{ "best_metric": null, "best_model_checkpoint": null, "epoch": 4.0, "global_step": 58936, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 8.483203257550051e-09, "loss": 7.9156, "step": 5 }, { "epoch": 0.0, "learning_rate": 1.6966406515100103e-08, "loss": 8.0891, "step": 10 }, { "epoch": 0.0, "learning_rate": 2.5449609772650154e-08, "loss": 8.3758, "step": 15 }, { "epoch": 0.0, "learning_rate": 3.3932813030200205e-08, "loss": 7.9437, "step": 20 }, { "epoch": 0.0, "learning_rate": 4.2416016287750254e-08, "loss": 8.1453, "step": 25 }, { "epoch": 0.0, "learning_rate": 5.089921954530031e-08, "loss": 7.868, "step": 30 }, { "epoch": 0.0, "learning_rate": 5.9382422802850356e-08, "loss": 8.1453, "step": 35 }, { "epoch": 0.0, "learning_rate": 6.786562606040041e-08, "loss": 7.9242, "step": 40 }, { "epoch": 0.0, "learning_rate": 7.634882931795046e-08, "loss": 8.132, "step": 45 }, { "epoch": 0.0, "learning_rate": 8.483203257550051e-08, "loss": 8.0695, "step": 50 }, { "epoch": 0.0, "learning_rate": 9.331523583305057e-08, "loss": 7.7742, "step": 55 }, { "epoch": 0.0, "learning_rate": 1.0179843909060062e-07, "loss": 8.1312, "step": 60 }, { "epoch": 0.0, "learning_rate": 1.1028164234815068e-07, "loss": 7.7734, "step": 65 }, { "epoch": 0.0, "learning_rate": 1.1876484560570071e-07, "loss": 7.9766, "step": 70 }, { "epoch": 0.01, "learning_rate": 1.272480488632508e-07, "loss": 7.8602, "step": 75 }, { "epoch": 0.01, "learning_rate": 1.3573125212080082e-07, "loss": 7.807, "step": 80 }, { "epoch": 0.01, "learning_rate": 1.4421445537835088e-07, "loss": 8.0211, "step": 85 }, { "epoch": 0.01, "learning_rate": 1.5269765863590092e-07, "loss": 7.8328, "step": 90 }, { "epoch": 0.01, "learning_rate": 1.6118086189345098e-07, "loss": 7.8969, "step": 95 }, { "epoch": 0.01, "learning_rate": 1.6966406515100101e-07, "loss": 7.6344, "step": 100 }, { "epoch": 0.01, "learning_rate": 1.7814726840855108e-07, "loss": 7.9227, "step": 105 }, { "epoch": 0.01, "learning_rate": 1.8663047166610114e-07, "loss": 7.4844, "step": 110 }, { "epoch": 0.01, "learning_rate": 1.9511367492365117e-07, "loss": 8.0719, "step": 115 }, { "epoch": 0.01, "learning_rate": 2.0359687818120123e-07, "loss": 8.0414, "step": 120 }, { "epoch": 0.01, "learning_rate": 2.120800814387513e-07, "loss": 7.8, "step": 125 }, { "epoch": 0.01, "learning_rate": 2.2056328469630136e-07, "loss": 7.8133, "step": 130 }, { "epoch": 0.01, "learning_rate": 2.290464879538514e-07, "loss": 7.5117, "step": 135 }, { "epoch": 0.01, "learning_rate": 2.3752969121140143e-07, "loss": 7.5898, "step": 140 }, { "epoch": 0.01, "learning_rate": 2.460128944689515e-07, "loss": 7.2891, "step": 145 }, { "epoch": 0.01, "learning_rate": 2.544960977265016e-07, "loss": 7.8922, "step": 150 }, { "epoch": 0.01, "learning_rate": 2.629793009840516e-07, "loss": 7.5492, "step": 155 }, { "epoch": 0.01, "learning_rate": 2.7146250424160164e-07, "loss": 7.8094, "step": 160 }, { "epoch": 0.01, "learning_rate": 2.799457074991517e-07, "loss": 7.2875, "step": 165 }, { "epoch": 0.01, "learning_rate": 2.8842891075670177e-07, "loss": 7.7367, "step": 170 }, { "epoch": 0.01, "learning_rate": 2.969121140142518e-07, "loss": 7.4102, "step": 175 }, { "epoch": 0.01, "learning_rate": 3.0539531727180184e-07, "loss": 7.5016, "step": 180 }, { "epoch": 0.01, "learning_rate": 3.1387852052935195e-07, "loss": 7.4539, "step": 185 }, { "epoch": 0.01, "learning_rate": 3.2236172378690196e-07, "loss": 6.9125, "step": 190 }, { "epoch": 0.01, "learning_rate": 3.30844927044452e-07, "loss": 7.543, "step": 195 }, { "epoch": 0.01, "learning_rate": 3.3932813030200203e-07, "loss": 6.6602, "step": 200 }, { "epoch": 0.01, "learning_rate": 3.478113335595521e-07, "loss": 6.6344, "step": 205 }, { "epoch": 0.01, "learning_rate": 3.5629453681710215e-07, "loss": 7.7047, "step": 210 }, { "epoch": 0.01, "learning_rate": 3.647777400746522e-07, "loss": 6.95, "step": 215 }, { "epoch": 0.01, "learning_rate": 3.7326094333220227e-07, "loss": 7.2555, "step": 220 }, { "epoch": 0.02, "learning_rate": 3.8174414658975233e-07, "loss": 6.5016, "step": 225 }, { "epoch": 0.02, "learning_rate": 3.9022734984730234e-07, "loss": 6.9258, "step": 230 }, { "epoch": 0.02, "learning_rate": 3.987105531048524e-07, "loss": 6.2562, "step": 235 }, { "epoch": 0.02, "learning_rate": 4.0719375636240247e-07, "loss": 7.0047, "step": 240 }, { "epoch": 0.02, "learning_rate": 4.1567695961995253e-07, "loss": 6.8961, "step": 245 }, { "epoch": 0.02, "learning_rate": 4.241601628775026e-07, "loss": 6.2578, "step": 250 }, { "epoch": 0.02, "learning_rate": 4.3264336613505265e-07, "loss": 6.3023, "step": 255 }, { "epoch": 0.02, "learning_rate": 4.411265693926027e-07, "loss": 7.107, "step": 260 }, { "epoch": 0.02, "learning_rate": 4.4960977265015277e-07, "loss": 6.4141, "step": 265 }, { "epoch": 0.02, "learning_rate": 4.580929759077028e-07, "loss": 5.675, "step": 270 }, { "epoch": 0.02, "learning_rate": 4.6657617916525284e-07, "loss": 6.9406, "step": 275 }, { "epoch": 0.02, "learning_rate": 4.7505938242280285e-07, "loss": 5.6281, "step": 280 }, { "epoch": 0.02, "learning_rate": 4.83542585680353e-07, "loss": 7.3, "step": 285 }, { "epoch": 0.02, "learning_rate": 4.92025788937903e-07, "loss": 5.5508, "step": 290 }, { "epoch": 0.02, "learning_rate": 5.005089921954531e-07, "loss": 5.8953, "step": 295 }, { "epoch": 0.02, "learning_rate": 5.089921954530031e-07, "loss": 6.8297, "step": 300 }, { "epoch": 0.02, "learning_rate": 5.174753987105532e-07, "loss": 5.9555, "step": 305 }, { "epoch": 0.02, "learning_rate": 5.259586019681032e-07, "loss": 6.4516, "step": 310 }, { "epoch": 0.02, "learning_rate": 5.344418052256532e-07, "loss": 5.5438, "step": 315 }, { "epoch": 0.02, "learning_rate": 5.429250084832033e-07, "loss": 5.8797, "step": 320 }, { "epoch": 0.02, "learning_rate": 5.514082117407533e-07, "loss": 5.3199, "step": 325 }, { "epoch": 0.02, "learning_rate": 5.598914149983034e-07, "loss": 5.1945, "step": 330 }, { "epoch": 0.02, "learning_rate": 5.683746182558535e-07, "loss": 5.8508, "step": 335 }, { "epoch": 0.02, "learning_rate": 5.768578215134035e-07, "loss": 5.0687, "step": 340 }, { "epoch": 0.02, "learning_rate": 5.853410247709536e-07, "loss": 5.0836, "step": 345 }, { "epoch": 0.02, "learning_rate": 5.938242280285035e-07, "loss": 4.9924, "step": 350 }, { "epoch": 0.02, "learning_rate": 6.023074312860536e-07, "loss": 5.5008, "step": 355 }, { "epoch": 0.02, "learning_rate": 6.107906345436037e-07, "loss": 6.3445, "step": 360 }, { "epoch": 0.02, "learning_rate": 6.192738378011537e-07, "loss": 4.1828, "step": 365 }, { "epoch": 0.03, "learning_rate": 6.277570410587039e-07, "loss": 4.684, "step": 370 }, { "epoch": 0.03, "learning_rate": 6.362402443162539e-07, "loss": 4.6918, "step": 375 }, { "epoch": 0.03, "learning_rate": 6.447234475738039e-07, "loss": 5.4055, "step": 380 }, { "epoch": 0.03, "learning_rate": 6.53206650831354e-07, "loss": 4.475, "step": 385 }, { "epoch": 0.03, "learning_rate": 6.61689854088904e-07, "loss": 6.0508, "step": 390 }, { "epoch": 0.03, "learning_rate": 6.70173057346454e-07, "loss": 5.1324, "step": 395 }, { "epoch": 0.03, "learning_rate": 6.786562606040041e-07, "loss": 4.4174, "step": 400 }, { "epoch": 0.03, "learning_rate": 6.871394638615541e-07, "loss": 4.6125, "step": 405 }, { "epoch": 0.03, "learning_rate": 6.956226671191042e-07, "loss": 4.427, "step": 410 }, { "epoch": 0.03, "learning_rate": 7.041058703766542e-07, "loss": 4.227, "step": 415 }, { "epoch": 0.03, "learning_rate": 7.125890736342043e-07, "loss": 4.5535, "step": 420 }, { "epoch": 0.03, "learning_rate": 7.210722768917544e-07, "loss": 5.8156, "step": 425 }, { "epoch": 0.03, "learning_rate": 7.295554801493044e-07, "loss": 5.4328, "step": 430 }, { "epoch": 0.03, "learning_rate": 7.380386834068545e-07, "loss": 3.5109, "step": 435 }, { "epoch": 0.03, "learning_rate": 7.465218866644045e-07, "loss": 4.973, "step": 440 }, { "epoch": 0.03, "learning_rate": 7.550050899219546e-07, "loss": 4.2766, "step": 445 }, { "epoch": 0.03, "learning_rate": 7.634882931795047e-07, "loss": 4.6086, "step": 450 }, { "epoch": 0.03, "learning_rate": 7.719714964370546e-07, "loss": 4.8164, "step": 455 }, { "epoch": 0.03, "learning_rate": 7.804546996946047e-07, "loss": 4.2274, "step": 460 }, { "epoch": 0.03, "learning_rate": 7.889379029521547e-07, "loss": 3.1027, "step": 465 }, { "epoch": 0.03, "learning_rate": 7.974211062097048e-07, "loss": 4.1688, "step": 470 }, { "epoch": 0.03, "learning_rate": 8.059043094672549e-07, "loss": 4.3023, "step": 475 }, { "epoch": 0.03, "learning_rate": 8.143875127248049e-07, "loss": 3.7992, "step": 480 }, { "epoch": 0.03, "learning_rate": 8.22870715982355e-07, "loss": 4.5901, "step": 485 }, { "epoch": 0.03, "learning_rate": 8.313539192399051e-07, "loss": 5.0078, "step": 490 }, { "epoch": 0.03, "learning_rate": 8.398371224974551e-07, "loss": 5.569, "step": 495 }, { "epoch": 0.03, "learning_rate": 8.483203257550052e-07, "loss": 4.348, "step": 500 }, { "epoch": 0.03, "learning_rate": 8.568035290125552e-07, "loss": 2.9572, "step": 505 }, { "epoch": 0.03, "learning_rate": 8.652867322701053e-07, "loss": 5.2523, "step": 510 }, { "epoch": 0.03, "learning_rate": 8.737699355276554e-07, "loss": 4.6254, "step": 515 }, { "epoch": 0.04, "learning_rate": 8.822531387852054e-07, "loss": 4.4797, "step": 520 }, { "epoch": 0.04, "learning_rate": 8.907363420427555e-07, "loss": 5.9328, "step": 525 }, { "epoch": 0.04, "learning_rate": 8.992195453003055e-07, "loss": 3.643, "step": 530 }, { "epoch": 0.04, "learning_rate": 9.077027485578556e-07, "loss": 3.6324, "step": 535 }, { "epoch": 0.04, "learning_rate": 9.161859518154056e-07, "loss": 3.4383, "step": 540 }, { "epoch": 0.04, "learning_rate": 9.246691550729556e-07, "loss": 4.268, "step": 545 }, { "epoch": 0.04, "learning_rate": 9.331523583305057e-07, "loss": 3.7355, "step": 550 }, { "epoch": 0.04, "learning_rate": 9.416355615880556e-07, "loss": 5.0988, "step": 555 }, { "epoch": 0.04, "learning_rate": 9.501187648456057e-07, "loss": 4.5812, "step": 560 }, { "epoch": 0.04, "learning_rate": 9.586019681031559e-07, "loss": 4.841, "step": 565 }, { "epoch": 0.04, "learning_rate": 9.67085171360706e-07, "loss": 5.7066, "step": 570 }, { "epoch": 0.04, "learning_rate": 9.75568374618256e-07, "loss": 4.8145, "step": 575 }, { "epoch": 0.04, "learning_rate": 9.84051577875806e-07, "loss": 3.1859, "step": 580 }, { "epoch": 0.04, "learning_rate": 9.925347811333561e-07, "loss": 4.943, "step": 585 }, { "epoch": 0.04, "learning_rate": 1.0010179843909062e-06, "loss": 4.8863, "step": 590 }, { "epoch": 0.04, "learning_rate": 1.0095011876484562e-06, "loss": 4.6398, "step": 595 }, { "epoch": 0.04, "learning_rate": 1.0179843909060063e-06, "loss": 5.141, "step": 600 }, { "epoch": 0.04, "learning_rate": 1.0264675941635564e-06, "loss": 3.3991, "step": 605 }, { "epoch": 0.04, "learning_rate": 1.0349507974211064e-06, "loss": 3.7336, "step": 610 }, { "epoch": 0.04, "learning_rate": 1.0434340006786565e-06, "loss": 3.9914, "step": 615 }, { "epoch": 0.04, "learning_rate": 1.0519172039362063e-06, "loss": 3.0272, "step": 620 }, { "epoch": 0.04, "learning_rate": 1.0604004071937564e-06, "loss": 4.2422, "step": 625 }, { "epoch": 0.04, "learning_rate": 1.0688836104513065e-06, "loss": 4.4305, "step": 630 }, { "epoch": 0.04, "learning_rate": 1.0773668137088565e-06, "loss": 3.766, "step": 635 }, { "epoch": 0.04, "learning_rate": 1.0858500169664066e-06, "loss": 3.7687, "step": 640 }, { "epoch": 0.04, "learning_rate": 1.0943332202239566e-06, "loss": 4.7512, "step": 645 }, { "epoch": 0.04, "learning_rate": 1.1028164234815067e-06, "loss": 4.407, "step": 650 }, { "epoch": 0.04, "learning_rate": 1.1112996267390568e-06, "loss": 4.5865, "step": 655 }, { "epoch": 0.04, "learning_rate": 1.1197828299966068e-06, "loss": 4.8359, "step": 660 }, { "epoch": 0.05, "learning_rate": 1.1282660332541569e-06, "loss": 4.4465, "step": 665 }, { "epoch": 0.05, "learning_rate": 1.136749236511707e-06, "loss": 3.898, "step": 670 }, { "epoch": 0.05, "learning_rate": 1.145232439769257e-06, "loss": 5.5387, "step": 675 }, { "epoch": 0.05, "learning_rate": 1.153715643026807e-06, "loss": 3.4188, "step": 680 }, { "epoch": 0.05, "learning_rate": 1.1621988462843571e-06, "loss": 2.153, "step": 685 }, { "epoch": 0.05, "learning_rate": 1.1706820495419072e-06, "loss": 4.5308, "step": 690 }, { "epoch": 0.05, "learning_rate": 1.1791652527994572e-06, "loss": 3.648, "step": 695 }, { "epoch": 0.05, "learning_rate": 1.187648456057007e-06, "loss": 4.0102, "step": 700 }, { "epoch": 0.05, "learning_rate": 1.1961316593145572e-06, "loss": 5.2145, "step": 705 }, { "epoch": 0.05, "learning_rate": 1.2046148625721072e-06, "loss": 3.2852, "step": 710 }, { "epoch": 0.05, "learning_rate": 1.2130980658296573e-06, "loss": 3.0016, "step": 715 }, { "epoch": 0.05, "learning_rate": 1.2215812690872073e-06, "loss": 3.7799, "step": 720 }, { "epoch": 0.05, "learning_rate": 1.2300644723447574e-06, "loss": 4.2594, "step": 725 }, { "epoch": 0.05, "learning_rate": 1.2385476756023075e-06, "loss": 4.2043, "step": 730 }, { "epoch": 0.05, "learning_rate": 1.2470308788598575e-06, "loss": 4.5301, "step": 735 }, { "epoch": 0.05, "learning_rate": 1.2555140821174078e-06, "loss": 3.1496, "step": 740 }, { "epoch": 0.05, "learning_rate": 1.2639972853749577e-06, "loss": 3.4547, "step": 745 }, { "epoch": 0.05, "learning_rate": 1.2724804886325077e-06, "loss": 5.0543, "step": 750 }, { "epoch": 0.05, "learning_rate": 1.2809636918900578e-06, "loss": 4.2965, "step": 755 }, { "epoch": 0.05, "learning_rate": 1.2894468951476078e-06, "loss": 4.5699, "step": 760 }, { "epoch": 0.05, "learning_rate": 1.297930098405158e-06, "loss": 4.4707, "step": 765 }, { "epoch": 0.05, "learning_rate": 1.306413301662708e-06, "loss": 3.7352, "step": 770 }, { "epoch": 0.05, "learning_rate": 1.314896504920258e-06, "loss": 3.8457, "step": 775 }, { "epoch": 0.05, "learning_rate": 1.323379708177808e-06, "loss": 2.6625, "step": 780 }, { "epoch": 0.05, "learning_rate": 1.3318629114353581e-06, "loss": 3.4609, "step": 785 }, { "epoch": 0.05, "learning_rate": 1.340346114692908e-06, "loss": 4.4027, "step": 790 }, { "epoch": 0.05, "learning_rate": 1.3488293179504583e-06, "loss": 3.1041, "step": 795 }, { "epoch": 0.05, "learning_rate": 1.3573125212080081e-06, "loss": 3.7766, "step": 800 }, { "epoch": 0.05, "learning_rate": 1.3657957244655584e-06, "loss": 2.6936, "step": 805 }, { "epoch": 0.05, "learning_rate": 1.3742789277231082e-06, "loss": 4.0422, "step": 810 }, { "epoch": 0.06, "learning_rate": 1.3827621309806585e-06, "loss": 3.9448, "step": 815 }, { "epoch": 0.06, "learning_rate": 1.3912453342382084e-06, "loss": 3.9325, "step": 820 }, { "epoch": 0.06, "learning_rate": 1.3997285374957584e-06, "loss": 2.9727, "step": 825 }, { "epoch": 0.06, "learning_rate": 1.4082117407533085e-06, "loss": 4.7302, "step": 830 }, { "epoch": 0.06, "learning_rate": 1.4166949440108585e-06, "loss": 3.9946, "step": 835 }, { "epoch": 0.06, "learning_rate": 1.4251781472684086e-06, "loss": 4.4797, "step": 840 }, { "epoch": 0.06, "learning_rate": 1.4336613505259587e-06, "loss": 3.1293, "step": 845 }, { "epoch": 0.06, "learning_rate": 1.4421445537835087e-06, "loss": 2.9921, "step": 850 }, { "epoch": 0.06, "learning_rate": 1.4506277570410588e-06, "loss": 3.6516, "step": 855 }, { "epoch": 0.06, "learning_rate": 1.4591109602986088e-06, "loss": 2.5961, "step": 860 }, { "epoch": 0.06, "learning_rate": 1.467594163556159e-06, "loss": 3.5898, "step": 865 }, { "epoch": 0.06, "learning_rate": 1.476077366813709e-06, "loss": 2.7932, "step": 870 }, { "epoch": 0.06, "learning_rate": 1.484560570071259e-06, "loss": 4.2238, "step": 875 }, { "epoch": 0.06, "learning_rate": 1.493043773328809e-06, "loss": 3.75, "step": 880 }, { "epoch": 0.06, "learning_rate": 1.5015269765863592e-06, "loss": 3.6043, "step": 885 }, { "epoch": 0.06, "learning_rate": 1.5100101798439092e-06, "loss": 3.5203, "step": 890 }, { "epoch": 0.06, "learning_rate": 1.5184933831014593e-06, "loss": 3.1963, "step": 895 }, { "epoch": 0.06, "learning_rate": 1.5269765863590093e-06, "loss": 3.0615, "step": 900 }, { "epoch": 0.06, "learning_rate": 1.5354597896165594e-06, "loss": 2.8456, "step": 905 }, { "epoch": 0.06, "learning_rate": 1.5439429928741092e-06, "loss": 2.5942, "step": 910 }, { "epoch": 0.06, "learning_rate": 1.5524261961316595e-06, "loss": 2.1173, "step": 915 }, { "epoch": 0.06, "learning_rate": 1.5609093993892094e-06, "loss": 3.3949, "step": 920 }, { "epoch": 0.06, "learning_rate": 1.5693926026467596e-06, "loss": 2.9775, "step": 925 }, { "epoch": 0.06, "learning_rate": 1.5778758059043095e-06, "loss": 2.8984, "step": 930 }, { "epoch": 0.06, "learning_rate": 1.5863590091618598e-06, "loss": 4.4586, "step": 935 }, { "epoch": 0.06, "learning_rate": 1.5948422124194096e-06, "loss": 2.5443, "step": 940 }, { "epoch": 0.06, "learning_rate": 1.6033254156769599e-06, "loss": 3.3309, "step": 945 }, { "epoch": 0.06, "learning_rate": 1.6118086189345097e-06, "loss": 2.7556, "step": 950 }, { "epoch": 0.06, "learning_rate": 1.62029182219206e-06, "loss": 3.0418, "step": 955 }, { "epoch": 0.07, "learning_rate": 1.6287750254496099e-06, "loss": 3.6102, "step": 960 }, { "epoch": 0.07, "learning_rate": 1.6372582287071601e-06, "loss": 4.184, "step": 965 }, { "epoch": 0.07, "learning_rate": 1.64574143196471e-06, "loss": 2.7004, "step": 970 }, { "epoch": 0.07, "learning_rate": 1.6542246352222603e-06, "loss": 2.9357, "step": 975 }, { "epoch": 0.07, "learning_rate": 1.6627078384798101e-06, "loss": 2.9758, "step": 980 }, { "epoch": 0.07, "learning_rate": 1.6711910417373604e-06, "loss": 3.477, "step": 985 }, { "epoch": 0.07, "learning_rate": 1.6796742449949102e-06, "loss": 4.2041, "step": 990 }, { "epoch": 0.07, "learning_rate": 1.68815744825246e-06, "loss": 3.4752, "step": 995 }, { "epoch": 0.07, "learning_rate": 1.6966406515100104e-06, "loss": 3.6211, "step": 1000 }, { "epoch": 0.07, "learning_rate": 1.7051238547675602e-06, "loss": 2.5957, "step": 1005 }, { "epoch": 0.07, "learning_rate": 1.7136070580251105e-06, "loss": 4.4422, "step": 1010 }, { "epoch": 0.07, "learning_rate": 1.7220902612826603e-06, "loss": 3.1666, "step": 1015 }, { "epoch": 0.07, "learning_rate": 1.7305734645402106e-06, "loss": 3.4514, "step": 1020 }, { "epoch": 0.07, "learning_rate": 1.7390566677977604e-06, "loss": 3.3086, "step": 1025 }, { "epoch": 0.07, "learning_rate": 1.7475398710553107e-06, "loss": 4.9145, "step": 1030 }, { "epoch": 0.07, "learning_rate": 1.7560230743128606e-06, "loss": 3.2045, "step": 1035 }, { "epoch": 0.07, "learning_rate": 1.7645062775704108e-06, "loss": 3.0834, "step": 1040 }, { "epoch": 0.07, "learning_rate": 1.7729894808279607e-06, "loss": 3.2967, "step": 1045 }, { "epoch": 0.07, "learning_rate": 1.781472684085511e-06, "loss": 2.8278, "step": 1050 }, { "epoch": 0.07, "learning_rate": 1.7899558873430608e-06, "loss": 2.8638, "step": 1055 }, { "epoch": 0.07, "learning_rate": 1.798439090600611e-06, "loss": 3.173, "step": 1060 }, { "epoch": 0.07, "learning_rate": 1.806922293858161e-06, "loss": 3.9258, "step": 1065 }, { "epoch": 0.07, "learning_rate": 1.8154054971157112e-06, "loss": 3.1956, "step": 1070 }, { "epoch": 0.07, "learning_rate": 1.823888700373261e-06, "loss": 4.3852, "step": 1075 }, { "epoch": 0.07, "learning_rate": 1.8323719036308111e-06, "loss": 3.2195, "step": 1080 }, { "epoch": 0.07, "learning_rate": 1.8408551068883612e-06, "loss": 3.5129, "step": 1085 }, { "epoch": 0.07, "learning_rate": 1.8493383101459112e-06, "loss": 3.2395, "step": 1090 }, { "epoch": 0.07, "learning_rate": 1.8578215134034613e-06, "loss": 1.917, "step": 1095 }, { "epoch": 0.07, "learning_rate": 1.8663047166610114e-06, "loss": 4.0586, "step": 1100 }, { "epoch": 0.07, "learning_rate": 1.8747879199185614e-06, "loss": 2.9253, "step": 1105 }, { "epoch": 0.08, "learning_rate": 1.8832711231761113e-06, "loss": 3.3449, "step": 1110 }, { "epoch": 0.08, "learning_rate": 1.8917543264336616e-06, "loss": 4.3992, "step": 1115 }, { "epoch": 0.08, "learning_rate": 1.9002375296912114e-06, "loss": 2.4142, "step": 1120 }, { "epoch": 0.08, "learning_rate": 1.908720732948762e-06, "loss": 1.7545, "step": 1125 }, { "epoch": 0.08, "learning_rate": 1.9172039362063117e-06, "loss": 4.3359, "step": 1130 }, { "epoch": 0.08, "learning_rate": 1.925687139463862e-06, "loss": 3.8922, "step": 1135 }, { "epoch": 0.08, "learning_rate": 1.934170342721412e-06, "loss": 4.0648, "step": 1140 }, { "epoch": 0.08, "learning_rate": 1.942653545978962e-06, "loss": 2.757, "step": 1145 }, { "epoch": 0.08, "learning_rate": 1.951136749236512e-06, "loss": 2.5563, "step": 1150 }, { "epoch": 0.08, "learning_rate": 1.959619952494062e-06, "loss": 3.4707, "step": 1155 }, { "epoch": 0.08, "learning_rate": 1.968103155751612e-06, "loss": 3.1654, "step": 1160 }, { "epoch": 0.08, "learning_rate": 1.976586359009162e-06, "loss": 3.7863, "step": 1165 }, { "epoch": 0.08, "learning_rate": 1.9850695622667122e-06, "loss": 2.0713, "step": 1170 }, { "epoch": 0.08, "learning_rate": 1.993552765524262e-06, "loss": 3.5117, "step": 1175 }, { "epoch": 0.08, "learning_rate": 2.0020359687818123e-06, "loss": 2.095, "step": 1180 }, { "epoch": 0.08, "learning_rate": 2.010519172039362e-06, "loss": 1.9218, "step": 1185 }, { "epoch": 0.08, "learning_rate": 2.0190023752969125e-06, "loss": 3.3027, "step": 1190 }, { "epoch": 0.08, "learning_rate": 2.0274855785544623e-06, "loss": 1.8242, "step": 1195 }, { "epoch": 0.08, "learning_rate": 2.0359687818120126e-06, "loss": 2.563, "step": 1200 }, { "epoch": 0.08, "learning_rate": 2.0444519850695624e-06, "loss": 1.9848, "step": 1205 }, { "epoch": 0.08, "learning_rate": 2.0529351883271127e-06, "loss": 2.2807, "step": 1210 }, { "epoch": 0.08, "learning_rate": 2.0614183915846626e-06, "loss": 2.5801, "step": 1215 }, { "epoch": 0.08, "learning_rate": 2.069901594842213e-06, "loss": 2.7947, "step": 1220 }, { "epoch": 0.08, "learning_rate": 2.0783847980997627e-06, "loss": 2.4934, "step": 1225 }, { "epoch": 0.08, "learning_rate": 2.086868001357313e-06, "loss": 2.1461, "step": 1230 }, { "epoch": 0.08, "learning_rate": 2.095351204614863e-06, "loss": 4.1415, "step": 1235 }, { "epoch": 0.08, "learning_rate": 2.1038344078724127e-06, "loss": 2.7672, "step": 1240 }, { "epoch": 0.08, "learning_rate": 2.112317611129963e-06, "loss": 2.557, "step": 1245 }, { "epoch": 0.08, "learning_rate": 2.1208008143875128e-06, "loss": 2.7213, "step": 1250 }, { "epoch": 0.09, "learning_rate": 2.129284017645063e-06, "loss": 2.1246, "step": 1255 }, { "epoch": 0.09, "learning_rate": 2.137767220902613e-06, "loss": 2.34, "step": 1260 }, { "epoch": 0.09, "learning_rate": 2.146250424160163e-06, "loss": 3.4021, "step": 1265 }, { "epoch": 0.09, "learning_rate": 2.154733627417713e-06, "loss": 3.1895, "step": 1270 }, { "epoch": 0.09, "learning_rate": 2.1632168306752633e-06, "loss": 3.5379, "step": 1275 }, { "epoch": 0.09, "learning_rate": 2.171700033932813e-06, "loss": 2.5277, "step": 1280 }, { "epoch": 0.09, "learning_rate": 2.1801832371903634e-06, "loss": 2.7035, "step": 1285 }, { "epoch": 0.09, "learning_rate": 2.1886664404479133e-06, "loss": 2.6008, "step": 1290 }, { "epoch": 0.09, "learning_rate": 2.1971496437054635e-06, "loss": 2.7986, "step": 1295 }, { "epoch": 0.09, "learning_rate": 2.2056328469630134e-06, "loss": 2.1992, "step": 1300 }, { "epoch": 0.09, "learning_rate": 2.2141160502205637e-06, "loss": 2.9004, "step": 1305 }, { "epoch": 0.09, "learning_rate": 2.2225992534781135e-06, "loss": 1.8229, "step": 1310 }, { "epoch": 0.09, "learning_rate": 2.231082456735664e-06, "loss": 3.165, "step": 1315 }, { "epoch": 0.09, "learning_rate": 2.2395656599932136e-06, "loss": 3.0629, "step": 1320 }, { "epoch": 0.09, "learning_rate": 2.2480488632507635e-06, "loss": 3.2926, "step": 1325 }, { "epoch": 0.09, "learning_rate": 2.2565320665083138e-06, "loss": 3.2838, "step": 1330 }, { "epoch": 0.09, "learning_rate": 2.2650152697658636e-06, "loss": 3.6594, "step": 1335 }, { "epoch": 0.09, "learning_rate": 2.273498473023414e-06, "loss": 1.6854, "step": 1340 }, { "epoch": 0.09, "learning_rate": 2.2819816762809637e-06, "loss": 2.0272, "step": 1345 }, { "epoch": 0.09, "learning_rate": 2.290464879538514e-06, "loss": 1.8784, "step": 1350 }, { "epoch": 0.09, "learning_rate": 2.298948082796064e-06, "loss": 3.775, "step": 1355 }, { "epoch": 0.09, "learning_rate": 2.307431286053614e-06, "loss": 2.1449, "step": 1360 }, { "epoch": 0.09, "learning_rate": 2.315914489311164e-06, "loss": 2.2746, "step": 1365 }, { "epoch": 0.09, "learning_rate": 2.3243976925687143e-06, "loss": 3.4684, "step": 1370 }, { "epoch": 0.09, "learning_rate": 2.332880895826264e-06, "loss": 2.7242, "step": 1375 }, { "epoch": 0.09, "learning_rate": 2.3413640990838144e-06, "loss": 2.6856, "step": 1380 }, { "epoch": 0.09, "learning_rate": 2.3498473023413642e-06, "loss": 3.2816, "step": 1385 }, { "epoch": 0.09, "learning_rate": 2.3583305055989145e-06, "loss": 2.8984, "step": 1390 }, { "epoch": 0.09, "learning_rate": 2.3668137088564643e-06, "loss": 2.9697, "step": 1395 }, { "epoch": 0.1, "learning_rate": 2.375296912114014e-06, "loss": 2.7007, "step": 1400 }, { "epoch": 0.1, "learning_rate": 2.3837801153715645e-06, "loss": 3.7279, "step": 1405 }, { "epoch": 0.1, "learning_rate": 2.3922633186291143e-06, "loss": 3.626, "step": 1410 }, { "epoch": 0.1, "learning_rate": 2.4007465218866646e-06, "loss": 2.6211, "step": 1415 }, { "epoch": 0.1, "learning_rate": 2.4092297251442144e-06, "loss": 2.8492, "step": 1420 }, { "epoch": 0.1, "learning_rate": 2.4177129284017647e-06, "loss": 3.2515, "step": 1425 }, { "epoch": 0.1, "learning_rate": 2.4261961316593146e-06, "loss": 2.6645, "step": 1430 }, { "epoch": 0.1, "learning_rate": 2.434679334916865e-06, "loss": 2.2677, "step": 1435 }, { "epoch": 0.1, "learning_rate": 2.4431625381744147e-06, "loss": 2.5316, "step": 1440 }, { "epoch": 0.1, "learning_rate": 2.451645741431965e-06, "loss": 1.8498, "step": 1445 }, { "epoch": 0.1, "learning_rate": 2.460128944689515e-06, "loss": 1.5558, "step": 1450 }, { "epoch": 0.1, "learning_rate": 2.468612147947065e-06, "loss": 2.5236, "step": 1455 }, { "epoch": 0.1, "learning_rate": 2.477095351204615e-06, "loss": 3.0307, "step": 1460 }, { "epoch": 0.1, "learning_rate": 2.485578554462165e-06, "loss": 1.6035, "step": 1465 }, { "epoch": 0.1, "learning_rate": 2.494061757719715e-06, "loss": 2.6655, "step": 1470 }, { "epoch": 0.1, "learning_rate": 2.502544960977265e-06, "loss": 3.7184, "step": 1475 }, { "epoch": 0.1, "learning_rate": 2.5110281642348156e-06, "loss": 2.1453, "step": 1480 }, { "epoch": 0.1, "learning_rate": 2.5195113674923655e-06, "loss": 2.3629, "step": 1485 }, { "epoch": 0.1, "learning_rate": 2.5279945707499153e-06, "loss": 1.9499, "step": 1490 }, { "epoch": 0.1, "learning_rate": 2.536477774007465e-06, "loss": 2.6299, "step": 1495 }, { "epoch": 0.1, "learning_rate": 2.5449609772650154e-06, "loss": 3.5132, "step": 1500 }, { "epoch": 0.1, "learning_rate": 2.5534441805225657e-06, "loss": 2.4019, "step": 1505 }, { "epoch": 0.1, "learning_rate": 2.5619273837801155e-06, "loss": 2.1061, "step": 1510 }, { "epoch": 0.1, "learning_rate": 2.5704105870376654e-06, "loss": 2.5825, "step": 1515 }, { "epoch": 0.1, "learning_rate": 2.5788937902952157e-06, "loss": 1.6013, "step": 1520 }, { "epoch": 0.1, "learning_rate": 2.587376993552766e-06, "loss": 3.059, "step": 1525 }, { "epoch": 0.1, "learning_rate": 2.595860196810316e-06, "loss": 2.733, "step": 1530 }, { "epoch": 0.1, "learning_rate": 2.6043434000678656e-06, "loss": 3.1348, "step": 1535 }, { "epoch": 0.1, "learning_rate": 2.612826603325416e-06, "loss": 1.8278, "step": 1540 }, { "epoch": 0.1, "learning_rate": 2.621309806582966e-06, "loss": 2.5457, "step": 1545 }, { "epoch": 0.11, "learning_rate": 2.629793009840516e-06, "loss": 2.0575, "step": 1550 }, { "epoch": 0.11, "learning_rate": 2.638276213098066e-06, "loss": 2.6969, "step": 1555 }, { "epoch": 0.11, "learning_rate": 2.646759416355616e-06, "loss": 2.543, "step": 1560 }, { "epoch": 0.11, "learning_rate": 2.655242619613166e-06, "loss": 2.2389, "step": 1565 }, { "epoch": 0.11, "learning_rate": 2.6637258228707163e-06, "loss": 2.369, "step": 1570 }, { "epoch": 0.11, "learning_rate": 2.672209026128266e-06, "loss": 2.2446, "step": 1575 }, { "epoch": 0.11, "learning_rate": 2.680692229385816e-06, "loss": 2.9564, "step": 1580 }, { "epoch": 0.11, "learning_rate": 2.6891754326433663e-06, "loss": 3.0527, "step": 1585 }, { "epoch": 0.11, "learning_rate": 2.6976586359009165e-06, "loss": 3.4801, "step": 1590 }, { "epoch": 0.11, "learning_rate": 2.7061418391584664e-06, "loss": 1.8436, "step": 1595 }, { "epoch": 0.11, "learning_rate": 2.7146250424160162e-06, "loss": 2.5486, "step": 1600 }, { "epoch": 0.11, "learning_rate": 2.7231082456735665e-06, "loss": 1.811, "step": 1605 }, { "epoch": 0.11, "learning_rate": 2.7315914489311168e-06, "loss": 2.3609, "step": 1610 }, { "epoch": 0.11, "learning_rate": 2.7400746521886666e-06, "loss": 2.3194, "step": 1615 }, { "epoch": 0.11, "learning_rate": 2.7485578554462165e-06, "loss": 2.4229, "step": 1620 }, { "epoch": 0.11, "learning_rate": 2.7570410587037667e-06, "loss": 2.9533, "step": 1625 }, { "epoch": 0.11, "learning_rate": 2.765524261961317e-06, "loss": 1.9473, "step": 1630 }, { "epoch": 0.11, "learning_rate": 2.774007465218867e-06, "loss": 2.5977, "step": 1635 }, { "epoch": 0.11, "learning_rate": 2.7824906684764167e-06, "loss": 1.9589, "step": 1640 }, { "epoch": 0.11, "learning_rate": 2.790973871733967e-06, "loss": 2.4553, "step": 1645 }, { "epoch": 0.11, "learning_rate": 2.799457074991517e-06, "loss": 2.8572, "step": 1650 }, { "epoch": 0.11, "learning_rate": 2.807940278249067e-06, "loss": 2.5994, "step": 1655 }, { "epoch": 0.11, "learning_rate": 2.816423481506617e-06, "loss": 3.3534, "step": 1660 }, { "epoch": 0.11, "learning_rate": 2.8249066847641672e-06, "loss": 2.5073, "step": 1665 }, { "epoch": 0.11, "learning_rate": 2.833389888021717e-06, "loss": 3.5906, "step": 1670 }, { "epoch": 0.11, "learning_rate": 2.8418730912792674e-06, "loss": 2.3825, "step": 1675 }, { "epoch": 0.11, "learning_rate": 2.850356294536817e-06, "loss": 2.418, "step": 1680 }, { "epoch": 0.11, "learning_rate": 2.8588394977943675e-06, "loss": 1.7972, "step": 1685 }, { "epoch": 0.11, "learning_rate": 2.8673227010519173e-06, "loss": 1.9553, "step": 1690 }, { "epoch": 0.12, "learning_rate": 2.8758059043094676e-06, "loss": 3.2061, "step": 1695 }, { "epoch": 0.12, "learning_rate": 2.8842891075670175e-06, "loss": 3.2529, "step": 1700 }, { "epoch": 0.12, "learning_rate": 2.8927723108245677e-06, "loss": 2.2563, "step": 1705 }, { "epoch": 0.12, "learning_rate": 2.9012555140821176e-06, "loss": 3.4211, "step": 1710 }, { "epoch": 0.12, "learning_rate": 2.909738717339668e-06, "loss": 2.365, "step": 1715 }, { "epoch": 0.12, "learning_rate": 2.9182219205972177e-06, "loss": 2.8943, "step": 1720 }, { "epoch": 0.12, "learning_rate": 2.926705123854768e-06, "loss": 1.7405, "step": 1725 }, { "epoch": 0.12, "learning_rate": 2.935188327112318e-06, "loss": 2.3562, "step": 1730 }, { "epoch": 0.12, "learning_rate": 2.9436715303698677e-06, "loss": 2.2449, "step": 1735 }, { "epoch": 0.12, "learning_rate": 2.952154733627418e-06, "loss": 2.5305, "step": 1740 }, { "epoch": 0.12, "learning_rate": 2.9606379368849682e-06, "loss": 2.5051, "step": 1745 }, { "epoch": 0.12, "learning_rate": 2.969121140142518e-06, "loss": 3.0462, "step": 1750 }, { "epoch": 0.12, "learning_rate": 2.977604343400068e-06, "loss": 3.4081, "step": 1755 }, { "epoch": 0.12, "learning_rate": 2.986087546657618e-06, "loss": 1.685, "step": 1760 }, { "epoch": 0.12, "learning_rate": 2.9945707499151685e-06, "loss": 2.5854, "step": 1765 }, { "epoch": 0.12, "learning_rate": 3.0030539531727183e-06, "loss": 1.9157, "step": 1770 }, { "epoch": 0.12, "learning_rate": 3.011537156430268e-06, "loss": 2.4659, "step": 1775 }, { "epoch": 0.12, "learning_rate": 3.0200203596878184e-06, "loss": 1.8663, "step": 1780 }, { "epoch": 0.12, "learning_rate": 3.0285035629453687e-06, "loss": 1.8562, "step": 1785 }, { "epoch": 0.12, "learning_rate": 3.0369867662029186e-06, "loss": 2.1948, "step": 1790 }, { "epoch": 0.12, "learning_rate": 3.0454699694604684e-06, "loss": 1.448, "step": 1795 }, { "epoch": 0.12, "learning_rate": 3.0539531727180187e-06, "loss": 2.8048, "step": 1800 }, { "epoch": 0.12, "learning_rate": 3.062436375975569e-06, "loss": 1.7508, "step": 1805 }, { "epoch": 0.12, "learning_rate": 3.070919579233119e-06, "loss": 2.5616, "step": 1810 }, { "epoch": 0.12, "learning_rate": 3.0794027824906686e-06, "loss": 3.1537, "step": 1815 }, { "epoch": 0.12, "learning_rate": 3.0878859857482185e-06, "loss": 2.6078, "step": 1820 }, { "epoch": 0.12, "learning_rate": 3.096369189005769e-06, "loss": 1.8651, "step": 1825 }, { "epoch": 0.12, "learning_rate": 3.104852392263319e-06, "loss": 2.8188, "step": 1830 }, { "epoch": 0.12, "learning_rate": 3.113335595520869e-06, "loss": 2.5123, "step": 1835 }, { "epoch": 0.12, "learning_rate": 3.1218187987784187e-06, "loss": 2.9451, "step": 1840 }, { "epoch": 0.13, "learning_rate": 3.1303020020359694e-06, "loss": 2.735, "step": 1845 }, { "epoch": 0.13, "learning_rate": 3.1387852052935193e-06, "loss": 2.084, "step": 1850 }, { "epoch": 0.13, "learning_rate": 3.147268408551069e-06, "loss": 2.9102, "step": 1855 }, { "epoch": 0.13, "learning_rate": 3.155751611808619e-06, "loss": 3.2244, "step": 1860 }, { "epoch": 0.13, "learning_rate": 3.1642348150661693e-06, "loss": 2.2596, "step": 1865 }, { "epoch": 0.13, "learning_rate": 3.1727180183237195e-06, "loss": 2.5423, "step": 1870 }, { "epoch": 0.13, "learning_rate": 3.1812012215812694e-06, "loss": 1.9754, "step": 1875 }, { "epoch": 0.13, "learning_rate": 3.1896844248388192e-06, "loss": 2.8066, "step": 1880 }, { "epoch": 0.13, "learning_rate": 3.1981676280963695e-06, "loss": 2.5214, "step": 1885 }, { "epoch": 0.13, "learning_rate": 3.2066508313539198e-06, "loss": 4.0949, "step": 1890 }, { "epoch": 0.13, "learning_rate": 3.2151340346114696e-06, "loss": 3.7504, "step": 1895 }, { "epoch": 0.13, "learning_rate": 3.2236172378690195e-06, "loss": 1.4069, "step": 1900 }, { "epoch": 0.13, "learning_rate": 3.2321004411265693e-06, "loss": 0.8493, "step": 1905 }, { "epoch": 0.13, "learning_rate": 3.24058364438412e-06, "loss": 2.5564, "step": 1910 }, { "epoch": 0.13, "learning_rate": 3.24906684764167e-06, "loss": 1.3595, "step": 1915 }, { "epoch": 0.13, "learning_rate": 3.2575500508992197e-06, "loss": 2.1167, "step": 1920 }, { "epoch": 0.13, "learning_rate": 3.2660332541567696e-06, "loss": 2.593, "step": 1925 }, { "epoch": 0.13, "learning_rate": 3.2745164574143203e-06, "loss": 2.746, "step": 1930 }, { "epoch": 0.13, "learning_rate": 3.28299966067187e-06, "loss": 1.7012, "step": 1935 }, { "epoch": 0.13, "learning_rate": 3.29148286392942e-06, "loss": 2.3498, "step": 1940 }, { "epoch": 0.13, "learning_rate": 3.29996606718697e-06, "loss": 2.3722, "step": 1945 }, { "epoch": 0.13, "learning_rate": 3.3084492704445205e-06, "loss": 3.0887, "step": 1950 }, { "epoch": 0.13, "learning_rate": 3.3169324737020704e-06, "loss": 1.9668, "step": 1955 }, { "epoch": 0.13, "learning_rate": 3.3254156769596202e-06, "loss": 2.7149, "step": 1960 }, { "epoch": 0.13, "learning_rate": 3.33389888021717e-06, "loss": 1.6986, "step": 1965 }, { "epoch": 0.13, "learning_rate": 3.3423820834747208e-06, "loss": 2.0309, "step": 1970 }, { "epoch": 0.13, "learning_rate": 3.3508652867322706e-06, "loss": 1.9683, "step": 1975 }, { "epoch": 0.13, "learning_rate": 3.3593484899898205e-06, "loss": 1.9985, "step": 1980 }, { "epoch": 0.13, "learning_rate": 3.3678316932473703e-06, "loss": 2.2535, "step": 1985 }, { "epoch": 0.14, "learning_rate": 3.37631489650492e-06, "loss": 1.2407, "step": 1990 }, { "epoch": 0.14, "learning_rate": 3.384798099762471e-06, "loss": 1.709, "step": 1995 }, { "epoch": 0.14, "learning_rate": 3.3932813030200207e-06, "loss": 1.9041, "step": 2000 }, { "epoch": 0.14, "learning_rate": 3.4017645062775706e-06, "loss": 3.1398, "step": 2005 }, { "epoch": 0.14, "learning_rate": 3.4102477095351204e-06, "loss": 2.7436, "step": 2010 }, { "epoch": 0.14, "learning_rate": 3.418730912792671e-06, "loss": 2.3188, "step": 2015 }, { "epoch": 0.14, "learning_rate": 3.427214116050221e-06, "loss": 1.5684, "step": 2020 }, { "epoch": 0.14, "learning_rate": 3.435697319307771e-06, "loss": 1.049, "step": 2025 }, { "epoch": 0.14, "learning_rate": 3.4441805225653207e-06, "loss": 2.0626, "step": 2030 }, { "epoch": 0.14, "learning_rate": 3.4526637258228713e-06, "loss": 1.9922, "step": 2035 }, { "epoch": 0.14, "learning_rate": 3.461146929080421e-06, "loss": 3.0748, "step": 2040 }, { "epoch": 0.14, "learning_rate": 3.469630132337971e-06, "loss": 2.5918, "step": 2045 }, { "epoch": 0.14, "learning_rate": 3.478113335595521e-06, "loss": 1.6462, "step": 2050 }, { "epoch": 0.14, "learning_rate": 3.4865965388530716e-06, "loss": 2.4569, "step": 2055 }, { "epoch": 0.14, "learning_rate": 3.4950797421106214e-06, "loss": 2.707, "step": 2060 }, { "epoch": 0.14, "learning_rate": 3.5035629453681713e-06, "loss": 2.1187, "step": 2065 }, { "epoch": 0.14, "learning_rate": 3.512046148625721e-06, "loss": 2.4191, "step": 2070 }, { "epoch": 0.14, "learning_rate": 3.520529351883271e-06, "loss": 2.2025, "step": 2075 }, { "epoch": 0.14, "learning_rate": 3.5290125551408217e-06, "loss": 1.5461, "step": 2080 }, { "epoch": 0.14, "learning_rate": 3.5374957583983715e-06, "loss": 2.4469, "step": 2085 }, { "epoch": 0.14, "learning_rate": 3.5459789616559214e-06, "loss": 3.0078, "step": 2090 }, { "epoch": 0.14, "learning_rate": 3.5544621649134712e-06, "loss": 0.597, "step": 2095 }, { "epoch": 0.14, "learning_rate": 3.562945368171022e-06, "loss": 1.3024, "step": 2100 }, { "epoch": 0.14, "learning_rate": 3.5714285714285718e-06, "loss": 2.9811, "step": 2105 }, { "epoch": 0.14, "learning_rate": 3.5799117746861216e-06, "loss": 1.418, "step": 2110 }, { "epoch": 0.14, "learning_rate": 3.5883949779436715e-06, "loss": 1.7629, "step": 2115 }, { "epoch": 0.14, "learning_rate": 3.596878181201222e-06, "loss": 2.3645, "step": 2120 }, { "epoch": 0.14, "learning_rate": 3.605361384458772e-06, "loss": 1.9825, "step": 2125 }, { "epoch": 0.14, "learning_rate": 3.613844587716322e-06, "loss": 3.0057, "step": 2130 }, { "epoch": 0.14, "learning_rate": 3.6223277909738717e-06, "loss": 2.4173, "step": 2135 }, { "epoch": 0.15, "learning_rate": 3.6308109942314224e-06, "loss": 2.5712, "step": 2140 }, { "epoch": 0.15, "learning_rate": 3.6392941974889723e-06, "loss": 2.2702, "step": 2145 }, { "epoch": 0.15, "learning_rate": 3.647777400746522e-06, "loss": 1.4728, "step": 2150 }, { "epoch": 0.15, "learning_rate": 3.656260604004072e-06, "loss": 1.6808, "step": 2155 }, { "epoch": 0.15, "learning_rate": 3.6647438072616222e-06, "loss": 2.3837, "step": 2160 }, { "epoch": 0.15, "learning_rate": 3.6732270105191725e-06, "loss": 1.6418, "step": 2165 }, { "epoch": 0.15, "learning_rate": 3.6817102137767224e-06, "loss": 1.2494, "step": 2170 }, { "epoch": 0.15, "learning_rate": 3.6901934170342722e-06, "loss": 2.8943, "step": 2175 }, { "epoch": 0.15, "learning_rate": 3.6986766202918225e-06, "loss": 1.6094, "step": 2180 }, { "epoch": 0.15, "learning_rate": 3.7071598235493728e-06, "loss": 1.7042, "step": 2185 }, { "epoch": 0.15, "learning_rate": 3.7156430268069226e-06, "loss": 2.0766, "step": 2190 }, { "epoch": 0.15, "learning_rate": 3.7241262300644725e-06, "loss": 2.6871, "step": 2195 }, { "epoch": 0.15, "learning_rate": 3.7326094333220227e-06, "loss": 2.1834, "step": 2200 }, { "epoch": 0.15, "learning_rate": 3.741092636579573e-06, "loss": 2.1081, "step": 2205 }, { "epoch": 0.15, "learning_rate": 3.749575839837123e-06, "loss": 1.2553, "step": 2210 }, { "epoch": 0.15, "learning_rate": 3.7580590430946727e-06, "loss": 2.7459, "step": 2215 }, { "epoch": 0.15, "learning_rate": 3.7665422463522226e-06, "loss": 1.7495, "step": 2220 }, { "epoch": 0.15, "learning_rate": 3.775025449609773e-06, "loss": 2.8184, "step": 2225 }, { "epoch": 0.15, "learning_rate": 3.783508652867323e-06, "loss": 1.4649, "step": 2230 }, { "epoch": 0.15, "learning_rate": 3.791991856124873e-06, "loss": 2.5297, "step": 2235 }, { "epoch": 0.15, "learning_rate": 3.800475059382423e-06, "loss": 1.656, "step": 2240 }, { "epoch": 0.15, "learning_rate": 3.808958262639973e-06, "loss": 1.263, "step": 2245 }, { "epoch": 0.15, "learning_rate": 3.817441465897524e-06, "loss": 1.5835, "step": 2250 }, { "epoch": 0.15, "learning_rate": 3.825924669155074e-06, "loss": 1.1275, "step": 2255 }, { "epoch": 0.15, "learning_rate": 3.8344078724126235e-06, "loss": 1.7566, "step": 2260 }, { "epoch": 0.15, "learning_rate": 3.842891075670173e-06, "loss": 1.2293, "step": 2265 }, { "epoch": 0.15, "learning_rate": 3.851374278927724e-06, "loss": 3.0621, "step": 2270 }, { "epoch": 0.15, "learning_rate": 3.859857482185274e-06, "loss": 0.9832, "step": 2275 }, { "epoch": 0.15, "learning_rate": 3.868340685442824e-06, "loss": 2.1425, "step": 2280 }, { "epoch": 0.16, "learning_rate": 3.8768238887003736e-06, "loss": 2.3215, "step": 2285 }, { "epoch": 0.16, "learning_rate": 3.885307091957924e-06, "loss": 2.2917, "step": 2290 }, { "epoch": 0.16, "learning_rate": 3.893790295215474e-06, "loss": 2.3424, "step": 2295 }, { "epoch": 0.16, "learning_rate": 3.902273498473024e-06, "loss": 1.1075, "step": 2300 }, { "epoch": 0.16, "learning_rate": 3.910756701730574e-06, "loss": 1.6467, "step": 2305 }, { "epoch": 0.16, "learning_rate": 3.919239904988124e-06, "loss": 0.8982, "step": 2310 }, { "epoch": 0.16, "learning_rate": 3.927723108245674e-06, "loss": 1.8119, "step": 2315 }, { "epoch": 0.16, "learning_rate": 3.936206311503224e-06, "loss": 2.3027, "step": 2320 }, { "epoch": 0.16, "learning_rate": 3.944689514760774e-06, "loss": 2.2491, "step": 2325 }, { "epoch": 0.16, "learning_rate": 3.953172718018324e-06, "loss": 1.0377, "step": 2330 }, { "epoch": 0.16, "learning_rate": 3.961655921275875e-06, "loss": 2.0285, "step": 2335 }, { "epoch": 0.16, "learning_rate": 3.9701391245334245e-06, "loss": 1.5259, "step": 2340 }, { "epoch": 0.16, "learning_rate": 3.978622327790974e-06, "loss": 1.6128, "step": 2345 }, { "epoch": 0.16, "learning_rate": 3.987105531048524e-06, "loss": 2.4332, "step": 2350 }, { "epoch": 0.16, "learning_rate": 3.995588734306075e-06, "loss": 2.3947, "step": 2355 }, { "epoch": 0.16, "learning_rate": 4.004071937563625e-06, "loss": 1.9687, "step": 2360 }, { "epoch": 0.16, "learning_rate": 4.0125551408211745e-06, "loss": 1.0648, "step": 2365 }, { "epoch": 0.16, "learning_rate": 4.021038344078724e-06, "loss": 1.4948, "step": 2370 }, { "epoch": 0.16, "learning_rate": 4.029521547336275e-06, "loss": 1.5695, "step": 2375 }, { "epoch": 0.16, "learning_rate": 4.038004750593825e-06, "loss": 1.3968, "step": 2380 }, { "epoch": 0.16, "learning_rate": 4.046487953851375e-06, "loss": 2.0636, "step": 2385 }, { "epoch": 0.16, "learning_rate": 4.054971157108925e-06, "loss": 1.5876, "step": 2390 }, { "epoch": 0.16, "learning_rate": 4.0634543603664745e-06, "loss": 2.0669, "step": 2395 }, { "epoch": 0.16, "learning_rate": 4.071937563624025e-06, "loss": 1.9425, "step": 2400 }, { "epoch": 0.16, "learning_rate": 4.080420766881575e-06, "loss": 1.4441, "step": 2405 }, { "epoch": 0.16, "learning_rate": 4.088903970139125e-06, "loss": 2.2615, "step": 2410 }, { "epoch": 0.16, "learning_rate": 4.097387173396675e-06, "loss": 1.3319, "step": 2415 }, { "epoch": 0.16, "learning_rate": 4.1058703766542254e-06, "loss": 1.3957, "step": 2420 }, { "epoch": 0.16, "learning_rate": 4.114353579911775e-06, "loss": 2.5342, "step": 2425 }, { "epoch": 0.16, "learning_rate": 4.122836783169325e-06, "loss": 2.7473, "step": 2430 }, { "epoch": 0.17, "learning_rate": 4.131319986426875e-06, "loss": 1.3764, "step": 2435 }, { "epoch": 0.17, "learning_rate": 4.139803189684426e-06, "loss": 2.2689, "step": 2440 }, { "epoch": 0.17, "learning_rate": 4.1482863929419755e-06, "loss": 1.6064, "step": 2445 }, { "epoch": 0.17, "learning_rate": 4.156769596199525e-06, "loss": 2.6199, "step": 2450 }, { "epoch": 0.17, "learning_rate": 4.165252799457075e-06, "loss": 1.0541, "step": 2455 }, { "epoch": 0.17, "learning_rate": 4.173736002714626e-06, "loss": 1.6228, "step": 2460 }, { "epoch": 0.17, "learning_rate": 4.182219205972176e-06, "loss": 1.5295, "step": 2465 }, { "epoch": 0.17, "learning_rate": 4.190702409229726e-06, "loss": 2.0926, "step": 2470 }, { "epoch": 0.17, "learning_rate": 4.1991856124872755e-06, "loss": 1.9096, "step": 2475 }, { "epoch": 0.17, "learning_rate": 4.207668815744825e-06, "loss": 2.0809, "step": 2480 }, { "epoch": 0.17, "learning_rate": 4.216152019002376e-06, "loss": 2.308, "step": 2485 }, { "epoch": 0.17, "learning_rate": 4.224635222259926e-06, "loss": 1.6226, "step": 2490 }, { "epoch": 0.17, "learning_rate": 4.233118425517476e-06, "loss": 2.1655, "step": 2495 }, { "epoch": 0.17, "learning_rate": 4.2416016287750256e-06, "loss": 0.9795, "step": 2500 }, { "epoch": 0.17, "learning_rate": 4.250084832032576e-06, "loss": 2.8162, "step": 2505 }, { "epoch": 0.17, "learning_rate": 4.258568035290126e-06, "loss": 0.8841, "step": 2510 }, { "epoch": 0.17, "learning_rate": 4.267051238547676e-06, "loss": 2.4994, "step": 2515 }, { "epoch": 0.17, "learning_rate": 4.275534441805226e-06, "loss": 2.657, "step": 2520 }, { "epoch": 0.17, "learning_rate": 4.2840176450627765e-06, "loss": 1.0529, "step": 2525 }, { "epoch": 0.17, "learning_rate": 4.292500848320326e-06, "loss": 2.6045, "step": 2530 }, { "epoch": 0.17, "learning_rate": 4.300984051577876e-06, "loss": 0.941, "step": 2535 }, { "epoch": 0.17, "learning_rate": 4.309467254835426e-06, "loss": 1.7609, "step": 2540 }, { "epoch": 0.17, "learning_rate": 4.317950458092977e-06, "loss": 1.6094, "step": 2545 }, { "epoch": 0.17, "learning_rate": 4.326433661350527e-06, "loss": 1.236, "step": 2550 }, { "epoch": 0.17, "learning_rate": 4.3349168646080765e-06, "loss": 1.2183, "step": 2555 }, { "epoch": 0.17, "learning_rate": 4.343400067865626e-06, "loss": 1.0345, "step": 2560 }, { "epoch": 0.17, "learning_rate": 4.351883271123176e-06, "loss": 2.9158, "step": 2565 }, { "epoch": 0.17, "learning_rate": 4.360366474380727e-06, "loss": 1.8309, "step": 2570 }, { "epoch": 0.17, "learning_rate": 4.368849677638277e-06, "loss": 1.9239, "step": 2575 }, { "epoch": 0.18, "learning_rate": 4.3773328808958265e-06, "loss": 1.0226, "step": 2580 }, { "epoch": 0.18, "learning_rate": 4.385816084153376e-06, "loss": 1.489, "step": 2585 }, { "epoch": 0.18, "learning_rate": 4.394299287410927e-06, "loss": 1.2979, "step": 2590 }, { "epoch": 0.18, "learning_rate": 4.402782490668477e-06, "loss": 1.8196, "step": 2595 }, { "epoch": 0.18, "learning_rate": 4.411265693926027e-06, "loss": 2.213, "step": 2600 }, { "epoch": 0.18, "learning_rate": 4.419748897183577e-06, "loss": 1.6345, "step": 2605 }, { "epoch": 0.18, "learning_rate": 4.428232100441127e-06, "loss": 1.5949, "step": 2610 }, { "epoch": 0.18, "learning_rate": 4.436715303698677e-06, "loss": 1.4351, "step": 2615 }, { "epoch": 0.18, "learning_rate": 4.445198506956227e-06, "loss": 1.2113, "step": 2620 }, { "epoch": 0.18, "learning_rate": 4.453681710213777e-06, "loss": 1.9119, "step": 2625 }, { "epoch": 0.18, "learning_rate": 4.462164913471328e-06, "loss": 1.9472, "step": 2630 }, { "epoch": 0.18, "learning_rate": 4.4706481167288774e-06, "loss": 2.0322, "step": 2635 }, { "epoch": 0.18, "learning_rate": 4.479131319986427e-06, "loss": 1.967, "step": 2640 }, { "epoch": 0.18, "learning_rate": 4.487614523243977e-06, "loss": 1.9979, "step": 2645 }, { "epoch": 0.18, "learning_rate": 4.496097726501527e-06, "loss": 0.9028, "step": 2650 }, { "epoch": 0.18, "learning_rate": 4.504580929759078e-06, "loss": 1.8506, "step": 2655 }, { "epoch": 0.18, "learning_rate": 4.5130641330166275e-06, "loss": 2.5645, "step": 2660 }, { "epoch": 0.18, "learning_rate": 4.521547336274177e-06, "loss": 1.3375, "step": 2665 }, { "epoch": 0.18, "learning_rate": 4.530030539531727e-06, "loss": 2.7363, "step": 2670 }, { "epoch": 0.18, "learning_rate": 4.538513742789278e-06, "loss": 1.208, "step": 2675 }, { "epoch": 0.18, "learning_rate": 4.546996946046828e-06, "loss": 2.023, "step": 2680 }, { "epoch": 0.18, "learning_rate": 4.555480149304378e-06, "loss": 1.6252, "step": 2685 }, { "epoch": 0.18, "learning_rate": 4.5639633525619275e-06, "loss": 1.6091, "step": 2690 }, { "epoch": 0.18, "learning_rate": 4.572446555819478e-06, "loss": 1.476, "step": 2695 }, { "epoch": 0.18, "learning_rate": 4.580929759077028e-06, "loss": 1.1194, "step": 2700 }, { "epoch": 0.18, "learning_rate": 4.589412962334578e-06, "loss": 2.1469, "step": 2705 }, { "epoch": 0.18, "learning_rate": 4.597896165592128e-06, "loss": 3.116, "step": 2710 }, { "epoch": 0.18, "learning_rate": 4.606379368849678e-06, "loss": 2.3248, "step": 2715 }, { "epoch": 0.18, "learning_rate": 4.614862572107228e-06, "loss": 1.0995, "step": 2720 }, { "epoch": 0.18, "learning_rate": 4.623345775364778e-06, "loss": 1.5326, "step": 2725 }, { "epoch": 0.19, "learning_rate": 4.631828978622328e-06, "loss": 1.3002, "step": 2730 }, { "epoch": 0.19, "learning_rate": 4.640312181879878e-06, "loss": 1.9356, "step": 2735 }, { "epoch": 0.19, "learning_rate": 4.6487953851374285e-06, "loss": 2.543, "step": 2740 }, { "epoch": 0.19, "learning_rate": 4.657278588394978e-06, "loss": 2.7094, "step": 2745 }, { "epoch": 0.19, "learning_rate": 4.665761791652528e-06, "loss": 2.0845, "step": 2750 }, { "epoch": 0.19, "learning_rate": 4.674244994910078e-06, "loss": 1.9102, "step": 2755 }, { "epoch": 0.19, "learning_rate": 4.682728198167629e-06, "loss": 1.508, "step": 2760 }, { "epoch": 0.19, "learning_rate": 4.691211401425179e-06, "loss": 1.8106, "step": 2765 }, { "epoch": 0.19, "learning_rate": 4.6996946046827285e-06, "loss": 2.2835, "step": 2770 }, { "epoch": 0.19, "learning_rate": 4.708177807940278e-06, "loss": 2.684, "step": 2775 }, { "epoch": 0.19, "learning_rate": 4.716661011197829e-06, "loss": 1.3701, "step": 2780 }, { "epoch": 0.19, "learning_rate": 4.725144214455379e-06, "loss": 1.6018, "step": 2785 }, { "epoch": 0.19, "learning_rate": 4.733627417712929e-06, "loss": 1.0261, "step": 2790 }, { "epoch": 0.19, "learning_rate": 4.7421106209704785e-06, "loss": 2.0012, "step": 2795 }, { "epoch": 0.19, "learning_rate": 4.750593824228028e-06, "loss": 2.5041, "step": 2800 }, { "epoch": 0.19, "learning_rate": 4.759077027485579e-06, "loss": 1.3477, "step": 2805 }, { "epoch": 0.19, "learning_rate": 4.767560230743129e-06, "loss": 1.9226, "step": 2810 }, { "epoch": 0.19, "learning_rate": 4.776043434000679e-06, "loss": 1.5209, "step": 2815 }, { "epoch": 0.19, "learning_rate": 4.784526637258229e-06, "loss": 0.4898, "step": 2820 }, { "epoch": 0.19, "learning_rate": 4.793009840515779e-06, "loss": 1.8473, "step": 2825 }, { "epoch": 0.19, "learning_rate": 4.801493043773329e-06, "loss": 1.8141, "step": 2830 }, { "epoch": 0.19, "learning_rate": 4.809976247030879e-06, "loss": 1.5838, "step": 2835 }, { "epoch": 0.19, "learning_rate": 4.818459450288429e-06, "loss": 1.4553, "step": 2840 }, { "epoch": 0.19, "learning_rate": 4.82694265354598e-06, "loss": 1.6626, "step": 2845 }, { "epoch": 0.19, "learning_rate": 4.8354258568035294e-06, "loss": 0.5478, "step": 2850 }, { "epoch": 0.19, "learning_rate": 4.843909060061079e-06, "loss": 1.8611, "step": 2855 }, { "epoch": 0.19, "learning_rate": 4.852392263318629e-06, "loss": 0.7929, "step": 2860 }, { "epoch": 0.19, "learning_rate": 4.86087546657618e-06, "loss": 0.8796, "step": 2865 }, { "epoch": 0.19, "learning_rate": 4.86935866983373e-06, "loss": 2.4407, "step": 2870 }, { "epoch": 0.2, "learning_rate": 4.8778418730912795e-06, "loss": 1.1799, "step": 2875 }, { "epoch": 0.2, "learning_rate": 4.886325076348829e-06, "loss": 0.9514, "step": 2880 }, { "epoch": 0.2, "learning_rate": 4.894808279606379e-06, "loss": 1.8889, "step": 2885 }, { "epoch": 0.2, "learning_rate": 4.90329148286393e-06, "loss": 1.4458, "step": 2890 }, { "epoch": 0.2, "learning_rate": 4.91177468612148e-06, "loss": 1.6151, "step": 2895 }, { "epoch": 0.2, "learning_rate": 4.92025788937903e-06, "loss": 1.9205, "step": 2900 }, { "epoch": 0.2, "learning_rate": 4.9287410926365795e-06, "loss": 1.1976, "step": 2905 }, { "epoch": 0.2, "learning_rate": 4.93722429589413e-06, "loss": 1.4111, "step": 2910 }, { "epoch": 0.2, "learning_rate": 4.94570749915168e-06, "loss": 1.3603, "step": 2915 }, { "epoch": 0.2, "learning_rate": 4.95419070240923e-06, "loss": 1.0981, "step": 2920 }, { "epoch": 0.2, "learning_rate": 4.96267390566678e-06, "loss": 1.1766, "step": 2925 }, { "epoch": 0.2, "learning_rate": 4.97115710892433e-06, "loss": 1.8744, "step": 2930 }, { "epoch": 0.2, "learning_rate": 4.97964031218188e-06, "loss": 1.6937, "step": 2935 }, { "epoch": 0.2, "learning_rate": 4.98812351543943e-06, "loss": 1.2975, "step": 2940 }, { "epoch": 0.2, "learning_rate": 4.99660671869698e-06, "loss": 1.3911, "step": 2945 }, { "epoch": 0.2, "learning_rate": 5.00508992195453e-06, "loss": 1.6568, "step": 2950 }, { "epoch": 0.2, "learning_rate": 5.01357312521208e-06, "loss": 1.4335, "step": 2955 }, { "epoch": 0.2, "learning_rate": 5.022056328469631e-06, "loss": 1.0243, "step": 2960 }, { "epoch": 0.2, "learning_rate": 5.030539531727181e-06, "loss": 1.4638, "step": 2965 }, { "epoch": 0.2, "learning_rate": 5.039022734984731e-06, "loss": 0.7201, "step": 2970 }, { "epoch": 0.2, "learning_rate": 5.047505938242281e-06, "loss": 1.2977, "step": 2975 }, { "epoch": 0.2, "learning_rate": 5.055989141499831e-06, "loss": 1.5895, "step": 2980 }, { "epoch": 0.2, "learning_rate": 5.0644723447573805e-06, "loss": 1.3562, "step": 2985 }, { "epoch": 0.2, "learning_rate": 5.07295554801493e-06, "loss": 0.961, "step": 2990 }, { "epoch": 0.2, "learning_rate": 5.08143875127248e-06, "loss": 1.2128, "step": 2995 }, { "epoch": 0.2, "learning_rate": 5.089921954530031e-06, "loss": 1.141, "step": 3000 }, { "epoch": 0.2, "learning_rate": 5.0984051577875815e-06, "loss": 0.8891, "step": 3005 }, { "epoch": 0.2, "learning_rate": 5.106888361045131e-06, "loss": 1.096, "step": 3010 }, { "epoch": 0.2, "learning_rate": 5.115371564302681e-06, "loss": 1.4094, "step": 3015 }, { "epoch": 0.2, "learning_rate": 5.123854767560231e-06, "loss": 1.3625, "step": 3020 }, { "epoch": 0.21, "learning_rate": 5.132337970817781e-06, "loss": 0.8649, "step": 3025 }, { "epoch": 0.21, "learning_rate": 5.140821174075331e-06, "loss": 2.3539, "step": 3030 }, { "epoch": 0.21, "learning_rate": 5.149304377332881e-06, "loss": 1.0066, "step": 3035 }, { "epoch": 0.21, "learning_rate": 5.157787580590431e-06, "loss": 1.5793, "step": 3040 }, { "epoch": 0.21, "learning_rate": 5.166270783847982e-06, "loss": 1.6785, "step": 3045 }, { "epoch": 0.21, "learning_rate": 5.174753987105532e-06, "loss": 2.3337, "step": 3050 }, { "epoch": 0.21, "learning_rate": 5.183237190363082e-06, "loss": 2.3434, "step": 3055 }, { "epoch": 0.21, "learning_rate": 5.191720393620632e-06, "loss": 0.9265, "step": 3060 }, { "epoch": 0.21, "learning_rate": 5.2002035968781814e-06, "loss": 1.1198, "step": 3065 }, { "epoch": 0.21, "learning_rate": 5.208686800135731e-06, "loss": 1.9313, "step": 3070 }, { "epoch": 0.21, "learning_rate": 5.217170003393281e-06, "loss": 2.392, "step": 3075 }, { "epoch": 0.21, "learning_rate": 5.225653206650832e-06, "loss": 1.3066, "step": 3080 }, { "epoch": 0.21, "learning_rate": 5.234136409908382e-06, "loss": 1.1845, "step": 3085 }, { "epoch": 0.21, "learning_rate": 5.242619613165932e-06, "loss": 1.6427, "step": 3090 }, { "epoch": 0.21, "learning_rate": 5.251102816423482e-06, "loss": 1.1975, "step": 3095 }, { "epoch": 0.21, "learning_rate": 5.259586019681032e-06, "loss": 1.7835, "step": 3100 }, { "epoch": 0.21, "learning_rate": 5.268069222938582e-06, "loss": 0.8183, "step": 3105 }, { "epoch": 0.21, "learning_rate": 5.276552426196132e-06, "loss": 1.1198, "step": 3110 }, { "epoch": 0.21, "learning_rate": 5.285035629453682e-06, "loss": 1.8344, "step": 3115 }, { "epoch": 0.21, "learning_rate": 5.293518832711232e-06, "loss": 1.1431, "step": 3120 }, { "epoch": 0.21, "learning_rate": 5.302002035968782e-06, "loss": 1.6724, "step": 3125 }, { "epoch": 0.21, "learning_rate": 5.310485239226332e-06, "loss": 0.8941, "step": 3130 }, { "epoch": 0.21, "learning_rate": 5.318968442483883e-06, "loss": 0.9279, "step": 3135 }, { "epoch": 0.21, "learning_rate": 5.3274516457414326e-06, "loss": 1.9874, "step": 3140 }, { "epoch": 0.21, "learning_rate": 5.335934848998982e-06, "loss": 2.1222, "step": 3145 }, { "epoch": 0.21, "learning_rate": 5.344418052256532e-06, "loss": 2.0073, "step": 3150 }, { "epoch": 0.21, "learning_rate": 5.352901255514082e-06, "loss": 1.797, "step": 3155 }, { "epoch": 0.21, "learning_rate": 5.361384458771632e-06, "loss": 1.4789, "step": 3160 }, { "epoch": 0.21, "learning_rate": 5.369867662029183e-06, "loss": 1.5716, "step": 3165 }, { "epoch": 0.22, "learning_rate": 5.3783508652867325e-06, "loss": 1.6938, "step": 3170 }, { "epoch": 0.22, "learning_rate": 5.386834068544283e-06, "loss": 1.0938, "step": 3175 }, { "epoch": 0.22, "learning_rate": 5.395317271801833e-06, "loss": 2.4726, "step": 3180 }, { "epoch": 0.22, "learning_rate": 5.403800475059383e-06, "loss": 0.8226, "step": 3185 }, { "epoch": 0.22, "learning_rate": 5.412283678316933e-06, "loss": 1.9163, "step": 3190 }, { "epoch": 0.22, "learning_rate": 5.420766881574483e-06, "loss": 0.5866, "step": 3195 }, { "epoch": 0.22, "learning_rate": 5.4292500848320325e-06, "loss": 1.1063, "step": 3200 }, { "epoch": 0.22, "learning_rate": 5.437733288089583e-06, "loss": 1.028, "step": 3205 }, { "epoch": 0.22, "learning_rate": 5.446216491347133e-06, "loss": 1.4023, "step": 3210 }, { "epoch": 0.22, "learning_rate": 5.454699694604683e-06, "loss": 1.1842, "step": 3215 }, { "epoch": 0.22, "learning_rate": 5.4631828978622335e-06, "loss": 1.4007, "step": 3220 }, { "epoch": 0.22, "learning_rate": 5.471666101119783e-06, "loss": 1.6613, "step": 3225 }, { "epoch": 0.22, "learning_rate": 5.480149304377333e-06, "loss": 1.8898, "step": 3230 }, { "epoch": 0.22, "learning_rate": 5.488632507634883e-06, "loss": 1.7857, "step": 3235 }, { "epoch": 0.22, "learning_rate": 5.497115710892433e-06, "loss": 0.899, "step": 3240 }, { "epoch": 0.22, "learning_rate": 5.505598914149984e-06, "loss": 1.2759, "step": 3245 }, { "epoch": 0.22, "learning_rate": 5.5140821174075335e-06, "loss": 1.3707, "step": 3250 }, { "epoch": 0.22, "learning_rate": 5.522565320665083e-06, "loss": 1.9325, "step": 3255 }, { "epoch": 0.22, "learning_rate": 5.531048523922634e-06, "loss": 0.8879, "step": 3260 }, { "epoch": 0.22, "learning_rate": 5.539531727180184e-06, "loss": 1.8365, "step": 3265 }, { "epoch": 0.22, "learning_rate": 5.548014930437734e-06, "loss": 1.3064, "step": 3270 }, { "epoch": 0.22, "learning_rate": 5.556498133695284e-06, "loss": 2.3359, "step": 3275 }, { "epoch": 0.22, "learning_rate": 5.5649813369528334e-06, "loss": 0.6104, "step": 3280 }, { "epoch": 0.22, "learning_rate": 5.573464540210384e-06, "loss": 0.7511, "step": 3285 }, { "epoch": 0.22, "learning_rate": 5.581947743467934e-06, "loss": 1.8905, "step": 3290 }, { "epoch": 0.22, "learning_rate": 5.590430946725484e-06, "loss": 1.48, "step": 3295 }, { "epoch": 0.22, "learning_rate": 5.598914149983034e-06, "loss": 0.6241, "step": 3300 }, { "epoch": 0.22, "learning_rate": 5.607397353240584e-06, "loss": 1.6211, "step": 3305 }, { "epoch": 0.22, "learning_rate": 5.615880556498134e-06, "loss": 1.2394, "step": 3310 }, { "epoch": 0.22, "learning_rate": 5.624363759755684e-06, "loss": 1.8439, "step": 3315 }, { "epoch": 0.23, "learning_rate": 5.632846963013234e-06, "loss": 1.4875, "step": 3320 }, { "epoch": 0.23, "learning_rate": 5.641330166270785e-06, "loss": 1.1221, "step": 3325 }, { "epoch": 0.23, "learning_rate": 5.6498133695283345e-06, "loss": 2.24, "step": 3330 }, { "epoch": 0.23, "learning_rate": 5.658296572785884e-06, "loss": 2.1019, "step": 3335 }, { "epoch": 0.23, "learning_rate": 5.666779776043434e-06, "loss": 1.132, "step": 3340 }, { "epoch": 0.23, "learning_rate": 5.675262979300985e-06, "loss": 1.1682, "step": 3345 }, { "epoch": 0.23, "learning_rate": 5.683746182558535e-06, "loss": 0.8908, "step": 3350 }, { "epoch": 0.23, "learning_rate": 5.6922293858160846e-06, "loss": 1.6654, "step": 3355 }, { "epoch": 0.23, "learning_rate": 5.700712589073634e-06, "loss": 0.921, "step": 3360 }, { "epoch": 0.23, "learning_rate": 5.709195792331185e-06, "loss": 1.4066, "step": 3365 }, { "epoch": 0.23, "learning_rate": 5.717678995588735e-06, "loss": 1.1992, "step": 3370 }, { "epoch": 0.23, "learning_rate": 5.726162198846285e-06, "loss": 0.7609, "step": 3375 }, { "epoch": 0.23, "learning_rate": 5.734645402103835e-06, "loss": 1.2701, "step": 3380 }, { "epoch": 0.23, "learning_rate": 5.7431286053613845e-06, "loss": 0.5474, "step": 3385 }, { "epoch": 0.23, "learning_rate": 5.751611808618935e-06, "loss": 1.3052, "step": 3390 }, { "epoch": 0.23, "learning_rate": 5.760095011876485e-06, "loss": 1.3759, "step": 3395 }, { "epoch": 0.23, "learning_rate": 5.768578215134035e-06, "loss": 1.8962, "step": 3400 }, { "epoch": 0.23, "learning_rate": 5.777061418391586e-06, "loss": 1.6858, "step": 3405 }, { "epoch": 0.23, "learning_rate": 5.7855446216491354e-06, "loss": 2.2, "step": 3410 }, { "epoch": 0.23, "learning_rate": 5.794027824906685e-06, "loss": 1.1616, "step": 3415 }, { "epoch": 0.23, "learning_rate": 5.802511028164235e-06, "loss": 1.7921, "step": 3420 }, { "epoch": 0.23, "learning_rate": 5.810994231421785e-06, "loss": 1.458, "step": 3425 }, { "epoch": 0.23, "learning_rate": 5.819477434679336e-06, "loss": 1.078, "step": 3430 }, { "epoch": 0.23, "learning_rate": 5.8279606379368855e-06, "loss": 2.1138, "step": 3435 }, { "epoch": 0.23, "learning_rate": 5.836443841194435e-06, "loss": 1.3623, "step": 3440 }, { "epoch": 0.23, "learning_rate": 5.844927044451985e-06, "loss": 1.4557, "step": 3445 }, { "epoch": 0.23, "learning_rate": 5.853410247709536e-06, "loss": 0.8037, "step": 3450 }, { "epoch": 0.23, "learning_rate": 5.861893450967086e-06, "loss": 1.2813, "step": 3455 }, { "epoch": 0.23, "learning_rate": 5.870376654224636e-06, "loss": 0.6385, "step": 3460 }, { "epoch": 0.24, "learning_rate": 5.8788598574821855e-06, "loss": 1.1968, "step": 3465 }, { "epoch": 0.24, "learning_rate": 5.887343060739735e-06, "loss": 1.2694, "step": 3470 }, { "epoch": 0.24, "learning_rate": 5.895826263997286e-06, "loss": 1.2377, "step": 3475 }, { "epoch": 0.24, "learning_rate": 5.904309467254836e-06, "loss": 1.8212, "step": 3480 }, { "epoch": 0.24, "learning_rate": 5.912792670512386e-06, "loss": 1.1553, "step": 3485 }, { "epoch": 0.24, "learning_rate": 5.9212758737699364e-06, "loss": 1.2088, "step": 3490 }, { "epoch": 0.24, "learning_rate": 5.929759077027486e-06, "loss": 1.9489, "step": 3495 }, { "epoch": 0.24, "learning_rate": 5.938242280285036e-06, "loss": 0.8979, "step": 3500 }, { "epoch": 0.24, "learning_rate": 5.946725483542586e-06, "loss": 1.3113, "step": 3505 }, { "epoch": 0.24, "learning_rate": 5.955208686800136e-06, "loss": 0.9942, "step": 3510 }, { "epoch": 0.24, "learning_rate": 5.9636918900576865e-06, "loss": 1.4722, "step": 3515 }, { "epoch": 0.24, "learning_rate": 5.972175093315236e-06, "loss": 0.8761, "step": 3520 }, { "epoch": 0.24, "learning_rate": 5.980658296572786e-06, "loss": 0.9498, "step": 3525 }, { "epoch": 0.24, "learning_rate": 5.989141499830337e-06, "loss": 0.5611, "step": 3530 }, { "epoch": 0.24, "learning_rate": 5.997624703087887e-06, "loss": 1.5969, "step": 3535 }, { "epoch": 0.24, "learning_rate": 6.006107906345437e-06, "loss": 1.1235, "step": 3540 }, { "epoch": 0.24, "learning_rate": 6.0145911096029865e-06, "loss": 0.9523, "step": 3545 }, { "epoch": 0.24, "learning_rate": 6.023074312860536e-06, "loss": 0.8676, "step": 3550 }, { "epoch": 0.24, "learning_rate": 6.031557516118086e-06, "loss": 1.5819, "step": 3555 }, { "epoch": 0.24, "learning_rate": 6.040040719375637e-06, "loss": 1.5395, "step": 3560 }, { "epoch": 0.24, "learning_rate": 6.048523922633187e-06, "loss": 0.9856, "step": 3565 }, { "epoch": 0.24, "learning_rate": 6.057007125890737e-06, "loss": 1.2469, "step": 3570 }, { "epoch": 0.24, "learning_rate": 6.065490329148287e-06, "loss": 1.523, "step": 3575 }, { "epoch": 0.24, "learning_rate": 6.073973532405837e-06, "loss": 0.8435, "step": 3580 }, { "epoch": 0.24, "learning_rate": 6.082456735663387e-06, "loss": 1.4979, "step": 3585 }, { "epoch": 0.24, "learning_rate": 6.090939938920937e-06, "loss": 0.7208, "step": 3590 }, { "epoch": 0.24, "learning_rate": 6.099423142178487e-06, "loss": 1.3292, "step": 3595 }, { "epoch": 0.24, "learning_rate": 6.107906345436037e-06, "loss": 1.9606, "step": 3600 }, { "epoch": 0.24, "learning_rate": 6.116389548693587e-06, "loss": 1.5435, "step": 3605 }, { "epoch": 0.25, "learning_rate": 6.124872751951138e-06, "loss": 1.1229, "step": 3610 }, { "epoch": 0.25, "learning_rate": 6.133355955208688e-06, "loss": 1.3258, "step": 3615 }, { "epoch": 0.25, "learning_rate": 6.141839158466238e-06, "loss": 1.3257, "step": 3620 }, { "epoch": 0.25, "learning_rate": 6.1503223617237874e-06, "loss": 0.9961, "step": 3625 }, { "epoch": 0.25, "learning_rate": 6.158805564981337e-06, "loss": 1.0207, "step": 3630 }, { "epoch": 0.25, "learning_rate": 6.167288768238887e-06, "loss": 0.9849, "step": 3635 }, { "epoch": 0.25, "learning_rate": 6.175771971496437e-06, "loss": 1.0277, "step": 3640 }, { "epoch": 0.25, "learning_rate": 6.184255174753988e-06, "loss": 1.4158, "step": 3645 }, { "epoch": 0.25, "learning_rate": 6.192738378011538e-06, "loss": 2.0445, "step": 3650 }, { "epoch": 0.25, "learning_rate": 6.201221581269088e-06, "loss": 1.8341, "step": 3655 }, { "epoch": 0.25, "learning_rate": 6.209704784526638e-06, "loss": 1.6492, "step": 3660 }, { "epoch": 0.25, "learning_rate": 6.218187987784188e-06, "loss": 1.0158, "step": 3665 }, { "epoch": 0.25, "learning_rate": 6.226671191041738e-06, "loss": 2.2922, "step": 3670 }, { "epoch": 0.25, "learning_rate": 6.235154394299288e-06, "loss": 1.0189, "step": 3675 }, { "epoch": 0.25, "learning_rate": 6.2436375975568375e-06, "loss": 1.269, "step": 3680 }, { "epoch": 0.25, "learning_rate": 6.252120800814388e-06, "loss": 1.3801, "step": 3685 }, { "epoch": 0.25, "learning_rate": 6.260604004071939e-06, "loss": 1.0805, "step": 3690 }, { "epoch": 0.25, "learning_rate": 6.269087207329489e-06, "loss": 1.4821, "step": 3695 }, { "epoch": 0.25, "learning_rate": 6.277570410587039e-06, "loss": 1.1766, "step": 3700 }, { "epoch": 0.25, "learning_rate": 6.2860536138445884e-06, "loss": 0.6716, "step": 3705 }, { "epoch": 0.25, "learning_rate": 6.294536817102138e-06, "loss": 1.0789, "step": 3710 }, { "epoch": 0.25, "learning_rate": 6.303020020359688e-06, "loss": 1.6728, "step": 3715 }, { "epoch": 0.25, "learning_rate": 6.311503223617238e-06, "loss": 1.4255, "step": 3720 }, { "epoch": 0.25, "learning_rate": 6.319986426874788e-06, "loss": 1.8246, "step": 3725 }, { "epoch": 0.25, "learning_rate": 6.3284696301323385e-06, "loss": 0.9952, "step": 3730 }, { "epoch": 0.25, "learning_rate": 6.336952833389889e-06, "loss": 0.5115, "step": 3735 }, { "epoch": 0.25, "learning_rate": 6.345436036647439e-06, "loss": 1.6422, "step": 3740 }, { "epoch": 0.25, "learning_rate": 6.353919239904989e-06, "loss": 1.0897, "step": 3745 }, { "epoch": 0.25, "learning_rate": 6.362402443162539e-06, "loss": 0.8986, "step": 3750 }, { "epoch": 0.25, "learning_rate": 6.370885646420089e-06, "loss": 1.3525, "step": 3755 }, { "epoch": 0.26, "learning_rate": 6.3793688496776385e-06, "loss": 1.0828, "step": 3760 }, { "epoch": 0.26, "learning_rate": 6.387852052935188e-06, "loss": 1.5402, "step": 3765 }, { "epoch": 0.26, "learning_rate": 6.396335256192739e-06, "loss": 0.7276, "step": 3770 }, { "epoch": 0.26, "learning_rate": 6.40481845945029e-06, "loss": 1.0127, "step": 3775 }, { "epoch": 0.26, "learning_rate": 6.4133016627078396e-06, "loss": 2.0883, "step": 3780 }, { "epoch": 0.26, "learning_rate": 6.421784865965389e-06, "loss": 1.2361, "step": 3785 }, { "epoch": 0.26, "learning_rate": 6.430268069222939e-06, "loss": 1.1879, "step": 3790 }, { "epoch": 0.26, "learning_rate": 6.438751272480489e-06, "loss": 1.381, "step": 3795 }, { "epoch": 0.26, "learning_rate": 6.447234475738039e-06, "loss": 1.7445, "step": 3800 }, { "epoch": 0.26, "learning_rate": 6.455717678995589e-06, "loss": 0.4484, "step": 3805 }, { "epoch": 0.26, "learning_rate": 6.464200882253139e-06, "loss": 1.5171, "step": 3810 }, { "epoch": 0.26, "learning_rate": 6.47268408551069e-06, "loss": 1.5603, "step": 3815 }, { "epoch": 0.26, "learning_rate": 6.48116728876824e-06, "loss": 1.6784, "step": 3820 }, { "epoch": 0.26, "learning_rate": 6.48965049202579e-06, "loss": 0.9899, "step": 3825 }, { "epoch": 0.26, "learning_rate": 6.49813369528334e-06, "loss": 1.2871, "step": 3830 }, { "epoch": 0.26, "learning_rate": 6.50661689854089e-06, "loss": 1.4208, "step": 3835 }, { "epoch": 0.26, "learning_rate": 6.5151001017984395e-06, "loss": 1.6445, "step": 3840 }, { "epoch": 0.26, "learning_rate": 6.523583305055989e-06, "loss": 0.6135, "step": 3845 }, { "epoch": 0.26, "learning_rate": 6.532066508313539e-06, "loss": 0.9972, "step": 3850 }, { "epoch": 0.26, "learning_rate": 6.540549711571091e-06, "loss": 1.3882, "step": 3855 }, { "epoch": 0.26, "learning_rate": 6.5490329148286405e-06, "loss": 1.0141, "step": 3860 }, { "epoch": 0.26, "learning_rate": 6.55751611808619e-06, "loss": 0.6109, "step": 3865 }, { "epoch": 0.26, "learning_rate": 6.56599932134374e-06, "loss": 0.9762, "step": 3870 }, { "epoch": 0.26, "learning_rate": 6.57448252460129e-06, "loss": 1.1288, "step": 3875 }, { "epoch": 0.26, "learning_rate": 6.58296572785884e-06, "loss": 1.3785, "step": 3880 }, { "epoch": 0.26, "learning_rate": 6.59144893111639e-06, "loss": 1.5421, "step": 3885 }, { "epoch": 0.26, "learning_rate": 6.59993213437394e-06, "loss": 1.3764, "step": 3890 }, { "epoch": 0.26, "learning_rate": 6.6084153376314895e-06, "loss": 1.29, "step": 3895 }, { "epoch": 0.26, "learning_rate": 6.616898540889041e-06, "loss": 1.3895, "step": 3900 }, { "epoch": 0.27, "learning_rate": 6.625381744146591e-06, "loss": 1.1098, "step": 3905 }, { "epoch": 0.27, "learning_rate": 6.633864947404141e-06, "loss": 1.0641, "step": 3910 }, { "epoch": 0.27, "learning_rate": 6.642348150661691e-06, "loss": 0.912, "step": 3915 }, { "epoch": 0.27, "learning_rate": 6.6508313539192404e-06, "loss": 1.8388, "step": 3920 }, { "epoch": 0.27, "learning_rate": 6.65931455717679e-06, "loss": 0.9016, "step": 3925 }, { "epoch": 0.27, "learning_rate": 6.66779776043434e-06, "loss": 1.1841, "step": 3930 }, { "epoch": 0.27, "learning_rate": 6.67628096369189e-06, "loss": 0.7185, "step": 3935 }, { "epoch": 0.27, "learning_rate": 6.6847641669494415e-06, "loss": 1.5295, "step": 3940 }, { "epoch": 0.27, "learning_rate": 6.693247370206991e-06, "loss": 1.324, "step": 3945 }, { "epoch": 0.27, "learning_rate": 6.701730573464541e-06, "loss": 0.5553, "step": 3950 }, { "epoch": 0.27, "learning_rate": 6.710213776722091e-06, "loss": 0.7274, "step": 3955 }, { "epoch": 0.27, "learning_rate": 6.718696979979641e-06, "loss": 0.7096, "step": 3960 }, { "epoch": 0.27, "learning_rate": 6.727180183237191e-06, "loss": 1.2942, "step": 3965 }, { "epoch": 0.27, "learning_rate": 6.735663386494741e-06, "loss": 1.3646, "step": 3970 }, { "epoch": 0.27, "learning_rate": 6.7441465897522905e-06, "loss": 1.0235, "step": 3975 }, { "epoch": 0.27, "learning_rate": 6.75262979300984e-06, "loss": 1.7108, "step": 3980 }, { "epoch": 0.27, "learning_rate": 6.761112996267392e-06, "loss": 0.9902, "step": 3985 }, { "epoch": 0.27, "learning_rate": 6.769596199524942e-06, "loss": 0.5722, "step": 3990 }, { "epoch": 0.27, "learning_rate": 6.7780794027824916e-06, "loss": 1.2101, "step": 3995 }, { "epoch": 0.27, "learning_rate": 6.786562606040041e-06, "loss": 2.0852, "step": 4000 }, { "epoch": 0.27, "learning_rate": 6.795045809297591e-06, "loss": 0.5256, "step": 4005 }, { "epoch": 0.27, "learning_rate": 6.803529012555141e-06, "loss": 0.8904, "step": 4010 }, { "epoch": 0.27, "learning_rate": 6.812012215812691e-06, "loss": 1.8157, "step": 4015 }, { "epoch": 0.27, "learning_rate": 6.820495419070241e-06, "loss": 1.1117, "step": 4020 }, { "epoch": 0.27, "learning_rate": 6.828978622327792e-06, "loss": 1.2766, "step": 4025 }, { "epoch": 0.27, "learning_rate": 6.837461825585342e-06, "loss": 1.433, "step": 4030 }, { "epoch": 0.27, "learning_rate": 6.845945028842892e-06, "loss": 1.1645, "step": 4035 }, { "epoch": 0.27, "learning_rate": 6.854428232100442e-06, "loss": 1.4974, "step": 4040 }, { "epoch": 0.27, "learning_rate": 6.862911435357992e-06, "loss": 0.9521, "step": 4045 }, { "epoch": 0.27, "learning_rate": 6.871394638615542e-06, "loss": 1.6964, "step": 4050 }, { "epoch": 0.28, "learning_rate": 6.8798778418730915e-06, "loss": 1.0765, "step": 4055 }, { "epoch": 0.28, "learning_rate": 6.888361045130641e-06, "loss": 1.1792, "step": 4060 }, { "epoch": 0.28, "learning_rate": 6.896844248388191e-06, "loss": 1.1674, "step": 4065 }, { "epoch": 0.28, "learning_rate": 6.905327451645743e-06, "loss": 2.0898, "step": 4070 }, { "epoch": 0.28, "learning_rate": 6.9138106549032925e-06, "loss": 1.2683, "step": 4075 }, { "epoch": 0.28, "learning_rate": 6.922293858160842e-06, "loss": 1.4892, "step": 4080 }, { "epoch": 0.28, "learning_rate": 6.930777061418392e-06, "loss": 1.3453, "step": 4085 }, { "epoch": 0.28, "learning_rate": 6.939260264675942e-06, "loss": 1.6244, "step": 4090 }, { "epoch": 0.28, "learning_rate": 6.947743467933492e-06, "loss": 1.1722, "step": 4095 }, { "epoch": 0.28, "learning_rate": 6.956226671191042e-06, "loss": 0.6802, "step": 4100 }, { "epoch": 0.28, "learning_rate": 6.964709874448592e-06, "loss": 1.3419, "step": 4105 }, { "epoch": 0.28, "learning_rate": 6.973193077706143e-06, "loss": 1.3268, "step": 4110 }, { "epoch": 0.28, "learning_rate": 6.981676280963693e-06, "loss": 0.857, "step": 4115 }, { "epoch": 0.28, "learning_rate": 6.990159484221243e-06, "loss": 0.4912, "step": 4120 }, { "epoch": 0.28, "learning_rate": 6.998642687478793e-06, "loss": 1.5939, "step": 4125 }, { "epoch": 0.28, "learning_rate": 7.007125890736343e-06, "loss": 2.1314, "step": 4130 }, { "epoch": 0.28, "learning_rate": 7.0156090939938924e-06, "loss": 0.9476, "step": 4135 }, { "epoch": 0.28, "learning_rate": 7.024092297251442e-06, "loss": 1.1579, "step": 4140 }, { "epoch": 0.28, "learning_rate": 7.032575500508992e-06, "loss": 1.2311, "step": 4145 }, { "epoch": 0.28, "learning_rate": 7.041058703766542e-06, "loss": 0.5252, "step": 4150 }, { "epoch": 0.28, "learning_rate": 7.0495419070240935e-06, "loss": 1.8583, "step": 4155 }, { "epoch": 0.28, "learning_rate": 7.058025110281643e-06, "loss": 0.659, "step": 4160 }, { "epoch": 0.28, "learning_rate": 7.066508313539193e-06, "loss": 0.5518, "step": 4165 }, { "epoch": 0.28, "learning_rate": 7.074991516796743e-06, "loss": 1.4587, "step": 4170 }, { "epoch": 0.28, "learning_rate": 7.083474720054293e-06, "loss": 1.5751, "step": 4175 }, { "epoch": 0.28, "learning_rate": 7.091957923311843e-06, "loss": 1.1595, "step": 4180 }, { "epoch": 0.28, "learning_rate": 7.100441126569393e-06, "loss": 1.0283, "step": 4185 }, { "epoch": 0.28, "learning_rate": 7.1089243298269425e-06, "loss": 1.2127, "step": 4190 }, { "epoch": 0.28, "learning_rate": 7.117407533084494e-06, "loss": 1.2047, "step": 4195 }, { "epoch": 0.29, "learning_rate": 7.125890736342044e-06, "loss": 1.7436, "step": 4200 }, { "epoch": 0.29, "learning_rate": 7.134373939599594e-06, "loss": 1.7921, "step": 4205 }, { "epoch": 0.29, "learning_rate": 7.1428571428571436e-06, "loss": 1.725, "step": 4210 }, { "epoch": 0.29, "learning_rate": 7.151340346114693e-06, "loss": 1.4423, "step": 4215 }, { "epoch": 0.29, "learning_rate": 7.159823549372243e-06, "loss": 1.4232, "step": 4220 }, { "epoch": 0.29, "learning_rate": 7.168306752629793e-06, "loss": 1.506, "step": 4225 }, { "epoch": 0.29, "learning_rate": 7.176789955887343e-06, "loss": 1.0637, "step": 4230 }, { "epoch": 0.29, "learning_rate": 7.185273159144893e-06, "loss": 1.5499, "step": 4235 }, { "epoch": 0.29, "learning_rate": 7.193756362402444e-06, "loss": 0.8301, "step": 4240 }, { "epoch": 0.29, "learning_rate": 7.202239565659994e-06, "loss": 0.8183, "step": 4245 }, { "epoch": 0.29, "learning_rate": 7.210722768917544e-06, "loss": 1.3893, "step": 4250 }, { "epoch": 0.29, "learning_rate": 7.219205972175094e-06, "loss": 0.9792, "step": 4255 }, { "epoch": 0.29, "learning_rate": 7.227689175432644e-06, "loss": 1.1432, "step": 4260 }, { "epoch": 0.29, "learning_rate": 7.236172378690194e-06, "loss": 2.3494, "step": 4265 }, { "epoch": 0.29, "learning_rate": 7.2446555819477435e-06, "loss": 0.9083, "step": 4270 }, { "epoch": 0.29, "learning_rate": 7.253138785205293e-06, "loss": 1.5375, "step": 4275 }, { "epoch": 0.29, "learning_rate": 7.261621988462845e-06, "loss": 1.3849, "step": 4280 }, { "epoch": 0.29, "learning_rate": 7.270105191720395e-06, "loss": 1.3707, "step": 4285 }, { "epoch": 0.29, "learning_rate": 7.2785883949779445e-06, "loss": 1.9158, "step": 4290 }, { "epoch": 0.29, "learning_rate": 7.287071598235494e-06, "loss": 1.3006, "step": 4295 }, { "epoch": 0.29, "learning_rate": 7.295554801493044e-06, "loss": 1.727, "step": 4300 }, { "epoch": 0.29, "learning_rate": 7.304038004750594e-06, "loss": 0.8958, "step": 4305 }, { "epoch": 0.29, "learning_rate": 7.312521208008144e-06, "loss": 1.2932, "step": 4310 }, { "epoch": 0.29, "learning_rate": 7.321004411265694e-06, "loss": 1.2966, "step": 4315 }, { "epoch": 0.29, "learning_rate": 7.3294876145232445e-06, "loss": 0.744, "step": 4320 }, { "epoch": 0.29, "learning_rate": 7.337970817780795e-06, "loss": 0.8491, "step": 4325 }, { "epoch": 0.29, "learning_rate": 7.346454021038345e-06, "loss": 0.9314, "step": 4330 }, { "epoch": 0.29, "learning_rate": 7.354937224295895e-06, "loss": 1.4566, "step": 4335 }, { "epoch": 0.29, "learning_rate": 7.363420427553445e-06, "loss": 0.8223, "step": 4340 }, { "epoch": 0.29, "learning_rate": 7.371903630810995e-06, "loss": 1.4462, "step": 4345 }, { "epoch": 0.3, "learning_rate": 7.3803868340685444e-06, "loss": 1.0659, "step": 4350 }, { "epoch": 0.3, "learning_rate": 7.388870037326094e-06, "loss": 0.7032, "step": 4355 }, { "epoch": 0.3, "learning_rate": 7.397353240583645e-06, "loss": 0.401, "step": 4360 }, { "epoch": 0.3, "learning_rate": 7.405836443841195e-06, "loss": 1.1384, "step": 4365 }, { "epoch": 0.3, "learning_rate": 7.4143196470987455e-06, "loss": 1.041, "step": 4370 }, { "epoch": 0.3, "learning_rate": 7.422802850356295e-06, "loss": 1.0491, "step": 4375 }, { "epoch": 0.3, "learning_rate": 7.431286053613845e-06, "loss": 0.8703, "step": 4380 }, { "epoch": 0.3, "learning_rate": 7.439769256871395e-06, "loss": 0.9189, "step": 4385 }, { "epoch": 0.3, "learning_rate": 7.448252460128945e-06, "loss": 1.0514, "step": 4390 }, { "epoch": 0.3, "learning_rate": 7.456735663386495e-06, "loss": 1.3295, "step": 4395 }, { "epoch": 0.3, "learning_rate": 7.4652188666440455e-06, "loss": 1.0568, "step": 4400 }, { "epoch": 0.3, "learning_rate": 7.473702069901595e-06, "loss": 0.9981, "step": 4405 }, { "epoch": 0.3, "learning_rate": 7.482185273159146e-06, "loss": 1.1471, "step": 4410 }, { "epoch": 0.3, "learning_rate": 7.490668476416696e-06, "loss": 0.7857, "step": 4415 }, { "epoch": 0.3, "learning_rate": 7.499151679674246e-06, "loss": 1.4914, "step": 4420 }, { "epoch": 0.3, "learning_rate": 7.5076348829317956e-06, "loss": 1.0736, "step": 4425 }, { "epoch": 0.3, "learning_rate": 7.516118086189345e-06, "loss": 1.3457, "step": 4430 }, { "epoch": 0.3, "learning_rate": 7.524601289446895e-06, "loss": 1.2175, "step": 4435 }, { "epoch": 0.3, "learning_rate": 7.533084492704445e-06, "loss": 1.0467, "step": 4440 }, { "epoch": 0.3, "learning_rate": 7.541567695961996e-06, "loss": 1.4331, "step": 4445 }, { "epoch": 0.3, "learning_rate": 7.550050899219546e-06, "loss": 1.0356, "step": 4450 }, { "epoch": 0.3, "learning_rate": 7.558534102477096e-06, "loss": 1.0948, "step": 4455 }, { "epoch": 0.3, "learning_rate": 7.567017305734646e-06, "loss": 1.251, "step": 4460 }, { "epoch": 0.3, "learning_rate": 7.575500508992196e-06, "loss": 0.9264, "step": 4465 }, { "epoch": 0.3, "learning_rate": 7.583983712249746e-06, "loss": 1.2275, "step": 4470 }, { "epoch": 0.3, "learning_rate": 7.592466915507296e-06, "loss": 1.0832, "step": 4475 }, { "epoch": 0.3, "learning_rate": 7.600950118764846e-06, "loss": 0.649, "step": 4480 }, { "epoch": 0.3, "learning_rate": 7.609433322022396e-06, "loss": 1.2939, "step": 4485 }, { "epoch": 0.3, "learning_rate": 7.617916525279946e-06, "loss": 1.7802, "step": 4490 }, { "epoch": 0.31, "learning_rate": 7.626399728537497e-06, "loss": 1.931, "step": 4495 }, { "epoch": 0.31, "learning_rate": 7.634882931795048e-06, "loss": 0.8046, "step": 4500 }, { "epoch": 0.31, "learning_rate": 7.643366135052597e-06, "loss": 0.3878, "step": 4505 }, { "epoch": 0.31, "learning_rate": 7.651849338310147e-06, "loss": 0.6635, "step": 4510 }, { "epoch": 0.31, "learning_rate": 7.660332541567697e-06, "loss": 0.7715, "step": 4515 }, { "epoch": 0.31, "learning_rate": 7.668815744825247e-06, "loss": 0.4234, "step": 4520 }, { "epoch": 0.31, "learning_rate": 7.677298948082797e-06, "loss": 0.9609, "step": 4525 }, { "epoch": 0.31, "learning_rate": 7.685782151340347e-06, "loss": 1.9072, "step": 4530 }, { "epoch": 0.31, "learning_rate": 7.694265354597896e-06, "loss": 0.7468, "step": 4535 }, { "epoch": 0.31, "learning_rate": 7.702748557855448e-06, "loss": 0.8884, "step": 4540 }, { "epoch": 0.31, "learning_rate": 7.711231761112998e-06, "loss": 1.0844, "step": 4545 }, { "epoch": 0.31, "learning_rate": 7.719714964370548e-06, "loss": 1.6957, "step": 4550 }, { "epoch": 0.31, "learning_rate": 7.728198167628098e-06, "loss": 0.4958, "step": 4555 }, { "epoch": 0.31, "learning_rate": 7.736681370885647e-06, "loss": 1.0773, "step": 4560 }, { "epoch": 0.31, "learning_rate": 7.745164574143197e-06, "loss": 1.5331, "step": 4565 }, { "epoch": 0.31, "learning_rate": 7.753647777400747e-06, "loss": 1.556, "step": 4570 }, { "epoch": 0.31, "learning_rate": 7.762130980658297e-06, "loss": 0.9948, "step": 4575 }, { "epoch": 0.31, "learning_rate": 7.770614183915849e-06, "loss": 1.1346, "step": 4580 }, { "epoch": 0.31, "learning_rate": 7.779097387173398e-06, "loss": 0.7772, "step": 4585 }, { "epoch": 0.31, "learning_rate": 7.787580590430948e-06, "loss": 0.985, "step": 4590 }, { "epoch": 0.31, "learning_rate": 7.796063793688498e-06, "loss": 0.7617, "step": 4595 }, { "epoch": 0.31, "learning_rate": 7.804546996946048e-06, "loss": 0.6683, "step": 4600 }, { "epoch": 0.31, "learning_rate": 7.813030200203598e-06, "loss": 0.913, "step": 4605 }, { "epoch": 0.31, "learning_rate": 7.821513403461148e-06, "loss": 0.9196, "step": 4610 }, { "epoch": 0.31, "learning_rate": 7.829996606718697e-06, "loss": 1.6425, "step": 4615 }, { "epoch": 0.31, "learning_rate": 7.838479809976247e-06, "loss": 1.0549, "step": 4620 }, { "epoch": 0.31, "learning_rate": 7.846963013233799e-06, "loss": 0.3757, "step": 4625 }, { "epoch": 0.31, "learning_rate": 7.855446216491349e-06, "loss": 0.5971, "step": 4630 }, { "epoch": 0.31, "learning_rate": 7.863929419748899e-06, "loss": 1.4618, "step": 4635 }, { "epoch": 0.31, "learning_rate": 7.872412623006448e-06, "loss": 1.1199, "step": 4640 }, { "epoch": 0.32, "learning_rate": 7.880895826263998e-06, "loss": 1.0543, "step": 4645 }, { "epoch": 0.32, "learning_rate": 7.889379029521548e-06, "loss": 1.3487, "step": 4650 }, { "epoch": 0.32, "learning_rate": 7.897862232779098e-06, "loss": 1.4562, "step": 4655 }, { "epoch": 0.32, "learning_rate": 7.906345436036648e-06, "loss": 0.9313, "step": 4660 }, { "epoch": 0.32, "learning_rate": 7.9148286392942e-06, "loss": 1.29, "step": 4665 }, { "epoch": 0.32, "learning_rate": 7.92331184255175e-06, "loss": 1.0396, "step": 4670 }, { "epoch": 0.32, "learning_rate": 7.931795045809299e-06, "loss": 1.8735, "step": 4675 }, { "epoch": 0.32, "learning_rate": 7.940278249066849e-06, "loss": 1.6321, "step": 4680 }, { "epoch": 0.32, "learning_rate": 7.948761452324399e-06, "loss": 0.9099, "step": 4685 }, { "epoch": 0.32, "learning_rate": 7.957244655581949e-06, "loss": 0.7121, "step": 4690 }, { "epoch": 0.32, "learning_rate": 7.965727858839498e-06, "loss": 1.2892, "step": 4695 }, { "epoch": 0.32, "learning_rate": 7.974211062097048e-06, "loss": 1.1701, "step": 4700 }, { "epoch": 0.32, "learning_rate": 7.982694265354598e-06, "loss": 0.9023, "step": 4705 }, { "epoch": 0.32, "learning_rate": 7.99117746861215e-06, "loss": 0.5501, "step": 4710 }, { "epoch": 0.32, "learning_rate": 7.9996606718697e-06, "loss": 0.4681, "step": 4715 }, { "epoch": 0.32, "learning_rate": 8.00814387512725e-06, "loss": 1.4513, "step": 4720 }, { "epoch": 0.32, "learning_rate": 8.0166270783848e-06, "loss": 1.245, "step": 4725 }, { "epoch": 0.32, "learning_rate": 8.025110281642349e-06, "loss": 0.8893, "step": 4730 }, { "epoch": 0.32, "learning_rate": 8.033593484899899e-06, "loss": 1.1177, "step": 4735 }, { "epoch": 0.32, "learning_rate": 8.042076688157449e-06, "loss": 1.4634, "step": 4740 }, { "epoch": 0.32, "learning_rate": 8.050559891414999e-06, "loss": 1.3794, "step": 4745 }, { "epoch": 0.32, "learning_rate": 8.05904309467255e-06, "loss": 1.9229, "step": 4750 }, { "epoch": 0.32, "learning_rate": 8.0675262979301e-06, "loss": 0.6185, "step": 4755 }, { "epoch": 0.32, "learning_rate": 8.07600950118765e-06, "loss": 0.5039, "step": 4760 }, { "epoch": 0.32, "learning_rate": 8.0844927044452e-06, "loss": 1.4695, "step": 4765 }, { "epoch": 0.32, "learning_rate": 8.09297590770275e-06, "loss": 0.2721, "step": 4770 }, { "epoch": 0.32, "learning_rate": 8.1014591109603e-06, "loss": 0.952, "step": 4775 }, { "epoch": 0.32, "learning_rate": 8.10994231421785e-06, "loss": 0.755, "step": 4780 }, { "epoch": 0.32, "learning_rate": 8.118425517475399e-06, "loss": 0.8615, "step": 4785 }, { "epoch": 0.33, "learning_rate": 8.126908720732949e-06, "loss": 0.6982, "step": 4790 }, { "epoch": 0.33, "learning_rate": 8.1353919239905e-06, "loss": 1.0192, "step": 4795 }, { "epoch": 0.33, "learning_rate": 8.14387512724805e-06, "loss": 0.8378, "step": 4800 }, { "epoch": 0.33, "learning_rate": 8.1523583305056e-06, "loss": 0.895, "step": 4805 }, { "epoch": 0.33, "learning_rate": 8.16084153376315e-06, "loss": 1.7439, "step": 4810 }, { "epoch": 0.33, "learning_rate": 8.1693247370207e-06, "loss": 1.2491, "step": 4815 }, { "epoch": 0.33, "learning_rate": 8.17780794027825e-06, "loss": 1.0434, "step": 4820 }, { "epoch": 0.33, "learning_rate": 8.1862911435358e-06, "loss": 0.8141, "step": 4825 }, { "epoch": 0.33, "learning_rate": 8.19477434679335e-06, "loss": 1.2393, "step": 4830 }, { "epoch": 0.33, "learning_rate": 8.203257550050901e-06, "loss": 1.4531, "step": 4835 }, { "epoch": 0.33, "learning_rate": 8.211740753308451e-06, "loss": 1.8671, "step": 4840 }, { "epoch": 0.33, "learning_rate": 8.220223956566e-06, "loss": 0.9743, "step": 4845 }, { "epoch": 0.33, "learning_rate": 8.22870715982355e-06, "loss": 2.8906, "step": 4850 }, { "epoch": 0.33, "learning_rate": 8.2371903630811e-06, "loss": 0.2948, "step": 4855 }, { "epoch": 0.33, "learning_rate": 8.24567356633865e-06, "loss": 0.8045, "step": 4860 }, { "epoch": 0.33, "learning_rate": 8.2541567695962e-06, "loss": 1.0559, "step": 4865 }, { "epoch": 0.33, "learning_rate": 8.26263997285375e-06, "loss": 0.8666, "step": 4870 }, { "epoch": 0.33, "learning_rate": 8.2711231761113e-06, "loss": 0.8417, "step": 4875 }, { "epoch": 0.33, "learning_rate": 8.279606379368851e-06, "loss": 1.0514, "step": 4880 }, { "epoch": 0.33, "learning_rate": 8.288089582626401e-06, "loss": 0.6586, "step": 4885 }, { "epoch": 0.33, "learning_rate": 8.296572785883951e-06, "loss": 1.5072, "step": 4890 }, { "epoch": 0.33, "learning_rate": 8.305055989141501e-06, "loss": 0.7325, "step": 4895 }, { "epoch": 0.33, "learning_rate": 8.31353919239905e-06, "loss": 0.4437, "step": 4900 }, { "epoch": 0.33, "learning_rate": 8.3220223956566e-06, "loss": 0.6306, "step": 4905 }, { "epoch": 0.33, "learning_rate": 8.33050559891415e-06, "loss": 1.9037, "step": 4910 }, { "epoch": 0.33, "learning_rate": 8.3389888021717e-06, "loss": 1.4545, "step": 4915 }, { "epoch": 0.33, "learning_rate": 8.347472005429252e-06, "loss": 0.6715, "step": 4920 }, { "epoch": 0.33, "learning_rate": 8.355955208686802e-06, "loss": 1.762, "step": 4925 }, { "epoch": 0.33, "learning_rate": 8.364438411944352e-06, "loss": 0.7989, "step": 4930 }, { "epoch": 0.33, "learning_rate": 8.372921615201901e-06, "loss": 0.7071, "step": 4935 }, { "epoch": 0.34, "learning_rate": 8.381404818459451e-06, "loss": 1.0598, "step": 4940 }, { "epoch": 0.34, "learning_rate": 8.389888021717001e-06, "loss": 1.3475, "step": 4945 }, { "epoch": 0.34, "learning_rate": 8.398371224974551e-06, "loss": 0.6949, "step": 4950 }, { "epoch": 0.34, "learning_rate": 8.4068544282321e-06, "loss": 0.9866, "step": 4955 }, { "epoch": 0.34, "learning_rate": 8.41533763148965e-06, "loss": 0.4891, "step": 4960 }, { "epoch": 0.34, "learning_rate": 8.423820834747202e-06, "loss": 1.2161, "step": 4965 }, { "epoch": 0.34, "learning_rate": 8.432304038004752e-06, "loss": 0.8265, "step": 4970 }, { "epoch": 0.34, "learning_rate": 8.440787241262302e-06, "loss": 1.0787, "step": 4975 }, { "epoch": 0.34, "learning_rate": 8.449270444519852e-06, "loss": 1.304, "step": 4980 }, { "epoch": 0.34, "learning_rate": 8.457753647777402e-06, "loss": 1.1246, "step": 4985 }, { "epoch": 0.34, "learning_rate": 8.466236851034951e-06, "loss": 0.7383, "step": 4990 }, { "epoch": 0.34, "learning_rate": 8.474720054292501e-06, "loss": 1.1045, "step": 4995 }, { "epoch": 0.34, "learning_rate": 8.483203257550051e-06, "loss": 0.4583, "step": 5000 }, { "epoch": 0.34, "learning_rate": 8.491686460807603e-06, "loss": 0.5527, "step": 5005 }, { "epoch": 0.34, "learning_rate": 8.500169664065153e-06, "loss": 0.7046, "step": 5010 }, { "epoch": 0.34, "learning_rate": 8.508652867322702e-06, "loss": 0.562, "step": 5015 }, { "epoch": 0.34, "learning_rate": 8.517136070580252e-06, "loss": 0.7779, "step": 5020 }, { "epoch": 0.34, "learning_rate": 8.525619273837802e-06, "loss": 1.5646, "step": 5025 }, { "epoch": 0.34, "learning_rate": 8.534102477095352e-06, "loss": 0.9264, "step": 5030 }, { "epoch": 0.34, "learning_rate": 8.542585680352902e-06, "loss": 0.7966, "step": 5035 }, { "epoch": 0.34, "learning_rate": 8.551068883610452e-06, "loss": 0.4214, "step": 5040 }, { "epoch": 0.34, "learning_rate": 8.559552086868001e-06, "loss": 1.1515, "step": 5045 }, { "epoch": 0.34, "learning_rate": 8.568035290125553e-06, "loss": 0.6125, "step": 5050 }, { "epoch": 0.34, "learning_rate": 8.576518493383103e-06, "loss": 0.472, "step": 5055 }, { "epoch": 0.34, "learning_rate": 8.585001696640653e-06, "loss": 1.3757, "step": 5060 }, { "epoch": 0.34, "learning_rate": 8.593484899898203e-06, "loss": 0.8116, "step": 5065 }, { "epoch": 0.34, "learning_rate": 8.601968103155752e-06, "loss": 0.8079, "step": 5070 }, { "epoch": 0.34, "learning_rate": 8.610451306413302e-06, "loss": 1.2483, "step": 5075 }, { "epoch": 0.34, "learning_rate": 8.618934509670852e-06, "loss": 0.719, "step": 5080 }, { "epoch": 0.35, "learning_rate": 8.627417712928402e-06, "loss": 1.0955, "step": 5085 }, { "epoch": 0.35, "learning_rate": 8.635900916185954e-06, "loss": 1.4092, "step": 5090 }, { "epoch": 0.35, "learning_rate": 8.644384119443503e-06, "loss": 1.2491, "step": 5095 }, { "epoch": 0.35, "learning_rate": 8.652867322701053e-06, "loss": 0.925, "step": 5100 }, { "epoch": 0.35, "learning_rate": 8.661350525958603e-06, "loss": 1.6221, "step": 5105 }, { "epoch": 0.35, "learning_rate": 8.669833729216153e-06, "loss": 0.9173, "step": 5110 }, { "epoch": 0.35, "learning_rate": 8.678316932473703e-06, "loss": 1.3397, "step": 5115 }, { "epoch": 0.35, "learning_rate": 8.686800135731253e-06, "loss": 0.8918, "step": 5120 }, { "epoch": 0.35, "learning_rate": 8.695283338988802e-06, "loss": 1.3732, "step": 5125 }, { "epoch": 0.35, "learning_rate": 8.703766542246352e-06, "loss": 0.7354, "step": 5130 }, { "epoch": 0.35, "learning_rate": 8.712249745503904e-06, "loss": 0.5546, "step": 5135 }, { "epoch": 0.35, "learning_rate": 8.720732948761454e-06, "loss": 0.7543, "step": 5140 }, { "epoch": 0.35, "learning_rate": 8.729216152019004e-06, "loss": 0.596, "step": 5145 }, { "epoch": 0.35, "learning_rate": 8.737699355276553e-06, "loss": 0.9578, "step": 5150 }, { "epoch": 0.35, "learning_rate": 8.746182558534103e-06, "loss": 0.629, "step": 5155 }, { "epoch": 0.35, "learning_rate": 8.754665761791653e-06, "loss": 0.4197, "step": 5160 }, { "epoch": 0.35, "learning_rate": 8.763148965049203e-06, "loss": 0.5523, "step": 5165 }, { "epoch": 0.35, "learning_rate": 8.771632168306753e-06, "loss": 1.0335, "step": 5170 }, { "epoch": 0.35, "learning_rate": 8.780115371564304e-06, "loss": 0.79, "step": 5175 }, { "epoch": 0.35, "learning_rate": 8.788598574821854e-06, "loss": 2.2763, "step": 5180 }, { "epoch": 0.35, "learning_rate": 8.797081778079404e-06, "loss": 1.1138, "step": 5185 }, { "epoch": 0.35, "learning_rate": 8.805564981336954e-06, "loss": 0.5377, "step": 5190 }, { "epoch": 0.35, "learning_rate": 8.814048184594504e-06, "loss": 1.0254, "step": 5195 }, { "epoch": 0.35, "learning_rate": 8.822531387852054e-06, "loss": 0.7949, "step": 5200 }, { "epoch": 0.35, "learning_rate": 8.831014591109603e-06, "loss": 0.4494, "step": 5205 }, { "epoch": 0.35, "learning_rate": 8.839497794367153e-06, "loss": 1.0987, "step": 5210 }, { "epoch": 0.35, "learning_rate": 8.847980997624703e-06, "loss": 1.8193, "step": 5215 }, { "epoch": 0.35, "learning_rate": 8.856464200882255e-06, "loss": 0.6839, "step": 5220 }, { "epoch": 0.35, "learning_rate": 8.864947404139805e-06, "loss": 0.3925, "step": 5225 }, { "epoch": 0.35, "learning_rate": 8.873430607397354e-06, "loss": 0.9795, "step": 5230 }, { "epoch": 0.36, "learning_rate": 8.881913810654904e-06, "loss": 0.6632, "step": 5235 }, { "epoch": 0.36, "learning_rate": 8.890397013912454e-06, "loss": 0.6987, "step": 5240 }, { "epoch": 0.36, "learning_rate": 8.898880217170004e-06, "loss": 0.2713, "step": 5245 }, { "epoch": 0.36, "learning_rate": 8.907363420427554e-06, "loss": 0.5676, "step": 5250 }, { "epoch": 0.36, "learning_rate": 8.915846623685104e-06, "loss": 0.8742, "step": 5255 }, { "epoch": 0.36, "learning_rate": 8.924329826942655e-06, "loss": 0.4303, "step": 5260 }, { "epoch": 0.36, "learning_rate": 8.932813030200205e-06, "loss": 1.725, "step": 5265 }, { "epoch": 0.36, "learning_rate": 8.941296233457755e-06, "loss": 1.0686, "step": 5270 }, { "epoch": 0.36, "learning_rate": 8.949779436715305e-06, "loss": 1.3809, "step": 5275 }, { "epoch": 0.36, "learning_rate": 8.958262639972855e-06, "loss": 1.5011, "step": 5280 }, { "epoch": 0.36, "learning_rate": 8.966745843230404e-06, "loss": 0.7444, "step": 5285 }, { "epoch": 0.36, "learning_rate": 8.975229046487954e-06, "loss": 1.5068, "step": 5290 }, { "epoch": 0.36, "learning_rate": 8.983712249745504e-06, "loss": 0.8037, "step": 5295 }, { "epoch": 0.36, "learning_rate": 8.992195453003054e-06, "loss": 0.4176, "step": 5300 }, { "epoch": 0.36, "learning_rate": 9.000678656260606e-06, "loss": 1.2049, "step": 5305 }, { "epoch": 0.36, "learning_rate": 9.009161859518155e-06, "loss": 0.7887, "step": 5310 }, { "epoch": 0.36, "learning_rate": 9.017645062775705e-06, "loss": 1.6319, "step": 5315 }, { "epoch": 0.36, "learning_rate": 9.026128266033255e-06, "loss": 0.9672, "step": 5320 }, { "epoch": 0.36, "learning_rate": 9.034611469290805e-06, "loss": 1.0172, "step": 5325 }, { "epoch": 0.36, "learning_rate": 9.043094672548355e-06, "loss": 1.7704, "step": 5330 }, { "epoch": 0.36, "learning_rate": 9.051577875805905e-06, "loss": 0.492, "step": 5335 }, { "epoch": 0.36, "learning_rate": 9.060061079063454e-06, "loss": 1.1826, "step": 5340 }, { "epoch": 0.36, "learning_rate": 9.068544282321006e-06, "loss": 1.1053, "step": 5345 }, { "epoch": 0.36, "learning_rate": 9.077027485578556e-06, "loss": 0.782, "step": 5350 }, { "epoch": 0.36, "learning_rate": 9.085510688836106e-06, "loss": 0.9388, "step": 5355 }, { "epoch": 0.36, "learning_rate": 9.093993892093656e-06, "loss": 0.6971, "step": 5360 }, { "epoch": 0.36, "learning_rate": 9.102477095351205e-06, "loss": 0.6562, "step": 5365 }, { "epoch": 0.36, "learning_rate": 9.110960298608755e-06, "loss": 0.9881, "step": 5370 }, { "epoch": 0.36, "learning_rate": 9.119443501866305e-06, "loss": 1.5203, "step": 5375 }, { "epoch": 0.37, "learning_rate": 9.127926705123855e-06, "loss": 0.6598, "step": 5380 }, { "epoch": 0.37, "learning_rate": 9.136409908381405e-06, "loss": 0.7245, "step": 5385 }, { "epoch": 0.37, "learning_rate": 9.144893111638956e-06, "loss": 1.1115, "step": 5390 }, { "epoch": 0.37, "learning_rate": 9.153376314896506e-06, "loss": 0.9858, "step": 5395 }, { "epoch": 0.37, "learning_rate": 9.161859518154056e-06, "loss": 0.8913, "step": 5400 }, { "epoch": 0.37, "learning_rate": 9.170342721411606e-06, "loss": 1.102, "step": 5405 }, { "epoch": 0.37, "learning_rate": 9.178825924669156e-06, "loss": 1.1261, "step": 5410 }, { "epoch": 0.37, "learning_rate": 9.187309127926706e-06, "loss": 0.9821, "step": 5415 }, { "epoch": 0.37, "learning_rate": 9.195792331184255e-06, "loss": 0.7531, "step": 5420 }, { "epoch": 0.37, "learning_rate": 9.204275534441805e-06, "loss": 0.7423, "step": 5425 }, { "epoch": 0.37, "learning_rate": 9.212758737699357e-06, "loss": 0.8295, "step": 5430 }, { "epoch": 0.37, "learning_rate": 9.221241940956907e-06, "loss": 0.7916, "step": 5435 }, { "epoch": 0.37, "learning_rate": 9.229725144214457e-06, "loss": 1.2781, "step": 5440 }, { "epoch": 0.37, "learning_rate": 9.238208347472006e-06, "loss": 0.6712, "step": 5445 }, { "epoch": 0.37, "learning_rate": 9.246691550729556e-06, "loss": 1.3994, "step": 5450 }, { "epoch": 0.37, "learning_rate": 9.255174753987106e-06, "loss": 0.8901, "step": 5455 }, { "epoch": 0.37, "learning_rate": 9.263657957244656e-06, "loss": 0.8833, "step": 5460 }, { "epoch": 0.37, "learning_rate": 9.272141160502206e-06, "loss": 1.744, "step": 5465 }, { "epoch": 0.37, "learning_rate": 9.280624363759756e-06, "loss": 0.962, "step": 5470 }, { "epoch": 0.37, "learning_rate": 9.289107567017307e-06, "loss": 0.9846, "step": 5475 }, { "epoch": 0.37, "learning_rate": 9.297590770274857e-06, "loss": 0.5495, "step": 5480 }, { "epoch": 0.37, "learning_rate": 9.306073973532407e-06, "loss": 1.1324, "step": 5485 }, { "epoch": 0.37, "learning_rate": 9.314557176789957e-06, "loss": 1.3536, "step": 5490 }, { "epoch": 0.37, "learning_rate": 9.323040380047507e-06, "loss": 0.4835, "step": 5495 }, { "epoch": 0.37, "learning_rate": 9.331523583305056e-06, "loss": 0.3695, "step": 5500 }, { "epoch": 0.37, "learning_rate": 9.340006786562606e-06, "loss": 0.9273, "step": 5505 }, { "epoch": 0.37, "learning_rate": 9.348489989820156e-06, "loss": 0.8492, "step": 5510 }, { "epoch": 0.37, "learning_rate": 9.356973193077708e-06, "loss": 1.2052, "step": 5515 }, { "epoch": 0.37, "learning_rate": 9.365456396335258e-06, "loss": 0.7919, "step": 5520 }, { "epoch": 0.37, "learning_rate": 9.373939599592807e-06, "loss": 0.9396, "step": 5525 }, { "epoch": 0.38, "learning_rate": 9.382422802850357e-06, "loss": 1.3733, "step": 5530 }, { "epoch": 0.38, "learning_rate": 9.390906006107907e-06, "loss": 1.8256, "step": 5535 }, { "epoch": 0.38, "learning_rate": 9.399389209365457e-06, "loss": 0.9621, "step": 5540 }, { "epoch": 0.38, "learning_rate": 9.407872412623007e-06, "loss": 0.7146, "step": 5545 }, { "epoch": 0.38, "learning_rate": 9.416355615880557e-06, "loss": 0.5633, "step": 5550 }, { "epoch": 0.38, "learning_rate": 9.424838819138106e-06, "loss": 1.097, "step": 5555 }, { "epoch": 0.38, "learning_rate": 9.433322022395658e-06, "loss": 0.9205, "step": 5560 }, { "epoch": 0.38, "learning_rate": 9.441805225653208e-06, "loss": 1.0267, "step": 5565 }, { "epoch": 0.38, "learning_rate": 9.450288428910758e-06, "loss": 0.5468, "step": 5570 }, { "epoch": 0.38, "learning_rate": 9.458771632168308e-06, "loss": 1.4946, "step": 5575 }, { "epoch": 0.38, "learning_rate": 9.467254835425857e-06, "loss": 0.9944, "step": 5580 }, { "epoch": 0.38, "learning_rate": 9.475738038683407e-06, "loss": 0.4883, "step": 5585 }, { "epoch": 0.38, "learning_rate": 9.484221241940957e-06, "loss": 0.7073, "step": 5590 }, { "epoch": 0.38, "learning_rate": 9.492704445198507e-06, "loss": 1.2007, "step": 5595 }, { "epoch": 0.38, "learning_rate": 9.501187648456057e-06, "loss": 0.5861, "step": 5600 }, { "epoch": 0.38, "learning_rate": 9.509670851713608e-06, "loss": 0.9652, "step": 5605 }, { "epoch": 0.38, "learning_rate": 9.518154054971158e-06, "loss": 1.3257, "step": 5610 }, { "epoch": 0.38, "learning_rate": 9.526637258228708e-06, "loss": 1.0473, "step": 5615 }, { "epoch": 0.38, "learning_rate": 9.535120461486258e-06, "loss": 0.8353, "step": 5620 }, { "epoch": 0.38, "learning_rate": 9.543603664743808e-06, "loss": 1.008, "step": 5625 }, { "epoch": 0.38, "learning_rate": 9.552086868001358e-06, "loss": 1.0544, "step": 5630 }, { "epoch": 0.38, "learning_rate": 9.560570071258907e-06, "loss": 0.8825, "step": 5635 }, { "epoch": 0.38, "learning_rate": 9.569053274516457e-06, "loss": 1.0643, "step": 5640 }, { "epoch": 0.38, "learning_rate": 9.577536477774009e-06, "loss": 1.552, "step": 5645 }, { "epoch": 0.38, "learning_rate": 9.586019681031559e-06, "loss": 0.8746, "step": 5650 }, { "epoch": 0.38, "learning_rate": 9.594502884289109e-06, "loss": 0.653, "step": 5655 }, { "epoch": 0.38, "learning_rate": 9.602986087546658e-06, "loss": 1.0789, "step": 5660 }, { "epoch": 0.38, "learning_rate": 9.611469290804208e-06, "loss": 0.7455, "step": 5665 }, { "epoch": 0.38, "learning_rate": 9.619952494061758e-06, "loss": 1.4258, "step": 5670 }, { "epoch": 0.39, "learning_rate": 9.628435697319308e-06, "loss": 0.8333, "step": 5675 }, { "epoch": 0.39, "learning_rate": 9.636918900576858e-06, "loss": 0.8894, "step": 5680 }, { "epoch": 0.39, "learning_rate": 9.645402103834408e-06, "loss": 1.4503, "step": 5685 }, { "epoch": 0.39, "learning_rate": 9.65388530709196e-06, "loss": 1.1401, "step": 5690 }, { "epoch": 0.39, "learning_rate": 9.662368510349509e-06, "loss": 0.8047, "step": 5695 }, { "epoch": 0.39, "learning_rate": 9.670851713607059e-06, "loss": 0.6484, "step": 5700 }, { "epoch": 0.39, "learning_rate": 9.679334916864609e-06, "loss": 0.8004, "step": 5705 }, { "epoch": 0.39, "learning_rate": 9.687818120122159e-06, "loss": 1.2449, "step": 5710 }, { "epoch": 0.39, "learning_rate": 9.696301323379708e-06, "loss": 0.7005, "step": 5715 }, { "epoch": 0.39, "learning_rate": 9.704784526637258e-06, "loss": 0.6057, "step": 5720 }, { "epoch": 0.39, "learning_rate": 9.713267729894808e-06, "loss": 1.6501, "step": 5725 }, { "epoch": 0.39, "learning_rate": 9.72175093315236e-06, "loss": 1.2677, "step": 5730 }, { "epoch": 0.39, "learning_rate": 9.73023413640991e-06, "loss": 0.4781, "step": 5735 }, { "epoch": 0.39, "learning_rate": 9.73871733966746e-06, "loss": 0.8355, "step": 5740 }, { "epoch": 0.39, "learning_rate": 9.74720054292501e-06, "loss": 0.9568, "step": 5745 }, { "epoch": 0.39, "learning_rate": 9.755683746182559e-06, "loss": 0.5848, "step": 5750 }, { "epoch": 0.39, "learning_rate": 9.764166949440109e-06, "loss": 0.778, "step": 5755 }, { "epoch": 0.39, "learning_rate": 9.772650152697659e-06, "loss": 1.127, "step": 5760 }, { "epoch": 0.39, "learning_rate": 9.781133355955209e-06, "loss": 0.687, "step": 5765 }, { "epoch": 0.39, "learning_rate": 9.789616559212758e-06, "loss": 1.7625, "step": 5770 }, { "epoch": 0.39, "learning_rate": 9.79809976247031e-06, "loss": 0.7685, "step": 5775 }, { "epoch": 0.39, "learning_rate": 9.80658296572786e-06, "loss": 0.586, "step": 5780 }, { "epoch": 0.39, "learning_rate": 9.81506616898541e-06, "loss": 1.1315, "step": 5785 }, { "epoch": 0.39, "learning_rate": 9.82354937224296e-06, "loss": 0.6479, "step": 5790 }, { "epoch": 0.39, "learning_rate": 9.83203257550051e-06, "loss": 0.1812, "step": 5795 }, { "epoch": 0.39, "learning_rate": 9.84051577875806e-06, "loss": 0.8863, "step": 5800 }, { "epoch": 0.39, "learning_rate": 9.848998982015609e-06, "loss": 2.0463, "step": 5805 }, { "epoch": 0.39, "learning_rate": 9.857482185273159e-06, "loss": 0.827, "step": 5810 }, { "epoch": 0.39, "learning_rate": 9.86596538853071e-06, "loss": 0.7887, "step": 5815 }, { "epoch": 0.4, "learning_rate": 9.87444859178826e-06, "loss": 0.6683, "step": 5820 }, { "epoch": 0.4, "learning_rate": 9.88293179504581e-06, "loss": 0.6108, "step": 5825 }, { "epoch": 0.4, "learning_rate": 9.89141499830336e-06, "loss": 0.8801, "step": 5830 }, { "epoch": 0.4, "learning_rate": 9.89989820156091e-06, "loss": 1.2398, "step": 5835 }, { "epoch": 0.4, "learning_rate": 9.90838140481846e-06, "loss": 0.4713, "step": 5840 }, { "epoch": 0.4, "learning_rate": 9.91686460807601e-06, "loss": 1.2452, "step": 5845 }, { "epoch": 0.4, "learning_rate": 9.92534781133356e-06, "loss": 0.886, "step": 5850 }, { "epoch": 0.4, "learning_rate": 9.93383101459111e-06, "loss": 1.3425, "step": 5855 }, { "epoch": 0.4, "learning_rate": 9.94231421784866e-06, "loss": 0.7112, "step": 5860 }, { "epoch": 0.4, "learning_rate": 9.95079742110621e-06, "loss": 0.792, "step": 5865 }, { "epoch": 0.4, "learning_rate": 9.95928062436376e-06, "loss": 1.0448, "step": 5870 }, { "epoch": 0.4, "learning_rate": 9.96776382762131e-06, "loss": 1.2078, "step": 5875 }, { "epoch": 0.4, "learning_rate": 9.97624703087886e-06, "loss": 0.8671, "step": 5880 }, { "epoch": 0.4, "learning_rate": 9.98473023413641e-06, "loss": 0.7805, "step": 5885 }, { "epoch": 0.4, "learning_rate": 9.99321343739396e-06, "loss": 0.277, "step": 5890 }, { "epoch": 0.4, "learning_rate": 9.999811470155728e-06, "loss": 0.988, "step": 5895 }, { "epoch": 0.4, "learning_rate": 9.998868820934354e-06, "loss": 0.376, "step": 5900 }, { "epoch": 0.4, "learning_rate": 9.997926171712983e-06, "loss": 0.8591, "step": 5905 }, { "epoch": 0.4, "learning_rate": 9.99698352249161e-06, "loss": 0.6865, "step": 5910 }, { "epoch": 0.4, "learning_rate": 9.99604087327024e-06, "loss": 0.8485, "step": 5915 }, { "epoch": 0.4, "learning_rate": 9.995098224048868e-06, "loss": 0.7608, "step": 5920 }, { "epoch": 0.4, "learning_rate": 9.994155574827497e-06, "loss": 0.799, "step": 5925 }, { "epoch": 0.4, "learning_rate": 9.993212925606125e-06, "loss": 0.8182, "step": 5930 }, { "epoch": 0.4, "learning_rate": 9.992270276384752e-06, "loss": 0.646, "step": 5935 }, { "epoch": 0.4, "learning_rate": 9.991327627163382e-06, "loss": 0.9654, "step": 5940 }, { "epoch": 0.4, "learning_rate": 9.99038497794201e-06, "loss": 0.6468, "step": 5945 }, { "epoch": 0.4, "learning_rate": 9.989442328720639e-06, "loss": 0.7318, "step": 5950 }, { "epoch": 0.4, "learning_rate": 9.988499679499264e-06, "loss": 0.6423, "step": 5955 }, { "epoch": 0.4, "learning_rate": 9.987557030277894e-06, "loss": 1.7815, "step": 5960 }, { "epoch": 0.4, "learning_rate": 9.986614381056521e-06, "loss": 0.6116, "step": 5965 }, { "epoch": 0.41, "learning_rate": 9.98567173183515e-06, "loss": 0.4615, "step": 5970 }, { "epoch": 0.41, "learning_rate": 9.984729082613778e-06, "loss": 0.5407, "step": 5975 }, { "epoch": 0.41, "learning_rate": 9.983786433392408e-06, "loss": 0.992, "step": 5980 }, { "epoch": 0.41, "learning_rate": 9.982843784171035e-06, "loss": 0.9708, "step": 5985 }, { "epoch": 0.41, "learning_rate": 9.981901134949663e-06, "loss": 1.6125, "step": 5990 }, { "epoch": 0.41, "learning_rate": 9.98095848572829e-06, "loss": 1.7771, "step": 5995 }, { "epoch": 0.41, "learning_rate": 9.98001583650692e-06, "loss": 1.5413, "step": 6000 }, { "epoch": 0.41, "learning_rate": 9.979073187285548e-06, "loss": 1.1026, "step": 6005 }, { "epoch": 0.41, "learning_rate": 9.978130538064177e-06, "loss": 0.7229, "step": 6010 }, { "epoch": 0.41, "learning_rate": 9.977187888842805e-06, "loss": 0.6937, "step": 6015 }, { "epoch": 0.41, "learning_rate": 9.976245239621432e-06, "loss": 0.9506, "step": 6020 }, { "epoch": 0.41, "learning_rate": 9.975302590400062e-06, "loss": 0.7454, "step": 6025 }, { "epoch": 0.41, "learning_rate": 9.97435994117869e-06, "loss": 0.9027, "step": 6030 }, { "epoch": 0.41, "learning_rate": 9.973417291957319e-06, "loss": 0.4322, "step": 6035 }, { "epoch": 0.41, "learning_rate": 9.972474642735946e-06, "loss": 0.8788, "step": 6040 }, { "epoch": 0.41, "learning_rate": 9.971531993514574e-06, "loss": 1.2602, "step": 6045 }, { "epoch": 0.41, "learning_rate": 9.970589344293201e-06, "loss": 0.7465, "step": 6050 }, { "epoch": 0.41, "learning_rate": 9.96964669507183e-06, "loss": 0.6622, "step": 6055 }, { "epoch": 0.41, "learning_rate": 9.968704045850458e-06, "loss": 1.8235, "step": 6060 }, { "epoch": 0.41, "learning_rate": 9.967761396629088e-06, "loss": 0.9081, "step": 6065 }, { "epoch": 0.41, "learning_rate": 9.966818747407715e-06, "loss": 1.0773, "step": 6070 }, { "epoch": 0.41, "learning_rate": 9.965876098186343e-06, "loss": 0.7462, "step": 6075 }, { "epoch": 0.41, "learning_rate": 9.964933448964972e-06, "loss": 0.4128, "step": 6080 }, { "epoch": 0.41, "learning_rate": 9.9639907997436e-06, "loss": 0.8235, "step": 6085 }, { "epoch": 0.41, "learning_rate": 9.96304815052223e-06, "loss": 1.6523, "step": 6090 }, { "epoch": 0.41, "learning_rate": 9.962105501300857e-06, "loss": 0.9573, "step": 6095 }, { "epoch": 0.41, "learning_rate": 9.961162852079485e-06, "loss": 0.5877, "step": 6100 }, { "epoch": 0.41, "learning_rate": 9.960220202858112e-06, "loss": 0.5464, "step": 6105 }, { "epoch": 0.41, "learning_rate": 9.959277553636742e-06, "loss": 0.8579, "step": 6110 }, { "epoch": 0.42, "learning_rate": 9.95833490441537e-06, "loss": 0.7508, "step": 6115 }, { "epoch": 0.42, "learning_rate": 9.957392255193999e-06, "loss": 1.2709, "step": 6120 }, { "epoch": 0.42, "learning_rate": 9.956449605972626e-06, "loss": 1.4461, "step": 6125 }, { "epoch": 0.42, "learning_rate": 9.955506956751254e-06, "loss": 1.1453, "step": 6130 }, { "epoch": 0.42, "learning_rate": 9.954564307529883e-06, "loss": 1.4768, "step": 6135 }, { "epoch": 0.42, "learning_rate": 9.953621658308511e-06, "loss": 0.2804, "step": 6140 }, { "epoch": 0.42, "learning_rate": 9.95267900908714e-06, "loss": 1.5882, "step": 6145 }, { "epoch": 0.42, "learning_rate": 9.951736359865768e-06, "loss": 1.2771, "step": 6150 }, { "epoch": 0.42, "learning_rate": 9.950793710644397e-06, "loss": 0.781, "step": 6155 }, { "epoch": 0.42, "learning_rate": 9.949851061423023e-06, "loss": 0.3397, "step": 6160 }, { "epoch": 0.42, "learning_rate": 9.948908412201652e-06, "loss": 0.9141, "step": 6165 }, { "epoch": 0.42, "learning_rate": 9.94796576298028e-06, "loss": 0.6929, "step": 6170 }, { "epoch": 0.42, "learning_rate": 9.94702311375891e-06, "loss": 0.4914, "step": 6175 }, { "epoch": 0.42, "learning_rate": 9.946080464537537e-06, "loss": 1.1609, "step": 6180 }, { "epoch": 0.42, "learning_rate": 9.945137815316165e-06, "loss": 1.0277, "step": 6185 }, { "epoch": 0.42, "learning_rate": 9.944195166094794e-06, "loss": 1.2514, "step": 6190 }, { "epoch": 0.42, "learning_rate": 9.943252516873422e-06, "loss": 1.0548, "step": 6195 }, { "epoch": 0.42, "learning_rate": 9.942309867652051e-06, "loss": 0.8746, "step": 6200 }, { "epoch": 0.42, "learning_rate": 9.941367218430679e-06, "loss": 0.6566, "step": 6205 }, { "epoch": 0.42, "learning_rate": 9.940424569209306e-06, "loss": 1.1599, "step": 6210 }, { "epoch": 0.42, "learning_rate": 9.939481919987934e-06, "loss": 1.1763, "step": 6215 }, { "epoch": 0.42, "learning_rate": 9.938539270766563e-06, "loss": 0.7576, "step": 6220 }, { "epoch": 0.42, "learning_rate": 9.937596621545191e-06, "loss": 0.7003, "step": 6225 }, { "epoch": 0.42, "learning_rate": 9.93665397232382e-06, "loss": 1.1385, "step": 6230 }, { "epoch": 0.42, "learning_rate": 9.935711323102448e-06, "loss": 0.5443, "step": 6235 }, { "epoch": 0.42, "learning_rate": 9.934768673881077e-06, "loss": 0.7912, "step": 6240 }, { "epoch": 0.42, "learning_rate": 9.933826024659705e-06, "loss": 0.777, "step": 6245 }, { "epoch": 0.42, "learning_rate": 9.932883375438333e-06, "loss": 0.5917, "step": 6250 }, { "epoch": 0.42, "learning_rate": 9.93194072621696e-06, "loss": 1.7153, "step": 6255 }, { "epoch": 0.42, "learning_rate": 9.93099807699559e-06, "loss": 0.4518, "step": 6260 }, { "epoch": 0.43, "learning_rate": 9.930055427774217e-06, "loss": 0.7887, "step": 6265 }, { "epoch": 0.43, "learning_rate": 9.929112778552846e-06, "loss": 0.7703, "step": 6270 }, { "epoch": 0.43, "learning_rate": 9.928170129331474e-06, "loss": 1.004, "step": 6275 }, { "epoch": 0.43, "learning_rate": 9.927227480110102e-06, "loss": 1.0693, "step": 6280 }, { "epoch": 0.43, "learning_rate": 9.926284830888731e-06, "loss": 0.6039, "step": 6285 }, { "epoch": 0.43, "learning_rate": 9.925342181667359e-06, "loss": 0.5168, "step": 6290 }, { "epoch": 0.43, "learning_rate": 9.924399532445988e-06, "loss": 0.3902, "step": 6295 }, { "epoch": 0.43, "learning_rate": 9.923456883224614e-06, "loss": 0.7502, "step": 6300 }, { "epoch": 0.43, "learning_rate": 9.922514234003243e-06, "loss": 1.0979, "step": 6305 }, { "epoch": 0.43, "learning_rate": 9.921571584781871e-06, "loss": 0.9006, "step": 6310 }, { "epoch": 0.43, "learning_rate": 9.9206289355605e-06, "loss": 0.5772, "step": 6315 }, { "epoch": 0.43, "learning_rate": 9.919686286339128e-06, "loss": 0.521, "step": 6320 }, { "epoch": 0.43, "learning_rate": 9.918743637117757e-06, "loss": 0.6505, "step": 6325 }, { "epoch": 0.43, "learning_rate": 9.917800987896385e-06, "loss": 0.7096, "step": 6330 }, { "epoch": 0.43, "learning_rate": 9.916858338675013e-06, "loss": 1.1079, "step": 6335 }, { "epoch": 0.43, "learning_rate": 9.915915689453642e-06, "loss": 0.6465, "step": 6340 }, { "epoch": 0.43, "learning_rate": 9.91497304023227e-06, "loss": 0.9689, "step": 6345 }, { "epoch": 0.43, "learning_rate": 9.914030391010899e-06, "loss": 1.3404, "step": 6350 }, { "epoch": 0.43, "learning_rate": 9.913087741789527e-06, "loss": 0.6054, "step": 6355 }, { "epoch": 0.43, "learning_rate": 9.912145092568154e-06, "loss": 1.3919, "step": 6360 }, { "epoch": 0.43, "learning_rate": 9.911202443346782e-06, "loss": 0.8436, "step": 6365 }, { "epoch": 0.43, "learning_rate": 9.910259794125411e-06, "loss": 0.7655, "step": 6370 }, { "epoch": 0.43, "learning_rate": 9.909317144904039e-06, "loss": 0.7314, "step": 6375 }, { "epoch": 0.43, "learning_rate": 9.908374495682668e-06, "loss": 1.1413, "step": 6380 }, { "epoch": 0.43, "learning_rate": 9.907431846461296e-06, "loss": 0.4117, "step": 6385 }, { "epoch": 0.43, "learning_rate": 9.906489197239923e-06, "loss": 0.6796, "step": 6390 }, { "epoch": 0.43, "learning_rate": 9.905546548018553e-06, "loss": 0.6411, "step": 6395 }, { "epoch": 0.43, "learning_rate": 9.90460389879718e-06, "loss": 0.7729, "step": 6400 }, { "epoch": 0.43, "learning_rate": 9.90366124957581e-06, "loss": 0.4727, "step": 6405 }, { "epoch": 0.44, "learning_rate": 9.902718600354437e-06, "loss": 1.253, "step": 6410 }, { "epoch": 0.44, "learning_rate": 9.901775951133065e-06, "loss": 0.8367, "step": 6415 }, { "epoch": 0.44, "learning_rate": 9.900833301911693e-06, "loss": 1.2505, "step": 6420 }, { "epoch": 0.44, "learning_rate": 9.899890652690322e-06, "loss": 0.6117, "step": 6425 }, { "epoch": 0.44, "learning_rate": 9.89894800346895e-06, "loss": 0.5446, "step": 6430 }, { "epoch": 0.44, "learning_rate": 9.898005354247579e-06, "loss": 0.8091, "step": 6435 }, { "epoch": 0.44, "learning_rate": 9.897062705026207e-06, "loss": 0.7265, "step": 6440 }, { "epoch": 0.44, "learning_rate": 9.896120055804834e-06, "loss": 0.3654, "step": 6445 }, { "epoch": 0.44, "learning_rate": 9.895177406583464e-06, "loss": 0.9622, "step": 6450 }, { "epoch": 0.44, "learning_rate": 9.894234757362091e-06, "loss": 0.7965, "step": 6455 }, { "epoch": 0.44, "learning_rate": 9.893292108140719e-06, "loss": 0.4324, "step": 6460 }, { "epoch": 0.44, "learning_rate": 9.892349458919348e-06, "loss": 1.3143, "step": 6465 }, { "epoch": 0.44, "learning_rate": 9.891406809697976e-06, "loss": 0.4852, "step": 6470 }, { "epoch": 0.44, "learning_rate": 9.890464160476603e-06, "loss": 1.6383, "step": 6475 }, { "epoch": 0.44, "learning_rate": 9.889521511255233e-06, "loss": 0.2983, "step": 6480 }, { "epoch": 0.44, "learning_rate": 9.88857886203386e-06, "loss": 1.0196, "step": 6485 }, { "epoch": 0.44, "learning_rate": 9.88763621281249e-06, "loss": 0.502, "step": 6490 }, { "epoch": 0.44, "learning_rate": 9.886693563591117e-06, "loss": 0.8654, "step": 6495 }, { "epoch": 0.44, "learning_rate": 9.885750914369745e-06, "loss": 0.586, "step": 6500 }, { "epoch": 0.44, "learning_rate": 9.884808265148373e-06, "loss": 1.0781, "step": 6505 }, { "epoch": 0.44, "learning_rate": 9.883865615927002e-06, "loss": 1.0514, "step": 6510 }, { "epoch": 0.44, "learning_rate": 9.88292296670563e-06, "loss": 0.7407, "step": 6515 }, { "epoch": 0.44, "learning_rate": 9.881980317484259e-06, "loss": 1.0283, "step": 6520 }, { "epoch": 0.44, "learning_rate": 9.881037668262887e-06, "loss": 0.5502, "step": 6525 }, { "epoch": 0.44, "learning_rate": 9.880095019041514e-06, "loss": 0.4663, "step": 6530 }, { "epoch": 0.44, "learning_rate": 9.879152369820144e-06, "loss": 0.9316, "step": 6535 }, { "epoch": 0.44, "learning_rate": 9.878209720598771e-06, "loss": 0.6606, "step": 6540 }, { "epoch": 0.44, "learning_rate": 9.8772670713774e-06, "loss": 0.5616, "step": 6545 }, { "epoch": 0.44, "learning_rate": 9.876324422156028e-06, "loss": 0.8021, "step": 6550 }, { "epoch": 0.44, "learning_rate": 9.875381772934658e-06, "loss": 0.9607, "step": 6555 }, { "epoch": 0.45, "learning_rate": 9.874439123713283e-06, "loss": 0.5125, "step": 6560 }, { "epoch": 0.45, "learning_rate": 9.873496474491913e-06, "loss": 0.4531, "step": 6565 }, { "epoch": 0.45, "learning_rate": 9.87255382527054e-06, "loss": 0.5896, "step": 6570 }, { "epoch": 0.45, "learning_rate": 9.87161117604917e-06, "loss": 0.8416, "step": 6575 }, { "epoch": 0.45, "learning_rate": 9.870668526827797e-06, "loss": 0.7469, "step": 6580 }, { "epoch": 0.45, "learning_rate": 9.869725877606427e-06, "loss": 0.4114, "step": 6585 }, { "epoch": 0.45, "learning_rate": 9.868783228385054e-06, "loss": 0.6022, "step": 6590 }, { "epoch": 0.45, "learning_rate": 9.867840579163682e-06, "loss": 0.5907, "step": 6595 }, { "epoch": 0.45, "learning_rate": 9.866897929942311e-06, "loss": 1.3983, "step": 6600 }, { "epoch": 0.45, "learning_rate": 9.865955280720939e-06, "loss": 1.1922, "step": 6605 }, { "epoch": 0.45, "learning_rate": 9.865012631499568e-06, "loss": 1.4288, "step": 6610 }, { "epoch": 0.45, "learning_rate": 9.864069982278194e-06, "loss": 0.6755, "step": 6615 }, { "epoch": 0.45, "learning_rate": 9.863127333056824e-06, "loss": 0.94, "step": 6620 }, { "epoch": 0.45, "learning_rate": 9.862184683835451e-06, "loss": 0.8654, "step": 6625 }, { "epoch": 0.45, "learning_rate": 9.86124203461408e-06, "loss": 0.5205, "step": 6630 }, { "epoch": 0.45, "learning_rate": 9.860299385392708e-06, "loss": 0.7568, "step": 6635 }, { "epoch": 0.45, "learning_rate": 9.859356736171338e-06, "loss": 1.0904, "step": 6640 }, { "epoch": 0.45, "learning_rate": 9.858414086949965e-06, "loss": 0.31, "step": 6645 }, { "epoch": 0.45, "learning_rate": 9.857471437728593e-06, "loss": 1.6787, "step": 6650 }, { "epoch": 0.45, "learning_rate": 9.856528788507222e-06, "loss": 0.4832, "step": 6655 }, { "epoch": 0.45, "learning_rate": 9.85558613928585e-06, "loss": 0.4121, "step": 6660 }, { "epoch": 0.45, "learning_rate": 9.854643490064477e-06, "loss": 0.7737, "step": 6665 }, { "epoch": 0.45, "learning_rate": 9.853700840843107e-06, "loss": 1.1114, "step": 6670 }, { "epoch": 0.45, "learning_rate": 9.852758191621734e-06, "loss": 0.2202, "step": 6675 }, { "epoch": 0.45, "learning_rate": 9.851815542400362e-06, "loss": 0.7079, "step": 6680 }, { "epoch": 0.45, "learning_rate": 9.850872893178991e-06, "loss": 0.6946, "step": 6685 }, { "epoch": 0.45, "learning_rate": 9.849930243957619e-06, "loss": 0.125, "step": 6690 }, { "epoch": 0.45, "learning_rate": 9.848987594736248e-06, "loss": 0.4598, "step": 6695 }, { "epoch": 0.45, "learning_rate": 9.848044945514876e-06, "loss": 1.6811, "step": 6700 }, { "epoch": 0.46, "learning_rate": 9.847102296293504e-06, "loss": 0.9009, "step": 6705 }, { "epoch": 0.46, "learning_rate": 9.846159647072131e-06, "loss": 0.4181, "step": 6710 }, { "epoch": 0.46, "learning_rate": 9.84521699785076e-06, "loss": 0.8075, "step": 6715 }, { "epoch": 0.46, "learning_rate": 9.844274348629388e-06, "loss": 0.5539, "step": 6720 }, { "epoch": 0.46, "learning_rate": 9.843331699408018e-06, "loss": 0.5026, "step": 6725 }, { "epoch": 0.46, "learning_rate": 9.842389050186645e-06, "loss": 1.214, "step": 6730 }, { "epoch": 0.46, "learning_rate": 9.841446400965273e-06, "loss": 0.4885, "step": 6735 }, { "epoch": 0.46, "learning_rate": 9.840503751743902e-06, "loss": 0.4925, "step": 6740 }, { "epoch": 0.46, "learning_rate": 9.83956110252253e-06, "loss": 0.7404, "step": 6745 }, { "epoch": 0.46, "learning_rate": 9.83861845330116e-06, "loss": 0.4439, "step": 6750 }, { "epoch": 0.46, "learning_rate": 9.837675804079787e-06, "loss": 1.1039, "step": 6755 }, { "epoch": 0.46, "learning_rate": 9.836733154858415e-06, "loss": 0.6431, "step": 6760 }, { "epoch": 0.46, "learning_rate": 9.835790505637042e-06, "loss": 0.7552, "step": 6765 }, { "epoch": 0.46, "learning_rate": 9.834847856415671e-06, "loss": 0.9154, "step": 6770 }, { "epoch": 0.46, "learning_rate": 9.833905207194299e-06, "loss": 1.2786, "step": 6775 }, { "epoch": 0.46, "learning_rate": 9.832962557972928e-06, "loss": 1.0242, "step": 6780 }, { "epoch": 0.46, "learning_rate": 9.832019908751556e-06, "loss": 1.3229, "step": 6785 }, { "epoch": 0.46, "learning_rate": 9.831077259530184e-06, "loss": 0.5537, "step": 6790 }, { "epoch": 0.46, "learning_rate": 9.830134610308813e-06, "loss": 0.5086, "step": 6795 }, { "epoch": 0.46, "learning_rate": 9.82919196108744e-06, "loss": 0.5434, "step": 6800 }, { "epoch": 0.46, "learning_rate": 9.82824931186607e-06, "loss": 0.7779, "step": 6805 }, { "epoch": 0.46, "learning_rate": 9.827306662644698e-06, "loss": 0.9756, "step": 6810 }, { "epoch": 0.46, "learning_rate": 9.826364013423327e-06, "loss": 0.3555, "step": 6815 }, { "epoch": 0.46, "learning_rate": 9.825421364201953e-06, "loss": 1.4772, "step": 6820 }, { "epoch": 0.46, "learning_rate": 9.824478714980582e-06, "loss": 0.2986, "step": 6825 }, { "epoch": 0.46, "learning_rate": 9.82353606575921e-06, "loss": 0.6126, "step": 6830 }, { "epoch": 0.46, "learning_rate": 9.82259341653784e-06, "loss": 0.4115, "step": 6835 }, { "epoch": 0.46, "learning_rate": 9.821650767316467e-06, "loss": 0.7273, "step": 6840 }, { "epoch": 0.46, "learning_rate": 9.820708118095095e-06, "loss": 0.7221, "step": 6845 }, { "epoch": 0.46, "learning_rate": 9.819765468873724e-06, "loss": 0.9196, "step": 6850 }, { "epoch": 0.47, "learning_rate": 9.818822819652352e-06, "loss": 1.2807, "step": 6855 }, { "epoch": 0.47, "learning_rate": 9.817880170430981e-06, "loss": 0.9923, "step": 6860 }, { "epoch": 0.47, "learning_rate": 9.816937521209608e-06, "loss": 0.6361, "step": 6865 }, { "epoch": 0.47, "learning_rate": 9.815994871988236e-06, "loss": 1.0015, "step": 6870 }, { "epoch": 0.47, "learning_rate": 9.815052222766864e-06, "loss": 1.5571, "step": 6875 }, { "epoch": 0.47, "learning_rate": 9.814109573545493e-06, "loss": 1.0959, "step": 6880 }, { "epoch": 0.47, "learning_rate": 9.81316692432412e-06, "loss": 0.5978, "step": 6885 }, { "epoch": 0.47, "learning_rate": 9.81222427510275e-06, "loss": 0.6464, "step": 6890 }, { "epoch": 0.47, "learning_rate": 9.811281625881378e-06, "loss": 0.4129, "step": 6895 }, { "epoch": 0.47, "learning_rate": 9.810338976660007e-06, "loss": 0.6321, "step": 6900 }, { "epoch": 0.47, "learning_rate": 9.809396327438635e-06, "loss": 0.3249, "step": 6905 }, { "epoch": 0.47, "learning_rate": 9.808453678217262e-06, "loss": 0.4265, "step": 6910 }, { "epoch": 0.47, "learning_rate": 9.80751102899589e-06, "loss": 0.206, "step": 6915 }, { "epoch": 0.47, "learning_rate": 9.80656837977452e-06, "loss": 0.5894, "step": 6920 }, { "epoch": 0.47, "learning_rate": 9.805625730553147e-06, "loss": 1.0164, "step": 6925 }, { "epoch": 0.47, "learning_rate": 9.804683081331776e-06, "loss": 0.8824, "step": 6930 }, { "epoch": 0.47, "learning_rate": 9.803740432110404e-06, "loss": 0.6342, "step": 6935 }, { "epoch": 0.47, "learning_rate": 9.802797782889032e-06, "loss": 0.9928, "step": 6940 }, { "epoch": 0.47, "learning_rate": 9.801855133667661e-06, "loss": 0.9913, "step": 6945 }, { "epoch": 0.47, "learning_rate": 9.800912484446289e-06, "loss": 1.3143, "step": 6950 }, { "epoch": 0.47, "learning_rate": 9.799969835224918e-06, "loss": 0.5476, "step": 6955 }, { "epoch": 0.47, "learning_rate": 9.799027186003544e-06, "loss": 1.2251, "step": 6960 }, { "epoch": 0.47, "learning_rate": 9.798084536782173e-06, "loss": 0.6634, "step": 6965 }, { "epoch": 0.47, "learning_rate": 9.7971418875608e-06, "loss": 1.0099, "step": 6970 }, { "epoch": 0.47, "learning_rate": 9.79619923833943e-06, "loss": 0.77, "step": 6975 }, { "epoch": 0.47, "learning_rate": 9.795256589118058e-06, "loss": 1.1927, "step": 6980 }, { "epoch": 0.47, "learning_rate": 9.794313939896687e-06, "loss": 1.1924, "step": 6985 }, { "epoch": 0.47, "learning_rate": 9.793371290675315e-06, "loss": 0.9006, "step": 6990 }, { "epoch": 0.47, "learning_rate": 9.792428641453942e-06, "loss": 1.1186, "step": 6995 }, { "epoch": 0.48, "learning_rate": 9.791485992232572e-06, "loss": 0.8349, "step": 7000 }, { "epoch": 0.48, "learning_rate": 9.7905433430112e-06, "loss": 1.221, "step": 7005 }, { "epoch": 0.48, "learning_rate": 9.789600693789829e-06, "loss": 0.5771, "step": 7010 }, { "epoch": 0.48, "learning_rate": 9.788658044568456e-06, "loss": 0.4662, "step": 7015 }, { "epoch": 0.48, "learning_rate": 9.787715395347084e-06, "loss": 0.6975, "step": 7020 }, { "epoch": 0.48, "learning_rate": 9.786772746125712e-06, "loss": 0.5269, "step": 7025 }, { "epoch": 0.48, "learning_rate": 9.785830096904341e-06, "loss": 0.6436, "step": 7030 }, { "epoch": 0.48, "learning_rate": 9.784887447682969e-06, "loss": 0.7849, "step": 7035 }, { "epoch": 0.48, "learning_rate": 9.783944798461598e-06, "loss": 0.8534, "step": 7040 }, { "epoch": 0.48, "learning_rate": 9.783002149240226e-06, "loss": 0.2178, "step": 7045 }, { "epoch": 0.48, "learning_rate": 9.782059500018853e-06, "loss": 1.186, "step": 7050 }, { "epoch": 0.48, "learning_rate": 9.781116850797483e-06, "loss": 0.7697, "step": 7055 }, { "epoch": 0.48, "learning_rate": 9.78017420157611e-06, "loss": 0.359, "step": 7060 }, { "epoch": 0.48, "learning_rate": 9.77923155235474e-06, "loss": 1.01, "step": 7065 }, { "epoch": 0.48, "learning_rate": 9.778288903133367e-06, "loss": 1.2608, "step": 7070 }, { "epoch": 0.48, "learning_rate": 9.777346253911995e-06, "loss": 0.8817, "step": 7075 }, { "epoch": 0.48, "learning_rate": 9.776403604690622e-06, "loss": 0.3101, "step": 7080 }, { "epoch": 0.48, "learning_rate": 9.775460955469252e-06, "loss": 0.4256, "step": 7085 }, { "epoch": 0.48, "learning_rate": 9.77451830624788e-06, "loss": 1.0153, "step": 7090 }, { "epoch": 0.48, "learning_rate": 9.773575657026509e-06, "loss": 1.1794, "step": 7095 }, { "epoch": 0.48, "learning_rate": 9.772633007805136e-06, "loss": 0.2896, "step": 7100 }, { "epoch": 0.48, "learning_rate": 9.771690358583764e-06, "loss": 0.699, "step": 7105 }, { "epoch": 0.48, "learning_rate": 9.770747709362393e-06, "loss": 0.8165, "step": 7110 }, { "epoch": 0.48, "learning_rate": 9.769805060141021e-06, "loss": 0.8991, "step": 7115 }, { "epoch": 0.48, "learning_rate": 9.768862410919649e-06, "loss": 1.0218, "step": 7120 }, { "epoch": 0.48, "learning_rate": 9.767919761698278e-06, "loss": 0.5677, "step": 7125 }, { "epoch": 0.48, "learning_rate": 9.766977112476906e-06, "loss": 0.3838, "step": 7130 }, { "epoch": 0.48, "learning_rate": 9.766034463255533e-06, "loss": 0.6518, "step": 7135 }, { "epoch": 0.48, "learning_rate": 9.765091814034163e-06, "loss": 1.1201, "step": 7140 }, { "epoch": 0.48, "learning_rate": 9.76414916481279e-06, "loss": 1.0007, "step": 7145 }, { "epoch": 0.49, "learning_rate": 9.76320651559142e-06, "loss": 0.4935, "step": 7150 }, { "epoch": 0.49, "learning_rate": 9.762263866370047e-06, "loss": 0.5882, "step": 7155 }, { "epoch": 0.49, "learning_rate": 9.761321217148677e-06, "loss": 0.4001, "step": 7160 }, { "epoch": 0.49, "learning_rate": 9.760378567927302e-06, "loss": 1.0297, "step": 7165 }, { "epoch": 0.49, "learning_rate": 9.759435918705932e-06, "loss": 1.1161, "step": 7170 }, { "epoch": 0.49, "learning_rate": 9.75849326948456e-06, "loss": 0.404, "step": 7175 }, { "epoch": 0.49, "learning_rate": 9.757550620263189e-06, "loss": 0.9635, "step": 7180 }, { "epoch": 0.49, "learning_rate": 9.756607971041816e-06, "loss": 0.8341, "step": 7185 }, { "epoch": 0.49, "learning_rate": 9.755665321820444e-06, "loss": 0.6313, "step": 7190 }, { "epoch": 0.49, "learning_rate": 9.754722672599073e-06, "loss": 0.7072, "step": 7195 }, { "epoch": 0.49, "learning_rate": 9.753780023377701e-06, "loss": 0.2242, "step": 7200 }, { "epoch": 0.49, "learning_rate": 9.75283737415633e-06, "loss": 1.0796, "step": 7205 }, { "epoch": 0.49, "learning_rate": 9.751894724934958e-06, "loss": 1.103, "step": 7210 }, { "epoch": 0.49, "learning_rate": 9.750952075713587e-06, "loss": 0.7751, "step": 7215 }, { "epoch": 0.49, "learning_rate": 9.750009426492213e-06, "loss": 0.3469, "step": 7220 }, { "epoch": 0.49, "learning_rate": 9.749066777270843e-06, "loss": 0.6355, "step": 7225 }, { "epoch": 0.49, "learning_rate": 9.74812412804947e-06, "loss": 1.406, "step": 7230 }, { "epoch": 0.49, "learning_rate": 9.7471814788281e-06, "loss": 0.2664, "step": 7235 }, { "epoch": 0.49, "learning_rate": 9.746238829606727e-06, "loss": 0.7651, "step": 7240 }, { "epoch": 0.49, "learning_rate": 9.745296180385357e-06, "loss": 0.4736, "step": 7245 }, { "epoch": 0.49, "learning_rate": 9.744353531163984e-06, "loss": 0.8595, "step": 7250 }, { "epoch": 0.49, "learning_rate": 9.743410881942612e-06, "loss": 1.1605, "step": 7255 }, { "epoch": 0.49, "learning_rate": 9.742468232721241e-06, "loss": 1.0379, "step": 7260 }, { "epoch": 0.49, "learning_rate": 9.741525583499869e-06, "loss": 0.7088, "step": 7265 }, { "epoch": 0.49, "learning_rate": 9.740582934278498e-06, "loss": 0.6329, "step": 7270 }, { "epoch": 0.49, "learning_rate": 9.739640285057126e-06, "loss": 0.8322, "step": 7275 }, { "epoch": 0.49, "learning_rate": 9.738697635835753e-06, "loss": 0.3529, "step": 7280 }, { "epoch": 0.49, "learning_rate": 9.737754986614381e-06, "loss": 0.6055, "step": 7285 }, { "epoch": 0.49, "learning_rate": 9.73681233739301e-06, "loss": 1.2936, "step": 7290 }, { "epoch": 0.5, "learning_rate": 9.735869688171638e-06, "loss": 1.7163, "step": 7295 }, { "epoch": 0.5, "learning_rate": 9.734927038950267e-06, "loss": 0.66, "step": 7300 }, { "epoch": 0.5, "learning_rate": 9.733984389728895e-06, "loss": 1.1418, "step": 7305 }, { "epoch": 0.5, "learning_rate": 9.733041740507523e-06, "loss": 0.8489, "step": 7310 }, { "epoch": 0.5, "learning_rate": 9.732099091286152e-06, "loss": 0.4426, "step": 7315 }, { "epoch": 0.5, "learning_rate": 9.73115644206478e-06, "loss": 1.0069, "step": 7320 }, { "epoch": 0.5, "learning_rate": 9.730213792843409e-06, "loss": 0.3099, "step": 7325 }, { "epoch": 0.5, "learning_rate": 9.729271143622037e-06, "loss": 0.3766, "step": 7330 }, { "epoch": 0.5, "learning_rate": 9.728328494400664e-06, "loss": 0.4578, "step": 7335 }, { "epoch": 0.5, "learning_rate": 9.727385845179292e-06, "loss": 0.238, "step": 7340 }, { "epoch": 0.5, "learning_rate": 9.726443195957921e-06, "loss": 0.3499, "step": 7345 }, { "epoch": 0.5, "learning_rate": 9.725500546736549e-06, "loss": 0.9894, "step": 7350 }, { "epoch": 0.5, "learning_rate": 9.724557897515178e-06, "loss": 0.6388, "step": 7355 }, { "epoch": 0.5, "learning_rate": 9.723615248293806e-06, "loss": 0.6721, "step": 7360 }, { "epoch": 0.5, "learning_rate": 9.722672599072434e-06, "loss": 0.3964, "step": 7365 }, { "epoch": 0.5, "learning_rate": 9.721729949851061e-06, "loss": 0.3763, "step": 7370 }, { "epoch": 0.5, "learning_rate": 9.72078730062969e-06, "loss": 0.5021, "step": 7375 }, { "epoch": 0.5, "learning_rate": 9.719844651408318e-06, "loss": 0.8643, "step": 7380 }, { "epoch": 0.5, "learning_rate": 9.718902002186947e-06, "loss": 1.1715, "step": 7385 }, { "epoch": 0.5, "learning_rate": 9.717959352965575e-06, "loss": 0.9817, "step": 7390 }, { "epoch": 0.5, "learning_rate": 9.717016703744203e-06, "loss": 0.898, "step": 7395 }, { "epoch": 0.5, "learning_rate": 9.716074054522832e-06, "loss": 0.7389, "step": 7400 }, { "epoch": 0.5, "learning_rate": 9.71513140530146e-06, "loss": 0.7768, "step": 7405 }, { "epoch": 0.5, "learning_rate": 9.714188756080089e-06, "loss": 1.4694, "step": 7410 }, { "epoch": 0.5, "learning_rate": 9.713246106858717e-06, "loss": 0.9666, "step": 7415 }, { "epoch": 0.5, "learning_rate": 9.712303457637344e-06, "loss": 0.7048, "step": 7420 }, { "epoch": 0.5, "learning_rate": 9.711360808415972e-06, "loss": 1.5417, "step": 7425 }, { "epoch": 0.5, "learning_rate": 9.710418159194601e-06, "loss": 1.074, "step": 7430 }, { "epoch": 0.5, "learning_rate": 9.709475509973229e-06, "loss": 1.0551, "step": 7435 }, { "epoch": 0.5, "learning_rate": 9.708532860751858e-06, "loss": 0.6274, "step": 7440 }, { "epoch": 0.51, "learning_rate": 9.707590211530486e-06, "loss": 0.5223, "step": 7445 }, { "epoch": 0.51, "learning_rate": 9.706647562309114e-06, "loss": 0.714, "step": 7450 }, { "epoch": 0.51, "learning_rate": 