LLAMA-3.2-1B-MathInstrcu-LORA / trainer_log.jsonl
SriSanth2345's picture
Upload 224 files
242ef39 verified
{"current_steps": 5, "total_steps": 234, "loss": 1.0617, "lr": 1.9977477585156252e-05, "epoch": 0.064, "percentage": 2.14, "elapsed_time": "0:00:51", "remaining_time": "0:39:31"}
{"current_steps": 5, "total_steps": 1875, "loss": 1.2049, "lr": 0.0001999964908278481, "epoch": 0.008, "percentage": 0.27, "elapsed_time": "0:00:05", "remaining_time": "0:34:31"}
{"current_steps": 10, "total_steps": 1875, "loss": 0.9333, "lr": 0.00019998596355767805, "epoch": 0.016, "percentage": 0.53, "elapsed_time": "0:00:13", "remaining_time": "0:40:38"}
{"current_steps": 15, "total_steps": 1875, "loss": 0.8671, "lr": 0.00019996841892833, "epoch": 0.024, "percentage": 0.8, "elapsed_time": "0:00:19", "remaining_time": "0:41:10"}
{"current_steps": 20, "total_steps": 1875, "loss": 0.7979, "lr": 0.00019994385817114646, "epoch": 0.032, "percentage": 1.07, "elapsed_time": "0:00:25", "remaining_time": "0:39:02"}
{"current_steps": 25, "total_steps": 1875, "loss": 0.7662, "lr": 0.00019991228300988585, "epoch": 0.04, "percentage": 1.33, "elapsed_time": "0:00:31", "remaining_time": "0:38:18"}
{"current_steps": 30, "total_steps": 1875, "loss": 0.7929, "lr": 0.00019987369566060176, "epoch": 0.048, "percentage": 1.6, "elapsed_time": "0:00:37", "remaining_time": "0:38:54"}
{"current_steps": 35, "total_steps": 1875, "loss": 0.7683, "lr": 0.00019982809883148722, "epoch": 0.056, "percentage": 1.87, "elapsed_time": "0:00:44", "remaining_time": "0:38:41"}
{"current_steps": 40, "total_steps": 1875, "loss": 0.8667, "lr": 0.00019977549572268468, "epoch": 0.064, "percentage": 2.13, "elapsed_time": "0:00:50", "remaining_time": "0:38:42"}
{"current_steps": 45, "total_steps": 1875, "loss": 0.8446, "lr": 0.0001997158900260614, "epoch": 0.072, "percentage": 2.4, "elapsed_time": "0:00:58", "remaining_time": "0:39:46"}
{"current_steps": 50, "total_steps": 1875, "loss": 0.9051, "lr": 0.00019964928592495045, "epoch": 0.08, "percentage": 2.67, "elapsed_time": "0:01:03", "remaining_time": "0:38:40"}
{"current_steps": 55, "total_steps": 1875, "loss": 0.7235, "lr": 0.00019957568809385694, "epoch": 0.088, "percentage": 2.93, "elapsed_time": "0:01:08", "remaining_time": "0:37:45"}
{"current_steps": 60, "total_steps": 1875, "loss": 0.8169, "lr": 0.00019949510169813003, "epoch": 0.096, "percentage": 3.2, "elapsed_time": "0:01:12", "remaining_time": "0:36:28"}
{"current_steps": 65, "total_steps": 1875, "loss": 0.8266, "lr": 0.00019940753239360047, "epoch": 0.104, "percentage": 3.47, "elapsed_time": "0:01:18", "remaining_time": "0:36:17"}
{"current_steps": 70, "total_steps": 1875, "loss": 0.758, "lr": 0.00019931298632618356, "epoch": 0.112, "percentage": 3.73, "elapsed_time": "0:01:23", "remaining_time": "0:35:52"}
{"current_steps": 75, "total_steps": 1875, "loss": 0.7759, "lr": 0.0001992114701314478, "epoch": 0.12, "percentage": 4.0, "elapsed_time": "0:01:30", "remaining_time": "0:36:07"}
{"current_steps": 80, "total_steps": 1875, "loss": 0.7797, "lr": 0.0001991029909341493, "epoch": 0.128, "percentage": 4.27, "elapsed_time": "0:01:37", "remaining_time": "0:36:20"}
{"current_steps": 85, "total_steps": 1875, "loss": 0.7437, "lr": 0.00019898755634773158, "epoch": 0.136, "percentage": 4.53, "elapsed_time": "0:01:43", "remaining_time": "0:36:21"}
{"current_steps": 90, "total_steps": 1875, "loss": 0.8043, "lr": 0.0001988651744737914, "epoch": 0.144, "percentage": 4.8, "elapsed_time": "0:01:50", "remaining_time": "0:36:26"}
{"current_steps": 95, "total_steps": 1875, "loss": 0.7701, "lr": 0.00019873585390151003, "epoch": 0.152, "percentage": 5.07, "elapsed_time": "0:01:57", "remaining_time": "0:36:34"}
{"current_steps": 100, "total_steps": 1875, "loss": 0.709, "lr": 0.0001985996037070505, "epoch": 0.16, "percentage": 5.33, "elapsed_time": "0:02:04", "remaining_time": "0:36:46"}
{"current_steps": 105, "total_steps": 1875, "loss": 0.7377, "lr": 0.00019845643345292054, "epoch": 0.168, "percentage": 5.6, "elapsed_time": "0:02:11", "remaining_time": "0:36:57"}
{"current_steps": 110, "total_steps": 1875, "loss": 0.8352, "lr": 0.00019830635318730154, "epoch": 0.176, "percentage": 5.87, "elapsed_time": "0:02:17", "remaining_time": "0:36:45"}
{"current_steps": 115, "total_steps": 1875, "loss": 0.7738, "lr": 0.0001981493734433433, "epoch": 0.184, "percentage": 6.13, "elapsed_time": "0:02:24", "remaining_time": "0:36:55"}
{"current_steps": 120, "total_steps": 1875, "loss": 0.8067, "lr": 0.0001979855052384247, "epoch": 0.192, "percentage": 6.4, "elapsed_time": "0:02:29", "remaining_time": "0:36:32"}
{"current_steps": 125, "total_steps": 1875, "loss": 0.7456, "lr": 0.00019781476007338058, "epoch": 0.2, "percentage": 6.67, "elapsed_time": "0:02:35", "remaining_time": "0:36:13"}
{"current_steps": 130, "total_steps": 1875, "loss": 0.758, "lr": 0.00019763714993169452, "epoch": 0.208, "percentage": 6.93, "elapsed_time": "0:02:42", "remaining_time": "0:36:20"}
{"current_steps": 135, "total_steps": 1875, "loss": 0.7895, "lr": 0.00019745268727865774, "epoch": 0.216, "percentage": 7.2, "elapsed_time": "0:02:48", "remaining_time": "0:36:10"}
{"current_steps": 140, "total_steps": 1875, "loss": 0.7302, "lr": 0.00019726138506049438, "epoch": 0.224, "percentage": 7.47, "elapsed_time": "0:02:56", "remaining_time": "0:36:26"}
{"current_steps": 145, "total_steps": 1875, "loss": 0.8152, "lr": 0.00019706325670345275, "epoch": 0.232, "percentage": 7.73, "elapsed_time": "0:03:01", "remaining_time": "0:36:11"}
{"current_steps": 150, "total_steps": 1875, "loss": 0.8461, "lr": 0.0001968583161128631, "epoch": 0.24, "percentage": 8.0, "elapsed_time": "0:03:06", "remaining_time": "0:35:45"}
{"current_steps": 155, "total_steps": 1875, "loss": 0.7787, "lr": 0.00019664657767216176, "epoch": 0.248, "percentage": 8.27, "elapsed_time": "0:03:12", "remaining_time": "0:35:34"}
{"current_steps": 160, "total_steps": 1875, "loss": 0.7574, "lr": 0.00019642805624188147, "epoch": 0.256, "percentage": 8.53, "elapsed_time": "0:03:18", "remaining_time": "0:35:27"}
{"current_steps": 165, "total_steps": 1875, "loss": 0.8487, "lr": 0.0001962027671586086, "epoch": 0.264, "percentage": 8.8, "elapsed_time": "0:03:22", "remaining_time": "0:35:00"}
{"current_steps": 170, "total_steps": 1875, "loss": 0.6611, "lr": 0.00019597072623390668, "epoch": 0.272, "percentage": 9.07, "elapsed_time": "0:03:29", "remaining_time": "0:34:56"}
{"current_steps": 175, "total_steps": 1875, "loss": 0.7802, "lr": 0.00019573194975320673, "epoch": 0.28, "percentage": 9.33, "elapsed_time": "0:03:35", "remaining_time": "0:34:55"}
{"current_steps": 180, "total_steps": 1875, "loss": 0.6727, "lr": 0.00019548645447466431, "epoch": 0.288, "percentage": 9.6, "elapsed_time": "0:03:42", "remaining_time": "0:34:56"}
{"current_steps": 185, "total_steps": 1875, "loss": 0.7502, "lr": 0.00019523425762798329, "epoch": 0.296, "percentage": 9.87, "elapsed_time": "0:03:47", "remaining_time": "0:34:38"}
{"current_steps": 190, "total_steps": 1875, "loss": 0.8401, "lr": 0.00019497537691320668, "epoch": 0.304, "percentage": 10.13, "elapsed_time": "0:03:53", "remaining_time": "0:34:30"}
{"current_steps": 195, "total_steps": 1875, "loss": 0.7494, "lr": 0.00019470983049947444, "epoch": 0.312, "percentage": 10.4, "elapsed_time": "0:03:58", "remaining_time": "0:34:16"}
{"current_steps": 200, "total_steps": 1875, "loss": 0.7842, "lr": 0.00019443763702374812, "epoch": 0.32, "percentage": 10.67, "elapsed_time": "0:04:04", "remaining_time": "0:34:11"}
{"current_steps": 205, "total_steps": 1875, "loss": 0.8082, "lr": 0.00019415881558950302, "epoch": 0.328, "percentage": 10.93, "elapsed_time": "0:04:12", "remaining_time": "0:34:15"}
{"current_steps": 210, "total_steps": 1875, "loss": 0.7883, "lr": 0.00019387338576538744, "epoch": 0.336, "percentage": 11.2, "elapsed_time": "0:04:17", "remaining_time": "0:34:01"}
{"current_steps": 215, "total_steps": 1875, "loss": 0.7356, "lr": 0.00019358136758384912, "epoch": 0.344, "percentage": 11.47, "elapsed_time": "0:04:25", "remaining_time": "0:34:13"}
{"current_steps": 220, "total_steps": 1875, "loss": 0.7891, "lr": 0.00019328278153972947, "epoch": 0.352, "percentage": 11.73, "elapsed_time": "0:04:31", "remaining_time": "0:34:04"}
{"current_steps": 225, "total_steps": 1875, "loss": 0.7671, "lr": 0.00019297764858882514, "epoch": 0.36, "percentage": 12.0, "elapsed_time": "0:04:39", "remaining_time": "0:34:06"}
{"current_steps": 230, "total_steps": 1875, "loss": 0.6608, "lr": 0.0001926659901464172, "epoch": 0.368, "percentage": 12.27, "elapsed_time": "0:04:45", "remaining_time": "0:34:04"}
