chchen's picture
Training in progress, step 125
e46faac verified
{"current_steps": 1, "total_steps": 125, "loss": 0.9492, "lr": 7.692307692307694e-06, "epoch": 0.03950617283950617, "percentage": 0.8, "elapsed_time": "0:00:38", "remaining_time": "1:19:49"}
{"current_steps": 2, "total_steps": 125, "loss": 1.002, "lr": 1.5384615384615387e-05, "epoch": 0.07901234567901234, "percentage": 1.6, "elapsed_time": "0:01:07", "remaining_time": "1:08:57"}
{"current_steps": 3, "total_steps": 125, "loss": 1.071, "lr": 2.307692307692308e-05, "epoch": 0.11851851851851852, "percentage": 2.4, "elapsed_time": "0:01:34", "remaining_time": "1:04:10"}
{"current_steps": 4, "total_steps": 125, "loss": 1.1055, "lr": 3.0769230769230774e-05, "epoch": 0.1580246913580247, "percentage": 3.2, "elapsed_time": "0:02:06", "remaining_time": "1:03:36"}
{"current_steps": 5, "total_steps": 125, "loss": 1.1171, "lr": 3.846153846153846e-05, "epoch": 0.19753086419753085, "percentage": 4.0, "elapsed_time": "0:02:36", "remaining_time": "1:02:32"}
{"current_steps": 6, "total_steps": 125, "loss": 1.0424, "lr": 4.615384615384616e-05, "epoch": 0.23703703703703705, "percentage": 4.8, "elapsed_time": "0:03:06", "remaining_time": "1:01:43"}
{"current_steps": 7, "total_steps": 125, "loss": 0.9371, "lr": 5.384615384615385e-05, "epoch": 0.2765432098765432, "percentage": 5.6, "elapsed_time": "0:03:39", "remaining_time": "1:01:37"}
{"current_steps": 8, "total_steps": 125, "loss": 0.9213, "lr": 6.153846153846155e-05, "epoch": 0.3160493827160494, "percentage": 6.4, "elapsed_time": "0:04:09", "remaining_time": "1:00:43"}
{"current_steps": 9, "total_steps": 125, "loss": 0.9886, "lr": 6.923076923076924e-05, "epoch": 0.35555555555555557, "percentage": 7.2, "elapsed_time": "0:04:36", "remaining_time": "0:59:29"}
{"current_steps": 10, "total_steps": 125, "loss": 0.7837, "lr": 7.692307692307693e-05, "epoch": 0.3950617283950617, "percentage": 8.0, "elapsed_time": "0:05:05", "remaining_time": "0:58:30"}
{"current_steps": 10, "total_steps": 125, "eval_loss": 0.7542588114738464, "epoch": 0.3950617283950617, "percentage": 8.0, "elapsed_time": "0:05:27", "remaining_time": "1:02:49"}
{"current_steps": 11, "total_steps": 125, "loss": 0.6732, "lr": 8.461538461538461e-05, "epoch": 0.4345679012345679, "percentage": 8.8, "elapsed_time": "0:05:57", "remaining_time": "1:01:42"}
{"current_steps": 12, "total_steps": 125, "loss": 0.6115, "lr": 9.230769230769232e-05, "epoch": 0.4740740740740741, "percentage": 9.6, "elapsed_time": "0:06:32", "remaining_time": "1:01:35"}
{"current_steps": 13, "total_steps": 125, "loss": 0.5413, "lr": 0.0001, "epoch": 0.5135802469135803, "percentage": 10.4, "elapsed_time": "0:07:00", "remaining_time": "1:00:18"}
{"current_steps": 14, "total_steps": 125, "loss": 0.4075, "lr": 9.998033131915266e-05, "epoch": 0.5530864197530864, "percentage": 11.2, "elapsed_time": "0:07:26", "remaining_time": "0:59:03"}
{"current_steps": 15, "total_steps": 125, "loss": 0.347, "lr": 9.992134075089084e-05, "epoch": 0.5925925925925926, "percentage": 12.0, "elapsed_time": "0:07:55", "remaining_time": "0:58:06"}
{"current_steps": 16, "total_steps": 125, "loss": 0.4062, "lr": 9.982307470588098e-05, "epoch": 0.6320987654320988, "percentage": 12.8, "elapsed_time": "0:08:27", "remaining_time": "0:57:40"}