9.705704913087743e-06, "loss": 0.9301, "step": 7455 }, { "epoch": 0.51, "learning_rate": 9.70476226386637e-06, "loss": 0.6944, "step": 7460 }, { "epoch": 0.51, "learning_rate": 9.703819614645e-06, "loss": 0.5436, "step": 7465 }, { "epoch": 0.51, "learning_rate": 9.702876965423628e-06, "loss": 0.516, "step": 7470 }, { "epoch": 0.51, "learning_rate": 9.701934316202257e-06, "loss": 0.314, "step": 7475 }, { "epoch": 0.51, "learning_rate": 9.700991666980883e-06, "loss": 0.563, "step": 7480 }, { "epoch": 0.51, "learning_rate": 9.700049017759512e-06, "loss": 0.6551, "step": 7485 }, { "epoch": 0.51, "learning_rate": 9.69910636853814e-06, "loss": 0.5574, "step": 7490 }, { "epoch": 0.51, "learning_rate": 9.698163719316769e-06, "loss": 0.4433, "step": 7495 }, { "epoch": 0.51, "learning_rate": 9.697221070095397e-06, "loss": 1.1968, "step": 7500 }, { "epoch": 0.51, "learning_rate": 9.696278420874024e-06, "loss": 1.946, "step": 7505 }, { "epoch": 0.51, "learning_rate": 9.695335771652654e-06, "loss": 1.1762, "step": 7510 }, { "epoch": 0.51, "learning_rate": 9.694393122431281e-06, "loss": 0.8282, "step": 7515 }, { "epoch": 0.51, "learning_rate": 9.69345047320991e-06, "loss": 0.7371, "step": 7520 }, { "epoch": 0.51, "learning_rate": 9.692507823988538e-06, "loss": 0.8885, "step": 7525 }, { "epoch": 0.51, "learning_rate": 9.691565174767168e-06, "loss": 1.3681, "step": 7530 }, { "epoch": 0.51, "learning_rate": 9.690622525545794e-06, "loss": 0.787, "step": 7535 }, { "epoch": 0.51, "learning_rate": 9.689679876324423e-06, "loss": 0.7462, "step": 7540 }, { "epoch": 0.51, "learning_rate": 9.68873722710305e-06, "loss": 1.0022, "step": 7545 }, { "epoch": 0.51, "learning_rate": 9.68779457788168e-06, "loss": 0.7136, "step": 7550 }, { "epoch": 0.51, "learning_rate": 9.686851928660308e-06, "loss": 0.8358, "step": 7555 }, { "epoch": 0.51, "learning_rate": 9.685909279438937e-06, "loss": 1.0002, "step": 7560 }, { "epoch": 0.51, "learning_rate": 9.684966630217565e-06, "loss": 0.6543, "step": 7565 }, { "epoch": 0.51, "learning_rate": 9.684023980996192e-06, "loss": 0.9131, "step": 7570 }, { "epoch": 0.51, "learning_rate": 9.683081331774822e-06, "loss": 0.8808, "step": 7575 }, { "epoch": 0.51, "learning_rate": 9.682138682553449e-06, "loss": 0.9327, "step": 7580 }, { "epoch": 0.51, "learning_rate": 9.681196033332077e-06, "loss": 0.7643, "step": 7585 }, { "epoch": 0.52, "learning_rate": 9.680253384110706e-06, "loss": 1.0122, "step": 7590 }, { "epoch": 0.52, "learning_rate": 9.679310734889334e-06, "loss": 0.5702, "step": 7595 }, { "epoch": 0.52, "learning_rate": 9.678368085667961e-06, "loss": 0.6138, "step": 7600 }, { "epoch": 0.52, "learning_rate": 9.67742543644659e-06, "loss": 0.8666, "step": 7605 }, { "epoch": 0.52, "learning_rate": 9.676482787225218e-06, "loss": 0.8854, "step": 7610 }, { "epoch": 0.52, "learning_rate": 9.675540138003848e-06, "loss": 2.1837, "step": 7615 }, { "epoch": 0.52, "learning_rate": 9.674597488782475e-06, "loss": 0.2919, "step": 7620 }, { "epoch": 0.52, "learning_rate": 9.673654839561103e-06, "loss": 1.1508, "step": 7625 }, { "epoch": 0.52, "learning_rate": 9.67271219033973e-06, "loss": 1.1709, "step": 7630 }, { "epoch": 0.52, "learning_rate": 9.67176954111836e-06, "loss": 1.0398, "step": 7635 }, { "epoch": 0.52, "learning_rate": 9.670826891896988e-06, "loss": 0.5613, "step": 7640 }, { "epoch": 0.52, "learning_rate": 9.669884242675617e-06, "loss": 1.0232, "step": 7645 }, { "epoch": 0.52, "learning_rate": 9.668941593454245e-06, "loss": 0.8415, "step": 7650 }, { "epoch": 0.52, "learning_rate": 9.667998944232872e-06, "loss": 1.3732, "step": 7655 }, { "epoch": 0.52, "learning_rate": 9.667056295011502e-06, "loss": 0.6431, "step": 7660 }, { "epoch": 0.52, "learning_rate": 9.66611364579013e-06, "loss": 1.2167, "step": 7665 }, { "epoch": 0.52, "learning_rate": 9.665170996568759e-06, "loss": 1.127, "step": 7670 }, { "epoch": 0.52, "learning_rate": 9.664228347347386e-06, "loss": 0.9503, "step": 7675 }, { "epoch": 0.52, "learning_rate": 9.663285698126014e-06, "loss": 0.6425, "step": 7680 }, { "epoch": 0.52, "learning_rate": 9.662343048904641e-06, "loss": 0.9962, "step": 7685 }, { "epoch": 0.52, "learning_rate": 9.66140039968327e-06, "loss": 0.9098, "step": 7690 }, { "epoch": 0.52, "learning_rate": 9.660457750461898e-06, "loss": 0.6096, "step": 7695 }, { "epoch": 0.52, "learning_rate": 9.659515101240528e-06, "loss": 0.3893, "step": 7700 }, { "epoch": 0.52, "learning_rate": 9.658572452019155e-06, "loss": 1.1583, "step": 7705 }, { "epoch": 0.52, "learning_rate": 9.657629802797783e-06, "loss": 0.793, "step": 7710 }, { "epoch": 0.52, "learning_rate": 9.656687153576412e-06, "loss": 0.6989, "step": 7715 }, { "epoch": 0.52, "learning_rate": 9.65574450435504e-06, "loss": 1.1703, "step": 7720 }, { "epoch": 0.52, "learning_rate": 9.65480185513367e-06, "loss": 0.9384, "step": 7725 }, { "epoch": 0.52, "learning_rate": 9.653859205912297e-06, "loss": 0.7218, "step": 7730 }, { "epoch": 0.52, "learning_rate": 9.652916556690925e-06, "loss": 0.5897, "step": 7735 }, { "epoch": 0.53, "learning_rate": 9.651973907469552e-06, "loss": 0.8512, "step": 7740 }, { "epoch": 0.53, "learning_rate": 9.651031258248182e-06, "loss": 0.9472, "step": 7745 }, { "epoch": 0.53, "learning_rate": 9.65008860902681e-06, "loss": 0.8709, "step": 7750 }, { "epoch": 0.53, "learning_rate": 9.649145959805439e-06, "loss": 1.1843, "step": 7755 }, { "epoch": 0.53, "learning_rate": 9.648203310584066e-06, "loss": 0.9818, "step": 7760 }, { "epoch": 0.53, "learning_rate": 9.647260661362694e-06, "loss": 0.6926, "step": 7765 }, { "epoch": 0.53, "learning_rate": 9.646318012141323e-06, "loss": 0.7125, "step": 7770 }, { "epoch": 0.53, "learning_rate": 9.64537536291995e-06, "loss": 0.7874, "step": 7775 }, { "epoch": 0.53, "learning_rate": 9.64443271369858e-06, "loss": 0.6357, "step": 7780 }, { "epoch": 0.53, "learning_rate": 9.643490064477208e-06, "loss": 0.5344, "step": 7785 }, { "epoch": 0.53, "learning_rate": 9.642547415255835e-06, "loss": 0.4024, "step": 7790 }, { "epoch": 0.53, "learning_rate": 9.641604766034463e-06, "loss": 0.3847, "step": 7795 }, { "epoch": 0.53, "learning_rate": 9.640662116813092e-06, "loss": 0.4975, "step": 7800 }, { "epoch": 0.53, "learning_rate": 9.63971946759172e-06, "loss": 1.0251, "step": 7805 }, { "epoch": 0.53, "learning_rate": 9.63877681837035e-06, "loss": 1.101, "step": 7810 }, { "epoch": 0.53, "learning_rate": 9.637834169148977e-06, "loss": 1.2719, "step": 7815 }, { "epoch": 0.53, "learning_rate": 9.636891519927606e-06, "loss": 0.5741, "step": 7820 }, { "epoch": 0.53, "learning_rate": 9.635948870706234e-06, "loss": 1.0066, "step": 7825 }, { "epoch": 0.53, "learning_rate": 9.635006221484862e-06, "loss": 0.8564, "step": 7830 }, { "epoch": 0.53, "learning_rate": 9.63406357226349e-06, "loss": 1.1401, "step": 7835 }, { "epoch": 0.53, "learning_rate": 9.633120923042119e-06, "loss": 0.3237, "step": 7840 }, { "epoch": 0.53, "learning_rate": 9.632178273820746e-06, "loss": 0.2197, "step": 7845 }, { "epoch": 0.53, "learning_rate": 9.631235624599374e-06, "loss": 1.0384, "step": 7850 }, { "epoch": 0.53, "learning_rate": 9.630292975378003e-06, "loss": 0.6831, "step": 7855 }, { "epoch": 0.53, "learning_rate": 9.629350326156631e-06, "loss": 0.2769, "step": 7860 }, { "epoch": 0.53, "learning_rate": 9.62840767693526e-06, "loss": 0.9527, "step": 7865 }, { "epoch": 0.53, "learning_rate": 9.627465027713888e-06, "loss": 0.5204, "step": 7870 }, { "epoch": 0.53, "learning_rate": 9.626522378492517e-06, "loss": 0.6507, "step": 7875 }, { "epoch": 0.53, "learning_rate": 9.625579729271143e-06, "loss": 0.3867, "step": 7880 }, { "epoch": 0.54, "learning_rate": 9.624637080049772e-06, "loss": 0.1681, "step": 7885 }, { "epoch": 0.54, "learning_rate": 9.6236944308284e-06, "loss": 0.5787, "step": 7890 }, { "epoch": 0.54, "learning_rate": 9.62275178160703e-06, "loss": 0.5972, "step": 7895 }, { "epoch": 0.54, "learning_rate": 9.621809132385657e-06, "loss": 0.4133, "step": 7900 }, { "epoch": 0.54, "learning_rate": 9.620866483164286e-06, "loss": 0.8102, "step": 7905 }, { "epoch": 0.54, "learning_rate": 9.619923833942914e-06, "loss": 1.5462, "step": 7910 }, { "epoch": 0.54, "learning_rate": 9.618981184721542e-06, "loss": 0.5578, "step": 7915 }, { "epoch": 0.54, "learning_rate": 9.618038535500171e-06, "loss": 0.6278, "step": 7920 }, { "epoch": 0.54, "learning_rate": 9.617095886278799e-06, "loss": 0.8735, "step": 7925 }, { "epoch": 0.54, "learning_rate": 9.616153237057428e-06, "loss": 0.5903, "step": 7930 }, { "epoch": 0.54, "learning_rate": 9.615210587836056e-06, "loss": 0.4325, "step": 7935 }, { "epoch": 0.54, "learning_rate": 9.614267938614683e-06, "loss": 0.9277, "step": 7940 }, { "epoch": 0.54, "learning_rate": 9.613325289393311e-06, "loss": 0.9635, "step": 7945 }, { "epoch": 0.54, "learning_rate": 9.61238264017194e-06, "loss": 0.7308, "step": 7950 }, { "epoch": 0.54, "learning_rate": 9.611439990950568e-06, "loss": 1.2898, "step": 7955 }, { "epoch": 0.54, "learning_rate": 9.610497341729197e-06, "loss": 0.637, "step": 7960 }, { "epoch": 0.54, "learning_rate": 9.609554692507825e-06, "loss": 1.278, "step": 7965 }, { "epoch": 0.54, "learning_rate": 9.608612043286453e-06, "loss": 0.7364, "step": 7970 }, { "epoch": 0.54, "learning_rate": 9.607669394065082e-06, "loss": 0.2788, "step": 7975 }, { "epoch": 0.54, "learning_rate": 9.60672674484371e-06, "loss": 1.1307, "step": 7980 }, { "epoch": 0.54, "learning_rate": 9.605784095622339e-06, "loss": 0.8766, "step": 7985 }, { "epoch": 0.54, "learning_rate": 9.604841446400966e-06, "loss": 0.6887, "step": 7990 }, { "epoch": 0.54, "learning_rate": 9.603898797179594e-06, "loss": 1.2531, "step": 7995 }, { "epoch": 0.54, "learning_rate": 9.602956147958222e-06, "loss": 0.522, "step": 8000 }, { "epoch": 0.54, "learning_rate": 9.602013498736851e-06, "loss": 0.7283, "step": 8005 }, { "epoch": 0.54, "learning_rate": 9.601070849515479e-06, "loss": 0.5963, "step": 8010 }, { "epoch": 0.54, "learning_rate": 9.600128200294108e-06, "loss": 0.7722, "step": 8015 }, { "epoch": 0.54, "learning_rate": 9.599185551072736e-06, "loss": 0.8785, "step": 8020 }, { "epoch": 0.54, "learning_rate": 9.598242901851363e-06, "loss": 0.6146, "step": 8025 }, { "epoch": 0.54, "learning_rate": 9.597300252629993e-06, "loss": 0.4031, "step": 8030 }, { "epoch": 0.55, "learning_rate": 9.59635760340862e-06, "loss": 0.7064, "step": 8035 }, { "epoch": 0.55, "learning_rate": 9.595414954187248e-06, "loss": 0.8904, "step": 8040 }, { "epoch": 0.55, "learning_rate": 9.594472304965877e-06, "loss": 1.0205, "step": 8045 }, { "epoch": 0.55, "learning_rate": 9.593529655744505e-06, "loss": 0.7206, "step": 8050 }, { "epoch": 0.55, "learning_rate": 9.592587006523133e-06, "loss": 0.3653, "step": 8055 }, { "epoch": 0.55, "learning_rate": 9.591644357301762e-06, "loss": 0.6096, "step": 8060 }, { "epoch": 0.55, "learning_rate": 9.59070170808039e-06, "loss": 0.4454, "step": 8065 }, { "epoch": 0.55, "learning_rate": 9.589759058859019e-06, "loss": 0.8194, "step": 8070 }, { "epoch": 0.55, "learning_rate": 9.588816409637647e-06, "loss": 0.6394, "step": 8075 }, { "epoch": 0.55, "learning_rate": 9.587873760416274e-06, "loss": 0.35, "step": 8080 }, { "epoch": 0.55, "learning_rate": 9.586931111194902e-06, "loss": 1.0713, "step": 8085 }, { "epoch": 0.55, "learning_rate": 9.585988461973531e-06, "loss": 0.7802, "step": 8090 }, { "epoch": 0.55, "learning_rate": 9.585045812752159e-06, "loss": 0.4639, "step": 8095 }, { "epoch": 0.55, "learning_rate": 9.584103163530788e-06, "loss": 0.7082, "step": 8100 }, { "epoch": 0.55, "learning_rate": 9.583160514309416e-06, "loss": 0.2567, "step": 8105 }, { "epoch": 0.55, "learning_rate": 9.582217865088043e-06, "loss": 1.0491, "step": 8110 }, { "epoch": 0.55, "learning_rate": 9.581275215866673e-06, "loss": 0.9032, "step": 8115 }, { "epoch": 0.55, "learning_rate": 9.5803325666453e-06, "loss": 0.6374, "step": 8120 }, { "epoch": 0.55, "learning_rate": 9.57938991742393e-06, "loss": 0.4996, "step": 8125 }, { "epoch": 0.55, "learning_rate": 9.578447268202557e-06, "loss": 0.7772, "step": 8130 }, { "epoch": 0.55, "learning_rate": 9.577504618981187e-06, "loss": 1.556, "step": 8135 }, { "epoch": 0.55, "learning_rate": 9.576561969759813e-06, "loss": 0.8642, "step": 8140 }, { "epoch": 0.55, "learning_rate": 9.575619320538442e-06, "loss": 0.5141, "step": 8145 }, { "epoch": 0.55, "learning_rate": 9.57467667131707e-06, "loss": 0.7785, "step": 8150 }, { "epoch": 0.55, "learning_rate": 9.573734022095699e-06, "loss": 0.6164, "step": 8155 }, { "epoch": 0.55, "learning_rate": 9.572791372874327e-06, "loss": 0.8964, "step": 8160 }, { "epoch": 0.55, "learning_rate": 9.571848723652956e-06, "loss": 0.6809, "step": 8165 }, { "epoch": 0.55, "learning_rate": 9.570906074431584e-06, "loss": 1.0431, "step": 8170 }, { "epoch": 0.55, "learning_rate": 9.569963425210211e-06, "loss": 0.6417, "step": 8175 }, { "epoch": 0.56, "learning_rate": 9.56902077598884e-06, "loss": 0.7418, "step": 8180 }, { "epoch": 0.56, "learning_rate": 9.568078126767468e-06, "loss": 0.3866, "step": 8185 }, { "epoch": 0.56, "learning_rate": 9.567135477546097e-06, "loss": 0.2988, "step": 8190 }, { "epoch": 0.56, "learning_rate": 9.566192828324723e-06, "loss": 1.1566, "step": 8195 }, { "epoch": 0.56, "learning_rate": 9.565250179103353e-06, "loss": 0.3677, "step": 8200 }, { "epoch": 0.56, "learning_rate": 9.56430752988198e-06, "loss": 0.5658, "step": 8205 }, { "epoch": 0.56, "learning_rate": 9.56336488066061e-06, "loss": 0.968, "step": 8210 }, { "epoch": 0.56, "learning_rate": 9.562422231439237e-06, "loss": 0.655, "step": 8215 }, { "epoch": 0.56, "learning_rate": 9.561479582217867e-06, "loss": 0.6444, "step": 8220 }, { "epoch": 0.56, "learning_rate": 9.560536932996494e-06, "loss": 0.7406, "step": 8225 }, { "epoch": 0.56, "learning_rate": 9.559594283775122e-06, "loss": 0.5359, "step": 8230 }, { "epoch": 0.56, "learning_rate": 9.558651634553751e-06, "loss": 0.5425, "step": 8235 }, { "epoch": 0.56, "learning_rate": 9.557708985332379e-06, "loss": 0.5257, "step": 8240 }, { "epoch": 0.56, "learning_rate": 9.556766336111007e-06, "loss": 0.9725, "step": 8245 }, { "epoch": 0.56, "learning_rate": 9.555823686889636e-06, "loss": 0.539, "step": 8250 }, { "epoch": 0.56, "learning_rate": 9.554881037668264e-06, "loss": 0.5204, "step": 8255 }, { "epoch": 0.56, "learning_rate": 9.553938388446891e-06, "loss": 0.7691, "step": 8260 }, { "epoch": 0.56, "learning_rate": 9.55299573922552e-06, "loss": 0.9838, "step": 8265 }, { "epoch": 0.56, "learning_rate": 9.552053090004148e-06, "loss": 0.3412, "step": 8270 }, { "epoch": 0.56, "learning_rate": 9.551110440782778e-06, "loss": 0.8682, "step": 8275 }, { "epoch": 0.56, "learning_rate": 9.550167791561405e-06, "loss": 0.7232, "step": 8280 }, { "epoch": 0.56, "learning_rate": 9.549225142340033e-06, "loss": 0.6043, "step": 8285 }, { "epoch": 0.56, "learning_rate": 9.54828249311866e-06, "loss": 0.6277, "step": 8290 }, { "epoch": 0.56, "learning_rate": 9.54733984389729e-06, "loss": 0.7065, "step": 8295 }, { "epoch": 0.56, "learning_rate": 9.546397194675917e-06, "loss": 0.4923, "step": 8300 }, { "epoch": 0.56, "learning_rate": 9.545454545454547e-06, "loss": 0.4866, "step": 8305 }, { "epoch": 0.56, "learning_rate": 9.544511896233174e-06, "loss": 1.1497, "step": 8310 }, { "epoch": 0.56, "learning_rate": 9.543569247011802e-06, "loss": 0.2948, "step": 8315 }, { "epoch": 0.56, "learning_rate": 9.542626597790431e-06, "loss": 1.0316, "step": 8320 }, { "epoch": 0.57, "learning_rate": 9.541683948569059e-06, "loss": 0.4104, "step": 8325 }, { "epoch": 0.57, "learning_rate": 9.540741299347688e-06, "loss": 0.5597, "step": 8330 }, { "epoch": 0.57, "learning_rate": 9.539798650126316e-06, "loss": 0.8399, "step": 8335 }, { "epoch": 0.57, "learning_rate": 9.538856000904944e-06, "loss": 0.448, "step": 8340 }, { "epoch": 0.57, "learning_rate": 9.537913351683571e-06, "loss": 0.7359, "step": 8345 }, { "epoch": 0.57, "learning_rate": 9.5369707024622e-06, "loss": 1.335, "step": 8350 }, { "epoch": 0.57, "learning_rate": 9.536028053240828e-06, "loss": 0.5481, "step": 8355 }, { "epoch": 0.57, "learning_rate": 9.535085404019458e-06, "loss": 1.1661, "step": 8360 }, { "epoch": 0.57, "learning_rate": 9.534142754798085e-06, "loss": 0.8165, "step": 8365 }, { "epoch": 0.57, "learning_rate": 9.533200105576713e-06, "loss": 1.0099, "step": 8370 }, { "epoch": 0.57, "learning_rate": 9.532257456355342e-06, "loss": 1.4125, "step": 8375 }, { "epoch": 0.57, "learning_rate": 9.53131480713397e-06, "loss": 0.7941, "step": 8380 }, { "epoch": 0.57, "learning_rate": 9.5303721579126e-06, "loss": 0.354, "step": 8385 }, { "epoch": 0.57, "learning_rate": 9.529429508691227e-06, "loss": 1.6591, "step": 8390 }, { "epoch": 0.57, "learning_rate": 9.528486859469856e-06, "loss": 0.9348, "step": 8395 }, { "epoch": 0.57, "learning_rate": 9.527544210248482e-06, "loss": 1.3696, "step": 8400 }, { "epoch": 0.57, "learning_rate": 9.526601561027111e-06, "loss": 0.6246, "step": 8405 }, { "epoch": 0.57, "learning_rate": 9.525658911805739e-06, "loss": 0.4945, "step": 8410 }, { "epoch": 0.57, "learning_rate": 9.524716262584368e-06, "loss": 1.0697, "step": 8415 }, { "epoch": 0.57, "learning_rate": 9.523773613362996e-06, "loss": 0.2341, "step": 8420 }, { "epoch": 0.57, "learning_rate": 9.522830964141624e-06, "loss": 0.7271, "step": 8425 }, { "epoch": 0.57, "learning_rate": 9.521888314920253e-06, "loss": 0.7074, "step": 8430 }, { "epoch": 0.57, "learning_rate": 9.52094566569888e-06, "loss": 0.4784, "step": 8435 }, { "epoch": 0.57, "learning_rate": 9.52000301647751e-06, "loss": 1.0385, "step": 8440 }, { "epoch": 0.57, "learning_rate": 9.519060367256138e-06, "loss": 0.6117, "step": 8445 }, { "epoch": 0.57, "learning_rate": 9.518117718034767e-06, "loss": 0.8461, "step": 8450 }, { "epoch": 0.57, "learning_rate": 9.517175068813393e-06, "loss": 0.2352, "step": 8455 }, { "epoch": 0.57, "learning_rate": 9.516232419592022e-06, "loss": 0.5398, "step": 8460 }, { "epoch": 0.57, "learning_rate": 9.51528977037065e-06, "loss": 0.6751, "step": 8465 }, { "epoch": 0.57, "learning_rate": 9.51434712114928e-06, "loss": 0.7143, "step": 8470 }, { "epoch": 0.58, "learning_rate": 9.513404471927907e-06, "loss": 0.8898, "step": 8475 }, { "epoch": 0.58, "learning_rate": 9.512461822706536e-06, "loss": 0.5872, "step": 8480 }, { "epoch": 0.58, "learning_rate": 9.511519173485164e-06, "loss": 1.2499, "step": 8485 }, { "epoch": 0.58, "learning_rate": 9.510576524263791e-06, "loss": 0.4065, "step": 8490 }, { "epoch": 0.58, "learning_rate": 9.509633875042419e-06, "loss": 0.4466, "step": 8495 }, { "epoch": 0.58, "learning_rate": 9.508691225821048e-06, "loss": 0.3468, "step": 8500 }, { "epoch": 0.58, "learning_rate": 9.507748576599676e-06, "loss": 0.8698, "step": 8505 }, { "epoch": 0.58, "learning_rate": 9.506805927378304e-06, "loss": 0.7624, "step": 8510 }, { "epoch": 0.58, "learning_rate": 9.505863278156933e-06, "loss": 1.0432, "step": 8515 }, { "epoch": 0.58, "learning_rate": 9.50492062893556e-06, "loss": 0.4361, "step": 8520 }, { "epoch": 0.58, "learning_rate": 9.50397797971419e-06, "loss": 1.3389, "step": 8525 }, { "epoch": 0.58, "learning_rate": 9.503035330492818e-06, "loss": 0.781, "step": 8530 }, { "epoch": 0.58, "learning_rate": 9.502092681271447e-06, "loss": 0.7357, "step": 8535 }, { "epoch": 0.58, "learning_rate": 9.501150032050073e-06, "loss": 0.5755, "step": 8540 }, { "epoch": 0.58, "learning_rate": 9.500207382828702e-06, "loss": 0.7979, "step": 8545 }, { "epoch": 0.58, "learning_rate": 9.49926473360733e-06, "loss": 1.5733, "step": 8550 }, { "epoch": 0.58, "learning_rate": 9.49832208438596e-06, "loss": 1.5157, "step": 8555 }, { "epoch": 0.58, "learning_rate": 9.497379435164587e-06, "loss": 0.5594, "step": 8560 }, { "epoch": 0.58, "learning_rate": 9.496436785943216e-06, "loss": 0.808, "step": 8565 }, { "epoch": 0.58, "learning_rate": 9.495494136721844e-06, "loss": 0.7213, "step": 8570 }, { "epoch": 0.58, "learning_rate": 9.494551487500472e-06, "loss": 0.8077, "step": 8575 }, { "epoch": 0.58, "learning_rate": 9.493608838279101e-06, "loss": 0.624, "step": 8580 }, { "epoch": 0.58, "learning_rate": 9.492666189057729e-06, "loss": 0.5865, "step": 8585 }, { "epoch": 0.58, "learning_rate": 9.491723539836358e-06, "loss": 0.7451, "step": 8590 }, { "epoch": 0.58, "learning_rate": 9.490780890614985e-06, "loss": 0.6105, "step": 8595 }, { "epoch": 0.58, "learning_rate": 9.489838241393613e-06, "loss": 0.8022, "step": 8600 }, { "epoch": 0.58, "learning_rate": 9.48889559217224e-06, "loss": 0.781, "step": 8605 }, { "epoch": 0.58, "learning_rate": 9.48795294295087e-06, "loss": 0.5778, "step": 8610 }, { "epoch": 0.58, "learning_rate": 9.487010293729498e-06, "loss": 0.9254, "step": 8615 }, { "epoch": 0.59, "learning_rate": 9.486067644508127e-06, "loss": 0.9518, "step": 8620 }, { "epoch": 0.59, "learning_rate": 9.485124995286755e-06, "loss": 1.2889, "step": 8625 }, { "epoch": 0.59, "learning_rate": 9.484182346065382e-06, "loss": 0.3521, "step": 8630 }, { "epoch": 0.59, "learning_rate": 9.483239696844012e-06, "loss": 0.8582, "step": 8635 }, { "epoch": 0.59, "learning_rate": 9.48229704762264e-06, "loss": 1.1483, "step": 8640 }, { "epoch": 0.59, "learning_rate": 9.481354398401269e-06, "loss": 0.998, "step": 8645 }, { "epoch": 0.59, "learning_rate": 9.480411749179896e-06, "loss": 1.0128, "step": 8650 }, { "epoch": 0.59, "learning_rate": 9.479469099958524e-06, "loss": 0.7746, "step": 8655 }, { "epoch": 0.59, "learning_rate": 9.478526450737152e-06, "loss": 0.4254, "step": 8660 }, { "epoch": 0.59, "learning_rate": 9.477583801515781e-06, "loss": 0.6051, "step": 8665 }, { "epoch": 0.59, "learning_rate": 9.476641152294409e-06, "loss": 0.5037, "step": 8670 }, { "epoch": 0.59, "learning_rate": 9.475698503073038e-06, "loss": 0.871, "step": 8675 }, { "epoch": 0.59, "learning_rate": 9.474755853851666e-06, "loss": 0.6774, "step": 8680 }, { "epoch": 0.59, "learning_rate": 9.473813204630293e-06, "loss": 0.5852, "step": 8685 }, { "epoch": 0.59, "learning_rate": 9.472870555408923e-06, "loss": 1.3319, "step": 8690 }, { "epoch": 0.59, "learning_rate": 9.47192790618755e-06, "loss": 0.6744, "step": 8695 }, { "epoch": 0.59, "learning_rate": 9.47098525696618e-06, "loss": 1.0484, "step": 8700 }, { "epoch": 0.59, "learning_rate": 9.470042607744807e-06, "loss": 0.6731, "step": 8705 }, { "epoch": 0.59, "learning_rate": 9.469099958523435e-06, "loss": 1.9003, "step": 8710 }, { "epoch": 0.59, "learning_rate": 9.468157309302062e-06, "loss": 0.7518, "step": 8715 }, { "epoch": 0.59, "learning_rate": 9.467214660080692e-06, "loss": 0.7239, "step": 8720 }, { "epoch": 0.59, "learning_rate": 9.46627201085932e-06, "loss": 0.6714, "step": 8725 }, { "epoch": 0.59, "learning_rate": 9.465329361637949e-06, "loss": 0.824, "step": 8730 }, { "epoch": 0.59, "learning_rate": 9.464386712416576e-06, "loss": 0.7103, "step": 8735 }, { "epoch": 0.59, "learning_rate": 9.463444063195204e-06, "loss": 1.3432, "step": 8740 }, { "epoch": 0.59, "learning_rate": 9.462501413973833e-06, "loss": 1.0145, "step": 8745 }, { "epoch": 0.59, "learning_rate": 9.461558764752461e-06, "loss": 0.7046, "step": 8750 }, { "epoch": 0.59, "learning_rate": 9.460616115531089e-06, "loss": 0.9844, "step": 8755 }, { "epoch": 0.59, "learning_rate": 9.459673466309718e-06, "loss": 0.594, "step": 8760 }, { "epoch": 0.59, "learning_rate": 9.458730817088346e-06, "loss": 0.4107, "step": 8765 }, { "epoch": 0.6, "learning_rate": 9.457788167866973e-06, "loss": 0.8204, "step": 8770 }, { "epoch": 0.6, "learning_rate": 9.456845518645603e-06, "loss": 0.5973, "step": 8775 }, { "epoch": 0.6, "learning_rate": 9.45590286942423e-06, "loss": 0.7489, "step": 8780 }, { "epoch": 0.6, "learning_rate": 9.45496022020286e-06, "loss": 0.8589, "step": 8785 }, { "epoch": 0.6, "learning_rate": 9.454017570981487e-06, "loss": 0.5674, "step": 8790 }, { "epoch": 0.6, "learning_rate": 9.453074921760117e-06, "loss": 0.2216, "step": 8795 }, { "epoch": 0.6, "learning_rate": 9.452132272538742e-06, "loss": 0.8728, "step": 8800 }, { "epoch": 0.6, "learning_rate": 9.451189623317372e-06, "loss": 0.8135, "step": 8805 }, { "epoch": 0.6, "learning_rate": 9.450246974096e-06, "loss": 1.2375, "step": 8810 }, { "epoch": 0.6, "learning_rate": 9.449304324874629e-06, "loss": 0.3844, "step": 8815 }, { "epoch": 0.6, "learning_rate": 9.448361675653256e-06, "loss": 0.3128, "step": 8820 }, { "epoch": 0.6, "learning_rate": 9.447419026431886e-06, "loss": 1.0977, "step": 8825 }, { "epoch": 0.6, "learning_rate": 9.446476377210513e-06, "loss": 0.9319, "step": 8830 }, { "epoch": 0.6, "learning_rate": 9.445533727989141e-06, "loss": 0.5246, "step": 8835 }, { "epoch": 0.6, "learning_rate": 9.44459107876777e-06, "loss": 0.7486, "step": 8840 }, { "epoch": 0.6, "learning_rate": 9.443648429546398e-06, "loss": 0.8336, "step": 8845 }, { "epoch": 0.6, "learning_rate": 9.442705780325027e-06, "loss": 0.2352, "step": 8850 }, { "epoch": 0.6, "learning_rate": 9.441763131103653e-06, "loss": 0.1619, "step": 8855 }, { "epoch": 0.6, "learning_rate": 9.440820481882283e-06, "loss": 0.5887, "step": 8860 }, { "epoch": 0.6, "learning_rate": 9.43987783266091e-06, "loss": 0.8237, "step": 8865 }, { "epoch": 0.6, "learning_rate": 9.43893518343954e-06, "loss": 0.2864, "step": 8870 }, { "epoch": 0.6, "learning_rate": 9.437992534218167e-06, "loss": 1.1197, "step": 8875 }, { "epoch": 0.6, "learning_rate": 9.437049884996797e-06, "loss": 0.7354, "step": 8880 }, { "epoch": 0.6, "learning_rate": 9.436107235775424e-06, "loss": 0.844, "step": 8885 }, { "epoch": 0.6, "learning_rate": 9.435164586554052e-06, "loss": 1.0101, "step": 8890 }, { "epoch": 0.6, "learning_rate": 9.434221937332681e-06, "loss": 0.8033, "step": 8895 }, { "epoch": 0.6, "learning_rate": 9.433279288111309e-06, "loss": 0.9671, "step": 8900 }, { "epoch": 0.6, "learning_rate": 9.432336638889938e-06, "loss": 0.3929, "step": 8905 }, { "epoch": 0.6, "learning_rate": 9.431393989668566e-06, "loss": 0.7295, "step": 8910 }, { "epoch": 0.61, "learning_rate": 9.430451340447193e-06, "loss": 1.1527, "step": 8915 }, { "epoch": 0.61, "learning_rate": 9.429508691225821e-06, "loss": 1.2564, "step": 8920 }, { "epoch": 0.61, "learning_rate": 9.42856604200445e-06, "loss": 0.4985, "step": 8925 }, { "epoch": 0.61, "learning_rate": 9.427623392783078e-06, "loss": 0.6666, "step": 8930 }, { "epoch": 0.61, "learning_rate": 9.426680743561707e-06, "loss": 0.7432, "step": 8935 }, { "epoch": 0.61, "learning_rate": 9.425738094340335e-06, "loss": 0.432, "step": 8940 }, { "epoch": 0.61, "learning_rate": 9.424795445118963e-06, "loss": 0.8378, "step": 8945 }, { "epoch": 0.61, "learning_rate": 9.423852795897592e-06, "loss": 0.7596, "step": 8950 }, { "epoch": 0.61, "learning_rate": 9.42291014667622e-06, "loss": 0.9971, "step": 8955 }, { "epoch": 0.61, "learning_rate": 9.421967497454847e-06, "loss": 0.6194, "step": 8960 }, { "epoch": 0.61, "learning_rate": 9.421024848233477e-06, "loss": 0.8544, "step": 8965 }, { "epoch": 0.61, "learning_rate": 9.420082199012104e-06, "loss": 0.8496, "step": 8970 }, { "epoch": 0.61, "learning_rate": 9.419139549790732e-06, "loss": 1.0214, "step": 8975 }, { "epoch": 0.61, "learning_rate": 9.418196900569361e-06, "loss": 0.5407, "step": 8980 }, { "epoch": 0.61, "learning_rate": 9.417254251347989e-06, "loss": 0.62, "step": 8985 }, { "epoch": 0.61, "learning_rate": 9.416311602126618e-06, "loss": 0.8147, "step": 8990 }, { "epoch": 0.61, "learning_rate": 9.415368952905246e-06, "loss": 1.458, "step": 8995 }, { "epoch": 0.61, "learning_rate": 9.414426303683873e-06, "loss": 1.246, "step": 9000 }, { "epoch": 0.61, "learning_rate": 9.413483654462501e-06, "loss": 0.3795, "step": 9005 }, { "epoch": 0.61, "learning_rate": 9.41254100524113e-06, "loss": 0.4181, "step": 9010 }, { "epoch": 0.61, "learning_rate": 9.411598356019758e-06, "loss": 1.1659, "step": 9015 }, { "epoch": 0.61, "learning_rate": 9.410655706798387e-06, "loss": 0.8234, "step": 9020 }, { "epoch": 0.61, "learning_rate": 9.409713057577015e-06, "loss": 0.8613, "step": 9025 }, { "epoch": 0.61, "learning_rate": 9.408770408355643e-06, "loss": 0.4, "step": 9030 }, { "epoch": 0.61, "learning_rate": 9.407827759134272e-06, "loss": 2.2914, "step": 9035 }, { "epoch": 0.61, "learning_rate": 9.4068851099129e-06, "loss": 0.7317, "step": 9040 }, { "epoch": 0.61, "learning_rate": 9.405942460691529e-06, "loss": 0.679, "step": 9045 }, { "epoch": 0.61, "learning_rate": 9.404999811470157e-06, "loss": 1.0238, "step": 9050 }, { "epoch": 0.61, "learning_rate": 9.404057162248786e-06, "loss": 0.4876, "step": 9055 }, { "epoch": 0.61, "learning_rate": 9.403114513027412e-06, "loss": 0.3674, "step": 9060 }, { "epoch": 0.62, "learning_rate": 9.402171863806041e-06, "loss": 0.3583, "step": 9065 }, { "epoch": 0.62, "learning_rate": 9.401229214584669e-06, "loss": 1.4944, "step": 9070 }, { "epoch": 0.62, "learning_rate": 9.400286565363298e-06, "loss": 0.6188, "step": 9075 }, { "epoch": 0.62, "learning_rate": 9.399343916141926e-06, "loss": 0.4455, "step": 9080 }, { "epoch": 0.62, "learning_rate": 9.398401266920554e-06, "loss": 0.6602, "step": 9085 }, { "epoch": 0.62, "learning_rate": 9.397458617699183e-06, "loss": 0.8668, "step": 9090 }, { "epoch": 0.62, "learning_rate": 9.39651596847781e-06, "loss": 0.5231, "step": 9095 }, { "epoch": 0.62, "learning_rate": 9.39557331925644e-06, "loss": 0.6494, "step": 9100 }, { "epoch": 0.62, "learning_rate": 9.394630670035067e-06, "loss": 0.3555, "step": 9105 }, { "epoch": 0.62, "learning_rate": 9.393688020813697e-06, "loss": 1.2174, "step": 9110 }, { "epoch": 0.62, "learning_rate": 9.392745371592323e-06, "loss": 0.4135, "step": 9115 }, { "epoch": 0.62, "learning_rate": 9.391802722370952e-06, "loss": 0.5848, "step": 9120 }, { "epoch": 0.62, "learning_rate": 9.39086007314958e-06, "loss": 1.0867, "step": 9125 }, { "epoch": 0.62, "learning_rate": 9.389917423928209e-06, "loss": 1.2763, "step": 9130 }, { "epoch": 0.62, "learning_rate": 9.388974774706837e-06, "loss": 0.5589, "step": 9135 }, { "epoch": 0.62, "learning_rate": 9.388032125485466e-06, "loss": 0.8557, "step": 9140 }, { "epoch": 0.62, "learning_rate": 9.387089476264094e-06, "loss": 0.6947, "step": 9145 }, { "epoch": 0.62, "learning_rate": 9.386146827042721e-06, "loss": 0.7499, "step": 9150 }, { "epoch": 0.62, "learning_rate": 9.38520417782135e-06, "loss": 0.5659, "step": 9155 }, { "epoch": 0.62, "learning_rate": 9.384261528599978e-06, "loss": 0.6573, "step": 9160 }, { "epoch": 0.62, "learning_rate": 9.383318879378606e-06, "loss": 0.269, "step": 9165 }, { "epoch": 0.62, "learning_rate": 9.382376230157235e-06, "loss": 0.9964, "step": 9170 }, { "epoch": 0.62, "learning_rate": 9.381433580935863e-06, "loss": 0.6308, "step": 9175 }, { "epoch": 0.62, "learning_rate": 9.38049093171449e-06, "loss": 1.5251, "step": 9180 }, { "epoch": 0.62, "learning_rate": 9.37954828249312e-06, "loss": 0.8937, "step": 9185 }, { "epoch": 0.62, "learning_rate": 9.378605633271748e-06, "loss": 0.4419, "step": 9190 }, { "epoch": 0.62, "learning_rate": 9.377662984050377e-06, "loss": 0.8119, "step": 9195 }, { "epoch": 0.62, "learning_rate": 9.376720334829004e-06, "loss": 0.477, "step": 9200 }, { "epoch": 0.62, "learning_rate": 9.375777685607632e-06, "loss": 0.8325, "step": 9205 }, { "epoch": 0.63, "learning_rate": 9.37483503638626e-06, "loss": 0.4839, "step": 9210 }, { "epoch": 0.63, "learning_rate": 9.373892387164889e-06, "loss": 0.2054, "step": 9215 }, { "epoch": 0.63, "learning_rate": 9.372949737943517e-06, "loss": 0.7799, "step": 9220 }, { "epoch": 0.63, "learning_rate": 9.372007088722146e-06, "loss": 0.9283, "step": 9225 }, { "epoch": 0.63, "learning_rate": 9.371064439500774e-06, "loss": 0.523, "step": 9230 }, { "epoch": 0.63, "learning_rate": 9.370121790279401e-06, "loss": 0.7192, "step": 9235 }, { "epoch": 0.63, "learning_rate": 9.36917914105803e-06, "loss": 1.3952, "step": 9240 }, { "epoch": 0.63, "learning_rate": 9.368236491836658e-06, "loss": 0.9993, "step": 9245 }, { "epoch": 0.63, "learning_rate": 9.367293842615288e-06, "loss": 0.7293, "step": 9250 }, { "epoch": 0.63, "learning_rate": 9.366351193393915e-06, "loss": 0.6415, "step": 9255 }, { "epoch": 0.63, "learning_rate": 9.365408544172543e-06, "loss": 0.4279, "step": 9260 }, { "epoch": 0.63, "learning_rate": 9.36446589495117e-06, "loss": 0.6285, "step": 9265 }, { "epoch": 0.63, "learning_rate": 9.3635232457298e-06, "loss": 1.3708, "step": 9270 }, { "epoch": 0.63, "learning_rate": 9.362580596508428e-06, "loss": 0.8702, "step": 9275 }, { "epoch": 0.63, "learning_rate": 9.361637947287057e-06, "loss": 0.511, "step": 9280 }, { "epoch": 0.63, "learning_rate": 9.360695298065685e-06, "loss": 0.7852, "step": 9285 }, { "epoch": 0.63, "learning_rate": 9.359752648844312e-06, "loss": 0.6054, "step": 9290 }, { "epoch": 0.63, "learning_rate": 9.358809999622942e-06, "loss": 0.9154, "step": 9295 }, { "epoch": 0.63, "learning_rate": 9.35786735040157e-06, "loss": 0.6351, "step": 9300 }, { "epoch": 0.63, "learning_rate": 9.356924701180198e-06, "loss": 0.6161, "step": 9305 }, { "epoch": 0.63, "learning_rate": 9.355982051958826e-06, "loss": 0.9811, "step": 9310 }, { "epoch": 0.63, "learning_rate": 9.355039402737454e-06, "loss": 0.3546, "step": 9315 }, { "epoch": 0.63, "learning_rate": 9.354096753516081e-06, "loss": 0.2324, "step": 9320 }, { "epoch": 0.63, "learning_rate": 9.35315410429471e-06, "loss": 0.6491, "step": 9325 }, { "epoch": 0.63, "learning_rate": 9.352211455073338e-06, "loss": 0.7754, "step": 9330 }, { "epoch": 0.63, "learning_rate": 9.351268805851968e-06, "loss": 0.8258, "step": 9335 }, { "epoch": 0.63, "learning_rate": 9.350326156630595e-06, "loss": 0.5766, "step": 9340 }, { "epoch": 0.63, "learning_rate": 9.349383507409223e-06, "loss": 0.6281, "step": 9345 }, { "epoch": 0.63, "learning_rate": 9.348440858187852e-06, "loss": 1.0672, "step": 9350 }, { "epoch": 0.63, "learning_rate": 9.34749820896648e-06, "loss": 0.6502, "step": 9355 }, { "epoch": 0.64, "learning_rate": 9.34655555974511e-06, "loss": 0.737, "step": 9360 }, { "epoch": 0.64, "learning_rate": 9.345612910523737e-06, "loss": 0.617, "step": 9365 }, { "epoch": 0.64, "learning_rate": 9.344670261302365e-06, "loss": 0.4386, "step": 9370 }, { "epoch": 0.64, "learning_rate": 9.343727612080992e-06, "loss": 0.966, "step": 9375 }, { "epoch": 0.64, "learning_rate": 9.342784962859622e-06, "loss": 0.6289, "step": 9380 }, { "epoch": 0.64, "learning_rate": 9.34184231363825e-06, "loss": 0.4161, "step": 9385 }, { "epoch": 0.64, "learning_rate": 9.340899664416879e-06, "loss": 0.4114, "step": 9390 }, { "epoch": 0.64, "learning_rate": 9.339957015195506e-06, "loss": 1.1567, "step": 9395 }, { "epoch": 0.64, "learning_rate": 9.339014365974136e-06, "loss": 0.4488, "step": 9400 }, { "epoch": 0.64, "learning_rate": 9.338071716752763e-06, "loss": 0.6996, "step": 9405 }, { "epoch": 0.64, "learning_rate": 9.33712906753139e-06, "loss": 0.6485, "step": 9410 }, { "epoch": 0.64, "learning_rate": 9.336186418310018e-06, "loss": 1.0928, "step": 9415 }, { "epoch": 0.64, "learning_rate": 9.335243769088648e-06, "loss": 0.5773, "step": 9420 }, { "epoch": 0.64, "learning_rate": 9.334301119867275e-06, "loss": 0.6732, "step": 9425 }, { "epoch": 0.64, "learning_rate": 9.333358470645903e-06, "loss": 0.2451, "step": 9430 }, { "epoch": 0.64, "learning_rate": 9.332415821424532e-06, "loss": 0.6542, "step": 9435 }, { "epoch": 0.64, "learning_rate": 9.33147317220316e-06, "loss": 1.1153, "step": 9440 }, { "epoch": 0.64, "learning_rate": 9.33053052298179e-06, "loss": 0.3873, "step": 9445 }, { "epoch": 0.64, "learning_rate": 9.329587873760417e-06, "loss": 1.1026, "step": 9450 }, { "epoch": 0.64, "learning_rate": 9.328645224539046e-06, "loss": 0.9049, "step": 9455 }, { "epoch": 0.64, "learning_rate": 9.327702575317672e-06, "loss": 0.4381, "step": 9460 }, { "epoch": 0.64, "learning_rate": 9.326759926096302e-06, "loss": 0.4466, "step": 9465 }, { "epoch": 0.64, "learning_rate": 9.32581727687493e-06, "loss": 0.6223, "step": 9470 }, { "epoch": 0.64, "learning_rate": 9.324874627653559e-06, "loss": 0.4218, "step": 9475 }, { "epoch": 0.64, "learning_rate": 9.323931978432186e-06, "loss": 0.8633, "step": 9480 }, { "epoch": 0.64, "learning_rate": 9.322989329210816e-06, "loss": 1.1941, "step": 9485 }, { "epoch": 0.64, "learning_rate": 9.322046679989443e-06, "loss": 0.5448, "step": 9490 }, { "epoch": 0.64, "learning_rate": 9.321104030768071e-06, "loss": 0.5271, "step": 9495 }, { "epoch": 0.64, "learning_rate": 9.3201613815467e-06, "loss": 2.1169, "step": 9500 }, { "epoch": 0.65, "learning_rate": 9.319218732325328e-06, "loss": 0.5491, "step": 9505 }, { "epoch": 0.65, "learning_rate": 9.318276083103957e-06, "loss": 0.4972, "step": 9510 }, { "epoch": 0.65, "learning_rate": 9.317333433882583e-06, "loss": 0.782, "step": 9515 }, { "epoch": 0.65, "learning_rate": 9.316390784661212e-06, "loss": 0.6501, "step": 9520 }, { "epoch": 0.65, "learning_rate": 9.31544813543984e-06, "loss": 0.9136, "step": 9525 }, { "epoch": 0.65, "learning_rate": 9.31450548621847e-06, "loss": 0.3521, "step": 9530 }, { "epoch": 0.65, "learning_rate": 9.313562836997097e-06, "loss": 1.1939, "step": 9535 }, { "epoch": 0.65, "learning_rate": 9.312620187775726e-06, "loss": 1.0228, "step": 9540 }, { "epoch": 0.65, "learning_rate": 9.311677538554354e-06, "loss": 0.452, "step": 9545 }, { "epoch": 0.65, "learning_rate": 9.310734889332982e-06, "loss": 0.8828, "step": 9550 }, { "epoch": 0.65, "learning_rate": 9.309792240111611e-06, "loss": 0.8953, "step": 9555 }, { "epoch": 0.65, "learning_rate": 9.308849590890239e-06, "loss": 1.0051, "step": 9560 }, { "epoch": 0.65, "learning_rate": 9.307906941668868e-06, "loss": 0.7116, "step": 9565 }, { "epoch": 0.65, "learning_rate": 9.306964292447496e-06, "loss": 1.0593, "step": 9570 }, { "epoch": 0.65, "learning_rate": 9.306021643226123e-06, "loss": 0.4728, "step": 9575 }, { "epoch": 0.65, "learning_rate": 9.305078994004751e-06, "loss": 0.3535, "step": 9580 }, { "epoch": 0.65, "learning_rate": 9.30413634478338e-06, "loss": 0.7036, "step": 9585 }, { "epoch": 0.65, "learning_rate": 9.303193695562008e-06, "loss": 0.5127, "step": 9590 }, { "epoch": 0.65, "learning_rate": 9.302251046340637e-06, "loss": 0.5859, "step": 9595 }, { "epoch": 0.65, "learning_rate": 9.301308397119265e-06, "loss": 0.4864, "step": 9600 }, { "epoch": 0.65, "learning_rate": 9.300365747897892e-06, "loss": 0.5002, "step": 9605 }, { "epoch": 0.65, "learning_rate": 9.299423098676522e-06, "loss": 0.9247, "step": 9610 }, { "epoch": 0.65, "learning_rate": 9.29848044945515e-06, "loss": 0.4426, "step": 9615 }, { "epoch": 0.65, "learning_rate": 9.297537800233777e-06, "loss": 0.9075, "step": 9620 }, { "epoch": 0.65, "learning_rate": 9.296595151012406e-06, "loss": 0.3506, "step": 9625 }, { "epoch": 0.65, "learning_rate": 9.295652501791034e-06, "loss": 0.7969, "step": 9630 }, { "epoch": 0.65, "learning_rate": 9.294709852569662e-06, "loss": 0.9718, "step": 9635 }, { "epoch": 0.65, "learning_rate": 9.293767203348291e-06, "loss": 0.4412, "step": 9640 }, { "epoch": 0.65, "learning_rate": 9.292824554126919e-06, "loss": 0.874, "step": 9645 }, { "epoch": 0.65, "learning_rate": 9.291881904905548e-06, "loss": 0.7242, "step": 9650 }, { "epoch": 0.66, "learning_rate": 9.290939255684176e-06, "loss": 0.4016, "step": 9655 }, { "epoch": 0.66, "learning_rate": 9.289996606462803e-06, "loss": 0.8977, "step": 9660 }, { "epoch": 0.66, "learning_rate": 9.289053957241431e-06, "loss": 0.4434, "step": 9665 }, { "epoch": 0.66, "learning_rate": 9.28811130802006e-06, "loss": 0.9108, "step": 9670 }, { "epoch": 0.66, "learning_rate": 9.287168658798688e-06, "loss": 0.6625, "step": 9675 }, { "epoch": 0.66, "learning_rate": 9.286226009577317e-06, "loss": 1.272, "step": 9680 }, { "epoch": 0.66, "learning_rate": 9.285283360355945e-06, "loss": 1.0608, "step": 9685 }, { "epoch": 0.66, "learning_rate": 9.284340711134573e-06, "loss": 0.543, "step": 9690 }, { "epoch": 0.66, "learning_rate": 9.283398061913202e-06, "loss": 0.6103, "step": 9695 }, { "epoch": 0.66, "learning_rate": 9.28245541269183e-06, "loss": 0.6475, "step": 9700 }, { "epoch": 0.66, "learning_rate": 9.281512763470459e-06, "loss": 0.5489, "step": 9705 }, { "epoch": 0.66, "learning_rate": 9.280570114249086e-06, "loss": 1.2492, "step": 9710 }, { "epoch": 0.66, "learning_rate": 9.279627465027716e-06, "loss": 1.477, "step": 9715 }, { "epoch": 0.66, "learning_rate": 9.278684815806342e-06, "loss": 1.0493, "step": 9720 }, { "epoch": 0.66, "learning_rate": 9.277742166584971e-06, "loss": 0.8024, "step": 9725 }, { "epoch": 0.66, "learning_rate": 9.276799517363599e-06, "loss": 0.4832, "step": 9730 }, { "epoch": 0.66, "learning_rate": 9.275856868142228e-06, "loss": 0.3757, "step": 9735 }, { "epoch": 0.66, "learning_rate": 9.274914218920856e-06, "loss": 1.5943, "step": 9740 }, { "epoch": 0.66, "learning_rate": 9.273971569699483e-06, "loss": 0.4603, "step": 9745 }, { "epoch": 0.66, "learning_rate": 9.273028920478113e-06, "loss": 0.4245, "step": 9750 }, { "epoch": 0.66, "learning_rate": 9.27208627125674e-06, "loss": 0.5603, "step": 9755 }, { "epoch": 0.66, "learning_rate": 9.27114362203537e-06, "loss": 0.4885, "step": 9760 }, { "epoch": 0.66, "learning_rate": 9.270200972813997e-06, "loss": 0.9033, "step": 9765 }, { "epoch": 0.66, "learning_rate": 9.269258323592627e-06, "loss": 0.5804, "step": 9770 }, { "epoch": 0.66, "learning_rate": 9.268315674371253e-06, "loss": 1.199, "step": 9775 }, { "epoch": 0.66, "learning_rate": 9.267373025149882e-06, "loss": 0.6974, "step": 9780 }, { "epoch": 0.66, "learning_rate": 9.26643037592851e-06, "loss": 0.7421, "step": 9785 }, { "epoch": 0.66, "learning_rate": 9.265487726707139e-06, "loss": 0.643, "step": 9790 }, { "epoch": 0.66, "learning_rate": 9.264545077485767e-06, "loss": 0.406, "step": 9795 }, { "epoch": 0.67, "learning_rate": 9.263602428264396e-06, "loss": 0.6884, "step": 9800 }, { "epoch": 0.67, "learning_rate": 9.262659779043024e-06, "loss": 0.7092, "step": 9805 }, { "epoch": 0.67, "learning_rate": 9.261717129821651e-06, "loss": 0.4274, "step": 9810 }, { "epoch": 0.67, "learning_rate": 9.26077448060028e-06, "loss": 0.5984, "step": 9815 }, { "epoch": 0.67, "learning_rate": 9.259831831378908e-06, "loss": 0.3293, "step": 9820 }, { "epoch": 0.67, "learning_rate": 9.258889182157537e-06, "loss": 0.2114, "step": 9825 }, { "epoch": 0.67, "learning_rate": 9.257946532936165e-06, "loss": 1.3534, "step": 9830 }, { "epoch": 0.67, "learning_rate": 9.257003883714793e-06, "loss": 0.6627, "step": 9835 }, { "epoch": 0.67, "learning_rate": 9.25606123449342e-06, "loss": 0.7, "step": 9840 }, { "epoch": 0.67, "learning_rate": 9.25511858527205e-06, "loss": 0.6726, "step": 9845 }, { "epoch": 0.67, "learning_rate": 9.254175936050677e-06, "loss": 0.5711, "step": 9850 }, { "epoch": 0.67, "learning_rate": 9.253233286829307e-06, "loss": 1.0069, "step": 9855 }, { "epoch": 0.67, "learning_rate": 9.252290637607934e-06, "loss": 0.3239, "step": 9860 }, { "epoch": 0.67, "learning_rate": 9.251347988386562e-06, "loss": 0.7822, "step": 9865 }, { "epoch": 0.67, "learning_rate": 9.250405339165191e-06, "loss": 1.2489, "step": 9870 }, { "epoch": 0.67, "learning_rate": 9.249462689943819e-06, "loss": 0.3549, "step": 9875 }, { "epoch": 0.67, "learning_rate": 9.248520040722447e-06, "loss": 0.7068, "step": 9880 }, { "epoch": 0.67, "learning_rate": 9.247577391501076e-06, "loss": 1.5338, "step": 9885 }, { "epoch": 0.67, "learning_rate": 9.246634742279704e-06, "loss": 1.0679, "step": 9890 }, { "epoch": 0.67, "learning_rate": 9.245692093058331e-06, "loss": 0.3098, "step": 9895 }, { "epoch": 0.67, "learning_rate": 9.24474944383696e-06, "loss": 0.9854, "step": 9900 }, { "epoch": 0.67, "learning_rate": 9.243806794615588e-06, "loss": 0.673, "step": 9905 }, { "epoch": 0.67, "learning_rate": 9.242864145394218e-06, "loss": 1.1656, "step": 9910 }, { "epoch": 0.67, "learning_rate": 9.241921496172845e-06, "loss": 0.9117, "step": 9915 }, { "epoch": 0.67, "learning_rate": 9.240978846951473e-06, "loss": 0.8016, "step": 9920 }, { "epoch": 0.67, "learning_rate": 9.2400361977301e-06, "loss": 0.6401, "step": 9925 }, { "epoch": 0.67, "learning_rate": 9.23909354850873e-06, "loss": 0.9533, "step": 9930 }, { "epoch": 0.67, "learning_rate": 9.238150899287357e-06, "loss": 0.3988, "step": 9935 }, { "epoch": 0.67, "learning_rate": 9.237208250065987e-06, "loss": 0.615, "step": 9940 }, { "epoch": 0.67, "learning_rate": 9.236265600844614e-06, "loss": 0.5366, "step": 9945 }, { "epoch": 0.68, "learning_rate": 9.235322951623242e-06, "loss": 0.3744, "step": 9950 }, { "epoch": 0.68, "learning_rate": 9.234380302401871e-06, "loss": 0.8185, "step": 9955 }, { "epoch": 0.68, "learning_rate": 9.233437653180499e-06, "loss": 0.6745, "step": 9960 }, { "epoch": 0.68, "learning_rate": 9.232495003959128e-06, "loss": 0.968, "step": 9965 }, { "epoch": 0.68, "learning_rate": 9.231552354737756e-06, "loss": 0.8106, "step": 9970 }, { "epoch": 0.68, "learning_rate": 9.230609705516384e-06, "loss": 0.9602, "step": 9975 }, { "epoch": 0.68, "learning_rate": 9.229667056295011e-06, "loss": 0.5687, "step": 9980 }, { "epoch": 0.68, "learning_rate": 9.22872440707364e-06, "loss": 0.1705, "step": 9985 }, { "epoch": 0.68, "learning_rate": 9.227781757852268e-06, "loss": 0.441, "step": 9990 }, { "epoch": 0.68, "learning_rate": 9.226839108630898e-06, "loss": 0.4099, "step": 9995 }, { "epoch": 0.68, "learning_rate": 9.225896459409525e-06, "loss": 0.9835, "step": 10000 }, { "epoch": 0.68, "learning_rate": 9.224953810188153e-06, "loss": 0.9062, "step": 10005 }, { "epoch": 0.68, "learning_rate": 9.224011160966782e-06, "loss": 0.7714, "step": 10010 }, { "epoch": 0.68, "learning_rate": 9.22306851174541e-06, "loss": 0.6902, "step": 10015 }, { "epoch": 0.68, "learning_rate": 9.222125862524039e-06, "loss": 0.5503, "step": 10020 }, { "epoch": 0.68, "learning_rate": 9.221183213302667e-06, "loss": 0.2009, "step": 10025 }, { "epoch": 0.68, "learning_rate": 9.220240564081296e-06, "loss": 0.2202, "step": 10030 }, { "epoch": 0.68, "learning_rate": 9.219297914859922e-06, "loss": 1.5577, "step": 10035 }, { "epoch": 0.68, "learning_rate": 9.218355265638551e-06, "loss": 0.2124, "step": 10040 }, { "epoch": 0.68, "learning_rate": 9.217412616417179e-06, "loss": 1.1054, "step": 10045 }, { "epoch": 0.68, "learning_rate": 9.216469967195808e-06, "loss": 0.3191, "step": 10050 }, { "epoch": 0.68, "learning_rate": 9.215527317974436e-06, "loss": 0.4928, "step": 10055 }, { "epoch": 0.68, "learning_rate": 9.214584668753065e-06, "loss": 1.2832, "step": 10060 }, { "epoch": 0.68, "learning_rate": 9.213642019531693e-06, "loss": 0.1889, "step": 10065 }, { "epoch": 0.68, "learning_rate": 9.21269937031032e-06, "loss": 0.495, "step": 10070 }, { "epoch": 0.68, "learning_rate": 9.21175672108895e-06, "loss": 1.1576, "step": 10075 }, { "epoch": 0.68, "learning_rate": 9.210814071867578e-06, "loss": 0.8407, "step": 10080 }, { "epoch": 0.68, "learning_rate": 9.209871422646205e-06, "loss": 0.8642, "step": 10085 }, { "epoch": 0.68, "learning_rate": 9.208928773424833e-06, "loss": 0.5322, "step": 10090 }, { "epoch": 0.69, "learning_rate": 9.207986124203462e-06, "loss": 0.7761, "step": 10095 }, { "epoch": 0.69, "learning_rate": 9.20704347498209e-06, "loss": 1.613, "step": 10100 }, { "epoch": 0.69, "learning_rate": 9.20610082576072e-06, "loss": 0.8281, "step": 10105 }, { "epoch": 0.69, "learning_rate": 9.205158176539347e-06, "loss": 0.4419, "step": 10110 }, { "epoch": 0.69, "learning_rate": 9.204215527317976e-06, "loss": 0.8499, "step": 10115 }, { "epoch": 0.69, "learning_rate": 9.203272878096604e-06, "loss": 0.7873, "step": 10120 }, { "epoch": 0.69, "learning_rate": 9.202330228875231e-06, "loss": 0.5038, "step": 10125 }, { "epoch": 0.69, "learning_rate": 9.201387579653859e-06, "loss": 0.4539, "step": 10130 }, { "epoch": 0.69, "learning_rate": 9.200444930432488e-06, "loss": 0.9937, "step": 10135 }, { "epoch": 0.69, "learning_rate": 9.199502281211116e-06, "loss": 0.5087, "step": 10140 }, { "epoch": 0.69, "learning_rate": 9.198559631989745e-06, "loss": 0.5091, "step": 10145 }, { "epoch": 0.69, "learning_rate": 9.197616982768373e-06, "loss": 1.1807, "step": 10150 }, { "epoch": 0.69, "learning_rate": 9.196674333547e-06, "loss": 0.3365, "step": 10155 }, { "epoch": 0.69, "learning_rate": 9.19573168432563e-06, "loss": 1.0015, "step": 10160 }, { "epoch": 0.69, "learning_rate": 9.194789035104258e-06, "loss": 0.673, "step": 10165 }, { "epoch": 0.69, "learning_rate": 9.193846385882887e-06, "loss": 0.8411, "step": 10170 }, { "epoch": 0.69, "learning_rate": 9.192903736661515e-06, "loss": 0.8806, "step": 10175 }, { "epoch": 0.69, "learning_rate": 9.191961087440142e-06, "loss": 0.7037, "step": 10180 }, { "epoch": 0.69, "learning_rate": 9.19101843821877e-06, "loss": 0.3566, "step": 10185 }, { "epoch": 0.69, "learning_rate": 9.1900757889974e-06, "loss": 0.7612, "step": 10190 }, { "epoch": 0.69, "learning_rate": 9.189133139776027e-06, "loss": 0.8694, "step": 10195 }, { "epoch": 0.69, "learning_rate": 9.188190490554656e-06, "loss": 1.1111, "step": 10200 }, { "epoch": 0.69, "learning_rate": 9.187247841333284e-06, "loss": 0.8438, "step": 10205 }, { "epoch": 0.69, "learning_rate": 9.186305192111911e-06, "loss": 0.7281, "step": 10210 }, { "epoch": 0.69, "learning_rate": 9.18536254289054e-06, "loss": 0.3617, "step": 10215 }, { "epoch": 0.69, "learning_rate": 9.184419893669168e-06, "loss": 0.886, "step": 10220 }, { "epoch": 0.69, "learning_rate": 9.183477244447798e-06, "loss": 0.5814, "step": 10225 }, { "epoch": 0.69, "learning_rate": 9.182534595226425e-06, "loss": 0.8027, "step": 10230 }, { "epoch": 0.69, "learning_rate": 9.181591946005053e-06, "loss": 0.5195, "step": 10235 }, { "epoch": 0.69, "learning_rate": 9.18064929678368e-06, "loss": 0.4146, "step": 10240 }, { "epoch": 0.7, "learning_rate": 9.17970664756231e-06, "loss": 1.1208, "step": 10245 }, { "epoch": 0.7, "learning_rate": 9.178763998340938e-06, "loss": 0.9771, "step": 10250 }, { "epoch": 0.7, "learning_rate": 9.177821349119567e-06, "loss": 0.4112, "step": 10255 }, { "epoch": 0.7, "learning_rate": 9.176878699898195e-06, "loss": 0.628, "step": 10260 }, { "epoch": 0.7, "learning_rate": 9.175936050676822e-06, "loss": 0.9982, "step": 10265 }, { "epoch": 0.7, "learning_rate": 9.174993401455452e-06, "loss": 0.7261, "step": 10270 }, { "epoch": 0.7, "learning_rate": 9.17405075223408e-06, "loss": 0.5281, "step": 10275 }, { "epoch": 0.7, "learning_rate": 9.173108103012709e-06, "loss": 0.6544, "step": 10280 }, { "epoch": 0.7, "learning_rate": 9.172165453791336e-06, "loss": 0.2009, "step": 10285 }, { "epoch": 0.7, "learning_rate": 9.171222804569964e-06, "loss": 1.0509, "step": 10290 }, { "epoch": 0.7, "learning_rate": 9.170280155348592e-06, "loss": 0.7644, "step": 10295 }, { "epoch": 0.7, "learning_rate": 9.169337506127221e-06, "loss": 0.7649, "step": 10300 }, { "epoch": 0.7, "learning_rate": 9.168394856905849e-06, "loss": 0.4874, "step": 10305 }, { "epoch": 0.7, "learning_rate": 9.167452207684478e-06, "loss": 0.4921, "step": 10310 }, { "epoch": 0.7, "learning_rate": 9.166509558463105e-06, "loss": 0.539, "step": 10315 }, { "epoch": 0.7, "learning_rate": 9.165566909241733e-06, "loss": 0.5996, "step": 10320 }, { "epoch": 0.7, "learning_rate": 9.164624260020362e-06, "loss": 1.0399, "step": 10325 }, { "epoch": 0.7, "learning_rate": 9.16368161079899e-06, "loss": 0.8599, "step": 10330 }, { "epoch": 0.7, "learning_rate": 9.162738961577618e-06, "loss": 0.9676, "step": 10335 }, { "epoch": 0.7, "learning_rate": 9.161796312356247e-06, "loss": 0.6327, "step": 10340 }, { "epoch": 0.7, "learning_rate": 9.160853663134875e-06, "loss": 0.3911, "step": 10345 }, { "epoch": 0.7, "learning_rate": 9.159911013913502e-06, "loss": 1.0681, "step": 10350 }, { "epoch": 0.7, "learning_rate": 9.158968364692132e-06, "loss": 0.3923, "step": 10355 }, { "epoch": 0.7, "learning_rate": 9.15802571547076e-06, "loss": 0.7329, "step": 10360 }, { "epoch": 0.7, "learning_rate": 9.157083066249389e-06, "loss": 1.1218, "step": 10365 }, { "epoch": 0.7, "learning_rate": 9.156140417028016e-06, "loss": 1.2073, "step": 10370 }, { "epoch": 0.7, "learning_rate": 9.155197767806646e-06, "loss": 0.6062, "step": 10375 }, { "epoch": 0.7, "learning_rate": 9.154255118585272e-06, "loss": 0.6526, "step": 10380 }, { "epoch": 0.7, "learning_rate": 9.153312469363901e-06, "loss": 0.9093, "step": 10385 }, { "epoch": 0.71, "learning_rate": 9.152369820142529e-06, "loss": 0.7632, "step": 10390 }, { "epoch": 0.71, "learning_rate": 9.151427170921158e-06, "loss": 0.4745, "step": 10395 }, { "epoch": 0.71, "learning_rate": 9.150484521699786e-06, "loss": 1.0283, "step": 10400 }, { "epoch": 0.71, "learning_rate": 9.149541872478415e-06, "loss": 0.704, "step": 10405 }, { "epoch": 0.71, "learning_rate": 9.148599223257043e-06, "loss": 0.2013, "step": 10410 }, { "epoch": 0.71, "learning_rate": 9.14765657403567e-06, "loss": 0.2504, "step": 10415 }, { "epoch": 0.71, "learning_rate": 9.1467139248143e-06, "loss": 0.6696, "step": 10420 }, { "epoch": 0.71, "learning_rate": 9.145771275592927e-06, "loss": 0.8648, "step": 10425 }, { "epoch": 0.71, "learning_rate": 9.144828626371556e-06, "loss": 0.8974, "step": 10430 }, { "epoch": 0.71, "learning_rate": 9.143885977150182e-06, "loss": 0.3244, "step": 10435 }, { "epoch": 0.71, "learning_rate": 9.142943327928812e-06, "loss": 0.4399, "step": 10440 }, { "epoch": 0.71, "learning_rate": 9.14200067870744e-06, "loss": 0.3137, "step": 10445 }, { "epoch": 0.71, "learning_rate": 9.141058029486069e-06, "loss": 0.6147, "step": 10450 }, { "epoch": 0.71, "learning_rate": 9.140115380264696e-06, "loss": 0.5862, "step": 10455 }, { "epoch": 0.71, "learning_rate": 9.139172731043326e-06, "loss": 1.1713, "step": 10460 }, { "epoch": 0.71, "learning_rate": 9.138230081821953e-06, "loss": 0.5921, "step": 10465 }, { "epoch": 0.71, "learning_rate": 9.137287432600581e-06, "loss": 0.4345, "step": 10470 }, { "epoch": 0.71, "learning_rate": 9.13634478337921e-06, "loss": 1.0882, "step": 10475 }, { "epoch": 0.71, "learning_rate": 9.135402134157838e-06, "loss": 0.6073, "step": 10480 }, { "epoch": 0.71, "learning_rate": 9.134459484936467e-06, "loss": 1.2061, "step": 10485 }, { "epoch": 0.71, "learning_rate": 9.133516835715095e-06, "loss": 0.6553, "step": 10490 }, { "epoch": 0.71, "learning_rate": 9.132574186493723e-06, "loss": 0.6838, "step": 10495 }, { "epoch": 0.71, "learning_rate": 9.13163153727235e-06, "loss": 0.1482, "step": 10500 }, { "epoch": 0.71, "learning_rate": 9.13068888805098e-06, "loss": 0.5958, "step": 10505 }, { "epoch": 0.71, "learning_rate": 9.129746238829607e-06, "loss": 0.4726, "step": 10510 }, { "epoch": 0.71, "learning_rate": 9.128803589608237e-06, "loss": 0.8645, "step": 10515 }, { "epoch": 0.71, "learning_rate": 9.127860940386864e-06, "loss": 0.43, "step": 10520 }, { "epoch": 0.71, "learning_rate": 9.126918291165492e-06, "loss": 0.8064, "step": 10525 }, { "epoch": 0.71, "learning_rate": 9.125975641944121e-06, "loss": 0.2376, "step": 10530 }, { "epoch": 0.72, "learning_rate": 9.125032992722749e-06, "loss": 0.9526, "step": 10535 }, { "epoch": 0.72, "learning_rate": 9.124090343501376e-06, "loss": 0.6905, "step": 10540 }, { "epoch": 0.72, "learning_rate": 9.123147694280006e-06, "loss": 0.8559, "step": 10545 }, { "epoch": 0.72, "learning_rate": 9.122205045058633e-06, "loss": 0.3049, "step": 10550 }, { "epoch": 0.72, "learning_rate": 9.121262395837261e-06, "loss": 0.3265, "step": 10555 }, { "epoch": 0.72, "learning_rate": 9.12031974661589e-06, "loss": 0.9429, "step": 10560 }, { "epoch": 0.72, "learning_rate": 9.119377097394518e-06, "loss": 0.7491, "step": 10565 }, { "epoch": 0.72, "learning_rate": 9.118434448173147e-06, "loss": 0.4667, "step": 10570 }, { "epoch": 0.72, "learning_rate": 9.117491798951775e-06, "loss": 0.6065, "step": 10575 }, { "epoch": 0.72, "learning_rate": 9.116549149730403e-06, "loss": 0.4125, "step": 10580 }, { "epoch": 0.72, "learning_rate": 9.11560650050903e-06, "loss": 0.8141, "step": 10585 }, { "epoch": 0.72, "learning_rate": 9.11466385128766e-06, "loss": 1.177, "step": 10590 }, { "epoch": 0.72, "learning_rate": 9.113721202066287e-06, "loss": 1.2779, "step": 10595 }, { "epoch": 0.72, "learning_rate": 9.112778552844917e-06, "loss": 1.02, "step": 10600 }, { "epoch": 0.72, "learning_rate": 9.111835903623544e-06, "loss": 0.2619, "step": 10605 }, { "epoch": 0.72, "learning_rate": 9.110893254402172e-06, "loss": 0.6135, "step": 10610 }, { "epoch": 0.72, "learning_rate": 9.109950605180801e-06, "loss": 0.3264, "step": 10615 }, { "epoch": 0.72, "learning_rate": 9.109007955959429e-06, "loss": 0.7113, "step": 10620 }, { "epoch": 0.72, "learning_rate": 9.108065306738058e-06, "loss": 0.5766, "step": 10625 }, { "epoch": 0.72, "learning_rate": 9.107122657516686e-06, "loss": 0.6882, "step": 10630 }, { "epoch": 0.72, "learning_rate": 9.106180008295313e-06, "loss": 0.9573, "step": 10635 }, { "epoch": 0.72, "learning_rate": 9.105237359073941e-06, "loss": 0.4452, "step": 10640 }, { "epoch": 0.72, "learning_rate": 9.10429470985257e-06, "loss": 0.441, "step": 10645 }, { "epoch": 0.72, "learning_rate": 9.103352060631198e-06, "loss": 0.79, "step": 10650 }, { "epoch": 0.72, "learning_rate": 9.102409411409827e-06, "loss": 0.9407, "step": 10655 }, { "epoch": 0.72, "learning_rate": 9.101466762188455e-06, "loss": 0.4594, "step": 10660 }, { "epoch": 0.72, "learning_rate": 9.100524112967083e-06, "loss": 1.5738, "step": 10665 }, { "epoch": 0.72, "learning_rate": 9.099581463745712e-06, "loss": 0.3442, "step": 10670 }, { "epoch": 0.72, "learning_rate": 9.09863881452434e-06, "loss": 1.138, "step": 10675 }, { "epoch": 0.72, "learning_rate": 9.097696165302969e-06, "loss": 1.0154, "step": 10680 }, { "epoch": 0.73, "learning_rate": 9.096753516081597e-06, "loss": 0.8608, "step": 10685 }, { "epoch": 0.73, "learning_rate": 9.095810866860226e-06, "loss": 0.5652, "step": 10690 }, { "epoch": 0.73, "learning_rate": 9.094868217638852e-06, "loss": 0.8198, "step": 10695 }, { "epoch": 0.73, "learning_rate": 9.093925568417481e-06, "loss": 1.0083, "step": 10700 }, { "epoch": 0.73, "learning_rate": 9.092982919196109e-06, "loss": 0.4243, "step": 10705 }, { "epoch": 0.73, "learning_rate": 9.092040269974738e-06, "loss": 0.464, "step": 10710 }, { "epoch": 0.73, "learning_rate": 9.091097620753366e-06, "loss": 0.6639, "step": 10715 }, { "epoch": 0.73, "learning_rate": 9.090154971531995e-06, "loss": 0.6563, "step": 10720 }, { "epoch": 0.73, "learning_rate": 9.089212322310623e-06, "loss": 1.2343, "step": 10725 }, { "epoch": 0.73, "learning_rate": 9.08826967308925e-06, "loss": 0.8304, "step": 10730 }, { "epoch": 0.73, "learning_rate": 9.08732702386788e-06, "loss": 0.4908, "step": 10735 }, { "epoch": 0.73, "learning_rate": 9.086384374646507e-06, "loss": 0.5263, "step": 10740 }, { "epoch": 0.73, "learning_rate": 9.085441725425135e-06, "loss": 0.9624, "step": 10745 }, { "epoch": 0.73, "learning_rate": 9.084499076203763e-06, "loss": 0.9152, "step": 10750 }, { "epoch": 0.73, "learning_rate": 9.083556426982392e-06, "loss": 1.0656, "step": 10755 }, { "epoch": 0.73, "learning_rate": 9.08261377776102e-06, "loss": 0.3498, "step": 10760 }, { "epoch": 0.73, "learning_rate": 9.081671128539649e-06, "loss": 0.7918, "step": 10765 }, { "epoch": 0.73, "learning_rate": 9.080728479318277e-06, "loss": 0.3473, "step": 10770 }, { "epoch": 0.73, "learning_rate": 9.079785830096906e-06, "loss": 0.7213, "step": 10775 }, { "epoch": 0.73, "learning_rate": 9.078843180875534e-06, "loss": 1.0031, "step": 10780 }, { "epoch": 0.73, "learning_rate": 9.077900531654161e-06, "loss": 1.2128, "step": 10785 }, { "epoch": 0.73, "learning_rate": 9.076957882432789e-06, "loss": 0.9947, "step": 10790 }, { "epoch": 0.73, "learning_rate": 9.076015233211418e-06, "loss": 0.3336, "step": 10795 }, { "epoch": 0.73, "learning_rate": 9.075072583990046e-06, "loss": 0.4723, "step": 10800 }, { "epoch": 0.73, "learning_rate": 9.074129934768675e-06, "loss": 0.323, "step": 10805 }, { "epoch": 0.73, "learning_rate": 9.073187285547303e-06, "loss": 0.484, "step": 10810 }, { "epoch": 0.73, "learning_rate": 9.07224463632593e-06, "loss": 0.8045, "step": 10815 }, { "epoch": 0.73, "learning_rate": 9.07130198710456e-06, "loss": 0.8252, "step": 10820 }, { "epoch": 0.73, "learning_rate": 9.070359337883187e-06, "loss": 0.2518, "step": 10825 }, { "epoch": 0.74, "learning_rate": 9.069416688661817e-06, "loss": 0.4373, "step": 10830 }, { "epoch": 0.74, "learning_rate": 9.068474039440444e-06, "loss": 0.5441, "step": 10835 }, { "epoch": 0.74, "learning_rate": 9.067531390219072e-06, "loss": 1.1478, "step": 10840 }, { "epoch": 0.74, "learning_rate": 9.0665887409977e-06, "loss": 1.0773, "step": 10845 }, { "epoch": 0.74, "learning_rate": 9.065646091776329e-06, "loss": 0.8231, "step": 10850 }, { "epoch": 0.74, "learning_rate": 9.064703442554957e-06, "loss": 0.4107, "step": 10855 }, { "epoch": 0.74, "learning_rate": 9.063760793333586e-06, "loss": 1.1631, "step": 10860 }, { "epoch": 0.74, "learning_rate": 9.062818144112214e-06, "loss": 0.8468, "step": 10865 }, { "epoch": 0.74, "learning_rate": 9.061875494890841e-06, "loss": 0.9972, "step": 10870 }, { "epoch": 0.74, "learning_rate": 9.06093284566947e-06, "loss": 0.3358, "step": 10875 }, { "epoch": 0.74, "learning_rate": 9.059990196448098e-06, "loss": 0.4177, "step": 10880 }, { "epoch": 0.74, "learning_rate": 9.059047547226728e-06, "loss": 0.3721, "step": 10885 }, { "epoch": 0.74, "learning_rate": 9.058104898005355e-06, "loss": 0.7092, "step": 10890 }, { "epoch": 0.74, "learning_rate": 9.057162248783983e-06, "loss": 0.7469, "step": 10895 }, { "epoch": 0.74, "learning_rate": 9.05621959956261e-06, "loss": 1.0159, "step": 10900 }, { "epoch": 0.74, "learning_rate": 9.05527695034124e-06, "loss": 0.5166, "step": 10905 }, { "epoch": 0.74, "learning_rate": 9.054334301119868e-06, "loss": 0.4925, "step": 10910 }, { "epoch": 0.74, "learning_rate": 9.053391651898497e-06, "loss": 0.9807, "step": 10915 }, { "epoch": 0.74, "learning_rate": 9.052449002677125e-06, "loss": 0.8672, "step": 10920 }, { "epoch": 0.74, "learning_rate": 9.051506353455752e-06, "loss": 1.0843, "step": 10925 }, { "epoch": 0.74, "learning_rate": 9.050563704234381e-06, "loss": 0.5444, "step": 10930 }, { "epoch": 0.74, "learning_rate": 9.049621055013009e-06, "loss": 0.3123, "step": 10935 }, { "epoch": 0.74, "learning_rate": 9.048678405791638e-06, "loss": 1.0143, "step": 10940 }, { "epoch": 0.74, "learning_rate": 9.047735756570266e-06, "loss": 0.6676, "step": 10945 }, { "epoch": 0.74, "learning_rate": 9.046793107348895e-06, "loss": 0.6093, "step": 10950 }, { "epoch": 0.74, "learning_rate": 9.045850458127521e-06, "loss": 0.4833, "step": 10955 }, { "epoch": 0.74, "learning_rate": 9.04490780890615e-06, "loss": 0.4374, "step": 10960 }, { "epoch": 0.74, "learning_rate": 9.043965159684778e-06, "loss": 0.3129, "step": 10965 }, { "epoch": 0.74, "learning_rate": 9.043022510463408e-06, "loss": 1.0796, "step": 10970 }, { "epoch": 0.74, "learning_rate": 9.042079861242035e-06, "loss": 0.3646, "step": 10975 }, { "epoch": 0.75, "learning_rate": 9.041137212020663e-06, "loss": 0.8634, "step": 10980 }, { "epoch": 0.75, "learning_rate": 9.040194562799292e-06, "loss": 0.461, "step": 10985 }, { "epoch": 0.75, "learning_rate": 9.03925191357792e-06, "loss": 0.7184, "step": 10990 }, { "epoch": 0.75, "learning_rate": 9.038309264356548e-06, "loss": 0.3724, "step": 10995 }, { "epoch": 0.75, "learning_rate": 9.037366615135177e-06, "loss": 0.6876, "step": 11000 }, { "epoch": 0.75, "learning_rate": 9.036423965913805e-06, "loss": 0.8629, "step": 11005 }, { "epoch": 0.75, "learning_rate": 9.035481316692432e-06, "loss": 0.7403, "step": 11010 }, { "epoch": 0.75, "learning_rate": 9.034538667471062e-06, "loss": 0.4703, "step": 11015 }, { "epoch": 0.75, "learning_rate": 9.03359601824969e-06, "loss": 0.7165, "step": 11020 }, { "epoch": 0.75, "learning_rate": 9.032653369028319e-06, "loss": 1.4685, "step": 11025 }, { "epoch": 0.75, "learning_rate": 9.031710719806946e-06, "loss": 0.2833, "step": 11030 }, { "epoch": 0.75, "learning_rate": 9.030768070585575e-06, "loss": 0.2791, "step": 11035 }, { "epoch": 0.75, "learning_rate": 9.029825421364201e-06, "loss": 0.4272, "step": 11040 }, { "epoch": 0.75, "learning_rate": 9.02888277214283e-06, "loss": 0.5302, "step": 11045 }, { "epoch": 0.75, "learning_rate": 9.027940122921458e-06, "loss": 0.3281, "step": 11050 }, { "epoch": 0.75, "learning_rate": 9.026997473700088e-06, "loss": 0.6699, "step": 11055 }, { "epoch": 0.75, "learning_rate": 9.026054824478715e-06, "loss": 1.1451, "step": 11060 }, { "epoch": 0.75, "learning_rate": 9.025112175257345e-06, "loss": 0.5441, "step": 11065 }, { "epoch": 0.75, "learning_rate": 9.024169526035972e-06, "loss": 1.1977, "step": 11070 }, { "epoch": 0.75, "learning_rate": 9.0232268768146e-06, "loss": 1.1769, "step": 11075 }, { "epoch": 0.75, "learning_rate": 9.02228422759323e-06, "loss": 0.1905, "step": 11080 }, { "epoch": 0.75, "learning_rate": 9.021341578371857e-06, "loss": 0.8302, "step": 11085 }, { "epoch": 0.75, "learning_rate": 9.020398929150486e-06, "loss": 0.6155, "step": 11090 }, { "epoch": 0.75, "learning_rate": 9.019456279929112e-06, "loss": 0.3651, "step": 11095 }, { "epoch": 0.75, "learning_rate": 9.018513630707742e-06, "loss": 1.0039, "step": 11100 }, { "epoch": 0.75, "learning_rate": 9.01757098148637e-06, "loss": 1.2849, "step": 11105 }, { "epoch": 0.75, "learning_rate": 9.016628332264999e-06, "loss": 0.8357, "step": 11110 }, { "epoch": 0.75, "learning_rate": 9.015685683043626e-06, "loss": 0.4947, "step": 11115 }, { "epoch": 0.75, "learning_rate": 9.014743033822256e-06, "loss": 0.9295, "step": 11120 }, { "epoch": 0.76, "learning_rate": 9.013800384600883e-06, "loss": 0.791, "step": 11125 }, { "epoch": 0.76, "learning_rate": 9.01285773537951e-06, "loss": 0.8757, "step": 11130 }, { "epoch": 0.76, "learning_rate": 9.01191508615814e-06, "loss": 0.2829, "step": 11135 }, { "epoch": 0.76, "learning_rate": 9.010972436936768e-06, "loss": 0.8045, "step": 11140 }, { "epoch": 0.76, "learning_rate": 9.010029787715397e-06, "loss": 1.2368, "step": 11145 }, { "epoch": 0.76, "learning_rate": 9.009087138494025e-06, "loss": 1.103, "step": 11150 }, { "epoch": 0.76, "learning_rate": 9.008144489272652e-06, "loss": 1.0309, "step": 11155 }, { "epoch": 0.76, "learning_rate": 9.00720184005128e-06, "loss": 0.8621, "step": 11160 }, { "epoch": 0.76, "learning_rate": 9.00625919082991e-06, "loss": 0.5928, "step": 11165 }, { "epoch": 0.76, "learning_rate": 9.005316541608537e-06, "loss": 1.2381, "step": 11170 }, { "epoch": 0.76, "learning_rate": 9.004373892387166e-06, "loss": 0.662, "step": 11175 }, { "epoch": 0.76, "learning_rate": 9.003431243165794e-06, "loss": 0.7491, "step": 11180 }, { "epoch": 0.76, "learning_rate": 9.002488593944422e-06, "loss": 0.5634, "step": 11185 }, { "epoch": 0.76, "learning_rate": 9.001545944723051e-06, "loss": 0.4959, "step": 11190 }, { "epoch": 0.76, "learning_rate": 9.000603295501679e-06, "loss": 1.0253, "step": 11195 }, { "epoch": 0.76, "learning_rate": 8.999660646280308e-06, "loss": 0.889, "step": 11200 }, { "epoch": 0.76, "learning_rate": 8.998717997058936e-06, "loss": 0.8044, "step": 11205 }, { "epoch": 0.76, "learning_rate": 8.997775347837563e-06, "loss": 0.3398, "step": 11210 }, { "epoch": 0.76, "learning_rate": 8.996832698616191e-06, "loss": 0.7005, "step": 11215 }, { "epoch": 0.76, "learning_rate": 8.99589004939482e-06, "loss": 0.3267, "step": 11220 }, { "epoch": 0.76, "learning_rate": 8.994947400173448e-06, "loss": 0.1414, "step": 11225 }, { "epoch": 0.76, "learning_rate": 8.994004750952077e-06, "loss": 0.9521, "step": 11230 }, { "epoch": 0.76, "learning_rate": 8.993062101730705e-06, "loss": 1.5084, "step": 11235 }, { "epoch": 0.76, "learning_rate": 8.992119452509332e-06, "loss": 0.769, "step": 11240 }, { "epoch": 0.76, "learning_rate": 8.991176803287962e-06, "loss": 0.6541, "step": 11245 }, { "epoch": 0.76, "learning_rate": 8.99023415406659e-06, "loss": 0.9189, "step": 11250 }, { "epoch": 0.76, "learning_rate": 8.989291504845217e-06, "loss": 0.7104, "step": 11255 }, { "epoch": 0.76, "learning_rate": 8.988348855623846e-06, "loss": 0.5865, "step": 11260 }, { "epoch": 0.76, "learning_rate": 8.987406206402474e-06, "loss": 0.3651, "step": 11265 }, { "epoch": 0.76, "learning_rate": 8.986463557181102e-06, "loss": 0.9239, "step": 11270 }, { "epoch": 0.77, "learning_rate": 8.985520907959731e-06, "loss": 0.6487, "step": 11275 }, { "epoch": 0.77, "learning_rate": 8.984578258738359e-06, "loss": 0.3945, "step": 11280 }, { "epoch": 0.77, "learning_rate": 8.983635609516988e-06, "loss": 0.7878, "step": 11285 }, { "epoch": 0.77, "learning_rate": 8.982692960295616e-06, "loss": 0.9292, "step": 11290 }, { "epoch": 0.77, "learning_rate": 8.981750311074245e-06, "loss": 0.5678, "step": 11295 }, { "epoch": 0.77, "learning_rate": 8.980807661852871e-06, "loss": 0.6782, "step": 11300 }, { "epoch": 0.77, "learning_rate": 8.9798650126315e-06, "loss": 0.8696, "step": 11305 }, { "epoch": 0.77, "learning_rate": 8.978922363410128e-06, "loss": 0.8648, "step": 11310 }, { "epoch": 0.77, "learning_rate": 8.977979714188757e-06, "loss": 0.966, "step": 11315 }, { "epoch": 0.77, "learning_rate": 8.977037064967385e-06, "loss": 0.4034, "step": 11320 }, { "epoch": 0.77, "learning_rate": 8.976094415746012e-06, "loss": 0.7589, "step": 11325 }, { "epoch": 0.77, "learning_rate": 8.975151766524642e-06, "loss": 1.7138, "step": 11330 }, { "epoch": 0.77, "learning_rate": 8.97420911730327e-06, "loss": 0.9079, "step": 11335 }, { "epoch": 0.77, "learning_rate": 8.973266468081899e-06, "loss": 0.9567, "step": 11340 }, { "epoch": 0.77, "learning_rate": 8.972323818860526e-06, "loss": 1.0319, "step": 11345 }, { "epoch": 0.77, "learning_rate": 8.971381169639156e-06, "loss": 0.6987, "step": 11350 }, { "epoch": 0.77, "learning_rate": 8.970438520417782e-06, "loss": 0.5155, "step": 11355 }, { "epoch": 0.77, "learning_rate": 8.969495871196411e-06, "loss": 0.8658, "step": 11360 }, { "epoch": 0.77, "learning_rate": 8.968553221975039e-06, "loss": 0.711, "step": 11365 }, { "epoch": 0.77, "learning_rate": 8.967610572753668e-06, "loss": 0.3569, "step": 11370 }, { "epoch": 0.77, "learning_rate": 8.966667923532296e-06, "loss": 0.7046, "step": 11375 }, { "epoch": 0.77, "learning_rate": 8.965725274310925e-06, "loss": 0.79, "step": 11380 }, { "epoch": 0.77, "learning_rate": 8.964782625089553e-06, "loss": 0.7162, "step": 11385 }, { "epoch": 0.77, "learning_rate": 8.96383997586818e-06, "loss": 0.1846, "step": 11390 }, { "epoch": 0.77, "learning_rate": 8.96289732664681e-06, "loss": 0.6203, "step": 11395 }, { "epoch": 0.77, "learning_rate": 8.961954677425437e-06, "loss": 1.0789, "step": 11400 }, { "epoch": 0.77, "learning_rate": 8.961012028204067e-06, "loss": 0.4848, "step": 11405 }, { "epoch": 0.77, "learning_rate": 8.960069378982694e-06, "loss": 0.5676, "step": 11410 }, { "epoch": 0.77, "learning_rate": 8.959126729761322e-06, "loss": 1.3169, "step": 11415 }, { "epoch": 0.78, "learning_rate": 8.95818408053995e-06, "loss": 0.885, "step": 11420 }, { "epoch": 0.78, "learning_rate": 8.957241431318579e-06, "loss": 0.1969, "step": 11425 }, { "epoch": 0.78, "learning_rate": 8.956298782097206e-06, "loss": 0.4178, "step": 11430 }, { "epoch": 0.78, "learning_rate": 8.955356132875836e-06, "loss": 1.5142, "step": 11435 }, { "epoch": 0.78, "learning_rate": 8.954413483654463e-06, "loss": 0.2937, "step": 11440 }, { "epoch": 0.78, "learning_rate": 8.953470834433091e-06, "loss": 0.8687, "step": 11445 }, { "epoch": 0.78, "learning_rate": 8.95252818521172e-06, "loss": 0.3019, "step": 11450 }, { "epoch": 0.78, "learning_rate": 8.951585535990348e-06, "loss": 0.8141, "step": 11455 }, { "epoch": 0.78, "learning_rate": 8.950642886768976e-06, "loss": 0.4221, "step": 11460 }, { "epoch": 0.78, "learning_rate": 8.949700237547605e-06, "loss": 0.2963, "step": 11465 }, { "epoch": 0.78, "learning_rate": 8.948757588326233e-06, "loss": 0.9262, "step": 11470 }, { "epoch": 0.78, "learning_rate": 8.94781493910486e-06, "loss": 0.9211, "step": 11475 }, { "epoch": 0.78, "learning_rate": 8.94687228988349e-06, "loss": 0.9083, "step": 11480 }, { "epoch": 0.78, "learning_rate": 8.945929640662117e-06, "loss": 0.4952, "step": 11485 }, { "epoch": 0.78, "learning_rate": 8.944986991440747e-06, "loss": 0.6914, "step": 11490 }, { "epoch": 0.78, "learning_rate": 8.944044342219374e-06, "loss": 1.0369, "step": 11495 }, { "epoch": 0.78, "learning_rate": 8.943101692998002e-06, "loss": 0.2747, "step": 11500 }, { "epoch": 0.78, "learning_rate": 8.94215904377663e-06, "loss": 0.443, "step": 11505 }, { "epoch": 0.78, "learning_rate": 8.941216394555259e-06, "loss": 0.6963, "step": 11510 }, { "epoch": 0.78, "learning_rate": 8.940273745333887e-06, "loss": 0.684, "step": 11515 }, { "epoch": 0.78, "learning_rate": 8.939331096112516e-06, "loss": 0.5379, "step": 11520 }, { "epoch": 0.78, "learning_rate": 8.938388446891144e-06, "loss": 0.3395, "step": 11525 }, { "epoch": 0.78, "learning_rate": 8.937445797669771e-06, "loss": 0.5962, "step": 11530 }, { "epoch": 0.78, "learning_rate": 8.9365031484484e-06, "loss": 0.6924, "step": 11535 }, { "epoch": 0.78, "learning_rate": 8.935560499227028e-06, "loss": 0.6077, "step": 11540 }, { "epoch": 0.78, "learning_rate": 8.934617850005657e-06, "loss": 0.5359, "step": 11545 }, { "epoch": 0.78, "learning_rate": 8.933675200784285e-06, "loss": 0.5502, "step": 11550 }, { "epoch": 0.78, "learning_rate": 8.932732551562913e-06, "loss": 0.9623, "step": 11555 }, { "epoch": 0.78, "learning_rate": 8.93178990234154e-06, "loss": 0.4609, "step": 11560 }, { "epoch": 0.78, "learning_rate": 8.93084725312017e-06, "loss": 0.7225, "step": 11565 }, { "epoch": 0.79, "learning_rate": 8.929904603898797e-06, "loss": 0.4541, "step": 11570 }, { "epoch": 0.79, "learning_rate": 8.928961954677427e-06, "loss": 1.5065, "step": 11575 }, { "epoch": 0.79, "learning_rate": 8.928019305456054e-06, "loss": 0.7204, "step": 11580 }, { "epoch": 0.79, "learning_rate": 8.927076656234682e-06, "loss": 0.4522, "step": 11585 }, { "epoch": 0.79, "learning_rate": 8.926134007013311e-06, "loss": 0.6097, "step": 11590 }, { "epoch": 0.79, "learning_rate": 8.925191357791939e-06, "loss": 0.6333, "step": 11595 }, { "epoch": 0.79, "learning_rate": 8.924248708570568e-06, "loss": 0.6544, "step": 11600 }, { "epoch": 0.79, "learning_rate": 8.923306059349196e-06, "loss": 0.8849, "step": 11605 }, { "epoch": 0.79, "learning_rate": 8.922363410127825e-06, "loss": 0.784, "step": 11610 }, { "epoch": 0.79, "learning_rate": 8.921420760906451e-06, "loss": 0.7346, "step": 11615 }, { "epoch": 0.79, "learning_rate": 8.92047811168508e-06, "loss": 0.5805, "step": 11620 }, { "epoch": 0.79, "learning_rate": 8.919535462463708e-06, "loss": 0.9022, "step": 11625 }, { "epoch": 0.79, "learning_rate": 8.918592813242338e-06, "loss": 0.6335, "step": 11630 }, { "epoch": 0.79, "learning_rate": 8.917650164020965e-06, "loss": 1.4028, "step": 11635 }, { "epoch": 0.79, "learning_rate": 8.916707514799593e-06, "loss": 0.3359, "step": 11640 }, { "epoch": 0.79, "learning_rate": 8.915764865578222e-06, "loss": 0.4342, "step": 11645 }, { "epoch": 0.79, "learning_rate": 8.91482221635685e-06, "loss": 0.4769, "step": 11650 }, { "epoch": 0.79, "learning_rate": 8.913879567135479e-06, "loss": 0.6378, "step": 11655 }, { "epoch": 0.79, "learning_rate": 8.912936917914107e-06, "loss": 0.9402, "step": 11660 }, { "epoch": 0.79, "learning_rate": 8.911994268692734e-06, "loss": 0.3705, "step": 11665 }, { "epoch": 0.79, "learning_rate": 8.911051619471362e-06, "loss": 0.7259, "step": 11670 }, { "epoch": 0.79, "learning_rate": 8.910108970249991e-06, "loss": 0.7848, "step": 11675 }, { "epoch": 0.79, "learning_rate": 8.909166321028619e-06, "loss": 0.9481, "step": 11680 }, { "epoch": 0.79, "learning_rate": 8.908223671807248e-06, "loss": 0.7351, "step": 11685 }, { "epoch": 0.79, "learning_rate": 8.907281022585876e-06, "loss": 0.5202, "step": 11690 }, { "epoch": 0.79, "learning_rate": 8.906338373364505e-06, "loss": 0.5663, "step": 11695 }, { "epoch": 0.79, "learning_rate": 8.905395724143133e-06, "loss": 0.7489, "step": 11700 }, { "epoch": 0.79, "learning_rate": 8.90445307492176e-06, "loss": 0.465, "step": 11705 }, { "epoch": 0.79, "learning_rate": 8.903510425700388e-06, "loss": 0.5752, "step": 11710 }, { "epoch": 0.8, "learning_rate": 8.902567776479018e-06, "loss": 0.8985, "step": 11715 }, { "epoch": 0.8, "learning_rate": 8.901625127257645e-06, "loss": 0.358, "step": 11720 }, { "epoch": 0.8, "learning_rate": 8.900682478036275e-06, "loss": 0.6622, "step": 11725 }, { "epoch": 0.8, "learning_rate": 8.899739828814902e-06, "loss": 0.7712, "step": 11730 }, { "epoch": 0.8, "learning_rate": 8.89879717959353e-06, "loss": 0.5845, "step": 11735 }, { "epoch": 0.8, "learning_rate": 8.89785453037216e-06, "loss": 1.2897, "step": 11740 }, { "epoch": 0.8, "learning_rate": 8.896911881150787e-06, "loss": 0.9597, "step": 11745 }, { "epoch": 0.8, "learning_rate": 8.895969231929416e-06, "loss": 0.5038, "step": 11750 }, { "epoch": 0.8, "learning_rate": 8.895026582708042e-06, "loss": 0.7093, "step": 11755 }, { "epoch": 0.8, "learning_rate": 8.894083933486671e-06, "loss": 0.2552, "step": 11760 }, { "epoch": 0.8, "learning_rate": 8.893141284265299e-06, "loss": 0.8362, "step": 11765 }, { "epoch": 0.8, "learning_rate": 8.892198635043928e-06, "loss": 0.3254, "step": 11770 }, { "epoch": 0.8, "learning_rate": 8.891255985822556e-06, "loss": 1.0243, "step": 11775 }, { "epoch": 0.8, "learning_rate": 8.890313336601185e-06, "loss": 0.4572, "step": 11780 }, { "epoch": 0.8, "learning_rate": 8.889370687379813e-06, "loss": 1.2924, "step": 11785 }, { "epoch": 0.8, "learning_rate": 8.88842803815844e-06, "loss": 0.4856, "step": 11790 }, { "epoch": 0.8, "learning_rate": 8.88748538893707e-06, "loss": 0.4193, "step": 11795 }, { "epoch": 0.8, "learning_rate": 8.886542739715698e-06, "loss": 0.8603, "step": 11800 }, { "epoch": 0.8, "learning_rate": 8.885600090494327e-06, "loss": 0.7926, "step": 11805 }, { "epoch": 0.8, "learning_rate": 8.884657441272955e-06, "loss": 1.2419, "step": 11810 }, { "epoch": 0.8, "learning_rate": 8.883714792051582e-06, "loss": 0.8258, "step": 11815 }, { "epoch": 0.8, "learning_rate": 8.88277214283021e-06, "loss": 0.5828, "step": 11820 }, { "epoch": 0.8, "learning_rate": 8.88182949360884e-06, "loss": 0.591, "step": 11825 }, { "epoch": 0.8, "learning_rate": 8.880886844387467e-06, "loss": 0.7686, "step": 11830 }, { "epoch": 0.8, "learning_rate": 8.879944195166096e-06, "loss": 0.4003, "step": 11835 }, { "epoch": 0.8, "learning_rate": 8.879001545944724e-06, "loss": 0.7593, "step": 11840 }, { "epoch": 0.8, "learning_rate": 8.878058896723351e-06, "loss": 1.1583, "step": 11845 }, { "epoch": 0.8, "learning_rate": 8.87711624750198e-06, "loss": 0.3327, "step": 11850 }, { "epoch": 0.8, "learning_rate": 8.876173598280608e-06, "loss": 1.4757, "step": 11855 }, { "epoch": 0.8, "learning_rate": 8.875230949059238e-06, "loss": 0.3094, "step": 11860 }, { "epoch": 0.81, "learning_rate": 8.874288299837865e-06, "loss": 0.6283, "step": 11865 }, { "epoch": 0.81, "learning_rate": 8.873345650616493e-06, "loss": 0.4245, "step": 11870 }, { "epoch": 0.81, "learning_rate": 8.87240300139512e-06, "loss": 0.8895, "step": 11875 }, { "epoch": 0.81, "learning_rate": 8.87146035217375e-06, "loss": 1.3122, "step": 11880 }, { "epoch": 0.81, "learning_rate": 8.870517702952378e-06, "loss": 0.5076, "step": 11885 }, { "epoch": 0.81, "learning_rate": 8.869575053731007e-06, "loss": 0.6082, "step": 11890 }, { "epoch": 0.81, "learning_rate": 8.868632404509635e-06, "loss": 0.4662, "step": 11895 }, { "epoch": 0.81, "learning_rate": 8.867689755288262e-06, "loss": 0.5031, "step": 11900 }, { "epoch": 0.81, "learning_rate": 8.866747106066892e-06, "loss": 0.2811, "step": 11905 }, { "epoch": 0.81, "learning_rate": 8.86580445684552e-06, "loss": 0.9105, "step": 11910 }, { "epoch": 0.81, "learning_rate": 8.864861807624147e-06, "loss": 0.4693, "step": 11915 }, { "epoch": 0.81, "learning_rate": 8.863919158402776e-06, "loss": 0.3719, "step": 11920 }, { "epoch": 0.81, "learning_rate": 8.862976509181404e-06, "loss": 0.5125, "step": 11925 }, { "epoch": 0.81, "learning_rate": 8.862033859960032e-06, "loss": 1.1143, "step": 11930 }, { "epoch": 0.81, "learning_rate": 8.86109121073866e-06, "loss": 0.5331, "step": 11935 }, { "epoch": 0.81, "learning_rate": 8.860148561517288e-06, "loss": 0.548, "step": 11940 }, { "epoch": 0.81, "learning_rate": 8.859205912295918e-06, "loss": 0.5306, "step": 11945 }, { "epoch": 0.81, "learning_rate": 8.858263263074545e-06, "loss": 0.5504, "step": 11950 }, { "epoch": 0.81, "learning_rate": 8.857320613853175e-06, "loss": 0.9018, "step": 11955 }, { "epoch": 0.81, "learning_rate": 8.8563779646318e-06, "loss": 0.7444, "step": 11960 }, { "epoch": 0.81, "learning_rate": 8.85543531541043e-06, "loss": 0.8821, "step": 11965 }, { "epoch": 0.81, "learning_rate": 8.854492666189058e-06, "loss": 0.7041, "step": 11970 }, { "epoch": 0.81, "learning_rate": 8.853550016967687e-06, "loss": 0.9485, "step": 11975 }, { "epoch": 0.81, "learning_rate": 8.852607367746315e-06, "loss": 0.86, "step": 11980 }, { "epoch": 0.81, "learning_rate": 8.851664718524942e-06, "loss": 0.2254, "step": 11985 }, { "epoch": 0.81, "learning_rate": 8.850722069303572e-06, "loss": 0.4916, "step": 11990 }, { "epoch": 0.81, "learning_rate": 8.8497794200822e-06, "loss": 0.8599, "step": 11995 }, { "epoch": 0.81, "learning_rate": 8.848836770860829e-06, "loss": 1.1857, "step": 12000 }, { "epoch": 0.81, "learning_rate": 8.847894121639456e-06, "loss": 0.7802, "step": 12005 }, { "epoch": 0.82, "learning_rate": 8.846951472418086e-06, "loss": 0.3433, "step": 12010 }, { "epoch": 0.82, "learning_rate": 8.846008823196712e-06, "loss": 0.3594, "step": 12015 }, { "epoch": 0.82, "learning_rate": 8.845066173975341e-06, "loss": 0.4222, "step": 12020 }, { "epoch": 0.82, "learning_rate": 8.844123524753969e-06, "loss": 1.0927, "step": 12025 }, { "epoch": 0.82, "learning_rate": 8.843180875532598e-06, "loss": 0.1423, "step": 12030 }, { "epoch": 0.82, "learning_rate": 8.842238226311226e-06, "loss": 0.6786, "step": 12035 }, { "epoch": 0.82, "learning_rate": 8.841295577089855e-06, "loss": 0.535, "step": 12040 }, { "epoch": 0.82, "learning_rate": 8.840352927868482e-06, "loss": 1.5672, "step": 12045 }, { "epoch": 0.82, "learning_rate": 8.83941027864711e-06, "loss": 0.3825, "step": 12050 }, { "epoch": 0.82, "learning_rate": 8.83846762942574e-06, "loss": 0.2775, "step": 12055 }, { "epoch": 0.82, "learning_rate": 8.837524980204367e-06, "loss": 0.7624, "step": 12060 }, { "epoch": 0.82, "learning_rate": 8.836582330982996e-06, "loss": 0.6615, "step": 12065 }, { "epoch": 0.82, "learning_rate": 8.835639681761624e-06, "loss": 0.8048, "step": 12070 }, { "epoch": 0.82, "learning_rate": 8.834697032540252e-06, "loss": 0.0934, "step": 12075 }, { "epoch": 0.82, "learning_rate": 8.83375438331888e-06, "loss": 0.2578, "step": 12080 }, { "epoch": 0.82, "learning_rate": 8.832811734097509e-06, "loss": 0.8265, "step": 12085 }, { "epoch": 0.82, "learning_rate": 8.831869084876136e-06, "loss": 0.5603, "step": 12090 }, { "epoch": 0.82, "learning_rate": 8.830926435654766e-06, "loss": 0.7058, "step": 12095 }, { "epoch": 0.82, "learning_rate": 8.829983786433393e-06, "loss": 1.0014, "step": 12100 }, { "epoch": 0.82, "learning_rate": 8.829041137212021e-06, "loss": 0.4932, "step": 12105 }, { "epoch": 0.82, "learning_rate": 8.82809848799065e-06, "loss": 0.6496, "step": 12110 }, { "epoch": 0.82, "learning_rate": 8.827155838769278e-06, "loss": 0.1943, "step": 12115 }, { "epoch": 0.82, "learning_rate": 8.826213189547906e-06, "loss": 0.7896, "step": 12120 }, { "epoch": 0.82, "learning_rate": 8.825270540326535e-06, "loss": 0.729, "step": 12125 }, { "epoch": 0.82, "learning_rate": 8.824327891105163e-06, "loss": 0.182, "step": 12130 }, { "epoch": 0.82, "learning_rate": 8.82338524188379e-06, "loss": 0.6827, "step": 12135 }, { "epoch": 0.82, "learning_rate": 8.82244259266242e-06, "loss": 0.4801, "step": 12140 }, { "epoch": 0.82, "learning_rate": 8.821499943441047e-06, "loss": 0.5476, "step": 12145 }, { "epoch": 0.82, "learning_rate": 8.820557294219676e-06, "loss": 1.0469, "step": 12150 }, { "epoch": 0.82, "learning_rate": 8.819614644998304e-06, "loss": 0.3187, "step": 12155 }, { "epoch": 0.83, "learning_rate": 8.818671995776932e-06, "loss": 0.7925, "step": 12160 }, { "epoch": 0.83, "learning_rate": 8.81772934655556e-06, "loss": 0.269, "step": 12165 }, { "epoch": 0.83, "learning_rate": 8.816786697334189e-06, "loss": 0.8216, "step": 12170 }, { "epoch": 0.83, "learning_rate": 8.815844048112816e-06, "loss": 0.5324, "step": 12175 }, { "epoch": 0.83, "learning_rate": 8.814901398891446e-06, "loss": 0.561, "step": 12180 }, { "epoch": 0.83, "learning_rate": 8.813958749670073e-06, "loss": 0.7164, "step": 12185 }, { "epoch": 0.83, "learning_rate": 8.813016100448701e-06, "loss": 1.1727, "step": 12190 }, { "epoch": 0.83, "learning_rate": 8.81207345122733e-06, "loss": 0.3075, "step": 12195 }, { "epoch": 0.83, "learning_rate": 8.811130802005958e-06, "loss": 0.3828, "step": 12200 }, { "epoch": 0.83, "learning_rate": 8.810188152784587e-06, "loss": 0.6689, "step": 12205 }, { "epoch": 0.83, "learning_rate": 8.809245503563215e-06, "loss": 0.3874, "step": 12210 }, { "epoch": 0.83, "learning_rate": 8.808302854341843e-06, "loss": 0.9715, "step": 12215 }, { "epoch": 0.83, "learning_rate": 8.80736020512047e-06, "loss": 0.7573, "step": 12220 }, { "epoch": 0.83, "learning_rate": 8.8064175558991e-06, "loss": 0.6881, "step": 12225 }, { "epoch": 0.83, "learning_rate": 8.805474906677727e-06, "loss": 0.4806, "step": 12230 }, { "epoch": 0.83, "learning_rate": 8.804532257456357e-06, "loss": 0.7377, "step": 12235 }, { "epoch": 0.83, "learning_rate": 8.803589608234984e-06, "loss": 0.4991, "step": 12240 }, { "epoch": 0.83, "learning_rate": 8.802646959013612e-06, "loss": 0.5667, "step": 12245 }, { "epoch": 0.83, "learning_rate": 8.801704309792241e-06, "loss": 1.0571, "step": 12250 }, { "epoch": 0.83, "learning_rate": 8.800761660570869e-06, "loss": 0.5946, "step": 12255 }, { "epoch": 0.83, "learning_rate": 8.799819011349498e-06, "loss": 0.5101, "step": 12260 }, { "epoch": 0.83, "learning_rate": 8.798876362128126e-06, "loss": 0.531, "step": 12265 }, { "epoch": 0.83, "learning_rate": 8.797933712906755e-06, "loss": 0.5919, "step": 12270 }, { "epoch": 0.83, "learning_rate": 8.796991063685381e-06, "loss": 0.5255, "step": 12275 }, { "epoch": 0.83, "learning_rate": 8.79604841446401e-06, "loss": 0.6581, "step": 12280 }, { "epoch": 0.83, "learning_rate": 8.795105765242638e-06, "loss": 1.0993, "step": 12285 }, { "epoch": 0.83, "learning_rate": 8.794163116021267e-06, "loss": 1.007, "step": 12290 }, { "epoch": 0.83, "learning_rate": 8.793220466799895e-06, "loss": 0.6195, "step": 12295 }, { "epoch": 0.83, "learning_rate": 8.792277817578524e-06, "loss": 1.6434, "step": 12300 }, { "epoch": 0.84, "learning_rate": 8.791335168357152e-06, "loss": 1.2636, "step": 12305 }, { "epoch": 0.84, "learning_rate": 8.79039251913578e-06, "loss": 0.2217, "step": 12310 }, { "epoch": 0.84, "learning_rate": 8.789449869914409e-06, "loss": 0.7188, "step": 12315 }, { "epoch": 0.84, "learning_rate": 8.788507220693037e-06, "loss": 0.8698, "step": 12320 }, { "epoch": 0.84, "learning_rate": 8.787564571471666e-06, "loss": 0.5007, "step": 12325 }, { "epoch": 0.84, "learning_rate": 8.786621922250292e-06, "loss": 0.2634, "step": 12330 }, { "epoch": 0.84, "learning_rate": 8.785679273028921e-06, "loss": 0.6587, "step": 12335 }, { "epoch": 0.84, "learning_rate": 8.784736623807549e-06, "loss": 0.64, "step": 12340 }, { "epoch": 0.84, "learning_rate": 8.783793974586178e-06, "loss": 1.2306, "step": 12345 }, { "epoch": 0.84, "learning_rate": 8.782851325364806e-06, "loss": 0.6529, "step": 12350 }, { "epoch": 0.84, "learning_rate": 8.781908676143435e-06, "loss": 0.5158, "step": 12355 }, { "epoch": 0.84, "learning_rate": 8.780966026922063e-06, "loss": 0.8235, "step": 12360 }, { "epoch": 0.84, "learning_rate": 8.78002337770069e-06, "loss": 0.5935, "step": 12365 }, { "epoch": 0.84, "learning_rate": 8.77908072847932e-06, "loss": 0.4927, "step": 12370 }, { "epoch": 0.84, "learning_rate": 8.778138079257947e-06, "loss": 0.9768, "step": 12375 }, { "epoch": 0.84, "learning_rate": 8.777195430036575e-06, "loss": 0.9048, "step": 12380 }, { "epoch": 0.84, "learning_rate": 8.776252780815204e-06, "loss": 0.2555, "step": 12385 }, { "epoch": 0.84, "learning_rate": 8.775310131593832e-06, "loss": 0.8429, "step": 12390 }, { "epoch": 0.84, "learning_rate": 8.77436748237246e-06, "loss": 0.6406, "step": 12395 }, { "epoch": 0.84, "learning_rate": 8.773424833151089e-06, "loss": 0.3018, "step": 12400 }, { "epoch": 0.84, "learning_rate": 8.772482183929717e-06, "loss": 1.0924, "step": 12405 }, { "epoch": 0.84, "learning_rate": 8.771539534708346e-06, "loss": 0.7848, "step": 12410 }, { "epoch": 0.84, "learning_rate": 8.770596885486974e-06, "loss": 0.2075, "step": 12415 }, { "epoch": 0.84, "learning_rate": 8.769654236265601e-06, "loss": 0.4082, "step": 12420 }, { "epoch": 0.84, "learning_rate": 8.768711587044229e-06, "loss": 0.4279, "step": 12425 }, { "epoch": 0.84, "learning_rate": 8.767768937822858e-06, "loss": 0.468, "step": 12430 }, { "epoch": 0.84, "learning_rate": 8.766826288601486e-06, "loss": 1.0148, "step": 12435 }, { "epoch": 0.84, "learning_rate": 8.765883639380115e-06, "loss": 1.1391, "step": 12440 }, { "epoch": 0.84, "learning_rate": 8.764940990158743e-06, "loss": 0.5873, "step": 12445 }, { "epoch": 0.84, "learning_rate": 8.76399834093737e-06, "loss": 0.8432, "step": 12450 }, { "epoch": 0.85, "learning_rate": 8.763055691716e-06, "loss": 0.6549, "step": 12455 }, { "epoch": 0.85, "learning_rate": 8.762113042494627e-06, "loss": 0.5306, "step": 12460 }, { "epoch": 0.85, "learning_rate": 8.761170393273257e-06, "loss": 0.7328, "step": 12465 }, { "epoch": 0.85, "learning_rate": 8.760227744051884e-06, "loss": 0.8202, "step": 12470 }, { "epoch": 0.85, "learning_rate": 8.759285094830512e-06, "loss": 0.4419, "step": 12475 }, { "epoch": 0.85, "learning_rate": 8.75834244560914e-06, "loss": 0.1326, "step": 12480 }, { "epoch": 0.85, "learning_rate": 8.757399796387769e-06, "loss": 0.1445, "step": 12485 }, { "epoch": 0.85, "learning_rate": 8.756457147166397e-06, "loss": 0.2262, "step": 12490 }, { "epoch": 0.85, "learning_rate": 8.755514497945026e-06, "loss": 1.0146, "step": 12495 }, { "epoch": 0.85, "learning_rate": 8.754571848723654e-06, "loss": 0.5099, "step": 12500 }, { "epoch": 0.85, "learning_rate": 8.753629199502281e-06, "loss": 0.9468, "step": 12505 }, { "epoch": 0.85, "learning_rate": 8.75268655028091e-06, "loss": 0.2625, "step": 12510 }, { "epoch": 0.85, "learning_rate": 8.751743901059538e-06, "loss": 0.3564, "step": 12515 }, { "epoch": 0.85, "learning_rate": 8.750801251838168e-06, "loss": 1.1402, "step": 12520 }, { "epoch": 0.85, "learning_rate": 8.749858602616795e-06, "loss": 0.7567, "step": 12525 }, { "epoch": 0.85, "learning_rate": 8.748915953395423e-06, "loss": 1.2731, "step": 12530 }, { "epoch": 0.85, "learning_rate": 8.74797330417405e-06, "loss": 0.5927, "step": 12535 }, { "epoch": 0.85, "learning_rate": 8.74703065495268e-06, "loss": 0.6256, "step": 12540 }, { "epoch": 0.85, "learning_rate": 8.746088005731307e-06, "loss": 0.1906, "step": 12545 }, { "epoch": 0.85, "learning_rate": 8.745145356509937e-06, "loss": 0.7068, "step": 12550 }, { "epoch": 0.85, "learning_rate": 8.744202707288564e-06, "loss": 0.287, "step": 12555 }, { "epoch": 0.85, "learning_rate": 8.743260058067192e-06, "loss": 0.7423, "step": 12560 }, { "epoch": 0.85, "learning_rate": 8.742317408845821e-06, "loss": 0.9762, "step": 12565 }, { "epoch": 0.85, "learning_rate": 8.741374759624449e-06, "loss": 0.2808, "step": 12570 }, { "epoch": 0.85, "learning_rate": 8.740432110403078e-06, "loss": 0.7672, "step": 12575 }, { "epoch": 0.85, "learning_rate": 8.739489461181706e-06, "loss": 0.4524, "step": 12580 }, { "epoch": 0.85, "learning_rate": 8.738546811960334e-06, "loss": 0.713, "step": 12585 }, { "epoch": 0.85, "learning_rate": 8.737604162738961e-06, "loss": 0.8783, "step": 12590 }, { "epoch": 0.85, "learning_rate": 8.73666151351759e-06, "loss": 0.4818, "step": 12595 }, { "epoch": 0.86, "learning_rate": 8.735718864296218e-06, "loss": 0.6013, "step": 12600 }, { "epoch": 0.86, "learning_rate": 8.734776215074848e-06, "loss": 0.9867, "step": 12605 }, { "epoch": 0.86, "learning_rate": 8.733833565853475e-06, "loss": 0.5315, "step": 12610 }, { "epoch": 0.86, "learning_rate": 8.732890916632105e-06, "loss": 0.6156, "step": 12615 }, { "epoch": 0.86, "learning_rate": 8.731948267410732e-06, "loss": 0.4219, "step": 12620 }, { "epoch": 0.86, "learning_rate": 8.73100561818936e-06, "loss": 0.4141, "step": 12625 }, { "epoch": 0.86, "learning_rate": 8.730062968967988e-06, "loss": 0.651, "step": 12630 }, { "epoch": 0.86, "learning_rate": 8.729120319746617e-06, "loss": 0.3558, "step": 12635 }, { "epoch": 0.86, "learning_rate": 8.728177670525245e-06, "loss": 0.5185, "step": 12640 }, { "epoch": 0.86, "learning_rate": 8.727235021303872e-06, "loss": 0.9598, "step": 12645 }, { "epoch": 0.86, "learning_rate": 8.726292372082501e-06, "loss": 0.6358, "step": 12650 }, { "epoch": 0.86, "learning_rate": 8.725349722861129e-06, "loss": 0.3973, "step": 12655 }, { "epoch": 0.86, "learning_rate": 8.724407073639758e-06, "loss": 0.5321, "step": 12660 }, { "epoch": 0.86, "learning_rate": 8.723464424418386e-06, "loss": 1.069, "step": 12665 }, { "epoch": 0.86, "learning_rate": 8.722521775197015e-06, "loss": 0.5828, "step": 12670 }, { "epoch": 0.86, "learning_rate": 8.721579125975641e-06, "loss": 0.9287, "step": 12675 }, { "epoch": 0.86, "learning_rate": 8.72063647675427e-06, "loss": 0.7377, "step": 12680 }, { "epoch": 0.86, "learning_rate": 8.719693827532898e-06, "loss": 0.3747, "step": 12685 }, { "epoch": 0.86, "learning_rate": 8.718751178311528e-06, "loss": 0.6276, "step": 12690 }, { "epoch": 0.86, "learning_rate": 8.717808529090155e-06, "loss": 0.5886, "step": 12695 }, { "epoch": 0.86, "learning_rate": 8.716865879868785e-06, "loss": 0.6549, "step": 12700 }, { "epoch": 0.86, "learning_rate": 8.715923230647412e-06, "loss": 0.4993, "step": 12705 }, { "epoch": 0.86, "learning_rate": 8.71498058142604e-06, "loss": 0.638, "step": 12710 }, { "epoch": 0.86, "learning_rate": 8.71403793220467e-06, "loss": 0.7208, "step": 12715 }, { "epoch": 0.86, "learning_rate": 8.713095282983297e-06, "loss": 0.4029, "step": 12720 }, { "epoch": 0.86, "learning_rate": 8.712152633761926e-06, "loss": 0.7109, "step": 12725 }, { "epoch": 0.86, "learning_rate": 8.711209984540554e-06, "loss": 0.3882, "step": 12730 }, { "epoch": 0.86, "learning_rate": 8.710267335319182e-06, "loss": 0.3254, "step": 12735 }, { "epoch": 0.86, "learning_rate": 8.70932468609781e-06, "loss": 0.8185, "step": 12740 }, { "epoch": 0.87, "learning_rate": 8.708382036876439e-06, "loss": 0.8616, "step": 12745 }, { "epoch": 0.87, "learning_rate": 8.707439387655066e-06, "loss": 1.7938, "step": 12750 }, { "epoch": 0.87, "learning_rate": 8.706496738433695e-06, "loss": 0.2792, "step": 12755 }, { "epoch": 0.87, "learning_rate": 8.705554089212323e-06, "loss": 1.3503, "step": 12760 }, { "epoch": 0.87, "learning_rate": 8.70461143999095e-06, "loss": 0.7114, "step": 12765 }, { "epoch": 0.87, "learning_rate": 8.70366879076958e-06, "loss": 1.1214, "step": 12770 }, { "epoch": 0.87, "learning_rate": 8.702726141548208e-06, "loss": 0.83, "step": 12775 }, { "epoch": 0.87, "learning_rate": 8.701783492326837e-06, "loss": 0.4601, "step": 12780 }, { "epoch": 0.87, "learning_rate": 8.700840843105465e-06, "loss": 0.8552, "step": 12785 }, { "epoch": 0.87, "learning_rate": 8.699898193884092e-06, "loss": 0.8497, "step": 12790 }, { "epoch": 0.87, "learning_rate": 8.69895554466272e-06, "loss": 0.9924, "step": 12795 }, { "epoch": 0.87, "learning_rate": 8.69801289544135e-06, "loss": 0.7504, "step": 12800 }, { "epoch": 0.87, "learning_rate": 8.697070246219977e-06, "loss": 0.7532, "step": 12805 }, { "epoch": 0.87, "learning_rate": 8.696127596998606e-06, "loss": 0.8307, "step": 12810 }, { "epoch": 0.87, "learning_rate": 8.695184947777234e-06, "loss": 1.0456, "step": 12815 }, { "epoch": 0.87, "learning_rate": 8.694242298555862e-06, "loss": 0.3038, "step": 12820 }, { "epoch": 0.87, "learning_rate": 8.693299649334491e-06, "loss": 0.4999, "step": 12825 }, { "epoch": 0.87, "learning_rate": 8.692357000113119e-06, "loss": 0.4575, "step": 12830 }, { "epoch": 0.87, "learning_rate": 8.691414350891746e-06, "loss": 0.5352, "step": 12835 }, { "epoch": 0.87, "learning_rate": 8.690471701670376e-06, "loss": 0.518, "step": 12840 }, { "epoch": 0.87, "learning_rate": 8.689529052449003e-06, "loss": 0.5217, "step": 12845 }, { "epoch": 0.87, "learning_rate": 8.68858640322763e-06, "loss": 0.5917, "step": 12850 }, { "epoch": 0.87, "learning_rate": 8.68764375400626e-06, "loss": 0.3823, "step": 12855 }, { "epoch": 0.87, "learning_rate": 8.686701104784888e-06, "loss": 0.5344, "step": 12860 }, { "epoch": 0.87, "learning_rate": 8.685758455563517e-06, "loss": 0.4248, "step": 12865 }, { "epoch": 0.87, "learning_rate": 8.684815806342145e-06, "loss": 0.823, "step": 12870 }, { "epoch": 0.87, "learning_rate": 8.683873157120772e-06, "loss": 0.4562, "step": 12875 }, { "epoch": 0.87, "learning_rate": 8.6829305078994e-06, "loss": 0.5191, "step": 12880 }, { "epoch": 0.87, "learning_rate": 8.68198785867803e-06, "loss": 1.1357, "step": 12885 }, { "epoch": 0.87, "learning_rate": 8.681045209456657e-06, "loss": 0.6027, "step": 12890 }, { "epoch": 0.88, "learning_rate": 8.680102560235286e-06, "loss": 0.3835, "step": 12895 }, { "epoch": 0.88, "learning_rate": 8.679159911013914e-06, "loss": 1.1907, "step": 12900 }, { "epoch": 0.88, "learning_rate": 8.678217261792542e-06, "loss": 0.7276, "step": 12905 }, { "epoch": 0.88, "learning_rate": 8.677274612571171e-06, "loss": 0.3138, "step": 12910 }, { "epoch": 0.88, "learning_rate": 8.676331963349799e-06, "loss": 0.6754, "step": 12915 }, { "epoch": 0.88, "learning_rate": 8.675389314128428e-06, "loss": 0.7514, "step": 12920 }, { "epoch": 0.88, "learning_rate": 8.674446664907056e-06, "loss": 0.3541, "step": 12925 }, { "epoch": 0.88, "learning_rate": 8.673504015685685e-06, "loss": 1.4923, "step": 12930 }, { "epoch": 0.88, "learning_rate": 8.672561366464311e-06, "loss": 0.6007, "step": 12935 }, { "epoch": 0.88, "learning_rate": 8.67161871724294e-06, "loss": 0.8905, "step": 12940 }, { "epoch": 0.88, "learning_rate": 8.670676068021568e-06, "loss": 0.6574, "step": 12945 }, { "epoch": 0.88, "learning_rate": 8.669733418800197e-06, "loss": 0.2893, "step": 12950 }, { "epoch": 0.88, "learning_rate": 8.668790769578825e-06, "loss": 0.9948, "step": 12955 }, { "epoch": 0.88, "learning_rate": 8.667848120357454e-06, "loss": 0.7909, "step": 12960 }, { "epoch": 0.88, "learning_rate": 8.666905471136082e-06, "loss": 0.1325, "step": 12965 }, { "epoch": 0.88, "learning_rate": 8.66596282191471e-06, "loss": 0.5675, "step": 12970 }, { "epoch": 0.88, "learning_rate": 8.665020172693339e-06, "loss": 1.1716, "step": 12975 }, { "epoch": 0.88, "learning_rate": 8.664077523471966e-06, "loss": 1.2745, "step": 12980 }, { "epoch": 0.88, "learning_rate": 8.663134874250596e-06, "loss": 0.7446, "step": 12985 }, { "epoch": 0.88, "learning_rate": 8.662192225029222e-06, "loss": 0.2367, "step": 12990 }, { "epoch": 0.88, "learning_rate": 8.661249575807851e-06, "loss": 1.3518, "step": 12995 }, { "epoch": 0.88, "learning_rate": 8.660306926586479e-06, "loss": 0.6423, "step": 13000 }, { "epoch": 0.88, "learning_rate": 8.659364277365108e-06, "loss": 0.5429, "step": 13005 }, { "epoch": 0.88, "learning_rate": 8.658421628143736e-06, "loss": 1.1396, "step": 13010 }, { "epoch": 0.88, "learning_rate": 8.657478978922365e-06, "loss": 0.2504, "step": 13015 }, { "epoch": 0.88, "learning_rate": 8.656536329700993e-06, "loss": 0.441, "step": 13020 }, { "epoch": 0.88, "learning_rate": 8.65559368047962e-06, "loss": 0.7039, "step": 13025 }, { "epoch": 0.88, "learning_rate": 8.65465103125825e-06, "loss": 0.2988, "step": 13030 }, { "epoch": 0.88, "learning_rate": 8.653708382036877e-06, "loss": 0.8877, "step": 13035 }, { "epoch": 0.89, "learning_rate": 8.652765732815505e-06, "loss": 0.5089, "step": 13040 }, { "epoch": 0.89, "learning_rate": 8.651823083594134e-06, "loss": 1.2743, "step": 13045 }, { "epoch": 0.89, "learning_rate": 8.650880434372762e-06, "loss": 0.956, "step": 13050 }, { "epoch": 0.89, "learning_rate": 8.64993778515139e-06, "loss": 0.8431, "step": 13055 }, { "epoch": 0.89, "learning_rate": 8.648995135930019e-06, "loss": 0.2818, "step": 13060 }, { "epoch": 0.89, "learning_rate": 8.648052486708646e-06, "loss": 0.8689, "step": 13065 }, { "epoch": 0.89, "learning_rate": 8.647109837487276e-06, "loss": 0.8936, "step": 13070 }, { "epoch": 0.89, "learning_rate": 8.646167188265903e-06, "loss": 0.8616, "step": 13075 }, { "epoch": 0.89, "learning_rate": 8.645224539044531e-06, "loss": 1.0607, "step": 13080 }, { "epoch": 0.89, "learning_rate": 8.644281889823159e-06, "loss": 0.8964, "step": 13085 }, { "epoch": 0.89, "learning_rate": 8.643339240601788e-06, "loss": 0.3529, "step": 13090 }, { "epoch": 0.89, "learning_rate": 8.642396591380416e-06, "loss": 0.6915, "step": 13095 }, { "epoch": 0.89, "learning_rate": 8.641453942159045e-06, "loss": 0.7771, "step": 13100 }, { "epoch": 0.89, "learning_rate": 8.640511292937673e-06, "loss": 0.6895, "step": 13105 }, { "epoch": 0.89, "learning_rate": 8.6395686437163e-06, "loss": 0.4368, "step": 13110 }, { "epoch": 0.89, "learning_rate": 8.63862599449493e-06, "loss": 0.5227, "step": 13115 }, { "epoch": 0.89, "learning_rate": 8.637683345273557e-06, "loss": 0.2118, "step": 13120 }, { "epoch": 0.89, "learning_rate": 8.636740696052187e-06, "loss": 0.8611, "step": 13125 }, { "epoch": 0.89, "learning_rate": 8.635798046830814e-06, "loss": 0.3388, "step": 13130 }, { "epoch": 0.89, "learning_rate": 8.634855397609442e-06, "loss": 1.0737, "step": 13135 }, { "epoch": 0.89, "learning_rate": 8.63391274838807e-06, "loss": 0.2654, "step": 13140 }, { "epoch": 0.89, "learning_rate": 8.632970099166699e-06, "loss": 0.7499, "step": 13145 }, { "epoch": 0.89, "learning_rate": 8.632027449945327e-06, "loss": 0.6788, "step": 13150 }, { "epoch": 0.89, "learning_rate": 8.631084800723956e-06, "loss": 1.5847, "step": 13155 }, { "epoch": 0.89, "learning_rate": 8.630142151502583e-06, "loss": 0.9303, "step": 13160 }, { "epoch": 0.89, "learning_rate": 8.629199502281211e-06, "loss": 0.7496, "step": 13165 }, { "epoch": 0.89, "learning_rate": 8.62825685305984e-06, "loss": 0.8021, "step": 13170 }, { "epoch": 0.89, "learning_rate": 8.627314203838468e-06, "loss": 1.196, "step": 13175 }, { "epoch": 0.89, "learning_rate": 8.626371554617097e-06, "loss": 0.8192, "step": 13180 }, { "epoch": 0.89, "learning_rate": 8.625428905395725e-06, "loss": 0.6608, "step": 13185 }, { "epoch": 0.9, "learning_rate": 8.624486256174354e-06, "loss": 0.7831, "step": 13190 }, { "epoch": 0.9, "learning_rate": 8.62354360695298e-06, "loss": 0.4969, "step": 13195 }, { "epoch": 0.9, "learning_rate": 8.62260095773161e-06, "loss": 0.3853, "step": 13200 }, { "epoch": 0.9, "learning_rate": 8.621658308510237e-06, "loss": 0.7112, "step": 13205 }, { "epoch": 0.9, "learning_rate": 8.620715659288867e-06, "loss": 0.6993, "step": 13210 }, { "epoch": 0.9, "learning_rate": 8.619773010067494e-06, "loss": 0.4271, "step": 13215 }, { "epoch": 0.9, "learning_rate": 8.618830360846122e-06, "loss": 0.877, "step": 13220 }, { "epoch": 0.9, "learning_rate": 8.617887711624751e-06, "loss": 0.2456, "step": 13225 }, { "epoch": 0.9, "learning_rate": 8.616945062403379e-06, "loss": 1.2524, "step": 13230 }, { "epoch": 0.9, "learning_rate": 8.616002413182008e-06, "loss": 0.9597, "step": 13235 }, { "epoch": 0.9, "learning_rate": 8.615059763960636e-06, "loss": 0.824, "step": 13240 }, { "epoch": 0.9, "learning_rate": 8.614117114739264e-06, "loss": 0.9993, "step": 13245 }, { "epoch": 0.9, "learning_rate": 8.613174465517891e-06, "loss": 0.3629, "step": 13250 }, { "epoch": 0.9, "learning_rate": 8.61223181629652e-06, "loss": 0.6324, "step": 13255 }, { "epoch": 0.9, "learning_rate": 8.611289167075148e-06, "loss": 0.8319, "step": 13260 }, { "epoch": 0.9, "learning_rate": 8.610346517853777e-06, "loss": 0.6856, "step": 13265 }, { "epoch": 0.9, "learning_rate": 8.609403868632405e-06, "loss": 0.9606, "step": 13270 }, { "epoch": 0.9, "learning_rate": 8.608461219411034e-06, "loss": 0.4922, "step": 13275 }, { "epoch": 0.9, "learning_rate": 8.607518570189662e-06, "loss": 0.793, "step": 13280 }, { "epoch": 0.9, "learning_rate": 8.60657592096829e-06, "loss": 1.2104, "step": 13285 }, { "epoch": 0.9, "learning_rate": 8.605633271746917e-06, "loss": 0.5897, "step": 13290 }, { "epoch": 0.9, "learning_rate": 8.604690622525547e-06, "loss": 0.8613, "step": 13295 }, { "epoch": 0.9, "learning_rate": 8.603747973304174e-06, "loss": 0.75, "step": 13300 }, { "epoch": 0.9, "learning_rate": 8.602805324082804e-06, "loss": 1.203, "step": 13305 }, { "epoch": 0.9, "learning_rate": 8.601862674861431e-06, "loss": 0.446, "step": 13310 }, { "epoch": 0.9, "learning_rate": 8.600920025640059e-06, "loss": 0.6578, "step": 13315 }, { "epoch": 0.9, "learning_rate": 8.599977376418688e-06, "loss": 1.9589, "step": 13320 }, { "epoch": 0.9, "learning_rate": 8.599034727197316e-06, "loss": 0.4437, "step": 13325 }, { "epoch": 0.9, "learning_rate": 8.598092077975945e-06, "loss": 0.7211, "step": 13330 }, { "epoch": 0.91, "learning_rate": 8.597149428754571e-06, "loss": 0.4813, "step": 13335 }, { "epoch": 0.91, "learning_rate": 8.5962067795332e-06, "loss": 0.3274, "step": 13340 }, { "epoch": 0.91, "learning_rate": 8.595264130311828e-06, "loss": 0.6894, "step": 13345 }, { "epoch": 0.91, "learning_rate": 8.594321481090458e-06, "loss": 0.8031, "step": 13350 }, { "epoch": 0.91, "learning_rate": 8.593378831869085e-06, "loss": 0.9878, "step": 13355 }, { "epoch": 0.91, "learning_rate": 8.592436182647715e-06, "loss": 0.8833, "step": 13360 }, { "epoch": 0.91, "learning_rate": 8.591493533426342e-06, "loss": 0.5497, "step": 13365 }, { "epoch": 0.91, "learning_rate": 8.59055088420497e-06, "loss": 0.4515, "step": 13370 }, { "epoch": 0.91, "learning_rate": 8.589608234983599e-06, "loss": 0.2323, "step": 13375 }, { "epoch": 0.91, "learning_rate": 8.588665585762227e-06, "loss": 0.3501, "step": 13380 }, { "epoch": 0.91, "learning_rate": 8.587722936540856e-06, "loss": 0.7372, "step": 13385 }, { "epoch": 0.91, "learning_rate": 8.586780287319484e-06, "loss": 0.571, "step": 13390 }, { "epoch": 0.91, "learning_rate": 8.585837638098111e-06, "loss": 0.9524, "step": 13395 }, { "epoch": 0.91, "learning_rate": 8.584894988876739e-06, "loss": 0.426, "step": 13400 }, { "epoch": 0.91, "learning_rate": 8.583952339655368e-06, "loss": 1.1622, "step": 13405 }, { "epoch": 0.91, "learning_rate": 8.583009690433996e-06, "loss": 0.9369, "step": 13410 }, { "epoch": 0.91, "learning_rate": 8.582067041212625e-06, "loss": 0.59, "step": 13415 }, { "epoch": 0.91, "learning_rate": 8.581124391991253e-06, "loss": 0.2353, "step": 13420 }, { "epoch": 0.91, "learning_rate": 8.58018174276988e-06, "loss": 1.0636, "step": 13425 }, { "epoch": 0.91, "learning_rate": 8.57923909354851e-06, "loss": 0.7893, "step": 13430 }, { "epoch": 0.91, "learning_rate": 8.578296444327138e-06, "loss": 0.4964, "step": 13435 }, { "epoch": 0.91, "learning_rate": 8.577353795105767e-06, "loss": 0.1668, "step": 13440 }, { "epoch": 0.91, "learning_rate": 8.576411145884395e-06, "loss": 0.6563, "step": 13445 }, { "epoch": 0.91, "learning_rate": 8.575468496663022e-06, "loss": 0.4315, "step": 13450 }, { "epoch": 0.91, "learning_rate": 8.57452584744165e-06, "loss": 0.4007, "step": 13455 }, { "epoch": 0.91, "learning_rate": 8.57358319822028e-06, "loss": 0.5772, "step": 13460 }, { "epoch": 0.91, "learning_rate": 8.572640548998907e-06, "loss": 0.4775, "step": 13465 }, { "epoch": 0.91, "learning_rate": 8.571697899777536e-06, "loss": 0.6788, "step": 13470 }, { "epoch": 0.91, "learning_rate": 8.570755250556164e-06, "loss": 0.4183, "step": 13475 }, { "epoch": 0.91, "learning_rate": 8.569812601334791e-06, "loss": 0.5501, "step": 13480 }, { "epoch": 0.92, "learning_rate": 8.56886995211342e-06, "loss": 0.6782, "step": 13485 }, { "epoch": 0.92, "learning_rate": 8.567927302892048e-06, "loss": 0.4743, "step": 13490 }, { "epoch": 0.92, "learning_rate": 8.566984653670676e-06, "loss": 0.4347, "step": 13495 }, { "epoch": 0.92, "learning_rate": 8.566042004449305e-06, "loss": 0.4591, "step": 13500 }, { "epoch": 0.92, "learning_rate": 8.565099355227933e-06, "loss": 0.2694, "step": 13505 }, { "epoch": 0.92, "learning_rate": 8.56415670600656e-06, "loss": 1.0194, "step": 13510 }, { "epoch": 0.92, "learning_rate": 8.56321405678519e-06, "loss": 0.1494, "step": 13515 }, { "epoch": 0.92, "learning_rate": 8.562271407563818e-06, "loss": 0.4185, "step": 13520 }, { "epoch": 0.92, "learning_rate": 8.561328758342447e-06, "loss": 0.6264, "step": 13525 }, { "epoch": 0.92, "learning_rate": 8.560386109121075e-06, "loss": 0.6483, "step": 13530 }, { "epoch": 0.92, "learning_rate": 8.559443459899704e-06, "loss": 0.8161, "step": 13535 }, { "epoch": 0.92, "learning_rate": 8.55850081067833e-06, "loss": 0.6231, "step": 13540 }, { "epoch": 0.92, "learning_rate": 8.55755816145696e-06, "loss": 0.7272, "step": 13545 }, { "epoch": 0.92, "learning_rate": 8.556615512235587e-06, "loss": 0.3486, "step": 13550 }, { "epoch": 0.92, "learning_rate": 8.555672863014216e-06, "loss": 0.8305, "step": 13555 }, { "epoch": 0.92, "learning_rate": 8.554730213792844e-06, "loss": 0.955, "step": 13560 }, { "epoch": 0.92, "learning_rate": 8.553787564571471e-06, "loss": 0.2918, "step": 13565 }, { "epoch": 0.92, "learning_rate": 8.5528449153501e-06, "loss": 0.6445, "step": 13570 }, { "epoch": 0.92, "learning_rate": 8.551902266128728e-06, "loss": 0.7872, "step": 13575 }, { "epoch": 0.92, "learning_rate": 8.550959616907358e-06, "loss": 0.9941, "step": 13580 }, { "epoch": 0.92, "learning_rate": 8.550016967685985e-06, "loss": 0.7962, "step": 13585 }, { "epoch": 0.92, "learning_rate": 8.549074318464615e-06, "loss": 0.6487, "step": 13590 }, { "epoch": 0.92, "learning_rate": 8.54813166924324e-06, "loss": 0.4875, "step": 13595 }, { "epoch": 0.92, "learning_rate": 8.54718902002187e-06, "loss": 0.6909, "step": 13600 }, { "epoch": 0.92, "learning_rate": 8.546246370800498e-06, "loss": 0.287, "step": 13605 }, { "epoch": 0.92, "learning_rate": 8.545303721579127e-06, "loss": 0.721, "step": 13610 }, { "epoch": 0.92, "learning_rate": 8.544361072357755e-06, "loss": 0.3946, "step": 13615 }, { "epoch": 0.92, "learning_rate": 8.543418423136384e-06, "loss": 1.0114, "step": 13620 }, { "epoch": 0.92, "learning_rate": 8.542475773915012e-06, "loss": 0.6169, "step": 13625 }, { "epoch": 0.93, "learning_rate": 8.54153312469364e-06, "loss": 0.5135, "step": 13630 }, { "epoch": 0.93, "learning_rate": 8.540590475472269e-06, "loss": 0.7616, "step": 13635 }, { "epoch": 0.93, "learning_rate": 8.539647826250896e-06, "loss": 0.5411, "step": 13640 }, { "epoch": 0.93, "learning_rate": 8.538705177029526e-06, "loss": 1.3114, "step": 13645 }, { "epoch": 0.93, "learning_rate": 8.537762527808152e-06, "loss": 0.9958, "step": 13650 }, { "epoch": 0.93, "learning_rate": 8.536819878586781e-06, "loss": 0.5603, "step": 13655 }, { "epoch": 0.93, "learning_rate": 8.535877229365408e-06, "loss": 0.1583, "step": 13660 }, { "epoch": 0.93, "learning_rate": 8.534934580144038e-06, "loss": 0.3367, "step": 13665 }, { "epoch": 0.93, "learning_rate": 8.533991930922665e-06, "loss": 0.5981, "step": 13670 }, { "epoch": 0.93, "learning_rate": 8.533049281701295e-06, "loss": 0.5796, "step": 13675 }, { "epoch": 0.93, "learning_rate": 8.532106632479922e-06, "loss": 0.571, "step": 13680 }, { "epoch": 0.93, "learning_rate": 8.53116398325855e-06, "loss": 0.5872, "step": 13685 }, { "epoch": 0.93, "learning_rate": 8.53022133403718e-06, "loss": 0.8973, "step": 13690 }, { "epoch": 0.93, "learning_rate": 8.529278684815807e-06, "loss": 0.7924, "step": 13695 }, { "epoch": 0.93, "learning_rate": 8.528336035594436e-06, "loss": 0.7796, "step": 13700 }, { "epoch": 0.93, "learning_rate": 8.527393386373064e-06, "loss": 0.5846, "step": 13705 }, { "epoch": 0.93, "learning_rate": 8.526450737151692e-06, "loss": 0.9736, "step": 13710 }, { "epoch": 0.93, "learning_rate": 8.52550808793032e-06, "loss": 0.6598, "step": 13715 }, { "epoch": 0.93, "learning_rate": 8.524565438708949e-06, "loss": 0.2076, "step": 13720 }, { "epoch": 0.93, "learning_rate": 8.523622789487576e-06, "loss": 0.4914, "step": 13725 }, { "epoch": 0.93, "learning_rate": 8.522680140266206e-06, "loss": 0.4532, "step": 13730 }, { "epoch": 0.93, "learning_rate": 8.521737491044833e-06, "loss": 0.4673, "step": 13735 }, { "epoch": 0.93, "learning_rate": 8.520794841823461e-06, "loss": 0.8018, "step": 13740 }, { "epoch": 0.93, "learning_rate": 8.51985219260209e-06, "loss": 1.6557, "step": 13745 }, { "epoch": 0.93, "learning_rate": 8.518909543380718e-06, "loss": 1.4372, "step": 13750 }, { "epoch": 0.93, "learning_rate": 8.517966894159346e-06, "loss": 0.7557, "step": 13755 }, { "epoch": 0.93, "learning_rate": 8.517024244937975e-06, "loss": 0.7128, "step": 13760 }, { "epoch": 0.93, "learning_rate": 8.516081595716602e-06, "loss": 0.6092, "step": 13765 }, { "epoch": 0.93, "learning_rate": 8.51513894649523e-06, "loss": 1.4208, "step": 13770 }, { "epoch": 0.93, "learning_rate": 8.51419629727386e-06, "loss": 0.9739, "step": 13775 }, { "epoch": 0.94, "learning_rate": 8.513253648052487e-06, "loss": 0.784, "step": 13780 }, { "epoch": 0.94, "learning_rate": 8.512310998831116e-06, "loss": 0.3215, "step": 13785 }, { "epoch": 0.94, "learning_rate": 8.511368349609744e-06, "loss": 0.8359, "step": 13790 }, { "epoch": 0.94, "learning_rate": 8.510425700388372e-06, "loss": 0.8841, "step": 13795 }, { "epoch": 0.94, "learning_rate": 8.509483051167e-06, "loss": 0.5918, "step": 13800 }, { "epoch": 0.94, "learning_rate": 8.508540401945629e-06, "loss": 1.0178, "step": 13805 }, { "epoch": 0.94, "learning_rate": 8.507597752724256e-06, "loss": 0.509, "step": 13810 }, { "epoch": 0.94, "learning_rate": 8.506655103502886e-06, "loss": 0.3594, "step": 13815 }, { "epoch": 0.94, "learning_rate": 8.505712454281513e-06, "loss": 2.2063, "step": 13820 }, { "epoch": 0.94, "learning_rate": 8.504769805060141e-06, "loss": 0.7891, "step": 13825 }, { "epoch": 0.94, "learning_rate": 8.50382715583877e-06, "loss": 0.3713, "step": 13830 }, { "epoch": 0.94, "learning_rate": 8.502884506617398e-06, "loss": 0.5252, "step": 13835 }, { "epoch": 0.94, "learning_rate": 8.501941857396027e-06, "loss": 0.7195, "step": 13840 }, { "epoch": 0.94, "learning_rate": 8.500999208174655e-06, "loss": 0.5613, "step": 13845 }, { "epoch": 0.94, "learning_rate": 8.500056558953284e-06, "loss": 1.1279, "step": 13850 }, { "epoch": 0.94, "learning_rate": 8.49911390973191e-06, "loss": 0.4109, "step": 13855 }, { "epoch": 0.94, "learning_rate": 8.49817126051054e-06, "loss": 0.7238, "step": 13860 }, { "epoch": 0.94, "learning_rate": 8.497228611289167e-06, "loss": 0.171, "step": 13865 }, { "epoch": 0.94, "learning_rate": 8.496285962067796e-06, "loss": 0.4666, "step": 13870 }, { "epoch": 0.94, "learning_rate": 8.495343312846424e-06, "loss": 0.4337, "step": 13875 }, { "epoch": 0.94, "learning_rate": 8.494400663625052e-06, "loss": 0.6498, "step": 13880 }, { "epoch": 0.94, "learning_rate": 8.493458014403681e-06, "loss": 0.6096, "step": 13885 }, { "epoch": 0.94, "learning_rate": 8.492515365182309e-06, "loss": 0.3758, "step": 13890 }, { "epoch": 0.94, "learning_rate": 8.491572715960938e-06, "loss": 0.7763, "step": 13895 }, { "epoch": 0.94, "learning_rate": 8.490630066739566e-06, "loss": 0.8753, "step": 13900 }, { "epoch": 0.94, "learning_rate": 8.489687417518195e-06, "loss": 0.887, "step": 13905 }, { "epoch": 0.94, "learning_rate": 8.488744768296821e-06, "loss": 0.4597, "step": 13910 }, { "epoch": 0.94, "learning_rate": 8.48780211907545e-06, "loss": 0.2366, "step": 13915 }, { "epoch": 0.94, "learning_rate": 8.486859469854078e-06, "loss": 0.682, "step": 13920 }, { "epoch": 0.95, "learning_rate": 8.485916820632707e-06, "loss": 0.4717, "step": 13925 }, { "epoch": 0.95, "learning_rate": 8.484974171411335e-06, "loss": 0.7946, "step": 13930 }, { "epoch": 0.95, "learning_rate": 8.484031522189964e-06, "loss": 1.0232, "step": 13935 }, { "epoch": 0.95, "learning_rate": 8.483088872968592e-06, "loss": 0.8397, "step": 13940 }, { "epoch": 0.95, "learning_rate": 8.48214622374722e-06, "loss": 0.7816, "step": 13945 }, { "epoch": 0.95, "learning_rate": 8.481203574525849e-06, "loss": 0.3983, "step": 13950 }, { "epoch": 0.95, "learning_rate": 8.480260925304477e-06, "loss": 0.8635, "step": 13955 }, { "epoch": 0.95, "learning_rate": 8.479318276083104e-06, "loss": 0.4284, "step": 13960 }, { "epoch": 0.95, "learning_rate": 8.478375626861734e-06, "loss": 0.3146, "step": 13965 }, { "epoch": 0.95, "learning_rate": 8.477432977640361e-06, "loss": 0.6261, "step": 13970 }, { "epoch": 0.95, "learning_rate": 8.476490328418989e-06, "loss": 0.8141, "step": 13975 }, { "epoch": 0.95, "learning_rate": 8.475547679197618e-06, "loss": 0.3623, "step": 13980 }, { "epoch": 0.95, "learning_rate": 8.474605029976246e-06, "loss": 0.4596, "step": 13985 }, { "epoch": 0.95, "learning_rate": 8.473662380754875e-06, "loss": 0.5472, "step": 13990 }, { "epoch": 0.95, "learning_rate": 8.472719731533503e-06, "loss": 1.4119, "step": 13995 }, { "epoch": 0.95, "learning_rate": 8.47177708231213e-06, "loss": 0.661, "step": 14000 }, { "epoch": 0.95, "learning_rate": 8.470834433090758e-06, "loss": 0.3563, "step": 14005 }, { "epoch": 0.95, "learning_rate": 8.469891783869387e-06, "loss": 0.1825, "step": 14010 }, { "epoch": 0.95, "learning_rate": 8.468949134648015e-06, "loss": 0.8003, "step": 14015 }, { "epoch": 0.95, "learning_rate": 8.468006485426644e-06, "loss": 0.8077, "step": 14020 }, { "epoch": 0.95, "learning_rate": 8.467063836205272e-06, "loss": 0.3069, "step": 14025 }, { "epoch": 0.95, "learning_rate": 8.4661211869839e-06, "loss": 0.7142, "step": 14030 }, { "epoch": 0.95, "learning_rate": 8.465178537762529e-06, "loss": 0.3277, "step": 14035 }, { "epoch": 0.95, "learning_rate": 8.464235888541157e-06, "loss": 0.9492, "step": 14040 }, { "epoch": 0.95, "learning_rate": 8.463293239319786e-06, "loss": 1.4833, "step": 14045 }, { "epoch": 0.95, "learning_rate": 8.462350590098414e-06, "loss": 0.3946, "step": 14050 }, { "epoch": 0.95, "learning_rate": 8.461407940877041e-06, "loss": 0.5927, "step": 14055 }, { "epoch": 0.95, "learning_rate": 8.460465291655669e-06, "loss": 0.4949, "step": 14060 }, { "epoch": 0.95, "learning_rate": 8.459522642434298e-06, "loss": 0.2983, "step": 14065 }, { "epoch": 0.95, "learning_rate": 8.458579993212926e-06, "loss": 0.3598, "step": 14070 }, { "epoch": 0.96, "learning_rate": 8.457637343991555e-06, "loss": 0.6961, "step": 14075 }, { "epoch": 0.96, "learning_rate": 8.456694694770183e-06, "loss": 0.8943, "step": 14080 }, { "epoch": 0.96, "learning_rate": 8.45575204554881e-06, "loss": 0.7204, "step": 14085 }, { "epoch": 0.96, "learning_rate": 8.45480939632744e-06, "loss": 0.7584, "step": 14090 }, { "epoch": 0.96, "learning_rate": 8.453866747106067e-06, "loss": 0.387, "step": 14095 }, { "epoch": 0.96, "learning_rate": 8.452924097884697e-06, "loss": 1.2293, "step": 14100 }, { "epoch": 0.96, "learning_rate": 8.451981448663324e-06, "loss": 0.7628, "step": 14105 }, { "epoch": 0.96, "learning_rate": 8.451038799441952e-06, "loss": 0.8642, "step": 14110 }, { "epoch": 0.96, "learning_rate": 8.45009615022058e-06, "loss": 0.1514, "step": 14115 }, { "epoch": 0.96, "learning_rate": 8.449153500999209e-06, "loss": 1.1298, "step": 14120 }, { "epoch": 0.96, "learning_rate": 8.448210851777837e-06, "loss": 0.7164, "step": 14125 }, { "epoch": 0.96, "learning_rate": 8.447268202556466e-06, "loss": 0.3778, "step": 14130 }, { "epoch": 0.96, "learning_rate": 8.446325553335094e-06, "loss": 0.4841, "step": 14135 }, { "epoch": 0.96, "learning_rate": 8.445382904113721e-06, "loss": 1.1663, "step": 14140 }, { "epoch": 0.96, "learning_rate": 8.44444025489235e-06, "loss": 0.2422, "step": 14145 }, { "epoch": 0.96, "learning_rate": 8.443497605670978e-06, "loss": 0.4416, "step": 14150 }, { "epoch": 0.96, "learning_rate": 8.442554956449608e-06, "loss": 0.3698, "step": 14155 }, { "epoch": 0.96, "learning_rate": 8.441612307228235e-06, "loss": 1.0096, "step": 14160 }, { "epoch": 0.96, "learning_rate": 8.440669658006863e-06, "loss": 1.2363, "step": 14165 }, { "epoch": 0.96, "learning_rate": 8.43972700878549e-06, "loss": 0.3329, "step": 14170 }, { "epoch": 0.96, "learning_rate": 8.43878435956412e-06, "loss": 1.0507, "step": 14175 }, { "epoch": 0.96, "learning_rate": 8.437841710342747e-06, "loss": 0.6, "step": 14180 }, { "epoch": 0.96, "learning_rate": 8.436899061121377e-06, "loss": 1.0505, "step": 14185 }, { "epoch": 0.96, "learning_rate": 8.435956411900004e-06, "loss": 0.5213, "step": 14190 }, { "epoch": 0.96, "learning_rate": 8.435013762678634e-06, "loss": 0.3845, "step": 14195 }, { "epoch": 0.96, "learning_rate": 8.434071113457261e-06, "loss": 0.7099, "step": 14200 }, { "epoch": 0.96, "learning_rate": 8.433128464235889e-06, "loss": 0.6503, "step": 14205 }, { "epoch": 0.96, "learning_rate": 8.432185815014517e-06, "loss": 0.3626, "step": 14210 }, { "epoch": 0.96, "learning_rate": 8.431243165793146e-06, "loss": 1.5224, "step": 14215 }, { "epoch": 0.97, "learning_rate": 8.430300516571774e-06, "loss": 0.7883, "step": 14220 }, { "epoch": 0.97, "learning_rate": 8.429357867350401e-06, "loss": 0.6949, "step": 14225 }, { "epoch": 0.97, "learning_rate": 8.42841521812903e-06, "loss": 0.3799, "step": 14230 }, { "epoch": 0.97, "learning_rate": 8.427472568907658e-06, "loss": 0.6777, "step": 14235 }, { "epoch": 0.97, "learning_rate": 8.426529919686288e-06, "loss": 0.9019, "step": 14240 }, { "epoch": 0.97, "learning_rate": 8.425587270464915e-06, "loss": 0.2298, "step": 14245 }, { "epoch": 0.97, "learning_rate": 8.424644621243545e-06, "loss": 0.8235, "step": 14250 }, { "epoch": 0.97, "learning_rate": 8.42370197202217e-06, "loss": 0.2257, "step": 14255 }, { "epoch": 0.97, "learning_rate": 8.4227593228008e-06, "loss": 0.5713, "step": 14260 }, { "epoch": 0.97, "learning_rate": 8.421816673579428e-06, "loss": 0.8682, "step": 14265 }, { "epoch": 0.97, "learning_rate": 8.420874024358057e-06, "loss": 0.8163, "step": 14270 }, { "epoch": 0.97, "learning_rate": 8.419931375136684e-06, "loss": 0.2027, "step": 14275 }, { "epoch": 0.97, "learning_rate": 8.418988725915314e-06, "loss": 0.5043, "step": 14280 }, { "epoch": 0.97, "learning_rate": 8.418046076693941e-06, "loss": 0.7207, "step": 14285 }, { "epoch": 0.97, "learning_rate": 8.417103427472569e-06, "loss": 0.8128, "step": 14290 }, { "epoch": 0.97, "learning_rate": 8.416160778251198e-06, "loss": 1.0057, "step": 14295 }, { "epoch": 0.97, "learning_rate": 8.415218129029826e-06, "loss": 1.0036, "step": 14300 }, { "epoch": 0.97, "learning_rate": 8.414275479808455e-06, "loss": 0.4512, "step": 14305 }, { "epoch": 0.97, "learning_rate": 8.413332830587083e-06, "loss": 0.6484, "step": 14310 }, { "epoch": 0.97, "learning_rate": 8.41239018136571e-06, "loss": 1.1379, "step": 14315 }, { "epoch": 0.97, "learning_rate": 8.411447532144338e-06, "loss": 0.5299, "step": 14320 }, { "epoch": 0.97, "learning_rate": 8.410504882922968e-06, "loss": 0.0888, "step": 14325 }, { "epoch": 0.97, "learning_rate": 8.409562233701595e-06, "loss": 0.874, "step": 14330 }, { "epoch": 0.97, "learning_rate": 8.408619584480225e-06, "loss": 0.6876, "step": 14335 }, { "epoch": 0.97, "learning_rate": 8.407676935258852e-06, "loss": 0.1909, "step": 14340 }, { "epoch": 0.97, "learning_rate": 8.40673428603748e-06, "loss": 0.2639, "step": 14345 }, { "epoch": 0.97, "learning_rate": 8.40579163681611e-06, "loss": 0.9478, "step": 14350 }, { "epoch": 0.97, "learning_rate": 8.404848987594737e-06, "loss": 0.4722, "step": 14355 }, { "epoch": 0.97, "learning_rate": 8.403906338373366e-06, "loss": 0.7867, "step": 14360 }, { "epoch": 0.97, "learning_rate": 8.402963689151994e-06, "loss": 0.7043, "step": 14365 }, { "epoch": 0.98, "learning_rate": 8.402021039930622e-06, "loss": 0.2313, "step": 14370 }, { "epoch": 0.98, "learning_rate": 8.401078390709249e-06, "loss": 0.5678, "step": 14375 }, { "epoch": 0.98, "learning_rate": 8.400135741487878e-06, "loss": 0.6388, "step": 14380 }, { "epoch": 0.98, "learning_rate": 8.399193092266506e-06, "loss": 0.2248, "step": 14385 }, { "epoch": 0.98, "learning_rate": 8.398250443045135e-06, "loss": 0.5287, "step": 14390 }, { "epoch": 0.98, "learning_rate": 8.397307793823763e-06, "loss": 0.713, "step": 14395 }, { "epoch": 0.98, "learning_rate": 8.39636514460239e-06, "loss": 0.5034, "step": 14400 }, { "epoch": 0.98, "learning_rate": 8.39542249538102e-06, "loss": 0.4549, "step": 14405 }, { "epoch": 0.98, "learning_rate": 8.394479846159648e-06, "loss": 1.0408, "step": 14410 }, { "epoch": 0.98, "learning_rate": 8.393537196938275e-06, "loss": 0.4869, "step": 14415 }, { "epoch": 0.98, "learning_rate": 8.392594547716905e-06, "loss": 0.767, "step": 14420 }, { "epoch": 0.98, "learning_rate": 8.391651898495532e-06, "loss": 0.2074, "step": 14425 }, { "epoch": 0.98, "learning_rate": 8.39070924927416e-06, "loss": 0.4904, "step": 14430 }, { "epoch": 0.98, "learning_rate": 8.38976660005279e-06, "loss": 0.4163, "step": 14435 }, { "epoch": 0.98, "learning_rate": 8.388823950831417e-06, "loss": 0.7459, "step": 14440 }, { "epoch": 0.98, "learning_rate": 8.387881301610046e-06, "loss": 0.2773, "step": 14445 }, { "epoch": 0.98, "learning_rate": 8.386938652388674e-06, "loss": 0.918, "step": 14450 }, { "epoch": 0.98, "learning_rate": 8.385996003167302e-06, "loss": 1.093, "step": 14455 }, { "epoch": 0.98, "learning_rate": 8.38505335394593e-06, "loss": 1.0566, "step": 14460 }, { "epoch": 0.98, "learning_rate": 8.384110704724559e-06, "loss": 0.3786, "step": 14465 }, { "epoch": 0.98, "learning_rate": 8.383168055503186e-06, "loss": 1.0359, "step": 14470 }, { "epoch": 0.98, "learning_rate": 8.382225406281816e-06, "loss": 0.5592, "step": 14475 }, { "epoch": 0.98, "learning_rate": 8.381282757060443e-06, "loss": 1.0469, "step": 14480 }, { "epoch": 0.98, "learning_rate": 8.38034010783907e-06, "loss": 0.4298, "step": 14485 }, { "epoch": 0.98, "learning_rate": 8.3793974586177e-06, "loss": 1.2273, "step": 14490 }, { "epoch": 0.98, "learning_rate": 8.378454809396328e-06, "loss": 0.4233, "step": 14495 }, { "epoch": 0.98, "learning_rate": 8.377512160174957e-06, "loss": 0.3443, "step": 14500 }, { "epoch": 0.98, "learning_rate": 8.376569510953585e-06, "loss": 0.2362, "step": 14505 }, { "epoch": 0.98, "learning_rate": 8.375626861732214e-06, "loss": 1.1261, "step": 14510 }, { "epoch": 0.99, "learning_rate": 8.37468421251084e-06, "loss": 0.3437, "step": 14515 }, { "epoch": 0.99, "learning_rate": 8.37374156328947e-06, "loss": 0.5877, "step": 14520 }, { "epoch": 0.99, "learning_rate": 8.372798914068097e-06, "loss": 0.7611, "step": 14525 }, { "epoch": 0.99, "learning_rate": 8.371856264846726e-06, "loss": 0.2696, "step": 14530 }, { "epoch": 0.99, "learning_rate": 8.370913615625354e-06, "loss": 0.4808, "step": 14535 }, { "epoch": 0.99, "learning_rate": 8.369970966403983e-06, "loss": 1.025, "step": 14540 }, { "epoch": 0.99, "learning_rate": 8.369028317182611e-06, "loss": 0.5356, "step": 14545 }, { "epoch": 0.99, "learning_rate": 8.368085667961239e-06, "loss": 0.6566, "step": 14550 }, { "epoch": 0.99, "learning_rate": 8.367143018739868e-06, "loss": 0.3901, "step": 14555 }, { "epoch": 0.99, "learning_rate": 8.366200369518496e-06, "loss": 0.6937, "step": 14560 }, { "epoch": 0.99, "learning_rate": 8.365257720297125e-06, "loss": 0.3972, "step": 14565 }, { "epoch": 0.99, "learning_rate": 8.36431507107575e-06, "loss": 1.1767, "step": 14570 }, { "epoch": 0.99, "learning_rate": 8.36337242185438e-06, "loss": 0.5208, "step": 14575 }, { "epoch": 0.99, "learning_rate": 8.362429772633008e-06, "loss": 0.6339, "step": 14580 }, { "epoch": 0.99, "learning_rate": 8.361487123411637e-06, "loss": 0.7108, "step": 14585 }, { "epoch": 0.99, "learning_rate": 8.360544474190265e-06, "loss": 0.7375, "step": 14590 }, { "epoch": 0.99, "learning_rate": 8.359601824968894e-06, "loss": 0.9679, "step": 14595 }, { "epoch": 0.99, "learning_rate": 8.358659175747522e-06, "loss": 1.0126, "step": 14600 }, { "epoch": 0.99, "learning_rate": 8.35771652652615e-06, "loss": 0.5572, "step": 14605 }, { "epoch": 0.99, "learning_rate": 8.356773877304779e-06, "loss": 0.229, "step": 14610 }, { "epoch": 0.99, "learning_rate": 8.355831228083406e-06, "loss": 0.2556, "step": 14615 }, { "epoch": 0.99, "learning_rate": 8.354888578862034e-06, "loss": 0.536, "step": 14620 }, { "epoch": 0.99, "learning_rate": 8.353945929640663e-06, "loss": 0.3826, "step": 14625 }, { "epoch": 0.99, "learning_rate": 8.353003280419291e-06, "loss": 0.48, "step": 14630 }, { "epoch": 0.99, "learning_rate": 8.352060631197919e-06, "loss": 0.7679, "step": 14635 }, { "epoch": 0.99, "learning_rate": 8.351117981976548e-06, "loss": 0.4478, "step": 14640 }, { "epoch": 0.99, "learning_rate": 8.350175332755176e-06, "loss": 0.5214, "step": 14645 }, { "epoch": 0.99, "learning_rate": 8.349232683533805e-06, "loss": 0.2901, "step": 14650 }, { "epoch": 0.99, "learning_rate": 8.348290034312433e-06, "loss": 0.6027, "step": 14655 }, { "epoch": 0.99, "learning_rate": 8.34734738509106e-06, "loss": 0.5406, "step": 14660 }, { "epoch": 1.0, "learning_rate": 8.346404735869688e-06, "loss": 0.5026, "step": 14665 }, { "epoch": 1.0, "learning_rate": 8.345462086648317e-06, "loss": 0.8803, "step": 14670 }, { "epoch": 1.0, "learning_rate": 8.344519437426945e-06, "loss": 0.2445, "step": 14675 }, { "epoch": 1.0, "learning_rate": 8.343576788205574e-06, "loss": 1.2849, "step": 14680 }, { "epoch": 1.0, "learning_rate": 8.342634138984202e-06, "loss": 0.68, "step": 14685 }, { "epoch": 1.0, "learning_rate": 8.34169148976283e-06, "loss": 0.5576, "step": 14690 }, { "epoch": 1.0, "learning_rate": 8.340748840541459e-06, "loss": 0.2507, "step": 14695 }, { "epoch": 1.0, "learning_rate": 8.339806191320086e-06, "loss": 0.4113, "step": 14700 }, { "epoch": 1.0, "learning_rate": 8.338863542098716e-06, "loss": 0.5147, "step": 14705 }, { "epoch": 1.0, "learning_rate": 8.337920892877343e-06, "loss": 0.9777, "step": 14710 }, { "epoch": 1.0, "learning_rate": 8.336978243655971e-06, "loss": 0.3532, "step": 14715 }, { "epoch": 1.0, "learning_rate": 8.336035594434599e-06, "loss": 0.6967, "step": 14720 }, { "epoch": 1.0, "learning_rate": 8.335092945213228e-06, "loss": 0.4511, "step": 14725 }, { "epoch": 1.0, "learning_rate": 8.334150295991856e-06, "loss": 0.3211, "step": 14730 }, { "epoch": 1.0, "learning_rate": 8.333207646770485e-06, "loss": 0.7885, "step": 14735 }, { "epoch": 1.0, "learning_rate": 8.332264997549113e-06, "loss": 0.2098, "step": 14740 }, { "epoch": 1.0, "learning_rate": 8.33132234832774e-06, "loss": 0.5034, "step": 14745 }, { "epoch": 1.0, "learning_rate": 8.33037969910637e-06, "loss": 0.7443, "step": 14750 }, { "epoch": 1.0, "learning_rate": 8.329437049884997e-06, "loss": 0.1834, "step": 14755 }, { "epoch": 1.0, "learning_rate": 8.328494400663627e-06, "loss": 0.2211, "step": 14760 }, { "epoch": 1.0, "learning_rate": 8.327551751442254e-06, "loss": 0.7329, "step": 14765 }, { "epoch": 1.0, "learning_rate": 8.326609102220882e-06, "loss": 0.6499, "step": 14770 }, { "epoch": 1.0, "learning_rate": 8.32566645299951e-06, "loss": 0.7174, "step": 14775 }, { "epoch": 1.0, "learning_rate": 8.324723803778139e-06, "loss": 0.3339, "step": 14780 }, { "epoch": 1.0, "learning_rate": 8.323781154556766e-06, "loss": 0.5936, "step": 14785 }, { "epoch": 1.0, "learning_rate": 8.322838505335396e-06, "loss": 0.7135, "step": 14790 }, { "epoch": 1.0, "learning_rate": 8.321895856114023e-06, "loss": 0.6997, "step": 14795 }, { "epoch": 1.0, "learning_rate": 8.320953206892651e-06, "loss": 0.1339, "step": 14800 }, { "epoch": 1.0, "learning_rate": 8.32001055767128e-06, "loss": 0.2519, "step": 14805 }, { "epoch": 1.01, "learning_rate": 8.319067908449908e-06, "loss": 0.9293, "step": 14810 }, { "epoch": 1.01, "learning_rate": 8.318125259228537e-06, "loss": 0.3939, "step": 14815 }, { "epoch": 1.01, "learning_rate": 8.317182610007165e-06, "loss": 0.6797, "step": 14820 }, { "epoch": 1.01, "learning_rate": 8.316239960785794e-06, "loss": 0.3225, "step": 14825 }, { "epoch": 1.01, "learning_rate": 8.31529731156442e-06, "loss": 0.4534, "step": 14830 }, { "epoch": 1.01, "learning_rate": 8.31435466234305e-06, "loss": 0.4999, "step": 14835 }, { "epoch": 1.01, "learning_rate": 8.313412013121677e-06, "loss": 0.3344, "step": 14840 }, { "epoch": 1.01, "learning_rate": 8.312469363900307e-06, "loss": 0.6157, "step": 14845 }, { "epoch": 1.01, "learning_rate": 8.311526714678934e-06, "loss": 0.4415, "step": 14850 }, { "epoch": 1.01, "learning_rate": 8.310584065457564e-06, "loss": 0.538, "step": 14855 }, { "epoch": 1.01, "learning_rate": 8.309641416236191e-06, "loss": 0.6304, "step": 14860 }, { "epoch": 1.01, "learning_rate": 8.308698767014819e-06, "loss": 0.5964, "step": 14865 }, { "epoch": 