{ "best_metric": null, "best_model_checkpoint": null, "epoch": 1.999976612016746, "global_step": 42756, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 1.9997661146973525e-05, "loss": 0.5663, "step": 5 }, { "epoch": 0.0, "learning_rate": 1.999532229394705e-05, "loss": 0.2184, "step": 10 }, { "epoch": 0.0, "learning_rate": 1.9992983440920575e-05, "loss": 0.2297, "step": 15 }, { "epoch": 0.0, "learning_rate": 1.9990644587894098e-05, "loss": 0.2685, "step": 20 }, { "epoch": 0.0, "learning_rate": 1.9988305734867625e-05, "loss": 0.2173, "step": 25 }, { "epoch": 0.0, "learning_rate": 1.9985966881841145e-05, "loss": 0.1281, "step": 30 }, { "epoch": 0.0, "learning_rate": 1.998362802881467e-05, "loss": 0.1408, "step": 35 }, { "epoch": 0.0, "learning_rate": 1.9981289175788194e-05, "loss": 0.182, "step": 40 }, { "epoch": 0.0, "learning_rate": 1.9978950322761718e-05, "loss": 0.2657, "step": 45 }, { "epoch": 0.0, "learning_rate": 1.9976611469735244e-05, "loss": 0.1563, "step": 50 }, { "epoch": 0.0, "learning_rate": 1.9974272616708768e-05, "loss": 0.1469, "step": 55 }, { "epoch": 0.0, "learning_rate": 1.997193376368229e-05, "loss": 0.2043, "step": 60 }, { "epoch": 0.0, "learning_rate": 1.9969594910655818e-05, "loss": 0.189, "step": 65 }, { "epoch": 0.0, "learning_rate": 1.996725605762934e-05, "loss": 0.2313, "step": 70 }, { "epoch": 0.0, "learning_rate": 1.9964917204602864e-05, "loss": 0.1986, "step": 75 }, { "epoch": 0.0, "learning_rate": 1.996257835157639e-05, "loss": 0.2241, "step": 80 }, { "epoch": 0.0, "learning_rate": 1.9960239498549914e-05, "loss": 0.2198, "step": 85 }, { "epoch": 0.0, "learning_rate": 1.9957900645523437e-05, "loss": 0.1641, "step": 90 }, { "epoch": 0.0, "learning_rate": 1.995556179249696e-05, "loss": 0.1678, "step": 95 }, { "epoch": 0.0, "learning_rate": 1.9953222939470484e-05, "loss": 0.2344, "step": 100 }, { "epoch": 0.0, "learning_rate": 1.995088408644401e-05, "loss": 0.1122, "step": 105 }, { "epoch": 0.01, "learning_rate": 1.9948545233417534e-05, "loss": 0.1928, "step": 110 }, { "epoch": 0.01, "learning_rate": 1.9946206380391057e-05, "loss": 0.1773, "step": 115 }, { "epoch": 0.01, "learning_rate": 1.9943867527364584e-05, "loss": 0.1878, "step": 120 }, { "epoch": 0.01, "learning_rate": 1.9941528674338107e-05, "loss": 0.166, "step": 125 }, { "epoch": 0.01, "learning_rate": 1.993918982131163e-05, "loss": 0.2206, "step": 130 }, { "epoch": 0.01, "learning_rate": 1.9936850968285157e-05, "loss": 0.105, "step": 135 }, { "epoch": 0.01, "learning_rate": 1.993451211525868e-05, "loss": 0.0942, "step": 140 }, { "epoch": 0.01, "learning_rate": 1.9932173262232203e-05, "loss": 0.2196, "step": 145 }, { "epoch": 0.01, "learning_rate": 1.9929834409205726e-05, "loss": 0.1726, "step": 150 }, { "epoch": 0.01, "learning_rate": 1.992749555617925e-05, "loss": 0.1978, "step": 155 }, { "epoch": 0.01, "learning_rate": 1.9925156703152776e-05, "loss": 0.1458, "step": 160 }, { "epoch": 0.01, "learning_rate": 1.99228178501263e-05, "loss": 0.1802, "step": 165 }, { "epoch": 0.01, "learning_rate": 1.9920478997099823e-05, "loss": 0.1705, "step": 170 }, { "epoch": 0.01, "learning_rate": 1.991814014407335e-05, "loss": 0.2382, "step": 175 }, { "epoch": 0.01, "learning_rate": 1.9915801291046873e-05, "loss": 0.1319, "step": 180 }, { "epoch": 0.01, "learning_rate": 1.9913462438020396e-05, "loss": 0.1764, "step": 185 }, { "epoch": 0.01, "learning_rate": 1.9911123584993923e-05, "loss": 0.089, "step": 190 }, { "epoch": 0.01, "learning_rate": 1.9908784731967446e-05, "loss": 0.1913, "step": 195 }, { "epoch": 0.01, "learning_rate": 1.990644587894097e-05, "loss": 0.1484, "step": 200 }, { "epoch": 0.01, "learning_rate": 1.9904107025914496e-05, "loss": 0.0797, "step": 205 }, { "epoch": 0.01, "learning_rate": 1.9901768172888016e-05, "loss": 0.1276, "step": 210 }, { "epoch": 0.01, "learning_rate": 1.9899429319861542e-05, "loss": 0.1643, "step": 215 }, { "epoch": 0.01, "learning_rate": 1.9897090466835066e-05, "loss": 0.1313, "step": 220 }, { "epoch": 0.01, "learning_rate": 1.989475161380859e-05, "loss": 0.1637, "step": 225 }, { "epoch": 0.01, "learning_rate": 1.9892412760782112e-05, "loss": 0.1306, "step": 230 }, { "epoch": 0.01, "learning_rate": 1.989007390775564e-05, "loss": 0.1551, "step": 235 }, { "epoch": 0.01, "learning_rate": 1.9887735054729162e-05, "loss": 0.2301, "step": 240 }, { "epoch": 0.01, "learning_rate": 1.9885396201702685e-05, "loss": 0.1992, "step": 245 }, { "epoch": 0.01, "learning_rate": 1.9883057348676212e-05, "loss": 0.1415, "step": 250 }, { "epoch": 0.01, "learning_rate": 1.9880718495649735e-05, "loss": 0.1786, "step": 255 }, { "epoch": 0.01, "learning_rate": 1.987837964262326e-05, "loss": 0.1837, "step": 260 }, { "epoch": 0.01, "learning_rate": 1.9876040789596785e-05, "loss": 0.1724, "step": 265 }, { "epoch": 0.01, "learning_rate": 1.9873701936570305e-05, "loss": 0.1758, "step": 270 }, { "epoch": 0.01, "learning_rate": 1.987136308354383e-05, "loss": 0.0815, "step": 275 }, { "epoch": 0.01, "learning_rate": 1.9869024230517355e-05, "loss": 0.1571, "step": 280 }, { "epoch": 0.01, "learning_rate": 1.9866685377490878e-05, "loss": 0.0577, "step": 285 }, { "epoch": 0.01, "learning_rate": 1.9864346524464405e-05, "loss": 0.1624, "step": 290 }, { "epoch": 0.01, "learning_rate": 1.9862007671437928e-05, "loss": 0.1213, "step": 295 }, { "epoch": 0.01, "learning_rate": 1.985966881841145e-05, "loss": 0.0848, "step": 300 }, { "epoch": 0.01, "learning_rate": 1.9857329965384978e-05, "loss": 0.1484, "step": 305 }, { "epoch": 0.01, "learning_rate": 1.98549911123585e-05, "loss": 0.1814, "step": 310 }, { "epoch": 0.01, "learning_rate": 1.9852652259332024e-05, "loss": 0.1326, "step": 315 }, { "epoch": 0.01, "learning_rate": 1.985031340630555e-05, "loss": 0.0791, "step": 320 }, { "epoch": 0.02, "learning_rate": 1.9847974553279074e-05, "loss": 0.2211, "step": 325 }, { "epoch": 0.02, "learning_rate": 1.9845635700252598e-05, "loss": 0.1878, "step": 330 }, { "epoch": 0.02, "learning_rate": 1.984329684722612e-05, "loss": 0.1731, "step": 335 }, { "epoch": 0.02, "learning_rate": 1.9840957994199644e-05, "loss": 0.1982, "step": 340 }, { "epoch": 0.02, "learning_rate": 1.983861914117317e-05, "loss": 0.2487, "step": 345 }, { "epoch": 0.02, "learning_rate": 1.9836280288146694e-05, "loss": 0.2042, "step": 350 }, { "epoch": 0.02, "learning_rate": 1.9833941435120217e-05, "loss": 0.1497, "step": 355 }, { "epoch": 0.02, "learning_rate": 1.9831602582093744e-05, "loss": 0.1681, "step": 360 }, { "epoch": 0.02, "learning_rate": 1.9829263729067267e-05, "loss": 0.1641, "step": 365 }, { "epoch": 0.02, "learning_rate": 1.982692487604079e-05, "loss": 0.1213, "step": 370 }, { "epoch": 0.02, "learning_rate": 1.9824586023014317e-05, "loss": 0.12, "step": 375 }, { "epoch": 0.02, "learning_rate": 1.982224716998784e-05, "loss": 0.1017, "step": 380 }, { "epoch": 0.02, "learning_rate": 1.9819908316961364e-05, "loss": 0.1507, "step": 385 }, { "epoch": 0.02, "learning_rate": 1.9817569463934887e-05, "loss": 0.1323, "step": 390 }, { "epoch": 0.02, "learning_rate": 1.981523061090841e-05, "loss": 0.1741, "step": 395 }, { "epoch": 0.02, "learning_rate": 1.9812891757881937e-05, "loss": 0.1562, "step": 400 }, { "epoch": 0.02, "learning_rate": 1.981055290485546e-05, "loss": 0.1865, "step": 405 }, { "epoch": 0.02, "learning_rate": 1.9808214051828983e-05, "loss": 0.163, "step": 410 }, { "epoch": 0.02, "learning_rate": 1.980587519880251e-05, "loss": 0.1357, "step": 415 }, { "epoch": 0.02, "learning_rate": 1.9803536345776033e-05, "loss": 0.2273, "step": 420 }, { "epoch": 0.02, "learning_rate": 1.9801197492749556e-05, "loss": 0.2059, "step": 425 }, { "epoch": 0.02, "learning_rate": 1.9798858639723083e-05, "loss": 0.1755, "step": 430 }, { "epoch": 0.02, "learning_rate": 1.9796519786696606e-05, "loss": 0.1474, "step": 435 }, { "epoch": 0.02, "learning_rate": 1.979418093367013e-05, "loss": 0.1692, "step": 440 }, { "epoch": 0.02, "learning_rate": 1.9791842080643653e-05, "loss": 0.1754, "step": 445 }, { "epoch": 0.02, "learning_rate": 1.9789503227617176e-05, "loss": 0.1265, "step": 450 }, { "epoch": 0.02, "learning_rate": 1.9787164374590703e-05, "loss": 0.1522, "step": 455 }, { "epoch": 0.02, "learning_rate": 1.9784825521564226e-05, "loss": 0.1628, "step": 460 }, { "epoch": 0.02, "learning_rate": 1.978248666853775e-05, "loss": 0.0948, "step": 465 }, { "epoch": 0.02, "learning_rate": 1.9780147815511276e-05, "loss": 0.0878, "step": 470 }, { "epoch": 0.02, "learning_rate": 1.97778089624848e-05, "loss": 0.1663, "step": 475 }, { "epoch": 0.02, "learning_rate": 1.9775470109458322e-05, "loss": 0.1411, "step": 480 }, { "epoch": 0.02, "learning_rate": 1.977313125643185e-05, "loss": 0.1296, "step": 485 }, { "epoch": 0.02, "learning_rate": 1.9770792403405372e-05, "loss": 0.1273, "step": 490 }, { "epoch": 0.02, "learning_rate": 1.9768453550378896e-05, "loss": 0.0774, "step": 495 }, { "epoch": 0.02, "learning_rate": 1.9766114697352422e-05, "loss": 0.1353, "step": 500 }, { "epoch": 0.02, "learning_rate": 1.9763775844325942e-05, "loss": 0.1887, "step": 505 }, { "epoch": 0.02, "learning_rate": 1.976143699129947e-05, "loss": 0.1163, "step": 510 }, { "epoch": 0.02, "learning_rate": 1.9759098138272992e-05, "loss": 0.1604, "step": 515 }, { "epoch": 0.02, "learning_rate": 1.9756759285246515e-05, "loss": 0.1838, "step": 520 }, { "epoch": 0.02, "learning_rate": 1.9754420432220042e-05, "loss": 0.1546, "step": 525 }, { "epoch": 0.02, "learning_rate": 1.9752081579193565e-05, "loss": 0.1304, "step": 530 }, { "epoch": 0.03, "learning_rate": 1.974974272616709e-05, "loss": 0.1017, "step": 535 }, { "epoch": 0.03, "learning_rate": 1.9747403873140615e-05, "loss": 0.1382, "step": 540 }, { "epoch": 0.03, "learning_rate": 1.974506502011414e-05, "loss": 0.187, "step": 545 }, { "epoch": 0.03, "learning_rate": 1.974272616708766e-05, "loss": 0.1176, "step": 550 }, { "epoch": 0.03, "learning_rate": 1.9740387314061188e-05, "loss": 0.1455, "step": 555 }, { "epoch": 0.03, "learning_rate": 1.973804846103471e-05, "loss": 0.1547, "step": 560 }, { "epoch": 0.03, "learning_rate": 1.9735709608008235e-05, "loss": 0.1295, "step": 565 }, { "epoch": 0.03, "learning_rate": 1.9733370754981758e-05, "loss": 0.123, "step": 570 }, { "epoch": 0.03, "learning_rate": 1.973103190195528e-05, "loss": 0.1043, "step": 575 }, { "epoch": 0.03, "learning_rate": 1.9728693048928808e-05, "loss": 0.1027, "step": 580 }, { "epoch": 0.03, "learning_rate": 1.972635419590233e-05, "loss": 0.2441, "step": 585 }, { "epoch": 0.03, "learning_rate": 1.9724015342875854e-05, "loss": 0.1166, "step": 590 }, { "epoch": 0.03, "learning_rate": 1.972167648984938e-05, "loss": 0.1963, "step": 595 }, { "epoch": 0.03, "learning_rate": 1.9719337636822904e-05, "loss": 0.1192, "step": 600 }, { "epoch": 0.03, "learning_rate": 1.9716998783796428e-05, "loss": 0.0962, "step": 605 }, { "epoch": 0.03, "learning_rate": 1.9714659930769954e-05, "loss": 0.1793, "step": 610 }, { "epoch": 0.03, "learning_rate": 1.9712321077743477e-05, "loss": 0.1316, "step": 615 }, { "epoch": 0.03, "learning_rate": 1.9709982224717e-05, "loss": 0.1477, "step": 620 }, { "epoch": 0.03, "learning_rate": 1.9707643371690524e-05, "loss": 0.1905, "step": 625 }, { "epoch": 0.03, "learning_rate": 1.9705304518664047e-05, "loss": 0.1575, "step": 630 }, { "epoch": 0.03, "learning_rate": 1.9702965665637574e-05, "loss": 0.1656, "step": 635 }, { "epoch": 0.03, "learning_rate": 1.9700626812611097e-05, "loss": 0.136, "step": 640 }, { "epoch": 0.03, "learning_rate": 1.969828795958462e-05, "loss": 0.1598, "step": 645 }, { "epoch": 0.03, "learning_rate": 1.9695949106558147e-05, "loss": 0.1324, "step": 650 }, { "epoch": 0.03, "learning_rate": 1.969361025353167e-05, "loss": 0.1865, "step": 655 }, { "epoch": 0.03, "learning_rate": 1.9691271400505194e-05, "loss": 0.1363, "step": 660 }, { "epoch": 0.03, "learning_rate": 1.9688932547478717e-05, "loss": 0.1377, "step": 665 }, { "epoch": 0.03, "learning_rate": 1.9686593694452243e-05, "loss": 0.1757, "step": 670 }, { "epoch": 0.03, "learning_rate": 1.9684254841425767e-05, "loss": 0.1383, "step": 675 }, { "epoch": 0.03, "learning_rate": 1.968191598839929e-05, "loss": 0.1216, "step": 680 }, { "epoch": 0.03, "learning_rate": 1.9679577135372813e-05, "loss": 0.161, "step": 685 }, { "epoch": 0.03, "learning_rate": 1.9677238282346337e-05, "loss": 0.1899, "step": 690 }, { "epoch": 0.03, "learning_rate": 1.9674899429319863e-05, "loss": 0.123, "step": 695 }, { "epoch": 0.03, "learning_rate": 1.9672560576293386e-05, "loss": 0.0901, "step": 700 }, { "epoch": 0.03, "learning_rate": 1.967022172326691e-05, "loss": 0.1054, "step": 705 }, { "epoch": 0.03, "learning_rate": 1.9667882870240436e-05, "loss": 0.1769, "step": 710 }, { "epoch": 0.03, "learning_rate": 1.966554401721396e-05, "loss": 0.1008, "step": 715 }, { "epoch": 0.03, "learning_rate": 1.9663205164187483e-05, "loss": 0.1116, "step": 720 }, { "epoch": 0.03, "learning_rate": 1.966086631116101e-05, "loss": 0.094, "step": 725 }, { "epoch": 0.03, "learning_rate": 1.9658527458134533e-05, "loss": 0.1042, "step": 730 }, { "epoch": 0.03, "learning_rate": 1.9656188605108056e-05, "loss": 0.169, "step": 735 }, { "epoch": 0.03, "learning_rate": 1.9653849752081583e-05, "loss": 0.1004, "step": 740 }, { "epoch": 0.03, "learning_rate": 1.9651510899055103e-05, "loss": 0.1478, "step": 745 }, { "epoch": 0.04, "learning_rate": 1.964917204602863e-05, "loss": 0.1328, "step": 750 }, { "epoch": 0.04, "learning_rate": 1.9646833193002152e-05, "loss": 0.1385, "step": 755 }, { "epoch": 0.04, "learning_rate": 1.9644494339975676e-05, "loss": 0.122, "step": 760 }, { "epoch": 0.04, "learning_rate": 1.9642155486949202e-05, "loss": 0.1553, "step": 765 }, { "epoch": 0.04, "learning_rate": 1.9639816633922726e-05, "loss": 0.0874, "step": 770 }, { "epoch": 0.04, "learning_rate": 1.963747778089625e-05, "loss": 0.0952, "step": 775 }, { "epoch": 0.04, "learning_rate": 1.9635138927869775e-05, "loss": 0.1158, "step": 780 }, { "epoch": 0.04, "learning_rate": 1.96328000748433e-05, "loss": 0.1283, "step": 785 }, { "epoch": 0.04, "learning_rate": 1.9630461221816822e-05, "loss": 0.1925, "step": 790 }, { "epoch": 0.04, "learning_rate": 1.962812236879035e-05, "loss": 0.1361, "step": 795 }, { "epoch": 0.04, "learning_rate": 1.9625783515763872e-05, "loss": 0.1544, "step": 800 }, { "epoch": 0.04, "learning_rate": 1.9623444662737395e-05, "loss": 0.1148, "step": 805 }, { "epoch": 0.04, "learning_rate": 1.962110580971092e-05, "loss": 0.162, "step": 810 }, { "epoch": 0.04, "learning_rate": 1.961876695668444e-05, "loss": 0.1602, "step": 815 }, { "epoch": 0.04, "learning_rate": 1.961642810365797e-05, "loss": 0.1729, "step": 820 }, { "epoch": 0.04, "learning_rate": 1.961408925063149e-05, "loss": 0.1759, "step": 825 }, { "epoch": 0.04, "learning_rate": 1.9611750397605015e-05, "loss": 0.1637, "step": 830 }, { "epoch": 0.04, "learning_rate": 1.960941154457854e-05, "loss": 0.1466, "step": 835 }, { "epoch": 0.04, "learning_rate": 1.9607072691552065e-05, "loss": 0.1042, "step": 840 }, { "epoch": 0.04, "learning_rate": 1.9604733838525588e-05, "loss": 0.1013, "step": 845 }, { "epoch": 0.04, "learning_rate": 1.9602394985499115e-05, "loss": 0.0974, "step": 850 }, { "epoch": 0.04, "learning_rate": 1.9600056132472638e-05, "loss": 0.173, "step": 855 }, { "epoch": 0.04, "learning_rate": 1.959771727944616e-05, "loss": 0.1042, "step": 860 }, { "epoch": 0.04, "learning_rate": 1.9595378426419684e-05, "loss": 0.1578, "step": 865 }, { "epoch": 0.04, "learning_rate": 1.9593039573393208e-05, "loss": 0.1351, "step": 870 }, { "epoch": 0.04, "learning_rate": 1.9590700720366734e-05, "loss": 0.141, "step": 875 }, { "epoch": 0.04, "learning_rate": 1.9588361867340258e-05, "loss": 0.1046, "step": 880 }, { "epoch": 0.04, "learning_rate": 1.958602301431378e-05, "loss": 0.1145, "step": 885 }, { "epoch": 0.04, "learning_rate": 1.9583684161287307e-05, "loss": 0.1254, "step": 890 }, { "epoch": 0.04, "learning_rate": 1.958134530826083e-05, "loss": 0.1733, "step": 895 }, { "epoch": 0.04, "learning_rate": 1.9579006455234354e-05, "loss": 0.1183, "step": 900 }, { "epoch": 0.04, "learning_rate": 1.957666760220788e-05, "loss": 0.1137, "step": 905 }, { "epoch": 0.04, "learning_rate": 1.9574328749181404e-05, "loss": 0.1662, "step": 910 }, { "epoch": 0.04, "learning_rate": 1.9571989896154927e-05, "loss": 0.1179, "step": 915 }, { "epoch": 0.04, "learning_rate": 1.956965104312845e-05, "loss": 0.0957, "step": 920 }, { "epoch": 0.04, "learning_rate": 1.9567312190101974e-05, "loss": 0.1093, "step": 925 }, { "epoch": 0.04, "learning_rate": 1.95649733370755e-05, "loss": 0.1303, "step": 930 }, { "epoch": 0.04, "learning_rate": 1.9562634484049024e-05, "loss": 0.1566, "step": 935 }, { "epoch": 0.04, "learning_rate": 1.9560295631022547e-05, "loss": 0.078, "step": 940 }, { "epoch": 0.04, "learning_rate": 1.9557956777996073e-05, "loss": 0.167, "step": 945 }, { "epoch": 0.04, "learning_rate": 1.9555617924969597e-05, "loss": 0.2242, "step": 950 }, { "epoch": 0.04, "learning_rate": 1.955327907194312e-05, "loss": 0.1237, "step": 955 }, { "epoch": 0.04, "learning_rate": 1.9550940218916647e-05, "loss": 0.125, "step": 960 }, { "epoch": 0.05, "learning_rate": 1.954860136589017e-05, "loss": 0.1247, "step": 965 }, { "epoch": 0.05, "learning_rate": 1.9546262512863693e-05, "loss": 0.24, "step": 970 }, { "epoch": 0.05, "learning_rate": 1.954392365983722e-05, "loss": 0.1153, "step": 975 }, { "epoch": 0.05, "learning_rate": 1.954158480681074e-05, "loss": 0.1986, "step": 980 }, { "epoch": 0.05, "learning_rate": 1.9539245953784266e-05, "loss": 0.122, "step": 985 }, { "epoch": 0.05, "learning_rate": 1.953690710075779e-05, "loss": 0.1753, "step": 990 }, { "epoch": 0.05, "learning_rate": 1.9534568247731313e-05, "loss": 0.091, "step": 995 }, { "epoch": 0.05, "learning_rate": 1.953222939470484e-05, "loss": 0.1572, "step": 1000 }, { "epoch": 0.05, "learning_rate": 1.9529890541678363e-05, "loss": 0.2004, "step": 1005 }, { "epoch": 0.05, "learning_rate": 1.9527551688651886e-05, "loss": 0.1389, "step": 1010 }, { "epoch": 0.05, "learning_rate": 1.9525212835625413e-05, "loss": 0.1671, "step": 1015 }, { "epoch": 0.05, "learning_rate": 1.9522873982598936e-05, "loss": 0.1531, "step": 1020 }, { "epoch": 0.05, "learning_rate": 1.952053512957246e-05, "loss": 0.1761, "step": 1025 }, { "epoch": 0.05, "learning_rate": 1.9518196276545986e-05, "loss": 0.188, "step": 1030 }, { "epoch": 0.05, "learning_rate": 1.951585742351951e-05, "loss": 0.1407, "step": 1035 }, { "epoch": 0.05, "learning_rate": 1.9513518570493032e-05, "loss": 0.1215, "step": 1040 }, { "epoch": 0.05, "learning_rate": 1.9511179717466556e-05, "loss": 0.1339, "step": 1045 }, { "epoch": 0.05, "learning_rate": 1.950884086444008e-05, "loss": 0.1822, "step": 1050 }, { "epoch": 0.05, "learning_rate": 1.9506502011413605e-05, "loss": 0.1972, "step": 1055 }, { "epoch": 0.05, "learning_rate": 1.950416315838713e-05, "loss": 0.108, "step": 1060 }, { "epoch": 0.05, "learning_rate": 1.9501824305360652e-05, "loss": 0.1337, "step": 1065 }, { "epoch": 0.05, "learning_rate": 1.949948545233418e-05, "loss": 0.0637, "step": 1070 }, { "epoch": 0.05, "learning_rate": 1.9497146599307702e-05, "loss": 0.1238, "step": 1075 }, { "epoch": 0.05, "learning_rate": 1.9494807746281225e-05, "loss": 0.1621, "step": 1080 }, { "epoch": 0.05, "learning_rate": 1.9492468893254752e-05, "loss": 0.1271, "step": 1085 }, { "epoch": 0.05, "learning_rate": 1.9490130040228275e-05, "loss": 0.1343, "step": 1090 }, { "epoch": 0.05, "learning_rate": 1.9487791187201798e-05, "loss": 0.0983, "step": 1095 }, { "epoch": 0.05, "learning_rate": 1.948545233417532e-05, "loss": 0.1535, "step": 1100 }, { "epoch": 0.05, "learning_rate": 1.9483113481148845e-05, "loss": 0.1653, "step": 1105 }, { "epoch": 0.05, "learning_rate": 1.9480774628122368e-05, "loss": 0.0767, "step": 1110 }, { "epoch": 0.05, "learning_rate": 1.9478435775095895e-05, "loss": 0.1884, "step": 1115 }, { "epoch": 0.05, "learning_rate": 1.9476096922069418e-05, "loss": 0.2059, "step": 1120 }, { "epoch": 0.05, "learning_rate": 1.947375806904294e-05, "loss": 0.1234, "step": 1125 }, { "epoch": 0.05, "learning_rate": 1.9471419216016468e-05, "loss": 0.225, "step": 1130 }, { "epoch": 0.05, "learning_rate": 1.946908036298999e-05, "loss": 0.1506, "step": 1135 }, { "epoch": 0.05, "learning_rate": 1.9466741509963514e-05, "loss": 0.1444, "step": 1140 }, { "epoch": 0.05, "learning_rate": 1.946440265693704e-05, "loss": 0.0865, "step": 1145 }, { "epoch": 0.05, "learning_rate": 1.9462063803910564e-05, "loss": 0.1494, "step": 1150 }, { "epoch": 0.05, "learning_rate": 1.9459724950884088e-05, "loss": 0.2087, "step": 1155 }, { "epoch": 0.05, "learning_rate": 1.945738609785761e-05, "loss": 0.1273, "step": 1160 }, { "epoch": 0.05, "learning_rate": 1.9455047244831134e-05, "loss": 0.1575, "step": 1165 }, { "epoch": 0.05, "learning_rate": 1.945270839180466e-05, "loss": 0.199, "step": 1170 }, { "epoch": 0.05, "learning_rate": 1.9450369538778184e-05, "loss": 0.1533, "step": 1175 }, { "epoch": 0.06, "learning_rate": 1.9448030685751707e-05, "loss": 0.096, "step": 1180 }, { "epoch": 0.06, "learning_rate": 1.9445691832725234e-05, "loss": 0.1809, "step": 1185 }, { "epoch": 0.06, "learning_rate": 1.9443352979698757e-05, "loss": 0.1168, "step": 1190 }, { "epoch": 0.06, "learning_rate": 1.944101412667228e-05, "loss": 0.1655, "step": 1195 }, { "epoch": 0.06, "learning_rate": 1.9438675273645807e-05, "loss": 0.2384, "step": 1200 }, { "epoch": 0.06, "learning_rate": 1.943633642061933e-05, "loss": 0.1089, "step": 1205 }, { "epoch": 0.06, "learning_rate": 1.9433997567592854e-05, "loss": 0.1232, "step": 1210 }, { "epoch": 0.06, "learning_rate": 1.943165871456638e-05, "loss": 0.1303, "step": 1215 }, { "epoch": 0.06, "learning_rate": 1.94293198615399e-05, "loss": 0.1231, "step": 1220 }, { "epoch": 0.06, "learning_rate": 1.9426981008513427e-05, "loss": 0.1165, "step": 1225 }, { "epoch": 0.06, "learning_rate": 1.942464215548695e-05, "loss": 0.1202, "step": 1230 }, { "epoch": 0.06, "learning_rate": 1.9422303302460473e-05, "loss": 0.1863, "step": 1235 }, { "epoch": 0.06, "learning_rate": 1.9419964449434e-05, "loss": 0.1697, "step": 1240 }, { "epoch": 0.06, "learning_rate": 1.9417625596407523e-05, "loss": 0.1264, "step": 1245 }, { "epoch": 0.06, "learning_rate": 1.9415286743381046e-05, "loss": 0.1516, "step": 1250 }, { "epoch": 0.06, "learning_rate": 1.9412947890354573e-05, "loss": 0.2865, "step": 1255 }, { "epoch": 0.06, "learning_rate": 1.9410609037328096e-05, "loss": 0.1398, "step": 1260 }, { "epoch": 0.06, "learning_rate": 1.940827018430162e-05, "loss": 0.1555, "step": 1265 }, { "epoch": 0.06, "learning_rate": 1.9405931331275146e-05, "loss": 0.0779, "step": 1270 }, { "epoch": 0.06, "learning_rate": 1.940359247824867e-05, "loss": 0.0841, "step": 1275 }, { "epoch": 0.06, "learning_rate": 1.9401253625222193e-05, "loss": 0.1114, "step": 1280 }, { "epoch": 0.06, "learning_rate": 1.9398914772195716e-05, "loss": 0.1498, "step": 1285 }, { "epoch": 0.06, "learning_rate": 1.939657591916924e-05, "loss": 0.1624, "step": 1290 }, { "epoch": 0.06, "learning_rate": 1.9394237066142766e-05, "loss": 0.1274, "step": 1295 }, { "epoch": 0.06, "learning_rate": 1.939189821311629e-05, "loss": 0.1891, "step": 1300 }, { "epoch": 0.06, "learning_rate": 1.9389559360089812e-05, "loss": 0.1806, "step": 1305 }, { "epoch": 0.06, "learning_rate": 1.938722050706334e-05, "loss": 0.1948, "step": 1310 }, { "epoch": 0.06, "learning_rate": 1.9384881654036862e-05, "loss": 0.1099, "step": 1315 }, { "epoch": 0.06, "learning_rate": 1.9382542801010386e-05, "loss": 0.1731, "step": 1320 }, { "epoch": 0.06, "learning_rate": 1.9380203947983912e-05, "loss": 0.1572, "step": 1325 }, { "epoch": 0.06, "learning_rate": 1.9377865094957435e-05, "loss": 0.1115, "step": 1330 }, { "epoch": 0.06, "learning_rate": 1.937552624193096e-05, "loss": 0.1329, "step": 1335 }, { "epoch": 0.06, "learning_rate": 1.9373187388904482e-05, "loss": 0.1319, "step": 1340 }, { "epoch": 0.06, "learning_rate": 1.9370848535878005e-05, "loss": 0.1158, "step": 1345 }, { "epoch": 0.06, "learning_rate": 1.9368509682851532e-05, "loss": 0.2136, "step": 1350 }, { "epoch": 0.06, "learning_rate": 1.9366170829825055e-05, "loss": 0.1088, "step": 1355 }, { "epoch": 0.06, "learning_rate": 1.936383197679858e-05, "loss": 0.1762, "step": 1360 }, { "epoch": 0.06, "learning_rate": 1.9361493123772105e-05, "loss": 0.108, "step": 1365 }, { "epoch": 0.06, "learning_rate": 1.9359154270745628e-05, "loss": 0.0951, "step": 1370 }, { "epoch": 0.06, "learning_rate": 1.935681541771915e-05, "loss": 0.1531, "step": 1375 }, { "epoch": 0.06, "learning_rate": 1.9354476564692678e-05, "loss": 0.1503, "step": 1380 }, { "epoch": 0.06, "learning_rate": 1.93521377116662e-05, "loss": 0.1596, "step": 1385 }, { "epoch": 0.07, "learning_rate": 1.9349798858639725e-05, "loss": 0.1947, "step": 1390 }, { "epoch": 0.07, "learning_rate": 1.9347460005613248e-05, "loss": 0.1303, "step": 1395 }, { "epoch": 0.07, "learning_rate": 1.934512115258677e-05, "loss": 0.0974, "step": 1400 }, { "epoch": 0.07, "learning_rate": 1.9342782299560298e-05, "loss": 0.0917, "step": 1405 }, { "epoch": 0.07, "learning_rate": 1.934044344653382e-05, "loss": 0.1741, "step": 1410 }, { "epoch": 0.07, "learning_rate": 1.9338104593507344e-05, "loss": 0.084, "step": 1415 }, { "epoch": 0.07, "learning_rate": 1.933576574048087e-05, "loss": 0.1293, "step": 1420 }, { "epoch": 0.07, "learning_rate": 1.9333426887454394e-05, "loss": 0.168, "step": 1425 }, { "epoch": 0.07, "learning_rate": 1.9331088034427918e-05, "loss": 0.1288, "step": 1430 }, { "epoch": 0.07, "learning_rate": 1.9328749181401444e-05, "loss": 0.1375, "step": 1435 }, { "epoch": 0.07, "learning_rate": 1.9326410328374967e-05, "loss": 0.1979, "step": 1440 }, { "epoch": 0.07, "learning_rate": 1.932407147534849e-05, "loss": 0.231, "step": 1445 }, { "epoch": 0.07, "learning_rate": 1.9321732622322017e-05, "loss": 0.0967, "step": 1450 }, { "epoch": 0.07, "learning_rate": 1.9319393769295537e-05, "loss": 0.1523, "step": 1455 }, { "epoch": 0.07, "learning_rate": 1.9317054916269064e-05, "loss": 0.1322, "step": 1460 }, { "epoch": 0.07, "learning_rate": 1.9314716063242587e-05, "loss": 0.1292, "step": 1465 }, { "epoch": 0.07, "learning_rate": 1.931237721021611e-05, "loss": 0.14, "step": 1470 }, { "epoch": 0.07, "learning_rate": 1.9310038357189637e-05, "loss": 0.1323, "step": 1475 }, { "epoch": 0.07, "learning_rate": 1.930769950416316e-05, "loss": 0.1737, "step": 1480 }, { "epoch": 0.07, "learning_rate": 1.9305360651136684e-05, "loss": 0.1393, "step": 1485 }, { "epoch": 0.07, "learning_rate": 1.930302179811021e-05, "loss": 0.1425, "step": 1490 }, { "epoch": 0.07, "learning_rate": 1.9300682945083733e-05, "loss": 0.2035, "step": 1495 }, { "epoch": 0.07, "learning_rate": 1.9298344092057257e-05, "loss": 0.1039, "step": 1500 }, { "epoch": 0.07, "learning_rate": 1.9296005239030783e-05, "loss": 0.1608, "step": 1505 }, { "epoch": 0.07, "learning_rate": 1.9293666386004307e-05, "loss": 0.1792, "step": 1510 }, { "epoch": 0.07, "learning_rate": 1.929132753297783e-05, "loss": 0.1682, "step": 1515 }, { "epoch": 0.07, "learning_rate": 1.9288988679951353e-05, "loss": 0.1491, "step": 1520 }, { "epoch": 0.07, "learning_rate": 1.9286649826924876e-05, "loss": 0.1204, "step": 1525 }, { "epoch": 0.07, "learning_rate": 1.9284310973898403e-05, "loss": 0.0899, "step": 1530 }, { "epoch": 0.07, "learning_rate": 1.9281972120871926e-05, "loss": 0.1087, "step": 1535 }, { "epoch": 0.07, "learning_rate": 1.927963326784545e-05, "loss": 0.1392, "step": 1540 }, { "epoch": 0.07, "learning_rate": 1.9277294414818973e-05, "loss": 0.1253, "step": 1545 }, { "epoch": 0.07, "learning_rate": 1.92749555617925e-05, "loss": 0.097, "step": 1550 }, { "epoch": 0.07, "learning_rate": 1.9272616708766023e-05, "loss": 0.1605, "step": 1555 }, { "epoch": 0.07, "learning_rate": 1.9270277855739546e-05, "loss": 0.1322, "step": 1560 }, { "epoch": 0.07, "learning_rate": 1.9267939002713073e-05, "loss": 0.1268, "step": 1565 }, { "epoch": 0.07, "learning_rate": 1.9265600149686596e-05, "loss": 0.0761, "step": 1570 }, { "epoch": 0.07, "learning_rate": 1.926326129666012e-05, "loss": 0.0882, "step": 1575 }, { "epoch": 0.07, "learning_rate": 1.9260922443633642e-05, "loss": 0.1902, "step": 1580 }, { "epoch": 0.07, "learning_rate": 1.9258583590607166e-05, "loss": 0.0748, "step": 1585 }, { "epoch": 0.07, "learning_rate": 1.9256244737580692e-05, "loss": 0.1409, "step": 1590 }, { "epoch": 0.07, "learning_rate": 1.9253905884554216e-05, "loss": 0.1405, "step": 1595 }, { "epoch": 0.07, "learning_rate": 1.925156703152774e-05, "loss": 0.1436, "step": 1600 }, { "epoch": 0.08, "learning_rate": 1.9249228178501265e-05, "loss": 0.1698, "step": 1605 }, { "epoch": 0.08, "learning_rate": 1.924688932547479e-05, "loss": 0.109, "step": 1610 }, { "epoch": 0.08, "learning_rate": 1.9244550472448312e-05, "loss": 0.1635, "step": 1615 }, { "epoch": 0.08, "learning_rate": 1.924221161942184e-05, "loss": 0.1047, "step": 1620 }, { "epoch": 0.08, "learning_rate": 1.9239872766395362e-05, "loss": 0.1078, "step": 1625 }, { "epoch": 0.08, "learning_rate": 1.9237533913368885e-05, "loss": 0.145, "step": 1630 }, { "epoch": 0.08, "learning_rate": 1.923519506034241e-05, "loss": 0.0902, "step": 1635 }, { "epoch": 0.08, "learning_rate": 1.923285620731593e-05, "loss": 0.1171, "step": 1640 }, { "epoch": 0.08, "learning_rate": 1.9230517354289458e-05, "loss": 0.0963, "step": 1645 }, { "epoch": 0.08, "learning_rate": 1.922817850126298e-05, "loss": 0.1586, "step": 1650 }, { "epoch": 0.08, "learning_rate": 1.9225839648236505e-05, "loss": 0.155, "step": 1655 }, { "epoch": 0.08, "learning_rate": 1.922350079521003e-05, "loss": 0.172, "step": 1660 }, { "epoch": 0.08, "learning_rate": 1.9221161942183555e-05, "loss": 0.1931, "step": 1665 }, { "epoch": 0.08, "learning_rate": 1.9218823089157078e-05, "loss": 0.1285, "step": 1670 }, { "epoch": 0.08, "learning_rate": 1.9216484236130605e-05, "loss": 0.1524, "step": 1675 }, { "epoch": 0.08, "learning_rate": 1.9214145383104128e-05, "loss": 0.1465, "step": 1680 }, { "epoch": 0.08, "learning_rate": 1.921180653007765e-05, "loss": 0.1021, "step": 1685 }, { "epoch": 0.08, "learning_rate": 1.9209467677051178e-05, "loss": 0.0814, "step": 1690 }, { "epoch": 0.08, "learning_rate": 1.9207128824024698e-05, "loss": 0.123, "step": 1695 }, { "epoch": 0.08, "learning_rate": 1.9204789970998224e-05, "loss": 0.1639, "step": 1700 }, { "epoch": 0.08, "learning_rate": 1.9202451117971748e-05, "loss": 0.1966, "step": 1705 }, { "epoch": 0.08, "learning_rate": 1.920011226494527e-05, "loss": 0.1318, "step": 1710 }, { "epoch": 0.08, "learning_rate": 1.9197773411918797e-05, "loss": 0.1688, "step": 1715 }, { "epoch": 0.08, "learning_rate": 1.919543455889232e-05, "loss": 0.1215, "step": 1720 }, { "epoch": 0.08, "learning_rate": 1.9193095705865844e-05, "loss": 0.1878, "step": 1725 }, { "epoch": 0.08, "learning_rate": 1.919075685283937e-05, "loss": 0.1264, "step": 1730 }, { "epoch": 0.08, "learning_rate": 1.9188417999812894e-05, "loss": 0.1062, "step": 1735 }, { "epoch": 0.08, "learning_rate": 1.9186079146786417e-05, "loss": 0.1533, "step": 1740 }, { "epoch": 0.08, "learning_rate": 1.9183740293759944e-05, "loss": 0.0927, "step": 1745 }, { "epoch": 0.08, "learning_rate": 1.9181401440733464e-05, "loss": 0.1449, "step": 1750 }, { "epoch": 0.08, "learning_rate": 1.917906258770699e-05, "loss": 0.0942, "step": 1755 }, { "epoch": 0.08, "learning_rate": 1.9176723734680514e-05, "loss": 0.0983, "step": 1760 }, { "epoch": 0.08, "learning_rate": 1.9174384881654037e-05, "loss": 0.0938, "step": 1765 }, { "epoch": 0.08, "learning_rate": 1.9172046028627563e-05, "loss": 0.1672, "step": 1770 }, { "epoch": 0.08, "learning_rate": 1.9169707175601087e-05, "loss": 0.1242, "step": 1775 }, { "epoch": 0.08, "learning_rate": 1.916736832257461e-05, "loss": 0.1105, "step": 1780 }, { "epoch": 0.08, "learning_rate": 1.9165029469548137e-05, "loss": 0.07, "step": 1785 }, { "epoch": 0.08, "learning_rate": 1.916269061652166e-05, "loss": 0.0587, "step": 1790 }, { "epoch": 0.08, "learning_rate": 1.9160351763495183e-05, "loss": 0.1305, "step": 1795 }, { "epoch": 0.08, "learning_rate": 1.915801291046871e-05, "loss": 0.1494, "step": 1800 }, { "epoch": 0.08, "learning_rate": 1.9155674057442233e-05, "loss": 0.1309, "step": 1805 }, { "epoch": 0.08, "learning_rate": 1.9153335204415756e-05, "loss": 0.1766, "step": 1810 }, { "epoch": 0.08, "learning_rate": 1.915099635138928e-05, "loss": 0.0923, "step": 1815 }, { "epoch": 0.09, "learning_rate": 1.9148657498362803e-05, "loss": 0.0688, "step": 1820 }, { "epoch": 0.09, "learning_rate": 1.914631864533633e-05, "loss": 0.2169, "step": 1825 }, { "epoch": 0.09, "learning_rate": 1.9143979792309853e-05, "loss": 0.0635, "step": 1830 }, { "epoch": 0.09, "learning_rate": 1.9141640939283376e-05, "loss": 0.1536, "step": 1835 }, { "epoch": 0.09, "learning_rate": 1.9139302086256903e-05, "loss": 0.1329, "step": 1840 }, { "epoch": 0.09, "learning_rate": 1.9136963233230426e-05, "loss": 0.114, "step": 1845 }, { "epoch": 0.09, "learning_rate": 1.913462438020395e-05, "loss": 0.1494, "step": 1850 }, { "epoch": 0.09, "learning_rate": 1.9132285527177476e-05, "loss": 0.1439, "step": 1855 }, { "epoch": 0.09, "learning_rate": 1.9129946674151e-05, "loss": 0.1966, "step": 1860 }, { "epoch": 0.09, "learning_rate": 1.9127607821124522e-05, "loss": 0.1515, "step": 1865 }, { "epoch": 0.09, "learning_rate": 1.9125268968098046e-05, "loss": 0.1078, "step": 1870 }, { "epoch": 0.09, "learning_rate": 1.912293011507157e-05, "loss": 0.1284, "step": 1875 }, { "epoch": 0.09, "learning_rate": 1.9120591262045095e-05, "loss": 0.078, "step": 1880 }, { "epoch": 0.09, "learning_rate": 1.911825240901862e-05, "loss": 0.0872, "step": 1885 }, { "epoch": 0.09, "learning_rate": 1.9115913555992142e-05, "loss": 0.0754, "step": 1890 }, { "epoch": 0.09, "learning_rate": 1.911357470296567e-05, "loss": 0.1363, "step": 1895 }, { "epoch": 0.09, "learning_rate": 1.9111235849939192e-05, "loss": 0.2519, "step": 1900 }, { "epoch": 0.09, "learning_rate": 1.9108896996912715e-05, "loss": 0.1174, "step": 1905 }, { "epoch": 0.09, "learning_rate": 1.9106558143886242e-05, "loss": 0.1119, "step": 1910 }, { "epoch": 0.09, "learning_rate": 1.9104219290859765e-05, "loss": 0.1222, "step": 1915 }, { "epoch": 0.09, "learning_rate": 1.9101880437833288e-05, "loss": 0.1837, "step": 1920 }, { "epoch": 0.09, "learning_rate": 1.9099541584806815e-05, "loss": 0.1341, "step": 1925 }, { "epoch": 0.09, "learning_rate": 1.9097202731780335e-05, "loss": 0.0936, "step": 1930 }, { "epoch": 0.09, "learning_rate": 1.909486387875386e-05, "loss": 0.1525, "step": 1935 }, { "epoch": 0.09, "learning_rate": 1.9092525025727385e-05, "loss": 0.0985, "step": 1940 }, { "epoch": 0.09, "learning_rate": 1.9090186172700908e-05, "loss": 0.1269, "step": 1945 }, { "epoch": 0.09, "learning_rate": 1.9087847319674435e-05, "loss": 0.1252, "step": 1950 }, { "epoch": 0.09, "learning_rate": 1.9085508466647958e-05, "loss": 0.1033, "step": 1955 }, { "epoch": 0.09, "learning_rate": 1.908316961362148e-05, "loss": 0.1332, "step": 1960 }, { "epoch": 0.09, "learning_rate": 1.9080830760595008e-05, "loss": 0.1824, "step": 1965 }, { "epoch": 0.09, "learning_rate": 1.907849190756853e-05, "loss": 0.1707, "step": 1970 }, { "epoch": 0.09, "learning_rate": 1.9076153054542054e-05, "loss": 0.0979, "step": 1975 }, { "epoch": 0.09, "learning_rate": 1.9073814201515577e-05, "loss": 0.1479, "step": 1980 }, { "epoch": 0.09, "learning_rate": 1.9071475348489104e-05, "loss": 0.1435, "step": 1985 }, { "epoch": 0.09, "learning_rate": 1.9069136495462624e-05, "loss": 0.1715, "step": 1990 }, { "epoch": 0.09, "learning_rate": 1.906679764243615e-05, "loss": 0.0874, "step": 1995 }, { "epoch": 0.09, "learning_rate": 1.9064458789409674e-05, "loss": 0.1321, "step": 2000 }, { "epoch": 0.09, "learning_rate": 1.9062119936383197e-05, "loss": 0.1344, "step": 2005 }, { "epoch": 0.09, "learning_rate": 1.9059781083356724e-05, "loss": 0.1467, "step": 2010 }, { "epoch": 0.09, "learning_rate": 1.9057442230330247e-05, "loss": 0.0991, "step": 2015 }, { "epoch": 0.09, "learning_rate": 1.905510337730377e-05, "loss": 0.1198, "step": 2020 }, { "epoch": 0.09, "learning_rate": 1.9052764524277297e-05, "loss": 0.0972, "step": 2025 }, { "epoch": 0.09, "learning_rate": 1.905042567125082e-05, "loss": 0.0864, "step": 2030 }, { "epoch": 0.1, "learning_rate": 1.9048086818224343e-05, "loss": 0.1343, "step": 2035 }, { "epoch": 0.1, "learning_rate": 1.904574796519787e-05, "loss": 0.1037, "step": 2040 }, { "epoch": 0.1, "learning_rate": 1.9043409112171393e-05, "loss": 0.1093, "step": 2045 }, { "epoch": 0.1, "learning_rate": 1.9041070259144917e-05, "loss": 0.1216, "step": 2050 }, { "epoch": 0.1, "learning_rate": 1.903873140611844e-05, "loss": 0.206, "step": 2055 }, { "epoch": 0.1, "learning_rate": 1.9036392553091963e-05, "loss": 0.1355, "step": 2060 }, { "epoch": 0.1, "learning_rate": 1.903405370006549e-05, "loss": 0.1187, "step": 2065 }, { "epoch": 0.1, "learning_rate": 1.9031714847039013e-05, "loss": 0.0993, "step": 2070 }, { "epoch": 0.1, "learning_rate": 1.9029375994012536e-05, "loss": 0.1286, "step": 2075 }, { "epoch": 0.1, "learning_rate": 1.9027037140986063e-05, "loss": 0.1331, "step": 2080 }, { "epoch": 0.1, "learning_rate": 1.9024698287959586e-05, "loss": 0.1931, "step": 2085 }, { "epoch": 0.1, "learning_rate": 1.902235943493311e-05, "loss": 0.0952, "step": 2090 }, { "epoch": 0.1, "learning_rate": 1.9020020581906636e-05, "loss": 0.1776, "step": 2095 }, { "epoch": 0.1, "learning_rate": 1.901768172888016e-05, "loss": 0.215, "step": 2100 }, { "epoch": 0.1, "learning_rate": 1.9015342875853683e-05, "loss": 0.1457, "step": 2105 }, { "epoch": 0.1, "learning_rate": 1.9013004022827206e-05, "loss": 0.1489, "step": 2110 }, { "epoch": 0.1, "learning_rate": 1.901066516980073e-05, "loss": 0.1577, "step": 2115 }, { "epoch": 0.1, "learning_rate": 1.9008326316774256e-05, "loss": 0.1466, "step": 2120 }, { "epoch": 0.1, "learning_rate": 1.900598746374778e-05, "loss": 0.1128, "step": 2125 }, { "epoch": 0.1, "learning_rate": 1.9003648610721302e-05, "loss": 0.1427, "step": 2130 }, { "epoch": 0.1, "learning_rate": 1.900130975769483e-05, "loss": 0.1308, "step": 2135 }, { "epoch": 0.1, "learning_rate": 1.8998970904668352e-05, "loss": 0.1049, "step": 2140 }, { "epoch": 0.1, "learning_rate": 1.8996632051641875e-05, "loss": 0.2216, "step": 2145 }, { "epoch": 0.1, "learning_rate": 1.8994293198615402e-05, "loss": 0.1416, "step": 2150 }, { "epoch": 0.1, "learning_rate": 1.8991954345588925e-05, "loss": 0.1156, "step": 2155 }, { "epoch": 0.1, "learning_rate": 1.898961549256245e-05, "loss": 0.1219, "step": 2160 }, { "epoch": 0.1, "learning_rate": 1.8987276639535975e-05, "loss": 0.238, "step": 2165 }, { "epoch": 0.1, "learning_rate": 1.8984937786509495e-05, "loss": 0.1216, "step": 2170 }, { "epoch": 0.1, "learning_rate": 1.8982598933483022e-05, "loss": 0.1015, "step": 2175 }, { "epoch": 0.1, "learning_rate": 1.8980260080456545e-05, "loss": 0.1485, "step": 2180 }, { "epoch": 0.1, "learning_rate": 1.897792122743007e-05, "loss": 0.1398, "step": 2185 }, { "epoch": 0.1, "learning_rate": 1.8975582374403595e-05, "loss": 0.1864, "step": 2190 }, { "epoch": 0.1, "learning_rate": 1.8973243521377118e-05, "loss": 0.1223, "step": 2195 }, { "epoch": 0.1, "learning_rate": 1.897090466835064e-05, "loss": 0.1881, "step": 2200 }, { "epoch": 0.1, "learning_rate": 1.8968565815324168e-05, "loss": 0.1189, "step": 2205 }, { "epoch": 0.1, "learning_rate": 1.896622696229769e-05, "loss": 0.1593, "step": 2210 }, { "epoch": 0.1, "learning_rate": 1.8963888109271215e-05, "loss": 0.1015, "step": 2215 }, { "epoch": 0.1, "learning_rate": 1.896154925624474e-05, "loss": 0.1166, "step": 2220 }, { "epoch": 0.1, "learning_rate": 1.895921040321826e-05, "loss": 0.1306, "step": 2225 }, { "epoch": 0.1, "learning_rate": 1.8956871550191788e-05, "loss": 0.1622, "step": 2230 }, { "epoch": 0.1, "learning_rate": 1.895453269716531e-05, "loss": 0.1442, "step": 2235 }, { "epoch": 0.1, "learning_rate": 1.8952193844138834e-05, "loss": 0.1141, "step": 2240 }, { "epoch": 0.11, "learning_rate": 1.894985499111236e-05, "loss": 0.1362, "step": 2245 }, { "epoch": 0.11, "learning_rate": 1.8947516138085884e-05, "loss": 0.178, "step": 2250 }, { "epoch": 0.11, "learning_rate": 1.8945177285059407e-05, "loss": 0.1447, "step": 2255 }, { "epoch": 0.11, "learning_rate": 1.8942838432032934e-05, "loss": 0.1377, "step": 2260 }, { "epoch": 0.11, "learning_rate": 1.8940499579006457e-05, "loss": 0.0793, "step": 2265 }, { "epoch": 0.11, "learning_rate": 1.893816072597998e-05, "loss": 0.157, "step": 2270 }, { "epoch": 0.11, "learning_rate": 1.8935821872953507e-05, "loss": 0.1021, "step": 2275 }, { "epoch": 0.11, "learning_rate": 1.893348301992703e-05, "loss": 0.1293, "step": 2280 }, { "epoch": 0.11, "learning_rate": 1.8931144166900554e-05, "loss": 0.1174, "step": 2285 }, { "epoch": 0.11, "learning_rate": 1.8928805313874077e-05, "loss": 0.1218, "step": 2290 }, { "epoch": 0.11, "learning_rate": 1.89264664608476e-05, "loss": 0.0728, "step": 2295 }, { "epoch": 0.11, "learning_rate": 1.8924127607821127e-05, "loss": 0.2162, "step": 2300 }, { "epoch": 0.11, "learning_rate": 1.892178875479465e-05, "loss": 0.1955, "step": 2305 }, { "epoch": 0.11, "learning_rate": 1.8919449901768173e-05, "loss": 0.1108, "step": 2310 }, { "epoch": 0.11, "learning_rate": 1.89171110487417e-05, "loss": 0.2064, "step": 2315 }, { "epoch": 0.11, "learning_rate": 1.8914772195715223e-05, "loss": 0.0842, "step": 2320 }, { "epoch": 0.11, "learning_rate": 1.8912433342688747e-05, "loss": 0.1397, "step": 2325 }, { "epoch": 0.11, "learning_rate": 1.8910094489662273e-05, "loss": 0.1096, "step": 2330 }, { "epoch": 0.11, "learning_rate": 1.8907755636635797e-05, "loss": 0.1451, "step": 2335 }, { "epoch": 0.11, "learning_rate": 1.890541678360932e-05, "loss": 0.1588, "step": 2340 }, { "epoch": 0.11, "learning_rate": 1.8903077930582843e-05, "loss": 0.1626, "step": 2345 }, { "epoch": 0.11, "learning_rate": 1.8900739077556366e-05, "loss": 0.1248, "step": 2350 }, { "epoch": 0.11, "learning_rate": 1.8898400224529893e-05, "loss": 0.0639, "step": 2355 }, { "epoch": 0.11, "learning_rate": 1.8896061371503416e-05, "loss": 0.0695, "step": 2360 }, { "epoch": 0.11, "learning_rate": 1.889372251847694e-05, "loss": 0.1199, "step": 2365 }, { "epoch": 0.11, "learning_rate": 1.8891383665450466e-05, "loss": 0.176, "step": 2370 }, { "epoch": 0.11, "learning_rate": 1.888904481242399e-05, "loss": 0.1518, "step": 2375 }, { "epoch": 0.11, "learning_rate": 1.8886705959397513e-05, "loss": 0.1264, "step": 2380 }, { "epoch": 0.11, "learning_rate": 1.888436710637104e-05, "loss": 0.1685, "step": 2385 }, { "epoch": 0.11, "learning_rate": 1.8882028253344563e-05, "loss": 0.1178, "step": 2390 }, { "epoch": 0.11, "learning_rate": 1.8879689400318086e-05, "loss": 0.1357, "step": 2395 }, { "epoch": 0.11, "learning_rate": 1.8877350547291612e-05, "loss": 0.1321, "step": 2400 }, { "epoch": 0.11, "learning_rate": 1.8875011694265132e-05, "loss": 0.1305, "step": 2405 }, { "epoch": 0.11, "learning_rate": 1.887267284123866e-05, "loss": 0.114, "step": 2410 }, { "epoch": 0.11, "learning_rate": 1.8870333988212182e-05, "loss": 0.1013, "step": 2415 }, { "epoch": 0.11, "learning_rate": 1.8867995135185705e-05, "loss": 0.0785, "step": 2420 }, { "epoch": 0.11, "learning_rate": 1.886565628215923e-05, "loss": 0.1163, "step": 2425 }, { "epoch": 0.11, "learning_rate": 1.8863317429132755e-05, "loss": 0.1107, "step": 2430 }, { "epoch": 0.11, "learning_rate": 1.886097857610628e-05, "loss": 0.1306, "step": 2435 }, { "epoch": 0.11, "learning_rate": 1.8858639723079802e-05, "loss": 0.1314, "step": 2440 }, { "epoch": 0.11, "learning_rate": 1.885630087005333e-05, "loss": 0.0996, "step": 2445 }, { "epoch": 0.11, "learning_rate": 1.8853962017026852e-05, "loss": 0.1431, "step": 2450 }, { "epoch": 0.11, "learning_rate": 1.8851623164000375e-05, "loss": 0.1919, "step": 2455 }, { "epoch": 0.12, "learning_rate": 1.88492843109739e-05, "loss": 0.1631, "step": 2460 }, { "epoch": 0.12, "learning_rate": 1.884694545794742e-05, "loss": 0.1285, "step": 2465 }, { "epoch": 0.12, "learning_rate": 1.8844606604920948e-05, "loss": 0.1527, "step": 2470 }, { "epoch": 0.12, "learning_rate": 1.884226775189447e-05, "loss": 0.0932, "step": 2475 }, { "epoch": 0.12, "learning_rate": 1.8839928898867995e-05, "loss": 0.1288, "step": 2480 }, { "epoch": 0.12, "learning_rate": 1.883759004584152e-05, "loss": 0.1051, "step": 2485 }, { "epoch": 0.12, "learning_rate": 1.8835251192815045e-05, "loss": 0.1557, "step": 2490 }, { "epoch": 0.12, "learning_rate": 1.8832912339788568e-05, "loss": 0.1017, "step": 2495 }, { "epoch": 0.12, "learning_rate": 1.8830573486762095e-05, "loss": 0.15, "step": 2500 }, { "epoch": 0.12, "learning_rate": 1.8828234633735618e-05, "loss": 0.0733, "step": 2505 }, { "epoch": 0.12, "learning_rate": 1.882589578070914e-05, "loss": 0.1155, "step": 2510 }, { "epoch": 0.12, "learning_rate": 1.8823556927682668e-05, "loss": 0.1397, "step": 2515 }, { "epoch": 0.12, "learning_rate": 1.882121807465619e-05, "loss": 0.2046, "step": 2520 }, { "epoch": 0.12, "learning_rate": 1.8818879221629714e-05, "loss": 0.0848, "step": 2525 }, { "epoch": 0.12, "learning_rate": 1.8816540368603237e-05, "loss": 0.1114, "step": 2530 }, { "epoch": 0.12, "learning_rate": 1.881420151557676e-05, "loss": 0.0933, "step": 2535 }, { "epoch": 0.12, "learning_rate": 1.8811862662550287e-05, "loss": 0.1211, "step": 2540 }, { "epoch": 0.12, "learning_rate": 1.880952380952381e-05, "loss": 0.1325, "step": 2545 }, { "epoch": 0.12, "learning_rate": 1.8807184956497334e-05, "loss": 0.1117, "step": 2550 }, { "epoch": 0.12, "learning_rate": 1.880484610347086e-05, "loss": 0.0755, "step": 2555 }, { "epoch": 0.12, "learning_rate": 1.8802507250444384e-05, "loss": 0.1492, "step": 2560 }, { "epoch": 0.12, "learning_rate": 1.8800168397417907e-05, "loss": 0.1815, "step": 2565 }, { "epoch": 0.12, "learning_rate": 1.8797829544391434e-05, "loss": 0.0701, "step": 2570 }, { "epoch": 0.12, "learning_rate": 1.8795490691364957e-05, "loss": 0.1432, "step": 2575 }, { "epoch": 0.12, "learning_rate": 1.879315183833848e-05, "loss": 0.0987, "step": 2580 }, { "epoch": 0.12, "learning_rate": 1.8790812985312003e-05, "loss": 0.164, "step": 2585 }, { "epoch": 0.12, "learning_rate": 1.8788474132285527e-05, "loss": 0.1283, "step": 2590 }, { "epoch": 0.12, "learning_rate": 1.8786135279259053e-05, "loss": 0.1175, "step": 2595 }, { "epoch": 0.12, "learning_rate": 1.8783796426232577e-05, "loss": 0.1052, "step": 2600 }, { "epoch": 0.12, "learning_rate": 1.87814575732061e-05, "loss": 0.1178, "step": 2605 }, { "epoch": 0.12, "learning_rate": 1.8779118720179627e-05, "loss": 0.1317, "step": 2610 }, { "epoch": 0.12, "learning_rate": 1.877677986715315e-05, "loss": 0.1882, "step": 2615 }, { "epoch": 0.12, "learning_rate": 1.8774441014126673e-05, "loss": 0.143, "step": 2620 }, { "epoch": 0.12, "learning_rate": 1.87721021611002e-05, "loss": 0.0915, "step": 2625 }, { "epoch": 0.12, "learning_rate": 1.8769763308073723e-05, "loss": 0.1779, "step": 2630 }, { "epoch": 0.12, "learning_rate": 1.8767424455047246e-05, "loss": 0.2211, "step": 2635 }, { "epoch": 0.12, "learning_rate": 1.876508560202077e-05, "loss": 0.0754, "step": 2640 }, { "epoch": 0.12, "learning_rate": 1.8762746748994293e-05, "loss": 0.0829, "step": 2645 }, { "epoch": 0.12, "learning_rate": 1.876040789596782e-05, "loss": 0.1624, "step": 2650 }, { "epoch": 0.12, "learning_rate": 1.8758069042941343e-05, "loss": 0.1225, "step": 2655 }, { "epoch": 0.12, "learning_rate": 1.8755730189914866e-05, "loss": 0.1008, "step": 2660 }, { "epoch": 0.12, "learning_rate": 1.8753391336888393e-05, "loss": 0.113, "step": 2665 }, { "epoch": 0.12, "learning_rate": 1.8751052483861916e-05, "loss": 0.0977, "step": 2670 }, { "epoch": 0.13, "learning_rate": 1.874871363083544e-05, "loss": 0.0846, "step": 2675 }, { "epoch": 0.13, "learning_rate": 1.8746374777808966e-05, "loss": 0.0849, "step": 2680 }, { "epoch": 0.13, "learning_rate": 1.874403592478249e-05, "loss": 0.0925, "step": 2685 }, { "epoch": 0.13, "learning_rate": 1.8741697071756012e-05, "loss": 0.0715, "step": 2690 }, { "epoch": 0.13, "learning_rate": 1.873935821872954e-05, "loss": 0.0879, "step": 2695 }, { "epoch": 0.13, "learning_rate": 1.873701936570306e-05, "loss": 0.1571, "step": 2700 }, { "epoch": 0.13, "learning_rate": 1.8734680512676585e-05, "loss": 0.1045, "step": 2705 }, { "epoch": 0.13, "learning_rate": 1.873234165965011e-05, "loss": 0.138, "step": 2710 }, { "epoch": 0.13, "learning_rate": 1.8730002806623632e-05, "loss": 0.1424, "step": 2715 }, { "epoch": 0.13, "learning_rate": 1.872766395359716e-05, "loss": 0.1596, "step": 2720 }, { "epoch": 0.13, "learning_rate": 1.8725325100570682e-05, "loss": 0.146, "step": 2725 }, { "epoch": 0.13, "learning_rate": 1.8722986247544205e-05, "loss": 0.1831, "step": 2730 }, { "epoch": 0.13, "learning_rate": 1.872064739451773e-05, "loss": 0.1174, "step": 2735 }, { "epoch": 0.13, "learning_rate": 1.8718308541491255e-05, "loss": 0.19, "step": 2740 }, { "epoch": 0.13, "learning_rate": 1.8715969688464778e-05, "loss": 0.1164, "step": 2745 }, { "epoch": 0.13, "learning_rate": 1.8713630835438305e-05, "loss": 0.1308, "step": 2750 }, { "epoch": 0.13, "learning_rate": 1.8711291982411828e-05, "loss": 0.0996, "step": 2755 }, { "epoch": 0.13, "learning_rate": 1.870895312938535e-05, "loss": 0.1737, "step": 2760 }, { "epoch": 0.13, "learning_rate": 1.8706614276358875e-05, "loss": 0.0956, "step": 2765 }, { "epoch": 0.13, "learning_rate": 1.8704275423332398e-05, "loss": 0.0846, "step": 2770 }, { "epoch": 0.13, "learning_rate": 1.8701936570305925e-05, "loss": 0.2156, "step": 2775 }, { "epoch": 0.13, "learning_rate": 1.8699597717279448e-05, "loss": 0.0606, "step": 2780 }, { "epoch": 0.13, "learning_rate": 1.869725886425297e-05, "loss": 0.1254, "step": 2785 }, { "epoch": 0.13, "learning_rate": 1.8694920011226498e-05, "loss": 0.0823, "step": 2790 }, { "epoch": 0.13, "learning_rate": 1.869258115820002e-05, "loss": 0.1394, "step": 2795 }, { "epoch": 0.13, "learning_rate": 1.8690242305173544e-05, "loss": 0.1569, "step": 2800 }, { "epoch": 0.13, "learning_rate": 1.868790345214707e-05, "loss": 0.1627, "step": 2805 }, { "epoch": 0.13, "learning_rate": 1.8685564599120594e-05, "loss": 0.1266, "step": 2810 }, { "epoch": 0.13, "learning_rate": 1.8683225746094117e-05, "loss": 0.1596, "step": 2815 }, { "epoch": 0.13, "learning_rate": 1.868088689306764e-05, "loss": 0.1057, "step": 2820 }, { "epoch": 0.13, "learning_rate": 1.8678548040041164e-05, "loss": 0.1739, "step": 2825 }, { "epoch": 0.13, "learning_rate": 1.867620918701469e-05, "loss": 0.1128, "step": 2830 }, { "epoch": 0.13, "learning_rate": 1.8673870333988214e-05, "loss": 0.1727, "step": 2835 }, { "epoch": 0.13, "learning_rate": 1.8671531480961737e-05, "loss": 0.1094, "step": 2840 }, { "epoch": 0.13, "learning_rate": 1.8669192627935264e-05, "loss": 0.0889, "step": 2845 }, { "epoch": 0.13, "learning_rate": 1.8666853774908787e-05, "loss": 0.1302, "step": 2850 }, { "epoch": 0.13, "learning_rate": 1.866451492188231e-05, "loss": 0.1156, "step": 2855 }, { "epoch": 0.13, "learning_rate": 1.8662176068855833e-05, "loss": 0.1225, "step": 2860 }, { "epoch": 0.13, "learning_rate": 1.865983721582936e-05, "loss": 0.0836, "step": 2865 }, { "epoch": 0.13, "learning_rate": 1.8657498362802883e-05, "loss": 0.1166, "step": 2870 }, { "epoch": 0.13, "learning_rate": 1.8655159509776407e-05, "loss": 0.1058, "step": 2875 }, { "epoch": 0.13, "learning_rate": 1.865282065674993e-05, "loss": 0.1036, "step": 2880 }, { "epoch": 0.13, "learning_rate": 1.8650481803723453e-05, "loss": 0.0873, "step": 2885 }, { "epoch": 0.14, "learning_rate": 1.864814295069698e-05, "loss": 0.1835, "step": 2890 }, { "epoch": 0.14, "learning_rate": 1.8645804097670503e-05, "loss": 0.1387, "step": 2895 }, { "epoch": 0.14, "learning_rate": 1.8643465244644026e-05, "loss": 0.123, "step": 2900 }, { "epoch": 0.14, "learning_rate": 1.8641126391617553e-05, "loss": 0.1831, "step": 2905 }, { "epoch": 0.14, "learning_rate": 1.8638787538591076e-05, "loss": 0.1033, "step": 2910 }, { "epoch": 0.14, "learning_rate": 1.86364486855646e-05, "loss": 0.0832, "step": 2915 }, { "epoch": 0.14, "learning_rate": 1.8634109832538126e-05, "loss": 0.1226, "step": 2920 }, { "epoch": 0.14, "learning_rate": 1.863177097951165e-05, "loss": 0.1141, "step": 2925 }, { "epoch": 0.14, "learning_rate": 1.8629432126485173e-05, "loss": 0.1195, "step": 2930 }, { "epoch": 0.14, "learning_rate": 1.86270932734587e-05, "loss": 0.1299, "step": 2935 }, { "epoch": 0.14, "learning_rate": 1.862475442043222e-05, "loss": 0.1077, "step": 2940 }, { "epoch": 0.14, "learning_rate": 1.8622415567405746e-05, "loss": 0.1104, "step": 2945 }, { "epoch": 0.14, "learning_rate": 1.862007671437927e-05, "loss": 0.1667, "step": 2950 }, { "epoch": 0.14, "learning_rate": 1.8617737861352792e-05, "loss": 0.133, "step": 2955 }, { "epoch": 0.14, "learning_rate": 1.861539900832632e-05, "loss": 0.1028, "step": 2960 }, { "epoch": 0.14, "learning_rate": 1.8613060155299842e-05, "loss": 0.1187, "step": 2965 }, { "epoch": 0.14, "learning_rate": 1.8610721302273365e-05, "loss": 0.1111, "step": 2970 }, { "epoch": 0.14, "learning_rate": 1.8608382449246892e-05, "loss": 0.1305, "step": 2975 }, { "epoch": 0.14, "learning_rate": 1.8606043596220415e-05, "loss": 0.133, "step": 2980 }, { "epoch": 0.14, "learning_rate": 1.860370474319394e-05, "loss": 0.1333, "step": 2985 }, { "epoch": 0.14, "learning_rate": 1.8601365890167465e-05, "loss": 0.1071, "step": 2990 }, { "epoch": 0.14, "learning_rate": 1.859902703714099e-05, "loss": 0.148, "step": 2995 }, { "epoch": 0.14, "learning_rate": 1.8596688184114512e-05, "loss": 0.1233, "step": 3000 }, { "epoch": 0.14, "learning_rate": 1.8594349331088035e-05, "loss": 0.0685, "step": 3005 }, { "epoch": 0.14, "learning_rate": 1.8592010478061558e-05, "loss": 0.1001, "step": 3010 }, { "epoch": 0.14, "learning_rate": 1.8589671625035085e-05, "loss": 0.1511, "step": 3015 }, { "epoch": 0.14, "learning_rate": 1.8587332772008608e-05, "loss": 0.0812, "step": 3020 }, { "epoch": 0.14, "learning_rate": 1.858499391898213e-05, "loss": 0.194, "step": 3025 }, { "epoch": 0.14, "learning_rate": 1.8582655065955658e-05, "loss": 0.1355, "step": 3030 }, { "epoch": 0.14, "learning_rate": 1.858031621292918e-05, "loss": 0.1753, "step": 3035 }, { "epoch": 0.14, "learning_rate": 1.8577977359902705e-05, "loss": 0.1098, "step": 3040 }, { "epoch": 0.14, "learning_rate": 1.857563850687623e-05, "loss": 0.1362, "step": 3045 }, { "epoch": 0.14, "learning_rate": 1.8573299653849754e-05, "loss": 0.1482, "step": 3050 }, { "epoch": 0.14, "learning_rate": 1.8570960800823278e-05, "loss": 0.0959, "step": 3055 }, { "epoch": 0.14, "learning_rate": 1.85686219477968e-05, "loss": 0.0932, "step": 3060 }, { "epoch": 0.14, "learning_rate": 1.8566283094770324e-05, "loss": 0.1285, "step": 3065 }, { "epoch": 0.14, "learning_rate": 1.856394424174385e-05, "loss": 0.1446, "step": 3070 }, { "epoch": 0.14, "learning_rate": 1.8561605388717374e-05, "loss": 0.1884, "step": 3075 }, { "epoch": 0.14, "learning_rate": 1.8559266535690897e-05, "loss": 0.141, "step": 3080 }, { "epoch": 0.14, "learning_rate": 1.8556927682664424e-05, "loss": 0.1722, "step": 3085 }, { "epoch": 0.14, "learning_rate": 1.8554588829637947e-05, "loss": 0.1066, "step": 3090 }, { "epoch": 0.14, "learning_rate": 1.855224997661147e-05, "loss": 0.0894, "step": 3095 }, { "epoch": 0.15, "learning_rate": 1.8549911123584997e-05, "loss": 0.1097, "step": 3100 }, { "epoch": 0.15, "learning_rate": 1.854757227055852e-05, "loss": 0.1214, "step": 3105 }, { "epoch": 0.15, "learning_rate": 1.8545233417532044e-05, "loss": 0.0483, "step": 3110 }, { "epoch": 0.15, "learning_rate": 1.8542894564505567e-05, "loss": 0.0953, "step": 3115 }, { "epoch": 0.15, "learning_rate": 1.854055571147909e-05, "loss": 0.1342, "step": 3120 }, { "epoch": 0.15, "learning_rate": 1.8538216858452617e-05, "loss": 0.1203, "step": 3125 }, { "epoch": 0.15, "learning_rate": 1.853587800542614e-05, "loss": 0.1318, "step": 3130 }, { "epoch": 0.15, "learning_rate": 1.8533539152399663e-05, "loss": 0.1074, "step": 3135 }, { "epoch": 0.15, "learning_rate": 1.853120029937319e-05, "loss": 0.0661, "step": 3140 }, { "epoch": 0.15, "learning_rate": 1.8528861446346713e-05, "loss": 0.1727, "step": 3145 }, { "epoch": 0.15, "learning_rate": 1.8526522593320237e-05, "loss": 0.1582, "step": 3150 }, { "epoch": 0.15, "learning_rate": 1.8524183740293763e-05, "loss": 0.1145, "step": 3155 }, { "epoch": 0.15, "learning_rate": 1.8521844887267286e-05, "loss": 0.1292, "step": 3160 }, { "epoch": 0.15, "learning_rate": 1.851950603424081e-05, "loss": 0.0479, "step": 3165 }, { "epoch": 0.15, "learning_rate": 1.8517167181214336e-05, "loss": 0.1114, "step": 3170 }, { "epoch": 0.15, "learning_rate": 1.8514828328187856e-05, "loss": 0.121, "step": 3175 }, { "epoch": 0.15, "learning_rate": 1.8512489475161383e-05, "loss": 0.1273, "step": 3180 }, { "epoch": 0.15, "learning_rate": 1.8510150622134906e-05, "loss": 0.1476, "step": 3185 }, { "epoch": 0.15, "learning_rate": 1.850781176910843e-05, "loss": 0.0929, "step": 3190 }, { "epoch": 0.15, "learning_rate": 1.8505472916081956e-05, "loss": 0.132, "step": 3195 }, { "epoch": 0.15, "learning_rate": 1.850313406305548e-05, "loss": 0.1143, "step": 3200 }, { "epoch": 0.15, "learning_rate": 1.8500795210029003e-05, "loss": 0.0828, "step": 3205 }, { "epoch": 0.15, "learning_rate": 1.849845635700253e-05, "loss": 0.1017, "step": 3210 }, { "epoch": 0.15, "learning_rate": 1.8496117503976052e-05, "loss": 0.1337, "step": 3215 }, { "epoch": 0.15, "learning_rate": 1.8493778650949576e-05, "loss": 0.1269, "step": 3220 }, { "epoch": 0.15, "learning_rate": 1.8491439797923102e-05, "loss": 0.1335, "step": 3225 }, { "epoch": 0.15, "learning_rate": 1.8489100944896626e-05, "loss": 0.1117, "step": 3230 }, { "epoch": 0.15, "learning_rate": 1.848676209187015e-05, "loss": 0.1362, "step": 3235 }, { "epoch": 0.15, "learning_rate": 1.8484423238843672e-05, "loss": 0.1031, "step": 3240 }, { "epoch": 0.15, "learning_rate": 1.8482084385817195e-05, "loss": 0.1365, "step": 3245 }, { "epoch": 0.15, "learning_rate": 1.8479745532790722e-05, "loss": 0.0931, "step": 3250 }, { "epoch": 0.15, "learning_rate": 1.8477406679764245e-05, "loss": 0.1161, "step": 3255 }, { "epoch": 0.15, "learning_rate": 1.847506782673777e-05, "loss": 0.1133, "step": 3260 }, { "epoch": 0.15, "learning_rate": 1.8472728973711295e-05, "loss": 0.1278, "step": 3265 }, { "epoch": 0.15, "learning_rate": 1.847039012068482e-05, "loss": 0.1973, "step": 3270 }, { "epoch": 0.15, "learning_rate": 1.8468051267658342e-05, "loss": 0.1248, "step": 3275 }, { "epoch": 0.15, "learning_rate": 1.846571241463187e-05, "loss": 0.1644, "step": 3280 }, { "epoch": 0.15, "learning_rate": 1.846337356160539e-05, "loss": 0.063, "step": 3285 }, { "epoch": 0.15, "learning_rate": 1.8461034708578915e-05, "loss": 0.1047, "step": 3290 }, { "epoch": 0.15, "learning_rate": 1.8458695855552438e-05, "loss": 0.1553, "step": 3295 }, { "epoch": 0.15, "learning_rate": 1.845635700252596e-05, "loss": 0.0467, "step": 3300 }, { "epoch": 0.15, "learning_rate": 1.8454018149499485e-05, "loss": 0.1385, "step": 3305 }, { "epoch": 0.15, "learning_rate": 1.845167929647301e-05, "loss": 0.1622, "step": 3310 }, { "epoch": 0.16, "learning_rate": 1.8449340443446535e-05, "loss": 0.0839, "step": 3315 }, { "epoch": 0.16, "learning_rate": 1.8447001590420058e-05, "loss": 0.0869, "step": 3320 }, { "epoch": 0.16, "learning_rate": 1.8444662737393584e-05, "loss": 0.1341, "step": 3325 }, { "epoch": 0.16, "learning_rate": 1.8442323884367108e-05, "loss": 0.2328, "step": 3330 }, { "epoch": 0.16, "learning_rate": 1.843998503134063e-05, "loss": 0.1192, "step": 3335 }, { "epoch": 0.16, "learning_rate": 1.8437646178314158e-05, "loss": 0.1716, "step": 3340 }, { "epoch": 0.16, "learning_rate": 1.843530732528768e-05, "loss": 0.1071, "step": 3345 }, { "epoch": 0.16, "learning_rate": 1.8432968472261204e-05, "loss": 0.1357, "step": 3350 }, { "epoch": 0.16, "learning_rate": 1.8430629619234727e-05, "loss": 0.1985, "step": 3355 }, { "epoch": 0.16, "learning_rate": 1.842829076620825e-05, "loss": 0.1238, "step": 3360 }, { "epoch": 0.16, "learning_rate": 1.8425951913181777e-05, "loss": 0.1297, "step": 3365 }, { "epoch": 0.16, "learning_rate": 1.84236130601553e-05, "loss": 0.1447, "step": 3370 }, { "epoch": 0.16, "learning_rate": 1.8421274207128824e-05, "loss": 0.1107, "step": 3375 }, { "epoch": 0.16, "learning_rate": 1.841893535410235e-05, "loss": 0.1642, "step": 3380 }, { "epoch": 0.16, "learning_rate": 1.8416596501075874e-05, "loss": 0.1383, "step": 3385 }, { "epoch": 0.16, "learning_rate": 1.8414257648049397e-05, "loss": 0.1196, "step": 3390 }, { "epoch": 0.16, "learning_rate": 1.8411918795022924e-05, "loss": 0.1293, "step": 3395 }, { "epoch": 0.16, "learning_rate": 1.8409579941996447e-05, "loss": 0.1155, "step": 3400 }, { "epoch": 0.16, "learning_rate": 1.840724108896997e-05, "loss": 0.1308, "step": 3405 }, { "epoch": 0.16, "learning_rate": 1.8404902235943497e-05, "loss": 0.0783, "step": 3410 }, { "epoch": 0.16, "learning_rate": 1.8402563382917017e-05, "loss": 0.121, "step": 3415 }, { "epoch": 0.16, "learning_rate": 1.8400224529890543e-05, "loss": 0.1285, "step": 3420 }, { "epoch": 0.16, "learning_rate": 1.8397885676864067e-05, "loss": 0.0908, "step": 3425 }, { "epoch": 0.16, "learning_rate": 1.839554682383759e-05, "loss": 0.2183, "step": 3430 }, { "epoch": 0.16, "learning_rate": 1.8393207970811116e-05, "loss": 0.1148, "step": 3435 }, { "epoch": 0.16, "learning_rate": 1.839086911778464e-05, "loss": 0.1198, "step": 3440 }, { "epoch": 0.16, "learning_rate": 1.8388530264758163e-05, "loss": 0.1035, "step": 3445 }, { "epoch": 0.16, "learning_rate": 1.838619141173169e-05, "loss": 0.1127, "step": 3450 }, { "epoch": 0.16, "learning_rate": 1.8383852558705213e-05, "loss": 0.135, "step": 3455 }, { "epoch": 0.16, "learning_rate": 1.8381513705678736e-05, "loss": 0.1423, "step": 3460 }, { "epoch": 0.16, "learning_rate": 1.8379174852652263e-05, "loss": 0.1079, "step": 3465 }, { "epoch": 0.16, "learning_rate": 1.8376835999625786e-05, "loss": 0.19, "step": 3470 }, { "epoch": 0.16, "learning_rate": 1.837449714659931e-05, "loss": 0.1258, "step": 3475 }, { "epoch": 0.16, "learning_rate": 1.8372158293572833e-05, "loss": 0.1984, "step": 3480 }, { "epoch": 0.16, "learning_rate": 1.8369819440546356e-05, "loss": 0.1469, "step": 3485 }, { "epoch": 0.16, "learning_rate": 1.8367480587519882e-05, "loss": 0.1605, "step": 3490 }, { "epoch": 0.16, "learning_rate": 1.8365141734493406e-05, "loss": 0.0779, "step": 3495 }, { "epoch": 0.16, "learning_rate": 1.836280288146693e-05, "loss": 0.1357, "step": 3500 }, { "epoch": 0.16, "learning_rate": 1.8360464028440456e-05, "loss": 0.056, "step": 3505 }, { "epoch": 0.16, "learning_rate": 1.835812517541398e-05, "loss": 0.0809, "step": 3510 }, { "epoch": 0.16, "learning_rate": 1.8355786322387502e-05, "loss": 0.1556, "step": 3515 }, { "epoch": 0.16, "learning_rate": 1.835344746936103e-05, "loss": 0.1075, "step": 3520 }, { "epoch": 0.16, "learning_rate": 1.8351108616334552e-05, "loss": 0.0716, "step": 3525 }, { "epoch": 0.17, "learning_rate": 1.8348769763308075e-05, "loss": 0.1385, "step": 3530 }, { "epoch": 0.17, "learning_rate": 1.83464309102816e-05, "loss": 0.1447, "step": 3535 }, { "epoch": 0.17, "learning_rate": 1.8344092057255122e-05, "loss": 0.103, "step": 3540 }, { "epoch": 0.17, "learning_rate": 1.834175320422865e-05, "loss": 0.1616, "step": 3545 }, { "epoch": 0.17, "learning_rate": 1.8339414351202172e-05, "loss": 0.1023, "step": 3550 }, { "epoch": 0.17, "learning_rate": 1.8337075498175695e-05, "loss": 0.1414, "step": 3555 }, { "epoch": 0.17, "learning_rate": 1.833473664514922e-05, "loss": 0.1268, "step": 3560 }, { "epoch": 0.17, "learning_rate": 1.8332397792122745e-05, "loss": 0.1379, "step": 3565 }, { "epoch": 0.17, "learning_rate": 1.8330058939096268e-05, "loss": 0.0781, "step": 3570 }, { "epoch": 0.17, "learning_rate": 1.8327720086069795e-05, "loss": 0.1641, "step": 3575 }, { "epoch": 0.17, "learning_rate": 1.8325381233043318e-05, "loss": 0.0812, "step": 3580 }, { "epoch": 0.17, "learning_rate": 1.832304238001684e-05, "loss": 0.1015, "step": 3585 }, { "epoch": 0.17, "learning_rate": 1.8320703526990365e-05, "loss": 0.0848, "step": 3590 }, { "epoch": 0.17, "learning_rate": 1.8318364673963888e-05, "loss": 0.1228, "step": 3595 }, { "epoch": 0.17, "learning_rate": 1.8316025820937414e-05, "loss": 0.081, "step": 3600 }, { "epoch": 0.17, "learning_rate": 1.8313686967910938e-05, "loss": 0.1635, "step": 3605 }, { "epoch": 0.17, "learning_rate": 1.831134811488446e-05, "loss": 0.1284, "step": 3610 }, { "epoch": 0.17, "learning_rate": 1.8309009261857988e-05, "loss": 0.1373, "step": 3615 }, { "epoch": 0.17, "learning_rate": 1.830667040883151e-05, "loss": 0.1296, "step": 3620 }, { "epoch": 0.17, "learning_rate": 1.8304331555805034e-05, "loss": 0.1537, "step": 3625 }, { "epoch": 0.17, "learning_rate": 1.830199270277856e-05, "loss": 0.1356, "step": 3630 }, { "epoch": 0.17, "learning_rate": 1.8299653849752084e-05, "loss": 0.1054, "step": 3635 }, { "epoch": 0.17, "learning_rate": 1.8297314996725607e-05, "loss": 0.1219, "step": 3640 }, { "epoch": 0.17, "learning_rate": 1.8294976143699134e-05, "loss": 0.1388, "step": 3645 }, { "epoch": 0.17, "learning_rate": 1.8292637290672654e-05, "loss": 0.1088, "step": 3650 }, { "epoch": 0.17, "learning_rate": 1.829029843764618e-05, "loss": 0.1289, "step": 3655 }, { "epoch": 0.17, "learning_rate": 1.8287959584619704e-05, "loss": 0.155, "step": 3660 }, { "epoch": 0.17, "learning_rate": 1.8285620731593227e-05, "loss": 0.0771, "step": 3665 }, { "epoch": 0.17, "learning_rate": 1.8283281878566754e-05, "loss": 0.1082, "step": 3670 }, { "epoch": 0.17, "learning_rate": 1.8280943025540277e-05, "loss": 0.1152, "step": 3675 }, { "epoch": 0.17, "learning_rate": 1.82786041725138e-05, "loss": 0.0644, "step": 3680 }, { "epoch": 0.17, "learning_rate": 1.8276265319487327e-05, "loss": 0.1141, "step": 3685 }, { "epoch": 0.17, "learning_rate": 1.827392646646085e-05, "loss": 0.1753, "step": 3690 }, { "epoch": 0.17, "learning_rate": 1.8271587613434373e-05, "loss": 0.1112, "step": 3695 }, { "epoch": 0.17, "learning_rate": 1.82692487604079e-05, "loss": 0.1086, "step": 3700 }, { "epoch": 0.17, "learning_rate": 1.8266909907381423e-05, "loss": 0.1168, "step": 3705 }, { "epoch": 0.17, "learning_rate": 1.8264571054354946e-05, "loss": 0.1288, "step": 3710 }, { "epoch": 0.17, "learning_rate": 1.826223220132847e-05, "loss": 0.1047, "step": 3715 }, { "epoch": 0.17, "learning_rate": 1.8259893348301993e-05, "loss": 0.0559, "step": 3720 }, { "epoch": 0.17, "learning_rate": 1.825755449527552e-05, "loss": 0.1355, "step": 3725 }, { "epoch": 0.17, "learning_rate": 1.8255215642249043e-05, "loss": 0.0669, "step": 3730 }, { "epoch": 0.17, "learning_rate": 1.8252876789222566e-05, "loss": 0.0687, "step": 3735 }, { "epoch": 0.17, "learning_rate": 1.825053793619609e-05, "loss": 0.1592, "step": 3740 }, { "epoch": 0.18, "learning_rate": 1.8248199083169616e-05, "loss": 0.0916, "step": 3745 }, { "epoch": 0.18, "learning_rate": 1.824586023014314e-05, "loss": 0.1259, "step": 3750 }, { "epoch": 0.18, "learning_rate": 1.8243521377116663e-05, "loss": 0.1585, "step": 3755 }, { "epoch": 0.18, "learning_rate": 1.824118252409019e-05, "loss": 0.0993, "step": 3760 }, { "epoch": 0.18, "learning_rate": 1.8238843671063712e-05, "loss": 0.1076, "step": 3765 }, { "epoch": 0.18, "learning_rate": 1.8236504818037236e-05, "loss": 0.1587, "step": 3770 }, { "epoch": 0.18, "learning_rate": 1.823416596501076e-05, "loss": 0.1403, "step": 3775 }, { "epoch": 0.18, "learning_rate": 1.8231827111984282e-05, "loss": 0.0644, "step": 3780 }, { "epoch": 0.18, "learning_rate": 1.822948825895781e-05, "loss": 0.1536, "step": 3785 }, { "epoch": 0.18, "learning_rate": 1.8227149405931332e-05, "loss": 0.1252, "step": 3790 }, { "epoch": 0.18, "learning_rate": 1.8224810552904855e-05, "loss": 0.082, "step": 3795 }, { "epoch": 0.18, "learning_rate": 1.8222471699878382e-05, "loss": 0.1657, "step": 3800 }, { "epoch": 0.18, "learning_rate": 1.8220132846851905e-05, "loss": 0.1385, "step": 3805 }, { "epoch": 0.18, "learning_rate": 1.821779399382543e-05, "loss": 0.1064, "step": 3810 }, { "epoch": 0.18, "learning_rate": 1.8215455140798955e-05, "loss": 0.1243, "step": 3815 }, { "epoch": 0.18, "learning_rate": 1.821311628777248e-05, "loss": 0.093, "step": 3820 }, { "epoch": 0.18, "learning_rate": 1.8210777434746e-05, "loss": 0.1143, "step": 3825 }, { "epoch": 0.18, "learning_rate": 1.8208438581719525e-05, "loss": 0.1347, "step": 3830 }, { "epoch": 0.18, "learning_rate": 1.8206099728693048e-05, "loss": 0.0832, "step": 3835 }, { "epoch": 0.18, "learning_rate": 1.8203760875666575e-05, "loss": 0.1635, "step": 3840 }, { "epoch": 0.18, "learning_rate": 1.8201422022640098e-05, "loss": 0.0989, "step": 3845 }, { "epoch": 0.18, "learning_rate": 1.819908316961362e-05, "loss": 0.1277, "step": 3850 }, { "epoch": 0.18, "learning_rate": 1.8196744316587148e-05, "loss": 0.0732, "step": 3855 }, { "epoch": 0.18, "learning_rate": 1.819440546356067e-05, "loss": 0.0971, "step": 3860 }, { "epoch": 0.18, "learning_rate": 1.8192066610534195e-05, "loss": 0.081, "step": 3865 }, { "epoch": 0.18, "learning_rate": 1.818972775750772e-05, "loss": 0.1705, "step": 3870 }, { "epoch": 0.18, "learning_rate": 1.8187388904481244e-05, "loss": 0.1299, "step": 3875 }, { "epoch": 0.18, "learning_rate": 1.8185050051454768e-05, "loss": 0.1243, "step": 3880 }, { "epoch": 0.18, "learning_rate": 1.8182711198428294e-05, "loss": 0.1392, "step": 3885 }, { "epoch": 0.18, "learning_rate": 1.8180372345401814e-05, "loss": 0.1307, "step": 3890 }, { "epoch": 0.18, "learning_rate": 1.817803349237534e-05, "loss": 0.1534, "step": 3895 }, { "epoch": 0.18, "learning_rate": 1.8175694639348864e-05, "loss": 0.1598, "step": 3900 }, { "epoch": 0.18, "learning_rate": 1.8173355786322387e-05, "loss": 0.0835, "step": 3905 }, { "epoch": 0.18, "learning_rate": 1.8171016933295914e-05, "loss": 0.1285, "step": 3910 }, { "epoch": 0.18, "learning_rate": 1.8168678080269437e-05, "loss": 0.1664, "step": 3915 }, { "epoch": 0.18, "learning_rate": 1.816633922724296e-05, "loss": 0.1304, "step": 3920 }, { "epoch": 0.18, "learning_rate": 1.8164000374216487e-05, "loss": 0.0643, "step": 3925 }, { "epoch": 0.18, "learning_rate": 1.816166152119001e-05, "loss": 0.1228, "step": 3930 }, { "epoch": 0.18, "learning_rate": 1.8159322668163534e-05, "loss": 0.0916, "step": 3935 }, { "epoch": 0.18, "learning_rate": 1.815698381513706e-05, "loss": 0.0957, "step": 3940 }, { "epoch": 0.18, "learning_rate": 1.815464496211058e-05, "loss": 0.0692, "step": 3945 }, { "epoch": 0.18, "learning_rate": 1.8152306109084107e-05, "loss": 0.174, "step": 3950 }, { "epoch": 0.18, "learning_rate": 1.814996725605763e-05, "loss": 0.1266, "step": 3955 }, { "epoch": 0.19, "learning_rate": 1.8147628403031153e-05, "loss": 0.1465, "step": 3960 }, { "epoch": 0.19, "learning_rate": 1.814528955000468e-05, "loss": 0.1487, "step": 3965 }, { "epoch": 0.19, "learning_rate": 1.8142950696978203e-05, "loss": 0.0847, "step": 3970 }, { "epoch": 0.19, "learning_rate": 1.8140611843951727e-05, "loss": 0.1383, "step": 3975 }, { "epoch": 0.19, "learning_rate": 1.8138272990925253e-05, "loss": 0.169, "step": 3980 }, { "epoch": 0.19, "learning_rate": 1.8135934137898776e-05, "loss": 0.127, "step": 3985 }, { "epoch": 0.19, "learning_rate": 1.81335952848723e-05, "loss": 0.067, "step": 3990 }, { "epoch": 0.19, "learning_rate": 1.8131256431845826e-05, "loss": 0.124, "step": 3995 }, { "epoch": 0.19, "learning_rate": 1.812891757881935e-05, "loss": 0.1234, "step": 4000 }, { "epoch": 0.19, "learning_rate": 1.8126578725792873e-05, "loss": 0.1561, "step": 4005 }, { "epoch": 0.19, "learning_rate": 1.8124239872766396e-05, "loss": 0.1503, "step": 4010 }, { "epoch": 0.19, "learning_rate": 1.812190101973992e-05, "loss": 0.119, "step": 4015 }, { "epoch": 0.19, "learning_rate": 1.8119562166713446e-05, "loss": 0.1357, "step": 4020 }, { "epoch": 0.19, "learning_rate": 1.811722331368697e-05, "loss": 0.0933, "step": 4025 }, { "epoch": 0.19, "learning_rate": 1.8114884460660493e-05, "loss": 0.1375, "step": 4030 }, { "epoch": 0.19, "learning_rate": 1.811254560763402e-05, "loss": 0.1049, "step": 4035 }, { "epoch": 0.19, "learning_rate": 1.8110206754607542e-05, "loss": 0.1056, "step": 4040 }, { "epoch": 0.19, "learning_rate": 1.8107867901581066e-05, "loss": 0.2034, "step": 4045 }, { "epoch": 0.19, "learning_rate": 1.8105529048554592e-05, "loss": 0.1122, "step": 4050 }, { "epoch": 0.19, "learning_rate": 1.8103190195528116e-05, "loss": 0.1135, "step": 4055 }, { "epoch": 0.19, "learning_rate": 1.810085134250164e-05, "loss": 0.1326, "step": 4060 }, { "epoch": 0.19, "learning_rate": 1.8098512489475162e-05, "loss": 0.1629, "step": 4065 }, { "epoch": 0.19, "learning_rate": 1.8096173636448685e-05, "loss": 0.1253, "step": 4070 }, { "epoch": 0.19, "learning_rate": 1.8093834783422212e-05, "loss": 0.1617, "step": 4075 }, { "epoch": 0.19, "learning_rate": 1.8091495930395735e-05, "loss": 0.1193, "step": 4080 }, { "epoch": 0.19, "learning_rate": 1.808915707736926e-05, "loss": 0.0785, "step": 4085 }, { "epoch": 0.19, "learning_rate": 1.8086818224342785e-05, "loss": 0.0822, "step": 4090 }, { "epoch": 0.19, "learning_rate": 1.808447937131631e-05, "loss": 0.1377, "step": 4095 }, { "epoch": 0.19, "learning_rate": 1.808214051828983e-05, "loss": 0.1525, "step": 4100 }, { "epoch": 0.19, "learning_rate": 1.807980166526336e-05, "loss": 0.1234, "step": 4105 }, { "epoch": 0.19, "learning_rate": 1.807746281223688e-05, "loss": 0.1394, "step": 4110 }, { "epoch": 0.19, "learning_rate": 1.8075123959210405e-05, "loss": 0.1008, "step": 4115 }, { "epoch": 0.19, "learning_rate": 1.807278510618393e-05, "loss": 0.0873, "step": 4120 }, { "epoch": 0.19, "learning_rate": 1.807044625315745e-05, "loss": 0.0969, "step": 4125 }, { "epoch": 0.19, "learning_rate": 1.8068107400130978e-05, "loss": 0.1186, "step": 4130 }, { "epoch": 0.19, "learning_rate": 1.80657685471045e-05, "loss": 0.0683, "step": 4135 }, { "epoch": 0.19, "learning_rate": 1.8063429694078025e-05, "loss": 0.1297, "step": 4140 }, { "epoch": 0.19, "learning_rate": 1.806109084105155e-05, "loss": 0.0833, "step": 4145 }, { "epoch": 0.19, "learning_rate": 1.8058751988025074e-05, "loss": 0.1274, "step": 4150 }, { "epoch": 0.19, "learning_rate": 1.8056413134998598e-05, "loss": 0.1006, "step": 4155 }, { "epoch": 0.19, "learning_rate": 1.8054074281972124e-05, "loss": 0.1563, "step": 4160 }, { "epoch": 0.19, "learning_rate": 1.8051735428945648e-05, "loss": 0.1737, "step": 4165 }, { "epoch": 0.2, "learning_rate": 1.804939657591917e-05, "loss": 0.0953, "step": 4170 }, { "epoch": 0.2, "learning_rate": 1.8047057722892694e-05, "loss": 0.1262, "step": 4175 }, { "epoch": 0.2, "learning_rate": 1.804471886986622e-05, "loss": 0.1112, "step": 4180 }, { "epoch": 0.2, "learning_rate": 1.804238001683974e-05, "loss": 0.1112, "step": 4185 }, { "epoch": 0.2, "learning_rate": 1.8040041163813267e-05, "loss": 0.1493, "step": 4190 }, { "epoch": 0.2, "learning_rate": 1.803770231078679e-05, "loss": 0.0781, "step": 4195 }, { "epoch": 0.2, "learning_rate": 1.8035363457760314e-05, "loss": 0.2134, "step": 4200 }, { "epoch": 0.2, "learning_rate": 1.803302460473384e-05, "loss": 0.1373, "step": 4205 }, { "epoch": 0.2, "learning_rate": 1.8030685751707364e-05, "loss": 0.1049, "step": 4210 }, { "epoch": 0.2, "learning_rate": 1.8028346898680887e-05, "loss": 0.1257, "step": 4215 }, { "epoch": 0.2, "learning_rate": 1.8026008045654414e-05, "loss": 0.1058, "step": 4220 }, { "epoch": 0.2, "learning_rate": 1.8023669192627937e-05, "loss": 0.1033, "step": 4225 }, { "epoch": 0.2, "learning_rate": 1.802133033960146e-05, "loss": 0.1469, "step": 4230 }, { "epoch": 0.2, "learning_rate": 1.8018991486574987e-05, "loss": 0.0971, "step": 4235 }, { "epoch": 0.2, "learning_rate": 1.801665263354851e-05, "loss": 0.1188, "step": 4240 }, { "epoch": 0.2, "learning_rate": 1.8014313780522033e-05, "loss": 0.1392, "step": 4245 }, { "epoch": 0.2, "learning_rate": 1.8011974927495556e-05, "loss": 0.097, "step": 4250 }, { "epoch": 0.2, "learning_rate": 1.800963607446908e-05, "loss": 0.0957, "step": 4255 }, { "epoch": 0.2, "learning_rate": 1.8007297221442606e-05, "loss": 0.0916, "step": 4260 }, { "epoch": 0.2, "learning_rate": 1.800495836841613e-05, "loss": 0.0556, "step": 4265 }, { "epoch": 0.2, "learning_rate": 1.8002619515389653e-05, "loss": 0.1311, "step": 4270 }, { "epoch": 0.2, "learning_rate": 1.800028066236318e-05, "loss": 0.1505, "step": 4275 }, { "epoch": 0.2, "learning_rate": 1.7997941809336703e-05, "loss": 0.0575, "step": 4280 }, { "epoch": 0.2, "learning_rate": 1.7995602956310226e-05, "loss": 0.122, "step": 4285 }, { "epoch": 0.2, "learning_rate": 1.7993264103283753e-05, "loss": 0.1375, "step": 4290 }, { "epoch": 0.2, "learning_rate": 1.7990925250257276e-05, "loss": 0.0841, "step": 4295 }, { "epoch": 0.2, "learning_rate": 1.79885863972308e-05, "loss": 0.1127, "step": 4300 }, { "epoch": 0.2, "learning_rate": 1.7986247544204322e-05, "loss": 0.1274, "step": 4305 }, { "epoch": 0.2, "learning_rate": 1.7983908691177846e-05, "loss": 0.1248, "step": 4310 }, { "epoch": 0.2, "learning_rate": 1.7981569838151372e-05, "loss": 0.0843, "step": 4315 }, { "epoch": 0.2, "learning_rate": 1.7979230985124896e-05, "loss": 0.1499, "step": 4320 }, { "epoch": 0.2, "learning_rate": 1.797689213209842e-05, "loss": 0.1764, "step": 4325 }, { "epoch": 0.2, "learning_rate": 1.7974553279071946e-05, "loss": 0.1389, "step": 4330 }, { "epoch": 0.2, "learning_rate": 1.797221442604547e-05, "loss": 0.1168, "step": 4335 }, { "epoch": 0.2, "learning_rate": 1.7969875573018992e-05, "loss": 0.1899, "step": 4340 }, { "epoch": 0.2, "learning_rate": 1.796753671999252e-05, "loss": 0.0941, "step": 4345 }, { "epoch": 0.2, "learning_rate": 1.7965197866966042e-05, "loss": 0.1663, "step": 4350 }, { "epoch": 0.2, "learning_rate": 1.7962859013939565e-05, "loss": 0.129, "step": 4355 }, { "epoch": 0.2, "learning_rate": 1.796052016091309e-05, "loss": 0.0811, "step": 4360 }, { "epoch": 0.2, "learning_rate": 1.7958181307886612e-05, "loss": 0.1711, "step": 4365 }, { "epoch": 0.2, "learning_rate": 1.795584245486014e-05, "loss": 0.1424, "step": 4370 }, { "epoch": 0.2, "learning_rate": 1.795350360183366e-05, "loss": 0.0968, "step": 4375 }, { "epoch": 0.2, "learning_rate": 1.7951164748807185e-05, "loss": 0.1495, "step": 4380 }, { "epoch": 0.21, "learning_rate": 1.794882589578071e-05, "loss": 0.0733, "step": 4385 }, { "epoch": 0.21, "learning_rate": 1.7946487042754235e-05, "loss": 0.0997, "step": 4390 }, { "epoch": 0.21, "learning_rate": 1.7944148189727758e-05, "loss": 0.1032, "step": 4395 }, { "epoch": 0.21, "learning_rate": 1.7941809336701285e-05, "loss": 0.0798, "step": 4400 }, { "epoch": 0.21, "learning_rate": 1.7939470483674808e-05, "loss": 0.1015, "step": 4405 }, { "epoch": 0.21, "learning_rate": 1.793713163064833e-05, "loss": 0.1311, "step": 4410 }, { "epoch": 0.21, "learning_rate": 1.7934792777621858e-05, "loss": 0.1128, "step": 4415 }, { "epoch": 0.21, "learning_rate": 1.7932453924595378e-05, "loss": 0.149, "step": 4420 }, { "epoch": 0.21, "learning_rate": 1.7930115071568904e-05, "loss": 0.1305, "step": 4425 }, { "epoch": 0.21, "learning_rate": 1.7927776218542428e-05, "loss": 0.115, "step": 4430 }, { "epoch": 0.21, "learning_rate": 1.792543736551595e-05, "loss": 0.0927, "step": 4435 }, { "epoch": 0.21, "learning_rate": 1.7923098512489478e-05, "loss": 0.1078, "step": 4440 }, { "epoch": 0.21, "learning_rate": 1.7920759659463e-05, "loss": 0.0896, "step": 4445 }, { "epoch": 0.21, "learning_rate": 1.7918420806436524e-05, "loss": 0.0739, "step": 4450 }, { "epoch": 0.21, "learning_rate": 1.791608195341005e-05, "loss": 0.0919, "step": 4455 }, { "epoch": 0.21, "learning_rate": 1.7913743100383574e-05, "loss": 0.1123, "step": 4460 }, { "epoch": 0.21, "learning_rate": 1.7911404247357097e-05, "loss": 0.1644, "step": 4465 }, { "epoch": 0.21, "learning_rate": 1.7909065394330624e-05, "loss": 0.0923, "step": 4470 }, { "epoch": 0.21, "learning_rate": 1.7906726541304147e-05, "loss": 0.0952, "step": 4475 }, { "epoch": 0.21, "learning_rate": 1.790438768827767e-05, "loss": 0.1195, "step": 4480 }, { "epoch": 0.21, "learning_rate": 1.7902048835251194e-05, "loss": 0.1171, "step": 4485 }, { "epoch": 0.21, "learning_rate": 1.7899709982224717e-05, "loss": 0.1275, "step": 4490 }, { "epoch": 0.21, "learning_rate": 1.7897371129198244e-05, "loss": 0.1636, "step": 4495 }, { "epoch": 0.21, "learning_rate": 1.7895032276171767e-05, "loss": 0.1268, "step": 4500 }, { "epoch": 0.21, "learning_rate": 1.789269342314529e-05, "loss": 0.1569, "step": 4505 }, { "epoch": 0.21, "learning_rate": 1.7890354570118817e-05, "loss": 0.1515, "step": 4510 }, { "epoch": 0.21, "learning_rate": 1.788801571709234e-05, "loss": 0.15, "step": 4515 }, { "epoch": 0.21, "learning_rate": 1.7885676864065863e-05, "loss": 0.0983, "step": 4520 }, { "epoch": 0.21, "learning_rate": 1.788333801103939e-05, "loss": 0.1313, "step": 4525 }, { "epoch": 0.21, "learning_rate": 1.7880999158012913e-05, "loss": 0.1725, "step": 4530 }, { "epoch": 0.21, "learning_rate": 1.7878660304986436e-05, "loss": 0.1258, "step": 4535 }, { "epoch": 0.21, "learning_rate": 1.787632145195996e-05, "loss": 0.1287, "step": 4540 }, { "epoch": 0.21, "learning_rate": 1.7873982598933483e-05, "loss": 0.1125, "step": 4545 }, { "epoch": 0.21, "learning_rate": 1.787164374590701e-05, "loss": 0.1579, "step": 4550 }, { "epoch": 0.21, "learning_rate": 1.7869304892880533e-05, "loss": 0.1188, "step": 4555 }, { "epoch": 0.21, "learning_rate": 1.7866966039854056e-05, "loss": 0.1489, "step": 4560 }, { "epoch": 0.21, "learning_rate": 1.7864627186827583e-05, "loss": 0.081, "step": 4565 }, { "epoch": 0.21, "learning_rate": 1.7862288333801106e-05, "loss": 0.1369, "step": 4570 }, { "epoch": 0.21, "learning_rate": 1.785994948077463e-05, "loss": 0.1077, "step": 4575 }, { "epoch": 0.21, "learning_rate": 1.7857610627748156e-05, "loss": 0.138, "step": 4580 }, { "epoch": 0.21, "learning_rate": 1.785527177472168e-05, "loss": 0.1042, "step": 4585 }, { "epoch": 0.21, "learning_rate": 1.7852932921695202e-05, "loss": 0.1259, "step": 4590 }, { "epoch": 0.21, "learning_rate": 1.785059406866873e-05, "loss": 0.1924, "step": 4595 }, { "epoch": 0.22, "learning_rate": 1.784825521564225e-05, "loss": 0.1183, "step": 4600 }, { "epoch": 0.22, "learning_rate": 1.7845916362615776e-05, "loss": 0.1184, "step": 4605 }, { "epoch": 0.22, "learning_rate": 1.78435775095893e-05, "loss": 0.0826, "step": 4610 }, { "epoch": 0.22, "learning_rate": 1.7841238656562822e-05, "loss": 0.1225, "step": 4615 }, { "epoch": 0.22, "learning_rate": 1.7838899803536345e-05, "loss": 0.1492, "step": 4620 }, { "epoch": 0.22, "learning_rate": 1.7836560950509872e-05, "loss": 0.1043, "step": 4625 }, { "epoch": 0.22, "learning_rate": 1.7834222097483395e-05, "loss": 0.0866, "step": 4630 }, { "epoch": 0.22, "learning_rate": 1.783188324445692e-05, "loss": 0.1255, "step": 4635 }, { "epoch": 0.22, "learning_rate": 1.7829544391430445e-05, "loss": 0.1031, "step": 4640 }, { "epoch": 0.22, "learning_rate": 1.782720553840397e-05, "loss": 0.1019, "step": 4645 }, { "epoch": 0.22, "learning_rate": 1.782486668537749e-05, "loss": 0.0695, "step": 4650 }, { "epoch": 0.22, "learning_rate": 1.7822527832351018e-05, "loss": 0.0578, "step": 4655 }, { "epoch": 0.22, "learning_rate": 1.7820188979324538e-05, "loss": 0.1347, "step": 4660 }, { "epoch": 0.22, "learning_rate": 1.7817850126298065e-05, "loss": 0.0815, "step": 4665 }, { "epoch": 0.22, "learning_rate": 1.7815511273271588e-05, "loss": 0.0806, "step": 4670 }, { "epoch": 0.22, "learning_rate": 1.781317242024511e-05, "loss": 0.1466, "step": 4675 }, { "epoch": 0.22, "learning_rate": 1.7810833567218638e-05, "loss": 0.1098, "step": 4680 }, { "epoch": 0.22, "learning_rate": 1.780849471419216e-05, "loss": 0.0972, "step": 4685 }, { "epoch": 0.22, "learning_rate": 1.7806155861165684e-05, "loss": 0.1361, "step": 4690 }, { "epoch": 0.22, "learning_rate": 1.780381700813921e-05, "loss": 0.0688, "step": 4695 }, { "epoch": 0.22, "learning_rate": 1.7801478155112734e-05, "loss": 0.1728, "step": 4700 }, { "epoch": 0.22, "learning_rate": 1.7799139302086258e-05, "loss": 0.1742, "step": 4705 }, { "epoch": 0.22, "learning_rate": 1.7796800449059784e-05, "loss": 0.1282, "step": 4710 }, { "epoch": 0.22, "learning_rate": 1.7794461596033308e-05, "loss": 0.0982, "step": 4715 }, { "epoch": 0.22, "learning_rate": 1.779212274300683e-05, "loss": 0.1703, "step": 4720 }, { "epoch": 0.22, "learning_rate": 1.7789783889980354e-05, "loss": 0.0942, "step": 4725 }, { "epoch": 0.22, "learning_rate": 1.7787445036953877e-05, "loss": 0.09, "step": 4730 }, { "epoch": 0.22, "learning_rate": 1.7785106183927404e-05, "loss": 0.1237, "step": 4735 }, { "epoch": 0.22, "learning_rate": 1.7782767330900927e-05, "loss": 0.1577, "step": 4740 }, { "epoch": 0.22, "learning_rate": 1.778042847787445e-05, "loss": 0.1252, "step": 4745 }, { "epoch": 0.22, "learning_rate": 1.7778089624847977e-05, "loss": 0.093, "step": 4750 }, { "epoch": 0.22, "learning_rate": 1.77757507718215e-05, "loss": 0.1253, "step": 4755 }, { "epoch": 0.22, "learning_rate": 1.7773411918795024e-05, "loss": 0.1586, "step": 4760 }, { "epoch": 0.22, "learning_rate": 1.777107306576855e-05, "loss": 0.1646, "step": 4765 }, { "epoch": 0.22, "learning_rate": 1.7768734212742074e-05, "loss": 0.1328, "step": 4770 }, { "epoch": 0.22, "learning_rate": 1.7766395359715597e-05, "loss": 0.1065, "step": 4775 }, { "epoch": 0.22, "learning_rate": 1.776405650668912e-05, "loss": 0.1695, "step": 4780 }, { "epoch": 0.22, "learning_rate": 1.7761717653662643e-05, "loss": 0.1317, "step": 4785 }, { "epoch": 0.22, "learning_rate": 1.775937880063617e-05, "loss": 0.1162, "step": 4790 }, { "epoch": 0.22, "learning_rate": 1.7757039947609693e-05, "loss": 0.1261, "step": 4795 }, { "epoch": 0.22, "learning_rate": 1.7754701094583216e-05, "loss": 0.0726, "step": 4800 }, { "epoch": 0.22, "learning_rate": 1.7752362241556743e-05, "loss": 0.1409, "step": 4805 }, { "epoch": 0.22, "learning_rate": 1.7750023388530266e-05, "loss": 0.0454, "step": 4810 }, { "epoch": 0.23, "learning_rate": 1.774768453550379e-05, "loss": 0.1701, "step": 4815 }, { "epoch": 0.23, "learning_rate": 1.7745345682477316e-05, "loss": 0.1057, "step": 4820 }, { "epoch": 0.23, "learning_rate": 1.774300682945084e-05, "loss": 0.1143, "step": 4825 }, { "epoch": 0.23, "learning_rate": 1.7740667976424363e-05, "loss": 0.0993, "step": 4830 }, { "epoch": 0.23, "learning_rate": 1.7738329123397886e-05, "loss": 0.1731, "step": 4835 }, { "epoch": 0.23, "learning_rate": 1.773599027037141e-05, "loss": 0.1137, "step": 4840 }, { "epoch": 0.23, "learning_rate": 1.7733651417344936e-05, "loss": 0.1801, "step": 4845 }, { "epoch": 0.23, "learning_rate": 1.773131256431846e-05, "loss": 0.1054, "step": 4850 }, { "epoch": 0.23, "learning_rate": 1.7728973711291982e-05, "loss": 0.1565, "step": 4855 }, { "epoch": 0.23, "learning_rate": 1.772663485826551e-05, "loss": 0.153, "step": 4860 }, { "epoch": 0.23, "learning_rate": 1.7724296005239032e-05, "loss": 0.1503, "step": 4865 }, { "epoch": 0.23, "learning_rate": 1.7721957152212556e-05, "loss": 0.1035, "step": 4870 }, { "epoch": 0.23, "learning_rate": 1.7719618299186082e-05, "loss": 0.1643, "step": 4875 }, { "epoch": 0.23, "learning_rate": 1.7717279446159606e-05, "loss": 0.1929, "step": 4880 }, { "epoch": 0.23, "learning_rate": 1.771494059313313e-05, "loss": 0.1551, "step": 4885 }, { "epoch": 0.23, "learning_rate": 1.7712601740106655e-05, "loss": 0.0839, "step": 4890 }, { "epoch": 0.23, "learning_rate": 1.7710262887080175e-05, "loss": 0.0868, "step": 4895 }, { "epoch": 0.23, "learning_rate": 1.7707924034053702e-05, "loss": 0.1192, "step": 4900 }, { "epoch": 0.23, "learning_rate": 1.7705585181027225e-05, "loss": 0.1132, "step": 4905 }, { "epoch": 0.23, "learning_rate": 1.770324632800075e-05, "loss": 0.0825, "step": 4910 }, { "epoch": 0.23, "learning_rate": 1.7700907474974275e-05, "loss": 0.1731, "step": 4915 }, { "epoch": 0.23, "learning_rate": 1.76985686219478e-05, "loss": 0.1113, "step": 4920 }, { "epoch": 0.23, "learning_rate": 1.769622976892132e-05, "loss": 0.0888, "step": 4925 }, { "epoch": 0.23, "learning_rate": 1.7693890915894848e-05, "loss": 0.1536, "step": 4930 }, { "epoch": 0.23, "learning_rate": 1.769155206286837e-05, "loss": 0.1077, "step": 4935 }, { "epoch": 0.23, "learning_rate": 1.7689213209841895e-05, "loss": 0.0803, "step": 4940 }, { "epoch": 0.23, "learning_rate": 1.768687435681542e-05, "loss": 0.148, "step": 4945 }, { "epoch": 0.23, "learning_rate": 1.7684535503788945e-05, "loss": 0.1425, "step": 4950 }, { "epoch": 0.23, "learning_rate": 1.7682196650762468e-05, "loss": 0.1337, "step": 4955 }, { "epoch": 0.23, "learning_rate": 1.767985779773599e-05, "loss": 0.1075, "step": 4960 }, { "epoch": 0.23, "learning_rate": 1.7677518944709514e-05, "loss": 0.1791, "step": 4965 }, { "epoch": 0.23, "learning_rate": 1.767518009168304e-05, "loss": 0.1247, "step": 4970 }, { "epoch": 0.23, "learning_rate": 1.7672841238656564e-05, "loss": 0.0917, "step": 4975 }, { "epoch": 0.23, "learning_rate": 1.7670502385630088e-05, "loss": 0.1155, "step": 4980 }, { "epoch": 0.23, "learning_rate": 1.7668163532603614e-05, "loss": 0.0847, "step": 4985 }, { "epoch": 0.23, "learning_rate": 1.7665824679577138e-05, "loss": 0.0918, "step": 4990 }, { "epoch": 0.23, "learning_rate": 1.766348582655066e-05, "loss": 0.1418, "step": 4995 }, { "epoch": 0.23, "learning_rate": 1.7661146973524187e-05, "loss": 0.1466, "step": 5000 }, { "epoch": 0.23, "learning_rate": 1.765880812049771e-05, "loss": 0.1182, "step": 5005 }, { "epoch": 0.23, "learning_rate": 1.7656469267471234e-05, "loss": 0.107, "step": 5010 }, { "epoch": 0.23, "learning_rate": 1.7654130414444757e-05, "loss": 0.1323, "step": 5015 }, { "epoch": 0.23, "learning_rate": 1.765179156141828e-05, "loss": 0.1, "step": 5020 }, { "epoch": 0.24, "learning_rate": 1.7649452708391807e-05, "loss": 0.0947, "step": 5025 }, { "epoch": 0.24, "learning_rate": 1.764711385536533e-05, "loss": 0.143, "step": 5030 }, { "epoch": 0.24, "learning_rate": 1.7644775002338854e-05, "loss": 0.1764, "step": 5035 }, { "epoch": 0.24, "learning_rate": 1.764243614931238e-05, "loss": 0.0787, "step": 5040 }, { "epoch": 0.24, "learning_rate": 1.7640097296285904e-05, "loss": 0.0942, "step": 5045 }, { "epoch": 0.24, "learning_rate": 1.7637758443259427e-05, "loss": 0.101, "step": 5050 }, { "epoch": 0.24, "learning_rate": 1.763541959023295e-05, "loss": 0.1248, "step": 5055 }, { "epoch": 0.24, "learning_rate": 1.7633080737206477e-05, "loss": 0.0954, "step": 5060 }, { "epoch": 0.24, "learning_rate": 1.763074188418e-05, "loss": 0.0869, "step": 5065 }, { "epoch": 0.24, "learning_rate": 1.7628403031153523e-05, "loss": 0.1164, "step": 5070 }, { "epoch": 0.24, "learning_rate": 1.7626064178127046e-05, "loss": 0.0928, "step": 5075 }, { "epoch": 0.24, "learning_rate": 1.762372532510057e-05, "loss": 0.1077, "step": 5080 }, { "epoch": 0.24, "learning_rate": 1.7621386472074096e-05, "loss": 0.1036, "step": 5085 }, { "epoch": 0.24, "learning_rate": 1.761904761904762e-05, "loss": 0.1382, "step": 5090 }, { "epoch": 0.24, "learning_rate": 1.7616708766021143e-05, "loss": 0.1546, "step": 5095 }, { "epoch": 0.24, "learning_rate": 1.761436991299467e-05, "loss": 0.1282, "step": 5100 }, { "epoch": 0.24, "learning_rate": 1.7612031059968193e-05, "loss": 0.1039, "step": 5105 }, { "epoch": 0.24, "learning_rate": 1.7609692206941716e-05, "loss": 0.1616, "step": 5110 }, { "epoch": 0.24, "learning_rate": 1.7607353353915243e-05, "loss": 0.1063, "step": 5115 }, { "epoch": 0.24, "learning_rate": 1.7605014500888766e-05, "loss": 0.1389, "step": 5120 }, { "epoch": 0.24, "learning_rate": 1.760267564786229e-05, "loss": 0.1135, "step": 5125 }, { "epoch": 0.24, "learning_rate": 1.7600336794835816e-05, "loss": 0.1436, "step": 5130 }, { "epoch": 0.24, "learning_rate": 1.7597997941809336e-05, "loss": 0.0571, "step": 5135 }, { "epoch": 0.24, "learning_rate": 1.7595659088782862e-05, "loss": 0.1486, "step": 5140 }, { "epoch": 0.24, "learning_rate": 1.7593320235756386e-05, "loss": 0.1827, "step": 5145 }, { "epoch": 0.24, "learning_rate": 1.759098138272991e-05, "loss": 0.1175, "step": 5150 }, { "epoch": 0.24, "learning_rate": 1.7588642529703436e-05, "loss": 0.1092, "step": 5155 }, { "epoch": 0.24, "learning_rate": 1.758630367667696e-05, "loss": 0.0851, "step": 5160 }, { "epoch": 0.24, "learning_rate": 1.7583964823650482e-05, "loss": 0.0785, "step": 5165 }, { "epoch": 0.24, "learning_rate": 1.758162597062401e-05, "loss": 0.1049, "step": 5170 }, { "epoch": 0.24, "learning_rate": 1.7579287117597532e-05, "loss": 0.1154, "step": 5175 }, { "epoch": 0.24, "learning_rate": 1.7576948264571055e-05, "loss": 0.114, "step": 5180 }, { "epoch": 0.24, "learning_rate": 1.7574609411544582e-05, "loss": 0.1515, "step": 5185 }, { "epoch": 0.24, "learning_rate": 1.7572270558518105e-05, "loss": 0.1141, "step": 5190 }, { "epoch": 0.24, "learning_rate": 1.756993170549163e-05, "loss": 0.1063, "step": 5195 }, { "epoch": 0.24, "learning_rate": 1.756759285246515e-05, "loss": 0.107, "step": 5200 }, { "epoch": 0.24, "learning_rate": 1.7565253999438675e-05, "loss": 0.172, "step": 5205 }, { "epoch": 0.24, "learning_rate": 1.75629151464122e-05, "loss": 0.1203, "step": 5210 }, { "epoch": 0.24, "learning_rate": 1.7560576293385725e-05, "loss": 0.0806, "step": 5215 }, { "epoch": 0.24, "learning_rate": 1.7558237440359248e-05, "loss": 0.1466, "step": 5220 }, { "epoch": 0.24, "learning_rate": 1.7555898587332775e-05, "loss": 0.1089, "step": 5225 }, { "epoch": 0.24, "learning_rate": 1.7553559734306298e-05, "loss": 0.135, "step": 5230 }, { "epoch": 0.24, "learning_rate": 1.755122088127982e-05, "loss": 0.1353, "step": 5235 }, { "epoch": 0.25, "learning_rate": 1.7548882028253348e-05, "loss": 0.0992, "step": 5240 }, { "epoch": 0.25, "learning_rate": 1.754654317522687e-05, "loss": 0.1288, "step": 5245 }, { "epoch": 0.25, "learning_rate": 1.7544204322200394e-05, "loss": 0.1325, "step": 5250 }, { "epoch": 0.25, "learning_rate": 1.7541865469173918e-05, "loss": 0.0805, "step": 5255 }, { "epoch": 0.25, "learning_rate": 1.753952661614744e-05, "loss": 0.0975, "step": 5260 }, { "epoch": 0.25, "learning_rate": 1.7537187763120967e-05, "loss": 0.0697, "step": 5265 }, { "epoch": 0.25, "learning_rate": 1.753484891009449e-05, "loss": 0.1271, "step": 5270 }, { "epoch": 0.25, "learning_rate": 1.7532510057068014e-05, "loss": 0.0822, "step": 5275 }, { "epoch": 0.25, "learning_rate": 1.753017120404154e-05, "loss": 0.1214, "step": 5280 }, { "epoch": 0.25, "learning_rate": 1.7527832351015064e-05, "loss": 0.0888, "step": 5285 }, { "epoch": 0.25, "learning_rate": 1.7525493497988587e-05, "loss": 0.1375, "step": 5290 }, { "epoch": 0.25, "learning_rate": 1.7523154644962114e-05, "loss": 0.0619, "step": 5295 }, { "epoch": 0.25, "learning_rate": 1.7520815791935637e-05, "loss": 0.0926, "step": 5300 }, { "epoch": 0.25, "learning_rate": 1.751847693890916e-05, "loss": 0.124, "step": 5305 }, { "epoch": 0.25, "learning_rate": 1.7516138085882684e-05, "loss": 0.1148, "step": 5310 }, { "epoch": 0.25, "learning_rate": 1.7513799232856207e-05, "loss": 0.097, "step": 5315 }, { "epoch": 0.25, "learning_rate": 1.7511460379829733e-05, "loss": 0.1452, "step": 5320 }, { "epoch": 0.25, "learning_rate": 1.7509121526803257e-05, "loss": 0.1757, "step": 5325 }, { "epoch": 0.25, "learning_rate": 1.750678267377678e-05, "loss": 0.1331, "step": 5330 }, { "epoch": 0.25, "learning_rate": 1.7504443820750307e-05, "loss": 0.1457, "step": 5335 }, { "epoch": 0.25, "learning_rate": 1.750210496772383e-05, "loss": 0.1478, "step": 5340 }, { "epoch": 0.25, "learning_rate": 1.7499766114697353e-05, "loss": 0.119, "step": 5345 }, { "epoch": 0.25, "learning_rate": 1.749742726167088e-05, "loss": 0.0874, "step": 5350 }, { "epoch": 0.25, "learning_rate": 1.7495088408644403e-05, "loss": 0.1057, "step": 5355 }, { "epoch": 0.25, "learning_rate": 1.7492749555617926e-05, "loss": 0.0654, "step": 5360 }, { "epoch": 0.25, "learning_rate": 1.7490410702591453e-05, "loss": 0.0415, "step": 5365 }, { "epoch": 0.25, "learning_rate": 1.7488071849564973e-05, "loss": 0.1035, "step": 5370 }, { "epoch": 0.25, "learning_rate": 1.74857329965385e-05, "loss": 0.1103, "step": 5375 }, { "epoch": 0.25, "learning_rate": 1.7483394143512023e-05, "loss": 0.0893, "step": 5380 }, { "epoch": 0.25, "learning_rate": 1.7481055290485546e-05, "loss": 0.1883, "step": 5385 }, { "epoch": 0.25, "learning_rate": 1.7478716437459073e-05, "loss": 0.1523, "step": 5390 }, { "epoch": 0.25, "learning_rate": 1.7476377584432596e-05, "loss": 0.1882, "step": 5395 }, { "epoch": 0.25, "learning_rate": 1.747403873140612e-05, "loss": 0.1279, "step": 