{"current_steps": 17, "total_steps": 125, "loss": 0.2423, "lr": 9.968561049466214e-05, "epoch": 0.671604938271605, "percentage": 13.6, "elapsed_time": "0:08:59", "remaining_time": "0:57:04"}
{"current_steps": 18, "total_steps": 125, "loss": 0.2523, "lr": 9.950905626682228e-05, "epoch": 0.7111111111111111, "percentage": 14.4, "elapsed_time": "0:09:24", "remaining_time": "0:55:55"}
{"current_steps": 19, "total_steps": 125, "loss": 0.3278, "lr": 9.92935509259118e-05, "epoch": 0.7506172839506173, "percentage": 15.2, "elapsed_time": "0:09:54", "remaining_time": "0:55:19"}
{"current_steps": 20, "total_steps": 125, "loss": 0.2931, "lr": 9.903926402016153e-05, "epoch": 0.7901234567901234, "percentage": 16.0, "elapsed_time": "0:10:20", "remaining_time": "0:54:17"}
{"current_steps": 20, "total_steps": 125, "eval_loss": 0.2810686230659485, "epoch": 0.7901234567901234, "percentage": 16.0, "elapsed_time": "0:10:42", "remaining_time": "0:56:15"}
{"current_steps": 21, "total_steps": 125, "loss": 0.2341, "lr": 9.874639560909117e-05, "epoch": 0.8296296296296296, "percentage": 16.8, "elapsed_time": "0:11:15", "remaining_time": "0:55:45"}
{"current_steps": 22, "total_steps": 125, "loss": 0.253, "lr": 9.841517610611309e-05, "epoch": 0.8691358024691358, "percentage": 17.6, "elapsed_time": "0:11:43", "remaining_time": "0:54:52"}
{"current_steps": 23, "total_steps": 125, "loss": 0.1967, "lr": 9.804586609725499e-05, "epoch": 0.908641975308642, "percentage": 18.4, "elapsed_time": "0:12:12", "remaining_time": "0:54:07"}
{"current_steps": 24, "total_steps": 125, "loss": 0.2832, "lr": 9.763875613614482e-05, "epoch": 0.9481481481481482, "percentage": 19.2, "elapsed_time": "0:12:41", "remaining_time": "0:53:26"}
{"current_steps": 25, "total_steps": 125, "loss": 0.2855, "lr": 9.719416651541839e-05, "epoch": 0.9876543209876543, "percentage": 20.0, "elapsed_time": "0:13:08", "remaining_time": "0:52:35"}
{"current_steps": 26, "total_steps": 125, "loss": 0.3211, "lr": 9.671244701472999e-05, "epoch": 1.0271604938271606, "percentage": 20.8, "elapsed_time": "0:13:36", "remaining_time": "0:51:48"}
{"current_steps": 27, "total_steps": 125, "loss": 0.2032, "lr": 9.619397662556435e-05, "epoch": 1.0666666666666667, "percentage": 21.6, "elapsed_time": "0:14:04", "remaining_time": "0:51:04"}
{"current_steps": 28, "total_steps": 125, "loss": 0.2985, "lr": 9.563916325306594e-05, "epoch": 1.106172839506173, "percentage": 22.4, "elapsed_time": "0:14:31", "remaining_time": "0:50:19"}
{"current_steps": 29, "total_steps": 125, "loss": 0.1901, "lr": 9.504844339512095e-05, "epoch": 1.145679012345679, "percentage": 23.2, "elapsed_time": "0:15:04", "remaining_time": "0:49:53"}
{"current_steps": 30, "total_steps": 125, "loss": 0.1974, "lr": 9.442228179894362e-05, "epoch": 1.1851851851851851, "percentage": 24.0, "elapsed_time": "0:15:30", "remaining_time": "0:49:07"}
{"current_steps": 30, "total_steps": 125, "eval_loss": 0.2092229425907135, "epoch": 1.1851851851851851, "percentage": 24.0, "elapsed_time": "0:15:53", "remaining_time": "0:50:19"}
{"current_steps": 31, "total_steps": 125, "loss": 0.1991, "lr": 9.376117109543769e-05, "epoch": 1.2246913580246914, "percentage": 24.8, "elapsed_time": "0:16:23", "remaining_time": "0:49:42"}