1.01, "learning_rate": 8.307756117793448e-06, "loss": 1.0059, "step": 14870 }, { "epoch": 1.01, "learning_rate": 8.306813468572076e-06, "loss": 0.4937, "step": 14875 }, { "epoch": 1.01, "learning_rate": 8.305870819350703e-06, "loss": 0.368, "step": 14880 }, { "epoch": 1.01, "learning_rate": 8.304928170129331e-06, "loss": 0.4406, "step": 14885 }, { "epoch": 1.01, "learning_rate": 8.30398552090796e-06, "loss": 0.1294, "step": 14890 }, { "epoch": 1.01, "learning_rate": 8.303042871686588e-06, "loss": 0.3919, "step": 14895 }, { "epoch": 1.01, "learning_rate": 8.302100222465217e-06, "loss": 0.8144, "step": 14900 }, { "epoch": 1.01, "learning_rate": 8.301157573243845e-06, "loss": 0.5939, "step": 14905 }, { "epoch": 1.01, "learning_rate": 8.300214924022474e-06, "loss": 0.2133, "step": 14910 }, { "epoch": 1.01, "learning_rate": 8.2992722748011e-06, "loss": 0.4457, "step": 14915 }, { "epoch": 1.01, "learning_rate": 8.29832962557973e-06, "loss": 0.8174, "step": 14920 }, { "epoch": 1.01, "learning_rate": 8.297386976358357e-06, "loss": 0.4785, "step": 14925 }, { "epoch": 1.01, "learning_rate": 8.296444327136987e-06, "loss": 0.4848, "step": 14930 }, { "epoch": 1.01, "learning_rate": 8.295501677915614e-06, "loss": 0.5356, "step": 14935 }, { "epoch": 1.01, "learning_rate": 8.294559028694244e-06, "loss": 0.6153, "step": 14940 }, { "epoch": 1.01, "learning_rate": 8.293616379472871e-06, "loss": 0.5007, "step": 14945 }, { "epoch": 1.01, "learning_rate": 8.292673730251499e-06, "loss": 0.8025, "step": 14950 }, { "epoch": 1.01, "learning_rate": 8.291731081030128e-06, "loss": 0.3245, "step": 14955 }, { "epoch": 1.02, "learning_rate": 8.290788431808756e-06, "loss": 0.3483, "step": 14960 }, { "epoch": 1.02, "learning_rate": 8.289845782587385e-06, "loss": 0.2688, "step": 14965 }, { "epoch": 1.02, "learning_rate": 8.288903133366013e-06, "loss": 0.3907, "step": 14970 }, { "epoch": 1.02, "learning_rate": 8.28796048414464e-06, "loss": 0.4554, "step": 14975 }, { "epoch": 1.02, "learning_rate": 8.287017834923268e-06, "loss": 0.8898, "step": 14980 }, { "epoch": 1.02, "learning_rate": 8.286075185701897e-06, "loss": 0.7271, "step": 14985 }, { "epoch": 1.02, "learning_rate": 8.285132536480525e-06, "loss": 1.0673, "step": 14990 }, { "epoch": 1.02, "learning_rate": 8.284189887259154e-06, "loss": 0.4145, "step": 14995 }, { "epoch": 1.02, "learning_rate": 8.283247238037782e-06, "loss": 0.7714, "step": 15000 }, { "epoch": 1.02, "learning_rate": 8.28230458881641e-06, "loss": 0.4382, "step": 15005 }, { "epoch": 1.02, "learning_rate": 8.281361939595039e-06, "loss": 0.6274, "step": 15010 }, { "epoch": 1.02, "learning_rate": 8.280419290373667e-06, "loss": 0.1367, "step": 15015 }, { "epoch": 1.02, "learning_rate": 8.279476641152296e-06, "loss": 0.572, "step": 15020 }, { "epoch": 1.02, "learning_rate": 8.278533991930924e-06, "loss": 0.3749, "step": 15025 }, { "epoch": 1.02, "learning_rate": 8.277591342709551e-06, "loss": 0.1642, "step": 15030 }, { "epoch": 1.02, "learning_rate": 8.276648693488179e-06, "loss": 0.8131, "step": 15035 }, { "epoch": 1.02, "learning_rate": 8.275706044266808e-06, "loss": 0.4753, "step": 15040 }, { "epoch": 1.02, "learning_rate": 8.274763395045436e-06, "loss": 0.8536, "step": 15045 }, { "epoch": 1.02, "learning_rate": 8.273820745824065e-06, "loss": 0.2961, "step": 15050 }, { "epoch": 1.02, "learning_rate": 8.272878096602693e-06, "loss": 0.4488, "step": 15055 }, { "epoch": 1.02, "learning_rate": 8.27193544738132e-06, "loss": 0.0869, "step": 15060 }, { "epoch": 1.02, "learning_rate": 8.27099279815995e-06, "loss": 0.8472, "step": 15065 }, { "epoch": 1.02, "learning_rate": 8.270050148938578e-06, "loss": 0.3106, "step": 15070 }, { "epoch": 1.02, "learning_rate": 8.269107499717207e-06, "loss": 0.516, "step": 15075 }, { "epoch": 1.02, "learning_rate": 8.268164850495835e-06, "loss": 0.3399, "step": 15080 }, { "epoch": 1.02, "learning_rate": 8.267222201274462e-06, "loss": 0.6346, "step": 15085 }, { "epoch": 1.02, "learning_rate": 8.26627955205309e-06, "loss": 0.3665, "step": 15090 }, { "epoch": 1.02, "learning_rate": 8.265336902831719e-06, "loss": 0.3769, "step": 15095 }, { "epoch": 1.02, "learning_rate": 8.264394253610347e-06, "loss": 0.2557, "step": 15100 }, { "epoch": 1.03, "learning_rate": 8.263451604388976e-06, "loss": 0.7448, "step": 15105 }, { "epoch": 1.03, "learning_rate": 8.262508955167604e-06, "loss": 0.6179, "step": 15110 }, { "epoch": 1.03, "learning_rate": 8.261566305946231e-06, "loss": 0.7832, "step": 15115 }, { "epoch": 1.03, "learning_rate": 8.26062365672486e-06, "loss": 0.2303, "step": 15120 }, { "epoch": 1.03, "learning_rate": 8.259681007503488e-06, "loss": 0.7044, "step": 15125 }, { "epoch": 1.03, "learning_rate": 8.258738358282116e-06, "loss": 0.1139, "step": 15130 }, { "epoch": 1.03, "learning_rate": 8.257795709060745e-06, "loss": 0.2104, "step": 15135 }, { "epoch": 1.03, "learning_rate": 8.256853059839373e-06, "loss": 0.6128, "step": 15140 }, { "epoch": 1.03, "learning_rate": 8.255910410618e-06, "loss": 0.8891, "step": 15145 }, { "epoch": 1.03, "learning_rate": 8.25496776139663e-06, "loss": 0.4036, "step": 15150 }, { "epoch": 1.03, "learning_rate": 8.254025112175258e-06, "loss": 0.7825, "step": 15155 }, { "epoch": 1.03, "learning_rate": 8.253082462953887e-06, "loss": 0.6667, "step": 15160 }, { "epoch": 1.03, "learning_rate": 8.252139813732515e-06, "loss": 0.3344, "step": 15165 }, { "epoch": 1.03, "learning_rate": 8.251197164511144e-06, "loss": 0.5949, "step": 15170 }, { "epoch": 1.03, "learning_rate": 8.25025451528977e-06, "loss": 0.5897, "step": 15175 }, { "epoch": 1.03, "learning_rate": 8.2493118660684e-06, "loss": 0.5833, "step": 15180 }, { "epoch": 1.03, "learning_rate": 8.248369216847027e-06, "loss": 0.6816, "step": 15185 }, { "epoch": 1.03, "learning_rate": 8.247426567625656e-06, "loss": 0.6008, "step": 15190 }, { "epoch": 1.03, "learning_rate": 8.246483918404284e-06, "loss": 0.65, "step": 15195 }, { "epoch": 1.03, "learning_rate": 8.245541269182913e-06, "loss": 0.2779, "step": 15200 }, { "epoch": 1.03, "learning_rate": 8.24459861996154e-06, "loss": 0.4889, "step": 15205 }, { "epoch": 1.03, "learning_rate": 8.243655970740168e-06, "loss": 0.8607, "step": 15210 }, { "epoch": 1.03, "learning_rate": 8.242713321518798e-06, "loss": 0.7499, "step": 15215 }, { "epoch": 1.03, "learning_rate": 8.241770672297425e-06, "loss": 0.5889, "step": 15220 }, { "epoch": 1.03, "learning_rate": 8.240828023076055e-06, "loss": 0.4338, "step": 15225 }, { "epoch": 1.03, "learning_rate": 8.23988537385468e-06, "loss": 0.4952, "step": 15230 }, { "epoch": 1.03, "learning_rate": 8.23894272463331e-06, "loss": 0.7798, "step": 15235 }, { "epoch": 1.03, "learning_rate": 8.238000075411938e-06, "loss": 0.142, "step": 15240 }, { "epoch": 1.03, "learning_rate": 8.237057426190567e-06, "loss": 1.1121, "step": 15245 }, { "epoch": 1.04, "learning_rate": 8.236114776969195e-06, "loss": 0.4178, "step": 15250 }, { "epoch": 1.04, "learning_rate": 8.235172127747824e-06, "loss": 0.5219, "step": 15255 }, { "epoch": 1.04, "learning_rate": 8.234229478526452e-06, "loss": 1.1215, "step": 15260 }, { "epoch": 1.04, "learning_rate": 8.23328682930508e-06, "loss": 0.1924, "step": 15265 }, { "epoch": 1.04, "learning_rate": 8.232344180083709e-06, "loss": 0.4637, "step": 15270 }, { "epoch": 1.04, "learning_rate": 8.231401530862336e-06, "loss": 0.2006, "step": 15275 }, { "epoch": 1.04, "learning_rate": 8.230458881640966e-06, "loss": 0.7421, "step": 15280 }, { "epoch": 1.04, "learning_rate": 8.229516232419593e-06, "loss": 0.4941, "step": 15285 }, { "epoch": 1.04, "learning_rate": 8.22857358319822e-06, "loss": 0.2472, "step": 15290 }, { "epoch": 1.04, "learning_rate": 8.227630933976848e-06, "loss": 0.3313, "step": 15295 }, { "epoch": 1.04, "learning_rate": 8.226688284755478e-06, "loss": 0.6372, "step": 15300 }, { "epoch": 1.04, "learning_rate": 8.225745635534105e-06, "loss": 0.5278, "step": 15305 }, { "epoch": 1.04, "learning_rate": 8.224802986312735e-06, "loss": 0.4764, "step": 15310 }, { "epoch": 1.04, "learning_rate": 8.223860337091362e-06, "loss": 0.4099, "step": 15315 }, { "epoch": 1.04, "learning_rate": 8.22291768786999e-06, "loss": 0.9899, "step": 15320 }, { "epoch": 1.04, "learning_rate": 8.22197503864862e-06, "loss": 0.8766, "step": 15325 }, { "epoch": 1.04, "learning_rate": 8.221032389427247e-06, "loss": 0.8592, "step": 15330 }, { "epoch": 1.04, "learning_rate": 8.220089740205875e-06, "loss": 0.6828, "step": 15335 }, { "epoch": 1.04, "learning_rate": 8.219147090984504e-06, "loss": 0.7189, "step": 15340 }, { "epoch": 1.04, "learning_rate": 8.218204441763132e-06, "loss": 0.4241, "step": 15345 }, { "epoch": 1.04, "learning_rate": 8.21726179254176e-06, "loss": 0.3539, "step": 15350 }, { "epoch": 1.04, "learning_rate": 8.216319143320389e-06, "loss": 0.1228, "step": 15355 }, { "epoch": 1.04, "learning_rate": 8.215376494099016e-06, "loss": 0.3274, "step": 15360 }, { "epoch": 1.04, "learning_rate": 8.214433844877646e-06, "loss": 0.3519, "step": 15365 }, { "epoch": 1.04, "learning_rate": 8.213491195656273e-06, "loss": 0.917, "step": 15370 }, { "epoch": 1.04, "learning_rate": 8.212548546434901e-06, "loss": 0.6291, "step": 15375 }, { "epoch": 1.04, "learning_rate": 8.211605897213529e-06, "loss": 0.376, "step": 15380 }, { "epoch": 1.04, "learning_rate": 8.210663247992158e-06, "loss": 0.753, "step": 15385 }, { "epoch": 1.04, "learning_rate": 8.209720598770785e-06, "loss": 0.2032, "step": 15390 }, { "epoch": 1.04, "learning_rate": 8.208777949549415e-06, "loss": 0.4763, "step": 15395 }, { "epoch": 1.05, "learning_rate": 8.207835300328042e-06, "loss": 0.4535, "step": 15400 }, { "epoch": 1.05, "learning_rate": 8.20689265110667e-06, "loss": 0.449, "step": 15405 }, { "epoch": 1.05, "learning_rate": 8.2059500018853e-06, "loss": 0.5812, "step": 15410 }, { "epoch": 1.05, "learning_rate": 8.205007352663927e-06, "loss": 0.3322, "step": 15415 }, { "epoch": 1.05, "learning_rate": 8.204064703442556e-06, "loss": 0.9213, "step": 15420 }, { "epoch": 1.05, "learning_rate": 8.203122054221184e-06, "loss": 0.1632, "step": 15425 }, { "epoch": 1.05, "learning_rate": 8.202179404999813e-06, "loss": 0.8422, "step": 15430 }, { "epoch": 1.05, "learning_rate": 8.20123675577844e-06, "loss": 0.4985, "step": 15435 }, { "epoch": 1.05, "learning_rate": 8.200294106557069e-06, "loss": 0.209, "step": 15440 }, { "epoch": 1.05, "learning_rate": 8.199351457335696e-06, "loss": 0.4468, "step": 15445 }, { "epoch": 1.05, "learning_rate": 8.198408808114326e-06, "loss": 1.0566, "step": 15450 }, { "epoch": 1.05, "learning_rate": 8.197466158892953e-06, "loss": 0.6695, "step": 15455 }, { "epoch": 1.05, "learning_rate": 8.196523509671581e-06, "loss": 0.2329, "step": 15460 }, { "epoch": 1.05, "learning_rate": 8.19558086045021e-06, "loss": 0.4514, "step": 15465 }, { "epoch": 1.05, "learning_rate": 8.194638211228838e-06, "loss": 0.4444, "step": 15470 }, { "epoch": 1.05, "learning_rate": 8.193695562007467e-06, "loss": 0.3297, "step": 15475 }, { "epoch": 1.05, "learning_rate": 8.192752912786095e-06, "loss": 0.5114, "step": 15480 }, { "epoch": 1.05, "learning_rate": 8.191810263564724e-06, "loss": 0.3593, "step": 15485 }, { "epoch": 1.05, "learning_rate": 8.19086761434335e-06, "loss": 0.1719, "step": 15490 }, { "epoch": 1.05, "learning_rate": 8.18992496512198e-06, "loss": 0.244, "step": 15495 }, { "epoch": 1.05, "learning_rate": 8.188982315900607e-06, "loss": 0.3778, "step": 15500 }, { "epoch": 1.05, "learning_rate": 8.188039666679236e-06, "loss": 0.1648, "step": 15505 }, { "epoch": 1.05, "learning_rate": 8.187097017457864e-06, "loss": 0.3136, "step": 15510 }, { "epoch": 1.05, "learning_rate": 8.186154368236493e-06, "loss": 1.1807, "step": 15515 }, { "epoch": 1.05, "learning_rate": 8.185211719015121e-06, "loss": 0.4892, "step": 15520 }, { "epoch": 1.05, "learning_rate": 8.184269069793749e-06, "loss": 0.6401, "step": 15525 }, { "epoch": 1.05, "learning_rate": 8.183326420572378e-06, "loss": 0.4123, "step": 15530 }, { "epoch": 1.05, "learning_rate": 8.182383771351006e-06, "loss": 0.4657, "step": 15535 }, { "epoch": 1.05, "learning_rate": 8.181441122129633e-06, "loss": 0.2195, "step": 15540 }, { "epoch": 1.06, "learning_rate": 8.180498472908263e-06, "loss": 0.7465, "step": 15545 }, { "epoch": 1.06, "learning_rate": 8.17955582368689e-06, "loss": 0.2942, "step": 15550 }, { "epoch": 1.06, "learning_rate": 8.178613174465518e-06, "loss": 0.7463, "step": 15555 }, { "epoch": 1.06, "learning_rate": 8.177670525244147e-06, "loss": 0.4752, "step": 15560 }, { "epoch": 1.06, "learning_rate": 8.176727876022775e-06, "loss": 0.523, "step": 15565 }, { "epoch": 1.06, "learning_rate": 8.175785226801404e-06, "loss": 0.2562, "step": 15570 }, { "epoch": 1.06, "learning_rate": 8.174842577580032e-06, "loss": 0.3605, "step": 15575 }, { "epoch": 1.06, "learning_rate": 8.17389992835866e-06, "loss": 0.4296, "step": 15580 }, { "epoch": 1.06, "learning_rate": 8.172957279137287e-06, "loss": 0.3607, "step": 15585 }, { "epoch": 1.06, "learning_rate": 8.172014629915917e-06, "loss": 0.426, "step": 15590 }, { "epoch": 1.06, "learning_rate": 8.171071980694544e-06, "loss": 0.4226, "step": 15595 }, { "epoch": 1.06, "learning_rate": 8.170129331473173e-06, "loss": 0.1922, "step": 15600 }, { "epoch": 1.06, "learning_rate": 8.169186682251801e-06, "loss": 0.1816, "step": 15605 }, { "epoch": 1.06, "learning_rate": 8.168244033030429e-06, "loss": 0.6085, "step": 15610 }, { "epoch": 1.06, "learning_rate": 8.167301383809058e-06, "loss": 0.4085, "step": 15615 }, { "epoch": 1.06, "learning_rate": 8.166358734587686e-06, "loss": 0.5499, "step": 15620 }, { "epoch": 1.06, "learning_rate": 8.165416085366315e-06, "loss": 0.2573, "step": 15625 }, { "epoch": 1.06, "learning_rate": 8.164473436144943e-06, "loss": 0.7874, "step": 15630 }, { "epoch": 1.06, "learning_rate": 8.16353078692357e-06, "loss": 0.7267, "step": 15635 }, { "epoch": 1.06, "learning_rate": 8.162588137702198e-06, "loss": 0.1905, "step": 15640 }, { "epoch": 1.06, "learning_rate": 8.161645488480827e-06, "loss": 0.0808, "step": 15645 }, { "epoch": 1.06, "learning_rate": 8.160702839259455e-06, "loss": 0.5344, "step": 15650 }, { "epoch": 1.06, "learning_rate": 8.159760190038084e-06, "loss": 0.6337, "step": 15655 }, { "epoch": 1.06, "learning_rate": 8.158817540816712e-06, "loss": 0.2893, "step": 15660 }, { "epoch": 1.06, "learning_rate": 8.15787489159534e-06, "loss": 0.3462, "step": 15665 }, { "epoch": 1.06, "learning_rate": 8.156932242373969e-06, "loss": 0.3652, "step": 15670 }, { "epoch": 1.06, "learning_rate": 8.155989593152597e-06, "loss": 0.4903, "step": 15675 }, { "epoch": 1.06, "learning_rate": 8.155046943931226e-06, "loss": 0.549, "step": 15680 }, { "epoch": 1.06, "learning_rate": 8.154104294709854e-06, "loss": 0.5249, "step": 15685 }, { "epoch": 1.06, "learning_rate": 8.153161645488481e-06, "loss": 0.7577, "step": 15690 }, { "epoch": 1.07, "learning_rate": 8.152218996267109e-06, "loss": 0.699, "step": 15695 }, { "epoch": 1.07, "learning_rate": 8.151276347045738e-06, "loss": 1.1918, "step": 15700 }, { "epoch": 1.07, "learning_rate": 8.150333697824366e-06, "loss": 0.2193, "step": 15705 }, { "epoch": 1.07, "learning_rate": 8.149391048602995e-06, "loss": 0.6272, "step": 15710 }, { "epoch": 1.07, "learning_rate": 8.148448399381623e-06, "loss": 0.3417, "step": 15715 }, { "epoch": 1.07, "learning_rate": 8.14750575016025e-06, "loss": 0.7081, "step": 15720 }, { "epoch": 1.07, "learning_rate": 8.14656310093888e-06, "loss": 1.0656, "step": 15725 }, { "epoch": 1.07, "learning_rate": 8.145620451717507e-06, "loss": 0.3287, "step": 15730 }, { "epoch": 1.07, "learning_rate": 8.144677802496137e-06, "loss": 1.2168, "step": 15735 }, { "epoch": 1.07, "learning_rate": 8.143735153274764e-06, "loss": 0.6103, "step": 15740 }, { "epoch": 1.07, "learning_rate": 8.142792504053392e-06, "loss": 0.8528, "step": 15745 }, { "epoch": 1.07, "learning_rate": 8.14184985483202e-06, "loss": 0.6082, "step": 15750 }, { "epoch": 1.07, "learning_rate": 8.140907205610649e-06, "loss": 0.6008, "step": 15755 }, { "epoch": 1.07, "learning_rate": 8.139964556389277e-06, "loss": 0.2005, "step": 15760 }, { "epoch": 1.07, "learning_rate": 8.139021907167906e-06, "loss": 0.692, "step": 15765 }, { "epoch": 1.07, "learning_rate": 8.138079257946534e-06, "loss": 0.3461, "step": 15770 }, { "epoch": 1.07, "learning_rate": 8.137136608725161e-06, "loss": 0.8102, "step": 15775 }, { "epoch": 1.07, "learning_rate": 8.13619395950379e-06, "loss": 0.8104, "step": 15780 }, { "epoch": 1.07, "learning_rate": 8.135251310282418e-06, "loss": 0.4232, "step": 15785 }, { "epoch": 1.07, "learning_rate": 8.134308661061046e-06, "loss": 0.9787, "step": 15790 }, { "epoch": 1.07, "learning_rate": 8.133366011839675e-06, "loss": 0.0747, "step": 15795 }, { "epoch": 1.07, "learning_rate": 8.132423362618303e-06, "loss": 0.6953, "step": 15800 }, { "epoch": 1.07, "learning_rate": 8.13148071339693e-06, "loss": 0.6786, "step": 15805 }, { "epoch": 1.07, "learning_rate": 8.13053806417556e-06, "loss": 0.4292, "step": 15810 }, { "epoch": 1.07, "learning_rate": 8.129595414954187e-06, "loss": 0.3513, "step": 15815 }, { "epoch": 1.07, "learning_rate": 8.128652765732817e-06, "loss": 1.0794, "step": 15820 }, { "epoch": 1.07, "learning_rate": 8.127710116511444e-06, "loss": 0.4251, "step": 15825 }, { "epoch": 1.07, "learning_rate": 8.126767467290074e-06, "loss": 0.8846, "step": 15830 }, { "epoch": 1.07, "learning_rate": 8.1258248180687e-06, "loss": 0.7757, "step": 15835 }, { "epoch": 1.08, "learning_rate": 8.124882168847329e-06, "loss": 0.5581, "step": 15840 }, { "epoch": 1.08, "learning_rate": 8.123939519625957e-06, "loss": 0.5376, "step": 15845 }, { "epoch": 1.08, "learning_rate": 8.122996870404586e-06, "loss": 0.5579, "step": 15850 }, { "epoch": 1.08, "learning_rate": 8.122054221183214e-06, "loss": 0.8964, "step": 15855 }, { "epoch": 1.08, "learning_rate": 8.121111571961843e-06, "loss": 0.2161, "step": 15860 }, { "epoch": 1.08, "learning_rate": 8.12016892274047e-06, "loss": 0.8826, "step": 15865 }, { "epoch": 1.08, "learning_rate": 8.119226273519098e-06, "loss": 0.5975, "step": 15870 }, { "epoch": 1.08, "learning_rate": 8.118283624297728e-06, "loss": 0.621, "step": 15875 }, { "epoch": 1.08, "learning_rate": 8.117340975076355e-06, "loss": 0.8644, "step": 15880 }, { "epoch": 1.08, "learning_rate": 8.116398325854985e-06, "loss": 0.635, "step": 15885 }, { "epoch": 1.08, "learning_rate": 8.11545567663361e-06, "loss": 0.8446, "step": 15890 }, { "epoch": 1.08, "learning_rate": 8.11451302741224e-06, "loss": 0.4999, "step": 15895 }, { "epoch": 1.08, "learning_rate": 8.113570378190867e-06, "loss": 0.7761, "step": 15900 }, { "epoch": 1.08, "learning_rate": 8.112627728969497e-06, "loss": 0.3313, "step": 15905 }, { "epoch": 1.08, "learning_rate": 8.111685079748124e-06, "loss": 0.6212, "step": 15910 }, { "epoch": 1.08, "learning_rate": 8.110742430526754e-06, "loss": 0.5297, "step": 15915 }, { "epoch": 1.08, "learning_rate": 8.109799781305381e-06, "loss": 0.5836, "step": 15920 }, { "epoch": 1.08, "learning_rate": 8.108857132084009e-06, "loss": 0.4738, "step": 15925 }, { "epoch": 1.08, "learning_rate": 8.107914482862638e-06, "loss": 0.4074, "step": 15930 }, { "epoch": 1.08, "learning_rate": 8.106971833641266e-06, "loss": 0.198, "step": 15935 }, { "epoch": 1.08, "learning_rate": 8.106029184419895e-06, "loss": 0.4732, "step": 15940 }, { "epoch": 1.08, "learning_rate": 8.105086535198523e-06, "loss": 0.6353, "step": 15945 }, { "epoch": 1.08, "learning_rate": 8.10414388597715e-06, "loss": 0.8597, "step": 15950 }, { "epoch": 1.08, "learning_rate": 8.103201236755778e-06, "loss": 0.4417, "step": 15955 }, { "epoch": 1.08, "learning_rate": 8.102258587534408e-06, "loss": 0.5251, "step": 15960 }, { "epoch": 1.08, "learning_rate": 8.101315938313035e-06, "loss": 0.4959, "step": 15965 }, { "epoch": 1.08, "learning_rate": 8.100373289091665e-06, "loss": 0.4951, "step": 15970 }, { "epoch": 1.08, "learning_rate": 8.099430639870292e-06, "loss": 0.1976, "step": 15975 }, { "epoch": 1.08, "learning_rate": 8.09848799064892e-06, "loss": 0.8961, "step": 15980 }, { "epoch": 1.08, "learning_rate": 8.09754534142755e-06, "loss": 0.2527, "step": 15985 }, { "epoch": 1.09, "learning_rate": 8.096602692206177e-06, "loss": 0.4029, "step": 15990 }, { "epoch": 1.09, "learning_rate": 8.095660042984806e-06, "loss": 0.4654, "step": 15995 }, { "epoch": 1.09, "learning_rate": 8.094717393763434e-06, "loss": 0.2977, "step": 16000 }, { "epoch": 1.09, "learning_rate": 8.093774744542061e-06, "loss": 0.6518, "step": 16005 }, { "epoch": 1.09, "learning_rate": 8.092832095320689e-06, "loss": 0.9729, "step": 16010 }, { "epoch": 1.09, "learning_rate": 8.091889446099318e-06, "loss": 0.4921, "step": 16015 }, { "epoch": 1.09, "learning_rate": 8.090946796877946e-06, "loss": 0.5674, "step": 16020 }, { "epoch": 1.09, "learning_rate": 8.090004147656575e-06, "loss": 0.2978, "step": 16025 }, { "epoch": 1.09, "learning_rate": 8.089061498435203e-06, "loss": 0.3454, "step": 16030 }, { "epoch": 1.09, "learning_rate": 8.08811884921383e-06, "loss": 0.4594, "step": 16035 }, { "epoch": 1.09, "learning_rate": 8.087176199992458e-06, "loss": 0.511, "step": 16040 }, { "epoch": 1.09, "learning_rate": 8.086233550771088e-06, "loss": 0.3505, "step": 16045 }, { "epoch": 1.09, "learning_rate": 8.085290901549715e-06, "loss": 0.3863, "step": 16050 }, { "epoch": 1.09, "learning_rate": 8.084348252328345e-06, "loss": 0.3329, "step": 16055 }, { "epoch": 1.09, "learning_rate": 8.083405603106972e-06, "loss": 0.6335, "step": 16060 }, { "epoch": 1.09, "learning_rate": 8.0824629538856e-06, "loss": 0.3136, "step": 16065 }, { "epoch": 1.09, "learning_rate": 8.08152030466423e-06, "loss": 0.5825, "step": 16070 }, { "epoch": 1.09, "learning_rate": 8.080577655442857e-06, "loss": 0.5879, "step": 16075 }, { "epoch": 1.09, "learning_rate": 8.079635006221486e-06, "loss": 0.3154, "step": 16080 }, { "epoch": 1.09, "learning_rate": 8.078692357000114e-06, "loss": 1.1243, "step": 16085 }, { "epoch": 1.09, "learning_rate": 8.077749707778743e-06, "loss": 0.2925, "step": 16090 }, { "epoch": 1.09, "learning_rate": 8.07680705855737e-06, "loss": 0.6165, "step": 16095 }, { "epoch": 1.09, "learning_rate": 8.075864409335998e-06, "loss": 0.2121, "step": 16100 }, { "epoch": 1.09, "learning_rate": 8.074921760114626e-06, "loss": 0.2271, "step": 16105 }, { "epoch": 1.09, "learning_rate": 8.073979110893255e-06, "loss": 1.1623, "step": 16110 }, { "epoch": 1.09, "learning_rate": 8.073036461671883e-06, "loss": 0.9933, "step": 16115 }, { "epoch": 1.09, "learning_rate": 8.07209381245051e-06, "loss": 0.4139, "step": 16120 }, { "epoch": 1.09, "learning_rate": 8.07115116322914e-06, "loss": 0.9433, "step": 16125 }, { "epoch": 1.09, "learning_rate": 8.070208514007768e-06, "loss": 0.6025, "step": 16130 }, { "epoch": 1.1, "learning_rate": 8.069265864786397e-06, "loss": 0.2775, "step": 16135 }, { "epoch": 1.1, "learning_rate": 8.068323215565025e-06, "loss": 0.7024, "step": 16140 }, { "epoch": 1.1, "learning_rate": 8.067380566343654e-06, "loss": 0.5959, "step": 16145 }, { "epoch": 1.1, "learning_rate": 8.06643791712228e-06, "loss": 0.3791, "step": 16150 }, { "epoch": 1.1, "learning_rate": 8.06549526790091e-06, "loss": 0.5673, "step": 16155 }, { "epoch": 1.1, "learning_rate": 8.064552618679537e-06, "loss": 1.4324, "step": 16160 }, { "epoch": 1.1, "learning_rate": 8.063609969458166e-06, "loss": 0.7095, "step": 16165 }, { "epoch": 1.1, "learning_rate": 8.062667320236794e-06, "loss": 0.2077, "step": 16170 }, { "epoch": 1.1, "learning_rate": 8.061724671015423e-06, "loss": 0.5915, "step": 16175 }, { "epoch": 1.1, "learning_rate": 8.060782021794051e-06, "loss": 0.3157, "step": 16180 }, { "epoch": 1.1, "learning_rate": 8.059839372572679e-06, "loss": 0.3697, "step": 16185 }, { "epoch": 1.1, "learning_rate": 8.058896723351308e-06, "loss": 0.3663, "step": 16190 }, { "epoch": 1.1, "learning_rate": 8.057954074129936e-06, "loss": 0.0912, "step": 16195 }, { "epoch": 1.1, "learning_rate": 8.057011424908565e-06, "loss": 0.9167, "step": 16200 }, { "epoch": 1.1, "learning_rate": 8.056068775687192e-06, "loss": 0.2252, "step": 16205 }, { "epoch": 1.1, "learning_rate": 8.05512612646582e-06, "loss": 0.2577, "step": 16210 }, { "epoch": 1.1, "learning_rate": 8.054183477244448e-06, "loss": 1.2529, "step": 16215 }, { "epoch": 1.1, "learning_rate": 8.053240828023077e-06, "loss": 0.1273, "step": 16220 }, { "epoch": 1.1, "learning_rate": 8.052298178801705e-06, "loss": 0.5074, "step": 16225 }, { "epoch": 1.1, "learning_rate": 8.051355529580334e-06, "loss": 0.4422, "step": 16230 }, { "epoch": 1.1, "learning_rate": 8.050412880358962e-06, "loss": 0.3189, "step": 16235 }, { "epoch": 1.1, "learning_rate": 8.04947023113759e-06, "loss": 0.5726, "step": 16240 }, { "epoch": 1.1, "learning_rate": 8.048527581916219e-06, "loss": 0.2617, "step": 16245 }, { "epoch": 1.1, "learning_rate": 8.047584932694846e-06, "loss": 0.5122, "step": 16250 }, { "epoch": 1.1, "learning_rate": 8.046642283473474e-06, "loss": 0.4011, "step": 16255 }, { "epoch": 1.1, "learning_rate": 8.045699634252103e-06, "loss": 0.6239, "step": 16260 }, { "epoch": 1.1, "learning_rate": 8.044756985030731e-06, "loss": 0.6395, "step": 16265 }, { "epoch": 1.1, "learning_rate": 8.043814335809359e-06, "loss": 0.5336, "step": 16270 }, { "epoch": 1.1, "learning_rate": 8.042871686587988e-06, "loss": 0.7736, "step": 16275 }, { "epoch": 1.1, "learning_rate": 8.041929037366616e-06, "loss": 0.3725, "step": 16280 }, { "epoch": 1.11, "learning_rate": 8.040986388145245e-06, "loss": 0.6668, "step": 16285 }, { "epoch": 1.11, "learning_rate": 8.040043738923873e-06, "loss": 0.3568, "step": 16290 }, { "epoch": 1.11, "learning_rate": 8.0391010897025e-06, "loss": 0.8446, "step": 16295 }, { "epoch": 1.11, "learning_rate": 8.038158440481128e-06, "loss": 0.5101, "step": 16300 }, { "epoch": 1.11, "learning_rate": 8.037215791259757e-06, "loss": 0.6472, "step": 16305 }, { "epoch": 1.11, "learning_rate": 8.036273142038385e-06, "loss": 0.3554, "step": 16310 }, { "epoch": 1.11, "learning_rate": 8.035330492817014e-06, "loss": 0.3706, "step": 16315 }, { "epoch": 1.11, "learning_rate": 8.034387843595642e-06, "loss": 0.3581, "step": 16320 }, { "epoch": 1.11, "learning_rate": 8.03344519437427e-06, "loss": 0.3165, "step": 16325 }, { "epoch": 1.11, "learning_rate": 8.032502545152899e-06, "loss": 0.2678, "step": 16330 }, { "epoch": 1.11, "learning_rate": 8.031559895931526e-06, "loss": 0.456, "step": 16335 }, { "epoch": 1.11, "learning_rate": 8.030617246710156e-06, "loss": 0.3199, "step": 16340 }, { "epoch": 1.11, "learning_rate": 8.029674597488783e-06, "loss": 0.3083, "step": 16345 }, { "epoch": 1.11, "learning_rate": 8.028731948267411e-06, "loss": 0.4967, "step": 16350 }, { "epoch": 1.11, "learning_rate": 8.027789299046039e-06, "loss": 0.0937, "step": 16355 }, { "epoch": 1.11, "learning_rate": 8.026846649824668e-06, "loss": 0.7596, "step": 16360 }, { "epoch": 1.11, "learning_rate": 8.025904000603296e-06, "loss": 1.4712, "step": 16365 }, { "epoch": 1.11, "learning_rate": 8.024961351381925e-06, "loss": 0.5752, "step": 16370 }, { "epoch": 1.11, "learning_rate": 8.024018702160553e-06, "loss": 0.6853, "step": 16375 }, { "epoch": 1.11, "learning_rate": 8.02307605293918e-06, "loss": 0.7109, "step": 16380 }, { "epoch": 1.11, "learning_rate": 8.02213340371781e-06, "loss": 0.5927, "step": 16385 }, { "epoch": 1.11, "learning_rate": 8.021190754496437e-06, "loss": 0.7634, "step": 16390 }, { "epoch": 1.11, "learning_rate": 8.020248105275067e-06, "loss": 0.2528, "step": 16395 }, { "epoch": 1.11, "learning_rate": 8.019305456053694e-06, "loss": 0.5179, "step": 16400 }, { "epoch": 1.11, "learning_rate": 8.018362806832324e-06, "loss": 1.0913, "step": 16405 }, { "epoch": 1.11, "learning_rate": 8.01742015761095e-06, "loss": 0.5507, "step": 16410 }, { "epoch": 1.11, "learning_rate": 8.016477508389579e-06, "loss": 0.3475, "step": 16415 }, { "epoch": 1.11, "learning_rate": 8.015534859168206e-06, "loss": 0.1902, "step": 16420 }, { "epoch": 1.11, "learning_rate": 8.014592209946836e-06, "loss": 0.6739, "step": 16425 }, { "epoch": 1.12, "learning_rate": 8.013649560725463e-06, "loss": 0.4167, "step": 16430 }, { "epoch": 1.12, "learning_rate": 8.012706911504093e-06, "loss": 0.4855, "step": 16435 }, { "epoch": 1.12, "learning_rate": 8.01176426228272e-06, "loss": 0.7499, "step": 16440 }, { "epoch": 1.12, "learning_rate": 8.010821613061348e-06, "loss": 0.5491, "step": 16445 }, { "epoch": 1.12, "learning_rate": 8.009878963839977e-06, "loss": 0.3764, "step": 16450 }, { "epoch": 1.12, "learning_rate": 8.008936314618605e-06, "loss": 0.2042, "step": 16455 }, { "epoch": 1.12, "learning_rate": 8.007993665397233e-06, "loss": 0.3237, "step": 16460 }, { "epoch": 1.12, "learning_rate": 8.00705101617586e-06, "loss": 0.3285, "step": 16465 }, { "epoch": 1.12, "learning_rate": 8.00610836695449e-06, "loss": 0.4416, "step": 16470 }, { "epoch": 1.12, "learning_rate": 8.005165717733117e-06, "loss": 0.2495, "step": 16475 }, { "epoch": 1.12, "learning_rate": 8.004223068511747e-06, "loss": 0.4077, "step": 16480 }, { "epoch": 1.12, "learning_rate": 8.003280419290374e-06, "loss": 0.7066, "step": 16485 }, { "epoch": 1.12, "learning_rate": 8.002337770069004e-06, "loss": 0.3622, "step": 16490 }, { "epoch": 1.12, "learning_rate": 8.001395120847631e-06, "loss": 0.4071, "step": 16495 }, { "epoch": 1.12, "learning_rate": 8.000452471626259e-06, "loss": 0.2715, "step": 16500 }, { "epoch": 1.12, "learning_rate": 7.999509822404886e-06, "loss": 0.4994, "step": 16505 }, { "epoch": 1.12, "learning_rate": 7.998567173183516e-06, "loss": 0.5467, "step": 16510 }, { "epoch": 1.12, "learning_rate": 7.997624523962143e-06, "loss": 0.1999, "step": 16515 }, { "epoch": 1.12, "learning_rate": 7.996681874740773e-06, "loss": 0.6452, "step": 16520 }, { "epoch": 1.12, "learning_rate": 7.9957392255194e-06, "loss": 0.5271, "step": 16525 }, { "epoch": 1.12, "learning_rate": 7.994796576298028e-06, "loss": 0.5094, "step": 16530 }, { "epoch": 1.12, "learning_rate": 7.993853927076657e-06, "loss": 1.2215, "step": 16535 }, { "epoch": 1.12, "learning_rate": 7.992911277855285e-06, "loss": 0.6983, "step": 16540 }, { "epoch": 1.12, "learning_rate": 7.991968628633914e-06, "loss": 0.673, "step": 16545 }, { "epoch": 1.12, "learning_rate": 7.991025979412542e-06, "loss": 0.5661, "step": 16550 }, { "epoch": 1.12, "learning_rate": 7.99008333019117e-06, "loss": 0.4316, "step": 16555 }, { "epoch": 1.12, "learning_rate": 7.989140680969797e-06, "loss": 0.4874, "step": 16560 }, { "epoch": 1.12, "learning_rate": 7.988198031748427e-06, "loss": 1.0443, "step": 16565 }, { "epoch": 1.12, "learning_rate": 7.987255382527054e-06, "loss": 0.4391, "step": 16570 }, { "epoch": 1.12, "learning_rate": 7.986312733305684e-06, "loss": 0.6046, "step": 16575 }, { "epoch": 1.13, "learning_rate": 7.985370084084311e-06, "loss": 0.0919, "step": 16580 }, { "epoch": 1.13, "learning_rate": 7.984427434862939e-06, "loss": 0.7854, "step": 16585 }, { "epoch": 1.13, "learning_rate": 7.983484785641568e-06, "loss": 0.7643, "step": 16590 }, { "epoch": 1.13, "learning_rate": 7.982542136420196e-06, "loss": 0.3194, "step": 16595 }, { "epoch": 1.13, "learning_rate": 7.981599487198825e-06, "loss": 1.2053, "step": 16600 }, { "epoch": 1.13, "learning_rate": 7.980656837977453e-06, "loss": 0.42, "step": 16605 }, { "epoch": 1.13, "learning_rate": 7.97971418875608e-06, "loss": 0.6292, "step": 16610 }, { "epoch": 1.13, "learning_rate": 7.978771539534708e-06, "loss": 0.4315, "step": 16615 }, { "epoch": 1.13, "learning_rate": 7.977828890313337e-06, "loss": 0.3409, "step": 16620 }, { "epoch": 1.13, "learning_rate": 7.976886241091965e-06, "loss": 0.1349, "step": 16625 }, { "epoch": 1.13, "learning_rate": 7.975943591870594e-06, "loss": 0.8779, "step": 16630 }, { "epoch": 1.13, "learning_rate": 7.975000942649222e-06, "loss": 0.293, "step": 16635 }, { "epoch": 1.13, "learning_rate": 7.97405829342785e-06, "loss": 0.1848, "step": 16640 }, { "epoch": 1.13, "learning_rate": 7.973115644206479e-06, "loss": 0.5339, "step": 16645 }, { "epoch": 1.13, "learning_rate": 7.972172994985107e-06, "loss": 0.2054, "step": 16650 }, { "epoch": 1.13, "learning_rate": 7.971230345763736e-06, "loss": 1.128, "step": 16655 }, { "epoch": 1.13, "learning_rate": 7.970287696542364e-06, "loss": 0.625, "step": 16660 }, { "epoch": 1.13, "learning_rate": 7.969345047320991e-06, "loss": 0.4142, "step": 16665 }, { "epoch": 1.13, "learning_rate": 7.968402398099619e-06, "loss": 0.4516, "step": 16670 }, { "epoch": 1.13, "learning_rate": 7.967459748878248e-06, "loss": 0.4425, "step": 16675 }, { "epoch": 1.13, "learning_rate": 7.966517099656876e-06, "loss": 0.5572, "step": 16680 }, { "epoch": 1.13, "learning_rate": 7.965574450435505e-06, "loss": 0.0844, "step": 16685 }, { "epoch": 1.13, "learning_rate": 7.964631801214133e-06, "loss": 0.4492, "step": 16690 }, { "epoch": 1.13, "learning_rate": 7.96368915199276e-06, "loss": 0.1568, "step": 16695 }, { "epoch": 1.13, "learning_rate": 7.96274650277139e-06, "loss": 0.7485, "step": 16700 }, { "epoch": 1.13, "learning_rate": 7.961803853550018e-06, "loss": 0.2816, "step": 16705 }, { "epoch": 1.13, "learning_rate": 7.960861204328645e-06, "loss": 0.7725, "step": 16710 }, { "epoch": 1.13, "learning_rate": 7.959918555107274e-06, "loss": 0.2602, "step": 16715 }, { "epoch": 1.13, "learning_rate": 7.958975905885902e-06, "loss": 0.1582, "step": 16720 }, { "epoch": 1.14, "learning_rate": 7.95803325666453e-06, "loss": 0.3441, "step": 16725 }, { "epoch": 1.14, "learning_rate": 7.957090607443159e-06, "loss": 0.2368, "step": 16730 }, { "epoch": 1.14, "learning_rate": 7.956147958221787e-06, "loss": 0.1932, "step": 16735 }, { "epoch": 1.14, "learning_rate": 7.955205309000416e-06, "loss": 0.6864, "step": 16740 }, { "epoch": 1.14, "learning_rate": 7.954262659779044e-06, "loss": 0.7815, "step": 16745 }, { "epoch": 1.14, "learning_rate": 7.953320010557673e-06, "loss": 0.601, "step": 16750 }, { "epoch": 1.14, "learning_rate": 7.952377361336299e-06, "loss": 0.2908, "step": 16755 }, { "epoch": 1.14, "learning_rate": 7.951434712114928e-06, "loss": 0.3628, "step": 16760 }, { "epoch": 1.14, "learning_rate": 7.950492062893556e-06, "loss": 0.4777, "step": 16765 }, { "epoch": 1.14, "learning_rate": 7.949549413672185e-06, "loss": 0.3119, "step": 16770 }, { "epoch": 1.14, "learning_rate": 7.948606764450813e-06, "loss": 0.2429, "step": 16775 }, { "epoch": 1.14, "learning_rate": 7.94766411522944e-06, "loss": 0.5856, "step": 16780 }, { "epoch": 1.14, "learning_rate": 7.94672146600807e-06, "loss": 0.358, "step": 16785 }, { "epoch": 1.14, "learning_rate": 7.945778816786698e-06, "loss": 0.3699, "step": 16790 }, { "epoch": 1.14, "learning_rate": 7.944836167565327e-06, "loss": 0.6782, "step": 16795 }, { "epoch": 1.14, "learning_rate": 7.943893518343955e-06, "loss": 0.7273, "step": 16800 }, { "epoch": 1.14, "learning_rate": 7.942950869122584e-06, "loss": 0.262, "step": 16805 }, { "epoch": 1.14, "learning_rate": 7.94200821990121e-06, "loss": 0.2787, "step": 16810 }, { "epoch": 1.14, "learning_rate": 7.941065570679839e-06, "loss": 0.6174, "step": 16815 }, { "epoch": 1.14, "learning_rate": 7.940122921458467e-06, "loss": 0.245, "step": 16820 }, { "epoch": 1.14, "learning_rate": 7.939180272237096e-06, "loss": 0.821, "step": 16825 }, { "epoch": 1.14, "learning_rate": 7.938237623015724e-06, "loss": 0.726, "step": 16830 }, { "epoch": 1.14, "learning_rate": 7.937294973794353e-06, "loss": 0.5326, "step": 16835 }, { "epoch": 1.14, "learning_rate": 7.93635232457298e-06, "loss": 0.63, "step": 16840 }, { "epoch": 1.14, "learning_rate": 7.935409675351608e-06, "loss": 0.4678, "step": 16845 }, { "epoch": 1.14, "learning_rate": 7.934467026130238e-06, "loss": 0.2463, "step": 16850 }, { "epoch": 1.14, "learning_rate": 7.933524376908865e-06, "loss": 0.4663, "step": 16855 }, { "epoch": 1.14, "learning_rate": 7.932581727687495e-06, "loss": 0.4202, "step": 16860 }, { "epoch": 1.14, "learning_rate": 7.931639078466122e-06, "loss": 0.327, "step": 16865 }, { "epoch": 1.14, "learning_rate": 7.93069642924475e-06, "loss": 0.2783, "step": 16870 }, { "epoch": 1.15, "learning_rate": 7.929753780023378e-06, "loss": 0.4329, "step": 16875 }, { "epoch": 1.15, "learning_rate": 7.928811130802007e-06, "loss": 0.802, "step": 16880 }, { "epoch": 1.15, "learning_rate": 7.927868481580635e-06, "loss": 0.7448, "step": 16885 }, { "epoch": 1.15, "learning_rate": 7.926925832359264e-06, "loss": 0.4057, "step": 16890 }, { "epoch": 1.15, "learning_rate": 7.925983183137892e-06, "loss": 0.4411, "step": 16895 }, { "epoch": 1.15, "learning_rate": 7.92504053391652e-06, "loss": 0.5636, "step": 16900 }, { "epoch": 1.15, "learning_rate": 7.924097884695149e-06, "loss": 0.6652, "step": 16905 }, { "epoch": 1.15, "learning_rate": 7.923155235473776e-06, "loss": 0.7367, "step": 16910 }, { "epoch": 1.15, "learning_rate": 7.922212586252404e-06, "loss": 1.2892, "step": 16915 }, { "epoch": 1.15, "learning_rate": 7.921269937031033e-06, "loss": 0.507, "step": 16920 }, { "epoch": 1.15, "learning_rate": 7.92032728780966e-06, "loss": 0.4469, "step": 16925 }, { "epoch": 1.15, "learning_rate": 7.919384638588288e-06, "loss": 0.5969, "step": 16930 }, { "epoch": 1.15, "learning_rate": 7.918441989366918e-06, "loss": 0.8503, "step": 16935 }, { "epoch": 1.15, "learning_rate": 7.917499340145545e-06, "loss": 0.5537, "step": 16940 }, { "epoch": 1.15, "learning_rate": 7.916556690924175e-06, "loss": 0.7089, "step": 16945 }, { "epoch": 1.15, "learning_rate": 7.915614041702802e-06, "loss": 0.543, "step": 16950 }, { "epoch": 1.15, "learning_rate": 7.91467139248143e-06, "loss": 0.1004, "step": 16955 }, { "epoch": 1.15, "learning_rate": 7.913728743260058e-06, "loss": 0.4524, "step": 16960 }, { "epoch": 1.15, "learning_rate": 7.912786094038687e-06, "loss": 0.371, "step": 16965 }, { "epoch": 1.15, "learning_rate": 7.911843444817315e-06, "loss": 0.7315, "step": 16970 }, { "epoch": 1.15, "learning_rate": 7.910900795595944e-06, "loss": 0.4427, "step": 16975 }, { "epoch": 1.15, "learning_rate": 7.909958146374572e-06, "loss": 0.4753, "step": 16980 }, { "epoch": 1.15, "learning_rate": 7.9090154971532e-06, "loss": 0.3972, "step": 16985 }, { "epoch": 1.15, "learning_rate": 7.908072847931829e-06, "loss": 0.3593, "step": 16990 }, { "epoch": 1.15, "learning_rate": 7.907130198710456e-06, "loss": 0.8498, "step": 16995 }, { "epoch": 1.15, "learning_rate": 7.906187549489086e-06, "loss": 0.6821, "step": 17000 }, { "epoch": 1.15, "learning_rate": 7.905244900267713e-06, "loss": 0.6449, "step": 17005 }, { "epoch": 1.15, "learning_rate": 7.90430225104634e-06, "loss": 0.2659, "step": 17010 }, { "epoch": 1.15, "learning_rate": 7.903359601824968e-06, "loss": 0.2998, "step": 17015 }, { "epoch": 1.16, "learning_rate": 7.902416952603598e-06, "loss": 0.2823, "step": 17020 }, { "epoch": 1.16, "learning_rate": 7.901474303382225e-06, "loss": 0.2911, "step": 17025 }, { "epoch": 1.16, "learning_rate": 7.900531654160855e-06, "loss": 0.2247, "step": 17030 }, { "epoch": 1.16, "learning_rate": 7.899589004939482e-06, "loss": 0.1552, "step": 17035 }, { "epoch": 1.16, "learning_rate": 7.89864635571811e-06, "loss": 0.1957, "step": 17040 }, { "epoch": 1.16, "learning_rate": 7.89770370649674e-06, "loss": 1.2068, "step": 17045 }, { "epoch": 1.16, "learning_rate": 7.896761057275367e-06, "loss": 0.3141, "step": 17050 }, { "epoch": 1.16, "learning_rate": 7.895818408053996e-06, "loss": 0.9511, "step": 17055 }, { "epoch": 1.16, "learning_rate": 7.894875758832624e-06, "loss": 0.1584, "step": 17060 }, { "epoch": 1.16, "learning_rate": 7.893933109611253e-06, "loss": 0.6955, "step": 17065 }, { "epoch": 1.16, "learning_rate": 7.89299046038988e-06, "loss": 0.3727, "step": 17070 }, { "epoch": 1.16, "learning_rate": 7.892047811168509e-06, "loss": 0.5076, "step": 17075 }, { "epoch": 1.16, "learning_rate": 7.891105161947136e-06, "loss": 0.2726, "step": 17080 }, { "epoch": 1.16, "learning_rate": 7.890162512725766e-06, "loss": 0.5306, "step": 17085 }, { "epoch": 1.16, "learning_rate": 7.889219863504393e-06, "loss": 0.3653, "step": 17090 }, { "epoch": 1.16, "learning_rate": 7.888277214283023e-06, "loss": 0.4207, "step": 17095 }, { "epoch": 1.16, "learning_rate": 7.88733456506165e-06, "loss": 0.8752, "step": 17100 }, { "epoch": 1.16, "learning_rate": 7.886391915840278e-06, "loss": 0.0541, "step": 17105 }, { "epoch": 1.16, "learning_rate": 7.885449266618907e-06, "loss": 0.3012, "step": 17110 }, { "epoch": 1.16, "learning_rate": 7.884506617397535e-06, "loss": 0.6945, "step": 17115 }, { "epoch": 1.16, "learning_rate": 7.883563968176162e-06, "loss": 0.2624, "step": 17120 }, { "epoch": 1.16, "learning_rate": 7.88262131895479e-06, "loss": 0.1619, "step": 17125 }, { "epoch": 1.16, "learning_rate": 7.88167866973342e-06, "loss": 0.7491, "step": 17130 }, { "epoch": 1.16, "learning_rate": 7.880736020512047e-06, "loss": 0.2646, "step": 17135 }, { "epoch": 1.16, "learning_rate": 7.879793371290676e-06, "loss": 0.1552, "step": 17140 }, { "epoch": 1.16, "learning_rate": 7.878850722069304e-06, "loss": 0.0443, "step": 17145 }, { "epoch": 1.16, "learning_rate": 7.877908072847933e-06, "loss": 0.3716, "step": 17150 }, { "epoch": 1.16, "learning_rate": 7.876965423626561e-06, "loss": 0.4073, "step": 17155 }, { "epoch": 1.16, "learning_rate": 7.876022774405189e-06, "loss": 0.6588, "step": 17160 }, { "epoch": 1.16, "learning_rate": 7.875080125183816e-06, "loss": 0.6334, "step": 17165 }, { "epoch": 1.17, "learning_rate": 7.874137475962446e-06, "loss": 0.4197, "step": 17170 }, { "epoch": 1.17, "learning_rate": 7.873194826741073e-06, "loss": 0.1943, "step": 17175 }, { "epoch": 1.17, "learning_rate": 7.872252177519703e-06, "loss": 0.6161, "step": 17180 }, { "epoch": 1.17, "learning_rate": 7.87130952829833e-06, "loss": 0.3596, "step": 17185 }, { "epoch": 1.17, "learning_rate": 7.870366879076958e-06, "loss": 0.5153, "step": 17190 }, { "epoch": 1.17, "learning_rate": 7.869424229855587e-06, "loss": 0.179, "step": 17195 }, { "epoch": 1.17, "learning_rate": 7.868481580634215e-06, "loss": 0.4422, "step": 17200 }, { "epoch": 1.17, "learning_rate": 7.867538931412844e-06, "loss": 0.0928, "step": 17205 }, { "epoch": 1.17, "learning_rate": 7.866596282191472e-06, "loss": 0.1833, "step": 17210 }, { "epoch": 1.17, "learning_rate": 7.8656536329701e-06, "loss": 0.3872, "step": 17215 }, { "epoch": 1.17, "learning_rate": 7.864710983748727e-06, "loss": 0.4705, "step": 17220 }, { "epoch": 1.17, "learning_rate": 7.863768334527356e-06, "loss": 0.5395, "step": 17225 }, { "epoch": 1.17, "learning_rate": 7.862825685305984e-06, "loss": 0.622, "step": 17230 }, { "epoch": 1.17, "learning_rate": 7.861883036084613e-06, "loss": 0.9597, "step": 17235 }, { "epoch": 1.17, "learning_rate": 7.860940386863241e-06, "loss": 0.4239, "step": 17240 }, { "epoch": 1.17, "learning_rate": 7.859997737641869e-06, "loss": 0.671, "step": 17245 }, { "epoch": 1.17, "learning_rate": 7.859055088420498e-06, "loss": 0.5444, "step": 17250 }, { "epoch": 1.17, "learning_rate": 7.858112439199126e-06, "loss": 0.7291, "step": 17255 }, { "epoch": 1.17, "learning_rate": 7.857169789977755e-06, "loss": 0.5648, "step": 17260 }, { "epoch": 1.17, "learning_rate": 7.856227140756383e-06, "loss": 0.2709, "step": 17265 }, { "epoch": 1.17, "learning_rate": 7.85528449153501e-06, "loss": 0.7035, "step": 17270 }, { "epoch": 1.17, "learning_rate": 7.854341842313638e-06, "loss": 0.7002, "step": 17275 }, { "epoch": 1.17, "learning_rate": 7.853399193092267e-06, "loss": 0.5926, "step": 17280 }, { "epoch": 1.17, "learning_rate": 7.852456543870895e-06, "loss": 0.5707, "step": 17285 }, { "epoch": 1.17, "learning_rate": 7.851513894649524e-06, "loss": 0.3541, "step": 17290 }, { "epoch": 1.17, "learning_rate": 7.850571245428152e-06, "loss": 0.3119, "step": 17295 }, { "epoch": 1.17, "learning_rate": 7.84962859620678e-06, "loss": 0.4485, "step": 17300 }, { "epoch": 1.17, "learning_rate": 7.848685946985409e-06, "loss": 0.1587, "step": 17305 }, { "epoch": 1.17, "learning_rate": 7.847743297764037e-06, "loss": 0.5735, "step": 17310 }, { "epoch": 1.18, "learning_rate": 7.846800648542666e-06, "loss": 0.5651, "step": 17315 }, { "epoch": 1.18, "learning_rate": 7.845857999321293e-06, "loss": 0.3201, "step": 17320 }, { "epoch": 1.18, "learning_rate": 7.844915350099923e-06, "loss": 0.5419, "step": 17325 }, { "epoch": 1.18, "learning_rate": 7.843972700878549e-06, "loss": 0.5422, "step": 17330 }, { "epoch": 1.18, "learning_rate": 7.843030051657178e-06, "loss": 0.5649, "step": 17335 }, { "epoch": 1.18, "learning_rate": 7.842087402435806e-06, "loss": 0.4917, "step": 17340 }, { "epoch": 1.18, "learning_rate": 7.841144753214435e-06, "loss": 0.8255, "step": 17345 }, { "epoch": 1.18, "learning_rate": 7.840202103993063e-06, "loss": 0.3385, "step": 17350 }, { "epoch": 1.18, "learning_rate": 7.83925945477169e-06, "loss": 0.5303, "step": 17355 }, { "epoch": 1.18, "learning_rate": 7.83831680555032e-06, "loss": 0.5654, "step": 17360 }, { "epoch": 1.18, "learning_rate": 7.837374156328947e-06, "loss": 0.7063, "step": 17365 }, { "epoch": 1.18, "learning_rate": 7.836431507107577e-06, "loss": 0.374, "step": 17370 }, { "epoch": 1.18, "learning_rate": 7.835488857886204e-06, "loss": 0.4542, "step": 17375 }, { "epoch": 1.18, "learning_rate": 7.834546208664832e-06, "loss": 0.3315, "step": 17380 }, { "epoch": 1.18, "learning_rate": 7.83360355944346e-06, "loss": 0.4176, "step": 17385 }, { "epoch": 1.18, "learning_rate": 7.832660910222089e-06, "loss": 0.3392, "step": 17390 }, { "epoch": 1.18, "learning_rate": 7.831718261000717e-06, "loss": 0.8547, "step": 17395 }, { "epoch": 1.18, "learning_rate": 7.830775611779346e-06, "loss": 0.2647, "step": 17400 }, { "epoch": 1.18, "learning_rate": 7.829832962557974e-06, "loss": 0.6143, "step": 17405 }, { "epoch": 1.18, "learning_rate": 7.828890313336603e-06, "loss": 0.425, "step": 17410 }, { "epoch": 1.18, "learning_rate": 7.827947664115229e-06, "loss": 0.2509, "step": 17415 }, { "epoch": 1.18, "learning_rate": 7.827005014893858e-06, "loss": 0.3048, "step": 17420 }, { "epoch": 1.18, "learning_rate": 7.826062365672486e-06, "loss": 0.671, "step": 17425 }, { "epoch": 1.18, "learning_rate": 7.825119716451115e-06, "loss": 0.391, "step": 17430 }, { "epoch": 1.18, "learning_rate": 7.824177067229743e-06, "loss": 0.2453, "step": 17435 }, { "epoch": 1.18, "learning_rate": 7.823234418008372e-06, "loss": 0.4699, "step": 17440 }, { "epoch": 1.18, "learning_rate": 7.822291768787e-06, "loss": 0.5454, "step": 17445 }, { "epoch": 1.18, "learning_rate": 7.821349119565627e-06, "loss": 0.7461, "step": 17450 }, { "epoch": 1.18, "learning_rate": 7.820406470344257e-06, "loss": 0.7236, "step": 17455 }, { "epoch": 1.19, "learning_rate": 7.819463821122884e-06, "loss": 0.814, "step": 17460 }, { "epoch": 1.19, "learning_rate": 7.818521171901514e-06, "loss": 0.6757, "step": 17465 }, { "epoch": 1.19, "learning_rate": 7.81757852268014e-06, "loss": 0.1983, "step": 17470 }, { "epoch": 1.19, "learning_rate": 7.816635873458769e-06, "loss": 0.2957, "step": 17475 }, { "epoch": 1.19, "learning_rate": 7.815693224237397e-06, "loss": 0.2818, "step": 17480 }, { "epoch": 1.19, "learning_rate": 7.814750575016026e-06, "loss": 0.4241, "step": 17485 }, { "epoch": 1.19, "learning_rate": 7.813807925794654e-06, "loss": 0.8867, "step": 17490 }, { "epoch": 1.19, "learning_rate": 7.812865276573283e-06, "loss": 0.1927, "step": 17495 }, { "epoch": 1.19, "learning_rate": 7.81192262735191e-06, "loss": 0.2334, "step": 17500 }, { "epoch": 1.19, "learning_rate": 7.810979978130538e-06, "loss": 0.2678, "step": 17505 }, { "epoch": 1.19, "learning_rate": 7.810037328909168e-06, "loss": 0.8029, "step": 17510 }, { "epoch": 1.19, "learning_rate": 7.809094679687795e-06, "loss": 0.2815, "step": 17515 }, { "epoch": 1.19, "learning_rate": 7.808152030466425e-06, "loss": 0.4307, "step": 17520 }, { "epoch": 1.19, "learning_rate": 7.807209381245052e-06, "loss": 1.1649, "step": 17525 }, { "epoch": 1.19, "learning_rate": 7.80626673202368e-06, "loss": 0.6797, "step": 17530 }, { "epoch": 1.19, "learning_rate": 7.805324082802307e-06, "loss": 0.8063, "step": 17535 }, { "epoch": 1.19, "learning_rate": 7.804381433580937e-06, "loss": 1.4064, "step": 17540 }, { "epoch": 1.19, "learning_rate": 7.803438784359564e-06, "loss": 0.45, "step": 17545 }, { "epoch": 1.19, "learning_rate": 7.802496135138194e-06, "loss": 0.1929, "step": 17550 }, { "epoch": 1.19, "learning_rate": 7.801553485916821e-06, "loss": 0.5864, "step": 17555 }, { "epoch": 1.19, "learning_rate": 7.800610836695449e-06, "loss": 0.444, "step": 17560 }, { "epoch": 1.19, "learning_rate": 7.799668187474078e-06, "loss": 0.6795, "step": 17565 }, { "epoch": 1.19, "learning_rate": 7.798725538252706e-06, "loss": 0.3018, "step": 17570 }, { "epoch": 1.19, "learning_rate": 7.797782889031335e-06, "loss": 0.4682, "step": 17575 }, { "epoch": 1.19, "learning_rate": 7.796840239809963e-06, "loss": 0.5294, "step": 17580 }, { "epoch": 1.19, "learning_rate": 7.79589759058859e-06, "loss": 0.2546, "step": 17585 }, { "epoch": 1.19, "learning_rate": 7.794954941367218e-06, "loss": 0.694, "step": 17590 }, { "epoch": 1.19, "learning_rate": 7.794012292145848e-06, "loss": 0.8012, "step": 17595 }, { "epoch": 1.19, "learning_rate": 7.793069642924475e-06, "loss": 0.5264, "step": 17600 }, { "epoch": 1.19, "learning_rate": 7.792126993703105e-06, "loss": 0.6661, "step": 17605 }, { "epoch": 1.2, "learning_rate": 7.791184344481732e-06, "loss": 0.1779, "step": 17610 }, { "epoch": 1.2, "learning_rate": 7.79024169526036e-06, "loss": 0.1648, "step": 17615 }, { "epoch": 1.2, "learning_rate": 7.78929904603899e-06, "loss": 0.3144, "step": 17620 }, { "epoch": 1.2, "learning_rate": 7.788356396817617e-06, "loss": 0.7303, "step": 17625 }, { "epoch": 1.2, "learning_rate": 7.787413747596244e-06, "loss": 0.2929, "step": 17630 }, { "epoch": 1.2, "learning_rate": 7.786471098374874e-06, "loss": 0.4952, "step": 17635 }, { "epoch": 1.2, "learning_rate": 7.785528449153501e-06, "loss": 0.459, "step": 17640 }, { "epoch": 1.2, "learning_rate": 7.784585799932129e-06, "loss": 0.1244, "step": 17645 }, { "epoch": 1.2, "learning_rate": 7.783643150710758e-06, "loss": 0.3436, "step": 17650 }, { "epoch": 1.2, "learning_rate": 7.782700501489386e-06, "loss": 0.6107, "step": 17655 }, { "epoch": 1.2, "learning_rate": 7.781757852268015e-06, "loss": 0.3702, "step": 17660 }, { "epoch": 1.2, "learning_rate": 7.780815203046643e-06, "loss": 0.7482, "step": 17665 }, { "epoch": 1.2, "learning_rate": 7.77987255382527e-06, "loss": 0.4322, "step": 17670 }, { "epoch": 1.2, "learning_rate": 7.778929904603898e-06, "loss": 1.0618, "step": 17675 }, { "epoch": 1.2, "learning_rate": 7.777987255382528e-06, "loss": 0.3814, "step": 17680 }, { "epoch": 1.2, "learning_rate": 7.777044606161155e-06, "loss": 0.4011, "step": 17685 }, { "epoch": 1.2, "learning_rate": 7.776101956939785e-06, "loss": 0.1582, "step": 17690 }, { "epoch": 1.2, "learning_rate": 7.775159307718412e-06, "loss": 0.7107, "step": 17695 }, { "epoch": 1.2, "learning_rate": 7.77421665849704e-06, "loss": 0.4384, "step": 17700 }, { "epoch": 1.2, "learning_rate": 7.77327400927567e-06, "loss": 0.5034, "step": 17705 }, { "epoch": 1.2, "learning_rate": 7.772331360054297e-06, "loss": 0.307, "step": 17710 }, { "epoch": 1.2, "learning_rate": 7.771388710832926e-06, "loss": 0.4842, "step": 17715 }, { "epoch": 1.2, "learning_rate": 7.770446061611554e-06, "loss": 0.3946, "step": 17720 }, { "epoch": 1.2, "learning_rate": 7.769503412390183e-06, "loss": 0.6761, "step": 17725 }, { "epoch": 1.2, "learning_rate": 7.768560763168809e-06, "loss": 0.3252, "step": 17730 }, { "epoch": 1.2, "learning_rate": 7.767618113947438e-06, "loss": 1.1483, "step": 17735 }, { "epoch": 1.2, "learning_rate": 7.766675464726066e-06, "loss": 0.4827, "step": 17740 }, { "epoch": 1.2, "learning_rate": 7.765732815504695e-06, "loss": 0.2613, "step": 17745 }, { "epoch": 1.2, "learning_rate": 7.764790166283323e-06, "loss": 0.2151, "step": 17750 }, { "epoch": 1.21, "learning_rate": 7.763847517061952e-06, "loss": 0.2002, "step": 17755 }, { "epoch": 1.21, "learning_rate": 7.76290486784058e-06, "loss": 0.7482, "step": 17760 }, { "epoch": 1.21, "learning_rate": 7.761962218619208e-06, "loss": 0.598, "step": 17765 }, { "epoch": 1.21, "learning_rate": 7.761019569397837e-06, "loss": 1.122, "step": 17770 }, { "epoch": 1.21, "learning_rate": 7.760076920176465e-06, "loss": 0.6053, "step": 17775 }, { "epoch": 1.21, "learning_rate": 7.759134270955094e-06, "loss": 0.6259, "step": 17780 }, { "epoch": 1.21, "learning_rate": 7.75819162173372e-06, "loss": 0.4161, "step": 17785 }, { "epoch": 1.21, "learning_rate": 7.75724897251235e-06, "loss": 1.271, "step": 17790 }, { "epoch": 1.21, "learning_rate": 7.756306323290977e-06, "loss": 1.0077, "step": 17795 }, { "epoch": 1.21, "learning_rate": 7.755363674069606e-06, "loss": 0.6881, "step": 17800 }, { "epoch": 1.21, "learning_rate": 7.754421024848234e-06, "loss": 0.2676, "step": 17805 }, { "epoch": 1.21, "learning_rate": 7.753478375626863e-06, "loss": 0.5407, "step": 17810 }, { "epoch": 1.21, "learning_rate": 7.752535726405491e-06, "loss": 0.3912, "step": 17815 }, { "epoch": 1.21, "learning_rate": 7.751593077184119e-06, "loss": 0.5621, "step": 17820 }, { "epoch": 1.21, "learning_rate": 7.750650427962748e-06, "loss": 0.8024, "step": 17825 }, { "epoch": 1.21, "learning_rate": 7.749707778741375e-06, "loss": 0.2347, "step": 17830 }, { "epoch": 1.21, "learning_rate": 7.748765129520003e-06, "loss": 1.1405, "step": 17835 }, { "epoch": 1.21, "learning_rate": 7.747822480298632e-06, "loss": 0.3624, "step": 17840 }, { "epoch": 1.21, "learning_rate": 7.74687983107726e-06, "loss": 0.4071, "step": 17845 }, { "epoch": 1.21, "learning_rate": 7.745937181855888e-06, "loss": 0.8581, "step": 17850 }, { "epoch": 1.21, "learning_rate": 7.744994532634517e-06, "loss": 0.0966, "step": 17855 }, { "epoch": 1.21, "learning_rate": 7.744051883413145e-06, "loss": 0.4632, "step": 17860 }, { "epoch": 1.21, "learning_rate": 7.743109234191774e-06, "loss": 0.8084, "step": 17865 }, { "epoch": 1.21, "learning_rate": 7.742166584970402e-06, "loss": 0.6635, "step": 17870 }, { "epoch": 1.21, "learning_rate": 7.74122393574903e-06, "loss": 0.5738, "step": 17875 }, { "epoch": 1.21, "learning_rate": 7.740281286527657e-06, "loss": 0.639, "step": 17880 }, { "epoch": 1.21, "learning_rate": 7.739338637306286e-06, "loss": 0.6344, "step": 17885 }, { "epoch": 1.21, "learning_rate": 7.738395988084914e-06, "loss": 0.1582, "step": 17890 }, { "epoch": 1.21, "learning_rate": 7.737453338863543e-06, "loss": 0.517, "step": 17895 }, { "epoch": 1.21, "learning_rate": 7.736510689642171e-06, "loss": 0.6791, "step": 17900 }, { "epoch": 1.22, "learning_rate": 7.735568040420799e-06, "loss": 0.3687, "step": 17905 }, { "epoch": 1.22, "learning_rate": 7.734625391199428e-06, "loss": 0.4864, "step": 17910 }, { "epoch": 1.22, "learning_rate": 7.733682741978056e-06, "loss": 1.1996, "step": 17915 }, { "epoch": 1.22, "learning_rate": 7.732740092756685e-06, "loss": 0.6105, "step": 17920 }, { "epoch": 1.22, "learning_rate": 7.731797443535313e-06, "loss": 0.4077, "step": 17925 }, { "epoch": 1.22, "learning_rate": 7.73085479431394e-06, "loss": 0.5325, "step": 17930 }, { "epoch": 1.22, "learning_rate": 7.729912145092568e-06, "loss": 0.8004, "step": 17935 }, { "epoch": 1.22, "learning_rate": 7.728969495871197e-06, "loss": 0.9367, "step": 17940 }, { "epoch": 1.22, "learning_rate": 7.728026846649825e-06, "loss": 0.708, "step": 17945 }, { "epoch": 1.22, "learning_rate": 7.727084197428454e-06, "loss": 0.4869, "step": 17950 }, { "epoch": 1.22, "learning_rate": 7.726141548207082e-06, "loss": 0.5806, "step": 17955 }, { "epoch": 1.22, "learning_rate": 7.72519889898571e-06, "loss": 0.6514, "step": 17960 }, { "epoch": 1.22, "learning_rate": 7.724256249764339e-06, "loss": 0.0705, "step": 17965 }, { "epoch": 1.22, "learning_rate": 7.723313600542966e-06, "loss": 0.1811, "step": 17970 }, { "epoch": 1.22, "learning_rate": 7.722370951321596e-06, "loss": 0.9915, "step": 17975 }, { "epoch": 1.22, "learning_rate": 7.721428302100223e-06, "loss": 0.3107, "step": 17980 }, { "epoch": 1.22, "learning_rate": 7.720485652878853e-06, "loss": 0.1214, "step": 17985 }, { "epoch": 1.22, "learning_rate": 7.719543003657479e-06, "loss": 0.4551, "step": 17990 }, { "epoch": 1.22, "learning_rate": 7.718600354436108e-06, "loss": 0.4034, "step": 17995 }, { "epoch": 1.22, "learning_rate": 7.717657705214736e-06, "loss": 0.4236, "step": 18000 }, { "epoch": 1.22, "learning_rate": 7.716715055993365e-06, "loss": 0.9911, "step": 18005 }, { "epoch": 1.22, "learning_rate": 7.715772406771993e-06, "loss": 0.2336, "step": 18010 }, { "epoch": 1.22, "learning_rate": 7.71482975755062e-06, "loss": 0.5987, "step": 18015 }, { "epoch": 1.22, "learning_rate": 7.71388710832925e-06, "loss": 0.9314, "step": 18020 }, { "epoch": 1.22, "learning_rate": 7.712944459107877e-06, "loss": 0.7834, "step": 18025 }, { "epoch": 1.22, "learning_rate": 7.712001809886507e-06, "loss": 0.5984, "step": 18030 }, { "epoch": 1.22, "learning_rate": 7.711059160665134e-06, "loss": 0.4387, "step": 18035 }, { "epoch": 1.22, "learning_rate": 7.710116511443762e-06, "loss": 0.4195, "step": 18040 }, { "epoch": 1.22, "learning_rate": 7.70917386222239e-06, "loss": 0.1822, "step": 18045 }, { "epoch": 1.23, "learning_rate": 7.708231213001019e-06, "loss": 0.1412, "step": 18050 }, { "epoch": 1.23, "learning_rate": 7.707288563779646e-06, "loss": 1.0008, "step": 18055 }, { "epoch": 1.23, "learning_rate": 7.706345914558276e-06, "loss": 0.1965, "step": 18060 }, { "epoch": 1.23, "learning_rate": 7.705403265336903e-06, "loss": 0.6633, "step": 18065 }, { "epoch": 1.23, "learning_rate": 7.704460616115533e-06, "loss": 0.28, "step": 18070 }, { "epoch": 1.23, "learning_rate": 7.70351796689416e-06, "loss": 0.337, "step": 18075 }, { "epoch": 1.23, "learning_rate": 7.702575317672788e-06, "loss": 0.8039, "step": 18080 }, { "epoch": 1.23, "learning_rate": 7.701632668451416e-06, "loss": 0.49, "step": 18085 }, { "epoch": 1.23, "learning_rate": 7.700690019230045e-06, "loss": 0.7796, "step": 18090 }, { "epoch": 1.23, "learning_rate": 7.699747370008673e-06, "loss": 0.2087, "step": 18095 }, { "epoch": 1.23, "learning_rate": 7.698804720787302e-06, "loss": 0.3797, "step": 18100 }, { "epoch": 1.23, "learning_rate": 7.69786207156593e-06, "loss": 0.4062, "step": 18105 }, { "epoch": 1.23, "learning_rate": 7.696919422344557e-06, "loss": 0.8465, "step": 18110 }, { "epoch": 1.23, "learning_rate": 7.695976773123187e-06, "loss": 0.372, "step": 18115 }, { "epoch": 1.23, "learning_rate": 7.695034123901814e-06, "loss": 0.353, "step": 18120 }, { "epoch": 1.23, "learning_rate": 7.694091474680444e-06, "loss": 0.3386, "step": 18125 }, { "epoch": 1.23, "learning_rate": 7.69314882545907e-06, "loss": 0.1697, "step": 18130 }, { "epoch": 1.23, "learning_rate": 7.692206176237699e-06, "loss": 0.412, "step": 18135 }, { "epoch": 1.23, "learning_rate": 7.691263527016326e-06, "loss": 0.4878, "step": 18140 }, { "epoch": 1.23, "learning_rate": 7.690320877794956e-06, "loss": 0.4316, "step": 18145 }, { "epoch": 1.23, "learning_rate": 7.689378228573583e-06, "loss": 0.7525, "step": 18150 }, { "epoch": 1.23, "learning_rate": 7.688435579352213e-06, "loss": 0.3726, "step": 18155 }, { "epoch": 1.23, "learning_rate": 7.68749293013084e-06, "loss": 0.6033, "step": 18160 }, { "epoch": 1.23, "learning_rate": 7.686550280909468e-06, "loss": 0.4793, "step": 18165 }, { "epoch": 1.23, "learning_rate": 7.685607631688097e-06, "loss": 0.3845, "step": 18170 }, { "epoch": 1.23, "learning_rate": 7.684664982466725e-06, "loss": 0.335, "step": 18175 }, { "epoch": 1.23, "learning_rate": 7.683722333245354e-06, "loss": 0.4606, "step": 18180 }, { "epoch": 1.23, "learning_rate": 7.682779684023982e-06, "loss": 0.2598, "step": 18185 }, { "epoch": 1.23, "learning_rate": 7.68183703480261e-06, "loss": 0.4629, "step": 18190 }, { "epoch": 1.23, "learning_rate": 7.680894385581237e-06, "loss": 0.5715, "step": 18195 }, { "epoch": 1.24, "learning_rate": 7.679951736359867e-06, "loss": 0.2641, "step": 18200 }, { "epoch": 1.24, "learning_rate": 7.679009087138494e-06, "loss": 0.5734, "step": 18205 }, { "epoch": 1.24, "learning_rate": 7.678066437917124e-06, "loss": 0.8867, "step": 18210 }, { "epoch": 1.24, "learning_rate": 7.677123788695751e-06, "loss": 0.5148, "step": 18215 }, { "epoch": 1.24, "learning_rate": 7.676181139474379e-06, "loss": 0.5501, "step": 18220 }, { "epoch": 1.24, "learning_rate": 7.675238490253008e-06, "loss": 0.7532, "step": 18225 }, { "epoch": 1.24, "learning_rate": 7.674295841031636e-06, "loss": 0.7647, "step": 18230 }, { "epoch": 1.24, "learning_rate": 7.673353191810265e-06, "loss": 0.3495, "step": 18235 }, { "epoch": 1.24, "learning_rate": 7.672410542588893e-06, "loss": 0.6217, "step": 18240 }, { "epoch": 1.24, "learning_rate": 7.67146789336752e-06, "loss": 0.3801, "step": 18245 }, { "epoch": 1.24, "learning_rate": 7.670525244146148e-06, "loss": 0.0591, "step": 18250 }, { "epoch": 1.24, "learning_rate": 7.669582594924777e-06, "loss": 0.6671, "step": 18255 }, { "epoch": 1.24, "learning_rate": 7.668639945703405e-06, "loss": 0.3679, "step": 18260 }, { "epoch": 1.24, "learning_rate": 7.667697296482034e-06, "loss": 0.2677, "step": 18265 }, { "epoch": 1.24, "learning_rate": 7.666754647260662e-06, "loss": 0.526, "step": 18270 }, { "epoch": 1.24, "learning_rate": 7.66581199803929e-06, "loss": 0.4326, "step": 18275 }, { "epoch": 1.24, "learning_rate": 7.664869348817919e-06, "loss": 0.1909, "step": 18280 }, { "epoch": 1.24, "learning_rate": 7.663926699596547e-06, "loss": 1.345, "step": 18285 }, { "epoch": 1.24, "learning_rate": 7.662984050375174e-06, "loss": 0.6532, "step": 18290 }, { "epoch": 1.24, "learning_rate": 7.662041401153804e-06, "loss": 0.6343, "step": 18295 }, { "epoch": 1.24, "learning_rate": 7.661098751932431e-06, "loss": 0.5549, "step": 18300 }, { "epoch": 1.24, "learning_rate": 7.660156102711059e-06, "loss": 0.8892, "step": 18305 }, { "epoch": 1.24, "learning_rate": 7.659213453489688e-06, "loss": 0.4278, "step": 18310 }, { "epoch": 1.24, "learning_rate": 7.658270804268316e-06, "loss": 0.2823, "step": 18315 }, { "epoch": 1.24, "learning_rate": 7.657328155046945e-06, "loss": 0.8842, "step": 18320 }, { "epoch": 1.24, "learning_rate": 7.656385505825573e-06, "loss": 0.763, "step": 18325 }, { "epoch": 1.24, "learning_rate": 7.655442856604202e-06, "loss": 0.7212, "step": 18330 }, { "epoch": 1.24, "learning_rate": 7.654500207382828e-06, "loss": 0.4701, "step": 18335 }, { "epoch": 1.24, "learning_rate": 7.653557558161457e-06, "loss": 0.6229, "step": 18340 }, { "epoch": 1.25, "learning_rate": 7.652614908940085e-06, "loss": 0.2413, "step": 18345 }, { "epoch": 1.25, "learning_rate": 7.651672259718714e-06, "loss": 0.2549, "step": 18350 }, { "epoch": 1.25, "learning_rate": 7.650729610497342e-06, "loss": 0.5234, "step": 18355 }, { "epoch": 1.25, "learning_rate": 7.64978696127597e-06, "loss": 0.6604, "step": 18360 }, { "epoch": 1.25, "learning_rate": 7.648844312054599e-06, "loss": 0.3579, "step": 18365 }, { "epoch": 1.25, "learning_rate": 7.647901662833227e-06, "loss": 0.5377, "step": 18370 }, { "epoch": 1.25, "learning_rate": 7.646959013611856e-06, "loss": 0.4507, "step": 18375 }, { "epoch": 1.25, "learning_rate": 7.646016364390484e-06, "loss": 0.2084, "step": 18380 }, { "epoch": 1.25, "learning_rate": 7.645073715169113e-06, "loss": 0.7307, "step": 18385 }, { "epoch": 1.25, "learning_rate": 7.644131065947739e-06, "loss": 0.153, "step": 18390 }, { "epoch": 1.25, "learning_rate": 7.643188416726368e-06, "loss": 0.4823, "step": 18395 }, { "epoch": 1.25, "learning_rate": 7.642245767504996e-06, "loss": 0.4504, "step": 18400 }, { "epoch": 1.25, "learning_rate": 7.641303118283625e-06, "loss": 0.9433, "step": 18405 }, { "epoch": 1.25, "learning_rate": 7.640360469062253e-06, "loss": 0.4936, "step": 18410 }, { "epoch": 1.25, "learning_rate": 7.639417819840882e-06, "loss": 0.509, "step": 18415 }, { "epoch": 1.25, "learning_rate": 7.63847517061951e-06, "loss": 0.2582, "step": 18420 }, { "epoch": 1.25, "learning_rate": 7.637532521398138e-06, "loss": 1.0841, "step": 18425 }, { "epoch": 1.25, "learning_rate": 7.636589872176767e-06, "loss": 0.8327, "step": 18430 }, { "epoch": 1.25, "learning_rate": 7.635647222955394e-06, "loss": 0.4551, "step": 18435 }, { "epoch": 1.25, "learning_rate": 7.634704573734024e-06, "loss": 0.3673, "step": 18440 }, { "epoch": 1.25, "learning_rate": 7.633761924512651e-06, "loss": 0.3978, "step": 18445 }, { "epoch": 1.25, "learning_rate": 7.632819275291279e-06, "loss": 0.754, "step": 18450 }, { "epoch": 1.25, "learning_rate": 7.631876626069907e-06, "loss": 0.35, "step": 18455 }, { "epoch": 1.25, "learning_rate": 7.630933976848536e-06, "loss": 0.6389, "step": 18460 }, { "epoch": 1.25, "learning_rate": 7.629991327627164e-06, "loss": 0.4749, "step": 18465 }, { "epoch": 1.25, "learning_rate": 7.629048678405793e-06, "loss": 0.9234, "step": 18470 }, { "epoch": 1.25, "learning_rate": 7.62810602918442e-06, "loss": 0.6268, "step": 18475 }, { "epoch": 1.25, "learning_rate": 7.627163379963048e-06, "loss": 0.5896, "step": 18480 }, { "epoch": 1.25, "learning_rate": 7.626220730741677e-06, "loss": 0.2743, "step": 18485 }, { "epoch": 1.25, "learning_rate": 7.625278081520305e-06, "loss": 0.2428, "step": 18490 }, { "epoch": 1.26, "learning_rate": 7.624335432298934e-06, "loss": 0.7858, "step": 18495 }, { "epoch": 1.26, "learning_rate": 7.623392783077562e-06, "loss": 0.9477, "step": 18500 }, { "epoch": 1.26, "learning_rate": 7.62245013385619e-06, "loss": 0.5775, "step": 18505 }, { "epoch": 1.26, "learning_rate": 7.621507484634818e-06, "loss": 0.8218, "step": 18510 }, { "epoch": 1.26, "learning_rate": 7.620564835413447e-06, "loss": 0.1551, "step": 18515 }, { "epoch": 1.26, "learning_rate": 7.6196221861920745e-06, "loss": 0.3088, "step": 18520 }, { "epoch": 1.26, "learning_rate": 7.618679536970703e-06, "loss": 0.8151, "step": 18525 }, { "epoch": 1.26, "learning_rate": 7.6177368877493315e-06, "loss": 0.534, "step": 18530 }, { "epoch": 1.26, "learning_rate": 7.616794238527959e-06, "loss": 0.3865, "step": 18535 }, { "epoch": 1.26, "learning_rate": 7.615851589306588e-06, "loss": 0.537, "step": 18540 }, { "epoch": 1.26, "learning_rate": 7.614908940085216e-06, "loss": 0.3245, "step": 18545 }, { "epoch": 1.26, "learning_rate": 7.613966290863845e-06, "loss": 0.7904, "step": 18550 }, { "epoch": 1.26, "learning_rate": 7.613023641642473e-06, "loss": 0.9347, "step": 18555 }, { "epoch": 1.26, "learning_rate": 7.612080992421102e-06, "loss": 1.3489, "step": 18560 }, { "epoch": 1.26, "learning_rate": 7.611138343199728e-06, "loss": 0.2884, "step": 18565 }, { "epoch": 1.26, "learning_rate": 7.610195693978357e-06, "loss": 0.1725, "step": 18570 }, { "epoch": 1.26, "learning_rate": 7.609253044756985e-06, "loss": 0.4801, "step": 18575 }, { "epoch": 1.26, "learning_rate": 7.608310395535614e-06, "loss": 0.6122, "step": 18580 }, { "epoch": 1.26, "learning_rate": 7.607367746314242e-06, "loss": 0.7214, "step": 18585 }, { "epoch": 1.26, "learning_rate": 7.60642509709287e-06, "loss": 0.4656, "step": 18590 }, { "epoch": 1.26, "learning_rate": 7.6054824478714985e-06, "loss": 0.3959, "step": 18595 }, { "epoch": 1.26, "learning_rate": 7.604539798650127e-06, "loss": 0.642, "step": 18600 }, { "epoch": 1.26, "learning_rate": 7.6035971494287554e-06, "loss": 0.1762, "step": 18605 }, { "epoch": 1.26, "learning_rate": 7.602654500207384e-06, "loss": 0.5221, "step": 18610 }, { "epoch": 1.26, "learning_rate": 7.601711850986012e-06, "loss": 0.2609, "step": 18615 }, { "epoch": 1.26, "learning_rate": 7.600769201764639e-06, "loss": 0.5922, "step": 18620 }, { "epoch": 1.26, "learning_rate": 7.599826552543268e-06, "loss": 0.4877, "step": 18625 }, { "epoch": 1.26, "learning_rate": 7.598883903321896e-06, "loss": 0.9111, "step": 18630 }, { "epoch": 1.26, "learning_rate": 7.597941254100525e-06, "loss": 0.6583, "step": 18635 }, { "epoch": 1.27, "learning_rate": 7.596998604879153e-06, "loss": 0.2035, "step": 18640 }, { "epoch": 1.27, "learning_rate": 7.596055955657782e-06, "loss": 0.6724, "step": 18645 }, { "epoch": 1.27, "learning_rate": 7.595113306436409e-06, "loss": 0.49, "step": 18650 }, { "epoch": 1.27, "learning_rate": 7.594170657215038e-06, "loss": 0.264, "step": 18655 }, { "epoch": 1.27, "learning_rate": 7.593228007993666e-06, "loss": 0.0884, "step": 18660 }, { "epoch": 1.27, "learning_rate": 7.592285358772295e-06, "loss": 0.4405, "step": 18665 }, { "epoch": 1.27, "learning_rate": 7.591342709550923e-06, "loss": 0.3409, "step": 18670 }, { "epoch": 1.27, "learning_rate": 7.59040006032955e-06, "loss": 0.3189, "step": 18675 }, { "epoch": 1.27, "learning_rate": 7.5894574111081785e-06, "loss": 0.4781, "step": 18680 }, { "epoch": 1.27, "learning_rate": 7.588514761886807e-06, "loss": 0.1663, "step": 18685 }, { "epoch": 1.27, "learning_rate": 7.5875721126654355e-06, "loss": 0.5862, "step": 18690 }, { "epoch": 1.27, "learning_rate": 7.586629463444064e-06, "loss": 0.7403, "step": 18695 }, { "epoch": 1.27, "learning_rate": 7.5856868142226925e-06, "loss": 0.5439, "step": 18700 }, { "epoch": 1.27, "learning_rate": 7.58474416500132e-06, "loss": 0.5688, "step": 18705 }, { "epoch": 1.27, "learning_rate": 7.583801515779949e-06, "loss": 0.4591, "step": 18710 }, { "epoch": 1.27, "learning_rate": 7.582858866558577e-06, "loss": 1.0883, "step": 18715 }, { "epoch": 1.27, "learning_rate": 7.5819162173372056e-06, "loss": 0.6005, "step": 18720 }, { "epoch": 1.27, "learning_rate": 7.580973568115833e-06, "loss": 0.3014, "step": 18725 }, { "epoch": 1.27, "learning_rate": 7.580030918894462e-06, "loss": 0.6414, "step": 18730 }, { "epoch": 1.27, "learning_rate": 7.579088269673089e-06, "loss": 0.3868, "step": 18735 }, { "epoch": 1.27, "learning_rate": 7.578145620451718e-06, "loss": 0.3834, "step": 18740 }, { "epoch": 1.27, "learning_rate": 7.577202971230346e-06, "loss": 0.544, "step": 18745 }, { "epoch": 1.27, "learning_rate": 7.576260322008975e-06, "loss": 0.2406, "step": 18750 }, { "epoch": 1.27, "learning_rate": 7.575317672787603e-06, "loss": 0.4176, "step": 18755 }, { "epoch": 1.27, "learning_rate": 7.574375023566232e-06, "loss": 1.393, "step": 18760 }, { "epoch": 1.27, "learning_rate": 7.573432374344859e-06, "loss": 0.1577, "step": 18765 }, { "epoch": 1.27, "learning_rate": 7.572489725123487e-06, "loss": 0.2364, "step": 18770 }, { "epoch": 1.27, "learning_rate": 7.5715470759021155e-06, "loss": 0.4829, "step": 18775 }, { "epoch": 1.27, "learning_rate": 7.570604426680744e-06, "loss": 0.4602, "step": 18780 }, { "epoch": 1.27, "learning_rate": 7.5696617774593725e-06, "loss": 0.1927, "step": 18785 }, { "epoch": 1.28, "learning_rate": 7.568719128238e-06, "loss": 0.2949, "step": 18790 }, { "epoch": 1.28, "learning_rate": 7.567776479016629e-06, "loss": 0.58, "step": 18795 }, { "epoch": 1.28, "learning_rate": 7.566833829795257e-06, "loss": 0.334, "step": 18800 }, { "epoch": 1.28, "learning_rate": 7.565891180573886e-06, "loss": 0.5017, "step": 18805 }, { "epoch": 1.28, "learning_rate": 7.564948531352514e-06, "loss": 0.8919, "step": 18810 }, { "epoch": 1.28, "learning_rate": 7.564005882131143e-06, "loss": 0.2725, "step": 18815 }, { "epoch": 1.28, "learning_rate": 7.563063232909769e-06, "loss": 0.5517, "step": 18820 }, { "epoch": 1.28, "learning_rate": 7.562120583688398e-06, "loss": 0.3355, "step": 18825 }, { "epoch": 1.28, "learning_rate": 7.561177934467026e-06, "loss": 0.7296, "step": 18830 }, { "epoch": 1.28, "learning_rate": 7.560235285245655e-06, "loss": 0.4335, "step": 18835 }, { "epoch": 1.28, "learning_rate": 7.559292636024283e-06, "loss": 0.2989, "step": 18840 }, { "epoch": 1.28, "learning_rate": 7.558349986802912e-06, "loss": 0.4849, "step": 18845 }, { "epoch": 1.28, "learning_rate": 7.5574073375815395e-06, "loss": 0.4084, "step": 18850 }, { "epoch": 1.28, "learning_rate": 7.556464688360168e-06, "loss": 0.3558, "step": 18855 }, { "epoch": 1.28, "learning_rate": 7.5555220391387964e-06, "loss": 0.676, "step": 18860 }, { "epoch": 1.28, "learning_rate": 7.554579389917425e-06, "loss": 0.4014, "step": 18865 }, { "epoch": 1.28, "learning_rate": 7.553636740696053e-06, "loss": 0.9397, "step": 18870 }, { "epoch": 1.28, "learning_rate": 7.552694091474682e-06, "loss": 0.5003, "step": 18875 }, { "epoch": 1.28, "learning_rate": 7.551751442253309e-06, "loss": 0.2395, "step": 18880 }, { "epoch": 1.28, "learning_rate": 7.550808793031937e-06, "loss": 0.3631, "step": 18885 }, { "epoch": 1.28, "learning_rate": 7.549866143810566e-06, "loss": 1.0124, "step": 18890 }, { "epoch": 1.28, "learning_rate": 7.548923494589194e-06, "loss": 0.4161, "step": 18895 }, { "epoch": 1.28, "learning_rate": 7.547980845367823e-06, "loss": 0.6308, "step": 18900 }, { "epoch": 1.28, "learning_rate": 7.54703819614645e-06, "loss": 0.3659, "step": 18905 }, { "epoch": 1.28, "learning_rate": 7.546095546925079e-06, "loss": 0.3033, "step": 18910 }, { "epoch": 1.28, "learning_rate": 7.545152897703707e-06, "loss": 0.4162, "step": 18915 }, { "epoch": 1.28, "learning_rate": 7.544210248482336e-06, "loss": 0.3977, "step": 18920 }, { "epoch": 1.28, "learning_rate": 7.543267599260964e-06, "loss": 0.3519, "step": 18925 }, { "epoch": 1.28, "learning_rate": 7.542324950039593e-06, "loss": 0.4326, "step": 18930 }, { "epoch": 1.29, "learning_rate": 7.5413823008182195e-06, "loss": 0.2248, "step": 18935 }, { "epoch": 1.29, "learning_rate": 7.540439651596848e-06, "loss": 0.7115, "step": 18940 }, { "epoch": 1.29, "learning_rate": 7.5394970023754765e-06, "loss": 0.2581, "step": 18945 }, { "epoch": 1.29, "learning_rate": 7.538554353154105e-06, "loss": 0.4236, "step": 18950 }, { "epoch": 1.29, "learning_rate": 7.5376117039327335e-06, "loss": 0.7916, "step": 18955 }, { "epoch": 1.29, "learning_rate": 7.536669054711362e-06, "loss": 0.4832, "step": 18960 }, { "epoch": 1.29, "learning_rate": 7.53572640548999e-06, "loss": 0.5083, "step": 18965 }, { "epoch": 1.29, "learning_rate": 7.534783756268618e-06, "loss": 0.7667, "step": 18970 }, { "epoch": 1.29, "learning_rate": 7.5338411070472466e-06, "loss": 0.1782, "step": 18975 }, { "epoch": 1.29, "learning_rate": 7.532898457825874e-06, "loss": 0.4758, "step": 18980 }, { "epoch": 1.29, "learning_rate": 7.531955808604503e-06, "loss": 0.1213, "step": 18985 }, { "epoch": 1.29, "learning_rate": 7.531013159383131e-06, "loss": 1.0011, "step": 18990 }, { "epoch": 1.29, "learning_rate": 7.530070510161759e-06, "loss": 0.5557, "step": 18995 }, { "epoch": 1.29, "learning_rate": 7.529127860940387e-06, "loss": 0.361, "step": 19000 }, { "epoch": 1.29, "learning_rate": 7.528185211719016e-06, "loss": 0.8981, "step": 19005 }, { "epoch": 1.29, "learning_rate": 7.527242562497644e-06, "loss": 0.2512, "step": 19010 }, { "epoch": 1.29, "learning_rate": 7.526299913276273e-06, "loss": 0.4212, "step": 19015 }, { "epoch": 1.29, "learning_rate": 7.5253572640548996e-06, "loss": 0.3058, "step": 19020 }, { "epoch": 1.29, "learning_rate": 7.524414614833528e-06, "loss": 0.2497, "step": 19025 }, { "epoch": 1.29, "learning_rate": 7.5234719656121565e-06, "loss": 0.1558, "step": 19030 }, { "epoch": 1.29, "learning_rate": 7.522529316390785e-06, "loss": 0.5969, "step": 19035 }, { "epoch": 1.29, "learning_rate": 7.5215866671694135e-06, "loss": 0.8039, "step": 19040 }, { "epoch": 1.29, "learning_rate": 7.520644017948042e-06, "loss": 0.3657, "step": 19045 }, { "epoch": 1.29, "learning_rate": 7.51970136872667e-06, "loss": 1.1065, "step": 19050 }, { "epoch": 1.29, "learning_rate": 7.518758719505298e-06, "loss": 0.2912, "step": 19055 }, { "epoch": 1.29, "learning_rate": 7.517816070283927e-06, "loss": 0.6734, "step": 19060 }, { "epoch": 1.29, "learning_rate": 7.516873421062555e-06, "loss": 0.2017, "step": 19065 }, { "epoch": 1.29, "learning_rate": 7.515930771841184e-06, "loss": 0.8504, "step": 19070 }, { "epoch": 1.29, "learning_rate": 7.514988122619812e-06, "loss": 0.4194, "step": 19075 }, { "epoch": 1.29, "learning_rate": 7.514045473398439e-06, "loss": 0.5145, "step": 19080 }, { "epoch": 1.3, "learning_rate": 7.513102824177067e-06, "loss": 0.498, "step": 19085 }, { "epoch": 1.3, "learning_rate": 7.512160174955696e-06, "loss": 0.8758, "step": 19090 }, { "epoch": 1.3, "learning_rate": 7.511217525734324e-06, "loss": 0.2636, "step": 19095 }, { "epoch": 1.3, "learning_rate": 7.510274876512953e-06, "loss": 0.3057, "step": 19100 }, { "epoch": 1.3, "learning_rate": 7.509332227291581e-06, "loss": 0.1932, "step": 19105 }, { "epoch": 1.3, "learning_rate": 7.508389578070209e-06, "loss": 0.5597, "step": 19110 }, { "epoch": 1.3, "learning_rate": 7.5074469288488374e-06, "loss": 0.5634, "step": 19115 }, { "epoch": 1.3, "learning_rate": 7.506504279627466e-06, "loss": 0.0994, "step": 19120 }, { "epoch": 1.3, "learning_rate": 7.505561630406094e-06, "loss": 0.5109, "step": 19125 }, { "epoch": 1.3, "learning_rate": 7.504618981184723e-06, "loss": 0.7352, "step": 19130 }, { "epoch": 1.3, "learning_rate": 7.50367633196335e-06, "loss": 0.2563, "step": 19135 }, { "epoch": 1.3, "learning_rate": 7.502733682741978e-06, "loss": 0.2703, "step": 19140 }, { "epoch": 1.3, "learning_rate": 7.501791033520607e-06, "loss": 0.1868, "step": 19145 }, { "epoch": 1.3, "learning_rate": 7.500848384299235e-06, "loss": 0.5561, "step": 19150 }, { "epoch": 1.3, "learning_rate": 7.499905735077864e-06, "loss": 0.5397, "step": 19155 }, { "epoch": 1.3, "learning_rate": 7.498963085856492e-06, "loss": 0.8371, "step": 19160 }, { "epoch": 1.3, "learning_rate": 7.49802043663512e-06, "loss": 1.1385, "step": 19165 }, { "epoch": 1.3, "learning_rate": 7.497077787413748e-06, "loss": 0.7894, "step": 19170 }, { "epoch": 1.3, "learning_rate": 7.496135138192377e-06, "loss": 0.3311, "step": 19175 }, { "epoch": 1.3, "learning_rate": 7.495192488971005e-06, "loss": 0.419, "step": 19180 }, { "epoch": 1.3, "learning_rate": 7.494249839749633e-06, "loss": 1.035, "step": 19185 }, { "epoch": 1.3, "learning_rate": 7.493307190528261e-06, "loss": 0.4944, "step": 19190 }, { "epoch": 1.3, "learning_rate": 7.492364541306889e-06, "loss": 0.2078, "step": 19195 }, { "epoch": 1.3, "learning_rate": 7.4914218920855175e-06, "loss": 0.4693, "step": 19200 }, { "epoch": 1.3, "learning_rate": 7.490479242864146e-06, "loss": 0.5285, "step": 19205 }, { "epoch": 1.3, "learning_rate": 7.4895365936427745e-06, "loss": 0.112, "step": 19210 }, { "epoch": 1.3, "learning_rate": 7.488593944421403e-06, "loss": 0.2567, "step": 19215 }, { "epoch": 1.3, "learning_rate": 7.4876512952000314e-06, "loss": 0.34, "step": 19220 }, { "epoch": 1.3, "learning_rate": 7.486708645978659e-06, "loss": 0.9371, "step": 19225 }, { "epoch": 1.31, "learning_rate": 7.485765996757287e-06, "loss": 0.1278, "step": 19230 }, { "epoch": 1.31, "learning_rate": 7.484823347535915e-06, "loss": 0.7256, "step": 19235 }, { "epoch": 1.31, "learning_rate": 7.483880698314544e-06, "loss": 0.6257, "step": 19240 }, { "epoch": 1.31, "learning_rate": 7.482938049093172e-06, "loss": 0.7321, "step": 19245 }, { "epoch": 1.31, "learning_rate": 7.4819953998718e-06, "loss": 0.4116, "step": 19250 }, { "epoch": 1.31, "learning_rate": 7.481052750650428e-06, "loss": 0.6935, "step": 19255 }, { "epoch": 1.31, "learning_rate": 7.480110101429057e-06, "loss": 0.624, "step": 19260 }, { "epoch": 1.31, "learning_rate": 7.479167452207685e-06, "loss": 0.4442, "step": 19265 }, { "epoch": 1.31, "learning_rate": 7.478224802986314e-06, "loss": 0.3053, "step": 19270 }, { "epoch": 1.31, "learning_rate": 7.477282153764942e-06, "loss": 0.5096, "step": 19275 }, { "epoch": 1.31, "learning_rate": 7.476339504543569e-06, "loss": 0.6125, "step": 19280 }, { "epoch": 1.31, "learning_rate": 7.4753968553221975e-06, "loss": 0.7366, "step": 19285 }, { "epoch": 1.31, "learning_rate": 7.474454206100826e-06, "loss": 0.775, "step": 19290 }, { "epoch": 1.31, "learning_rate": 7.4735115568794545e-06, "loss": 0.6759, "step": 19295 }, { "epoch": 1.31, "learning_rate": 7.472568907658083e-06, "loss": 0.8298, "step": 19300 }, { "epoch": 1.31, "learning_rate": 7.4716262584367115e-06, "loss": 0.2696, "step": 19305 }, { "epoch": 1.31, "learning_rate": 7.470683609215339e-06, "loss": 0.2718, "step": 19310 }, { "epoch": 1.31, "learning_rate": 7.469740959993968e-06, "loss": 1.0794, "step": 19315 }, { "epoch": 1.31, "learning_rate": 7.468798310772596e-06, "loss": 0.8607, "step": 19320 }, { "epoch": 1.31, "learning_rate": 7.467855661551225e-06, "loss": 0.4681, "step": 19325 }, { "epoch": 1.31, "learning_rate": 7.466913012329853e-06, "loss": 0.4874, "step": 19330 }, { "epoch": 1.31, "learning_rate": 7.4659703631084816e-06, "loss": 0.2457, "step": 19335 }, { "epoch": 1.31, "learning_rate": 7.465027713887108e-06, "loss": 0.5501, "step": 19340 }, { "epoch": 1.31, "learning_rate": 7.464085064665737e-06, "loss": 0.4595, "step": 19345 }, { "epoch": 1.31, "learning_rate": 7.463142415444365e-06, "loss": 0.3393, "step": 19350 }, { "epoch": 1.31, "learning_rate": 7.462199766222994e-06, "loss": 0.2882, "step": 19355 }, { "epoch": 1.31, "learning_rate": 7.461257117001622e-06, "loss": 0.1226, "step": 19360 }, { "epoch": 1.31, "learning_rate": 7.46031446778025e-06, "loss": 0.5412, "step": 19365 }, { "epoch": 1.31, "learning_rate": 7.459371818558878e-06, "loss": 0.555, "step": 19370 }, { "epoch": 1.31, "learning_rate": 7.458429169337507e-06, "loss": 0.3073, "step": 19375 }, { "epoch": 1.32, "learning_rate": 7.457486520116135e-06, "loss": 0.6705, "step": 19380 }, { "epoch": 1.32, "learning_rate": 7.456543870894764e-06, "loss": 0.2989, "step": 19385 }, { "epoch": 1.32, "learning_rate": 7.455601221673392e-06, "loss": 0.5807, "step": 19390 }, { "epoch": 1.32, "learning_rate": 7.454658572452019e-06, "loss": 0.6034, "step": 19395 }, { "epoch": 1.32, "learning_rate": 7.453715923230648e-06, "loss": 0.6888, "step": 19400 }, { "epoch": 1.32, "learning_rate": 7.452773274009276e-06, "loss": 1.1144, "step": 19405 }, { "epoch": 1.32, "learning_rate": 7.451830624787905e-06, "loss": 0.3946, "step": 19410 }, { "epoch": 1.32, "learning_rate": 7.450887975566533e-06, "loss": 0.4153, "step": 19415 }, { "epoch": 1.32, "learning_rate": 7.449945326345162e-06, "loss": 0.3807, "step": 19420 }, { "epoch": 1.32, "learning_rate": 7.449002677123789e-06, "loss": 0.3779, "step": 19425 }, { "epoch": 1.32, "learning_rate": 7.448060027902418e-06, "loss": 0.568, "step": 19430 }, { "epoch": 1.32, "learning_rate": 7.447117378681045e-06, "loss": 0.4533, "step": 19435 }, { "epoch": 1.32, "learning_rate": 7.446174729459674e-06, "loss": 0.3928, "step": 19440 }, { "epoch": 1.32, "learning_rate": 7.445232080238302e-06, "loss": 0.5361, "step": 19445 }, { "epoch": 1.32, "learning_rate": 7.444289431016931e-06, "loss": 0.433, "step": 19450 }, { "epoch": 1.32, "learning_rate": 7.4433467817955585e-06, "loss": 0.4467, "step": 19455 }, { "epoch": 1.32, "learning_rate": 7.442404132574187e-06, "loss": 0.3663, "step": 19460 }, { "epoch": 1.32, "learning_rate": 7.4414614833528154e-06, "loss": 0.4464, "step": 19465 }, { "epoch": 1.32, "learning_rate": 7.440518834131444e-06, "loss": 0.3487, "step": 19470 }, { "epoch": 1.32, "learning_rate": 7.439576184910072e-06, "loss": 0.1945, "step": 19475 }, { "epoch": 1.32, "learning_rate": 7.438633535688699e-06, "loss": 0.2119, "step": 19480 }, { "epoch": 1.32, "learning_rate": 7.437690886467328e-06, "loss": 0.6611, "step": 19485 }, { "epoch": 1.32, "learning_rate": 7.436748237245956e-06, "loss": 0.622, "step": 19490 }, { "epoch": 1.32, "learning_rate": 7.435805588024585e-06, "loss": 0.7158, "step": 19495 }, { "epoch": 1.32, "learning_rate": 7.434862938803213e-06, "loss": 0.4672, "step": 19500 }, { "epoch": 1.32, "learning_rate": 7.433920289581842e-06, "loss": 0.5488, "step": 19505 }, { "epoch": 1.32, "learning_rate": 7.432977640360469e-06, "loss": 0.6371, "step": 19510 }, { "epoch": 1.32, "learning_rate": 7.432034991139098e-06, "loss": 0.3567, "step": 19515 }, { "epoch": 1.32, "learning_rate": 7.431092341917726e-06, "loss": 0.4571, "step": 19520 }, { "epoch": 1.33, "learning_rate": 7.430149692696355e-06, "loss": 0.8455, "step": 19525 }, { "epoch": 1.33, "learning_rate": 7.429207043474983e-06, "loss": 0.6937, "step": 19530 }, { "epoch": 1.33, "learning_rate": 7.428264394253612e-06, "loss": 0.4884, "step": 19535 }, { "epoch": 1.33, "learning_rate": 7.4273217450322385e-06, "loss": 0.1942, "step": 19540 }, { "epoch": 1.33, "learning_rate": 7.426379095810867e-06, "loss": 0.4257, "step": 19545 }, { "epoch": 1.33, "learning_rate": 7.4254364465894955e-06, "loss": 0.7739, "step": 19550 }, { "epoch": 1.33, "learning_rate": 7.424493797368124e-06, "loss": 0.9548, "step": 19555 }, { "epoch": 1.33, "learning_rate": 7.4235511481467525e-06, "loss": 0.5398, "step": 19560 }, { "epoch": 1.33, "learning_rate": 7.422608498925381e-06, "loss": 0.5165, "step": 19565 }, { "epoch": 1.33, "learning_rate": 7.421665849704009e-06, "loss": 0.6196, "step": 19570 }, { "epoch": 1.33, "learning_rate": 7.420723200482637e-06, "loss": 0.4452, "step": 19575 }, { "epoch": 1.33, "learning_rate": 7.4197805512612656e-06, "loss": 1.0094, "step": 19580 }, { "epoch": 1.33, "learning_rate": 7.418837902039894e-06, "loss": 0.2722, "step": 19585 }, { "epoch": 1.33, "learning_rate": 7.4178952528185225e-06, "loss": 0.3773, "step": 19590 }, { "epoch": 1.33, "learning_rate": 7.416952603597149e-06, "loss": 0.5023, "step": 19595 }, { "epoch": 1.33, "learning_rate": 7.416009954375778e-06, "loss": 0.6216, "step": 19600 }, { "epoch": 1.33, "learning_rate": 7.415067305154406e-06, "loss": 0.4056, "step": 19605 }, { "epoch": 1.33, "learning_rate": 7.414124655933035e-06, "loss": 0.5794, "step": 19610 }, { "epoch": 1.33, "learning_rate": 7.413182006711663e-06, "loss": 0.3532, "step": 19615 }, { "epoch": 1.33, "learning_rate": 7.412239357490292e-06, "loss": 1.1962, "step": 19620 }, { "epoch": 1.33, "learning_rate": 7.411296708268919e-06, "loss": 0.1635, "step": 19625 }, { "epoch": 1.33, "learning_rate": 7.410354059047548e-06, "loss": 0.3934, "step": 19630 }, { "epoch": 1.33, "learning_rate": 7.409411409826176e-06, "loss": 0.7658, "step": 19635 }, { "epoch": 1.33, "learning_rate": 7.408468760604805e-06, "loss": 0.6721, "step": 19640 }, { "epoch": 1.33, "learning_rate": 7.4075261113834325e-06, "loss": 0.5665, "step": 19645 }, { "epoch": 1.33, "learning_rate": 7.406583462162061e-06, "loss": 0.4114, "step": 19650 }, { "epoch": 1.33, "learning_rate": 7.405640812940689e-06, "loss": 0.4563, "step": 19655 }, { "epoch": 1.33, "learning_rate": 7.404698163719317e-06, "loss": 0.0552, "step": 19660 }, { "epoch": 1.33, "learning_rate": 7.403755514497946e-06, "loss": 0.4983, "step": 19665 }, { "epoch": 1.34, "learning_rate": 7.402812865276574e-06, "loss": 0.2808, "step": 19670 }, { "epoch": 1.34, "learning_rate": 7.401870216055203e-06, "loss": 0.3929, "step": 19675 }, { "epoch": 1.34, "learning_rate": 7.40092756683383e-06, "loss": 0.2782, "step": 19680 }, { "epoch": 1.34, "learning_rate": 7.399984917612458e-06, "loss": 0.7043, "step": 19685 }, { "epoch": 1.34, "learning_rate": 7.399042268391086e-06, "loss": 0.2879, "step": 19690 }, { "epoch": 1.34, "learning_rate": 7.398099619169715e-06, "loss": 0.5807, "step": 19695 }, { "epoch": 1.34, "learning_rate": 7.397156969948343e-06, "loss": 0.6165, "step": 19700 }, { "epoch": 1.34, "learning_rate": 7.396214320726972e-06, "loss": 0.5546, "step": 19705 }, { "epoch": 1.34, "learning_rate": 7.3952716715055995e-06, "loss": 0.3814, "step": 19710 }, { "epoch": 1.34, "learning_rate": 7.394329022284228e-06, "loss": 0.7119, "step": 19715 }, { "epoch": 1.34, "learning_rate": 7.3933863730628564e-06, "loss": 0.3437, "step": 19720 }, { "epoch": 1.34, "learning_rate": 7.392443723841485e-06, "loss": 0.8145, "step": 19725 }, { "epoch": 1.34, "learning_rate": 7.391501074620113e-06, "loss": 0.8966, "step": 19730 }, { "epoch": 1.34, "learning_rate": 7.390558425398742e-06, "loss": 0.3718, "step": 19735 }, { "epoch": 1.34, "learning_rate": 7.389615776177369e-06, "loss": 0.2998, "step": 19740 }, { "epoch": 1.34, "learning_rate": 7.388673126955997e-06, "loss": 0.6154, "step": 19745 }, { "epoch": 1.34, "learning_rate": 7.387730477734626e-06, "loss": 0.824, "step": 19750 }, { "epoch": 1.34, "learning_rate": 7.386787828513254e-06, "loss": 0.4992, "step": 19755 }, { "epoch": 1.34, "learning_rate": 7.385845179291883e-06, "loss": 0.2781, "step": 19760 }, { "epoch": 1.34, "learning_rate": 7.384902530070511e-06, "loss": 0.4366, "step": 19765 }, { "epoch": 1.34, "learning_rate": 7.383959880849139e-06, "loss": 0.0616, "step": 19770 }, { "epoch": 1.34, "learning_rate": 7.383017231627767e-06, "loss": 0.5773, "step": 19775 }, { "epoch": 1.34, "learning_rate": 7.382074582406396e-06, "loss": 0.5773, "step": 19780 }, { "epoch": 1.34, "learning_rate": 7.381131933185024e-06, "loss": 0.2459, "step": 19785 }, { "epoch": 1.34, "learning_rate": 7.380189283963653e-06, "loss": 0.2256, "step": 19790 }, { "epoch": 1.34, "learning_rate": 7.3792466347422795e-06, "loss": 0.6363, "step": 19795 }, { "epoch": 1.34, "learning_rate": 7.378303985520908e-06, "loss": 0.469, "step": 19800 }, { "epoch": 1.34, "learning_rate": 7.3773613362995365e-06, "loss": 0.6783, "step": 19805 }, { "epoch": 1.34, "learning_rate": 7.376418687078165e-06, "loss": 0.7607, "step": 19810 }, { "epoch": 1.34, "learning_rate": 7.3754760378567935e-06, "loss": 0.3396, "step": 19815 }, { "epoch": 1.35, "learning_rate": 7.374533388635422e-06, "loss": 0.2102, "step": 19820 }, { "epoch": 1.35, "learning_rate": 7.37359073941405e-06, "loss": 0.2712, "step": 19825 }, { "epoch": 1.35, "learning_rate": 7.372648090192678e-06, "loss": 0.665, "step": 19830 }, { "epoch": 1.35, "learning_rate": 7.3717054409713066e-06, "loss": 0.4861, "step": 19835 }, { "epoch": 1.35, "learning_rate": 7.370762791749935e-06, "loss": 0.916, "step": 19840 }, { "epoch": 1.35, "learning_rate": 7.3698201425285635e-06, "loss": 0.6419, "step": 19845 }, { "epoch": 1.35, "learning_rate": 7.368877493307191e-06, "loss": 0.2463, "step": 19850 }, { "epoch": 1.35, "learning_rate": 7.367934844085819e-06, "loss": 0.5743, "step": 19855 }, { "epoch": 1.35, "learning_rate": 7.366992194864447e-06, "loss": 0.4414, "step": 19860 }, { "epoch": 1.35, "learning_rate": 7.366049545643076e-06, "loss": 0.3973, "step": 19865 }, { "epoch": 1.35, "learning_rate": 7.365106896421704e-06, "loss": 0.3305, "step": 19870 }, { "epoch": 1.35, "learning_rate": 7.364164247200333e-06, "loss": 0.7068, "step": 19875 }, { "epoch": 1.35, "learning_rate": 7.363221597978961e-06, "loss": 0.6963, "step": 19880 }, { "epoch": 1.35, "learning_rate": 7.362278948757589e-06, "loss": 0.7565, "step": 19885 }, { "epoch": 1.35, "learning_rate": 7.361336299536217e-06, "loss": 0.0817, "step": 19890 }, { "epoch": 1.35, "learning_rate": 7.360393650314845e-06, "loss": 0.8806, "step": 19895 }, { "epoch": 1.35, "learning_rate": 7.3594510010934735e-06, "loss": 0.3355, "step": 19900 }, { "epoch": 1.35, "learning_rate": 7.358508351872102e-06, "loss": 0.3773, "step": 19905 }, { "epoch": 1.35, "learning_rate": 7.35756570265073e-06, "loss": 0.2061, "step": 19910 }, { "epoch": 1.35, "learning_rate": 7.356623053429358e-06, "loss": 0.5983, "step": 19915 }, { "epoch": 1.35, "learning_rate": 7.355680404207987e-06, "loss": 0.4248, "step": 19920 }, { "epoch": 1.35, "learning_rate": 7.354737754986615e-06, "loss": 1.2827, "step": 19925 }, { "epoch": 1.35, "learning_rate": 7.353795105765244e-06, "loss": 0.515, "step": 19930 }, { "epoch": 1.35, "learning_rate": 7.352852456543872e-06, "loss": 0.6709, "step": 19935 }, { "epoch": 1.35, "learning_rate": 7.351909807322499e-06, "loss": 0.6831, "step": 19940 }, { "epoch": 1.35, "learning_rate": 7.350967158101127e-06, "loss": 0.4354, "step": 19945 }, { "epoch": 1.35, "learning_rate": 7.350024508879756e-06, "loss": 0.3827, "step": 19950 }, { "epoch": 1.35, "learning_rate": 7.349081859658384e-06, "loss": 0.1892, "step": 19955 }, { "epoch": 1.35, "learning_rate": 7.348139210437013e-06, "loss": 0.3665, "step": 19960 }, { "epoch": 1.36, "learning_rate": 7.347196561215641e-06, "loss": 0.5508, "step": 19965 }, { "epoch": 1.36, "learning_rate": 7.346253911994269e-06, "loss": 0.3198, "step": 19970 }, { "epoch": 1.36, "learning_rate": 7.3453112627728974e-06, "loss": 1.1657, "step": 19975 }, { "epoch": 1.36, "learning_rate": 7.344368613551526e-06, "loss": 0.61, "step": 19980 }, { "epoch": 1.36, "learning_rate": 7.343425964330154e-06, "loss": 0.7886, "step": 19985 }, { "epoch": 1.36, "learning_rate": 7.342483315108783e-06, "loss": 0.5062, "step": 19990 }, { "epoch": 1.36, "learning_rate": 7.341540665887411e-06, "loss": 0.4853, "step": 19995 }, { "epoch": 1.36, "learning_rate": 7.340598016666038e-06, "loss": 0.5526, "step": 20000 }, { "epoch": 1.36, "learning_rate": 7.339655367444667e-06, "loss": 0.204, "step": 20005 }, { "epoch": 1.36, "learning_rate": 7.338712718223295e-06, "loss": 0.6995, "step": 20010 }, { "epoch": 1.36, "learning_rate": 7.337770069001924e-06, "loss": 0.12, "step": 20015 }, { "epoch": 1.36, "learning_rate": 7.336827419780552e-06, "loss": 0.2326, "step": 20020 }, { "epoch": 1.36, "learning_rate": 7.33588477055918e-06, "loss": 0.1784, "step": 20025 }, { "epoch": 1.36, "learning_rate": 7.334942121337808e-06, "loss": 0.4201, "step": 20030 }, { "epoch": 1.36, "learning_rate": 7.333999472116437e-06, "loss": 0.2071, "step": 20035 }, { "epoch": 1.36, "learning_rate": 7.333056822895065e-06, "loss": 0.4678, "step": 20040 }, { "epoch": 1.36, "learning_rate": 7.332114173673694e-06, "loss": 0.9292, "step": 20045 }, { "epoch": 1.36, "learning_rate": 7.331171524452322e-06, "loss": 0.3093, "step": 20050 }, { "epoch": 1.36, "learning_rate": 7.330228875230949e-06, "loss": 0.6351, "step": 20055 }, { "epoch": 1.36, "learning_rate": 7.3292862260095775e-06, "loss": 0.7443, "step": 20060 }, { "epoch": 1.36, "learning_rate": 7.328343576788206e-06, "loss": 0.9599, "step": 20065 }, { "epoch": 1.36, "learning_rate": 7.3274009275668345e-06, "loss": 0.2697, "step": 20070 }, { "epoch": 1.36, "learning_rate": 7.326458278345463e-06, "loss": 0.2679, "step": 20075 }, { "epoch": 1.36, "learning_rate": 7.3255156291240914e-06, "loss": 0.5958, "step": 20080 }, { "epoch": 1.36, "learning_rate": 7.324572979902719e-06, "loss": 0.3744, "step": 20085 }, { "epoch": 1.36, "learning_rate": 7.3236303306813476e-06, "loss": 0.7547, "step": 20090 }, { "epoch": 1.36, "learning_rate": 7.322687681459976e-06, "loss": 0.8909, "step": 20095 }, { "epoch": 1.36, "learning_rate": 7.3217450322386045e-06, "loss": 0.2582, "step": 20100 }, { "epoch": 1.36, "learning_rate": 7.320802383017232e-06, "loss": 0.6175, "step": 20105 }, { "epoch": 1.36, "learning_rate": 7.319859733795861e-06, "loss": 0.4105, "step": 20110 }, { "epoch": 1.37, "learning_rate": 7.318917084574488e-06, "loss": 0.3629, "step": 20115 }, { "epoch": 1.37, "learning_rate": 7.317974435353117e-06, "loss": 0.12, "step": 20120 }, { "epoch": 1.37, "learning_rate": 7.317031786131745e-06, "loss": 1.1451, "step": 20125 }, { "epoch": 1.37, "learning_rate": 7.316089136910374e-06, "loss": 0.3099, "step": 20130 }, { "epoch": 1.37, "learning_rate": 7.315146487689002e-06, "loss": 0.4207, "step": 20135 }, { "epoch": 1.37, "learning_rate": 7.31420383846763e-06, "loss": 0.5298, "step": 20140 }, { "epoch": 1.37, "learning_rate": 7.3132611892462575e-06, "loss": 0.2511, "step": 20145 }, { "epoch": 1.37, "learning_rate": 7.312318540024886e-06, "loss": 0.304, "step": 20150 }, { "epoch": 1.37, "learning_rate": 7.3113758908035145e-06, "loss": 0.5411, "step": 20155 }, { "epoch": 1.37, "learning_rate": 7.310433241582143e-06, "loss": 0.2642, "step": 20160 }, { "epoch": 1.37, "learning_rate": 7.3094905923607715e-06, "loss": 0.7268, "step": 20165 }, { "epoch": 1.37, "learning_rate": 7.308547943139399e-06, "loss": 0.7332, "step": 20170 }, { "epoch": 1.37, "learning_rate": 7.307605293918028e-06, "loss": 0.92, "step": 20175 }, { "epoch": 1.37, "learning_rate": 7.306662644696656e-06, "loss": 0.2463, "step": 20180 }, { "epoch": 1.37, "learning_rate": 7.305719995475285e-06, "loss": 0.5579, "step": 20185 }, { "epoch": 1.37, "learning_rate": 7.304777346253913e-06, "loss": 0.5464, "step": 20190 }, { "epoch": 1.37, "learning_rate": 7.3038346970325416e-06, "loss": 0.1951, "step": 20195 }, { "epoch": 1.37, "learning_rate": 7.302892047811168e-06, "loss": 0.3612, "step": 20200 }, { "epoch": 1.37, "learning_rate": 7.301949398589797e-06, "loss": 0.2898, "step": 20205 }, { "epoch": 1.37, "learning_rate": 7.301006749368425e-06, "loss": 0.3621, "step": 20210 }, { "epoch": 1.37, "learning_rate": 7.300064100147054e-06, "loss": 0.867, "step": 20215 }, { "epoch": 1.37, "learning_rate": 7.299121450925682e-06, "loss": 0.626, "step": 20220 }, { "epoch": 1.37, "learning_rate": 7.298178801704311e-06, "loss": 0.0429, "step": 20225 }, { "epoch": 1.37, "learning_rate": 7.2972361524829384e-06, "loss": 0.7548, "step": 20230 }, { "epoch": 1.37, "learning_rate": 7.296293503261567e-06, "loss": 0.7219, "step": 20235 }, { "epoch": 1.37, "learning_rate": 7.295350854040195e-06, "loss": 0.296, "step": 20240 }, { "epoch": 1.37, "learning_rate": 7.294408204818824e-06, "loss": 0.2777, "step": 20245 }, { "epoch": 1.37, "learning_rate": 7.293465555597452e-06, "loss": 0.7902, "step": 20250 }, { "epoch": 1.37, "learning_rate": 7.292522906376079e-06, "loss": 0.3427, "step": 20255 }, { "epoch": 1.38, "learning_rate": 7.291580257154708e-06, "loss": 0.4943, "step": 20260 }, { "epoch": 1.38, "learning_rate": 7.290637607933336e-06, "loss": 0.2145, "step": 20265 }, { "epoch": 1.38, "learning_rate": 7.289694958711965e-06, "loss": 0.3128, "step": 20270 }, { "epoch": 1.38, "learning_rate": 7.288752309490593e-06, "loss": 0.4503, "step": 20275 }, { "epoch": 1.38, "learning_rate": 7.287809660269222e-06, "loss": 1.0004, "step": 20280 }, { "epoch": 1.38, "learning_rate": 7.286867011047849e-06, "loss": 0.3848, "step": 20285 }, { "epoch": 1.38, "learning_rate": 7.285924361826478e-06, "loss": 0.2762, "step": 20290 }, { "epoch": 1.38, "learning_rate": 7.284981712605106e-06, "loss": 0.1958, "step": 20295 }, { "epoch": 1.38, "learning_rate": 7.284039063383735e-06, "loss": 1.2534, "step": 20300 }, { "epoch": 1.38, "learning_rate": 7.283096414162363e-06, "loss": 0.2257, "step": 20305 }, { "epoch": 1.38, "learning_rate": 7.282153764940991e-06, "loss": 0.7732, "step": 20310 }, { "epoch": 1.38, "learning_rate": 7.2812111157196185e-06, "loss": 0.4653, "step": 20315 }, { "epoch": 1.38, "learning_rate": 7.280268466498247e-06, "loss": 0.9073, "step": 20320 }, { "epoch": 1.38, "learning_rate": 7.2793258172768755e-06, "loss": 0.392, "step": 20325 }, { "epoch": 1.38, "learning_rate": 7.278383168055504e-06, "loss": 0.237, "step": 20330 }, { "epoch": 1.38, "learning_rate": 7.2774405188341324e-06, "loss": 0.6805, "step": 20335 }, { "epoch": 1.38, "learning_rate": 7.276497869612761e-06, "loss": 0.4367, "step": 20340 }, { "epoch": 1.38, "learning_rate": 7.2755552203913886e-06, "loss": 0.2383, "step": 20345 }, { "epoch": 1.38, "learning_rate": 7.274612571170017e-06, "loss": 0.7937, "step": 20350 }, { "epoch": 1.38, "learning_rate": 7.273669921948645e-06, "loss": 0.4289, "step": 20355 }, { "epoch": 1.38, "learning_rate": 7.272727272727273e-06, "loss": 0.1912, "step": 20360 }, { "epoch": 1.38, "learning_rate": 7.271784623505902e-06, "loss": 0.3553, "step": 20365 }, { "epoch": 1.38, "learning_rate": 7.270841974284529e-06, "loss": 0.5724, "step": 20370 }, { "epoch": 1.38, "learning_rate": 7.269899325063158e-06, "loss": 0.1417, "step": 20375 }, { "epoch": 1.38, "learning_rate": 7.268956675841786e-06, "loss": 0.3328, "step": 20380 }, { "epoch": 1.38, "learning_rate": 7.268014026620415e-06, "loss": 0.4842, "step": 20385 }, { "epoch": 1.38, "learning_rate": 7.267071377399043e-06, "loss": 0.4732, "step": 20390 }, { "epoch": 1.38, "learning_rate": 7.266128728177672e-06, "loss": 0.7237, "step": 20395 }, { "epoch": 1.38, "learning_rate": 7.2651860789562985e-06, "loss": 0.5881, "step": 20400 }, { "epoch": 1.38, "learning_rate": 7.264243429734927e-06, "loss": 0.422, "step": 20405 }, { "epoch": 1.39, "learning_rate": 7.2633007805135555e-06, "loss": 0.686, "step": 20410 }, { "epoch": 1.39, "learning_rate": 7.262358131292184e-06, "loss": 0.3846, "step": 20415 }, { "epoch": 1.39, "learning_rate": 7.2614154820708125e-06, "loss": 0.7786, "step": 20420 }, { "epoch": 1.39, "learning_rate": 7.260472832849441e-06, "loss": 0.7803, "step": 20425 }, { "epoch": 1.39, "learning_rate": 7.259530183628069e-06, "loss": 0.6266, "step": 20430 }, { "epoch": 1.39, "learning_rate": 7.258587534406697e-06, "loss": 0.3519, "step": 20435 }, { "epoch": 1.39, "learning_rate": 7.257644885185326e-06, "loss": 0.316, "step": 20440 }, { "epoch": 1.39, "learning_rate": 7.256702235963954e-06, "loss": 0.1356, "step": 20445 }, { "epoch": 1.39, "learning_rate": 7.2557595867425826e-06, "loss": 0.4139, "step": 20450 }, { "epoch": 1.39, "learning_rate": 7.254816937521211e-06, "loss": 0.9877, "step": 20455 }, { "epoch": 1.39, "learning_rate": 7.253874288299838e-06, "loss": 1.0875, "step": 20460 }, { "epoch": 1.39, "learning_rate": 7.252931639078466e-06, "loss": 0.198, "step": 20465 }, { "epoch": 1.39, "learning_rate": 7.251988989857095e-06, "loss": 0.2636, "step": 20470 }, { "epoch": 1.39, "learning_rate": 7.251046340635723e-06, "loss": 0.3832, "step": 20475 }, { "epoch": 1.39, "learning_rate": 7.250103691414352e-06, "loss": 0.2341, "step": 20480 }, { "epoch": 1.39, "learning_rate": 7.249161042192979e-06, "loss": 0.6024, "step": 20485 }, { "epoch": 1.39, "learning_rate": 7.248218392971608e-06, "loss": 0.2097, "step": 20490 }, { "epoch": 1.39, "learning_rate": 7.247275743750236e-06, "loss": 0.5911, "step": 20495 }, { "epoch": 1.39, "learning_rate": 7.246333094528865e-06, "loss": 0.6718, "step": 20500 }, { "epoch": 1.39, "learning_rate": 7.245390445307493e-06, "loss": 0.9064, "step": 20505 }, { "epoch": 1.39, "learning_rate": 7.244447796086122e-06, "loss": 0.6014, "step": 20510 }, { "epoch": 1.39, "learning_rate": 7.243505146864749e-06, "loss": 0.8024, "step": 20515 }, { "epoch": 1.39, "learning_rate": 7.242562497643377e-06, "loss": 0.5295, "step": 20520 }, { "epoch": 1.39, "learning_rate": 7.241619848422006e-06, "loss": 0.3736, "step": 20525 }, { "epoch": 1.39, "learning_rate": 7.240677199200634e-06, "loss": 0.4175, "step": 20530 }, { "epoch": 1.39, "learning_rate": 7.239734549979263e-06, "loss": 0.6118, "step": 20535 }, { "epoch": 1.39, "learning_rate": 7.238791900757891e-06, "loss": 0.5203, "step": 20540 }, { "epoch": 1.39, "learning_rate": 7.237849251536519e-06, "loss": 0.3473, "step": 20545 }, { "epoch": 1.39, "learning_rate": 7.236906602315147e-06, "loss": 0.6199, "step": 20550 }, { "epoch": 1.4, "learning_rate": 7.235963953093776e-06, "loss": 0.5529, "step": 20555 }, { "epoch": 1.4, "learning_rate": 7.235021303872403e-06, "loss": 0.7706, "step": 20560 }, { "epoch": 1.4, "learning_rate": 7.234078654651032e-06, "loss": 0.7093, "step": 20565 }, { "epoch": 1.4, "learning_rate": 7.23313600542966e-06, "loss": 0.3576, "step": 20570 }, { "epoch": 1.4, "learning_rate": 7.232193356208288e-06, "loss": 0.5035, "step": 20575 }, { "epoch": 1.4, "learning_rate": 7.2312507069869164e-06, "loss": 0.6088, "step": 20580 }, { "epoch": 1.4, "learning_rate": 7.230308057765545e-06, "loss": 0.3932, "step": 20585 }, { "epoch": 1.4, "learning_rate": 7.229365408544173e-06, "loss": 0.7038, "step": 20590 }, { "epoch": 1.4, "learning_rate": 7.228422759322802e-06, "loss": 0.681, "step": 20595 }, { "epoch": 1.4, "learning_rate": 7.2274801101014295e-06, "loss": 0.2248, "step": 20600 }, { "epoch": 1.4, "learning_rate": 7.226537460880057e-06, "loss": 0.5016, "step": 20605 }, { "epoch": 1.4, "learning_rate": 7.225594811658686e-06, "loss": 0.0293, "step": 20610 }, { "epoch": 1.4, "learning_rate": 7.224652162437314e-06, "loss": 0.6585, "step": 20615 }, { "epoch": 1.4, "learning_rate": 7.223709513215943e-06, "loss": 0.6707, "step": 20620 }, { "epoch": 1.4, "learning_rate": 7.222766863994571e-06, "loss": 0.2589, "step": 20625 }, { "epoch": 1.4, "learning_rate": 7.221824214773199e-06, "loss": 0.4277, "step": 20630 }, { "epoch": 1.4, "learning_rate": 7.220881565551827e-06, "loss": 0.6704, "step": 20635 }, { "epoch": 1.4, "learning_rate": 7.219938916330456e-06, "loss": 0.58, "step": 20640 }, { "epoch": 1.4, "learning_rate": 7.218996267109084e-06, "loss": 0.1689, "step": 20645 }, { "epoch": 1.4, "learning_rate": 7.218053617887713e-06, "loss": 0.4518, "step": 20650 }, { "epoch": 1.4, "learning_rate": 7.217110968666341e-06, "loss": 0.2881, "step": 20655 }, { "epoch": 1.4, "learning_rate": 7.216168319444968e-06, "loss": 0.2078, "step": 20660 }, { "epoch": 1.4, "learning_rate": 7.2152256702235965e-06, "loss": 1.1272, "step": 20665 }, { "epoch": 1.4, "learning_rate": 7.214283021002225e-06, "loss": 0.3162, "step": 20670 }, { "epoch": 1.4, "learning_rate": 7.2133403717808535e-06, "loss": 1.1067, "step": 20675 }, { "epoch": 1.4, "learning_rate": 7.212397722559482e-06, "loss": 0.251, "step": 20680 }, { "epoch": 1.4, "learning_rate": 7.21145507333811e-06, "loss": 0.5883, "step": 20685 }, { "epoch": 1.4, "learning_rate": 7.210512424116738e-06, "loss": 0.6746, "step": 20690 }, { "epoch": 1.4, "learning_rate": 7.2095697748953666e-06, "loss": 0.5512, "step": 20695 }, { "epoch": 1.4, "learning_rate": 7.208627125673995e-06, "loss": 0.4761, "step": 20700 }, { "epoch": 1.41, "learning_rate": 7.2076844764526235e-06, "loss": 0.1252, "step": 20705 }, { "epoch": 1.41, "learning_rate": 7.206741827231252e-06, "loss": 0.4494, "step": 20710 }, { "epoch": 1.41, "learning_rate": 7.205799178009879e-06, "loss": 0.219, "step": 20715 }, { "epoch": 1.41, "learning_rate": 7.204856528788507e-06, "loss": 0.5262, "step": 20720 }, { "epoch": 1.41, "learning_rate": 7.203913879567136e-06, "loss": 1.0943, "step": 20725 }, { "epoch": 1.41, "learning_rate": 7.202971230345764e-06, "loss": 0.8096, "step": 20730 }, { "epoch": 1.41, "learning_rate": 7.202028581124393e-06, "loss": 0.3206, "step": 20735 }, { "epoch": 1.41, "learning_rate": 7.201085931903021e-06, "loss": 0.5876, "step": 20740 }, { "epoch": 1.41, "learning_rate": 7.200143282681649e-06, "loss": 0.3535, "step": 20745 }, { "epoch": 1.41, "learning_rate": 7.199200633460277e-06, "loss": 0.4324, "step": 20750 }, { "epoch": 1.41, "learning_rate": 7.198257984238906e-06, "loss": 0.7878, "step": 20755 }, { "epoch": 1.41, "learning_rate": 7.197315335017534e-06, "loss": 0.7503, "step": 20760 }, { "epoch": 1.41, "learning_rate": 7.196372685796163e-06, "loss": 0.4871, "step": 20765 }, { "epoch": 1.41, "learning_rate": 7.1954300365747905e-06, "loss": 0.2922, "step": 20770 }, { "epoch": 1.41, "learning_rate": 7.194487387353418e-06, "loss": 0.6307, "step": 20775 }, { "epoch": 1.41, "learning_rate": 7.193544738132047e-06, "loss": 0.1217, "step": 20780 }, { "epoch": 1.41, "learning_rate": 7.192602088910675e-06, "loss": 0.3873, "step": 20785 }, { "epoch": 1.41, "learning_rate": 7.191659439689304e-06, "loss": 0.4453, "step": 20790 }, { "epoch": 1.41, "learning_rate": 7.190716790467932e-06, "loss": 0.469, "step": 20795 }, { "epoch": 1.41, "learning_rate": 7.18977414124656e-06, "loss": 0.3903, "step": 20800 }, { "epoch": 1.41, "learning_rate": 7.188831492025188e-06, "loss": 0.162, "step": 20805 }, { "epoch": 1.41, "learning_rate": 7.187888842803816e-06, "loss": 0.2809, "step": 20810 }, { "epoch": 1.41, "learning_rate": 7.186946193582444e-06, "loss": 0.4526, "step": 20815 }, { "epoch": 1.41, "learning_rate": 7.186003544361073e-06, "loss": 0.518, "step": 20820 }, { "epoch": 1.41, "learning_rate": 7.185060895139701e-06, "loss": 0.3671, "step": 20825 }, { "epoch": 1.41, "learning_rate": 7.184118245918329e-06, "loss": 0.9555, "step": 20830 }, { "epoch": 1.41, "learning_rate": 7.1831755966969574e-06, "loss": 0.9823, "step": 20835 }, { "epoch": 1.41, "learning_rate": 7.182232947475586e-06, "loss": 0.2496, "step": 20840 }, { "epoch": 1.41, "learning_rate": 7.181290298254214e-06, "loss": 0.3268, "step": 20845 }, { "epoch": 1.42, "learning_rate": 7.180347649032843e-06, "loss": 0.3745, "step": 20850 }, { "epoch": 1.42, "learning_rate": 7.179404999811471e-06, "loss": 0.6601, "step": 20855 }, { "epoch": 1.42, "learning_rate": 7.178462350590098e-06, "loss": 0.5971, "step": 20860 }, { "epoch": 1.42, "learning_rate": 7.177519701368727e-06, "loss": 0.8641, "step": 20865 }, { "epoch": 1.42, "learning_rate": 7.176577052147355e-06, "loss": 0.9438, "step": 20870 }, { "epoch": 1.42, "learning_rate": 7.175634402925984e-06, "loss": 0.4518, "step": 20875 }, { "epoch": 1.42, "learning_rate": 7.174691753704612e-06, "loss": 0.4806, "step": 20880 }, { "epoch": 1.42, "learning_rate": 7.173749104483241e-06, "loss": 0.8518, "step": 20885 }, { "epoch": 1.42, "learning_rate": 7.172806455261868e-06, "loss": 0.7469, "step": 20890 }, { "epoch": 1.42, "learning_rate": 7.171863806040497e-06, "loss": 0.6406, "step": 20895 }, { "epoch": 1.42, "learning_rate": 7.170921156819125e-06, "loss": 0.3017, "step": 20900 }, { "epoch": 1.42, "learning_rate": 7.169978507597754e-06, "loss": 0.2988, "step": 20905 }, { "epoch": 1.42, "learning_rate": 7.169035858376382e-06, "loss": 0.4068, "step": 20910 }, { "epoch": 1.42, "learning_rate": 7.168093209155009e-06, "loss": 1.0146, "step": 20915 }, { "epoch": 1.42, "learning_rate": 7.1671505599336375e-06, "loss": 0.5333, "step": 20920 }, { "epoch": 1.42, "learning_rate": 7.166207910712266e-06, "loss": 0.7495, "step": 20925 }, { "epoch": 1.42, "learning_rate": 7.1652652614908945e-06, "loss": 0.307, "step": 20930 }, { "epoch": 1.42, "learning_rate": 7.164322612269523e-06, "loss": 0.4271, "step": 20935 }, { "epoch": 1.42, "learning_rate": 7.1633799630481514e-06, "loss": 0.198, "step": 20940 }, { "epoch": 1.42, "learning_rate": 7.162437313826779e-06, "loss": 0.1672, "step": 20945 }, { "epoch": 1.42, "learning_rate": 7.1614946646054076e-06, "loss": 0.3027, "step": 20950 }, { "epoch": 1.42, "learning_rate": 7.160552015384036e-06, "loss": 0.5098, "step": 20955 }, { "epoch": 1.42, "learning_rate": 7.1596093661626645e-06, "loss": 0.2284, "step": 20960 }, { "epoch": 1.42, "learning_rate": 7.158666716941293e-06, "loss": 0.2329, "step": 20965 }, { "epoch": 1.42, "learning_rate": 7.1577240677199215e-06, "loss": 0.5251, "step": 20970 }, { "epoch": 1.42, "learning_rate": 7.156781418498548e-06, "loss": 0.9577, "step": 20975 }, { "epoch": 1.42, "learning_rate": 7.155838769277177e-06, "loss": 0.4531, "step": 20980 }, { "epoch": 1.42, "learning_rate": 7.154896120055805e-06, "loss": 0.5127, "step": 20985 }, { "epoch": 1.42, "learning_rate": 7.153953470834434e-06, "loss": 0.6528, "step": 20990 }, { "epoch": 1.42, "learning_rate": 7.153010821613062e-06, "loss": 0.1543, "step": 20995 }, { "epoch": 1.43, "learning_rate": 7.152068172391691e-06, "loss": 0.3321, "step": 21000 }, { "epoch": 1.43, "learning_rate": 7.151125523170318e-06, "loss": 1.2704, "step": 21005 }, { "epoch": 1.43, "learning_rate": 7.150182873948947e-06, "loss": 0.6487, "step": 21010 }, { "epoch": 1.43, "learning_rate": 7.149240224727575e-06, "loss": 0.6396, "step": 21015 }, { "epoch": 1.43, "learning_rate": 7.148297575506203e-06, "loss": 0.4092, "step": 21020 }, { "epoch": 1.43, "learning_rate": 7.1473549262848315e-06, "loss": 0.1767, "step": 21025 }, { "epoch": 1.43, "learning_rate": 7.146412277063459e-06, "loss": 0.6592, "step": 21030 }, { "epoch": 1.43, "learning_rate": 7.145469627842088e-06, "loss": 0.3484, "step": 21035 }, { "epoch": 1.43, "learning_rate": 7.144526978620716e-06, "loss": 0.9509, "step": 21040 }, { "epoch": 1.43, "learning_rate": 7.143584329399345e-06, "loss": 0.2233, "step": 21045 }, { "epoch": 1.43, "learning_rate": 7.142641680177973e-06, "loss": 0.4415, "step": 21050 }, { "epoch": 1.43, "learning_rate": 7.1416990309566016e-06, "loss": 0.2523, "step": 21055 }, { "epoch": 1.43, "learning_rate": 7.140756381735229e-06, "loss": 0.4992, "step": 21060 }, { "epoch": 1.43, "learning_rate": 7.139813732513857e-06, "loss": 1.1138, "step": 21065 }, { "epoch": 1.43, "learning_rate": 7.138871083292485e-06, "loss": 0.7965, "step": 21070 }, { "epoch": 1.43, "learning_rate": 7.137928434071114e-06, "loss": 0.8679, "step": 21075 }, { "epoch": 1.43, "learning_rate": 7.136985784849742e-06, "loss": 0.3905, "step": 21080 }, { "epoch": 1.43, "learning_rate": 7.136043135628371e-06, "loss": 0.3024, "step": 21085 }, { "epoch": 1.43, "learning_rate": 7.1351004864069984e-06, "loss": 0.2557, "step": 21090 }, { "epoch": 1.43, "learning_rate": 7.134157837185627e-06, "loss": 0.4953, "step": 21095 }, { "epoch": 1.43, "learning_rate": 7.133215187964255e-06, "loss": 0.4191, "step": 21100 }, { "epoch": 1.43, "learning_rate": 7.132272538742884e-06, "loss": 0.2979, "step": 21105 }, { "epoch": 1.43, "learning_rate": 7.131329889521512e-06, "loss": 0.2655, "step": 21110 }, { "epoch": 1.43, "learning_rate": 7.130387240300141e-06, "loss": 0.5858, "step": 21115 }, { "epoch": 1.43, "learning_rate": 7.129444591078768e-06, "loss": 0.5009, "step": 21120 }, { "epoch": 1.43, "learning_rate": 7.128501941857396e-06, "loss": 0.3769, "step": 21125 }, { "epoch": 1.43, "learning_rate": 7.127559292636025e-06, "loss": 0.2423, "step": 21130 }, { "epoch": 1.43, "learning_rate": 7.126616643414653e-06, "loss": 0.5069, "step": 21135 }, { "epoch": 1.43, "learning_rate": 7.125673994193282e-06, "loss": 0.4572, "step": 21140 }, { "epoch": 1.44, "learning_rate": 7.124731344971909e-06, "loss": 0.1285, "step": 21145 }, { "epoch": 1.44, "learning_rate": 7.123788695750538e-06, "loss": 0.2802, "step": 21150 }, { "epoch": 1.44, "learning_rate": 7.122846046529166e-06, "loss": 0.3921, "step": 21155 }, { "epoch": 1.44, "learning_rate": 7.121903397307795e-06, "loss": 0.4689, "step": 21160 }, { "epoch": 1.44, "learning_rate": 7.120960748086423e-06, "loss": 0.2457, "step": 21165 }, { "epoch": 1.44, "learning_rate": 7.120018098865052e-06, "loss": 0.8328, "step": 21170 }, { "epoch": 1.44, "learning_rate": 7.1190754496436785e-06, "loss": 0.2861, "step": 21175 }, { "epoch": 1.44, "learning_rate": 7.118132800422307e-06, "loss": 0.1524, "step": 21180 }, { "epoch": 1.44, "learning_rate": 7.1171901512009355e-06, "loss": 0.0874, "step": 21185 }, { "epoch": 1.44, "learning_rate": 7.116247501979564e-06, "loss": 0.1685, "step": 21190 }, { "epoch": 1.44, "learning_rate": 7.1153048527581924e-06, "loss": 0.5178, "step": 21195 }, { "epoch": 1.44, "learning_rate": 7.114362203536821e-06, "loss": 0.148, "step": 21200 }, { "epoch": 1.44, "learning_rate": 7.1134195543154486e-06, "loss": 0.4577, "step": 21205 }, { "epoch": 1.44, "learning_rate": 7.112476905094077e-06, "loss": 0.6125, "step": 21210 }, { "epoch": 1.44, "learning_rate": 7.1115342558727055e-06, "loss": 0.2232, "step": 21215 }, { "epoch": 1.44, "learning_rate": 7.110591606651334e-06, "loss": 0.7509, "step": 21220 }, { "epoch": 1.44, "learning_rate": 7.109648957429962e-06, "loss": 0.3792, "step": 21225 }, { "epoch": 1.44, "learning_rate": 7.10870630820859e-06, "loss": 0.3273, "step": 21230 }, { "epoch": 1.44, "learning_rate": 7.107763658987218e-06, "loss": 0.6505, "step": 21235 }, { "epoch": 1.44, "learning_rate": 7.106821009765846e-06, "loss": 0.3118, "step": 21240 }, { "epoch": 1.44, "learning_rate": 7.105878360544475e-06, "loss": 0.3164, "step": 21245 }, { "epoch": 1.44, "learning_rate": 7.104935711323103e-06, "loss": 0.3738, "step": 21250 }, { "epoch": 1.44, "learning_rate": 7.103993062101732e-06, "loss": 0.5132, "step": 21255 }, { "epoch": 1.44, "learning_rate": 7.103050412880359e-06, "loss": 0.2036, "step": 21260 }, { "epoch": 1.44, "learning_rate": 7.102107763658988e-06, "loss": 0.6006, "step": 21265 }, { "epoch": 1.44, "learning_rate": 7.1011651144376155e-06, "loss": 0.7105, "step": 21270 }, { "epoch": 1.44, "learning_rate": 7.100222465216244e-06, "loss": 0.3097, "step": 21275 }, { "epoch": 1.44, "learning_rate": 7.0992798159948725e-06, "loss": 0.1656, "step": 21280 }, { "epoch": 1.44, "learning_rate": 7.098337166773501e-06, "loss": 0.3816, "step": 21285 }, { "epoch": 1.44, "learning_rate": 7.097394517552129e-06, "loss": 0.2446, "step": 21290 }, { "epoch": 1.45, "learning_rate": 7.096451868330757e-06, "loss": 0.2908, "step": 21295 }, { "epoch": 1.45, "learning_rate": 7.095509219109386e-06, "loss": 0.7023, "step": 21300 }, { "epoch": 1.45, "learning_rate": 7.094566569888014e-06, "loss": 0.2858, "step": 21305 }, { "epoch": 1.45, "learning_rate": 7.0936239206666426e-06, "loss": 0.4482, "step": 21310 }, { "epoch": 1.45, "learning_rate": 7.092681271445271e-06, "loss": 0.5359, "step": 21315 }, { "epoch": 1.45, "learning_rate": 7.091738622223898e-06, "loss": 0.5147, "step": 21320 }, { "epoch": 1.45, "learning_rate": 7.090795973002526e-06, "loss": 0.772, "step": 21325 }, { "epoch": 1.45, "learning_rate": 7.089853323781155e-06, "loss": 0.3371, "step": 21330 }, { "epoch": 1.45, "learning_rate": 7.088910674559783e-06, "loss": 0.7885, "step": 21335 }, { "epoch": 1.45, "learning_rate": 7.087968025338412e-06, "loss": 0.185, "step": 21340 }, { "epoch": 1.45, "learning_rate": 7.08702537611704e-06, "loss": 0.7219, "step": 21345 }, { "epoch": 1.45, "learning_rate": 7.086082726895668e-06, "loss": 0.3321, "step": 21350 }, { "epoch": 1.45, "learning_rate": 7.085140077674296e-06, "loss": 0.501, "step": 21355 }, { "epoch": 1.45, "learning_rate": 7.084197428452925e-06, "loss": 1.137, "step": 21360 }, { "epoch": 1.45, "learning_rate": 7.083254779231553e-06, "loss": 0.2976, "step": 21365 }, { "epoch": 1.45, "learning_rate": 7.082312130010182e-06, "loss": 0.7258, "step": 21370 }, { "epoch": 1.45, "learning_rate": 7.081369480788809e-06, "loss": 0.5085, "step": 21375 }, { "epoch": 1.45, "learning_rate": 7.080426831567437e-06, "loss": 0.1552, "step": 21380 }, { "epoch": 1.45, "learning_rate": 7.079484182346066e-06, "loss": 0.9874, "step": 21385 }, { "epoch": 1.45, "learning_rate": 7.078541533124694e-06, "loss": 0.8066, "step": 21390 }, { "epoch": 1.45, "learning_rate": 7.077598883903323e-06, "loss": 0.7399, "step": 21395 }, { "epoch": 1.45, "learning_rate": 7.076656234681951e-06, "loss": 0.5503, "step": 21400 }, { "epoch": 1.45, "learning_rate": 7.075713585460579e-06, "loss": 0.5455, "step": 21405 }, { "epoch": 1.45, "learning_rate": 7.074770936239207e-06, "loss": 0.6419, "step": 21410 }, { "epoch": 1.45, "learning_rate": 7.073828287017836e-06, "loss": 0.1231, "step": 21415 }, { "epoch": 1.45, "learning_rate": 7.072885637796464e-06, "loss": 0.3961, "step": 21420 }, { "epoch": 1.45, "learning_rate": 7.071942988575093e-06, "loss": 0.4844, "step": 21425 }, { "epoch": 1.45, "learning_rate": 7.071000339353721e-06, "loss": 0.6519, "step": 21430 }, { "epoch": 1.45, "learning_rate": 7.070057690132348e-06, "loss": 0.2697, "step": 21435 }, { "epoch": 1.46, "learning_rate": 7.0691150409109765e-06, "loss": 0.4023, "step": 21440 }, { "epoch": 1.46, "learning_rate": 7.068172391689605e-06, "loss": 0.4238, "step": 21445 }, { "epoch": 1.46, "learning_rate": 7.0672297424682334e-06, "loss": 0.4303, "step": 21450 }, { "epoch": 1.46, "learning_rate": 7.066287093246862e-06, "loss": 0.2976, "step": 21455 }, { "epoch": 1.46, "learning_rate": 7.06534444402549e-06, "loss": 0.3805, "step": 21460 }, { "epoch": 1.46, "learning_rate": 7.064401794804118e-06, "loss": 0.2114, "step": 21465 }, { "epoch": 1.46, "learning_rate": 7.0634591455827465e-06, "loss": 0.2962, "step": 21470 }, { "epoch": 1.46, "learning_rate": 7.062516496361375e-06, "loss": 0.4526, "step": 21475 }, { "epoch": 1.46, "learning_rate": 7.061573847140003e-06, "loss": 0.3221, "step": 21480 }, { "epoch": 1.46, "learning_rate": 7.060631197918631e-06, "loss": 0.2946, "step": 21485 }, { "epoch": 1.46, "learning_rate": 7.059688548697259e-06, "loss": 0.8096, "step": 21490 }, { "epoch": 1.46, "learning_rate": 7.058745899475887e-06, "loss": 0.7905, "step": 21495 }, { "epoch": 1.46, "learning_rate": 7.057803250254516e-06, "loss": 0.9609, "step": 21500 }, { "epoch": 1.46, "learning_rate": 7.056860601033144e-06, "loss": 0.5677, "step": 21505 }, { "epoch": 1.46, "learning_rate": 7.055917951811773e-06, "loss": 0.6828, "step": 21510 }, { "epoch": 1.46, "learning_rate": 7.054975302590401e-06, "loss": 0.8666, "step": 21515 }, { "epoch": 1.46, "learning_rate": 7.054032653369028e-06, "loss": 0.2943, "step": 21520 }, { "epoch": 1.46, "learning_rate": 7.0530900041476565e-06, "loss": 0.7725, "step": 21525 }, { "epoch": 1.46, "learning_rate": 7.052147354926285e-06, "loss": 0.3274, "step": 21530 }, { "epoch": 1.46, "learning_rate": 7.0512047057049135e-06, "loss": 0.2891, "step": 21535 }, { "epoch": 1.46, "learning_rate": 7.050262056483542e-06, "loss": 0.7425, "step": 21540 }, { "epoch": 1.46, "learning_rate": 7.0493194072621704e-06, "loss": 0.159, "step": 21545 }, { "epoch": 1.46, "learning_rate": 7.048376758040798e-06, "loss": 0.3075, "step": 21550 }, { "epoch": 1.46, "learning_rate": 7.047434108819427e-06, "loss": 0.3474, "step": 21555 }, { "epoch": 1.46, "learning_rate": 7.046491459598055e-06, "loss": 0.4834, "step": 21560 }, { "epoch": 1.46, "learning_rate": 7.0455488103766836e-06, "loss": 0.5123, "step": 21565 }, { "epoch": 1.46, "learning_rate": 7.044606161155312e-06, "loss": 0.3604, "step": 21570 }, { "epoch": 1.46, "learning_rate": 7.0436635119339405e-06, "loss": 0.386, "step": 21575 }, { "epoch": 1.46, "learning_rate": 7.042720862712567e-06, "loss": 0.4198, "step": 21580 }, { "epoch": 1.46, "learning_rate": 7.041778213491196e-06, "loss": 0.4086, "step": 21585 }, { "epoch": 1.47, "learning_rate": 7.040835564269824e-06, "loss": 0.4279, "step": 21590 }, { "epoch": 1.47, "learning_rate": 7.039892915048453e-06, "loss": 0.8039, "step": 21595 }, { "epoch": 1.47, "learning_rate": 7.038950265827081e-06, "loss": 0.495, "step": 21600 }, { "epoch": 1.47, "learning_rate": 7.038007616605709e-06, "loss": 0.5293, "step": 21605 }, { "epoch": 1.47, "learning_rate": 7.037064967384337e-06, "loss": 0.6079, "step": 21610 }, { "epoch": 1.47, "learning_rate": 7.036122318162966e-06, "loss": 0.6067, "step": 21615 }, { "epoch": 1.47, "learning_rate": 7.035179668941594e-06, "loss": 0.813, "step": 21620 }, { "epoch": 1.47, "learning_rate": 7.034237019720223e-06, "loss": 0.5561, "step": 21625 }, { "epoch": 1.47, "learning_rate": 7.033294370498851e-06, "loss": 0.2053, "step": 21630 }, { "epoch": 1.47, "learning_rate": 7.032351721277478e-06, "loss": 0.1336, "step": 21635 }, { "epoch": 1.47, "learning_rate": 7.031409072056107e-06, "loss": 0.8584, "step": 21640 }, { "epoch": 1.47, "learning_rate": 7.030466422834735e-06, "loss": 0.3216, "step": 21645 }, { "epoch": 1.47, "learning_rate": 7.029523773613364e-06, "loss": 0.8621, "step": 21650 }, { "epoch": 1.47, "learning_rate": 7.028581124391992e-06, "loss": 0.4819, "step": 21655 }, { "epoch": 1.47, "learning_rate": 7.0276384751706206e-06, "loss": 0.3044, "step": 21660 }, { "epoch": 1.47, "learning_rate": 7.026695825949248e-06, "loss": 0.6861, "step": 21665 }, { "epoch": 1.47, "learning_rate": 7.025753176727877e-06, "loss": 0.6653, "step": 21670 }, { "epoch": 1.47, "learning_rate": 7.024810527506505e-06, "loss": 0.1813, "step": 21675 }, { "epoch": 1.47, "learning_rate": 7.023867878285134e-06, "loss": 0.4911, "step": 21680 }, { "epoch": 1.47, "learning_rate": 7.022925229063761e-06, "loss": 0.5148, "step": 21685 }, { "epoch": 1.47, "learning_rate": 7.02198257984239e-06, "loss": 0.5541, "step": 21690 }, { "epoch": 1.47, "learning_rate": 7.0210399306210174e-06, "loss": 0.1075, "step": 21695 }, { "epoch": 1.47, "learning_rate": 7.020097281399646e-06, "loss": 0.4381, "step": 21700 }, { "epoch": 1.47, "learning_rate": 7.019154632178274e-06, "loss": 0.8458, "step": 21705 }, { "epoch": 1.47, "learning_rate": 7.018211982956903e-06, "loss": 0.2475, "step": 21710 }, { "epoch": 1.47, "learning_rate": 7.017269333735531e-06, "loss": 0.5375, "step": 21715 }, { "epoch": 1.47, "learning_rate": 7.016326684514159e-06, "loss": 0.3242, "step": 21720 }, { "epoch": 1.47, "learning_rate": 7.0153840352927875e-06, "loss": 0.821, "step": 21725 }, { "epoch": 1.47, "learning_rate": 7.014441386071415e-06, "loss": 0.4187, "step": 21730 }, { "epoch": 1.48, "learning_rate": 7.013498736850044e-06, "loss": 0.7294, "step": 21735 }, { "epoch": 1.48, "learning_rate": 7.012556087628672e-06, "loss": 0.7721, "step": 21740 }, { "epoch": 1.48, "learning_rate": 7.011613438407301e-06, "loss": 0.7484, "step": 21745 }, { "epoch": 1.48, "learning_rate": 7.010670789185928e-06, "loss": 0.3133, "step": 21750 }, { "epoch": 1.48, "learning_rate": 7.009728139964557e-06, "loss": 0.5098, "step": 21755 }, { "epoch": 1.48, "learning_rate": 7.008785490743185e-06, "loss": 0.233, "step": 21760 }, { "epoch": 1.48, "learning_rate": 7.007842841521814e-06, "loss": 0.1419, "step": 21765 }, { "epoch": 1.48, "learning_rate": 7.006900192300442e-06, "loss": 0.6572, "step": 21770 }, { "epoch": 1.48, "learning_rate": 7.005957543079071e-06, "loss": 0.1762, "step": 21775 }, { "epoch": 1.48, "learning_rate": 7.0050148938576975e-06, "loss": 0.3083, "step": 21780 }, { "epoch": 1.48, "learning_rate": 7.004072244636326e-06, "loss": 0.3535, "step": 21785 }, { "epoch": 1.48, "learning_rate": 7.0031295954149545e-06, "loss": 0.3839, "step": 21790 }, { "epoch": 1.48, "learning_rate": 7.002186946193583e-06, "loss": 0.9276, "step": 21795 }, { "epoch": 1.48, "learning_rate": 7.0012442969722114e-06, "loss": 0.6135, "step": 21800 }, { "epoch": 1.48, "learning_rate": 7.000301647750839e-06, "loss": 0.7405, "step": 21805 }, { "epoch": 1.48, "learning_rate": 6.9993589985294676e-06, "loss": 0.6664, "step": 21810 }, { "epoch": 1.48, "learning_rate": 6.998416349308096e-06, "loss": 0.5939, "step": 21815 }, { "epoch": 1.48, "learning_rate": 6.9974737000867245e-06, "loss": 0.3187, "step": 21820 }, { "epoch": 1.48, "learning_rate": 6.996531050865353e-06, "loss": 0.2537, "step": 21825 }, { "epoch": 1.48, "learning_rate": 6.9955884016439815e-06, "loss": 0.2809, "step": 21830 }, { "epoch": 1.48, "learning_rate": 6.994645752422608e-06, "loss": 0.5749, "step": 21835 }, { "epoch": 1.48, "learning_rate": 6.993703103201237e-06, "loss": 0.2696, "step": 21840 }, { "epoch": 1.48, "learning_rate": 6.992760453979865e-06, "loss": 0.3351, "step": 21845 }, { "epoch": 1.48, "learning_rate": 6.991817804758494e-06, "loss": 0.5691, "step": 21850 }, { "epoch": 1.48, "learning_rate": 6.990875155537122e-06, "loss": 0.2836, "step": 21855 }, { "epoch": 1.48, "learning_rate": 6.989932506315751e-06, "loss": 0.3399, "step": 21860 }, { "epoch": 1.48, "learning_rate": 6.988989857094378e-06, "loss": 0.6076, "step": 21865 }, { "epoch": 1.48, "learning_rate": 6.988047207873007e-06, "loss": 0.808, "step": 21870 }, { "epoch": 1.48, "learning_rate": 6.987104558651635e-06, "loss": 0.3674, "step": 21875 }, { "epoch": 1.49, "learning_rate": 6.986161909430264e-06, "loss": 0.4373, "step": 21880 }, { "epoch": 1.49, "learning_rate": 6.985219260208892e-06, "loss": 0.5639, "step": 21885 }, { "epoch": 1.49, "learning_rate": 6.984276610987521e-06, "loss": 0.2256, "step": 21890 }, { "epoch": 1.49, "learning_rate": 6.983333961766148e-06, "loss": 0.6902, "step": 21895 }, { "epoch": 1.49, "learning_rate": 6.982391312544776e-06, "loss": 0.4076, "step": 21900 }, { "epoch": 1.49, "learning_rate": 6.981448663323405e-06, "loss": 0.3471, "step": 21905 }, { "epoch": 1.49, "learning_rate": 6.980506014102033e-06, "loss": 0.5951, "step": 21910 }, { "epoch": 1.49, "learning_rate": 6.9795633648806616e-06, "loss": 0.3967, "step": 21915 }, { "epoch": 1.49, "learning_rate": 6.978620715659289e-06, "loss": 0.4869, "step": 21920 }, { "epoch": 1.49, "learning_rate": 6.977678066437918e-06, "loss": 0.468, "step": 21925 }, { "epoch": 1.49, "learning_rate": 6.976735417216546e-06, "loss": 0.2252, "step": 21930 }, { "epoch": 1.49, "learning_rate": 6.975792767995174e-06, "loss": 0.2887, "step": 21935 }, { "epoch": 1.49, "learning_rate": 6.974850118773802e-06, "loss": 0.1331, "step": 21940 }, { "epoch": 1.49, "learning_rate": 6.973907469552431e-06, "loss": 0.719, "step": 21945 }, { "epoch": 1.49, "learning_rate": 6.9729648203310584e-06, "loss": 0.4564, "step": 21950 }, { "epoch": 1.49, "learning_rate": 6.972022171109687e-06, "loss": 0.4297, "step": 21955 }, { "epoch": 1.49, "learning_rate": 6.971079521888315e-06, "loss": 0.199, "step": 21960 }, { "epoch": 1.49, "learning_rate": 6.970136872666944e-06, "loss": 0.2889, "step": 21965 }, { "epoch": 1.49, "learning_rate": 6.969194223445572e-06, "loss": 0.407, "step": 21970 }, { "epoch": 1.49, "learning_rate": 6.968251574224201e-06, "loss": 0.3291, "step": 21975 }, { "epoch": 1.49, "learning_rate": 6.967308925002828e-06, "loss": 0.3792, "step": 21980 }, { "epoch": 1.49, "learning_rate": 6.966366275781456e-06, "loss": 0.7368, "step": 21985 }, { "epoch": 1.49, "learning_rate": 6.965423626560085e-06, "loss": 0.145, "step": 21990 }, { "epoch": 1.49, "learning_rate": 6.964480977338713e-06, "loss": 0.1799, "step": 21995 }, { "epoch": 1.49, "learning_rate": 6.963538328117342e-06, "loss": 0.6867, "step": 22000 }, { "epoch": 1.49, "learning_rate": 6.96259567889597e-06, "loss": 0.5744, "step": 22005 }, { "epoch": 1.49, "learning_rate": 6.961653029674598e-06, "loss": 0.1995, "step": 22010 }, { "epoch": 1.49, "learning_rate": 6.960710380453226e-06, "loss": 0.0484, "step": 22015 }, { "epoch": 1.49, "learning_rate": 6.959767731231855e-06, "loss": 0.345, "step": 22020 }, { "epoch": 1.49, "learning_rate": 6.958825082010483e-06, "loss": 0.6653, "step": 22025 }, { "epoch": 1.5, "learning_rate": 6.957882432789112e-06, "loss": 1.1178, "step": 22030 }, { "epoch": 1.5, "learning_rate": 6.9569397835677385e-06, "loss": 1.4802, "step": 22035 }, { "epoch": 1.5, "learning_rate": 6.955997134346367e-06, "loss": 0.2121, "step": 22040 }, { "epoch": 1.5, "learning_rate": 6.9550544851249955e-06, "loss": 0.6874, "step": 22045 }, { "epoch": 1.5, "learning_rate": 6.954111835903624e-06, "loss": 0.7539, "step": 22050 }, { "epoch": 1.5, "learning_rate": 6.9531691866822524e-06, "loss": 0.5086, "step": 22055 }, { "epoch": 1.5, "learning_rate": 6.952226537460881e-06, "loss": 0.2935, "step": 22060 }, { "epoch": 1.5, "learning_rate": 6.9512838882395086e-06, "loss": 0.2781, "step": 22065 }, { "epoch": 1.5, "learning_rate": 6.950341239018137e-06, "loss": 0.2003, "step": 22070 }, { "epoch": 1.5, "learning_rate": 6.9493985897967655e-06, "loss": 1.0801, "step": 22075 }, { "epoch": 1.5, "learning_rate": 6.948455940575394e-06, "loss": 0.3873, "step": 22080 }, { "epoch": 1.5, "learning_rate": 6.9475132913540225e-06, "loss": 0.4549, "step": 22085 }, { "epoch": 1.5, "learning_rate": 6.946570642132651e-06, "loss": 0.4638, "step": 22090 }, { "epoch": 1.5, "learning_rate": 6.945627992911278e-06, "loss": 0.3562, "step": 22095 }, { "epoch": 1.5, "learning_rate": 6.944685343689906e-06, "loss": 0.2379, "step": 22100 }, { "epoch": 1.5, "learning_rate": 6.943742694468535e-06, "loss": 0.2873, "step": 22105 }, { "epoch": 1.5, "learning_rate": 6.942800045247163e-06, "loss": 0.3456, "step": 22110 }, { "epoch": 1.5, "learning_rate": 6.941857396025792e-06, "loss": 0.274, "step": 22115 }, { "epoch": 1.5, "learning_rate": 6.94091474680442e-06, "loss": 0.3902, "step": 22120 }, { "epoch": 1.5, "learning_rate": 6.939972097583048e-06, "loss": 0.3245, "step": 22125 }, { "epoch": 1.5, "learning_rate": 6.939029448361676e-06, "loss": 1.0194, "step": 22130 }, { "epoch": 1.5, "learning_rate": 6.938086799140305e-06, "loss": 0.4997, "step": 22135 }, { "epoch": 1.5, "learning_rate": 6.937144149918933e-06, "loss": 0.3939, "step": 22140 }, { "epoch": 1.5, "learning_rate": 6.936201500697561e-06, "loss": 0.6771, "step": 22145 }, { "epoch": 1.5, "learning_rate": 6.935258851476189e-06, "loss": 0.2779, "step": 22150 }, { "epoch": 1.5, "learning_rate": 6.934316202254817e-06, "loss": 0.461, "step": 22155 }, { "epoch": 1.5, "learning_rate": 6.933373553033446e-06, "loss": 0.4567, "step": 22160 }, { "epoch": 1.5, "learning_rate": 6.932430903812074e-06, "loss": 0.5874, "step": 22165 }, { "epoch": 1.5, "learning_rate": 6.9314882545907026e-06, "loss": 0.5663, "step": 22170 }, { "epoch": 1.51, "learning_rate": 6.930545605369331e-06, "loss": 0.9542, "step": 22175 }, { "epoch": 1.51, "learning_rate": 6.929602956147959e-06, "loss": 0.3387, "step": 22180 }, { "epoch": 1.51, "learning_rate": 6.928660306926586e-06, "loss": 0.4668, "step": 22185 }, { "epoch": 1.51, "learning_rate": 6.927717657705215e-06, "loss": 0.2333, "step": 22190 }, { "epoch": 1.51, "learning_rate": 6.926775008483843e-06, "loss": 0.3756, "step": 22195 }, { "epoch": 1.51, "learning_rate": 6.925832359262472e-06, "loss": 0.2413, "step": 22200 }, { "epoch": 1.51, "learning_rate": 6.9248897100411e-06, "loss": 0.8745, "step": 22205 }, { "epoch": 1.51, "learning_rate": 6.923947060819728e-06, "loss": 0.3501, "step": 22210 }, { "epoch": 1.51, "learning_rate": 6.923004411598356e-06, "loss": 0.0822, "step": 22215 }, { "epoch": 1.51, "learning_rate": 6.922061762376985e-06, "loss": 0.3592, "step": 22220 }, { "epoch": 1.51, "learning_rate": 6.921119113155613e-06, "loss": 0.2616, "step": 22225 }, { "epoch": 1.51, "learning_rate": 6.920176463934242e-06, "loss": 0.225, "step": 22230 }, { "epoch": 1.51, "learning_rate": 6.91923381471287e-06, "loss": 0.1437, "step": 22235 }, { "epoch": 