{"current_steps": 235, "total_steps": 1875, "loss": 0.647, "lr": 0.00019234782808576824, "epoch": 0.376, "percentage": 12.53, "elapsed_time": "0:04:54", "remaining_time": "0:34:17"}
{"current_steps": 240, "total_steps": 1875, "loss": 0.729, "lr": 0.00019202318473658705, "epoch": 0.384, "percentage": 12.8, "elapsed_time": "0:05:02", "remaining_time": "0:34:18"}
{"current_steps": 245, "total_steps": 1875, "loss": 0.6713, "lr": 0.00019169208288346166, "epoch": 0.392, "percentage": 13.07, "elapsed_time": "0:05:08", "remaining_time": "0:34:09"}
{"current_steps": 250, "total_steps": 1875, "loss": 0.7049, "lr": 0.0001913545457642601, "epoch": 0.4, "percentage": 13.33, "elapsed_time": "0:05:16", "remaining_time": "0:34:19"}
{"current_steps": 255, "total_steps": 1875, "loss": 0.7419, "lr": 0.00019101059706849957, "epoch": 0.408, "percentage": 13.6, "elapsed_time": "0:05:24", "remaining_time": "0:34:20"}
{"current_steps": 260, "total_steps": 1875, "loss": 0.7148, "lr": 0.00019066026093568378, "epoch": 0.416, "percentage": 13.87, "elapsed_time": "0:05:29", "remaining_time": "0:34:08"}
{"current_steps": 265, "total_steps": 1875, "loss": 0.7493, "lr": 0.00019030356195360874, "epoch": 0.424, "percentage": 14.13, "elapsed_time": "0:05:37", "remaining_time": "0:34:10"}
{"current_steps": 270, "total_steps": 1875, "loss": 0.7652, "lr": 0.0001899405251566371, "epoch": 0.432, "percentage": 14.4, "elapsed_time": "0:05:43", "remaining_time": "0:34:01"}
{"current_steps": 275, "total_steps": 1875, "loss": 0.7438, "lr": 0.0001895711760239413, "epoch": 0.44, "percentage": 14.67, "elapsed_time": "0:05:50", "remaining_time": "0:33:59"}
{"current_steps": 280, "total_steps": 1875, "loss": 0.7683, "lr": 0.0001891955404777151, "epoch": 0.448, "percentage": 14.93, "elapsed_time": "0:05:55", "remaining_time": "0:33:44"}
{"current_steps": 285, "total_steps": 1875, "loss": 0.8115, "lr": 0.00018881364488135448, "epoch": 0.456, "percentage": 15.2, "elapsed_time": "0:06:00", "remaining_time": "0:33:30"}
{"current_steps": 290, "total_steps": 1875, "loss": 0.8335, "lr": 0.00018842551603760724, "epoch": 0.464, "percentage": 15.47, "elapsed_time": "0:06:06", "remaining_time": "0:33:25"}
{"current_steps": 295, "total_steps": 1875, "loss": 0.6933, "lr": 0.00018803118118669202, "epoch": 0.472, "percentage": 15.73, "elapsed_time": "0:06:13", "remaining_time": "0:33:20"}
{"current_steps": 300, "total_steps": 1875, "loss": 0.7515, "lr": 0.00018763066800438636, "epoch": 0.48, "percentage": 16.0, "elapsed_time": "0:06:17", "remaining_time": "0:33:04"}
{"current_steps": 305, "total_steps": 1875, "loss": 0.6931, "lr": 0.0001872240046000844, "epoch": 0.488, "percentage": 16.27, "elapsed_time": "0:06:25", "remaining_time": "0:33:05"}
{"current_steps": 310, "total_steps": 1875, "loss": 0.782, "lr": 0.00018681121951482393, "epoch": 0.496, "percentage": 16.53, "elapsed_time": "0:06:31", "remaining_time": "0:32:58"}
{"current_steps": 315, "total_steps": 1875, "loss": 0.7361, "lr": 0.00018639234171928353, "epoch": 0.504, "percentage": 16.8, "elapsed_time": "0:06:38", "remaining_time": "0:32:52"}
{"current_steps": 320, "total_steps": 1875, "loss": 0.7443, "lr": 0.0001859674006117491, "epoch": 0.512, "percentage": 17.07, "elapsed_time": "0:06:44", "remaining_time": "0:32:43"}
{"current_steps": 325, "total_steps": 1875, "loss": 0.7221, "lr": 0.00018553642601605068, "epoch": 0.52, "percentage": 17.33, "elapsed_time": "0:06:51", "remaining_time": "0:32:41"}
{"current_steps": 330, "total_steps": 1875, "loss": 0.7622, "lr": 0.00018509944817946922, "epoch": 0.528, "percentage": 17.6, "elapsed_time": "0:06:57", "remaining_time": "0:32:33"}
{"current_steps": 335, "total_steps": 1875, "loss": 0.8556, "lr": 0.0001846564977706138, "epoch": 0.536, "percentage": 17.87, "elapsed_time": "0:07:03", "remaining_time": "0:32:25"}
{"current_steps": 340, "total_steps": 1875, "loss": 0.7814, "lr": 0.00018420760587726923, "epoch": 0.544, "percentage": 18.13, "elapsed_time": "0:07:09", "remaining_time": "0:32:17"}
{"current_steps": 345, "total_steps": 1875, "loss": 0.722, "lr": 0.0001837528040042142, "epoch": 0.552, "percentage": 18.4, "elapsed_time": "0:07:18", "remaining_time": "0:32:24"}
{"current_steps": 350, "total_steps": 1875, "loss": 0.7903, "lr": 0.00018329212407100994, "epoch": 0.56, "percentage": 18.67, "elapsed_time": "0:07:22", "remaining_time": "0:32:09"}
{"current_steps": 355, "total_steps": 1875, "loss": 0.6996, "lr": 0.00018282559840976042, "epoch": 0.568, "percentage": 18.93, "elapsed_time": "0:07:28", "remaining_time": "0:31:58"}
{"current_steps": 360, "total_steps": 1875, "loss": 0.773, "lr": 0.00018235325976284275, "epoch": 0.576, "percentage": 19.2, "elapsed_time": "0:07:33", "remaining_time": "0:31:49"}
{"current_steps": 365, "total_steps": 1875, "loss": 0.728, "lr": 0.00018187514128060946, "epoch": 0.584, "percentage": 19.47, "elapsed_time": "0:07:40", "remaining_time": "0:31:43"}
{"current_steps": 370, "total_steps": 1875, "loss": 0.7659, "lr": 0.00018139127651906184, "epoch": 0.592, "percentage": 19.73, "elapsed_time": "0:07:45", "remaining_time": "0:31:33"}
{"current_steps": 375, "total_steps": 1875, "loss": 0.7039, "lr": 0.00018090169943749476, "epoch": 0.6, "percentage": 20.0, "elapsed_time": "0:07:50", "remaining_time": "0:31:21"}
{"current_steps": 380, "total_steps": 1875, "loss": 0.7125, "lr": 0.00018040644439611348, "epoch": 0.608, "percentage": 20.27, "elapsed_time": "0:07:56", "remaining_time": "0:31:14"}
{"current_steps": 385, "total_steps": 1875, "loss": 0.698, "lr": 0.00017990554615362198, "epoch": 0.616, "percentage": 20.53, "elapsed_time": "0:08:03", "remaining_time": "0:31:11"}
{"current_steps": 390, "total_steps": 1875, "loss": 0.8255, "lr": 0.00017939903986478355, "epoch": 0.624, "percentage": 20.8, "elapsed_time": "0:08:08", "remaining_time": "0:30:59"}
{"current_steps": 395, "total_steps": 1875, "loss": 0.6616, "lr": 0.00017888696107795342, "epoch": 0.632, "percentage": 21.07, "elapsed_time": "0:08:13", "remaining_time": "0:30:50"}
{"current_steps": 400, "total_steps": 1875, "loss": 0.7452, "lr": 0.000178369345732584, "epoch": 0.64, "percentage": 21.33, "elapsed_time": "0:08:18", "remaining_time": "0:30:39"}
{"current_steps": 405, "total_steps": 1875, "loss": 0.7652, "lr": 0.00017784623015670238, "epoch": 0.648, "percentage": 21.6, "elapsed_time": "0:08:27", "remaining_time": "0:30:40"}
{"current_steps": 410, "total_steps": 1875, "loss": 0.7793, "lr": 0.00017731765106436073, "epoch": 0.656, "percentage": 21.87, "elapsed_time": "0:08:34", "remaining_time": "0:30:39"}
{"current_steps": 415, "total_steps": 1875, "loss": 0.6875, "lr": 0.00017678364555305978, "epoch": 0.664, "percentage": 22.13, "elapsed_time": "0:08:40", "remaining_time": "0:30:29"}
{"current_steps": 420, "total_steps": 1875, "loss": 0.7465, "lr": 0.0001762442511011448, "epoch": 0.672, "percentage": 22.4, "elapsed_time": "0:08:48", "remaining_time": "0:30:30"}
{"current_steps": 425, "total_steps": 1875, "loss": 0.7205, "lr": 0.00017569950556517566, "epoch": 0.68, "percentage": 22.67, "elapsed_time": "0:08:55", "remaining_time": "0:30:27"}
{"current_steps": 430, "total_steps": 1875, "loss": 0.6589, "lr": 0.00017514944717726962, "epoch": 0.688, "percentage": 22.93, "elapsed_time": "0:09:02", "remaining_time": "0:30:22"}
{"current_steps": 435, "total_steps": 1875, "loss": 0.7035, "lr": 0.00017459411454241822, "epoch": 0.696, "percentage": 23.2, "elapsed_time": "0:09:08", "remaining_time": "0:30:14"}
{"current_steps": 440, "total_steps": 1875, "loss": 0.787, "lr": 0.00017403354663577783, "epoch": 0.704, "percentage": 23.47, "elapsed_time": "0:09:15", "remaining_time": "0:30:10"}
{"current_steps": 445, "total_steps": 1875, "loss": 0.7515, "lr": 0.00017346778279993415, "epoch": 0.712, "percentage": 23.73, "elapsed_time": "0:09:23", "remaining_time": "0:30:09"}
{"current_steps": 450, "total_steps": 1875, "loss": 0.7199, "lr": 0.00017289686274214118, "epoch": 0.72, "percentage": 24.0, "elapsed_time": "0:09:28", "remaining_time": "0:30:01"}
{"current_steps": 455, "total_steps": 1875, "loss": 0.8037, "lr": 0.00017232082653153422, "epoch": 0.728, "percentage": 24.27, "elapsed_time": "0:09:35", "remaining_time": "0:29:55"}
{"current_steps": 460, "total_steps": 1875, "loss": 0.7502, "lr": 0.00017173971459631787, "epoch": 0.736, "percentage": 24.53, "elapsed_time": "0:09:40", "remaining_time": "0:29:46"}
{"current_steps": 465, "total_steps": 1875, "loss": 0.7446, "lr": 0.00017115356772092857, "epoch": 0.744, "percentage": 24.8, "elapsed_time": "0:09:47", "remaining_time": "0:29:42"}
{"current_steps": 470, "total_steps": 1875, "loss": 0.6507, "lr": 0.0001705624270431721, "epoch": 0.752, "percentage": 25.07, "elapsed_time": "0:09:54", "remaining_time": "0:29:36"}
