 9.92, "elapsed_time": "0:56:20", "remaining_time": "8:31:24"}
{"current_steps": 330, "total_steps": 3225, "loss": 0.0499, "lr": 9.999856438185238e-05, "epoch": 0.5116279069767442, "percentage": 10.23, "elapsed_time": "0:57:35", "remaining_time": "8:25:15"}
{"current_steps": 340, "total_steps": 3225, "loss": 0.0558, "lr": 9.999153298122152e-05, "epoch": 0.5271317829457365, "percentage": 10.54, "elapsed_time": "0:58:51", "remaining_time": "8:19:23"}
{"current_steps": 350, "total_steps": 3225, "loss": 0.0474, "lr": 9.997864293614054e-05, "epoch": 0.5426356589147286, "percentage": 10.85, "elapsed_time": "1:00:07", "remaining_time": "8:13:49"}
{"current_steps": 350, "total_steps": 3225, "eval_loss": 0.048954565078020096, "epoch": 0.5426356589147286, "percentage": 10.85, "elapsed_time": "1:02:40", "remaining_time": "8:34:48"}
{"current_steps": 360, "total_steps": 3225, "loss": 0.0462, "lr": 9.995989575722902e-05, "epoch": 0.5581395348837209, "percentage": 11.16, "elapsed_time": "1:04:00", "remaining_time": "8:29:26"}
{"current_steps": 370, "total_steps": 3225, "loss": 0.063, "lr": 9.993529364152018e-05, "epoch": 0.5736434108527132, "percentage": 11.47, "elapsed_time": "1:05:16", "remaining_time": "8:23:43"}
{"current_steps": 380, "total_steps": 3225, "loss": 0.0606, "lr": 9.990483947220319e-05, "epoch": 0.5891472868217055, "percentage": 11.78, "elapsed_time": "1:06:32", "remaining_time": "8:18:12"}
{"current_steps": 390, "total_steps": 3225, "loss": 0.0531, "lr": 9.986853681828546e-05, "epoch": 0.6046511627906976, "percentage": 12.09, "elapsed_time": "1:07:48", "remaining_time": "8:12:56"}
{"current_steps": 400, "total_steps": 3225, "loss": 0.0546, "lr": 9.982638993417425e-05, "epoch": 0.6201550387596899, "percentage": 12.4, "elapsed_time": "1:09:04", "remaining_time": "8:07:50"}
{"current_steps": 400, "total_steps": 3225, "eval_loss": 0.0430627167224884, "epoch": 0.6201550387596899, "percentage": 12.4, "elapsed_time": "1:11:37", "remaining_time": "8:25:52"}
{"current_steps": 410, "total_steps": 3225, "loss": 0.0509, "lr": 9.977840375917817e-05, "epoch": 0.6356589147286822, "percentage": 12.71, "elapsed_time": "1:12:58", "remaining_time": "8:20:58"}
{"current_steps": 420, "total_steps": 3225, "loss": 0.06, "lr": 9.972458391692827e-05, "epoch": 0.6511627906976745, "percentage": 13.02, "elapsed_time": "1:14:14", "remaining_time": "8:15:47"}
{"current_steps": 430, "total_steps": 3225, "loss": 0.0507, "lr": 9.966493671471904e-05, "epoch": 0.6666666666666666, "percentage": 13.33, "elapsed_time": "1:15:30", "remaining_time": "8:10:46"}
{"current_steps": 440, "total_steps": 3225, "loss": 0.0597, "lr": 9.959946914276922e-05, "epoch": 0.6821705426356589, "percentage": 13.64, "elapsed_time": "1:16:46", "remaining_time": "8:05:56"}
{"current_steps": 450, "total_steps": 3225, "loss": 0.0432, "lr": 9.952818887340257e-05, "epoch": 0.6976744186046512, "percentage": 13.95, "elapsed_time": "1:18:02", "remaining_time": "8:01:12"}
{"current_steps": 450, "total_steps": 3225, "eval_loss": 0.03933678939938545, "epoch": 0.6976744186046512, "percentage": 13.95, "elapsed_time": "1:20:35", "remaining_time": "8:16:57"}
{"current_steps": 460, "total_steps": 3225, "loss": 0.0521, "lr": 9.945110426014878e-05, "epoch": 0.7131782945736435, "percentage": 14.26, "elapsed_time": "1:21:55", "remaining_time": "8:12:26"}
{"current_steps": 470, "total_steps": 3225, "loss": 0.0431, "lr": 9.936822433676444e-05, "epoch": 0.7286821705426356, "percentage": 14.57, "elapsed_time": "1:23:11", "remaining_time": "8:07:37"}
{"current_steps": 480, "total_steps": 3225, "loss": 0.0555, "lr": 9.927955881617444e-05, "epoch": 0.7441860465116279, "percentage": 14.88, "elapsed_time": "1:24:27", "remaining_time": "8:02:58"}
{"current_steps": 490, "total_steps": 3225, "loss": 0.0435, "lr": 9.918511808933358e-05, "epoch": 0.7596899224806202, "percentage": 15.19, "elapsed_time": "1:25:43", "remaining_time": "7:58:27"}
{"current_steps": 500, "total_steps": 3225, "loss": 0.0526, "lr": 9.908491322400885e-05, "epoch": 0.7751937984496124, "percentage": 15.5, "elapsed_time": "1:26:59", "remaining_time": "7:54:06"}
{"current_steps": 500, "total_steps": 3225, "eval_loss": 0.04014807939529419, "epoch": 0.7751937984496124, "percentage": 15.5, "elapsed_time": "1:29:32", "remaining_time": "8:08:00"}
{"current_steps": 510, "total_steps": 3225, "loss": 0.0466, "lr": 9.897895596348247e-05, "epoch": 0.7906976744186046, "percentage": 15.81, "elapsed_time": "1:30:52", "remaining_time": "8:03:45"}
{"current_steps": 520, "total_steps": 3225, "loss": 0.0425, "lr": 9.886725872517552e-05, "epoch": 0.8062015503875969, "percentage": 16.12, "elapsed_time": "1:32:08", "remaining_time": "7:59:16"}
{"current_steps": 530, "total_steps": 3225, "loss": 0.0495, "lr": 9.874983459919277e-05, "epoch": 0.8217054263565892, "percentage": 16.43, "elapsed_time": "1:33:23", "remaining_time": "7:54:54"}
{"current_steps": 540, "total_steps": 3225, "loss": 0.0371, "lr": 9.862669734678867e-05, "epoch": 0.8372093023255814, "percentage": 16.74, "elapsed_time": "1:34:39", "remaining_time": "7:50:40"}
{"current_steps": 550, "total_steps": 3225, "loss": 0.0506, "lr": 9.849786139875452e-05, "epoch": 0.8527131782945736, "percentage": 17.05, "elapsed_time": "1:35:55", "remaining_time": "7:46:32"}
{"current_steps": 550, "total_steps": 3225, "eval_loss": 0.0400104746222496, "epoch": 0.8527131782945736, "percentage": 17.05, "elapsed_time": "1:38:28", "remaining_time": "7:58:56"}
{"current_steps": 560, "total_steps": 3225, "loss": 0.0531, "lr": 9.836334185372738e-05, "epoch": 0.8682170542635659, "percentage": 17.36, "elapsed_time": "1:39:48", "remaining_time": "7:54:59"}
{"current_steps": 570, "total_steps": 3225, "loss": 0.0434, "lr": 9.822315447642056e-05, "epoch": 0.8837209302325582, "percentage": 17.67, "elapsed_time": "1:41:04", "remaining_time": "7:50:47"}
{"current_steps": 580, "total_steps": 3225, "loss": 0.0372, "lr": 9.807731569577615e-05, "epoch": 0.8992248062015504, "percentage": 17.98, "elapsed_time": "1:42:20", "remaining_time": "7:46:41"}
{"current_steps": 590, "total_steps": 3225, "loss": 0.0397, "lr": 9.792584260303964e-05, "epoch": 0.9147286821705426, "percentage": 18.29, "elapsed_time": "1:43:35", "remaining_time": "7:42:41"}
{"current_steps": 600, "total_steps": 3225, "loss": 0.0622, "lr": 9.776875294975698e-05, "epoch": 0.9302325581395349, "percentage": 18.6, "elapsed_time": "1:44:51", "remaining_time": "7:38:46"}
{"current_steps": 600, "total_steps": 3225, "eval_loss": 0.04188257083296776, "epoch": 0.9302325581395349, "percentage": 18.6, "elapsed_time": "1:47:24", "remaining_time": "7:49:56"}
{"current_steps": 610, "total_steps": 3225, "loss": 0.0454, "lr": 9.760606514569416e-05, "epoch": 0.9457364341085271, "percentage": 18.91, "elapsed_time": "1:48:45", "remaining_time": "7:46:13"}
{"current_steps": 620, "total_steps": 3225, "loss": 0.0413, "lr": 9.743779825667984e-05, "epoch": 0.9612403100775194, "percentage": 19.22, "elapsed_time": "1:50:01", "remaining_time": "7:42:17"}
{"current_steps": 630, "total_steps": 3225, "loss": 0.0404, "lr": 9.726397200237085e-05, "epoch": 0.9767441860465116, "percentage": 19.53, "elapsed_time": "1:51:17", "remaining_time": "7:38:24"}
{"current_steps": 640, "total_steps": 3225, "loss": 0.0549, "lr": 9.708460675394126e-05, "epoch": 0.9922480620155039, "percentage": 19.84, "elapsed_time": "1:52:33", "remaining_time": "7:34:37"}
{"current_steps": 650, "total_steps": 3225, "loss": 0.0363, "lr": 9.689972353169499e-05, "epoch": 1.0077519379844961, "percentage": 20.16, "elapsed_time": "1:53:49", "remaining_time": "7:30:54"}
{"current_steps": 650, "total_steps": 3225, "eval_loss": 0.03797760233283043, "epoch": 1.0077519379844961, "percentage": 20.16, "elapsed_time": "1:56:22", "remaining_time": "7:41:00"}
{"current_steps": 660, "total_steps": 3225, "loss": 0.0361, "lr": 9.67093440026024e-05, "epoch": 1.0232558139534884, "percentage": 20.47, "elapsed_time": "1:57:42", "remaining_time": "7:37:27"}
{"current_steps": 670, "total_steps": 3225, "loss": 0.0465, "lr": 9.651349047776112e-05, "epoch": 1.0387596899224807, "percentage": 20.78, "elapsed_time": "1:58:58", "remaining_time": "7:33:42"}
{"current_steps": 680, "total_steps": 3225, "loss": 0.0278, "lr": 9.631218590978126e-05, "epoch": 1.054263565891473, "percentage": 21.09, "elapsed_time": "2:00:14", "remaining_time": "7:30:00"}