1.51, "learning_rate": 6.918291165491497e-06, "loss": 0.8345, "step": 22240 }, { "epoch": 1.51, "learning_rate": 6.917348516270126e-06, "loss": 0.4794, "step": 22245 }, { "epoch": 1.51, "learning_rate": 6.916405867048754e-06, "loss": 0.4901, "step": 22250 }, { "epoch": 1.51, "learning_rate": 6.915463217827383e-06, "loss": 0.7335, "step": 22255 }, { "epoch": 1.51, "learning_rate": 6.914520568606011e-06, "loss": 0.5289, "step": 22260 }, { "epoch": 1.51, "learning_rate": 6.913577919384639e-06, "loss": 0.7607, "step": 22265 }, { "epoch": 1.51, "learning_rate": 6.912635270163267e-06, "loss": 0.8984, "step": 22270 }, { "epoch": 1.51, "learning_rate": 6.911692620941896e-06, "loss": 0.7864, "step": 22275 }, { "epoch": 1.51, "learning_rate": 6.910749971720524e-06, "loss": 0.3781, "step": 22280 }, { "epoch": 1.51, "learning_rate": 6.909807322499153e-06, "loss": 0.206, "step": 22285 }, { "epoch": 1.51, "learning_rate": 6.908864673277781e-06, "loss": 0.6367, "step": 22290 }, { "epoch": 1.51, "learning_rate": 6.907922024056408e-06, "loss": 0.4655, "step": 22295 }, { "epoch": 1.51, "learning_rate": 6.9069793748350365e-06, "loss": 0.4346, "step": 22300 }, { "epoch": 1.51, "learning_rate": 6.906036725613665e-06, "loss": 0.2958, "step": 22305 }, { "epoch": 1.51, "learning_rate": 6.9050940763922934e-06, "loss": 0.1498, "step": 22310 }, { "epoch": 1.51, "learning_rate": 6.904151427170922e-06, "loss": 0.5209, "step": 22315 }, { "epoch": 1.51, "learning_rate": 6.90320877794955e-06, "loss": 0.2531, "step": 22320 }, { "epoch": 1.52, "learning_rate": 6.902266128728178e-06, "loss": 0.5329, "step": 22325 }, { "epoch": 1.52, "learning_rate": 6.9013234795068065e-06, "loss": 0.6496, "step": 22330 }, { "epoch": 1.52, "learning_rate": 6.900380830285435e-06, "loss": 0.425, "step": 22335 }, { "epoch": 1.52, "learning_rate": 6.8994381810640635e-06, "loss": 0.3524, "step": 22340 }, { "epoch": 1.52, "learning_rate": 6.898495531842692e-06, "loss": 0.3309, "step": 22345 }, { "epoch": 1.52, "learning_rate": 6.89755288262132e-06, "loss": 0.6914, "step": 22350 }, { "epoch": 1.52, "learning_rate": 6.896610233399947e-06, "loss": 0.2623, "step": 22355 }, { "epoch": 1.52, "learning_rate": 6.895667584178576e-06, "loss": 0.6872, "step": 22360 }, { "epoch": 1.52, "learning_rate": 6.894724934957204e-06, "loss": 0.2586, "step": 22365 }, { "epoch": 1.52, "learning_rate": 6.893782285735833e-06, "loss": 0.7007, "step": 22370 }, { "epoch": 1.52, "learning_rate": 6.892839636514461e-06, "loss": 0.8109, "step": 22375 }, { "epoch": 1.52, "learning_rate": 6.891896987293089e-06, "loss": 0.6688, "step": 22380 }, { "epoch": 1.52, "learning_rate": 6.890954338071717e-06, "loss": 0.1755, "step": 22385 }, { "epoch": 1.52, "learning_rate": 6.890011688850346e-06, "loss": 0.3183, "step": 22390 }, { "epoch": 1.52, "learning_rate": 6.8890690396289735e-06, "loss": 0.6462, "step": 22395 }, { "epoch": 1.52, "learning_rate": 6.888126390407602e-06, "loss": 0.5275, "step": 22400 }, { "epoch": 1.52, "learning_rate": 6.8871837411862305e-06, "loss": 0.4459, "step": 22405 }, { "epoch": 1.52, "learning_rate": 6.886241091964858e-06, "loss": 0.5022, "step": 22410 }, { "epoch": 1.52, "learning_rate": 6.885298442743487e-06, "loss": 0.2154, "step": 22415 }, { "epoch": 1.52, "learning_rate": 6.884355793522115e-06, "loss": 0.2752, "step": 22420 }, { "epoch": 1.52, "learning_rate": 6.8834131443007436e-06, "loss": 0.3836, "step": 22425 }, { "epoch": 1.52, "learning_rate": 6.882470495079372e-06, "loss": 0.7532, "step": 22430 }, { "epoch": 1.52, "learning_rate": 6.8815278458580005e-06, "loss": 0.2777, "step": 22435 }, { "epoch": 1.52, "learning_rate": 6.880585196636627e-06, "loss": 0.3822, "step": 22440 }, { "epoch": 1.52, "learning_rate": 6.879642547415256e-06, "loss": 0.631, "step": 22445 }, { "epoch": 1.52, "learning_rate": 6.878699898193884e-06, "loss": 0.4356, "step": 22450 }, { "epoch": 1.52, "learning_rate": 6.877757248972513e-06, "loss": 0.4125, "step": 22455 }, { "epoch": 1.52, "learning_rate": 6.876814599751141e-06, "loss": 0.3668, "step": 22460 }, { "epoch": 1.52, "learning_rate": 6.87587195052977e-06, "loss": 0.1894, "step": 22465 }, { "epoch": 1.53, "learning_rate": 6.874929301308397e-06, "loss": 0.6103, "step": 22470 }, { "epoch": 1.53, "learning_rate": 6.873986652087026e-06, "loss": 0.2631, "step": 22475 }, { "epoch": 1.53, "learning_rate": 6.873044002865654e-06, "loss": 0.4656, "step": 22480 }, { "epoch": 1.53, "learning_rate": 6.872101353644283e-06, "loss": 0.1859, "step": 22485 }, { "epoch": 1.53, "learning_rate": 6.871158704422911e-06, "loss": 0.533, "step": 22490 }, { "epoch": 1.53, "learning_rate": 6.870216055201538e-06, "loss": 0.5676, "step": 22495 }, { "epoch": 1.53, "learning_rate": 6.869273405980167e-06, "loss": 0.4474, "step": 22500 }, { "epoch": 1.53, "learning_rate": 6.868330756758795e-06, "loss": 0.6175, "step": 22505 }, { "epoch": 1.53, "learning_rate": 6.867388107537424e-06, "loss": 0.2994, "step": 22510 }, { "epoch": 1.53, "learning_rate": 6.866445458316052e-06, "loss": 0.1583, "step": 22515 }, { "epoch": 1.53, "learning_rate": 6.865502809094681e-06, "loss": 0.5928, "step": 22520 }, { "epoch": 1.53, "learning_rate": 6.864560159873308e-06, "loss": 0.5138, "step": 22525 }, { "epoch": 1.53, "learning_rate": 6.863617510651937e-06, "loss": 0.3021, "step": 22530 }, { "epoch": 1.53, "learning_rate": 6.862674861430565e-06, "loss": 0.367, "step": 22535 }, { "epoch": 1.53, "learning_rate": 6.861732212209194e-06, "loss": 0.1168, "step": 22540 }, { "epoch": 1.53, "learning_rate": 6.860789562987822e-06, "loss": 0.6539, "step": 22545 }, { "epoch": 1.53, "learning_rate": 6.859846913766451e-06, "loss": 0.4015, "step": 22550 }, { "epoch": 1.53, "learning_rate": 6.8589042645450775e-06, "loss": 0.7454, "step": 22555 }, { "epoch": 1.53, "learning_rate": 6.857961615323706e-06, "loss": 0.7391, "step": 22560 }, { "epoch": 1.53, "learning_rate": 6.8570189661023344e-06, "loss": 0.5324, "step": 22565 }, { "epoch": 1.53, "learning_rate": 6.856076316880963e-06, "loss": 1.4592, "step": 22570 }, { "epoch": 1.53, "learning_rate": 6.855133667659591e-06, "loss": 0.4582, "step": 22575 }, { "epoch": 1.53, "learning_rate": 6.85419101843822e-06, "loss": 0.5153, "step": 22580 }, { "epoch": 1.53, "learning_rate": 6.8532483692168475e-06, "loss": 0.4576, "step": 22585 }, { "epoch": 1.53, "learning_rate": 6.852305719995476e-06, "loss": 0.7571, "step": 22590 }, { "epoch": 1.53, "learning_rate": 6.8513630707741045e-06, "loss": 0.5231, "step": 22595 }, { "epoch": 1.53, "learning_rate": 6.850420421552733e-06, "loss": 0.9037, "step": 22600 }, { "epoch": 1.53, "learning_rate": 6.849477772331361e-06, "loss": 0.6855, "step": 22605 }, { "epoch": 1.53, "learning_rate": 6.848535123109988e-06, "loss": 0.8627, "step": 22610 }, { "epoch": 1.53, "learning_rate": 6.847592473888617e-06, "loss": 0.5285, "step": 22615 }, { "epoch": 1.54, "learning_rate": 6.846649824667245e-06, "loss": 0.5219, "step": 22620 }, { "epoch": 1.54, "learning_rate": 6.845707175445874e-06, "loss": 0.1276, "step": 22625 }, { "epoch": 1.54, "learning_rate": 6.844764526224502e-06, "loss": 1.0993, "step": 22630 }, { "epoch": 1.54, "learning_rate": 6.843821877003131e-06, "loss": 0.6114, "step": 22635 }, { "epoch": 1.54, "learning_rate": 6.842879227781758e-06, "loss": 0.1719, "step": 22640 }, { "epoch": 1.54, "learning_rate": 6.841936578560386e-06, "loss": 0.7053, "step": 22645 }, { "epoch": 1.54, "learning_rate": 6.8409939293390145e-06, "loss": 0.3973, "step": 22650 }, { "epoch": 1.54, "learning_rate": 6.840051280117643e-06, "loss": 1.2115, "step": 22655 }, { "epoch": 1.54, "learning_rate": 6.8391086308962714e-06, "loss": 0.3221, "step": 22660 }, { "epoch": 1.54, "learning_rate": 6.8381659816749e-06, "loss": 0.8988, "step": 22665 }, { "epoch": 1.54, "learning_rate": 6.837223332453528e-06, "loss": 0.5723, "step": 22670 }, { "epoch": 1.54, "learning_rate": 6.836280683232156e-06, "loss": 0.7423, "step": 22675 }, { "epoch": 1.54, "learning_rate": 6.8353380340107846e-06, "loss": 1.1727, "step": 22680 }, { "epoch": 1.54, "learning_rate": 6.834395384789413e-06, "loss": 0.3548, "step": 22685 }, { "epoch": 1.54, "learning_rate": 6.8334527355680415e-06, "loss": 0.4453, "step": 22690 }, { "epoch": 1.54, "learning_rate": 6.83251008634667e-06, "loss": 0.4054, "step": 22695 }, { "epoch": 1.54, "learning_rate": 6.831567437125297e-06, "loss": 0.2555, "step": 22700 }, { "epoch": 1.54, "learning_rate": 6.830624787903925e-06, "loss": 0.3334, "step": 22705 }, { "epoch": 1.54, "learning_rate": 6.829682138682554e-06, "loss": 0.489, "step": 22710 }, { "epoch": 1.54, "learning_rate": 6.828739489461182e-06, "loss": 0.63, "step": 22715 }, { "epoch": 1.54, "learning_rate": 6.827796840239811e-06, "loss": 0.3017, "step": 22720 }, { "epoch": 1.54, "learning_rate": 6.826854191018438e-06, "loss": 0.5958, "step": 22725 }, { "epoch": 1.54, "learning_rate": 6.825911541797067e-06, "loss": 0.3123, "step": 22730 }, { "epoch": 1.54, "learning_rate": 6.824968892575695e-06, "loss": 0.1525, "step": 22735 }, { "epoch": 1.54, "learning_rate": 6.824026243354324e-06, "loss": 0.3306, "step": 22740 }, { "epoch": 1.54, "learning_rate": 6.823083594132952e-06, "loss": 0.4256, "step": 22745 }, { "epoch": 1.54, "learning_rate": 6.822140944911581e-06, "loss": 0.744, "step": 22750 }, { "epoch": 1.54, "learning_rate": 6.821198295690208e-06, "loss": 0.508, "step": 22755 }, { "epoch": 1.54, "learning_rate": 6.820255646468836e-06, "loss": 0.7067, "step": 22760 }, { "epoch": 1.55, "learning_rate": 6.819312997247465e-06, "loss": 0.0838, "step": 22765 }, { "epoch": 1.55, "learning_rate": 6.818370348026093e-06, "loss": 0.8071, "step": 22770 }, { "epoch": 1.55, "learning_rate": 6.8174276988047216e-06, "loss": 0.9132, "step": 22775 }, { "epoch": 1.55, "learning_rate": 6.81648504958335e-06, "loss": 0.6635, "step": 22780 }, { "epoch": 1.55, "learning_rate": 6.815542400361978e-06, "loss": 0.4851, "step": 22785 }, { "epoch": 1.55, "learning_rate": 6.814599751140606e-06, "loss": 0.6937, "step": 22790 }, { "epoch": 1.55, "learning_rate": 6.813657101919235e-06, "loss": 0.402, "step": 22795 }, { "epoch": 1.55, "learning_rate": 6.812714452697863e-06, "loss": 0.931, "step": 22800 }, { "epoch": 1.55, "learning_rate": 6.811771803476492e-06, "loss": 0.3877, "step": 22805 }, { "epoch": 1.55, "learning_rate": 6.8108291542551184e-06, "loss": 0.4167, "step": 22810 }, { "epoch": 1.55, "learning_rate": 6.809886505033747e-06, "loss": 0.2256, "step": 22815 }, { "epoch": 1.55, "learning_rate": 6.808943855812375e-06, "loss": 0.3983, "step": 22820 }, { "epoch": 1.55, "learning_rate": 6.808001206591004e-06, "loss": 0.625, "step": 22825 }, { "epoch": 1.55, "learning_rate": 6.807058557369632e-06, "loss": 0.6054, "step": 22830 }, { "epoch": 1.55, "learning_rate": 6.806115908148261e-06, "loss": 0.2843, "step": 22835 }, { "epoch": 1.55, "learning_rate": 6.8051732589268885e-06, "loss": 0.428, "step": 22840 }, { "epoch": 1.55, "learning_rate": 6.804230609705517e-06, "loss": 0.2901, "step": 22845 }, { "epoch": 1.55, "learning_rate": 6.8032879604841455e-06, "loss": 0.41, "step": 22850 }, { "epoch": 1.55, "learning_rate": 6.802345311262773e-06, "loss": 0.0903, "step": 22855 }, { "epoch": 1.55, "learning_rate": 6.801402662041402e-06, "loss": 1.0493, "step": 22860 }, { "epoch": 1.55, "learning_rate": 6.80046001282003e-06, "loss": 0.6596, "step": 22865 }, { "epoch": 1.55, "learning_rate": 6.799517363598658e-06, "loss": 0.5133, "step": 22870 }, { "epoch": 1.55, "learning_rate": 6.798574714377286e-06, "loss": 0.5725, "step": 22875 }, { "epoch": 1.55, "learning_rate": 6.797632065155915e-06, "loss": 0.1898, "step": 22880 }, { "epoch": 1.55, "learning_rate": 6.796689415934543e-06, "loss": 0.6803, "step": 22885 }, { "epoch": 1.55, "learning_rate": 6.795746766713172e-06, "loss": 0.1902, "step": 22890 }, { "epoch": 1.55, "learning_rate": 6.7948041174918e-06, "loss": 0.6748, "step": 22895 }, { "epoch": 1.55, "learning_rate": 6.793861468270427e-06, "loss": 0.4288, "step": 22900 }, { "epoch": 1.55, "learning_rate": 6.7929188190490555e-06, "loss": 1.1252, "step": 22905 }, { "epoch": 1.55, "learning_rate": 6.791976169827684e-06, "loss": 0.4575, "step": 22910 }, { "epoch": 1.56, "learning_rate": 6.7910335206063124e-06, "loss": 0.4255, "step": 22915 }, { "epoch": 1.56, "learning_rate": 6.790090871384941e-06, "loss": 0.6396, "step": 22920 }, { "epoch": 1.56, "learning_rate": 6.7891482221635686e-06, "loss": 0.6899, "step": 22925 }, { "epoch": 1.56, "learning_rate": 6.788205572942197e-06, "loss": 0.3797, "step": 22930 }, { "epoch": 1.56, "learning_rate": 6.7872629237208255e-06, "loss": 0.3938, "step": 22935 }, { "epoch": 1.56, "learning_rate": 6.786320274499454e-06, "loss": 0.3914, "step": 22940 }, { "epoch": 1.56, "learning_rate": 6.7853776252780825e-06, "loss": 0.3127, "step": 22945 }, { "epoch": 1.56, "learning_rate": 6.784434976056711e-06, "loss": 0.1288, "step": 22950 }, { "epoch": 1.56, "learning_rate": 6.783492326835338e-06, "loss": 0.6657, "step": 22955 }, { "epoch": 1.56, "learning_rate": 6.782549677613966e-06, "loss": 0.6266, "step": 22960 }, { "epoch": 1.56, "learning_rate": 6.781607028392595e-06, "loss": 0.2202, "step": 22965 }, { "epoch": 1.56, "learning_rate": 6.780664379171223e-06, "loss": 0.1574, "step": 22970 }, { "epoch": 1.56, "learning_rate": 6.779721729949852e-06, "loss": 0.9541, "step": 22975 }, { "epoch": 1.56, "learning_rate": 6.77877908072848e-06, "loss": 0.6021, "step": 22980 }, { "epoch": 1.56, "learning_rate": 6.777836431507108e-06, "loss": 0.4914, "step": 22985 }, { "epoch": 1.56, "learning_rate": 6.776893782285736e-06, "loss": 0.6188, "step": 22990 }, { "epoch": 1.56, "learning_rate": 6.775951133064365e-06, "loss": 1.1908, "step": 22995 }, { "epoch": 1.56, "learning_rate": 6.775008483842993e-06, "loss": 0.1118, "step": 23000 }, { "epoch": 1.56, "learning_rate": 6.774065834621622e-06, "loss": 0.7657, "step": 23005 }, { "epoch": 1.56, "learning_rate": 6.77312318540025e-06, "loss": 0.7559, "step": 23010 }, { "epoch": 1.56, "learning_rate": 6.772180536178877e-06, "loss": 0.6818, "step": 23015 }, { "epoch": 1.56, "learning_rate": 6.771237886957506e-06, "loss": 0.5651, "step": 23020 }, { "epoch": 1.56, "learning_rate": 6.770295237736134e-06, "loss": 0.2011, "step": 23025 }, { "epoch": 1.56, "learning_rate": 6.7693525885147626e-06, "loss": 0.2323, "step": 23030 }, { "epoch": 1.56, "learning_rate": 6.768409939293391e-06, "loss": 0.2257, "step": 23035 }, { "epoch": 1.56, "learning_rate": 6.767467290072019e-06, "loss": 0.6232, "step": 23040 }, { "epoch": 1.56, "learning_rate": 6.766524640850647e-06, "loss": 0.4353, "step": 23045 }, { "epoch": 1.56, "learning_rate": 6.765581991629276e-06, "loss": 0.6177, "step": 23050 }, { "epoch": 1.56, "learning_rate": 6.764639342407904e-06, "loss": 0.4829, "step": 23055 }, { "epoch": 1.57, "learning_rate": 6.763696693186532e-06, "loss": 0.6099, "step": 23060 }, { "epoch": 1.57, "learning_rate": 6.76275404396516e-06, "loss": 0.4721, "step": 23065 }, { "epoch": 1.57, "learning_rate": 6.761811394743788e-06, "loss": 0.4512, "step": 23070 }, { "epoch": 1.57, "learning_rate": 6.760868745522416e-06, "loss": 0.545, "step": 23075 }, { "epoch": 1.57, "learning_rate": 6.759926096301045e-06, "loss": 0.2912, "step": 23080 }, { "epoch": 1.57, "learning_rate": 6.758983447079673e-06, "loss": 0.2378, "step": 23085 }, { "epoch": 1.57, "learning_rate": 6.758040797858302e-06, "loss": 0.8431, "step": 23090 }, { "epoch": 1.57, "learning_rate": 6.75709814863693e-06, "loss": 0.5093, "step": 23095 }, { "epoch": 1.57, "learning_rate": 6.756155499415558e-06, "loss": 0.6209, "step": 23100 }, { "epoch": 1.57, "learning_rate": 6.755212850194186e-06, "loss": 0.6385, "step": 23105 }, { "epoch": 1.57, "learning_rate": 6.754270200972814e-06, "loss": 0.7233, "step": 23110 }, { "epoch": 1.57, "learning_rate": 6.753327551751443e-06, "loss": 0.4551, "step": 23115 }, { "epoch": 1.57, "learning_rate": 6.752384902530071e-06, "loss": 0.678, "step": 23120 }, { "epoch": 1.57, "learning_rate": 6.7514422533087e-06, "loss": 0.2775, "step": 23125 }, { "epoch": 1.57, "learning_rate": 6.750499604087327e-06, "loss": 0.7312, "step": 23130 }, { "epoch": 1.57, "learning_rate": 6.749556954865956e-06, "loss": 0.4738, "step": 23135 }, { "epoch": 1.57, "learning_rate": 6.748614305644584e-06, "loss": 0.4434, "step": 23140 }, { "epoch": 1.57, "learning_rate": 6.747671656423213e-06, "loss": 1.0739, "step": 23145 }, { "epoch": 1.57, "learning_rate": 6.746729007201841e-06, "loss": 0.4835, "step": 23150 }, { "epoch": 1.57, "learning_rate": 6.745786357980468e-06, "loss": 0.7441, "step": 23155 }, { "epoch": 1.57, "learning_rate": 6.7448437087590965e-06, "loss": 0.3874, "step": 23160 }, { "epoch": 1.57, "learning_rate": 6.743901059537725e-06, "loss": 0.2785, "step": 23165 }, { "epoch": 1.57, "learning_rate": 6.7429584103163534e-06, "loss": 0.5623, "step": 23170 }, { "epoch": 1.57, "learning_rate": 6.742015761094982e-06, "loss": 0.2335, "step": 23175 }, { "epoch": 1.57, "learning_rate": 6.74107311187361e-06, "loss": 0.4776, "step": 23180 }, { "epoch": 1.57, "learning_rate": 6.740130462652238e-06, "loss": 0.2559, "step": 23185 }, { "epoch": 1.57, "learning_rate": 6.7391878134308665e-06, "loss": 0.4884, "step": 23190 }, { "epoch": 1.57, "learning_rate": 6.738245164209495e-06, "loss": 0.7178, "step": 23195 }, { "epoch": 1.57, "learning_rate": 6.7373025149881235e-06, "loss": 1.3875, "step": 23200 }, { "epoch": 1.57, "learning_rate": 6.736359865766752e-06, "loss": 0.8537, "step": 23205 }, { "epoch": 1.58, "learning_rate": 6.7354172165453805e-06, "loss": 0.1593, "step": 23210 }, { "epoch": 1.58, "learning_rate": 6.734474567324007e-06, "loss": 0.5494, "step": 23215 }, { "epoch": 1.58, "learning_rate": 6.733531918102636e-06, "loss": 0.2526, "step": 23220 }, { "epoch": 1.58, "learning_rate": 6.732589268881264e-06, "loss": 1.0167, "step": 23225 }, { "epoch": 1.58, "learning_rate": 6.731646619659893e-06, "loss": 0.9034, "step": 23230 }, { "epoch": 1.58, "learning_rate": 6.730703970438521e-06, "loss": 0.1647, "step": 23235 }, { "epoch": 1.58, "learning_rate": 6.72976132121715e-06, "loss": 0.1854, "step": 23240 }, { "epoch": 1.58, "learning_rate": 6.728818671995777e-06, "loss": 0.5057, "step": 23245 }, { "epoch": 1.58, "learning_rate": 6.727876022774406e-06, "loss": 0.5797, "step": 23250 }, { "epoch": 1.58, "learning_rate": 6.726933373553034e-06, "loss": 0.2007, "step": 23255 }, { "epoch": 1.58, "learning_rate": 6.725990724331663e-06, "loss": 0.4473, "step": 23260 }, { "epoch": 1.58, "learning_rate": 6.725048075110291e-06, "loss": 0.7252, "step": 23265 }, { "epoch": 1.58, "learning_rate": 6.724105425888918e-06, "loss": 0.2837, "step": 23270 }, { "epoch": 1.58, "learning_rate": 6.723162776667547e-06, "loss": 0.6562, "step": 23275 }, { "epoch": 1.58, "learning_rate": 6.722220127446175e-06, "loss": 0.3193, "step": 23280 }, { "epoch": 1.58, "learning_rate": 6.7212774782248036e-06, "loss": 0.7598, "step": 23285 }, { "epoch": 1.58, "learning_rate": 6.720334829003432e-06, "loss": 0.3059, "step": 23290 }, { "epoch": 1.58, "learning_rate": 6.7193921797820605e-06, "loss": 0.2802, "step": 23295 }, { "epoch": 1.58, "learning_rate": 6.718449530560688e-06, "loss": 0.2562, "step": 23300 }, { "epoch": 1.58, "learning_rate": 6.717506881339317e-06, "loss": 0.6376, "step": 23305 }, { "epoch": 1.58, "learning_rate": 6.716564232117944e-06, "loss": 0.3412, "step": 23310 }, { "epoch": 1.58, "learning_rate": 6.715621582896573e-06, "loss": 0.718, "step": 23315 }, { "epoch": 1.58, "learning_rate": 6.714678933675201e-06, "loss": 0.5322, "step": 23320 }, { "epoch": 1.58, "learning_rate": 6.71373628445383e-06, "loss": 0.5646, "step": 23325 }, { "epoch": 1.58, "learning_rate": 6.712793635232457e-06, "loss": 0.3528, "step": 23330 }, { "epoch": 1.58, "learning_rate": 6.711850986011086e-06, "loss": 0.318, "step": 23335 }, { "epoch": 1.58, "learning_rate": 6.710908336789714e-06, "loss": 0.4989, "step": 23340 }, { "epoch": 1.58, "learning_rate": 6.709965687568343e-06, "loss": 1.0065, "step": 23345 }, { "epoch": 1.58, "learning_rate": 6.709023038346971e-06, "loss": 0.7024, "step": 23350 }, { "epoch": 1.59, "learning_rate": 6.7080803891256e-06, "loss": 0.4197, "step": 23355 }, { "epoch": 1.59, "learning_rate": 6.707137739904227e-06, "loss": 0.4871, "step": 23360 }, { "epoch": 1.59, "learning_rate": 6.706195090682855e-06, "loss": 0.5237, "step": 23365 }, { "epoch": 1.59, "learning_rate": 6.705252441461484e-06, "loss": 0.1891, "step": 23370 }, { "epoch": 1.59, "learning_rate": 6.704309792240112e-06, "loss": 0.5439, "step": 23375 }, { "epoch": 1.59, "learning_rate": 6.703367143018741e-06, "loss": 0.3386, "step": 23380 }, { "epoch": 1.59, "learning_rate": 6.702424493797368e-06, "loss": 0.3768, "step": 23385 }, { "epoch": 1.59, "learning_rate": 6.701481844575997e-06, "loss": 0.2647, "step": 23390 }, { "epoch": 1.59, "learning_rate": 6.700539195354625e-06, "loss": 0.8901, "step": 23395 }, { "epoch": 1.59, "learning_rate": 6.699596546133254e-06, "loss": 0.1128, "step": 23400 }, { "epoch": 1.59, "learning_rate": 6.698653896911882e-06, "loss": 0.5504, "step": 23405 }, { "epoch": 1.59, "learning_rate": 6.697711247690511e-06, "loss": 0.3543, "step": 23410 }, { "epoch": 1.59, "learning_rate": 6.6967685984691375e-06, "loss": 0.1296, "step": 23415 }, { "epoch": 1.59, "learning_rate": 6.695825949247766e-06, "loss": 0.3042, "step": 23420 }, { "epoch": 1.59, "learning_rate": 6.6948833000263944e-06, "loss": 0.459, "step": 23425 }, { "epoch": 1.59, "learning_rate": 6.693940650805023e-06, "loss": 0.639, "step": 23430 }, { "epoch": 1.59, "learning_rate": 6.692998001583651e-06, "loss": 0.8374, "step": 23435 }, { "epoch": 1.59, "learning_rate": 6.69205535236228e-06, "loss": 0.8634, "step": 23440 }, { "epoch": 1.59, "learning_rate": 6.6911127031409075e-06, "loss": 0.4081, "step": 23445 }, { "epoch": 1.59, "learning_rate": 6.690170053919536e-06, "loss": 0.6399, "step": 23450 }, { "epoch": 1.59, "learning_rate": 6.6892274046981645e-06, "loss": 0.9373, "step": 23455 }, { "epoch": 1.59, "learning_rate": 6.688284755476793e-06, "loss": 0.4056, "step": 23460 }, { "epoch": 1.59, "learning_rate": 6.6873421062554215e-06, "loss": 0.4712, "step": 23465 }, { "epoch": 1.59, "learning_rate": 6.68639945703405e-06, "loss": 0.2603, "step": 23470 }, { "epoch": 1.59, "learning_rate": 6.685456807812677e-06, "loss": 0.192, "step": 23475 }, { "epoch": 1.59, "learning_rate": 6.684514158591305e-06, "loss": 0.6095, "step": 23480 }, { "epoch": 1.59, "learning_rate": 6.683571509369934e-06, "loss": 0.4999, "step": 23485 }, { "epoch": 1.59, "learning_rate": 6.682628860148562e-06, "loss": 0.2467, "step": 23490 }, { "epoch": 1.59, "learning_rate": 6.681686210927191e-06, "loss": 0.3588, "step": 23495 }, { "epoch": 1.59, "learning_rate": 6.680743561705818e-06, "loss": 0.6562, "step": 23500 }, { "epoch": 1.6, "learning_rate": 6.679800912484447e-06, "loss": 0.709, "step": 23505 }, { "epoch": 1.6, "learning_rate": 6.678858263263075e-06, "loss": 0.5645, "step": 23510 }, { "epoch": 1.6, "learning_rate": 6.677915614041704e-06, "loss": 0.8957, "step": 23515 }, { "epoch": 1.6, "learning_rate": 6.6769729648203315e-06, "loss": 0.7183, "step": 23520 }, { "epoch": 1.6, "learning_rate": 6.67603031559896e-06, "loss": 0.8389, "step": 23525 }, { "epoch": 1.6, "learning_rate": 6.675087666377588e-06, "loss": 0.1525, "step": 23530 }, { "epoch": 1.6, "learning_rate": 6.674145017156216e-06, "loss": 0.1144, "step": 23535 }, { "epoch": 1.6, "learning_rate": 6.6732023679348446e-06, "loss": 0.371, "step": 23540 }, { "epoch": 1.6, "learning_rate": 6.672259718713473e-06, "loss": 0.4836, "step": 23545 }, { "epoch": 1.6, "learning_rate": 6.6713170694921015e-06, "loss": 0.6911, "step": 23550 }, { "epoch": 1.6, "learning_rate": 6.67037442027073e-06, "loss": 0.671, "step": 23555 }, { "epoch": 1.6, "learning_rate": 6.669431771049358e-06, "loss": 1.0146, "step": 23560 }, { "epoch": 1.6, "learning_rate": 6.668489121827985e-06, "loss": 0.1805, "step": 23565 }, { "epoch": 1.6, "learning_rate": 6.667546472606614e-06, "loss": 0.1968, "step": 23570 }, { "epoch": 1.6, "learning_rate": 6.666603823385242e-06, "loss": 0.5428, "step": 23575 }, { "epoch": 1.6, "learning_rate": 6.665661174163871e-06, "loss": 0.5325, "step": 23580 }, { "epoch": 1.6, "learning_rate": 6.664718524942499e-06, "loss": 0.542, "step": 23585 }, { "epoch": 1.6, "learning_rate": 6.663775875721127e-06, "loss": 1.0886, "step": 23590 }, { "epoch": 1.6, "learning_rate": 6.662833226499755e-06, "loss": 0.5365, "step": 23595 }, { "epoch": 1.6, "learning_rate": 6.661890577278384e-06, "loss": 0.6096, "step": 23600 }, { "epoch": 1.6, "learning_rate": 6.660947928057012e-06, "loss": 1.7355, "step": 23605 }, { "epoch": 1.6, "learning_rate": 6.660005278835641e-06, "loss": 0.3334, "step": 23610 }, { "epoch": 1.6, "learning_rate": 6.659062629614268e-06, "loss": 0.6212, "step": 23615 }, { "epoch": 1.6, "learning_rate": 6.658119980392896e-06, "loss": 0.3824, "step": 23620 }, { "epoch": 1.6, "learning_rate": 6.657177331171525e-06, "loss": 0.4668, "step": 23625 }, { "epoch": 1.6, "learning_rate": 6.656234681950153e-06, "loss": 0.8165, "step": 23630 }, { "epoch": 1.6, "learning_rate": 6.655292032728782e-06, "loss": 0.7663, "step": 23635 }, { "epoch": 1.6, "learning_rate": 6.65434938350741e-06, "loss": 0.8045, "step": 23640 }, { "epoch": 1.6, "learning_rate": 6.653406734286038e-06, "loss": 0.4672, "step": 23645 }, { "epoch": 1.61, "learning_rate": 6.652464085064666e-06, "loss": 0.4201, "step": 23650 }, { "epoch": 1.61, "learning_rate": 6.651521435843295e-06, "loss": 0.7216, "step": 23655 }, { "epoch": 1.61, "learning_rate": 6.650578786621923e-06, "loss": 0.3454, "step": 23660 }, { "epoch": 1.61, "learning_rate": 6.649636137400552e-06, "loss": 0.3305, "step": 23665 }, { "epoch": 1.61, "learning_rate": 6.64869348817918e-06, "loss": 0.6103, "step": 23670 }, { "epoch": 1.61, "learning_rate": 6.647750838957807e-06, "loss": 0.7974, "step": 23675 }, { "epoch": 1.61, "learning_rate": 6.6468081897364354e-06, "loss": 0.8687, "step": 23680 }, { "epoch": 1.61, "learning_rate": 6.645865540515064e-06, "loss": 0.202, "step": 23685 }, { "epoch": 1.61, "learning_rate": 6.644922891293692e-06, "loss": 0.1279, "step": 23690 }, { "epoch": 1.61, "learning_rate": 6.643980242072321e-06, "loss": 0.2497, "step": 23695 }, { "epoch": 1.61, "learning_rate": 6.643037592850949e-06, "loss": 0.2094, "step": 23700 }, { "epoch": 1.61, "learning_rate": 6.642094943629577e-06, "loss": 0.3577, "step": 23705 }, { "epoch": 1.61, "learning_rate": 6.6411522944082055e-06, "loss": 0.3489, "step": 23710 }, { "epoch": 1.61, "learning_rate": 6.640209645186834e-06, "loss": 0.2997, "step": 23715 }, { "epoch": 1.61, "learning_rate": 6.6392669959654625e-06, "loss": 0.6476, "step": 23720 }, { "epoch": 1.61, "learning_rate": 6.638324346744091e-06, "loss": 0.1467, "step": 23725 }, { "epoch": 1.61, "learning_rate": 6.637381697522718e-06, "loss": 0.728, "step": 23730 }, { "epoch": 1.61, "learning_rate": 6.636439048301346e-06, "loss": 0.3977, "step": 23735 }, { "epoch": 1.61, "learning_rate": 6.635496399079975e-06, "loss": 0.876, "step": 23740 }, { "epoch": 1.61, "learning_rate": 6.634553749858603e-06, "loss": 0.2429, "step": 23745 }, { "epoch": 1.61, "learning_rate": 6.633611100637232e-06, "loss": 0.4037, "step": 23750 }, { "epoch": 1.61, "learning_rate": 6.63266845141586e-06, "loss": 0.6759, "step": 23755 }, { "epoch": 1.61, "learning_rate": 6.631725802194488e-06, "loss": 0.7937, "step": 23760 }, { "epoch": 1.61, "learning_rate": 6.630783152973116e-06, "loss": 0.5123, "step": 23765 }, { "epoch": 1.61, "learning_rate": 6.629840503751744e-06, "loss": 0.9424, "step": 23770 }, { "epoch": 1.61, "learning_rate": 6.6288978545303724e-06, "loss": 0.3707, "step": 23775 }, { "epoch": 1.61, "learning_rate": 6.627955205309001e-06, "loss": 1.1441, "step": 23780 }, { "epoch": 1.61, "learning_rate": 6.6270125560876294e-06, "loss": 1.4599, "step": 23785 }, { "epoch": 1.61, "learning_rate": 6.626069906866257e-06, "loss": 0.9819, "step": 23790 }, { "epoch": 1.61, "learning_rate": 6.6251272576448856e-06, "loss": 0.3946, "step": 23795 }, { "epoch": 1.62, "learning_rate": 6.624184608423514e-06, "loss": 1.1199, "step": 23800 }, { "epoch": 1.62, "learning_rate": 6.6232419592021425e-06, "loss": 0.6033, "step": 23805 }, { "epoch": 1.62, "learning_rate": 6.622299309980771e-06, "loss": 0.5806, "step": 23810 }, { "epoch": 1.62, "learning_rate": 6.621356660759398e-06, "loss": 0.8048, "step": 23815 }, { "epoch": 1.62, "learning_rate": 6.620414011538026e-06, "loss": 0.112, "step": 23820 }, { "epoch": 1.62, "learning_rate": 6.619471362316655e-06, "loss": 0.6895, "step": 23825 }, { "epoch": 1.62, "learning_rate": 6.618528713095283e-06, "loss": 0.3927, "step": 23830 }, { "epoch": 1.62, "learning_rate": 6.617586063873912e-06, "loss": 0.4933, "step": 23835 }, { "epoch": 1.62, "learning_rate": 6.61664341465254e-06, "loss": 0.4094, "step": 23840 }, { "epoch": 1.62, "learning_rate": 6.615700765431168e-06, "loss": 0.1374, "step": 23845 }, { "epoch": 1.62, "learning_rate": 6.614758116209796e-06, "loss": 0.3794, "step": 23850 }, { "epoch": 1.62, "learning_rate": 6.613815466988425e-06, "loss": 1.0345, "step": 23855 }, { "epoch": 1.62, "learning_rate": 6.612872817767053e-06, "loss": 0.6668, "step": 23860 }, { "epoch": 1.62, "learning_rate": 6.611930168545682e-06, "loss": 0.2544, "step": 23865 }, { "epoch": 1.62, "learning_rate": 6.61098751932431e-06, "loss": 0.223, "step": 23870 }, { "epoch": 1.62, "learning_rate": 6.610044870102937e-06, "loss": 0.3312, "step": 23875 }, { "epoch": 1.62, "learning_rate": 6.609102220881566e-06, "loss": 0.4567, "step": 23880 }, { "epoch": 1.62, "learning_rate": 6.608159571660194e-06, "loss": 0.5405, "step": 23885 }, { "epoch": 1.62, "learning_rate": 6.6072169224388226e-06, "loss": 1.0696, "step": 23890 }, { "epoch": 1.62, "learning_rate": 6.606274273217451e-06, "loss": 0.4311, "step": 23895 }, { "epoch": 1.62, "learning_rate": 6.6053316239960796e-06, "loss": 0.2296, "step": 23900 }, { "epoch": 1.62, "learning_rate": 6.604388974774707e-06, "loss": 0.2919, "step": 23905 }, { "epoch": 1.62, "learning_rate": 6.603446325553336e-06, "loss": 0.5744, "step": 23910 }, { "epoch": 1.62, "learning_rate": 6.602503676331964e-06, "loss": 0.4693, "step": 23915 }, { "epoch": 1.62, "learning_rate": 6.601561027110593e-06, "loss": 0.3369, "step": 23920 }, { "epoch": 1.62, "learning_rate": 6.600618377889221e-06, "loss": 0.8534, "step": 23925 }, { "epoch": 1.62, "learning_rate": 6.599675728667848e-06, "loss": 0.3917, "step": 23930 }, { "epoch": 1.62, "learning_rate": 6.598733079446476e-06, "loss": 0.2041, "step": 23935 }, { "epoch": 1.62, "learning_rate": 6.597790430225105e-06, "loss": 0.6459, "step": 23940 }, { "epoch": 1.63, "learning_rate": 6.596847781003733e-06, "loss": 0.5978, "step": 23945 }, { "epoch": 1.63, "learning_rate": 6.595905131782362e-06, "loss": 0.2217, "step": 23950 }, { "epoch": 1.63, "learning_rate": 6.59496248256099e-06, "loss": 0.6244, "step": 23955 }, { "epoch": 1.63, "learning_rate": 6.594019833339618e-06, "loss": 0.7154, "step": 23960 }, { "epoch": 1.63, "learning_rate": 6.5930771841182465e-06, "loss": 0.3028, "step": 23965 }, { "epoch": 1.63, "learning_rate": 6.592134534896875e-06, "loss": 0.4778, "step": 23970 }, { "epoch": 1.63, "learning_rate": 6.5911918856755035e-06, "loss": 0.3379, "step": 23975 }, { "epoch": 1.63, "learning_rate": 6.590249236454131e-06, "loss": 0.5895, "step": 23980 }, { "epoch": 1.63, "learning_rate": 6.58930658723276e-06, "loss": 0.3345, "step": 23985 }, { "epoch": 1.63, "learning_rate": 6.588363938011387e-06, "loss": 0.614, "step": 23990 }, { "epoch": 1.63, "learning_rate": 6.587421288790016e-06, "loss": 0.2736, "step": 23995 }, { "epoch": 1.63, "learning_rate": 6.586478639568644e-06, "loss": 0.6427, "step": 24000 }, { "epoch": 1.63, "learning_rate": 6.585535990347273e-06, "loss": 0.1656, "step": 24005 }, { "epoch": 1.63, "learning_rate": 6.584593341125901e-06, "loss": 1.0922, "step": 24010 }, { "epoch": 1.63, "learning_rate": 6.58365069190453e-06, "loss": 0.3398, "step": 24015 }, { "epoch": 1.63, "learning_rate": 6.5827080426831565e-06, "loss": 0.5387, "step": 24020 }, { "epoch": 1.63, "learning_rate": 6.581765393461785e-06, "loss": 0.7299, "step": 24025 }, { "epoch": 1.63, "learning_rate": 6.5808227442404134e-06, "loss": 0.3392, "step": 24030 }, { "epoch": 1.63, "learning_rate": 6.579880095019042e-06, "loss": 0.1925, "step": 24035 }, { "epoch": 1.63, "learning_rate": 6.57893744579767e-06, "loss": 0.5199, "step": 24040 }, { "epoch": 1.63, "learning_rate": 6.577994796576298e-06, "loss": 0.2445, "step": 24045 }, { "epoch": 1.63, "learning_rate": 6.5770521473549265e-06, "loss": 0.3615, "step": 24050 }, { "epoch": 1.63, "learning_rate": 6.576109498133555e-06, "loss": 0.8264, "step": 24055 }, { "epoch": 1.63, "learning_rate": 6.5751668489121835e-06, "loss": 0.4666, "step": 24060 }, { "epoch": 1.63, "learning_rate": 6.574224199690812e-06, "loss": 0.473, "step": 24065 }, { "epoch": 1.63, "learning_rate": 6.5732815504694405e-06, "loss": 0.9836, "step": 24070 }, { "epoch": 1.63, "learning_rate": 6.572338901248067e-06, "loss": 0.2182, "step": 24075 }, { "epoch": 1.63, "learning_rate": 6.571396252026696e-06, "loss": 0.5643, "step": 24080 }, { "epoch": 1.63, "learning_rate": 6.570453602805324e-06, "loss": 0.2659, "step": 24085 }, { "epoch": 1.63, "learning_rate": 6.569510953583953e-06, "loss": 0.4707, "step": 24090 }, { "epoch": 1.64, "learning_rate": 6.568568304362581e-06, "loss": 0.6083, "step": 24095 }, { "epoch": 1.64, "learning_rate": 6.56762565514121e-06, "loss": 0.5852, "step": 24100 }, { "epoch": 1.64, "learning_rate": 6.566683005919837e-06, "loss": 0.5314, "step": 24105 }, { "epoch": 1.64, "learning_rate": 6.565740356698466e-06, "loss": 0.1954, "step": 24110 }, { "epoch": 1.64, "learning_rate": 6.564797707477094e-06, "loss": 0.6076, "step": 24115 }, { "epoch": 1.64, "learning_rate": 6.563855058255723e-06, "loss": 0.9447, "step": 24120 }, { "epoch": 1.64, "learning_rate": 6.562912409034351e-06, "loss": 0.1575, "step": 24125 }, { "epoch": 1.64, "learning_rate": 6.56196975981298e-06, "loss": 0.3267, "step": 24130 }, { "epoch": 1.64, "learning_rate": 6.561027110591607e-06, "loss": 0.2738, "step": 24135 }, { "epoch": 1.64, "learning_rate": 6.560084461370235e-06, "loss": 0.5989, "step": 24140 }, { "epoch": 1.64, "learning_rate": 6.5591418121488636e-06, "loss": 1.4123, "step": 24145 }, { "epoch": 1.64, "learning_rate": 6.558199162927492e-06, "loss": 0.9121, "step": 24150 }, { "epoch": 1.64, "learning_rate": 6.5572565137061205e-06, "loss": 0.3553, "step": 24155 }, { "epoch": 1.64, "learning_rate": 6.556313864484748e-06, "loss": 0.1238, "step": 24160 }, { "epoch": 1.64, "learning_rate": 6.555371215263377e-06, "loss": 0.5794, "step": 24165 }, { "epoch": 1.64, "learning_rate": 6.554428566042005e-06, "loss": 0.4791, "step": 24170 }, { "epoch": 1.64, "learning_rate": 6.553485916820634e-06, "loss": 0.2209, "step": 24175 }, { "epoch": 1.64, "learning_rate": 6.552543267599262e-06, "loss": 1.1209, "step": 24180 }, { "epoch": 1.64, "learning_rate": 6.55160061837789e-06, "loss": 0.58, "step": 24185 }, { "epoch": 1.64, "learning_rate": 6.550657969156517e-06, "loss": 0.1507, "step": 24190 }, { "epoch": 1.64, "learning_rate": 6.549715319935146e-06, "loss": 0.9891, "step": 24195 }, { "epoch": 1.64, "learning_rate": 6.548772670713774e-06, "loss": 0.4582, "step": 24200 }, { "epoch": 1.64, "learning_rate": 6.547830021492403e-06, "loss": 0.5086, "step": 24205 }, { "epoch": 1.64, "learning_rate": 6.546887372271031e-06, "loss": 0.5719, "step": 24210 }, { "epoch": 1.64, "learning_rate": 6.54594472304966e-06, "loss": 0.4986, "step": 24215 }, { "epoch": 1.64, "learning_rate": 6.5450020738282875e-06, "loss": 0.4771, "step": 24220 }, { "epoch": 1.64, "learning_rate": 6.544059424606916e-06, "loss": 0.6148, "step": 24225 }, { "epoch": 1.64, "learning_rate": 6.543116775385544e-06, "loss": 0.1345, "step": 24230 }, { "epoch": 1.64, "learning_rate": 6.542174126164172e-06, "loss": 0.1873, "step": 24235 }, { "epoch": 1.65, "learning_rate": 6.541231476942801e-06, "loss": 0.0841, "step": 24240 }, { "epoch": 1.65, "learning_rate": 6.540288827721429e-06, "loss": 0.5049, "step": 24245 }, { "epoch": 1.65, "learning_rate": 6.539346178500057e-06, "loss": 0.3501, "step": 24250 }, { "epoch": 1.65, "learning_rate": 6.538403529278685e-06, "loss": 0.9101, "step": 24255 }, { "epoch": 1.65, "learning_rate": 6.537460880057314e-06, "loss": 0.1547, "step": 24260 }, { "epoch": 1.65, "learning_rate": 6.536518230835942e-06, "loss": 0.5681, "step": 24265 }, { "epoch": 1.65, "learning_rate": 6.535575581614571e-06, "loss": 0.5397, "step": 24270 }, { "epoch": 1.65, "learning_rate": 6.5346329323931975e-06, "loss": 0.4412, "step": 24275 }, { "epoch": 1.65, "learning_rate": 6.533690283171826e-06, "loss": 0.5208, "step": 24280 }, { "epoch": 1.65, "learning_rate": 6.5327476339504544e-06, "loss": 0.2854, "step": 24285 }, { "epoch": 1.65, "learning_rate": 6.531804984729083e-06, "loss": 0.3997, "step": 24290 }, { "epoch": 1.65, "learning_rate": 6.530862335507711e-06, "loss": 0.3832, "step": 24295 }, { "epoch": 1.65, "learning_rate": 6.52991968628634e-06, "loss": 0.424, "step": 24300 }, { "epoch": 1.65, "learning_rate": 6.5289770370649675e-06, "loss": 0.3826, "step": 24305 }, { "epoch": 1.65, "learning_rate": 6.528034387843596e-06, "loss": 0.3248, "step": 24310 }, { "epoch": 1.65, "learning_rate": 6.5270917386222245e-06, "loss": 0.3421, "step": 24315 }, { "epoch": 1.65, "learning_rate": 6.526149089400853e-06, "loss": 0.4583, "step": 24320 }, { "epoch": 1.65, "learning_rate": 6.5252064401794815e-06, "loss": 0.4692, "step": 24325 }, { "epoch": 1.65, "learning_rate": 6.52426379095811e-06, "loss": 0.4715, "step": 24330 }, { "epoch": 1.65, "learning_rate": 6.523321141736737e-06, "loss": 0.1925, "step": 24335 }, { "epoch": 1.65, "learning_rate": 6.522378492515365e-06, "loss": 0.7224, "step": 24340 }, { "epoch": 1.65, "learning_rate": 6.521435843293994e-06, "loss": 0.5018, "step": 24345 }, { "epoch": 1.65, "learning_rate": 6.520493194072622e-06, "loss": 0.1456, "step": 24350 }, { "epoch": 1.65, "learning_rate": 6.519550544851251e-06, "loss": 0.5136, "step": 24355 }, { "epoch": 1.65, "learning_rate": 6.518607895629879e-06, "loss": 0.6648, "step": 24360 }, { "epoch": 1.65, "learning_rate": 6.517665246408507e-06, "loss": 0.6236, "step": 24365 }, { "epoch": 1.65, "learning_rate": 6.516722597187135e-06, "loss": 0.6538, "step": 24370 }, { "epoch": 1.65, "learning_rate": 6.515779947965764e-06, "loss": 0.4558, "step": 24375 }, { "epoch": 1.65, "learning_rate": 6.514837298744392e-06, "loss": 0.5854, "step": 24380 }, { "epoch": 1.66, "learning_rate": 6.513894649523021e-06, "loss": 0.5875, "step": 24385 }, { "epoch": 1.66, "learning_rate": 6.512952000301648e-06, "loss": 0.5381, "step": 24390 }, { "epoch": 1.66, "learning_rate": 6.512009351080276e-06, "loss": 0.7036, "step": 24395 }, { "epoch": 1.66, "learning_rate": 6.5110667018589046e-06, "loss": 0.2723, "step": 24400 }, { "epoch": 1.66, "learning_rate": 6.510124052637533e-06, "loss": 0.2767, "step": 24405 }, { "epoch": 1.66, "learning_rate": 6.5091814034161615e-06, "loss": 0.3395, "step": 24410 }, { "epoch": 1.66, "learning_rate": 6.50823875419479e-06, "loss": 0.7971, "step": 24415 }, { "epoch": 1.66, "learning_rate": 6.507296104973418e-06, "loss": 0.4618, "step": 24420 }, { "epoch": 1.66, "learning_rate": 6.506353455752046e-06, "loss": 0.4859, "step": 24425 }, { "epoch": 1.66, "learning_rate": 6.505410806530675e-06, "loss": 1.1754, "step": 24430 }, { "epoch": 1.66, "learning_rate": 6.504468157309302e-06, "loss": 0.5274, "step": 24435 }, { "epoch": 1.66, "learning_rate": 6.503525508087931e-06, "loss": 0.687, "step": 24440 }, { "epoch": 1.66, "learning_rate": 6.502582858866559e-06, "loss": 0.7419, "step": 24445 }, { "epoch": 1.66, "learning_rate": 6.501640209645187e-06, "loss": 0.2512, "step": 24450 }, { "epoch": 1.66, "learning_rate": 6.500697560423815e-06, "loss": 0.4665, "step": 24455 }, { "epoch": 1.66, "learning_rate": 6.499754911202444e-06, "loss": 0.6291, "step": 24460 }, { "epoch": 1.66, "learning_rate": 6.498812261981072e-06, "loss": 0.784, "step": 24465 }, { "epoch": 1.66, "learning_rate": 6.497869612759701e-06, "loss": 0.4807, "step": 24470 }, { "epoch": 1.66, "learning_rate": 6.496926963538329e-06, "loss": 0.5709, "step": 24475 }, { "epoch": 1.66, "learning_rate": 6.495984314316956e-06, "loss": 0.3326, "step": 24480 }, { "epoch": 1.66, "learning_rate": 6.495041665095585e-06, "loss": 0.6239, "step": 24485 }, { "epoch": 1.66, "learning_rate": 6.494099015874213e-06, "loss": 0.8097, "step": 24490 }, { "epoch": 1.66, "learning_rate": 6.493156366652842e-06, "loss": 0.8873, "step": 24495 }, { "epoch": 1.66, "learning_rate": 6.49221371743147e-06, "loss": 0.2127, "step": 24500 }, { "epoch": 1.66, "learning_rate": 6.491271068210098e-06, "loss": 0.6912, "step": 24505 }, { "epoch": 1.66, "learning_rate": 6.490328418988726e-06, "loss": 0.2158, "step": 24510 }, { "epoch": 1.66, "learning_rate": 6.489385769767355e-06, "loss": 0.5467, "step": 24515 }, { "epoch": 1.66, "learning_rate": 6.488443120545983e-06, "loss": 0.9609, "step": 24520 }, { "epoch": 1.66, "learning_rate": 6.487500471324612e-06, "loss": 0.3491, "step": 24525 }, { "epoch": 1.66, "learning_rate": 6.48655782210324e-06, "loss": 0.4369, "step": 24530 }, { "epoch": 1.67, "learning_rate": 6.485615172881867e-06, "loss": 0.8917, "step": 24535 }, { "epoch": 1.67, "learning_rate": 6.4846725236604954e-06, "loss": 0.3091, "step": 24540 }, { "epoch": 1.67, "learning_rate": 6.483729874439124e-06, "loss": 0.628, "step": 24545 }, { "epoch": 1.67, "learning_rate": 6.482787225217752e-06, "loss": 0.3169, "step": 24550 }, { "epoch": 1.67, "learning_rate": 6.481844575996381e-06, "loss": 0.2755, "step": 24555 }, { "epoch": 1.67, "learning_rate": 6.480901926775009e-06, "loss": 0.3543, "step": 24560 }, { "epoch": 1.67, "learning_rate": 6.479959277553637e-06, "loss": 0.8194, "step": 24565 }, { "epoch": 1.67, "learning_rate": 6.4790166283322655e-06, "loss": 0.3829, "step": 24570 }, { "epoch": 1.67, "learning_rate": 6.478073979110894e-06, "loss": 1.0866, "step": 24575 }, { "epoch": 1.67, "learning_rate": 6.4771313298895225e-06, "loss": 0.1569, "step": 24580 }, { "epoch": 1.67, "learning_rate": 6.476188680668151e-06, "loss": 0.4275, "step": 24585 }, { "epoch": 1.67, "learning_rate": 6.4752460314467795e-06, "loss": 0.2282, "step": 24590 }, { "epoch": 1.67, "learning_rate": 6.474303382225406e-06, "loss": 0.5024, "step": 24595 }, { "epoch": 1.67, "learning_rate": 6.473360733004035e-06, "loss": 0.2318, "step": 24600 }, { "epoch": 1.67, "learning_rate": 6.472418083782663e-06, "loss": 0.3589, "step": 24605 }, { "epoch": 1.67, "learning_rate": 6.471475434561292e-06, "loss": 0.4653, "step": 24610 }, { "epoch": 1.67, "learning_rate": 6.47053278533992e-06, "loss": 0.6638, "step": 24615 }, { "epoch": 1.67, "learning_rate": 6.469590136118548e-06, "loss": 0.2512, "step": 24620 }, { "epoch": 1.67, "learning_rate": 6.468647486897176e-06, "loss": 1.2118, "step": 24625 }, { "epoch": 1.67, "learning_rate": 6.467704837675805e-06, "loss": 0.4936, "step": 24630 }, { "epoch": 1.67, "learning_rate": 6.466762188454433e-06, "loss": 0.3387, "step": 24635 }, { "epoch": 1.67, "learning_rate": 6.465819539233062e-06, "loss": 0.2668, "step": 24640 }, { "epoch": 1.67, "learning_rate": 6.4648768900116894e-06, "loss": 0.3437, "step": 24645 }, { "epoch": 1.67, "learning_rate": 6.463934240790317e-06, "loss": 1.0374, "step": 24650 }, { "epoch": 1.67, "learning_rate": 6.4629915915689456e-06, "loss": 0.283, "step": 24655 }, { "epoch": 1.67, "learning_rate": 6.462048942347574e-06, "loss": 0.5473, "step": 24660 }, { "epoch": 1.67, "learning_rate": 6.4611062931262025e-06, "loss": 0.4852, "step": 24665 }, { "epoch": 1.67, "learning_rate": 6.460163643904831e-06, "loss": 0.4036, "step": 24670 }, { "epoch": 1.67, "learning_rate": 6.4592209946834595e-06, "loss": 0.2909, "step": 24675 }, { "epoch": 1.68, "learning_rate": 6.458278345462087e-06, "loss": 0.3943, "step": 24680 }, { "epoch": 1.68, "learning_rate": 6.457335696240716e-06, "loss": 0.3682, "step": 24685 }, { "epoch": 1.68, "learning_rate": 6.456393047019343e-06, "loss": 0.3653, "step": 24690 }, { "epoch": 1.68, "learning_rate": 6.455450397797972e-06, "loss": 0.2185, "step": 24695 }, { "epoch": 1.68, "learning_rate": 6.4545077485766e-06, "loss": 0.39, "step": 24700 }, { "epoch": 1.68, "learning_rate": 6.453565099355229e-06, "loss": 0.4651, "step": 24705 }, { "epoch": 1.68, "learning_rate": 6.452622450133856e-06, "loss": 0.4245, "step": 24710 }, { "epoch": 1.68, "learning_rate": 6.451679800912485e-06, "loss": 0.9782, "step": 24715 }, { "epoch": 1.68, "learning_rate": 6.450737151691113e-06, "loss": 0.4613, "step": 24720 }, { "epoch": 1.68, "learning_rate": 6.449794502469742e-06, "loss": 0.5931, "step": 24725 }, { "epoch": 1.68, "learning_rate": 6.44885185324837e-06, "loss": 0.4662, "step": 24730 }, { "epoch": 1.68, "learning_rate": 6.447909204026997e-06, "loss": 0.5936, "step": 24735 }, { "epoch": 1.68, "learning_rate": 6.446966554805626e-06, "loss": 0.3571, "step": 24740 }, { "epoch": 1.68, "learning_rate": 6.446023905584254e-06, "loss": 0.225, "step": 24745 }, { "epoch": 1.68, "learning_rate": 6.445081256362883e-06, "loss": 0.3446, "step": 24750 }, { "epoch": 1.68, "learning_rate": 6.444138607141511e-06, "loss": 0.6384, "step": 24755 }, { "epoch": 1.68, "learning_rate": 6.4431959579201396e-06, "loss": 0.8358, "step": 24760 }, { "epoch": 1.68, "learning_rate": 6.442253308698767e-06, "loss": 0.4866, "step": 24765 }, { "epoch": 1.68, "learning_rate": 6.441310659477396e-06, "loss": 0.2276, "step": 24770 }, { "epoch": 1.68, "learning_rate": 6.440368010256024e-06, "loss": 0.4613, "step": 24775 }, { "epoch": 1.68, "learning_rate": 6.439425361034653e-06, "loss": 0.3257, "step": 24780 }, { "epoch": 1.68, "learning_rate": 6.438482711813281e-06, "loss": 0.48, "step": 24785 }, { "epoch": 1.68, "learning_rate": 6.43754006259191e-06, "loss": 0.8631, "step": 24790 }, { "epoch": 1.68, "learning_rate": 6.4365974133705364e-06, "loss": 0.3053, "step": 24795 }, { "epoch": 1.68, "learning_rate": 6.435654764149165e-06, "loss": 0.6934, "step": 24800 }, { "epoch": 1.68, "learning_rate": 6.434712114927793e-06, "loss": 0.3272, "step": 24805 }, { "epoch": 1.68, "learning_rate": 6.433769465706422e-06, "loss": 0.2391, "step": 24810 }, { "epoch": 1.68, "learning_rate": 6.43282681648505e-06, "loss": 0.4686, "step": 24815 }, { "epoch": 1.68, "learning_rate": 6.431884167263678e-06, "loss": 0.2473, "step": 24820 }, { "epoch": 1.68, "learning_rate": 6.4309415180423065e-06, "loss": 0.4653, "step": 24825 }, { "epoch": 1.69, "learning_rate": 6.429998868820935e-06, "loss": 0.9697, "step": 24830 }, { "epoch": 1.69, "learning_rate": 6.4290562195995635e-06, "loss": 0.3338, "step": 24835 }, { "epoch": 1.69, "learning_rate": 6.428113570378192e-06, "loss": 1.0423, "step": 24840 }, { "epoch": 1.69, "learning_rate": 6.4271709211568205e-06, "loss": 0.7329, "step": 24845 }, { "epoch": 1.69, "learning_rate": 6.426228271935447e-06, "loss": 1.0984, "step": 24850 }, { "epoch": 1.69, "learning_rate": 6.425285622714076e-06, "loss": 0.4458, "step": 24855 }, { "epoch": 1.69, "learning_rate": 6.424342973492704e-06, "loss": 0.6108, "step": 24860 }, { "epoch": 1.69, "learning_rate": 6.423400324271333e-06, "loss": 0.3527, "step": 24865 }, { "epoch": 1.69, "learning_rate": 6.422457675049961e-06, "loss": 0.6026, "step": 24870 }, { "epoch": 1.69, "learning_rate": 6.42151502582859e-06, "loss": 0.4498, "step": 24875 }, { "epoch": 1.69, "learning_rate": 6.420572376607217e-06, "loss": 0.1594, "step": 24880 }, { "epoch": 1.69, "learning_rate": 6.419629727385846e-06, "loss": 0.1984, "step": 24885 }, { "epoch": 1.69, "learning_rate": 6.418687078164474e-06, "loss": 0.1962, "step": 24890 }, { "epoch": 1.69, "learning_rate": 6.417744428943102e-06, "loss": 0.6516, "step": 24895 }, { "epoch": 1.69, "learning_rate": 6.4168017797217304e-06, "loss": 0.4326, "step": 24900 }, { "epoch": 1.69, "learning_rate": 6.415859130500359e-06, "loss": 0.2883, "step": 24905 }, { "epoch": 1.69, "learning_rate": 6.4149164812789866e-06, "loss": 0.5452, "step": 24910 }, { "epoch": 1.69, "learning_rate": 6.413973832057615e-06, "loss": 0.2509, "step": 24915 }, { "epoch": 1.69, "learning_rate": 6.4130311828362435e-06, "loss": 0.178, "step": 24920 }, { "epoch": 1.69, "learning_rate": 6.412088533614872e-06, "loss": 0.5382, "step": 24925 }, { "epoch": 1.69, "learning_rate": 6.4111458843935005e-06, "loss": 1.7844, "step": 24930 }, { "epoch": 1.69, "learning_rate": 6.410203235172128e-06, "loss": 0.6545, "step": 24935 }, { "epoch": 1.69, "learning_rate": 6.409260585950756e-06, "loss": 0.3145, "step": 24940 }, { "epoch": 1.69, "learning_rate": 6.408317936729384e-06, "loss": 0.0772, "step": 24945 }, { "epoch": 1.69, "learning_rate": 6.407375287508013e-06, "loss": 0.4331, "step": 24950 }, { "epoch": 1.69, "learning_rate": 6.406432638286641e-06, "loss": 0.5427, "step": 24955 }, { "epoch": 1.69, "learning_rate": 6.40548998906527e-06, "loss": 0.5628, "step": 24960 }, { "epoch": 1.69, "learning_rate": 6.404547339843897e-06, "loss": 0.2642, "step": 24965 }, { "epoch": 1.69, "learning_rate": 6.403604690622526e-06, "loss": 0.2638, "step": 24970 }, { "epoch": 1.7, "learning_rate": 6.402662041401154e-06, "loss": 0.7598, "step": 24975 }, { "epoch": 1.7, "learning_rate": 6.401719392179783e-06, "loss": 0.6955, "step": 24980 }, { "epoch": 1.7, "learning_rate": 6.400776742958411e-06, "loss": 0.4721, "step": 24985 }, { "epoch": 1.7, "learning_rate": 6.39983409373704e-06, "loss": 0.726, "step": 24990 }, { "epoch": 1.7, "learning_rate": 6.398891444515667e-06, "loss": 0.4438, "step": 24995 }, { "epoch": 1.7, "learning_rate": 6.397948795294295e-06, "loss": 0.4541, "step": 25000 }, { "epoch": 1.7, "learning_rate": 6.3970061460729236e-06, "loss": 0.6359, "step": 25005 }, { "epoch": 1.7, "learning_rate": 6.396063496851552e-06, "loss": 0.7128, "step": 25010 }, { "epoch": 1.7, "learning_rate": 6.3951208476301806e-06, "loss": 0.5582, "step": 25015 }, { "epoch": 1.7, "learning_rate": 6.394178198408809e-06, "loss": 0.1869, "step": 25020 }, { "epoch": 1.7, "learning_rate": 6.393235549187437e-06, "loss": 0.5594, "step": 25025 }, { "epoch": 1.7, "learning_rate": 6.392292899966065e-06, "loss": 0.8561, "step": 25030 }, { "epoch": 1.7, "learning_rate": 6.391350250744694e-06, "loss": 0.3656, "step": 25035 }, { "epoch": 1.7, "learning_rate": 6.390407601523322e-06, "loss": 0.862, "step": 25040 }, { "epoch": 1.7, "learning_rate": 6.389464952301951e-06, "loss": 0.5855, "step": 25045 }, { "epoch": 1.7, "learning_rate": 6.388522303080577e-06, "loss": 0.2775, "step": 25050 }, { "epoch": 1.7, "learning_rate": 6.387579653859206e-06, "loss": 0.3312, "step": 25055 }, { "epoch": 1.7, "learning_rate": 6.386637004637834e-06, "loss": 0.9354, "step": 25060 }, { "epoch": 1.7, "learning_rate": 6.385694355416463e-06, "loss": 0.4231, "step": 25065 }, { "epoch": 1.7, "learning_rate": 6.384751706195091e-06, "loss": 0.3698, "step": 25070 }, { "epoch": 1.7, "learning_rate": 6.38380905697372e-06, "loss": 0.2164, "step": 25075 }, { "epoch": 1.7, "learning_rate": 6.3828664077523475e-06, "loss": 0.2303, "step": 25080 }, { "epoch": 1.7, "learning_rate": 6.381923758530976e-06, "loss": 0.9198, "step": 25085 }, { "epoch": 1.7, "learning_rate": 6.3809811093096045e-06, "loss": 0.3471, "step": 25090 }, { "epoch": 1.7, "learning_rate": 6.380038460088233e-06, "loss": 0.5646, "step": 25095 }, { "epoch": 1.7, "learning_rate": 6.3790958108668614e-06, "loss": 0.127, "step": 25100 }, { "epoch": 1.7, "learning_rate": 6.378153161645489e-06, "loss": 0.4326, "step": 25105 }, { "epoch": 1.7, "learning_rate": 6.377210512424117e-06, "loss": 0.2577, "step": 25110 }, { "epoch": 1.7, "learning_rate": 6.376267863202745e-06, "loss": 0.5603, "step": 25115 }, { "epoch": 1.7, "learning_rate": 6.375325213981374e-06, "loss": 0.4965, "step": 25120 }, { "epoch": 1.71, "learning_rate": 6.374382564760002e-06, "loss": 0.7139, "step": 25125 }, { "epoch": 1.71, "learning_rate": 6.373439915538631e-06, "loss": 0.5483, "step": 25130 }, { "epoch": 1.71, "learning_rate": 6.372497266317259e-06, "loss": 0.952, "step": 25135 }, { "epoch": 1.71, "learning_rate": 6.371554617095887e-06, "loss": 0.6531, "step": 25140 }, { "epoch": 1.71, "learning_rate": 6.3706119678745144e-06, "loss": 0.5981, "step": 25145 }, { "epoch": 1.71, "learning_rate": 6.369669318653143e-06, "loss": 1.0031, "step": 25150 }, { "epoch": 1.71, "learning_rate": 6.368726669431771e-06, "loss": 0.2228, "step": 25155 }, { "epoch": 1.71, "learning_rate": 6.3677840202104e-06, "loss": 0.1992, "step": 25160 }, { "epoch": 1.71, "learning_rate": 6.3668413709890275e-06, "loss": 0.5676, "step": 25165 }, { "epoch": 1.71, "learning_rate": 6.365898721767656e-06, "loss": 0.2699, "step": 25170 }, { "epoch": 1.71, "learning_rate": 6.3649560725462845e-06, "loss": 0.2297, "step": 25175 }, { "epoch": 1.71, "learning_rate": 6.364013423324913e-06, "loss": 0.9433, "step": 25180 }, { "epoch": 1.71, "learning_rate": 6.3630707741035415e-06, "loss": 0.3135, "step": 25185 }, { "epoch": 1.71, "learning_rate": 6.36212812488217e-06, "loss": 0.3846, "step": 25190 }, { "epoch": 1.71, "learning_rate": 6.361185475660797e-06, "loss": 0.3128, "step": 25195 }, { "epoch": 1.71, "learning_rate": 6.360242826439425e-06, "loss": 0.6986, "step": 25200 }, { "epoch": 1.71, "learning_rate": 6.359300177218054e-06, "loss": 0.2539, "step": 25205 }, { "epoch": 1.71, "learning_rate": 6.358357527996682e-06, "loss": 0.3253, "step": 25210 }, { "epoch": 1.71, "learning_rate": 6.357414878775311e-06, "loss": 0.3036, "step": 25215 }, { "epoch": 1.71, "learning_rate": 6.356472229553939e-06, "loss": 0.6668, "step": 25220 }, { "epoch": 1.71, "learning_rate": 6.355529580332567e-06, "loss": 0.5166, "step": 25225 }, { "epoch": 1.71, "learning_rate": 6.354586931111195e-06, "loss": 0.5924, "step": 25230 }, { "epoch": 1.71, "learning_rate": 6.353644281889824e-06, "loss": 0.2745, "step": 25235 }, { "epoch": 1.71, "learning_rate": 6.352701632668452e-06, "loss": 0.5367, "step": 25240 }, { "epoch": 1.71, "learning_rate": 6.351758983447081e-06, "loss": 0.3291, "step": 25245 }, { "epoch": 1.71, "learning_rate": 6.350816334225709e-06, "loss": 0.3563, "step": 25250 }, { "epoch": 1.71, "learning_rate": 6.349873685004336e-06, "loss": 0.6286, "step": 25255 }, { "epoch": 1.71, "learning_rate": 6.3489310357829646e-06, "loss": 0.287, "step": 25260 }, { "epoch": 1.71, "learning_rate": 6.347988386561593e-06, "loss": 0.1309, "step": 25265 }, { "epoch": 1.72, "learning_rate": 6.3470457373402215e-06, "loss": 0.646, "step": 25270 }, { "epoch": 1.72, "learning_rate": 6.34610308811885e-06, "loss": 0.3788, "step": 25275 }, { "epoch": 1.72, "learning_rate": 6.345160438897478e-06, "loss": 1.2442, "step": 25280 }, { "epoch": 1.72, "learning_rate": 6.344217789676106e-06, "loss": 0.5989, "step": 25285 }, { "epoch": 1.72, "learning_rate": 6.343275140454735e-06, "loss": 0.2181, "step": 25290 }, { "epoch": 1.72, "learning_rate": 6.342332491233363e-06, "loss": 0.8704, "step": 25295 }, { "epoch": 1.72, "learning_rate": 6.341389842011992e-06, "loss": 0.6657, "step": 25300 }, { "epoch": 1.72, "learning_rate": 6.34044719279062e-06, "loss": 0.3144, "step": 25305 }, { "epoch": 1.72, "learning_rate": 6.339504543569247e-06, "loss": 0.5882, "step": 25310 }, { "epoch": 1.72, "learning_rate": 6.338561894347875e-06, "loss": 0.3919, "step": 25315 }, { "epoch": 1.72, "learning_rate": 6.337619245126504e-06, "loss": 0.2289, "step": 25320 }, { "epoch": 1.72, "learning_rate": 6.336676595905132e-06, "loss": 0.5663, "step": 25325 }, { "epoch": 1.72, "learning_rate": 6.335733946683761e-06, "loss": 0.4103, "step": 25330 }, { "epoch": 1.72, "learning_rate": 6.334791297462389e-06, "loss": 0.7127, "step": 25335 }, { "epoch": 1.72, "learning_rate": 6.333848648241017e-06, "loss": 0.698, "step": 25340 }, { "epoch": 1.72, "learning_rate": 6.3329059990196455e-06, "loss": 0.6498, "step": 25345 }, { "epoch": 1.72, "learning_rate": 6.331963349798274e-06, "loss": 0.1634, "step": 25350 }, { "epoch": 1.72, "learning_rate": 6.331020700576902e-06, "loss": 0.3236, "step": 25355 }, { "epoch": 1.72, "learning_rate": 6.33007805135553e-06, "loss": 0.3747, "step": 25360 }, { "epoch": 1.72, "learning_rate": 6.3291354021341586e-06, "loss": 0.4671, "step": 25365 }, { "epoch": 1.72, "learning_rate": 6.328192752912786e-06, "loss": 0.5257, "step": 25370 }, { "epoch": 1.72, "learning_rate": 6.327250103691415e-06, "loss": 0.22, "step": 25375 }, { "epoch": 1.72, "learning_rate": 6.326307454470043e-06, "loss": 0.1781, "step": 25380 }, { "epoch": 1.72, "learning_rate": 6.325364805248672e-06, "loss": 0.2601, "step": 25385 }, { "epoch": 1.72, "learning_rate": 6.3244221560273e-06, "loss": 0.6262, "step": 25390 }, { "epoch": 1.72, "learning_rate": 6.323479506805927e-06, "loss": 0.3696, "step": 25395 }, { "epoch": 1.72, "learning_rate": 6.3225368575845554e-06, "loss": 0.2229, "step": 25400 }, { "epoch": 1.72, "learning_rate": 6.321594208363184e-06, "loss": 0.6602, "step": 25405 }, { "epoch": 1.72, "learning_rate": 6.320651559141812e-06, "loss": 0.1363, "step": 25410 }, { "epoch": 1.72, "learning_rate": 6.319708909920441e-06, "loss": 0.5005, "step": 25415 }, { "epoch": 1.73, "learning_rate": 6.318766260699069e-06, "loss": 0.3168, "step": 25420 }, { "epoch": 1.73, "learning_rate": 6.317823611477697e-06, "loss": 0.1618, "step": 25425 }, { "epoch": 1.73, "learning_rate": 6.3168809622563255e-06, "loss": 0.1246, "step": 25430 }, { "epoch": 1.73, "learning_rate": 6.315938313034954e-06, "loss": 0.6992, "step": 25435 }, { "epoch": 1.73, "learning_rate": 6.3149956638135825e-06, "loss": 1.0764, "step": 25440 }, { "epoch": 1.73, "learning_rate": 6.314053014592211e-06, "loss": 0.5081, "step": 25445 }, { "epoch": 1.73, "learning_rate": 6.3131103653708395e-06, "loss": 0.3117, "step": 25450 }, { "epoch": 1.73, "learning_rate": 6.312167716149466e-06, "loss": 0.3167, "step": 25455 }, { "epoch": 1.73, "learning_rate": 6.311225066928095e-06, "loss": 0.6863, "step": 25460 }, { "epoch": 1.73, "learning_rate": 6.310282417706723e-06, "loss": 0.6994, "step": 25465 }, { "epoch": 1.73, "learning_rate": 6.309339768485352e-06, "loss": 0.5411, "step": 25470 }, { "epoch": 1.73, "learning_rate": 6.30839711926398e-06, "loss": 0.8766, "step": 25475 }, { "epoch": 1.73, "learning_rate": 6.307454470042609e-06, "loss": 0.4403, "step": 25480 }, { "epoch": 1.73, "learning_rate": 6.306511820821236e-06, "loss": 0.4244, "step": 25485 }, { "epoch": 1.73, "learning_rate": 6.305569171599865e-06, "loss": 0.674, "step": 25490 }, { "epoch": 1.73, "learning_rate": 6.304626522378493e-06, "loss": 0.6874, "step": 25495 }, { "epoch": 1.73, "learning_rate": 6.303683873157122e-06, "loss": 0.3747, "step": 25500 }, { "epoch": 1.73, "learning_rate": 6.30274122393575e-06, "loss": 0.491, "step": 25505 }, { "epoch": 1.73, "learning_rate": 6.301798574714377e-06, "loss": 0.3255, "step": 25510 }, { "epoch": 1.73, "learning_rate": 6.3008559254930056e-06, "loss": 0.6639, "step": 25515 }, { "epoch": 1.73, "learning_rate": 6.299913276271634e-06, "loss": 0.1321, "step": 25520 }, { "epoch": 1.73, "learning_rate": 6.2989706270502625e-06, "loss": 0.1423, "step": 25525 }, { "epoch": 1.73, "learning_rate": 6.298027977828891e-06, "loss": 0.2444, "step": 25530 }, { "epoch": 1.73, "learning_rate": 6.2970853286075195e-06, "loss": 0.6951, "step": 25535 }, { "epoch": 1.73, "learning_rate": 6.296142679386147e-06, "loss": 0.6313, "step": 25540 }, { "epoch": 1.73, "learning_rate": 6.295200030164776e-06, "loss": 0.5039, "step": 25545 }, { "epoch": 1.73, "learning_rate": 6.294257380943404e-06, "loss": 0.1048, "step": 25550 }, { "epoch": 1.73, "learning_rate": 6.293314731722033e-06, "loss": 0.2642, "step": 25555 }, { "epoch": 1.73, "learning_rate": 6.29237208250066e-06, "loss": 0.4799, "step": 25560 }, { "epoch": 1.74, "learning_rate": 6.291429433279289e-06, "loss": 0.3693, "step": 25565 }, { "epoch": 1.74, "learning_rate": 6.290486784057916e-06, "loss": 0.3511, "step": 25570 }, { "epoch": 1.74, "learning_rate": 6.289544134836545e-06, "loss": 0.6486, "step": 25575 }, { "epoch": 1.74, "learning_rate": 6.288601485615173e-06, "loss": 0.1131, "step": 25580 }, { "epoch": 1.74, "learning_rate": 6.287658836393802e-06, "loss": 0.3826, "step": 25585 }, { "epoch": 1.74, "learning_rate": 6.28671618717243e-06, "loss": 0.2161, "step": 25590 }, { "epoch": 1.74, "learning_rate": 6.285773537951059e-06, "loss": 0.2468, "step": 25595 }, { "epoch": 1.74, "learning_rate": 6.2848308887296865e-06, "loss": 0.2504, "step": 25600 }, { "epoch": 1.74, "learning_rate": 6.283888239508314e-06, "loss": 1.0933, "step": 25605 }, { "epoch": 1.74, "learning_rate": 6.282945590286943e-06, "loss": 0.395, "step": 25610 }, { "epoch": 1.74, "learning_rate": 6.282002941065571e-06, "loss": 0.2959, "step": 25615 }, { "epoch": 1.74, "learning_rate": 6.2810602918441996e-06, "loss": 0.9286, "step": 25620 }, { "epoch": 1.74, "learning_rate": 6.280117642622827e-06, "loss": 0.7068, "step": 25625 }, { "epoch": 1.74, "learning_rate": 6.279174993401456e-06, "loss": 0.1888, "step": 25630 }, { "epoch": 1.74, "learning_rate": 6.278232344180084e-06, "loss": 0.6191, "step": 25635 }, { "epoch": 1.74, "learning_rate": 6.277289694958713e-06, "loss": 0.1661, "step": 25640 }, { "epoch": 1.74, "learning_rate": 6.276347045737341e-06, "loss": 0.5276, "step": 25645 }, { "epoch": 1.74, "learning_rate": 6.27540439651597e-06, "loss": 0.4259, "step": 25650 }, { "epoch": 1.74, "learning_rate": 6.2744617472945964e-06, "loss": 0.2628, "step": 25655 }, { "epoch": 1.74, "learning_rate": 6.273519098073225e-06, "loss": 0.3323, "step": 25660 }, { "epoch": 1.74, "learning_rate": 6.272576448851853e-06, "loss": 0.596, "step": 25665 }, { "epoch": 1.74, "learning_rate": 6.271633799630482e-06, "loss": 0.4461, "step": 25670 }, { "epoch": 1.74, "learning_rate": 6.27069115040911e-06, "loss": 0.3919, "step": 25675 }, { "epoch": 1.74, "learning_rate": 6.269748501187739e-06, "loss": 0.9775, "step": 25680 }, { "epoch": 1.74, "learning_rate": 6.2688058519663665e-06, "loss": 0.4884, "step": 25685 }, { "epoch": 1.74, "learning_rate": 6.267863202744995e-06, "loss": 0.5235, "step": 25690 }, { "epoch": 1.74, "learning_rate": 6.2669205535236235e-06, "loss": 0.1928, "step": 25695 }, { "epoch": 1.74, "learning_rate": 6.265977904302252e-06, "loss": 0.525, "step": 25700 }, { "epoch": 1.74, "learning_rate": 6.2650352550808805e-06, "loss": 0.2291, "step": 25705 }, { "epoch": 1.74, "learning_rate": 6.264092605859509e-06, "loss": 0.7739, "step": 25710 }, { "epoch": 1.75, "learning_rate": 6.263149956638136e-06, "loss": 0.454, "step": 25715 }, { "epoch": 1.75, "learning_rate": 6.262207307416764e-06, "loss": 0.2858, "step": 25720 }, { "epoch": 1.75, "learning_rate": 6.261264658195393e-06, "loss": 0.094, "step": 25725 }, { "epoch": 1.75, "learning_rate": 6.260322008974021e-06, "loss": 0.4589, "step": 25730 }, { "epoch": 1.75, "learning_rate": 6.25937935975265e-06, "loss": 0.2639, "step": 25735 }, { "epoch": 1.75, "learning_rate": 6.258436710531277e-06, "loss": 0.4144, "step": 25740 }, { "epoch": 1.75, "learning_rate": 6.257494061309906e-06, "loss": 0.4672, "step": 25745 }, { "epoch": 1.75, "learning_rate": 6.256551412088534e-06, "loss": 0.3279, "step": 25750 }, { "epoch": 1.75, "learning_rate": 6.255608762867163e-06, "loss": 0.3185, "step": 25755 }, { "epoch": 1.75, "learning_rate": 6.254666113645791e-06, "loss": 0.9438, "step": 25760 }, { "epoch": 1.75, "learning_rate": 6.25372346442442e-06, "loss": 0.2441, "step": 25765 }, { "epoch": 1.75, "learning_rate": 6.2527808152030466e-06, "loss": 0.4233, "step": 25770 }, { "epoch": 1.75, "learning_rate": 6.251838165981675e-06, "loss": 0.8309, "step": 25775 }, { "epoch": 1.75, "learning_rate": 6.2508955167603035e-06, "loss": 0.399, "step": 25780 }, { "epoch": 1.75, "learning_rate": 6.249952867538932e-06, "loss": 0.3858, "step": 25785 }, { "epoch": 1.75, "learning_rate": 6.2490102183175605e-06, "loss": 0.6716, "step": 25790 }, { "epoch": 1.75, "learning_rate": 6.248067569096189e-06, "loss": 0.4751, "step": 25795 }, { "epoch": 1.75, "learning_rate": 6.247124919874817e-06, "loss": 0.2708, "step": 25800 }, { "epoch": 1.75, "learning_rate": 6.246182270653445e-06, "loss": 0.2509, "step": 25805 }, { "epoch": 1.75, "learning_rate": 6.245239621432073e-06, "loss": 0.1318, "step": 25810 }, { "epoch": 1.75, "learning_rate": 6.244296972210701e-06, "loss": 0.347, "step": 25815 }, { "epoch": 1.75, "learning_rate": 6.24335432298933e-06, "loss": 0.3365, "step": 25820 }, { "epoch": 1.75, "learning_rate": 6.242411673767957e-06, "loss": 0.3046, "step": 25825 }, { "epoch": 1.75, "learning_rate": 6.241469024546586e-06, "loss": 0.8504, "step": 25830 }, { "epoch": 1.75, "learning_rate": 6.240526375325214e-06, "loss": 0.4137, "step": 25835 }, { "epoch": 1.75, "learning_rate": 6.239583726103843e-06, "loss": 0.3646, "step": 25840 }, { "epoch": 1.75, "learning_rate": 6.238641076882471e-06, "loss": 0.8007, "step": 25845 }, { "epoch": 1.75, "learning_rate": 6.2376984276611e-06, "loss": 0.4474, "step": 25850 }, { "epoch": 1.75, "learning_rate": 6.236755778439727e-06, "loss": 0.6261, "step": 25855 }, { "epoch": 1.76, "learning_rate": 6.235813129218355e-06, "loss": 0.5901, "step": 25860 }, { "epoch": 1.76, "learning_rate": 6.234870479996984e-06, "loss": 0.7931, "step": 25865 }, { "epoch": 1.76, "learning_rate": 6.233927830775612e-06, "loss": 0.436, "step": 25870 }, { "epoch": 1.76, "learning_rate": 6.2329851815542406e-06, "loss": 0.5776, "step": 25875 }, { "epoch": 1.76, "learning_rate": 6.232042532332869e-06, "loss": 0.8768, "step": 25880 }, { "epoch": 1.76, "learning_rate": 6.231099883111497e-06, "loss": 0.4385, "step": 25885 }, { "epoch": 1.76, "learning_rate": 6.230157233890125e-06, "loss": 0.2725, "step": 25890 }, { "epoch": 1.76, "learning_rate": 6.229214584668754e-06, "loss": 0.3061, "step": 25895 }, { "epoch": 1.76, "learning_rate": 6.228271935447382e-06, "loss": 0.2488, "step": 25900 }, { "epoch": 1.76, "learning_rate": 6.227329286226011e-06, "loss": 0.1854, "step": 25905 }, { "epoch": 1.76, "learning_rate": 6.226386637004639e-06, "loss": 0.4802, "step": 25910 }, { "epoch": 1.76, "learning_rate": 6.225443987783266e-06, "loss": 0.5253, "step": 25915 }, { "epoch": 1.76, "learning_rate": 6.224501338561894e-06, "loss": 0.2515, "step": 25920 }, { "epoch": 1.76, "learning_rate": 6.223558689340523e-06, "loss": 0.6632, "step": 25925 }, { "epoch": 1.76, "learning_rate": 6.222616040119151e-06, "loss": 0.3639, "step": 25930 }, { "epoch": 1.76, "learning_rate": 6.22167339089778e-06, "loss": 0.492, "step": 25935 }, { "epoch": 1.76, "learning_rate": 6.2207307416764075e-06, "loss": 0.5521, "step": 25940 }, { "epoch": 1.76, "learning_rate": 6.219788092455036e-06, "loss": 0.6307, "step": 25945 }, { "epoch": 1.76, "learning_rate": 6.2188454432336645e-06, "loss": 0.5174, "step": 25950 }, { "epoch": 1.76, "learning_rate": 6.217902794012293e-06, "loss": 0.3294, "step": 25955 }, { "epoch": 1.76, "learning_rate": 6.2169601447909215e-06, "loss": 0.5963, "step": 25960 }, { "epoch": 1.76, "learning_rate": 6.21601749556955e-06, "loss": 0.3734, "step": 25965 }, { "epoch": 1.76, "learning_rate": 6.215074846348177e-06, "loss": 0.5345, "step": 25970 }, { "epoch": 1.76, "learning_rate": 6.214132197126805e-06, "loss": 0.3043, "step": 25975 }, { "epoch": 1.76, "learning_rate": 6.213189547905434e-06, "loss": 0.4789, "step": 25980 }, { "epoch": 1.76, "learning_rate": 6.212246898684062e-06, "loss": 0.5223, "step": 25985 }, { "epoch": 1.76, "learning_rate": 6.211304249462691e-06, "loss": 0.2457, "step": 25990 }, { "epoch": 1.76, "learning_rate": 6.210361600241319e-06, "loss": 0.9208, "step": 25995 }, { "epoch": 1.76, "learning_rate": 6.209418951019947e-06, "loss": 0.4516, "step": 26000 }, { "epoch": 1.76, "learning_rate": 6.208476301798575e-06, "loss": 0.5778, "step": 26005 }, { "epoch": 1.77, "learning_rate": 6.207533652577204e-06, "loss": 0.5663, "step": 26010 }, { "epoch": 1.77, "learning_rate": 6.206591003355832e-06, "loss": 0.2863, "step": 26015 }, { "epoch": 1.77, "learning_rate": 6.20564835413446e-06, "loss": 0.6751, "step": 26020 }, { "epoch": 1.77, "learning_rate": 6.204705704913088e-06, "loss": 0.152, "step": 26025 }, { "epoch": 1.77, "learning_rate": 6.203763055691716e-06, "loss": 0.2361, "step": 26030 }, { "epoch": 1.77, "learning_rate": 6.2028204064703445e-06, "loss": 0.5507, "step": 26035 }, { "epoch": 1.77, "learning_rate": 6.201877757248973e-06, "loss": 1.24, "step": 26040 }, { "epoch": 1.77, "learning_rate": 6.2009351080276015e-06, "loss": 0.2377, "step": 26045 }, { "epoch": 1.77, "learning_rate": 6.19999245880623e-06, "loss": 0.2987, "step": 26050 }, { "epoch": 1.77, "learning_rate": 6.199049809584858e-06, "loss": 0.5267, "step": 26055 }, { "epoch": 1.77, "learning_rate": 6.198107160363486e-06, "loss": 0.3473, "step": 26060 }, { "epoch": 1.77, "learning_rate": 6.197164511142114e-06, "loss": 0.4092, "step": 26065 }, { "epoch": 1.77, "learning_rate": 6.196221861920742e-06, "loss": 1.0141, "step": 26070 }, { "epoch": 1.77, "learning_rate": 6.195279212699371e-06, "loss": 0.4683, "step": 26075 }, { "epoch": 1.77, "learning_rate": 6.194336563477999e-06, "loss": 0.2254, "step": 26080 }, { "epoch": 1.77, "learning_rate": 6.193393914256627e-06, "loss": 0.2808, "step": 26085 }, { "epoch": 1.77, "learning_rate": 6.192451265035255e-06, "loss": 0.0832, "step": 26090 }, { "epoch": 1.77, "learning_rate": 6.191508615813884e-06, "loss": 0.3611, "step": 26095 }, { "epoch": 1.77, "learning_rate": 6.190565966592512e-06, "loss": 0.7727, "step": 26100 }, { "epoch": 1.77, "learning_rate": 6.189623317371141e-06, "loss": 0.2429, "step": 26105 }, { "epoch": 1.77, "learning_rate": 6.188680668149769e-06, "loss": 0.4916, "step": 26110 }, { "epoch": 1.77, "learning_rate": 6.187738018928396e-06, "loss": 0.8408, "step": 26115 }, { "epoch": 1.77, "learning_rate": 6.1867953697070246e-06, "loss": 0.12, "step": 26120 }, { "epoch": 1.77, "learning_rate": 6.185852720485653e-06, "loss": 0.3849, "step": 26125 }, { "epoch": 1.77, "learning_rate": 6.1849100712642816e-06, "loss": 0.5376, "step": 26130 }, { "epoch": 1.77, "learning_rate": 6.18396742204291e-06, "loss": 0.2479, "step": 26135 }, { "epoch": 1.77, "learning_rate": 6.1830247728215385e-06, "loss": 0.3274, "step": 26140 }, { "epoch": 1.77, "learning_rate": 6.182082123600166e-06, "loss": 0.1391, "step": 26145 }, { "epoch": 1.77, "learning_rate": 6.181139474378795e-06, "loss": 0.7958, "step": 26150 }, { "epoch": 1.78, "learning_rate": 6.180196825157423e-06, "loss": 0.4878, "step": 26155 }, { "epoch": 1.78, "learning_rate": 6.179254175936052e-06, "loss": 0.17, "step": 26160 }, { "epoch": 1.78, "learning_rate": 6.17831152671468e-06, "loss": 0.3456, "step": 26165 }, { "epoch": 1.78, "learning_rate": 6.177368877493307e-06, "loss": 0.8288, "step": 26170 }, { "epoch": 1.78, "learning_rate": 6.176426228271935e-06, "loss": 0.7553, "step": 26175 }, { "epoch": 1.78, "learning_rate": 6.175483579050564e-06, "loss": 0.87, "step": 26180 }, { "epoch": 1.78, "learning_rate": 6.174540929829192e-06, "loss": 0.4969, "step": 26185 }, { "epoch": 1.78, "learning_rate": 6.173598280607821e-06, "loss": 0.0662, "step": 26190 }, { "epoch": 1.78, "learning_rate": 6.172655631386449e-06, "loss": 0.3443, "step": 26195 }, { "epoch": 1.78, "learning_rate": 6.171712982165077e-06, "loss": 0.3504, "step": 26200 }, { "epoch": 1.78, "learning_rate": 6.1707703329437055e-06, "loss": 0.2032, "step": 26205 }, { "epoch": 1.78, "learning_rate": 6.169827683722334e-06, "loss": 0.8166, "step": 26210 }, { "epoch": 1.78, "learning_rate": 6.1688850345009624e-06, "loss": 0.5476, "step": 26215 }, { "epoch": 1.78, "learning_rate": 6.167942385279591e-06, "loss": 0.5893, "step": 26220 }, { "epoch": 1.78, "learning_rate": 6.166999736058219e-06, "loss": 0.1051, "step": 26225 }, { "epoch": 1.78, "learning_rate": 6.166057086836846e-06, "loss": 0.435, "step": 26230 }, { "epoch": 1.78, "learning_rate": 6.165114437615475e-06, "loss": 0.2612, "step": 26235 }, { "epoch": 1.78, "learning_rate": 6.164171788394103e-06, "loss": 0.6213, "step": 26240 }, { "epoch": 1.78, "learning_rate": 6.163229139172732e-06, "loss": 0.3818, "step": 26245 }, { "epoch": 1.78, "learning_rate": 6.16228648995136e-06, "loss": 0.4777, "step": 26250 }, { "epoch": 1.78, "learning_rate": 6.161343840729989e-06, "loss": 0.6639, "step": 26255 }, { "epoch": 1.78, "learning_rate": 6.160401191508616e-06, "loss": 0.3077, "step": 26260 }, { "epoch": 1.78, "learning_rate": 6.159458542287245e-06, "loss": 0.3957, "step": 26265 }, { "epoch": 1.78, "learning_rate": 6.158515893065872e-06, "loss": 0.1483, "step": 26270 }, { "epoch": 1.78, "learning_rate": 6.157573243844501e-06, "loss": 0.6477, "step": 26275 }, { "epoch": 1.78, "learning_rate": 6.156630594623129e-06, "loss": 0.3201, "step": 26280 }, { "epoch": 1.78, "learning_rate": 6.155687945401757e-06, "loss": 0.6412, "step": 26285 }, { "epoch": 1.78, "learning_rate": 6.1547452961803855e-06, "loss": 0.2917, "step": 26290 }, { "epoch": 1.78, "learning_rate": 6.153802646959014e-06, "loss": 0.1844, "step": 26295 }, { "epoch": 1.78, "learning_rate": 6.1528599977376425e-06, "loss": 0.4135, "step": 26300 }, { "epoch": 1.79, "learning_rate": 6.151917348516271e-06, "loss": 0.7452, "step": 26305 }, { "epoch": 1.79, "learning_rate": 6.1509746992948995e-06, "loss": 0.3234, "step": 26310 }, { "epoch": 1.79, "learning_rate": 6.150032050073526e-06, "loss": 0.191, "step": 26315 }, { "epoch": 1.79, "learning_rate": 6.149089400852155e-06, "loss": 0.2234, "step": 26320 }, { "epoch": 1.79, "learning_rate": 6.148146751630783e-06, "loss": 0.4684, "step": 26325 }, { "epoch": 1.79, "learning_rate": 6.147204102409412e-06, "loss": 0.9339, "step": 26330 }, { "epoch": 1.79, "learning_rate": 6.14626145318804e-06, "loss": 0.3662, "step": 26335 }, { "epoch": 1.79, "learning_rate": 6.145318803966669e-06, "loss": 0.3206, "step": 26340 }, { "epoch": 1.79, "learning_rate": 6.144376154745296e-06, "loss": 0.5563, "step": 26345 }, { "epoch": 1.79, "learning_rate": 6.143433505523925e-06, "loss": 0.1732, "step": 26350 }, { "epoch": 1.79, "learning_rate": 6.142490856302553e-06, "loss": 0.4255, "step": 26355 }, { "epoch": 1.79, "learning_rate": 6.141548207081182e-06, "loss": 0.6897, "step": 26360 }, { "epoch": 1.79, "learning_rate": 6.14060555785981e-06, "loss": 0.4857, "step": 26365 }, { "epoch": 1.79, "learning_rate": 6.139662908638439e-06, "loss": 0.5542, "step": 26370 }, { "epoch": 1.79, "learning_rate": 6.1387202594170656e-06, "loss": 0.3979, "step": 26375 }, { "epoch": 1.79, "learning_rate": 6.137777610195694e-06, "loss": 0.1665, "step": 26380 }, { "epoch": 1.79, "learning_rate": 6.1368349609743225e-06, "loss": 0.8826, "step": 26385 }, { "epoch": 1.79, "learning_rate": 6.135892311752951e-06, "loss": 0.1985, "step": 26390 }, { "epoch": 1.79, "learning_rate": 6.1349496625315795e-06, "loss": 0.7572, "step": 26395 }, { "epoch": 1.79, "learning_rate": 6.134007013310207e-06, "loss": 0.3189, "step": 26400 }, { "epoch": 1.79, "learning_rate": 6.133064364088836e-06, "loss": 0.4131, "step": 26405 }, { "epoch": 1.79, "learning_rate": 6.132121714867464e-06, "loss": 0.2524, "step": 26410 }, { "epoch": 1.79, "learning_rate": 6.131179065646093e-06, "loss": 0.2577, "step": 26415 }, { "epoch": 1.79, "learning_rate": 6.130236416424721e-06, "loss": 0.8614, "step": 26420 }, { "epoch": 1.79, "learning_rate": 6.12929376720335e-06, "loss": 0.5322, "step": 26425 }, { "epoch": 1.79, "learning_rate": 6.128351117981976e-06, "loss": 0.4863, "step": 26430 }, { "epoch": 1.79, "learning_rate": 6.127408468760605e-06, "loss": 0.463, "step": 26435 }, { "epoch": 1.79, "learning_rate": 6.126465819539233e-06, "loss": 0.569, "step": 26440 }, { "epoch": 1.79, "learning_rate": 6.125523170317862e-06, "loss": 0.6353, "step": 26445 }, { "epoch": 1.8, "learning_rate": 6.12458052109649e-06, "loss": 0.4114, "step": 26450 }, { "epoch": 1.8, "learning_rate": 6.123637871875119e-06, "loss": 0.91, "step": 26455 }, { "epoch": 1.8, "learning_rate": 6.1226952226537465e-06, "loss": 0.7557, "step": 26460 }, { "epoch": 1.8, "learning_rate": 6.121752573432375e-06, "loss": 0.9923, "step": 26465 }, { "epoch": 1.8, "learning_rate": 6.1208099242110034e-06, "loss": 0.4592, "step": 26470 }, { "epoch": 1.8, "learning_rate": 6.119867274989632e-06, "loss": 0.3864, "step": 26475 }, { "epoch": 1.8, "learning_rate": 6.1189246257682596e-06, "loss": 0.433, "step": 26480 }, { "epoch": 1.8, "learning_rate": 6.117981976546888e-06, "loss": 0.2102, "step": 26485 }, { "epoch": 1.8, "learning_rate": 6.117039327325516e-06, "loss": 1.0859, "step": 26490 }, { "epoch": 1.8, "learning_rate": 6.116096678104144e-06, "loss": 0.4988, "step": 26495 }, { "epoch": 1.8, "learning_rate": 6.115154028882773e-06, "loss": 0.5053, "step": 26500 }, { "epoch": 1.8, "learning_rate": 6.114211379661401e-06, "loss": 0.2825, "step": 26505 }, { "epoch": 1.8, "learning_rate": 6.11326873044003e-06, "loss": 0.3766, "step": 26510 }, { "epoch": 1.8, "learning_rate": 6.112326081218657e-06, "loss": 0.302, "step": 26515 }, { "epoch": 1.8, "learning_rate": 6.111383431997285e-06, "loss": 0.2511, "step": 26520 }, { "epoch": 1.8, "learning_rate": 6.110440782775913e-06, "loss": 0.1809, "step": 26525 }, { "epoch": 1.8, "learning_rate": 6.109498133554542e-06, "loss": 0.3511, "step": 26530 }, { "epoch": 1.8, "learning_rate": 6.10855548433317e-06, "loss": 0.4644, "step": 26535 }, { "epoch": 1.8, "learning_rate": 6.107612835111799e-06, "loss": 0.5643, "step": 26540 }, { "epoch": 1.8, "learning_rate": 6.1066701858904265e-06, "loss": 0.4153, "step": 26545 }, { "epoch": 1.8, "learning_rate": 6.105727536669055e-06, "loss": 0.3846, "step": 26550 }, { "epoch": 1.8, "learning_rate": 6.1047848874476835e-06, "loss": 0.9776, "step": 26555 }, { "epoch": 1.8, "learning_rate": 6.103842238226312e-06, "loss": 0.3038, "step": 26560 }, { "epoch": 1.8, "learning_rate": 6.1028995890049405e-06, "loss": 0.9514, "step": 26565 }, { "epoch": 1.8, "learning_rate": 6.101956939783569e-06, "loss": 0.2417, "step": 26570 }, { "epoch": 1.8, "learning_rate": 6.101014290562196e-06, "loss": 0.4299, "step": 26575 }, { "epoch": 1.8, "learning_rate": 6.100071641340824e-06, "loss": 0.767, "step": 26580 }, { "epoch": 1.8, "learning_rate": 6.099128992119453e-06, "loss": 1.0964, "step": 26585 }, { "epoch": 1.8, "learning_rate": 6.098186342898081e-06, "loss": 0.0988, "step": 26590 }, { "epoch": 1.81, "learning_rate": 6.09724369367671e-06, "loss": 0.8638, "step": 26595 }, { "epoch": 1.81, "learning_rate": 6.096301044455338e-06, "loss": 0.8039, "step": 26600 }, { "epoch": 1.81, "learning_rate": 6.095358395233966e-06, "loss": 0.2626, "step": 26605 }, { "epoch": 1.81, "learning_rate": 6.094415746012594e-06, "loss": 0.2726, "step": 26610 }, { "epoch": 1.81, "learning_rate": 6.093473096791223e-06, "loss": 0.3964, "step": 26615 }, { "epoch": 1.81, "learning_rate": 6.092530447569851e-06, "loss": 0.4488, "step": 26620 }, { "epoch": 1.81, "learning_rate": 6.09158779834848e-06, "loss": 0.648, "step": 26625 }, { "epoch": 1.81, "learning_rate": 6.0906451491271066e-06, "loss": 0.7935, "step": 26630 }, { "epoch": 1.81, "learning_rate": 6.089702499905735e-06, "loss": 0.5263, "step": 26635 }, { "epoch": 1.81, "learning_rate": 6.0887598506843635e-06, "loss": 0.6317, "step": 26640 }, { "epoch": 1.81, "learning_rate": 6.087817201462992e-06, "loss": 0.0898, "step": 26645 }, { "epoch": 1.81, "learning_rate": 6.0868745522416205e-06, "loss": 0.2262, "step": 26650 }, { "epoch": 1.81, "learning_rate": 6.085931903020249e-06, "loss": 0.135, "step": 26655 }, { "epoch": 1.81, "learning_rate": 6.084989253798877e-06, "loss": 0.3661, "step": 26660 }, { "epoch": 1.81, "learning_rate": 6.084046604577505e-06, "loss": 0.2746, "step": 26665 }, { "epoch": 1.81, "learning_rate": 6.083103955356134e-06, "loss": 0.6333, "step": 26670 }, { "epoch": 1.81, "learning_rate": 6.082161306134762e-06, "loss": 0.7988, "step": 26675 }, { "epoch": 1.81, "learning_rate": 6.081218656913391e-06, "loss": 0.6328, "step": 26680 }, { "epoch": 1.81, "learning_rate": 6.080276007692018e-06, "loss": 0.6003, "step": 26685 }, { "epoch": 1.81, "learning_rate": 6.079333358470646e-06, "loss": 1.0098, "step": 26690 }, { "epoch": 1.81, "learning_rate": 6.078390709249274e-06, "loss": 0.5481, "step": 26695 }, { "epoch": 1.81, "learning_rate": 6.077448060027903e-06, "loss": 0.3674, "step": 26700 }, { "epoch": 1.81, "learning_rate": 6.076505410806531e-06, "loss": 0.6124, "step": 26705 }, { "epoch": 1.81, "learning_rate": 6.07556276158516e-06, "loss": 0.725, "step": 26710 }, { "epoch": 1.81, "learning_rate": 6.074620112363788e-06, "loss": 0.5183, "step": 26715 }, { "epoch": 1.81, "learning_rate": 6.073677463142416e-06, "loss": 0.5239, "step": 26720 }, { "epoch": 1.81, "learning_rate": 6.0727348139210444e-06, "loss": 0.5148, "step": 26725 }, { "epoch": 1.81, "learning_rate": 6.071792164699672e-06, "loss": 0.224, "step": 26730 }, { "epoch": 1.81, "learning_rate": 6.0708495154783006e-06, "loss": 0.5937, "step": 26735 }, { "epoch": 1.81, "learning_rate": 6.069906866256929e-06, "loss": 0.1864, "step": 26740 }, { "epoch": 1.82, "learning_rate": 6.068964217035557e-06, "loss": 0.3922, "step": 26745 }, { "epoch": 1.82, "learning_rate": 6.068021567814185e-06, "loss": 0.3829, "step": 26750 }, { "epoch": 1.82, "learning_rate": 6.067078918592814e-06, "loss": 0.6012, "step": 26755 }, { "epoch": 1.82, "learning_rate": 6.066136269371442e-06, "loss": 0.3137, "step": 26760 }, { "epoch": 1.82, "learning_rate": 6.065193620150071e-06, "loss": 0.9905, "step": 26765 }, { "epoch": 1.82, "learning_rate": 6.064250970928699e-06, "loss": 0.72, "step": 26770 }, { "epoch": 1.82, "learning_rate": 6.063308321707326e-06, "loss": 0.5738, "step": 26775 }, { "epoch": 1.82, "learning_rate": 6.062365672485954e-06, "loss": 0.2279, "step": 26780 }, { "epoch": 1.82, "learning_rate": 6.061423023264583e-06, "loss": 0.4951, "step": 26785 }, { "epoch": 1.82, "learning_rate": 6.060480374043211e-06, "loss": 0.6036, "step": 26790 }, { "epoch": 1.82, "learning_rate": 6.05953772482184e-06, "loss": 0.2204, "step": 26795 }, { "epoch": 1.82, "learning_rate": 6.058595075600468e-06, "loss": 0.8334, "step": 26800 }, { "epoch": 1.82, "learning_rate": 6.057652426379096e-06, "loss": 0.1063, "step": 26805 }, { "epoch": 1.82, "learning_rate": 6.0567097771577245e-06, "loss": 0.7729, "step": 26810 }, { "epoch": 1.82, "learning_rate": 6.055767127936353e-06, "loss": 0.6895, "step": 26815 }, { "epoch": 1.82, "learning_rate": 6.0548244787149815e-06, "loss": 0.9117, "step": 26820 }, { "epoch": 1.82, "learning_rate": 6.05388182949361e-06, "loss": 0.1434, "step": 26825 }, { "epoch": 1.82, "learning_rate": 6.052939180272237e-06, "loss": 0.759, "step": 26830 }, { "epoch": 1.82, "learning_rate": 6.051996531050865e-06, "loss": 0.1847, "step": 26835 }, { "epoch": 1.82, "learning_rate": 6.051053881829494e-06, "loss": 0.1169, "step": 26840 }, { "epoch": 1.82, "learning_rate": 6.050111232608122e-06, "loss": 0.5363, "step": 26845 }, { "epoch": 1.82, "learning_rate": 6.049168583386751e-06, "loss": 0.3812, "step": 26850 }, { "epoch": 1.82, "learning_rate": 6.048225934165379e-06, "loss": 0.5209, "step": 26855 }, { "epoch": 1.82, "learning_rate": 6.047283284944007e-06, "loss": 0.3441, "step": 26860 }, { "epoch": 1.82, "learning_rate": 6.046340635722635e-06, "loss": 0.3332, "step": 26865 }, { "epoch": 1.82, "learning_rate": 6.045397986501264e-06, "loss": 0.9159, "step": 26870 }, { "epoch": 1.82, "learning_rate": 6.044455337279892e-06, "loss": 0.4663, "step": 26875 }, { "epoch": 1.82, "learning_rate": 6.043512688058521e-06, "loss": 0.2718, "step": 26880 }, { "epoch": 1.82, "learning_rate": 6.042570038837149e-06, "loss": 0.4625, "step": 26885 }, { "epoch": 1.83, "learning_rate": 6.041627389615776e-06, "loss": 0.554, "step": 26890 }, { "epoch": 1.83, "learning_rate": 6.0406847403944045e-06, "loss": 0.4717, "step": 26895 }, { "epoch": 1.83, "learning_rate": 6.039742091173033e-06, "loss": 0.3419, "step": 26900 }, { "epoch": 1.83, "learning_rate": 6.0387994419516615e-06, "loss": 0.6297, "step": 26905 }, { "epoch": 1.83, "learning_rate": 6.03785679273029e-06, "loss": 0.2588, "step": 26910 }, { "epoch": 1.83, "learning_rate": 6.0369141435089185e-06, "loss": 0.6232, "step": 26915 }, { "epoch": 1.83, "learning_rate": 6.035971494287546e-06, "loss": 0.3064, "step": 26920 }, { "epoch": 1.83, "learning_rate": 6.035028845066175e-06, "loss": 1.1339, "step": 26925 }, { "epoch": 1.83, "learning_rate": 6.034086195844803e-06, "loss": 0.8352, "step": 26930 }, { "epoch": 1.83, "learning_rate": 6.033143546623431e-06, "loss": 0.3606, "step": 26935 }, { "epoch": 1.83, "learning_rate": 6.032200897402059e-06, "loss": 0.8663, "step": 26940 }, { "epoch": 1.83, "learning_rate": 6.031258248180687e-06, "loss": 0.5645, "step": 26945 }, { "epoch": 1.83, "learning_rate": 6.030315598959315e-06, "loss": 0.135, "step": 26950 }, { "epoch": 1.83, "learning_rate": 6.029372949737944e-06, "loss": 0.8502, "step": 26955 }, { "epoch": 1.83, "learning_rate": 6.028430300516572e-06, "loss": 0.7168, "step": 26960 }, { "epoch": 1.83, "learning_rate": 6.027487651295201e-06, "loss": 0.1995, "step": 26965 }, { "epoch": 1.83, "learning_rate": 6.026545002073829e-06, "loss": 0.4244, "step": 26970 }, { "epoch": 1.83, "learning_rate": 6.025602352852457e-06, "loss": 0.7175, "step": 26975 }, { "epoch": 1.83, "learning_rate": 6.024659703631085e-06, "loss": 0.288, "step": 26980 }, { "epoch": 1.83, "learning_rate": 6.023717054409713e-06, "loss": 0.5423, "step": 26985 }, { "epoch": 1.83, "learning_rate": 6.0227744051883416e-06, "loss": 0.4069, "step": 26990 }, { "epoch": 1.83, "learning_rate": 6.02183175596697e-06, "loss": 0.3809, "step": 26995 }, { "epoch": 1.83, "learning_rate": 6.0208891067455985e-06, "loss": 0.7868, "step": 27000 }, { "epoch": 1.83, "learning_rate": 6.019946457524226e-06, "loss": 0.2529, "step": 27005 }, { "epoch": 1.83, "learning_rate": 6.019003808302855e-06, "loss": 0.2633, "step": 27010 }, { "epoch": 1.83, "learning_rate": 6.018061159081483e-06, "loss": 0.242, "step": 27015 }, { "epoch": 1.83, "learning_rate": 6.017118509860112e-06, "loss": 0.5046, "step": 27020 }, { "epoch": 1.83, "learning_rate": 6.01617586063874e-06, "loss": 0.8679, "step": 27025 }, { "epoch": 1.83, "learning_rate": 6.015233211417369e-06, "loss": 0.6174, "step": 27030 }, { "epoch": 1.83, "learning_rate": 6.014290562195995e-06, "loss": 1.0642, "step": 27035 }, { "epoch": 1.84, "learning_rate": 6.013347912974624e-06, "loss": 0.3183, "step": 27040 }, { "epoch": 1.84, "learning_rate": 6.012405263753252e-06, "loss": 0.3602, "step": 27045 }, { "epoch": 1.84, "learning_rate": 6.011462614531881e-06, "loss": 0.1967, "step": 27050 }, { "epoch": 1.84, "learning_rate": 6.010519965310509e-06, "loss": 0.2344, "step": 27055 }, { "epoch": 1.84, "learning_rate": 6.009577316089137e-06, "loss": 0.3822, "step": 27060 }, { "epoch": 1.84, "learning_rate": 6.0086346668677655e-06, "loss": 0.3988, "step": 27065 }, { "epoch": 1.84, "learning_rate": 6.007692017646394e-06, "loss": 0.5564, "step": 27070 }, { "epoch": 1.84, "learning_rate": 6.0067493684250225e-06, "loss": 0.3443, "step": 27075 }, { "epoch": 1.84, "learning_rate": 6.005806719203651e-06, "loss": 0.4623, "step": 27080 }, { "epoch": 1.84, "learning_rate": 6.0048640699822794e-06, "loss": 1.0047, "step": 27085 }, { "epoch": 1.84, "learning_rate": 6.003921420760906e-06, "loss": 0.2691, "step": 27090 }, { "epoch": 1.84, "learning_rate": 6.002978771539535e-06, "loss": 1.0063, "step": 27095 }, { "epoch": 1.84, "learning_rate": 6.002036122318163e-06, "loss": 0.6098, "step": 27100 }, { "epoch": 1.84, "learning_rate": 6.001093473096792e-06, "loss": 0.123, "step": 27105 }, { "epoch": 1.84, "learning_rate": 6.00015082387542e-06, "loss": 0.6735, "step": 27110 }, { "epoch": 1.84, "learning_rate": 5.999208174654049e-06, "loss": 0.788, "step": 27115 }, { "epoch": 1.84, "learning_rate": 5.998265525432676e-06, "loss": 0.2868, "step": 27120 }, { "epoch": 1.84, "learning_rate": 5.997322876211305e-06, "loss": 0.2774, "step": 27125 }, { "epoch": 1.84, "learning_rate": 5.996380226989933e-06, "loss": 0.3581, "step": 27130 }, { "epoch": 1.84, "learning_rate": 5.995437577768562e-06, "loss": 0.1678, "step": 27135 }, { "epoch": 1.84, "learning_rate": 5.99449492854719e-06, "loss": 0.3324, "step": 27140 }, { "epoch": 1.84, "learning_rate": 5.993552279325818e-06, "loss": 0.1265, "step": 27145 }, { "epoch": 1.84, "learning_rate": 5.9926096301044455e-06, "loss": 0.5499, "step": 27150 }, { "epoch": 1.84, "learning_rate": 5.991666980883074e-06, "loss": 0.5793, "step": 27155 }, { "epoch": 1.84, "learning_rate": 5.9907243316617025e-06, "loss": 0.2802, "step": 27160 }, { "epoch": 1.84, "learning_rate": 5.989781682440331e-06, "loss": 0.4907, "step": 27165 }, { "epoch": 1.84, "learning_rate": 5.9888390332189595e-06, "loss": 0.2699, "step": 27170 }, { "epoch": 1.84, "learning_rate": 5.987896383997587e-06, "loss": 0.6053, "step": 27175 }, { "epoch": 1.84, "learning_rate": 5.986953734776216e-06, "loss": 0.404, "step": 27180 }, { "epoch": 1.85, "learning_rate": 5.986011085554844e-06, "loss": 0.4035, "step": 27185 }, { "epoch": 1.85, "learning_rate": 5.985068436333472e-06, "loss": 0.1641, "step": 27190 }, { "epoch": 1.85, "learning_rate": 5.9841257871121e-06, "loss": 0.5352, "step": 27195 }, { "epoch": 1.85, "learning_rate": 5.983183137890729e-06, "loss": 0.4976, "step": 27200 }, { "epoch": 1.85, "learning_rate": 5.982240488669356e-06, "loss": 0.7454, "step": 27205 }, { "epoch": 1.85, "learning_rate": 5.981297839447985e-06, "loss": 0.2536, "step": 27210 }, { "epoch": 1.85, "learning_rate": 5.980355190226613e-06, "loss": 0.337, "step": 27215 }, { "epoch": 1.85, "learning_rate": 5.979412541005242e-06, "loss": 0.6152, "step": 27220 }, { "epoch": 1.85, "learning_rate": 5.97846989178387e-06, "loss": 0.3238, "step": 27225 }, { "epoch": 1.85, "learning_rate": 5.977527242562499e-06, "loss": 0.6731, "step": 27230 }, { "epoch": 1.85, "learning_rate": 5.9765845933411256e-06, "loss": 0.3078, "step": 27235 }, { "epoch": 1.85, "learning_rate": 5.975641944119754e-06, "loss": 0.6122, "step": 27240 }, { "epoch": 1.85, "learning_rate": 5.9746992948983826e-06, "loss": 0.6697, "step": 27245 }, { "epoch": 1.85, "learning_rate": 5.973756645677011e-06, "loss": 0.9488, "step": 27250 }, { "epoch": 1.85, "learning_rate": 5.9728139964556395e-06, "loss": 0.2063, "step": 27255 }, { "epoch": 1.85, "learning_rate": 5.971871347234268e-06, "loss": 0.8443, "step": 27260 }, { "epoch": 1.85, "learning_rate": 5.970928698012896e-06, "loss": 0.2954, "step": 27265 }, { "epoch": 1.85, "learning_rate": 5.969986048791524e-06, "loss": 0.4923, "step": 27270 }, { "epoch": 1.85, "learning_rate": 5.969043399570153e-06, "loss": 0.5696, "step": 27275 }, { "epoch": 1.85, "learning_rate": 5.968100750348781e-06, "loss": 0.8005, "step": 27280 }, { "epoch": 1.85, "learning_rate": 5.96715810112741e-06, "loss": 0.5612, "step": 27285 }, { "epoch": 1.85, "learning_rate": 5.966215451906036e-06, "loss": 0.6907, "step": 27290 }, { "epoch": 1.85, "learning_rate": 5.965272802684665e-06, "loss": 0.5352, "step": 27295 }, { "epoch": 1.85, "learning_rate": 5.964330153463293e-06, "loss": 0.5701, "step": 27300 }, { "epoch": 1.85, "learning_rate": 5.963387504241922e-06, "loss": 0.1478, "step": 27305 }, { "epoch": 1.85, "learning_rate": 5.96244485502055e-06, "loss": 0.9349, "step": 27310 }, { "epoch": 1.85, "learning_rate": 5.961502205799179e-06, "loss": 0.363, "step": 27315 }, { "epoch": 1.85, "learning_rate": 5.9605595565778065e-06, "loss": 0.6713, "step": 27320 }, { "epoch": 1.85, "learning_rate": 5.959616907356435e-06, "loss": 0.7095, "step": 27325 }, { "epoch": 1.85, "learning_rate": 5.9586742581350634e-06, "loss": 0.2915, "step": 27330 }, { "epoch": 1.86, "learning_rate": 5.957731608913692e-06, "loss": 0.9005, "step": 27335 }, { "epoch": 1.86, "learning_rate": 5.95678895969232e-06, "loss": 0.61, "step": 27340 }, { "epoch": 1.86, "learning_rate": 5.955846310470949e-06, "loss": 0.6157, "step": 27345 }, { "epoch": 1.86, "learning_rate": 5.954903661249576e-06, "loss": 0.1324, "step": 27350 }, { "epoch": 1.86, "learning_rate": 5.953961012028204e-06, "loss": 0.5081, "step": 27355 }, { "epoch": 1.86, "learning_rate": 5.953018362806833e-06, "loss": 0.4677, "step": 27360 }, { "epoch": 1.86, "learning_rate": 5.952075713585461e-06, "loss": 0.1958, "step": 27365 }, { "epoch": 1.86, "learning_rate": 5.95113306436409e-06, "loss": 0.2633, "step": 27370 }, { "epoch": 1.86, "learning_rate": 5.950190415142718e-06, "loss": 0.5049, "step": 27375 }, { "epoch": 1.86, "learning_rate": 5.949247765921346e-06, "loss": 0.4974, "step": 27380 }, { "epoch": 1.86, "learning_rate": 5.948305116699974e-06, "loss": 0.4868, "step": 27385 }, { "epoch": 1.86, "learning_rate": 5.947362467478603e-06, "loss": 0.458, "step": 27390 }, { "epoch": 1.86, "learning_rate": 5.94641981825723e-06, "loss": 0.2021, "step": 27395 }, { "epoch": 1.86, "learning_rate": 5.945477169035859e-06, "loss": 0.5314, "step": 27400 }, { "epoch": 1.86, "learning_rate": 5.9445345198144865e-06, "loss": 0.5416, "step": 27405 }, { "epoch": 1.86, "learning_rate": 5.943591870593115e-06, "loss": 0.7032, "step": 27410 }, { "epoch": 1.86, "learning_rate": 5.9426492213717435e-06, "loss": 0.8311, "step": 27415 }, { "epoch": 1.86, "learning_rate": 5.941706572150372e-06, "loss": 0.2117, "step": 27420 }, { "epoch": 1.86, "learning_rate": 5.9407639229290005e-06, "loss": 0.249, "step": 27425 }, { "epoch": 1.86, "learning_rate": 5.939821273707629e-06, "loss": 0.6894, "step": 27430 }, { "epoch": 1.86, "learning_rate": 5.938878624486257e-06, "loss": 0.1653, "step": 27435 }, { "epoch": 1.86, "learning_rate": 5.937935975264884e-06, "loss": 0.48, "step": 27440 }, { "epoch": 1.86, "learning_rate": 5.936993326043513e-06, "loss": 0.7069, "step": 27445 }, { "epoch": 1.86, "learning_rate": 5.936050676822141e-06, "loss": 0.28, "step": 27450 }, { "epoch": 1.86, "learning_rate": 5.93510802760077e-06, "loss": 0.5721, "step": 27455 }, { "epoch": 1.86, "learning_rate": 5.934165378379398e-06, "loss": 0.7787, "step": 27460 }, { "epoch": 1.86, "learning_rate": 5.933222729158026e-06, "loss": 0.4371, "step": 27465 }, { "epoch": 1.86, "learning_rate": 5.932280079936654e-06, "loss": 0.518, "step": 27470 }, { "epoch": 1.86, "learning_rate": 5.931337430715283e-06, "loss": 0.2124, "step": 27475 }, { "epoch": 1.87, "learning_rate": 5.930394781493911e-06, "loss": 0.2719, "step": 27480 }, { "epoch": 1.87, "learning_rate": 5.92945213227254e-06, "loss": 0.5885, "step": 27485 }, { "epoch": 1.87, "learning_rate": 5.928509483051168e-06, "loss": 0.4364, "step": 27490 }, { "epoch": 1.87, "learning_rate": 5.927566833829795e-06, "loss": 0.1732, "step": 27495 }, { "epoch": 1.87, "learning_rate": 5.9266241846084235e-06, "loss": 0.6311, "step": 27500 }, { "epoch": 1.87, "learning_rate": 5.925681535387052e-06, "loss": 0.3928, "step": 27505 }, { "epoch": 1.87, "learning_rate": 5.9247388861656805e-06, "loss": 0.501, "step": 27510 }, { "epoch": 1.87, "learning_rate": 5.923796236944309e-06, "loss": 1.1019, "step": 27515 }, { "epoch": 1.87, "learning_rate": 5.922853587722937e-06, "loss": 0.4316, "step": 27520 }, { "epoch": 1.87, "learning_rate": 5.921910938501565e-06, "loss": 0.302, "step": 27525 }, { "epoch": 1.87, "learning_rate": 5.920968289280194e-06, "loss": 0.1053, "step": 27530 }, { "epoch": 1.87, "learning_rate": 5.920025640058822e-06, "loss": 0.2126, "step": 27535 }, { "epoch": 1.87, "learning_rate": 5.919082990837451e-06, "loss": 0.8633, "step": 27540 }, { "epoch": 1.87, "learning_rate": 5.918140341616079e-06, "loss": 0.2003, "step": 27545 }, { "epoch": 1.87, "learning_rate": 5.917197692394706e-06, "loss": 0.5769, "step": 27550 }, { "epoch": 1.87, "learning_rate": 5.916255043173334e-06, "loss": 1.1904, "step": 27555 }, { "epoch": 1.87, "learning_rate": 5.915312393951963e-06, "loss": 0.1159, "step": 27560 }, { "epoch": 1.87, "learning_rate": 5.914369744730591e-06, "loss": 0.1559, "step": 27565 }, { "epoch": 1.87, "learning_rate": 5.91342709550922e-06, "loss": 0.795, "step": 27570 }, { "epoch": 1.87, "learning_rate": 5.912484446287848e-06, "loss": 0.4405, "step": 27575 }, { "epoch": 1.87, "learning_rate": 5.911541797066476e-06, "loss": 0.3762, "step": 27580 }, { "epoch": 1.87, "learning_rate": 5.9105991478451044e-06, "loss": 0.5839, "step": 27585 }, { "epoch": 1.87, "learning_rate": 5.909656498623733e-06, "loss": 0.7678, "step": 27590 }, { "epoch": 1.87, "learning_rate": 5.908713849402361e-06, "loss": 0.5555, "step": 27595 }, { "epoch": 1.87, "learning_rate": 5.90777120018099e-06, "loss": 0.9933, "step": 27600 }, { "epoch": 1.87, "learning_rate": 5.9068285509596175e-06, "loss": 0.4127, "step": 27605 }, { "epoch": 1.87, "learning_rate": 5.905885901738245e-06, "loss": 0.5214, "step": 27610 }, { "epoch": 1.87, "learning_rate": 5.904943252516874e-06, "loss": 0.2347, "step": 27615 }, { "epoch": 1.87, "learning_rate": 5.904000603295502e-06, "loss": 0.7673, "step": 27620 }, { "epoch": 1.87, "learning_rate": 5.903057954074131e-06, "loss": 0.7651, "step": 27625 }, { "epoch": 1.88, "learning_rate": 5.902115304852759e-06, "loss": 0.1467, "step": 27630 }, { "epoch": 1.88, "learning_rate": 5.901172655631387e-06, "loss": 0.4417, "step": 27635 }, { "epoch": 1.88, "learning_rate": 5.900230006410015e-06, "loss": 0.2754, "step": 27640 }, { "epoch": 1.88, "learning_rate": 5.899287357188643e-06, "loss": 0.5884, "step": 27645 }, { "epoch": 1.88, "learning_rate": 5.898344707967271e-06, "loss": 0.3925, "step": 27650 }, { "epoch": 1.88, "learning_rate": 5.8974020587459e-06, "loss": 0.3663, "step": 27655 }, { "epoch": 1.88, "learning_rate": 5.896459409524528e-06, "loss": 0.6044, "step": 27660 }, { "epoch": 1.88, "learning_rate": 5.895516760303156e-06, "loss": 0.3551, "step": 27665 }, { "epoch": 1.88, "learning_rate": 5.8945741110817845e-06, "loss": 0.3727, "step": 27670 }, { "epoch": 1.88, "learning_rate": 5.893631461860413e-06, "loss": 0.6176, "step": 27675 }, { "epoch": 1.88, "learning_rate": 5.8926888126390415e-06, "loss": 0.5507, "step": 27680 }, { "epoch": 1.88, "learning_rate": 5.89174616341767e-06, "loss": 1.0475, "step": 27685 }, { "epoch": 1.88, "learning_rate": 5.8908035141962984e-06, "loss": 0.3249, "step": 27690 }, { "epoch": 1.88, "learning_rate": 5.889860864974925e-06, "loss": 0.143, "step": 27695 }, { "epoch": 1.88, "learning_rate": 5.888918215753554e-06, "loss": 0.3235, "step": 27700 }, { "epoch": 1.88, "learning_rate": 5.887975566532182e-06, "loss": 0.6612, "step": 27705 }, { "epoch": 1.88, "learning_rate": 5.887032917310811e-06, "loss": 0.2644, "step": 27710 }, { "epoch": 1.88, "learning_rate": 5.886090268089439e-06, "loss": 0.5081, "step": 27715 }, { "epoch": 1.88, "learning_rate": 5.885147618868068e-06, "loss": 0.2071, "step": 27720 }, { "epoch": 1.88, "learning_rate": 5.884204969646695e-06, "loss": 0.4134, "step": 27725 }, { "epoch": 1.88, "learning_rate": 5.883262320425324e-06, "loss": 0.7594, "step": 27730 }, { "epoch": 1.88, "learning_rate": 5.882319671203952e-06, "loss": 0.4322, "step": 27735 }, { "epoch": 1.88, "learning_rate": 5.881377021982581e-06, "loss": 0.817, "step": 27740 }, { "epoch": 1.88, "learning_rate": 5.880434372761209e-06, "loss": 0.4296, "step": 27745 }, { "epoch": 1.88, "learning_rate": 5.879491723539836e-06, "loss": 0.2153, "step": 27750 }, { "epoch": 1.88, "learning_rate": 5.8785490743184645e-06, "loss": 0.5634, "step": 27755 }, { "epoch": 1.88, "learning_rate": 5.877606425097093e-06, "loss": 0.2247, "step": 27760 }, { "epoch": 1.88, "learning_rate": 5.8766637758757215e-06, "loss": 0.2132, "step": 27765 }, { "epoch": 1.88, "learning_rate": 5.87572112665435e-06, "loss": 0.3357, "step": 27770 }, { "epoch": 1.89, "learning_rate": 5.8747784774329785e-06, "loss": 0.4661, "step": 27775 }, { "epoch": 1.89, "learning_rate": 5.873835828211606e-06, "loss": 0.8499, "step": 27780 }, { "epoch": 1.89, "learning_rate": 5.872893178990235e-06, "loss": 0.5638, "step": 27785 }, { "epoch": 1.89, "learning_rate": 5.871950529768863e-06, "loss": 0.8345, "step": 27790 }, { "epoch": 1.89, "learning_rate": 5.871007880547492e-06, "loss": 0.6416, "step": 27795 }, { "epoch": 1.89, "learning_rate": 5.87006523132612e-06, "loss": 0.2917, "step": 27800 }, { "epoch": 1.89, "learning_rate": 5.8691225821047486e-06, "loss": 0.2319, "step": 27805 }, { "epoch": 1.89, "learning_rate": 5.868179932883375e-06, "loss": 0.4131, "step": 27810 }, { "epoch": 1.89, "learning_rate": 5.867237283662004e-06, "loss": 0.4818, "step": 27815 }, { "epoch": 1.89, "learning_rate": 5.866294634440632e-06, "loss": 0.1951, "step": 27820 }, { "epoch": 1.89, "learning_rate": 5.865351985219261e-06, "loss": 0.5289, "step": 27825 }, { "epoch": 1.89, "learning_rate": 5.864409335997889e-06, "loss": 0.6074, "step": 27830 }, { "epoch": 1.89, "learning_rate": 5.863466686776517e-06, "loss": 0.8415, "step": 27835 }, { "epoch": 1.89, "learning_rate": 5.8625240375551454e-06, "loss": 0.5084, "step": 27840 }, { "epoch": 1.89, "learning_rate": 5.861581388333774e-06, "loss": 0.9302, "step": 27845 }, { "epoch": 1.89, "learning_rate": 5.860638739112402e-06, "loss": 0.5526, "step": 27850 }, { "epoch": 1.89, "learning_rate": 5.85969608989103e-06, "loss": 0.2385, "step": 27855 }, { "epoch": 1.89, "learning_rate": 5.8587534406696585e-06, "loss": 0.1957, "step": 27860 }, { "epoch": 1.89, "learning_rate": 5.857810791448286e-06, "loss": 0.3513, "step": 27865 }, { "epoch": 1.89, "learning_rate": 5.856868142226915e-06, "loss": 0.9237, "step": 27870 }, { "epoch": 1.89, "learning_rate": 5.855925493005543e-06, "loss": 0.2815, "step": 27875 }, { "epoch": 1.89, "learning_rate": 5.854982843784172e-06, "loss": 0.168, "step": 27880 }, { "epoch": 1.89, "learning_rate": 5.8540401945628e-06, "loss": 0.3153, "step": 27885 }, { "epoch": 1.89, "learning_rate": 5.853097545341429e-06, "loss": 0.3408, "step": 27890 }, { "epoch": 1.89, "learning_rate": 5.852154896120055e-06, "loss": 0.3503, "step": 27895 }, { "epoch": 1.89, "learning_rate": 5.851212246898684e-06, "loss": 0.6463, "step": 27900 }, { "epoch": 1.89, "learning_rate": 5.850269597677312e-06, "loss": 0.1798, "step": 27905 }, { "epoch": 1.89, "learning_rate": 5.849326948455941e-06, "loss": 0.2545, "step": 27910 }, { "epoch": 1.89, "learning_rate": 5.848384299234569e-06, "loss": 0.2689, "step": 27915 }, { "epoch": 1.89, "learning_rate": 5.847441650013198e-06, "loss": 0.719, "step": 27920 }, { "epoch": 1.9, "learning_rate": 5.8464990007918255e-06, "loss": 0.5271, "step": 27925 }, { "epoch": 1.9, "learning_rate": 5.845556351570454e-06, "loss": 0.255, "step": 27930 }, { "epoch": 1.9, "learning_rate": 5.8446137023490825e-06, "loss": 0.3652, "step": 27935 }, { "epoch": 1.9, "learning_rate": 5.843671053127711e-06, "loss": 0.6515, "step": 27940 }, { "epoch": 1.9, "learning_rate": 5.8427284039063394e-06, "loss": 0.1736, "step": 27945 }, { "epoch": 1.9, "learning_rate": 5.841785754684966e-06, "loss": 0.2901, "step": 27950 }, { "epoch": 1.9, "learning_rate": 5.840843105463595e-06, "loss": 0.2803, "step": 27955 }, { "epoch": 1.9, "learning_rate": 5.839900456242223e-06, "loss": 0.2648, "step": 27960 }, { "epoch": 1.9, "learning_rate": 5.838957807020852e-06, "loss": 0.4323, "step": 27965 }, { "epoch": 1.9, "learning_rate": 5.83801515779948e-06, "loss": 0.2219, "step": 27970 }, { "epoch": 1.9, "learning_rate": 5.837072508578109e-06, "loss": 0.4793, "step": 27975 }, { "epoch": 1.9, "learning_rate": 5.836129859356736e-06, "loss": 0.8686, "step": 27980 }, { "epoch": 1.9, "learning_rate": 5.835187210135365e-06, "loss": 0.2369, "step": 27985 }, { "epoch": 1.9, "learning_rate": 5.834244560913993e-06, "loss": 0.4083, "step": 27990 }, { "epoch": 1.9, "learning_rate": 5.833301911692622e-06, "loss": 0.7591, "step": 27995 }, { "epoch": 1.9, "learning_rate": 5.83235926247125e-06, "loss": 0.6209, "step": 28000 }, { "epoch": 1.9, "learning_rate": 5.831416613249879e-06, "loss": 0.2365, "step": 28005 }, { "epoch": 1.9, "learning_rate": 5.8304739640285055e-06, "loss": 0.3802, "step": 28010 }, { "epoch": 1.9, "learning_rate": 5.829531314807134e-06, "loss": 0.858, "step": 28015 }, { "epoch": 1.9, "learning_rate": 5.8285886655857625e-06, "loss": 0.598, "step": 28020 }, { "epoch": 1.9, "learning_rate": 5.827646016364391e-06, "loss": 0.2769, "step": 28025 }, { "epoch": 1.9, "learning_rate": 5.8267033671430195e-06, "loss": 0.8469, "step": 28030 }, { "epoch": 1.9, "learning_rate": 5.825760717921648e-06, "loss": 0.816, "step": 28035 }, { "epoch": 1.9, "learning_rate": 5.824818068700276e-06, "loss": 0.2482, "step": 28040 }, { "epoch": 1.9, "learning_rate": 5.823875419478904e-06, "loss": 0.4872, "step": 28045 }, { "epoch": 1.9, "learning_rate": 5.822932770257533e-06, "loss": 0.6412, "step": 28050 }, { "epoch": 1.9, "learning_rate": 5.821990121036161e-06, "loss": 0.6169, "step": 28055 }, { "epoch": 1.9, "learning_rate": 5.821047471814789e-06, "loss": 0.4138, "step": 28060 }, { "epoch": 1.9, "learning_rate": 5.820104822593416e-06, "loss": 0.4332, "step": 28065 }, { "epoch": 1.91, "learning_rate": 5.819162173372045e-06, "loss": 0.2293, "step": 28070 }, { "epoch": 1.91, "learning_rate": 5.818219524150673e-06, "loss": 0.4924, "step": 28075 }, { "epoch": 1.91, "learning_rate": 5.817276874929302e-06, "loss": 0.5927, "step": 28080 }, { "epoch": 1.91, "learning_rate": 5.81633422570793e-06, "loss": 0.4907, "step": 28085 }, { "epoch": 1.91, "learning_rate": 5.815391576486559e-06, "loss": 0.253, "step": 28090 }, { "epoch": 1.91, "learning_rate": 5.8144489272651864e-06, "loss": 0.7777, "step": 28095 }, { "epoch": 1.91, "learning_rate": 5.813506278043815e-06, "loss": 0.516, "step": 28100 }, { "epoch": 1.91, "learning_rate": 5.8125636288224426e-06, "loss": 0.3979, "step": 28105 }, { "epoch": 1.91, "learning_rate": 5.811620979601071e-06, "loss": 0.5502, "step": 28110 }, { "epoch": 1.91, "learning_rate": 5.8106783303796995e-06, "loss": 1.0718, "step": 28115 }, { "epoch": 1.91, "learning_rate": 5.809735681158328e-06, "loss": 0.3048, "step": 28120 }, { "epoch": 1.91, "learning_rate": 5.808793031936956e-06, "loss": 0.1875, "step": 28125 }, { "epoch": 1.91, "learning_rate": 5.807850382715584e-06, "loss": 0.496, "step": 28130 }, { "epoch": 1.91, "learning_rate": 5.806907733494213e-06, "loss": 0.5404, "step": 28135 }, { "epoch": 1.91, "learning_rate": 5.805965084272841e-06, "loss": 0.2816, "step": 28140 }, { "epoch": 1.91, "learning_rate": 5.80502243505147e-06, "loss": 0.6197, "step": 28145 }, { "epoch": 1.91, "learning_rate": 5.804079785830098e-06, "loss": 0.4902, "step": 28150 }, { "epoch": 1.91, "learning_rate": 5.803137136608725e-06, "loss": 0.1852, "step": 28155 }, { "epoch": 1.91, "learning_rate": 5.802194487387353e-06, "loss": 0.1477, "step": 28160 }, { "epoch": 1.91, "learning_rate": 5.801251838165982e-06, "loss": 0.7472, "step": 28165 }, { "epoch": 1.91, "learning_rate": 5.80030918894461e-06, "loss": 0.361, "step": 28170 }, { "epoch": 1.91, "learning_rate": 5.799366539723239e-06, "loss": 0.5793, "step": 28175 }, { "epoch": 1.91, "learning_rate": 5.7984238905018665e-06, "loss": 0.3844, "step": 28180 }, { "epoch": 1.91, "learning_rate": 5.797481241280495e-06, "loss": 0.0345, "step": 28185 }, { "epoch": 1.91, "learning_rate": 5.7965385920591235e-06, "loss": 0.1993, "step": 28190 }, { "epoch": 1.91, "learning_rate": 5.795595942837752e-06, "loss": 0.8184, "step": 28195 }, { "epoch": 1.91, "learning_rate": 5.7946532936163804e-06, "loss": 0.9079, "step": 28200 }, { "epoch": 1.91, "learning_rate": 5.793710644395009e-06, "loss": 0.5583, "step": 28205 }, { "epoch": 1.91, "learning_rate": 5.792767995173636e-06, "loss": 0.3934, "step": 28210 }, { "epoch": 1.91, "learning_rate": 5.791825345952264e-06, "loss": 0.5291, "step": 28215 }, { "epoch": 1.92, "learning_rate": 5.790882696730893e-06, "loss": 0.3187, "step": 28220 }, { "epoch": 1.92, "learning_rate": 5.789940047509521e-06, "loss": 0.4342, "step": 28225 }, { "epoch": 1.92, "learning_rate": 5.78899739828815e-06, "loss": 0.1907, "step": 28230 }, { "epoch": 1.92, "learning_rate": 5.788054749066778e-06, "loss": 0.4798, "step": 28235 }, { "epoch": 1.92, "learning_rate": 5.787112099845406e-06, "loss": 1.1958, "step": 28240 }, { "epoch": 1.92, "learning_rate": 5.786169450624034e-06, "loss": 0.4814, "step": 28245 }, { "epoch": 1.92, "learning_rate": 5.785226801402663e-06, "loss": 0.516, "step": 28250 }, { "epoch": 1.92, "learning_rate": 5.784284152181291e-06, "loss": 0.1874, "step": 28255 }, { "epoch": 1.92, "learning_rate": 5.78334150295992e-06, "loss": 0.4834, "step": 28260 }, { "epoch": 1.92, "learning_rate": 5.782398853738548e-06, "loss": 0.3757, "step": 28265 }, { "epoch": 1.92, "learning_rate": 5.781456204517175e-06, "loss": 0.6562, "step": 28270 }, { "epoch": 1.92, "learning_rate": 5.7805135552958035e-06, "loss": 0.7466, "step": 28275 }, { "epoch": 1.92, "learning_rate": 5.779570906074432e-06, "loss": 0.4493, "step": 28280 }, { "epoch": 1.92, "learning_rate": 5.7786282568530605e-06, "loss": 0.1611, "step": 28285 }, { "epoch": 1.92, "learning_rate": 5.777685607631689e-06, "loss": 0.5062, "step": 28290 }, { "epoch": 1.92, "learning_rate": 5.776742958410317e-06, "loss": 0.4989, "step": 28295 }, { "epoch": 1.92, "learning_rate": 5.775800309188945e-06, "loss": 0.4718, "step": 28300 }, { "epoch": 1.92, "learning_rate": 5.774857659967574e-06, "loss": 0.2976, "step": 28305 }, { "epoch": 1.92, "learning_rate": 5.773915010746202e-06, "loss": 0.8792, "step": 28310 }, { "epoch": 1.92, "learning_rate": 5.77297236152483e-06, "loss": 0.1378, "step": 28315 }, { "epoch": 1.92, "learning_rate": 5.772029712303458e-06, "loss": 0.2422, "step": 28320 }, { "epoch": 1.92, "learning_rate": 5.771087063082086e-06, "loss": 0.6178, "step": 28325 }, { "epoch": 1.92, "learning_rate": 5.770144413860714e-06, "loss": 0.402, "step": 28330 }, { "epoch": 1.92, "learning_rate": 5.769201764639343e-06, "loss": 0.1459, "step": 28335 }, { "epoch": 1.92, "learning_rate": 5.768259115417971e-06, "loss": 0.0599, "step": 28340 }, { "epoch": 1.92, "learning_rate": 5.7673164661966e-06, "loss": 0.3616, "step": 28345 }, { "epoch": 1.92, "learning_rate": 5.766373816975228e-06, "loss": 0.6553, "step": 28350 }, { "epoch": 1.92, "learning_rate": 5.765431167753855e-06, "loss": 0.4128, "step": 28355 }, { "epoch": 1.92, "learning_rate": 5.7644885185324836e-06, "loss": 0.8936, "step": 28360 }, { "epoch": 1.93, "learning_rate": 5.763545869311112e-06, "loss": 0.3547, "step": 28365 }, { "epoch": 1.93, "learning_rate": 5.7626032200897405e-06, "loss": 0.3841, "step": 28370 }, { "epoch": 1.93, "learning_rate": 5.761660570868369e-06, "loss": 0.6581, "step": 28375 }, { "epoch": 1.93, "learning_rate": 5.7607179216469975e-06, "loss": 0.565, "step": 28380 }, { "epoch": 1.93, "learning_rate": 5.759775272425625e-06, "loss": 0.1379, "step": 28385 }, { "epoch": 1.93, "learning_rate": 5.758832623204254e-06, "loss": 0.7595, "step": 28390 }, { "epoch": 1.93, "learning_rate": 5.757889973982882e-06, "loss": 0.5746, "step": 28395 }, { "epoch": 1.93, "learning_rate": 5.756947324761511e-06, "loss": 0.514, "step": 28400 }, { "epoch": 1.93, "learning_rate": 5.756004675540139e-06, "loss": 0.4235, "step": 28405 }, { "epoch": 1.93, "learning_rate": 5.755062026318766e-06, "loss": 0.3076, "step": 28410 }, { "epoch": 1.93, "learning_rate": 5.754119377097394e-06, "loss": 0.7172, "step": 28415 }, { "epoch": 1.93, "learning_rate": 5.753176727876023e-06, "loss": 0.2947, "step": 28420 }, { "epoch": 1.93, "learning_rate": 5.752234078654651e-06, "loss": 0.4503, "step": 28425 }, { "epoch": 1.93, "learning_rate": 5.75129142943328e-06, "loss": 0.2343, "step": 28430 }, { "epoch": 1.93, "learning_rate": 5.750348780211908e-06, "loss": 0.4272, "step": 28435 }, { "epoch": 1.93, "learning_rate": 5.749406130990536e-06, "loss": 0.3881, "step": 28440 }, { "epoch": 1.93, "learning_rate": 5.7484634817691644e-06, "loss": 0.3845, "step": 28445 }, { "epoch": 1.93, "learning_rate": 5.747520832547793e-06, "loss": 0.8651, "step": 28450 }, { "epoch": 1.93, "learning_rate": 5.746578183326421e-06, "loss": 0.711, "step": 28455 }, { "epoch": 1.93, "learning_rate": 5.74563553410505e-06, "loss": 0.8783, "step": 28460 }, { "epoch": 1.93, "learning_rate": 5.744692884883678e-06, "loss": 0.1185, "step": 28465 }, { "epoch": 1.93, "learning_rate": 5.743750235662305e-06, "loss": 0.7762, "step": 28470 }, { "epoch": 1.93, "learning_rate": 5.742807586440934e-06, "loss": 0.5603, "step": 28475 }, { "epoch": 1.93, "learning_rate": 5.741864937219562e-06, "loss": 0.144, "step": 28480 }, { "epoch": 1.93, "learning_rate": 5.740922287998191e-06, "loss": 0.2866, "step": 28485 }, { "epoch": 1.93, "learning_rate": 5.739979638776819e-06, "loss": 0.3949, "step": 28490 }, { "epoch": 1.93, "learning_rate": 5.739036989555448e-06, "loss": 0.5101, "step": 28495 }, { "epoch": 1.93, "learning_rate": 5.738094340334075e-06, "loss": 0.2471, "step": 28500 }, { "epoch": 1.93, "learning_rate": 5.737151691112704e-06, "loss": 0.9142, "step": 28505 }, { "epoch": 1.93, "learning_rate": 5.736209041891332e-06, "loss": 0.504, "step": 28510 }, { "epoch": 1.94, "learning_rate": 5.735266392669961e-06, "loss": 0.286, "step": 28515 }, { "epoch": 1.94, "learning_rate": 5.734323743448588e-06, "loss": 0.7291, "step": 28520 }, { "epoch": 1.94, "learning_rate": 5.733381094227216e-06, "loss": 0.464, "step": 28525 }, { "epoch": 1.94, "learning_rate": 5.7324384450058445e-06, "loss": 0.6168, "step": 28530 }, { "epoch": 1.94, "learning_rate": 5.731495795784473e-06, "loss": 0.2601, "step": 28535 }, { "epoch": 1.94, "learning_rate": 5.7305531465631015e-06, "loss": 0.3178, "step": 28540 }, { "epoch": 1.94, "learning_rate": 5.72961049734173e-06, "loss": 0.4407, "step": 28545 }, { "epoch": 1.94, "learning_rate": 5.7286678481203584e-06, "loss": 0.5674, "step": 28550 }, { "epoch": 1.94, "learning_rate": 5.727725198898986e-06, "loss": 1.0393, "step": 28555 }, { "epoch": 1.94, "learning_rate": 5.7267825496776146e-06, "loss": 0.6656, "step": 28560 }, { "epoch": 1.94, "learning_rate": 5.725839900456242e-06, "loss": 0.2552, "step": 28565 }, { "epoch": 1.94, "learning_rate": 5.724897251234871e-06, "loss": 0.4262, "step": 28570 }, { "epoch": 1.94, "learning_rate": 5.723954602013499e-06, "loss": 0.6399, "step": 28575 }, { "epoch": 1.94, "learning_rate": 5.723011952792128e-06, "loss": 0.7965, "step": 28580 }, { "epoch": 1.94, "learning_rate": 5.722069303570755e-06, "loss": 0.4201, "step": 28585 }, { "epoch": 1.94, "learning_rate": 5.721126654349384e-06, "loss": 0.6174, "step": 28590 }, { "epoch": 1.94, "learning_rate": 5.720184005128012e-06, "loss": 0.4031, "step": 28595 }, { "epoch": 1.94, "learning_rate": 5.719241355906641e-06, "loss": 0.1161, "step": 28600 }, { "epoch": 1.94, "learning_rate": 5.718298706685269e-06, "loss": 0.2771, "step": 28605 }, { "epoch": 1.94, "learning_rate": 5.717356057463898e-06, "loss": 1.0797, "step": 28610 }, { "epoch": 1.94, "learning_rate": 5.7164134082425245e-06, "loss": 0.1273, "step": 28615 }, { "epoch": 1.94, "learning_rate": 5.715470759021153e-06, "loss": 0.7788, "step": 28620 }, { "epoch": 1.94, "learning_rate": 5.7145281097997815e-06, "loss": 0.7352, "step": 28625 }, { "epoch": 1.94, "learning_rate": 5.71358546057841e-06, "loss": 0.8256, "step": 28630 }, { "epoch": 1.94, "learning_rate": 5.7126428113570385e-06, "loss": 0.3134, "step": 28635 }, { "epoch": 1.94, "learning_rate": 5.711700162135666e-06, "loss": 0.3414, "step": 28640 }, { "epoch": 1.94, "learning_rate": 5.710757512914295e-06, "loss": 0.2627, "step": 28645 }, { "epoch": 1.94, "learning_rate": 5.709814863692923e-06, "loss": 0.3614, "step": 28650 }, { "epoch": 1.94, "learning_rate": 5.708872214471552e-06, "loss": 0.9101, "step": 28655 }, { "epoch": 1.95, "learning_rate": 5.70792956525018e-06, "loss": 0.1659, "step": 28660 }, { "epoch": 1.95, "learning_rate": 5.7069869160288086e-06, "loss": 0.3586, "step": 28665 }, { "epoch": 1.95, "learning_rate": 5.706044266807435e-06, "loss": 1.3505, "step": 28670 }, { "epoch": 1.95, "learning_rate": 5.705101617586064e-06, "loss": 0.595, "step": 28675 }, { "epoch": 1.95, "learning_rate": 5.704158968364692e-06, "loss": 0.4485, "step": 28680 }, { "epoch": 1.95, "learning_rate": 5.703216319143321e-06, "loss": 0.2878, "step": 28685 }, { "epoch": 1.95, "learning_rate": 5.702273669921949e-06, "loss": 0.3454, "step": 28690 }, { "epoch": 1.95, "learning_rate": 5.701331020700578e-06, "loss": 0.8996, "step": 28695 }, { "epoch": 1.95, "learning_rate": 5.7003883714792054e-06, "loss": 0.2799, "step": 28700 }, { "epoch": 1.95, "learning_rate": 5.699445722257834e-06, "loss": 0.1786, "step": 28705 }, { "epoch": 1.95, "learning_rate": 5.698503073036462e-06, "loss": 0.2912, "step": 28710 }, { "epoch": 1.95, "learning_rate": 5.697560423815091e-06, "loss": 0.1501, "step": 28715 }, { "epoch": 1.95, "learning_rate": 5.696617774593719e-06, "loss": 0.4875, "step": 28720 }, { "epoch": 1.95, "learning_rate": 5.695675125372348e-06, "loss": 0.6775, "step": 28725 }, { "epoch": 1.95, "learning_rate": 5.694732476150975e-06, "loss": 0.6065, "step": 28730 }, { "epoch": 1.95, "learning_rate": 5.693789826929603e-06, "loss": 0.1524, "step": 28735 }, { "epoch": 1.95, "learning_rate": 5.692847177708232e-06, "loss": 0.1652, "step": 28740 }, { "epoch": 1.95, "learning_rate": 5.69190452848686e-06, "loss": 0.133, "step": 28745 }, { "epoch": 1.95, "learning_rate": 5.690961879265489e-06, "loss": 0.5086, "step": 28750 }, { "epoch": 1.95, "learning_rate": 5.690019230044116e-06, "loss": 0.9846, "step": 28755 }, { "epoch": 1.95, "learning_rate": 5.689076580822745e-06, "loss": 0.6516, "step": 28760 }, { "epoch": 1.95, "learning_rate": 5.688133931601373e-06, "loss": 0.8017, "step": 28765 }, { "epoch": 1.95, "learning_rate": 5.687191282380001e-06, "loss": 0.5429, "step": 28770 }, { "epoch": 1.95, "learning_rate": 5.686248633158629e-06, "loss": 0.6296, "step": 28775 }, { "epoch": 1.95, "learning_rate": 5.685305983937258e-06, "loss": 0.7759, "step": 28780 }, { "epoch": 1.95, "learning_rate": 5.6843633347158855e-06, "loss": 0.3572, "step": 28785 }, { "epoch": 1.95, "learning_rate": 5.683420685494514e-06, "loss": 0.1969, "step": 28790 }, { "epoch": 1.95, "learning_rate": 5.6824780362731425e-06, "loss": 0.3109, "step": 28795 }, { "epoch": 1.95, "learning_rate": 5.681535387051771e-06, "loss": 0.2814, "step": 28800 }, { "epoch": 1.96, "learning_rate": 5.6805927378303994e-06, "loss": 0.4773, "step": 28805 }, { "epoch": 1.96, "learning_rate": 5.679650088609028e-06, "loss": 0.7867, "step": 28810 }, { "epoch": 1.96, "learning_rate": 5.678707439387655e-06, "loss": 0.4264, "step": 28815 }, { "epoch": 1.96, "learning_rate": 5.677764790166283e-06, "loss": 0.4226, "step": 28820 }, { "epoch": 1.96, "learning_rate": 5.676822140944912e-06, "loss": 0.9841, "step": 28825 }, { "epoch": 1.96, "learning_rate": 5.67587949172354e-06, "loss": 0.5124, "step": 28830 }, { "epoch": 1.96, "learning_rate": 5.674936842502169e-06, "loss": 0.649, "step": 28835 }, { "epoch": 1.96, "learning_rate": 5.673994193280797e-06, "loss": 0.7121, "step": 28840 }, { "epoch": 1.96, "learning_rate": 5.673051544059425e-06, "loss": 0.2559, "step": 28845 }, { "epoch": 1.96, "learning_rate": 5.672108894838053e-06, "loss": 0.477, "step": 28850 }, { "epoch": 1.96, "learning_rate": 5.671166245616682e-06, "loss": 0.2584, "step": 28855 }, { "epoch": 1.96, "learning_rate": 5.67022359639531e-06, "loss": 0.2693, "step": 28860 }, { "epoch": 1.96, "learning_rate": 5.669280947173939e-06, "loss": 0.3449, "step": 28865 }, { "epoch": 1.96, "learning_rate": 5.6683382979525655e-06, "loss": 0.3719, "step": 28870 }, { "epoch": 1.96, "learning_rate": 5.667395648731194e-06, "loss": 0.7684, "step": 28875 }, { "epoch": 1.96, "learning_rate": 5.6664529995098225e-06, "loss": 0.2354, "step": 28880 }, { "epoch": 1.96, "learning_rate": 5.665510350288451e-06, "loss": 0.2887, "step": 28885 }, { "epoch": 1.96, "learning_rate": 5.6645677010670795e-06, "loss": 0.2799, "step": 28890 }, { "epoch": 1.96, "learning_rate": 5.663625051845708e-06, "loss": 0.415, "step": 28895 }, { "epoch": 1.96, "learning_rate": 5.662682402624336e-06, "loss": 0.3196, "step": 28900 }, { "epoch": 1.96, "learning_rate": 5.661739753402964e-06, "loss": 0.4245, "step": 28905 }, { "epoch": 1.96, "learning_rate": 5.660797104181593e-06, "loss": 0.648, "step": 28910 }, { "epoch": 1.96, "learning_rate": 5.659854454960221e-06, "loss": 0.8731, "step": 28915 }, { "epoch": 1.96, "learning_rate": 5.6589118057388496e-06, "loss": 0.2521, "step": 28920 }, { "epoch": 1.96, "learning_rate": 5.657969156517478e-06, "loss": 0.2943, "step": 28925 }, { "epoch": 1.96, "learning_rate": 5.657026507296105e-06, "loss": 0.5225, "step": 28930 }, { "epoch": 1.96, "learning_rate": 5.656083858074733e-06, "loss": 0.5997, "step": 28935 }, { "epoch": 1.96, "learning_rate": 5.655141208853362e-06, "loss": 0.4309, "step": 28940 }, { "epoch": 1.96, "learning_rate": 5.65419855963199e-06, "loss": 0.833, "step": 28945 }, { "epoch": 1.96, "learning_rate": 5.653255910410619e-06, "loss": 0.5138, "step": 28950 }, { "epoch": 1.97, "learning_rate": 5.6523132611892464e-06, "loss": 0.2486, "step": 28955 }, { "epoch": 1.97, "learning_rate": 5.651370611967875e-06, "loss": 0.2466, "step": 28960 }, { "epoch": 1.97, "learning_rate": 5.650427962746503e-06, "loss": 0.0419, "step": 28965 }, { "epoch": 1.97, "learning_rate": 5.649485313525132e-06, "loss": 0.5017, "step": 28970 }, { "epoch": 1.97, "learning_rate": 5.64854266430376e-06, "loss": 0.5371, "step": 28975 }, { "epoch": 1.97, "learning_rate": 5.647600015082388e-06, "loss": 1.0224, "step": 28980 }, { "epoch": 1.97, "learning_rate": 5.646657365861016e-06, "loss": 0.307, "step": 28985 }, { "epoch": 1.97, "learning_rate": 5.645714716639644e-06, "loss": 0.4815, "step": 28990 }, { "epoch": 1.97, "learning_rate": 5.644772067418273e-06, "loss": 0.8251, "step": 28995 }, { "epoch": 1.97, "learning_rate": 5.643829418196901e-06, "loss": 0.6143, "step": 29000 }, { "epoch": 1.97, "learning_rate": 5.64288676897553e-06, "loss": 0.53, "step": 29005 }, { "epoch": 1.97, "learning_rate": 5.641944119754158e-06, "loss": 0.2455, "step": 29010 }, { "epoch": 1.97, "learning_rate": 5.641001470532786e-06, "loss": 0.4834, "step": 29015 }, { "epoch": 1.97, "learning_rate": 5.640058821311413e-06, "loss": 0.3333, "step": 29020 }, { "epoch": 1.97, "learning_rate": 5.639116172090042e-06, "loss": 0.5582, "step": 29025 }, { "epoch": 1.97, "learning_rate": 5.63817352286867e-06, "loss": 0.3202, "step": 29030 }, { "epoch": 1.97, "learning_rate": 5.637230873647299e-06, "loss": 0.658, "step": 29035 }, { "epoch": 1.97, "learning_rate": 5.636288224425927e-06, "loss": 1.1244, "step": 29040 }, { "epoch": 1.97, "learning_rate": 5.635345575204555e-06, "loss": 0.3772, "step": 29045 }, { "epoch": 1.97, "learning_rate": 5.6344029259831835e-06, "loss": 0.3251, "step": 29050 }, { "epoch": 1.97, "learning_rate": 5.633460276761812e-06, "loss": 0.3329, "step": 29055 }, { "epoch": 1.97, "learning_rate": 5.6325176275404404e-06, "loss": 0.6516, "step": 29060 }, { "epoch": 1.97, "learning_rate": 5.631574978319069e-06, "loss": 0.3696, "step": 29065 }, { "epoch": 1.97, "learning_rate": 5.630632329097696e-06, "loss": 0.6268, "step": 29070 }, { "epoch": 1.97, "learning_rate": 5.629689679876324e-06, "loss": 0.3452, "step": 29075 }, { "epoch": 1.97, "learning_rate": 5.628747030654953e-06, "loss": 0.385, "step": 29080 }, { "epoch": 1.97, "learning_rate": 5.627804381433581e-06, "loss": 0.5503, "step": 29085 }, { "epoch": 1.97, "learning_rate": 5.62686173221221e-06, "loss": 0.7274, "step": 29090 }, { "epoch": 1.97, "learning_rate": 5.625919082990838e-06, "loss": 0.7852, "step": 29095 }, { "epoch": 1.98, "learning_rate": 5.624976433769466e-06, "loss": 0.6286, "step": 29100 }, { "epoch": 1.98, "learning_rate": 5.624033784548094e-06, "loss": 0.4891, "step": 29105 }, { "epoch": 1.98, "learning_rate": 5.623091135326723e-06, "loss": 0.1451, "step": 29110 }, { "epoch": 1.98, "learning_rate": 5.622148486105351e-06, "loss": 0.9131, "step": 29115 }, { "epoch": 1.98, "learning_rate": 5.62120583688398e-06, "loss": 0.3076, "step": 29120 }, { "epoch": 1.98, "learning_rate": 5.620263187662608e-06, "loss": 0.4434, "step": 29125 }, { "epoch": 1.98, "learning_rate": 5.619320538441235e-06, "loss": 0.5422, "step": 29130 }, { "epoch": 1.98, "learning_rate": 5.6183778892198635e-06, "loss": 0.3581, "step": 29135 }, { "epoch": 1.98, "learning_rate": 5.617435239998492e-06, "loss": 0.2497, "step": 29140 }, { "epoch": 1.98, "learning_rate": 5.6164925907771205e-06, "loss": 0.7555, "step": 29145 }, { "epoch": 1.98, "learning_rate": 5.615549941555749e-06, "loss": 0.6422, "step": 29150 }, { "epoch": 1.98, "learning_rate": 5.6146072923343775e-06, "loss": 0.556, "step": 29155 }, { "epoch": 1.98, "learning_rate": 5.613664643113005e-06, "loss": 0.8316, "step": 29160 }, { "epoch": 1.98, "learning_rate": 5.612721993891634e-06, "loss": 0.27, "step": 29165 }, { "epoch": 1.98, "learning_rate": 5.611779344670262e-06, "loss": 0.4, "step": 29170 }, { "epoch": 1.98, "learning_rate": 5.6108366954488906e-06, "loss": 0.2412, "step": 29175 }, { "epoch": 1.98, "learning_rate": 5.609894046227519e-06, "loss": 0.3217, "step": 29180 }, { "epoch": 1.98, "learning_rate": 5.608951397006146e-06, "loss": 0.4716, "step": 29185 }, { "epoch": 1.98, "learning_rate": 5.608008747784774e-06, "loss": 0.4753, "step": 29190 }, { "epoch": 1.98, "learning_rate": 5.607066098563403e-06, "loss": 0.3405, "step": 29195 }, { "epoch": 1.98, "learning_rate": 5.606123449342031e-06, "loss": 0.2675, "step": 29200 }, { "epoch": 1.98, "learning_rate": 5.60518080012066e-06, "loss": 0.1977, "step": 29205 }, { "epoch": 1.98, "learning_rate": 5.604238150899288e-06, "loss": 0.7221, "step": 29210 }, { "epoch": 1.98, "learning_rate": 5.603295501677916e-06, "loss": 0.4123, "step": 29215 }, { "epoch": 1.98, "learning_rate": 5.602352852456544e-06, "loss": 0.388, "step": 29220 }, { "epoch": 1.98, "learning_rate": 5.601410203235173e-06, "loss": 0.2561, "step": 29225 }, { "epoch": 1.98, "learning_rate": 5.6004675540138005e-06, "loss": 0.2179, "step": 29230 }, { "epoch": 1.98, "learning_rate": 5.599524904792429e-06, "loss": 0.372, "step": 29235 }, { "epoch": 1.98, "learning_rate": 5.5985822555710575e-06, "loss": 0.7911, "step": 29240 }, { "epoch": 1.98, "learning_rate": 5.597639606349685e-06, "loss": 0.175, "step": 29245 }, { "epoch": 1.99, "learning_rate": 5.596696957128314e-06, "loss": 0.4967, "step": 29250 }, { "epoch": 1.99, "learning_rate": 5.595754307906942e-06, "loss": 0.1373, "step": 29255 }, { "epoch": 1.99, "learning_rate": 5.594811658685571e-06, "loss": 0.1968, "step": 29260 }, { "epoch": 1.99, "learning_rate": 5.593869009464199e-06, "loss": 0.3389, "step": 29265 }, { "epoch": 1.99, "learning_rate": 5.592926360242828e-06, "loss": 0.9544, "step": 29270 }, { "epoch": 1.99, "learning_rate": 5.591983711021454e-06, "loss": 0.323, "step": 29275 }, { "epoch": 1.99, "learning_rate": 5.591041061800083e-06, "loss": 0.1535, "step": 29280 }, { "epoch": 1.99, "learning_rate": 5.590098412578711e-06, "loss": 0.2818, "step": 29285 }, { "epoch": 1.99, "learning_rate": 5.58915576335734e-06, "loss": 0.6299, "step": 29290 }, { "epoch": 1.99, "learning_rate": 5.588213114135968e-06, "loss": 0.5517, "step": 29295 }, { "epoch": 1.99, "learning_rate": 5.587270464914596e-06, "loss": 0.8154, "step": 29300 }, { "epoch": 1.99, "learning_rate": 5.5863278156932245e-06, "loss": 0.1786, "step": 29305 }, { "epoch": 1.99, "learning_rate": 5.585385166471853e-06, "loss": 0.7436, "step": 29310 }, { "epoch": 1.99, "learning_rate": 5.5844425172504814e-06, "loss": 0.7508, "step": 29315 }, { "epoch": 1.99, "learning_rate": 5.58349986802911e-06, "loss": 1.0144, "step": 29320 }, { "epoch": 1.99, "learning_rate": 5.582557218807738e-06, "loss": 0.78, "step": 29325 }, { "epoch": 1.99, "learning_rate": 5.581614569586365e-06, "loss": 0.3043, "step": 29330 }, { "epoch": 1.99, "learning_rate": 5.580671920364994e-06, "loss": 0.4421, "step": 29335 }, { "epoch": 1.99, "learning_rate": 5.579729271143622e-06, "loss": 0.4774, "step": 29340 }, { "epoch": 1.99, "learning_rate": 5.578786621922251e-06, "loss": 0.3028, "step": 29345 }, { "epoch": 1.99, "learning_rate": 5.577843972700879e-06, "loss": 0.453, "step": 29350 }, { "epoch": 1.99, "learning_rate": 5.576901323479508e-06, "loss": 1.1441, "step": 29355 }, { "epoch": 1.99, "learning_rate": 5.575958674258135e-06, "loss": 0.3016, "step": 29360 }, { "epoch": 1.99, "learning_rate": 5.575016025036764e-06, "loss": 0.6229, "step": 29365 }, { "epoch": 1.99, "learning_rate": 5.574073375815392e-06, "loss": 0.3847, "step": 29370 }, { "epoch": 1.99, "learning_rate": 5.573130726594021e-06, "loss": 0.53, "step": 29375 }, { "epoch": 1.99, "learning_rate": 5.572188077372649e-06, "loss": 0.2705, "step": 29380 }, { "epoch": 1.99, "learning_rate": 5.571245428151278e-06, "loss": 0.5126, "step": 29385 }, { "epoch": 1.99, "learning_rate": 5.5703027789299045e-06, "loss": 0.603, "step": 29390 }, { "epoch": 2.0, "learning_rate": 5.569360129708533e-06, "loss": 0.3376, "step": 29395 }, { "epoch": 2.0, "learning_rate": 5.5684174804871615e-06, "loss": 0.4645, "step": 29400 }, { "epoch": 2.0, "learning_rate": 5.56747483126579e-06, "loss": 1.034, "step": 29405 }, { "epoch": 2.0, "learning_rate": 5.5665321820444185e-06, "loss": 0.2528, "step": 29410 }, { "epoch": 2.0, "learning_rate": 5.565589532823046e-06, "loss": 0.6928, "step": 29415 }, { "epoch": 2.0, "learning_rate": 5.564646883601675e-06, "loss": 0.3486, "step": 29420 }, { "epoch": 2.0, "learning_rate": 5.563704234380303e-06, "loss": 0.3523, "step": 29425 }, { "epoch": 2.0, "learning_rate": 5.5627615851589316e-06, "loss": 0.2581, "step": 29430 }, { "epoch": 2.0, "learning_rate": 5.561818935937559e-06, "loss": 1.0959, "step": 29435 }, { "epoch": 2.0, "learning_rate": 5.560876286716188e-06, "loss": 0.3929, "step": 29440 }, { "epoch": 2.0, "learning_rate": 5.559933637494815e-06, "loss": 0.2157, "step": 29445 }, { "epoch": 2.0, "learning_rate": 5.558990988273444e-06, "loss": 0.4301, "step": 29450 }, { "epoch": 2.0, "learning_rate": 5.558048339052072e-06, "loss": 0.4489, "step": 29455 }, { "epoch": 2.0, "learning_rate": 5.557105689830701e-06, "loss": 0.1768, "step": 29460 }, { "epoch": 2.0, "learning_rate": 5.556163040609329e-06, "loss": 1.1158, "step": 29465 }, { "epoch": 2.0, "learning_rate": 5.555220391387958e-06, "loss": 0.2996, "step": 29470 }, { "epoch": 2.0, "learning_rate": 5.554277742166585e-06, "loss": 0.3464, "step": 29475 }, { "epoch": 2.0, "learning_rate": 5.553335092945213e-06, "loss": 0.5099, "step": 29480 }, { "epoch": 2.0, "learning_rate": 5.5523924437238415e-06, "loss": 0.2633, "step": 29485 }, { "epoch": 2.0, "learning_rate": 5.55144979450247e-06, "loss": 0.0823, "step": 29490 }, { "epoch": 2.0, "learning_rate": 5.5505071452810985e-06, "loss": 0.2695, "step": 29495 }, { "epoch": 2.0, "learning_rate": 5.549564496059727e-06, "loss": 0.301, "step": 29500 }, { "epoch": 2.0, "learning_rate": 5.548621846838355e-06, "loss": 0.749, "step": 29505 }, { "epoch": 2.0, "learning_rate": 5.547679197616983e-06, "loss": 0.6345, "step": 29510 }, { "epoch": 2.0, "learning_rate": 5.546736548395612e-06, "loss": 0.2541, "step": 29515 }, { "epoch": 2.0, "learning_rate": 5.54579389917424e-06, "loss": 0.1712, "step": 29520 }, { "epoch": 2.0, "learning_rate": 5.544851249952869e-06, "loss": 0.105, "step": 29525 }, { "epoch": 2.0, "learning_rate": 5.543908600731495e-06, "loss": 0.2276, "step": 29530 }, { "epoch": 2.0, "learning_rate": 5.542965951510124e-06, "loss": 0.6777, "step": 29535 }, { "epoch": 2.0, "learning_rate": 5.542023302288752e-06, "loss": 0.5056, "step": 29540 }, { "epoch": 2.01, "learning_rate": 5.541080653067381e-06, "loss": 0.4542, "step": 29545 }, { "epoch": 2.01, "learning_rate": 5.540138003846009e-06, "loss": 0.1474, "step": 29550 }, { "epoch": 2.01, "learning_rate": 5.539195354624638e-06, "loss": 0.2467, "step": 29555 }, { "epoch": 2.01, "learning_rate": 5.5382527054032654e-06, "loss": 0.2999, "step": 29560 }, { "epoch": 2.01, "learning_rate": 5.537310056181894e-06, "loss": 0.4612, "step": 29565 }, { "epoch": 2.01, "learning_rate": 5.536367406960522e-06, "loss": 0.2673, "step": 29570 }, { "epoch": 2.01, "learning_rate": 5.535424757739151e-06, "loss": 0.6722, "step": 29575 }, { "epoch": 2.01, "learning_rate": 5.534482108517779e-06, "loss": 0.5454, "step": 29580 }, { "epoch": 2.01, "learning_rate": 5.533539459296408e-06, "loss": 0.3696, "step": 29585 }, { "epoch": 2.01, "learning_rate": 5.532596810075035e-06, "loss": 0.3181, "step": 29590 }, { "epoch": 2.01, "learning_rate": 5.531654160853663e-06, "loss": 0.8532, "step": 29595 }, { "epoch": 2.01, "learning_rate": 5.530711511632292e-06, "loss": 