{"current_steps": 475, "total_steps": 1875, "loss": 0.7164, "lr": 0.00016996633405133655, "epoch": 0.76, "percentage": 25.33, "elapsed_time": "0:10:00", "remaining_time": "0:29:29"}
{"current_steps": 480, "total_steps": 1875, "loss": 0.7621, "lr": 0.0001693653305812805, "epoch": 0.768, "percentage": 25.6, "elapsed_time": "0:10:06", "remaining_time": "0:29:23"}
{"current_steps": 485, "total_steps": 1875, "loss": 0.7623, "lr": 0.00016875945881349676, "epoch": 0.776, "percentage": 25.87, "elapsed_time": "0:10:12", "remaining_time": "0:29:16"}
{"current_steps": 490, "total_steps": 1875, "loss": 0.6606, "lr": 0.000168148761270152, "epoch": 0.784, "percentage": 26.13, "elapsed_time": "0:10:20", "remaining_time": "0:29:14"}
{"current_steps": 495, "total_steps": 1875, "loss": 0.6941, "lr": 0.00016753328081210245, "epoch": 0.792, "percentage": 26.4, "elapsed_time": "0:10:28", "remaining_time": "0:29:13"}
{"current_steps": 500, "total_steps": 1875, "loss": 0.6841, "lr": 0.00016691306063588583, "epoch": 0.8, "percentage": 26.67, "elapsed_time": "0:10:33", "remaining_time": "0:29:02"}
{"current_steps": 505, "total_steps": 1875, "loss": 0.6996, "lr": 0.00016628814427068953, "epoch": 0.808, "percentage": 26.93, "elapsed_time": "0:10:41", "remaining_time": "0:28:59"}
{"current_steps": 510, "total_steps": 1875, "loss": 0.7542, "lr": 0.00016565857557529566, "epoch": 0.816, "percentage": 27.2, "elapsed_time": "0:10:48", "remaining_time": "0:28:55"}
{"current_steps": 515, "total_steps": 1875, "loss": 0.7175, "lr": 0.00016502439873500289, "epoch": 0.824, "percentage": 27.47, "elapsed_time": "0:10:54", "remaining_time": "0:28:47"}
{"current_steps": 520, "total_steps": 1875, "loss": 0.7565, "lr": 0.0001643856582585254, "epoch": 0.832, "percentage": 27.73, "elapsed_time": "0:10:59", "remaining_time": "0:28:37"}
{"current_steps": 525, "total_steps": 1875, "loss": 0.7339, "lr": 0.000163742398974869, "epoch": 0.84, "percentage": 28.0, "elapsed_time": "0:11:04", "remaining_time": "0:28:29"}
{"current_steps": 530, "total_steps": 1875, "loss": 0.569, "lr": 0.00016309466603018496, "epoch": 0.848, "percentage": 28.27, "elapsed_time": "0:11:10", "remaining_time": "0:28:22"}
{"current_steps": 535, "total_steps": 1875, "loss": 0.7556, "lr": 0.00016244250488460158, "epoch": 0.856, "percentage": 28.53, "elapsed_time": "0:11:15", "remaining_time": "0:28:11"}
{"current_steps": 540, "total_steps": 1875, "loss": 0.7084, "lr": 0.00016178596130903344, "epoch": 0.864, "percentage": 28.8, "elapsed_time": "0:11:19", "remaining_time": "0:28:00"}
{"current_steps": 545, "total_steps": 1875, "loss": 0.6935, "lr": 0.00016112508138196917, "epoch": 0.872, "percentage": 29.07, "elapsed_time": "0:11:25", "remaining_time": "0:27:53"}
{"current_steps": 550, "total_steps": 1875, "loss": 0.7076, "lr": 0.0001604599114862375, "epoch": 0.88, "percentage": 29.33, "elapsed_time": "0:11:31", "remaining_time": "0:27:44"}
{"current_steps": 555, "total_steps": 1875, "loss": 0.7151, "lr": 0.0001597904983057519, "epoch": 0.888, "percentage": 29.6, "elapsed_time": "0:11:36", "remaining_time": "0:27:37"}
{"current_steps": 560, "total_steps": 1875, "loss": 0.7001, "lr": 0.0001591168888222342, "epoch": 0.896, "percentage": 29.87, "elapsed_time": "0:11:42", "remaining_time": "0:27:29"}
{"current_steps": 565, "total_steps": 1875, "loss": 0.7285, "lr": 0.00015843913031191723, "epoch": 0.904, "percentage": 30.13, "elapsed_time": "0:11:47", "remaining_time": "0:27:19"}
{"current_steps": 570, "total_steps": 1875, "loss": 0.8041, "lr": 0.00015775727034222675, "epoch": 0.912, "percentage": 30.4, "elapsed_time": "0:11:53", "remaining_time": "0:27:13"}
{"current_steps": 575, "total_steps": 1875, "loss": 0.7353, "lr": 0.0001570713567684432, "epoch": 0.92, "percentage": 30.67, "elapsed_time": "0:12:00", "remaining_time": "0:27:08"}
{"current_steps": 580, "total_steps": 1875, "loss": 0.7792, "lr": 0.00015638143773034267, "epoch": 0.928, "percentage": 30.93, "elapsed_time": "0:12:05", "remaining_time": "0:27:00"}
{"current_steps": 585, "total_steps": 1875, "loss": 1.0121, "lr": 0.00015568756164881882, "epoch": 0.936, "percentage": 31.2, "elapsed_time": "0:12:12", "remaining_time": "0:26:55"}
{"current_steps": 590, "total_steps": 1875, "loss": 0.7727, "lr": 0.000154989777222484, "epoch": 0.944, "percentage": 31.47, "elapsed_time": "0:12:18", "remaining_time": "0:26:48"}
{"current_steps": 595, "total_steps": 1875, "loss": 0.741, "lr": 0.00015428813342425177, "epoch": 0.952, "percentage": 31.73, "elapsed_time": "0:12:24", "remaining_time": "0:26:41"}
{"current_steps": 600, "total_steps": 1875, "loss": 0.6919, "lr": 0.00015358267949789966, "epoch": 0.96, "percentage": 32.0, "elapsed_time": "0:12:30", "remaining_time": "0:26:34"}
{"current_steps": 605, "total_steps": 1875, "loss": 0.7163, "lr": 0.00015287346495461315, "epoch": 0.968, "percentage": 32.27, "elapsed_time": "0:12:38", "remaining_time": "0:26:31"}
{"current_steps": 610, "total_steps": 1875, "loss": 0.8152, "lr": 0.0001521605395695108, "epoch": 0.976, "percentage": 32.53, "elapsed_time": "0:12:42", "remaining_time": "0:26:21"}
{"current_steps": 615, "total_steps": 1875, "loss": 0.6709, "lr": 0.00015144395337815064, "epoch": 0.984, "percentage": 32.8, "elapsed_time": "0:12:48", "remaining_time": "0:26:15"}
{"current_steps": 620, "total_steps": 1875, "loss": 0.6527, "lr": 0.00015072375667301893, "epoch": 0.992, "percentage": 33.07, "elapsed_time": "0:12:54", "remaining_time": "0:26:07"}
{"current_steps": 625, "total_steps": 1875, "loss": 0.8194, "lr": 0.00015000000000000001, "epoch": 1.0, "percentage": 33.33, "elapsed_time": "0:12:59", "remaining_time": "0:25:58"}
{"current_steps": 630, "total_steps": 1875, "loss": 0.6627, "lr": 0.00014927273415482915, "epoch": 1.008, "percentage": 33.6, "elapsed_time": "0:13:04", "remaining_time": "0:25:51"}
{"current_steps": 635, "total_steps": 1875, "loss": 0.6366, "lr": 0.0001485420101795274, "epoch": 1.016, "percentage": 33.87, "elapsed_time": "0:13:10", "remaining_time": "0:25:44"}
{"current_steps": 640, "total_steps": 1875, "loss": 0.6717, "lr": 0.00014780787935881923, "epoch": 1.024, "percentage": 34.13, "elapsed_time": "0:13:15", "remaining_time": "0:25:35"}
{"current_steps": 645, "total_steps": 1875, "loss": 0.6483, "lr": 0.0001470703932165333, "epoch": 1.032, "percentage": 34.4, "elapsed_time": "0:13:20", "remaining_time": "0:25:26"}
{"current_steps": 650, "total_steps": 1875, "loss": 0.6151, "lr": 0.00014632960351198618, "epoch": 1.04, "percentage": 34.67, "elapsed_time": "0:13:25", "remaining_time": "0:25:18"}
{"current_steps": 655, "total_steps": 1875, "loss": 0.6707, "lr": 0.00014558556223635003, "epoch": 1.048, "percentage": 34.93, "elapsed_time": "0:13:32", "remaining_time": "0:25:13"}
{"current_steps": 660, "total_steps": 1875, "loss": 0.6125, "lr": 0.00014483832160900326, "epoch": 1.056, "percentage": 35.2, "elapsed_time": "0:13:37", "remaining_time": "0:25:04"}
{"current_steps": 665, "total_steps": 1875, "loss": 0.6206, "lr": 0.00014408793407386588, "epoch": 1.064, "percentage": 35.47, "elapsed_time": "0:13:42", "remaining_time": "0:24:56"}
{"current_steps": 670, "total_steps": 1875, "loss": 0.6161, "lr": 0.00014333445229571873, "epoch": 1.072, "percentage": 35.73, "elapsed_time": "0:13:49", "remaining_time": "0:24:50"}
{"current_steps": 675, "total_steps": 1875, "loss": 0.6583, "lr": 0.00014257792915650728, "epoch": 1.08, "percentage": 36.0, "elapsed_time": "0:13:54", "remaining_time": "0:24:43"}
{"current_steps": 680, "total_steps": 1875, "loss": 0.6222, "lr": 0.00014181841775163013, "epoch": 1.088, "percentage": 36.27, "elapsed_time": "0:13:59", "remaining_time": "0:24:34"}
{"current_steps": 685, "total_steps": 1875, "loss": 0.716, "lr": 0.0001410559713862128, "epoch": 1.096, "percentage": 36.53, "elapsed_time": "0:14:04", "remaining_time": "0:24:26"}
{"current_steps": 690, "total_steps": 1875, "loss": 0.6198, "lr": 0.00014029064357136628, "epoch": 1.104, "percentage": 36.8, "elapsed_time": "0:14:10", "remaining_time": "0:24:21"}
{"current_steps": 695, "total_steps": 1875, "loss": 0.6389, "lr": 0.00013952248802043165, "epoch": 1.112, "percentage": 37.07, "elapsed_time": "0:14:18", "remaining_time": "0:24:17"}
{"current_steps": 700, "total_steps": 1875, "loss": 0.6842, "lr": 0.0001387515586452103, "epoch": 1.12, "percentage": 37.33, "elapsed_time": "0:14:24", "remaining_time": "0:24:10"}
{"current_steps": 705, "total_steps": 1875, "loss": 0.6071, "lr": 0.00013797790955218014, "epoch": 1.1280000000000001, "percentage": 37.6, "elapsed_time": "0:14:32", "remaining_time": "0:24:08"}