5400 }, { "epoch": 0.25, "learning_rate": 1.7471699878379646e-05, "loss": 0.1413, "step": 5405 }, { "epoch": 0.25, "learning_rate": 1.746936102535317e-05, "loss": 0.1176, "step": 5410 }, { "epoch": 0.25, "learning_rate": 1.7467022172326692e-05, "loss": 0.0781, "step": 5415 }, { "epoch": 0.25, "learning_rate": 1.746468331930022e-05, "loss": 0.1473, "step": 5420 }, { "epoch": 0.25, "learning_rate": 1.7462344466273742e-05, "loss": 0.1153, "step": 5425 }, { "epoch": 0.25, "learning_rate": 1.7460005613247265e-05, "loss": 0.1158, "step": 5430 }, { "epoch": 0.25, "learning_rate": 1.745766676022079e-05, "loss": 0.0891, "step": 5435 }, { "epoch": 0.25, "learning_rate": 1.7455327907194312e-05, "loss": 0.1278, "step": 5440 }, { "epoch": 0.25, "learning_rate": 1.745298905416784e-05, "loss": 0.1346, "step": 5445 }, { "epoch": 0.25, "learning_rate": 1.7450650201141362e-05, "loss": 0.1437, "step": 5450 }, { "epoch": 0.26, "learning_rate": 1.7448311348114885e-05, "loss": 0.0956, "step": 5455 }, { "epoch": 0.26, "learning_rate": 1.7445972495088412e-05, "loss": 0.1683, "step": 5460 }, { "epoch": 0.26, "learning_rate": 1.7443633642061935e-05, "loss": 0.2007, "step": 5465 }, { "epoch": 0.26, "learning_rate": 1.744129478903546e-05, "loss": 0.0701, "step": 5470 }, { "epoch": 0.26, "learning_rate": 1.7438955936008985e-05, "loss": 0.0696, "step": 5475 }, { "epoch": 0.26, "learning_rate": 1.7436617082982508e-05, "loss": 0.1246, "step": 5480 }, { "epoch": 0.26, "learning_rate": 1.743427822995603e-05, "loss": 0.1755, "step": 5485 }, { "epoch": 0.26, "learning_rate": 1.7431939376929555e-05, "loss": 0.1508, "step": 5490 }, { "epoch": 0.26, "learning_rate": 1.7429600523903078e-05, "loss": 0.0875, "step": 5495 }, { "epoch": 0.26, "learning_rate": 1.74272616708766e-05, "loss": 0.1298, "step": 5500 }, { "epoch": 0.26, "learning_rate": 1.7424922817850128e-05, "loss": 0.136, "step": 5505 }, { "epoch": 0.26, "learning_rate": 1.742258396482365e-05, "loss": 0.1281, "step": 5510 }, { "epoch": 0.26, "learning_rate": 1.7420245111797174e-05, "loss": 0.1575, "step": 5515 }, { "epoch": 0.26, "learning_rate": 1.74179062587707e-05, "loss": 0.0959, "step": 5520 }, { "epoch": 0.26, "learning_rate": 1.7415567405744224e-05, "loss": 0.0926, "step": 5525 }, { "epoch": 0.26, "learning_rate": 1.7413228552717748e-05, "loss": 0.1645, "step": 5530 }, { "epoch": 0.26, "learning_rate": 1.7410889699691274e-05, "loss": 0.1309, "step": 5535 }, { "epoch": 0.26, "learning_rate": 1.7408550846664797e-05, "loss": 0.1207, "step": 5540 }, { "epoch": 0.26, "learning_rate": 1.740621199363832e-05, "loss": 0.1133, "step": 5545 }, { "epoch": 0.26, "learning_rate": 1.7403873140611844e-05, "loss": 0.1985, "step": 5550 }, { "epoch": 0.26, "learning_rate": 1.7401534287585367e-05, "loss": 0.1112, "step": 5555 }, { "epoch": 0.26, "learning_rate": 1.7399195434558894e-05, "loss": 0.1586, "step": 5560 }, { "epoch": 0.26, "learning_rate": 1.7396856581532417e-05, "loss": 0.1485, "step": 5565 }, { "epoch": 0.26, "learning_rate": 1.739451772850594e-05, "loss": 0.0552, "step": 5570 }, { "epoch": 0.26, "learning_rate": 1.7392178875479467e-05, "loss": 0.1322, "step": 5575 }, { "epoch": 0.26, "learning_rate": 1.738984002245299e-05, "loss": 0.1337, "step": 5580 }, { "epoch": 0.26, "learning_rate": 1.7387501169426514e-05, "loss": 0.1056, "step": 5585 }, { "epoch": 0.26, "learning_rate": 1.738516231640004e-05, "loss": 0.0598, "step": 5590 }, { "epoch": 0.26, "learning_rate": 1.7382823463373563e-05, "loss": 0.1176, "step": 5595 }, { "epoch": 0.26, "learning_rate": 1.7380484610347087e-05, "loss": 0.1019, "step": 5600 }, { "epoch": 0.26, "learning_rate": 1.7378145757320613e-05, "loss": 0.0997, "step": 5605 }, { "epoch": 0.26, "learning_rate": 1.7375806904294133e-05, "loss": 0.1055, "step": 5610 }, { "epoch": 0.26, "learning_rate": 1.737346805126766e-05, "loss": 0.1538, "step": 5615 }, { "epoch": 0.26, "learning_rate": 1.7371129198241183e-05, "loss": 0.1287, "step": 5620 }, { "epoch": 0.26, "learning_rate": 1.7368790345214706e-05, "loss": 0.0969, "step": 5625 }, { "epoch": 0.26, "learning_rate": 1.7366451492188233e-05, "loss": 0.1526, "step": 5630 }, { "epoch": 0.26, "learning_rate": 1.7364112639161756e-05, "loss": 0.0525, "step": 5635 }, { "epoch": 0.26, "learning_rate": 1.736177378613528e-05, "loss": 0.1308, "step": 5640 }, { "epoch": 0.26, "learning_rate": 1.7359434933108806e-05, "loss": 0.1068, "step": 5645 }, { "epoch": 0.26, "learning_rate": 1.735709608008233e-05, "loss": 0.0967, "step": 5650 }, { "epoch": 0.26, "learning_rate": 1.7354757227055853e-05, "loss": 0.1, "step": 5655 }, { "epoch": 0.26, "learning_rate": 1.735241837402938e-05, "loss": 0.1307, "step": 5660 }, { "epoch": 0.26, "learning_rate": 1.73500795210029e-05, "loss": 0.1552, "step": 5665 }, { "epoch": 0.27, "learning_rate": 1.7347740667976426e-05, "loss": 0.0989, "step": 5670 }, { "epoch": 0.27, "learning_rate": 1.734540181494995e-05, "loss": 0.0696, "step": 5675 }, { "epoch": 0.27, "learning_rate": 1.7343062961923472e-05, "loss": 0.1387, "step": 5680 }, { "epoch": 0.27, "learning_rate": 1.7340724108897e-05, "loss": 0.135, "step": 5685 }, { "epoch": 0.27, "learning_rate": 1.7338385255870522e-05, "loss": 0.1399, "step": 5690 }, { "epoch": 0.27, "learning_rate": 1.7336046402844046e-05, "loss": 0.2022, "step": 5695 }, { "epoch": 0.27, "learning_rate": 1.7333707549817572e-05, "loss": 0.1442, "step": 5700 }, { "epoch": 0.27, "learning_rate": 1.7331368696791095e-05, "loss": 0.1411, "step": 5705 }, { "epoch": 0.27, "learning_rate": 1.732902984376462e-05, "loss": 0.1003, "step": 5710 }, { "epoch": 0.27, "learning_rate": 1.7326690990738145e-05, "loss": 0.1166, "step": 5715 }, { "epoch": 0.27, "learning_rate": 1.732435213771167e-05, "loss": 0.0883, "step": 5720 }, { "epoch": 0.27, "learning_rate": 1.7322013284685192e-05, "loss": 0.1536, "step": 5725 }, { "epoch": 0.27, "learning_rate": 1.7319674431658715e-05, "loss": 0.0654, "step": 5730 }, { "epoch": 0.27, "learning_rate": 1.731733557863224e-05, "loss": 0.1386, "step": 5735 }, { "epoch": 0.27, "learning_rate": 1.7314996725605765e-05, "loss": 0.1133, "step": 5740 }, { "epoch": 0.27, "learning_rate": 1.7312657872579288e-05, "loss": 0.0963, "step": 5745 }, { "epoch": 0.27, "learning_rate": 1.731031901955281e-05, "loss": 0.1467, "step": 5750 }, { "epoch": 0.27, "learning_rate": 1.7307980166526338e-05, "loss": 0.0699, "step": 5755 }, { "epoch": 0.27, "learning_rate": 1.730564131349986e-05, "loss": 0.066, "step": 5760 }, { "epoch": 0.27, "learning_rate": 1.7303302460473385e-05, "loss": 0.108, "step": 5765 }, { "epoch": 0.27, "learning_rate": 1.730096360744691e-05, "loss": 0.1083, "step": 5770 }, { "epoch": 0.27, "learning_rate": 1.7298624754420435e-05, "loss": 0.1805, "step": 5775 }, { "epoch": 0.27, "learning_rate": 1.7296285901393958e-05, "loss": 0.1206, "step": 5780 }, { "epoch": 0.27, "learning_rate": 1.729394704836748e-05, "loss": 0.0955, "step": 5785 }, { "epoch": 0.27, "learning_rate": 1.7291608195341004e-05, "loss": 0.075, "step": 5790 }, { "epoch": 0.27, "learning_rate": 1.728926934231453e-05, "loss": 0.1464, "step": 5795 }, { "epoch": 0.27, "learning_rate": 1.7286930489288054e-05, "loss": 0.0878, "step": 5800 }, { "epoch": 0.27, "learning_rate": 1.7284591636261578e-05, "loss": 0.1774, "step": 5805 }, { "epoch": 0.27, "learning_rate": 1.7282252783235104e-05, "loss": 0.1253, "step": 5810 }, { "epoch": 0.27, "learning_rate": 1.7279913930208627e-05, "loss": 0.0741, "step": 5815 }, { "epoch": 0.27, "learning_rate": 1.727757507718215e-05, "loss": 0.0869, "step": 5820 }, { "epoch": 0.27, "learning_rate": 1.7275236224155677e-05, "loss": 0.1486, "step": 5825 }, { "epoch": 0.27, "learning_rate": 1.72728973711292e-05, "loss": 0.0865, "step": 5830 }, { "epoch": 0.27, "learning_rate": 1.7270558518102724e-05, "loss": 0.102, "step": 5835 }, { "epoch": 0.27, "learning_rate": 1.726821966507625e-05, "loss": 0.1176, "step": 5840 }, { "epoch": 0.27, "learning_rate": 1.726588081204977e-05, "loss": 0.1067, "step": 5845 }, { "epoch": 0.27, "learning_rate": 1.7263541959023297e-05, "loss": 0.1332, "step": 5850 }, { "epoch": 0.27, "learning_rate": 1.726120310599682e-05, "loss": 0.1009, "step": 5855 }, { "epoch": 0.27, "learning_rate": 1.7258864252970344e-05, "loss": 0.13, "step": 5860 }, { "epoch": 0.27, "learning_rate": 1.725652539994387e-05, "loss": 0.1148, "step": 5865 }, { "epoch": 0.27, "learning_rate": 1.7254186546917393e-05, "loss": 0.0918, "step": 5870 }, { "epoch": 0.27, "learning_rate": 1.7251847693890917e-05, "loss": 0.1268, "step": 5875 }, { "epoch": 0.28, "learning_rate": 1.7249508840864443e-05, "loss": 0.0616, "step": 5880 }, { "epoch": 0.28, "learning_rate": 1.7247169987837967e-05, "loss": 0.1652, "step": 5885 }, { "epoch": 0.28, "learning_rate": 1.724483113481149e-05, "loss": 0.1001, "step": 5890 }, { "epoch": 0.28, "learning_rate": 1.7242492281785017e-05, "loss": 0.114, "step": 5895 }, { "epoch": 0.28, "learning_rate": 1.724015342875854e-05, "loss": 0.0718, "step": 5900 }, { "epoch": 0.28, "learning_rate": 1.7237814575732063e-05, "loss": 0.0961, "step": 5905 }, { "epoch": 0.28, "learning_rate": 1.7235475722705586e-05, "loss": 0.0633, "step": 5910 }, { "epoch": 0.28, "learning_rate": 1.723313686967911e-05, "loss": 0.1145, "step": 5915 }, { "epoch": 0.28, "learning_rate": 1.7230798016652636e-05, "loss": 0.0594, "step": 5920 }, { "epoch": 0.28, "learning_rate": 1.722845916362616e-05, "loss": 0.1123, "step": 5925 }, { "epoch": 0.28, "learning_rate": 1.7226120310599683e-05, "loss": 0.1208, "step": 5930 }, { "epoch": 0.28, "learning_rate": 1.7223781457573206e-05, "loss": 0.1043, "step": 5935 }, { "epoch": 0.28, "learning_rate": 1.7221442604546733e-05, "loss": 0.1756, "step": 5940 }, { "epoch": 0.28, "learning_rate": 1.7219103751520256e-05, "loss": 0.1142, "step": 5945 }, { "epoch": 0.28, "learning_rate": 1.721676489849378e-05, "loss": 0.1042, "step": 5950 }, { "epoch": 0.28, "learning_rate": 1.7214426045467306e-05, "loss": 0.0884, "step": 5955 }, { "epoch": 0.28, "learning_rate": 1.721208719244083e-05, "loss": 0.0857, "step": 5960 }, { "epoch": 0.28, "learning_rate": 1.7209748339414352e-05, "loss": 0.0994, "step": 5965 }, { "epoch": 0.28, "learning_rate": 1.7207409486387876e-05, "loss": 0.1152, "step": 5970 }, { "epoch": 0.28, "learning_rate": 1.72050706333614e-05, "loss": 0.118, "step": 5975 }, { "epoch": 0.28, "learning_rate": 1.7202731780334925e-05, "loss": 0.0837, "step": 5980 }, { "epoch": 0.28, "learning_rate": 1.720039292730845e-05, "loss": 0.1563, "step": 5985 }, { "epoch": 0.28, "learning_rate": 1.7198054074281972e-05, "loss": 0.0947, "step": 5990 }, { "epoch": 0.28, "learning_rate": 1.71957152212555e-05, "loss": 0.0774, "step": 5995 }, { "epoch": 0.28, "learning_rate": 1.7193376368229022e-05, "loss": 0.1069, "step": 6000 }, { "epoch": 0.28, "learning_rate": 1.7191037515202545e-05, "loss": 0.1107, "step": 6005 }, { "epoch": 0.28, "learning_rate": 1.7188698662176072e-05, "loss": 0.1061, "step": 6010 }, { "epoch": 0.28, "learning_rate": 1.7186359809149595e-05, "loss": 0.0676, "step": 6015 }, { "epoch": 0.28, "learning_rate": 1.7184020956123118e-05, "loss": 0.1244, "step": 6020 }, { "epoch": 0.28, "learning_rate": 1.718168210309664e-05, "loss": 0.1618, "step": 6025 }, { "epoch": 0.28, "learning_rate": 1.7179343250070165e-05, "loss": 0.0755, "step": 6030 }, { "epoch": 0.28, "learning_rate": 1.717700439704369e-05, "loss": 0.058, "step": 6035 }, { "epoch": 0.28, "learning_rate": 1.7174665544017215e-05, "loss": 0.1103, "step": 6040 }, { "epoch": 0.28, "learning_rate": 1.7172326690990738e-05, "loss": 0.1731, "step": 6045 }, { "epoch": 0.28, "learning_rate": 1.7169987837964265e-05, "loss": 0.0725, "step": 6050 }, { "epoch": 0.28, "learning_rate": 1.7167648984937788e-05, "loss": 0.1329, "step": 6055 }, { "epoch": 0.28, "learning_rate": 1.716531013191131e-05, "loss": 0.1526, "step": 6060 }, { "epoch": 0.28, "learning_rate": 1.7162971278884838e-05, "loss": 0.1348, "step": 6065 }, { "epoch": 0.28, "learning_rate": 1.716063242585836e-05, "loss": 0.117, "step": 6070 }, { "epoch": 0.28, "learning_rate": 1.7158293572831884e-05, "loss": 0.1164, "step": 6075 }, { "epoch": 0.28, "learning_rate": 1.715595471980541e-05, "loss": 0.2138, "step": 6080 }, { "epoch": 0.28, "learning_rate": 1.715361586677893e-05, "loss": 0.0881, "step": 6085 }, { "epoch": 0.28, "learning_rate": 1.7151277013752457e-05, "loss": 0.1379, "step": 6090 }, { "epoch": 0.29, "learning_rate": 1.714893816072598e-05, "loss": 0.0973, "step": 6095 }, { "epoch": 0.29, "learning_rate": 1.7146599307699504e-05, "loss": 0.1211, "step": 6100 }, { "epoch": 0.29, "learning_rate": 1.714426045467303e-05, "loss": 0.1099, "step": 6105 }, { "epoch": 0.29, "learning_rate": 1.7141921601646554e-05, "loss": 0.129, "step": 6110 }, { "epoch": 0.29, "learning_rate": 1.7139582748620077e-05, "loss": 0.0707, "step": 6115 }, { "epoch": 0.29, "learning_rate": 1.7137243895593604e-05, "loss": 0.0985, "step": 6120 }, { "epoch": 0.29, "learning_rate": 1.7134905042567127e-05, "loss": 0.1493, "step": 6125 }, { "epoch": 0.29, "learning_rate": 1.713256618954065e-05, "loss": 0.1654, "step": 6130 }, { "epoch": 0.29, "learning_rate": 1.7130227336514177e-05, "loss": 0.0996, "step": 6135 }, { "epoch": 0.29, "learning_rate": 1.7127888483487697e-05, "loss": 0.1362, "step": 6140 }, { "epoch": 0.29, "learning_rate": 1.7125549630461223e-05, "loss": 0.1113, "step": 6145 }, { "epoch": 0.29, "learning_rate": 1.7123210777434747e-05, "loss": 0.1451, "step": 6150 }, { "epoch": 0.29, "learning_rate": 1.712087192440827e-05, "loss": 0.079, "step": 6155 }, { "epoch": 0.29, "learning_rate": 1.7118533071381797e-05, "loss": 0.1259, "step": 6160 }, { "epoch": 0.29, "learning_rate": 1.711619421835532e-05, "loss": 0.0783, "step": 6165 }, { "epoch": 0.29, "learning_rate": 1.7113855365328843e-05, "loss": 0.1182, "step": 6170 }, { "epoch": 0.29, "learning_rate": 1.711151651230237e-05, "loss": 0.1178, "step": 6175 }, { "epoch": 0.29, "learning_rate": 1.7109177659275893e-05, "loss": 0.1901, "step": 6180 }, { "epoch": 0.29, "learning_rate": 1.7106838806249416e-05, "loss": 0.0821, "step": 6185 }, { "epoch": 0.29, "learning_rate": 1.7104499953222943e-05, "loss": 0.1033, "step": 6190 }, { "epoch": 0.29, "learning_rate": 1.7102161100196466e-05, "loss": 0.1073, "step": 6195 }, { "epoch": 0.29, "learning_rate": 1.709982224716999e-05, "loss": 0.1668, "step": 6200 }, { "epoch": 0.29, "learning_rate": 1.7097483394143513e-05, "loss": 0.1303, "step": 6205 }, { "epoch": 0.29, "learning_rate": 1.7095144541117036e-05, "loss": 0.1596, "step": 6210 }, { "epoch": 0.29, "learning_rate": 1.7092805688090563e-05, "loss": 0.1232, "step": 6215 }, { "epoch": 0.29, "learning_rate": 1.7090466835064086e-05, "loss": 0.1055, "step": 6220 }, { "epoch": 0.29, "learning_rate": 1.708812798203761e-05, "loss": 0.1478, "step": 6225 }, { "epoch": 0.29, "learning_rate": 1.7085789129011136e-05, "loss": 0.0999, "step": 6230 }, { "epoch": 0.29, "learning_rate": 1.708345027598466e-05, "loss": 0.1485, "step": 6235 }, { "epoch": 0.29, "learning_rate": 1.7081111422958182e-05, "loss": 0.0865, "step": 6240 }, { "epoch": 0.29, "learning_rate": 1.707877256993171e-05, "loss": 0.1416, "step": 6245 }, { "epoch": 0.29, "learning_rate": 1.7076433716905232e-05, "loss": 0.1124, "step": 6250 }, { "epoch": 0.29, "learning_rate": 1.7074094863878755e-05, "loss": 0.1306, "step": 6255 }, { "epoch": 0.29, "learning_rate": 1.707175601085228e-05, "loss": 0.1203, "step": 6260 }, { "epoch": 0.29, "learning_rate": 1.7069417157825802e-05, "loss": 0.1906, "step": 6265 }, { "epoch": 0.29, "learning_rate": 1.706707830479933e-05, "loss": 0.1044, "step": 6270 }, { "epoch": 0.29, "learning_rate": 1.7064739451772852e-05, "loss": 0.1054, "step": 6275 }, { "epoch": 0.29, "learning_rate": 1.7062400598746375e-05, "loss": 0.0809, "step": 6280 }, { "epoch": 0.29, "learning_rate": 1.7060061745719902e-05, "loss": 0.1401, "step": 6285 }, { "epoch": 0.29, "learning_rate": 1.7057722892693425e-05, "loss": 0.1101, "step": 6290 }, { "epoch": 0.29, "learning_rate": 1.7055384039666948e-05, "loss": 0.0773, "step": 6295 }, { "epoch": 0.29, "learning_rate": 1.7053045186640475e-05, "loss": 0.1327, "step": 6300 }, { "epoch": 0.29, "learning_rate": 1.7050706333613998e-05, "loss": 0.1239, "step": 6305 }, { "epoch": 0.3, "learning_rate": 1.704836748058752e-05, "loss": 0.1174, "step": 6310 }, { "epoch": 0.3, "learning_rate": 1.7046028627561048e-05, "loss": 0.0752, "step": 6315 }, { "epoch": 0.3, "learning_rate": 1.7043689774534568e-05, "loss": 0.1357, "step": 6320 }, { "epoch": 0.3, "learning_rate": 1.7041350921508095e-05, "loss": 0.1011, "step": 6325 }, { "epoch": 0.3, "learning_rate": 1.7039012068481618e-05, "loss": 0.0797, "step": 6330 }, { "epoch": 0.3, "learning_rate": 1.703667321545514e-05, "loss": 0.1442, "step": 6335 }, { "epoch": 0.3, "learning_rate": 1.7034334362428668e-05, "loss": 0.1638, "step": 6340 }, { "epoch": 0.3, "learning_rate": 1.703199550940219e-05, "loss": 0.0858, "step": 6345 }, { "epoch": 0.3, "learning_rate": 1.7029656656375714e-05, "loss": 0.0907, "step": 6350 }, { "epoch": 0.3, "learning_rate": 1.702731780334924e-05, "loss": 0.1137, "step": 6355 }, { "epoch": 0.3, "learning_rate": 1.7024978950322764e-05, "loss": 0.1825, "step": 6360 }, { "epoch": 0.3, "learning_rate": 1.7022640097296287e-05, "loss": 0.164, "step": 6365 }, { "epoch": 0.3, "learning_rate": 1.702030124426981e-05, "loss": 0.1306, "step": 6370 }, { "epoch": 0.3, "learning_rate": 1.7017962391243337e-05, "loss": 0.0746, "step": 6375 }, { "epoch": 0.3, "learning_rate": 1.7015623538216857e-05, "loss": 0.1084, "step": 6380 }, { "epoch": 0.3, "learning_rate": 1.7013284685190384e-05, "loss": 0.1527, "step": 6385 }, { "epoch": 0.3, "learning_rate": 1.7010945832163907e-05, "loss": 0.152, "step": 6390 }, { "epoch": 0.3, "learning_rate": 1.700860697913743e-05, "loss": 0.0531, "step": 6395 }, { "epoch": 0.3, "learning_rate": 1.7006268126110957e-05, "loss": 0.1367, "step": 6400 }, { "epoch": 0.3, "learning_rate": 1.700392927308448e-05, "loss": 0.0784, "step": 6405 }, { "epoch": 0.3, "learning_rate": 1.7001590420058004e-05, "loss": 0.1001, "step": 6410 }, { "epoch": 0.3, "learning_rate": 1.699925156703153e-05, "loss": 0.0921, "step": 6415 }, { "epoch": 0.3, "learning_rate": 1.6996912714005053e-05, "loss": 0.095, "step": 6420 }, { "epoch": 0.3, "learning_rate": 1.6994573860978577e-05, "loss": 0.0843, "step": 6425 }, { "epoch": 0.3, "learning_rate": 1.6992235007952103e-05, "loss": 0.0942, "step": 6430 }, { "epoch": 0.3, "learning_rate": 1.6989896154925627e-05, "loss": 0.1745, "step": 6435 }, { "epoch": 0.3, "learning_rate": 1.698755730189915e-05, "loss": 0.1337, "step": 6440 }, { "epoch": 0.3, "learning_rate": 1.6985218448872673e-05, "loss": 0.1091, "step": 6445 }, { "epoch": 0.3, "learning_rate": 1.6982879595846196e-05, "loss": 0.1444, "step": 6450 }, { "epoch": 0.3, "learning_rate": 1.6980540742819723e-05, "loss": 0.1325, "step": 6455 }, { "epoch": 0.3, "learning_rate": 1.6978201889793246e-05, "loss": 0.0729, "step": 6460 }, { "epoch": 0.3, "learning_rate": 1.697586303676677e-05, "loss": 0.0934, "step": 6465 }, { "epoch": 0.3, "learning_rate": 1.6973524183740296e-05, "loss": 0.1482, "step": 6470 }, { "epoch": 0.3, "learning_rate": 1.697118533071382e-05, "loss": 0.0763, "step": 6475 }, { "epoch": 0.3, "learning_rate": 1.6968846477687343e-05, "loss": 0.1131, "step": 6480 }, { "epoch": 0.3, "learning_rate": 1.696650762466087e-05, "loss": 0.1233, "step": 6485 }, { "epoch": 0.3, "learning_rate": 1.6964168771634393e-05, "loss": 0.1175, "step": 6490 }, { "epoch": 0.3, "learning_rate": 1.6961829918607916e-05, "loss": 0.064, "step": 6495 }, { "epoch": 0.3, "learning_rate": 1.695949106558144e-05, "loss": 0.0881, "step": 6500 }, { "epoch": 0.3, "learning_rate": 1.6957152212554962e-05, "loss": 0.108, "step": 6505 }, { "epoch": 0.3, "learning_rate": 1.695481335952849e-05, "loss": 0.0967, "step": 6510 }, { "epoch": 0.3, "learning_rate": 1.6952474506502012e-05, "loss": 0.0808, "step": 6515 }, { "epoch": 0.3, "learning_rate": 1.6950135653475536e-05, "loss": 0.0836, "step": 6520 }, { "epoch": 0.31, "learning_rate": 1.6947796800449062e-05, "loss": 0.0775, "step": 6525 }, { "epoch": 0.31, "learning_rate": 1.6945457947422585e-05, "loss": 0.1074, "step": 6530 }, { "epoch": 0.31, "learning_rate": 1.694311909439611e-05, "loss": 0.1301, "step": 6535 }, { "epoch": 0.31, "learning_rate": 1.6940780241369635e-05, "loss": 0.0963, "step": 6540 }, { "epoch": 0.31, "learning_rate": 1.693844138834316e-05, "loss": 0.1559, "step": 6545 }, { "epoch": 0.31, "learning_rate": 1.6936102535316682e-05, "loss": 0.1704, "step": 6550 }, { "epoch": 0.31, "learning_rate": 1.6933763682290205e-05, "loss": 0.1196, "step": 6555 }, { "epoch": 0.31, "learning_rate": 1.693142482926373e-05, "loss": 0.0846, "step": 6560 }, { "epoch": 0.31, "learning_rate": 1.6929085976237255e-05, "loss": 0.0845, "step": 6565 }, { "epoch": 0.31, "learning_rate": 1.6926747123210778e-05, "loss": 0.078, "step": 6570 }, { "epoch": 0.31, "learning_rate": 1.69244082701843e-05, "loss": 0.1057, "step": 6575 }, { "epoch": 0.31, "learning_rate": 1.6922069417157828e-05, "loss": 0.1126, "step": 6580 }, { "epoch": 0.31, "learning_rate": 1.691973056413135e-05, "loss": 0.122, "step": 6585 }, { "epoch": 0.31, "learning_rate": 1.6917391711104875e-05, "loss": 0.121, "step": 6590 }, { "epoch": 0.31, "learning_rate": 1.69150528580784e-05, "loss": 0.1518, "step": 6595 }, { "epoch": 0.31, "learning_rate": 1.6912714005051925e-05, "loss": 0.1011, "step": 6600 }, { "epoch": 0.31, "learning_rate": 1.6910375152025448e-05, "loss": 0.2477, "step": 6605 }, { "epoch": 0.31, "learning_rate": 1.6908036298998974e-05, "loss": 0.1324, "step": 6610 }, { "epoch": 0.31, "learning_rate": 1.6905697445972494e-05, "loss": 0.1319, "step": 6615 }, { "epoch": 0.31, "learning_rate": 1.690335859294602e-05, "loss": 0.0989, "step": 6620 }, { "epoch": 0.31, "learning_rate": 1.6901019739919544e-05, "loss": 0.1261, "step": 6625 }, { "epoch": 0.31, "learning_rate": 1.6898680886893067e-05, "loss": 0.1384, "step": 6630 }, { "epoch": 0.31, "learning_rate": 1.6896342033866594e-05, "loss": 0.039, "step": 6635 }, { "epoch": 0.31, "learning_rate": 1.6894003180840117e-05, "loss": 0.0692, "step": 6640 }, { "epoch": 0.31, "learning_rate": 1.689166432781364e-05, "loss": 0.052, "step": 6645 }, { "epoch": 0.31, "learning_rate": 1.6889325474787167e-05, "loss": 0.0998, "step": 6650 }, { "epoch": 0.31, "learning_rate": 1.688698662176069e-05, "loss": 0.0728, "step": 6655 }, { "epoch": 0.31, "learning_rate": 1.6884647768734214e-05, "loss": 0.0753, "step": 6660 }, { "epoch": 0.31, "learning_rate": 1.688230891570774e-05, "loss": 0.1005, "step": 6665 }, { "epoch": 0.31, "learning_rate": 1.6879970062681264e-05, "loss": 0.1094, "step": 6670 }, { "epoch": 0.31, "learning_rate": 1.6877631209654787e-05, "loss": 0.1506, "step": 6675 }, { "epoch": 0.31, "learning_rate": 1.687529235662831e-05, "loss": 0.0978, "step": 6680 }, { "epoch": 0.31, "learning_rate": 1.6872953503601833e-05, "loss": 0.1005, "step": 6685 }, { "epoch": 0.31, "learning_rate": 1.687061465057536e-05, "loss": 0.1367, "step": 6690 }, { "epoch": 0.31, "learning_rate": 1.6868275797548883e-05, "loss": 0.0825, "step": 6695 }, { "epoch": 0.31, "learning_rate": 1.6865936944522407e-05, "loss": 0.1156, "step": 6700 }, { "epoch": 0.31, "learning_rate": 1.6863598091495933e-05, "loss": 0.0973, "step": 6705 }, { "epoch": 0.31, "learning_rate": 1.6861259238469457e-05, "loss": 0.0917, "step": 6710 }, { "epoch": 0.31, "learning_rate": 1.685892038544298e-05, "loss": 0.1191, "step": 6715 }, { "epoch": 0.31, "learning_rate": 1.6856581532416506e-05, "loss": 0.1183, "step": 6720 }, { "epoch": 0.31, "learning_rate": 1.685424267939003e-05, "loss": 0.0776, "step": 6725 }, { "epoch": 0.31, "learning_rate": 1.6851903826363553e-05, "loss": 0.1647, "step": 6730 }, { "epoch": 0.32, "learning_rate": 1.6849564973337076e-05, "loss": 0.1275, "step": 6735 }, { "epoch": 0.32, "learning_rate": 1.68472261203106e-05, "loss": 0.1078, "step": 6740 }, { "epoch": 0.32, "learning_rate": 1.6844887267284126e-05, "loss": 0.1499, "step": 6745 }, { "epoch": 0.32, "learning_rate": 1.684254841425765e-05, "loss": 0.1472, "step": 6750 }, { "epoch": 0.32, "learning_rate": 1.6840209561231173e-05, "loss": 0.1581, "step": 6755 }, { "epoch": 0.32, "learning_rate": 1.68378707082047e-05, "loss": 0.1225, "step": 6760 }, { "epoch": 0.32, "learning_rate": 1.6835531855178223e-05, "loss": 0.1242, "step": 6765 }, { "epoch": 0.32, "learning_rate": 1.6833193002151746e-05, "loss": 0.1198, "step": 6770 }, { "epoch": 0.32, "learning_rate": 1.6830854149125272e-05, "loss": 0.107, "step": 6775 }, { "epoch": 0.32, "learning_rate": 1.6828515296098796e-05, "loss": 0.1078, "step": 6780 }, { "epoch": 0.32, "learning_rate": 1.682617644307232e-05, "loss": 0.1537, "step": 6785 }, { "epoch": 0.32, "learning_rate": 1.6823837590045846e-05, "loss": 0.0783, "step": 6790 }, { "epoch": 0.32, "learning_rate": 1.6821498737019365e-05, "loss": 0.1759, "step": 6795 }, { "epoch": 0.32, "learning_rate": 1.6819159883992892e-05, "loss": 0.165, "step": 6800 }, { "epoch": 0.32, "learning_rate": 1.6816821030966415e-05, "loss": 0.1316, "step": 6805 }, { "epoch": 0.32, "learning_rate": 1.681448217793994e-05, "loss": 0.1952, "step": 6810 }, { "epoch": 0.32, "learning_rate": 1.6812143324913462e-05, "loss": 0.0746, "step": 6815 }, { "epoch": 0.32, "learning_rate": 1.680980447188699e-05, "loss": 0.141, "step": 6820 }, { "epoch": 0.32, "learning_rate": 1.6807465618860512e-05, "loss": 0.1123, "step": 6825 }, { "epoch": 0.32, "learning_rate": 1.6805126765834035e-05, "loss": 0.1331, "step": 6830 }, { "epoch": 0.32, "learning_rate": 1.6802787912807562e-05, "loss": 0.0295, "step": 6835 }, { "epoch": 0.32, "learning_rate": 1.6800449059781085e-05, "loss": 0.1059, "step": 6840 }, { "epoch": 0.32, "learning_rate": 1.6798110206754608e-05, "loss": 0.1534, "step": 6845 }, { "epoch": 0.32, "learning_rate": 1.6795771353728135e-05, "loss": 0.1055, "step": 6850 }, { "epoch": 0.32, "learning_rate": 1.6793432500701655e-05, "loss": 0.0758, "step": 6855 }, { "epoch": 0.32, "learning_rate": 1.679109364767518e-05, "loss": 0.1005, "step": 6860 }, { "epoch": 0.32, "learning_rate": 1.6788754794648705e-05, "loss": 0.1825, "step": 6865 }, { "epoch": 0.32, "learning_rate": 1.6786415941622228e-05, "loss": 0.0905, "step": 6870 }, { "epoch": 0.32, "learning_rate": 1.6784077088595755e-05, "loss": 0.0901, "step": 6875 }, { "epoch": 0.32, "learning_rate": 1.6781738235569278e-05, "loss": 0.1331, "step": 6880 }, { "epoch": 0.32, "learning_rate": 1.67793993825428e-05, "loss": 0.0973, "step": 6885 }, { "epoch": 0.32, "learning_rate": 1.6777060529516328e-05, "loss": 0.1261, "step": 6890 }, { "epoch": 0.32, "learning_rate": 1.677472167648985e-05, "loss": 0.089, "step": 6895 }, { "epoch": 0.32, "learning_rate": 1.6772382823463374e-05, "loss": 0.1836, "step": 6900 }, { "epoch": 0.32, "learning_rate": 1.67700439704369e-05, "loss": 0.1246, "step": 6905 }, { "epoch": 0.32, "learning_rate": 1.6767705117410424e-05, "loss": 0.0625, "step": 6910 }, { "epoch": 0.32, "learning_rate": 1.6765366264383947e-05, "loss": 0.0843, "step": 6915 }, { "epoch": 0.32, "learning_rate": 1.676302741135747e-05, "loss": 0.1371, "step": 6920 }, { "epoch": 0.32, "learning_rate": 1.6760688558330994e-05, "loss": 0.0911, "step": 6925 }, { "epoch": 0.32, "learning_rate": 1.675834970530452e-05, "loss": 0.1658, "step": 6930 }, { "epoch": 0.32, "learning_rate": 1.6756010852278044e-05, "loss": 0.0313, "step": 6935 }, { "epoch": 0.32, "learning_rate": 1.6753671999251567e-05, "loss": 0.0969, "step": 6940 }, { "epoch": 0.32, "learning_rate": 1.6751333146225094e-05, "loss": 0.0845, "step": 6945 }, { "epoch": 0.33, "learning_rate": 1.6748994293198617e-05, "loss": 0.1152, "step": 6950 }, { "epoch": 0.33, "learning_rate": 1.674665544017214e-05, "loss": 0.0804, "step": 6955 }, { "epoch": 0.33, "learning_rate": 1.6744316587145667e-05, "loss": 0.1222, "step": 6960 }, { "epoch": 0.33, "learning_rate": 1.674197773411919e-05, "loss": 0.0843, "step": 6965 }, { "epoch": 0.33, "learning_rate": 1.6739638881092713e-05, "loss": 0.0908, "step": 6970 }, { "epoch": 0.33, "learning_rate": 1.6737300028066237e-05, "loss": 0.0862, "step": 6975 }, { "epoch": 0.33, "learning_rate": 1.673496117503976e-05, "loss": 0.0951, "step": 6980 }, { "epoch": 0.33, "learning_rate": 1.6732622322013287e-05, "loss": 0.1671, "step": 6985 }, { "epoch": 0.33, "learning_rate": 1.673028346898681e-05, "loss": 0.1157, "step": 6990 }, { "epoch": 0.33, "learning_rate": 1.6727944615960333e-05, "loss": 0.1142, "step": 6995 }, { "epoch": 0.33, "learning_rate": 1.672560576293386e-05, "loss": 0.1189, "step": 7000 }, { "epoch": 0.33, "learning_rate": 1.6723266909907383e-05, "loss": 0.1331, "step": 7005 }, { "epoch": 0.33, "learning_rate": 1.6720928056880906e-05, "loss": 0.1074, "step": 7010 }, { "epoch": 0.33, "learning_rate": 1.6718589203854433e-05, "loss": 0.1619, "step": 7015 }, { "epoch": 0.33, "learning_rate": 1.6716250350827956e-05, "loss": 0.0592, "step": 7020 }, { "epoch": 0.33, "learning_rate": 1.671391149780148e-05, "loss": 0.1238, "step": 7025 }, { "epoch": 0.33, "learning_rate": 1.6711572644775003e-05, "loss": 0.1353, "step": 7030 }, { "epoch": 0.33, "learning_rate": 1.6709233791748526e-05, "loss": 0.1394, "step": 7035 }, { "epoch": 0.33, "learning_rate": 1.6706894938722053e-05, "loss": 0.1172, "step": 7040 }, { "epoch": 0.33, "learning_rate": 1.6704556085695576e-05, "loss": 0.132, "step": 7045 }, { "epoch": 0.33, "learning_rate": 1.67022172326691e-05, "loss": 0.0762, "step": 7050 }, { "epoch": 0.33, "learning_rate": 1.6699878379642626e-05, "loss": 0.0823, "step": 7055 }, { "epoch": 0.33, "learning_rate": 1.669753952661615e-05, "loss": 0.1337, "step": 7060 }, { "epoch": 0.33, "learning_rate": 1.6695200673589672e-05, "loss": 0.0845, "step": 7065 }, { "epoch": 0.33, "learning_rate": 1.66928618205632e-05, "loss": 0.0804, "step": 7070 }, { "epoch": 0.33, "learning_rate": 1.6690522967536722e-05, "loss": 0.0697, "step": 7075 }, { "epoch": 0.33, "learning_rate": 1.6688184114510245e-05, "loss": 0.1212, "step": 7080 }, { "epoch": 0.33, "learning_rate": 1.6685845261483772e-05, "loss": 0.1088, "step": 7085 }, { "epoch": 0.33, "learning_rate": 1.6683506408457292e-05, "loss": 0.0584, "step": 7090 }, { "epoch": 0.33, "learning_rate": 1.668116755543082e-05, "loss": 0.0641, "step": 7095 }, { "epoch": 0.33, "learning_rate": 1.6678828702404342e-05, "loss": 0.1179, "step": 7100 }, { "epoch": 0.33, "learning_rate": 1.6676489849377865e-05, "loss": 0.0692, "step": 7105 }, { "epoch": 0.33, "learning_rate": 1.667415099635139e-05, "loss": 0.0987, "step": 7110 }, { "epoch": 0.33, "learning_rate": 1.6671812143324915e-05, "loss": 0.1067, "step": 7115 }, { "epoch": 0.33, "learning_rate": 1.6669473290298438e-05, "loss": 0.0741, "step": 7120 }, { "epoch": 0.33, "learning_rate": 1.6667134437271965e-05, "loss": 0.1072, "step": 7125 }, { "epoch": 0.33, "learning_rate": 1.6664795584245488e-05, "loss": 0.0647, "step": 7130 }, { "epoch": 0.33, "learning_rate": 1.666245673121901e-05, "loss": 0.17, "step": 7135 }, { "epoch": 0.33, "learning_rate": 1.6660117878192538e-05, "loss": 0.172, "step": 7140 }, { "epoch": 0.33, "learning_rate": 1.665777902516606e-05, "loss": 0.1293, "step": 7145 }, { "epoch": 0.33, "learning_rate": 1.6655440172139585e-05, "loss": 0.0621, "step": 7150 }, { "epoch": 0.33, "learning_rate": 1.6653101319113108e-05, "loss": 0.1039, "step": 7155 }, { "epoch": 0.33, "learning_rate": 1.665076246608663e-05, "loss": 0.1302, "step": 7160 }, { "epoch": 0.34, "learning_rate": 1.6648423613060158e-05, "loss": 0.1325, "step": 7165 }, { "epoch": 0.34, "learning_rate": 1.664608476003368e-05, "loss": 0.0625, "step": 7170 }, { "epoch": 0.34, "learning_rate": 1.6643745907007204e-05, "loss": 0.147, "step": 7175 }, { "epoch": 0.34, "learning_rate": 1.664140705398073e-05, "loss": 0.1237, "step": 7180 }, { "epoch": 0.34, "learning_rate": 1.6639068200954254e-05, "loss": 0.0546, "step": 7185 }, { "epoch": 0.34, "learning_rate": 1.6636729347927777e-05, "loss": 0.087, "step": 7190 }, { "epoch": 0.34, "learning_rate": 1.6634390494901304e-05, "loss": 0.1146, "step": 7195 }, { "epoch": 0.34, "learning_rate": 1.6632051641874827e-05, "loss": 0.0899, "step": 7200 }, { "epoch": 0.34, "learning_rate": 1.662971278884835e-05, "loss": 0.1752, "step": 7205 }, { "epoch": 0.34, "learning_rate": 1.6627373935821874e-05, "loss": 0.0597, "step": 7210 }, { "epoch": 0.34, "learning_rate": 1.6625035082795397e-05, "loss": 0.0775, "step": 7215 }, { "epoch": 0.34, "learning_rate": 1.6622696229768924e-05, "loss": 0.222, "step": 7220 }, { "epoch": 0.34, "learning_rate": 1.6620357376742447e-05, "loss": 0.1343, "step": 7225 }, { "epoch": 0.34, "learning_rate": 1.661801852371597e-05, "loss": 0.0691, "step": 7230 }, { "epoch": 0.34, "learning_rate": 1.6615679670689497e-05, "loss": 0.1277, "step": 7235 }, { "epoch": 0.34, "learning_rate": 1.661334081766302e-05, "loss": 0.0701, "step": 7240 }, { "epoch": 0.34, "learning_rate": 1.6611001964636543e-05, "loss": 0.1087, "step": 7245 }, { "epoch": 0.34, "learning_rate": 1.6608663111610067e-05, "loss": 0.0683, "step": 7250 }, { "epoch": 0.34, "learning_rate": 1.6606324258583593e-05, "loss": 0.1054, "step": 7255 }, { "epoch": 0.34, "learning_rate": 1.6603985405557117e-05, "loss": 0.0391, "step": 7260 }, { "epoch": 0.34, "learning_rate": 1.660164655253064e-05, "loss": 0.158, "step": 7265 }, { "epoch": 0.34, "learning_rate": 1.6599307699504163e-05, "loss": 0.1668, "step": 7270 }, { "epoch": 0.34, "learning_rate": 1.6596968846477686e-05, "loss": 0.0696, "step": 7275 }, { "epoch": 0.34, "learning_rate": 1.6594629993451213e-05, "loss": 0.0751, "step": 7280 }, { "epoch": 0.34, "learning_rate": 1.6592291140424736e-05, "loss": 0.0933, "step": 7285 }, { "epoch": 0.34, "learning_rate": 1.658995228739826e-05, "loss": 0.0718, "step": 7290 }, { "epoch": 0.34, "learning_rate": 1.6587613434371786e-05, "loss": 0.0881, "step": 7295 }, { "epoch": 0.34, "learning_rate": 1.658527458134531e-05, "loss": 0.1467, "step": 7300 }, { "epoch": 0.34, "learning_rate": 1.6582935728318833e-05, "loss": 0.1072, "step": 7305 }, { "epoch": 0.34, "learning_rate": 1.658059687529236e-05, "loss": 0.1191, "step": 7310 }, { "epoch": 0.34, "learning_rate": 1.6578258022265883e-05, "loss": 0.0577, "step": 7315 }, { "epoch": 0.34, "learning_rate": 1.6575919169239406e-05, "loss": 0.1761, "step": 7320 }, { "epoch": 0.34, "learning_rate": 1.6573580316212932e-05, "loss": 0.0711, "step": 7325 }, { "epoch": 0.34, "learning_rate": 1.6571241463186452e-05, "loss": 0.1358, "step": 7330 }, { "epoch": 0.34, "learning_rate": 1.656890261015998e-05, "loss": 0.1086, "step": 7335 }, { "epoch": 0.34, "learning_rate": 1.6566563757133502e-05, "loss": 0.0779, "step": 7340 }, { "epoch": 0.34, "learning_rate": 1.6564224904107025e-05, "loss": 0.134, "step": 7345 }, { "epoch": 0.34, "learning_rate": 1.6561886051080552e-05, "loss": 0.1032, "step": 7350 }, { "epoch": 0.34, "learning_rate": 1.6559547198054075e-05, "loss": 0.1301, "step": 7355 }, { "epoch": 0.34, "learning_rate": 1.65572083450276e-05, "loss": 0.0735, "step": 7360 }, { "epoch": 0.34, "learning_rate": 1.6554869492001125e-05, "loss": 0.1312, "step": 7365 }, { "epoch": 0.34, "learning_rate": 1.655253063897465e-05, "loss": 0.0782, "step": 7370 }, { "epoch": 0.34, "learning_rate": 1.6550191785948172e-05, "loss": 0.1487, "step": 7375 }, { "epoch": 0.35, "learning_rate": 1.65478529329217e-05, "loss": 0.1562, "step": 7380 }, { "epoch": 0.35, "learning_rate": 1.654551407989522e-05, "loss": 0.0838, "step": 7385 }, { "epoch": 0.35, "learning_rate": 1.6543175226868745e-05, "loss": 0.1075, "step": 7390 }, { "epoch": 0.35, "learning_rate": 1.6540836373842268e-05, "loss": 0.1126, "step": 7395 }, { "epoch": 0.35, "learning_rate": 1.653849752081579e-05, "loss": 0.1062, "step": 7400 }, { "epoch": 0.35, "learning_rate": 1.6536158667789318e-05, "loss": 0.1369, "step": 7405 }, { "epoch": 0.35, "learning_rate": 1.653381981476284e-05, "loss": 0.1552, "step": 7410 }, { "epoch": 0.35, "learning_rate": 1.6531480961736365e-05, "loss": 0.0623, "step": 7415 }, { "epoch": 0.35, "learning_rate": 1.652914210870989e-05, "loss": 0.1052, "step": 7420 }, { "epoch": 0.35, "learning_rate": 1.6526803255683415e-05, "loss": 0.1487, "step": 7425 }, { "epoch": 0.35, "learning_rate": 1.6524464402656938e-05, "loss": 0.1124, "step": 7430 }, { "epoch": 0.35, "learning_rate": 1.6522125549630464e-05, "loss": 0.1289, "step": 7435 }, { "epoch": 0.35, "learning_rate": 1.6519786696603988e-05, "loss": 0.0657, "step": 7440 }, { "epoch": 0.35, "learning_rate": 1.651744784357751e-05, "loss": 0.1621, "step": 7445 }, { "epoch": 0.35, "learning_rate": 1.6515108990551034e-05, "loss": 0.0733, "step": 7450 }, { "epoch": 0.35, "learning_rate": 1.6512770137524557e-05, "loss": 0.099, "step": 7455 }, { "epoch": 0.35, "learning_rate": 1.6510431284498084e-05, "loss": 0.0429, "step": 7460 }, { "epoch": 0.35, "learning_rate": 1.6508092431471607e-05, "loss": 0.1633, "step": 7465 }, { "epoch": 0.35, "learning_rate": 1.650575357844513e-05, "loss": 0.1196, "step": 7470 }, { "epoch": 0.35, "learning_rate": 1.6503414725418657e-05, "loss": 0.1196, "step": 7475 }, { "epoch": 0.35, "learning_rate": 1.650107587239218e-05, "loss": 0.1298, "step": 7480 }, { "epoch": 0.35, "learning_rate": 1.6498737019365704e-05, "loss": 0.1224, "step": 7485 }, { "epoch": 0.35, "learning_rate": 1.649639816633923e-05, "loss": 0.0375, "step": 7490 }, { "epoch": 0.35, "learning_rate": 1.6494059313312754e-05, "loss": 0.1402, "step": 7495 }, { "epoch": 0.35, "learning_rate": 1.6491720460286277e-05, "loss": 0.0861, "step": 7500 }, { "epoch": 0.35, "learning_rate": 1.64893816072598e-05, "loss": 0.1483, "step": 7505 }, { "epoch": 0.35, "learning_rate": 1.6487042754233323e-05, "loss": 0.1182, "step": 7510 }, { "epoch": 0.35, "learning_rate": 1.648470390120685e-05, "loss": 0.0874, "step": 7515 }, { "epoch": 0.35, "learning_rate": 1.6482365048180373e-05, "loss": 0.0808, "step": 7520 }, { "epoch": 0.35, "learning_rate": 1.6480026195153897e-05, "loss": 0.1351, "step": 7525 }, { "epoch": 0.35, "learning_rate": 1.6477687342127423e-05, "loss": 0.0757, "step": 7530 }, { "epoch": 0.35, "learning_rate": 1.6475348489100946e-05, "loss": 0.1187, "step": 7535 }, { "epoch": 0.35, "learning_rate": 1.647300963607447e-05, "loss": 0.1618, "step": 7540 }, { "epoch": 0.35, "learning_rate": 1.6470670783047996e-05, "loss": 0.0691, "step": 7545 }, { "epoch": 0.35, "learning_rate": 1.646833193002152e-05, "loss": 0.129, "step": 7550 }, { "epoch": 0.35, "learning_rate": 1.6465993076995043e-05, "loss": 0.1007, "step": 7555 }, { "epoch": 0.35, "learning_rate": 1.646365422396857e-05, "loss": 0.1052, "step": 7560 }, { "epoch": 0.35, "learning_rate": 1.646131537094209e-05, "loss": 0.1604, "step": 7565 }, { "epoch": 0.35, "learning_rate": 1.6458976517915616e-05, "loss": 0.1459, "step": 7570 }, { "epoch": 0.35, "learning_rate": 1.645663766488914e-05, "loss": 0.0804, "step": 7575 }, { "epoch": 0.35, "learning_rate": 1.6454298811862663e-05, "loss": 0.079, "step": 7580 }, { "epoch": 0.35, "learning_rate": 1.645195995883619e-05, "loss": 0.137, "step": 7585 }, { "epoch": 0.36, "learning_rate": 1.6449621105809712e-05, "loss": 0.1022, "step": 7590 }, { "epoch": 0.36, "learning_rate": 1.6447282252783236e-05, "loss": 0.203, "step": 7595 }, { "epoch": 0.36, "learning_rate": 1.6444943399756762e-05, "loss": 0.1801, "step": 7600 }, { "epoch": 0.36, "learning_rate": 1.6442604546730286e-05, "loss": 0.0844, "step": 7605 }, { "epoch": 0.36, "learning_rate": 1.644026569370381e-05, "loss": 0.0855, "step": 7610 }, { "epoch": 0.36, "learning_rate": 1.6437926840677336e-05, "loss": 0.1366, "step": 7615 }, { "epoch": 0.36, "learning_rate": 1.643558798765086e-05, "loss": 0.1631, "step": 7620 }, { "epoch": 0.36, "learning_rate": 1.6433249134624382e-05, "loss": 0.1212, "step": 7625 }, { "epoch": 0.36, "learning_rate": 1.6430910281597905e-05, "loss": 0.1088, "step": 7630 }, { "epoch": 0.36, "learning_rate": 1.642857142857143e-05, "loss": 0.0866, "step": 7635 }, { "epoch": 0.36, "learning_rate": 1.6426232575544955e-05, "loss": 0.0825, "step": 7640 }, { "epoch": 0.36, "learning_rate": 1.642389372251848e-05, "loss": 0.1468, "step": 7645 }, { "epoch": 0.36, "learning_rate": 1.6421554869492002e-05, "loss": 0.0575, "step": 7650 }, { "epoch": 0.36, "learning_rate": 1.641921601646553e-05, "loss": 0.0575, "step": 7655 }, { "epoch": 0.36, "learning_rate": 1.641687716343905e-05, "loss": 0.1131, "step": 7660 }, { "epoch": 0.36, "learning_rate": 1.6414538310412575e-05, "loss": 0.0993, "step": 7665 }, { "epoch": 0.36, "learning_rate": 1.64121994573861e-05, "loss": 0.1554, "step": 7670 }, { "epoch": 0.36, "learning_rate": 1.6409860604359625e-05, "loss": 0.1423, "step": 7675 }, { "epoch": 0.36, "learning_rate": 1.6407521751333148e-05, "loss": 0.1265, "step": 7680 }, { "epoch": 0.36, "learning_rate": 1.640518289830667e-05, "loss": 0.0592, "step": 7685 }, { "epoch": 0.36, "learning_rate": 1.6402844045280195e-05, "loss": 0.0759, "step": 7690 }, { "epoch": 0.36, "learning_rate": 1.6400505192253718e-05, "loss": 0.1562, "step": 7695 }, { "epoch": 0.36, "learning_rate": 1.6398166339227244e-05, "loss": 0.0847, "step": 7700 }, { "epoch": 0.36, "learning_rate": 1.6395827486200768e-05, "loss": 0.1418, "step": 7705 }, { "epoch": 0.36, "learning_rate": 1.639348863317429e-05, "loss": 0.1047, "step": 7710 }, { "epoch": 0.36, "learning_rate": 1.6391149780147818e-05, "loss": 0.1024, "step": 7715 }, { "epoch": 0.36, "learning_rate": 1.638881092712134e-05, "loss": 0.0775, "step": 7720 }, { "epoch": 0.36, "learning_rate": 1.6386472074094864e-05, "loss": 0.1209, "step": 7725 }, { "epoch": 0.36, "learning_rate": 1.638413322106839e-05, "loss": 0.0958, "step": 7730 }, { "epoch": 0.36, "learning_rate": 1.6381794368041914e-05, "loss": 0.1357, "step": 7735 }, { "epoch": 0.36, "learning_rate": 1.6379455515015437e-05, "loss": 0.053, "step": 7740 }, { "epoch": 0.36, "learning_rate": 1.637711666198896e-05, "loss": 0.1169, "step": 7745 }, { "epoch": 0.36, "learning_rate": 1.6374777808962484e-05, "loss": 0.0811, "step": 7750 }, { "epoch": 0.36, "learning_rate": 1.637243895593601e-05, "loss": 0.0466, "step": 7755 }, { "epoch": 0.36, "learning_rate": 1.6370100102909534e-05, "loss": 0.1068, "step": 7760 }, { "epoch": 0.36, "learning_rate": 1.6367761249883057e-05, "loss": 0.0984, "step": 7765 }, { "epoch": 0.36, "learning_rate": 1.6365422396856584e-05, "loss": 0.1488, "step": 7770 }, { "epoch": 0.36, "learning_rate": 1.6363083543830107e-05, "loss": 0.1227, "step": 7775 }, { "epoch": 0.36, "learning_rate": 1.636074469080363e-05, "loss": 0.1157, "step": 7780 }, { "epoch": 0.36, "learning_rate": 1.6358405837777157e-05, "loss": 0.0999, "step": 7785 }, { "epoch": 0.36, "learning_rate": 1.635606698475068e-05, "loss": 0.0554, "step": 7790 }, { "epoch": 0.36, "learning_rate": 1.6353728131724203e-05, "loss": 0.1406, "step": 7795 }, { "epoch": 0.36, "learning_rate": 1.635138927869773e-05, "loss": 0.0833, "step": 7800 }, { "epoch": 0.37, "learning_rate": 1.634905042567125e-05, "loss": 0.0472, "step": 7805 }, { "epoch": 0.37, "learning_rate": 1.6346711572644776e-05, "loss": 0.1029, "step": 7810 }, { "epoch": 0.37, "learning_rate": 1.63443727196183e-05, "loss": 0.0789, "step": 7815 }, { "epoch": 0.37, "learning_rate": 1.6342033866591823e-05, "loss": 0.1559, "step": 7820 }, { "epoch": 0.37, "learning_rate": 1.633969501356535e-05, "loss": 0.1535, "step": 7825 }, { "epoch": 0.37, "learning_rate": 1.6337356160538873e-05, "loss": 0.1119, "step": 7830 }, { "epoch": 0.37, "learning_rate": 1.6335017307512396e-05, "loss": 0.1343, "step": 7835 }, { "epoch": 0.37, "learning_rate": 1.6332678454485923e-05, "loss": 0.0998, "step": 7840 }, { "epoch": 0.37, "learning_rate": 1.6330339601459446e-05, "loss": 0.1374, "step": 7845 }, { "epoch": 0.37, "learning_rate": 1.632800074843297e-05, "loss": 0.1796, "step": 7850 }, { "epoch": 0.37, "learning_rate": 1.6325661895406496e-05, "loss": 0.1027, "step": 7855 }, { "epoch": 0.37, "learning_rate": 1.6323323042380016e-05, "loss": 0.1236, "step": 7860 }, { "epoch": 0.37, "learning_rate": 1.6320984189353542e-05, "loss": 0.1228, "step": 7865 }, { "epoch": 0.37, "learning_rate": 1.6318645336327066e-05, "loss": 0.1788, "step": 7870 }, { "epoch": 0.37, "learning_rate": 1.631630648330059e-05, "loss": 0.1449, "step": 7875 }, { "epoch": 0.37, "learning_rate": 1.6313967630274116e-05, "loss": 0.0679, "step": 7880 }, { "epoch": 0.37, "learning_rate": 1.631162877724764e-05, "loss": 0.1119, "step": 7885 }, { "epoch": 0.37, "learning_rate": 1.6309289924221162e-05, "loss": 0.2058, "step": 7890 }, { "epoch": 0.37, "learning_rate": 1.630695107119469e-05, "loss": 0.0679, "step": 7895 }, { "epoch": 0.37, "learning_rate": 1.6304612218168212e-05, "loss": 0.1276, "step": 7900 }, { "epoch": 0.37, "learning_rate": 1.6302273365141735e-05, "loss": 0.1167, "step": 7905 }, { "epoch": 0.37, "learning_rate": 1.6299934512115262e-05, "loss": 0.1218, "step": 7910 }, { "epoch": 0.37, "learning_rate": 1.6297595659088785e-05, "loss": 0.0362, "step": 7915 }, { "epoch": 0.37, "learning_rate": 1.629525680606231e-05, "loss": 0.0688, "step": 7920 }, { "epoch": 0.37, "learning_rate": 1.6292917953035832e-05, "loss": 0.0673, "step": 7925 }, { "epoch": 0.37, "learning_rate": 1.6290579100009355e-05, "loss": 0.0768, "step": 7930 }, { "epoch": 0.37, "learning_rate": 1.628824024698288e-05, "loss": 0.1718, "step": 7935 }, { "epoch": 0.37, "learning_rate": 1.6285901393956405e-05, "loss": 0.0794, "step": 7940 }, { "epoch": 0.37, "learning_rate": 1.6283562540929928e-05, "loss": 0.1075, "step": 7945 }, { "epoch": 0.37, "learning_rate": 1.6281223687903455e-05, "loss": 0.1623, "step": 7950 }, { "epoch": 0.37, "learning_rate": 1.6278884834876978e-05, "loss": 0.1379, "step": 7955 }, { "epoch": 0.37, "learning_rate": 1.62765459818505e-05, "loss": 0.1353, "step": 7960 }, { "epoch": 0.37, "learning_rate": 1.6274207128824028e-05, "loss": 0.0963, "step": 7965 }, { "epoch": 0.37, "learning_rate": 1.627186827579755e-05, "loss": 0.0803, "step": 7970 }, { "epoch": 0.37, "learning_rate": 1.6269529422771074e-05, "loss": 0.1212, "step": 7975 }, { "epoch": 0.37, "learning_rate": 1.6267190569744598e-05, "loss": 0.0668, "step": 7980 }, { "epoch": 0.37, "learning_rate": 1.626485171671812e-05, "loss": 0.0834, "step": 7985 }, { "epoch": 0.37, "learning_rate": 1.6262512863691648e-05, "loss": 0.1094, "step": 7990 }, { "epoch": 0.37, "learning_rate": 1.626017401066517e-05, "loss": 0.1138, "step": 7995 }, { "epoch": 0.37, "learning_rate": 1.6257835157638694e-05, "loss": 0.0879, "step": 8000 }, { "epoch": 0.37, "learning_rate": 1.625549630461222e-05, "loss": 0.1509, "step": 8005 }, { "epoch": 0.37, "learning_rate": 1.6253157451585744e-05, "loss": 0.1189, "step": 8010 }, { "epoch": 0.37, "learning_rate": 1.6250818598559267e-05, "loss": 0.1283, "step": 8015 }, { "epoch": 0.38, "learning_rate": 1.6248479745532794e-05, "loss": 0.1045, "step": 8020 }, { "epoch": 0.38, "learning_rate": 1.6246140892506317e-05, "loss": 0.0973, "step": 8025 }, { "epoch": 0.38, "learning_rate": 1.624380203947984e-05, "loss": 0.1792, "step": 8030 }, { "epoch": 0.38, "learning_rate": 1.6241463186453367e-05, "loss": 0.0901, "step": 8035 }, { "epoch": 0.38, "learning_rate": 1.6239124333426887e-05, "loss": 0.1172, "step": 8040 }, { "epoch": 0.38, "learning_rate": 1.6236785480400414e-05, "loss": 0.1337, "step": 8045 }, { "epoch": 0.38, "learning_rate": 1.6234446627373937e-05, "loss": 0.1312, "step": 8050 }, { "epoch": 0.38, "learning_rate": 1.623210777434746e-05, "loss": 0.0783, "step": 8055 }, { "epoch": 0.38, "learning_rate": 1.6229768921320987e-05, "loss": 0.0864, "step": 8060 }, { "epoch": 0.38, "learning_rate": 1.622743006829451e-05, "loss": 0.1001, "step": 8065 }, { "epoch": 0.38, "learning_rate": 1.6225091215268033e-05, "loss": 0.1223, "step": 8070 }, { "epoch": 0.38, "learning_rate": 1.622275236224156e-05, "loss": 0.1328, "step": 8075 }, { "epoch": 0.38, "learning_rate": 1.6220413509215083e-05, "loss": 0.0673, "step": 8080 }, { "epoch": 0.38, "learning_rate": 1.6218074656188606e-05, "loss": 0.1455, "step": 8085 }, { "epoch": 0.38, "learning_rate": 1.6215735803162133e-05, "loss": 0.1224, "step": 8090 }, { "epoch": 0.38, "learning_rate": 1.6213396950135656e-05, "loss": 0.0645, "step": 8095 }, { "epoch": 0.38, "learning_rate": 1.621105809710918e-05, "loss": 0.0723, "step": 8100 }, { "epoch": 0.38, "learning_rate": 1.6208719244082703e-05, "loss": 0.046, "step": 8105 }, { "epoch": 0.38, "learning_rate": 1.6206380391056226e-05, "loss": 0.0819, "step": 8110 }, { "epoch": 0.38, "learning_rate": 1.6204041538029753e-05, "loss": 0.0949, "step": 8115 }, { "epoch": 0.38, "learning_rate": 1.6201702685003276e-05, "loss": 0.1096, "step": 8120 }, { "epoch": 0.38, "learning_rate": 1.61993638319768e-05, "loss": 0.0435, "step": 8125 }, { "epoch": 0.38, "learning_rate": 1.6197024978950323e-05, "loss": 0.0781, "step": 8130 }, { "epoch": 0.38, "learning_rate": 1.619468612592385e-05, "loss": 0.1571, "step": 8135 }, { "epoch": 0.38, "learning_rate": 1.6192347272897372e-05, "loss": 0.0999, "step": 8140 }, { "epoch": 0.38, "learning_rate": 1.6190008419870896e-05, "loss": 0.1843, "step": 8145 }, { "epoch": 0.38, "learning_rate": 1.6187669566844422e-05, "loss": 0.1174, "step": 8150 }, { "epoch": 0.38, "learning_rate": 1.6185330713817946e-05, "loss": 0.1529, "step": 8155 }, { "epoch": 0.38, "learning_rate": 1.618299186079147e-05, "loss": 0.228, "step": 8160 }, { "epoch": 0.38, "learning_rate": 1.6180653007764992e-05, "loss": 0.0651, "step": 8165 }, { "epoch": 0.38, "learning_rate": 1.6178314154738515e-05, "loss": 0.1202, "step": 8170 }, { "epoch": 0.38, "learning_rate": 1.6175975301712042e-05, "loss": 0.104, "step": 8175 }, { "epoch": 0.38, "learning_rate": 1.6173636448685565e-05, "loss": 0.0726, "step": 8180 }, { "epoch": 0.38, "learning_rate": 1.617129759565909e-05, "loss": 0.1289, "step": 8185 }, { "epoch": 0.38, "learning_rate": 1.6168958742632615e-05, "loss": 0.0721, "step": 8190 }, { "epoch": 0.38, "learning_rate": 1.616661988960614e-05, "loss": 0.1209, "step": 8195 }, { "epoch": 0.38, "learning_rate": 1.6164281036579662e-05, "loss": 0.1233, "step": 8200 }, { "epoch": 0.38, "learning_rate": 1.616194218355319e-05, "loss": 0.0949, "step": 8205 }, { "epoch": 0.38, "learning_rate": 1.615960333052671e-05, "loss": 0.0999, "step": 8210 }, { "epoch": 0.38, "learning_rate": 1.6157264477500235e-05, "loss": 0.1743, "step": 8215 }, { "epoch": 0.38, "learning_rate": 1.6154925624473758e-05, "loss": 0.1293, "step": 8220 }, { "epoch": 0.38, "learning_rate": 1.615258677144728e-05, "loss": 0.0842, "step": 8225 }, { "epoch": 0.38, "learning_rate": 1.6150247918420808e-05, "loss": 0.1785, "step": 8230 }, { "epoch": 0.39, "learning_rate": 1.614790906539433e-05, "loss": 0.108, "step": 8235 }, { "epoch": 0.39, "learning_rate": 1.6145570212367855e-05, "loss": 0.0787, "step": 8240 }, { "epoch": 0.39, "learning_rate": 1.614323135934138e-05, "loss": 0.097, "step": 8245 }, { "epoch": 0.39, "learning_rate": 1.6140892506314904e-05, "loss": 0.1809, "step": 8250 }, { "epoch": 0.39, "learning_rate": 1.6138553653288428e-05, "loss": 0.0847, "step": 8255 }, { "epoch": 0.39, "learning_rate": 1.6136214800261954e-05, "loss": 0.0883, "step": 8260 }, { "epoch": 0.39, "learning_rate": 1.6133875947235478e-05, "loss": 0.0786, "step": 8265 }, { "epoch": 0.39, "learning_rate": 1.6131537094209e-05, "loss": 0.1175, "step": 8270 }, { "epoch": 0.39, "learning_rate": 1.6129198241182524e-05, "loss": 0.1252, "step": 8275 }, { "epoch": 0.39, "learning_rate": 1.6126859388156047e-05, "loss": 0.1285, "step": 8280 }, { "epoch": 0.39, "learning_rate": 1.6124520535129574e-05, "loss": 0.0952, "step": 8285 }, { "epoch": 0.39, "learning_rate": 1.6122181682103097e-05, "loss": 0.1094, "step": 8290 }, { "epoch": 0.39, "learning_rate": 1.611984282907662e-05, "loss": 0.1183, "step": 8295 }, { "epoch": 0.39, "learning_rate": 1.6117503976050147e-05, "loss": 0.1283, "step": 8300 }, { "epoch": 0.39, "learning_rate": 1.611516512302367e-05, "loss": 0.1181, "step": 8305 }, { "epoch": 0.39, "learning_rate": 1.6112826269997194e-05, "loss": 0.1241, "step": 8310 }, { "epoch": 0.39, "learning_rate": 1.611048741697072e-05, "loss": 0.0698, "step": 8315 }, { "epoch": 0.39, "learning_rate": 1.6108148563944244e-05, "loss": 0.1226, "step": 8320 }, { "epoch": 0.39, "learning_rate": 1.6105809710917767e-05, "loss": 0.0844, "step": 8325 }, { "epoch": 0.39, "learning_rate": 1.6103470857891294e-05, "loss": 0.1387, "step": 8330 }, { "epoch": 0.39, "learning_rate": 1.6101132004864813e-05, "loss": 0.1195, "step": 8335 }, { "epoch": 0.39, "learning_rate": 1.609879315183834e-05, "loss": 0.0623, "step": 8340 }, { "epoch": 0.39, "learning_rate": 1.6096454298811863e-05, "loss": 0.1597, "step": 8345 }, { "epoch": 0.39, "learning_rate": 1.6094115445785387e-05, "loss": 0.1297, "step": 8350 }, { "epoch": 0.39, "learning_rate": 1.6091776592758913e-05, "loss": 0.1391, "step": 8355 }, { "epoch": 0.39, "learning_rate": 1.6089437739732436e-05, "loss": 0.0677, "step": 8360 }, { "epoch": 0.39, "learning_rate": 1.608709888670596e-05, "loss": 0.0878, "step": 8365 }, { "epoch": 0.39, "learning_rate": 1.6084760033679486e-05, "loss": 0.1305, "step": 8370 }, { "epoch": 0.39, "learning_rate": 1.608242118065301e-05, "loss": 0.083, "step": 8375 }, { "epoch": 0.39, "learning_rate": 1.6080082327626533e-05, "loss": 0.1179, "step": 8380 }, { "epoch": 0.39, "learning_rate": 1.607774347460006e-05, "loss": 0.1449, "step": 8385 }, { "epoch": 0.39, "learning_rate": 1.6075404621573583e-05, "loss": 0.1213, "step": 8390 }, { "epoch": 0.39, "learning_rate": 1.6073065768547106e-05, "loss": 0.0718, "step": 8395 }, { "epoch": 0.39, "learning_rate": 1.607072691552063e-05, "loss": 0.0987, "step": 8400 }, { "epoch": 0.39, "learning_rate": 1.6068388062494153e-05, "loss": 0.1246, "step": 8405 }, { "epoch": 0.39, "learning_rate": 1.606604920946768e-05, "loss": 0.1396, "step": 8410 }, { "epoch": 0.39, "learning_rate": 1.6063710356441202e-05, "loss": 0.0901, "step": 8415 }, { "epoch": 0.39, "learning_rate": 1.6061371503414726e-05, "loss": 0.1063, "step": 8420 }, { "epoch": 0.39, "learning_rate": 1.6059032650388252e-05, "loss": 0.1305, "step": 8425 }, { "epoch": 0.39, "learning_rate": 1.6056693797361776e-05, "loss": 0.0608, "step": 8430 }, { "epoch": 0.39, "learning_rate": 1.60543549443353e-05, "loss": 0.1023, "step": 8435 }, { "epoch": 0.39, "learning_rate": 1.6052016091308825e-05, "loss": 0.0447, "step": 8440 }, { "epoch": 0.4, "learning_rate": 1.604967723828235e-05, "loss": 0.1234, "step": 8445 }, { "epoch": 0.4, "learning_rate": 1.6047338385255872e-05, "loss": 0.1449, "step": 8450 }, { "epoch": 0.4, "learning_rate": 1.6044999532229395e-05, "loss": 0.0927, "step": 8455 }, { "epoch": 0.4, "learning_rate": 1.604266067920292e-05, "loss": 0.1493, "step": 8460 }, { "epoch": 0.4, "learning_rate": 1.6040321826176445e-05, "loss": 0.1089, "step": 8465 }, { "epoch": 0.4, "learning_rate": 1.603798297314997e-05, "loss": 0.2196, "step": 8470 }, { "epoch": 0.4, "learning_rate": 1.603564412012349e-05, "loss": 0.0723, "step": 8475 }, { "epoch": 0.4, "learning_rate": 1.603330526709702e-05, "loss": 0.1013, "step": 8480 }, { "epoch": 0.4, "learning_rate": 1.603096641407054e-05, "loss": 0.0871, "step": 8485 }, { "epoch": 0.4, "learning_rate": 1.6028627561044065e-05, "loss": 0.0453, "step": 8490 }, { "epoch": 0.4, "learning_rate": 1.602628870801759e-05, "loss": 0.1366, "step": 8495 }, { "epoch": 0.4, "learning_rate": 1.6023949854991115e-05, "loss": 0.1308, "step": 8500 }, { "epoch": 0.4, "learning_rate": 1.6021611001964638e-05, "loss": 0.0758, "step": 8505 }, { "epoch": 0.4, "learning_rate": 1.6019272148938165e-05, "loss": 0.0794, "step": 8510 }, { "epoch": 0.4, "learning_rate": 1.6016933295911685e-05, "loss": 0.1209, "step": 8515 }, { "epoch": 0.4, "learning_rate": 1.601459444288521e-05, "loss": 0.1171, "step": 8520 }, { "epoch": 0.4, "learning_rate": 1.6012255589858734e-05, "loss": 0.1446, "step": 8525 }, { "epoch": 0.4, "learning_rate": 1.6009916736832258e-05, "loss": 0.1056, "step": 8530 }, { "epoch": 0.4, "learning_rate": 1.6007577883805784e-05, "loss": 0.0988, "step": 8535 }, { "epoch": 0.4, "learning_rate": 1.6005239030779308e-05, "loss": 0.1102, "step": 8540 }, { "epoch": 0.4, "learning_rate": 1.600290017775283e-05, "loss": 0.1084, "step": 8545 }, { "epoch": 0.4, "learning_rate": 1.6000561324726357e-05, "loss": 0.1526, "step": 8550 }, { "epoch": 0.4, "learning_rate": 1.599822247169988e-05, "loss": 0.1556, "step": 8555 }, { "epoch": 0.4, "learning_rate": 1.5995883618673404e-05, "loss": 0.1229, "step": 8560 }, { "epoch": 0.4, "learning_rate": 1.5993544765646927e-05, "loss": 0.0635, "step": 8565 }, { "epoch": 0.4, "learning_rate": 1.5991205912620454e-05, "loss": 0.1141, "step": 8570 }, { "epoch": 0.4, "learning_rate": 1.5988867059593974e-05, "loss": 0.0664, "step": 8575 }, { "epoch": 0.4, "learning_rate": 1.59865282065675e-05, "loss": 0.1838, "step": 8580 }, { "epoch": 0.4, "learning_rate": 1.5984189353541024e-05, "loss": 0.0544, "step": 8585 }, { "epoch": 0.4, "learning_rate": 1.5981850500514547e-05, "loss": 0.1693, "step": 8590 }, { "epoch": 0.4, "learning_rate": 1.5979511647488074e-05, "loss": 0.1472, "step": 8595 }, { "epoch": 0.4, "learning_rate": 1.5977172794461597e-05, "loss": 0.1055, "step": 8600 }, { "epoch": 0.4, "learning_rate": 1.597483394143512e-05, "loss": 0.1873, "step": 8605 }, { "epoch": 0.4, "learning_rate": 1.5972495088408647e-05, "loss": 0.1496, "step": 8610 }, { "epoch": 0.4, "learning_rate": 1.597015623538217e-05, "loss": 0.1636, "step": 8615 }, { "epoch": 0.4, "learning_rate": 1.5967817382355693e-05, "loss": 0.0892, "step": 8620 }, { "epoch": 0.4, "learning_rate": 1.596547852932922e-05, "loss": 0.132, "step": 8625 }, { "epoch": 0.4, "learning_rate": 1.5963139676302743e-05, "loss": 0.0997, "step": 8630 }, { "epoch": 0.4, "learning_rate": 1.5960800823276266e-05, "loss": 0.1659, "step": 8635 }, { "epoch": 0.4, "learning_rate": 1.595846197024979e-05, "loss": 0.135, "step": 8640 }, { "epoch": 0.4, "learning_rate": 1.5956123117223313e-05, "loss": 0.0909, "step": 8645 }, { "epoch": 0.4, "learning_rate": 1.595378426419684e-05, "loss": 0.1051, "step": 8650 }, { "epoch": 0.4, "learning_rate": 1.5951445411170363e-05, "loss": 0.1009, "step": 8655 }, { "epoch": 0.41, "learning_rate": 1.5949106558143886e-05, "loss": 0.1427, "step": 8660 }, { "epoch": 0.41, "learning_rate": 1.5946767705117413e-05, "loss": 0.1142, "step": 8665 }, { "epoch": 0.41, "learning_rate": 1.5944428852090936e-05, "loss": 0.0823, "step": 8670 }, { "epoch": 0.41, "learning_rate": 1.594208999906446e-05, "loss": 0.0836, "step": 8675 }, { "epoch": 0.41, "learning_rate": 1.5939751146037986e-05, "loss": 0.0889, "step": 8680 }, { "epoch": 0.41, "learning_rate": 1.593741229301151e-05, "loss": 0.0729, "step": 8685 }, { "epoch": 0.41, "learning_rate": 1.5935073439985032e-05, "loss": 0.1189, "step": 8690 }, { "epoch": 0.41, "learning_rate": 1.5932734586958556e-05, "loss": 0.1298, "step": 8695 }, { "epoch": 0.41, "learning_rate": 1.593039573393208e-05, "loss": 0.1052, "step": 8700 }, { "epoch": 0.41, "learning_rate": 1.5928056880905606e-05, "loss": 0.0708, "step": 8705 }, { "epoch": 0.41, "learning_rate": 1.592571802787913e-05, "loss": 0.0615, "step": 8710 }, { "epoch": 0.41, "learning_rate": 1.5923379174852652e-05, "loss": 0.071, "step": 8715 }, { "epoch": 0.41, "learning_rate": 1.592104032182618e-05, "loss": 0.0892, "step": 8720 }, { "epoch": 0.41, "learning_rate": 1.5918701468799702e-05, "loss": 0.0306, "step": 8725 }, { "epoch": 0.41, "learning_rate": 1.5916362615773225e-05, "loss": 0.0891, "step": 8730 }, { "epoch": 0.41, "learning_rate": 1.5914023762746752e-05, "loss": 0.1454, "step": 8735 }, { "epoch": 0.41, "learning_rate": 1.5911684909720275e-05, "loss": 0.1364, "step": 8740 }, { "epoch": 0.41, "learning_rate": 1.59093460566938e-05, "loss": 0.0757, "step": 8745 }, { "epoch": 0.41, "learning_rate": 1.590700720366732e-05, "loss": 0.096, "step": 8750 }, { "epoch": 0.41, "learning_rate": 1.5904668350640845e-05, "loss": 0.0953, "step": 8755 }, { "epoch": 0.41, "learning_rate": 1.590232949761437e-05, "loss": 0.132, "step": 8760 }, { "epoch": 0.41, "learning_rate": 1.5899990644587895e-05, "loss": 0.1256, "step": 8765 }, { "epoch": 0.41, "learning_rate": 1.5897651791561418e-05, "loss": 0.084, "step": 8770 }, { "epoch": 0.41, "learning_rate": 1.5895312938534945e-05, "loss": 0.1199, "step": 8775 }, { "epoch": 0.41, "learning_rate": 1.5892974085508468e-05, "loss": 0.1576, "step": 8780 }, { "epoch": 0.41, "learning_rate": 1.589063523248199e-05, "loss": 0.0992, "step": 8785 }, { "epoch": 0.41, "learning_rate": 1.5888296379455518e-05, "loss": 0.1266, "step": 8790 }, { "epoch": 0.41, "learning_rate": 1.588595752642904e-05, "loss": 0.1397, "step": 8795 }, { "epoch": 0.41, "learning_rate": 1.5883618673402564e-05, "loss": 0.1034, "step": 8800 }, { "epoch": 0.41, "learning_rate": 1.588127982037609e-05, "loss": 0.0974, "step": 8805 }, { "epoch": 0.41, "learning_rate": 1.587894096734961e-05, "loss": 0.0762, "step": 8810 }, { "epoch": 0.41, "learning_rate": 1.5876602114323138e-05, "loss": 0.1275, "step": 8815 }, { "epoch": 0.41, "learning_rate": 1.587426326129666e-05, "loss": 0.1555, "step": 8820 }, { "epoch": 0.41, "learning_rate": 1.5871924408270184e-05, "loss": 0.0678, "step": 8825 }, { "epoch": 0.41, "learning_rate": 1.586958555524371e-05, "loss": 0.074, "step": 8830 }, { "epoch": 0.41, "learning_rate": 1.5867246702217234e-05, "loss": 0.0726, "step": 8835 }, { "epoch": 0.41, "learning_rate": 1.5864907849190757e-05, "loss": 0.0975, "step": 8840 }, { "epoch": 0.41, "learning_rate": 1.5862568996164284e-05, "loss": 0.1079, "step": 8845 }, { "epoch": 0.41, "learning_rate": 1.5860230143137807e-05, "loss": 0.1164, "step": 8850 }, { "epoch": 0.41, "learning_rate": 1.585789129011133e-05, "loss": 0.0416, "step": 8855 }, { "epoch": 0.41, "learning_rate": 1.5855552437084857e-05, "loss": 0.1409, "step": 8860 }, { "epoch": 0.41, "learning_rate": 1.585321358405838e-05, "loss": 0.1926, "step": 8865 }, { "epoch": 0.41, "learning_rate": 1.5850874731031904e-05, "loss": 0.1056, "step": 8870 }, { "epoch": 0.42, "learning_rate": 1.5848535878005427e-05, "loss": 0.0421, "step": 8875 }, { "epoch": 0.42, "learning_rate": 1.584619702497895e-05, "loss": 0.0807, "step": 8880 }, { "epoch": 0.42, "learning_rate": 1.5843858171952477e-05, "loss": 0.0976, "step": 8885 }, { "epoch": 0.42, "learning_rate": 1.5841519318926e-05, "loss": 0.1432, "step": 8890 }, { "epoch": 0.42, "learning_rate": 1.5839180465899523e-05, "loss": 0.1411, "step": 8895 }, { "epoch": 0.42, "learning_rate": 1.583684161287305e-05, "loss": 0.1555, "step": 8900 }, { "epoch": 0.42, "learning_rate": 1.5834502759846573e-05, "loss": 0.1474, "step": 8905 }, { "epoch": 0.42, "learning_rate": 1.5832163906820096e-05, "loss": 0.0391, "step": 8910 }, { "epoch": 0.42, "learning_rate": 1.5829825053793623e-05, "loss": 0.0867, "step": 8915 }, { "epoch": 0.42, "learning_rate": 1.5827486200767146e-05, "loss": 0.1154, "step": 8920 }, { "epoch": 0.42, "learning_rate": 1.582514734774067e-05, "loss": 0.0938, "step": 8925 }, { "epoch": 0.42, "learning_rate": 1.5822808494714193e-05, "loss": 0.1142, "step": 8930 }, { "epoch": 0.42, "learning_rate": 1.5820469641687716e-05, "loss": 0.065, "step": 8935 }, { "epoch": 0.42, "learning_rate": 1.5818130788661243e-05, "loss": 0.0914, "step": 8940 }, { "epoch": 0.42, "learning_rate": 1.5815791935634766e-05, "loss": 0.119, "step": 8945 }, { "epoch": 0.42, "learning_rate": 1.581345308260829e-05, "loss": 0.1138, "step": 8950 }, { "epoch": 0.42, "learning_rate": 1.5811114229581816e-05, "loss": 0.1073, "step": 8955 }, { "epoch": 0.42, "learning_rate": 1.580877537655534e-05, "loss": 0.2016, "step": 8960 }, { "epoch": 0.42, "learning_rate": 1.5806436523528862e-05, "loss": 0.077, "step": 8965 }, { "epoch": 0.42, "learning_rate": 1.580409767050239e-05, "loss": 0.1081, "step": 8970 }, { "epoch": 0.42, "learning_rate": 1.5801758817475912e-05, "loss": 0.139, "step": 8975 }, { "epoch": 0.42, "learning_rate": 1.5799419964449436e-05, "loss": 0.1, "step": 8980 }, { "epoch": 0.42, "learning_rate": 1.5797081111422962e-05, "loss": 0.0761, "step": 8985 }, { "epoch": 0.42, "learning_rate": 1.5794742258396482e-05, "loss": 0.1015, "step": 8990 }, { "epoch": 0.42, "learning_rate": 1.579240340537001e-05, "loss": 0.0964, "step": 8995 }, { "epoch": 0.42, "learning_rate": 1.5790064552343532e-05, "loss": 0.0806, "step": 9000 }, { "epoch": 0.42, "learning_rate": 1.5787725699317055e-05, "loss": 0.0793, "step": 9005 }, { "epoch": 0.42, "learning_rate": 1.578538684629058e-05, "loss": 0.1217, "step": 9010 }, { "epoch": 0.42, "learning_rate": 1.5783047993264105e-05, "loss": 0.1003, "step": 9015 }, { "epoch": 0.42, "learning_rate": 1.578070914023763e-05, "loss": 0.1059, "step": 9020 }, { "epoch": 0.42, "learning_rate": 1.577837028721115e-05, "loss": 0.1212, "step": 9025 }, { "epoch": 0.42, "learning_rate": 1.5776031434184678e-05, "loss": 0.1841, "step": 9030 }, { "epoch": 0.42, "learning_rate": 1.57736925811582e-05, "loss": 0.1336, "step": 9035 }, { "epoch": 0.42, "learning_rate": 1.5771353728131725e-05, "loss": 0.0551, "step": 9040 }, { "epoch": 0.42, "learning_rate": 1.576901487510525e-05, "loss": 0.1533, "step": 9045 }, { "epoch": 0.42, "learning_rate": 1.576667602207877e-05, "loss": 0.1459, "step": 9050 }, { "epoch": 0.42, "learning_rate": 1.5764337169052298e-05, "loss": 0.1649, "step": 9055 }, { "epoch": 0.42, "learning_rate": 1.576199831602582e-05, "loss": 0.1112, "step": 9060 }, { "epoch": 0.42, "learning_rate": 1.5759659462999344e-05, "loss": 0.0754, "step": 9065 }, { "epoch": 0.42, "learning_rate": 1.575732060997287e-05, "loss": 0.0468, "step": 9070 }, { "epoch": 0.42, "learning_rate": 1.5754981756946394e-05, "loss": 0.1003, "step": 9075 }, { "epoch": 0.42, "learning_rate": 1.5752642903919918e-05, "loss": 0.1319, "step": 9080 }, { "epoch": 0.42, "learning_rate": 1.5750304050893444e-05, "loss": 0.1623, "step": 9085 }, { "epoch": 0.43, "learning_rate": 1.5747965197866968e-05, "loss": 0.1235, "step": 9090 }, { "epoch": 0.43, "learning_rate": 1.574562634484049e-05, "loss": 0.1048, "step": 9095 }, { "epoch": 0.43, "learning_rate": 1.5743287491814017e-05, "loss": 0.1145, "step": 9100 }, { "epoch": 0.43, "learning_rate": 1.574094863878754e-05, "loss": 0.1087, "step": 9105 }, { "epoch": 0.43, "learning_rate": 1.5738609785761064e-05, "loss": 0.1087, "step": 9110 }, { "epoch": 0.43, "learning_rate": 1.5736270932734587e-05, "loss": 0.1598, "step": 9115 }, { "epoch": 0.43, "learning_rate": 1.573393207970811e-05, "loss": 0.1308, "step": 9120 }, { "epoch": 0.43, "learning_rate": 1.5731593226681637e-05, "loss": 0.0941, "step": 9125 }, { "epoch": 0.43, "learning_rate": 1.572925437365516e-05, "loss": 0.1108, "step": 9130 }, { "epoch": 0.43, "learning_rate": 1.5726915520628684e-05, "loss": 0.1437, "step": 9135 }, { "epoch": 0.43, "learning_rate": 1.572457666760221e-05, "loss": 0.1188, "step": 9140 }, { "epoch": 0.43, "learning_rate": 1.5722237814575734e-05, "loss": 0.0604, "step": 9145 }, { "epoch": 0.43, "learning_rate": 1.5719898961549257e-05, "loss": 0.0968, "step": 9150 }, { "epoch": 0.43, "learning_rate": 1.5717560108522783e-05, "loss": 0.1395, "step": 9155 }, { "epoch": 0.43, "learning_rate": 1.5715221255496307e-05, "loss": 0.097, "step": 9160 }, { "epoch": 0.43, "learning_rate": 1.571288240246983e-05, "loss": 0.1234, "step": 9165 }, { "epoch": 0.43, "learning_rate": 1.5710543549443353e-05, "loss": 0.0881, "step": 9170 }, { "epoch": 0.43, "learning_rate": 1.5708204696416876e-05, "loss": 0.1082, "step": 9175 }, { "epoch": 0.43, "learning_rate": 1.5705865843390403e-05, "loss": 0.0824, "step": 9180 }, { "epoch": 0.43, "learning_rate": 1.5703526990363926e-05, "loss": 0.1381, "step": 9185 }, { "epoch": 0.43, "learning_rate": 1.570118813733745e-05, "loss": 0.0965, "step": 9190 }, { "epoch": 0.43, "learning_rate": 1.5698849284310976e-05, "loss": 0.0327, "step": 9195 }, { "epoch": 0.43, "learning_rate": 1.56965104312845e-05, "loss": 0.0946, "step": 9200 }, { "epoch": 0.43, "learning_rate": 1.5694171578258023e-05, "loss": 0.1084, "step": 9205 }, { "epoch": 0.43, "learning_rate": 1.569183272523155e-05, "loss": 0.137, "step": 9210 }, { "epoch": 0.43, "learning_rate": 1.5689493872205073e-05, "loss": 0.0703, "step": 9215 }, { "epoch": 0.43, "learning_rate": 1.5687155019178596e-05, "loss": 0.0876, "step": 9220 }, { "epoch": 0.43, "learning_rate": 1.568481616615212e-05, "loss": 0.1168, "step": 9225 }, { "epoch": 0.43, "learning_rate": 1.5682477313125642e-05, "loss": 0.1047, "step": 9230 }, { "epoch": 0.43, "learning_rate": 1.568013846009917e-05, "loss": 0.0553, "step": 9235 }, { "epoch": 0.43, "learning_rate": 1.5677799607072692e-05, "loss": 0.1296, "step": 9240 }, { "epoch": 0.43, "learning_rate": 1.5675460754046216e-05, "loss": 0.11, "step": 9245 }, { "epoch": 0.43, "learning_rate": 1.5673121901019742e-05, "loss": 0.1078, "step": 9250 }, { "epoch": 0.43, "learning_rate": 1.5670783047993266e-05, "loss": 0.0949, "step": 9255 }, { "epoch": 0.43, "learning_rate": 1.566844419496679e-05, "loss": 0.1241, "step": 9260 }, { "epoch": 0.43, "learning_rate": 1.5666105341940315e-05, "loss": 0.0706, "step": 9265 }, { "epoch": 0.43, "learning_rate": 1.566376648891384e-05, "loss": 0.0727, "step": 9270 }, { "epoch": 0.43, "learning_rate": 1.5661427635887362e-05, "loss": 0.138, "step": 9275 }, { "epoch": 0.43, "learning_rate": 1.565908878286089e-05, "loss": 0.0608, "step": 9280 }, { "epoch": 0.43, "learning_rate": 1.565674992983441e-05, "loss": 0.1041, "step": 9285 }, { "epoch": 0.43, "learning_rate": 1.5654411076807935e-05, "loss": 0.0956, "step": 9290 }, { "epoch": 0.43, "learning_rate": 1.565207222378146e-05, "loss": 0.1063, "step": 9295 }, { "epoch": 0.44, "learning_rate": 1.564973337075498e-05, "loss": 0.0606, "step": 9300 }, { "epoch": 0.44, "learning_rate": 1.5647394517728508e-05, "loss": 0.1061, "step": 9305 }, { "epoch": 0.44, "learning_rate": 1.564505566470203e-05, "loss": 0.1413, "step": 9310 }, { "epoch": 0.44, "learning_rate": 1.5642716811675555e-05, "loss": 0.0936, "step": 9315 }, { "epoch": 0.44, "learning_rate": 1.564037795864908e-05, "loss": 0.1047, "step": 9320 }, { "epoch": 0.44, "learning_rate": 1.5638039105622605e-05, "loss": 0.0876, "step": 9325 }, { "epoch": 0.44, "learning_rate": 1.5635700252596128e-05, "loss": 0.0817, "step": 9330 }, { "epoch": 0.44, "learning_rate": 1.5633361399569655e-05, "loss": 0.0812, "step": 9335 }, { "epoch": 0.44, "learning_rate": 1.5631022546543178e-05, "loss": 0.1132, "step": 9340 }, { "epoch": 0.44, "learning_rate": 1.56286836935167e-05, "loss": 0.119, "step": 9345 }, { "epoch": 0.44, "learning_rate": 1.5626344840490224e-05, "loss": 0.1461, "step": 9350 }, { "epoch": 0.44, "learning_rate": 1.5624005987463748e-05, "loss": 0.087, "step": 9355 }, { "epoch": 0.44, "learning_rate": 1.5621667134437274e-05, "loss": 0.0923, "step": 9360 }, { "epoch": 0.44, "learning_rate": 1.5619328281410798e-05, "loss": 0.0887, "step": 9365 }, { "epoch": 0.44, "learning_rate": 1.561698942838432e-05, "loss": 0.053, "step": 9370 }, { "epoch": 0.44, "learning_rate": 1.5614650575357847e-05, "loss": 0.0749, "step": 9375 }, { "epoch": 0.44, "learning_rate": 1.561231172233137e-05, "loss": 0.1482, "step": 9380 }, { "epoch": 0.44, "learning_rate": 1.5609972869304894e-05, "loss": 0.1071, "step": 9385 }, { "epoch": 0.44, "learning_rate": 1.560763401627842e-05, "loss": 0.0859, "step": 9390 }, { "epoch": 0.44, "learning_rate": 1.5605295163251944e-05, "loss": 0.0712, "step": 9395 }, { "epoch": 0.44, "learning_rate": 1.5602956310225467e-05, "loss": 0.0593, "step": 9400 }, { "epoch": 0.44, "learning_rate": 1.560061745719899e-05, "loss": 0.1071, "step": 9405 }, { "epoch": 0.44, "learning_rate": 1.5598278604172514e-05, "loss": 0.1191, "step": 9410 }, { "epoch": 0.44, "learning_rate": 1.559593975114604e-05, "loss": 0.0886, "step": 9415 }, { "epoch": 0.44, "learning_rate": 1.5593600898119564e-05, "loss": 0.0874, "step": 9420 }, { "epoch": 0.44, "learning_rate": 1.5591262045093087e-05, "loss": 0.1072, "step": 9425 }, { "epoch": 0.44, "learning_rate": 1.5588923192066613e-05, "loss": 0.1047, "step": 9430 }, { "epoch": 0.44, "learning_rate": 1.5586584339040137e-05, "loss": 0.0712, "step": 9435 }, { "epoch": 0.44, "learning_rate": 1.558424548601366e-05, "loss": 0.1663, "step": 9440 }, { "epoch": 0.44, "learning_rate": 1.5581906632987183e-05, "loss": 0.084, "step": 9445 }, { "epoch": 0.44, "learning_rate": 1.557956777996071e-05, "loss": 0.0595, "step": 9450 }, { "epoch": 0.44, "learning_rate": 1.5577228926934233e-05, "loss": 0.1238, "step": 9455 }, { "epoch": 0.44, "learning_rate": 1.5574890073907756e-05, "loss": 0.1318, "step": 9460 }, { "epoch": 0.44, "learning_rate": 1.557255122088128e-05, "loss": 0.107, "step": 9465 }, { "epoch": 0.44, "learning_rate": 1.5570212367854803e-05, "loss": 0.1358, "step": 9470 }, { "epoch": 0.44, "learning_rate": 1.556787351482833e-05, "loss": 0.1278, "step": 9475 }, { "epoch": 0.44, "learning_rate": 1.5565534661801853e-05, "loss": 0.0806, "step": 9480 }, { "epoch": 0.44, "learning_rate": 1.5563195808775376e-05, "loss": 0.1278, "step": 9485 }, { "epoch": 0.44, "learning_rate": 1.5560856955748903e-05, "loss": 0.1578, "step": 9490 }, { "epoch": 0.44, "learning_rate": 1.5558518102722426e-05, "loss": 0.146, "step": 9495 }, { "epoch": 0.44, "learning_rate": 1.555617924969595e-05, "loss": 0.1313, "step": 9500 }, { "epoch": 0.44, "learning_rate": 1.5553840396669476e-05, "loss": 0.0902, "step": 9505 }, { "epoch": 0.44, "learning_rate": 1.5551501543643e-05, "loss": 0.0564, "step": 9510 }, { "epoch": 0.45, "learning_rate": 1.5549162690616522e-05, "loss": 0.1041, "step": 9515 }, { "epoch": 0.45, "learning_rate": 1.554682383759005e-05, "loss": 0.1081, "step": 9520 }, { "epoch": 0.45, "learning_rate": 1.554448498456357e-05, "loss": 0.1226, "step": 9525 }, { "epoch": 0.45, "learning_rate": 1.5542146131537096e-05, "loss": 0.0911, "step": 9530 }, { "epoch": 0.45, "learning_rate": 1.553980727851062e-05, "loss": 0.0838, "step": 9535 }, { "epoch": 0.45, "learning_rate": 1.5537468425484142e-05, "loss": 0.0927, "step": 9540 }, { "epoch": 0.45, "learning_rate": 1.553512957245767e-05, "loss": 0.1573, "step": 9545 }, { "epoch": 0.45, "learning_rate": 1.5532790719431192e-05, "loss": 0.1143, "step": 9550 }, { "epoch": 0.45, "learning_rate": 1.5530451866404715e-05, "loss": 0.0773, "step": 9555 }, { "epoch": 0.45, "learning_rate": 1.5528113013378242e-05, "loss": 0.1376, "step": 9560 }, { "epoch": 0.45, "learning_rate": 1.5525774160351765e-05, "loss": 0.1292, "step": 9565 }, { "epoch": 0.45, "learning_rate": 1.552343530732529e-05, "loss": 0.0615, "step": 9570 }, { "epoch": 0.45, "learning_rate": 1.5521096454298815e-05, "loss": 0.1316, "step": 9575 }, { "epoch": 0.45, "learning_rate": 1.5518757601272335e-05, "loss": 0.1217, "step": 9580 }, { "epoch": 0.45, "learning_rate": 1.551641874824586e-05, "loss": 0.1288, "step": 9585 }, { "epoch": 0.45, "learning_rate": 1.5514079895219385e-05, "loss": 0.1328, "step": 9590 }, { "epoch": 0.45, "learning_rate": 1.5511741042192908e-05, "loss": 0.1006, "step": 9595 }, { "epoch": 0.45, "learning_rate": 1.5509402189166435e-05, "loss": 0.0632, "step": 9600 }, { "epoch": 0.45, "learning_rate": 1.5507063336139958e-05, "loss": 0.168, "step": 9605 }, { "epoch": 0.45, "learning_rate": 1.550472448311348e-05, "loss": 0.2259, "step": 9610 }, { "epoch": 0.45, "learning_rate": 1.5502385630087008e-05, "loss": 0.1803, "step": 9615 }, { "epoch": 0.45, "learning_rate": 1.550004677706053e-05, "loss": 0.0725, "step": 9620 }, { "epoch": 0.45, "learning_rate": 1.5497707924034054e-05, "loss": 0.1154, "step": 9625 }, { "epoch": 0.45, "learning_rate": 1.549536907100758e-05, "loss": 0.0994, "step": 9630 }, { "epoch": 0.45, "learning_rate": 1.5493030217981104e-05, "loss": 0.0654, "step": 9635 }, { "epoch": 0.45, "learning_rate": 1.5490691364954628e-05, "loss": 0.0815, "step": 9640 }, { "epoch": 0.45, "learning_rate": 1.548835251192815e-05, "loss": 0.1461, "step": 9645 }, { "epoch": 0.45, "learning_rate": 1.5486013658901674e-05, "loss": 0.1448, "step": 9650 }, { "epoch": 0.45, "learning_rate": 1.54836748058752e-05, "loss": 0.1643, "step": 9655 }, { "epoch": 0.45, "learning_rate": 1.5481335952848724e-05, "loss": 0.0845, "step": 9660 }, { "epoch": 0.45, "learning_rate": 1.5478997099822247e-05, "loss": 0.1019, "step": 9665 }, { "epoch": 0.45, "learning_rate": 1.5476658246795774e-05, "loss": 0.0821, "step": 9670 }, { "epoch": 0.45, "learning_rate": 1.5474319393769297e-05, "loss": 0.0909, "step": 9675 }, { "epoch": 0.45, "learning_rate": 1.547198054074282e-05, "loss": 0.1155, "step": 9680 }, { "epoch": 0.45, "learning_rate": 1.5469641687716347e-05, "loss": 0.1202, "step": 9685 }, { "epoch": 0.45, "learning_rate": 1.546730283468987e-05, "loss": 0.0678, "step": 9690 }, { "epoch": 0.45, "learning_rate": 1.5464963981663394e-05, "loss": 0.1086, "step": 9695 }, { "epoch": 0.45, "learning_rate": 1.5462625128636917e-05, "loss": 0.0837, "step": 9700 }, { "epoch": 0.45, "learning_rate": 1.546028627561044e-05, "loss": 0.0829, "step": 9705 }, { "epoch": 0.45, "learning_rate": 1.5457947422583967e-05, "loss": 0.0961, "step": 9710 }, { "epoch": 0.45, "learning_rate": 1.545560856955749e-05, "loss": 0.1039, "step": 9715 }, { "epoch": 0.45, "learning_rate": 1.5453269716531013e-05, "loss": 0.0804, "step": 9720 }, { "epoch": 0.45, "learning_rate": 1.545093086350454e-05, "loss": 0.0498, "step": 9725 }, { "epoch": 0.46, "learning_rate": 1.5448592010478063e-05, "loss": 0.136, "step": 9730 }, { "epoch": 0.46, "learning_rate": 1.5446253157451586e-05, "loss": 0.1193, "step": 9735 }, { "epoch": 0.46, "learning_rate": 1.5443914304425113e-05, "loss": 0.0495, "step": 9740 }, { "epoch": 0.46, "learning_rate": 1.5441575451398636e-05, "loss": 0.1363, "step": 9745 }, { "epoch": 0.46, "learning_rate": 1.543923659837216e-05, "loss": 0.1475, "step": 9750 }, { "epoch": 0.46, "learning_rate": 1.5436897745345686e-05, "loss": 0.1202, "step": 9755 }, { "epoch": 0.46, "learning_rate": 1.5434558892319206e-05, "loss": 0.0801, "step": 9760 }, { "epoch": 0.46, "learning_rate": 1.5432220039292733e-05, "loss": 0.1098, "step": 9765 }, { "epoch": 0.46, "learning_rate": 1.5429881186266256e-05, "loss": 0.0786, "step": 9770 }, { "epoch": 0.46, "learning_rate": 1.542754233323978e-05, "loss": 0.1172, "step": 9775 }, { "epoch": 0.46, "learning_rate": 1.5425203480213306e-05, "loss": 0.0755, "step": 9780 }, { "epoch": 0.46, "learning_rate": 1.542286462718683e-05, "loss": 0.1005, "step": 9785 }, { "epoch": 0.46, "learning_rate": 1.5420525774160352e-05, "loss": 0.0711, "step": 9790 }, { "epoch": 0.46, "learning_rate": 1.541818692113388e-05, "loss": 0.0232, "step": 9795 }, { "epoch": 0.46, "learning_rate": 1.5415848068107402e-05, "loss": 0.0538, "step": 9800 }, { "epoch": 0.46, "learning_rate": 1.5413509215080925e-05, "loss": 0.0691, "step": 9805 }, { "epoch": 0.46, "learning_rate": 1.5411170362054452e-05, "loss": 0.1248, "step": 9810 }, { "epoch": 0.46, "learning_rate": 1.5408831509027975e-05, "loss": 0.1091, "step": 9815 }, { "epoch": 0.46, "learning_rate": 1.54064926560015e-05, "loss": 0.1382, "step": 9820 }, { "epoch": 0.46, "learning_rate": 1.5404153802975022e-05, "loss": 0.129, "step": 9825 }, { "epoch": 0.46, "learning_rate": 1.5401814949948545e-05, "loss": 0.0904, "step": 9830 }, { "epoch": 0.46, "learning_rate": 1.5399476096922072e-05, "loss": 0.0664, "step": 9835 }, { "epoch": 0.46, "learning_rate": 1.5397137243895595e-05, "loss": 0.1025, "step": 9840 }, { "epoch": 0.46, "learning_rate": 1.539479839086912e-05, "loss": 0.0931, "step": 9845 }, { "epoch": 0.46, "learning_rate": 1.5392459537842645e-05, "loss": 0.1852, "step": 9850 }, { "epoch": 0.46, "learning_rate": 1.5390120684816168e-05, "loss": 0.1145, "step": 9855 }, { "epoch": 0.46, "learning_rate": 1.538778183178969e-05, "loss": 0.1037, "step": 9860 }, { "epoch": 0.46, "learning_rate": 1.5385442978763218e-05, "loss": 0.1115, "step": 9865 }, { "epoch": 0.46, "learning_rate": 1.538310412573674e-05, "loss": 0.077, "step": 9870 }, { "epoch": 0.46, "learning_rate": 1.5380765272710265e-05, "loss": 0.0704, "step": 9875 }, { "epoch": 0.46, "learning_rate": 1.5378426419683788e-05, "loss": 0.1298, "step": 9880 }, { "epoch": 0.46, "learning_rate": 1.537608756665731e-05, "loss": 0.0848, "step": 9885 }, { "epoch": 0.46, "learning_rate": 1.5373748713630834e-05, "loss": 0.1644, "step": 9890 }, { "epoch": 0.46, "learning_rate": 1.537140986060436e-05, "loss": 0.1392, "step": 9895 }, { "epoch": 0.46, "learning_rate": 1.5369071007577884e-05, "loss": 0.1098, "step": 9900 }, { "epoch": 0.46, "learning_rate": 1.5366732154551408e-05, "loss": 0.1371, "step": 9905 }, { "epoch": 0.46, "learning_rate": 1.5364393301524934e-05, "loss": 0.0748, "step": 9910 }, { "epoch": 0.46, "learning_rate": 1.5362054448498457e-05, "loss": 0.1203, "step": 9915 }, { "epoch": 0.46, "learning_rate": 1.535971559547198e-05, "loss": 0.1818, "step": 9920 }, { "epoch": 0.46, "learning_rate": 1.5357376742445507e-05, "loss": 0.1057, "step": 9925 }, { "epoch": 0.46, "learning_rate": 1.535503788941903e-05, "loss": 0.1589, "step": 9930 }, { "epoch": 0.46, "learning_rate": 1.5352699036392554e-05, "loss": 0.1239, "step": 9935 }, { "epoch": 0.46, "learning_rate": 1.5350360183366077e-05, "loss": 0.0612, "step": 9940 }, { "epoch": 0.47, "learning_rate": 1.53480213303396e-05, "loss": 0.0954, "step": 9945 }, { "epoch": 0.47, "learning_rate": 1.5345682477313127e-05, "loss": 0.0606, "step": 9950 }, { "epoch": 0.47, "learning_rate": 1.534334362428665e-05, "loss": 0.0589, "step": 9955 }, { "epoch": 0.47, "learning_rate": 1.5341004771260174e-05, "loss": 0.0698, "step": 9960 }, { "epoch": 0.47, "learning_rate": 1.53386659182337e-05, "loss": 0.0445, "step": 9965 }, { "epoch": 0.47, "learning_rate": 1.5336327065207223e-05, "loss": 0.0852, "step": 9970 }, { "epoch": 0.47, "learning_rate": 1.5333988212180747e-05, "loss": 0.1319, "step": 9975 }, { "epoch": 0.47, "learning_rate": 1.5331649359154273e-05, "loss": 0.1578, "step": 9980 }, { "epoch": 0.47, "learning_rate": 1.5329310506127797e-05, "loss": 0.1136, "step": 9985 }, { "epoch": 0.47, "learning_rate": 1.532697165310132e-05, "loss": 0.0786, "step": 9990 }, { "epoch": 0.47, "learning_rate": 1.5324632800074843e-05, "loss": 0.0958, "step": 9995 }, { "epoch": 0.47, "learning_rate": 1.5322293947048366e-05, "loss": 0.084, "step": 10000 }, { "epoch": 0.47, "learning_rate": 1.5319955094021893e-05, "loss": 0.1414, "step": 10005 }, { "epoch": 0.47, "learning_rate": 1.5317616240995416e-05, "loss": 0.0962, "step": 10010 }, { "epoch": 0.47, "learning_rate": 1.531527738796894e-05, "loss": 0.1253, "step": 10015 }, { "epoch": 0.47, "learning_rate": 1.5312938534942466e-05, "loss": 0.0965, "step": 10020 }, { "epoch": 0.47, "learning_rate": 1.531059968191599e-05, "loss": 0.1636, "step": 10025 }, { "epoch": 0.47, "learning_rate": 1.5308260828889513e-05, "loss": 0.1289, "step": 10030 }, { "epoch": 0.47, "learning_rate": 1.530592197586304e-05, "loss": 0.1565, "step": 10035 }, { "epoch": 0.47, "learning_rate": 1.5303583122836563e-05, "loss": 0.071, "step": 10040 }, { "epoch": 0.47, "learning_rate": 1.5301244269810086e-05, "loss": 0.0626, "step": 10045 }, { "epoch": 0.47, "learning_rate": 1.5298905416783613e-05, "loss": 0.1034, "step": 10050 }, { "epoch": 0.47, "learning_rate": 1.5296566563757132e-05, "loss": 0.1091, "step": 10055 }, { "epoch": 0.47, "learning_rate": 1.529422771073066e-05, "loss": 0.0756, "step": 10060 }, { "epoch": 0.47, "learning_rate": 1.5291888857704182e-05, "loss": 0.0999, "step": 10065 }, { "epoch": 0.47, "learning_rate": 1.5289550004677706e-05, "loss": 0.1319, "step": 10070 }, { "epoch": 0.47, "learning_rate": 1.5287211151651232e-05, "loss": 0.0999, "step": 10075 }, { "epoch": 0.47, "learning_rate": 1.5284872298624755e-05, "loss": 0.087, "step": 10080 }, { "epoch": 0.47, "learning_rate": 1.528253344559828e-05, "loss": 0.0732, "step": 10085 }, { "epoch": 0.47, "learning_rate": 1.5280194592571805e-05, "loss": 0.1176, "step": 10090 }, { "epoch": 0.47, "learning_rate": 1.527785573954533e-05, "loss": 0.0525, "step": 10095 }, { "epoch": 0.47, "learning_rate": 1.5275516886518852e-05, "loss": 0.106, "step": 10100 }, { "epoch": 0.47, "learning_rate": 1.527317803349238e-05, "loss": 0.0766, "step": 10105 }, { "epoch": 0.47, "learning_rate": 1.5270839180465902e-05, "loss": 0.0719, "step": 10110 }, { "epoch": 0.47, "learning_rate": 1.5268500327439425e-05, "loss": 0.0892, "step": 10115 }, { "epoch": 0.47, "learning_rate": 1.526616147441295e-05, "loss": 0.0817, "step": 10120 }, { "epoch": 0.47, "learning_rate": 1.526382262138647e-05, "loss": 0.1294, "step": 10125 }, { "epoch": 0.47, "learning_rate": 1.5261483768359998e-05, "loss": 0.1542, "step": 10130 }, { "epoch": 0.47, "learning_rate": 1.525914491533352e-05, "loss": 0.1247, "step": 10135 }, { "epoch": 0.47, "learning_rate": 1.5256806062307046e-05, "loss": 0.0947, "step": 10140 }, { "epoch": 0.47, "learning_rate": 1.525446720928057e-05, "loss": 0.1036, "step": 10145 }, { "epoch": 0.47, "learning_rate": 1.5252128356254095e-05, "loss": 0.0799, "step": 10150 }, { "epoch": 0.48, "learning_rate": 1.524978950322762e-05, "loss": 0.1518, "step": 10155 }, { "epoch": 0.48, "learning_rate": 1.5247450650201143e-05, "loss": 0.2048, "step": 10160 }, { "epoch": 0.48, "learning_rate": 1.5245111797174668e-05, "loss": 0.0909, "step": 10165 }, { "epoch": 0.48, "learning_rate": 1.5242772944148193e-05, "loss": 0.1434, "step": 10170 }, { "epoch": 0.48, "learning_rate": 1.5240434091121714e-05, "loss": 0.0788, "step": 10175 }, { "epoch": 0.48, "learning_rate": 1.523809523809524e-05, "loss": 0.0627, "step": 10180 }, { "epoch": 0.48, "learning_rate": 1.5235756385068763e-05, "loss": 0.081, "step": 10185 }, { "epoch": 0.48, "learning_rate": 1.5233417532042287e-05, "loss": 0.1024, "step": 10190 }, { "epoch": 0.48, "learning_rate": 1.5231078679015812e-05, "loss": 0.0921, "step": 10195 }, { "epoch": 0.48, "learning_rate": 1.5228739825989336e-05, "loss": 0.082, "step": 10200 }, { "epoch": 0.48, "learning_rate": 1.522640097296286e-05, "loss": 0.0906, "step": 10205 }, { "epoch": 0.48, "learning_rate": 1.5224062119936384e-05, "loss": 0.1533, "step": 10210 }, { "epoch": 0.48, "learning_rate": 1.5221723266909909e-05, "loss": 0.1345, "step": 10215 }, { "epoch": 0.48, "learning_rate": 1.5219384413883434e-05, "loss": 0.0823, "step": 10220 }, { "epoch": 0.48, "learning_rate": 1.5217045560856957e-05, "loss": 0.1362, "step": 10225 }, { "epoch": 0.48, "learning_rate": 1.5214706707830482e-05, "loss": 0.1098, "step": 10230 }, { "epoch": 0.48, "learning_rate": 1.5212367854804004e-05, "loss": 0.0636, "step": 10235 }, { "epoch": 0.48, "learning_rate": 1.5210029001777529e-05, "loss": 0.0504, "step": 10240 }, { "epoch": 0.48, "learning_rate": 1.5207690148751053e-05, "loss": 0.1478, "step": 10245 }, { "epoch": 0.48, "learning_rate": 1.5205351295724577e-05, "loss": 0.0871, "step": 10250 }, { "epoch": 0.48, "learning_rate": 1.5203012442698102e-05, "loss": 0.1124, "step": 10255 }, { "epoch": 0.48, "learning_rate": 1.5200673589671627e-05, "loss": 0.1273, "step": 10260 }, { "epoch": 0.48, "learning_rate": 1.519833473664515e-05, "loss": 0.0571, "step": 10265 }, { "epoch": 0.48, "learning_rate": 1.5195995883618675e-05, "loss": 0.0468, "step": 10270 }, { "epoch": 0.48, "learning_rate": 1.51936570305922e-05, "loss": 0.1113, "step": 10275 }, { "epoch": 0.48, "learning_rate": 1.5191318177565723e-05, "loss": 0.2059, "step": 10280 }, { "epoch": 0.48, "learning_rate": 1.5188979324539248e-05, "loss": 0.0702, "step": 10285 }, { "epoch": 0.48, "learning_rate": 1.5186640471512773e-05, "loss": 0.1162, "step": 10290 }, { "epoch": 0.48, "learning_rate": 1.5184301618486295e-05, "loss": 0.1341, "step": 10295 }, { "epoch": 0.48, "learning_rate": 1.518196276545982e-05, "loss": 0.0714, "step": 10300 }, { "epoch": 0.48, "learning_rate": 1.5179623912433343e-05, "loss": 0.1074, "step": 10305 }, { "epoch": 0.48, "learning_rate": 1.5177285059406868e-05, "loss": 0.1128, "step": 10310 }, { "epoch": 0.48, "learning_rate": 1.5174946206380393e-05, "loss": 0.077, "step": 10315 }, { "epoch": 0.48, "learning_rate": 1.5172607353353916e-05, "loss": 0.1127, "step": 10320 }, { "epoch": 0.48, "learning_rate": 1.517026850032744e-05, "loss": 0.0935, "step": 10325 }, { "epoch": 0.48, "learning_rate": 1.5167929647300966e-05, "loss": 0.0761, "step": 10330 }, { "epoch": 0.48, "learning_rate": 1.5165590794274489e-05, "loss": 0.0849, "step": 10335 }, { "epoch": 0.48, "learning_rate": 1.5163251941248014e-05, "loss": 0.0946, "step": 10340 }, { "epoch": 0.48, "learning_rate": 1.5160913088221539e-05, "loss": 0.1212, "step": 10345 }, { "epoch": 0.48, "learning_rate": 1.5158574235195062e-05, "loss": 0.1173, "step": 10350 }, { "epoch": 0.48, "learning_rate": 1.5156235382168585e-05, "loss": 0.0674, "step": 10355 }, { "epoch": 0.48, "learning_rate": 1.5153896529142109e-05, "loss": 0.0864, "step": 10360 }, { "epoch": 0.48, "learning_rate": 1.5151557676115634e-05, "loss": 0.1502, "step": 10365 }, { "epoch": 0.49, "learning_rate": 1.5149218823089159e-05, "loss": 0.0668, "step": 10370 }, { "epoch": 0.49, "learning_rate": 1.5146879970062682e-05, "loss": 0.1041, "step": 10375 }, { "epoch": 0.49, "learning_rate": 1.5144541117036207e-05, "loss": 0.1571, "step": 10380 }, { "epoch": 0.49, "learning_rate": 1.5142202264009732e-05, "loss": 0.0849, "step": 10385 }, { "epoch": 0.49, "learning_rate": 1.5139863410983255e-05, "loss": 0.0536, "step": 10390 }, { "epoch": 0.49, "learning_rate": 1.513752455795678e-05, "loss": 0.1016, "step": 10395 }, { "epoch": 0.49, "learning_rate": 1.5135185704930305e-05, "loss": 0.1556, "step": 10400 }, { "epoch": 0.49, "learning_rate": 1.5132846851903828e-05, "loss": 0.1046, "step": 10405 }, { "epoch": 0.49, "learning_rate": 1.5130507998877353e-05, "loss": 0.0857, "step": 10410 }, { "epoch": 0.49, "learning_rate": 1.5128169145850875e-05, "loss": 0.092, "step": 10415 }, { "epoch": 0.49, "learning_rate": 1.51258302928244e-05, "loss": 0.1437, "step": 10420 }, { "epoch": 0.49, "learning_rate": 1.5123491439797925e-05, "loss": 0.1206, "step": 10425 }, { "epoch": 0.49, "learning_rate": 1.5121152586771448e-05, "loss": 0.0926, "step": 10430 }, { "epoch": 0.49, "learning_rate": 1.5118813733744973e-05, "loss": 0.1042, "step": 10435 }, { "epoch": 0.49, "learning_rate": 1.5116474880718496e-05, "loss": 0.0743, "step": 10440 }, { "epoch": 0.49, "learning_rate": 1.5114136027692021e-05, "loss": 0.1055, "step": 10445 }, { "epoch": 0.49, "learning_rate": 1.5111797174665546e-05, "loss": 0.0955, "step": 10450 }, { "epoch": 0.49, "learning_rate": 1.510945832163907e-05, "loss": 0.1172, "step": 10455 }, { "epoch": 0.49, "learning_rate": 1.5107119468612594e-05, "loss": 0.1117, "step": 10460 }, { "epoch": 0.49, "learning_rate": 1.5104780615586119e-05, "loss": 0.1383, "step": 10465 }, { "epoch": 0.49, "learning_rate": 1.510244176255964e-05, "loss": 0.1049, "step": 10470 }, { "epoch": 0.49, "learning_rate": 1.5100102909533166e-05, "loss": 0.0803, "step": 10475 }, { "epoch": 0.49, "learning_rate": 1.5097764056506689e-05, "loss": 0.0872, "step": 10480 }, { "epoch": 0.49, "learning_rate": 1.5095425203480214e-05, "loss": 0.114, "step": 10485 }, { "epoch": 0.49, "learning_rate": 1.5093086350453739e-05, "loss": 0.0798, "step": 10490 }, { "epoch": 0.49, "learning_rate": 1.5090747497427262e-05, "loss": 0.1309, "step": 10495 }, { "epoch": 0.49, "learning_rate": 1.5088408644400787e-05, "loss": 0.1353, "step": 10500 }, { "epoch": 0.49, "learning_rate": 1.5086069791374312e-05, "loss": 0.1029, "step": 10505 }, { "epoch": 0.49, "learning_rate": 1.5083730938347835e-05, "loss": 0.0952, "step": 10510 }, { "epoch": 0.49, "learning_rate": 1.508139208532136e-05, "loss": 0.1327, "step": 10515 }, { "epoch": 0.49, "learning_rate": 1.5079053232294885e-05, "loss": 0.0528, "step": 10520 }, { "epoch": 0.49, "learning_rate": 1.5076714379268408e-05, "loss": 0.1496, "step": 10525 }, { "epoch": 0.49, "learning_rate": 1.5074375526241932e-05, "loss": 0.0663, "step": 10530 }, { "epoch": 0.49, "learning_rate": 1.5072036673215455e-05, "loss": 0.1341, "step": 10535 }, { "epoch": 0.49, "learning_rate": 1.506969782018898e-05, "loss": 0.1041, "step": 10540 }, { "epoch": 0.49, "learning_rate": 1.5067358967162505e-05, "loss": 0.0908, "step": 10545 }, { "epoch": 0.49, "learning_rate": 1.5065020114136028e-05, "loss": 0.096, "step": 10550 }, { "epoch": 0.49, "learning_rate": 1.5062681261109553e-05, "loss": 0.0887, "step": 10555 }, { "epoch": 0.49, "learning_rate": 1.5060342408083078e-05, "loss": 0.0747, "step": 10560 }, { "epoch": 0.49, "learning_rate": 1.5058003555056601e-05, "loss": 0.1425, "step": 10565 }, { "epoch": 0.49, "learning_rate": 1.5055664702030126e-05, "loss": 0.1207, "step": 10570 }, { "epoch": 0.49, "learning_rate": 1.5053325849003651e-05, "loss": 0.0969, "step": 10575 }, { "epoch": 0.49, "learning_rate": 1.5050986995977174e-05, "loss": 0.0857, "step": 10580 }, { "epoch": 0.5, "learning_rate": 1.50486481429507e-05, "loss": 0.1518, "step": 10585 }, { "epoch": 0.5, "learning_rate": 1.5046309289924221e-05, "loss": 0.0683, "step": 10590 }, { "epoch": 0.5, "learning_rate": 1.5043970436897746e-05, "loss": 0.0976, "step": 10595 }, { "epoch": 0.5, "learning_rate": 1.504163158387127e-05, "loss": 0.1279, "step": 10600 }, { "epoch": 0.5, "learning_rate": 1.5039292730844794e-05, "loss": 0.0778, "step": 10605 }, { "epoch": 0.5, "learning_rate": 1.5036953877818319e-05, "loss": 0.0923, "step": 10610 }, { "epoch": 0.5, "learning_rate": 1.5034615024791844e-05, "loss": 0.1061, "step": 10615 }, { "epoch": 0.5, "learning_rate": 1.5032276171765367e-05, "loss": 0.1261, "step": 10620 }, { "epoch": 0.5, "learning_rate": 1.5029937318738892e-05, "loss": 0.1069, "step": 10625 }, { "epoch": 0.5, "learning_rate": 1.5027598465712417e-05, "loss": 0.1434, "step": 10630 }, { "epoch": 0.5, "learning_rate": 1.502525961268594e-05, "loss": 0.0896, "step": 10635 }, { "epoch": 0.5, "learning_rate": 1.5022920759659465e-05, "loss": 0.1174, "step": 10640 }, { "epoch": 0.5, "learning_rate": 1.5020581906632989e-05, "loss": 0.1514, "step": 10645 }, { "epoch": 0.5, "learning_rate": 1.5018243053606512e-05, "loss": 0.0736, "step": 10650 }, { "epoch": 0.5, "learning_rate": 1.5015904200580035e-05, "loss": 0.1307, "step": 10655 }, { "epoch": 0.5, "learning_rate": 1.501356534755356e-05, "loss": 0.1424, "step": 10660 }, { "epoch": 0.5, "learning_rate": 1.5011226494527085e-05, "loss": 0.0673, "step": 10665 }, { "epoch": 0.5, "learning_rate": 1.5008887641500608e-05, "loss": 0.0907, "step": 10670 }, { "epoch": 0.5, "learning_rate": 1.5006548788474133e-05, "loss": 0.0934, "step": 10675 }, { "epoch": 0.5, "learning_rate": 1.5004209935447658e-05, "loss": 0.1248, "step": 10680 }, { "epoch": 0.5, "learning_rate": 1.5001871082421181e-05, "loss": 0.0917, "step": 10685 }, { "epoch": 0.5, "learning_rate": 1.4999532229394706e-05, "loss": 0.1272, "step": 10690 }, { "epoch": 0.5, "learning_rate": 1.4997193376368231e-05, "loss": 0.0446, "step": 10695 }, { "epoch": 0.5, "learning_rate": 1.4994854523341755e-05, "loss": 0.1066, "step": 10700 }, { "epoch": 0.5, "learning_rate": 1.499251567031528e-05, "loss": 0.1435, "step": 10705 }, { "epoch": 0.5, "learning_rate": 1.4990176817288801e-05, "loss": 0.1685, "step": 10710 }, { "epoch": 0.5, "learning_rate": 1.4987837964262326e-05, "loss": 0.1059, "step": 10715 }, { "epoch": 0.5, "learning_rate": 1.4985499111235851e-05, "loss": 0.0819, "step": 10720 }, { "epoch": 0.5, "learning_rate": 1.4983160258209374e-05, "loss": 0.0605, "step": 10725 }, { "epoch": 0.5, "learning_rate": 1.49808214051829e-05, "loss": 0.0969, "step": 10730 }, { "epoch": 0.5, "learning_rate": 1.4978482552156424e-05, "loss": 0.0607, "step": 10735 }, { "epoch": 0.5, "learning_rate": 1.4976143699129947e-05, "loss": 0.1158, "step": 10740 }, { "epoch": 0.5, "learning_rate": 1.4973804846103472e-05, "loss": 0.1535, "step": 10745 }, { "epoch": 0.5, "learning_rate": 1.4971465993076997e-05, "loss": 0.1084, "step": 10750 }, { "epoch": 0.5, "learning_rate": 1.496912714005052e-05, "loss": 0.0964, "step": 10755 }, { "epoch": 0.5, "learning_rate": 1.4966788287024046e-05, "loss": 0.0811, "step": 10760 }, { "epoch": 0.5, "learning_rate": 1.496444943399757e-05, "loss": 0.1412, "step": 10765 }, { "epoch": 0.5, "learning_rate": 1.4962110580971092e-05, "loss": 0.1206, "step": 10770 }, { "epoch": 0.5, "learning_rate": 1.4959771727944617e-05, "loss": 0.112, "step": 10775 }, { "epoch": 0.5, "learning_rate": 1.495743287491814e-05, "loss": 0.0522, "step": 10780 }, { "epoch": 0.5, "learning_rate": 1.4955094021891665e-05, "loss": 