{"current_steps": 690, "total_steps": 3225, "loss": 0.0408, "lr": 9.610545389009562e-05, "epoch": 1.069767441860465, "percentage": 21.4, "elapsed_time": "2:01:30", "remaining_time": "7:26:23"}
{"current_steps": 700, "total_steps": 3225, "loss": 0.032, "lr": 9.589331864619486e-05, "epoch": 1.0852713178294573, "percentage": 21.71, "elapsed_time": "2:02:45", "remaining_time": "7:22:50"}
{"current_steps": 700, "total_steps": 3225, "eval_loss": 0.03769562765955925, "epoch": 1.0852713178294573, "percentage": 21.71, "elapsed_time": "2:05:19", "remaining_time": "7:32:02"}
{"current_steps": 710, "total_steps": 3225, "loss": 0.0285, "lr": 9.567580503878833e-05, "epoch": 1.1007751937984496, "percentage": 22.02, "elapsed_time": "2:06:39", "remaining_time": "7:28:37"}
{"current_steps": 720, "total_steps": 3225, "loss": 0.035, "lr": 9.545293855889043e-05, "epoch": 1.1162790697674418, "percentage": 22.33, "elapsed_time": "2:07:54", "remaining_time": "7:25:02"}
{"current_steps": 730, "total_steps": 3225, "loss": 0.0371, "lr": 9.522474532483337e-05, "epoch": 1.1317829457364341, "percentage": 22.64, "elapsed_time": "2:09:10", "remaining_time": "7:21:29"}
{"current_steps": 740, "total_steps": 3225, "loss": 0.0443, "lr": 9.499125207920622e-05, "epoch": 1.1472868217054264, "percentage": 22.95, "elapsed_time": "2:10:26", "remaining_time": "7:18:02"}
{"current_steps": 750, "total_steps": 3225, "loss": 0.0361, "lr": 9.475248618572096e-05, "epoch": 1.1627906976744187, "percentage": 23.26, "elapsed_time": "2:11:42", "remaining_time": "7:14:37"}
{"current_steps": 750, "total_steps": 3225, "eval_loss": 0.04346157982945442, "epoch": 1.1627906976744187, "percentage": 23.26, "elapsed_time": "2:14:15", "remaining_time": "7:23:02"}
{"current_steps": 760, "total_steps": 3225, "loss": 0.0396, "lr": 9.45084756260055e-05, "epoch": 1.178294573643411, "percentage": 23.57, "elapsed_time": "2:15:37", "remaining_time": "7:19:51"}
{"current_steps": 770, "total_steps": 3225, "loss": 0.0336, "lr": 9.425924899632458e-05, "epoch": 1.193798449612403, "percentage": 23.88, "elapsed_time": "2:16:52", "remaining_time": "7:16:24"}
{"current_steps": 780, "total_steps": 3225, "loss": 0.0299, "lr": 9.400483550422845e-05, "epoch": 1.2093023255813953, "percentage": 24.19, "elapsed_time": "2:18:08", "remaining_time": "7:13:01"}
{"current_steps": 790, "total_steps": 3225, "loss": 0.0434, "lr": 9.374526496512992e-05, "epoch": 1.2248062015503876, "percentage": 24.5, "elapsed_time": "2:19:24", "remaining_time": "7:09:41"}
{"current_steps": 800, "total_steps": 3225, "loss": 0.0256, "lr": 9.348056779881025e-05, "epoch": 1.2403100775193798, "percentage": 24.81, "elapsed_time": "2:20:40", "remaining_time": "7:06:25"}
{"current_steps": 800, "total_steps": 3225, "eval_loss": 0.03653433918952942, "epoch": 1.2403100775193798, "percentage": 24.81, "elapsed_time": "2:23:13", "remaining_time": "7:14:09"}
{"current_steps": 810, "total_steps": 3225, "loss": 0.0277, "lr": 9.321077502585417e-05, "epoch": 1.255813953488372, "percentage": 25.12, "elapsed_time": "2:24:33", "remaining_time": "7:11:00"}
{"current_steps": 820, "total_steps": 3225, "loss": 0.0389, "lr": 9.293591826401451e-05, "epoch": 1.2713178294573644, "percentage": 25.43, "elapsed_time": "2:25:49", "remaining_time": "7:07:41"}
{"current_steps": 830, "total_steps": 3225, "loss": 0.0368, "lr": 9.265602972450673e-05, "epoch": 1.2868217054263567, "percentage": 25.74, "elapsed_time": "2:27:05", "remaining_time": "7:04:25"}
{"current_steps": 840, "total_steps": 3225, "loss": 0.0324, "lr": 9.237114220823413e-05, "epoch": 1.302325581395349, "percentage": 26.05, "elapsed_time": "2:28:21", "remaining_time": "7:01:13"}
{"current_steps": 850, "total_steps": 3225, "loss": 0.0361, "lr": 9.208128910194377e-05, "epoch": 1.3178294573643412, "percentage": 26.36, "elapsed_time": "2:29:37", "remaining_time": "6:58:03"}
{"current_steps": 850, "total_steps": 3225, "eval_loss": 0.0356990210711956, "epoch": 1.3178294573643412, "percentage": 26.36, "elapsed_time": "2:32:10", "remaining_time": "7:05:10"}
{"current_steps": 860, "total_steps": 3225, "loss": 0.0408, "lr": 9.17865043743138e-05, "epoch": 1.3333333333333333, "percentage": 26.67, "elapsed_time": "2:33:30", "remaining_time": "7:02:07"}
{"current_steps": 870, "total_steps": 3225, "loss": 0.0276, "lr": 9.148682257197245e-05, "epoch": 1.3488372093023255, "percentage": 26.98, "elapsed_time": "2:34:45", "remaining_time": "6:58:55"}
{"current_steps": 880, "total_steps": 3225, "loss": 0.0485, "lr": 9.118227881544967e-05, "epoch": 1.3643410852713178, "percentage": 27.29, "elapsed_time": "2:36:01", "remaining_time": "6:55:46"}
{"current_steps": 890, "total_steps": 3225, "loss": 0.0324, "lr": 9.087290879506104e-05, "epoch": 1.37984496124031, "percentage": 27.6, "elapsed_time": "2:37:17", "remaining_time": "6:52:40"}
{"current_steps": 900, "total_steps": 3225, "loss": 0.0428, "lr": 9.055874876672519e-05, "epoch": 1.3953488372093024, "percentage": 27.91, "elapsed_time": "2:38:33", "remaining_time": "6:49:36"}
{"current_steps": 900, "total_steps": 3225, "eval_loss": 0.03691842034459114, "epoch": 1.3953488372093024, "percentage": 27.91, "elapsed_time": "2:41:06", "remaining_time": "6:56:11"}
{"current_steps": 910, "total_steps": 3225, "loss": 0.0314, "lr": 9.023983554771492e-05, "epoch": 1.4108527131782946, "percentage": 28.22, "elapsed_time": "2:42:30", "remaining_time": "6:53:23"}
{"current_steps": 920, "total_steps": 3225, "loss": 0.0287, "lr": 8.991620651234242e-05, "epoch": 1.4263565891472867, "percentage": 28.53, "elapsed_time": "2:43:45", "remaining_time": "6:50:17"}
{"current_steps": 930, "total_steps": 3225, "loss": 0.0325, "lr": 8.958789958757928e-05, "epoch": 1.441860465116279, "percentage": 28.84, "elapsed_time": "2:45:01", "remaining_time": "6:47:14"}
{"current_steps": 940, "total_steps": 3225, "loss": 0.034, "lr": 8.925495324861177e-05, "epoch": 1.4573643410852712, "percentage": 29.15, "elapsed_time": "2:46:17", "remaining_time": "6:44:14"}
{"current_steps": 950, "total_steps": 3225, "loss": 0.0423, "lr": 8.89174065143318e-05, "epoch": 1.4728682170542635, "percentage": 29.46, "elapsed_time": "2:47:33", "remaining_time": "6:41:16"}
{"current_steps": 950, "total_steps": 3225, "eval_loss": 0.036706067621707916, "epoch": 1.4728682170542635, "percentage": 29.46, "elapsed_time": "2:50:06", "remaining_time": "6:47:22"}
{"current_steps": 960, "total_steps": 3225, "loss": 0.0351, "lr": 8.857529894276412e-05, "epoch": 1.4883720930232558, "percentage": 29.77, "elapsed_time": "2:51:26", "remaining_time": "6:44:30"}
{"current_steps": 970, "total_steps": 3225, "loss": 0.032, "lr": 8.822867062643059e-05, "epoch": 1.503875968992248, "percentage": 30.08, "elapsed_time": "2:52:42", "remaining_time": "6:41:29"}
{"current_steps": 980, "total_steps": 3225, "loss": 0.0348, "lr": 8.78775621876514e-05, "epoch": 1.5193798449612403, "percentage": 30.39, "elapsed_time": "2:53:58", "remaining_time": "6:38:31"}
{"current_steps": 990, "total_steps": 3225, "loss": 0.03, "lr": 8.752201477378459e-05, "epoch": 1.5348837209302326, "percentage": 30.7, "elapsed_time": "2:55:13", "remaining_time": "6:35:36"}
{"current_steps": 1000, "total_steps": 3225, "loss": 0.0298, "lr": 8.716207005240382e-05, "epoch": 1.550387596899225, "percentage": 31.01, "elapsed_time": "2:56:29", "remaining_time": "6:32:42"}
{"current_steps": 1000, "total_steps": 3225, "eval_loss": 0.03816875442862511, "epoch": 1.550387596899225, "percentage": 31.01, "elapsed_time": "2:59:02", "remaining_time": "6:38:22"}
{"current_steps": 1010, "total_steps": 3225, "loss": 0.0411, "lr": 8.679777020641525e-05, "epoch": 1.5658914728682172, "percentage": 31.32, "elapsed_time": "3:00:23", "remaining_time": "6:35:36"}
{"current_steps": 1020, "total_steps": 3225, "loss": 0.0428, "lr": 8.6429157929114e-05, "epoch": 1.5813953488372094, "percentage": 31.63, "elapsed_time": "3:01:38", "remaining_time": "6:32:40"}
{"current_steps": 1030, "total_steps": 3225, "loss": 0.0352, "lr": 8.60562764191808e-05, "epoch": 1.5968992248062015, "percentage": 31.94, "elapsed_time": "3:02:55", "remaining_time": "6:29:48"}
{"current_steps": 1040, "total_steps": 3225, "loss": 0.0338, "lr": 8.567916937561944e-05, "epoch": 1.6124031007751938, "percentage": 32.25, "elapsed_time": "3:04:10", "remaining_time": "6:26:57"}
{"current_steps": 1050, "total_steps": 3225, "loss": 0.0357, "lr": 8.529788099263554e-05, "epoch": 1.627906976744186, "percentage": 32.56, "elapsed_time": "3:05:26", "remaining_time": "6:24:07"}