{"current_steps": 710, "total_steps": 1875, "loss": 0.5915, "lr": 0.00013720159503869815, "epoch": 1.1360000000000001, "percentage": 37.87, "elapsed_time": "0:14:40", "remaining_time": "0:24:04"}
{"current_steps": 715, "total_steps": 1875, "loss": 0.6794, "lr": 0.00013642266958918984, "epoch": 1.144, "percentage": 38.13, "elapsed_time": "0:14:46", "remaining_time": "0:23:57"}
{"current_steps": 720, "total_steps": 1875, "loss": 0.6773, "lr": 0.00013564118787132506, "epoch": 1.152, "percentage": 38.4, "elapsed_time": "0:14:51", "remaining_time": "0:23:50"}
{"current_steps": 725, "total_steps": 1875, "loss": 0.668, "lr": 0.00013485720473218154, "epoch": 1.16, "percentage": 38.67, "elapsed_time": "0:14:59", "remaining_time": "0:23:46"}
{"current_steps": 730, "total_steps": 1875, "loss": 0.6997, "lr": 0.0001340707751943952, "epoch": 1.168, "percentage": 38.93, "elapsed_time": "0:15:06", "remaining_time": "0:23:41"}
{"current_steps": 735, "total_steps": 1875, "loss": 0.831, "lr": 0.00013328195445229868, "epoch": 1.176, "percentage": 39.2, "elapsed_time": "0:15:15", "remaining_time": "0:23:39"}
{"current_steps": 740, "total_steps": 1875, "loss": 0.6378, "lr": 0.00013249079786804765, "epoch": 1.184, "percentage": 39.47, "elapsed_time": "0:15:21", "remaining_time": "0:23:33"}
{"current_steps": 745, "total_steps": 1875, "loss": 0.6547, "lr": 0.0001316973609677352, "epoch": 1.192, "percentage": 39.73, "elapsed_time": "0:15:26", "remaining_time": "0:23:25"}
{"current_steps": 750, "total_steps": 1875, "loss": 0.5808, "lr": 0.00013090169943749476, "epoch": 1.2, "percentage": 40.0, "elapsed_time": "0:15:31", "remaining_time": "0:23:17"}
{"current_steps": 755, "total_steps": 1875, "loss": 0.5582, "lr": 0.00013010386911959206, "epoch": 1.208, "percentage": 40.27, "elapsed_time": "0:15:37", "remaining_time": "0:23:10"}
{"current_steps": 760, "total_steps": 1875, "loss": 0.5801, "lr": 0.00012930392600850573, "epoch": 1.216, "percentage": 40.53, "elapsed_time": "0:15:44", "remaining_time": "0:23:05"}
{"current_steps": 765, "total_steps": 1875, "loss": 0.65, "lr": 0.0001285019262469976, "epoch": 1.224, "percentage": 40.8, "elapsed_time": "0:15:50", "remaining_time": "0:22:59"}
{"current_steps": 770, "total_steps": 1875, "loss": 0.6627, "lr": 0.00012769792612217224, "epoch": 1.232, "percentage": 41.07, "elapsed_time": "0:15:57", "remaining_time": "0:22:53"}
{"current_steps": 775, "total_steps": 1875, "loss": 0.5603, "lr": 0.00012689198206152657, "epoch": 1.24, "percentage": 41.33, "elapsed_time": "0:16:03", "remaining_time": "0:22:46"}
{"current_steps": 780, "total_steps": 1875, "loss": 0.6525, "lr": 0.00012608415062898972, "epoch": 1.248, "percentage": 41.6, "elapsed_time": "0:16:08", "remaining_time": "0:22:38"}
{"current_steps": 785, "total_steps": 1875, "loss": 0.6731, "lr": 0.00012527448852095295, "epoch": 1.256, "percentage": 41.87, "elapsed_time": "0:16:14", "remaining_time": "0:22:33"}
{"current_steps": 790, "total_steps": 1875, "loss": 0.6255, "lr": 0.00012446305256229073, "epoch": 1.264, "percentage": 42.13, "elapsed_time": "0:16:21", "remaining_time": "0:22:27"}
{"current_steps": 795, "total_steps": 1875, "loss": 0.6585, "lr": 0.00012364989970237248, "epoch": 1.272, "percentage": 42.4, "elapsed_time": "0:16:28", "remaining_time": "0:22:22"}
{"current_steps": 800, "total_steps": 1875, "loss": 0.5996, "lr": 0.00012283508701106557, "epoch": 1.28, "percentage": 42.67, "elapsed_time": "0:16:35", "remaining_time": "0:22:17"}
{"current_steps": 805, "total_steps": 1875, "loss": 0.6355, "lr": 0.00012201867167473015, "epoch": 1.288, "percentage": 42.93, "elapsed_time": "0:16:44", "remaining_time": "0:22:15"}
{"current_steps": 810, "total_steps": 1875, "loss": 0.6615, "lr": 0.00012120071099220549, "epoch": 1.296, "percentage": 43.2, "elapsed_time": "0:16:50", "remaining_time": "0:22:08"}
{"current_steps": 815, "total_steps": 1875, "loss": 0.6096, "lr": 0.0001203812623707885, "epoch": 1.304, "percentage": 43.47, "elapsed_time": "0:16:58", "remaining_time": "0:22:04"}
{"current_steps": 820, "total_steps": 1875, "loss": 0.5984, "lr": 0.00011956038332220483, "epoch": 1.312, "percentage": 43.73, "elapsed_time": "0:17:05", "remaining_time": "0:22:00"}
{"current_steps": 825, "total_steps": 1875, "loss": 0.5569, "lr": 0.00011873813145857249, "epoch": 1.32, "percentage": 44.0, "elapsed_time": "0:17:14", "remaining_time": "0:21:56"}
{"current_steps": 830, "total_steps": 1875, "loss": 0.7088, "lr": 0.00011791456448835825, "epoch": 1.328, "percentage": 44.27, "elapsed_time": "0:17:20", "remaining_time": "0:21:50"}
{"current_steps": 835, "total_steps": 1875, "loss": 0.6731, "lr": 0.00011708974021232769, "epoch": 1.336, "percentage": 44.53, "elapsed_time": "0:17:26", "remaining_time": "0:21:43"}
{"current_steps": 840, "total_steps": 1875, "loss": 0.6188, "lr": 0.00011626371651948838, "epoch": 1.3439999999999999, "percentage": 44.8, "elapsed_time": "0:17:32", "remaining_time": "0:21:36"}
{"current_steps": 845, "total_steps": 1875, "loss": 0.7004, "lr": 0.00011543655138302714, "epoch": 1.3519999999999999, "percentage": 45.07, "elapsed_time": "0:17:37", "remaining_time": "0:21:29"}
{"current_steps": 850, "total_steps": 1875, "loss": 0.5884, "lr": 0.00011460830285624118, "epoch": 1.3599999999999999, "percentage": 45.33, "elapsed_time": "0:17:45", "remaining_time": "0:21:25"}
{"current_steps": 855, "total_steps": 1875, "loss": 0.5739, "lr": 0.0001137790290684638, "epoch": 1.3679999999999999, "percentage": 45.6, "elapsed_time": "0:17:54", "remaining_time": "0:21:22"}
{"current_steps": 860, "total_steps": 1875, "loss": 0.6435, "lr": 0.00011294878822098469, "epoch": 1.376, "percentage": 45.87, "elapsed_time": "0:17:59", "remaining_time": "0:21:14"}
{"current_steps": 865, "total_steps": 1875, "loss": 0.6897, "lr": 0.00011211763858296507, "epoch": 1.384, "percentage": 46.13, "elapsed_time": "0:18:06", "remaining_time": "0:21:08"}
{"current_steps": 870, "total_steps": 1875, "loss": 0.6641, "lr": 0.00011128563848734816, "epoch": 1.392, "percentage": 46.4, "elapsed_time": "0:18:10", "remaining_time": "0:21:00"}
{"current_steps": 875, "total_steps": 1875, "loss": 0.6273, "lr": 0.00011045284632676536, "epoch": 1.4, "percentage": 46.67, "elapsed_time": "0:18:16", "remaining_time": "0:20:53"}
{"current_steps": 880, "total_steps": 1875, "loss": 0.6437, "lr": 0.00010961932054943778, "epoch": 1.408, "percentage": 46.93, "elapsed_time": "0:18:24", "remaining_time": "0:20:49"}
{"current_steps": 885, "total_steps": 1875, "loss": 0.6345, "lr": 0.00010878511965507434, "epoch": 1.416, "percentage": 47.2, "elapsed_time": "0:18:31", "remaining_time": "0:20:43"}
{"current_steps": 890, "total_steps": 1875, "loss": 0.5913, "lr": 0.00010795030219076599, "epoch": 1.424, "percentage": 47.47, "elapsed_time": "0:18:39", "remaining_time": "0:20:39"}
{"current_steps": 895, "total_steps": 1875, "loss": 0.6482, "lr": 0.00010711492674687671, "epoch": 1.432, "percentage": 47.73, "elapsed_time": "0:18:45", "remaining_time": "0:20:32"}
{"current_steps": 900, "total_steps": 1875, "loss": 0.6165, "lr": 0.00010627905195293135, "epoch": 1.44, "percentage": 48.0, "elapsed_time": "0:18:52", "remaining_time": "0:20:26"}
{"current_steps": 905, "total_steps": 1875, "loss": 0.634, "lr": 0.00010544273647350092, "epoch": 1.448, "percentage": 48.27, "elapsed_time": "0:18:59", "remaining_time": "0:20:21"}
{"current_steps": 910, "total_steps": 1875, "loss": 0.6509, "lr": 0.00010460603900408523, "epoch": 1.456, "percentage": 48.53, "elapsed_time": "0:19:04", "remaining_time": "0:20:13"}
{"current_steps": 915, "total_steps": 1875, "loss": 0.6212, "lr": 0.00010376901826699348, "epoch": 1.464, "percentage": 48.8, "elapsed_time": "0:19:11", "remaining_time": "0:20:07"}
{"current_steps": 920, "total_steps": 1875, "loss": 0.7305, "lr": 0.00010293173300722285, "epoch": 1.472, "percentage": 49.07, "elapsed_time": "0:19:17", "remaining_time": "0:20:01"}
{"current_steps": 925, "total_steps": 1875, "loss": 0.6685, "lr": 0.0001020942419883357, "epoch": 1.48, "percentage": 49.33, "elapsed_time": "0:19:23", "remaining_time": "0:19:54"}
{"current_steps": 930, "total_steps": 1875, "loss": 0.6214, "lr": 0.00010125660398833528, "epoch": 1.488, "percentage": 49.6, "elapsed_time": "0:19:29", "remaining_time": "0:19:48"}
{"current_steps": 935, "total_steps": 1875, "loss": 0.6035, "lr": 0.0001004188777955404, "epoch": 1.496, "percentage": 49.87, "elapsed_time": "0:19:37", "remaining_time": "0:19:43"}
{"current_steps": 940, "total_steps": 1875, "loss": 0.5868, "lr": 9.958112220445963e-05, "epoch": 1.504, "percentage": 50.13, "elapsed_time": "0:19:44", "remaining_time": "0:19:37"}