0.3366, "step": 29600 }, { "epoch": 2.01, "learning_rate": 5.52976886241092e-06, "loss": 0.2932, "step": 29605 }, { "epoch": 2.01, "learning_rate": 5.528826213189549e-06, "loss": 0.5021, "step": 29610 }, { "epoch": 2.01, "learning_rate": 5.527883563968177e-06, "loss": 0.5062, "step": 29615 }, { "epoch": 2.01, "learning_rate": 5.526940914746805e-06, "loss": 0.6556, "step": 29620 }, { "epoch": 2.01, "learning_rate": 5.525998265525433e-06, "loss": 0.3588, "step": 29625 }, { "epoch": 2.01, "learning_rate": 5.525055616304062e-06, "loss": 0.3402, "step": 29630 }, { "epoch": 2.01, "learning_rate": 5.52411296708269e-06, "loss": 0.2857, "step": 29635 }, { "epoch": 2.01, "learning_rate": 5.523170317861319e-06, "loss": 0.8358, "step": 29640 }, { "epoch": 2.01, "learning_rate": 5.5222276686399455e-06, "loss": 0.315, "step": 29645 }, { "epoch": 2.01, "learning_rate": 5.521285019418574e-06, "loss": 0.4784, "step": 29650 }, { "epoch": 2.01, "learning_rate": 5.5203423701972025e-06, "loss": 0.4517, "step": 29655 }, { "epoch": 2.01, "learning_rate": 5.519399720975831e-06, "loss": 0.3592, "step": 29660 }, { "epoch": 2.01, "learning_rate": 5.5184570717544594e-06, "loss": 0.3794, "step": 29665 }, { "epoch": 2.01, "learning_rate": 5.517514422533088e-06, "loss": 0.5659, "step": 29670 }, { "epoch": 2.01, "learning_rate": 5.5165717733117156e-06, "loss": 0.41, "step": 29675 }, { "epoch": 2.01, "learning_rate": 5.515629124090344e-06, "loss": 0.2494, "step": 29680 }, { "epoch": 2.01, "learning_rate": 5.5146864748689725e-06, "loss": 0.5743, "step": 29685 }, { "epoch": 2.02, "learning_rate": 5.5137438256476e-06, "loss": 0.2562, "step": 29690 }, { "epoch": 2.02, "learning_rate": 5.512801176426229e-06, "loss": 0.4024, "step": 29695 }, { "epoch": 2.02, "learning_rate": 5.511858527204857e-06, "loss": 0.0843, "step": 29700 }, { "epoch": 2.02, "learning_rate": 5.510915877983485e-06, "loss": 0.1413, "step": 29705 }, { "epoch": 2.02, "learning_rate": 5.509973228762113e-06, "loss": 0.5743, "step": 29710 }, { "epoch": 2.02, "learning_rate": 5.509030579540742e-06, "loss": 0.2167, "step": 29715 }, { "epoch": 2.02, "learning_rate": 5.50808793031937e-06, "loss": 0.2895, "step": 29720 }, { "epoch": 2.02, "learning_rate": 5.507145281097999e-06, "loss": 0.0774, "step": 29725 }, { "epoch": 2.02, "learning_rate": 5.506202631876627e-06, "loss": 0.0643, "step": 29730 }, { "epoch": 2.02, "learning_rate": 5.505259982655254e-06, "loss": 0.7191, "step": 29735 }, { "epoch": 2.02, "learning_rate": 5.5043173334338825e-06, "loss": 0.2202, "step": 29740 }, { "epoch": 2.02, "learning_rate": 5.503374684212511e-06, "loss": 0.2631, "step": 29745 }, { "epoch": 2.02, "learning_rate": 5.5024320349911395e-06, "loss": 0.2966, "step": 29750 }, { "epoch": 2.02, "learning_rate": 5.501489385769768e-06, "loss": 0.5004, "step": 29755 }, { "epoch": 2.02, "learning_rate": 5.500546736548396e-06, "loss": 0.5923, "step": 29760 }, { "epoch": 2.02, "learning_rate": 5.499604087327024e-06, "loss": 0.3399, "step": 29765 }, { "epoch": 2.02, "learning_rate": 5.498661438105653e-06, "loss": 0.5039, "step": 29770 }, { "epoch": 2.02, "learning_rate": 5.497718788884281e-06, "loss": 0.5177, "step": 29775 }, { "epoch": 2.02, "learning_rate": 5.4967761396629096e-06, "loss": 0.2066, "step": 29780 }, { "epoch": 2.02, "learning_rate": 5.495833490441538e-06, "loss": 0.1624, "step": 29785 }, { "epoch": 2.02, "learning_rate": 5.494890841220165e-06, "loss": 0.8919, "step": 29790 }, { "epoch": 2.02, "learning_rate": 5.493948191998793e-06, "loss": 0.581, "step": 29795 }, { "epoch": 2.02, "learning_rate": 5.493005542777422e-06, "loss": 0.1116, "step": 29800 }, { "epoch": 2.02, "learning_rate": 5.49206289355605e-06, "loss": 0.3106, "step": 29805 }, { "epoch": 2.02, "learning_rate": 5.491120244334679e-06, "loss": 0.4878, "step": 29810 }, { "epoch": 2.02, "learning_rate": 5.490177595113307e-06, "loss": 0.5373, "step": 29815 }, { "epoch": 2.02, "learning_rate": 5.489234945891935e-06, "loss": 0.3232, "step": 29820 }, { "epoch": 2.02, "learning_rate": 5.488292296670563e-06, "loss": 0.5546, "step": 29825 }, { "epoch": 2.02, "learning_rate": 5.487349647449192e-06, "loss": 0.5811, "step": 29830 }, { "epoch": 2.02, "learning_rate": 5.48640699822782e-06, "loss": 0.4095, "step": 29835 }, { "epoch": 2.03, "learning_rate": 5.485464349006449e-06, "loss": 0.2814, "step": 29840 }, { "epoch": 2.03, "learning_rate": 5.484521699785077e-06, "loss": 0.489, "step": 29845 }, { "epoch": 2.03, "learning_rate": 5.483579050563704e-06, "loss": 0.2532, "step": 29850 }, { "epoch": 2.03, "learning_rate": 5.482636401342333e-06, "loss": 0.4475, "step": 29855 }, { "epoch": 2.03, "learning_rate": 5.481693752120961e-06, "loss": 0.2589, "step": 29860 }, { "epoch": 2.03, "learning_rate": 5.48075110289959e-06, "loss": 0.81, "step": 29865 }, { "epoch": 2.03, "learning_rate": 5.479808453678218e-06, "loss": 0.2727, "step": 29870 }, { "epoch": 2.03, "learning_rate": 5.478865804456846e-06, "loss": 0.3982, "step": 29875 }, { "epoch": 2.03, "learning_rate": 5.477923155235474e-06, "loss": 0.3839, "step": 29880 }, { "epoch": 2.03, "learning_rate": 5.476980506014103e-06, "loss": 0.1059, "step": 29885 }, { "epoch": 2.03, "learning_rate": 5.476037856792731e-06, "loss": 0.4923, "step": 29890 }, { "epoch": 2.03, "learning_rate": 5.475095207571359e-06, "loss": 0.2795, "step": 29895 }, { "epoch": 2.03, "learning_rate": 5.474152558349987e-06, "loss": 0.5632, "step": 29900 }, { "epoch": 2.03, "learning_rate": 5.473209909128615e-06, "loss": 0.2031, "step": 29905 }, { "epoch": 2.03, "learning_rate": 5.4722672599072435e-06, "loss": 0.1997, "step": 29910 }, { "epoch": 2.03, "learning_rate": 5.471324610685872e-06, "loss": 0.0914, "step": 29915 }, { "epoch": 2.03, "learning_rate": 5.4703819614645004e-06, "loss": 0.4671, "step": 29920 }, { "epoch": 2.03, "learning_rate": 5.469439312243129e-06, "loss": 0.2806, "step": 29925 }, { "epoch": 2.03, "learning_rate": 5.468496663021757e-06, "loss": 0.323, "step": 29930 }, { "epoch": 2.03, "learning_rate": 5.467554013800385e-06, "loss": 0.4095, "step": 29935 }, { "epoch": 2.03, "learning_rate": 5.466611364579013e-06, "loss": 0.2175, "step": 29940 }, { "epoch": 2.03, "learning_rate": 5.465668715357641e-06, "loss": 0.4979, "step": 29945 }, { "epoch": 2.03, "learning_rate": 5.46472606613627e-06, "loss": 0.3257, "step": 29950 }, { "epoch": 2.03, "learning_rate": 5.463783416914898e-06, "loss": 0.1577, "step": 29955 }, { "epoch": 2.03, "learning_rate": 5.462840767693526e-06, "loss": 0.2263, "step": 29960 }, { "epoch": 2.03, "learning_rate": 5.461898118472154e-06, "loss": 0.4389, "step": 29965 }, { "epoch": 2.03, "learning_rate": 5.460955469250783e-06, "loss": 0.2794, "step": 29970 }, { "epoch": 2.03, "learning_rate": 5.460012820029411e-06, "loss": 0.7112, "step": 29975 }, { "epoch": 2.03, "learning_rate": 5.45907017080804e-06, "loss": 0.7832, "step": 29980 }, { "epoch": 2.04, "learning_rate": 5.458127521586668e-06, "loss": 0.5347, "step": 29985 }, { "epoch": 2.04, "learning_rate": 5.457184872365295e-06, "loss": 0.2314, "step": 29990 }, { "epoch": 2.04, "learning_rate": 5.4562422231439235e-06, "loss": 0.1333, "step": 29995 }, { "epoch": 2.04, "learning_rate": 5.455299573922552e-06, "loss": 0.1672, "step": 30000 }, { "epoch": 2.04, "learning_rate": 5.4543569247011805e-06, "loss": 0.3241, "step": 30005 }, { "epoch": 2.04, "learning_rate": 5.453414275479809e-06, "loss": 0.4328, "step": 30010 }, { "epoch": 2.04, "learning_rate": 5.4524716262584375e-06, "loss": 0.2248, "step": 30015 }, { "epoch": 2.04, "learning_rate": 5.451528977037065e-06, "loss": 0.5419, "step": 30020 }, { "epoch": 2.04, "learning_rate": 5.450586327815694e-06, "loss": 0.2094, "step": 30025 }, { "epoch": 2.04, "learning_rate": 5.449643678594322e-06, "loss": 0.5253, "step": 30030 }, { "epoch": 2.04, "learning_rate": 5.4487010293729506e-06, "loss": 0.5197, "step": 30035 }, { "epoch": 2.04, "learning_rate": 5.447758380151579e-06, "loss": 0.4735, "step": 30040 }, { "epoch": 2.04, "learning_rate": 5.4468157309302075e-06, "loss": 0.3698, "step": 30045 }, { "epoch": 2.04, "learning_rate": 5.445873081708834e-06, "loss": 0.4872, "step": 30050 }, { "epoch": 2.04, "learning_rate": 5.444930432487463e-06, "loss": 0.0907, "step": 30055 }, { "epoch": 2.04, "learning_rate": 5.443987783266091e-06, "loss": 0.1711, "step": 30060 }, { "epoch": 2.04, "learning_rate": 5.44304513404472e-06, "loss": 0.7335, "step": 30065 }, { "epoch": 2.04, "learning_rate": 5.442102484823348e-06, "loss": 0.31, "step": 30070 }, { "epoch": 2.04, "learning_rate": 5.441159835601976e-06, "loss": 0.6056, "step": 30075 }, { "epoch": 2.04, "learning_rate": 5.440217186380604e-06, "loss": 0.4064, "step": 30080 }, { "epoch": 2.04, "learning_rate": 5.439274537159233e-06, "loss": 0.5793, "step": 30085 }, { "epoch": 2.04, "learning_rate": 5.438331887937861e-06, "loss": 0.2161, "step": 30090 }, { "epoch": 2.04, "learning_rate": 5.43738923871649e-06, "loss": 0.249, "step": 30095 }, { "epoch": 2.04, "learning_rate": 5.436446589495118e-06, "loss": 0.3374, "step": 30100 }, { "epoch": 2.04, "learning_rate": 5.435503940273745e-06, "loss": 0.6263, "step": 30105 }, { "epoch": 2.04, "learning_rate": 5.434561291052374e-06, "loss": 1.0616, "step": 30110 }, { "epoch": 2.04, "learning_rate": 5.433618641831002e-06, "loss": 0.1132, "step": 30115 }, { "epoch": 2.04, "learning_rate": 5.432675992609631e-06, "loss": 0.6119, "step": 30120 }, { "epoch": 2.04, "learning_rate": 5.431733343388259e-06, "loss": 0.1674, "step": 30125 }, { "epoch": 2.04, "learning_rate": 5.430790694166888e-06, "loss": 0.34, "step": 30130 }, { "epoch": 2.05, "learning_rate": 5.429848044945515e-06, "loss": 0.6313, "step": 30135 }, { "epoch": 2.05, "learning_rate": 5.428905395724144e-06, "loss": 0.2612, "step": 30140 }, { "epoch": 2.05, "learning_rate": 5.427962746502771e-06, "loss": 0.2711, "step": 30145 }, { "epoch": 2.05, "learning_rate": 5.4270200972814e-06, "loss": 0.3286, "step": 30150 }, { "epoch": 2.05, "learning_rate": 5.426077448060028e-06, "loss": 0.2054, "step": 30155 }, { "epoch": 2.05, "learning_rate": 5.425134798838657e-06, "loss": 0.1956, "step": 30160 }, { "epoch": 2.05, "learning_rate": 5.4241921496172845e-06, "loss": 0.4355, "step": 30165 }, { "epoch": 2.05, "learning_rate": 5.423249500395913e-06, "loss": 0.2841, "step": 30170 }, { "epoch": 2.05, "learning_rate": 5.4223068511745414e-06, "loss": 0.3329, "step": 30175 }, { "epoch": 2.05, "learning_rate": 5.42136420195317e-06, "loss": 0.1931, "step": 30180 }, { "epoch": 2.05, "learning_rate": 5.420421552731798e-06, "loss": 0.4329, "step": 30185 }, { "epoch": 2.05, "learning_rate": 5.419478903510425e-06, "loss": 0.5999, "step": 30190 }, { "epoch": 2.05, "learning_rate": 5.418536254289054e-06, "loss": 0.0914, "step": 30195 }, { "epoch": 2.05, "learning_rate": 5.417593605067682e-06, "loss": 0.9429, "step": 30200 }, { "epoch": 2.05, "learning_rate": 5.416650955846311e-06, "loss": 0.5853, "step": 30205 }, { "epoch": 2.05, "learning_rate": 5.415708306624939e-06, "loss": 0.1236, "step": 30210 }, { "epoch": 2.05, "learning_rate": 5.414765657403568e-06, "loss": 0.2692, "step": 30215 }, { "epoch": 2.05, "learning_rate": 5.413823008182195e-06, "loss": 0.4872, "step": 30220 }, { "epoch": 2.05, "learning_rate": 5.412880358960824e-06, "loss": 0.5603, "step": 30225 }, { "epoch": 2.05, "learning_rate": 5.411937709739452e-06, "loss": 0.4814, "step": 30230 }, { "epoch": 2.05, "learning_rate": 5.410995060518081e-06, "loss": 0.2372, "step": 30235 }, { "epoch": 2.05, "learning_rate": 5.410052411296709e-06, "loss": 0.5769, "step": 30240 }, { "epoch": 2.05, "learning_rate": 5.409109762075338e-06, "loss": 0.3151, "step": 30245 }, { "epoch": 2.05, "learning_rate": 5.4081671128539645e-06, "loss": 0.0487, "step": 30250 }, { "epoch": 2.05, "learning_rate": 5.407224463632593e-06, "loss": 0.2293, "step": 30255 }, { "epoch": 2.05, "learning_rate": 5.4062818144112215e-06, "loss": 0.6499, "step": 30260 }, { "epoch": 2.05, "learning_rate": 5.40533916518985e-06, "loss": 0.2793, "step": 30265 }, { "epoch": 2.05, "learning_rate": 5.4043965159684785e-06, "loss": 0.5935, "step": 30270 }, { "epoch": 2.05, "learning_rate": 5.403453866747107e-06, "loss": 0.1803, "step": 30275 }, { "epoch": 2.06, "learning_rate": 5.402511217525735e-06, "loss": 0.5204, "step": 30280 }, { "epoch": 2.06, "learning_rate": 5.401568568304363e-06, "loss": 0.3548, "step": 30285 }, { "epoch": 2.06, "learning_rate": 5.4006259190829916e-06, "loss": 0.5655, "step": 30290 }, { "epoch": 2.06, "learning_rate": 5.39968326986162e-06, "loss": 0.4802, "step": 30295 }, { "epoch": 2.06, "learning_rate": 5.3987406206402485e-06, "loss": 0.21, "step": 30300 }, { "epoch": 2.06, "learning_rate": 5.397797971418875e-06, "loss": 0.0338, "step": 30305 }, { "epoch": 2.06, "learning_rate": 5.396855322197504e-06, "loss": 0.1222, "step": 30310 }, { "epoch": 2.06, "learning_rate": 5.395912672976132e-06, "loss": 0.2514, "step": 30315 }, { "epoch": 2.06, "learning_rate": 5.394970023754761e-06, "loss": 0.2508, "step": 30320 }, { "epoch": 2.06, "learning_rate": 5.394027374533389e-06, "loss": 0.1312, "step": 30325 }, { "epoch": 2.06, "learning_rate": 5.393084725312018e-06, "loss": 0.551, "step": 30330 }, { "epoch": 2.06, "learning_rate": 5.392142076090645e-06, "loss": 0.2195, "step": 30335 }, { "epoch": 2.06, "learning_rate": 5.391199426869274e-06, "loss": 0.9408, "step": 30340 }, { "epoch": 2.06, "learning_rate": 5.390256777647902e-06, "loss": 0.3447, "step": 30345 }, { "epoch": 2.06, "learning_rate": 5.389314128426531e-06, "loss": 0.1287, "step": 30350 }, { "epoch": 2.06, "learning_rate": 5.3883714792051585e-06, "loss": 0.864, "step": 30355 }, { "epoch": 2.06, "learning_rate": 5.387428829983787e-06, "loss": 0.5017, "step": 30360 }, { "epoch": 2.06, "learning_rate": 5.386486180762415e-06, "loss": 0.1856, "step": 30365 }, { "epoch": 2.06, "learning_rate": 5.385543531541043e-06, "loss": 0.4133, "step": 30370 }, { "epoch": 2.06, "learning_rate": 5.384600882319672e-06, "loss": 0.233, "step": 30375 }, { "epoch": 2.06, "learning_rate": 5.3836582330983e-06, "loss": 0.4409, "step": 30380 }, { "epoch": 2.06, "learning_rate": 5.382715583876929e-06, "loss": 0.1992, "step": 30385 }, { "epoch": 2.06, "learning_rate": 5.381772934655557e-06, "loss": 0.0567, "step": 30390 }, { "epoch": 2.06, "learning_rate": 5.380830285434184e-06, "loss": 0.6952, "step": 30395 }, { "epoch": 2.06, "learning_rate": 5.379887636212812e-06, "loss": 0.3032, "step": 30400 }, { "epoch": 2.06, "learning_rate": 5.378944986991441e-06, "loss": 0.537, "step": 30405 }, { "epoch": 2.06, "learning_rate": 5.378002337770069e-06, "loss": 0.2762, "step": 30410 }, { "epoch": 2.06, "learning_rate": 5.377059688548698e-06, "loss": 0.2759, "step": 30415 }, { "epoch": 2.06, "learning_rate": 5.3761170393273255e-06, "loss": 0.1109, "step": 30420 }, { "epoch": 2.06, "learning_rate": 5.375174390105954e-06, "loss": 0.3417, "step": 30425 }, { "epoch": 2.07, "learning_rate": 5.3742317408845824e-06, "loss": 0.255, "step": 30430 }, { "epoch": 2.07, "learning_rate": 5.373289091663211e-06, "loss": 0.3163, "step": 30435 }, { "epoch": 2.07, "learning_rate": 5.372346442441839e-06, "loss": 0.244, "step": 30440 }, { "epoch": 2.07, "learning_rate": 5.371403793220468e-06, "loss": 0.205, "step": 30445 }, { "epoch": 2.07, "learning_rate": 5.370461143999095e-06, "loss": 0.2972, "step": 30450 }, { "epoch": 2.07, "learning_rate": 5.369518494777723e-06, "loss": 0.143, "step": 30455 }, { "epoch": 2.07, "learning_rate": 5.368575845556352e-06, "loss": 0.2695, "step": 30460 }, { "epoch": 2.07, "learning_rate": 5.36763319633498e-06, "loss": 0.5588, "step": 30465 }, { "epoch": 2.07, "learning_rate": 5.366690547113609e-06, "loss": 0.2903, "step": 30470 }, { "epoch": 2.07, "learning_rate": 5.365747897892237e-06, "loss": 0.5516, "step": 30475 }, { "epoch": 2.07, "learning_rate": 5.364805248670865e-06, "loss": 0.5069, "step": 30480 }, { "epoch": 2.07, "learning_rate": 5.363862599449493e-06, "loss": 0.3675, "step": 30485 }, { "epoch": 2.07, "learning_rate": 5.362919950228122e-06, "loss": 0.9551, "step": 30490 }, { "epoch": 2.07, "learning_rate": 5.36197730100675e-06, "loss": 0.3767, "step": 30495 }, { "epoch": 2.07, "learning_rate": 5.361034651785379e-06, "loss": 0.3767, "step": 30500 }, { "epoch": 2.07, "learning_rate": 5.360092002564007e-06, "loss": 0.292, "step": 30505 }, { "epoch": 2.07, "learning_rate": 5.359149353342634e-06, "loss": 0.1538, "step": 30510 }, { "epoch": 2.07, "learning_rate": 5.3582067041212625e-06, "loss": 0.2246, "step": 30515 }, { "epoch": 2.07, "learning_rate": 5.357264054899891e-06, "loss": 0.2365, "step": 30520 }, { "epoch": 2.07, "learning_rate": 5.3563214056785195e-06, "loss": 0.4902, "step": 30525 }, { "epoch": 2.07, "learning_rate": 5.355378756457148e-06, "loss": 0.2767, "step": 30530 }, { "epoch": 2.07, "learning_rate": 5.354436107235776e-06, "loss": 0.1317, "step": 30535 }, { "epoch": 2.07, "learning_rate": 5.353493458014404e-06, "loss": 0.1876, "step": 30540 }, { "epoch": 2.07, "learning_rate": 5.3525508087930326e-06, "loss": 0.5293, "step": 30545 }, { "epoch": 2.07, "learning_rate": 5.351608159571661e-06, "loss": 0.2906, "step": 30550 }, { "epoch": 2.07, "learning_rate": 5.3506655103502895e-06, "loss": 0.2603, "step": 30555 }, { "epoch": 2.07, "learning_rate": 5.349722861128917e-06, "loss": 0.4959, "step": 30560 }, { "epoch": 2.07, "learning_rate": 5.348780211907545e-06, "loss": 0.1542, "step": 30565 }, { "epoch": 2.07, "learning_rate": 5.347837562686173e-06, "loss": 0.5161, "step": 30570 }, { "epoch": 2.08, "learning_rate": 5.346894913464802e-06, "loss": 0.1879, "step": 30575 }, { "epoch": 2.08, "learning_rate": 5.34595226424343e-06, "loss": 0.0572, "step": 30580 }, { "epoch": 2.08, "learning_rate": 5.345009615022059e-06, "loss": 0.2322, "step": 30585 }, { "epoch": 2.08, "learning_rate": 5.344066965800687e-06, "loss": 0.163, "step": 30590 }, { "epoch": 2.08, "learning_rate": 5.343124316579315e-06, "loss": 0.4312, "step": 30595 }, { "epoch": 2.08, "learning_rate": 5.342181667357943e-06, "loss": 0.5335, "step": 30600 }, { "epoch": 2.08, "learning_rate": 5.341239018136571e-06, "loss": 0.3298, "step": 30605 }, { "epoch": 2.08, "learning_rate": 5.3402963689151995e-06, "loss": 0.5247, "step": 30610 }, { "epoch": 2.08, "learning_rate": 5.339353719693828e-06, "loss": 0.3154, "step": 30615 }, { "epoch": 2.08, "learning_rate": 5.3384110704724565e-06, "loss": 0.6789, "step": 30620 }, { "epoch": 2.08, "learning_rate": 5.337468421251084e-06, "loss": 0.6056, "step": 30625 }, { "epoch": 2.08, "learning_rate": 5.336525772029713e-06, "loss": 0.1582, "step": 30630 }, { "epoch": 2.08, "learning_rate": 5.335583122808341e-06, "loss": 0.0914, "step": 30635 }, { "epoch": 2.08, "learning_rate": 5.33464047358697e-06, "loss": 0.4679, "step": 30640 }, { "epoch": 2.08, "learning_rate": 5.333697824365598e-06, "loss": 0.2996, "step": 30645 }, { "epoch": 2.08, "learning_rate": 5.332755175144225e-06, "loss": 0.4019, "step": 30650 }, { "epoch": 2.08, "learning_rate": 5.331812525922853e-06, "loss": 0.4033, "step": 30655 }, { "epoch": 2.08, "learning_rate": 5.330869876701482e-06, "loss": 0.2648, "step": 30660 }, { "epoch": 2.08, "learning_rate": 5.32992722748011e-06, "loss": 0.2041, "step": 30665 }, { "epoch": 2.08, "learning_rate": 5.328984578258739e-06, "loss": 0.7168, "step": 30670 }, { "epoch": 2.08, "learning_rate": 5.328041929037367e-06, "loss": 0.1999, "step": 30675 }, { "epoch": 2.08, "learning_rate": 5.327099279815995e-06, "loss": 0.1133, "step": 30680 }, { "epoch": 2.08, "learning_rate": 5.326156630594623e-06, "loss": 0.3526, "step": 30685 }, { "epoch": 2.08, "learning_rate": 5.325213981373252e-06, "loss": 0.132, "step": 30690 }, { "epoch": 2.08, "learning_rate": 5.32427133215188e-06, "loss": 0.3914, "step": 30695 }, { "epoch": 2.08, "learning_rate": 5.323328682930509e-06, "loss": 0.0651, "step": 30700 }, { "epoch": 2.08, "learning_rate": 5.322386033709137e-06, "loss": 0.2083, "step": 30705 }, { "epoch": 2.08, "learning_rate": 5.321443384487764e-06, "loss": 0.5316, "step": 30710 }, { "epoch": 2.08, "learning_rate": 5.320500735266393e-06, "loss": 0.4792, "step": 30715 }, { "epoch": 2.08, "learning_rate": 5.319558086045021e-06, "loss": 0.136, "step": 30720 }, { "epoch": 2.09, "learning_rate": 5.31861543682365e-06, "loss": 0.3658, "step": 30725 }, { "epoch": 2.09, "learning_rate": 5.317672787602278e-06, "loss": 0.4763, "step": 30730 }, { "epoch": 2.09, "learning_rate": 5.316730138380907e-06, "loss": 0.5411, "step": 30735 }, { "epoch": 2.09, "learning_rate": 5.315787489159534e-06, "loss": 0.2177, "step": 30740 }, { "epoch": 2.09, "learning_rate": 5.314844839938163e-06, "loss": 0.1153, "step": 30745 }, { "epoch": 2.09, "learning_rate": 5.313902190716791e-06, "loss": 0.2693, "step": 30750 }, { "epoch": 2.09, "learning_rate": 5.31295954149542e-06, "loss": 0.4142, "step": 30755 }, { "epoch": 2.09, "learning_rate": 5.312016892274048e-06, "loss": 0.1469, "step": 30760 }, { "epoch": 2.09, "learning_rate": 5.311074243052675e-06, "loss": 0.3364, "step": 30765 }, { "epoch": 2.09, "learning_rate": 5.3101315938313035e-06, "loss": 0.7296, "step": 30770 }, { "epoch": 2.09, "learning_rate": 5.309188944609932e-06, "loss": 0.5527, "step": 30775 }, { "epoch": 2.09, "learning_rate": 5.3082462953885604e-06, "loss": 0.6795, "step": 30780 }, { "epoch": 2.09, "learning_rate": 5.307303646167189e-06, "loss": 0.0951, "step": 30785 }, { "epoch": 2.09, "learning_rate": 5.306360996945817e-06, "loss": 0.3422, "step": 30790 }, { "epoch": 2.09, "learning_rate": 5.305418347724445e-06, "loss": 0.7025, "step": 30795 }, { "epoch": 2.09, "learning_rate": 5.3044756985030735e-06, "loss": 0.2414, "step": 30800 }, { "epoch": 2.09, "learning_rate": 5.303533049281702e-06, "loss": 0.4909, "step": 30805 }, { "epoch": 2.09, "learning_rate": 5.3025904000603305e-06, "loss": 0.3199, "step": 30810 }, { "epoch": 2.09, "learning_rate": 5.301647750838958e-06, "loss": 0.0827, "step": 30815 }, { "epoch": 2.09, "learning_rate": 5.300705101617587e-06, "loss": 0.313, "step": 30820 }, { "epoch": 2.09, "learning_rate": 5.299762452396214e-06, "loss": 0.1831, "step": 30825 }, { "epoch": 2.09, "learning_rate": 5.298819803174843e-06, "loss": 0.154, "step": 30830 }, { "epoch": 2.09, "learning_rate": 5.297877153953471e-06, "loss": 0.8138, "step": 30835 }, { "epoch": 2.09, "learning_rate": 5.2969345047321e-06, "loss": 0.4769, "step": 30840 }, { "epoch": 2.09, "learning_rate": 5.295991855510728e-06, "loss": 0.1915, "step": 30845 }, { "epoch": 2.09, "learning_rate": 5.295049206289357e-06, "loss": 0.3151, "step": 30850 }, { "epoch": 2.09, "learning_rate": 5.2941065570679835e-06, "loss": 0.5616, "step": 30855 }, { "epoch": 2.09, "learning_rate": 5.293163907846612e-06, "loss": 0.3095, "step": 30860 }, { "epoch": 2.09, "learning_rate": 5.2922212586252405e-06, "loss": 0.5989, "step": 30865 }, { "epoch": 2.1, "learning_rate": 5.291278609403869e-06, "loss": 0.479, "step": 30870 }, { "epoch": 2.1, "learning_rate": 5.2903359601824975e-06, "loss": 0.1541, "step": 30875 }, { "epoch": 2.1, "learning_rate": 5.289393310961125e-06, "loss": 0.3597, "step": 30880 }, { "epoch": 2.1, "learning_rate": 5.288450661739754e-06, "loss": 0.3015, "step": 30885 }, { "epoch": 2.1, "learning_rate": 5.287508012518382e-06, "loss": 0.1118, "step": 30890 }, { "epoch": 2.1, "learning_rate": 5.2865653632970106e-06, "loss": 0.3926, "step": 30895 }, { "epoch": 2.1, "learning_rate": 5.285622714075639e-06, "loss": 0.3775, "step": 30900 }, { "epoch": 2.1, "learning_rate": 5.2846800648542675e-06, "loss": 0.4084, "step": 30905 }, { "epoch": 2.1, "learning_rate": 5.283737415632894e-06, "loss": 0.9365, "step": 30910 }, { "epoch": 2.1, "learning_rate": 5.282794766411523e-06, "loss": 0.4162, "step": 30915 }, { "epoch": 2.1, "learning_rate": 5.281852117190151e-06, "loss": 0.4439, "step": 30920 }, { "epoch": 2.1, "learning_rate": 5.28090946796878e-06, "loss": 0.1325, "step": 30925 }, { "epoch": 2.1, "learning_rate": 5.279966818747408e-06, "loss": 0.352, "step": 30930 }, { "epoch": 2.1, "learning_rate": 5.279024169526037e-06, "loss": 0.887, "step": 30935 }, { "epoch": 2.1, "learning_rate": 5.278081520304664e-06, "loss": 0.0521, "step": 30940 }, { "epoch": 2.1, "learning_rate": 5.277138871083293e-06, "loss": 0.3107, "step": 30945 }, { "epoch": 2.1, "learning_rate": 5.276196221861921e-06, "loss": 0.3394, "step": 30950 }, { "epoch": 2.1, "learning_rate": 5.27525357264055e-06, "loss": 0.406, "step": 30955 }, { "epoch": 2.1, "learning_rate": 5.274310923419178e-06, "loss": 0.5161, "step": 30960 }, { "epoch": 2.1, "learning_rate": 5.273368274197805e-06, "loss": 0.4271, "step": 30965 }, { "epoch": 2.1, "learning_rate": 5.272425624976434e-06, "loss": 0.4, "step": 30970 }, { "epoch": 2.1, "learning_rate": 5.271482975755062e-06, "loss": 0.1363, "step": 30975 }, { "epoch": 2.1, "learning_rate": 5.270540326533691e-06, "loss": 0.2492, "step": 30980 }, { "epoch": 2.1, "learning_rate": 5.269597677312319e-06, "loss": 0.1961, "step": 30985 }, { "epoch": 2.1, "learning_rate": 5.268655028090948e-06, "loss": 0.3317, "step": 30990 }, { "epoch": 2.1, "learning_rate": 5.267712378869575e-06, "loss": 0.3128, "step": 30995 }, { "epoch": 2.1, "learning_rate": 5.266769729648204e-06, "loss": 0.2243, "step": 31000 }, { "epoch": 2.1, "learning_rate": 5.265827080426832e-06, "loss": 0.2352, "step": 31005 }, { "epoch": 2.1, "learning_rate": 5.264884431205461e-06, "loss": 0.1003, "step": 31010 }, { "epoch": 2.1, "learning_rate": 5.263941781984089e-06, "loss": 0.2393, "step": 31015 }, { "epoch": 2.11, "learning_rate": 5.262999132762717e-06, "loss": 0.1071, "step": 31020 }, { "epoch": 2.11, "learning_rate": 5.2620564835413445e-06, "loss": 0.6006, "step": 31025 }, { "epoch": 2.11, "learning_rate": 5.261113834319973e-06, "loss": 0.1672, "step": 31030 }, { "epoch": 2.11, "learning_rate": 5.2601711850986014e-06, "loss": 0.0744, "step": 31035 }, { "epoch": 2.11, "learning_rate": 5.25922853587723e-06, "loss": 0.3331, "step": 31040 }, { "epoch": 2.11, "learning_rate": 5.258285886655858e-06, "loss": 0.4694, "step": 31045 }, { "epoch": 2.11, "learning_rate": 5.257343237434487e-06, "loss": 0.2722, "step": 31050 }, { "epoch": 2.11, "learning_rate": 5.2564005882131145e-06, "loss": 0.2666, "step": 31055 }, { "epoch": 2.11, "learning_rate": 5.255457938991743e-06, "loss": 0.7259, "step": 31060 }, { "epoch": 2.11, "learning_rate": 5.254515289770371e-06, "loss": 0.2419, "step": 31065 }, { "epoch": 2.11, "learning_rate": 5.253572640548999e-06, "loss": 0.1411, "step": 31070 }, { "epoch": 2.11, "learning_rate": 5.252629991327628e-06, "loss": 0.3178, "step": 31075 }, { "epoch": 2.11, "learning_rate": 5.251687342106255e-06, "loss": 0.5122, "step": 31080 }, { "epoch": 2.11, "learning_rate": 5.250744692884884e-06, "loss": 0.6185, "step": 31085 }, { "epoch": 2.11, "learning_rate": 5.249802043663512e-06, "loss": 0.4163, "step": 31090 }, { "epoch": 2.11, "learning_rate": 5.248859394442141e-06, "loss": 0.7635, "step": 31095 }, { "epoch": 2.11, "learning_rate": 5.247916745220769e-06, "loss": 0.2449, "step": 31100 }, { "epoch": 2.11, "learning_rate": 5.246974095999398e-06, "loss": 0.1701, "step": 31105 }, { "epoch": 2.11, "learning_rate": 5.2460314467780245e-06, "loss": 0.593, "step": 31110 }, { "epoch": 2.11, "learning_rate": 5.245088797556653e-06, "loss": 0.3141, "step": 31115 }, { "epoch": 2.11, "learning_rate": 5.2441461483352815e-06, "loss": 0.2179, "step": 31120 }, { "epoch": 2.11, "learning_rate": 5.24320349911391e-06, "loss": 0.222, "step": 31125 }, { "epoch": 2.11, "learning_rate": 5.2422608498925385e-06, "loss": 0.7221, "step": 31130 }, { "epoch": 2.11, "learning_rate": 5.241318200671167e-06, "loss": 0.2256, "step": 31135 }, { "epoch": 2.11, "learning_rate": 5.240375551449795e-06, "loss": 0.1459, "step": 31140 }, { "epoch": 2.11, "learning_rate": 5.239432902228423e-06, "loss": 0.4087, "step": 31145 }, { "epoch": 2.11, "learning_rate": 5.2384902530070516e-06, "loss": 0.2322, "step": 31150 }, { "epoch": 2.11, "learning_rate": 5.23754760378568e-06, "loss": 0.3597, "step": 31155 }, { "epoch": 2.11, "learning_rate": 5.2366049545643085e-06, "loss": 0.401, "step": 31160 }, { "epoch": 2.12, "learning_rate": 5.235662305342937e-06, "loss": 0.7286, "step": 31165 }, { "epoch": 2.12, "learning_rate": 5.234719656121564e-06, "loss": 0.558, "step": 31170 }, { "epoch": 2.12, "learning_rate": 5.233777006900192e-06, "loss": 0.531, "step": 31175 }, { "epoch": 2.12, "learning_rate": 5.232834357678821e-06, "loss": 0.386, "step": 31180 }, { "epoch": 2.12, "learning_rate": 5.231891708457449e-06, "loss": 0.2674, "step": 31185 }, { "epoch": 2.12, "learning_rate": 5.230949059236078e-06, "loss": 0.1115, "step": 31190 }, { "epoch": 2.12, "learning_rate": 5.230006410014705e-06, "loss": 0.2001, "step": 31195 }, { "epoch": 2.12, "learning_rate": 5.229063760793334e-06, "loss": 0.4837, "step": 31200 }, { "epoch": 2.12, "learning_rate": 5.228121111571962e-06, "loss": 0.1645, "step": 31205 }, { "epoch": 2.12, "learning_rate": 5.227178462350591e-06, "loss": 0.7235, "step": 31210 }, { "epoch": 2.12, "learning_rate": 5.226235813129219e-06, "loss": 0.3718, "step": 31215 }, { "epoch": 2.12, "learning_rate": 5.225293163907848e-06, "loss": 0.2253, "step": 31220 }, { "epoch": 2.12, "learning_rate": 5.224350514686475e-06, "loss": 0.575, "step": 31225 }, { "epoch": 2.12, "learning_rate": 5.223407865465103e-06, "loss": 0.2408, "step": 31230 }, { "epoch": 2.12, "learning_rate": 5.222465216243732e-06, "loss": 0.1297, "step": 31235 }, { "epoch": 2.12, "learning_rate": 5.22152256702236e-06, "loss": 0.4595, "step": 31240 }, { "epoch": 2.12, "learning_rate": 5.220579917800989e-06, "loss": 0.6153, "step": 31245 }, { "epoch": 2.12, "learning_rate": 5.219637268579617e-06, "loss": 0.4951, "step": 31250 }, { "epoch": 2.12, "learning_rate": 5.218694619358245e-06, "loss": 0.3792, "step": 31255 }, { "epoch": 2.12, "learning_rate": 5.217751970136873e-06, "loss": 0.5987, "step": 31260 }, { "epoch": 2.12, "learning_rate": 5.216809320915502e-06, "loss": 0.3019, "step": 31265 }, { "epoch": 2.12, "learning_rate": 5.215866671694129e-06, "loss": 0.1478, "step": 31270 }, { "epoch": 2.12, "learning_rate": 5.214924022472758e-06, "loss": 1.2301, "step": 31275 }, { "epoch": 2.12, "learning_rate": 5.213981373251386e-06, "loss": 0.5821, "step": 31280 }, { "epoch": 2.12, "learning_rate": 5.213038724030014e-06, "loss": 0.2409, "step": 31285 }, { "epoch": 2.12, "learning_rate": 5.2120960748086424e-06, "loss": 0.4378, "step": 31290 }, { "epoch": 2.12, "learning_rate": 5.211153425587271e-06, "loss": 0.1338, "step": 31295 }, { "epoch": 2.12, "learning_rate": 5.210210776365899e-06, "loss": 0.112, "step": 31300 }, { "epoch": 2.12, "learning_rate": 5.209268127144528e-06, "loss": 0.3581, "step": 31305 }, { "epoch": 2.13, "learning_rate": 5.2083254779231555e-06, "loss": 0.4028, "step": 31310 }, { "epoch": 2.13, "learning_rate": 5.207382828701783e-06, "loss": 0.321, "step": 31315 }, { "epoch": 2.13, "learning_rate": 5.206440179480412e-06, "loss": 0.426, "step": 31320 }, { "epoch": 2.13, "learning_rate": 5.20549753025904e-06, "loss": 0.1308, "step": 31325 }, { "epoch": 2.13, "learning_rate": 5.204554881037669e-06, "loss": 0.4277, "step": 31330 }, { "epoch": 2.13, "learning_rate": 5.203612231816297e-06, "loss": 0.4226, "step": 31335 }, { "epoch": 2.13, "learning_rate": 5.202669582594925e-06, "loss": 0.2404, "step": 31340 }, { "epoch": 2.13, "learning_rate": 5.201726933373553e-06, "loss": 0.4169, "step": 31345 }, { "epoch": 2.13, "learning_rate": 5.200784284152182e-06, "loss": 0.6414, "step": 31350 }, { "epoch": 2.13, "learning_rate": 5.19984163493081e-06, "loss": 0.4765, "step": 31355 }, { "epoch": 2.13, "learning_rate": 5.198898985709439e-06, "loss": 0.236, "step": 31360 }, { "epoch": 2.13, "learning_rate": 5.197956336488067e-06, "loss": 0.3439, "step": 31365 }, { "epoch": 2.13, "learning_rate": 5.197013687266694e-06, "loss": 0.7234, "step": 31370 }, { "epoch": 2.13, "learning_rate": 5.1960710380453225e-06, "loss": 0.1462, "step": 31375 }, { "epoch": 2.13, "learning_rate": 5.195128388823951e-06, "loss": 0.857, "step": 31380 }, { "epoch": 2.13, "learning_rate": 5.1941857396025795e-06, "loss": 0.4299, "step": 31385 }, { "epoch": 2.13, "learning_rate": 5.193243090381208e-06, "loss": 0.4046, "step": 31390 }, { "epoch": 2.13, "learning_rate": 5.1923004411598364e-06, "loss": 0.3911, "step": 31395 }, { "epoch": 2.13, "learning_rate": 5.191357791938464e-06, "loss": 0.1106, "step": 31400 }, { "epoch": 2.13, "learning_rate": 5.1904151427170926e-06, "loss": 0.2245, "step": 31405 }, { "epoch": 2.13, "learning_rate": 5.189472493495721e-06, "loss": 0.5903, "step": 31410 }, { "epoch": 2.13, "learning_rate": 5.1885298442743495e-06, "loss": 0.5678, "step": 31415 }, { "epoch": 2.13, "learning_rate": 5.187587195052978e-06, "loss": 0.1915, "step": 31420 }, { "epoch": 2.13, "learning_rate": 5.186644545831605e-06, "loss": 0.2926, "step": 31425 }, { "epoch": 2.13, "learning_rate": 5.185701896610233e-06, "loss": 0.3663, "step": 31430 }, { "epoch": 2.13, "learning_rate": 5.184759247388862e-06, "loss": 0.5848, "step": 31435 }, { "epoch": 2.13, "learning_rate": 5.18381659816749e-06, "loss": 0.1276, "step": 31440 }, { "epoch": 2.13, "learning_rate": 5.182873948946119e-06, "loss": 0.3842, "step": 31445 }, { "epoch": 2.13, "learning_rate": 5.181931299724747e-06, "loss": 0.3697, "step": 31450 }, { "epoch": 2.13, "learning_rate": 5.180988650503375e-06, "loss": 0.3716, "step": 31455 }, { "epoch": 2.14, "learning_rate": 5.180046001282003e-06, "loss": 0.3129, "step": 31460 }, { "epoch": 2.14, "learning_rate": 5.179103352060632e-06, "loss": 0.3419, "step": 31465 }, { "epoch": 2.14, "learning_rate": 5.17816070283926e-06, "loss": 0.307, "step": 31470 }, { "epoch": 2.14, "learning_rate": 5.177218053617889e-06, "loss": 0.4186, "step": 31475 }, { "epoch": 2.14, "learning_rate": 5.1762754043965165e-06, "loss": 0.4639, "step": 31480 }, { "epoch": 2.14, "learning_rate": 5.175332755175144e-06, "loss": 0.1886, "step": 31485 }, { "epoch": 2.14, "learning_rate": 5.174390105953773e-06, "loss": 0.7364, "step": 31490 }, { "epoch": 2.14, "learning_rate": 5.173447456732401e-06, "loss": 0.345, "step": 31495 }, { "epoch": 2.14, "learning_rate": 5.17250480751103e-06, "loss": 0.1131, "step": 31500 }, { "epoch": 2.14, "learning_rate": 5.171562158289658e-06, "loss": 0.48, "step": 31505 }, { "epoch": 2.14, "learning_rate": 5.1706195090682866e-06, "loss": 0.1848, "step": 31510 }, { "epoch": 2.14, "learning_rate": 5.169676859846914e-06, "loss": 0.6722, "step": 31515 }, { "epoch": 2.14, "learning_rate": 5.168734210625542e-06, "loss": 0.8093, "step": 31520 }, { "epoch": 2.14, "learning_rate": 5.16779156140417e-06, "loss": 0.687, "step": 31525 }, { "epoch": 2.14, "learning_rate": 5.166848912182799e-06, "loss": 0.3764, "step": 31530 }, { "epoch": 2.14, "learning_rate": 5.165906262961427e-06, "loss": 0.8042, "step": 31535 }, { "epoch": 2.14, "learning_rate": 5.164963613740055e-06, "loss": 0.3834, "step": 31540 }, { "epoch": 2.14, "learning_rate": 5.1640209645186834e-06, "loss": 0.4815, "step": 31545 }, { "epoch": 2.14, "learning_rate": 5.163078315297312e-06, "loss": 0.6582, "step": 31550 }, { "epoch": 2.14, "learning_rate": 5.16213566607594e-06, "loss": 0.2576, "step": 31555 }, { "epoch": 2.14, "learning_rate": 5.161193016854569e-06, "loss": 0.3292, "step": 31560 }, { "epoch": 2.14, "learning_rate": 5.160250367633197e-06, "loss": 0.9476, "step": 31565 }, { "epoch": 2.14, "learning_rate": 5.159307718411824e-06, "loss": 0.2258, "step": 31570 }, { "epoch": 2.14, "learning_rate": 5.158365069190453e-06, "loss": 0.2719, "step": 31575 }, { "epoch": 2.14, "learning_rate": 5.157422419969081e-06, "loss": 0.5753, "step": 31580 }, { "epoch": 2.14, "learning_rate": 5.15647977074771e-06, "loss": 0.1719, "step": 31585 }, { "epoch": 2.14, "learning_rate": 5.155537121526338e-06, "loss": 0.3168, "step": 31590 }, { "epoch": 2.14, "learning_rate": 5.154594472304967e-06, "loss": 0.558, "step": 31595 }, { "epoch": 2.14, "learning_rate": 5.153651823083594e-06, "loss": 0.2141, "step": 31600 }, { "epoch": 2.15, "learning_rate": 5.152709173862223e-06, "loss": 0.1676, "step": 31605 }, { "epoch": 2.15, "learning_rate": 5.151766524640851e-06, "loss": 0.082, "step": 31610 }, { "epoch": 2.15, "learning_rate": 5.15082387541948e-06, "loss": 0.2544, "step": 31615 }, { "epoch": 2.15, "learning_rate": 5.149881226198108e-06, "loss": 0.7685, "step": 31620 }, { "epoch": 2.15, "learning_rate": 5.148938576976737e-06, "loss": 0.4517, "step": 31625 }, { "epoch": 2.15, "learning_rate": 5.1479959277553635e-06, "loss": 0.5286, "step": 31630 }, { "epoch": 2.15, "learning_rate": 5.147053278533992e-06, "loss": 0.1438, "step": 31635 }, { "epoch": 2.15, "learning_rate": 5.1461106293126205e-06, "loss": 0.68, "step": 31640 }, { "epoch": 2.15, "learning_rate": 5.145167980091249e-06, "loss": 0.1998, "step": 31645 }, { "epoch": 2.15, "learning_rate": 5.1442253308698774e-06, "loss": 0.5046, "step": 31650 }, { "epoch": 2.15, "learning_rate": 5.143282681648505e-06, "loss": 0.326, "step": 31655 }, { "epoch": 2.15, "learning_rate": 5.1423400324271336e-06, "loss": 0.2027, "step": 31660 }, { "epoch": 2.15, "learning_rate": 5.141397383205762e-06, "loss": 0.0758, "step": 31665 }, { "epoch": 2.15, "learning_rate": 5.1404547339843905e-06, "loss": 0.1859, "step": 31670 }, { "epoch": 2.15, "learning_rate": 5.139512084763019e-06, "loss": 0.3653, "step": 31675 }, { "epoch": 2.15, "learning_rate": 5.1385694355416475e-06, "loss": 0.0543, "step": 31680 }, { "epoch": 2.15, "learning_rate": 5.137626786320274e-06, "loss": 0.2528, "step": 31685 }, { "epoch": 2.15, "learning_rate": 5.136684137098903e-06, "loss": 0.3958, "step": 31690 }, { "epoch": 2.15, "learning_rate": 5.135741487877531e-06, "loss": 0.1332, "step": 31695 }, { "epoch": 2.15, "learning_rate": 5.13479883865616e-06, "loss": 0.3, "step": 31700 }, { "epoch": 2.15, "learning_rate": 5.133856189434788e-06, "loss": 0.1381, "step": 31705 }, { "epoch": 2.15, "learning_rate": 5.132913540213417e-06, "loss": 0.5364, "step": 31710 }, { "epoch": 2.15, "learning_rate": 5.131970890992044e-06, "loss": 0.3175, "step": 31715 }, { "epoch": 2.15, "learning_rate": 5.131028241770673e-06, "loss": 0.2254, "step": 31720 }, { "epoch": 2.15, "learning_rate": 5.130085592549301e-06, "loss": 0.1883, "step": 31725 }, { "epoch": 2.15, "learning_rate": 5.129142943327929e-06, "loss": 0.4803, "step": 31730 }, { "epoch": 2.15, "learning_rate": 5.1282002941065575e-06, "loss": 0.8103, "step": 31735 }, { "epoch": 2.15, "learning_rate": 5.127257644885186e-06, "loss": 0.7739, "step": 31740 }, { "epoch": 2.15, "learning_rate": 5.126314995663814e-06, "loss": 0.4028, "step": 31745 }, { "epoch": 2.15, "learning_rate": 5.125372346442442e-06, "loss": 0.3439, "step": 31750 }, { "epoch": 2.16, "learning_rate": 5.124429697221071e-06, "loss": 0.7899, "step": 31755 }, { "epoch": 2.16, "learning_rate": 5.123487047999699e-06, "loss": 0.2033, "step": 31760 }, { "epoch": 2.16, "learning_rate": 5.1225443987783276e-06, "loss": 0.3601, "step": 31765 }, { "epoch": 2.16, "learning_rate": 5.121601749556955e-06, "loss": 0.1735, "step": 31770 }, { "epoch": 2.16, "learning_rate": 5.120659100335583e-06, "loss": 0.3108, "step": 31775 }, { "epoch": 2.16, "learning_rate": 5.119716451114211e-06, "loss": 0.8145, "step": 31780 }, { "epoch": 2.16, "learning_rate": 5.11877380189284e-06, "loss": 0.7913, "step": 31785 }, { "epoch": 2.16, "learning_rate": 5.117831152671468e-06, "loss": 0.2176, "step": 31790 }, { "epoch": 2.16, "learning_rate": 5.116888503450097e-06, "loss": 0.0574, "step": 31795 }, { "epoch": 2.16, "learning_rate": 5.115945854228724e-06, "loss": 0.0476, "step": 31800 }, { "epoch": 2.16, "learning_rate": 5.115003205007353e-06, "loss": 0.6562, "step": 31805 }, { "epoch": 2.16, "learning_rate": 5.114060555785981e-06, "loss": 0.3539, "step": 31810 }, { "epoch": 2.16, "learning_rate": 5.11311790656461e-06, "loss": 0.5083, "step": 31815 }, { "epoch": 2.16, "learning_rate": 5.112175257343238e-06, "loss": 0.5529, "step": 31820 }, { "epoch": 2.16, "learning_rate": 5.111232608121867e-06, "loss": 0.3342, "step": 31825 }, { "epoch": 2.16, "learning_rate": 5.110289958900494e-06, "loss": 0.1824, "step": 31830 }, { "epoch": 2.16, "learning_rate": 5.109347309679122e-06, "loss": 1.2727, "step": 31835 }, { "epoch": 2.16, "learning_rate": 5.108404660457751e-06, "loss": 0.3929, "step": 31840 }, { "epoch": 2.16, "learning_rate": 5.107462011236379e-06, "loss": 0.1445, "step": 31845 }, { "epoch": 2.16, "learning_rate": 5.106519362015008e-06, "loss": 0.4282, "step": 31850 }, { "epoch": 2.16, "learning_rate": 5.105576712793636e-06, "loss": 0.5814, "step": 31855 }, { "epoch": 2.16, "learning_rate": 5.104634063572264e-06, "loss": 0.2068, "step": 31860 }, { "epoch": 2.16, "learning_rate": 5.103691414350892e-06, "loss": 0.2308, "step": 31865 }, { "epoch": 2.16, "learning_rate": 5.102748765129521e-06, "loss": 0.105, "step": 31870 }, { "epoch": 2.16, "learning_rate": 5.101806115908149e-06, "loss": 0.3268, "step": 31875 }, { "epoch": 2.16, "learning_rate": 5.100863466686778e-06, "loss": 0.3443, "step": 31880 }, { "epoch": 2.16, "learning_rate": 5.0999208174654045e-06, "loss": 0.1981, "step": 31885 }, { "epoch": 2.16, "learning_rate": 5.098978168244033e-06, "loss": 0.2388, "step": 31890 }, { "epoch": 2.16, "learning_rate": 5.0980355190226614e-06, "loss": 0.4464, "step": 31895 }, { "epoch": 2.17, "learning_rate": 5.09709286980129e-06, "loss": 0.2811, "step": 31900 }, { "epoch": 2.17, "learning_rate": 5.096150220579918e-06, "loss": 0.28, "step": 31905 }, { "epoch": 2.17, "learning_rate": 5.095207571358547e-06, "loss": 0.4955, "step": 31910 }, { "epoch": 2.17, "learning_rate": 5.0942649221371745e-06, "loss": 0.1101, "step": 31915 }, { "epoch": 2.17, "learning_rate": 5.093322272915803e-06, "loss": 0.1626, "step": 31920 }, { "epoch": 2.17, "learning_rate": 5.0923796236944315e-06, "loss": 0.3347, "step": 31925 }, { "epoch": 2.17, "learning_rate": 5.09143697447306e-06, "loss": 0.0973, "step": 31930 }, { "epoch": 2.17, "learning_rate": 5.090494325251688e-06, "loss": 0.1825, "step": 31935 }, { "epoch": 2.17, "learning_rate": 5.089551676030316e-06, "loss": 0.3307, "step": 31940 }, { "epoch": 2.17, "learning_rate": 5.088609026808944e-06, "loss": 0.3811, "step": 31945 }, { "epoch": 2.17, "learning_rate": 5.087666377587572e-06, "loss": 0.1669, "step": 31950 }, { "epoch": 2.17, "learning_rate": 5.086723728366201e-06, "loss": 0.9056, "step": 31955 }, { "epoch": 2.17, "learning_rate": 5.085781079144829e-06, "loss": 0.1497, "step": 31960 }, { "epoch": 2.17, "learning_rate": 5.084838429923458e-06, "loss": 0.2055, "step": 31965 }, { "epoch": 2.17, "learning_rate": 5.083895780702085e-06, "loss": 0.4126, "step": 31970 }, { "epoch": 2.17, "learning_rate": 5.082953131480714e-06, "loss": 0.6094, "step": 31975 }, { "epoch": 2.17, "learning_rate": 5.0820104822593415e-06, "loss": 0.2922, "step": 31980 }, { "epoch": 2.17, "learning_rate": 5.08106783303797e-06, "loss": 0.3171, "step": 31985 }, { "epoch": 2.17, "learning_rate": 5.0801251838165985e-06, "loss": 0.5456, "step": 31990 }, { "epoch": 2.17, "learning_rate": 5.079182534595227e-06, "loss": 0.3482, "step": 31995 }, { "epoch": 2.17, "learning_rate": 5.078239885373855e-06, "loss": 0.618, "step": 32000 }, { "epoch": 2.17, "learning_rate": 5.077297236152483e-06, "loss": 0.8902, "step": 32005 }, { "epoch": 2.17, "learning_rate": 5.0763545869311116e-06, "loss": 0.6072, "step": 32010 }, { "epoch": 2.17, "learning_rate": 5.07541193770974e-06, "loss": 0.222, "step": 32015 }, { "epoch": 2.17, "learning_rate": 5.0744692884883685e-06, "loss": 0.1229, "step": 32020 }, { "epoch": 2.17, "learning_rate": 5.073526639266997e-06, "loss": 0.2773, "step": 32025 }, { "epoch": 2.17, "learning_rate": 5.072583990045624e-06, "loss": 0.2604, "step": 32030 }, { "epoch": 2.17, "learning_rate": 5.071641340824252e-06, "loss": 0.1368, "step": 32035 }, { "epoch": 2.17, "learning_rate": 5.070698691602881e-06, "loss": 0.1663, "step": 32040 }, { "epoch": 2.17, "learning_rate": 5.069756042381509e-06, "loss": 0.1905, "step": 32045 }, { "epoch": 2.18, "learning_rate": 5.068813393160138e-06, "loss": 0.4444, "step": 32050 }, { "epoch": 2.18, "learning_rate": 5.067870743938766e-06, "loss": 0.2827, "step": 32055 }, { "epoch": 2.18, "learning_rate": 5.066928094717394e-06, "loss": 0.5584, "step": 32060 }, { "epoch": 2.18, "learning_rate": 5.065985445496022e-06, "loss": 0.2022, "step": 32065 }, { "epoch": 2.18, "learning_rate": 5.065042796274651e-06, "loss": 0.1352, "step": 32070 }, { "epoch": 2.18, "learning_rate": 5.064100147053279e-06, "loss": 0.4155, "step": 32075 }, { "epoch": 2.18, "learning_rate": 5.063157497831908e-06, "loss": 0.3003, "step": 32080 }, { "epoch": 2.18, "learning_rate": 5.062214848610535e-06, "loss": 0.3429, "step": 32085 }, { "epoch": 2.18, "learning_rate": 5.061272199389163e-06, "loss": 0.3331, "step": 32090 }, { "epoch": 2.18, "learning_rate": 5.060329550167792e-06, "loss": 0.3103, "step": 32095 }, { "epoch": 2.18, "learning_rate": 5.05938690094642e-06, "loss": 0.1157, "step": 32100 }, { "epoch": 2.18, "learning_rate": 5.058444251725049e-06, "loss": 0.0768, "step": 32105 }, { "epoch": 2.18, "learning_rate": 5.057501602503677e-06, "loss": 0.7783, "step": 32110 }, { "epoch": 2.18, "learning_rate": 5.056558953282305e-06, "loss": 0.2959, "step": 32115 }, { "epoch": 2.18, "learning_rate": 5.055616304060933e-06, "loss": 0.3253, "step": 32120 }, { "epoch": 2.18, "learning_rate": 5.054673654839562e-06, "loss": 0.5197, "step": 32125 }, { "epoch": 2.18, "learning_rate": 5.05373100561819e-06, "loss": 0.2545, "step": 32130 }, { "epoch": 2.18, "learning_rate": 5.052788356396819e-06, "loss": 0.4992, "step": 32135 }, { "epoch": 2.18, "learning_rate": 5.051845707175447e-06, "loss": 0.8818, "step": 32140 }, { "epoch": 2.18, "learning_rate": 5.050903057954074e-06, "loss": 0.1159, "step": 32145 }, { "epoch": 2.18, "learning_rate": 5.0499604087327024e-06, "loss": 0.2656, "step": 32150 }, { "epoch": 2.18, "learning_rate": 5.049017759511331e-06, "loss": 0.183, "step": 32155 }, { "epoch": 2.18, "learning_rate": 5.048075110289959e-06, "loss": 0.6025, "step": 32160 }, { "epoch": 2.18, "learning_rate": 5.047132461068588e-06, "loss": 0.1877, "step": 32165 }, { "epoch": 2.18, "learning_rate": 5.046189811847216e-06, "loss": 0.2825, "step": 32170 }, { "epoch": 2.18, "learning_rate": 5.045247162625844e-06, "loss": 1.0475, "step": 32175 }, { "epoch": 2.18, "learning_rate": 5.0443045134044725e-06, "loss": 0.6104, "step": 32180 }, { "epoch": 2.18, "learning_rate": 5.043361864183101e-06, "loss": 0.4076, "step": 32185 }, { "epoch": 2.18, "learning_rate": 5.042419214961729e-06, "loss": 0.5974, "step": 32190 }, { "epoch": 2.19, "learning_rate": 5.041476565740357e-06, "loss": 0.5077, "step": 32195 }, { "epoch": 2.19, "learning_rate": 5.040533916518985e-06, "loss": 0.1663, "step": 32200 }, { "epoch": 2.19, "learning_rate": 5.039591267297613e-06, "loss": 0.1491, "step": 32205 }, { "epoch": 2.19, "learning_rate": 5.038648618076242e-06, "loss": 0.1659, "step": 32210 }, { "epoch": 2.19, "learning_rate": 5.03770596885487e-06, "loss": 0.4803, "step": 32215 }, { "epoch": 2.19, "learning_rate": 5.036763319633499e-06, "loss": 0.1937, "step": 32220 }, { "epoch": 2.19, "learning_rate": 5.035820670412127e-06, "loss": 0.2265, "step": 32225 }, { "epoch": 2.19, "learning_rate": 5.034878021190754e-06, "loss": 0.3504, "step": 32230 }, { "epoch": 2.19, "learning_rate": 5.0339353719693825e-06, "loss": 0.2838, "step": 32235 }, { "epoch": 2.19, "learning_rate": 5.032992722748011e-06, "loss": 0.6434, "step": 32240 }, { "epoch": 2.19, "learning_rate": 5.0320500735266395e-06, "loss": 0.3999, "step": 32245 }, { "epoch": 2.19, "learning_rate": 5.031107424305268e-06, "loss": 0.12, "step": 32250 }, { "epoch": 2.19, "learning_rate": 5.0301647750838964e-06, "loss": 0.1274, "step": 32255 }, { "epoch": 2.19, "learning_rate": 5.029222125862524e-06, "loss": 0.5939, "step": 32260 }, { "epoch": 2.19, "learning_rate": 5.0282794766411526e-06, "loss": 0.2478, "step": 32265 }, { "epoch": 2.19, "learning_rate": 5.027336827419781e-06, "loss": 0.7118, "step": 32270 }, { "epoch": 2.19, "learning_rate": 5.0263941781984095e-06, "loss": 0.4167, "step": 32275 }, { "epoch": 2.19, "learning_rate": 5.025451528977038e-06, "loss": 0.2738, "step": 32280 }, { "epoch": 2.19, "learning_rate": 5.0245088797556665e-06, "loss": 0.1578, "step": 32285 }, { "epoch": 2.19, "learning_rate": 5.023566230534293e-06, "loss": 0.3033, "step": 32290 }, { "epoch": 2.19, "learning_rate": 5.022623581312922e-06, "loss": 0.5476, "step": 32295 }, { "epoch": 2.19, "learning_rate": 5.02168093209155e-06, "loss": 0.1821, "step": 32300 }, { "epoch": 2.19, "learning_rate": 5.020738282870179e-06, "loss": 0.5573, "step": 32305 }, { "epoch": 2.19, "learning_rate": 5.019795633648807e-06, "loss": 0.1247, "step": 32310 }, { "epoch": 2.19, "learning_rate": 5.018852984427435e-06, "loss": 0.2245, "step": 32315 }, { "epoch": 2.19, "learning_rate": 5.017910335206063e-06, "loss": 0.0337, "step": 32320 }, { "epoch": 2.19, "learning_rate": 5.016967685984692e-06, "loss": 0.21, "step": 32325 }, { "epoch": 2.19, "learning_rate": 5.01602503676332e-06, "loss": 0.4921, "step": 32330 }, { "epoch": 2.19, "learning_rate": 5.015082387541949e-06, "loss": 0.4177, "step": 32335 }, { "epoch": 2.19, "learning_rate": 5.014139738320577e-06, "loss": 0.5315, "step": 32340 }, { "epoch": 2.2, "learning_rate": 5.013197089099204e-06, "loss": 0.3688, "step": 32345 }, { "epoch": 2.2, "learning_rate": 5.012254439877833e-06, "loss": 0.0942, "step": 32350 }, { "epoch": 2.2, "learning_rate": 5.011311790656461e-06, "loss": 0.6022, "step": 32355 }, { "epoch": 2.2, "learning_rate": 5.01036914143509e-06, "loss": 0.1973, "step": 32360 }, { "epoch": 2.2, "learning_rate": 5.009426492213718e-06, "loss": 0.1737, "step": 32365 }, { "epoch": 2.2, "learning_rate": 5.0084838429923466e-06, "loss": 0.5242, "step": 32370 }, { "epoch": 2.2, "learning_rate": 5.007541193770974e-06, "loss": 0.3171, "step": 32375 }, { "epoch": 2.2, "learning_rate": 5.006598544549603e-06, "loss": 0.2044, "step": 32380 }, { "epoch": 2.2, "learning_rate": 5.005655895328231e-06, "loss": 0.3018, "step": 32385 }, { "epoch": 2.2, "learning_rate": 5.00471324610686e-06, "loss": 0.2063, "step": 32390 }, { "epoch": 2.2, "learning_rate": 5.003770596885487e-06, "loss": 0.6096, "step": 32395 }, { "epoch": 2.2, "learning_rate": 5.002827947664116e-06, "loss": 0.4307, "step": 32400 }, { "epoch": 2.2, "learning_rate": 5.0018852984427434e-06, "loss": 0.3464, "step": 32405 }, { "epoch": 2.2, "learning_rate": 5.000942649221372e-06, "loss": 0.3944, "step": 32410 }, { "epoch": 2.2, "learning_rate": 5e-06, "loss": 0.1198, "step": 32415 }, { "epoch": 2.2, "learning_rate": 4.999057350778629e-06, "loss": 0.577, "step": 32420 }, { "epoch": 2.2, "learning_rate": 4.9981147015572565e-06, "loss": 0.1287, "step": 32425 }, { "epoch": 2.2, "learning_rate": 4.997172052335885e-06, "loss": 0.5734, "step": 32430 }, { "epoch": 2.2, "learning_rate": 4.9962294031145135e-06, "loss": 0.8562, "step": 32435 }, { "epoch": 2.2, "learning_rate": 4.995286753893141e-06, "loss": 0.2328, "step": 32440 }, { "epoch": 2.2, "learning_rate": 4.99434410467177e-06, "loss": 0.2098, "step": 32445 }, { "epoch": 2.2, "learning_rate": 4.993401455450398e-06, "loss": 0.3336, "step": 32450 }, { "epoch": 2.2, "learning_rate": 4.992458806229027e-06, "loss": 0.1324, "step": 32455 }, { "epoch": 2.2, "learning_rate": 4.991516157007655e-06, "loss": 0.2715, "step": 32460 }, { "epoch": 2.2, "learning_rate": 4.990573507786284e-06, "loss": 0.1186, "step": 32465 }, { "epoch": 2.2, "learning_rate": 4.989630858564911e-06, "loss": 0.2695, "step": 32470 }, { "epoch": 2.2, "learning_rate": 4.98868820934354e-06, "loss": 0.1308, "step": 32475 }, { "epoch": 2.2, "learning_rate": 4.987745560122167e-06, "loss": 0.2919, "step": 32480 }, { "epoch": 2.2, "learning_rate": 4.986802910900796e-06, "loss": 0.1063, "step": 32485 }, { "epoch": 2.21, "learning_rate": 4.985860261679424e-06, "loss": 0.286, "step": 32490 }, { "epoch": 2.21, "learning_rate": 4.984917612458052e-06, "loss": 0.7319, "step": 32495 }, { "epoch": 2.21, "learning_rate": 4.9839749632366805e-06, "loss": 0.2809, "step": 32500 }, { "epoch": 2.21, "learning_rate": 4.983032314015309e-06, "loss": 0.3311, "step": 32505 }, { "epoch": 2.21, "learning_rate": 4.9820896647939374e-06, "loss": 0.1449, "step": 32510 }, { "epoch": 2.21, "learning_rate": 4.981147015572566e-06, "loss": 0.5335, "step": 32515 }, { "epoch": 2.21, "learning_rate": 4.9802043663511936e-06, "loss": 0.325, "step": 32520 }, { "epoch": 2.21, "learning_rate": 4.979261717129822e-06, "loss": 0.4122, "step": 32525 }, { "epoch": 2.21, "learning_rate": 4.9783190679084505e-06, "loss": 1.1134, "step": 32530 }, { "epoch": 2.21, "learning_rate": 4.977376418687079e-06, "loss": 0.1703, "step": 32535 }, { "epoch": 2.21, "learning_rate": 4.976433769465707e-06, "loss": 0.5579, "step": 32540 }, { "epoch": 2.21, "learning_rate": 4.975491120244335e-06, "loss": 0.457, "step": 32545 }, { "epoch": 2.21, "learning_rate": 4.974548471022964e-06, "loss": 0.5539, "step": 32550 }, { "epoch": 2.21, "learning_rate": 4.973605821801591e-06, "loss": 0.2989, "step": 32555 }, { "epoch": 2.21, "learning_rate": 4.97266317258022e-06, "loss": 0.4681, "step": 32560 }, { "epoch": 2.21, "learning_rate": 4.971720523358848e-06, "loss": 0.9115, "step": 32565 }, { "epoch": 2.21, "learning_rate": 4.970777874137476e-06, "loss": 0.1903, "step": 32570 }, { "epoch": 2.21, "learning_rate": 4.969835224916104e-06, "loss": 0.3156, "step": 32575 }, { "epoch": 2.21, "learning_rate": 4.968892575694733e-06, "loss": 0.2679, "step": 32580 }, { "epoch": 2.21, "learning_rate": 4.967949926473361e-06, "loss": 0.4788, "step": 32585 }, { "epoch": 2.21, "learning_rate": 4.96700727725199e-06, "loss": 0.2457, "step": 32590 }, { "epoch": 2.21, "learning_rate": 4.9660646280306175e-06, "loss": 0.2277, "step": 32595 }, { "epoch": 2.21, "learning_rate": 4.965121978809246e-06, "loss": 0.2384, "step": 32600 }, { "epoch": 2.21, "learning_rate": 4.9641793295878745e-06, "loss": 0.3266, "step": 32605 }, { "epoch": 2.21, "learning_rate": 4.963236680366502e-06, "loss": 0.2094, "step": 32610 }, { "epoch": 2.21, "learning_rate": 4.962294031145131e-06, "loss": 0.2975, "step": 32615 }, { "epoch": 2.21, "learning_rate": 4.961351381923759e-06, "loss": 0.5426, "step": 32620 }, { "epoch": 2.21, "learning_rate": 4.960408732702387e-06, "loss": 0.6079, "step": 32625 }, { "epoch": 2.21, "learning_rate": 4.959466083481015e-06, "loss": 0.4813, "step": 32630 }, { "epoch": 2.21, "learning_rate": 4.958523434259644e-06, "loss": 0.3527, "step": 32635 }, { "epoch": 2.22, "learning_rate": 4.957580785038272e-06, "loss": 0.2471, "step": 32640 }, { "epoch": 2.22, "learning_rate": 4.9566381358169e-06, "loss": 0.3357, "step": 32645 }, { "epoch": 2.22, "learning_rate": 4.955695486595528e-06, "loss": 0.6103, "step": 32650 }, { "epoch": 2.22, "learning_rate": 4.954752837374157e-06, "loss": 0.408, "step": 32655 }, { "epoch": 2.22, "learning_rate": 4.953810188152785e-06, "loss": 0.5031, "step": 32660 }, { "epoch": 2.22, "learning_rate": 4.952867538931414e-06, "loss": 0.6527, "step": 32665 }, { "epoch": 2.22, "learning_rate": 4.951924889710041e-06, "loss": 0.398, "step": 32670 }, { "epoch": 2.22, "learning_rate": 4.95098224048867e-06, "loss": 0.3252, "step": 32675 }, { "epoch": 2.22, "learning_rate": 4.950039591267298e-06, "loss": 0.2975, "step": 32680 }, { "epoch": 2.22, "learning_rate": 4.949096942045926e-06, "loss": 0.2173, "step": 32685 }, { "epoch": 2.22, "learning_rate": 4.9481542928245545e-06, "loss": 0.405, "step": 32690 }, { "epoch": 2.22, "learning_rate": 4.947211643603182e-06, "loss": 0.3644, "step": 32695 }, { "epoch": 2.22, "learning_rate": 4.946268994381811e-06, "loss": 0.3752, "step": 32700 }, { "epoch": 2.22, "learning_rate": 4.945326345160439e-06, "loss": 0.3451, "step": 32705 }, { "epoch": 2.22, "learning_rate": 4.944383695939068e-06, "loss": 0.1471, "step": 32710 }, { "epoch": 2.22, "learning_rate": 4.943441046717696e-06, "loss": 0.1578, "step": 32715 }, { "epoch": 2.22, "learning_rate": 4.942498397496325e-06, "loss": 0.2265, "step": 32720 }, { "epoch": 2.22, "learning_rate": 4.941555748274952e-06, "loss": 0.4017, "step": 32725 }, { "epoch": 2.22, "learning_rate": 4.940613099053581e-06, "loss": 0.3188, "step": 32730 }, { "epoch": 2.22, "learning_rate": 4.939670449832209e-06, "loss": 0.1929, "step": 32735 }, { "epoch": 2.22, "learning_rate": 4.938727800610837e-06, "loss": 0.734, "step": 32740 }, { "epoch": 2.22, "learning_rate": 4.937785151389465e-06, "loss": 0.3194, "step": 32745 }, { "epoch": 2.22, "learning_rate": 4.936842502168094e-06, "loss": 0.1843, "step": 32750 }, { "epoch": 2.22, "learning_rate": 4.9358998529467215e-06, "loss": 0.0467, "step": 32755 }, { "epoch": 2.22, "learning_rate": 4.93495720372535e-06, "loss": 0.3693, "step": 32760 }, { "epoch": 2.22, "learning_rate": 4.9340145545039784e-06, "loss": 0.5357, "step": 32765 }, { "epoch": 2.22, "learning_rate": 4.933071905282606e-06, "loss": 0.0907, "step": 32770 }, { "epoch": 2.22, "learning_rate": 4.9321292560612346e-06, "loss": 0.4931, "step": 32775 }, { "epoch": 2.22, "learning_rate": 4.931186606839863e-06, "loss": 0.3833, "step": 32780 }, { "epoch": 2.23, "learning_rate": 4.9302439576184915e-06, "loss": 0.4142, "step": 32785 }, { "epoch": 2.23, "learning_rate": 4.92930130839712e-06, "loss": 0.0533, "step": 32790 }, { "epoch": 2.23, "learning_rate": 4.9283586591757485e-06, "loss": 0.5963, "step": 32795 }, { "epoch": 2.23, "learning_rate": 4.927416009954376e-06, "loss": 0.5306, "step": 32800 }, { "epoch": 2.23, "learning_rate": 4.926473360733005e-06, "loss": 0.2439, "step": 32805 }, { "epoch": 2.23, "learning_rate": 4.925530711511632e-06, "loss": 0.4742, "step": 32810 }, { "epoch": 2.23, "learning_rate": 4.924588062290261e-06, "loss": 0.2456, "step": 32815 }, { "epoch": 2.23, "learning_rate": 4.923645413068889e-06, "loss": 0.4554, "step": 32820 }, { "epoch": 2.23, "learning_rate": 4.922702763847517e-06, "loss": 0.2899, "step": 32825 }, { "epoch": 2.23, "learning_rate": 4.921760114626145e-06, "loss": 0.1141, "step": 32830 }, { "epoch": 2.23, "learning_rate": 4.920817465404774e-06, "loss": 0.2014, "step": 32835 }, { "epoch": 2.23, "learning_rate": 4.919874816183402e-06, "loss": 0.2863, "step": 32840 }, { "epoch": 2.23, "learning_rate": 4.918932166962031e-06, "loss": 0.6107, "step": 32845 }, { "epoch": 2.23, "learning_rate": 4.917989517740659e-06, "loss": 0.4309, "step": 32850 }, { "epoch": 2.23, "learning_rate": 4.917046868519287e-06, "loss": 0.2111, "step": 32855 }, { "epoch": 2.23, "learning_rate": 4.9161042192979155e-06, "loss": 0.3037, "step": 32860 }, { "epoch": 2.23, "learning_rate": 4.915161570076544e-06, "loss": 0.2626, "step": 32865 }, { "epoch": 2.23, "learning_rate": 4.914218920855172e-06, "loss": 0.2146, "step": 32870 }, { "epoch": 2.23, "learning_rate": 4.9132762716338e-06, "loss": 0.296, "step": 32875 }, { "epoch": 2.23, "learning_rate": 4.9123336224124286e-06, "loss": 0.3857, "step": 32880 }, { "epoch": 2.23, "learning_rate": 4.911390973191056e-06, "loss": 0.2911, "step": 32885 }, { "epoch": 2.23, "learning_rate": 4.910448323969685e-06, "loss": 0.4046, "step": 32890 }, { "epoch": 2.23, "learning_rate": 4.909505674748313e-06, "loss": 0.8276, "step": 32895 }, { "epoch": 2.23, "learning_rate": 4.908563025526941e-06, "loss": 0.512, "step": 32900 }, { "epoch": 2.23, "learning_rate": 4.907620376305569e-06, "loss": 0.4731, "step": 32905 }, { "epoch": 2.23, "learning_rate": 4.906677727084198e-06, "loss": 0.5286, "step": 32910 }, { "epoch": 2.23, "learning_rate": 4.905735077862826e-06, "loss": 0.6047, "step": 32915 }, { "epoch": 2.23, "learning_rate": 4.904792428641455e-06, "loss": 0.9019, "step": 32920 }, { "epoch": 2.23, "learning_rate": 4.903849779420082e-06, "loss": 0.2618, "step": 32925 }, { "epoch": 2.23, "learning_rate": 4.902907130198711e-06, "loss": 0.3531, "step": 32930 }, { "epoch": 2.24, "learning_rate": 4.901964480977339e-06, "loss": 0.1698, "step": 32935 }, { "epoch": 2.24, "learning_rate": 4.901021831755967e-06, "loss": 0.4107, "step": 32940 }, { "epoch": 2.24, "learning_rate": 4.9000791825345955e-06, "loss": 0.0831, "step": 32945 }, { "epoch": 2.24, "learning_rate": 4.899136533313224e-06, "loss": 0.0636, "step": 32950 }, { "epoch": 2.24, "learning_rate": 4.898193884091852e-06, "loss": 0.3203, "step": 32955 }, { "epoch": 2.24, "learning_rate": 4.89725123487048e-06, "loss": 0.3654, "step": 32960 }, { "epoch": 2.24, "learning_rate": 4.896308585649109e-06, "loss": 0.2337, "step": 32965 }, { "epoch": 2.24, "learning_rate": 4.895365936427737e-06, "loss": 0.2549, "step": 32970 }, { "epoch": 2.24, "learning_rate": 4.8944232872063656e-06, "loss": 0.1202, "step": 32975 }, { "epoch": 2.24, "learning_rate": 4.893480637984993e-06, "loss": 0.7369, "step": 32980 }, { "epoch": 2.24, "learning_rate": 4.892537988763622e-06, "loss": 0.2527, "step": 32985 }, { "epoch": 2.24, "learning_rate": 4.89159533954225e-06, "loss": 0.4674, "step": 32990 }, { "epoch": 2.24, "learning_rate": 4.890652690320879e-06, "loss": 0.6058, "step": 32995 }, { "epoch": 2.24, "learning_rate": 4.889710041099506e-06, "loss": 0.4187, "step": 33000 }, { "epoch": 2.24, "learning_rate": 4.888767391878135e-06, "loss": 0.4603, "step": 33005 }, { "epoch": 2.24, "learning_rate": 4.887824742656763e-06, "loss": 0.6074, "step": 33010 }, { "epoch": 2.24, "learning_rate": 4.886882093435391e-06, "loss": 0.1378, "step": 33015 }, { "epoch": 2.24, "learning_rate": 4.885939444214019e-06, "loss": 0.7596, "step": 33020 }, { "epoch": 2.24, "learning_rate": 4.884996794992648e-06, "loss": 0.3838, "step": 33025 }, { "epoch": 2.24, "learning_rate": 4.8840541457712755e-06, "loss": 0.1589, "step": 33030 }, { "epoch": 2.24, "learning_rate": 4.883111496549904e-06, "loss": 0.5481, "step": 33035 }, { "epoch": 2.24, "learning_rate": 4.8821688473285325e-06, "loss": 0.2168, "step": 33040 }, { "epoch": 2.24, "learning_rate": 4.881226198107161e-06, "loss": 0.3484, "step": 33045 }, { "epoch": 2.24, "learning_rate": 4.8802835488857895e-06, "loss": 0.6015, "step": 33050 }, { "epoch": 2.24, "learning_rate": 4.879340899664417e-06, "loss": 0.2056, "step": 33055 }, { "epoch": 2.24, "learning_rate": 4.878398250443046e-06, "loss": 0.6924, "step": 33060 }, { "epoch": 2.24, "learning_rate": 4.877455601221674e-06, "loss": 0.1199, "step": 33065 }, { "epoch": 2.24, "learning_rate": 4.876512952000302e-06, "loss": 0.1298, "step": 33070 }, { "epoch": 2.24, "learning_rate": 4.87557030277893e-06, "loss": 0.2215, "step": 33075 }, { "epoch": 2.25, "learning_rate": 4.874627653557559e-06, "loss": 0.6749, "step": 33080 }, { "epoch": 2.25, "learning_rate": 4.873685004336186e-06, "loss": 0.3902, "step": 33085 }, { "epoch": 2.25, "learning_rate": 4.872742355114815e-06, "loss": 0.4365, "step": 33090 }, { "epoch": 2.25, "learning_rate": 4.871799705893443e-06, "loss": 0.5115, "step": 33095 }, { "epoch": 2.25, "learning_rate": 4.870857056672072e-06, "loss": 0.5075, "step": 33100 }, { "epoch": 2.25, "learning_rate": 4.8699144074506995e-06, "loss": 0.1233, "step": 33105 }, { "epoch": 2.25, "learning_rate": 4.868971758229328e-06, "loss": 0.6792, "step": 33110 }, { "epoch": 2.25, "learning_rate": 4.8680291090079564e-06, "loss": 0.2792, "step": 33115 }, { "epoch": 2.25, "learning_rate": 4.867086459786585e-06, "loss": 0.3004, "step": 33120 }, { "epoch": 2.25, "learning_rate": 4.866143810565213e-06, "loss": 0.3377, "step": 33125 }, { "epoch": 2.25, "learning_rate": 4.865201161343841e-06, "loss": 0.5964, "step": 33130 }, { "epoch": 2.25, "learning_rate": 4.8642585121224695e-06, "loss": 0.7626, "step": 33135 }, { "epoch": 2.25, "learning_rate": 4.863315862901098e-06, "loss": 0.2919, "step": 33140 }, { "epoch": 2.25, "learning_rate": 4.862373213679726e-06, "loss": 0.0773, "step": 33145 }, { "epoch": 2.25, "learning_rate": 4.861430564458354e-06, "loss": 0.2537, "step": 33150 }, { "epoch": 2.25, "learning_rate": 4.860487915236982e-06, "loss": 0.3533, "step": 33155 }, { "epoch": 2.25, "learning_rate": 4.85954526601561e-06, "loss": 0.0505, "step": 33160 }, { "epoch": 2.25, "learning_rate": 4.858602616794239e-06, "loss": 0.3015, "step": 33165 }, { "epoch": 2.25, "learning_rate": 4.857659967572867e-06, "loss": 0.2361, "step": 33170 }, { "epoch": 2.25, "learning_rate": 4.856717318351496e-06, "loss": 0.257, "step": 33175 }, { "epoch": 2.25, "learning_rate": 4.855774669130124e-06, "loss": 0.2545, "step": 33180 }, { "epoch": 2.25, "learning_rate": 4.854832019908752e-06, "loss": 0.7439, "step": 33185 }, { "epoch": 2.25, "learning_rate": 4.85388937068738e-06, "loss": 0.5787, "step": 33190 }, { "epoch": 2.25, "learning_rate": 4.852946721466009e-06, "loss": 0.0893, "step": 33195 }, { "epoch": 2.25, "learning_rate": 4.8520040722446365e-06, "loss": 0.1777, "step": 33200 }, { "epoch": 2.25, "learning_rate": 4.851061423023265e-06, "loss": 0.4261, "step": 33205 }, { "epoch": 2.25, "learning_rate": 4.8501187738018935e-06, "loss": 0.6944, "step": 33210 }, { "epoch": 2.25, "learning_rate": 4.849176124580521e-06, "loss": 0.3933, "step": 33215 }, { "epoch": 2.25, "learning_rate": 4.84823347535915e-06, "loss": 0.2202, "step": 33220 }, { "epoch": 2.25, "learning_rate": 4.847290826137778e-06, "loss": 0.2355, "step": 33225 }, { "epoch": 2.26, "learning_rate": 4.846348176916406e-06, "loss": 0.2552, "step": 33230 }, { "epoch": 2.26, "learning_rate": 4.845405527695034e-06, "loss": 0.3055, "step": 33235 }, { "epoch": 2.26, "learning_rate": 4.844462878473663e-06, "loss": 0.3296, "step": 33240 }, { "epoch": 2.26, "learning_rate": 4.843520229252291e-06, "loss": 0.3975, "step": 33245 }, { "epoch": 2.26, "learning_rate": 4.84257758003092e-06, "loss": 0.1414, "step": 33250 }, { "epoch": 2.26, "learning_rate": 4.841634930809547e-06, "loss": 0.3138, "step": 33255 }, { "epoch": 2.26, "learning_rate": 4.840692281588176e-06, "loss": 0.9167, "step": 33260 }, { "epoch": 2.26, "learning_rate": 4.839749632366804e-06, "loss": 0.163, "step": 33265 }, { "epoch": 2.26, "learning_rate": 4.838806983145432e-06, "loss": 0.7181, "step": 33270 }, { "epoch": 2.26, "learning_rate": 4.83786433392406e-06, "loss": 0.6799, "step": 33275 }, { "epoch": 2.26, "learning_rate": 4.836921684702689e-06, "loss": 0.1517, "step": 33280 }, { "epoch": 2.26, "learning_rate": 4.8359790354813165e-06, "loss": 0.8512, "step": 33285 }, { "epoch": 2.26, "learning_rate": 4.835036386259945e-06, "loss": 0.2654, "step": 33290 }, { "epoch": 2.26, "learning_rate": 4.8340937370385735e-06, "loss": 0.2933, "step": 33295 }, { "epoch": 2.26, "learning_rate": 4.833151087817202e-06, "loss": 0.6196, "step": 33300 }, { "epoch": 2.26, "learning_rate": 4.8322084385958305e-06, "loss": 0.2736, "step": 33305 }, { "epoch": 2.26, "learning_rate": 4.831265789374459e-06, "loss": 0.2597, "step": 33310 }, { "epoch": 2.26, "learning_rate": 4.830323140153087e-06, "loss": 0.2459, "step": 33315 }, { "epoch": 2.26, "learning_rate": 4.829380490931715e-06, "loss": 0.4325, "step": 33320 }, { "epoch": 2.26, "learning_rate": 4.828437841710344e-06, "loss": 0.293, "step": 33325 }, { "epoch": 2.26, "learning_rate": 4.827495192488971e-06, "loss": 0.4196, "step": 33330 }, { "epoch": 2.26, "learning_rate": 4.8265525432676e-06, "loss": 0.2493, "step": 33335 }, { "epoch": 2.26, "learning_rate": 4.825609894046228e-06, "loss": 0.601, "step": 33340 }, { "epoch": 2.26, "learning_rate": 4.824667244824856e-06, "loss": 0.7324, "step": 33345 }, { "epoch": 2.26, "learning_rate": 4.823724595603484e-06, "loss": 0.2991, "step": 33350 }, { "epoch": 2.26, "learning_rate": 4.822781946382113e-06, "loss": 0.4063, "step": 33355 }, { "epoch": 2.26, "learning_rate": 4.8218392971607405e-06, "loss": 0.9512, "step": 33360 }, { "epoch": 2.26, "learning_rate": 4.820896647939369e-06, "loss": 0.0684, "step": 33365 }, { "epoch": 2.26, "learning_rate": 4.8199539987179974e-06, "loss": 0.7883, "step": 33370 }, { "epoch": 2.27, "learning_rate": 4.819011349496626e-06, "loss": 0.6464, "step": 33375 }, { "epoch": 2.27, "learning_rate": 4.818068700275254e-06, "loss": 0.9516, "step": 33380 }, { "epoch": 2.27, "learning_rate": 4.817126051053882e-06, "loss": 0.1581, "step": 33385 }, { "epoch": 2.27, "learning_rate": 4.8161834018325105e-06, "loss": 0.1259, "step": 33390 }, { "epoch": 2.27, "learning_rate": 4.815240752611139e-06, "loss": 0.3962, "step": 33395 }, { "epoch": 2.27, "learning_rate": 4.814298103389767e-06, "loss": 0.5012, "step": 33400 }, { "epoch": 2.27, "learning_rate": 4.813355454168395e-06, "loss": 0.1223, "step": 33405 }, { "epoch": 2.27, "learning_rate": 4.812412804947024e-06, "loss": 0.2796, "step": 33410 }, { "epoch": 2.27, "learning_rate": 4.811470155725651e-06, "loss": 0.2267, "step": 33415 }, { "epoch": 2.27, "learning_rate": 4.81052750650428e-06, "loss": 0.1598, "step": 33420 }, { "epoch": 2.27, "learning_rate": 4.809584857282908e-06, "loss": 0.8189, "step": 33425 }, { "epoch": 2.27, "learning_rate": 4.808642208061537e-06, "loss": 0.2812, "step": 33430 }, { "epoch": 2.27, "learning_rate": 4.807699558840165e-06, "loss": 0.369, "step": 33435 }, { "epoch": 2.27, "learning_rate": 4.806756909618793e-06, "loss": 0.2572, "step": 33440 }, { "epoch": 2.27, "learning_rate": 4.805814260397421e-06, "loss": 0.2867, "step": 33445 }, { "epoch": 2.27, "learning_rate": 4.80487161117605e-06, "loss": 0.2935, "step": 33450 }, { "epoch": 2.27, "learning_rate": 4.803928961954678e-06, "loss": 0.3172, "step": 33455 }, { "epoch": 2.27, "learning_rate": 4.802986312733306e-06, "loss": 0.2253, "step": 33460 }, { "epoch": 2.27, "learning_rate": 4.8020436635119345e-06, "loss": 0.3477, "step": 33465 }, { "epoch": 2.27, "learning_rate": 4.801101014290563e-06, "loss": 0.5188, "step": 33470 }, { "epoch": 2.27, "learning_rate": 4.800158365069191e-06, "loss": 1.0361, "step": 33475 }, { "epoch": 2.27, "learning_rate": 4.799215715847819e-06, "loss": 0.403, "step": 33480 }, { "epoch": 2.27, "learning_rate": 4.798273066626447e-06, "loss": 0.2628, "step": 33485 }, { "epoch": 2.27, "learning_rate": 4.797330417405075e-06, "loss": 0.6947, "step": 33490 }, { "epoch": 2.27, "learning_rate": 4.796387768183704e-06, "loss": 0.2793, "step": 33495 }, { "epoch": 2.27, "learning_rate": 4.795445118962332e-06, "loss": 0.2411, "step": 33500 }, { "epoch": 2.27, "learning_rate": 4.794502469740961e-06, "loss": 0.3428, "step": 33505 }, { "epoch": 2.27, "learning_rate": 4.793559820519589e-06, "loss": 0.2184, "step": 33510 }, { "epoch": 2.27, "learning_rate": 4.792617171298217e-06, "loss": 0.3533, "step": 33515 }, { "epoch": 2.28, "learning_rate": 4.791674522076845e-06, "loss": 0.5774, "step": 33520 }, { "epoch": 2.28, "learning_rate": 4.790731872855474e-06, "loss": 0.345, "step": 33525 }, { "epoch": 2.28, "learning_rate": 4.789789223634101e-06, "loss": 0.1358, "step": 33530 }, { "epoch": 2.28, "learning_rate": 4.78884657441273e-06, "loss": 0.1006, "step": 33535 }, { "epoch": 2.28, "learning_rate": 4.787903925191358e-06, "loss": 0.3497, "step": 33540 }, { "epoch": 2.28, "learning_rate": 4.786961275969986e-06, "loss": 0.28, "step": 33545 }, { "epoch": 2.28, "learning_rate": 4.7860186267486145e-06, "loss": 0.5188, "step": 33550 }, { "epoch": 2.28, "learning_rate": 4.785075977527243e-06, "loss": 0.2533, "step": 33555 }, { "epoch": 2.28, "learning_rate": 4.7841333283058715e-06, "loss": 0.328, "step": 33560 }, { "epoch": 2.28, "learning_rate": 4.783190679084499e-06, "loss": 0.1895, "step": 33565 }, { "epoch": 2.28, "learning_rate": 4.782248029863128e-06, "loss": 0.5131, "step": 33570 }, { "epoch": 2.28, "learning_rate": 4.781305380641756e-06, "loss": 0.5473, "step": 33575 }, { "epoch": 2.28, "learning_rate": 4.780362731420385e-06, "loss": 0.2974, "step": 33580 }, { "epoch": 2.28, "learning_rate": 4.779420082199013e-06, "loss": 0.6921, "step": 33585 }, { "epoch": 2.28, "learning_rate": 4.778477432977641e-06, "loss": 0.1092, "step": 33590 }, { "epoch": 2.28, "learning_rate": 4.777534783756269e-06, "loss": 0.4439, "step": 33595 }, { "epoch": 2.28, "learning_rate": 4.776592134534897e-06, "loss": 0.5171, "step": 33600 }, { "epoch": 2.28, "learning_rate": 4.775649485313525e-06, "loss": 0.3887, "step": 33605 }, { "epoch": 2.28, "learning_rate": 4.774706836092154e-06, "loss": 0.3009, "step": 33610 }, { "epoch": 2.28, "learning_rate": 4.7737641868707815e-06, "loss": 0.2358, "step": 33615 }, { "epoch": 2.28, "learning_rate": 4.77282153764941e-06, "loss": 0.428, "step": 33620 }, { "epoch": 2.28, "learning_rate": 4.7718788884280384e-06, "loss": 0.319, "step": 33625 }, { "epoch": 2.28, "learning_rate": 4.770936239206667e-06, "loss": 0.5082, "step": 33630 }, { "epoch": 2.28, "learning_rate": 4.769993589985295e-06, "loss": 0.444, "step": 33635 }, { "epoch": 2.28, "learning_rate": 4.769050940763924e-06, "loss": 0.2052, "step": 33640 }, { "epoch": 2.28, "learning_rate": 4.7681082915425515e-06, "loss": 0.2716, "step": 33645 }, { "epoch": 2.28, "learning_rate": 4.76716564232118e-06, "loss": 0.3236, "step": 33650 }, { "epoch": 2.28, "learning_rate": 4.7662229930998085e-06, "loss": 0.6624, "step": 33655 }, { "epoch": 2.28, "learning_rate": 4.765280343878436e-06, "loss": 0.2378, "step": 33660 }, { "epoch": 2.28, "learning_rate": 4.764337694657065e-06, "loss": 0.2715, "step": 33665 }, { "epoch": 2.29, "learning_rate": 4.763395045435693e-06, "loss": 0.6114, "step": 33670 }, { "epoch": 2.29, "learning_rate": 4.762452396214321e-06, "loss": 0.2839, "step": 33675 }, { "epoch": 2.29, "learning_rate": 4.761509746992949e-06, "loss": 0.2977, "step": 33680 }, { "epoch": 2.29, "learning_rate": 4.760567097771578e-06, "loss": 0.244, "step": 33685 }, { "epoch": 2.29, "learning_rate": 4.759624448550205e-06, "loss": 0.595, "step": 33690 }, { "epoch": 2.29, "learning_rate": 4.758681799328834e-06, "loss": 0.2017, "step": 33695 }, { "epoch": 2.29, "learning_rate": 4.757739150107462e-06, "loss": 0.2492, "step": 33700 }, { "epoch": 2.29, "learning_rate": 4.756796500886091e-06, "loss": 0.4804, "step": 33705 }, { "epoch": 2.29, "learning_rate": 4.755853851664719e-06, "loss": 0.3111, "step": 33710 }, { "epoch": 2.29, "learning_rate": 4.754911202443347e-06, "loss": 0.1662, "step": 33715 }, { "epoch": 2.29, "learning_rate": 4.7539685532219755e-06, "loss": 0.6624, "step": 33720 }, { "epoch": 2.29, "learning_rate": 4.753025904000604e-06, "loss": 0.3487, "step": 33725 }, { "epoch": 2.29, "learning_rate": 4.752083254779232e-06, "loss": 0.2192, "step": 33730 }, { "epoch": 2.29, "learning_rate": 4.75114060555786e-06, "loss": 0.4497, "step": 33735 }, { "epoch": 2.29, "learning_rate": 4.7501979563364886e-06, "loss": 0.8822, "step": 33740 }, { "epoch": 2.29, "learning_rate": 4.749255307115116e-06, "loss": 0.464, "step": 33745 }, { "epoch": 2.29, "learning_rate": 4.748312657893745e-06, "loss": 0.5722, "step": 33750 }, { "epoch": 2.29, "learning_rate": 4.747370008672373e-06, "loss": 0.4323, "step": 33755 }, { "epoch": 2.29, "learning_rate": 4.746427359451002e-06, "loss": 0.2976, "step": 33760 }, { "epoch": 2.29, "learning_rate": 4.74548471022963e-06, "loss": 0.5433, "step": 33765 }, { "epoch": 2.29, "learning_rate": 4.744542061008258e-06, "loss": 0.2384, "step": 33770 }, { "epoch": 2.29, "learning_rate": 4.743599411786886e-06, "loss": 0.1433, "step": 33775 }, { "epoch": 2.29, "learning_rate": 4.742656762565515e-06, "loss": 0.5419, "step": 33780 }, { "epoch": 2.29, "learning_rate": 4.741714113344143e-06, "loss": 0.2049, "step": 33785 }, { "epoch": 2.29, "learning_rate": 4.740771464122771e-06, "loss": 0.4841, "step": 33790 }, { "epoch": 2.29, "learning_rate": 4.739828814901399e-06, "loss": 1.1731, "step": 33795 }, { "epoch": 2.29, "learning_rate": 4.738886165680028e-06, "loss": 0.1621, "step": 33800 }, { "epoch": 2.29, "learning_rate": 4.7379435164586555e-06, "loss": 0.2944, "step": 33805 }, { "epoch": 2.29, "learning_rate": 4.737000867237284e-06, "loss": 0.5394, "step": 33810 }, { "epoch": 2.3, "learning_rate": 4.736058218015912e-06, "loss": 0.4303, "step": 33815 }, { "epoch": 2.3, "learning_rate": 4.73511556879454e-06, "loss": 1.0052, "step": 33820 }, { "epoch": 2.3, "learning_rate": 4.734172919573169e-06, "loss": 0.2685, "step": 33825 }, { "epoch": 2.3, "learning_rate": 4.733230270351797e-06, "loss": 0.4228, "step": 33830 }, { "epoch": 2.3, "learning_rate": 4.732287621130426e-06, "loss": 0.2806, "step": 33835 }, { "epoch": 2.3, "learning_rate": 4.731344971909054e-06, "loss": 0.4288, "step": 33840 }, { "epoch": 2.3, "learning_rate": 4.730402322687682e-06, "loss": 0.0856, "step": 33845 }, { "epoch": 2.3, "learning_rate": 4.72945967346631e-06, "loss": 0.5087, "step": 33850 }, { "epoch": 2.3, "learning_rate": 4.728517024244939e-06, "loss": 0.0873, "step": 33855 }, { "epoch": 2.3, "learning_rate": 4.727574375023566e-06, "loss": 0.4468, "step": 33860 }, { "epoch": 2.3, "learning_rate": 4.726631725802195e-06, "loss": 0.4525, "step": 33865 }, { "epoch": 2.3, "learning_rate": 4.725689076580823e-06, "loss": 0.4678, "step": 33870 }, { "epoch": 2.3, "learning_rate": 4.724746427359451e-06, "loss": 0.6936, "step": 33875 }, { "epoch": 2.3, "learning_rate": 4.7238037781380794e-06, "loss": 0.4047, "step": 33880 }, { "epoch": 2.3, "learning_rate": 4.722861128916708e-06, "loss": 0.2006, "step": 33885 }, { "epoch": 2.3, "learning_rate": 4.721918479695336e-06, "loss": 0.2818, "step": 33890 }, { "epoch": 2.3, "learning_rate": 4.720975830473964e-06, "loss": 0.0878, "step": 33895 }, { "epoch": 2.3, "learning_rate": 4.7200331812525925e-06, "loss": 0.365, "step": 33900 }, { "epoch": 2.3, "learning_rate": 4.719090532031221e-06, "loss": 0.1584, "step": 33905 }, { "epoch": 2.3, "learning_rate": 4.7181478828098495e-06, "loss": 0.5146, "step": 33910 }, { "epoch": 2.3, "learning_rate": 4.717205233588478e-06, "loss": 0.2826, "step": 33915 }, { "epoch": 2.3, "learning_rate": 4.716262584367106e-06, "loss": 0.1824, "step": 33920 }, { "epoch": 2.3, "learning_rate": 4.715319935145734e-06, "loss": 0.5811, "step": 33925 }, { "epoch": 2.3, "learning_rate": 4.714377285924362e-06, "loss": 0.4936, "step": 33930 }, { "epoch": 2.3, "learning_rate": 4.71343463670299e-06, "loss": 0.7001, "step": 33935 }, { "epoch": 2.3, "learning_rate": 4.712491987481619e-06, "loss": 0.2706, "step": 33940 }, { "epoch": 2.3, "learning_rate": 4.711549338260246e-06, "loss": 0.4323, "step": 33945 }, { "epoch": 2.3, "learning_rate": 4.710606689038875e-06, "loss": 0.5151, "step": 33950 }, { "epoch": 2.3, "learning_rate": 4.709664039817503e-06, "loss": 0.5489, "step": 33955 }, { "epoch": 2.3, "learning_rate": 4.708721390596132e-06, "loss": 0.1366, "step": 33960 }, { "epoch": 2.31, "learning_rate": 4.70777874137476e-06, "loss": 0.3719, "step": 33965 }, { "epoch": 2.31, "learning_rate": 4.706836092153389e-06, "loss": 0.4209, "step": 33970 }, { "epoch": 2.31, "learning_rate": 4.7058934429320165e-06, "loss": 0.2493, "step": 33975 }, { "epoch": 2.31, "learning_rate": 4.704950793710645e-06, "loss": 0.2844, "step": 33980 }, { "epoch": 2.31, "learning_rate": 4.7040081444892734e-06, "loss": 0.447, "step": 33985 }, { "epoch": 2.31, "learning_rate": 4.703065495267901e-06, "loss": 0.6133, "step": 33990 }, { "epoch": 2.31, "learning_rate": 4.7021228460465296e-06, "loss": 0.2062, "step": 33995 }, { "epoch": 2.31, "learning_rate": 4.701180196825158e-06, "loss": 0.0714, "step": 34000 }, { "epoch": 2.31, "learning_rate": 4.700237547603786e-06, "loss": 0.3591, "step": 34005 }, { "epoch": 2.31, "learning_rate": 4.699294898382414e-06, "loss": 0.1123, "step": 34010 }, { "epoch": 2.31, "learning_rate": 4.698352249161043e-06, "loss": 0.6236, "step": 34015 }, { "epoch": 2.31, "learning_rate": 4.69740959993967e-06, "loss": 0.2716, "step": 34020 }, { "epoch": 2.31, "learning_rate": 4.696466950718299e-06, "loss": 0.332, "step": 34025 }, { "epoch": 2.31, "learning_rate": 4.695524301496927e-06, "loss": 0.3296, "step": 34030 }, { "epoch": 2.31, "learning_rate": 4.694581652275556e-06, "loss": 0.525, "step": 34035 }, { "epoch": 2.31, "learning_rate": 4.693639003054184e-06, "loss": 0.3065, "step": 34040 }, { "epoch": 2.31, "learning_rate": 4.692696353832812e-06, "loss": 0.3787, "step": 34045 }, { "epoch": 2.31, "learning_rate": 4.69175370461144e-06, "loss": 0.2713, "step": 34050 }, { "epoch": 2.31, "learning_rate": 4.690811055390069e-06, "loss": 0.1857, "step": 34055 }, { "epoch": 2.31, "learning_rate": 4.6898684061686965e-06, "loss": 0.3143, "step": 34060 }, { "epoch": 2.31, "learning_rate": 4.688925756947325e-06, "loss": 0.0902, "step": 34065 }, { "epoch": 2.31, "learning_rate": 4.6879831077259535e-06, "loss": 0.1791, "step": 34070 }, { "epoch": 2.31, "learning_rate": 4.687040458504581e-06, "loss": 0.2709, "step": 34075 }, { "epoch": 2.31, "learning_rate": 4.68609780928321e-06, "loss": 0.2953, "step": 34080 }, { "epoch": 2.31, "learning_rate": 4.685155160061838e-06, "loss": 0.2008, "step": 34085 }, { "epoch": 2.31, "learning_rate": 4.6842125108404666e-06, "loss": 0.5876, "step": 34090 }, { "epoch": 2.31, "learning_rate": 4.683269861619095e-06, "loss": 0.2559, "step": 34095 }, { "epoch": 2.31, "learning_rate": 4.6823272123977236e-06, "loss": 0.2715, "step": 34100 }, { "epoch": 2.31, "learning_rate": 4.681384563176351e-06, "loss": 0.5997, "step": 34105 }, { "epoch": 2.32, "learning_rate": 4.68044191395498e-06, "loss": 0.3044, "step": 34110 }, { "epoch": 2.32, "learning_rate": 4.679499264733608e-06, "loss": 0.7928, "step": 34115 }, { "epoch": 2.32, "learning_rate": 4.678556615512236e-06, "loss": 0.3927, "step": 34120 }, { "epoch": 2.32, "learning_rate": 4.677613966290864e-06, "loss": 0.497, "step": 34125 }, { "epoch": 2.32, "learning_rate": 4.676671317069493e-06, "loss": 0.1923, "step": 34130 }, { "epoch": 2.32, "learning_rate": 4.67572866784812e-06, "loss": 0.0378, "step": 34135 }, { "epoch": 2.32, "learning_rate": 4.674786018626749e-06, "loss": 0.3834, "step": 34140 }, { "epoch": 2.32, "learning_rate": 4.673843369405377e-06, "loss": 0.1008, "step": 34145 }, { "epoch": 2.32, "learning_rate": 4.672900720184005e-06, "loss": 0.2246, "step": 34150 }, { "epoch": 2.32, "learning_rate": 4.6719580709626335e-06, "loss": 0.4201, "step": 34155 }, { "epoch": 2.32, "learning_rate": 4.671015421741262e-06, "loss": 0.3978, "step": 34160 }, { "epoch": 2.32, "learning_rate": 4.6700727725198905e-06, "loss": 0.6313, "step": 34165 }, { "epoch": 2.32, "learning_rate": 4.669130123298519e-06, "loss": 0.1674, "step": 34170 }, { "epoch": 2.32, "learning_rate": 4.668187474077147e-06, "loss": 0.4621, "step": 34175 }, { "epoch": 2.32, "learning_rate": 4.667244824855775e-06, "loss": 0.8003, "step": 34180 }, { "epoch": 2.32, "learning_rate": 4.666302175634404e-06, "loss": 0.4116, "step": 34185 }, { "epoch": 2.32, "learning_rate": 4.665359526413031e-06, "loss": 0.1102, "step": 34190 }, { "epoch": 2.32, "learning_rate": 4.66441687719166e-06, "loss": 0.4579, "step": 34195 }, { "epoch": 2.32, "learning_rate": 4.663474227970288e-06, "loss": 0.5125, "step": 34200 }, { "epoch": 2.32, "learning_rate": 4.662531578748916e-06, "loss": 0.2477, "step": 34205 }, { "epoch": 2.32, "learning_rate": 4.661588929527544e-06, "loss": 0.3079, "step": 34210 }, { "epoch": 2.32, "learning_rate": 4.660646280306173e-06, "loss": 0.4088, "step": 34215 }, { "epoch": 2.32, "learning_rate": 4.659703631084801e-06, "loss": 0.3143, "step": 34220 }, { "epoch": 2.32, "learning_rate": 4.65876098186343e-06, "loss": 0.4726, "step": 34225 }, { "epoch": 2.32, "learning_rate": 4.6578183326420574e-06, "loss": 0.1927, "step": 34230 }, { "epoch": 2.32, "learning_rate": 4.656875683420686e-06, "loss": 0.2748, "step": 34235 }, { "epoch": 2.32, "learning_rate": 4.655933034199314e-06, "loss": 0.1751, "step": 34240 }, { "epoch": 2.32, "learning_rate": 4.654990384977943e-06, "loss": 0.0454, "step": 34245 }, { "epoch": 2.32, "learning_rate": 4.6540477357565705e-06, "loss": 0.216, "step": 34250 }, { "epoch": 2.32, "learning_rate": 4.653105086535199e-06, "loss": 0.5102, "step": 34255 }, { "epoch": 2.33, "learning_rate": 4.652162437313827e-06, "loss": 0.5298, "step": 34260 }, { "epoch": 2.33, "learning_rate": 4.651219788092455e-06, "loss": 0.7154, "step": 34265 }, { "epoch": 2.33, "learning_rate": 4.650277138871084e-06, "loss": 0.225, "step": 34270 }, { "epoch": 2.33, "learning_rate": 4.649334489649711e-06, "loss": 0.4415, "step": 34275 }, { "epoch": 2.33, "learning_rate": 4.64839184042834e-06, "loss": 0.1746, "step": 34280 }, { "epoch": 2.33, "learning_rate": 4.647449191206968e-06, "loss": 0.3881, "step": 34285 }, { "epoch": 2.33, "learning_rate": 4.646506541985597e-06, "loss": 0.7261, "step": 34290 }, { "epoch": 2.33, "learning_rate": 4.645563892764225e-06, "loss": 0.1766, "step": 34295 }, { "epoch": 2.33, "learning_rate": 4.644621243542854e-06, "loss": 0.3599, "step": 34300 }, { "epoch": 2.33, "learning_rate": 4.643678594321481e-06, "loss": 0.712, "step": 34305 }, { "epoch": 2.33, "learning_rate": 4.64273594510011e-06, "loss": 0.1815, "step": 34310 }, { "epoch": 2.33, "learning_rate": 4.641793295878738e-06, "loss": 0.2217, "step": 34315 }, { "epoch": 2.33, "learning_rate": 4.640850646657366e-06, "loss": 0.4808, "step": 34320 }, { "epoch": 2.33, "learning_rate": 4.6399079974359945e-06, "loss": 0.1453, "step": 34325 }, { "epoch": 2.33, "learning_rate": 4.638965348214623e-06, "loss": 0.1628, "step": 34330 }, { "epoch": 2.33, "learning_rate": 4.638022698993251e-06, "loss": 0.1814, "step": 34335 }, { "epoch": 2.33, "learning_rate": 4.637080049771879e-06, "loss": 0.0353, "step": 34340 }, { "epoch": 2.33, "learning_rate": 4.6361374005505076e-06, "loss": 0.2569, "step": 34345 }, { "epoch": 2.33, "learning_rate": 4.635194751329136e-06, "loss": 0.3675, "step": 34350 }, { "epoch": 2.33, "learning_rate": 4.634252102107764e-06, "loss": 0.3858, "step": 34355 }, { "epoch": 2.33, "learning_rate": 4.633309452886392e-06, "loss": 0.5747, "step": 34360 }, { "epoch": 2.33, "learning_rate": 4.632366803665021e-06, "loss": 0.066, "step": 34365 }, { "epoch": 2.33, "learning_rate": 4.631424154443649e-06, "loss": 0.4512, "step": 34370 }, { "epoch": 2.33, "learning_rate": 4.630481505222277e-06, "loss": 0.4869, "step": 34375 }, { "epoch": 2.33, "learning_rate": 4.629538856000905e-06, "loss": 0.418, "step": 34380 }, { "epoch": 2.33, "learning_rate": 4.628596206779534e-06, "loss": 0.6334, "step": 34385 }, { "epoch": 2.33, "learning_rate": 4.627653557558161e-06, "loss": 0.5041, "step": 34390 }, { "epoch": 2.33, "learning_rate": 4.62671090833679e-06, "loss": 0.2801, "step": 34395 }, { "epoch": 2.33, "learning_rate": 4.625768259115418e-06, "loss": 0.5192, "step": 34400 }, { "epoch": 2.34, "learning_rate": 4.624825609894046e-06, "loss": 0.6825, "step": 34405 }, { "epoch": 2.34, "learning_rate": 4.6238829606726745e-06, "loss": 0.1339, "step": 34410 }, { "epoch": 2.34, "learning_rate": 4.622940311451303e-06, "loss": 0.7188, "step": 34415 }, { "epoch": 2.34, "learning_rate": 4.6219976622299315e-06, "loss": 0.4647, "step": 34420 }, { "epoch": 2.34, "learning_rate": 4.62105501300856e-06, "loss": 0.1355, "step": 34425 }, { "epoch": 2.34, "learning_rate": 4.6201123637871885e-06, "loss": 0.3997, "step": 34430 }, { "epoch": 2.34, "learning_rate": 4.619169714565816e-06, "loss": 0.3452, "step": 34435 }, { "epoch": 2.34, "learning_rate": 4.618227065344445e-06, "loss": 0.1719, "step": 34440 }, { "epoch": 2.34, "learning_rate": 4.617284416123073e-06, "loss": 0.2295, "step": 34445 }, { "epoch": 2.34, "learning_rate": 4.616341766901701e-06, "loss": 0.2545, "step": 34450 }, { "epoch": 2.34, "learning_rate": 4.615399117680329e-06, "loss": 0.2557, "step": 34455 }, { "epoch": 2.34, "learning_rate": 4.614456468458958e-06, "loss": 0.4678, "step": 34460 }, { "epoch": 2.34, "learning_rate": 4.613513819237585e-06, "loss": 0.441, "step": 34465 }, { "epoch": 2.34, "learning_rate": 4.612571170016214e-06, "loss": 0.4005, "step": 34470 }, { "epoch": 2.34, "learning_rate": 4.611628520794842e-06, "loss": 0.529, "step": 34475 }, { "epoch": 2.34, "learning_rate": 4.61068587157347e-06, "loss": 0.2253, "step": 34480 }, { "epoch": 2.34, "learning_rate": 4.6097432223520984e-06, "loss": 0.6498, "step": 34485 }, { "epoch": 2.34, "learning_rate": 4.608800573130727e-06, "loss": 0.3342, "step": 34490 }, { "epoch": 2.34, "learning_rate": 4.607857923909355e-06, "loss": 0.5123, "step": 34495 }, { "epoch": 2.34, "learning_rate": 4.606915274687984e-06, "loss": 0.418, "step": 34500 }, { "epoch": 2.34, "learning_rate": 4.6059726254666115e-06, "loss": 0.1136, "step": 34505 }, { "epoch": 2.34, "learning_rate": 4.60502997624524e-06, "loss": 0.5726, "step": 34510 }, { "epoch": 2.34, "learning_rate": 4.6040873270238685e-06, "loss": 0.1463, "step": 34515 }, { "epoch": 2.34, "learning_rate": 4.603144677802496e-06, "loss": 0.2412, "step": 34520 }, { "epoch": 2.34, "learning_rate": 4.602202028581125e-06, "loss": 0.995, "step": 34525 }, { "epoch": 2.34, "learning_rate": 4.601259379359753e-06, "loss": 0.4288, "step": 34530 }, { "epoch": 2.34, "learning_rate": 4.600316730138381e-06, "loss": 0.1262, "step": 34535 }, { "epoch": 2.34, "learning_rate": 4.599374080917009e-06, "loss": 0.1573, "step": 34540 }, { "epoch": 2.34, "learning_rate": 4.598431431695638e-06, "loss": 0.2098, "step": 34545 }, { "epoch": 2.34, "learning_rate": 4.597488782474266e-06, "loss": 0.302, "step": 34550 }, { "epoch": 2.35, "learning_rate": 4.596546133252895e-06, "loss": 0.0797, "step": 34555 }, { "epoch": 2.35, "learning_rate": 4.595603484031523e-06, "loss": 0.3394, "step": 34560 }, { "epoch": 2.35, "learning_rate": 4.594660834810151e-06, "loss": 0.2836, "step": 34565 }, { "epoch": 2.35, "learning_rate": 4.593718185588779e-06, "loss": 0.2657, "step": 34570 }, { "epoch": 2.35, "learning_rate": 4.592775536367408e-06, "loss": 0.3329, "step": 34575 }, { "epoch": 2.35, "learning_rate": 4.5918328871460355e-06, "loss": 0.3289, "step": 34580 }, { "epoch": 2.35, "learning_rate": 4.590890237924664e-06, "loss": 0.5125, "step": 34585 }, { "epoch": 2.35, "learning_rate": 4.5899475887032924e-06, "loss": 0.7514, "step": 34590 }, { "epoch": 2.35, "learning_rate": 4.58900493948192e-06, "loss": 0.1318, "step": 34595 }, { "epoch": 2.35, "learning_rate": 4.5880622902605486e-06, "loss": 0.3867, "step": 34600 }, { "epoch": 2.35, "learning_rate": 4.587119641039176e-06, "loss": 0.4396, "step": 34605 }, { "epoch": 2.35, "learning_rate": 4.586176991817805e-06, "loss": 0.4034, "step": 34610 }, { "epoch": 2.35, "learning_rate": 4.585234342596433e-06, "loss": 0.2582, "step": 34615 }, { "epoch": 2.35, "learning_rate": 4.584291693375062e-06, "loss": 0.2829, "step": 34620 }, { "epoch": 2.35, "learning_rate": 4.58334904415369e-06, "loss": 0.1473, "step": 34625 }, { "epoch": 2.35, "learning_rate": 4.582406394932319e-06, "loss": 0.5233, "step": 34630 }, { "epoch": 2.35, "learning_rate": 4.581463745710946e-06, "loss": 0.8052, "step": 34635 }, { "epoch": 2.35, "learning_rate": 4.580521096489575e-06, "loss": 0.366, "step": 34640 }, { "epoch": 2.35, "learning_rate": 4.579578447268203e-06, "loss": 0.503, "step": 34645 }, { "epoch": 2.35, "learning_rate": 4.578635798046831e-06, "loss": 0.2468, "step": 34650 }, { "epoch": 2.35, "learning_rate": 4.577693148825459e-06, "loss": 0.2792, "step": 34655 }, { "epoch": 2.35, "learning_rate": 4.576750499604088e-06, "loss": 0.319, "step": 34660 }, { "epoch": 2.35, "learning_rate": 4.5758078503827155e-06, "loss": 0.2019, "step": 34665 }, { "epoch": 2.35, "learning_rate": 4.574865201161344e-06, "loss": 0.3851, "step": 34670 }, { "epoch": 2.35, "learning_rate": 4.5739225519399725e-06, "loss": 0.3388, "step": 34675 }, { "epoch": 2.35, "learning_rate": 4.572979902718601e-06, "loss": 0.5639, "step": 34680 }, { "epoch": 2.35, "learning_rate": 4.5720372534972295e-06, "loss": 0.1711, "step": 34685 }, { "epoch": 2.35, "learning_rate": 4.571094604275857e-06, "loss": 0.3129, "step": 34690 }, { "epoch": 2.35, "learning_rate": 4.570151955054486e-06, "loss": 0.1636, "step": 34695 }, { "epoch": 2.36, "learning_rate": 4.569209305833114e-06, "loss": 0.2423, "step": 34700 }, { "epoch": 2.36, "learning_rate": 4.5682666566117426e-06, "loss": 0.1334, "step": 34705 }, { "epoch": 2.36, "learning_rate": 4.56732400739037e-06, "loss": 0.3116, "step": 34710 }, { "epoch": 2.36, "learning_rate": 4.566381358168999e-06, "loss": 0.0495, "step": 34715 }, { "epoch": 2.36, "learning_rate": 4.565438708947626e-06, "loss": 0.1339, "step": 34720 }, { "epoch": 2.36, "learning_rate": 4.564496059726255e-06, "loss": 0.6664, "step": 34725 }, { "epoch": 2.36, "learning_rate": 4.563553410504883e-06, "loss": 0.3277, "step": 34730 }, { "epoch": 2.36, "learning_rate": 4.562610761283511e-06, "loss": 0.289, "step": 34735 }, { "epoch": 2.36, "learning_rate": 4.5616681120621394e-06, "loss": 0.3131, "step": 34740 }, { "epoch": 2.36, "learning_rate": 4.560725462840768e-06, "loss": 0.3268, "step": 34745 }, { "epoch": 2.36, "learning_rate": 4.559782813619396e-06, "loss": 0.2203, "step": 34750 }, { "epoch": 2.36, "learning_rate": 4.558840164398025e-06, "loss": 0.3112, "step": 34755 }, { "epoch": 2.36, "learning_rate": 4.557897515176653e-06, "loss": 0.7848, "step": 34760 }, { "epoch": 2.36, "learning_rate": 4.556954865955281e-06, "loss": 0.5566, "step": 34765 }, { "epoch": 2.36, "learning_rate": 4.5560122167339095e-06, "loss": 0.4606, "step": 34770 }, { "epoch": 2.36, "learning_rate": 4.555069567512538e-06, "loss": 0.6584, "step": 34775 }, { "epoch": 2.36, "learning_rate": 4.554126918291166e-06, "loss": 0.2581, "step": 34780 }, { "epoch": 2.36, "learning_rate": 4.553184269069794e-06, "loss": 0.3403, "step": 34785 }, { "epoch": 2.36, "learning_rate": 4.552241619848423e-06, "loss": 0.0976, "step": 34790 }, { "epoch": 2.36, "learning_rate": 4.55129897062705e-06, "loss": 0.6698, "step": 34795 }, { "epoch": 2.36, "learning_rate": 4.550356321405679e-06, "loss": 0.0757, "step": 34800 }, { "epoch": 2.36, "learning_rate": 4.549413672184307e-06, "loss": 0.2651, "step": 34805 }, { "epoch": 2.36, "learning_rate": 4.548471022962936e-06, "loss": 0.1529, "step": 34810 }, { "epoch": 2.36, "learning_rate": 4.547528373741563e-06, "loss": 0.1687, "step": 34815 }, { "epoch": 2.36, "learning_rate": 4.546585724520192e-06, "loss": 0.5038, "step": 34820 }, { "epoch": 2.36, "learning_rate": 4.54564307529882e-06, "loss": 0.1787, "step": 34825 }, { "epoch": 2.36, "learning_rate": 4.544700426077449e-06, "loss": 0.2349, "step": 34830 }, { "epoch": 2.36, "learning_rate": 4.5437577768560765e-06, "loss": 0.3061, "step": 34835 }, { "epoch": 2.36, "learning_rate": 4.542815127634705e-06, "loss": 0.4092, "step": 34840 }, { "epoch": 2.36, "learning_rate": 4.5418724784133334e-06, "loss": 0.2483, "step": 34845 }, { "epoch": 2.37, "learning_rate": 4.540929829191961e-06, "loss": 0.477, "step": 34850 }, { "epoch": 2.37, "learning_rate": 4.5399871799705896e-06, "loss": 0.1936, "step": 34855 }, { "epoch": 2.37, "learning_rate": 4.539044530749218e-06, "loss": 0.2463, "step": 34860 }, { "epoch": 2.37, "learning_rate": 4.538101881527846e-06, "loss": 0.3477, "step": 34865 }, { "epoch": 2.37, "learning_rate": 4.537159232306474e-06, "loss": 0.1691, "step": 34870 }, { "epoch": 2.37, "learning_rate": 4.536216583085103e-06, "loss": 0.4186, "step": 34875 }, { "epoch": 2.37, "learning_rate": 4.535273933863731e-06, "loss": 0.6149, "step": 34880 }, { "epoch": 2.37, "learning_rate": 4.53433128464236e-06, "loss": 0.5102, "step": 34885 }, { "epoch": 2.37, "learning_rate": 4.533388635420988e-06, "loss": 0.1362, "step": 34890 }, { "epoch": 2.37, "learning_rate": 4.532445986199616e-06, "loss": 0.5699, "step": 34895 }, { "epoch": 2.37, "learning_rate": 4.531503336978244e-06, "loss": 0.2283, "step": 34900 }, { "epoch": 2.37, "learning_rate": 4.530560687756873e-06, "loss": 0.3589, "step": 34905 }, { "epoch": 2.37, "learning_rate": 4.5296180385355e-06, "loss": 0.2451, "step": 34910 }, { "epoch": 2.37, "learning_rate": 4.528675389314129e-06, "loss": 0.5428, "step": 34915 }, { "epoch": 2.37, "learning_rate": 4.527732740092757e-06, "loss": 0.1661, "step": 34920 }, { "epoch": 2.37, "learning_rate": 4.526790090871385e-06, "loss": 0.3574, "step": 34925 }, { "epoch": 2.37, "learning_rate": 4.5258474416500135e-06, "loss": 0.2317, "step": 34930 }, { "epoch": 2.37, "learning_rate": 4.524904792428642e-06, "loss": 0.1154, "step": 34935 }, { "epoch": 2.37, "learning_rate": 4.52396214320727e-06, "loss": 0.2757, "step": 34940 }, { "epoch": 2.37, "learning_rate": 4.523019493985898e-06, "loss": 0.4734, "step": 34945 }, { "epoch": 2.37, "learning_rate": 4.522076844764527e-06, "loss": 0.3195, "step": 34950 }, { "epoch": 2.37, "learning_rate": 4.521134195543155e-06, "loss": 0.293, "step": 34955 }, { "epoch": 2.37, "learning_rate": 4.5201915463217836e-06, "loss": 0.5801, "step": 34960 }, { "epoch": 2.37, "learning_rate": 4.519248897100411e-06, "loss": 0.9488, "step": 34965 }, { "epoch": 2.37, "learning_rate": 4.51830624787904e-06, "loss": 0.732, "step": 34970 }, { "epoch": 2.37, "learning_rate": 4.517363598657668e-06, "loss": 0.3147, "step": 34975 }, { "epoch": 2.37, "learning_rate": 4.516420949436296e-06, "loss": 0.358, "step": 34980 }, { "epoch": 2.37, "learning_rate": 4.515478300214924e-06, "loss": 0.6594, "step": 34985 }, { "epoch": 2.37, "learning_rate": 4.514535650993553e-06, "loss": 0.4163, "step": 34990 }, { "epoch": 2.38, "learning_rate": 4.5135930017721804e-06, "loss": 0.1754, "step": 34995 }, { "epoch": 2.38, "learning_rate": 4.512650352550809e-06, "loss": 0.2978, "step": 35000 }, { "epoch": 2.38, "learning_rate": 4.511707703329437e-06, "loss": 0.2258, "step": 35005 }, { "epoch": 2.38, "learning_rate": 4.510765054108066e-06, "loss": 0.0988, "step": 35010 }, { "epoch": 2.38, "learning_rate": 4.509822404886694e-06, "loss": 0.1159, "step": 35015 }, { "epoch": 2.38, "learning_rate": 4.508879755665322e-06, "loss": 0.1658, "step": 35020 }, { "epoch": 2.38, "learning_rate": 4.5079371064439505e-06, "loss": 0.5163, "step": 35025 }, { "epoch": 2.38, "learning_rate": 4.506994457222579e-06, "loss": 0.3977, "step": 35030 }, { "epoch": 2.38, "learning_rate": 4.5060518080012075e-06, "loss": 0.2666, "step": 35035 }, { "epoch": 2.38, "learning_rate": 4.505109158779835e-06, "loss": 0.2921, "step": 35040 }, { "epoch": 2.38, "learning_rate": 4.504166509558464e-06, "loss": 0.5331, "step": 35045 }, { "epoch": 2.38, "learning_rate": 4.503223860337091e-06, "loss": 0.6206, "step": 35050 }, { "epoch": 2.38, "learning_rate": 4.50228121111572e-06, "loss": 0.264, "step": 35055 }, { "epoch": 2.38, "learning_rate": 4.501338561894348e-06, "loss": 0.0726, "step": 35060 }, { "epoch": 2.38, "learning_rate": 4.500395912672976e-06, "loss": 0.6032, "step": 35065 }, { "epoch": 2.38, "learning_rate": 4.499453263451604e-06, "loss": 0.4372, "step": 35070 }, { "epoch": 2.38, "learning_rate": 4.498510614230233e-06, "loss": 0.5228, "step": 35075 }, { "epoch": 2.38, "learning_rate": 4.497567965008861e-06, "loss": 0.8809, "step": 35080 }, { "epoch": 2.38, "learning_rate": 4.49662531578749e-06, "loss": 0.4363, "step": 35085 }, { "epoch": 2.38, "learning_rate": 4.495682666566118e-06, "loss": 0.3346, "step": 35090 }, { "epoch": 2.38, "learning_rate": 4.494740017344746e-06, "loss": 0.3199, "step": 35095 }, { "epoch": 2.38, "learning_rate": 4.4937973681233744e-06, "loss": 0.1657, "step": 35100 }, { "epoch": 2.38, "learning_rate": 4.492854718902003e-06, "loss": 0.3694, "step": 35105 }, { "epoch": 2.38, "learning_rate": 4.4919120696806306e-06, "loss": 0.2735, "step": 35110 }, { "epoch": 2.38, "learning_rate": 4.490969420459259e-06, "loss": 0.507, "step": 35115 }, { "epoch": 2.38, "learning_rate": 4.4900267712378875e-06, "loss": 0.0963, "step": 35120 }, { "epoch": 2.38, "learning_rate": 4.489084122016515e-06, "loss": 0.1237, "step": 35125 }, { "epoch": 2.38, "learning_rate": 4.488141472795144e-06, "loss": 0.0475, "step": 35130 }, { "epoch": 2.38, "learning_rate": 4.487198823573772e-06, "loss": 0.1348, "step": 35135 }, { "epoch": 2.38, "learning_rate": 4.486256174352401e-06, "loss": 0.0368, "step": 35140 }, { "epoch": 2.39, "learning_rate": 4.485313525131028e-06, "loss": 0.4115, "step": 35145 }, { "epoch": 2.39, "learning_rate": 4.484370875909657e-06, "loss": 0.3138, "step": 35150 }, { "epoch": 2.39, "learning_rate": 4.483428226688285e-06, "loss": 0.1098, "step": 35155 }, { "epoch": 2.39, "learning_rate": 4.482485577466914e-06, "loss": 0.1655, "step": 35160 }, { "epoch": 2.39, "learning_rate": 4.481542928245541e-06, "loss": 0.5282, "step": 35165 }, { "epoch": 2.39, "learning_rate": 4.48060027902417e-06, "loss": 0.5575, "step": 35170 }, { "epoch": 2.39, "learning_rate": 4.479657629802798e-06, "loss": 0.4411, "step": 35175 }, { "epoch": 2.39, "learning_rate": 4.478714980581426e-06, "loss": 0.5893, "step": 35180 }, { "epoch": 2.39, "learning_rate": 4.4777723313600545e-06, "loss": 0.1165, "step": 35185 }, { "epoch": 2.39, "learning_rate": 4.476829682138683e-06, "loss": 0.6494, "step": 35190 }, { "epoch": 2.39, "learning_rate": 4.475887032917311e-06, "loss": 0.25, "step": 35195 }, { "epoch": 2.39, "learning_rate": 4.474944383695939e-06, "loss": 0.2382, "step": 35200 }, { "epoch": 2.39, "learning_rate": 4.474001734474568e-06, "loss": 0.1682, "step": 35205 }, { "epoch": 2.39, "learning_rate": 4.473059085253196e-06, "loss": 0.1228, "step": 35210 }, { "epoch": 2.39, "learning_rate": 4.4721164360318246e-06, "loss": 0.3722, "step": 35215 }, { "epoch": 2.39, "learning_rate": 4.471173786810453e-06, "loss": 0.2456, "step": 35220 }, { "epoch": 2.39, "learning_rate": 4.470231137589081e-06, "loss": 1.1771, "step": 35225 }, { "epoch": 2.39, "learning_rate": 4.469288488367709e-06, "loss": 0.3196, "step": 35230 }, { "epoch": 2.39, "learning_rate": 4.468345839146338e-06, "loss": 0.1718, "step": 35235 }, { "epoch": 2.39, "learning_rate": 4.467403189924965e-06, "loss": 0.6429, "step": 35240 }, { "epoch": 2.39, "learning_rate": 4.466460540703594e-06, "loss": 0.1834, "step": 35245 }, { "epoch": 2.39, "learning_rate": 4.465517891482222e-06, "loss": 0.2481, "step": 35250 }, { "epoch": 2.39, "learning_rate": 4.46457524226085e-06, "loss": 0.2308, "step": 35255 }, { "epoch": 2.39, "learning_rate": 4.463632593039478e-06, "loss": 0.4439, "step": 35260 }, { "epoch": 2.39, "learning_rate": 4.462689943818107e-06, "loss": 0.4038, "step": 35265 }, { "epoch": 2.39, "learning_rate": 4.4617472945967345e-06, "loss": 0.386, "step": 35270 }, { "epoch": 2.39, "learning_rate": 4.460804645375363e-06, "loss": 0.199, "step": 35275 }, { "epoch": 2.39, "learning_rate": 4.4598619961539915e-06, "loss": 0.338, "step": 35280 }, { "epoch": 2.39, "learning_rate": 4.45891934693262e-06, "loss": 0.5108, "step": 35285 }, { "epoch": 2.4, "learning_rate": 4.4579766977112485e-06, "loss": 0.3355, "step": 35290 }, { "epoch": 2.4, "learning_rate": 4.457034048489876e-06, "loss": 0.1229, "step": 35295 }, { "epoch": 2.4, "learning_rate": 4.456091399268505e-06, "loss": 0.6628, "step": 35300 }, { "epoch": 2.4, "learning_rate": 4.455148750047133e-06, "loss": 0.5086, "step": 35305 }, { "epoch": 2.4, "learning_rate": 4.454206100825761e-06, "loss": 0.4192, "step": 35310 }, { "epoch": 2.4, "learning_rate": 4.453263451604389e-06, "loss": 0.1793, "step": 35315 }, { "epoch": 2.4, "learning_rate": 4.452320802383018e-06, "loss": 0.2877, "step": 35320 }, { "epoch": 2.4, "learning_rate": 4.451378153161645e-06, "loss": 0.1354, "step": 35325 }, { "epoch": 2.4, "learning_rate": 4.450435503940274e-06, "loss": 0.3091, "step": 35330 }, { "epoch": 2.4, "learning_rate": 4.449492854718902e-06, "loss": 0.5316, "step": 35335 }, { "epoch": 2.4, "learning_rate": 4.448550205497531e-06, "loss": 0.6745, "step": 35340 }, { "epoch": 2.4, "learning_rate": 4.447607556276159e-06, "loss": 0.362, "step": 35345 }, { "epoch": 2.4, "learning_rate": 4.446664907054788e-06, "loss": 0.0679, "step": 35350 }, { "epoch": 2.4, "learning_rate": 4.445722257833415e-06, "loss": 0.3387, "step": 35355 }, { "epoch": 2.4, "learning_rate": 4.444779608612044e-06, "loss": 0.4706, "step": 35360 }, { "epoch": 2.4, "learning_rate": 4.443836959390672e-06, "loss": 0.097, "step": 35365 }, { "epoch": 2.4, "learning_rate": 4.4428943101693e-06, "loss": 0.1164, "step": 35370 }, { "epoch": 2.4, "learning_rate": 4.4419516609479285e-06, "loss": 0.8243, "step": 35375 }, { "epoch": 2.4, "learning_rate": 4.441009011726556e-06, "loss": 0.5173, "step": 35380 }, { "epoch": 2.4, "learning_rate": 4.440066362505185e-06, "loss": 0.6614, "step": 35385 }, { "epoch": 2.4, "learning_rate": 4.439123713283813e-06, "loss": 0.4461, "step": 35390 }, { "epoch": 2.4, "learning_rate": 4.438181064062442e-06, "loss": 0.0999, "step": 35395 }, { "epoch": 2.4, "learning_rate": 4.437238414841069e-06, "loss": 0.3168, "step": 35400 }, { "epoch": 2.4, "learning_rate": 4.436295765619698e-06, "loss": 0.3391, "step": 35405 }, { "epoch": 2.4, "learning_rate": 4.435353116398326e-06, "loss": 0.7473, "step": 35410 }, { "epoch": 2.4, "learning_rate": 4.434410467176955e-06, "loss": 0.4593, "step": 35415 }, { "epoch": 2.4, "learning_rate": 4.433467817955583e-06, "loss": 0.0999, "step": 35420 }, { "epoch": 2.4, "learning_rate": 4.432525168734211e-06, "loss": 0.2726, "step": 35425 }, { "epoch": 2.4, "learning_rate": 4.431582519512839e-06, "loss": 0.6851, "step": 35430 }, { "epoch": 2.4, "learning_rate": 4.430639870291468e-06, "loss": 0.5601, "step": 35435 }, { "epoch": 2.41, "learning_rate": 4.4296972210700955e-06, "loss": 0.2769, "step": 35440 }, { "epoch": 2.41, "learning_rate": 4.428754571848724e-06, "loss": 0.3687, "step": 35445 }, { "epoch": 2.41, "learning_rate": 4.4278119226273524e-06, "loss": 0.2795, "step": 35450 }, { "epoch": 2.41, "learning_rate": 4.42686927340598e-06, "loss": 0.2771, "step": 35455 }, { "epoch": 2.41, "learning_rate": 4.4259266241846086e-06, "loss": 0.5207, "step": 35460 }, { "epoch": 2.41, "learning_rate": 4.424983974963237e-06, "loss": 0.7316, "step": 35465 }, { "epoch": 2.41, "learning_rate": 4.4240413257418655e-06, "loss": 0.1718, "step": 35470 }, { "epoch": 2.41, "learning_rate": 4.423098676520494e-06, "loss": 0.3072, "step": 35475 }, { "epoch": 2.41, "learning_rate": 4.422156027299122e-06, "loss": 0.5817, "step": 35480 }, { "epoch": 2.41, "learning_rate": 4.42121337807775e-06, "loss": 0.0449, "step": 35485 }, { "epoch": 2.41, "learning_rate": 4.420270728856379e-06, "loss": 0.205, "step": 35490 }, { "epoch": 2.41, "learning_rate": 4.419328079635006e-06, "loss": 0.2262, "step": 35495 }, { "epoch": 2.41, "learning_rate": 4.418385430413635e-06, "loss": 0.5684, "step": 35500 }, { "epoch": 2.41, "learning_rate": 4.417442781192263e-06, "loss": 0.4657, "step": 35505 }, { "epoch": 2.41, "learning_rate": 4.416500131970891e-06, "loss": 0.4693, "step": 35510 }, { "epoch": 2.41, "learning_rate": 4.415557482749519e-06, "loss": 0.4248, "step": 35515 }, { "epoch": 2.41, "learning_rate": 4.414614833528148e-06, "loss": 0.1766, "step": 35520 }, { "epoch": 2.41, "learning_rate": 4.4136721843067755e-06, "loss": 0.5219, "step": 35525 }, { "epoch": 2.41, "learning_rate": 4.412729535085404e-06, "loss": 0.5081, "step": 35530 }, { "epoch": 2.41, "learning_rate": 4.4117868858640325e-06, "loss": 0.4902, "step": 35535 }, { "epoch": 2.41, "learning_rate": 4.410844236642661e-06, "loss": 0.2671, "step": 35540 }, { "epoch": 2.41, "learning_rate": 4.4099015874212895e-06, "loss": 0.3089, "step": 35545 }, { "epoch": 2.41, "learning_rate": 4.408958938199918e-06, "loss": 0.305, "step": 35550 }, { "epoch": 2.41, "learning_rate": 4.408016288978546e-06, "loss": 0.2623, "step": 35555 }, { "epoch": 2.41, "learning_rate": 4.407073639757174e-06, "loss": 0.2752, "step": 35560 }, { "epoch": 2.41, "learning_rate": 4.4061309905358026e-06, "loss": 0.4316, "step": 35565 }, { "epoch": 2.41, "learning_rate": 4.40518834131443e-06, "loss": 0.1803, "step": 35570 }, { "epoch": 2.41, "learning_rate": 4.404245692093059e-06, "loss": 0.3109, "step": 35575 }, { "epoch": 2.41, "learning_rate": 4.403303042871687e-06, "loss": 0.6729, "step": 35580 }, { "epoch": 2.42, "learning_rate": 4.402360393650315e-06, "loss": 0.3966, "step": 35585 }, { "epoch": 2.42, "learning_rate": 4.401417744428943e-06, "loss": 0.9918, "step": 35590 }, { "epoch": 2.42, "learning_rate": 4.400475095207572e-06, "loss": 0.4729, "step": 35595 }, { "epoch": 2.42, "learning_rate": 4.3995324459862e-06, "loss": 0.318, "step": 35600 }, { "epoch": 2.42, "learning_rate": 4.398589796764828e-06, "loss": 0.7088, "step": 35605 }, { "epoch": 2.42, "learning_rate": 4.397647147543456e-06, "loss": 0.3306, "step": 35610 }, { "epoch": 2.42, "learning_rate": 4.396704498322085e-06, "loss": 0.1755, "step": 35615 }, { "epoch": 2.42, "learning_rate": 4.395761849100713e-06, "loss": 0.0751, "step": 35620 }, { "epoch": 2.42, "learning_rate": 4.394819199879341e-06, "loss": 0.1309, "step": 35625 }, { "epoch": 2.42, "learning_rate": 4.3938765506579695e-06, "loss": 0.5867, "step": 35630 }, { "epoch": 2.42, "learning_rate": 4.392933901436598e-06, "loss": 0.3058, "step": 35635 }, { "epoch": 2.42, "learning_rate": 4.391991252215226e-06, "loss": 0.3437, "step": 35640 }, { "epoch": 2.42, "learning_rate": 4.391048602993854e-06, "loss": 0.281, "step": 35645 }, { "epoch": 2.42, "learning_rate": 4.390105953772483e-06, "loss": 0.3088, "step": 35650 }, { "epoch": 2.42, "learning_rate": 4.38916330455111e-06, "loss": 0.0851, "step": 35655 }, { "epoch": 2.42, "learning_rate": 4.388220655329739e-06, "loss": 0.4945, "step": 35660 }, { "epoch": 2.42, "learning_rate": 4.387278006108367e-06, "loss": 0.295, "step": 35665 }, { "epoch": 2.42, "learning_rate": 4.386335356886996e-06, "loss": 0.322, "step": 35670 }, { "epoch": 2.42, "learning_rate": 4.385392707665624e-06, "loss": 0.642, "step": 35675 }, { "epoch": 2.42, "learning_rate": 4.384450058444253e-06, "loss": 0.2846, "step": 35680 }, { "epoch": 2.42, "learning_rate": 4.38350740922288e-06, "loss": 0.4239, "step": 35685 }, { "epoch": 2.42, "learning_rate": 4.382564760001509e-06, "loss": 0.1604, "step": 35690 }, { "epoch": 2.42, "learning_rate": 4.381622110780137e-06, "loss": 0.0549, "step": 35695 }, { "epoch": 2.42, "learning_rate": 4.380679461558765e-06, "loss": 0.4991, "step": 35700 }, { "epoch": 2.42, "learning_rate": 4.3797368123373934e-06, "loss": 0.1128, "step": 35705 }, { "epoch": 2.42, "learning_rate": 4.378794163116022e-06, "loss": 0.2137, "step": 35710 }, { "epoch": 2.42, "learning_rate": 4.3778515138946496e-06, "loss": 0.2162, "step": 35715 }, { "epoch": 2.42, "learning_rate": 4.376908864673278e-06, "loss": 0.2849, "step": 35720 }, { "epoch": 2.42, "learning_rate": 4.3759662154519065e-06, "loss": 0.2308, "step": 35725 }, { "epoch": 2.43, "learning_rate": 4.375023566230534e-06, "loss": 0.2286, "step": 35730 }, { "epoch": 2.43, "learning_rate": 4.374080917009163e-06, "loss": 0.4094, "step": 35735 }, { "epoch": 2.43, "learning_rate": 4.373138267787791e-06, "loss": 0.3752, "step": 35740 }, { "epoch": 2.43, "learning_rate": 4.37219561856642e-06, "loss": 0.3332, "step": 35745 }, { "epoch": 2.43, "learning_rate": 4.371252969345048e-06, "loss": 0.1846, "step": 35750 }, { "epoch": 2.43, "learning_rate": 4.370310320123676e-06, "loss": 0.0952, "step": 35755 }, { "epoch": 2.43, "learning_rate": 4.369367670902304e-06, "loss": 0.4955, "step": 35760 }, { "epoch": 2.43, "learning_rate": 4.368425021680933e-06, "loss": 0.6221, "step": 35765 }, { "epoch": 2.43, "learning_rate": 4.36748237245956e-06, "loss": 0.0872, "step": 35770 }, { "epoch": 2.43, "learning_rate": 4.366539723238189e-06, "loss": 0.3794, "step": 35775 }, { "epoch": 2.43, "learning_rate": 4.365597074016817e-06, "loss": 0.244, "step": 35780 }, { "epoch": 2.43, "learning_rate": 4.364654424795445e-06, "loss": 0.2808, "step": 35785 }, { "epoch": 2.43, "learning_rate": 4.3637117755740735e-06, "loss": 0.4984, "step": 35790 }, { "epoch": 2.43, "learning_rate": 4.362769126352702e-06, "loss": 0.2507, "step": 35795 }, { "epoch": 2.43, "learning_rate": 4.3618264771313305e-06, "loss": 0.4387, "step": 35800 }, { "epoch": 2.43, "learning_rate": 4.360883827909959e-06, "loss": 0.3067, "step": 35805 }, { "epoch": 2.43, "learning_rate": 4.3599411786885874e-06, "loss": 0.2502, "step": 35810 }, { "epoch": 2.43, "learning_rate": 4.358998529467215e-06, "loss": 0.1347, "step": 35815 }, { "epoch": 2.43, "learning_rate": 4.3580558802458436e-06, "loss": 0.435, "step": 35820 }, { "epoch": 2.43, "learning_rate": 4.357113231024471e-06, "loss": 0.5817, "step": 35825 }, { "epoch": 2.43, "learning_rate": 4.3561705818031e-06, "loss": 0.2346, "step": 35830 }, { "epoch": 2.43, "learning_rate": 4.355227932581728e-06, "loss": 0.2983, "step": 35835 }, { "epoch": 2.43, "learning_rate": 4.354285283360356e-06, "loss": 0.2783, "step": 35840 }, { "epoch": 2.43, "learning_rate": 4.353342634138984e-06, "loss": 0.3351, "step": 35845 }, { "epoch": 2.43, "learning_rate": 4.352399984917613e-06, "loss": 0.4834, "step": 35850 }, { "epoch": 2.43, "learning_rate": 4.3514573356962404e-06, "loss": 0.3181, "step": 35855 }, { "epoch": 2.43, "learning_rate": 4.350514686474869e-06, "loss": 0.3219, "step": 35860 }, { "epoch": 2.43, "learning_rate": 4.349572037253497e-06, "loss": 0.3531, "step": 35865 }, { "epoch": 2.43, "learning_rate": 4.348629388032126e-06, "loss": 0.5498, "step": 35870 }, { "epoch": 2.43, "learning_rate": 4.347686738810754e-06, "loss": 0.147, "step": 35875 }, { "epoch": 2.44, "learning_rate": 4.346744089589383e-06, "loss": 0.5576, "step": 35880 }, { "epoch": 2.44, "learning_rate": 4.3458014403680105e-06, "loss": 0.1681, "step": 35885 }, { "epoch": 2.44, "learning_rate": 4.344858791146639e-06, "loss": 0.2778, "step": 35890 }, { "epoch": 2.44, "learning_rate": 4.3439161419252675e-06, "loss": 0.1372, "step": 35895 }, { "epoch": 2.44, "learning_rate": 4.342973492703895e-06, "loss": 0.2629, "step": 35900 }, { "epoch": 2.44, "learning_rate": 4.342030843482524e-06, "loss": 0.3437, "step": 35905 }, { "epoch": 2.44, "learning_rate": 4.341088194261152e-06, "loss": 0.4806, "step": 35910 }, { "epoch": 2.44, "learning_rate": 4.34014554503978e-06, "loss": 0.4287, "step": 35915 }, { "epoch": 2.44, "learning_rate": 4.339202895818408e-06, "loss": 0.2532, "step": 35920 }, { "epoch": 2.44, "learning_rate": 4.338260246597037e-06, "loss": 0.482, "step": 35925 }, { "epoch": 2.44, "learning_rate": 4.337317597375665e-06, "loss": 0.2279, "step": 35930 }, { "epoch": 2.44, "learning_rate": 4.336374948154294e-06, "loss": 0.4897, "step": 35935 }, { "epoch": 2.44, "learning_rate": 4.335432298932921e-06, "loss": 0.3183, "step": 35940 }, { "epoch": 2.44, "learning_rate": 4.33448964971155e-06, "loss": 0.021, "step": 35945 }, { "epoch": 2.44, "learning_rate": 4.333547000490178e-06, "loss": 0.0582, "step": 35950 }, { "epoch": 2.44, "learning_rate": 4.332604351268806e-06, "loss": 0.1471, "step": 35955 }, { "epoch": 2.44, "learning_rate": 4.3316617020474344e-06, "loss": 0.8275, "step": 35960 }, { "epoch": 2.44, "learning_rate": 4.330719052826063e-06, "loss": 0.1571, "step": 35965 }, { "epoch": 2.44, "learning_rate": 4.3297764036046906e-06, "loss": 0.8444, "step": 35970 }, { "epoch": 2.44, "learning_rate": 4.328833754383319e-06, "loss": 0.2972, "step": 35975 }, { "epoch": 2.44, "learning_rate": 4.3278911051619475e-06, "loss": 0.243, "step": 35980 }, { "epoch": 2.44, "learning_rate": 4.326948455940575e-06, "loss": 0.361, "step": 35985 }, { "epoch": 2.44, "learning_rate": 4.326005806719204e-06, "loss": 0.1536, "step": 35990 }, { "epoch": 2.44, "learning_rate": 4.325063157497832e-06, "loss": 0.4261, "step": 35995 }, { "epoch": 2.44, "learning_rate": 4.324120508276461e-06, "loss": 0.6069, "step": 36000 }, { "epoch": 2.44, "learning_rate": 4.323177859055089e-06, "loss": 0.0743, "step": 36005 }, { "epoch": 2.44, "learning_rate": 4.322235209833718e-06, "loss": 0.3129, "step": 36010 }, { "epoch": 2.44, "learning_rate": 4.321292560612345e-06, "loss": 0.6187, "step": 36015 }, { "epoch": 2.44, "learning_rate": 4.320349911390974e-06, "loss": 0.3796, "step": 36020 }, { "epoch": 2.45, "learning_rate": 4.319407262169602e-06, "loss": 0.2834, "step": 36025 }, { "epoch": 2.45, "learning_rate": 4.31846461294823e-06, "loss": 0.2265, "step": 36030 }, { "epoch": 2.45, "learning_rate": 4.317521963726858e-06, "loss": 0.4793, "step": 36035 }, { "epoch": 2.45, "learning_rate": 4.316579314505487e-06, "loss": 0.3849, "step": 36040 }, { "epoch": 2.45, "learning_rate": 4.3156366652841145e-06, "loss": 0.2662, "step": 36045 }, { "epoch": 2.45, "learning_rate": 4.314694016062743e-06, "loss": 0.284, "step": 36050 }, { "epoch": 2.45, "learning_rate": 4.3137513668413715e-06, "loss": 0.1462, "step": 36055 }, { "epoch": 2.45, "learning_rate": 4.31280871762e-06, "loss": 0.2765, "step": 36060 }, { "epoch": 2.45, "learning_rate": 4.311866068398628e-06, "loss": 0.3555, "step": 36065 }, { "epoch": 2.45, "learning_rate": 4.310923419177256e-06, "loss": 0.394, "step": 36070 }, { "epoch": 2.45, "learning_rate": 4.3099807699558846e-06, "loss": 0.3233, "step": 36075 }, { "epoch": 2.45, "learning_rate": 4.309038120734513e-06, "loss": 0.3489, "step": 36080 }, { "epoch": 2.45, "learning_rate": 4.308095471513141e-06, "loss": 0.2975, "step": 36085 }, { "epoch": 2.45, "learning_rate": 4.307152822291769e-06, "loss": 0.2589, "step": 36090 }, { "epoch": 2.45, "learning_rate": 4.306210173070398e-06, "loss": 0.4637, "step": 36095 }, { "epoch": 2.45, "learning_rate": 4.305267523849025e-06, "loss": 0.5153, "step": 36100 }, { "epoch": 2.45, "learning_rate": 4.304324874627654e-06, "loss": 1.1813, "step": 36105 }, { "epoch": 2.45, "learning_rate": 4.303382225406282e-06, "loss": 0.2935, "step": 36110 }, { "epoch": 2.45, "learning_rate": 4.30243957618491e-06, "loss": 0.2062, "step": 36115 }, { "epoch": 2.45, "learning_rate": 4.301496926963538e-06, "loss": 0.3571, "step": 36120 }, { "epoch": 2.45, "learning_rate": 4.300554277742167e-06, "loss": 0.2987, "step": 36125 }, { "epoch": 2.45, "learning_rate": 4.299611628520795e-06, "loss": 0.409, "step": 36130 }, { "epoch": 2.45, "learning_rate": 4.298668979299424e-06, "loss": 0.0896, "step": 36135 }, { "epoch": 2.45, "learning_rate": 4.297726330078052e-06, "loss": 0.284, "step": 36140 }, { "epoch": 2.45, "learning_rate": 4.29678368085668e-06, "loss": 0.0965, "step": 36145 }, { "epoch": 2.45, "learning_rate": 4.2958410316353085e-06, "loss": 0.8389, "step": 36150 }, { "epoch": 2.45, "learning_rate": 4.294898382413937e-06, "loss": 0.4035, "step": 36155 }, { "epoch": 2.45, "learning_rate": 4.293955733192565e-06, "loss": 0.2524, "step": 36160 }, { "epoch": 2.45, "learning_rate": 4.293013083971193e-06, "loss": 0.4317, "step": 36165 }, { "epoch": 2.45, "learning_rate": 4.292070434749821e-06, "loss": 0.5829, "step": 36170 }, { "epoch": 2.46, "learning_rate": 4.291127785528449e-06, "loss": 0.2995, "step": 36175 }, { "epoch": 2.46, "learning_rate": 4.290185136307078e-06, "loss": 0.4237, "step": 36180 }, { "epoch": 2.46, "learning_rate": 4.289242487085706e-06, "loss": 0.4212, "step": 36185 }, { "epoch": 2.46, "learning_rate": 4.288299837864334e-06, "loss": 0.529, "step": 36190 }, { "epoch": 2.46, "learning_rate": 4.287357188642962e-06, "loss": 0.2764, "step": 36195 }, { "epoch": 2.46, "learning_rate": 4.286414539421591e-06, "loss": 0.5739, "step": 36200 }, { "epoch": 2.46, "learning_rate": 4.285471890200219e-06, "loss": 0.3571, "step": 36205 }, { "epoch": 2.46, "learning_rate": 4.284529240978848e-06, "loss": 0.5794, "step": 36210 }, { "epoch": 2.46, "learning_rate": 4.2835865917574754e-06, "loss": 0.0503, "step": 36215 }, { "epoch": 2.46, "learning_rate": 4.282643942536104e-06, "loss": 1.4129, "step": 36220 }, { "epoch": 2.46, "learning_rate": 4.281701293314732e-06, "loss": 0.3018, "step": 36225 }, { "epoch": 2.46, "learning_rate": 4.28075864409336e-06, "loss": 0.1098, "step": 36230 }, { "epoch": 2.46, "learning_rate": 4.2798159948719885e-06, "loss": 0.1076, "step": 36235 }, { "epoch": 2.46, "learning_rate": 4.278873345650617e-06, "loss": 0.1462, "step": 36240 }, { "epoch": 2.46, "learning_rate": 4.277930696429245e-06, "loss": 0.2318, "step": 36245 }, { "epoch": 2.46, "learning_rate": 4.276988047207873e-06, "loss": 0.6601, "step": 36250 }, { "epoch": 2.46, "learning_rate": 4.276045397986502e-06, "loss": 0.2525, "step": 36255 }, { "epoch": 2.46, "learning_rate": 4.27510274876513e-06, "loss": 0.3245, "step": 36260 }, { "epoch": 2.46, "learning_rate": 4.274160099543759e-06, "loss": 0.6966, "step": 36265 }, { "epoch": 2.46, "learning_rate": 4.273217450322386e-06, "loss": 0.7008, "step": 36270 }, { "epoch": 2.46, "learning_rate": 4.272274801101015e-06, "loss": 0.3302, "step": 36275 }, { "epoch": 2.46, "learning_rate": 4.271332151879643e-06, "loss": 0.3579, "step": 36280 }, { "epoch": 2.46, "learning_rate": 4.270389502658271e-06, "loss": 0.3021, "step": 36285 }, { "epoch": 2.46, "learning_rate": 4.269446853436899e-06, "loss": 0.7056, "step": 36290 }, { "epoch": 2.46, "learning_rate": 4.268504204215528e-06, "loss": 0.8557, "step": 36295 }, { "epoch": 2.46, "learning_rate": 4.2675615549941555e-06, "loss": 0.2031, "step": 36300 }, { "epoch": 2.46, "learning_rate": 4.266618905772784e-06, "loss": 0.2595, "step": 36305 }, { "epoch": 2.46, "learning_rate": 4.2656762565514124e-06, "loss": 0.1246, "step": 36310 }, { "epoch": 2.46, "learning_rate": 4.26473360733004e-06, "loss": 0.0457, "step": 36315 }, { "epoch": 2.47, "learning_rate": 4.263790958108669e-06, "loss": 0.1615, "step": 36320 }, { "epoch": 2.47, "learning_rate": 4.262848308887297e-06, "loss": 0.2342, "step": 36325 }, { "epoch": 2.47, "learning_rate": 4.2619056596659256e-06, "loss": 0.052, "step": 36330 }, { "epoch": 2.47, "learning_rate": 4.260963010444554e-06, "loss": 0.2878, "step": 36335 }, { "epoch": 2.47, "learning_rate": 4.2600203612231825e-06, "loss": 0.3574, "step": 36340 }, { "epoch": 2.47, "learning_rate": 4.25907771200181e-06, "loss": 0.2498, "step": 36345 }, { "epoch": 2.47, "learning_rate": 4.258135062780439e-06, "loss": 0.4931, "step": 36350 }, { "epoch": 2.47, "learning_rate": 4.257192413559067e-06, "loss": 0.4682, "step": 36355 }, { "epoch": 2.47, "learning_rate": 4.256249764337695e-06, "loss": 0.4749, "step": 36360 }, { "epoch": 2.47, "learning_rate": 4.255307115116323e-06, "loss": 0.2453, "step": 36365 }, { "epoch": 2.47, "learning_rate": 4.254364465894952e-06, "loss": 0.7597, "step": 36370 }, { "epoch": 2.47, "learning_rate": 4.253421816673579e-06, "loss": 0.3227, "step": 36375 }, { "epoch": 2.47, "learning_rate": 4.252479167452208e-06, "loss": 0.2897, "step": 36380 }, { "epoch": 2.47, "learning_rate": 4.251536518230836e-06, "loss": 0.7677, "step": 36385 }, { "epoch": 2.47, "learning_rate": 4.250593869009465e-06, "loss": 0.4411, "step": 36390 }, { "epoch": 2.47, "learning_rate": 4.2496512197880925e-06, "loss": 0.5336, "step": 36395 }, { "epoch": 2.47, "learning_rate": 4.248708570566721e-06, "loss": 0.366, "step": 36400 }, { "epoch": 2.47, "learning_rate": 4.2477659213453495e-06, "loss": 0.2832, "step": 36405 }, { "epoch": 2.47, "learning_rate": 4.246823272123978e-06, "loss": 0.8229, "step": 36410 }, { "epoch": 2.47, "learning_rate": 4.245880622902606e-06, "loss": 0.4142, "step": 36415 }, { "epoch": 2.47, "learning_rate": 4.244937973681234e-06, "loss": 0.2356, "step": 36420 }, { "epoch": 2.47, "learning_rate": 4.2439953244598626e-06, "loss": 0.6743, "step": 36425 }, { "epoch": 2.47, "learning_rate": 4.24305267523849e-06, "loss": 0.236, "step": 36430 }, { "epoch": 2.47, "learning_rate": 4.242110026017119e-06, "loss": 0.3158, "step": 36435 }, { "epoch": 2.47, "learning_rate": 4.241167376795747e-06, "loss": 0.2232, "step": 36440 }, { "epoch": 2.47, "learning_rate": 4.240224727574375e-06, "loss": 0.1547, "step": 36445 }, { "epoch": 2.47, "learning_rate": 4.239282078353003e-06, "loss": 0.3851, "step": 36450 }, { "epoch": 2.47, "learning_rate": 4.238339429131632e-06, "loss": 0.2864, "step": 36455 }, { "epoch": 2.47, "learning_rate": 4.23739677991026e-06, "loss": 0.295, "step": 36460 }, { "epoch": 2.47, "learning_rate": 4.236454130688889e-06, "loss": 0.3103, "step": 36465 }, { "epoch": 2.48, "learning_rate": 4.235511481467517e-06, "loss": 0.2878, "step": 36470 }, { "epoch": 2.48, "learning_rate": 4.234568832246145e-06, "loss": 0.7864, "step": 36475 }, { "epoch": 2.48, "learning_rate": 4.233626183024773e-06, "loss": 0.204, "step": 36480 }, { "epoch": 2.48, "learning_rate": 4.232683533803402e-06, "loss": 0.3238, "step": 36485 }, { "epoch": 2.48, "learning_rate": 4.2317408845820295e-06, "loss": 0.6915, "step": 36490 }, { "epoch": 2.48, "learning_rate": 4.230798235360658e-06, "loss": 0.1773, "step": 36495 }, { "epoch": 2.48, "learning_rate": 4.229855586139286e-06, "loss": 0.1235, "step": 36500 }, { "epoch": 2.48, "learning_rate": 4.228912936917914e-06, "loss": 0.2032, "step": 36505 }, { "epoch": 2.48, "learning_rate": 4.227970287696543e-06, "loss": 0.3789, "step": 36510 }, { "epoch": 2.48, "learning_rate": 4.227027638475171e-06, "loss": 0.1941, "step": 36515 }, { "epoch": 2.48, "learning_rate": 4.226084989253799e-06, "loss": 0.4317, "step": 36520 }, { "epoch": 2.48, "learning_rate": 4.225142340032427e-06, "loss": 0.1381, "step": 36525 }, { "epoch": 2.48, "learning_rate": 4.224199690811056e-06, "loss": 0.7204, "step": 36530 }, { "epoch": 2.48, "learning_rate": 4.223257041589684e-06, "loss": 0.6464, "step": 36535 }, { "epoch": 2.48, "learning_rate": 4.222314392368313e-06, "loss": 0.2415, "step": 36540 }, { "epoch": 2.48, "learning_rate": 4.22137174314694e-06, "loss": 0.1965, "step": 36545 }, { "epoch": 2.48, "learning_rate": 4.220429093925569e-06, "loss": 0.6986, "step": 36550 }, { "epoch": 2.48, "learning_rate": 4.219486444704197e-06, "loss": 0.4017, "step": 36555 }, { "epoch": 2.48, "learning_rate": 4.218543795482825e-06, "loss": 0.4318, "step": 36560 }, { "epoch": 2.48, "learning_rate": 4.2176011462614534e-06, "loss": 0.0815, "step": 36565 }, { "epoch": 2.48, "learning_rate": 4.216658497040082e-06, "loss": 0.2897, "step": 36570 }, { "epoch": 2.48, "learning_rate": 4.2157158478187096e-06, "loss": 0.1573, "step": 36575 }, { "epoch": 2.48, "learning_rate": 4.214773198597338e-06, "loss": 0.1982, "step": 36580 }, { "epoch": 2.48, "learning_rate": 4.2138305493759665e-06, "loss": 0.3611, "step": 36585 }, { "epoch": 2.48, "learning_rate": 4.212887900154595e-06, "loss": 0.2211, "step": 36590 }, { "epoch": 2.48, "learning_rate": 4.2119452509332235e-06, "loss": 0.4679, "step": 36595 }, { "epoch": 2.48, "learning_rate": 4.211002601711852e-06, "loss": 0.5852, "step": 36600 }, { "epoch": 2.48, "learning_rate": 4.21005995249048e-06, "loss": 0.7813, "step": 36605 }, { "epoch": 2.48, "learning_rate": 4.209117303269108e-06, "loss": 0.3676, "step": 36610 }, { "epoch": 2.49, "learning_rate": 4.208174654047736e-06, "loss": 0.4234, "step": 36615 }, { "epoch": 2.49, "learning_rate": 4.207232004826364e-06, "loss": 0.4473, "step": 36620 }, { "epoch": 2.49, "learning_rate": 4.206289355604993e-06, "loss": 0.1735, "step": 36625 }, { "epoch": 2.49, "learning_rate": 4.20534670638362e-06, "loss": 0.4524, "step": 36630 }, { "epoch": 2.49, "learning_rate": 4.204404057162249e-06, "loss": 0.2177, "step": 36635 }, { "epoch": 2.49, "learning_rate": 4.203461407940877e-06, "loss": 0.386, "step": 36640 }, { "epoch": 2.49, "learning_rate": 4.202518758719506e-06, "loss": 0.1584, "step": 36645 }, { "epoch": 2.49, "learning_rate": 4.2015761094981335e-06, "loss": 0.1861, "step": 36650 }, { "epoch": 2.49, "learning_rate": 4.200633460276762e-06, "loss": 0.375, "step": 36655 }, { "epoch": 2.49, "learning_rate": 4.1996908110553905e-06, "loss": 0.3888, "step": 36660 }, { "epoch": 2.49, "learning_rate": 4.198748161834019e-06, "loss": 0.7795, "step": 36665 }, { "epoch": 2.49, "learning_rate": 4.1978055126126474e-06, "loss": 0.3452, "step": 36670 }, { "epoch": 2.49, "learning_rate": 4.196862863391275e-06, "loss": 0.3754, "step": 36675 }, { "epoch": 2.49, "learning_rate": 4.1959202141699036e-06, "loss": 0.2266, "step": 36680 }, { "epoch": 2.49, "learning_rate": 4.194977564948532e-06, "loss": 0.5571, "step": 36685 }, { "epoch": 2.49, "learning_rate": 4.19403491572716e-06, "loss": 0.2735, "step": 36690 }, { "epoch": 2.49, "learning_rate": 4.193092266505788e-06, "loss": 0.6055, "step": 36695 }, { "epoch": 2.49, "learning_rate": 4.192149617284417e-06, "loss": 0.4789, "step": 36700 }, { "epoch": 2.49, "learning_rate": 4.191206968063044e-06, "loss": 0.1055, "step": 36705 }, { "epoch": 2.49, "learning_rate": 4.190264318841673e-06, "loss": 0.5621, "step": 36710 }, { "epoch": 2.49, "learning_rate": 4.189321669620301e-06, "loss": 0.2059, "step": 36715 }, { "epoch": 2.49, "learning_rate": 4.18837902039893e-06, "loss": 0.3759, "step": 36720 }, { "epoch": 2.49, "learning_rate": 4.187436371177558e-06, "loss": 0.3149, "step": 36725 }, { "epoch": 2.49, "learning_rate": 4.186493721956186e-06, "loss": 0.6232, "step": 36730 }, { "epoch": 2.49, "learning_rate": 4.185551072734814e-06, "loss": 0.5626, "step": 36735 }, { "epoch": 2.49, "learning_rate": 4.184608423513443e-06, "loss": 0.671, "step": 36740 }, { "epoch": 2.49, "learning_rate": 4.1836657742920705e-06, "loss": 0.2251, "step": 36745 }, { "epoch": 2.49, "learning_rate": 4.182723125070699e-06, "loss": 0.4435, "step": 36750 }, { "epoch": 2.49, "learning_rate": 4.1817804758493275e-06, "loss": 0.1487, "step": 36755 }, { "epoch": 2.49, "learning_rate": 4.180837826627955e-06, "loss": 0.0894, "step": 36760 }, { "epoch": 2.5, "learning_rate": 4.179895177406584e-06, "loss": 0.6278, "step": 36765 }, { "epoch": 2.5, "learning_rate": 4.178952528185212e-06, "loss": 0.1485, "step": 36770 }, { "epoch": 2.5, "learning_rate": 4.17800987896384e-06, "loss": 0.7204, "step": 36775 }, { "epoch": 2.5, "learning_rate": 4.177067229742468e-06, "loss": 0.5614, "step": 36780 }, { "epoch": 2.5, "learning_rate": 4.176124580521097e-06, "loss": 0.1594, "step": 36785 }, { "epoch": 2.5, "learning_rate": 4.175181931299725e-06, "loss": 0.4166, "step": 36790 }, { "epoch": 2.5, "learning_rate": 4.174239282078354e-06, "loss": 0.2881, "step": 36795 }, { "epoch": 2.5, "learning_rate": 4.173296632856982e-06, "loss": 0.4875, "step": 36800 }, { "epoch": 2.5, "learning_rate": 4.17235398363561e-06, "loss": 0.1355, "step": 36805 }, { "epoch": 2.5, "learning_rate": 4.171411334414238e-06, "loss": 0.1732, "step": 36810 }, { "epoch": 2.5, "learning_rate": 4.170468685192867e-06, "loss": 0.0537, "step": 36815 }, { "epoch": 2.5, "learning_rate": 4.1695260359714944e-06, "loss": 0.1117, "step": 36820 }, { "epoch": 2.5, "learning_rate": 4.168583386750123e-06, "loss": 0.6638, "step": 36825 }, { "epoch": 2.5, "learning_rate": 4.1676407375287506e-06, "loss": 0.034, "step": 36830 }, { "epoch": 2.5, "learning_rate": 4.166698088307379e-06, "loss": 0.251, "step": 36835 }, { "epoch": 2.5, "learning_rate": 4.1657554390860075e-06, "loss": 0.262, "step": 36840 }, { "epoch": 2.5, "learning_rate": 4.164812789864636e-06, "loss": 0.1506, "step": 36845 }, { "epoch": 2.5, "learning_rate": 4.1638701406432645e-06, "loss": 0.1551, "step": 36850 }, { "epoch": 2.5, "learning_rate": 4.162927491421892e-06, "loss": 0.351, "step": 36855 }, { "epoch": 2.5, "learning_rate": 4.161984842200521e-06, "loss": 0.2604, "step": 36860 }, { "epoch": 2.5, "learning_rate": 4.161042192979149e-06, "loss": 0.1365, "step": 36865 }, { "epoch": 2.5, "learning_rate": 4.160099543757778e-06, "loss": 0.4422, "step": 36870 }, { "epoch": 2.5, "learning_rate": 4.159156894536405e-06, "loss": 0.5662, "step": 36875 }, { "epoch": 2.5, "learning_rate": 4.158214245315034e-06, "loss": 0.6683, "step": 36880 }, { "epoch": 2.5, "learning_rate": 4.157271596093662e-06, "loss": 0.3725, "step": 36885 }, { "epoch": 2.5, "learning_rate": 4.15632894687229e-06, "loss": 0.2224, "step": 36890 }, { "epoch": 2.5, "learning_rate": 4.155386297650918e-06, "loss": 0.5603, "step": 36895 }, { "epoch": 2.5, "learning_rate": 4.154443648429547e-06, "loss": 0.575, "step": 36900 }, { "epoch": 2.5, "learning_rate": 4.1535009992081745e-06, "loss": 0.1271, "step": 36905 }, { "epoch": 2.51, "learning_rate": 4.152558349986803e-06, "loss": 0.1672, "step": 36910 }, { "epoch": 2.51, "learning_rate": 4.1516157007654315e-06, "loss": 0.2954, "step": 36915 }, { "epoch": 2.51, "learning_rate": 4.15067305154406e-06, "loss": 0.1497, "step": 36920 }, { "epoch": 2.51, "learning_rate": 4.1497304023226884e-06, "loss": 0.3944, "step": 36925 }, { "epoch": 2.51, "learning_rate": 4.148787753101317e-06, "loss": 0.3627, "step": 36930 }, { "epoch": 2.51, "learning_rate": 4.1478451038799446e-06, "loss": 0.2499, "step": 36935 }, { "epoch": 2.51, "learning_rate": 4.146902454658573e-06, "loss": 0.1998, "step": 36940 }, { "epoch": 2.51, "learning_rate": 4.145959805437201e-06, "loss": 0.5911, "step": 36945 }, { "epoch": 2.51, "learning_rate": 4.145017156215829e-06, "loss": 0.2045, "step": 36950 }, { "epoch": 2.51, "learning_rate": 4.144074506994458e-06, "loss": 0.5331, "step": 36955 }, { "epoch": 2.51, "learning_rate": 4.143131857773085e-06, "loss": 0.2278, "step": 36960 }, { "epoch": 2.51, "learning_rate": 4.142189208551714e-06, "loss": 