{"current_steps": 1050, "total_steps": 3225, "eval_loss": 0.03658704832196236, "epoch": 1.627906976744186, "percentage": 32.56, "elapsed_time": "3:07:59", "remaining_time": "6:29:24"}
{"current_steps": 1060, "total_steps": 3225, "loss": 0.0297, "lr": 8.49124559544573e-05, "epoch": 1.6434108527131783, "percentage": 32.87, "elapsed_time": "3:09:20", "remaining_time": "6:26:42"}
{"current_steps": 1070, "total_steps": 3225, "loss": 0.0299, "lr": 8.452293943009889e-05, "epoch": 1.6589147286821704, "percentage": 33.18, "elapsed_time": "3:10:36", "remaining_time": "6:23:52"}
{"current_steps": 1080, "total_steps": 3225, "loss": 0.0374, "lr": 8.412937706806692e-05, "epoch": 1.6744186046511627, "percentage": 33.49, "elapsed_time": "3:11:52", "remaining_time": "6:21:05"}
{"current_steps": 1090, "total_steps": 3225, "loss": 0.0355, "lr": 8.373181499101077e-05, "epoch": 1.689922480620155, "percentage": 33.8, "elapsed_time": "3:13:08", "remaining_time": "6:18:17"}
{"current_steps": 1100, "total_steps": 3225, "loss": 0.0271, "lr": 8.333029979031737e-05, "epoch": 1.7054263565891472, "percentage": 34.11, "elapsed_time": "3:14:23", "remaining_time": "6:15:32"}
{"current_steps": 1100, "total_steps": 3225, "eval_loss": 0.037469059228897095, "epoch": 1.7054263565891472, "percentage": 34.11, "elapsed_time": "3:16:56", "remaining_time": "6:20:27"}
{"current_steps": 1110, "total_steps": 3225, "loss": 0.0478, "lr": 8.292487852065104e-05, "epoch": 1.7209302325581395, "percentage": 34.42, "elapsed_time": "3:18:17", "remaining_time": "6:17:48"}
{"current_steps": 1120, "total_steps": 3225, "loss": 0.0342, "lr": 8.251559869443897e-05, "epoch": 1.7364341085271318, "percentage": 34.73, "elapsed_time": "3:19:33", "remaining_time": "6:15:03"}
{"current_steps": 1130, "total_steps": 3225, "loss": 0.0328, "lr": 8.210250827630313e-05, "epoch": 1.751937984496124, "percentage": 35.04, "elapsed_time": "3:20:48", "remaining_time": "6:12:17"}
{"current_steps": 1140, "total_steps": 3225, "loss": 0.029, "lr": 8.168565567743924e-05, "epoch": 1.7674418604651163, "percentage": 35.35, "elapsed_time": "3:22:04", "remaining_time": "6:09:34"}
{"current_steps": 1150, "total_steps": 3225, "loss": 0.0325, "lr": 8.126508974994321e-05, "epoch": 1.7829457364341086, "percentage": 35.66, "elapsed_time": "3:23:20", "remaining_time": "6:06:53"}
{"current_steps": 1150, "total_steps": 3225, "eval_loss": 0.03696885332465172, "epoch": 1.7829457364341086, "percentage": 35.66, "elapsed_time": "3:25:53", "remaining_time": "6:11:29"}
{"current_steps": 1160, "total_steps": 3225, "loss": 0.0202, "lr": 8.084085978108611e-05, "epoch": 1.7984496124031009, "percentage": 35.97, "elapsed_time": "3:27:13", "remaining_time": "6:08:53"}
{"current_steps": 1170, "total_steps": 3225, "loss": 0.0299, "lr": 8.041301548753804e-05, "epoch": 1.8139534883720931, "percentage": 36.28, "elapsed_time": "3:28:29", "remaining_time": "6:06:11"}
{"current_steps": 1180, "total_steps": 3225, "loss": 0.0248, "lr": 7.998160700954165e-05, "epoch": 1.8294573643410854, "percentage": 36.59, "elapsed_time": "3:29:45", "remaining_time": "6:03:30"}
{"current_steps": 1190, "total_steps": 3225, "loss": 0.0362, "lr": 7.954668490503621e-05, "epoch": 1.8449612403100775, "percentage": 36.9, "elapsed_time": "3:31:00", "remaining_time": "6:00:50"}
{"current_steps": 1200, "total_steps": 3225, "loss": 0.0328, "lr": 7.910830014373237e-05, "epoch": 1.8604651162790697, "percentage": 37.21, "elapsed_time": "3:32:16", "remaining_time": "5:58:13"}
{"current_steps": 1200, "total_steps": 3225, "eval_loss": 0.03458752855658531, "epoch": 1.8604651162790697, "percentage": 37.21, "elapsed_time": "3:34:50", "remaining_time": "6:02:32"}
{"current_steps": 1210, "total_steps": 3225, "loss": 0.0288, "lr": 7.866650410113907e-05, "epoch": 1.875968992248062, "percentage": 37.52, "elapsed_time": "3:36:10", "remaining_time": "5:59:59"}
{"current_steps": 1220, "total_steps": 3225, "loss": 0.038, "lr": 7.822134855254263e-05, "epoch": 1.8914728682170543, "percentage": 37.83, "elapsed_time": "3:37:25", "remaining_time": "5:57:20"}
{"current_steps": 1230, "total_steps": 3225, "loss": 0.0316, "lr": 7.777288566693905e-05, "epoch": 1.9069767441860463, "percentage": 38.14, "elapsed_time": "3:38:41", "remaining_time": "5:54:42"}
{"current_steps": 1240, "total_steps": 3225, "loss": 0.0345, "lr": 7.732116800092018e-05, "epoch": 1.9224806201550386, "percentage": 38.45, "elapsed_time": "3:39:57", "remaining_time": "5:52:06"}
{"current_steps": 1250, "total_steps": 3225, "loss": 0.0373, "lr": 7.686624849251453e-05, "epoch": 1.937984496124031, "percentage": 38.76, "elapsed_time": "3:41:13", "remaining_time": "5:49:32"}
{"current_steps": 1250, "total_steps": 3225, "eval_loss": 0.034580811858177185, "epoch": 1.937984496124031, "percentage": 38.76, "elapsed_time": "3:43:46", "remaining_time": "5:53:33"}
{"current_steps": 1260, "total_steps": 3225, "loss": 0.0427, "lr": 7.640818045498324e-05, "epoch": 1.9534883720930232, "percentage": 39.07, "elapsed_time": "3:45:06", "remaining_time": "5:51:03"}
{"current_steps": 1270, "total_steps": 3225, "loss": 0.0353, "lr": 7.59470175705722e-05, "epoch": 1.9689922480620154, "percentage": 39.38, "elapsed_time": "3:46:22", "remaining_time": "5:48:28"}
{"current_steps": 1280, "total_steps": 3225, "loss": 0.036, "lr": 7.548281388422088e-05, "epoch": 1.9844961240310077, "percentage": 39.69, "elapsed_time": "3:47:37", "remaining_time": "5:45:53"}
{"current_steps": 1290, "total_steps": 3225, "loss": 0.0282, "lr": 7.50156237972286e-05, "epoch": 2.0, "percentage": 40.0, "elapsed_time": "3:48:53", "remaining_time": "5:43:20"}
{"current_steps": 1300, "total_steps": 3225, "loss": 0.0219, "lr": 7.454550206087921e-05, "epoch": 2.0155038759689923, "percentage": 40.31, "elapsed_time": "3:50:09", "remaining_time": "5:40:48"}
{"current_steps": 1300, "total_steps": 3225, "eval_loss": 0.03509150817990303, "epoch": 2.0155038759689923, "percentage": 40.31, "elapsed_time": "3:52:42", "remaining_time": "5:44:35"}
{"current_steps": 1310, "total_steps": 3225, "loss": 0.0196, "lr": 7.407250377002451e-05, "epoch": 2.0310077519379846, "percentage": 40.62, "elapsed_time": "3:54:02", "remaining_time": "5:42:08"}
{"current_steps": 1320, "total_steps": 3225, "loss": 0.0237, "lr": 7.359668435662757e-05, "epoch": 2.046511627906977, "percentage": 40.93, "elapsed_time": "3:55:18", "remaining_time": "5:39:35"}
{"current_steps": 1330, "total_steps": 3225, "loss": 0.017, "lr": 7.311809958326659e-05, "epoch": 2.062015503875969, "percentage": 41.24, "elapsed_time": "3:56:34", "remaining_time": "5:37:04"}
{"current_steps": 1340, "total_steps": 3225, "loss": 0.0198, "lr": 7.263680553659983e-05, "epoch": 2.0775193798449614, "percentage": 41.55, "elapsed_time": "3:57:50", "remaining_time": "5:34:34"}
{"current_steps": 1350, "total_steps": 3225, "loss": 0.0179, "lr": 7.21528586207927e-05, "epoch": 2.0930232558139537, "percentage": 41.86, "elapsed_time": "3:59:06", "remaining_time": "5:32:05"}
{"current_steps": 1350, "total_steps": 3225, "eval_loss": 0.038030773401260376, "epoch": 2.0930232558139537, "percentage": 41.86, "elapsed_time": "4:01:39", "remaining_time": "5:35:37"}
{"current_steps": 1360, "total_steps": 3225, "loss": 0.0242, "lr": 7.166631555090759e-05, "epoch": 2.108527131782946, "percentage": 42.17, "elapsed_time": "4:02:59", "remaining_time": "5:33:12"}
{"current_steps": 1370, "total_steps": 3225, "loss": 0.0204, "lr": 7.117723334625731e-05, "epoch": 2.124031007751938, "percentage": 42.48, "elapsed_time": "4:04:15", "remaining_time": "5:30:43"}
{"current_steps": 1380, "total_steps": 3225, "loss": 0.0197, "lr": 7.068566932372278e-05, "epoch": 2.13953488372093, "percentage": 42.79, "elapsed_time": "4:05:31", "remaining_time": "5:28:14"}
{"current_steps": 1390, "total_steps": 3225, "loss": 0.0173, "lr": 7.019168109103599e-05, "epoch": 2.1550387596899223, "percentage": 43.1, "elapsed_time": "4:06:47", "remaining_time": "5:25:48"}
{"current_steps": 1400, "total_steps": 3225, "loss": 0.018, "lr": 6.96953265400287e-05, "epoch": 2.1705426356589146, "percentage": 43.41, "elapsed_time": "4:08:03", "remaining_time": "5:23:21"}
{"current_steps": 1400, "total_steps": 3225, "eval_loss": 0.03982286527752876, "epoch": 2.1705426356589146, "percentage": 43.41, "elapsed_time": "4:10:36", "remaining_time": "5:26:41"}
{"current_steps": 1410, "total_steps": 3225, "loss": 0.0157, "lr": 6.9196663839848e-05, "epoch": 2.186046511627907, "percentage": 43.72, "elapsed_time": "4:11:57", "remaining_time": "5:24:19"}