{"current_steps": 945, "total_steps": 1875, "loss": 0.6003, "lr": 9.874339601166473e-05, "epoch": 1.512, "percentage": 50.4, "elapsed_time": "0:19:50", "remaining_time": "0:19:31"}
{"current_steps": 950, "total_steps": 1875, "loss": 0.5854, "lr": 9.790575801166432e-05, "epoch": 1.52, "percentage": 50.67, "elapsed_time": "0:19:55", "remaining_time": "0:19:24"}
{"current_steps": 955, "total_steps": 1875, "loss": 0.5882, "lr": 9.706826699277718e-05, "epoch": 1.528, "percentage": 50.93, "elapsed_time": "0:20:03", "remaining_time": "0:19:19"}
{"current_steps": 960, "total_steps": 1875, "loss": 0.7187, "lr": 9.623098173300654e-05, "epoch": 1.536, "percentage": 51.2, "elapsed_time": "0:20:08", "remaining_time": "0:19:11"}
{"current_steps": 965, "total_steps": 1875, "loss": 0.6156, "lr": 9.539396099591476e-05, "epoch": 1.544, "percentage": 51.47, "elapsed_time": "0:20:13", "remaining_time": "0:19:04"}
{"current_steps": 970, "total_steps": 1875, "loss": 0.6488, "lr": 9.455726352649911e-05, "epoch": 1.552, "percentage": 51.73, "elapsed_time": "0:20:17", "remaining_time": "0:18:56"}
{"current_steps": 975, "total_steps": 1875, "loss": 0.6601, "lr": 9.372094804706867e-05, "epoch": 1.56, "percentage": 52.0, "elapsed_time": "0:20:24", "remaining_time": "0:18:50"}
{"current_steps": 980, "total_steps": 1875, "loss": 0.5968, "lr": 9.288507325312335e-05, "epoch": 1.568, "percentage": 52.27, "elapsed_time": "0:20:31", "remaining_time": "0:18:44"}
{"current_steps": 985, "total_steps": 1875, "loss": 0.7034, "lr": 9.204969780923403e-05, "epoch": 1.576, "percentage": 52.53, "elapsed_time": "0:20:36", "remaining_time": "0:18:37"}
{"current_steps": 990, "total_steps": 1875, "loss": 0.5973, "lr": 9.121488034492569e-05, "epoch": 1.584, "percentage": 52.8, "elapsed_time": "0:20:41", "remaining_time": "0:18:29"}
{"current_steps": 995, "total_steps": 1875, "loss": 0.7877, "lr": 9.038067945056227e-05, "epoch": 1.592, "percentage": 53.07, "elapsed_time": "0:20:45", "remaining_time": "0:18:21"}
{"current_steps": 1000, "total_steps": 1875, "loss": 0.644, "lr": 8.954715367323468e-05, "epoch": 1.6, "percentage": 53.33, "elapsed_time": "0:20:51", "remaining_time": "0:18:15"}
{"current_steps": 1005, "total_steps": 1875, "loss": 0.6678, "lr": 8.871436151265184e-05, "epoch": 1.608, "percentage": 53.6, "elapsed_time": "0:20:58", "remaining_time": "0:18:09"}
{"current_steps": 1010, "total_steps": 1875, "loss": 0.6088, "lr": 8.788236141703498e-05, "epoch": 1.616, "percentage": 53.87, "elapsed_time": "0:21:03", "remaining_time": "0:18:02"}
{"current_steps": 1015, "total_steps": 1875, "loss": 0.6219, "lr": 8.705121177901532e-05, "epoch": 1.624, "percentage": 54.13, "elapsed_time": "0:21:10", "remaining_time": "0:17:56"}
{"current_steps": 1020, "total_steps": 1875, "loss": 0.6698, "lr": 8.62209709315362e-05, "epoch": 1.6320000000000001, "percentage": 54.4, "elapsed_time": "0:21:16", "remaining_time": "0:17:49"}
{"current_steps": 1025, "total_steps": 1875, "loss": 0.6207, "lr": 8.539169714375885e-05, "epoch": 1.6400000000000001, "percentage": 54.67, "elapsed_time": "0:21:23", "remaining_time": "0:17:44"}
{"current_steps": 1030, "total_steps": 1875, "loss": 0.626, "lr": 8.456344861697289e-05, "epoch": 1.6480000000000001, "percentage": 54.93, "elapsed_time": "0:21:30", "remaining_time": "0:17:39"}
{"current_steps": 1035, "total_steps": 1875, "loss": 0.6972, "lr": 8.373628348051165e-05, "epoch": 1.6560000000000001, "percentage": 55.2, "elapsed_time": "0:21:36", "remaining_time": "0:17:32"}
{"current_steps": 1040, "total_steps": 1875, "loss": 0.6282, "lr": 8.291025978767235e-05, "epoch": 1.6640000000000001, "percentage": 55.47, "elapsed_time": "0:21:41", "remaining_time": "0:17:25"}
{"current_steps": 1045, "total_steps": 1875, "loss": 0.6219, "lr": 8.208543551164178e-05, "epoch": 1.6720000000000002, "percentage": 55.73, "elapsed_time": "0:21:47", "remaining_time": "0:17:18"}
{"current_steps": 1050, "total_steps": 1875, "loss": 0.622, "lr": 8.126186854142752e-05, "epoch": 1.6800000000000002, "percentage": 56.0, "elapsed_time": "0:21:54", "remaining_time": "0:17:12"}
{"current_steps": 1055, "total_steps": 1875, "loss": 0.5801, "lr": 8.04396166777952e-05, "epoch": 1.688, "percentage": 56.27, "elapsed_time": "0:22:01", "remaining_time": "0:17:06"}
{"current_steps": 1060, "total_steps": 1875, "loss": 0.598, "lr": 7.961873762921153e-05, "epoch": 1.696, "percentage": 56.53, "elapsed_time": "0:22:07", "remaining_time": "0:17:00"}
{"current_steps": 1065, "total_steps": 1875, "loss": 0.699, "lr": 7.879928900779456e-05, "epoch": 1.704, "percentage": 56.8, "elapsed_time": "0:22:12", "remaining_time": "0:16:53"}
{"current_steps": 1070, "total_steps": 1875, "loss": 0.5882, "lr": 7.798132832526986e-05, "epoch": 1.712, "percentage": 57.07, "elapsed_time": "0:22:17", "remaining_time": "0:16:46"}
{"current_steps": 1075, "total_steps": 1875, "loss": 0.5321, "lr": 7.716491298893442e-05, "epoch": 1.72, "percentage": 57.33, "elapsed_time": "0:22:25", "remaining_time": "0:16:41"}
{"current_steps": 1080, "total_steps": 1875, "loss": 0.6647, "lr": 7.635010029762756e-05, "epoch": 1.728, "percentage": 57.6, "elapsed_time": "0:22:33", "remaining_time": "0:16:36"}
{"current_steps": 1085, "total_steps": 1875, "loss": 0.628, "lr": 7.553694743770928e-05, "epoch": 1.736, "percentage": 57.87, "elapsed_time": "0:22:40", "remaining_time": "0:16:30"}
{"current_steps": 1090, "total_steps": 1875, "loss": 0.6262, "lr": 7.472551147904708e-05, "epoch": 1.744, "percentage": 58.13, "elapsed_time": "0:22:46", "remaining_time": "0:16:24"}
{"current_steps": 1095, "total_steps": 1875, "loss": 0.6131, "lr": 7.391584937101033e-05, "epoch": 1.752, "percentage": 58.4, "elapsed_time": "0:22:53", "remaining_time": "0:16:18"}
{"current_steps": 1100, "total_steps": 1875, "loss": 0.6494, "lr": 7.310801793847344e-05, "epoch": 1.76, "percentage": 58.67, "elapsed_time": "0:22:58", "remaining_time": "0:16:11"}
{"current_steps": 1105, "total_steps": 1875, "loss": 0.5514, "lr": 7.230207387782776e-05, "epoch": 1.768, "percentage": 58.93, "elapsed_time": "0:23:06", "remaining_time": "0:16:06"}
{"current_steps": 1110, "total_steps": 1875, "loss": 0.5823, "lr": 7.149807375300239e-05, "epoch": 1.776, "percentage": 59.2, "elapsed_time": "0:23:14", "remaining_time": "0:16:01"}
{"current_steps": 1115, "total_steps": 1875, "loss": 0.7207, "lr": 7.069607399149428e-05, "epoch": 1.784, "percentage": 59.47, "elapsed_time": "0:23:20", "remaining_time": "0:15:54"}
{"current_steps": 1120, "total_steps": 1875, "loss": 0.7006, "lr": 6.989613088040796e-05, "epoch": 1.792, "percentage": 59.73, "elapsed_time": "0:23:26", "remaining_time": "0:15:48"}
{"current_steps": 1125, "total_steps": 1875, "loss": 0.6222, "lr": 6.909830056250527e-05, "epoch": 1.8, "percentage": 60.0, "elapsed_time": "0:23:32", "remaining_time": "0:15:41"}
{"current_steps": 1130, "total_steps": 1875, "loss": 0.6569, "lr": 6.830263903226483e-05, "epoch": 1.808, "percentage": 60.27, "elapsed_time": "0:23:37", "remaining_time": "0:15:34"}
{"current_steps": 1135, "total_steps": 1875, "loss": 0.543, "lr": 6.750920213195238e-05, "epoch": 1.8159999999999998, "percentage": 60.53, "elapsed_time": "0:23:45", "remaining_time": "0:15:29"}
{"current_steps": 1140, "total_steps": 1875, "loss": 0.6334, "lr": 6.671804554770135e-05, "epoch": 1.8239999999999998, "percentage": 60.8, "elapsed_time": "0:23:51", "remaining_time": "0:15:23"}
{"current_steps": 1145, "total_steps": 1875, "loss": 0.6701, "lr": 6.592922480560483e-05, "epoch": 1.8319999999999999, "percentage": 61.07, "elapsed_time": "0:24:00", "remaining_time": "0:15:18"}
{"current_steps": 1150, "total_steps": 1875, "loss": 0.6216, "lr": 6.51427952678185e-05, "epoch": 1.8399999999999999, "percentage": 61.33, "elapsed_time": "0:24:07", "remaining_time": "0:15:12"}
{"current_steps": 1155, "total_steps": 1875, "loss": 0.5877, "lr": 6.435881212867493e-05, "epoch": 1.8479999999999999, "percentage": 61.6, "elapsed_time": "0:24:14", "remaining_time": "0:15:06"}
{"current_steps": 1160, "total_steps": 1875, "loss": 0.6256, "lr": 6.357733041081018e-05, "epoch": 1.8559999999999999, "percentage": 61.87, "elapsed_time": "0:24:21", "remaining_time": "0:15:00"}
{"current_steps": 1165, "total_steps": 1875, "loss": 0.7062, "lr": 6.27984049613019e-05, "epoch": 1.8639999999999999, "percentage": 62.13, "elapsed_time": "0:24:26", "remaining_time": "0:14:53"}
{"current_steps": 1170, "total_steps": 1875, "loss": 0.6304, "lr": 6.20220904478199e-05, "epoch": 1.8719999999999999, "percentage": 62.4, "elapsed_time": "0:24:32", "remaining_time": "0:14:47"}
{"current_steps": 1175, "total_steps": 1875, "loss": 0.7695, "lr": 6.12484413547897e-05, "epoch": 1.88, "percentage": 62.67, "elapsed_time": "0:24:36", "remaining_time": "0:14:39"}