{"current_steps": 32, "total_steps": 125, "loss": 0.2061, "lr": 9.306563141162046e-05, "epoch": 1.2641975308641975, "percentage": 25.6, "elapsed_time": "0:16:55", "remaining_time": "0:49:10"}
{"current_steps": 33, "total_steps": 125, "loss": 0.203, "lr": 9.233620996141421e-05, "epoch": 1.3037037037037038, "percentage": 26.4, "elapsed_time": "0:17:25", "remaining_time": "0:48:35"}
{"current_steps": 34, "total_steps": 125, "loss": 0.1553, "lr": 9.157348061512727e-05, "epoch": 1.34320987654321, "percentage": 27.2, "elapsed_time": "0:17:49", "remaining_time": "0:47:43"}
{"current_steps": 35, "total_steps": 125, "loss": 0.2012, "lr": 9.077804344796302e-05, "epoch": 1.382716049382716, "percentage": 28.0, "elapsed_time": "0:18:18", "remaining_time": "0:47:03"}
{"current_steps": 36, "total_steps": 125, "loss": 0.1661, "lr": 8.995052426791247e-05, "epoch": 1.4222222222222223, "percentage": 28.8, "elapsed_time": "0:18:46", "remaining_time": "0:46:24"}
{"current_steps": 37, "total_steps": 125, "loss": 0.1218, "lr": 8.90915741234015e-05, "epoch": 1.4617283950617284, "percentage": 29.6, "elapsed_time": "0:19:12", "remaining_time": "0:45:42"}
{"current_steps": 38, "total_steps": 125, "loss": 0.1414, "lr": 8.820186879108038e-05, "epoch": 1.5012345679012347, "percentage": 30.4, "elapsed_time": "0:19:43", "remaining_time": "0:45:09"}
{"current_steps": 39, "total_steps": 125, "loss": 0.1725, "lr": 8.728210824415827e-05, "epoch": 1.5407407407407407, "percentage": 31.2, "elapsed_time": "0:20:12", "remaining_time": "0:44:32"}
{"current_steps": 40, "total_steps": 125, "loss": 0.144, "lr": 8.633301610170135e-05, "epoch": 1.5802469135802468, "percentage": 32.0, "elapsed_time": "0:20:41", "remaining_time": "0:43:57"}
{"current_steps": 40, "total_steps": 125, "eval_loss": 0.1874631941318512, "epoch": 1.5802469135802468, "percentage": 32.0, "elapsed_time": "0:21:03", "remaining_time": "0:44:45"}
{"current_steps": 41, "total_steps": 125, "loss": 0.1841, "lr": 8.535533905932738e-05, "epoch": 1.6197530864197531, "percentage": 32.8, "elapsed_time": "0:21:36", "remaining_time": "0:44:15"}
{"current_steps": 42, "total_steps": 125, "loss": 0.1598, "lr": 8.434984630174509e-05, "epoch": 1.6592592592592592, "percentage": 33.6, "elapsed_time": "0:22:08", "remaining_time": "0:43:45"}
{"current_steps": 43, "total_steps": 125, "loss": 0.1499, "lr": 8.33173288976002e-05, "epoch": 1.6987654320987655, "percentage": 34.4, "elapsed_time": "0:22:39", "remaining_time": "0:43:13"}
{"current_steps": 44, "total_steps": 125, "loss": 0.1356, "lr": 8.225859917710439e-05, "epoch": 1.7382716049382716, "percentage": 35.2, "elapsed_time": "0:23:12", "remaining_time": "0:42:43"}
{"current_steps": 45, "total_steps": 125, "loss": 0.2259, "lr": 8.117449009293668e-05, "epoch": 1.7777777777777777, "percentage": 36.0, "elapsed_time": "0:23:40", "remaining_time": "0:42:05"}
{"current_steps": 46, "total_steps": 125, "loss": 0.157, "lr": 8.006585456492029e-05, "epoch": 1.817283950617284, "percentage": 36.8, "elapsed_time": "0:24:08", "remaining_time": "0:41:28"}
{"current_steps": 47, "total_steps": 125, "loss": 0.1731, "lr": 7.89335648089903e-05, "epoch": 1.8567901234567903, "percentage": 37.6, "elapsed_time": "0:24:38", "remaining_time": "0:40:53"}