0.0999, "step": 10785 }, { "epoch": 0.5, "learning_rate": 1.495275516886519e-05, "loss": 0.1485, "step": 10790 }, { "epoch": 0.5, "learning_rate": 1.4950416315838713e-05, "loss": 0.1121, "step": 10795 }, { "epoch": 0.51, "learning_rate": 1.4948077462812238e-05, "loss": 0.0993, "step": 10800 }, { "epoch": 0.51, "learning_rate": 1.4945738609785763e-05, "loss": 0.098, "step": 10805 }, { "epoch": 0.51, "learning_rate": 1.4943399756759287e-05, "loss": 0.1589, "step": 10810 }, { "epoch": 0.51, "learning_rate": 1.4941060903732812e-05, "loss": 0.0882, "step": 10815 }, { "epoch": 0.51, "learning_rate": 1.4938722050706336e-05, "loss": 0.0948, "step": 10820 }, { "epoch": 0.51, "learning_rate": 1.493638319767986e-05, "loss": 0.0555, "step": 10825 }, { "epoch": 0.51, "learning_rate": 1.4934044344653383e-05, "loss": 0.1384, "step": 10830 }, { "epoch": 0.51, "learning_rate": 1.4931705491626906e-05, "loss": 0.126, "step": 10835 }, { "epoch": 0.51, "learning_rate": 1.4929366638600431e-05, "loss": 0.0949, "step": 10840 }, { "epoch": 0.51, "learning_rate": 1.4927027785573956e-05, "loss": 0.0936, "step": 10845 }, { "epoch": 0.51, "learning_rate": 1.492468893254748e-05, "loss": 0.0686, "step": 10850 }, { "epoch": 0.51, "learning_rate": 1.4922350079521004e-05, "loss": 0.0891, "step": 10855 }, { "epoch": 0.51, "learning_rate": 1.492001122649453e-05, "loss": 0.1152, "step": 10860 }, { "epoch": 0.51, "learning_rate": 1.4917672373468053e-05, "loss": 0.1583, "step": 10865 }, { "epoch": 0.51, "learning_rate": 1.4915333520441578e-05, "loss": 0.1358, "step": 10870 }, { "epoch": 0.51, "learning_rate": 1.49129946674151e-05, "loss": 0.0744, "step": 10875 }, { "epoch": 0.51, "learning_rate": 1.4910655814388626e-05, "loss": 0.0883, "step": 10880 }, { "epoch": 0.51, "learning_rate": 1.4908316961362147e-05, "loss": 0.1071, "step": 10885 }, { "epoch": 0.51, "learning_rate": 1.4905978108335672e-05, "loss": 0.0844, "step": 10890 }, { "epoch": 0.51, "learning_rate": 1.4903639255309197e-05, "loss": 0.1164, "step": 10895 }, { "epoch": 0.51, "learning_rate": 1.490130040228272e-05, "loss": 0.1155, "step": 10900 }, { "epoch": 0.51, "learning_rate": 1.4898961549256245e-05, "loss": 0.0973, "step": 10905 }, { "epoch": 0.51, "learning_rate": 1.489662269622977e-05, "loss": 0.2084, "step": 10910 }, { "epoch": 0.51, "learning_rate": 1.4894283843203294e-05, "loss": 0.1858, "step": 10915 }, { "epoch": 0.51, "learning_rate": 1.4891944990176819e-05, "loss": 0.2636, "step": 10920 }, { "epoch": 0.51, "learning_rate": 1.4889606137150344e-05, "loss": 0.1674, "step": 10925 }, { "epoch": 0.51, "learning_rate": 1.4887267284123867e-05, "loss": 0.0947, "step": 10930 }, { "epoch": 0.51, "learning_rate": 1.4884928431097392e-05, "loss": 0.1792, "step": 10935 }, { "epoch": 0.51, "learning_rate": 1.4882589578070917e-05, "loss": 0.0888, "step": 10940 }, { "epoch": 0.51, "learning_rate": 1.4880250725044438e-05, "loss": 0.0489, "step": 10945 }, { "epoch": 0.51, "learning_rate": 1.4877911872017963e-05, "loss": 0.1081, "step": 10950 }, { "epoch": 0.51, "learning_rate": 1.4875573018991486e-05, "loss": 0.1027, "step": 10955 }, { "epoch": 0.51, "learning_rate": 1.4873234165965011e-05, "loss": 0.0966, "step": 10960 }, { "epoch": 0.51, "learning_rate": 1.4870895312938536e-05, "loss": 0.0891, "step": 10965 }, { "epoch": 0.51, "learning_rate": 1.486855645991206e-05, "loss": 0.1841, "step": 10970 }, { "epoch": 0.51, "learning_rate": 1.4866217606885585e-05, "loss": 0.2594, "step": 10975 }, { "epoch": 0.51, "learning_rate": 1.486387875385911e-05, "loss": 0.1812, "step": 10980 }, { "epoch": 0.51, "learning_rate": 1.4861539900832633e-05, "loss": 0.1241, "step": 10985 }, { "epoch": 0.51, "learning_rate": 1.4859201047806158e-05, "loss": 0.084, "step": 10990 }, { "epoch": 0.51, "learning_rate": 1.4856862194779683e-05, "loss": 0.0534, "step": 10995 }, { "epoch": 0.51, "learning_rate": 1.4854523341753206e-05, "loss": 0.1291, "step": 11000 }, { "epoch": 0.51, "learning_rate": 1.485218448872673e-05, "loss": 0.2053, "step": 11005 }, { "epoch": 0.52, "learning_rate": 1.4849845635700252e-05, "loss": 0.0485, "step": 11010 }, { "epoch": 0.52, "learning_rate": 1.4847506782673777e-05, "loss": 0.1043, "step": 11015 }, { "epoch": 0.52, "learning_rate": 1.4845167929647302e-05, "loss": 0.0646, "step": 11020 }, { "epoch": 0.52, "learning_rate": 1.4842829076620826e-05, "loss": 0.0599, "step": 11025 }, { "epoch": 0.52, "learning_rate": 1.484049022359435e-05, "loss": 0.1333, "step": 11030 }, { "epoch": 0.52, "learning_rate": 1.4838151370567876e-05, "loss": 0.1564, "step": 11035 }, { "epoch": 0.52, "learning_rate": 1.4835812517541399e-05, "loss": 0.067, "step": 11040 }, { "epoch": 0.52, "learning_rate": 1.4833473664514924e-05, "loss": 0.1569, "step": 11045 }, { "epoch": 0.52, "learning_rate": 1.4831134811488449e-05, "loss": 0.1389, "step": 11050 }, { "epoch": 0.52, "learning_rate": 1.4828795958461972e-05, "loss": 0.0612, "step": 11055 }, { "epoch": 0.52, "learning_rate": 1.4826457105435497e-05, "loss": 0.062, "step": 11060 }, { "epoch": 0.52, "learning_rate": 1.4824118252409018e-05, "loss": 0.0748, "step": 11065 }, { "epoch": 0.52, "learning_rate": 1.4821779399382543e-05, "loss": 0.1921, "step": 11070 }, { "epoch": 0.52, "learning_rate": 1.4819440546356068e-05, "loss": 0.117, "step": 11075 }, { "epoch": 0.52, "learning_rate": 1.4817101693329592e-05, "loss": 0.0868, "step": 11080 }, { "epoch": 0.52, "learning_rate": 1.4814762840303117e-05, "loss": 0.0849, "step": 11085 }, { "epoch": 0.52, "learning_rate": 1.481242398727664e-05, "loss": 0.0558, "step": 11090 }, { "epoch": 0.52, "learning_rate": 1.4810085134250165e-05, "loss": 0.0979, "step": 11095 }, { "epoch": 0.52, "learning_rate": 1.480774628122369e-05, "loss": 0.1408, "step": 11100 }, { "epoch": 0.52, "learning_rate": 1.4805407428197213e-05, "loss": 0.1615, "step": 11105 }, { "epoch": 0.52, "learning_rate": 1.4803068575170738e-05, "loss": 0.0631, "step": 11110 }, { "epoch": 0.52, "learning_rate": 1.4800729722144263e-05, "loss": 0.0917, "step": 11115 }, { "epoch": 0.52, "learning_rate": 1.4798390869117786e-05, "loss": 0.1083, "step": 11120 }, { "epoch": 0.52, "learning_rate": 1.479605201609131e-05, "loss": 0.1281, "step": 11125 }, { "epoch": 0.52, "learning_rate": 1.4793713163064833e-05, "loss": 0.0967, "step": 11130 }, { "epoch": 0.52, "learning_rate": 1.4791374310038358e-05, "loss": 0.1201, "step": 11135 }, { "epoch": 0.52, "learning_rate": 1.4789035457011883e-05, "loss": 0.1335, "step": 11140 }, { "epoch": 0.52, "learning_rate": 1.4786696603985406e-05, "loss": 0.091, "step": 11145 }, { "epoch": 0.52, "learning_rate": 1.478435775095893e-05, "loss": 0.0885, "step": 11150 }, { "epoch": 0.52, "learning_rate": 1.4782018897932456e-05, "loss": 0.0666, "step": 11155 }, { "epoch": 0.52, "learning_rate": 1.4779680044905979e-05, "loss": 0.1271, "step": 11160 }, { "epoch": 0.52, "learning_rate": 1.4777341191879504e-05, "loss": 0.0801, "step": 11165 }, { "epoch": 0.52, "learning_rate": 1.4775002338853029e-05, "loss": 0.0779, "step": 11170 }, { "epoch": 0.52, "learning_rate": 1.4772663485826552e-05, "loss": 0.0651, "step": 11175 }, { "epoch": 0.52, "learning_rate": 1.4770324632800077e-05, "loss": 0.0483, "step": 11180 }, { "epoch": 0.52, "learning_rate": 1.4767985779773599e-05, "loss": 0.0954, "step": 11185 }, { "epoch": 0.52, "learning_rate": 1.4765646926747124e-05, "loss": 0.1127, "step": 11190 }, { "epoch": 0.52, "learning_rate": 1.4763308073720649e-05, "loss": 0.1149, "step": 11195 }, { "epoch": 0.52, "learning_rate": 1.4760969220694172e-05, "loss": 0.0719, "step": 11200 }, { "epoch": 0.52, "learning_rate": 1.4758630367667697e-05, "loss": 0.1329, "step": 11205 }, { "epoch": 0.52, "learning_rate": 1.4756291514641222e-05, "loss": 0.0478, "step": 11210 }, { "epoch": 0.52, "learning_rate": 1.4753952661614745e-05, "loss": 0.0556, "step": 11215 }, { "epoch": 0.52, "learning_rate": 1.475161380858827e-05, "loss": 0.0816, "step": 11220 }, { "epoch": 0.53, "learning_rate": 1.4749274955561795e-05, "loss": 0.1263, "step": 11225 }, { "epoch": 0.53, "learning_rate": 1.4746936102535318e-05, "loss": 0.1259, "step": 11230 }, { "epoch": 0.53, "learning_rate": 1.4744597249508843e-05, "loss": 0.0852, "step": 11235 }, { "epoch": 0.53, "learning_rate": 1.4742258396482368e-05, "loss": 0.0972, "step": 11240 }, { "epoch": 0.53, "learning_rate": 1.473991954345589e-05, "loss": 0.1433, "step": 11245 }, { "epoch": 0.53, "learning_rate": 1.4737580690429415e-05, "loss": 0.0585, "step": 11250 }, { "epoch": 0.53, "learning_rate": 1.4735241837402938e-05, "loss": 0.0423, "step": 11255 }, { "epoch": 0.53, "learning_rate": 1.4732902984376463e-05, "loss": 0.1229, "step": 11260 }, { "epoch": 0.53, "learning_rate": 1.4730564131349988e-05, "loss": 0.0853, "step": 11265 }, { "epoch": 0.53, "learning_rate": 1.4728225278323511e-05, "loss": 0.1066, "step": 11270 }, { "epoch": 0.53, "learning_rate": 1.4725886425297036e-05, "loss": 0.0644, "step": 11275 }, { "epoch": 0.53, "learning_rate": 1.4723547572270561e-05, "loss": 0.1375, "step": 11280 }, { "epoch": 0.53, "learning_rate": 1.4721208719244084e-05, "loss": 0.149, "step": 11285 }, { "epoch": 0.53, "learning_rate": 1.4718869866217609e-05, "loss": 0.0719, "step": 11290 }, { "epoch": 0.53, "learning_rate": 1.4716531013191134e-05, "loss": 0.0818, "step": 11295 }, { "epoch": 0.53, "learning_rate": 1.4714192160164656e-05, "loss": 0.0907, "step": 11300 }, { "epoch": 0.53, "learning_rate": 1.471185330713818e-05, "loss": 0.1606, "step": 11305 }, { "epoch": 0.53, "learning_rate": 1.4709514454111704e-05, "loss": 0.0888, "step": 11310 }, { "epoch": 0.53, "learning_rate": 1.4707175601085229e-05, "loss": 0.0872, "step": 11315 }, { "epoch": 0.53, "learning_rate": 1.4704836748058752e-05, "loss": 0.0868, "step": 11320 }, { "epoch": 0.53, "learning_rate": 1.4702497895032277e-05, "loss": 0.0689, "step": 11325 }, { "epoch": 0.53, "learning_rate": 1.4700159042005802e-05, "loss": 0.0714, "step": 11330 }, { "epoch": 0.53, "learning_rate": 1.4697820188979325e-05, "loss": 0.1051, "step": 11335 }, { "epoch": 0.53, "learning_rate": 1.469548133595285e-05, "loss": 0.1064, "step": 11340 }, { "epoch": 0.53, "learning_rate": 1.4693142482926375e-05, "loss": 0.0641, "step": 11345 }, { "epoch": 0.53, "learning_rate": 1.4690803629899898e-05, "loss": 0.1005, "step": 11350 }, { "epoch": 0.53, "learning_rate": 1.4688464776873423e-05, "loss": 0.0673, "step": 11355 }, { "epoch": 0.53, "learning_rate": 1.4686125923846945e-05, "loss": 0.0736, "step": 11360 }, { "epoch": 0.53, "learning_rate": 1.468378707082047e-05, "loss": 0.1207, "step": 11365 }, { "epoch": 0.53, "learning_rate": 1.4681448217793995e-05, "loss": 0.1393, "step": 11370 }, { "epoch": 0.53, "learning_rate": 1.4679109364767518e-05, "loss": 0.1332, "step": 11375 }, { "epoch": 0.53, "learning_rate": 1.4676770511741043e-05, "loss": 0.0469, "step": 11380 }, { "epoch": 0.53, "learning_rate": 1.4674431658714568e-05, "loss": 0.093, "step": 11385 }, { "epoch": 0.53, "learning_rate": 1.4672092805688091e-05, "loss": 0.1461, "step": 11390 }, { "epoch": 0.53, "learning_rate": 1.4669753952661616e-05, "loss": 0.0758, "step": 11395 }, { "epoch": 0.53, "learning_rate": 1.4667415099635141e-05, "loss": 0.0973, "step": 11400 }, { "epoch": 0.53, "learning_rate": 1.4665076246608664e-05, "loss": 0.1023, "step": 11405 }, { "epoch": 0.53, "learning_rate": 1.466273739358219e-05, "loss": 0.0891, "step": 11410 }, { "epoch": 0.53, "learning_rate": 1.4660398540555714e-05, "loss": 0.0852, "step": 11415 }, { "epoch": 0.53, "learning_rate": 1.4658059687529236e-05, "loss": 0.0828, "step": 11420 }, { "epoch": 0.53, "learning_rate": 1.465572083450276e-05, "loss": 0.1169, "step": 11425 }, { "epoch": 0.53, "learning_rate": 1.4653381981476284e-05, "loss": 0.0582, "step": 11430 }, { "epoch": 0.53, "learning_rate": 1.4651043128449809e-05, "loss": 0.1309, "step": 11435 }, { "epoch": 0.54, "learning_rate": 1.4648704275423334e-05, "loss": 0.1105, "step": 11440 }, { "epoch": 0.54, "learning_rate": 1.4646365422396857e-05, "loss": 0.0724, "step": 11445 }, { "epoch": 0.54, "learning_rate": 1.4644026569370382e-05, "loss": 0.0845, "step": 11450 }, { "epoch": 0.54, "learning_rate": 1.4641687716343907e-05, "loss": 0.1398, "step": 11455 }, { "epoch": 0.54, "learning_rate": 1.463934886331743e-05, "loss": 0.1493, "step": 11460 }, { "epoch": 0.54, "learning_rate": 1.4637010010290955e-05, "loss": 0.0811, "step": 11465 }, { "epoch": 0.54, "learning_rate": 1.463467115726448e-05, "loss": 0.1707, "step": 11470 }, { "epoch": 0.54, "learning_rate": 1.4632332304238003e-05, "loss": 0.1542, "step": 11475 }, { "epoch": 0.54, "learning_rate": 1.4629993451211527e-05, "loss": 0.1235, "step": 11480 }, { "epoch": 0.54, "learning_rate": 1.462765459818505e-05, "loss": 0.1394, "step": 11485 }, { "epoch": 0.54, "learning_rate": 1.4625315745158575e-05, "loss": 0.0852, "step": 11490 }, { "epoch": 0.54, "learning_rate": 1.46229768921321e-05, "loss": 0.133, "step": 11495 }, { "epoch": 0.54, "learning_rate": 1.4620638039105623e-05, "loss": 0.1435, "step": 11500 }, { "epoch": 0.54, "learning_rate": 1.4618299186079148e-05, "loss": 0.1132, "step": 11505 }, { "epoch": 0.54, "learning_rate": 1.4615960333052673e-05, "loss": 0.1086, "step": 11510 }, { "epoch": 0.54, "learning_rate": 1.4613621480026196e-05, "loss": 0.0818, "step": 11515 }, { "epoch": 0.54, "learning_rate": 1.4611282626999721e-05, "loss": 0.0616, "step": 11520 }, { "epoch": 0.54, "learning_rate": 1.4608943773973245e-05, "loss": 0.1639, "step": 11525 }, { "epoch": 0.54, "learning_rate": 1.460660492094677e-05, "loss": 0.1101, "step": 11530 }, { "epoch": 0.54, "learning_rate": 1.4604266067920294e-05, "loss": 0.117, "step": 11535 }, { "epoch": 0.54, "learning_rate": 1.4601927214893816e-05, "loss": 0.1003, "step": 11540 }, { "epoch": 0.54, "learning_rate": 1.4599588361867341e-05, "loss": 0.0739, "step": 11545 }, { "epoch": 0.54, "learning_rate": 1.4597249508840864e-05, "loss": 0.0581, "step": 11550 }, { "epoch": 0.54, "learning_rate": 1.459491065581439e-05, "loss": 0.1454, "step": 11555 }, { "epoch": 0.54, "learning_rate": 1.4592571802787914e-05, "loss": 0.0702, "step": 11560 }, { "epoch": 0.54, "learning_rate": 1.4590232949761437e-05, "loss": 0.0715, "step": 11565 }, { "epoch": 0.54, "learning_rate": 1.4587894096734962e-05, "loss": 0.085, "step": 11570 }, { "epoch": 0.54, "learning_rate": 1.4585555243708487e-05, "loss": 0.1397, "step": 11575 }, { "epoch": 0.54, "learning_rate": 1.458321639068201e-05, "loss": 0.0925, "step": 11580 }, { "epoch": 0.54, "learning_rate": 1.4580877537655535e-05, "loss": 0.0231, "step": 11585 }, { "epoch": 0.54, "learning_rate": 1.457853868462906e-05, "loss": 0.1002, "step": 11590 }, { "epoch": 0.54, "learning_rate": 1.4576199831602584e-05, "loss": 0.104, "step": 11595 }, { "epoch": 0.54, "learning_rate": 1.4573860978576107e-05, "loss": 0.0693, "step": 11600 }, { "epoch": 0.54, "learning_rate": 1.457152212554963e-05, "loss": 0.076, "step": 11605 }, { "epoch": 0.54, "learning_rate": 1.4569183272523155e-05, "loss": 0.1401, "step": 11610 }, { "epoch": 0.54, "learning_rate": 1.456684441949668e-05, "loss": 0.1094, "step": 11615 }, { "epoch": 0.54, "learning_rate": 1.4564505566470203e-05, "loss": 0.0987, "step": 11620 }, { "epoch": 0.54, "learning_rate": 1.4562166713443728e-05, "loss": 0.0624, "step": 11625 }, { "epoch": 0.54, "learning_rate": 1.4559827860417253e-05, "loss": 0.0949, "step": 11630 }, { "epoch": 0.54, "learning_rate": 1.4557489007390777e-05, "loss": 0.1162, "step": 11635 }, { "epoch": 0.54, "learning_rate": 1.4555150154364301e-05, "loss": 0.0903, "step": 11640 }, { "epoch": 0.54, "learning_rate": 1.4552811301337826e-05, "loss": 0.1303, "step": 11645 }, { "epoch": 0.54, "learning_rate": 1.455047244831135e-05, "loss": 0.1112, "step": 11650 }, { "epoch": 0.55, "learning_rate": 1.4548133595284875e-05, "loss": 0.0858, "step": 11655 }, { "epoch": 0.55, "learning_rate": 1.4545794742258396e-05, "loss": 0.0945, "step": 11660 }, { "epoch": 0.55, "learning_rate": 1.4543455889231921e-05, "loss": 0.1437, "step": 11665 }, { "epoch": 0.55, "learning_rate": 1.4541117036205446e-05, "loss": 0.1306, "step": 11670 }, { "epoch": 0.55, "learning_rate": 1.453877818317897e-05, "loss": 0.1456, "step": 11675 }, { "epoch": 0.55, "learning_rate": 1.4536439330152494e-05, "loss": 0.1259, "step": 11680 }, { "epoch": 0.55, "learning_rate": 1.453410047712602e-05, "loss": 0.0952, "step": 11685 }, { "epoch": 0.55, "learning_rate": 1.4531761624099543e-05, "loss": 0.0611, "step": 11690 }, { "epoch": 0.55, "learning_rate": 1.4529422771073067e-05, "loss": 0.1069, "step": 11695 }, { "epoch": 0.55, "learning_rate": 1.4527083918046592e-05, "loss": 0.0773, "step": 11700 }, { "epoch": 0.55, "learning_rate": 1.4524745065020116e-05, "loss": 0.0821, "step": 11705 }, { "epoch": 0.55, "learning_rate": 1.452240621199364e-05, "loss": 0.09, "step": 11710 }, { "epoch": 0.55, "learning_rate": 1.4520067358967166e-05, "loss": 0.1144, "step": 11715 }, { "epoch": 0.55, "learning_rate": 1.4517728505940687e-05, "loss": 0.0933, "step": 11720 }, { "epoch": 0.55, "learning_rate": 1.4515389652914212e-05, "loss": 0.1218, "step": 11725 }, { "epoch": 0.55, "learning_rate": 1.4513050799887735e-05, "loss": 0.1308, "step": 11730 }, { "epoch": 0.55, "learning_rate": 1.451071194686126e-05, "loss": 0.1322, "step": 11735 }, { "epoch": 0.55, "learning_rate": 1.4508373093834785e-05, "loss": 0.1244, "step": 11740 }, { "epoch": 0.55, "learning_rate": 1.4506034240808309e-05, "loss": 0.1162, "step": 11745 }, { "epoch": 0.55, "learning_rate": 1.4503695387781833e-05, "loss": 0.1, "step": 11750 }, { "epoch": 0.55, "learning_rate": 1.4501356534755357e-05, "loss": 0.0437, "step": 11755 }, { "epoch": 0.55, "learning_rate": 1.4499017681728882e-05, "loss": 0.1219, "step": 11760 }, { "epoch": 0.55, "learning_rate": 1.4496678828702407e-05, "loss": 0.1078, "step": 11765 }, { "epoch": 0.55, "learning_rate": 1.449433997567593e-05, "loss": 0.0848, "step": 11770 }, { "epoch": 0.55, "learning_rate": 1.4492001122649453e-05, "loss": 0.1037, "step": 11775 }, { "epoch": 0.55, "learning_rate": 1.4489662269622976e-05, "loss": 0.139, "step": 11780 }, { "epoch": 0.55, "learning_rate": 1.4487323416596501e-05, "loss": 0.0893, "step": 11785 }, { "epoch": 0.55, "learning_rate": 1.4484984563570026e-05, "loss": 0.21, "step": 11790 }, { "epoch": 0.55, "learning_rate": 1.448264571054355e-05, "loss": 0.1352, "step": 11795 }, { "epoch": 0.55, "learning_rate": 1.4480306857517075e-05, "loss": 0.1628, "step": 11800 }, { "epoch": 0.55, "learning_rate": 1.44779680044906e-05, "loss": 0.1576, "step": 11805 }, { "epoch": 0.55, "learning_rate": 1.4475629151464123e-05, "loss": 0.154, "step": 11810 }, { "epoch": 0.55, "learning_rate": 1.4473290298437648e-05, "loss": 0.2199, "step": 11815 }, { "epoch": 0.55, "learning_rate": 1.4470951445411173e-05, "loss": 0.2449, "step": 11820 }, { "epoch": 0.55, "learning_rate": 1.4468612592384696e-05, "loss": 0.1473, "step": 11825 }, { "epoch": 0.55, "learning_rate": 1.446627373935822e-05, "loss": 0.1223, "step": 11830 }, { "epoch": 0.55, "learning_rate": 1.4463934886331742e-05, "loss": 0.0824, "step": 11835 }, { "epoch": 0.55, "learning_rate": 1.4461596033305267e-05, "loss": 0.1177, "step": 11840 }, { "epoch": 0.55, "learning_rate": 1.4459257180278792e-05, "loss": 0.1507, "step": 11845 }, { "epoch": 0.55, "learning_rate": 1.4456918327252316e-05, "loss": 0.1748, "step": 11850 }, { "epoch": 0.55, "learning_rate": 1.445457947422584e-05, "loss": 0.1416, "step": 11855 }, { "epoch": 0.55, "learning_rate": 1.4452240621199365e-05, "loss": 0.0923, "step": 11860 }, { "epoch": 0.55, "learning_rate": 1.4449901768172889e-05, "loss": 0.079, "step": 11865 }, { "epoch": 0.56, "learning_rate": 1.4447562915146414e-05, "loss": 0.1002, "step": 11870 }, { "epoch": 0.56, "learning_rate": 1.4445224062119939e-05, "loss": 0.1119, "step": 11875 }, { "epoch": 0.56, "learning_rate": 1.4442885209093462e-05, "loss": 0.0854, "step": 11880 }, { "epoch": 0.56, "learning_rate": 1.4440546356066987e-05, "loss": 0.0764, "step": 11885 }, { "epoch": 0.56, "learning_rate": 1.4438207503040512e-05, "loss": 0.1322, "step": 11890 }, { "epoch": 0.56, "learning_rate": 1.4435868650014033e-05, "loss": 0.084, "step": 11895 }, { "epoch": 0.56, "learning_rate": 1.4433529796987558e-05, "loss": 0.0468, "step": 11900 }, { "epoch": 0.56, "learning_rate": 1.4431190943961082e-05, "loss": 0.0627, "step": 11905 }, { "epoch": 0.56, "learning_rate": 1.4428852090934607e-05, "loss": 0.1264, "step": 11910 }, { "epoch": 0.56, "learning_rate": 1.4426513237908131e-05, "loss": 0.1407, "step": 11915 }, { "epoch": 0.56, "learning_rate": 1.4424174384881655e-05, "loss": 0.1878, "step": 11920 }, { "epoch": 0.56, "learning_rate": 1.442183553185518e-05, "loss": 0.1416, "step": 11925 }, { "epoch": 0.56, "learning_rate": 1.4419496678828705e-05, "loss": 0.2002, "step": 11930 }, { "epoch": 0.56, "learning_rate": 1.4417157825802228e-05, "loss": 0.2219, "step": 11935 }, { "epoch": 0.56, "learning_rate": 1.4414818972775753e-05, "loss": 0.1219, "step": 11940 }, { "epoch": 0.56, "learning_rate": 1.4412480119749278e-05, "loss": 0.1104, "step": 11945 }, { "epoch": 0.56, "learning_rate": 1.4410141266722801e-05, "loss": 0.0887, "step": 11950 }, { "epoch": 0.56, "learning_rate": 1.4407802413696324e-05, "loss": 0.0626, "step": 11955 }, { "epoch": 0.56, "learning_rate": 1.4405463560669848e-05, "loss": 0.1051, "step": 11960 }, { "epoch": 0.56, "learning_rate": 1.4403124707643373e-05, "loss": 0.1364, "step": 11965 }, { "epoch": 0.56, "learning_rate": 1.4400785854616896e-05, "loss": 0.1109, "step": 11970 }, { "epoch": 0.56, "learning_rate": 1.439844700159042e-05, "loss": 0.1182, "step": 11975 }, { "epoch": 0.56, "learning_rate": 1.4396108148563946e-05, "loss": 0.0737, "step": 11980 }, { "epoch": 0.56, "learning_rate": 1.4393769295537469e-05, "loss": 0.1243, "step": 11985 }, { "epoch": 0.56, "learning_rate": 1.4391430442510994e-05, "loss": 0.1282, "step": 11990 }, { "epoch": 0.56, "learning_rate": 1.4389091589484519e-05, "loss": 0.0854, "step": 11995 }, { "epoch": 0.56, "learning_rate": 1.4386752736458042e-05, "loss": 0.1595, "step": 12000 }, { "epoch": 0.56, "learning_rate": 1.4384413883431567e-05, "loss": 0.1269, "step": 12005 }, { "epoch": 0.56, "learning_rate": 1.4382075030405092e-05, "loss": 0.0864, "step": 12010 }, { "epoch": 0.56, "learning_rate": 1.4379736177378614e-05, "loss": 0.116, "step": 12015 }, { "epoch": 0.56, "learning_rate": 1.4377397324352139e-05, "loss": 0.0826, "step": 12020 }, { "epoch": 0.56, "learning_rate": 1.4375058471325662e-05, "loss": 0.1174, "step": 12025 }, { "epoch": 0.56, "learning_rate": 1.4372719618299187e-05, "loss": 0.1136, "step": 12030 }, { "epoch": 0.56, "learning_rate": 1.4370380765272712e-05, "loss": 0.0742, "step": 12035 }, { "epoch": 0.56, "learning_rate": 1.4368041912246235e-05, "loss": 0.1072, "step": 12040 }, { "epoch": 0.56, "learning_rate": 1.436570305921976e-05, "loss": 0.0828, "step": 12045 }, { "epoch": 0.56, "learning_rate": 1.4363364206193285e-05, "loss": 0.1183, "step": 12050 }, { "epoch": 0.56, "learning_rate": 1.4361025353166808e-05, "loss": 0.0886, "step": 12055 }, { "epoch": 0.56, "learning_rate": 1.4358686500140333e-05, "loss": 0.1029, "step": 12060 }, { "epoch": 0.56, "learning_rate": 1.4356347647113858e-05, "loss": 0.0654, "step": 12065 }, { "epoch": 0.56, "learning_rate": 1.4354008794087381e-05, "loss": 0.1529, "step": 12070 }, { "epoch": 0.56, "learning_rate": 1.4351669941060904e-05, "loss": 0.0916, "step": 12075 }, { "epoch": 0.57, "learning_rate": 1.4349331088034428e-05, "loss": 0.0617, "step": 12080 }, { "epoch": 0.57, "learning_rate": 1.4346992235007953e-05, "loss": 0.0362, "step": 12085 }, { "epoch": 0.57, "learning_rate": 1.4344653381981478e-05, "loss": 0.1454, "step": 12090 }, { "epoch": 0.57, "learning_rate": 1.4342314528955001e-05, "loss": 0.1213, "step": 12095 }, { "epoch": 0.57, "learning_rate": 1.4339975675928526e-05, "loss": 0.1472, "step": 12100 }, { "epoch": 0.57, "learning_rate": 1.433763682290205e-05, "loss": 0.1024, "step": 12105 }, { "epoch": 0.57, "learning_rate": 1.4335297969875574e-05, "loss": 0.1555, "step": 12110 }, { "epoch": 0.57, "learning_rate": 1.4332959116849099e-05, "loss": 0.1242, "step": 12115 }, { "epoch": 0.57, "learning_rate": 1.4330620263822624e-05, "loss": 0.0687, "step": 12120 }, { "epoch": 0.57, "learning_rate": 1.4328281410796147e-05, "loss": 0.0887, "step": 12125 }, { "epoch": 0.57, "learning_rate": 1.4325942557769672e-05, "loss": 0.0921, "step": 12130 }, { "epoch": 0.57, "learning_rate": 1.4323603704743194e-05, "loss": 0.1177, "step": 12135 }, { "epoch": 0.57, "learning_rate": 1.4321264851716719e-05, "loss": 0.0637, "step": 12140 }, { "epoch": 0.57, "learning_rate": 1.4318925998690244e-05, "loss": 0.1065, "step": 12145 }, { "epoch": 0.57, "learning_rate": 1.4316587145663767e-05, "loss": 0.1345, "step": 12150 }, { "epoch": 0.57, "learning_rate": 1.4314248292637292e-05, "loss": 0.1313, "step": 12155 }, { "epoch": 0.57, "learning_rate": 1.4311909439610817e-05, "loss": 0.0961, "step": 12160 }, { "epoch": 0.57, "learning_rate": 1.430957058658434e-05, "loss": 0.1472, "step": 12165 }, { "epoch": 0.57, "learning_rate": 1.4307231733557865e-05, "loss": 0.1089, "step": 12170 }, { "epoch": 0.57, "learning_rate": 1.430489288053139e-05, "loss": 0.1578, "step": 12175 }, { "epoch": 0.57, "learning_rate": 1.4302554027504913e-05, "loss": 0.0664, "step": 12180 }, { "epoch": 0.57, "learning_rate": 1.4300215174478438e-05, "loss": 0.1316, "step": 12185 }, { "epoch": 0.57, "learning_rate": 1.429787632145196e-05, "loss": 0.1307, "step": 12190 }, { "epoch": 0.57, "learning_rate": 1.4295537468425485e-05, "loss": 0.0709, "step": 12195 }, { "epoch": 0.57, "learning_rate": 1.4293198615399008e-05, "loss": 0.163, "step": 12200 }, { "epoch": 0.57, "learning_rate": 1.4290859762372533e-05, "loss": 0.0895, "step": 12205 }, { "epoch": 0.57, "learning_rate": 1.4288520909346058e-05, "loss": 0.1338, "step": 12210 }, { "epoch": 0.57, "learning_rate": 1.4286182056319581e-05, "loss": 0.1012, "step": 12215 }, { "epoch": 0.57, "learning_rate": 1.4283843203293106e-05, "loss": 0.1163, "step": 12220 }, { "epoch": 0.57, "learning_rate": 1.4281504350266631e-05, "loss": 0.1006, "step": 12225 }, { "epoch": 0.57, "learning_rate": 1.4279165497240154e-05, "loss": 0.1316, "step": 12230 }, { "epoch": 0.57, "learning_rate": 1.427682664421368e-05, "loss": 0.0341, "step": 12235 }, { "epoch": 0.57, "learning_rate": 1.4274487791187204e-05, "loss": 0.0679, "step": 12240 }, { "epoch": 0.57, "learning_rate": 1.4272148938160727e-05, "loss": 0.1265, "step": 12245 }, { "epoch": 0.57, "learning_rate": 1.426981008513425e-05, "loss": 0.0539, "step": 12250 }, { "epoch": 0.57, "learning_rate": 1.4267471232107774e-05, "loss": 0.1337, "step": 12255 }, { "epoch": 0.57, "learning_rate": 1.4265132379081299e-05, "loss": 0.1536, "step": 12260 }, { "epoch": 0.57, "learning_rate": 1.4262793526054824e-05, "loss": 0.1163, "step": 12265 }, { "epoch": 0.57, "learning_rate": 1.4260454673028347e-05, "loss": 0.146, "step": 12270 }, { "epoch": 0.57, "learning_rate": 1.4258115820001872e-05, "loss": 0.1271, "step": 12275 }, { "epoch": 0.57, "learning_rate": 1.4255776966975397e-05, "loss": 0.1186, "step": 12280 }, { "epoch": 0.57, "learning_rate": 1.425343811394892e-05, "loss": 0.0804, "step": 12285 }, { "epoch": 0.57, "learning_rate": 1.4251099260922445e-05, "loss": 0.0932, "step": 12290 }, { "epoch": 0.58, "learning_rate": 1.424876040789597e-05, "loss": 0.0806, "step": 12295 }, { "epoch": 0.58, "learning_rate": 1.4246421554869493e-05, "loss": 0.0387, "step": 12300 }, { "epoch": 0.58, "learning_rate": 1.4244082701843018e-05, "loss": 0.1047, "step": 12305 }, { "epoch": 0.58, "learning_rate": 1.424174384881654e-05, "loss": 0.1525, "step": 12310 }, { "epoch": 0.58, "learning_rate": 1.4239404995790065e-05, "loss": 0.0855, "step": 12315 }, { "epoch": 0.58, "learning_rate": 1.423706614276359e-05, "loss": 0.0567, "step": 12320 }, { "epoch": 0.58, "learning_rate": 1.4234727289737113e-05, "loss": 0.1196, "step": 12325 }, { "epoch": 0.58, "learning_rate": 1.4232388436710638e-05, "loss": 0.1143, "step": 12330 }, { "epoch": 0.58, "learning_rate": 1.4230049583684163e-05, "loss": 0.0977, "step": 12335 }, { "epoch": 0.58, "learning_rate": 1.4227710730657686e-05, "loss": 0.0597, "step": 12340 }, { "epoch": 0.58, "learning_rate": 1.4225371877631211e-05, "loss": 0.1006, "step": 12345 }, { "epoch": 0.58, "learning_rate": 1.4223033024604736e-05, "loss": 0.1035, "step": 12350 }, { "epoch": 0.58, "learning_rate": 1.422069417157826e-05, "loss": 0.083, "step": 12355 }, { "epoch": 0.58, "learning_rate": 1.4218355318551784e-05, "loss": 0.0602, "step": 12360 }, { "epoch": 0.58, "learning_rate": 1.421601646552531e-05, "loss": 0.1214, "step": 12365 }, { "epoch": 0.58, "learning_rate": 1.4213677612498831e-05, "loss": 0.1355, "step": 12370 }, { "epoch": 0.58, "learning_rate": 1.4211338759472356e-05, "loss": 0.1228, "step": 12375 }, { "epoch": 0.58, "learning_rate": 1.4208999906445879e-05, "loss": 0.0616, "step": 12380 }, { "epoch": 0.58, "learning_rate": 1.4206661053419404e-05, "loss": 0.1002, "step": 12385 }, { "epoch": 0.58, "learning_rate": 1.4204322200392929e-05, "loss": 0.04, "step": 12390 }, { "epoch": 0.58, "learning_rate": 1.4201983347366452e-05, "loss": 0.0746, "step": 12395 }, { "epoch": 0.58, "learning_rate": 1.4199644494339977e-05, "loss": 0.0431, "step": 12400 }, { "epoch": 0.58, "learning_rate": 1.41973056413135e-05, "loss": 0.0952, "step": 12405 }, { "epoch": 0.58, "learning_rate": 1.4194966788287025e-05, "loss": 0.069, "step": 12410 }, { "epoch": 0.58, "learning_rate": 1.419262793526055e-05, "loss": 0.0862, "step": 12415 }, { "epoch": 0.58, "learning_rate": 1.4190289082234074e-05, "loss": 0.1612, "step": 12420 }, { "epoch": 0.58, "learning_rate": 1.4187950229207599e-05, "loss": 0.1165, "step": 12425 }, { "epoch": 0.58, "learning_rate": 1.418561137618112e-05, "loss": 0.1546, "step": 12430 }, { "epoch": 0.58, "learning_rate": 1.4183272523154645e-05, "loss": 0.0946, "step": 12435 }, { "epoch": 0.58, "learning_rate": 1.418093367012817e-05, "loss": 0.1093, "step": 12440 }, { "epoch": 0.58, "learning_rate": 1.4178594817101693e-05, "loss": 0.0971, "step": 12445 }, { "epoch": 0.58, "learning_rate": 1.4176255964075218e-05, "loss": 0.1568, "step": 12450 }, { "epoch": 0.58, "learning_rate": 1.4173917111048743e-05, "loss": 0.0513, "step": 12455 }, { "epoch": 0.58, "learning_rate": 1.4171578258022266e-05, "loss": 0.0402, "step": 12460 }, { "epoch": 0.58, "learning_rate": 1.4169239404995791e-05, "loss": 0.1155, "step": 12465 }, { "epoch": 0.58, "learning_rate": 1.4166900551969316e-05, "loss": 0.0857, "step": 12470 }, { "epoch": 0.58, "learning_rate": 1.416456169894284e-05, "loss": 0.0863, "step": 12475 }, { "epoch": 0.58, "learning_rate": 1.4162222845916365e-05, "loss": 0.0923, "step": 12480 }, { "epoch": 0.58, "learning_rate": 1.415988399288989e-05, "loss": 0.102, "step": 12485 }, { "epoch": 0.58, "learning_rate": 1.4157545139863411e-05, "loss": 0.098, "step": 12490 }, { "epoch": 0.58, "learning_rate": 1.4155206286836936e-05, "loss": 0.1164, "step": 12495 }, { "epoch": 0.58, "learning_rate": 1.415286743381046e-05, "loss": 0.0985, "step": 12500 }, { "epoch": 0.58, "learning_rate": 1.4150528580783984e-05, "loss": 0.102, "step": 12505 }, { "epoch": 0.59, "learning_rate": 1.414818972775751e-05, "loss": 0.0855, "step": 12510 }, { "epoch": 0.59, "learning_rate": 1.4145850874731032e-05, "loss": 0.0934, "step": 12515 }, { "epoch": 0.59, "learning_rate": 1.4143512021704557e-05, "loss": 0.0639, "step": 12520 }, { "epoch": 0.59, "learning_rate": 1.4141173168678082e-05, "loss": 0.0933, "step": 12525 }, { "epoch": 0.59, "learning_rate": 1.4138834315651606e-05, "loss": 0.075, "step": 12530 }, { "epoch": 0.59, "learning_rate": 1.413649546262513e-05, "loss": 0.1371, "step": 12535 }, { "epoch": 0.59, "learning_rate": 1.4134156609598656e-05, "loss": 0.18, "step": 12540 }, { "epoch": 0.59, "learning_rate": 1.4131817756572179e-05, "loss": 0.0584, "step": 12545 }, { "epoch": 0.59, "learning_rate": 1.4129478903545702e-05, "loss": 0.0841, "step": 12550 }, { "epoch": 0.59, "learning_rate": 1.4127140050519225e-05, "loss": 0.1125, "step": 12555 }, { "epoch": 0.59, "learning_rate": 1.412480119749275e-05, "loss": 0.0879, "step": 12560 }, { "epoch": 0.59, "learning_rate": 1.4122462344466275e-05, "loss": 0.0708, "step": 12565 }, { "epoch": 0.59, "learning_rate": 1.4120123491439798e-05, "loss": 0.1288, "step": 12570 }, { "epoch": 0.59, "learning_rate": 1.4117784638413323e-05, "loss": 0.0793, "step": 12575 }, { "epoch": 0.59, "learning_rate": 1.4115445785386848e-05, "loss": 0.0818, "step": 12580 }, { "epoch": 0.59, "learning_rate": 1.4113106932360372e-05, "loss": 0.1018, "step": 12585 }, { "epoch": 0.59, "learning_rate": 1.4110768079333897e-05, "loss": 0.0986, "step": 12590 }, { "epoch": 0.59, "learning_rate": 1.4108429226307422e-05, "loss": 0.0771, "step": 12595 }, { "epoch": 0.59, "learning_rate": 1.4106090373280945e-05, "loss": 0.0733, "step": 12600 }, { "epoch": 0.59, "learning_rate": 1.4103751520254468e-05, "loss": 0.1046, "step": 12605 }, { "epoch": 0.59, "learning_rate": 1.4101412667227991e-05, "loss": 0.0748, "step": 12610 }, { "epoch": 0.59, "learning_rate": 1.4099073814201516e-05, "loss": 0.0737, "step": 12615 }, { "epoch": 0.59, "learning_rate": 1.4096734961175041e-05, "loss": 0.1071, "step": 12620 }, { "epoch": 0.59, "learning_rate": 1.4094396108148564e-05, "loss": 0.11, "step": 12625 }, { "epoch": 0.59, "learning_rate": 1.409205725512209e-05, "loss": 0.0701, "step": 12630 }, { "epoch": 0.59, "learning_rate": 1.4089718402095613e-05, "loss": 0.0884, "step": 12635 }, { "epoch": 0.59, "learning_rate": 1.4087379549069138e-05, "loss": 0.1603, "step": 12640 }, { "epoch": 0.59, "learning_rate": 1.4085040696042663e-05, "loss": 0.1181, "step": 12645 }, { "epoch": 0.59, "learning_rate": 1.4082701843016186e-05, "loss": 0.105, "step": 12650 }, { "epoch": 0.59, "learning_rate": 1.408036298998971e-05, "loss": 0.1308, "step": 12655 }, { "epoch": 0.59, "learning_rate": 1.4078024136963236e-05, "loss": 0.0678, "step": 12660 }, { "epoch": 0.59, "learning_rate": 1.4075685283936757e-05, "loss": 0.1427, "step": 12665 }, { "epoch": 0.59, "learning_rate": 1.4073346430910282e-05, "loss": 0.1513, "step": 12670 }, { "epoch": 0.59, "learning_rate": 1.4071007577883806e-05, "loss": 0.087, "step": 12675 }, { "epoch": 0.59, "learning_rate": 1.406866872485733e-05, "loss": 0.0976, "step": 12680 }, { "epoch": 0.59, "learning_rate": 1.4066329871830855e-05, "loss": 0.0389, "step": 12685 }, { "epoch": 0.59, "learning_rate": 1.4063991018804379e-05, "loss": 0.0843, "step": 12690 }, { "epoch": 0.59, "learning_rate": 1.4061652165777904e-05, "loss": 0.1198, "step": 12695 }, { "epoch": 0.59, "learning_rate": 1.4059313312751429e-05, "loss": 0.1053, "step": 12700 }, { "epoch": 0.59, "learning_rate": 1.4056974459724952e-05, "loss": 0.1005, "step": 12705 }, { "epoch": 0.59, "learning_rate": 1.4054635606698477e-05, "loss": 0.1058, "step": 12710 }, { "epoch": 0.59, "learning_rate": 1.4052296753672002e-05, "loss": 0.1178, "step": 12715 }, { "epoch": 0.59, "learning_rate": 1.4049957900645525e-05, "loss": 0.1111, "step": 12720 }, { "epoch": 0.6, "learning_rate": 1.4047619047619048e-05, "loss": 0.0752, "step": 12725 }, { "epoch": 0.6, "learning_rate": 1.4045280194592572e-05, "loss": 0.1475, "step": 12730 }, { "epoch": 0.6, "learning_rate": 1.4042941341566096e-05, "loss": 0.0444, "step": 12735 }, { "epoch": 0.6, "learning_rate": 1.4040602488539621e-05, "loss": 0.08, "step": 12740 }, { "epoch": 0.6, "learning_rate": 1.4038263635513145e-05, "loss": 0.1943, "step": 12745 }, { "epoch": 0.6, "learning_rate": 1.403592478248667e-05, "loss": 0.1206, "step": 12750 }, { "epoch": 0.6, "learning_rate": 1.4033585929460195e-05, "loss": 0.1151, "step": 12755 }, { "epoch": 0.6, "learning_rate": 1.4031247076433718e-05, "loss": 0.0787, "step": 12760 }, { "epoch": 0.6, "learning_rate": 1.4028908223407243e-05, "loss": 0.1334, "step": 12765 }, { "epoch": 0.6, "learning_rate": 1.4026569370380768e-05, "loss": 0.1264, "step": 12770 }, { "epoch": 0.6, "learning_rate": 1.4024230517354291e-05, "loss": 0.1195, "step": 12775 }, { "epoch": 0.6, "learning_rate": 1.4021891664327816e-05, "loss": 0.138, "step": 12780 }, { "epoch": 0.6, "learning_rate": 1.4019552811301337e-05, "loss": 0.1098, "step": 12785 }, { "epoch": 0.6, "learning_rate": 1.4017213958274862e-05, "loss": 0.1208, "step": 12790 }, { "epoch": 0.6, "learning_rate": 1.4014875105248387e-05, "loss": 0.1064, "step": 12795 }, { "epoch": 0.6, "learning_rate": 1.401253625222191e-05, "loss": 0.0974, "step": 12800 }, { "epoch": 0.6, "learning_rate": 1.4010197399195436e-05, "loss": 0.0873, "step": 12805 }, { "epoch": 0.6, "learning_rate": 1.400785854616896e-05, "loss": 0.1429, "step": 12810 }, { "epoch": 0.6, "learning_rate": 1.4005519693142484e-05, "loss": 0.1463, "step": 12815 }, { "epoch": 0.6, "learning_rate": 1.4003180840116009e-05, "loss": 0.0518, "step": 12820 }, { "epoch": 0.6, "learning_rate": 1.4000841987089534e-05, "loss": 0.0791, "step": 12825 }, { "epoch": 0.6, "learning_rate": 1.3998503134063057e-05, "loss": 0.0923, "step": 12830 }, { "epoch": 0.6, "learning_rate": 1.3996164281036582e-05, "loss": 0.1651, "step": 12835 }, { "epoch": 0.6, "learning_rate": 1.3993825428010105e-05, "loss": 0.1516, "step": 12840 }, { "epoch": 0.6, "learning_rate": 1.3991486574983628e-05, "loss": 0.1212, "step": 12845 }, { "epoch": 0.6, "learning_rate": 1.3989147721957152e-05, "loss": 0.0727, "step": 12850 }, { "epoch": 0.6, "learning_rate": 1.3986808868930677e-05, "loss": 0.123, "step": 12855 }, { "epoch": 0.6, "learning_rate": 1.3984470015904202e-05, "loss": 0.0884, "step": 12860 }, { "epoch": 0.6, "learning_rate": 1.3982131162877725e-05, "loss": 0.1279, "step": 12865 }, { "epoch": 0.6, "learning_rate": 1.397979230985125e-05, "loss": 0.1263, "step": 12870 }, { "epoch": 0.6, "learning_rate": 1.3977453456824775e-05, "loss": 0.0965, "step": 12875 }, { "epoch": 0.6, "learning_rate": 1.3975114603798298e-05, "loss": 0.0621, "step": 12880 }, { "epoch": 0.6, "learning_rate": 1.3972775750771823e-05, "loss": 0.1048, "step": 12885 }, { "epoch": 0.6, "learning_rate": 1.3970436897745348e-05, "loss": 0.1436, "step": 12890 }, { "epoch": 0.6, "learning_rate": 1.3968098044718871e-05, "loss": 0.0545, "step": 12895 }, { "epoch": 0.6, "learning_rate": 1.3965759191692396e-05, "loss": 0.1045, "step": 12900 }, { "epoch": 0.6, "learning_rate": 1.3963420338665918e-05, "loss": 0.1168, "step": 12905 }, { "epoch": 0.6, "learning_rate": 1.3961081485639443e-05, "loss": 0.0567, "step": 12910 }, { "epoch": 0.6, "learning_rate": 1.3958742632612968e-05, "loss": 0.1001, "step": 12915 }, { "epoch": 0.6, "learning_rate": 1.3956403779586491e-05, "loss": 0.1489, "step": 12920 }, { "epoch": 0.6, "learning_rate": 1.3954064926560016e-05, "loss": 0.1015, "step": 12925 }, { "epoch": 0.6, "learning_rate": 1.395172607353354e-05, "loss": 0.0885, "step": 12930 }, { "epoch": 0.61, "learning_rate": 1.3949387220507064e-05, "loss": 0.1491, "step": 12935 }, { "epoch": 0.61, "learning_rate": 1.3947048367480589e-05, "loss": 0.0787, "step": 12940 }, { "epoch": 0.61, "learning_rate": 1.3944709514454114e-05, "loss": 0.0888, "step": 12945 }, { "epoch": 0.61, "learning_rate": 1.3942370661427637e-05, "loss": 0.1261, "step": 12950 }, { "epoch": 0.61, "learning_rate": 1.3940031808401162e-05, "loss": 0.0784, "step": 12955 }, { "epoch": 0.61, "learning_rate": 1.3937692955374687e-05, "loss": 0.1347, "step": 12960 }, { "epoch": 0.61, "learning_rate": 1.3935354102348209e-05, "loss": 0.0789, "step": 12965 }, { "epoch": 0.61, "learning_rate": 1.3933015249321734e-05, "loss": 0.1259, "step": 12970 }, { "epoch": 0.61, "learning_rate": 1.3930676396295257e-05, "loss": 0.0334, "step": 12975 }, { "epoch": 0.61, "learning_rate": 1.3928337543268782e-05, "loss": 0.0496, "step": 12980 }, { "epoch": 0.61, "learning_rate": 1.3925998690242307e-05, "loss": 0.0705, "step": 12985 }, { "epoch": 0.61, "learning_rate": 1.392365983721583e-05, "loss": 0.1079, "step": 12990 }, { "epoch": 0.61, "learning_rate": 1.3921320984189355e-05, "loss": 0.1645, "step": 12995 }, { "epoch": 0.61, "learning_rate": 1.391898213116288e-05, "loss": 0.1334, "step": 13000 }, { "epoch": 0.61, "learning_rate": 1.3916643278136403e-05, "loss": 0.0936, "step": 13005 }, { "epoch": 0.61, "learning_rate": 1.3914304425109928e-05, "loss": 0.0546, "step": 13010 }, { "epoch": 0.61, "learning_rate": 1.3911965572083453e-05, "loss": 0.1013, "step": 13015 }, { "epoch": 0.61, "learning_rate": 1.3909626719056976e-05, "loss": 0.1446, "step": 13020 }, { "epoch": 0.61, "learning_rate": 1.39072878660305e-05, "loss": 0.0827, "step": 13025 }, { "epoch": 0.61, "learning_rate": 1.3904949013004023e-05, "loss": 0.0952, "step": 13030 }, { "epoch": 0.61, "learning_rate": 1.3902610159977548e-05, "loss": 0.1814, "step": 13035 }, { "epoch": 0.61, "learning_rate": 1.3900271306951073e-05, "loss": 0.1492, "step": 13040 }, { "epoch": 0.61, "learning_rate": 1.3897932453924596e-05, "loss": 0.0996, "step": 13045 }, { "epoch": 0.61, "learning_rate": 1.3895593600898121e-05, "loss": 0.1994, "step": 13050 }, { "epoch": 0.61, "learning_rate": 1.3893254747871646e-05, "loss": 0.116, "step": 13055 }, { "epoch": 0.61, "learning_rate": 1.389091589484517e-05, "loss": 0.0864, "step": 13060 }, { "epoch": 0.61, "learning_rate": 1.3888577041818694e-05, "loss": 0.086, "step": 13065 }, { "epoch": 0.61, "learning_rate": 1.3886238188792217e-05, "loss": 0.0811, "step": 13070 }, { "epoch": 0.61, "learning_rate": 1.3883899335765742e-05, "loss": 0.1079, "step": 13075 }, { "epoch": 0.61, "learning_rate": 1.3881560482739264e-05, "loss": 0.1336, "step": 13080 }, { "epoch": 0.61, "learning_rate": 1.3879221629712789e-05, "loss": 0.0922, "step": 13085 }, { "epoch": 0.61, "learning_rate": 1.3876882776686314e-05, "loss": 0.092, "step": 13090 }, { "epoch": 0.61, "learning_rate": 1.3874543923659837e-05, "loss": 0.0746, "step": 13095 }, { "epoch": 0.61, "learning_rate": 1.3872205070633362e-05, "loss": 0.0864, "step": 13100 }, { "epoch": 0.61, "learning_rate": 1.3869866217606887e-05, "loss": 0.0791, "step": 13105 }, { "epoch": 0.61, "learning_rate": 1.386752736458041e-05, "loss": 0.1383, "step": 13110 }, { "epoch": 0.61, "learning_rate": 1.3865188511553935e-05, "loss": 0.053, "step": 13115 }, { "epoch": 0.61, "learning_rate": 1.386284965852746e-05, "loss": 0.0774, "step": 13120 }, { "epoch": 0.61, "learning_rate": 1.3860510805500983e-05, "loss": 0.2083, "step": 13125 }, { "epoch": 0.61, "learning_rate": 1.3858171952474508e-05, "loss": 0.0972, "step": 13130 }, { "epoch": 0.61, "learning_rate": 1.3855833099448033e-05, "loss": 0.0937, "step": 13135 }, { "epoch": 0.61, "learning_rate": 1.3853494246421555e-05, "loss": 0.095, "step": 13140 }, { "epoch": 0.61, "learning_rate": 1.385115539339508e-05, "loss": 0.0698, "step": 13145 }, { "epoch": 0.62, "learning_rate": 1.3848816540368603e-05, "loss": 0.0855, "step": 13150 }, { "epoch": 0.62, "learning_rate": 1.3846477687342128e-05, "loss": 0.1071, "step": 13155 }, { "epoch": 0.62, "learning_rate": 1.3844138834315653e-05, "loss": 0.1164, "step": 13160 }, { "epoch": 0.62, "learning_rate": 1.3841799981289176e-05, "loss": 0.0835, "step": 13165 }, { "epoch": 0.62, "learning_rate": 1.3839461128262701e-05, "loss": 0.0385, "step": 13170 }, { "epoch": 0.62, "learning_rate": 1.3837122275236226e-05, "loss": 0.0452, "step": 13175 }, { "epoch": 0.62, "learning_rate": 1.383478342220975e-05, "loss": 0.1313, "step": 13180 }, { "epoch": 0.62, "learning_rate": 1.3832444569183274e-05, "loss": 0.1223, "step": 13185 }, { "epoch": 0.62, "learning_rate": 1.38301057161568e-05, "loss": 0.1508, "step": 13190 }, { "epoch": 0.62, "learning_rate": 1.3827766863130323e-05, "loss": 0.0573, "step": 13195 }, { "epoch": 0.62, "learning_rate": 1.3825428010103846e-05, "loss": 0.0654, "step": 13200 }, { "epoch": 0.62, "learning_rate": 1.3823089157077369e-05, "loss": 0.0736, "step": 13205 }, { "epoch": 0.62, "learning_rate": 1.3820750304050894e-05, "loss": 0.0575, "step": 13210 }, { "epoch": 0.62, "learning_rate": 1.3818411451024419e-05, "loss": 0.0669, "step": 13215 }, { "epoch": 0.62, "learning_rate": 1.3816072597997942e-05, "loss": 0.0834, "step": 13220 }, { "epoch": 0.62, "learning_rate": 1.3813733744971467e-05, "loss": 0.1713, "step": 13225 }, { "epoch": 0.62, "learning_rate": 1.3811394891944992e-05, "loss": 0.1124, "step": 13230 }, { "epoch": 0.62, "learning_rate": 1.3809056038918515e-05, "loss": 0.1282, "step": 13235 }, { "epoch": 0.62, "learning_rate": 1.380671718589204e-05, "loss": 0.112, "step": 13240 }, { "epoch": 0.62, "learning_rate": 1.3804378332865565e-05, "loss": 0.1564, "step": 13245 }, { "epoch": 0.62, "learning_rate": 1.3802039479839089e-05, "loss": 0.087, "step": 13250 }, { "epoch": 0.62, "learning_rate": 1.3799700626812613e-05, "loss": 0.0865, "step": 13255 }, { "epoch": 0.62, "learning_rate": 1.3797361773786135e-05, "loss": 0.0958, "step": 13260 }, { "epoch": 0.62, "learning_rate": 1.379502292075966e-05, "loss": 0.1286, "step": 13265 }, { "epoch": 0.62, "learning_rate": 1.3792684067733185e-05, "loss": 0.1329, "step": 13270 }, { "epoch": 0.62, "learning_rate": 1.3790345214706708e-05, "loss": 0.0646, "step": 13275 }, { "epoch": 0.62, "learning_rate": 1.3788006361680233e-05, "loss": 0.0878, "step": 13280 }, { "epoch": 0.62, "learning_rate": 1.3785667508653756e-05, "loss": 0.0805, "step": 13285 }, { "epoch": 0.62, "learning_rate": 1.3783328655627281e-05, "loss": 0.1333, "step": 13290 }, { "epoch": 0.62, "learning_rate": 1.3780989802600806e-05, "loss": 0.1244, "step": 13295 }, { "epoch": 0.62, "learning_rate": 1.377865094957433e-05, "loss": 0.054, "step": 13300 }, { "epoch": 0.62, "learning_rate": 1.3776312096547855e-05, "loss": 0.1203, "step": 13305 }, { "epoch": 0.62, "learning_rate": 1.377397324352138e-05, "loss": 0.0905, "step": 13310 }, { "epoch": 0.62, "learning_rate": 1.3771634390494903e-05, "loss": 0.1187, "step": 13315 }, { "epoch": 0.62, "learning_rate": 1.3769295537468426e-05, "loss": 0.1042, "step": 13320 }, { "epoch": 0.62, "learning_rate": 1.376695668444195e-05, "loss": 0.0878, "step": 13325 }, { "epoch": 0.62, "learning_rate": 1.3764617831415474e-05, "loss": 0.0915, "step": 13330 }, { "epoch": 0.62, "learning_rate": 1.3762278978388999e-05, "loss": 0.0847, "step": 13335 }, { "epoch": 0.62, "learning_rate": 1.3759940125362522e-05, "loss": 0.0455, "step": 13340 }, { "epoch": 0.62, "learning_rate": 1.3757601272336047e-05, "loss": 0.1412, "step": 13345 }, { "epoch": 0.62, "learning_rate": 1.3755262419309572e-05, "loss": 0.1106, "step": 13350 }, { "epoch": 0.62, "learning_rate": 1.3752923566283096e-05, "loss": 0.12, "step": 13355 }, { "epoch": 0.62, "learning_rate": 1.375058471325662e-05, "loss": 0.1108, "step": 13360 }, { "epoch": 0.63, "learning_rate": 1.3748245860230145e-05, "loss": 0.0697, "step": 13365 }, { "epoch": 0.63, "learning_rate": 1.3745907007203669e-05, "loss": 0.1095, "step": 13370 }, { "epoch": 0.63, "learning_rate": 1.3743568154177194e-05, "loss": 0.0694, "step": 13375 }, { "epoch": 0.63, "learning_rate": 1.3741229301150715e-05, "loss": 0.0996, "step": 13380 }, { "epoch": 0.63, "learning_rate": 1.373889044812424e-05, "loss": 0.1151, "step": 13385 }, { "epoch": 0.63, "learning_rate": 1.3736551595097765e-05, "loss": 0.0769, "step": 13390 }, { "epoch": 0.63, "learning_rate": 1.3734212742071288e-05, "loss": 0.1024, "step": 13395 }, { "epoch": 0.63, "learning_rate": 1.3731873889044813e-05, "loss": 0.071, "step": 13400 }, { "epoch": 0.63, "learning_rate": 1.3729535036018338e-05, "loss": 0.1451, "step": 13405 }, { "epoch": 0.63, "learning_rate": 1.3727196182991862e-05, "loss": 0.1256, "step": 13410 }, { "epoch": 0.63, "learning_rate": 1.3724857329965387e-05, "loss": 0.0801, "step": 13415 }, { "epoch": 0.63, "learning_rate": 1.3722518476938911e-05, "loss": 0.0966, "step": 13420 }, { "epoch": 0.63, "learning_rate": 1.3720179623912435e-05, "loss": 0.1461, "step": 13425 }, { "epoch": 0.63, "learning_rate": 1.371784077088596e-05, "loss": 0.1268, "step": 13430 }, { "epoch": 0.63, "learning_rate": 1.3715501917859485e-05, "loss": 0.1011, "step": 13435 }, { "epoch": 0.63, "learning_rate": 1.3713163064833006e-05, "loss": 0.0785, "step": 13440 }, { "epoch": 0.63, "learning_rate": 1.3710824211806531e-05, "loss": 0.1036, "step": 13445 }, { "epoch": 0.63, "learning_rate": 1.3708485358780054e-05, "loss": 0.1055, "step": 13450 }, { "epoch": 0.63, "learning_rate": 1.370614650575358e-05, "loss": 0.0863, "step": 13455 }, { "epoch": 0.63, "learning_rate": 1.3703807652727104e-05, "loss": 0.1314, "step": 13460 }, { "epoch": 0.63, "learning_rate": 1.3701468799700628e-05, "loss": 0.1051, "step": 13465 }, { "epoch": 0.63, "learning_rate": 1.3699129946674153e-05, "loss": 0.0889, "step": 13470 }, { "epoch": 0.63, "learning_rate": 1.3696791093647677e-05, "loss": 0.0807, "step": 13475 }, { "epoch": 0.63, "learning_rate": 1.36944522406212e-05, "loss": 0.1037, "step": 13480 }, { "epoch": 0.63, "learning_rate": 1.3692113387594726e-05, "loss": 0.1404, "step": 13485 }, { "epoch": 0.63, "learning_rate": 1.368977453456825e-05, "loss": 0.0836, "step": 13490 }, { "epoch": 0.63, "learning_rate": 1.3687435681541772e-05, "loss": 0.0501, "step": 13495 }, { "epoch": 0.63, "learning_rate": 1.3685096828515297e-05, "loss": 0.0806, "step": 13500 }, { "epoch": 0.63, "learning_rate": 1.368275797548882e-05, "loss": 0.1016, "step": 13505 }, { "epoch": 0.63, "learning_rate": 1.3680419122462345e-05, "loss": 0.0826, "step": 13510 }, { "epoch": 0.63, "learning_rate": 1.3678080269435869e-05, "loss": 0.0943, "step": 13515 }, { "epoch": 0.63, "learning_rate": 1.3675741416409394e-05, "loss": 0.0509, "step": 13520 }, { "epoch": 0.63, "learning_rate": 1.3673402563382919e-05, "loss": 0.1476, "step": 13525 }, { "epoch": 0.63, "learning_rate": 1.3671063710356442e-05, "loss": 0.0659, "step": 13530 }, { "epoch": 0.63, "learning_rate": 1.3668724857329967e-05, "loss": 0.075, "step": 13535 }, { "epoch": 0.63, "learning_rate": 1.3666386004303492e-05, "loss": 0.0943, "step": 13540 }, { "epoch": 0.63, "learning_rate": 1.3664047151277015e-05, "loss": 0.1037, "step": 13545 }, { "epoch": 0.63, "learning_rate": 1.366170829825054e-05, "loss": 0.0791, "step": 13550 }, { "epoch": 0.63, "learning_rate": 1.3659369445224061e-05, "loss": 0.0944, "step": 13555 }, { "epoch": 0.63, "learning_rate": 1.3657030592197586e-05, "loss": 0.1556, "step": 13560 }, { "epoch": 0.63, "learning_rate": 1.3654691739171111e-05, "loss": 0.065, "step": 13565 }, { "epoch": 0.63, "learning_rate": 1.3652352886144635e-05, "loss": 0.093, "step": 13570 }, { "epoch": 0.63, "learning_rate": 1.365001403311816e-05, "loss": 0.0982, "step": 13575 }, { "epoch": 0.64, "learning_rate": 1.3647675180091685e-05, "loss": 0.1087, "step": 13580 }, { "epoch": 0.64, "learning_rate": 1.3645336327065208e-05, "loss": 0.1367, "step": 13585 }, { "epoch": 0.64, "learning_rate": 1.3642997474038733e-05, "loss": 0.1176, "step": 13590 }, { "epoch": 0.64, "learning_rate": 1.3640658621012258e-05, "loss": 0.1269, "step": 13595 }, { "epoch": 0.64, "learning_rate": 1.3638319767985781e-05, "loss": 0.0661, "step": 13600 }, { "epoch": 0.64, "learning_rate": 1.3635980914959306e-05, "loss": 0.1806, "step": 13605 }, { "epoch": 0.64, "learning_rate": 1.363364206193283e-05, "loss": 0.0577, "step": 13610 }, { "epoch": 0.64, "learning_rate": 1.3631303208906352e-05, "loss": 0.0634, "step": 13615 }, { "epoch": 0.64, "learning_rate": 1.3628964355879877e-05, "loss": 0.1057, "step": 13620 }, { "epoch": 0.64, "learning_rate": 1.36266255028534e-05, "loss": 0.1215, "step": 13625 }, { "epoch": 0.64, "learning_rate": 1.3624286649826926e-05, "loss": 0.0952, "step": 13630 }, { "epoch": 0.64, "learning_rate": 1.362194779680045e-05, "loss": 0.0987, "step": 13635 }, { "epoch": 0.64, "learning_rate": 1.3619608943773974e-05, "loss": 0.0811, "step": 13640 }, { "epoch": 0.64, "learning_rate": 1.3617270090747499e-05, "loss": 0.1348, "step": 13645 }, { "epoch": 0.64, "learning_rate": 1.3614931237721024e-05, "loss": 0.067, "step": 13650 }, { "epoch": 0.64, "learning_rate": 1.3612592384694547e-05, "loss": 0.0767, "step": 13655 }, { "epoch": 0.64, "learning_rate": 1.3610253531668072e-05, "loss": 0.0932, "step": 13660 }, { "epoch": 0.64, "learning_rate": 1.3607914678641597e-05, "loss": 0.1198, "step": 13665 }, { "epoch": 0.64, "learning_rate": 1.360557582561512e-05, "loss": 0.1162, "step": 13670 }, { "epoch": 0.64, "learning_rate": 1.3603236972588643e-05, "loss": 0.0673, "step": 13675 }, { "epoch": 0.64, "learning_rate": 1.3600898119562167e-05, "loss": 0.0686, "step": 13680 }, { "epoch": 0.64, "learning_rate": 1.3598559266535692e-05, "loss": 0.1469, "step": 13685 }, { "epoch": 0.64, "learning_rate": 1.3596220413509216e-05, "loss": 0.0806, "step": 13690 }, { "epoch": 0.64, "learning_rate": 1.359388156048274e-05, "loss": 0.1128, "step": 13695 }, { "epoch": 0.64, "learning_rate": 1.3591542707456265e-05, "loss": 0.0802, "step": 13700 }, { "epoch": 0.64, "learning_rate": 1.358920385442979e-05, "loss": 0.1264, "step": 13705 }, { "epoch": 0.64, "learning_rate": 1.3586865001403313e-05, "loss": 0.0474, "step": 13710 }, { "epoch": 0.64, "learning_rate": 1.3584526148376838e-05, "loss": 0.1173, "step": 13715 }, { "epoch": 0.64, "learning_rate": 1.3582187295350361e-05, "loss": 0.1177, "step": 13720 }, { "epoch": 0.64, "learning_rate": 1.3579848442323886e-05, "loss": 0.0725, "step": 13725 }, { "epoch": 0.64, "learning_rate": 1.3577509589297411e-05, "loss": 0.1467, "step": 13730 }, { "epoch": 0.64, "learning_rate": 1.3575170736270933e-05, "loss": 0.0603, "step": 13735 }, { "epoch": 0.64, "learning_rate": 1.3572831883244458e-05, "loss": 0.0663, "step": 13740 }, { "epoch": 0.64, "learning_rate": 1.357049303021798e-05, "loss": 0.1005, "step": 13745 }, { "epoch": 0.64, "learning_rate": 1.3568154177191506e-05, "loss": 0.0763, "step": 13750 }, { "epoch": 0.64, "learning_rate": 1.356581532416503e-05, "loss": 0.073, "step": 13755 }, { "epoch": 0.64, "learning_rate": 1.3563476471138554e-05, "loss": 0.0712, "step": 13760 }, { "epoch": 0.64, "learning_rate": 1.3561137618112079e-05, "loss": 0.0689, "step": 13765 }, { "epoch": 0.64, "learning_rate": 1.3558798765085604e-05, "loss": 0.1863, "step": 13770 }, { "epoch": 0.64, "learning_rate": 1.3556459912059127e-05, "loss": 0.1001, "step": 13775 }, { "epoch": 0.64, "learning_rate": 1.3554121059032652e-05, "loss": 0.1111, "step": 13780 }, { "epoch": 0.64, "learning_rate": 1.3551782206006177e-05, "loss": 0.0353, "step": 13785 }, { "epoch": 0.65, "learning_rate": 1.35494433529797e-05, "loss": 0.0723, "step": 13790 }, { "epoch": 0.65, "learning_rate": 1.3547104499953224e-05, "loss": 0.1414, "step": 13795 }, { "epoch": 0.65, "learning_rate": 1.3544765646926747e-05, "loss": 0.0843, "step": 13800 }, { "epoch": 0.65, "learning_rate": 1.3542426793900272e-05, "loss": 0.1425, "step": 13805 }, { "epoch": 0.65, "learning_rate": 1.3540087940873797e-05, "loss": 0.0692, "step": 13810 }, { "epoch": 0.65, "learning_rate": 1.353774908784732e-05, "loss": 0.1083, "step": 13815 }, { "epoch": 0.65, "learning_rate": 1.3535410234820845e-05, "loss": 0.0669, "step": 13820 }, { "epoch": 0.65, "learning_rate": 1.353307138179437e-05, "loss": 0.1085, "step": 13825 }, { "epoch": 0.65, "learning_rate": 1.3530732528767893e-05, "loss": 0.1647, "step": 13830 }, { "epoch": 0.65, "learning_rate": 1.3528393675741418e-05, "loss": 0.0867, "step": 13835 }, { "epoch": 0.65, "learning_rate": 1.3526054822714943e-05, "loss": 0.1352, "step": 13840 }, { "epoch": 0.65, "learning_rate": 1.3523715969688466e-05, "loss": 0.0879, "step": 13845 }, { "epoch": 0.65, "learning_rate": 1.3521377116661991e-05, "loss": 0.1144, "step": 13850 }, { "epoch": 0.65, "learning_rate": 1.3519038263635513e-05, "loss": 0.1418, "step": 13855 }, { "epoch": 0.65, "learning_rate": 1.3516699410609038e-05, "loss": 0.0534, "step": 13860 }, { "epoch": 0.65, "learning_rate": 1.3514360557582563e-05, "loss": 0.0774, "step": 13865 }, { "epoch": 0.65, "learning_rate": 1.3512021704556086e-05, "loss": 0.1109, "step": 13870 }, { "epoch": 0.65, "learning_rate": 1.3509682851529611e-05, "loss": 0.1149, "step": 13875 }, { "epoch": 0.65, "learning_rate": 1.3507343998503136e-05, "loss": 0.1119, "step": 13880 }, { "epoch": 0.65, "learning_rate": 1.3505005145476659e-05, "loss": 0.1585, "step": 13885 }, { "epoch": 0.65, "learning_rate": 1.3502666292450184e-05, "loss": 0.1812, "step": 13890 }, { "epoch": 0.65, "learning_rate": 1.3500327439423709e-05, "loss": 0.0414, "step": 13895 }, { "epoch": 0.65, "learning_rate": 1.3497988586397232e-05, "loss": 0.1353, "step": 13900 }, { "epoch": 0.65, "learning_rate": 1.3495649733370757e-05, "loss": 0.0872, "step": 13905 }, { "epoch": 0.65, "learning_rate": 1.3493310880344279e-05, "loss": 0.1133, "step": 13910 }, { "epoch": 0.65, "learning_rate": 1.3490972027317804e-05, "loss": 0.077, "step": 13915 }, { "epoch": 0.65, "learning_rate": 1.3488633174291329e-05, "loss": 0.1049, "step": 13920 }, { "epoch": 0.65, "learning_rate": 1.3486294321264852e-05, "loss": 0.0991, "step": 13925 }, { "epoch": 0.65, "learning_rate": 1.3483955468238377e-05, "loss": 0.0895, "step": 13930 }, { "epoch": 0.65, "learning_rate": 1.3481616615211902e-05, "loss": 0.0879, "step": 13935 }, { "epoch": 0.65, "learning_rate": 1.3479277762185425e-05, "loss": 0.0742, "step": 13940 }, { "epoch": 0.65, "learning_rate": 1.347693890915895e-05, "loss": 0.0929, "step": 13945 }, { "epoch": 0.65, "learning_rate": 1.3474600056132473e-05, "loss": 0.1608, "step": 13950 }, { "epoch": 0.65, "learning_rate": 1.3472261203105998e-05, "loss": 0.1105, "step": 13955 }, { "epoch": 0.65, "learning_rate": 1.3469922350079523e-05, "loss": 0.0699, "step": 13960 }, { "epoch": 0.65, "learning_rate": 1.3467583497053046e-05, "loss": 0.0553, "step": 13965 }, { "epoch": 0.65, "learning_rate": 1.346524464402657e-05, "loss": 0.1278, "step": 13970 }, { "epoch": 0.65, "learning_rate": 1.3462905791000093e-05, "loss": 0.0822, "step": 13975 }, { "epoch": 0.65, "learning_rate": 1.3460566937973618e-05, "loss": 0.0697, "step": 13980 }, { "epoch": 0.65, "learning_rate": 1.3458228084947143e-05, "loss": 0.0873, "step": 13985 }, { "epoch": 0.65, "learning_rate": 1.3455889231920666e-05, "loss": 0.1223, "step": 13990 }, { "epoch": 0.65, "learning_rate": 1.3453550378894191e-05, "loss": 0.1095, "step": 13995 }, { "epoch": 0.65, "learning_rate": 1.3451211525867716e-05, "loss": 0.0717, "step": 14000 }, { "epoch": 0.66, "learning_rate": 1.344887267284124e-05, "loss": 0.0846, "step": 14005 }, { "epoch": 0.66, "learning_rate": 1.3446533819814764e-05, "loss": 0.1046, "step": 14010 }, { "epoch": 0.66, "learning_rate": 1.344419496678829e-05, "loss": 0.0896, "step": 14015 }, { "epoch": 0.66, "learning_rate": 1.3441856113761812e-05, "loss": 0.1101, "step": 14020 }, { "epoch": 0.66, "learning_rate": 1.3439517260735337e-05, "loss": 0.0837, "step": 14025 }, { "epoch": 0.66, "learning_rate": 1.3437178407708859e-05, "loss": 0.0626, "step": 14030 }, { "epoch": 0.66, "learning_rate": 1.3434839554682384e-05, "loss": 0.0875, "step": 14035 }, { "epoch": 0.66, "learning_rate": 1.3432500701655909e-05, "loss": 0.0967, "step": 14040 }, { "epoch": 0.66, "learning_rate": 1.3430161848629432e-05, "loss": 0.0397, "step": 14045 }, { "epoch": 0.66, "learning_rate": 1.3427822995602957e-05, "loss": 0.0632, "step": 14050 }, { "epoch": 0.66, "learning_rate": 1.3425484142576482e-05, "loss": 0.095, "step": 14055 }, { "epoch": 0.66, "learning_rate": 1.3423145289550005e-05, "loss": 0.1119, "step": 14060 }, { "epoch": 0.66, "learning_rate": 1.342080643652353e-05, "loss": 0.0738, "step": 14065 }, { "epoch": 0.66, "learning_rate": 1.3418467583497055e-05, "loss": 0.0651, "step": 14070 }, { "epoch": 0.66, "learning_rate": 1.3416128730470578e-05, "loss": 0.0518, "step": 14075 }, { "epoch": 0.66, "learning_rate": 1.3413789877444103e-05, "loss": 0.0905, "step": 14080 }, { "epoch": 0.66, "learning_rate": 1.3411451024417628e-05, "loss": 0.0822, "step": 14085 }, { "epoch": 0.66, "learning_rate": 1.340911217139115e-05, "loss": 0.049, "step": 14090 }, { "epoch": 0.66, "learning_rate": 1.3406773318364675e-05, "loss": 0.1632, "step": 14095 }, { "epoch": 0.66, "learning_rate": 1.3404434465338198e-05, "loss": 0.098, "step": 14100 }, { "epoch": 0.66, "learning_rate": 1.3402095612311723e-05, "loss": 0.0841, "step": 14105 }, { "epoch": 0.66, "learning_rate": 1.3399756759285248e-05, "loss": 0.085, "step": 14110 }, { "epoch": 0.66, "learning_rate": 1.3397417906258771e-05, "loss": 0.0674, "step": 14115 }, { "epoch": 0.66, "learning_rate": 1.3395079053232296e-05, "loss": 0.1063, "step": 14120 }, { "epoch": 0.66, "learning_rate": 1.3392740200205821e-05, "loss": 0.0817, "step": 14125 }, { "epoch": 0.66, "learning_rate": 1.3390401347179344e-05, "loss": 0.1501, "step": 14130 }, { "epoch": 0.66, "learning_rate": 1.338806249415287e-05, "loss": 0.148, "step": 14135 }, { "epoch": 0.66, "learning_rate": 1.3385723641126394e-05, "loss": 0.1359, "step": 14140 }, { "epoch": 0.66, "learning_rate": 1.3383384788099918e-05, "loss": 0.0939, "step": 14145 }, { "epoch": 0.66, "learning_rate": 1.3381045935073441e-05, "loss": 0.0813, "step": 14150 }, { "epoch": 0.66, "learning_rate": 1.3378707082046964e-05, "loss": 0.0677, "step": 14155 }, { "epoch": 0.66, "learning_rate": 1.3376368229020489e-05, "loss": 0.0873, "step": 14160 }, { "epoch": 0.66, "learning_rate": 1.3374029375994012e-05, "loss": 0.0782, "step": 14165 }, { "epoch": 0.66, "learning_rate": 1.3371690522967537e-05, "loss": 0.1321, "step": 14170 }, { "epoch": 0.66, "learning_rate": 1.3369351669941062e-05, "loss": 0.0475, "step": 14175 }, { "epoch": 0.66, "learning_rate": 1.3367012816914586e-05, "loss": 0.1603, "step": 14180 }, { "epoch": 0.66, "learning_rate": 1.336467396388811e-05, "loss": 0.1105, "step": 14185 }, { "epoch": 0.66, "learning_rate": 1.3362335110861635e-05, "loss": 0.1259, "step": 14190 }, { "epoch": 0.66, "learning_rate": 1.3359996257835159e-05, "loss": 0.1077, "step": 14195 }, { "epoch": 0.66, "learning_rate": 1.3357657404808684e-05, "loss": 0.127, "step": 14200 }, { "epoch": 0.66, "learning_rate": 1.3355318551782209e-05, "loss": 0.0458, "step": 14205 }, { "epoch": 0.66, "learning_rate": 1.335297969875573e-05, "loss": 0.1041, "step": 14210 }, { "epoch": 0.66, "learning_rate": 1.3350640845729255e-05, "loss": 0.0352, "step": 14215 }, { "epoch": 0.67, "learning_rate": 1.3348301992702778e-05, "loss": 0.0691, "step": 14220 }, { "epoch": 0.67, "learning_rate": 1.3345963139676303e-05, "loss": 0.0826, "step": 14225 }, { "epoch": 0.67, "learning_rate": 1.3343624286649828e-05, "loss": 0.0786, "step": 14230 }, { "epoch": 0.67, "learning_rate": 1.3341285433623352e-05, "loss": 0.1278, "step": 14235 }, { "epoch": 0.67, "learning_rate": 1.3338946580596876e-05, "loss": 0.1114, "step": 14240 }, { "epoch": 0.67, "learning_rate": 1.3336607727570401e-05, "loss": 0.0788, "step": 14245 }, { "epoch": 0.67, "learning_rate": 1.3334268874543925e-05, "loss": 0.1448, "step": 14250 }, { "epoch": 0.67, "learning_rate": 1.333193002151745e-05, "loss": 0.0968, "step": 14255 }, { "epoch": 0.67, "learning_rate": 1.3329591168490975e-05, "loss": 0.0659, "step": 14260 }, { "epoch": 0.67, "learning_rate": 1.3327252315464498e-05, "loss": 0.1104, "step": 14265 }, { "epoch": 0.67, "learning_rate": 1.3324913462438021e-05, "loss": 0.1511, "step": 14270 }, { "epoch": 0.67, "learning_rate": 1.3322574609411544e-05, "loss": 0.0528, "step": 14275 }, { "epoch": 0.67, "learning_rate": 1.332023575638507e-05, "loss": 0.0848, "step": 14280 }, { "epoch": 0.67, "learning_rate": 1.3317896903358594e-05, "loss": 0.0679, "step": 14285 }, { "epoch": 0.67, "learning_rate": 1.3315558050332118e-05, "loss": 0.113, "step": 14290 }, { "epoch": 0.67, "learning_rate": 1.3313219197305642e-05, "loss": 0.0946, "step": 14295 }, { "epoch": 0.67, "learning_rate": 1.3310880344279167e-05, "loss": 0.1186, "step": 14300 }, { "epoch": 0.67, "learning_rate": 1.330854149125269e-05, "loss": 0.0792, "step": 14305 }, { "epoch": 0.67, "learning_rate": 1.3306202638226216e-05, "loss": 0.0853, "step": 14310 }, { "epoch": 0.67, "learning_rate": 1.330386378519974e-05, "loss": 0.1394, "step": 14315 }, { "epoch": 0.67, "learning_rate": 1.3301524932173264e-05, "loss": 0.0858, "step": 14320 }, { "epoch": 0.67, "learning_rate": 1.3299186079146789e-05, "loss": 0.0579, "step": 14325 }, { "epoch": 0.67, "learning_rate": 1.329684722612031e-05, "loss": 0.0737, "step": 14330 }, { "epoch": 0.67, "learning_rate": 1.3294508373093835e-05, "loss": 0.1794, "step": 14335 }, { "epoch": 0.67, "learning_rate": 1.329216952006736e-05, "loss": 0.0797, "step": 14340 }, { "epoch": 0.67, "learning_rate": 1.3289830667040884e-05, "loss": 0.0776, "step": 14345 }, { "epoch": 0.67, "learning_rate": 1.3287491814014408e-05, "loss": 0.0938, "step": 14350 }, { "epoch": 0.67, "learning_rate": 1.3285152960987933e-05, "loss": 0.0657, "step": 14355 }, { "epoch": 0.67, "learning_rate": 1.3282814107961457e-05, "loss": 0.0754, "step": 14360 }, { "epoch": 0.67, "learning_rate": 1.3280475254934982e-05, "loss": 0.0839, "step": 14365 }, { "epoch": 0.67, "learning_rate": 1.3278136401908507e-05, "loss": 0.1484, "step": 14370 }, { "epoch": 0.67, "learning_rate": 1.327579754888203e-05, "loss": 0.1003, "step": 14375 }, { "epoch": 0.67, "learning_rate": 1.3273458695855555e-05, "loss": 0.1202, "step": 14380 }, { "epoch": 0.67, "learning_rate": 1.3271119842829076e-05, "loss": 0.149, "step": 14385 }, { "epoch": 0.67, "learning_rate": 1.3268780989802601e-05, "loss": 0.1023, "step": 14390 }, { "epoch": 0.67, "learning_rate": 1.3266442136776125e-05, "loss": 0.1575, "step": 14395 }, { "epoch": 0.67, "learning_rate": 1.326410328374965e-05, "loss": 0.0587, "step": 14400 }, { "epoch": 0.67, "learning_rate": 1.3261764430723174e-05, "loss": 0.1092, "step": 14405 }, { "epoch": 0.67, "learning_rate": 1.3259425577696698e-05, "loss": 0.1309, "step": 14410 }, { "epoch": 0.67, "learning_rate": 1.3257086724670223e-05, "loss": 0.0848, "step": 14415 }, { "epoch": 0.67, "learning_rate": 1.3254747871643748e-05, "loss": 0.0629, "step": 14420 }, { "epoch": 0.67, "learning_rate": 1.3252409018617271e-05, "loss": 0.0717, "step": 14425 }, { "epoch": 0.67, "learning_rate": 1.3250070165590796e-05, "loss": 0.0859, "step": 14430 }, { "epoch": 0.68, "learning_rate": 1.324773131256432e-05, "loss": 0.0946, "step": 14435 }, { "epoch": 0.68, "learning_rate": 1.3245392459537844e-05, "loss": 0.1149, "step": 14440 }, { "epoch": 0.68, "learning_rate": 1.3243053606511367e-05, "loss": 0.0818, "step": 14445 }, { "epoch": 0.68, "learning_rate": 1.324071475348489e-05, "loss": 0.155, "step": 14450 }, { "epoch": 0.68, "learning_rate": 1.3238375900458415e-05, "loss": 0.1674, "step": 14455 }, { "epoch": 0.68, "learning_rate": 1.323603704743194e-05, "loss": 0.1018, "step": 14460 }, { "epoch": 0.68, "learning_rate": 1.3233698194405464e-05, "loss": 0.1048, "step": 14465 }, { "epoch": 0.68, "learning_rate": 1.3231359341378989e-05, "loss": 0.1706, "step": 14470 }, { "epoch": 0.68, "learning_rate": 1.3229020488352514e-05, "loss": 0.1445, "step": 14475 }, { "epoch": 0.68, "learning_rate": 1.3226681635326037e-05, "loss": 0.0984, "step": 14480 }, { "epoch": 0.68, "learning_rate": 1.3224342782299562e-05, "loss": 0.07, "step": 14485 }, { "epoch": 0.68, "learning_rate": 1.3222003929273087e-05, "loss": 0.0617, "step": 14490 }, { "epoch": 0.68, "learning_rate": 1.321966507624661e-05, "loss": 0.1035, "step": 14495 }, { "epoch": 0.68, "learning_rate": 1.3217326223220135e-05, "loss": 0.1428, "step": 14500 }, { "epoch": 0.68, "learning_rate": 1.3214987370193657e-05, "loss": 0.1163, "step": 14505 }, { "epoch": 0.68, "learning_rate": 1.3212648517167181e-05, "loss": 0.139, "step": 14510 }, { "epoch": 0.68, "learning_rate": 1.3210309664140706e-05, "loss": 0.0793, "step": 14515 }, { "epoch": 0.68, "learning_rate": 1.320797081111423e-05, "loss": 0.1098, "step": 14520 }, { "epoch": 0.68, "learning_rate": 1.3205631958087755e-05, "loss": 0.132, "step": 14525 }, { "epoch": 0.68, "learning_rate": 1.320329310506128e-05, "loss": 0.1776, "step": 14530 }, { "epoch": 0.68, "learning_rate": 1.3200954252034803e-05, "loss": 0.0857, "step": 14535 }, { "epoch": 0.68, "learning_rate": 1.3198615399008328e-05, "loss": 0.0888, "step": 14540 }, { "epoch": 0.68, "learning_rate": 1.3196276545981853e-05, "loss": 0.0928, "step": 14545 }, { "epoch": 0.68, "learning_rate": 1.3193937692955376e-05, "loss": 0.1408, "step": 14550 }, { "epoch": 0.68, "learning_rate": 1.3191598839928901e-05, "loss": 0.0841, "step": 14555 }, { "epoch": 0.68, "learning_rate": 1.3189259986902426e-05, "loss": 0.1313, "step": 14560 }, { "epoch": 0.68, "learning_rate": 1.3186921133875947e-05, "loss": 0.0735, "step": 14565 }, { "epoch": 0.68, "learning_rate": 1.3184582280849472e-05, "loss": 0.0925, "step": 14570 }, { "epoch": 0.68, "learning_rate": 1.3182243427822996e-05, "loss": 0.0574, "step": 14575 }, { "epoch": 0.68, "learning_rate": 1.317990457479652e-05, "loss": 0.1284, "step": 14580 }, { "epoch": 0.68, "learning_rate": 1.3177565721770046e-05, "loss": 0.1429, "step": 14585 }, { "epoch": 0.68, "learning_rate": 1.3175226868743569e-05, "loss": 0.1423, "step": 14590 }, { "epoch": 0.68, "learning_rate": 1.3172888015717094e-05, "loss": 0.0798, "step": 14595 }, { "epoch": 0.68, "learning_rate": 1.3170549162690617e-05, "loss": 0.1433, "step": 14600 }, { "epoch": 0.68, "learning_rate": 1.3168210309664142e-05, "loss": 0.1324, "step": 14605 }, { "epoch": 0.68, "learning_rate": 1.3165871456637667e-05, "loss": 0.1119, "step": 14610 }, { "epoch": 0.68, "learning_rate": 1.316353260361119e-05, "loss": 0.0752, "step": 14615 }, { "epoch": 0.68, "learning_rate": 1.3161193750584715e-05, "loss": 0.0775, "step": 14620 }, { "epoch": 0.68, "learning_rate": 1.3158854897558237e-05, "loss": 0.0639, "step": 14625 }, { "epoch": 0.68, "learning_rate": 1.3156516044531762e-05, "loss": 0.0519, "step": 14630 }, { "epoch": 0.68, "learning_rate": 1.3154177191505287e-05, "loss": 0.1375, "step": 14635 }, { "epoch": 0.68, "learning_rate": 1.315183833847881e-05, "loss": 0.0991, "step": 14640 }, { "epoch": 0.69, "learning_rate": 1.3149499485452335e-05, "loss": 0.136, "step": 14645 }, { "epoch": 0.69, "learning_rate": 1.314716063242586e-05, "loss": 0.1041, "step": 14650 }, { "epoch": 0.69, "learning_rate": 1.3144821779399383e-05, "loss": 0.0786, "step": 14655 }, { "epoch": 0.69, "learning_rate": 1.3142482926372908e-05, "loss": 0.0695, "step": 14660 }, { "epoch": 0.69, "learning_rate": 1.3140144073346433e-05, "loss": 0.1017, "step": 14665 }, { "epoch": 0.69, "learning_rate": 1.3137805220319956e-05, "loss": 0.0766, "step": 14670 }, { "epoch": 0.69, "learning_rate": 1.3135466367293481e-05, "loss": 0.1315, "step": 14675 }, { "epoch": 0.69, "learning_rate": 1.3133127514267006e-05, "loss": 0.1489, "step": 14680 }, { "epoch": 0.69, "learning_rate": 1.3130788661240528e-05, "loss": 0.044, "step": 14685 }, { "epoch": 0.69, "learning_rate": 1.3128449808214053e-05, "loss": 0.0993, "step": 14690 }, { "epoch": 0.69, "learning_rate": 1.3126110955187576e-05, "loss": 0.1215, "step": 14695 }, { "epoch": 0.69, "learning_rate": 1.3123772102161101e-05, "loss": 0.061, "step": 14700 }, { "epoch": 0.69, "learning_rate": 1.3121433249134626e-05, "loss": 0.1467, "step": 14705 }, { "epoch": 0.69, "learning_rate": 1.3119094396108149e-05, "loss": 0.1216, "step": 14710 }, { "epoch": 0.69, "learning_rate": 1.3116755543081674e-05, "loss": 0.0582, "step": 14715 }, { "epoch": 0.69, "learning_rate": 1.3114416690055199e-05, "loss": 0.1215, "step": 14720 }, { "epoch": 0.69, "learning_rate": 1.3112077837028722e-05, "loss": 0.1332, "step": 14725 }, { "epoch": 0.69, "learning_rate": 1.3109738984002247e-05, "loss": 0.0703, "step": 14730 }, { "epoch": 0.69, "learning_rate": 1.3107400130975772e-05, "loss": 0.07, "step": 14735 }, { "epoch": 0.69, "learning_rate": 1.3105061277949295e-05, "loss": 0.0761, "step": 14740 }, { "epoch": 0.69, "learning_rate": 1.3102722424922819e-05, "loss": 0.1961, "step": 14745 }, { "epoch": 0.69, "learning_rate": 1.3100383571896342e-05, "loss": 0.105, "step": 14750 }, { "epoch": 0.69, "learning_rate": 1.3098044718869867e-05, "loss": 0.1054, "step": 14755 }, { "epoch": 0.69, "learning_rate": 1.3095705865843392e-05, "loss": 0.0782, "step": 14760 }, { "epoch": 0.69, "learning_rate": 1.3093367012816915e-05, "loss": 0.0531, "step": 14765 }, { "epoch": 0.69, "learning_rate": 1.309102815979044e-05, "loss": 0.0776, "step": 14770 }, { "epoch": 0.69, "learning_rate": 1.3088689306763965e-05, "loss": 0.0943, "step": 14775 }, { "epoch": 0.69, "learning_rate": 1.3086350453737488e-05, "loss": 0.0848, "step": 14780 }, { "epoch": 0.69, "learning_rate": 1.3084011600711013e-05, "loss": 0.0965, "step": 14785 }, { "epoch": 0.69, "learning_rate": 1.3081672747684538e-05, "loss": 0.1633, "step": 14790 }, { "epoch": 0.69, "learning_rate": 1.3079333894658061e-05, "loss": 0.1017, "step": 14795 }, { "epoch": 0.69, "learning_rate": 1.3076995041631585e-05, "loss": 0.0702, "step": 14800 }, { "epoch": 0.69, "learning_rate": 1.3074656188605108e-05, "loss": 0.0598, "step": 14805 }, { "epoch": 0.69, "learning_rate": 1.3072317335578633e-05, "loss": 0.0576, "step": 14810 }, { "epoch": 0.69, "learning_rate": 1.3069978482552158e-05, "loss": 0.1455, "step": 14815 }, { "epoch": 0.69, "learning_rate": 1.3067639629525681e-05, "loss": 0.0937, "step": 14820 }, { "epoch": 0.69, "learning_rate": 1.3065300776499206e-05, "loss": 0.0752, "step": 14825 }, { "epoch": 0.69, "learning_rate": 1.306296192347273e-05, "loss": 0.1113, "step": 14830 }, { "epoch": 0.69, "learning_rate": 1.3060623070446254e-05, "loss": 0.1081, "step": 14835 }, { "epoch": 0.69, "learning_rate": 1.305828421741978e-05, "loss": 0.085, "step": 14840 }, { "epoch": 0.69, "learning_rate": 1.3055945364393302e-05, "loss": 0.0868, "step": 14845 }, { "epoch": 0.69, "learning_rate": 1.3053606511366827e-05, "loss": 0.0855, "step": 14850 }, { "epoch": 0.69, "learning_rate": 1.3051267658340352e-05, "loss": 0.1057, "step": 14855 }, { "epoch": 0.7, "learning_rate": 1.3048928805313874e-05, "loss": 0.1104, "step": 14860 }, { "epoch": 0.7, "learning_rate": 1.3046589952287399e-05, "loss": 0.1206, "step": 14865 }, { "epoch": 0.7, "learning_rate": 1.3044251099260922e-05, "loss": 0.0854, "step": 14870 }, { "epoch": 0.7, "learning_rate": 1.3041912246234447e-05, "loss": 0.1089, "step": 14875 }, { "epoch": 0.7, "learning_rate": 1.3039573393207972e-05, "loss": 0.0565, "step": 14880 }, { "epoch": 0.7, "learning_rate": 1.3037234540181495e-05, "loss": 0.0715, "step": 14885 }, { "epoch": 0.7, "learning_rate": 1.303489568715502e-05, "loss": 0.0569, "step": 14890 }, { "epoch": 0.7, "learning_rate": 1.3032556834128545e-05, "loss": 0.0711, "step": 14895 }, { "epoch": 0.7, "learning_rate": 1.3030217981102068e-05, "loss": 0.056, "step": 14900 }, { "epoch": 0.7, "learning_rate": 1.3027879128075593e-05, "loss": 0.159, "step": 14905 }, { "epoch": 0.7, "learning_rate": 1.3025540275049118e-05, "loss": 0.0591, "step": 14910 }, { "epoch": 0.7, "learning_rate": 1.3023201422022642e-05, "loss": 0.1011, "step": 14915 }, { "epoch": 0.7, "learning_rate": 1.3020862568996165e-05, "loss": 0.0609, "step": 14920 }, { "epoch": 0.7, "learning_rate": 1.3018523715969688e-05, "loss": 0.1277, "step": 14925 }, { "epoch": 0.7, "learning_rate": 1.3016184862943213e-05, "loss": 0.0922, "step": 14930 }, { "epoch": 0.7, "learning_rate": 1.3013846009916738e-05, "loss": 0.1087, "step": 14935 }, { "epoch": 0.7, "learning_rate": 1.3011507156890261e-05, "loss": 0.137, "step": 14940 }, { "epoch": 0.7, "learning_rate": 1.3009168303863786e-05, "loss": 0.0945, "step": 14945 }, { "epoch": 0.7, "learning_rate": 1.3006829450837311e-05, "loss": 0.0323, "step": 14950 }, { "epoch": 0.7, "learning_rate": 1.3004490597810834e-05, "loss": 0.1042, "step": 14955 }, { "epoch": 0.7, "learning_rate": 1.300215174478436e-05, "loss": 0.1088, "step": 14960 }, { "epoch": 0.7, "learning_rate": 1.2999812891757884e-05, "loss": 0.0743, "step": 14965 }, { "epoch": 0.7, "learning_rate": 1.2997474038731408e-05, "loss": 0.107, "step": 14970 }, { "epoch": 0.7, "learning_rate": 1.2995135185704933e-05, "loss": 0.1071, "step": 14975 }, { "epoch": 0.7, "learning_rate": 1.2992796332678454e-05, "loss": 0.1564, "step": 14980 }, { "epoch": 0.7, "learning_rate": 1.2990457479651979e-05, "loss": 0.1462, "step": 14985 }, { "epoch": 0.7, "learning_rate": 1.2988118626625504e-05, "loss": 0.0758, "step": 14990 }, { "epoch": 0.7, "learning_rate": 1.2985779773599027e-05, "loss": 0.0619, "step": 14995 }, { "epoch": 0.7, "learning_rate": 1.2983440920572552e-05, "loss": 0.0871, "step": 15000 }, { "epoch": 0.7, "learning_rate": 1.2981102067546077e-05, "loss": 0.0647, "step": 15005 }, { "epoch": 0.7, "learning_rate": 1.29787632145196e-05, "loss": 0.0839, "step": 15010 }, { "epoch": 0.7, "learning_rate": 1.2976424361493125e-05, "loss": 0.1019, "step": 15015 }, { "epoch": 0.7, "learning_rate": 1.297408550846665e-05, "loss": 0.0673, "step": 15020 }, { "epoch": 0.7, "learning_rate": 1.2971746655440174e-05, "loss": 0.1423, "step": 15025 }, { "epoch": 0.7, "learning_rate": 1.2969407802413699e-05, "loss": 0.125, "step": 15030 }, { "epoch": 0.7, "learning_rate": 1.2967068949387222e-05, "loss": 0.1525, "step": 15035 }, { "epoch": 0.7, "learning_rate": 1.2964730096360745e-05, "loss": 0.1292, "step": 15040 }, { "epoch": 0.7, "learning_rate": 1.2962391243334268e-05, "loss": 0.0955, "step": 15045 }, { "epoch": 0.7, "learning_rate": 1.2960052390307793e-05, "loss": 0.1133, "step": 15050 }, { "epoch": 0.7, "learning_rate": 1.2957713537281318e-05, "loss": 0.0599, "step": 15055 }, { "epoch": 0.7, "learning_rate": 1.2955374684254841e-05, "loss": 0.1252, "step": 15060 }, { "epoch": 0.7, "learning_rate": 1.2953035831228366e-05, "loss": 0.1039, "step": 15065 }, { "epoch": 0.7, "learning_rate": 1.2950696978201891e-05, "loss": 0.051, "step": 15070 }, { "epoch": 0.71, "learning_rate": 1.2948358125175415e-05, "loss": 0.0905, "step": 15075 }, { "epoch": 0.71, "learning_rate": 1.294601927214894e-05, "loss": 0.1108, "step": 15080 }, { "epoch": 0.71, "learning_rate": 1.2943680419122465e-05, "loss": 0.1123, "step": 15085 }, { "epoch": 0.71, "learning_rate": 1.2941341566095988e-05, "loss": 0.0646, "step": 15090 }, { "epoch": 0.71, "learning_rate": 1.2939002713069513e-05, "loss": 0.1403, "step": 15095 }, { "epoch": 0.71, "learning_rate": 1.2936663860043034e-05, "loss": 0.1187, "step": 15100 }, { "epoch": 0.71, "learning_rate": 1.293432500701656e-05, "loss": 0.0772, "step": 15105 }, { "epoch": 0.71, "learning_rate": 1.2931986153990084e-05, "loss": 0.135, "step": 15110 }, { "epoch": 0.71, "learning_rate": 1.2929647300963607e-05, "loss": 0.1047, "step": 15115 }, { "epoch": 0.71, "learning_rate": 1.2927308447937132e-05, "loss": 0.1199, "step": 15120 }, { "epoch": 0.71, "learning_rate": 1.2924969594910657e-05, "loss": 0.0815, "step": 15125 }, { "epoch": 0.71, "learning_rate": 1.292263074188418e-05, "loss": 0.1469, "step": 15130 }, { "epoch": 0.71, "learning_rate": 1.2920291888857706e-05, "loss": 0.1092, "step": 15135 }, { "epoch": 0.71, "learning_rate": 1.291795303583123e-05, "loss": 0.0949, "step": 15140 }, { "epoch": 0.71, "learning_rate": 1.2915614182804754e-05, "loss": 0.0702, "step": 15145 }, { "epoch": 0.71, "learning_rate": 1.2913275329778279e-05, "loss": 0.0887, "step": 15150 }, { "epoch": 0.71, "learning_rate": 1.2910936476751804e-05, "loss": 0.1177, "step": 15155 }, { "epoch": 0.71, "learning_rate": 1.2908597623725325e-05, "loss": 0.0459, "step": 15160 }, { "epoch": 0.71, "learning_rate": 1.290625877069885e-05, "loss": 0.0712, "step": 15165 }, { "epoch": 0.71, "learning_rate": 1.2903919917672373e-05, "loss": 0.1188, "step": 15170 }, { "epoch": 0.71, "learning_rate": 1.2901581064645898e-05, "loss": 0.1444, "step": 15175 }, { "epoch": 0.71, "learning_rate": 1.2899242211619423e-05, "loss": 0.1456, "step": 15180 }, { "epoch": 0.71, "learning_rate": 1.2896903358592947e-05, "loss": 0.0976, "step": 15185 }, { "epoch": 0.71, "learning_rate": 1.2894564505566472e-05, "loss": 0.0563, "step": 15190 }, { "epoch": 0.71, "learning_rate": 1.2892225652539997e-05, "loss": 0.1084, "step": 15195 }, { "epoch": 0.71, "learning_rate": 1.288988679951352e-05, "loss": 0.0585, "step": 15200 }, { "epoch": 0.71, "learning_rate": 1.2887547946487045e-05, "loss": 0.1033, "step": 15205 }, { "epoch": 0.71, "learning_rate": 1.288520909346057e-05, "loss": 0.0794, "step": 15210 }, { "epoch": 0.71, "learning_rate": 1.2882870240434091e-05, "loss": 0.0602, "step": 15215 }, { "epoch": 0.71, "learning_rate": 1.2880531387407616e-05, "loss": 0.0328, "step": 15220 }, { "epoch": 0.71, "learning_rate": 1.287819253438114e-05, "loss": 0.1002, "step": 15225 }, { "epoch": 0.71, "learning_rate": 1.2875853681354664e-05, "loss": 0.1528, "step": 15230 }, { "epoch": 0.71, "learning_rate": 1.287351482832819e-05, "loss": 0.0739, "step": 15235 }, { "epoch": 0.71, "learning_rate": 1.2871175975301713e-05, "loss": 0.0842, "step": 15240 }, { "epoch": 0.71, "learning_rate": 1.2868837122275238e-05, "loss": 0.0914, "step": 15245 }, { "epoch": 0.71, "learning_rate": 1.2866498269248763e-05, "loss": 0.0726, "step": 15250 }, { "epoch": 0.71, "learning_rate": 1.2864159416222286e-05, "loss": 0.1649, "step": 15255 }, { "epoch": 0.71, "learning_rate": 1.286182056319581e-05, "loss": 0.1769, "step": 15260 }, { "epoch": 0.71, "learning_rate": 1.2859481710169334e-05, "loss": 0.1376, "step": 15265 }, { "epoch": 0.71, "learning_rate": 1.2857142857142859e-05, "loss": 0.0935, "step": 15270 }, { "epoch": 0.71, "learning_rate": 1.285480400411638e-05, "loss": 0.1099, "step": 15275 }, { "epoch": 0.71, "learning_rate": 1.2852465151089905e-05, "loss": 0.1018, "step": 15280 }, { "epoch": 0.71, "learning_rate": 1.285012629806343e-05, "loss": 0.1131, "step": 15285 }, { "epoch": 0.72, "learning_rate": 1.2847787445036954e-05, "loss": 0.1171, "step": 15290 }, { "epoch": 0.72, "learning_rate": 1.2845448592010479e-05, "loss": 0.0505, "step": 15295 }, { "epoch": 0.72, "learning_rate": 1.2843109738984004e-05, "loss": 0.1161, "step": 15300 }, { "epoch": 0.72, "learning_rate": 1.2840770885957527e-05, "loss": 0.093, "step": 15305 }, { "epoch": 0.72, "learning_rate": 1.2838432032931052e-05, "loss": 0.0515, "step": 15310 }, { "epoch": 0.72, "learning_rate": 1.2836093179904577e-05, "loss": 0.0836, "step": 15315 }, { "epoch": 0.72, "learning_rate": 1.28337543268781e-05, "loss": 0.1019, "step": 15320 }, { "epoch": 0.72, "learning_rate": 1.2831415473851625e-05, "loss": 0.0799, "step": 15325 }, { "epoch": 0.72, "learning_rate": 1.282907662082515e-05, "loss": 0.0959, "step": 15330 }, { "epoch": 0.72, "learning_rate": 1.2826737767798671e-05, "loss": 0.0884, "step": 15335 }, { "epoch": 0.72, "learning_rate": 1.2824398914772196e-05, "loss": 0.1093, "step": 15340 }, { "epoch": 0.72, "learning_rate": 1.282206006174572e-05, "loss": 0.066, "step": 15345 }, { "epoch": 0.72, "learning_rate": 1.2819721208719245e-05, "loss": 0.1311, "step": 15350 }, { "epoch": 0.72, "learning_rate": 1.281738235569277e-05, "loss": 0.0727, "step": 15355 }, { "epoch": 0.72, "learning_rate": 1.2815043502666293e-05, "loss": 0.118, "step": 15360 }, { "epoch": 0.72, "learning_rate": 1.2812704649639818e-05, "loss": 0.0653, "step": 15365 }, { "epoch": 0.72, "learning_rate": 1.2810365796613343e-05, "loss": 0.0623, "step": 15370 }, { "epoch": 0.72, "learning_rate": 1.2808026943586866e-05, "loss": 0.0413, "step": 15375 }, { "epoch": 0.72, "learning_rate": 1.2805688090560391e-05, "loss": 0.0572, "step": 15380 }, { "epoch": 0.72, "learning_rate": 1.2803349237533916e-05, "loss": 0.0654, "step": 15385 }, { "epoch": 0.72, "learning_rate": 1.2801010384507439e-05, "loss": 0.131, "step": 15390 }, { "epoch": 0.72, "learning_rate": 1.2798671531480962e-05, "loss": 0.1123, "step": 15395 }, { "epoch": 0.72, "learning_rate": 1.2796332678454486e-05, "loss": 0.073, "step": 15400 }, { "epoch": 0.72, "learning_rate": 1.279399382542801e-05, "loss": 0.1136, "step": 15405 }, { "epoch": 0.72, "learning_rate": 1.2791654972401536e-05, "loss": 0.1227, "step": 15410 }, { "epoch": 0.72, "learning_rate": 1.2789316119375059e-05, "loss": 0.1049, "step": 15415 }, { "epoch": 0.72, "learning_rate": 1.2786977266348584e-05, "loss": 0.0935, "step": 15420 }, { "epoch": 0.72, "learning_rate": 1.2784638413322109e-05, "loss": 0.1141, "step": 15425 }, { "epoch": 0.72, "learning_rate": 1.2782299560295632e-05, "loss": 0.1017, "step": 15430 }, { "epoch": 0.72, "learning_rate": 1.2779960707269157e-05, "loss": 0.0558, "step": 15435 }, { "epoch": 0.72, "learning_rate": 1.2777621854242682e-05, "loss": 0.0761, "step": 15440 }, { "epoch": 0.72, "learning_rate": 1.2775283001216205e-05, "loss": 0.0683, "step": 15445 }, { "epoch": 0.72, "learning_rate": 1.277294414818973e-05, "loss": 0.0848, "step": 15450 }, { "epoch": 0.72, "learning_rate": 1.2770605295163252e-05, "loss": 0.09, "step": 15455 }, { "epoch": 0.72, "learning_rate": 1.2768266442136777e-05, "loss": 0.1442, "step": 15460 }, { "epoch": 0.72, "learning_rate": 1.2765927589110302e-05, "loss": 0.0432, "step": 15465 }, { "epoch": 0.72, "learning_rate": 1.2763588736083825e-05, "loss": 0.1035, "step": 15470 }, { "epoch": 0.72, "learning_rate": 1.276124988305735e-05, "loss": 0.1066, "step": 15475 }, { "epoch": 0.72, "learning_rate": 1.2758911030030873e-05, "loss": 0.0743, "step": 15480 }, { "epoch": 0.72, "learning_rate": 1.2756572177004398e-05, "loss": 0.1201, "step": 15485 }, { "epoch": 0.72, "learning_rate": 1.2754233323977923e-05, "loss": 0.0661, "step": 15490 }, { "epoch": 0.72, "learning_rate": 1.2751894470951446e-05, "loss": 0.0581, "step": 15495 }, { "epoch": 0.73, "learning_rate": 1.2749555617924971e-05, "loss": 0.0692, "step": 15500 }, { "epoch": 0.73, "learning_rate": 1.2747216764898496e-05, "loss": 0.0582, "step": 15505 }, { "epoch": 0.73, "learning_rate": 1.274487791187202e-05, "loss": 0.1104, "step": 15510 }, { "epoch": 0.73, "learning_rate": 1.2742539058845543e-05, "loss": 0.1353, "step": 15515 }, { "epoch": 0.73, "learning_rate": 1.2740200205819066e-05, "loss": 0.0584, "step": 15520 }, { "epoch": 0.73, "learning_rate": 1.273786135279259e-05, "loss": 0.1304, "step": 15525 }, { "epoch": 0.73, "learning_rate": 1.2735522499766116e-05, "loss": 0.1162, "step": 15530 }, { "epoch": 0.73, "learning_rate": 1.2733183646739639e-05, "loss": 0.1221, "step": 15535 }, { "epoch": 0.73, "learning_rate": 1.2730844793713164e-05, "loss": 0.1084, "step": 15540 }, { "epoch": 0.73, "learning_rate": 1.2728505940686689e-05, "loss": 0.1653, "step": 15545 }, { "epoch": 0.73, "learning_rate": 1.2726167087660212e-05, "loss": 0.0788, "step": 15550 }, { "epoch": 0.73, "learning_rate": 1.2723828234633737e-05, "loss": 0.1382, "step": 15555 }, { "epoch": 0.73, "learning_rate": 1.2721489381607262e-05, "loss": 0.0739, "step": 15560 }, { "epoch": 0.73, "learning_rate": 1.2719150528580785e-05, "loss": 0.1001, "step": 15565 }, { "epoch": 0.73, "learning_rate": 1.271681167555431e-05, "loss": 0.0846, "step": 15570 }, { "epoch": 0.73, "learning_rate": 1.2714472822527832e-05, "loss": 0.1087, "step": 15575 }, { "epoch": 0.73, "learning_rate": 1.2712133969501357e-05, "loss": 0.1161, "step": 15580 }, { "epoch": 0.73, "learning_rate": 1.2709795116474882e-05, "loss": 0.1035, "step": 15585 }, { "epoch": 0.73, "learning_rate": 1.2707456263448405e-05, "loss": 0.0381, "step": 15590 }, { "epoch": 0.73, "learning_rate": 1.270511741042193e-05, "loss": 0.0846, "step": 15595 }, { "epoch": 0.73, "learning_rate": 1.2702778557395455e-05, "loss": 0.0663, "step": 15600 }, { "epoch": 0.73, "learning_rate": 1.2700439704368978e-05, "loss": 0.0495, "step": 15605 }, { "epoch": 0.73, "learning_rate": 1.2698100851342503e-05, "loss": 0.0624, "step": 15610 }, { "epoch": 0.73, "learning_rate": 1.2695761998316028e-05, "loss": 0.0905, "step": 15615 }, { "epoch": 0.73, "learning_rate": 1.2693423145289551e-05, "loss": 0.0737, "step": 15620 }, { "epoch": 0.73, "learning_rate": 1.2691084292263076e-05, "loss": 0.1139, "step": 15625 }, { "epoch": 0.73, "learning_rate": 1.2688745439236601e-05, "loss": 0.103, "step": 15630 }, { "epoch": 0.73, "learning_rate": 1.2686406586210123e-05, "loss": 0.1478, "step": 15635 }, { "epoch": 0.73, "learning_rate": 1.2684067733183648e-05, "loss": 0.0609, "step": 15640 }, { "epoch": 0.73, "learning_rate": 1.2681728880157171e-05, "loss": 0.0951, "step": 15645 }, { "epoch": 0.73, "learning_rate": 1.2679390027130696e-05, "loss": 0.1002, "step": 15650 }, { "epoch": 0.73, "learning_rate": 1.2677051174104221e-05, "loss": 0.0846, "step": 15655 }, { "epoch": 0.73, "learning_rate": 1.2674712321077744e-05, "loss": 0.0699, "step": 15660 }, { "epoch": 0.73, "learning_rate": 1.2672373468051269e-05, "loss": 0.0463, "step": 15665 }, { "epoch": 0.73, "learning_rate": 1.2670034615024794e-05, "loss": 0.1092, "step": 15670 }, { "epoch": 0.73, "learning_rate": 1.2667695761998317e-05, "loss": 0.13, "step": 15675 }, { "epoch": 0.73, "learning_rate": 1.2665356908971842e-05, "loss": 0.126, "step": 15680 }, { "epoch": 0.73, "learning_rate": 1.2663018055945367e-05, "loss": 0.1265, "step": 15685 }, { "epoch": 0.73, "learning_rate": 1.2660679202918889e-05, "loss": 0.1312, "step": 15690 }, { "epoch": 0.73, "learning_rate": 1.2658340349892414e-05, "loss": 0.0812, "step": 15695 }, { "epoch": 0.73, "learning_rate": 1.2656001496865937e-05, "loss": 0.0568, "step": 15700 }, { "epoch": 0.73, "learning_rate": 1.2653662643839462e-05, "loss": 0.1126, "step": 15705 }, { "epoch": 0.73, "learning_rate": 1.2651323790812985e-05, "loss": 0.0832, "step": 15710 }, { "epoch": 0.74, "learning_rate": 1.264898493778651e-05, "loss": 0.0878, "step": 15715 }, { "epoch": 0.74, "learning_rate": 1.2646646084760035e-05, "loss": 0.1365, "step": 15720 }, { "epoch": 0.74, "learning_rate": 1.2644307231733558e-05, "loss": 0.1434, "step": 15725 }, { "epoch": 0.74, "learning_rate": 1.2641968378707083e-05, "loss": 0.1098, "step": 15730 }, { "epoch": 0.74, "learning_rate": 1.2639629525680608e-05, "loss": 0.1069, "step": 15735 }, { "epoch": 0.74, "learning_rate": 1.2637290672654132e-05, "loss": 0.1264, "step": 15740 }, { "epoch": 0.74, "learning_rate": 1.2634951819627656e-05, "loss": 0.0631, "step": 15745 }, { "epoch": 0.74, "learning_rate": 1.2632612966601178e-05, "loss": 0.0985, "step": 15750 }, { "epoch": 0.74, "learning_rate": 1.2630274113574703e-05, "loss": 0.0727, "step": 15755 }, { "epoch": 0.74, "learning_rate": 1.2627935260548228e-05, "loss": 0.1382, "step": 15760 }, { "epoch": 0.74, "learning_rate": 1.2625596407521751e-05, "loss": 0.1189, "step": 15765 }, { "epoch": 0.74, "learning_rate": 1.2623257554495276e-05, "loss": 0.0871, "step": 15770 }, { "epoch": 0.74, "learning_rate": 1.2620918701468801e-05, "loss": 0.093, "step": 15775 }, { "epoch": 0.74, "learning_rate": 1.2618579848442324e-05, "loss": 0.0543, "step": 15780 }, { "epoch": 0.74, "learning_rate": 1.261624099541585e-05, "loss": 0.0893, "step": 15785 }, { "epoch": 0.74, "learning_rate": 1.2613902142389374e-05, "loss": 0.0482, "step": 15790 }, { "epoch": 0.74, "learning_rate": 1.2611563289362898e-05, "loss": 0.0644, "step": 15795 }, { "epoch": 0.74, "learning_rate": 1.2609224436336422e-05, "loss": 0.0515, "step": 15800 }, { "epoch": 0.74, "learning_rate": 1.2606885583309947e-05, "loss": 0.0891, "step": 15805 }, { "epoch": 0.74, "learning_rate": 1.2604546730283469e-05, "loss": 0.1041, "step": 15810 }, { "epoch": 0.74, "learning_rate": 1.2602207877256994e-05, "loss": 0.1027, "step": 15815 }, { "epoch": 0.74, "learning_rate": 1.2599869024230517e-05, "loss": 0.0912, "step": 15820 }, { "epoch": 0.74, "learning_rate": 1.2597530171204042e-05, "loss": 0.0812, "step": 15825 }, { "epoch": 0.74, "learning_rate": 1.2595191318177567e-05, "loss": 0.0946, "step": 15830 }, { "epoch": 0.74, "learning_rate": 1.259285246515109e-05, "loss": 0.1184, "step": 15835 }, { "epoch": 0.74, "learning_rate": 1.2590513612124615e-05, "loss": 0.0843, "step": 15840 }, { "epoch": 0.74, "learning_rate": 1.258817475909814e-05, "loss": 0.082, "step": 15845 }, { "epoch": 0.74, "learning_rate": 1.2585835906071664e-05, "loss": 0.0391, "step": 15850 }, { "epoch": 0.74, "learning_rate": 1.2583497053045188e-05, "loss": 0.1046, "step": 15855 }, { "epoch": 0.74, "learning_rate": 1.2581158200018713e-05, "loss": 0.147, "step": 15860 }, { "epoch": 0.74, "learning_rate": 1.2578819346992237e-05, "loss": 0.0837, "step": 15865 }, { "epoch": 0.74, "learning_rate": 1.257648049396576e-05, "loss": 0.0803, "step": 15870 }, { "epoch": 0.74, "learning_rate": 1.2574141640939283e-05, "loss": 0.1199, "step": 15875 }, { "epoch": 0.74, "learning_rate": 1.2571802787912808e-05, "loss": 0.049, "step": 15880 }, { "epoch": 0.74, "learning_rate": 1.2569463934886333e-05, "loss": 0.1025, "step": 15885 }, { "epoch": 0.74, "learning_rate": 1.2567125081859856e-05, "loss": 0.1266, "step": 15890 }, { "epoch": 0.74, "learning_rate": 1.2564786228833381e-05, "loss": 0.0637, "step": 15895 }, { "epoch": 0.74, "learning_rate": 1.2562447375806906e-05, "loss": 0.0751, "step": 15900 }, { "epoch": 0.74, "learning_rate": 1.256010852278043e-05, "loss": 0.07, "step": 15905 }, { "epoch": 0.74, "learning_rate": 1.2557769669753954e-05, "loss": 0.122, "step": 15910 }, { "epoch": 0.74, "learning_rate": 1.2555430816727478e-05, "loss": 0.1434, "step": 15915 }, { "epoch": 0.74, "learning_rate": 1.2553091963701003e-05, "loss": 0.0963, "step": 15920 }, { "epoch": 0.74, "learning_rate": 1.2550753110674528e-05, "loss": 0.0712, "step": 15925 }, { "epoch": 0.75, "learning_rate": 1.254841425764805e-05, "loss": 0.107, "step": 15930 }, { "epoch": 0.75, "learning_rate": 1.2546075404621574e-05, "loss": 0.1091, "step": 15935 }, { "epoch": 0.75, "learning_rate": 1.2543736551595097e-05, "loss": 0.1548, "step": 15940 }, { "epoch": 0.75, "learning_rate": 1.2541397698568622e-05, "loss": 0.07, "step": 15945 }, { "epoch": 0.75, "learning_rate": 1.2539058845542147e-05, "loss": 0.1334, "step": 15950 }, { "epoch": 0.75, "learning_rate": 1.253671999251567e-05, "loss": 0.1325, "step": 15955 }, { "epoch": 0.75, "learning_rate": 1.2534381139489196e-05, "loss": 0.1285, "step": 15960 }, { "epoch": 0.75, "learning_rate": 1.253204228646272e-05, "loss": 0.094, "step": 15965 }, { "epoch": 0.75, "learning_rate": 1.2529703433436244e-05, "loss": 0.0721, "step": 15970 }, { "epoch": 0.75, "learning_rate": 1.2527364580409769e-05, "loss": 0.0702, "step": 15975 }, { "epoch": 0.75, "learning_rate": 1.2525025727383294e-05, "loss": 0.0621, "step": 15980 }, { "epoch": 0.75, "learning_rate": 1.2522686874356817e-05, "loss": 0.1022, "step": 15985 }, { "epoch": 0.75, "learning_rate": 1.252034802133034e-05, "loss": 0.1387, "step": 15990 }, { "epoch": 0.75, "learning_rate": 1.2518009168303863e-05, "loss": 0.0765, "step": 15995 }, { "epoch": 0.75, "learning_rate": 1.2515670315277388e-05, "loss": 0.0597, "step": 16000 }, { "epoch": 0.75, "learning_rate": 1.2513331462250913e-05, "loss": 0.0949, "step": 16005 }, { "epoch": 0.75, "learning_rate": 1.2510992609224437e-05, "loss": 0.0387, "step": 16010 }, { "epoch": 0.75, "learning_rate": 1.2508653756197961e-05, "loss": 0.1368, "step": 16015 }, { "epoch": 0.75, "learning_rate": 1.2506314903171486e-05, "loss": 0.101, "step": 16020 }, { "epoch": 0.75, "learning_rate": 1.250397605014501e-05, "loss": 0.0625, "step": 16025 }, { "epoch": 0.75, "learning_rate": 1.2501637197118535e-05, "loss": 0.1068, "step": 16030 }, { "epoch": 0.75, "learning_rate": 1.249929834409206e-05, "loss": 0.0889, "step": 16035 }, { "epoch": 0.75, "learning_rate": 1.2496959491065583e-05, "loss": 0.0886, "step": 16040 }, { "epoch": 0.75, "learning_rate": 1.2494620638039108e-05, "loss": 0.0439, "step": 16045 }, { "epoch": 0.75, "learning_rate": 1.249228178501263e-05, "loss": 0.1181, "step": 16050 }, { "epoch": 0.75, "learning_rate": 1.2489942931986154e-05, "loss": 0.1245, "step": 16055 }, { "epoch": 0.75, "learning_rate": 1.248760407895968e-05, "loss": 0.1035, "step": 16060 }, { "epoch": 0.75, "learning_rate": 1.2485265225933203e-05, "loss": 0.1156, "step": 16065 }, { "epoch": 0.75, "learning_rate": 1.2482926372906727e-05, "loss": 0.0881, "step": 16070 }, { "epoch": 0.75, "learning_rate": 1.2480587519880252e-05, "loss": 0.0476, "step": 16075 }, { "epoch": 0.75, "learning_rate": 1.2478248666853776e-05, "loss": 0.0666, "step": 16080 }, { "epoch": 0.75, "learning_rate": 1.24759098138273e-05, "loss": 0.0859, "step": 16085 }, { "epoch": 0.75, "learning_rate": 1.2473570960800826e-05, "loss": 0.0916, "step": 16090 }, { "epoch": 0.75, "learning_rate": 1.2471232107774349e-05, "loss": 0.0575, "step": 16095 }, { "epoch": 0.75, "learning_rate": 1.2468893254747874e-05, "loss": 0.0641, "step": 16100 }, { "epoch": 0.75, "learning_rate": 1.2466554401721395e-05, "loss": 0.1079, "step": 16105 }, { "epoch": 