{"current_steps": 1180, "total_steps": 1875, "loss": 0.5723, "lr": 6.047751197956838e-05, "epoch": 1.888, "percentage": 62.93, "elapsed_time": "0:24:43", "remaining_time": "0:14:33"}
{"current_steps": 1185, "total_steps": 1875, "loss": 0.6847, "lr": 5.9709356428633746e-05, "epoch": 1.896, "percentage": 63.2, "elapsed_time": "0:24:48", "remaining_time": "0:14:26"}
{"current_steps": 1190, "total_steps": 1875, "loss": 0.6618, "lr": 5.8944028613787206e-05, "epoch": 1.904, "percentage": 63.47, "elapsed_time": "0:24:55", "remaining_time": "0:14:20"}
{"current_steps": 1195, "total_steps": 1875, "loss": 0.6275, "lr": 5.818158224836987e-05, "epoch": 1.912, "percentage": 63.73, "elapsed_time": "0:25:01", "remaining_time": "0:14:14"}
{"current_steps": 1200, "total_steps": 1875, "loss": 0.5617, "lr": 5.7422070843492734e-05, "epoch": 1.92, "percentage": 64.0, "elapsed_time": "0:25:09", "remaining_time": "0:14:08"}
{"current_steps": 1205, "total_steps": 1875, "loss": 0.6278, "lr": 5.666554770428129e-05, "epoch": 1.928, "percentage": 64.27, "elapsed_time": "0:25:15", "remaining_time": "0:14:02"}
{"current_steps": 1210, "total_steps": 1875, "loss": 0.6713, "lr": 5.591206592613416e-05, "epoch": 1.936, "percentage": 64.53, "elapsed_time": "0:25:20", "remaining_time": "0:13:55"}
{"current_steps": 1215, "total_steps": 1875, "loss": 0.6113, "lr": 5.5161678390996796e-05, "epoch": 1.944, "percentage": 64.8, "elapsed_time": "0:25:25", "remaining_time": "0:13:48"}
{"current_steps": 1220, "total_steps": 1875, "loss": 0.5587, "lr": 5.441443776365003e-05, "epoch": 1.952, "percentage": 65.07, "elapsed_time": "0:25:31", "remaining_time": "0:13:42"}
{"current_steps": 1225, "total_steps": 1875, "loss": 0.5601, "lr": 5.3670396488013854e-05, "epoch": 1.96, "percentage": 65.33, "elapsed_time": "0:25:38", "remaining_time": "0:13:36"}
{"current_steps": 1230, "total_steps": 1875, "loss": 0.5941, "lr": 5.292960678346675e-05, "epoch": 1.968, "percentage": 65.6, "elapsed_time": "0:25:43", "remaining_time": "0:13:29"}
{"current_steps": 1235, "total_steps": 1875, "loss": 0.6285, "lr": 5.2192120641180786e-05, "epoch": 1.976, "percentage": 65.87, "elapsed_time": "0:25:49", "remaining_time": "0:13:22"}
{"current_steps": 1240, "total_steps": 1875, "loss": 0.6516, "lr": 5.145798982047261e-05, "epoch": 1.984, "percentage": 66.13, "elapsed_time": "0:25:55", "remaining_time": "0:13:16"}
{"current_steps": 1245, "total_steps": 1875, "loss": 0.5904, "lr": 5.072726584517086e-05, "epoch": 1.992, "percentage": 66.4, "elapsed_time": "0:26:05", "remaining_time": "0:13:12"}
{"current_steps": 1250, "total_steps": 1875, "loss": 0.619, "lr": 5.000000000000002e-05, "epoch": 2.0, "percentage": 66.67, "elapsed_time": "0:26:11", "remaining_time": "0:13:05"}
{"current_steps": 1255, "total_steps": 1875, "loss": 0.6058, "lr": 4.927624332698109e-05, "epoch": 2.008, "percentage": 66.93, "elapsed_time": "0:26:15", "remaining_time": "0:12:58"}
{"current_steps": 1260, "total_steps": 1875, "loss": 0.6248, "lr": 4.8556046621849346e-05, "epoch": 2.016, "percentage": 67.2, "elapsed_time": "0:26:20", "remaining_time": "0:12:51"}
{"current_steps": 1265, "total_steps": 1875, "loss": 0.5247, "lr": 4.783946043048923e-05, "epoch": 2.024, "percentage": 67.47, "elapsed_time": "0:26:27", "remaining_time": "0:12:45"}
{"current_steps": 1270, "total_steps": 1875, "loss": 0.5439, "lr": 4.712653504538683e-05, "epoch": 2.032, "percentage": 67.73, "elapsed_time": "0:26:34", "remaining_time": "0:12:39"}
{"current_steps": 1275, "total_steps": 1875, "loss": 0.4491, "lr": 4.6417320502100316e-05, "epoch": 2.04, "percentage": 68.0, "elapsed_time": "0:26:40", "remaining_time": "0:12:33"}
{"current_steps": 1280, "total_steps": 1875, "loss": 0.52, "lr": 4.5711866575748276e-05, "epoch": 2.048, "percentage": 68.27, "elapsed_time": "0:26:48", "remaining_time": "0:12:27"}
{"current_steps": 1285, "total_steps": 1875, "loss": 0.5259, "lr": 4.501022277751602e-05, "epoch": 2.056, "percentage": 68.53, "elapsed_time": "0:26:55", "remaining_time": "0:12:21"}
{"current_steps": 1290, "total_steps": 1875, "loss": 0.5025, "lr": 4.431243835118124e-05, "epoch": 2.064, "percentage": 68.8, "elapsed_time": "0:27:02", "remaining_time": "0:12:15"}
{"current_steps": 1295, "total_steps": 1875, "loss": 0.4772, "lr": 4.361856226965733e-05, "epoch": 2.072, "percentage": 69.07, "elapsed_time": "0:27:08", "remaining_time": "0:12:09"}
{"current_steps": 1300, "total_steps": 1875, "loss": 0.5945, "lr": 4.2928643231556844e-05, "epoch": 2.08, "percentage": 69.33, "elapsed_time": "0:27:14", "remaining_time": "0:12:02"}
{"current_steps": 1305, "total_steps": 1875, "loss": 0.4813, "lr": 4.224272965777326e-05, "epoch": 2.088, "percentage": 69.6, "elapsed_time": "0:27:22", "remaining_time": "0:11:57"}
{"current_steps": 1310, "total_steps": 1875, "loss": 0.5315, "lr": 4.15608696880828e-05, "epoch": 2.096, "percentage": 69.87, "elapsed_time": "0:27:28", "remaining_time": "0:11:51"}
{"current_steps": 1315, "total_steps": 1875, "loss": 0.5591, "lr": 4.08831111777658e-05, "epoch": 2.104, "percentage": 70.13, "elapsed_time": "0:27:34", "remaining_time": "0:11:44"}
{"current_steps": 1320, "total_steps": 1875, "loss": 0.605, "lr": 4.020950169424815e-05, "epoch": 2.112, "percentage": 70.4, "elapsed_time": "0:27:40", "remaining_time": "0:11:37"}
{"current_steps": 1325, "total_steps": 1875, "loss": 0.4955, "lr": 3.954008851376252e-05, "epoch": 2.12, "percentage": 70.67, "elapsed_time": "0:27:47", "remaining_time": "0:11:32"}
{"current_steps": 1330, "total_steps": 1875, "loss": 0.5757, "lr": 3.887491861803085e-05, "epoch": 2.128, "percentage": 70.93, "elapsed_time": "0:27:52", "remaining_time": "0:11:25"}
{"current_steps": 1335, "total_steps": 1875, "loss": 0.5313, "lr": 3.821403869096658e-05, "epoch": 2.136, "percentage": 71.2, "elapsed_time": "0:27:59", "remaining_time": "0:11:19"}
{"current_steps": 1340, "total_steps": 1875, "loss": 0.5904, "lr": 3.755749511539845e-05, "epoch": 2.144, "percentage": 71.47, "elapsed_time": "0:28:06", "remaining_time": "0:11:13"}
{"current_steps": 1345, "total_steps": 1875, "loss": 0.4679, "lr": 3.690533396981504e-05, "epoch": 2.152, "percentage": 71.73, "elapsed_time": "0:28:12", "remaining_time": "0:11:06"}
{"current_steps": 1350, "total_steps": 1875, "loss": 0.5235, "lr": 3.6257601025131026e-05, "epoch": 2.16, "percentage": 72.0, "elapsed_time": "0:28:17", "remaining_time": "0:10:59"}
{"current_steps": 1355, "total_steps": 1875, "loss": 0.5797, "lr": 3.561434174147463e-05, "epoch": 2.168, "percentage": 72.27, "elapsed_time": "0:28:21", "remaining_time": "0:10:52"}
{"current_steps": 1360, "total_steps": 1875, "loss": 0.5772, "lr": 3.497560126499709e-05, "epoch": 2.176, "percentage": 72.53, "elapsed_time": "0:28:26", "remaining_time": "0:10:46"}
{"current_steps": 1365, "total_steps": 1875, "loss": 0.5316, "lr": 3.4341424424704375e-05, "epoch": 2.184, "percentage": 72.8, "elapsed_time": "0:28:32", "remaining_time": "0:10:39"}
{"current_steps": 1370, "total_steps": 1875, "loss": 0.5646, "lr": 3.371185572931048e-05, "epoch": 2.192, "percentage": 73.07, "elapsed_time": "0:28:38", "remaining_time": "0:10:33"}
{"current_steps": 1375, "total_steps": 1875, "loss": 0.5431, "lr": 3.308693936411421e-05, "epoch": 2.2, "percentage": 73.33, "elapsed_time": "0:28:43", "remaining_time": "0:10:26"}
{"current_steps": 1380, "total_steps": 1875, "loss": 0.5403, "lr": 3.246671918789755e-05, "epoch": 2.208, "percentage": 73.6, "elapsed_time": "0:28:48", "remaining_time": "0:10:20"}
{"current_steps": 1385, "total_steps": 1875, "loss": 0.5329, "lr": 3.1851238729848034e-05, "epoch": 2.216, "percentage": 73.87, "elapsed_time": "0:28:53", "remaining_time": "0:10:13"}
{"current_steps": 1390, "total_steps": 1875, "loss": 0.5696, "lr": 3.124054118650327e-05, "epoch": 2.224, "percentage": 74.13, "elapsed_time": "0:29:01", "remaining_time": "0:10:07"}
{"current_steps": 1395, "total_steps": 1875, "loss": 0.59, "lr": 3.063466941871952e-05, "epoch": 2.232, "percentage": 74.4, "elapsed_time": "0:29:07", "remaining_time": "0:10:01"}
{"current_steps": 1400, "total_steps": 1875, "loss": 0.5116, "lr": 3.0033665948663448e-05, "epoch": 2.24, "percentage": 74.67, "elapsed_time": "0:29:14", "remaining_time": "0:09:55"}
{"current_steps": 1405, "total_steps": 1875, "loss": 0.5783, "lr": 2.9437572956827964e-05, "epoch": 2.248, "percentage": 74.93, "elapsed_time": "0:29:21", "remaining_time": "0:09:49"}
{"current_steps": 1410, "total_steps": 1875, "loss": 0.6259, "lr": 2.8846432279071467e-05, "epoch": 2.2560000000000002, "percentage": 75.2, "elapsed_time": "0:29:27", "remaining_time": "0:09:42"}