{"current_steps": 48, "total_steps": 125, "loss": 0.1294, "lr": 7.777851165098012e-05, "epoch": 1.8962962962962964, "percentage": 38.4, "elapsed_time": "0:25:06", "remaining_time": "0:40:16"}
{"current_steps": 49, "total_steps": 125, "loss": 0.2119, "lr": 7.660160382576683e-05, "epoch": 1.9358024691358025, "percentage": 39.2, "elapsed_time": "0:25:38", "remaining_time": "0:39:46"}
{"current_steps": 50, "total_steps": 125, "loss": 0.1558, "lr": 7.540376726232648e-05, "epoch": 1.9753086419753085, "percentage": 40.0, "elapsed_time": "0:26:09", "remaining_time": "0:39:14"}
{"current_steps": 50, "total_steps": 125, "eval_loss": 0.17156915366649628, "epoch": 1.9753086419753085, "percentage": 40.0, "elapsed_time": "0:26:32", "remaining_time": "0:39:48"}
{"current_steps": 51, "total_steps": 125, "loss": 0.2625, "lr": 7.4185944355262e-05, "epoch": 2.0148148148148146, "percentage": 40.8, "elapsed_time": "0:27:04", "remaining_time": "0:39:17"}
{"current_steps": 52, "total_steps": 125, "loss": 0.1106, "lr": 7.294909322337689e-05, "epoch": 2.054320987654321, "percentage": 41.6, "elapsed_time": "0:27:34", "remaining_time": "0:38:42"}
{"current_steps": 53, "total_steps": 125, "loss": 0.1197, "lr": 7.169418695587791e-05, "epoch": 2.093827160493827, "percentage": 42.4, "elapsed_time": "0:28:02", "remaining_time": "0:38:05"}
{"current_steps": 54, "total_steps": 125, "loss": 0.1626, "lr": 7.042221284679982e-05, "epoch": 2.1333333333333333, "percentage": 43.2, "elapsed_time": "0:28:31", "remaining_time": "0:37:30"}
{"current_steps": 55, "total_steps": 125, "loss": 0.1342, "lr": 6.91341716182545e-05, "epoch": 2.1728395061728394, "percentage": 44.0, "elapsed_time": "0:28:59", "remaining_time": "0:36:53"}
{"current_steps": 56, "total_steps": 125, "loss": 0.1537, "lr": 6.783107663311565e-05, "epoch": 2.212345679012346, "percentage": 44.8, "elapsed_time": "0:29:28", "remaining_time": "0:36:18"}
{"current_steps": 57, "total_steps": 125, "loss": 0.1264, "lr": 6.651395309775837e-05, "epoch": 2.251851851851852, "percentage": 45.6, "elapsed_time": "0:30:00", "remaining_time": "0:35:48"}
{"current_steps": 58, "total_steps": 125, "loss": 0.1229, "lr": 6.518383725548074e-05, "epoch": 2.291358024691358, "percentage": 46.4, "elapsed_time": "0:30:29", "remaining_time": "0:35:12"}
{"current_steps": 59, "total_steps": 125, "loss": 0.1136, "lr": 6.384177557124247e-05, "epoch": 2.330864197530864, "percentage": 47.2, "elapsed_time": "0:30:55", "remaining_time": "0:34:35"}
{"current_steps": 60, "total_steps": 125, "loss": 0.0987, "lr": 6.248882390836135e-05, "epoch": 2.3703703703703702, "percentage": 48.0, "elapsed_time": "0:31:25", "remaining_time": "0:34:02"}
{"current_steps": 60, "total_steps": 125, "eval_loss": 0.1600760817527771, "epoch": 2.3703703703703702, "percentage": 48.0, "elapsed_time": "0:31:47", "remaining_time": "0:34:26"}
{"current_steps": 61, "total_steps": 125, "loss": 0.1161, "lr": 6.112604669781572e-05, "epoch": 2.4098765432098768, "percentage": 48.8, "elapsed_time": "0:32:21", "remaining_time": "0:33:57"}
{"current_steps": 62, "total_steps": 125, "loss": 0.1828, "lr": 5.9754516100806423e-05, "epoch": 2.449382716049383, "percentage": 49.6, "elapsed_time": "0:32:49", "remaining_time": "0:33:21"}
{"current_steps": 63, "total_steps": 125, "loss": 0.1433, "lr": 5.837531116523682e-05, "epoch": 2.488888888888889, "percentage": 50.4, "elapsed_time": "0:33:19", "remaining_time": "0:32:47"}