{"current_steps": 1420, "total_steps": 3225, "loss": 0.0164, "lr": 6.869575143013934e-05, "epoch": 2.201550387596899, "percentage": 44.03, "elapsed_time": "4:13:12", "remaining_time": "5:21:52"}
{"current_steps": 1430, "total_steps": 3225, "loss": 0.0305, "lr": 6.819264801419771e-05, "epoch": 2.2170542635658914, "percentage": 44.34, "elapsed_time": "4:14:28", "remaining_time": "5:19:26"}
{"current_steps": 1440, "total_steps": 3225, "loss": 0.0195, "lr": 6.768741255208813e-05, "epoch": 2.2325581395348837, "percentage": 44.65, "elapsed_time": "4:15:44", "remaining_time": "5:17:00"}
{"current_steps": 1450, "total_steps": 3225, "loss": 0.0203, "lr": 6.718010425373599e-05, "epoch": 2.248062015503876, "percentage": 44.96, "elapsed_time": "4:17:00", "remaining_time": "5:14:37"}
{"current_steps": 1450, "total_steps": 3225, "eval_loss": 0.038189034909009933, "epoch": 2.248062015503876, "percentage": 44.96, "elapsed_time": "4:19:33", "remaining_time": "5:17:44"}
{"current_steps": 1460, "total_steps": 3225, "loss": 0.0188, "lr": 6.667078257198796e-05, "epoch": 2.2635658914728682, "percentage": 45.27, "elapsed_time": "4:20:53", "remaining_time": "5:15:24"}
{"current_steps": 1470, "total_steps": 3225, "loss": 0.0163, "lr": 6.615950719564466e-05, "epoch": 2.2790697674418605, "percentage": 45.58, "elapsed_time": "4:22:09", "remaining_time": "5:12:59"}
{"current_steps": 1480, "total_steps": 3225, "loss": 0.0295, "lr": 6.56463380424655e-05, "epoch": 2.294573643410853, "percentage": 45.89, "elapsed_time": "4:23:26", "remaining_time": "5:10:36"}
{"current_steps": 1490, "total_steps": 3225, "loss": 0.0157, "lr": 6.51313352521468e-05, "epoch": 2.310077519379845, "percentage": 46.2, "elapsed_time": "4:24:41", "remaining_time": "5:08:13"}
{"current_steps": 1500, "total_steps": 3225, "loss": 0.0257, "lr": 6.46145591792738e-05, "epoch": 2.3255813953488373, "percentage": 46.51, "elapsed_time": "4:25:57", "remaining_time": "5:05:51"}
{"current_steps": 1500, "total_steps": 3225, "eval_loss": 0.040530748665332794, "epoch": 2.3255813953488373, "percentage": 46.51, "elapsed_time": "4:28:30", "remaining_time": "5:08:47"}
{"current_steps": 1510, "total_steps": 3225, "loss": 0.0165, "lr": 6.409607038624759e-05, "epoch": 2.3410852713178296, "percentage": 46.82, "elapsed_time": "4:29:50", "remaining_time": "5:06:28"}
{"current_steps": 1520, "total_steps": 3225, "loss": 0.028, "lr": 6.357592963618761e-05, "epoch": 2.356589147286822, "percentage": 47.13, "elapsed_time": "4:31:06", "remaining_time": "5:04:06"}
{"current_steps": 1530, "total_steps": 3225, "loss": 0.0211, "lr": 6.305419788581067e-05, "epoch": 2.3720930232558137, "percentage": 47.44, "elapsed_time": "4:32:22", "remaining_time": "5:01:44"}
{"current_steps": 1540, "total_steps": 3225, "loss": 0.0195, "lr": 6.253093627828724e-05, "epoch": 2.387596899224806, "percentage": 47.75, "elapsed_time": "4:33:37", "remaining_time": "4:59:23"}
{"current_steps": 1550, "total_steps": 3225, "loss": 0.0165, "lr": 6.200620613607597e-05, "epoch": 2.4031007751937983, "percentage": 48.06, "elapsed_time": "4:34:53", "remaining_time": "4:57:03"}
{"current_steps": 1550, "total_steps": 3225, "eval_loss": 0.0382014699280262, "epoch": 2.4031007751937983, "percentage": 48.06, "elapsed_time": "4:37:26", "remaining_time": "4:59:49"}
{"current_steps": 1560, "total_steps": 3225, "loss": 0.0202, "lr": 6.148006895373706e-05, "epoch": 2.4186046511627906, "percentage": 48.37, "elapsed_time": "4:38:46", "remaining_time": "4:57:32"}
{"current_steps": 1570, "total_steps": 3225, "loss": 0.0181, "lr": 6.0952586390725644e-05, "epoch": 2.434108527131783, "percentage": 48.68, "elapsed_time": "4:40:02", "remaining_time": "4:55:12"}
{"current_steps": 1580, "total_steps": 3225, "loss": 0.0179, "lr": 6.042382026416563e-05, "epoch": 2.449612403100775, "percentage": 48.99, "elapsed_time": "4:41:18", "remaining_time": "4:52:52"}
{"current_steps": 1590, "total_steps": 3225, "loss": 0.0203, "lr": 5.989383254160529e-05, "epoch": 2.4651162790697674, "percentage": 49.3, "elapsed_time": "4:42:34", "remaining_time": "4:50:33"}
{"current_steps": 1600, "total_steps": 3225, "loss": 0.0212, "lr": 5.936268533375506e-05, "epoch": 2.4806201550387597, "percentage": 49.61, "elapsed_time": "4:43:49", "remaining_time": "4:48:15"}
{"current_steps": 1600, "total_steps": 3225, "eval_loss": 0.037529706954956055, "epoch": 2.4806201550387597, "percentage": 49.61, "elapsed_time": "4:46:22", "remaining_time": "4:50:50"}
{"current_steps": 1610, "total_steps": 3225, "loss": 0.0262, "lr": 5.883044088720865e-05, "epoch": 2.496124031007752, "percentage": 49.92, "elapsed_time": "4:47:43", "remaining_time": "4:48:37"}
{"current_steps": 1620, "total_steps": 3225, "loss": 0.0235, "lr": 5.829716157714814e-05, "epoch": 2.511627906976744, "percentage": 50.23, "elapsed_time": "4:48:59", "remaining_time": "4:46:18"}
{"current_steps": 1630, "total_steps": 3225, "loss": 0.0279, "lr": 5.77629099000341e-05, "epoch": 2.5271317829457365, "percentage": 50.54, "elapsed_time": "4:50:15", "remaining_time": "4:44:01"}
{"current_steps": 1640, "total_steps": 3225, "loss": 0.0257, "lr": 5.7227748466281486e-05, "epoch": 2.5426356589147288, "percentage": 50.85, "elapsed_time": "4:51:31", "remaining_time": "4:41:44"}
{"current_steps": 1650, "total_steps": 3225, "loss": 0.0315, "lr": 5.669173999292208e-05, "epoch": 2.558139534883721, "percentage": 51.16, "elapsed_time": "4:52:47", "remaining_time": "4:39:28"}
{"current_steps": 1650, "total_steps": 3225, "eval_loss": 0.03725634142756462, "epoch": 2.558139534883721, "percentage": 51.16, "elapsed_time": "4:55:20", "remaining_time": "4:41:54"}
{"current_steps": 1660, "total_steps": 3225, "loss": 0.0232, "lr": 5.615494729625458e-05, "epoch": 2.5736434108527133, "percentage": 51.47, "elapsed_time": "4:56:40", "remaining_time": "4:39:42"}
{"current_steps": 1670, "total_steps": 3225, "loss": 0.0294, "lr": 5.561743328448296e-05, "epoch": 2.5891472868217056, "percentage": 51.78, "elapsed_time": "4:57:56", "remaining_time": "4:37:25"}
{"current_steps": 1680, "total_steps": 3225, "loss": 0.0197, "lr": 5.5079260950344035e-05, "epoch": 2.604651162790698, "percentage": 52.09, "elapsed_time": "4:59:12", "remaining_time": "4:35:09"}
{"current_steps": 1690, "total_steps": 3225, "loss": 0.0197, "lr": 5.454049336372531e-05, "epoch": 2.62015503875969, "percentage": 52.4, "elapsed_time": "5:00:28", "remaining_time": "4:32:54"}
{"current_steps": 1700, "total_steps": 3225, "loss": 0.0155, "lr": 5.4001193664273454e-05, "epoch": 2.6356589147286824, "percentage": 52.71, "elapsed_time": "5:01:44", "remaining_time": "4:30:40"}
{"current_steps": 1700, "total_steps": 3225, "eval_loss": 0.03793378174304962, "epoch": 2.6356589147286824, "percentage": 52.71, "elapsed_time": "5:04:17", "remaining_time": "4:32:57"}
{"current_steps": 1710, "total_steps": 3225, "loss": 0.0173, "lr": 5.346142505399495e-05, "epoch": 2.6511627906976747, "percentage": 53.02, "elapsed_time": "5:05:37", "remaining_time": "4:30:46"}
{"current_steps": 1720, "total_steps": 3225, "loss": 0.018, "lr": 5.292125078984925e-05, "epoch": 2.6666666666666665, "percentage": 53.33, "elapsed_time": "5:06:53", "remaining_time": "4:28:31"}
{"current_steps": 1730, "total_steps": 3225, "loss": 0.0184, "lr": 5.2380734176335425e-05, "epoch": 2.682170542635659, "percentage": 53.64, "elapsed_time": "5:08:09", "remaining_time": "4:26:17"}
{"current_steps": 1740, "total_steps": 3225, "loss": 0.0238, "lr": 5.183993855807343e-05, "epoch": 2.697674418604651, "percentage": 53.95, "elapsed_time": "5:09:25", "remaining_time": "4:24:04"}
{"current_steps": 1750, "total_steps": 3225, "loss": 0.0188, "lr": 5.1298927312380586e-05, "epoch": 2.7131782945736433, "percentage": 54.26, "elapsed_time": "5:10:40", "remaining_time": "4:21:51"}
{"current_steps": 1750, "total_steps": 3225, "eval_loss": 0.03785843402147293, "epoch": 2.7131782945736433, "percentage": 54.26, "elapsed_time": "5:13:13", "remaining_time": "4:24:00"}
{"current_steps": 1760, "total_steps": 3225, "loss": 0.0271, "lr": 5.075776384184411e-05, "epoch": 2.7286821705426356, "percentage": 54.57, "elapsed_time": "5:14:33", "remaining_time": "4:21:50"}
{"current_steps": 1770, "total_steps": 3225, "loss": 0.0249, "lr": 5.021651156689094e-05, "epoch": 2.744186046511628, "percentage": 54.88, "elapsed_time": "5:15:49", "remaining_time": "4:19:37"}