{"current_steps": 1415, "total_steps": 1875, "loss": 0.5759, "lr": 2.826028540368215e-05, "epoch": 2.2640000000000002, "percentage": 75.47, "elapsed_time": "0:29:35", "remaining_time": "0:09:37"}
{"current_steps": 1420, "total_steps": 1875, "loss": 0.497, "lr": 2.7679173468465812e-05, "epoch": 2.2720000000000002, "percentage": 75.73, "elapsed_time": "0:29:41", "remaining_time": "0:09:30"}
{"current_steps": 1425, "total_steps": 1875, "loss": 0.619, "lr": 2.7103137257858868e-05, "epoch": 2.2800000000000002, "percentage": 76.0, "elapsed_time": "0:29:47", "remaining_time": "0:09:24"}
{"current_steps": 1430, "total_steps": 1875, "loss": 0.5858, "lr": 2.6532217200065858e-05, "epoch": 2.288, "percentage": 76.27, "elapsed_time": "0:29:52", "remaining_time": "0:09:17"}
{"current_steps": 1435, "total_steps": 1875, "loss": 0.6291, "lr": 2.5966453364222186e-05, "epoch": 2.296, "percentage": 76.53, "elapsed_time": "0:29:58", "remaining_time": "0:09:11"}
{"current_steps": 1440, "total_steps": 1875, "loss": 0.6418, "lr": 2.540588545758179e-05, "epoch": 2.304, "percentage": 76.8, "elapsed_time": "0:30:04", "remaining_time": "0:09:05"}
{"current_steps": 1445, "total_steps": 1875, "loss": 0.5483, "lr": 2.48505528227304e-05, "epoch": 2.312, "percentage": 77.07, "elapsed_time": "0:30:11", "remaining_time": "0:08:59"}
{"current_steps": 1450, "total_steps": 1875, "loss": 0.6071, "lr": 2.4300494434824373e-05, "epoch": 2.32, "percentage": 77.33, "elapsed_time": "0:30:18", "remaining_time": "0:08:53"}
{"current_steps": 1455, "total_steps": 1875, "loss": 0.5099, "lr": 2.37557488988552e-05, "epoch": 2.328, "percentage": 77.6, "elapsed_time": "0:30:25", "remaining_time": "0:08:47"}
{"current_steps": 1460, "total_steps": 1875, "loss": 0.5186, "lr": 2.321635444694028e-05, "epoch": 2.336, "percentage": 77.87, "elapsed_time": "0:30:30", "remaining_time": "0:08:40"}
{"current_steps": 1465, "total_steps": 1875, "loss": 0.5043, "lr": 2.2682348935639274e-05, "epoch": 2.344, "percentage": 78.13, "elapsed_time": "0:30:36", "remaining_time": "0:08:33"}
{"current_steps": 1470, "total_steps": 1875, "loss": 0.61, "lr": 2.2153769843297667e-05, "epoch": 2.352, "percentage": 78.4, "elapsed_time": "0:30:42", "remaining_time": "0:08:27"}
{"current_steps": 1475, "total_steps": 1875, "loss": 0.5987, "lr": 2.163065426741603e-05, "epoch": 2.36, "percentage": 78.67, "elapsed_time": "0:30:50", "remaining_time": "0:08:21"}
{"current_steps": 1480, "total_steps": 1875, "loss": 0.5212, "lr": 2.1113038922046602e-05, "epoch": 2.368, "percentage": 78.93, "elapsed_time": "0:30:57", "remaining_time": "0:08:15"}
{"current_steps": 1485, "total_steps": 1875, "loss": 0.4796, "lr": 2.0600960135216462e-05, "epoch": 2.376, "percentage": 79.2, "elapsed_time": "0:31:04", "remaining_time": "0:08:09"}
{"current_steps": 1490, "total_steps": 1875, "loss": 0.4844, "lr": 2.009445384637805e-05, "epoch": 2.384, "percentage": 79.47, "elapsed_time": "0:31:11", "remaining_time": "0:08:03"}
{"current_steps": 1495, "total_steps": 1875, "loss": 0.5085, "lr": 1.9593555603886538e-05, "epoch": 2.392, "percentage": 79.73, "elapsed_time": "0:31:16", "remaining_time": "0:07:57"}
{"current_steps": 1500, "total_steps": 1875, "loss": 0.4839, "lr": 1.9098300562505266e-05, "epoch": 2.4, "percentage": 80.0, "elapsed_time": "0:31:23", "remaining_time": "0:07:50"}
{"current_steps": 1505, "total_steps": 1875, "loss": 0.5715, "lr": 1.8608723480938206e-05, "epoch": 2.408, "percentage": 80.27, "elapsed_time": "0:31:32", "remaining_time": "0:07:45"}
{"current_steps": 1510, "total_steps": 1875, "loss": 0.5266, "lr": 1.812485871939056e-05, "epoch": 2.416, "percentage": 80.53, "elapsed_time": "0:31:36", "remaining_time": "0:07:38"}
{"current_steps": 1515, "total_steps": 1875, "loss": 0.5422, "lr": 1.7646740237157256e-05, "epoch": 2.424, "percentage": 80.8, "elapsed_time": "0:31:43", "remaining_time": "0:07:32"}
{"current_steps": 1520, "total_steps": 1875, "loss": 0.5553, "lr": 1.7174401590239587e-05, "epoch": 2.432, "percentage": 81.07, "elapsed_time": "0:31:49", "remaining_time": "0:07:25"}
{"current_steps": 1525, "total_steps": 1875, "loss": 0.5765, "lr": 1.6707875928990058e-05, "epoch": 2.44, "percentage": 81.33, "elapsed_time": "0:31:57", "remaining_time": "0:07:20"}
{"current_steps": 1530, "total_steps": 1875, "loss": 0.549, "lr": 1.6247195995785837e-05, "epoch": 2.448, "percentage": 81.6, "elapsed_time": "0:32:03", "remaining_time": "0:07:13"}
{"current_steps": 1535, "total_steps": 1875, "loss": 0.4876, "lr": 1.579239412273078e-05, "epoch": 2.456, "percentage": 81.87, "elapsed_time": "0:32:08", "remaining_time": "0:07:07"}
{"current_steps": 1540, "total_steps": 1875, "loss": 0.5546, "lr": 1.5343502229386207e-05, "epoch": 2.464, "percentage": 82.13, "elapsed_time": "0:32:13", "remaining_time": "0:07:00"}
{"current_steps": 1545, "total_steps": 1875, "loss": 0.5356, "lr": 1.4900551820530828e-05, "epoch": 2.472, "percentage": 82.4, "elapsed_time": "0:32:18", "remaining_time": "0:06:54"}
{"current_steps": 1550, "total_steps": 1875, "loss": 0.5142, "lr": 1.4463573983949341e-05, "epoch": 2.48, "percentage": 82.67, "elapsed_time": "0:32:25", "remaining_time": "0:06:47"}
{"current_steps": 1555, "total_steps": 1875, "loss": 0.6054, "lr": 1.40325993882509e-05, "epoch": 2.488, "percentage": 82.93, "elapsed_time": "0:32:29", "remaining_time": "0:06:41"}
{"current_steps": 1560, "total_steps": 1875, "loss": 0.5294, "lr": 1.3607658280716473e-05, "epoch": 2.496, "percentage": 83.2, "elapsed_time": "0:32:35", "remaining_time": "0:06:34"}
{"current_steps": 1565, "total_steps": 1875, "loss": 0.5294, "lr": 1.3188780485176088e-05, "epoch": 2.504, "percentage": 83.47, "elapsed_time": "0:32:40", "remaining_time": "0:06:28"}
{"current_steps": 1570, "total_steps": 1875, "loss": 0.4905, "lr": 1.2775995399915631e-05, "epoch": 2.512, "percentage": 83.73, "elapsed_time": "0:32:47", "remaining_time": "0:06:22"}
{"current_steps": 1575, "total_steps": 1875, "loss": 0.5186, "lr": 1.2369331995613665e-05, "epoch": 2.52, "percentage": 84.0, "elapsed_time": "0:32:54", "remaining_time": "0:06:16"}
{"current_steps": 1580, "total_steps": 1875, "loss": 0.4909, "lr": 1.196881881330798e-05, "epoch": 2.528, "percentage": 84.27, "elapsed_time": "0:33:00", "remaining_time": "0:06:09"}
{"current_steps": 1585, "total_steps": 1875, "loss": 0.5303, "lr": 1.1574483962392767e-05, "epoch": 2.536, "percentage": 84.53, "elapsed_time": "0:33:05", "remaining_time": "0:06:03"}
{"current_steps": 1590, "total_steps": 1875, "loss": 0.5169, "lr": 1.1186355118645554e-05, "epoch": 2.544, "percentage": 84.8, "elapsed_time": "0:33:11", "remaining_time": "0:05:57"}
{"current_steps": 1595, "total_steps": 1875, "loss": 0.5339, "lr": 1.0804459522284926e-05, "epoch": 2.552, "percentage": 85.07, "elapsed_time": "0:33:19", "remaining_time": "0:05:50"}
{"current_steps": 1600, "total_steps": 1875, "loss": 0.5283, "lr": 1.042882397605871e-05, "epoch": 2.56, "percentage": 85.33, "elapsed_time": "0:33:23", "remaining_time": "0:05:44"}
{"current_steps": 1605, "total_steps": 1875, "loss": 0.5576, "lr": 1.0059474843362892e-05, "epoch": 2.568, "percentage": 85.6, "elapsed_time": "0:33:30", "remaining_time": "0:05:38"}
{"current_steps": 1610, "total_steps": 1875, "loss": 0.5136, "lr": 9.696438046391288e-06, "epoch": 2.576, "percentage": 85.87, "elapsed_time": "0:33:36", "remaining_time": "0:05:31"}
{"current_steps": 1615, "total_steps": 1875, "loss": 0.5885, "lr": 9.339739064316233e-06, "epoch": 2.584, "percentage": 86.13, "elapsed_time": "0:33:43", "remaining_time": "0:05:25"}
{"current_steps": 1620, "total_steps": 1875, "loss": 0.5, "lr": 8.989402931500434e-06, "epoch": 2.592, "percentage": 86.4, "elapsed_time": "0:33:51", "remaining_time": "0:05:19"}
{"current_steps": 1625, "total_steps": 1875, "loss": 0.5325, "lr": 8.645454235739903e-06, "epoch": 2.6, "percentage": 86.67, "elapsed_time": "0:33:58", "remaining_time": "0:05:13"}
{"current_steps": 1630, "total_steps": 1875, "loss": 0.5772, "lr": 8.307917116538378e-06, "epoch": 2.608, "percentage": 86.93, "elapsed_time": "0:34:04", "remaining_time": "0:05:07"}
{"current_steps": 1635, "total_steps": 1875, "loss": 0.5736, "lr": 7.976815263412963e-06, "epoch": 2.616, "percentage": 87.2, "elapsed_time": "0:34:11", "remaining_time": "0:05:01"}
{"current_steps": 1640, "total_steps": 1875, "loss": 0.5199, "lr": 7.652171914231776e-06, "epoch": 2.624, "percentage": 87.47, "elapsed_time": "0:34:18", "remaining_time": "0:04:54"}
{"current_steps": 1645, "total_steps": 1875, "loss": 0.5753, "lr": 7.3340098535827905e-06, "epoch": 2.632, "percentage": 87.73, "elapsed_time": "0:34:24", "remaining_time": "0:04:48"}