0.75, "learning_rate": 1.246421554869492e-05, "loss": 0.0955, "step": 16110 }, { "epoch": 0.75, "learning_rate": 1.2461876695668445e-05, "loss": 0.129, "step": 16115 }, { "epoch": 0.75, "learning_rate": 1.2459537842641969e-05, "loss": 0.1019, "step": 16120 }, { "epoch": 0.75, "learning_rate": 1.2457198989615493e-05, "loss": 0.112, "step": 16125 }, { "epoch": 0.75, "learning_rate": 1.2454860136589018e-05, "loss": 0.137, "step": 16130 }, { "epoch": 0.75, "learning_rate": 1.2452521283562542e-05, "loss": 0.0809, "step": 16135 }, { "epoch": 0.75, "learning_rate": 1.2450182430536067e-05, "loss": 0.0958, "step": 16140 }, { "epoch": 0.76, "learning_rate": 1.244784357750959e-05, "loss": 0.1466, "step": 16145 }, { "epoch": 0.76, "learning_rate": 1.2445504724483115e-05, "loss": 0.0729, "step": 16150 }, { "epoch": 0.76, "learning_rate": 1.244316587145664e-05, "loss": 0.0955, "step": 16155 }, { "epoch": 0.76, "learning_rate": 1.2440827018430163e-05, "loss": 0.1072, "step": 16160 }, { "epoch": 0.76, "learning_rate": 1.2438488165403686e-05, "loss": 0.0736, "step": 16165 }, { "epoch": 0.76, "learning_rate": 1.243614931237721e-05, "loss": 0.0987, "step": 16170 }, { "epoch": 0.76, "learning_rate": 1.2433810459350735e-05, "loss": 0.074, "step": 16175 }, { "epoch": 0.76, "learning_rate": 1.243147160632426e-05, "loss": 0.0846, "step": 16180 }, { "epoch": 0.76, "learning_rate": 1.2429132753297783e-05, "loss": 0.0602, "step": 16185 }, { "epoch": 0.76, "learning_rate": 1.2426793900271308e-05, "loss": 0.1266, "step": 16190 }, { "epoch": 0.76, "learning_rate": 1.2424455047244833e-05, "loss": 0.1038, "step": 16195 }, { "epoch": 0.76, "learning_rate": 1.2422116194218356e-05, "loss": 0.054, "step": 16200 }, { "epoch": 0.76, "learning_rate": 1.2419777341191881e-05, "loss": 0.0723, "step": 16205 }, { "epoch": 0.76, "learning_rate": 1.2417438488165406e-05, "loss": 0.0956, "step": 16210 }, { "epoch": 0.76, "learning_rate": 1.2415099635138929e-05, "loss": 0.0909, "step": 16215 }, { "epoch": 0.76, "learning_rate": 1.2412760782112454e-05, "loss": 0.0899, "step": 16220 }, { "epoch": 0.76, "learning_rate": 1.2410421929085976e-05, "loss": 0.0992, "step": 16225 }, { "epoch": 0.76, "learning_rate": 1.24080830760595e-05, "loss": 0.0771, "step": 16230 }, { "epoch": 0.76, "learning_rate": 1.2405744223033025e-05, "loss": 0.095, "step": 16235 }, { "epoch": 0.76, "learning_rate": 1.2403405370006549e-05, "loss": 0.0908, "step": 16240 }, { "epoch": 0.76, "learning_rate": 1.2401066516980074e-05, "loss": 0.0628, "step": 16245 }, { "epoch": 0.76, "learning_rate": 1.2398727663953599e-05, "loss": 0.0836, "step": 16250 }, { "epoch": 0.76, "learning_rate": 1.2396388810927122e-05, "loss": 0.135, "step": 16255 }, { "epoch": 0.76, "learning_rate": 1.2394049957900647e-05, "loss": 0.0608, "step": 16260 }, { "epoch": 0.76, "learning_rate": 1.2391711104874172e-05, "loss": 0.0932, "step": 16265 }, { "epoch": 0.76, "learning_rate": 1.2389372251847695e-05, "loss": 0.1032, "step": 16270 }, { "epoch": 0.76, "learning_rate": 1.238703339882122e-05, "loss": 0.1108, "step": 16275 }, { "epoch": 0.76, "learning_rate": 1.2384694545794745e-05, "loss": 0.0933, "step": 16280 }, { "epoch": 0.76, "learning_rate": 1.2382355692768267e-05, "loss": 0.1218, "step": 16285 }, { "epoch": 0.76, "learning_rate": 1.2380016839741791e-05, "loss": 0.0453, "step": 16290 }, { "epoch": 0.76, "learning_rate": 1.2377677986715315e-05, "loss": 0.1054, "step": 16295 }, { "epoch": 0.76, "learning_rate": 1.237533913368884e-05, "loss": 0.131, "step": 16300 }, { "epoch": 0.76, "learning_rate": 1.2373000280662365e-05, "loss": 0.1, "step": 16305 }, { "epoch": 0.76, "learning_rate": 1.2370661427635888e-05, "loss": 0.0777, "step": 16310 }, { "epoch": 0.76, "learning_rate": 1.2368322574609413e-05, "loss": 0.092, "step": 16315 }, { "epoch": 0.76, "learning_rate": 1.2365983721582938e-05, "loss": 0.0538, "step": 16320 }, { "epoch": 0.76, "learning_rate": 1.2363644868556461e-05, "loss": 0.07, "step": 16325 }, { "epoch": 0.76, "learning_rate": 1.2361306015529986e-05, "loss": 0.1153, "step": 16330 }, { "epoch": 0.76, "learning_rate": 1.2358967162503511e-05, "loss": 0.1016, "step": 16335 }, { "epoch": 0.76, "learning_rate": 1.2356628309477034e-05, "loss": 0.0881, "step": 16340 }, { "epoch": 0.76, "learning_rate": 1.2354289456450557e-05, "loss": 0.137, "step": 16345 }, { "epoch": 0.76, "learning_rate": 1.235195060342408e-05, "loss": 0.069, "step": 16350 }, { "epoch": 0.77, "learning_rate": 1.2349611750397606e-05, "loss": 0.0399, "step": 16355 }, { "epoch": 0.77, "learning_rate": 1.2347272897371129e-05, "loss": 0.0553, "step": 16360 }, { "epoch": 0.77, "learning_rate": 1.2344934044344654e-05, "loss": 0.1124, "step": 16365 }, { "epoch": 0.77, "learning_rate": 1.2342595191318179e-05, "loss": 0.0735, "step": 16370 }, { "epoch": 0.77, "learning_rate": 1.2340256338291702e-05, "loss": 0.1152, "step": 16375 }, { "epoch": 0.77, "learning_rate": 1.2337917485265227e-05, "loss": 0.107, "step": 16380 }, { "epoch": 0.77, "learning_rate": 1.2335578632238752e-05, "loss": 0.1569, "step": 16385 }, { "epoch": 0.77, "learning_rate": 1.2333239779212275e-05, "loss": 0.0879, "step": 16390 }, { "epoch": 0.77, "learning_rate": 1.23309009261858e-05, "loss": 0.125, "step": 16395 }, { "epoch": 0.77, "learning_rate": 1.2328562073159325e-05, "loss": 0.1202, "step": 16400 }, { "epoch": 0.77, "learning_rate": 1.2326223220132847e-05, "loss": 0.0975, "step": 16405 }, { "epoch": 0.77, "learning_rate": 1.2323884367106372e-05, "loss": 0.0405, "step": 16410 }, { "epoch": 0.77, "learning_rate": 1.2321545514079895e-05, "loss": 0.1115, "step": 16415 }, { "epoch": 0.77, "learning_rate": 1.231920666105342e-05, "loss": 0.0889, "step": 16420 }, { "epoch": 0.77, "learning_rate": 1.2316867808026945e-05, "loss": 0.1133, "step": 16425 }, { "epoch": 0.77, "learning_rate": 1.2314528955000468e-05, "loss": 0.0836, "step": 16430 }, { "epoch": 0.77, "learning_rate": 1.2312190101973993e-05, "loss": 0.0509, "step": 16435 }, { "epoch": 0.77, "learning_rate": 1.2309851248947518e-05, "loss": 0.0557, "step": 16440 }, { "epoch": 0.77, "learning_rate": 1.2307512395921041e-05, "loss": 0.1588, "step": 16445 }, { "epoch": 0.77, "learning_rate": 1.2305173542894566e-05, "loss": 0.0487, "step": 16450 }, { "epoch": 0.77, "learning_rate": 1.2302834689868091e-05, "loss": 0.0509, "step": 16455 }, { "epoch": 0.77, "learning_rate": 1.2300495836841614e-05, "loss": 0.1126, "step": 16460 }, { "epoch": 0.77, "learning_rate": 1.2298156983815138e-05, "loss": 0.0756, "step": 16465 }, { "epoch": 0.77, "learning_rate": 1.2295818130788661e-05, "loss": 0.0836, "step": 16470 }, { "epoch": 0.77, "learning_rate": 1.2293479277762186e-05, "loss": 0.1018, "step": 16475 }, { "epoch": 0.77, "learning_rate": 1.229114042473571e-05, "loss": 0.0678, "step": 16480 }, { "epoch": 0.77, "learning_rate": 1.2288801571709234e-05, "loss": 0.0688, "step": 16485 }, { "epoch": 0.77, "learning_rate": 1.2286462718682759e-05, "loss": 0.1116, "step": 16490 }, { "epoch": 0.77, "learning_rate": 1.2284123865656284e-05, "loss": 0.087, "step": 16495 }, { "epoch": 0.77, "learning_rate": 1.2281785012629807e-05, "loss": 0.1549, "step": 16500 }, { "epoch": 0.77, "learning_rate": 1.2279446159603332e-05, "loss": 0.1059, "step": 16505 }, { "epoch": 0.77, "learning_rate": 1.2277107306576857e-05, "loss": 0.0624, "step": 16510 }, { "epoch": 0.77, "learning_rate": 1.227476845355038e-05, "loss": 0.0775, "step": 16515 }, { "epoch": 0.77, "learning_rate": 1.2272429600523904e-05, "loss": 0.1172, "step": 16520 }, { "epoch": 0.77, "learning_rate": 1.2270090747497427e-05, "loss": 0.0913, "step": 16525 }, { "epoch": 0.77, "learning_rate": 1.2267751894470952e-05, "loss": 0.1598, "step": 16530 }, { "epoch": 0.77, "learning_rate": 1.2265413041444477e-05, "loss": 0.0868, "step": 16535 }, { "epoch": 0.77, "learning_rate": 1.2263074188418e-05, "loss": 0.0434, "step": 16540 }, { "epoch": 0.77, "learning_rate": 1.2260735335391525e-05, "loss": 0.0707, "step": 16545 }, { "epoch": 0.77, "learning_rate": 1.225839648236505e-05, "loss": 0.0858, "step": 16550 }, { "epoch": 0.77, "learning_rate": 1.2256057629338573e-05, "loss": 0.0517, "step": 16555 }, { "epoch": 0.77, "learning_rate": 1.2253718776312098e-05, "loss": 0.1294, "step": 16560 }, { "epoch": 0.77, "learning_rate": 1.2251379923285623e-05, "loss": 0.1305, "step": 16565 }, { "epoch": 0.78, "learning_rate": 1.2249041070259146e-05, "loss": 0.1064, "step": 16570 }, { "epoch": 0.78, "learning_rate": 1.2246702217232671e-05, "loss": 0.1235, "step": 16575 }, { "epoch": 0.78, "learning_rate": 1.2244363364206193e-05, "loss": 0.0719, "step": 16580 }, { "epoch": 0.78, "learning_rate": 1.2242024511179718e-05, "loss": 0.1028, "step": 16585 }, { "epoch": 0.78, "learning_rate": 1.2239685658153241e-05, "loss": 0.0763, "step": 16590 }, { "epoch": 0.78, "learning_rate": 1.2237346805126766e-05, "loss": 0.0808, "step": 16595 }, { "epoch": 0.78, "learning_rate": 1.2235007952100291e-05, "loss": 0.0645, "step": 16600 }, { "epoch": 0.78, "learning_rate": 1.2232669099073814e-05, "loss": 0.1425, "step": 16605 }, { "epoch": 0.78, "learning_rate": 1.223033024604734e-05, "loss": 0.0729, "step": 16610 }, { "epoch": 0.78, "learning_rate": 1.2227991393020864e-05, "loss": 0.0827, "step": 16615 }, { "epoch": 0.78, "learning_rate": 1.2225652539994387e-05, "loss": 0.1242, "step": 16620 }, { "epoch": 0.78, "learning_rate": 1.2223313686967912e-05, "loss": 0.0983, "step": 16625 }, { "epoch": 0.78, "learning_rate": 1.2220974833941437e-05, "loss": 0.1284, "step": 16630 }, { "epoch": 0.78, "learning_rate": 1.221863598091496e-05, "loss": 0.1084, "step": 16635 }, { "epoch": 0.78, "learning_rate": 1.2216297127888484e-05, "loss": 0.1063, "step": 16640 }, { "epoch": 0.78, "learning_rate": 1.2213958274862007e-05, "loss": 0.1032, "step": 16645 }, { "epoch": 0.78, "learning_rate": 1.2211619421835532e-05, "loss": 0.0981, "step": 16650 }, { "epoch": 0.78, "learning_rate": 1.2209280568809057e-05, "loss": 0.1094, "step": 16655 }, { "epoch": 0.78, "learning_rate": 1.220694171578258e-05, "loss": 0.0664, "step": 16660 }, { "epoch": 0.78, "learning_rate": 1.2204602862756105e-05, "loss": 0.0624, "step": 16665 }, { "epoch": 0.78, "learning_rate": 1.220226400972963e-05, "loss": 0.0953, "step": 16670 }, { "epoch": 0.78, "learning_rate": 1.2199925156703153e-05, "loss": 0.1645, "step": 16675 }, { "epoch": 0.78, "learning_rate": 1.2197586303676678e-05, "loss": 0.0719, "step": 16680 }, { "epoch": 0.78, "learning_rate": 1.2195247450650203e-05, "loss": 0.1032, "step": 16685 }, { "epoch": 0.78, "learning_rate": 1.2192908597623727e-05, "loss": 0.0633, "step": 16690 }, { "epoch": 0.78, "learning_rate": 1.2190569744597252e-05, "loss": 0.0898, "step": 16695 }, { "epoch": 0.78, "learning_rate": 1.2188230891570773e-05, "loss": 0.0377, "step": 16700 }, { "epoch": 0.78, "learning_rate": 1.2185892038544298e-05, "loss": 0.0559, "step": 16705 }, { "epoch": 0.78, "learning_rate": 1.2183553185517823e-05, "loss": 0.067, "step": 16710 }, { "epoch": 0.78, "learning_rate": 1.2181214332491346e-05, "loss": 0.0997, "step": 16715 }, { "epoch": 0.78, "learning_rate": 1.2178875479464871e-05, "loss": 0.1329, "step": 16720 }, { "epoch": 0.78, "learning_rate": 1.2176536626438396e-05, "loss": 0.0634, "step": 16725 }, { "epoch": 0.78, "learning_rate": 1.217419777341192e-05, "loss": 0.065, "step": 16730 }, { "epoch": 0.78, "learning_rate": 1.2171858920385444e-05, "loss": 0.0614, "step": 16735 }, { "epoch": 0.78, "learning_rate": 1.216952006735897e-05, "loss": 0.0864, "step": 16740 }, { "epoch": 0.78, "learning_rate": 1.2167181214332493e-05, "loss": 0.145, "step": 16745 }, { "epoch": 0.78, "learning_rate": 1.2164842361306018e-05, "loss": 0.0968, "step": 16750 }, { "epoch": 0.78, "learning_rate": 1.2162503508279543e-05, "loss": 0.102, "step": 16755 }, { "epoch": 0.78, "learning_rate": 1.2160164655253064e-05, "loss": 0.1029, "step": 16760 }, { "epoch": 0.78, "learning_rate": 1.2157825802226589e-05, "loss": 0.0538, "step": 16765 }, { "epoch": 0.78, "learning_rate": 1.2155486949200112e-05, "loss": 0.1237, "step": 16770 }, { "epoch": 0.78, "learning_rate": 1.2153148096173637e-05, "loss": 0.078, "step": 16775 }, { "epoch": 0.78, "learning_rate": 1.2150809243147162e-05, "loss": 0.092, "step": 16780 }, { "epoch": 0.79, "learning_rate": 1.2148470390120685e-05, "loss": 0.0628, "step": 16785 }, { "epoch": 0.79, "learning_rate": 1.214613153709421e-05, "loss": 0.0768, "step": 16790 }, { "epoch": 0.79, "learning_rate": 1.2143792684067734e-05, "loss": 0.0587, "step": 16795 }, { "epoch": 0.79, "learning_rate": 1.2141453831041259e-05, "loss": 0.11, "step": 16800 }, { "epoch": 0.79, "learning_rate": 1.2139114978014784e-05, "loss": 0.1538, "step": 16805 }, { "epoch": 0.79, "learning_rate": 1.2136776124988307e-05, "loss": 0.1048, "step": 16810 }, { "epoch": 0.79, "learning_rate": 1.2134437271961832e-05, "loss": 0.0875, "step": 16815 }, { "epoch": 0.79, "learning_rate": 1.2132098418935353e-05, "loss": 0.1, "step": 16820 }, { "epoch": 0.79, "learning_rate": 1.2129759565908878e-05, "loss": 0.081, "step": 16825 }, { "epoch": 0.79, "learning_rate": 1.2127420712882403e-05, "loss": 0.0983, "step": 16830 }, { "epoch": 0.79, "learning_rate": 1.2125081859855926e-05, "loss": 0.0635, "step": 16835 }, { "epoch": 0.79, "learning_rate": 1.2122743006829451e-05, "loss": 0.0833, "step": 16840 }, { "epoch": 0.79, "learning_rate": 1.2120404153802976e-05, "loss": 0.125, "step": 16845 }, { "epoch": 0.79, "learning_rate": 1.21180653007765e-05, "loss": 0.0836, "step": 16850 }, { "epoch": 0.79, "learning_rate": 1.2115726447750025e-05, "loss": 0.0601, "step": 16855 }, { "epoch": 0.79, "learning_rate": 1.211338759472355e-05, "loss": 0.1336, "step": 16860 }, { "epoch": 0.79, "learning_rate": 1.2111048741697073e-05, "loss": 0.0321, "step": 16865 }, { "epoch": 0.79, "learning_rate": 1.2108709888670598e-05, "loss": 0.1419, "step": 16870 }, { "epoch": 0.79, "learning_rate": 1.2106371035644123e-05, "loss": 0.1425, "step": 16875 }, { "epoch": 0.79, "learning_rate": 1.2104032182617644e-05, "loss": 0.1204, "step": 16880 }, { "epoch": 0.79, "learning_rate": 1.210169332959117e-05, "loss": 0.1104, "step": 16885 }, { "epoch": 0.79, "learning_rate": 1.2099354476564692e-05, "loss": 0.0978, "step": 16890 }, { "epoch": 0.79, "learning_rate": 1.2097015623538217e-05, "loss": 0.0854, "step": 16895 }, { "epoch": 0.79, "learning_rate": 1.2094676770511742e-05, "loss": 0.1137, "step": 16900 }, { "epoch": 0.79, "learning_rate": 1.2092337917485266e-05, "loss": 0.1093, "step": 16905 }, { "epoch": 0.79, "learning_rate": 1.208999906445879e-05, "loss": 0.0797, "step": 16910 }, { "epoch": 0.79, "learning_rate": 1.2087660211432316e-05, "loss": 0.0456, "step": 16915 }, { "epoch": 0.79, "learning_rate": 1.2085321358405839e-05, "loss": 0.0975, "step": 16920 }, { "epoch": 0.79, "learning_rate": 1.2082982505379364e-05, "loss": 0.1036, "step": 16925 }, { "epoch": 0.79, "learning_rate": 1.2080643652352889e-05, "loss": 0.0533, "step": 16930 }, { "epoch": 0.79, "learning_rate": 1.2078304799326412e-05, "loss": 0.083, "step": 16935 }, { "epoch": 0.79, "learning_rate": 1.2075965946299935e-05, "loss": 0.1256, "step": 16940 }, { "epoch": 0.79, "learning_rate": 1.2073627093273458e-05, "loss": 0.0609, "step": 16945 }, { "epoch": 0.79, "learning_rate": 1.2071288240246983e-05, "loss": 0.0581, "step": 16950 }, { "epoch": 0.79, "learning_rate": 1.2068949387220508e-05, "loss": 0.1125, "step": 16955 }, { "epoch": 0.79, "learning_rate": 1.2066610534194032e-05, "loss": 0.0965, "step": 16960 }, { "epoch": 0.79, "learning_rate": 1.2064271681167557e-05, "loss": 0.1401, "step": 16965 }, { "epoch": 0.79, "learning_rate": 1.2061932828141082e-05, "loss": 0.1243, "step": 16970 }, { "epoch": 0.79, "learning_rate": 1.2059593975114605e-05, "loss": 0.098, "step": 16975 }, { "epoch": 0.79, "learning_rate": 1.205725512208813e-05, "loss": 0.1175, "step": 16980 }, { "epoch": 0.79, "learning_rate": 1.2054916269061655e-05, "loss": 0.0678, "step": 16985 }, { "epoch": 0.79, "learning_rate": 1.2052577416035178e-05, "loss": 0.1041, "step": 16990 }, { "epoch": 0.79, "learning_rate": 1.2050238563008701e-05, "loss": 0.1601, "step": 16995 }, { "epoch": 0.8, "learning_rate": 1.2047899709982224e-05, "loss": 0.0941, "step": 17000 }, { "epoch": 0.8, "learning_rate": 1.204556085695575e-05, "loss": 0.1031, "step": 17005 }, { "epoch": 0.8, "learning_rate": 1.2043222003929274e-05, "loss": 0.1261, "step": 17010 }, { "epoch": 0.8, "learning_rate": 1.2040883150902798e-05, "loss": 0.1136, "step": 17015 }, { "epoch": 0.8, "learning_rate": 1.2038544297876323e-05, "loss": 0.1198, "step": 17020 }, { "epoch": 0.8, "learning_rate": 1.2036205444849846e-05, "loss": 0.0859, "step": 17025 }, { "epoch": 0.8, "learning_rate": 1.203386659182337e-05, "loss": 0.1094, "step": 17030 }, { "epoch": 0.8, "learning_rate": 1.2031527738796896e-05, "loss": 0.0737, "step": 17035 }, { "epoch": 0.8, "learning_rate": 1.2029188885770419e-05, "loss": 0.0744, "step": 17040 }, { "epoch": 0.8, "learning_rate": 1.2026850032743944e-05, "loss": 0.1232, "step": 17045 }, { "epoch": 0.8, "learning_rate": 1.2024511179717469e-05, "loss": 0.0652, "step": 17050 }, { "epoch": 0.8, "learning_rate": 1.202217232669099e-05, "loss": 0.0661, "step": 17055 }, { "epoch": 0.8, "learning_rate": 1.2019833473664515e-05, "loss": 0.0939, "step": 17060 }, { "epoch": 0.8, "learning_rate": 1.2017494620638039e-05, "loss": 0.0564, "step": 17065 }, { "epoch": 0.8, "learning_rate": 1.2015155767611564e-05, "loss": 0.1114, "step": 17070 }, { "epoch": 0.8, "learning_rate": 1.2012816914585089e-05, "loss": 0.1201, "step": 17075 }, { "epoch": 0.8, "learning_rate": 1.2010478061558612e-05, "loss": 0.0744, "step": 17080 }, { "epoch": 0.8, "learning_rate": 1.2008139208532137e-05, "loss": 0.067, "step": 17085 }, { "epoch": 0.8, "learning_rate": 1.2005800355505662e-05, "loss": 0.0986, "step": 17090 }, { "epoch": 0.8, "learning_rate": 1.2003461502479185e-05, "loss": 0.0718, "step": 17095 }, { "epoch": 0.8, "learning_rate": 1.200112264945271e-05, "loss": 0.1098, "step": 17100 }, { "epoch": 0.8, "learning_rate": 1.1998783796426235e-05, "loss": 0.1451, "step": 17105 }, { "epoch": 0.8, "learning_rate": 1.1996444943399758e-05, "loss": 0.0771, "step": 17110 }, { "epoch": 0.8, "learning_rate": 1.1994106090373281e-05, "loss": 0.1533, "step": 17115 }, { "epoch": 0.8, "learning_rate": 1.1991767237346805e-05, "loss": 0.0934, "step": 17120 }, { "epoch": 0.8, "learning_rate": 1.198942838432033e-05, "loss": 0.1294, "step": 17125 }, { "epoch": 0.8, "learning_rate": 1.1987089531293855e-05, "loss": 0.1363, "step": 17130 }, { "epoch": 0.8, "learning_rate": 1.1984750678267378e-05, "loss": 0.0637, "step": 17135 }, { "epoch": 0.8, "learning_rate": 1.1982411825240903e-05, "loss": 0.0324, "step": 17140 }, { "epoch": 0.8, "learning_rate": 1.1980072972214428e-05, "loss": 0.1105, "step": 17145 }, { "epoch": 0.8, "learning_rate": 1.1977734119187951e-05, "loss": 0.1072, "step": 17150 }, { "epoch": 0.8, "learning_rate": 1.1975395266161476e-05, "loss": 0.0754, "step": 17155 }, { "epoch": 0.8, "learning_rate": 1.1973056413135001e-05, "loss": 0.1054, "step": 17160 }, { "epoch": 0.8, "learning_rate": 1.1970717560108524e-05, "loss": 0.1608, "step": 17165 }, { "epoch": 0.8, "learning_rate": 1.1968378707082049e-05, "loss": 0.144, "step": 17170 }, { "epoch": 0.8, "learning_rate": 1.196603985405557e-05, "loss": 0.0959, "step": 17175 }, { "epoch": 0.8, "learning_rate": 1.1963701001029096e-05, "loss": 0.1044, "step": 17180 }, { "epoch": 0.8, "learning_rate": 1.196136214800262e-05, "loss": 0.1153, "step": 17185 }, { "epoch": 0.8, "learning_rate": 1.1959023294976144e-05, "loss": 0.1168, "step": 17190 }, { "epoch": 0.8, "learning_rate": 1.1956684441949669e-05, "loss": 0.1351, "step": 17195 }, { "epoch": 0.8, "learning_rate": 1.1954345588923194e-05, "loss": 0.0746, "step": 17200 }, { "epoch": 0.8, "learning_rate": 1.1952006735896717e-05, "loss": 0.0909, "step": 17205 }, { "epoch": 0.81, "learning_rate": 1.1949667882870242e-05, "loss": 0.0341, "step": 17210 }, { "epoch": 0.81, "learning_rate": 1.1947329029843767e-05, "loss": 0.1577, "step": 17215 }, { "epoch": 0.81, "learning_rate": 1.194499017681729e-05, "loss": 0.1064, "step": 17220 }, { "epoch": 0.81, "learning_rate": 1.1942651323790815e-05, "loss": 0.1162, "step": 17225 }, { "epoch": 0.81, "learning_rate": 1.1940312470764338e-05, "loss": 0.0766, "step": 17230 }, { "epoch": 0.81, "learning_rate": 1.1937973617737862e-05, "loss": 0.0493, "step": 17235 }, { "epoch": 0.81, "learning_rate": 1.1935634764711385e-05, "loss": 0.0763, "step": 17240 }, { "epoch": 0.81, "learning_rate": 1.193329591168491e-05, "loss": 0.0569, "step": 17245 }, { "epoch": 0.81, "learning_rate": 1.1930957058658435e-05, "loss": 0.148, "step": 17250 }, { "epoch": 0.81, "learning_rate": 1.1928618205631958e-05, "loss": 0.0846, "step": 17255 }, { "epoch": 0.81, "learning_rate": 1.1926279352605483e-05, "loss": 0.0475, "step": 17260 }, { "epoch": 0.81, "learning_rate": 1.1923940499579008e-05, "loss": 0.0736, "step": 17265 }, { "epoch": 0.81, "learning_rate": 1.1921601646552531e-05, "loss": 0.0955, "step": 17270 }, { "epoch": 0.81, "learning_rate": 1.1919262793526056e-05, "loss": 0.0728, "step": 17275 }, { "epoch": 0.81, "learning_rate": 1.1916923940499581e-05, "loss": 0.1016, "step": 17280 }, { "epoch": 0.81, "learning_rate": 1.1914585087473104e-05, "loss": 0.075, "step": 17285 }, { "epoch": 0.81, "learning_rate": 1.191224623444663e-05, "loss": 0.1026, "step": 17290 }, { "epoch": 0.81, "learning_rate": 1.1909907381420151e-05, "loss": 0.1353, "step": 17295 }, { "epoch": 0.81, "learning_rate": 1.1907568528393676e-05, "loss": 0.1296, "step": 17300 }, { "epoch": 0.81, "learning_rate": 1.19052296753672e-05, "loss": 0.1029, "step": 17305 }, { "epoch": 0.81, "learning_rate": 1.1902890822340724e-05, "loss": 0.1318, "step": 17310 }, { "epoch": 0.81, "learning_rate": 1.1900551969314249e-05, "loss": 0.0923, "step": 17315 }, { "epoch": 0.81, "learning_rate": 1.1898213116287774e-05, "loss": 0.0964, "step": 17320 }, { "epoch": 0.81, "learning_rate": 1.1895874263261297e-05, "loss": 0.059, "step": 17325 }, { "epoch": 0.81, "learning_rate": 1.1893535410234822e-05, "loss": 0.1162, "step": 17330 }, { "epoch": 0.81, "learning_rate": 1.1891196557208347e-05, "loss": 0.169, "step": 17335 }, { "epoch": 0.81, "learning_rate": 1.188885770418187e-05, "loss": 0.0861, "step": 17340 }, { "epoch": 0.81, "learning_rate": 1.1886518851155395e-05, "loss": 0.0979, "step": 17345 }, { "epoch": 0.81, "learning_rate": 1.188417999812892e-05, "loss": 0.1305, "step": 17350 }, { "epoch": 0.81, "learning_rate": 1.1881841145102442e-05, "loss": 0.0991, "step": 17355 }, { "epoch": 0.81, "learning_rate": 1.1879502292075967e-05, "loss": 0.0835, "step": 17360 }, { "epoch": 0.81, "learning_rate": 1.187716343904949e-05, "loss": 0.1228, "step": 17365 }, { "epoch": 0.81, "learning_rate": 1.1874824586023015e-05, "loss": 0.1021, "step": 17370 }, { "epoch": 0.81, "learning_rate": 1.187248573299654e-05, "loss": 0.136, "step": 17375 }, { "epoch": 0.81, "learning_rate": 1.1870146879970063e-05, "loss": 0.0863, "step": 17380 }, { "epoch": 0.81, "learning_rate": 1.1867808026943588e-05, "loss": 0.0785, "step": 17385 }, { "epoch": 0.81, "learning_rate": 1.1865469173917113e-05, "loss": 0.0998, "step": 17390 }, { "epoch": 0.81, "learning_rate": 1.1863130320890636e-05, "loss": 0.1124, "step": 17395 }, { "epoch": 0.81, "learning_rate": 1.1860791467864161e-05, "loss": 0.0651, "step": 17400 }, { "epoch": 0.81, "learning_rate": 1.1858452614837686e-05, "loss": 0.0333, "step": 17405 }, { "epoch": 0.81, "learning_rate": 1.1856113761811208e-05, "loss": 0.1088, "step": 17410 }, { "epoch": 0.81, "learning_rate": 1.1853774908784733e-05, "loss": 0.0674, "step": 17415 }, { "epoch": 0.81, "learning_rate": 1.1851436055758256e-05, "loss": 0.0643, "step": 17420 }, { "epoch": 0.82, "learning_rate": 1.1849097202731781e-05, "loss": 0.0564, "step": 17425 }, { "epoch": 0.82, "learning_rate": 1.1846758349705306e-05, "loss": 0.1513, "step": 17430 }, { "epoch": 0.82, "learning_rate": 1.184441949667883e-05, "loss": 0.0865, "step": 17435 }, { "epoch": 0.82, "learning_rate": 1.1842080643652354e-05, "loss": 0.066, "step": 17440 }, { "epoch": 0.82, "learning_rate": 1.1839741790625879e-05, "loss": 0.1179, "step": 17445 }, { "epoch": 0.82, "learning_rate": 1.1837402937599402e-05, "loss": 0.0858, "step": 17450 }, { "epoch": 0.82, "learning_rate": 1.1835064084572927e-05, "loss": 0.0816, "step": 17455 }, { "epoch": 0.82, "learning_rate": 1.183272523154645e-05, "loss": 0.0527, "step": 17460 }, { "epoch": 0.82, "learning_rate": 1.1830386378519976e-05, "loss": 0.0651, "step": 17465 }, { "epoch": 0.82, "learning_rate": 1.1828047525493497e-05, "loss": 0.0854, "step": 17470 }, { "epoch": 0.82, "learning_rate": 1.1825708672467022e-05, "loss": 0.1057, "step": 17475 }, { "epoch": 0.82, "learning_rate": 1.1823369819440547e-05, "loss": 0.0915, "step": 17480 }, { "epoch": 0.82, "learning_rate": 1.182103096641407e-05, "loss": 0.0837, "step": 17485 }, { "epoch": 0.82, "learning_rate": 1.1818692113387595e-05, "loss": 0.1236, "step": 17490 }, { "epoch": 0.82, "learning_rate": 1.181635326036112e-05, "loss": 0.0855, "step": 17495 }, { "epoch": 0.82, "learning_rate": 1.1814014407334643e-05, "loss": 0.1257, "step": 17500 }, { "epoch": 0.82, "learning_rate": 1.1811675554308168e-05, "loss": 0.1278, "step": 17505 }, { "epoch": 0.82, "learning_rate": 1.1809336701281693e-05, "loss": 0.0765, "step": 17510 }, { "epoch": 0.82, "learning_rate": 1.1806997848255217e-05, "loss": 0.0633, "step": 17515 }, { "epoch": 0.82, "learning_rate": 1.1804658995228742e-05, "loss": 0.1346, "step": 17520 }, { "epoch": 0.82, "learning_rate": 1.1802320142202266e-05, "loss": 0.1128, "step": 17525 }, { "epoch": 0.82, "learning_rate": 1.1799981289175788e-05, "loss": 0.0907, "step": 17530 }, { "epoch": 0.82, "learning_rate": 1.1797642436149313e-05, "loss": 0.118, "step": 17535 }, { "epoch": 0.82, "learning_rate": 1.1795303583122836e-05, "loss": 0.1187, "step": 17540 }, { "epoch": 0.82, "learning_rate": 1.1792964730096361e-05, "loss": 0.0908, "step": 17545 }, { "epoch": 0.82, "learning_rate": 1.1790625877069886e-05, "loss": 0.1092, "step": 17550 }, { "epoch": 0.82, "learning_rate": 1.178828702404341e-05, "loss": 0.1153, "step": 17555 }, { "epoch": 0.82, "learning_rate": 1.1785948171016934e-05, "loss": 0.0434, "step": 17560 }, { "epoch": 0.82, "learning_rate": 1.178360931799046e-05, "loss": 0.0967, "step": 17565 }, { "epoch": 0.82, "learning_rate": 1.1781270464963983e-05, "loss": 0.0984, "step": 17570 }, { "epoch": 0.82, "learning_rate": 1.1778931611937508e-05, "loss": 0.1308, "step": 17575 }, { "epoch": 0.82, "learning_rate": 1.1776592758911032e-05, "loss": 0.1017, "step": 17580 }, { "epoch": 0.82, "learning_rate": 1.1774253905884556e-05, "loss": 0.1519, "step": 17585 }, { "epoch": 0.82, "learning_rate": 1.1771915052858079e-05, "loss": 0.1084, "step": 17590 }, { "epoch": 0.82, "learning_rate": 1.1769576199831602e-05, "loss": 0.1135, "step": 17595 }, { "epoch": 0.82, "learning_rate": 1.1767237346805127e-05, "loss": 0.0645, "step": 17600 }, { "epoch": 0.82, "learning_rate": 1.1764898493778652e-05, "loss": 0.1167, "step": 17605 }, { "epoch": 0.82, "learning_rate": 1.1762559640752175e-05, "loss": 0.0826, "step": 17610 }, { "epoch": 0.82, "learning_rate": 1.17602207877257e-05, "loss": 0.1062, "step": 17615 }, { "epoch": 0.82, "learning_rate": 1.1757881934699225e-05, "loss": 0.0544, "step": 17620 }, { "epoch": 0.82, "learning_rate": 1.1755543081672749e-05, "loss": 0.0914, "step": 17625 }, { "epoch": 0.82, "learning_rate": 1.1753204228646273e-05, "loss": 0.0827, "step": 17630 }, { "epoch": 0.82, "learning_rate": 1.1750865375619798e-05, "loss": 0.0706, "step": 17635 }, { "epoch": 0.83, "learning_rate": 1.1748526522593322e-05, "loss": 0.0989, "step": 17640 }, { "epoch": 0.83, "learning_rate": 1.1746187669566847e-05, "loss": 0.0917, "step": 17645 }, { "epoch": 0.83, "learning_rate": 1.1743848816540368e-05, "loss": 0.1388, "step": 17650 }, { "epoch": 0.83, "learning_rate": 1.1741509963513893e-05, "loss": 0.1618, "step": 17655 }, { "epoch": 0.83, "learning_rate": 1.1739171110487418e-05, "loss": 0.0714, "step": 17660 }, { "epoch": 0.83, "learning_rate": 1.1736832257460941e-05, "loss": 0.1154, "step": 17665 }, { "epoch": 0.83, "learning_rate": 1.1734493404434466e-05, "loss": 0.073, "step": 17670 }, { "epoch": 0.83, "learning_rate": 1.173215455140799e-05, "loss": 0.1598, "step": 17675 }, { "epoch": 0.83, "learning_rate": 1.1729815698381515e-05, "loss": 0.075, "step": 17680 }, { "epoch": 0.83, "learning_rate": 1.172747684535504e-05, "loss": 0.0766, "step": 17685 }, { "epoch": 0.83, "learning_rate": 1.1725137992328563e-05, "loss": 0.0532, "step": 17690 }, { "epoch": 0.83, "learning_rate": 1.1722799139302088e-05, "loss": 0.0703, "step": 17695 }, { "epoch": 0.83, "learning_rate": 1.1720460286275613e-05, "loss": 0.0772, "step": 17700 }, { "epoch": 0.83, "learning_rate": 1.1718121433249136e-05, "loss": 0.0675, "step": 17705 }, { "epoch": 0.83, "learning_rate": 1.171578258022266e-05, "loss": 0.0577, "step": 17710 }, { "epoch": 0.83, "learning_rate": 1.1713443727196182e-05, "loss": 0.1198, "step": 17715 }, { "epoch": 0.83, "learning_rate": 1.1711104874169707e-05, "loss": 0.0862, "step": 17720 }, { "epoch": 0.83, "learning_rate": 1.1708766021143232e-05, "loss": 0.0748, "step": 17725 }, { "epoch": 0.83, "learning_rate": 1.1706427168116756e-05, "loss": 0.1188, "step": 17730 }, { "epoch": 0.83, "learning_rate": 1.170408831509028e-05, "loss": 0.1287, "step": 17735 }, { "epoch": 0.83, "learning_rate": 1.1701749462063805e-05, "loss": 0.0285, "step": 17740 }, { "epoch": 0.83, "learning_rate": 1.1699410609037329e-05, "loss": 0.1117, "step": 17745 }, { "epoch": 0.83, "learning_rate": 1.1697071756010854e-05, "loss": 0.0452, "step": 17750 }, { "epoch": 0.83, "learning_rate": 1.1694732902984379e-05, "loss": 0.1536, "step": 17755 }, { "epoch": 0.83, "learning_rate": 1.1692394049957902e-05, "loss": 0.0504, "step": 17760 }, { "epoch": 0.83, "learning_rate": 1.1690055196931427e-05, "loss": 0.1118, "step": 17765 }, { "epoch": 0.83, "learning_rate": 1.1687716343904948e-05, "loss": 0.1358, "step": 17770 }, { "epoch": 0.83, "learning_rate": 1.1685377490878473e-05, "loss": 0.0597, "step": 17775 }, { "epoch": 0.83, "learning_rate": 1.1683038637851998e-05, "loss": 0.0897, "step": 17780 }, { "epoch": 0.83, "learning_rate": 1.1680699784825522e-05, "loss": 0.098, "step": 17785 }, { "epoch": 0.83, "learning_rate": 1.1678360931799047e-05, "loss": 0.0873, "step": 17790 }, { "epoch": 0.83, "learning_rate": 1.1676022078772571e-05, "loss": 0.1082, "step": 17795 }, { "epoch": 0.83, "learning_rate": 1.1673683225746095e-05, "loss": 0.0672, "step": 17800 }, { "epoch": 0.83, "learning_rate": 1.167134437271962e-05, "loss": 0.1282, "step": 17805 }, { "epoch": 0.83, "learning_rate": 1.1669005519693145e-05, "loss": 0.0417, "step": 17810 }, { "epoch": 0.83, "learning_rate": 1.1666666666666668e-05, "loss": 0.1137, "step": 17815 }, { "epoch": 0.83, "learning_rate": 1.1664327813640193e-05, "loss": 0.0703, "step": 17820 }, { "epoch": 0.83, "learning_rate": 1.1661988960613714e-05, "loss": 0.0909, "step": 17825 }, { "epoch": 0.83, "learning_rate": 1.165965010758724e-05, "loss": 0.1064, "step": 17830 }, { "epoch": 0.83, "learning_rate": 1.1657311254560764e-05, "loss": 0.0788, "step": 17835 }, { "epoch": 0.83, "learning_rate": 1.1654972401534288e-05, "loss": 0.1134, "step": 17840 }, { "epoch": 0.83, "learning_rate": 1.1652633548507813e-05, "loss": 0.1218, "step": 17845 }, { "epoch": 0.83, "learning_rate": 1.1650294695481337e-05, "loss": 0.0777, "step": 17850 }, { "epoch": 0.84, "learning_rate": 1.164795584245486e-05, "loss": 0.1106, "step": 17855 }, { "epoch": 0.84, "learning_rate": 1.1645616989428386e-05, "loss": 0.0465, "step": 17860 }, { "epoch": 0.84, "learning_rate": 1.164327813640191e-05, "loss": 0.0413, "step": 17865 }, { "epoch": 0.84, "learning_rate": 1.1640939283375434e-05, "loss": 0.137, "step": 17870 }, { "epoch": 0.84, "learning_rate": 1.1638600430348959e-05, "loss": 0.0815, "step": 17875 }, { "epoch": 0.84, "learning_rate": 1.1636261577322484e-05, "loss": 0.153, "step": 17880 }, { "epoch": 0.84, "learning_rate": 1.1633922724296005e-05, "loss": 0.0902, "step": 17885 }, { "epoch": 0.84, "learning_rate": 1.163158387126953e-05, "loss": 0.0789, "step": 17890 }, { "epoch": 0.84, "learning_rate": 1.1629245018243054e-05, "loss": 0.1073, "step": 17895 }, { "epoch": 0.84, "learning_rate": 1.1626906165216579e-05, "loss": 0.0686, "step": 17900 }, { "epoch": 0.84, "learning_rate": 1.1624567312190102e-05, "loss": 0.0974, "step": 17905 }, { "epoch": 0.84, "learning_rate": 1.1622228459163627e-05, "loss": 0.0895, "step": 17910 }, { "epoch": 0.84, "learning_rate": 1.1619889606137152e-05, "loss": 0.171, "step": 17915 }, { "epoch": 0.84, "learning_rate": 1.1617550753110675e-05, "loss": 0.1088, "step": 17920 }, { "epoch": 0.84, "learning_rate": 1.16152119000842e-05, "loss": 0.048, "step": 17925 }, { "epoch": 0.84, "learning_rate": 1.1612873047057725e-05, "loss": 0.0825, "step": 17930 }, { "epoch": 0.84, "learning_rate": 1.1610534194031248e-05, "loss": 0.0995, "step": 17935 }, { "epoch": 0.84, "learning_rate": 1.1608195341004773e-05, "loss": 0.0833, "step": 17940 }, { "epoch": 0.84, "learning_rate": 1.1605856487978295e-05, "loss": 0.0687, "step": 17945 }, { "epoch": 0.84, "learning_rate": 1.160351763495182e-05, "loss": 0.0588, "step": 17950 }, { "epoch": 0.84, "learning_rate": 1.1601178781925345e-05, "loss": 0.0813, "step": 17955 }, { "epoch": 0.84, "learning_rate": 1.1598839928898868e-05, "loss": 0.1216, "step": 17960 }, { "epoch": 0.84, "learning_rate": 1.1596501075872393e-05, "loss": 0.0565, "step": 17965 }, { "epoch": 0.84, "learning_rate": 1.1594162222845918e-05, "loss": 0.0737, "step": 17970 }, { "epoch": 0.84, "learning_rate": 1.1591823369819441e-05, "loss": 0.0551, "step": 17975 }, { "epoch": 0.84, "learning_rate": 1.1589484516792966e-05, "loss": 0.1595, "step": 17980 }, { "epoch": 0.84, "learning_rate": 1.158714566376649e-05, "loss": 0.0919, "step": 17985 }, { "epoch": 0.84, "learning_rate": 1.1584806810740014e-05, "loss": 0.0785, "step": 17990 }, { "epoch": 0.84, "learning_rate": 1.1582467957713539e-05, "loss": 0.1223, "step": 17995 }, { "epoch": 0.84, "learning_rate": 1.1580129104687064e-05, "loss": 0.0646, "step": 18000 }, { "epoch": 0.84, "learning_rate": 1.1577790251660586e-05, "loss": 0.1309, "step": 18005 }, { "epoch": 0.84, "learning_rate": 1.157545139863411e-05, "loss": 0.077, "step": 18010 }, { "epoch": 0.84, "learning_rate": 1.1573112545607634e-05, "loss": 0.0637, "step": 18015 }, { "epoch": 0.84, "learning_rate": 1.1570773692581159e-05, "loss": 0.1298, "step": 18020 }, { "epoch": 0.84, "learning_rate": 1.1568434839554684e-05, "loss": 0.0522, "step": 18025 }, { "epoch": 0.84, "learning_rate": 1.1566095986528207e-05, "loss": 0.0729, "step": 18030 }, { "epoch": 0.84, "learning_rate": 1.1563757133501732e-05, "loss": 0.0798, "step": 18035 }, { "epoch": 0.84, "learning_rate": 1.1561418280475257e-05, "loss": 0.0754, "step": 18040 }, { "epoch": 0.84, "learning_rate": 1.155907942744878e-05, "loss": 0.0977, "step": 18045 }, { "epoch": 0.84, "learning_rate": 1.1556740574422305e-05, "loss": 0.1174, "step": 18050 }, { "epoch": 0.84, "learning_rate": 1.155440172139583e-05, "loss": 0.096, "step": 18055 }, { "epoch": 0.84, "learning_rate": 1.1552062868369353e-05, "loss": 0.175, "step": 18060 }, { "epoch": 0.85, "learning_rate": 1.1549724015342877e-05, "loss": 0.0709, "step": 18065 }, { "epoch": 0.85, "learning_rate": 1.15473851623164e-05, "loss": 0.1569, "step": 18070 }, { "epoch": 0.85, "learning_rate": 1.1545046309289925e-05, "loss": 0.1221, "step": 18075 }, { "epoch": 0.85, "learning_rate": 1.154270745626345e-05, "loss": 0.0778, "step": 18080 }, { "epoch": 0.85, "learning_rate": 1.1540368603236973e-05, "loss": 0.1624, "step": 18085 }, { "epoch": 0.85, "learning_rate": 1.1538029750210498e-05, "loss": 0.0582, "step": 18090 }, { "epoch": 0.85, "learning_rate": 1.1535690897184023e-05, "loss": 0.0787, "step": 18095 }, { "epoch": 0.85, "learning_rate": 1.1533352044157546e-05, "loss": 0.085, "step": 18100 }, { "epoch": 0.85, "learning_rate": 1.1531013191131071e-05, "loss": 0.0655, "step": 18105 }, { "epoch": 0.85, "learning_rate": 1.1528674338104594e-05, "loss": 0.0709, "step": 18110 }, { "epoch": 0.85, "learning_rate": 1.152633548507812e-05, "loss": 0.0871, "step": 18115 }, { "epoch": 0.85, "learning_rate": 1.1523996632051644e-05, "loss": 0.1151, "step": 18120 }, { "epoch": 0.85, "learning_rate": 1.1521657779025166e-05, "loss": 0.0445, "step": 18125 }, { "epoch": 0.85, "learning_rate": 1.151931892599869e-05, "loss": 0.1249, "step": 18130 }, { "epoch": 0.85, "learning_rate": 1.1516980072972214e-05, "loss": 0.151, "step": 18135 }, { "epoch": 0.85, "learning_rate": 1.1514641219945739e-05, "loss": 0.0795, "step": 18140 }, { "epoch": 0.85, "learning_rate": 1.1512302366919264e-05, "loss": 0.1177, "step": 18145 }, { "epoch": 0.85, "learning_rate": 1.1509963513892787e-05, "loss": 0.082, "step": 18150 }, { "epoch": 0.85, "learning_rate": 1.1507624660866312e-05, "loss": 0.0985, "step": 18155 }, { "epoch": 0.85, "learning_rate": 1.1505285807839837e-05, "loss": 0.1144, "step": 18160 }, { "epoch": 0.85, "learning_rate": 1.150294695481336e-05, "loss": 0.112, "step": 18165 }, { "epoch": 0.85, "learning_rate": 1.1500608101786885e-05, "loss": 0.091, "step": 18170 }, { "epoch": 0.85, "learning_rate": 1.149826924876041e-05, "loss": 0.052, "step": 18175 }, { "epoch": 0.85, "learning_rate": 1.1495930395733933e-05, "loss": 0.1055, "step": 18180 }, { "epoch": 0.85, "learning_rate": 1.1493591542707457e-05, "loss": 0.1473, "step": 18185 }, { "epoch": 0.85, "learning_rate": 1.149125268968098e-05, "loss": 0.1194, "step": 18190 }, { "epoch": 0.85, "learning_rate": 1.1488913836654505e-05, "loss": 0.0574, "step": 18195 }, { "epoch": 0.85, "learning_rate": 1.148657498362803e-05, "loss": 0.1079, "step": 18200 }, { "epoch": 0.85, "learning_rate": 1.1484236130601553e-05, "loss": 0.0901, "step": 18205 }, { "epoch": 0.85, "learning_rate": 1.1481897277575078e-05, "loss": 0.0949, "step": 18210 }, { "epoch": 0.85, "learning_rate": 1.1479558424548603e-05, "loss": 0.2026, "step": 18215 }, { "epoch": 0.85, "learning_rate": 1.1477219571522126e-05, "loss": 0.0613, "step": 18220 }, { "epoch": 0.85, "learning_rate": 1.1474880718495651e-05, "loss": 0.1269, "step": 18225 }, { "epoch": 0.85, "learning_rate": 1.1472541865469176e-05, "loss": 0.0728, "step": 18230 }, { "epoch": 0.85, "learning_rate": 1.14702030124427e-05, "loss": 0.0831, "step": 18235 }, { "epoch": 0.85, "learning_rate": 1.1467864159416224e-05, "loss": 0.0576, "step": 18240 }, { "epoch": 0.85, "learning_rate": 1.1465525306389746e-05, "loss": 0.1411, "step": 18245 }, { "epoch": 0.85, "learning_rate": 1.1463186453363271e-05, "loss": 0.0644, "step": 18250 }, { "epoch": 0.85, "learning_rate": 1.1460847600336796e-05, "loss": 0.1219, "step": 18255 }, { "epoch": 0.85, "learning_rate": 1.1458508747310319e-05, "loss": 0.1079, "step": 18260 }, { "epoch": 0.85, "learning_rate": 1.1456169894283844e-05, "loss": 0.0572, "step": 18265 }, { "epoch": 0.85, "learning_rate": 1.1453831041257369e-05, "loss": 0.1056, "step": 18270 }, { "epoch": 0.85, "learning_rate": 1.1451492188230892e-05, "loss": 0.0855, "step": 18275 }, { "epoch": 0.86, "learning_rate": 1.1449153335204417e-05, "loss": 0.0605, "step": 18280 }, { "epoch": 0.86, "learning_rate": 1.1446814482177942e-05, "loss": 0.0691, "step": 18285 }, { "epoch": 0.86, "learning_rate": 1.1444475629151465e-05, "loss": 0.0644, "step": 18290 }, { "epoch": 0.86, "learning_rate": 1.144213677612499e-05, "loss": 0.0958, "step": 18295 }, { "epoch": 0.86, "learning_rate": 1.1439797923098512e-05, "loss": 0.0542, "step": 18300 }, { "epoch": 0.86, "learning_rate": 1.1437459070072037e-05, "loss": 0.1094, "step": 18305 }, { "epoch": 0.86, "learning_rate": 1.1435120217045562e-05, "loss": 0.0647, "step": 18310 }, { "epoch": 0.86, "learning_rate": 1.1432781364019085e-05, "loss": 0.0814, "step": 18315 }, { "epoch": 0.86, "learning_rate": 1.143044251099261e-05, "loss": 0.0218, "step": 18320 }, { "epoch": 0.86, "learning_rate": 1.1428103657966135e-05, "loss": 0.0583, "step": 18325 }, { "epoch": 0.86, "learning_rate": 1.1425764804939658e-05, "loss": 0.1478, "step": 18330 }, { "epoch": 0.86, "learning_rate": 1.1423425951913183e-05, "loss": 0.1476, "step": 18335 }, { "epoch": 0.86, "learning_rate": 1.1421087098886706e-05, "loss": 0.0496, "step": 18340 }, { "epoch": 0.86, "learning_rate": 1.1418748245860231e-05, "loss": 0.0912, "step": 18345 }, { "epoch": 0.86, "learning_rate": 1.1416409392833756e-05, "loss": 0.0869, "step": 18350 }, { "epoch": 0.86, "learning_rate": 1.141407053980728e-05, "loss": 0.1339, "step": 18355 }, { "epoch": 0.86, "learning_rate": 1.1411731686780803e-05, "loss": 0.1253, "step": 18360 }, { "epoch": 0.86, "learning_rate": 1.1409392833754326e-05, "loss": 0.1121, "step": 18365 }, { "epoch": 0.86, "learning_rate": 1.1407053980727851e-05, "loss": 0.127, "step": 18370 }, { "epoch": 0.86, "learning_rate": 1.1404715127701376e-05, "loss": 0.1248, "step": 18375 }, { "epoch": 0.86, "learning_rate": 1.14023762746749e-05, "loss": 0.1458, "step": 18380 }, { "epoch": 0.86, "learning_rate": 1.1400037421648424e-05, "loss": 0.0553, "step": 18385 }, { "epoch": 0.86, "learning_rate": 1.139769856862195e-05, "loss": 0.072, "step": 18390 }, { "epoch": 0.86, "learning_rate": 1.1395359715595472e-05, "loss": 0.1342, "step": 18395 }, { "epoch": 0.86, "learning_rate": 1.1393020862568997e-05, "loss": 0.0841, "step": 18400 }, { "epoch": 0.86, "learning_rate": 1.1390682009542522e-05, "loss": 0.05, "step": 18405 }, { "epoch": 0.86, "learning_rate": 1.1388343156516046e-05, "loss": 0.0639, "step": 18410 }, { "epoch": 0.86, "learning_rate": 1.138600430348957e-05, "loss": 0.1224, "step": 18415 }, { "epoch": 0.86, "learning_rate": 1.1383665450463092e-05, "loss": 0.0816, "step": 18420 }, { "epoch": 0.86, "learning_rate": 1.1381326597436617e-05, "loss": 0.086, "step": 18425 }, { "epoch": 0.86, "learning_rate": 1.1378987744410142e-05, "loss": 0.1134, "step": 18430 }, { "epoch": 0.86, "learning_rate": 1.1376648891383665e-05, "loss": 0.1204, "step": 18435 }, { "epoch": 0.86, "learning_rate": 1.137431003835719e-05, "loss": 0.0948, "step": 18440 }, { "epoch": 0.86, "learning_rate": 1.1371971185330715e-05, "loss": 0.0913, "step": 18445 }, { "epoch": 0.86, "learning_rate": 1.1369632332304238e-05, "loss": 0.1132, "step": 18450 }, { "epoch": 0.86, "learning_rate": 1.1367293479277763e-05, "loss": 0.1088, "step": 18455 }, { "epoch": 0.86, "learning_rate": 1.1364954626251288e-05, "loss": 0.127, "step": 18460 }, { "epoch": 0.86, "learning_rate": 1.1362615773224812e-05, "loss": 0.0526, "step": 18465 }, { "epoch": 0.86, "learning_rate": 1.1360276920198337e-05, "loss": 0.0677, "step": 18470 }, { "epoch": 0.86, "learning_rate": 1.1357938067171862e-05, "loss": 0.076, "step": 18475 }, { "epoch": 0.86, "learning_rate": 1.1355599214145383e-05, "loss": 0.0857, "step": 18480 }, { "epoch": 0.86, "learning_rate": 1.1353260361118908e-05, "loss": 0.1549, "step": 18485 }, { "epoch": 0.86, "learning_rate": 1.1350921508092431e-05, "loss": 0.0859, "step": 18490 }, { "epoch": 0.87, "learning_rate": 1.1348582655065956e-05, "loss": 0.0472, "step": 18495 }, { "epoch": 0.87, "learning_rate": 1.1346243802039481e-05, "loss": 0.0629, "step": 18500 }, { "epoch": 0.87, "learning_rate": 1.1343904949013004e-05, "loss": 0.0639, "step": 18505 }, { "epoch": 0.87, "learning_rate": 1.134156609598653e-05, "loss": 0.0777, "step": 18510 }, { "epoch": 0.87, "learning_rate": 1.1339227242960054e-05, "loss": 0.096, "step": 18515 }, { "epoch": 0.87, "learning_rate": 1.1336888389933578e-05, "loss": 0.0919, "step": 18520 }, { "epoch": 0.87, "learning_rate": 1.1334549536907103e-05, "loss": 0.0677, "step": 18525 }, { "epoch": 0.87, "learning_rate": 1.1332210683880628e-05, "loss": 0.0852, "step": 18530 }, { "epoch": 0.87, "learning_rate": 1.132987183085415e-05, "loss": 0.1762, "step": 18535 }, { "epoch": 0.87, "learning_rate": 1.1327532977827674e-05, "loss": 0.066, "step": 18540 }, { "epoch": 0.87, "learning_rate": 1.1325194124801197e-05, "loss": 0.1151, "step": 18545 }, { "epoch": 0.87, "learning_rate": 1.1322855271774722e-05, "loss": 0.1442, "step": 18550 }, { "epoch": 0.87, "learning_rate": 1.1320516418748246e-05, "loss": 0.0561, "step": 18555 }, { "epoch": 0.87, "learning_rate": 1.131817756572177e-05, "loss": 0.0746, "step": 18560 }, { "epoch": 0.87, "learning_rate": 1.1315838712695295e-05, "loss": 0.0492, "step": 18565 }, { "epoch": 0.87, "learning_rate": 1.1313499859668819e-05, "loss": 0.0885, "step": 18570 }, { "epoch": 0.87, "learning_rate": 1.1311161006642344e-05, "loss": 0.0929, "step": 18575 }, { "epoch": 0.87, "learning_rate": 1.1308822153615869e-05, "loss": 0.0358, "step": 18580 }, { "epoch": 0.87, "learning_rate": 1.1306483300589392e-05, "loss": 0.1108, "step": 18585 }, { "epoch": 0.87, "learning_rate": 1.1304144447562917e-05, "loss": 0.0667, "step": 18590 }, { "epoch": 0.87, "learning_rate": 1.1301805594536442e-05, "loss": 0.098, "step": 18595 }, { "epoch": 0.87, "learning_rate": 1.1299466741509963e-05, "loss": 0.1242, "step": 18600 }, { "epoch": 0.87, "learning_rate": 1.1297127888483488e-05, "loss": 0.0862, "step": 18605 }, { "epoch": 0.87, "learning_rate": 1.1294789035457012e-05, "loss": 0.0955, "step": 18610 }, { "epoch": 0.87, "learning_rate": 1.1292450182430536e-05, "loss": 0.1408, "step": 18615 }, { "epoch": 0.87, "learning_rate": 1.1290111329404061e-05, "loss": 0.1013, "step": 18620 }, { "epoch": 0.87, "learning_rate": 1.1287772476377585e-05, "loss": 0.0927, "step": 18625 }, { "epoch": 0.87, "learning_rate": 1.128543362335111e-05, "loss": 0.1422, "step": 18630 }, { "epoch": 0.87, "learning_rate": 1.1283094770324635e-05, "loss": 0.1213, "step": 18635 }, { "epoch": 0.87, "learning_rate": 1.1280755917298158e-05, "loss": 0.1156, "step": 18640 }, { "epoch": 0.87, "learning_rate": 1.1278417064271683e-05, "loss": 0.0956, "step": 18645 }, { "epoch": 0.87, "learning_rate": 1.1276078211245208e-05, "loss": 0.0543, "step": 18650 }, { "epoch": 0.87, "learning_rate": 1.1273739358218731e-05, "loss": 0.0852, "step": 18655 }, { "epoch": 0.87, "learning_rate": 1.1271400505192254e-05, "loss": 0.1327, "step": 18660 }, { "epoch": 0.87, "learning_rate": 1.1269061652165778e-05, "loss": 0.1744, "step": 18665 }, { "epoch": 0.87, "learning_rate": 1.1266722799139302e-05, "loss": 0.1164, "step": 18670 }, { "epoch": 0.87, "learning_rate": 1.1264383946112827e-05, "loss": 0.0888, "step": 18675 }, { "epoch": 0.87, "learning_rate": 1.126204509308635e-05, "loss": 0.1207, "step": 18680 }, { "epoch": 0.87, "learning_rate": 1.1259706240059876e-05, "loss": 0.1232, "step": 18685 }, { "epoch": 0.87, "learning_rate": 1.12573673870334e-05, "loss": 0.0781, "step": 18690 }, { "epoch": 0.87, "learning_rate": 1.1255028534006924e-05, "loss": 0.0643, "step": 18695 }, { "epoch": 0.87, "learning_rate": 1.1252689680980449e-05, "loss": 0.0997, "step": 18700 }, { "epoch": 0.87, "learning_rate": 1.1250350827953974e-05, "loss": 0.0791, "step": 18705 }, { "epoch": 0.88, "learning_rate": 1.1248011974927497e-05, "loss": 0.1107, "step": 18710 }, { "epoch": 0.88, "learning_rate": 1.124567312190102e-05, "loss": 0.0825, "step": 18715 }, { "epoch": 0.88, "learning_rate": 1.1243334268874544e-05, "loss": 0.1185, "step": 18720 }, { "epoch": 0.88, "learning_rate": 1.1240995415848068e-05, "loss": 0.1389, "step": 18725 }, { "epoch": 0.88, "learning_rate": 1.1238656562821593e-05, "loss": 0.0775, "step": 18730 }, { "epoch": 0.88, "learning_rate": 1.1236317709795117e-05, "loss": 0.0459, "step": 18735 }, { "epoch": 0.88, "learning_rate": 1.1233978856768642e-05, "loss": 0.1072, "step": 18740 }, { "epoch": 0.88, "learning_rate": 1.1231640003742167e-05, "loss": 0.0546, "step": 18745 }, { "epoch": 0.88, "learning_rate": 1.122930115071569e-05, "loss": 0.0609, "step": 18750 }, { "epoch": 0.88, "learning_rate": 1.1226962297689215e-05, "loss": 0.0904, "step": 18755 }, { "epoch": 0.88, "learning_rate": 1.122462344466274e-05, "loss": 0.0536, "step": 18760 }, { "epoch": 0.88, "learning_rate": 1.1222284591636263e-05, "loss": 0.1135, "step": 18765 }, { "epoch": 0.88, "learning_rate": 1.1219945738609788e-05, "loss": 0.1241, "step": 18770 }, { "epoch": 0.88, "learning_rate": 1.121760688558331e-05, "loss": 0.1592, "step": 18775 }, { "epoch": 0.88, "learning_rate": 1.1215268032556834e-05, "loss": 0.1126, "step": 18780 }, { "epoch": 0.88, "learning_rate": 1.1212929179530358e-05, "loss": 0.1145, "step": 18785 }, { "epoch": 0.88, "learning_rate": 1.1210590326503883e-05, "loss": 0.082, "step": 18790 }, { "epoch": 0.88, "learning_rate": 1.1208251473477408e-05, "loss": 0.1135, "step": 18795 }, { "epoch": 0.88, "learning_rate": 1.1205912620450931e-05, "loss": 0.1629, "step": 18800 }, { "epoch": 0.88, "learning_rate": 1.1203573767424456e-05, "loss": 0.097, "step": 18805 }, { "epoch": 0.88, "learning_rate": 1.120123491439798e-05, "loss": 0.108, "step": 18810 }, { "epoch": 0.88, "learning_rate": 1.1198896061371504e-05, "loss": 0.0411, "step": 18815 }, { "epoch": 0.88, "learning_rate": 1.1196557208345029e-05, "loss": 0.1242, "step": 18820 }, { "epoch": 0.88, "learning_rate": 1.1194218355318554e-05, "loss": 0.0802, "step": 18825 }, { "epoch": 0.88, "learning_rate": 1.1191879502292077e-05, "loss": 0.1184, "step": 18830 }, { "epoch": 0.88, "learning_rate": 1.11895406492656e-05, "loss": 0.1335, "step": 18835 }, { "epoch": 0.88, "learning_rate": 1.1187201796239124e-05, "loss": 0.1434, "step": 18840 }, { "epoch": 0.88, "learning_rate": 1.1184862943212649e-05, "loss": 0.2039, "step": 18845 }, { "epoch": 0.88, "learning_rate": 1.1182524090186174e-05, "loss": 0.0793, "step": 18850 }, { "epoch": 0.88, "learning_rate": 1.1180185237159697e-05, "loss": 0.1304, "step": 18855 }, { "epoch": 0.88, "learning_rate": 1.1177846384133222e-05, "loss": 0.0302, "step": 18860 }, { "epoch": 0.88, "learning_rate": 1.1175507531106747e-05, "loss": 0.0498, "step": 18865 }, { "epoch": 0.88, "learning_rate": 1.117316867808027e-05, "loss": 0.0989, "step": 18870 }, { "epoch": 0.88, "learning_rate": 1.1170829825053795e-05, "loss": 0.0637, "step": 18875 }, { "epoch": 0.88, "learning_rate": 1.116849097202732e-05, "loss": 0.0714, "step": 18880 }, { "epoch": 0.88, "learning_rate": 1.1166152119000843e-05, "loss": 0.0564, "step": 18885 }, { "epoch": 0.88, "learning_rate": 1.1163813265974368e-05, "loss": 0.0866, "step": 18890 }, { "epoch": 0.88, "learning_rate": 1.116147441294789e-05, "loss": 0.0748, "step": 18895 }, { "epoch": 0.88, "learning_rate": 1.1159135559921415e-05, "loss": 0.1421, "step": 18900 }, { "epoch": 0.88, "learning_rate": 1.115679670689494e-05, "loss": 0.06, "step": 18905 }, { "epoch": 0.88, "learning_rate": 1.1154457853868463e-05, "loss": 0.1055, "step": 18910 }, { "epoch": 0.88, "learning_rate": 1.1152119000841988e-05, "loss": 0.057, "step": 18915 }, { "epoch": 0.89, "learning_rate": 1.1149780147815513e-05, "loss": 0.0737, "step": 18920 }, { "epoch": 0.89, "learning_rate": 1.1147441294789036e-05, "loss": 0.1876, "step": 18925 }, { "epoch": 0.89, "learning_rate": 1.1145102441762561e-05, "loss": 0.1373, "step": 18930 }, { "epoch": 0.89, "learning_rate": 1.1142763588736086e-05, "loss": 0.0819, "step": 18935 }, { "epoch": 0.89, "learning_rate": 1.114042473570961e-05, "loss": 0.0946, "step": 18940 }, { "epoch": 0.89, "learning_rate": 1.1138085882683134e-05, "loss": 0.0738, "step": 18945 }, { "epoch": 0.89, "learning_rate": 1.1135747029656659e-05, "loss": 0.086, "step": 18950 }, { "epoch": 0.89, "learning_rate": 1.113340817663018e-05, "loss": 0.1339, "step": 18955 }, { "epoch": 0.89, "learning_rate": 1.1131069323603706e-05, "loss": 0.1296, "step": 18960 }, { "epoch": 0.89, "learning_rate": 1.1128730470577229e-05, "loss": 0.0666, "step": 18965 }, { "epoch": 0.89, "learning_rate": 1.1126391617550754e-05, "loss": 0.1154, "step": 18970 }, { "epoch": 0.89, "learning_rate": 1.1124052764524279e-05, "loss": 0.0672, "step": 18975 }, { "epoch": 0.89, "learning_rate": 1.1121713911497802e-05, "loss": 0.0912, "step": 18980 }, { "epoch": 0.89, "learning_rate": 1.1119375058471327e-05, "loss": 0.1286, "step": 18985 }, { "epoch": 0.89, "learning_rate": 1.111703620544485e-05, "loss": 0.0985, "step": 18990 }, { "epoch": 0.89, "learning_rate": 1.1114697352418375e-05, "loss": 0.1221, "step": 18995 }, { "epoch": 0.89, "learning_rate": 1.11123584993919e-05, "loss": 0.1175, "step": 19000 }, { "epoch": 0.89, "learning_rate": 1.1110019646365423e-05, "loss": 0.1126, "step": 19005 }, { "epoch": 0.89, "learning_rate": 1.1107680793338948e-05, "loss": 0.1029, "step": 19010 }, { "epoch": 0.89, "learning_rate": 1.110534194031247e-05, "loss": 0.0579, "step": 19015 }, { "epoch": 0.89, "learning_rate": 1.1103003087285995e-05, "loss": 0.069, "step": 19020 }, { "epoch": 0.89, "learning_rate": 1.110066423425952e-05, "loss": 0.1001, "step": 19025 }, { "epoch": 0.89, "learning_rate": 1.1098325381233043e-05, "loss": 0.1069, "step": 19030 }, { "epoch": 0.89, "learning_rate": 1.1095986528206568e-05, "loss": 0.1205, "step": 19035 }, { "epoch": 0.89, "learning_rate": 1.1093647675180093e-05, "loss": 0.059, "step": 19040 }, { "epoch": 0.89, "learning_rate": 1.1091308822153616e-05, "loss": 0.1218, "step": 19045 }, { "epoch": 0.89, "learning_rate": 1.1088969969127141e-05, "loss": 0.1059, "step": 19050 }, { "epoch": 0.89, "learning_rate": 1.1086631116100666e-05, "loss": 0.0973, "step": 19055 }, { "epoch": 0.89, "learning_rate": 1.108429226307419e-05, "loss": 0.0592, "step": 19060 }, { "epoch": 0.89, "learning_rate": 1.1081953410047714e-05, "loss": 0.0611, "step": 19065 }, { "epoch": 0.89, "learning_rate": 1.107961455702124e-05, "loss": 0.1634, "step": 19070 }, { "epoch": 0.89, "learning_rate": 1.1077275703994761e-05, "loss": 0.0888, "step": 19075 }, { "epoch": 0.89, "learning_rate": 1.1074936850968286e-05, "loss": 0.0535, "step": 19080 }, { "epoch": 0.89, "learning_rate": 1.1072597997941809e-05, "loss": 0.1486, "step": 19085 }, { "epoch": 0.89, "learning_rate": 1.1070259144915334e-05, "loss": 0.1297, "step": 19090 }, { "epoch": 0.89, "learning_rate": 1.1067920291888859e-05, "loss": 0.0833, "step": 19095 }, { "epoch": 0.89, "learning_rate": 1.1065581438862382e-05, "loss": 0.0679, "step": 19100 }, { "epoch": 0.89, "learning_rate": 1.1063242585835907e-05, "loss": 0.0853, "step": 19105 }, { "epoch": 0.89, "learning_rate": 1.1060903732809432e-05, "loss": 0.0966, "step": 19110 }, { "epoch": 0.89, "learning_rate": 1.1058564879782955e-05, "loss": 0.0666, "step": 19115 }, { "epoch": 0.89, "learning_rate": 1.105622602675648e-05, "loss": 0.0596, "step": 19120 }, { "epoch": 0.89, "learning_rate": 1.1053887173730005e-05, "loss": 0.0664, "step": 19125 }, { "epoch": 0.89, "learning_rate": 1.1051548320703527e-05, "loss": 0.1222, "step": 19130 }, { "epoch": 0.9, "learning_rate": 1.1049209467677052e-05, "loss": 0.1107, "step": 19135 }, { "epoch": 0.9, "learning_rate": 1.1046870614650575e-05, "loss": 0.1463, "step": 19140 }, { "epoch": 0.9, "learning_rate": 1.10445317616241e-05, "loss": 0.059, "step": 19145 }, { "epoch": 0.9, "learning_rate": 1.1042192908597625e-05, "loss": 0.1119, "step": 19150 }, { "epoch": 0.9, "learning_rate": 1.1039854055571148e-05, "loss": 0.0504, "step": 19155 }, { "epoch": 0.9, "learning_rate": 1.1037515202544673e-05, "loss": 0.0991, "step": 19160 }, { "epoch": 0.9, "learning_rate": 1.1035176349518198e-05, "loss": 0.0398, "step": 19165 }, { "epoch": 0.9, "learning_rate": 1.1032837496491721e-05, "loss": 0.1371, "step": 19170 }, { "epoch": 0.9, "learning_rate": 1.1030498643465246e-05, "loss": 0.0955, "step": 19175 }, { "epoch": 0.9, "learning_rate": 1.1028159790438771e-05, "loss": 0.0892, "step": 19180 }, { "epoch": 0.9, "learning_rate": 1.1025820937412295e-05, "loss": 0.1511, "step": 19185 }, { "epoch": 0.9, "learning_rate": 1.1023482084385818e-05, "loss": 0.0764, "step": 19190 }, { "epoch": 0.9, "learning_rate": 1.1021143231359341e-05, "loss": 0.0693, "step": 19195 }, { "epoch": 0.9, "learning_rate": 1.1018804378332866e-05, "loss": 0.0854, "step": 19200 }, { "epoch": 0.9, "learning_rate": 1.1016465525306391e-05, "loss": 0.0643, "step": 19205 }, { "epoch": 0.9, "learning_rate": 1.1014126672279914e-05, "loss": 0.0787, "step": 19210 }, { "epoch": 0.9, "learning_rate": 1.101178781925344e-05, "loss": 0.115, "step": 19215 }, { "epoch": 0.9, "learning_rate": 1.1009448966226962e-05, "loss": 0.0891, "step": 19220 }, { "epoch": 0.9, "learning_rate": 1.1007110113200487e-05, "loss": 0.1192, "step": 19225 }, { "epoch": 0.9, "learning_rate": 1.1004771260174012e-05, "loss": 0.0285, "step": 19230 }, { "epoch": 0.9, "learning_rate": 1.1002432407147536e-05, "loss": 0.0903, "step": 19235 }, { "epoch": 0.9, "learning_rate": 1.100009355412106e-05, "loss": 0.0725, "step": 19240 }, { "epoch": 0.9, "learning_rate": 1.0997754701094585e-05, "loss": 0.0839, "step": 19245 }, { "epoch": 0.9, "learning_rate": 1.0995415848068107e-05, "loss": 0.0725, "step": 19250 }, { "epoch": 0.9, "learning_rate": 1.0993076995041632e-05, "loss": 0.1057, "step": 19255 }, { "epoch": 0.9, "learning_rate": 1.0990738142015155e-05, "loss": 0.1011, "step": 19260 }, { "epoch": 0.9, "learning_rate": 1.098839928898868e-05, "loss": 0.0429, "step": 19265 }, { "epoch": 0.9, "learning_rate": 1.0986060435962205e-05, "loss": 0.0722, "step": 19270 }, { "epoch": 0.9, "learning_rate": 1.0983721582935728e-05, "loss": 0.0855, "step": 19275 }, { "epoch": 0.9, "learning_rate": 1.0981382729909253e-05, "loss": 0.0463, "step": 19280 }, { "epoch": 0.9, "learning_rate": 1.0979043876882778e-05, "loss": 0.0783, "step": 19285 }, { "epoch": 0.9, "learning_rate": 1.0976705023856302e-05, "loss": 0.0658, "step": 19290 }, { "epoch": 0.9, "learning_rate": 1.0974366170829827e-05, "loss": 0.0856, "step": 19295 }, { "epoch": 0.9, "learning_rate": 1.0972027317803351e-05, "loss": 0.1278, "step": 19300 }, { "epoch": 0.9, "learning_rate": 1.0969688464776875e-05, "loss": 0.0615, "step": 19305 }, { "epoch": 0.9, "learning_rate": 1.0967349611750398e-05, "loss": 0.0759, "step": 19310 }, { "epoch": 0.9, "learning_rate": 1.0965010758723921e-05, "loss": 0.1054, "step": 19315 }, { "epoch": 0.9, "learning_rate": 1.0962671905697446e-05, "loss": 0.1163, "step": 19320 }, { "epoch": 0.9, "learning_rate": 1.0960333052670971e-05, "loss": 0.1179, "step": 19325 }, { "epoch": 0.9, "learning_rate": 1.0957994199644494e-05, "loss": 0.1549, "step": 19330 }, { "epoch": 0.9, "learning_rate": 1.095565534661802e-05, "loss": 0.112, "step": 19335 }, { "epoch": 0.9, "learning_rate": 1.0953316493591544e-05, "loss": 0.0764, "step": 19340 }, { "epoch": 0.9, "learning_rate": 1.0950977640565068e-05, "loss": 0.0663, "step": 19345 }, { "epoch": 0.91, "learning_rate": 1.0948638787538593e-05, "loss": 0.1591, "step": 19350 }, { "epoch": 0.91, "learning_rate": 1.0946299934512117e-05, "loss": 0.1117, "step": 19355 }, { "epoch": 0.91, "learning_rate": 1.094396108148564e-05, "loss": 0.0778, "step": 19360 }, { "epoch": 0.91, "learning_rate": 1.0941622228459166e-05, "loss": 0.1248, "step": 19365 }, { "epoch": 0.91, "learning_rate": 1.0939283375432687e-05, "loss": 0.1461, "step": 19370 }, { "epoch": 0.91, "learning_rate": 1.0936944522406212e-05, "loss": 0.0646, "step": 19375 }, { "epoch": 0.91, "learning_rate": 1.0934605669379737e-05, "loss": 0.0293, "step": 19380 }, { "epoch": 0.91, "learning_rate": 1.093226681635326e-05, "loss": 0.0664, "step": 19385 }, { "epoch": 0.91, "learning_rate": 1.0929927963326785e-05, "loss": 0.0969, "step": 19390 }, { "epoch": 0.91, "learning_rate": 1.092758911030031e-05, "loss": 0.0933, "step": 19395 }, { "epoch": 0.91, "learning_rate": 1.0925250257273834e-05, "loss": 0.0952, "step": 19400 }, { "epoch": 0.91, "learning_rate": 1.0922911404247359e-05, "loss": 0.1041, "step": 19405 }, { "epoch": 0.91, "learning_rate": 1.0920572551220883e-05, "loss": 0.0879, "step": 19410 }, { "epoch": 0.91, "learning_rate": 1.0918233698194407e-05, "loss": 0.0572, "step": 19415 }, { "epoch": 0.91, "learning_rate": 1.0915894845167932e-05, "loss": 0.107, "step": 19420 }, { "epoch": 0.91, "learning_rate": 1.0913555992141455e-05, "loss": 0.0588, "step": 19425 }, { "epoch": 0.91, "learning_rate": 1.0911217139114978e-05, "loss": 0.1151, "step": 19430 }, { "epoch": 0.91, "learning_rate": 1.0908878286088501e-05, "loss": 0.1368, "step": 19435 }, { "epoch": 0.91, "learning_rate": 1.0906539433062026e-05, "loss": 0.1405, "step": 19440 }, { "epoch": 0.91, "learning_rate": 1.0904200580035551e-05, "loss": 0.1184, "step": 19445 }, { "epoch": 0.91, "learning_rate": 1.0901861727009075e-05, "loss": 0.1225, "step": 19450 }, { "epoch": 0.91, "learning_rate": 1.08995228739826e-05, "loss": 0.0945, "step": 19455 }, { "epoch": 0.91, "learning_rate": 1.0897184020956125e-05, "loss": 0.0675, "step": 19460 }, { "epoch": 0.91, "learning_rate": 1.0894845167929648e-05, "loss": 0.0646, "step": 19465 }, { "epoch": 0.91, "learning_rate": 1.0892506314903173e-05, "loss": 0.0504, "step": 19470 }, { "epoch": 0.91, "learning_rate": 1.0890167461876698e-05, "loss": 0.0874, "step": 19475 }, { "epoch": 0.91, "learning_rate": 1.0887828608850221e-05, "loss": 0.0451, "step": 19480 }, { "epoch": 0.91, "learning_rate": 1.0885489755823746e-05, "loss": 0.0342, "step": 19485 }, { "epoch": 0.91, "learning_rate": 1.0883150902797267e-05, "loss": 0.0864, "step": 19490 }, { "epoch": 0.91, "learning_rate": 1.0880812049770792e-05, "loss": 0.0404, "step": 19495 }, { "epoch": 0.91, "learning_rate": 1.0878473196744317e-05, "loss": 0.0996, "step": 19500 }, { "epoch": 0.91, "learning_rate": 1.087613434371784e-05, "loss": 0.1137, "step": 19505 }, { "epoch": 0.91, "learning_rate": 1.0873795490691366e-05, "loss": 0.1451, "step": 19510 }, { "epoch": 0.91, "learning_rate": 1.087145663766489e-05, "loss": 0.1013, "step": 19515 }, { "epoch": 0.91, "learning_rate": 1.0869117784638414e-05, "loss": 0.0598, "step": 19520 }, { "epoch": 0.91, "learning_rate": 1.0866778931611939e-05, "loss": 0.0629, "step": 19525 }, { "epoch": 0.91, "learning_rate": 1.0864440078585464e-05, "loss": 0.1282, "step": 19530 }, { "epoch": 0.91, "learning_rate": 1.0862101225558987e-05, "loss": 0.0766, "step": 19535 }, { "epoch": 0.91, "learning_rate": 1.0859762372532512e-05, "loss": 0.078, "step": 19540 }, { "epoch": 0.91, "learning_rate": 1.0857423519506037e-05, "loss": 0.1015, "step": 19545 }, { "epoch": 0.91, "learning_rate": 1.0855084666479558e-05, "loss": 0.0998, "step": 19550 }, { "epoch": 0.91, "learning_rate": 1.0852745813453083e-05, "loss": 0.1109, "step": 19555 }, { "epoch": 0.91, "learning_rate": 1.0850406960426607e-05, "loss": 0.0816, "step": 19560 }, { "epoch": 0.92, "learning_rate": 1.0848068107400132e-05, "loss": 0.1421, "step": 19565 }, { "epoch": 0.92, "learning_rate": 1.0845729254373657e-05, "loss": 0.079, "step": 19570 }, { "epoch": 0.92, "learning_rate": 1.084339040134718e-05, "loss": 0.1091, "step": 19575 }, { "epoch": 0.92, "learning_rate": 1.0841051548320705e-05, "loss": 0.1016, "step": 19580 }, { "epoch": 0.92, "learning_rate": 1.083871269529423e-05, "loss": 0.0659, "step": 19585 }, { "epoch": 0.92, "learning_rate": 1.0836373842267753e-05, "loss": 0.0269, "step": 19590 }, { "epoch": 0.92, "learning_rate": 1.0834034989241278e-05, "loss": 0.0973, "step": 19595 }, { "epoch": 0.92, "learning_rate": 1.0831696136214803e-05, "loss": 0.0881, "step": 19600 }, { "epoch": 0.92, "learning_rate": 1.0829357283188324e-05, "loss": 0.0995, "step": 19605 }, { "epoch": 0.92, "learning_rate": 1.082701843016185e-05, "loss": 0.0833, "step": 19610 }, { "epoch": 0.92, "learning_rate": 1.0824679577135373e-05, "loss": 0.0799, "step": 19615 }, { "epoch": 0.92, "learning_rate": 1.0822340724108898e-05, "loss": 0.087, "step": 19620 }, { "epoch": 0.92, "learning_rate": 1.0820001871082423e-05, "loss": 0.0691, "step": 19625 }, { "epoch": 0.92, "learning_rate": 1.0817663018055946e-05, "loss": 0.0481, "step": 19630 }, { "epoch": 0.92, "learning_rate": 1.081532416502947e-05, "loss": 0.0773, "step": 19635 }, { "epoch": 0.92, "learning_rate": 1.0812985312002996e-05, "loss": 0.0447, "step": 19640 }, { "epoch": 0.92, "learning_rate": 1.0810646458976519e-05, "loss": 0.1031, "step": 19645 }, { "epoch": 0.92, "learning_rate": 1.0808307605950044e-05, "loss": 0.147, "step": 19650 }, { "epoch": 0.92, "learning_rate": 1.0805968752923567e-05, "loss": 0.0688, "step": 19655 }, { "epoch": 0.92, "learning_rate": 1.0803629899897092e-05, "loss": 0.0759, "step": 19660 }, { "epoch": 0.92, "learning_rate": 1.0801291046870614e-05, "loss": 0.0459, "step": 19665 }, { "epoch": 0.92, "learning_rate": 1.0798952193844139e-05, "loss": 0.1187, "step": 19670 }, { "epoch": 0.92, "learning_rate": 1.0796613340817664e-05, "loss": 0.0592, "step": 19675 }, { "epoch": 0.92, "learning_rate": 1.0794274487791187e-05, "loss": 0.079, "step": 19680 }, { "epoch": 0.92, "learning_rate": 1.0791935634764712e-05, "loss": 0.0912, "step": 19685 }, { "epoch": 0.92, "learning_rate": 1.0789596781738237e-05, "loss": 0.1246, "step": 19690 }, { "epoch": 0.92, "learning_rate": 1.078725792871176e-05, "loss": 0.0618, "step": 19695 }, { "epoch": 0.92, "learning_rate": 1.0784919075685285e-05, "loss": 0.0636, "step": 19700 }, { "epoch": 0.92, "learning_rate": 1.078258022265881e-05, "loss": 0.1082, "step": 19705 }, { "epoch": 0.92, "learning_rate": 1.0780241369632333e-05, "loss": 0.1167, "step": 19710 }, { "epoch": 0.92, "learning_rate": 1.0777902516605858e-05, "loss": 0.0675, "step": 19715 }, { "epoch": 0.92, "learning_rate": 1.0775563663579383e-05, "loss": 0.0681, "step": 19720 }, { "epoch": 0.92, "learning_rate": 1.0773224810552905e-05, "loss": 0.1265, "step": 19725 }, { "epoch": 0.92, "learning_rate": 1.077088595752643e-05, "loss": 0.102, "step": 19730 }, { "epoch": 0.92, "learning_rate": 1.0768547104499953e-05, "loss": 0.0933, "step": 19735 }, { "epoch": 0.92, "learning_rate": 1.0766208251473478e-05, "loss": 0.1251, "step": 19740 }, { "epoch": 0.92, "learning_rate": 1.0763869398447003e-05, "loss": 0.0976, "step": 19745 }, { "epoch": 0.92, "learning_rate": 1.0761530545420526e-05, "loss": 0.0769, "step": 19750 }, { "epoch": 0.92, "learning_rate": 1.0759191692394051e-05, "loss": 0.0272, "step": 19755 }, { "epoch": 0.92, "learning_rate": 1.0756852839367576e-05, "loss": 0.0779, "step": 19760 }, { "epoch": 0.92, "learning_rate": 1.0754513986341099e-05, "loss": 0.132, "step": 19765 }, { "epoch": 0.92, "learning_rate": 1.0752175133314624e-05, "loss": 0.0873, "step": 19770 }, { "epoch": 0.92, "learning_rate": 1.0749836280288149e-05, "loss": 0.083, "step": 19775 }, { "epoch": 0.93, "learning_rate": 1.0747497427261672e-05, "loss": 0.1058, "step": 19780 }, { "epoch": 0.93, "learning_rate": 1.0745158574235196e-05, "loss": 0.0837, "step": 19785 }, { "epoch": 0.93, "learning_rate": 1.0742819721208719e-05, "loss": 0.094, "step": 19790 }, { "epoch": 0.93, "learning_rate": 1.0740480868182244e-05, "loss": 0.0605, "step": 19795 }, { "epoch": 0.93, "learning_rate": 1.0738142015155769e-05, "loss": 0.075, "step": 19800 }, { "epoch": 0.93, "learning_rate": 1.0735803162129292e-05, "loss": 0.0962, "step": 19805 }, { "epoch": 0.93, "learning_rate": 1.0733464309102817e-05, "loss": 0.0596, "step": 19810 }, { "epoch": 0.93, "learning_rate": 1.0731125456076342e-05, "loss": 0.09, "step": 19815 }, { "epoch": 0.93, "learning_rate": 1.0728786603049865e-05, "loss": 0.1578, "step": 19820 }, { "epoch": 0.93, "learning_rate": 1.072644775002339e-05, "loss": 0.0797, "step": 19825 }, { "epoch": 0.93, "learning_rate": 1.0724108896996915e-05, "loss": 0.1001, "step": 19830 }, { "epoch": 0.93, "learning_rate": 1.0721770043970438e-05, "loss": 0.1014, "step": 19835 }, { "epoch": 0.93, "learning_rate": 1.0719431190943963e-05, "loss": 0.1189, "step": 19840 }, { "epoch": 0.93, "learning_rate": 1.0717092337917485e-05, "loss": 0.1417, "step": 19845 }, { "epoch": 0.93, "learning_rate": 1.071475348489101e-05, "loss": 0.0807, "step": 19850 }, { "epoch": 0.93, "learning_rate": 1.0712414631864535e-05, "loss": 0.0607, "step": 19855 }, { "epoch": 0.93, "learning_rate": 1.0710075778838058e-05, "loss": 0.1294, "step": 19860 }, { "epoch": 0.93, "learning_rate": 1.0707736925811583e-05, "loss": 0.061, "step": 19865 }, { "epoch": 0.93, "learning_rate": 1.0705398072785106e-05, "loss": 0.1302, "step": 19870 }, { "epoch": 0.93, "learning_rate": 1.0703059219758631e-05, "loss": 0.0815, "step": 19875 }, { "epoch": 0.93, "learning_rate": 1.0700720366732156e-05, "loss": 0.1327, "step": 19880 }, { "epoch": 0.93, "learning_rate": 1.069838151370568e-05, "loss": 0.0769, "step": 19885 }, { "epoch": 0.93, "learning_rate": 1.0696042660679204e-05, "loss": 0.1137, "step": 19890 }, { "epoch": 0.93, "learning_rate": 1.069370380765273e-05, "loss": 0.1381, "step": 19895 }, { "epoch": 0.93, "learning_rate": 1.0691364954626252e-05, "loss": 0.0929, "step": 19900 }, { "epoch": 0.93, "learning_rate": 1.0689026101599776e-05, "loss": 0.0979, "step": 19905 }, { "epoch": 0.93, "learning_rate": 1.0686687248573299e-05, "loss": 0.0818, "step": 19910 }, { "epoch": 0.93, "learning_rate": 1.0684348395546824e-05, "loss": 0.1017, "step": 19915 }, { "epoch": 0.93, "learning_rate": 1.0682009542520349e-05, "loss": 0.056, "step": 19920 }, { "epoch": 0.93, "learning_rate": 1.0679670689493872e-05, "loss": 0.0279, "step": 19925 }, { "epoch": 0.93, "learning_rate": 1.0677331836467397e-05, "loss": 0.0643, "step": 19930 }, { "epoch": 0.93, "learning_rate": 1.0674992983440922e-05, "loss": 0.0783, "step": 19935 }, { "epoch": 0.93, "learning_rate": 1.0672654130414445e-05, "loss": 0.0692, "step": 19940 }, { "epoch": 0.93, "learning_rate": 1.067031527738797e-05, "loss": 0.1072, "step": 19945 }, { "epoch": 0.93, "learning_rate": 1.0667976424361495e-05, "loss": 0.1443, "step": 19950 }, { "epoch": 0.93, "learning_rate": 1.0665637571335018e-05, "loss": 0.0928, "step": 19955 }, { "epoch": 0.93, "learning_rate": 1.0663298718308543e-05, "loss": 0.0896, "step": 19960 }, { "epoch": 0.93, "learning_rate": 1.0660959865282065e-05, "loss": 0.077, "step": 19965 }, { "epoch": 0.93, "learning_rate": 1.065862101225559e-05, "loss": 0.1395, "step": 19970 }, { "epoch": 0.93, "learning_rate": 1.0656282159229115e-05, "loss": 0.0863, "step": 19975 }, { "epoch": 0.93, "learning_rate": 1.0653943306202638e-05, "loss": 0.0627, "step": 19980 }, { "epoch": 0.93, "learning_rate": 1.0651604453176163e-05, "loss": 0.1182, "step": 19985 }, { "epoch": 0.94, "learning_rate": 1.0649265600149688e-05, "loss": 0.0538, "step": 19990 }, { "epoch": 0.94, "learning_rate": 1.0646926747123211e-05, "loss": 0.095, "step": 19995 }, { "epoch": 0.94, "learning_rate": 1.0644587894096736e-05, "loss": 0.092, "step": 20000 }, { "epoch": 0.94, "learning_rate": 1.0642249041070261e-05, "loss": 0.1234, "step": 20005 }, { "epoch": 0.94, "learning_rate": 1.0639910188043784e-05, "loss": 0.1159, "step": 20010 }, { "epoch": 0.94, "learning_rate": 1.063757133501731e-05, "loss": 0.076, "step": 20015 }, { "epoch": 0.94, "learning_rate": 1.0635232481990831e-05, "loss": 0.1202, "step": 20020 }, { "epoch": 0.94, "learning_rate": 1.0632893628964356e-05, "loss": 0.1022, "step": 20025 }, { "epoch": 0.94, "learning_rate": 1.0630554775937881e-05, "loss": 0.0751, "step": 20030 }, { "epoch": 0.94, "learning_rate": 1.0628215922911404e-05, "loss": 0.0903, "step": 20035 }, { "epoch": 0.94, "learning_rate": 1.0625877069884929e-05, "loss": 0.1467, "step": 20040 }, { "epoch": 0.94, "learning_rate": 1.0623538216858454e-05, "loss": 0.1355, "step": 20045 }, { "epoch": 0.94, "learning_rate": 1.0621199363831977e-05, "loss": 0.0618, "step": 20050 }, { "epoch": 0.94, "learning_rate": 1.0618860510805502e-05, "loss": 0.1238, "step": 20055 }, { "epoch": 0.94, "learning_rate": 1.0616521657779027e-05, "loss": 0.1104, "step": 20060 }, { "epoch": 0.94, "learning_rate": 1.061418280475255e-05, "loss": 0.0859, "step": 20065 }, { "epoch": 0.94, "learning_rate": 1.0611843951726075e-05, "loss": 0.13, "step": 20070 }, { "epoch": 0.94, "learning_rate": 1.06095050986996e-05, "loss": 0.0501, "step": 20075 }, { "epoch": 0.94, "learning_rate": 1.0607166245673122e-05, "loss": 0.0986, "step": 20080 }, { "epoch": 0.94, "learning_rate": 1.0604827392646647e-05, "loss": 0.1678, "step": 20085 }, { "epoch": 0.94, "learning_rate": 1.060248853962017e-05, "loss": 0.1188, "step": 20090 }, { "epoch": 0.94, "learning_rate": 1.0600149686593695e-05, "loss": 0.1133, "step": 20095 }, { "epoch": 0.94, "learning_rate": 1.0597810833567218e-05, "loss": 0.1101, "step": 20100 }, { "epoch": 0.94, "learning_rate": 1.0595471980540743e-05, "loss": 0.0521, "step": 20105 }, { "epoch": 0.94, "learning_rate": 1.0593133127514268e-05, "loss": 0.1136, "step": 20110 }, { "epoch": 0.94, "learning_rate": 1.0590794274487792e-05, "loss": 0.0781, "step": 20115 }, { "epoch": 0.94, "learning_rate": 1.0588455421461316e-05, "loss": 0.11, "step": 20120 }, { "epoch": 0.94, "learning_rate": 1.0586116568434841e-05, "loss": 0.0912, "step": 20125 }, { "epoch": 0.94, "learning_rate": 1.0583777715408365e-05, "loss": 0.0713, "step": 20130 }, { "epoch": 0.94, "learning_rate": 1.058143886238189e-05, "loss": 0.0677, "step": 20135 }, { "epoch": 0.94, "learning_rate": 1.0579100009355411e-05, "loss": 0.0806, "step": 20140 }, { "epoch": 0.94, "learning_rate": 1.0576761156328936e-05, "loss": 0.1166, "step": 20145 }, { "epoch": 0.94, "learning_rate": 1.0574422303302461e-05, "loss": 0.1234, "step": 20150 }, { "epoch": 0.94, "learning_rate": 1.0572083450275984e-05, "loss": 0.0979, "step": 20155 }, { "epoch": 0.94, "learning_rate": 1.056974459724951e-05, "loss": 0.1226, "step": 20160 }, { "epoch": 0.94, "learning_rate": 1.0567405744223034e-05, "loss": 0.0633, "step": 20165 }, { "epoch": 0.94, "learning_rate": 1.0565066891196558e-05, "loss": 0.0707, "step": 20170 }, { "epoch": 0.94, "learning_rate": 1.0562728038170082e-05, "loss": 0.1154, "step": 20175 }, { "epoch": 0.94, "learning_rate": 1.0560389185143607e-05, "loss": 0.1018, "step": 20180 }, { "epoch": 0.94, "learning_rate": 1.055805033211713e-05, "loss": 0.1206, "step": 20185 }, { "epoch": 0.94, "learning_rate": 1.0555711479090656e-05, "loss": 0.1032, "step": 20190 }, { "epoch": 0.94, "learning_rate": 1.055337262606418e-05, "loss": 0.1035, "step": 20195 }, { "epoch": 0.94, "learning_rate": 1.0551033773037702e-05, "loss": 0.0688, "step": 20200 }, { "epoch": 0.95, "learning_rate": 1.0548694920011227e-05, "loss": 0.0776, "step": 20205 }, { "epoch": 0.95, "learning_rate": 1.054635606698475e-05, "loss": 0.1502, "step": 20210 }, { "epoch": 0.95, "learning_rate": 1.0544017213958275e-05, "loss": 0.1261, "step": 20215 }, { "epoch": 0.95, "learning_rate": 1.05416783609318e-05, "loss": 0.1407, "step": 20220 }, { "epoch": 0.95, "learning_rate": 1.0539339507905324e-05, "loss": 0.1267, "step": 20225 }, { "epoch": 0.95, "learning_rate": 1.0537000654878848e-05, "loss": 0.1091, "step": 20230 }, { "epoch": 0.95, "learning_rate": 1.0534661801852373e-05, "loss": 0.1016, "step": 20235 }, { "epoch": 0.95, "learning_rate": 1.0532322948825897e-05, "loss": 0.0936, "step": 20240 }, { "epoch": 0.95, "learning_rate": 1.0529984095799422e-05, "loss": 0.0791, "step": 20245 }, { "epoch": 0.95, "learning_rate": 1.0527645242772947e-05, "loss": 0.0763, "step": 20250 }, { "epoch": 0.95, "learning_rate": 1.052530638974647e-05, "loss": 0.0673, "step": 20255 }, { "epoch": 0.95, "learning_rate": 1.0522967536719993e-05, "loss": 0.0463, "step": 20260 }, { "epoch": 0.95, "learning_rate": 1.0520628683693516e-05, "loss": 0.1478, "step": 20265 }, { "epoch": 0.95, "learning_rate": 1.0518289830667041e-05, "loss": 0.0601, "step": 20270 }, { "epoch": 0.95, "learning_rate": 1.0515950977640566e-05, "loss": 0.0986, "step": 20275 }, { "epoch": 0.95, "learning_rate": 1.051361212461409e-05, "loss": 0.1086, "step": 20280 }, { "epoch": 0.95, "learning_rate": 1.0511273271587614e-05, "loss": 0.1013, "step": 20285 }, { "epoch": 0.95, "learning_rate": 1.050893441856114e-05, "loss": 0.0529, "step": 20290 }, { "epoch": 0.95, "learning_rate": 1.0506595565534663e-05, "loss": 0.0759, "step": 20295 }, { "epoch": 0.95, "learning_rate": 1.0504256712508188e-05, "loss": 0.0985, "step": 20300 }, { "epoch": 0.95, "learning_rate": 1.0501917859481711e-05, "loss": 0.0831, "step": 20305 }, { "epoch": 0.95, "learning_rate": 1.0499579006455236e-05, "loss": 0.1405, "step": 20310 }, { "epoch": 0.95, "learning_rate": 1.049724015342876e-05, "loss": 0.1071, "step": 20315 }, { "epoch": 0.95, "learning_rate": 1.0494901300402282e-05, "loss": 0.108, "step": 20320 }, { "epoch": 0.95, "learning_rate": 1.0492562447375807e-05, "loss": 0.063, "step": 20325 }, { "epoch": 0.95, "learning_rate": 1.049022359434933e-05, "loss": 0.0981, "step": 20330 }, { "epoch": 0.95, "learning_rate": 1.0487884741322856e-05, "loss": 0.1066, "step": 20335 }, { "epoch": 0.95, "learning_rate": 1.048554588829638e-05, "loss": 0.1124, "step": 20340 }, { "epoch": 0.95, "learning_rate": 1.0483207035269904e-05, "loss": 0.0794, "step": 20345 }, { "epoch": 0.95, "learning_rate": 1.0480868182243429e-05, "loss": 0.0978, "step": 20350 }, { "epoch": 0.95, "learning_rate": 1.0478529329216954e-05, "loss": 0.0566, "step": 20355 }, { "epoch": 0.95, "learning_rate": 1.0476190476190477e-05, "loss": 0.1315, "step": 20360 }, { "epoch": 0.95, "learning_rate": 1.0473851623164002e-05, "loss": 0.0787, "step": 20365 }, { "epoch": 0.95, "learning_rate": 1.0471512770137527e-05, "loss": 0.0668, "step": 20370 }, { "epoch": 0.95, "learning_rate": 1.046917391711105e-05, "loss": 0.0794, "step": 20375 }, { "epoch": 0.95, "learning_rate": 1.0466835064084573e-05, "loss": 0.0689, "step": 20380 }, { "epoch": 0.95, "learning_rate": 1.0464496211058097e-05, "loss": 0.0663, "step": 20385 }, { "epoch": 0.95, "learning_rate": 1.0462157358031622e-05, "loss": 0.088, "step": 20390 }, { "epoch": 0.95, "learning_rate": 1.0459818505005146e-05, "loss": 0.1196, "step": 20395 }, { "epoch": 0.95, "learning_rate": 1.045747965197867e-05, "loss": 0.1021, "step": 20400 }, { "epoch": 0.95, "learning_rate": 1.0455140798952195e-05, "loss": 0.1288, "step": 20405 }, { "epoch": 0.95, "learning_rate": 1.045280194592572e-05, "loss": 0.0653, "step": 20410 }, { "epoch": 0.95, "learning_rate": 1.0450463092899243e-05, "loss": 0.0712, "step": 20415 }, { "epoch": 0.96, "learning_rate": 1.0448124239872768e-05, "loss": 0.0891, "step": 20420 }, { "epoch": 0.96, "learning_rate": 1.0445785386846293e-05, "loss": 0.0941, "step": 20425 }, { "epoch": 0.96, "learning_rate": 1.0443446533819816e-05, "loss": 0.159, "step": 20430 }, { "epoch": 0.96, "learning_rate": 1.044110768079334e-05, "loss": 0.0267, "step": 20435 }, { "epoch": 0.96, "learning_rate": 1.0438768827766863e-05, "loss": 0.0759, "step": 20440 }, { "epoch": 0.96, "learning_rate": 1.0436429974740388e-05, "loss": 0.0721, "step": 20445 }, { "epoch": 0.96, "learning_rate": 1.0434091121713912e-05, "loss": 0.0705, "step": 20450 }, { "epoch": 0.96, "learning_rate": 1.0431752268687436e-05, "loss": 0.0792, "step": 20455 }, { "epoch": 0.96, "learning_rate": 1.042941341566096e-05, "loss": 0.0997, "step": 20460 }, { "epoch": 0.96, "learning_rate": 1.0427074562634486e-05, "loss": 0.1233, "step": 20465 }, { "epoch": 0.96, "learning_rate": 1.0424735709608009e-05, "loss": 0.0647, "step": 20470 }, { "epoch": 0.96, "learning_rate": 1.0422396856581534e-05, "loss": 0.0871, "step": 20475 }, { "epoch": 0.96, "learning_rate": 1.0420058003555059e-05, "loss": 0.0738, "step": 20480 }, { "epoch": 0.96, "learning_rate": 1.0417719150528582e-05, "loss": 0.0843, "step": 20485 }, { "epoch": 0.96, "learning_rate": 1.0415380297502107e-05, "loss": 0.0965, "step": 20490 }, { "epoch": 0.96, "learning_rate": 1.0413041444475629e-05, "loss": 0.1096, "step": 20495 }, { "epoch": 0.96, "learning_rate": 1.0410702591449154e-05, "loss": 0.0608, "step": 20500 }, { "epoch": 0.96, "learning_rate": 1.0408363738422678e-05, "loss": 0.07, "step": 20505 }, { "epoch": 0.96, "learning_rate": 1.0406024885396202e-05, "loss": 0.1171, "step": 20510 }, { "epoch": 0.96, "learning_rate": 1.0403686032369727e-05, "loss": 0.0677, "step": 20515 }, { "epoch": 0.96, "learning_rate": 1.0401347179343252e-05, "loss": 0.111, "step": 20520 }, { "epoch": 0.96, "learning_rate": 1.0399008326316775e-05, "loss": 0.0811, "step": 20525 }, { "epoch": 0.96, "learning_rate": 1.03966694732903e-05, "loss": 0.0432, "step": 20530 }, { "epoch": 0.96, "learning_rate": 1.0394330620263823e-05, "loss": 0.0836, "step": 20535 }, { "epoch": 0.96, "learning_rate": 1.0391991767237348e-05, "loss": 0.0564, "step": 20540 }, { "epoch": 0.96, "learning_rate": 1.0389652914210873e-05, "loss": 0.1134, "step": 20545 }, { "epoch": 0.96, "learning_rate": 1.0387314061184396e-05, "loss": 0.1389, "step": 20550 }, { "epoch": 0.96, "learning_rate": 1.038497520815792e-05, "loss": 0.1277, "step": 20555 }, { "epoch": 0.96, "learning_rate": 1.0382636355131443e-05, "loss": 0.0832, "step": 20560 }, { "epoch": 0.96, "learning_rate": 1.0380297502104968e-05, "loss": 0.0605, "step": 20565 }, { "epoch": 0.96, "learning_rate": 1.0377958649078493e-05, "loss": 0.1218, "step": 20570 }, { "epoch": 0.96, "learning_rate": 1.0375619796052016e-05, "loss": 0.1076, "step": 20575 }, { "epoch": 0.96, "learning_rate": 1.0373280943025541e-05, "loss": 0.1317, "step": 20580 }, { "epoch": 0.96, "learning_rate": 1.0370942089999066e-05, "loss": 0.0615, "step": 20585 }, { "epoch": 0.96, "learning_rate": 1.0368603236972589e-05, "loss": 0.0388, "step": 20590 }, { "epoch": 0.96, "learning_rate": 1.0366264383946114e-05, "loss": 0.071, "step": 20595 }, { "epoch": 0.96, "learning_rate": 1.0363925530919639e-05, "loss": 0.1254, "step": 20600 }, { "epoch": 0.96, "learning_rate": 1.0361586677893162e-05, "loss": 0.0705, "step": 20605 }, { "epoch": 0.96, "learning_rate": 1.0359247824866687e-05, "loss": 0.0623, "step": 20610 }, { "epoch": 0.96, "learning_rate": 1.0356908971840209e-05, "loss": 0.101, "step": 20615 }, { "epoch": 0.96, "learning_rate": 1.0354570118813734e-05, "loss": 0.0706, "step": 20620 }, { "epoch": 0.96, "learning_rate": 1.0352231265787259e-05, "loss": 0.1002, "step": 20625 }, { "epoch": 0.96, "learning_rate": 1.0349892412760782e-05, "loss": 0.0678, "step": 20630 }, { "epoch": 0.97, "learning_rate": 1.0347553559734307e-05, "loss": 0.1305, "step": 20635 }, { "epoch": 0.97, "learning_rate": 1.0345214706707832e-05, "loss": 0.132, "step": 20640 }, { "epoch": 0.97, "learning_rate": 1.0342875853681355e-05, "loss": 0.0863, "step": 20645 }, { "epoch": 0.97, "learning_rate": 1.034053700065488e-05, "loss": 0.1398, "step": 20650 }, { "epoch": 0.97, "learning_rate": 1.0338198147628405e-05, "loss": 0.0854, "step": 20655 }, { "epoch": 0.97, "learning_rate": 1.0335859294601928e-05, "loss": 0.0803, "step": 20660 }, { "epoch": 0.97, "learning_rate": 1.0333520441575453e-05, "loss": 0.0684, "step": 20665 }, { "epoch": 0.97, "learning_rate": 1.0331181588548978e-05, "loss": 0.0685, "step": 20670 }, { "epoch": 0.97, "learning_rate": 1.03288427355225e-05, "loss": 0.1241, "step": 20675 }, { "epoch": 0.97, "learning_rate": 1.0326503882496025e-05, "loss": 0.0856, "step": 20680 }, { "epoch": 0.97, "learning_rate": 1.0324165029469548e-05, "loss": 0.0919, "step": 20685 }, { "epoch": 0.97, "learning_rate": 1.0321826176443073e-05, "loss": 0.0784, "step": 20690 }, { "epoch": 0.97, "learning_rate": 1.0319487323416598e-05, "loss": 0.0827, "step": 20695 }, { "epoch": 0.97, "learning_rate": 1.0317148470390121e-05, "loss": 0.143, "step": 20700 }, { "epoch": 0.97, "learning_rate": 1.0314809617363646e-05, "loss": 0.0437, "step": 20705 }, { "epoch": 0.97, "learning_rate": 1.0312470764337171e-05, "loss": 0.0607, "step": 20710 }, { "epoch": 0.97, "learning_rate": 1.0310131911310694e-05, "loss": 0.0452, "step": 20715 }, { "epoch": 0.97, "learning_rate": 1.030779305828422e-05, "loss": 0.0912, "step": 20720 }, { "epoch": 0.97, "learning_rate": 1.0305454205257744e-05, "loss": 0.0552, "step": 20725 }, { "epoch": 0.97, "learning_rate": 1.0303115352231267e-05, "loss": 0.112, "step": 20730 }, { "epoch": 0.97, "learning_rate": 1.030077649920479e-05, "loss": 0.0792, "step": 20735 }, { "epoch": 0.97, "learning_rate": 1.0298437646178314e-05, "loss": 0.1013, "step": 20740 }, { "epoch": 0.97, "learning_rate": 1.0296098793151839e-05, "loss": 0.1513, "step": 20745 }, { "epoch": 0.97, "learning_rate": 1.0293759940125362e-05, "loss": 0.0864, "step": 20750 }, { "epoch": 0.97, "learning_rate": 1.0291421087098887e-05, "loss": 0.1186, "step": 20755 }, { "epoch": 0.97, "learning_rate": 1.0289082234072412e-05, "loss": 0.0917, "step": 20760 }, { "epoch": 0.97, "learning_rate": 1.0286743381045935e-05, "loss": 0.016, "step": 20765 }, { "epoch": 0.97, "learning_rate": 1.028440452801946e-05, "loss": 0.1073, "step": 20770 }, { "epoch": 0.97, "learning_rate": 1.0282065674992985e-05, "loss": 0.1102, "step": 20775 }, { "epoch": 0.97, "learning_rate": 1.0279726821966508e-05, "loss": 0.0533, "step": 20780 }, { "epoch": 0.97, "learning_rate": 1.0277387968940033e-05, "loss": 0.0529, "step": 20785 }, { "epoch": 0.97, "learning_rate": 1.0275049115913558e-05, "loss": 0.0956, "step": 20790 }, { "epoch": 0.97, "learning_rate": 1.027271026288708e-05, "loss": 0.0319, "step": 20795 }, { "epoch": 0.97, "learning_rate": 1.0270371409860605e-05, "loss": 0.0805, "step": 20800 }, { "epoch": 0.97, "learning_rate": 1.0268032556834128e-05, "loss": 0.1009, "step": 20805 }, { "epoch": 0.97, "learning_rate": 1.0265693703807653e-05, "loss": 0.0697, "step": 20810 }, { "epoch": 0.97, "learning_rate": 1.0263354850781178e-05, "loss": 0.0805, "step": 20815 }, { "epoch": 0.97, "learning_rate": 1.0261015997754701e-05, "loss": 0.0687, "step": 20820 }, { "epoch": 0.97, "learning_rate": 1.0258677144728226e-05, "loss": 0.0767, "step": 20825 }, { "epoch": 0.97, "learning_rate": 1.0256338291701751e-05, "loss": 0.1317, "step": 20830 }, { "epoch": 0.97, "learning_rate": 1.0253999438675274e-05, "loss": 0.1202, "step": 20835 }, { "epoch": 0.97, "learning_rate": 1.02516605856488e-05, "loss": 0.0761, "step": 20840 }, { "epoch": 0.98, "learning_rate": 1.0249321732622324e-05, "loss": 0.0832, "step": 20845 }, { "epoch": 0.98, "learning_rate": 1.0246982879595848e-05, "loss": 0.1443, "step": 20850 }, { "epoch": 0.98, "learning_rate": 1.0244644026569371e-05, "loss": 0.0916, "step": 20855 }, { "epoch": 0.98, "learning_rate": 1.0242305173542894e-05, "loss": 0.1694, "step": 20860 }, { "epoch": 0.98, "learning_rate": 1.0239966320516419e-05, "loss": 0.0862, "step": 20865 }, { "epoch": 0.98, "learning_rate": 1.0237627467489944e-05, "loss": 0.1538, "step": 20870 }, { "epoch": 0.98, "learning_rate": 1.0235288614463467e-05, "loss": 0.1407, "step": 20875 }, { "epoch": 0.98, "learning_rate": 1.0232949761436992e-05, "loss": 0.0386, "step": 20880 }, { "epoch": 0.98, "learning_rate": 1.0230610908410517e-05, "loss": 0.1137, "step": 20885 }, { "epoch": 0.98, "learning_rate": 1.022827205538404e-05, "loss": 0.1101, "step": 20890 }, { "epoch": 0.98, "learning_rate": 1.0225933202357565e-05, "loss": 0.0977, "step": 20895 }, { "epoch": 0.98, "learning_rate": 1.022359434933109e-05, "loss": 0.0945, "step": 20900 }, { "epoch": 0.98, "learning_rate": 1.0221255496304614e-05, "loss": 0.0786, "step": 20905 }, { "epoch": 0.98, "learning_rate": 1.0218916643278137e-05, "loss": 0.0964, "step": 20910 }, { "epoch": 0.98, "learning_rate": 1.021657779025166e-05, "loss": 0.0876, "step": 20915 }, { "epoch": 0.98, "learning_rate": 1.0214238937225185e-05, "loss": 0.058, "step": 20920 }, { "epoch": 0.98, "learning_rate": 1.021190008419871e-05, "loss": 0.0467, "step": 20925 }, { "epoch": 0.98, "learning_rate": 1.0209561231172233e-05, "loss": 0.0826, "step": 20930 }, { "epoch": 0.98, "learning_rate": 1.0207222378145758e-05, "loss": 0.0571, "step": 20935 }, { "epoch": 0.98, "learning_rate": 1.0204883525119283e-05, "loss": 0.0863, "step": 20940 }, { "epoch": 0.98, "learning_rate": 1.0202544672092806e-05, "loss": 0.0764, "step": 20945 }, { "epoch": 0.98, "learning_rate": 1.0200205819066331e-05, "loss": 0.0683, "step": 20950 }, { "epoch": 0.98, "learning_rate": 1.0197866966039856e-05, "loss": 0.109, "step": 20955 }, { "epoch": 0.98, "learning_rate": 1.019552811301338e-05, "loss": 0.0997, "step": 20960 }, { "epoch": 0.98, "learning_rate": 1.0193189259986905e-05, "loss": 0.0585, "step": 20965 }, { "epoch": 0.98, "learning_rate": 1.0190850406960426e-05, "loss": 0.0768, "step": 20970 }, { "epoch": 0.98, "learning_rate": 1.0188511553933951e-05, "loss": 0.1044, "step": 20975 }, { "epoch": 0.98, "learning_rate": 1.0186172700907474e-05, "loss": 0.0872, "step": 20980 }, { "epoch": 0.98, "learning_rate": 1.0183833847881e-05, "loss": 0.1024, "step": 20985 }, { "epoch": 0.98, "learning_rate": 1.0181494994854524e-05, "loss": 0.0609, "step": 20990 }, { "epoch": 0.98, "learning_rate": 1.0179156141828047e-05, "loss": 0.1353, "step": 20995 }, { "epoch": 0.98, "learning_rate": 1.0176817288801572e-05, "loss": 0.1472, "step": 21000 }, { "epoch": 0.98, "learning_rate": 1.0174478435775097e-05, "loss": 0.0688, "step": 21005 }, { "epoch": 0.98, "learning_rate": 1.017213958274862e-05, "loss": 0.0844, "step": 21010 }, { "epoch": 0.98, "learning_rate": 1.0169800729722146e-05, "loss": 0.0506, "step": 21015 }, { "epoch": 0.98, "learning_rate": 1.016746187669567e-05, "loss": 0.0905, "step": 21020 }, { "epoch": 0.98, "learning_rate": 1.0165123023669194e-05, "loss": 0.1108, "step": 21025 }, { "epoch": 0.98, "learning_rate": 1.0162784170642717e-05, "loss": 0.1111, "step": 21030 }, { "epoch": 0.98, "learning_rate": 1.016044531761624e-05, "loss": 0.0725, "step": 21035 }, { "epoch": 0.98, "learning_rate": 1.0158106464589765e-05, "loss": 0.0552, "step": 21040 }, { "epoch": 0.98, "learning_rate": 1.015576761156329e-05, "loss": 0.096, "step": 21045 }, { "epoch": 0.98, "learning_rate": 1.0153428758536813e-05, "loss": 0.0889, "step": 21050 }, { "epoch": 0.98, "learning_rate": 1.0151089905510338e-05, "loss": 0.0766, "step": 21055 }, { "epoch": 0.99, "learning_rate": 1.0148751052483863e-05, "loss": 0.1207, "step": 21060 }, { "epoch": 0.99, "learning_rate": 1.0146412199457387e-05, "loss": 0.1143, "step": 21065 }, { "epoch": 0.99, "learning_rate": 1.0144073346430912e-05, "loss": 0.0747, "step": 21070 }, { "epoch": 0.99, "learning_rate": 1.0141734493404437e-05, "loss": 0.0573, "step": 21075 }, { "epoch": 0.99, "learning_rate": 1.013939564037796e-05, "loss": 0.0728, "step": 21080 }, { "epoch": 0.99, "learning_rate": 1.0137056787351485e-05, "loss": 0.0804, "step": 21085 }, { "epoch": 0.99, "learning_rate": 1.0134717934325006e-05, "loss": 0.1244, "step": 21090 }, { "epoch": 0.99, "learning_rate": 1.0132379081298531e-05, "loss": 0.0953, "step": 21095 }, { "epoch": 0.99, "learning_rate": 1.0130040228272056e-05, "loss": 0.0994, "step": 21100 }, { "epoch": 0.99, "learning_rate": 1.012770137524558e-05, "loss": 0.0899, "step": 21105 }, { "epoch": 0.99, "learning_rate": 1.0125362522219104e-05, "loss": 0.0654, "step": 21110 }, { "epoch": 0.99, "learning_rate": 1.012302366919263e-05, "loss": 0.063, "step": 21115 }, { "epoch": 0.99, "learning_rate": 1.0120684816166153e-05, "loss": 0.0447, "step": 21120 }, { "epoch": 0.99, "learning_rate": 1.0118345963139678e-05, "loss": 0.0493, "step": 21125 }, { "epoch": 0.99, "learning_rate": 1.0116007110113203e-05, "loss": 0.1001, "step": 21130 }, { "epoch": 0.99, "learning_rate": 1.0113668257086726e-05, "loss": 0.1156, "step": 21135 }, { "epoch": 0.99, "learning_rate": 1.011132940406025e-05, "loss": 0.1033, "step": 21140 }, { "epoch": 0.99, "learning_rate": 1.0108990551033776e-05, "loss": 0.0609, "step": 21145 }, { "epoch": 0.99, "learning_rate": 1.0106651698007297e-05, "loss": 0.0656, "step": 21150 }, { "epoch": 0.99, "learning_rate": 1.0104312844980822e-05, "loss": 0.1627, "step": 21155 }, { "epoch": 0.99, "learning_rate": 1.0101973991954345e-05, "loss": 0.0779, "step": 21160 }, { "epoch": 0.99, "learning_rate": 1.009963513892787e-05, "loss": 0.0882, "step": 21165 }, { "epoch": 0.99, "learning_rate": 1.0097296285901395e-05, "loss": 0.0343, "step": 21170 }, { "epoch": 0.99, "learning_rate": 1.0094957432874919e-05, "loss": 0.0625, "step": 21175 }, { "epoch": 0.99, "learning_rate": 1.0092618579848444e-05, "loss": 0.1352, "step": 21180 }, { "epoch": 0.99, "learning_rate": 1.0090279726821967e-05, "loss": 0.0933, "step": 21185 }, { "epoch": 0.99, "learning_rate": 1.0087940873795492e-05, "loss": 0.0901, "step": 21190 }, { "epoch": 0.99, "learning_rate": 1.0085602020769017e-05, "loss": 0.1208, "step": 21195 }, { "epoch": 0.99, "learning_rate": 1.008326316774254e-05, "loss": 0.1149, "step": 21200 }, { "epoch": 0.99, "learning_rate": 1.0080924314716065e-05, "loss": 0.0644, "step": 21205 }, { "epoch": 0.99, "learning_rate": 1.0078585461689587e-05, "loss": 0.0503, "step": 21210 }, { "epoch": 0.99, "learning_rate": 1.0076246608663111e-05, "loss": 0.0913, "step": 21215 }, { "epoch": 0.99, "learning_rate": 1.0073907755636636e-05, "loss": 0.1226, "step": 21220 }, { "epoch": 0.99, "learning_rate": 1.007156890261016e-05, "loss": 0.1281, "step": 21225 }, { "epoch": 0.99, "learning_rate": 1.0069230049583685e-05, "loss": 0.1136, "step": 21230 }, { "epoch": 0.99, "learning_rate": 1.006689119655721e-05, "loss": 0.0894, "step": 21235 }, { "epoch": 0.99, "learning_rate": 1.0064552343530733e-05, "loss": 0.1818, "step": 21240 }, { "epoch": 0.99, "learning_rate": 1.0062213490504258e-05, "loss": 0.097, "step": 21245 }, { "epoch": 0.99, "learning_rate": 1.0059874637477783e-05, "loss": 0.0852, "step": 21250 }, { "epoch": 0.99, "learning_rate": 1.0057535784451306e-05, "loss": 0.0341, "step": 21255 }, { "epoch": 0.99, "learning_rate": 1.0055196931424831e-05, "loss": 0.0925, "step": 21260 }, { "epoch": 0.99, "learning_rate": 1.0052858078398356e-05, "loss": 0.0938, "step": 21265 }, { "epoch": 0.99, "learning_rate": 1.0050519225371877e-05, "loss": 0.0433, "step": 21270 }, { "epoch": 1.0, "learning_rate": 1.0048180372345402e-05, "loss": 0.0331, "step": 21275 }, { "epoch": 1.0, "learning_rate": 1.0045841519318926e-05, "loss": 0.0986, "step": 21280 }, { "epoch": 1.0, "learning_rate": 1.004350266629245e-05, "loss": 0.1132, "step": 21285 }, { "epoch": 1.0, "learning_rate": 1.0041163813265976e-05, "loss": 0.1202, "step": 21290 }, { "epoch": 1.0, "learning_rate": 1.0038824960239499e-05, "loss": 0.0801, "step": 21295 }, { "epoch": 1.0, "learning_rate": 1.0036486107213024e-05, "loss": 0.0892, "step": 21300 }, { "epoch": 1.0, "learning_rate": 1.0034147254186549e-05, "loss": 0.0999, "step": 21305 }, { "epoch": 1.0, "learning_rate": 1.0031808401160072e-05, "loss": 0.0952, "step": 21310 }, { "epoch": 1.0, "learning_rate": 1.0029469548133597e-05, "loss": 0.1392, "step": 21315 }, { "epoch": 1.0, "learning_rate": 1.0027130695107122e-05, "loss": 0.1003, "step": 21320 }, { "epoch": 1.0, "learning_rate": 1.0024791842080643e-05, "loss": 0.0822, "step": 21325 }, { "epoch": 1.0, "learning_rate": 1.0022452989054168e-05, "loss": 0.0711, "step": 21330 }, { "epoch": 1.0, "learning_rate": 1.0020114136027692e-05, "loss": 0.0362, "step": 21335 }, { "epoch": 1.0, "learning_rate": 1.0017775283001217e-05, "loss": 0.0576, "step": 21340 }, { "epoch": 1.0, "learning_rate": 1.0015436429974742e-05, "loss": 0.1084, "step": 21345 }, { "epoch": 1.0, "learning_rate": 1.0013097576948265e-05, "loss": 0.1131, "step": 21350 }, { "epoch": 1.0, "learning_rate": 1.001075872392179e-05, "loss": 0.1067, "step": 21355 }, { "epoch": 1.0, "learning_rate": 1.0008419870895315e-05, "loss": 0.0266, "step": 21360 }, { "epoch": 1.0, "learning_rate": 1.0006081017868838e-05, "loss": 0.1252, "step": 21365 }, { "epoch": 1.0, "learning_rate": 1.0003742164842363e-05, "loss": 0.0816, "step": 21370 }, { "epoch": 1.0, "learning_rate": 1.0001403311815888e-05, "loss": 0.12, "step": 21375 }, { "epoch": 1.0, "eval_method_name": "train_2023-04-14_171922_bbb7108a-dad7-11ed-934b-3cecef280fae", "eval_missing_count": 0, "eval_test_avg": 81.23166666666667, "eval_test_cite__map": 91.76, "eval_test_cite__ndcg": 96.49, "eval_test_co-cite__map": 90.28, "eval_test_co-cite__ndcg": 95.81, "eval_test_co-read__map": 86.78, "eval_test_co-read__ndcg": 93.46, "eval_test_co-view__map": 84.75, "eval_test_co-view__ndcg": 92.06, "eval_test_mag__f1": 79.95, "eval_test_mesh__f1": 89.85, "eval_test_recomm__adj-NDCG": 54.23, "eval_test_recomm__adj-P@1": 19.36, "eval_val_avg": 81.05083333333333, "eval_val_cite__map": 91.19, "eval_val_cite__ndcg": 96.21, "eval_val_co-cite__map": 89.98, "eval_val_co-cite__ndcg": 95.66, "eval_val_co-read__map": 86.79, "eval_val_co-read__ndcg": 93.39, "eval_val_co-view__map": 83.33, "eval_val_co-view__ndcg": 91.04, "eval_val_mag__f1": 79.4, "eval_val_mesh__f1": 90.25, "eval_val_recomm__adj-NDCG": 54.52, "eval_val_recomm__adj-P@1": 20.85, "step": 21378 }, { "epoch": 1.0, "learning_rate": 9.999064458789411e-06, "loss": 0.065, "step": 21380 }, { "epoch": 1.0, "learning_rate": 9.996725605762934e-06, "loss": 0.0771, "step": 21385 }, { "epoch": 1.0, "learning_rate": 9.99438675273646e-06, "loss": 0.0212, "step": 21390 }, { "epoch": 1.0, "learning_rate": 9.992047899709984e-06, "loss": 0.075, "step": 21395 }, { "epoch": 1.0, "learning_rate": 9.989709046683508e-06, "loss": 0.0207, "step": 21400 }, { "epoch": 1.0, "learning_rate": 9.98737019365703e-06, "loss": 0.042, "step": 21405 }, { "epoch": 1.0, "learning_rate": 9.985031340630556e-06, "loss": 0.0392, "step": 21410 }, { "epoch": 1.0, "learning_rate": 9.982692487604079e-06, "loss": 0.0779, "step": 21415 }, { "epoch": 1.0, "learning_rate": 9.980353634577604e-06, "loss": 0.0679, "step": 21420 }, { "epoch": 1.0, "learning_rate": 9.978014781551129e-06, "loss": 0.0228, "step": 21425 }, { "epoch": 1.0, "learning_rate": 9.975675928524652e-06, "loss": 0.0468, "step": 21430 }, { "epoch": 1.0, "learning_rate": 9.973337075498175e-06, "loss": 0.0454, "step": 21435 }, { "epoch": 1.0, "learning_rate": 9.9709982224717e-06, "loss": 0.0785, "step": 21440 }, { "epoch": 1.0, "learning_rate": 9.968659369445225e-06, "loss": 0.0896, "step": 21445 }, { "epoch": 1.0, "learning_rate": 9.966320516418749e-06, "loss": 0.0674, "step": 21450 }, { "epoch": 1.0, "learning_rate": 9.963981663392274e-06, "loss": 0.079, "step": 21455 }, { "epoch": 1.0, "learning_rate": 9.961642810365797e-06, "loss": 0.0607, "step": 21460 }, { "epoch": 1.0, "learning_rate": 9.959303957339322e-06, "loss": 0.0463, "step": 21465 }, { "epoch": 1.0, "learning_rate": 9.956965104312845e-06, "loss": 0.0323, "step": 21470 }, { "epoch": 1.0, "learning_rate": 9.95462625128637e-06, "loss": 0.0726, "step": 21475 }, { "epoch": 1.0, "learning_rate": 9.952287398259895e-06, "loss": 0.0538, "step": 21480 }, { "epoch": 1.01, "learning_rate": 9.949948545233418e-06, "loss": 0.0736, "step": 21485 }, { "epoch": 1.01, "learning_rate": 9.947609692206941e-06, "loss": 0.0231, "step": 21490 }, { "epoch": 1.01, "learning_rate": 9.945270839180466e-06, "loss": 0.051, "step": 21495 }, { "epoch": 1.01, "learning_rate": 9.942931986153991e-06, "loss": 0.0484, "step": 21500 }, { "epoch": 1.01, "learning_rate": 9.940593133127515e-06, "loss": 0.0784, "step": 21505 }, { "epoch": 1.01, "learning_rate": 9.93825428010104e-06, "loss": 0.0934, "step": 21510 }, { "epoch": 1.01, "learning_rate": 9.935915427074564e-06, "loss": 0.0309, "step": 21515 }, { "epoch": 1.01, "learning_rate": 9.933576574048088e-06, "loss": 0.0512, "step": 21520 }, { "epoch": 1.01, "learning_rate": 9.931237721021611e-06, "loss": 0.059, "step": 21525 }, { "epoch": 1.01, "learning_rate": 9.928898867995136e-06, "loss": 0.0462, "step": 21530 }, { "epoch": 1.01, "learning_rate": 9.926560014968661e-06, "loss": 0.0665, "step": 21535 }, { "epoch": 1.01, "learning_rate": 9.924221161942184e-06, "loss": 0.0923, "step": 21540 }, { "epoch": 1.01, "learning_rate": 9.921882308915709e-06, "loss": 0.101, "step": 21545 }, { "epoch": 1.01, "learning_rate": 9.919543455889232e-06, "loss": 0.0452, "step": 21550 }, { "epoch": 1.01, "learning_rate": 9.917204602862757e-06, "loss": 0.0604, "step": 21555 }, { "epoch": 1.01, "learning_rate": 9.91486574983628e-06, "loss": 0.0509, "step": 21560 }, { "epoch": 1.01, "learning_rate": 9.912526896809806e-06, "loss": 0.0679, "step": 21565 }, { "epoch": 1.01, "learning_rate": 9.91018804378333e-06, "loss": 0.0381, "step": 21570 }, { "epoch": 1.01, "learning_rate": 9.907849190756854e-06, "loss": 0.0778, "step": 21575 }, { "epoch": 1.01, "learning_rate": 9.905510337730377e-06, "loss": 0.0737, "step": 21580 }, { "epoch": 1.01, "learning_rate": 9.903171484703902e-06, "loss": 0.0363, "step": 21585 }, { "epoch": 1.01, "learning_rate": 9.900832631677427e-06, "loss": 0.0646, "step": 21590 }, { "epoch": 1.01, "learning_rate": 9.89849377865095e-06, "loss": 0.044, "step": 21595 }, { "epoch": 1.01, "learning_rate": 9.896154925624475e-06, "loss": 0.0899, "step": 21600 }, { "epoch": 1.01, "learning_rate": 9.893816072598e-06, "loss": 0.0677, "step": 21605 }, { "epoch": 1.01, "learning_rate": 9.891477219571523e-06, "loss": 0.0922, "step": 21610 }, { "epoch": 1.01, "learning_rate": 9.889138366545047e-06, "loss": 0.0387, "step": 21615 }, { "epoch": 1.01, "learning_rate": 9.886799513518572e-06, "loss": 0.0338, "step": 21620 }, { "epoch": 1.01, "learning_rate": 9.884460660492095e-06, "loss": 0.0531, "step": 21625 }, { "epoch": 1.01, "learning_rate": 9.88212180746562e-06, "loss": 0.0718, "step": 21630 }, { "epoch": 1.01, "learning_rate": 9.879782954439145e-06, "loss": 0.0522, "step": 21635 }, { "epoch": 1.01, "learning_rate": 9.877444101412668e-06, "loss": 0.0519, "step": 21640 }, { "epoch": 1.01, "learning_rate": 9.875105248386191e-06, "loss": 0.0939, "step": 21645 }, { "epoch": 1.01, "learning_rate": 9.872766395359716e-06, "loss": 0.076, "step": 21650 }, { "epoch": 1.01, "learning_rate": 9.870427542333241e-06, "loss": 0.0701, "step": 21655 }, { "epoch": 1.01, "learning_rate": 9.868088689306764e-06, "loss": 0.0329, "step": 21660 }, { "epoch": 1.01, "learning_rate": 9.86574983628029e-06, "loss": 0.0755, "step": 21665 }, { "epoch": 1.01, "learning_rate": 9.863410983253813e-06, "loss": 0.0638, "step": 21670 }, { "epoch": 1.01, "learning_rate": 9.861072130227338e-06, "loss": 0.1068, "step": 21675 }, { "epoch": 1.01, "learning_rate": 9.85873327720086e-06, "loss": 0.0369, "step": 21680 }, { "epoch": 1.01, "learning_rate": 9.856394424174386e-06, "loss": 0.0209, "step": 21685 }, { "epoch": 1.01, "learning_rate": 9.85405557114791e-06, "loss": 0.0498, "step": 21690 }, { "epoch": 1.01, "learning_rate": 9.851716718121434e-06, "loss": 0.0809, "step": 21695 }, { "epoch": 1.02, "learning_rate": 9.849377865094957e-06, "loss": 0.0337, "step": 21700 }, { "epoch": 1.02, "learning_rate": 9.847039012068482e-06, "loss": 0.0359, "step": 21705 }, { "epoch": 1.02, "learning_rate": 9.844700159042007e-06, "loss": 0.0898, "step": 21710 }, { "epoch": 1.02, "learning_rate": 9.84236130601553e-06, "loss": 0.0525, "step": 21715 }, { "epoch": 1.02, "learning_rate": 9.840022452989055e-06, "loss": 0.0453, "step": 21720 }, { "epoch": 1.02, "learning_rate": 9.83768359996258e-06, "loss": 0.049, "step": 21725 }, { "epoch": 1.02, "learning_rate": 9.835344746936104e-06, "loss": 0.0231, "step": 21730 }, { "epoch": 1.02, "learning_rate": 9.833005893909627e-06, "loss": 0.0453, "step": 21735 }, { "epoch": 1.02, "learning_rate": 9.830667040883152e-06, "loss": 0.0288, "step": 21740 }, { "epoch": 1.02, "learning_rate": 9.828328187856677e-06, "loss": 0.0709, "step": 21745 }, { "epoch": 1.02, "learning_rate": 9.8259893348302e-06, "loss": 0.0413, "step": 21750 }, { "epoch": 1.02, "learning_rate": 9.823650481803723e-06, "loss": 0.0382, "step": 21755 }, { "epoch": 1.02, "learning_rate": 9.821311628777248e-06, "loss": 0.0839, "step": 21760 }, { "epoch": 1.02, "learning_rate": 9.818972775750773e-06, "loss": 0.0709, "step": 21765 }, { "epoch": 1.02, "learning_rate": 9.816633922724296e-06, "loss": 0.0636, "step": 21770 }, { "epoch": 1.02, "learning_rate": 9.814295069697821e-06, "loss": 0.0365, "step": 21775 }, { "epoch": 1.02, "learning_rate": 9.811956216671346e-06, "loss": 0.03, "step": 21780 }, { "epoch": 1.02, "learning_rate": 9.80961736364487e-06, "loss": 0.029, "step": 21785 }, { "epoch": 1.02, "learning_rate": 9.807278510618393e-06, "loss": 0.0811, "step": 21790 }, { "epoch": 1.02, "learning_rate": 9.804939657591918e-06, "loss": 0.0621, "step": 21795 }, { "epoch": 1.02, "learning_rate": 9.802600804565443e-06, "loss": 0.055, "step": 21800 }, { "epoch": 1.02, "learning_rate": 9.800261951538966e-06, "loss": 0.0677, "step": 21805 }, { "epoch": 1.02, "learning_rate": 9.797923098512491e-06, "loss": 0.0315, "step": 21810 }, { "epoch": 1.02, "learning_rate": 9.795584245486014e-06, "loss": 0.0681, "step": 21815 }, { "epoch": 1.02, "learning_rate": 9.793245392459539e-06, "loss": 0.0841, "step": 21820 }, { "epoch": 1.02, "learning_rate": 9.790906539433062e-06, "loss": 0.0438, "step": 21825 }, { "epoch": 1.02, "learning_rate": 9.788567686406587e-06, "loss": 0.0768, "step": 21830 }, { "epoch": 1.02, "learning_rate": 9.786228833380112e-06, "loss": 0.0764, "step": 21835 }, { "epoch": 1.02, "learning_rate": 9.783889980353636e-06, "loss": 0.1343, "step": 21840 }, { "epoch": 1.02, "learning_rate": 9.781551127327159e-06, "loss": 0.0624, "step": 21845 }, { "epoch": 1.02, "learning_rate": 9.779212274300684e-06, "loss": 0.0857, "step": 21850 }, { "epoch": 1.02, "learning_rate": 9.776873421274207e-06, "loss": 0.0875, "step": 21855 }, { "epoch": 1.02, "learning_rate": 9.774534568247732e-06, "loss": 0.0339, "step": 21860 }, { "epoch": 1.02, "learning_rate": 9.772195715221257e-06, "loss": 0.0201, "step": 21865 }, { "epoch": 1.02, "learning_rate": 9.76985686219478e-06, "loss": 0.0734, "step": 21870 }, { "epoch": 1.02, "learning_rate": 9.767518009168303e-06, "loss": 0.0319, "step": 21875 }, { "epoch": 1.02, "learning_rate": 9.765179156141828e-06, "loss": 0.0524, "step": 21880 }, { "epoch": 1.02, "learning_rate": 9.762840303115353e-06, "loss": 0.0475, "step": 21885 }, { "epoch": 1.02, "learning_rate": 9.760501450088877e-06, "loss": 0.03, "step": 21890 }, { "epoch": 1.02, "learning_rate": 9.758162597062402e-06, "loss": 0.0599, "step": 21895 }, { "epoch": 1.02, "learning_rate": 9.755823744035926e-06, "loss": 0.073, "step": 21900 }, { "epoch": 1.02, "learning_rate": 9.75348489100945e-06, "loss": 0.0919, "step": 21905 }, { "epoch": 1.02, "learning_rate": 9.751146037982973e-06, "loss": 0.0805, "step": 21910 }, { "epoch": 1.03, "learning_rate": 9.748807184956498e-06, "loss": 0.0527, "step": 21915 }, { "epoch": 1.03, "learning_rate": 9.746468331930023e-06, "loss": 0.1114, "step": 21920 }, { "epoch": 1.03, "learning_rate": 9.744129478903546e-06, "loss": 0.1029, "step": 21925 }, { "epoch": 1.03, "learning_rate": 9.741790625877071e-06, "loss": 0.0235, "step": 21930 }, { "epoch": 1.03, "learning_rate": 9.739451772850594e-06, "loss": 0.039, "step": 21935 }, { "epoch": 1.03, "learning_rate": 9.73711291982412e-06, "loss": 0.0502, "step": 21940 }, { "epoch": 1.03, "learning_rate": 9.734774066797643e-06, "loss": 0.0488, "step": 21945 }, { "epoch": 1.03, "learning_rate": 9.732435213771168e-06, "loss": 0.023, "step": 21950 }, { "epoch": 1.03, "learning_rate": 9.730096360744692e-06, "loss": 0.1572, "step": 21955 }, { "epoch": 1.03, "learning_rate": 9.727757507718216e-06, "loss": 0.0871, "step": 21960 }, { "epoch": 1.03, "learning_rate": 9.725418654691739e-06, "loss": 0.0443, "step": 21965 }, { "epoch": 1.03, "learning_rate": 9.723079801665264e-06, "loss": 0.0239, "step": 21970 }, { "epoch": 1.03, "learning_rate": 9.720740948638789e-06, "loss": 0.0607, "step": 21975 }, { "epoch": 1.03, "learning_rate": 9.718402095612312e-06, "loss": 0.0807, "step": 21980 }, { "epoch": 1.03, "learning_rate": 9.716063242585837e-06, "loss": 0.0555, "step": 21985 }, { "epoch": 1.03, "learning_rate": 9.713724389559362e-06, "loss": 0.0604, "step": 21990 }, { "epoch": 1.03, "learning_rate": 9.711385536532885e-06, "loss": 0.0255, "step": 21995 }, { "epoch": 1.03, "learning_rate": 9.709046683506409e-06, "loss": 0.0347, "step": 22000 }, { "epoch": 1.03, "learning_rate": 9.706707830479934e-06, "loss": 0.0561, "step": 22005 }, { "epoch": 1.03, "learning_rate": 9.704368977453458e-06, "loss": 0.0794, "step": 22010 }, { "epoch": 1.03, "learning_rate": 9.702030124426982e-06, "loss": 0.0765, "step": 22015 }, { "epoch": 1.03, "learning_rate": 9.699691271400507e-06, "loss": 0.1234, "step": 22020 }, { "epoch": 1.03, "learning_rate": 9.69735241837403e-06, "loss": 0.0618, "step": 22025 }, { "epoch": 1.03, "learning_rate": 9.695013565347555e-06, "loss": 0.0239, "step": 22030 }, { "epoch": 1.03, "learning_rate": 9.692674712321078e-06, "loss": 0.0458, "step": 22035 }, { "epoch": 1.03, "learning_rate": 9.690335859294603e-06, "loss": 0.0605, "step": 22040 }, { "epoch": 1.03, "learning_rate": 9.687997006268128e-06, "loss": 0.0959, "step": 22045 }, { "epoch": 1.03, "learning_rate": 9.685658153241651e-06, "loss": 0.0469, "step": 22050 }, { "epoch": 1.03, "learning_rate": 9.683319300215175e-06, "loss": 0.0303, "step": 22055 }, { "epoch": 1.03, "learning_rate": 9.6809804471887e-06, "loss": 0.0308, "step": 22060 }, { "epoch": 1.03, "learning_rate": 9.678641594162223e-06, "loss": 0.0601, "step": 22065 }, { "epoch": 1.03, "learning_rate": 9.676302741135748e-06, "loss": 0.0226, "step": 22070 }, { "epoch": 1.03, "learning_rate": 9.673963888109273e-06, "loss": 0.0594, "step": 22075 }, { "epoch": 1.03, "learning_rate": 9.671625035082796e-06, "loss": 0.0172, "step": 22080 }, { "epoch": 1.03, "learning_rate": 9.66928618205632e-06, "loss": 0.071, "step": 22085 }, { "epoch": 1.03, "learning_rate": 9.666947329029844e-06, "loss": 0.0866, "step": 22090 }, { "epoch": 1.03, "learning_rate": 9.664608476003369e-06, "loss": 0.0691, "step": 22095 }, { "epoch": 1.03, "learning_rate": 9.662269622976892e-06, "loss": 0.0324, "step": 22100 }, { "epoch": 1.03, "learning_rate": 9.659930769950417e-06, "loss": 0.0845, "step": 22105 }, { "epoch": 1.03, "learning_rate": 9.657591916923942e-06, "loss": 0.0928, "step": 22110 }, { "epoch": 1.03, "learning_rate": 9.655253063897466e-06, "loss": 0.0368, "step": 22115 }, { "epoch": 1.03, "learning_rate": 9.652914210870989e-06, "loss": 0.0634, "step": 22120 }, { "epoch": 1.03, "learning_rate": 9.650575357844514e-06, "loss": 0.0401, "step": 22125 }, { "epoch": 1.04, "learning_rate": 9.648236504818039e-06, "loss": 0.0538, "step": 22130 }, { "epoch": 1.04, "learning_rate": 9.645897651791562e-06, "loss": 0.0291, "step": 22135 }, { "epoch": 1.04, "learning_rate": 9.643558798765087e-06, "loss": 0.0195, "step": 22140 }, { "epoch": 1.04, "learning_rate": 9.64121994573861e-06, "loss": 0.0519, "step": 22145 }, { "epoch": 1.04, "learning_rate": 9.638881092712135e-06, "loss": 0.1182, "step": 22150 }, { "epoch": 1.04, "learning_rate": 9.636542239685658e-06, "loss": 0.0482, "step": 22155 }, { "epoch": 1.04, "learning_rate": 9.634203386659183e-06, "loss": 0.0457, "step": 22160 }, { "epoch": 1.04, "learning_rate": 9.631864533632708e-06, "loss": 0.0726, "step": 22165 }, { "epoch": 1.04, "learning_rate": 9.629525680606232e-06, "loss": 0.0389, "step": 22170 }, { "epoch": 1.04, "learning_rate": 9.627186827579755e-06, "loss": 0.1145, "step": 22175 }, { "epoch": 1.04, "learning_rate": 9.62484797455328e-06, "loss": 0.0527, "step": 22180 }, { "epoch": 1.04, "learning_rate": 9.622509121526805e-06, "loss": 0.063, "step": 22185 }, { "epoch": 1.04, "learning_rate": 9.620170268500328e-06, "loss": 0.0344, "step": 22190 }, { "epoch": 1.04, "learning_rate": 9.617831415473853e-06, "loss": 0.0665, "step": 22195 }, { "epoch": 1.04, "learning_rate": 9.615492562447376e-06, "loss": 0.0519, "step": 22200 }, { "epoch": 1.04, "learning_rate": 9.613153709420901e-06, "loss": 0.0482, "step": 22205 }, { "epoch": 1.04, "learning_rate": 9.610814856394424e-06, "loss": 0.0255, "step": 22210 }, { "epoch": 1.04, "learning_rate": 9.60847600336795e-06, "loss": 0.0768, "step": 22215 }, { "epoch": 1.04, "learning_rate": 9.606137150341474e-06, "loss": 0.0592, "step": 22220 }, { "epoch": 1.04, "learning_rate": 9.603798297314997e-06, "loss": 0.0384, "step": 22225 }, { "epoch": 1.04, "learning_rate": 9.60145944428852e-06, "loss": 0.0408, "step": 22230 }, { "epoch": 1.04, "learning_rate": 9.599120591262046e-06, "loss": 0.0165, "step": 22235 }, { "epoch": 1.04, "learning_rate": 9.59678173823557e-06, "loss": 0.0651, "step": 22240 }, { "epoch": 1.04, "learning_rate": 9.594442885209094e-06, "loss": 0.0537, "step": 22245 }, { "epoch": 1.04, "learning_rate": 9.592104032182619e-06, "loss": 0.0652, "step": 22250 }, { "epoch": 1.04, "learning_rate": 9.589765179156144e-06, "loss": 0.059, "step": 22255 }, { "epoch": 1.04, "learning_rate": 9.587426326129667e-06, "loss": 0.0506, "step": 22260 }, { "epoch": 1.04, "learning_rate": 9.58508747310319e-06, "loss": 0.096, "step": 22265 }, { "epoch": 1.04, "learning_rate": 9.582748620076715e-06, "loss": 0.0866, "step": 22270 }, { "epoch": 1.04, "learning_rate": 9.58040976705024e-06, "loss": 0.0291, "step": 22275 }, { "epoch": 1.04, "learning_rate": 9.578070914023763e-06, "loss": 0.0819, "step": 22280 }, { "epoch": 1.04, "learning_rate": 9.575732060997288e-06, "loss": 0.0476, "step": 22285 }, { "epoch": 1.04, "learning_rate": 9.573393207970812e-06, "loss": 0.0788, "step": 22290 }, { "epoch": 1.04, "learning_rate": 9.571054354944335e-06, "loss": 0.0676, "step": 22295 }, { "epoch": 1.04, "learning_rate": 9.56871550191786e-06, "loss": 0.0736, "step": 22300 }, { "epoch": 1.04, "learning_rate": 9.566376648891385e-06, "loss": 0.0774, "step": 22305 }, { "epoch": 1.04, "learning_rate": 9.564037795864908e-06, "loss": 0.0721, "step": 22310 }, { "epoch": 1.04, "learning_rate": 9.561698942838433e-06, "loss": 0.0709, "step": 22315 }, { "epoch": 1.04, "learning_rate": 9.559360089811956e-06, "loss": 0.0479, "step": 22320 }, { "epoch": 1.04, "learning_rate": 9.557021236785481e-06, "loss": 0.0366, "step": 22325 }, { "epoch": 1.04, "learning_rate": 9.554682383759005e-06, "loss": 0.0239, "step": 22330 }, { "epoch": 1.04, "learning_rate": 9.55234353073253e-06, "loss": 0.0636, "step": 22335 }, { "epoch": 1.04, "learning_rate": 9.550004677706054e-06, "loss": 0.0467, "step": 22340 }, { "epoch": 1.05, "learning_rate": 9.547665824679578e-06, "loss": 0.0546, "step": 22345 }, { "epoch": 1.05, "learning_rate": 9.545326971653101e-06, "loss": 0.0235, "step": 22350 }, { "epoch": 1.05, "learning_rate": 9.542988118626626e-06, "loss": 0.0629, "step": 22355 }, { "epoch": 1.05, "learning_rate": 9.540649265600151e-06, "loss": 0.0753, "step": 22360 }, { "epoch": 1.05, "learning_rate": 9.538310412573674e-06, "loss": 0.0405, "step": 22365 }, { "epoch": 1.05, "learning_rate": 9.535971559547199e-06, "loss": 0.0471, "step": 22370 }, { "epoch": 1.05, "learning_rate": 9.533632706520724e-06, "loss": 0.0501, "step": 22375 }, { "epoch": 1.05, "learning_rate": 9.531293853494247e-06, "loss": 0.1478, "step": 22380 }, { "epoch": 1.05, "learning_rate": 9.52895500046777e-06, "loss": 0.0198, "step": 22385 }, { "epoch": 1.05, "learning_rate": 9.526616147441295e-06, "loss": 0.0411, "step": 22390 }, { "epoch": 1.05, "learning_rate": 9.52427729441482e-06, "loss": 0.0788, "step": 22395 }, { "epoch": 1.05, "learning_rate": 9.521938441388344e-06, "loss": 0.0536, "step": 22400 }, { "epoch": 1.05, "learning_rate": 9.519599588361869e-06, "loss": 0.0674, "step": 22405 }, { "epoch": 1.05, "learning_rate": 9.517260735335392e-06, "loss": 0.0779, "step": 22410 }, { "epoch": 1.05, "learning_rate": 9.514921882308917e-06, "loss": 0.1621, "step": 22415 }, { "epoch": 1.05, "learning_rate": 9.51258302928244e-06, "loss": 0.0571, "step": 22420 }, { "epoch": 1.05, "learning_rate": 9.510244176255965e-06, "loss": 0.0517, "step": 22425 }, { "epoch": 1.05, "learning_rate": 9.50790532322949e-06, "loss": 0.0405, "step": 22430 }, { "epoch": 1.05, "learning_rate": 9.505566470203013e-06, "loss": 0.0821, "step": 22435 }, { "epoch": 1.05, "learning_rate": 9.503227617176537e-06, "loss": 0.085, "step": 22440 }, { "epoch": 1.05, "learning_rate": 9.500888764150061e-06, "loss": 0.0594, "step": 22445 }, { "epoch": 1.05, "learning_rate": 9.498549911123586e-06, "loss": 0.0462, "step": 22450 }, { "epoch": 1.05, "learning_rate": 9.49621105809711e-06, "loss": 0.0641, "step": 22455 }, { "epoch": 1.05, "learning_rate": 9.493872205070635e-06, "loss": 0.0598, "step": 22460 }, { "epoch": 1.05, "learning_rate": 9.49153335204416e-06, "loss": 0.0472, "step": 22465 }, { "epoch": 1.05, "learning_rate": 9.489194499017683e-06, "loss": 0.0589, "step": 22470 }, { "epoch": 1.05, "learning_rate": 9.486855645991206e-06, "loss": 0.0332, "step": 22475 }, { "epoch": 1.05, "learning_rate": 9.484516792964731e-06, "loss": 0.0373, "step": 22480 }, { "epoch": 1.05, "learning_rate": 9.482177939938256e-06, "loss": 0.0279, "step": 22485 }, { "epoch": 1.05, "learning_rate": 9.47983908691178e-06, "loss": 0.0456, "step": 22490 }, { "epoch": 1.05, "learning_rate": 9.477500233885304e-06, "loss": 0.1008, "step": 22495 }, { "epoch": 1.05, "learning_rate": 9.475161380858827e-06, "loss": 0.078, "step": 22500 }, { "epoch": 1.05, "learning_rate": 9.47282252783235e-06, "loss": 0.0589, "step": 22505 }, { "epoch": 1.05, "learning_rate": 9.470483674805876e-06, "loss": 0.0533, "step": 22510 }, { "epoch": 1.05, "learning_rate": 9.4681448217794e-06, "loss": 0.0764, "step": 22515 }, { "epoch": 1.05, "learning_rate": 9.465805968752924e-06, "loss": 0.0715, "step": 22520 }, { "epoch": 1.05, "learning_rate": 9.463467115726449e-06, "loss": 0.0976, "step": 22525 }, { "epoch": 1.05, "learning_rate": 9.461128262699972e-06, "loss": 0.0739, "step": 22530 }, { "epoch": 1.05, "learning_rate": 9.458789409673497e-06, "loss": 0.0399, "step": 22535 }, { "epoch": 1.05, "learning_rate": 9.45645055664702e-06, "loss": 0.0545, "step": 22540 }, { "epoch": 1.05, "learning_rate": 9.454111703620545e-06, "loss": 0.0542, "step": 22545 }, { "epoch": 1.05, "learning_rate": 9.45177285059407e-06, "loss": 0.0545, "step": 22550 }, { "epoch": 1.06, "learning_rate": 9.449433997567593e-06, "loss": 0.0572, "step": 22555 }, { "epoch": 1.06, "learning_rate": 9.447095144541117e-06, "loss": 0.131, "step": 22560 }, { "epoch": 1.06, "learning_rate": 9.444756291514642e-06, "loss": 0.0469, "step": 22565 }, { "epoch": 1.06, "learning_rate": 9.442417438488167e-06, "loss": 0.024, "step": 22570 }, { "epoch": 1.06, "learning_rate": 9.44007858546169e-06, "loss": 0.0451, "step": 22575 }, { "epoch": 1.06, "learning_rate": 9.437739732435215e-06, "loss": 0.0668, "step": 22580 }, { "epoch": 1.06, "learning_rate": 9.43540087940874e-06, "loss": 0.0683, "step": 22585 }, { "epoch": 1.06, "learning_rate": 9.433062026382263e-06, "loss": 0.0891, "step": 22590 }, { "epoch": 1.06, "learning_rate": 9.430723173355786e-06, "loss": 0.0553, "step": 22595 }, { "epoch": 1.06, "learning_rate": 9.428384320329311e-06, "loss": 0.0542, "step": 22600 }, { "epoch": 1.06, "learning_rate": 9.426045467302836e-06, "loss": 0.0443, "step": 22605 }, { "epoch": 1.06, "learning_rate": 9.42370661427636e-06, "loss": 0.0753, "step": 22610 }, { "epoch": 1.06, "learning_rate": 9.421367761249883e-06, "loss": 0.0829, "step": 22615 }, { "epoch": 1.06, "learning_rate": 9.419028908223408e-06, "loss": 0.0793, "step": 22620 }, { "epoch": 1.06, "learning_rate": 9.416690055196933e-06, "loss": 0.0633, "step": 22625 }, { "epoch": 1.06, "learning_rate": 9.414351202170456e-06, "loss": 0.0633, "step": 22630 }, { "epoch": 1.06, "learning_rate": 9.41201234914398e-06, "loss": 0.0827, "step": 22635 }, { "epoch": 1.06, "learning_rate": 9.409673496117506e-06, "loss": 0.0244, "step": 22640 }, { "epoch": 1.06, "learning_rate": 9.407334643091029e-06, "loss": 0.02, "step": 22645 }, { "epoch": 1.06, "learning_rate": 9.404995790064552e-06, "loss": 0.0059, "step": 22650 }, { "epoch": 1.06, "learning_rate": 9.402656937038077e-06, "loss": 0.0538, "step": 22655 }, { "epoch": 1.06, "learning_rate": 9.400318084011602e-06, "loss": 0.054, "step": 22660 }, { "epoch": 1.06, "learning_rate": 9.397979230985125e-06, "loss": 0.0618, "step": 22665 }, { "epoch": 1.06, "learning_rate": 9.39564037795865e-06, "loss": 0.1001, "step": 22670 }, { "epoch": 1.06, "learning_rate": 9.393301524932174e-06, "loss": 0.0471, "step": 22675 }, { "epoch": 1.06, "learning_rate": 9.390962671905699e-06, "loss": 0.0303, "step": 22680 }, { "epoch": 1.06, "learning_rate": 9.388623818879222e-06, "loss": 0.0297, "step": 22685 }, { "epoch": 1.06, "learning_rate": 9.386284965852747e-06, "loss": 0.0672, "step": 22690 }, { "epoch": 1.06, "learning_rate": 9.383946112826272e-06, "loss": 0.0414, "step": 22695 }, { "epoch": 1.06, "learning_rate": 9.381607259799795e-06, "loss": 0.0802, "step": 22700 }, { "epoch": 1.06, "learning_rate": 9.379268406773318e-06, "loss": 0.0521, "step": 22705 }, { "epoch": 1.06, "learning_rate": 9.376929553746843e-06, "loss": 0.0455, "step": 22710 }, { "epoch": 1.06, "learning_rate": 9.374590700720368e-06, "loss": 0.0672, "step": 22715 }, { "epoch": 1.06, "learning_rate": 9.372251847693891e-06, "loss": 0.1026, "step": 22720 }, { "epoch": 1.06, "learning_rate": 9.369912994667416e-06, "loss": 0.048, "step": 22725 }, { "epoch": 1.06, "learning_rate": 9.36757414164094e-06, "loss": 0.0373, "step": 22730 }, { "epoch": 1.06, "learning_rate": 9.365235288614463e-06, "loss": 0.0747, "step": 22735 }, { "epoch": 1.06, "learning_rate": 9.362896435587988e-06, "loss": 0.0727, "step": 22740 }, { "epoch": 1.06, "learning_rate": 9.360557582561513e-06, "loss": 0.041, "step": 22745 }, { "epoch": 1.06, "learning_rate": 9.358218729535036e-06, "loss": 0.0737, "step": 22750 }, { "epoch": 1.06, "learning_rate": 9.355879876508561e-06, "loss": 0.0873, "step": 22755 }, { "epoch": 1.06, "learning_rate": 9.353541023482086e-06, "loss": 0.0237, "step": 22760 }, { "epoch": 1.06, "learning_rate": 9.35120217045561e-06, "loss": 0.0779, "step": 22765 }, { "epoch": 1.07, "learning_rate": 9.348863317429133e-06, "loss": 0.0751, "step": 22770 }, { "epoch": 1.07, "learning_rate": 9.346524464402657e-06, "loss": 0.1039, "step": 22775 }, { "epoch": 1.07, "learning_rate": 9.344185611376182e-06, "loss": 0.0483, "step": 22780 }, { "epoch": 1.07, "learning_rate": 9.341846758349706e-06, "loss": 0.051, "step": 22785 }, { "epoch": 1.07, "learning_rate": 9.33950790532323e-06, "loss": 0.0696, "step": 22790 }, { "epoch": 1.07, "learning_rate": 9.337169052296754e-06, "loss": 0.0494, "step": 22795 }, { "epoch": 1.07, "learning_rate": 9.334830199270279e-06, "loss": 0.0269, "step": 22800 }, { "epoch": 1.07, "learning_rate": 9.332491346243802e-06, "loss": 0.0895, "step": 22805 }, { "epoch": 1.07, "learning_rate": 9.330152493217327e-06, "loss": 0.0719, "step": 22810 }, { "epoch": 1.07, "learning_rate": 9.327813640190852e-06, "loss": 0.1167, "step": 22815 }, { "epoch": 1.07, "learning_rate": 9.325474787164375e-06, "loss": 0.0571, "step": 22820 }, { "epoch": 1.07, "learning_rate": 9.323135934137899e-06, "loss": 0.0374, "step": 22825 }, { "epoch": 1.07, "learning_rate": 9.320797081111423e-06, "loss": 0.0445, "step": 22830 }, { "epoch": 1.07, "learning_rate": 9.318458228084948e-06, "loss": 0.0401, "step": 22835 }, { "epoch": 1.07, "learning_rate": 9.316119375058472e-06, "loss": 0.0483, "step": 22840 }, { "epoch": 1.07, "learning_rate": 9.313780522031997e-06, "loss": 0.0711, "step": 22845 }, { "epoch": 1.07, "learning_rate": 9.311441669005522e-06, "loss": 0.085, "step": 22850 }, { "epoch": 1.07, "learning_rate": 9.309102815979045e-06, "loss": 0.041, "step": 22855 }, { "epoch": 1.07, "learning_rate": 9.306763962952568e-06, "loss": 0.0434, "step": 22860 }, { "epoch": 1.07, "learning_rate": 9.304425109926093e-06, "loss": 0.0475, "step": 22865 }, { "epoch": 1.07, "learning_rate": 9.302086256899618e-06, "loss": 0.0176, "step": 22870 }, { "epoch": 1.07, "learning_rate": 9.299747403873141e-06, "loss": 0.038, "step": 22875 }, { "epoch": 1.07, "learning_rate": 9.297408550846666e-06, "loss": 0.0535, "step": 22880 }, { "epoch": 1.07, "learning_rate": 9.29506969782019e-06, "loss": 0.0161, "step": 22885 }, { "epoch": 1.07, "learning_rate": 9.292730844793714e-06, "loss": 0.0377, "step": 22890 }, { "epoch": 1.07, "learning_rate": 9.290391991767238e-06, "loss": 0.1081, "step": 22895 }, { "epoch": 1.07, "learning_rate": 9.288053138740763e-06, "loss": 0.0629, "step": 22900 }, { "epoch": 1.07, "learning_rate": 9.285714285714288e-06, "loss": 0.044, "step": 22905 }, { "epoch": 1.07, "learning_rate": 9.28337543268781e-06, "loss": 0.0851, "step": 22910 }, { "epoch": 1.07, "learning_rate": 9.281036579661334e-06, "loss": 0.0353, "step": 22915 }, { "epoch": 1.07, "learning_rate": 9.278697726634859e-06, "loss": 0.0613, "step": 22920 }, { "epoch": 1.07, "learning_rate": 9.276358873608384e-06, "loss": 0.0548, "step": 22925 }, { "epoch": 1.07, "learning_rate": 9.274020020581907e-06, "loss": 0.0637, "step": 22930 }, { "epoch": 1.07, "learning_rate": 9.271681167555432e-06, "loss": 0.088, "step": 22935 }, { "epoch": 1.07, "learning_rate": 9.269342314528955e-06, "loss": 0.1039, "step": 22940 }, { "epoch": 1.07, "learning_rate": 9.267003461502479e-06, "loss": 0.0836, "step": 22945 }, { "epoch": 1.07, "learning_rate": 9.264664608476004e-06, "loss": 0.0504, "step": 22950 }, { "epoch": 1.07, "learning_rate": 9.262325755449529e-06, "loss": 0.0513, "step": 22955 }, { "epoch": 1.07, "learning_rate": 9.259986902423052e-06, "loss": 0.0427, "step": 22960 }, { "epoch": 1.07, "learning_rate": 9.257648049396577e-06, "loss": 0.0581, "step": 22965 }, { "epoch": 1.07, "learning_rate": 9.255309196370102e-06, "loss": 0.0392, "step": 22970 }, { "epoch": 1.07, "learning_rate": 9.252970343343625e-06, "loss": 0.1036, "step": 22975 }, { "epoch": 1.07, "learning_rate": 9.250631490317148e-06, "loss": 0.0835, "step": 22980 }, { "epoch": 1.08, "learning_rate": 9.248292637290673e-06, "loss": 0.1046, "step": 22985 }, { "epoch": 1.08, "learning_rate": 9.245953784264198e-06, "loss": 0.0791, "step": 22990 }, { "epoch": 1.08, "learning_rate": 9.243614931237721e-06, "loss": 0.0761, "step": 22995 }, { "epoch": 1.08, "learning_rate": 9.241276078211246e-06, "loss": 0.0894, "step": 23000 }, { "epoch": 1.08, "learning_rate": 9.23893722518477e-06, "loss": 0.0605, "step": 23005 }, { "epoch": 1.08, "learning_rate": 9.236598372158295e-06, "loss": 0.0416, "step": 23010 }, { "epoch": 1.08, "learning_rate": 9.234259519131818e-06, "loss": 0.0467, "step": 23015 }, { "epoch": 1.08, "learning_rate": 9.231920666105343e-06, "loss": 0.068, "step": 23020 }, { "epoch": 1.08, "learning_rate": 9.229581813078868e-06, "loss": 0.0978, "step": 23025 }, { "epoch": 1.08, "learning_rate": 9.227242960052391e-06, "loss": 0.0479, "step": 23030 }, { "epoch": 1.08, "learning_rate": 9.224904107025914e-06, "loss": 0.0703, "step": 23035 }, { "epoch": 1.08, "learning_rate": 9.22256525399944e-06, "loss": 0.0423, "step": 23040 }, { "epoch": 1.08, "learning_rate": 9.220226400972964e-06, "loss": 0.0395, "step": 23045 }, { "epoch": 1.08, "learning_rate": 9.217887547946487e-06, "loss": 0.0665, "step": 23050 }, { "epoch": 1.08, "learning_rate": 9.215548694920012e-06, "loss": 0.0499, "step": 23055 }, { "epoch": 1.08, "learning_rate": 9.213209841893536e-06, "loss": 0.0479, "step": 23060 }, { "epoch": 1.08, "learning_rate": 9.21087098886706e-06, "loss": 0.0418, "step": 23065 }, { "epoch": 1.08, "learning_rate": 9.208532135840584e-06, "loss": 0.0429, "step": 23070 }, { "epoch": 1.08, "learning_rate": 9.206193282814109e-06, "loss": 0.0933, "step": 23075 }, { "epoch": 1.08, "learning_rate": 9.203854429787634e-06, "loss": 0.059, "step": 23080 }, { "epoch": 1.08, "learning_rate": 9.201515576761157e-06, "loss": 0.0506, "step": 23085 }, { "epoch": 1.08, "learning_rate": 9.19917672373468e-06, "loss": 0.0746, "step": 23090 }, { "epoch": 1.08, "learning_rate": 9.196837870708205e-06, "loss": 0.0171, "step": 23095 }, { "epoch": 1.08, "learning_rate": 9.19449901768173e-06, "loss": 0.146, "step": 23100 }, { "epoch": 1.08, "learning_rate": 9.192160164655253e-06, "loss": 0.0502, "step": 23105 }, { "epoch": 1.08, "learning_rate": 9.189821311628778e-06, "loss": 0.0138, "step": 23110 }, { "epoch": 1.08, "learning_rate": 9.187482458602303e-06, "loss": 0.0715, "step": 23115 }, { "epoch": 1.08, "learning_rate": 9.185143605575827e-06, "loss": 0.0342, "step": 23120 }, { "epoch": 1.08, "learning_rate": 9.18280475254935e-06, "loss": 0.046, "step": 23125 }, { "epoch": 1.08, "learning_rate": 9.180465899522875e-06, "loss": 0.0821, "step": 23130 }, { "epoch": 1.08, "learning_rate": 9.1781270464964e-06, "loss": 0.1063, "step": 23135 }, { "epoch": 1.08, "learning_rate": 9.175788193469923e-06, "loss": 0.0765, "step": 23140 }, { "epoch": 1.08, "learning_rate": 9.173449340443448e-06, "loss": 0.0469, "step": 23145 }, { "epoch": 1.08, "learning_rate": 9.171110487416971e-06, "loss": 0.032, "step": 23150 }, { "epoch": 1.08, "learning_rate": 9.168771634390496e-06, "loss": 0.0993, "step": 23155 }, { "epoch": 1.08, "learning_rate": 9.16643278136402e-06, "loss": 0.0328, "step": 23160 }, { "epoch": 1.08, "learning_rate": 9.164093928337544e-06, "loss": 0.1154, "step": 23165 }, { "epoch": 1.08, "learning_rate": 9.161755075311068e-06, "loss": 0.064, "step": 23170 }, { "epoch": 1.08, "learning_rate": 9.159416222284593e-06, "loss": 0.0322, "step": 23175 }, { "epoch": 1.08, "learning_rate": 9.157077369258116e-06, "loss": 0.0419, "step": 23180 }, { "epoch": 1.08, "learning_rate": 9.15473851623164e-06, "loss": 0.0519, "step": 23185 }, { "epoch": 1.08, "learning_rate": 9.152399663205164e-06, "loss": 0.0853, "step": 23190 }, { "epoch": 1.08, "learning_rate": 9.150060810178689e-06, "loss": 0.0565, "step": 23195 }, { "epoch": 1.09, "learning_rate": 9.147721957152214e-06, "loss": 0.1326, "step": 23200 }, { "epoch": 1.09, "learning_rate": 9.145383104125737e-06, "loss": 0.0678, "step": 23205 }, { "epoch": 1.09, "learning_rate": 9.14304425109926e-06, "loss": 0.0676, "step": 23210 }, { "epoch": 1.09, "learning_rate": 9.140705398072785e-06, "loss": 0.0515, "step": 23215 }, { "epoch": 1.09, "learning_rate": 9.13836654504631e-06, "loss": 0.0547, "step": 23220 }, { "epoch": 1.09, "learning_rate": 9.136027692019834e-06, "loss": 0.1057, "step": 23225 }, { "epoch": 1.09, "learning_rate": 9.133688838993359e-06, "loss": 0.0583, "step": 23230 }, { "epoch": 1.09, "learning_rate": 9.131349985966884e-06, "loss": 0.0648, "step": 23235 }, { "epoch": 1.09, "learning_rate": 9.129011132940407e-06, "loss": 0.0911, "step": 23240 }, { "epoch": 1.09, "learning_rate": 9.12667227991393e-06, "loss": 0.0319, "step": 23245 }, { "epoch": 1.09, "learning_rate": 9.124333426887455e-06, "loss": 0.0166, "step": 23250 }, { "epoch": 1.09, "learning_rate": 9.12199457386098e-06, "loss": 0.0523, "step": 23255 }, { "epoch": 1.09, "learning_rate": 9.119655720834503e-06, "loss": 0.0462, "step": 23260 }, { "epoch": 1.09, "learning_rate": 9.117316867808028e-06, "loss": 0.0878, "step": 23265 }, { "epoch": 1.09, "learning_rate": 9.114978014781551e-06, "loss": 0.0968, "step": 23270 }, { "epoch": 1.09, "learning_rate": 9.112639161755076e-06, "loss": 0.0614, "step": 23275 }, { "epoch": 1.09, "learning_rate": 9.1103003087286e-06, "loss": 0.0407, "step": 23280 }, { "epoch": 1.09, "learning_rate": 9.107961455702125e-06, "loss": 0.0729, "step": 23285 }, { "epoch": 1.09, "learning_rate": 9.10562260267565e-06, "loss": 0.0718, "step": 23290 }, { "epoch": 1.09, "learning_rate": 9.103283749649173e-06, "loss": 0.0756, "step": 23295 }, { "epoch": 1.09, "learning_rate": 9.100944896622696e-06, "loss": 0.0367, "step": 23300 }, { "epoch": 1.09, "learning_rate": 9.098606043596221e-06, "loss": 0.0704, "step": 23305 }, { "epoch": 1.09, "learning_rate": 9.096267190569746e-06, "loss": 0.0424, "step": 23310 }, { "epoch": 1.09, "learning_rate": 9.09392833754327e-06, "loss": 0.0566, "step": 23315 }, { "epoch": 1.09, "learning_rate": 9.091589484516794e-06, "loss": 0.0749, "step": 23320 }, { "epoch": 1.09, "learning_rate": 9.089250631490319e-06, "loss": 0.0477, "step": 23325 }, { "epoch": 1.09, "learning_rate": 9.086911778463842e-06, "loss": 0.0605, "step": 23330 }, { "epoch": 1.09, "learning_rate": 9.084572925437366e-06, "loss": 0.0548, "step": 23335 }, { "epoch": 1.09, "learning_rate": 9.08223407241089e-06, "loss": 0.0849, "step": 23340 }, { "epoch": 1.09, "learning_rate": 9.079895219384416e-06, "loss": 0.0585, "step": 23345 }, { "epoch": 1.09, "learning_rate": 9.077556366357939e-06, "loss": 0.0883, "step": 23350 }, { "epoch": 1.09, "learning_rate": 9.075217513331464e-06, "loss": 0.0547, "step": 23355 }, { "epoch": 1.09, "learning_rate": 9.072878660304987e-06, "loss": 0.059, "step": 23360 }, { "epoch": 1.09, "learning_rate": 9.070539807278512e-06, "loss": 0.077, "step": 23365 }, { "epoch": 1.09, "learning_rate": 9.068200954252035e-06, "loss": 0.0807, "step": 23370 }, { "epoch": 1.09, "learning_rate": 9.06586210122556e-06, "loss": 0.0831, "step": 23375 }, { "epoch": 1.09, "learning_rate": 9.063523248199083e-06, "loss": 0.047, "step": 23380 }, { "epoch": 1.09, "learning_rate": 9.061184395172608e-06, "loss": 0.1323, "step": 23385 }, { "epoch": 1.09, "learning_rate": 9.058845542146132e-06, "loss": 0.08, "step": 23390 }, { "epoch": 1.09, "learning_rate": 9.056506689119657e-06, "loss": 0.0306, "step": 23395 }, { "epoch": 1.09, "learning_rate": 9.05416783609318e-06, "loss": 0.0471, "step": 23400 }, { "epoch": 1.09, "learning_rate": 9.051828983066705e-06, "loss": 0.0672, "step": 23405 }, { "epoch": 1.1, "learning_rate": 9.04949013004023e-06, "loss": 0.0403, "step": 23410 }, { "epoch": 1.1, "learning_rate": 9.047151277013753e-06, "loss": 0.0786, "step": 23415 }, { "epoch": 1.1, "learning_rate": 9.044812423987276e-06, "loss": 0.0749, "step": 23420 }, { "epoch": 1.1, "learning_rate": 9.042473570960801e-06, "loss": 0.0347, "step": 23425 }, { "epoch": 1.1, "learning_rate": 9.040134717934326e-06, "loss": 0.0545, "step": 23430 }, { "epoch": 1.1, "learning_rate": 9.03779586490785e-06, "loss": 0.049, "step": 23435 }, { "epoch": 1.1, "learning_rate": 9.035457011881374e-06, "loss": 0.0341, "step": 23440 }, { "epoch": 1.1, "learning_rate": 9.0331181588549e-06, "loss": 0.091, "step": 23445 }, { "epoch": 1.1, "learning_rate": 9.030779305828423e-06, "loss": 0.0242, "step": 23450 }, { "epoch": 1.1, "learning_rate": 9.028440452801946e-06, "loss": 0.0456, "step": 23455 }, { "epoch": 1.1, "learning_rate": 9.02610159977547e-06, "loss": 0.0699, "step": 23460 }, { "epoch": 1.1, "learning_rate": 9.023762746748996e-06, "loss": 0.1035, "step": 23465 }, { "epoch": 1.1, "learning_rate": 9.021423893722519e-06, "loss": 0.0879, "step": 23470 }, { "epoch": 1.1, "learning_rate": 9.019085040696042e-06, "loss": 0.0242, "step": 23475 }, { "epoch": 1.1, "learning_rate": 9.016746187669567e-06, "loss": 0.0543, "step": 23480 }, { "epoch": 1.1, "learning_rate": 9.014407334643092e-06, "loss": 0.1253, "step": 23485 }, { "epoch": 1.1, "learning_rate": 9.012068481616615e-06, "loss": 0.0529, "step": 23490 }, { "epoch": 1.1, "learning_rate": 9.00972962859014e-06, "loss": 0.0998, "step": 23495 }, { "epoch": 1.1, "learning_rate": 9.007390775563665e-06, "loss": 0.0544, "step": 23500 }, { "epoch": 1.1, "learning_rate": 9.005051922537189e-06, "loss": 0.0485, "step": 23505 }, { "epoch": 1.1, "learning_rate": 9.002713069510712e-06, "loss": 0.0819, "step": 23510 }, { "epoch": 1.1, "learning_rate": 9.000374216484237e-06, "loss": 0.0548, "step": 23515 }, { "epoch": 1.1, "learning_rate": 8.998035363457762e-06, "loss": 0.0393, "step": 23520 }, { "epoch": 1.1, "learning_rate": 8.995696510431285e-06, "loss": 0.1064, "step": 23525 }, { "epoch": 1.1, "learning_rate": 8.99335765740481e-06, "loss": 0.073, "step": 23530 }, { "epoch": 1.1, "learning_rate": 8.991018804378333e-06, "loss": 0.0695, "step": 23535 }, { "epoch": 1.1, "learning_rate": 8.988679951351858e-06, "loss": 0.1238, "step": 23540 }, { "epoch": 1.1, "learning_rate": 8.986341098325381e-06, "loss": 0.0479, "step": 23545 }, { "epoch": 1.1, "learning_rate": 8.984002245298906e-06, "loss": 0.0457, "step": 23550 }, { "epoch": 1.1, "learning_rate": 8.981663392272431e-06, "loss": 0.0586, "step": 23555 }, { "epoch": 1.1, "learning_rate": 8.979324539245955e-06, "loss": 0.0639, "step": 23560 }, { "epoch": 1.1, "learning_rate": 8.976985686219478e-06, "loss": 0.0358, "step": 23565 }, { "epoch": 1.1, "learning_rate": 8.974646833193003e-06, "loss": 0.0849, "step": 23570 }, { "epoch": 1.1, "learning_rate": 8.972307980166528e-06, "loss": 0.0417, "step": 23575 }, { "epoch": 1.1, "learning_rate": 8.969969127140051e-06, "loss": 0.0712, "step": 23580 }, { "epoch": 1.1, "learning_rate": 8.967630274113576e-06, "loss": 0.057, "step": 23585 }, { "epoch": 1.1, "learning_rate": 8.965291421087101e-06, "loss": 0.0773, "step": 23590 }, { "epoch": 1.1, "learning_rate": 8.962952568060624e-06, "loss": 0.0457, "step": 23595 }, { "epoch": 1.1, "learning_rate": 8.960613715034147e-06, "loss": 0.0305, "step": 23600 }, { "epoch": 1.1, "learning_rate": 8.958274862007672e-06, "loss": 0.0538, "step": 23605 }, { "epoch": 1.1, "learning_rate": 8.955936008981196e-06, "loss": 0.0645, "step": 23610 }, { "epoch": 1.1, "learning_rate": 8.95359715595472e-06, "loss": 0.0543, "step": 23615 }, { "epoch": 1.1, "learning_rate": 8.951258302928246e-06, "loss": 0.0647, "step": 23620 }, { "epoch": 1.11, "learning_rate": 8.948919449901769e-06, "loss": 0.1186, "step": 23625 }, { "epoch": 1.11, "learning_rate": 8.946580596875292e-06, "loss": 0.0712, "step": 23630 }, { "epoch": 1.11, "learning_rate": 8.944241743848817e-06, "loss": 0.0572, "step": 23635 }, { "epoch": 1.11, "learning_rate": 8.941902890822342e-06, "loss": 0.0589, "step": 23640 }, { "epoch": 1.11, "learning_rate": 8.939564037795865e-06, "loss": 0.0309, "step": 23645 }, { "epoch": 1.11, "learning_rate": 8.93722518476939e-06, "loss": 0.0818, "step": 23650 }, { "epoch": 1.11, "learning_rate": 8.934886331742913e-06, "loss": 0.0552, "step": 23655 }, { "epoch": 1.11, "learning_rate": 8.932547478716438e-06, "loss": 0.0966, "step": 23660 }, { "epoch": 1.11, "learning_rate": 8.930208625689962e-06, "loss": 0.0262, "step": 23665 }, { "epoch": 1.11, "learning_rate": 8.927869772663487e-06, "loss": 0.0133, "step": 23670 }, { "epoch": 1.11, "learning_rate": 8.925530919637012e-06, "loss": 0.0726, "step": 23675 }, { "epoch": 1.11, "learning_rate": 8.923192066610535e-06, "loss": 0.0372, "step": 23680 }, { "epoch": 1.11, "learning_rate": 8.920853213584058e-06, "loss": 0.0845, "step": 23685 }, { "epoch": 1.11, "learning_rate": 8.918514360557583e-06, "loss": 0.0768, "step": 23690 }, { "epoch": 1.11, "learning_rate": 8.916175507531108e-06, "loss": 0.1496, "step": 23695 }, { "epoch": 1.11, "learning_rate": 8.913836654504631e-06, "loss": 0.0308, "step": 23700 }, { "epoch": 1.11, "learning_rate": 8.911497801478156e-06, "loss": 0.0648, "step": 23705 }, { "epoch": 1.11, "learning_rate": 8.909158948451681e-06, "loss": 0.0615, "step": 23710 }, { "epoch": 1.11, "learning_rate": 8.906820095425204e-06, "loss": 0.0684, "step": 23715 }, { "epoch": 1.11, "learning_rate": 8.904481242398728e-06, "loss": 0.0561, "step": 23720 }, { "epoch": 1.11, "learning_rate": 8.902142389372253e-06, "loss": 0.1224, "step": 23725 }, { "epoch": 1.11, "learning_rate": 8.899803536345778e-06, "loss": 0.0887, "step": 23730 }, { "epoch": 1.11, "learning_rate": 8.8974646833193e-06, "loss": 0.0618, "step": 23735 }, { "epoch": 1.11, "learning_rate": 8.895125830292826e-06, "loss": 0.0219, "step": 23740 }, { "epoch": 1.11, "learning_rate": 8.892786977266349e-06, "loss": 0.0372, "step": 23745 }, { "epoch": 1.11, "learning_rate": 8.890448124239874e-06, "loss": 0.0938, "step": 23750 }, { "epoch": 1.11, "learning_rate": 8.888109271213397e-06, "loss": 0.0703, "step": 23755 }, { "epoch": 1.11, "learning_rate": 8.885770418186922e-06, "loss": 0.0414, "step": 23760 }, { "epoch": 1.11, "learning_rate": 8.883431565160447e-06, "loss": 0.0654, "step": 23765 }, { "epoch": 1.11, "learning_rate": 8.88109271213397e-06, "loss": 0.0392, "step": 23770 }, { "epoch": 1.11, "learning_rate": 8.878753859107494e-06, "loss": 0.0308, "step": 23775 }, { "epoch": 1.11, "learning_rate": 8.876415006081019e-06, "loss": 0.033, "step": 23780 }, { "epoch": 1.11, "learning_rate": 8.874076153054544e-06, "loss": 0.0148, "step": 23785 }, { "epoch": 1.11, "learning_rate": 8.871737300028067e-06, "loss": 0.0538, "step": 23790 }, { "epoch": 1.11, "learning_rate": 8.869398447001592e-06, "loss": 0.0559, "step": 23795 }, { "epoch": 1.11, "learning_rate": 8.867059593975117e-06, "loss": 0.0492, "step": 23800 }, { "epoch": 1.11, "learning_rate": 8.86472074094864e-06, "loss": 0.0644, "step": 23805 }, { "epoch": 1.11, "learning_rate": 8.862381887922163e-06, "loss": 0.0797, "step": 23810 }, { "epoch": 1.11, "learning_rate": 8.860043034895688e-06, "loss": 0.0794, "step": 23815 }, { "epoch": 1.11, "learning_rate": 8.857704181869211e-06, "loss": 0.0877, "step": 23820 }, { "epoch": 1.11, "learning_rate": 8.855365328842736e-06, "loss": 0.0791, "step": 23825 }, { "epoch": 1.11, "learning_rate": 8.853026475816261e-06, "loss": 0.1079, "step": 23830 }, { "epoch": 1.11, "learning_rate": 8.850687622789785e-06, "loss": 0.0837, "step": 23835 }, { "epoch": 1.12, "learning_rate": 8.848348769763308e-06, "loss": 0.102, "step": 23840 }, { "epoch": 1.12, "learning_rate": 8.846009916736833e-06, "loss": 0.1248, "step": 23845 }, { "epoch": 1.12, "learning_rate": 8.843671063710358e-06, "loss": 0.0514, "step": 23850 }, { "epoch": 1.12, "learning_rate": 8.841332210683881e-06, "loss": 0.0746, "step": 23855 }, { "epoch": 1.12, "learning_rate": 8.838993357657406e-06, "loss": 0.0499, "step": 23860 }, { "epoch": 1.12, "learning_rate": 8.83665450463093e-06, "loss": 0.0715, "step": 23865 }, { "epoch": 1.12, "learning_rate": 8.834315651604454e-06, "loss": 0.0564, "step": 23870 }, { "epoch": 1.12, "learning_rate": 8.831976798577977e-06, "loss": 0.0881, "step": 23875 }, { "epoch": 1.12, "learning_rate": 8.829637945551502e-06, "loss": 0.0292, "step": 23880 }, { "epoch": 1.12, "learning_rate": 8.827299092525027e-06, "loss": 0.0365, "step": 23885 }, { "epoch": 1.12, "learning_rate": 8.82496023949855e-06, "loss": 0.0229, "step": 23890 }, { "epoch": 1.12, "learning_rate": 8.822621386472074e-06, "loss": 0.0492, "step": 23895 }, { "epoch": 1.12, "learning_rate": 8.820282533445599e-06, "loss": 0.0543, "step": 23900 }, { "epoch": 1.12, "learning_rate": 8.817943680419124e-06, "loss": 0.0472, "step": 23905 }, { "epoch": 1.12, "learning_rate": 8.815604827392647e-06, "loss": 0.06, "step": 23910 }, { "epoch": 1.12, "learning_rate": 8.813265974366172e-06, "loss": 0.0512, "step": 23915 }, { "epoch": 1.12, "learning_rate": 8.810927121339695e-06, "loss": 0.0349, "step": 23920 }, { "epoch": 1.12, "learning_rate": 8.80858826831322e-06, "loss": 0.0565, "step": 23925 }, { "epoch": 1.12, "learning_rate": 8.806249415286743e-06, "loss": 0.0466, "step": 23930 }, { "epoch": 1.12, "learning_rate": 8.803910562260268e-06, "loss": 0.0338, "step": 23935 }, { "epoch": 1.12, "learning_rate": 8.801571709233793e-06, "loss": 0.0463, "step": 23940 }, { "epoch": 1.12, "learning_rate": 8.799232856207317e-06, "loss": 0.0532, "step": 23945 }, { "epoch": 1.12, "learning_rate": 8.79689400318084e-06, "loss": 0.0617, "step": 23950 }, { "epoch": 1.12, "learning_rate": 8.794555150154365e-06, "loss": 0.1646, "step": 23955 }, { "epoch": 1.12, "learning_rate": 8.79221629712789e-06, "loss": 0.0743, "step": 23960 }, { "epoch": 1.12, "learning_rate": 8.789877444101413e-06, "loss": 0.0521, "step": 23965 }, { "epoch": 1.12, "learning_rate": 8.787538591074938e-06, "loss": 0.0474, "step": 23970 }, { "epoch": 1.12, "learning_rate": 8.785199738048463e-06, "loss": 0.0316, "step": 23975 }, { "epoch": 1.12, "learning_rate": 8.782860885021986e-06, "loss": 0.0354, "step": 23980 }, { "epoch": 1.12, "learning_rate": 8.78052203199551e-06, "loss": 0.0814, "step": 23985 }, { "epoch": 1.12, "learning_rate": 8.778183178969034e-06, "loss": 0.0345, "step": 23990 }, { "epoch": 1.12, "learning_rate": 8.77584432594256e-06, "loss": 0.0401, "step": 23995 }, { "epoch": 1.12, "learning_rate": 8.773505472916083e-06, "loss": 0.0551, "step": 24000 }, { "epoch": 1.12, "learning_rate": 8.771166619889607e-06, "loss": 0.0428, "step": 24005 }, { "epoch": 1.12, "learning_rate": 8.76882776686313e-06, "loss": 0.069, "step": 24010 }, { "epoch": 1.12, "learning_rate": 8.766488913836656e-06, "loss": 0.0705, "step": 24015 }, { "epoch": 1.12, "learning_rate": 8.764150060810179e-06, "loss": 0.0232, "step": 24020 }, { "epoch": 1.12, "learning_rate": 8.761811207783704e-06, "loss": 0.0491, "step": 24025 }, { "epoch": 1.12, "learning_rate": 8.759472354757229e-06, "loss": 0.0643, "step": 24030 }, { "epoch": 1.12, "learning_rate": 8.757133501730752e-06, "loss": 0.0872, "step": 24035 }, { "epoch": 1.12, "learning_rate": 8.754794648704275e-06, "loss": 0.0177, "step": 24040 }, { "epoch": 1.12, "learning_rate": 8.7524557956778e-06, "loss": 0.0398, "step": 24045 }, { "epoch": 1.12, "learning_rate": 8.750116942651324e-06, "loss": 0.1002, "step": 24050 }, { "epoch": 1.13, "learning_rate": 8.747778089624849e-06, "loss": 0.085, "step": 24055 }, { "epoch": 1.13, "learning_rate": 8.745439236598373e-06, "loss": 0.0421, "step": 24060 }, { "epoch": 1.13, "learning_rate": 8.743100383571897e-06, "loss": 0.066, "step": 24065 }, { "epoch": 1.13, "learning_rate": 8.74076153054542e-06, "loss": 0.0259, "step": 24070 }, { "epoch": 1.13, "learning_rate": 8.738422677518945e-06, "loss": 0.1182, "step": 24075 }, { "epoch": 1.13, "learning_rate": 8.73608382449247e-06, "loss": 0.0475, "step": 24080 }, { "epoch": 1.13, "learning_rate": 8.733744971465993e-06, "loss": 0.0598, "step": 24085 }, { "epoch": 1.13, "learning_rate": 8.731406118439518e-06, "loss": 0.0518, "step": 24090 }, { "epoch": 1.13, "learning_rate": 8.729067265413043e-06, "loss": 0.0135, "step": 24095 }, { "epoch": 1.13, "learning_rate": 8.726728412386566e-06, "loss": 0.0783, "step": 24100 }, { "epoch": 1.13, "learning_rate": 8.72438955936009e-06, "loss": 0.0553, "step": 24105 }, { "epoch": 1.13, "learning_rate": 8.722050706333615e-06, "loss": 0.0777, "step": 24110 }, { "epoch": 1.13, "learning_rate": 8.71971185330714e-06, "loss": 0.0373, "step": 24115 }, { "epoch": 1.13, "learning_rate": 8.717373000280663e-06, "loss": 0.0291, "step": 24120 }, { "epoch": 1.13, "learning_rate": 8.715034147254188e-06, "loss": 0.0546, "step": 24125 }, { "epoch": 1.13, "learning_rate": 8.712695294227711e-06, "loss": 0.0273, "step": 24130 }, { "epoch": 1.13, "learning_rate": 8.710356441201236e-06, "loss": 0.071, "step": 24135 }, { "epoch": 1.13, "learning_rate": 8.70801758817476e-06, "loss": 0.017, "step": 24140 }, { "epoch": 1.13, "learning_rate": 8.705678735148284e-06, "loss": 0.0631, "step": 24145 }, { "epoch": 1.13, "learning_rate": 8.703339882121809e-06, "loss": 0.0702, "step": 24150 }, { "epoch": 1.13, "learning_rate": 8.701001029095332e-06, "loss": 0.148, "step": 24155 }, { "epoch": 1.13, "learning_rate": 8.698662176068856e-06, "loss": 0.0604, "step": 24160 }, { "epoch": 1.13, "learning_rate": 8.69632332304238e-06, "loss": 0.0437, "step": 24165 }, { "epoch": 1.13, "learning_rate": 8.693984470015905e-06, "loss": 0.038, "step": 24170 }, { "epoch": 1.13, "learning_rate": 8.691645616989429e-06, "loss": 0.0527, "step": 24175 }, { "epoch": 1.13, "learning_rate": 8.689306763962954e-06, "loss": 0.0522, "step": 24180 }, { "epoch": 1.13, "learning_rate": 8.686967910936479e-06, "loss": 0.0331, "step": 24185 }, { "epoch": 1.13, "learning_rate": 8.684629057910002e-06, "loss": 0.1011, "step": 24190 }, { "epoch": 1.13, "learning_rate": 8.682290204883525e-06, "loss": 0.0714, "step": 24195 }, { "epoch": 1.13, "learning_rate": 8.67995135185705e-06, "loss": 0.0965, "step": 24200 }, { "epoch": 1.13, "learning_rate": 8.677612498830575e-06, "loss": 0.0637, "step": 24205 }, { "epoch": 1.13, "learning_rate": 8.675273645804098e-06, "loss": 0.0505, "step": 24210 }, { "epoch": 1.13, "learning_rate": 8.672934792777623e-06, "loss": 0.0821, "step": 24215 }, { "epoch": 1.13, "learning_rate": 8.670595939751147e-06, "loss": 0.051, "step": 24220 }, { "epoch": 1.13, "learning_rate": 8.668257086724671e-06, "loss": 0.081, "step": 24225 }, { "epoch": 1.13, "learning_rate": 8.665918233698195e-06, "loss": 0.0308, "step": 24230 }, { "epoch": 1.13, "learning_rate": 8.66357938067172e-06, "loss": 0.0359, "step": 24235 }, { "epoch": 1.13, "learning_rate": 8.661240527645245e-06, "loss": 0.0688, "step": 24240 }, { "epoch": 1.13, "learning_rate": 8.658901674618768e-06, "loss": 0.0446, "step": 24245 }, { "epoch": 1.13, "learning_rate": 8.656562821592291e-06, "loss": 0.0588, "step": 24250 }, { "epoch": 1.13, "learning_rate": 8.654223968565816e-06, "loss": 0.0628, "step": 24255 }, { "epoch": 1.13, "learning_rate": 8.65188511553934e-06, "loss": 0.0728, "step": 24260 }, { "epoch": 1.14, "learning_rate": 8.649546262512864e-06, "loss": 0.0563, "step": 24265 }, { "epoch": 1.14, "learning_rate": 8.64720740948639e-06, "loss": 0.0694, "step": 24270 }, { "epoch": 1.14, "learning_rate": 8.644868556459913e-06, "loss": 0.0385, "step": 24275 }, { "epoch": 1.14, "learning_rate": 8.642529703433436e-06, "loss": 0.0509, "step": 24280 }, { "epoch": 1.14, "learning_rate": 8.64019085040696e-06, "loss": 0.0295, "step": 24285 }, { "epoch": 1.14, "learning_rate": 8.637851997380486e-06, "loss": 0.0875, "step": 24290 }, { "epoch": 1.14, "learning_rate": 8.635513144354009e-06, "loss": 0.0192, "step": 24295 }, { "epoch": 1.14, "learning_rate": 8.633174291327534e-06, "loss": 0.0739, "step": 24300 }, { "epoch": 1.14, "learning_rate": 8.630835438301059e-06, "loss": 0.1037, "step": 24305 }, { "epoch": 1.14, "learning_rate": 8.628496585274582e-06, "loss": 0.0588, "step": 24310 }, { "epoch": 1.14, "learning_rate": 8.626157732248105e-06, "loss": 0.1076, "step": 24315 }, { "epoch": 1.14, "learning_rate": 8.62381887922163e-06, "loss": 0.0298, "step": 24320 }, { "epoch": 1.14, "learning_rate": 8.621480026195155e-06, "loss": 0.0338, "step": 24325 }, { "epoch": 1.14, "learning_rate": 8.619141173168679e-06, "loss": 0.0542, "step": 24330 }, { "epoch": 1.14, "learning_rate": 8.616802320142202e-06, "loss": 0.0192, "step": 24335 }, { "epoch": 1.14, "learning_rate": 8.614463467115727e-06, "loss": 0.0257, "step": 24340 }, { "epoch": 1.14, "learning_rate": 8.612124614089252e-06, "loss": 0.0262, "step": 24345 }, { "epoch": 1.14, "learning_rate": 8.609785761062775e-06, "loss": 0.0799, "step": 24350 }, { "epoch": 1.14, "learning_rate": 8.6074469080363e-06, "loss": 0.039, "step": 24355 }, { "epoch": 1.14, "learning_rate": 8.605108055009825e-06, "loss": 0.0607, "step": 24360 }, { "epoch": 1.14, "learning_rate": 8.602769201983348e-06, "loss": 0.0646, "step": 24365 }, { "epoch": 1.14, "learning_rate": 8.600430348956871e-06, "loss": 0.042, "step": 24370 }, { "epoch": 1.14, "learning_rate": 8.598091495930396e-06, "loss": 0.0671, "step": 24375 }, { "epoch": 1.14, "learning_rate": 8.595752642903921e-06, "loss": 0.1056, "step": 24380 }, { "epoch": 1.14, "learning_rate": 8.593413789877445e-06, "loss": 0.0492, "step": 24385 }, { "epoch": 1.14, "learning_rate": 8.59107493685097e-06, "loss": 0.0643, "step": 24390 }, { "epoch": 1.14, "learning_rate": 8.588736083824493e-06, "loss": 0.0546, "step": 24395 }, { "epoch": 1.14, "learning_rate": 8.586397230798018e-06, "loss": 0.0974, "step": 24400 }, { "epoch": 1.14, "learning_rate": 8.584058377771541e-06, "loss": 0.0791, "step": 24405 }, { "epoch": 1.14, "learning_rate": 8.581719524745066e-06, "loss": 0.061, "step": 24410 }, { "epoch": 1.14, "learning_rate": 8.57938067171859e-06, "loss": 0.0443, "step": 24415 }, { "epoch": 1.14, "learning_rate": 8.577041818692114e-06, "loss": 0.0691, "step": 24420 }, { "epoch": 1.14, "learning_rate": 8.574702965665637e-06, "loss": 0.0695, "step": 24425 }, { "epoch": 1.14, "learning_rate": 8.572364112639162e-06, "loss": 0.0666, "step": 24430 }, { "epoch": 1.14, "learning_rate": 8.570025259612687e-06, "loss": 0.0364, "step": 24435 }, { "epoch": 1.14, "learning_rate": 8.56768640658621e-06, "loss": 0.057, "step": 24440 }, { "epoch": 1.14, "learning_rate": 8.565347553559735e-06, "loss": 0.0342, "step": 24445 }, { "epoch": 1.14, "learning_rate": 8.56300870053326e-06, "loss": 0.0771, "step": 24450 }, { "epoch": 1.14, "learning_rate": 8.560669847506784e-06, "loss": 0.0466, "step": 24455 }, { "epoch": 1.14, "learning_rate": 8.558330994480307e-06, "loss": 0.0503, "step": 24460 }, { "epoch": 1.14, "learning_rate": 8.555992141453832e-06, "loss": 0.055, "step": 24465 }, { "epoch": 1.14, "learning_rate": 8.553653288427357e-06, "loss": 0.0396, "step": 24470 }, { "epoch": 1.14, "learning_rate": 8.55131443540088e-06, "loss": 0.0357, "step": 24475 }, { "epoch": 1.15, "learning_rate": 8.548975582374405e-06, "loss": 0.0655, "step": 24480 }, { "epoch": 1.15, "learning_rate": 8.546636729347928e-06, "loss": 0.0292, "step": 24485 }, { "epoch": 1.15, "learning_rate": 8.544297876321452e-06, "loss": 0.0588, "step": 24490 }, { "epoch": 1.15, "learning_rate": 8.541959023294976e-06, "loss": 0.0307, "step": 24495 }, { "epoch": 1.15, "learning_rate": 8.539620170268501e-06, "loss": 0.0984, "step": 24500 }, { "epoch": 1.15, "learning_rate": 8.537281317242025e-06, "loss": 0.0642, "step": 24505 }, { "epoch": 1.15, "learning_rate": 8.53494246421555e-06, "loss": 0.0543, "step": 24510 }, { "epoch": 1.15, "learning_rate": 8.532603611189073e-06, "loss": 0.0471, "step": 24515 }, { "epoch": 1.15, "learning_rate": 8.530264758162598e-06, "loss": 0.0657, "step": 24520 }, { "epoch": 1.15, "learning_rate": 8.527925905136121e-06, "loss": 0.0724, "step": 24525 }, { "epoch": 1.15, "learning_rate": 8.525587052109646e-06, "loss": 0.0835, "step": 24530 }, { "epoch": 1.15, "learning_rate": 8.523248199083171e-06, "loss": 0.0706, "step": 24535 }, { "epoch": 1.15, "learning_rate": 8.520909346056694e-06, "loss": 0.071, "step": 24540 }, { "epoch": 1.15, "learning_rate": 8.518570493030218e-06, "loss": 0.0871, "step": 24545 }, { "epoch": 1.15, "learning_rate": 8.516231640003742e-06, "loss": 0.0207, "step": 24550 }, { "epoch": 1.15, "learning_rate": 8.513892786977267e-06, "loss": 0.0536, "step": 24555 }, { "epoch": 1.15, "learning_rate": 8.51155393395079e-06, "loss": 0.0613, "step": 24560 }, { "epoch": 1.15, "learning_rate": 8.509215080924316e-06, "loss": 0.0531, "step": 24565 }, { "epoch": 1.15, "learning_rate": 8.50687622789784e-06, "loss": 0.027, "step": 24570 }, { "epoch": 1.15, "learning_rate": 8.504537374871364e-06, "loss": 0.0535, "step": 24575 }, { "epoch": 1.15, "learning_rate": 8.502198521844887e-06, "loss": 0.0428, "step": 24580 }, { "epoch": 1.15, "learning_rate": 8.499859668818412e-06, "loss": 0.0631, "step": 24585 }, { "epoch": 1.15, "learning_rate": 8.497520815791937e-06, "loss": 0.1581, "step": 24590 }, { "epoch": 1.15, "learning_rate": 8.49518196276546e-06, "loss": 0.0481, "step": 24595 }, { "epoch": 1.15, "learning_rate": 8.492843109738985e-06, "loss": 0.081, "step": 24600 }, { "epoch": 1.15, "learning_rate": 8.490504256712508e-06, "loss": 0.1209, "step": 24605 }, { "epoch": 1.15, "learning_rate": 8.488165403686033e-06, "loss": 0.0533, "step": 24610 }, { "epoch": 1.15, "learning_rate": 8.485826550659557e-06, "loss": 0.0452, "step": 24615 }, { "epoch": 1.15, "learning_rate": 8.483487697633082e-06, "loss": 0.0678, "step": 24620 }, { "epoch": 1.15, "learning_rate": 8.481148844606607e-06, "loss": 0.0426, "step": 24625 }, { "epoch": 1.15, "learning_rate": 8.47880999158013e-06, "loss": 0.0643, "step": 24630 }, { "epoch": 1.15, "learning_rate": 8.476471138553653e-06, "loss": 0.0407, "step": 24635 }, { "epoch": 1.15, "learning_rate": 8.474132285527178e-06, "loss": 0.0475, "step": 24640 }, { "epoch": 1.15, "learning_rate": 8.471793432500703e-06, "loss": 0.0425, "step": 24645 }, { "epoch": 1.15, "learning_rate": 8.469454579474226e-06, "loss": 0.1148, "step": 24650 }, { "epoch": 1.15, "learning_rate": 8.467115726447751e-06, "loss": 0.0829, "step": 24655 }, { "epoch": 1.15, "learning_rate": 8.464776873421276e-06, "loss": 0.056, "step": 24660 }, { "epoch": 1.15, "learning_rate": 8.4624380203948e-06, "loss": 0.0832, "step": 24665 }, { "epoch": 1.15, "learning_rate": 8.460099167368323e-06, "loss": 0.0732, "step": 24670 }, { "epoch": 1.15, "learning_rate": 8.457760314341848e-06, "loss": 0.1042, "step": 24675 }, { "epoch": 1.15, "learning_rate": 8.455421461315373e-06, "loss": 0.1119, "step": 24680 }, { "epoch": 1.15, "learning_rate": 8.453082608288896e-06, "loss": 0.0669, "step": 24685 }, { "epoch": 1.15, "learning_rate": 8.45074375526242e-06, "loss": 0.0396, "step": 24690 }, { "epoch": 1.16, "learning_rate": 8.448404902235944e-06, "loss": 0.0839, "step": 24695 }, { "epoch": 1.16, "learning_rate": 8.446066049209467e-06, "loss": 0.0701, "step": 24700 }, { "epoch": 1.16, "learning_rate": 8.443727196182992e-06, "loss": 0.0617, "step": 24705 }, { "epoch": 1.16, "learning_rate": 8.441388343156517e-06, "loss": 0.0902, "step": 24710 }, { "epoch": 1.16, "learning_rate": 8.43904949013004e-06, "loss": 0.0826, "step": 24715 }, { "epoch": 1.16, "learning_rate": 8.436710637103565e-06, "loss": 0.0529, "step": 24720 }, { "epoch": 1.16, "learning_rate": 8.434371784077089e-06, "loss": 0.0377, "step": 24725 }, { "epoch": 1.16, "learning_rate": 8.432032931050614e-06, "loss": 0.031, "step": 24730 }, { "epoch": 1.16, "learning_rate": 8.429694078024137e-06, "loss": 0.0336, "step": 24735 }, { "epoch": 1.16, "learning_rate": 8.427355224997662e-06, "loss": 0.0728, "step": 24740 }, { "epoch": 1.16, "learning_rate": 8.425016371971187e-06, "loss": 0.0784, "step": 24745 }, { "epoch": 1.16, "learning_rate": 8.42267751894471e-06, "loss": 0.0703, "step": 24750 }, { "epoch": 1.16, "learning_rate": 8.420338665918233e-06, "loss": 0.1074, "step": 24755 }, { "epoch": 1.16, "learning_rate": 8.417999812891758e-06, "loss": 0.0904, "step": 24760 }, { "epoch": 1.16, "learning_rate": 8.415660959865283e-06, "loss": 0.0587, "step": 24765 }, { "epoch": 1.16, "learning_rate": 8.413322106838806e-06, "loss": 0.0811, "step": 24770 }, { "epoch": 1.16, "learning_rate": 8.410983253812331e-06, "loss": 0.065, "step": 24775 }, { "epoch": 1.16, "learning_rate": 8.408644400785855e-06, "loss": 0.0274, "step": 24780 }, { "epoch": 1.16, "learning_rate": 8.40630554775938e-06, "loss": 0.0686, "step": 24785 }, { "epoch": 1.16, "learning_rate": 8.403966694732903e-06, "loss": 0.0867, "step": 24790 }, { "epoch": 1.16, "learning_rate": 8.401627841706428e-06, "loss": 0.0476, "step": 24795 }, { "epoch": 1.16, "learning_rate": 8.399288988679953e-06, "loss": 0.0711, "step": 24800 }, { "epoch": 1.16, "learning_rate": 8.396950135653476e-06, "loss": 0.0848, "step": 24805 }, { "epoch": 1.16, "learning_rate": 8.394611282627e-06, "loss": 0.0694, "step": 24810 }, { "epoch": 1.16, "learning_rate": 8.392272429600524e-06, "loss": 0.0432, "step": 24815 }, { "epoch": 1.16, "learning_rate": 8.38993357657405e-06, "loss": 0.05, "step": 24820 }, { "epoch": 1.16, "learning_rate": 8.387594723547572e-06, "loss": 0.0583, "step": 24825 }, { "epoch": 1.16, "learning_rate": 8.385255870521097e-06, "loss": 0.0378, "step": 24830 }, { "epoch": 1.16, "learning_rate": 8.382917017494622e-06, "loss": 0.0483, "step": 24835 }, { "epoch": 1.16, "learning_rate": 8.380578164468146e-06, "loss": 0.0738, "step": 24840 }, { "epoch": 1.16, "learning_rate": 8.378239311441669e-06, "loss": 0.0574, "step": 24845 }, { "epoch": 1.16, "learning_rate": 8.375900458415194e-06, "loss": 0.0985, "step": 24850 }, { "epoch": 1.16, "learning_rate": 8.373561605388719e-06, "loss": 0.0177, "step": 24855 }, { "epoch": 1.16, "learning_rate": 8.371222752362242e-06, "loss": 0.009, "step": 24860 }, { "epoch": 1.16, "learning_rate": 8.368883899335767e-06, "loss": 0.0419, "step": 24865 }, { "epoch": 1.16, "learning_rate": 8.36654504630929e-06, "loss": 0.0895, "step": 24870 }, { "epoch": 1.16, "learning_rate": 8.364206193282815e-06, "loss": 0.0422, "step": 24875 }, { "epoch": 1.16, "learning_rate": 8.361867340256338e-06, "loss": 0.0729, "step": 24880 }, { "epoch": 1.16, "learning_rate": 8.359528487229863e-06, "loss": 0.027, "step": 24885 }, { "epoch": 1.16, "learning_rate": 8.357189634203388e-06, "loss": 0.0488, "step": 24890 }, { "epoch": 1.16, "learning_rate": 8.354850781176912e-06, "loss": 0.042, "step": 24895 }, { "epoch": 1.16, "learning_rate": 8.352511928150435e-06, "loss": 0.0504, "step": 24900 }, { "epoch": 1.16, "learning_rate": 8.35017307512396e-06, "loss": 0.0413, "step": 24905 }, { "epoch": 1.17, "learning_rate": 8.347834222097485e-06, "loss": 0.0798, "step": 24910 }, { "epoch": 1.17, "learning_rate": 8.345495369071008e-06, "loss": 0.0916, "step": 24915 }, { "epoch": 1.17, "learning_rate": 8.343156516044533e-06, "loss": 0.0806, "step": 24920 }, { "epoch": 1.17, "learning_rate": 8.340817663018056e-06, "loss": 0.0763, "step": 24925 }, { "epoch": 1.17, "learning_rate": 8.33847880999158e-06, "loss": 0.0632, "step": 24930 }, { "epoch": 1.17, "learning_rate": 8.336139956965104e-06, "loss": 0.0899, "step": 24935 }, { "epoch": 1.17, "learning_rate": 8.33380110393863e-06, "loss": 0.1116, "step": 24940 }, { "epoch": 1.17, "learning_rate": 8.331462250912153e-06, "loss": 0.0401, "step": 24945 }, { "epoch": 1.17, "learning_rate": 8.329123397885678e-06, "loss": 0.1033, "step": 24950 }, { "epoch": 1.17, "learning_rate": 8.326784544859203e-06, "loss": 0.0741, "step": 24955 }, { "epoch": 1.17, "learning_rate": 8.324445691832726e-06, "loss": 0.0469, "step": 24960 }, { "epoch": 1.17, "learning_rate": 8.322106838806249e-06, "loss": 0.0505, "step": 24965 }, { "epoch": 1.17, "learning_rate": 8.319767985779774e-06, "loss": 0.0575, "step": 24970 }, { "epoch": 1.17, "learning_rate": 8.317429132753299e-06, "loss": 0.0924, "step": 24975 }, { "epoch": 1.17, "learning_rate": 8.315090279726822e-06, "loss": 0.039, "step": 24980 }, { "epoch": 1.17, "learning_rate": 8.312751426700347e-06, "loss": 0.0275, "step": 24985 }, { "epoch": 1.17, "learning_rate": 8.31041257367387e-06, "loss": 0.0402, "step": 24990 }, { "epoch": 1.17, "learning_rate": 8.308073720647395e-06, "loss": 0.0435, "step": 24995 }, { "epoch": 1.17, "learning_rate": 8.305734867620919e-06, "loss": 0.0516, "step": 25000 }, { "epoch": 1.17, "learning_rate": 8.303396014594444e-06, "loss": 0.0394, "step": 25005 }, { "epoch": 1.17, "learning_rate": 8.301057161567969e-06, "loss": 0.0636, "step": 25010 }, { "epoch": 1.17, "learning_rate": 8.298718308541492e-06, "loss": 0.0255, "step": 25015 }, { "epoch": 1.17, "learning_rate": 8.296379455515015e-06, "loss": 0.0914, "step": 25020 }, { "epoch": 1.17, "learning_rate": 8.29404060248854e-06, "loss": 0.0488, "step": 25025 }, { "epoch": 1.17, "learning_rate": 8.291701749462065e-06, "loss": 0.0416, "step": 25030 }, { "epoch": 1.17, "learning_rate": 8.289362896435588e-06, "loss": 0.0366, "step": 25035 }, { "epoch": 1.17, "learning_rate": 8.287024043409113e-06, "loss": 0.0709, "step": 25040 }, { "epoch": 1.17, "learning_rate": 8.284685190382638e-06, "loss": 0.0506, "step": 25045 }, { "epoch": 1.17, "learning_rate": 8.282346337356161e-06, "loss": 0.0891, "step": 25050 }, { "epoch": 1.17, "learning_rate": 8.280007484329685e-06, "loss": 0.0732, "step": 25055 }, { "epoch": 1.17, "learning_rate": 8.27766863130321e-06, "loss": 0.0339, "step": 25060 }, { "epoch": 1.17, "learning_rate": 8.275329778276735e-06, "loss": 0.0316, "step": 25065 }, { "epoch": 1.17, "learning_rate": 8.272990925250258e-06, "loss": 0.0665, "step": 25070 }, { "epoch": 1.17, "learning_rate": 8.270652072223783e-06, "loss": 0.0835, "step": 25075 }, { "epoch": 1.17, "learning_rate": 8.268313219197306e-06, "loss": 0.0376, "step": 25080 }, { "epoch": 1.17, "learning_rate": 8.265974366170831e-06, "loss": 0.1208, "step": 25085 }, { "epoch": 1.17, "learning_rate": 8.263635513144354e-06, "loss": 0.0368, "step": 25090 }, { "epoch": 1.17, "learning_rate": 8.26129666011788e-06, "loss": 0.0491, "step": 25095 }, { "epoch": 1.17, "learning_rate": 8.258957807091404e-06, "loss": 0.0611, "step": 25100 }, { "epoch": 1.17, "learning_rate": 8.256618954064927e-06, "loss": 0.0436, "step": 25105 }, { "epoch": 1.17, "learning_rate": 8.25428010103845e-06, "loss": 0.0525, "step": 25110 }, { "epoch": 1.17, "learning_rate": 8.251941248011976e-06, "loss": 0.0644, "step": 25115 }, { "epoch": 1.18, "learning_rate": 8.2496023949855e-06, "loss": 0.0617, "step": 25120 }, { "epoch": 1.18, "learning_rate": 8.247263541959024e-06, "loss": 0.0363, "step": 25125 }, { "epoch": 1.18, "learning_rate": 8.244924688932549e-06, "loss": 0.0329, "step": 25130 }, { "epoch": 1.18, "learning_rate": 8.242585835906072e-06, "loss": 0.0594, "step": 25135 }, { "epoch": 1.18, "learning_rate": 8.240246982879595e-06, "loss": 0.0359, "step": 25140 }, { "epoch": 1.18, "learning_rate": 8.23790812985312e-06, "loss": 0.0989, "step": 25145 }, { "epoch": 1.18, "learning_rate": 8.235569276826645e-06, "loss": 0.0841, "step": 25150 }, { "epoch": 1.18, "learning_rate": 8.233230423800168e-06, "loss": 0.1119, "step": 25155 }, { "epoch": 1.18, "learning_rate": 8.230891570773693e-06, "loss": 0.0681, "step": 25160 }, { "epoch": 1.18, "learning_rate": 8.228552717747218e-06, "loss": 0.063, "step": 25165 }, { "epoch": 1.18, "learning_rate": 8.226213864720742e-06, "loss": 0.0491, "step": 25170 }, { "epoch": 1.18, "learning_rate": 8.223875011694265e-06, "loss": 0.0478, "step": 25175 }, { "epoch": 1.18, "learning_rate": 8.22153615866779e-06, "loss": 0.0759, "step": 25180 }, { "epoch": 1.18, "learning_rate": 8.219197305641315e-06, "loss": 0.0728, "step": 25185 }, { "epoch": 1.18, "learning_rate": 8.216858452614838e-06, "loss": 0.0309, "step": 25190 }, { "epoch": 1.18, "learning_rate": 8.214519599588363e-06, "loss": 0.0911, "step": 25195 }, { "epoch": 1.18, "learning_rate": 8.212180746561886e-06, "loss": 0.071, "step": 25200 }, { "epoch": 1.18, "learning_rate": 8.209841893535411e-06, "loss": 0.0572, "step": 25205 }, { "epoch": 1.18, "learning_rate": 8.207503040508934e-06, "loss": 0.0648, "step": 25210 }, { "epoch": 1.18, "learning_rate": 8.20516418748246e-06, "loss": 0.0783, "step": 25215 }, { "epoch": 1.18, "learning_rate": 8.202825334455984e-06, "loss": 0.1251, "step": 25220 }, { "epoch": 1.18, "learning_rate": 8.200486481429508e-06, "loss": 0.0624, "step": 25225 }, { "epoch": 1.18, "learning_rate": 8.198147628403031e-06, "loss": 0.0514, "step": 25230 }, { "epoch": 1.18, "learning_rate": 8.195808775376556e-06, "loss": 0.1126, "step": 25235 }, { "epoch": 1.18, "learning_rate": 8.19346992235008e-06, "loss": 0.0731, "step": 25240 }, { "epoch": 1.18, "learning_rate": 8.191131069323604e-06, "loss": 0.0201, "step": 25245 }, { "epoch": 1.18, "learning_rate": 8.188792216297129e-06, "loss": 0.051, "step": 25250 }, { "epoch": 1.18, "learning_rate": 8.186453363270652e-06, "loss": 0.0185, "step": 25255 }, { "epoch": 1.18, "learning_rate": 8.184114510244177e-06, "loss": 0.0283, "step": 25260 }, { "epoch": 1.18, "learning_rate": 8.1817756572177e-06, "loss": 0.0725, "step": 25265 }, { "epoch": 1.18, "learning_rate": 8.179436804191225e-06, "loss": 0.0418, "step": 25270 }, { "epoch": 1.18, "learning_rate": 8.17709795116475e-06, "loss": 0.0785, "step": 25275 }, { "epoch": 1.18, "learning_rate": 8.174759098138274e-06, "loss": 0.0871, "step": 25280 }, { "epoch": 1.18, "learning_rate": 8.172420245111797e-06, "loss": 0.0818, "step": 25285 }, { "epoch": 1.18, "learning_rate": 8.170081392085322e-06, "loss": 0.0962, "step": 25290 }, { "epoch": 1.18, "learning_rate": 8.167742539058847e-06, "loss": 0.0476, "step": 25295 }, { "epoch": 1.18, "learning_rate": 8.16540368603237e-06, "loss": 0.052, "step": 25300 }, { "epoch": 1.18, "learning_rate": 8.163064833005895e-06, "loss": 0.0799, "step": 25305 }, { "epoch": 1.18, "learning_rate": 8.16072597997942e-06, "loss": 0.0483, "step": 25310 }, { "epoch": 1.18, "learning_rate": 8.158387126952943e-06, "loss": 0.082, "step": 25315 }, { "epoch": 1.18, "learning_rate": 8.156048273926466e-06, "loss": 0.0714, "step": 25320 }, { "epoch": 1.18, "learning_rate": 8.153709420899991e-06, "loss": 0.0544, "step": 25325 }, { "epoch": 1.18, "learning_rate": 8.151370567873516e-06, "loss": 0.0746, "step": 25330 }, { "epoch": 1.19, "learning_rate": 8.14903171484704e-06, "loss": 0.0886, "step": 25335 }, { "epoch": 1.19, "learning_rate": 8.146692861820565e-06, "loss": 0.042, "step": 25340 }, { "epoch": 1.19, "learning_rate": 8.144354008794088e-06, "loss": 0.0156, "step": 25345 }, { "epoch": 1.19, "learning_rate": 8.142015155767613e-06, "loss": 0.0485, "step": 25350 }, { "epoch": 1.19, "learning_rate": 8.139676302741136e-06, "loss": 0.0961, "step": 25355 }, { "epoch": 1.19, "learning_rate": 8.137337449714661e-06, "loss": 0.0157, "step": 25360 }, { "epoch": 1.19, "learning_rate": 8.134998596688184e-06, "loss": 0.0437, "step": 25365 }, { "epoch": 1.19, "learning_rate": 8.13265974366171e-06, "loss": 0.0346, "step": 25370 }, { "epoch": 1.19, "learning_rate": 8.130320890635232e-06, "loss": 0.0492, "step": 25375 }, { "epoch": 1.19, "learning_rate": 8.127982037608757e-06, "loss": 0.057, "step": 25380 }, { "epoch": 1.19, "learning_rate": 8.12564318458228e-06, "loss": 0.0567, "step": 25385 }, { "epoch": 1.19, "learning_rate": 8.123304331555806e-06, "loss": 0.115, "step": 25390 }, { "epoch": 1.19, "learning_rate": 8.12096547852933e-06, "loss": 0.0356, "step": 25395 }, { "epoch": 1.19, "learning_rate": 8.118626625502854e-06, "loss": 0.0965, "step": 25400 }, { "epoch": 1.19, "learning_rate": 8.116287772476377e-06, "loss": 0.0875, "step": 25405 }, { "epoch": 1.19, "learning_rate": 8.113948919449902e-06, "loss": 0.0723, "step": 25410 }, { "epoch": 1.19, "learning_rate": 8.111610066423427e-06, "loss": 0.0525, "step": 25415 }, { "epoch": 1.19, "learning_rate": 8.10927121339695e-06, "loss": 0.0567, "step": 25420 }, { "epoch": 1.19, "learning_rate": 8.106932360370475e-06, "loss": 0.0859, "step": 25425 }, { "epoch": 1.19, "learning_rate": 8.104593507344e-06, "loss": 0.034, "step": 25430 }, { "epoch": 1.19, "learning_rate": 8.102254654317523e-06, "loss": 0.0767, "step": 25435 }, { "epoch": 1.19, "learning_rate": 8.099915801291047e-06, "loss": 0.08, "step": 25440 }, { "epoch": 1.19, "learning_rate": 8.097576948264572e-06, "loss": 0.0889, "step": 25445 }, { "epoch": 1.19, "learning_rate": 8.095238095238097e-06, "loss": 0.0507, "step": 25450 }, { "epoch": 1.19, "learning_rate": 8.09289924221162e-06, "loss": 0.0365, "step": 25455 }, { "epoch": 1.19, "learning_rate": 8.090560389185145e-06, "loss": 0.0924, "step": 25460 }, { "epoch": 1.19, "learning_rate": 8.088221536158668e-06, "loss": 0.0619, "step": 25465 }, { "epoch": 1.19, "learning_rate": 8.085882683132193e-06, "loss": 0.0804, "step": 25470 }, { "epoch": 1.19, "learning_rate": 8.083543830105716e-06, "loss": 0.0864, "step": 25475 }, { "epoch": 1.19, "learning_rate": 8.081204977079241e-06, "loss": 0.0582, "step": 25480 }, { "epoch": 1.19, "learning_rate": 8.078866124052766e-06, "loss": 0.0659, "step": 25485 }, { "epoch": 1.19, "learning_rate": 8.07652727102629e-06, "loss": 0.0405, "step": 25490 }, { "epoch": 1.19, "learning_rate": 8.074188417999813e-06, "loss": 0.054, "step": 25495 }, { "epoch": 1.19, "learning_rate": 8.071849564973338e-06, "loss": 0.0388, "step": 25500 }, { "epoch": 1.19, "learning_rate": 8.069510711946863e-06, "loss": 0.031, "step": 25505 }, { "epoch": 1.19, "learning_rate": 8.067171858920386e-06, "loss": 0.0288, "step": 25510 }, { "epoch": 1.19, "learning_rate": 8.06483300589391e-06, "loss": 0.0782, "step": 25515 }, { "epoch": 1.19, "learning_rate": 8.062494152867436e-06, "loss": 0.0768, "step": 25520 }, { "epoch": 1.19, "learning_rate": 8.060155299840959e-06, "loss": 0.0463, "step": 25525 }, { "epoch": 1.19, "learning_rate": 8.057816446814482e-06, "loss": 0.0535, "step": 25530 }, { "epoch": 1.19, "learning_rate": 8.055477593788007e-06, "loss": 0.0393, "step": 25535 }, { "epoch": 1.19, "learning_rate": 8.053138740761532e-06, "loss": 0.1152, "step": 25540 }, { "epoch": 1.19, "learning_rate": 8.050799887735055e-06, "loss": 0.0298, "step": 25545 }, { "epoch": 1.2, "learning_rate": 8.04846103470858e-06, "loss": 0.0327, "step": 25550 }, { "epoch": 1.2, "learning_rate": 8.046122181682104e-06, "loss": 0.0572, "step": 25555 }, { "epoch": 1.2, "learning_rate": 8.043783328655629e-06, "loss": 0.0658, "step": 25560 }, { "epoch": 1.2, "learning_rate": 8.041444475629152e-06, "loss": 0.0488, "step": 25565 }, { "epoch": 1.2, "learning_rate": 8.039105622602677e-06, "loss": 0.072, "step": 25570 }, { "epoch": 1.2, "learning_rate": 8.0367667695762e-06, "loss": 0.0227, "step": 25575 }, { "epoch": 1.2, "learning_rate": 8.034427916549725e-06, "loss": 0.0523, "step": 25580 }, { "epoch": 1.2, "learning_rate": 8.032089063523248e-06, "loss": 0.0308, "step": 25585 }, { "epoch": 1.2, "learning_rate": 8.029750210496773e-06, "loss": 0.0556, "step": 25590 }, { "epoch": 1.2, "learning_rate": 8.027411357470296e-06, "loss": 0.0173, "step": 25595 }, { "epoch": 1.2, "learning_rate": 8.025072504443821e-06, "loss": 0.0304, "step": 25600 }, { "epoch": 1.2, "learning_rate": 8.022733651417346e-06, "loss": 0.0477, "step": 25605 }, { "epoch": 1.2, "learning_rate": 8.02039479839087e-06, "loss": 0.0748, "step": 25610 }, { "epoch": 1.2, "learning_rate": 8.018055945364393e-06, "loss": 0.0591, "step": 25615 }, { "epoch": 1.2, "learning_rate": 8.015717092337918e-06, "loss": 0.056, "step": 25620 }, { "epoch": 1.2, "learning_rate": 8.013378239311443e-06, "loss": 0.033, "step": 25625 }, { "epoch": 1.2, "learning_rate": 8.011039386284966e-06, "loss": 0.1086, "step": 25630 }, { "epoch": 1.2, "learning_rate": 8.008700533258491e-06, "loss": 0.0339, "step": 25635 }, { "epoch": 1.2, "learning_rate": 8.006361680232014e-06, "loss": 0.077, "step": 25640 }, { "epoch": 1.2, "learning_rate": 8.00402282720554e-06, "loss": 0.0598, "step": 25645 }, { "epoch": 1.2, "learning_rate": 8.001683974179062e-06, "loss": 0.0364, "step": 25650 }, { "epoch": 1.2, "learning_rate": 7.999345121152587e-06, "loss": 0.0874, "step": 25655 }, { "epoch": 1.2, "learning_rate": 7.997006268126112e-06, "loss": 0.0403, "step": 25660 }, { "epoch": 1.2, "learning_rate": 7.994667415099636e-06, "loss": 0.0637, "step": 25665 }, { "epoch": 1.2, "learning_rate": 7.992328562073159e-06, "loss": 0.0886, "step": 25670 }, { "epoch": 1.2, "learning_rate": 7.989989709046684e-06, "loss": 0.0863, "step": 25675 }, { "epoch": 1.2, "learning_rate": 7.987650856020209e-06, "loss": 0.0766, "step": 25680 }, { "epoch": 1.2, "learning_rate": 7.985312002993732e-06, "loss": 0.0836, "step": 25685 }, { "epoch": 1.2, "learning_rate": 7.982973149967257e-06, "loss": 0.046, "step": 25690 }, { "epoch": 1.2, "learning_rate": 7.980634296940782e-06, "loss": 0.0508, "step": 25695 }, { "epoch": 1.2, "learning_rate": 7.978295443914305e-06, "loss": 0.0785, "step": 25700 }, { "epoch": 1.2, "learning_rate": 7.975956590887828e-06, "loss": 0.0738, "step": 25705 }, { "epoch": 1.2, "learning_rate": 7.973617737861353e-06, "loss": 0.0384, "step": 25710 }, { "epoch": 1.2, "learning_rate": 7.971278884834878e-06, "loss": 0.0674, "step": 25715 }, { "epoch": 1.2, "learning_rate": 7.968940031808402e-06, "loss": 0.0653, "step": 25720 }, { "epoch": 1.2, "learning_rate": 7.966601178781927e-06, "loss": 0.0272, "step": 25725 }, { "epoch": 1.2, "learning_rate": 7.96426232575545e-06, "loss": 0.0546, "step": 25730 }, { "epoch": 1.2, "learning_rate": 7.961923472728975e-06, "loss": 0.0394, "step": 25735 }, { "epoch": 1.2, "learning_rate": 7.959584619702498e-06, "loss": 0.1115, "step": 25740 }, { "epoch": 1.2, "learning_rate": 7.957245766676023e-06, "loss": 0.1154, "step": 25745 }, { "epoch": 1.2, "learning_rate": 7.954906913649548e-06, "loss": 0.044, "step": 25750 }, { "epoch": 1.2, "learning_rate": 7.952568060623071e-06, "loss": 0.0255, "step": 25755 }, { "epoch": 1.2, "learning_rate": 7.950229207596594e-06, "loss": 0.0466, "step": 25760 }, { "epoch": 1.21, "learning_rate": 7.94789035457012e-06, "loss": 0.1063, "step": 25765 }, { "epoch": 1.21, "learning_rate": 7.945551501543644e-06, "loss": 0.0608, "step": 25770 }, { "epoch": 1.21, "learning_rate": 7.943212648517168e-06, "loss": 0.1214, "step": 25775 }, { "epoch": 1.21, "learning_rate": 7.940873795490693e-06, "loss": 0.0856, "step": 25780 }, { "epoch": 1.21, "learning_rate": 7.938534942464217e-06, "loss": 0.0419, "step": 25785 }, { "epoch": 1.21, "learning_rate": 7.93619608943774e-06, "loss": 0.0712, "step": 25790 }, { "epoch": 1.21, "learning_rate": 7.933857236411264e-06, "loss": 0.0611, "step": 25795 }, { "epoch": 1.21, "learning_rate": 7.931518383384789e-06, "loss": 0.0481, "step": 25800 }, { "epoch": 1.21, "learning_rate": 7.929179530358312e-06, "loss": 0.0483, "step": 25805 }, { "epoch": 1.21, "learning_rate": 7.926840677331837e-06, "loss": 0.0409, "step": 25810 }, { "epoch": 1.21, "learning_rate": 7.924501824305362e-06, "loss": 0.047, "step": 25815 }, { "epoch": 1.21, "learning_rate": 7.922162971278885e-06, "loss": 0.0204, "step": 25820 }, { "epoch": 1.21, "learning_rate": 7.919824118252409e-06, "loss": 0.0322, "step": 25825 }, { "epoch": 1.21, "learning_rate": 7.917485265225934e-06, "loss": 0.0477, "step": 25830 }, { "epoch": 1.21, "learning_rate": 7.915146412199459e-06, "loss": 0.0255, "step": 25835 }, { "epoch": 1.21, "learning_rate": 7.912807559172982e-06, "loss": 0.0517, "step": 25840 }, { "epoch": 1.21, "learning_rate": 7.910468706146507e-06, "loss": 0.0673, "step": 25845 }, { "epoch": 1.21, "learning_rate": 7.90812985312003e-06, "loss": 0.0686, "step": 25850 }, { "epoch": 1.21, "learning_rate": 7.905791000093555e-06, "loss": 0.0211, "step": 25855 }, { "epoch": 1.21, "learning_rate": 7.903452147067078e-06, "loss": 0.0735, "step": 25860 }, { "epoch": 1.21, "learning_rate": 7.901113294040603e-06, "loss": 0.1585, "step": 25865 }, { "epoch": 1.21, "learning_rate": 7.898774441014128e-06, "loss": 0.1054, "step": 25870 }, { "epoch": 1.21, "learning_rate": 7.896435587987651e-06, "loss": 0.0726, "step": 25875 }, { "epoch": 1.21, "learning_rate": 7.894096734961175e-06, "loss": 0.0511, "step": 25880 }, { "epoch": 1.21, "learning_rate": 7.8917578819347e-06, "loss": 0.0851, "step": 25885 }, { "epoch": 1.21, "learning_rate": 7.889419028908225e-06, "loss": 0.0607, "step": 25890 }, { "epoch": 1.21, "learning_rate": 7.887080175881748e-06, "loss": 0.0768, "step": 25895 }, { "epoch": 1.21, "learning_rate": 7.884741322855273e-06, "loss": 0.0466, "step": 25900 }, { "epoch": 1.21, "learning_rate": 7.882402469828798e-06, "loss": 0.1284, "step": 25905 }, { "epoch": 1.21, "learning_rate": 7.880063616802321e-06, "loss": 0.0826, "step": 25910 }, { "epoch": 1.21, "learning_rate": 7.877724763775844e-06, "loss": 0.0723, "step": 25915 }, { "epoch": 1.21, "learning_rate": 7.875385910749369e-06, "loss": 0.0418, "step": 25920 }, { "epoch": 1.21, "learning_rate": 7.873047057722894e-06, "loss": 0.0432, "step": 25925 }, { "epoch": 1.21, "learning_rate": 7.870708204696417e-06, "loss": 0.0537, "step": 25930 }, { "epoch": 1.21, "learning_rate": 7.868369351669942e-06, "loss": 0.0438, "step": 25935 }, { "epoch": 1.21, "learning_rate": 7.866030498643466e-06, "loss": 0.0409, "step": 25940 }, { "epoch": 1.21, "learning_rate": 7.86369164561699e-06, "loss": 0.0685, "step": 25945 }, { "epoch": 1.21, "learning_rate": 7.861352792590514e-06, "loss": 0.0256, "step": 25950 }, { "epoch": 1.21, "learning_rate": 7.859013939564039e-06, "loss": 0.0687, "step": 25955 }, { "epoch": 1.21, "learning_rate": 7.856675086537564e-06, "loss": 0.0561, "step": 25960 }, { "epoch": 1.21, "learning_rate": 7.854336233511087e-06, "loss": 0.0377, "step": 25965 }, { "epoch": 1.21, "learning_rate": 7.85199738048461e-06, "loss": 0.1035, "step": 25970 }, { "epoch": 1.22, "learning_rate": 7.849658527458135e-06, "loss": 0.0692, "step": 25975 }, { "epoch": 1.22, "learning_rate": 7.84731967443166e-06, "loss": 0.093, "step": 25980 }, { "epoch": 1.22, "learning_rate": 7.844980821405183e-06, "loss": 0.0192, "step": 25985 }, { "epoch": 1.22, "learning_rate": 7.842641968378708e-06, "loss": 0.0398, "step": 25990 }, { "epoch": 1.22, "learning_rate": 7.840303115352233e-06, "loss": 0.0468, "step": 25995 }, { "epoch": 1.22, "learning_rate": 7.837964262325757e-06, "loss": 0.0391, "step": 26000 }, { "epoch": 1.22, "learning_rate": 7.83562540929928e-06, "loss": 0.0676, "step": 26005 }, { "epoch": 1.22, "learning_rate": 7.833286556272805e-06, "loss": 0.0745, "step": 26010 }, { "epoch": 1.22, "learning_rate": 7.830947703246328e-06, "loss": 0.0262, "step": 26015 }, { "epoch": 1.22, "learning_rate": 7.828608850219853e-06, "loss": 0.043, "step": 26020 }, { "epoch": 1.22, "learning_rate": 7.826269997193378e-06, "loss": 0.091, "step": 26025 }, { "epoch": 1.22, "learning_rate": 7.823931144166901e-06, "loss": 0.0881, "step": 26030 }, { "epoch": 1.22, "learning_rate": 7.821592291140424e-06, "loss": 0.0597, "step": 26035 }, { "epoch": 1.22, "learning_rate": 7.81925343811395e-06, "loss": 0.0312, "step": 26040 }, { "epoch": 1.22, "learning_rate": 7.816914585087474e-06, "loss": 0.0196, "step": 26045 }, { "epoch": 1.22, "learning_rate": 7.814575732060998e-06, "loss": 0.0863, "step": 26050 }, { "epoch": 1.22, "learning_rate": 7.812236879034523e-06, "loss": 0.0435, "step": 26055 }, { "epoch": 1.22, "learning_rate": 7.809898026008046e-06, "loss": 0.0837, "step": 26060 }, { "epoch": 1.22, "learning_rate": 7.80755917298157e-06, "loss": 0.0796, "step": 26065 }, { "epoch": 1.22, "learning_rate": 7.805220319955094e-06, "loss": 0.123, "step": 26070 }, { "epoch": 1.22, "learning_rate": 7.802881466928619e-06, "loss": 0.0376, "step": 26075 }, { "epoch": 1.22, "learning_rate": 7.800542613902144e-06, "loss": 0.0895, "step": 26080 }, { "epoch": 1.22, "learning_rate": 7.798203760875667e-06, "loss": 0.0484, "step": 26085 }, { "epoch": 1.22, "learning_rate": 7.79586490784919e-06, "loss": 0.0576, "step": 26090 }, { "epoch": 1.22, "learning_rate": 7.793526054822715e-06, "loss": 0.0528, "step": 26095 }, { "epoch": 1.22, "learning_rate": 7.79118720179624e-06, "loss": 0.1078, "step": 26100 }, { "epoch": 1.22, "learning_rate": 7.788848348769764e-06, "loss": 0.0382, "step": 26105 }, { "epoch": 1.22, "learning_rate": 7.786509495743288e-06, "loss": 0.0232, "step": 26110 }, { "epoch": 1.22, "learning_rate": 7.784170642716812e-06, "loss": 0.0199, "step": 26115 }, { "epoch": 1.22, "learning_rate": 7.781831789690337e-06, "loss": 0.0739, "step": 26120 }, { "epoch": 1.22, "learning_rate": 7.77949293666386e-06, "loss": 0.0263, "step": 26125 }, { "epoch": 1.22, "learning_rate": 7.777154083637385e-06, "loss": 0.061, "step": 26130 }, { "epoch": 1.22, "learning_rate": 7.77481523061091e-06, "loss": 0.0665, "step": 26135 }, { "epoch": 1.22, "learning_rate": 7.772476377584433e-06, "loss": 0.0462, "step": 26140 }, { "epoch": 1.22, "learning_rate": 7.770137524557956e-06, "loss": 0.0479, "step": 26145 }, { "epoch": 1.22, "learning_rate": 7.767798671531481e-06, "loss": 0.0784, "step": 26150 }, { "epoch": 1.22, "learning_rate": 7.765459818505006e-06, "loss": 0.0576, "step": 26155 }, { "epoch": 1.22, "learning_rate": 7.76312096547853e-06, "loss": 0.0525, "step": 26160 }, { "epoch": 1.22, "learning_rate": 7.760782112452054e-06, "loss": 0.0185, "step": 26165 }, { "epoch": 1.22, "learning_rate": 7.75844325942558e-06, "loss": 0.0563, "step": 26170 }, { "epoch": 1.22, "learning_rate": 7.756104406399103e-06, "loss": 0.0385, "step": 26175 }, { "epoch": 1.22, "learning_rate": 7.753765553372626e-06, "loss": 0.0255, "step": 26180 }, { "epoch": 1.22, "learning_rate": 7.751426700346151e-06, "loss": 0.0572, "step": 26185 }, { "epoch": 1.23, "learning_rate": 7.749087847319676e-06, "loss": 0.0769, "step": 26190 }, { "epoch": 1.23, "learning_rate": 7.746748994293199e-06, "loss": 0.0063, "step": 26195 }, { "epoch": 1.23, "learning_rate": 7.744410141266724e-06, "loss": 0.0529, "step": 26200 }, { "epoch": 1.23, "learning_rate": 7.742071288240247e-06, "loss": 0.0263, "step": 26205 }, { "epoch": 1.23, "learning_rate": 7.739732435213772e-06, "loss": 0.0478, "step": 26210 }, { "epoch": 1.23, "learning_rate": 7.737393582187296e-06, "loss": 0.1, "step": 26215 }, { "epoch": 1.23, "learning_rate": 7.73505472916082e-06, "loss": 0.0465, "step": 26220 }, { "epoch": 1.23, "learning_rate": 7.732715876134345e-06, "loss": 0.0147, "step": 26225 }, { "epoch": 1.23, "learning_rate": 7.730377023107869e-06, "loss": 0.0263, "step": 26230 }, { "epoch": 1.23, "learning_rate": 7.728038170081392e-06, "loss": 0.0658, "step": 26235 }, { "epoch": 1.23, "learning_rate": 7.725699317054917e-06, "loss": 0.0427, "step": 26240 }, { "epoch": 1.23, "learning_rate": 7.72336046402844e-06, "loss": 0.0584, "step": 26245 }, { "epoch": 1.23, "learning_rate": 7.721021611001965e-06, "loss": 0.0412, "step": 26250 }, { "epoch": 1.23, "learning_rate": 7.71868275797549e-06, "loss": 0.0438, "step": 26255 }, { "epoch": 1.23, "learning_rate": 7.716343904949013e-06, "loss": 0.0389, "step": 26260 }, { "epoch": 1.23, "learning_rate": 7.714005051922537e-06, "loss": 0.0573, "step": 26265 }, { "epoch": 1.23, "learning_rate": 7.711666198896062e-06, "loss": 0.0579, "step": 26270 }, { "epoch": 1.23, "learning_rate": 7.709327345869586e-06, "loss": 0.0438, "step": 26275 }, { "epoch": 1.23, "learning_rate": 7.70698849284311e-06, "loss": 0.0845, "step": 26280 }, { "epoch": 1.23, "learning_rate": 7.704649639816635e-06, "loss": 0.082, "step": 26285 }, { "epoch": 1.23, "learning_rate": 7.70231078679016e-06, "loss": 0.1284, "step": 26290 }, { "epoch": 1.23, "learning_rate": 7.699971933763683e-06, "loss": 0.057, "step": 26295 }, { "epoch": 1.23, "learning_rate": 7.697633080737206e-06, "loss": 0.0224, "step": 26300 }, { "epoch": 1.23, "learning_rate": 7.695294227710731e-06, "loss": 0.0708, "step": 26305 }, { "epoch": 1.23, "learning_rate": 7.692955374684256e-06, "loss": 0.0546, "step": 26310 }, { "epoch": 1.23, "learning_rate": 7.69061652165778e-06, "loss": 0.0476, "step": 26315 }, { "epoch": 1.23, "learning_rate": 7.688277668631304e-06, "loss": 0.0739, "step": 26320 }, { "epoch": 1.23, "learning_rate": 7.685938815604828e-06, "loss": 0.0773, "step": 26325 }, { "epoch": 1.23, "learning_rate": 7.683599962578352e-06, "loss": 0.095, "step": 26330 }, { "epoch": 1.23, "learning_rate": 7.681261109551876e-06, "loss": 0.0744, "step": 26335 }, { "epoch": 1.23, "learning_rate": 7.6789222565254e-06, "loss": 0.0614, "step": 26340 }, { "epoch": 1.23, "learning_rate": 7.676583403498926e-06, "loss": 0.0537, "step": 26345 }, { "epoch": 1.23, "learning_rate": 7.674244550472449e-06, "loss": 0.0725, "step": 26350 }, { "epoch": 1.23, "learning_rate": 7.671905697445972e-06, "loss": 0.0377, "step": 26355 }, { "epoch": 1.23, "learning_rate": 7.669566844419497e-06, "loss": 0.0746, "step": 26360 }, { "epoch": 1.23, "learning_rate": 7.667227991393022e-06, "loss": 0.05, "step": 26365 }, { "epoch": 1.23, "learning_rate": 7.664889138366545e-06, "loss": 0.0128, "step": 26370 }, { "epoch": 1.23, "learning_rate": 7.66255028534007e-06, "loss": 0.0504, "step": 26375 }, { "epoch": 1.23, "learning_rate": 7.660211432313595e-06, "loss": 0.082, "step": 26380 }, { "epoch": 1.23, "learning_rate": 7.657872579287118e-06, "loss": 0.0392, "step": 26385 }, { "epoch": 1.23, "learning_rate": 7.655533726260642e-06, "loss": 0.0799, "step": 26390 }, { "epoch": 1.23, "learning_rate": 7.653194873234167e-06, "loss": 0.0762, "step": 26395 }, { "epoch": 1.23, "learning_rate": 7.650856020207692e-06, "loss": 0.0314, "step": 26400 }, { "epoch": 1.24, "learning_rate": 7.648517167181215e-06, "loss": 0.0425, "step": 26405 }, { "epoch": 1.24, "learning_rate": 7.64617831415474e-06, "loss": 0.073, "step": 26410 }, { "epoch": 1.24, "learning_rate": 7.643839461128263e-06, "loss": 0.06, "step": 26415 }, { "epoch": 1.24, "learning_rate": 7.641500608101788e-06, "loss": 0.0476, "step": 26420 }, { "epoch": 1.24, "learning_rate": 7.639161755075311e-06, "loss": 0.0324, "step": 26425 }, { "epoch": 1.24, "learning_rate": 7.636822902048836e-06, "loss": 0.0561, "step": 26430 }, { "epoch": 1.24, "learning_rate": 7.634484049022361e-06, "loss": 0.0602, "step": 26435 }, { "epoch": 1.24, "learning_rate": 7.632145195995884e-06, "loss": 0.0376, "step": 26440 }, { "epoch": 1.24, "learning_rate": 7.629806342969408e-06, "loss": 0.0896, "step": 26445 }, { "epoch": 1.24, "learning_rate": 7.627467489942933e-06, "loss": 0.0475, "step": 26450 }, { "epoch": 1.24, "learning_rate": 7.625128636916457e-06, "loss": 0.058, "step": 26455 }, { "epoch": 1.24, "learning_rate": 7.622789783889981e-06, "loss": 0.0511, "step": 26460 }, { "epoch": 1.24, "learning_rate": 7.620450930863506e-06, "loss": 0.037, "step": 26465 }, { "epoch": 1.24, "learning_rate": 7.61811207783703e-06, "loss": 0.1029, "step": 26470 }, { "epoch": 1.24, "learning_rate": 7.615773224810553e-06, "loss": 0.0783, "step": 26475 }, { "epoch": 1.24, "learning_rate": 7.613434371784077e-06, "loss": 0.0817, "step": 26480 }, { "epoch": 1.24, "learning_rate": 7.611095518757602e-06, "loss": 0.0447, "step": 26485 }, { "epoch": 1.24, "learning_rate": 7.608756665731126e-06, "loss": 0.0529, "step": 26490 }, { "epoch": 1.24, "learning_rate": 7.6064178127046505e-06, "loss": 0.0507, "step": 26495 }, { "epoch": 1.24, "learning_rate": 7.604078959678175e-06, "loss": 0.079, "step": 26500 }, { "epoch": 1.24, "learning_rate": 7.601740106651699e-06, "loss": 0.0571, "step": 26505 }, { "epoch": 1.24, "learning_rate": 7.599401253625223e-06, "loss": 0.0923, "step": 26510 }, { "epoch": 1.24, "learning_rate": 7.597062400598747e-06, "loss": 0.0925, "step": 26515 }, { "epoch": 1.24, "learning_rate": 7.594723547572271e-06, "loss": 0.0634, "step": 26520 }, { "epoch": 1.24, "learning_rate": 7.592384694545796e-06, "loss": 0.0995, "step": 26525 }, { "epoch": 1.24, "learning_rate": 7.590045841519319e-06, "loss": 0.1, "step": 26530 }, { "epoch": 1.24, "learning_rate": 7.587706988492843e-06, "loss": 0.0793, "step": 26535 }, { "epoch": 1.24, "learning_rate": 7.585368135466367e-06, "loss": 0.0608, "step": 26540 }, { "epoch": 1.24, "learning_rate": 7.583029282439892e-06, "loss": 0.0871, "step": 26545 }, { "epoch": 1.24, "learning_rate": 7.5806904294134165e-06, "loss": 0.0764, "step": 26550 }, { "epoch": 1.24, "learning_rate": 7.5783515763869406e-06, "loss": 0.0556, "step": 26555 }, { "epoch": 1.24, "learning_rate": 7.576012723360464e-06, "loss": 0.0725, "step": 26560 }, { "epoch": 1.24, "learning_rate": 7.573673870333989e-06, "loss": 0.052, "step": 26565 }, { "epoch": 1.24, "learning_rate": 7.571335017307513e-06, "loss": 0.0549, "step": 26570 }, { "epoch": 1.24, "learning_rate": 7.568996164281037e-06, "loss": 0.0552, "step": 26575 }, { "epoch": 1.24, "learning_rate": 7.566657311254562e-06, "loss": 0.0483, "step": 26580 }, { "epoch": 1.24, "learning_rate": 7.564318458228086e-06, "loss": 0.0533, "step": 26585 }, { "epoch": 1.24, "learning_rate": 7.561979605201609e-06, "loss": 0.0926, "step": 26590 }, { "epoch": 1.24, "learning_rate": 7.559640752175133e-06, "loss": 0.036, "step": 26595 }, { "epoch": 1.24, "learning_rate": 7.557301899148658e-06, "loss": 0.045, "step": 26600 }, { "epoch": 1.24, "learning_rate": 7.5549630461221825e-06, "loss": 0.0526, "step": 26605 }, { "epoch": 1.24, "learning_rate": 7.5526241930957066e-06, "loss": 0.1506, "step": 26610 }, { "epoch": 1.24, "learning_rate": 7.550285340069231e-06, "loss": 0.0387, "step": 26615 }, { "epoch": 1.25, "learning_rate": 7.547946487042754e-06, "loss": 0.0855, "step": 26620 }, { "epoch": 1.25, "learning_rate": 7.545607634016279e-06, "loss": 0.0638, "step": 26625 }, { "epoch": 1.25, "learning_rate": 7.543268780989803e-06, "loss": 0.0618, "step": 26630 }, { "epoch": 1.25, "learning_rate": 7.540929927963327e-06, "loss": 0.0379, "step": 26635 }, { "epoch": 1.25, "learning_rate": 7.538591074936852e-06, "loss": 0.0658, "step": 26640 }, { "epoch": 1.25, "learning_rate": 7.536252221910376e-06, "loss": 0.0781, "step": 26645 }, { "epoch": 1.25, "learning_rate": 7.533913368883899e-06, "loss": 0.0165, "step": 26650 }, { "epoch": 1.25, "learning_rate": 7.5315745158574235e-06, "loss": 0.0367, "step": 26655 }, { "epoch": 1.25, "learning_rate": 7.5292356628309485e-06, "loss": 0.0784, "step": 26660 }, { "epoch": 1.25, "learning_rate": 7.5268968098044726e-06, "loss": 0.0466, "step": 26665 }, { "epoch": 1.25, "learning_rate": 7.524557956777997e-06, "loss": 0.0729, "step": 26670 }, { "epoch": 1.25, "learning_rate": 7.522219103751522e-06, "loss": 0.0661, "step": 26675 }, { "epoch": 1.25, "learning_rate": 7.519880250725045e-06, "loss": 0.0474, "step": 26680 }, { "epoch": 1.25, "learning_rate": 7.517541397698569e-06, "loss": 0.0458, "step": 26685 }, { "epoch": 1.25, "learning_rate": 7.515202544672093e-06, "loss": 0.0398, "step": 26690 }, { "epoch": 1.25, "learning_rate": 7.512863691645618e-06, "loss": 0.0551, "step": 26695 }, { "epoch": 1.25, "learning_rate": 7.510524838619142e-06, "loss": 0.04, "step": 26700 }, { "epoch": 1.25, "learning_rate": 7.508185985592666e-06, "loss": 0.1272, "step": 26705 }, { "epoch": 1.25, "learning_rate": 7.5058471325661895e-06, "loss": 0.0465, "step": 26710 }, { "epoch": 1.25, "learning_rate": 7.5035082795397145e-06, "loss": 0.0209, "step": 26715 }, { "epoch": 1.25, "learning_rate": 7.5011694265132386e-06, "loss": 0.0909, "step": 26720 }, { "epoch": 1.25, "learning_rate": 7.498830573486763e-06, "loss": 0.0905, "step": 26725 }, { "epoch": 1.25, "learning_rate": 7.496491720460287e-06, "loss": 0.0803, "step": 26730 }, { "epoch": 1.25, "learning_rate": 7.494152867433812e-06, "loss": 0.0279, "step": 26735 }, { "epoch": 1.25, "learning_rate": 7.491814014407335e-06, "loss": 0.0481, "step": 26740 }, { "epoch": 1.25, "learning_rate": 7.489475161380859e-06, "loss": 0.0523, "step": 26745 }, { "epoch": 1.25, "learning_rate": 7.487136308354383e-06, "loss": 0.0473, "step": 26750 }, { "epoch": 1.25, "learning_rate": 7.484797455327908e-06, "loss": 0.0607, "step": 26755 }, { "epoch": 1.25, "learning_rate": 7.482458602301432e-06, "loss": 