{"current_steps": 64, "total_steps": 125, "loss": 0.1724, "lr": 5.698951697677498e-05, "epoch": 2.528395061728395, "percentage": 51.2, "elapsed_time": "0:33:44", "remaining_time": "0:32:09"}
{"current_steps": 65, "total_steps": 125, "loss": 0.1229, "lr": 5.559822380516539e-05, "epoch": 2.567901234567901, "percentage": 52.0, "elapsed_time": "0:34:21", "remaining_time": "0:31:43"}
{"current_steps": 66, "total_steps": 125, "loss": 0.1181, "lr": 5.420252624646238e-05, "epoch": 2.6074074074074076, "percentage": 52.8, "elapsed_time": "0:34:47", "remaining_time": "0:31:06"}
{"current_steps": 67, "total_steps": 125, "loss": 0.1191, "lr": 5.2803522361859594e-05, "epoch": 2.6469135802469137, "percentage": 53.6, "elapsed_time": "0:35:19", "remaining_time": "0:30:34"}
{"current_steps": 68, "total_steps": 125, "loss": 0.1013, "lr": 5.140231281379345e-05, "epoch": 2.68641975308642, "percentage": 54.4, "elapsed_time": "0:35:46", "remaining_time": "0:29:59"}
{"current_steps": 69, "total_steps": 125, "loss": 0.0987, "lr": 5e-05, "epoch": 2.725925925925926, "percentage": 55.2, "elapsed_time": "0:36:11", "remaining_time": "0:29:22"}
{"current_steps": 70, "total_steps": 125, "loss": 0.1066, "lr": 4.859768718620656e-05, "epoch": 2.765432098765432, "percentage": 56.0, "elapsed_time": "0:36:37", "remaining_time": "0:28:46"}
{"current_steps": 70, "total_steps": 125, "eval_loss": 0.15221436321735382, "epoch": 2.765432098765432, "percentage": 56.0, "elapsed_time": "0:36:59", "remaining_time": "0:29:04"}
{"current_steps": 71, "total_steps": 125, "loss": 0.0965, "lr": 4.7196477638140404e-05, "epoch": 2.8049382716049385, "percentage": 56.8, "elapsed_time": "0:37:29", "remaining_time": "0:28:31"}
{"current_steps": 72, "total_steps": 125, "loss": 0.1518, "lr": 4.579747375353763e-05, "epoch": 2.8444444444444446, "percentage": 57.6, "elapsed_time": "0:38:04", "remaining_time": "0:28:01"}
{"current_steps": 73, "total_steps": 125, "loss": 0.1678, "lr": 4.4401776194834613e-05, "epoch": 2.8839506172839506, "percentage": 58.4, "elapsed_time": "0:38:35", "remaining_time": "0:27:29"}
{"current_steps": 74, "total_steps": 125, "loss": 0.1033, "lr": 4.3010483023225045e-05, "epoch": 2.9234567901234567, "percentage": 59.2, "elapsed_time": "0:39:04", "remaining_time": "0:26:55"}
{"current_steps": 75, "total_steps": 125, "loss": 0.1493, "lr": 4.162468883476319e-05, "epoch": 2.962962962962963, "percentage": 60.0, "elapsed_time": "0:39:32", "remaining_time": "0:26:21"}
{"current_steps": 76, "total_steps": 125, "loss": 0.2777, "lr": 4.0245483899193595e-05, "epoch": 3.0024691358024693, "percentage": 60.8, "elapsed_time": "0:40:05", "remaining_time": "0:25:50"}
{"current_steps": 77, "total_steps": 125, "loss": 0.0958, "lr": 3.887395330218429e-05, "epoch": 3.0419753086419754, "percentage": 61.6, "elapsed_time": "0:40:37", "remaining_time": "0:25:19"}
{"current_steps": 78, "total_steps": 125, "loss": 0.104, "lr": 3.7511176091638653e-05, "epoch": 3.0814814814814815, "percentage": 62.4, "elapsed_time": "0:41:05", "remaining_time": "0:24:45"}
{"current_steps": 79, "total_steps": 125, "loss": 0.131, "lr": 3.6158224428757535e-05, "epoch": 3.1209876543209876, "percentage": 63.2, "elapsed_time": "0:41:31", "remaining_time": "0:24:11"}