0.7312, "step": 36965 }, { "epoch": 2.51, "learning_rate": 4.141246559330342e-06, "loss": 0.1856, "step": 36970 }, { "epoch": 2.51, "learning_rate": 4.140303910108971e-06, "loss": 0.8428, "step": 36975 }, { "epoch": 2.51, "learning_rate": 4.139361260887598e-06, "loss": 0.7999, "step": 36980 }, { "epoch": 2.51, "learning_rate": 4.138418611666227e-06, "loss": 0.2323, "step": 36985 }, { "epoch": 2.51, "learning_rate": 4.137475962444855e-06, "loss": 0.1344, "step": 36990 }, { "epoch": 2.51, "learning_rate": 4.136533313223484e-06, "loss": 0.6604, "step": 36995 }, { "epoch": 2.51, "learning_rate": 4.135590664002112e-06, "loss": 0.2324, "step": 37000 }, { "epoch": 2.51, "learning_rate": 4.13464801478074e-06, "loss": 0.3526, "step": 37005 }, { "epoch": 2.51, "learning_rate": 4.1337053655593685e-06, "loss": 0.5245, "step": 37010 }, { "epoch": 2.51, "learning_rate": 4.132762716337997e-06, "loss": 0.6092, "step": 37015 }, { "epoch": 2.51, "learning_rate": 4.131820067116625e-06, "loss": 0.4616, "step": 37020 }, { "epoch": 2.51, "learning_rate": 4.130877417895253e-06, "loss": 0.2092, "step": 37025 }, { "epoch": 2.51, "learning_rate": 4.129934768673882e-06, "loss": 0.5098, "step": 37030 }, { "epoch": 2.51, "learning_rate": 4.128992119452509e-06, "loss": 0.2506, "step": 37035 }, { "epoch": 2.51, "learning_rate": 4.128049470231138e-06, "loss": 0.4317, "step": 37040 }, { "epoch": 2.51, "learning_rate": 4.127106821009766e-06, "loss": 0.2915, "step": 37045 }, { "epoch": 2.51, "learning_rate": 4.126164171788395e-06, "loss": 0.2945, "step": 37050 }, { "epoch": 2.51, "learning_rate": 4.125221522567023e-06, "loss": 0.1591, "step": 37055 }, { "epoch": 2.52, "learning_rate": 4.124278873345651e-06, "loss": 0.6516, "step": 37060 }, { "epoch": 2.52, "learning_rate": 4.123336224124279e-06, "loss": 0.0406, "step": 37065 }, { "epoch": 2.52, "learning_rate": 4.122393574902908e-06, "loss": 0.2454, "step": 37070 }, { "epoch": 2.52, "learning_rate": 4.1214509256815354e-06, "loss": 0.5431, "step": 37075 }, { "epoch": 2.52, "learning_rate": 4.120508276460164e-06, "loss": 0.4021, "step": 37080 }, { "epoch": 2.52, "learning_rate": 4.119565627238792e-06, "loss": 0.4629, "step": 37085 }, { "epoch": 2.52, "learning_rate": 4.11862297801742e-06, "loss": 0.234, "step": 37090 }, { "epoch": 2.52, "learning_rate": 4.1176803287960485e-06, "loss": 0.1513, "step": 37095 }, { "epoch": 2.52, "learning_rate": 4.116737679574677e-06, "loss": 0.2482, "step": 37100 }, { "epoch": 2.52, "learning_rate": 4.115795030353305e-06, "loss": 0.4538, "step": 37105 }, { "epoch": 2.52, "learning_rate": 4.114852381131933e-06, "loss": 0.517, "step": 37110 }, { "epoch": 2.52, "learning_rate": 4.113909731910562e-06, "loss": 0.765, "step": 37115 }, { "epoch": 2.52, "learning_rate": 4.11296708268919e-06, "loss": 0.2385, "step": 37120 }, { "epoch": 2.52, "learning_rate": 4.112024433467819e-06, "loss": 0.2159, "step": 37125 }, { "epoch": 2.52, "learning_rate": 4.111081784246447e-06, "loss": 0.2766, "step": 37130 }, { "epoch": 2.52, "learning_rate": 4.110139135025075e-06, "loss": 0.1327, "step": 37135 }, { "epoch": 2.52, "learning_rate": 4.109196485803703e-06, "loss": 0.3043, "step": 37140 }, { "epoch": 2.52, "learning_rate": 4.108253836582332e-06, "loss": 0.6389, "step": 37145 }, { "epoch": 2.52, "learning_rate": 4.107311187360959e-06, "loss": 0.213, "step": 37150 }, { "epoch": 2.52, "learning_rate": 4.106368538139588e-06, "loss": 0.5485, "step": 37155 }, { "epoch": 2.52, "learning_rate": 4.105425888918216e-06, "loss": 0.2604, "step": 37160 }, { "epoch": 2.52, "learning_rate": 4.104483239696844e-06, "loss": 0.5747, "step": 37165 }, { "epoch": 2.52, "learning_rate": 4.1035405904754725e-06, "loss": 0.4325, "step": 37170 }, { "epoch": 2.52, "learning_rate": 4.102597941254101e-06, "loss": 0.4144, "step": 37175 }, { "epoch": 2.52, "learning_rate": 4.1016552920327294e-06, "loss": 0.1281, "step": 37180 }, { "epoch": 2.52, "learning_rate": 4.100712642811358e-06, "loss": 0.39, "step": 37185 }, { "epoch": 2.52, "learning_rate": 4.0997699935899856e-06, "loss": 0.2065, "step": 37190 }, { "epoch": 2.52, "learning_rate": 4.098827344368614e-06, "loss": 0.3837, "step": 37195 }, { "epoch": 2.52, "learning_rate": 4.0978846951472425e-06, "loss": 0.2725, "step": 37200 }, { "epoch": 2.53, "learning_rate": 4.09694204592587e-06, "loss": 0.1203, "step": 37205 }, { "epoch": 2.53, "learning_rate": 4.095999396704499e-06, "loss": 0.2867, "step": 37210 }, { "epoch": 2.53, "learning_rate": 4.095056747483127e-06, "loss": 0.2671, "step": 37215 }, { "epoch": 2.53, "learning_rate": 4.094114098261755e-06, "loss": 0.4695, "step": 37220 }, { "epoch": 2.53, "learning_rate": 4.093171449040383e-06, "loss": 0.7251, "step": 37225 }, { "epoch": 2.53, "learning_rate": 4.092228799819012e-06, "loss": 0.3453, "step": 37230 }, { "epoch": 2.53, "learning_rate": 4.091286150597639e-06, "loss": 0.3934, "step": 37235 }, { "epoch": 2.53, "learning_rate": 4.090343501376268e-06, "loss": 0.3152, "step": 37240 }, { "epoch": 2.53, "learning_rate": 4.089400852154896e-06, "loss": 0.3822, "step": 37245 }, { "epoch": 2.53, "learning_rate": 4.088458202933525e-06, "loss": 0.5684, "step": 37250 }, { "epoch": 2.53, "learning_rate": 4.087515553712153e-06, "loss": 0.3009, "step": 37255 }, { "epoch": 2.53, "learning_rate": 4.086572904490782e-06, "loss": 0.2682, "step": 37260 }, { "epoch": 2.53, "learning_rate": 4.0856302552694095e-06, "loss": 0.5703, "step": 37265 }, { "epoch": 2.53, "learning_rate": 4.084687606048038e-06, "loss": 0.3643, "step": 37270 }, { "epoch": 2.53, "learning_rate": 4.083744956826666e-06, "loss": 0.4654, "step": 37275 }, { "epoch": 2.53, "learning_rate": 4.082802307605294e-06, "loss": 0.3946, "step": 37280 }, { "epoch": 2.53, "learning_rate": 4.081859658383923e-06, "loss": 0.3596, "step": 37285 }, { "epoch": 2.53, "learning_rate": 4.08091700916255e-06, "loss": 0.3827, "step": 37290 }, { "epoch": 2.53, "learning_rate": 4.079974359941179e-06, "loss": 0.632, "step": 37295 }, { "epoch": 2.53, "learning_rate": 4.079031710719807e-06, "loss": 0.1831, "step": 37300 }, { "epoch": 2.53, "learning_rate": 4.078089061498436e-06, "loss": 0.1567, "step": 37305 }, { "epoch": 2.53, "learning_rate": 4.077146412277064e-06, "loss": 0.1626, "step": 37310 }, { "epoch": 2.53, "learning_rate": 4.076203763055692e-06, "loss": 0.4478, "step": 37315 }, { "epoch": 2.53, "learning_rate": 4.07526111383432e-06, "loss": 0.2649, "step": 37320 }, { "epoch": 2.53, "learning_rate": 4.074318464612949e-06, "loss": 0.3371, "step": 37325 }, { "epoch": 2.53, "learning_rate": 4.073375815391577e-06, "loss": 0.6072, "step": 37330 }, { "epoch": 2.53, "learning_rate": 4.072433166170205e-06, "loss": 0.1816, "step": 37335 }, { "epoch": 2.53, "learning_rate": 4.071490516948833e-06, "loss": 0.2069, "step": 37340 }, { "epoch": 2.53, "learning_rate": 4.070547867727462e-06, "loss": 0.154, "step": 37345 }, { "epoch": 2.53, "learning_rate": 4.0696052185060895e-06, "loss": 0.3667, "step": 37350 }, { "epoch": 2.54, "learning_rate": 4.068662569284718e-06, "loss": 0.0937, "step": 37355 }, { "epoch": 2.54, "learning_rate": 4.0677199200633465e-06, "loss": 0.2375, "step": 37360 }, { "epoch": 2.54, "learning_rate": 4.066777270841974e-06, "loss": 0.3405, "step": 37365 }, { "epoch": 2.54, "learning_rate": 4.065834621620603e-06, "loss": 0.5003, "step": 37370 }, { "epoch": 2.54, "learning_rate": 4.064891972399231e-06, "loss": 0.3087, "step": 37375 }, { "epoch": 2.54, "learning_rate": 4.06394932317786e-06, "loss": 0.5471, "step": 37380 }, { "epoch": 2.54, "learning_rate": 4.063006673956488e-06, "loss": 0.15, "step": 37385 }, { "epoch": 2.54, "learning_rate": 4.062064024735116e-06, "loss": 0.2913, "step": 37390 }, { "epoch": 2.54, "learning_rate": 4.061121375513744e-06, "loss": 0.3728, "step": 37395 }, { "epoch": 2.54, "learning_rate": 4.060178726292373e-06, "loss": 0.1009, "step": 37400 }, { "epoch": 2.54, "learning_rate": 4.059236077071e-06, "loss": 0.4433, "step": 37405 }, { "epoch": 2.54, "learning_rate": 4.058293427849629e-06, "loss": 0.3078, "step": 37410 }, { "epoch": 2.54, "learning_rate": 4.057350778628257e-06, "loss": 0.2833, "step": 37415 }, { "epoch": 2.54, "learning_rate": 4.056408129406885e-06, "loss": 0.5667, "step": 37420 }, { "epoch": 2.54, "learning_rate": 4.0554654801855134e-06, "loss": 1.1504, "step": 37425 }, { "epoch": 2.54, "learning_rate": 4.054522830964142e-06, "loss": 0.2486, "step": 37430 }, { "epoch": 2.54, "learning_rate": 4.0535801817427704e-06, "loss": 0.2238, "step": 37435 }, { "epoch": 2.54, "learning_rate": 4.052637532521398e-06, "loss": 0.534, "step": 37440 }, { "epoch": 2.54, "learning_rate": 4.0516948833000266e-06, "loss": 0.2653, "step": 37445 }, { "epoch": 2.54, "learning_rate": 4.050752234078655e-06, "loss": 0.4616, "step": 37450 }, { "epoch": 2.54, "learning_rate": 4.0498095848572835e-06, "loss": 0.127, "step": 37455 }, { "epoch": 2.54, "learning_rate": 4.048866935635912e-06, "loss": 0.1433, "step": 37460 }, { "epoch": 2.54, "learning_rate": 4.04792428641454e-06, "loss": 0.3391, "step": 37465 }, { "epoch": 2.54, "learning_rate": 4.046981637193168e-06, "loss": 0.3054, "step": 37470 }, { "epoch": 2.54, "learning_rate": 4.046038987971797e-06, "loss": 1.0128, "step": 37475 }, { "epoch": 2.54, "learning_rate": 4.045096338750424e-06, "loss": 0.2252, "step": 37480 }, { "epoch": 2.54, "learning_rate": 4.044153689529053e-06, "loss": 0.7364, "step": 37485 }, { "epoch": 2.54, "learning_rate": 4.043211040307681e-06, "loss": 0.4528, "step": 37490 }, { "epoch": 2.54, "learning_rate": 4.042268391086309e-06, "loss": 0.5354, "step": 37495 }, { "epoch": 2.55, "learning_rate": 4.041325741864937e-06, "loss": 0.3535, "step": 37500 }, { "epoch": 2.55, "learning_rate": 4.040383092643566e-06, "loss": 0.4198, "step": 37505 }, { "epoch": 2.55, "learning_rate": 4.039440443422194e-06, "loss": 0.208, "step": 37510 }, { "epoch": 2.55, "learning_rate": 4.038497794200823e-06, "loss": 0.6209, "step": 37515 }, { "epoch": 2.55, "learning_rate": 4.0375551449794505e-06, "loss": 0.4231, "step": 37520 }, { "epoch": 2.55, "learning_rate": 4.036612495758079e-06, "loss": 0.1134, "step": 37525 }, { "epoch": 2.55, "learning_rate": 4.0356698465367074e-06, "loss": 0.4878, "step": 37530 }, { "epoch": 2.55, "learning_rate": 4.034727197315335e-06, "loss": 0.2955, "step": 37535 }, { "epoch": 2.55, "learning_rate": 4.0337845480939636e-06, "loss": 0.1072, "step": 37540 }, { "epoch": 2.55, "learning_rate": 4.032841898872592e-06, "loss": 0.3261, "step": 37545 }, { "epoch": 2.55, "learning_rate": 4.03189924965122e-06, "loss": 0.2868, "step": 37550 }, { "epoch": 2.55, "learning_rate": 4.030956600429848e-06, "loss": 0.2607, "step": 37555 }, { "epoch": 2.55, "learning_rate": 4.030013951208477e-06, "loss": 0.1619, "step": 37560 }, { "epoch": 2.55, "learning_rate": 4.029071301987104e-06, "loss": 0.1994, "step": 37565 }, { "epoch": 2.55, "learning_rate": 4.028128652765733e-06, "loss": 0.1286, "step": 37570 }, { "epoch": 2.55, "learning_rate": 4.027186003544361e-06, "loss": 0.7404, "step": 37575 }, { "epoch": 2.55, "learning_rate": 4.02624335432299e-06, "loss": 0.502, "step": 37580 }, { "epoch": 2.55, "learning_rate": 4.025300705101618e-06, "loss": 0.109, "step": 37585 }, { "epoch": 2.55, "learning_rate": 4.024358055880247e-06, "loss": 0.1605, "step": 37590 }, { "epoch": 2.55, "learning_rate": 4.023415406658874e-06, "loss": 0.4058, "step": 37595 }, { "epoch": 2.55, "learning_rate": 4.022472757437503e-06, "loss": 0.8509, "step": 37600 }, { "epoch": 2.55, "learning_rate": 4.021530108216131e-06, "loss": 0.1729, "step": 37605 }, { "epoch": 2.55, "learning_rate": 4.020587458994759e-06, "loss": 0.2891, "step": 37610 }, { "epoch": 2.55, "learning_rate": 4.0196448097733875e-06, "loss": 0.0938, "step": 37615 }, { "epoch": 2.55, "learning_rate": 4.018702160552015e-06, "loss": 0.1231, "step": 37620 }, { "epoch": 2.55, "learning_rate": 4.017759511330644e-06, "loss": 0.2309, "step": 37625 }, { "epoch": 2.55, "learning_rate": 4.016816862109272e-06, "loss": 0.6852, "step": 37630 }, { "epoch": 2.55, "learning_rate": 4.015874212887901e-06, "loss": 0.1053, "step": 37635 }, { "epoch": 2.55, "learning_rate": 4.014931563666529e-06, "loss": 0.3487, "step": 37640 }, { "epoch": 2.55, "learning_rate": 4.013988914445157e-06, "loss": 0.2984, "step": 37645 }, { "epoch": 2.56, "learning_rate": 4.013046265223785e-06, "loss": 0.4321, "step": 37650 }, { "epoch": 2.56, "learning_rate": 4.012103616002414e-06, "loss": 0.5282, "step": 37655 }, { "epoch": 2.56, "learning_rate": 4.011160966781042e-06, "loss": 0.3315, "step": 37660 }, { "epoch": 2.56, "learning_rate": 4.01021831755967e-06, "loss": 0.4927, "step": 37665 }, { "epoch": 2.56, "learning_rate": 4.009275668338298e-06, "loss": 0.1599, "step": 37670 }, { "epoch": 2.56, "learning_rate": 4.008333019116927e-06, "loss": 0.144, "step": 37675 }, { "epoch": 2.56, "learning_rate": 4.0073903698955544e-06, "loss": 0.3611, "step": 37680 }, { "epoch": 2.56, "learning_rate": 4.006447720674183e-06, "loss": 0.4017, "step": 37685 }, { "epoch": 2.56, "learning_rate": 4.005505071452811e-06, "loss": 0.273, "step": 37690 }, { "epoch": 2.56, "learning_rate": 4.004562422231439e-06, "loss": 0.1627, "step": 37695 }, { "epoch": 2.56, "learning_rate": 4.0036197730100675e-06, "loss": 0.1202, "step": 37700 }, { "epoch": 2.56, "learning_rate": 4.002677123788696e-06, "loss": 0.2438, "step": 37705 }, { "epoch": 2.56, "learning_rate": 4.0017344745673245e-06, "loss": 0.3029, "step": 37710 }, { "epoch": 2.56, "learning_rate": 4.000791825345953e-06, "loss": 0.4087, "step": 37715 }, { "epoch": 2.56, "learning_rate": 3.9998491761245815e-06, "loss": 0.1054, "step": 37720 }, { "epoch": 2.56, "learning_rate": 3.998906526903209e-06, "loss": 0.3913, "step": 37725 }, { "epoch": 2.56, "learning_rate": 3.997963877681838e-06, "loss": 0.2148, "step": 37730 }, { "epoch": 2.56, "learning_rate": 3.997021228460465e-06, "loss": 0.2346, "step": 37735 }, { "epoch": 2.56, "learning_rate": 3.996078579239094e-06, "loss": 0.3422, "step": 37740 }, { "epoch": 2.56, "learning_rate": 3.995135930017722e-06, "loss": 0.5591, "step": 37745 }, { "epoch": 2.56, "learning_rate": 3.99419328079635e-06, "loss": 0.1657, "step": 37750 }, { "epoch": 2.56, "learning_rate": 3.993250631574978e-06, "loss": 0.4652, "step": 37755 }, { "epoch": 2.56, "learning_rate": 3.992307982353607e-06, "loss": 0.1815, "step": 37760 }, { "epoch": 2.56, "learning_rate": 3.991365333132235e-06, "loss": 0.2112, "step": 37765 }, { "epoch": 2.56, "learning_rate": 3.990422683910864e-06, "loss": 0.4696, "step": 37770 }, { "epoch": 2.56, "learning_rate": 3.9894800346894915e-06, "loss": 0.314, "step": 37775 }, { "epoch": 2.56, "learning_rate": 3.98853738546812e-06, "loss": 0.5319, "step": 37780 }, { "epoch": 2.56, "learning_rate": 3.9875947362467484e-06, "loss": 1.0839, "step": 37785 }, { "epoch": 2.56, "learning_rate": 3.986652087025377e-06, "loss": 0.0563, "step": 37790 }, { "epoch": 2.57, "learning_rate": 3.9857094378040046e-06, "loss": 0.1446, "step": 37795 }, { "epoch": 2.57, "learning_rate": 3.984766788582633e-06, "loss": 0.3333, "step": 37800 }, { "epoch": 2.57, "learning_rate": 3.9838241393612615e-06, "loss": 0.6192, "step": 37805 }, { "epoch": 2.57, "learning_rate": 3.982881490139889e-06, "loss": 0.6836, "step": 37810 }, { "epoch": 2.57, "learning_rate": 3.981938840918518e-06, "loss": 0.2504, "step": 37815 }, { "epoch": 2.57, "learning_rate": 3.980996191697146e-06, "loss": 0.5594, "step": 37820 }, { "epoch": 2.57, "learning_rate": 3.980053542475774e-06, "loss": 0.1953, "step": 37825 }, { "epoch": 2.57, "learning_rate": 3.979110893254402e-06, "loss": 0.0908, "step": 37830 }, { "epoch": 2.57, "learning_rate": 3.978168244033031e-06, "loss": 0.2366, "step": 37835 }, { "epoch": 2.57, "learning_rate": 3.977225594811659e-06, "loss": 0.52, "step": 37840 }, { "epoch": 2.57, "learning_rate": 3.976282945590288e-06, "loss": 0.2849, "step": 37845 }, { "epoch": 2.57, "learning_rate": 3.975340296368915e-06, "loss": 0.1314, "step": 37850 }, { "epoch": 2.57, "learning_rate": 3.974397647147544e-06, "loss": 0.2679, "step": 37855 }, { "epoch": 2.57, "learning_rate": 3.973454997926172e-06, "loss": 0.3983, "step": 37860 }, { "epoch": 2.57, "learning_rate": 3.9725123487048e-06, "loss": 0.112, "step": 37865 }, { "epoch": 2.57, "learning_rate": 3.9715696994834285e-06, "loss": 0.4298, "step": 37870 }, { "epoch": 2.57, "learning_rate": 3.970627050262057e-06, "loss": 0.4053, "step": 37875 }, { "epoch": 2.57, "learning_rate": 3.969684401040685e-06, "loss": 0.4015, "step": 37880 }, { "epoch": 2.57, "learning_rate": 3.968741751819313e-06, "loss": 0.3714, "step": 37885 }, { "epoch": 2.57, "learning_rate": 3.967799102597942e-06, "loss": 0.6128, "step": 37890 }, { "epoch": 2.57, "learning_rate": 3.96685645337657e-06, "loss": 0.3135, "step": 37895 }, { "epoch": 2.57, "learning_rate": 3.965913804155198e-06, "loss": 0.0984, "step": 37900 }, { "epoch": 2.57, "learning_rate": 3.964971154933826e-06, "loss": 0.4049, "step": 37905 }, { "epoch": 2.57, "learning_rate": 3.964028505712455e-06, "loss": 0.3696, "step": 37910 }, { "epoch": 2.57, "learning_rate": 3.963085856491083e-06, "loss": 0.3382, "step": 37915 }, { "epoch": 2.57, "learning_rate": 3.962143207269712e-06, "loss": 0.4867, "step": 37920 }, { "epoch": 2.57, "learning_rate": 3.961200558048339e-06, "loss": 0.3929, "step": 37925 }, { "epoch": 2.57, "learning_rate": 3.960257908826968e-06, "loss": 0.1167, "step": 37930 }, { "epoch": 2.57, "learning_rate": 3.959315259605596e-06, "loss": 0.0793, "step": 37935 }, { "epoch": 2.57, "learning_rate": 3.958372610384224e-06, "loss": 0.3935, "step": 37940 }, { "epoch": 2.58, "learning_rate": 3.957429961162852e-06, "loss": 0.3935, "step": 37945 }, { "epoch": 2.58, "learning_rate": 3.95648731194148e-06, "loss": 0.2654, "step": 37950 }, { "epoch": 2.58, "learning_rate": 3.9555446627201085e-06, "loss": 0.5924, "step": 37955 }, { "epoch": 2.58, "learning_rate": 3.954602013498737e-06, "loss": 0.1939, "step": 37960 }, { "epoch": 2.58, "learning_rate": 3.9536593642773655e-06, "loss": 0.179, "step": 37965 }, { "epoch": 2.58, "learning_rate": 3.952716715055994e-06, "loss": 0.47, "step": 37970 }, { "epoch": 2.58, "learning_rate": 3.9517740658346225e-06, "loss": 0.3099, "step": 37975 }, { "epoch": 2.58, "learning_rate": 3.95083141661325e-06, "loss": 0.629, "step": 37980 }, { "epoch": 2.58, "learning_rate": 3.949888767391879e-06, "loss": 0.2873, "step": 37985 }, { "epoch": 2.58, "learning_rate": 3.948946118170507e-06, "loss": 0.3307, "step": 37990 }, { "epoch": 2.58, "learning_rate": 3.948003468949135e-06, "loss": 0.4685, "step": 37995 }, { "epoch": 2.58, "learning_rate": 3.947060819727763e-06, "loss": 0.1947, "step": 38000 }, { "epoch": 2.58, "learning_rate": 3.946118170506392e-06, "loss": 0.2565, "step": 38005 }, { "epoch": 2.58, "learning_rate": 3.945175521285019e-06, "loss": 0.3158, "step": 38010 }, { "epoch": 2.58, "learning_rate": 3.944232872063648e-06, "loss": 0.1312, "step": 38015 }, { "epoch": 2.58, "learning_rate": 3.943290222842276e-06, "loss": 0.3568, "step": 38020 }, { "epoch": 2.58, "learning_rate": 3.942347573620904e-06, "loss": 0.4487, "step": 38025 }, { "epoch": 2.58, "learning_rate": 3.9414049243995325e-06, "loss": 0.1122, "step": 38030 }, { "epoch": 2.58, "learning_rate": 3.940462275178161e-06, "loss": 0.334, "step": 38035 }, { "epoch": 2.58, "learning_rate": 3.9395196259567894e-06, "loss": 0.1359, "step": 38040 }, { "epoch": 2.58, "learning_rate": 3.938576976735418e-06, "loss": 1.0632, "step": 38045 }, { "epoch": 2.58, "learning_rate": 3.937634327514046e-06, "loss": 0.111, "step": 38050 }, { "epoch": 2.58, "learning_rate": 3.936691678292674e-06, "loss": 0.0698, "step": 38055 }, { "epoch": 2.58, "learning_rate": 3.9357490290713025e-06, "loss": 0.388, "step": 38060 }, { "epoch": 2.58, "learning_rate": 3.93480637984993e-06, "loss": 0.1644, "step": 38065 }, { "epoch": 2.58, "learning_rate": 3.933863730628559e-06, "loss": 0.2146, "step": 38070 }, { "epoch": 2.58, "learning_rate": 3.932921081407187e-06, "loss": 0.5113, "step": 38075 }, { "epoch": 2.58, "learning_rate": 3.931978432185815e-06, "loss": 0.2565, "step": 38080 }, { "epoch": 2.58, "learning_rate": 3.931035782964443e-06, "loss": 0.3568, "step": 38085 }, { "epoch": 2.59, "learning_rate": 3.930093133743072e-06, "loss": 0.3593, "step": 38090 }, { "epoch": 2.59, "learning_rate": 3.9291504845217e-06, "loss": 0.1083, "step": 38095 }, { "epoch": 2.59, "learning_rate": 3.928207835300329e-06, "loss": 0.679, "step": 38100 }, { "epoch": 2.59, "learning_rate": 3.927265186078956e-06, "loss": 0.2466, "step": 38105 }, { "epoch": 2.59, "learning_rate": 3.926322536857585e-06, "loss": 0.3047, "step": 38110 }, { "epoch": 2.59, "learning_rate": 3.925379887636213e-06, "loss": 0.2895, "step": 38115 }, { "epoch": 2.59, "learning_rate": 3.924437238414842e-06, "loss": 0.5332, "step": 38120 }, { "epoch": 2.59, "learning_rate": 3.9234945891934695e-06, "loss": 0.5925, "step": 38125 }, { "epoch": 2.59, "learning_rate": 3.922551939972098e-06, "loss": 0.5401, "step": 38130 }, { "epoch": 2.59, "learning_rate": 3.9216092907507265e-06, "loss": 0.5859, "step": 38135 }, { "epoch": 2.59, "learning_rate": 3.920666641529354e-06, "loss": 0.1414, "step": 38140 }, { "epoch": 2.59, "learning_rate": 3.919723992307983e-06, "loss": 0.9053, "step": 38145 }, { "epoch": 2.59, "learning_rate": 3.918781343086611e-06, "loss": 0.1995, "step": 38150 }, { "epoch": 2.59, "learning_rate": 3.917838693865239e-06, "loss": 0.5987, "step": 38155 }, { "epoch": 2.59, "learning_rate": 3.916896044643867e-06, "loss": 0.2295, "step": 38160 }, { "epoch": 2.59, "learning_rate": 3.915953395422496e-06, "loss": 0.6632, "step": 38165 }, { "epoch": 2.59, "learning_rate": 3.915010746201124e-06, "loss": 0.458, "step": 38170 }, { "epoch": 2.59, "learning_rate": 3.914068096979753e-06, "loss": 0.1485, "step": 38175 }, { "epoch": 2.59, "learning_rate": 3.91312544775838e-06, "loss": 1.0603, "step": 38180 }, { "epoch": 2.59, "learning_rate": 3.912182798537009e-06, "loss": 0.1797, "step": 38185 }, { "epoch": 2.59, "learning_rate": 3.911240149315637e-06, "loss": 0.6011, "step": 38190 }, { "epoch": 2.59, "learning_rate": 3.910297500094265e-06, "loss": 0.489, "step": 38195 }, { "epoch": 2.59, "learning_rate": 3.909354850872893e-06, "loss": 0.1424, "step": 38200 }, { "epoch": 2.59, "learning_rate": 3.908412201651522e-06, "loss": 0.3336, "step": 38205 }, { "epoch": 2.59, "learning_rate": 3.9074695524301495e-06, "loss": 0.635, "step": 38210 }, { "epoch": 2.59, "learning_rate": 3.906526903208778e-06, "loss": 0.5081, "step": 38215 }, { "epoch": 2.59, "learning_rate": 3.9055842539874065e-06, "loss": 0.2005, "step": 38220 }, { "epoch": 2.59, "learning_rate": 3.904641604766035e-06, "loss": 0.2552, "step": 38225 }, { "epoch": 2.59, "learning_rate": 3.903698955544663e-06, "loss": 0.5043, "step": 38230 }, { "epoch": 2.6, "learning_rate": 3.902756306323291e-06, "loss": 0.2227, "step": 38235 }, { "epoch": 2.6, "learning_rate": 3.90181365710192e-06, "loss": 0.4577, "step": 38240 }, { "epoch": 2.6, "learning_rate": 3.900871007880548e-06, "loss": 0.1271, "step": 38245 }, { "epoch": 2.6, "learning_rate": 3.899928358659177e-06, "loss": 0.6662, "step": 38250 }, { "epoch": 2.6, "learning_rate": 3.898985709437804e-06, "loss": 0.321, "step": 38255 }, { "epoch": 2.6, "learning_rate": 3.898043060216433e-06, "loss": 0.2337, "step": 38260 }, { "epoch": 2.6, "learning_rate": 3.897100410995061e-06, "loss": 0.7, "step": 38265 }, { "epoch": 2.6, "learning_rate": 3.896157761773689e-06, "loss": 0.3409, "step": 38270 }, { "epoch": 2.6, "learning_rate": 3.895215112552317e-06, "loss": 0.2076, "step": 38275 }, { "epoch": 2.6, "learning_rate": 3.894272463330946e-06, "loss": 0.9796, "step": 38280 }, { "epoch": 2.6, "learning_rate": 3.8933298141095735e-06, "loss": 0.2743, "step": 38285 }, { "epoch": 2.6, "learning_rate": 3.892387164888202e-06, "loss": 0.3888, "step": 38290 }, { "epoch": 2.6, "learning_rate": 3.8914445156668304e-06, "loss": 0.1664, "step": 38295 }, { "epoch": 2.6, "learning_rate": 3.890501866445459e-06, "loss": 0.5601, "step": 38300 }, { "epoch": 2.6, "learning_rate": 3.889559217224087e-06, "loss": 0.1471, "step": 38305 }, { "epoch": 2.6, "learning_rate": 3.888616568002715e-06, "loss": 0.1658, "step": 38310 }, { "epoch": 2.6, "learning_rate": 3.8876739187813435e-06, "loss": 0.8208, "step": 38315 }, { "epoch": 2.6, "learning_rate": 3.886731269559972e-06, "loss": 0.4018, "step": 38320 }, { "epoch": 2.6, "learning_rate": 3.8857886203386e-06, "loss": 0.2261, "step": 38325 }, { "epoch": 2.6, "learning_rate": 3.884845971117228e-06, "loss": 0.1418, "step": 38330 }, { "epoch": 2.6, "learning_rate": 3.883903321895857e-06, "loss": 0.3771, "step": 38335 }, { "epoch": 2.6, "learning_rate": 3.882960672674484e-06, "loss": 0.31, "step": 38340 }, { "epoch": 2.6, "learning_rate": 3.882018023453113e-06, "loss": 0.3683, "step": 38345 }, { "epoch": 2.6, "learning_rate": 3.881075374231741e-06, "loss": 0.082, "step": 38350 }, { "epoch": 2.6, "learning_rate": 3.880132725010369e-06, "loss": 0.3566, "step": 38355 }, { "epoch": 2.6, "learning_rate": 3.879190075788997e-06, "loss": 0.2758, "step": 38360 }, { "epoch": 2.6, "learning_rate": 3.878247426567626e-06, "loss": 0.3097, "step": 38365 }, { "epoch": 2.6, "learning_rate": 3.877304777346254e-06, "loss": 0.1202, "step": 38370 }, { "epoch": 2.6, "learning_rate": 3.876362128124883e-06, "loss": 0.3872, "step": 38375 }, { "epoch": 2.6, "learning_rate": 3.875419478903511e-06, "loss": 0.7778, "step": 38380 }, { "epoch": 2.61, "learning_rate": 3.874476829682139e-06, "loss": 0.3927, "step": 38385 }, { "epoch": 2.61, "learning_rate": 3.8735341804607675e-06, "loss": 0.1229, "step": 38390 }, { "epoch": 2.61, "learning_rate": 3.872591531239395e-06, "loss": 0.129, "step": 38395 }, { "epoch": 2.61, "learning_rate": 3.871648882018024e-06, "loss": 0.2673, "step": 38400 }, { "epoch": 2.61, "learning_rate": 3.870706232796652e-06, "loss": 0.1333, "step": 38405 }, { "epoch": 2.61, "learning_rate": 3.86976358357528e-06, "loss": 0.3887, "step": 38410 }, { "epoch": 2.61, "learning_rate": 3.868820934353908e-06, "loss": 0.4749, "step": 38415 }, { "epoch": 2.61, "learning_rate": 3.867878285132537e-06, "loss": 0.3595, "step": 38420 }, { "epoch": 2.61, "learning_rate": 3.866935635911165e-06, "loss": 0.5449, "step": 38425 }, { "epoch": 2.61, "learning_rate": 3.865992986689794e-06, "loss": 0.5765, "step": 38430 }, { "epoch": 2.61, "learning_rate": 3.865050337468422e-06, "loss": 0.2452, "step": 38435 }, { "epoch": 2.61, "learning_rate": 3.86410768824705e-06, "loss": 0.5018, "step": 38440 }, { "epoch": 2.61, "learning_rate": 3.863165039025678e-06, "loss": 0.2305, "step": 38445 }, { "epoch": 2.61, "learning_rate": 3.862222389804307e-06, "loss": 0.3255, "step": 38450 }, { "epoch": 2.61, "learning_rate": 3.861279740582934e-06, "loss": 0.4775, "step": 38455 }, { "epoch": 2.61, "learning_rate": 3.860337091361563e-06, "loss": 0.2898, "step": 38460 }, { "epoch": 2.61, "learning_rate": 3.859394442140191e-06, "loss": 0.2279, "step": 38465 }, { "epoch": 2.61, "learning_rate": 3.858451792918819e-06, "loss": 0.4755, "step": 38470 }, { "epoch": 2.61, "learning_rate": 3.8575091436974475e-06, "loss": 0.3817, "step": 38475 }, { "epoch": 2.61, "learning_rate": 3.856566494476076e-06, "loss": 0.6423, "step": 38480 }, { "epoch": 2.61, "learning_rate": 3.855623845254704e-06, "loss": 0.3875, "step": 38485 }, { "epoch": 2.61, "learning_rate": 3.854681196033332e-06, "loss": 0.3158, "step": 38490 }, { "epoch": 2.61, "learning_rate": 3.853738546811961e-06, "loss": 0.3526, "step": 38495 }, { "epoch": 2.61, "learning_rate": 3.852795897590589e-06, "loss": 0.3443, "step": 38500 }, { "epoch": 2.61, "learning_rate": 3.851853248369218e-06, "loss": 0.0975, "step": 38505 }, { "epoch": 2.61, "learning_rate": 3.850910599147845e-06, "loss": 0.191, "step": 38510 }, { "epoch": 2.61, "learning_rate": 3.849967949926474e-06, "loss": 0.7189, "step": 38515 }, { "epoch": 2.61, "learning_rate": 3.849025300705102e-06, "loss": 0.076, "step": 38520 }, { "epoch": 2.61, "learning_rate": 3.84808265148373e-06, "loss": 0.2375, "step": 38525 }, { "epoch": 2.62, "learning_rate": 3.847140002262358e-06, "loss": 0.6041, "step": 38530 }, { "epoch": 2.62, "learning_rate": 3.846197353040987e-06, "loss": 0.3157, "step": 38535 }, { "epoch": 2.62, "learning_rate": 3.8452547038196144e-06, "loss": 0.3003, "step": 38540 }, { "epoch": 2.62, "learning_rate": 3.844312054598243e-06, "loss": 0.2188, "step": 38545 }, { "epoch": 2.62, "learning_rate": 3.8433694053768714e-06, "loss": 0.3353, "step": 38550 }, { "epoch": 2.62, "learning_rate": 3.8424267561555e-06, "loss": 0.1497, "step": 38555 }, { "epoch": 2.62, "learning_rate": 3.841484106934128e-06, "loss": 0.3728, "step": 38560 }, { "epoch": 2.62, "learning_rate": 3.840541457712756e-06, "loss": 0.2169, "step": 38565 }, { "epoch": 2.62, "learning_rate": 3.8395988084913845e-06, "loss": 0.501, "step": 38570 }, { "epoch": 2.62, "learning_rate": 3.838656159270013e-06, "loss": 0.1992, "step": 38575 }, { "epoch": 2.62, "learning_rate": 3.8377135100486415e-06, "loss": 0.2826, "step": 38580 }, { "epoch": 2.62, "learning_rate": 3.836770860827269e-06, "loss": 0.2341, "step": 38585 }, { "epoch": 2.62, "learning_rate": 3.835828211605898e-06, "loss": 0.3042, "step": 38590 }, { "epoch": 2.62, "learning_rate": 3.834885562384526e-06, "loss": 0.3996, "step": 38595 }, { "epoch": 2.62, "learning_rate": 3.833942913163154e-06, "loss": 0.4064, "step": 38600 }, { "epoch": 2.62, "learning_rate": 3.833000263941782e-06, "loss": 0.0912, "step": 38605 }, { "epoch": 2.62, "learning_rate": 3.832057614720411e-06, "loss": 0.5893, "step": 38610 }, { "epoch": 2.62, "learning_rate": 3.831114965499038e-06, "loss": 0.3709, "step": 38615 }, { "epoch": 2.62, "learning_rate": 3.830172316277667e-06, "loss": 0.2571, "step": 38620 }, { "epoch": 2.62, "learning_rate": 3.829229667056295e-06, "loss": 0.1271, "step": 38625 }, { "epoch": 2.62, "learning_rate": 3.828287017834924e-06, "loss": 0.4047, "step": 38630 }, { "epoch": 2.62, "learning_rate": 3.827344368613552e-06, "loss": 0.3315, "step": 38635 }, { "epoch": 2.62, "learning_rate": 3.82640171939218e-06, "loss": 0.307, "step": 38640 }, { "epoch": 2.62, "learning_rate": 3.8254590701708084e-06, "loss": 0.3291, "step": 38645 }, { "epoch": 2.62, "learning_rate": 3.824516420949437e-06, "loss": 0.3291, "step": 38650 }, { "epoch": 2.62, "learning_rate": 3.8235737717280646e-06, "loss": 0.4478, "step": 38655 }, { "epoch": 2.62, "learning_rate": 3.822631122506693e-06, "loss": 0.0562, "step": 38660 }, { "epoch": 2.62, "learning_rate": 3.8216884732853216e-06, "loss": 0.2876, "step": 38665 }, { "epoch": 2.62, "learning_rate": 3.820745824063949e-06, "loss": 0.1496, "step": 38670 }, { "epoch": 2.62, "learning_rate": 3.819803174842578e-06, "loss": 0.2004, "step": 38675 }, { "epoch": 2.63, "learning_rate": 3.818860525621206e-06, "loss": 0.3532, "step": 38680 }, { "epoch": 2.63, "learning_rate": 3.817917876399835e-06, "loss": 0.4744, "step": 38685 }, { "epoch": 2.63, "learning_rate": 3.816975227178462e-06, "loss": 0.3085, "step": 38690 }, { "epoch": 2.63, "learning_rate": 3.816032577957091e-06, "loss": 0.4628, "step": 38695 }, { "epoch": 2.63, "learning_rate": 3.815089928735719e-06, "loss": 0.075, "step": 38700 }, { "epoch": 2.63, "learning_rate": 3.8141472795143473e-06, "loss": 0.2202, "step": 38705 }, { "epoch": 2.63, "learning_rate": 3.813204630292976e-06, "loss": 0.4951, "step": 38710 }, { "epoch": 2.63, "learning_rate": 3.812261981071604e-06, "loss": 0.6812, "step": 38715 }, { "epoch": 2.63, "learning_rate": 3.8113193318502324e-06, "loss": 0.2442, "step": 38720 }, { "epoch": 2.63, "learning_rate": 3.810376682628861e-06, "loss": 0.3302, "step": 38725 }, { "epoch": 2.63, "learning_rate": 3.8094340334074885e-06, "loss": 0.3403, "step": 38730 }, { "epoch": 2.63, "learning_rate": 3.808491384186117e-06, "loss": 0.5736, "step": 38735 }, { "epoch": 2.63, "learning_rate": 3.807548734964745e-06, "loss": 0.6401, "step": 38740 }, { "epoch": 2.63, "learning_rate": 3.8066060857433735e-06, "loss": 0.2712, "step": 38745 }, { "epoch": 2.63, "learning_rate": 3.805663436522002e-06, "loss": 0.3503, "step": 38750 }, { "epoch": 2.63, "learning_rate": 3.8047207873006297e-06, "loss": 0.22, "step": 38755 }, { "epoch": 2.63, "learning_rate": 3.803778138079258e-06, "loss": 0.3722, "step": 38760 }, { "epoch": 2.63, "learning_rate": 3.8028354888578866e-06, "loss": 0.3778, "step": 38765 }, { "epoch": 2.63, "learning_rate": 3.8018928396365147e-06, "loss": 0.8937, "step": 38770 }, { "epoch": 2.63, "learning_rate": 3.800950190415143e-06, "loss": 0.0965, "step": 38775 }, { "epoch": 2.63, "learning_rate": 3.8000075411937717e-06, "loss": 0.0426, "step": 38780 }, { "epoch": 2.63, "learning_rate": 3.7990648919723993e-06, "loss": 0.3444, "step": 38785 }, { "epoch": 2.63, "learning_rate": 3.798122242751028e-06, "loss": 0.6724, "step": 38790 }, { "epoch": 2.63, "learning_rate": 3.7971795935296563e-06, "loss": 0.079, "step": 38795 }, { "epoch": 2.63, "learning_rate": 3.7962369443082844e-06, "loss": 0.2922, "step": 38800 }, { "epoch": 2.63, "learning_rate": 3.7952942950869124e-06, "loss": 0.9332, "step": 38805 }, { "epoch": 2.63, "learning_rate": 3.794351645865541e-06, "loss": 0.4172, "step": 38810 }, { "epoch": 2.63, "learning_rate": 3.793408996644169e-06, "loss": 0.2069, "step": 38815 }, { "epoch": 2.63, "learning_rate": 3.7924663474227975e-06, "loss": 0.3416, "step": 38820 }, { "epoch": 2.64, "learning_rate": 3.791523698201426e-06, "loss": 0.466, "step": 38825 }, { "epoch": 2.64, "learning_rate": 3.7905810489800536e-06, "loss": 0.2705, "step": 38830 }, { "epoch": 2.64, "learning_rate": 3.789638399758682e-06, "loss": 0.0661, "step": 38835 }, { "epoch": 2.64, "learning_rate": 3.78869575053731e-06, "loss": 0.6849, "step": 38840 }, { "epoch": 2.64, "learning_rate": 3.7877531013159386e-06, "loss": 0.1608, "step": 38845 }, { "epoch": 2.64, "learning_rate": 3.786810452094567e-06, "loss": 0.3304, "step": 38850 }, { "epoch": 2.64, "learning_rate": 3.7858678028731948e-06, "loss": 0.5846, "step": 38855 }, { "epoch": 2.64, "learning_rate": 3.7849251536518232e-06, "loss": 0.3784, "step": 38860 }, { "epoch": 2.64, "learning_rate": 3.7839825044304517e-06, "loss": 0.2973, "step": 38865 }, { "epoch": 2.64, "learning_rate": 3.78303985520908e-06, "loss": 0.447, "step": 38870 }, { "epoch": 2.64, "learning_rate": 3.7820972059877083e-06, "loss": 0.3828, "step": 38875 }, { "epoch": 2.64, "learning_rate": 3.7811545567663368e-06, "loss": 0.3259, "step": 38880 }, { "epoch": 2.64, "learning_rate": 3.7802119075449644e-06, "loss": 0.6883, "step": 38885 }, { "epoch": 2.64, "learning_rate": 3.779269258323593e-06, "loss": 0.6823, "step": 38890 }, { "epoch": 2.64, "learning_rate": 3.7783266091022214e-06, "loss": 0.5203, "step": 38895 }, { "epoch": 2.64, "learning_rate": 3.7773839598808494e-06, "loss": 0.5639, "step": 38900 }, { "epoch": 2.64, "learning_rate": 3.776441310659478e-06, "loss": 0.0766, "step": 38905 }, { "epoch": 2.64, "learning_rate": 3.775498661438106e-06, "loss": 0.3707, "step": 38910 }, { "epoch": 2.64, "learning_rate": 3.774556012216734e-06, "loss": 0.2233, "step": 38915 }, { "epoch": 2.64, "learning_rate": 3.7736133629953625e-06, "loss": 0.4756, "step": 38920 }, { "epoch": 2.64, "learning_rate": 3.772670713773991e-06, "loss": 0.1755, "step": 38925 }, { "epoch": 2.64, "learning_rate": 3.7717280645526187e-06, "loss": 0.2234, "step": 38930 }, { "epoch": 2.64, "learning_rate": 3.770785415331247e-06, "loss": 0.071, "step": 38935 }, { "epoch": 2.64, "learning_rate": 3.7698427661098756e-06, "loss": 0.4252, "step": 38940 }, { "epoch": 2.64, "learning_rate": 3.7689001168885037e-06, "loss": 0.2884, "step": 38945 }, { "epoch": 2.64, "learning_rate": 3.767957467667132e-06, "loss": 0.1121, "step": 38950 }, { "epoch": 2.64, "learning_rate": 3.76701481844576e-06, "loss": 0.1711, "step": 38955 }, { "epoch": 2.64, "learning_rate": 3.7660721692243883e-06, "loss": 0.1215, "step": 38960 }, { "epoch": 2.64, "learning_rate": 3.765129520003017e-06, "loss": 0.8436, "step": 38965 }, { "epoch": 2.64, "learning_rate": 3.764186870781645e-06, "loss": 0.4827, "step": 38970 }, { "epoch": 2.65, "learning_rate": 3.7632442215602734e-06, "loss": 0.4726, "step": 38975 }, { "epoch": 2.65, "learning_rate": 3.762301572338902e-06, "loss": 0.3275, "step": 38980 }, { "epoch": 2.65, "learning_rate": 3.7613589231175295e-06, "loss": 0.3801, "step": 38985 }, { "epoch": 2.65, "learning_rate": 3.760416273896158e-06, "loss": 0.2046, "step": 38990 }, { "epoch": 2.65, "learning_rate": 3.7594736246747865e-06, "loss": 0.5934, "step": 38995 }, { "epoch": 2.65, "learning_rate": 3.7585309754534145e-06, "loss": 0.9644, "step": 39000 }, { "epoch": 2.65, "learning_rate": 3.757588326232043e-06, "loss": 0.1913, "step": 39005 }, { "epoch": 2.65, "learning_rate": 3.756645677010671e-06, "loss": 0.4509, "step": 39010 }, { "epoch": 2.65, "learning_rate": 3.755703027789299e-06, "loss": 0.3466, "step": 39015 }, { "epoch": 2.65, "learning_rate": 3.7547603785679276e-06, "loss": 0.6505, "step": 39020 }, { "epoch": 2.65, "learning_rate": 3.753817729346556e-06, "loss": 0.5754, "step": 39025 }, { "epoch": 2.65, "learning_rate": 3.752875080125184e-06, "loss": 0.1237, "step": 39030 }, { "epoch": 2.65, "learning_rate": 3.7519324309038122e-06, "loss": 0.0693, "step": 39035 }, { "epoch": 2.65, "learning_rate": 3.7509897816824407e-06, "loss": 0.6386, "step": 39040 }, { "epoch": 2.65, "learning_rate": 3.750047132461069e-06, "loss": 0.4199, "step": 39045 }, { "epoch": 2.65, "learning_rate": 3.7491044832396973e-06, "loss": 0.3814, "step": 39050 }, { "epoch": 2.65, "learning_rate": 3.7481618340183258e-06, "loss": 0.4156, "step": 39055 }, { "epoch": 2.65, "learning_rate": 3.7472191847969534e-06, "loss": 0.0433, "step": 39060 }, { "epoch": 2.65, "learning_rate": 3.746276535575582e-06, "loss": 0.1676, "step": 39065 }, { "epoch": 2.65, "learning_rate": 3.74533388635421e-06, "loss": 0.1516, "step": 39070 }, { "epoch": 2.65, "learning_rate": 3.7443912371328385e-06, "loss": 0.2262, "step": 39075 }, { "epoch": 2.65, "learning_rate": 3.743448587911467e-06, "loss": 0.4844, "step": 39080 }, { "epoch": 2.65, "learning_rate": 3.7425059386900946e-06, "loss": 0.4646, "step": 39085 }, { "epoch": 2.65, "learning_rate": 3.741563289468723e-06, "loss": 0.3696, "step": 39090 }, { "epoch": 2.65, "learning_rate": 3.7406206402473516e-06, "loss": 0.1338, "step": 39095 }, { "epoch": 2.65, "learning_rate": 3.7396779910259796e-06, "loss": 0.4183, "step": 39100 }, { "epoch": 2.65, "learning_rate": 3.738735341804608e-06, "loss": 0.5794, "step": 39105 }, { "epoch": 2.65, "learning_rate": 3.7377926925832366e-06, "loss": 0.4316, "step": 39110 }, { "epoch": 2.65, "learning_rate": 3.7368500433618642e-06, "loss": 0.1309, "step": 39115 }, { "epoch": 2.66, "learning_rate": 3.7359073941404927e-06, "loss": 0.1636, "step": 39120 }, { "epoch": 2.66, "learning_rate": 3.734964744919121e-06, "loss": 0.0987, "step": 39125 }, { "epoch": 2.66, "learning_rate": 3.7340220956977493e-06, "loss": 0.337, "step": 39130 }, { "epoch": 2.66, "learning_rate": 3.7330794464763773e-06, "loss": 0.3431, "step": 39135 }, { "epoch": 2.66, "learning_rate": 3.732136797255006e-06, "loss": 0.168, "step": 39140 }, { "epoch": 2.66, "learning_rate": 3.731194148033634e-06, "loss": 0.3468, "step": 39145 }, { "epoch": 2.66, "learning_rate": 3.7302514988122624e-06, "loss": 0.0838, "step": 39150 }, { "epoch": 2.66, "learning_rate": 3.729308849590891e-06, "loss": 0.4492, "step": 39155 }, { "epoch": 2.66, "learning_rate": 3.7283662003695185e-06, "loss": 0.1359, "step": 39160 }, { "epoch": 2.66, "learning_rate": 3.727423551148147e-06, "loss": 0.4504, "step": 39165 }, { "epoch": 2.66, "learning_rate": 3.7264809019267755e-06, "loss": 0.3049, "step": 39170 }, { "epoch": 2.66, "learning_rate": 3.7255382527054035e-06, "loss": 0.1377, "step": 39175 }, { "epoch": 2.66, "learning_rate": 3.724595603484032e-06, "loss": 0.4313, "step": 39180 }, { "epoch": 2.66, "learning_rate": 3.7236529542626597e-06, "loss": 0.5268, "step": 39185 }, { "epoch": 2.66, "learning_rate": 3.722710305041288e-06, "loss": 0.3745, "step": 39190 }, { "epoch": 2.66, "learning_rate": 3.7217676558199166e-06, "loss": 0.4887, "step": 39195 }, { "epoch": 2.66, "learning_rate": 3.7208250065985447e-06, "loss": 0.1636, "step": 39200 }, { "epoch": 2.66, "learning_rate": 3.719882357377173e-06, "loss": 0.1482, "step": 39205 }, { "epoch": 2.66, "learning_rate": 3.7189397081558017e-06, "loss": 0.4952, "step": 39210 }, { "epoch": 2.66, "learning_rate": 3.7179970589344293e-06, "loss": 0.5742, "step": 39215 }, { "epoch": 2.66, "learning_rate": 3.717054409713058e-06, "loss": 0.0683, "step": 39220 }, { "epoch": 2.66, "learning_rate": 3.7161117604916863e-06, "loss": 0.1632, "step": 39225 }, { "epoch": 2.66, "learning_rate": 3.7151691112703144e-06, "loss": 0.4825, "step": 39230 }, { "epoch": 2.66, "learning_rate": 3.714226462048943e-06, "loss": 0.4015, "step": 39235 }, { "epoch": 2.66, "learning_rate": 3.713283812827571e-06, "loss": 0.8929, "step": 39240 }, { "epoch": 2.66, "learning_rate": 3.712341163606199e-06, "loss": 0.4955, "step": 39245 }, { "epoch": 2.66, "learning_rate": 3.7113985143848275e-06, "loss": 0.7629, "step": 39250 }, { "epoch": 2.66, "learning_rate": 3.710455865163456e-06, "loss": 0.2902, "step": 39255 }, { "epoch": 2.66, "learning_rate": 3.709513215942084e-06, "loss": 0.2953, "step": 39260 }, { "epoch": 2.66, "learning_rate": 3.708570566720712e-06, "loss": 0.5119, "step": 39265 }, { "epoch": 2.67, "learning_rate": 3.7076279174993406e-06, "loss": 0.535, "step": 39270 }, { "epoch": 2.67, "learning_rate": 3.7066852682779686e-06, "loss": 0.1437, "step": 39275 }, { "epoch": 2.67, "learning_rate": 3.705742619056597e-06, "loss": 0.2139, "step": 39280 }, { "epoch": 2.67, "learning_rate": 3.7047999698352256e-06, "loss": 0.3232, "step": 39285 }, { "epoch": 2.67, "learning_rate": 3.7038573206138532e-06, "loss": 0.3717, "step": 39290 }, { "epoch": 2.67, "learning_rate": 3.7029146713924817e-06, "loss": 0.2472, "step": 39295 }, { "epoch": 2.67, "learning_rate": 3.70197202217111e-06, "loss": 0.5135, "step": 39300 }, { "epoch": 2.67, "learning_rate": 3.7010293729497383e-06, "loss": 0.3236, "step": 39305 }, { "epoch": 2.67, "learning_rate": 3.7000867237283668e-06, "loss": 0.5634, "step": 39310 }, { "epoch": 2.67, "learning_rate": 3.6991440745069944e-06, "loss": 0.1989, "step": 39315 }, { "epoch": 2.67, "learning_rate": 3.698201425285623e-06, "loss": 0.2472, "step": 39320 }, { "epoch": 2.67, "learning_rate": 3.6972587760642514e-06, "loss": 0.6778, "step": 39325 }, { "epoch": 2.67, "learning_rate": 3.6963161268428794e-06, "loss": 0.3838, "step": 39330 }, { "epoch": 2.67, "learning_rate": 3.695373477621508e-06, "loss": 0.3032, "step": 39335 }, { "epoch": 2.67, "learning_rate": 3.6944308284001364e-06, "loss": 0.4679, "step": 39340 }, { "epoch": 2.67, "learning_rate": 3.693488179178764e-06, "loss": 0.1062, "step": 39345 }, { "epoch": 2.67, "learning_rate": 3.6925455299573925e-06, "loss": 0.5748, "step": 39350 }, { "epoch": 2.67, "learning_rate": 3.691602880736021e-06, "loss": 0.2267, "step": 39355 }, { "epoch": 2.67, "learning_rate": 3.690660231514649e-06, "loss": 0.1287, "step": 39360 }, { "epoch": 2.67, "learning_rate": 3.689717582293277e-06, "loss": 0.2146, "step": 39365 }, { "epoch": 2.67, "learning_rate": 3.6887749330719056e-06, "loss": 0.2469, "step": 39370 }, { "epoch": 2.67, "learning_rate": 3.6878322838505337e-06, "loss": 0.2323, "step": 39375 }, { "epoch": 2.67, "learning_rate": 3.686889634629162e-06, "loss": 0.2554, "step": 39380 }, { "epoch": 2.67, "learning_rate": 3.6859469854077907e-06, "loss": 0.2761, "step": 39385 }, { "epoch": 2.67, "learning_rate": 3.6850043361864183e-06, "loss": 0.1858, "step": 39390 }, { "epoch": 2.67, "learning_rate": 3.684061686965047e-06, "loss": 0.1264, "step": 39395 }, { "epoch": 2.67, "learning_rate": 3.683119037743675e-06, "loss": 0.7231, "step": 39400 }, { "epoch": 2.67, "learning_rate": 3.6821763885223034e-06, "loss": 0.5147, "step": 39405 }, { "epoch": 2.67, "learning_rate": 3.681233739300932e-06, "loss": 0.0468, "step": 39410 }, { "epoch": 2.68, "learning_rate": 3.6802910900795595e-06, "loss": 0.2048, "step": 39415 }, { "epoch": 2.68, "learning_rate": 3.679348440858188e-06, "loss": 0.2841, "step": 39420 }, { "epoch": 2.68, "learning_rate": 3.6784057916368165e-06, "loss": 0.5238, "step": 39425 }, { "epoch": 2.68, "learning_rate": 3.6774631424154445e-06, "loss": 0.3583, "step": 39430 }, { "epoch": 2.68, "learning_rate": 3.676520493194073e-06, "loss": 0.4009, "step": 39435 }, { "epoch": 2.68, "learning_rate": 3.6755778439727015e-06, "loss": 0.4446, "step": 39440 }, { "epoch": 2.68, "learning_rate": 3.674635194751329e-06, "loss": 0.6018, "step": 39445 }, { "epoch": 2.68, "learning_rate": 3.6736925455299576e-06, "loss": 0.6574, "step": 39450 }, { "epoch": 2.68, "learning_rate": 3.672749896308586e-06, "loss": 0.3513, "step": 39455 }, { "epoch": 2.68, "learning_rate": 3.671807247087214e-06, "loss": 0.6092, "step": 39460 }, { "epoch": 2.68, "learning_rate": 3.6708645978658427e-06, "loss": 0.2729, "step": 39465 }, { "epoch": 2.68, "learning_rate": 3.6699219486444707e-06, "loss": 0.5145, "step": 39470 }, { "epoch": 2.68, "learning_rate": 3.668979299423099e-06, "loss": 0.3908, "step": 39475 }, { "epoch": 2.68, "learning_rate": 3.6680366502017273e-06, "loss": 0.1408, "step": 39480 }, { "epoch": 2.68, "learning_rate": 3.6670940009803558e-06, "loss": 0.2104, "step": 39485 }, { "epoch": 2.68, "learning_rate": 3.6661513517589834e-06, "loss": 0.1726, "step": 39490 }, { "epoch": 2.68, "learning_rate": 3.665208702537612e-06, "loss": 0.7019, "step": 39495 }, { "epoch": 2.68, "learning_rate": 3.6642660533162404e-06, "loss": 0.6676, "step": 39500 }, { "epoch": 2.68, "learning_rate": 3.6633234040948685e-06, "loss": 0.4226, "step": 39505 }, { "epoch": 2.68, "learning_rate": 3.662380754873497e-06, "loss": 0.1741, "step": 39510 }, { "epoch": 2.68, "learning_rate": 3.6614381056521246e-06, "loss": 0.4933, "step": 39515 }, { "epoch": 2.68, "learning_rate": 3.660495456430753e-06, "loss": 0.5985, "step": 39520 }, { "epoch": 2.68, "learning_rate": 3.6595528072093816e-06, "loss": 0.1622, "step": 39525 }, { "epoch": 2.68, "learning_rate": 3.6586101579880096e-06, "loss": 0.4112, "step": 39530 }, { "epoch": 2.68, "learning_rate": 3.657667508766638e-06, "loss": 0.1904, "step": 39535 }, { "epoch": 2.68, "learning_rate": 3.6567248595452666e-06, "loss": 0.2285, "step": 39540 }, { "epoch": 2.68, "learning_rate": 3.6557822103238942e-06, "loss": 0.2385, "step": 39545 }, { "epoch": 2.68, "learning_rate": 3.6548395611025227e-06, "loss": 0.2613, "step": 39550 }, { "epoch": 2.68, "learning_rate": 3.653896911881151e-06, "loss": 0.1671, "step": 39555 }, { "epoch": 2.68, "learning_rate": 3.6529542626597793e-06, "loss": 0.242, "step": 39560 }, { "epoch": 2.69, "learning_rate": 3.6520116134384078e-06, "loss": 0.2203, "step": 39565 }, { "epoch": 2.69, "learning_rate": 3.6510689642170362e-06, "loss": 0.1904, "step": 39570 }, { "epoch": 2.69, "learning_rate": 3.650126314995664e-06, "loss": 0.2052, "step": 39575 }, { "epoch": 2.69, "learning_rate": 3.6491836657742924e-06, "loss": 0.2695, "step": 39580 }, { "epoch": 2.69, "learning_rate": 3.648241016552921e-06, "loss": 0.0934, "step": 39585 }, { "epoch": 2.69, "learning_rate": 3.647298367331549e-06, "loss": 0.3988, "step": 39590 }, { "epoch": 2.69, "learning_rate": 3.646355718110177e-06, "loss": 0.3069, "step": 39595 }, { "epoch": 2.69, "learning_rate": 3.6454130688888055e-06, "loss": 0.147, "step": 39600 }, { "epoch": 2.69, "learning_rate": 3.6444704196674335e-06, "loss": 0.0892, "step": 39605 }, { "epoch": 2.69, "learning_rate": 3.643527770446062e-06, "loss": 0.1411, "step": 39610 }, { "epoch": 2.69, "learning_rate": 3.6425851212246905e-06, "loss": 0.3319, "step": 39615 }, { "epoch": 2.69, "learning_rate": 3.641642472003318e-06, "loss": 0.1329, "step": 39620 }, { "epoch": 2.69, "learning_rate": 3.6406998227819466e-06, "loss": 0.2815, "step": 39625 }, { "epoch": 2.69, "learning_rate": 3.6397571735605747e-06, "loss": 0.5862, "step": 39630 }, { "epoch": 2.69, "learning_rate": 3.638814524339203e-06, "loss": 0.3127, "step": 39635 }, { "epoch": 2.69, "learning_rate": 3.6378718751178317e-06, "loss": 0.3291, "step": 39640 }, { "epoch": 2.69, "learning_rate": 3.6369292258964593e-06, "loss": 0.3963, "step": 39645 }, { "epoch": 2.69, "learning_rate": 3.635986576675088e-06, "loss": 0.1917, "step": 39650 }, { "epoch": 2.69, "learning_rate": 3.6350439274537163e-06, "loss": 0.1533, "step": 39655 }, { "epoch": 2.69, "learning_rate": 3.6341012782323444e-06, "loss": 0.1588, "step": 39660 }, { "epoch": 2.69, "learning_rate": 3.633158629010973e-06, "loss": 0.3292, "step": 39665 }, { "epoch": 2.69, "learning_rate": 3.6322159797896013e-06, "loss": 0.0977, "step": 39670 }, { "epoch": 2.69, "learning_rate": 3.631273330568229e-06, "loss": 0.2699, "step": 39675 }, { "epoch": 2.69, "learning_rate": 3.6303306813468575e-06, "loss": 0.4795, "step": 39680 }, { "epoch": 2.69, "learning_rate": 3.629388032125486e-06, "loss": 0.4888, "step": 39685 }, { "epoch": 2.69, "learning_rate": 3.628445382904114e-06, "loss": 0.4286, "step": 39690 }, { "epoch": 2.69, "learning_rate": 3.6275027336827425e-06, "loss": 0.1636, "step": 39695 }, { "epoch": 2.69, "learning_rate": 3.6265600844613706e-06, "loss": 0.1478, "step": 39700 }, { "epoch": 2.69, "learning_rate": 3.6256174352399986e-06, "loss": 1.0474, "step": 39705 }, { "epoch": 2.7, "learning_rate": 3.624674786018627e-06, "loss": 0.4846, "step": 39710 }, { "epoch": 2.7, "learning_rate": 3.6237321367972556e-06, "loss": 0.2854, "step": 39715 }, { "epoch": 2.7, "learning_rate": 3.6227894875758832e-06, "loss": 0.4814, "step": 39720 }, { "epoch": 2.7, "learning_rate": 3.6218468383545117e-06, "loss": 0.4656, "step": 39725 }, { "epoch": 2.7, "learning_rate": 3.6209041891331402e-06, "loss": 0.1492, "step": 39730 }, { "epoch": 2.7, "learning_rate": 3.6199615399117683e-06, "loss": 0.4455, "step": 39735 }, { "epoch": 2.7, "learning_rate": 3.6190188906903968e-06, "loss": 0.2803, "step": 39740 }, { "epoch": 2.7, "learning_rate": 3.6180762414690244e-06, "loss": 0.6364, "step": 39745 }, { "epoch": 2.7, "learning_rate": 3.617133592247653e-06, "loss": 0.5687, "step": 39750 }, { "epoch": 2.7, "learning_rate": 3.6161909430262814e-06, "loss": 0.7864, "step": 39755 }, { "epoch": 2.7, "learning_rate": 3.6152482938049094e-06, "loss": 0.3584, "step": 39760 }, { "epoch": 2.7, "learning_rate": 3.614305644583538e-06, "loss": 0.3253, "step": 39765 }, { "epoch": 2.7, "learning_rate": 3.6133629953621664e-06, "loss": 0.2373, "step": 39770 }, { "epoch": 2.7, "learning_rate": 3.612420346140794e-06, "loss": 0.1383, "step": 39775 }, { "epoch": 2.7, "learning_rate": 3.6114776969194226e-06, "loss": 0.221, "step": 39780 }, { "epoch": 2.7, "learning_rate": 3.610535047698051e-06, "loss": 0.3709, "step": 39785 }, { "epoch": 2.7, "learning_rate": 3.609592398476679e-06, "loss": 0.4476, "step": 39790 }, { "epoch": 2.7, "learning_rate": 3.6086497492553076e-06, "loss": 0.2507, "step": 39795 }, { "epoch": 2.7, "learning_rate": 3.607707100033936e-06, "loss": 0.3343, "step": 39800 }, { "epoch": 2.7, "learning_rate": 3.6067644508125637e-06, "loss": 0.1776, "step": 39805 }, { "epoch": 2.7, "learning_rate": 3.605821801591192e-06, "loss": 0.3347, "step": 39810 }, { "epoch": 2.7, "learning_rate": 3.6048791523698207e-06, "loss": 0.4297, "step": 39815 }, { "epoch": 2.7, "learning_rate": 3.6039365031484488e-06, "loss": 0.2707, "step": 39820 }, { "epoch": 2.7, "learning_rate": 3.602993853927077e-06, "loss": 0.2897, "step": 39825 }, { "epoch": 2.7, "learning_rate": 3.6020512047057053e-06, "loss": 0.1978, "step": 39830 }, { "epoch": 2.7, "learning_rate": 3.6011085554843334e-06, "loss": 0.1737, "step": 39835 }, { "epoch": 2.7, "learning_rate": 3.600165906262962e-06, "loss": 0.251, "step": 39840 }, { "epoch": 2.7, "learning_rate": 3.5992232570415895e-06, "loss": 0.423, "step": 39845 }, { "epoch": 2.7, "learning_rate": 3.598280607820218e-06, "loss": 0.5044, "step": 39850 }, { "epoch": 2.7, "learning_rate": 3.5973379585988465e-06, "loss": 0.1362, "step": 39855 }, { "epoch": 2.71, "learning_rate": 3.5963953093774745e-06, "loss": 0.305, "step": 39860 }, { "epoch": 2.71, "learning_rate": 3.595452660156103e-06, "loss": 0.3933, "step": 39865 }, { "epoch": 2.71, "learning_rate": 3.5945100109347315e-06, "loss": 0.6603, "step": 39870 }, { "epoch": 2.71, "learning_rate": 3.593567361713359e-06, "loss": 0.3075, "step": 39875 }, { "epoch": 2.71, "learning_rate": 3.5926247124919876e-06, "loss": 0.2321, "step": 39880 }, { "epoch": 2.71, "learning_rate": 3.591682063270616e-06, "loss": 0.126, "step": 39885 }, { "epoch": 2.71, "learning_rate": 3.590739414049244e-06, "loss": 0.2416, "step": 39890 }, { "epoch": 2.71, "learning_rate": 3.5897967648278727e-06, "loss": 0.2514, "step": 39895 }, { "epoch": 2.71, "learning_rate": 3.588854115606501e-06, "loss": 0.4095, "step": 39900 }, { "epoch": 2.71, "learning_rate": 3.587911466385129e-06, "loss": 0.1907, "step": 39905 }, { "epoch": 2.71, "learning_rate": 3.5869688171637573e-06, "loss": 0.6177, "step": 39910 }, { "epoch": 2.71, "learning_rate": 3.5860261679423858e-06, "loss": 0.3593, "step": 39915 }, { "epoch": 2.71, "learning_rate": 3.585083518721014e-06, "loss": 0.293, "step": 39920 }, { "epoch": 2.71, "learning_rate": 3.5841408694996423e-06, "loss": 0.1358, "step": 39925 }, { "epoch": 2.71, "learning_rate": 3.5831982202782704e-06, "loss": 0.1514, "step": 39930 }, { "epoch": 2.71, "learning_rate": 3.5822555710568985e-06, "loss": 0.5875, "step": 39935 }, { "epoch": 2.71, "learning_rate": 3.581312921835527e-06, "loss": 0.2786, "step": 39940 }, { "epoch": 2.71, "learning_rate": 3.5803702726141554e-06, "loss": 0.286, "step": 39945 }, { "epoch": 2.71, "learning_rate": 3.579427623392783e-06, "loss": 0.2073, "step": 39950 }, { "epoch": 2.71, "learning_rate": 3.5784849741714116e-06, "loss": 0.282, "step": 39955 }, { "epoch": 2.71, "learning_rate": 3.5775423249500396e-06, "loss": 0.2847, "step": 39960 }, { "epoch": 2.71, "learning_rate": 3.576599675728668e-06, "loss": 0.2078, "step": 39965 }, { "epoch": 2.71, "learning_rate": 3.5756570265072966e-06, "loss": 0.2302, "step": 39970 }, { "epoch": 2.71, "learning_rate": 3.5747143772859242e-06, "loss": 0.4602, "step": 39975 }, { "epoch": 2.71, "learning_rate": 3.5737717280645527e-06, "loss": 0.4755, "step": 39980 }, { "epoch": 2.71, "learning_rate": 3.5728290788431812e-06, "loss": 0.1442, "step": 39985 }, { "epoch": 2.71, "learning_rate": 3.5718864296218093e-06, "loss": 0.6373, "step": 39990 }, { "epoch": 2.71, "learning_rate": 3.5709437804004378e-06, "loss": 0.5303, "step": 39995 }, { "epoch": 2.71, "learning_rate": 3.5700011311790663e-06, "loss": 0.2184, "step": 40000 }, { "epoch": 2.72, "learning_rate": 3.569058481957694e-06, "loss": 0.1308, "step": 40005 }, { "epoch": 2.72, "learning_rate": 3.5681158327363224e-06, "loss": 0.2015, "step": 40010 }, { "epoch": 2.72, "learning_rate": 3.567173183514951e-06, "loss": 0.4195, "step": 40015 }, { "epoch": 2.72, "learning_rate": 3.566230534293579e-06, "loss": 0.1215, "step": 40020 }, { "epoch": 2.72, "learning_rate": 3.5652878850722074e-06, "loss": 0.6886, "step": 40025 }, { "epoch": 2.72, "learning_rate": 3.564345235850836e-06, "loss": 0.3383, "step": 40030 }, { "epoch": 2.72, "learning_rate": 3.5634025866294635e-06, "loss": 0.0874, "step": 40035 }, { "epoch": 2.72, "learning_rate": 3.562459937408092e-06, "loss": 0.3135, "step": 40040 }, { "epoch": 2.72, "learning_rate": 3.5615172881867205e-06, "loss": 0.3425, "step": 40045 }, { "epoch": 2.72, "learning_rate": 3.5605746389653486e-06, "loss": 0.175, "step": 40050 }, { "epoch": 2.72, "learning_rate": 3.5596319897439766e-06, "loss": 0.223, "step": 40055 }, { "epoch": 2.72, "learning_rate": 3.558689340522605e-06, "loss": 0.4816, "step": 40060 }, { "epoch": 2.72, "learning_rate": 3.557746691301233e-06, "loss": 0.6312, "step": 40065 }, { "epoch": 2.72, "learning_rate": 3.5568040420798617e-06, "loss": 0.2615, "step": 40070 }, { "epoch": 2.72, "learning_rate": 3.5558613928584893e-06, "loss": 0.1054, "step": 40075 }, { "epoch": 2.72, "learning_rate": 3.554918743637118e-06, "loss": 0.1932, "step": 40080 }, { "epoch": 2.72, "learning_rate": 3.5539760944157463e-06, "loss": 0.4654, "step": 40085 }, { "epoch": 2.72, "learning_rate": 3.5530334451943744e-06, "loss": 0.1257, "step": 40090 }, { "epoch": 2.72, "learning_rate": 3.552090795973003e-06, "loss": 0.15, "step": 40095 }, { "epoch": 2.72, "learning_rate": 3.5511481467516313e-06, "loss": 0.3402, "step": 40100 }, { "epoch": 2.72, "learning_rate": 3.550205497530259e-06, "loss": 0.1288, "step": 40105 }, { "epoch": 2.72, "learning_rate": 3.5492628483088875e-06, "loss": 0.7226, "step": 40110 }, { "epoch": 2.72, "learning_rate": 3.548320199087516e-06, "loss": 0.3599, "step": 40115 }, { "epoch": 2.72, "learning_rate": 3.547377549866144e-06, "loss": 0.5727, "step": 40120 }, { "epoch": 2.72, "learning_rate": 3.5464349006447725e-06, "loss": 0.315, "step": 40125 }, { "epoch": 2.72, "learning_rate": 3.545492251423401e-06, "loss": 0.3243, "step": 40130 }, { "epoch": 2.72, "learning_rate": 3.5445496022020286e-06, "loss": 0.2096, "step": 40135 }, { "epoch": 2.72, "learning_rate": 3.543606952980657e-06, "loss": 0.1406, "step": 40140 }, { "epoch": 2.72, "learning_rate": 3.5426643037592856e-06, "loss": 0.1962, "step": 40145 }, { "epoch": 2.72, "learning_rate": 3.5417216545379137e-06, "loss": 0.1521, "step": 40150 }, { "epoch": 2.73, "learning_rate": 3.540779005316542e-06, "loss": 0.4844, "step": 40155 }, { "epoch": 2.73, "learning_rate": 3.5398363560951702e-06, "loss": 0.07, "step": 40160 }, { "epoch": 2.73, "learning_rate": 3.5388937068737983e-06, "loss": 0.2961, "step": 40165 }, { "epoch": 2.73, "learning_rate": 3.5379510576524268e-06, "loss": 0.5397, "step": 40170 }, { "epoch": 2.73, "learning_rate": 3.5370084084310553e-06, "loss": 0.4096, "step": 40175 }, { "epoch": 2.73, "learning_rate": 3.536065759209683e-06, "loss": 0.4451, "step": 40180 }, { "epoch": 2.73, "learning_rate": 3.5351231099883114e-06, "loss": 0.2196, "step": 40185 }, { "epoch": 2.73, "learning_rate": 3.5341804607669395e-06, "loss": 0.692, "step": 40190 }, { "epoch": 2.73, "learning_rate": 3.533237811545568e-06, "loss": 0.3022, "step": 40195 }, { "epoch": 2.73, "learning_rate": 3.5322951623241964e-06, "loss": 0.4786, "step": 40200 }, { "epoch": 2.73, "learning_rate": 3.531352513102824e-06, "loss": 0.5909, "step": 40205 }, { "epoch": 2.73, "learning_rate": 3.5304098638814526e-06, "loss": 0.2113, "step": 40210 }, { "epoch": 2.73, "learning_rate": 3.529467214660081e-06, "loss": 0.1774, "step": 40215 }, { "epoch": 2.73, "learning_rate": 3.528524565438709e-06, "loss": 0.7004, "step": 40220 }, { "epoch": 2.73, "learning_rate": 3.5275819162173376e-06, "loss": 0.8964, "step": 40225 }, { "epoch": 2.73, "learning_rate": 3.526639266995966e-06, "loss": 0.0801, "step": 40230 }, { "epoch": 2.73, "learning_rate": 3.5256966177745937e-06, "loss": 0.3093, "step": 40235 }, { "epoch": 2.73, "learning_rate": 3.524753968553222e-06, "loss": 0.4303, "step": 40240 }, { "epoch": 2.73, "learning_rate": 3.5238113193318507e-06, "loss": 0.3595, "step": 40245 }, { "epoch": 2.73, "learning_rate": 3.5228686701104788e-06, "loss": 0.2215, "step": 40250 }, { "epoch": 2.73, "learning_rate": 3.5219260208891072e-06, "loss": 0.341, "step": 40255 }, { "epoch": 2.73, "learning_rate": 3.5209833716677353e-06, "loss": 0.0588, "step": 40260 }, { "epoch": 2.73, "learning_rate": 3.5200407224463634e-06, "loss": 0.2541, "step": 40265 }, { "epoch": 2.73, "learning_rate": 3.519098073224992e-06, "loss": 0.809, "step": 40270 }, { "epoch": 2.73, "learning_rate": 3.5181554240036203e-06, "loss": 0.13, "step": 40275 }, { "epoch": 2.73, "learning_rate": 3.5172127747822484e-06, "loss": 0.2198, "step": 40280 }, { "epoch": 2.73, "learning_rate": 3.5162701255608765e-06, "loss": 0.4066, "step": 40285 }, { "epoch": 2.73, "learning_rate": 3.515327476339505e-06, "loss": 0.317, "step": 40290 }, { "epoch": 2.73, "learning_rate": 3.514384827118133e-06, "loss": 0.1354, "step": 40295 }, { "epoch": 2.74, "learning_rate": 3.5134421778967615e-06, "loss": 0.5315, "step": 40300 }, { "epoch": 2.74, "learning_rate": 3.512499528675389e-06, "loss": 0.3392, "step": 40305 }, { "epoch": 2.74, "learning_rate": 3.5115568794540176e-06, "loss": 0.4766, "step": 40310 }, { "epoch": 2.74, "learning_rate": 3.510614230232646e-06, "loss": 0.3047, "step": 40315 }, { "epoch": 2.74, "learning_rate": 3.509671581011274e-06, "loss": 0.1378, "step": 40320 }, { "epoch": 2.74, "learning_rate": 3.5087289317899027e-06, "loss": 0.5603, "step": 40325 }, { "epoch": 2.74, "learning_rate": 3.507786282568531e-06, "loss": 0.3228, "step": 40330 }, { "epoch": 2.74, "learning_rate": 3.506843633347159e-06, "loss": 0.1212, "step": 40335 }, { "epoch": 2.74, "learning_rate": 3.5059009841257873e-06, "loss": 0.2284, "step": 40340 }, { "epoch": 2.74, "learning_rate": 3.5049583349044158e-06, "loss": 0.3656, "step": 40345 }, { "epoch": 2.74, "learning_rate": 3.504015685683044e-06, "loss": 0.3283, "step": 40350 }, { "epoch": 2.74, "learning_rate": 3.5030730364616723e-06, "loss": 0.7588, "step": 40355 }, { "epoch": 2.74, "learning_rate": 3.502130387240301e-06, "loss": 0.3881, "step": 40360 }, { "epoch": 2.74, "learning_rate": 3.5011877380189285e-06, "loss": 0.2662, "step": 40365 }, { "epoch": 2.74, "learning_rate": 3.500245088797557e-06, "loss": 0.0879, "step": 40370 }, { "epoch": 2.74, "learning_rate": 3.4993024395761854e-06, "loss": 0.3355, "step": 40375 }, { "epoch": 2.74, "learning_rate": 3.4983597903548135e-06, "loss": 0.3665, "step": 40380 }, { "epoch": 2.74, "learning_rate": 3.4974171411334416e-06, "loss": 0.4044, "step": 40385 }, { "epoch": 2.74, "learning_rate": 3.49647449191207e-06, "loss": 0.3688, "step": 40390 }, { "epoch": 2.74, "learning_rate": 3.495531842690698e-06, "loss": 0.3411, "step": 40395 }, { "epoch": 2.74, "learning_rate": 3.4945891934693266e-06, "loss": 0.11, "step": 40400 }, { "epoch": 2.74, "learning_rate": 3.4936465442479547e-06, "loss": 0.4655, "step": 40405 }, { "epoch": 2.74, "learning_rate": 3.4927038950265827e-06, "loss": 0.4609, "step": 40410 }, { "epoch": 2.74, "learning_rate": 3.4917612458052112e-06, "loss": 0.4783, "step": 40415 }, { "epoch": 2.74, "learning_rate": 3.4908185965838393e-06, "loss": 0.557, "step": 40420 }, { "epoch": 2.74, "learning_rate": 3.4898759473624678e-06, "loss": 0.4776, "step": 40425 }, { "epoch": 2.74, "learning_rate": 3.4889332981410963e-06, "loss": 0.6358, "step": 40430 }, { "epoch": 2.74, "learning_rate": 3.487990648919724e-06, "loss": 0.3389, "step": 40435 }, { "epoch": 2.74, "learning_rate": 3.4870479996983524e-06, "loss": 0.4607, "step": 40440 }, { "epoch": 2.75, "learning_rate": 3.486105350476981e-06, "loss": 0.1528, "step": 40445 }, { "epoch": 2.75, "learning_rate": 3.485162701255609e-06, "loss": 0.2019, "step": 40450 }, { "epoch": 2.75, "learning_rate": 3.4842200520342374e-06, "loss": 0.1856, "step": 40455 }, { "epoch": 2.75, "learning_rate": 3.483277402812866e-06, "loss": 0.3374, "step": 40460 }, { "epoch": 2.75, "learning_rate": 3.4823347535914935e-06, "loss": 0.3883, "step": 40465 }, { "epoch": 2.75, "learning_rate": 3.481392104370122e-06, "loss": 0.0989, "step": 40470 }, { "epoch": 2.75, "learning_rate": 3.4804494551487505e-06, "loss": 0.9293, "step": 40475 }, { "epoch": 2.75, "learning_rate": 3.4795068059273786e-06, "loss": 0.1407, "step": 40480 }, { "epoch": 2.75, "learning_rate": 3.478564156706007e-06, "loss": 0.1436, "step": 40485 }, { "epoch": 2.75, "learning_rate": 3.477621507484635e-06, "loss": 0.4245, "step": 40490 }, { "epoch": 2.75, "learning_rate": 3.476678858263263e-06, "loss": 0.2819, "step": 40495 }, { "epoch": 2.75, "learning_rate": 3.4757362090418917e-06, "loss": 0.258, "step": 40500 }, { "epoch": 2.75, "learning_rate": 3.47479355982052e-06, "loss": 0.1433, "step": 40505 }, { "epoch": 2.75, "learning_rate": 3.4738509105991482e-06, "loss": 0.1342, "step": 40510 }, { "epoch": 2.75, "learning_rate": 3.4729082613777763e-06, "loss": 0.7648, "step": 40515 }, { "epoch": 2.75, "learning_rate": 3.4719656121564044e-06, "loss": 0.1976, "step": 40520 }, { "epoch": 2.75, "learning_rate": 3.471022962935033e-06, "loss": 0.3332, "step": 40525 }, { "epoch": 2.75, "learning_rate": 3.4700803137136613e-06, "loss": 0.2861, "step": 40530 }, { "epoch": 2.75, "learning_rate": 3.469137664492289e-06, "loss": 0.2193, "step": 40535 }, { "epoch": 2.75, "learning_rate": 3.4681950152709175e-06, "loss": 0.397, "step": 40540 }, { "epoch": 2.75, "learning_rate": 3.467252366049546e-06, "loss": 0.1685, "step": 40545 }, { "epoch": 2.75, "learning_rate": 3.466309716828174e-06, "loss": 0.2389, "step": 40550 }, { "epoch": 2.75, "learning_rate": 3.4653670676068025e-06, "loss": 0.3239, "step": 40555 }, { "epoch": 2.75, "learning_rate": 3.464424418385431e-06, "loss": 0.187, "step": 40560 }, { "epoch": 2.75, "learning_rate": 3.4634817691640586e-06, "loss": 0.1492, "step": 40565 }, { "epoch": 2.75, "learning_rate": 3.462539119942687e-06, "loss": 0.419, "step": 40570 }, { "epoch": 2.75, "learning_rate": 3.4615964707213156e-06, "loss": 0.1806, "step": 40575 }, { "epoch": 2.75, "learning_rate": 3.4606538214999437e-06, "loss": 0.2343, "step": 40580 }, { "epoch": 2.75, "learning_rate": 3.459711172278572e-06, "loss": 0.3161, "step": 40585 }, { "epoch": 2.75, "learning_rate": 3.4587685230572006e-06, "loss": 0.0718, "step": 40590 }, { "epoch": 2.76, "learning_rate": 3.4578258738358283e-06, "loss": 0.3224, "step": 40595 }, { "epoch": 2.76, "learning_rate": 3.4568832246144568e-06, "loss": 0.4302, "step": 40600 }, { "epoch": 2.76, "learning_rate": 3.4559405753930853e-06, "loss": 0.4043, "step": 40605 }, { "epoch": 2.76, "learning_rate": 3.4549979261717133e-06, "loss": 0.1456, "step": 40610 }, { "epoch": 2.76, "learning_rate": 3.4540552769503414e-06, "loss": 1.0834, "step": 40615 }, { "epoch": 2.76, "learning_rate": 3.45311262772897e-06, "loss": 0.6223, "step": 40620 }, { "epoch": 2.76, "learning_rate": 3.452169978507598e-06, "loss": 0.4242, "step": 40625 }, { "epoch": 2.76, "learning_rate": 3.4512273292862264e-06, "loss": 0.2809, "step": 40630 }, { "epoch": 2.76, "learning_rate": 3.4502846800648545e-06, "loss": 0.2503, "step": 40635 }, { "epoch": 2.76, "learning_rate": 3.4493420308434826e-06, "loss": 0.2587, "step": 40640 }, { "epoch": 2.76, "learning_rate": 3.448399381622111e-06, "loss": 0.2806, "step": 40645 }, { "epoch": 2.76, "learning_rate": 3.447456732400739e-06, "loss": 0.3321, "step": 40650 }, { "epoch": 2.76, "learning_rate": 3.4465140831793676e-06, "loss": 0.5768, "step": 40655 }, { "epoch": 2.76, "learning_rate": 3.445571433957996e-06, "loss": 0.5697, "step": 40660 }, { "epoch": 2.76, "learning_rate": 3.4446287847366237e-06, "loss": 0.6882, "step": 40665 }, { "epoch": 2.76, "learning_rate": 3.443686135515252e-06, "loss": 0.1801, "step": 40670 }, { "epoch": 2.76, "learning_rate": 3.4427434862938807e-06, "loss": 0.1925, "step": 40675 }, { "epoch": 2.76, "learning_rate": 3.4418008370725088e-06, "loss": 0.608, "step": 40680 }, { "epoch": 2.76, "learning_rate": 3.4408581878511372e-06, "loss": 0.0429, "step": 40685 }, { "epoch": 2.76, "learning_rate": 3.4399155386297657e-06, "loss": 0.7437, "step": 40690 }, { "epoch": 2.76, "learning_rate": 3.4389728894083934e-06, "loss": 0.2537, "step": 40695 }, { "epoch": 2.76, "learning_rate": 3.438030240187022e-06, "loss": 0.2907, "step": 40700 }, { "epoch": 2.76, "learning_rate": 3.4370875909656504e-06, "loss": 0.4981, "step": 40705 }, { "epoch": 2.76, "learning_rate": 3.4361449417442784e-06, "loss": 0.2848, "step": 40710 }, { "epoch": 2.76, "learning_rate": 3.435202292522907e-06, "loss": 0.1622, "step": 40715 }, { "epoch": 2.76, "learning_rate": 3.434259643301535e-06, "loss": 0.2822, "step": 40720 }, { "epoch": 2.76, "learning_rate": 3.433316994080163e-06, "loss": 0.3638, "step": 40725 }, { "epoch": 2.76, "learning_rate": 3.4323743448587915e-06, "loss": 0.1024, "step": 40730 }, { "epoch": 2.76, "learning_rate": 3.43143169563742e-06, "loss": 0.4719, "step": 40735 }, { "epoch": 2.77, "learning_rate": 3.4304890464160476e-06, "loss": 0.5002, "step": 40740 }, { "epoch": 2.77, "learning_rate": 3.429546397194676e-06, "loss": 0.1112, "step": 40745 }, { "epoch": 2.77, "learning_rate": 3.428603747973304e-06, "loss": 0.3177, "step": 40750 }, { "epoch": 2.77, "learning_rate": 3.4276610987519327e-06, "loss": 0.306, "step": 40755 }, { "epoch": 2.77, "learning_rate": 3.426718449530561e-06, "loss": 0.2399, "step": 40760 }, { "epoch": 2.77, "learning_rate": 3.425775800309189e-06, "loss": 1.0306, "step": 40765 }, { "epoch": 2.77, "learning_rate": 3.4248331510878173e-06, "loss": 0.3106, "step": 40770 }, { "epoch": 2.77, "learning_rate": 3.4238905018664458e-06, "loss": 0.564, "step": 40775 }, { "epoch": 2.77, "learning_rate": 3.422947852645074e-06, "loss": 0.6481, "step": 40780 }, { "epoch": 2.77, "learning_rate": 3.4220052034237023e-06, "loss": 0.1394, "step": 40785 }, { "epoch": 2.77, "learning_rate": 3.421062554202331e-06, "loss": 0.3244, "step": 40790 }, { "epoch": 2.77, "learning_rate": 3.4201199049809585e-06, "loss": 0.4833, "step": 40795 }, { "epoch": 2.77, "learning_rate": 3.419177255759587e-06, "loss": 0.3863, "step": 40800 }, { "epoch": 2.77, "learning_rate": 3.4182346065382154e-06, "loss": 0.4814, "step": 40805 }, { "epoch": 2.77, "learning_rate": 3.4172919573168435e-06, "loss": 0.6131, "step": 40810 }, { "epoch": 2.77, "learning_rate": 3.416349308095472e-06, "loss": 0.1628, "step": 40815 }, { "epoch": 2.77, "learning_rate": 3.4154066588741005e-06, "loss": 1.1613, "step": 40820 }, { "epoch": 2.77, "learning_rate": 3.414464009652728e-06, "loss": 0.8634, "step": 40825 }, { "epoch": 2.77, "learning_rate": 3.4135213604313566e-06, "loss": 0.2999, "step": 40830 }, { "epoch": 2.77, "learning_rate": 3.412578711209985e-06, "loss": 0.2189, "step": 40835 }, { "epoch": 2.77, "learning_rate": 3.411636061988613e-06, "loss": 0.3164, "step": 40840 }, { "epoch": 2.77, "learning_rate": 3.4106934127672412e-06, "loss": 0.3192, "step": 40845 }, { "epoch": 2.77, "learning_rate": 3.4097507635458693e-06, "loss": 0.7948, "step": 40850 }, { "epoch": 2.77, "learning_rate": 3.4088081143244978e-06, "loss": 0.2583, "step": 40855 }, { "epoch": 2.77, "learning_rate": 3.4078654651031263e-06, "loss": 0.3743, "step": 40860 }, { "epoch": 2.77, "learning_rate": 3.406922815881754e-06, "loss": 0.0898, "step": 40865 }, { "epoch": 2.77, "learning_rate": 3.4059801666603824e-06, "loss": 0.8679, "step": 40870 }, { "epoch": 2.77, "learning_rate": 3.405037517439011e-06, "loss": 0.0464, "step": 40875 }, { "epoch": 2.77, "learning_rate": 3.404094868217639e-06, "loss": 0.1474, "step": 40880 }, { "epoch": 2.77, "learning_rate": 3.4031522189962674e-06, "loss": 0.4589, "step": 40885 }, { "epoch": 2.78, "learning_rate": 3.402209569774896e-06, "loss": 0.1202, "step": 40890 }, { "epoch": 2.78, "learning_rate": 3.4012669205535236e-06, "loss": 0.2002, "step": 40895 }, { "epoch": 2.78, "learning_rate": 3.400324271332152e-06, "loss": 0.4564, "step": 40900 }, { "epoch": 2.78, "learning_rate": 3.3993816221107805e-06, "loss": 0.405, "step": 40905 }, { "epoch": 2.78, "learning_rate": 3.3984389728894086e-06, "loss": 0.1707, "step": 40910 }, { "epoch": 2.78, "learning_rate": 3.397496323668037e-06, "loss": 0.185, "step": 40915 }, { "epoch": 2.78, "learning_rate": 3.3965536744466656e-06, "loss": 0.0954, "step": 40920 }, { "epoch": 2.78, "learning_rate": 3.395611025225293e-06, "loss": 0.0743, "step": 40925 }, { "epoch": 2.78, "learning_rate": 3.3946683760039217e-06, "loss": 0.3823, "step": 40930 }, { "epoch": 2.78, "learning_rate": 3.39372572678255e-06, "loss": 0.2949, "step": 40935 }, { "epoch": 2.78, "learning_rate": 3.3927830775611782e-06, "loss": 0.2606, "step": 40940 }, { "epoch": 2.78, "learning_rate": 3.3918404283398067e-06, "loss": 0.3314, "step": 40945 }, { "epoch": 2.78, "learning_rate": 3.390897779118435e-06, "loss": 0.9138, "step": 40950 }, { "epoch": 2.78, "learning_rate": 3.389955129897063e-06, "loss": 0.1833, "step": 40955 }, { "epoch": 2.78, "learning_rate": 3.3890124806756913e-06, "loss": 0.244, "step": 40960 }, { "epoch": 2.78, "learning_rate": 3.3880698314543194e-06, "loss": 0.3627, "step": 40965 }, { "epoch": 2.78, "learning_rate": 3.3871271822329475e-06, "loss": 0.5597, "step": 40970 }, { "epoch": 2.78, "learning_rate": 3.386184533011576e-06, "loss": 0.3312, "step": 40975 }, { "epoch": 2.78, "learning_rate": 3.385241883790204e-06, "loss": 0.3976, "step": 40980 }, { "epoch": 2.78, "learning_rate": 3.3842992345688325e-06, "loss": 0.3231, "step": 40985 }, { "epoch": 2.78, "learning_rate": 3.383356585347461e-06, "loss": 0.3074, "step": 40990 }, { "epoch": 2.78, "learning_rate": 3.3824139361260886e-06, "loss": 0.2668, "step": 40995 }, { "epoch": 2.78, "learning_rate": 3.381471286904717e-06, "loss": 1.0458, "step": 41000 }, { "epoch": 2.78, "learning_rate": 3.3805286376833456e-06, "loss": 0.2262, "step": 41005 }, { "epoch": 2.78, "learning_rate": 3.3795859884619737e-06, "loss": 0.2802, "step": 41010 }, { "epoch": 2.78, "learning_rate": 3.378643339240602e-06, "loss": 0.3299, "step": 41015 }, { "epoch": 2.78, "learning_rate": 3.3777006900192307e-06, "loss": 0.6301, "step": 41020 }, { "epoch": 2.78, "learning_rate": 3.3767580407978583e-06, "loss": 0.2421, "step": 41025 }, { "epoch": 2.78, "learning_rate": 3.3758153915764868e-06, "loss": 0.4784, "step": 41030 }, { "epoch": 2.79, "learning_rate": 3.3748727423551153e-06, "loss": 0.668, "step": 41035 }, { "epoch": 2.79, "learning_rate": 3.3739300931337433e-06, "loss": 0.5538, "step": 41040 }, { "epoch": 2.79, "learning_rate": 3.372987443912372e-06, "loss": 0.1428, "step": 41045 }, { "epoch": 2.79, "learning_rate": 3.3720447946910003e-06, "loss": 0.734, "step": 41050 }, { "epoch": 2.79, "learning_rate": 3.371102145469628e-06, "loss": 0.3781, "step": 41055 }, { "epoch": 2.79, "learning_rate": 3.3701594962482564e-06, "loss": 0.1142, "step": 41060 }, { "epoch": 2.79, "learning_rate": 3.369216847026885e-06, "loss": 0.3946, "step": 41065 }, { "epoch": 2.79, "learning_rate": 3.368274197805513e-06, "loss": 0.3543, "step": 41070 }, { "epoch": 2.79, "learning_rate": 3.367331548584141e-06, "loss": 0.2403, "step": 41075 }, { "epoch": 2.79, "learning_rate": 3.366388899362769e-06, "loss": 0.1331, "step": 41080 }, { "epoch": 2.79, "learning_rate": 3.3654462501413976e-06, "loss": 0.2609, "step": 41085 }, { "epoch": 2.79, "learning_rate": 3.364503600920026e-06, "loss": 0.4011, "step": 41090 }, { "epoch": 2.79, "learning_rate": 3.3635609516986537e-06, "loss": 0.3368, "step": 41095 }, { "epoch": 2.79, "learning_rate": 3.3626183024772822e-06, "loss": 1.0873, "step": 41100 }, { "epoch": 2.79, "learning_rate": 3.3616756532559107e-06, "loss": 0.4037, "step": 41105 }, { "epoch": 2.79, "learning_rate": 3.3607330040345388e-06, "loss": 0.1619, "step": 41110 }, { "epoch": 2.79, "learning_rate": 3.3597903548131673e-06, "loss": 0.4367, "step": 41115 }, { "epoch": 2.79, "learning_rate": 3.3588477055917957e-06, "loss": 0.7432, "step": 41120 }, { "epoch": 2.79, "learning_rate": 3.3579050563704234e-06, "loss": 0.0985, "step": 41125 }, { "epoch": 2.79, "learning_rate": 3.356962407149052e-06, "loss": 0.2411, "step": 41130 }, { "epoch": 2.79, "learning_rate": 3.3560197579276804e-06, "loss": 0.3969, "step": 41135 }, { "epoch": 2.79, "learning_rate": 3.3550771087063084e-06, "loss": 0.2376, "step": 41140 }, { "epoch": 2.79, "learning_rate": 3.354134459484937e-06, "loss": 0.5688, "step": 41145 }, { "epoch": 2.79, "learning_rate": 3.3531918102635654e-06, "loss": 0.1361, "step": 41150 }, { "epoch": 2.79, "learning_rate": 3.352249161042193e-06, "loss": 0.3247, "step": 41155 }, { "epoch": 2.79, "learning_rate": 3.3513065118208215e-06, "loss": 0.7859, "step": 41160 }, { "epoch": 2.79, "learning_rate": 3.35036386259945e-06, "loss": 0.6087, "step": 41165 }, { "epoch": 2.79, "learning_rate": 3.349421213378078e-06, "loss": 0.3227, "step": 41170 }, { "epoch": 2.79, "learning_rate": 3.3484785641567066e-06, "loss": 0.2534, "step": 41175 }, { "epoch": 2.79, "learning_rate": 3.3475359149353346e-06, "loss": 0.1682, "step": 41180 }, { "epoch": 2.8, "learning_rate": 3.3465932657139627e-06, "loss": 0.585, "step": 41185 }, { "epoch": 2.8, "learning_rate": 3.345650616492591e-06, "loss": 0.671, "step": 41190 }, { "epoch": 2.8, "learning_rate": 3.3447079672712192e-06, "loss": 0.1519, "step": 41195 }, { "epoch": 2.8, "learning_rate": 3.3437653180498473e-06, "loss": 0.1989, "step": 41200 }, { "epoch": 2.8, "learning_rate": 3.3428226688284758e-06, "loss": 0.1259, "step": 41205 }, { "epoch": 2.8, "learning_rate": 3.341880019607104e-06, "loss": 0.6367, "step": 41210 }, { "epoch": 2.8, "learning_rate": 3.3409373703857323e-06, "loss": 0.5847, "step": 41215 }, { "epoch": 2.8, "learning_rate": 3.339994721164361e-06, "loss": 0.4296, "step": 41220 }, { "epoch": 2.8, "learning_rate": 3.3390520719429885e-06, "loss": 1.1072, "step": 41225 }, { "epoch": 2.8, "learning_rate": 3.338109422721617e-06, "loss": 0.2479, "step": 41230 }, { "epoch": 2.8, "learning_rate": 3.3371667735002454e-06, "loss": 0.1688, "step": 41235 }, { "epoch": 2.8, "learning_rate": 3.3362241242788735e-06, "loss": 0.9774, "step": 41240 }, { "epoch": 2.8, "learning_rate": 3.335281475057502e-06, "loss": 0.1826, "step": 41245 }, { "epoch": 2.8, "learning_rate": 3.3343388258361305e-06, "loss": 0.2754, "step": 41250 }, { "epoch": 2.8, "learning_rate": 3.333396176614758e-06, "loss": 0.4557, "step": 41255 }, { "epoch": 2.8, "learning_rate": 3.3324535273933866e-06, "loss": 0.1752, "step": 41260 }, { "epoch": 2.8, "learning_rate": 3.331510878172015e-06, "loss": 0.8964, "step": 41265 }, { "epoch": 2.8, "learning_rate": 3.330568228950643e-06, "loss": 0.2007, "step": 41270 }, { "epoch": 2.8, "learning_rate": 3.3296255797292716e-06, "loss": 0.3194, "step": 41275 }, { "epoch": 2.8, "learning_rate": 3.3286829305079e-06, "loss": 0.1369, "step": 41280 }, { "epoch": 2.8, "learning_rate": 3.3277402812865278e-06, "loss": 0.4156, "step": 41285 }, { "epoch": 2.8, "learning_rate": 3.3267976320651563e-06, "loss": 0.185, "step": 41290 }, { "epoch": 2.8, "learning_rate": 3.3258549828437847e-06, "loss": 0.2747, "step": 41295 }, { "epoch": 2.8, "learning_rate": 3.324912333622413e-06, "loss": 0.4621, "step": 41300 }, { "epoch": 2.8, "learning_rate": 3.323969684401041e-06, "loss": 0.583, "step": 41305 }, { "epoch": 2.8, "learning_rate": 3.323027035179669e-06, "loss": 0.3605, "step": 41310 }, { "epoch": 2.8, "learning_rate": 3.3220843859582974e-06, "loss": 0.2009, "step": 41315 }, { "epoch": 2.8, "learning_rate": 3.321141736736926e-06, "loss": 0.2688, "step": 41320 }, { "epoch": 2.8, "learning_rate": 3.3201990875155536e-06, "loss": 0.2097, "step": 41325 }, { "epoch": 2.81, "learning_rate": 3.319256438294182e-06, "loss": 0.2921, "step": 41330 }, { "epoch": 2.81, "learning_rate": 3.3183137890728105e-06, "loss": 0.3151, "step": 41335 }, { "epoch": 2.81, "learning_rate": 3.3173711398514386e-06, "loss": 0.2697, "step": 41340 }, { "epoch": 2.81, "learning_rate": 3.316428490630067e-06, "loss": 0.0797, "step": 41345 }, { "epoch": 2.81, "learning_rate": 3.3154858414086956e-06, "loss": 0.3663, "step": 41350 }, { "epoch": 2.81, "learning_rate": 3.314543192187323e-06, "loss": 0.2705, "step": 41355 }, { "epoch": 2.81, "learning_rate": 3.3136005429659517e-06, "loss": 0.3076, "step": 41360 }, { "epoch": 2.81, "learning_rate": 3.31265789374458e-06, "loss": 0.4309, "step": 41365 }, { "epoch": 2.81, "learning_rate": 3.3117152445232082e-06, "loss": 0.6015, "step": 41370 }, { "epoch": 2.81, "learning_rate": 3.3107725953018367e-06, "loss": 0.431, "step": 41375 }, { "epoch": 2.81, "learning_rate": 3.3098299460804652e-06, "loss": 0.3729, "step": 41380 }, { "epoch": 2.81, "learning_rate": 3.308887296859093e-06, "loss": 0.2915, "step": 41385 }, { "epoch": 2.81, "learning_rate": 3.3079446476377213e-06, "loss": 0.2323, "step": 41390 }, { "epoch": 2.81, "learning_rate": 3.30700199841635e-06, "loss": 0.1323, "step": 41395 }, { "epoch": 2.81, "learning_rate": 3.306059349194978e-06, "loss": 0.2678, "step": 41400 }, { "epoch": 2.81, "learning_rate": 3.3051166999736064e-06, "loss": 0.185, "step": 41405 }, { "epoch": 2.81, "learning_rate": 3.304174050752234e-06, "loss": 0.3766, "step": 41410 }, { "epoch": 2.81, "learning_rate": 3.3032314015308625e-06, "loss": 0.4991, "step": 41415 }, { "epoch": 2.81, "learning_rate": 3.302288752309491e-06, "loss": 0.2537, "step": 41420 }, { "epoch": 2.81, "learning_rate": 3.301346103088119e-06, "loss": 0.4171, "step": 41425 }, { "epoch": 2.81, "learning_rate": 3.300403453866747e-06, "loss": 0.419, "step": 41430 }, { "epoch": 2.81, "learning_rate": 3.2994608046453756e-06, "loss": 0.5459, "step": 41435 }, { "epoch": 2.81, "learning_rate": 3.2985181554240037e-06, "loss": 0.1106, "step": 41440 }, { "epoch": 2.81, "learning_rate": 3.297575506202632e-06, "loss": 0.3534, "step": 41445 }, { "epoch": 2.81, "learning_rate": 3.2966328569812607e-06, "loss": 0.3897, "step": 41450 }, { "epoch": 2.81, "learning_rate": 3.2956902077598883e-06, "loss": 0.4867, "step": 41455 }, { "epoch": 2.81, "learning_rate": 3.2947475585385168e-06, "loss": 0.443, "step": 41460 }, { "epoch": 2.81, "learning_rate": 3.2938049093171453e-06, "loss": 0.3791, "step": 41465 }, { "epoch": 2.81, "learning_rate": 3.2928622600957733e-06, "loss": 0.3325, "step": 41470 }, { "epoch": 2.81, "learning_rate": 3.291919610874402e-06, "loss": 0.4328, "step": 41475 }, { "epoch": 2.82, "learning_rate": 3.2909769616530303e-06, "loss": 0.5464, "step": 41480 }, { "epoch": 2.82, "learning_rate": 3.290034312431658e-06, "loss": 0.2726, "step": 41485 }, { "epoch": 2.82, "learning_rate": 3.2890916632102864e-06, "loss": 0.5938, "step": 41490 }, { "epoch": 2.82, "learning_rate": 3.288149013988915e-06, "loss": 0.164, "step": 41495 }, { "epoch": 2.82, "learning_rate": 3.287206364767543e-06, "loss": 0.3038, "step": 41500 }, { "epoch": 2.82, "learning_rate": 3.2862637155461715e-06, "loss": 0.3153, "step": 41505 }, { "epoch": 2.82, "learning_rate": 3.2853210663247995e-06, "loss": 0.3395, "step": 41510 }, { "epoch": 2.82, "learning_rate": 3.2843784171034276e-06, "loss": 0.3993, "step": 41515 }, { "epoch": 2.82, "learning_rate": 3.283435767882056e-06, "loss": 0.5091, "step": 41520 }, { "epoch": 2.82, "learning_rate": 3.282493118660684e-06, "loss": 0.6163, "step": 41525 }, { "epoch": 2.82, "learning_rate": 3.2815504694393126e-06, "loss": 0.1085, "step": 41530 }, { "epoch": 2.82, "learning_rate": 3.2806078202179407e-06, "loss": 0.3405, "step": 41535 }, { "epoch": 2.82, "learning_rate": 3.2796651709965688e-06, "loss": 0.2236, "step": 41540 }, { "epoch": 2.82, "learning_rate": 3.2787225217751973e-06, "loss": 0.7479, "step": 41545 }, { "epoch": 2.82, "learning_rate": 3.2777798725538257e-06, "loss": 0.2009, "step": 41550 }, { "epoch": 2.82, "learning_rate": 3.2768372233324534e-06, "loss": 0.6106, "step": 41555 }, { "epoch": 2.82, "learning_rate": 3.275894574111082e-06, "loss": 0.4091, "step": 41560 }, { "epoch": 2.82, "learning_rate": 3.2749519248897104e-06, "loss": 0.1812, "step": 41565 }, { "epoch": 2.82, "learning_rate": 3.2740092756683384e-06, "loss": 0.6415, "step": 41570 }, { "epoch": 2.82, "learning_rate": 3.273066626446967e-06, "loss": 0.1366, "step": 41575 }, { "epoch": 2.82, "learning_rate": 3.2721239772255954e-06, "loss": 0.2681, "step": 41580 }, { "epoch": 2.82, "learning_rate": 3.271181328004223e-06, "loss": 0.2942, "step": 41585 }, { "epoch": 2.82, "learning_rate": 3.2702386787828515e-06, "loss": 0.6926, "step": 41590 }, { "epoch": 2.82, "learning_rate": 3.26929602956148e-06, "loss": 0.1374, "step": 41595 }, { "epoch": 2.82, "learning_rate": 3.268353380340108e-06, "loss": 0.2138, "step": 41600 }, { "epoch": 2.82, "learning_rate": 3.2674107311187366e-06, "loss": 0.3344, "step": 41605 }, { "epoch": 2.82, "learning_rate": 3.266468081897365e-06, "loss": 0.0742, "step": 41610 }, { "epoch": 2.82, "learning_rate": 3.2655254326759927e-06, "loss": 0.3257, "step": 41615 }, { "epoch": 2.82, "learning_rate": 3.264582783454621e-06, "loss": 0.2933, "step": 41620 }, { "epoch": 2.83, "learning_rate": 3.2636401342332497e-06, "loss": 0.2918, "step": 41625 }, { "epoch": 2.83, "learning_rate": 3.2626974850118777e-06, "loss": 0.4409, "step": 41630 }, { "epoch": 2.83, "learning_rate": 3.2617548357905062e-06, "loss": 0.2184, "step": 41635 }, { "epoch": 2.83, "learning_rate": 3.260812186569134e-06, "loss": 0.1841, "step": 41640 }, { "epoch": 2.83, "learning_rate": 3.2598695373477623e-06, "loss": 0.7372, "step": 41645 }, { "epoch": 2.83, "learning_rate": 3.258926888126391e-06, "loss": 0.3016, "step": 41650 }, { "epoch": 2.83, "learning_rate": 3.257984238905019e-06, "loss": 0.4127, "step": 41655 }, { "epoch": 2.83, "learning_rate": 3.257041589683647e-06, "loss": 0.0596, "step": 41660 }, { "epoch": 2.83, "learning_rate": 3.2560989404622754e-06, "loss": 0.3749, "step": 41665 }, { "epoch": 2.83, "learning_rate": 3.2551562912409035e-06, "loss": 0.2084, "step": 41670 }, { "epoch": 2.83, "learning_rate": 3.254213642019532e-06, "loss": 0.1383, "step": 41675 }, { "epoch": 2.83, "learning_rate": 3.2532709927981605e-06, "loss": 0.4512, "step": 41680 }, { "epoch": 2.83, "learning_rate": 3.252328343576788e-06, "loss": 0.5224, "step": 41685 }, { "epoch": 2.83, "learning_rate": 3.2513856943554166e-06, "loss": 0.4462, "step": 41690 }, { "epoch": 2.83, "learning_rate": 3.250443045134045e-06, "loss": 0.6828, "step": 41695 }, { "epoch": 2.83, "learning_rate": 3.249500395912673e-06, "loss": 0.608, "step": 41700 }, { "epoch": 2.83, "learning_rate": 3.2485577466913016e-06, "loss": 0.1994, "step": 41705 }, { "epoch": 2.83, "learning_rate": 3.24761509746993e-06, "loss": 0.3449, "step": 41710 }, { "epoch": 2.83, "learning_rate": 3.2466724482485578e-06, "loss": 0.1104, "step": 41715 }, { "epoch": 2.83, "learning_rate": 3.2457297990271863e-06, "loss": 0.6116, "step": 41720 }, { "epoch": 2.83, "learning_rate": 3.2447871498058148e-06, "loss": 0.2576, "step": 41725 }, { "epoch": 2.83, "learning_rate": 3.243844500584443e-06, "loss": 0.4802, "step": 41730 }, { "epoch": 2.83, "learning_rate": 3.2429018513630713e-06, "loss": 0.3662, "step": 41735 }, { "epoch": 2.83, "learning_rate": 3.2419592021416994e-06, "loss": 0.4927, "step": 41740 }, { "epoch": 2.83, "learning_rate": 3.2410165529203274e-06, "loss": 0.8229, "step": 41745 }, { "epoch": 2.83, "learning_rate": 3.240073903698956e-06, "loss": 0.2822, "step": 41750 }, { "epoch": 2.83, "learning_rate": 3.239131254477584e-06, "loss": 0.2416, "step": 41755 }, { "epoch": 2.83, "learning_rate": 3.2381886052562125e-06, "loss": 0.3586, "step": 41760 }, { "epoch": 2.83, "learning_rate": 3.2372459560348405e-06, "loss": 0.0598, "step": 41765 }, { "epoch": 2.83, "learning_rate": 3.2363033068134686e-06, "loss": 0.1282, "step": 41770 }, { "epoch": 2.84, "learning_rate": 3.235360657592097e-06, "loss": 0.1832, "step": 41775 }, { "epoch": 2.84, "learning_rate": 3.2344180083707256e-06, "loss": 0.1289, "step": 41780 }, { "epoch": 2.84, "learning_rate": 3.233475359149353e-06, "loss": 0.5273, "step": 41785 }, { "epoch": 2.84, "learning_rate": 3.2325327099279817e-06, "loss": 0.9435, "step": 41790 }, { "epoch": 2.84, "learning_rate": 3.23159006070661e-06, "loss": 0.2788, "step": 41795 }, { "epoch": 2.84, "learning_rate": 3.2306474114852382e-06, "loss": 0.3169, "step": 41800 }, { "epoch": 2.84, "learning_rate": 3.2297047622638667e-06, "loss": 0.7296, "step": 41805 }, { "epoch": 2.84, "learning_rate": 3.2287621130424952e-06, "loss": 0.3334, "step": 41810 }, { "epoch": 2.84, "learning_rate": 3.227819463821123e-06, "loss": 0.1601, "step": 41815 }, { "epoch": 2.84, "learning_rate": 3.2268768145997514e-06, "loss": 0.2358, "step": 41820 }, { "epoch": 2.84, "learning_rate": 3.22593416537838e-06, "loss": 0.292, "step": 41825 }, { "epoch": 2.84, "learning_rate": 3.224991516157008e-06, "loss": 0.3227, "step": 41830 }, { "epoch": 2.84, "learning_rate": 3.2240488669356364e-06, "loss": 0.54, "step": 41835 }, { "epoch": 2.84, "learning_rate": 3.223106217714265e-06, "loss": 0.5762, "step": 41840 }, { "epoch": 2.84, "learning_rate": 3.2221635684928925e-06, "loss": 0.1807, "step": 41845 }, { "epoch": 2.84, "learning_rate": 3.221220919271521e-06, "loss": 0.6744, "step": 41850 }, { "epoch": 2.84, "learning_rate": 3.220278270050149e-06, "loss": 0.2654, "step": 41855 }, { "epoch": 2.84, "learning_rate": 3.2193356208287776e-06, "loss": 0.4365, "step": 41860 }, { "epoch": 2.84, "learning_rate": 3.2183929716074056e-06, "loss": 0.2586, "step": 41865 }, { "epoch": 2.84, "learning_rate": 3.2174503223860337e-06, "loss": 0.32, "step": 41870 }, { "epoch": 2.84, "learning_rate": 3.216507673164662e-06, "loss": 0.3259, "step": 41875 }, { "epoch": 2.84, "learning_rate": 3.2155650239432907e-06, "loss": 0.1352, "step": 41880 }, { "epoch": 2.84, "learning_rate": 3.2146223747219187e-06, "loss": 0.636, "step": 41885 }, { "epoch": 2.84, "learning_rate": 3.2136797255005468e-06, "loss": 1.0012, "step": 41890 }, { "epoch": 2.84, "learning_rate": 3.2127370762791753e-06, "loss": 0.2173, "step": 41895 }, { "epoch": 2.84, "learning_rate": 3.2117944270578033e-06, "loss": 0.2198, "step": 41900 }, { "epoch": 2.84, "learning_rate": 3.210851777836432e-06, "loss": 0.1126, "step": 41905 }, { "epoch": 2.84, "learning_rate": 3.2099091286150603e-06, "loss": 0.1933, "step": 41910 }, { "epoch": 2.84, "learning_rate": 3.208966479393688e-06, "loss": 0.6729, "step": 41915 }, { "epoch": 2.85, "learning_rate": 3.2080238301723164e-06, "loss": 0.1225, "step": 41920 }, { "epoch": 2.85, "learning_rate": 3.207081180950945e-06, "loss": 0.1844, "step": 41925 }, { "epoch": 2.85, "learning_rate": 3.206138531729573e-06, "loss": 0.1697, "step": 41930 }, { "epoch": 2.85, "learning_rate": 3.2051958825082015e-06, "loss": 0.2393, "step": 41935 }, { "epoch": 2.85, "learning_rate": 3.20425323328683e-06, "loss": 0.2966, "step": 41940 }, { "epoch": 2.85, "learning_rate": 3.2033105840654576e-06, "loss": 0.219, "step": 41945 }, { "epoch": 2.85, "learning_rate": 3.202367934844086e-06, "loss": 0.5566, "step": 41950 }, { "epoch": 2.85, "learning_rate": 3.2014252856227146e-06, "loss": 0.2284, "step": 41955 }, { "epoch": 2.85, "learning_rate": 3.2004826364013426e-06, "loss": 0.4253, "step": 41960 }, { "epoch": 2.85, "learning_rate": 3.199539987179971e-06, "loss": 0.2652, "step": 41965 }, { "epoch": 2.85, "learning_rate": 3.1985973379585988e-06, "loss": 0.1311, "step": 41970 }, { "epoch": 2.85, "learning_rate": 3.1976546887372273e-06, "loss": 0.3962, "step": 41975 }, { "epoch": 2.85, "learning_rate": 3.1967120395158557e-06, "loss": 0.7012, "step": 41980 }, { "epoch": 2.85, "learning_rate": 3.195769390294484e-06, "loss": 0.7025, "step": 41985 }, { "epoch": 2.85, "learning_rate": 3.194826741073112e-06, "loss": 0.1509, "step": 41990 }, { "epoch": 2.85, "learning_rate": 3.1938840918517404e-06, "loss": 0.0961, "step": 41995 }, { "epoch": 2.85, "learning_rate": 3.1929414426303684e-06, "loss": 0.4874, "step": 42000 }, { "epoch": 2.85, "learning_rate": 3.191998793408997e-06, "loss": 0.1313, "step": 42005 }, { "epoch": 2.85, "learning_rate": 3.1910561441876254e-06, "loss": 0.2582, "step": 42010 }, { "epoch": 2.85, "learning_rate": 3.190113494966253e-06, "loss": 0.5387, "step": 42015 }, { "epoch": 2.85, "learning_rate": 3.1891708457448815e-06, "loss": 0.3526, "step": 42020 }, { "epoch": 2.85, "learning_rate": 3.18822819652351e-06, "loss": 0.0613, "step": 42025 }, { "epoch": 2.85, "learning_rate": 3.187285547302138e-06, "loss": 0.4592, "step": 42030 }, { "epoch": 2.85, "learning_rate": 3.1863428980807666e-06, "loss": 0.0265, "step": 42035 }, { "epoch": 2.85, "learning_rate": 3.185400248859395e-06, "loss": 0.4952, "step": 42040 }, { "epoch": 2.85, "learning_rate": 3.1844575996380227e-06, "loss": 0.0551, "step": 42045 }, { "epoch": 2.85, "learning_rate": 3.183514950416651e-06, "loss": 0.2768, "step": 42050 }, { "epoch": 2.85, "learning_rate": 3.1825723011952797e-06, "loss": 0.2532, "step": 42055 }, { "epoch": 2.85, "learning_rate": 3.1816296519739077e-06, "loss": 0.1574, "step": 42060 }, { "epoch": 2.85, "learning_rate": 3.1806870027525362e-06, "loss": 0.2854, "step": 42065 }, { "epoch": 2.86, "learning_rate": 3.1797443535311647e-06, "loss": 0.1064, "step": 42070 }, { "epoch": 2.86, "learning_rate": 3.1788017043097923e-06, "loss": 0.1385, "step": 42075 }, { "epoch": 2.86, "learning_rate": 3.177859055088421e-06, "loss": 0.3873, "step": 42080 }, { "epoch": 2.86, "learning_rate": 3.176916405867049e-06, "loss": 0.4277, "step": 42085 }, { "epoch": 2.86, "learning_rate": 3.1759737566456774e-06, "loss": 0.3967, "step": 42090 }, { "epoch": 2.86, "learning_rate": 3.1750311074243054e-06, "loss": 0.4899, "step": 42095 }, { "epoch": 2.86, "learning_rate": 3.1740884582029335e-06, "loss": 0.2733, "step": 42100 }, { "epoch": 2.86, "learning_rate": 3.173145808981562e-06, "loss": 0.6472, "step": 42105 }, { "epoch": 2.86, "learning_rate": 3.1722031597601905e-06, "loss": 0.2817, "step": 42110 }, { "epoch": 2.86, "learning_rate": 3.1712605105388185e-06, "loss": 0.226, "step": 42115 }, { "epoch": 2.86, "learning_rate": 3.1703178613174466e-06, "loss": 0.124, "step": 42120 }, { "epoch": 2.86, "learning_rate": 3.169375212096075e-06, "loss": 0.1033, "step": 42125 }, { "epoch": 2.86, "learning_rate": 3.168432562874703e-06, "loss": 0.7909, "step": 42130 }, { "epoch": 2.86, "learning_rate": 3.1674899136533317e-06, "loss": 0.0324, "step": 42135 }, { "epoch": 2.86, "learning_rate": 3.16654726443196e-06, "loss": 0.1252, "step": 42140 }, { "epoch": 2.86, "learning_rate": 3.1656046152105878e-06, "loss": 0.2593, "step": 42145 }, { "epoch": 2.86, "learning_rate": 3.1646619659892163e-06, "loss": 0.1314, "step": 42150 }, { "epoch": 2.86, "learning_rate": 3.1637193167678448e-06, "loss": 0.1717, "step": 42155 }, { "epoch": 2.86, "learning_rate": 3.162776667546473e-06, "loss": 0.5813, "step": 42160 }, { "epoch": 2.86, "learning_rate": 3.1618340183251013e-06, "loss": 0.5039, "step": 42165 }, { "epoch": 2.86, "learning_rate": 3.16089136910373e-06, "loss": 0.3278, "step": 42170 }, { "epoch": 2.86, "learning_rate": 3.1599487198823574e-06, "loss": 0.4248, "step": 42175 }, { "epoch": 2.86, "learning_rate": 3.159006070660986e-06, "loss": 0.2105, "step": 42180 }, { "epoch": 2.86, "learning_rate": 3.1580634214396144e-06, "loss": 0.1619, "step": 42185 }, { "epoch": 2.86, "learning_rate": 3.1571207722182425e-06, "loss": 0.4739, "step": 42190 }, { "epoch": 2.86, "learning_rate": 3.156178122996871e-06, "loss": 0.4078, "step": 42195 }, { "epoch": 2.86, "learning_rate": 3.1552354737754986e-06, "loss": 0.3672, "step": 42200 }, { "epoch": 2.86, "learning_rate": 3.154292824554127e-06, "loss": 0.1991, "step": 42205 }, { "epoch": 2.86, "learning_rate": 3.1533501753327556e-06, "loss": 0.4675, "step": 42210 }, { "epoch": 2.87, "learning_rate": 3.1524075261113836e-06, "loss": 0.604, "step": 42215 }, { "epoch": 2.87, "learning_rate": 3.1514648768900117e-06, "loss": 0.2738, "step": 42220 }, { "epoch": 2.87, "learning_rate": 3.15052222766864e-06, "loss": 0.1469, "step": 42225 }, { "epoch": 2.87, "learning_rate": 3.1495795784472683e-06, "loss": 0.2997, "step": 42230 }, { "epoch": 2.87, "learning_rate": 3.1486369292258967e-06, "loss": 0.6661, "step": 42235 }, { "epoch": 2.87, "learning_rate": 3.1476942800045252e-06, "loss": 0.3522, "step": 42240 }, { "epoch": 2.87, "learning_rate": 3.146751630783153e-06, "loss": 0.3224, "step": 42245 }, { "epoch": 2.87, "learning_rate": 3.1458089815617814e-06, "loss": 0.4432, "step": 42250 }, { "epoch": 2.87, "learning_rate": 3.14486633234041e-06, "loss": 0.3535, "step": 42255 }, { "epoch": 2.87, "learning_rate": 3.143923683119038e-06, "loss": 0.1884, "step": 42260 }, { "epoch": 2.87, "learning_rate": 3.1429810338976664e-06, "loss": 0.2924, "step": 42265 }, { "epoch": 2.87, "learning_rate": 3.142038384676295e-06, "loss": 0.3269, "step": 42270 }, { "epoch": 2.87, "learning_rate": 3.1410957354549225e-06, "loss": 0.6022, "step": 42275 }, { "epoch": 2.87, "learning_rate": 3.140153086233551e-06, "loss": 0.6335, "step": 42280 }, { "epoch": 2.87, "learning_rate": 3.1392104370121795e-06, "loss": 0.4, "step": 42285 }, { "epoch": 2.87, "learning_rate": 3.1382677877908076e-06, "loss": 0.5037, "step": 42290 }, { "epoch": 2.87, "learning_rate": 3.137325138569436e-06, "loss": 0.7732, "step": 42295 }, { "epoch": 2.87, "learning_rate": 3.1363824893480645e-06, "loss": 0.5511, "step": 42300 }, { "epoch": 2.87, "learning_rate": 3.135439840126692e-06, "loss": 0.3044, "step": 42305 }, { "epoch": 2.87, "learning_rate": 3.1344971909053207e-06, "loss": 0.3883, "step": 42310 }, { "epoch": 2.87, "learning_rate": 3.1335545416839487e-06, "loss": 0.1379, "step": 42315 }, { "epoch": 2.87, "learning_rate": 3.132611892462577e-06, "loss": 0.4629, "step": 42320 }, { "epoch": 2.87, "learning_rate": 3.1316692432412053e-06, "loss": 0.1046, "step": 42325 }, { "epoch": 2.87, "learning_rate": 3.1307265940198333e-06, "loss": 0.2246, "step": 42330 }, { "epoch": 2.87, "learning_rate": 3.129783944798462e-06, "loss": 0.2387, "step": 42335 }, { "epoch": 2.87, "learning_rate": 3.1288412955770903e-06, "loss": 0.2718, "step": 42340 }, { "epoch": 2.87, "learning_rate": 3.127898646355718e-06, "loss": 0.2277, "step": 42345 }, { "epoch": 2.87, "learning_rate": 3.1269559971343464e-06, "loss": 0.867, "step": 42350 }, { "epoch": 2.87, "learning_rate": 3.126013347912975e-06, "loss": 0.7086, "step": 42355 }, { "epoch": 2.87, "learning_rate": 3.125070698691603e-06, "loss": 0.0574, "step": 42360 }, { "epoch": 2.88, "learning_rate": 3.1241280494702315e-06, "loss": 0.2551, "step": 42365 }, { "epoch": 2.88, "learning_rate": 3.12318540024886e-06, "loss": 0.2499, "step": 42370 }, { "epoch": 2.88, "learning_rate": 3.1222427510274876e-06, "loss": 0.2615, "step": 42375 }, { "epoch": 2.88, "learning_rate": 3.121300101806116e-06, "loss": 0.2065, "step": 42380 }, { "epoch": 2.88, "learning_rate": 3.1203574525847446e-06, "loss": 0.2182, "step": 42385 }, { "epoch": 2.88, "learning_rate": 3.1194148033633726e-06, "loss": 0.1665, "step": 42390 }, { "epoch": 2.88, "learning_rate": 3.118472154142001e-06, "loss": 0.7541, "step": 42395 }, { "epoch": 2.88, "learning_rate": 3.1175295049206296e-06, "loss": 0.3994, "step": 42400 }, { "epoch": 2.88, "learning_rate": 3.1165868556992573e-06, "loss": 0.4155, "step": 42405 }, { "epoch": 2.88, "learning_rate": 3.1156442064778857e-06, "loss": 0.6997, "step": 42410 }, { "epoch": 2.88, "learning_rate": 3.114701557256514e-06, "loss": 0.3536, "step": 42415 }, { "epoch": 2.88, "learning_rate": 3.1137589080351423e-06, "loss": 0.4511, "step": 42420 }, { "epoch": 2.88, "learning_rate": 3.1128162588137708e-06, "loss": 0.0964, "step": 42425 }, { "epoch": 2.88, "learning_rate": 3.1118736095923984e-06, "loss": 0.5893, "step": 42430 }, { "epoch": 2.88, "learning_rate": 3.110930960371027e-06, "loss": 0.2897, "step": 42435 }, { "epoch": 2.88, "learning_rate": 3.1099883111496554e-06, "loss": 0.0635, "step": 42440 }, { "epoch": 2.88, "learning_rate": 3.1090456619282835e-06, "loss": 0.5981, "step": 42445 }, { "epoch": 2.88, "learning_rate": 3.1081030127069115e-06, "loss": 0.3989, "step": 42450 }, { "epoch": 2.88, "learning_rate": 3.10716036348554e-06, "loss": 0.5795, "step": 42455 }, { "epoch": 2.88, "learning_rate": 3.106217714264168e-06, "loss": 0.0864, "step": 42460 }, { "epoch": 2.88, "learning_rate": 3.1052750650427966e-06, "loss": 0.4033, "step": 42465 }, { "epoch": 2.88, "learning_rate": 3.104332415821425e-06, "loss": 0.1589, "step": 42470 }, { "epoch": 2.88, "learning_rate": 3.1033897666000527e-06, "loss": 0.2933, "step": 42475 }, { "epoch": 2.88, "learning_rate": 3.102447117378681e-06, "loss": 0.2033, "step": 42480 }, { "epoch": 2.88, "learning_rate": 3.1015044681573097e-06, "loss": 0.2819, "step": 42485 }, { "epoch": 2.88, "learning_rate": 3.1005618189359377e-06, "loss": 0.3004, "step": 42490 }, { "epoch": 2.88, "learning_rate": 3.0996191697145662e-06, "loss": 0.6232, "step": 42495 }, { "epoch": 2.88, "learning_rate": 3.0986765204931947e-06, "loss": 0.2355, "step": 42500 }, { "epoch": 2.88, "learning_rate": 3.0977338712718223e-06, "loss": 0.2301, "step": 42505 }, { "epoch": 2.89, "learning_rate": 3.096791222050451e-06, "loss": 0.1341, "step": 42510 }, { "epoch": 2.89, "learning_rate": 3.0958485728290793e-06, "loss": 0.3375, "step": 42515 }, { "epoch": 2.89, "learning_rate": 3.0949059236077074e-06, "loss": 0.1052, "step": 42520 }, { "epoch": 2.89, "learning_rate": 3.093963274386336e-06, "loss": 0.4861, "step": 42525 }, { "epoch": 2.89, "learning_rate": 3.0930206251649635e-06, "loss": 0.2522, "step": 42530 }, { "epoch": 2.89, "learning_rate": 3.092077975943592e-06, "loss": 0.291, "step": 42535 }, { "epoch": 2.89, "learning_rate": 3.0911353267222205e-06, "loss": 0.343, "step": 42540 }, { "epoch": 2.89, "learning_rate": 3.0901926775008486e-06, "loss": 0.5901, "step": 42545 }, { "epoch": 2.89, "learning_rate": 3.089250028279477e-06, "loss": 0.6252, "step": 42550 }, { "epoch": 2.89, "learning_rate": 3.088307379058105e-06, "loss": 0.5352, "step": 42555 }, { "epoch": 2.89, "learning_rate": 3.087364729836733e-06, "loss": 0.6647, "step": 42560 }, { "epoch": 2.89, "learning_rate": 3.0864220806153617e-06, "loss": 0.332, "step": 42565 }, { "epoch": 2.89, "learning_rate": 3.08547943139399e-06, "loss": 0.375, "step": 42570 }, { "epoch": 2.89, "learning_rate": 3.0845367821726178e-06, "loss": 0.5627, "step": 42575 }, { "epoch": 2.89, "learning_rate": 3.0835941329512463e-06, "loss": 0.3222, "step": 42580 }, { "epoch": 2.89, "learning_rate": 3.0826514837298748e-06, "loss": 0.2443, "step": 42585 }, { "epoch": 2.89, "learning_rate": 3.081708834508503e-06, "loss": 0.1018, "step": 42590 }, { "epoch": 2.89, "learning_rate": 3.0807661852871313e-06, "loss": 0.1683, "step": 42595 }, { "epoch": 2.89, "learning_rate": 3.07982353606576e-06, "loss": 0.1576, "step": 42600 }, { "epoch": 2.89, "learning_rate": 3.0788808868443874e-06, "loss": 0.277, "step": 42605 }, { "epoch": 2.89, "learning_rate": 3.077938237623016e-06, "loss": 0.3853, "step": 42610 }, { "epoch": 2.89, "learning_rate": 3.0769955884016444e-06, "loss": 0.3821, "step": 42615 }, { "epoch": 2.89, "learning_rate": 3.0760529391802725e-06, "loss": 0.3819, "step": 42620 }, { "epoch": 2.89, "learning_rate": 3.075110289958901e-06, "loss": 0.4362, "step": 42625 }, { "epoch": 2.89, "learning_rate": 3.0741676407375294e-06, "loss": 0.4703, "step": 42630 }, { "epoch": 2.89, "learning_rate": 3.073224991516157e-06, "loss": 0.5839, "step": 42635 }, { "epoch": 2.89, "learning_rate": 3.0722823422947856e-06, "loss": 0.1927, "step": 42640 }, { "epoch": 2.89, "learning_rate": 3.0713396930734136e-06, "loss": 0.3878, "step": 42645 }, { "epoch": 2.89, "learning_rate": 3.070397043852042e-06, "loss": 0.3164, "step": 42650 }, { "epoch": 2.9, "learning_rate": 3.0694543946306706e-06, "loss": 0.391, "step": 42655 }, { "epoch": 2.9, "learning_rate": 3.0685117454092983e-06, "loss": 0.9931, "step": 42660 }, { "epoch": 2.9, "learning_rate": 3.0675690961879267e-06, "loss": 0.1206, "step": 42665 }, { "epoch": 2.9, "learning_rate": 3.0666264469665552e-06, "loss": 0.2234, "step": 42670 }, { "epoch": 2.9, "learning_rate": 3.0656837977451833e-06, "loss": 0.0988, "step": 42675 }, { "epoch": 2.9, "learning_rate": 3.0647411485238114e-06, "loss": 0.4443, "step": 42680 }, { "epoch": 2.9, "learning_rate": 3.06379849930244e-06, "loss": 0.5153, "step": 42685 }, { "epoch": 2.9, "learning_rate": 3.062855850081068e-06, "loss": 0.5212, "step": 42690 }, { "epoch": 2.9, "learning_rate": 3.0619132008596964e-06, "loss": 0.5186, "step": 42695 }, { "epoch": 2.9, "learning_rate": 3.060970551638325e-06, "loss": 0.2425, "step": 42700 }, { "epoch": 2.9, "learning_rate": 3.0600279024169525e-06, "loss": 0.3433, "step": 42705 }, { "epoch": 2.9, "learning_rate": 3.059085253195581e-06, "loss": 0.3881, "step": 42710 }, { "epoch": 2.9, "learning_rate": 3.0581426039742095e-06, "loss": 0.7007, "step": 42715 }, { "epoch": 2.9, "learning_rate": 3.0571999547528376e-06, "loss": 0.3546, "step": 42720 }, { "epoch": 2.9, "learning_rate": 3.056257305531466e-06, "loss": 0.4596, "step": 42725 }, { "epoch": 2.9, "learning_rate": 3.0553146563100945e-06, "loss": 0.4107, "step": 42730 }, { "epoch": 2.9, "learning_rate": 3.054372007088722e-06, "loss": 0.2083, "step": 42735 }, { "epoch": 2.9, "learning_rate": 3.0534293578673507e-06, "loss": 0.7282, "step": 42740 }, { "epoch": 2.9, "learning_rate": 3.052486708645979e-06, "loss": 0.1826, "step": 42745 }, { "epoch": 2.9, "learning_rate": 3.0515440594246072e-06, "loss": 0.2486, "step": 42750 }, { "epoch": 2.9, "learning_rate": 3.0506014102032357e-06, "loss": 0.2182, "step": 42755 }, { "epoch": 2.9, "learning_rate": 3.0496587609818633e-06, "loss": 0.4783, "step": 42760 }, { "epoch": 2.9, "learning_rate": 3.048716111760492e-06, "loss": 0.2414, "step": 42765 }, { "epoch": 2.9, "learning_rate": 3.0477734625391203e-06, "loss": 0.4359, "step": 42770 }, { "epoch": 2.9, "learning_rate": 3.0468308133177484e-06, "loss": 0.6376, "step": 42775 }, { "epoch": 2.9, "learning_rate": 3.045888164096377e-06, "loss": 0.1696, "step": 42780 }, { "epoch": 2.9, "learning_rate": 3.044945514875005e-06, "loss": 0.4266, "step": 42785 }, { "epoch": 2.9, "learning_rate": 3.044002865653633e-06, "loss": 0.378, "step": 42790 }, { "epoch": 2.9, "learning_rate": 3.0430602164322615e-06, "loss": 0.0914, "step": 42795 }, { "epoch": 2.9, "learning_rate": 3.04211756721089e-06, "loss": 0.3683, "step": 42800 }, { "epoch": 2.91, "learning_rate": 3.0411749179895176e-06, "loss": 0.4274, "step": 42805 }, { "epoch": 2.91, "learning_rate": 3.040232268768146e-06, "loss": 0.2925, "step": 42810 }, { "epoch": 2.91, "learning_rate": 3.0392896195467746e-06, "loss": 0.1212, "step": 42815 }, { "epoch": 2.91, "learning_rate": 3.0383469703254026e-06, "loss": 0.0659, "step": 42820 }, { "epoch": 2.91, "learning_rate": 3.037404321104031e-06, "loss": 0.1918, "step": 42825 }, { "epoch": 2.91, "learning_rate": 3.0364616718826596e-06, "loss": 0.3934, "step": 42830 }, { "epoch": 2.91, "learning_rate": 3.0355190226612873e-06, "loss": 0.5327, "step": 42835 }, { "epoch": 2.91, "learning_rate": 3.0345763734399158e-06, "loss": 0.5691, "step": 42840 }, { "epoch": 2.91, "learning_rate": 3.0336337242185442e-06, "loss": 0.7265, "step": 42845 }, { "epoch": 2.91, "learning_rate": 3.0326910749971723e-06, "loss": 0.4016, "step": 42850 }, { "epoch": 2.91, "learning_rate": 3.031748425775801e-06, "loss": 0.2775, "step": 42855 }, { "epoch": 2.91, "learning_rate": 3.0308057765544293e-06, "loss": 0.5518, "step": 42860 }, { "epoch": 2.91, "learning_rate": 3.029863127333057e-06, "loss": 0.3083, "step": 42865 }, { "epoch": 2.91, "learning_rate": 3.0289204781116854e-06, "loss": 0.5571, "step": 42870 }, { "epoch": 2.91, "learning_rate": 3.0279778288903135e-06, "loss": 0.1918, "step": 42875 }, { "epoch": 2.91, "learning_rate": 3.027035179668942e-06, "loss": 0.269, "step": 42880 }, { "epoch": 2.91, "learning_rate": 3.0260925304475704e-06, "loss": 0.3518, "step": 42885 }, { "epoch": 2.91, "learning_rate": 3.025149881226198e-06, "loss": 0.3794, "step": 42890 }, { "epoch": 2.91, "learning_rate": 3.0242072320048266e-06, "loss": 0.3821, "step": 42895 }, { "epoch": 2.91, "learning_rate": 3.023264582783455e-06, "loss": 0.1621, "step": 42900 }, { "epoch": 2.91, "learning_rate": 3.022321933562083e-06, "loss": 0.4273, "step": 42905 }, { "epoch": 2.91, "learning_rate": 3.021379284340711e-06, "loss": 0.388, "step": 42910 }, { "epoch": 2.91, "learning_rate": 3.0204366351193397e-06, "loss": 0.6626, "step": 42915 }, { "epoch": 2.91, "learning_rate": 3.0194939858979677e-06, "loss": 0.3266, "step": 42920 }, { "epoch": 2.91, "learning_rate": 3.0185513366765962e-06, "loss": 0.5418, "step": 42925 }, { "epoch": 2.91, "learning_rate": 3.0176086874552247e-06, "loss": 0.2792, "step": 42930 }, { "epoch": 2.91, "learning_rate": 3.0166660382338524e-06, "loss": 0.366, "step": 42935 }, { "epoch": 2.91, "learning_rate": 3.015723389012481e-06, "loss": 0.3892, "step": 42940 }, { "epoch": 2.91, "learning_rate": 3.0147807397911093e-06, "loss": 0.2967, "step": 42945 }, { "epoch": 2.92, "learning_rate": 3.0138380905697374e-06, "loss": 0.1754, "step": 42950 }, { "epoch": 2.92, "learning_rate": 3.012895441348366e-06, "loss": 0.7544, "step": 42955 }, { "epoch": 2.92, "learning_rate": 3.0119527921269944e-06, "loss": 0.5736, "step": 42960 }, { "epoch": 2.92, "learning_rate": 3.011010142905622e-06, "loss": 0.7034, "step": 42965 }, { "epoch": 2.92, "learning_rate": 3.0100674936842505e-06, "loss": 0.3718, "step": 42970 }, { "epoch": 2.92, "learning_rate": 3.0091248444628786e-06, "loss": 0.9012, "step": 42975 }, { "epoch": 2.92, "learning_rate": 3.008182195241507e-06, "loss": 0.4631, "step": 42980 }, { "epoch": 2.92, "learning_rate": 3.0072395460201355e-06, "loss": 0.4436, "step": 42985 }, { "epoch": 2.92, "learning_rate": 3.006296896798763e-06, "loss": 0.5724, "step": 42990 }, { "epoch": 2.92, "learning_rate": 3.0053542475773917e-06, "loss": 0.087, "step": 42995 }, { "epoch": 2.92, "learning_rate": 3.00441159835602e-06, "loss": 0.6014, "step": 43000 }, { "epoch": 2.92, "learning_rate": 3.003468949134648e-06, "loss": 0.4647, "step": 43005 }, { "epoch": 2.92, "learning_rate": 3.0025262999132767e-06, "loss": 0.2505, "step": 43010 }, { "epoch": 2.92, "learning_rate": 3.0015836506919048e-06, "loss": 0.0991, "step": 43015 }, { "epoch": 2.92, "learning_rate": 3.000641001470533e-06, "loss": 0.2384, "step": 43020 }, { "epoch": 2.92, "learning_rate": 2.9996983522491613e-06, "loss": 0.4031, "step": 43025 }, { "epoch": 2.92, "learning_rate": 2.99875570302779e-06, "loss": 0.2515, "step": 43030 }, { "epoch": 2.92, "learning_rate": 2.9978130538064174e-06, "loss": 0.2743, "step": 43035 }, { "epoch": 2.92, "learning_rate": 2.996870404585046e-06, "loss": 0.3394, "step": 43040 }, { "epoch": 2.92, "learning_rate": 2.9959277553636744e-06, "loss": 0.0849, "step": 43045 }, { "epoch": 2.92, "learning_rate": 2.9949851061423025e-06, "loss": 1.002, "step": 43050 }, { "epoch": 2.92, "learning_rate": 2.994042456920931e-06, "loss": 0.1407, "step": 43055 }, { "epoch": 2.92, "learning_rate": 2.9930998076995595e-06, "loss": 0.2865, "step": 43060 }, { "epoch": 2.92, "learning_rate": 2.992157158478187e-06, "loss": 0.1298, "step": 43065 }, { "epoch": 2.92, "learning_rate": 2.9912145092568156e-06, "loss": 0.5667, "step": 43070 }, { "epoch": 2.92, "learning_rate": 2.990271860035444e-06, "loss": 0.2904, "step": 43075 }, { "epoch": 2.92, "learning_rate": 2.989329210814072e-06, "loss": 0.3908, "step": 43080 }, { "epoch": 2.92, "learning_rate": 2.9883865615927006e-06, "loss": 0.2619, "step": 43085 }, { "epoch": 2.92, "learning_rate": 2.9874439123713283e-06, "loss": 0.2376, "step": 43090 }, { "epoch": 2.92, "learning_rate": 2.9865012631499567e-06, "loss": 0.7888, "step": 43095 }, { "epoch": 2.93, "learning_rate": 2.9855586139285852e-06, "loss": 0.3727, "step": 43100 }, { "epoch": 2.93, "learning_rate": 2.9846159647072133e-06, "loss": 0.4394, "step": 43105 }, { "epoch": 2.93, "learning_rate": 2.9836733154858418e-06, "loss": 0.2151, "step": 43110 }, { "epoch": 2.93, "learning_rate": 2.98273066626447e-06, "loss": 0.0776, "step": 43115 }, { "epoch": 2.93, "learning_rate": 2.981788017043098e-06, "loss": 0.2563, "step": 43120 }, { "epoch": 2.93, "learning_rate": 2.9808453678217264e-06, "loss": 0.1483, "step": 43125 }, { "epoch": 2.93, "learning_rate": 2.979902718600355e-06, "loss": 0.1704, "step": 43130 }, { "epoch": 2.93, "learning_rate": 2.978960069378983e-06, "loss": 0.357, "step": 43135 }, { "epoch": 2.93, "learning_rate": 2.978017420157611e-06, "loss": 0.3956, "step": 43140 }, { "epoch": 2.93, "learning_rate": 2.9770747709362395e-06, "loss": 0.2902, "step": 43145 }, { "epoch": 2.93, "learning_rate": 2.9761321217148676e-06, "loss": 0.2802, "step": 43150 }, { "epoch": 2.93, "learning_rate": 2.975189472493496e-06, "loss": 0.7189, "step": 43155 }, { "epoch": 2.93, "learning_rate": 2.9742468232721245e-06, "loss": 0.1308, "step": 43160 }, { "epoch": 2.93, "learning_rate": 2.973304174050752e-06, "loss": 0.3966, "step": 43165 }, { "epoch": 2.93, "learning_rate": 2.9723615248293807e-06, "loss": 0.7708, "step": 43170 }, { "epoch": 2.93, "learning_rate": 2.971418875608009e-06, "loss": 0.2094, "step": 43175 }, { "epoch": 2.93, "learning_rate": 2.9704762263866372e-06, "loss": 0.3405, "step": 43180 }, { "epoch": 2.93, "learning_rate": 2.9695335771652657e-06, "loss": 0.2964, "step": 43185 }, { "epoch": 2.93, "learning_rate": 2.968590927943894e-06, "loss": 0.0505, "step": 43190 }, { "epoch": 2.93, "learning_rate": 2.967648278722522e-06, "loss": 0.5986, "step": 43195 }, { "epoch": 2.93, "learning_rate": 2.9667056295011503e-06, "loss": 0.3953, "step": 43200 }, { "epoch": 2.93, "learning_rate": 2.9657629802797784e-06, "loss": 0.3347, "step": 43205 }, { "epoch": 2.93, "learning_rate": 2.964820331058407e-06, "loss": 0.2086, "step": 43210 }, { "epoch": 2.93, "learning_rate": 2.9638776818370354e-06, "loss": 0.1765, "step": 43215 }, { "epoch": 2.93, "learning_rate": 2.962935032615663e-06, "loss": 0.4609, "step": 43220 }, { "epoch": 2.93, "learning_rate": 2.9619923833942915e-06, "loss": 0.4638, "step": 43225 }, { "epoch": 2.93, "learning_rate": 2.96104973417292e-06, "loss": 0.0766, "step": 43230 }, { "epoch": 2.93, "learning_rate": 2.960107084951548e-06, "loss": 0.1981, "step": 43235 }, { "epoch": 2.93, "learning_rate": 2.959164435730176e-06, "loss": 0.1467, "step": 43240 }, { "epoch": 2.94, "learning_rate": 2.9582217865088046e-06, "loss": 0.263, "step": 43245 }, { "epoch": 2.94, "learning_rate": 2.9572791372874327e-06, "loss": 0.5826, "step": 43250 }, { "epoch": 2.94, "learning_rate": 2.956336488066061e-06, "loss": 0.2061, "step": 43255 }, { "epoch": 2.94, "learning_rate": 2.9553938388446896e-06, "loss": 0.4735, "step": 43260 }, { "epoch": 2.94, "learning_rate": 2.9544511896233173e-06, "loss": 0.117, "step": 43265 }, { "epoch": 2.94, "learning_rate": 2.9535085404019458e-06, "loss": 0.1467, "step": 43270 }, { "epoch": 2.94, "learning_rate": 2.9525658911805742e-06, "loss": 0.4769, "step": 43275 }, { "epoch": 2.94, "learning_rate": 2.9516232419592023e-06, "loss": 0.174, "step": 43280 }, { "epoch": 2.94, "learning_rate": 2.950680592737831e-06, "loss": 0.0916, "step": 43285 }, { "epoch": 2.94, "learning_rate": 2.9497379435164593e-06, "loss": 0.2001, "step": 43290 }, { "epoch": 2.94, "learning_rate": 2.948795294295087e-06, "loss": 0.202, "step": 43295 }, { "epoch": 2.94, "learning_rate": 2.9478526450737154e-06, "loss": 0.1967, "step": 43300 }, { "epoch": 2.94, "learning_rate": 2.946909995852344e-06, "loss": 0.2636, "step": 43305 }, { "epoch": 2.94, "learning_rate": 2.945967346630972e-06, "loss": 0.43, "step": 43310 }, { "epoch": 2.94, "learning_rate": 2.9450246974096004e-06, "loss": 0.7955, "step": 43315 }, { "epoch": 2.94, "learning_rate": 2.944082048188228e-06, "loss": 0.2182, "step": 43320 }, { "epoch": 2.94, "learning_rate": 2.9431393989668566e-06, "loss": 0.5183, "step": 43325 }, { "epoch": 2.94, "learning_rate": 2.942196749745485e-06, "loss": 0.2139, "step": 43330 }, { "epoch": 2.94, "learning_rate": 2.941254100524113e-06, "loss": 0.3259, "step": 43335 }, { "epoch": 2.94, "learning_rate": 2.9403114513027416e-06, "loss": 0.6379, "step": 43340 }, { "epoch": 2.94, "learning_rate": 2.9393688020813697e-06, "loss": 0.7372, "step": 43345 }, { "epoch": 2.94, "learning_rate": 2.9384261528599977e-06, "loss": 0.4211, "step": 43350 }, { "epoch": 2.94, "learning_rate": 2.9374835036386262e-06, "loss": 0.4075, "step": 43355 }, { "epoch": 2.94, "learning_rate": 2.9365408544172547e-06, "loss": 0.4562, "step": 43360 }, { "epoch": 2.94, "learning_rate": 2.9355982051958828e-06, "loss": 0.3917, "step": 43365 }, { "epoch": 2.94, "learning_rate": 2.934655555974511e-06, "loss": 0.8, "step": 43370 }, { "epoch": 2.94, "learning_rate": 2.9337129067531393e-06, "loss": 0.3164, "step": 43375 }, { "epoch": 2.94, "learning_rate": 2.9327702575317674e-06, "loss": 0.7172, "step": 43380 }, { "epoch": 2.94, "learning_rate": 2.931827608310396e-06, "loss": 0.3014, "step": 43385 }, { "epoch": 2.94, "learning_rate": 2.9308849590890244e-06, "loss": 0.3141, "step": 43390 }, { "epoch": 2.95, "learning_rate": 2.929942309867652e-06, "loss": 0.3142, "step": 43395 }, { "epoch": 2.95, "learning_rate": 2.9289996606462805e-06, "loss": 0.2821, "step": 43400 }, { "epoch": 2.95, "learning_rate": 2.928057011424909e-06, "loss": 0.8478, "step": 43405 }, { "epoch": 2.95, "learning_rate": 2.927114362203537e-06, "loss": 0.5656, "step": 43410 }, { "epoch": 2.95, "learning_rate": 2.9261717129821655e-06, "loss": 0.3772, "step": 43415 }, { "epoch": 2.95, "learning_rate": 2.925229063760793e-06, "loss": 0.198, "step": 43420 }, { "epoch": 2.95, "learning_rate": 2.9242864145394217e-06, "loss": 0.4162, "step": 43425 }, { "epoch": 2.95, "learning_rate": 2.92334376531805e-06, "loss": 0.3198, "step": 43430 }, { "epoch": 2.95, "learning_rate": 2.922401116096678e-06, "loss": 0.0813, "step": 43435 }, { "epoch": 2.95, "learning_rate": 2.9214584668753067e-06, "loss": 0.2776, "step": 43440 }, { "epoch": 2.95, "learning_rate": 2.920515817653935e-06, "loss": 0.7292, "step": 43445 }, { "epoch": 2.95, "learning_rate": 2.919573168432563e-06, "loss": 0.4953, "step": 43450 }, { "epoch": 2.95, "learning_rate": 2.9186305192111913e-06, "loss": 0.3857, "step": 43455 }, { "epoch": 2.95, "learning_rate": 2.91768786998982e-06, "loss": 0.6215, "step": 43460 }, { "epoch": 2.95, "learning_rate": 2.916745220768448e-06, "loss": 0.4388, "step": 43465 }, { "epoch": 2.95, "learning_rate": 2.915802571547076e-06, "loss": 0.2113, "step": 43470 }, { "epoch": 2.95, "learning_rate": 2.9148599223257044e-06, "loss": 0.1958, "step": 43475 }, { "epoch": 2.95, "learning_rate": 2.9139172731043325e-06, "loss": 0.4731, "step": 43480 }, { "epoch": 2.95, "learning_rate": 2.912974623882961e-06, "loss": 0.3073, "step": 43485 }, { "epoch": 2.95, "learning_rate": 2.9120319746615895e-06, "loss": 0.6752, "step": 43490 }, { "epoch": 2.95, "learning_rate": 2.911089325440217e-06, "loss": 0.595, "step": 43495 }, { "epoch": 2.95, "learning_rate": 2.9101466762188456e-06, "loss": 0.3138, "step": 43500 }, { "epoch": 2.95, "learning_rate": 2.909204026997474e-06, "loss": 0.4176, "step": 43505 }, { "epoch": 2.95, "learning_rate": 2.908261377776102e-06, "loss": 0.2354, "step": 43510 }, { "epoch": 2.95, "learning_rate": 2.9073187285547306e-06, "loss": 0.1147, "step": 43515 }, { "epoch": 2.95, "learning_rate": 2.906376079333359e-06, "loss": 0.2138, "step": 43520 }, { "epoch": 2.95, "learning_rate": 2.9054334301119867e-06, "loss": 0.2564, "step": 43525 }, { "epoch": 2.95, "learning_rate": 2.9044907808906152e-06, "loss": 0.4879, "step": 43530 }, { "epoch": 2.95, "learning_rate": 2.9035481316692433e-06, "loss": 0.2632, "step": 43535 }, { "epoch": 2.96, "learning_rate": 2.9026054824478718e-06, "loss": 0.4417, "step": 43540 }, { "epoch": 2.96, "learning_rate": 2.9016628332265003e-06, "loss": 0.3048, "step": 43545 }, { "epoch": 2.96, "learning_rate": 2.900720184005128e-06, "loss": 0.3258, "step": 43550 }, { "epoch": 2.96, "learning_rate": 2.8997775347837564e-06, "loss": 0.3268, "step": 43555 }, { "epoch": 2.96, "learning_rate": 2.898834885562385e-06, "loss": 0.3078, "step": 43560 }, { "epoch": 2.96, "learning_rate": 2.897892236341013e-06, "loss": 0.3533, "step": 43565 }, { "epoch": 2.96, "learning_rate": 2.8969495871196414e-06, "loss": 0.3594, "step": 43570 }, { "epoch": 2.96, "learning_rate": 2.8960069378982695e-06, "loss": 0.2213, "step": 43575 }, { "epoch": 2.96, "learning_rate": 2.8950642886768976e-06, "loss": 0.6822, "step": 43580 }, { "epoch": 2.96, "learning_rate": 2.894121639455526e-06, "loss": 0.4805, "step": 43585 }, { "epoch": 2.96, "learning_rate": 2.8931789902341545e-06, "loss": 0.4587, "step": 43590 }, { "epoch": 2.96, "learning_rate": 2.892236341012782e-06, "loss": 0.1053, "step": 43595 }, { "epoch": 2.96, "learning_rate": 2.8912936917914107e-06, "loss": 0.4565, "step": 43600 }, { "epoch": 2.96, "learning_rate": 2.890351042570039e-06, "loss": 0.0704, "step": 43605 }, { "epoch": 2.96, "learning_rate": 2.8894083933486672e-06, "loss": 0.3888, "step": 43610 }, { "epoch": 2.96, "learning_rate": 2.8884657441272957e-06, "loss": 0.506, "step": 43615 }, { "epoch": 2.96, "learning_rate": 2.887523094905924e-06, "loss": 0.1148, "step": 43620 }, { "epoch": 2.96, "learning_rate": 2.886580445684552e-06, "loss": 0.7856, "step": 43625 }, { "epoch": 2.96, "learning_rate": 2.8856377964631803e-06, "loss": 0.6157, "step": 43630 }, { "epoch": 2.96, "learning_rate": 2.884695147241809e-06, "loss": 0.3524, "step": 43635 }, { "epoch": 2.96, "learning_rate": 2.883752498020437e-06, "loss": 0.3627, "step": 43640 }, { "epoch": 2.96, "learning_rate": 2.8828098487990654e-06, "loss": 0.2634, "step": 43645 }, { "epoch": 2.96, "learning_rate": 2.881867199577693e-06, "loss": 0.2396, "step": 43650 }, { "epoch": 2.96, "learning_rate": 2.8809245503563215e-06, "loss": 0.1438, "step": 43655 }, { "epoch": 2.96, "learning_rate": 2.87998190113495e-06, "loss": 0.2962, "step": 43660 }, { "epoch": 2.96, "learning_rate": 2.879039251913578e-06, "loss": 0.0863, "step": 43665 }, { "epoch": 2.96, "learning_rate": 2.8780966026922065e-06, "loss": 0.7328, "step": 43670 }, { "epoch": 2.96, "learning_rate": 2.877153953470835e-06, "loss": 0.4879, "step": 43675 }, { "epoch": 2.96, "learning_rate": 2.8762113042494627e-06, "loss": 0.2142, "step": 43680 }, { "epoch": 2.96, "learning_rate": 2.875268655028091e-06, "loss": 0.2483, "step": 43685 }, { "epoch": 2.97, "learning_rate": 2.8743260058067196e-06, "loss": 0.5407, "step": 43690 }, { "epoch": 2.97, "learning_rate": 2.8733833565853477e-06, "loss": 0.2513, "step": 43695 }, { "epoch": 2.97, "learning_rate": 2.8724407073639758e-06, "loss": 0.3981, "step": 43700 }, { "epoch": 2.97, "learning_rate": 2.8714980581426042e-06, "loss": 0.1889, "step": 43705 }, { "epoch": 2.97, "learning_rate": 2.8705554089212323e-06, "loss": 0.1249, "step": 43710 }, { "epoch": 2.97, "learning_rate": 2.869612759699861e-06, "loss": 0.2879, "step": 43715 }, { "epoch": 2.97, "learning_rate": 2.8686701104784893e-06, "loss": 0.4152, "step": 43720 }, { "epoch": 2.97, "learning_rate": 2.867727461257117e-06, "loss": 0.2431, "step": 43725 }, { "epoch": 2.97, "learning_rate": 2.8667848120357454e-06, "loss": 0.2719, "step": 43730 }, { "epoch": 2.97, "learning_rate": 2.865842162814374e-06, "loss": 0.3156, "step": 43735 }, { "epoch": 2.97, "learning_rate": 2.864899513593002e-06, "loss": 0.4735, "step": 43740 }, { "epoch": 2.97, "learning_rate": 2.8639568643716304e-06, "loss": 0.1485, "step": 43745 }, { "epoch": 2.97, "learning_rate": 2.863014215150259e-06, "loss": 0.3237, "step": 43750 }, { "epoch": 2.97, "learning_rate": 2.8620715659288866e-06, "loss": 0.2746, "step": 43755 }, { "epoch": 2.97, "learning_rate": 2.861128916707515e-06, "loss": 0.409, "step": 43760 }, { "epoch": 2.97, "learning_rate": 2.860186267486143e-06, "loss": 0.3215, "step": 43765 }, { "epoch": 2.97, "learning_rate": 2.8592436182647716e-06, "loss": 0.1573, "step": 43770 }, { "epoch": 2.97, "learning_rate": 2.8583009690434e-06, "loss": 0.3516, "step": 43775 }, { "epoch": 2.97, "learning_rate": 2.8573583198220277e-06, "loss": 0.1262, "step": 43780 }, { "epoch": 2.97, "learning_rate": 2.8564156706006562e-06, "loss": 0.4993, "step": 43785 }, { "epoch": 2.97, "learning_rate": 2.8554730213792847e-06, "loss": 0.2251, "step": 43790 }, { "epoch": 2.97, "learning_rate": 2.8545303721579128e-06, "loss": 0.1242, "step": 43795 }, { "epoch": 2.97, "learning_rate": 2.8535877229365413e-06, "loss": 0.2622, "step": 43800 }, { "epoch": 2.97, "learning_rate": 2.8526450737151693e-06, "loss": 0.4469, "step": 43805 }, { "epoch": 2.97, "learning_rate": 2.8517024244937974e-06, "loss": 0.8787, "step": 43810 }, { "epoch": 2.97, "learning_rate": 2.850759775272426e-06, "loss": 0.1562, "step": 43815 }, { "epoch": 2.97, "learning_rate": 2.8498171260510544e-06, "loss": 0.1853, "step": 43820 }, { "epoch": 2.97, "learning_rate": 2.848874476829682e-06, "loss": 0.4571, "step": 43825 }, { "epoch": 2.97, "learning_rate": 2.8479318276083105e-06, "loss": 0.4289, "step": 43830 }, { "epoch": 2.98, "learning_rate": 2.846989178386939e-06, "loss": 0.2397, "step": 43835 }, { "epoch": 2.98, "learning_rate": 2.846046529165567e-06, "loss": 0.365, "step": 43840 }, { "epoch": 2.98, "learning_rate": 2.8451038799441955e-06, "loss": 0.4299, "step": 43845 }, { "epoch": 2.98, "learning_rate": 2.844161230722824e-06, "loss": 0.3092, "step": 43850 }, { "epoch": 2.98, "learning_rate": 2.8432185815014517e-06, "loss": 0.2993, "step": 43855 }, { "epoch": 2.98, "learning_rate": 2.84227593228008e-06, "loss": 0.283, "step": 43860 }, { "epoch": 2.98, "learning_rate": 2.8413332830587086e-06, "loss": 0.3564, "step": 43865 }, { "epoch": 2.98, "learning_rate": 2.8403906338373367e-06, "loss": 0.2139, "step": 43870 }, { "epoch": 2.98, "learning_rate": 2.839447984615965e-06, "loss": 0.1677, "step": 43875 }, { "epoch": 2.98, "learning_rate": 2.838505335394593e-06, "loss": 0.0694, "step": 43880 }, { "epoch": 2.98, "learning_rate": 2.8375626861732213e-06, "loss": 0.404, "step": 43885 }, { "epoch": 2.98, "learning_rate": 2.83662003695185e-06, "loss": 0.3269, "step": 43890 }, { "epoch": 2.98, "learning_rate": 2.835677387730478e-06, "loss": 0.4929, "step": 43895 }, { "epoch": 2.98, "learning_rate": 2.8347347385091064e-06, "loss": 0.3326, "step": 43900 }, { "epoch": 2.98, "learning_rate": 2.833792089287735e-06, "loss": 0.3191, "step": 43905 }, { "epoch": 2.98, "learning_rate": 2.8328494400663625e-06, "loss": 0.451, "step": 43910 }, { "epoch": 2.98, "learning_rate": 2.831906790844991e-06, "loss": 0.4283, "step": 43915 }, { "epoch": 2.98, "learning_rate": 2.8309641416236195e-06, "loss": 0.3646, "step": 43920 }, { "epoch": 2.98, "learning_rate": 2.8300214924022475e-06, "loss": 0.3928, "step": 43925 }, { "epoch": 2.98, "learning_rate": 2.8290788431808756e-06, "loss": 0.0762, "step": 43930 }, { "epoch": 2.98, "learning_rate": 2.828136193959504e-06, "loss": 0.2375, "step": 43935 }, { "epoch": 2.98, "learning_rate": 2.827193544738132e-06, "loss": 0.2683, "step": 43940 }, { "epoch": 2.98, "learning_rate": 2.8262508955167606e-06, "loss": 0.226, "step": 43945 }, { "epoch": 2.98, "learning_rate": 2.825308246295389e-06, "loss": 0.5644, "step": 43950 }, { "epoch": 2.98, "learning_rate": 2.8243655970740168e-06, "loss": 0.1224, "step": 43955 }, { "epoch": 2.98, "learning_rate": 2.8234229478526452e-06, "loss": 0.4153, "step": 43960 }, { "epoch": 2.98, "learning_rate": 2.8224802986312737e-06, "loss": 0.3225, "step": 43965 }, { "epoch": 2.98, "learning_rate": 2.821537649409902e-06, "loss": 0.3288, "step": 43970 }, { "epoch": 2.98, "learning_rate": 2.8205950001885303e-06, "loss": 0.3334, "step": 43975 }, { "epoch": 2.98, "learning_rate": 2.819652350967158e-06, "loss": 0.4765, "step": 43980 }, { "epoch": 2.99, "learning_rate": 2.8187097017457864e-06, "loss": 0.1478, "step": 43985 }, { "epoch": 2.99, "learning_rate": 2.817767052524415e-06, "loss": 0.1983, "step": 43990 }, { "epoch": 2.99, "learning_rate": 2.816824403303043e-06, "loss": 0.4342, "step": 43995 }, { "epoch": 2.99, "learning_rate": 2.8158817540816714e-06, "loss": 0.7268, "step": 44000 }, { "epoch": 2.99, "learning_rate": 2.8149391048603e-06, "loss": 0.2012, "step": 44005 }, { "epoch": 2.99, "learning_rate": 2.8139964556389276e-06, "loss": 0.2465, "step": 44010 }, { "epoch": 2.99, "learning_rate": 2.813053806417556e-06, "loss": 0.2474, "step": 44015 }, { "epoch": 2.99, "learning_rate": 2.8121111571961845e-06, "loss": 0.2688, "step": 44020 }, { "epoch": 2.99, "learning_rate": 2.8111685079748126e-06, "loss": 0.2021, "step": 44025 }, { "epoch": 2.99, "learning_rate": 2.810225858753441e-06, "loss": 0.6286, "step": 44030 }, { "epoch": 2.99, "learning_rate": 2.809283209532069e-06, "loss": 0.3136, "step": 44035 }, { "epoch": 2.99, "learning_rate": 2.8083405603106972e-06, "loss": 0.3415, "step": 44040 }, { "epoch": 2.99, "learning_rate": 2.8073979110893257e-06, "loss": 0.3477, "step": 44045 }, { "epoch": 2.99, "learning_rate": 2.806455261867954e-06, "loss": 0.2076, "step": 44050 }, { "epoch": 2.99, "learning_rate": 2.805512612646582e-06, "loss": 0.4156, "step": 44055 }, { "epoch": 2.99, "learning_rate": 2.8045699634252103e-06, "loss": 0.092, "step": 44060 }, { "epoch": 2.99, "learning_rate": 2.803627314203839e-06, "loss": 0.2413, "step": 44065 }, { "epoch": 2.99, "learning_rate": 2.802684664982467e-06, "loss": 0.268, "step": 44070 }, { "epoch": 2.99, "learning_rate": 2.8017420157610954e-06, "loss": 0.3546, "step": 44075 }, { "epoch": 2.99, "learning_rate": 2.800799366539724e-06, "loss": 1.1391, "step": 44080 }, { "epoch": 2.99, "learning_rate": 2.7998567173183515e-06, "loss": 0.3103, "step": 44085 }, { "epoch": 2.99, "learning_rate": 2.79891406809698e-06, "loss": 0.1014, "step": 44090 }, { "epoch": 2.99, "learning_rate": 2.797971418875608e-06, "loss": 0.2956, "step": 44095 }, { "epoch": 2.99, "learning_rate": 2.7970287696542365e-06, "loss": 0.3236, "step": 44100 }, { "epoch": 2.99, "learning_rate": 2.796086120432865e-06, "loss": 0.8308, "step": 44105 }, { "epoch": 2.99, "learning_rate": 2.7951434712114927e-06, "loss": 0.3037, "step": 44110 }, { "epoch": 2.99, "learning_rate": 2.794200821990121e-06, "loss": 0.3608, "step": 44115 }, { "epoch": 2.99, "learning_rate": 2.7932581727687496e-06, "loss": 0.3378, "step": 44120 }, { "epoch": 2.99, "learning_rate": 2.7923155235473777e-06, "loss": 0.4344, "step": 44125 }, { "epoch": 3.0, "learning_rate": 2.791372874326006e-06, "loss": 0.0788, "step": 44130 }, { "epoch": 3.0, "learning_rate": 2.7904302251046347e-06, "loss": 0.4996, "step": 44135 }, { "epoch": 3.0, "learning_rate": 2.7894875758832623e-06, "loss": 0.6367, "step": 44140 }, { "epoch": 3.0, "learning_rate": 2.788544926661891e-06, "loss": 0.1372, "step": 44145 }, { "epoch": 3.0, "learning_rate": 2.7876022774405193e-06, "loss": 0.2572, "step": 44150 }, { "epoch": 3.0, "learning_rate": 2.7866596282191473e-06, "loss": 0.3698, "step": 44155 }, { "epoch": 3.0, "learning_rate": 2.7857169789977754e-06, "loss": 0.4932, "step": 44160 }, { "epoch": 3.0, "learning_rate": 2.784774329776404e-06, "loss": 0.5183, "step": 44165 }, { "epoch": 3.0, "learning_rate": 2.783831680555032e-06, "loss": 0.3283, "step": 44170 }, { "epoch": 3.0, "learning_rate": 2.7828890313336605e-06, "loss": 0.4469, "step": 44175 }, { "epoch": 3.0, "learning_rate": 2.781946382112289e-06, "loss": 0.469, "step": 44180 }, { "epoch": 3.0, "learning_rate": 2.7810037328909166e-06, "loss": 0.3633, "step": 44185 }, { "epoch": 3.0, "learning_rate": 2.780061083669545e-06, "loss": 0.1606, "step": 44190 }, { "epoch": 3.0, "learning_rate": 2.7791184344481736e-06, "loss": 0.2958, "step": 44195 }, { "epoch": 3.0, "learning_rate": 2.7781757852268016e-06, "loss": 0.7895, "step": 44200 }, { "epoch": 3.0, "learning_rate": 2.77723313600543e-06, "loss": 0.4639, "step": 44205 }, { "epoch": 3.0, "learning_rate": 2.7762904867840577e-06, "loss": 0.1837, "step": 44210 }, { "epoch": 3.0, "learning_rate": 2.7753478375626862e-06, "loss": 0.2639, "step": 44215 }, { "epoch": 3.0, "learning_rate": 2.7744051883413147e-06, "loss": 0.9869, "step": 44220 }, { "epoch": 3.0, "learning_rate": 2.7734625391199428e-06, "loss": 0.4077, "step": 44225 }, { "epoch": 3.0, "learning_rate": 2.7725198898985713e-06, "loss": 0.3521, "step": 44230 }, { "epoch": 3.0, "learning_rate": 2.7715772406771998e-06, "loss": 0.5713, "step": 44235 }, { "epoch": 3.0, "learning_rate": 2.7706345914558274e-06, "loss": 0.568, "step": 44240 }, { "epoch": 3.0, "learning_rate": 2.769691942234456e-06, "loss": 0.2053, "step": 44245 }, { "epoch": 3.0, "learning_rate": 2.7687492930130844e-06, "loss": 0.1397, "step": 44250 }, { "epoch": 3.0, "learning_rate": 2.7678066437917124e-06, "loss": 0.6295, "step": 44255 }, { "epoch": 3.0, "learning_rate": 2.766863994570341e-06, "loss": 0.2096, "step": 44260 }, { "epoch": 3.0, "learning_rate": 2.765921345348969e-06, "loss": 0.1987, "step": 44265 }, { "epoch": 3.0, "learning_rate": 2.764978696127597e-06, "loss": 0.1754, "step": 44270 }, { "epoch": 3.0, "learning_rate": 2.7640360469062255e-06, "loss": 0.1999, "step": 44275 }, { "epoch": 3.01, "learning_rate": 2.763093397684854e-06, "loss": 0.6859, "step": 44280 }, { "epoch": 3.01, "learning_rate": 2.7621507484634817e-06, "loss": 0.1207, "step": 44285 }, { "epoch": 3.01, "learning_rate": 2.76120809924211e-06, "loss": 0.3178, "step": 44290 }, { "epoch": 3.01, "learning_rate": 2.7602654500207386e-06, "loss": 0.1851, "step": 44295 }, { "epoch": 