0.0996, "step": 26760 }, { "epoch": 1.25, "learning_rate": 7.480119749274956e-06, "loss": 0.0357, "step": 26765 }, { "epoch": 1.25, "learning_rate": 7.47778089624848e-06, "loss": 0.1165, "step": 26770 }, { "epoch": 1.25, "learning_rate": 7.4754420432220046e-06, "loss": 0.0453, "step": 26775 }, { "epoch": 1.25, "learning_rate": 7.473103190195529e-06, "loss": 0.0522, "step": 26780 }, { "epoch": 1.25, "learning_rate": 7.470764337169053e-06, "loss": 0.0455, "step": 26785 }, { "epoch": 1.25, "learning_rate": 7.468425484142578e-06, "loss": 0.0488, "step": 26790 }, { "epoch": 1.25, "learning_rate": 7.466086631116102e-06, "loss": 0.0681, "step": 26795 }, { "epoch": 1.25, "learning_rate": 7.463747778089625e-06, "loss": 0.057, "step": 26800 }, { "epoch": 1.25, "learning_rate": 7.461408925063149e-06, "loss": 0.0596, "step": 26805 }, { "epoch": 1.25, "learning_rate": 7.459070072036674e-06, "loss": 0.0925, "step": 26810 }, { "epoch": 1.25, "learning_rate": 7.456731219010198e-06, "loss": 0.0758, "step": 26815 }, { "epoch": 1.25, "learning_rate": 7.454392365983722e-06, "loss": 0.0857, "step": 26820 }, { "epoch": 1.25, "learning_rate": 7.452053512957247e-06, "loss": 0.0396, "step": 26825 }, { "epoch": 1.26, "learning_rate": 7.4497146599307706e-06, "loss": 0.0534, "step": 26830 }, { "epoch": 1.26, "learning_rate": 7.447375806904295e-06, "loss": 0.032, "step": 26835 }, { "epoch": 1.26, "learning_rate": 7.445036953877819e-06, "loss": 0.065, "step": 26840 }, { "epoch": 1.26, "learning_rate": 7.442698100851343e-06, "loss": 0.0422, "step": 26845 }, { "epoch": 1.26, "learning_rate": 7.440359247824868e-06, "loss": 0.1478, "step": 26850 }, { "epoch": 1.26, "learning_rate": 7.438020394798392e-06, "loss": 0.0556, "step": 26855 }, { "epoch": 1.26, "learning_rate": 7.435681541771915e-06, "loss": 0.0691, "step": 26860 }, { "epoch": 1.26, "learning_rate": 7.433342688745439e-06, "loss": 0.0193, "step": 26865 }, { "epoch": 1.26, "learning_rate": 7.431003835718964e-06, "loss": 0.0603, "step": 26870 }, { "epoch": 1.26, "learning_rate": 7.428664982692488e-06, "loss": 0.0622, "step": 26875 }, { "epoch": 1.26, "learning_rate": 7.4263261296660124e-06, "loss": 0.0911, "step": 26880 }, { "epoch": 1.26, "learning_rate": 7.423987276639537e-06, "loss": 0.0625, "step": 26885 }, { "epoch": 1.26, "learning_rate": 7.421648423613061e-06, "loss": 0.0775, "step": 26890 }, { "epoch": 1.26, "learning_rate": 7.419309570586585e-06, "loss": 0.0877, "step": 26895 }, { "epoch": 1.26, "learning_rate": 7.416970717560109e-06, "loss": 0.0406, "step": 26900 }, { "epoch": 1.26, "learning_rate": 7.414631864533634e-06, "loss": 0.0285, "step": 26905 }, { "epoch": 1.26, "learning_rate": 7.412293011507158e-06, "loss": 0.0636, "step": 26910 }, { "epoch": 1.26, "learning_rate": 7.409954158480682e-06, "loss": 0.0706, "step": 26915 }, { "epoch": 1.26, "learning_rate": 7.407615305454205e-06, "loss": 0.0918, "step": 26920 }, { "epoch": 1.26, "learning_rate": 7.40527645242773e-06, "loss": 0.0562, "step": 26925 }, { "epoch": 1.26, "learning_rate": 7.402937599401254e-06, "loss": 0.0591, "step": 26930 }, { "epoch": 1.26, "learning_rate": 7.4005987463747784e-06, "loss": 0.066, "step": 26935 }, { "epoch": 1.26, "learning_rate": 7.3982598933483025e-06, "loss": 0.0761, "step": 26940 }, { "epoch": 1.26, "learning_rate": 7.395921040321827e-06, "loss": 0.0785, "step": 26945 }, { "epoch": 1.26, "learning_rate": 7.393582187295351e-06, "loss": 0.0619, "step": 26950 }, { "epoch": 1.26, "learning_rate": 7.391243334268875e-06, "loss": 0.0858, "step": 26955 }, { "epoch": 1.26, "learning_rate": 7.388904481242399e-06, "loss": 0.05, "step": 26960 }, { "epoch": 1.26, "learning_rate": 7.386565628215924e-06, "loss": 0.0457, "step": 26965 }, { "epoch": 1.26, "learning_rate": 7.384226775189448e-06, "loss": 0.036, "step": 26970 }, { "epoch": 1.26, "learning_rate": 7.381887922162971e-06, "loss": 0.0196, "step": 26975 }, { "epoch": 1.26, "learning_rate": 7.379549069136495e-06, "loss": 0.0788, "step": 26980 }, { "epoch": 1.26, "learning_rate": 7.37721021611002e-06, "loss": 0.081, "step": 26985 }, { "epoch": 1.26, "learning_rate": 7.3748713630835444e-06, "loss": 0.0855, "step": 26990 }, { "epoch": 1.26, "learning_rate": 7.3725325100570685e-06, "loss": 0.0441, "step": 26995 }, { "epoch": 1.26, "learning_rate": 7.3701936570305935e-06, "loss": 0.0676, "step": 27000 }, { "epoch": 1.26, "learning_rate": 7.367854804004117e-06, "loss": 0.0461, "step": 27005 }, { "epoch": 1.26, "learning_rate": 7.365515950977641e-06, "loss": 0.0402, "step": 27010 }, { "epoch": 1.26, "learning_rate": 7.363177097951165e-06, "loss": 0.0267, "step": 27015 }, { "epoch": 1.26, "learning_rate": 7.36083824492469e-06, "loss": 0.1015, "step": 27020 }, { "epoch": 1.26, "learning_rate": 7.358499391898214e-06, "loss": 0.0257, "step": 27025 }, { "epoch": 1.26, "learning_rate": 7.356160538871738e-06, "loss": 0.0705, "step": 27030 }, { "epoch": 1.26, "learning_rate": 7.353821685845261e-06, "loss": 0.0752, "step": 27035 }, { "epoch": 1.26, "learning_rate": 7.351482832818786e-06, "loss": 0.0638, "step": 27040 }, { "epoch": 1.27, "learning_rate": 7.3491439797923104e-06, "loss": 0.0181, "step": 27045 }, { "epoch": 1.27, "learning_rate": 7.3468051267658345e-06, "loss": 0.0816, "step": 27050 }, { "epoch": 1.27, "learning_rate": 7.344466273739359e-06, "loss": 0.1121, "step": 27055 }, { "epoch": 1.27, "learning_rate": 7.342127420712884e-06, "loss": 0.0678, "step": 27060 }, { "epoch": 1.27, "learning_rate": 7.339788567686407e-06, "loss": 0.0319, "step": 27065 }, { "epoch": 1.27, "learning_rate": 7.337449714659931e-06, "loss": 0.0877, "step": 27070 }, { "epoch": 1.27, "learning_rate": 7.335110861633455e-06, "loss": 0.0309, "step": 27075 }, { "epoch": 1.27, "learning_rate": 7.33277200860698e-06, "loss": 0.0509, "step": 27080 }, { "epoch": 1.27, "learning_rate": 7.330433155580504e-06, "loss": 0.0232, "step": 27085 }, { "epoch": 1.27, "learning_rate": 7.328094302554028e-06, "loss": 0.0501, "step": 27090 }, { "epoch": 1.27, "learning_rate": 7.3257554495275515e-06, "loss": 0.0537, "step": 27095 }, { "epoch": 1.27, "learning_rate": 7.3234165965010764e-06, "loss": 0.0264, "step": 27100 }, { "epoch": 1.27, "learning_rate": 7.3210777434746005e-06, "loss": 0.0282, "step": 27105 }, { "epoch": 1.27, "learning_rate": 7.318738890448125e-06, "loss": 0.043, "step": 27110 }, { "epoch": 1.27, "learning_rate": 7.31640003742165e-06, "loss": 0.08, "step": 27115 }, { "epoch": 1.27, "learning_rate": 7.314061184395174e-06, "loss": 0.0763, "step": 27120 }, { "epoch": 1.27, "learning_rate": 7.311722331368697e-06, "loss": 0.0723, "step": 27125 }, { "epoch": 1.27, "learning_rate": 7.309383478342221e-06, "loss": 0.034, "step": 27130 }, { "epoch": 1.27, "learning_rate": 7.307044625315746e-06, "loss": 0.0319, "step": 27135 }, { "epoch": 1.27, "learning_rate": 7.30470577228927e-06, "loss": 0.0533, "step": 27140 }, { "epoch": 1.27, "learning_rate": 7.302366919262794e-06, "loss": 0.0392, "step": 27145 }, { "epoch": 1.27, "learning_rate": 7.300028066236319e-06, "loss": 0.0276, "step": 27150 }, { "epoch": 1.27, "learning_rate": 7.2976892132098424e-06, "loss": 0.0278, "step": 27155 }, { "epoch": 1.27, "learning_rate": 7.2953503601833665e-06, "loss": 0.0871, "step": 27160 }, { "epoch": 1.27, "learning_rate": 7.293011507156891e-06, "loss": 0.0147, "step": 27165 }, { "epoch": 1.27, "learning_rate": 7.290672654130415e-06, "loss": 0.0048, "step": 27170 }, { "epoch": 1.27, "learning_rate": 7.28833380110394e-06, "loss": 0.0609, "step": 27175 }, { "epoch": 1.27, "learning_rate": 7.285994948077464e-06, "loss": 0.0883, "step": 27180 }, { "epoch": 1.27, "learning_rate": 7.283656095050987e-06, "loss": 0.0526, "step": 27185 }, { "epoch": 1.27, "learning_rate": 7.281317242024511e-06, "loss": 0.0566, "step": 27190 }, { "epoch": 1.27, "learning_rate": 7.278978388998036e-06, "loss": 0.0934, "step": 27195 }, { "epoch": 1.27, "learning_rate": 7.27663953597156e-06, "loss": 0.0264, "step": 27200 }, { "epoch": 1.27, "learning_rate": 7.274300682945084e-06, "loss": 0.0805, "step": 27205 }, { "epoch": 1.27, "learning_rate": 7.271961829918609e-06, "loss": 0.0883, "step": 27210 }, { "epoch": 1.27, "learning_rate": 7.2696229768921325e-06, "loss": 0.0809, "step": 27215 }, { "epoch": 1.27, "learning_rate": 7.267284123865657e-06, "loss": 0.07, "step": 27220 }, { "epoch": 1.27, "learning_rate": 7.264945270839181e-06, "loss": 0.0845, "step": 27225 }, { "epoch": 1.27, "learning_rate": 7.262606417812706e-06, "loss": 0.0537, "step": 27230 }, { "epoch": 1.27, "learning_rate": 7.26026756478623e-06, "loss": 0.1174, "step": 27235 }, { "epoch": 1.27, "learning_rate": 7.257928711759754e-06, "loss": 0.0891, "step": 27240 }, { "epoch": 1.27, "learning_rate": 7.255589858733277e-06, "loss": 0.0616, "step": 27245 }, { "epoch": 1.27, "learning_rate": 7.253251005706802e-06, "loss": 0.0309, "step": 27250 }, { "epoch": 1.27, "learning_rate": 7.250912152680326e-06, "loss": 0.0255, "step": 27255 }, { "epoch": 1.28, "learning_rate": 7.24857329965385e-06, "loss": 0.0342, "step": 27260 }, { "epoch": 1.28, "learning_rate": 7.246234446627375e-06, "loss": 0.0728, "step": 27265 }, { "epoch": 1.28, "learning_rate": 7.243895593600899e-06, "loss": 0.0711, "step": 27270 }, { "epoch": 1.28, "learning_rate": 7.241556740574423e-06, "loss": 0.0338, "step": 27275 }, { "epoch": 1.28, "learning_rate": 7.239217887547947e-06, "loss": 0.0466, "step": 27280 }, { "epoch": 1.28, "learning_rate": 7.236879034521471e-06, "loss": 0.0449, "step": 27285 }, { "epoch": 1.28, "learning_rate": 7.234540181494996e-06, "loss": 0.0582, "step": 27290 }, { "epoch": 1.28, "learning_rate": 7.23220132846852e-06, "loss": 0.0124, "step": 27295 }, { "epoch": 1.28, "learning_rate": 7.229862475442044e-06, "loss": 0.0358, "step": 27300 }, { "epoch": 1.28, "learning_rate": 7.227523622415567e-06, "loss": 0.0119, "step": 27305 }, { "epoch": 1.28, "learning_rate": 7.225184769389092e-06, "loss": 0.0744, "step": 27310 }, { "epoch": 1.28, "learning_rate": 7.222845916362616e-06, "loss": 0.113, "step": 27315 }, { "epoch": 1.28, "learning_rate": 7.22050706333614e-06, "loss": 0.0455, "step": 27320 }, { "epoch": 1.28, "learning_rate": 7.218168210309665e-06, "loss": 0.0895, "step": 27325 }, { "epoch": 1.28, "learning_rate": 7.2158293572831895e-06, "loss": 0.103, "step": 27330 }, { "epoch": 1.28, "learning_rate": 7.213490504256713e-06, "loss": 0.0979, "step": 27335 }, { "epoch": 1.28, "learning_rate": 7.211151651230237e-06, "loss": 0.0752, "step": 27340 }, { "epoch": 1.28, "learning_rate": 7.208812798203762e-06, "loss": 0.1092, "step": 27345 }, { "epoch": 1.28, "learning_rate": 7.206473945177286e-06, "loss": 0.0602, "step": 27350 }, { "epoch": 1.28, "learning_rate": 7.20413509215081e-06, "loss": 0.0729, "step": 27355 }, { "epoch": 1.28, "learning_rate": 7.201796239124335e-06, "loss": 0.0766, "step": 27360 }, { "epoch": 1.28, "learning_rate": 7.199457386097858e-06, "loss": 0.0367, "step": 27365 }, { "epoch": 1.28, "learning_rate": 7.197118533071382e-06, "loss": 0.0708, "step": 27370 }, { "epoch": 1.28, "learning_rate": 7.194779680044906e-06, "loss": 0.0326, "step": 27375 }, { "epoch": 1.28, "learning_rate": 7.1924408270184305e-06, "loss": 0.0606, "step": 27380 }, { "epoch": 1.28, "learning_rate": 7.1901019739919555e-06, "loss": 0.0495, "step": 27385 }, { "epoch": 1.28, "learning_rate": 7.187763120965479e-06, "loss": 0.1103, "step": 27390 }, { "epoch": 1.28, "learning_rate": 7.185424267939003e-06, "loss": 0.0548, "step": 27395 }, { "epoch": 1.28, "learning_rate": 7.183085414912527e-06, "loss": 0.0676, "step": 27400 }, { "epoch": 1.28, "learning_rate": 7.180746561886052e-06, "loss": 0.0398, "step": 27405 }, { "epoch": 1.28, "learning_rate": 7.178407708859576e-06, "loss": 0.0862, "step": 27410 }, { "epoch": 1.28, "learning_rate": 7.1760688558331e-06, "loss": 0.0454, "step": 27415 }, { "epoch": 1.28, "learning_rate": 7.173730002806623e-06, "loss": 0.036, "step": 27420 }, { "epoch": 1.28, "learning_rate": 7.171391149780148e-06, "loss": 0.0949, "step": 27425 }, { "epoch": 1.28, "learning_rate": 7.169052296753672e-06, "loss": 0.0677, "step": 27430 }, { "epoch": 1.28, "learning_rate": 7.1667134437271965e-06, "loss": 0.0343, "step": 27435 }, { "epoch": 1.28, "learning_rate": 7.1643745907007215e-06, "loss": 0.0545, "step": 27440 }, { "epoch": 1.28, "learning_rate": 7.1620357376742456e-06, "loss": 0.0604, "step": 27445 }, { "epoch": 1.28, "learning_rate": 7.159696884647769e-06, "loss": 0.0545, "step": 27450 }, { "epoch": 1.28, "learning_rate": 7.157358031621293e-06, "loss": 0.1586, "step": 27455 }, { "epoch": 1.28, "learning_rate": 7.155019178594818e-06, "loss": 0.1048, "step": 27460 }, { "epoch": 1.28, "learning_rate": 7.152680325568342e-06, "loss": 0.0461, "step": 27465 }, { "epoch": 1.28, "learning_rate": 7.150341472541866e-06, "loss": 0.028, "step": 27470 }, { "epoch": 1.29, "learning_rate": 7.148002619515391e-06, "loss": 0.0658, "step": 27475 }, { "epoch": 1.29, "learning_rate": 7.145663766488914e-06, "loss": 0.0914, "step": 27480 }, { "epoch": 1.29, "learning_rate": 7.143324913462438e-06, "loss": 0.0489, "step": 27485 }, { "epoch": 1.29, "learning_rate": 7.1409860604359625e-06, "loss": 0.0212, "step": 27490 }, { "epoch": 1.29, "learning_rate": 7.138647207409487e-06, "loss": 0.074, "step": 27495 }, { "epoch": 1.29, "learning_rate": 7.1363083543830116e-06, "loss": 0.0636, "step": 27500 }, { "epoch": 1.29, "learning_rate": 7.133969501356536e-06, "loss": 0.0401, "step": 27505 }, { "epoch": 1.29, "learning_rate": 7.131630648330059e-06, "loss": 0.0945, "step": 27510 }, { "epoch": 1.29, "learning_rate": 7.129291795303583e-06, "loss": 0.0471, "step": 27515 }, { "epoch": 1.29, "learning_rate": 7.126952942277108e-06, "loss": 0.0631, "step": 27520 }, { "epoch": 1.29, "learning_rate": 7.124614089250632e-06, "loss": 0.0905, "step": 27525 }, { "epoch": 1.29, "learning_rate": 7.122275236224156e-06, "loss": 0.0761, "step": 27530 }, { "epoch": 1.29, "learning_rate": 7.119936383197681e-06, "loss": 0.0443, "step": 27535 }, { "epoch": 1.29, "learning_rate": 7.117597530171204e-06, "loss": 0.0532, "step": 27540 }, { "epoch": 1.29, "learning_rate": 7.1152586771447285e-06, "loss": 0.0551, "step": 27545 }, { "epoch": 1.29, "learning_rate": 7.112919824118253e-06, "loss": 0.0216, "step": 27550 }, { "epoch": 1.29, "learning_rate": 7.1105809710917776e-06, "loss": 0.0722, "step": 27555 }, { "epoch": 1.29, "learning_rate": 7.108242118065302e-06, "loss": 0.0373, "step": 27560 }, { "epoch": 1.29, "learning_rate": 7.105903265038826e-06, "loss": 0.0381, "step": 27565 }, { "epoch": 1.29, "learning_rate": 7.103564412012349e-06, "loss": 0.0843, "step": 27570 }, { "epoch": 1.29, "learning_rate": 7.101225558985874e-06, "loss": 0.0921, "step": 27575 }, { "epoch": 1.29, "learning_rate": 7.098886705959398e-06, "loss": 0.083, "step": 27580 }, { "epoch": 1.29, "learning_rate": 7.096547852932922e-06, "loss": 0.089, "step": 27585 }, { "epoch": 1.29, "learning_rate": 7.094208999906447e-06, "loss": 0.0621, "step": 27590 }, { "epoch": 1.29, "learning_rate": 7.091870146879971e-06, "loss": 0.0824, "step": 27595 }, { "epoch": 1.29, "learning_rate": 7.0895312938534945e-06, "loss": 0.0893, "step": 27600 }, { "epoch": 1.29, "learning_rate": 7.087192440827019e-06, "loss": 0.0989, "step": 27605 }, { "epoch": 1.29, "learning_rate": 7.084853587800543e-06, "loss": 0.059, "step": 27610 }, { "epoch": 1.29, "learning_rate": 7.082514734774068e-06, "loss": 0.0341, "step": 27615 }, { "epoch": 1.29, "learning_rate": 7.080175881747592e-06, "loss": 0.0492, "step": 27620 }, { "epoch": 1.29, "learning_rate": 7.077837028721116e-06, "loss": 0.0355, "step": 27625 }, { "epoch": 1.29, "learning_rate": 7.075498175694639e-06, "loss": 0.0761, "step": 27630 }, { "epoch": 1.29, "learning_rate": 7.073159322668164e-06, "loss": 0.0412, "step": 27635 }, { "epoch": 1.29, "learning_rate": 7.070820469641688e-06, "loss": 0.0126, "step": 27640 }, { "epoch": 1.29, "learning_rate": 7.068481616615212e-06, "loss": 0.0424, "step": 27645 }, { "epoch": 1.29, "learning_rate": 7.066142763588737e-06, "loss": 0.089, "step": 27650 }, { "epoch": 1.29, "learning_rate": 7.063803910562261e-06, "loss": 0.0338, "step": 27655 }, { "epoch": 1.29, "learning_rate": 7.061465057535785e-06, "loss": 0.0644, "step": 27660 }, { "epoch": 1.29, "learning_rate": 7.059126204509309e-06, "loss": 0.0752, "step": 27665 }, { "epoch": 1.29, "learning_rate": 7.056787351482834e-06, "loss": 0.0637, "step": 27670 }, { "epoch": 1.29, "learning_rate": 7.054448498456358e-06, "loss": 0.0668, "step": 27675 }, { "epoch": 1.29, "learning_rate": 7.052109645429882e-06, "loss": 0.0842, "step": 27680 }, { "epoch": 1.3, "learning_rate": 7.049770792403407e-06, "loss": 0.0737, "step": 27685 }, { "epoch": 1.3, "learning_rate": 7.04743193937693e-06, "loss": 0.0846, "step": 27690 }, { "epoch": 1.3, "learning_rate": 7.045093086350454e-06, "loss": 0.1457, "step": 27695 }, { "epoch": 1.3, "learning_rate": 7.042754233323978e-06, "loss": 0.1041, "step": 27700 }, { "epoch": 1.3, "learning_rate": 7.040415380297503e-06, "loss": 0.084, "step": 27705 }, { "epoch": 1.3, "learning_rate": 7.038076527271027e-06, "loss": 0.0162, "step": 27710 }, { "epoch": 1.3, "learning_rate": 7.0357376742445514e-06, "loss": 0.0383, "step": 27715 }, { "epoch": 1.3, "learning_rate": 7.033398821218075e-06, "loss": 0.0154, "step": 27720 }, { "epoch": 1.3, "learning_rate": 7.031059968191599e-06, "loss": 0.0841, "step": 27725 }, { "epoch": 1.3, "learning_rate": 7.028721115165124e-06, "loss": 0.0386, "step": 27730 }, { "epoch": 1.3, "learning_rate": 7.026382262138648e-06, "loss": 0.0635, "step": 27735 }, { "epoch": 1.3, "learning_rate": 7.024043409112172e-06, "loss": 0.1026, "step": 27740 }, { "epoch": 1.3, "learning_rate": 7.021704556085697e-06, "loss": 0.0755, "step": 27745 }, { "epoch": 1.3, "learning_rate": 7.01936570305922e-06, "loss": 0.0628, "step": 27750 }, { "epoch": 1.3, "learning_rate": 7.017026850032744e-06, "loss": 0.0211, "step": 27755 }, { "epoch": 1.3, "learning_rate": 7.014687997006268e-06, "loss": 0.0797, "step": 27760 }, { "epoch": 1.3, "learning_rate": 7.012349143979793e-06, "loss": 0.0548, "step": 27765 }, { "epoch": 1.3, "learning_rate": 7.0100102909533174e-06, "loss": 0.0448, "step": 27770 }, { "epoch": 1.3, "learning_rate": 7.0076714379268415e-06, "loss": 0.0675, "step": 27775 }, { "epoch": 1.3, "learning_rate": 7.005332584900365e-06, "loss": 0.0376, "step": 27780 }, { "epoch": 1.3, "learning_rate": 7.00299373187389e-06, "loss": 0.0658, "step": 27785 }, { "epoch": 1.3, "learning_rate": 7.000654878847414e-06, "loss": 0.0777, "step": 27790 }, { "epoch": 1.3, "learning_rate": 6.998316025820938e-06, "loss": 0.0672, "step": 27795 }, { "epoch": 1.3, "learning_rate": 6.995977172794463e-06, "loss": 0.0709, "step": 27800 }, { "epoch": 1.3, "learning_rate": 6.993638319767987e-06, "loss": 0.084, "step": 27805 }, { "epoch": 1.3, "learning_rate": 6.99129946674151e-06, "loss": 0.0438, "step": 27810 }, { "epoch": 1.3, "learning_rate": 6.988960613715034e-06, "loss": 0.0346, "step": 27815 }, { "epoch": 1.3, "learning_rate": 6.9866217606885585e-06, "loss": 0.1015, "step": 27820 }, { "epoch": 1.3, "learning_rate": 6.9842829076620834e-06, "loss": 0.1077, "step": 27825 }, { "epoch": 1.3, "learning_rate": 6.9819440546356075e-06, "loss": 0.0313, "step": 27830 }, { "epoch": 1.3, "learning_rate": 6.979605201609131e-06, "loss": 0.0675, "step": 27835 }, { "epoch": 1.3, "learning_rate": 6.977266348582655e-06, "loss": 0.0461, "step": 27840 }, { "epoch": 1.3, "learning_rate": 6.97492749555618e-06, "loss": 0.0457, "step": 27845 }, { "epoch": 1.3, "learning_rate": 6.972588642529704e-06, "loss": 0.0438, "step": 27850 }, { "epoch": 1.3, "learning_rate": 6.970249789503228e-06, "loss": 0.0612, "step": 27855 }, { "epoch": 1.3, "learning_rate": 6.967910936476753e-06, "loss": 0.0295, "step": 27860 }, { "epoch": 1.3, "learning_rate": 6.965572083450276e-06, "loss": 0.0563, "step": 27865 }, { "epoch": 1.3, "learning_rate": 6.9632332304238e-06, "loss": 0.0341, "step": 27870 }, { "epoch": 1.3, "learning_rate": 6.9608943773973245e-06, "loss": 0.0332, "step": 27875 }, { "epoch": 1.3, "learning_rate": 6.9585555243708494e-06, "loss": 0.057, "step": 27880 }, { "epoch": 1.3, "learning_rate": 6.9562166713443735e-06, "loss": 0.0458, "step": 27885 }, { "epoch": 1.3, "learning_rate": 6.953877818317898e-06, "loss": 0.1176, "step": 27890 }, { "epoch": 1.3, "learning_rate": 6.951538965291421e-06, "loss": 0.0795, "step": 27895 }, { "epoch": 1.31, "learning_rate": 6.949200112264946e-06, "loss": 0.0896, "step": 27900 }, { "epoch": 1.31, "learning_rate": 6.94686125923847e-06, "loss": 0.0881, "step": 27905 }, { "epoch": 1.31, "learning_rate": 6.944522406211994e-06, "loss": 0.0209, "step": 27910 }, { "epoch": 1.31, "learning_rate": 6.942183553185519e-06, "loss": 0.025, "step": 27915 }, { "epoch": 1.31, "learning_rate": 6.939844700159043e-06, "loss": 0.0504, "step": 27920 }, { "epoch": 1.31, "learning_rate": 6.937505847132566e-06, "loss": 0.0297, "step": 27925 }, { "epoch": 1.31, "learning_rate": 6.9351669941060905e-06, "loss": 0.0722, "step": 27930 }, { "epoch": 1.31, "learning_rate": 6.932828141079615e-06, "loss": 0.0591, "step": 27935 }, { "epoch": 1.31, "learning_rate": 6.9304892880531395e-06, "loss": 0.0284, "step": 27940 }, { "epoch": 1.31, "learning_rate": 6.928150435026664e-06, "loss": 0.0696, "step": 27945 }, { "epoch": 1.31, "learning_rate": 6.925811582000188e-06, "loss": 0.0725, "step": 27950 }, { "epoch": 1.31, "learning_rate": 6.923472728973711e-06, "loss": 0.0424, "step": 27955 }, { "epoch": 1.31, "learning_rate": 6.921133875947236e-06, "loss": 0.0808, "step": 27960 }, { "epoch": 1.31, "learning_rate": 6.91879502292076e-06, "loss": 0.0715, "step": 27965 }, { "epoch": 1.31, "learning_rate": 6.916456169894284e-06, "loss": 0.107, "step": 27970 }, { "epoch": 1.31, "learning_rate": 6.914117316867809e-06, "loss": 0.0434, "step": 27975 }, { "epoch": 1.31, "learning_rate": 6.911778463841333e-06, "loss": 0.1104, "step": 27980 }, { "epoch": 1.31, "learning_rate": 6.9094396108148565e-06, "loss": 0.0581, "step": 27985 }, { "epoch": 1.31, "learning_rate": 6.907100757788381e-06, "loss": 0.0344, "step": 27990 }, { "epoch": 1.31, "learning_rate": 6.9047619047619055e-06, "loss": 0.071, "step": 27995 }, { "epoch": 1.31, "learning_rate": 6.90242305173543e-06, "loss": 0.065, "step": 28000 }, { "epoch": 1.31, "learning_rate": 6.900084198708954e-06, "loss": 0.0796, "step": 28005 }, { "epoch": 1.31, "learning_rate": 6.897745345682479e-06, "loss": 0.0497, "step": 28010 }, { "epoch": 1.31, "learning_rate": 6.895406492656002e-06, "loss": 0.0539, "step": 28015 }, { "epoch": 1.31, "learning_rate": 6.893067639629526e-06, "loss": 0.1042, "step": 28020 }, { "epoch": 1.31, "learning_rate": 6.89072878660305e-06, "loss": 0.0417, "step": 28025 }, { "epoch": 1.31, "learning_rate": 6.888389933576575e-06, "loss": 0.0586, "step": 28030 }, { "epoch": 1.31, "learning_rate": 6.886051080550099e-06, "loss": 0.0693, "step": 28035 }, { "epoch": 1.31, "learning_rate": 6.883712227523623e-06, "loss": 0.0501, "step": 28040 }, { "epoch": 1.31, "learning_rate": 6.8813733744971466e-06, "loss": 0.0385, "step": 28045 }, { "epoch": 1.31, "learning_rate": 6.879034521470671e-06, "loss": 0.0805, "step": 28050 }, { "epoch": 1.31, "learning_rate": 6.876695668444196e-06, "loss": 0.0681, "step": 28055 }, { "epoch": 1.31, "learning_rate": 6.87435681541772e-06, "loss": 0.0152, "step": 28060 }, { "epoch": 1.31, "learning_rate": 6.872017962391244e-06, "loss": 0.0239, "step": 28065 }, { "epoch": 1.31, "learning_rate": 6.869679109364769e-06, "loss": 0.0853, "step": 28070 }, { "epoch": 1.31, "learning_rate": 6.867340256338292e-06, "loss": 0.0484, "step": 28075 }, { "epoch": 1.31, "learning_rate": 6.865001403311816e-06, "loss": 0.0364, "step": 28080 }, { "epoch": 1.31, "learning_rate": 6.86266255028534e-06, "loss": 0.0437, "step": 28085 }, { "epoch": 1.31, "learning_rate": 6.860323697258865e-06, "loss": 0.0474, "step": 28090 }, { "epoch": 1.31, "learning_rate": 6.857984844232389e-06, "loss": 0.0916, "step": 28095 }, { "epoch": 1.31, "learning_rate": 6.855645991205913e-06, "loss": 0.0806, "step": 28100 }, { "epoch": 1.31, "learning_rate": 6.853307138179437e-06, "loss": 0.0819, "step": 28105 }, { "epoch": 1.31, "learning_rate": 6.850968285152962e-06, "loss": 0.0798, "step": 28110 }, { "epoch": 1.32, "learning_rate": 6.848629432126486e-06, "loss": 0.0124, "step": 28115 }, { "epoch": 1.32, "learning_rate": 6.84629057910001e-06, "loss": 0.0429, "step": 28120 }, { "epoch": 1.32, "learning_rate": 6.843951726073535e-06, "loss": 0.0621, "step": 28125 }, { "epoch": 1.32, "learning_rate": 6.841612873047059e-06, "loss": 0.0501, "step": 28130 }, { "epoch": 1.32, "learning_rate": 6.839274020020582e-06, "loss": 0.0438, "step": 28135 }, { "epoch": 1.32, "learning_rate": 6.836935166994106e-06, "loss": 0.0182, "step": 28140 }, { "epoch": 1.32, "learning_rate": 6.834596313967631e-06, "loss": 0.0522, "step": 28145 }, { "epoch": 1.32, "learning_rate": 6.832257460941155e-06, "loss": 0.0978, "step": 28150 }, { "epoch": 1.32, "learning_rate": 6.829918607914679e-06, "loss": 0.081, "step": 28155 }, { "epoch": 1.32, "learning_rate": 6.8275797548882035e-06, "loss": 0.08, "step": 28160 }, { "epoch": 1.32, "learning_rate": 6.825240901861727e-06, "loss": 0.1132, "step": 28165 }, { "epoch": 1.32, "learning_rate": 6.822902048835252e-06, "loss": 0.0717, "step": 28170 }, { "epoch": 1.32, "learning_rate": 6.820563195808776e-06, "loss": 0.0522, "step": 28175 }, { "epoch": 1.32, "learning_rate": 6.8182243427823e-06, "loss": 0.0646, "step": 28180 }, { "epoch": 1.32, "learning_rate": 6.815885489755825e-06, "loss": 0.0197, "step": 28185 }, { "epoch": 1.32, "learning_rate": 6.813546636729349e-06, "loss": 0.0445, "step": 28190 }, { "epoch": 1.32, "learning_rate": 6.811207783702872e-06, "loss": 0.0775, "step": 28195 }, { "epoch": 1.32, "learning_rate": 6.808868930676396e-06, "loss": 0.0835, "step": 28200 }, { "epoch": 1.32, "learning_rate": 6.806530077649921e-06, "loss": 0.0709, "step": 28205 }, { "epoch": 1.32, "learning_rate": 6.804191224623445e-06, "loss": 0.0655, "step": 28210 }, { "epoch": 1.32, "learning_rate": 6.8018523715969695e-06, "loss": 0.0443, "step": 28215 }, { "epoch": 1.32, "learning_rate": 6.7995135185704945e-06, "loss": 0.0449, "step": 28220 }, { "epoch": 1.32, "learning_rate": 6.797174665544018e-06, "loss": 0.0458, "step": 28225 }, { "epoch": 1.32, "learning_rate": 6.794835812517542e-06, "loss": 0.0567, "step": 28230 }, { "epoch": 1.32, "learning_rate": 6.792496959491066e-06, "loss": 0.041, "step": 28235 }, { "epoch": 1.32, "learning_rate": 6.790158106464591e-06, "loss": 0.0489, "step": 28240 }, { "epoch": 1.32, "learning_rate": 6.787819253438115e-06, "loss": 0.058, "step": 28245 }, { "epoch": 1.32, "learning_rate": 6.785480400411638e-06, "loss": 0.0515, "step": 28250 }, { "epoch": 1.32, "learning_rate": 6.783141547385162e-06, "loss": 0.0856, "step": 28255 }, { "epoch": 1.32, "learning_rate": 6.780802694358687e-06, "loss": 0.1197, "step": 28260 }, { "epoch": 1.32, "learning_rate": 6.778463841332211e-06, "loss": 0.035, "step": 28265 }, { "epoch": 1.32, "learning_rate": 6.7761249883057355e-06, "loss": 0.0711, "step": 28270 }, { "epoch": 1.32, "learning_rate": 6.77378613527926e-06, "loss": 0.0425, "step": 28275 }, { "epoch": 1.32, "learning_rate": 6.771447282252783e-06, "loss": 0.0647, "step": 28280 }, { "epoch": 1.32, "learning_rate": 6.769108429226308e-06, "loss": 0.0706, "step": 28285 }, { "epoch": 1.32, "learning_rate": 6.766769576199832e-06, "loss": 0.0858, "step": 28290 }, { "epoch": 1.32, "learning_rate": 6.764430723173356e-06, "loss": 0.139, "step": 28295 }, { "epoch": 1.32, "learning_rate": 6.762091870146881e-06, "loss": 0.1014, "step": 28300 }, { "epoch": 1.32, "learning_rate": 6.759753017120405e-06, "loss": 0.0237, "step": 28305 }, { "epoch": 1.32, "learning_rate": 6.757414164093928e-06, "loss": 0.0597, "step": 28310 }, { "epoch": 1.32, "learning_rate": 6.7550753110674524e-06, "loss": 0.0775, "step": 28315 }, { "epoch": 1.32, "learning_rate": 6.752736458040977e-06, "loss": 0.0503, "step": 28320 }, { "epoch": 1.32, "learning_rate": 6.7503976050145015e-06, "loss": 0.0486, "step": 28325 }, { "epoch": 1.33, "learning_rate": 6.748058751988026e-06, "loss": 0.059, "step": 28330 }, { "epoch": 1.33, "learning_rate": 6.7457198989615506e-06, "loss": 0.0627, "step": 28335 }, { "epoch": 1.33, "learning_rate": 6.743381045935074e-06, "loss": 0.0527, "step": 28340 }, { "epoch": 1.33, "learning_rate": 6.741042192908598e-06, "loss": 0.0351, "step": 28345 }, { "epoch": 1.33, "learning_rate": 6.738703339882122e-06, "loss": 0.1009, "step": 28350 }, { "epoch": 1.33, "learning_rate": 6.736364486855647e-06, "loss": 0.0332, "step": 28355 }, { "epoch": 1.33, "learning_rate": 6.734025633829171e-06, "loss": 0.0926, "step": 28360 }, { "epoch": 1.33, "learning_rate": 6.731686780802695e-06, "loss": 0.0576, "step": 28365 }, { "epoch": 1.33, "learning_rate": 6.7293479277762184e-06, "loss": 0.0933, "step": 28370 }, { "epoch": 1.33, "learning_rate": 6.7270090747497425e-06, "loss": 0.0688, "step": 28375 }, { "epoch": 1.33, "learning_rate": 6.7246702217232675e-06, "loss": 0.0261, "step": 28380 }, { "epoch": 1.33, "learning_rate": 6.722331368696792e-06, "loss": 0.0299, "step": 28385 }, { "epoch": 1.33, "learning_rate": 6.719992515670316e-06, "loss": 0.0711, "step": 28390 }, { "epoch": 1.33, "learning_rate": 6.717653662643841e-06, "loss": 0.029, "step": 28395 }, { "epoch": 1.33, "learning_rate": 6.715314809617364e-06, "loss": 0.0885, "step": 28400 }, { "epoch": 1.33, "learning_rate": 6.712975956590888e-06, "loss": 0.0758, "step": 28405 }, { "epoch": 1.33, "learning_rate": 6.710637103564412e-06, "loss": 0.0703, "step": 28410 }, { "epoch": 1.33, "learning_rate": 6.708298250537937e-06, "loss": 0.0511, "step": 28415 }, { "epoch": 1.33, "learning_rate": 6.705959397511461e-06, "loss": 0.0416, "step": 28420 }, { "epoch": 1.33, "learning_rate": 6.703620544484985e-06, "loss": 0.0871, "step": 28425 }, { "epoch": 1.33, "learning_rate": 6.7012816914585085e-06, "loss": 0.0807, "step": 28430 }, { "epoch": 1.33, "learning_rate": 6.6989428384320335e-06, "loss": 0.0704, "step": 28435 }, { "epoch": 1.33, "learning_rate": 6.696603985405558e-06, "loss": 0.0685, "step": 28440 }, { "epoch": 1.33, "learning_rate": 6.694265132379082e-06, "loss": 0.0721, "step": 28445 }, { "epoch": 1.33, "learning_rate": 6.691926279352607e-06, "loss": 0.074, "step": 28450 }, { "epoch": 1.33, "learning_rate": 6.689587426326131e-06, "loss": 0.0855, "step": 28455 }, { "epoch": 1.33, "learning_rate": 6.687248573299654e-06, "loss": 0.0491, "step": 28460 }, { "epoch": 1.33, "learning_rate": 6.684909720273178e-06, "loss": 0.0405, "step": 28465 }, { "epoch": 1.33, "learning_rate": 6.682570867246703e-06, "loss": 0.0473, "step": 28470 }, { "epoch": 1.33, "learning_rate": 6.680232014220227e-06, "loss": 0.0721, "step": 28475 }, { "epoch": 1.33, "learning_rate": 6.677893161193751e-06, "loss": 0.0865, "step": 28480 }, { "epoch": 1.33, "learning_rate": 6.675554308167275e-06, "loss": 0.0471, "step": 28485 }, { "epoch": 1.33, "learning_rate": 6.673215455140799e-06, "loss": 0.023, "step": 28490 }, { "epoch": 1.33, "learning_rate": 6.670876602114324e-06, "loss": 0.0826, "step": 28495 }, { "epoch": 1.33, "learning_rate": 6.668537749087848e-06, "loss": 0.0527, "step": 28500 }, { "epoch": 1.33, "learning_rate": 6.666198896061372e-06, "loss": 0.0799, "step": 28505 }, { "epoch": 1.33, "learning_rate": 6.663860043034897e-06, "loss": 0.0861, "step": 28510 }, { "epoch": 1.33, "learning_rate": 6.661521190008421e-06, "loss": 0.0465, "step": 28515 }, { "epoch": 1.33, "learning_rate": 6.659182336981944e-06, "loss": 0.087, "step": 28520 }, { "epoch": 1.33, "learning_rate": 6.656843483955468e-06, "loss": 0.0861, "step": 28525 }, { "epoch": 1.33, "learning_rate": 6.654504630928993e-06, "loss": 0.062, "step": 28530 }, { "epoch": 1.33, "learning_rate": 6.652165777902517e-06, "loss": 0.0364, "step": 28535 }, { "epoch": 1.34, "learning_rate": 6.649826924876041e-06, "loss": 0.0424, "step": 28540 }, { "epoch": 1.34, "learning_rate": 6.647488071849566e-06, "loss": 0.0689, "step": 28545 }, { "epoch": 1.34, "learning_rate": 6.64514921882309e-06, "loss": 0.0631, "step": 28550 }, { "epoch": 1.34, "learning_rate": 6.642810365796614e-06, "loss": 0.0687, "step": 28555 }, { "epoch": 1.34, "learning_rate": 6.640471512770138e-06, "loss": 0.0522, "step": 28560 }, { "epoch": 1.34, "learning_rate": 6.638132659743663e-06, "loss": 0.01, "step": 28565 }, { "epoch": 1.34, "learning_rate": 6.635793806717187e-06, "loss": 0.064, "step": 28570 }, { "epoch": 1.34, "learning_rate": 6.633454953690711e-06, "loss": 0.0531, "step": 28575 }, { "epoch": 1.34, "learning_rate": 6.631116100664234e-06, "loss": 0.0902, "step": 28580 }, { "epoch": 1.34, "learning_rate": 6.628777247637759e-06, "loss": 0.0307, "step": 28585 }, { "epoch": 1.34, "learning_rate": 6.626438394611283e-06, "loss": 0.0261, "step": 28590 }, { "epoch": 1.34, "learning_rate": 6.624099541584807e-06, "loss": 0.037, "step": 28595 }, { "epoch": 1.34, "learning_rate": 6.6217606885583315e-06, "loss": 0.1058, "step": 28600 }, { "epoch": 1.34, "learning_rate": 6.6194218355318564e-06, "loss": 0.0625, "step": 28605 }, { "epoch": 1.34, "learning_rate": 6.61708298250538e-06, "loss": 0.0465, "step": 28610 }, { "epoch": 1.34, "learning_rate": 6.614744129478904e-06, "loss": 0.0482, "step": 28615 }, { "epoch": 1.34, "learning_rate": 6.612405276452428e-06, "loss": 0.0609, "step": 28620 }, { "epoch": 1.34, "learning_rate": 6.610066423425953e-06, "loss": 0.0679, "step": 28625 }, { "epoch": 1.34, "learning_rate": 6.607727570399477e-06, "loss": 0.0528, "step": 28630 }, { "epoch": 1.34, "learning_rate": 6.605388717373001e-06, "loss": 0.0926, "step": 28635 }, { "epoch": 1.34, "learning_rate": 6.603049864346524e-06, "loss": 0.0992, "step": 28640 }, { "epoch": 1.34, "learning_rate": 6.600711011320049e-06, "loss": 0.0542, "step": 28645 }, { "epoch": 1.34, "learning_rate": 6.598372158293573e-06, "loss": 0.1137, "step": 28650 }, { "epoch": 1.34, "learning_rate": 6.5960333052670975e-06, "loss": 0.0689, "step": 28655 }, { "epoch": 1.34, "learning_rate": 6.5936944522406224e-06, "loss": 0.0552, "step": 28660 }, { "epoch": 1.34, "learning_rate": 6.5913555992141465e-06, "loss": 0.0575, "step": 28665 }, { "epoch": 1.34, "learning_rate": 6.58901674618767e-06, "loss": 0.0646, "step": 28670 }, { "epoch": 1.34, "learning_rate": 6.586677893161194e-06, "loss": 0.0564, "step": 28675 }, { "epoch": 1.34, "learning_rate": 6.584339040134719e-06, "loss": 0.0428, "step": 28680 }, { "epoch": 1.34, "learning_rate": 6.582000187108243e-06, "loss": 0.0255, "step": 28685 }, { "epoch": 1.34, "learning_rate": 6.579661334081767e-06, "loss": 0.0577, "step": 28690 }, { "epoch": 1.34, "learning_rate": 6.57732248105529e-06, "loss": 0.0334, "step": 28695 }, { "epoch": 1.34, "learning_rate": 6.574983628028815e-06, "loss": 0.0268, "step": 28700 }, { "epoch": 1.34, "learning_rate": 6.572644775002339e-06, "loss": 0.0749, "step": 28705 }, { "epoch": 1.34, "learning_rate": 6.5703059219758635e-06, "loss": 0.0376, "step": 28710 }, { "epoch": 1.34, "learning_rate": 6.567967068949388e-06, "loss": 0.0818, "step": 28715 }, { "epoch": 1.34, "learning_rate": 6.5656282159229125e-06, "loss": 0.1419, "step": 28720 }, { "epoch": 1.34, "learning_rate": 6.563289362896436e-06, "loss": 0.0342, "step": 28725 }, { "epoch": 1.34, "learning_rate": 6.56095050986996e-06, "loss": 0.1045, "step": 28730 }, { "epoch": 1.34, "learning_rate": 6.558611656843484e-06, "loss": 0.074, "step": 28735 }, { "epoch": 1.34, "learning_rate": 6.556272803817009e-06, "loss": 0.0457, "step": 28740 }, { "epoch": 1.34, "learning_rate": 6.553933950790533e-06, "loss": 0.0454, "step": 28745 }, { "epoch": 1.34, "learning_rate": 6.551595097764057e-06, "loss": 0.0193, "step": 28750 }, { "epoch": 1.35, "learning_rate": 6.54925624473758e-06, "loss": 0.0966, "step": 28755 }, { "epoch": 1.35, "learning_rate": 6.546917391711105e-06, "loss": 0.0737, "step": 28760 }, { "epoch": 1.35, "learning_rate": 6.5445785386846295e-06, "loss": 0.0756, "step": 28765 }, { "epoch": 1.35, "learning_rate": 6.542239685658154e-06, "loss": 0.0209, "step": 28770 }, { "epoch": 1.35, "learning_rate": 6.5399008326316785e-06, "loss": 0.0266, "step": 28775 }, { "epoch": 1.35, "learning_rate": 6.537561979605203e-06, "loss": 0.0205, "step": 28780 }, { "epoch": 1.35, "learning_rate": 6.535223126578726e-06, "loss": 0.0458, "step": 28785 }, { "epoch": 1.35, "learning_rate": 6.53288427355225e-06, "loss": 0.0705, "step": 28790 }, { "epoch": 1.35, "learning_rate": 6.530545420525775e-06, "loss": 0.0536, "step": 28795 }, { "epoch": 1.35, "learning_rate": 6.528206567499299e-06, "loss": 0.0306, "step": 28800 }, { "epoch": 1.35, "learning_rate": 6.525867714472823e-06, "loss": 0.0641, "step": 28805 }, { "epoch": 1.35, "learning_rate": 6.523528861446347e-06, "loss": 0.06, "step": 28810 }, { "epoch": 1.35, "learning_rate": 6.5211900084198705e-06, "loss": 0.0463, "step": 28815 }, { "epoch": 1.35, "learning_rate": 6.5188511553933955e-06, "loss": 0.0704, "step": 28820 }, { "epoch": 1.35, "learning_rate": 6.5165123023669196e-06, "loss": 0.0269, "step": 28825 }, { "epoch": 1.35, "learning_rate": 6.514173449340444e-06, "loss": 0.0424, "step": 28830 }, { "epoch": 1.35, "learning_rate": 6.511834596313969e-06, "loss": 0.0973, "step": 28835 }, { "epoch": 1.35, "learning_rate": 6.509495743287493e-06, "loss": 0.0349, "step": 28840 }, { "epoch": 1.35, "learning_rate": 6.507156890261016e-06, "loss": 0.0241, "step": 28845 }, { "epoch": 1.35, "learning_rate": 6.50481803723454e-06, "loss": 0.0747, "step": 28850 }, { "epoch": 1.35, "learning_rate": 6.502479184208065e-06, "loss": 0.0238, "step": 28855 }, { "epoch": 1.35, "learning_rate": 6.500140331181589e-06, "loss": 0.0192, "step": 28860 }, { "epoch": 1.35, "learning_rate": 6.497801478155113e-06, "loss": 0.0551, "step": 28865 }, { "epoch": 1.35, "learning_rate": 6.495462625128638e-06, "loss": 0.0611, "step": 28870 }, { "epoch": 1.35, "learning_rate": 6.4931237721021615e-06, "loss": 0.0336, "step": 28875 }, { "epoch": 1.35, "learning_rate": 6.4907849190756856e-06, "loss": 0.0529, "step": 28880 }, { "epoch": 1.35, "learning_rate": 6.48844606604921e-06, "loss": 0.0623, "step": 28885 }, { "epoch": 1.35, "learning_rate": 6.486107213022735e-06, "loss": 0.0711, "step": 28890 }, { "epoch": 1.35, "learning_rate": 6.483768359996259e-06, "loss": 0.0925, "step": 28895 }, { "epoch": 1.35, "learning_rate": 6.481429506969783e-06, "loss": 0.0614, "step": 28900 }, { "epoch": 1.35, "learning_rate": 6.479090653943306e-06, "loss": 0.1142, "step": 28905 }, { "epoch": 1.35, "learning_rate": 6.476751800916831e-06, "loss": 0.0357, "step": 28910 }, { "epoch": 1.35, "learning_rate": 6.474412947890355e-06, "loss": 0.1033, "step": 28915 }, { "epoch": 1.35, "learning_rate": 6.472074094863879e-06, "loss": 0.0368, "step": 28920 }, { "epoch": 1.35, "learning_rate": 6.469735241837403e-06, "loss": 0.0508, "step": 28925 }, { "epoch": 1.35, "learning_rate": 6.467396388810928e-06, "loss": 0.0339, "step": 28930 }, { "epoch": 1.35, "learning_rate": 6.4650575357844516e-06, "loss": 0.0519, "step": 28935 }, { "epoch": 1.35, "learning_rate": 6.462718682757976e-06, "loss": 0.0153, "step": 28940 }, { "epoch": 1.35, "learning_rate": 6.4603798297315e-06, "loss": 0.0413, "step": 28945 }, { "epoch": 1.35, "learning_rate": 6.458040976705025e-06, "loss": 0.0884, "step": 28950 }, { "epoch": 1.35, "learning_rate": 6.455702123678549e-06, "loss": 0.0465, "step": 28955 }, { "epoch": 1.35, "learning_rate": 6.453363270652073e-06, "loss": 0.0757, "step": 28960 }, { "epoch": 1.35, "learning_rate": 6.451024417625596e-06, "loss": 0.0356, "step": 28965 }, { "epoch": 1.36, "learning_rate": 6.448685564599121e-06, "loss": 0.0459, "step": 28970 }, { "epoch": 1.36, "learning_rate": 6.446346711572645e-06, "loss": 0.0887, "step": 28975 }, { "epoch": 1.36, "learning_rate": 6.444007858546169e-06, "loss": 0.0307, "step": 28980 }, { "epoch": 1.36, "learning_rate": 6.441669005519694e-06, "loss": 0.0877, "step": 28985 }, { "epoch": 1.36, "learning_rate": 6.439330152493218e-06, "loss": 0.1154, "step": 28990 }, { "epoch": 1.36, "learning_rate": 6.436991299466742e-06, "loss": 0.0546, "step": 28995 }, { "epoch": 1.36, "learning_rate": 6.434652446440266e-06, "loss": 0.0939, "step": 29000 }, { "epoch": 1.36, "learning_rate": 6.432313593413791e-06, "loss": 0.035, "step": 29005 }, { "epoch": 1.36, "learning_rate": 6.429974740387315e-06, "loss": 0.0281, "step": 29010 }, { "epoch": 1.36, "learning_rate": 6.427635887360839e-06, "loss": 0.0365, "step": 29015 }, { "epoch": 1.36, "learning_rate": 6.425297034334364e-06, "loss": 0.0428, "step": 29020 }, { "epoch": 1.36, "learning_rate": 6.422958181307887e-06, "loss": 0.0899, "step": 29025 }, { "epoch": 1.36, "learning_rate": 6.420619328281411e-06, "loss": 0.0295, "step": 29030 }, { "epoch": 1.36, "learning_rate": 6.418280475254935e-06, "loss": 0.03, "step": 29035 }, { "epoch": 1.36, "learning_rate": 6.4159416222284595e-06, "loss": 0.0799, "step": 29040 }, { "epoch": 1.36, "learning_rate": 6.413602769201984e-06, "loss": 0.0909, "step": 29045 }, { "epoch": 1.36, "learning_rate": 6.4112639161755085e-06, "loss": 0.0929, "step": 29050 }, { "epoch": 1.36, "learning_rate": 6.408925063149032e-06, "loss": 0.0443, "step": 29055 }, { "epoch": 1.36, "learning_rate": 6.406586210122556e-06, "loss": 0.0285, "step": 29060 }, { "epoch": 1.36, "learning_rate": 6.404247357096081e-06, "loss": 0.0756, "step": 29065 }, { "epoch": 1.36, "learning_rate": 6.401908504069605e-06, "loss": 0.0241, "step": 29070 }, { "epoch": 1.36, "learning_rate": 6.399569651043129e-06, "loss": 0.0987, "step": 29075 }, { "epoch": 1.36, "learning_rate": 6.397230798016654e-06, "loss": 0.0573, "step": 29080 }, { "epoch": 1.36, "learning_rate": 6.394891944990177e-06, "loss": 0.0146, "step": 29085 }, { "epoch": 1.36, "learning_rate": 6.392553091963701e-06, "loss": 0.0687, "step": 29090 }, { "epoch": 1.36, "learning_rate": 6.3902142389372255e-06, "loss": 0.0605, "step": 29095 }, { "epoch": 1.36, "learning_rate": 6.38787538591075e-06, "loss": 0.0408, "step": 29100 }, { "epoch": 1.36, "learning_rate": 6.3855365328842745e-06, "loss": 0.0389, "step": 29105 }, { "epoch": 1.36, "learning_rate": 6.383197679857799e-06, "loss": 0.085, "step": 29110 }, { "epoch": 1.36, "learning_rate": 6.380858826831322e-06, "loss": 0.0584, "step": 29115 }, { "epoch": 1.36, "learning_rate": 6.378519973804847e-06, "loss": 0.0358, "step": 29120 }, { "epoch": 1.36, "learning_rate": 6.376181120778371e-06, "loss": 0.1079, "step": 29125 }, { "epoch": 1.36, "learning_rate": 6.373842267751895e-06, "loss": 0.0351, "step": 29130 }, { "epoch": 1.36, "learning_rate": 6.371503414725419e-06, "loss": 0.0232, "step": 29135 }, { "epoch": 1.36, "learning_rate": 6.369164561698943e-06, "loss": 0.0356, "step": 29140 }, { "epoch": 1.36, "learning_rate": 6.366825708672467e-06, "loss": 0.0755, "step": 29145 }, { "epoch": 1.36, "learning_rate": 6.3644868556459914e-06, "loss": 0.049, "step": 29150 }, { "epoch": 1.36, "learning_rate": 6.3621480026195156e-06, "loss": 0.052, "step": 29155 }, { "epoch": 1.36, "learning_rate": 6.3598091495930405e-06, "loss": 0.0111, "step": 29160 }, { "epoch": 1.36, "learning_rate": 6.357470296566565e-06, "loss": 0.0545, "step": 29165 }, { "epoch": 1.36, "learning_rate": 6.355131443540088e-06, "loss": 0.0324, "step": 29170 }, { "epoch": 1.36, "learning_rate": 6.352792590513612e-06, "loss": 0.0889, "step": 29175 }, { "epoch": 1.36, "learning_rate": 6.350453737487137e-06, "loss": 0.0308, "step": 29180 }, { "epoch": 1.37, "learning_rate": 6.348114884460661e-06, "loss": 0.1047, "step": 29185 }, { "epoch": 1.37, "learning_rate": 6.345776031434185e-06, "loss": 0.0879, "step": 29190 }, { "epoch": 1.37, "learning_rate": 6.34343717840771e-06, "loss": 0.0571, "step": 29195 }, { "epoch": 1.37, "learning_rate": 6.341098325381233e-06, "loss": 0.0368, "step": 29200 }, { "epoch": 1.37, "learning_rate": 6.3387594723547574e-06, "loss": 0.0375, "step": 29205 }, { "epoch": 1.37, "learning_rate": 6.3364206193282815e-06, "loss": 0.0649, "step": 29210 }, { "epoch": 1.37, "learning_rate": 6.3340817663018065e-06, "loss": 0.1517, "step": 29215 }, { "epoch": 1.37, "learning_rate": 6.331742913275331e-06, "loss": 0.0191, "step": 29220 }, { "epoch": 1.37, "learning_rate": 6.329404060248855e-06, "loss": 0.0752, "step": 29225 }, { "epoch": 1.37, "learning_rate": 6.327065207222378e-06, "loss": 0.029, "step": 29230 }, { "epoch": 1.37, "learning_rate": 6.324726354195903e-06, "loss": 0.0523, "step": 29235 }, { "epoch": 1.37, "learning_rate": 6.322387501169427e-06, "loss": 0.0716, "step": 29240 }, { "epoch": 1.37, "learning_rate": 6.320048648142951e-06, "loss": 0.0839, "step": 29245 }, { "epoch": 1.37, "learning_rate": 6.317709795116475e-06, "loss": 0.0633, "step": 29250 }, { "epoch": 1.37, "learning_rate": 6.31537094209e-06, "loss": 0.0557, "step": 29255 }, { "epoch": 1.37, "learning_rate": 6.3130320890635234e-06, "loss": 0.0579, "step": 29260 }, { "epoch": 1.37, "learning_rate": 6.3106932360370475e-06, "loss": 0.0474, "step": 29265 }, { "epoch": 1.37, "learning_rate": 6.308354383010572e-06, "loss": 0.1019, "step": 29270 }, { "epoch": 1.37, "learning_rate": 6.306015529984097e-06, "loss": 0.0644, "step": 29275 }, { "epoch": 1.37, "learning_rate": 6.303676676957621e-06, "loss": 0.0909, "step": 29280 }, { "epoch": 1.37, "learning_rate": 6.301337823931145e-06, "loss": 0.0951, "step": 29285 }, { "epoch": 1.37, "learning_rate": 6.298998970904668e-06, "loss": 0.0586, "step": 29290 }, { "epoch": 1.37, "learning_rate": 6.296660117878193e-06, "loss": 0.1213, "step": 29295 }, { "epoch": 1.37, "learning_rate": 6.294321264851717e-06, "loss": 0.085, "step": 29300 }, { "epoch": 1.37, "learning_rate": 6.291982411825241e-06, "loss": 0.016, "step": 29305 }, { "epoch": 1.37, "learning_rate": 6.289643558798766e-06, "loss": 0.0534, "step": 29310 }, { "epoch": 1.37, "learning_rate": 6.28730470577229e-06, "loss": 0.0842, "step": 29315 }, { "epoch": 1.37, "learning_rate": 6.2849658527458135e-06, "loss": 0.0931, "step": 29320 }, { "epoch": 1.37, "learning_rate": 6.282626999719338e-06, "loss": 0.0097, "step": 29325 }, { "epoch": 1.37, "learning_rate": 6.280288146692863e-06, "loss": 0.0181, "step": 29330 }, { "epoch": 1.37, "learning_rate": 6.277949293666387e-06, "loss": 0.0795, "step": 29335 }, { "epoch": 1.37, "learning_rate": 6.275610440639911e-06, "loss": 0.0529, "step": 29340 }, { "epoch": 1.37, "learning_rate": 6.273271587613436e-06, "loss": 0.0273, "step": 29345 }, { "epoch": 1.37, "learning_rate": 6.270932734586959e-06, "loss": 0.066, "step": 29350 }, { "epoch": 1.37, "learning_rate": 6.268593881560483e-06, "loss": 0.0387, "step": 29355 }, { "epoch": 1.37, "learning_rate": 6.266255028534007e-06, "loss": 0.1135, "step": 29360 }, { "epoch": 1.37, "learning_rate": 6.263916175507531e-06, "loss": 0.0449, "step": 29365 }, { "epoch": 1.37, "learning_rate": 6.261577322481056e-06, "loss": 0.0971, "step": 29370 }, { "epoch": 1.37, "learning_rate": 6.25923846945458e-06, "loss": 0.052, "step": 29375 }, { "epoch": 1.37, "learning_rate": 6.256899616428104e-06, "loss": 0.0204, "step": 29380 }, { "epoch": 1.37, "learning_rate": 6.254560763401628e-06, "loss": 0.0345, "step": 29385 }, { "epoch": 1.37, "learning_rate": 6.252221910375153e-06, "loss": 0.0466, "step": 29390 }, { "epoch": 1.38, "learning_rate": 6.249883057348677e-06, "loss": 0.0401, "step": 29395 }, { "epoch": 1.38, "learning_rate": 6.247544204322201e-06, "loss": 0.1113, "step": 29400 }, { "epoch": 1.38, "learning_rate": 6.245205351295726e-06, "loss": 0.0312, "step": 29405 }, { "epoch": 1.38, "learning_rate": 6.242866498269249e-06, "loss": 0.0835, "step": 29410 }, { "epoch": 1.38, "learning_rate": 6.240527645242773e-06, "loss": 0.0731, "step": 29415 }, { "epoch": 1.38, "learning_rate": 6.238188792216297e-06, "loss": 0.0534, "step": 29420 }, { "epoch": 1.38, "learning_rate": 6.235849939189822e-06, "loss": 0.113, "step": 29425 }, { "epoch": 1.38, "learning_rate": 6.233511086163346e-06, "loss": 0.0733, "step": 29430 }, { "epoch": 1.38, "learning_rate": 6.2311722331368705e-06, "loss": 0.0335, "step": 29435 }, { "epoch": 1.38, "learning_rate": 6.228833380110394e-06, "loss": 0.0635, "step": 29440 }, { "epoch": 1.38, "learning_rate": 6.226494527083919e-06, "loss": 0.0331, "step": 29445 }, { "epoch": 1.38, "learning_rate": 6.224155674057443e-06, "loss": 0.029, "step": 29450 }, { "epoch": 1.38, "learning_rate": 6.221816821030967e-06, "loss": 0.0685, "step": 29455 }, { "epoch": 1.38, "learning_rate": 6.219477968004492e-06, "loss": 0.0857, "step": 29460 }, { "epoch": 1.38, "learning_rate": 6.217139114978016e-06, "loss": 0.0685, "step": 29465 }, { "epoch": 1.38, "learning_rate": 6.214800261951539e-06, "loss": 0.0509, "step": 29470 }, { "epoch": 1.38, "learning_rate": 6.212461408925063e-06, "loss": 0.0853, "step": 29475 }, { "epoch": 1.38, "learning_rate": 6.2101225558985874e-06, "loss": 0.0872, "step": 29480 }, { "epoch": 1.38, "learning_rate": 6.207783702872112e-06, "loss": 0.0257, "step": 29485 }, { "epoch": 1.38, "learning_rate": 6.2054448498456365e-06, "loss": 0.0281, "step": 29490 }, { "epoch": 1.38, "learning_rate": 6.203105996819161e-06, "loss": 0.0845, "step": 29495 }, { "epoch": 1.38, "learning_rate": 6.200767143792684e-06, "loss": 0.0581, "step": 29500 }, { "epoch": 1.38, "learning_rate": 6.198428290766209e-06, "loss": 0.0668, "step": 29505 }, { "epoch": 1.38, "learning_rate": 6.196089437739733e-06, "loss": 0.1214, "step": 29510 }, { "epoch": 1.38, "learning_rate": 6.193750584713257e-06, "loss": 0.0488, "step": 29515 }, { "epoch": 1.38, "learning_rate": 6.191411731686782e-06, "loss": 0.0435, "step": 29520 }, { "epoch": 1.38, "learning_rate": 6.189072878660306e-06, "loss": 0.0272, "step": 29525 }, { "epoch": 1.38, "learning_rate": 6.186734025633829e-06, "loss": 0.0523, "step": 29530 }, { "epoch": 1.38, "learning_rate": 6.184395172607353e-06, "loss": 0.0711, "step": 29535 }, { "epoch": 1.38, "learning_rate": 6.182056319580878e-06, "loss": 0.128, "step": 29540 }, { "epoch": 1.38, "learning_rate": 6.1797174665544025e-06, "loss": 0.0, "step": 29545 }, { "epoch": 1.38, "learning_rate": 6.177378613527927e-06, "loss": 0.0901, "step": 29550 }, { "epoch": 1.38, "learning_rate": 6.17503976050145e-06, "loss": 0.074, "step": 29555 }, { "epoch": 1.38, "learning_rate": 6.172700907474975e-06, "loss": 0.0407, "step": 29560 }, { "epoch": 1.38, "learning_rate": 6.170362054448499e-06, "loss": 0.0551, "step": 29565 }, { "epoch": 1.38, "learning_rate": 6.168023201422023e-06, "loss": 0.114, "step": 29570 }, { "epoch": 1.38, "learning_rate": 6.165684348395547e-06, "loss": 0.04, "step": 29575 }, { "epoch": 1.38, "learning_rate": 6.163345495369072e-06, "loss": 0.0614, "step": 29580 }, { "epoch": 1.38, "learning_rate": 6.161006642342595e-06, "loss": 0.0279, "step": 29585 }, { "epoch": 1.38, "learning_rate": 6.158667789316119e-06, "loss": 0.0495, "step": 29590 }, { "epoch": 1.38, "learning_rate": 6.1563289362896435e-06, "loss": 0.0667, "step": 29595 }, { "epoch": 1.38, "learning_rate": 6.1539900832631685e-06, "loss": 0.026, "step": 29600 }, { "epoch": 1.38, "learning_rate": 6.151651230236693e-06, "loss": 0.0915, "step": 29605 }, { "epoch": 1.39, "learning_rate": 6.149312377210217e-06, "loss": 0.0607, "step": 29610 }, { "epoch": 1.39, "learning_rate": 6.14697352418374e-06, "loss": 0.0353, "step": 29615 }, { "epoch": 1.39, "learning_rate": 6.144634671157265e-06, "loss": 0.0485, "step": 29620 }, { "epoch": 1.39, "learning_rate": 6.142295818130789e-06, "loss": 0.0239, "step": 29625 }, { "epoch": 1.39, "learning_rate": 6.139956965104313e-06, "loss": 0.0269, "step": 29630 }, { "epoch": 1.39, "learning_rate": 6.137618112077838e-06, "loss": 0.0413, "step": 29635 }, { "epoch": 1.39, "learning_rate": 6.135279259051362e-06, "loss": 0.0358, "step": 29640 }, { "epoch": 1.39, "learning_rate": 6.132940406024885e-06, "loss": 0.0209, "step": 29645 }, { "epoch": 1.39, "learning_rate": 6.1306015529984095e-06, "loss": 0.0267, "step": 29650 }, { "epoch": 1.39, "learning_rate": 6.1282626999719345e-06, "loss": 0.0751, "step": 29655 }, { "epoch": 1.39, "learning_rate": 6.1259238469454586e-06, "loss": 0.037, "step": 29660 }, { "epoch": 1.39, "learning_rate": 6.123584993918983e-06, "loss": 0.0424, "step": 29665 }, { "epoch": 1.39, "learning_rate": 6.121246140892508e-06, "loss": 0.0618, "step": 29670 }, { "epoch": 1.39, "learning_rate": 6.118907287866031e-06, "loss": 0.0307, "step": 29675 }, { "epoch": 1.39, "learning_rate": 6.116568434839555e-06, "loss": 0.0847, "step": 29680 }, { "epoch": 1.39, "learning_rate": 6.114229581813079e-06, "loss": 0.0651, "step": 29685 }, { "epoch": 1.39, "learning_rate": 6.111890728786603e-06, "loss": 0.0224, "step": 29690 }, { "epoch": 1.39, "learning_rate": 6.109551875760128e-06, "loss": 0.0684, "step": 29695 }, { "epoch": 1.39, "learning_rate": 6.107213022733652e-06, "loss": 0.0512, "step": 29700 }, { "epoch": 1.39, "learning_rate": 6.1048741697071755e-06, "loss": 0.0594, "step": 29705 }, { "epoch": 1.39, "learning_rate": 6.1025353166807e-06, "loss": 0.0738, "step": 29710 }, { "epoch": 1.39, "learning_rate": 6.1001964636542246e-06, "loss": 0.032, "step": 29715 }, { "epoch": 1.39, "learning_rate": 6.097857610627749e-06, "loss": 0.0358, "step": 29720 }, { "epoch": 1.39, "learning_rate": 6.095518757601273e-06, "loss": 0.0738, "step": 29725 }, { "epoch": 1.39, "learning_rate": 6.093179904574798e-06, "loss": 0.0257, "step": 29730 }, { "epoch": 1.39, "learning_rate": 6.090841051548321e-06, "loss": 0.0679, "step": 29735 }, { "epoch": 1.39, "learning_rate": 6.088502198521845e-06, "loss": 0.0591, "step": 29740 }, { "epoch": 1.39, "learning_rate": 6.086163345495369e-06, "loss": 0.0706, "step": 29745 }, { "epoch": 1.39, "learning_rate": 6.083824492468894e-06, "loss": 0.0211, "step": 29750 }, { "epoch": 1.39, "learning_rate": 6.081485639442418e-06, "loss": 0.0077, "step": 29755 }, { "epoch": 1.39, "learning_rate": 6.079146786415942e-06, "loss": 0.0289, "step": 29760 }, { "epoch": 1.39, "learning_rate": 6.076807933389466e-06, "loss": 0.0928, "step": 29765 }, { "epoch": 1.39, "learning_rate": 6.0744690803629906e-06, "loss": 0.0514, "step": 29770 }, { "epoch": 1.39, "learning_rate": 6.072130227336515e-06, "loss": 0.0586, "step": 29775 }, { "epoch": 1.39, "learning_rate": 6.069791374310039e-06, "loss": 0.054, "step": 29780 }, { "epoch": 1.39, "learning_rate": 6.067452521283564e-06, "loss": 0.0159, "step": 29785 }, { "epoch": 1.39, "learning_rate": 6.065113668257088e-06, "loss": 0.0248, "step": 29790 }, { "epoch": 1.39, "learning_rate": 6.062774815230611e-06, "loss": 0.0791, "step": 29795 }, { "epoch": 1.39, "learning_rate": 6.060435962204135e-06, "loss": 0.0693, "step": 29800 }, { "epoch": 1.39, "learning_rate": 6.058097109177659e-06, "loss": 0.0218, "step": 29805 }, { "epoch": 1.39, "learning_rate": 6.055758256151184e-06, "loss": 0.0795, "step": 29810 }, { "epoch": 1.39, "learning_rate": 6.053419403124708e-06, "loss": 0.0259, "step": 29815 }, { "epoch": 1.39, "learning_rate": 6.0510805500982325e-06, "loss": 0.0668, "step": 29820 }, { "epoch": 1.4, "learning_rate": 6.048741697071756e-06, "loss": 0.0668, "step": 29825 }, { "epoch": 1.4, "learning_rate": 6.046402844045281e-06, "loss": 0.0176, "step": 29830 }, { "epoch": 1.4, "learning_rate": 6.044063991018805e-06, "loss": 0.0616, "step": 29835 }, { "epoch": 1.4, "learning_rate": 6.041725137992329e-06, "loss": 0.0402, "step": 29840 }, { "epoch": 1.4, "learning_rate": 6.039386284965854e-06, "loss": 0.094, "step": 29845 }, { "epoch": 1.4, "learning_rate": 6.037047431939378e-06, "loss": 0.0527, "step": 29850 }, { "epoch": 1.4, "learning_rate": 6.034708578912901e-06, "loss": 0.0425, "step": 29855 }, { "epoch": 1.4, "learning_rate": 6.032369725886425e-06, "loss": 0.0458, "step": 29860 }, { "epoch": 1.4, "learning_rate": 6.03003087285995e-06, "loss": 0.1441, "step": 29865 }, { "epoch": 1.4, "learning_rate": 6.027692019833474e-06, "loss": 0.0234, "step": 29870 }, { "epoch": 1.4, "learning_rate": 6.0253531668069985e-06, "loss": 0.0151, "step": 29875 }, { "epoch": 1.4, "learning_rate": 6.023014313780523e-06, "loss": 0.0485, "step": 29880 }, { "epoch": 1.4, "learning_rate": 6.020675460754047e-06, "loss": 0.0418, "step": 29885 }, { "epoch": 1.4, "learning_rate": 6.018336607727571e-06, "loss": 0.0375, "step": 29890 }, { "epoch": 1.4, "learning_rate": 6.015997754701095e-06, "loss": 0.0384, "step": 29895 }, { "epoch": 1.4, "learning_rate": 6.01365890167462e-06, "loss": 0.0306, "step": 29900 }, { "epoch": 1.4, "learning_rate": 6.011320048648144e-06, "loss": 0.0653, "step": 29905 }, { "epoch": 1.4, "learning_rate": 6.008981195621668e-06, "loss": 0.0602, "step": 29910 }, { "epoch": 1.4, "learning_rate": 6.006642342595191e-06, "loss": 0.0808, "step": 29915 }, { "epoch": 1.4, "learning_rate": 6.004303489568715e-06, "loss": 0.0393, "step": 29920 }, { "epoch": 1.4, "learning_rate": 6.00196463654224e-06, "loss": 0.0535, "step": 29925 }, { "epoch": 1.4, "learning_rate": 5.9996257835157644e-06, "loss": 0.0756, "step": 29930 }, { "epoch": 1.4, "learning_rate": 5.9972869304892886e-06, "loss": 0.0451, "step": 29935 }, { "epoch": 1.4, "learning_rate": 5.9949480774628135e-06, "loss": 0.0574, "step": 29940 }, { "epoch": 1.4, "learning_rate": 5.992609224436337e-06, "loss": 0.0949, "step": 29945 }, { "epoch": 1.4, "learning_rate": 5.990270371409861e-06, "loss": 0.0774, "step": 29950 }, { "epoch": 1.4, "learning_rate": 5.987931518383385e-06, "loss": 0.1074, "step": 29955 }, { "epoch": 1.4, "learning_rate": 5.98559266535691e-06, "loss": 0.0832, "step": 29960 }, { "epoch": 1.4, "learning_rate": 5.983253812330434e-06, "loss": 0.0474, "step": 29965 }, { "epoch": 1.4, "learning_rate": 5.980914959303958e-06, "loss": 0.0243, "step": 29970 }, { "epoch": 1.4, "learning_rate": 5.978576106277481e-06, "loss": 0.0745, "step": 29975 }, { "epoch": 1.4, "learning_rate": 5.976237253251006e-06, "loss": 0.0584, "step": 29980 }, { "epoch": 1.4, "learning_rate": 5.9738984002245304e-06, "loss": 0.0889, "step": 29985 }, { "epoch": 1.4, "learning_rate": 5.9715595471980545e-06, "loss": 0.0732, "step": 29990 }, { "epoch": 1.4, "learning_rate": 5.9692206941715795e-06, "loss": 0.0455, "step": 29995 }, { "epoch": 1.4, "learning_rate": 5.966881841145103e-06, "loss": 0.0563, "step": 30000 }, { "epoch": 1.4, "learning_rate": 5.964542988118627e-06, "loss": 0.0407, "step": 30005 }, { "epoch": 1.4, "learning_rate": 5.962204135092151e-06, "loss": 0.0384, "step": 30010 }, { "epoch": 1.4, "learning_rate": 5.959865282065675e-06, "loss": 0.0494, "step": 30015 }, { "epoch": 1.4, "learning_rate": 5.9575264290392e-06, "loss": 0.0815, "step": 30020 }, { "epoch": 1.4, "learning_rate": 5.955187576012724e-06, "loss": 0.0753, "step": 30025 }, { "epoch": 1.4, "learning_rate": 5.952848722986247e-06, "loss": 0.0436, "step": 30030 }, { "epoch": 1.4, "learning_rate": 5.9505098699597715e-06, "loss": 0.0375, "step": 30035 }, { "epoch": 1.41, "learning_rate": 5.9481710169332964e-06, "loss": 0.056, "step": 30040 }, { "epoch": 1.41, "learning_rate": 5.9458321639068205e-06, "loss": 0.0675, "step": 30045 }, { "epoch": 1.41, "learning_rate": 5.943493310880345e-06, "loss": 0.0518, "step": 30050 }, { "epoch": 1.41, "learning_rate": 5.94115445785387e-06, "loss": 0.0427, "step": 30055 }, { "epoch": 1.41, "learning_rate": 5.938815604827393e-06, "loss": 0.0784, "step": 30060 }, { "epoch": 1.41, "learning_rate": 5.936476751800917e-06, "loss": 0.0999, "step": 30065 }, { "epoch": 1.41, "learning_rate": 5.934137898774441e-06, "loss": 0.0471, "step": 30070 }, { "epoch": 1.41, "learning_rate": 5.931799045747966e-06, "loss": 0.0537, "step": 30075 }, { "epoch": 1.41, "learning_rate": 5.92946019272149e-06, "loss": 0.0704, "step": 30080 }, { "epoch": 1.41, "learning_rate": 5.927121339695014e-06, "loss": 0.058, "step": 30085 }, { "epoch": 1.41, "learning_rate": 5.9247824866685375e-06, "loss": 0.0635, "step": 30090 }, { "epoch": 1.41, "learning_rate": 5.9224436336420624e-06, "loss": 0.1067, "step": 30095 }, { "epoch": 1.41, "learning_rate": 5.9201047806155865e-06, "loss": 0.0993, "step": 30100 }, { "epoch": 1.41, "learning_rate": 5.917765927589111e-06, "loss": 0.0811, "step": 30105 }, { "epoch": 1.41, "learning_rate": 5.915427074562636e-06, "loss": 0.0569, "step": 30110 }, { "epoch": 1.41, "learning_rate": 5.91308822153616e-06, "loss": 0.0859, "step": 30115 }, { "epoch": 1.41, "learning_rate": 5.910749368509683e-06, "loss": 0.0554, "step": 30120 }, { "epoch": 1.41, "learning_rate": 5.908410515483207e-06, "loss": 0.1321, "step": 30125 }, { "epoch": 1.41, "learning_rate": 5.906071662456731e-06, "loss": 0.0407, "step": 30130 }, { "epoch": 1.41, "learning_rate": 5.903732809430256e-06, "loss": 0.0381, "step": 30135 }, { "epoch": 1.41, "learning_rate": 5.90139395640378e-06, "loss": 0.0573, "step": 30140 }, { "epoch": 1.41, "learning_rate": 5.899055103377304e-06, "loss": 0.0729, "step": 30145 }, { "epoch": 1.41, "learning_rate": 5.896716250350828e-06, "loss": 0.0917, "step": 30150 }, { "epoch": 1.41, "learning_rate": 5.8943773973243525e-06, "loss": 0.0507, "step": 30155 }, { "epoch": 1.41, "learning_rate": 5.892038544297877e-06, "loss": 0.1158, "step": 30160 }, { "epoch": 1.41, "learning_rate": 5.889699691271401e-06, "loss": 0.0634, "step": 30165 }, { "epoch": 1.41, "learning_rate": 5.887360838244926e-06, "loss": 0.053, "step": 30170 }, { "epoch": 1.41, "learning_rate": 5.88502198521845e-06, "loss": 0.0556, "step": 30175 }, { "epoch": 1.41, "learning_rate": 5.882683132191973e-06, "loss": 0.0645, "step": 30180 }, { "epoch": 1.41, "learning_rate": 5.880344279165497e-06, "loss": 0.0595, "step": 30185 }, { "epoch": 1.41, "learning_rate": 5.878005426139022e-06, "loss": 0.0386, "step": 30190 }, { "epoch": 1.41, "learning_rate": 5.875666573112546e-06, "loss": 0.0325, "step": 30195 }, { "epoch": 1.41, "learning_rate": 5.87332772008607e-06, "loss": 0.0407, "step": 30200 }, { "epoch": 1.41, "learning_rate": 5.870988867059595e-06, "loss": 0.0266, "step": 30205 }, { "epoch": 1.41, "learning_rate": 5.8686500140331185e-06, "loss": 0.0172, "step": 30210 }, { "epoch": 1.41, "learning_rate": 5.866311161006643e-06, "loss": 0.0582, "step": 30215 }, { "epoch": 1.41, "learning_rate": 5.863972307980167e-06, "loss": 0.0427, "step": 30220 }, { "epoch": 1.41, "learning_rate": 5.861633454953692e-06, "loss": 0.0552, "step": 30225 }, { "epoch": 1.41, "learning_rate": 5.859294601927216e-06, "loss": 0.038, "step": 30230 }, { "epoch": 1.41, "learning_rate": 5.85695574890074e-06, "loss": 0.0402, "step": 30235 }, { "epoch": 1.41, "learning_rate": 5.854616895874263e-06, "loss": 0.0374, "step": 30240 }, { "epoch": 1.41, "learning_rate": 5.852278042847787e-06, "loss": 0.0471, "step": 30245 }, { "epoch": 1.41, "learning_rate": 5.849939189821312e-06, "loss": 0.0522, "step": 30250 }, { "epoch": 1.42, "learning_rate": 5.847600336794836e-06, "loss": 0.0624, "step": 30255 }, { "epoch": 1.42, "learning_rate": 5.8452614837683604e-06, "loss": 0.0346, "step": 30260 }, { "epoch": 1.42, "learning_rate": 5.842922630741885e-06, "loss": 0.049, "step": 30265 }, { "epoch": 1.42, "learning_rate": 5.840583777715409e-06, "loss": 0.0451, "step": 30270 }, { "epoch": 1.42, "learning_rate": 5.838244924688933e-06, "loss": 0.1263, "step": 30275 }, { "epoch": 1.42, "learning_rate": 5.835906071662457e-06, "loss": 0.0425, "step": 30280 }, { "epoch": 1.42, "learning_rate": 5.833567218635982e-06, "loss": 0.0498, "step": 30285 }, { "epoch": 1.42, "learning_rate": 5.831228365609506e-06, "loss": 0.0373, "step": 30290 }, { "epoch": 1.42, "learning_rate": 5.82888951258303e-06, "loss": 0.0698, "step": 30295 }, { "epoch": 1.42, "learning_rate": 5.826550659556553e-06, "loss": 0.0476, "step": 30300 }, { "epoch": 1.42, "learning_rate": 5.824211806530078e-06, "loss": 0.0566, "step": 30305 }, { "epoch": 1.42, "learning_rate": 5.821872953503602e-06, "loss": 0.0811, "step": 30310 }, { "epoch": 1.42, "learning_rate": 5.819534100477126e-06, "loss": 0.0452, "step": 30315 }, { "epoch": 1.42, "learning_rate": 5.817195247450651e-06, "loss": 0.0649, "step": 30320 }, { "epoch": 1.42, "learning_rate": 5.8148563944241755e-06, "loss": 0.0408, "step": 30325 }, { "epoch": 1.42, "learning_rate": 5.812517541397699e-06, "loss": 0.0525, "step": 30330 }, { "epoch": 1.42, "learning_rate": 5.810178688371223e-06, "loss": 0.0474, "step": 30335 }, { "epoch": 1.42, "learning_rate": 5.807839835344748e-06, "loss": 0.0745, "step": 30340 }, { "epoch": 1.42, "learning_rate": 5.805500982318272e-06, "loss": 0.0563, "step": 30345 }, { "epoch": 1.42, "learning_rate": 5.803162129291796e-06, "loss": 0.071, "step": 30350 }, { "epoch": 1.42, "learning_rate": 5.80082327626532e-06, "loss": 0.0319, "step": 30355 }, { "epoch": 1.42, "learning_rate": 5.798484423238843e-06, "loss": 0.0826, "step": 30360 }, { "epoch": 1.42, "learning_rate": 5.796145570212368e-06, "loss": 0.0528, "step": 30365 }, { "epoch": 1.42, "learning_rate": 5.793806717185892e-06, "loss": 0.0688, "step": 30370 }, { "epoch": 1.42, "learning_rate": 5.7914678641594165e-06, "loss": 0.0142, "step": 30375 }, { "epoch": 1.42, "learning_rate": 5.7891290111329415e-06, "loss": 0.0749, "step": 30380 }, { "epoch": 1.42, "learning_rate": 5.786790158106466e-06, "loss": 0.0257, "step": 30385 }, { "epoch": 1.42, "learning_rate": 5.784451305079989e-06, "loss": 0.0818, "step": 30390 }, { "epoch": 1.42, "learning_rate": 5.782112452053513e-06, "loss": 0.0434, "step": 30395 }, { "epoch": 1.42, "learning_rate": 5.779773599027038e-06, "loss": 0.034, "step": 30400 }, { "epoch": 1.42, "learning_rate": 5.777434746000562e-06, "loss": 0.0835, "step": 30405 }, { "epoch": 1.42, "learning_rate": 5.775095892974086e-06, "loss": 0.0406, "step": 30410 }, { "epoch": 1.42, "learning_rate": 5.772757039947611e-06, "loss": 0.0596, "step": 30415 }, { "epoch": 1.42, "learning_rate": 5.770418186921134e-06, "loss": 0.0982, "step": 30420 }, { "epoch": 1.42, "learning_rate": 5.768079333894658e-06, "loss": 0.0842, "step": 30425 }, { "epoch": 1.42, "learning_rate": 5.7657404808681825e-06, "loss": 0.079, "step": 30430 }, { "epoch": 1.42, "learning_rate": 5.7634016278417075e-06, "loss": 0.1011, "step": 30435 }, { "epoch": 1.42, "learning_rate": 5.7610627748152316e-06, "loss": 0.0434, "step": 30440 }, { "epoch": 1.42, "learning_rate": 5.758723921788755e-06, "loss": 0.0361, "step": 30445 }, { "epoch": 1.42, "learning_rate": 5.756385068762279e-06, "loss": 0.0404, "step": 30450 }, { "epoch": 1.42, "learning_rate": 5.754046215735803e-06, "loss": 0.0479, "step": 30455 }, { "epoch": 1.42, "learning_rate": 5.751707362709328e-06, "loss": 0.0533, "step": 30460 }, { "epoch": 1.43, "learning_rate": 5.749368509682852e-06, "loss": 0.1245, "step": 30465 }, { "epoch": 1.43, "learning_rate": 5.747029656656376e-06, "loss": 0.0616, "step": 30470 }, { "epoch": 1.43, "learning_rate": 5.7446908036298995e-06, "loss": 0.0516, "step": 30475 }, { "epoch": 1.43, "learning_rate": 5.742351950603424e-06, "loss": 0.0447, "step": 30480 }, { "epoch": 1.43, "learning_rate": 5.7400130975769485e-06, "loss": 0.0336, "step": 30485 }, { "epoch": 1.43, "learning_rate": 5.737674244550473e-06, "loss": 0.0516, "step": 30490 }, { "epoch": 1.43, "learning_rate": 5.7353353915239976e-06, "loss": 0.044, "step": 30495 }, { "epoch": 1.43, "learning_rate": 5.732996538497522e-06, "loss": 0.0827, "step": 30500 }, { "epoch": 1.43, "learning_rate": 5.730657685471045e-06, "loss": 0.0503, "step": 30505 }, { "epoch": 1.43, "learning_rate": 5.728318832444569e-06, "loss": 0.0583, "step": 30510 }, { "epoch": 1.43, "learning_rate": 5.725979979418094e-06, "loss": 0.0597, "step": 30515 }, { "epoch": 1.43, "learning_rate": 5.723641126391618e-06, "loss": 0.064, "step": 30520 }, { "epoch": 1.43, "learning_rate": 5.721302273365142e-06, "loss": 0.0769, "step": 30525 }, { "epoch": 1.43, "learning_rate": 5.718963420338667e-06, "loss": 0.0465, "step": 30530 }, { "epoch": 1.43, "learning_rate": 5.71662456731219e-06, "loss": 0.0605, "step": 30535 }, { "epoch": 1.43, "learning_rate": 5.7142857142857145e-06, "loss": 0.0518, "step": 30540 }, { "epoch": 1.43, "learning_rate": 5.711946861259239e-06, "loss": 0.0604, "step": 30545 }, { "epoch": 1.43, "learning_rate": 5.7096080082327636e-06, "loss": 0.0839, "step": 30550 }, { "epoch": 1.43, "learning_rate": 5.707269155206288e-06, "loss": 0.0941, "step": 30555 }, { "epoch": 1.43, "learning_rate": 5.704930302179812e-06, "loss": 0.0859, "step": 30560 }, { "epoch": 1.43, "learning_rate": 5.702591449153335e-06, "loss": 0.0223, "step": 30565 }, { "epoch": 1.43, "learning_rate": 5.700252596126859e-06, "loss": 0.0398, "step": 30570 }, { "epoch": 1.43, "learning_rate": 5.697913743100384e-06, "loss": 0.0428, "step": 30575 }, { "epoch": 1.43, "learning_rate": 5.695574890073908e-06, "loss": 0.0622, "step": 30580 }, { "epoch": 1.43, "learning_rate": 5.693236037047432e-06, "loss": 0.0363, "step": 30585 }, { "epoch": 1.43, "learning_rate": 5.690897184020957e-06, "loss": 0.0563, "step": 30590 }, { "epoch": 1.43, "learning_rate": 5.6885583309944805e-06, "loss": 0.0588, "step": 30595 }, { "epoch": 1.43, "learning_rate": 5.686219477968005e-06, "loss": 0.0638, "step": 30600 }, { "epoch": 1.43, "learning_rate": 5.683880624941529e-06, "loss": 0.0718, "step": 30605 }, { "epoch": 1.43, "learning_rate": 5.681541771915054e-06, "loss": 0.061, "step": 30610 }, { "epoch": 1.43, "learning_rate": 5.679202918888578e-06, "loss": 0.0703, "step": 30615 }, { "epoch": 1.43, "learning_rate": 5.676864065862102e-06, "loss": 0.0369, "step": 30620 }, { "epoch": 1.43, "learning_rate": 5.674525212835625e-06, "loss": 0.0511, "step": 30625 }, { "epoch": 1.43, "learning_rate": 5.67218635980915e-06, "loss": 0.0595, "step": 30630 }, { "epoch": 1.43, "learning_rate": 5.669847506782674e-06, "loss": 0.0377, "step": 30635 }, { "epoch": 1.43, "learning_rate": 5.667508653756198e-06, "loss": 0.0306, "step": 30640 }, { "epoch": 1.43, "learning_rate": 5.665169800729723e-06, "loss": 0.0354, "step": 30645 }, { "epoch": 1.43, "learning_rate": 5.662830947703247e-06, "loss": 0.0247, "step": 30650 }, { "epoch": 1.43, "learning_rate": 5.660492094676771e-06, "loss": 0.0526, "step": 30655 }, { "epoch": 1.43, "learning_rate": 5.658153241650295e-06, "loss": 0.0808, "step": 30660 }, { "epoch": 1.43, "learning_rate": 5.65581438862382e-06, "loss": 0.0639, "step": 30665 }, { "epoch": 1.43, "learning_rate": 5.653475535597344e-06, "loss": 0.0278, "step": 30670 }, { "epoch": 1.43, "learning_rate": 5.651136682570868e-06, "loss": 0.0386, "step": 30675 }, { "epoch": 1.44, "learning_rate": 5.648797829544392e-06, "loss": 0.053, "step": 30680 }, { "epoch": 1.44, "learning_rate": 5.646458976517915e-06, "loss": 0.0641, "step": 30685 }, { "epoch": 1.44, "learning_rate": 5.64412012349144e-06, "loss": 0.0798, "step": 30690 }, { "epoch": 1.44, "learning_rate": 5.641781270464964e-06, "loss": 0.0225, "step": 30695 }, { "epoch": 1.44, "learning_rate": 5.639442417438488e-06, "loss": 0.0804, "step": 30700 }, { "epoch": 1.44, "learning_rate": 5.637103564412013e-06, "loss": 0.0742, "step": 30705 }, { "epoch": 1.44, "learning_rate": 5.6347647113855375e-06, "loss": 0.0411, "step": 30710 }, { "epoch": 1.44, "learning_rate": 5.632425858359061e-06, "loss": 0.0579, "step": 30715 }, { "epoch": 1.44, "learning_rate": 5.630087005332585e-06, "loss": 0.0329, "step": 30720 }, { "epoch": 1.44, "learning_rate": 5.62774815230611e-06, "loss": 0.0839, "step": 30725 }, { "epoch": 1.44, "learning_rate": 5.625409299279634e-06, "loss": 0.0722, "step": 30730 }, { "epoch": 1.44, "learning_rate": 5.623070446253158e-06, "loss": 0.0307, "step": 30735 }, { "epoch": 1.44, "learning_rate": 5.620731593226683e-06, "loss": 0.0395, "step": 30740 }, { "epoch": 1.44, "learning_rate": 5.618392740200206e-06, "loss": 0.0568, "step": 30745 }, { "epoch": 1.44, "learning_rate": 5.61605388717373e-06, "loss": 0.0492, "step": 30750 }, { "epoch": 1.44, "learning_rate": 5.613715034147254e-06, "loss": 0.0791, "step": 30755 }, { "epoch": 1.44, "learning_rate": 5.611376181120779e-06, "loss": 0.0646, "step": 30760 }, { "epoch": 1.44, "learning_rate": 5.6090373280943034e-06, "loss": 0.0546, "step": 30765 }, { "epoch": 1.44, "learning_rate": 5.6066984750678276e-06, "loss": 0.0244, "step": 30770 }, { "epoch": 1.44, "learning_rate": 5.604359622041351e-06, "loss": 0.0149, "step": 30775 }, { "epoch": 1.44, "learning_rate": 5.602020769014876e-06, "loss": 0.0687, "step": 30780 }, { "epoch": 1.44, "learning_rate": 5.5996819159884e-06, "loss": 0.0943, "step": 30785 }, { "epoch": 1.44, "learning_rate": 5.597343062961924e-06, "loss": 0.0447, "step": 30790 }, { "epoch": 1.44, "learning_rate": 5.595004209935448e-06, "loss": 0.0802, "step": 30795 }, { "epoch": 1.44, "learning_rate": 5.592665356908973e-06, "loss": 0.0122, "step": 30800 }, { "epoch": 1.44, "learning_rate": 5.590326503882496e-06, "loss": 0.0791, "step": 30805 }, { "epoch": 1.44, "learning_rate": 5.58798765085602e-06, "loss": 0.0521, "step": 30810 }, { "epoch": 1.44, "learning_rate": 5.5856487978295445e-06, "loss": 0.0359, "step": 30815 }, { "epoch": 1.44, "learning_rate": 5.5833099448030694e-06, "loss": 0.0373, "step": 30820 }, { "epoch": 1.44, "learning_rate": 5.5809710917765935e-06, "loss": 0.0364, "step": 30825 }, { "epoch": 1.44, "learning_rate": 5.578632238750118e-06, "loss": 0.0318, "step": 30830 }, { "epoch": 1.44, "learning_rate": 5.576293385723641e-06, "loss": 0.089, "step": 30835 }, { "epoch": 1.44, "learning_rate": 5.573954532697166e-06, "loss": 0.0795, "step": 30840 }, { "epoch": 1.44, "learning_rate": 5.57161567967069e-06, "loss": 0.0625, "step": 30845 }, { "epoch": 1.44, "learning_rate": 5.569276826644214e-06, "loss": 0.0426, "step": 30850 }, { "epoch": 1.44, "learning_rate": 5.566937973617739e-06, "loss": 0.0954, "step": 30855 }, { "epoch": 1.44, "learning_rate": 5.564599120591262e-06, "loss": 0.0283, "step": 30860 }, { "epoch": 1.44, "learning_rate": 5.562260267564786e-06, "loss": 0.0383, "step": 30865 }, { "epoch": 1.44, "learning_rate": 5.5599214145383105e-06, "loss": 0.0677, "step": 30870 }, { "epoch": 1.44, "learning_rate": 5.5575825615118354e-06, "loss": 0.0324, "step": 30875 }, { "epoch": 1.44, "learning_rate": 5.5552437084853595e-06, "loss": 0.0726, "step": 30880 }, { "epoch": 1.44, "learning_rate": 5.552904855458884e-06, "loss": 0.0926, "step": 30885 }, { "epoch": 1.44, "learning_rate": 5.550566002432407e-06, "loss": 0.0593, "step": 30890 }, { "epoch": 1.45, "learning_rate": 5.548227149405932e-06, "loss": 0.0759, "step": 30895 }, { "epoch": 1.45, "learning_rate": 5.545888296379456e-06, "loss": 0.053, "step": 30900 }, { "epoch": 1.45, "learning_rate": 5.54354944335298e-06, "loss": 0.0932, "step": 30905 }, { "epoch": 1.45, "learning_rate": 5.541210590326504e-06, "loss": 0.0904, "step": 30910 }, { "epoch": 1.45, "learning_rate": 5.538871737300029e-06, "loss": 0.0726, "step": 30915 }, { "epoch": 1.45, "learning_rate": 5.536532884273552e-06, "loss": 0.0368, "step": 30920 }, { "epoch": 1.45, "learning_rate": 5.5341940312470765e-06, "loss": 0.1114, "step": 30925 }, { "epoch": 1.45, "learning_rate": 5.531855178220601e-06, "loss": 0.0563, "step": 30930 }, { "epoch": 1.45, "learning_rate": 5.5295163251941255e-06, "loss": 0.0681, "step": 30935 }, { "epoch": 1.45, "learning_rate": 5.52717747216765e-06, "loss": 0.0669, "step": 30940 }, { "epoch": 1.45, "learning_rate": 5.524838619141174e-06, "loss": 0.0696, "step": 30945 }, { "epoch": 1.45, "learning_rate": 5.522499766114697e-06, "loss": 0.0675, "step": 30950 }, { "epoch": 1.45, "learning_rate": 5.520160913088222e-06, "loss": 0.1082, "step": 30955 }, { "epoch": 1.45, "learning_rate": 5.517822060061746e-06, "loss": 0.075, "step": 30960 }, { "epoch": 1.45, "learning_rate": 5.51548320703527e-06, "loss": 0.0627, "step": 30965 }, { "epoch": 1.45, "learning_rate": 5.513144354008795e-06, "loss": 0.0709, "step": 30970 }, { "epoch": 1.45, "learning_rate": 5.510805500982319e-06, "loss": 0.0257, "step": 30975 }, { "epoch": 1.45, "learning_rate": 5.5084666479558425e-06, "loss": 0.0275, "step": 30980 }, { "epoch": 1.45, "learning_rate": 5.506127794929367e-06, "loss": 0.0705, "step": 30985 }, { "epoch": 1.45, "learning_rate": 5.5037889419028915e-06, "loss": 0.0476, "step": 30990 }, { "epoch": 1.45, "learning_rate": 5.501450088876416e-06, "loss": 0.112, "step": 30995 }, { "epoch": 1.45, "learning_rate": 5.49911123584994e-06, "loss": 0.0592, "step": 31000 }, { "epoch": 1.45, "learning_rate": 5.496772382823464e-06, "loss": 0.0596, "step": 31005 }, { "epoch": 1.45, "learning_rate": 5.494433529796987e-06, "loss": 0.065, "step": 31010 }, { "epoch": 1.45, "learning_rate": 5.492094676770512e-06, "loss": 0.0709, "step": 31015 }, { "epoch": 1.45, "learning_rate": 5.489755823744036e-06, "loss": 0.0812, "step": 31020 }, { "epoch": 1.45, "learning_rate": 5.48741697071756e-06, "loss": 0.0661, "step": 31025 }, { "epoch": 1.45, "learning_rate": 5.485078117691085e-06, "loss": 0.0536, "step": 31030 }, { "epoch": 1.45, "learning_rate": 5.482739264664609e-06, "loss": 0.0547, "step": 31035 }, { "epoch": 1.45, "learning_rate": 5.480400411638133e-06, "loss": 0.0264, "step": 31040 }, { "epoch": 1.45, "learning_rate": 5.478061558611657e-06, "loss": 0.1327, "step": 31045 }, { "epoch": 1.45, "learning_rate": 5.475722705585182e-06, "loss": 0.0726, "step": 31050 }, { "epoch": 1.45, "learning_rate": 5.473383852558706e-06, "loss": 0.0703, "step": 31055 }, { "epoch": 1.45, "learning_rate": 5.47104499953223e-06, "loss": 0.0453, "step": 31060 }, { "epoch": 1.45, "learning_rate": 5.468706146505755e-06, "loss": 0.0489, "step": 31065 }, { "epoch": 1.45, "learning_rate": 5.466367293479278e-06, "loss": 0.0639, "step": 31070 }, { "epoch": 1.45, "learning_rate": 5.464028440452802e-06, "loss": 0.0509, "step": 31075 }, { "epoch": 1.45, "learning_rate": 5.461689587426326e-06, "loss": 0.0478, "step": 31080 }, { "epoch": 1.45, "learning_rate": 5.459350734399851e-06, "loss": 0.0442, "step": 31085 }, { "epoch": 1.45, "learning_rate": 5.457011881373375e-06, "loss": 0.0714, "step": 31090 }, { "epoch": 1.45, "learning_rate": 5.4546730283468994e-06, "loss": 0.0337, "step": 31095 }, { "epoch": 1.45, "learning_rate": 5.452334175320423e-06, "loss": 0.0329, "step": 31100 }, { "epoch": 1.45, "learning_rate": 5.449995322293948e-06, "loss": 0.0586, "step": 31105 }, { "epoch": 1.46, "learning_rate": 5.447656469267472e-06, "loss": 0.0517, "step": 31110 }, { "epoch": 1.46, "learning_rate": 5.445317616240996e-06, "loss": 0.0479, "step": 31115 }, { "epoch": 1.46, "learning_rate": 5.44297876321452e-06, "loss": 0.0514, "step": 31120 }, { "epoch": 1.46, "learning_rate": 5.440639910188045e-06, "loss": 0.0816, "step": 31125 }, { "epoch": 1.46, "learning_rate": 5.438301057161568e-06, "loss": 0.0525, "step": 31130 }, { "epoch": 1.46, "learning_rate": 5.435962204135092e-06, "loss": 0.0656, "step": 31135 }, { "epoch": 1.46, "learning_rate": 5.433623351108616e-06, "loss": 0.0262, "step": 31140 }, { "epoch": 1.46, "learning_rate": 5.431284498082141e-06, "loss": 0.023, "step": 31145 }, { "epoch": 1.46, "learning_rate": 5.428945645055665e-06, "loss": 0.0558, "step": 31150 }, { "epoch": 1.46, "learning_rate": 5.4266067920291895e-06, "loss": 0.0652, "step": 31155 }, { "epoch": 1.46, "learning_rate": 5.424267939002713e-06, "loss": 0.046, "step": 31160 }, { "epoch": 1.46, "learning_rate": 5.421929085976238e-06, "loss": 0.0614, "step": 31165 }, { "epoch": 1.46, "learning_rate": 5.419590232949762e-06, "loss": 0.0537, "step": 31170 }, { "epoch": 1.46, "learning_rate": 5.417251379923286e-06, "loss": 0.0818, "step": 31175 }, { "epoch": 1.46, "learning_rate": 5.414912526896811e-06, "loss": 0.0531, "step": 31180 }, { "epoch": 1.46, "learning_rate": 5.412573673870335e-06, "loss": 0.0434, "step": 31185 }, { "epoch": 1.46, "learning_rate": 5.410234820843858e-06, "loss": 0.0427, "step": 31190 }, { "epoch": 1.46, "learning_rate": 5.407895967817382e-06, "loss": 0.0435, "step": 31195 }, { "epoch": 1.46, "learning_rate": 5.405557114790907e-06, "loss": 0.0643, "step": 31200 }, { "epoch": 1.46, "learning_rate": 5.403218261764431e-06, "loss": 0.0657, "step": 31205 }, { "epoch": 1.46, "learning_rate": 5.4008794087379555e-06, "loss": 0.0347, "step": 31210 }, { "epoch": 1.46, "learning_rate": 5.3985405557114805e-06, "loss": 0.0258, "step": 31215 }, { "epoch": 1.46, "learning_rate": 5.396201702685004e-06, "loss": 0.0137, "step": 31220 }, { "epoch": 1.46, "learning_rate": 5.393862849658528e-06, "loss": 0.0439, "step": 31225 }, { "epoch": 1.46, "learning_rate": 5.391523996632052e-06, "loss": 0.0767, "step": 31230 }, { "epoch": 1.46, "learning_rate": 5.389185143605576e-06, "loss": 0.0444, "step": 31235 }, { "epoch": 1.46, "learning_rate": 5.386846290579101e-06, "loss": 0.084, "step": 31240 }, { "epoch": 1.46, "learning_rate": 5.384507437552625e-06, "loss": 0.0461, "step": 31245 }, { "epoch": 1.46, "learning_rate": 5.382168584526148e-06, "loss": 0.0599, "step": 31250 }, { "epoch": 1.46, "learning_rate": 5.3798297314996725e-06, "loss": 0.0604, "step": 31255 }, { "epoch": 1.46, "learning_rate": 5.377490878473197e-06, "loss": 0.0712, "step": 31260 }, { "epoch": 1.46, "learning_rate": 5.3751520254467215e-06, "loss": 0.0358, "step": 31265 }, { "epoch": 1.46, "learning_rate": 5.372813172420246e-06, "loss": 0.0563, "step": 31270 }, { "epoch": 1.46, "learning_rate": 5.3704743193937706e-06, "loss": 0.0715, "step": 31275 }, { "epoch": 1.46, "learning_rate": 5.368135466367294e-06, "loss": 0.0472, "step": 31280 }, { "epoch": 1.46, "learning_rate": 5.365796613340818e-06, "loss": 0.0796, "step": 31285 }, { "epoch": 1.46, "learning_rate": 5.363457760314342e-06, "loss": 0.083, "step": 31290 }, { "epoch": 1.46, "learning_rate": 5.361118907287867e-06, "loss": 0.0342, "step": 31295 }, { "epoch": 1.46, "learning_rate": 5.358780054261391e-06, "loss": 0.0668, "step": 31300 }, { "epoch": 1.46, "learning_rate": 5.356441201234914e-06, "loss": 0.0198, "step": 31305 }, { "epoch": 1.46, "learning_rate": 5.3541023482084385e-06, "loss": 0.0498, "step": 31310 }, { "epoch": 1.46, "learning_rate": 5.351763495181963e-06, "loss": 0.0262, "step": 31315 }, { "epoch": 1.47, "learning_rate": 5.3494246421554875e-06, "loss": 0.0522, "step": 31320 }, { "epoch": 1.47, "learning_rate": 5.347085789129012e-06, "loss": 0.0105, "step": 31325 }, { "epoch": 1.47, "learning_rate": 5.344746936102536e-06, "loss": 0.0733, "step": 31330 }, { "epoch": 1.47, "learning_rate": 5.34240808307606e-06, "loss": 0.037, "step": 31335 }, { "epoch": 1.47, "learning_rate": 5.340069230049584e-06, "loss": 0.0559, "step": 31340 }, { "epoch": 1.47, "learning_rate": 5.337730377023108e-06, "loss": 0.0867, "step": 31345 }, { "epoch": 1.47, "learning_rate": 5.335391523996632e-06, "loss": 0.0435, "step": 31350 }, { "epoch": 1.47, "learning_rate": 5.333052670970157e-06, "loss": 0.0621, "step": 31355 }, { "epoch": 1.47, "learning_rate": 5.330713817943681e-06, "loss": 0.0768, "step": 31360 }, { "epoch": 1.47, "learning_rate": 5.3283749649172045e-06, "loss": 0.0611, "step": 31365 }, { "epoch": 1.47, "learning_rate": 5.3260361118907286e-06, "loss": 0.0613, "step": 31370 }, { "epoch": 1.47, "learning_rate": 5.3236972588642535e-06, "loss": 0.0679, "step": 31375 }, { "epoch": 1.47, "learning_rate": 5.321358405837778e-06, "loss": 0.0187, "step": 31380 }, { "epoch": 1.47, "learning_rate": 5.319019552811302e-06, "loss": 0.0338, "step": 31385 }, { "epoch": 1.47, "learning_rate": 5.316680699784827e-06, "loss": 0.038, "step": 31390 }, { "epoch": 1.47, "learning_rate": 5.31434184675835e-06, "loss": 0.053, "step": 31395 }, { "epoch": 1.47, "learning_rate": 5.312002993731874e-06, "loss": 0.0473, "step": 31400 }, { "epoch": 1.47, "learning_rate": 5.309664140705398e-06, "loss": 0.0955, "step": 31405 }, { "epoch": 1.47, "learning_rate": 5.307325287678923e-06, "loss": 0.0184, "step": 31410 }, { "epoch": 1.47, "learning_rate": 5.304986434652447e-06, "loss": 0.0433, "step": 31415 }, { "epoch": 1.47, "learning_rate": 5.302647581625971e-06, "loss": 0.0845, "step": 31420 }, { "epoch": 1.47, "learning_rate": 5.3003087285994946e-06, "loss": 0.0342, "step": 31425 }, { "epoch": 1.47, "learning_rate": 5.2979698755730195e-06, "loss": 0.097, "step": 31430 }, { "epoch": 1.47, "learning_rate": 5.295631022546544e-06, "loss": 0.0872, "step": 31435 }, { "epoch": 1.47, "learning_rate": 5.293292169520068e-06, "loss": 0.0597, "step": 31440 }, { "epoch": 1.47, "learning_rate": 5.290953316493592e-06, "loss": 0.0923, "step": 31445 }, { "epoch": 1.47, "learning_rate": 5.288614463467117e-06, "loss": 0.0757, "step": 31450 }, { "epoch": 1.47, "learning_rate": 5.28627561044064e-06, "loss": 0.1333, "step": 31455 }, { "epoch": 1.47, "learning_rate": 5.283936757414164e-06, "loss": 0.0295, "step": 31460 }, { "epoch": 1.47, "learning_rate": 5.281597904387688e-06, "loss": 0.0434, "step": 31465 }, { "epoch": 1.47, "learning_rate": 5.279259051361213e-06, "loss": 0.0621, "step": 31470 }, { "epoch": 1.47, "learning_rate": 5.276920198334737e-06, "loss": 0.0593, "step": 31475 }, { "epoch": 1.47, "learning_rate": 5.274581345308261e-06, "loss": 0.0561, "step": 31480 }, { "epoch": 1.47, "learning_rate": 5.272242492281785e-06, "loss": 0.0603, "step": 31485 }, { "epoch": 1.47, "learning_rate": 5.26990363925531e-06, "loss": 0.0885, "step": 31490 }, { "epoch": 1.47, "learning_rate": 5.267564786228834e-06, "loss": 0.0489, "step": 31495 }, { "epoch": 1.47, "learning_rate": 5.265225933202358e-06, "loss": 0.0561, "step": 31500 }, { "epoch": 1.47, "learning_rate": 5.262887080175883e-06, "loss": 0.0379, "step": 31505 }, { "epoch": 1.47, "learning_rate": 5.260548227149407e-06, "loss": 0.0459, "step": 31510 }, { "epoch": 1.47, "learning_rate": 5.25820937412293e-06, "loss": 0.079, "step": 31515 }, { "epoch": 1.47, "learning_rate": 5.255870521096454e-06, "loss": 0.1162, "step": 31520 }, { "epoch": 1.47, "learning_rate": 5.253531668069979e-06, "loss": 0.0702, "step": 31525 }, { "epoch": 1.47, "learning_rate": 5.251192815043503e-06, "loss": 0.0284, "step": 31530 }, { "epoch": 1.48, "learning_rate": 5.248853962017027e-06, "loss": 0.0391, "step": 31535 }, { "epoch": 1.48, "learning_rate": 5.246515108990552e-06, "loss": 0.0591, "step": 31540 }, { "epoch": 1.48, "learning_rate": 5.244176255964076e-06, "loss": 0.0283, "step": 31545 }, { "epoch": 1.48, "learning_rate": 5.2418374029376e-06, "loss": 0.0642, "step": 31550 }, { "epoch": 1.48, "learning_rate": 5.239498549911124e-06, "loss": 0.0844, "step": 31555 }, { "epoch": 1.48, "learning_rate": 5.237159696884648e-06, "loss": 0.0456, "step": 31560 }, { "epoch": 1.48, "learning_rate": 5.234820843858173e-06, "loss": 0.0286, "step": 31565 }, { "epoch": 1.48, "learning_rate": 5.232481990831697e-06, "loss": 0.0383, "step": 31570 }, { "epoch": 1.48, "learning_rate": 5.23014313780522e-06, "loss": 0.0606, "step": 31575 }, { "epoch": 1.48, "learning_rate": 5.227804284778744e-06, "loss": 0.099, "step": 31580 }, { "epoch": 1.48, "learning_rate": 5.225465431752269e-06, "loss": 0.0381, "step": 31585 }, { "epoch": 1.48, "learning_rate": 5.223126578725793e-06, "loss": 0.0292, "step": 31590 }, { "epoch": 1.48, "learning_rate": 5.2207877256993175e-06, "loss": 0.085, "step": 31595 }, { "epoch": 1.48, "learning_rate": 5.2184488726728424e-06, "loss": 0.0531, "step": 31600 }, { "epoch": 1.48, "learning_rate": 5.216110019646366e-06, "loss": 0.0831, "step": 31605 }, { "epoch": 1.48, "learning_rate": 5.21377116661989e-06, "loss": 0.0588, "step": 31610 }, { "epoch": 1.48, "learning_rate": 5.211432313593414e-06, "loss": 0.067, "step": 31615 }, { "epoch": 1.48, "learning_rate": 5.209093460566939e-06, "loss": 0.0495, "step": 31620 }, { "epoch": 1.48, "learning_rate": 5.206754607540463e-06, "loss": 0.0239, "step": 31625 }, { "epoch": 1.48, "learning_rate": 5.204415754513987e-06, "loss": 0.0277, "step": 31630 }, { "epoch": 1.48, "learning_rate": 5.20207690148751e-06, "loss": 0.0576, "step": 31635 }, { "epoch": 1.48, "learning_rate": 5.199738048461035e-06, "loss": 0.0265, "step": 31640 }, { "epoch": 1.48, "learning_rate": 5.197399195434559e-06, "loss": 0.0481, "step": 31645 }, { "epoch": 1.48, "learning_rate": 5.1950603424080835e-06, "loss": 0.0772, "step": 31650 }, { "epoch": 1.48, "learning_rate": 5.1927214893816084e-06, "loss": 0.0401, "step": 31655 }, { "epoch": 1.48, "learning_rate": 5.1903826363551325e-06, "loss": 0.0505, "step": 31660 }, { "epoch": 1.48, "learning_rate": 5.188043783328656e-06, "loss": 0.0568, "step": 31665 }, { "epoch": 1.48, "learning_rate": 5.18570493030218e-06, "loss": 0.0788, "step": 31670 }, { "epoch": 1.48, "learning_rate": 5.183366077275704e-06, "loss": 0.0316, "step": 31675 }, { "epoch": 1.48, "learning_rate": 5.181027224249229e-06, "loss": 0.0877, "step": 31680 }, { "epoch": 1.48, "learning_rate": 5.178688371222753e-06, "loss": 0.063, "step": 31685 }, { "epoch": 1.48, "learning_rate": 5.176349518196277e-06, "loss": 0.018, "step": 31690 }, { "epoch": 1.48, "learning_rate": 5.1740106651698004e-06, "loss": 0.0642, "step": 31695 }, { "epoch": 1.48, "learning_rate": 5.171671812143325e-06, "loss": 0.0647, "step": 31700 }, { "epoch": 1.48, "learning_rate": 5.1693329591168495e-06, "loss": 0.0691, "step": 31705 }, { "epoch": 1.48, "learning_rate": 5.166994106090374e-06, "loss": 0.0243, "step": 31710 }, { "epoch": 1.48, "learning_rate": 5.1646552530638985e-06, "loss": 0.0858, "step": 31715 }, { "epoch": 1.48, "learning_rate": 5.162316400037422e-06, "loss": 0.0525, "step": 31720 }, { "epoch": 1.48, "learning_rate": 5.159977547010946e-06, "loss": 0.0085, "step": 31725 }, { "epoch": 1.48, "learning_rate": 5.15763869398447e-06, "loss": 0.0221, "step": 31730 }, { "epoch": 1.48, "learning_rate": 5.155299840957995e-06, "loss": 0.0262, "step": 31735 }, { "epoch": 1.48, "learning_rate": 5.152960987931519e-06, "loss": 0.0875, "step": 31740 }, { "epoch": 1.48, "learning_rate": 5.150622134905043e-06, "loss": 0.0315, "step": 31745 }, { "epoch": 1.49, "learning_rate": 5.1482832818785664e-06, "loss": 0.0296, "step": 31750 }, { "epoch": 1.49, "learning_rate": 5.145944428852091e-06, "loss": 0.0441, "step": 31755 }, { "epoch": 1.49, "learning_rate": 5.1436055758256155e-06, "loss": 0.0209, "step": 31760 }, { "epoch": 1.49, "learning_rate": 5.14126672279914e-06, "loss": 0.0376, "step": 31765 }, { "epoch": 1.49, "learning_rate": 5.138927869772664e-06, "loss": 0.0951, "step": 31770 }, { "epoch": 1.49, "learning_rate": 5.136589016746189e-06, "loss": 0.0696, "step": 31775 }, { "epoch": 1.49, "learning_rate": 5.134250163719712e-06, "loss": 0.0632, "step": 31780 }, { "epoch": 1.49, "learning_rate": 5.131911310693236e-06, "loss": 0.0228, "step": 31785 }, { "epoch": 1.49, "learning_rate": 5.12957245766676e-06, "loss": 0.0398, "step": 31790 }, { "epoch": 1.49, "learning_rate": 5.127233604640285e-06, "loss": 0.0512, "step": 31795 }, { "epoch": 1.49, "learning_rate": 5.124894751613809e-06, "loss": 0.0576, "step": 31800 }, { "epoch": 1.49, "learning_rate": 5.122555898587333e-06, "loss": 0.0751, "step": 31805 }, { "epoch": 1.49, "learning_rate": 5.1202170455608565e-06, "loss": 0.0891, "step": 31810 }, { "epoch": 1.49, "learning_rate": 5.1178781925343815e-06, "loss": 0.0329, "step": 31815 }, { "epoch": 1.49, "learning_rate": 5.115539339507906e-06, "loss": 0.0337, "step": 31820 }, { "epoch": 1.49, "learning_rate": 5.11320048648143e-06, "loss": 0.0415, "step": 31825 }, { "epoch": 1.49, "learning_rate": 5.110861633454955e-06, "loss": 0.0727, "step": 31830 }, { "epoch": 1.49, "learning_rate": 5.108522780428479e-06, "loss": 0.0435, "step": 31835 }, { "epoch": 1.49, "learning_rate": 5.106183927402002e-06, "loss": 0.0637, "step": 31840 }, { "epoch": 1.49, "learning_rate": 5.103845074375526e-06, "loss": 0.0515, "step": 31845 }, { "epoch": 1.49, "learning_rate": 5.101506221349051e-06, "loss": 0.0594, "step": 31850 }, { "epoch": 1.49, "learning_rate": 5.099167368322575e-06, "loss": 0.0589, "step": 31855 }, { "epoch": 1.49, "learning_rate": 5.096828515296099e-06, "loss": 0.051, "step": 31860 }, { "epoch": 1.49, "learning_rate": 5.094489662269624e-06, "loss": 0.0637, "step": 31865 }, { "epoch": 1.49, "learning_rate": 5.0921508092431475e-06, "loss": 0.0149, "step": 31870 }, { "epoch": 1.49, "learning_rate": 5.089811956216672e-06, "loss": 0.027, "step": 31875 }, { "epoch": 1.49, "learning_rate": 5.087473103190196e-06, "loss": 0.1286, "step": 31880 }, { "epoch": 1.49, "learning_rate": 5.08513425016372e-06, "loss": 0.0621, "step": 31885 }, { "epoch": 1.49, "learning_rate": 5.082795397137245e-06, "loss": 0.0922, "step": 31890 }, { "epoch": 1.49, "learning_rate": 5.080456544110769e-06, "loss": 0.066, "step": 31895 }, { "epoch": 1.49, "learning_rate": 5.078117691084292e-06, "loss": 0.0466, "step": 31900 }, { "epoch": 1.49, "learning_rate": 5.075778838057816e-06, "loss": 0.0476, "step": 31905 }, { "epoch": 1.49, "learning_rate": 5.073439985031341e-06, "loss": 0.0231, "step": 31910 }, { "epoch": 1.49, "learning_rate": 5.071101132004865e-06, "loss": 0.0737, "step": 31915 }, { "epoch": 1.49, "learning_rate": 5.068762278978389e-06, "loss": 0.0453, "step": 31920 }, { "epoch": 1.49, "learning_rate": 5.066423425951914e-06, "loss": 0.0745, "step": 31925 }, { "epoch": 1.49, "learning_rate": 5.0640845729254376e-06, "loss": 0.0205, "step": 31930 }, { "epoch": 1.49, "learning_rate": 5.061745719898962e-06, "loss": 0.1242, "step": 31935 }, { "epoch": 1.49, "learning_rate": 5.059406866872486e-06, "loss": 0.0469, "step": 31940 }, { "epoch": 1.49, "learning_rate": 5.057068013846011e-06, "loss": 0.0789, "step": 31945 }, { "epoch": 1.49, "learning_rate": 5.054729160819535e-06, "loss": 0.0534, "step": 31950 }, { "epoch": 1.49, "learning_rate": 5.052390307793059e-06, "loss": 0.055, "step": 31955 }, { "epoch": 1.49, "learning_rate": 5.050051454766582e-06, "loss": 0.0735, "step": 31960 }, { "epoch": 1.5, "learning_rate": 5.047712601740107e-06, "loss": 0.0679, "step": 31965 }, { "epoch": 1.5, "learning_rate": 5.045373748713631e-06, "loss": 0.0383, "step": 31970 }, { "epoch": 1.5, "learning_rate": 5.043034895687155e-06, "loss": 0.053, "step": 31975 }, { "epoch": 1.5, "learning_rate": 5.04069604266068e-06, "loss": 0.1331, "step": 31980 }, { "epoch": 1.5, "learning_rate": 5.038357189634204e-06, "loss": 0.0596, "step": 31985 }, { "epoch": 1.5, "learning_rate": 5.036018336607728e-06, "loss": 0.0689, "step": 31990 }, { "epoch": 1.5, "learning_rate": 5.033679483581252e-06, "loss": 0.0913, "step": 31995 }, { "epoch": 1.5, "learning_rate": 5.031340630554776e-06, "loss": 0.0458, "step": 32000 }, { "epoch": 1.5, "learning_rate": 5.029001777528301e-06, "loss": 0.0982, "step": 32005 }, { "epoch": 1.5, "learning_rate": 5.026662924501825e-06, "loss": 0.0614, "step": 32010 }, { "epoch": 1.5, "learning_rate": 5.024324071475349e-06, "loss": 0.0596, "step": 32015 }, { "epoch": 1.5, "learning_rate": 5.021985218448872e-06, "loss": 0.0442, "step": 32020 }, { "epoch": 1.5, "learning_rate": 5.019646365422397e-06, "loss": 0.0354, "step": 32025 }, { "epoch": 1.5, "learning_rate": 5.017307512395921e-06, "loss": 0.0693, "step": 32030 }, { "epoch": 1.5, "learning_rate": 5.0149686593694455e-06, "loss": 0.071, "step": 32035 }, { "epoch": 1.5, "learning_rate": 5.01262980634297e-06, "loss": 0.0326, "step": 32040 }, { "epoch": 1.5, "learning_rate": 5.0102909533164945e-06, "loss": 0.049, "step": 32045 }, { "epoch": 1.5, "learning_rate": 5.007952100290018e-06, "loss": 0.0904, "step": 32050 }, { "epoch": 1.5, "learning_rate": 5.005613247263542e-06, "loss": 0.0408, "step": 32055 }, { "epoch": 1.5, "learning_rate": 5.003274394237067e-06, "loss": 0.1029, "step": 32060 }, { "epoch": 1.5, "learning_rate": 5.000935541210591e-06, "loss": 0.0803, "step": 32065 }, { "epoch": 1.5, "learning_rate": 4.998596688184115e-06, "loss": 0.0438, "step": 32070 }, { "epoch": 1.5, "learning_rate": 4.996257835157639e-06, "loss": 0.0693, "step": 32075 }, { "epoch": 1.5, "learning_rate": 4.993918982131163e-06, "loss": 0.0662, "step": 32080 }, { "epoch": 1.5, "learning_rate": 4.991580129104687e-06, "loss": 0.0329, "step": 32085 }, { "epoch": 1.5, "learning_rate": 4.9892412760782115e-06, "loss": 0.0797, "step": 32090 }, { "epoch": 1.5, "learning_rate": 4.986902423051736e-06, "loss": 0.0565, "step": 32095 }, { "epoch": 1.5, "learning_rate": 4.98456357002526e-06, "loss": 0.0478, "step": 32100 }, { "epoch": 1.5, "learning_rate": 4.982224716998784e-06, "loss": 0.0546, "step": 32105 }, { "epoch": 1.5, "learning_rate": 4.979885863972309e-06, "loss": 0.0543, "step": 32110 }, { "epoch": 1.5, "learning_rate": 4.977547010945832e-06, "loss": 0.0858, "step": 32115 }, { "epoch": 1.5, "learning_rate": 4.975208157919357e-06, "loss": 0.0694, "step": 32120 }, { "epoch": 1.5, "learning_rate": 4.972869304892881e-06, "loss": 0.0486, "step": 32125 }, { "epoch": 1.5, "learning_rate": 