{"current_steps": 80, "total_steps": 125, "loss": 0.0932, "lr": 3.4816162744519263e-05, "epoch": 3.1604938271604937, "percentage": 64.0, "elapsed_time": "0:41:59", "remaining_time": "0:23:37"}
{"current_steps": 80, "total_steps": 125, "eval_loss": 0.1546574831008911, "epoch": 3.1604938271604937, "percentage": 64.0, "elapsed_time": "0:42:21", "remaining_time": "0:23:49"}
{"current_steps": 81, "total_steps": 125, "loss": 0.0958, "lr": 3.3486046902241664e-05, "epoch": 3.2, "percentage": 64.8, "elapsed_time": "0:42:54", "remaining_time": "0:23:18"}
{"current_steps": 82, "total_steps": 125, "loss": 0.0959, "lr": 3.216892336688435e-05, "epoch": 3.2395061728395063, "percentage": 65.6, "elapsed_time": "0:43:27", "remaining_time": "0:22:47"}
{"current_steps": 83, "total_steps": 125, "loss": 0.1014, "lr": 3.086582838174551e-05, "epoch": 3.2790123456790123, "percentage": 66.4, "elapsed_time": "0:43:56", "remaining_time": "0:22:14"}
{"current_steps": 84, "total_steps": 125, "loss": 0.1139, "lr": 2.9577787153200197e-05, "epoch": 3.3185185185185184, "percentage": 67.2, "elapsed_time": "0:44:25", "remaining_time": "0:21:41"}
{"current_steps": 85, "total_steps": 125, "loss": 0.1066, "lr": 2.8305813044122097e-05, "epoch": 3.3580246913580245, "percentage": 68.0, "elapsed_time": "0:44:53", "remaining_time": "0:21:07"}
{"current_steps": 86, "total_steps": 125, "loss": 0.108, "lr": 2.705090677662311e-05, "epoch": 3.397530864197531, "percentage": 68.8, "elapsed_time": "0:45:26", "remaining_time": "0:20:36"}
{"current_steps": 87, "total_steps": 125, "loss": 0.1205, "lr": 2.581405564473801e-05, "epoch": 3.437037037037037, "percentage": 69.6, "elapsed_time": "0:45:57", "remaining_time": "0:20:04"}
{"current_steps": 88, "total_steps": 125, "loss": 0.1064, "lr": 2.459623273767354e-05, "epoch": 3.476543209876543, "percentage": 70.4, "elapsed_time": "0:46:28", "remaining_time": "0:19:32"}
{"current_steps": 89, "total_steps": 125, "loss": 0.1362, "lr": 2.3398396174233178e-05, "epoch": 3.5160493827160493, "percentage": 71.2, "elapsed_time": "0:46:57", "remaining_time": "0:18:59"}
{"current_steps": 90, "total_steps": 125, "loss": 0.0981, "lr": 2.2221488349019903e-05, "epoch": 3.5555555555555554, "percentage": 72.0, "elapsed_time": "0:47:24", "remaining_time": "0:18:26"}
{"current_steps": 90, "total_steps": 125, "eval_loss": 0.15438534319400787, "epoch": 3.5555555555555554, "percentage": 72.0, "elapsed_time": "0:47:46", "remaining_time": "0:18:34"}
{"current_steps": 91, "total_steps": 125, "loss": 0.1463, "lr": 2.1066435191009715e-05, "epoch": 3.595061728395062, "percentage": 72.8, "elapsed_time": "0:48:18", "remaining_time": "0:18:03"}
{"current_steps": 92, "total_steps": 125, "loss": 0.1175, "lr": 1.9934145435079702e-05, "epoch": 3.634567901234568, "percentage": 73.6, "elapsed_time": "0:48:48", "remaining_time": "0:17:30"}
{"current_steps": 93, "total_steps": 125, "loss": 0.127, "lr": 1.8825509907063327e-05, "epoch": 3.674074074074074, "percentage": 74.4, "elapsed_time": "0:49:19", "remaining_time": "0:16:58"}
{"current_steps": 94, "total_steps": 125, "loss": 0.1193, "lr": 1.774140082289563e-05, "epoch": 3.71358024691358, "percentage": 75.2, "elapsed_time": "0:49:53", "remaining_time": "0:16:27"}