{"current_steps": 1780, "total_steps": 3225, "loss": 0.0234, "lr": 4.967523391835521e-05, "epoch": 2.75968992248062, "percentage": 55.19, "elapsed_time": "5:17:05", "remaining_time": "4:17:25"}
{"current_steps": 1790, "total_steps": 3225, "loss": 0.0253, "lr": 4.9133994330044644e-05, "epoch": 2.7751937984496124, "percentage": 55.5, "elapsed_time": "5:18:21", "remaining_time": "4:15:13"}
{"current_steps": 1800, "total_steps": 3225, "loss": 0.0195, "lr": 4.85928562313066e-05, "epoch": 2.7906976744186047, "percentage": 55.81, "elapsed_time": "5:19:37", "remaining_time": "4:13:01"}
{"current_steps": 1800, "total_steps": 3225, "eval_loss": 0.03974982351064682, "epoch": 2.7906976744186047, "percentage": 55.81, "elapsed_time": "5:22:10", "remaining_time": "4:15:02"}
{"current_steps": 1810, "total_steps": 3225, "loss": 0.017, "lr": 4.8051883039594616e-05, "epoch": 2.806201550387597, "percentage": 56.12, "elapsed_time": "5:23:30", "remaining_time": "4:12:54"}
{"current_steps": 1820, "total_steps": 3225, "loss": 0.0262, "lr": 4.751113815303624e-05, "epoch": 2.8217054263565893, "percentage": 56.43, "elapsed_time": "5:24:45", "remaining_time": "4:10:42"}
{"current_steps": 1830, "total_steps": 3225, "loss": 0.0187, "lr": 4.697068494300343e-05, "epoch": 2.8372093023255816, "percentage": 56.74, "elapsed_time": "5:26:01", "remaining_time": "4:08:31"}
{"current_steps": 1840, "total_steps": 3225, "loss": 0.0218, "lr": 4.6430586746685724e-05, "epoch": 2.8527131782945734, "percentage": 57.05, "elapsed_time": "5:27:17", "remaining_time": "4:06:21"}
{"current_steps": 1850, "total_steps": 3225, "loss": 0.0213, "lr": 4.589090685966758e-05, "epoch": 2.8682170542635657, "percentage": 57.36, "elapsed_time": "5:28:34", "remaining_time": "4:04:12"}
{"current_steps": 1850, "total_steps": 3225, "eval_loss": 0.0373452827334404, "epoch": 2.8682170542635657, "percentage": 57.36, "elapsed_time": "5:31:06", "remaining_time": "4:06:05"}
{"current_steps": 1860, "total_steps": 3225, "loss": 0.0331, "lr": 4.535170852851073e-05, "epoch": 2.883720930232558, "percentage": 57.67, "elapsed_time": "5:32:27", "remaining_time": "4:03:59"}
{"current_steps": 1870, "total_steps": 3225, "loss": 0.0188, "lr": 4.481305494334201e-05, "epoch": 2.89922480620155, "percentage": 57.98, "elapsed_time": "5:33:43", "remaining_time": "4:01:48"}
{"current_steps": 1880, "total_steps": 3225, "loss": 0.0194, "lr": 4.427500923044801e-05, "epoch": 2.9147286821705425, "percentage": 58.29, "elapsed_time": "5:34:59", "remaining_time": "3:59:39"}
{"current_steps": 1890, "total_steps": 3225, "loss": 0.0187, "lr": 4.373763444487705e-05, "epoch": 2.9302325581395348, "percentage": 58.6, "elapsed_time": "5:36:15", "remaining_time": "3:57:30"}
{"current_steps": 1900, "total_steps": 3225, "loss": 0.0171, "lr": 4.3200993563049725e-05, "epoch": 2.945736434108527, "percentage": 58.91, "elapsed_time": "5:37:30", "remaining_time": "3:55:22"}
{"current_steps": 1900, "total_steps": 3225, "eval_loss": 0.037420857697725296, "epoch": 2.945736434108527, "percentage": 58.91, "elapsed_time": "5:40:03", "remaining_time": "3:57:08"}
{"current_steps": 1910, "total_steps": 3225, "loss": 0.0206, "lr": 4.266514947537839e-05, "epoch": 2.9612403100775193, "percentage": 59.22, "elapsed_time": "5:41:24", "remaining_time": "3:55:02"}
{"current_steps": 1920, "total_steps": 3225, "loss": 0.0202, "lr": 4.2130164978896916e-05, "epoch": 2.9767441860465116, "percentage": 59.53, "elapsed_time": "5:42:39", "remaining_time": "3:52:54"}
{"current_steps": 1930, "total_steps": 3225, "loss": 0.0145, "lr": 4.159610276990137e-05, "epoch": 2.992248062015504, "percentage": 59.84, "elapsed_time": "5:43:55", "remaining_time": "3:50:46"}
{"current_steps": 1940, "total_steps": 3225, "loss": 0.0161, "lr": 4.106302543660235e-05, "epoch": 3.007751937984496, "percentage": 60.16, "elapsed_time": "5:45:11", "remaining_time": "3:48:38"}
{"current_steps": 1950, "total_steps": 3225, "loss": 0.0108, "lr": 4.053099545179028e-05, "epoch": 3.0232558139534884, "percentage": 60.47, "elapsed_time": "5:46:27", "remaining_time": "3:46:31"}
{"current_steps": 1950, "total_steps": 3225, "eval_loss": 0.03898989409208298, "epoch": 3.0232558139534884, "percentage": 60.47, "elapsed_time": "5:49:00", "remaining_time": "3:48:11"}
{"current_steps": 1960, "total_steps": 3225, "loss": 0.0127, "lr": 4.0000075165513845e-05, "epoch": 3.0387596899224807, "percentage": 60.78, "elapsed_time": "5:50:20", "remaining_time": "3:46:06"}
{"current_steps": 1970, "total_steps": 3225, "loss": 0.009, "lr": 3.9470326797773216e-05, "epoch": 3.054263565891473, "percentage": 61.09, "elapsed_time": "5:51:36", "remaining_time": "3:43:59"}
{"current_steps": 1980, "total_steps": 3225, "loss": 0.0098, "lr": 3.8941812431228166e-05, "epoch": 3.0697674418604652, "percentage": 61.4, "elapsed_time": "5:52:51", "remaining_time": "3:41:52"}
{"current_steps": 1990, "total_steps": 3225, "loss": 0.0128, "lr": 3.8414594003922515e-05, "epoch": 3.0852713178294575, "percentage": 61.71, "elapsed_time": "5:54:07", "remaining_time": "3:39:46"}
{"current_steps": 2000, "total_steps": 3225, "loss": 0.0125, "lr": 3.788873330202544e-05, "epoch": 3.10077519379845, "percentage": 62.02, "elapsed_time": "5:55:23", "remaining_time": "3:37:40"}
{"current_steps": 2000, "total_steps": 3225, "eval_loss": 0.04365137219429016, "epoch": 3.10077519379845, "percentage": 62.02, "elapsed_time": "5:57:56", "remaining_time": "3:39:14"}
{"current_steps": 2010, "total_steps": 3225, "loss": 0.0059, "lr": 3.736429195259051e-05, "epoch": 3.116279069767442, "percentage": 62.33, "elapsed_time": "5:59:17", "remaining_time": "3:37:10"}
{"current_steps": 2020, "total_steps": 3225, "loss": 0.0118, "lr": 3.684133141633358e-05, "epoch": 3.1317829457364343, "percentage": 62.64, "elapsed_time": "6:00:33", "remaining_time": "3:35:05"}
{"current_steps": 2030, "total_steps": 3225, "loss": 0.0099, "lr": 3.6319912980429846e-05, "epoch": 3.147286821705426, "percentage": 62.95, "elapsed_time": "6:01:49", "remaining_time": "3:32:59"}
{"current_steps": 2040, "total_steps": 3225, "loss": 0.0082, "lr": 3.580009775133168e-05, "epoch": 3.1627906976744184, "percentage": 63.26, "elapsed_time": "6:03:05", "remaining_time": "3:30:54"}
{"current_steps": 2050, "total_steps": 3225, "loss": 0.0046, "lr": 3.528194664760714e-05, "epoch": 3.1782945736434107, "percentage": 63.57, "elapsed_time": "6:04:21", "remaining_time": "3:28:50"}
{"current_steps": 2050, "total_steps": 3225, "eval_loss": 0.04593827575445175, "epoch": 3.1782945736434107, "percentage": 63.57, "elapsed_time": "6:06:53", "remaining_time": "3:30:17"}
{"current_steps": 2060, "total_steps": 3225, "loss": 0.0095, "lr": 3.476552039280096e-05, "epoch": 3.193798449612403, "percentage": 63.88, "elapsed_time": "6:08:13", "remaining_time": "3:28:14"}
{"current_steps": 2070, "total_steps": 3225, "loss": 0.007, "lr": 3.42508795083181e-05, "epoch": 3.2093023255813953, "percentage": 64.19, "elapsed_time": "6:09:29", "remaining_time": "3:26:10"}
{"current_steps": 2080, "total_steps": 3225, "loss": 0.0054, "lr": 3.373808430633106e-05, "epoch": 3.2248062015503876, "percentage": 64.5, "elapsed_time": "6:10:45", "remaining_time": "3:24:05"}
{"current_steps": 2090, "total_steps": 3225, "loss": 0.0103, "lr": 3.32271948827118e-05, "epoch": 3.24031007751938, "percentage": 64.81, "elapsed_time": "6:12:01", "remaining_time": "3:22:01"}
{"current_steps": 2100, "total_steps": 3225, "loss": 0.0059, "lr": 3.2718271109988863e-05, "epoch": 3.255813953488372, "percentage": 65.12, "elapsed_time": "6:13:16", "remaining_time": "3:19:58"}
{"current_steps": 2100, "total_steps": 3225, "eval_loss": 0.047852758318185806, "epoch": 3.255813953488372, "percentage": 65.12, "elapsed_time": "6:15:49", "remaining_time": "3:21:20"}
{"current_steps": 2110, "total_steps": 3225, "loss": 0.0069, "lr": 3.2211372630330835e-05, "epoch": 3.2713178294573644, "percentage": 65.43, "elapsed_time": "6:17:09", "remaining_time": "3:19:18"}
{"current_steps": 2120, "total_steps": 3225, "loss": 0.0124, "lr": 3.170655884855661e-05, "epoch": 3.2868217054263567, "percentage": 65.74, "elapsed_time": "6:18:24", "remaining_time": "3:17:14"}
{"current_steps": 2130, "total_steps": 3225, "loss": 0.0195, "lr": 3.120388892517368e-05, "epoch": 3.302325581395349, "percentage": 66.05, "elapsed_time": "6:19:40", "remaining_time": "3:15:10"}
{"current_steps": 2140, "total_steps": 3225, "loss": 0.0146, "lr": 3.070342176944494e-05, "epoch": 3.317829457364341, "percentage": 66.36, "elapsed_time": "6:20:55", "remaining_time": "3:13:08"}