4.970530451866405e-06, "loss": 0.0427, "step": 32130 }, { "epoch": 1.5, "learning_rate": 4.968191598839929e-06, "loss": 0.1175, "step": 32135 }, { "epoch": 1.5, "learning_rate": 4.965852745813453e-06, "loss": 0.0647, "step": 32140 }, { "epoch": 1.5, "learning_rate": 4.9635138927869775e-06, "loss": 0.0429, "step": 32145 }, { "epoch": 1.5, "learning_rate": 4.9611750397605016e-06, "loss": 0.0399, "step": 32150 }, { "epoch": 1.5, "learning_rate": 4.9588361867340265e-06, "loss": 0.0672, "step": 32155 }, { "epoch": 1.5, "learning_rate": 4.95649733370755e-06, "loss": 0.0509, "step": 32160 }, { "epoch": 1.5, "learning_rate": 4.954158480681075e-06, "loss": 0.0355, "step": 32165 }, { "epoch": 1.5, "learning_rate": 4.951819627654599e-06, "loss": 0.0528, "step": 32170 }, { "epoch": 1.51, "learning_rate": 4.949480774628123e-06, "loss": 0.086, "step": 32175 }, { "epoch": 1.51, "learning_rate": 4.947141921601647e-06, "loss": 0.0549, "step": 32180 }, { "epoch": 1.51, "learning_rate": 4.944803068575171e-06, "loss": 0.0419, "step": 32185 }, { "epoch": 1.51, "learning_rate": 4.942464215548695e-06, "loss": 0.0079, "step": 32190 }, { "epoch": 1.51, "learning_rate": 4.940125362522219e-06, "loss": 0.0276, "step": 32195 }, { "epoch": 1.51, "learning_rate": 4.937786509495744e-06, "loss": 0.03, "step": 32200 }, { "epoch": 1.51, "learning_rate": 4.9354476564692676e-06, "loss": 0.0197, "step": 32205 }, { "epoch": 1.51, "learning_rate": 4.933108803442792e-06, "loss": 0.0589, "step": 32210 }, { "epoch": 1.51, "learning_rate": 4.930769950416317e-06, "loss": 0.0421, "step": 32215 }, { "epoch": 1.51, "learning_rate": 4.92843109738984e-06, "loss": 0.0428, "step": 32220 }, { "epoch": 1.51, "learning_rate": 4.926092244363365e-06, "loss": 0.0516, "step": 32225 }, { "epoch": 1.51, "learning_rate": 4.923753391336889e-06, "loss": 0.0796, "step": 32230 }, { "epoch": 1.51, "learning_rate": 4.921414538310413e-06, "loss": 0.026, "step": 32235 }, { "epoch": 1.51, "learning_rate": 4.919075685283937e-06, "loss": 0.0934, "step": 32240 }, { "epoch": 1.51, "learning_rate": 4.916736832257461e-06, "loss": 0.0899, "step": 32245 }, { "epoch": 1.51, "learning_rate": 4.914397979230985e-06, "loss": 0.0322, "step": 32250 }, { "epoch": 1.51, "learning_rate": 4.9120591262045094e-06, "loss": 0.0665, "step": 32255 }, { "epoch": 1.51, "learning_rate": 4.909720273178034e-06, "loss": 0.075, "step": 32260 }, { "epoch": 1.51, "learning_rate": 4.907381420151558e-06, "loss": 0.0873, "step": 32265 }, { "epoch": 1.51, "learning_rate": 4.905042567125083e-06, "loss": 0.0236, "step": 32270 }, { "epoch": 1.51, "learning_rate": 4.902703714098607e-06, "loss": 0.0643, "step": 32275 }, { "epoch": 1.51, "learning_rate": 4.900364861072131e-06, "loss": 0.0385, "step": 32280 }, { "epoch": 1.51, "learning_rate": 4.898026008045655e-06, "loss": 0.0221, "step": 32285 }, { "epoch": 1.51, "learning_rate": 4.895687155019179e-06, "loss": 0.0975, "step": 32290 }, { "epoch": 1.51, "learning_rate": 4.893348301992703e-06, "loss": 0.0587, "step": 32295 }, { "epoch": 1.51, "learning_rate": 4.891009448966227e-06, "loss": 0.0661, "step": 32300 }, { "epoch": 1.51, "learning_rate": 4.888670595939752e-06, "loss": 0.0415, "step": 32305 }, { "epoch": 1.51, "learning_rate": 4.8863317429132754e-06, "loss": 0.1224, "step": 32310 }, { "epoch": 1.51, "learning_rate": 4.8839928898868e-06, "loss": 0.0374, "step": 32315 }, { "epoch": 1.51, "learning_rate": 4.8816540368603245e-06, "loss": 0.0492, "step": 32320 }, { "epoch": 1.51, "learning_rate": 4.879315183833848e-06, "loss": 0.048, "step": 32325 }, { "epoch": 1.51, "learning_rate": 4.876976330807373e-06, "loss": 0.0428, "step": 32330 }, { "epoch": 1.51, "learning_rate": 4.874637477780897e-06, "loss": 0.0273, "step": 32335 }, { "epoch": 1.51, "learning_rate": 4.872298624754421e-06, "loss": 0.0414, "step": 32340 }, { "epoch": 1.51, "learning_rate": 4.869959771727945e-06, "loss": 0.0181, "step": 32345 }, { "epoch": 1.51, "learning_rate": 4.867620918701469e-06, "loss": 0.0784, "step": 32350 }, { "epoch": 1.51, "learning_rate": 4.865282065674993e-06, "loss": 0.0547, "step": 32355 }, { "epoch": 1.51, "learning_rate": 4.862943212648517e-06, "loss": 0.0627, "step": 32360 }, { "epoch": 1.51, "learning_rate": 4.860604359622042e-06, "loss": 0.0321, "step": 32365 }, { "epoch": 1.51, "learning_rate": 4.8582655065955655e-06, "loss": 0.0329, "step": 32370 }, { "epoch": 1.51, "learning_rate": 4.8559266535690905e-06, "loss": 0.0103, "step": 32375 }, { "epoch": 1.51, "learning_rate": 4.853587800542614e-06, "loss": 0.0465, "step": 32380 }, { "epoch": 1.51, "learning_rate": 4.851248947516139e-06, "loss": 0.0661, "step": 32385 }, { "epoch": 1.52, "learning_rate": 4.848910094489663e-06, "loss": 0.1013, "step": 32390 }, { "epoch": 1.52, "learning_rate": 4.846571241463187e-06, "loss": 0.0556, "step": 32395 }, { "epoch": 1.52, "learning_rate": 4.844232388436711e-06, "loss": 0.0643, "step": 32400 }, { "epoch": 1.52, "learning_rate": 4.841893535410235e-06, "loss": 0.0621, "step": 32405 }, { "epoch": 1.52, "learning_rate": 4.839554682383759e-06, "loss": 0.0443, "step": 32410 }, { "epoch": 1.52, "learning_rate": 4.837215829357283e-06, "loss": 0.0921, "step": 32415 }, { "epoch": 1.52, "learning_rate": 4.834876976330808e-06, "loss": 0.0733, "step": 32420 }, { "epoch": 1.52, "learning_rate": 4.8325381233043315e-06, "loss": 0.0609, "step": 32425 }, { "epoch": 1.52, "learning_rate": 4.830199270277856e-06, "loss": 0.017, "step": 32430 }, { "epoch": 1.52, "learning_rate": 4.827860417251381e-06, "loss": 0.0402, "step": 32435 }, { "epoch": 1.52, "learning_rate": 4.825521564224904e-06, "loss": 0.0496, "step": 32440 }, { "epoch": 1.52, "learning_rate": 4.823182711198429e-06, "loss": 0.0432, "step": 32445 }, { "epoch": 1.52, "learning_rate": 4.820843858171953e-06, "loss": 0.0243, "step": 32450 }, { "epoch": 1.52, "learning_rate": 4.818505005145477e-06, "loss": 0.0464, "step": 32455 }, { "epoch": 1.52, "learning_rate": 4.816166152119001e-06, "loss": 0.0297, "step": 32460 }, { "epoch": 1.52, "learning_rate": 4.813827299092525e-06, "loss": 0.0545, "step": 32465 }, { "epoch": 1.52, "learning_rate": 4.811488446066049e-06, "loss": 0.0392, "step": 32470 }, { "epoch": 1.52, "learning_rate": 4.8091495930395734e-06, "loss": 0.0646, "step": 32475 }, { "epoch": 1.52, "learning_rate": 4.806810740013098e-06, "loss": 0.0928, "step": 32480 }, { "epoch": 1.52, "learning_rate": 4.804471886986622e-06, "loss": 0.067, "step": 32485 }, { "epoch": 1.52, "learning_rate": 4.802133033960147e-06, "loss": 0.0404, "step": 32490 }, { "epoch": 1.52, "learning_rate": 4.799794180933671e-06, "loss": 0.1058, "step": 32495 }, { "epoch": 1.52, "learning_rate": 4.797455327907195e-06, "loss": 0.0293, "step": 32500 }, { "epoch": 1.52, "learning_rate": 4.795116474880719e-06, "loss": 0.0508, "step": 32505 }, { "epoch": 1.52, "learning_rate": 4.792777621854243e-06, "loss": 0.0341, "step": 32510 }, { "epoch": 1.52, "learning_rate": 4.790438768827767e-06, "loss": 0.0496, "step": 32515 }, { "epoch": 1.52, "learning_rate": 4.788099915801291e-06, "loss": 0.0591, "step": 32520 }, { "epoch": 1.52, "learning_rate": 4.785761062774816e-06, "loss": 0.0539, "step": 32525 }, { "epoch": 1.52, "learning_rate": 4.7834222097483394e-06, "loss": 0.0698, "step": 32530 }, { "epoch": 1.52, "learning_rate": 4.781083356721864e-06, "loss": 0.0248, "step": 32535 }, { "epoch": 1.52, "learning_rate": 4.7787445036953885e-06, "loss": 0.099, "step": 32540 }, { "epoch": 1.52, "learning_rate": 4.776405650668912e-06, "loss": 0.053, "step": 32545 }, { "epoch": 1.52, "learning_rate": 4.774066797642437e-06, "loss": 0.0575, "step": 32550 }, { "epoch": 1.52, "learning_rate": 4.771727944615961e-06, "loss": 0.0651, "step": 32555 }, { "epoch": 1.52, "learning_rate": 4.769389091589485e-06, "loss": 0.1007, "step": 32560 }, { "epoch": 1.52, "learning_rate": 4.767050238563009e-06, "loss": 0.0427, "step": 32565 }, { "epoch": 1.52, "learning_rate": 4.764711385536533e-06, "loss": 0.0656, "step": 32570 }, { "epoch": 1.52, "learning_rate": 4.762372532510057e-06, "loss": 0.0336, "step": 32575 }, { "epoch": 1.52, "learning_rate": 4.760033679483581e-06, "loss": 0.0782, "step": 32580 }, { "epoch": 1.52, "learning_rate": 4.757694826457106e-06, "loss": 0.1083, "step": 32585 }, { "epoch": 1.52, "learning_rate": 4.7553559734306295e-06, "loss": 0.0516, "step": 32590 }, { "epoch": 1.52, "learning_rate": 4.7530171204041545e-06, "loss": 0.0901, "step": 32595 }, { "epoch": 1.52, "learning_rate": 4.750678267377679e-06, "loss": 0.062, "step": 32600 }, { "epoch": 1.53, "learning_rate": 4.748339414351203e-06, "loss": 0.0488, "step": 32605 }, { "epoch": 1.53, "learning_rate": 4.746000561324727e-06, "loss": 0.0722, "step": 32610 }, { "epoch": 1.53, "learning_rate": 4.743661708298251e-06, "loss": 0.0165, "step": 32615 }, { "epoch": 1.53, "learning_rate": 4.741322855271775e-06, "loss": 0.0696, "step": 32620 }, { "epoch": 1.53, "learning_rate": 4.738984002245299e-06, "loss": 0.056, "step": 32625 }, { "epoch": 1.53, "learning_rate": 4.736645149218824e-06, "loss": 0.0734, "step": 32630 }, { "epoch": 1.53, "learning_rate": 4.734306296192347e-06, "loss": 0.1025, "step": 32635 }, { "epoch": 1.53, "learning_rate": 4.731967443165872e-06, "loss": 0.0638, "step": 32640 }, { "epoch": 1.53, "learning_rate": 4.729628590139396e-06, "loss": 0.0565, "step": 32645 }, { "epoch": 1.53, "learning_rate": 4.72728973711292e-06, "loss": 0.0251, "step": 32650 }, { "epoch": 1.53, "learning_rate": 4.724950884086445e-06, "loss": 0.0784, "step": 32655 }, { "epoch": 1.53, "learning_rate": 4.722612031059969e-06, "loss": 0.0307, "step": 32660 }, { "epoch": 1.53, "learning_rate": 4.720273178033493e-06, "loss": 0.1599, "step": 32665 }, { "epoch": 1.53, "learning_rate": 4.717934325007017e-06, "loss": 0.0418, "step": 32670 }, { "epoch": 1.53, "learning_rate": 4.715595471980541e-06, "loss": 0.0661, "step": 32675 }, { "epoch": 1.53, "learning_rate": 4.713256618954065e-06, "loss": 0.0598, "step": 32680 }, { "epoch": 1.53, "learning_rate": 4.710917765927589e-06, "loss": 0.0508, "step": 32685 }, { "epoch": 1.53, "learning_rate": 4.708578912901114e-06, "loss": 0.0444, "step": 32690 }, { "epoch": 1.53, "learning_rate": 4.706240059874637e-06, "loss": 0.0525, "step": 32695 }, { "epoch": 1.53, "learning_rate": 4.703901206848162e-06, "loss": 0.0323, "step": 32700 }, { "epoch": 1.53, "learning_rate": 4.7015623538216865e-06, "loss": 0.0705, "step": 32705 }, { "epoch": 1.53, "learning_rate": 4.699223500795211e-06, "loss": 0.0479, "step": 32710 }, { "epoch": 1.53, "learning_rate": 4.696884647768735e-06, "loss": 0.0876, "step": 32715 }, { "epoch": 1.53, "learning_rate": 4.694545794742259e-06, "loss": 0.1073, "step": 32720 }, { "epoch": 1.53, "learning_rate": 4.692206941715783e-06, "loss": 0.0701, "step": 32725 }, { "epoch": 1.53, "learning_rate": 4.689868088689307e-06, "loss": 0.1039, "step": 32730 }, { "epoch": 1.53, "learning_rate": 4.687529235662832e-06, "loss": 0.0979, "step": 32735 }, { "epoch": 1.53, "learning_rate": 4.685190382636355e-06, "loss": 0.0471, "step": 32740 }, { "epoch": 1.53, "learning_rate": 4.68285152960988e-06, "loss": 0.066, "step": 32745 }, { "epoch": 1.53, "learning_rate": 4.680512676583404e-06, "loss": 0.0254, "step": 32750 }, { "epoch": 1.53, "learning_rate": 4.678173823556928e-06, "loss": 0.0548, "step": 32755 }, { "epoch": 1.53, "learning_rate": 4.6758349705304525e-06, "loss": 0.0555, "step": 32760 }, { "epoch": 1.53, "learning_rate": 4.6734961175039766e-06, "loss": 0.0598, "step": 32765 }, { "epoch": 1.53, "learning_rate": 4.671157264477501e-06, "loss": 0.1119, "step": 32770 }, { "epoch": 1.53, "learning_rate": 4.668818411451025e-06, "loss": 0.069, "step": 32775 }, { "epoch": 1.53, "learning_rate": 4.666479558424549e-06, "loss": 0.0775, "step": 32780 }, { "epoch": 1.53, "learning_rate": 4.664140705398073e-06, "loss": 0.0437, "step": 32785 }, { "epoch": 1.53, "learning_rate": 4.661801852371597e-06, "loss": 0.0869, "step": 32790 }, { "epoch": 1.53, "learning_rate": 4.659462999345122e-06, "loss": 0.0191, "step": 32795 }, { "epoch": 1.53, "learning_rate": 4.657124146318645e-06, "loss": 0.0355, "step": 32800 }, { "epoch": 1.53, "learning_rate": 4.65478529329217e-06, "loss": 0.0979, "step": 32805 }, { "epoch": 1.53, "learning_rate": 4.6524464402656935e-06, "loss": 0.0676, "step": 32810 }, { "epoch": 1.53, "learning_rate": 4.6501075872392185e-06, "loss": 0.0243, "step": 32815 }, { "epoch": 1.54, "learning_rate": 4.6477687342127426e-06, "loss": 0.0489, "step": 32820 }, { "epoch": 1.54, "learning_rate": 4.645429881186267e-06, "loss": 0.0827, "step": 32825 }, { "epoch": 1.54, "learning_rate": 4.643091028159791e-06, "loss": 0.034, "step": 32830 }, { "epoch": 1.54, "learning_rate": 4.640752175133315e-06, "loss": 0.0361, "step": 32835 }, { "epoch": 1.54, "learning_rate": 4.638413322106839e-06, "loss": 0.0576, "step": 32840 }, { "epoch": 1.54, "learning_rate": 4.636074469080363e-06, "loss": 0.0547, "step": 32845 }, { "epoch": 1.54, "learning_rate": 4.633735616053888e-06, "loss": 0.0358, "step": 32850 }, { "epoch": 1.54, "learning_rate": 4.631396763027411e-06, "loss": 0.0459, "step": 32855 }, { "epoch": 1.54, "learning_rate": 4.629057910000936e-06, "loss": 0.0695, "step": 32860 }, { "epoch": 1.54, "learning_rate": 4.62671905697446e-06, "loss": 0.0595, "step": 32865 }, { "epoch": 1.54, "learning_rate": 4.624380203947984e-06, "loss": 0.0405, "step": 32870 }, { "epoch": 1.54, "learning_rate": 4.6220413509215086e-06, "loss": 0.0261, "step": 32875 }, { "epoch": 1.54, "learning_rate": 4.619702497895033e-06, "loss": 0.0382, "step": 32880 }, { "epoch": 1.54, "learning_rate": 4.617363644868557e-06, "loss": 0.0842, "step": 32885 }, { "epoch": 1.54, "learning_rate": 4.615024791842081e-06, "loss": 0.0573, "step": 32890 }, { "epoch": 1.54, "learning_rate": 4.612685938815605e-06, "loss": 0.0709, "step": 32895 }, { "epoch": 1.54, "learning_rate": 4.610347085789129e-06, "loss": 0.0388, "step": 32900 }, { "epoch": 1.54, "learning_rate": 4.608008232762653e-06, "loss": 0.0992, "step": 32905 }, { "epoch": 1.54, "learning_rate": 4.605669379736178e-06, "loss": 0.0273, "step": 32910 }, { "epoch": 1.54, "learning_rate": 4.603330526709701e-06, "loss": 0.0608, "step": 32915 }, { "epoch": 1.54, "learning_rate": 4.600991673683226e-06, "loss": 0.0576, "step": 32920 }, { "epoch": 1.54, "learning_rate": 4.5986528206567505e-06, "loss": 0.0949, "step": 32925 }, { "epoch": 1.54, "learning_rate": 4.5963139676302746e-06, "loss": 0.0423, "step": 32930 }, { "epoch": 1.54, "learning_rate": 4.593975114603799e-06, "loss": 0.0699, "step": 32935 }, { "epoch": 1.54, "learning_rate": 4.591636261577323e-06, "loss": 0.0325, "step": 32940 }, { "epoch": 1.54, "learning_rate": 4.589297408550847e-06, "loss": 0.0625, "step": 32945 }, { "epoch": 1.54, "learning_rate": 4.586958555524371e-06, "loss": 0.0818, "step": 32950 }, { "epoch": 1.54, "learning_rate": 4.584619702497896e-06, "loss": 0.0321, "step": 32955 }, { "epoch": 1.54, "learning_rate": 4.582280849471419e-06, "loss": 0.0636, "step": 32960 }, { "epoch": 1.54, "learning_rate": 4.579941996444944e-06, "loss": 0.0878, "step": 32965 }, { "epoch": 1.54, "learning_rate": 4.577603143418468e-06, "loss": 0.0419, "step": 32970 }, { "epoch": 1.54, "learning_rate": 4.575264290391992e-06, "loss": 0.0891, "step": 32975 }, { "epoch": 1.54, "learning_rate": 4.5729254373655165e-06, "loss": 0.0717, "step": 32980 }, { "epoch": 1.54, "learning_rate": 4.5705865843390406e-06, "loss": 0.0613, "step": 32985 }, { "epoch": 1.54, "learning_rate": 4.568247731312565e-06, "loss": 0.0307, "step": 32990 }, { "epoch": 1.54, "learning_rate": 4.565908878286089e-06, "loss": 0.0735, "step": 32995 }, { "epoch": 1.54, "learning_rate": 4.563570025259613e-06, "loss": 0.0609, "step": 33000 }, { "epoch": 1.54, "learning_rate": 4.561231172233137e-06, "loss": 0.0791, "step": 33005 }, { "epoch": 1.54, "learning_rate": 4.558892319206661e-06, "loss": 0.029, "step": 33010 }, { "epoch": 1.54, "learning_rate": 4.556553466180186e-06, "loss": 0.0311, "step": 33015 }, { "epoch": 1.54, "learning_rate": 4.554214613153709e-06, "loss": 0.0618, "step": 33020 }, { "epoch": 1.54, "learning_rate": 4.551875760127234e-06, "loss": 0.0456, "step": 33025 }, { "epoch": 1.55, "learning_rate": 4.549536907100758e-06, "loss": 0.0549, "step": 33030 }, { "epoch": 1.55, "learning_rate": 4.5471980540742824e-06, "loss": 0.0346, "step": 33035 }, { "epoch": 1.55, "learning_rate": 4.5448592010478066e-06, "loss": 0.0468, "step": 33040 }, { "epoch": 1.55, "learning_rate": 4.542520348021331e-06, "loss": 0.0405, "step": 33045 }, { "epoch": 1.55, "learning_rate": 4.540181494994855e-06, "loss": 0.1004, "step": 33050 }, { "epoch": 1.55, "learning_rate": 4.537842641968379e-06, "loss": 0.0592, "step": 33055 }, { "epoch": 1.55, "learning_rate": 4.535503788941904e-06, "loss": 0.0472, "step": 33060 }, { "epoch": 1.55, "learning_rate": 4.533164935915427e-06, "loss": 0.0447, "step": 33065 }, { "epoch": 1.55, "learning_rate": 4.530826082888952e-06, "loss": 0.0996, "step": 33070 }, { "epoch": 1.55, "learning_rate": 4.528487229862476e-06, "loss": 0.0584, "step": 33075 }, { "epoch": 1.55, "learning_rate": 4.526148376836e-06, "loss": 0.0229, "step": 33080 }, { "epoch": 1.55, "learning_rate": 4.523809523809524e-06, "loss": 0.0664, "step": 33085 }, { "epoch": 1.55, "learning_rate": 4.5214706707830484e-06, "loss": 0.0486, "step": 33090 }, { "epoch": 1.55, "learning_rate": 4.5191318177565725e-06, "loss": 0.0292, "step": 33095 }, { "epoch": 1.55, "learning_rate": 4.516792964730097e-06, "loss": 0.0399, "step": 33100 }, { "epoch": 1.55, "learning_rate": 4.514454111703621e-06, "loss": 0.08, "step": 33105 }, { "epoch": 1.55, "learning_rate": 4.512115258677145e-06, "loss": 0.0642, "step": 33110 }, { "epoch": 1.55, "learning_rate": 4.509776405650669e-06, "loss": 0.0395, "step": 33115 }, { "epoch": 1.55, "learning_rate": 4.507437552624194e-06, "loss": 0.0833, "step": 33120 }, { "epoch": 1.55, "learning_rate": 4.505098699597717e-06, "loss": 0.0808, "step": 33125 }, { "epoch": 1.55, "learning_rate": 4.502759846571242e-06, "loss": 0.0227, "step": 33130 }, { "epoch": 1.55, "learning_rate": 4.500420993544766e-06, "loss": 0.0129, "step": 33135 }, { "epoch": 1.55, "learning_rate": 4.49808214051829e-06, "loss": 0.0513, "step": 33140 }, { "epoch": 1.55, "learning_rate": 4.4957432874918144e-06, "loss": 0.0556, "step": 33145 }, { "epoch": 1.55, "learning_rate": 4.4934044344653385e-06, "loss": 0.0564, "step": 33150 }, { "epoch": 1.55, "learning_rate": 4.491065581438863e-06, "loss": 0.0373, "step": 33155 }, { "epoch": 1.55, "learning_rate": 4.488726728412387e-06, "loss": 0.0527, "step": 33160 }, { "epoch": 1.55, "learning_rate": 4.486387875385912e-06, "loss": 0.0396, "step": 33165 }, { "epoch": 1.55, "learning_rate": 4.484049022359435e-06, "loss": 0.0758, "step": 33170 }, { "epoch": 1.55, "learning_rate": 4.48171016933296e-06, "loss": 0.1083, "step": 33175 }, { "epoch": 1.55, "learning_rate": 4.479371316306484e-06, "loss": 0.0338, "step": 33180 }, { "epoch": 1.55, "learning_rate": 4.477032463280008e-06, "loss": 0.0595, "step": 33185 }, { "epoch": 1.55, "learning_rate": 4.474693610253532e-06, "loss": 0.0235, "step": 33190 }, { "epoch": 1.55, "learning_rate": 4.472354757227056e-06, "loss": 0.0612, "step": 33195 }, { "epoch": 1.55, "learning_rate": 4.4700159042005804e-06, "loss": 0.0601, "step": 33200 }, { "epoch": 1.55, "learning_rate": 4.4676770511741045e-06, "loss": 0.0667, "step": 33205 }, { "epoch": 1.55, "learning_rate": 4.465338198147629e-06, "loss": 0.0589, "step": 33210 }, { "epoch": 1.55, "learning_rate": 4.462999345121153e-06, "loss": 0.06, "step": 33215 }, { "epoch": 1.55, "learning_rate": 4.460660492094677e-06, "loss": 0.0448, "step": 33220 }, { "epoch": 1.55, "learning_rate": 4.458321639068202e-06, "loss": 0.0619, "step": 33225 }, { "epoch": 1.55, "learning_rate": 4.455982786041725e-06, "loss": 0.0285, "step": 33230 }, { "epoch": 1.55, "learning_rate": 4.45364393301525e-06, "loss": 0.0875, "step": 33235 }, { "epoch": 1.55, "learning_rate": 4.451305079988774e-06, "loss": 0.0424, "step": 33240 }, { "epoch": 1.56, "learning_rate": 4.448966226962298e-06, "loss": 0.0801, "step": 33245 }, { "epoch": 1.56, "learning_rate": 4.446627373935822e-06, "loss": 0.0408, "step": 33250 }, { "epoch": 1.56, "learning_rate": 4.4442885209093464e-06, "loss": 0.094, "step": 33255 }, { "epoch": 1.56, "learning_rate": 4.4419496678828705e-06, "loss": 0.0434, "step": 33260 }, { "epoch": 1.56, "learning_rate": 4.439610814856395e-06, "loss": 0.0759, "step": 33265 }, { "epoch": 1.56, "learning_rate": 4.437271961829919e-06, "loss": 0.1136, "step": 33270 }, { "epoch": 1.56, "learning_rate": 4.434933108803443e-06, "loss": 0.0544, "step": 33275 }, { "epoch": 1.56, "learning_rate": 4.432594255776968e-06, "loss": 0.0478, "step": 33280 }, { "epoch": 1.56, "learning_rate": 4.430255402750491e-06, "loss": 0.0647, "step": 33285 }, { "epoch": 1.56, "learning_rate": 4.427916549724016e-06, "loss": 0.062, "step": 33290 }, { "epoch": 1.56, "learning_rate": 4.42557769669754e-06, "loss": 0.0417, "step": 33295 }, { "epoch": 1.56, "learning_rate": 4.423238843671064e-06, "loss": 0.0376, "step": 33300 }, { "epoch": 1.56, "learning_rate": 4.420899990644588e-06, "loss": 0.1115, "step": 33305 }, { "epoch": 1.56, "learning_rate": 4.4185611376181124e-06, "loss": 0.0437, "step": 33310 }, { "epoch": 1.56, "learning_rate": 4.4162222845916365e-06, "loss": 0.0055, "step": 33315 }, { "epoch": 1.56, "learning_rate": 4.413883431565161e-06, "loss": 0.0337, "step": 33320 }, { "epoch": 1.56, "learning_rate": 4.411544578538685e-06, "loss": 0.0382, "step": 33325 }, { "epoch": 1.56, "learning_rate": 4.409205725512209e-06, "loss": 0.0717, "step": 33330 }, { "epoch": 1.56, "learning_rate": 4.406866872485733e-06, "loss": 0.0725, "step": 33335 }, { "epoch": 1.56, "learning_rate": 4.404528019459258e-06, "loss": 0.0534, "step": 33340 }, { "epoch": 1.56, "learning_rate": 4.402189166432781e-06, "loss": 0.029, "step": 33345 }, { "epoch": 1.56, "learning_rate": 4.399850313406306e-06, "loss": 0.044, "step": 33350 }, { "epoch": 1.56, "learning_rate": 4.39751146037983e-06, "loss": 0.0216, "step": 33355 }, { "epoch": 1.56, "learning_rate": 4.395172607353354e-06, "loss": 0.0515, "step": 33360 }, { "epoch": 1.56, "learning_rate": 4.3928337543268784e-06, "loss": 0.0572, "step": 33365 }, { "epoch": 1.56, "learning_rate": 4.3904949013004025e-06, "loss": 0.0952, "step": 33370 }, { "epoch": 1.56, "learning_rate": 4.388156048273927e-06, "loss": 0.0777, "step": 33375 }, { "epoch": 1.56, "learning_rate": 4.385817195247451e-06, "loss": 0.0928, "step": 33380 }, { "epoch": 1.56, "learning_rate": 4.383478342220976e-06, "loss": 0.0644, "step": 33385 }, { "epoch": 1.56, "learning_rate": 4.381139489194499e-06, "loss": 0.0668, "step": 33390 }, { "epoch": 1.56, "learning_rate": 4.378800636168024e-06, "loss": 0.0387, "step": 33395 }, { "epoch": 1.56, "learning_rate": 4.376461783141548e-06, "loss": 0.0659, "step": 33400 }, { "epoch": 1.56, "learning_rate": 4.374122930115072e-06, "loss": 0.0243, "step": 33405 }, { "epoch": 1.56, "learning_rate": 4.371784077088596e-06, "loss": 0.0484, "step": 33410 }, { "epoch": 1.56, "learning_rate": 4.36944522406212e-06, "loss": 0.0982, "step": 33415 }, { "epoch": 1.56, "learning_rate": 4.367106371035644e-06, "loss": 0.0413, "step": 33420 }, { "epoch": 1.56, "learning_rate": 4.3647675180091685e-06, "loss": 0.0269, "step": 33425 }, { "epoch": 1.56, "learning_rate": 4.362428664982693e-06, "loss": 0.0459, "step": 33430 }, { "epoch": 1.56, "learning_rate": 4.360089811956217e-06, "loss": 0.0221, "step": 33435 }, { "epoch": 1.56, "learning_rate": 4.357750958929741e-06, "loss": 0.0555, "step": 33440 }, { "epoch": 1.56, "learning_rate": 4.355412105903266e-06, "loss": 0.0171, "step": 33445 }, { "epoch": 1.56, "learning_rate": 4.353073252876789e-06, "loss": 0.0809, "step": 33450 }, { "epoch": 1.56, "learning_rate": 4.350734399850314e-06, "loss": 0.0327, "step": 33455 }, { "epoch": 1.57, "learning_rate": 4.348395546823838e-06, "loss": 0.033, "step": 33460 }, { "epoch": 1.57, "learning_rate": 4.346056693797362e-06, "loss": 0.0647, "step": 33465 }, { "epoch": 1.57, "learning_rate": 4.343717840770886e-06, "loss": 0.038, "step": 33470 }, { "epoch": 1.57, "learning_rate": 4.34137898774441e-06, "loss": 0.0497, "step": 33475 }, { "epoch": 1.57, "learning_rate": 4.3390401347179345e-06, "loss": 0.0224, "step": 33480 }, { "epoch": 1.57, "learning_rate": 4.336701281691459e-06, "loss": 0.0605, "step": 33485 }, { "epoch": 1.57, "learning_rate": 4.334362428664984e-06, "loss": 0.0587, "step": 33490 }, { "epoch": 1.57, "learning_rate": 4.332023575638507e-06, "loss": 0.0323, "step": 33495 }, { "epoch": 1.57, "learning_rate": 4.329684722612032e-06, "loss": 0.049, "step": 33500 }, { "epoch": 1.57, "learning_rate": 4.327345869585556e-06, "loss": 0.061, "step": 33505 }, { "epoch": 1.57, "learning_rate": 4.32500701655908e-06, "loss": 0.0704, "step": 33510 }, { "epoch": 1.57, "learning_rate": 4.322668163532604e-06, "loss": 0.055, "step": 33515 }, { "epoch": 1.57, "learning_rate": 4.320329310506128e-06, "loss": 0.0599, "step": 33520 }, { "epoch": 1.57, "learning_rate": 4.317990457479652e-06, "loss": 0.0651, "step": 33525 }, { "epoch": 1.57, "learning_rate": 4.315651604453176e-06, "loss": 0.0522, "step": 33530 }, { "epoch": 1.57, "learning_rate": 4.3133127514267005e-06, "loss": 0.0525, "step": 33535 }, { "epoch": 1.57, "learning_rate": 4.310973898400225e-06, "loss": 0.0406, "step": 33540 }, { "epoch": 1.57, "learning_rate": 4.308635045373749e-06, "loss": 0.0296, "step": 33545 }, { "epoch": 1.57, "learning_rate": 4.306296192347274e-06, "loss": 0.0501, "step": 33550 }, { "epoch": 1.57, "learning_rate": 4.303957339320797e-06, "loss": 0.0956, "step": 33555 }, { "epoch": 1.57, "learning_rate": 4.301618486294322e-06, "loss": 0.0172, "step": 33560 }, { "epoch": 1.57, "learning_rate": 4.299279633267846e-06, "loss": 0.0091, "step": 33565 }, { "epoch": 1.57, "learning_rate": 4.29694078024137e-06, "loss": 0.0792, "step": 33570 }, { "epoch": 1.57, "learning_rate": 4.294601927214894e-06, "loss": 0.0167, "step": 33575 }, { "epoch": 1.57, "learning_rate": 4.292263074188418e-06, "loss": 0.0807, "step": 33580 }, { "epoch": 1.57, "learning_rate": 4.289924221161942e-06, "loss": 0.0804, "step": 33585 }, { "epoch": 1.57, "learning_rate": 4.2875853681354665e-06, "loss": 0.0569, "step": 33590 }, { "epoch": 1.57, "learning_rate": 4.2852465151089915e-06, "loss": 0.0451, "step": 33595 }, { "epoch": 1.57, "learning_rate": 4.282907662082515e-06, "loss": 0.0515, "step": 33600 }, { "epoch": 1.57, "learning_rate": 4.28056880905604e-06, "loss": 0.0663, "step": 33605 }, { "epoch": 1.57, "learning_rate": 4.278229956029564e-06, "loss": 0.0442, "step": 33610 }, { "epoch": 1.57, "learning_rate": 4.275891103003088e-06, "loss": 0.0369, "step": 33615 }, { "epoch": 1.57, "learning_rate": 4.273552249976612e-06, "loss": 0.0416, "step": 33620 }, { "epoch": 1.57, "learning_rate": 4.271213396950136e-06, "loss": 0.1308, "step": 33625 }, { "epoch": 1.57, "learning_rate": 4.26887454392366e-06, "loss": 0.0434, "step": 33630 }, { "epoch": 1.57, "learning_rate": 4.266535690897184e-06, "loss": 0.0393, "step": 33635 }, { "epoch": 1.57, "learning_rate": 4.264196837870708e-06, "loss": 0.0207, "step": 33640 }, { "epoch": 1.57, "learning_rate": 4.2618579848442325e-06, "loss": 0.0411, "step": 33645 }, { "epoch": 1.57, "learning_rate": 4.259519131817757e-06, "loss": 0.0726, "step": 33650 }, { "epoch": 1.57, "learning_rate": 4.2571802787912816e-06, "loss": 0.0691, "step": 33655 }, { "epoch": 1.57, "learning_rate": 4.254841425764805e-06, "loss": 0.0344, "step": 33660 }, { "epoch": 1.57, "learning_rate": 4.25250257273833e-06, "loss": 0.0458, "step": 33665 }, { "epoch": 1.57, "learning_rate": 4.250163719711854e-06, "loss": 0.0159, "step": 33670 }, { "epoch": 1.58, "learning_rate": 4.247824866685378e-06, "loss": 0.0586, "step": 33675 }, { "epoch": 1.58, "learning_rate": 4.245486013658902e-06, "loss": 0.0451, "step": 33680 }, { "epoch": 1.58, "learning_rate": 4.243147160632426e-06, "loss": 0.0938, "step": 33685 }, { "epoch": 1.58, "learning_rate": 4.24080830760595e-06, "loss": 0.0737, "step": 33690 }, { "epoch": 1.58, "learning_rate": 4.238469454579474e-06, "loss": 0.04, "step": 33695 }, { "epoch": 1.58, "learning_rate": 4.2361306015529985e-06, "loss": 0.0696, "step": 33700 }, { "epoch": 1.58, "learning_rate": 4.233791748526523e-06, "loss": 0.0638, "step": 33705 }, { "epoch": 1.58, "learning_rate": 4.2314528955000476e-06, "loss": 0.0394, "step": 33710 }, { "epoch": 1.58, "learning_rate": 4.229114042473571e-06, "loss": 0.0992, "step": 33715 }, { "epoch": 1.58, "learning_rate": 4.226775189447096e-06, "loss": 0.0867, "step": 33720 }, { "epoch": 1.58, "learning_rate": 4.22443633642062e-06, "loss": 0.0618, "step": 33725 }, { "epoch": 1.58, "learning_rate": 4.222097483394144e-06, "loss": 0.0461, "step": 33730 }, { "epoch": 1.58, "learning_rate": 4.219758630367668e-06, "loss": 0.0367, "step": 33735 }, { "epoch": 1.58, "learning_rate": 4.217419777341192e-06, "loss": 0.0452, "step": 33740 }, { "epoch": 1.58, "learning_rate": 4.215080924314716e-06, "loss": 0.0423, "step": 33745 }, { "epoch": 1.58, "learning_rate": 4.21274207128824e-06, "loss": 0.035, "step": 33750 }, { "epoch": 1.58, "learning_rate": 4.2104032182617645e-06, "loss": 0.0494, "step": 33755 }, { "epoch": 1.58, "learning_rate": 4.208064365235289e-06, "loss": 0.0398, "step": 33760 }, { "epoch": 1.58, "learning_rate": 4.205725512208813e-06, "loss": 0.0875, "step": 33765 }, { "epoch": 1.58, "learning_rate": 4.203386659182338e-06, "loss": 0.0441, "step": 33770 }, { "epoch": 1.58, "learning_rate": 4.201047806155861e-06, "loss": 0.0191, "step": 33775 }, { "epoch": 1.58, "learning_rate": 4.198708953129386e-06, "loss": 0.0953, "step": 33780 }, { "epoch": 1.58, "learning_rate": 4.19637010010291e-06, "loss": 0.06, "step": 33785 }, { "epoch": 1.58, "learning_rate": 4.194031247076434e-06, "loss": 0.09, "step": 33790 }, { "epoch": 1.58, "learning_rate": 4.191692394049958e-06, "loss": 0.0389, "step": 33795 }, { "epoch": 1.58, "learning_rate": 4.189353541023482e-06, "loss": 0.0815, "step": 33800 }, { "epoch": 1.58, "learning_rate": 4.187014687997006e-06, "loss": 0.0238, "step": 33805 }, { "epoch": 1.58, "learning_rate": 4.1846758349705305e-06, "loss": 0.0195, "step": 33810 }, { "epoch": 1.58, "learning_rate": 4.1823369819440555e-06, "loss": 0.0946, "step": 33815 }, { "epoch": 1.58, "learning_rate": 4.179998128917579e-06, "loss": 0.0315, "step": 33820 }, { "epoch": 1.58, "learning_rate": 4.177659275891104e-06, "loss": 0.0423, "step": 33825 }, { "epoch": 1.58, "learning_rate": 4.175320422864628e-06, "loss": 0.0576, "step": 33830 }, { "epoch": 1.58, "learning_rate": 4.172981569838152e-06, "loss": 0.0437, "step": 33835 }, { "epoch": 1.58, "learning_rate": 4.170642716811676e-06, "loss": 0.0823, "step": 33840 }, { "epoch": 1.58, "learning_rate": 4.1683038637852e-06, "loss": 0.0583, "step": 33845 }, { "epoch": 1.58, "learning_rate": 4.165965010758724e-06, "loss": 0.0321, "step": 33850 }, { "epoch": 1.58, "learning_rate": 4.163626157732248e-06, "loss": 0.0412, "step": 33855 }, { "epoch": 1.58, "learning_rate": 4.161287304705772e-06, "loss": 0.0504, "step": 33860 }, { "epoch": 1.58, "learning_rate": 4.1589484516792965e-06, "loss": 0.0513, "step": 33865 }, { "epoch": 1.58, "learning_rate": 4.156609598652821e-06, "loss": 0.054, "step": 33870 }, { "epoch": 1.58, "learning_rate": 4.1542707456263456e-06, "loss": 0.0625, "step": 33875 }, { "epoch": 1.58, "learning_rate": 4.151931892599869e-06, "loss": 0.0577, "step": 33880 }, { "epoch": 1.59, "learning_rate": 4.149593039573394e-06, "loss": 0.0943, "step": 33885 }, { "epoch": 1.59, "learning_rate": 4.147254186546918e-06, "loss": 0.0126, "step": 33890 }, { "epoch": 1.59, "learning_rate": 4.144915333520442e-06, "loss": 0.0073, "step": 33895 }, { "epoch": 1.59, "learning_rate": 4.142576480493966e-06, "loss": 0.0884, "step": 33900 }, { "epoch": 1.59, "learning_rate": 4.14023762746749e-06, "loss": 0.0811, "step": 33905 }, { "epoch": 1.59, "learning_rate": 4.137898774441014e-06, "loss": 0.0325, "step": 33910 }, { "epoch": 1.59, "learning_rate": 4.135559921414538e-06, "loss": 0.0812, "step": 33915 }, { "epoch": 1.59, "learning_rate": 4.133221068388063e-06, "loss": 0.0156, "step": 33920 }, { "epoch": 1.59, "learning_rate": 4.130882215361587e-06, "loss": 0.0752, "step": 33925 }, { "epoch": 1.59, "learning_rate": 4.1285433623351115e-06, "loss": 0.0667, "step": 33930 }, { "epoch": 1.59, "learning_rate": 4.126204509308636e-06, "loss": 0.0362, "step": 33935 }, { "epoch": 1.59, "learning_rate": 4.12386565628216e-06, "loss": 0.0559, "step": 33940 }, { "epoch": 1.59, "learning_rate": 4.121526803255684e-06, "loss": 0.0545, "step": 33945 }, { "epoch": 1.59, "learning_rate": 4.119187950229208e-06, "loss": 0.0267, "step": 33950 }, { "epoch": 1.59, "learning_rate": 4.116849097202732e-06, "loss": 0.025, "step": 33955 }, { "epoch": 1.59, "learning_rate": 4.114510244176256e-06, "loss": 0.0642, "step": 33960 }, { "epoch": 1.59, "learning_rate": 4.11217139114978e-06, "loss": 0.0852, "step": 33965 }, { "epoch": 1.59, "learning_rate": 4.109832538123304e-06, "loss": 0.0817, "step": 33970 }, { "epoch": 1.59, "learning_rate": 4.1074936850968285e-06, "loss": 0.0653, "step": 33975 }, { "epoch": 1.59, "learning_rate": 4.1051548320703534e-06, "loss": 0.1083, "step": 33980 }, { "epoch": 1.59, "learning_rate": 4.102815979043877e-06, "loss": 0.0695, "step": 33985 }, { "epoch": 1.59, "learning_rate": 4.100477126017402e-06, "loss": 0.0398, "step": 33990 }, { "epoch": 1.59, "learning_rate": 4.098138272990926e-06, "loss": 0.0518, "step": 33995 }, { "epoch": 1.59, "learning_rate": 4.09579941996445e-06, "loss": 0.0203, "step": 34000 }, { "epoch": 1.59, "learning_rate": 4.093460566937974e-06, "loss": 0.0674, "step": 34005 }, { "epoch": 1.59, "learning_rate": 4.091121713911498e-06, "loss": 0.033, "step": 34010 }, { "epoch": 1.59, "learning_rate": 4.088782860885022e-06, "loss": 0.024, "step": 34015 }, { "epoch": 1.59, "learning_rate": 4.086444007858546e-06, "loss": 0.0618, "step": 34020 }, { "epoch": 1.59, "learning_rate": 4.084105154832071e-06, "loss": 0.0531, "step": 34025 }, { "epoch": 1.59, "learning_rate": 4.0817663018055945e-06, "loss": 0.0415, "step": 34030 }, { "epoch": 1.59, "learning_rate": 4.0794274487791194e-06, "loss": 0.0406, "step": 34035 }, { "epoch": 1.59, "learning_rate": 4.0770885957526435e-06, "loss": 0.0489, "step": 34040 }, { "epoch": 1.59, "learning_rate": 4.074749742726168e-06, "loss": 0.0578, "step": 34045 }, { "epoch": 1.59, "learning_rate": 4.072410889699692e-06, "loss": 0.0749, "step": 34050 }, { "epoch": 1.59, "learning_rate": 4.070072036673216e-06, "loss": 0.0379, "step": 34055 }, { "epoch": 1.59, "learning_rate": 4.06773318364674e-06, "loss": 0.0391, "step": 34060 }, { "epoch": 1.59, "learning_rate": 4.065394330620264e-06, "loss": 0.0301, "step": 34065 }, { "epoch": 1.59, "learning_rate": 4.063055477593789e-06, "loss": 0.0472, "step": 34070 }, { "epoch": 1.59, "learning_rate": 4.060716624567312e-06, "loss": 0.0525, "step": 34075 }, { "epoch": 1.59, "learning_rate": 4.058377771540836e-06, "loss": 0.0272, "step": 34080 }, { "epoch": 1.59, "learning_rate": 4.056038918514361e-06, "loss": 0.0767, "step": 34085 }, { "epoch": 1.59, "learning_rate": 4.053700065487885e-06, "loss": 0.0313, "step": 34090 }, { "epoch": 1.59, "learning_rate": 4.0513612124614095e-06, "loss": 0.0589, "step": 34095 }, { "epoch": 1.6, "learning_rate": 4.049022359434934e-06, "loss": 0.0323, "step": 34100 }, { "epoch": 1.6, "learning_rate": 4.046683506408458e-06, "loss": 0.0543, "step": 34105 }, { "epoch": 1.6, "learning_rate": 4.044344653381982e-06, "loss": 0.043, "step": 34110 }, { "epoch": 1.6, "learning_rate": 4.042005800355506e-06, "loss": 0.0726, "step": 34115 }, { "epoch": 1.6, "learning_rate": 4.03966694732903e-06, "loss": 0.0111, "step": 34120 }, { "epoch": 1.6, "learning_rate": 4.037328094302554e-06, "loss": 0.034, "step": 34125 }, { "epoch": 1.6, "learning_rate": 4.034989241276078e-06, "loss": 0.024, "step": 34130 }, { "epoch": 1.6, "learning_rate": 4.032650388249602e-06, "loss": 0.0269, "step": 34135 }, { "epoch": 1.6, "learning_rate": 4.030311535223127e-06, "loss": 0.0581, "step": 34140 }, { "epoch": 1.6, "learning_rate": 4.027972682196651e-06, "loss": 0.013, "step": 34145 }, { "epoch": 1.6, "learning_rate": 4.0256338291701755e-06, "loss": 0.0398, "step": 34150 }, { "epoch": 1.6, "learning_rate": 4.0232949761437e-06, "loss": 0.05, "step": 34155 }, { "epoch": 1.6, "learning_rate": 4.020956123117224e-06, "loss": 0.0316, "step": 34160 }, { "epoch": 1.6, "learning_rate": 4.018617270090748e-06, "loss": 0.0505, "step": 34165 }, { "epoch": 1.6, "learning_rate": 4.016278417064272e-06, "loss": 0.0297, "step": 34170 }, { "epoch": 1.6, "learning_rate": 4.013939564037796e-06, "loss": 0.07, "step": 34175 }, { "epoch": 1.6, "learning_rate": 4.01160071101132e-06, "loss": 0.0808, "step": 34180 }, { "epoch": 1.6, "learning_rate": 4.009261857984844e-06, "loss": 0.0515, "step": 34185 }, { "epoch": 1.6, "learning_rate": 4.006923004958368e-06, "loss": 0.0506, "step": 34190 }, { "epoch": 1.6, "learning_rate": 4.0045841519318925e-06, "loss": 0.0201, "step": 34195 }, { "epoch": 1.6, "learning_rate": 4.0022452989054174e-06, "loss": 0.0667, "step": 34200 }, { "epoch": 1.6, "learning_rate": 3.999906445878941e-06, "loss": 0.058, "step": 34205 }, { "epoch": 1.6, "learning_rate": 3.997567592852466e-06, "loss": 0.0759, "step": 34210 }, { "epoch": 1.6, "learning_rate": 3.99522873982599e-06, "loss": 0.0229, "step": 34215 }, { "epoch": 1.6, "learning_rate": 3.992889886799514e-06, "loss": 0.0827, "step": 34220 }, { "epoch": 1.6, "learning_rate": 3.990551033773038e-06, "loss": 0.0517, "step": 34225 }, { "epoch": 1.6, "learning_rate": 3.988212180746562e-06, "loss": 0.0108, "step": 34230 }, { "epoch": 1.6, "learning_rate": 3.985873327720086e-06, "loss": 0.0462, "step": 34235 }, { "epoch": 1.6, "learning_rate": 3.98353447469361e-06, "loss": 0.0574, "step": 34240 }, { "epoch": 1.6, "learning_rate": 3.981195621667135e-06, "loss": 0.0313, "step": 34245 }, { "epoch": 1.6, "learning_rate": 3.9788567686406585e-06, "loss": 0.0261, "step": 34250 }, { "epoch": 1.6, "learning_rate": 3.976517915614183e-06, "loss": 0.0582, "step": 34255 }, { "epoch": 1.6, "learning_rate": 3.9741790625877075e-06, "loss": 0.0245, "step": 34260 }, { "epoch": 1.6, "learning_rate": 3.971840209561232e-06, "loss": 0.053, "step": 34265 }, { "epoch": 1.6, "learning_rate": 3.969501356534756e-06, "loss": 0.0449, "step": 34270 }, { "epoch": 1.6, "learning_rate": 3.96716250350828e-06, "loss": 0.0658, "step": 34275 }, { "epoch": 1.6, "learning_rate": 3.964823650481804e-06, "loss": 0.0941, "step": 34280 }, { "epoch": 1.6, "learning_rate": 3.962484797455328e-06, "loss": 0.0592, "step": 34285 }, { "epoch": 1.6, "learning_rate": 3.960145944428853e-06, "loss": 0.1036, "step": 34290 }, { "epoch": 1.6, "learning_rate": 3.957807091402376e-06, "loss": 0.0778, "step": 34295 }, { "epoch": 1.6, "learning_rate": 3.9554682383759e-06, "loss": 0.0658, "step": 34300 }, { "epoch": 1.6, "learning_rate": 3.953129385349425e-06, "loss": 0.057, "step": 34305 }, { "epoch": 1.6, "learning_rate": 3.9507905323229486e-06, "loss": 0.0428, "step": 34310 }, { "epoch": 1.61, "learning_rate": 3.9484516792964735e-06, "loss": 0.0463, "step": 34315 }, { "epoch": 1.61, "learning_rate": 3.946112826269998e-06, "loss": 0.0837, "step": 34320 }, { "epoch": 1.61, "learning_rate": 3.943773973243522e-06, "loss": 0.0071, "step": 34325 }, { "epoch": 1.61, "learning_rate": 3.941435120217046e-06, "loss": 0.0548, "step": 34330 }, { "epoch": 1.61, "learning_rate": 3.93909626719057e-06, "loss": 0.0312, "step": 34335 }, { "epoch": 1.61, "learning_rate": 3.936757414164094e-06, "loss": 0.0797, "step": 34340 }, { "epoch": 1.61, "learning_rate": 3.934418561137618e-06, "loss": 0.0098, "step": 34345 }, { "epoch": 1.61, "learning_rate": 3.932079708111143e-06, "loss": 0.0598, "step": 34350 }, { "epoch": 1.61, "learning_rate": 3.929740855084666e-06, "loss": 0.0671, "step": 34355 }, { "epoch": 1.61, "learning_rate": 3.927402002058191e-06, "loss": 0.0472, "step": 34360 }, { "epoch": 1.61, "learning_rate": 3.925063149031715e-06, "loss": 0.0406, "step": 34365 }, { "epoch": 1.61, "learning_rate": 3.9227242960052395e-06, "loss": 0.0661, "step": 34370 }, { "epoch": 1.61, "learning_rate": 3.920385442978764e-06, "loss": 0.0478, "step": 34375 }, { "epoch": 1.61, "learning_rate": 3.918046589952288e-06, "loss": 0.0525, "step": 34380 }, { "epoch": 1.61, "learning_rate": 3.915707736925812e-06, "loss": 0.0421, "step": 34385 }, { "epoch": 1.61, "learning_rate": 3.913368883899336e-06, "loss": 0.0641, "step": 34390 }, { "epoch": 1.61, "learning_rate": 3.911030030872861e-06, "loss": 0.1013, "step": 34395 }, { "epoch": 1.61, "learning_rate": 3.908691177846384e-06, "loss": 0.0518, "step": 34400 }, { "epoch": 1.61, "learning_rate": 3.906352324819908e-06, "loss": 0.0456, "step": 34405 }, { "epoch": 1.61, "learning_rate": 3.904013471793433e-06, "loss": 0.0827, "step": 34410 }, { "epoch": 1.61, "learning_rate": 3.9016746187669565e-06, "loss": 0.0698, "step": 34415 }, { "epoch": 1.61, "learning_rate": 3.899335765740481e-06, "loss": 0.0587, "step": 34420 }, { "epoch": 1.61, "learning_rate": 3.8969969127140055e-06, "loss": 0.0506, "step": 34425 }, { "epoch": 1.61, "learning_rate": 3.89465805968753e-06, "loss": 0.0475, "step": 34430 }, { "epoch": 1.61, "learning_rate": 3.892319206661054e-06, "loss": 0.0866, "step": 34435 }, { "epoch": 1.61, "learning_rate": 3.889980353634578e-06, "loss": 0.0457, "step": 34440 }, { "epoch": 1.61, "learning_rate": 3.887641500608102e-06, "loss": 0.0315, "step": 34445 }, { "epoch": 1.61, "learning_rate": 3.885302647581626e-06, "loss": 0.0401, "step": 34450 }, { "epoch": 1.61, "learning_rate": 3.882963794555151e-06, "loss": 0.1177, "step": 34455 }, { "epoch": 1.61, "learning_rate": 3.880624941528674e-06, "loss": 0.0817, "step": 34460 }, { "epoch": 1.61, "learning_rate": 3.878286088502199e-06, "loss": 0.0544, "step": 34465 }, { "epoch": 1.61, "learning_rate": 3.875947235475723e-06, "loss": 0.0654, "step": 34470 }, { "epoch": 1.61, "learning_rate": 3.873608382449247e-06, "loss": 0.0278, "step": 34475 }, { "epoch": 1.61, "learning_rate": 3.8712695294227715e-06, "loss": 0.049, "step": 34480 }, { "epoch": 1.61, "learning_rate": 3.868930676396296e-06, "loss": 0.0607, "step": 34485 }, { "epoch": 1.61, "learning_rate": 3.86659182336982e-06, "loss": 0.0638, "step": 34490 }, { "epoch": 1.61, "learning_rate": 3.864252970343344e-06, "loss": 0.0222, "step": 34495 }, { "epoch": 1.61, "learning_rate": 3.861914117316869e-06, "loss": 0.0472, "step": 34500 }, { "epoch": 1.61, "learning_rate": 3.859575264290392e-06, "loss": 0.0646, "step": 34505 }, { "epoch": 1.61, "learning_rate": 3.857236411263917e-06, "loss": 0.0565, "step": 34510 }, { "epoch": 1.61, "learning_rate": 3.854897558237441e-06, "loss": 0.0736, "step": 34515 }, { "epoch": 1.61, "learning_rate": 3.852558705210964e-06, "loss": 0.0543, "step": 34520 }, { "epoch": 1.61, "learning_rate": 3.850219852184489e-06, "loss": 0.0267, "step": 34525 }, { "epoch": 1.62, "learning_rate": 3.847880999158013e-06, "loss": 0.0569, "step": 34530 }, { "epoch": 1.62, "learning_rate": 3.8455421461315375e-06, "loss": 0.0444, "step": 34535 }, { "epoch": 1.62, "learning_rate": 3.843203293105062e-06, "loss": 0.0894, "step": 34540 }, { "epoch": 1.62, "learning_rate": 3.840864440078586e-06, "loss": 0.0465, "step": 34545 }, { "epoch": 1.62, "learning_rate": 3.83852558705211e-06, "loss": 0.0845, "step": 34550 }, { "epoch": 1.62, "learning_rate": 3.836186734025634e-06, "loss": 0.0662, "step": 34555 }, { "epoch": 1.62, "learning_rate": 3.833847880999158e-06, "loss": 0.032, "step": 34560 }, { "epoch": 1.62, "learning_rate": 3.831509027972682e-06, "loss": 0.0764, "step": 34565 }, { "epoch": 1.62, "learning_rate": 3.829170174946207e-06, "loss": 0.0596, "step": 34570 }, { "epoch": 1.62, "learning_rate": 3.82683132191973e-06, "loss": 0.0566, "step": 34575 }, { "epoch": 1.62, "learning_rate": 3.824492468893255e-06, "loss": 0.07, "step": 34580 }, { "epoch": 1.62, "learning_rate": 3.822153615866779e-06, "loss": 0.0488, "step": 34585 }, { "epoch": 1.62, "learning_rate": 3.8198147628403035e-06, "loss": 0.0371, "step": 34590 }, { "epoch": 1.62, "learning_rate": 3.817475909813828e-06, "loss": 0.083, "step": 34595 }, { "epoch": 1.62, "learning_rate": 3.815137056787352e-06, "loss": 0.0481, "step": 34600 }, { "epoch": 1.62, "learning_rate": 3.812798203760876e-06, "loss": 0.0285, "step": 34605 }, { "epoch": 1.62, "learning_rate": 3.8104593507344e-06, "loss": 0.1085, "step": 34610 }, { "epoch": 1.62, "learning_rate": 3.8081204977079244e-06, "loss": 0.0491, "step": 34615 }, { "epoch": 1.62, "learning_rate": 3.805781644681448e-06, "loss": 0.0387, "step": 34620 }, { "epoch": 1.62, "learning_rate": 3.8034427916549727e-06, "loss": 0.0315, "step": 34625 }, { "epoch": 1.62, "learning_rate": 3.801103938628497e-06, "loss": 0.0437, "step": 34630 }, { "epoch": 1.62, "learning_rate": 3.798765085602021e-06, "loss": 0.0285, "step": 34635 }, { "epoch": 1.62, "learning_rate": 3.7964262325755454e-06, "loss": 0.0686, "step": 34640 }, { "epoch": 1.62, "learning_rate": 3.7940873795490695e-06, "loss": 0.0867, "step": 34645 }, { "epoch": 1.62, "learning_rate": 3.7917485265225936e-06, "loss": 0.0623, "step": 34650 }, { "epoch": 1.62, "learning_rate": 3.7894096734961177e-06, "loss": 0.0124, "step": 34655 }, { "epoch": 1.62, "learning_rate": 3.7870708204696422e-06, "loss": 0.0416, "step": 34660 }, { "epoch": 1.62, "learning_rate": 3.784731967443166e-06, "loss": 0.0482, "step": 34665 }, { "epoch": 1.62, "learning_rate": 3.7823931144166904e-06, "loss": 0.0912, "step": 34670 }, { "epoch": 1.62, "learning_rate": 3.7800542613902145e-06, "loss": 0.0398, "step": 34675 }, { "epoch": 1.62, "learning_rate": 3.7777154083637387e-06, "loss": 0.0484, "step": 34680 }, { "epoch": 1.62, "learning_rate": 3.7753765553372628e-06, "loss": 0.0261, "step": 34685 }, { "epoch": 1.62, "learning_rate": 3.7730377023107873e-06, "loss": 0.0306, "step": 34690 }, { "epoch": 1.62, "learning_rate": 3.770698849284311e-06, "loss": 0.0593, "step": 34695 }, { "epoch": 1.62, "learning_rate": 3.7683599962578355e-06, "loss": 0.0403, "step": 34700 }, { "epoch": 1.62, "learning_rate": 3.76602114323136e-06, "loss": 0.0361, "step": 34705 }, { "epoch": 1.62, "learning_rate": 3.7636822902048837e-06, "loss": 0.0302, "step": 34710 }, { "epoch": 1.62, "learning_rate": 3.761343437178408e-06, "loss": 0.0645, "step": 34715 }, { "epoch": 1.62, "learning_rate": 3.7590045841519323e-06, "loss": 0.1207, "step": 34720 }, { "epoch": 1.62, "learning_rate": 3.756665731125456e-06, "loss": 0.0769, "step": 34725 }, { "epoch": 1.62, "learning_rate": 3.7543268780989805e-06, "loss": 0.0294, "step": 34730 }, { "epoch": 1.62, "learning_rate": 3.751988025072505e-06, "loss": 0.0576, "step": 34735 }, { "epoch": 1.63, "learning_rate": 3.7496491720460288e-06, "loss": 0.0543, "step": 34740 }, { "epoch": 1.63, "learning_rate": 3.7473103190195533e-06, "loss": 0.0602, "step": 34745 }, { "epoch": 1.63, "learning_rate": 3.7449714659930774e-06, "loss": 0.0645, "step": 34750 }, { "epoch": 1.63, "learning_rate": 3.7426326129666015e-06, "loss": 0.0449, "step": 34755 }, { "epoch": 1.63, "learning_rate": 3.7402937599401256e-06, "loss": 0.0596, "step": 34760 }, { "epoch": 1.63, "learning_rate": 3.73795490691365e-06, "loss": 0.0736, "step": 34765 }, { "epoch": 1.63, "learning_rate": 3.735616053887174e-06, "loss": 0.0564, "step": 34770 }, { "epoch": 1.63, "learning_rate": 3.7332772008606983e-06, "loss": 0.0389, "step": 34775 }, { "epoch": 1.63, "learning_rate": 3.7309383478342224e-06, "loss": 0.0585, "step": 34780 }, { "epoch": 1.63, "learning_rate": 3.7285994948077465e-06, "loss": 0.0581, "step": 34785 }, { "epoch": 1.63, "learning_rate": 3.7262606417812706e-06, "loss": 0.0478, "step": 34790 }, { "epoch": 1.63, "learning_rate": 3.723921788754795e-06, "loss": 0.0509, "step": 34795 }, { "epoch": 1.63, "learning_rate": 3.721582935728319e-06, "loss": 0.0736, "step": 34800 }, { "epoch": 1.63, "learning_rate": 3.7192440827018434e-06, "loss": 0.0109, "step": 34805 }, { "epoch": 1.63, "learning_rate": 3.716905229675368e-06, "loss": 0.1256, "step": 34810 }, { "epoch": 1.63, "learning_rate": 3.7145663766488916e-06, "loss": 0.0321, "step": 34815 }, { "epoch": 1.63, "learning_rate": 3.7122275236224157e-06, "loss": 0.0586, "step": 34820 }, { "epoch": 1.63, "learning_rate": 3.7098886705959402e-06, "loss": 0.0402, "step": 34825 }, { "epoch": 1.63, "learning_rate": 3.707549817569464e-06, "loss": 0.0696, "step": 34830 }, { "epoch": 1.63, "learning_rate": 3.7052109645429884e-06, "loss": 0.0697, "step": 34835 }, { "epoch": 1.63, "learning_rate": 3.702872111516513e-06, "loss": 0.0389, "step": 34840 }, { "epoch": 1.63, "learning_rate": 3.7005332584900366e-06, "loss": 0.0743, "step": 34845 }, { "epoch": 1.63, "learning_rate": 3.698194405463561e-06, "loss": 0.043, "step": 34850 }, { "epoch": 1.63, "learning_rate": 3.6958555524370853e-06, "loss": 0.0482, "step": 34855 }, { "epoch": 1.63, "learning_rate": 3.6935166994106094e-06, "loss": 0.0358, "step": 34860 }, { "epoch": 1.63, "learning_rate": 3.6911778463841335e-06, "loss": 0.068, "step": 34865 }, { "epoch": 1.63, "learning_rate": 3.688838993357658e-06, "loss": 0.0437, "step": 34870 }, { "epoch": 1.63, "learning_rate": 3.6865001403311817e-06, "loss": 0.0396, "step": 34875 }, { "epoch": 1.63, "learning_rate": 3.6841612873047062e-06, "loss": 0.0382, "step": 34880 }, { "epoch": 1.63, "learning_rate": 3.6818224342782303e-06, "loss": 0.027, "step": 34885 }, { "epoch": 1.63, "learning_rate": 3.6794835812517544e-06, "loss": 0.0312, "step": 34890 }, { "epoch": 1.63, "learning_rate": 3.6771447282252785e-06, "loss": 0.0617, "step": 34895 }, { "epoch": 1.63, "learning_rate": 3.674805875198803e-06, "loss": 0.0455, "step": 34900 }, { "epoch": 1.63, "learning_rate": 3.6724670221723267e-06, "loss": 0.0452, "step": 34905 }, { "epoch": 1.63, "learning_rate": 3.6701281691458513e-06, "loss": 0.0597, "step": 34910 }, { "epoch": 1.63, "learning_rate": 3.667789316119376e-06, "loss": 0.0291, "step": 34915 }, { "epoch": 1.63, "learning_rate": 3.6654504630928995e-06, "loss": 0.0594, "step": 34920 }, { "epoch": 1.63, "learning_rate": 3.663111610066424e-06, "loss": 0.108, "step": 34925 }, { "epoch": 1.63, "learning_rate": 3.660772757039948e-06, "loss": 0.0496, "step": 34930 }, { "epoch": 1.63, "learning_rate": 3.658433904013472e-06, "loss": 0.0393, "step": 34935 }, { "epoch": 1.63, "learning_rate": 3.6560950509869963e-06, "loss": 0.0257, "step": 34940 }, { "epoch": 1.63, "learning_rate": 3.653756197960521e-06, "loss": 0.0261, "step": 34945 }, { "epoch": 1.63, "learning_rate": 3.6514173449340445e-06, "loss": 0.0431, "step": 34950 }, { "epoch": 1.64, "learning_rate": 3.649078491907569e-06, "loss": 0.0361, "step": 34955 }, { "epoch": 1.64, "learning_rate": 3.646739638881093e-06, "loss": 0.0816, "step": 34960 }, { "epoch": 1.64, "learning_rate": 3.6444007858546173e-06, "loss": 0.0486, "step": 34965 }, { "epoch": 1.64, "learning_rate": 3.6420619328281414e-06, "loss": 0.0227, "step": 34970 }, { "epoch": 1.64, "learning_rate": 3.639723079801666e-06, "loss": 0.0653, "step": 34975 }, { "epoch": 1.64, "learning_rate": 3.6373842267751896e-06, "loss": 0.0341, "step": 34980 }, { "epoch": 1.64, "learning_rate": 3.635045373748714e-06, "loss": 0.0442, "step": 34985 }, { "epoch": 1.64, "learning_rate": 3.6327065207222378e-06, "loss": 0.0353, "step": 34990 }, { "epoch": 1.64, "learning_rate": 3.6303676676957623e-06, "loss": 0.0555, "step": 34995 }, { "epoch": 1.64, "learning_rate": 3.6280288146692864e-06, "loss": 0.0711, "step": 35000 }, { "epoch": 1.64, "learning_rate": 3.6256899616428105e-06, "loss": 0.0337, "step": 35005 }, { "epoch": 1.64, "learning_rate": 3.6233511086163346e-06, "loss": 0.0826, "step": 35010 }, { "epoch": 1.64, "learning_rate": 3.621012255589859e-06, "loss": 0.0579, "step": 35015 }, { "epoch": 1.64, "learning_rate": 3.618673402563383e-06, "loss": 0.0739, "step": 35020 }, { "epoch": 1.64, "learning_rate": 3.6163345495369074e-06, "loss": 0.0409, "step": 35025 }, { "epoch": 1.64, "learning_rate": 3.613995696510432e-06, "loss": 0.0908, "step": 35030 }, { "epoch": 1.64, "learning_rate": 3.6116568434839556e-06, "loss": 0.0936, "step": 35035 }, { "epoch": 1.64, "learning_rate": 3.6093179904574797e-06, "loss": 0.0311, "step": 35040 }, { "epoch": 1.64, "learning_rate": 3.606979137431004e-06, "loss": 0.0803, "step": 35045 }, { "epoch": 1.64, "learning_rate": 3.604640284404528e-06, "loss": 0.0326, "step": 35050 }, { "epoch": 1.64, "learning_rate": 3.6023014313780524e-06, "loss": 0.0328, "step": 35055 }, { "epoch": 1.64, "learning_rate": 3.599962578351577e-06, "loss": 0.0433, "step": 35060 }, { "epoch": 1.64, "learning_rate": 3.5976237253251006e-06, "loss": 0.0593, "step": 35065 }, { "epoch": 1.64, "learning_rate": 3.595284872298625e-06, "loss": 0.063, "step": 35070 }, { "epoch": 1.64, "learning_rate": 3.5929460192721493e-06, "loss": 0.0652, "step": 35075 }, { "epoch": 1.64, "learning_rate": 3.5906071662456734e-06, "loss": 0.0313, "step": 35080 }, { "epoch": 1.64, "learning_rate": 3.5882683132191975e-06, "loss": 0.0699, "step": 35085 }, { "epoch": 1.64, "learning_rate": 3.585929460192722e-06, "loss": 0.0633, "step": 35090 }, { "epoch": 1.64, "learning_rate": 3.5835906071662457e-06, "loss": 0.0565, "step": 35095 }, { "epoch": 1.64, "learning_rate": 3.58125175413977e-06, "loss": 0.039, "step": 35100 }, { "epoch": 1.64, "learning_rate": 3.5789129011132943e-06, "loss": 0.0564, "step": 35105 }, { "epoch": 1.64, "learning_rate": 3.5765740480868184e-06, "loss": 0.0816, "step": 35110 }, { "epoch": 1.64, "learning_rate": 3.5742351950603425e-06, "loss": 0.0475, "step": 35115 }, { "epoch": 1.64, "learning_rate": 3.571896342033867e-06, "loss": 0.0549, "step": 35120 }, { "epoch": 1.64, "learning_rate": 3.5695574890073907e-06, "loss": 0.0263, "step": 35125 }, { "epoch": 1.64, "learning_rate": 3.5672186359809153e-06, "loss": 0.0491, "step": 35130 }, { "epoch": 1.64, "learning_rate": 3.5648797829544398e-06, "loss": 0.0159, "step": 35135 }, { "epoch": 1.64, "learning_rate": 3.5625409299279635e-06, "loss": 0.0564, "step": 35140 }, { "epoch": 1.64, "learning_rate": 3.560202076901488e-06, "loss": 0.029, "step": 35145 }, { "epoch": 1.64, "learning_rate": 3.557863223875012e-06, "loss": 0.0372, "step": 35150 }, { "epoch": 1.64, "learning_rate": 3.5555243708485358e-06, "loss": 0.0584, "step": 35155 }, { "epoch": 1.64, "learning_rate": 3.5531855178220603e-06, "loss": 0.0806, "step": 35160 }, { "epoch": 1.64, "learning_rate": 3.550846664795585e-06, "loss": 0.0863, "step": 35165 }, { "epoch": 1.65, "learning_rate": 3.5485078117691085e-06, "loss": 0.0162, "step": 35170 }, { "epoch": 1.65, "learning_rate": 3.546168958742633e-06, "loss": 0.0589, "step": 35175 }, { "epoch": 1.65, "learning_rate": 3.543830105716157e-06, "loss": 0.0315, "step": 35180 }, { "epoch": 1.65, "learning_rate": 3.5414912526896813e-06, "loss": 0.0285, "step": 35185 }, { "epoch": 1.65, "learning_rate": 3.5391523996632054e-06, "loss": 0.032, "step": 35190 }, { "epoch": 1.65, "learning_rate": 3.53681354663673e-06, "loss": 0.1009, "step": 35195 }, { "epoch": 1.65, "learning_rate": 3.5344746936102536e-06, "loss": 0.0298, "step": 35200 }, { "epoch": 1.65, "learning_rate": 3.532135840583778e-06, "loss": 0.0824, "step": 35205 }, { "epoch": 1.65, "learning_rate": 3.529796987557302e-06, "loss": 0.0762, "step": 35210 }, { "epoch": 1.65, "learning_rate": 3.5274581345308263e-06, "loss": 0.0473, "step": 35215 }, { "epoch": 1.65, "learning_rate": 3.5251192815043504e-06, "loss": 0.0538, "step": 35220 }, { "epoch": 1.65, "learning_rate": 3.522780428477875e-06, "loss": 0.055, "step": 35225 }, { "epoch": 1.65, "learning_rate": 3.5204415754513986e-06, "loss": 0.0639, "step": 35230 }, { "epoch": 1.65, "learning_rate": 3.518102722424923e-06, "loss": 0.0735, "step": 35235 }, { "epoch": 1.65, "learning_rate": 3.5157638693984477e-06, "loss": 0.0629, "step": 35240 }, { "epoch": 1.65, "learning_rate": 3.5134250163719714e-06, "loss": 0.0429, "step": 35245 }, { "epoch": 1.65, "learning_rate": 3.511086163345496e-06, "loss": 0.0468, "step": 35250 }, { "epoch": 1.65, "learning_rate": 3.50874731031902e-06, "loss": 0.0459, "step": 35255 }, { "epoch": 1.65, "learning_rate": 3.5064084572925437e-06, "loss": 0.0046, "step": 35260 }, { "epoch": 1.65, "learning_rate": 3.504069604266068e-06, "loss": 0.0364, "step": 35265 }, { "epoch": 1.65, "learning_rate": 3.5017307512395927e-06, "loss": 0.0827, "step": 35270 }, { "epoch": 1.65, "learning_rate": 3.4993918982131164e-06, "loss": 0.0484, "step": 35275 }, { "epoch": 1.65, "learning_rate": 3.497053045186641e-06, "loss": 0.0632, "step": 35280 }, { "epoch": 1.65, "learning_rate": 3.494714192160165e-06, "loss": 0.0292, "step": 35285 }, { "epoch": 1.65, "learning_rate": 3.492375339133689e-06, "loss": 0.0693, "step": 35290 }, { "epoch": 1.65, "learning_rate": 3.4900364861072132e-06, "loss": 0.0323, "step": 35295 }, { "epoch": 1.65, "learning_rate": 3.4876976330807378e-06, "loss": 0.0716, "step": 35300 }, { "epoch": 1.65, "learning_rate": 3.4853587800542615e-06, "loss": 0.0873, "step": 35305 }, { "epoch": 1.65, "learning_rate": 3.483019927027786e-06, "loss": 0.0356, "step": 35310 }, { "epoch": 1.65, "learning_rate": 3.48068107400131e-06, "loss": 0.0164, "step": 35315 }, { "epoch": 1.65, "learning_rate": 3.478342220974834e-06, "loss": 0.0556, "step": 35320 }, { "epoch": 1.65, "learning_rate": 3.4760033679483583e-06, "loss": 0.0626, "step": 35325 }, { "epoch": 1.65, "learning_rate": 3.473664514921883e-06, "loss": 0.0782, "step": 35330 }, { "epoch": 1.65, "learning_rate": 3.4713256618954065e-06, "loss": 0.0742, "step": 35335 }, { "epoch": 1.65, "learning_rate": 3.468986808868931e-06, "loss": 0.0435, "step": 35340 }, { "epoch": 1.65, "learning_rate": 3.4666479558424556e-06, "loss": 0.0301, "step": 35345 }, { "epoch": 1.65, "learning_rate": 3.4643091028159792e-06, "loss": 0.0638, "step": 35350 }, { "epoch": 1.65, "learning_rate": 3.4619702497895038e-06, "loss": 0.04, "step": 35355 }, { "epoch": 1.65, "learning_rate": 3.459631396763028e-06, "loss": 0.0637, "step": 35360 }, { "epoch": 1.65, "learning_rate": 3.457292543736552e-06, "loss": 0.0498, "step": 35365 }, { "epoch": 1.65, "learning_rate": 3.454953690710076e-06, "loss": 0.0581, "step": 35370 }, { "epoch": 1.65, "learning_rate": 3.4526148376836006e-06, "loss": 0.1048, "step": 35375 }, { "epoch": 1.65, "learning_rate": 3.4502759846571243e-06, "loss": 0.0555, "step": 35380 }, { "epoch": 1.66, "learning_rate": 3.447937131630649e-06, "loss": 0.036, "step": 35385 }, { "epoch": 1.66, "learning_rate": 3.445598278604173e-06, "loss": 0.0362, "step": 35390 }, { "epoch": 1.66, "learning_rate": 3.443259425577697e-06, "loss": 0.0613, "step": 35395 }, { "epoch": 1.66, "learning_rate": 3.440920572551221e-06, "loss": 0.089, "step": 35400 }, { "epoch": 1.66, "learning_rate": 3.4385817195247457e-06, "loss": 0.0078, "step": 35405 }, { "epoch": 1.66, "learning_rate": 3.4362428664982693e-06, "loss": 0.0185, "step": 35410 }, { "epoch": 1.66, "learning_rate": 3.433904013471794e-06, "loss": 0.0212, "step": 35415 }, { "epoch": 1.66, "learning_rate": 3.4315651604453175e-06, "loss": 0.0677, "step": 35420 }, { "epoch": 1.66, "learning_rate": 3.429226307418842e-06, "loss": 0.036, "step": 35425 }, { "epoch": 1.66, "learning_rate": 3.426887454392366e-06, "loss": 0.0845, "step": 35430 }, { "epoch": 1.66, "learning_rate": 3.4245486013658903e-06, "loss": 0.1139, "step": 35435 }, { "epoch": 1.66, "learning_rate": 3.4222097483394144e-06, "loss": 0.0896, "step": 35440 }, { "epoch": 1.66, "learning_rate": 3.419870895312939e-06, "loss": 0.068, "step": 35445 }, { "epoch": 1.66, "learning_rate": 3.4175320422864626e-06, "loss": 0.0515, "step": 35450 }, { "epoch": 1.66, "learning_rate": 3.415193189259987e-06, "loss": 0.0819, "step": 35455 }, { "epoch": 1.66, "learning_rate": 3.4128543362335117e-06, "loss": 0.0642, "step": 35460 }, { "epoch": 1.66, "learning_rate": 3.4105154832070353e-06, "loss": 0.0845, "step": 35465 }, { "epoch": 1.66, "learning_rate": 3.40817663018056e-06, "loss": 0.0947, "step": 35470 }, { "epoch": 1.66, "learning_rate": 3.405837777154084e-06, "loss": 0.0537, "step": 35475 }, { "epoch": 1.66, "learning_rate": 3.4034989241276076e-06, "loss": 0.0903, "step": 35480 }, { "epoch": 1.66, "learning_rate": 3.401160071101132e-06, "loss": 0.0105, "step": 35485 }, { "epoch": 1.66, "learning_rate": 3.3988212180746567e-06, "loss": 0.0793, "step": 35490 }, { "epoch": 1.66, "learning_rate": 3.3964823650481804e-06, "loss": 0.0381, "step": 35495 }, { "epoch": 1.66, "learning_rate": 3.394143512021705e-06, "loss": 0.054, "step": 35500 }, { "epoch": 1.66, "learning_rate": 3.391804658995229e-06, "loss": 0.0431, "step": 35505 }, { "epoch": 1.66, "learning_rate": 3.389465805968753e-06, "loss": 0.081, "step": 35510 }, { "epoch": 1.66, "learning_rate": 3.3871269529422772e-06, "loss": 0.0621, "step": 35515 }, { "epoch": 1.66, "learning_rate": 3.3847880999158018e-06, "loss": 0.0285, "step": 35520 }, { "epoch": 1.66, "learning_rate": 3.3824492468893254e-06, "loss": 0.0543, "step": 35525 }, { "epoch": 1.66, "learning_rate": 3.38011039386285e-06, "loss": 0.0644, "step": 35530 }, { "epoch": 1.66, "learning_rate": 3.377771540836374e-06, "loss": 0.0621, "step": 35535 }, { "epoch": 1.66, "learning_rate": 3.375432687809898e-06, "loss": 0.0505, "step": 35540 }, { "epoch": 1.66, "learning_rate": 3.3730938347834223e-06, "loss": 0.0338, "step": 35545 }, { "epoch": 1.66, "learning_rate": 3.370754981756947e-06, "loss": 0.0298, "step": 35550 }, { "epoch": 1.66, "learning_rate": 3.3684161287304705e-06, "loss": 0.0522, "step": 35555 }, { "epoch": 1.66, "learning_rate": 3.366077275703995e-06, "loss": 0.0517, "step": 35560 }, { "epoch": 1.66, "learning_rate": 3.3637384226775195e-06, "loss": 0.0586, "step": 35565 }, { "epoch": 1.66, "learning_rate": 3.3613995696510432e-06, "loss": 0.082, "step": 35570 }, { "epoch": 1.66, "learning_rate": 3.3590607166245678e-06, "loss": 0.0528, "step": 35575 }, { "epoch": 1.66, "learning_rate": 3.356721863598092e-06, "loss": 0.0528, "step": 35580 }, { "epoch": 1.66, "learning_rate": 3.354383010571616e-06, "loss": 0.0293, "step": 35585 }, { "epoch": 1.66, "learning_rate": 3.35204415754514e-06, "loss": 0.0542, "step": 35590 }, { "epoch": 1.67, "learning_rate": 3.3497053045186646e-06, "loss": 0.0838, "step": 35595 }, { "epoch": 1.67, "learning_rate": 3.3473664514921883e-06, "loss": 0.0901, "step": 35600 }, { "epoch": 1.67, "learning_rate": 3.345027598465713e-06, "loss": 0.0436, "step": 35605 }, { "epoch": 1.67, "learning_rate": 3.342688745439237e-06, "loss": 0.0761, "step": 35610 }, { "epoch": 1.67, "learning_rate": 3.340349892412761e-06, "loss": 0.0521, "step": 35615 }, { "epoch": 1.67, "learning_rate": 3.338011039386285e-06, "loss": 0.0187, "step": 35620 }, { "epoch": 1.67, "learning_rate": 3.3356721863598096e-06, "loss": 0.0416, "step": 35625 }, { "epoch": 1.67, "learning_rate": 3.3333333333333333e-06, "loss": 0.0726, "step": 35630 }, { "epoch": 1.67, "learning_rate": 3.330994480306858e-06, "loss": 0.0468, "step": 35635 }, { "epoch": 1.67, "learning_rate": 3.328655627280382e-06, "loss": 0.0932, "step": 35640 }, { "epoch": 1.67, "learning_rate": 3.326316774253906e-06, "loss": 0.037, "step": 35645 }, { "epoch": 1.67, "learning_rate": 3.32397792122743e-06, "loss": 0.0282, "step": 35650 }, { "epoch": 1.67, "learning_rate": 3.3216390682009547e-06, "loss": 0.0724, "step": 35655 }, { "epoch": 1.67, "learning_rate": 3.3193002151744784e-06, "loss": 0.0468, "step": 35660 }, { "epoch": 1.67, "learning_rate": 3.316961362148003e-06, "loss": 0.0312, "step": 35665 }, { "epoch": 1.67, "learning_rate": 3.3146225091215274e-06, "loss": 0.0655, "step": 35670 }, { "epoch": 1.67, "learning_rate": 3.312283656095051e-06, "loss": 0.0454, "step": 35675 }, { "epoch": 1.67, "learning_rate": 3.3099448030685756e-06, "loss": 0.036, "step": 35680 }, { "epoch": 1.67, "learning_rate": 3.3076059500420997e-06, "loss": 0.0504, "step": 35685 }, { "epoch": 1.67, "learning_rate": 3.305267097015624e-06, "loss": 0.0905, "step": 35690 }, { "epoch": 1.67, "learning_rate": 3.302928243989148e-06, "loss": 0.062, "step": 35695 }, { "epoch": 1.67, "learning_rate": 3.3005893909626725e-06, "loss": 0.0134, "step": 35700 }, { "epoch": 1.67, "learning_rate": 3.298250537936196e-06, "loss": 0.0246, "step": 35705 }, { "epoch": 1.67, "learning_rate": 3.2959116849097207e-06, "loss": 0.0598, "step": 35710 }, { "epoch": 1.67, "learning_rate": 3.293572831883245e-06, "loss": 0.1007, "step": 35715 }, { "epoch": 1.67, "learning_rate": 3.291233978856769e-06, "loss": 0.04, "step": 35720 }, { "epoch": 1.67, "learning_rate": 3.288895125830293e-06, "loss": 0.0931, "step": 35725 }, { "epoch": 1.67, "learning_rate": 3.2865562728038175e-06, "loss": 0.0404, "step": 35730 }, { "epoch": 1.67, "learning_rate": 3.284217419777341e-06, "loss": 0.0108, "step": 35735 }, { "epoch": 1.67, "learning_rate": 3.2818785667508657e-06, "loss": 0.0335, "step": 35740 }, { "epoch": 1.67, "learning_rate": 3.2795397137243903e-06, "loss": 0.0528, "step": 35745 }, { "epoch": 1.67, "learning_rate": 3.277200860697914e-06, "loss": 0.0252, "step": 35750 }, { "epoch": 1.67, "learning_rate": 3.274862007671438e-06, "loss": 0.0388, "step": 35755 }, { "epoch": 1.67, "learning_rate": 3.2725231546449626e-06, "loss": 0.0282, "step": 35760 }, { "epoch": 1.67, "learning_rate": 3.2701843016184863e-06, "loss": 0.0441, "step": 35765 }, { "epoch": 1.67, "learning_rate": 3.267845448592011e-06, "loss": 0.0518, "step": 35770 }, { "epoch": 1.67, "learning_rate": 3.2655065955655353e-06, "loss": 0.0285, "step": 35775 }, { "epoch": 1.67, "learning_rate": 3.263167742539059e-06, "loss": 0.036, "step": 35780 }, { "epoch": 1.67, "learning_rate": 3.2608288895125835e-06, "loss": 0.0529, "step": 35785 }, { "epoch": 1.67, "learning_rate": 3.2584900364861076e-06, "loss": 0.0489, "step": 35790 }, { "epoch": 1.67, "learning_rate": 3.2561511834596317e-06, "loss": 0.0361, "step": 35795 }, { "epoch": 1.67, "learning_rate": 3.253812330433156e-06, "loss": 0.0393, "step": 35800 }, { "epoch": 1.67, "learning_rate": 3.2514734774066804e-06, "loss": 0.0399, "step": 35805 }, { "epoch": 1.68, "learning_rate": 3.249134624380204e-06, "loss": 0.0513, "step": 35810 }, { "epoch": 1.68, "learning_rate": 3.2467957713537286e-06, "loss": 0.0208, "step": 35815 }, { "epoch": 1.68, "learning_rate": 3.2444569183272527e-06, "loss": 0.058, "step": 35820 }, { "epoch": 1.68, "learning_rate": 3.2421180653007768e-06, "loss": 0.0627, "step": 35825 }, { "epoch": 1.68, "learning_rate": 3.239779212274301e-06, "loss": 0.0635, "step": 35830 }, { "epoch": 1.68, "learning_rate": 3.2374403592478254e-06, "loss": 0.0488, "step": 35835 }, { "epoch": 1.68, "learning_rate": 3.235101506221349e-06, "loss": 0.076, "step": 35840 }, { "epoch": 1.68, "learning_rate": 3.2327626531948736e-06, "loss": 0.078, "step": 35845 }, { "epoch": 1.68, "learning_rate": 3.2304238001683973e-06, "loss": 0.0581, "step": 35850 }, { "epoch": 1.68, "learning_rate": 3.228084947141922e-06, "loss": 0.0559, "step": 35855 }, { "epoch": 1.68, "learning_rate": 3.225746094115446e-06, "loss": 0.0368, "step": 35860 }, { "epoch": 1.68, "learning_rate": 3.22340724108897e-06, "loss": 0.0348, "step": 35865 }, { "epoch": 1.68, "learning_rate": 3.221068388062494e-06, "loss": 0.0876, "step": 35870 }, { "epoch": 1.68, "learning_rate": 3.2187295350360187e-06, "loss": 0.0292, "step": 35875 }, { "epoch": 1.68, "learning_rate": 3.2163906820095424e-06, "loss": 0.0441, "step": 35880 }, { "epoch": 1.68, "learning_rate": 3.214051828983067e-06, "loss": 0.033, "step": 35885 }, { "epoch": 1.68, "learning_rate": 3.2117129759565914e-06, "loss": 0.0595, "step": 35890 }, { "epoch": 1.68, "learning_rate": 3.209374122930115e-06, "loss": 0.1033, "step": 35895 }, { "epoch": 1.68, "learning_rate": 3.2070352699036396e-06, "loss": 0.0501, "step": 35900 }, { "epoch": 1.68, "learning_rate": 3.2046964168771637e-06, "loss": 0.1049, "step": 35905 }, { "epoch": 1.68, "learning_rate": 3.202357563850688e-06, "loss": 0.0502, "step": 35910 }, { "epoch": 1.68, "learning_rate": 3.200018710824212e-06, "loss": 0.0456, "step": 35915 }, { "epoch": 1.68, "learning_rate": 3.1976798577977365e-06, "loss": 0.0267, "step": 35920 }, { "epoch": 1.68, "learning_rate": 3.19534100477126e-06, "loss": 0.0974, "step": 35925 }, { "epoch": 1.68, "learning_rate": 3.1930021517447847e-06, "loss": 0.0401, "step": 35930 }, { "epoch": 1.68, "learning_rate": 3.1906632987183088e-06, "loss": 0.0234, "step": 35935 }, { "epoch": 1.68, "learning_rate": 3.188324445691833e-06, "loss": 0.0357, "step": 35940 }, { "epoch": 1.68, "learning_rate": 3.185985592665357e-06, "loss": 0.0758, "step": 35945 }, { "epoch": 1.68, "learning_rate": 3.1836467396388815e-06, "loss": 0.0643, "step": 35950 }, { "epoch": 1.68, "learning_rate": 3.181307886612405e-06, "loss": 0.0709, "step": 35955 }, { "epoch": 1.68, "learning_rate": 3.1789690335859297e-06, "loss": 0.0508, "step": 35960 }, { "epoch": 1.68, "learning_rate": 3.1766301805594543e-06, "loss": 0.0425, "step": 35965 }, { "epoch": 1.68, "learning_rate": 3.174291327532978e-06, "loss": 0.0928, "step": 35970 }, { "epoch": 1.68, "learning_rate": 3.171952474506502e-06, "loss": 0.069, "step": 35975 }, { "epoch": 1.68, "learning_rate": 3.1696136214800266e-06, "loss": 0.0429, "step": 35980 }, { "epoch": 1.68, "learning_rate": 3.1672747684535502e-06, "loss": 0.1265, "step": 35985 }, { "epoch": 1.68, "learning_rate": 3.1649359154270748e-06, "loss": 0.0417, "step": 35990 }, { "epoch": 1.68, "learning_rate": 3.1625970624005993e-06, "loss": 0.1079, "step": 35995 }, { "epoch": 1.68, "learning_rate": 3.160258209374123e-06, "loss": 0.0317, "step": 36000 }, { "epoch": 1.68, "learning_rate": 3.1579193563476475e-06, "loss": 0.0138, "step": 36005 }, { "epoch": 1.68, "learning_rate": 3.1555805033211716e-06, "loss": 0.0287, "step": 36010 }, { "epoch": 1.68, "learning_rate": 3.1532416502946957e-06, "loss": 0.0796, "step": 36015 }, { "epoch": 1.68, "learning_rate": 3.15090279726822e-06, "loss": 0.0597, "step": 36020 }, { "epoch": 1.69, "learning_rate": 3.1485639442417444e-06, "loss": 0.126, "step": 36025 }, { "epoch": 1.69, "learning_rate": 3.146225091215268e-06, "loss": 0.0737, "step": 36030 }, { "epoch": 1.69, "learning_rate": 3.1438862381887926e-06, "loss": 0.0772, "step": 36035 }, { "epoch": 1.69, "learning_rate": 3.1415473851623167e-06, "loss": 0.0583, "step": 36040 }, { "epoch": 1.69, "learning_rate": 3.1392085321358408e-06, "loss": 0.0552, "step": 36045 }, { "epoch": 1.69, "learning_rate": 3.136869679109365e-06, "loss": 0.0239, "step": 36050 }, { "epoch": 1.69, "learning_rate": 3.1345308260828894e-06, "loss": 0.0398, "step": 36055 }, { "epoch": 1.69, "learning_rate": 3.132191973056413e-06, "loss": 0.0532, "step": 36060 }, { "epoch": 1.69, "learning_rate": 3.1298531200299376e-06, "loss": 0.0552, "step": 36065 }, { "epoch": 1.69, "learning_rate": 3.127514267003462e-06, "loss": 0.0419, "step": 36070 }, { "epoch": 1.69, "learning_rate": 3.125175413976986e-06, "loss": 0.0547, "step": 36075 }, { "epoch": 1.69, "learning_rate": 3.12283656095051e-06, "loss": 0.036, "step": 36080 }, { "epoch": 1.69, "learning_rate": 3.1204977079240345e-06, "loss": 0.0604, "step": 36085 }, { "epoch": 1.69, "learning_rate": 3.118158854897558e-06, "loss": 0.0493, "step": 36090 }, { "epoch": 1.69, "learning_rate": 3.1158200018710827e-06, "loss": 0.0823, "step": 36095 }, { "epoch": 1.69, "learning_rate": 3.113481148844607e-06, "loss": 0.0146, "step": 36100 }, { "epoch": 1.69, "learning_rate": 3.111142295818131e-06, "loss": 0.0527, "step": 36105 }, { "epoch": 1.69, "learning_rate": 3.1088034427916554e-06, "loss": 0.0608, "step": 36110 }, { "epoch": 1.69, "learning_rate": 3.1064645897651795e-06, "loss": 0.0874, "step": 36115 }, { "epoch": 1.69, "learning_rate": 3.1041257367387036e-06, "loss": 0.0261, "step": 36120 }, { "epoch": 1.69, "learning_rate": 3.1017868837122277e-06, "loss": 0.0297, "step": 36125 }, { "epoch": 1.69, "learning_rate": 3.0994480306857522e-06, "loss": 0.0836, "step": 36130 }, { "epoch": 1.69, "learning_rate": 3.097109177659276e-06, "loss": 0.0754, "step": 36135 }, { "epoch": 1.69, "learning_rate": 3.0947703246328004e-06, "loss": 0.0226, "step": 36140 }, { "epoch": 1.69, "learning_rate": 3.0924314716063246e-06, "loss": 0.1219, "step": 36145 }, { "epoch": 1.69, "learning_rate": 3.0900926185798487e-06, "loss": 0.0273, "step": 36150 }, { "epoch": 1.69, "learning_rate": 3.0877537655533728e-06, "loss": 0.0723, "step": 36155 }, { "epoch": 1.69, "learning_rate": 3.0854149125268973e-06, "loss": 0.0512, "step": 36160 }, { "epoch": 1.69, "learning_rate": 3.083076059500421e-06, "loss": 0.0209, "step": 36165 }, { "epoch": 1.69, "learning_rate": 3.0807372064739455e-06, "loss": 0.0662, "step": 36170 }, { "epoch": 1.69, "learning_rate": 3.07839835344747e-06, "loss": 0.0708, "step": 36175 }, { "epoch": 1.69, "learning_rate": 3.0760595004209937e-06, "loss": 0.0722, "step": 36180 }, { "epoch": 1.69, "learning_rate": 3.0737206473945182e-06, "loss": 0.0945, "step": 36185 }, { "epoch": 1.69, "learning_rate": 3.0713817943680423e-06, "loss": 0.0461, "step": 36190 }, { "epoch": 1.69, "learning_rate": 3.069042941341566e-06, "loss": 0.0599, "step": 36195 }, { "epoch": 1.69, "learning_rate": 3.0667040883150905e-06, "loss": 0.0342, "step": 36200 }, { "epoch": 1.69, "learning_rate": 3.064365235288615e-06, "loss": 0.017, "step": 36205 }, { "epoch": 1.69, "learning_rate": 3.0620263822621388e-06, "loss": 0.0532, "step": 36210 }, { "epoch": 1.69, "learning_rate": 3.0596875292356633e-06, "loss": 0.0648, "step": 36215 }, { "epoch": 1.69, "learning_rate": 3.0573486762091874e-06, "loss": 0.0308, "step": 36220 }, { "epoch": 1.69, "learning_rate": 3.0550098231827115e-06, "loss": 0.0167, "step": 36225 }, { "epoch": 1.69, "learning_rate": 3.0526709701562356e-06, "loss": 0.0457, "step": 36230 }, { "epoch": 1.69, "learning_rate": 3.05033211712976e-06, "loss": 0.0549, "step": 36235 }, { "epoch": 1.7, "learning_rate": 3.047993264103284e-06, "loss": 0.0686, "step": 36240 }, { "epoch": 1.7, "learning_rate": 3.0456544110768083e-06, "loss": 0.1081, "step": 36245 }, { "epoch": 1.7, "learning_rate": 3.0433155580503324e-06, "loss": 0.0893, "step": 36250 }, { "epoch": 1.7, "learning_rate": 3.0409767050238565e-06, "loss": 0.0275, "step": 36255 }, { "epoch": 1.7, "learning_rate": 3.0386378519973807e-06, "loss": 0.0475, "step": 36260 }, { "epoch": 1.7, "learning_rate": 3.036298998970905e-06, "loss": 0.0836, "step": 36265 }, { "epoch": 1.7, "learning_rate": 3.033960145944429e-06, "loss": 0.0312, "step": 36270 }, { "epoch": 1.7, "learning_rate": 3.0316212929179534e-06, "loss": 0.0576, "step": 36275 }, { "epoch": 1.7, "learning_rate": 3.029282439891478e-06, "loss": 0.083, "step": 36280 }, { "epoch": 1.7, "learning_rate": 3.0269435868650016e-06, "loss": 0.0623, "step": 36285 }, { "epoch": 1.7, "learning_rate": 3.024604733838526e-06, "loss": 0.049, "step": 36290 }, { "epoch": 1.7, "learning_rate": 3.02226588081205e-06, "loss": 0.0477, "step": 36295 }, { "epoch": 1.7, "learning_rate": 3.019927027785574e-06, "loss": 0.0831, "step": 36300 }, { "epoch": 1.7, "learning_rate": 3.0175881747590984e-06, "loss": 0.0729, "step": 36305 }, { "epoch": 1.7, "learning_rate": 3.015249321732622e-06, "loss": 0.1028, "step": 36310 }, { "epoch": 1.7, "learning_rate": 3.0129104687061466e-06, "loss": 0.1266, "step": 36315 }, { "epoch": 1.7, "learning_rate": 3.010571615679671e-06, "loss": 0.0403, "step": 36320 }, { "epoch": 1.7, "learning_rate": 3.008232762653195e-06, "loss": 0.0734, "step": 36325 }, { "epoch": 1.7, "learning_rate": 3.0058939096267194e-06, "loss": 0.0145, "step": 36330 }, { "epoch": 1.7, "learning_rate": 3.0035550566002435e-06, "loss": 0.0911, "step": 36335 }, { "epoch": 1.7, "learning_rate": 3.0012162035737676e-06, "loss": 0.0581, "step": 36340 }, { "epoch": 1.7, "learning_rate": 2.9988773505472917e-06, "loss": 0.0117, "step": 36345 }, { "epoch": 1.7, "learning_rate": 2.9965384975208162e-06, "loss": 0.0414, "step": 36350 }, { "epoch": 1.7, "learning_rate": 2.99419964449434e-06, "loss": 0.0304, "step": 36355 }, { "epoch": 1.7, "learning_rate": 2.9918607914678644e-06, "loss": 0.0527, "step": 36360 }, { "epoch": 1.7, "learning_rate": 2.9895219384413885e-06, "loss": 0.015, "step": 36365 }, { "epoch": 1.7, "learning_rate": 2.9871830854149126e-06, "loss": 0.0277, "step": 36370 }, { "epoch": 1.7, "learning_rate": 2.9848442323884367e-06, "loss": 0.084, "step": 36375 }, { "epoch": 1.7, "learning_rate": 2.9825053793619613e-06, "loss": 0.0132, "step": 36380 }, { "epoch": 1.7, "learning_rate": 2.980166526335485e-06, "loss": 0.0732, "step": 36385 }, { "epoch": 1.7, "learning_rate": 2.9778276733090095e-06, "loss": 0.078, "step": 36390 }, { "epoch": 1.7, "learning_rate": 2.975488820282534e-06, "loss": 0.1021, "step": 36395 }, { "epoch": 1.7, "learning_rate": 2.9731499672560577e-06, "loss": 0.0828, "step": 36400 }, { "epoch": 1.7, "learning_rate": 2.9708111142295822e-06, "loss": 0.0428, "step": 36405 }, { "epoch": 1.7, "learning_rate": 2.9684722612031063e-06, "loss": 0.0414, "step": 36410 }, { "epoch": 1.7, "learning_rate": 2.96613340817663e-06, "loss": 0.0306, "step": 36415 }, { "epoch": 1.7, "learning_rate": 2.9637945551501545e-06, "loss": 0.0635, "step": 36420 }, { "epoch": 1.7, "learning_rate": 2.961455702123679e-06, "loss": 0.0621, "step": 36425 }, { "epoch": 1.7, "learning_rate": 2.9591168490972027e-06, "loss": 0.0813, "step": 36430 }, { "epoch": 1.7, "learning_rate": 2.9567779960707273e-06, "loss": 0.0737, "step": 36435 }, { "epoch": 1.7, "learning_rate": 2.9544391430442514e-06, "loss": 0.0577, "step": 36440 }, { "epoch": 1.7, "learning_rate": 2.9521002900177755e-06, "loss": 0.0631, "step": 36445 }, { "epoch": 1.71, "learning_rate": 2.9497614369912996e-06, "loss": 0.0376, "step": 36450 }, { "epoch": 1.71, "learning_rate": 2.947422583964824e-06, "loss": 0.0436, "step": 36455 }, { "epoch": 1.71, "learning_rate": 2.945083730938348e-06, "loss": 0.0612, "step": 36460 }, { "epoch": 1.71, "learning_rate": 2.9427448779118723e-06, "loss": 0.0687, "step": 36465 }, { "epoch": 1.71, "learning_rate": 2.9404060248853964e-06, "loss": 0.0889, "step": 36470 }, { "epoch": 1.71, "learning_rate": 2.9380671718589205e-06, "loss": 0.0604, "step": 36475 }, { "epoch": 1.71, "learning_rate": 2.9357283188324446e-06, "loss": 0.0205, "step": 36480 }, { "epoch": 1.71, "learning_rate": 2.933389465805969e-06, "loss": 0.0598, "step": 36485 }, { "epoch": 1.71, "learning_rate": 2.931050612779493e-06, "loss": 0.0178, "step": 36490 }, { "epoch": 1.71, "learning_rate": 2.9287117597530174e-06, "loss": 0.0582, "step": 36495 }, { "epoch": 1.71, "learning_rate": 2.926372906726542e-06, "loss": 0.04, "step": 36500 }, { "epoch": 1.71, "learning_rate": 2.9240340537000656e-06, "loss": 0.0388, "step": 36505 }, { "epoch": 1.71, "learning_rate": 2.92169520067359e-06, "loss": 0.0561, "step": 36510 }, { "epoch": 1.71, "learning_rate": 2.9193563476471142e-06, "loss": 0.0027, "step": 36515 }, { "epoch": 1.71, "learning_rate": 2.917017494620638e-06, "loss": 0.0047, "step": 36520 }, { "epoch": 1.71, "learning_rate": 2.9146786415941624e-06, "loss": 0.0512, "step": 36525 }, { "epoch": 1.71, "learning_rate": 2.912339788567687e-06, "loss": 0.08, "step": 36530 }, { "epoch": 1.71, "learning_rate": 2.9100009355412106e-06, "loss": 0.1021, "step": 36535 }, { "epoch": 1.71, "learning_rate": 2.907662082514735e-06, "loss": 0.0421, "step": 36540 }, { "epoch": 1.71, "learning_rate": 2.9053232294882593e-06, "loss": 0.0754, "step": 36545 }, { "epoch": 1.71, "learning_rate": 2.9029843764617834e-06, "loss": 0.0962, "step": 36550 }, { "epoch": 1.71, "learning_rate": 2.9006455234353075e-06, "loss": 0.0268, "step": 36555 }, { "epoch": 1.71, "learning_rate": 2.898306670408832e-06, "loss": 0.0781, "step": 36560 }, { "epoch": 1.71, "learning_rate": 2.8959678173823557e-06, "loss": 0.0675, "step": 36565 }, { "epoch": 1.71, "learning_rate": 2.89362896435588e-06, "loss": 0.063, "step": 36570 }, { "epoch": 1.71, "learning_rate": 2.8912901113294043e-06, "loss": 0.0373, "step": 36575 }, { "epoch": 1.71, "learning_rate": 2.8889512583029284e-06, "loss": 0.0705, "step": 36580 }, { "epoch": 1.71, "learning_rate": 2.8866124052764525e-06, "loss": 0.0587, "step": 36585 }, { "epoch": 1.71, "learning_rate": 2.884273552249977e-06, "loss": 0.0604, "step": 36590 }, { "epoch": 1.71, "learning_rate": 2.8819346992235007e-06, "loss": 0.0352, "step": 36595 }, { "epoch": 1.71, "learning_rate": 2.8795958461970253e-06, "loss": 0.0421, "step": 36600 }, { "epoch": 1.71, "learning_rate": 2.8772569931705498e-06, "loss": 0.0365, "step": 36605 }, { "epoch": 1.71, "learning_rate": 2.8749181401440735e-06, "loss": 0.0314, "step": 36610 }, { "epoch": 1.71, "learning_rate": 2.872579287117598e-06, "loss": 0.0393, "step": 36615 }, { "epoch": 1.71, "learning_rate": 2.870240434091122e-06, "loss": 0.061, "step": 36620 }, { "epoch": 1.71, "learning_rate": 2.867901581064646e-06, "loss": 0.0792, "step": 36625 }, { "epoch": 1.71, "learning_rate": 2.8655627280381703e-06, "loss": 0.0642, "step": 36630 }, { "epoch": 1.71, "learning_rate": 2.863223875011695e-06, "loss": 0.0618, "step": 36635 }, { "epoch": 1.71, "learning_rate": 2.8608850219852185e-06, "loss": 0.0416, "step": 36640 }, { "epoch": 1.71, "learning_rate": 2.858546168958743e-06, "loss": 0.0703, "step": 36645 }, { "epoch": 1.71, "learning_rate": 2.856207315932267e-06, "loss": 0.0217, "step": 36650 }, { "epoch": 1.71, "learning_rate": 2.8538684629057913e-06, "loss": 0.0614, "step": 36655 }, { "epoch": 1.71, "learning_rate": 2.8515296098793154e-06, "loss": 0.0179, "step": 36660 }, { "epoch": 1.72, "learning_rate": 2.84919075685284e-06, "loss": 0.0544, "step": 36665 }, { "epoch": 1.72, "learning_rate": 2.8468519038263636e-06, "loss": 0.0066, "step": 36670 }, { "epoch": 1.72, "learning_rate": 2.844513050799888e-06, "loss": 0.021, "step": 36675 }, { "epoch": 1.72, "learning_rate": 2.842174197773412e-06, "loss": 0.0369, "step": 36680 }, { "epoch": 1.72, "learning_rate": 2.8398353447469363e-06, "loss": 0.0784, "step": 36685 }, { "epoch": 1.72, "learning_rate": 2.8374964917204604e-06, "loss": 0.0393, "step": 36690 }, { "epoch": 1.72, "learning_rate": 2.835157638693985e-06, "loss": 0.0345, "step": 36695 }, { "epoch": 1.72, "learning_rate": 2.8328187856675086e-06, "loss": 0.0407, "step": 36700 }, { "epoch": 1.72, "learning_rate": 2.830479932641033e-06, "loss": 0.0471, "step": 36705 }, { "epoch": 1.72, "learning_rate": 2.8281410796145577e-06, "loss": 0.0371, "step": 36710 }, { "epoch": 1.72, "learning_rate": 2.8258022265880814e-06, "loss": 0.0522, "step": 36715 }, { "epoch": 1.72, "learning_rate": 2.823463373561606e-06, "loss": 0.0389, "step": 36720 }, { "epoch": 1.72, "learning_rate": 2.8211245205351296e-06, "loss": 0.0246, "step": 36725 }, { "epoch": 1.72, "learning_rate": 2.818785667508654e-06, "loss": 0.0493, "step": 36730 }, { "epoch": 1.72, "learning_rate": 2.816446814482178e-06, "loss": 0.0709, "step": 36735 }, { "epoch": 1.72, "learning_rate": 2.814107961455702e-06, "loss": 0.0448, "step": 36740 }, { "epoch": 1.72, "learning_rate": 2.8117691084292264e-06, "loss": 0.0489, "step": 36745 }, { "epoch": 1.72, "learning_rate": 2.809430255402751e-06, "loss": 0.0263, "step": 36750 }, { "epoch": 1.72, "learning_rate": 2.8070914023762746e-06, "loss": 0.0509, "step": 36755 }, { "epoch": 1.72, "learning_rate": 2.804752549349799e-06, "loss": 0.022, "step": 36760 }, { "epoch": 1.72, "learning_rate": 2.8024136963233232e-06, "loss": 0.0713, "step": 36765 }, { "epoch": 1.72, "learning_rate": 2.8000748432968474e-06, "loss": 0.0413, "step": 36770 }, { "epoch": 1.72, "learning_rate": 2.7977359902703715e-06, "loss": 0.0399, "step": 36775 }, { "epoch": 1.72, "learning_rate": 2.795397137243896e-06, "loss": 0.0662, "step": 36780 }, { "epoch": 1.72, "learning_rate": 2.7930582842174197e-06, "loss": 0.0657, "step": 36785 }, { "epoch": 1.72, "learning_rate": 2.790719431190944e-06, "loss": 0.009, "step": 36790 }, { "epoch": 1.72, "learning_rate": 2.7883805781644683e-06, "loss": 0.0343, "step": 36795 }, { "epoch": 1.72, "learning_rate": 2.7860417251379924e-06, "loss": 0.0907, "step": 36800 }, { "epoch": 1.72, "learning_rate": 2.7837028721115165e-06, "loss": 0.1045, "step": 36805 }, { "epoch": 1.72, "learning_rate": 2.781364019085041e-06, "loss": 0.095, "step": 36810 }, { "epoch": 1.72, "learning_rate": 2.7790251660585647e-06, "loss": 0.0273, "step": 36815 }, { "epoch": 1.72, "learning_rate": 2.7766863130320892e-06, "loss": 0.0505, "step": 36820 }, { "epoch": 1.72, "learning_rate": 2.7743474600056138e-06, "loss": 0.0456, "step": 36825 }, { "epoch": 1.72, "learning_rate": 2.7720086069791375e-06, "loss": 0.0413, "step": 36830 }, { "epoch": 1.72, "learning_rate": 2.769669753952662e-06, "loss": 0.066, "step": 36835 }, { "epoch": 1.72, "learning_rate": 2.767330900926186e-06, "loss": 0.0386, "step": 36840 }, { "epoch": 1.72, "learning_rate": 2.76499204789971e-06, "loss": 0.018, "step": 36845 }, { "epoch": 1.72, "learning_rate": 2.7626531948732343e-06, "loss": 0.026, "step": 36850 }, { "epoch": 1.72, "learning_rate": 2.760314341846759e-06, "loss": 0.0317, "step": 36855 }, { "epoch": 1.72, "learning_rate": 2.7579754888202825e-06, "loss": 0.0967, "step": 36860 }, { "epoch": 1.72, "learning_rate": 2.755636635793807e-06, "loss": 0.0411, "step": 36865 }, { "epoch": 1.72, "learning_rate": 2.753297782767331e-06, "loss": 0.017, "step": 36870 }, { "epoch": 1.72, "learning_rate": 2.7509589297408552e-06, "loss": 0.0328, "step": 36875 }, { "epoch": 1.73, "learning_rate": 2.7486200767143793e-06, "loss": 0.0648, "step": 36880 }, { "epoch": 1.73, "learning_rate": 2.746281223687904e-06, "loss": 0.0562, "step": 36885 }, { "epoch": 1.73, "learning_rate": 2.7439423706614276e-06, "loss": 0.0661, "step": 36890 }, { "epoch": 1.73, "learning_rate": 2.741603517634952e-06, "loss": 0.0909, "step": 36895 }, { "epoch": 1.73, "learning_rate": 2.7392646646084766e-06, "loss": 0.0832, "step": 36900 }, { "epoch": 1.73, "learning_rate": 2.7369258115820003e-06, "loss": 0.052, "step": 36905 }, { "epoch": 1.73, "learning_rate": 2.7345869585555244e-06, "loss": 0.0771, "step": 36910 }, { "epoch": 1.73, "learning_rate": 2.732248105529049e-06, "loss": 0.0826, "step": 36915 }, { "epoch": 1.73, "learning_rate": 2.7299092525025726e-06, "loss": 0.0502, "step": 36920 }, { "epoch": 1.73, "learning_rate": 2.727570399476097e-06, "loss": 0.0362, "step": 36925 }, { "epoch": 1.73, "learning_rate": 2.7252315464496217e-06, "loss": 0.0451, "step": 36930 }, { "epoch": 1.73, "learning_rate": 2.7228926934231453e-06, "loss": 0.07, "step": 36935 }, { "epoch": 1.73, "learning_rate": 2.72055384039667e-06, "loss": 0.015, "step": 36940 }, { "epoch": 1.73, "learning_rate": 2.718214987370194e-06, "loss": 0.1023, "step": 36945 }, { "epoch": 1.73, "learning_rate": 2.715876134343718e-06, "loss": 0.0686, "step": 36950 }, { "epoch": 1.73, "learning_rate": 2.713537281317242e-06, "loss": 0.0708, "step": 36955 }, { "epoch": 1.73, "learning_rate": 2.7111984282907667e-06, "loss": 0.0761, "step": 36960 }, { "epoch": 1.73, "learning_rate": 2.7088595752642904e-06, "loss": 0.0483, "step": 36965 }, { "epoch": 1.73, "learning_rate": 2.706520722237815e-06, "loss": 0.0416, "step": 36970 }, { "epoch": 1.73, "learning_rate": 2.704181869211339e-06, "loss": 0.0832, "step": 36975 }, { "epoch": 1.73, "learning_rate": 2.701843016184863e-06, "loss": 0.046, "step": 36980 }, { "epoch": 1.73, "learning_rate": 2.6995041631583872e-06, "loss": 0.0536, "step": 36985 }, { "epoch": 1.73, "learning_rate": 2.6971653101319118e-06, "loss": 0.044, "step": 36990 }, { "epoch": 1.73, "learning_rate": 2.6948264571054354e-06, "loss": 0.0614, "step": 36995 }, { "epoch": 1.73, "learning_rate": 2.69248760407896e-06, "loss": 0.0496, "step": 37000 }, { "epoch": 1.73, "learning_rate": 2.6901487510524845e-06, "loss": 0.0764, "step": 37005 }, { "epoch": 1.73, "learning_rate": 2.687809898026008e-06, "loss": 0.0405, "step": 37010 }, { "epoch": 1.73, "learning_rate": 2.6854710449995323e-06, "loss": 0.0753, "step": 37015 }, { "epoch": 1.73, "learning_rate": 2.683132191973057e-06, "loss": 0.0117, "step": 37020 }, { "epoch": 1.73, "learning_rate": 2.6807933389465805e-06, "loss": 0.0677, "step": 37025 }, { "epoch": 1.73, "learning_rate": 2.678454485920105e-06, "loss": 0.056, "step": 37030 }, { "epoch": 1.73, "learning_rate": 2.6761156328936295e-06, "loss": 0.0837, "step": 37035 }, { "epoch": 1.73, "learning_rate": 2.6737767798671532e-06, "loss": 0.0204, "step": 37040 }, { "epoch": 1.73, "learning_rate": 2.6714379268406778e-06, "loss": 0.0524, "step": 37045 }, { "epoch": 1.73, "learning_rate": 2.669099073814202e-06, "loss": 0.0342, "step": 37050 }, { "epoch": 1.73, "learning_rate": 2.666760220787726e-06, "loss": 0.0927, "step": 37055 }, { "epoch": 1.73, "learning_rate": 2.66442136776125e-06, "loss": 0.0456, "step": 37060 }, { "epoch": 1.73, "learning_rate": 2.6620825147347746e-06, "loss": 0.0318, "step": 37065 }, { "epoch": 1.73, "learning_rate": 2.6597436617082983e-06, "loss": 0.051, "step": 37070 }, { "epoch": 1.73, "learning_rate": 2.657404808681823e-06, "loss": 0.026, "step": 37075 }, { "epoch": 1.73, "learning_rate": 2.655065955655347e-06, "loss": 0.061, "step": 37080 }, { "epoch": 1.73, "learning_rate": 2.652727102628871e-06, "loss": 0.0583, "step": 37085 }, { "epoch": 1.73, "learning_rate": 2.650388249602395e-06, "loss": 0.0321, "step": 37090 }, { "epoch": 1.74, "learning_rate": 2.6480493965759196e-06, "loss": 0.0404, "step": 37095 }, { "epoch": 1.74, "learning_rate": 2.6457105435494433e-06, "loss": 0.0257, "step": 37100 }, { "epoch": 1.74, "learning_rate": 2.643371690522968e-06, "loss": 0.0487, "step": 37105 }, { "epoch": 1.74, "learning_rate": 2.6410328374964924e-06, "loss": 0.0106, "step": 37110 }, { "epoch": 1.74, "learning_rate": 2.638693984470016e-06, "loss": 0.0369, "step": 37115 }, { "epoch": 1.74, "learning_rate": 2.6363551314435406e-06, "loss": 0.0677, "step": 37120 }, { "epoch": 1.74, "learning_rate": 2.6340162784170647e-06, "loss": 0.056, "step": 37125 }, { "epoch": 1.74, "learning_rate": 2.6316774253905884e-06, "loss": 0.0611, "step": 37130 }, { "epoch": 1.74, "learning_rate": 2.629338572364113e-06, "loss": 0.0418, "step": 37135 }, { "epoch": 1.74, "learning_rate": 2.6269997193376374e-06, "loss": 0.0811, "step": 37140 }, { "epoch": 1.74, "learning_rate": 2.624660866311161e-06, "loss": 0.0442, "step": 37145 }, { "epoch": 1.74, "learning_rate": 2.6223220132846856e-06, "loss": 0.0287, "step": 37150 }, { "epoch": 1.74, "learning_rate": 2.6199831602582093e-06, "loss": 0.0286, "step": 37155 }, { "epoch": 1.74, "learning_rate": 2.617644307231734e-06, "loss": 0.1016, "step": 37160 }, { "epoch": 1.74, "learning_rate": 2.615305454205258e-06, "loss": 0.0621, "step": 37165 }, { "epoch": 1.74, "learning_rate": 2.612966601178782e-06, "loss": 0.035, "step": 37170 }, { "epoch": 1.74, "learning_rate": 2.610627748152306e-06, "loss": 0.0603, "step": 37175 }, { "epoch": 1.74, "learning_rate": 2.6082888951258307e-06, "loss": 0.046, "step": 37180 }, { "epoch": 1.74, "learning_rate": 2.6059500420993544e-06, "loss": 0.0195, "step": 37185 }, { "epoch": 1.74, "learning_rate": 2.603611189072879e-06, "loss": 0.1258, "step": 37190 }, { "epoch": 1.74, "learning_rate": 2.601272336046403e-06, "loss": 0.0161, "step": 37195 }, { "epoch": 1.74, "learning_rate": 2.598933483019927e-06, "loss": 0.0736, "step": 37200 }, { "epoch": 1.74, "learning_rate": 2.5965946299934512e-06, "loss": 0.0344, "step": 37205 }, { "epoch": 1.74, "learning_rate": 2.5942557769669757e-06, "loss": 0.0471, "step": 37210 }, { "epoch": 1.74, "learning_rate": 2.5919169239404994e-06, "loss": 0.0422, "step": 37215 }, { "epoch": 1.74, "learning_rate": 2.589578070914024e-06, "loss": 0.0721, "step": 37220 }, { "epoch": 1.74, "learning_rate": 2.5872392178875485e-06, "loss": 0.0669, "step": 37225 }, { "epoch": 1.74, "learning_rate": 2.584900364861072e-06, "loss": 0.0945, "step": 37230 }, { "epoch": 1.74, "learning_rate": 2.5825615118345963e-06, "loss": 0.1361, "step": 37235 }, { "epoch": 1.74, "learning_rate": 2.580222658808121e-06, "loss": 0.0587, "step": 37240 }, { "epoch": 1.74, "learning_rate": 2.5778838057816445e-06, "loss": 0.0277, "step": 37245 }, { "epoch": 1.74, "learning_rate": 2.575544952755169e-06, "loss": 0.0336, "step": 37250 }, { "epoch": 1.74, "learning_rate": 2.5732060997286935e-06, "loss": 0.1024, "step": 37255 }, { "epoch": 1.74, "learning_rate": 2.5708672467022172e-06, "loss": 0.0897, "step": 37260 }, { "epoch": 1.74, "learning_rate": 2.5685283936757417e-06, "loss": 0.0458, "step": 37265 }, { "epoch": 1.74, "learning_rate": 2.566189540649266e-06, "loss": 0.1161, "step": 37270 }, { "epoch": 1.74, "learning_rate": 2.56385068762279e-06, "loss": 0.0177, "step": 37275 }, { "epoch": 1.74, "learning_rate": 2.561511834596314e-06, "loss": 0.0598, "step": 37280 }, { "epoch": 1.74, "learning_rate": 2.5591729815698386e-06, "loss": 0.0731, "step": 37285 }, { "epoch": 1.74, "learning_rate": 2.5568341285433623e-06, "loss": 0.0835, "step": 37290 }, { "epoch": 1.74, "learning_rate": 2.554495275516887e-06, "loss": 0.0516, "step": 37295 }, { "epoch": 1.74, "learning_rate": 2.552156422490411e-06, "loss": 0.0262, "step": 37300 }, { "epoch": 1.75, "learning_rate": 2.549817569463935e-06, "loss": 0.071, "step": 37305 }, { "epoch": 1.75, "learning_rate": 2.547478716437459e-06, "loss": 0.0657, "step": 37310 }, { "epoch": 1.75, "learning_rate": 2.5451398634109836e-06, "loss": 0.0375, "step": 37315 }, { "epoch": 1.75, "learning_rate": 2.5428010103845073e-06, "loss": 0.0301, "step": 37320 }, { "epoch": 1.75, "learning_rate": 2.540462157358032e-06, "loss": 0.0315, "step": 37325 }, { "epoch": 1.75, "learning_rate": 2.5381233043315564e-06, "loss": 0.0763, "step": 37330 }, { "epoch": 1.75, "learning_rate": 2.53578445130508e-06, "loss": 0.0325, "step": 37335 }, { "epoch": 1.75, "learning_rate": 2.5334455982786046e-06, "loss": 0.0746, "step": 37340 }, { "epoch": 1.75, "learning_rate": 2.5311067452521287e-06, "loss": 0.0258, "step": 37345 }, { "epoch": 1.75, "learning_rate": 2.5287678922256524e-06, "loss": 0.023, "step": 37350 }, { "epoch": 1.75, "learning_rate": 2.526429039199177e-06, "loss": 0.0357, "step": 37355 }, { "epoch": 1.75, "learning_rate": 2.5240901861727014e-06, "loss": 0.018, "step": 37360 }, { "epoch": 1.75, "learning_rate": 2.521751333146225e-06, "loss": 0.0354, "step": 37365 }, { "epoch": 1.75, "learning_rate": 2.5194124801197496e-06, "loss": 0.0371, "step": 37370 }, { "epoch": 1.75, "learning_rate": 2.5170736270932737e-06, "loss": 0.0629, "step": 37375 }, { "epoch": 1.75, "learning_rate": 2.514734774066798e-06, "loss": 0.0765, "step": 37380 }, { "epoch": 1.75, "learning_rate": 2.512395921040322e-06, "loss": 0.0563, "step": 37385 }, { "epoch": 1.75, "learning_rate": 2.5100570680138465e-06, "loss": 0.0484, "step": 37390 }, { "epoch": 1.75, "learning_rate": 2.50771821498737e-06, "loss": 0.0275, "step": 37395 }, { "epoch": 1.75, "learning_rate": 2.5053793619608947e-06, "loss": 0.0308, "step": 37400 }, { "epoch": 1.75, "learning_rate": 2.5030405089344188e-06, "loss": 0.0125, "step": 37405 }, { "epoch": 1.75, "learning_rate": 2.500701655907943e-06, "loss": 0.0377, "step": 37410 }, { "epoch": 1.75, "learning_rate": 2.498362802881467e-06, "loss": 0.0332, "step": 37415 }, { "epoch": 1.75, "learning_rate": 2.496023949854991e-06, "loss": 0.0125, "step": 37420 }, { "epoch": 1.75, "learning_rate": 2.4936850968285156e-06, "loss": 0.071, "step": 37425 }, { "epoch": 1.75, "learning_rate": 2.4913462438020397e-06, "loss": 0.0338, "step": 37430 }, { "epoch": 1.75, "learning_rate": 2.489007390775564e-06, "loss": 0.0014, "step": 37435 }, { "epoch": 1.75, "learning_rate": 2.4866685377490884e-06, "loss": 0.0588, "step": 37440 }, { "epoch": 1.75, "learning_rate": 2.4843296847226125e-06, "loss": 0.0582, "step": 37445 }, { "epoch": 1.75, "learning_rate": 2.4819908316961366e-06, "loss": 0.0261, "step": 37450 }, { "epoch": 1.75, "learning_rate": 2.4796519786696607e-06, "loss": 0.0254, "step": 37455 }, { "epoch": 1.75, "learning_rate": 2.4773131256431848e-06, "loss": 0.0256, "step": 37460 }, { "epoch": 1.75, "learning_rate": 2.474974272616709e-06, "loss": 0.0687, "step": 37465 }, { "epoch": 1.75, "learning_rate": 2.4726354195902334e-06, "loss": 0.0346, "step": 37470 }, { "epoch": 1.75, "learning_rate": 2.4702965665637575e-06, "loss": 0.0637, "step": 37475 }, { "epoch": 1.75, "learning_rate": 2.4679577135372816e-06, "loss": 0.0676, "step": 37480 }, { "epoch": 