3.01, "learning_rate": 2.7593228007993667e-06, "loss": 0.1554, "step": 44300 }, { "epoch": 3.01, "learning_rate": 2.758380151577995e-06, "loss": 0.1815, "step": 44305 }, { "epoch": 3.01, "learning_rate": 2.7574375023566237e-06, "loss": 0.2206, "step": 44310 }, { "epoch": 3.01, "learning_rate": 2.7564948531352513e-06, "loss": 0.1355, "step": 44315 }, { "epoch": 3.01, "learning_rate": 2.75555220391388e-06, "loss": 0.243, "step": 44320 }, { "epoch": 3.01, "learning_rate": 2.754609554692508e-06, "loss": 0.1762, "step": 44325 }, { "epoch": 3.01, "learning_rate": 2.7536669054711364e-06, "loss": 0.5259, "step": 44330 }, { "epoch": 3.01, "learning_rate": 2.752724256249765e-06, "loss": 0.3282, "step": 44335 }, { "epoch": 3.01, "learning_rate": 2.7517816070283925e-06, "loss": 0.4663, "step": 44340 }, { "epoch": 3.01, "learning_rate": 2.750838957807021e-06, "loss": 0.9455, "step": 44345 }, { "epoch": 3.01, "learning_rate": 2.7498963085856495e-06, "loss": 0.6423, "step": 44350 }, { "epoch": 3.01, "learning_rate": 2.7489536593642775e-06, "loss": 0.4395, "step": 44355 }, { "epoch": 3.01, "learning_rate": 2.748011010142906e-06, "loss": 0.3378, "step": 44360 }, { "epoch": 3.01, "learning_rate": 2.747068360921534e-06, "loss": 0.1452, "step": 44365 }, { "epoch": 3.01, "learning_rate": 2.746125711700162e-06, "loss": 0.1708, "step": 44370 }, { "epoch": 3.01, "learning_rate": 2.7451830624787906e-06, "loss": 0.1848, "step": 44375 }, { "epoch": 3.01, "learning_rate": 2.744240413257419e-06, "loss": 0.0388, "step": 44380 }, { "epoch": 3.01, "learning_rate": 2.743297764036047e-06, "loss": 0.2787, "step": 44385 }, { "epoch": 3.01, "learning_rate": 2.7423551148146752e-06, "loss": 0.1378, "step": 44390 }, { "epoch": 3.01, "learning_rate": 2.7414124655933037e-06, "loss": 0.2775, "step": 44395 }, { "epoch": 3.01, "learning_rate": 2.740469816371932e-06, "loss": 0.207, "step": 44400 }, { "epoch": 3.01, "learning_rate": 2.7395271671505603e-06, "loss": 0.1868, "step": 44405 }, { "epoch": 3.01, "learning_rate": 2.7385845179291888e-06, "loss": 0.0824, "step": 44410 }, { "epoch": 3.01, "learning_rate": 2.7376418687078164e-06, "loss": 0.2541, "step": 44415 }, { "epoch": 3.01, "learning_rate": 2.736699219486445e-06, "loss": 0.1642, "step": 44420 }, { "epoch": 3.02, "learning_rate": 2.735756570265073e-06, "loss": 0.2042, "step": 44425 }, { "epoch": 3.02, "learning_rate": 2.7348139210437014e-06, "loss": 0.0849, "step": 44430 }, { "epoch": 3.02, "learning_rate": 2.73387127182233e-06, "loss": 0.4309, "step": 44435 }, { "epoch": 3.02, "learning_rate": 2.7329286226009576e-06, "loss": 0.2397, "step": 44440 }, { "epoch": 3.02, "learning_rate": 2.731985973379586e-06, "loss": 0.1182, "step": 44445 }, { "epoch": 3.02, "learning_rate": 2.7310433241582145e-06, "loss": 0.3511, "step": 44450 }, { "epoch": 3.02, "learning_rate": 2.7301006749368426e-06, "loss": 0.5588, "step": 44455 }, { "epoch": 3.02, "learning_rate": 2.729158025715471e-06, "loss": 0.0995, "step": 44460 }, { "epoch": 3.02, "learning_rate": 2.7282153764940996e-06, "loss": 0.5296, "step": 44465 }, { "epoch": 3.02, "learning_rate": 2.7272727272727272e-06, "loss": 0.1435, "step": 44470 }, { "epoch": 3.02, "learning_rate": 2.7263300780513557e-06, "loss": 0.2204, "step": 44475 }, { "epoch": 3.02, "learning_rate": 2.725387428829984e-06, "loss": 0.1239, "step": 44480 }, { "epoch": 3.02, "learning_rate": 2.7244447796086123e-06, "loss": 0.0727, "step": 44485 }, { "epoch": 3.02, "learning_rate": 2.7235021303872403e-06, "loss": 0.2453, "step": 44490 }, { "epoch": 3.02, "learning_rate": 2.722559481165869e-06, "loss": 0.2281, "step": 44495 }, { "epoch": 3.02, "learning_rate": 2.721616831944497e-06, "loss": 0.164, "step": 44500 }, { "epoch": 3.02, "learning_rate": 2.7206741827231254e-06, "loss": 0.5298, "step": 44505 }, { "epoch": 3.02, "learning_rate": 2.719731533501754e-06, "loss": 0.0246, "step": 44510 }, { "epoch": 3.02, "learning_rate": 2.7187888842803815e-06, "loss": 0.3689, "step": 44515 }, { "epoch": 3.02, "learning_rate": 2.71784623505901e-06, "loss": 0.309, "step": 44520 }, { "epoch": 3.02, "learning_rate": 2.7169035858376385e-06, "loss": 0.353, "step": 44525 }, { "epoch": 3.02, "learning_rate": 2.7159609366162665e-06, "loss": 0.0677, "step": 44530 }, { "epoch": 3.02, "learning_rate": 2.715018287394895e-06, "loss": 0.488, "step": 44535 }, { "epoch": 3.02, "learning_rate": 2.7140756381735227e-06, "loss": 0.0847, "step": 44540 }, { "epoch": 3.02, "learning_rate": 2.713132988952151e-06, "loss": 0.3621, "step": 44545 }, { "epoch": 3.02, "learning_rate": 2.7121903397307796e-06, "loss": 0.1996, "step": 44550 }, { "epoch": 3.02, "learning_rate": 2.7112476905094077e-06, "loss": 0.2441, "step": 44555 }, { "epoch": 3.02, "learning_rate": 2.710305041288036e-06, "loss": 0.0617, "step": 44560 }, { "epoch": 3.02, "learning_rate": 2.7093623920666647e-06, "loss": 0.4589, "step": 44565 }, { "epoch": 3.02, "learning_rate": 2.7084197428452923e-06, "loss": 0.0981, "step": 44570 }, { "epoch": 3.03, "learning_rate": 2.707477093623921e-06, "loss": 0.1716, "step": 44575 }, { "epoch": 3.03, "learning_rate": 2.7065344444025493e-06, "loss": 0.233, "step": 44580 }, { "epoch": 3.03, "learning_rate": 2.7055917951811774e-06, "loss": 0.5328, "step": 44585 }, { "epoch": 3.03, "learning_rate": 2.704649145959806e-06, "loss": 0.0707, "step": 44590 }, { "epoch": 3.03, "learning_rate": 2.703706496738434e-06, "loss": 0.4105, "step": 44595 }, { "epoch": 3.03, "learning_rate": 2.702763847517062e-06, "loss": 0.1395, "step": 44600 }, { "epoch": 3.03, "learning_rate": 2.7018211982956905e-06, "loss": 0.3535, "step": 44605 }, { "epoch": 3.03, "learning_rate": 2.700878549074319e-06, "loss": 0.0667, "step": 44610 }, { "epoch": 3.03, "learning_rate": 2.699935899852947e-06, "loss": 0.4927, "step": 44615 }, { "epoch": 3.03, "learning_rate": 2.698993250631575e-06, "loss": 0.2705, "step": 44620 }, { "epoch": 3.03, "learning_rate": 2.6980506014102036e-06, "loss": 0.525, "step": 44625 }, { "epoch": 3.03, "learning_rate": 2.6971079521888316e-06, "loss": 0.4497, "step": 44630 }, { "epoch": 3.03, "learning_rate": 2.69616530296746e-06, "loss": 0.3484, "step": 44635 }, { "epoch": 3.03, "learning_rate": 2.6952226537460886e-06, "loss": 0.5191, "step": 44640 }, { "epoch": 3.03, "learning_rate": 2.6942800045247162e-06, "loss": 0.1324, "step": 44645 }, { "epoch": 3.03, "learning_rate": 2.6933373553033447e-06, "loss": 0.1445, "step": 44650 }, { "epoch": 3.03, "learning_rate": 2.6923947060819728e-06, "loss": 0.0853, "step": 44655 }, { "epoch": 3.03, "learning_rate": 2.6914520568606013e-06, "loss": 0.1999, "step": 44660 }, { "epoch": 3.03, "learning_rate": 2.6905094076392298e-06, "loss": 0.2964, "step": 44665 }, { "epoch": 3.03, "learning_rate": 2.6895667584178574e-06, "loss": 0.0991, "step": 44670 }, { "epoch": 3.03, "learning_rate": 2.688624109196486e-06, "loss": 0.3881, "step": 44675 }, { "epoch": 3.03, "learning_rate": 2.6876814599751144e-06, "loss": 0.3978, "step": 44680 }, { "epoch": 3.03, "learning_rate": 2.6867388107537424e-06, "loss": 0.6139, "step": 44685 }, { "epoch": 3.03, "learning_rate": 2.685796161532371e-06, "loss": 0.095, "step": 44690 }, { "epoch": 3.03, "learning_rate": 2.6848535123109994e-06, "loss": 0.2744, "step": 44695 }, { "epoch": 3.03, "learning_rate": 2.683910863089627e-06, "loss": 0.5536, "step": 44700 }, { "epoch": 3.03, "learning_rate": 2.6829682138682555e-06, "loss": 0.409, "step": 44705 }, { "epoch": 3.03, "learning_rate": 2.682025564646884e-06, "loss": 0.3467, "step": 44710 }, { "epoch": 3.03, "learning_rate": 2.681082915425512e-06, "loss": 0.126, "step": 44715 }, { "epoch": 3.04, "learning_rate": 2.68014026620414e-06, "loss": 0.1198, "step": 44720 }, { "epoch": 3.04, "learning_rate": 2.6791976169827686e-06, "loss": 0.2898, "step": 44725 }, { "epoch": 3.04, "learning_rate": 2.6782549677613967e-06, "loss": 0.4528, "step": 44730 }, { "epoch": 3.04, "learning_rate": 2.677312318540025e-06, "loss": 0.1331, "step": 44735 }, { "epoch": 3.04, "learning_rate": 2.6763696693186537e-06, "loss": 0.3187, "step": 44740 }, { "epoch": 3.04, "learning_rate": 2.6754270200972813e-06, "loss": 0.4427, "step": 44745 }, { "epoch": 3.04, "learning_rate": 2.67448437087591e-06, "loss": 0.35, "step": 44750 }, { "epoch": 3.04, "learning_rate": 2.6735417216545383e-06, "loss": 0.4158, "step": 44755 }, { "epoch": 3.04, "learning_rate": 2.6725990724331664e-06, "loss": 0.3313, "step": 44760 }, { "epoch": 3.04, "learning_rate": 2.671656423211795e-06, "loss": 0.3442, "step": 44765 }, { "epoch": 3.04, "learning_rate": 2.6707137739904225e-06, "loss": 0.1218, "step": 44770 }, { "epoch": 3.04, "learning_rate": 2.669771124769051e-06, "loss": 0.2592, "step": 44775 }, { "epoch": 3.04, "learning_rate": 2.6688284755476795e-06, "loss": 0.3101, "step": 44780 }, { "epoch": 3.04, "learning_rate": 2.6678858263263075e-06, "loss": 0.1545, "step": 44785 }, { "epoch": 3.04, "learning_rate": 2.666943177104936e-06, "loss": 0.1437, "step": 44790 }, { "epoch": 3.04, "learning_rate": 2.6660005278835645e-06, "loss": 0.2044, "step": 44795 }, { "epoch": 3.04, "learning_rate": 2.665057878662192e-06, "loss": 0.2486, "step": 44800 }, { "epoch": 3.04, "learning_rate": 2.6641152294408206e-06, "loss": 0.3892, "step": 44805 }, { "epoch": 3.04, "learning_rate": 2.663172580219449e-06, "loss": 0.3841, "step": 44810 }, { "epoch": 3.04, "learning_rate": 2.662229930998077e-06, "loss": 0.2371, "step": 44815 }, { "epoch": 3.04, "learning_rate": 2.6612872817767057e-06, "loss": 0.0894, "step": 44820 }, { "epoch": 3.04, "learning_rate": 2.6603446325553337e-06, "loss": 0.0777, "step": 44825 }, { "epoch": 3.04, "learning_rate": 2.659401983333962e-06, "loss": 0.2456, "step": 44830 }, { "epoch": 3.04, "learning_rate": 2.6584593341125903e-06, "loss": 0.1577, "step": 44835 }, { "epoch": 3.04, "learning_rate": 2.6575166848912188e-06, "loss": 0.1745, "step": 44840 }, { "epoch": 3.04, "learning_rate": 2.6565740356698464e-06, "loss": 0.0884, "step": 44845 }, { "epoch": 3.04, "learning_rate": 2.655631386448475e-06, "loss": 0.1661, "step": 44850 }, { "epoch": 3.04, "learning_rate": 2.6546887372271034e-06, "loss": 0.1175, "step": 44855 }, { "epoch": 3.04, "learning_rate": 2.6537460880057314e-06, "loss": 0.3104, "step": 44860 }, { "epoch": 3.04, "learning_rate": 2.65280343878436e-06, "loss": 0.9257, "step": 44865 }, { "epoch": 3.05, "learning_rate": 2.6518607895629884e-06, "loss": 0.2318, "step": 44870 }, { "epoch": 3.05, "learning_rate": 2.650918140341616e-06, "loss": 0.272, "step": 44875 }, { "epoch": 3.05, "learning_rate": 2.6499754911202446e-06, "loss": 0.022, "step": 44880 }, { "epoch": 3.05, "learning_rate": 2.6490328418988726e-06, "loss": 0.196, "step": 44885 }, { "epoch": 3.05, "learning_rate": 2.648090192677501e-06, "loss": 0.1976, "step": 44890 }, { "epoch": 3.05, "learning_rate": 2.6471475434561296e-06, "loss": 0.3117, "step": 44895 }, { "epoch": 3.05, "learning_rate": 2.6462048942347572e-06, "loss": 0.0902, "step": 44900 }, { "epoch": 3.05, "learning_rate": 2.6452622450133857e-06, "loss": 0.0602, "step": 44905 }, { "epoch": 3.05, "learning_rate": 2.644319595792014e-06, "loss": 0.3312, "step": 44910 }, { "epoch": 3.05, "learning_rate": 2.6433769465706423e-06, "loss": 0.3057, "step": 44915 }, { "epoch": 3.05, "learning_rate": 2.6424342973492708e-06, "loss": 0.2275, "step": 44920 }, { "epoch": 3.05, "learning_rate": 2.6414916481278992e-06, "loss": 0.5387, "step": 44925 }, { "epoch": 3.05, "learning_rate": 2.640548998906527e-06, "loss": 0.4122, "step": 44930 }, { "epoch": 3.05, "learning_rate": 2.6396063496851554e-06, "loss": 0.5253, "step": 44935 }, { "epoch": 3.05, "learning_rate": 2.638663700463784e-06, "loss": 0.088, "step": 44940 }, { "epoch": 3.05, "learning_rate": 2.637721051242412e-06, "loss": 0.105, "step": 44945 }, { "epoch": 3.05, "learning_rate": 2.63677840202104e-06, "loss": 0.095, "step": 44950 }, { "epoch": 3.05, "learning_rate": 2.6358357527996685e-06, "loss": 0.4403, "step": 44955 }, { "epoch": 3.05, "learning_rate": 2.6348931035782965e-06, "loss": 0.087, "step": 44960 }, { "epoch": 3.05, "learning_rate": 2.633950454356925e-06, "loss": 0.2296, "step": 44965 }, { "epoch": 3.05, "learning_rate": 2.6330078051355535e-06, "loss": 0.261, "step": 44970 }, { "epoch": 3.05, "learning_rate": 2.632065155914181e-06, "loss": 0.1804, "step": 44975 }, { "epoch": 3.05, "learning_rate": 2.6311225066928096e-06, "loss": 0.2832, "step": 44980 }, { "epoch": 3.05, "learning_rate": 2.6301798574714377e-06, "loss": 0.7695, "step": 44985 }, { "epoch": 3.05, "learning_rate": 2.629237208250066e-06, "loss": 0.3621, "step": 44990 }, { "epoch": 3.05, "learning_rate": 2.6282945590286947e-06, "loss": 0.4528, "step": 44995 }, { "epoch": 3.05, "learning_rate": 2.6273519098073223e-06, "loss": 1.0017, "step": 45000 }, { "epoch": 3.05, "learning_rate": 2.626409260585951e-06, "loss": 0.0883, "step": 45005 }, { "epoch": 3.05, "learning_rate": 2.6254666113645793e-06, "loss": 0.3656, "step": 45010 }, { "epoch": 3.06, "learning_rate": 2.6245239621432074e-06, "loss": 0.5492, "step": 45015 }, { "epoch": 3.06, "learning_rate": 2.623581312921836e-06, "loss": 0.1268, "step": 45020 }, { "epoch": 3.06, "learning_rate": 2.6226386637004643e-06, "loss": 0.1643, "step": 45025 }, { "epoch": 3.06, "learning_rate": 2.621696014479092e-06, "loss": 0.3195, "step": 45030 }, { "epoch": 3.06, "learning_rate": 2.6207533652577205e-06, "loss": 0.4886, "step": 45035 }, { "epoch": 3.06, "learning_rate": 2.619810716036349e-06, "loss": 0.3189, "step": 45040 }, { "epoch": 3.06, "learning_rate": 2.618868066814977e-06, "loss": 0.1585, "step": 45045 }, { "epoch": 3.06, "learning_rate": 2.6179254175936055e-06, "loss": 0.1174, "step": 45050 }, { "epoch": 3.06, "learning_rate": 2.6169827683722336e-06, "loss": 0.1154, "step": 45055 }, { "epoch": 3.06, "learning_rate": 2.6160401191508616e-06, "loss": 0.5941, "step": 45060 }, { "epoch": 3.06, "learning_rate": 2.61509746992949e-06, "loss": 0.1794, "step": 45065 }, { "epoch": 3.06, "learning_rate": 2.6141548207081186e-06, "loss": 0.1657, "step": 45070 }, { "epoch": 3.06, "learning_rate": 2.6132121714867462e-06, "loss": 0.244, "step": 45075 }, { "epoch": 3.06, "learning_rate": 2.6122695222653747e-06, "loss": 0.3741, "step": 45080 }, { "epoch": 3.06, "learning_rate": 2.6113268730440032e-06, "loss": 0.0739, "step": 45085 }, { "epoch": 3.06, "learning_rate": 2.6103842238226313e-06, "loss": 0.1122, "step": 45090 }, { "epoch": 3.06, "learning_rate": 2.6094415746012598e-06, "loss": 0.1762, "step": 45095 }, { "epoch": 3.06, "learning_rate": 2.6084989253798874e-06, "loss": 0.2941, "step": 45100 }, { "epoch": 3.06, "learning_rate": 2.607556276158516e-06, "loss": 0.1459, "step": 45105 }, { "epoch": 3.06, "learning_rate": 2.6066136269371444e-06, "loss": 0.1333, "step": 45110 }, { "epoch": 3.06, "learning_rate": 2.6056709777157724e-06, "loss": 0.0954, "step": 45115 }, { "epoch": 3.06, "learning_rate": 2.604728328494401e-06, "loss": 0.1997, "step": 45120 }, { "epoch": 3.06, "learning_rate": 2.6037856792730294e-06, "loss": 0.3618, "step": 45125 }, { "epoch": 3.06, "learning_rate": 2.602843030051657e-06, "loss": 0.0596, "step": 45130 }, { "epoch": 3.06, "learning_rate": 2.6019003808302855e-06, "loss": 0.3017, "step": 45135 }, { "epoch": 3.06, "learning_rate": 2.600957731608914e-06, "loss": 0.1645, "step": 45140 }, { "epoch": 3.06, "learning_rate": 2.600015082387542e-06, "loss": 0.4093, "step": 45145 }, { "epoch": 3.06, "learning_rate": 2.5990724331661706e-06, "loss": 0.0702, "step": 45150 }, { "epoch": 3.06, "learning_rate": 2.598129783944799e-06, "loss": 0.2731, "step": 45155 }, { "epoch": 3.07, "learning_rate": 2.5971871347234267e-06, "loss": 0.2759, "step": 45160 }, { "epoch": 3.07, "learning_rate": 2.596244485502055e-06, "loss": 0.4928, "step": 45165 }, { "epoch": 3.07, "learning_rate": 2.5953018362806837e-06, "loss": 1.0005, "step": 45170 }, { "epoch": 3.07, "learning_rate": 2.5943591870593117e-06, "loss": 0.239, "step": 45175 }, { "epoch": 3.07, "learning_rate": 2.59341653783794e-06, "loss": 0.1288, "step": 45180 }, { "epoch": 3.07, "learning_rate": 2.5924738886165683e-06, "loss": 0.0476, "step": 45185 }, { "epoch": 3.07, "learning_rate": 2.5915312393951964e-06, "loss": 0.3048, "step": 45190 }, { "epoch": 3.07, "learning_rate": 2.590588590173825e-06, "loss": 0.4444, "step": 45195 }, { "epoch": 3.07, "learning_rate": 2.5896459409524533e-06, "loss": 0.398, "step": 45200 }, { "epoch": 3.07, "learning_rate": 2.588703291731081e-06, "loss": 0.163, "step": 45205 }, { "epoch": 3.07, "learning_rate": 2.5877606425097095e-06, "loss": 0.1769, "step": 45210 }, { "epoch": 3.07, "learning_rate": 2.5868179932883375e-06, "loss": 0.2648, "step": 45215 }, { "epoch": 3.07, "learning_rate": 2.585875344066966e-06, "loss": 0.0827, "step": 45220 }, { "epoch": 3.07, "learning_rate": 2.5849326948455945e-06, "loss": 0.3073, "step": 45225 }, { "epoch": 3.07, "learning_rate": 2.583990045624222e-06, "loss": 0.1338, "step": 45230 }, { "epoch": 3.07, "learning_rate": 2.5830473964028506e-06, "loss": 0.3778, "step": 45235 }, { "epoch": 3.07, "learning_rate": 2.582104747181479e-06, "loss": 0.2847, "step": 45240 }, { "epoch": 3.07, "learning_rate": 2.581162097960107e-06, "loss": 0.0531, "step": 45245 }, { "epoch": 3.07, "learning_rate": 2.5802194487387357e-06, "loss": 0.2433, "step": 45250 }, { "epoch": 3.07, "learning_rate": 2.579276799517364e-06, "loss": 0.0569, "step": 45255 }, { "epoch": 3.07, "learning_rate": 2.578334150295992e-06, "loss": 0.2538, "step": 45260 }, { "epoch": 3.07, "learning_rate": 2.5773915010746203e-06, "loss": 0.213, "step": 45265 }, { "epoch": 3.07, "learning_rate": 2.5764488518532488e-06, "loss": 0.7141, "step": 45270 }, { "epoch": 3.07, "learning_rate": 2.575506202631877e-06, "loss": 0.1453, "step": 45275 }, { "epoch": 3.07, "learning_rate": 2.5745635534105053e-06, "loss": 0.4872, "step": 45280 }, { "epoch": 3.07, "learning_rate": 2.5736209041891334e-06, "loss": 0.3091, "step": 45285 }, { "epoch": 3.07, "learning_rate": 2.5726782549677615e-06, "loss": 0.3606, "step": 45290 }, { "epoch": 3.07, "learning_rate": 2.57173560574639e-06, "loss": 0.353, "step": 45295 }, { "epoch": 3.07, "learning_rate": 2.5707929565250184e-06, "loss": 0.1744, "step": 45300 }, { "epoch": 3.07, "learning_rate": 2.569850307303646e-06, "loss": 0.372, "step": 45305 }, { "epoch": 3.08, "learning_rate": 2.5689076580822746e-06, "loss": 0.3066, "step": 45310 }, { "epoch": 3.08, "learning_rate": 2.567965008860903e-06, "loss": 0.5472, "step": 45315 }, { "epoch": 3.08, "learning_rate": 2.567022359639531e-06, "loss": 0.5201, "step": 45320 }, { "epoch": 3.08, "learning_rate": 2.5660797104181596e-06, "loss": 0.1644, "step": 45325 }, { "epoch": 3.08, "learning_rate": 2.5651370611967872e-06, "loss": 0.3295, "step": 45330 }, { "epoch": 3.08, "learning_rate": 2.5641944119754157e-06, "loss": 0.5075, "step": 45335 }, { "epoch": 3.08, "learning_rate": 2.563251762754044e-06, "loss": 0.5679, "step": 45340 }, { "epoch": 3.08, "learning_rate": 2.5623091135326723e-06, "loss": 0.2441, "step": 45345 }, { "epoch": 3.08, "learning_rate": 2.5613664643113008e-06, "loss": 0.1428, "step": 45350 }, { "epoch": 3.08, "learning_rate": 2.5604238150899292e-06, "loss": 0.1881, "step": 45355 }, { "epoch": 3.08, "learning_rate": 2.559481165868557e-06, "loss": 0.293, "step": 45360 }, { "epoch": 3.08, "learning_rate": 2.5585385166471854e-06, "loss": 0.6984, "step": 45365 }, { "epoch": 3.08, "learning_rate": 2.557595867425814e-06, "loss": 0.5464, "step": 45370 }, { "epoch": 3.08, "learning_rate": 2.556653218204442e-06, "loss": 0.0663, "step": 45375 }, { "epoch": 3.08, "learning_rate": 2.5557105689830704e-06, "loss": 0.0078, "step": 45380 }, { "epoch": 3.08, "learning_rate": 2.554767919761699e-06, "loss": 0.2477, "step": 45385 }, { "epoch": 3.08, "learning_rate": 2.5538252705403265e-06, "loss": 0.0928, "step": 45390 }, { "epoch": 3.08, "learning_rate": 2.552882621318955e-06, "loss": 0.4855, "step": 45395 }, { "epoch": 3.08, "learning_rate": 2.5519399720975835e-06, "loss": 0.1628, "step": 45400 }, { "epoch": 3.08, "learning_rate": 2.5509973228762116e-06, "loss": 0.2944, "step": 45405 }, { "epoch": 3.08, "learning_rate": 2.5500546736548396e-06, "loss": 0.1558, "step": 45410 }, { "epoch": 3.08, "learning_rate": 2.549112024433468e-06, "loss": 0.7192, "step": 45415 }, { "epoch": 3.08, "learning_rate": 2.548169375212096e-06, "loss": 0.3111, "step": 45420 }, { "epoch": 3.08, "learning_rate": 2.5472267259907247e-06, "loss": 0.1824, "step": 45425 }, { "epoch": 3.08, "learning_rate": 2.5462840767693523e-06, "loss": 0.2067, "step": 45430 }, { "epoch": 3.08, "learning_rate": 2.545341427547981e-06, "loss": 0.4474, "step": 45435 }, { "epoch": 3.08, "learning_rate": 2.5443987783266093e-06, "loss": 0.2523, "step": 45440 }, { "epoch": 3.08, "learning_rate": 2.5434561291052374e-06, "loss": 0.1351, "step": 45445 }, { "epoch": 3.08, "learning_rate": 2.542513479883866e-06, "loss": 0.3952, "step": 45450 }, { "epoch": 3.09, "learning_rate": 2.5415708306624943e-06, "loss": 0.1428, "step": 45455 }, { "epoch": 3.09, "learning_rate": 2.540628181441122e-06, "loss": 0.0737, "step": 45460 }, { "epoch": 3.09, "learning_rate": 2.5396855322197505e-06, "loss": 0.2502, "step": 45465 }, { "epoch": 3.09, "learning_rate": 2.538742882998379e-06, "loss": 0.5905, "step": 45470 }, { "epoch": 3.09, "learning_rate": 2.537800233777007e-06, "loss": 0.1052, "step": 45475 }, { "epoch": 3.09, "learning_rate": 2.5368575845556355e-06, "loss": 0.1121, "step": 45480 }, { "epoch": 3.09, "learning_rate": 2.535914935334264e-06, "loss": 0.0872, "step": 45485 }, { "epoch": 3.09, "learning_rate": 2.5349722861128916e-06, "loss": 0.4934, "step": 45490 }, { "epoch": 3.09, "learning_rate": 2.53402963689152e-06, "loss": 0.2505, "step": 45495 }, { "epoch": 3.09, "learning_rate": 2.5330869876701486e-06, "loss": 0.1977, "step": 45500 }, { "epoch": 3.09, "learning_rate": 2.5321443384487767e-06, "loss": 0.2149, "step": 45505 }, { "epoch": 3.09, "learning_rate": 2.531201689227405e-06, "loss": 0.1362, "step": 45510 }, { "epoch": 3.09, "learning_rate": 2.5302590400060332e-06, "loss": 0.4498, "step": 45515 }, { "epoch": 3.09, "learning_rate": 2.5293163907846613e-06, "loss": 0.0642, "step": 45520 }, { "epoch": 3.09, "learning_rate": 2.5283737415632898e-06, "loss": 0.3921, "step": 45525 }, { "epoch": 3.09, "learning_rate": 2.5274310923419183e-06, "loss": 0.5383, "step": 45530 }, { "epoch": 3.09, "learning_rate": 2.526488443120546e-06, "loss": 0.06, "step": 45535 }, { "epoch": 3.09, "learning_rate": 2.5255457938991744e-06, "loss": 0.4702, "step": 45540 }, { "epoch": 3.09, "learning_rate": 2.5246031446778024e-06, "loss": 0.3452, "step": 45545 }, { "epoch": 3.09, "learning_rate": 2.523660495456431e-06, "loss": 0.2013, "step": 45550 }, { "epoch": 3.09, "learning_rate": 2.5227178462350594e-06, "loss": 0.1223, "step": 45555 }, { "epoch": 3.09, "learning_rate": 2.521775197013687e-06, "loss": 0.4405, "step": 45560 }, { "epoch": 3.09, "learning_rate": 2.5208325477923155e-06, "loss": 0.195, "step": 45565 }, { "epoch": 3.09, "learning_rate": 2.519889898570944e-06, "loss": 1.2884, "step": 45570 }, { "epoch": 3.09, "learning_rate": 2.518947249349572e-06, "loss": 0.0585, "step": 45575 }, { "epoch": 3.09, "learning_rate": 2.5180046001282006e-06, "loss": 0.1382, "step": 45580 }, { "epoch": 3.09, "learning_rate": 2.517061950906829e-06, "loss": 0.1477, "step": 45585 }, { "epoch": 3.09, "learning_rate": 2.5161193016854567e-06, "loss": 0.5035, "step": 45590 }, { "epoch": 3.09, "learning_rate": 2.515176652464085e-06, "loss": 0.313, "step": 45595 }, { "epoch": 3.09, "learning_rate": 2.5142340032427137e-06, "loss": 0.3318, "step": 45600 }, { "epoch": 3.1, "learning_rate": 2.5132913540213418e-06, "loss": 0.6838, "step": 45605 }, { "epoch": 3.1, "learning_rate": 2.5123487047999702e-06, "loss": 0.3234, "step": 45610 }, { "epoch": 3.1, "learning_rate": 2.5114060555785983e-06, "loss": 0.7253, "step": 45615 }, { "epoch": 3.1, "learning_rate": 2.5104634063572264e-06, "loss": 0.2961, "step": 45620 }, { "epoch": 3.1, "learning_rate": 2.509520757135855e-06, "loss": 0.7618, "step": 45625 }, { "epoch": 3.1, "learning_rate": 2.5085781079144833e-06, "loss": 0.2445, "step": 45630 }, { "epoch": 3.1, "learning_rate": 2.5076354586931114e-06, "loss": 0.4633, "step": 45635 }, { "epoch": 3.1, "learning_rate": 2.5066928094717395e-06, "loss": 0.312, "step": 45640 }, { "epoch": 3.1, "learning_rate": 2.505750160250368e-06, "loss": 0.2966, "step": 45645 }, { "epoch": 3.1, "learning_rate": 2.504807511028996e-06, "loss": 0.1674, "step": 45650 }, { "epoch": 3.1, "learning_rate": 2.5038648618076245e-06, "loss": 0.0363, "step": 45655 }, { "epoch": 3.1, "learning_rate": 2.502922212586252e-06, "loss": 0.2962, "step": 45660 }, { "epoch": 3.1, "learning_rate": 2.5019795633648806e-06, "loss": 0.2579, "step": 45665 }, { "epoch": 3.1, "learning_rate": 2.501036914143509e-06, "loss": 0.2865, "step": 45670 }, { "epoch": 3.1, "learning_rate": 2.500094264922137e-06, "loss": 0.2487, "step": 45675 }, { "epoch": 3.1, "learning_rate": 2.4991516157007657e-06, "loss": 0.148, "step": 45680 }, { "epoch": 3.1, "learning_rate": 2.4982089664793937e-06, "loss": 0.1621, "step": 45685 }, { "epoch": 3.1, "learning_rate": 2.4972663172580222e-06, "loss": 0.1644, "step": 45690 }, { "epoch": 3.1, "learning_rate": 2.4963236680366503e-06, "loss": 0.1719, "step": 45695 }, { "epoch": 3.1, "learning_rate": 2.4953810188152784e-06, "loss": 0.2815, "step": 45700 }, { "epoch": 3.1, "learning_rate": 2.494438369593907e-06, "loss": 0.0321, "step": 45705 }, { "epoch": 3.1, "learning_rate": 2.4934957203725353e-06, "loss": 0.5604, "step": 45710 }, { "epoch": 3.1, "learning_rate": 2.4925530711511634e-06, "loss": 0.2208, "step": 45715 }, { "epoch": 3.1, "learning_rate": 2.491610421929792e-06, "loss": 0.451, "step": 45720 }, { "epoch": 3.1, "learning_rate": 2.49066777270842e-06, "loss": 0.1368, "step": 45725 }, { "epoch": 3.1, "learning_rate": 2.489725123487048e-06, "loss": 0.1853, "step": 45730 }, { "epoch": 3.1, "learning_rate": 2.4887824742656765e-06, "loss": 0.3069, "step": 45735 }, { "epoch": 3.1, "learning_rate": 2.4878398250443046e-06, "loss": 0.351, "step": 45740 }, { "epoch": 3.1, "learning_rate": 2.486897175822933e-06, "loss": 0.2272, "step": 45745 }, { "epoch": 3.11, "learning_rate": 2.485954526601561e-06, "loss": 0.4203, "step": 45750 }, { "epoch": 3.11, "learning_rate": 2.4850118773801896e-06, "loss": 0.206, "step": 45755 }, { "epoch": 3.11, "learning_rate": 2.4840692281588177e-06, "loss": 0.0246, "step": 45760 }, { "epoch": 3.11, "learning_rate": 2.4831265789374457e-06, "loss": 0.5183, "step": 45765 }, { "epoch": 3.11, "learning_rate": 2.482183929716074e-06, "loss": 0.0757, "step": 45770 }, { "epoch": 3.11, "learning_rate": 2.4812412804947027e-06, "loss": 0.1345, "step": 45775 }, { "epoch": 3.11, "learning_rate": 2.4802986312733308e-06, "loss": 0.1668, "step": 45780 }, { "epoch": 3.11, "learning_rate": 2.4793559820519592e-06, "loss": 0.1872, "step": 45785 }, { "epoch": 3.11, "learning_rate": 2.4784133328305873e-06, "loss": 0.6603, "step": 45790 }, { "epoch": 3.11, "learning_rate": 2.4774706836092154e-06, "loss": 0.1789, "step": 45795 }, { "epoch": 3.11, "learning_rate": 2.476528034387844e-06, "loss": 0.1356, "step": 45800 }, { "epoch": 3.11, "learning_rate": 2.475585385166472e-06, "loss": 0.1939, "step": 45805 }, { "epoch": 3.11, "learning_rate": 2.4746427359451004e-06, "loss": 0.2205, "step": 45810 }, { "epoch": 3.11, "learning_rate": 2.4737000867237285e-06, "loss": 0.1321, "step": 45815 }, { "epoch": 3.11, "learning_rate": 2.472757437502357e-06, "loss": 0.2218, "step": 45820 }, { "epoch": 3.11, "learning_rate": 2.471814788280985e-06, "loss": 0.5335, "step": 45825 }, { "epoch": 3.11, "learning_rate": 2.470872139059613e-06, "loss": 0.2086, "step": 45830 }, { "epoch": 3.11, "learning_rate": 2.4699294898382416e-06, "loss": 0.1456, "step": 45835 }, { "epoch": 3.11, "learning_rate": 2.46898684061687e-06, "loss": 0.3995, "step": 45840 }, { "epoch": 3.11, "learning_rate": 2.468044191395498e-06, "loss": 0.2217, "step": 45845 }, { "epoch": 3.11, "learning_rate": 2.467101542174126e-06, "loss": 0.1308, "step": 45850 }, { "epoch": 3.11, "learning_rate": 2.4661588929527547e-06, "loss": 0.2086, "step": 45855 }, { "epoch": 3.11, "learning_rate": 2.4652162437313827e-06, "loss": 0.1581, "step": 45860 }, { "epoch": 3.11, "learning_rate": 2.4642735945100112e-06, "loss": 0.1488, "step": 45865 }, { "epoch": 3.11, "learning_rate": 2.4633309452886393e-06, "loss": 0.0263, "step": 45870 }, { "epoch": 3.11, "learning_rate": 2.4623882960672678e-06, "loss": 0.1685, "step": 45875 }, { "epoch": 3.11, "learning_rate": 2.461445646845896e-06, "loss": 0.1616, "step": 45880 }, { "epoch": 3.11, "learning_rate": 2.4605029976245243e-06, "loss": 0.3522, "step": 45885 }, { "epoch": 3.11, "learning_rate": 2.4595603484031524e-06, "loss": 0.3598, "step": 45890 }, { "epoch": 3.11, "learning_rate": 2.4586176991817805e-06, "loss": 0.1123, "step": 45895 }, { "epoch": 3.12, "learning_rate": 2.457675049960409e-06, "loss": 0.0607, "step": 45900 }, { "epoch": 3.12, "learning_rate": 2.4567324007390374e-06, "loss": 0.2492, "step": 45905 }, { "epoch": 3.12, "learning_rate": 2.4557897515176655e-06, "loss": 0.5905, "step": 45910 }, { "epoch": 3.12, "learning_rate": 2.4548471022962936e-06, "loss": 0.1949, "step": 45915 }, { "epoch": 3.12, "learning_rate": 2.453904453074922e-06, "loss": 0.6413, "step": 45920 }, { "epoch": 3.12, "learning_rate": 2.45296180385355e-06, "loss": 0.1221, "step": 45925 }, { "epoch": 3.12, "learning_rate": 2.452019154632178e-06, "loss": 0.0832, "step": 45930 }, { "epoch": 3.12, "learning_rate": 2.4510765054108067e-06, "loss": 0.5806, "step": 45935 }, { "epoch": 3.12, "learning_rate": 2.450133856189435e-06, "loss": 0.9002, "step": 45940 }, { "epoch": 3.12, "learning_rate": 2.4491912069680632e-06, "loss": 0.0967, "step": 45945 }, { "epoch": 3.12, "learning_rate": 2.4482485577466917e-06, "loss": 0.431, "step": 45950 }, { "epoch": 3.12, "learning_rate": 2.4473059085253198e-06, "loss": 0.7363, "step": 45955 }, { "epoch": 3.12, "learning_rate": 2.446363259303948e-06, "loss": 0.1148, "step": 45960 }, { "epoch": 3.12, "learning_rate": 2.4454206100825763e-06, "loss": 0.2483, "step": 45965 }, { "epoch": 3.12, "learning_rate": 2.4444779608612044e-06, "loss": 0.3513, "step": 45970 }, { "epoch": 3.12, "learning_rate": 2.443535311639833e-06, "loss": 0.2113, "step": 45975 }, { "epoch": 3.12, "learning_rate": 2.442592662418461e-06, "loss": 0.048, "step": 45980 }, { "epoch": 3.12, "learning_rate": 2.4416500131970894e-06, "loss": 0.6843, "step": 45985 }, { "epoch": 3.12, "learning_rate": 2.4407073639757175e-06, "loss": 0.3943, "step": 45990 }, { "epoch": 3.12, "learning_rate": 2.4397647147543456e-06, "loss": 0.3991, "step": 45995 }, { "epoch": 3.12, "learning_rate": 2.438822065532974e-06, "loss": 0.443, "step": 46000 }, { "epoch": 3.12, "learning_rate": 2.4378794163116025e-06, "loss": 0.5783, "step": 46005 }, { "epoch": 3.12, "learning_rate": 2.4369367670902306e-06, "loss": 0.4452, "step": 46010 }, { "epoch": 3.12, "learning_rate": 2.435994117868859e-06, "loss": 0.1554, "step": 46015 }, { "epoch": 3.12, "learning_rate": 2.435051468647487e-06, "loss": 0.2013, "step": 46020 }, { "epoch": 3.12, "learning_rate": 2.434108819426115e-06, "loss": 0.5633, "step": 46025 }, { "epoch": 3.12, "learning_rate": 2.4331661702047437e-06, "loss": 0.0265, "step": 46030 }, { "epoch": 3.12, "learning_rate": 2.4322235209833718e-06, "loss": 0.1092, "step": 46035 }, { "epoch": 3.12, "learning_rate": 2.4312808717620002e-06, "loss": 0.152, "step": 46040 }, { "epoch": 3.13, "learning_rate": 2.4303382225406283e-06, "loss": 0.2148, "step": 46045 }, { "epoch": 3.13, "learning_rate": 2.429395573319257e-06, "loss": 0.2691, "step": 46050 }, { "epoch": 3.13, "learning_rate": 2.428452924097885e-06, "loss": 0.3439, "step": 46055 }, { "epoch": 3.13, "learning_rate": 2.427510274876513e-06, "loss": 0.4465, "step": 46060 }, { "epoch": 3.13, "learning_rate": 2.4265676256551414e-06, "loss": 0.4065, "step": 46065 }, { "epoch": 3.13, "learning_rate": 2.42562497643377e-06, "loss": 0.5844, "step": 46070 }, { "epoch": 3.13, "learning_rate": 2.424682327212398e-06, "loss": 0.2634, "step": 46075 }, { "epoch": 3.13, "learning_rate": 2.423739677991026e-06, "loss": 0.1152, "step": 46080 }, { "epoch": 3.13, "learning_rate": 2.4227970287696545e-06, "loss": 0.5169, "step": 46085 }, { "epoch": 3.13, "learning_rate": 2.4218543795482826e-06, "loss": 0.151, "step": 46090 }, { "epoch": 3.13, "learning_rate": 2.4209117303269106e-06, "loss": 0.7305, "step": 46095 }, { "epoch": 3.13, "learning_rate": 2.419969081105539e-06, "loss": 0.0434, "step": 46100 }, { "epoch": 3.13, "learning_rate": 2.4190264318841676e-06, "loss": 0.4367, "step": 46105 }, { "epoch": 3.13, "learning_rate": 2.4180837826627957e-06, "loss": 0.4087, "step": 46110 }, { "epoch": 3.13, "learning_rate": 2.417141133441424e-06, "loss": 0.4985, "step": 46115 }, { "epoch": 3.13, "learning_rate": 2.4161984842200522e-06, "loss": 0.2577, "step": 46120 }, { "epoch": 3.13, "learning_rate": 2.4152558349986803e-06, "loss": 0.104, "step": 46125 }, { "epoch": 3.13, "learning_rate": 2.4143131857773088e-06, "loss": 0.3161, "step": 46130 }, { "epoch": 3.13, "learning_rate": 2.4133705365559373e-06, "loss": 0.2751, "step": 46135 }, { "epoch": 3.13, "learning_rate": 2.4124278873345653e-06, "loss": 0.4031, "step": 46140 }, { "epoch": 3.13, "learning_rate": 2.4114852381131934e-06, "loss": 0.0808, "step": 46145 }, { "epoch": 3.13, "learning_rate": 2.410542588891822e-06, "loss": 0.3177, "step": 46150 }, { "epoch": 3.13, "learning_rate": 2.40959993967045e-06, "loss": 0.1655, "step": 46155 }, { "epoch": 3.13, "learning_rate": 2.408657290449078e-06, "loss": 0.2503, "step": 46160 }, { "epoch": 3.13, "learning_rate": 2.4077146412277065e-06, "loss": 0.1909, "step": 46165 }, { "epoch": 3.13, "learning_rate": 2.406771992006335e-06, "loss": 0.536, "step": 46170 }, { "epoch": 3.13, "learning_rate": 2.405829342784963e-06, "loss": 0.0729, "step": 46175 }, { "epoch": 3.13, "learning_rate": 2.4048866935635915e-06, "loss": 0.0695, "step": 46180 }, { "epoch": 3.13, "learning_rate": 2.4039440443422196e-06, "loss": 0.059, "step": 46185 }, { "epoch": 3.13, "learning_rate": 2.4030013951208477e-06, "loss": 0.8619, "step": 46190 }, { "epoch": 3.14, "learning_rate": 2.402058745899476e-06, "loss": 0.0431, "step": 46195 }, { "epoch": 3.14, "learning_rate": 2.4011160966781042e-06, "loss": 0.0598, "step": 46200 }, { "epoch": 3.14, "learning_rate": 2.4001734474567327e-06, "loss": 0.113, "step": 46205 }, { "epoch": 3.14, "learning_rate": 2.3992307982353608e-06, "loss": 0.3326, "step": 46210 }, { "epoch": 3.14, "learning_rate": 2.3982881490139893e-06, "loss": 0.2056, "step": 46215 }, { "epoch": 3.14, "learning_rate": 2.3973454997926173e-06, "loss": 0.162, "step": 46220 }, { "epoch": 3.14, "learning_rate": 2.3964028505712454e-06, "loss": 0.2788, "step": 46225 }, { "epoch": 3.14, "learning_rate": 2.395460201349874e-06, "loss": 0.0803, "step": 46230 }, { "epoch": 3.14, "learning_rate": 2.3945175521285024e-06, "loss": 0.3255, "step": 46235 }, { "epoch": 3.14, "learning_rate": 2.3935749029071304e-06, "loss": 0.237, "step": 46240 }, { "epoch": 3.14, "learning_rate": 2.3926322536857585e-06, "loss": 0.1551, "step": 46245 }, { "epoch": 3.14, "learning_rate": 2.391689604464387e-06, "loss": 0.3801, "step": 46250 }, { "epoch": 3.14, "learning_rate": 2.390746955243015e-06, "loss": 0.3407, "step": 46255 }, { "epoch": 3.14, "learning_rate": 2.3898043060216435e-06, "loss": 0.3685, "step": 46260 }, { "epoch": 3.14, "learning_rate": 2.3888616568002716e-06, "loss": 0.2218, "step": 46265 }, { "epoch": 3.14, "learning_rate": 2.3879190075789e-06, "loss": 0.3628, "step": 46270 }, { "epoch": 3.14, "learning_rate": 2.386976358357528e-06, "loss": 0.3683, "step": 46275 }, { "epoch": 3.14, "learning_rate": 2.3860337091361566e-06, "loss": 0.2137, "step": 46280 }, { "epoch": 3.14, "learning_rate": 2.3850910599147847e-06, "loss": 0.3364, "step": 46285 }, { "epoch": 3.14, "learning_rate": 2.3841484106934127e-06, "loss": 0.0579, "step": 46290 }, { "epoch": 3.14, "learning_rate": 2.3832057614720412e-06, "loss": 0.1414, "step": 46295 }, { "epoch": 3.14, "learning_rate": 2.3822631122506697e-06, "loss": 0.3405, "step": 46300 }, { "epoch": 3.14, "learning_rate": 2.3813204630292978e-06, "loss": 0.3109, "step": 46305 }, { "epoch": 3.14, "learning_rate": 2.380377813807926e-06, "loss": 0.3188, "step": 46310 }, { "epoch": 3.14, "learning_rate": 2.3794351645865543e-06, "loss": 0.6179, "step": 46315 }, { "epoch": 3.14, "learning_rate": 2.3784925153651824e-06, "loss": 0.277, "step": 46320 }, { "epoch": 3.14, "learning_rate": 2.3775498661438105e-06, "loss": 0.0681, "step": 46325 }, { "epoch": 3.14, "learning_rate": 2.376607216922439e-06, "loss": 0.618, "step": 46330 }, { "epoch": 3.14, "learning_rate": 2.3756645677010674e-06, "loss": 0.2597, "step": 46335 }, { "epoch": 3.15, "learning_rate": 2.3747219184796955e-06, "loss": 0.1495, "step": 46340 }, { "epoch": 3.15, "learning_rate": 2.373779269258324e-06, "loss": 0.1249, "step": 46345 }, { "epoch": 3.15, "learning_rate": 2.372836620036952e-06, "loss": 0.4271, "step": 46350 }, { "epoch": 3.15, "learning_rate": 2.37189397081558e-06, "loss": 0.3298, "step": 46355 }, { "epoch": 3.15, "learning_rate": 2.3709513215942086e-06, "loss": 0.4921, "step": 46360 }, { "epoch": 3.15, "learning_rate": 2.370008672372837e-06, "loss": 0.8793, "step": 46365 }, { "epoch": 3.15, "learning_rate": 2.369066023151465e-06, "loss": 0.4253, "step": 46370 }, { "epoch": 3.15, "learning_rate": 2.3681233739300932e-06, "loss": 0.3214, "step": 46375 }, { "epoch": 3.15, "learning_rate": 2.3671807247087217e-06, "loss": 0.2286, "step": 46380 }, { "epoch": 3.15, "learning_rate": 2.3662380754873498e-06, "loss": 0.0985, "step": 46385 }, { "epoch": 3.15, "learning_rate": 2.365295426265978e-06, "loss": 0.2078, "step": 46390 }, { "epoch": 3.15, "learning_rate": 2.3643527770446063e-06, "loss": 0.2906, "step": 46395 }, { "epoch": 3.15, "learning_rate": 2.363410127823235e-06, "loss": 0.2227, "step": 46400 }, { "epoch": 3.15, "learning_rate": 2.362467478601863e-06, "loss": 0.3666, "step": 46405 }, { "epoch": 3.15, "learning_rate": 2.361524829380491e-06, "loss": 0.32, "step": 46410 }, { "epoch": 3.15, "learning_rate": 2.3605821801591194e-06, "loss": 0.065, "step": 46415 }, { "epoch": 3.15, "learning_rate": 2.3596395309377475e-06, "loss": 0.2352, "step": 46420 }, { "epoch": 3.15, "learning_rate": 2.358696881716376e-06, "loss": 0.2735, "step": 46425 }, { "epoch": 3.15, "learning_rate": 2.357754232495004e-06, "loss": 0.2636, "step": 46430 }, { "epoch": 3.15, "learning_rate": 2.3568115832736325e-06, "loss": 0.3089, "step": 46435 }, { "epoch": 3.15, "learning_rate": 2.3558689340522606e-06, "loss": 0.1073, "step": 46440 }, { "epoch": 3.15, "learning_rate": 2.354926284830889e-06, "loss": 0.3163, "step": 46445 }, { "epoch": 3.15, "learning_rate": 2.353983635609517e-06, "loss": 0.3998, "step": 46450 }, { "epoch": 3.15, "learning_rate": 2.353040986388145e-06, "loss": 0.571, "step": 46455 }, { "epoch": 3.15, "learning_rate": 2.3520983371667737e-06, "loss": 0.2288, "step": 46460 }, { "epoch": 3.15, "learning_rate": 2.351155687945402e-06, "loss": 0.3255, "step": 46465 }, { "epoch": 3.15, "learning_rate": 2.3502130387240302e-06, "loss": 0.0489, "step": 46470 }, { "epoch": 3.15, "learning_rate": 2.3492703895026583e-06, "loss": 0.0338, "step": 46475 }, { "epoch": 3.15, "learning_rate": 2.348327740281287e-06, "loss": 0.1404, "step": 46480 }, { "epoch": 3.15, "learning_rate": 2.347385091059915e-06, "loss": 0.0838, "step": 46485 }, { "epoch": 3.16, "learning_rate": 2.3464424418385433e-06, "loss": 0.2846, "step": 46490 }, { "epoch": 3.16, "learning_rate": 2.3454997926171714e-06, "loss": 0.2384, "step": 46495 }, { "epoch": 3.16, "learning_rate": 2.3445571433958e-06, "loss": 0.6553, "step": 46500 }, { "epoch": 3.16, "learning_rate": 2.343614494174428e-06, "loss": 0.1469, "step": 46505 }, { "epoch": 3.16, "learning_rate": 2.3426718449530564e-06, "loss": 0.0674, "step": 46510 }, { "epoch": 3.16, "learning_rate": 2.3417291957316845e-06, "loss": 0.2927, "step": 46515 }, { "epoch": 3.16, "learning_rate": 2.3407865465103126e-06, "loss": 0.207, "step": 46520 }, { "epoch": 3.16, "learning_rate": 2.339843897288941e-06, "loss": 0.2462, "step": 46525 }, { "epoch": 3.16, "learning_rate": 2.3389012480675696e-06, "loss": 0.1673, "step": 46530 }, { "epoch": 3.16, "learning_rate": 2.3379585988461976e-06, "loss": 0.1648, "step": 46535 }, { "epoch": 3.16, "learning_rate": 2.3370159496248257e-06, "loss": 0.531, "step": 46540 }, { "epoch": 3.16, "learning_rate": 2.336073300403454e-06, "loss": 0.2255, "step": 46545 }, { "epoch": 3.16, "learning_rate": 2.3351306511820822e-06, "loss": 0.412, "step": 46550 }, { "epoch": 3.16, "learning_rate": 2.3341880019607103e-06, "loss": 0.3934, "step": 46555 }, { "epoch": 3.16, "learning_rate": 2.3332453527393388e-06, "loss": 0.2514, "step": 46560 }, { "epoch": 3.16, "learning_rate": 2.3323027035179673e-06, "loss": 0.2808, "step": 46565 }, { "epoch": 3.16, "learning_rate": 2.3313600542965953e-06, "loss": 0.5729, "step": 46570 }, { "epoch": 3.16, "learning_rate": 2.3304174050752234e-06, "loss": 0.2719, "step": 46575 }, { "epoch": 3.16, "learning_rate": 2.329474755853852e-06, "loss": 0.0243, "step": 46580 }, { "epoch": 3.16, "learning_rate": 2.32853210663248e-06, "loss": 0.0684, "step": 46585 }, { "epoch": 3.16, "learning_rate": 2.3275894574111084e-06, "loss": 0.3107, "step": 46590 }, { "epoch": 3.16, "learning_rate": 2.3266468081897365e-06, "loss": 0.4064, "step": 46595 }, { "epoch": 3.16, "learning_rate": 2.325704158968365e-06, "loss": 0.0836, "step": 46600 }, { "epoch": 3.16, "learning_rate": 2.324761509746993e-06, "loss": 0.1108, "step": 46605 }, { "epoch": 3.16, "learning_rate": 2.3238188605256215e-06, "loss": 0.1373, "step": 46610 }, { "epoch": 3.16, "learning_rate": 2.3228762113042496e-06, "loss": 0.2268, "step": 46615 }, { "epoch": 3.16, "learning_rate": 2.3219335620828777e-06, "loss": 0.2056, "step": 46620 }, { "epoch": 3.16, "learning_rate": 2.320990912861506e-06, "loss": 0.3575, "step": 46625 }, { "epoch": 3.16, "learning_rate": 2.3200482636401346e-06, "loss": 0.1965, "step": 46630 }, { "epoch": 3.17, "learning_rate": 2.3191056144187627e-06, "loss": 0.2145, "step": 46635 }, { "epoch": 3.17, "learning_rate": 2.3181629651973908e-06, "loss": 0.3193, "step": 46640 }, { "epoch": 3.17, "learning_rate": 2.3172203159760193e-06, "loss": 0.0744, "step": 46645 }, { "epoch": 3.17, "learning_rate": 2.3162776667546473e-06, "loss": 0.1545, "step": 46650 }, { "epoch": 3.17, "learning_rate": 2.315335017533276e-06, "loss": 0.6094, "step": 46655 }, { "epoch": 3.17, "learning_rate": 2.314392368311904e-06, "loss": 0.6075, "step": 46660 }, { "epoch": 3.17, "learning_rate": 2.3134497190905324e-06, "loss": 0.4523, "step": 46665 }, { "epoch": 3.17, "learning_rate": 2.3125070698691604e-06, "loss": 0.1926, "step": 46670 }, { "epoch": 3.17, "learning_rate": 2.311564420647789e-06, "loss": 0.4292, "step": 46675 }, { "epoch": 3.17, "learning_rate": 2.310621771426417e-06, "loss": 0.1973, "step": 46680 }, { "epoch": 3.17, "learning_rate": 2.309679122205045e-06, "loss": 0.1086, "step": 46685 }, { "epoch": 3.17, "learning_rate": 2.3087364729836735e-06, "loss": 0.2929, "step": 46690 }, { "epoch": 3.17, "learning_rate": 2.307793823762302e-06, "loss": 0.4424, "step": 46695 }, { "epoch": 3.17, "learning_rate": 2.30685117454093e-06, "loss": 0.7703, "step": 46700 }, { "epoch": 3.17, "learning_rate": 2.305908525319558e-06, "loss": 0.2581, "step": 46705 }, { "epoch": 3.17, "learning_rate": 2.3049658760981866e-06, "loss": 0.8143, "step": 46710 }, { "epoch": 3.17, "learning_rate": 2.3040232268768147e-06, "loss": 0.1299, "step": 46715 }, { "epoch": 3.17, "learning_rate": 2.3030805776554428e-06, "loss": 0.6804, "step": 46720 }, { "epoch": 3.17, "learning_rate": 2.3021379284340712e-06, "loss": 0.4638, "step": 46725 }, { "epoch": 3.17, "learning_rate": 2.3011952792126997e-06, "loss": 0.5744, "step": 46730 }, { "epoch": 3.17, "learning_rate": 2.300252629991328e-06, "loss": 0.4629, "step": 46735 }, { "epoch": 3.17, "learning_rate": 2.2993099807699563e-06, "loss": 0.1442, "step": 46740 }, { "epoch": 3.17, "learning_rate": 2.2983673315485843e-06, "loss": 0.1628, "step": 46745 }, { "epoch": 3.17, "learning_rate": 2.2974246823272124e-06, "loss": 0.1513, "step": 46750 }, { "epoch": 3.17, "learning_rate": 2.296482033105841e-06, "loss": 0.1331, "step": 46755 }, { "epoch": 3.17, "learning_rate": 2.2955393838844694e-06, "loss": 0.2414, "step": 46760 }, { "epoch": 3.17, "learning_rate": 2.2945967346630974e-06, "loss": 0.2164, "step": 46765 }, { "epoch": 3.17, "learning_rate": 2.2936540854417255e-06, "loss": 0.1348, "step": 46770 }, { "epoch": 3.17, "learning_rate": 2.292711436220354e-06, "loss": 0.3332, "step": 46775 }, { "epoch": 3.17, "learning_rate": 2.291768786998982e-06, "loss": 0.4437, "step": 46780 }, { "epoch": 3.18, "learning_rate": 2.29082613777761e-06, "loss": 0.4698, "step": 46785 }, { "epoch": 3.18, "learning_rate": 2.2898834885562386e-06, "loss": 0.1978, "step": 46790 }, { "epoch": 3.18, "learning_rate": 2.288940839334867e-06, "loss": 0.34, "step": 46795 }, { "epoch": 3.18, "learning_rate": 2.287998190113495e-06, "loss": 0.4479, "step": 46800 }, { "epoch": 3.18, "learning_rate": 2.2870555408921232e-06, "loss": 0.1189, "step": 46805 }, { "epoch": 3.18, "learning_rate": 2.2861128916707517e-06, "loss": 0.587, "step": 46810 }, { "epoch": 3.18, "learning_rate": 2.2851702424493798e-06, "loss": 0.1151, "step": 46815 }, { "epoch": 3.18, "learning_rate": 2.2842275932280083e-06, "loss": 0.4735, "step": 46820 }, { "epoch": 3.18, "learning_rate": 2.2832849440066363e-06, "loss": 0.2604, "step": 46825 }, { "epoch": 3.18, "learning_rate": 2.282342294785265e-06, "loss": 0.2792, "step": 46830 }, { "epoch": 3.18, "learning_rate": 2.281399645563893e-06, "loss": 0.0482, "step": 46835 }, { "epoch": 3.18, "learning_rate": 2.2804569963425214e-06, "loss": 0.3374, "step": 46840 }, { "epoch": 3.18, "learning_rate": 2.2795143471211494e-06, "loss": 0.4132, "step": 46845 }, { "epoch": 3.18, "learning_rate": 2.2785716978997775e-06, "loss": 0.5003, "step": 46850 }, { "epoch": 3.18, "learning_rate": 2.277629048678406e-06, "loss": 0.0964, "step": 46855 }, { "epoch": 3.18, "learning_rate": 2.2766863994570345e-06, "loss": 0.2917, "step": 46860 }, { "epoch": 3.18, "learning_rate": 2.2757437502356625e-06, "loss": 0.2075, "step": 46865 }, { "epoch": 3.18, "learning_rate": 2.2748011010142906e-06, "loss": 0.2737, "step": 46870 }, { "epoch": 3.18, "learning_rate": 2.273858451792919e-06, "loss": 0.3703, "step": 46875 }, { "epoch": 3.18, "learning_rate": 2.272915802571547e-06, "loss": 0.1023, "step": 46880 }, { "epoch": 3.18, "learning_rate": 2.2719731533501756e-06, "loss": 0.1006, "step": 46885 }, { "epoch": 3.18, "learning_rate": 2.2710305041288037e-06, "loss": 0.3234, "step": 46890 }, { "epoch": 3.18, "learning_rate": 2.270087854907432e-06, "loss": 0.2374, "step": 46895 }, { "epoch": 3.18, "learning_rate": 2.2691452056860602e-06, "loss": 0.094, "step": 46900 }, { "epoch": 3.18, "learning_rate": 2.2682025564646887e-06, "loss": 0.3572, "step": 46905 }, { "epoch": 3.18, "learning_rate": 2.267259907243317e-06, "loss": 0.1611, "step": 46910 }, { "epoch": 3.18, "learning_rate": 2.266317258021945e-06, "loss": 0.1058, "step": 46915 }, { "epoch": 3.18, "learning_rate": 2.2653746088005734e-06, "loss": 0.2854, "step": 46920 }, { "epoch": 3.18, "learning_rate": 2.264431959579202e-06, "loss": 0.1061, "step": 46925 }, { "epoch": 3.19, "learning_rate": 2.26348931035783e-06, "loss": 0.3168, "step": 46930 }, { "epoch": 3.19, "learning_rate": 2.262546661136458e-06, "loss": 0.2591, "step": 46935 }, { "epoch": 3.19, "learning_rate": 2.2616040119150865e-06, "loss": 0.1225, "step": 46940 }, { "epoch": 3.19, "learning_rate": 2.2606613626937145e-06, "loss": 0.1187, "step": 46945 }, { "epoch": 3.19, "learning_rate": 2.2597187134723426e-06, "loss": 0.2936, "step": 46950 }, { "epoch": 3.19, "learning_rate": 2.258776064250971e-06, "loss": 0.0867, "step": 46955 }, { "epoch": 3.19, "learning_rate": 2.2578334150295996e-06, "loss": 0.4433, "step": 46960 }, { "epoch": 3.19, "learning_rate": 2.2568907658082276e-06, "loss": 0.4628, "step": 46965 }, { "epoch": 3.19, "learning_rate": 2.2559481165868557e-06, "loss": 0.2271, "step": 46970 }, { "epoch": 3.19, "learning_rate": 2.255005467365484e-06, "loss": 0.1583, "step": 46975 }, { "epoch": 3.19, "learning_rate": 2.2540628181441122e-06, "loss": 0.4108, "step": 46980 }, { "epoch": 3.19, "learning_rate": 2.2531201689227407e-06, "loss": 0.1414, "step": 46985 }, { "epoch": 3.19, "learning_rate": 2.252177519701369e-06, "loss": 0.2666, "step": 46990 }, { "epoch": 3.19, "learning_rate": 2.2512348704799973e-06, "loss": 0.2438, "step": 46995 }, { "epoch": 3.19, "learning_rate": 2.2502922212586253e-06, "loss": 0.5837, "step": 47000 }, { "epoch": 3.19, "learning_rate": 2.249349572037254e-06, "loss": 0.2052, "step": 47005 }, { "epoch": 3.19, "learning_rate": 2.248406922815882e-06, "loss": 0.1216, "step": 47010 }, { "epoch": 3.19, "learning_rate": 2.24746427359451e-06, "loss": 0.1619, "step": 47015 }, { "epoch": 3.19, "learning_rate": 2.2465216243731384e-06, "loss": 0.1577, "step": 47020 }, { "epoch": 3.19, "learning_rate": 2.245578975151767e-06, "loss": 0.2006, "step": 47025 }, { "epoch": 3.19, "learning_rate": 2.244636325930395e-06, "loss": 0.1699, "step": 47030 }, { "epoch": 3.19, "learning_rate": 2.243693676709023e-06, "loss": 0.5258, "step": 47035 }, { "epoch": 3.19, "learning_rate": 2.2427510274876515e-06, "loss": 0.256, "step": 47040 }, { "epoch": 3.19, "learning_rate": 2.2418083782662796e-06, "loss": 0.1122, "step": 47045 }, { "epoch": 3.19, "learning_rate": 2.240865729044908e-06, "loss": 0.1833, "step": 47050 }, { "epoch": 3.19, "learning_rate": 2.239923079823536e-06, "loss": 0.3129, "step": 47055 }, { "epoch": 3.19, "learning_rate": 2.2389804306021646e-06, "loss": 0.0865, "step": 47060 }, { "epoch": 3.19, "learning_rate": 2.2380377813807927e-06, "loss": 0.1531, "step": 47065 }, { "epoch": 3.19, "learning_rate": 2.237095132159421e-06, "loss": 0.365, "step": 47070 }, { "epoch": 3.19, "learning_rate": 2.2361524829380493e-06, "loss": 0.2635, "step": 47075 }, { "epoch": 3.2, "learning_rate": 2.2352098337166773e-06, "loss": 0.2284, "step": 47080 }, { "epoch": 3.2, "learning_rate": 2.234267184495306e-06, "loss": 0.6261, "step": 47085 }, { "epoch": 3.2, "learning_rate": 2.2333245352739343e-06, "loss": 0.5188, "step": 47090 }, { "epoch": 3.2, "learning_rate": 2.2323818860525624e-06, "loss": 0.45, "step": 47095 }, { "epoch": 3.2, "learning_rate": 2.2314392368311904e-06, "loss": 0.2165, "step": 47100 }, { "epoch": 3.2, "learning_rate": 2.230496587609819e-06, "loss": 0.1178, "step": 47105 }, { "epoch": 3.2, "learning_rate": 2.229553938388447e-06, "loss": 0.2352, "step": 47110 }, { "epoch": 3.2, "learning_rate": 2.2286112891670755e-06, "loss": 0.107, "step": 47115 }, { "epoch": 3.2, "learning_rate": 2.2276686399457035e-06, "loss": 0.281, "step": 47120 }, { "epoch": 3.2, "learning_rate": 2.226725990724332e-06, "loss": 0.202, "step": 47125 }, { "epoch": 3.2, "learning_rate": 2.22578334150296e-06, "loss": 0.0805, "step": 47130 }, { "epoch": 3.2, "learning_rate": 2.224840692281588e-06, "loss": 0.1501, "step": 47135 }, { "epoch": 3.2, "learning_rate": 2.2238980430602166e-06, "loss": 0.2946, "step": 47140 }, { "epoch": 3.2, "learning_rate": 2.2229553938388447e-06, "loss": 0.6768, "step": 47145 }, { "epoch": 3.2, "learning_rate": 2.222012744617473e-06, "loss": 0.0596, "step": 47150 }, { "epoch": 3.2, "learning_rate": 2.2210700953961017e-06, "loss": 0.1023, "step": 47155 }, { "epoch": 3.2, "learning_rate": 2.2201274461747297e-06, "loss": 0.1307, "step": 47160 }, { "epoch": 3.2, "learning_rate": 2.219184796953358e-06, "loss": 0.2138, "step": 47165 }, { "epoch": 3.2, "learning_rate": 2.2182421477319863e-06, "loss": 0.0933, "step": 47170 }, { "epoch": 3.2, "learning_rate": 2.2172994985106143e-06, "loss": 0.1383, "step": 47175 }, { "epoch": 3.2, "learning_rate": 2.2163568492892424e-06, "loss": 0.1971, "step": 47180 }, { "epoch": 3.2, "learning_rate": 2.215414200067871e-06, "loss": 0.3644, "step": 47185 }, { "epoch": 3.2, "learning_rate": 2.2144715508464994e-06, "loss": 0.3138, "step": 47190 }, { "epoch": 3.2, "learning_rate": 2.2135289016251274e-06, "loss": 0.6291, "step": 47195 }, { "epoch": 3.2, "learning_rate": 2.2125862524037555e-06, "loss": 0.4221, "step": 47200 }, { "epoch": 3.2, "learning_rate": 2.211643603182384e-06, "loss": 0.3063, "step": 47205 }, { "epoch": 3.2, "learning_rate": 2.210700953961012e-06, "loss": 0.1117, "step": 47210 }, { "epoch": 3.2, "learning_rate": 2.2097583047396405e-06, "loss": 0.3362, "step": 47215 }, { "epoch": 3.2, "learning_rate": 2.2088156555182686e-06, "loss": 0.1964, "step": 47220 }, { "epoch": 3.21, "learning_rate": 2.207873006296897e-06, "loss": 0.3718, "step": 47225 }, { "epoch": 3.21, "learning_rate": 2.206930357075525e-06, "loss": 0.1532, "step": 47230 }, { "epoch": 3.21, "learning_rate": 2.2059877078541537e-06, "loss": 0.178, "step": 47235 }, { "epoch": 3.21, "learning_rate": 2.2050450586327817e-06, "loss": 0.1566, "step": 47240 }, { "epoch": 3.21, "learning_rate": 2.2041024094114098e-06, "loss": 0.3862, "step": 47245 }, { "epoch": 3.21, "learning_rate": 2.2031597601900383e-06, "loss": 0.057, "step": 47250 }, { "epoch": 3.21, "learning_rate": 2.2022171109686668e-06, "loss": 0.0512, "step": 47255 }, { "epoch": 3.21, "learning_rate": 2.201274461747295e-06, "loss": 0.2572, "step": 47260 }, { "epoch": 3.21, "learning_rate": 2.200331812525923e-06, "loss": 0.1118, "step": 47265 }, { "epoch": 3.21, "learning_rate": 2.1993891633045514e-06, "loss": 0.253, "step": 47270 }, { "epoch": 3.21, "learning_rate": 2.1984465140831794e-06, "loss": 0.0318, "step": 47275 }, { "epoch": 3.21, "learning_rate": 2.197503864861808e-06, "loss": 0.1452, "step": 47280 }, { "epoch": 3.21, "learning_rate": 2.196561215640436e-06, "loss": 0.2721, "step": 47285 }, { "epoch": 3.21, "learning_rate": 2.1956185664190645e-06, "loss": 0.2407, "step": 47290 }, { "epoch": 3.21, "learning_rate": 2.1946759171976925e-06, "loss": 0.2628, "step": 47295 }, { "epoch": 3.21, "learning_rate": 2.193733267976321e-06, "loss": 0.3246, "step": 47300 }, { "epoch": 3.21, "learning_rate": 2.192790618754949e-06, "loss": 0.3422, "step": 47305 }, { "epoch": 3.21, "learning_rate": 2.191847969533577e-06, "loss": 0.3464, "step": 47310 }, { "epoch": 3.21, "learning_rate": 2.1909053203122056e-06, "loss": 0.0516, "step": 47315 }, { "epoch": 3.21, "learning_rate": 2.189962671090834e-06, "loss": 0.3127, "step": 47320 }, { "epoch": 3.21, "learning_rate": 2.189020021869462e-06, "loss": 0.4221, "step": 47325 }, { "epoch": 3.21, "learning_rate": 2.1880773726480903e-06, "loss": 0.0185, "step": 47330 }, { "epoch": 3.21, "learning_rate": 2.1871347234267187e-06, "loss": 0.366, "step": 47335 }, { "epoch": 3.21, "learning_rate": 2.186192074205347e-06, "loss": 0.4412, "step": 47340 }, { "epoch": 3.21, "learning_rate": 2.185249424983975e-06, "loss": 0.3237, "step": 47345 }, { "epoch": 3.21, "learning_rate": 2.1843067757626034e-06, "loss": 0.3145, "step": 47350 }, { "epoch": 3.21, "learning_rate": 2.183364126541232e-06, "loss": 0.1601, "step": 47355 }, { "epoch": 3.21, "learning_rate": 2.18242147731986e-06, "loss": 0.0508, "step": 47360 }, { "epoch": 3.21, "learning_rate": 2.181478828098488e-06, "loss": 0.5221, "step": 47365 }, { "epoch": 3.22, "learning_rate": 2.1805361788771165e-06, "loss": 0.7554, "step": 47370 }, { "epoch": 3.22, "learning_rate": 2.1795935296557445e-06, "loss": 0.2475, "step": 47375 }, { "epoch": 3.22, "learning_rate": 2.178650880434373e-06, "loss": 0.6646, "step": 47380 }, { "epoch": 3.22, "learning_rate": 2.1777082312130015e-06, "loss": 0.196, "step": 47385 }, { "epoch": 3.22, "learning_rate": 2.1767655819916296e-06, "loss": 0.3379, "step": 47390 }, { "epoch": 3.22, "learning_rate": 2.1758229327702576e-06, "loss": 0.4224, "step": 47395 }, { "epoch": 3.22, "learning_rate": 2.174880283548886e-06, "loss": 0.5476, "step": 47400 }, { "epoch": 3.22, "learning_rate": 2.173937634327514e-06, "loss": 0.1497, "step": 47405 }, { "epoch": 3.22, "learning_rate": 2.1729949851061422e-06, "loss": 0.4228, "step": 47410 }, { "epoch": 3.22, "learning_rate": 2.1720523358847707e-06, "loss": 0.2041, "step": 47415 }, { "epoch": 3.22, "learning_rate": 2.171109686663399e-06, "loss": 0.2347, "step": 47420 }, { "epoch": 3.22, "learning_rate": 2.1701670374420273e-06, "loss": 0.0872, "step": 47425 }, { "epoch": 3.22, "learning_rate": 2.1692243882206553e-06, "loss": 0.2494, "step": 47430 }, { "epoch": 3.22, "learning_rate": 2.168281738999284e-06, "loss": 0.2657, "step": 47435 }, { "epoch": 3.22, "learning_rate": 2.167339089777912e-06, "loss": 0.4244, "step": 47440 }, { "epoch": 3.22, "learning_rate": 2.1663964405565404e-06, "loss": 0.0891, "step": 47445 }, { "epoch": 3.22, "learning_rate": 2.1654537913351684e-06, "loss": 0.2505, "step": 47450 }, { "epoch": 3.22, "learning_rate": 2.164511142113797e-06, "loss": 0.373, "step": 47455 }, { "epoch": 3.22, "learning_rate": 2.163568492892425e-06, "loss": 0.6066, "step": 47460 }, { "epoch": 3.22, "learning_rate": 2.1626258436710535e-06, "loss": 0.1036, "step": 47465 }, { "epoch": 3.22, "learning_rate": 2.1616831944496815e-06, "loss": 0.6706, "step": 47470 }, { "epoch": 3.22, "learning_rate": 2.1607405452283096e-06, "loss": 0.1795, "step": 47475 }, { "epoch": 3.22, "learning_rate": 2.159797896006938e-06, "loss": 0.7042, "step": 47480 }, { "epoch": 3.22, "learning_rate": 2.1588552467855666e-06, "loss": 0.411, "step": 47485 }, { "epoch": 3.22, "learning_rate": 2.1579125975641946e-06, "loss": 0.3683, "step": 47490 }, { "epoch": 3.22, "learning_rate": 2.1569699483428227e-06, "loss": 0.0628, "step": 47495 }, { "epoch": 3.22, "learning_rate": 2.156027299121451e-06, "loss": 0.0832, "step": 47500 }, { "epoch": 3.22, "learning_rate": 2.1550846499000793e-06, "loss": 0.1211, "step": 47505 }, { "epoch": 3.22, "learning_rate": 2.1541420006787077e-06, "loss": 0.3865, "step": 47510 }, { "epoch": 3.22, "learning_rate": 2.153199351457336e-06, "loss": 0.3464, "step": 47515 }, { "epoch": 3.23, "learning_rate": 2.1522567022359643e-06, "loss": 0.4578, "step": 47520 }, { "epoch": 3.23, "learning_rate": 2.1513140530145924e-06, "loss": 0.1081, "step": 47525 }, { "epoch": 3.23, "learning_rate": 2.1503714037932204e-06, "loss": 0.2959, "step": 47530 }, { "epoch": 3.23, "learning_rate": 2.149428754571849e-06, "loss": 0.4799, "step": 47535 }, { "epoch": 3.23, "learning_rate": 2.148486105350477e-06, "loss": 0.29, "step": 47540 }, { "epoch": 3.23, "learning_rate": 2.1475434561291055e-06, "loss": 0.446, "step": 47545 }, { "epoch": 3.23, "learning_rate": 2.146600806907734e-06, "loss": 0.419, "step": 47550 }, { "epoch": 3.23, "learning_rate": 2.145658157686362e-06, "loss": 0.135, "step": 47555 }, { "epoch": 3.23, "learning_rate": 2.14471550846499e-06, "loss": 0.4123, "step": 47560 }, { "epoch": 3.23, "learning_rate": 2.1437728592436186e-06, "loss": 0.4658, "step": 47565 }, { "epoch": 3.23, "learning_rate": 2.1428302100222466e-06, "loss": 0.0911, "step": 47570 }, { "epoch": 3.23, "learning_rate": 2.1418875608008747e-06, "loss": 0.0586, "step": 47575 }, { "epoch": 3.23, "learning_rate": 2.140944911579503e-06, "loss": 0.2399, "step": 47580 }, { "epoch": 3.23, "learning_rate": 2.1400022623581317e-06, "loss": 0.435, "step": 47585 }, { "epoch": 3.23, "learning_rate": 2.1390596131367597e-06, "loss": 0.1252, "step": 47590 }, { "epoch": 3.23, "learning_rate": 2.138116963915388e-06, "loss": 0.8035, "step": 47595 }, { "epoch": 3.23, "learning_rate": 2.1371743146940163e-06, "loss": 0.7262, "step": 47600 }, { "epoch": 3.23, "learning_rate": 2.1362316654726443e-06, "loss": 0.195, "step": 47605 }, { "epoch": 3.23, "learning_rate": 2.135289016251273e-06, "loss": 0.2747, "step": 47610 }, { "epoch": 3.23, "learning_rate": 2.1343463670299013e-06, "loss": 0.2157, "step": 47615 }, { "epoch": 3.23, "learning_rate": 2.1334037178085294e-06, "loss": 0.3673, "step": 47620 }, { "epoch": 3.23, "learning_rate": 2.1324610685871575e-06, "loss": 0.5888, "step": 47625 }, { "epoch": 3.23, "learning_rate": 2.131518419365786e-06, "loss": 0.0893, "step": 47630 }, { "epoch": 3.23, "learning_rate": 2.130575770144414e-06, "loss": 0.5778, "step": 47635 }, { "epoch": 3.23, "learning_rate": 2.129633120923042e-06, "loss": 0.0294, "step": 47640 }, { "epoch": 3.23, "learning_rate": 2.1286904717016706e-06, "loss": 0.3374, "step": 47645 }, { "epoch": 3.23, "learning_rate": 2.127747822480299e-06, "loss": 0.2338, "step": 47650 }, { "epoch": 3.23, "learning_rate": 2.126805173258927e-06, "loss": 0.9482, "step": 47655 }, { "epoch": 3.23, "learning_rate": 2.125862524037555e-06, "loss": 0.1358, "step": 47660 }, { "epoch": 3.24, "learning_rate": 2.1249198748161837e-06, "loss": 0.3974, "step": 47665 }, { "epoch": 3.24, "learning_rate": 2.1239772255948117e-06, "loss": 0.2252, "step": 47670 }, { "epoch": 3.24, "learning_rate": 2.12303457637344e-06, "loss": 0.1143, "step": 47675 }, { "epoch": 3.24, "learning_rate": 2.1220919271520683e-06, "loss": 0.4625, "step": 47680 }, { "epoch": 3.24, "learning_rate": 2.1211492779306968e-06, "loss": 0.274, "step": 47685 }, { "epoch": 3.24, "learning_rate": 2.120206628709325e-06, "loss": 0.0684, "step": 47690 }, { "epoch": 3.24, "learning_rate": 2.119263979487953e-06, "loss": 0.1942, "step": 47695 }, { "epoch": 3.24, "learning_rate": 2.1183213302665814e-06, "loss": 0.1434, "step": 47700 }, { "epoch": 3.24, "learning_rate": 2.1173786810452094e-06, "loss": 0.4326, "step": 47705 }, { "epoch": 3.24, "learning_rate": 2.116436031823838e-06, "loss": 0.3829, "step": 47710 }, { "epoch": 3.24, "learning_rate": 2.1154933826024664e-06, "loss": 0.5175, "step": 47715 }, { "epoch": 3.24, "learning_rate": 2.1145507333810945e-06, "loss": 0.3527, "step": 47720 }, { "epoch": 3.24, "learning_rate": 2.1136080841597225e-06, "loss": 0.1419, "step": 47725 }, { "epoch": 3.24, "learning_rate": 2.112665434938351e-06, "loss": 0.8455, "step": 47730 }, { "epoch": 3.24, "learning_rate": 2.111722785716979e-06, "loss": 0.1751, "step": 47735 }, { "epoch": 3.24, "learning_rate": 2.1107801364956076e-06, "loss": 0.8139, "step": 47740 }, { "epoch": 3.24, "learning_rate": 2.1098374872742356e-06, "loss": 0.4176, "step": 47745 }, { "epoch": 3.24, "learning_rate": 2.108894838052864e-06, "loss": 0.1199, "step": 47750 }, { "epoch": 3.24, "learning_rate": 2.107952188831492e-06, "loss": 0.0503, "step": 47755 }, { "epoch": 3.24, "learning_rate": 2.1070095396101203e-06, "loss": 0.199, "step": 47760 }, { "epoch": 3.24, "learning_rate": 2.1060668903887487e-06, "loss": 0.4947, "step": 47765 }, { "epoch": 3.24, "learning_rate": 2.105124241167377e-06, "loss": 0.5458, "step": 47770 }, { "epoch": 3.24, "learning_rate": 2.1041815919460053e-06, "loss": 0.2839, "step": 47775 }, { "epoch": 3.24, "learning_rate": 2.1032389427246338e-06, "loss": 0.2451, "step": 47780 }, { "epoch": 3.24, "learning_rate": 2.102296293503262e-06, "loss": 0.3163, "step": 47785 }, { "epoch": 3.24, "learning_rate": 2.10135364428189e-06, "loss": 0.1015, "step": 47790 }, { "epoch": 3.24, "learning_rate": 2.1004109950605184e-06, "loss": 0.1593, "step": 47795 }, { "epoch": 3.24, "learning_rate": 2.0994683458391465e-06, "loss": 0.2541, "step": 47800 }, { "epoch": 3.24, "learning_rate": 2.0985256966177745e-06, "loss": 0.1743, "step": 47805 }, { "epoch": 3.24, "learning_rate": 2.097583047396403e-06, "loss": 0.1748, "step": 47810 }, { "epoch": 3.25, "learning_rate": 2.0966403981750315e-06, "loss": 0.6749, "step": 47815 }, { "epoch": 3.25, "learning_rate": 2.0956977489536596e-06, "loss": 0.5514, "step": 47820 }, { "epoch": 3.25, "learning_rate": 2.0947550997322876e-06, "loss": 0.448, "step": 47825 }, { "epoch": 3.25, "learning_rate": 2.093812450510916e-06, "loss": 0.2796, "step": 47830 }, { "epoch": 3.25, "learning_rate": 2.092869801289544e-06, "loss": 0.7717, "step": 47835 }, { "epoch": 3.25, "learning_rate": 2.0919271520681727e-06, "loss": 0.1369, "step": 47840 }, { "epoch": 3.25, "learning_rate": 2.0909845028468007e-06, "loss": 0.2826, "step": 47845 }, { "epoch": 3.25, "learning_rate": 2.0900418536254292e-06, "loss": 0.0283, "step": 47850 }, { "epoch": 3.25, "learning_rate": 2.0890992044040573e-06, "loss": 0.3066, "step": 47855 }, { "epoch": 3.25, "learning_rate": 2.0881565551826853e-06, "loss": 0.2414, "step": 47860 }, { "epoch": 3.25, "learning_rate": 2.087213905961314e-06, "loss": 0.0892, "step": 47865 }, { "epoch": 3.25, "learning_rate": 2.086271256739942e-06, "loss": 0.0842, "step": 47870 }, { "epoch": 3.25, "learning_rate": 2.0853286075185704e-06, "loss": 0.2145, "step": 47875 }, { "epoch": 3.25, "learning_rate": 2.084385958297199e-06, "loss": 0.0847, "step": 47880 }, { "epoch": 3.25, "learning_rate": 2.083443309075827e-06, "loss": 0.1473, "step": 47885 }, { "epoch": 3.25, "learning_rate": 2.082500659854455e-06, "loss": 0.304, "step": 47890 }, { "epoch": 3.25, "learning_rate": 2.0815580106330835e-06, "loss": 0.2324, "step": 47895 }, { "epoch": 3.25, "learning_rate": 2.0806153614117115e-06, "loss": 0.1723, "step": 47900 }, { "epoch": 3.25, "learning_rate": 2.07967271219034e-06, "loss": 0.3195, "step": 47905 }, { "epoch": 3.25, "learning_rate": 2.078730062968968e-06, "loss": 0.3052, "step": 47910 }, { "epoch": 3.25, "learning_rate": 2.0777874137475966e-06, "loss": 0.25, "step": 47915 }, { "epoch": 3.25, "learning_rate": 2.0768447645262246e-06, "loss": 0.5637, "step": 47920 }, { "epoch": 3.25, "learning_rate": 2.0759021153048527e-06, "loss": 0.3505, "step": 47925 }, { "epoch": 3.25, "learning_rate": 2.074959466083481e-06, "loss": 0.1167, "step": 47930 }, { "epoch": 3.25, "learning_rate": 2.0740168168621093e-06, "loss": 0.1096, "step": 47935 }, { "epoch": 3.25, "learning_rate": 2.0730741676407378e-06, "loss": 0.3891, "step": 47940 }, { "epoch": 3.25, "learning_rate": 2.0721315184193662e-06, "loss": 0.2623, "step": 47945 }, { "epoch": 3.25, "learning_rate": 2.0711888691979943e-06, "loss": 0.2346, "step": 47950 }, { "epoch": 3.25, "learning_rate": 2.0702462199766224e-06, "loss": 0.3212, "step": 47955 }, { "epoch": 3.26, "learning_rate": 2.069303570755251e-06, "loss": 0.1938, "step": 47960 }, { "epoch": 3.26, "learning_rate": 2.068360921533879e-06, "loss": 0.1363, "step": 47965 }, { "epoch": 3.26, "learning_rate": 2.067418272312507e-06, "loss": 0.279, "step": 47970 }, { "epoch": 3.26, "learning_rate": 2.0664756230911355e-06, "loss": 0.412, "step": 47975 }, { "epoch": 3.26, "learning_rate": 2.065532973869764e-06, "loss": 0.2528, "step": 47980 }, { "epoch": 3.26, "learning_rate": 2.064590324648392e-06, "loss": 0.2288, "step": 47985 }, { "epoch": 3.26, "learning_rate": 2.06364767542702e-06, "loss": 0.3225, "step": 47990 }, { "epoch": 3.26, "learning_rate": 2.0627050262056486e-06, "loss": 0.2473, "step": 47995 }, { "epoch": 3.26, "learning_rate": 2.0617623769842766e-06, "loss": 0.1588, "step": 48000 }, { "epoch": 3.26, "learning_rate": 2.060819727762905e-06, "loss": 0.4253, "step": 48005 }, { "epoch": 3.26, "learning_rate": 2.0598770785415336e-06, "loss": 0.3115, "step": 48010 }, { "epoch": 3.26, "learning_rate": 2.0589344293201617e-06, "loss": 0.4208, "step": 48015 }, { "epoch": 3.26, "learning_rate": 2.0579917800987897e-06, "loss": 0.2296, "step": 48020 }, { "epoch": 3.26, "learning_rate": 2.0570491308774182e-06, "loss": 0.3163, "step": 48025 }, { "epoch": 3.26, "learning_rate": 2.0561064816560463e-06, "loss": 0.1906, "step": 48030 }, { "epoch": 3.26, "learning_rate": 2.0551638324346744e-06, "loss": 0.1848, "step": 48035 }, { "epoch": 3.26, "learning_rate": 2.054221183213303e-06, "loss": 0.2478, "step": 48040 }, { "epoch": 3.26, "learning_rate": 2.0532785339919313e-06, "loss": 0.2808, "step": 48045 }, { "epoch": 3.26, "learning_rate": 2.0523358847705594e-06, "loss": 0.5364, "step": 48050 }, { "epoch": 3.26, "learning_rate": 2.0513932355491875e-06, "loss": 0.2618, "step": 48055 }, { "epoch": 3.26, "learning_rate": 2.050450586327816e-06, "loss": 0.4551, "step": 48060 }, { "epoch": 3.26, "learning_rate": 2.049507937106444e-06, "loss": 0.2251, "step": 48065 }, { "epoch": 3.26, "learning_rate": 2.0485652878850725e-06, "loss": 0.1936, "step": 48070 }, { "epoch": 3.26, "learning_rate": 2.0476226386637006e-06, "loss": 0.1327, "step": 48075 }, { "epoch": 3.26, "learning_rate": 2.046679989442329e-06, "loss": 0.3615, "step": 48080 }, { "epoch": 3.26, "learning_rate": 2.045737340220957e-06, "loss": 0.5467, "step": 48085 }, { "epoch": 3.26, "learning_rate": 2.044794690999585e-06, "loss": 0.2411, "step": 48090 }, { "epoch": 3.26, "learning_rate": 2.0438520417782137e-06, "loss": 0.1447, "step": 48095 }, { "epoch": 3.26, "learning_rate": 2.0429093925568417e-06, "loss": 0.0336, "step": 48100 }, { "epoch": 3.26, "learning_rate": 2.04196674333547e-06, "loss": 0.0729, "step": 48105 }, { "epoch": 3.27, "learning_rate": 2.0410240941140987e-06, "loss": 0.502, "step": 48110 }, { "epoch": 3.27, "learning_rate": 2.0400814448927268e-06, "loss": 0.4047, "step": 48115 }, { "epoch": 3.27, "learning_rate": 2.039138795671355e-06, "loss": 0.0974, "step": 48120 }, { "epoch": 3.27, "learning_rate": 2.0381961464499833e-06, "loss": 0.5508, "step": 48125 }, { "epoch": 3.27, "learning_rate": 2.0372534972286114e-06, "loss": 0.6879, "step": 48130 }, { "epoch": 3.27, "learning_rate": 2.03631084800724e-06, "loss": 0.1361, "step": 48135 }, { "epoch": 3.27, "learning_rate": 2.035368198785868e-06, "loss": 0.1343, "step": 48140 }, { "epoch": 3.27, "learning_rate": 2.0344255495644964e-06, "loss": 0.4151, "step": 48145 }, { "epoch": 3.27, "learning_rate": 2.0334829003431245e-06, "loss": 0.4289, "step": 48150 }, { "epoch": 3.27, "learning_rate": 2.0325402511217525e-06, "loss": 0.4917, "step": 48155 }, { "epoch": 3.27, "learning_rate": 2.031597601900381e-06, "loss": 0.4623, "step": 48160 }, { "epoch": 3.27, "learning_rate": 2.030654952679009e-06, "loss": 0.2235, "step": 48165 }, { "epoch": 3.27, "learning_rate": 2.0297123034576376e-06, "loss": 0.3257, "step": 48170 }, { "epoch": 3.27, "learning_rate": 2.028769654236266e-06, "loss": 0.4676, "step": 48175 }, { "epoch": 3.27, "learning_rate": 2.027827005014894e-06, "loss": 0.2321, "step": 48180 }, { "epoch": 3.27, "learning_rate": 2.026884355793522e-06, "loss": 0.1273, "step": 48185 }, { "epoch": 3.27, "learning_rate": 2.0259417065721507e-06, "loss": 0.2125, "step": 48190 }, { "epoch": 3.27, "learning_rate": 2.0249990573507787e-06, "loss": 1.0617, "step": 48195 }, { "epoch": 3.27, "learning_rate": 2.024056408129407e-06, "loss": 0.5758, "step": 48200 }, { "epoch": 3.27, "learning_rate": 2.0231137589080353e-06, "loss": 0.3488, "step": 48205 }, { "epoch": 3.27, "learning_rate": 2.0221711096866638e-06, "loss": 0.4435, "step": 48210 }, { "epoch": 3.27, "learning_rate": 2.021228460465292e-06, "loss": 0.0956, "step": 48215 }, { "epoch": 3.27, "learning_rate": 2.02028581124392e-06, "loss": 0.309, "step": 48220 }, { "epoch": 3.27, "learning_rate": 2.0193431620225484e-06, "loss": 0.3213, "step": 48225 }, { "epoch": 3.27, "learning_rate": 2.0184005128011765e-06, "loss": 0.1697, "step": 48230 }, { "epoch": 3.27, "learning_rate": 2.017457863579805e-06, "loss": 0.3973, "step": 48235 }, { "epoch": 3.27, "learning_rate": 2.0165152143584334e-06, "loss": 0.245, "step": 48240 }, { "epoch": 3.27, "learning_rate": 2.0155725651370615e-06, "loss": 0.0999, "step": 48245 }, { "epoch": 3.27, "learning_rate": 2.0146299159156896e-06, "loss": 0.9817, "step": 48250 }, { "epoch": 3.28, "learning_rate": 2.0136872666943176e-06, "loss": 0.1308, "step": 48255 }, { "epoch": 3.28, "learning_rate": 2.012744617472946e-06, "loss": 0.3288, "step": 48260 }, { "epoch": 3.28, "learning_rate": 2.011801968251574e-06, "loss": 0.2246, "step": 48265 }, { "epoch": 3.28, "learning_rate": 2.0108593190302027e-06, "loss": 0.1089, "step": 48270 }, { "epoch": 3.28, "learning_rate": 2.009916669808831e-06, "loss": 0.1341, "step": 48275 }, { "epoch": 3.28, "learning_rate": 2.0089740205874592e-06, "loss": 0.4569, "step": 48280 }, { "epoch": 3.28, "learning_rate": 2.0080313713660873e-06, "loss": 0.0459, "step": 48285 }, { "epoch": 3.28, "learning_rate": 2.0070887221447158e-06, "loss": 0.2922, "step": 48290 }, { "epoch": 3.28, "learning_rate": 2.006146072923344e-06, "loss": 0.0686, "step": 48295 }, { "epoch": 3.28, "learning_rate": 2.0052034237019723e-06, "loss": 0.1329, "step": 48300 }, { "epoch": 3.28, "learning_rate": 2.0042607744806004e-06, "loss": 0.1695, "step": 48305 }, { "epoch": 3.28, "learning_rate": 2.003318125259229e-06, "loss": 0.406, "step": 48310 }, { "epoch": 3.28, "learning_rate": 2.002375476037857e-06, "loss": 0.3809, "step": 48315 }, { "epoch": 3.28, "learning_rate": 2.001432826816485e-06, "loss": 0.2309, "step": 48320 }, { "epoch": 3.28, "learning_rate": 2.0004901775951135e-06, "loss": 0.1174, "step": 48325 }, { "epoch": 3.28, "learning_rate": 1.9995475283737415e-06, "loss": 0.2022, "step": 48330 }, { "epoch": 3.28, "learning_rate": 1.99860487915237e-06, "loss": 0.2596, "step": 48335 }, { "epoch": 3.28, "learning_rate": 1.9976622299309985e-06, "loss": 0.3424, "step": 48340 }, { "epoch": 3.28, "learning_rate": 1.9967195807096266e-06, "loss": 0.3713, "step": 48345 }, { "epoch": 3.28, "learning_rate": 1.9957769314882547e-06, "loss": 0.4874, "step": 48350 }, { "epoch": 3.28, "learning_rate": 1.994834282266883e-06, "loss": 0.2412, "step": 48355 }, { "epoch": 3.28, "learning_rate": 1.993891633045511e-06, "loss": 0.1617, "step": 48360 }, { "epoch": 3.28, "learning_rate": 1.9929489838241397e-06, "loss": 0.1458, "step": 48365 }, { "epoch": 3.28, "learning_rate": 1.9920063346027678e-06, "loss": 0.8794, "step": 48370 }, { "epoch": 3.28, "learning_rate": 1.9910636853813962e-06, "loss": 0.1795, "step": 48375 }, { "epoch": 3.28, "learning_rate": 1.9901210361600243e-06, "loss": 0.1429, "step": 48380 }, { "epoch": 3.28, "learning_rate": 1.9891783869386524e-06, "loss": 0.2326, "step": 48385 }, { "epoch": 3.28, "learning_rate": 1.988235737717281e-06, "loss": 0.3014, "step": 48390 }, { "epoch": 3.28, "learning_rate": 1.987293088495909e-06, "loss": 0.2048, "step": 48395 }, { "epoch": 3.28, "learning_rate": 1.9863504392745374e-06, "loss": 0.0221, "step": 48400 }, { "epoch": 3.29, "learning_rate": 1.985407790053166e-06, "loss": 0.299, "step": 48405 }, { "epoch": 3.29, "learning_rate": 1.984465140831794e-06, "loss": 0.1205, "step": 48410 }, { "epoch": 3.29, "learning_rate": 1.983522491610422e-06, "loss": 0.7553, "step": 48415 }, { "epoch": 3.29, "learning_rate": 1.98257984238905e-06, "loss": 0.0965, "step": 48420 }, { "epoch": 3.29, "learning_rate": 1.9816371931676786e-06, "loss": 0.2303, "step": 48425 }, { "epoch": 3.29, "learning_rate": 1.9806945439463066e-06, "loss": 0.1795, "step": 48430 }, { "epoch": 3.29, "learning_rate": 1.979751894724935e-06, "loss": 0.3276, "step": 48435 }, { "epoch": 3.29, "learning_rate": 1.9788092455035636e-06, "loss": 0.1541, "step": 48440 }, { "epoch": 3.29, "learning_rate": 1.9778665962821917e-06, "loss": 0.1473, "step": 48445 }, { "epoch": 3.29, "learning_rate": 1.9769239470608197e-06, "loss": 0.1744, "step": 48450 }, { "epoch": 3.29, "learning_rate": 1.9759812978394482e-06, "loss": 0.1405, "step": 48455 }, { "epoch": 3.29, "learning_rate": 1.9750386486180763e-06, "loss": 0.216, "step": 48460 }, { "epoch": 3.29, "learning_rate": 1.9740959993967048e-06, "loss": 0.198, "step": 48465 }, { "epoch": 3.29, "learning_rate": 1.973153350175333e-06, "loss": 0.3163, "step": 48470 }, { "epoch": 3.29, "learning_rate": 1.9722107009539613e-06, "loss": 0.0253, "step": 48475 }, { "epoch": 3.29, "learning_rate": 1.9712680517325894e-06, "loss": 0.2299, "step": 48480 }, { "epoch": 3.29, "learning_rate": 1.9703254025112175e-06, "loss": 0.2756, "step": 48485 }, { "epoch": 3.29, "learning_rate": 1.969382753289846e-06, "loss": 0.0903, "step": 48490 }, { "epoch": 3.29, "learning_rate": 1.968440104068474e-06, "loss": 0.2221, "step": 48495 }, { "epoch": 3.29, "learning_rate": 1.9674974548471025e-06, "loss": 0.4716, "step": 48500 }, { "epoch": 3.29, "learning_rate": 1.966554805625731e-06, "loss": 0.2263, "step": 48505 }, { "epoch": 3.29, "learning_rate": 1.965612156404359e-06, "loss": 0.5982, "step": 48510 }, { "epoch": 3.29, "learning_rate": 1.964669507182987e-06, "loss": 0.2447, "step": 48515 }, { "epoch": 3.29, "learning_rate": 1.9637268579616156e-06, "loss": 0.132, "step": 48520 }, { "epoch": 3.29, "learning_rate": 1.9627842087402437e-06, "loss": 0.3045, "step": 48525 }, { "epoch": 3.29, "learning_rate": 1.961841559518872e-06, "loss": 0.2796, "step": 48530 }, { "epoch": 3.29, "learning_rate": 1.9608989102975e-06, "loss": 0.3105, "step": 48535 }, { "epoch": 3.29, "learning_rate": 1.9599562610761287e-06, "loss": 0.3136, "step": 48540 }, { "epoch": 3.29, "learning_rate": 1.9590136118547568e-06, "loss": 0.1076, "step": 48545 }, { "epoch": 3.3, "learning_rate": 1.958070962633385e-06, "loss": 0.1093, "step": 48550 }, { "epoch": 3.3, "learning_rate": 1.9571283134120133e-06, "loss": 0.2448, "step": 48555 }, { "epoch": 3.3, "learning_rate": 1.9561856641906414e-06, "loss": 0.1493, "step": 48560 }, { "epoch": 3.3, "learning_rate": 1.95524301496927e-06, "loss": 0.0692, "step": 48565 }, { "epoch": 3.3, "learning_rate": 1.9543003657478984e-06, "loss": 0.0964, "step": 48570 }, { "epoch": 3.3, "learning_rate": 1.9533577165265264e-06, "loss": 0.4785, "step": 48575 }, { "epoch": 3.3, "learning_rate": 1.9524150673051545e-06, "loss": 0.0742, "step": 48580 }, { "epoch": 3.3, "learning_rate": 1.951472418083783e-06, "loss": 0.1981, "step": 48585 }, { "epoch": 3.3, "learning_rate": 1.950529768862411e-06, "loss": 0.2846, "step": 48590 }, { "epoch": 3.3, "learning_rate": 1.949587119641039e-06, "loss": 0.2724, "step": 48595 }, { "epoch": 3.3, "learning_rate": 1.9486444704196676e-06, "loss": 0.2298, "step": 48600 }, { "epoch": 3.3, "learning_rate": 1.947701821198296e-06, "loss": 0.261, "step": 48605 }, { "epoch": 3.3, "learning_rate": 1.946759171976924e-06, "loss": 0.1264, "step": 48610 }, { "epoch": 3.3, "learning_rate": 1.945816522755552e-06, "loss": 0.2412, "step": 48615 }, { "epoch": 3.3, "learning_rate": 1.9448738735341807e-06, "loss": 0.1758, "step": 48620 }, { "epoch": 3.3, "learning_rate": 1.9439312243128087e-06, "loss": 0.1163, "step": 48625 }, { "epoch": 3.3, "learning_rate": 1.9429885750914372e-06, "loss": 0.1472, "step": 48630 }, { "epoch": 3.3, "learning_rate": 1.9420459258700657e-06, "loss": 0.1372, "step": 48635 }, { "epoch": 3.3, "learning_rate": 1.9411032766486938e-06, "loss": 0.5421, "step": 48640 }, { "epoch": 3.3, "learning_rate": 1.940160627427322e-06, "loss": 0.1958, "step": 48645 }, { "epoch": 3.3, "learning_rate": 1.93921797820595e-06, "loss": 0.3117, "step": 48650 }, { "epoch": 3.3, "learning_rate": 1.9382753289845784e-06, "loss": 0.1234, "step": 48655 }, { "epoch": 3.3, "learning_rate": 1.9373326797632065e-06, "loss": 0.5428, "step": 48660 }, { "epoch": 3.3, "learning_rate": 1.936390030541835e-06, "loss": 0.1026, "step": 48665 }, { "epoch": 3.3, "learning_rate": 1.9354473813204634e-06, "loss": 0.4545, "step": 48670 }, { "epoch": 3.3, "learning_rate": 1.9345047320990915e-06, "loss": 0.1486, "step": 48675 }, { "epoch": 3.3, "learning_rate": 1.9335620828777196e-06, "loss": 0.2124, "step": 48680 }, { "epoch": 3.3, "learning_rate": 1.932619433656348e-06, "loss": 0.2651, "step": 48685 }, { "epoch": 3.3, "learning_rate": 1.931676784434976e-06, "loss": 0.4355, "step": 48690 }, { "epoch": 3.3, "learning_rate": 1.9307341352136046e-06, "loss": 0.1873, "step": 48695 }, { "epoch": 3.31, "learning_rate": 1.9297914859922327e-06, "loss": 0.5885, "step": 48700 }, { "epoch": 3.31, "learning_rate": 1.928848836770861e-06, "loss": 0.0911, "step": 48705 }, { "epoch": 3.31, "learning_rate": 1.9279061875494892e-06, "loss": 0.1669, "step": 48710 }, { "epoch": 3.31, "learning_rate": 1.9269635383281173e-06, "loss": 0.152, "step": 48715 }, { "epoch": 3.31, "learning_rate": 1.9260208891067458e-06, "loss": 0.1734, "step": 48720 }, { "epoch": 3.31, "learning_rate": 1.925078239885374e-06, "loss": 0.0881, "step": 48725 }, { "epoch": 3.31, "learning_rate": 1.9241355906640023e-06, "loss": 0.1461, "step": 48730 }, { "epoch": 3.31, "learning_rate": 1.923192941442631e-06, "loss": 0.1678, "step": 48735 }, { "epoch": 3.31, "learning_rate": 1.922250292221259e-06, "loss": 0.2226, "step": 48740 }, { "epoch": 3.31, "learning_rate": 1.921307642999887e-06, "loss": 0.0679, "step": 48745 }, { "epoch": 3.31, "learning_rate": 1.9203649937785154e-06, "loss": 0.1458, "step": 48750 }, { "epoch": 3.31, "learning_rate": 1.9194223445571435e-06, "loss": 0.2377, "step": 48755 }, { "epoch": 3.31, "learning_rate": 1.918479695335772e-06, "loss": 0.2898, "step": 48760 }, { "epoch": 3.31, "learning_rate": 1.9175370461144e-06, "loss": 0.2841, "step": 48765 }, { "epoch": 3.31, "learning_rate": 1.9165943968930285e-06, "loss": 0.0972, "step": 48770 }, { "epoch": 3.31, "learning_rate": 1.9156517476716566e-06, "loss": 0.3791, "step": 48775 }, { "epoch": 3.31, "learning_rate": 1.9147090984502847e-06, "loss": 0.3714, "step": 48780 }, { "epoch": 3.31, "learning_rate": 1.913766449228913e-06, "loss": 0.1122, "step": 48785 }, { "epoch": 3.31, "learning_rate": 1.912823800007541e-06, "loss": 0.4114, "step": 48790 }, { "epoch": 3.31, "learning_rate": 1.9118811507861697e-06, "loss": 0.4521, "step": 48795 }, { "epoch": 3.31, "learning_rate": 1.910938501564798e-06, "loss": 0.4708, "step": 48800 }, { "epoch": 3.31, "learning_rate": 1.9099958523434262e-06, "loss": 0.0596, "step": 48805 }, { "epoch": 3.31, "learning_rate": 1.9090532031220543e-06, "loss": 0.3607, "step": 48810 }, { "epoch": 3.31, "learning_rate": 1.9081105539006824e-06, "loss": 0.2646, "step": 48815 }, { "epoch": 3.31, "learning_rate": 1.9071679046793109e-06, "loss": 0.3527, "step": 48820 }, { "epoch": 3.31, "learning_rate": 1.9062252554579391e-06, "loss": 0.3921, "step": 48825 }, { "epoch": 3.31, "learning_rate": 1.9052826062365672e-06, "loss": 0.1522, "step": 48830 }, { "epoch": 3.31, "learning_rate": 1.9043399570151957e-06, "loss": 0.2291, "step": 48835 }, { "epoch": 3.31, "learning_rate": 1.903397307793824e-06, "loss": 0.4481, "step": 48840 }, { "epoch": 3.32, "learning_rate": 1.902454658572452e-06, "loss": 0.253, "step": 48845 }, { "epoch": 3.32, "learning_rate": 1.9015120093510805e-06, "loss": 0.1304, "step": 48850 }, { "epoch": 3.32, "learning_rate": 1.9005693601297088e-06, "loss": 0.6102, "step": 48855 }, { "epoch": 3.32, "learning_rate": 1.8996267109083369e-06, "loss": 0.2376, "step": 48860 }, { "epoch": 3.32, "learning_rate": 1.8986840616869651e-06, "loss": 0.3666, "step": 48865 }, { "epoch": 3.32, "learning_rate": 1.8977414124655936e-06, "loss": 0.0795, "step": 48870 }, { "epoch": 3.32, "learning_rate": 1.8967987632442217e-06, "loss": 0.4033, "step": 48875 }, { "epoch": 3.32, "learning_rate": 1.89585611402285e-06, "loss": 0.1797, "step": 48880 }, { "epoch": 3.32, "learning_rate": 1.8949134648014782e-06, "loss": 0.7549, "step": 48885 }, { "epoch": 3.32, "learning_rate": 1.8939708155801065e-06, "loss": 0.3267, "step": 48890 }, { "epoch": 3.32, "learning_rate": 1.8930281663587346e-06, "loss": 0.1987, "step": 48895 }, { "epoch": 3.32, "learning_rate": 1.892085517137363e-06, "loss": 0.2035, "step": 48900 }, { "epoch": 3.32, "learning_rate": 1.8911428679159913e-06, "loss": 0.3079, "step": 48905 }, { "epoch": 3.32, "learning_rate": 1.8902002186946194e-06, "loss": 0.2606, "step": 48910 }, { "epoch": 3.32, "learning_rate": 1.8892575694732479e-06, "loss": 0.0988, "step": 48915 }, { "epoch": 3.32, "learning_rate": 1.8883149202518762e-06, "loss": 0.7423, "step": 48920 }, { "epoch": 3.32, "learning_rate": 1.8873722710305042e-06, "loss": 0.0723, "step": 48925 }, { "epoch": 3.32, "learning_rate": 1.8864296218091325e-06, "loss": 0.3496, "step": 48930 }, { "epoch": 3.32, "learning_rate": 1.8854869725877608e-06, "loss": 0.1347, "step": 48935 }, { "epoch": 3.32, "learning_rate": 1.884544323366389e-06, "loss": 0.6524, "step": 48940 }, { "epoch": 3.32, "learning_rate": 1.8836016741450171e-06, "loss": 0.1885, "step": 48945 }, { "epoch": 3.32, "learning_rate": 1.8826590249236456e-06, "loss": 0.1783, "step": 48950 }, { "epoch": 3.32, "learning_rate": 1.8817163757022739e-06, "loss": 0.3218, "step": 48955 }, { "epoch": 3.32, "learning_rate": 1.880773726480902e-06, "loss": 0.3734, "step": 48960 }, { "epoch": 3.32, "learning_rate": 1.8798310772595304e-06, "loss": 0.1239, "step": 48965 }, { "epoch": 3.32, "learning_rate": 1.8788884280381587e-06, "loss": 0.4685, "step": 48970 }, { "epoch": 3.32, "learning_rate": 1.8779457788167868e-06, "loss": 0.3152, "step": 48975 }, { "epoch": 3.32, "learning_rate": 1.877003129595415e-06, "loss": 0.5043, "step": 48980 }, { "epoch": 3.32, "learning_rate": 1.8760604803740435e-06, "loss": 0.7728, "step": 48985 }, { "epoch": 3.32, "learning_rate": 1.8751178311526716e-06, "loss": 0.2184, "step": 48990 }, { "epoch": 3.33, "learning_rate": 1.8741751819312999e-06, "loss": 0.7666, "step": 48995 }, { "epoch": 3.33, "learning_rate": 1.8732325327099281e-06, "loss": 0.0822, "step": 49000 }, { "epoch": 3.33, "learning_rate": 1.8722898834885564e-06, "loss": 0.152, "step": 49005 }, { "epoch": 3.33, "learning_rate": 1.8713472342671845e-06, "loss": 0.7214, "step": 49010 }, { "epoch": 3.33, "learning_rate": 1.870404585045813e-06, "loss": 0.2031, "step": 49015 }, { "epoch": 3.33, "learning_rate": 1.8694619358244412e-06, "loss": 0.1099, "step": 49020 }, { "epoch": 3.33, "learning_rate": 1.8685192866030693e-06, "loss": 0.2185, "step": 49025 }, { "epoch": 3.33, "learning_rate": 1.8675766373816978e-06, "loss": 0.3327, "step": 49030 }, { "epoch": 3.33, "learning_rate": 1.866633988160326e-06, "loss": 0.6008, "step": 49035 }, { "epoch": 3.33, "learning_rate": 1.8656913389389541e-06, "loss": 0.2848, "step": 49040 }, { "epoch": 3.33, "learning_rate": 1.8647486897175824e-06, "loss": 0.0578, "step": 49045 }, { "epoch": 3.33, "learning_rate": 1.8638060404962107e-06, "loss": 0.2581, "step": 49050 }, { "epoch": 3.33, "learning_rate": 1.862863391274839e-06, "loss": 0.586, "step": 49055 }, { "epoch": 3.33, "learning_rate": 1.861920742053467e-06, "loss": 0.5914, "step": 49060 }, { "epoch": 3.33, "learning_rate": 1.8609780928320955e-06, "loss": 0.1668, "step": 49065 }, { "epoch": 3.33, "learning_rate": 1.8600354436107238e-06, "loss": 0.2013, "step": 49070 }, { "epoch": 3.33, "learning_rate": 1.8590927943893519e-06, "loss": 0.5553, "step": 49075 }, { "epoch": 3.33, "learning_rate": 1.8581501451679803e-06, "loss": 0.603, "step": 49080 }, { "epoch": 3.33, "learning_rate": 1.8572074959466086e-06, "loss": 0.1894, "step": 49085 }, { "epoch": 3.33, "learning_rate": 1.8562648467252367e-06, "loss": 0.2334, "step": 49090 }, { "epoch": 3.33, "learning_rate": 1.855322197503865e-06, "loss": 0.24, "step": 49095 }, { "epoch": 3.33, "learning_rate": 1.8543795482824932e-06, "loss": 0.3759, "step": 49100 }, { "epoch": 3.33, "learning_rate": 1.8534368990611215e-06, "loss": 0.0809, "step": 49105 }, { "epoch": 3.33, "learning_rate": 1.8524942498397498e-06, "loss": 0.153, "step": 49110 }, { "epoch": 3.33, "learning_rate": 1.851551600618378e-06, "loss": 0.1534, "step": 49115 }, { "epoch": 3.33, "learning_rate": 1.8506089513970063e-06, "loss": 0.2915, "step": 49120 }, { "epoch": 3.33, "learning_rate": 1.8496663021756344e-06, "loss": 0.2337, "step": 49125 }, { "epoch": 3.33, "learning_rate": 1.8487236529542629e-06, "loss": 0.1468, "step": 49130 }, { "epoch": 3.33, "learning_rate": 1.8477810037328912e-06, "loss": 0.299, "step": 49135 }, { "epoch": 3.34, "learning_rate": 1.8468383545115192e-06, "loss": 0.1123, "step": 49140 }, { "epoch": 3.34, "learning_rate": 1.8458957052901475e-06, "loss": 0.2471, "step": 49145 }, { "epoch": 3.34, "learning_rate": 1.844953056068776e-06, "loss": 0.5031, "step": 49150 }, { "epoch": 3.34, "learning_rate": 1.844010406847404e-06, "loss": 0.3087, "step": 49155 }, { "epoch": 3.34, "learning_rate": 1.8430677576260323e-06, "loss": 0.3391, "step": 49160 }, { "epoch": 3.34, "learning_rate": 1.8421251084046606e-06, "loss": 0.7817, "step": 49165 }, { "epoch": 3.34, "learning_rate": 1.8411824591832889e-06, "loss": 0.2595, "step": 49170 }, { "epoch": 3.34, "learning_rate": 1.840239809961917e-06, "loss": 0.0673, "step": 49175 }, { "epoch": 3.34, "learning_rate": 1.8392971607405454e-06, "loss": 0.4073, "step": 49180 }, { "epoch": 3.34, "learning_rate": 1.8383545115191737e-06, "loss": 0.1293, "step": 49185 }, { "epoch": 3.34, "learning_rate": 1.8374118622978018e-06, "loss": 0.1813, "step": 49190 }, { "epoch": 3.34, "learning_rate": 1.8364692130764303e-06, "loss": 0.107, "step": 49195 }, { "epoch": 3.34, "learning_rate": 1.8355265638550585e-06, "loss": 0.1648, "step": 49200 }, { "epoch": 3.34, "learning_rate": 1.8345839146336866e-06, "loss": 0.1292, "step": 49205 }, { "epoch": 3.34, "learning_rate": 1.8336412654123149e-06, "loss": 0.2429, "step": 49210 }, { "epoch": 3.34, "learning_rate": 1.8326986161909431e-06, "loss": 0.1011, "step": 49215 }, { "epoch": 3.34, "learning_rate": 1.8317559669695714e-06, "loss": 0.3528, "step": 49220 }, { "epoch": 3.34, "learning_rate": 1.8308133177481997e-06, "loss": 0.2085, "step": 49225 }, { "epoch": 3.34, "learning_rate": 1.829870668526828e-06, "loss": 0.3868, "step": 49230 }, { "epoch": 3.34, "learning_rate": 1.8289280193054562e-06, "loss": 0.2315, "step": 49235 }, { "epoch": 3.34, "learning_rate": 1.8279853700840843e-06, "loss": 0.1071, "step": 49240 }, { "epoch": 3.34, "learning_rate": 1.8270427208627128e-06, "loss": 0.1716, "step": 49245 }, { "epoch": 3.34, "learning_rate": 1.826100071641341e-06, "loss": 0.1822, "step": 49250 }, { "epoch": 3.34, "learning_rate": 1.8251574224199691e-06, "loss": 0.3966, "step": 49255 }, { "epoch": 3.34, "learning_rate": 1.8242147731985974e-06, "loss": 0.5017, "step": 49260 }, { "epoch": 3.34, "learning_rate": 1.823272123977226e-06, "loss": 0.1884, "step": 49265 }, { "epoch": 3.34, "learning_rate": 1.822329474755854e-06, "loss": 0.1097, "step": 49270 }, { "epoch": 3.34, "learning_rate": 1.8213868255344822e-06, "loss": 0.1423, "step": 49275 }, { "epoch": 3.34, "learning_rate": 1.8204441763131105e-06, "loss": 0.3475, "step": 49280 }, { "epoch": 3.34, "learning_rate": 1.8195015270917388e-06, "loss": 0.2017, "step": 49285 }, { "epoch": 3.35, "learning_rate": 1.8185588778703669e-06, "loss": 0.043, "step": 49290 }, { "epoch": 3.35, "learning_rate": 1.8176162286489953e-06, "loss": 0.2314, "step": 49295 }, { "epoch": 3.35, "learning_rate": 1.8166735794276236e-06, "loss": 0.4082, "step": 49300 }, { "epoch": 3.35, "learning_rate": 1.8157309302062517e-06, "loss": 0.1755, "step": 49305 }, { "epoch": 3.35, "learning_rate": 1.8147882809848802e-06, "loss": 0.3938, "step": 49310 }, { "epoch": 3.35, "learning_rate": 1.8138456317635084e-06, "loss": 0.1718, "step": 49315 }, { "epoch": 3.35, "learning_rate": 1.8129029825421365e-06, "loss": 0.1431, "step": 49320 }, { "epoch": 3.35, "learning_rate": 1.8119603333207648e-06, "loss": 0.5111, "step": 49325 }, { "epoch": 3.35, "learning_rate": 1.811017684099393e-06, "loss": 0.0347, "step": 49330 }, { "epoch": 3.35, "learning_rate": 1.8100750348780213e-06, "loss": 0.1476, "step": 49335 }, { "epoch": 3.35, "learning_rate": 1.8091323856566494e-06, "loss": 0.577, "step": 49340 }, { "epoch": 3.35, "learning_rate": 1.8081897364352779e-06, "loss": 0.1312, "step": 49345 }, { "epoch": 3.35, "learning_rate": 1.8072470872139062e-06, "loss": 0.5682, "step": 49350 }, { "epoch": 3.35, "learning_rate": 1.8063044379925342e-06, "loss": 0.1342, "step": 49355 }, { "epoch": 3.35, "learning_rate": 1.8053617887711627e-06, "loss": 0.2359, "step": 49360 }, { "epoch": 3.35, "learning_rate": 1.804419139549791e-06, "loss": 0.2539, "step": 49365 }, { "epoch": 3.35, "learning_rate": 1.803476490328419e-06, "loss": 0.0438, "step": 49370 }, { "epoch": 3.35, "learning_rate": 1.8025338411070473e-06, "loss": 0.0851, "step": 49375 }, { "epoch": 3.35, "learning_rate": 1.8015911918856758e-06, "loss": 0.0786, "step": 49380 }, { "epoch": 3.35, "learning_rate": 1.8006485426643039e-06, "loss": 0.2986, "step": 49385 }, { "epoch": 3.35, "learning_rate": 1.7997058934429322e-06, "loss": 0.1489, "step": 49390 }, { "epoch": 3.35, "learning_rate": 1.7987632442215604e-06, "loss": 0.0641, "step": 49395 }, { "epoch": 3.35, "learning_rate": 1.7978205950001887e-06, "loss": 0.1777, "step": 49400 }, { "epoch": 3.35, "learning_rate": 1.7968779457788168e-06, "loss": 0.32, "step": 49405 }, { "epoch": 3.35, "learning_rate": 1.7959352965574453e-06, "loss": 0.1956, "step": 49410 }, { "epoch": 3.35, "learning_rate": 1.7949926473360735e-06, "loss": 0.0899, "step": 49415 }, { "epoch": 3.35, "learning_rate": 1.7940499981147016e-06, "loss": 0.0517, "step": 49420 }, { "epoch": 3.35, "learning_rate": 1.7931073488933299e-06, "loss": 0.3103, "step": 49425 }, { "epoch": 3.35, "learning_rate": 1.7921646996719584e-06, "loss": 0.1475, "step": 49430 }, { "epoch": 3.36, "learning_rate": 1.7912220504505864e-06, "loss": 0.1704, "step": 49435 }, { "epoch": 3.36, "learning_rate": 1.7902794012292147e-06, "loss": 0.2206, "step": 49440 }, { "epoch": 3.36, "learning_rate": 1.789336752007843e-06, "loss": 0.556, "step": 49445 }, { "epoch": 3.36, "learning_rate": 1.7883941027864712e-06, "loss": 0.1886, "step": 49450 }, { "epoch": 3.36, "learning_rate": 1.7874514535650993e-06, "loss": 0.3493, "step": 49455 }, { "epoch": 3.36, "learning_rate": 1.7865088043437278e-06, "loss": 0.3324, "step": 49460 }, { "epoch": 3.36, "learning_rate": 1.785566155122356e-06, "loss": 0.0742, "step": 49465 }, { "epoch": 3.36, "learning_rate": 1.7846235059009841e-06, "loss": 0.5949, "step": 49470 }, { "epoch": 3.36, "learning_rate": 1.7836808566796126e-06, "loss": 0.1474, "step": 49475 }, { "epoch": 3.36, "learning_rate": 1.782738207458241e-06, "loss": 0.1846, "step": 49480 }, { "epoch": 3.36, "learning_rate": 1.781795558236869e-06, "loss": 0.1147, "step": 49485 }, { "epoch": 3.36, "learning_rate": 1.7808529090154972e-06, "loss": 0.58, "step": 49490 }, { "epoch": 3.36, "learning_rate": 1.7799102597941257e-06, "loss": 0.3239, "step": 49495 }, { "epoch": 3.36, "learning_rate": 1.7789676105727538e-06, "loss": 0.5284, "step": 49500 }, { "epoch": 3.36, "learning_rate": 1.778024961351382e-06, "loss": 0.0677, "step": 49505 }, { "epoch": 3.36, "learning_rate": 1.7770823121300103e-06, "loss": 0.9343, "step": 49510 }, { "epoch": 3.36, "learning_rate": 1.7761396629086386e-06, "loss": 0.333, "step": 49515 }, { "epoch": 3.36, "learning_rate": 1.7751970136872667e-06, "loss": 0.3559, "step": 49520 }, { "epoch": 3.36, "learning_rate": 1.7742543644658952e-06, "loss": 0.5219, "step": 49525 }, { "epoch": 3.36, "learning_rate": 1.7733117152445234e-06, "loss": 0.0652, "step": 49530 }, { "epoch": 3.36, "learning_rate": 1.7723690660231515e-06, "loss": 0.1017, "step": 49535 }, { "epoch": 3.36, "learning_rate": 1.7714264168017798e-06, "loss": 0.1398, "step": 49540 }, { "epoch": 3.36, "learning_rate": 1.7704837675804083e-06, "loss": 0.3462, "step": 49545 }, { "epoch": 3.36, "learning_rate": 1.7695411183590363e-06, "loss": 0.3168, "step": 49550 }, { "epoch": 3.36, "learning_rate": 1.7685984691376646e-06, "loss": 0.15, "step": 49555 }, { "epoch": 3.36, "learning_rate": 1.7676558199162929e-06, "loss": 0.1218, "step": 49560 }, { "epoch": 3.36, "learning_rate": 1.7667131706949212e-06, "loss": 0.1348, "step": 49565 }, { "epoch": 3.36, "learning_rate": 1.7657705214735492e-06, "loss": 0.1199, "step": 49570 }, { "epoch": 3.36, "learning_rate": 1.7648278722521777e-06, "loss": 0.3698, "step": 49575 }, { "epoch": 3.37, "learning_rate": 1.763885223030806e-06, "loss": 0.0935, "step": 49580 }, { "epoch": 3.37, "learning_rate": 1.762942573809434e-06, "loss": 0.4223, "step": 49585 }, { "epoch": 3.37, "learning_rate": 1.7619999245880625e-06, "loss": 0.0978, "step": 49590 }, { "epoch": 3.37, "learning_rate": 1.7610572753666908e-06, "loss": 0.2549, "step": 49595 }, { "epoch": 3.37, "learning_rate": 1.7601146261453189e-06, "loss": 0.0572, "step": 49600 }, { "epoch": 3.37, "learning_rate": 1.7591719769239472e-06, "loss": 0.1578, "step": 49605 }, { "epoch": 3.37, "learning_rate": 1.7582293277025756e-06, "loss": 0.035, "step": 49610 }, { "epoch": 3.37, "learning_rate": 1.7572866784812037e-06, "loss": 0.5919, "step": 49615 }, { "epoch": 3.37, "learning_rate": 1.756344029259832e-06, "loss": 0.2932, "step": 49620 }, { "epoch": 3.37, "learning_rate": 1.7554013800384603e-06, "loss": 0.0882, "step": 49625 }, { "epoch": 3.37, "learning_rate": 1.7544587308170885e-06, "loss": 0.299, "step": 49630 }, { "epoch": 3.37, "learning_rate": 1.7535160815957166e-06, "loss": 0.4755, "step": 49635 }, { "epoch": 3.37, "learning_rate": 1.752573432374345e-06, "loss": 0.4729, "step": 49640 }, { "epoch": 3.37, "learning_rate": 1.7516307831529734e-06, "loss": 0.2327, "step": 49645 }, { "epoch": 3.37, "learning_rate": 1.7506881339316014e-06, "loss": 0.3486, "step": 49650 }, { "epoch": 3.37, "learning_rate": 1.7497454847102297e-06, "loss": 0.4686, "step": 49655 }, { "epoch": 3.37, "learning_rate": 1.7488028354888582e-06, "loss": 0.1099, "step": 49660 }, { "epoch": 3.37, "learning_rate": 1.7478601862674863e-06, "loss": 0.1941, "step": 49665 }, { "epoch": 3.37, "learning_rate": 1.7469175370461145e-06, "loss": 0.0748, "step": 49670 }, { "epoch": 3.37, "learning_rate": 1.7459748878247428e-06, "loss": 0.1351, "step": 49675 }, { "epoch": 3.37, "learning_rate": 1.745032238603371e-06, "loss": 0.1107, "step": 49680 }, { "epoch": 3.37, "learning_rate": 1.7440895893819991e-06, "loss": 0.2171, "step": 49685 }, { "epoch": 3.37, "learning_rate": 1.7431469401606276e-06, "loss": 0.2232, "step": 49690 }, { "epoch": 3.37, "learning_rate": 1.742204290939256e-06, "loss": 0.1883, "step": 49695 }, { "epoch": 3.37, "learning_rate": 1.741261641717884e-06, "loss": 0.1079, "step": 49700 }, { "epoch": 3.37, "learning_rate": 1.7403189924965122e-06, "loss": 0.2148, "step": 49705 }, { "epoch": 3.37, "learning_rate": 1.7393763432751407e-06, "loss": 0.2239, "step": 49710 }, { "epoch": 3.37, "learning_rate": 1.7384336940537688e-06, "loss": 0.2107, "step": 49715 }, { "epoch": 3.37, "learning_rate": 1.737491044832397e-06, "loss": 0.2583, "step": 49720 }, { "epoch": 3.37, "learning_rate": 1.7365483956110253e-06, "loss": 0.2519, "step": 49725 }, { "epoch": 3.38, "learning_rate": 1.7356057463896536e-06, "loss": 0.5496, "step": 49730 }, { "epoch": 3.38, "learning_rate": 1.734663097168282e-06, "loss": 0.2046, "step": 49735 }, { "epoch": 3.38, "learning_rate": 1.7337204479469102e-06, "loss": 0.2589, "step": 49740 }, { "epoch": 3.38, "learning_rate": 1.7327777987255384e-06, "loss": 0.2313, "step": 49745 }, { "epoch": 3.38, "learning_rate": 1.7318351495041665e-06, "loss": 0.4079, "step": 49750 }, { "epoch": 3.38, "learning_rate": 1.730892500282795e-06, "loss": 0.2193, "step": 49755 }, { "epoch": 3.38, "learning_rate": 1.7299498510614233e-06, "loss": 0.3092, "step": 49760 }, { "epoch": 3.38, "learning_rate": 1.7290072018400513e-06, "loss": 0.5196, "step": 49765 }, { "epoch": 3.38, "learning_rate": 1.7280645526186796e-06, "loss": 0.339, "step": 49770 }, { "epoch": 3.38, "learning_rate": 1.727121903397308e-06, "loss": 0.663, "step": 49775 }, { "epoch": 3.38, "learning_rate": 1.7261792541759362e-06, "loss": 0.2852, "step": 49780 }, { "epoch": 3.38, "learning_rate": 1.7252366049545644e-06, "loss": 0.5524, "step": 49785 }, { "epoch": 3.38, "learning_rate": 1.7242939557331927e-06, "loss": 0.1314, "step": 49790 }, { "epoch": 3.38, "learning_rate": 1.723351306511821e-06, "loss": 0.1547, "step": 49795 }, { "epoch": 3.38, "learning_rate": 1.722408657290449e-06, "loss": 0.2159, "step": 49800 }, { "epoch": 3.38, "learning_rate": 1.7214660080690775e-06, "loss": 0.5836, "step": 49805 }, { "epoch": 3.38, "learning_rate": 1.7205233588477058e-06, "loss": 0.3443, "step": 49810 }, { "epoch": 3.38, "learning_rate": 1.7195807096263339e-06, "loss": 0.1953, "step": 49815 }, { "epoch": 3.38, "learning_rate": 1.7186380604049622e-06, "loss": 0.6415, "step": 49820 }, { "epoch": 3.38, "learning_rate": 1.7176954111835906e-06, "loss": 0.2528, "step": 49825 }, { "epoch": 3.38, "learning_rate": 1.7167527619622187e-06, "loss": 0.2361, "step": 49830 }, { "epoch": 3.38, "learning_rate": 1.715810112740847e-06, "loss": 0.3123, "step": 49835 }, { "epoch": 3.38, "learning_rate": 1.7148674635194753e-06, "loss": 0.3884, "step": 49840 }, { "epoch": 3.38, "learning_rate": 1.7139248142981035e-06, "loss": 0.1512, "step": 49845 }, { "epoch": 3.38, "learning_rate": 1.7129821650767318e-06, "loss": 0.1146, "step": 49850 }, { "epoch": 3.38, "learning_rate": 1.71203951585536e-06, "loss": 0.6317, "step": 49855 }, { "epoch": 3.38, "learning_rate": 1.7110968666339884e-06, "loss": 0.3339, "step": 49860 }, { "epoch": 3.38, "learning_rate": 1.7101542174126164e-06, "loss": 0.3411, "step": 49865 }, { "epoch": 3.38, "learning_rate": 1.709211568191245e-06, "loss": 0.1758, "step": 49870 }, { "epoch": 3.39, "learning_rate": 1.7082689189698732e-06, "loss": 0.2897, "step": 49875 }, { "epoch": 3.39, "learning_rate": 1.7073262697485013e-06, "loss": 0.3304, "step": 49880 }, { "epoch": 3.39, "learning_rate": 1.7063836205271295e-06, "loss": 0.5316, "step": 49885 }, { "epoch": 3.39, "learning_rate": 1.705440971305758e-06, "loss": 0.4346, "step": 49890 }, { "epoch": 3.39, "learning_rate": 1.704498322084386e-06, "loss": 0.4332, "step": 49895 }, { "epoch": 3.39, "learning_rate": 1.7035556728630144e-06, "loss": 0.2022, "step": 49900 }, { "epoch": 3.39, "learning_rate": 1.7026130236416426e-06, "loss": 0.1447, "step": 49905 }, { "epoch": 3.39, "learning_rate": 1.701670374420271e-06, "loss": 0.1096, "step": 49910 }, { "epoch": 3.39, "learning_rate": 1.700727725198899e-06, "loss": 0.3077, "step": 49915 }, { "epoch": 3.39, "learning_rate": 1.6997850759775275e-06, "loss": 0.0837, "step": 49920 }, { "epoch": 3.39, "learning_rate": 1.6988424267561557e-06, "loss": 0.1719, "step": 49925 }, { "epoch": 3.39, "learning_rate": 1.6978997775347838e-06, "loss": 0.1935, "step": 49930 }, { "epoch": 3.39, "learning_rate": 1.696957128313412e-06, "loss": 0.1673, "step": 49935 }, { "epoch": 3.39, "learning_rate": 1.6960144790920406e-06, "loss": 0.3518, "step": 49940 }, { "epoch": 3.39, "learning_rate": 1.6950718298706686e-06, "loss": 0.0764, "step": 49945 }, { "epoch": 3.39, "learning_rate": 1.694129180649297e-06, "loss": 0.1736, "step": 49950 }, { "epoch": 3.39, "learning_rate": 1.6931865314279252e-06, "loss": 0.2242, "step": 49955 }, { "epoch": 3.39, "learning_rate": 1.6922438822065534e-06, "loss": 0.1774, "step": 49960 }, { "epoch": 3.39, "learning_rate": 1.6913012329851815e-06, "loss": 0.5426, "step": 49965 }, { "epoch": 3.39, "learning_rate": 1.69035858376381e-06, "loss": 0.3411, "step": 49970 }, { "epoch": 3.39, "learning_rate": 1.6894159345424383e-06, "loss": 0.221, "step": 49975 }, { "epoch": 3.39, "learning_rate": 1.6884732853210663e-06, "loss": 0.3322, "step": 49980 }, { "epoch": 3.39, "learning_rate": 1.6875306360996946e-06, "loss": 0.283, "step": 49985 }, { "epoch": 3.39, "learning_rate": 1.686587986878323e-06, "loss": 0.2072, "step": 49990 }, { "epoch": 3.39, "learning_rate": 1.6856453376569512e-06, "loss": 0.0783, "step": 49995 }, { "epoch": 3.39, "learning_rate": 1.6847026884355794e-06, "loss": 0.2359, "step": 50000 }, { "epoch": 3.39, "learning_rate": 1.683760039214208e-06, "loss": 0.5176, "step": 50005 }, { "epoch": 3.39, "learning_rate": 1.682817389992836e-06, "loss": 0.7033, "step": 50010 }, { "epoch": 3.39, "learning_rate": 1.6818747407714643e-06, "loss": 0.1654, "step": 50015 }, { "epoch": 3.39, "learning_rate": 1.6809320915500925e-06, "loss": 0.0645, "step": 50020 }, { "epoch": 3.4, "learning_rate": 1.6799894423287208e-06, "loss": 0.1161, "step": 50025 }, { "epoch": 3.4, "learning_rate": 1.6790467931073489e-06, "loss": 0.1242, "step": 50030 }, { "epoch": 3.4, "learning_rate": 1.6781041438859774e-06, "loss": 0.3807, "step": 50035 }, { "epoch": 3.4, "learning_rate": 1.6771614946646056e-06, "loss": 0.1249, "step": 50040 }, { "epoch": 3.4, "learning_rate": 1.6762188454432337e-06, "loss": 0.0432, "step": 50045 }, { "epoch": 3.4, "learning_rate": 1.675276196221862e-06, "loss": 0.1159, "step": 50050 }, { "epoch": 3.4, "learning_rate": 1.6743335470004905e-06, "loss": 0.236, "step": 50055 }, { "epoch": 3.4, "learning_rate": 1.6733908977791185e-06, "loss": 0.1172, "step": 50060 }, { "epoch": 3.4, "learning_rate": 1.6724482485577468e-06, "loss": 0.3638, "step": 50065 }, { "epoch": 3.4, "learning_rate": 1.671505599336375e-06, "loss": 0.2544, "step": 50070 }, { "epoch": 3.4, "learning_rate": 1.6705629501150034e-06, "loss": 0.5919, "step": 50075 }, { "epoch": 3.4, "learning_rate": 1.6696203008936314e-06, "loss": 0.0696, "step": 50080 }, { "epoch": 3.4, "learning_rate": 1.66867765167226e-06, "loss": 0.3762, "step": 50085 }, { "epoch": 3.4, "learning_rate": 1.6677350024508882e-06, "loss": 0.1375, "step": 50090 }, { "epoch": 3.4, "learning_rate": 1.6667923532295163e-06, "loss": 0.6193, "step": 50095 }, { "epoch": 3.4, "learning_rate": 1.6658497040081445e-06, "loss": 0.3923, "step": 50100 }, { "epoch": 3.4, "learning_rate": 1.664907054786773e-06, "loss": 0.3498, "step": 50105 }, { "epoch": 3.4, "learning_rate": 1.663964405565401e-06, "loss": 0.0948, "step": 50110 }, { "epoch": 3.4, "learning_rate": 1.6630217563440294e-06, "loss": 0.1103, "step": 50115 }, { "epoch": 3.4, "learning_rate": 1.6620791071226578e-06, "loss": 0.3172, "step": 50120 }, { "epoch": 3.4, "learning_rate": 1.661136457901286e-06, "loss": 0.1622, "step": 50125 }, { "epoch": 3.4, "learning_rate": 1.6601938086799142e-06, "loss": 0.3731, "step": 50130 }, { "epoch": 3.4, "learning_rate": 1.6592511594585425e-06, "loss": 0.2142, "step": 50135 }, { "epoch": 3.4, "learning_rate": 1.6583085102371707e-06, "loss": 0.5155, "step": 50140 }, { "epoch": 3.4, "learning_rate": 1.6573658610157988e-06, "loss": 0.6158, "step": 50145 }, { "epoch": 3.4, "learning_rate": 1.656423211794427e-06, "loss": 0.0796, "step": 50150 }, { "epoch": 3.4, "learning_rate": 1.6554805625730556e-06, "loss": 0.2775, "step": 50155 }, { "epoch": 3.4, "learning_rate": 1.6545379133516836e-06, "loss": 0.4582, "step": 50160 }, { "epoch": 3.4, "learning_rate": 1.653595264130312e-06, "loss": 0.4552, "step": 50165 }, { "epoch": 3.41, "learning_rate": 1.6526526149089404e-06, "loss": 0.2115, "step": 50170 }, { "epoch": 3.41, "learning_rate": 1.6517099656875685e-06, "loss": 0.0657, "step": 50175 }, { "epoch": 3.41, "learning_rate": 1.6507673164661967e-06, "loss": 0.9539, "step": 50180 }, { "epoch": 3.41, "learning_rate": 1.649824667244825e-06, "loss": 0.2604, "step": 50185 }, { "epoch": 3.41, "learning_rate": 1.6488820180234533e-06, "loss": 0.0484, "step": 50190 }, { "epoch": 3.41, "learning_rate": 1.6479393688020813e-06, "loss": 0.5723, "step": 50195 }, { "epoch": 3.41, "learning_rate": 1.6469967195807098e-06, "loss": 0.2216, "step": 50200 }, { "epoch": 3.41, "learning_rate": 1.646054070359338e-06, "loss": 0.3689, "step": 50205 }, { "epoch": 3.41, "learning_rate": 1.6451114211379662e-06, "loss": 0.3679, "step": 50210 }, { "epoch": 3.41, "learning_rate": 1.6441687719165944e-06, "loss": 0.2328, "step": 50215 }, { "epoch": 3.41, "learning_rate": 1.643226122695223e-06, "loss": 0.2638, "step": 50220 }, { "epoch": 3.41, "learning_rate": 1.642283473473851e-06, "loss": 0.6649, "step": 50225 }, { "epoch": 3.41, "learning_rate": 1.6413408242524793e-06, "loss": 0.5127, "step": 50230 }, { "epoch": 3.41, "learning_rate": 1.6403981750311078e-06, "loss": 0.3576, "step": 50235 }, { "epoch": 3.41, "learning_rate": 1.6394555258097358e-06, "loss": 0.5654, "step": 50240 }, { "epoch": 3.41, "learning_rate": 1.638512876588364e-06, "loss": 0.2483, "step": 50245 }, { "epoch": 3.41, "learning_rate": 1.6375702273669924e-06, "loss": 0.2135, "step": 50250 }, { "epoch": 3.41, "learning_rate": 1.6366275781456206e-06, "loss": 0.2261, "step": 50255 }, { "epoch": 3.41, "learning_rate": 1.6356849289242487e-06, "loss": 0.126, "step": 50260 }, { "epoch": 3.41, "learning_rate": 1.634742279702877e-06, "loss": 0.1311, "step": 50265 }, { "epoch": 3.41, "learning_rate": 1.6337996304815055e-06, "loss": 0.2905, "step": 50270 }, { "epoch": 3.41, "learning_rate": 1.6328569812601335e-06, "loss": 0.294, "step": 50275 }, { "epoch": 3.41, "learning_rate": 1.6319143320387618e-06, "loss": 0.3556, "step": 50280 }, { "epoch": 3.41, "learning_rate": 1.6309716828173903e-06, "loss": 0.2461, "step": 50285 }, { "epoch": 3.41, "learning_rate": 1.6300290335960184e-06, "loss": 0.3763, "step": 50290 }, { "epoch": 3.41, "learning_rate": 1.6290863843746466e-06, "loss": 0.2275, "step": 50295 }, { "epoch": 3.41, "learning_rate": 1.628143735153275e-06, "loss": 0.1591, "step": 50300 }, { "epoch": 3.41, "learning_rate": 1.6272010859319032e-06, "loss": 0.2289, "step": 50305 }, { "epoch": 3.41, "learning_rate": 1.6262584367105313e-06, "loss": 0.3503, "step": 50310 }, { "epoch": 3.41, "learning_rate": 1.6253157874891597e-06, "loss": 0.1568, "step": 50315 }, { "epoch": 3.42, "learning_rate": 1.624373138267788e-06, "loss": 0.2712, "step": 50320 }, { "epoch": 3.42, "learning_rate": 1.623430489046416e-06, "loss": 0.2309, "step": 50325 }, { "epoch": 3.42, "learning_rate": 1.6224878398250444e-06, "loss": 0.448, "step": 50330 }, { "epoch": 3.42, "learning_rate": 1.6215451906036728e-06, "loss": 0.5351, "step": 50335 }, { "epoch": 3.42, "learning_rate": 1.620602541382301e-06, "loss": 0.5116, "step": 50340 }, { "epoch": 3.42, "learning_rate": 1.6196598921609292e-06, "loss": 0.1772, "step": 50345 }, { "epoch": 3.42, "learning_rate": 1.6187172429395575e-06, "loss": 0.1259, "step": 50350 }, { "epoch": 3.42, "learning_rate": 1.6177745937181857e-06, "loss": 0.1419, "step": 50355 }, { "epoch": 3.42, "learning_rate": 1.616831944496814e-06, "loss": 0.095, "step": 50360 }, { "epoch": 3.42, "learning_rate": 1.6158892952754423e-06, "loss": 0.304, "step": 50365 }, { "epoch": 3.42, "learning_rate": 1.6149466460540706e-06, "loss": 0.2971, "step": 50370 }, { "epoch": 3.42, "learning_rate": 1.6140039968326986e-06, "loss": 0.1341, "step": 50375 }, { "epoch": 3.42, "learning_rate": 1.613061347611327e-06, "loss": 0.3253, "step": 50380 }, { "epoch": 3.42, "learning_rate": 1.6121186983899554e-06, "loss": 0.1685, "step": 50385 }, { "epoch": 3.42, "learning_rate": 1.6111760491685835e-06, "loss": 0.313, "step": 50390 }, { "epoch": 3.42, "learning_rate": 1.6102333999472117e-06, "loss": 0.2669, "step": 50395 }, { "epoch": 3.42, "learning_rate": 1.6092907507258402e-06, "loss": 0.2181, "step": 50400 }, { "epoch": 3.42, "learning_rate": 1.6083481015044683e-06, "loss": 0.1723, "step": 50405 }, { "epoch": 3.42, "learning_rate": 1.6074054522830966e-06, "loss": 0.1457, "step": 50410 }, { "epoch": 3.42, "learning_rate": 1.6064628030617248e-06, "loss": 0.2714, "step": 50415 }, { "epoch": 3.42, "learning_rate": 1.605520153840353e-06, "loss": 0.1114, "step": 50420 }, { "epoch": 3.42, "learning_rate": 1.6045775046189812e-06, "loss": 0.0696, "step": 50425 }, { "epoch": 3.42, "learning_rate": 1.6036348553976094e-06, "loss": 0.2554, "step": 50430 }, { "epoch": 3.42, "learning_rate": 1.602692206176238e-06, "loss": 0.3308, "step": 50435 }, { "epoch": 3.42, "learning_rate": 1.601749556954866e-06, "loss": 0.5241, "step": 50440 }, { "epoch": 3.42, "learning_rate": 1.6008069077334943e-06, "loss": 0.1178, "step": 50445 }, { "epoch": 3.42, "learning_rate": 1.5998642585121228e-06, "loss": 0.3736, "step": 50450 }, { "epoch": 3.42, "learning_rate": 1.5989216092907508e-06, "loss": 0.4926, "step": 50455 }, { "epoch": 3.42, "learning_rate": 1.597978960069379e-06, "loss": 0.3332, "step": 50460 }, { "epoch": 3.43, "learning_rate": 1.5970363108480074e-06, "loss": 0.384, "step": 50465 }, { "epoch": 3.43, "learning_rate": 1.5960936616266356e-06, "loss": 0.1818, "step": 50470 }, { "epoch": 3.43, "learning_rate": 1.595151012405264e-06, "loss": 0.122, "step": 50475 }, { "epoch": 3.43, "learning_rate": 1.5942083631838922e-06, "loss": 0.0167, "step": 50480 }, { "epoch": 3.43, "learning_rate": 1.5932657139625205e-06, "loss": 0.0335, "step": 50485 }, { "epoch": 3.43, "learning_rate": 1.5923230647411485e-06, "loss": 0.0477, "step": 50490 }, { "epoch": 3.43, "learning_rate": 1.5913804155197768e-06, "loss": 0.5298, "step": 50495 }, { "epoch": 3.43, "learning_rate": 1.5904377662984053e-06, "loss": 0.2776, "step": 50500 }, { "epoch": 3.43, "learning_rate": 1.5894951170770334e-06, "loss": 0.451, "step": 50505 }, { "epoch": 3.43, "learning_rate": 1.5885524678556616e-06, "loss": 0.129, "step": 50510 }, { "epoch": 3.43, "learning_rate": 1.5876098186342901e-06, "loss": 0.1523, "step": 50515 }, { "epoch": 3.43, "learning_rate": 1.5866671694129182e-06, "loss": 0.3546, "step": 50520 }, { "epoch": 3.43, "learning_rate": 1.5857245201915465e-06, "loss": 0.3726, "step": 50525 }, { "epoch": 3.43, "learning_rate": 1.5847818709701747e-06, "loss": 0.1691, "step": 50530 }, { "epoch": 3.43, "learning_rate": 1.583839221748803e-06, "loss": 0.3129, "step": 50535 }, { "epoch": 3.43, "learning_rate": 1.582896572527431e-06, "loss": 0.3486, "step": 50540 }, { "epoch": 3.43, "learning_rate": 1.5819539233060594e-06, "loss": 0.2413, "step": 50545 }, { "epoch": 3.43, "learning_rate": 1.5810112740846878e-06, "loss": 0.1681, "step": 50550 }, { "epoch": 3.43, "learning_rate": 1.580068624863316e-06, "loss": 0.1778, "step": 50555 }, { "epoch": 3.43, "learning_rate": 1.5791259756419442e-06, "loss": 0.3106, "step": 50560 }, { "epoch": 3.43, "learning_rate": 1.5781833264205727e-06, "loss": 0.1108, "step": 50565 }, { "epoch": 3.43, "learning_rate": 1.5772406771992007e-06, "loss": 0.1696, "step": 50570 }, { "epoch": 3.43, "learning_rate": 1.576298027977829e-06, "loss": 0.322, "step": 50575 }, { "epoch": 3.43, "learning_rate": 1.5753553787564573e-06, "loss": 1.0356, "step": 50580 }, { "epoch": 3.43, "learning_rate": 1.5744127295350856e-06, "loss": 0.3526, "step": 50585 }, { "epoch": 3.43, "learning_rate": 1.5734700803137136e-06, "loss": 0.5157, "step": 50590 }, { "epoch": 3.43, "learning_rate": 1.5725274310923421e-06, "loss": 0.2459, "step": 50595 }, { "epoch": 3.43, "learning_rate": 1.5715847818709704e-06, "loss": 0.1444, "step": 50600 }, { "epoch": 3.43, "learning_rate": 1.5706421326495985e-06, "loss": 0.343, "step": 50605 }, { "epoch": 3.43, "learning_rate": 1.5696994834282267e-06, "loss": 0.3215, "step": 50610 }, { "epoch": 3.44, "learning_rate": 1.5687568342068552e-06, "loss": 0.2049, "step": 50615 }, { "epoch": 3.44, "learning_rate": 1.5678141849854833e-06, "loss": 0.4485, "step": 50620 }, { "epoch": 3.44, "learning_rate": 1.5668715357641116e-06, "loss": 0.1633, "step": 50625 }, { "epoch": 3.44, "learning_rate": 1.56592888654274e-06, "loss": 0.1253, "step": 50630 }, { "epoch": 3.44, "learning_rate": 1.5649862373213681e-06, "loss": 0.3752, "step": 50635 }, { "epoch": 3.44, "learning_rate": 1.5640435880999964e-06, "loss": 0.2002, "step": 50640 }, { "epoch": 3.44, "learning_rate": 1.5631009388786247e-06, "loss": 0.7329, "step": 50645 }, { "epoch": 3.44, "learning_rate": 1.562158289657253e-06, "loss": 0.2534, "step": 50650 }, { "epoch": 3.44, "learning_rate": 1.561215640435881e-06, "loss": 0.0307, "step": 50655 }, { "epoch": 3.44, "learning_rate": 1.5602729912145093e-06, "loss": 0.3164, "step": 50660 }, { "epoch": 3.44, "learning_rate": 1.5593303419931378e-06, "loss": 0.1794, "step": 50665 }, { "epoch": 3.44, "learning_rate": 1.5583876927717658e-06, "loss": 0.2657, "step": 50670 }, { "epoch": 3.44, "learning_rate": 1.557445043550394e-06, "loss": 0.4732, "step": 50675 }, { "epoch": 3.44, "learning_rate": 1.5565023943290226e-06, "loss": 0.3006, "step": 50680 }, { "epoch": 3.44, "learning_rate": 1.5555597451076507e-06, "loss": 0.0363, "step": 50685 }, { "epoch": 3.44, "learning_rate": 1.554617095886279e-06, "loss": 1.203, "step": 50690 }, { "epoch": 3.44, "learning_rate": 1.5536744466649072e-06, "loss": 0.0687, "step": 50695 }, { "epoch": 3.44, "learning_rate": 1.5527317974435355e-06, "loss": 0.2262, "step": 50700 }, { "epoch": 3.44, "learning_rate": 1.5517891482221635e-06, "loss": 0.1878, "step": 50705 }, { "epoch": 3.44, "learning_rate": 1.5508464990007918e-06, "loss": 0.2212, "step": 50710 }, { "epoch": 3.44, "learning_rate": 1.5499038497794203e-06, "loss": 0.5875, "step": 50715 }, { "epoch": 3.44, "learning_rate": 1.5489612005580484e-06, "loss": 0.3236, "step": 50720 }, { "epoch": 3.44, "learning_rate": 1.5480185513366766e-06, "loss": 0.2183, "step": 50725 }, { "epoch": 3.44, "learning_rate": 1.5470759021153051e-06, "loss": 0.343, "step": 50730 }, { "epoch": 3.44, "learning_rate": 1.5461332528939332e-06, "loss": 0.2178, "step": 50735 }, { "epoch": 3.44, "learning_rate": 1.5451906036725615e-06, "loss": 0.0676, "step": 50740 }, { "epoch": 3.44, "learning_rate": 1.54424795445119e-06, "loss": 0.2013, "step": 50745 }, { "epoch": 3.44, "learning_rate": 1.543305305229818e-06, "loss": 0.6448, "step": 50750 }, { "epoch": 3.44, "learning_rate": 1.5423626560084463e-06, "loss": 0.4393, "step": 50755 }, { "epoch": 3.45, "learning_rate": 1.5414200067870746e-06, "loss": 0.2989, "step": 50760 }, { "epoch": 3.45, "learning_rate": 1.5404773575657028e-06, "loss": 0.0646, "step": 50765 }, { "epoch": 3.45, "learning_rate": 1.539534708344331e-06, "loss": 0.8388, "step": 50770 }, { "epoch": 3.45, "learning_rate": 1.5385920591229592e-06, "loss": 0.1471, "step": 50775 }, { "epoch": 3.45, "learning_rate": 1.5376494099015877e-06, "loss": 0.4562, "step": 50780 }, { "epoch": 3.45, "learning_rate": 1.5367067606802157e-06, "loss": 0.5683, "step": 50785 }, { "epoch": 3.45, "learning_rate": 1.535764111458844e-06, "loss": 0.3888, "step": 50790 }, { "epoch": 3.45, "learning_rate": 1.5348214622374725e-06, "loss": 0.0701, "step": 50795 }, { "epoch": 3.45, "learning_rate": 1.5338788130161006e-06, "loss": 0.0893, "step": 50800 }, { "epoch": 3.45, "learning_rate": 1.5329361637947288e-06, "loss": 0.5164, "step": 50805 }, { "epoch": 3.45, "learning_rate": 1.5319935145733571e-06, "loss": 0.1648, "step": 50810 }, { "epoch": 3.45, "learning_rate": 1.5310508653519854e-06, "loss": 0.2466, "step": 50815 }, { "epoch": 3.45, "learning_rate": 1.5301082161306135e-06, "loss": 0.4723, "step": 50820 }, { "epoch": 3.45, "learning_rate": 1.5291655669092417e-06, "loss": 0.0874, "step": 50825 }, { "epoch": 3.45, "learning_rate": 1.5282229176878702e-06, "loss": 0.455, "step": 50830 }, { "epoch": 3.45, "learning_rate": 1.5272802684664983e-06, "loss": 0.2231, "step": 50835 }, { "epoch": 3.45, "learning_rate": 1.5263376192451266e-06, "loss": 0.4395, "step": 50840 }, { "epoch": 3.45, "learning_rate": 1.525394970023755e-06, "loss": 0.1388, "step": 50845 }, { "epoch": 3.45, "learning_rate": 1.5244523208023831e-06, "loss": 0.1902, "step": 50850 }, { "epoch": 3.45, "learning_rate": 1.5235096715810114e-06, "loss": 0.1925, "step": 50855 }, { "epoch": 3.45, "learning_rate": 1.5225670223596399e-06, "loss": 0.3318, "step": 50860 }, { "epoch": 3.45, "learning_rate": 1.521624373138268e-06, "loss": 0.1184, "step": 50865 }, { "epoch": 3.45, "learning_rate": 1.5206817239168962e-06, "loss": 0.026, "step": 50870 }, { "epoch": 3.45, "learning_rate": 1.5197390746955245e-06, "loss": 0.6246, "step": 50875 }, { "epoch": 3.45, "learning_rate": 1.5187964254741528e-06, "loss": 0.1353, "step": 50880 }, { "epoch": 3.45, "learning_rate": 1.5178537762527808e-06, "loss": 0.0979, "step": 50885 }, { "epoch": 3.45, "learning_rate": 1.516911127031409e-06, "loss": 0.2533, "step": 50890 }, { "epoch": 3.45, "learning_rate": 1.5159684778100376e-06, "loss": 0.5064, "step": 50895 }, { "epoch": 3.45, "learning_rate": 1.5150258285886657e-06, "loss": 0.3991, "step": 50900 }, { "epoch": 3.45, "learning_rate": 1.514083179367294e-06, "loss": 0.2567, "step": 50905 }, { "epoch": 3.46, "learning_rate": 1.5131405301459224e-06, "loss": 0.1311, "step": 50910 }, { "epoch": 3.46, "learning_rate": 1.5121978809245505e-06, "loss": 0.5052, "step": 50915 }, { "epoch": 3.46, "learning_rate": 1.5112552317031788e-06, "loss": 0.1033, "step": 50920 }, { "epoch": 3.46, "learning_rate": 1.510312582481807e-06, "loss": 0.1773, "step": 50925 }, { "epoch": 3.46, "learning_rate": 1.5093699332604353e-06, "loss": 0.07, "step": 50930 }, { "epoch": 3.46, "learning_rate": 1.5084272840390634e-06, "loss": 0.3204, "step": 50935 }, { "epoch": 3.46, "learning_rate": 1.5074846348176916e-06, "loss": 0.1574, "step": 50940 }, { "epoch": 3.46, "learning_rate": 1.5065419855963201e-06, "loss": 0.3076, "step": 50945 }, { "epoch": 3.46, "learning_rate": 1.5055993363749482e-06, "loss": 0.3525, "step": 50950 }, { "epoch": 3.46, "learning_rate": 1.5046566871535765e-06, "loss": 0.674, "step": 50955 }, { "epoch": 3.46, "learning_rate": 1.503714037932205e-06, "loss": 0.2627, "step": 50960 }, { "epoch": 3.46, "learning_rate": 1.502771388710833e-06, "loss": 0.3635, "step": 50965 }, { "epoch": 3.46, "learning_rate": 1.5018287394894613e-06, "loss": 0.1137, "step": 50970 }, { "epoch": 3.46, "learning_rate": 1.5008860902680898e-06, "loss": 0.0954, "step": 50975 }, { "epoch": 3.46, "learning_rate": 1.4999434410467178e-06, "loss": 0.1343, "step": 50980 }, { "epoch": 3.46, "learning_rate": 1.4990007918253461e-06, "loss": 0.3752, "step": 50985 }, { "epoch": 3.46, "learning_rate": 1.4980581426039742e-06, "loss": 0.1896, "step": 50990 }, { "epoch": 3.46, "learning_rate": 1.4971154933826027e-06, "loss": 0.206, "step": 50995 }, { "epoch": 3.46, "learning_rate": 1.4961728441612307e-06, "loss": 0.0587, "step": 51000 }, { "epoch": 3.46, "learning_rate": 1.495230194939859e-06, "loss": 0.2257, "step": 51005 }, { "epoch": 3.46, "learning_rate": 1.4942875457184875e-06, "loss": 0.1328, "step": 51010 }, { "epoch": 3.46, "learning_rate": 1.4933448964971156e-06, "loss": 0.257, "step": 51015 }, { "epoch": 3.46, "learning_rate": 1.4924022472757438e-06, "loss": 0.6323, "step": 51020 }, { "epoch": 3.46, "learning_rate": 1.4914595980543723e-06, "loss": 0.165, "step": 51025 }, { "epoch": 3.46, "learning_rate": 1.4905169488330004e-06, "loss": 0.0133, "step": 51030 }, { "epoch": 3.46, "learning_rate": 1.4895742996116287e-06, "loss": 0.3206, "step": 51035 }, { "epoch": 3.46, "learning_rate": 1.488631650390257e-06, "loss": 0.2996, "step": 51040 }, { "epoch": 3.46, "learning_rate": 1.4876890011688852e-06, "loss": 0.1507, "step": 51045 }, { "epoch": 3.46, "learning_rate": 1.4867463519475133e-06, "loss": 0.205, "step": 51050 }, { "epoch": 3.47, "learning_rate": 1.4858037027261416e-06, "loss": 0.1123, "step": 51055 }, { "epoch": 3.47, "learning_rate": 1.48486105350477e-06, "loss": 0.1768, "step": 51060 }, { "epoch": 3.47, "learning_rate": 1.4839184042833981e-06, "loss": 0.2019, "step": 51065 }, { "epoch": 3.47, "learning_rate": 1.4829757550620264e-06, "loss": 0.3612, "step": 51070 }, { "epoch": 3.47, "learning_rate": 1.4820331058406549e-06, "loss": 0.3799, "step": 51075 }, { "epoch": 3.47, "learning_rate": 1.481090456619283e-06, "loss": 0.674, "step": 51080 }, { "epoch": 3.47, "learning_rate": 1.4801478073979112e-06, "loss": 0.4519, "step": 51085 }, { "epoch": 3.47, "learning_rate": 1.4792051581765395e-06, "loss": 0.1876, "step": 51090 }, { "epoch": 3.47, "learning_rate": 1.4782625089551678e-06, "loss": 0.2248, "step": 51095 }, { "epoch": 3.47, "learning_rate": 1.477319859733796e-06, "loss": 0.0935, "step": 51100 }, { "epoch": 3.47, "learning_rate": 1.476377210512424e-06, "loss": 0.3231, "step": 51105 }, { "epoch": 3.47, "learning_rate": 1.4754345612910526e-06, "loss": 0.2929, "step": 51110 }, { "epoch": 3.47, "learning_rate": 1.4744919120696807e-06, "loss": 0.6341, "step": 51115 }, { "epoch": 3.47, "learning_rate": 1.473549262848309e-06, "loss": 0.2781, "step": 51120 }, { "epoch": 3.47, "learning_rate": 1.4726066136269374e-06, "loss": 0.1925, "step": 51125 }, { "epoch": 3.47, "learning_rate": 1.4716639644055655e-06, "loss": 0.2417, "step": 51130 }, { "epoch": 3.47, "learning_rate": 1.4707213151841938e-06, "loss": 0.5188, "step": 51135 }, { "epoch": 3.47, "learning_rate": 1.4697786659628222e-06, "loss": 0.2371, "step": 51140 }, { "epoch": 3.47, "learning_rate": 1.4688360167414503e-06, "loss": 0.0442, "step": 51145 }, { "epoch": 3.47, "learning_rate": 1.4678933675200786e-06, "loss": 0.0174, "step": 51150 }, { "epoch": 3.47, "learning_rate": 1.4669507182987069e-06, "loss": 0.076, "step": 51155 }, { "epoch": 3.47, "learning_rate": 1.4660080690773351e-06, "loss": 0.048, "step": 51160 }, { "epoch": 3.47, "learning_rate": 1.4650654198559632e-06, "loss": 0.3418, "step": 51165 }, { "epoch": 3.47, "learning_rate": 1.4641227706345915e-06, "loss": 0.2435, "step": 51170 }, { "epoch": 3.47, "learning_rate": 1.46318012141322e-06, "loss": 0.239, "step": 51175 }, { "epoch": 3.47, "learning_rate": 1.462237472191848e-06, "loss": 0.3538, "step": 51180 }, { "epoch": 3.47, "learning_rate": 1.4612948229704763e-06, "loss": 0.2191, "step": 51185 }, { "epoch": 3.47, "learning_rate": 1.4603521737491048e-06, "loss": 0.134, "step": 51190 }, { "epoch": 3.47, "learning_rate": 1.4594095245277329e-06, "loss": 0.1776, "step": 51195 }, { "epoch": 3.47, "learning_rate": 1.4584668753063611e-06, "loss": 0.6006, "step": 51200 }, { "epoch": 3.48, "learning_rate": 1.4575242260849894e-06, "loss": 0.0701, "step": 51205 }, { "epoch": 3.48, "learning_rate": 1.4565815768636177e-06, "loss": 0.2191, "step": 51210 }, { "epoch": 3.48, "learning_rate": 1.455638927642246e-06, "loss": 0.5998, "step": 51215 }, { "epoch": 3.48, "learning_rate": 1.454696278420874e-06, "loss": 0.0897, "step": 51220 }, { "epoch": 3.48, "learning_rate": 1.4537536291995025e-06, "loss": 0.1804, "step": 51225 }, { "epoch": 3.48, "learning_rate": 1.4528109799781306e-06, "loss": 0.0679, "step": 51230 }, { "epoch": 3.48, "learning_rate": 1.4518683307567588e-06, "loss": 0.3416, "step": 51235 }, { "epoch": 3.48, "learning_rate": 1.4509256815353873e-06, "loss": 0.3678, "step": 51240 }, { "epoch": 3.48, "learning_rate": 1.4499830323140154e-06, "loss": 0.2158, "step": 51245 }, { "epoch": 3.48, "learning_rate": 1.4490403830926437e-06, "loss": 0.1835, "step": 51250 }, { "epoch": 3.48, "learning_rate": 1.4480977338712722e-06, "loss": 0.2682, "step": 51255 }, { "epoch": 3.48, "learning_rate": 1.4471550846499002e-06, "loss": 0.1262, "step": 51260 }, { "epoch": 3.48, "learning_rate": 1.4462124354285285e-06, "loss": 0.2706, "step": 51265 }, { "epoch": 3.48, "learning_rate": 1.4452697862071566e-06, "loss": 0.4063, "step": 51270 }, { "epoch": 3.48, "learning_rate": 1.444327136985785e-06, "loss": 0.0419, "step": 51275 }, { "epoch": 3.48, "learning_rate": 1.4433844877644131e-06, "loss": 0.5831, "step": 51280 }, { "epoch": 3.48, "learning_rate": 1.4424418385430414e-06, "loss": 0.0368, "step": 51285 }, { "epoch": 3.48, "learning_rate": 1.4414991893216699e-06, "loss": 0.4795, "step": 51290 }, { "epoch": 3.48, "learning_rate": 1.440556540100298e-06, "loss": 0.6859, "step": 51295 }, { "epoch": 3.48, "learning_rate": 1.4396138908789262e-06, "loss": 0.206, "step": 51300 }, { "epoch": 3.48, "learning_rate": 1.4386712416575547e-06, "loss": 0.316, "step": 51305 }, { "epoch": 3.48, "learning_rate": 1.4377285924361828e-06, "loss": 0.2953, "step": 51310 }, { "epoch": 3.48, "learning_rate": 1.436785943214811e-06, "loss": 0.3802, "step": 51315 }, { "epoch": 3.48, "learning_rate": 1.4358432939934393e-06, "loss": 0.3948, "step": 51320 }, { "epoch": 3.48, "learning_rate": 1.4349006447720676e-06, "loss": 0.1524, "step": 51325 }, { "epoch": 3.48, "learning_rate": 1.4339579955506957e-06, "loss": 0.0683, "step": 51330 }, { "epoch": 3.48, "learning_rate": 1.433015346329324e-06, "loss": 0.0803, "step": 51335 }, { "epoch": 3.48, "learning_rate": 1.4320726971079524e-06, "loss": 0.0367, "step": 51340 }, { "epoch": 3.48, "learning_rate": 1.4311300478865805e-06, "loss": 0.2204, "step": 51345 }, { "epoch": 3.49, "learning_rate": 1.4301873986652088e-06, "loss": 0.2424, "step": 51350 }, { "epoch": 3.49, "learning_rate": 1.4292447494438372e-06, "loss": 0.3249, "step": 51355 }, { "epoch": 3.49, "learning_rate": 1.4283021002224653e-06, "loss": 0.3191, "step": 51360 }, { "epoch": 3.49, "learning_rate": 1.4273594510010936e-06, "loss": 0.1801, "step": 51365 }, { "epoch": 3.49, "learning_rate": 1.426416801779722e-06, "loss": 0.158, "step": 51370 }, { "epoch": 3.49, "learning_rate": 1.4254741525583501e-06, "loss": 0.1328, "step": 51375 }, { "epoch": 3.49, "learning_rate": 1.4245315033369784e-06, "loss": 0.2648, "step": 51380 }, { "epoch": 3.49, "learning_rate": 1.4235888541156065e-06, "loss": 0.2115, "step": 51385 }, { "epoch": 3.49, "learning_rate": 1.422646204894235e-06, "loss": 0.0202, "step": 51390 }, { "epoch": 3.49, "learning_rate": 1.421703555672863e-06, "loss": 0.2793, "step": 51395 }, { "epoch": 3.49, "learning_rate": 1.4207609064514913e-06, "loss": 0.2235, "step": 51400 }, { "epoch": 3.49, "learning_rate": 1.4198182572301198e-06, "loss": 0.4677, "step": 51405 }, { "epoch": 3.49, "learning_rate": 1.4188756080087479e-06, "loss": 0.2459, "step": 51410 }, { "epoch": 3.49, "learning_rate": 1.4179329587873761e-06, "loss": 0.3281, "step": 51415 }, { "epoch": 3.49, "learning_rate": 1.4169903095660046e-06, "loss": 0.0985, "step": 51420 }, { "epoch": 3.49, "learning_rate": 1.4160476603446327e-06, "loss": 0.1778, "step": 51425 }, { "epoch": 3.49, "learning_rate": 1.415105011123261e-06, "loss": 0.1836, "step": 51430 }, { "epoch": 3.49, "learning_rate": 1.414162361901889e-06, "loss": 0.3389, "step": 51435 }, { "epoch": 3.49, "learning_rate": 1.4132197126805175e-06, "loss": 0.53, "step": 51440 }, { "epoch": 3.49, "learning_rate": 1.4122770634591456e-06, "loss": 0.0985, "step": 51445 }, { "epoch": 3.49, "learning_rate": 1.4113344142377738e-06, "loss": 0.3999, "step": 51450 }, { "epoch": 3.49, "learning_rate": 1.4103917650164023e-06, "loss": 0.4649, "step": 51455 }, { "epoch": 3.49, "learning_rate": 1.4094491157950304e-06, "loss": 0.1374, "step": 51460 }, { "epoch": 3.49, "learning_rate": 1.4085064665736587e-06, "loss": 0.2766, "step": 51465 }, { "epoch": 3.49, "learning_rate": 1.4075638173522872e-06, "loss": 0.4236, "step": 51470 }, { "epoch": 3.49, "learning_rate": 1.4066211681309152e-06, "loss": 0.1035, "step": 51475 }, { "epoch": 3.49, "learning_rate": 1.4056785189095435e-06, "loss": 0.2446, "step": 51480 }, { "epoch": 3.49, "learning_rate": 1.404735869688172e-06, "loss": 0.3592, "step": 51485 }, { "epoch": 3.49, "learning_rate": 1.4037932204668e-06, "loss": 0.1384, "step": 51490 }, { "epoch": 3.49, "learning_rate": 1.4028505712454283e-06, "loss": 0.1911, "step": 51495 }, { "epoch": 3.5, "learning_rate": 1.4019079220240564e-06, "loss": 0.1917, "step": 51500 }, { "epoch": 3.5, "learning_rate": 1.4009652728026849e-06, "loss": 0.2485, "step": 51505 }, { "epoch": 3.5, "learning_rate": 1.400022623581313e-06, "loss": 0.3397, "step": 51510 }, { "epoch": 3.5, "learning_rate": 1.3990799743599412e-06, "loss": 0.2912, "step": 51515 }, { "epoch": 3.5, "learning_rate": 1.3981373251385697e-06, "loss": 0.2196, "step": 51520 }, { "epoch": 3.5, "learning_rate": 1.3971946759171978e-06, "loss": 0.1548, "step": 51525 }, { "epoch": 3.5, "learning_rate": 1.396252026695826e-06, "loss": 0.3202, "step": 51530 }, { "epoch": 3.5, "learning_rate": 1.3953093774744545e-06, "loss": 0.137, "step": 51535 }, { "epoch": 3.5, "learning_rate": 1.3943667282530826e-06, "loss": 0.9814, "step": 51540 }, { "epoch": 3.5, "learning_rate": 1.3934240790317109e-06, "loss": 0.2089, "step": 51545 }, { "epoch": 3.5, "learning_rate": 1.392481429810339e-06, "loss": 0.4257, "step": 51550 }, { "epoch": 3.5, "learning_rate": 1.3915387805889674e-06, "loss": 0.2012, "step": 51555 }, { "epoch": 3.5, "learning_rate": 1.3905961313675955e-06, "loss": 0.1539, "step": 51560 }, { "epoch": 3.5, "learning_rate": 1.3896534821462238e-06, "loss": 0.1369, "step": 51565 }, { "epoch": 3.5, "learning_rate": 1.3887108329248522e-06, "loss": 0.3926, "step": 51570 }, { "epoch": 3.5, "learning_rate": 1.3877681837034803e-06, "loss": 0.794, "step": 51575 }, { "epoch": 3.5, "learning_rate": 1.3868255344821086e-06, "loss": 0.1686, "step": 51580 }, { "epoch": 3.5, "learning_rate": 1.385882885260737e-06, "loss": 0.3299, "step": 51585 }, { "epoch": 3.5, "learning_rate": 1.3849402360393651e-06, "loss": 0.4351, "step": 51590 }, { "epoch": 3.5, "learning_rate": 1.3839975868179934e-06, "loss": 0.1374, "step": 51595 }, { "epoch": 3.5, "learning_rate": 1.383054937596622e-06, "loss": 0.5321, "step": 51600 }, { "epoch": 3.5, "learning_rate": 1.38211228837525e-06, "loss": 0.1697, "step": 51605 }, { "epoch": 3.5, "learning_rate": 1.3811696391538782e-06, "loss": 0.1292, "step": 51610 }, { "epoch": 3.5, "learning_rate": 1.3802269899325063e-06, "loss": 