{"current_steps": 2150, "total_steps": 3225, "loss": 0.0088, "lr": 3.0205216032484805e-05, "epoch": 3.3333333333333335, "percentage": 66.67, "elapsed_time": "6:22:11", "remaining_time": "3:11:05"}
{"current_steps": 2150, "total_steps": 3225, "eval_loss": 0.043197184801101685, "epoch": 3.3333333333333335, "percentage": 66.67, "elapsed_time": "6:24:43", "remaining_time": "3:12:21"}
{"current_steps": 2160, "total_steps": 3225, "loss": 0.0107, "lr": 2.970933010038599e-05, "epoch": 3.3488372093023258, "percentage": 66.98, "elapsed_time": "6:26:04", "remaining_time": "3:10:21"}
{"current_steps": 2170, "total_steps": 3225, "loss": 0.0085, "lr": 2.921582208737681e-05, "epoch": 3.3643410852713176, "percentage": 67.29, "elapsed_time": "6:27:20", "remaining_time": "3:08:18"}
{"current_steps": 2180, "total_steps": 3225, "loss": 0.0169, "lr": 2.872474982901081e-05, "epoch": 3.37984496124031, "percentage": 67.6, "elapsed_time": "6:28:35", "remaining_time": "3:06:16"}
{"current_steps": 2190, "total_steps": 3225, "loss": 0.007, "lr": 2.8236170875388744e-05, "epoch": 3.395348837209302, "percentage": 67.91, "elapsed_time": "6:29:51", "remaining_time": "3:04:14"}
{"current_steps": 2200, "total_steps": 3225, "loss": 0.0074, "lr": 2.775014248441422e-05, "epoch": 3.4108527131782944, "percentage": 68.22, "elapsed_time": "6:31:07", "remaining_time": "3:02:13"}
{"current_steps": 2200, "total_steps": 3225, "eval_loss": 0.045535024255514145, "epoch": 3.4108527131782944, "percentage": 68.22, "elapsed_time": "6:33:39", "remaining_time": "3:03:24"}
{"current_steps": 2210, "total_steps": 3225, "loss": 0.0106, "lr": 2.726672161508341e-05, "epoch": 3.4263565891472867, "percentage": 68.53, "elapsed_time": "6:34:59", "remaining_time": "3:01:24"}
{"current_steps": 2220, "total_steps": 3225, "loss": 0.0056, "lr": 2.678596492080984e-05, "epoch": 3.441860465116279, "percentage": 68.84, "elapsed_time": "6:36:15", "remaining_time": "2:59:23"}
{"current_steps": 2230, "total_steps": 3225, "loss": 0.0072, "lr": 2.630792874278516e-05, "epoch": 3.4573643410852712, "percentage": 69.15, "elapsed_time": "6:37:31", "remaining_time": "2:57:22"}
{"current_steps": 2240, "total_steps": 3225, "loss": 0.0084, "lr": 2.583266910337624e-05, "epoch": 3.4728682170542635, "percentage": 69.46, "elapsed_time": "6:38:46", "remaining_time": "2:55:21"}
{"current_steps": 2250, "total_steps": 3225, "loss": 0.0105, "lr": 2.5360241699559816e-05, "epoch": 3.488372093023256, "percentage": 69.77, "elapsed_time": "6:40:02", "remaining_time": "2:53:20"}
{"current_steps": 2250, "total_steps": 3225, "eval_loss": 0.04931580275297165, "epoch": 3.488372093023256, "percentage": 69.77, "elapsed_time": "6:42:34", "remaining_time": "2:54:27"}
{"current_steps": 2260, "total_steps": 3225, "loss": 0.0066, "lr": 2.4890701896395146e-05, "epoch": 3.503875968992248, "percentage": 70.08, "elapsed_time": "6:43:54", "remaining_time": "2:52:28"}
{"current_steps": 2270, "total_steps": 3225, "loss": 0.0108, "lr": 2.4424104720535735e-05, "epoch": 3.5193798449612403, "percentage": 70.39, "elapsed_time": "6:45:10", "remaining_time": "2:50:27"}
{"current_steps": 2280, "total_steps": 3225, "loss": 0.0095, "lr": 2.3960504853780462e-05, "epoch": 3.5348837209302326, "percentage": 70.7, "elapsed_time": "6:46:26", "remaining_time": "2:48:27"}
{"current_steps": 2290, "total_steps": 3225, "loss": 0.0071, "lr": 2.349995662666547e-05, "epoch": 3.550387596899225, "percentage": 71.01, "elapsed_time": "6:47:41", "remaining_time": "2:46:27"}
{"current_steps": 2300, "total_steps": 3225, "loss": 0.0116, "lr": 2.3042514012096843e-05, "epoch": 3.565891472868217, "percentage": 71.32, "elapsed_time": "6:48:57", "remaining_time": "2:44:28"}
{"current_steps": 2300, "total_steps": 3225, "eval_loss": 0.05101173743605614, "epoch": 3.565891472868217, "percentage": 71.32, "elapsed_time": "6:51:29", "remaining_time": "2:45:29"}
{"current_steps": 2310, "total_steps": 3225, "loss": 0.0164, "lr": 2.2588230619025407e-05, "epoch": 3.5813953488372094, "percentage": 71.63, "elapsed_time": "6:52:49", "remaining_time": "2:43:31"}
{"current_steps": 2320, "total_steps": 3225, "loss": 0.0095, "lr": 2.213715968616425e-05, "epoch": 3.5968992248062017, "percentage": 71.94, "elapsed_time": "6:54:05", "remaining_time": "2:41:31"}
{"current_steps": 2330, "total_steps": 3225, "loss": 0.0076, "lr": 2.16893540757494e-05, "epoch": 3.612403100775194, "percentage": 72.25, "elapsed_time": "6:55:21", "remaining_time": "2:39:32"}
{"current_steps": 2340, "total_steps": 3225, "loss": 0.0081, "lr": 2.1244866267344866e-05, "epoch": 3.6279069767441863, "percentage": 72.56, "elapsed_time": "6:56:37", "remaining_time": "2:37:34"}
{"current_steps": 2350, "total_steps": 3225, "loss": 0.01, "lr": 2.080374835169235e-05, "epoch": 3.6434108527131785, "percentage": 72.87, "elapsed_time": "6:57:52", "remaining_time": "2:35:35"}
{"current_steps": 2350, "total_steps": 3225, "eval_loss": 0.04809526726603508, "epoch": 3.6434108527131785, "percentage": 72.87, "elapsed_time": "7:00:25", "remaining_time": "2:36:32"}
{"current_steps": 2360, "total_steps": 3225, "loss": 0.009, "lr": 2.0366052024606612e-05, "epoch": 3.6589147286821704, "percentage": 73.18, "elapsed_time": "7:01:45", "remaining_time": "2:34:34"}
{"current_steps": 2370, "total_steps": 3225, "loss": 0.0096, "lr": 1.9931828580917107e-05, "epoch": 3.6744186046511627, "percentage": 73.49, "elapsed_time": "7:03:00", "remaining_time": "2:32:36"}
{"current_steps": 2380, "total_steps": 3225, "loss": 0.0103, "lr": 1.9501128908456523e-05, "epoch": 3.689922480620155, "percentage": 73.8, "elapsed_time": "7:04:16", "remaining_time": "2:30:38"}
{"current_steps": 2390, "total_steps": 3225, "loss": 0.0068, "lr": 1.9074003482097258e-05, "epoch": 3.705426356589147, "percentage": 74.11, "elapsed_time": "7:05:32", "remaining_time": "2:28:40"}
{"current_steps": 2400, "total_steps": 3225, "loss": 0.0126, "lr": 1.8650502357835925e-05, "epoch": 3.7209302325581395, "percentage": 74.42, "elapsed_time": "7:06:48", "remaining_time": "2:26:42"}
{"current_steps": 2400, "total_steps": 3225, "eval_loss": 0.047415368258953094, "epoch": 3.7209302325581395, "percentage": 74.42, "elapsed_time": "7:09:20", "remaining_time": "2:27:35"}
{"current_steps": 2410, "total_steps": 3225, "loss": 0.0077, "lr": 1.82306751669274e-05, "epoch": 3.7364341085271318, "percentage": 74.73, "elapsed_time": "7:10:40", "remaining_time": "2:25:38"}
{"current_steps": 2420, "total_steps": 3225, "loss": 0.0143, "lr": 1.7814571110068135e-05, "epoch": 3.751937984496124, "percentage": 75.04, "elapsed_time": "7:11:56", "remaining_time": "2:23:41"}
{"current_steps": 2430, "total_steps": 3225, "loss": 0.009, "lr": 1.740223895163039e-05, "epoch": 3.7674418604651163, "percentage": 75.35, "elapsed_time": "7:13:12", "remaining_time": "2:21:43"}
{"current_steps": 2440, "total_steps": 3225, "loss": 0.0106, "lr": 1.6993727013947336e-05, "epoch": 3.7829457364341086, "percentage": 75.66, "elapsed_time": "7:14:27", "remaining_time": "2:19:46"}
{"current_steps": 2450, "total_steps": 3225, "loss": 0.0061, "lr": 1.6589083171649977e-05, "epoch": 3.798449612403101, "percentage": 75.97, "elapsed_time": "7:15:43", "remaining_time": "2:17:49"}
{"current_steps": 2450, "total_steps": 3225, "eval_loss": 0.047738756984472275, "epoch": 3.798449612403101, "percentage": 75.97, "elapsed_time": "7:18:15", "remaining_time": "2:18:38"}
{"current_steps": 2460, "total_steps": 3225, "loss": 0.0071, "lr": 1.6188354846056698e-05, "epoch": 3.813953488372093, "percentage": 76.28, "elapsed_time": "7:19:35", "remaining_time": "2:16:42"}
{"current_steps": 2470, "total_steps": 3225, "loss": 0.007, "lr": 1.579158899961575e-05, "epoch": 3.8294573643410854, "percentage": 76.59, "elapsed_time": "7:20:52", "remaining_time": "2:14:45"}
{"current_steps": 2480, "total_steps": 3225, "loss": 0.0082, "lr": 1.5398832130401637e-05, "epoch": 3.8449612403100772, "percentage": 76.9, "elapsed_time": "7:22:07", "remaining_time": "2:12:49"}
{"current_steps": 2490, "total_steps": 3225, "loss": 0.0126, "lr": 1.5010130266665807e-05, "epoch": 3.8604651162790695, "percentage": 77.21, "elapsed_time": "7:23:23", "remaining_time": "2:10:52"}
{"current_steps": 2500, "total_steps": 3225, "loss": 0.0088, "lr": 1.4625528961442591e-05, "epoch": 3.875968992248062, "percentage": 77.52, "elapsed_time": "7:24:39", "remaining_time": "2:08:57"}