{"current_steps": 1650, "total_steps": 1875, "loss": 0.5424, "lr": 7.022351411174866e-06, "epoch": 2.64, "percentage": 88.0, "elapsed_time": "0:34:29", "remaining_time": "0:04:42"}
{"current_steps": 1655, "total_steps": 1875, "loss": 0.5555, "lr": 6.717218460270536e-06, "epoch": 2.648, "percentage": 88.27, "elapsed_time": "0:34:37", "remaining_time": "0:04:36"}
{"current_steps": 1660, "total_steps": 1875, "loss": 0.4936, "lr": 6.418632416150927e-06, "epoch": 2.656, "percentage": 88.53, "elapsed_time": "0:34:44", "remaining_time": "0:04:29"}
{"current_steps": 1665, "total_steps": 1875, "loss": 0.6291, "lr": 6.126614234612593e-06, "epoch": 2.664, "percentage": 88.8, "elapsed_time": "0:34:50", "remaining_time": "0:04:23"}
{"current_steps": 1670, "total_steps": 1875, "loss": 0.5197, "lr": 5.8411844104969916e-06, "epoch": 2.672, "percentage": 89.07, "elapsed_time": "0:34:55", "remaining_time": "0:04:17"}
{"current_steps": 1675, "total_steps": 1875, "loss": 0.5398, "lr": 5.562362976251901e-06, "epoch": 2.68, "percentage": 89.33, "elapsed_time": "0:35:02", "remaining_time": "0:04:11"}
{"current_steps": 1680, "total_steps": 1875, "loss": 0.6059, "lr": 5.290169500525577e-06, "epoch": 2.6879999999999997, "percentage": 89.6, "elapsed_time": "0:35:08", "remaining_time": "0:04:04"}
{"current_steps": 1685, "total_steps": 1875, "loss": 0.531, "lr": 5.024623086793323e-06, "epoch": 2.6959999999999997, "percentage": 89.87, "elapsed_time": "0:35:14", "remaining_time": "0:03:58"}
{"current_steps": 1690, "total_steps": 1875, "loss": 0.6054, "lr": 4.765742372016735e-06, "epoch": 2.7039999999999997, "percentage": 90.13, "elapsed_time": "0:35:19", "remaining_time": "0:03:51"}
{"current_steps": 1695, "total_steps": 1875, "loss": 0.5173, "lr": 4.513545525335705e-06, "epoch": 2.7119999999999997, "percentage": 90.4, "elapsed_time": "0:35:25", "remaining_time": "0:03:45"}
{"current_steps": 1700, "total_steps": 1875, "loss": 0.4944, "lr": 4.268050246793276e-06, "epoch": 2.7199999999999998, "percentage": 90.67, "elapsed_time": "0:35:31", "remaining_time": "0:03:39"}
{"current_steps": 1705, "total_steps": 1875, "loss": 0.5183, "lr": 4.029273766093333e-06, "epoch": 2.7279999999999998, "percentage": 90.93, "elapsed_time": "0:35:38", "remaining_time": "0:03:33"}
{"current_steps": 1710, "total_steps": 1875, "loss": 0.668, "lr": 3.797232841391407e-06, "epoch": 2.7359999999999998, "percentage": 91.2, "elapsed_time": "0:35:44", "remaining_time": "0:03:26"}
{"current_steps": 1715, "total_steps": 1875, "loss": 0.5068, "lr": 3.5719437581185454e-06, "epoch": 2.7439999999999998, "percentage": 91.47, "elapsed_time": "0:35:50", "remaining_time": "0:03:20"}
{"current_steps": 1720, "total_steps": 1875, "loss": 0.5823, "lr": 3.3534223278382405e-06, "epoch": 2.752, "percentage": 91.73, "elapsed_time": "0:35:57", "remaining_time": "0:03:14"}
{"current_steps": 1725, "total_steps": 1875, "loss": 0.6111, "lr": 3.1416838871368924e-06, "epoch": 2.76, "percentage": 92.0, "elapsed_time": "0:36:02", "remaining_time": "0:03:08"}
{"current_steps": 1730, "total_steps": 1875, "loss": 0.5231, "lr": 2.936743296547273e-06, "epoch": 2.768, "percentage": 92.27, "elapsed_time": "0:36:07", "remaining_time": "0:03:01"}
{"current_steps": 1735, "total_steps": 1875, "loss": 0.5236, "lr": 2.738614939505646e-06, "epoch": 2.776, "percentage": 92.53, "elapsed_time": "0:36:14", "remaining_time": "0:02:55"}
{"current_steps": 1740, "total_steps": 1875, "loss": 0.5657, "lr": 2.5473127213422763e-06, "epoch": 2.784, "percentage": 92.8, "elapsed_time": "0:36:23", "remaining_time": "0:02:49"}
{"current_steps": 1745, "total_steps": 1875, "loss": 0.5518, "lr": 2.3628500683055222e-06, "epoch": 2.792, "percentage": 93.07, "elapsed_time": "0:36:28", "remaining_time": "0:02:43"}
{"current_steps": 1750, "total_steps": 1875, "loss": 0.4908, "lr": 2.1852399266194314e-06, "epoch": 2.8, "percentage": 93.33, "elapsed_time": "0:36:34", "remaining_time": "0:02:36"}
{"current_steps": 1755, "total_steps": 1875, "loss": 0.5459, "lr": 2.014494761575314e-06, "epoch": 2.808, "percentage": 93.6, "elapsed_time": "0:36:43", "remaining_time": "0:02:30"}
{"current_steps": 1760, "total_steps": 1875, "loss": 0.5208, "lr": 1.8506265566567094e-06, "epoch": 2.816, "percentage": 93.87, "elapsed_time": "0:36:50", "remaining_time": "0:02:24"}
{"current_steps": 1765, "total_steps": 1875, "loss": 0.5824, "lr": 1.6936468126984572e-06, "epoch": 2.824, "percentage": 94.13, "elapsed_time": "0:36:56", "remaining_time": "0:02:18"}
{"current_steps": 1770, "total_steps": 1875, "loss": 0.5512, "lr": 1.543566547079467e-06, "epoch": 2.832, "percentage": 94.4, "elapsed_time": "0:37:05", "remaining_time": "0:02:12"}
{"current_steps": 1775, "total_steps": 1875, "loss": 0.6327, "lr": 1.400396292949513e-06, "epoch": 2.84, "percentage": 94.67, "elapsed_time": "0:37:12", "remaining_time": "0:02:05"}
{"current_steps": 1780, "total_steps": 1875, "loss": 0.5292, "lr": 1.26414609848996e-06, "epoch": 2.848, "percentage": 94.93, "elapsed_time": "0:37:20", "remaining_time": "0:01:59"}
{"current_steps": 1785, "total_steps": 1875, "loss": 0.5692, "lr": 1.134825526208605e-06, "epoch": 2.856, "percentage": 95.2, "elapsed_time": "0:37:26", "remaining_time": "0:01:53"}
{"current_steps": 1790, "total_steps": 1875, "loss": 0.5532, "lr": 1.0124436522684243e-06, "epoch": 2.864, "percentage": 95.47, "elapsed_time": "0:37:33", "remaining_time": "0:01:47"}
{"current_steps": 1795, "total_steps": 1875, "loss": 0.5314, "lr": 8.970090658507291e-07, "epoch": 2.872, "percentage": 95.73, "elapsed_time": "0:37:38", "remaining_time": "0:01:40"}
{"current_steps": 1800, "total_steps": 1875, "loss": 0.524, "lr": 7.885298685522235e-07, "epoch": 2.88, "percentage": 96.0, "elapsed_time": "0:37:45", "remaining_time": "0:01:34"}
{"current_steps": 1805, "total_steps": 1875, "loss": 0.5227, "lr": 6.870136738164612e-07, "epoch": 2.888, "percentage": 96.27, "elapsed_time": "0:37:52", "remaining_time": "0:01:28"}
{"current_steps": 1810, "total_steps": 1875, "loss": 0.5993, "lr": 5.924676063995382e-07, "epoch": 2.896, "percentage": 96.53, "elapsed_time": "0:37:58", "remaining_time": "0:01:21"}
{"current_steps": 1815, "total_steps": 1875, "loss": 0.5618, "lr": 5.048983018699827e-07, "epoch": 2.904, "percentage": 96.8, "elapsed_time": "0:38:06", "remaining_time": "0:01:15"}
{"current_steps": 1820, "total_steps": 1875, "loss": 0.5504, "lr": 4.2431190614309335e-07, "epoch": 2.912, "percentage": 97.07, "elapsed_time": "0:38:12", "remaining_time": "0:01:09"}
{"current_steps": 1825, "total_steps": 1875, "loss": 0.5147, "lr": 3.50714075049563e-07, "epoch": 2.92, "percentage": 97.33, "elapsed_time": "0:38:19", "remaining_time": "0:01:02"}
{"current_steps": 1830, "total_steps": 1875, "loss": 0.5564, "lr": 2.841099739386066e-07, "epoch": 2.928, "percentage": 97.6, "elapsed_time": "0:38:26", "remaining_time": "0:00:56"}
{"current_steps": 1835, "total_steps": 1875, "loss": 0.5188, "lr": 2.2450427731534053e-07, "epoch": 2.936, "percentage": 97.87, "elapsed_time": "0:38:34", "remaining_time": "0:00:50"}
{"current_steps": 1840, "total_steps": 1875, "loss": 0.5438, "lr": 1.7190116851280026e-07, "epoch": 2.944, "percentage": 98.13, "elapsed_time": "0:38:39", "remaining_time": "0:00:44"}
{"current_steps": 1845, "total_steps": 1875, "loss": 0.4962, "lr": 1.2630433939825327e-07, "epoch": 2.952, "percentage": 98.4, "elapsed_time": "0:38:45", "remaining_time": "0:00:37"}
{"current_steps": 1850, "total_steps": 1875, "loss": 0.5021, "lr": 8.771699011416168e-08, "epoch": 2.96, "percentage": 98.67, "elapsed_time": "0:38:52", "remaining_time": "0:00:31"}
{"current_steps": 1855, "total_steps": 1875, "loss": 0.5277, "lr": 5.6141828853573106e-08, "epoch": 2.968, "percentage": 98.93, "elapsed_time": "0:38:58", "remaining_time": "0:00:25"}
{"current_steps": 1860, "total_steps": 1875, "loss": 0.5764, "lr": 3.1581071670006015e-08, "epoch": 2.976, "percentage": 99.2, "elapsed_time": "0:39:01", "remaining_time": "0:00:18"}
{"current_steps": 1865, "total_steps": 1875, "loss": 0.5408, "lr": 1.4036442321962995e-08, "epoch": 2.984, "percentage": 99.47, "elapsed_time": "0:39:08", "remaining_time": "0:00:12"}
{"current_steps": 1870, "total_steps": 1875, "loss": 0.5014, "lr": 3.509172151938689e-09, "epoch": 2.992, "percentage": 99.73, "elapsed_time": "0:39:17", "remaining_time": "0:00:06"}
{"current_steps": 1875, "total_steps": 1875, "loss": 0.6182, "lr": 0.0, "epoch": 3.0, "percentage": 100.0, "elapsed_time": "0:39:23", "remaining_time": "0:00:00"}
{"current_steps": 1875, "total_steps": 1875, "epoch": 3.0, "percentage": 100.0, "elapsed_time": "0:39:25", "remaining_time": "0:00:00"}