0.0338, "step": 51615 }, { "epoch": 3.5, "learning_rate": 1.3792843407111348e-06, "loss": 0.1106, "step": 51620 }, { "epoch": 3.5, "learning_rate": 1.3783416914897629e-06, "loss": 0.1336, "step": 51625 }, { "epoch": 3.5, "learning_rate": 1.3773990422683911e-06, "loss": 0.4338, "step": 51630 }, { "epoch": 3.5, "learning_rate": 1.3764563930470196e-06, "loss": 0.7099, "step": 51635 }, { "epoch": 3.5, "learning_rate": 1.3755137438256477e-06, "loss": 0.286, "step": 51640 }, { "epoch": 3.51, "learning_rate": 1.374571094604276e-06, "loss": 0.3723, "step": 51645 }, { "epoch": 3.51, "learning_rate": 1.3736284453829044e-06, "loss": 0.1056, "step": 51650 }, { "epoch": 3.51, "learning_rate": 1.3726857961615325e-06, "loss": 0.1232, "step": 51655 }, { "epoch": 3.51, "learning_rate": 1.3717431469401608e-06, "loss": 0.1398, "step": 51660 }, { "epoch": 3.51, "learning_rate": 1.3708004977187888e-06, "loss": 0.0671, "step": 51665 }, { "epoch": 3.51, "learning_rate": 1.3698578484974173e-06, "loss": 0.1884, "step": 51670 }, { "epoch": 3.51, "learning_rate": 1.3689151992760454e-06, "loss": 0.6774, "step": 51675 }, { "epoch": 3.51, "learning_rate": 1.3679725500546737e-06, "loss": 0.4832, "step": 51680 }, { "epoch": 3.51, "learning_rate": 1.3670299008333022e-06, "loss": 0.2536, "step": 51685 }, { "epoch": 3.51, "learning_rate": 1.3660872516119302e-06, "loss": 0.3188, "step": 51690 }, { "epoch": 3.51, "learning_rate": 1.3651446023905585e-06, "loss": 0.4051, "step": 51695 }, { "epoch": 3.51, "learning_rate": 1.364201953169187e-06, "loss": 0.3398, "step": 51700 }, { "epoch": 3.51, "learning_rate": 1.363259303947815e-06, "loss": 0.0885, "step": 51705 }, { "epoch": 3.51, "learning_rate": 1.3623166547264433e-06, "loss": 0.1575, "step": 51710 }, { "epoch": 3.51, "learning_rate": 1.3613740055050714e-06, "loss": 0.6502, "step": 51715 }, { "epoch": 3.51, "learning_rate": 1.3604313562836999e-06, "loss": 0.3539, "step": 51720 }, { "epoch": 3.51, "learning_rate": 1.3594887070623282e-06, "loss": 0.1763, "step": 51725 }, { "epoch": 3.51, "learning_rate": 1.3585460578409562e-06, "loss": 0.1952, "step": 51730 }, { "epoch": 3.51, "learning_rate": 1.3576034086195847e-06, "loss": 0.2094, "step": 51735 }, { "epoch": 3.51, "learning_rate": 1.3566607593982128e-06, "loss": 0.4781, "step": 51740 }, { "epoch": 3.51, "learning_rate": 1.355718110176841e-06, "loss": 0.05, "step": 51745 }, { "epoch": 3.51, "learning_rate": 1.3547754609554695e-06, "loss": 0.0668, "step": 51750 }, { "epoch": 3.51, "learning_rate": 1.3538328117340976e-06, "loss": 0.0595, "step": 51755 }, { "epoch": 3.51, "learning_rate": 1.3528901625127259e-06, "loss": 0.0965, "step": 51760 }, { "epoch": 3.51, "learning_rate": 1.3519475132913544e-06, "loss": 0.2273, "step": 51765 }, { "epoch": 3.51, "learning_rate": 1.3510048640699824e-06, "loss": 0.314, "step": 51770 }, { "epoch": 3.51, "learning_rate": 1.3500622148486107e-06, "loss": 0.4348, "step": 51775 }, { "epoch": 3.51, "learning_rate": 1.3491195656272388e-06, "loss": 0.0357, "step": 51780 }, { "epoch": 3.51, "learning_rate": 1.3481769164058672e-06, "loss": 0.2427, "step": 51785 }, { "epoch": 3.51, "learning_rate": 1.3472342671844953e-06, "loss": 0.1361, "step": 51790 }, { "epoch": 3.52, "learning_rate": 1.3462916179631236e-06, "loss": 0.2524, "step": 51795 }, { "epoch": 3.52, "learning_rate": 1.345348968741752e-06, "loss": 0.2476, "step": 51800 }, { "epoch": 3.52, "learning_rate": 1.3444063195203801e-06, "loss": 0.5077, "step": 51805 }, { "epoch": 3.52, "learning_rate": 1.3434636702990084e-06, "loss": 0.2606, "step": 51810 }, { "epoch": 3.52, "learning_rate": 1.342521021077637e-06, "loss": 0.1821, "step": 51815 }, { "epoch": 3.52, "learning_rate": 1.341578371856265e-06, "loss": 0.2411, "step": 51820 }, { "epoch": 3.52, "learning_rate": 1.3406357226348932e-06, "loss": 0.4233, "step": 51825 }, { "epoch": 3.52, "learning_rate": 1.3396930734135213e-06, "loss": 0.3582, "step": 51830 }, { "epoch": 3.52, "learning_rate": 1.3387504241921498e-06, "loss": 0.0908, "step": 51835 }, { "epoch": 3.52, "learning_rate": 1.337807774970778e-06, "loss": 0.2861, "step": 51840 }, { "epoch": 3.52, "learning_rate": 1.3368651257494061e-06, "loss": 0.5016, "step": 51845 }, { "epoch": 3.52, "learning_rate": 1.3359224765280346e-06, "loss": 0.1992, "step": 51850 }, { "epoch": 3.52, "learning_rate": 1.3349798273066627e-06, "loss": 0.1473, "step": 51855 }, { "epoch": 3.52, "learning_rate": 1.334037178085291e-06, "loss": 0.5295, "step": 51860 }, { "epoch": 3.52, "learning_rate": 1.3330945288639194e-06, "loss": 0.236, "step": 51865 }, { "epoch": 3.52, "learning_rate": 1.3321518796425475e-06, "loss": 0.0293, "step": 51870 }, { "epoch": 3.52, "learning_rate": 1.3312092304211758e-06, "loss": 0.0939, "step": 51875 }, { "epoch": 3.52, "learning_rate": 1.3302665811998043e-06, "loss": 0.2777, "step": 51880 }, { "epoch": 3.52, "learning_rate": 1.3293239319784323e-06, "loss": 0.155, "step": 51885 }, { "epoch": 3.52, "learning_rate": 1.3283812827570606e-06, "loss": 0.0966, "step": 51890 }, { "epoch": 3.52, "learning_rate": 1.3274386335356887e-06, "loss": 0.2963, "step": 51895 }, { "epoch": 3.52, "learning_rate": 1.3264959843143172e-06, "loss": 0.3023, "step": 51900 }, { "epoch": 3.52, "learning_rate": 1.3255533350929452e-06, "loss": 0.065, "step": 51905 }, { "epoch": 3.52, "learning_rate": 1.3246106858715735e-06, "loss": 0.1524, "step": 51910 }, { "epoch": 3.52, "learning_rate": 1.323668036650202e-06, "loss": 0.2626, "step": 51915 }, { "epoch": 3.52, "learning_rate": 1.32272538742883e-06, "loss": 0.3664, "step": 51920 }, { "epoch": 3.52, "learning_rate": 1.3217827382074583e-06, "loss": 0.3922, "step": 51925 }, { "epoch": 3.52, "learning_rate": 1.3208400889860868e-06, "loss": 0.0701, "step": 51930 }, { "epoch": 3.52, "learning_rate": 1.3198974397647149e-06, "loss": 0.1571, "step": 51935 }, { "epoch": 3.53, "learning_rate": 1.3189547905433432e-06, "loss": 0.2137, "step": 51940 }, { "epoch": 3.53, "learning_rate": 1.3180121413219712e-06, "loss": 0.3004, "step": 51945 }, { "epoch": 3.53, "learning_rate": 1.3170694921005997e-06, "loss": 0.167, "step": 51950 }, { "epoch": 3.53, "learning_rate": 1.3161268428792278e-06, "loss": 0.5093, "step": 51955 }, { "epoch": 3.53, "learning_rate": 1.315184193657856e-06, "loss": 0.2166, "step": 51960 }, { "epoch": 3.53, "learning_rate": 1.3142415444364845e-06, "loss": 0.084, "step": 51965 }, { "epoch": 3.53, "learning_rate": 1.3132988952151126e-06, "loss": 0.5597, "step": 51970 }, { "epoch": 3.53, "learning_rate": 1.3123562459937409e-06, "loss": 0.5927, "step": 51975 }, { "epoch": 3.53, "learning_rate": 1.3114135967723694e-06, "loss": 0.1517, "step": 51980 }, { "epoch": 3.53, "learning_rate": 1.3104709475509974e-06, "loss": 0.4107, "step": 51985 }, { "epoch": 3.53, "learning_rate": 1.3095282983296257e-06, "loss": 0.1307, "step": 51990 }, { "epoch": 3.53, "learning_rate": 1.3085856491082538e-06, "loss": 0.1406, "step": 51995 }, { "epoch": 3.53, "learning_rate": 1.3076429998868822e-06, "loss": 0.392, "step": 52000 }, { "epoch": 3.53, "learning_rate": 1.3067003506655105e-06, "loss": 0.0868, "step": 52005 }, { "epoch": 3.53, "learning_rate": 1.3057577014441386e-06, "loss": 0.4238, "step": 52010 }, { "epoch": 3.53, "learning_rate": 1.304815052222767e-06, "loss": 0.1052, "step": 52015 }, { "epoch": 3.53, "learning_rate": 1.3038724030013951e-06, "loss": 0.1665, "step": 52020 }, { "epoch": 3.53, "learning_rate": 1.3029297537800234e-06, "loss": 0.3866, "step": 52025 }, { "epoch": 3.53, "learning_rate": 1.301987104558652e-06, "loss": 0.4045, "step": 52030 }, { "epoch": 3.53, "learning_rate": 1.30104445533728e-06, "loss": 0.238, "step": 52035 }, { "epoch": 3.53, "learning_rate": 1.3001018061159082e-06, "loss": 0.1853, "step": 52040 }, { "epoch": 3.53, "learning_rate": 1.2991591568945367e-06, "loss": 0.1607, "step": 52045 }, { "epoch": 3.53, "learning_rate": 1.2982165076731648e-06, "loss": 0.4577, "step": 52050 }, { "epoch": 3.53, "learning_rate": 1.297273858451793e-06, "loss": 0.2363, "step": 52055 }, { "epoch": 3.53, "learning_rate": 1.2963312092304211e-06, "loss": 0.7251, "step": 52060 }, { "epoch": 3.53, "learning_rate": 1.2953885600090496e-06, "loss": 0.1401, "step": 52065 }, { "epoch": 3.53, "learning_rate": 1.2944459107876777e-06, "loss": 0.4875, "step": 52070 }, { "epoch": 3.53, "learning_rate": 1.293503261566306e-06, "loss": 0.0444, "step": 52075 }, { "epoch": 3.53, "learning_rate": 1.2925606123449344e-06, "loss": 0.2509, "step": 52080 }, { "epoch": 3.54, "learning_rate": 1.2916179631235625e-06, "loss": 0.3736, "step": 52085 }, { "epoch": 3.54, "learning_rate": 1.2906753139021908e-06, "loss": 0.1561, "step": 52090 }, { "epoch": 3.54, "learning_rate": 1.2897326646808193e-06, "loss": 0.2279, "step": 52095 }, { "epoch": 3.54, "learning_rate": 1.2887900154594473e-06, "loss": 0.8207, "step": 52100 }, { "epoch": 3.54, "learning_rate": 1.2878473662380756e-06, "loss": 0.1981, "step": 52105 }, { "epoch": 3.54, "learning_rate": 1.2869047170167037e-06, "loss": 0.0641, "step": 52110 }, { "epoch": 3.54, "learning_rate": 1.2859620677953322e-06, "loss": 0.2887, "step": 52115 }, { "epoch": 3.54, "learning_rate": 1.2850194185739604e-06, "loss": 0.182, "step": 52120 }, { "epoch": 3.54, "learning_rate": 1.2840767693525885e-06, "loss": 0.3874, "step": 52125 }, { "epoch": 3.54, "learning_rate": 1.283134120131217e-06, "loss": 0.6404, "step": 52130 }, { "epoch": 3.54, "learning_rate": 1.282191470909845e-06, "loss": 0.3738, "step": 52135 }, { "epoch": 3.54, "learning_rate": 1.2812488216884733e-06, "loss": 0.2144, "step": 52140 }, { "epoch": 3.54, "learning_rate": 1.2803061724671018e-06, "loss": 0.2224, "step": 52145 }, { "epoch": 3.54, "learning_rate": 1.2793635232457299e-06, "loss": 0.3628, "step": 52150 }, { "epoch": 3.54, "learning_rate": 1.2784208740243582e-06, "loss": 0.5181, "step": 52155 }, { "epoch": 3.54, "learning_rate": 1.2774782248029866e-06, "loss": 0.1193, "step": 52160 }, { "epoch": 3.54, "learning_rate": 1.2765355755816147e-06, "loss": 0.0932, "step": 52165 }, { "epoch": 3.54, "learning_rate": 1.275592926360243e-06, "loss": 0.3054, "step": 52170 }, { "epoch": 3.54, "learning_rate": 1.274650277138871e-06, "loss": 0.1421, "step": 52175 }, { "epoch": 3.54, "learning_rate": 1.2737076279174995e-06, "loss": 0.2468, "step": 52180 }, { "epoch": 3.54, "learning_rate": 1.2727649786961276e-06, "loss": 0.2145, "step": 52185 }, { "epoch": 3.54, "learning_rate": 1.2718223294747559e-06, "loss": 0.0686, "step": 52190 }, { "epoch": 3.54, "learning_rate": 1.2708796802533844e-06, "loss": 0.455, "step": 52195 }, { "epoch": 3.54, "learning_rate": 1.2699370310320124e-06, "loss": 0.2449, "step": 52200 }, { "epoch": 3.54, "learning_rate": 1.2689943818106407e-06, "loss": 0.4509, "step": 52205 }, { "epoch": 3.54, "learning_rate": 1.2680517325892692e-06, "loss": 0.1788, "step": 52210 }, { "epoch": 3.54, "learning_rate": 1.2671090833678973e-06, "loss": 0.341, "step": 52215 }, { "epoch": 3.54, "learning_rate": 1.2661664341465255e-06, "loss": 0.1545, "step": 52220 }, { "epoch": 3.54, "learning_rate": 1.2652237849251536e-06, "loss": 0.0718, "step": 52225 }, { "epoch": 3.54, "learning_rate": 1.264281135703782e-06, "loss": 0.2791, "step": 52230 }, { "epoch": 3.55, "learning_rate": 1.2633384864824104e-06, "loss": 0.5267, "step": 52235 }, { "epoch": 3.55, "learning_rate": 1.2623958372610384e-06, "loss": 0.2927, "step": 52240 }, { "epoch": 3.55, "learning_rate": 1.261453188039667e-06, "loss": 1.0534, "step": 52245 }, { "epoch": 3.55, "learning_rate": 1.260510538818295e-06, "loss": 0.4306, "step": 52250 }, { "epoch": 3.55, "learning_rate": 1.2595678895969232e-06, "loss": 0.0911, "step": 52255 }, { "epoch": 3.55, "learning_rate": 1.2586252403755517e-06, "loss": 0.0956, "step": 52260 }, { "epoch": 3.55, "learning_rate": 1.2576825911541798e-06, "loss": 0.0693, "step": 52265 }, { "epoch": 3.55, "learning_rate": 1.256739941932808e-06, "loss": 0.3343, "step": 52270 }, { "epoch": 3.55, "learning_rate": 1.2557972927114361e-06, "loss": 0.1396, "step": 52275 }, { "epoch": 3.55, "learning_rate": 1.2548546434900646e-06, "loss": 0.2389, "step": 52280 }, { "epoch": 3.55, "learning_rate": 1.253911994268693e-06, "loss": 0.3411, "step": 52285 }, { "epoch": 3.55, "learning_rate": 1.252969345047321e-06, "loss": 0.3703, "step": 52290 }, { "epoch": 3.55, "learning_rate": 1.2520266958259494e-06, "loss": 0.3107, "step": 52295 }, { "epoch": 3.55, "learning_rate": 1.2510840466045775e-06, "loss": 0.2892, "step": 52300 }, { "epoch": 3.55, "learning_rate": 1.2501413973832058e-06, "loss": 0.097, "step": 52305 }, { "epoch": 3.55, "learning_rate": 1.249198748161834e-06, "loss": 0.341, "step": 52310 }, { "epoch": 3.55, "learning_rate": 1.2482560989404623e-06, "loss": 0.3683, "step": 52315 }, { "epoch": 3.55, "learning_rate": 1.2473134497190906e-06, "loss": 0.6878, "step": 52320 }, { "epoch": 3.55, "learning_rate": 1.2463708004977189e-06, "loss": 0.2673, "step": 52325 }, { "epoch": 3.55, "learning_rate": 1.2454281512763472e-06, "loss": 0.2283, "step": 52330 }, { "epoch": 3.55, "learning_rate": 1.2444855020549754e-06, "loss": 0.3183, "step": 52335 }, { "epoch": 3.55, "learning_rate": 1.2435428528336037e-06, "loss": 0.1372, "step": 52340 }, { "epoch": 3.55, "learning_rate": 1.2426002036122318e-06, "loss": 0.1563, "step": 52345 }, { "epoch": 3.55, "learning_rate": 1.2416575543908603e-06, "loss": 0.4387, "step": 52350 }, { "epoch": 3.55, "learning_rate": 1.2407149051694885e-06, "loss": 0.0606, "step": 52355 }, { "epoch": 3.55, "learning_rate": 1.2397722559481166e-06, "loss": 0.0687, "step": 52360 }, { "epoch": 3.55, "learning_rate": 1.2388296067267449e-06, "loss": 0.167, "step": 52365 }, { "epoch": 3.55, "learning_rate": 1.2378869575053732e-06, "loss": 0.1295, "step": 52370 }, { "epoch": 3.55, "learning_rate": 1.2369443082840014e-06, "loss": 0.8023, "step": 52375 }, { "epoch": 3.56, "learning_rate": 1.2360016590626297e-06, "loss": 0.2515, "step": 52380 }, { "epoch": 3.56, "learning_rate": 1.235059009841258e-06, "loss": 0.1338, "step": 52385 }, { "epoch": 3.56, "learning_rate": 1.2341163606198863e-06, "loss": 0.3854, "step": 52390 }, { "epoch": 3.56, "learning_rate": 1.2331737113985145e-06, "loss": 0.1212, "step": 52395 }, { "epoch": 3.56, "learning_rate": 1.2322310621771428e-06, "loss": 0.1031, "step": 52400 }, { "epoch": 3.56, "learning_rate": 1.231288412955771e-06, "loss": 0.1964, "step": 52405 }, { "epoch": 3.56, "learning_rate": 1.2303457637343991e-06, "loss": 0.8053, "step": 52410 }, { "epoch": 3.56, "learning_rate": 1.2294031145130274e-06, "loss": 0.0436, "step": 52415 }, { "epoch": 3.56, "learning_rate": 1.228460465291656e-06, "loss": 0.0821, "step": 52420 }, { "epoch": 3.56, "learning_rate": 1.227517816070284e-06, "loss": 0.2158, "step": 52425 }, { "epoch": 3.56, "learning_rate": 1.2265751668489123e-06, "loss": 0.6218, "step": 52430 }, { "epoch": 3.56, "learning_rate": 1.2256325176275405e-06, "loss": 0.1625, "step": 52435 }, { "epoch": 3.56, "learning_rate": 1.2246898684061688e-06, "loss": 0.4451, "step": 52440 }, { "epoch": 3.56, "learning_rate": 1.223747219184797e-06, "loss": 0.1182, "step": 52445 }, { "epoch": 3.56, "learning_rate": 1.2228045699634254e-06, "loss": 0.268, "step": 52450 }, { "epoch": 3.56, "learning_rate": 1.2218619207420536e-06, "loss": 0.345, "step": 52455 }, { "epoch": 3.56, "learning_rate": 1.2209192715206817e-06, "loss": 0.1337, "step": 52460 }, { "epoch": 3.56, "learning_rate": 1.2199766222993102e-06, "loss": 0.1726, "step": 52465 }, { "epoch": 3.56, "learning_rate": 1.2190339730779385e-06, "loss": 0.0615, "step": 52470 }, { "epoch": 3.56, "learning_rate": 1.2180913238565665e-06, "loss": 0.1519, "step": 52475 }, { "epoch": 3.56, "learning_rate": 1.2171486746351948e-06, "loss": 0.3966, "step": 52480 }, { "epoch": 3.56, "learning_rate": 1.216206025413823e-06, "loss": 0.038, "step": 52485 }, { "epoch": 3.56, "learning_rate": 1.2152633761924513e-06, "loss": 0.1415, "step": 52490 }, { "epoch": 3.56, "learning_rate": 1.2143207269710796e-06, "loss": 0.2062, "step": 52495 }, { "epoch": 3.56, "learning_rate": 1.213378077749708e-06, "loss": 0.684, "step": 52500 }, { "epoch": 3.56, "learning_rate": 1.2124354285283362e-06, "loss": 0.3373, "step": 52505 }, { "epoch": 3.56, "learning_rate": 1.2114927793069642e-06, "loss": 0.1221, "step": 52510 }, { "epoch": 3.56, "learning_rate": 1.2105501300855927e-06, "loss": 0.1594, "step": 52515 }, { "epoch": 3.56, "learning_rate": 1.209607480864221e-06, "loss": 0.0583, "step": 52520 }, { "epoch": 3.56, "learning_rate": 1.208664831642849e-06, "loss": 0.5887, "step": 52525 }, { "epoch": 3.57, "learning_rate": 1.2077221824214773e-06, "loss": 0.3713, "step": 52530 }, { "epoch": 3.57, "learning_rate": 1.2067795332001058e-06, "loss": 0.5278, "step": 52535 }, { "epoch": 3.57, "learning_rate": 1.2058368839787339e-06, "loss": 0.3384, "step": 52540 }, { "epoch": 3.57, "learning_rate": 1.2048942347573622e-06, "loss": 0.222, "step": 52545 }, { "epoch": 3.57, "learning_rate": 1.2039515855359904e-06, "loss": 0.0936, "step": 52550 }, { "epoch": 3.57, "learning_rate": 1.2030089363146187e-06, "loss": 0.3898, "step": 52555 }, { "epoch": 3.57, "learning_rate": 1.202066287093247e-06, "loss": 0.1905, "step": 52560 }, { "epoch": 3.57, "learning_rate": 1.2011236378718753e-06, "loss": 0.0922, "step": 52565 }, { "epoch": 3.57, "learning_rate": 1.2001809886505035e-06, "loss": 0.1962, "step": 52570 }, { "epoch": 3.57, "learning_rate": 1.1992383394291316e-06, "loss": 0.4707, "step": 52575 }, { "epoch": 3.57, "learning_rate": 1.1982956902077599e-06, "loss": 0.292, "step": 52580 }, { "epoch": 3.57, "learning_rate": 1.1973530409863884e-06, "loss": 0.352, "step": 52585 }, { "epoch": 3.57, "learning_rate": 1.1964103917650164e-06, "loss": 0.6226, "step": 52590 }, { "epoch": 3.57, "learning_rate": 1.1954677425436447e-06, "loss": 0.2715, "step": 52595 }, { "epoch": 3.57, "learning_rate": 1.194525093322273e-06, "loss": 0.2079, "step": 52600 }, { "epoch": 3.57, "learning_rate": 1.1935824441009013e-06, "loss": 0.166, "step": 52605 }, { "epoch": 3.57, "learning_rate": 1.1926397948795295e-06, "loss": 0.2717, "step": 52610 }, { "epoch": 3.57, "learning_rate": 1.1916971456581578e-06, "loss": 0.3576, "step": 52615 }, { "epoch": 3.57, "learning_rate": 1.190754496436786e-06, "loss": 0.1619, "step": 52620 }, { "epoch": 3.57, "learning_rate": 1.1898118472154142e-06, "loss": 0.3747, "step": 52625 }, { "epoch": 3.57, "learning_rate": 1.1888691979940426e-06, "loss": 0.1326, "step": 52630 }, { "epoch": 3.57, "learning_rate": 1.187926548772671e-06, "loss": 0.1762, "step": 52635 }, { "epoch": 3.57, "learning_rate": 1.186983899551299e-06, "loss": 0.6468, "step": 52640 }, { "epoch": 3.57, "learning_rate": 1.1860412503299273e-06, "loss": 0.1323, "step": 52645 }, { "epoch": 3.57, "learning_rate": 1.1850986011085555e-06, "loss": 0.3599, "step": 52650 }, { "epoch": 3.57, "learning_rate": 1.1841559518871838e-06, "loss": 0.1496, "step": 52655 }, { "epoch": 3.57, "learning_rate": 1.183213302665812e-06, "loss": 0.2318, "step": 52660 }, { "epoch": 3.57, "learning_rate": 1.1822706534444404e-06, "loss": 0.2547, "step": 52665 }, { "epoch": 3.57, "learning_rate": 1.1813280042230686e-06, "loss": 0.069, "step": 52670 }, { "epoch": 3.58, "learning_rate": 1.180385355001697e-06, "loss": 0.2537, "step": 52675 }, { "epoch": 3.58, "learning_rate": 1.1794427057803252e-06, "loss": 0.3987, "step": 52680 }, { "epoch": 3.58, "learning_rate": 1.1785000565589535e-06, "loss": 0.0902, "step": 52685 }, { "epoch": 3.58, "learning_rate": 1.1775574073375815e-06, "loss": 0.0807, "step": 52690 }, { "epoch": 3.58, "learning_rate": 1.1766147581162098e-06, "loss": 0.4642, "step": 52695 }, { "epoch": 3.58, "learning_rate": 1.1756721088948383e-06, "loss": 0.4027, "step": 52700 }, { "epoch": 3.58, "learning_rate": 1.1747294596734663e-06, "loss": 0.0775, "step": 52705 }, { "epoch": 3.58, "learning_rate": 1.1737868104520946e-06, "loss": 0.1347, "step": 52710 }, { "epoch": 3.58, "learning_rate": 1.172844161230723e-06, "loss": 0.2317, "step": 52715 }, { "epoch": 3.58, "learning_rate": 1.1719015120093512e-06, "loss": 0.2915, "step": 52720 }, { "epoch": 3.58, "learning_rate": 1.1709588627879795e-06, "loss": 0.4707, "step": 52725 }, { "epoch": 3.58, "learning_rate": 1.1700162135666077e-06, "loss": 0.6215, "step": 52730 }, { "epoch": 3.58, "learning_rate": 1.169073564345236e-06, "loss": 0.2217, "step": 52735 }, { "epoch": 3.58, "learning_rate": 1.168130915123864e-06, "loss": 0.0322, "step": 52740 }, { "epoch": 3.58, "learning_rate": 1.1671882659024926e-06, "loss": 0.1231, "step": 52745 }, { "epoch": 3.58, "learning_rate": 1.1662456166811208e-06, "loss": 0.1715, "step": 52750 }, { "epoch": 3.58, "learning_rate": 1.1653029674597489e-06, "loss": 0.1881, "step": 52755 }, { "epoch": 3.58, "learning_rate": 1.1643603182383772e-06, "loss": 0.0463, "step": 52760 }, { "epoch": 3.58, "learning_rate": 1.1634176690170054e-06, "loss": 0.6266, "step": 52765 }, { "epoch": 3.58, "learning_rate": 1.1624750197956337e-06, "loss": 0.0653, "step": 52770 }, { "epoch": 3.58, "learning_rate": 1.161532370574262e-06, "loss": 0.6593, "step": 52775 }, { "epoch": 3.58, "learning_rate": 1.1605897213528903e-06, "loss": 0.2751, "step": 52780 }, { "epoch": 3.58, "learning_rate": 1.1596470721315185e-06, "loss": 0.162, "step": 52785 }, { "epoch": 3.58, "learning_rate": 1.1587044229101466e-06, "loss": 0.2299, "step": 52790 }, { "epoch": 3.58, "learning_rate": 1.157761773688775e-06, "loss": 0.4474, "step": 52795 }, { "epoch": 3.58, "learning_rate": 1.1568191244674034e-06, "loss": 0.2182, "step": 52800 }, { "epoch": 3.58, "learning_rate": 1.1558764752460314e-06, "loss": 0.2827, "step": 52805 }, { "epoch": 3.58, "learning_rate": 1.1549338260246597e-06, "loss": 0.3431, "step": 52810 }, { "epoch": 3.58, "learning_rate": 1.1539911768032882e-06, "loss": 0.7729, "step": 52815 }, { "epoch": 3.58, "learning_rate": 1.1530485275819163e-06, "loss": 0.2016, "step": 52820 }, { "epoch": 3.59, "learning_rate": 1.1521058783605445e-06, "loss": 0.4514, "step": 52825 }, { "epoch": 3.59, "learning_rate": 1.1511632291391728e-06, "loss": 0.1147, "step": 52830 }, { "epoch": 3.59, "learning_rate": 1.150220579917801e-06, "loss": 0.5678, "step": 52835 }, { "epoch": 3.59, "learning_rate": 1.1492779306964294e-06, "loss": 0.1994, "step": 52840 }, { "epoch": 3.59, "learning_rate": 1.1483352814750576e-06, "loss": 0.1944, "step": 52845 }, { "epoch": 3.59, "learning_rate": 1.147392632253686e-06, "loss": 0.0897, "step": 52850 }, { "epoch": 3.59, "learning_rate": 1.146449983032314e-06, "loss": 0.2482, "step": 52855 }, { "epoch": 3.59, "learning_rate": 1.1455073338109425e-06, "loss": 0.1972, "step": 52860 }, { "epoch": 3.59, "learning_rate": 1.1445646845895707e-06, "loss": 0.0453, "step": 52865 }, { "epoch": 3.59, "learning_rate": 1.1436220353681988e-06, "loss": 0.1582, "step": 52870 }, { "epoch": 3.59, "learning_rate": 1.142679386146827e-06, "loss": 0.0424, "step": 52875 }, { "epoch": 3.59, "learning_rate": 1.1417367369254554e-06, "loss": 0.1876, "step": 52880 }, { "epoch": 3.59, "learning_rate": 1.1407940877040836e-06, "loss": 0.1363, "step": 52885 }, { "epoch": 3.59, "learning_rate": 1.139851438482712e-06, "loss": 0.148, "step": 52890 }, { "epoch": 3.59, "learning_rate": 1.1389087892613402e-06, "loss": 0.448, "step": 52895 }, { "epoch": 3.59, "learning_rate": 1.1379661400399685e-06, "loss": 0.2042, "step": 52900 }, { "epoch": 3.59, "learning_rate": 1.1370234908185965e-06, "loss": 0.2234, "step": 52905 }, { "epoch": 3.59, "learning_rate": 1.136080841597225e-06, "loss": 0.267, "step": 52910 }, { "epoch": 3.59, "learning_rate": 1.1351381923758533e-06, "loss": 0.209, "step": 52915 }, { "epoch": 3.59, "learning_rate": 1.1341955431544813e-06, "loss": 0.2495, "step": 52920 }, { "epoch": 3.59, "learning_rate": 1.1332528939331096e-06, "loss": 0.3215, "step": 52925 }, { "epoch": 3.59, "learning_rate": 1.132310244711738e-06, "loss": 0.1647, "step": 52930 }, { "epoch": 3.59, "learning_rate": 1.1313675954903662e-06, "loss": 0.1768, "step": 52935 }, { "epoch": 3.59, "learning_rate": 1.1304249462689945e-06, "loss": 0.5168, "step": 52940 }, { "epoch": 3.59, "learning_rate": 1.1294822970476227e-06, "loss": 0.2715, "step": 52945 }, { "epoch": 3.59, "learning_rate": 1.128539647826251e-06, "loss": 0.1828, "step": 52950 }, { "epoch": 3.59, "learning_rate": 1.1275969986048793e-06, "loss": 0.1724, "step": 52955 }, { "epoch": 3.59, "learning_rate": 1.1266543493835076e-06, "loss": 0.2188, "step": 52960 }, { "epoch": 3.59, "learning_rate": 1.1257117001621358e-06, "loss": 0.21, "step": 52965 }, { "epoch": 3.6, "learning_rate": 1.1247690509407639e-06, "loss": 0.1801, "step": 52970 }, { "epoch": 3.6, "learning_rate": 1.1238264017193924e-06, "loss": 0.121, "step": 52975 }, { "epoch": 3.6, "learning_rate": 1.1228837524980207e-06, "loss": 0.2545, "step": 52980 }, { "epoch": 3.6, "learning_rate": 1.1219411032766487e-06, "loss": 0.3643, "step": 52985 }, { "epoch": 3.6, "learning_rate": 1.120998454055277e-06, "loss": 0.3181, "step": 52990 }, { "epoch": 3.6, "learning_rate": 1.1200558048339053e-06, "loss": 0.4113, "step": 52995 }, { "epoch": 3.6, "learning_rate": 1.1191131556125335e-06, "loss": 0.4594, "step": 53000 }, { "epoch": 3.6, "learning_rate": 1.1181705063911618e-06, "loss": 0.5163, "step": 53005 }, { "epoch": 3.6, "learning_rate": 1.11722785716979e-06, "loss": 0.3802, "step": 53010 }, { "epoch": 3.6, "learning_rate": 1.1162852079484184e-06, "loss": 0.0893, "step": 53015 }, { "epoch": 3.6, "learning_rate": 1.1153425587270464e-06, "loss": 0.2756, "step": 53020 }, { "epoch": 3.6, "learning_rate": 1.114399909505675e-06, "loss": 0.9127, "step": 53025 }, { "epoch": 3.6, "learning_rate": 1.1134572602843032e-06, "loss": 0.1963, "step": 53030 }, { "epoch": 3.6, "learning_rate": 1.1125146110629313e-06, "loss": 0.0823, "step": 53035 }, { "epoch": 3.6, "learning_rate": 1.1115719618415595e-06, "loss": 0.3539, "step": 53040 }, { "epoch": 3.6, "learning_rate": 1.1106293126201878e-06, "loss": 0.1719, "step": 53045 }, { "epoch": 3.6, "learning_rate": 1.109686663398816e-06, "loss": 0.7453, "step": 53050 }, { "epoch": 3.6, "learning_rate": 1.1087440141774444e-06, "loss": 0.142, "step": 53055 }, { "epoch": 3.6, "learning_rate": 1.1078013649560726e-06, "loss": 0.2968, "step": 53060 }, { "epoch": 3.6, "learning_rate": 1.106858715734701e-06, "loss": 0.1849, "step": 53065 }, { "epoch": 3.6, "learning_rate": 1.105916066513329e-06, "loss": 0.1776, "step": 53070 }, { "epoch": 3.6, "learning_rate": 1.1049734172919575e-06, "loss": 0.0507, "step": 53075 }, { "epoch": 3.6, "learning_rate": 1.1040307680705857e-06, "loss": 0.3316, "step": 53080 }, { "epoch": 3.6, "learning_rate": 1.1030881188492138e-06, "loss": 0.1384, "step": 53085 }, { "epoch": 3.6, "learning_rate": 1.1021454696278423e-06, "loss": 0.2405, "step": 53090 }, { "epoch": 3.6, "learning_rate": 1.1012028204064706e-06, "loss": 0.1052, "step": 53095 }, { "epoch": 3.6, "learning_rate": 1.1002601711850986e-06, "loss": 0.2066, "step": 53100 }, { "epoch": 3.6, "learning_rate": 1.099317521963727e-06, "loss": 0.3073, "step": 53105 }, { "epoch": 3.6, "learning_rate": 1.0983748727423552e-06, "loss": 0.2094, "step": 53110 }, { "epoch": 3.6, "learning_rate": 1.0974322235209835e-06, "loss": 0.1638, "step": 53115 }, { "epoch": 3.61, "learning_rate": 1.0964895742996117e-06, "loss": 0.11, "step": 53120 }, { "epoch": 3.61, "learning_rate": 1.09554692507824e-06, "loss": 0.2081, "step": 53125 }, { "epoch": 3.61, "learning_rate": 1.0946042758568683e-06, "loss": 0.0381, "step": 53130 }, { "epoch": 3.61, "learning_rate": 1.0936616266354964e-06, "loss": 0.0897, "step": 53135 }, { "epoch": 3.61, "learning_rate": 1.0927189774141248e-06, "loss": 0.367, "step": 53140 }, { "epoch": 3.61, "learning_rate": 1.0917763281927531e-06, "loss": 0.2647, "step": 53145 }, { "epoch": 3.61, "learning_rate": 1.0908336789713812e-06, "loss": 0.1962, "step": 53150 }, { "epoch": 3.61, "learning_rate": 1.0898910297500095e-06, "loss": 0.3646, "step": 53155 }, { "epoch": 3.61, "learning_rate": 1.0889483805286377e-06, "loss": 0.4006, "step": 53160 }, { "epoch": 3.61, "learning_rate": 1.088005731307266e-06, "loss": 0.1923, "step": 53165 }, { "epoch": 3.61, "learning_rate": 1.0870630820858943e-06, "loss": 0.1559, "step": 53170 }, { "epoch": 3.61, "learning_rate": 1.0861204328645226e-06, "loss": 0.3325, "step": 53175 }, { "epoch": 3.61, "learning_rate": 1.0851777836431508e-06, "loss": 0.1515, "step": 53180 }, { "epoch": 3.61, "learning_rate": 1.084235134421779e-06, "loss": 0.2064, "step": 53185 }, { "epoch": 3.61, "learning_rate": 1.0832924852004074e-06, "loss": 0.6932, "step": 53190 }, { "epoch": 3.61, "learning_rate": 1.0823498359790357e-06, "loss": 0.2572, "step": 53195 }, { "epoch": 3.61, "learning_rate": 1.0814071867576637e-06, "loss": 0.2895, "step": 53200 }, { "epoch": 3.61, "learning_rate": 1.080464537536292e-06, "loss": 0.177, "step": 53205 }, { "epoch": 3.61, "learning_rate": 1.0795218883149203e-06, "loss": 0.3136, "step": 53210 }, { "epoch": 3.61, "learning_rate": 1.0785792390935485e-06, "loss": 0.4362, "step": 53215 }, { "epoch": 3.61, "learning_rate": 1.0776365898721768e-06, "loss": 0.4363, "step": 53220 }, { "epoch": 3.61, "learning_rate": 1.076693940650805e-06, "loss": 0.3503, "step": 53225 }, { "epoch": 3.61, "learning_rate": 1.0757512914294334e-06, "loss": 0.4632, "step": 53230 }, { "epoch": 3.61, "learning_rate": 1.0748086422080617e-06, "loss": 0.0732, "step": 53235 }, { "epoch": 3.61, "learning_rate": 1.07386599298669e-06, "loss": 0.6963, "step": 53240 }, { "epoch": 3.61, "learning_rate": 1.0729233437653182e-06, "loss": 0.2077, "step": 53245 }, { "epoch": 3.61, "learning_rate": 1.0719806945439463e-06, "loss": 0.2333, "step": 53250 }, { "epoch": 3.61, "learning_rate": 1.0710380453225748e-06, "loss": 0.1047, "step": 53255 }, { "epoch": 3.61, "learning_rate": 1.070095396101203e-06, "loss": 0.5793, "step": 53260 }, { "epoch": 3.62, "learning_rate": 1.069152746879831e-06, "loss": 0.0832, "step": 53265 }, { "epoch": 3.62, "learning_rate": 1.0682100976584594e-06, "loss": 0.1642, "step": 53270 }, { "epoch": 3.62, "learning_rate": 1.0672674484370876e-06, "loss": 0.1359, "step": 53275 }, { "epoch": 3.62, "learning_rate": 1.066324799215716e-06, "loss": 0.3171, "step": 53280 }, { "epoch": 3.62, "learning_rate": 1.0653821499943442e-06, "loss": 0.2816, "step": 53285 }, { "epoch": 3.62, "learning_rate": 1.0644395007729725e-06, "loss": 0.2663, "step": 53290 }, { "epoch": 3.62, "learning_rate": 1.0634968515516007e-06, "loss": 0.5066, "step": 53295 }, { "epoch": 3.62, "learning_rate": 1.0625542023302288e-06, "loss": 0.4605, "step": 53300 }, { "epoch": 3.62, "learning_rate": 1.0616115531088573e-06, "loss": 0.4544, "step": 53305 }, { "epoch": 3.62, "learning_rate": 1.0606689038874856e-06, "loss": 0.043, "step": 53310 }, { "epoch": 3.62, "learning_rate": 1.0597262546661136e-06, "loss": 0.1005, "step": 53315 }, { "epoch": 3.62, "learning_rate": 1.058783605444742e-06, "loss": 0.333, "step": 53320 }, { "epoch": 3.62, "learning_rate": 1.0578409562233702e-06, "loss": 0.087, "step": 53325 }, { "epoch": 3.62, "learning_rate": 1.0568983070019985e-06, "loss": 0.4512, "step": 53330 }, { "epoch": 3.62, "learning_rate": 1.0559556577806267e-06, "loss": 0.1251, "step": 53335 }, { "epoch": 3.62, "learning_rate": 1.055013008559255e-06, "loss": 0.2717, "step": 53340 }, { "epoch": 3.62, "learning_rate": 1.0540703593378833e-06, "loss": 0.5374, "step": 53345 }, { "epoch": 3.62, "learning_rate": 1.0531277101165116e-06, "loss": 0.1219, "step": 53350 }, { "epoch": 3.62, "learning_rate": 1.0521850608951398e-06, "loss": 0.3015, "step": 53355 }, { "epoch": 3.62, "learning_rate": 1.0512424116737681e-06, "loss": 0.3579, "step": 53360 }, { "epoch": 3.62, "learning_rate": 1.0502997624523962e-06, "loss": 0.1758, "step": 53365 }, { "epoch": 3.62, "learning_rate": 1.0493571132310247e-06, "loss": 0.6256, "step": 53370 }, { "epoch": 3.62, "learning_rate": 1.048414464009653e-06, "loss": 0.1043, "step": 53375 }, { "epoch": 3.62, "learning_rate": 1.047471814788281e-06, "loss": 0.4311, "step": 53380 }, { "epoch": 3.62, "learning_rate": 1.0465291655669093e-06, "loss": 0.5476, "step": 53385 }, { "epoch": 3.62, "learning_rate": 1.0455865163455376e-06, "loss": 0.1257, "step": 53390 }, { "epoch": 3.62, "learning_rate": 1.0446438671241658e-06, "loss": 0.186, "step": 53395 }, { "epoch": 3.62, "learning_rate": 1.0437012179027941e-06, "loss": 0.1178, "step": 53400 }, { "epoch": 3.62, "learning_rate": 1.0427585686814224e-06, "loss": 0.0343, "step": 53405 }, { "epoch": 3.62, "learning_rate": 1.0418159194600507e-06, "loss": 0.5345, "step": 53410 }, { "epoch": 3.63, "learning_rate": 1.0408732702386787e-06, "loss": 0.03, "step": 53415 }, { "epoch": 3.63, "learning_rate": 1.0399306210173072e-06, "loss": 0.1114, "step": 53420 }, { "epoch": 3.63, "learning_rate": 1.0389879717959355e-06, "loss": 0.0611, "step": 53425 }, { "epoch": 3.63, "learning_rate": 1.0380453225745635e-06, "loss": 0.2717, "step": 53430 }, { "epoch": 3.63, "learning_rate": 1.0371026733531918e-06, "loss": 0.476, "step": 53435 }, { "epoch": 3.63, "learning_rate": 1.03616002413182e-06, "loss": 0.156, "step": 53440 }, { "epoch": 3.63, "learning_rate": 1.0352173749104484e-06, "loss": 0.2972, "step": 53445 }, { "epoch": 3.63, "learning_rate": 1.0342747256890767e-06, "loss": 0.3028, "step": 53450 }, { "epoch": 3.63, "learning_rate": 1.033332076467705e-06, "loss": 0.0459, "step": 53455 }, { "epoch": 3.63, "learning_rate": 1.0323894272463332e-06, "loss": 0.1429, "step": 53460 }, { "epoch": 3.63, "learning_rate": 1.0314467780249615e-06, "loss": 0.4871, "step": 53465 }, { "epoch": 3.63, "learning_rate": 1.0305041288035898e-06, "loss": 0.2488, "step": 53470 }, { "epoch": 3.63, "learning_rate": 1.029561479582218e-06, "loss": 0.0961, "step": 53475 }, { "epoch": 3.63, "learning_rate": 1.028618830360846e-06, "loss": 0.4568, "step": 53480 }, { "epoch": 3.63, "learning_rate": 1.0276761811394746e-06, "loss": 0.4144, "step": 53485 }, { "epoch": 3.63, "learning_rate": 1.0267335319181026e-06, "loss": 0.347, "step": 53490 }, { "epoch": 3.63, "learning_rate": 1.025790882696731e-06, "loss": 0.0758, "step": 53495 }, { "epoch": 3.63, "learning_rate": 1.0248482334753592e-06, "loss": 0.1867, "step": 53500 }, { "epoch": 3.63, "learning_rate": 1.0239055842539875e-06, "loss": 0.2875, "step": 53505 }, { "epoch": 3.63, "learning_rate": 1.0229629350326157e-06, "loss": 0.3468, "step": 53510 }, { "epoch": 3.63, "learning_rate": 1.022020285811244e-06, "loss": 0.6202, "step": 53515 }, { "epoch": 3.63, "learning_rate": 1.0210776365898723e-06, "loss": 0.3449, "step": 53520 }, { "epoch": 3.63, "learning_rate": 1.0201349873685006e-06, "loss": 0.0171, "step": 53525 }, { "epoch": 3.63, "learning_rate": 1.0191923381471286e-06, "loss": 0.2152, "step": 53530 }, { "epoch": 3.63, "learning_rate": 1.0182496889257571e-06, "loss": 0.0784, "step": 53535 }, { "epoch": 3.63, "learning_rate": 1.0173070397043854e-06, "loss": 0.1064, "step": 53540 }, { "epoch": 3.63, "learning_rate": 1.0163643904830135e-06, "loss": 0.338, "step": 53545 }, { "epoch": 3.63, "learning_rate": 1.0154217412616417e-06, "loss": 0.1825, "step": 53550 }, { "epoch": 3.63, "learning_rate": 1.01447909204027e-06, "loss": 0.1849, "step": 53555 }, { "epoch": 3.64, "learning_rate": 1.0135364428188983e-06, "loss": 0.0255, "step": 53560 }, { "epoch": 3.64, "learning_rate": 1.0125937935975266e-06, "loss": 0.5307, "step": 53565 }, { "epoch": 3.64, "learning_rate": 1.0116511443761548e-06, "loss": 0.0771, "step": 53570 }, { "epoch": 3.64, "learning_rate": 1.0107084951547831e-06, "loss": 0.029, "step": 53575 }, { "epoch": 3.64, "learning_rate": 1.0097658459334114e-06, "loss": 0.2598, "step": 53580 }, { "epoch": 3.64, "learning_rate": 1.0088231967120397e-06, "loss": 0.2684, "step": 53585 }, { "epoch": 3.64, "learning_rate": 1.007880547490668e-06, "loss": 0.2622, "step": 53590 }, { "epoch": 3.64, "learning_rate": 1.006937898269296e-06, "loss": 0.3172, "step": 53595 }, { "epoch": 3.64, "learning_rate": 1.0059952490479245e-06, "loss": 0.1103, "step": 53600 }, { "epoch": 3.64, "learning_rate": 1.0050525998265526e-06, "loss": 0.1432, "step": 53605 }, { "epoch": 3.64, "learning_rate": 1.0041099506051808e-06, "loss": 0.0673, "step": 53610 }, { "epoch": 3.64, "learning_rate": 1.0031673013838091e-06, "loss": 0.3731, "step": 53615 }, { "epoch": 3.64, "learning_rate": 1.0022246521624374e-06, "loss": 0.2375, "step": 53620 }, { "epoch": 3.64, "learning_rate": 1.0012820029410657e-06, "loss": 0.1442, "step": 53625 }, { "epoch": 3.64, "learning_rate": 1.000339353719694e-06, "loss": 0.538, "step": 53630 }, { "epoch": 3.64, "learning_rate": 9.993967044983222e-07, "loss": 0.0574, "step": 53635 }, { "epoch": 3.64, "learning_rate": 9.984540552769505e-07, "loss": 0.6775, "step": 53640 }, { "epoch": 3.64, "learning_rate": 9.975114060555786e-07, "loss": 0.2103, "step": 53645 }, { "epoch": 3.64, "learning_rate": 9.96568756834207e-07, "loss": 0.414, "step": 53650 }, { "epoch": 3.64, "learning_rate": 9.95626107612835e-07, "loss": 0.2946, "step": 53655 }, { "epoch": 3.64, "learning_rate": 9.946834583914634e-07, "loss": 0.1409, "step": 53660 }, { "epoch": 3.64, "learning_rate": 9.937408091700917e-07, "loss": 1.1422, "step": 53665 }, { "epoch": 3.64, "learning_rate": 9.9279815994872e-07, "loss": 0.3178, "step": 53670 }, { "epoch": 3.64, "learning_rate": 9.918555107273482e-07, "loss": 0.1757, "step": 53675 }, { "epoch": 3.64, "learning_rate": 9.909128615059765e-07, "loss": 0.1882, "step": 53680 }, { "epoch": 3.64, "learning_rate": 9.899702122846048e-07, "loss": 0.2426, "step": 53685 }, { "epoch": 3.64, "learning_rate": 9.89027563063233e-07, "loss": 0.27, "step": 53690 }, { "epoch": 3.64, "learning_rate": 9.88084913841861e-07, "loss": 0.1559, "step": 53695 }, { "epoch": 3.64, "learning_rate": 9.871422646204896e-07, "loss": 0.3896, "step": 53700 }, { "epoch": 3.64, "learning_rate": 9.861996153991179e-07, "loss": 0.4852, "step": 53705 }, { "epoch": 3.65, "learning_rate": 9.85256966177746e-07, "loss": 0.1468, "step": 53710 }, { "epoch": 3.65, "learning_rate": 9.843143169563744e-07, "loss": 0.1766, "step": 53715 }, { "epoch": 3.65, "learning_rate": 9.833716677350025e-07, "loss": 0.1828, "step": 53720 }, { "epoch": 3.65, "learning_rate": 9.824290185136307e-07, "loss": 0.3367, "step": 53725 }, { "epoch": 3.65, "learning_rate": 9.81486369292259e-07, "loss": 0.0169, "step": 53730 }, { "epoch": 3.65, "learning_rate": 9.805437200708873e-07, "loss": 0.0964, "step": 53735 }, { "epoch": 3.65, "learning_rate": 9.796010708495156e-07, "loss": 0.2243, "step": 53740 }, { "epoch": 3.65, "learning_rate": 9.786584216281439e-07, "loss": 0.1627, "step": 53745 }, { "epoch": 3.65, "learning_rate": 9.777157724067721e-07, "loss": 0.3339, "step": 53750 }, { "epoch": 3.65, "learning_rate": 9.767731231854004e-07, "loss": 0.3101, "step": 53755 }, { "epoch": 3.65, "learning_rate": 9.758304739640285e-07, "loss": 0.0713, "step": 53760 }, { "epoch": 3.65, "learning_rate": 9.74887824742657e-07, "loss": 0.0888, "step": 53765 }, { "epoch": 3.65, "learning_rate": 9.73945175521285e-07, "loss": 0.1202, "step": 53770 }, { "epoch": 3.65, "learning_rate": 9.730025262999133e-07, "loss": 0.25, "step": 53775 }, { "epoch": 3.65, "learning_rate": 9.720598770785416e-07, "loss": 0.3233, "step": 53780 }, { "epoch": 3.65, "learning_rate": 9.711172278571698e-07, "loss": 0.2258, "step": 53785 }, { "epoch": 3.65, "learning_rate": 9.701745786357981e-07, "loss": 0.5142, "step": 53790 }, { "epoch": 3.65, "learning_rate": 9.692319294144264e-07, "loss": 0.6214, "step": 53795 }, { "epoch": 3.65, "learning_rate": 9.682892801930547e-07, "loss": 0.3044, "step": 53800 }, { "epoch": 3.65, "learning_rate": 9.67346630971683e-07, "loss": 0.7379, "step": 53805 }, { "epoch": 3.65, "learning_rate": 9.66403981750311e-07, "loss": 0.0377, "step": 53810 }, { "epoch": 3.65, "learning_rate": 9.654613325289395e-07, "loss": 0.1617, "step": 53815 }, { "epoch": 3.65, "learning_rate": 9.645186833075678e-07, "loss": 0.1778, "step": 53820 }, { "epoch": 3.65, "learning_rate": 9.635760340861958e-07, "loss": 0.3776, "step": 53825 }, { "epoch": 3.65, "learning_rate": 9.626333848648241e-07, "loss": 0.2011, "step": 53830 }, { "epoch": 3.65, "learning_rate": 9.616907356434524e-07, "loss": 0.6737, "step": 53835 }, { "epoch": 3.65, "learning_rate": 9.607480864220807e-07, "loss": 0.545, "step": 53840 }, { "epoch": 3.65, "learning_rate": 9.59805437200709e-07, "loss": 0.314, "step": 53845 }, { "epoch": 3.65, "learning_rate": 9.588627879793372e-07, "loss": 0.1044, "step": 53850 }, { "epoch": 3.66, "learning_rate": 9.579201387579655e-07, "loss": 0.153, "step": 53855 }, { "epoch": 3.66, "learning_rate": 9.569774895365938e-07, "loss": 0.2385, "step": 53860 }, { "epoch": 3.66, "learning_rate": 9.56034840315222e-07, "loss": 0.4267, "step": 53865 }, { "epoch": 3.66, "learning_rate": 9.550921910938503e-07, "loss": 0.1489, "step": 53870 }, { "epoch": 3.66, "learning_rate": 9.541495418724784e-07, "loss": 0.1179, "step": 53875 }, { "epoch": 3.66, "learning_rate": 9.532068926511068e-07, "loss": 0.0812, "step": 53880 }, { "epoch": 3.66, "learning_rate": 9.522642434297349e-07, "loss": 0.3742, "step": 53885 }, { "epoch": 3.66, "learning_rate": 9.513215942083632e-07, "loss": 0.3049, "step": 53890 }, { "epoch": 3.66, "learning_rate": 9.503789449869916e-07, "loss": 0.5246, "step": 53895 }, { "epoch": 3.66, "learning_rate": 9.494362957656198e-07, "loss": 0.1791, "step": 53900 }, { "epoch": 3.66, "learning_rate": 9.48493646544248e-07, "loss": 0.523, "step": 53905 }, { "epoch": 3.66, "learning_rate": 9.475509973228762e-07, "loss": 0.2606, "step": 53910 }, { "epoch": 3.66, "learning_rate": 9.466083481015046e-07, "loss": 0.1928, "step": 53915 }, { "epoch": 3.66, "learning_rate": 9.456656988801329e-07, "loss": 0.0683, "step": 53920 }, { "epoch": 3.66, "learning_rate": 9.44723049658761e-07, "loss": 0.3626, "step": 53925 }, { "epoch": 3.66, "learning_rate": 9.437804004373893e-07, "loss": 0.0874, "step": 53930 }, { "epoch": 3.66, "learning_rate": 9.428377512160175e-07, "loss": 0.5038, "step": 53935 }, { "epoch": 3.66, "learning_rate": 9.418951019946459e-07, "loss": 0.5191, "step": 53940 }, { "epoch": 3.66, "learning_rate": 9.409524527732741e-07, "loss": 0.2499, "step": 53945 }, { "epoch": 3.66, "learning_rate": 9.400098035519023e-07, "loss": 0.357, "step": 53950 }, { "epoch": 3.66, "learning_rate": 9.390671543305306e-07, "loss": 0.1555, "step": 53955 }, { "epoch": 3.66, "learning_rate": 9.38124505109159e-07, "loss": 0.1165, "step": 53960 }, { "epoch": 3.66, "learning_rate": 9.371818558877871e-07, "loss": 0.0954, "step": 53965 }, { "epoch": 3.66, "learning_rate": 9.362392066664154e-07, "loss": 0.2206, "step": 53970 }, { "epoch": 3.66, "learning_rate": 9.352965574450436e-07, "loss": 0.0647, "step": 53975 }, { "epoch": 3.66, "learning_rate": 9.343539082236718e-07, "loss": 0.1779, "step": 53980 }, { "epoch": 3.66, "learning_rate": 9.334112590023002e-07, "loss": 0.2642, "step": 53985 }, { "epoch": 3.66, "learning_rate": 9.324686097809284e-07, "loss": 0.0587, "step": 53990 }, { "epoch": 3.66, "learning_rate": 9.315259605595567e-07, "loss": 0.1589, "step": 53995 }, { "epoch": 3.66, "learning_rate": 9.305833113381848e-07, "loss": 0.0756, "step": 54000 }, { "epoch": 3.67, "learning_rate": 9.296406621168131e-07, "loss": 0.8699, "step": 54005 }, { "epoch": 3.67, "learning_rate": 9.286980128954415e-07, "loss": 0.4154, "step": 54010 }, { "epoch": 3.67, "learning_rate": 9.277553636740697e-07, "loss": 0.2338, "step": 54015 }, { "epoch": 3.67, "learning_rate": 9.268127144526979e-07, "loss": 0.2726, "step": 54020 }, { "epoch": 3.67, "learning_rate": 9.258700652313261e-07, "loss": 0.2269, "step": 54025 }, { "epoch": 3.67, "learning_rate": 9.249274160099545e-07, "loss": 0.7698, "step": 54030 }, { "epoch": 3.67, "learning_rate": 9.239847667885828e-07, "loss": 0.2602, "step": 54035 }, { "epoch": 3.67, "learning_rate": 9.230421175672109e-07, "loss": 0.3261, "step": 54040 }, { "epoch": 3.67, "learning_rate": 9.220994683458392e-07, "loss": 0.2531, "step": 54045 }, { "epoch": 3.67, "learning_rate": 9.211568191244674e-07, "loss": 0.1763, "step": 54050 }, { "epoch": 3.67, "learning_rate": 9.202141699030958e-07, "loss": 0.0772, "step": 54055 }, { "epoch": 3.67, "learning_rate": 9.19271520681724e-07, "loss": 0.4048, "step": 54060 }, { "epoch": 3.67, "learning_rate": 9.183288714603522e-07, "loss": 0.3803, "step": 54065 }, { "epoch": 3.67, "learning_rate": 9.173862222389805e-07, "loss": 0.2365, "step": 54070 }, { "epoch": 3.67, "learning_rate": 9.164435730176087e-07, "loss": 0.282, "step": 54075 }, { "epoch": 3.67, "learning_rate": 9.15500923796237e-07, "loss": 0.3077, "step": 54080 }, { "epoch": 3.67, "learning_rate": 9.145582745748653e-07, "loss": 0.1775, "step": 54085 }, { "epoch": 3.67, "learning_rate": 9.136156253534935e-07, "loss": 0.2465, "step": 54090 }, { "epoch": 3.67, "learning_rate": 9.126729761321218e-07, "loss": 0.3179, "step": 54095 }, { "epoch": 3.67, "learning_rate": 9.117303269107501e-07, "loss": 0.2468, "step": 54100 }, { "epoch": 3.67, "learning_rate": 9.107876776893783e-07, "loss": 0.1863, "step": 54105 }, { "epoch": 3.67, "learning_rate": 9.098450284680066e-07, "loss": 0.3779, "step": 54110 }, { "epoch": 3.67, "learning_rate": 9.089023792466348e-07, "loss": 0.2065, "step": 54115 }, { "epoch": 3.67, "learning_rate": 9.07959730025263e-07, "loss": 0.2995, "step": 54120 }, { "epoch": 3.67, "learning_rate": 9.070170808038914e-07, "loss": 0.6599, "step": 54125 }, { "epoch": 3.67, "learning_rate": 9.060744315825196e-07, "loss": 0.0503, "step": 54130 }, { "epoch": 3.67, "learning_rate": 9.051317823611479e-07, "loss": 0.3772, "step": 54135 }, { "epoch": 3.67, "learning_rate": 9.04189133139776e-07, "loss": 0.5122, "step": 54140 }, { "epoch": 3.67, "learning_rate": 9.032464839184044e-07, "loss": 0.2601, "step": 54145 }, { "epoch": 3.68, "learning_rate": 9.023038346970327e-07, "loss": 0.1301, "step": 54150 }, { "epoch": 3.68, "learning_rate": 9.013611854756609e-07, "loss": 0.4352, "step": 54155 }, { "epoch": 3.68, "learning_rate": 9.004185362542891e-07, "loss": 0.0986, "step": 54160 }, { "epoch": 3.68, "learning_rate": 8.994758870329173e-07, "loss": 0.0738, "step": 54165 }, { "epoch": 3.68, "learning_rate": 8.985332378115457e-07, "loss": 0.1494, "step": 54170 }, { "epoch": 3.68, "learning_rate": 8.97590588590174e-07, "loss": 0.5473, "step": 54175 }, { "epoch": 3.68, "learning_rate": 8.966479393688021e-07, "loss": 0.0462, "step": 54180 }, { "epoch": 3.68, "learning_rate": 8.957052901474304e-07, "loss": 0.2652, "step": 54185 }, { "epoch": 3.68, "learning_rate": 8.947626409260586e-07, "loss": 0.1621, "step": 54190 }, { "epoch": 3.68, "learning_rate": 8.93819991704687e-07, "loss": 0.3823, "step": 54195 }, { "epoch": 3.68, "learning_rate": 8.928773424833152e-07, "loss": 0.247, "step": 54200 }, { "epoch": 3.68, "learning_rate": 8.919346932619434e-07, "loss": 0.0786, "step": 54205 }, { "epoch": 3.68, "learning_rate": 8.909920440405717e-07, "loss": 0.0231, "step": 54210 }, { "epoch": 3.68, "learning_rate": 8.900493948191998e-07, "loss": 0.163, "step": 54215 }, { "epoch": 3.68, "learning_rate": 8.891067455978282e-07, "loss": 0.3001, "step": 54220 }, { "epoch": 3.68, "learning_rate": 8.881640963764565e-07, "loss": 0.0871, "step": 54225 }, { "epoch": 3.68, "learning_rate": 8.872214471550847e-07, "loss": 0.1794, "step": 54230 }, { "epoch": 3.68, "learning_rate": 8.86278797933713e-07, "loss": 0.3634, "step": 54235 }, { "epoch": 3.68, "learning_rate": 8.853361487123413e-07, "loss": 0.3742, "step": 54240 }, { "epoch": 3.68, "learning_rate": 8.843934994909695e-07, "loss": 0.1997, "step": 54245 }, { "epoch": 3.68, "learning_rate": 8.834508502695978e-07, "loss": 0.2429, "step": 54250 }, { "epoch": 3.68, "learning_rate": 8.825082010482259e-07, "loss": 0.2587, "step": 54255 }, { "epoch": 3.68, "learning_rate": 8.815655518268542e-07, "loss": 0.3441, "step": 54260 }, { "epoch": 3.68, "learning_rate": 8.806229026054826e-07, "loss": 0.1156, "step": 54265 }, { "epoch": 3.68, "learning_rate": 8.796802533841108e-07, "loss": 0.1734, "step": 54270 }, { "epoch": 3.68, "learning_rate": 8.78737604162739e-07, "loss": 0.5918, "step": 54275 }, { "epoch": 3.68, "learning_rate": 8.777949549413672e-07, "loss": 0.3945, "step": 54280 }, { "epoch": 3.68, "learning_rate": 8.768523057199956e-07, "loss": 0.203, "step": 54285 }, { "epoch": 3.68, "learning_rate": 8.759096564986239e-07, "loss": 0.4133, "step": 54290 }, { "epoch": 3.69, "learning_rate": 8.74967007277252e-07, "loss": 0.2772, "step": 54295 }, { "epoch": 3.69, "learning_rate": 8.740243580558803e-07, "loss": 0.0657, "step": 54300 }, { "epoch": 3.69, "learning_rate": 8.730817088345085e-07, "loss": 0.3073, "step": 54305 }, { "epoch": 3.69, "learning_rate": 8.721390596131369e-07, "loss": 0.1012, "step": 54310 }, { "epoch": 3.69, "learning_rate": 8.711964103917651e-07, "loss": 0.255, "step": 54315 }, { "epoch": 3.69, "learning_rate": 8.702537611703933e-07, "loss": 0.2837, "step": 54320 }, { "epoch": 3.69, "learning_rate": 8.693111119490216e-07, "loss": 0.1209, "step": 54325 }, { "epoch": 3.69, "learning_rate": 8.683684627276498e-07, "loss": 0.5981, "step": 54330 }, { "epoch": 3.69, "learning_rate": 8.674258135062781e-07, "loss": 0.2659, "step": 54335 }, { "epoch": 3.69, "learning_rate": 8.664831642849064e-07, "loss": 0.221, "step": 54340 }, { "epoch": 3.69, "learning_rate": 8.655405150635346e-07, "loss": 0.5575, "step": 54345 }, { "epoch": 3.69, "learning_rate": 8.645978658421629e-07, "loss": 0.1887, "step": 54350 }, { "epoch": 3.69, "learning_rate": 8.63655216620791e-07, "loss": 0.2999, "step": 54355 }, { "epoch": 3.69, "learning_rate": 8.627125673994194e-07, "loss": 0.2854, "step": 54360 }, { "epoch": 3.69, "learning_rate": 8.617699181780477e-07, "loss": 0.2076, "step": 54365 }, { "epoch": 3.69, "learning_rate": 8.608272689566759e-07, "loss": 0.3459, "step": 54370 }, { "epoch": 3.69, "learning_rate": 8.598846197353041e-07, "loss": 0.0695, "step": 54375 }, { "epoch": 3.69, "learning_rate": 8.589419705139325e-07, "loss": 0.0837, "step": 54380 }, { "epoch": 3.69, "learning_rate": 8.579993212925607e-07, "loss": 0.128, "step": 54385 }, { "epoch": 3.69, "learning_rate": 8.57056672071189e-07, "loss": 0.3121, "step": 54390 }, { "epoch": 3.69, "learning_rate": 8.561140228498171e-07, "loss": 0.0878, "step": 54395 }, { "epoch": 3.69, "learning_rate": 8.551713736284455e-07, "loss": 0.4382, "step": 54400 }, { "epoch": 3.69, "learning_rate": 8.542287244070738e-07, "loss": 0.2207, "step": 54405 }, { "epoch": 3.69, "learning_rate": 8.53286075185702e-07, "loss": 0.5784, "step": 54410 }, { "epoch": 3.69, "learning_rate": 8.523434259643302e-07, "loss": 0.2881, "step": 54415 }, { "epoch": 3.69, "learning_rate": 8.514007767429584e-07, "loss": 0.3356, "step": 54420 }, { "epoch": 3.69, "learning_rate": 8.504581275215868e-07, "loss": 0.227, "step": 54425 }, { "epoch": 3.69, "learning_rate": 8.495154783002151e-07, "loss": 0.2268, "step": 54430 }, { "epoch": 3.69, "learning_rate": 8.485728290788432e-07, "loss": 0.0799, "step": 54435 }, { "epoch": 3.69, "learning_rate": 8.476301798574715e-07, "loss": 0.271, "step": 54440 }, { "epoch": 3.7, "learning_rate": 8.466875306360997e-07, "loss": 0.1954, "step": 54445 }, { "epoch": 3.7, "learning_rate": 8.457448814147281e-07, "loss": 0.2248, "step": 54450 }, { "epoch": 3.7, "learning_rate": 8.448022321933563e-07, "loss": 0.2568, "step": 54455 }, { "epoch": 3.7, "learning_rate": 8.438595829719845e-07, "loss": 0.2508, "step": 54460 }, { "epoch": 3.7, "learning_rate": 8.429169337506128e-07, "loss": 0.1194, "step": 54465 }, { "epoch": 3.7, "learning_rate": 8.419742845292409e-07, "loss": 0.0998, "step": 54470 }, { "epoch": 3.7, "learning_rate": 8.410316353078693e-07, "loss": 0.2378, "step": 54475 }, { "epoch": 3.7, "learning_rate": 8.400889860864976e-07, "loss": 0.4702, "step": 54480 }, { "epoch": 3.7, "learning_rate": 8.391463368651258e-07, "loss": 0.1116, "step": 54485 }, { "epoch": 3.7, "learning_rate": 8.38203687643754e-07, "loss": 0.1643, "step": 54490 }, { "epoch": 3.7, "learning_rate": 8.372610384223822e-07, "loss": 0.1219, "step": 54495 }, { "epoch": 3.7, "learning_rate": 8.363183892010106e-07, "loss": 0.1954, "step": 54500 }, { "epoch": 3.7, "learning_rate": 8.353757399796389e-07, "loss": 0.3255, "step": 54505 }, { "epoch": 3.7, "learning_rate": 8.34433090758267e-07, "loss": 0.0621, "step": 54510 }, { "epoch": 3.7, "learning_rate": 8.334904415368953e-07, "loss": 0.3596, "step": 54515 }, { "epoch": 3.7, "learning_rate": 8.325477923155237e-07, "loss": 0.059, "step": 54520 }, { "epoch": 3.7, "learning_rate": 8.316051430941519e-07, "loss": 0.4428, "step": 54525 }, { "epoch": 3.7, "learning_rate": 8.306624938727801e-07, "loss": 0.0464, "step": 54530 }, { "epoch": 3.7, "learning_rate": 8.297198446514083e-07, "loss": 0.4116, "step": 54535 }, { "epoch": 3.7, "learning_rate": 8.287771954300367e-07, "loss": 0.1198, "step": 54540 }, { "epoch": 3.7, "learning_rate": 8.27834546208665e-07, "loss": 0.2402, "step": 54545 }, { "epoch": 3.7, "learning_rate": 8.268918969872931e-07, "loss": 0.5018, "step": 54550 }, { "epoch": 3.7, "learning_rate": 8.259492477659214e-07, "loss": 0.4849, "step": 54555 }, { "epoch": 3.7, "learning_rate": 8.250065985445496e-07, "loss": 0.5016, "step": 54560 }, { "epoch": 3.7, "learning_rate": 8.24063949323178e-07, "loss": 0.3024, "step": 54565 }, { "epoch": 3.7, "learning_rate": 8.231213001018062e-07, "loss": 0.1316, "step": 54570 }, { "epoch": 3.7, "learning_rate": 8.221786508804344e-07, "loss": 0.0715, "step": 54575 }, { "epoch": 3.7, "learning_rate": 8.212360016590627e-07, "loss": 0.7636, "step": 54580 }, { "epoch": 3.7, "learning_rate": 8.202933524376909e-07, "loss": 0.053, "step": 54585 }, { "epoch": 3.71, "learning_rate": 8.193507032163192e-07, "loss": 0.1064, "step": 54590 }, { "epoch": 3.71, "learning_rate": 8.184080539949475e-07, "loss": 0.3477, "step": 54595 }, { "epoch": 3.71, "learning_rate": 8.174654047735757e-07, "loss": 0.1852, "step": 54600 }, { "epoch": 3.71, "learning_rate": 8.16522755552204e-07, "loss": 0.5258, "step": 54605 }, { "epoch": 3.71, "learning_rate": 8.155801063308321e-07, "loss": 0.1107, "step": 54610 }, { "epoch": 3.71, "learning_rate": 8.146374571094605e-07, "loss": 0.17, "step": 54615 }, { "epoch": 3.71, "learning_rate": 8.136948078880888e-07, "loss": 0.2923, "step": 54620 }, { "epoch": 3.71, "learning_rate": 8.12752158666717e-07, "loss": 0.2985, "step": 54625 }, { "epoch": 3.71, "learning_rate": 8.118095094453452e-07, "loss": 0.2948, "step": 54630 }, { "epoch": 3.71, "learning_rate": 8.108668602239734e-07, "loss": 0.0591, "step": 54635 }, { "epoch": 3.71, "learning_rate": 8.099242110026018e-07, "loss": 0.0617, "step": 54640 }, { "epoch": 3.71, "learning_rate": 8.089815617812301e-07, "loss": 0.1038, "step": 54645 }, { "epoch": 3.71, "learning_rate": 8.080389125598582e-07, "loss": 0.1612, "step": 54650 }, { "epoch": 3.71, "learning_rate": 8.070962633384866e-07, "loss": 0.1031, "step": 54655 }, { "epoch": 3.71, "learning_rate": 8.061536141171149e-07, "loss": 0.6058, "step": 54660 }, { "epoch": 3.71, "learning_rate": 8.052109648957431e-07, "loss": 0.3462, "step": 54665 }, { "epoch": 3.71, "learning_rate": 8.042683156743713e-07, "loss": 0.1398, "step": 54670 }, { "epoch": 3.71, "learning_rate": 8.033256664529995e-07, "loss": 0.4762, "step": 54675 }, { "epoch": 3.71, "learning_rate": 8.023830172316279e-07, "loss": 0.1241, "step": 54680 }, { "epoch": 3.71, "learning_rate": 8.014403680102562e-07, "loss": 0.3454, "step": 54685 }, { "epoch": 3.71, "learning_rate": 8.004977187888843e-07, "loss": 0.2916, "step": 54690 }, { "epoch": 3.71, "learning_rate": 7.995550695675126e-07, "loss": 0.3976, "step": 54695 }, { "epoch": 3.71, "learning_rate": 7.986124203461408e-07, "loss": 0.1805, "step": 54700 }, { "epoch": 3.71, "learning_rate": 7.976697711247692e-07, "loss": 0.3738, "step": 54705 }, { "epoch": 3.71, "learning_rate": 7.967271219033974e-07, "loss": 0.2633, "step": 54710 }, { "epoch": 3.71, "learning_rate": 7.957844726820256e-07, "loss": 0.5575, "step": 54715 }, { "epoch": 3.71, "learning_rate": 7.948418234606539e-07, "loss": 0.2568, "step": 54720 }, { "epoch": 3.71, "learning_rate": 7.93899174239282e-07, "loss": 0.0178, "step": 54725 }, { "epoch": 3.71, "learning_rate": 7.929565250179104e-07, "loss": 0.265, "step": 54730 }, { "epoch": 3.71, "learning_rate": 7.920138757965387e-07, "loss": 0.2551, "step": 54735 }, { "epoch": 3.72, "learning_rate": 7.910712265751669e-07, "loss": 0.3257, "step": 54740 }, { "epoch": 3.72, "learning_rate": 7.901285773537951e-07, "loss": 0.2587, "step": 54745 }, { "epoch": 3.72, "learning_rate": 7.891859281324233e-07, "loss": 0.1773, "step": 54750 }, { "epoch": 3.72, "learning_rate": 7.882432789110517e-07, "loss": 0.2384, "step": 54755 }, { "epoch": 3.72, "learning_rate": 7.8730062968968e-07, "loss": 0.1868, "step": 54760 }, { "epoch": 3.72, "learning_rate": 7.863579804683081e-07, "loss": 0.3551, "step": 54765 }, { "epoch": 3.72, "learning_rate": 7.854153312469365e-07, "loss": 0.0606, "step": 54770 }, { "epoch": 3.72, "learning_rate": 7.844726820255647e-07, "loss": 0.1449, "step": 54775 }, { "epoch": 3.72, "learning_rate": 7.83530032804193e-07, "loss": 0.2634, "step": 54780 }, { "epoch": 3.72, "learning_rate": 7.825873835828212e-07, "loss": 0.1235, "step": 54785 }, { "epoch": 3.72, "learning_rate": 7.816447343614494e-07, "loss": 0.2247, "step": 54790 }, { "epoch": 3.72, "learning_rate": 7.807020851400778e-07, "loss": 0.2498, "step": 54795 }, { "epoch": 3.72, "learning_rate": 7.79759435918706e-07, "loss": 0.3033, "step": 54800 }, { "epoch": 3.72, "learning_rate": 7.788167866973342e-07, "loss": 0.2479, "step": 54805 }, { "epoch": 3.72, "learning_rate": 7.778741374759625e-07, "loss": 0.1625, "step": 54810 }, { "epoch": 3.72, "learning_rate": 7.769314882545907e-07, "loss": 0.1089, "step": 54815 }, { "epoch": 3.72, "learning_rate": 7.759888390332191e-07, "loss": 0.4358, "step": 54820 }, { "epoch": 3.72, "learning_rate": 7.750461898118473e-07, "loss": 0.3789, "step": 54825 }, { "epoch": 3.72, "learning_rate": 7.741035405904755e-07, "loss": 0.2105, "step": 54830 }, { "epoch": 3.72, "learning_rate": 7.731608913691038e-07, "loss": 0.4255, "step": 54835 }, { "epoch": 3.72, "learning_rate": 7.72218242147732e-07, "loss": 0.5505, "step": 54840 }, { "epoch": 3.72, "learning_rate": 7.712755929263603e-07, "loss": 0.4591, "step": 54845 }, { "epoch": 3.72, "learning_rate": 7.703329437049886e-07, "loss": 0.0738, "step": 54850 }, { "epoch": 3.72, "learning_rate": 7.693902944836168e-07, "loss": 0.6137, "step": 54855 }, { "epoch": 3.72, "learning_rate": 7.684476452622451e-07, "loss": 0.2944, "step": 54860 }, { "epoch": 3.72, "learning_rate": 7.675049960408732e-07, "loss": 0.3314, "step": 54865 }, { "epoch": 3.72, "learning_rate": 7.665623468195016e-07, "loss": 0.4993, "step": 54870 }, { "epoch": 3.72, "learning_rate": 7.656196975981299e-07, "loss": 0.1669, "step": 54875 }, { "epoch": 3.72, "learning_rate": 7.646770483767581e-07, "loss": 0.3374, "step": 54880 }, { "epoch": 3.73, "learning_rate": 7.637343991553863e-07, "loss": 0.4832, "step": 54885 }, { "epoch": 3.73, "learning_rate": 7.627917499340146e-07, "loss": 0.1122, "step": 54890 }, { "epoch": 3.73, "learning_rate": 7.618491007126429e-07, "loss": 0.7442, "step": 54895 }, { "epoch": 3.73, "learning_rate": 7.609064514912712e-07, "loss": 0.2812, "step": 54900 }, { "epoch": 3.73, "learning_rate": 7.599638022698993e-07, "loss": 0.3042, "step": 54905 }, { "epoch": 3.73, "learning_rate": 7.590211530485277e-07, "loss": 0.3222, "step": 54910 }, { "epoch": 3.73, "learning_rate": 7.580785038271559e-07, "loss": 0.1007, "step": 54915 }, { "epoch": 3.73, "learning_rate": 7.571358546057842e-07, "loss": 0.0749, "step": 54920 }, { "epoch": 3.73, "learning_rate": 7.561932053844124e-07, "loss": 0.5411, "step": 54925 }, { "epoch": 3.73, "learning_rate": 7.552505561630406e-07, "loss": 0.4371, "step": 54930 }, { "epoch": 3.73, "learning_rate": 7.54307906941669e-07, "loss": 0.1016, "step": 54935 }, { "epoch": 3.73, "learning_rate": 7.533652577202972e-07, "loss": 0.3606, "step": 54940 }, { "epoch": 3.73, "learning_rate": 7.524226084989254e-07, "loss": 0.1352, "step": 54945 }, { "epoch": 3.73, "learning_rate": 7.514799592775537e-07, "loss": 0.2859, "step": 54950 }, { "epoch": 3.73, "learning_rate": 7.505373100561819e-07, "loss": 0.2394, "step": 54955 }, { "epoch": 3.73, "learning_rate": 7.495946608348103e-07, "loss": 0.5369, "step": 54960 }, { "epoch": 3.73, "learning_rate": 7.486520116134385e-07, "loss": 0.4018, "step": 54965 }, { "epoch": 3.73, "learning_rate": 7.477093623920667e-07, "loss": 0.1662, "step": 54970 }, { "epoch": 3.73, "learning_rate": 7.46766713170695e-07, "loss": 0.1351, "step": 54975 }, { "epoch": 3.73, "learning_rate": 7.458240639493231e-07, "loss": 0.1182, "step": 54980 }, { "epoch": 3.73, "learning_rate": 7.448814147279515e-07, "loss": 0.3755, "step": 54985 }, { "epoch": 3.73, "learning_rate": 7.439387655065798e-07, "loss": 0.2916, "step": 54990 }, { "epoch": 3.73, "learning_rate": 7.42996116285208e-07, "loss": 0.46, "step": 54995 }, { "epoch": 3.73, "learning_rate": 7.420534670638362e-07, "loss": 0.2331, "step": 55000 }, { "epoch": 3.73, "learning_rate": 7.411108178424644e-07, "loss": 0.0847, "step": 55005 }, { "epoch": 3.73, "learning_rate": 7.401681686210928e-07, "loss": 0.4371, "step": 55010 }, { "epoch": 3.73, "learning_rate": 7.392255193997211e-07, "loss": 0.2557, "step": 55015 }, { "epoch": 3.73, "learning_rate": 7.382828701783492e-07, "loss": 0.2251, "step": 55020 }, { "epoch": 3.73, "learning_rate": 7.373402209569776e-07, "loss": 0.501, "step": 55025 }, { "epoch": 3.73, "learning_rate": 7.363975717356058e-07, "loss": 0.3087, "step": 55030 }, { "epoch": 3.74, "learning_rate": 7.354549225142341e-07, "loss": 0.0707, "step": 55035 }, { "epoch": 3.74, "learning_rate": 7.345122732928623e-07, "loss": 0.2754, "step": 55040 }, { "epoch": 3.74, "learning_rate": 7.335696240714905e-07, "loss": 0.3017, "step": 55045 }, { "epoch": 3.74, "learning_rate": 7.326269748501189e-07, "loss": 0.0412, "step": 55050 }, { "epoch": 3.74, "learning_rate": 7.316843256287471e-07, "loss": 0.1179, "step": 55055 }, { "epoch": 3.74, "learning_rate": 7.307416764073753e-07, "loss": 0.3144, "step": 55060 }, { "epoch": 3.74, "learning_rate": 7.297990271860036e-07, "loss": 0.2648, "step": 55065 }, { "epoch": 3.74, "learning_rate": 7.288563779646318e-07, "loss": 0.204, "step": 55070 }, { "epoch": 3.74, "learning_rate": 7.279137287432602e-07, "loss": 0.3304, "step": 55075 }, { "epoch": 3.74, "learning_rate": 7.269710795218883e-07, "loss": 0.6588, "step": 55080 }, { "epoch": 3.74, "learning_rate": 7.260284303005166e-07, "loss": 0.4677, "step": 55085 }, { "epoch": 3.74, "learning_rate": 7.250857810791449e-07, "loss": 0.0564, "step": 55090 }, { "epoch": 3.74, "learning_rate": 7.241431318577731e-07, "loss": 0.4691, "step": 55095 }, { "epoch": 3.74, "learning_rate": 7.232004826364014e-07, "loss": 0.3787, "step": 55100 }, { "epoch": 3.74, "learning_rate": 7.222578334150297e-07, "loss": 0.1963, "step": 55105 }, { "epoch": 3.74, "learning_rate": 7.213151841936579e-07, "loss": 0.0799, "step": 55110 }, { "epoch": 3.74, "learning_rate": 7.203725349722862e-07, "loss": 0.3657, "step": 55115 }, { "epoch": 3.74, "learning_rate": 7.194298857509143e-07, "loss": 0.415, "step": 55120 }, { "epoch": 3.74, "learning_rate": 7.184872365295427e-07, "loss": 0.3726, "step": 55125 }, { "epoch": 3.74, "learning_rate": 7.17544587308171e-07, "loss": 0.1766, "step": 55130 }, { "epoch": 3.74, "learning_rate": 7.166019380867992e-07, "loss": 0.4801, "step": 55135 }, { "epoch": 3.74, "learning_rate": 7.156592888654274e-07, "loss": 0.3843, "step": 55140 }, { "epoch": 3.74, "learning_rate": 7.147166396440557e-07, "loss": 0.2677, "step": 55145 }, { "epoch": 3.74, "learning_rate": 7.13773990422684e-07, "loss": 0.8082, "step": 55150 }, { "epoch": 3.74, "learning_rate": 7.128313412013123e-07, "loss": 0.4725, "step": 55155 }, { "epoch": 3.74, "learning_rate": 7.118886919799404e-07, "loss": 0.0655, "step": 55160 }, { "epoch": 3.74, "learning_rate": 7.109460427585688e-07, "loss": 0.1558, "step": 55165 }, { "epoch": 3.74, "learning_rate": 7.10003393537197e-07, "loss": 0.148, "step": 55170 }, { "epoch": 3.74, "learning_rate": 7.090607443158253e-07, "loss": 0.2232, "step": 55175 }, { "epoch": 3.75, "learning_rate": 7.081180950944535e-07, "loss": 0.137, "step": 55180 }, { "epoch": 3.75, "learning_rate": 7.071754458730817e-07, "loss": 0.2744, "step": 55185 }, { "epoch": 3.75, "learning_rate": 7.062327966517101e-07, "loss": 0.2994, "step": 55190 }, { "epoch": 3.75, "learning_rate": 7.052901474303383e-07, "loss": 0.4329, "step": 55195 }, { "epoch": 3.75, "learning_rate": 7.043474982089665e-07, "loss": 0.1419, "step": 55200 }, { "epoch": 3.75, "learning_rate": 7.034048489875948e-07, "loss": 0.1535, "step": 55205 }, { "epoch": 3.75, "learning_rate": 7.02462199766223e-07, "loss": 0.7291, "step": 55210 }, { "epoch": 3.75, "learning_rate": 7.015195505448514e-07, "loss": 0.1246, "step": 55215 }, { "epoch": 3.75, "learning_rate": 7.005769013234795e-07, "loss": 0.8049, "step": 55220 }, { "epoch": 3.75, "learning_rate": 6.996342521021078e-07, "loss": 0.1427, "step": 55225 }, { "epoch": 3.75, "learning_rate": 6.986916028807361e-07, "loss": 0.1606, "step": 55230 }, { "epoch": 3.75, "learning_rate": 6.977489536593642e-07, "loss": 0.2014, "step": 55235 }, { "epoch": 3.75, "learning_rate": 6.968063044379926e-07, "loss": 0.2289, "step": 55240 }, { "epoch": 3.75, "learning_rate": 6.958636552166209e-07, "loss": 0.5055, "step": 55245 }, { "epoch": 3.75, "learning_rate": 6.949210059952491e-07, "loss": 0.2769, "step": 55250 }, { "epoch": 3.75, "learning_rate": 6.939783567738773e-07, "loss": 0.1705, "step": 55255 }, { "epoch": 3.75, "learning_rate": 6.930357075525055e-07, "loss": 0.1376, "step": 55260 }, { "epoch": 3.75, "learning_rate": 6.920930583311339e-07, "loss": 0.1891, "step": 55265 }, { "epoch": 3.75, "learning_rate": 6.911504091097622e-07, "loss": 0.5385, "step": 55270 }, { "epoch": 3.75, "learning_rate": 6.902077598883903e-07, "loss": 0.3246, "step": 55275 }, { "epoch": 3.75, "learning_rate": 6.892651106670187e-07, "loss": 0.2207, "step": 55280 }, { "epoch": 3.75, "learning_rate": 6.883224614456469e-07, "loss": 0.2645, "step": 55285 }, { "epoch": 3.75, "learning_rate": 6.873798122242752e-07, "loss": 0.5491, "step": 55290 }, { "epoch": 3.75, "learning_rate": 6.864371630029034e-07, "loss": 0.256, "step": 55295 }, { "epoch": 3.75, "learning_rate": 6.854945137815316e-07, "loss": 0.2404, "step": 55300 }, { "epoch": 3.75, "learning_rate": 6.8455186456016e-07, "loss": 0.1556, "step": 55305 }, { "epoch": 3.75, "learning_rate": 6.836092153387882e-07, "loss": 0.2965, "step": 55310 }, { "epoch": 3.75, "learning_rate": 6.826665661174164e-07, "loss": 0.6541, "step": 55315 }, { "epoch": 3.75, "learning_rate": 6.817239168960447e-07, "loss": 0.5431, "step": 55320 }, { "epoch": 3.75, "learning_rate": 6.807812676746729e-07, "loss": 0.2203, "step": 55325 }, { "epoch": 3.76, "learning_rate": 6.798386184533013e-07, "loss": 0.156, "step": 55330 }, { "epoch": 3.76, "learning_rate": 6.788959692319294e-07, "loss": 0.5917, "step": 55335 }, { "epoch": 3.76, "learning_rate": 6.779533200105577e-07, "loss": 0.154, "step": 55340 }, { "epoch": 3.76, "learning_rate": 6.77010670789186e-07, "loss": 0.4562, "step": 55345 }, { "epoch": 3.76, "learning_rate": 6.760680215678142e-07, "loss": 0.3981, "step": 55350 }, { "epoch": 3.76, "learning_rate": 6.751253723464425e-07, "loss": 0.135, "step": 55355 }, { "epoch": 3.76, "learning_rate": 6.741827231250707e-07, "loss": 0.202, "step": 55360 }, { "epoch": 3.76, "learning_rate": 6.73240073903699e-07, "loss": 0.061, "step": 55365 }, { "epoch": 3.76, "learning_rate": 6.722974246823273e-07, "loss": 0.1369, "step": 55370 }, { "epoch": 3.76, "learning_rate": 6.713547754609554e-07, "loss": 0.2904, "step": 55375 }, { "epoch": 3.76, "learning_rate": 6.704121262395838e-07, "loss": 1.2528, "step": 55380 }, { "epoch": 3.76, "learning_rate": 6.694694770182121e-07, "loss": 0.3315, "step": 55385 }, { "epoch": 3.76, "learning_rate": 6.685268277968403e-07, "loss": 0.3841, "step": 55390 }, { "epoch": 3.76, "learning_rate": 6.675841785754686e-07, "loss": 0.1093, "step": 55395 }, { "epoch": 3.76, "learning_rate": 6.666415293540968e-07, "loss": 0.1463, "step": 55400 }, { "epoch": 3.76, "learning_rate": 6.656988801327251e-07, "loss": 0.4437, "step": 55405 }, { "epoch": 3.76, "learning_rate": 6.647562309113534e-07, "loss": 0.1458, "step": 55410 }, { "epoch": 3.76, "learning_rate": 6.638135816899815e-07, "loss": 0.5261, "step": 55415 }, { "epoch": 3.76, "learning_rate": 6.628709324686099e-07, "loss": 0.4517, "step": 55420 }, { "epoch": 3.76, "learning_rate": 6.619282832472381e-07, "loss": 0.539, "step": 55425 }, { "epoch": 3.76, "learning_rate": 6.609856340258664e-07, "loss": 0.1779, "step": 55430 }, { "epoch": 3.76, "learning_rate": 6.600429848044946e-07, "loss": 0.2944, "step": 55435 }, { "epoch": 3.76, "learning_rate": 6.591003355831228e-07, "loss": 0.3209, "step": 55440 }, { "epoch": 3.76, "learning_rate": 6.581576863617512e-07, "loss": 0.22, "step": 55445 }, { "epoch": 3.76, "learning_rate": 6.572150371403794e-07, "loss": 0.4846, "step": 55450 }, { "epoch": 3.76, "learning_rate": 6.562723879190076e-07, "loss": 0.1807, "step": 55455 }, { "epoch": 3.76, "learning_rate": 6.553297386976359e-07, "loss": 0.086, "step": 55460 }, { "epoch": 3.76, "learning_rate": 6.543870894762641e-07, "loss": 0.2841, "step": 55465 }, { "epoch": 3.76, "learning_rate": 6.534444402548925e-07, "loss": 0.1586, "step": 55470 }, { "epoch": 3.77, "learning_rate": 6.525017910335206e-07, "loss": 0.1246, "step": 55475 }, { "epoch": 3.77, "learning_rate": 6.515591418121489e-07, "loss": 0.6498, "step": 55480 }, { "epoch": 3.77, "learning_rate": 6.506164925907772e-07, "loss": 0.2548, "step": 55485 }, { "epoch": 3.77, "learning_rate": 6.496738433694053e-07, "loss": 0.1269, "step": 55490 }, { "epoch": 3.77, "learning_rate": 6.487311941480337e-07, "loss": 0.3067, "step": 55495 }, { "epoch": 3.77, "learning_rate": 6.477885449266619e-07, "loss": 0.1924, "step": 55500 }, { "epoch": 3.77, "learning_rate": 6.468458957052902e-07, "loss": 0.2622, "step": 55505 }, { "epoch": 3.77, "learning_rate": 6.459032464839184e-07, "loss": 0.1433, "step": 55510 }, { "epoch": 3.77, "learning_rate": 6.449605972625467e-07, "loss": 0.1691, "step": 55515 }, { "epoch": 3.77, "learning_rate": 6.44017948041175e-07, "loss": 0.4037, "step": 55520 }, { "epoch": 3.77, "learning_rate": 6.430752988198033e-07, "loss": 0.1343, "step": 55525 }, { "epoch": 3.77, "learning_rate": 6.421326495984314e-07, "loss": 0.8454, "step": 55530 }, { "epoch": 3.77, "learning_rate": 6.411900003770598e-07, "loss": 0.2964, "step": 55535 }, { "epoch": 3.77, "learning_rate": 6.40247351155688e-07, "loss": 0.2948, "step": 55540 }, { "epoch": 3.77, "learning_rate": 6.393047019343163e-07, "loss": 0.4977, "step": 55545 }, { "epoch": 3.77, "learning_rate": 6.383620527129445e-07, "loss": 0.452, "step": 55550 }, { "epoch": 3.77, "learning_rate": 6.374194034915727e-07, "loss": 0.3511, "step": 55555 }, { "epoch": 3.77, "learning_rate": 6.364767542702011e-07, "loss": 0.2307, "step": 55560 }, { "epoch": 3.77, "learning_rate": 6.355341050488293e-07, "loss": 0.3512, "step": 55565 }, { "epoch": 3.77, "learning_rate": 6.345914558274575e-07, "loss": 0.0636, "step": 55570 }, { "epoch": 3.77, "learning_rate": 6.336488066060858e-07, "loss": 0.3614, "step": 55575 }, { "epoch": 3.77, "learning_rate": 6.32706157384714e-07, "loss": 0.0775, "step": 55580 }, { "epoch": 3.77, "learning_rate": 6.317635081633424e-07, "loss": 0.0641, "step": 55585 }, { "epoch": 3.77, "learning_rate": 6.308208589419705e-07, "loss": 0.093, "step": 55590 }, { "epoch": 3.77, "learning_rate": 6.298782097205988e-07, "loss": 0.4084, "step": 55595 }, { "epoch": 3.77, "learning_rate": 6.289355604992271e-07, "loss": 0.4119, "step": 55600 }, { "epoch": 3.77, "learning_rate": 6.279929112778553e-07, "loss": 0.4647, "step": 55605 }, { "epoch": 3.77, "learning_rate": 6.270502620564836e-07, "loss": 0.1045, "step": 55610 }, { "epoch": 3.77, "learning_rate": 6.261076128351118e-07, "loss": 0.8842, "step": 55615 }, { "epoch": 3.77, "learning_rate": 6.251649636137401e-07, "loss": 0.1139, "step": 55620 }, { "epoch": 3.78, "learning_rate": 6.242223143923684e-07, "loss": 0.2819, "step": 55625 }, { "epoch": 3.78, "learning_rate": 6.232796651709966e-07, "loss": 0.1041, "step": 55630 }, { "epoch": 3.78, "learning_rate": 6.223370159496249e-07, "loss": 0.1343, "step": 55635 }, { "epoch": 3.78, "learning_rate": 6.213943667282532e-07, "loss": 0.4704, "step": 55640 }, { "epoch": 3.78, "learning_rate": 6.204517175068814e-07, "loss": 0.2259, "step": 55645 }, { "epoch": 3.78, "learning_rate": 6.195090682855096e-07, "loss": 0.0914, "step": 55650 }, { "epoch": 3.78, "learning_rate": 6.185664190641379e-07, "loss": 0.2398, "step": 55655 }, { "epoch": 3.78, "learning_rate": 6.176237698427662e-07, "loss": 0.1276, "step": 55660 }, { "epoch": 3.78, "learning_rate": 6.166811206213945e-07, "loss": 0.4749, "step": 55665 }, { "epoch": 3.78, "learning_rate": 6.157384714000226e-07, "loss": 0.1471, "step": 55670 }, { "epoch": 3.78, "learning_rate": 6.147958221786509e-07, "loss": 0.3194, "step": 55675 }, { "epoch": 3.78, "learning_rate": 6.138531729572792e-07, "loss": 0.5263, "step": 55680 }, { "epoch": 3.78, "learning_rate": 6.129105237359075e-07, "loss": 0.0279, "step": 55685 }, { "epoch": 3.78, "learning_rate": 6.119678745145357e-07, "loss": 0.2479, "step": 55690 }, { "epoch": 3.78, "learning_rate": 6.110252252931639e-07, "loss": 0.3177, "step": 55695 }, { "epoch": 3.78, "learning_rate": 6.100825760717923e-07, "loss": 0.2022, "step": 55700 }, { "epoch": 3.78, "learning_rate": 6.091399268504205e-07, "loss": 0.0778, "step": 55705 }, { "epoch": 3.78, "learning_rate": 6.081972776290487e-07, "loss": 0.2005, "step": 55710 }, { "epoch": 3.78, "learning_rate": 6.07254628407677e-07, "loss": 0.2864, "step": 55715 }, { "epoch": 3.78, "learning_rate": 6.063119791863052e-07, "loss": 0.8645, "step": 55720 }, { "epoch": 3.78, "learning_rate": 6.053693299649336e-07, "loss": 0.1852, "step": 55725 }, { "epoch": 3.78, "learning_rate": 6.044266807435617e-07, "loss": 0.1009, "step": 55730 }, { "epoch": 3.78, "learning_rate": 6.0348403152219e-07, "loss": 0.2637, "step": 55735 }, { "epoch": 3.78, "learning_rate": 6.025413823008183e-07, "loss": 0.2563, "step": 55740 }, { "epoch": 3.78, "learning_rate": 6.015987330794464e-07, "loss": 0.2129, "step": 55745 }, { "epoch": 3.78, "learning_rate": 6.006560838580748e-07, "loss": 0.064, "step": 55750 }, { "epoch": 3.78, "learning_rate": 5.99713434636703e-07, "loss": 0.12, "step": 55755 }, { "epoch": 3.78, "learning_rate": 5.987707854153313e-07, "loss": 0.3207, "step": 55760 }, { "epoch": 3.78, "learning_rate": 5.978281361939595e-07, "loss": 0.3253, "step": 55765 }, { "epoch": 3.79, "learning_rate": 5.968854869725878e-07, "loss": 0.5995, "step": 55770 }, { "epoch": 3.79, "learning_rate": 5.959428377512161e-07, "loss": 0.2975, "step": 55775 }, { "epoch": 3.79, "learning_rate": 5.950001885298444e-07, "loss": 0.2754, "step": 55780 }, { "epoch": 3.79, "learning_rate": 5.940575393084725e-07, "loss": 0.1716, "step": 55785 }, { "epoch": 3.79, "learning_rate": 5.931148900871008e-07, "loss": 0.374, "step": 55790 }, { "epoch": 3.79, "learning_rate": 5.921722408657291e-07, "loss": 0.2508, "step": 55795 }, { "epoch": 3.79, "learning_rate": 5.912295916443574e-07, "loss": 0.2055, "step": 55800 }, { "epoch": 3.79, "learning_rate": 5.902869424229856e-07, "loss": 0.1203, "step": 55805 }, { "epoch": 3.79, "learning_rate": 5.893442932016138e-07, "loss": 0.547, "step": 55810 }, { "epoch": 3.79, "learning_rate": 5.884016439802421e-07, "loss": 0.2965, "step": 55815 }, { "epoch": 3.79, "learning_rate": 5.874589947588704e-07, "loss": 0.4584, "step": 55820 }, { "epoch": 3.79, "learning_rate": 5.865163455374986e-07, "loss": 0.0772, "step": 55825 }, { "epoch": 3.79, "learning_rate": 5.855736963161269e-07, "loss": 0.4186, "step": 55830 }, { "epoch": 3.79, "learning_rate": 5.846310470947551e-07, "loss": 0.1853, "step": 55835 }, { "epoch": 3.79, "learning_rate": 5.836883978733835e-07, "loss": 0.39, "step": 55840 }, { "epoch": 3.79, "learning_rate": 5.827457486520116e-07, "loss": 0.1222, "step": 55845 }, { "epoch": 3.79, "learning_rate": 5.818030994306399e-07, "loss": 0.4029, "step": 55850 }, { "epoch": 3.79, "learning_rate": 5.808604502092682e-07, "loss": 0.39, "step": 55855 }, { "epoch": 3.79, "learning_rate": 5.799178009878964e-07, "loss": 0.3795, "step": 55860 }, { "epoch": 3.79, "learning_rate": 5.789751517665247e-07, "loss": 0.3589, "step": 55865 }, { "epoch": 3.79, "learning_rate": 5.780325025451529e-07, "loss": 0.0638, "step": 55870 }, { "epoch": 3.79, "learning_rate": 5.770898533237812e-07, "loss": 0.3621, "step": 55875 }, { "epoch": 3.79, "learning_rate": 5.761472041024095e-07, "loss": 0.0598, "step": 55880 }, { "epoch": 3.79, "learning_rate": 5.752045548810377e-07, "loss": 0.0923, "step": 55885 }, { "epoch": 3.79, "learning_rate": 5.74261905659666e-07, "loss": 0.0599, "step": 55890 }, { "epoch": 3.79, "learning_rate": 5.733192564382943e-07, "loss": 0.3322, "step": 55895 }, { "epoch": 3.79, "learning_rate": 5.723766072169225e-07, "loss": 0.1223, "step": 55900 }, { "epoch": 3.79, "learning_rate": 5.714339579955507e-07, "loss": 0.3916, "step": 55905 }, { "epoch": 3.79, "learning_rate": 5.70491308774179e-07, "loss": 0.0883, "step": 55910 }, { "epoch": 3.79, "learning_rate": 5.695486595528073e-07, "loss": 0.1597, "step": 55915 }, { "epoch": 3.8, "learning_rate": 5.686060103314356e-07, "loss": 0.212, "step": 55920 }, { "epoch": 3.8, "learning_rate": 5.676633611100637e-07, "loss": 0.2172, "step": 55925 }, { "epoch": 3.8, "learning_rate": 5.66720711888692e-07, "loss": 0.1958, "step": 55930 }, { "epoch": 3.8, "learning_rate": 5.657780626673203e-07, "loss": 0.2442, "step": 55935 }, { "epoch": 3.8, "learning_rate": 5.648354134459486e-07, "loss": 0.2401, "step": 55940 }, { "epoch": 3.8, "learning_rate": 5.638927642245768e-07, "loss": 0.3357, "step": 55945 }, { "epoch": 3.8, "learning_rate": 5.62950115003205e-07, "loss": 0.5052, "step": 55950 }, { "epoch": 3.8, "learning_rate": 5.620074657818333e-07, "loss": 0.3765, "step": 55955 }, { "epoch": 3.8, "learning_rate": 5.610648165604616e-07, "loss": 0.5569, "step": 55960 }, { "epoch": 3.8, "learning_rate": 5.601221673390898e-07, "loss": 0.3216, "step": 55965 }, { "epoch": 3.8, "learning_rate": 5.591795181177181e-07, "loss": 0.1111, "step": 55970 }, { "epoch": 3.8, "learning_rate": 5.582368688963463e-07, "loss": 0.4556, "step": 55975 }, { "epoch": 3.8, "learning_rate": 5.572942196749746e-07, "loss": 0.4542, "step": 55980 }, { "epoch": 3.8, "learning_rate": 5.563515704536028e-07, "loss": 0.0898, "step": 55985 }, { "epoch": 3.8, "learning_rate": 5.554089212322311e-07, "loss": 0.2165, "step": 55990 }, { "epoch": 3.8, "learning_rate": 5.544662720108594e-07, "loss": 0.4943, "step": 55995 }, { "epoch": 3.8, "learning_rate": 5.535236227894875e-07, "loss": 0.7668, "step": 56000 }, { "epoch": 3.8, "learning_rate": 5.525809735681159e-07, "loss": 0.0444, "step": 56005 }, { "epoch": 3.8, "learning_rate": 5.516383243467441e-07, "loss": 0.107, "step": 56010 }, { "epoch": 3.8, "learning_rate": 5.506956751253724e-07, "loss": 0.1374, "step": 56015 }, { "epoch": 3.8, "learning_rate": 5.497530259040006e-07, "loss": 0.1337, "step": 56020 }, { "epoch": 3.8, "learning_rate": 5.488103766826289e-07, "loss": 0.1046, "step": 56025 }, { "epoch": 3.8, "learning_rate": 5.478677274612572e-07, "loss": 0.1012, "step": 56030 }, { "epoch": 3.8, "learning_rate": 5.469250782398855e-07, "loss": 0.4169, "step": 56035 }, { "epoch": 3.8, "learning_rate": 5.459824290185136e-07, "loss": 0.0716, "step": 56040 }, { "epoch": 3.8, "learning_rate": 5.450397797971419e-07, "loss": 0.092, "step": 56045 }, { "epoch": 3.8, "learning_rate": 5.440971305757702e-07, "loss": 0.177, "step": 56050 }, { "epoch": 3.8, "learning_rate": 5.431544813543985e-07, "loss": 0.0955, "step": 56055 }, { "epoch": 3.8, "learning_rate": 5.422118321330267e-07, "loss": 0.1605, "step": 56060 }, { "epoch": 3.81, "learning_rate": 5.412691829116549e-07, "loss": 0.253, "step": 56065 }, { "epoch": 3.81, "learning_rate": 5.403265336902832e-07, "loss": 0.3163, "step": 56070 }, { "epoch": 3.81, "learning_rate": 5.393838844689115e-07, "loss": 0.4016, "step": 56075 }, { "epoch": 3.81, "learning_rate": 5.384412352475397e-07, "loss": 0.2396, "step": 56080 }, { "epoch": 3.81, "learning_rate": 5.37498586026168e-07, "loss": 0.3045, "step": 56085 }, { "epoch": 3.81, "learning_rate": 5.365559368047962e-07, "loss": 0.6473, "step": 56090 }, { "epoch": 3.81, "learning_rate": 5.356132875834245e-07, "loss": 0.2965, "step": 56095 }, { "epoch": 3.81, "learning_rate": 5.346706383620527e-07, "loss": 0.1686, "step": 56100 }, { "epoch": 3.81, "learning_rate": 5.33727989140681e-07, "loss": 0.1963, "step": 56105 }, { "epoch": 3.81, "learning_rate": 5.327853399193093e-07, "loss": 0.0384, "step": 56110 }, { "epoch": 3.81, "learning_rate": 5.318426906979375e-07, "loss": 0.5349, "step": 56115 }, { "epoch": 3.81, "learning_rate": 5.309000414765657e-07, "loss": 0.2941, "step": 56120 }, { "epoch": 3.81, "learning_rate": 5.29957392255194e-07, "loss": 0.0876, "step": 56125 }, { "epoch": 3.81, "learning_rate": 5.290147430338223e-07, "loss": 0.5118, "step": 56130 }, { "epoch": 3.81, "learning_rate": 5.280720938124506e-07, "loss": 0.107, "step": 56135 }, { "epoch": 3.81, "learning_rate": 5.271294445910788e-07, "loss": 0.3384, "step": 56140 }, { "epoch": 3.81, "learning_rate": 5.261867953697071e-07, "loss": 0.0962, "step": 56145 }, { "epoch": 3.81, "learning_rate": 5.252441461483354e-07, "loss": 0.4767, "step": 56150 }, { "epoch": 3.81, "learning_rate": 5.243014969269636e-07, "loss": 0.2266, "step": 56155 }, { "epoch": 3.81, "learning_rate": 5.233588477055918e-07, "loss": 0.4676, "step": 56160 }, { "epoch": 3.81, "learning_rate": 5.224161984842201e-07, "loss": 0.3423, "step": 56165 }, { "epoch": 3.81, "learning_rate": 5.214735492628484e-07, "loss": 0.3927, "step": 56170 }, { "epoch": 3.81, "learning_rate": 5.205309000414767e-07, "loss": 0.2168, "step": 56175 }, { "epoch": 3.81, "learning_rate": 5.195882508201048e-07, "loss": 0.2156, "step": 56180 }, { "epoch": 3.81, "learning_rate": 5.186456015987331e-07, "loss": 0.3224, "step": 56185 }, { "epoch": 3.81, "learning_rate": 5.177029523773614e-07, "loss": 0.2071, "step": 56190 }, { "epoch": 3.81, "learning_rate": 5.167603031559897e-07, "loss": 0.2068, "step": 56195 }, { "epoch": 3.81, "learning_rate": 5.158176539346179e-07, "loss": 0.3425, "step": 56200 }, { "epoch": 3.81, "learning_rate": 5.148750047132461e-07, "loss": 0.5884, "step": 56205 }, { "epoch": 3.81, "learning_rate": 5.139323554918744e-07, "loss": 0.2897, "step": 56210 }, { "epoch": 3.82, "learning_rate": 5.129897062705027e-07, "loss": 0.07, "step": 56215 }, { "epoch": 3.82, "learning_rate": 5.120470570491309e-07, "loss": 0.2945, "step": 56220 }, { "epoch": 3.82, "learning_rate": 5.111044078277592e-07, "loss": 0.3065, "step": 56225 }, { "epoch": 3.82, "learning_rate": 5.101617586063874e-07, "loss": 0.1029, "step": 56230 }, { "epoch": 3.82, "learning_rate": 5.092191093850157e-07, "loss": 0.1707, "step": 56235 }, { "epoch": 3.82, "learning_rate": 5.082764601636439e-07, "loss": 0.2763, "step": 56240 }, { "epoch": 3.82, "learning_rate": 5.073338109422722e-07, "loss": 0.1406, "step": 56245 }, { "epoch": 3.82, "learning_rate": 5.063911617209005e-07, "loss": 0.3222, "step": 56250 }, { "epoch": 3.82, "learning_rate": 5.054485124995286e-07, "loss": 0.0659, "step": 56255 }, { "epoch": 3.82, "learning_rate": 5.045058632781569e-07, "loss": 0.2897, "step": 56260 }, { "epoch": 3.82, "learning_rate": 5.035632140567853e-07, "loss": 0.0642, "step": 56265 }, { "epoch": 3.82, "learning_rate": 5.026205648354135e-07, "loss": 0.2617, "step": 56270 }, { "epoch": 3.82, "learning_rate": 5.016779156140417e-07, "loss": 0.3902, "step": 56275 }, { "epoch": 3.82, "learning_rate": 5.0073526639267e-07, "loss": 0.2635, "step": 56280 }, { "epoch": 3.82, "learning_rate": 4.997926171712983e-07, "loss": 0.0815, "step": 56285 }, { "epoch": 3.82, "learning_rate": 4.988499679499266e-07, "loss": 0.1634, "step": 56290 }, { "epoch": 3.82, "learning_rate": 4.979073187285547e-07, "loss": 0.4812, "step": 56295 }, { "epoch": 3.82, "learning_rate": 4.96964669507183e-07, "loss": 0.0936, "step": 56300 }, { "epoch": 3.82, "learning_rate": 4.960220202858113e-07, "loss": 0.0115, "step": 56305 }, { "epoch": 3.82, "learning_rate": 4.950793710644396e-07, "loss": 0.2126, "step": 56310 }, { "epoch": 3.82, "learning_rate": 4.941367218430678e-07, "loss": 0.4923, "step": 56315 }, { "epoch": 3.82, "learning_rate": 4.93194072621696e-07, "loss": 0.429, "step": 56320 }, { "epoch": 3.82, "learning_rate": 4.922514234003243e-07, "loss": 0.0534, "step": 56325 }, { "epoch": 3.82, "learning_rate": 4.913087741789526e-07, "loss": 0.6736, "step": 56330 }, { "epoch": 3.82, "learning_rate": 4.903661249575808e-07, "loss": 0.4531, "step": 56335 }, { "epoch": 3.82, "learning_rate": 4.894234757362091e-07, "loss": 0.0913, "step": 56340 }, { "epoch": 3.82, "learning_rate": 4.884808265148373e-07, "loss": 0.1457, "step": 56345 }, { "epoch": 3.82, "learning_rate": 4.875381772934656e-07, "loss": 0.2038, "step": 56350 }, { "epoch": 3.82, "learning_rate": 4.865955280720938e-07, "loss": 0.4035, "step": 56355 }, { "epoch": 3.83, "learning_rate": 4.856528788507221e-07, "loss": 0.2561, "step": 56360 }, { "epoch": 3.83, "learning_rate": 4.847102296293504e-07, "loss": 0.2342, "step": 56365 }, { "epoch": 3.83, "learning_rate": 4.837675804079786e-07, "loss": 0.3806, "step": 56370 }, { "epoch": 3.83, "learning_rate": 4.828249311866068e-07, "loss": 0.4324, "step": 56375 }, { "epoch": 3.83, "learning_rate": 4.818822819652351e-07, "loss": 0.8878, "step": 56380 }, { "epoch": 3.83, "learning_rate": 4.809396327438634e-07, "loss": 0.2904, "step": 56385 }, { "epoch": 3.83, "learning_rate": 4.799969835224917e-07, "loss": 0.1813, "step": 56390 }, { "epoch": 3.83, "learning_rate": 4.790543343011199e-07, "loss": 0.3127, "step": 56395 }, { "epoch": 3.83, "learning_rate": 4.781116850797481e-07, "loss": 0.156, "step": 56400 }, { "epoch": 3.83, "learning_rate": 4.771690358583765e-07, "loss": 0.4948, "step": 56405 }, { "epoch": 3.83, "learning_rate": 4.7622638663700466e-07, "loss": 0.6951, "step": 56410 }, { "epoch": 3.83, "learning_rate": 4.7528373741563293e-07, "loss": 0.2111, "step": 56415 }, { "epoch": 3.83, "learning_rate": 4.7434108819426116e-07, "loss": 0.3067, "step": 56420 }, { "epoch": 3.83, "learning_rate": 4.733984389728895e-07, "loss": 0.1062, "step": 56425 }, { "epoch": 3.83, "learning_rate": 4.724557897515177e-07, "loss": 0.1426, "step": 56430 }, { "epoch": 3.83, "learning_rate": 4.71513140530146e-07, "loss": 0.1082, "step": 56435 }, { "epoch": 3.83, "learning_rate": 4.705704913087742e-07, "loss": 0.3131, "step": 56440 }, { "epoch": 3.83, "learning_rate": 4.6962784208740243e-07, "loss": 0.3166, "step": 56445 }, { "epoch": 3.83, "learning_rate": 4.6868519286603076e-07, "loss": 0.196, "step": 56450 }, { "epoch": 3.83, "learning_rate": 4.67742543644659e-07, "loss": 0.1893, "step": 56455 }, { "epoch": 3.83, "learning_rate": 4.6679989442328726e-07, "loss": 0.0932, "step": 56460 }, { "epoch": 3.83, "learning_rate": 4.658572452019155e-07, "loss": 0.4179, "step": 56465 }, { "epoch": 3.83, "learning_rate": 4.649145959805437e-07, "loss": 0.206, "step": 56470 }, { "epoch": 3.83, "learning_rate": 4.6397194675917203e-07, "loss": 0.327, "step": 56475 }, { "epoch": 3.83, "learning_rate": 4.630292975378003e-07, "loss": 0.4695, "step": 56480 }, { "epoch": 3.83, "learning_rate": 4.6208664831642853e-07, "loss": 0.3591, "step": 56485 }, { "epoch": 3.83, "learning_rate": 4.6114399909505675e-07, "loss": 0.2468, "step": 56490 }, { "epoch": 3.83, "learning_rate": 4.602013498736851e-07, "loss": 0.204, "step": 56495 }, { "epoch": 3.83, "learning_rate": 4.592587006523133e-07, "loss": 0.308, "step": 56500 }, { "epoch": 3.84, "learning_rate": 4.583160514309416e-07, "loss": 0.1064, "step": 56505 }, { "epoch": 3.84, "learning_rate": 4.573734022095698e-07, "loss": 0.0567, "step": 56510 }, { "epoch": 3.84, "learning_rate": 4.56430752988198e-07, "loss": 0.208, "step": 56515 }, { "epoch": 3.84, "learning_rate": 4.5548810376682635e-07, "loss": 0.3779, "step": 56520 }, { "epoch": 3.84, "learning_rate": 4.5454545454545457e-07, "loss": 0.354, "step": 56525 }, { "epoch": 3.84, "learning_rate": 4.5360280532408285e-07, "loss": 0.1246, "step": 56530 }, { "epoch": 3.84, "learning_rate": 4.5266015610271107e-07, "loss": 0.0433, "step": 56535 }, { "epoch": 3.84, "learning_rate": 4.5171750688133935e-07, "loss": 0.3901, "step": 56540 }, { "epoch": 3.84, "learning_rate": 4.507748576599676e-07, "loss": 0.122, "step": 56545 }, { "epoch": 3.84, "learning_rate": 4.498322084385959e-07, "loss": 0.1039, "step": 56550 }, { "epoch": 3.84, "learning_rate": 4.488895592172241e-07, "loss": 0.5172, "step": 56555 }, { "epoch": 3.84, "learning_rate": 4.4794690999585234e-07, "loss": 0.5607, "step": 56560 }, { "epoch": 3.84, "learning_rate": 4.4700426077448067e-07, "loss": 0.1238, "step": 56565 }, { "epoch": 3.84, "learning_rate": 4.460616115531089e-07, "loss": 0.3733, "step": 56570 }, { "epoch": 3.84, "learning_rate": 4.4511896233173717e-07, "loss": 0.1064, "step": 56575 }, { "epoch": 3.84, "learning_rate": 4.441763131103654e-07, "loss": 0.3689, "step": 56580 }, { "epoch": 3.84, "learning_rate": 4.432336638889936e-07, "loss": 0.1019, "step": 56585 }, { "epoch": 3.84, "learning_rate": 4.4229101466762194e-07, "loss": 0.1658, "step": 56590 }, { "epoch": 3.84, "learning_rate": 4.4134836544625017e-07, "loss": 0.4492, "step": 56595 }, { "epoch": 3.84, "learning_rate": 4.4040571622487844e-07, "loss": 0.1664, "step": 56600 }, { "epoch": 3.84, "learning_rate": 4.3946306700350666e-07, "loss": 0.4597, "step": 56605 }, { "epoch": 3.84, "learning_rate": 4.3852041778213494e-07, "loss": 0.1985, "step": 56610 }, { "epoch": 3.84, "learning_rate": 4.375777685607632e-07, "loss": 0.5605, "step": 56615 }, { "epoch": 3.84, "learning_rate": 4.366351193393915e-07, "loss": 0.3586, "step": 56620 }, { "epoch": 3.84, "learning_rate": 4.356924701180197e-07, "loss": 0.5173, "step": 56625 }, { "epoch": 3.84, "learning_rate": 4.3474982089664794e-07, "loss": 0.7028, "step": 56630 }, { "epoch": 3.84, "learning_rate": 4.3380717167527626e-07, "loss": 0.1678, "step": 56635 }, { "epoch": 3.84, "learning_rate": 4.328645224539045e-07, "loss": 0.4892, "step": 56640 }, { "epoch": 3.84, "learning_rate": 4.3192187323253276e-07, "loss": 0.2301, "step": 56645 }, { "epoch": 3.84, "learning_rate": 4.30979224011161e-07, "loss": 0.3047, "step": 56650 }, { "epoch": 3.85, "learning_rate": 4.300365747897892e-07, "loss": 0.1166, "step": 56655 }, { "epoch": 3.85, "learning_rate": 4.2909392556841754e-07, "loss": 0.2874, "step": 56660 }, { "epoch": 3.85, "learning_rate": 4.281512763470458e-07, "loss": 0.1987, "step": 56665 }, { "epoch": 3.85, "learning_rate": 4.2720862712567403e-07, "loss": 0.0809, "step": 56670 }, { "epoch": 3.85, "learning_rate": 4.2626597790430226e-07, "loss": 0.107, "step": 56675 }, { "epoch": 3.85, "learning_rate": 4.2532332868293053e-07, "loss": 0.596, "step": 56680 }, { "epoch": 3.85, "learning_rate": 4.243806794615588e-07, "loss": 0.1521, "step": 56685 }, { "epoch": 3.85, "learning_rate": 4.234380302401871e-07, "loss": 0.096, "step": 56690 }, { "epoch": 3.85, "learning_rate": 4.224953810188153e-07, "loss": 0.2296, "step": 56695 }, { "epoch": 3.85, "learning_rate": 4.2155273179744353e-07, "loss": 0.1134, "step": 56700 }, { "epoch": 3.85, "learning_rate": 4.2061008257607186e-07, "loss": 0.2556, "step": 56705 }, { "epoch": 3.85, "learning_rate": 4.196674333547001e-07, "loss": 0.1941, "step": 56710 }, { "epoch": 3.85, "learning_rate": 4.1872478413332836e-07, "loss": 0.1428, "step": 56715 }, { "epoch": 3.85, "learning_rate": 4.177821349119566e-07, "loss": 0.0714, "step": 56720 }, { "epoch": 3.85, "learning_rate": 4.1683948569058485e-07, "loss": 0.4399, "step": 56725 }, { "epoch": 3.85, "learning_rate": 4.1589683646921313e-07, "loss": 0.2668, "step": 56730 }, { "epoch": 3.85, "learning_rate": 4.149541872478414e-07, "loss": 0.3195, "step": 56735 }, { "epoch": 3.85, "learning_rate": 4.1401153802646963e-07, "loss": 0.1804, "step": 56740 }, { "epoch": 3.85, "learning_rate": 4.1306888880509785e-07, "loss": 0.2299, "step": 56745 }, { "epoch": 3.85, "learning_rate": 4.121262395837261e-07, "loss": 0.2304, "step": 56750 }, { "epoch": 3.85, "learning_rate": 4.111835903623544e-07, "loss": 0.3024, "step": 56755 }, { "epoch": 3.85, "learning_rate": 4.102409411409827e-07, "loss": 0.1725, "step": 56760 }, { "epoch": 3.85, "learning_rate": 4.092982919196109e-07, "loss": 0.0563, "step": 56765 }, { "epoch": 3.85, "learning_rate": 4.083556426982391e-07, "loss": 0.4274, "step": 56770 }, { "epoch": 3.85, "learning_rate": 4.0741299347686745e-07, "loss": 0.1302, "step": 56775 }, { "epoch": 3.85, "learning_rate": 4.0647034425549567e-07, "loss": 0.2961, "step": 56780 }, { "epoch": 3.85, "learning_rate": 4.0552769503412395e-07, "loss": 0.3093, "step": 56785 }, { "epoch": 3.85, "learning_rate": 4.0458504581275217e-07, "loss": 0.0747, "step": 56790 }, { "epoch": 3.85, "learning_rate": 4.0364239659138045e-07, "loss": 0.3508, "step": 56795 }, { "epoch": 3.86, "learning_rate": 4.026997473700087e-07, "loss": 0.2372, "step": 56800 }, { "epoch": 3.86, "learning_rate": 4.01757098148637e-07, "loss": 0.0444, "step": 56805 }, { "epoch": 3.86, "learning_rate": 4.008144489272652e-07, "loss": 0.2165, "step": 56810 }, { "epoch": 3.86, "learning_rate": 3.9987179970589344e-07, "loss": 0.2448, "step": 56815 }, { "epoch": 3.86, "learning_rate": 3.989291504845217e-07, "loss": 0.5707, "step": 56820 }, { "epoch": 3.86, "learning_rate": 3.9798650126315e-07, "loss": 0.2782, "step": 56825 }, { "epoch": 3.86, "learning_rate": 3.9704385204177827e-07, "loss": 0.0907, "step": 56830 }, { "epoch": 3.86, "learning_rate": 3.961012028204065e-07, "loss": 0.0691, "step": 56835 }, { "epoch": 3.86, "learning_rate": 3.951585535990347e-07, "loss": 0.3553, "step": 56840 }, { "epoch": 3.86, "learning_rate": 3.9421590437766304e-07, "loss": 0.2042, "step": 56845 }, { "epoch": 3.86, "learning_rate": 3.9327325515629127e-07, "loss": 0.2158, "step": 56850 }, { "epoch": 3.86, "learning_rate": 3.9233060593491954e-07, "loss": 0.121, "step": 56855 }, { "epoch": 3.86, "learning_rate": 3.9138795671354776e-07, "loss": 0.1099, "step": 56860 }, { "epoch": 3.86, "learning_rate": 3.9044530749217604e-07, "loss": 0.3367, "step": 56865 }, { "epoch": 3.86, "learning_rate": 3.895026582708043e-07, "loss": 0.2455, "step": 56870 }, { "epoch": 3.86, "learning_rate": 3.885600090494326e-07, "loss": 0.1726, "step": 56875 }, { "epoch": 3.86, "learning_rate": 3.876173598280608e-07, "loss": 0.1954, "step": 56880 }, { "epoch": 3.86, "learning_rate": 3.8667471060668904e-07, "loss": 0.3945, "step": 56885 }, { "epoch": 3.86, "learning_rate": 3.857320613853173e-07, "loss": 0.1896, "step": 56890 }, { "epoch": 3.86, "learning_rate": 3.847894121639456e-07, "loss": 0.5026, "step": 56895 }, { "epoch": 3.86, "learning_rate": 3.8384676294257386e-07, "loss": 0.6469, "step": 56900 }, { "epoch": 3.86, "learning_rate": 3.829041137212021e-07, "loss": 0.3202, "step": 56905 }, { "epoch": 3.86, "learning_rate": 3.819614644998303e-07, "loss": 0.3683, "step": 56910 }, { "epoch": 3.86, "learning_rate": 3.8101881527845864e-07, "loss": 0.1879, "step": 56915 }, { "epoch": 3.86, "learning_rate": 3.800761660570869e-07, "loss": 0.0642, "step": 56920 }, { "epoch": 3.86, "learning_rate": 3.7913351683571513e-07, "loss": 0.4494, "step": 56925 }, { "epoch": 3.86, "learning_rate": 3.7819086761434336e-07, "loss": 0.1216, "step": 56930 }, { "epoch": 3.86, "learning_rate": 3.7724821839297163e-07, "loss": 0.1766, "step": 56935 }, { "epoch": 3.86, "learning_rate": 3.763055691715999e-07, "loss": 0.3005, "step": 56940 }, { "epoch": 3.86, "learning_rate": 3.753629199502282e-07, "loss": 0.2889, "step": 56945 }, { "epoch": 3.87, "learning_rate": 3.744202707288564e-07, "loss": 0.2339, "step": 56950 }, { "epoch": 3.87, "learning_rate": 3.7347762150748463e-07, "loss": 0.4587, "step": 56955 }, { "epoch": 3.87, "learning_rate": 3.725349722861129e-07, "loss": 0.572, "step": 56960 }, { "epoch": 3.87, "learning_rate": 3.715923230647412e-07, "loss": 0.2511, "step": 56965 }, { "epoch": 3.87, "learning_rate": 3.7064967384336946e-07, "loss": 0.1499, "step": 56970 }, { "epoch": 3.87, "learning_rate": 3.697070246219977e-07, "loss": 0.0938, "step": 56975 }, { "epoch": 3.87, "learning_rate": 3.6876437540062595e-07, "loss": 0.2772, "step": 56980 }, { "epoch": 3.87, "learning_rate": 3.6782172617925423e-07, "loss": 0.1924, "step": 56985 }, { "epoch": 3.87, "learning_rate": 3.668790769578825e-07, "loss": 0.4723, "step": 56990 }, { "epoch": 3.87, "learning_rate": 3.6593642773651073e-07, "loss": 0.2381, "step": 56995 }, { "epoch": 3.87, "learning_rate": 3.6499377851513895e-07, "loss": 0.2073, "step": 57000 }, { "epoch": 3.87, "learning_rate": 3.640511292937672e-07, "loss": 0.2808, "step": 57005 }, { "epoch": 3.87, "learning_rate": 3.631084800723955e-07, "loss": 0.4232, "step": 57010 }, { "epoch": 3.87, "learning_rate": 3.621658308510238e-07, "loss": 0.1822, "step": 57015 }, { "epoch": 3.87, "learning_rate": 3.61223181629652e-07, "loss": 0.0727, "step": 57020 }, { "epoch": 3.87, "learning_rate": 3.602805324082802e-07, "loss": 0.2026, "step": 57025 }, { "epoch": 3.87, "learning_rate": 3.593378831869085e-07, "loss": 0.8215, "step": 57030 }, { "epoch": 3.87, "learning_rate": 3.5839523396553677e-07, "loss": 0.0409, "step": 57035 }, { "epoch": 3.87, "learning_rate": 3.5745258474416505e-07, "loss": 0.1308, "step": 57040 }, { "epoch": 3.87, "learning_rate": 3.5650993552279327e-07, "loss": 0.2214, "step": 57045 }, { "epoch": 3.87, "learning_rate": 3.5556728630142155e-07, "loss": 0.1042, "step": 57050 }, { "epoch": 3.87, "learning_rate": 3.546246370800498e-07, "loss": 0.1759, "step": 57055 }, { "epoch": 3.87, "learning_rate": 3.536819878586781e-07, "loss": 0.0798, "step": 57060 }, { "epoch": 3.87, "learning_rate": 3.527393386373063e-07, "loss": 0.1303, "step": 57065 }, { "epoch": 3.87, "learning_rate": 3.5179668941593454e-07, "loss": 0.1796, "step": 57070 }, { "epoch": 3.87, "learning_rate": 3.508540401945628e-07, "loss": 0.8155, "step": 57075 }, { "epoch": 3.87, "learning_rate": 3.499113909731911e-07, "loss": 0.0786, "step": 57080 }, { "epoch": 3.87, "learning_rate": 3.4896874175181937e-07, "loss": 0.3097, "step": 57085 }, { "epoch": 3.87, "learning_rate": 3.480260925304476e-07, "loss": 0.2892, "step": 57090 }, { "epoch": 3.88, "learning_rate": 3.470834433090758e-07, "loss": 0.2642, "step": 57095 }, { "epoch": 3.88, "learning_rate": 3.461407940877041e-07, "loss": 0.4784, "step": 57100 }, { "epoch": 3.88, "learning_rate": 3.451981448663324e-07, "loss": 0.4522, "step": 57105 }, { "epoch": 3.88, "learning_rate": 3.4425549564496064e-07, "loss": 0.4765, "step": 57110 }, { "epoch": 3.88, "learning_rate": 3.4331284642358886e-07, "loss": 0.6694, "step": 57115 }, { "epoch": 3.88, "learning_rate": 3.4237019720221714e-07, "loss": 0.1542, "step": 57120 }, { "epoch": 3.88, "learning_rate": 3.4142754798084536e-07, "loss": 0.1172, "step": 57125 }, { "epoch": 3.88, "learning_rate": 3.404848987594737e-07, "loss": 0.3302, "step": 57130 }, { "epoch": 3.88, "learning_rate": 3.395422495381019e-07, "loss": 0.301, "step": 57135 }, { "epoch": 3.88, "learning_rate": 3.3859960031673014e-07, "loss": 0.0711, "step": 57140 }, { "epoch": 3.88, "learning_rate": 3.376569510953584e-07, "loss": 0.2476, "step": 57145 }, { "epoch": 3.88, "learning_rate": 3.367143018739867e-07, "loss": 0.7987, "step": 57150 }, { "epoch": 3.88, "learning_rate": 3.3577165265261496e-07, "loss": 0.245, "step": 57155 }, { "epoch": 3.88, "learning_rate": 3.348290034312432e-07, "loss": 0.4942, "step": 57160 }, { "epoch": 3.88, "learning_rate": 3.3388635420987146e-07, "loss": 0.1385, "step": 57165 }, { "epoch": 3.88, "learning_rate": 3.329437049884997e-07, "loss": 0.4202, "step": 57170 }, { "epoch": 3.88, "learning_rate": 3.32001055767128e-07, "loss": 0.2537, "step": 57175 }, { "epoch": 3.88, "learning_rate": 3.3105840654575623e-07, "loss": 0.6655, "step": 57180 }, { "epoch": 3.88, "learning_rate": 3.3011575732438446e-07, "loss": 0.2695, "step": 57185 }, { "epoch": 3.88, "learning_rate": 3.2917310810301273e-07, "loss": 0.2468, "step": 57190 }, { "epoch": 3.88, "learning_rate": 3.2823045888164096e-07, "loss": 0.127, "step": 57195 }, { "epoch": 3.88, "learning_rate": 3.272878096602693e-07, "loss": 0.0628, "step": 57200 }, { "epoch": 3.88, "learning_rate": 3.263451604388975e-07, "loss": 0.2016, "step": 57205 }, { "epoch": 3.88, "learning_rate": 3.2540251121752573e-07, "loss": 0.2547, "step": 57210 }, { "epoch": 3.88, "learning_rate": 3.24459861996154e-07, "loss": 0.2159, "step": 57215 }, { "epoch": 3.88, "learning_rate": 3.235172127747823e-07, "loss": 0.1523, "step": 57220 }, { "epoch": 3.88, "learning_rate": 3.2257456355341056e-07, "loss": 0.2118, "step": 57225 }, { "epoch": 3.88, "learning_rate": 3.216319143320388e-07, "loss": 0.1425, "step": 57230 }, { "epoch": 3.88, "learning_rate": 3.2068926511066705e-07, "loss": 1.0493, "step": 57235 }, { "epoch": 3.88, "learning_rate": 3.197466158892953e-07, "loss": 0.4132, "step": 57240 }, { "epoch": 3.89, "learning_rate": 3.188039666679236e-07, "loss": 0.3705, "step": 57245 }, { "epoch": 3.89, "learning_rate": 3.1786131744655183e-07, "loss": 0.1314, "step": 57250 }, { "epoch": 3.89, "learning_rate": 3.1691866822518005e-07, "loss": 0.127, "step": 57255 }, { "epoch": 3.89, "learning_rate": 3.159760190038083e-07, "loss": 1.0081, "step": 57260 }, { "epoch": 3.89, "learning_rate": 3.1503336978243655e-07, "loss": 0.2317, "step": 57265 }, { "epoch": 3.89, "learning_rate": 3.140907205610649e-07, "loss": 0.5468, "step": 57270 }, { "epoch": 3.89, "learning_rate": 3.131480713396931e-07, "loss": 0.3036, "step": 57275 }, { "epoch": 3.89, "learning_rate": 3.122054221183213e-07, "loss": 0.5257, "step": 57280 }, { "epoch": 3.89, "learning_rate": 3.112627728969496e-07, "loss": 0.1828, "step": 57285 }, { "epoch": 3.89, "learning_rate": 3.1032012367557787e-07, "loss": 0.6084, "step": 57290 }, { "epoch": 3.89, "learning_rate": 3.0937747445420615e-07, "loss": 0.1691, "step": 57295 }, { "epoch": 3.89, "learning_rate": 3.0843482523283437e-07, "loss": 0.134, "step": 57300 }, { "epoch": 3.89, "learning_rate": 3.0749217601146265e-07, "loss": 0.522, "step": 57305 }, { "epoch": 3.89, "learning_rate": 3.065495267900909e-07, "loss": 0.0806, "step": 57310 }, { "epoch": 3.89, "learning_rate": 3.0560687756871914e-07, "loss": 0.3573, "step": 57315 }, { "epoch": 3.89, "learning_rate": 3.046642283473474e-07, "loss": 0.0605, "step": 57320 }, { "epoch": 3.89, "learning_rate": 3.0372157912597564e-07, "loss": 0.3113, "step": 57325 }, { "epoch": 3.89, "learning_rate": 3.027789299046039e-07, "loss": 0.1312, "step": 57330 }, { "epoch": 3.89, "learning_rate": 3.018362806832322e-07, "loss": 0.5691, "step": 57335 }, { "epoch": 3.89, "learning_rate": 3.008936314618604e-07, "loss": 0.305, "step": 57340 }, { "epoch": 3.89, "learning_rate": 2.999509822404887e-07, "loss": 0.2218, "step": 57345 }, { "epoch": 3.89, "learning_rate": 2.9900833301911697e-07, "loss": 0.2754, "step": 57350 }, { "epoch": 3.89, "learning_rate": 2.9806568379774524e-07, "loss": 0.0772, "step": 57355 }, { "epoch": 3.89, "learning_rate": 2.9712303457637347e-07, "loss": 0.1571, "step": 57360 }, { "epoch": 3.89, "learning_rate": 2.9618038535500174e-07, "loss": 0.2637, "step": 57365 }, { "epoch": 3.89, "learning_rate": 2.9523773613362996e-07, "loss": 0.1751, "step": 57370 }, { "epoch": 3.89, "learning_rate": 2.9429508691225824e-07, "loss": 0.3702, "step": 57375 }, { "epoch": 3.89, "learning_rate": 2.933524376908865e-07, "loss": 0.2779, "step": 57380 }, { "epoch": 3.89, "learning_rate": 2.9240978846951474e-07, "loss": 0.1772, "step": 57385 }, { "epoch": 3.9, "learning_rate": 2.91467139248143e-07, "loss": 0.0428, "step": 57390 }, { "epoch": 3.9, "learning_rate": 2.9052449002677124e-07, "loss": 0.5385, "step": 57395 }, { "epoch": 3.9, "learning_rate": 2.895818408053995e-07, "loss": 0.4738, "step": 57400 }, { "epoch": 3.9, "learning_rate": 2.886391915840278e-07, "loss": 0.3743, "step": 57405 }, { "epoch": 3.9, "learning_rate": 2.87696542362656e-07, "loss": 0.4254, "step": 57410 }, { "epoch": 3.9, "learning_rate": 2.867538931412843e-07, "loss": 0.4135, "step": 57415 }, { "epoch": 3.9, "learning_rate": 2.8581124391991256e-07, "loss": 0.4299, "step": 57420 }, { "epoch": 3.9, "learning_rate": 2.8486859469854084e-07, "loss": 0.2179, "step": 57425 }, { "epoch": 3.9, "learning_rate": 2.8392594547716906e-07, "loss": 0.4008, "step": 57430 }, { "epoch": 3.9, "learning_rate": 2.8298329625579733e-07, "loss": 0.1033, "step": 57435 }, { "epoch": 3.9, "learning_rate": 2.8204064703442556e-07, "loss": 0.1359, "step": 57440 }, { "epoch": 3.9, "learning_rate": 2.8109799781305383e-07, "loss": 0.2316, "step": 57445 }, { "epoch": 3.9, "learning_rate": 2.801553485916821e-07, "loss": 0.2649, "step": 57450 }, { "epoch": 3.9, "learning_rate": 2.7921269937031033e-07, "loss": 0.182, "step": 57455 }, { "epoch": 3.9, "learning_rate": 2.782700501489386e-07, "loss": 0.5791, "step": 57460 }, { "epoch": 3.9, "learning_rate": 2.7732740092756683e-07, "loss": 0.4344, "step": 57465 }, { "epoch": 3.9, "learning_rate": 2.763847517061951e-07, "loss": 0.1554, "step": 57470 }, { "epoch": 3.9, "learning_rate": 2.754421024848234e-07, "loss": 0.2431, "step": 57475 }, { "epoch": 3.9, "learning_rate": 2.744994532634516e-07, "loss": 0.5289, "step": 57480 }, { "epoch": 3.9, "learning_rate": 2.735568040420799e-07, "loss": 0.3534, "step": 57485 }, { "epoch": 3.9, "learning_rate": 2.7261415482070815e-07, "loss": 0.2291, "step": 57490 }, { "epoch": 3.9, "learning_rate": 2.7167150559933643e-07, "loss": 0.2202, "step": 57495 }, { "epoch": 3.9, "learning_rate": 2.7072885637796465e-07, "loss": 0.0765, "step": 57500 }, { "epoch": 3.9, "learning_rate": 2.6978620715659293e-07, "loss": 0.5487, "step": 57505 }, { "epoch": 3.9, "learning_rate": 2.6884355793522115e-07, "loss": 0.3738, "step": 57510 }, { "epoch": 3.9, "learning_rate": 2.679009087138494e-07, "loss": 0.3624, "step": 57515 }, { "epoch": 3.9, "learning_rate": 2.669582594924777e-07, "loss": 0.2531, "step": 57520 }, { "epoch": 3.9, "learning_rate": 2.660156102711059e-07, "loss": 0.1536, "step": 57525 }, { "epoch": 3.9, "learning_rate": 2.650729610497342e-07, "loss": 0.3057, "step": 57530 }, { "epoch": 3.9, "learning_rate": 2.641303118283624e-07, "loss": 0.398, "step": 57535 }, { "epoch": 3.91, "learning_rate": 2.6318766260699075e-07, "loss": 0.1966, "step": 57540 }, { "epoch": 3.91, "learning_rate": 2.6224501338561897e-07, "loss": 0.2134, "step": 57545 }, { "epoch": 3.91, "learning_rate": 2.613023641642472e-07, "loss": 0.0977, "step": 57550 }, { "epoch": 3.91, "learning_rate": 2.6035971494287547e-07, "loss": 0.1326, "step": 57555 }, { "epoch": 3.91, "learning_rate": 2.5941706572150375e-07, "loss": 0.345, "step": 57560 }, { "epoch": 3.91, "learning_rate": 2.58474416500132e-07, "loss": 0.1004, "step": 57565 }, { "epoch": 3.91, "learning_rate": 2.5753176727876024e-07, "loss": 0.127, "step": 57570 }, { "epoch": 3.91, "learning_rate": 2.565891180573885e-07, "loss": 0.3265, "step": 57575 }, { "epoch": 3.91, "learning_rate": 2.5564646883601674e-07, "loss": 0.2945, "step": 57580 }, { "epoch": 3.91, "learning_rate": 2.54703819614645e-07, "loss": 0.3407, "step": 57585 }, { "epoch": 3.91, "learning_rate": 2.537611703932733e-07, "loss": 0.1155, "step": 57590 }, { "epoch": 3.91, "learning_rate": 2.528185211719015e-07, "loss": 0.3499, "step": 57595 }, { "epoch": 3.91, "learning_rate": 2.518758719505298e-07, "loss": 0.8351, "step": 57600 }, { "epoch": 3.91, "learning_rate": 2.5093322272915807e-07, "loss": 0.7884, "step": 57605 }, { "epoch": 3.91, "learning_rate": 2.4999057350778634e-07, "loss": 0.056, "step": 57610 }, { "epoch": 3.91, "learning_rate": 2.4904792428641457e-07, "loss": 0.2432, "step": 57615 }, { "epoch": 3.91, "learning_rate": 2.481052750650428e-07, "loss": 0.2644, "step": 57620 }, { "epoch": 3.91, "learning_rate": 2.4716262584367106e-07, "loss": 0.1096, "step": 57625 }, { "epoch": 3.91, "learning_rate": 2.4621997662229934e-07, "loss": 0.2878, "step": 57630 }, { "epoch": 3.91, "learning_rate": 2.452773274009276e-07, "loss": 0.2304, "step": 57635 }, { "epoch": 3.91, "learning_rate": 2.4433467817955584e-07, "loss": 0.2301, "step": 57640 }, { "epoch": 3.91, "learning_rate": 2.433920289581841e-07, "loss": 0.1287, "step": 57645 }, { "epoch": 3.91, "learning_rate": 2.4244937973681234e-07, "loss": 0.2648, "step": 57650 }, { "epoch": 3.91, "learning_rate": 2.415067305154406e-07, "loss": 0.2356, "step": 57655 }, { "epoch": 3.91, "learning_rate": 2.405640812940689e-07, "loss": 0.0693, "step": 57660 }, { "epoch": 3.91, "learning_rate": 2.396214320726971e-07, "loss": 0.1265, "step": 57665 }, { "epoch": 3.91, "learning_rate": 2.386787828513254e-07, "loss": 0.1712, "step": 57670 }, { "epoch": 3.91, "learning_rate": 2.3773613362995363e-07, "loss": 0.1376, "step": 57675 }, { "epoch": 3.91, "learning_rate": 2.3679348440858188e-07, "loss": 0.1522, "step": 57680 }, { "epoch": 3.92, "learning_rate": 2.3585083518721016e-07, "loss": 0.1654, "step": 57685 }, { "epoch": 3.92, "learning_rate": 2.349081859658384e-07, "loss": 0.2374, "step": 57690 }, { "epoch": 3.92, "learning_rate": 2.3396553674446668e-07, "loss": 0.1845, "step": 57695 }, { "epoch": 3.92, "learning_rate": 2.330228875230949e-07, "loss": 0.3808, "step": 57700 }, { "epoch": 3.92, "learning_rate": 2.3208023830172318e-07, "loss": 0.5547, "step": 57705 }, { "epoch": 3.92, "learning_rate": 2.3113758908035143e-07, "loss": 0.6479, "step": 57710 }, { "epoch": 3.92, "learning_rate": 2.3019493985897968e-07, "loss": 0.107, "step": 57715 }, { "epoch": 3.92, "learning_rate": 2.2925229063760796e-07, "loss": 0.4995, "step": 57720 }, { "epoch": 3.92, "learning_rate": 2.283096414162362e-07, "loss": 0.1266, "step": 57725 }, { "epoch": 3.92, "learning_rate": 2.2736699219486448e-07, "loss": 0.662, "step": 57730 }, { "epoch": 3.92, "learning_rate": 2.264243429734927e-07, "loss": 0.133, "step": 57735 }, { "epoch": 3.92, "learning_rate": 2.25481693752121e-07, "loss": 0.1056, "step": 57740 }, { "epoch": 3.92, "learning_rate": 2.2453904453074923e-07, "loss": 0.1663, "step": 57745 }, { "epoch": 3.92, "learning_rate": 2.2359639530937748e-07, "loss": 0.1744, "step": 57750 }, { "epoch": 3.92, "learning_rate": 2.2265374608800575e-07, "loss": 0.1794, "step": 57755 }, { "epoch": 3.92, "learning_rate": 2.21711096866634e-07, "loss": 0.4752, "step": 57760 }, { "epoch": 3.92, "learning_rate": 2.2076844764526228e-07, "loss": 0.3866, "step": 57765 }, { "epoch": 3.92, "learning_rate": 2.1982579842389053e-07, "loss": 0.1991, "step": 57770 }, { "epoch": 3.92, "learning_rate": 2.188831492025188e-07, "loss": 0.6396, "step": 57775 }, { "epoch": 3.92, "learning_rate": 2.1794049998114702e-07, "loss": 0.2466, "step": 57780 }, { "epoch": 3.92, "learning_rate": 2.1699785075977527e-07, "loss": 0.2443, "step": 57785 }, { "epoch": 3.92, "learning_rate": 2.1605520153840355e-07, "loss": 0.1156, "step": 57790 }, { "epoch": 3.92, "learning_rate": 2.151125523170318e-07, "loss": 0.2797, "step": 57795 }, { "epoch": 3.92, "learning_rate": 2.1416990309566007e-07, "loss": 0.1121, "step": 57800 }, { "epoch": 3.92, "learning_rate": 2.1322725387428832e-07, "loss": 0.1028, "step": 57805 }, { "epoch": 3.92, "learning_rate": 2.122846046529166e-07, "loss": 0.5056, "step": 57810 }, { "epoch": 3.92, "learning_rate": 2.1134195543154482e-07, "loss": 0.6224, "step": 57815 }, { "epoch": 3.92, "learning_rate": 2.1039930621017307e-07, "loss": 0.132, "step": 57820 }, { "epoch": 3.92, "learning_rate": 2.0945665698880134e-07, "loss": 0.4863, "step": 57825 }, { "epoch": 3.92, "learning_rate": 2.085140077674296e-07, "loss": 0.1175, "step": 57830 }, { "epoch": 3.93, "learning_rate": 2.0757135854605787e-07, "loss": 0.2402, "step": 57835 }, { "epoch": 3.93, "learning_rate": 2.0662870932468612e-07, "loss": 0.3997, "step": 57840 }, { "epoch": 3.93, "learning_rate": 2.056860601033144e-07, "loss": 0.4966, "step": 57845 }, { "epoch": 3.93, "learning_rate": 2.0474341088194262e-07, "loss": 0.1023, "step": 57850 }, { "epoch": 3.93, "learning_rate": 2.0380076166057087e-07, "loss": 0.2167, "step": 57855 }, { "epoch": 3.93, "learning_rate": 2.0285811243919914e-07, "loss": 0.1893, "step": 57860 }, { "epoch": 3.93, "learning_rate": 2.019154632178274e-07, "loss": 0.1849, "step": 57865 }, { "epoch": 3.93, "learning_rate": 2.0097281399645567e-07, "loss": 0.2931, "step": 57870 }, { "epoch": 3.93, "learning_rate": 2.0003016477508392e-07, "loss": 0.3609, "step": 57875 }, { "epoch": 3.93, "learning_rate": 1.990875155537122e-07, "loss": 0.2816, "step": 57880 }, { "epoch": 3.93, "learning_rate": 1.9814486633234041e-07, "loss": 0.4405, "step": 57885 }, { "epoch": 3.93, "learning_rate": 1.9720221711096866e-07, "loss": 0.5254, "step": 57890 }, { "epoch": 3.93, "learning_rate": 1.9625956788959694e-07, "loss": 0.1072, "step": 57895 }, { "epoch": 3.93, "learning_rate": 1.953169186682252e-07, "loss": 0.3081, "step": 57900 }, { "epoch": 3.93, "learning_rate": 1.9437426944685346e-07, "loss": 0.2781, "step": 57905 }, { "epoch": 3.93, "learning_rate": 1.934316202254817e-07, "loss": 0.2811, "step": 57910 }, { "epoch": 3.93, "learning_rate": 1.9248897100411e-07, "loss": 0.6941, "step": 57915 }, { "epoch": 3.93, "learning_rate": 1.915463217827382e-07, "loss": 0.4475, "step": 57920 }, { "epoch": 3.93, "learning_rate": 1.9060367256136646e-07, "loss": 0.4795, "step": 57925 }, { "epoch": 3.93, "learning_rate": 1.8966102333999473e-07, "loss": 0.3347, "step": 57930 }, { "epoch": 3.93, "learning_rate": 1.8871837411862298e-07, "loss": 0.1714, "step": 57935 }, { "epoch": 3.93, "learning_rate": 1.8777572489725126e-07, "loss": 0.6158, "step": 57940 }, { "epoch": 3.93, "learning_rate": 1.868330756758795e-07, "loss": 0.2021, "step": 57945 }, { "epoch": 3.93, "learning_rate": 1.8589042645450778e-07, "loss": 0.2664, "step": 57950 }, { "epoch": 3.93, "learning_rate": 1.84947777233136e-07, "loss": 0.3984, "step": 57955 }, { "epoch": 3.93, "learning_rate": 1.8400512801176426e-07, "loss": 0.5515, "step": 57960 }, { "epoch": 3.93, "learning_rate": 1.8306247879039253e-07, "loss": 0.5319, "step": 57965 }, { "epoch": 3.93, "learning_rate": 1.8211982956902078e-07, "loss": 0.1528, "step": 57970 }, { "epoch": 3.93, "learning_rate": 1.8117718034764906e-07, "loss": 0.2733, "step": 57975 }, { "epoch": 3.94, "learning_rate": 1.802345311262773e-07, "loss": 0.0868, "step": 57980 }, { "epoch": 3.94, "learning_rate": 1.7929188190490558e-07, "loss": 0.2839, "step": 57985 }, { "epoch": 3.94, "learning_rate": 1.7834923268353383e-07, "loss": 0.1953, "step": 57990 }, { "epoch": 3.94, "learning_rate": 1.7740658346216205e-07, "loss": 0.2244, "step": 57995 }, { "epoch": 3.94, "learning_rate": 1.7646393424079033e-07, "loss": 0.273, "step": 58000 }, { "epoch": 3.94, "learning_rate": 1.7552128501941858e-07, "loss": 0.1305, "step": 58005 }, { "epoch": 3.94, "learning_rate": 1.7457863579804685e-07, "loss": 0.423, "step": 58010 }, { "epoch": 3.94, "learning_rate": 1.736359865766751e-07, "loss": 0.4078, "step": 58015 }, { "epoch": 3.94, "learning_rate": 1.7269333735530338e-07, "loss": 0.1477, "step": 58020 }, { "epoch": 3.94, "learning_rate": 1.7175068813393163e-07, "loss": 0.221, "step": 58025 }, { "epoch": 3.94, "learning_rate": 1.7080803891255985e-07, "loss": 0.2951, "step": 58030 }, { "epoch": 3.94, "learning_rate": 1.6986538969118812e-07, "loss": 0.2365, "step": 58035 }, { "epoch": 3.94, "learning_rate": 1.6892274046981637e-07, "loss": 0.3783, "step": 58040 }, { "epoch": 3.94, "learning_rate": 1.6798009124844465e-07, "loss": 0.0603, "step": 58045 }, { "epoch": 3.94, "learning_rate": 1.670374420270729e-07, "loss": 0.1091, "step": 58050 }, { "epoch": 3.94, "learning_rate": 1.6609479280570117e-07, "loss": 0.2747, "step": 58055 }, { "epoch": 3.94, "learning_rate": 1.6515214358432942e-07, "loss": 0.1163, "step": 58060 }, { "epoch": 3.94, "learning_rate": 1.6420949436295764e-07, "loss": 0.2889, "step": 58065 }, { "epoch": 3.94, "learning_rate": 1.6326684514158592e-07, "loss": 0.1948, "step": 58070 }, { "epoch": 3.94, "learning_rate": 1.6232419592021417e-07, "loss": 0.198, "step": 58075 }, { "epoch": 3.94, "learning_rate": 1.6138154669884244e-07, "loss": 0.1607, "step": 58080 }, { "epoch": 3.94, "learning_rate": 1.604388974774707e-07, "loss": 0.6541, "step": 58085 }, { "epoch": 3.94, "learning_rate": 1.5949624825609897e-07, "loss": 0.5057, "step": 58090 }, { "epoch": 3.94, "learning_rate": 1.5855359903472722e-07, "loss": 0.5132, "step": 58095 }, { "epoch": 3.94, "learning_rate": 1.5761094981335544e-07, "loss": 0.1189, "step": 58100 }, { "epoch": 3.94, "learning_rate": 1.5666830059198372e-07, "loss": 0.5656, "step": 58105 }, { "epoch": 3.94, "learning_rate": 1.55725651370612e-07, "loss": 0.1844, "step": 58110 }, { "epoch": 3.94, "learning_rate": 1.5478300214924024e-07, "loss": 0.1502, "step": 58115 }, { "epoch": 3.94, "learning_rate": 1.538403529278685e-07, "loss": 0.1509, "step": 58120 }, { "epoch": 3.94, "learning_rate": 1.5289770370649674e-07, "loss": 0.3881, "step": 58125 }, { "epoch": 3.95, "learning_rate": 1.5195505448512502e-07, "loss": 0.1216, "step": 58130 }, { "epoch": 3.95, "learning_rate": 1.5101240526375326e-07, "loss": 0.2464, "step": 58135 }, { "epoch": 3.95, "learning_rate": 1.5006975604238151e-07, "loss": 0.1275, "step": 58140 }, { "epoch": 3.95, "learning_rate": 1.491271068210098e-07, "loss": 0.1613, "step": 58145 }, { "epoch": 3.95, "learning_rate": 1.4818445759963804e-07, "loss": 0.473, "step": 58150 }, { "epoch": 3.95, "learning_rate": 1.472418083782663e-07, "loss": 0.179, "step": 58155 }, { "epoch": 3.95, "learning_rate": 1.4629915915689454e-07, "loss": 0.2659, "step": 58160 }, { "epoch": 3.95, "learning_rate": 1.453565099355228e-07, "loss": 0.0563, "step": 58165 }, { "epoch": 3.95, "learning_rate": 1.4441386071415106e-07, "loss": 0.0722, "step": 58170 }, { "epoch": 3.95, "learning_rate": 1.4347121149277934e-07, "loss": 0.1456, "step": 58175 }, { "epoch": 3.95, "learning_rate": 1.4252856227140759e-07, "loss": 0.2902, "step": 58180 }, { "epoch": 3.95, "learning_rate": 1.4158591305003583e-07, "loss": 0.1523, "step": 58185 }, { "epoch": 3.95, "learning_rate": 1.4064326382866408e-07, "loss": 0.1231, "step": 58190 }, { "epoch": 3.95, "learning_rate": 1.3970061460729233e-07, "loss": 0.0604, "step": 58195 }, { "epoch": 3.95, "learning_rate": 1.387579653859206e-07, "loss": 0.1558, "step": 58200 }, { "epoch": 3.95, "learning_rate": 1.3781531616454886e-07, "loss": 0.2262, "step": 58205 }, { "epoch": 3.95, "learning_rate": 1.3687266694317713e-07, "loss": 0.1464, "step": 58210 }, { "epoch": 3.95, "learning_rate": 1.3593001772180538e-07, "loss": 0.3525, "step": 58215 }, { "epoch": 3.95, "learning_rate": 1.3498736850043363e-07, "loss": 0.2847, "step": 58220 }, { "epoch": 3.95, "learning_rate": 1.3404471927906188e-07, "loss": 0.4569, "step": 58225 }, { "epoch": 3.95, "learning_rate": 1.3310207005769013e-07, "loss": 0.5153, "step": 58230 }, { "epoch": 3.95, "learning_rate": 1.321594208363184e-07, "loss": 0.0572, "step": 58235 }, { "epoch": 3.95, "learning_rate": 1.3121677161494665e-07, "loss": 0.6108, "step": 58240 }, { "epoch": 3.95, "learning_rate": 1.3027412239357493e-07, "loss": 0.1312, "step": 58245 }, { "epoch": 3.95, "learning_rate": 1.2933147317220318e-07, "loss": 0.8513, "step": 58250 }, { "epoch": 3.95, "learning_rate": 1.2838882395083143e-07, "loss": 0.0988, "step": 58255 }, { "epoch": 3.95, "learning_rate": 1.2744617472945968e-07, "loss": 0.1892, "step": 58260 }, { "epoch": 3.95, "learning_rate": 1.2650352550808793e-07, "loss": 0.2318, "step": 58265 }, { "epoch": 3.95, "learning_rate": 1.255608762867162e-07, "loss": 0.3958, "step": 58270 }, { "epoch": 3.96, "learning_rate": 1.2461822706534445e-07, "loss": 0.4717, "step": 58275 }, { "epoch": 3.96, "learning_rate": 1.2367557784397273e-07, "loss": 0.1536, "step": 58280 }, { "epoch": 3.96, "learning_rate": 1.2273292862260097e-07, "loss": 0.3394, "step": 58285 }, { "epoch": 3.96, "learning_rate": 1.2179027940122922e-07, "loss": 0.3603, "step": 58290 }, { "epoch": 3.96, "learning_rate": 1.2084763017985747e-07, "loss": 0.2623, "step": 58295 }, { "epoch": 3.96, "learning_rate": 1.1990498095848575e-07, "loss": 0.3045, "step": 58300 }, { "epoch": 3.96, "learning_rate": 1.18962331737114e-07, "loss": 0.1207, "step": 58305 }, { "epoch": 3.96, "learning_rate": 1.1801968251574226e-07, "loss": 0.0951, "step": 58310 }, { "epoch": 3.96, "learning_rate": 1.1707703329437051e-07, "loss": 0.5925, "step": 58315 }, { "epoch": 3.96, "learning_rate": 1.1613438407299877e-07, "loss": 0.3163, "step": 58320 }, { "epoch": 3.96, "learning_rate": 1.1519173485162702e-07, "loss": 0.2402, "step": 58325 }, { "epoch": 3.96, "learning_rate": 1.1424908563025527e-07, "loss": 0.4243, "step": 58330 }, { "epoch": 3.96, "learning_rate": 1.1330643640888353e-07, "loss": 0.3374, "step": 58335 }, { "epoch": 3.96, "learning_rate": 1.123637871875118e-07, "loss": 0.1292, "step": 58340 }, { "epoch": 3.96, "learning_rate": 1.1142113796614006e-07, "loss": 0.0731, "step": 58345 }, { "epoch": 3.96, "learning_rate": 1.104784887447683e-07, "loss": 0.6484, "step": 58350 }, { "epoch": 3.96, "learning_rate": 1.0953583952339657e-07, "loss": 0.0896, "step": 58355 }, { "epoch": 3.96, "learning_rate": 1.0859319030202482e-07, "loss": 0.2711, "step": 58360 }, { "epoch": 3.96, "learning_rate": 1.0765054108065307e-07, "loss": 0.3058, "step": 58365 }, { "epoch": 3.96, "learning_rate": 1.0670789185928133e-07, "loss": 0.1223, "step": 58370 }, { "epoch": 3.96, "learning_rate": 1.0576524263790959e-07, "loss": 0.3842, "step": 58375 }, { "epoch": 3.96, "learning_rate": 1.0482259341653785e-07, "loss": 0.1773, "step": 58380 }, { "epoch": 3.96, "learning_rate": 1.0387994419516612e-07, "loss": 0.2771, "step": 58385 }, { "epoch": 3.96, "learning_rate": 1.0293729497379436e-07, "loss": 0.1309, "step": 58390 }, { "epoch": 3.96, "learning_rate": 1.0199464575242261e-07, "loss": 0.2192, "step": 58395 }, { "epoch": 3.96, "learning_rate": 1.0105199653105088e-07, "loss": 0.2914, "step": 58400 }, { "epoch": 3.96, "learning_rate": 1.0010934730967912e-07, "loss": 0.3124, "step": 58405 }, { "epoch": 3.96, "learning_rate": 9.916669808830739e-08, "loss": 0.3059, "step": 58410 }, { "epoch": 3.96, "learning_rate": 9.822404886693565e-08, "loss": 0.3103, "step": 58415 }, { "epoch": 3.96, "learning_rate": 9.728139964556391e-08, "loss": 0.5406, "step": 58420 }, { "epoch": 3.97, "learning_rate": 9.633875042419216e-08, "loss": 0.0844, "step": 58425 }, { "epoch": 3.97, "learning_rate": 9.539610120282041e-08, "loss": 0.3165, "step": 58430 }, { "epoch": 3.97, "learning_rate": 9.445345198144867e-08, "loss": 0.4764, "step": 58435 }, { "epoch": 3.97, "learning_rate": 9.351080276007692e-08, "loss": 0.073, "step": 58440 }, { "epoch": 3.97, "learning_rate": 9.256815353870518e-08, "loss": 0.3873, "step": 58445 }, { "epoch": 3.97, "learning_rate": 9.162550431733345e-08, "loss": 0.0565, "step": 58450 }, { "epoch": 3.97, "learning_rate": 9.068285509596171e-08, "loss": 0.1677, "step": 58455 }, { "epoch": 3.97, "learning_rate": 8.974020587458996e-08, "loss": 0.5196, "step": 58460 }, { "epoch": 3.97, "learning_rate": 8.87975566532182e-08, "loss": 0.467, "step": 58465 }, { "epoch": 3.97, "learning_rate": 8.785490743184647e-08, "loss": 0.3818, "step": 58470 }, { "epoch": 3.97, "learning_rate": 8.691225821047472e-08, "loss": 0.1353, "step": 58475 }, { "epoch": 3.97, "learning_rate": 8.596960898910298e-08, "loss": 0.1731, "step": 58480 }, { "epoch": 3.97, "learning_rate": 8.502695976773124e-08, "loss": 0.5404, "step": 58485 }, { "epoch": 3.97, "learning_rate": 8.40843105463595e-08, "loss": 0.2056, "step": 58490 }, { "epoch": 3.97, "learning_rate": 8.314166132498777e-08, "loss": 0.4737, "step": 58495 }, { "epoch": 3.97, "learning_rate": 8.2199012103616e-08, "loss": 0.4387, "step": 58500 }, { "epoch": 3.97, "learning_rate": 8.125636288224427e-08, "loss": 0.1828, "step": 58505 }, { "epoch": 3.97, "learning_rate": 8.031371366087253e-08, "loss": 0.0679, "step": 58510 }, { "epoch": 3.97, "learning_rate": 7.937106443950078e-08, "loss": 0.1638, "step": 58515 }, { "epoch": 3.97, "learning_rate": 7.842841521812904e-08, "loss": 0.0633, "step": 58520 }, { "epoch": 3.97, "learning_rate": 7.748576599675729e-08, "loss": 0.224, "step": 58525 }, { "epoch": 3.97, "learning_rate": 7.654311677538555e-08, "loss": 0.1506, "step": 58530 }, { "epoch": 3.97, "learning_rate": 7.560046755401381e-08, "loss": 0.3166, "step": 58535 }, { "epoch": 3.97, "learning_rate": 7.465781833264206e-08, "loss": 0.2529, "step": 58540 }, { "epoch": 3.97, "learning_rate": 7.371516911127032e-08, "loss": 0.0946, "step": 58545 }, { "epoch": 3.97, "learning_rate": 7.277251988989857e-08, "loss": 0.364, "step": 58550 }, { "epoch": 3.97, "learning_rate": 7.182987066852684e-08, "loss": 0.279, "step": 58555 }, { "epoch": 3.97, "learning_rate": 7.088722144715508e-08, "loss": 0.3012, "step": 58560 }, { "epoch": 3.97, "learning_rate": 6.994457222578335e-08, "loss": 0.0891, "step": 58565 }, { "epoch": 3.98, "learning_rate": 6.900192300441161e-08, "loss": 0.1695, "step": 58570 }, { "epoch": 3.98, "learning_rate": 6.805927378303986e-08, "loss": 0.2883, "step": 58575 }, { "epoch": 3.98, "learning_rate": 6.711662456166812e-08, "loss": 0.0557, "step": 58580 }, { "epoch": 3.98, "learning_rate": 6.617397534029637e-08, "loss": 0.2394, "step": 58585 }, { "epoch": 3.98, "learning_rate": 6.523132611892463e-08, "loss": 0.2239, "step": 58590 }, { "epoch": 3.98, "learning_rate": 6.428867689755288e-08, "loss": 0.2381, "step": 58595 }, { "epoch": 3.98, "learning_rate": 6.334602767618114e-08, "loss": 0.2443, "step": 58600 }, { "epoch": 3.98, "learning_rate": 6.24033784548094e-08, "loss": 1.0488, "step": 58605 }, { "epoch": 3.98, "learning_rate": 6.146072923343765e-08, "loss": 0.3267, "step": 58610 }, { "epoch": 3.98, "learning_rate": 6.051808001206592e-08, "loss": 0.2225, "step": 58615 }, { "epoch": 3.98, "learning_rate": 5.957543079069417e-08, "loss": 0.1347, "step": 58620 }, { "epoch": 3.98, "learning_rate": 5.8632781569322435e-08, "loss": 0.1312, "step": 58625 }, { "epoch": 3.98, "learning_rate": 5.7690132347950684e-08, "loss": 0.1163, "step": 58630 }, { "epoch": 3.98, "learning_rate": 5.674748312657894e-08, "loss": 0.0944, "step": 58635 }, { "epoch": 3.98, "learning_rate": 5.58048339052072e-08, "loss": 0.3498, "step": 58640 }, { "epoch": 3.98, "learning_rate": 5.486218468383545e-08, "loss": 0.3524, "step": 58645 }, { "epoch": 3.98, "learning_rate": 5.3919535462463713e-08, "loss": 0.2287, "step": 58650 }, { "epoch": 3.98, "learning_rate": 5.297688624109197e-08, "loss": 0.1393, "step": 58655 }, { "epoch": 3.98, "learning_rate": 5.203423701972023e-08, "loss": 0.2083, "step": 58660 }, { "epoch": 3.98, "learning_rate": 5.109158779834848e-08, "loss": 0.1193, "step": 58665 }, { "epoch": 3.98, "learning_rate": 5.014893857697674e-08, "loss": 0.4259, "step": 58670 }, { "epoch": 3.98, "learning_rate": 4.9206289355605e-08, "loss": 0.3811, "step": 58675 }, { "epoch": 3.98, "learning_rate": 4.826364013423325e-08, "loss": 0.2232, "step": 58680 }, { "epoch": 3.98, "learning_rate": 4.732099091286151e-08, "loss": 0.3072, "step": 58685 }, { "epoch": 3.98, "learning_rate": 4.6378341691489766e-08, "loss": 0.3508, "step": 58690 }, { "epoch": 3.98, "learning_rate": 4.543569247011802e-08, "loss": 0.3747, "step": 58695 }, { "epoch": 3.98, "learning_rate": 4.449304324874628e-08, "loss": 0.3619, "step": 58700 }, { "epoch": 3.98, "learning_rate": 4.355039402737454e-08, "loss": 0.2184, "step": 58705 }, { "epoch": 3.98, "learning_rate": 4.2607744806002795e-08, "loss": 0.3774, "step": 58710 }, { "epoch": 3.99, "learning_rate": 4.166509558463105e-08, "loss": 0.3539, "step": 58715 }, { "epoch": 3.99, "learning_rate": 4.0722446363259306e-08, "loss": 0.1258, "step": 58720 }, { "epoch": 3.99, "learning_rate": 3.977979714188757e-08, "loss": 0.3452, "step": 58725 }, { "epoch": 3.99, "learning_rate": 3.8837147920515824e-08, "loss": 0.9076, "step": 58730 }, { "epoch": 3.99, "learning_rate": 3.7894498699144073e-08, "loss": 0.0907, "step": 58735 }, { "epoch": 3.99, "learning_rate": 3.6951849477772336e-08, "loss": 0.2888, "step": 58740 }, { "epoch": 3.99, "learning_rate": 3.600920025640059e-08, "loss": 0.2106, "step": 58745 }, { "epoch": 3.99, "learning_rate": 3.506655103502885e-08, "loss": 0.1188, "step": 58750 }, { "epoch": 3.99, "learning_rate": 3.41239018136571e-08, "loss": 0.3379, "step": 58755 }, { "epoch": 3.99, "learning_rate": 3.318125259228536e-08, "loss": 0.1538, "step": 58760 }, { "epoch": 3.99, "learning_rate": 3.223860337091362e-08, "loss": 0.4153, "step": 58765 }, { "epoch": 3.99, "learning_rate": 3.1295954149541877e-08, "loss": 0.4839, "step": 58770 }, { "epoch": 3.99, "learning_rate": 3.035330492817013e-08, "loss": 0.213, "step": 58775 }, { "epoch": 3.99, "learning_rate": 2.9410655706798388e-08, "loss": 0.3235, "step": 58780 }, { "epoch": 3.99, "learning_rate": 2.8468006485426647e-08, "loss": 0.4162, "step": 58785 }, { "epoch": 3.99, "learning_rate": 2.7525357264054903e-08, "loss": 0.163, "step": 58790 }, { "epoch": 3.99, "learning_rate": 2.6582708042683158e-08, "loss": 0.4955, "step": 58795 }, { "epoch": 3.99, "learning_rate": 2.5640058821311417e-08, "loss": 0.1652, "step": 58800 }, { "epoch": 3.99, "learning_rate": 2.469740959993967e-08, "loss": 0.3972, "step": 58805 }, { "epoch": 3.99, "learning_rate": 2.3754760378567932e-08, "loss": 0.1278, "step": 58810 }, { "epoch": 3.99, "learning_rate": 2.2812111157196184e-08, "loss": 0.4284, "step": 58815 }, { "epoch": 3.99, "learning_rate": 2.1869461935824443e-08, "loss": 0.2897, "step": 58820 }, { "epoch": 3.99, "learning_rate": 2.09268127144527e-08, "loss": 0.3585, "step": 58825 }, { "epoch": 3.99, "learning_rate": 1.9984163493080955e-08, "loss": 0.1183, "step": 58830 }, { "epoch": 3.99, "learning_rate": 1.9041514271709214e-08, "loss": 0.48, "step": 58835 }, { "epoch": 3.99, "learning_rate": 1.809886505033747e-08, "loss": 0.2236, "step": 58840 }, { "epoch": 3.99, "learning_rate": 1.7156215828965725e-08, "loss": 0.1808, "step": 58845 }, { "epoch": 3.99, "learning_rate": 1.6213566607593984e-08, "loss": 0.3901, "step": 58850 }, { "epoch": 3.99, "learning_rate": 1.527091738622224e-08, "loss": 0.2787, "step": 58855 }, { "epoch": 3.99, "learning_rate": 1.4328268164850497e-08, "loss": 0.293, "step": 58860 }, { "epoch": 4.0, "learning_rate": 1.3385618943478755e-08, "loss": 0.8817, "step": 58865 }, { "epoch": 4.0, "learning_rate": 1.2442969722107012e-08, "loss": 0.4793, "step": 58870 }, { "epoch": 4.0, "learning_rate": 1.1500320500735266e-08, "loss": 0.224, "step": 58875 }, { "epoch": 4.0, "learning_rate": 1.0557671279363523e-08, "loss": 0.098, "step": 58880 }, { "epoch": 4.0, "learning_rate": 9.61502205799178e-09, "loss": 0.0942, "step": 58885 }, { "epoch": 4.0, "learning_rate": 8.672372836620038e-09, "loss": 0.3334, "step": 58890 }, { "epoch": 4.0, "learning_rate": 7.729723615248295e-09, "loss": 0.0226, "step": 58895 }, { "epoch": 4.0, "learning_rate": 6.787074393876551e-09, "loss": 0.1513, "step": 58900 }, { "epoch": 4.0, "learning_rate": 5.8444251725048076e-09, "loss": 0.3837, "step": 58905 }, { "epoch": 4.0, "learning_rate": 4.901775951133065e-09, "loss": 0.2993, "step": 58910 }, { "epoch": 4.0, "learning_rate": 3.9591267297613214e-09, "loss": 0.3691, "step": 58915 }, { "epoch": 4.0, "learning_rate": 3.016477508389578e-09, "loss": 0.2219, "step": 58920 }, { "epoch": 4.0, "learning_rate": 2.073828287017835e-09, "loss": 0.3615, "step": 58925 }, { "epoch": 4.0, "learning_rate": 1.1311790656460918e-09, "loss": 0.2802, "step": 58930 }, { "epoch": 4.0, "learning_rate": 1.8852984427434862e-10, "loss": 0.3439, "step": 58935 }, { "epoch": 4.0, "step": 58936, "total_flos": 1.3950467938185118e+20, "train_loss": 0.6123406731570221, "train_runtime": 29986.2288, "train_samples_per_second": 125.791, "train_steps_per_second": 1.965 } ], "max_steps": 58936, "num_train_epochs": 4, "total_flos": 1.3950467938185118e+20, "trial_name": null, "trial_params": null }