{"current_steps": 2500, "total_steps": 3225, "eval_loss": 0.04871485382318497, "epoch": 3.875968992248062, "percentage": 77.52, "elapsed_time": "7:27:11", "remaining_time": "2:09:41"}
{"current_steps": 2510, "total_steps": 3225, "loss": 0.0097, "lr": 1.4245073287210608e-05, "epoch": 3.891472868217054, "percentage": 77.83, "elapsed_time": "7:28:32", "remaining_time": "2:07:46"}
{"current_steps": 2520, "total_steps": 3225, "loss": 0.0147, "lr": 1.3868807830610653e-05, "epoch": 3.9069767441860463, "percentage": 78.14, "elapsed_time": "7:29:47", "remaining_time": "2:05:50"}
{"current_steps": 2530, "total_steps": 3225, "loss": 0.009, "lr": 1.3496776687220514e-05, "epoch": 3.9224806201550386, "percentage": 78.45, "elapsed_time": "7:31:03", "remaining_time": "2:03:54"}
{"current_steps": 2540, "total_steps": 3225, "loss": 0.0039, "lr": 1.3129023456387151e-05, "epoch": 3.937984496124031, "percentage": 78.76, "elapsed_time": "7:32:19", "remaining_time": "2:01:59"}
{"current_steps": 2550, "total_steps": 3225, "loss": 0.0074, "lr": 1.2765591236117324e-05, "epoch": 3.953488372093023, "percentage": 79.07, "elapsed_time": "7:33:34", "remaining_time": "2:00:03"}
{"current_steps": 2550, "total_steps": 3225, "eval_loss": 0.04882879555225372, "epoch": 3.953488372093023, "percentage": 79.07, "elapsed_time": "7:36:07", "remaining_time": "2:00:44"}
{"current_steps": 2560, "total_steps": 3225, "loss": 0.0077, "lr": 1.240652261802669e-05, "epoch": 3.9689922480620154, "percentage": 79.38, "elapsed_time": "7:37:27", "remaining_time": "1:58:49"}
{"current_steps": 2570, "total_steps": 3225, "loss": 0.0122, "lr": 1.205185968234847e-05, "epoch": 3.9844961240310077, "percentage": 79.69, "elapsed_time": "7:38:42", "remaining_time": "1:56:54"}
{"current_steps": 2580, "total_steps": 3225, "loss": 0.009, "lr": 1.1701643993001916e-05, "epoch": 4.0, "percentage": 80.0, "elapsed_time": "7:39:58", "remaining_time": "1:54:59"}
{"current_steps": 2590, "total_steps": 3225, "loss": 0.0035, "lr": 1.1355916592721316e-05, "epoch": 4.015503875968992, "percentage": 80.31, "elapsed_time": "7:41:14", "remaining_time": "1:53:05"}
{"current_steps": 2600, "total_steps": 3225, "loss": 0.0076, "lr": 1.1014717998246099e-05, "epoch": 4.0310077519379846, "percentage": 80.62, "elapsed_time": "7:42:30", "remaining_time": "1:51:10"}
{"current_steps": 2600, "total_steps": 3225, "eval_loss": 0.04985535144805908, "epoch": 4.0310077519379846, "percentage": 80.62, "elapsed_time": "7:45:02", "remaining_time": "1:51:47"}
{"current_steps": 2610, "total_steps": 3225, "loss": 0.0033, "lr": 1.0678088195572517e-05, "epoch": 4.046511627906977, "percentage": 80.93, "elapsed_time": "7:46:23", "remaining_time": "1:49:53"}
{"current_steps": 2620, "total_steps": 3225, "loss": 0.0025, "lr": 1.0346066635267676e-05, "epoch": 4.062015503875969, "percentage": 81.24, "elapsed_time": "7:47:39", "remaining_time": "1:47:59"}
{"current_steps": 2630, "total_steps": 3225, "loss": 0.0041, "lr": 1.001869222784611e-05, "epoch": 4.077519379844961, "percentage": 81.55, "elapsed_time": "7:48:55", "remaining_time": "1:46:05"}
{"current_steps": 2640, "total_steps": 3225, "loss": 0.0059, "lr": 9.696003339209819e-06, "epoch": 4.093023255813954, "percentage": 81.86, "elapsed_time": "7:50:10", "remaining_time": "1:44:11"}
{"current_steps": 2650, "total_steps": 3225, "loss": 0.0051, "lr": 9.378037786152055e-06, "epoch": 4.108527131782946, "percentage": 82.17, "elapsed_time": "7:51:26", "remaining_time": "1:42:17"}
{"current_steps": 2650, "total_steps": 3225, "eval_loss": 0.05235826596617699, "epoch": 4.108527131782946, "percentage": 82.17, "elapsed_time": "7:53:58", "remaining_time": "1:42:50"}
{"current_steps": 2660, "total_steps": 3225, "loss": 0.0037, "lr": 9.064832831925441e-06, "epoch": 4.124031007751938, "percentage": 82.48, "elapsed_time": "7:55:19", "remaining_time": "1:40:57"}
{"current_steps": 2670, "total_steps": 3225, "loss": 0.002, "lr": 8.756425181875028e-06, "epoch": 4.1395348837209305, "percentage": 82.79, "elapsed_time": "7:56:35", "remaining_time": "1:39:03"}
{"current_steps": 2680, "total_steps": 3225, "loss": 0.0024, "lr": 8.452850979136617e-06, "epoch": 4.155038759689923, "percentage": 83.1, "elapsed_time": "7:57:50", "remaining_time": "1:37:10"}
{"current_steps": 2690, "total_steps": 3225, "loss": 0.0017, "lr": 8.15414580040117e-06, "epoch": 4.170542635658915, "percentage": 83.41, "elapsed_time": "7:59:06", "remaining_time": "1:35:17"}
{"current_steps": 2700, "total_steps": 3225, "loss": 0.0038, "lr": 7.860344651745366e-06, "epoch": 4.186046511627907, "percentage": 83.72, "elapsed_time": "8:00:22", "remaining_time": "1:33:24"}
{"current_steps": 2700, "total_steps": 3225, "eval_loss": 0.05555489659309387, "epoch": 4.186046511627907, "percentage": 83.72, "elapsed_time": "8:02:54", "remaining_time": "1:33:54"}
{"current_steps": 2710, "total_steps": 3225, "loss": 0.003, "lr": 7.571481964529226e-06, "epoch": 4.2015503875969, "percentage": 84.03, "elapsed_time": "8:04:14", "remaining_time": "1:32:01"}
{"current_steps": 2720, "total_steps": 3225, "loss": 0.0014, "lr": 7.287591591360926e-06, "epoch": 4.217054263565892, "percentage": 84.34, "elapsed_time": "8:05:30", "remaining_time": "1:30:08"}
{"current_steps": 2730, "total_steps": 3225, "loss": 0.0055, "lr": 7.008706802129606e-06, "epoch": 4.232558139534884, "percentage": 84.65, "elapsed_time": "8:06:46", "remaining_time": "1:28:15"}
{"current_steps": 2740, "total_steps": 3225, "loss": 0.0019, "lr": 6.734860280106292e-06, "epoch": 4.248062015503876, "percentage": 84.96, "elapsed_time": "8:08:02", "remaining_time": "1:26:23"}
{"current_steps": 2750, "total_steps": 3225, "loss": 0.0031, "lr": 6.466084118113735e-06, "epoch": 4.263565891472869, "percentage": 85.27, "elapsed_time": "8:09:17", "remaining_time": "1:24:30"}
{"current_steps": 2750, "total_steps": 3225, "eval_loss": 0.05840018764138222, "epoch": 4.263565891472869, "percentage": 85.27, "elapsed_time": "8:11:50", "remaining_time": "1:24:57"}
{"current_steps": 2760, "total_steps": 3225, "loss": 0.0051, "lr": 6.202409814765325e-06, "epoch": 4.27906976744186, "percentage": 85.58, "elapsed_time": "8:13:11", "remaining_time": "1:23:05"}
{"current_steps": 2770, "total_steps": 3225, "loss": 0.0033, "lr": 5.9438682707736725e-06, "epoch": 4.294573643410852, "percentage": 85.89, "elapsed_time": "8:14:27", "remaining_time": "1:21:13"}
{"current_steps": 2780, "total_steps": 3225, "loss": 0.0031, "lr": 5.690489785329301e-06, "epoch": 4.310077519379845, "percentage": 86.2, "elapsed_time": "8:15:43", "remaining_time": "1:19:21"}
{"current_steps": 2790, "total_steps": 3225, "loss": 0.0029, "lr": 5.442304052549752e-06, "epoch": 4.325581395348837, "percentage": 86.51, "elapsed_time": "8:16:59", "remaining_time": "1:17:29"}
{"current_steps": 2800, "total_steps": 3225, "loss": 0.0028, "lr": 5.199340157999733e-06, "epoch": 4.341085271317829, "percentage": 86.82, "elapsed_time": "8:18:14", "remaining_time": "1:15:37"}
{"current_steps": 2800, "total_steps": 3225, "eval_loss": 0.06022586300969124, "epoch": 4.341085271317829, "percentage": 86.82, "elapsed_time": "8:20:47", "remaining_time": "1:16:00"}
{"current_steps": 2810, "total_steps": 3225, "loss": 0.0019, "lr": 4.961626575282396e-06, "epoch": 4.3565891472868215, "percentage": 87.13, "elapsed_time": "8:22:07", "remaining_time": "1:14:09"}
{"current_steps": 2820, "total_steps": 3225, "loss": 0.002, "lr": 4.7291911627025235e-06, "epoch": 4.372093023255814, "percentage": 87.44, "elapsed_time": "8:23:23", "remaining_time": "1:12:17"}
{"current_steps": 2830, "total_steps": 3225, "loss": 0.0021, "lr": 4.502061160001725e-06, "epoch": 4.387596899224806, "percentage": 87.75, "elapsed_time": "8:24:38", "remaining_time": "1:10:26"}
{"current_steps": 2840, "total_steps": 3225, "loss": 0.0037, "lr": 4.280263185166078e-06, "epoch": 4.403100775193798, "percentage": 88.06, "elapsed_time": "8:25:53", "remaining_time": "1:08:34"}
{"current_steps": 2850, "total_steps": 3225, "loss": 0.0037, "lr": 4.063823231306757e-06, "epoch": 4.4186046511627906, "percentage": 88.37, "elapsed_time": "8:27:09", "remaining_time": "1:06:43"}
{"current_steps": 2850, "total_steps": 3225, "eval_loss": 0.061209216713905334, "epoch": 4.4186046511627906, "percentage": 88.37, "elapsed_time": "8:29:42", "remaining_time": "1:07:03"}
{"current_steps": 2860, "total_steps": 3225, "loss": 0.0018, "lr": 3.8527666636137885e-06, "epoch": 4.434108527131783, "percentage": 88.68, "elapsed_time": "8:31:02", "remaining_time": "1:05:13"}