1.75, "learning_rate": 2.4656188605108057e-06, "loss": 0.0251, "step": 37485 }, { "epoch": 1.75, "learning_rate": 2.46328000748433e-06, "loss": 0.0933, "step": 37490 }, { "epoch": 1.75, "learning_rate": 2.460941154457854e-06, "loss": 0.0692, "step": 37495 }, { "epoch": 1.75, "learning_rate": 2.458602301431378e-06, "loss": 0.0574, "step": 37500 }, { "epoch": 1.75, "learning_rate": 2.4562634484049026e-06, "loss": 0.0748, "step": 37505 }, { "epoch": 1.75, "learning_rate": 2.4539245953784267e-06, "loss": 0.0411, "step": 37510 }, { "epoch": 1.75, "learning_rate": 2.4515857423519508e-06, "loss": 0.0219, "step": 37515 }, { "epoch": 1.76, "learning_rate": 2.449246889325475e-06, "loss": 0.0448, "step": 37520 }, { "epoch": 1.76, "learning_rate": 2.446908036298999e-06, "loss": 0.0861, "step": 37525 }, { "epoch": 1.76, "learning_rate": 2.444569183272523e-06, "loss": 0.0921, "step": 37530 }, { "epoch": 1.76, "learning_rate": 2.4422303302460476e-06, "loss": 0.0798, "step": 37535 }, { "epoch": 1.76, "learning_rate": 2.4398914772195717e-06, "loss": 0.073, "step": 37540 }, { "epoch": 1.76, "learning_rate": 2.437552624193096e-06, "loss": 0.0822, "step": 37545 }, { "epoch": 1.76, "learning_rate": 2.4352137711666204e-06, "loss": 0.088, "step": 37550 }, { "epoch": 1.76, "learning_rate": 2.4328749181401445e-06, "loss": 0.0623, "step": 37555 }, { "epoch": 1.76, "learning_rate": 2.4305360651136686e-06, "loss": 0.0234, "step": 37560 }, { "epoch": 1.76, "learning_rate": 2.4281972120871927e-06, "loss": 0.0753, "step": 37565 }, { "epoch": 1.76, "learning_rate": 2.4258583590607168e-06, "loss": 0.0501, "step": 37570 }, { "epoch": 1.76, "learning_rate": 2.423519506034241e-06, "loss": 0.0476, "step": 37575 }, { "epoch": 1.76, "learning_rate": 2.4211806530077654e-06, "loss": 0.0552, "step": 37580 }, { "epoch": 1.76, "learning_rate": 2.4188417999812895e-06, "loss": 0.0375, "step": 37585 }, { "epoch": 1.76, "learning_rate": 2.4165029469548136e-06, "loss": 0.0226, "step": 37590 }, { "epoch": 1.76, "learning_rate": 2.4141640939283377e-06, "loss": 0.0544, "step": 37595 }, { "epoch": 1.76, "learning_rate": 2.411825240901862e-06, "loss": 0.0753, "step": 37600 }, { "epoch": 1.76, "learning_rate": 2.409486387875386e-06, "loss": 0.0753, "step": 37605 }, { "epoch": 1.76, "learning_rate": 2.4071475348489105e-06, "loss": 0.0726, "step": 37610 }, { "epoch": 1.76, "learning_rate": 2.4048086818224346e-06, "loss": 0.031, "step": 37615 }, { "epoch": 1.76, "learning_rate": 2.4024698287959587e-06, "loss": 0.0604, "step": 37620 }, { "epoch": 1.76, "learning_rate": 2.4001309757694828e-06, "loss": 0.0857, "step": 37625 }, { "epoch": 1.76, "learning_rate": 2.397792122743007e-06, "loss": 0.0435, "step": 37630 }, { "epoch": 1.76, "learning_rate": 2.395453269716531e-06, "loss": 0.0217, "step": 37635 }, { "epoch": 1.76, "learning_rate": 2.3931144166900555e-06, "loss": 0.0275, "step": 37640 }, { "epoch": 1.76, "learning_rate": 2.3907755636635796e-06, "loss": 0.0295, "step": 37645 }, { "epoch": 1.76, "learning_rate": 2.3884367106371037e-06, "loss": 0.0612, "step": 37650 }, { "epoch": 1.76, "learning_rate": 2.3860978576106282e-06, "loss": 0.0488, "step": 37655 }, { "epoch": 1.76, "learning_rate": 2.3837590045841523e-06, "loss": 0.1216, "step": 37660 }, { "epoch": 1.76, "learning_rate": 2.3814201515576765e-06, "loss": 0.0198, "step": 37665 }, { "epoch": 1.76, "learning_rate": 2.3790812985312006e-06, "loss": 0.0566, "step": 37670 }, { "epoch": 1.76, "learning_rate": 2.3767424455047247e-06, "loss": 0.0349, "step": 37675 }, { "epoch": 1.76, "learning_rate": 2.3744035924782488e-06, "loss": 0.0575, "step": 37680 }, { "epoch": 1.76, "learning_rate": 2.3720647394517733e-06, "loss": 0.0338, "step": 37685 }, { "epoch": 1.76, "learning_rate": 2.3697258864252974e-06, "loss": 0.0216, "step": 37690 }, { "epoch": 1.76, "learning_rate": 2.3673870333988215e-06, "loss": 0.116, "step": 37695 }, { "epoch": 1.76, "learning_rate": 2.3650481803723456e-06, "loss": 0.0492, "step": 37700 }, { "epoch": 1.76, "learning_rate": 2.3627093273458697e-06, "loss": 0.0183, "step": 37705 }, { "epoch": 1.76, "learning_rate": 2.360370474319394e-06, "loss": 0.0363, "step": 37710 }, { "epoch": 1.76, "learning_rate": 2.358031621292918e-06, "loss": 0.0455, "step": 37715 }, { "epoch": 1.76, "learning_rate": 2.3556927682664424e-06, "loss": 0.0276, "step": 37720 }, { "epoch": 1.76, "learning_rate": 2.3533539152399666e-06, "loss": 0.0289, "step": 37725 }, { "epoch": 1.76, "learning_rate": 2.3510150622134907e-06, "loss": 0.0249, "step": 37730 }, { "epoch": 1.77, "learning_rate": 2.3486762091870148e-06, "loss": 0.0412, "step": 37735 }, { "epoch": 1.77, "learning_rate": 2.346337356160539e-06, "loss": 0.048, "step": 37740 }, { "epoch": 1.77, "learning_rate": 2.343998503134063e-06, "loss": 0.0347, "step": 37745 }, { "epoch": 1.77, "learning_rate": 2.3416596501075875e-06, "loss": 0.0678, "step": 37750 }, { "epoch": 1.77, "learning_rate": 2.3393207970811116e-06, "loss": 0.068, "step": 37755 }, { "epoch": 1.77, "learning_rate": 2.3369819440546357e-06, "loss": 0.0628, "step": 37760 }, { "epoch": 1.77, "learning_rate": 2.3346430910281602e-06, "loss": 0.0297, "step": 37765 }, { "epoch": 1.77, "learning_rate": 2.3323042380016843e-06, "loss": 0.0458, "step": 37770 }, { "epoch": 1.77, "learning_rate": 2.3299653849752084e-06, "loss": 0.0664, "step": 37775 }, { "epoch": 1.77, "learning_rate": 2.3276265319487325e-06, "loss": 0.033, "step": 37780 }, { "epoch": 1.77, "learning_rate": 2.3252876789222567e-06, "loss": 0.0034, "step": 37785 }, { "epoch": 1.77, "learning_rate": 2.3229488258957808e-06, "loss": 0.0253, "step": 37790 }, { "epoch": 1.77, "learning_rate": 2.3206099728693053e-06, "loss": 0.0031, "step": 37795 }, { "epoch": 1.77, "learning_rate": 2.3182711198428294e-06, "loss": 0.0999, "step": 37800 }, { "epoch": 1.77, "learning_rate": 2.3159322668163535e-06, "loss": 0.0492, "step": 37805 }, { "epoch": 1.77, "learning_rate": 2.3135934137898776e-06, "loss": 0.0542, "step": 37810 }, { "epoch": 1.77, "learning_rate": 2.3112545607634017e-06, "loss": 0.0217, "step": 37815 }, { "epoch": 1.77, "learning_rate": 2.308915707736926e-06, "loss": 0.033, "step": 37820 }, { "epoch": 1.77, "learning_rate": 2.3065768547104503e-06, "loss": 0.0807, "step": 37825 }, { "epoch": 1.77, "learning_rate": 2.3042380016839744e-06, "loss": 0.05, "step": 37830 }, { "epoch": 1.77, "learning_rate": 2.3018991486574985e-06, "loss": 0.0351, "step": 37835 }, { "epoch": 1.77, "learning_rate": 2.2995602956310226e-06, "loss": 0.093, "step": 37840 }, { "epoch": 1.77, "learning_rate": 2.2972214426045468e-06, "loss": 0.0772, "step": 37845 }, { "epoch": 1.77, "learning_rate": 2.294882589578071e-06, "loss": 0.0685, "step": 37850 }, { "epoch": 1.77, "learning_rate": 2.2925437365515954e-06, "loss": 0.0702, "step": 37855 }, { "epoch": 1.77, "learning_rate": 2.2902048835251195e-06, "loss": 0.0511, "step": 37860 }, { "epoch": 1.77, "learning_rate": 2.2878660304986436e-06, "loss": 0.0415, "step": 37865 }, { "epoch": 1.77, "learning_rate": 2.285527177472168e-06, "loss": 0.0793, "step": 37870 }, { "epoch": 1.77, "learning_rate": 2.2831883244456922e-06, "loss": 0.0548, "step": 37875 }, { "epoch": 1.77, "learning_rate": 2.2808494714192163e-06, "loss": 0.0724, "step": 37880 }, { "epoch": 1.77, "learning_rate": 2.2785106183927404e-06, "loss": 0.0698, "step": 37885 }, { "epoch": 1.77, "learning_rate": 2.2761717653662645e-06, "loss": 0.0638, "step": 37890 }, { "epoch": 1.77, "learning_rate": 2.2738329123397886e-06, "loss": 0.0457, "step": 37895 }, { "epoch": 1.77, "learning_rate": 2.271494059313313e-06, "loss": 0.0971, "step": 37900 }, { "epoch": 1.77, "learning_rate": 2.2691552062868373e-06, "loss": 0.0516, "step": 37905 }, { "epoch": 1.77, "learning_rate": 2.2668163532603614e-06, "loss": 0.1206, "step": 37910 }, { "epoch": 1.77, "learning_rate": 2.2644775002338855e-06, "loss": 0.065, "step": 37915 }, { "epoch": 1.77, "learning_rate": 2.2621386472074096e-06, "loss": 0.0192, "step": 37920 }, { "epoch": 1.77, "learning_rate": 2.2597997941809337e-06, "loss": 0.0441, "step": 37925 }, { "epoch": 1.77, "learning_rate": 2.257460941154458e-06, "loss": 0.0304, "step": 37930 }, { "epoch": 1.77, "learning_rate": 2.2551220881279823e-06, "loss": 0.0267, "step": 37935 }, { "epoch": 1.77, "learning_rate": 2.2527832351015064e-06, "loss": 0.047, "step": 37940 }, { "epoch": 1.77, "learning_rate": 2.2504443820750305e-06, "loss": 0.0941, "step": 37945 }, { "epoch": 1.78, "learning_rate": 2.2481055290485546e-06, "loss": 0.0955, "step": 37950 }, { "epoch": 1.78, "learning_rate": 2.2457666760220787e-06, "loss": 0.0157, "step": 37955 }, { "epoch": 1.78, "learning_rate": 2.243427822995603e-06, "loss": 0.0686, "step": 37960 }, { "epoch": 1.78, "learning_rate": 2.2410889699691274e-06, "loss": 0.0273, "step": 37965 }, { "epoch": 1.78, "learning_rate": 2.2387501169426515e-06, "loss": 0.0389, "step": 37970 }, { "epoch": 1.78, "learning_rate": 2.2364112639161756e-06, "loss": 0.1099, "step": 37975 }, { "epoch": 1.78, "learning_rate": 2.2340724108897e-06, "loss": 0.0411, "step": 37980 }, { "epoch": 1.78, "learning_rate": 2.2317335578632242e-06, "loss": 0.0525, "step": 37985 }, { "epoch": 1.78, "learning_rate": 2.2293947048367483e-06, "loss": 0.0434, "step": 37990 }, { "epoch": 1.78, "learning_rate": 2.2270558518102724e-06, "loss": 0.0377, "step": 37995 }, { "epoch": 1.78, "learning_rate": 2.2247169987837965e-06, "loss": 0.0682, "step": 38000 }, { "epoch": 1.78, "learning_rate": 2.2223781457573206e-06, "loss": 0.0616, "step": 38005 }, { "epoch": 1.78, "learning_rate": 2.220039292730845e-06, "loss": 0.0505, "step": 38010 }, { "epoch": 1.78, "learning_rate": 2.2177004397043693e-06, "loss": 0.0592, "step": 38015 }, { "epoch": 1.78, "learning_rate": 2.2153615866778934e-06, "loss": 0.032, "step": 38020 }, { "epoch": 1.78, "learning_rate": 2.2130227336514175e-06, "loss": 0.0321, "step": 38025 }, { "epoch": 1.78, "learning_rate": 2.2106838806249416e-06, "loss": 0.0431, "step": 38030 }, { "epoch": 1.78, "learning_rate": 2.2083450275984657e-06, "loss": 0.0547, "step": 38035 }, { "epoch": 1.78, "learning_rate": 2.2060061745719902e-06, "loss": 0.0408, "step": 38040 }, { "epoch": 1.78, "learning_rate": 2.2036673215455143e-06, "loss": 0.051, "step": 38045 }, { "epoch": 1.78, "learning_rate": 2.2013284685190384e-06, "loss": 0.0624, "step": 38050 }, { "epoch": 1.78, "learning_rate": 2.1989896154925625e-06, "loss": 0.0375, "step": 38055 }, { "epoch": 1.78, "learning_rate": 2.1966507624660866e-06, "loss": 0.058, "step": 38060 }, { "epoch": 1.78, "learning_rate": 2.1943119094396107e-06, "loss": 0.0486, "step": 38065 }, { "epoch": 1.78, "learning_rate": 2.1919730564131353e-06, "loss": 0.0329, "step": 38070 }, { "epoch": 1.78, "learning_rate": 2.1896342033866594e-06, "loss": 0.0385, "step": 38075 }, { "epoch": 1.78, "learning_rate": 2.1872953503601835e-06, "loss": 0.0653, "step": 38080 }, { "epoch": 1.78, "learning_rate": 2.184956497333708e-06, "loss": 0.0229, "step": 38085 }, { "epoch": 1.78, "learning_rate": 2.182617644307232e-06, "loss": 0.0561, "step": 38090 }, { "epoch": 1.78, "learning_rate": 2.180278791280756e-06, "loss": 0.0337, "step": 38095 }, { "epoch": 1.78, "learning_rate": 2.1779399382542803e-06, "loss": 0.0213, "step": 38100 }, { "epoch": 1.78, "learning_rate": 2.1756010852278044e-06, "loss": 0.0387, "step": 38105 }, { "epoch": 1.78, "learning_rate": 2.1732622322013285e-06, "loss": 0.0229, "step": 38110 }, { "epoch": 1.78, "learning_rate": 2.170923379174853e-06, "loss": 0.0778, "step": 38115 }, { "epoch": 1.78, "learning_rate": 2.168584526148377e-06, "loss": 0.0722, "step": 38120 }, { "epoch": 1.78, "learning_rate": 2.1662456731219013e-06, "loss": 0.0781, "step": 38125 }, { "epoch": 1.78, "learning_rate": 2.1639068200954254e-06, "loss": 0.0532, "step": 38130 }, { "epoch": 1.78, "learning_rate": 2.1615679670689495e-06, "loss": 0.0517, "step": 38135 }, { "epoch": 1.78, "learning_rate": 2.1592291140424736e-06, "loss": 0.0843, "step": 38140 }, { "epoch": 1.78, "learning_rate": 2.1568902610159977e-06, "loss": 0.0426, "step": 38145 }, { "epoch": 1.78, "learning_rate": 2.154551407989522e-06, "loss": 0.0304, "step": 38150 }, { "epoch": 1.78, "learning_rate": 2.1522125549630463e-06, "loss": 0.0378, "step": 38155 }, { "epoch": 1.78, "learning_rate": 2.1498737019365704e-06, "loss": 0.0155, "step": 38160 }, { "epoch": 1.79, "learning_rate": 2.1475348489100945e-06, "loss": 0.0108, "step": 38165 }, { "epoch": 1.79, "learning_rate": 2.1451959958836186e-06, "loss": 0.0558, "step": 38170 }, { "epoch": 1.79, "learning_rate": 2.1428571428571427e-06, "loss": 0.0262, "step": 38175 }, { "epoch": 1.79, "learning_rate": 2.1405182898306673e-06, "loss": 0.0457, "step": 38180 }, { "epoch": 1.79, "learning_rate": 2.1381794368041914e-06, "loss": 0.068, "step": 38185 }, { "epoch": 1.79, "learning_rate": 2.1358405837777155e-06, "loss": 0.0486, "step": 38190 }, { "epoch": 1.79, "learning_rate": 2.13350173075124e-06, "loss": 0.0534, "step": 38195 }, { "epoch": 1.79, "learning_rate": 2.131162877724764e-06, "loss": 0.067, "step": 38200 }, { "epoch": 1.79, "learning_rate": 2.128824024698288e-06, "loss": 0.0372, "step": 38205 }, { "epoch": 1.79, "learning_rate": 2.1264851716718123e-06, "loss": 0.067, "step": 38210 }, { "epoch": 1.79, "learning_rate": 2.1241463186453364e-06, "loss": 0.03, "step": 38215 }, { "epoch": 1.79, "learning_rate": 2.1218074656188605e-06, "loss": 0.0285, "step": 38220 }, { "epoch": 1.79, "learning_rate": 2.119468612592385e-06, "loss": 0.0651, "step": 38225 }, { "epoch": 1.79, "learning_rate": 2.117129759565909e-06, "loss": 0.0422, "step": 38230 }, { "epoch": 1.79, "learning_rate": 2.1147909065394333e-06, "loss": 0.12, "step": 38235 }, { "epoch": 1.79, "learning_rate": 2.1124520535129574e-06, "loss": 0.0736, "step": 38240 }, { "epoch": 1.79, "learning_rate": 2.1101132004864815e-06, "loss": 0.0487, "step": 38245 }, { "epoch": 1.79, "learning_rate": 2.1077743474600056e-06, "loss": 0.0912, "step": 38250 }, { "epoch": 1.79, "learning_rate": 2.10543549443353e-06, "loss": 0.033, "step": 38255 }, { "epoch": 1.79, "learning_rate": 2.103096641407054e-06, "loss": 0.1096, "step": 38260 }, { "epoch": 1.79, "learning_rate": 2.1007577883805783e-06, "loss": 0.038, "step": 38265 }, { "epoch": 1.79, "learning_rate": 2.098418935354103e-06, "loss": 0.061, "step": 38270 }, { "epoch": 1.79, "learning_rate": 2.0960800823276265e-06, "loss": 0.0104, "step": 38275 }, { "epoch": 1.79, "learning_rate": 2.0937412293011506e-06, "loss": 0.0594, "step": 38280 }, { "epoch": 1.79, "learning_rate": 2.091402376274675e-06, "loss": 0.0427, "step": 38285 }, { "epoch": 1.79, "learning_rate": 2.0890635232481993e-06, "loss": 0.0483, "step": 38290 }, { "epoch": 1.79, "learning_rate": 2.0867246702217234e-06, "loss": 0.0291, "step": 38295 }, { "epoch": 1.79, "learning_rate": 2.084385817195248e-06, "loss": 0.0239, "step": 38300 }, { "epoch": 1.79, "learning_rate": 2.082046964168772e-06, "loss": 0.046, "step": 38305 }, { "epoch": 1.79, "learning_rate": 2.079708111142296e-06, "loss": 0.0147, "step": 38310 }, { "epoch": 1.79, "learning_rate": 2.07736925811582e-06, "loss": 0.0332, "step": 38315 }, { "epoch": 1.79, "learning_rate": 2.0750304050893443e-06, "loss": 0.0792, "step": 38320 }, { "epoch": 1.79, "learning_rate": 2.0726915520628684e-06, "loss": 0.0445, "step": 38325 }, { "epoch": 1.79, "learning_rate": 2.070352699036393e-06, "loss": 0.0572, "step": 38330 }, { "epoch": 1.79, "learning_rate": 2.068013846009917e-06, "loss": 0.0488, "step": 38335 }, { "epoch": 1.79, "learning_rate": 2.065674992983441e-06, "loss": 0.0734, "step": 38340 }, { "epoch": 1.79, "learning_rate": 2.0633361399569652e-06, "loss": 0.0535, "step": 38345 }, { "epoch": 1.79, "learning_rate": 2.0609972869304894e-06, "loss": 0.0557, "step": 38350 }, { "epoch": 1.79, "learning_rate": 2.0586584339040135e-06, "loss": 0.0537, "step": 38355 }, { "epoch": 1.79, "learning_rate": 2.0563195808775376e-06, "loss": 0.0491, "step": 38360 }, { "epoch": 1.79, "learning_rate": 2.053980727851062e-06, "loss": 0.0668, "step": 38365 }, { "epoch": 1.79, "learning_rate": 2.051641874824586e-06, "loss": 0.0535, "step": 38370 }, { "epoch": 1.8, "learning_rate": 2.0493030217981103e-06, "loss": 0.0176, "step": 38375 }, { "epoch": 1.8, "learning_rate": 2.046964168771635e-06, "loss": 0.1218, "step": 38380 }, { "epoch": 1.8, "learning_rate": 2.0446253157451585e-06, "loss": 0.0471, "step": 38385 }, { "epoch": 1.8, "learning_rate": 2.0422864627186826e-06, "loss": 0.0494, "step": 38390 }, { "epoch": 1.8, "learning_rate": 2.039947609692207e-06, "loss": 0.0194, "step": 38395 }, { "epoch": 1.8, "learning_rate": 2.0376087566657312e-06, "loss": 0.0179, "step": 38400 }, { "epoch": 1.8, "learning_rate": 2.0352699036392553e-06, "loss": 0.0813, "step": 38405 }, { "epoch": 1.8, "learning_rate": 2.03293105061278e-06, "loss": 0.0144, "step": 38410 }, { "epoch": 1.8, "learning_rate": 2.030592197586304e-06, "loss": 0.0371, "step": 38415 }, { "epoch": 1.8, "learning_rate": 2.028253344559828e-06, "loss": 0.0496, "step": 38420 }, { "epoch": 1.8, "learning_rate": 2.025914491533352e-06, "loss": 0.1031, "step": 38425 }, { "epoch": 1.8, "learning_rate": 2.0235756385068763e-06, "loss": 0.0432, "step": 38430 }, { "epoch": 1.8, "learning_rate": 2.0212367854804004e-06, "loss": 0.031, "step": 38435 }, { "epoch": 1.8, "learning_rate": 2.018897932453925e-06, "loss": 0.0366, "step": 38440 }, { "epoch": 1.8, "learning_rate": 2.016559079427449e-06, "loss": 0.0095, "step": 38445 }, { "epoch": 1.8, "learning_rate": 2.014220226400973e-06, "loss": 0.0445, "step": 38450 }, { "epoch": 1.8, "learning_rate": 2.0118813733744972e-06, "loss": 0.0577, "step": 38455 }, { "epoch": 1.8, "learning_rate": 2.0095425203480213e-06, "loss": 0.0387, "step": 38460 }, { "epoch": 1.8, "learning_rate": 2.0072036673215454e-06, "loss": 0.0391, "step": 38465 }, { "epoch": 1.8, "learning_rate": 2.00486481429507e-06, "loss": 0.0733, "step": 38470 }, { "epoch": 1.8, "learning_rate": 2.002525961268594e-06, "loss": 0.0241, "step": 38475 }, { "epoch": 1.8, "learning_rate": 2.000187108242118e-06, "loss": 0.0206, "step": 38480 }, { "epoch": 1.8, "learning_rate": 1.9978482552156427e-06, "loss": 0.0342, "step": 38485 }, { "epoch": 1.8, "learning_rate": 1.995509402189167e-06, "loss": 0.0319, "step": 38490 }, { "epoch": 1.8, "learning_rate": 1.9931705491626905e-06, "loss": 0.0541, "step": 38495 }, { "epoch": 1.8, "learning_rate": 1.990831696136215e-06, "loss": 0.08, "step": 38500 }, { "epoch": 1.8, "learning_rate": 1.988492843109739e-06, "loss": 0.0716, "step": 38505 }, { "epoch": 1.8, "learning_rate": 1.9861539900832632e-06, "loss": 0.0877, "step": 38510 }, { "epoch": 1.8, "learning_rate": 1.9838151370567878e-06, "loss": 0.0688, "step": 38515 }, { "epoch": 1.8, "learning_rate": 1.981476284030312e-06, "loss": 0.0385, "step": 38520 }, { "epoch": 1.8, "learning_rate": 1.979137431003836e-06, "loss": 0.0367, "step": 38525 }, { "epoch": 1.8, "learning_rate": 1.97679857797736e-06, "loss": 0.0804, "step": 38530 }, { "epoch": 1.8, "learning_rate": 1.974459724950884e-06, "loss": 0.0555, "step": 38535 }, { "epoch": 1.8, "learning_rate": 1.9721208719244083e-06, "loss": 0.0087, "step": 38540 }, { "epoch": 1.8, "learning_rate": 1.969782018897933e-06, "loss": 0.0303, "step": 38545 }, { "epoch": 1.8, "learning_rate": 1.967443165871457e-06, "loss": 0.0534, "step": 38550 }, { "epoch": 1.8, "learning_rate": 1.965104312844981e-06, "loss": 0.07, "step": 38555 }, { "epoch": 1.8, "learning_rate": 1.962765459818505e-06, "loss": 0.0457, "step": 38560 }, { "epoch": 1.8, "learning_rate": 1.9604266067920292e-06, "loss": 0.0393, "step": 38565 }, { "epoch": 1.8, "learning_rate": 1.9580877537655533e-06, "loss": 0.0553, "step": 38570 }, { "epoch": 1.8, "learning_rate": 1.9557489007390774e-06, "loss": 0.0738, "step": 38575 }, { "epoch": 1.8, "learning_rate": 1.953410047712602e-06, "loss": 0.0289, "step": 38580 }, { "epoch": 1.8, "learning_rate": 1.951071194686126e-06, "loss": 0.0612, "step": 38585 }, { "epoch": 1.81, "learning_rate": 1.94873234165965e-06, "loss": 0.0356, "step": 38590 }, { "epoch": 1.81, "learning_rate": 1.9463934886331747e-06, "loss": 0.051, "step": 38595 }, { "epoch": 1.81, "learning_rate": 1.944054635606699e-06, "loss": 0.0105, "step": 38600 }, { "epoch": 1.81, "learning_rate": 1.9417157825802225e-06, "loss": 0.0931, "step": 38605 }, { "epoch": 1.81, "learning_rate": 1.939376929553747e-06, "loss": 0.0256, "step": 38610 }, { "epoch": 1.81, "learning_rate": 1.937038076527271e-06, "loss": 0.0239, "step": 38615 }, { "epoch": 1.81, "learning_rate": 1.9346992235007952e-06, "loss": 0.0046, "step": 38620 }, { "epoch": 1.81, "learning_rate": 1.9323603704743198e-06, "loss": 0.0365, "step": 38625 }, { "epoch": 1.81, "learning_rate": 1.930021517447844e-06, "loss": 0.0827, "step": 38630 }, { "epoch": 1.81, "learning_rate": 1.927682664421368e-06, "loss": 0.0188, "step": 38635 }, { "epoch": 1.81, "learning_rate": 1.925343811394892e-06, "loss": 0.0148, "step": 38640 }, { "epoch": 1.81, "learning_rate": 1.923004958368416e-06, "loss": 0.0599, "step": 38645 }, { "epoch": 1.81, "learning_rate": 1.9206661053419403e-06, "loss": 0.0766, "step": 38650 }, { "epoch": 1.81, "learning_rate": 1.918327252315465e-06, "loss": 0.1195, "step": 38655 }, { "epoch": 1.81, "learning_rate": 1.915988399288989e-06, "loss": 0.0896, "step": 38660 }, { "epoch": 1.81, "learning_rate": 1.913649546262513e-06, "loss": 0.0508, "step": 38665 }, { "epoch": 1.81, "learning_rate": 1.911310693236037e-06, "loss": 0.1044, "step": 38670 }, { "epoch": 1.81, "learning_rate": 1.9089718402095612e-06, "loss": 0.0434, "step": 38675 }, { "epoch": 1.81, "learning_rate": 1.9066329871830855e-06, "loss": 0.0369, "step": 38680 }, { "epoch": 1.81, "learning_rate": 1.9042941341566099e-06, "loss": 0.0528, "step": 38685 }, { "epoch": 1.81, "learning_rate": 1.901955281130134e-06, "loss": 0.0279, "step": 38690 }, { "epoch": 1.81, "learning_rate": 1.899616428103658e-06, "loss": 0.0828, "step": 38695 }, { "epoch": 1.81, "learning_rate": 1.8972775750771824e-06, "loss": 0.0983, "step": 38700 }, { "epoch": 1.81, "learning_rate": 1.8949387220507065e-06, "loss": 0.0286, "step": 38705 }, { "epoch": 1.81, "learning_rate": 1.8925998690242306e-06, "loss": 0.0585, "step": 38710 }, { "epoch": 1.81, "learning_rate": 1.890261015997755e-06, "loss": 0.0318, "step": 38715 }, { "epoch": 1.81, "learning_rate": 1.887922162971279e-06, "loss": 0.1023, "step": 38720 }, { "epoch": 1.81, "learning_rate": 1.8855833099448031e-06, "loss": 0.0437, "step": 38725 }, { "epoch": 1.81, "learning_rate": 1.8832444569183274e-06, "loss": 0.062, "step": 38730 }, { "epoch": 1.81, "learning_rate": 1.8809056038918515e-06, "loss": 0.0928, "step": 38735 }, { "epoch": 1.81, "learning_rate": 1.8785667508653756e-06, "loss": 0.0416, "step": 38740 }, { "epoch": 1.81, "learning_rate": 1.8762278978389002e-06, "loss": 0.0443, "step": 38745 }, { "epoch": 1.81, "learning_rate": 1.8738890448124243e-06, "loss": 0.0525, "step": 38750 }, { "epoch": 1.81, "learning_rate": 1.8715501917859482e-06, "loss": 0.0479, "step": 38755 }, { "epoch": 1.81, "learning_rate": 1.8692113387594727e-06, "loss": 0.0483, "step": 38760 }, { "epoch": 1.81, "learning_rate": 1.8668724857329968e-06, "loss": 0.0421, "step": 38765 }, { "epoch": 1.81, "learning_rate": 1.864533632706521e-06, "loss": 0.0303, "step": 38770 }, { "epoch": 1.81, "learning_rate": 1.8621947796800452e-06, "loss": 0.0595, "step": 38775 }, { "epoch": 1.81, "learning_rate": 1.8598559266535693e-06, "loss": 0.0302, "step": 38780 }, { "epoch": 1.81, "learning_rate": 1.8575170736270934e-06, "loss": 0.0179, "step": 38785 }, { "epoch": 1.81, "learning_rate": 1.8551782206006175e-06, "loss": 0.0137, "step": 38790 }, { "epoch": 1.81, "learning_rate": 1.8528393675741418e-06, "loss": 0.0725, "step": 38795 }, { "epoch": 1.81, "learning_rate": 1.850500514547666e-06, "loss": 0.0694, "step": 38800 }, { "epoch": 1.82, "learning_rate": 1.84816166152119e-06, "loss": 0.0246, "step": 38805 }, { "epoch": 1.82, "learning_rate": 1.8458228084947144e-06, "loss": 0.0162, "step": 38810 }, { "epoch": 1.82, "learning_rate": 1.8434839554682385e-06, "loss": 0.081, "step": 38815 }, { "epoch": 1.82, "learning_rate": 1.8411451024417626e-06, "loss": 0.0481, "step": 38820 }, { "epoch": 1.82, "learning_rate": 1.838806249415287e-06, "loss": 0.0385, "step": 38825 }, { "epoch": 1.82, "learning_rate": 1.836467396388811e-06, "loss": 0.0336, "step": 38830 }, { "epoch": 1.82, "learning_rate": 1.8341285433623351e-06, "loss": 0.0305, "step": 38835 }, { "epoch": 1.82, "learning_rate": 1.8317896903358594e-06, "loss": 0.0497, "step": 38840 }, { "epoch": 1.82, "learning_rate": 1.8294508373093835e-06, "loss": 0.0391, "step": 38845 }, { "epoch": 1.82, "learning_rate": 1.8271119842829076e-06, "loss": 0.059, "step": 38850 }, { "epoch": 1.82, "learning_rate": 1.8247731312564322e-06, "loss": 0.0578, "step": 38855 }, { "epoch": 1.82, "learning_rate": 1.8224342782299563e-06, "loss": 0.0094, "step": 38860 }, { "epoch": 1.82, "learning_rate": 1.8200954252034802e-06, "loss": 0.0333, "step": 38865 }, { "epoch": 1.82, "learning_rate": 1.8177565721770047e-06, "loss": 0.0585, "step": 38870 }, { "epoch": 1.82, "learning_rate": 1.8154177191505288e-06, "loss": 0.0263, "step": 38875 }, { "epoch": 1.82, "learning_rate": 1.813078866124053e-06, "loss": 0.034, "step": 38880 }, { "epoch": 1.82, "learning_rate": 1.8107400130975772e-06, "loss": 0.0515, "step": 38885 }, { "epoch": 1.82, "learning_rate": 1.8084011600711013e-06, "loss": 0.04, "step": 38890 }, { "epoch": 1.82, "learning_rate": 1.8060623070446254e-06, "loss": 0.0734, "step": 38895 }, { "epoch": 1.82, "learning_rate": 1.8037234540181497e-06, "loss": 0.0971, "step": 38900 }, { "epoch": 1.82, "learning_rate": 1.8013846009916738e-06, "loss": 0.058, "step": 38905 }, { "epoch": 1.82, "learning_rate": 1.799045747965198e-06, "loss": 0.0632, "step": 38910 }, { "epoch": 1.82, "learning_rate": 1.7967068949387223e-06, "loss": 0.0418, "step": 38915 }, { "epoch": 1.82, "learning_rate": 1.7943680419122464e-06, "loss": 0.0376, "step": 38920 }, { "epoch": 1.82, "learning_rate": 1.7920291888857705e-06, "loss": 0.0669, "step": 38925 }, { "epoch": 1.82, "learning_rate": 1.7896903358592948e-06, "loss": 0.065, "step": 38930 }, { "epoch": 1.82, "learning_rate": 1.7873514828328189e-06, "loss": 0.049, "step": 38935 }, { "epoch": 1.82, "learning_rate": 1.785012629806343e-06, "loss": 0.0505, "step": 38940 }, { "epoch": 1.82, "learning_rate": 1.7826737767798673e-06, "loss": 0.0599, "step": 38945 }, { "epoch": 1.82, "learning_rate": 1.7803349237533914e-06, "loss": 0.0559, "step": 38950 }, { "epoch": 1.82, "learning_rate": 1.7779960707269155e-06, "loss": 0.0613, "step": 38955 }, { "epoch": 1.82, "learning_rate": 1.77565721770044e-06, "loss": 0.0314, "step": 38960 }, { "epoch": 1.82, "learning_rate": 1.7733183646739642e-06, "loss": 0.0682, "step": 38965 }, { "epoch": 1.82, "learning_rate": 1.7709795116474883e-06, "loss": 0.0441, "step": 38970 }, { "epoch": 1.82, "learning_rate": 1.7686406586210126e-06, "loss": 0.0259, "step": 38975 }, { "epoch": 1.82, "learning_rate": 1.7663018055945367e-06, "loss": 0.0844, "step": 38980 }, { "epoch": 1.82, "learning_rate": 1.7639629525680608e-06, "loss": 0.0534, "step": 38985 }, { "epoch": 1.82, "learning_rate": 1.761624099541585e-06, "loss": 0.044, "step": 38990 }, { "epoch": 1.82, "learning_rate": 1.7592852465151092e-06, "loss": 0.0369, "step": 38995 }, { "epoch": 1.82, "learning_rate": 1.7569463934886333e-06, "loss": 0.0142, "step": 39000 }, { "epoch": 1.82, "learning_rate": 1.7546075404621574e-06, "loss": 0.0671, "step": 39005 }, { "epoch": 1.82, "learning_rate": 1.7522686874356817e-06, "loss": 0.0222, "step": 39010 }, { "epoch": 1.82, "learning_rate": 1.7499298344092058e-06, "loss": 0.0346, "step": 39015 }, { "epoch": 1.83, "learning_rate": 1.74759098138273e-06, "loss": 0.0305, "step": 39020 }, { "epoch": 1.83, "learning_rate": 1.7452521283562543e-06, "loss": 0.031, "step": 39025 }, { "epoch": 1.83, "learning_rate": 1.7429132753297784e-06, "loss": 0.0574, "step": 39030 }, { "epoch": 1.83, "learning_rate": 1.7405744223033025e-06, "loss": 0.0387, "step": 39035 }, { "epoch": 1.83, "learning_rate": 1.7382355692768268e-06, "loss": 0.0989, "step": 39040 }, { "epoch": 1.83, "learning_rate": 1.7358967162503509e-06, "loss": 0.0504, "step": 39045 }, { "epoch": 1.83, "learning_rate": 1.733557863223875e-06, "loss": 0.0536, "step": 39050 }, { "epoch": 1.83, "learning_rate": 1.7312190101973993e-06, "loss": 0.0216, "step": 39055 }, { "epoch": 1.83, "learning_rate": 1.7288801571709234e-06, "loss": 0.0331, "step": 39060 }, { "epoch": 1.83, "learning_rate": 1.7265413041444475e-06, "loss": 0.0407, "step": 39065 }, { "epoch": 1.83, "learning_rate": 1.724202451117972e-06, "loss": 0.0383, "step": 39070 }, { "epoch": 1.83, "learning_rate": 1.7218635980914961e-06, "loss": 0.1113, "step": 39075 }, { "epoch": 1.83, "learning_rate": 1.7195247450650202e-06, "loss": 0.0942, "step": 39080 }, { "epoch": 1.83, "learning_rate": 1.7171858920385446e-06, "loss": 0.1094, "step": 39085 }, { "epoch": 1.83, "learning_rate": 1.7148470390120687e-06, "loss": 0.0692, "step": 39090 }, { "epoch": 1.83, "learning_rate": 1.7125081859855928e-06, "loss": 0.0349, "step": 39095 }, { "epoch": 1.83, "learning_rate": 1.710169332959117e-06, "loss": 0.0488, "step": 39100 }, { "epoch": 1.83, "learning_rate": 1.7078304799326412e-06, "loss": 0.0086, "step": 39105 }, { "epoch": 1.83, "learning_rate": 1.7054916269061653e-06, "loss": 0.0108, "step": 39110 }, { "epoch": 1.83, "learning_rate": 1.7031527738796896e-06, "loss": 0.0409, "step": 39115 }, { "epoch": 1.83, "learning_rate": 1.7008139208532137e-06, "loss": 0.0525, "step": 39120 }, { "epoch": 1.83, "learning_rate": 1.6984750678267378e-06, "loss": 0.0255, "step": 39125 }, { "epoch": 1.83, "learning_rate": 1.6961362148002621e-06, "loss": 0.0421, "step": 39130 }, { "epoch": 1.83, "learning_rate": 1.6937973617737862e-06, "loss": 0.0731, "step": 39135 }, { "epoch": 1.83, "learning_rate": 1.6914585087473103e-06, "loss": 0.0302, "step": 39140 }, { "epoch": 1.83, "learning_rate": 1.6891196557208347e-06, "loss": 0.0405, "step": 39145 }, { "epoch": 1.83, "learning_rate": 1.6867808026943588e-06, "loss": 0.083, "step": 39150 }, { "epoch": 1.83, "learning_rate": 1.6844419496678829e-06, "loss": 0.0443, "step": 39155 }, { "epoch": 1.83, "learning_rate": 1.6821030966414074e-06, "loss": 0.0782, "step": 39160 }, { "epoch": 1.83, "learning_rate": 1.6797642436149313e-06, "loss": 0.017, "step": 39165 }, { "epoch": 1.83, "learning_rate": 1.6774253905884554e-06, "loss": 0.0074, "step": 39170 }, { "epoch": 1.83, "learning_rate": 1.67508653756198e-06, "loss": 0.0426, "step": 39175 }, { "epoch": 1.83, "learning_rate": 1.672747684535504e-06, "loss": 0.0671, "step": 39180 }, { "epoch": 1.83, "learning_rate": 1.6704088315090281e-06, "loss": 0.0168, "step": 39185 }, { "epoch": 1.83, "learning_rate": 1.6680699784825525e-06, "loss": 0.0347, "step": 39190 }, { "epoch": 1.83, "learning_rate": 1.6657311254560766e-06, "loss": 0.051, "step": 39195 }, { "epoch": 1.83, "learning_rate": 1.6633922724296007e-06, "loss": 0.0755, "step": 39200 }, { "epoch": 1.83, "learning_rate": 1.661053419403125e-06, "loss": 0.0247, "step": 39205 }, { "epoch": 1.83, "learning_rate": 1.658714566376649e-06, "loss": 0.0597, "step": 39210 }, { "epoch": 1.83, "learning_rate": 1.6563757133501732e-06, "loss": 0.0756, "step": 39215 }, { "epoch": 1.83, "learning_rate": 1.6540368603236973e-06, "loss": 0.016, "step": 39220 }, { "epoch": 1.83, "learning_rate": 1.6516980072972216e-06, "loss": 0.0448, "step": 39225 }, { "epoch": 1.84, "learning_rate": 1.6493591542707457e-06, "loss": 0.0348, "step": 39230 }, { "epoch": 1.84, "learning_rate": 1.6470203012442698e-06, "loss": 0.0189, "step": 39235 }, { "epoch": 1.84, "learning_rate": 1.6446814482177941e-06, "loss": 0.0532, "step": 39240 }, { "epoch": 1.84, "learning_rate": 1.6423425951913182e-06, "loss": 0.0599, "step": 39245 }, { "epoch": 1.84, "learning_rate": 1.6400037421648423e-06, "loss": 0.0568, "step": 39250 }, { "epoch": 1.84, "learning_rate": 1.6376648891383667e-06, "loss": 0.0237, "step": 39255 }, { "epoch": 1.84, "learning_rate": 1.6353260361118908e-06, "loss": 0.0473, "step": 39260 }, { "epoch": 1.84, "learning_rate": 1.6329871830854149e-06, "loss": 0.0209, "step": 39265 }, { "epoch": 1.84, "learning_rate": 1.6306483300589394e-06, "loss": 0.0262, "step": 39270 }, { "epoch": 1.84, "learning_rate": 1.6283094770324633e-06, "loss": 0.0314, "step": 39275 }, { "epoch": 1.84, "learning_rate": 1.6259706240059874e-06, "loss": 0.0261, "step": 39280 }, { "epoch": 1.84, "learning_rate": 1.623631770979512e-06, "loss": 0.0584, "step": 39285 }, { "epoch": 1.84, "learning_rate": 1.621292917953036e-06, "loss": 0.0822, "step": 39290 }, { "epoch": 1.84, "learning_rate": 1.6189540649265601e-06, "loss": 0.0269, "step": 39295 }, { "epoch": 1.84, "learning_rate": 1.6166152119000844e-06, "loss": 0.0489, "step": 39300 }, { "epoch": 1.84, "learning_rate": 1.6142763588736086e-06, "loss": 0.0332, "step": 39305 }, { "epoch": 1.84, "learning_rate": 1.6119375058471327e-06, "loss": 0.0434, "step": 39310 }, { "epoch": 1.84, "learning_rate": 1.609598652820657e-06, "loss": 0.0471, "step": 39315 }, { "epoch": 1.84, "learning_rate": 1.607259799794181e-06, "loss": 0.0446, "step": 39320 }, { "epoch": 1.84, "learning_rate": 1.6049209467677052e-06, "loss": 0.0428, "step": 39325 }, { "epoch": 1.84, "learning_rate": 1.6025820937412295e-06, "loss": 0.0586, "step": 39330 }, { "epoch": 1.84, "learning_rate": 1.6002432407147536e-06, "loss": 0.0789, "step": 39335 }, { "epoch": 1.84, "learning_rate": 1.5979043876882777e-06, "loss": 0.0365, "step": 39340 }, { "epoch": 1.84, "learning_rate": 1.595565534661802e-06, "loss": 0.0514, "step": 39345 }, { "epoch": 1.84, "learning_rate": 1.5932266816353261e-06, "loss": 0.0411, "step": 39350 }, { "epoch": 1.84, "learning_rate": 1.5908878286088502e-06, "loss": 0.0336, "step": 39355 }, { "epoch": 1.84, "learning_rate": 1.5885489755823745e-06, "loss": 0.0322, "step": 39360 }, { "epoch": 1.84, "learning_rate": 1.5862101225558987e-06, "loss": 0.0556, "step": 39365 }, { "epoch": 1.84, "learning_rate": 1.5838712695294228e-06, "loss": 0.057, "step": 39370 }, { "epoch": 1.84, "learning_rate": 1.5815324165029473e-06, "loss": 0.0674, "step": 39375 }, { "epoch": 1.84, "learning_rate": 1.5791935634764714e-06, "loss": 0.0624, "step": 39380 }, { "epoch": 1.84, "learning_rate": 1.5768547104499953e-06, "loss": 0.0395, "step": 39385 }, { "epoch": 1.84, "learning_rate": 1.5745158574235198e-06, "loss": 0.0422, "step": 39390 }, { "epoch": 1.84, "learning_rate": 1.572177004397044e-06, "loss": 0.0552, "step": 39395 }, { "epoch": 1.84, "learning_rate": 1.569838151370568e-06, "loss": 0.0235, "step": 39400 }, { "epoch": 1.84, "learning_rate": 1.5674992983440923e-06, "loss": 0.0526, "step": 39405 }, { "epoch": 1.84, "learning_rate": 1.5651604453176164e-06, "loss": 0.1032, "step": 39410 }, { "epoch": 1.84, "learning_rate": 1.5628215922911405e-06, "loss": 0.0401, "step": 39415 }, { "epoch": 1.84, "learning_rate": 1.5604827392646649e-06, "loss": 0.065, "step": 39420 }, { "epoch": 1.84, "learning_rate": 1.558143886238189e-06, "loss": 0.038, "step": 39425 }, { "epoch": 1.84, "learning_rate": 1.555805033211713e-06, "loss": 0.0643, "step": 39430 }, { "epoch": 1.84, "learning_rate": 1.5534661801852372e-06, "loss": 0.088, "step": 39435 }, { "epoch": 1.84, "learning_rate": 1.5511273271587615e-06, "loss": 0.0193, "step": 39440 }, { "epoch": 1.85, "learning_rate": 1.5487884741322856e-06, "loss": 0.0679, "step": 39445 }, { "epoch": 1.85, "learning_rate": 1.5464496211058097e-06, "loss": 0.034, "step": 39450 }, { "epoch": 1.85, "learning_rate": 1.544110768079334e-06, "loss": 0.0356, "step": 39455 }, { "epoch": 1.85, "learning_rate": 1.5417719150528581e-06, "loss": 0.0289, "step": 39460 }, { "epoch": 1.85, "learning_rate": 1.5394330620263822e-06, "loss": 0.0829, "step": 39465 }, { "epoch": 1.85, "learning_rate": 1.5370942089999065e-06, "loss": 0.069, "step": 39470 }, { "epoch": 1.85, "learning_rate": 1.5347553559734306e-06, "loss": 0.0484, "step": 39475 }, { "epoch": 1.85, "learning_rate": 1.5324165029469547e-06, "loss": 0.0394, "step": 39480 }, { "epoch": 1.85, "learning_rate": 1.5300776499204793e-06, "loss": 0.0354, "step": 39485 }, { "epoch": 1.85, "learning_rate": 1.5277387968940034e-06, "loss": 0.0735, "step": 39490 }, { "epoch": 1.85, "learning_rate": 1.5253999438675273e-06, "loss": 0.0411, "step": 39495 }, { "epoch": 1.85, "learning_rate": 1.5230610908410518e-06, "loss": 0.0252, "step": 39500 }, { "epoch": 1.85, "learning_rate": 1.520722237814576e-06, "loss": 0.0859, "step": 39505 }, { "epoch": 1.85, "learning_rate": 1.5183833847881e-06, "loss": 0.0456, "step": 39510 }, { "epoch": 1.85, "learning_rate": 1.5160445317616243e-06, "loss": 0.0101, "step": 39515 }, { "epoch": 1.85, "learning_rate": 1.5137056787351484e-06, "loss": 0.0237, "step": 39520 }, { "epoch": 1.85, "learning_rate": 1.5113668257086725e-06, "loss": 0.0761, "step": 39525 }, { "epoch": 1.85, "learning_rate": 1.5090279726821969e-06, "loss": 0.0254, "step": 39530 }, { "epoch": 1.85, "learning_rate": 1.506689119655721e-06, "loss": 0.0616, "step": 39535 }, { "epoch": 1.85, "learning_rate": 1.504350266629245e-06, "loss": 0.0478, "step": 39540 }, { "epoch": 1.85, "learning_rate": 1.5020114136027694e-06, "loss": 0.0673, "step": 39545 }, { "epoch": 1.85, "learning_rate": 1.4996725605762935e-06, "loss": 0.0635, "step": 39550 }, { "epoch": 1.85, "learning_rate": 1.4973337075498176e-06, "loss": 0.087, "step": 39555 }, { "epoch": 1.85, "learning_rate": 1.494994854523342e-06, "loss": 0.0767, "step": 39560 }, { "epoch": 1.85, "learning_rate": 1.492656001496866e-06, "loss": 0.0616, "step": 39565 }, { "epoch": 1.85, "learning_rate": 1.4903171484703901e-06, "loss": 0.0461, "step": 39570 }, { "epoch": 1.85, "learning_rate": 1.4879782954439146e-06, "loss": 0.0593, "step": 39575 }, { "epoch": 1.85, "learning_rate": 1.4856394424174385e-06, "loss": 0.0692, "step": 39580 }, { "epoch": 1.85, "learning_rate": 1.4833005893909626e-06, "loss": 0.0365, "step": 39585 }, { "epoch": 1.85, "learning_rate": 1.4809617363644872e-06, "loss": 0.0283, "step": 39590 }, { "epoch": 1.85, "learning_rate": 1.4786228833380113e-06, "loss": 0.0788, "step": 39595 }, { "epoch": 1.85, "learning_rate": 1.4762840303115354e-06, "loss": 0.0264, "step": 39600 }, { "epoch": 1.85, "learning_rate": 1.4739451772850597e-06, "loss": 0.0211, "step": 39605 }, { "epoch": 1.85, "learning_rate": 1.4716063242585838e-06, "loss": 0.0327, "step": 39610 }, { "epoch": 1.85, "learning_rate": 1.469267471232108e-06, "loss": 0.0372, "step": 39615 }, { "epoch": 1.85, "learning_rate": 1.4669286182056322e-06, "loss": 0.0876, "step": 39620 }, { "epoch": 1.85, "learning_rate": 1.4645897651791563e-06, "loss": 0.0218, "step": 39625 }, { "epoch": 1.85, "learning_rate": 1.4622509121526804e-06, "loss": 0.0676, "step": 39630 }, { "epoch": 1.85, "learning_rate": 1.4599120591262047e-06, "loss": 0.0235, "step": 39635 }, { "epoch": 1.85, "learning_rate": 1.4575732060997288e-06, "loss": 0.0842, "step": 39640 }, { "epoch": 1.85, "learning_rate": 1.455234353073253e-06, "loss": 0.0288, "step": 39645 }, { "epoch": 1.85, "learning_rate": 1.452895500046777e-06, "loss": 0.0226, "step": 39650 }, { "epoch": 1.85, "learning_rate": 1.4505566470203014e-06, "loss": 0.0273, "step": 39655 }, { "epoch": 1.86, "learning_rate": 1.4482177939938255e-06, "loss": 0.0529, "step": 39660 }, { "epoch": 1.86, "learning_rate": 1.4458789409673496e-06, "loss": 0.0863, "step": 39665 }, { "epoch": 1.86, "learning_rate": 1.4435400879408739e-06, "loss": 0.0432, "step": 39670 }, { "epoch": 1.86, "learning_rate": 1.441201234914398e-06, "loss": 0.0455, "step": 39675 }, { "epoch": 1.86, "learning_rate": 1.438862381887922e-06, "loss": 0.0248, "step": 39680 }, { "epoch": 1.86, "learning_rate": 1.4365235288614466e-06, "loss": 0.0633, "step": 39685 }, { "epoch": 1.86, "learning_rate": 1.4341846758349705e-06, "loss": 0.0594, "step": 39690 }, { "epoch": 1.86, "learning_rate": 1.4318458228084946e-06, "loss": 0.0619, "step": 39695 }, { "epoch": 1.86, "learning_rate": 1.4295069697820192e-06, "loss": 0.0487, "step": 39700 }, { "epoch": 1.86, "learning_rate": 1.4271681167555433e-06, "loss": 0.0624, "step": 39705 }, { "epoch": 1.86, "learning_rate": 1.4248292637290674e-06, "loss": 0.0463, "step": 39710 }, { "epoch": 1.86, "learning_rate": 1.4224904107025917e-06, "loss": 0.0317, "step": 39715 }, { "epoch": 1.86, "learning_rate": 1.4201515576761158e-06, "loss": 0.043, "step": 39720 }, { "epoch": 1.86, "learning_rate": 1.4178127046496399e-06, "loss": 0.0763, "step": 39725 }, { "epoch": 1.86, "learning_rate": 1.4154738516231642e-06, "loss": 0.0632, "step": 39730 }, { "epoch": 1.86, "learning_rate": 1.4131349985966883e-06, "loss": 0.0565, "step": 39735 }, { "epoch": 1.86, "learning_rate": 1.4107961455702124e-06, "loss": 0.0586, "step": 39740 }, { "epoch": 1.86, "learning_rate": 1.4084572925437367e-06, "loss": 0.0676, "step": 39745 }, { "epoch": 1.86, "learning_rate": 1.4061184395172608e-06, "loss": 0.0902, "step": 39750 }, { "epoch": 1.86, "learning_rate": 1.403779586490785e-06, "loss": 0.0158, "step": 39755 }, { "epoch": 1.86, "learning_rate": 1.4014407334643093e-06, "loss": 0.0502, "step": 39760 }, { "epoch": 1.86, "learning_rate": 1.3991018804378334e-06, "loss": 0.0526, "step": 39765 }, { "epoch": 1.86, "learning_rate": 1.3967630274113575e-06, "loss": 0.0377, "step": 39770 }, { "epoch": 1.86, "learning_rate": 1.3944241743848818e-06, "loss": 0.0454, "step": 39775 }, { "epoch": 1.86, "learning_rate": 1.3920853213584059e-06, "loss": 0.0408, "step": 39780 }, { "epoch": 1.86, "learning_rate": 1.38974646833193e-06, "loss": 0.0673, "step": 39785 }, { "epoch": 1.86, "learning_rate": 1.3874076153054545e-06, "loss": 0.0796, "step": 39790 }, { "epoch": 1.86, "learning_rate": 1.3850687622789786e-06, "loss": 0.0557, "step": 39795 }, { "epoch": 1.86, "learning_rate": 1.3827299092525025e-06, "loss": 0.0464, "step": 39800 }, { "epoch": 1.86, "learning_rate": 1.380391056226027e-06, "loss": 0.0652, "step": 39805 }, { "epoch": 1.86, "learning_rate": 1.3780522031995511e-06, "loss": 0.0413, "step": 39810 }, { "epoch": 1.86, "learning_rate": 1.3757133501730753e-06, "loss": 0.0382, "step": 39815 }, { "epoch": 1.86, "learning_rate": 1.3733744971465996e-06, "loss": 0.0547, "step": 39820 }, { "epoch": 1.86, "learning_rate": 1.3710356441201237e-06, "loss": 0.051, "step": 39825 }, { "epoch": 1.86, "learning_rate": 1.3686967910936478e-06, "loss": 0.0483, "step": 39830 }, { "epoch": 1.86, "learning_rate": 1.366357938067172e-06, "loss": 0.0459, "step": 39835 }, { "epoch": 1.86, "learning_rate": 1.3640190850406962e-06, "loss": 0.0843, "step": 39840 }, { "epoch": 1.86, "learning_rate": 1.3616802320142203e-06, "loss": 0.0416, "step": 39845 }, { "epoch": 1.86, "learning_rate": 1.3593413789877446e-06, "loss": 0.0518, "step": 39850 }, { "epoch": 1.86, "learning_rate": 1.3570025259612687e-06, "loss": 0.0277, "step": 39855 }, { "epoch": 1.86, "learning_rate": 1.3546636729347928e-06, "loss": 0.0828, "step": 39860 }, { "epoch": 1.86, "learning_rate": 1.3523248199083171e-06, "loss": 0.0454, "step": 39865 }, { "epoch": 1.86, "learning_rate": 1.3499859668818412e-06, "loss": 0.0552, "step": 39870 }, { "epoch": 1.87, "learning_rate": 1.3476471138553654e-06, "loss": 0.0452, "step": 39875 }, { "epoch": 1.87, "learning_rate": 1.3453082608288895e-06, "loss": 0.0586, "step": 39880 }, { "epoch": 1.87, "learning_rate": 1.3429694078024138e-06, "loss": 0.053, "step": 39885 }, { "epoch": 1.87, "learning_rate": 1.3406305547759379e-06, "loss": 0.0912, "step": 39890 }, { "epoch": 1.87, "learning_rate": 1.338291701749462e-06, "loss": 0.0716, "step": 39895 }, { "epoch": 1.87, "learning_rate": 1.3359528487229865e-06, "loss": 0.028, "step": 39900 }, { "epoch": 1.87, "learning_rate": 1.3336139956965106e-06, "loss": 0.0424, "step": 39905 }, { "epoch": 1.87, "learning_rate": 1.3312751426700345e-06, "loss": 0.0659, "step": 39910 }, { "epoch": 1.87, "learning_rate": 1.328936289643559e-06, "loss": 0.0222, "step": 39915 }, { "epoch": 1.87, "learning_rate": 1.3265974366170831e-06, "loss": 0.1103, "step": 39920 }, { "epoch": 1.87, "learning_rate": 1.3242585835906072e-06, "loss": 0.0697, "step": 39925 }, { "epoch": 1.87, "learning_rate": 1.3219197305641316e-06, "loss": 0.0705, "step": 39930 }, { "epoch": 1.87, "learning_rate": 1.3195808775376557e-06, "loss": 0.0505, "step": 39935 }, { "epoch": 1.87, "learning_rate": 1.3172420245111798e-06, "loss": 0.0268, "step": 39940 }, { "epoch": 1.87, "learning_rate": 1.314903171484704e-06, "loss": 0.0686, "step": 39945 }, { "epoch": 1.87, "learning_rate": 1.3125643184582282e-06, "loss": 0.0544, "step": 39950 }, { "epoch": 1.87, "learning_rate": 1.3102254654317523e-06, "loss": 0.0541, "step": 39955 }, { "epoch": 1.87, "learning_rate": 1.3078866124052766e-06, "loss": 0.0379, "step": 39960 }, { "epoch": 1.87, "learning_rate": 1.3055477593788007e-06, "loss": 0.045, "step": 39965 }, { "epoch": 1.87, "learning_rate": 1.3032089063523248e-06, "loss": 0.0914, "step": 39970 }, { "epoch": 1.87, "learning_rate": 1.3008700533258491e-06, "loss": 0.1219, "step": 39975 }, { "epoch": 1.87, "learning_rate": 1.2985312002993732e-06, "loss": 0.0432, "step": 39980 }, { "epoch": 1.87, "learning_rate": 1.2961923472728973e-06, "loss": 0.0513, "step": 39985 }, { "epoch": 1.87, "learning_rate": 1.2938534942464217e-06, "loss": 0.0465, "step": 39990 }, { "epoch": 1.87, "learning_rate": 1.2915146412199458e-06, "loss": 0.0539, "step": 39995 }, { "epoch": 1.87, "learning_rate": 1.2891757881934699e-06, "loss": 0.037, "step": 40000 }, { "epoch": 1.87, "learning_rate": 1.2868369351669944e-06, "loss": 0.0655, "step": 40005 }, { "epoch": 1.87, "learning_rate": 1.2844980821405185e-06, "loss": 0.025, "step": 40010 }, { "epoch": 1.87, "learning_rate": 1.2821592291140426e-06, "loss": 0.0285, "step": 40015 }, { "epoch": 1.87, "learning_rate": 1.279820376087567e-06, "loss": 0.0713, "step": 40020 }, { "epoch": 1.87, "learning_rate": 1.277481523061091e-06, "loss": 0.0547, "step": 40025 }, { "epoch": 1.87, "learning_rate": 1.2751426700346151e-06, "loss": 0.0621, "step": 40030 }, { "epoch": 1.87, "learning_rate": 1.2728038170081394e-06, "loss": 0.0305, "step": 40035 }, { "epoch": 1.87, "learning_rate": 1.2704649639816636e-06, "loss": 0.0352, "step": 40040 }, { "epoch": 1.87, "learning_rate": 1.2681261109551877e-06, "loss": 0.054, "step": 40045 }, { "epoch": 1.87, "learning_rate": 1.265787257928712e-06, "loss": 0.0377, "step": 40050 }, { "epoch": 1.87, "learning_rate": 1.263448404902236e-06, "loss": 0.0628, "step": 40055 }, { "epoch": 1.87, "learning_rate": 1.2611095518757602e-06, "loss": 0.0453, "step": 40060 }, { "epoch": 1.87, "learning_rate": 1.2587706988492845e-06, "loss": 0.0916, "step": 40065 }, { "epoch": 1.87, "learning_rate": 1.2564318458228086e-06, "loss": 0.0224, "step": 40070 }, { "epoch": 1.87, "learning_rate": 1.2540929927963327e-06, "loss": 0.0503, "step": 40075 }, { "epoch": 1.87, "learning_rate": 1.251754139769857e-06, "loss": 0.0642, "step": 40080 }, { "epoch": 1.88, "learning_rate": 1.2494152867433811e-06, "loss": 0.0319, "step": 40085 }, { "epoch": 1.88, "learning_rate": 1.2470764337169052e-06, "loss": 0.1147, "step": 40090 }, { "epoch": 1.88, "learning_rate": 1.2447375806904295e-06, "loss": 0.0311, "step": 40095 }, { "epoch": 1.88, "learning_rate": 1.2423987276639537e-06, "loss": 0.0271, "step": 40100 }, { "epoch": 1.88, "learning_rate": 1.2400598746374778e-06, "loss": 0.041, "step": 40105 }, { "epoch": 1.88, "learning_rate": 1.237721021611002e-06, "loss": 0.0292, "step": 40110 }, { "epoch": 1.88, "learning_rate": 1.2353821685845264e-06, "loss": 0.0479, "step": 40115 }, { "epoch": 1.88, "learning_rate": 1.2330433155580505e-06, "loss": 0.0477, "step": 40120 }, { "epoch": 1.88, "learning_rate": 1.2307044625315746e-06, "loss": 0.0309, "step": 40125 }, { "epoch": 1.88, "learning_rate": 1.228365609505099e-06, "loss": 0.0423, "step": 40130 }, { "epoch": 1.88, "learning_rate": 1.226026756478623e-06, "loss": 0.0646, "step": 40135 }, { "epoch": 1.88, "learning_rate": 1.2236879034521471e-06, "loss": 0.0344, "step": 40140 }, { "epoch": 1.88, "learning_rate": 1.2213490504256712e-06, "loss": 0.0289, "step": 40145 }, { "epoch": 1.88, "learning_rate": 1.2190101973991955e-06, "loss": 0.0727, "step": 40150 }, { "epoch": 1.88, "learning_rate": 1.2166713443727196e-06, "loss": 0.0903, "step": 40155 }, { "epoch": 1.88, "learning_rate": 1.2143324913462438e-06, "loss": 0.0904, "step": 40160 }, { "epoch": 1.88, "learning_rate": 1.211993638319768e-06, "loss": 0.0529, "step": 40165 }, { "epoch": 1.88, "learning_rate": 1.2096547852932924e-06, "loss": 0.082, "step": 40170 }, { "epoch": 1.88, "learning_rate": 1.2073159322668165e-06, "loss": 0.0338, "step": 40175 }, { "epoch": 1.88, "learning_rate": 1.2049770792403406e-06, "loss": 0.0239, "step": 40180 }, { "epoch": 1.88, "learning_rate": 1.202638226213865e-06, "loss": 0.0353, "step": 40185 }, { "epoch": 1.88, "learning_rate": 1.200299373187389e-06, "loss": 0.0136, "step": 40190 }, { "epoch": 1.88, "learning_rate": 1.1979605201609131e-06, "loss": 0.0904, "step": 40195 }, { "epoch": 1.88, "learning_rate": 1.1956216671344374e-06, "loss": 0.0365, "step": 40200 }, { "epoch": 1.88, "learning_rate": 1.1932828141079615e-06, "loss": 0.0537, "step": 40205 }, { "epoch": 1.88, "learning_rate": 1.1909439610814856e-06, "loss": 0.0883, "step": 40210 }, { "epoch": 1.88, "learning_rate": 1.18860510805501e-06, "loss": 0.0387, "step": 40215 }, { "epoch": 1.88, "learning_rate": 1.186266255028534e-06, "loss": 0.0506, "step": 40220 }, { "epoch": 1.88, "learning_rate": 1.1839274020020584e-06, "loss": 0.0437, "step": 40225 }, { "epoch": 1.88, "learning_rate": 1.1815885489755825e-06, "loss": 0.0351, "step": 40230 }, { "epoch": 1.88, "learning_rate": 1.1792496959491066e-06, "loss": 0.0931, "step": 40235 }, { "epoch": 1.88, "learning_rate": 1.176910842922631e-06, "loss": 0.0591, "step": 40240 }, { "epoch": 1.88, "learning_rate": 1.174571989896155e-06, "loss": 0.0379, "step": 40245 }, { "epoch": 1.88, "learning_rate": 1.1722331368696791e-06, "loss": 0.0225, "step": 40250 }, { "epoch": 1.88, "learning_rate": 1.1698942838432034e-06, "loss": 0.0441, "step": 40255 }, { "epoch": 1.88, "learning_rate": 1.1675554308167275e-06, "loss": 0.0402, "step": 40260 }, { "epoch": 1.88, "learning_rate": 1.1652165777902516e-06, "loss": 0.0275, "step": 40265 }, { "epoch": 1.88, "learning_rate": 1.162877724763776e-06, "loss": 0.0281, "step": 40270 }, { "epoch": 1.88, "learning_rate": 1.1605388717373e-06, "loss": 0.0372, "step": 40275 }, { "epoch": 1.88, "learning_rate": 1.1582000187108244e-06, "loss": 0.0412, "step": 40280 }, { "epoch": 1.88, "learning_rate": 1.1558611656843485e-06, "loss": 0.0379, "step": 40285 }, { "epoch": 1.88, "learning_rate": 1.1535223126578726e-06, "loss": 0.0623, "step": 40290 }, { "epoch": 1.88, "learning_rate": 1.151183459631397e-06, "loss": 0.0841, "step": 40295 }, { "epoch": 1.89, "learning_rate": 1.148844606604921e-06, "loss": 0.0486, "step": 40300 }, { "epoch": 1.89, "learning_rate": 1.1465057535784451e-06, "loss": 0.0454, "step": 40305 }, { "epoch": 1.89, "learning_rate": 1.1441669005519694e-06, "loss": 0.078, "step": 40310 }, { "epoch": 1.89, "learning_rate": 1.1418280475254937e-06, "loss": 0.0335, "step": 40315 }, { "epoch": 1.89, "learning_rate": 1.1394891944990176e-06, "loss": 0.1026, "step": 40320 }, { "epoch": 1.89, "learning_rate": 1.137150341472542e-06, "loss": 0.1181, "step": 40325 }, { "epoch": 1.89, "learning_rate": 1.1348114884460663e-06, "loss": 0.0812, "step": 40330 }, { "epoch": 1.89, "learning_rate": 1.1324726354195904e-06, "loss": 0.0494, "step": 40335 }, { "epoch": 1.89, "learning_rate": 1.1301337823931145e-06, "loss": 0.123, "step": 40340 }, { "epoch": 1.89, "learning_rate": 1.1277949293666388e-06, "loss": 0.0282, "step": 40345 }, { "epoch": 1.89, "learning_rate": 1.125456076340163e-06, "loss": 0.0915, "step": 40350 }, { "epoch": 1.89, "learning_rate": 1.123117223313687e-06, "loss": 0.0562, "step": 40355 }, { "epoch": 1.89, "learning_rate": 1.1207783702872111e-06, "loss": 0.0663, "step": 40360 }, { "epoch": 1.89, "learning_rate": 1.1184395172607354e-06, "loss": 0.0352, "step": 40365 }, { "epoch": 1.89, "learning_rate": 1.1161006642342597e-06, "loss": 0.0392, "step": 40370 }, { "epoch": 1.89, "learning_rate": 1.1137618112077836e-06, "loss": 0.0212, "step": 40375 }, { "epoch": 1.89, "learning_rate": 1.111422958181308e-06, "loss": 0.0784, "step": 40380 }, { "epoch": 1.89, "learning_rate": 1.1090841051548323e-06, "loss": 0.0728, "step": 40385 }, { "epoch": 1.89, "learning_rate": 1.1067452521283564e-06, "loss": 0.0597, "step": 40390 }, { "epoch": 1.89, "learning_rate": 1.1044063991018805e-06, "loss": 0.0667, "step": 40395 }, { "epoch": 1.89, "learning_rate": 1.1020675460754048e-06, "loss": 0.0981, "step": 40400 }, { "epoch": 1.89, "learning_rate": 1.099728693048929e-06, "loss": 0.0232, "step": 40405 }, { "epoch": 1.89, "learning_rate": 1.097389840022453e-06, "loss": 0.0465, "step": 40410 }, { "epoch": 1.89, "learning_rate": 1.0950509869959773e-06, "loss": 0.0383, "step": 40415 }, { "epoch": 1.89, "learning_rate": 1.0927121339695014e-06, "loss": 0.084, "step": 40420 }, { "epoch": 1.89, "learning_rate": 1.0903732809430257e-06, "loss": 0.0686, "step": 40425 }, { "epoch": 1.89, "learning_rate": 1.0880344279165498e-06, "loss": 0.0517, "step": 40430 }, { "epoch": 1.89, "learning_rate": 1.085695574890074e-06, "loss": 0.0152, "step": 40435 }, { "epoch": 1.89, "learning_rate": 1.0833567218635983e-06, "loss": 0.0591, "step": 40440 }, { "epoch": 1.89, "learning_rate": 1.0810178688371224e-06, "loss": 0.0375, "step": 40445 }, { "epoch": 1.89, "learning_rate": 1.0786790158106465e-06, "loss": 0.0579, "step": 40450 }, { "epoch": 1.89, "learning_rate": 1.0763401627841708e-06, "loss": 0.0943, "step": 40455 }, { "epoch": 1.89, "learning_rate": 1.0740013097576949e-06, "loss": 0.0378, "step": 40460 }, { "epoch": 1.89, "learning_rate": 1.071662456731219e-06, "loss": 0.0344, "step": 40465 }, { "epoch": 1.89, "learning_rate": 1.0693236037047433e-06, "loss": 0.0286, "step": 40470 }, { "epoch": 1.89, "learning_rate": 1.0669847506782674e-06, "loss": 0.0275, "step": 40475 }, { "epoch": 1.89, "learning_rate": 1.0646458976517917e-06, "loss": 0.0583, "step": 40480 }, { "epoch": 1.89, "learning_rate": 1.0623070446253158e-06, "loss": 0.061, "step": 40485 }, { "epoch": 1.89, "learning_rate": 1.05996819159884e-06, "loss": 0.0961, "step": 40490 }, { "epoch": 1.89, "learning_rate": 1.0576293385723643e-06, "loss": 0.0595, "step": 40495 }, { "epoch": 1.89, "learning_rate": 1.0552904855458884e-06, "loss": 0.0718, "step": 40500 }, { "epoch": 1.89, "learning_rate": 1.0529516325194125e-06, "loss": 0.081, "step": 40505 }, { "epoch": 1.89, "learning_rate": 1.0506127794929368e-06, "loss": 0.0289, "step": 40510 }, { "epoch": 1.9, "learning_rate": 1.0482739264664609e-06, "loss": 0.0238, "step": 40515 }, { "epoch": 1.9, "learning_rate": 1.045935073439985e-06, "loss": 0.0508, "step": 40520 }, { "epoch": 1.9, "learning_rate": 1.0435962204135093e-06, "loss": 0.0429, "step": 40525 }, { "epoch": 1.9, "learning_rate": 1.0412573673870336e-06, "loss": 0.0671, "step": 40530 }, { "epoch": 1.9, "learning_rate": 1.0389185143605577e-06, "loss": 0.0527, "step": 40535 }, { "epoch": 1.9, "learning_rate": 1.0365796613340818e-06, "loss": 0.0833, "step": 40540 }, { "epoch": 1.9, "learning_rate": 1.0342408083076062e-06, "loss": 0.0328, "step": 40545 }, { "epoch": 1.9, "learning_rate": 1.0319019552811303e-06, "loss": 0.0456, "step": 40550 }, { "epoch": 1.9, "learning_rate": 1.0295631022546544e-06, "loss": 0.0659, "step": 40555 }, { "epoch": 1.9, "learning_rate": 1.0272242492281787e-06, "loss": 0.069, "step": 40560 }, { "epoch": 1.9, "learning_rate": 1.0248853962017028e-06, "loss": 0.0288, "step": 40565 }, { "epoch": 1.9, "learning_rate": 1.0225465431752269e-06, "loss": 0.0238, "step": 40570 }, { "epoch": 1.9, "learning_rate": 1.020207690148751e-06, "loss": 0.0473, "step": 40575 }, { "epoch": 1.9, "learning_rate": 1.0178688371222753e-06, "loss": 0.0961, "step": 40580 }, { "epoch": 1.9, "learning_rate": 1.0155299840957996e-06, "loss": 0.0371, "step": 40585 }, { "epoch": 1.9, "learning_rate": 1.0131911310693237e-06, "loss": 0.1046, "step": 40590 }, { "epoch": 1.9, "learning_rate": 1.0108522780428478e-06, "loss": 0.1077, "step": 40595 }, { "epoch": 1.9, "learning_rate": 1.0085134250163721e-06, "loss": 0.0354, "step": 40600 }, { "epoch": 1.9, "learning_rate": 1.0061745719898963e-06, "loss": 0.0469, "step": 40605 }, { "epoch": 1.9, "learning_rate": 1.0038357189634204e-06, "loss": 0.0877, "step": 40610 }, { "epoch": 1.9, "learning_rate": 1.0014968659369447e-06, "loss": 0.0553, "step": 40615 }, { "epoch": 1.9, "learning_rate": 9.991580129104688e-07, "loss": 0.0698, "step": 40620 }, { "epoch": 1.9, "learning_rate": 9.968191598839929e-07, "loss": 0.0784, "step": 40625 }, { "epoch": 1.9, "learning_rate": 9.944803068575172e-07, "loss": 0.0387, "step": 40630 }, { "epoch": 1.9, "learning_rate": 9.921414538310413e-07, "loss": 0.0586, "step": 40635 }, { "epoch": 1.9, "learning_rate": 9.898026008045656e-07, "loss": 0.0713, "step": 40640 }, { "epoch": 1.9, "learning_rate": 9.874637477780897e-07, "loss": 0.0494, "step": 40645 }, { "epoch": 1.9, "learning_rate": 9.851248947516138e-07, "loss": 0.0788, "step": 40650 }, { "epoch": 1.9, "learning_rate": 9.827860417251381e-07, "loss": 0.0715, "step": 40655 }, { "epoch": 1.9, "learning_rate": 9.804471886986622e-07, "loss": 0.1431, "step": 40660 }, { "epoch": 1.9, "learning_rate": 9.781083356721864e-07, "loss": 0.0118, "step": 40665 }, { "epoch": 1.9, "learning_rate": 9.757694826457107e-07, "loss": 0.0794, "step": 40670 }, { "epoch": 1.9, "learning_rate": 9.734306296192348e-07, "loss": 0.0815, "step": 40675 }, { "epoch": 1.9, "learning_rate": 9.710917765927589e-07, "loss": 0.0795, "step": 40680 }, { "epoch": 1.9, "learning_rate": 9.687529235662832e-07, "loss": 0.0546, "step": 40685 }, { "epoch": 1.9, "learning_rate": 9.664140705398073e-07, "loss": 0.022, "step": 40690 }, { "epoch": 1.9, "learning_rate": 9.640752175133316e-07, "loss": 0.0559, "step": 40695 }, { "epoch": 1.9, "learning_rate": 9.617363644868557e-07, "loss": 0.0442, "step": 40700 }, { "epoch": 1.9, "learning_rate": 9.593975114603798e-07, "loss": 0.0752, "step": 40705 }, { "epoch": 1.9, "learning_rate": 9.570586584339041e-07, "loss": 0.0548, "step": 40710 }, { "epoch": 1.9, "learning_rate": 9.547198054074282e-07, "loss": 0.0529, "step": 40715 }, { "epoch": 1.9, "learning_rate": 9.523809523809525e-07, "loss": 0.0675, "step": 40720 }, { "epoch": 1.9, "learning_rate": 9.500420993544767e-07, "loss": 0.0321, "step": 40725 }, { "epoch": 1.91, "learning_rate": 9.477032463280009e-07, "loss": 0.0498, "step": 40730 }, { "epoch": 1.91, "learning_rate": 9.45364393301525e-07, "loss": 0.0649, "step": 40735 }, { "epoch": 1.91, "learning_rate": 9.430255402750492e-07, "loss": 0.0328, "step": 40740 }, { "epoch": 1.91, "learning_rate": 9.406866872485734e-07, "loss": 0.0415, "step": 40745 }, { "epoch": 1.91, "learning_rate": 9.383478342220975e-07, "loss": 0.0241, "step": 40750 }, { "epoch": 1.91, "learning_rate": 9.360089811956217e-07, "loss": 0.0819, "step": 40755 }, { "epoch": 1.91, "learning_rate": 9.33670128169146e-07, "loss": 0.1259, "step": 40760 }, { "epoch": 1.91, "learning_rate": 9.3133127514267e-07, "loss": 0.0376, "step": 40765 }, { "epoch": 1.91, "learning_rate": 9.289924221161943e-07, "loss": 0.0388, "step": 40770 }, { "epoch": 1.91, "learning_rate": 9.266535690897186e-07, "loss": 0.0725, "step": 40775 }, { "epoch": 1.91, "learning_rate": 9.243147160632427e-07, "loss": 0.0817, "step": 40780 }, { "epoch": 1.91, "learning_rate": 9.219758630367669e-07, "loss": 0.0731, "step": 40785 }, { "epoch": 1.91, "learning_rate": 9.19637010010291e-07, "loss": 0.0504, "step": 40790 }, { "epoch": 1.91, "learning_rate": 9.172981569838152e-07, "loss": 0.0301, "step": 40795 }, { "epoch": 1.91, "learning_rate": 9.149593039573394e-07, "loss": 0.0764, "step": 40800 }, { "epoch": 1.91, "learning_rate": 9.126204509308635e-07, "loss": 0.0327, "step": 40805 }, { "epoch": 1.91, "learning_rate": 9.102815979043877e-07, "loss": 0.0595, "step": 40810 }, { "epoch": 1.91, "learning_rate": 9.07942744877912e-07, "loss": 0.0394, "step": 40815 }, { "epoch": 1.91, "learning_rate": 9.05603891851436e-07, "loss": 0.0823, "step": 40820 }, { "epoch": 1.91, "learning_rate": 9.032650388249603e-07, "loss": 0.0531, "step": 40825 }, { "epoch": 1.91, "learning_rate": 9.009261857984846e-07, "loss": 0.0424, "step": 40830 }, { "epoch": 1.91, "learning_rate": 8.985873327720087e-07, "loss": 0.0451, "step": 40835 }, { "epoch": 1.91, "learning_rate": 8.962484797455329e-07, "loss": 0.0697, "step": 40840 }, { "epoch": 1.91, "learning_rate": 8.939096267190571e-07, "loss": 0.0451, "step": 40845 }, { "epoch": 1.91, "learning_rate": 8.915707736925812e-07, "loss": 0.0597, "step": 40850 }, { "epoch": 1.91, "learning_rate": 8.892319206661054e-07, "loss": 0.0542, "step": 40855 }, { "epoch": 1.91, "learning_rate": 8.868930676396296e-07, "loss": 0.0415, "step": 40860 }, { "epoch": 1.91, "learning_rate": 8.845542146131537e-07, "loss": 0.0263, "step": 40865 }, { "epoch": 1.91, "learning_rate": 8.82215361586678e-07, "loss": 0.0569, "step": 40870 }, { "epoch": 1.91, "learning_rate": 8.798765085602022e-07, "loss": 0.0422, "step": 40875 }, { "epoch": 1.91, "learning_rate": 8.775376555337263e-07, "loss": 0.041, "step": 40880 }, { "epoch": 1.91, "learning_rate": 8.751988025072505e-07, "loss": 0.0758, "step": 40885 }, { "epoch": 1.91, "learning_rate": 8.728599494807748e-07, "loss": 0.0433, "step": 40890 }, { "epoch": 1.91, "learning_rate": 8.705210964542989e-07, "loss": 0.0722, "step": 40895 }, { "epoch": 1.91, "learning_rate": 8.681822434278231e-07, "loss": 0.0309, "step": 40900 }, { "epoch": 1.91, "learning_rate": 8.658433904013472e-07, "loss": 0.0321, "step": 40905 }, { "epoch": 1.91, "learning_rate": 8.635045373748714e-07, "loss": 0.0202, "step": 40910 }, { "epoch": 1.91, "learning_rate": 8.611656843483956e-07, "loss": 0.0313, "step": 40915 }, { "epoch": 1.91, "learning_rate": 8.588268313219197e-07, "loss": 0.0839, "step": 40920 }, { "epoch": 1.91, "learning_rate": 8.56487978295444e-07, "loss": 0.0524, "step": 40925 }, { "epoch": 1.91, "learning_rate": 8.541491252689682e-07, "loss": 0.0159, "step": 40930 }, { "epoch": 1.91, "learning_rate": 8.518102722424923e-07, "loss": 0.0901, "step": 40935 }, { "epoch": 1.92, "learning_rate": 8.494714192160165e-07, "loss": 0.0425, "step": 40940 }, { "epoch": 1.92, "learning_rate": 8.471325661895408e-07, "loss": 0.017, "step": 40945 }, { "epoch": 1.92, "learning_rate": 8.447937131630649e-07, "loss": 0.0361, "step": 40950 }, { "epoch": 1.92, "learning_rate": 8.424548601365891e-07, "loss": 0.0406, "step": 40955 }, { "epoch": 1.92, "learning_rate": 8.401160071101133e-07, "loss": 0.0396, "step": 40960 }, { "epoch": 1.92, "learning_rate": 8.377771540836374e-07, "loss": 0.0627, "step": 40965 }, { "epoch": 1.92, "learning_rate": 8.354383010571616e-07, "loss": 0.0397, "step": 40970 }, { "epoch": 1.92, "learning_rate": 8.330994480306859e-07, "loss": 0.0586, "step": 40975 }, { "epoch": 1.92, "learning_rate": 8.3076059500421e-07, "loss": 0.0605, "step": 40980 }, { "epoch": 1.92, "learning_rate": 8.284217419777342e-07, "loss": 0.0524, "step": 40985 }, { "epoch": 1.92, "learning_rate": 8.260828889512584e-07, "loss": 0.0565, "step": 40990 }, { "epoch": 1.92, "learning_rate": 8.237440359247825e-07, "loss": 0.033, "step": 40995 }, { "epoch": 1.92, "learning_rate": 8.214051828983068e-07, "loss": 0.0238, "step": 41000 }, { "epoch": 1.92, "learning_rate": 8.19066329871831e-07, "loss": 0.0393, "step": 41005 }, { "epoch": 1.92, "learning_rate": 8.167274768453551e-07, "loss": 0.0534, "step": 41010 }, { "epoch": 1.92, "learning_rate": 8.143886238188793e-07, "loss": 0.0392, "step": 41015 }, { "epoch": 1.92, "learning_rate": 8.120497707924034e-07, "loss": 0.06, "step": 41020 }, { "epoch": 1.92, "learning_rate": 8.097109177659276e-07, "loss": 0.033, "step": 41025 }, { "epoch": 1.92, "learning_rate": 8.073720647394519e-07, "loss": 0.0194, "step": 41030 }, { "epoch": 1.92, "learning_rate": 8.05033211712976e-07, "loss": 0.0255, "step": 41035 }, { "epoch": 1.92, "learning_rate": 8.026943586865002e-07, "loss": 0.0322, "step": 41040 }, { "epoch": 1.92, "learning_rate": 8.003555056600244e-07, "loss": 0.0449, "step": 41045 }, { "epoch": 1.92, "learning_rate": 7.980166526335485e-07, "loss": 0.0746, "step": 41050 }, { "epoch": 1.92, "learning_rate": 7.956777996070727e-07, "loss": 0.0554, "step": 41055 }, { "epoch": 1.92, "learning_rate": 7.93338946580597e-07, "loss": 0.0507, "step": 41060 }, { "epoch": 1.92, "learning_rate": 7.910000935541211e-07, "loss": 0.0452, "step": 41065 }, { "epoch": 1.92, "learning_rate": 7.886612405276453e-07, "loss": 0.0639, "step": 41070 }, { "epoch": 1.92, "learning_rate": 7.863223875011696e-07, "loss": 0.0608, "step": 41075 }, { "epoch": 1.92, "learning_rate": 7.839835344746936e-07, "loss": 0.0576, "step": 41080 }, { "epoch": 1.92, "learning_rate": 7.816446814482179e-07, "loss": 0.033, "step": 41085 }, { "epoch": 1.92, "learning_rate": 7.793058284217421e-07, "loss": 0.0937, "step": 41090 }, { "epoch": 1.92, "learning_rate": 7.769669753952662e-07, "loss": 0.0703, "step": 41095 }, { "epoch": 1.92, "learning_rate": 7.746281223687904e-07, "loss": 0.0524, "step": 41100 }, { "epoch": 1.92, "learning_rate": 7.722892693423146e-07, "loss": 0.0195, "step": 41105 }, { "epoch": 1.92, "learning_rate": 7.699504163158387e-07, "loss": 0.0678, "step": 41110 }, { "epoch": 1.92, "learning_rate": 7.67611563289363e-07, "loss": 0.0311, "step": 41115 }, { "epoch": 1.92, "learning_rate": 7.652727102628871e-07, "loss": 0.0539, "step": 41120 }, { "epoch": 1.92, "learning_rate": 7.629338572364113e-07, "loss": 0.0124, "step": 41125 }, { "epoch": 1.92, "learning_rate": 7.605950042099356e-07, "loss": 0.0495, "step": 41130 }, { "epoch": 1.92, "learning_rate": 7.582561511834596e-07, "loss": 0.0374, "step": 41135 }, { "epoch": 1.92, "learning_rate": 7.559172981569839e-07, "loss": 0.041, "step": 41140 }, { "epoch": 1.92, "learning_rate": 7.535784451305081e-07, "loss": 0.0529, "step": 41145 }, { "epoch": 1.92, "learning_rate": 7.512395921040322e-07, "loss": 0.1148, "step": 41150 }, { "epoch": 1.93, "learning_rate": 7.489007390775564e-07, "loss": 0.1384, "step": 41155 }, { "epoch": 1.93, "learning_rate": 7.465618860510806e-07, "loss": 0.0361, "step": 41160 }, { "epoch": 1.93, "learning_rate": 7.442230330246047e-07, "loss": 0.0674, "step": 41165 }, { "epoch": 1.93, "learning_rate": 7.41884179998129e-07, "loss": 0.0599, "step": 41170 }, { "epoch": 1.93, "learning_rate": 7.395453269716532e-07, "loss": 0.0394, "step": 41175 }, { "epoch": 1.93, "learning_rate": 7.372064739451773e-07, "loss": 0.0286, "step": 41180 }, { "epoch": 1.93, "learning_rate": 7.348676209187016e-07, "loss": 0.0627, "step": 41185 }, { "epoch": 1.93, "learning_rate": 7.325287678922258e-07, "loss": 0.0423, "step": 41190 }, { "epoch": 1.93, "learning_rate": 7.301899148657499e-07, "loss": 0.035, "step": 41195 }, { "epoch": 1.93, "learning_rate": 7.278510618392741e-07, "loss": 0.0236, "step": 41200 }, { "epoch": 1.93, "learning_rate": 7.255122088127983e-07, "loss": 0.0425, "step": 41205 }, { "epoch": 1.93, "learning_rate": 7.231733557863224e-07, "loss": 0.0322, "step": 41210 }, { "epoch": 1.93, "learning_rate": 7.208345027598466e-07, "loss": 0.0549, "step": 41215 }, { "epoch": 1.93, "learning_rate": 7.184956497333708e-07, "loss": 0.0485, "step": 41220 }, { "epoch": 1.93, "learning_rate": 7.161567967068949e-07, "loss": 0.0462, "step": 41225 }, { "epoch": 1.93, "learning_rate": 7.138179436804192e-07, "loss": 0.0846, "step": 41230 }, { "epoch": 1.93, "learning_rate": 7.114790906539433e-07, "loss": 0.0188, "step": 41235 }, { "epoch": 1.93, "learning_rate": 7.091402376274676e-07, "loss": 0.0446, "step": 41240 }, { "epoch": 1.93, "learning_rate": 7.068013846009918e-07, "loss": 0.0986, "step": 41245 }, { "epoch": 1.93, "learning_rate": 7.044625315745159e-07, "loss": 0.0484, "step": 41250 }, { "epoch": 1.93, "learning_rate": 7.021236785480401e-07, "loss": 0.0897, "step": 41255 }, { "epoch": 1.93, "learning_rate": 6.997848255215643e-07, "loss": 0.0434, "step": 41260 }, { "epoch": 1.93, "learning_rate": 6.974459724950884e-07, "loss": 0.076, "step": 41265 }, { "epoch": 1.93, "learning_rate": 6.951071194686126e-07, "loss": 0.0687, "step": 41270 }, { "epoch": 1.93, "learning_rate": 6.927682664421368e-07, "loss": 0.0501, "step": 41275 }, { "epoch": 1.93, "learning_rate": 6.904294134156609e-07, "loss": 0.038, "step": 41280 }, { "epoch": 1.93, "learning_rate": 6.880905603891852e-07, "loss": 0.1153, "step": 41285 }, { "epoch": 1.93, "learning_rate": 6.857517073627095e-07, "loss": 0.0372, "step": 41290 }, { "epoch": 1.93, "learning_rate": 6.834128543362336e-07, "loss": 0.0484, "step": 41295 }, { "epoch": 1.93, "learning_rate": 6.810740013097578e-07, "loss": 0.0349, "step": 41300 }, { "epoch": 1.93, "learning_rate": 6.78735148283282e-07, "loss": 0.0696, "step": 41305 }, { "epoch": 1.93, "learning_rate": 6.763962952568061e-07, "loss": 0.0486, "step": 41310 }, { "epoch": 1.93, "learning_rate": 6.740574422303303e-07, "loss": 0.0672, "step": 41315 }, { "epoch": 1.93, "learning_rate": 6.717185892038545e-07, "loss": 0.0758, "step": 41320 }, { "epoch": 1.93, "learning_rate": 6.693797361773786e-07, "loss": 0.0416, "step": 41325 }, { "epoch": 1.93, "learning_rate": 6.670408831509028e-07, "loss": 0.0415, "step": 41330 }, { "epoch": 1.93, "learning_rate": 6.647020301244269e-07, "loss": 0.0565, "step": 41335 }, { "epoch": 1.93, "learning_rate": 6.623631770979511e-07, "loss": 0.0688, "step": 41340 }, { "epoch": 1.93, "learning_rate": 6.600243240714755e-07, "loss": 0.0332, "step": 41345 }, { "epoch": 1.93, "learning_rate": 6.576854710449996e-07, "loss": 0.0423, "step": 41350 }, { "epoch": 1.93, "learning_rate": 6.553466180185238e-07, "loss": 0.1141, "step": 41355 }, { "epoch": 1.93, "learning_rate": 6.53007764992048e-07, "loss": 0.0386, "step": 41360 }, { "epoch": 1.93, "learning_rate": 6.506689119655721e-07, "loss": 0.0413, "step": 41365 }, { "epoch": 1.94, "learning_rate": 6.483300589390963e-07, "loss": 0.0469, "step": 41370 }, { "epoch": 1.94, "learning_rate": 6.459912059126205e-07, "loss": 0.074, "step": 41375 }, { "epoch": 1.94, "learning_rate": 6.436523528861446e-07, "loss": 0.0465, "step": 41380 }, { "epoch": 1.94, "learning_rate": 6.413134998596688e-07, "loss": 0.0634, "step": 41385 }, { "epoch": 1.94, "learning_rate": 6.389746468331931e-07, "loss": 0.0367, "step": 41390 }, { "epoch": 1.94, "learning_rate": 6.366357938067171e-07, "loss": 0.1122, "step": 41395 }, { "epoch": 1.94, "learning_rate": 6.342969407802415e-07, "loss": 0.0345, "step": 41400 }, { "epoch": 1.94, "learning_rate": 6.319580877537657e-07, "loss": 0.0094, "step": 41405 }, { "epoch": 1.94, "learning_rate": 6.296192347272898e-07, "loss": 0.019, "step": 41410 }, { "epoch": 1.94, "learning_rate": 6.27280381700814e-07, "loss": 0.0478, "step": 41415 }, { "epoch": 1.94, "learning_rate": 6.249415286743382e-07, "loss": 0.0578, "step": 41420 }, { "epoch": 1.94, "learning_rate": 6.226026756478623e-07, "loss": 0.0354, "step": 41425 }, { "epoch": 1.94, "learning_rate": 6.202638226213865e-07, "loss": 0.0971, "step": 41430 }, { "epoch": 1.94, "learning_rate": 6.179249695949107e-07, "loss": 0.0764, "step": 41435 }, { "epoch": 1.94, "learning_rate": 6.155861165684348e-07, "loss": 0.0355, "step": 41440 }, { "epoch": 1.94, "learning_rate": 6.132472635419591e-07, "loss": 0.0362, "step": 41445 }, { "epoch": 1.94, "learning_rate": 6.109084105154832e-07, "loss": 0.1004, "step": 41450 }, { "epoch": 1.94, "learning_rate": 6.085695574890075e-07, "loss": 0.0489, "step": 41455 }, { "epoch": 1.94, "learning_rate": 6.062307044625317e-07, "loss": 0.0802, "step": 41460 }, { "epoch": 1.94, "learning_rate": 6.038918514360558e-07, "loss": 0.0112, "step": 41465 }, { "epoch": 1.94, "learning_rate": 6.0155299840958e-07, "loss": 0.0239, "step": 41470 }, { "epoch": 1.94, "learning_rate": 5.992141453831042e-07, "loss": 0.0297, "step": 41475 }, { "epoch": 1.94, "learning_rate": 5.968752923566284e-07, "loss": 0.026, "step": 41480 }, { "epoch": 1.94, "learning_rate": 5.945364393301525e-07, "loss": 0.0494, "step": 41485 }, { "epoch": 1.94, "learning_rate": 5.921975863036767e-07, "loss": 0.0662, "step": 41490 }, { "epoch": 1.94, "learning_rate": 5.898587332772009e-07, "loss": 0.0226, "step": 41495 }, { "epoch": 1.94, "learning_rate": 5.875198802507251e-07, "loss": 0.0311, "step": 41500 }, { "epoch": 1.94, "learning_rate": 5.851810272242492e-07, "loss": 0.0341, "step": 41505 }, { "epoch": 1.94, "learning_rate": 5.828421741977735e-07, "loss": 0.0211, "step": 41510 }, { "epoch": 1.94, "learning_rate": 5.805033211712977e-07, "loss": 0.0607, "step": 41515 }, { "epoch": 1.94, "learning_rate": 5.781644681448218e-07, "loss": 0.0758, "step": 41520 }, { "epoch": 1.94, "learning_rate": 5.75825615118346e-07, "loss": 0.0559, "step": 41525 }, { "epoch": 1.94, "learning_rate": 5.734867620918702e-07, "loss": 0.0534, "step": 41530 }, { "epoch": 1.94, "learning_rate": 5.711479090653944e-07, "loss": 0.0602, "step": 41535 }, { "epoch": 1.94, "learning_rate": 5.688090560389185e-07, "loss": 0.0684, "step": 41540 }, { "epoch": 1.94, "learning_rate": 5.664702030124427e-07, "loss": 0.02, "step": 41545 }, { "epoch": 1.94, "learning_rate": 5.641313499859669e-07, "loss": 0.0734, "step": 41550 }, { "epoch": 1.94, "learning_rate": 5.617924969594911e-07, "loss": 0.0314, "step": 41555 }, { "epoch": 1.94, "learning_rate": 5.594536439330153e-07, "loss": 0.0319, "step": 41560 }, { "epoch": 1.94, "learning_rate": 5.571147909065394e-07, "loss": 0.04, "step": 41565 }, { "epoch": 1.94, "learning_rate": 5.547759378800637e-07, "loss": 0.0717, "step": 41570 }, { "epoch": 1.94, "learning_rate": 5.524370848535879e-07, "loss": 0.0485, "step": 41575 }, { "epoch": 1.94, "learning_rate": 5.500982318271121e-07, "loss": 0.0581, "step": 41580 }, { "epoch": 1.95, "learning_rate": 5.477593788006362e-07, "loss": 0.0648, "step": 41585 }, { "epoch": 1.95, "learning_rate": 5.454205257741604e-07, "loss": 0.0191, "step": 41590 }, { "epoch": 1.95, "learning_rate": 5.430816727476846e-07, "loss": 0.0388, "step": 41595 }, { "epoch": 1.95, "learning_rate": 5.407428197212087e-07, "loss": 0.0285, "step": 41600 }, { "epoch": 1.95, "learning_rate": 5.384039666947329e-07, "loss": 0.0915, "step": 41605 }, { "epoch": 1.95, "learning_rate": 5.360651136682571e-07, "loss": 0.0708, "step": 41610 }, { "epoch": 1.95, "learning_rate": 5.337262606417813e-07, "loss": 0.0583, "step": 41615 }, { "epoch": 1.95, "learning_rate": 5.313874076153054e-07, "loss": 0.0489, "step": 41620 }, { "epoch": 1.95, "learning_rate": 5.290485545888298e-07, "loss": 0.047, "step": 41625 }, { "epoch": 1.95, "learning_rate": 5.267097015623539e-07, "loss": 0.0192, "step": 41630 }, { "epoch": 1.95, "learning_rate": 5.243708485358781e-07, "loss": 0.0237, "step": 41635 }, { "epoch": 1.95, "learning_rate": 5.220319955094022e-07, "loss": 0.0924, "step": 41640 }, { "epoch": 1.95, "learning_rate": 5.196931424829264e-07, "loss": 0.07, "step": 41645 }, { "epoch": 1.95, "learning_rate": 5.173542894564506e-07, "loss": 0.076, "step": 41650 }, { "epoch": 1.95, "learning_rate": 5.150154364299747e-07, "loss": 0.0617, "step": 41655 }, { "epoch": 1.95, "learning_rate": 5.12676583403499e-07, "loss": 0.0564, "step": 41660 }, { "epoch": 1.95, "learning_rate": 5.103377303770231e-07, "loss": 0.0524, "step": 41665 }, { "epoch": 1.95, "learning_rate": 5.079988773505473e-07, "loss": 0.0955, "step": 41670 }, { "epoch": 1.95, "learning_rate": 5.056600243240715e-07, "loss": 0.0472, "step": 41675 }, { "epoch": 1.95, "learning_rate": 5.033211712975958e-07, "loss": 0.0278, "step": 41680 }, { "epoch": 1.95, "learning_rate": 5.009823182711199e-07, "loss": 0.0221, "step": 41685 }, { "epoch": 1.95, "learning_rate": 4.986434652446441e-07, "loss": 0.0624, "step": 41690 }, { "epoch": 1.95, "learning_rate": 4.963046122181683e-07, "loss": 0.0526, "step": 41695 }, { "epoch": 1.95, "learning_rate": 4.939657591916924e-07, "loss": 0.033, "step": 41700 }, { "epoch": 1.95, "learning_rate": 4.916269061652166e-07, "loss": 0.0538, "step": 41705 }, { "epoch": 1.95, "learning_rate": 4.892880531387408e-07, "loss": 0.033, "step": 41710 }, { "epoch": 1.95, "learning_rate": 4.86949200112265e-07, "loss": 0.0588, "step": 41715 }, { "epoch": 1.95, "learning_rate": 4.846103470857891e-07, "loss": 0.076, "step": 41720 }, { "epoch": 1.95, "learning_rate": 4.822714940593133e-07, "loss": 0.0746, "step": 41725 }, { "epoch": 1.95, "learning_rate": 4.799326410328375e-07, "loss": 0.0421, "step": 41730 }, { "epoch": 1.95, "learning_rate": 4.775937880063618e-07, "loss": 0.0299, "step": 41735 }, { "epoch": 1.95, "learning_rate": 4.752549349798859e-07, "loss": 0.0635, "step": 41740 }, { "epoch": 1.95, "learning_rate": 4.729160819534101e-07, "loss": 0.0361, "step": 41745 }, { "epoch": 1.95, "learning_rate": 4.705772289269343e-07, "loss": 0.0221, "step": 41750 }, { "epoch": 1.95, "learning_rate": 4.6823837590045844e-07, "loss": 0.0371, "step": 41755 }, { "epoch": 1.95, "learning_rate": 4.6589952287398265e-07, "loss": 0.1144, "step": 41760 }, { "epoch": 1.95, "learning_rate": 4.635606698475068e-07, "loss": 0.0607, "step": 41765 }, { "epoch": 1.95, "learning_rate": 4.6122181682103096e-07, "loss": 0.0434, "step": 41770 }, { "epoch": 1.95, "learning_rate": 4.588829637945552e-07, "loss": 0.0681, "step": 41775 }, { "epoch": 1.95, "learning_rate": 4.565441107680794e-07, "loss": 0.0553, "step": 41780 }, { "epoch": 1.95, "learning_rate": 4.5420525774160354e-07, "loss": 0.0334, "step": 41785 }, { "epoch": 1.95, "learning_rate": 4.5186640471512775e-07, "loss": 0.0469, "step": 41790 }, { "epoch": 1.96, "learning_rate": 4.495275516886519e-07, "loss": 0.0533, "step": 41795 }, { "epoch": 1.96, "learning_rate": 4.471886986621761e-07, "loss": 0.0453, "step": 41800 }, { "epoch": 1.96, "learning_rate": 4.448498456357003e-07, "loss": 0.0215, "step": 41805 }, { "epoch": 1.96, "learning_rate": 4.425109926092245e-07, "loss": 0.0719, "step": 41810 }, { "epoch": 1.96, "learning_rate": 4.4017213958274864e-07, "loss": 0.0093, "step": 41815 }, { "epoch": 1.96, "learning_rate": 4.378332865562728e-07, "loss": 0.0753, "step": 41820 }, { "epoch": 1.96, "learning_rate": 4.3549443352979706e-07, "loss": 0.0919, "step": 41825 }, { "epoch": 1.96, "learning_rate": 4.331555805033212e-07, "loss": 0.0405, "step": 41830 }, { "epoch": 1.96, "learning_rate": 4.308167274768454e-07, "loss": 0.0742, "step": 41835 }, { "epoch": 1.96, "learning_rate": 4.284778744503696e-07, "loss": 0.0787, "step": 41840 }, { "epoch": 1.96, "learning_rate": 4.2613902142389375e-07, "loss": 0.0666, "step": 41845 }, { "epoch": 1.96, "learning_rate": 4.2380016839741796e-07, "loss": 0.046, "step": 41850 }, { "epoch": 1.96, "learning_rate": 4.214613153709421e-07, "loss": 0.0317, "step": 41855 }, { "epoch": 1.96, "learning_rate": 4.191224623444663e-07, "loss": 0.0168, "step": 41860 }, { "epoch": 1.96, "learning_rate": 4.167836093179905e-07, "loss": 0.0637, "step": 41865 }, { "epoch": 1.96, "learning_rate": 4.1444475629151464e-07, "loss": 0.0182, "step": 41870 }, { "epoch": 1.96, "learning_rate": 4.121059032650389e-07, "loss": 0.0481, "step": 41875 }, { "epoch": 1.96, "learning_rate": 4.0976705023856306e-07, "loss": 0.0102, "step": 41880 }, { "epoch": 1.96, "learning_rate": 4.074281972120872e-07, "loss": 0.0352, "step": 41885 }, { "epoch": 1.96, "learning_rate": 4.0508934418561143e-07, "loss": 0.0227, "step": 41890 }, { "epoch": 1.96, "learning_rate": 4.027504911591356e-07, "loss": 0.0739, "step": 41895 }, { "epoch": 1.96, "learning_rate": 4.0041163813265974e-07, "loss": 0.0032, "step": 41900 }, { "epoch": 1.96, "learning_rate": 3.9807278510618395e-07, "loss": 0.0974, "step": 41905 }, { "epoch": 1.96, "learning_rate": 3.9573393207970816e-07, "loss": 0.0294, "step": 41910 }, { "epoch": 1.96, "learning_rate": 3.933950790532323e-07, "loss": 0.0787, "step": 41915 }, { "epoch": 1.96, "learning_rate": 3.910562260267565e-07, "loss": 0.0552, "step": 41920 }, { "epoch": 1.96, "learning_rate": 3.8871737300028074e-07, "loss": 0.0449, "step": 41925 }, { "epoch": 1.96, "learning_rate": 3.863785199738049e-07, "loss": 0.107, "step": 41930 }, { "epoch": 1.96, "learning_rate": 3.8403966694732906e-07, "loss": 0.0453, "step": 41935 }, { "epoch": 1.96, "learning_rate": 3.8170081392085327e-07, "loss": 0.0786, "step": 41940 }, { "epoch": 1.96, "learning_rate": 3.793619608943774e-07, "loss": 0.0473, "step": 41945 }, { "epoch": 1.96, "learning_rate": 3.770231078679016e-07, "loss": 0.0229, "step": 41950 }, { "epoch": 1.96, "learning_rate": 3.7468425484142584e-07, "loss": 0.0569, "step": 41955 }, { "epoch": 1.96, "learning_rate": 3.7234540181495e-07, "loss": 0.0364, "step": 41960 }, { "epoch": 1.96, "learning_rate": 3.7000654878847416e-07, "loss": 0.0583, "step": 41965 }, { "epoch": 1.96, "learning_rate": 3.676676957619983e-07, "loss": 0.0379, "step": 41970 }, { "epoch": 1.96, "learning_rate": 3.6532884273552253e-07, "loss": 0.0228, "step": 41975 }, { "epoch": 1.96, "learning_rate": 3.6298998970904674e-07, "loss": 0.0325, "step": 41980 }, { "epoch": 1.96, "learning_rate": 3.606511366825709e-07, "loss": 0.0183, "step": 41985 }, { "epoch": 1.96, "learning_rate": 3.583122836560951e-07, "loss": 0.0744, "step": 41990 }, { "epoch": 1.96, "learning_rate": 3.5597343062961926e-07, "loss": 0.0516, "step": 41995 }, { "epoch": 1.96, "learning_rate": 3.536345776031434e-07, "loss": 0.0373, "step": 42000 }, { "epoch": 1.96, "learning_rate": 3.512957245766677e-07, "loss": 0.0509, "step": 42005 }, { "epoch": 1.97, "learning_rate": 3.4895687155019184e-07, "loss": 0.0837, "step": 42010 }, { "epoch": 1.97, "learning_rate": 3.46618018523716e-07, "loss": 0.0432, "step": 42015 }, { "epoch": 1.97, "learning_rate": 3.4427916549724016e-07, "loss": 0.0409, "step": 42020 }, { "epoch": 1.97, "learning_rate": 3.4194031247076437e-07, "loss": 0.0393, "step": 42025 }, { "epoch": 1.97, "learning_rate": 3.396014594442885e-07, "loss": 0.0398, "step": 42030 }, { "epoch": 1.97, "learning_rate": 3.3726260641781273e-07, "loss": 0.0123, "step": 42035 }, { "epoch": 1.97, "learning_rate": 3.3492375339133694e-07, "loss": 0.0391, "step": 42040 }, { "epoch": 1.97, "learning_rate": 3.325849003648611e-07, "loss": 0.0766, "step": 42045 }, { "epoch": 1.97, "learning_rate": 3.3024604733838526e-07, "loss": 0.0342, "step": 42050 }, { "epoch": 1.97, "learning_rate": 3.279071943119095e-07, "loss": 0.0614, "step": 42055 }, { "epoch": 1.97, "learning_rate": 3.255683412854337e-07, "loss": 0.0228, "step": 42060 }, { "epoch": 1.97, "learning_rate": 3.2322948825895784e-07, "loss": 0.0645, "step": 42065 }, { "epoch": 1.97, "learning_rate": 3.20890635232482e-07, "loss": 0.0435, "step": 42070 }, { "epoch": 1.97, "learning_rate": 3.185517822060062e-07, "loss": 0.0679, "step": 42075 }, { "epoch": 1.97, "learning_rate": 3.1621292917953036e-07, "loss": 0.0649, "step": 42080 }, { "epoch": 1.97, "learning_rate": 3.138740761530545e-07, "loss": 0.0415, "step": 42085 }, { "epoch": 1.97, "learning_rate": 3.1153522312657873e-07, "loss": 0.0291, "step": 42090 }, { "epoch": 1.97, "learning_rate": 3.0919637010010294e-07, "loss": 0.0217, "step": 42095 }, { "epoch": 1.97, "learning_rate": 3.0685751707362715e-07, "loss": 0.0257, "step": 42100 }, { "epoch": 1.97, "learning_rate": 3.045186640471513e-07, "loss": 0.0168, "step": 42105 }, { "epoch": 1.97, "learning_rate": 3.021798110206755e-07, "loss": 0.0328, "step": 42110 }, { "epoch": 1.97, "learning_rate": 2.998409579941997e-07, "loss": 0.0239, "step": 42115 }, { "epoch": 1.97, "learning_rate": 2.9750210496772383e-07, "loss": 0.109, "step": 42120 }, { "epoch": 1.97, "learning_rate": 2.9516325194124804e-07, "loss": 0.0051, "step": 42125 }, { "epoch": 1.97, "learning_rate": 2.928243989147722e-07, "loss": 0.0356, "step": 42130 }, { "epoch": 1.97, "learning_rate": 2.904855458882964e-07, "loss": 0.0709, "step": 42135 }, { "epoch": 1.97, "learning_rate": 2.8814669286182057e-07, "loss": 0.0263, "step": 42140 }, { "epoch": 1.97, "learning_rate": 2.858078398353448e-07, "loss": 0.0637, "step": 42145 }, { "epoch": 1.97, "learning_rate": 2.83468986808869e-07, "loss": 0.036, "step": 42150 }, { "epoch": 1.97, "learning_rate": 2.8113013378239315e-07, "loss": 0.0588, "step": 42155 }, { "epoch": 1.97, "learning_rate": 2.787912807559173e-07, "loss": 0.0645, "step": 42160 }, { "epoch": 1.97, "learning_rate": 2.764524277294415e-07, "loss": 0.0535, "step": 42165 }, { "epoch": 1.97, "learning_rate": 2.7411357470296567e-07, "loss": 0.0582, "step": 42170 }, { "epoch": 1.97, "learning_rate": 2.717747216764899e-07, "loss": 0.0198, "step": 42175 }, { "epoch": 1.97, "learning_rate": 2.6943586865001404e-07, "loss": 0.082, "step": 42180 }, { "epoch": 1.97, "learning_rate": 2.6709701562353825e-07, "loss": 0.0714, "step": 42185 }, { "epoch": 1.97, "learning_rate": 2.647581625970624e-07, "loss": 0.0371, "step": 42190 }, { "epoch": 1.97, "learning_rate": 2.624193095705866e-07, "loss": 0.04, "step": 42195 }, { "epoch": 1.97, "learning_rate": 2.600804565441108e-07, "loss": 0.0852, "step": 42200 }, { "epoch": 1.97, "learning_rate": 2.57741603517635e-07, "loss": 0.0461, "step": 42205 }, { "epoch": 1.97, "learning_rate": 2.5540275049115914e-07, "loss": 0.039, "step": 42210 }, { "epoch": 1.97, "learning_rate": 2.5306389746468335e-07, "loss": 0.0508, "step": 42215 }, { "epoch": 1.97, "learning_rate": 2.507250444382075e-07, "loss": 0.0186, "step": 42220 }, { "epoch": 1.98, "learning_rate": 2.483861914117317e-07, "loss": 0.0336, "step": 42225 }, { "epoch": 1.98, "learning_rate": 2.460473383852559e-07, "loss": 0.0403, "step": 42230 }, { "epoch": 1.98, "learning_rate": 2.437084853587801e-07, "loss": 0.0398, "step": 42235 }, { "epoch": 1.98, "learning_rate": 2.413696323323043e-07, "loss": 0.0784, "step": 42240 }, { "epoch": 1.98, "learning_rate": 2.3903077930582846e-07, "loss": 0.0399, "step": 42245 }, { "epoch": 1.98, "learning_rate": 2.3669192627935264e-07, "loss": 0.061, "step": 42250 }, { "epoch": 1.98, "learning_rate": 2.343530732528768e-07, "loss": 0.0542, "step": 42255 }, { "epoch": 1.98, "learning_rate": 2.3201422022640098e-07, "loss": 0.0492, "step": 42260 }, { "epoch": 1.98, "learning_rate": 2.296753671999252e-07, "loss": 0.0782, "step": 42265 }, { "epoch": 1.98, "learning_rate": 2.2733651417344935e-07, "loss": 0.0521, "step": 42270 }, { "epoch": 1.98, "learning_rate": 2.2499766114697356e-07, "loss": 0.0357, "step": 42275 }, { "epoch": 1.98, "learning_rate": 2.2265880812049772e-07, "loss": 0.0633, "step": 42280 }, { "epoch": 1.98, "learning_rate": 2.203199550940219e-07, "loss": 0.054, "step": 42285 }, { "epoch": 1.98, "learning_rate": 2.179811020675461e-07, "loss": 0.0527, "step": 42290 }, { "epoch": 1.98, "learning_rate": 2.1564224904107027e-07, "loss": 0.0799, "step": 42295 }, { "epoch": 1.98, "learning_rate": 2.1330339601459448e-07, "loss": 0.0911, "step": 42300 }, { "epoch": 1.98, "learning_rate": 2.1096454298811864e-07, "loss": 0.0387, "step": 42305 }, { "epoch": 1.98, "learning_rate": 2.0862568996164282e-07, "loss": 0.0195, "step": 42310 }, { "epoch": 1.98, "learning_rate": 2.0628683693516703e-07, "loss": 0.0514, "step": 42315 }, { "epoch": 1.98, "learning_rate": 2.039479839086912e-07, "loss": 0.0439, "step": 42320 }, { "epoch": 1.98, "learning_rate": 2.0160913088221537e-07, "loss": 0.0499, "step": 42325 }, { "epoch": 1.98, "learning_rate": 1.9927027785573956e-07, "loss": 0.0313, "step": 42330 }, { "epoch": 1.98, "learning_rate": 1.9693142482926374e-07, "loss": 0.0394, "step": 42335 }, { "epoch": 1.98, "learning_rate": 1.9459257180278795e-07, "loss": 0.0621, "step": 42340 }, { "epoch": 1.98, "learning_rate": 1.922537187763121e-07, "loss": 0.0634, "step": 42345 }, { "epoch": 1.98, "learning_rate": 1.899148657498363e-07, "loss": 0.0489, "step": 42350 }, { "epoch": 1.98, "learning_rate": 1.8757601272336047e-07, "loss": 0.029, "step": 42355 }, { "epoch": 1.98, "learning_rate": 1.8523715969688466e-07, "loss": 0.0287, "step": 42360 }, { "epoch": 1.98, "learning_rate": 1.8289830667040887e-07, "loss": 0.052, "step": 42365 }, { "epoch": 1.98, "learning_rate": 1.8055945364393303e-07, "loss": 0.0406, "step": 42370 }, { "epoch": 1.98, "learning_rate": 1.782206006174572e-07, "loss": 0.0574, "step": 42375 }, { "epoch": 1.98, "learning_rate": 1.7588174759098137e-07, "loss": 0.0654, "step": 42380 }, { "epoch": 1.98, "learning_rate": 1.7354289456450558e-07, "loss": 0.0327, "step": 42385 }, { "epoch": 1.98, "learning_rate": 1.712040415380298e-07, "loss": 0.0545, "step": 42390 }, { "epoch": 1.98, "learning_rate": 1.6886518851155395e-07, "loss": 0.0909, "step": 42395 }, { "epoch": 1.98, "learning_rate": 1.6652633548507813e-07, "loss": 0.0083, "step": 42400 }, { "epoch": 1.98, "learning_rate": 1.6418748245860234e-07, "loss": 0.0406, "step": 42405 }, { "epoch": 1.98, "learning_rate": 1.618486294321265e-07, "loss": 0.0336, "step": 42410 }, { "epoch": 1.98, "learning_rate": 1.5950977640565068e-07, "loss": 0.0516, "step": 42415 }, { "epoch": 1.98, "learning_rate": 1.5717092337917486e-07, "loss": 0.0902, "step": 42420 }, { "epoch": 1.98, "learning_rate": 1.5483207035269905e-07, "loss": 0.0515, "step": 42425 }, { "epoch": 1.98, "learning_rate": 1.5249321732622323e-07, "loss": 0.0461, "step": 42430 }, { "epoch": 1.98, "learning_rate": 1.5015436429974742e-07, "loss": 0.0157, "step": 42435 }, { "epoch": 1.99, "learning_rate": 1.478155112732716e-07, "loss": 0.0725, "step": 42440 }, { "epoch": 1.99, "learning_rate": 1.4547665824679578e-07, "loss": 0.0463, "step": 42445 }, { "epoch": 1.99, "learning_rate": 1.4313780522031997e-07, "loss": 0.0373, "step": 42450 }, { "epoch": 1.99, "learning_rate": 1.4079895219384415e-07, "loss": 0.0336, "step": 42455 }, { "epoch": 1.99, "learning_rate": 1.3846009916736834e-07, "loss": 0.0439, "step": 42460 }, { "epoch": 1.99, "learning_rate": 1.3612124614089252e-07, "loss": 0.0664, "step": 42465 }, { "epoch": 1.99, "learning_rate": 1.337823931144167e-07, "loss": 0.0627, "step": 42470 }, { "epoch": 1.99, "learning_rate": 1.314435400879409e-07, "loss": 0.0541, "step": 42475 }, { "epoch": 1.99, "learning_rate": 1.2910468706146507e-07, "loss": 0.0464, "step": 42480 }, { "epoch": 1.99, "learning_rate": 1.2676583403498926e-07, "loss": 0.0604, "step": 42485 }, { "epoch": 1.99, "learning_rate": 1.2442698100851344e-07, "loss": 0.0306, "step": 42490 }, { "epoch": 1.99, "learning_rate": 1.2208812798203762e-07, "loss": 0.0346, "step": 42495 }, { "epoch": 1.99, "learning_rate": 1.197492749555618e-07, "loss": 0.0407, "step": 42500 }, { "epoch": 1.99, "learning_rate": 1.1741042192908598e-07, "loss": 0.0599, "step": 42505 }, { "epoch": 1.99, "learning_rate": 1.1507156890261016e-07, "loss": 0.0641, "step": 42510 }, { "epoch": 1.99, "learning_rate": 1.1273271587613436e-07, "loss": 0.1112, "step": 42515 }, { "epoch": 1.99, "learning_rate": 1.1039386284965854e-07, "loss": 0.0754, "step": 42520 }, { "epoch": 1.99, "learning_rate": 1.0805500982318271e-07, "loss": 0.0558, "step": 42525 }, { "epoch": 1.99, "learning_rate": 1.057161567967069e-07, "loss": 0.074, "step": 42530 }, { "epoch": 1.99, "learning_rate": 1.0337730377023108e-07, "loss": 0.0875, "step": 42535 }, { "epoch": 1.99, "learning_rate": 1.0103845074375528e-07, "loss": 0.0444, "step": 42540 }, { "epoch": 1.99, "learning_rate": 9.869959771727946e-08, "loss": 0.0894, "step": 42545 }, { "epoch": 1.99, "learning_rate": 9.636074469080363e-08, "loss": 0.063, "step": 42550 }, { "epoch": 1.99, "learning_rate": 9.402189166432782e-08, "loss": 0.0566, "step": 42555 }, { "epoch": 1.99, "learning_rate": 9.168303863785201e-08, "loss": 0.0387, "step": 42560 }, { "epoch": 1.99, "learning_rate": 8.93441856113762e-08, "loss": 0.0743, "step": 42565 }, { "epoch": 1.99, "learning_rate": 8.700533258490037e-08, "loss": 0.0937, "step": 42570 }, { "epoch": 1.99, "learning_rate": 8.466647955842455e-08, "loss": 0.0673, "step": 42575 }, { "epoch": 1.99, "learning_rate": 8.232762653194874e-08, "loss": 0.0174, "step": 42580 }, { "epoch": 1.99, "learning_rate": 7.998877350547293e-08, "loss": 0.0472, "step": 42585 }, { "epoch": 1.99, "learning_rate": 7.76499204789971e-08, "loss": 0.0714, "step": 42590 }, { "epoch": 1.99, "learning_rate": 7.531106745252129e-08, "loss": 0.0445, "step": 42595 }, { "epoch": 1.99, "learning_rate": 7.297221442604547e-08, "loss": 0.0396, "step": 42600 }, { "epoch": 1.99, "learning_rate": 7.063336139956965e-08, "loss": 0.0706, "step": 42605 }, { "epoch": 1.99, "learning_rate": 6.829450837309384e-08, "loss": 0.0919, "step": 42610 }, { "epoch": 1.99, "learning_rate": 6.595565534661802e-08, "loss": 0.0428, "step": 42615 }, { "epoch": 1.99, "learning_rate": 6.36168023201422e-08, "loss": 0.1076, "step": 42620 }, { "epoch": 1.99, "learning_rate": 6.127794929366639e-08, "loss": 0.0581, "step": 42625 }, { "epoch": 1.99, "learning_rate": 5.8939096267190574e-08, "loss": 0.0519, "step": 42630 }, { "epoch": 1.99, "learning_rate": 5.660024324071475e-08, "loss": 0.0686, "step": 42635 }, { "epoch": 1.99, "learning_rate": 5.426139021423894e-08, "loss": 0.0581, "step": 42640 }, { "epoch": 1.99, "learning_rate": 5.1922537187763126e-08, "loss": 0.0367, "step": 42645 }, { "epoch": 2.0, "learning_rate": 4.958368416128731e-08, "loss": 0.0729, "step": 42650 }, { "epoch": 2.0, "learning_rate": 4.7244831134811494e-08, "loss": 0.0479, "step": 42655 }, { "epoch": 2.0, "learning_rate": 4.490597810833568e-08, "loss": 0.0918, "step": 42660 }, { "epoch": 2.0, "learning_rate": 4.256712508185986e-08, "loss": 0.0479, "step": 42665 }, { "epoch": 2.0, "learning_rate": 4.022827205538404e-08, "loss": 0.034, "step": 42670 }, { "epoch": 2.0, "learning_rate": 3.788941902890823e-08, "loss": 0.0738, "step": 42675 }, { "epoch": 2.0, "learning_rate": 3.5550566002432406e-08, "loss": 0.0183, "step": 42680 }, { "epoch": 2.0, "learning_rate": 3.321171297595659e-08, "loss": 0.0456, "step": 42685 }, { "epoch": 2.0, "learning_rate": 3.0872859949480774e-08, "loss": 0.0628, "step": 42690 }, { "epoch": 2.0, "learning_rate": 2.853400692300496e-08, "loss": 0.051, "step": 42695 }, { "epoch": 2.0, "learning_rate": 2.6195153896529145e-08, "loss": 0.0872, "step": 42700 }, { "epoch": 2.0, "learning_rate": 2.385630087005333e-08, "loss": 0.058, "step": 42705 }, { "epoch": 2.0, "learning_rate": 2.151744784357751e-08, "loss": 0.0951, "step": 42710 }, { "epoch": 2.0, "learning_rate": 1.9178594817101697e-08, "loss": 0.0281, "step": 42715 }, { "epoch": 2.0, "learning_rate": 1.6839741790625877e-08, "loss": 0.0346, "step": 42720 }, { "epoch": 2.0, "learning_rate": 1.4500888764150061e-08, "loss": 0.0391, "step": 42725 }, { "epoch": 2.0, "learning_rate": 1.2162035737674247e-08, "loss": 0.0293, "step": 42730 }, { "epoch": 2.0, "learning_rate": 9.823182711198429e-09, "loss": 0.1271, "step": 42735 }, { "epoch": 2.0, "learning_rate": 7.484329684722613e-09, "loss": 0.0258, "step": 42740 }, { "epoch": 2.0, "learning_rate": 5.145476658246796e-09, "loss": 0.0292, "step": 42745 }, { "epoch": 2.0, "learning_rate": 2.80662363177098e-09, "loss": 0.046, "step": 42750 }, { "epoch": 2.0, "learning_rate": 4.677706052951633e-10, "loss": 0.0366, "step": 42755 }, { "epoch": 2.0, "eval_method_name": "train_2023-04-15_005636_9bc68158-db17-11ed-934b-3cecef280fae", "eval_missing_count": 0, "eval_test_avg": 81.70416666666667, "eval_test_cite__map": 93.13, "eval_test_cite__ndcg": 97.21, "eval_test_co-cite__map": 91.08, "eval_test_co-cite__ndcg": 96.25, "eval_test_co-read__map": 87.22, "eval_test_co-read__ndcg": 93.72, "eval_test_co-view__map": 85.14, "eval_test_co-view__ndcg": 92.22, "eval_test_mag__f1": 79.37, "eval_test_mesh__f1": 90.05, "eval_test_recomm__adj-NDCG": 54.69, "eval_test_recomm__adj-P@1": 20.37, "eval_val_avg": 81.45833333333333, "eval_val_cite__map": 92.16, "eval_val_cite__ndcg": 96.64, "eval_val_co-cite__map": 90.9, "eval_val_co-cite__ndcg": 96.07, "eval_val_co-read__map": 87.38, "eval_val_co-read__ndcg": 93.72, "eval_val_co-view__map": 83.4, "eval_val_co-view__ndcg": 91.0, "eval_val_mag__f1": 79.78, "eval_val_mesh__f1": 89.74, "eval_val_recomm__adj-NDCG": 55.06, "eval_val_recomm__adj-P@1": 21.65, "step": 42756 }, { "epoch": 2.0, "step": 42756, "total_flos": 1.3804937887245926e+18, "train_runtime": 54936.6039, "train_samples_per_second": 0.778 } ], "max_steps": 42756, "num_train_epochs": 2, "total_flos": 1.3804937887245926e+18, "trial_name": null, "trial_params": null }