{"current_steps": 95, "total_steps": 125, "loss": 0.1098, "lr": 1.6682671102399805e-05, "epoch": 3.753086419753086, "percentage": 76.0, "elapsed_time": "0:50:19", "remaining_time": "0:15:53"}
{"current_steps": 96, "total_steps": 125, "loss": 0.089, "lr": 1.5650153698254916e-05, "epoch": 3.7925925925925927, "percentage": 76.8, "elapsed_time": "0:50:47", "remaining_time": "0:15:20"}
{"current_steps": 97, "total_steps": 125, "loss": 0.1019, "lr": 1.4644660940672627e-05, "epoch": 3.832098765432099, "percentage": 77.6, "elapsed_time": "0:51:17", "remaining_time": "0:14:48"}
{"current_steps": 98, "total_steps": 125, "loss": 0.0934, "lr": 1.3666983898298657e-05, "epoch": 3.871604938271605, "percentage": 78.4, "elapsed_time": "0:51:42", "remaining_time": "0:14:14"}
{"current_steps": 99, "total_steps": 125, "loss": 0.0942, "lr": 1.2717891755841722e-05, "epoch": 3.911111111111111, "percentage": 79.2, "elapsed_time": "0:52:09", "remaining_time": "0:13:41"}
{"current_steps": 100, "total_steps": 125, "loss": 0.1145, "lr": 1.1798131208919627e-05, "epoch": 3.950617283950617, "percentage": 80.0, "elapsed_time": "0:52:36", "remaining_time": "0:13:09"}
{"current_steps": 100, "total_steps": 125, "eval_loss": 0.14721176028251648, "epoch": 3.950617283950617, "percentage": 80.0, "elapsed_time": "0:52:59", "remaining_time": "0:13:14"}
{"current_steps": 101, "total_steps": 125, "loss": 0.1005, "lr": 1.090842587659851e-05, "epoch": 3.9901234567901236, "percentage": 80.8, "elapsed_time": "0:53:32", "remaining_time": "0:12:43"}
{"current_steps": 102, "total_steps": 125, "loss": 0.1571, "lr": 1.004947573208756e-05, "epoch": 4.029629629629629, "percentage": 81.6, "elapsed_time": "0:54:07", "remaining_time": "0:12:12"}
{"current_steps": 103, "total_steps": 125, "loss": 0.0844, "lr": 9.221956552036992e-06, "epoch": 4.069135802469136, "percentage": 82.4, "elapsed_time": "0:54:36", "remaining_time": "0:11:39"}
{"current_steps": 104, "total_steps": 125, "loss": 0.0907, "lr": 8.426519384872733e-06, "epoch": 4.108641975308642, "percentage": 83.2, "elapsed_time": "0:55:08", "remaining_time": "0:11:07"}
{"current_steps": 105, "total_steps": 125, "loss": 0.085, "lr": 7.663790038585793e-06, "epoch": 4.148148148148148, "percentage": 84.0, "elapsed_time": "0:55:39", "remaining_time": "0:10:36"}
{"current_steps": 106, "total_steps": 125, "loss": 0.0968, "lr": 6.934368588379553e-06, "epoch": 4.187654320987654, "percentage": 84.8, "elapsed_time": "0:56:07", "remaining_time": "0:10:03"}
{"current_steps": 107, "total_steps": 125, "loss": 0.0902, "lr": 6.238828904562316e-06, "epoch": 4.22716049382716, "percentage": 85.6, "elapsed_time": "0:56:36", "remaining_time": "0:09:31"}
{"current_steps": 108, "total_steps": 125, "loss": 0.1146, "lr": 5.577718201056392e-06, "epoch": 4.266666666666667, "percentage": 86.4, "elapsed_time": "0:57:03", "remaining_time": "0:08:58"}
{"current_steps": 109, "total_steps": 125, "loss": 0.096, "lr": 4.951556604879048e-06, "epoch": 4.306172839506173, "percentage": 87.2, "elapsed_time": "0:57:37", "remaining_time": "0:08:27"}
{"current_steps": 110, "total_steps": 125, "loss": 0.0768, "lr": 4.360836746934055e-06, "epoch": 4.345679012345679, "percentage": 88.0, "elapsed_time": "0:58:05", "remaining_time": "0:07:55"}
{"current_steps": 110, "total_steps": 125, "eval_loss": 0.1458897739648819, "epoch": 4.345679012345679, "percentage": 88.0, "elapsed_time": "0:58:28", "remaining_time": "0:07:58"}