{"current_steps": 2870, "total_steps": 3225, "loss": 0.0032, "lr": 3.6471182163834825e-06, "epoch": 4.449612403100775, "percentage": 88.99, "elapsed_time": "8:32:18", "remaining_time": "1:03:22"}
{"current_steps": 2880, "total_steps": 3225, "loss": 0.0018, "lr": 3.4469019901197054e-06, "epoch": 4.465116279069767, "percentage": 89.3, "elapsed_time": "8:33:33", "remaining_time": "1:01:31"}
{"current_steps": 2890, "total_steps": 3225, "loss": 0.0032, "lr": 3.252141448709495e-06, "epoch": 4.48062015503876, "percentage": 89.61, "elapsed_time": "8:34:49", "remaining_time": "0:59:40"}
{"current_steps": 2900, "total_steps": 3225, "loss": 0.0037, "lr": 3.0628594166732693e-06, "epoch": 4.496124031007752, "percentage": 89.92, "elapsed_time": "8:36:05", "remaining_time": "0:57:50"}
{"current_steps": 2900, "total_steps": 3225, "eval_loss": 0.06198548525571823, "epoch": 4.496124031007752, "percentage": 89.92, "elapsed_time": "8:38:37", "remaining_time": "0:58:07"}
{"current_steps": 2910, "total_steps": 3225, "loss": 0.0029, "lr": 2.8790780764899384e-06, "epoch": 4.511627906976744, "percentage": 90.23, "elapsed_time": "8:39:57", "remaining_time": "0:56:17"}
{"current_steps": 2920, "total_steps": 3225, "loss": 0.004, "lr": 2.700818965997315e-06, "epoch": 4.5271317829457365, "percentage": 90.54, "elapsed_time": "8:41:12", "remaining_time": "0:54:26"}
{"current_steps": 2930, "total_steps": 3225, "loss": 0.0007, "lr": 2.528102975867991e-06, "epoch": 4.542635658914729, "percentage": 90.85, "elapsed_time": "8:42:28", "remaining_time": "0:52:36"}
{"current_steps": 2940, "total_steps": 3225, "loss": 0.0014, "lr": 2.3609503471611284e-06, "epoch": 4.558139534883721, "percentage": 91.16, "elapsed_time": "8:43:44", "remaining_time": "0:50:46"}
{"current_steps": 2950, "total_steps": 3225, "loss": 0.0013, "lr": 2.1993806689503738e-06, "epoch": 4.573643410852713, "percentage": 91.47, "elapsed_time": "8:44:59", "remaining_time": "0:48:56"}
{"current_steps": 2950, "total_steps": 3225, "eval_loss": 0.06263046711683273, "epoch": 4.573643410852713, "percentage": 91.47, "elapsed_time": "8:47:32", "remaining_time": "0:49:10"}
{"current_steps": 2960, "total_steps": 3225, "loss": 0.001, "lr": 2.0434128760281167e-06, "epoch": 4.589147286821706, "percentage": 91.78, "elapsed_time": "8:48:53", "remaining_time": "0:47:20"}
{"current_steps": 2970, "total_steps": 3225, "loss": 0.0005, "lr": 1.893065246686504e-06, "epoch": 4.604651162790698, "percentage": 92.09, "elapsed_time": "8:50:08", "remaining_time": "0:45:31"}
{"current_steps": 2980, "total_steps": 3225, "loss": 0.004, "lr": 1.74835540057533e-06, "epoch": 4.62015503875969, "percentage": 92.4, "elapsed_time": "8:51:24", "remaining_time": "0:43:41"}
{"current_steps": 2990, "total_steps": 3225, "loss": 0.0013, "lr": 1.6093002966371617e-06, "epoch": 4.635658914728682, "percentage": 92.71, "elapsed_time": "8:52:40", "remaining_time": "0:41:51"}
{"current_steps": 3000, "total_steps": 3225, "loss": 0.0013, "lr": 1.4759162311198783e-06, "epoch": 4.651162790697675, "percentage": 93.02, "elapsed_time": "8:53:56", "remaining_time": "0:40:02"}
{"current_steps": 3000, "total_steps": 3225, "eval_loss": 0.06311814486980438, "epoch": 4.651162790697675, "percentage": 93.02, "elapsed_time": "8:56:28", "remaining_time": "0:40:14"}
{"current_steps": 3010, "total_steps": 3225, "loss": 0.003, "lr": 1.348218835666859e-06, "epoch": 4.666666666666667, "percentage": 93.33, "elapsed_time": "8:57:48", "remaining_time": "0:38:24"}
{"current_steps": 3020, "total_steps": 3225, "loss": 0.0029, "lr": 1.2262230754850445e-06, "epoch": 4.682170542635659, "percentage": 93.64, "elapsed_time": "8:59:04", "remaining_time": "0:36:35"}
{"current_steps": 3030, "total_steps": 3225, "loss": 0.0037, "lr": 1.109943247591172e-06, "epoch": 4.6976744186046515, "percentage": 93.95, "elapsed_time": "9:00:20", "remaining_time": "0:34:46"}
{"current_steps": 3040, "total_steps": 3225, "loss": 0.003, "lr": 9.993929791362323e-07, "epoch": 4.713178294573644, "percentage": 94.26, "elapsed_time": "9:01:36", "remaining_time": "0:32:57"}
{"current_steps": 3050, "total_steps": 3225, "loss": 0.0023, "lr": 8.945852258084863e-07, "epoch": 4.728682170542635, "percentage": 94.57, "elapsed_time": "9:02:51", "remaining_time": "0:31:08"}
{"current_steps": 3050, "total_steps": 3225, "eval_loss": 0.06340909004211426, "epoch": 4.728682170542635, "percentage": 94.57, "elapsed_time": "9:05:24", "remaining_time": "0:31:17"}
{"current_steps": 3060, "total_steps": 3225, "loss": 0.0051, "lr": 7.955322703151358e-07, "epoch": 4.7441860465116275, "percentage": 94.88, "elapsed_time": "9:06:44", "remaining_time": "0:29:28"}
{"current_steps": 3070, "total_steps": 3225, "loss": 0.0033, "lr": 7.022457209428901e-07, "epoch": 4.75968992248062, "percentage": 95.19, "elapsed_time": "9:07:59", "remaining_time": "0:27:40"}
{"current_steps": 3080, "total_steps": 3225, "loss": 0.0029, "lr": 6.147365101975666e-07, "epoch": 4.775193798449612, "percentage": 95.5, "elapsed_time": "9:09:15", "remaining_time": "0:25:51"}
{"current_steps": 3090, "total_steps": 3225, "loss": 0.003, "lr": 5.33014893522854e-07, "epoch": 4.790697674418604, "percentage": 95.81, "elapsed_time": "9:10:31", "remaining_time": "0:24:03"}
{"current_steps": 3100, "total_steps": 3225, "loss": 0.0042, "lr": 4.57090448098485e-07, "epoch": 4.8062015503875966, "percentage": 96.12, "elapsed_time": "9:11:46", "remaining_time": "0:22:14"}
{"current_steps": 3100, "total_steps": 3225, "eval_loss": 0.0634797140955925, "epoch": 4.8062015503875966, "percentage": 96.12, "elapsed_time": "9:14:19", "remaining_time": "0:22:21"}
{"current_steps": 3110, "total_steps": 3225, "loss": 0.0043, "lr": 3.8697207171781714e-07, "epoch": 4.821705426356589, "percentage": 96.43, "elapsed_time": "9:15:39", "remaining_time": "0:20:32"}
{"current_steps": 3120, "total_steps": 3225, "loss": 0.0043, "lr": 3.2266798174512837e-07, "epoch": 4.837209302325581, "percentage": 96.74, "elapsed_time": "9:16:54", "remaining_time": "0:18:44"}
{"current_steps": 3130, "total_steps": 3225, "loss": 0.0069, "lr": 2.6418571415255387e-07, "epoch": 4.852713178294573, "percentage": 97.05, "elapsed_time": "9:18:10", "remaining_time": "0:16:56"}
{"current_steps": 3140, "total_steps": 3225, "loss": 0.0026, "lr": 2.1153212263695378e-07, "epoch": 4.868217054263566, "percentage": 97.36, "elapsed_time": "9:19:25", "remaining_time": "0:15:08"}
{"current_steps": 3150, "total_steps": 3225, "loss": 0.0053, "lr": 1.6471337781669982e-07, "epoch": 4.883720930232558, "percentage": 97.67, "elapsed_time": "9:20:41", "remaining_time": "0:13:20"}
{"current_steps": 3150, "total_steps": 3225, "eval_loss": 0.06353544443845749, "epoch": 4.883720930232558, "percentage": 97.67, "elapsed_time": "9:23:13", "remaining_time": "0:13:24"}
{"current_steps": 3160, "total_steps": 3225, "loss": 0.0026, "lr": 1.237349665085097e-07, "epoch": 4.89922480620155, "percentage": 97.98, "elapsed_time": "9:24:34", "remaining_time": "0:11:36"}
{"current_steps": 3170, "total_steps": 3225, "loss": 0.0027, "lr": 8.86016910844667e-08, "epoch": 4.9147286821705425, "percentage": 98.29, "elapsed_time": "9:25:49", "remaining_time": "0:09:49"}
{"current_steps": 3180, "total_steps": 3225, "loss": 0.004, "lr": 5.9317668909192323e-08, "epoch": 4.930232558139535, "percentage": 98.6, "elapsed_time": "9:27:05", "remaining_time": "0:08:01"}
{"current_steps": 3190, "total_steps": 3225, "loss": 0.005, "lr": 3.588633185730994e-08, "epoch": 4.945736434108527, "percentage": 98.91, "elapsed_time": "9:28:21", "remaining_time": "0:06:14"}
{"current_steps": 3200, "total_steps": 3225, "loss": 0.0041, "lr": 1.8310425911294283e-08, "epoch": 4.961240310077519, "percentage": 99.22, "elapsed_time": "9:29:36", "remaining_time": "0:04:27"}
{"current_steps": 3200, "total_steps": 3225, "eval_loss": 0.06344824284315109, "epoch": 4.961240310077519, "percentage": 99.22, "elapsed_time": "9:32:09", "remaining_time": "0:04:28"}
{"current_steps": 3210, "total_steps": 3225, "loss": 0.0043, "lr": 6.592010839612251e-09, "epoch": 4.976744186046512, "percentage": 99.53, "elapsed_time": "9:33:29", "remaining_time": "0:02:40"}
{"current_steps": 3220, "total_steps": 3225, "loss": 0.0015, "lr": 7.324599553770739e-10, "epoch": 4.992248062015504, "percentage": 99.84, "elapsed_time": "9:34:45", "remaining_time": "0:00:53"}
{"current_steps": 3225, "total_steps": 3225, "epoch": 5.0, "percentage": 100.0, "elapsed_time": "9:35:28", "remaining_time": "0:00:00"}