{"current_steps": 111, "total_steps": 125, "loss": 0.1107, "lr": 3.8060233744356633e-06, "epoch": 4.385185185185185, "percentage": 88.8, "elapsed_time": "0:58:59", "remaining_time": "0:07:26"}
{"current_steps": 112, "total_steps": 125, "loss": 0.1183, "lr": 3.2875529852700147e-06, "epoch": 4.424691358024692, "percentage": 89.6, "elapsed_time": "0:59:28", "remaining_time": "0:06:54"}
{"current_steps": 113, "total_steps": 125, "loss": 0.0913, "lr": 2.8058334845816213e-06, "epoch": 4.4641975308641975, "percentage": 90.4, "elapsed_time": "0:59:53", "remaining_time": "0:06:21"}
{"current_steps": 114, "total_steps": 125, "loss": 0.0835, "lr": 2.361243863855184e-06, "epoch": 4.503703703703704, "percentage": 91.2, "elapsed_time": "1:00:20", "remaining_time": "0:05:49"}
{"current_steps": 115, "total_steps": 125, "loss": 0.1096, "lr": 1.9541339027450256e-06, "epoch": 4.54320987654321, "percentage": 92.0, "elapsed_time": "1:00:48", "remaining_time": "0:05:17"}
{"current_steps": 116, "total_steps": 125, "loss": 0.0949, "lr": 1.584823893886933e-06, "epoch": 4.582716049382716, "percentage": 92.8, "elapsed_time": "1:01:17", "remaining_time": "0:04:45"}
{"current_steps": 117, "total_steps": 125, "loss": 0.086, "lr": 1.2536043909088191e-06, "epoch": 4.622222222222222, "percentage": 93.6, "elapsed_time": "1:01:47", "remaining_time": "0:04:13"}
{"current_steps": 118, "total_steps": 125, "loss": 0.083, "lr": 9.607359798384785e-07, "epoch": 4.661728395061728, "percentage": 94.4, "elapsed_time": "1:02:13", "remaining_time": "0:03:41"}
{"current_steps": 119, "total_steps": 125, "loss": 0.1213, "lr": 7.064490740882057e-07, "epoch": 4.701234567901235, "percentage": 95.2, "elapsed_time": "1:02:41", "remaining_time": "0:03:09"}
{"current_steps": 120, "total_steps": 125, "loss": 0.0802, "lr": 4.909437331777179e-07, "epoch": 4.7407407407407405, "percentage": 96.0, "elapsed_time": "1:03:14", "remaining_time": "0:02:38"}
{"current_steps": 120, "total_steps": 125, "eval_loss": 0.14555972814559937, "epoch": 4.7407407407407405, "percentage": 96.0, "elapsed_time": "1:03:36", "remaining_time": "0:02:39"}
{"current_steps": 121, "total_steps": 125, "loss": 0.0878, "lr": 3.143895053378698e-07, "epoch": 4.780246913580247, "percentage": 96.8, "elapsed_time": "1:04:07", "remaining_time": "0:02:07"}
{"current_steps": 122, "total_steps": 125, "loss": 0.1139, "lr": 1.7692529411904578e-07, "epoch": 4.8197530864197535, "percentage": 97.6, "elapsed_time": "1:04:37", "remaining_time": "0:01:35"}
{"current_steps": 123, "total_steps": 125, "loss": 0.0863, "lr": 7.865924910916977e-08, "epoch": 4.859259259259259, "percentage": 98.4, "elapsed_time": "1:05:08", "remaining_time": "0:01:03"}
{"current_steps": 124, "total_steps": 125, "loss": 0.1279, "lr": 1.9668680847356735e-08, "epoch": 4.898765432098766, "percentage": 99.2, "elapsed_time": "1:05:36", "remaining_time": "0:00:31"}
{"current_steps": 125, "total_steps": 125, "loss": 0.1065, "lr": 0.0, "epoch": 4.938271604938271, "percentage": 100.0, "elapsed_time": "1:06:03", "remaining_time": "0:00:00"}
{"current_steps": 125, "total_steps": 125, "epoch": 4.938271604938271, "percentage": 100.0, "elapsed_time": "1:06:06", "remaining_time": "0:00:00"}